Limits and Infinite Series

Branko Ćurgus

February 15, 2022 18:27

 Contents

CHAPTER 1. Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1. Real Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1. Axioms for the Set R of Real Numbers . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2. Basic properties of the set of real numbers . . . . . . . . . . . . . . . . . . . 6
1.2. Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3. Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.1. The definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2. The sign and the unit step function . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.3. The floor and the ceiling function . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.4. A rounding function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.5. The absolute value function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.6. New functions from old . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
CHAPTER 2. Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1. Limit of a function as x approaches +∞ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.1. The definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.2. Examples for Definition 2.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.3. Negative results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.4. Infinite limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.5. Examples of infinite limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2. Limit of a function at a real number a . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.1. The definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.2. Examples for Definition 2.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2.3. Infinite limits at a real number a . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2.4. One-sided limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3. New limits from old . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.1. Squeeze theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.2. Four trigonometric limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.3. Algebra of limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4. Continuous functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4.1. The definition and examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4.2. General theorems about continuous functions . . . . . . . . . . . . . . . . . 46
CHAPTER 3. Infinite Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1. Sequences of real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1.1. Definitions and examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1.2. Convergent sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3
4 CONTENTS

3.1.3. Theorems about convergent sequences . . . . . . . . . . . . . . . . . . . . . . 52

3.1.4. The Monotone Convergence Theorem . . . . . . . . . . . . . . . . . . . . . . . 54
3.2. Infinite series of real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2.1. Definition and basic examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2.2. Geometric Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.3. How to recognize whether an infinite series is a geometric series? . . . 64
3.2.4. Harmonic Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2.5. Telescoping series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.2.6. Decimal numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2.7. Basic properties of infinite series . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.3. Convergence Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.3.1. Direct Comparison Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.3.2. Limit Comparison Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3.3. Integral Comparison Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.3.4. Ratio and root tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3.5. Alternating infinite series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.3.6. Absolute and Conditional Convergence . . . . . . . . . . . . . . . . . . . . . . 80
3.4. Infinite Series of functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.4.1. Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.4.2. Functions Represented as Power Series . . . . . . . . . . . . . . . . . . . . . . 88
3.4.3. Taylor series at 0 (Maclaurin series) . . . . . . . . . . . . . . . . . . . . . . . . 90
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
 CHAPTER 1

Preliminaries

1.1. Real Numbers

All numbers in these notes are real numbers. The set of all real numbers is denoted by
R. All mathematical proofs are constructed from axioms using the mathematical logic that
we reviewed in “A Brief Review of Mathematical Logic.” In the next subsection we present
Axioms of the set of the real numbers R.

1.1.1. Axioms for the Set R of Real Numbers.

Axiom 1 (AE: Addition exists). If a, b ∈ R, then the sum of a and b, denoted by a + b,
is a uniquely defined number in R.
Axiom 2 (AA: Addition is associative). For all a, b, c ∈ R we have a+(b+c) = (a+b)+c.
Axiom 3 (AC: Addition is commutative). For all a, b ∈ R we have a + b = b + a.
Axiom 4 (AZ: Addition has 0). There is an element 0 in R such that 0 + a = a + 0 = a
for all a ∈ R.
Axiom 5 (AO: Addition has opposites). If a ∈ R, then the equation a + x = 0 has a
solution −a ∈ R. The number −a is called the opposite of a.
Axiom 6 (ME: Multiplication exists). If a, b ∈ R, then the product of a and b, denoted
by ab, is a uniquely defined number in R.
Axiom 7 (MA: Multiplication is associative). For all a, b, c ∈ R we have a(bc) = (ab)c.
Axiom 8 (MC: Multiplication is commutative). For all a, b ∈ R we have ab = ba.
Axiom 9 (MO: Multiplication has 1). There is an element 1 6= 0 in R such that 1 · a =
a · 1 = a for all a ∈ R.
Axiom 10 (MR: Multiplication has reciprocals). If a ∈ R is such that a 6= 0, then the
1 1
equation a · x = 1 has a solution a−1 = in R. The number a−1 = is called the reciprocal
a a
of a.
Axiom 11 (DL: Distributive law, the connection between addition and multiplication).
For all a, b, c ∈ R we have a(b + c) = ab + ac.
Axiom 12 (OE: Order exists). Given any a, b ∈ R, exactly one of these statements is
true: a < b, a = b, or b < a. (The symbol a ≤ b stands for a < b or a = b.)
Axiom 13 (OT: Order is transitive). Given any a, b, c ∈ R, if a < b and b < c, then
a < c.
5
6 1. PRELIMINARIES

Axiom 14 (OA: Order respects addition). Given any a, b, c ∈ R, if a < b then a + c <
b + c.
Axiom 15 (OM: Order respects multiplication). Given any a, b, c ∈ R, if a < b and
0 < c, then ac < bc.
Axiom 16 (CA: Completeness Axiom). If A and B are nonempty subsets of R such
that for every a ∈ A and for every b ∈ B we have a ≤ b, then there exists c ∈ R such that
a ≤ c ≤ b for all a ∈ A and all b ∈ B.
1.1.2. Basic properties of the set of real numbers. In the next theorem we list
several most important properties of the real numbers that follow from the Axioms.
Theorem 1.1. Let a, b, c be real numbers. The following statements hold.
(i) a + c = b + c ⇔ a = b
(ii) −0 = 0
(iii) −a = (−1)a
(iv) ab = 0 ⇔ (a = 0) ∨ (b = 0)
(v) a < b ⇔ −b < −a
(vi) (a < b) ∧ (c < 0) ⇒ bc < ac
(vii) a 6= 0 ⇔ aa > 0
(viii) 0 < a ⇔ 0 < a1
(ix) If a and b are positive, the following equivalence holds a < b ⇔ 1b < a1 .
We can discuss proofs of these statements in Discussions on Canvas.
Now I will make a far-reaching statement: All properties about real numbers you learned
in high school algebra and precalculus courses can be deduced from the Axioms using
mathematical logic. Sometimes these deductions, more commonly known as proofs, are
tedious. That is probably why these proofs are not done formally in high school and
beginning college courses. However, we will assume that you have already learned a lot
of algebra. You can use all the algebra that you learned in your proofs. We will refer
to that knowledge as Background Knowledge. In each proof, it is a good idea to identify
the Background Knowledge you need and ensure that all the statements you are using are
true. The first step towards verifying the validity of Background Knowledge is a precise
formulation. Then you can draw graphs, try a few well-chosen cases, make up a real-life
illustration, or raise a question in Discussions on Canvas. If you find an algebraic property
particularly interesting, you can always challenge me to prove it from axioms.
The next theorem I call the Pizza-Party Inequality. It is often used in proofs.
Theorem 1.2. Let a, b, c and d are positive real numbers. The following implication
holds:
c d
(a ≤ b) ∧ (c ≤ d) ⇒ ≤ .
b a
From a pizza-party perspective, the implication in the preceding theorem is clear: If
your objective is getting more pizza, would you rather attend a smaller party that is sharing
in a larger pizza or a larger party sharing in a smaller pizza? Should this be taught in a
math class, or kids learn that in kindergarten? Ok, this is a math class, so we can prove it!

1.2. Sets
In this course we will use the standard set notation. We will be dealing with sets which
consist of real numbers. A set can be described by a clear statement such as “Let S be the
 1.2. SETS 7

set of real solutions of the equation x2 − x = 0.” A set can also be described by a listing of
all its elements; for example. In the preceding case: S = {0, 1}. To describe sets we often
use the set builder notation: 
S = x ∈ R : x2 = x .
The above expression is read as: “The set A is the set of all real numbers x such that
x2 = x.” Here the colon “:” is used as an abbreviation for the phrase “such that”. Instead
of colon many books use the vertical bar symbol | .
Pay attention to the usage of the braces (or curly brackets) { and } in the set notation.
The braces are used to delimit sets. Notice the difference between the symbols 0 and {0}.
The symbol 0 stands for the real number 0 and {0} denotes the set whose only element is
0.
Next we review some important subsets of the real numbers. The set of all integers is
denoted by Z. In the set notation it is written as

Z = . . . , −3, −2, −1, 0, 1, 2, 3, . . . .
Since we cannot list all the integers, we use the ellipses to indicate that the pattern continues
infinitely. The set of positive integers is denoted by N. In the set notation we can write
this set as follows  
N = 1, 2, 3, . . . = x ∈ Z : x > 0 .
The synonym for “positive integer” is “natural number”. A rational number is any real
number that can be expressed as a fraction p/q where p is an integer and q is a positive
integer. The set of rational numbers is denoted by Q. In the set notation we can write this
set as follows n o
p
Q = x ∈ R : x = where p ∈ Z, q ∈ N .
q
Further important subsets of R are intervals. Let a and b be real numbers such that
a < b. Below we list all possible intervals with endpoints a and b. The symbol ∧ denotes
the logical conjunction between two mathematical statements. We read it as and.

[a, b] = x ∈ R : (a ≤ x) ∧ (x ≤ b) is called a closed interval,

(a, b) = x ∈ R : (a < x) ∧ (x < b) is called an open interval,

[a, b) = x ∈ R : (a ≤ x) ∧ (x < b) is called a half-open interval,

(a, b] = x ∈ R : (a < x) ∧ (x ≤ b) is called a half-open interval.
The intervals above are called finite intervals. We also define four types of infinite intervals:

[a, +∞) = x ∈ R : a ≤ x is called a closed unbounded interval,

(a, +∞) = x ∈ R : a < x is called an open unbounded interval

(−∞, b] = x ∈ R : x ≤ b is called an unbounded closed interval,

(−∞, b) = x ∈ R : x < b is called an unbounded open interval.
Geometric illustrations of these intervals are given in Figures 1 through 8.
The infinity symbols −∞ and +∞ are used to indicate that the set is unbounded in
the negative (−∞) or positive (+∞) direction of the real number line. The symbols −∞
and +∞ are just symbols; they are not real numbers. Therefore we always exclude them
as endpoints by using parentheses.
The set R is also an infinite interval. Sometimes it is written as (−∞, +∞).
8 1. PRELIMINARIES

a b a b
Fig. 1. A closed interval Fig. 2. An open interval

a b a b
Fig. 3. A half-open interval Fig. 4. A half-open interval

a a
Fig. 5. A closed infinite interval Fig. 6. An open infinite interval

b b
Fig. 7. An infinite closed interval Fig. 8. An infinite open interval

Let S be a subset of R. If u is the smallest number in S, then u is called a minimum of

S and we write u = min S. If v is the greatest number in S, then v is called a maximum of S
and we write v = max S. More formally, we express these definitions as logical statements:
u = min S if and only if u∈S and u≤x for all x ∈ S,

v = max S if and only if v∈S and v ≥ x for all x ∈ S.

Notice that the set Z has neither a minimum nor a maximum. Also, the open interval
(a, b) has neither a minimum nor a maximum. The set N has no maximum and min N = 1.
Each finite subset of R has both a minimum and a maximum.

1.3. Functions
1.3.1. The definition. Next we review the definition of a function. Let A and B be
nonempty sets. A function f from A to B is a rule that assigns exactly one element
of B to each element in A. This relationship between the sets A and B and the rule f is
indicated by the following notation:
f : A → B.
This notation can be read as: “f maps the set A into the set B.” For x ∈ A the unique
element of B which is assigned to x by the function f is called the value of f at x. This
element is denoted by f (x). Sometimes this relationship between x and f (x) is emphasized
by the following notation:
x 7→ f (x) where x ∈ A.
This notation is particularly convenient when a function is given by a formula and it is not
given a letter name. For example,
x 7→ x2 where x ∈ [0, +∞).
 1.3. FUNCTIONS 9

Let f : A → B be a function. The set A is called the domain of f . The set B is called
the codomain of f . The subset

f (x) ∈ B : x ∈ A

of B is called the range of f .

In this class we are interested in functions for which both sets A and B are subsets of
the set of real numbers R. Some examples of such functions are given next.

1.3.2. The sign and the unit step function. Let sign : R → R be given by the
formula

 1
 for x > 0,
sign(x) = 0 for x = 0,


−1 for x < 0.
This function is called the sign function. A graph of the sign function is given in Fig. 9.
Notice the use of small circles and small disks on the graph of the sign function. The small
circles are placed at the points (0, −1) and (0, 1). They emphasise the fact that the value of
the sign function at 0 is neither −1, nor 1. The disk placed at the point (0, 0) emphasises
that the value of the sign function at 0 is 0. This is the standard notation used on the
graphs of piecewise defined function whenever a confusion could arise.
The domain of the sign function is the set R of real numbers. The range of the sign
function is the set {−1, 0, 1}.
Let us : R → R be given by the formula
(
1 for x ≥ 0,
us(x) =
0 for x < 0.

This function is called the unit step function. A graph of the unit step function is given in
Fig. 10. Notice that a disk is placed at the point (0, 1) and a circle is placed at (0, 0). This
notation emphasises that us(0) = 1.
The domain of the unit step function is the set R of real numbers. The range of the
unit step function is the set {0, 1}.

1 1

-2 -1 1 2 -2 -1 1 2

-1 -1

Fig. 9. The sign function Fig. 10. The unit step function

Exercise 1.3. Prove that max{u, v} = v + (u − v) us(u − v) for all u, v ∈ R.

10 1. PRELIMINARIES

1.3.3. The floor and the ceiling function. The floor function,
floor : R → R,
is defined by the formula

floor(x) = ⌊x⌋ = max k ∈ Z : k ≤ x .
A graph of the floor function is given in Fig. 11. Since the floor is piecewise defined function
and without disks and circles on its graph there could be confusion as of the exact values
at the integers, we placed disks at the following set of points:

(n, n) : n ∈ Z
and we placed circles at the following set of points:

(n − 1, n) : n ∈ Z .
It follows from the properties of the maximum that for an arbitrary x ∈ R we have the
following equivalence
m = ⌊x⌋ if and only if m≤x<m+1 and m ∈ Z.
It is important to notice the following equivalence: For all x ∈ R we have
⌊x⌋ ≤ x < ⌊x⌋ + 1 ⇔ x − 1 < ⌊x⌋ ≤ x.

3 3

2 2

1 1

-3 -2 -1 1 2 3 -3 -2 -1 1 2 3

-1 -1

-2 -2

-3 -3

Fig. 11. The floor function Fig. 12. The ceiling function

The ceiling function,

ceiling : R → R,
is defined by the formula
ceiling(x) = ⌈x⌉ = min{k ∈ Z : k ≥ x}.
A graph of the ceiling function is given in Fig. 12. In Fig. 12 we placed disks at the following
set of points: 
(n, n) : n ∈ Z
 1.3. FUNCTIONS 11

and we placed circles at the following set of points:


(n, n + 1) : n ∈ Z .
It follows from the properties of the minimum that for an arbitrary x ∈ R we have the
following equivalence
n = ⌈x⌉ if and only if n − 1 < x ≤ n and n ∈ Z.
Notice that the inequalities ⌈x⌉ − 1 < x ≤ ⌈x⌉ are equivalent to
x ≤ ⌈x⌉ < x + 1. (1.1)
Exercise 1.4. State clearly the domain and the range of the floor and the ceiling
function.
Exercise 1.5. Prove that for all x ∈ R we have
⌊2x⌋ = ⌊x⌋ + ⌊x + 21 ⌋.
Discover and prove the analogous identity for the ceiling function.
1.3.4. A rounding function. Since rounding function is probably the most used
function in everyday life, I wanted to include it in these notes. For a real number x by
⌈x⌋ we denote the closest integer to x. The previous statement is ambiguous for the odd
multiples of 1/2. To be specific we define that ⌈1/2 + m⌋ = m + 1 for all m ∈ Z, see Fig. 13.
This function can be expressed using the floor function:
 
⌊2x⌋
⌈x⌋ = ⌊x + 1/2⌋ = for all x ∈ R.
2
Or, explicitly,
1 1
∀x ∈ R ∀m ∈ Z m = ⌈x⌋ ⇔ m − ≤ x < m + . (1.2)
2 2

3 3

2 2

1 1

5 3 1 1 3 5 5 3 1 1 3 5
-3 - -2 - -1 - 1 2 3 -3 - -2 - -1 - 1 2 3
2 2 2 2 2 2 2 2 2 2 2 2

-1 -1

-2 -2

-3 -3

Fig. 13. The graph of our ⌈x⌋ Fig. 14. An alternative rounding

The Wikipedia page on rounding lists six ways of rounding to the nearest integer. The
advantage of the rounding shown in Fig. 14 is that the function is odd. That is, the rounding
12 1. PRELIMINARIES

shown in Fig. 14 treats positive and negative values symmetrically, −1/2 is rounded −1,
while 1/2 is rounded to 1. This rounding is used in commercial transactions.
1.3.5. The absolute value function.
Definition 1.6. Let abs : R → R be defined by the piecewise formula
(
x if x ≥ 0,
abs(x) = |x| =
−x if x < 0.
This function is called the absolute value function. For a given real number x the number
|x| is called the absolute value of x.

-3 -2 -1 1 2 3

Fig. 15. The absolute value function

From calculus you are familiar with the geometric representation of real numbers as
points on a straight line. This is done by choosing a point on the line to represent 0 and
another point to represent 1. Then, every real number will correspond to a point on this
line (called the number line), and every point on the number line will correspond to a real
number. This geometric representation might be very helpful in doing problems.
Geometrically, the absolute value of a represents the distance between 0 and a, or,
generally |a − b| is the distance between the real numbers a and b on the number line.
Exercise 1.7. In the following problems write your solution as a set.
(a) Find all values of x such that |5x − 3| = 4.
(b) Find all values of x such that |5x − 3| < 4.
(c) Find all values of x such that |5x − 3| > 4.
Exercise 1.8. In the following problems write your solution as a set.
(a) Find all values of x such that |7x + 3| = 5.
(b) Find all values of x such that |7x + 3| < 5.
(c) Find all values of x such that |7x + 3| > 5.
The basic properties of the absolute value are given in the following theorem.
Theorem 1.9. The following statements hold.
(i) For all x ∈ R we have |x| = max{x, −x}.
(ii) For all x ∈ R we have |x| ≥ 0.
(iii) For all x ∈ R we have |−x| = |x|.
(iv) for all x ∈ R we have −x ≤ |x| and x ≤ |x|.
 1.3. FUNCTIONS 13

(v) For all x, y ∈ R we have |xy| = |x||y|.

x |x|
(vi) For all x, y ∈ R with y 6= 0 we have = .
y |y|
Proof. To prove (i) we consider two cases. Case I. Assume x ≥ 0. Then −x ≤ 0. Since
−x ≤ 0 and 0 ≤ x, it follows that −x ≤ x. Therefore max{x, −x} = x. By Definition 1.6
for x ≥ 0 we have that abs(x) = x. Hence, we conclude that abs(x) = max{x, −x} in
this case. Case II. Assume x < 0. Then −x > 0. Since −x > 0 and 0 > x, it follows
that −x > x. Therefore max{x, −x} = −x. By Definition 1.6 for x < 0 we have that
abs(x) = −x. Hence, we conclude that abs(x) = max{x, −x} in this case.
Since Cases I and II cover all real numbers x, the equality abs(x) = max{x, −x} is
proved.
The statement (ii) can also be proved by considering two cases.
To prove (iii) note that by (i) |x| = max{x, −x} and also |−x| = max{−x, −(−x)} =
max{−x, x}. Since max{x, −x} = max{−x, x}, we conclude that |x| = |−x|.
By the definition of max we have max{a, b} ≥ a and max{a, b} ≥ b for any real numbers
a and b. Therefore max{x, −x} ≥ x and max{x, −x} ≥ −x. Using (i) we conclude |x| ≥ x
and |x| ≥ −x. This proves (iv).
The proof of (v) is by considering four cases. The proof of (vi) first considers the case
x = 1 by two cases and then applies (v). 
Exercise 1.10. Let x and y be real numbers. Prove that
1

max{x, y} = x + y + |x − y| .
2
Exercise 1.11. Let x ∈ R and a > 0. Prove that |x| < a if and only if −a < x < a.

Theorem 1.12. (Triangle Inequalities)

(a) For all a, b ∈ R we have |a + b| ≤ |a| + |b|.
(b) For all x, y, z ∈ R we have |x − y| ≤ |x − z| + |z − y|.
(c) For all x, y ∈ R we have |x| − |y| ≤ |x − y|.
Proof. Proof of (a). By Theorem 1.9 (iv) we know that a ≤ |a| and b ≤ |b|. Therefore
we conclude that
a + b ≤ |a| + |b|. (1.3)
By Theorem 1.9 (iv) we know that −a ≤ |a| and −b ≤ |b|. Therefore we conclude
− (a + b) = −a + (−b) ≤ |a| + |b|. (1.4)
The inequalities (1.3) and (1.4) imply
max{a + b, −(a + b)} ≤ |a| + |b|. (1.5)
By Theorem 1.9 (i) the inequality (1.5) yields |a + b| ≤ |a| + |b|. This proves (a).
Prove (b) and (c) as an exercise. 
The inequalities in Theorem 1.12 are called the Triangle Inequalities.
Exercise 1.13. Let a, b, c be real numbers such that a 6= 0 and c > 0. Write your
solution as a set.
(a) Find all values of x such that |ax + b| = c.
(b) Find all values of x such that |ax + b| < c.
14 1. PRELIMINARIES

(c) Find all values of x such that |ax + b| > c.

Exercise 1.14. Let a be a real number and let ǫ be a positive real number. Prove that
|x − a| < ǫ if and only if x ∈ (a − ǫ, a + ǫ).
1.3.6. New functions from old.
Definition 1.15. Given two functions f : A → B and g : A → B, with A, B ⊆ R, and
two real numbers α and β we form a new function αf + βg : A → B defined by
(αf + βg)(x) = a f (x) + β g(x), for all x ∈ A.
Notice that f (x) and g(x) are real numbers so that α f (x) and β g(x) in the above formula
is just a multiplication of real numbers. The function αf + βg is called a linear combination
of the functions f and g.
Definition 1.16. Given two functions f : A → B and g : A → B, with A, B ⊆ R we
form a new function f g : A → B defined by
(f g)(x) = f (x)g(x), for all x ∈ A.
Notice that f (x) and g(x) are real numbers so that f (x)g(x) in the above formula is just
a multiplication of real numbers. The function f g is called the product of the functions f
and g.
Definition 1.17. Given two functions f : A → B and g : B → C a new function
g ◦ f : A → C is defined by
(g ◦ f )(x) = g(f (x)), x ∈ A.
The function g ◦ f is called the composition of the functions f and g.
Applying these definitions to familiar functions gives rise to new, sometimes very inter-
esting functions.
Exercise 1.18. For each of the functions given below answer the following questions:
(a) What are the domain and the range of the function? (b) Plot the function using your
graphing calculator. Plot the function by hand emphasizing the details missed by your
graphing calculator.
(a) x 7→ x abs(x) (b) x 7→ x(1 − abs(x))
(c) x 7→ x sign(x) (d) x 7→ ceiling(x) − floor(x)
(e) x 7→ x − floor(x) (f) x 7→ x floor(1/x)
(g) x 7→ (1 + sign(x))/2 (h) x 7→ x us(x)
(i) x 7→ sign(abs(x)) (j) x 7→ abs(sign(x))
(k) x 7→ floor(abs(x)) (l) x 7→ ceiling(abs(x))
 CHAPTER 2

Limits

2.1. Limit of a function as x approaches +∞

2.1.1. The definition.
Definition 2.1. Let D be a subset of R and L ∈ R. A function f : D → R has the
limit L as x approaches +∞ if the following two conditions are satisfied:
(I) There exists X0 ∈ D such that [X0 , +∞) ⊆ D.
(II) For every real number ǫ > 0 there exists a real number X(ǫ) ≥ X0 such that
x > X(ǫ) ⇒ |f (x) − L| < ǫ.
If the conditions (I) and (II) in Definition 2.1 are satisfied we write
lim f (x) = L.
x→+∞

L+ǫ
L
L−ǫ

X(ǫ)
Fig. 1. An illustration for the condition (II) in Definition 2.1

2.1.2. Examples for Definition 2.1.

1
Example 2.2. Prove that lim √ = 0.
x→+∞ x−1
Solution. We have to show that the conditions (I) and (II) in Definition 2.1 are sat-
isfied. In this example L = 0 and we can take D = (1, +∞), since if x > 1, then x − 1 > 0
15
16 2. LIMITS
√
and 1/ x − 1 is defined. Next, we have to provide X0 . We can take X0 = 2, since clearly
[2, +∞) ⊆ (1, +∞).
Next we show that the condition (II) is satisfied. Let ǫ > 0 be given. We have to find a
real number X(ǫ) ≥ 2 such that
1
x > X(ǫ) ⇒ √ − 0 < ǫ. (2.1)
x−1
In some sense we have to solve the inequality
1
√ − 0 < ǫ.
x−1
for x. The first step is to simplify it. Clearly
1 1
√ −0 = √ for x ≥ 2.
x−1 x−1
Thus we need to solve
1
√ < ǫ.
x−1
This inequality is solved for x by using the following sequence of algebraic steps:
1 √ 1 1 1
√ <ǫ ⇔ x−1> ⇔ x − 1 > 2 ⇔ x > 2 + 1. (2.2)
x−1 ǫ ǫ ǫ
 
1
Since we need X(ǫ) ≥ 2, we choose X(ǫ) = max 2 + 1, 2 .
ǫ
It remains to prove that the implication (2.1) is satisfied. Assume that
x > X(ǫ). (2.3)
√
Since X(ǫ) ≥ 2, we conclude that x > 2. Therefore x − 1 > 0 and 1/ x − 1 is defined.
Since X(ǫ) ≥ 1/ǫ2 + 1, we conclude that
1
x > 2 + 1.
ǫ
Now the equivalences (2.2) imply that
1
√ < ǫ. (2.4)
x−1
√
Since 1/ x − 1 is positive we conclude that
1 1 1
√ = √ = √ −0 . (2.5)
x−1 x−1 x−1
Combining (2.4) and (2.5), yields
1
√ − 0 < ǫ. (2.6)
x−1
Thus, we have proved that the assumption (2.3) implies the inequality (2.6). This is exactly
the implication (2.1). 
⌈x⌉
Example 2.3. Determine the limit of the function x 7→ as x approaches +∞ and
x
prove your claim.
 2.1. LIMIT OF A FUNCTION AS x APPROACHES +∞ 17

Solution. In Subsection 1.3.3, see (1.1), we established that x ≤ ⌈x⌉ < x + 1 for
every real number x. Therefore, for large x, the value of ⌈x⌉ does not differ much from x.
Therefore it is reasonable to make the following claim
⌈x⌉
lim = 1.
x→+∞ x
⌈x⌉
Next we will prove this claim using Definition 2.1. The function x 7→ is defined for all
x
x 6= 0. Thus, we can take D = R\{0}, and X0 = 1. In this example L = 1.
Next we show that the condition (II) is satisfied. Let ǫ > 0 be given. We have to find a
real number X(ǫ) ≥ 1 such that
⌈x⌉
x > X(ǫ) ⇒ − 1 < ǫ. (2.7)
x
Solving for x the inequality
⌈x⌉
−1 <ǫ (2.8)
x
is not easy. To find solutions of this inequality we first need to simplify it. In this process
of simplification we can replace the expression
⌈x⌉
−1
x
with an expression which is greater or equal to it. By the definition of the ceiling function
we know that
x ≤ ⌈x⌉ < x + 1. (2.9)
Since we consider only x ≥ 1, we can divide by x in (2.9) and subtract 1 from each term to
get
⌈x⌉ x+1 1
0≤ −1< −1 = .
x x x
Therefore
⌈x⌉ 1
−1 ≤ for all x ≥ 1. (2.10)
x x
This inequality is the key step in this proof. Therefore I call it the BIg INequality, or BIN.
(Each of the proofs involving the definition of limit involves a BIN.) The importance of BIN
lies in the fact that instead of solving (2.8), we can solve for x the simpler inequality
1
< ǫ.
x
1
The solution of this inequality (remember x ≥ 1) is x > .
  ǫ
1
Now we can define X(ǫ) = max , 1 . With this X(ǫ) the implication (2.7) is true.
ǫ
It is easy to prove this claim: Assume that
 
1
x > X(ǫ) = max ,1 .
ǫ
18 2. LIMITS

1
Then x ≥ 1 and x > . Since x ≥ 1 the BIN inequality (see (2.10))
ǫ
⌈x⌉ 1
−1 ≤
x x
1
is true. Since also x > , we conclude that
ǫ
1
< ǫ.
x
The last two displayed inequalities imply that
⌈x⌉
− 1 < ǫ.
x
This proves the implication (2.7). 
Exercise 2.4. Determine whether the following functions have limits as x approaches
+∞. Prove your statements using the definition.
x 2x x + sin(x)
(a) x 7→ (b) x 7→ 2 (c) x 7→
3x − 2 x +x+1 x−1

x2 + x x3 − 2x2 + 1 √ √
(d) x 7→ (e) x 7→ (f) x 7→ x+1− x
x3 + 3 x3 + x + 101
 1/ ln x
x2 + x cos(x) 1 x2 − 1
(g) x 7→ (h) x 7→ (i) x 7→
x2 − x + 5 x x2 + 2x sin(x)
p
(j) x 7→ x − x2 − x
 
1 x
Exercise 2.5. Guess the limit of the function x 7→ ln 1 + and prove your guess.
x
Hint: 1) Use the rules for logarithms to simplify the expression. 2) Use the representation
of the logarithm function u 7→ ln(u) as an integral (area under the graph of the
x function
u 7→ 1/u) to find an upper and lower bound for the given function x 7→ ln 1 + x1 for large
values of x. The bounds should be very simple functions of x.
2.1.3. Negative results. How can we prove that lim f (x) = L is false? This means
x→+∞
that the condition (I) or the condition (II) in Definition 2.1 is not satisfied. Since the
condition (II) is the essence of the definition of limit we will focus on the negation of the
condition (II).
The negation of (II): There exists ǫ > 0 such that for every X ∈ R there exists x > X
such that |f (x) − L| ≥ ǫ.
Example 2.6. Prove that lim sin(x) = 0 is false.
x→+∞

Solution. Let ǫ = 1/2. For arbitrary X ∈ R we have

 
π X/π + π/2 > X
 
and, for x = π X/π + π/2, we have | sin(x)| = 1. Therefore
| sin(x) − 0| ≥ 1/2. 
 2.1. LIMIT OF A FUNCTION AS x APPROACHES +∞ 19

2
1.5
1
0.5
0
-0.5
-1
-1.5

0 π 2π 3π 4π 5π 6π 7π
Fig. 2. Illustration for the solution of Example 2.6

Now we consider the statement “ lim f (x) does not exist.”

x→+∞

This means that for every L ∈ R, lim f (x) = L is false.

x→+∞

Example 2.7. Prove that lim sin(x) does not exist.

x→+∞

Solution. Let L ∈ R be arbitrary. We need to prove that lim sin(x) = L is false.

x→+∞
Consider three cases L = 0, L < 0 and L > 0. The case L = 0 is done in Example 2.6. Now
assume L < 0. Let ǫ = 1/2. For arbitrary X ∈ R we have
 
X π
2π + >X
2π 2
X π
and, for x = 2π 2π + 2 , we have sin(x) = 1. Therefore
| sin(x) − L| = |1 − L| = 1 + |L| ≥ 1/2.
Do the case L > 0 as an exercise. 
2.1.4. Infinite limits.
Definition 2.8. Let D be a subset of R. A function f : D → R has the limit +∞ as x
approaches +∞ if the following two conditions are satisfied:
(I) There exists a real number X0 ∈ D such that [X0 , +∞) ⊆ D.
(II) For every real number M there exists a real number X(M ) ≥ X0 such that
x > X(M ) ⇒ f (x) > M.
The symbolic notation for this limit is
lim f (x) = +∞.
x→+∞

Definition 2.9. Let D be a subset of R. A function f : D → R has the limit −∞ as x

approaches +∞ if the following two conditions are satisfied:
20 2. LIMITS

(I) There exists a real number X0 ∈ D such that [X0 , +∞) ⊆ D.

(II) For every real number M there exists a real number X(M ) ≥ X0 such that
x > X(M ) ⇒ f (x) < M.
The symbolic notation for this limit is
lim f (x) = −∞.
x→+∞

2.1.5. Examples of infinite limits.

√ √
Example 2.10. Let f (x) = x. Prove that lim x = +∞.
x→+∞
√
Solution. The function · is defined for all x ≥ 0. Therefore we can take X0 = 0 in
the part (I) of the definition.
Now consider the part (II) of the definition. Let M ∈ R be arbitrary. we have to
determine a real number X(M ) such that
√
x > X(M ) ⇒ x > M.
√
This will be accomplished if we solve the inequality x > M . If M < 0, then all x ≥ 0
satisfy this inequality. If M ≥ 0 then the solution of the inequality is x > M 2 . Thus, we
can take (
M2 if M ≥ 0,
X(M ) =
0 if M < 0 .
Clearly, X(M ) ≥ 0 for all M ∈ R and
√
x > X(M ) ⇒ x > M. 
Example 2.11. Let f (x) = floor(x). Prove that lim floor(x) = +∞.
x→+∞

Solution. The function floor is defined for all x ∈ R. Therefore we can take X0 = 0
in the part (I) of the definition.
Now consider the part (II) of the definition. Let M ∈ R be arbitrary. We have to
determine a real number X(M ) ≥ X0 such that
x > X(M ) ⇒ floor(x) > M. (2.11)
This will be accomplished if we solve the inequality
floor(x) > M. (2.12)
Since we don’t know much about floor it is not easy to solve (2.12). To achieve the im-
plication (2.11), we can replace floor(x) in (2.12) with a smaller quantity g(x) such that
g(x) > M is easy to solve. Thus we need g(x) such that
(A) floor(x) ≥ g(x) for all x > X0 .
(B) g(x) > M is easy to solve.
By the definition of floor(x) we conclude that 0 ≤ x − floor(x) < 1 for all x ∈ R. Therefore
x − 1 < floor(x) for all x ∈ R. (2.13)
Clearly x − 1 > M is easy to solve: x > M + 1. Thus, we can take X(M ) = max{M + 1, 0}
in the part (II) of the definition. Clearly X(M ) ≥ X0 = 0. Let x > X(M ). Then x > M +1
and therefore x − 1 > M . By the inequality (2.13) we conclude that
floor(x) > x − 1 > M.
 2.2. LIMIT OF A FUNCTION AT A REAL NUMBER a 21

Thus x > X(M ) implies floor(x) > M . 

The key step in the solution of Example 2.11 was the discovery of the function g(x)
such that
(A) f (x) ≥ g(x) for all x > X0 .
(B) g(x) > M is easy to solve.
Most proofs about limits follow this same pattern. Therefore I refer to the discovery of the
function g as a Big Inequality or BIN for short.
Exercise 2.12. Determine whether the following functions have the limit +∞ when x
approaches +∞.
x2 √
(a) x 7→ , (b) x 7→ ln x, (c) x 7→ x − x,
2x + 1
x2 − x − 1 1
(d) x 7→ x − ln(x), (e) x 7→ √ , (f) x 7→ 1

,
x+2 x+1 sin x

(cos x)2 x
q p √ (2 + cos(x))x
(g) x 7→ x− x− x, (h) x 7→ √ , (j) x 7→ √ .
x + sin(x) x + sin(x)

2.2. Limit of a function at a real number a

2.2.1. The definition.
Definition 2.13. Let D be a subset of R and let a and L be real numbers. A function
f : D → R has the limit L as x approaches a if the following two conditions are satisfied:



(I) There exists a real number δ0 > 0 such that a − δ0 , a ∪ a, a + δ0 ⊆ D.
(II) For every real number ǫ > 0 there exists a real number δ(ǫ) such that 0 < δ(ǫ) ≤ δ0
and
0 < |x − a| < δ(ǫ) ⇒ |f (x) − L| < ǫ.
If the conditions (I) and (II) in Definition 2.13 are satisfied we write lim f (x) = L.
x→a
Figure 3 illustrates Definition 2.13.
Next we restate Definition 2.13 using the terminology of a calculator screen. The figure
below shows a fictional calculator screen with 35 pixels. We assume that ymin and ymax
are chosen in such a way that the number L is in the middle of the y-range and that xmin
and xmax are such that a is in the middle of the x-range.
In Definition 2.14 below we assume that the function f satisfies (I) in Definition 2.13.
We rephrase (II) from Definition 2.13 in terms of a calculator screen.
For the specific fictional calculator screen shown in Figure 4, the connection between
Definition 2.13 and Definition 2.14 is given by ǫ = (ymax − ymin)/8, xmin = a − δ(ǫ),
xmax = a + δ(ǫ) and δ(ǫ) = ∆.
The fictional screen in Figure 4 is chosen for its simplicity. The screen of TI-92 (see
the manual p. 321) is 239 pixels wide and 103 pixels tall; it has 24617 pixels. The screen
of TI-83 (see the manual p. 8-16) and of TI-82 is 95 pixels wide and 63 pixels tall; it has
5985 pixels. The screen of TI-85 (see the manual p. 4-13) is 127 pixels wide and 63 pixels
tall; it has 8001 pixels. The screen of TI-89 (see the manual p. 222) is 159 pixels wide and
77 pixels tall; it has 12243 pixels. Using these numbers you can calculate the connection
between ǫ and δ(ǫ) in Definition 2.13 and the screen of your calculator.
22 2. LIMITS

L+ǫ

L−ǫ

a − δ(ǫ) a a + δ(ǫ)

Fig. 3

Definition 2.14 (Calculator Screen). ymax

A function f has a limit L as x ap-
proaches a for every choice of ymin
and ymax there exists ∆ (which de-
L+ǫ
pends on ymin and ymax) such that
whenever we choose xmin and xmax
L ?
such that xmax − xmin < 2∆ the L−ǫ
graph of the function f will appear
to be a straight horizontal line on the
calculator screen with the only possi- ymin
ble exception at the pixel containing
x = a. xmin a xmax
a − δ(ǫ) a + δ(ǫ)
Fig. 4. A fictional calculator screen

2.2.2. Examples for Definition 2.13.

Example 2.15. Prove lim (3x − 1) = 5.
x→2

Solution. In this example a = 2, L = 5, D = R and f (x) = 3x − 1.

(I) We can take any positive number for δ0 . Since it might be useful to have a specific δ0
to work with, we set δ0 = 1.
(II) Let ǫ > 0 be given. Let δ(ǫ) = min{ǫ/3, 1}. Assume 0 < |x − 2| < δ(ǫ). Since
δ(ǫ) ≤ ǫ/3, we conclude that |x − 2| < ǫ/3. Next, we calculate
|(3x − 1) − 5| = |3x − 6| = 3 |x − 2| . (2.14)
It follows from the assumption 0 < |x − 2| < δ(ǫ) that |x − 2| < ǫ/3. Therefore we conclude
ǫ
|(3x − 1) − 5| = 3 |x − 2| < 3 = ǫ.
3
 2.2. LIMIT OF A FUNCTION AT A REAL NUMBER a 23

Thus we proved that

0 < |x − 2| < δ(ǫ) ⇒ |(3x − 1) − 5| < ǫ.
This is exactly the implication in (II) in Definition 2.13. Since ǫ > 0 was arbitrary this
completes the proof. 
Remark 2.16. How did we guess the formula for δ(ǫ) in the previous proof? We first
studied the implication in the statement (II) in Definition 2.13. The goal in that implication
is to prove
|(3x − 1) − 5| < ǫ.
To prove this inequality we need to assume something about |x − 2|. To find out what to
assume, we simplified the expression |(3x − 1) − 5| until |x − 2| appeared (see (2.14)). Then
we solved for |x − 2|. In this process of simplification we can afford to make the right-hand
side larger. This will be illustrated in the next example.


Example 2.17. Prove lim 3x2 − 2x − 1 = 7.
x→2

Solution. We will use Definition 2.13 to prove that the statement in the example is
correct. In this example a = 2, L = 7, D = R, and f (x) = 3x2 − 2x − 1, .
Next we prove (I). Since the given function is defined on R, we can take any positive
number for δ0 . In this example it is essential to specify δ0 , so, we put δ0 = 1. (Please pay
attention how this is used in an essential way in the proof below. Notice that this choice of
δ0 = 1 in essence implies that, from now on, we consider only in the values of x which are
in the set (1, 2) ∪ (2, 3).)
Next we will discover an inequality which will help us find a formula for δ(ǫ):
(3x2 − 2x − 1) − 7 = 3x2 − 2x − 8 = (3x + 4)(x − 2) = |3x + 4| |x − 2|.
Now we use the fact that we are considering only the values of x which are in the set
(1, 2) ∪ (2, 3). For x ∈ (1, 2) ∪ (2, 3) the value of |3x + 4| does not exceed 13. Therefore
|(3x2 − 2x − 1) − 7| ≤ 13 |x − 2| for all x ∈ (1, 2) ∪ (2, 3).
Let ǫ > 0 be given. The inequality 13 |x − 2| < ǫ is easy to solve for |x − 2|. The solution
is |x − 2| < ǫ/13. Now we define δ(ǫ):
n ǫ o
δ(ǫ) = min ,1 .
13
The remaining step of the proof is to prove the implication
|x − 2| < δ(ǫ) ⇒ |(3x2 − 2x − 1) − 7| < ǫ.
We hope that at this point you can prove this implication on your own. 
x3 − x − 4
Example 2.18. Prove lim = 2.
x→2 x−1
Solution. We will use Definition 2.13 to prove that the statement in the example is
correct. In this example a = 2, L = 2, D = R\{1} and f (x) = (x3 − x − 4)/(x − 1).
Next we prove (I). Notice that the function f (x) is defined on R \ {1}. In this proof
we are interested in the values of x near a = 2. Therefore, for δ0 we can take any positive
number which is smaller than 1. Since it is useful to have a specific number, we put
δ0 = 1/2. This implies that from now on we consider only the values of x which are in the
set (3/2, 2) ∪ (2, 5/2).
24 2. LIMITS

Next we will discover an inequality which will help us find a formula for δ(ǫ):
x3 − x − 4 x3 − 3x − 2 (x2 + 2x + 1)(x − 2) x2 + 2x + 1
−2 = = = |x−2|. (2.15)
x−1 x−1 x−1 x−1
Now remember that we are interested only in the values of x which are in the set (3/2, 2) ∪
(2, 5/2). For x ∈ (3/2, 2) ∪ (2, 5/2) we estimate
x2 + 2x + 1 x2 + 2x + 1 16
= ≤ = 32 for all x ∈ (3/2, 2) ∪ (2, 5/2). (2.16)
x−1 x−1 1/2
Combining (2.15) and (2.16) we get
x3 − x − 4
− 2 ≤ 32 |x − 2| for all x ∈ (3/2, 2) ∪ (2, 5/2).
x−1
Let ǫ > 0 be given. The inequality 32 |x − 2| < ǫ is very easy to solve for |x − 2|. The
solution is |x − 2| < ǫ/32. Now we define δ(ǫ):
 
ǫ 1
δ(ǫ) = min , .
32 2
The remaining piece of the proof is to prove the implication
x3 − x − 4
|x − 2| < δ(ǫ) ⇒ − 2 < ǫ.
x−1
We hope that at this point you can prove this on your own. Write down all the details of
your reasoning. 
√
Example 2.19. Prove lim x = 2.
x→4
√
Solution. In this example a = 4, L =√2, D = [0, +∞) and f (x) = x. We first deal
with (I). Notice that the function f (x) = x is defined on [0, +∞). We are interested in
the values of x near the point a = 4. Thus, for δ0 we can take any positive number which
is < 4. Since it is useful to have a specific number, we put δ0 = 1. (Notice that this implies
that from now on in this proof we are interested only in the values of x which are in the set
(3, 4) ∪ (4, 5).)
Next we will discover an inequality which will help us find a formula for δ(ǫ):
√ √
√ ( x − 2)( x + 2) x−4 1
x−2 = √ = √ = √ |x − 4|. (2.17)
x+2 x+2 x+2
Now remember that we are interested only in the values of x which are in the set (3, 4)∪(4, 5).
For x ∈ (3, 4) ∪ (4, 5) we estimate
1 1 1 1
√ =√ ≤√ ≤ for all x ∈ (3, 4) ∪ (4, 5). (2.18)
x+2 x+2 3+2 2
Combining (2.17) and (2.18) we get
√ 1
x − 2 ≤ |x − 4| for all x ∈ (3, 4) ∪ (4, 5).
2
Let ǫ > 0 be given. The inequality 12 |x − 4| < ǫ is easy to solve for |x − 4|. The solution
is |x − 4| < 2ǫ. Now define δ(ǫ):
δ(ǫ) = min {2ǫ, 1} .
 2.2. LIMIT OF A FUNCTION AT A REAL NUMBER a 25

The remaining step of the proof is to prove the implication

√
|x − 4| < min {2ǫ, 1} ⇒ x − 2 < ǫ.
We hope that at this point you can prove this on your own. As before, please do it and
write down the details of your reasoning. 
1 1
Example 2.20. Prove that for every a > 0, lim = .
x→a x a
Solution. Let a > 0 be arbitrary. In this example L = 1/a, f (x) = 1/x and D =
R \ {0}. Next, we deal with (I) in Definition 2.13. Since the function f (x) = 1/x is defined
on R \ {0} and we are interested in the values of x near the point a > 0, for δ0 we can take
any positive number which is smaller than a. Since it is useful to have a specific number,
we put δ0 = a/2. (Notice that this implies that from now on in this proof we are interested
only in the values of x which are in the set (a/2, a) ∪ (a, 3a/2).)
Next we will discover an inequality which will help us find a formula for δ(ǫ):
1 1 a−x |a − x| 1
− = = = |x − a| . (2.19)
x a xa xa xa
Now remember that we are interested only in the values of x which are in the set (a/2, a) ∪
(a, 3a/2). For x ∈ (a/2, a) ∪ (a, 3a/2) we estimate
1 1 2
≤ = 2 for all x ∈ (a/2, a) ∪ (a, 3a/2). (2.20)
xa (a/2)a a
Combining (2.19) and (2.20) we get
1 1 2
− ≤ 2 |x − a| for all x ∈ (a/2, a) ∪ (a, 3a/2).
x a a
2
Let ǫ > 0 be given. The inequality a2
|x − a| < ǫ is easy to solve for |x−a|. The solution
is |x − a| < (a2 /2)ǫ. Now define δ(ǫ):

a2 ǫ a
δ(ǫ) = min , .
2 2
The remaining step of the proof is to prove the implication
 2 
a ǫ a 1 1
|x − a| < min , ⇒ − < ǫ.
2 2 x a
We hope that at this point you can prove this on your own. Write down the details of your
reasoning. 
Exercise 2.21. Find each of the following limits. Prove your claims using Defini-
tion 2.13.


(a) lim (2x + 1) (b) lim (−3x − 7) (c) lim 4x2 + 3
x→3 x→1 x→1

x x2 − x + 2
(d) lim (e) lim (f) lim x1/3
x→2 x − 1 x→3 x+1 x→0
 3/ ln|x|
1 1 1
(g) lim (h) lim (i) lim
x→0 |x| x→3 x x→1 x2 +1
x
(j) lim 2
x→−2 x + 4 x + 3
26 2. LIMITS

x+1
Exercise 2.22. Let f (x) = . Does f have a limit at a = −1? Justify your answer.
x2 − 1
√ √
Exercise 2.23. Prove that for every a > 0, lim x = a.
x→a

2.2.3. Infinite limits at a real number a.

Definition 2.24. Let a ∈ R and let D ⊆ R. A function f : D → R has the limit +∞
as x approaches a if the following two conditions are satisfied:



(I) There exists a real number δ0 > 0 such that a − δ0 , a ∪ a, a + δ0 ⊆ D.
(II) For every real number M > 0 there exists a real number δ(M ) such that 0 <
δ(M ) ≤ δ0 and
0 < |x − a| < δ(M ) ⇒ f (x) > M.
If the conditions (I) and (II) in Definition 2.24 are satisfied we write lim f (x) = +∞.
x→a

Definition 2.25. Let a ∈ R and let D ⊆ R. A function f : D → R has the limit −∞

as x approaches a if the following two conditions are satisfied:



(I) There exists a real number δ0 > 0 such that a − δ0 , a ∪ a, a + δ0 ⊆ D.
(II) For every real number M < 0 there exists a real number δ(M ) such that 0 <
δ(M ) ≤ δ0 and
0 < |x − a| < δ(M ) ⇒ f (x) < M.
If the conditions (I) and (II) in Definition 2.25 are satisfied we write lim f (x) = −∞.
x→a

Exercise 2.26. Find each of the following limits. Prove your claims using the appro-
priate definition.
1 1 x−3
(a) lim (b) lim 2
(c) lim
x→0 |x| x→−3 (x + 3) x→2 x(x − 2)2

x x2 − x + 2 x2 − x
(d) lim (e) lim (f) lim
x→−1 (x + 1)4 x→+∞ x+1 x→+∞ 3−x
2.2.4. One-sided limits.
Definition 2.27. Let a, L ∈ R and let D ⊆ R. A function f : D → R has the limit L
as x approaches a from the left if the following two conditions are satisfied:


(I) There exists a real number δ0 > 0 such that a − δ0 , a ⊆ D.
(II) For every real number ǫ > 0 there exists a real number δ(ǫ) such that 0 < δ(ǫ) ≤ δ0
and
0 < a − x < δ(ǫ) ⇒ |f (x) − L| < ǫ.
If the conditions (I) and (II) in Definition 2.27 are satisfied we write lim f (x) = L.
x↑a

Definition 2.28. Let a, L ∈ R and let D ⊆ R. A function f : D → R has the limit

L ∈ R as x approaches a from the right if the following two conditions are satisfied:


(I) There exists a real number δ0 > 0 such that a, a + δ0 ⊆ D.
(II) For every real number ǫ > 0 there exists a real number δ(ǫ) such that 0 < δ(ǫ) ≤ δ0
and
0 < x − a < δ(ǫ) ⇒ |f (x) − L| < ǫ.
 2.2. LIMIT OF A FUNCTION AT A REAL NUMBER a 27

If the conditions (I) and (II) in Definition 2.28 are satisfied we write lim f (x) = L.
x↓a

Definition 2.29. Let a ∈ R and let D ⊆ R. A function f : D → R has the limit +∞

as x approaches a from the left if the following two conditions are satisfied:


(I) There exists a real number δ0 > 0 such that a − δ0 , a ⊆ D.
(II) For every real number M > 0 there exists a real number δ(M ) such that 0 <
δ(M ) ≤ δ0 and

0 < a − x < δ(M ) ⇒ f (x) > M.

If the conditions (I) and (II) in Definition 2.29 are satisfied we write lim f (x) = +∞.
x↑a

Definition 2.30. Let a ∈ R and let D ⊆ R. A function f : D → R has the limit +∞

as x approaches a from the right if the following two conditions are satisfied:


(I) There exists a real number δ0 > 0 such that a, a + δ0 ⊆ D.
(II) For every real number M > 0 there exists a real number δ(M ) such that 0 <
δ(M ) ≤ δ0 and

0 < x − a < δ(M ) ⇒ f (x) > M.

If the conditions (I) and (II) in Definition 2.30 are satisfied we write lim f (x) = +∞.
x↓a

Definition 2.31. Let a ∈ R and let D ⊆ R. A function f : D → R has the limit −∞

as x approaches a from the left if the following two conditions are satisfied:


(I) There exists a real number δ0 > 0 such that a − δ0 , a ⊆ D.
(II) For every real number M < 0 there exists a real number δ(M ) such that 0 <
δ(M ) ≤ δ0 and

0 < a − x < δ(M ) ⇒ f (x) < M.

If the conditions (I) and (II) in Definition 2.31 are satisfied we write lim f (x) = −∞.
x↑a

Definition 2.32. Let a ∈ R and let D ⊆ R. A function f : D → R has the limit −∞

as x approaches a from the right if the following two conditions are satisfied:


(I) There exists a real number δ0 > 0 such that a, a + δ0 ⊆ D.
(II) For every real number M < 0 there exists a real number δ(M ) such that 0 <
δ(M ) ≤ δ0 and

0 < x − a < δ(M ) ⇒ f (x) < M.

If the conditions (I) and (II) in Definition 2.32 are satisfied we write lim f (x) = −∞.
x↓a

Exercise 2.33. Find each of the following limits. Prove your claims using the appro-
priate definition.
28 2. LIMITS

3x − 15 3x − 15 x−3
(a) lim √ (b) lim √ (c) lim
x↑5 2
x − 10x + 25 x↓5 x2 − 10x + 25 x↑2 x(x − 2)
 
1 1 2 6
(d) lim − (e) lim √ (f) lim
x↓0 x x2 x↑5 5−x x↓5 5−x

x+3 x2 √

(g) lim (h) lim (i) lim x − x
x↑3 x2 − 9 x↑−3 x2 − 9 x↓0

x x2 √

(j) lim (k) lim (l) lim x− x
x→3 (x − 3)2 x↓−1 x+1 x→+∞

2.3. New limits from old

2.3.1. Squeeze theorems. In this section and in Section 2.3.3 we establish general
properties of limits which are based on the formal definition of limit. These properties are
stated as theorems.
Establishing theorems of this kind involves a major step forward in sophistication. Up
to this point we have been trying to show that limits exist directly from the definition. Now
for the first time we are going to assume that some limit exists (I refer to this in class as
a green limit.) and try to make use of this information to establish the existence of some
other limit (I refer to this in class as a red limit.). Remember that to establish the existence
of a limit, we had to come up with a procedure for finding δ(ǫ) that will work for any ǫ > 0
that is given. If we assume the existence of a limit, then we are assuming the existence of
such a procedure, though we may not know explicitly what it is. I refer to this as a green
δ(ǫ). It is this procedure we will need to use in order to construct a new procedure for the
limit whose existence we are trying to establish. I refer to this as a red δ(ǫ).
We start by considering squeeze theorems that resemble the role of BIN in previous
sections. The following theorem is the Sandwich Squeeze Theorem.
Theorem 2.34. Let f, g and h be given functions and let a and L be real numbers.
Suppose that the following three conditions are satisfied.
(1) lim f (x) = L.
x→a
(2) lim h(x) = L.
x→a


(3) There exists η0 > 0 such that f, g and h are defined on a − η0 , a ∪ a, a + η0 and



f (x) ≤ g(x) ≤ h(x) for all x ∈ a − η0 , a ∪ a, a + η0 .
Then
lim g(x) = L.
x→a

Proof. Here we have three functions and three definitions of limits, one for each func-
tion. Therefore we have to deal with three δ-s. We will give them appropriate names that
will distinguish them from each other. Let us name them δf , δg and δh .
In the theorem it is assumed that lim f (x) = L. This means that we are given the fact
x→a
that for every ǫ > 0 there exists δf (ǫ) > 0 (that is, we are given a function δf (ǫ)) such that
0 < |x − a| < δf (ǫ) ⇒ |f (x) − L| < ǫ. (2.21)
In class I refer to these as a green δf (·) and a green implication.
 2.3. NEW LIMITS FROM OLD 29

Since the theorem assumes that lim h(x) = L, we are also given that for every ǫ > 0
x→a
there exists δh (ǫ) > 0 such that
0 < |x − a| < δh (ǫ) ⇒ |h(x) − L| < ǫ. (2.22)
Again we refer to these as a green δh (·) and a green implication.
We need to prove that lim g(x) = L. Therefore, following the definition of limit, we
x→a
have to show that the following conditions are satisfied:
(I) There exists
a real number

δ0,g > 0 such that g(x) is defined for every x in the set
a − δ0,g , a ∪ a, a + δ0,g .
(II) For every real number ǫ > 0 there exists a real number δg (ǫ) such that 0 < δg (ǫ) ≤
δ0,g and such that
0 < |x − a| < δg (ǫ) ⇒ |g(x) − L| < ǫ. (2.23)
Since we have to produce δ0,g , δg (ǫ) and we have to prove the last implication, all of these
objects are red.
Notice that η0 in the theorem is green.
The objective here is to use the green objects to produce the red objects. We will do
that next. We put:
(I) δ0,g = η0 .
By the assumption

of the theorem g(x) is defined for every x in the set
a − η0 , a ∪ a, a + η0 .
(II) For every real number ǫ > 0, put

δg (ǫ) = min δf (ǫ), δh (ǫ), η0 .
This is a beautiful expression since the red object is expressed in terms of the green
objects.
It remains to prove the red implication (2.23) using the green implications and the
assumptions of the theorem.
To prove (2.23), assume that 0 < |x − a| < δg (ǫ). Then, clearly, 0 < |x − a| < η0 .
This is telling
me that x
6= a and that x is no further than η0 from a. Consequently,
x ∈ a − η0 , a ∪ a, a + η0 . Therefore, by the assumption of the theorem
f (x) ≤ g(x) ≤ h(x).
Subtracting L from each term in this inequality, we conclude that
f (x) − L ≤ g(x) − L ≤ h(x) − L.
Using the property of the absolute value that −|u| ≤ u ≤ |u| for every real number u, we
conclude that
− |f (x) − L| ≤ f (x) − L ≤ g(x) − L ≤ h(x) − L ≤ |h(x) − L|. (2.24)
From the assumption 0 < |x − a| < δg (ǫ), we conclude that 0 < |x − a| < δf (ǫ). By the
green implication (2.21), this implies that |f (x) − L| < ǫ and therefore
− ǫ < −|f (x) − L|. (2.25)
From the assumption 0 < |x − a| < δg (ǫ), we conclude that 0 < |x − a| < δh (ǫ). By the
green implication (2.22), this implies that
|h(x) − L| < ǫ. (2.26)
30 2. LIMITS

Putting together the inequalities (2.24), (2.25) and (2.26), we conclude that
− ǫ < g(x) − L < ǫ. (2.27)
The inequalities in (2.27) are equivalent to
|g(x) − L| < ǫ.
This proves that 0 < |x − a| < δg (ǫ) implies |g(x) − L| < ǫ and this is exactly the red
implication (2.23). This completes the proof. 
The following theorem is the Scissors Squeeze Theorem.
Theorem 2.35. Let f, g and h be given functions and let a ∈ R and L ∈ R. Assume
that
(1) lim f (x) = L.
x→a
(2) lim h(x) = L.
x→a


(3) There exists η0 > 0 such that f, g and h are defined on a − η0 , a ∪ a, a + η0 and


f (x) ≤ g(x) ≤ h(x) for all x ∈ a − η0 , a ,
and

h(x) ≤ g(x) ≤ f (x) for all x ∈ a, a + η0 .
Then
lim g(x) = L.
x→a

2.3.2. Four trigonometric limits. Figure 5 and the numbers that you can see on it
are essential for getting squeezes for limits involving trigonometric functions. The table to
the left of Figure 5 shows the numbers that you should be able to identify on the picture.

Geometric Associated 1
object number
D
⌢
Circular arc CB u B

Line segment OA cos u

Line segment AB sin u

C
Line segment AC 1 − cos u O A 1

Line segment CB You calculate

Line segment CD tan u

Line segment OB 1

Line segment OC 1
Fig. 5. The unit circle
 2.3. NEW LIMITS FROM OLD 31

Example 2.36. Prove that lim cos x = 1.

x→0

Solution. Set η0 = 1. Consider a positive u. Look at Figure 5. The triangle △ACB

is a right triangle. Therefore its hypothenuse, the line segment CB, is longer than its side
AC, whose length equals to 1 − cos u. Thus
1 − cos u = AC ≤ CB. (2.28)
The line segment CB is a segment of a straight line, therefore it is shorter than any
⌢
other curve joining C and B. In particular it is shorter than the circular arc CB joining
⌢
the points C and B. The length of the circular arc CB is u. Thus
CB ≤
⌢
CB ( = u ). (2.29)
Putting together the inequalities (2.28) and (2.29), we conclude that


1 − cos u ≤ u for all u ∈ 0, 1 . (2.30)
Since the length OA = cos u is smaller than 1, from (2.30) we conclude that


0 ≤ 1 − cos u ≤ u for all u ∈ 0, 1 ,
or, equivalently,

1 − u ≤ cos u ≤ 1 for all u ∈ 0, 1 ,
Now we substitute u = |x| and use the fact that cos(|x|) = cos x and the preceding inequality
becomes
1 − |x| ≤ cos x ≤ 1 for all x ∈ (−1, 1) . (2.31)


This is a sandwich squeeze for cos x. It is easy to prove that lim 1 = 1 and lim 1 − |x| = 1.
x→0 x→0
(Please prove this using the definition!) Now the Sandwich Squeeze Theorem implies that
lim cos x = 1.
x→0
At the end of the proof here we used the Sandwich Squeeze Theorem. However, we
could have also used the definition of limit. To use the definition, we observe that the
implication
1 − |x| ≤ cos x ≤ 1 ⇒ | cos x − 1| ≤ |x| (2.32)
is true and conclude that
∀x ∈ (−1, 1) we have | cos x − 1| ≤ |x|. (2.33)
Now, for an arbitrary ǫ > 0 we can prove the implication

0 < |x − 0| < min ǫ, 1 ⇒ | cos x − 1| < ǫ.
This provides a proof of this limit by using the definition of limit. 
Example 2.37. Prove that lim sin x = 0.
x→0

Solution. Set δ0 = 1. Consider a positive u. Look at Figure 5. The triangle △ACB

is a right triangle. Therefore its hypothenuse, the line segment CB, is longer than its side
AB which equals to sin u. Thus
sin u = AC ≤ CB.
As in Example 2.36 we have that CB < u. Therefore,


sin u ≤ u for all u ∈ 0, 1 . (2.34)
32 2. LIMITS


For x ∈ −1, 1 we substitute u = |x| and use the fact that sin(|x|) = | sin x| and (2.34)
becomes
| sin x| ≤ |x| for all x ∈ (−1, 1) .
With the last inequality we use the definition of limit to finish the proof. For an arbitrary
ǫ > 0 we can prove the implication

0 < |x − 0| < min ǫ, 1 ⇒ | sin x − 0| < ǫ.

sin x
Example 2.38. Prove that lim = 1.
x→0 x

Solution. To get a sandwich squeeze for this problem consider the following three
areas in Figure 5.
Area 1 The triangle △OCB .
Area 2 The sector of the unit disk bounded by the line segments CB and OB and the
⌢
circular arc CB joining the points C and B.
Area 3 The triangle △OCD .
The picture tells clearly the inequality between these areas. Write that inequality. Calculate
each area in terms of the numbers that appear in the table above. This will lead to the
inequality, which when simplified gives
sin u

cos u ≤ ≤1 for all u ∈ 0, 1 . (2.35)
u
Using the same idea as in the previous example, the inequality (2.35) leads to
sin x
cos x ≤ ≤1 for all x ∈ (−1, 0) ∪ (0, 1) . (2.36)
x
The inequality (2.36) is exactly what we need in the Sandwich Squeeze Theorem. Please
fill in all the details of the rest of the proof.
At the end of the proof here we used the Sandwich Squeeze Theorem. However, we could
have also used the definition of limit. To use the definition we need one more inequality.
We view inequality (2.36) as distances from 1 and conclude that the following inequality is
true:
sin x
− 1 ≤ | cos x − 1| for all x ∈ (−1, 0) ∪ (0, 1) .
x
Now we use inequality (2.33) from Example 2.36 and transitivity of order to conclude
sin x
− 1 ≤ |x| for all x ∈ (−1, 0) ∪ (0, 1) .
x
Now, for an arbitrary ǫ > 0 we can prove the implication
 sin x
0 < |x − 0| < min ǫ, 1 ⇒ − 1 < ǫ.
x
And this proves the stated limit. 
1 − cos x 1
Example 2.39. Prove that lim 2
= .
x→0 x 2
Solution. To establish squeeze inequalities consider three lengths:
Length 1 The line segment AB .
 2.3. NEW LIMITS FROM OLD 33

Length 2 The line segment CB .

⌢
Length 3 The length of the circular arc CB joining the points C and B.
The picture tells clearly the inequalities between these three lengths. Write these inequal-
ities. Calculate each length in terms of the numbers that appear in the table above. This
will lead to the inequalities, which, when simplified, give
 
1 sin u 2 1 − cos u 1

≤ ≤ for all u ∈ 0, 1 . (2.37)
2 u u2 2
As in the preceding three examples from inequality (2.37) we deduce
 
1 sin x 2 1 − cos x 1
≤ 2
≤ for all x ∈ (−1, 0) ∪ (0, 1) . (2.38)
2 x x 2
Next we recall two inequalities (2.31) and (2.36) to get
sin x
1 − |x| ≤ cos x ≤ for all x ∈ (−1, 0) ∪ (0, 1) .
x
For x ∈ (−1, 0) ∪ (0, 1) we have 1 − |x| ≥ 0, so we can square the first and the last term in
the preceding inequalities to get
 
2 sin x 2
(1 − |x|) ≤ for all x ∈ (−1, 0) ∪ (0, 1).
x
Finally, since (1 − |x|)2 = 1 − 2|x| + |x|2 ≥ 1 − 2|x| we conclude
 
sin x 2
1 − 2|x| ≤ for all x ∈ (−1, 0) ∪ (0, 1).
x
Substituting the last inequality in inequality (2.38) we get
1 1 − cos x 1
− |x| ≤ 2
≤ for all x ∈ (−1, 0) ∪ (0, 1).
2 x 2
This is a Sandwich Squeeze for the limit in this example. However, viewing the preceding
inequality as distances from 1/2 we deduce
1 − cos x 1
− ≤ |x| for all x ∈ (−1, 0) ∪ (0, 1).
x2 2
And the last inequality can be used to prove the following implication: for an arbitrary
ǫ > 0 we have
 1 − cos x 1
0 < |x − 0| < min ǫ, 1 ⇒ − < ǫ.
x2 2
proving the limit stated in this example. 
ln(1 + x)
Example 2.40. Prove that lim = 1.
x→0 x
Solution. The idea is to use the definition of ln as an integral and work with areas
to get squeeze inequalities. 
34 2. LIMITS

2.3.3. Algebra of limits. A nickname that I gave to a function which has a limit L
when x approaches a is: f is constantish L near a. If we are dealing with constant functions
f (x) = L and g(x) = K, then clearly the sum f + g of these two functions is a constant
function equal to L + K. The same is true for the product f g which is the constant function
equal to LK. Another question is whether we can talk about the reciprocal 1/f . If L 6= 0,
then the reciprocal of f is defined and it equals 1/L. In this section we will prove that all
these properties hold for constantish functions.
Theorem 2.41. Let f, g, and h, be functions with domain and range in R. Let a, K
and L be real numbers. Assume that
(1) lim f (x) = K,
x→a
(2) lim g(x) = L.
x→a
Then the following statements hold.
(A) If h = f + g, then lim h(x) = K + L.
x→a
(B) If h = f g, then lim h(x) = KL.
x→a
1 1
(C) If L 6= 0 and h = , then lim h(x) = .
g x→a L
f K
(D) If L 6= 0 and h = , then lim h(x) = .
g x→a L
Proof. The assumption lim f (x) = K implies that
x→a
green(I-f) There
exists (green!)
δ0,f > 0 such that f (x) is defined for all x in a −
δ0,f , a ∪ a, a + δ0,f ;
green(II-f) For every ǫ > 0 there exists (green!) δf (ǫ) such that 0 < δf (ǫ) ≤ δ0,f and such
that
0 < |x − a| < δf (ǫ) ⇒ |f (x) − K| < ǫ. (2.39)
The assumption lim g(x) = L implies that
x→a


green(I-g) There exists
(green!) δ0,g > 0 such that g(x) is defined for all x in a−δ0,g , a ∪
a, a + δ0,g ;
green(II-g) For every ǫ > 0 there exists (green!) δg (ǫ) such that 0 < δg (ǫ) ≤ δ0,g and such
that
0 < |x − a| < δg (ǫ) ⇒ |g(x) − L| < ǫ. (2.40)
Proof of the statement (A). Remember that h(x) = f (x) + g(x) here. First we list what is
red in this proof.


red(I-h) There exists
(red!) δ0,h > 0 such that h(x) is defined for all x in a − δ0,h , a ∪
a, a + δ0,h ;
red(II-h) For every ǫ > 0 there exists (red!) δh (ǫ) such that 0 < δh (ǫ) ≤ δ0,h and such
that
0 < |x − a| < δh (ǫ) ⇒ h(x) − (K + L) < ǫ. (2.41)
I will not elaborate here how I got the idea for δ0,h and δh (ǫ), I will just give formulas
and convince you that my choice is a correct one. The idea for the formulas comes from the
boxed paragraph on page 35. I invite you to enjoy the separation of colors in the following
formulas.
 2.3. NEW LIMITS FROM OLD 35

Let ǫ > 0 be given. Put

δ0,h = min {δ0,f , δ0,g }
n ǫ  ǫ o
δh (ǫ) = min δf , δg
2 2



Now we have to prove that
h(x) is defined

for every x ∈ a − δ0,h , a ∪ a, a + δ0,h .
Assume that x ∈ a − δ0,h , a ∪ a, a + δ0,h . Then
0 < |x − a| < δ0,h ≤ min {δ0,f , δ0,g } . (2.42)
It follows from (2.42) that
0 < |x − a| < δ0,f ,



and therefore x ∈ a − δ0,f , a ∪ a, a + δ0,f . Thus f (x) is defined. It also follows from
(2.42) that
0 < |x − a| < δ0,g ,



and therefore x ∈ a − δ0,g , a ∪ a, a + δ0,g .
Thus g(x) is
defined. Therefore h(x) =
f (x) + g(x) is defined for every x ∈ a − δ0,h , a ∪ a, a + δ0,h .
Now we will prove the red implication (2.41). Assume
n ǫ  ǫ o
0 < |x − a| < δh (ǫ) = min δf , δg . (2.43)
2 2
Then ǫ
0 < |x − a| < δf . (2.44)
2
The inequality (2.44) and the implication (2.39) allow me to conclude that
ǫ
|f (x) − K| < . (2.45)
2
It follows from (2.43) that ǫ
0 < |x − a| < δg . (2.46)
2
The inequality (2.46) and the implication (2.40) allow me to conclude that
ǫ
|g(x) − L| < . (2.47)
2
Now we remember that the absolute value has the property that |u + v| ≤ |u| + |v|.
We will apply this to the expression
|h(x) − (K + L)| = |f (x) + g(x) − K − L| = |(f (x) − K) + (g(x) − L)|
| {z } | {z }
u v
to get
|h(x) − (K + L)| ≤ |f (x) − K| + |g(x) − L|. (2.48)
This inequality plays a role of a BIN in this abstract proof. It has an unfriendly object
on the left and all friendly objects on the right.
The inequalities (2.45), (2.47) and (2.48) imply that
ǫ ǫ
|h(x) − (K + L)| < + = ǫ. (2.49)
2 2
Reviewing my reasoning above you should be convinced that based on the assumption (2.43)
we proved the inequality (2.49). This is exactly the implication (2.41). This completes the
proof of the statement (A).
36 2. LIMITS

Proof of the statement (B). Remember that h(x) = f (x)g(x) here. We first list what is red
in this proof.


red(I-h) There exists
(red!) δ0,h > 0 such that h(x) is defined for all x in a − δ0,h , a ∪
a, a + δ0,h ;
red(II-h) For every ǫ > 0 there exists (red!) δh (ǫ) such that 0 < δh (ǫ) ≤ δ0,h and such
that
0 < |x − a| < δh (ǫ) ⇒ |h(x) − KL| < ǫ. (2.50)
I will not elaborate how I got the idea for δ0,h and δh (ǫ), I will just give formulas and
convince you that my choice is a correct one. The idea for the formulas comes from the
boxed paragraph on page 37. Again, I invite you to enjoy the separation of colors in the
following formulas.
Let ǫ > 0 be given. Put
δ0,h = min {δ0,f , δg (1)}
    
ǫ ǫ
δh (ǫ) = min δf , δg .
2(|L| + 1) 2(|K| + 1)



Now we have to prove that
h(x) is defined

for every x ∈ a − δ0,h , a ∪ a, a + δ0,h .
Assume that x ∈ a − δ0,h , a ∪ a, a + δ0,h . Then
0 < |x − a| < δ0,h ≤ min {δ0,f , δg (1)} . (2.51)
It follows from (2.51) that
0 < |x − a| < δ0,f ,



and therefore x ∈ a − δ0,f , a ∪ a, a + δ0,f . Thus f (x) is defined. It also follows from
(2.51) that
0 < |x − a| < δg (1). (2.52)
Since by the assumption (II-g) we know that δg (1) ≤ δ0,g , the inequality (2.52) implies that
0 < |x − a| < δ0,g .



Therefore x ∈ a − δ0,g , a ∪ a, a +
δ0,g . Thus g(x)

is defined. Therefore h(x) = f (x)g(x)
is defined for every x ∈ a − δ0,h , a ∪ a, a + δ0,h .
At this point we will prove another consequence of the inequality (2.52). This inequality
and the implication (2.40) allow me to conclude that
|g(x) − L| < 1.
Therefore
−1 < g(x) − L < 1 ,
or, equivalently
−1 + L < g(x) < L + 1.
Multiplying the last inequality by −1, we conclude that
−1 − L < −g(x) < −L + 1.
From the last two inequalities we conclude that max{g(x), −g(x)} < max{L + 1, −L + 1} =
max{L, −L} + 1. Thus
|g(x)| < |L| + 1. (2.53)
 2.3. NEW LIMITS FROM OLD 37

Now we will prove the red implication (2.50). Assume

    
ǫ ǫ
0 < |x − a| < δh (ǫ) = min δf , δg . (2.54)
2(|L| + 1) 2(|K| + 1)
Then  
ǫ
0 < |x − a| < δf . (2.55)
2(|L| + 1)
The inequality (2.55) and the implication (2.39) allow me to conclude that
ǫ
|f (x) − K| < . (2.56)
2(|L| + 1)
It follows from (2.54) that
 
ǫ
0 < |x − a| < δg . (2.57)
2(|K| + 1)
The inequality (2.57) and the implication (2.40) allow me to conclude that
ǫ
|g(x) − L| < . (2.58)
2(|K| + 1)

Now we remember that the absolute value has the property that |u + v| ≤ |u| + |v|
and that |uv| = |u||v|. we will apply these properties to the expression



|h(x) − KL| = f (x)g(x) − KL = f (x)g(x) − Kg(x) + Kg(x) − KL
| {z } | {z }
u v


≤ f (x)g(x) − Kg(x) + Kg(x) − KL
≤ g(x) f (x) − K + |K| g(x) − L .
Summarizing
h(x) − KL ≤ g(x) f (x) − K + |K| g(x) − L . (2.59)
The inequalities (2.53) and (2.59) imply that


h(x) − KL ≤ |L| + 1 f (x) − K + |K| g(x) − L . (2.60)
This inequality plays a role of a BIN in this abstract proof. It has an unfriendly object
on the left and all friendly objects on the right.

The inequalities (2.56), (2.58) and (2.60) imply that

ǫ ǫ ǫ ǫ
|h(x) − LK| ≤ |L| + 1 + |K| < + = ǫ. (2.61)
2(|L| + 1) 2(|K| + 1) 2 2
I hope that my reasoning above convinces you that the assumption (2.54) implies the
inequality (2.61). This is exactly the implication (2.50). This completes the proof of the
part (B).
1
Proof of the statement (C). Here we assume that L 6= 0 and h(x) = . Next we list what
g(x)
is red in this proof.


red(I-h) There exists
(red!) δ0,h > 0 such that h(x) is defined for all x in a − δ0,h , a ∪
a, a + δ0,h ;
38 2. LIMITS

red(II-h) For every ǫ > 0 there exists (red!) δh (ǫ) such that 0 < δh (ǫ) ≤ δ0,h and such
that
1 1
0 < |x − a| < δh (ǫ) ⇒ − < ǫ. (2.62)
g(x) L
I will not elaborate how I got the idea for δ0,h and δh (ǫ), I will just give formulas and
convince you that my choice is a correct one. The idea for the formulas comes from the
boxed paragraph on page 39. Again, I invite you to enjoy the separation of colors in the
following formulas.
Let ǫ > 0 be given. Remember that it is assumed that |L| > 0. Put
 
|L|
δ0,h = δg
2
  2   
ǫL |L|
δh (ǫ) = min δg , δg .
2 2



Now we have to prove that
h(x) is defined

for every x ∈ a − δ0,h , a ∪ a, a + δ0,h .
Assume that x ∈ a − δ0,h , a ∪ a, a + δ0,h . Then
 
|L|
0 < |x − a| < δ0,h = δg .
2
This inequality and the implication (2.40) allow me to conclude that
|L|
|g(x) − L| < .
2
Therefore
|L| |L|
− < g(x) − L < ,
2 2
or, equivalently
|L| |L|
− + L < g(x) < L + .
2 2
Multiplying the last inequality by −1, we conclude that
|L| |L|
−L − < −g(x) < − L.
2 2
From the last two displayed relationships we conclude that
 
|L| |L| |L|
max{g(x), −g(x)} > max L − , −L − = max{L, −L} − .
2 2 2
Thus
|L| |L|
|g(x)| > |L| − = > 0. (2.63)
2 2
1

Consequently, g(x) 6= 0. Therefore, h(x) = is defined for all x ∈ a − δ0,h , a ∪ a, a +

g(x)
δ0,h .
Now we will prove the red implication (2.62). Assume
  2   
ǫL |L|
0 < |x − a| < δh (ǫ) = min δg , δg . (2.64)
2 2
Then  2
ǫL
0 < |x − a| < δg . (2.65)
2
 2.3. NEW LIMITS FROM OLD 39

The inequality (2.65) and the implication (2.40) allow me to conclude that
ǫL2
|g(x) − L| < . (2.66)
2
It also follows from (2.64) that
 
|L|
0 < |x − a| < δg .
2
We already proved that this inequality implies (2.63). Therefore
1 2
< . (2.67)
|g(x)| |L|
This inequality is used at the last step in the sequence of inequalities below. In some sense
this is an abstract version of a “pizza-party” play.
Using our standard tools, algebra, properties of the absolute value and the inequality
(2.67) we get
1 1 1 L − g(x) |L − g(x)|
h(x) − = − = =
L g(x) L g(x)L |g(x)| |L|
|g(x) − L| 1 |g(x) − L| 2 |g(x) − L|
= ≤ ≤ .
|g(x)| |L| |g(x)| |L| |L| |L|
Summarizing
1 1 2
− ≤ 2 |g(x) − L| . (2.68)
g(x) L L
This inequality plays a role of a BIN in this abstract proof. It has an unfriendly object
on the left and all friendly objects on the right.

The inequalities (2.66) and (2.68) imply that

1 1 2 ǫL2
− ≤ 2 = ǫ. (2.69)
g(x) L L 2
I hope that the reasoning above convinces you that the assumption (2.64) implies the
inequality (2.69). This is exactly the implication (2.62). This completes the proof of the
part (C).
f (x)
Proof of the statement (D). Here we assume that L 6= 0 and h(x) = . We can prove
g(x)
the statement (D) by using the universal power of the statements (B) and (C). First define
1
the functions g1 (x) = . Then, by the statement (C) we know
g(x)
1
.
lim g1 (x) = (2.70)
x→a L
Clearly, h(x) = f (x)g1 (x). Now we can apply the statement (B) to this function h. Taking
into account (2.70) the statement (B) implies
1 K
lim h(x) = K = .
x→a L L
This completes the proof of the statement (D). The theorem is proved. 
40 2. LIMITS

Exercise 2.42. Use the algebra of limits to give much simpler proofs for most of the
limits in the previous exercises and examples.

2.4. Continuous functions

2.4.1. The definition and examples. All this work about limits will now pay off
since we will be able to give mathematically rigorous definition of a continuous function.
Definition 2.43. Let D be a nonempty subset of R. A function f : D → R is continuous
at c if the following two conditions are satisfied:
(i) The function f is defined at c, that is c ∈ D.
(ii) lim f (x) = f (c).
x→a

To understand Definition 2.43 the reader needs to understand the concept of limit. Since
the concept of continuity is fundamental in mathematics it is important to understand the
definition of continuity directly, without appealing to the concept of limit.
Definition 2.44. Let D be a nonempty subset of R. A function f : D → R is continuous
at c if the following two conditions are satisfied:
(I) There exists a δ0 > 0 such that (a − δ0 , a + δ0 ) ⊆ D.
(II) For every ǫ > 0 there exists δ(ǫ) such that 0 < δ(ǫ) ≤ δ0 and such that
|x − c| < δ(ǫ) ⇒ |f (x) − f (c)| < ǫ.
Definition 2.44 is called ǫ-δ definition of continuity. (The symbol ǫ-δ is read “epsilon-
delta.”)
Definition 2.45. Let D be a nonempty subset of R. A function f : D → R is continuous
on D if it is continuous at each point in D.
A drawback of the Definition 2.45, together with Definition 2.44, is that it does not apply
to functions that are defined on closed intervals. For example, we cannot use Definition 2.44,
to prove that the square root function, that is defined on D = [0, +∞), is continuous at
c = 0. Why? Since for the square root function to be continuous at c = 0, Definition 2.44
requires that there exists δ0 > 0 such that
(0 − δ0 , 0 + δ0 ) = (−δ0 , δ0 ) ⊆ [0, +∞).
Such δ0 > 0 does not exist. So, in the sense of Definition 2.44, the square root function
is not continuous at c = 0. Since our intuitive sense of continuity expects that the square
root function is not continuous at c = 0, the above definition needs to be modified. A
modification is presented as Definition 2.46 below.
Notice that Definition 2.46 requires that a function is defined on an interval of real
numbers. Before stating Definition 2.46 we review nine kinds of intervals of real numbers
that one can encounter.
Recall that there are four kinds of finite intervals; with a, b ∈ R and a < b, the finite
intervals are:
(a, b), (a, b], [a, b), [a, b].
There are four kinds of infinite intervals; with a ∈ R, the infinite intervals are:
(a, +∞), [a, +∞), (−∞, a), (−∞, a];
and also R is an infinite interval, sometimes written as (−∞, +∞).
 2.4. CONTINUOUS FUNCTIONS 41

Definition 2.46. Let D ⊆ R be an interval. A function f : D → R is continuous on

D if the following condition is satisfied:

∀c ∈ D ∀ǫ > 0 ∃δ(ǫ, c) > 0 such that ∀x ∈ D we have

|x − c| < δ(ǫ, c) ⇒ |f (x) − f (c)| < ǫ.
Definition 2.46 is also an ǫ-δ definition of continuity. The advantage of Definition 2.46
is that it defines continuity of a function defined on an interval in one statement, not
two statements as in Definitions 2.44 and 2.45. If you are working with a function which is
defined on an open interval you can use either of the definitions. In fact, these two definitions
are equivalent if a function is defined on an open interval. However, in most cases doing
a proof using Definition 2.46 might be somewhat easier. It is a prudent proof strategy to
always have in mind both definitions. Then, when writing the final proof you write the
proof which will satisfy Definition 2.46. A good example of this strategy is Example 2.49
below.
Example 2.47. Let K be a real number and define f (x) = K for all x ∈ R. Use
Definition 2.46 to prove that f is continuous on R.
Example 2.48. Let f (x) = x for all x ∈ R. Use Definition 2.46 to prove that f is
continuous on R.
Example 2.49. Use ǫ-δ definition of continuity, that is Definition 2.46, to prove that
the function f (x) = 1/x is continuous on the interval (0, +∞).
Solution. It is interesting that we start this proof as if we are using Definition 2.44.
Then, after we find δ(ǫ, c) > 0, we do the final proof which proves the statement in Defini-
tion 2.46.
Let c ∈ (0, +∞), that is let c be an arbitrary positive number. Chose δ0 =
c/2. Since
c > 0, we conclude that c/2 > 0 and f (x) = 1/x is defined for all x ∈ c/2, 3c/2 .
Let ǫ > 0 be arbitrary. Now we have to solve
1 1
− <ǫ for |x − c|.
x c
First simplify the expression, using the fact that x > 0 and c > 0 and rules for the absolute
value:
1 1 c−x |c − x| |x − c|
− = = = .
x c xc |x| |c| xc
To get a larger expression which will be easy to solve we replace x in the denominator by
the smallest possible value for x. That value is c − c/2 = c/2. This gives me my BIN:
1 1 |x − c| |x − c| |x − c|
− = ≤ c =2 .
x c xc c c2
2
1 1 2

Thus my BIN is − ≤ 2 |x − c| valid for all x ∈ c/2, 3c/2 .
x c c
2
Next we solve the inequality 2 |x − c| < ǫ for |x − c|. Since c > 0 we have
c
2 c2 ǫ
|x − c| < ǫ ⇔ |x − c| < .
c2 2
42 2. LIMITS

We are ready to define  

c2 ǫ c
δ(ǫ) = min , .
2 2
To finish the proof, it remains to prove the implication
 2 
c ǫ c 1 1
∀c > 0 ∀x > 0 |x − c| < min , ⇒ − < ǫ.
2 2 x c
Using the BIN and the preceding displayed equivalence you can prove this implication as
an exercise. 
Example 2.50.√ Use ǫ-δ definition of continuity, that is Definition 2.46, to prove that
the function x 7→ x is continuous on the interval (0, +∞).
Solution. It is interesting that we start this proof as if we are using Definition 2.44.
Then, after we find δ(ǫ, c) > 0, we do the final proof which proves the statement in Defini-
tion 2.46.
c c
Let c ∈ (0, +∞). Chose δ0 = . Since c > 0, as before we conclude that > 0 and the
√ 2 2
function x 7→ x is defined for all x ∈ (c/2, 3c/2).
Let ǫ > 0 be arbitrary. Now we have to solve
√ √
x − c < ǫ for |x − c|.
First simplify algebraically the expression, using the fact that x > 0 and c > 0 and rules
for the absolute value and the Pizza-Party to get:
√ √
√ √ √ √
1 √ √
x+ c x−c
x− c = x− c = x− c √ √ = √ √
1 x+ c x+ c
|x − c| |x − c| |x − c|
= √ √ =√ √ ≤ √
| x + c| x+ c c
√ √ |x − c|
Thus the BIN is: x − c ≤ √ , valid for all x > 0 and all c > 0.
c
|x − c|
Next we solve √ < ǫ for |x − c|. The solution is the following equivalence: Since
c
c > 0 we have
|x − c| √
√ < ǫ ⇔ |x − c| < c ǫ. (2.71)
c
Since the BIN is valid for all c > 0 we can define
√
δ(ǫ) = c ǫ.
It remains to prove the implication
√ √ √
∀c > 0 ∀x > 0 |x − c| < cǫ ⇒ x− c < ǫ.
As usual, this is done using the BIN and
√ the equivalence in (2.71). Let c > 0 and x > 0
be arbitrary. Assume that |x − c| < c ǫ. By the equivalence in (2.71) we deduce that
|x − c| √ √ |x − c|
√ < ǫ holds. By the BIN we have x − c ≤ √ . By the transitivity of order
c √ √ c
from the last two inequalities we deduce that x − c < ǫ. 
1
Example 2.51. Let f (x) = for all x ∈ R. Use ǫ-δ definition to prove that f is
x2 +1
continuous on its domain.
 2.4. CONTINUOUS FUNCTIONS 43

Example 2.52. Let a, b, c be any real numbers. Let f (x) = ax2 + bx + c for all x ∈ R.
Let v be an arbitrary real number. Prove that f is continuous at v.
Example 2.53. Let f (x) = sin x for all x ∈ R. Prove that f is continuous at an
arbitrary real number a.
Example 2.54. Let f (x) = cos x for all x ∈ R. Prove that f is continuous at an
arbitrary real number a.
Hint for Exercises 2.53 and 2.54. Let A = (x1 , y1 ) and B = (x2 , y2 ) be two points in the
xy-plane. Then the length of the line segment AB is given by
p
AB = (x1 − x2 )2 + (y1 − y2 )2 .
Consequently
|x1 − x2 | ≤ AB and |y1 − y2 | ≤ AB.
Let u and v be real numbers and set A = (cos u, sin u), B = (cos v, sin v). The last displayed
inequalities now imply
| cos u − cos v| ≤ AB and | sin u − sin v| ≤ AB.
Recall that the points A and B are on the unit circle. Any two points on the unit circle
⌢
determine two arcs. Denote by AB the length of the shorter circular arc determined by
A and B. Since the shortest path between two points is a straight line we have that
⌢ ⌢
AB < AB . How is the arc length AB related to the numbers u and v? First, if |u − v| ≤ π,
⌢ ⌢
then AB = |u − v|. Second, if |u − v| > π, then AB ≤ π < |u − v|. Hence in each case
⌢
AB ≤ |u − v|. Thus we have established inequalities
⌢
| cos u − cos v| ≤ AB ≤ AB ≤ |u − v|,
⌢
| sin u − sin v| ≤ AB ≤ AB ≤ |u − v|,
for arbitrary real numbers u and v. These inequalities can be used to solve Exercises 2.53
and 2.54. The end of the Hint.
Example 2.55. Let f (x) = ln x for all x ∈ (0, +∞). Prove that f is continuous on its
domain.
Solution. First we recall the inequality
1
1 − ≤ ln v ≤ v − 1 valid for all v > 0, (2.72)
v
which we proved using the integral definition of ln.
An inequality for | ln v| will be useful in the proof of the continuity below. Such an
inequality can be obtained from the inequality in (2.72) by considering two cases:
( )
v−1 if 1 ≤ v
| ln v| ≤

− 1 − 1v if 0 < v < 1
( )
v − 1 if 1 ≤ v
=
− v−1
v if 0 < v < 1
( )
|v − 1| if 1 ≤ v
= |v−1|
.
v if 0 < v < 1
44 2. LIMITS

Next we will restrict v to the interval (1/2, 3/2). That is we assume v ∈ (1/2, 3/2).
Then we have that |v − 1|/v ≤ 2|v − 1|. Since always |v − 1| ≤ 2|v − 1|, we have that
| ln v| ≤ 2|v − 1| is valid for all v ∈ (1/2, 3/2). (2.73)
Let a > 0 be arbitrary. Let x ∈ (a/2, 3a/2). Then x/a ∈ (1/2, 3/2) and we can simplify
the expression | ln x−ln a| which appears in the definition of continuity. In the next sequence
of inequalities we first use a property of logarithm, then the inequality in (2.73) and simple
algebra to get:
x
| ln x − ln a| = ln
a
x
≤2 −1
a
x−a
=2
a
|x − a|
=2
a
2
= |x − a|.
a
Thus, we proved that
2
| ln x − ln a| ≤|x − a| is valid for all x ∈ (a/2, 3a/2). (2.74)
a
To finish the proof of continuity let ǫ > 0 be arbitrary and set
n aǫ a o
δ(ǫ) = min , .
2 2
Clearly δ(ǫ) > 0. Next we will prove the implication
n aǫ a o
|x − a| < min , ⇒ | ln x − ln a| < ǫ.
2 2
 aǫ a
Assume |x − a| < min 2 , 2 . Then |x − a| < aǫ a a
2 and |x − a| < 2 . Since |x − a| < 2 , we
have x ∈ (a/2, 3a/2) and therefore, by (2.74), we have
2
| ln x − ln a| ≤ |x − a|.
a
aǫ
Since |x − a| < 2 we have
2
|x − a| < ǫ.
a
The last two displayed inequalities yield
| ln x − ln a| < ǫ.
This completes the proof of the continuity of the logarithm function ln. 
Example 2.56. Let f (x) = ex for all x ∈ R. Prove that f is continuous at an arbitrary
real number a.
Solution. We first substitute v = exp u = eu in (2.72) to get
1
1− ≤ ln eu ≤ eu − 1 is valid for all u ∈ R.
eu
 2.4. CONTINUOUS FUNCTIONS 45

Simplifying we get
1
1− ≤ u ≤ eu − 1.
eu
We need a squeeze for eu . Above we already have one side of the squeeze. That is u+1 ≤ eu .
To get the other side we transform
1
1− u ≤u
e
to
1
1 − u ≤ u.
e
To get a useful inequality we need to take the reciprocals in the last inequality. For that
we need 1 − u > 0. That is we need to assume that u < 1. Assuming that u < 1 we have
1
eu ≤ .
1−u
Together with u + 1 ≤ eu , we proved that
1
u + 1 ≤ eu ≤ is valid for all u < 1. (2.75)
1−u
An inequality for |eu − 1| will be useful in the proof of the continuity below. The
inequalities in (2.75) yield that
u
u ≤ eu − 1 ≤ is valid for all u < 1.
1−u
To get an inequality for |eu − 1| we consider two cases:
 u 
u 1−u if 0 ≤ u < 1
|e − 1| ≤
−u if u < 0
 |u| 
if 0 ≤ u < 1
= 1−u
|u| if u < 0
Next we will restrict u to the interval (−1/2, 1/2). That is we assume u ∈ (−1/2, 1/2).
Then we have that |u|/(1 − u) ≤ 2|u|. Since always |u| ≤ 2|u|, we have that
|eu − 1| ≤ 2|u| is valid for all u ∈ (−1/2, 1/2). (2.76)
Let a > 0 be arbitrary. Let x ∈ (a − 1/2, a + 1/2). Then x − a ∈ (−1/2, 1/2) and we
can simplify the expression |ex − ea | which appears in the definition of continuity. For that
we use a property of the exponential function and (2.76) to get:
|ex − ea | = ea |e(x−a) − 1| ≤ 2ea |x − a|.
Thus, we proved that
|ex − ea | ≤ 2ea |x − a| is valid for all x ∈ (a − 1/2, a + 1/2). (2.77)
To finish the proof of the continuity let ǫ > 0 be arbitrary and set
 
ǫ 1
δ(ǫ) = min , .
2ea 2
Clearly δ(ǫ) > 0.
Next we will prove the implication
 
ǫ 1
|x − a| < min , ⇒ |ex − ea | < ǫ.
2ea 2
46 2. LIMITS

Assume |x − a| < min 2eǫa , 12 . Then |x − a| < 2eǫa and |x − a| < 12 . Since |x − a| < 12 , we
have x ∈ (a − 1/2, a + 1/2) and therefore, by (2.77), we have
|ex − ea | ≤ 2ea |x − a|.
ǫ
Since |x − a| < 2ea we have
2ea |x − a| < ǫ.
The last two displayed inequalities yield
|ex − ea | < ǫ.
This completes the proof of the continuity of of the exponential function exp. 

2.4.2. General theorems about continuous functions. The next theorem can be
deduced from Theorem 2.41.
Theorem 2.57 (Algebra of Continuous Functions). Let f and g be functions and let a
be a real number. Assume that f and g are continuous at the point a.
(a) If h = f + g, then h is continuous at a.
(b) If h = f g, then h is continuous at a.
f
(c) If h = and g(a) 6= 0, then h is continuous at a.
g
π π
Example 2.58. Let f (x) = tan x for all − < x < . Prove that f is continuous at
2 2
π π
an arbitrary real number a such that − < a < .
2 2
Solution. Use the algebra of continuous functions. 

The following theorem states that a composition of continuous functions is continuous.

Theorem 2.59. Let f and g be functions and let a be a real number. Assume that g is
continuous at a and that f is continuous at g(a). If h = f ◦ g, then h is continuous at a.
Proof. Assume that the function g is continuous at a. That is assume
(I-g) There exists a δ0,g > 0 such that g(x) is defined for all x ∈ (a − δ0,g , a + δ0,g ).
(II-g) For every ǫ > 0 there exists δg (ǫ) such that 0 < δg (ǫ) ≤ δ0,g and such that
|x − a| < δg (ǫ) ⇒ |g(x) − g(a)| < ǫ.
Also assume that the function f is continuous at g(a). That is assume


(I-f) There exists a δ0,f > 0 such that f (x) is defined for all x ∈ g(a) − δ0,f , g(a) + δ0,f .
(II-f) For every ǫ > 0 there exists δg (ǫ) such that 0 < δf (ǫ) ≤ δ0,f and such that
|u − g(a)| < δf (ǫ) ⇒ |f (u) − f (g(a))| < ǫ.
Let h = f ◦ g, that is h(x) = f (g(x)). I have to prove that h has the following properties:
(These items are red.)


(I-h) There exists a δ0,h > 0 such that h(x) is defined for all x ∈ a − δ0,h , a + δ0,h .
(II-h) For every ǫ > 0 there exists δh (ǫ) such that 0 < δh (ǫ) ≤ δ0,h and such that
|x − a| < δh (ǫ) ⇒ |h(x) − h(a)| < ǫ.
 2.4. CONTINUOUS FUNCTIONS 47

Where is h guaranteed to be defined? I must make sure that x is such that |g(x) − g(a)| <
δ0,f . We can achieve this

by using (II-g)!
Put δ0,h = δg δ0,f . Now assume that |x − a| < δ
0,h . By (II-g) it follows that |g(x) −
g(a)| < δ0,f . Therefore g(x) ∈ g(a) − δ0,f , g(a) + δ0,f . Hence, by (I-f), f (g(x)) is defined.
Thus we proved that f (g(x)) is defined whenever |x − a| < δ0,h .
Let ǫ > 0 be given. Put



δh (ǫ) = min δg δf (ǫ) , δg δ0,f .
Now we prove the red implication in (II-h).

Assume |x − a| < δh (ǫ). Then |x − a| < δg δf (ǫ) . By the green implication in (II-g),
we conclude that


|x − a| < δg δf (ǫ) ⇒ |g(x) − g(a)| < δf (ǫ).
Using the green implication in (II-f), we conclude that
|g(x) − g(a)| < δf (ǫ) ⇒ |f (g(x)) − f (g(a))| < ǫ.
Thus we proved that the assumption |x − a| < δh (ǫ) implies that
|h(x) − h(a)| = |f (g(x)) − f (g(a))| < ǫ.
This completes the proof. 
 CHAPTER 3

Infinite Series

3.1. Sequences of real numbers

3.1.1. Definitions and examples.
Definition 3.1. A sequence of real numbers is a real function whose domain is either
the set N of positive integers or the set N0 of nonnegative integers.
Let s : N → R be a sequence. Then the values of s are s(1), s(2), s(3), . . . , s(n), . . . . It is
customary to write sn instead of s(n) in this case. Sometimes a sequence will be specified
by listing its first few terms
s1 , s2 , s3 , s4 , . . . ,
and sometimes by listing of all its terms {sn }n∈N or simply {sn } since domain is clear. One
way of specifying a sequence is to give a formula, or recursion formula for its n−th term
sn . Notice that in this notation s is the “name” of the sequence and n is the variable.
Some examples of sequences follow.
Example 3.2. (a) 1, 0, −1, 0, 1, 0, −1, . . . ;
(b) 1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 7, 7, . . . ;
(c) 1, 1, 1, 1, 1, . . . ; (the constant sequence)
1 1 2 1 3 1 2 3 4 1 5 1 2 3 4 5 6 1 3 5 7
(d) , , , , , , , , , , , , , , , , , , , , , . . . ; (What
2 3 3 4 4 5 5 5 5 6 6 7 7 7 7 7 7 8 8 8 8
is the range of this sequence?)
Recursively defined sequences
xn
Example 3.3. (a) x1 = 1, xn+1 = 1 + , n ∈ N;
4
xn 1
(b) x1 = 2, xn+1 = + , n ∈ N;
√ 2√ xn
(c) a1 = 2, an+1 = √ 2 + an , n ∈ N;
(d) s1 = 1, sn+1 = 1 + sn , n ∈ N;
9 9 + xn
(e) x1 = , xn+1 = , n ∈ N.
10 10
1 1
(f) b1 = , bn+1 = p , n∈N
2 2 1 − b2n
(g) f0 = 1, fn = n · fn−1 , n ∈ N.
The standard notation for the terms of the sequence in (g) f : N0 → R is fn = n!, n ∈ N0 .
Below we present more examples of important sequences.
Example 3.4. (a) For c ∈ R set bn = c for all n ∈ N. This is a constant sequence.
(b) For a ∈ R recursively define the sequence p0 = 1 and pn = apn−1 for all n ∈ N.
This is the sequence of the powers of a commonly written as pn = an for all n ∈ N0 .
49
50 3. INFINITE SERIES

(c) A remarkable property of the two sequences given in this item is that they converge
to the same limit: the famous real number e. The first sequence is given by a
formula  
1 n
qn = 1 + for all n ∈ N,
n
and the second by a recursive definition
1 1
s0 = = 1, sn = sn−1 + with n ∈ N
0! n!
The n-th term of the preceding sequence is often written as
n
X 1 1 1 1 1
sn = = + + + ··· + .
n! 0! 1! 2! n!
k=0
We will hopefully have time to rigorously prove the claim in this item.
(d) A remarkable property of the two sequences given in this item is that they converge
x
to the same limit: the famous exponential function e . Let x ∈ R and define the
first sequence is given by a formula
 x n
qn = 1 + for all n ∈ N,
n
and the second by a recursive definition
1 xn
s0 = = 1, sn = sn−1 + with n ∈ N
0! n!
The n-th term of the preceding sequence is often written as
n
X xn 1 x x2 xn
sn = = + + + ··· + .
n! 0! 1! 2! n!
k=0
We might have time to explain why the claim in this item is true.
(e) The recursively defined sequences named s in the last two items are examples of a
general recursive pattern which we explain here. Let a : N0 → R be an arbitrary
sequence. An important recursively defined sequence associated with a : N0 → R
is the following sequence:
S0 = a0 , Sn = Sn−1 + an with n ∈ N.
The n-th term of the preceding sequence is often written as
Xn
Sn = an = a0 + a1 + a2 + · · · + an .
k=0

3.1.2. Convergent sequences.

Definition 3.5. A sequence s : N → R of real numbers converges to the real number L
if for every ǫ > 0 there exists a real number N (ǫ) such that
∀n ∈ N n > N (ǫ) ⇒ |sn − L| < ǫ.
If s : N → R converges to L we will write
lim sn = L or sn → L (n → +∞).
n→+∞
The number L is called the limit of the sequence s : N → R.
 3.1. SEQUENCES OF REAL NUMBERS 51

Definition 3.6. A sequence s : N → R converges if there exists L ∈ R such that

lim sn = L. In other words, a sequence s : N → R converges if
n→+∞

∃ L ∈ R s.t. ∀ ǫ > 0 ∃N (ǫ) ∈ R s.t. ∀n ∈ N n > N (ǫ) ⇒ |sn − L| < ǫ.

A sequence that does not converge is said to diverge.
Example 3.7. Let r be a real number such that |r| < 1. Prove that limn→+∞ r n = 0.
Solution. First note that if r = 0, then r n = 0 for all n ∈ N, so the given sequence is
a constant sequence. Therefore it converges to 0. Assume that r ∈ (−1, 0) ∪ (0, 1), that is
0 < |r| < 1 and let ǫ > 0 be arbitrary. We need to solve |r n − 0| < ǫ for n. First simplify
|r n − 0| = |r n | = |r|n . Now solve |r|n < ǫ by taking ln of both sides of the inequality (note
that ln is an increasing function)
ln |r|n = n ln |r| < ln ǫ.
Since 0 < |r| < 1, we conclude that ln |r| < 0. Therefore the solution is
ln ǫ
n> .
ln |r|
ln ǫ
Thus, with N (ǫ) = , the implication
ln |r|
∀n ∈ N n > N (ǫ) ⇒ |r n − 0| < ǫ
is valid. 
n2 − n − 1 1
Example 3.8. Prove that lim = .
n→+∞ 2n2 − 1 2
n2 − n − 1 1
Solution. Let ǫ > 0 be arbitrary. We need to solve − < ǫ for n. First
2n2 − 1 2
simplify:
n2 − n − 1 1 2 n2 − n − 1 1 2n2 − 1 −2n − 1 2n + 1
2
− = 2
− 2
= 2
= 2
2n − 1 2 2 2n − 1 2 2n − 1 2 (2n − 1) 4n − 2
Now invent the BIN:
2n + 1 2n + n 3n 3
2
≤ 2 2
= 2 = .
4n − 2 4n − 2n 2n 2n
Therefore the BIN is:
n2 − n − 1 1 3
2
− ≤ valid for all n ∈ N.
2n − 1 2 2n
Solving for n is now easy:
3 3
< ǫ. The solution is n > .
2n 2ǫ
3
Thus, with N (ǫ) = , the implication
2ǫ
n2 − n − 1 1
∀n ∈ R n > N (ǫ) ⇒ − <ǫ
2n2 − 1 2
is valid. (This implication is proved by using the BIN) 
52 3. INFINITE SERIES

3.1.3. Theorems about convergent sequences. The procedure of proving limits of

sequences is very similar to the procedure for proving limits of functions as x approaches
infinity. In fact the following two theorems are true.
Theorem 3.9. Let f : [1, +∞) → R be a function and define the sequence a : N → R by
an = f (n) for every n ∈ N.
If lim f (x) = L, then lim an = L.
x→+∞ n→+∞

Theorem 3.10. Let f : (0, 1] → R be a function which is defined for every x ∈ (0, 1].
Define the sequence a : N → R by
an = f (1/n) for every n ∈ N.
If lim f (x) = L, then lim an = L.
x↓0 n→+∞

The above two theorems are useful for proving limits of sequences which are defined by
a formula. For example you can prove the following limits by using these two theorems and
what we proved in previous sections.
Exercise 3.11. Find
 the following limits. Provide
  proofs.  
1 1 1
(a) lim sin (b) lim n sin (c) lim ln 1 +
n→+∞
n  n→+∞
 n  n→+∞
 n
1 1 1 1
(d) lim n ln 1 + (e) lim cos (f) lim cos
n→+∞ n n→+∞ n n→+∞ n n
In the following theorem we prove that the operation of taking the limit of a sequence
respects the algebra of real numbers. The theorem is called the Algebra of Limits Theorem.
Theorem 3.12. Let a : N → R, b : N → R and c : N → R be given sequences. Let K
and L be real numbers. Assume that
(1) lim an = K,
x→+∞
(2) lim bn = L.
x→+∞
Then the following statements hold.
(A) If cn = an + bn , n ∈ N, then lim cn = K + L.
x→+∞
(B) If cn = an bn , n ∈ N, then lim cn = KL.
x→+∞
an K
(C) If L 6= 0 and cn = , n ∈ N, then lim cn = .
bn x→+∞ L
Proof. To prove (A) assume that
lim an = K, lim bn = L, and ∀n ∈ N cn = an + bn .
x→+∞ x→+∞
By the definition of limit we have
∀ǫ > 0 ∃Na (ǫ) ∈ R such that ∀n ∈ N n > Na (ǫ) ⇒ |an − K| < ǫ (3.1)
and
∀ǫ > 0 ∃Nb (ǫ) ∈ R such that ∀n ∈ N n > Nb (ǫ) ⇒ |bn − L| < ǫ. (3.2)
Let ǫ > 0 be arbitrary. Define

Nc (ǫ) = max Na (ǫ/2), Nb (ǫ/2) .
 3.1. SEQUENCES OF REAL NUMBERS 53

Let n ∈ N be arbtrary. Assume

n > Nc (ǫ).
Then by the definition of Nc (ǫ) we have
n > Na (ǫ/2) and n > Nb (ǫ/2).
By (3.1) we have that
n > Na (ǫ/2) ⇒ |an − K| < ǫ/2.
By (3.2) we have that
n > Nb (ǫ/2) ⇒ |bn − L| < ǫ/2.
Therefore
n > Nc (ǫ) ⇒ |an − K| < ǫ/2 and |bn − L| < ǫ/2. (3.3)
By algebra and the Triangle Inequality we have that for all n ∈ N we have
cn − (K + L) = an + bn − K − L = (an − K) + (bn − L) ≤ an − K + bn − L (3.4)
From (3.3) and (3.4) and the transitivity of order we deduce that
n > Nc (ǫ) ⇒ |cn − (K + L)| < ǫ.
Thus we have proved that the following statement is true
∀ǫ > 0 ∃Nc (ǫ) ∈ R such that ∀n ∈ N n > Nc (ǫ) ⇒ |cn − (K + L)| < ǫ.
Therefore
lim cn = K + L. 
x→+∞

Theorem 3.13. Let a : N → R and b : N → R be given sequences. Let K and L be

real numbers. Assume that
(1) lim an = K.
x→+∞
(2) lim bn = L.
x→+∞
(3) There exists a positive integer n0 such that
∀n ∈ N such that n ≥ n0 we have a n ≤ bn .
Then K ≤ L.
Proof. Assume (1), (2) and (3). Let ǫ > 0 be arbitrary. Since lim an = K, there
x→+∞
exists Na (ǫ) such that
∀n ∈ N and n > Na (ǫ) ⇒ an − K < ǫ.
Since lim bn = L, there exists Nb (ǫ) such that
x→+∞

∀n ∈ N
and n > Nb (ǫ) ⇒ bn − L < ǫ.

Choose m ∈ N such that m > max n0 , Na (ǫ), Nb (ǫ) . Then
K − ǫ < am < K + ǫ
am ≤ bm
L − ǫ < bm < L + ǫ.
Consequently, by the transitivity axiom for real numbers we have
K − ǫ < am ≤ bm < L + ǫ.
54 3. INFINITE SERIES

Hence
K − L < 2ǫ.
Now recall that ǫ > 0 was arbitrary. Thus, the inequality K − L < 2ǫ holds for all ǫ > 0.
The following implication is true
∀ǫ > 0 K − L < 2ǫ ⇒ K − L ≤ 0.
(This implication is proved by proving its contrapositive.) Hence, we conclude that K −L ≤
0. 
Theorem 3.14. Let a : N → R, b : N → R and s : N → R be given sequences. Let L be
a real number. Assume the following
(1) The sequence a : N → R converges to L.
(2) The sequence b : N → R converges to L.
(3) There exists a positive integer n0 such that
a n ≤ s n ≤ bn for all n > n0 .
Then the sequence s : N → R converges to L.
Prove this theorem.
3.1.4. The Monotone Convergence Theorem. Many limits of sequences cannot
be found using theorems from the previous section. For example, the recursively defined
sequences (a), (b), (c), (d) and (e) in Example 3.3 converge but it cannot be proved using
the methods that we presented so far.
Definition 3.15. (1) A sequence s : N → R is bounded above if there exists a real
number M such that
∀ n ∈ N sn ≤ M.
A number M with the above property is called an upper bound of the sequence s.
(2) A sequence s : N → R is bounded below if there exists a real number m such that
∀n ∈ N m ≤ sn .
A number m with the above property is called a lower bound of the sequence s.
(3) A sequence s : N → R is bounded if it is bounded above and bounded below. In
other words, a sequence s : N → R is bounded if there exist real numbers m and
M such that
∀ n ∈ N m ≤ sn ≤ M.
Theorem 3.16. If a sequence converges, then it is bounded.
Proof. Assume that a sequence a : N → R converges to a real number L. We need to
prove that there exist real numbers m and M such that
∀n ∈ N we have m ≤ an ≤ M. (3.5)
Since a : N → R converges to L, Definition 3.5 yields that for every ǫ > 0 there exists a real
number N (ǫ) such that
∀n ∈ N n > N (ǫ) ⇒ |an − L| < ǫ.
In particular for ǫ = 1 > 0 there exists a real number N (1) such that
∀n ∈ N n > N (1) ⇒ |an − L| < 1.
 3.1. SEQUENCES OF REAL NUMBERS 55

Since |an − L| < 1 is equivalent to L − 1 < an < L + 1, the preceding implication can be
rewritten as
∀n ∈ N n > N (1) ⇒ L − 1 < an < L + 1. (3.6)
Case 1. Assume that N (1) < 1. Then for all n ∈ N we have n > N (1). Therefore (3.6)
yields
∀n ∈ N we have L − 1 < an < L + 1.
Thus, we can take m = L − 1 and M = L + 1 and (3.5) holds.
Case 2. Assume that N (1) ≥ 1. Set n0 = ⌊N (1)⌋. Then n0 is a positive integer with the
following property
n0 ≤ N (1) < n0 + 1. (3.7)
The preceding inequality suggests a partition of the set N in two disjoint sets
 
N = 1, 2, . . . , n0 ∪ k ∈ N : k > n0 . (3.8)

Thefirst set 1, 2, . . . , n0 in the preceding union is finite and has n0 elements. The second
set k ∈ N : k > n0 is infinite and consists of the positive integers n0 + 1, n0 + 2, . . .. It
follows from (3.7) that

∀n ∈ k ∈ N : k > n0 we have n > N (1).
Therefore (3.6) yields

∀n ∈ k ∈ N : k > n0 we have L − 1 < an < L + 1. (3.9)
The number L − 1 is not necessarily a lower bound and L + 1 is not necessarily an upper
bound for the sequence since we do not know whether the relation of L − 1 and L + 1 to
the terms
a1 , a2 , . . . , an0 .
Since every finite set has a minimum and a maximum we set

m = min a1 , a2 , . . . , an0 , L − 1

and

M = max a1 , a2 , . . . , an0 , L + 1 .
Now we can prove that m is a lower bound and M is an upper bound for the sequence
a : N → R. By the definitions of the minimum and the maximum we have

∀n ∈ 1, 2, . . . , n0 we have m ≤ an ≤ M.
The definitions of the minimum and the maximum also imply that m ≤ L−1 and L+1 ≤ M .
Using these inequalities, the transitivity of order and (3.9) we obtain

∀n ∈ k ∈ N : k > n0 we have m < an < M.
Because of (3.8) the preceding two displayed statements yield
∀n ∈ N we have m ≤ an ≤ M.
Hence the sequence a : N → R is bounded. 
56 3. INFINITE SERIES

Is the converse of Theorem 3.16 true? The converse is: If a sequence is bounded, then
it converges. This statement is not true since there exists a sequence that is bounded and
which does not converge. One such sequence is n 7→ (−1)n for all n ∈ N. This sequence is
bounded and it is not convergent. Thus, we found a counterexample to the implication: If
a sequence is bounded, then it converges.
The next question is whether boundedness and an additional property of a sequence
can guarantee convergence. It turns out that such an property is monotonicity defined in
the following definition.
Definition 3.17. A sequence s : N → R of real numbers is said to be:
non-decreasing if sn ≤ sn+1 for all n ∈ N,
non-increasing if sn ≥ sn+1 for all n ∈ N.
A sequence with either of these four properties is said to be monotonic.
The following theorem is the Monotone Convergence Theorem.
Theorem 3.18 (Monotone Convergence Theorem). A bounded monotonic sequence con-
verges.
To prove these theorems we have to resort to the most important property of the set of
real numbers: the Completeness Axiom.
The Completeness Axiom. If A and B are nonempty subsets of R such that for every
a ∈ A and for every b ∈ B we have a ≤ b, then there exists c ∈ R such that a ≤ c ≤ b for
all a ∈ A and all b ∈ B.
Proof of Theorem 3.18. Assume that s : N → R is a non-decreasing sequence and
that it is bounded above. Since s : N → R is non-decreasing we know that
s1 ≤ s2 ≤ s3 ≤ · · · ≤ sn−1 ≤ sn ≤ sn+1 ≤ · · · . (3.10)
Let A be the range of the sequence s : N → R. That is let

A = sn : n ∈ N .
Clearly A 6= ∅. Let B be the set of all upper bounds of the sequence s : N → R. Since
the sequence s : N → R is bounded above, the set B is not empty. Let b ∈ B be arbitrary.
Then b is an upper bound for s : N → R. Therefore
sn ≤ b for all n ∈ N.
By the definition of A this means
a≤b for all a ∈ A.
Since b ∈ B was arbitrary we have
a≤b for all a ∈ A and for all b ∈ B.
By the Completeness Axiom there exists c ∈ R such that
sn ≤ c ≤ b for all n ∈ N and for all b ∈ B. (3.11)
Thus c is an upper bound for s : N → R and also c ≤ b for all upper bounds b of the
sequence s : N → R. Therefore, for an arbitrary ǫ > 0 the number c − ǫ (which is < c) is not
 3.1. SEQUENCES OF REAL NUMBERS 57

an upper bound of the sequence s : N → R. Consequently, there exists a positive integer

N (ǫ) such that
c − ǫ < sN(ǫ) . (3.12)
Let n ∈ N be any positive integer which is > N (ǫ). Then the inequalities (3.10) imply
that
sN(ǫ) ≤ sn . (3.13)
By (3.11) the number c is an upper bound of s : N → R. Hence we have
sn ≤ c for all n ∈ N. (3.14)
Putting together the inequalities (3.12), (3.13) and (3.14) we conclude that
c − ǫ < sn ≤ c for all n ∈ N such that n > N (ǫ). (3.15)
The relationship (3.15) shows that for n ∈ N such that n > N (ǫ) the distance between
numbers sn and c is < ǫ. In other words
∀n ∈ N n > N (ǫ) ⇒ |sn − c| < ǫ.
This is exactly the implication in Definition 3.5. Thus, we proved that
lim sn = c. 
n→+∞

Example 3.19. Prove that the sequence in Example 3.3 (b) converges. That is, prove
xn 1
that the recursively defined sequence x1 = 2, xn+1 = + , n ∈ N, converges.
2 xn
Solution. It is useful to calculate the first few terms of this sequence:
3 17 577 665857 886731088897
x1 = 2, x2 = , x3 = , x4 = , x5 = , x6 = .
2 12 408 470832 627013566048
Notice that the formula xn+1 = x2n + x1n gives a positive output xn+1 whenever the
input xn is positive. Since x1 > 0 this guaranties that x2 > 0. In turn, the fact that x2 > 0
guaranties that x3 > 0, and so on. This reasoning justifies that xn > 0 for all n ∈ N. This
proves that the sequence {xn } is bounded below by 0.
Next we will prove that (xn )2 ≥ 2 for all n ∈ N. We consider two cases n = 1 and
n > 1. If n = 1, then (x1 )2 = 22 = 4 ≥ 2. Now assume that n > 1. Then n − 1 ∈ N and
xn = xn−1 1
2 + xn−1 . Therefore
x 1 2
n−1
(xn )2 = +
2 xn−1
(xn−1 )2 1
= +1+
4 (xn−1 )2
(xn−1 )2 1
=2+ −1+
4 (xn−1 )2
x 1 2 
n−1
=2+ −
2 xn−1
≥ 2.
Thus (xn )2 ≥ 2 for all n ∈ N.
Since xn > 0, (xn )2 ≥ 2 implies xn ≥ x2n . Further, dividing by 2 we get x2n ≥ x1n . Adding
xn xn 1
2 to the both sides of the last inequality we obtain xn ≥ 2 + xn . Thus xn ≥ xn+1 . Here
58 3. INFINITE SERIES

we have proved that (xn )2 ≥ 2 implies xn ≥ xn+1 . Since (xn )2 ≥ 2 is true for all n ∈ N, we
have proved that xn ≥ xn+1 is true for all n ∈ N.
To summarize, we have proved that xn > 0 for all n ∈ N and xn ≥ xn+1 is true for all
n ∈ N. That is, the sequence {xn } is bounded below and non-increasing. By the Monotone
Convergence Theorem this sequence converges. Denote the limit of {xn } by L.

2
Next we use the algebra of limits to calculate L. Since xn ≥ 2 for all n ∈ N, by
Theorems 3.12 and 3.13 we have L2 ≥ 2. Since xn > 0 for all n ∈ N, by Theorem 3.13
we have L ≥ 0. Since L2 ≥ 2 and L ≥ 0 we have L > 0. It is not difficult to prove
that limn→∞ xn+1 = L. This fact, Theorem 3.12 and the identity xn+1 = x2n + x1n imply
√
L = L2 + L1 . Hence L2 = 2. That is L = 2.
This example is in fact a proof that there exists a positive real number a such that
2
a = 2. 
n
X 1
Example 3.20. Prove that the sequence Sn = , n ∈ N, converges.
n!
k=0

Solution. Let n ∈ N. We first prove that for n > 1 we have

1 1 1 1
+ + + ··· + < 1.
2! 3! 4! n!
For this we use the fact that for k > 1 we have
1 1 1 1
≤ = −
k! (k − 1)k k−1 k
and therefore
1 1 1 1 1
+ + + ··· + +
2! 3! 4! (n − 1)! n!
         
1 1 1 1 1 1 1 1 1 1
≤ − + − + − + ··· + − + −
1 2 2 3 3 4 n−2 n−1 n−1 n
Since the right-hand side of the preceding displayed inequality simplifies to 1 − 1/n we have
1 1 1 1
+ + + ··· + < 1.
2! 3! 4! n!
Therefore for all n ∈ N
1 1 1 1 1 1
Sn = + + + + + ··· + < 3.
0! 1! 2! 3! 4! n!
This proves that the sequence {Sn } is bounded above. Since for every n ∈ N we have
1
Sn+1 − Sn = > 0, the sequence {Sn } is increasing. By the Monotone Convergence
(n + 1)!
Theorem {Sn } converges.
The limit of the sequence {Sn } is the famous number e. 
Example 3.21. Prove that the sequence
1 1 1 1
tn = 1 + + + + · · · + − ln n, n ∈ N,
2 3 4 n
converges.
 3.2. INFINITE SERIES OF REAL NUMBERS 59

Solution. Let n ∈ N. By the definition of the logarithm function

Z n
1
ln n = dx.
1 x
Since
1 1
≤ whenever k ≤ x ≤ k + 1,
x k
for n > 1 we have
Z 2 Z 3 Z n
1 1 1 1 1 1 1
ln n = dx + dx + · · · + dx < 1 + + + + · · · + .
1 x 2 x n−1 x 2 3 4 n − 1
Therefore,
1 1 1 1 1 1
tn = 1 + + + + ··· + + − ln n > > 0
2 3 4 n−1 n n
for all n ∈ N, n > 1. Since t1 = 1 > 0, this proves that the sequence {tn } is bounded below
by 0.
Next we prove that {tn } is decreasing. For arbitrary n ∈ N we have

1
tn − tn+1 = ln(n + 1) − ln n −
n+1
Z n+1
1 1
= dx −
n x n+1
Z n+1  
1 1
= − dx
n x n+1
> 0.
Hence tn > tn+1 for all n ∈ N.
Since {tn } is bounded below and decreasing it converges by the Monotone Convergence
Theorem.
The limit of the sequence {tn } is called Euler’s constant. It is denoted by γ. Its
approximate value to 50 decimal places is
γ ≈ 0.57721566490153286060651209008240243104215933593992.
It is not known whether γ is a rational or irrational number. 

3.2. Infinite series of real numbers

3.2.1. Definition and basic examples. The most important application of sequences
is the definition of convergence of an infinite series. From the elementary school you have
been dealing with addition of numbers. As you know the addition gets harder as you add
more and more numbers. For example it would take some time to add
S100 = 1 + 2 + 3 + 4 + 5 + · · · + 98 + 99 + 100
It gets much easier if you add two of these sums, and pair the numbers in a special way:
2 S100 = 1 + 2 + 3 + 4 + · · · + 97 + 98 + 99 + 100
100 + 99 + 98 + 97 + · · · + 4 + 3 + 2 + 1
60 3. INFINITE SERIES

A straightforward observation that each column on the right adds to 101 and that there are
100 such columns yields that
101 · 100
2 S100 = 101 · 100, that is S100 = = 5050.
2
This can be generalized to any positive integer n to get the formula
(n + 1) n
Sn = 1 + 2 + 3 + 4 + 5 + · · · + (n − 1) + n = .
2
This procedure indicates that it would be impossible to find the sum
1 + 2 + 3 + 4 + 5 + ··· + n + ···
where the last set of · · · indicates that we continue to add positive integers.
The situation is quite different if we consider the sequence
1 1 1 1 1
, , , , ..., n, ...
2 4 8 16 2
and start adding more and more consecutive terms of this sequence:
1 1 1
=1− =
2 2 2
1 1 1 3
+ =1− =
2 4 4 4
1 1 1 1 7
+ + =1− =
2 4 8 8 8
1 1 1 1 1 15
+ + + =1− =
2 4 8 16 16 16
1 1 1 1 1 1 31
+ + + + =1− =
2 4 8 16 32 32 32
1 1 1 1 1 1 1 63
+ + + + + =1− =
2 4 8 16 32 64 64 64

These sums are nicely illustrated in Fig. 1. The pictures in Fig. 1 strongly indicate that
the sum of infinitely many numbers 21 , 14 , 18 , . . . equals 1. That is
1 1 1 1 1
+ + + + ··· + n + ··· = 1
2 4 8 16 2
Why does this make sense? This makes sense since we have seen above that as we add
more and more terms of the sequence
1 1 1 1 1
, , , , ..., n, ...
2 4 8 16 2
we are getting closer and closer to 1. Indeed,
1 1 1 1 1 1
+ + + + ··· + n = 1 − n
2 4 8 16 2 2
and  
1
lim 1 − n = 1.
n→+∞ 2
This reasoning leads to the definition of convergence of an infinite series:
 3.2. INFINITE SERIES OF REAL NUMBERS 61

1 1 1

2 2 2

1 1

4 4

1 1

3  1

1 1 1 6

8 8 8

1 1 1

1
 

1 1 1

2 2 2

1 1 1

4 4 4

Fig. 1. In this example it seems natural to say that the sum of infinitely many numbers
1 1 1
2 , 4 , 8 , . . . equals 1

Definition 3.22. Let a : N → R be a given sequence. Then the expression

a1 + a2 + a3 + · · · + an + · · ·
is called an infinite series. We often abbreviate it by writing
+∞
X
a1 + a2 + a3 + · · · + an + · · · = an .
n=1
For each positive integer n we calculate the (finite) sum of the first n terms of the series
Sn = a1 + a2 + a3 + · · · + an .
+∞
X
We call Sn a partial sum of the infinite series an . (Notice that {Sn }+∞
n=1 is a new
n=1
sequence.) If the sequence {Sn }+∞
n=1 converges to a real number S, that is if
lim Sn = S,
n→+∞
P+∞
then the infinite series n=1 an is said to be convergent and we write
+∞
X
a1 + a2 + a3 + · · · + an + · · · = S or an = S.
n=1
The number S is called the sum of the series.
If the sequence of the partial sums S : N → R does not converge to a real number, then
the series is called divergent.
62 3. INFINITE SERIES

In the example above we have

 n
1 1
an =
= ,
2n 2
1 2n − 1
Sn = 1 − n =
2  2n 
1
lim Sn = lim 1 − n = 1.
n→+∞ n→+∞ 2
Therefore we say that the series
+∞
1 1 1 1 1 X 1
+ + + + ··· + n + ··· =
2 4 8 16 2 2n
n=1

+∞
X 1
converges and its sum is 1. We write = 1.
n=1
2n
In our opening example
an = n,
(n + 1)n
Sn = 1 + 2 + 3 + · · · + n =
2
(n + 1)n
lim does not exist.
n→+∞ 2
Therefore we say that the series
+∞
X
1 + 2 + 3 + 4 + ··· + n + ··· = n
n=1

diverges.

3.2.2. Geometric Series. Let a and r be real numbers. The most important infinite
series is
+∞
X
2 3 n
a + ar + ar + ar + ··· + ar + ··· = a rn (3.16)
n=0
This series is called a geometric series. To determine whether this series converges or not
we need to study its partial sums:
S0 = a, S1 = a + a r,
S2 = a + a r + a r 2 , S3 = a + a r + a r 2 + a r 3 ,
S4 = a + a r + a r 2 + a r 3 + a r 4 , S5 = a + a r + a r 2 + a r 3 + a r 4 + a r 5 ,
..
.
Sn = a + a r + a r 2 + · · · + a r n−1 + a r n
..
.
 3.2. INFINITE SERIES OF REAL NUMBERS 63

1
Notice that we have already studied the special case when a = 1 and r = . In this
2
special case we found a simple formula for Sn and then we evaluated lim Sn . It turns out
n→+∞
that we can find a simple formula for Sn in the general case as well.
First note that the case a = 0 is not interesting, since then all the terms of the geometric
series are equal to 0 and the series clearly converges and its sum is 0. Assume that a 6= 0.
If r = 1 then Sn = n a. Since we assume that a 6= 0, lim na does not exist. Thus for
n→+∞
r = 1 the series diverges.
Assume that r 6= 1. To find a simple formula for Sn , multiply the long formula for Sn
above by r to get:

Sn = a + ar + ar 2 + · · · + ar n−1 + ar n ,
rSn = ar + ar 2 + · · · + ar n + ar n+1 ;

now subtract,

Sn − r Sn = a − a r n+1 ,

and solve for Sn :

1 − r n+1
Sn = a .
1−r
We already proved that if |r| < 1, then lim r n+1 = 0. If |r| ≥ 1, then lim r n+1 does
n→+∞ n→+∞
not exist. Therefore we conclude that
1 − r n+1 1
lim Sn = lim a =a for |r| < 1,
n→+∞ n→+∞ 1−r 1−r
lim Sn does not converge to a real number for |r| ≥ 1.
n→+∞

In conclusion
+∞
X 1
• If |r| < 1, then the geometric series a r n converges and its sum is a .
n=0
1−r
+∞
X
• If |r| ≥ 1, then the geometric series a r n diverges.
n=0
Fig. 2 illustrates the sum of a geometric series with a > 0 and 0 < r < 1:
a
a + ar + ar 2 + · · · + ar n + · · · = .
1−r
In Fig. 2 the terms of a geometric series are represented as areas. As we can see in Fig. 2
the areas of the terms fill in the rectangle whose area is a/(1 − r).
In Fig. 3 we represent the terms of the geometric series by lengths of horizontal line
segments. The picture strongly indicates that the total length of infinitely many horizontal
line segments is a/(1 − r). The reason for this is that by construction the slope of the
hypothenuse CB of the right triangle ABC in Fig. 3 is (1 − r). Since the length of its
vertical cathetus AC is a, the length of its horizontal cathetus AB must be a/(1 − r).
64 3. INFINITE SERIES

ar7
ar5
ar3
ar ar6
ar4

ar2
a-ra
a
0 1
0 1
1-r
Fig. 2. The width of the rectangle is 1/(1− r) and its height is a. The slope of the diagonal
is (1 − r)a. The slope of the line above the diagonal is r(1 − r)a

C
a-ar

a
ar
ar2
ar3

A a ar ar2 ar3 ar4 ar5 ar6 ... B

Fig. 3. Consider the right triangle ABC. From the small left-topmost right triangle we
calculate that the slope of the hypothenuse CB is 1 − r = rise a−ar
run = a . Since the length of
the vertical cathetus AC is a we deduce that the length of the horizontal cathetus AB is
a/(1 − r).

3.2.3. How to recognize whether an infinite series is a geometric series?

+∞
X π n+2 π n+2
Consider for example the infinite series . Here a n = .
e2n−1 e2n−1
n=1
Looking at the formula (3.16) we note that the first term of the series is a and that the
ratio between any two consecutive terms is r.
π n+2
For an = 2n−1 given above we calculate
e
π n+1+2
an+1 2(n+1)−1 π n+3 e2n−1 π
= e n+2 = 2n+1 n+2 = 2 .
an π e π e
e2n−1
 3.2. INFINITE SERIES OF REAL NUMBERS 65

+∞
an+1 X π n+2
Since is constant, we conclude that the series is a geometric series with
an e2n−1
n=1

π2 π
a = a1 = and r= for all n ∈ N.
e e2
π
Since r = < 1, we conclude that the sum of this series is
e2
+∞
X π n+2 π2 1 π 2 e2 π2 e
= = = .
e2n−1 e 1− π e e2 − π e2 − π
n=1
e2
Thus, to verify whether a given infinite series is a geometric series calculate the ratio of
the consecutive terms and see whether it is a constant:
+∞
X an+1
an for which =r for all n∈N (3.17)
an
n=1

is a geometric series. In this case a = a1 (the first term of the series).

3.2.4. Harmonic Series. Harmonic series is the series

+∞
1 1 1 1 X 1
1+ + + + ··· + + ··· = .
2 3 4 n n
n=1

Again, to explore the convergence of this series we have to study its partial sums:
1
S1 = 1, S2 = 1 + ,
2
1 1 1 1 1
S3 = 1 + + , S4 = 1 + + + ,
2 3 2 3 4
1 1 1 1 1 1 1 1 1
S5 = 1 + + + + , S6 = 1 + + + + + ,
2 3 4 5 2 3 4 5 6
1 1 1 1 1 1 1 1 1 1 1 1 1
S7 = 1 + + + + + + , S8 = 1 + + + + + + + ,
2 3 4 5 6 7 2 3 4 5 6 7 8
..
.
1 1 1 1
Sn == 1 + + + ··· + +
2 3 n−1 n
..
.
1
Since Sn+1 − Sn = > 0 the sequence {Sn }+∞
n=1 is increasing.
n+1
+∞
Next we will prove that the sequence {Sn }n=1 is not bounded. We will consider only
the positive integers which are powers of 2: 2, 4, 8, . . . , 2k , . . . . The following inequalities
hold:
1 1 1
S2 = 1 + ≥ 1 + =1 + 1
2 2 2
1 1 1 1 1 1 1 1 1
S4 = 1 + + + ≥1+ + + =1+ +2 =1 + 2
2 3 4 2 4 4 2 4 2
66 3. INFINITE SERIES

1 1 1 1 1 1 1
S8 = 1 + + + + + + +
2 3 4 5 6 7 8
1 1 1 1 1 1 1 1 1 1 1
≥1+ + + + + + + =1+ +2 +4 =1 + 3
2 4 4 8 8 8 8 2 4 8 2
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
S16 = 1 ++ + + + + + + + + + + + + +
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 1 1 1 1 1 1 1 1 1 1 1
≥1+ + + + + + + =1+ +2 +4 +8 =1 + 4
2 4 4 8 8 8 8 2 4 8 16 2

Continuing this reasoning we conclude that for every k ∈ N the following formula holds:
1 1 1 1 1 1 1 1
S2k = 1 + + + + + · · · + + · · · + k−1 + k−1 + ··· + k
2 3 4 5 8 2 2 +1 2
1 1 1 1 k−1 1 1
≥1+ +2 +4 +8 + ··· + 2 =1 + k
2 4 8 16 2k 2
Thus
1
S2k ≥ 1 + k for all k ∈ N. (3.18)
2
This formula implies that the sequence{Sn }+∞n=1 is not bounded. Namely, let M be an
arbitrary real number. We put j = max 2 floor(M ), 1 . Then
j ≥ 2 floor(M ) > 2(M − 1).
Therefore,
1
1+j > M.
2
Together with the inequality (3.18) this implies that
S2j > M.
Thus for an arbitrary real number M there exists a positive integer n = 2j such that
Sn > M . This proves that the sequence {Sn }+∞
n=1 is not bounded and therefore it is not
convergent.
In conclusion:
• The harmonic series diverges.
3.2.5. Telescoping series. The next example is an example of a series for which
we can find a simple formula for the sequence of its partial sums and easily explore the
convergence of that sequence. Examples of this kind are called telescoping series.
+∞
X 1
Example 3.23. Prove that the series converges and find its sum.
n(n + 1)
n=1

Solution. We need to examine the series of partial sums of this series:

1 1 1 1
Sn = + + + ··· + , n ∈ N.
1·2 2·3 3·4 n(n + 1)
It turns out that it is easy to find the sum Sn if we use the partial fraction decomposition
for each of the terms of the series:
1 1 1
= − for all k ∈ N.
k(k + 1) k k+1
 3.2. INFINITE SERIES OF REAL NUMBERS 67

Now we calculate:
1 1 1 1
Sn = + + + ··· +
1·2 2·3 3·4 n (n + 1)
         
1 1 1 1 1 1 1 1 1 1
= − + − + − + ··· + − + −
1 2 2 3 3 4 n−1 n n n+1
1
=1 − .
n+1
1
Thus Sn = 1 − for all n ∈ N. Using the algebra of limits we conclude that
n+1
 
1
lim Sn = lim 1− = 1.
n→+∞ n→+∞ n+1
+∞
X 1
Therefore the series converges and its sum is 1:
n=1
n (n + 1)

+∞
X 1
= 1.
n=1
n (n + 1)

Exercise 3.24. Determine whether the series is convergent or divergent. If it is con-

vergent, find its sum.
+∞  n−1 +∞ +∞ √ n +∞ n+3
X 2 X (−2)n+3 X ( 2) X e
(a) 6 (b) (c) (d)
n=1
3 n=1
5n−1 n=0
2n+1 n=1
π n−1
+∞ 2n−1 +∞ +∞ +∞
X 2 X 5 X X 2
(e) (f) (g) (sin 1)n (h)
n=1
πn n=1
2n n=0 n=0
2
n + 4n + 3
+∞ +∞
X 2 +∞ +∞
X  
X X 1
(i) (cos 1)n (j) (k) (tan 1)n (l) ln 1 +
n=0 n=2
n2 − 1 n=0 n=1
n

3.2.6. Decimal numbers. A digit is an integer from the set D = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}.
Let d : N → D be a sequence of digits. A decimal number with decimal digits d1 , d2 , d3 , . . . , dn , . . .
is in fact the infinite series:
+∞
X dn
0.d1 d2 d3 . . . dn . . . = .
n=1
10n

Consider the partial sums:

n
X dk
Sn = where n ∈ N.
10k
k=1

Then, for all n ∈ N we have

dn+1
Sn+1 − Sn = ≥ 0.
10n+1
68 3. INFINITE SERIES

Hence Sn ≤ Sn+1 for all n ∈ N. Thus, the sequence {Sn } is nondecreasing. We now prove
that {Sn } is bounded above by 1. Since dk ≤ 9 for all k ∈ N we have
n n n−1
X dk X 9 9 X 1 9 1 − 101n 9 1
Sn = k
≤ k
= k
= 1 ≤ 1 = 1.
10 10 10 10 10 1 − 10 10 1 − 10
k=1 k=1 k=0
for all n ∈ N.
It turns out that each decimal number with digits that repeat leads to a geometric
series. We use the following abbreviation:
0.d1 d2 d3 . . . dk = 0.d1 d2 d3 . . . dk d1 d2 d3 . . . dk d1 d2 d3 . . . dk d1 d2 d3 . . . dk . . . .
Rather than proving this in general and finding to which rational number the preceding
series converges, we leave it to the reader to figure out several examples in the exercises.
Exercise 3.25. Express the following real nubers as ratios of positive integers.
(a) 0.9 = 0.999 . . . (b) 0.7 = 0.777 . . . (c) 0.712 (d) 0.5432
3.2.7. Basic properties of infinite series. An immediate consequence of the defi-
nition of a convergent series is the following theorem
+∞
X
Theorem 3.26. If a series an converges, then lim an = 0.
n→+∞
n=1
+∞
X
Proof. Assume that an is a convergent series. By the definition of convergence of
n=1
a series its sequence of partial sums {Sn }+∞
n=1 converges to some number S: lim Sn = S.
n→+∞
Then also lim Sn−1 = S. Now using the formula
n→+∞
an = Sn − Sn−1 , for all n ∈ N \ {1},
and the algebra of limits we conclude that
lim an = lim Sn − lim Sn−1 = S − S = 0.
n→+∞ n→+∞ n→+∞

Warning: The preceding theorem cannot be used to conclude that a particular series
+∞
X
converges. Notice that in this theorem it is assumed that an is a convergent.
n=1
On a positive note: Theorem 3.26 can be used to conclude that a given series diverges:
+∞
X
If we know that lim an = 0 is not true, then we can conclude that the series an
n→+∞
n=1
diverges. This is a useful test for divergence.
Theorem 3.27 (The Test for Divergence). If the sequence {an }+∞
n=1 does not converge
+∞
X
to 0, then the series an diverges.
n=1
+∞  
X 1
Example 3.28. Determine whether the infinite series cos converges or diverges.
n=1
n
 3.2. INFINITE SERIES OF REAL NUMBERS 69

Solution. Just perform the divergence test:

 
1
lim cos = 1 6= 0 .
n→+∞ n
+∞  
X 1
Therefore the series cos diverges. 
n=1
n

+∞ (−1)n
X n
Example 3.29. Determine whether the infinite series converges or diverges.
n=1
n+1
( n
)+∞
n(−1)
Solution. Consider the sequence :
n+1
n=1

1 2 1 4 1 6 1 8 1 10 1 12 1 2k
, , , , , , , , , , , , ..., , , ...
1 · 2 3 3 · 4 5 5 · 6 7 7 · 8 9 9 · 10 11 11 · 12 13 (2k − 1) · 2k 2k + 1
(3.19)
Without giving a formal proof we can tell that this sequence diverges. In my informal
language the sequence (3.19) is not constantish since it can not decide whether to be close
to 0 or 1.
+∞ (−1)n
X n
Therefore the series diverges. 
n+1
n=1

Remark 3.30. The divergence test can not be used to answer whether the series
+∞    
X 1 1
sin converges or diverges. It is clear that lim sin = 0. Thus we can
n n→+∞ n
n=1
not use the test for divergence.
+∞
X
Theorem 3.31 (The Algebra of Convergent Infinite Series). Assume that an and
n=1
+∞
X
bn are convergent series. Let c be a real number. Then the series
n=1

+∞
X +∞
X +∞
X



c an , an + bn , and a n − bn ,
n=1 n=1 n=1

are convergent series and the following formulas hold

+∞
X +∞
X
c an = c an ,
n=1 n=1
+∞
X +∞
X +∞
X


a n + bn = an + bn , and
n=1 n=1 n=1
+∞
X +∞
X +∞
X


a n − bn = an − bn .
n=1 n=1 n=1
70 3. INFINITE SERIES

+∞
X +∞
X
Remark 3.32. The fact that we write bn does not necessarily mean that bn is a
n=1 n=1
genuine infinite series.
For example, let m be a positive integer and assume that bn = 0 for all n > m. Then
+∞
X Xm +∞
X +∞
X
bn = bn . In this case the series bn is clearly convergent. If an is a convergent
n=1 n=1 n=1 n=1
+∞
X

(genuine) infinite series, then Theorem 3.31 implies that the infinite series an + bn is
n=1
convergent and
+∞
X +∞
X m
X


a n + bn = an + bn .
n=1 n=1 n=1

This in particular means that the nature of convergence of an infinite series can not be
changed by changing finitely many terms of the series.
For example, let m be a positive integer. Then:
+∞
X +∞
X
The series an converges if and only if the series am+k converges.
n=1 k=1

+∞
P
Moreover, if an converges, then the following formula holds
n=1

+∞
X m
X +∞
X
an = aj + am+k .
n=1 j=1 k=1

+∞  
X π 1
Example 3.33. Prove that the series − n converges and find its sum.
n(n + 1) 2
n=1

Exercise 3.34. Determine whether the series is convergent or divergent. If a series is

convergent find its sum.
+∞ +∞ +∞ n +∞  
X n X X 3 + 2n X 3 π
(a) (b) arctan n (c) (d) +
n=1
n+1 n=1 n=0
5n+1 n=2
n2 − 1 en
+∞ n +∞   +∞ +∞
X e + πn X 1 X (n + 1)2 X
(e) (f) n sin (g) (h) ((0.9)n + (0.1)n )
22n−1 n n2 + 1
n=0 n=1 n=0 n=0

Exercise 3.35. Express the following sums as ratios of positive integers and as repeating
decimal numbers.
(a) 0.47 + 0.5 (b) 0.499 + 0.47 (c) 0.499 + 0.503

3.3. Convergence Tests

Warning: All series in the next two subsections have positive terms! Do not
use the tests from these sections for series with some negative terms.
 3.3. CONVERGENCE TESTS 71

3.3.1. Direct Comparison Test. The convergence of the geometric series in Subsec-
tion 3.2.2 and the telescopic series in Subsection 3.2.5 was established by calculating the
limits of their partial sums. This is not possible for most series. For example we will soon
prove that the series
+∞
X 1
n=1
n2
π2
converges. To understand why the sum of this series is exactly you need to read a paper
6
that I posted on the class website. You have the background knowledge to understand this
proof, but the complete proof is on a longer side.
I hope that you have done your homework and that you proved that the series
+∞
X 1
n2 −1
n=2
converges and that you found its sum. If you didn’t here is a way to do it: (It turns out
that this is a telescoping series.)
Let
1 1 1 1
Sn = + + + ··· + 2 .
3 8 15 n −1
1
Since Sn+1 − Sn = > 0 the sequence {Sn }+∞
n=2 is increasing.
(n + 1)2 − 1
For every k ∈ N such that k > 1 we have the following partial fractions decomposition
 
1 1 1 1 1
= = − .
k2 − 1 (k − 1)(k + 1) 2 k−1 k+1
Next we use this formula to simplify the formula for the n-th partial sum
n n   n  
X 1 X 1 1 1 1X 1 1
Sn = = − = −
k2 − 1 2 k−1 k+1 2 k−1 k+1
k=2 k=2 k=2
         
1 1 1 1 1 1 1 1 1 1 1
= − + − + − + ··· + − + −
2 1 3 2 4 3 5 n−2 n n−1 n+1
 
1 1 1 1 1
= + − −
2 1 2 n n+1
 
1 3 2n + 1 3 2n + 1
= − = − .
2 2 n(n + 1) 4 2n(n + 1)
Using the algebra of limits we calculate
2n + 1 2 1
2n + 1 2
+ 2 0+0
lim = lim n = lim n n = = 0.
n→+∞ 2n(n + 1) n→+∞ 2n(n + 1) n→+∞ n + 1 2·1
2
n2 n
Therefore, using the algebra of limits again, we calculate
3 3
lim Sn = − 0 = .
n→+∞ 4 4
3
Clearly Sn < for all n ∈ N \ {1}.
4
72 3. INFINITE SERIES

Now consider the series

+∞
X 1 1 1 1 1
2
=1+ + + + ··· + 2 + ···
n=1
n 4 9 16 n
Let
1 1 1 1
Tn = 1 + + + + ··· + 2 .
4 9 16 n
1
The fact that Tn+1 − Tn = > 0 implies that the sequence {Tn }+∞
n=1 is increasing.
(n + 1)2
Since
1 1 1 1 1 1 1 1
< , < , < , ..., < 2 ,
4 3 9 8 16 15 n2 n −1
we conclude that
1 1 1 1 1 1 1 1 3
Tn = 1 + + + + ··· + 2 < 1 + + + + ··· + 2 = 1 + Sn < 1 + .
4 9 16 n 3 8 15 n −1 4
7
Thus Tn < for all n ∈ N \ {1}. Since the sequence {Tn }+∞ n=1 is increasing and bounded
4
+∞
X 1 7
above it converges by Theorem 3.18. Thus the series 2
converges and its sum is < .
n 4
n=1
The principle demonstrated in the above example is the core of the following comparison
theorem.
+∞
X +∞
X
Theorem 3.36 (The Direct Comparison Test). Let an and bn be infinite series
n=1 n=1
with positive terms. Assume that
an ≤ bn for all n ∈ N.
+∞
X +∞
X +∞
X +∞
X
(a) If bn converges, then an converges and an ≤ bn .
n=1 n=1 n=1 n=1
+∞
X +∞
X
(b) If an diverges, then bn diverges.
n=1 n=1

3.3.2. Limit Comparison Test. Sometimes the following comparison theorem is eas-
ier to use.
+∞
X +∞
X
Theorem 3.37 (The Limit Comparison Test). Let an and bn be infinite series
n=1 n=1
with positive terms. Assume that
an
lim = L.
n→+∞ bn
+∞
X +∞
X +∞
X +∞
X
If bn converges, then an converges. Or, equivalently, if an diverges, then bn
n=1 n=1 n=1 n=1
diverges.
+∞
X n+1
Example 3.38. Determine whether the series √ converges or diverges.
1 + n 6
n=1
 3.3. CONVERGENCE TESTS 73

Solution. √ The dominant term in the numerator is n and the dominant term in the
denominator is n6 = n3 . This suggests that this series behaves as the convergent series
+∞
X 1
. Since we are trying to prove convergence we will take
n2
n=1
n+1 1
an = √ and bn =
1 + n6 n2
in the Limit Comparison Test. Now calculate:
n+1 n2 (n + 1) 1
√ 2 1+
1+n 6 n (n + 1) 3
lim = lim √ = lim √ n = lim r n = 1.
n→+∞ 1 n→+∞ 1 + n6 n→+∞ 1 + n6 n→+∞ 1
n2 +1
n3 n6
In the last step we used the algebra of limits and the fact that
r
1
lim +1=1
n→+∞ n6
which needs a proof by definition.
n+1
√ +∞
1 + n6 X 1
Since we proved that lim = 1 and since we know that is convergent,
n→+∞ 1 n2
n=1
n2
+∞
X n+1
the Limit Comparison Test implies that the series √ converges. 
1 + n 6
n=1

3.3.3. Integral Comparison Test. In the next theorem we compare an infinite series
with an improper integral of a positive function. Here it is presumed that we know how to
determine the convergence or divergence of the improper integral involved.
Theorem 3.39 (The Integral Test). Suppose that x 7→ f (x) is a continuous positive,
decreasing function defined on the interval [1, +∞). Assume that an = f (n) for all n ∈ N.
Then the following statements are equivalent
Z +∞
(a) The integral f (x) dx converges.
1
+∞
X
(b) The series an converges.
n=1

At this point we assume that you are familiar with improper integrals and that you
know how to decide whether an improper integral converges or diverges.
We will use this test in two different forms:
Z +∞ +∞
X
• Prove that the integral f (x) dx converges. Conclude that the series an
1 n=1
converges.
Z +∞ +∞
X
• Prove that the integral f (x) dx diverges. Conclude that the series an
1 n=1
diverges.
74 3. INFINITE SERIES

+∞
X 1
Example 3.40 (Convergence of p-series). Let p be a real number. The p-series
np
n=1
is convergent if p > 1 and divergent if p ≤ 1.
Solution. Let n > 1. Then the function x 7→ nx is an increasing function. Therefore,
if p < 1, then np < n. Consequently,
1 1
p
> , for all n > 1 and p < 1 .
n n
+∞
X1 +∞
X 1
Since the series diverges, the Comparison Test implies that the series diverges
n np
n=1 n=1
for all p ≤ 1.
1
Now assume that p > 1. Consider the function f (x) = p , x > 0. This function is a
x
continuous, decreasing, positive function. Let me calculate the improper integral involved
in the Integral Test for convergence:
Z +∞ Z t t
1 1 1 1
dx = lim dx = lim
1 xp t→+∞ 1 xp t→+∞ 1 − p xp−1
1
!
1 1 1 1
= lim −1 = (−1) =
1 − p t→+∞ tp−1 1−p p−1
Thus this improper integral converges. Notice that the condition p > 1 was essential to
1 1
conclude that lim p−1 = 0. Since p = f (n) for all n ∈ N, the Integral Test implies that
t→+∞ t n
+∞
X 1
the series converges for p > 1. 
n=1
np
1
Remark 3.41. We have not proved this for all p > 1 the function f (x) = p , x > 0, is
x
continuous. One way to prove that for an arbitrary a ∈ R the function x 7→ xa , x > 0 is
continuous is to use the identity
xa = ea ln x , x > 0.
This identity shows that the function x 7→ xa ,
x > 0 is a composition of the function
exp(x) = ex , x ∈ R and the function x 7→ a ln x, x > 0. The later function is continuous by
the algebra of continuous functions: It is a product of a constant a and a continuous function
ln. We proved that exp is continuous. By Theorem 2.59 a composition of continuous
function is continuous. Consequently x 7→ xa , x > 0 is continuous.
Exercise 3.42. Determine whether the series is convergent or divergent.
+∞ +∞ +∞ +∞
X 1 X
−n2
X 1 X ln n
(a) √ (b) ne (c) (d) √
n=1
n n n=1 n=2
n ln n n=1
n n
+∞ +∞ +∞ +∞
X 1 X 1 X 1
X 1 1

(e) (f) (g) sin (h) sin
n=2
n(ln n)b n=1
n! n=1
n n=2
n n
+∞ +∞
X π + en +∞ +∞
X 1 1
X n! X n2 + 1
(i) cos (j) (k) (l) √
n=1
n n n=0
e + πn n=1
nn n=0
n7 + n3 + 1
 3.3. CONVERGENCE TESTS 75

For the series in (e) find all numbers b for which the series converges.
Exercise 3.43. A digit is a number from the set {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}. A decimal
number with digits d1 , d2 , d3 , . . . , dn , . . . is in fact the infinite series:
+∞
X dn
0.d1 d2 d3 . . . dn . . . = .
n=1
10n

Use a theorem from this section to prove that the series above always converges.
3.3.4. Ratio and root tests. Warning: All series in this section have positive
terms! Do not use the tests from this section for series with negative terms.
In Subsection 3.2.3 we pointed out (see (3.17)) that a series
+∞
X an+1
an for which =r for all ∈N
an
n=1

is a geometric series. Consequently, if |r| < 1 this series is convergent, and it is divergent
if |r| ≥ 1.
+∞
X 1
Testing the series n − 2n+1
using this criteria leads to the ratio
n=0
3
  n   n
1 n 2 2
n n+1 3 1−2 1−2
3n+1 − 2n+2 = 3 − 2 = 
3
 n  =
1 3
1 n+1 n+2  n+1
3 −2 2 3 2
3n+1 1 − 2 1−2
3n − 2n+1 3 3
which certainly is not constant, but it is “constantish.” I propose that series for which the
ratio an+1 /an is not constant but constantish, should be called “geometrish.” The following
theorem tells that convergence and divergence of these series is determined similarly to
geometric series.
+∞
X
Theorem 3.44 (The Ratio Test). Assume that an is a series with positive terms
n=1
and that
an+1
lim = R.
n→+∞ an

Then
(a) If R < 1, then the series converges.
(b) If R > 1, then the series diverges.
Another way to recognize a geometric series is:
+∞ r
X an+1
A series an for which n =r for all ∈N
a1
n=1

is a geometric series. Consequently, if |r| < 1 this series is convergent, and it is divergent
if |r| ≥ 1.
76 3. INFINITE SERIES

+∞  
X 1+n n
Testing the series using this criteria leads to the root
1 + 2n
n=0
s 
1+n n 1+n 1
+1
n
= = n1
1 + 2n 1 + 2n n +2
which certainly is not constant, but it is “constantish.”
+∞
X
Theorem 3.45 (The Root Test). Assume that an is a series with positive terms
n=1
and that √
lim n
an = R .
n→+∞
Then
(a) If R < 1, then the series converges.
(b) If R > 1, then the series diverges.
Remark 3.46. Notice that in both the ratio test and the root test if the limit R = 1
we can conclude neither divergence nor convergence. In this case the test is inconclusive.
Exercise 3.47. Determine whether the series is convergent or divergent.
+∞ +∞   +∞ +∞
X 1 X n+2 n X 4n X n!
(a) n
(b) (c) 2n−1
(d)
2 −3 2n − 1 3 1 · 3 · 5 · · · (2n − 1)
n=2 n=1 n=1 n=1
+∞
X 3n n2 +∞
X +∞
X e1/n +∞
X 2 · 4 · 6 · · · (2n)
(e) (f) e−n n! (g) (h)
n! n2 1 · 3 · 5 · · · (2n − 1)
n=1 n=1 n=1 n=1
+∞
X (n!)2 +∞
X 2 n2n +∞
X 23n +∞
X 1
(i) (j) 2 n
(k) 2n
(l)
n=1
(2n)! n=1
(3n + 1) n=1
3 n=1
(arctan n)n
+∞ 2 +∞ +∞ n +∞
X n X (n + 1)2 X a X 1 · 3 · 5 · · · (2n − 1)
(m) n
(n) n
(o) (p)
n=1
2 n=1
n2 n=1
n! n=1
2 · 4 · 6 · · · (2n)
For some of the problems you might need to use tests from previous sections.

I intentionally start a new page here.

 3.3. CONVERGENCE TESTS 77

3.3.5. Alternating infinite series. In the previous two sections we considered only
series with positive terms. In this section we consider series with both positive and negative
terms which alternate: positive, negative, positive, etc. Such series are called alternating
series. For example
+∞
1 1 1 1 1 1 X 1
1 − + − + − + · · · + (−1)n+1 + · · · = (−1)n+1 (3.20)
2 3 4 5 6 n n
n=1
+∞
1 1 1 1 1 1 1 1 1 1 X 4(−1)n+1
1−1+ − + − + − + − + − + ··· =
(3.21)
3 2 5 3 7 4 8 5 9 6
n=1
n 3 + (−1)n+1
+∞
3 4 5 6 7 n+1 X n+1
2 − + − + − + · · · + (−1)n+1 + ··· = (−1)n+1 (3.22)
2 3 4 5 6 n n
n=1

Theorem 3.48 (The Alternating Series Test). If the alternating series

+∞
X
a1 − a2 + a3 − a4 + · · · + (−1)n+1 an + · · · = (−1)n+1 an
n=1

satisfies the following three conditions:

(i) for all n ∈ N we have an > 0;
(ii) for all n ∈ N we have an+1 ≤ an ,
(iii) lim an = 0,
n→+∞
+∞
X
then the series (−1)n+1 an converges.
n=1

Proof. Assume that a sequence {an } satisfies (i), (ii) and (iii).
By the definition of convergence the assumption (iii) implies that for every ǫ > 0 there
exists Na (ǫ) such that
∀n ∈ N n > Na (ǫ) ⇒ |an − 0| < ǫ.
Since an > 0, the last implication can be simplified as follows
∀n ∈ N n > Na (ǫ) ⇒ an < ǫ. (3.23)
We need to show that the sequence of partial sums
Sn = a1 − a1 − a2 + a3 − a4 + · · · + (−1)n+1 an , n ∈ N,
converges.
0 S2 S4 S6 S8 S S7 S5 S3 S1
a8
a7
a6
a5
a4
a3
a2
a1

Fig. 4. The partial sums of an alternating series on a number line

78 3. INFINITE SERIES

As Figure 4 suggests, each even-indexed partial sum is less than each odd-indexed partial
sum. That is
∀j ∈ N ∀ k ∈ N S2j < S2k−1 (3.24)
Next we will prove this claim. Let k and j be arbitrary positive integers. First assume
k ≤ j. Then 2k − 1 < 2j and
2j
X


S2j − S2k−1 = (−1)i+1 ai = −a2k + a2k+1 + · · · + −a2j−2 + a2j−1 − a2j .
i=2k
In the last sum each of the numbers in parenthesis is nonpositive. Therefore,
2j
X
S2j − S2k−1 = (−1)i+1 ai ≤ −a2j < 0.
i=2k
Hence S2j < S2k−1 in this case. Now assume that k > j. Then 2k − 1 > 2j and
2k−1
X


S2k−1 − S2j = (−1)i+1 ai = a2j+1 − a2j+2 + · · · + a2k−3 − a2k−2 + a2k−1 .
i=2j+1
In the last sum each of the numbers in parenthesis is nonnegative. Therefore,
2k−1
X
S2k−1 − S2j = (−1)i+1 ai ≥ a2k−1 > 0.
i=2j+1

Hence S2j < S2k−1 in this case as well. This completes the proof of (3.24).
Define  
A = S2j : j ∈ N and B = S2k−1 : k ∈ N .
With this notation (3.24) can be restated as
∀a ∈ A ∀b ∈ B a < b.
Since clearly A 6= ∅ and B 6= ∅ we can apply the Completeness Axiom to the sets A and B.
By the Completeness Axiom there exists c ∈ R such that
∀a ∈ A ∀ b ∈ B a ≤ c ≤ b.
The last inequality in fact says
∀j ∈ N ∀ k ∈ N S2j ≤ c ≤ S2k−1 . (3.25)
Let n ∈ N be arbitrary. If n is even, then (3.25) yields
Sn ≤ c ≤ Sn+1 = Sn + an+1 .
Therefore c − Sn ≤ an+1 . If n is odd, then (3.25) yields
Sn − an+1 = Sn+1 ≤ c ≤ Sn .
Therefore, Sn − c ≤ an+1 . Thus, for all n ∈ N we have
|Sn − c| ≤ an+1 . (3.26)
Now, using (3.26) and (3.23) we can prove limn→+∞ Sn = c. Let ǫ > 0 be arbitrary.
Set N (ǫ) = Na (ǫ). Assume n ∈ N and n > N (ǫ) = Na (ǫ). Then also n + 1 ∈ N and
n + 1 > Na (ǫ). By the implication in (3.23) we conclude an+1 < ǫ. This inequlity, together
with (3.26), implies |Sn − c| < ǫ. Thus, we proved that
∀ǫ > 0 ∃ N (ǫ) ∈ R ∀n ∈ N n > N (ǫ) ⇒ |Sn − c| < ǫ.
 3.3. CONVERGENCE TESTS 79

This proves that the sequence {Sn }+∞

n=1 converges and hence the alternating series converges.

Example 3.49. Prove that the series in (3.20) converges. The series in (3.20) is called
the alternating harmonic series.
Solution. We verify three conditions of the Alternating Series Test. Here, an = 1/n,
n ∈ N. First, for all n ∈ N we have n > 0 and hence 1/n > 0. Second, since for all n ∈ N
we have n + 1 > n, by pizza-party principle 1/(n + 1) < 1/n. Third, lim (1/n) = 0 is easy
n→+∞
to prove. Thus the Alternating Series Test implies that the Alternating Harmonic Series
converges. 
Remark 3.50. There is a nice geometric argument that
+∞
X 1
(−1)k+1 = ln 2.
k
k=1
This argument is based on the fact that the even-indexed partial
R 2 sums of the Alternating
Harmonic Series are in fact right Riemann sums of the integral 1 (1/x)dx:
2n
X 1
S2n = (−1)k+1
k
k=1
n n
X 1 X 1
= −
2j − 1 2j
j=1 j=1
n n n
X 1 X 1 X 1
= + −2
2j − 1 2j 2j
j=1 j=1 j=1
2n n
X 1 X 1
= −
k j
k=1 j=1
n
X 1
=
n+k
k=1
n
X 1 1
= k
.
n 1+ n
k=1
And, similarly, the odd-indexed
R 2 partial sums of the Alternating Harmonic Series are left
Riemann sums of the integral 1 (1/x)dx:
2n−1
X n
X n−1
X n
X n−1
X n−1
X 2n−1
X n−1
X n−1
X n−1
X
S2n−1 = (−1)k+1 k1 = 1
2j−1 −
1
2j = 1
2j−1 +
1
2j −2
1
2j = 1
k−
1
j = 1
n+k = 1 1
k
n 1+ n .
k=1 j=1 j=1 j=1 j=1 j=1 k=1 j=1 k=0 k=0

The details of the argument I will post on the class website.

Remark 3.51. The Alternating Series Test does not apply to the series in (3.21) since
the sequence of numbers
1 1 1 1 1 1 1 1 1 1 4
1, 1, , , , , , , , , , , ...,
, . . .
3 2 5 3 7 4 8 5 9 6 n 3 + (−1)n+1
is not non-increasing. Further exploration of the series in (3.21) would show that it diverges.
80 3. INFINITE SERIES

The Alternating Series Test does not apply to the series in (3.22) since this series does
not satisfy the condition (ii):
n+1
lim = 1 6= 0 .
n→+∞ n
Again this series is divergent by the Test for Divergence.
Exercise 3.52. Determine whether the given series converges or diverges.
+∞   +∞  π +∞  
X 1 X X 1
(a) cos nπ + (b) sin n (c) sin nπ −
n 2 n
n=1 n=0 n=1
+∞   +∞   +∞
X 1 1 X (−1)n X 1  π
(d) cos nπ + (e) ln 1 − (f) sin n
n n n n 2
n=1 n=1 n=1
+∞   +∞ +∞
X π 1 X (−1)n+1 X (−1)n+1
(g) sin n + (h) (i)
n=1
2 n n=1
n − (−1)n n=1
2n − (−1)n

Several of the exercises in the next section use the Alternating Series Test for conver-
gence. Do those exercises as well.

3.3.6. Absolute and Conditional Convergence. In the previous section we proved

that the alternating harmonic series
+∞
1 1 1 1 1 1 X 1
1 − + − + − + · · · + (−1)n+1 + · · · = (−1)n+1 converges. (3.27)
2 3 4 5 6 n n
n=1

Later on we will see that the sum of this series is ln 2.

Talking about infinite series in class I have often used the analogy with an infinite
column in a spreadsheet and finding its sum. A series with positive and negative terms one
can interpret as balancing a checkbook with (infinitely) many deposits and withdrawals.
Looking at the alternating harmonic series (3.27) we see a sequence of alternating deposits
and withdrawals, infinitely many of them. What we proved in the previous section tells
that under two conditions on the deposits and withdrawals, although it has infinitely many
transactions, this checkbook can be balanced.
Now comes the first surprising fact! Let’s calculate how much has been deposited to
this account:
+∞
1 1 1 1 X 1
1 + + + + ··· + + ··· = .
3 5 7 2n − 1 n=1
2n − 1
Since for all k ∈ N it holds
1 1
< ,
2k 2k − 1
we have that
n n
1 1X1 X 1
Hn = <
2 2 k 2k − 1
k=1 k=1
for all n ∈ N. As the sequence {Hn } of harmonic numbers is unbounded, we deduce that
+∞
P 1
the sequence of partial sums of the infinite series is unbounded. Therefore this
n=1 2n − 1
series diverges.
 3.3. CONVERGENCE TESTS 81

Looking at the withdrawals we see

+∞
1 1 1 1 1X1
− − − − ··· − − ··· = − .
2 4 6 2n 2 n
n=1

Again this is a divergent series. In the language of a bank account, we are encountering a
suspicious situation: We have an account to which an “infinite” amount of money has been
deposited and an “infinite” amount of money has been withdrawn. A simpler way to look
at this is to look at the total amount of money that went through this account (one can
call this amount the total “activity” in the account):
+∞
X 1 1 1 1 1 1 1
(−1)n+1 = 1 + + + + + + ··· + + ··· (3.28)
n 2 3 4 5 6 n
n=1

This is the harmonic series which is divergent.

Since we know that an infinite amount of money has been deposited to this account
we might want to get into the spending mood sooner. So, we rearrange the deposits and
the withdrawals; we do two withdrawals after each deposit, keeping the amounts the same.
This results in the following infinite series:
1 1 1 1 1 1 1 1 1 1 1
1− − + − − + − − + − − + ··· . (3.29)
2 4 3 6 8 5 10 12 7 14 16
In any real life checking account just rearranging the deposits and the withdrawals might
result in an occasional low balance but the final balance will remain the same. Amazingly
this is not always the case with infinite series! (This is the second surprising fact!) For
example, the series in (3.29) and the series in (3.27) have identical terms which are arranged
differently; in Example 3.49 we proved that the series (3.27) converges and next we will show
that the series (3.29) also converges but to a different number.
To be specific, denote the terms of the series (3.29) by bn , n ∈ N. Then
1 1 1
b3k−2 = , b3k−1 = − , b3k = − , k ∈ N.
2k − 1 4k − 2 4k
It is clear that the series (3.29) has the same terms as the alternating harmonic series. The
terms of the alternating harmonic series have been reordered. For k ∈ N, the term at the
positions 2k − 1 (odd-indexed terms) in the alternating harmonic series is at the position
3k−2 in the series (3.29), the term which is at the position 4k−2 (a “half” of the even-index
terms) in the alternating harmonic series is at the position 3k −1 in the series (3.29) and the
term which is at the position 4k (another “half” of the even-index terms) in the alternating
harmonic series is at the position 3k in the series (3.29).
The following calculation indicates that the sum of the series in (3.29) is 1/2 of the sum
of the alternating harmonic series in (3.27). Let us calculate the 3n-th partial sum of the
series (3.29). Since this sum has 3n terms, one-third (exactly n) of them are positive and
two-thirds (exactly 2n) of them are negative. Since this is a finite sum we can rearrange
terms as we please. Here is the calculation
1 1 1 1 1 1 1 1 1 1 1
− + − − + −
S3n = 1 − − + ··· + − −
2 4 3 6 8 5 10 12 2n − 1 4n − 2 4n
1 1 1 1 1 1 1 1
= − + − + − + ··· + −
2 4 6 8 10 12 4n − 2 4n
82 3. INFINITE SERIES
 
1 1 1 1 1 1 1 1
= 1 − + − + − + ··· + −
2 2 3 4 5 6 2n − 1 2n
Hence, 3n-th partial sum of the series (3.29) is identical to one-half of the 2n-th partial sum
of the alternating harmonic series. Since the sum of the alternating harmonic series is ln 2
we have
ln 2
lim S3n = .
n→+∞ 2
Since
1 1 1 1
S3n+1 = S3n + and S3n+2 = S3n + − = S3n + ,
2n + 1 2n + 1 4n + 2 4n + 2
we conclude that
ln 2
lim S3n+1 = lim S3n+2 = lim S3n = .
n→+∞ n→+∞ n→+∞ 2
From the last three equalities one can prove rigorously that
ln 2
lim Sn = .
n→+∞ 2
This proves that the series (3.29) converges to (ln 2)/2. That is, just a rearrangement of
the terms changed the sum.
This is a remarkable observation: a change of order of summation can change the sum
of an infinite series. This feature is closely related to the fact that the total activity of
the account expressed in (3.28) is a divergent series. This is a motivation for the following
definition.
+∞
X
Definition 3.53. A convergent series an is called conditionally convergent if
n=1
+∞
X
the series of the absolute values of its terms |an | is divergent.
n=1
+∞
X
Definition 3.54. A series an is called absolutely convergent if the series of the
n=1
+∞
X
absolute values of its terms |an | is convergent.
n=1
Example 3.55. Prove that the series
+∞
1 1 1 1 1 1 X 1
1− + − + − + · · · + (−1)n+1 2 + · · · = (−1)n+1 2
4 9 16 25 36 n n=1
n
is absolutely convergent.
Solution. By the definition of absolute convergence we need to determine the conver-
gence of the series
+∞ +∞
X
n+1 1 X 1 1 1 1 1 1
(−1) 2
= 2
=1+ + + + + + ···
n n 4 9 16 25 36
n=1 n=1
This is a p-series with p = 2. Therefore this series converges. (Notice that in Subsection 3.3.1
we proved that this series converges by comparing it to a telescoping series.) 
 3.3. CONVERGENCE TESTS 83

Remark 3.56. One can interpreted the series in Example 3.55 as a checking account
with infinitely many alternating deposits and withdrawals. In this case the total activity of
the account is a convergent series. Consequently the total amount deposited
+∞
1 1 1 X 1
1+ + + ··· + 2
+ · · · = (3.30)
9 25 (2n − 1) (2n − 1)2
n=1

and the total amount withdrawn

+∞ +∞
1 1 1 1 X 1 1X 1
+ + + ··· + + · · · = = (3.31)
4 16 36 (2n)2 (2n)2 4 n2
n=1 n=1
are both convergent series. As we can see, the total amount withdrawn is 1/4 of the total
activity of the account. We mentioned before that (this is proved in a paper posted on the
class website)
+∞
X 1 1 1 1 1 1 π2
= 1 + + + + + + · · · = .
n=1
n2 4 9 16 25 36 6
Therefore
+∞
X 1 1 1 1 1 1 3 π2 1 π2 1 π2 π2
(−1)n+1 2 = 1 − + − + − + ··· = − = =
n=1
n 4 9 16 25 36 4 6 4 6 2 6 12
+∞
X
Theorem 3.57. If a series an is absolutely convergent, then it is convergent.
n=1
+∞
X +∞
X
Proof. Assume that an is absolutely convergent, that is assume that |an | is
n=1 n=1
+∞
X
convergent. Then the algebra of convergent series the series 2 |an | is convergent. Since
n=1
−|an | ≤ an ≤ |an |, we conclude that
0 ≤ an + |an | ≤ 2 |an | for all n ∈ N.
+∞
X

By the Comparison Test it follows that the series an + |an | is convergent. The algebra
n=1
of convergent series implies that the series
+∞ 
X  X+∞


an + |an | − |an | = an
n=1 n=1
is also convergent. 
The following stronger versions of the Ratio and the Root test can be applied to any
series to determine whether a series converges absolutely or it diverges.
+∞
X |an+1 |
Theorem 3.58 (The Ratio Test). Let an be a series for which lim = R.
n→+∞ |an |
n=1
Then
(a) If R < 1, then the series converges absolutely.
84 3. INFINITE SERIES

(b) If R > 1, then the series diverges.

+∞
X p
n
Theorem 3.59 (The Root Test). Let an be a series for which lim |an | = R.
n→+∞
n=1
Then
(a) If R < 1, then the series converges absolutely.
(b) If R > 1, then the series diverges.
Notice that if the root or the ratio test apply to a series, then series either converges
absolutely or diverges. This implies that if a series converges conditionally, then either
|an+1 | |an+1 |
lim = 1 or lim does not exist,
n→+∞ |an | n→+∞ |an |

and also
p
n
p
n
lim |an | = 1 or lim |an | does not exist.
n→+∞ n→+∞
In other words, the root and the ratio test cannot lead to a conclusion that a series converges
conditionally.
It turns out that our only tool which can be used to conclude conditional convergence
is the alternating series test.
Exercise 3.60. Determine whether the given series converges conditionally, converges
absolutely or diverges.
+∞ +∞ +∞ +∞
X cos(nπ) X sin(nπ/2) X (−1)n+1 X (−1)n+1
(a) (b) (c) √ (d) √
n=0
n2 + 1 n=0
n+1 n=1
n n=1
n n
+∞ +∞ +∞ +∞ √
X (−1)n+1 X
n+1 e
1/n X
n+1 n
n X
n+1 n
(e) p
(f) (−1) (g) (−1) (h) (−1)
n=1
n n=1
n n=1
n! n=1
n+1
+∞ n +∞ +∞ +∞
X (−1) X ln n X X n+1
(i) (j) (−1)n+1 (k) (−1)n+1 e1/n (l) (−1)n+1 ln
ln n n n
n=2 n=1 n=1 n=1
In problem (e) determine all the values of p for which the series converges absolutely,
converges conditionally and diverges.
Exercise 3.61. Determine whether the given series converges conditionally, converges
absolutely or diverges.
+∞ 2 +∞
n+1 (sin n) 4
X X
(a) (−1) 2
(b) (−1)n+1
n=1
n n=1
2n + 3 + (−1)n
+∞   +∞  
X
n+1 1 X
n+1 1
(c) (−1) cos (d) (−1) sin
n n
n=1 n=1
 3.4. INFINITE SERIES OF FUNCTIONS 85

3.4. Infinite Series of functions

3.4.1. Power Series. The most important series is the geometric series:
+∞
X
a + a r + a r2 + a r3 + · · · + a rn + · · · = a rn .
n=0
If −1 < r < 1 the geometric series converges. Moreover, we proved
+∞
X a
a rn = a + a r + a r2 + a r3 + · · · + a rn + · · · = for − 1 < r < 1. (3.32)
1−r
n=0
Replacing r by x and letting a = 1 we can rewrite the formula in (3.32) as
+∞
X 1
xn = 1 + x + x2 + x3 + · · · + xn + · · · = for − 1 < x < 1. (3.33)
n=0
1−x
The formula (3.33) can be viewed as a representation of the function
1
f (x) = , −1 < x < 1,
1−x
as an infinite series of powers of x: 1 = x0 , x, x2 , x3 , . . .:
+∞
1 X
= 1 + x + x2 + x3 + · · · + xn + · · · = xn for − 1 < x < 1.
1−x n=0
You will agree that the (non-negative) integer powers of x are very simple functions.
Therefore, it is natural to explore the following question:

Which functions can be represented as infinite series of

Q1:
constant multiples of (non-negative) integer powers of x?

In other words: Which functions x 7→ f (x) can be represented as

+∞
X
f (x) = a0 + a1 x + a2 x2 + a3 x3 + · · · + an xn + · · · = an x n for ? < x < ?.
n=0
The infinite series
+∞
X
2 3 n
a0 + a1 x + a2 x + a3 x + · · · + an x + · · · = an x n (3.34)
n=0
is called a power series.
The basic question to ask about a power series is:

Q2: For which real numbers x does the power series converge?

Since we are working with the powers of x and since there is no restriction on the signs
of an and x, we can use Theorems 3.58 and 3.59 (the ratio and root test) to determine the
absolute convergence of the power series (3.34). To apply Theorem 3.58 we calculate
|an+1 | |x|n+1 |an+1 | |x| |an+1 |
lim = lim = |x| lim .
n→+∞ |an | |x|n n→+∞ |an | n→+∞ |an |
86 3. INFINITE SERIES

Assume that
|an+1 |
lim = L. (3.35)
n→+∞ |an |

If L = 0, then Theorem 3.58 implies that the series (3.34) converges for all real numbers x.
If L > 0, then Theorem 3.58 implies that the series (3.34)
1 1
converges absolutely for |x| L < 1, that is for −<x<
L L
1 1
diverges for |x| L > 1, that is for x < − or x >
L L
If the limit in (3.35) does not exist, then no conclusion about the convergence or divergence
can be deduced.
To apply Theorem 3.59 we calculate
p p
lim n |an | |x|n = |x| lim n |an | .
n→+∞ n→+∞

Assume that p
n
lim |an | = L. (3.36)
n→+∞
If L = 0, then Theorem 3.59 implies that the series (3.34) converges for all real numbers x.
If L > 0, then Theorem 3.59 implies that the series (3.34)
1 1
converges absolutely for |x| L < 1, that is for −
<x<
L L
1 1
diverges for |x| L > 1, that is for x < − or x >
L L
If the limit in (3.36) does not exist, then no conclusion about the convergence or divergence
can be deduced.
Example 3.62. Consider the power series
∞
1 1 1 2 1 3 1 n X 1 n
+ x + x + x + ··· + x + ··· = x .
0! 1! 2! 3! n! n=0
n!
In this example an = 1/n!, n ∈ N ∪ {0}. We calculate
1
|an+1 | (n+1)! 1
L = lim = lim 1 = lim = 0.
n→+∞ |an | n→+∞ n→+∞ n+1
n!
Consequently the given power series converges absolutely for every x ∈ R.
Example 3.63. Consider the power series
∞
X
2 3 n
1 + 2x + 3 x + 4 x + · · · + (n + 1) x + · · · = (n + 1) xn .
n=0

Here an = n + 1, n ∈ N ∪ {0} and we calculate

|an+1 | n+2
L = lim = lim = 1.
n→+∞ |an | n→+∞ n + 1

Consequently the given power series converges absolutely for every x ∈ (−1, 1). Clearly the
series diverges for x = −1 and for x = 1.
 3.4. INFINITE SERIES OF FUNCTIONS 87

Example 3.64. Consider the power series

∞
1 2 1 3 1 4 1 X 1
x− x + x − x + · · · + (−1)n+1 xn + · · · = (−1)n+1 xn .
2 3 4 n n=0
n
Here a0 = 0 and an = (−1)n+1 1/n, for all n ∈ N. We calculate
1
|an+1 | n+1 n
L = lim = lim 1 = lim = 1.
n→+∞ |an | n→+∞ n→+∞ n + 1
n
Consequently the given power series converges absolutely for every x ∈ (−1, 1). Clearly the
series diverges for x = −1 and converges conditionally for x = 1.
Example 3.65. Consider the power series
∞
1 1 1 1 X 1
1 + x + 2 x2 + 3 x3 + · · · + n xn + · · · = xn . (3.37)
2 2 2 2 2n
n=0

Here an = 2−n , n ∈ N ∪ {0}. We calculate

1
|an+1 | 2n+1 1 1
L = lim = lim 1 = lim = .
n→+∞ |an | n→+∞ n→+∞ 2 2
2n
Consequently the given power series converges absolutely for every x ∈ (−2, 2). Clearly the
series diverges for x = −2 and for x = 2.
Notice that we can actually calculate the sum of this series using the following substi-
tution (or you can call this a trick). Substitute u = x/2 in (3.37). Then (3.37) becomes
∞
X
2 3 n
1 + u + u + u + ··· + u + ··· = un . (3.38)
n=0
We know that the sum of the series in (3.38) is 1/(1 − u) for u ∈ (−1, 1), that is,
∞
X 1
1 + u + u2 + u3 + · · · + un + · · · = un = , u ∈ (−1, 1).
1−u
n=0
Substituting back u = x/2 we get:
∞
1 1 1 1 X 1 n 2
1 + x + 2 x2 + 3 x3 + · · · + n xn + · · · = n
x = , x ∈ (−2, 2).
2 2 2 2 n=0
2 2−x

Example 3.66. Consider the power series

∞
1 1 1 1 X 1
x + x2 + x3 + · · · + 2 xn + · · · = xn .
1 4 9 n n2
n=1
We calculate
1
|an+1 | (n+1)2 n2
L = lim = lim 1 = lim = 1.
n→+∞ |an | n→+∞ n→+∞ (n + 1)2
n2
Consequently the given
P∞ power series converges absolutely for every x ∈ (−1, 1). For x = 1
1
we get the series n=1 n2 . Therefore, for x = 1 the given power series converges. For
x = −1 we get the alternating series which converges absolutely. Therefore the given power
series converges absolutely on [−1, 1].
88 3. INFINITE SERIES

The following theorem answers the question Q2 above.

Theorem 3.67. Let
+∞
X
2 3 n
a0 + a1 x + a2 x + a3 x + · · · + an x + · · · = an x n
n=0

be a power series. Then one of the following three cases holds.

(A) The power series converges absolutely for all x ∈ R.
(B) There exists r > 0 such that the power series converges absolutely for all x ∈
(−R, R) and diverges for all x such that |x| > R.
(C) The power series diverges for all x 6= 0. For x = 0 it is trivial that the power series
converges.
The set on which a power series converges is called the interval of convergence. The
number R > 0 in Theorem 3.67 (B) is called the radius of convergence. In the case (A) in
Theorem 3.67 we write R = +∞. In the case (C) in Theorem 3.67 we write R = 0.
Remark 3.68. In the case (B) in Theorem 3.67 the convergence of the power series at
the points x = R and x = −R must be determined by studying the infinite series
+∞
X +∞
X
an Rn and an (−R)n .
n=0 n=0

Examples in this section show that the interval of convergence of a power series can have
any of these four forms (−R, R), (−R, R], [−R, R) and [−R, R].
3.4.2. Functions Represented as Power Series. The following theorem lists prop-
erties of functions defined by a power series.
Theorem 3.69. Let R > 0 be the radius of convergence of the power series
+∞
X
a0 + a1 x + a2 x 2 + a3 x 3 + · · · + an x n + · · · = an xn .
n=0

Then the function f defined on (−R, R) by

+∞
X
f (x) := a0 + a1 x + a2 x2 + a3 x3 + · · · + an xn + · · · = an x n , −R < x < R,
n=0

has the following three properties:

(a) The function f is continuous on (−R, R).
(b) The derivative f ′ (x) exists for all x ∈ (−R, R) and
+∞
X
′ 2 n−1 n
f (x) = a1 + 2a2 x + 3a3 x + · · · + nan x + (n + 1)an+1 x + · · · = (n + 1)an+1 xn .
n=0

(c) The function f has derivatives of all orders 1, 2, 3, . . ., at all points of (−R, R). In
particular
f (0) = a0 , f ′ (0) = a1 , f ′′ (0) = 2 a2 , f ′′′ (0) = 3 · 2 a3 , . . . , f (n) (0) = n! an , . . . . (3.39)
 3.4. INFINITE SERIES OF FUNCTIONS 89

(d) For all x ∈ (−R, R) we have

Z x +∞
a1 2 a2 3 an−1 n an n+1
X an−1 n
f (t)dt = a0 x + x + x + ··· + x + x + ··· = x .
0 2 3 n n+1 n=1
n

Theorem 3.70. Let R > 0 be the radius of convergence of the power series
+∞
X
2 3 n
a0 + a1 x + a2 x + a3 x + · · · + an x + · · · = an xn .
n=0

Let f be the function defined on (−R, R) by

+∞
X
f (x) := a0 + a1 x + a2 x2 + a3 x3 + · · · + an xn + · · · = an x n , −R < x < R.
n=0
If the series
+∞
X
an Rn
n=0
converges, then the limit limx↑R f (x) exists and
+∞
X
lim f (x) = an R n .
x↑R
n=0
If the series
+∞
X
an (−R)n
n=0
converges, then the limit limx↓R f (x) exists and
+∞
X
lim f (x) = an (−R)n .
x↓R
n=0

Example 3.71. By (3.33) we have

1
= 1 + x + x2 + x3 + · · · + xn + · · · for all − 1 < x < 1. (3.40)
1−x
Thus the function f (x) = 1/(1 − x) defined for x ∈ (−1, 1) can be represented by a power
series. Applying Theorem 3.69 we get
1
= 1 + 2 x + 3 x2 + 4 x3 + · · · + n xn−1 + (n + 1) xn + · · · for all − 1 < x < 1.
(1 − x)2
Example 3.72. Substituting −x for x in (3.40) we get
1
= 1 − x + x2 − x3 + · · · + (−1)n xn + · · · for all − 1 < x < 1 . (3.41)
1+x
Thus the function f (x) = 1/(1 + x) defined for x ∈ (−1, 1) can be represented by a power
series. Applying Theorem 3.69 (d) we get
Z x
1 1 1 1 1
ln(1+x) = dt = x− x2 + x3 − x4 +· · ·+(−1)n+1 xn +· · · for all −1 < x < 1 .
0 1 + t 2 3 4 n
90 3. INFINITE SERIES

For x = 1 the above series is the alternating harmonic series which converges conditionally.
By Theorem 3.70 we have
+∞
X 1
lim ln(1 + x) = (−1)n+1 .
x↑1 n
n=0

Since the function ln(1 + x) is continuous at x = 1 we have

lim ln(1 + x) = ln 2.
x↑1

Thus we found the sum of the alternating harmonic series

+∞
X 1
(−1)n+1 = ln 2.
n=0
n

Example 3.73. Substituting x2 for x in (3.41) we get

1
= 1 − x2 + x4 − x6 + · · · + (−1)n x2n + · · · for all − 1 < x < 1.
1 + x2
Thus the function f (x) = 1/(1 + x2 ) defined for x ∈ (−1, 1) can be represented by a power
series. Applying Theorem 3.69 (d) for all x ∈ (−1, 1) we get
Z x
1 1 1 1 1
arctan(x) = 2
dt = x − x3 + x5 − x7 + · · · + (−1)n+1 x2n−1 + · · · .
0 1+t 3 5 7 2n − 1
With x = 1 the above series is
1 1 1 1
1 − + − + · · · + (−1)n+1 + ···
3 5 7 2n − 1
is a conditionally convergent alternating series. By Theorem 3.70 we have
+∞
X 1
lim arctan x = (−1)n−1 .
x↑1 2n − 1
n=1

We did not prove it, but it can be proved that arctan x is a continuous function. Therefore
π
lim arctan x = arctan 1 = .
x↑1 2
Thus we found a representation of π as an infinite sum:
+∞
π 1 1 1 1 X 1
= 1 − + − + · · · + (−1)n+1 + ··· = (−1)n−1 .
4 3 5 7 2n − 1 2n − 1
n=1

3.4.3. Taylor series at 0 (Maclaurin series). In the preceding section we found

power series representations for several well known functions. It turns out that all well
known functions can be represented as power series. The key step in finding the power series
representation of elementary functions are formulas (3.39) which establish the relationship
between the coefficients an , n = 0, 1, 2, . . . , of a power series and the derivatives of the
function f which is represented by that power series. We rewrite formulas (3.39) as
1 ′′ 1 (3) 1
a0 = f (0), a1 = f ′ (0), a2 = f (0), a3 = f (0), . . . , an = f (n) (0), . . . . (3.42)
2! 3! n!
 3.4. INFINITE SERIES OF FUNCTIONS 91

Let a > 0 and let f be a function defined on (−a, a). Assume that f has all derivatives
on (−a, a). Then the series power series
+inf
Xty 1
1 ′′ 1 1
f (0) + f ′ (0)x + f (0)x2 + f (3) (0)x3 + · · · + f (n) (0)xn + · · · = f (n) (0)xn
2! 3! n! n!
n=0
is called Taylor series at 0 or Maclaurin series of f .
Using formulas (3.42) it is not difficult to calculate a Maclaurin series for a given func-
tion. The difficulties arise in proving that the function defined by such power series is
identical to the given function. Fortunately this is true for all well known functions.
Example 3.74. Let f (x) = ex = exp(x), x ∈ R. Then f (n) (x) = ex for all n =
0, 1, 2, . . .. Therefore the coefficients of the Maclaurin series for the function exp are an =
1/n! and it can be proved that for all x ∈ R we have
1 1 1
ex = 1 + x + x2 + x3 + · · · + xn + · · · .
2! 3! n!
Example 3.75. Let f (x) = sin(x), x ∈ R. Then
f ′ (x) = cos(x), f ′′ (x) = − sin(x), f (3) (x) = − cos(x), f (4) (x) = sin(x).
Consequently,
f (2k) (0) = 0, f (2k+1) (0) = (−1)k , for all k ∈ N ∪ {0}.
Therefore the coefficients of the Maclaurin series for the function sin are
1
a2k = 0, a2k+1 = (−1)k , for all k ∈ N ∪ {0}.
(2k + 1)!
It can be proved that for all x ∈ R we have
1 1 1 1
sin(x) = x − x3 + x5 − x7 + · · · + (−1)k x2k+1 + · · · .
3! 5! 7! (2k + 1)!
Example 3.76. Let f (x) = cos(x), x ∈ R. Then
f ′ (x) = − sin(x), f ′′ (x) = − cos(x), f (3) (x) = sin(x), f (4) (x) = cos(x).
Consequently,
f (2k) (0) = (−1)k , f (2k+1) (0) = 0, for all k ∈ N ∪ {0}.
Therefore the coefficients of the Maclaurin series for the function cos are
1
a2k = (−1)k , a2k+1 = 0 for all k ∈ N ∪ {0}.
(2k)!
It can be proved that for all x ∈ R we have
1 1 1 1 2k
cos(x) = 1 − x2 + x4 − x6 + · · · + (−1)k x + ··· .
2! 4! 6! (2k)!
Example 3.77 (The Binomial Series). Let α ∈ R. Let f (x) = (1 + x)α , x ∈ (−1, 1).
Then
f ′ (x) = α(1 + x)α−1 ,
f ′′ (x) = α(α − 1)(1 + x)α−2 ,
f (3) (x) = α(α − 1)(α − 2)(1 + x)α−3 ,
92 3. INFINITE SERIES

..
.
f (n) (x) = α(α − 1) · · · (α − n + 1)(1 + x)α−n
..
.
Therefore the coefficients of the Maclaurin series for the function f are
α(α − 1) · · · (α − n + 1)
a0 = 1, an = , n ∈ N.
n!
It can be proved that for all x ∈ (−1, 1) we have
α α(α − 1) 2 α(α − 1)(α − 2) 3 α(α − 1) · · · (α − n + 1) n
(1+x)α = 1+ x+ x + x +· · ·+ x +· · · .
1! 2! 3! n!
This series is called binomial series. The reason for this name is that for α ∈ N the binomial
series becomes a polynomial:
(1 + x)1 = 1 + x
(1 + x)2 = 1 + 2x + x2
(1 + x)3 = 1 + 3x + 3x2 + x3
(1 + x)4 = 1 + 4x + 6x2 + 4x3 + x4
(1 + x)5 = 1 + 5x + 10x2 + 10x3 + 5x4 + x5
(1 + x)6 = 1 + 6x + 15x2 + 20x3 + 15x4 + 6x5 + x6
..
.
m    
m
X m k m m!
(1 + x) = x , were m ∈ N and =
k k k!(m − k)!
k=0
The last formula is called the binomial theorem. The coefficients
 
m m! m(m − 1) · · · (m − k + 1)
= = with m, k ∈ N, 0 ≤ k ≤ m,
k k!(m − k)! k!
are called binomial coefficients. With a general α ∈ R and k ∈ N the coefficients
 
α α(α − 1) · · · (α − k + 1)
=
k k!


are called generalized binomial coefficients. By definition α0 = 1. With this notation the
binomial series can be written as
+∞  
α
X α
(1 + x) = xk for qll x ∈ (−1, 1). (3.43)
k
k=0
Notice that formula (3.40) is a special case of (3.43), since
 
−1 (−1)(−2) · · · (−1 − k + 1) (−1)k k!
= = = (−1)k .
k k! k!
Notice also that differentiating (3.40) leads to
+∞
X
(1 + x)−2 = 1 + (−1)k (k + 1) xk for all − 1 < x < 1.
k=1
 3.4. INFINITE SERIES OF FUNCTIONS 93

This is a binomial series with α = −2. To verify this we calculate

 
−2 (−2)(−3) · · · (−2 − k + 1) (−1)k (k + 1)!
= = = (−1)k (k + 1).
k k! k!
For α = 1/2 the expression
  1




1/2 2 − 23 · · · 21 − k + 1
− 21
=
k k!
1 1




2 −2 − 2 · · · − 2k−3
3
2
=
k!
k−1
(−1) 1 · 3 · · · · (2k − 3)
=
2k k!
Thus
√ 1 1 1·3 1·3·5 4 1·3·5·7 5
1 + x = 1 + x − 2 x2 + 3 x3 − 4 x + x + · · · for all − 1 < x < 1 .
2 2 2! 2 3! 2 4! 25 5!
Example 3.78. Let f (x) = arcsin(x), x ∈ [−1, 1]. To calculate the Maclaurin series for
arcsin we notice that
d
1
arcsin(x) = √ , x ∈ (−1, 1).
dx 1 − x2
Now calculate the Maclaurin series for the last function using the binomial series with
α = −1/2. For α = −1/2 and k ∈ N, we calculate
 



−1/2 − 12 − 32 − 25 · · · − 12 − k + 1
=
k k!




− 2 − 2 − 2 · · · − 2k−1
1 3 5
2
=
k!
1 · 3 · · · · · (2k − 1)
= (−1)k
2k k!
Thus
1 1 1·3 1·3·5 3 1·3·5·7 4
√ = 1 − x + 2 x2 + 3 x − x + · · · for all − 1 < x < 1 ,
1+x 2 2 2! 2 3! 24 4!
that is,
+∞
1 X 1 · 3 · · · · · (2k − 1) k
√ =1+ (−1)k x ,
1+x 2k k!
k=1
or using the notation of double factorials
+∞
1 X (2k − 1)!! k
√ =1+ (−1)k x .
1+x (2k)!!
k=1
Substituting −x2 instead of x in the above formula we get
+∞
1 X (2k − 1)!! 2k
√ =1+ x , for all − 1 < x < 1.
1−x 2 (2k)!!
k=1
Since Z x
1
√ dt = arcsin(x),
0 1 − t2
94 3. INFINITE SERIES

integrating the last power series we get

+∞ +∞ 2k

X (2k − 1)!! 2k+1
X
arcsin(x) = x + x = k
x2k+1 , for all −1<x<1
(2k + 1)(2k)!! 4k (2k + 1)
k=1 k=0
It is interesting to note that the above expansion holds at both endpoints x = −1 and
x = 1. To prove this we need to recall Theorem 3.69 (a) and prove that the series
+∞
X (2k − 1)!!
1+
(2k + 1)(2k)!!
k=1
converges. This series converges by The Comparison Test. (Hint: Prove by mathematical
(2k − 1)!! 1
induction that < √3
for all k ∈ N.) As a consequence we obtain that
(2k)!! k
+∞ +∞ 2k


X (2k − 1)!! X
k π
1+ = k
= .
(2k + 1)(2k)!! 4 (2k + 1) 2
k=1 k=0
 Index

absolute value function, 12 range of a function, 9

real numbers, 5
codomain of a function, 9 rounding function, 11
domain of a function, 9

floor function, 10 sign function, 9

function, 8

limit at +∞, 15, 19

triangle inequality, 13
maximum of a set, 8
minimum of a set, 8

pizza-party inequality, 6 unit step function, 9

95