# 1

Assessment statements 2
1 Sequences, Limits and Improper Integrals 3
1.1 Introduction 3
1.2 Infnite sequences 3
1.3 Improper integrals and infnite limits of integration 18
Exercise 1.1 21
2 Series and Convergence 23
2.1 Introduction 23
2.2 Convergence of infnite series 33
2.3 Alternating series and absolute convergence 51
Exercise 2.1 60
3 Power Series 63
3.1 Introduction 63
3.2 Power series 64
3.3 Deriving power series – Taylor and Maclaurin series 70
3.4 Operations with power series 74
Exercise 3.1 82
Practice Questions (Chapters 1–3) 84
4 Differential Equations 93
4.1 Introduction 93
4.2 Slope felds 96
4.3 Separable equations 101
4.4 First order linear differential equations – use of integrating factor 105
4.5 Homogeneous differential equations 111
4.6 Euler’s method 117
Exercise 4.1 122
Practice Questions (Chapter 4) 126
Index 131
Contents
Series and Differential Equations
2
Assessment statements
10.1 Infnite sequences of real numbers.
Limit theorems as n approaches infnity.
Limit of a sequence.
Improper integrals of the type f x d
a
x ( )

.
The integral as a limit of a sum; lower sum and upper sum.
10.2 Convergence of infnite series.
Partial fractions and telescoping series (method of differences).
Tests for convergence: comparison test; limit comparison test; ratio
test; integral test.
The p-series,
1
n
p

.
Use of integrals to estimate sums of series.
10.3 Series that converge absolutely.
Series that converge conditionally.
Alternating series.
10.4 Power series: radius of convergence and interval of convergence.
Determination of the radius of convergence by the ratio test.
10.5 Taylor polynomials and series, including the error term.
Maclaurin series for e x x x x x
x
p
, sin , cos , arctan , ln , 1 1 + ( ) + ( ) . Use
of substitution to obtain other series.
The evaluation of limits of the
form lim
x a
f x
g x

( )
( )
using l’Hôpital’s rule and/or the Taylor series.
10.6 First order differential equations: geometric interpretation using slope
felds; numerical solution of
dy
dx
f x y = ( ) , using Euler’s method.
Homogenous differential equation
dy
dx
f
y
x
=

using the substitution
y vx = .
Solution of ′ + ( ) = ( ) y P x y Q x , using the integrating factor.
Series and
Differential
Equations
3
1.1 Introduction
Important concepts regarding sequences, series and limits were covered
in previous textbook chapters on the core syllabus. It would be helpful to
go back and read through the frst four sections of Chapter 4, especially
the material on infnite geometric series in Section 4.4. The frst section
in Chapter 13 includes an informal approach to limits of functions and
also covers properties of limits. Central to any discussion about sequences,
series and limits is the concept of a function. Thus, it may also prove
worthwhile to review some of the fundamental ideas, terminology and
notation for functions covered in the frst section of Chapter 2.
Arithmetic and geometric series, both fnite and infnite, were discussed in
Chapter 4. In this option topic we are primarily interested in characteristics
and applications of infnite series. As you will see, infnite series are
mathematically interesting and have very useful applications. Our
treatment of series in this option topic will require a more formal approach
than taken in Chapter 4. In order to develop a rigorous theory of infnite
series, we must frst consider infnite sequences of numbers.
1.2 Infnite sequences
Sequences occur in many areas of mathematics. For example, the positive
even numbers less than or equal to 10 form a sequence:
2, 4, 6, 8, 10.
This sequence is fnite because the list of numbers ends with a specifc
number, 10 in this case. If a sequence does not end, it is infnite. We will
be focusing on infnite sequences, so from now on if we use the word
‘sequence’ it is understood that we are referring to an infnite sequence.
From the defnition it is understood that an infnite sequence is a rule
that associates a number to each positive integer. The number associated
with the integer n is called the nth term of the sequence. Instead of
using the familiar function notation f n ( ) to represent the value (term)
Sequences, Limits
and Improper
Integrals
1
Sequences and series are closely
related, so you need to be careful
to apply these words correctly.
A sequence is an ordered list of
numbers commonly written out
with commas separating the
numbers. A series is a sum of a
sequence. The fnite sequence
1
1
2
1
4
1
8
, , , is an ordered list
whereas the closely related
fnite series 1
1
2
1
4
1
8
+ + + is a sum
that is precisely equal to the
number
15
8
.
Defnition of a sequence
A sequence of numbers is a discrete
function whose domain is the set of
positive integers, 
+
.
4
1
Sequences, Limits and Improper Integrals
of a sequence f for a certain positive integer n, it is customary to use a
subscripted letter, such as a
n
or u
n
. Hence, we will denote a sequence by
a a a a
n 1 2 2
, , , , ,   { }, or more simply with the notation a n
n
{ } ∈
+
,  .
It follows that a
n
is an explicit formula (sometimes called a closed
formula) that is a function whose domain, n, is the set of positive integers
and generates the value of the nth term of a sequence. The notation a
n
{ }
represents all the terms of a sequence, not just a single term. For example,
for the sequence formed by the reciprocals of the positive integers, we can
write a
n
{ } = 1
1
2
1
3
, , ,  and a
n
n
=
1
.
Example 1 – Listing the terms of a sequence
a) The terms of the sequence a
n
n
{ } =

1
1
Ϫ are 0
1
2
2
3
3
4
4
5
, , , , , . …
b) The terms of the sequence b
n
n
n
{ } =
( )

+
Ϫ1
1
are 1
1
2
1
3
1
4
1
5
, , , , , . Ϫ Ϫ …
c) The terms of the sequence c
n
n
n
{ } =
( )

2
1
1 Ϫ
Ϫ !
are 1
2
1
4
2
8
6
16
24
32
120
, , , , , , . …
The frst six terms of the sequence c
n
{ } can be simplifed to
1 2 2
4
3
2
3
4
15
, , , , , , . … This highlights the fact that although it is often
helpful to view some of the initial terms in an infnite sequence, knowing
the explicit formula for the value of the nth term is even more useful.
(Note: Evaluating the frst term in the sequence c
n
{ } required using the
defnition that 0 1 ! = .)
Example 2 – A sequence defned by a recursive formula
It is not necessary for a sequence to be defned by an explicit formula, as in
Example 1. The sequence a
n
{ } defned by
a a
1 2
1 = = , a a a
n n n + +
= +
2 1
for n ജ1
is a sequence that we saw in Chapter 4 of the book. The rule giving a
n + 2
in
terms of a
n + 1
and a
n
is an example of a recursion formula. It produces the
famous Fibonacci sequence:
1, 1, 2, 3, 5, 8, 13, … .
Our foremost concern with a sequence a
n
{ } is whether a
n
has a limit L as
n approaches infnity (n → ∞). If it does, we say that a
n
{ } converges to L;
otherwise we say that a
n
{ } diverges.
Since a sequence is a type of function, it seems appropriate that in our
investigation of limits of sequences, we can apply the same ideas from
our work with limits of functions in Chapter 13 of the book. A function
f whose domain is the half-open interval 1, ∞
[ [
can be converted into a
sequence by restricting its domain to the integers in that interval, i.e. the
Although a bit complicated,
an explicit formula exists for
the nth term of the Fibonacci
sequence. In general, the rules
for sequences and series in this
chapter will be explicit rather
than recursive. See Chapter 4
of the book for discussion of
explicit and recursive formulae
for sequences.
5
positive integers 
+
. Conversely, given a sequence a
n
{ }, it is often possible
to defne a function f on 1, ∞
[ [
such that f n a
n
( ) = for each integer n > 0.
Thus, if it was established that lim
x
f x L
→∞
( ) = , it would necessarily follow
that lim
n
n
a L
→∞
= . Therefore, results obtained in Chapter 13 of the book for
limits of functions are available for our work with limits of sequences. In our
development of the derivative through a limit process, we stated an informal
defnition of a limit of a function and fve properties of limits (Section 13.1).
Let’s start by using the informal defnition for the limit of a function (from
Section 13.1 of the book) to develop a more precise defnition for the limit
of a sequence.
Our earlier informal defnition of a limit of a function said that if f x ( )
becomes arbitrarily close to a unique fnite number L as x approaches
c from either side, then the limit of f x ( ) as x approaches c is L. What
precisely does ‘arbitrarily close’ mean? Let’s try to answer this question by
looking at a specifc sequence.
We have seen a few sequences already in this section that seem to have a
limit that is not infnite. One such limit is the one in Example 1, part a) –
namely a
n
n
{ } =

1
1
Ϫ which starts with 0
1
2
2
3
3
4
4
5
, , , , , . …. It appears that
lim
n
n
a
→∞
= 1. The graph of a
n
{ } is shown in Figure 1.1.
1.2
1
0.8
0.6
0.4
0.2
1 +
5 n
y
10 15 20
+
+
+
+
+
+
+
+
+ + + + + + + + + + +
+
1
1 
We can ensure that a
n
is within an arbitrarily small distance of 1 for all
suffciently large n. For example, 1
1
2
1
1
2
Ϫ < < + a
n
for all n > 2,
1
1
10
1
1
10
Ϫ < < + a
n
for all n > 10 , 1
1
100
1
1
100
Ϫ < < + a
n
for all n > 100,
etc. Indeed, if we graph a pair of horizontal lines, one at an arbitrarily
small distance of ε below 1 and the other the same distance of ε above 1,
then the points representing a
n
lie between these lines for all suffciently
large n, that is, 1 1 Ϫε ε < < + a
n
for all suffciently large n. The idea we are
developing here is that no matter how small we choose ε to be ( ε > 0 ), we
observe that 1 1 Ϫε ε < < + a
n
for all n >
1
ε
. If we choose N to be the
smallest integer such that N ജ
1
ε
, then 1 1 Ϫε ε < < + a
n
for all n N > .
The ffth letter in the Greek
alphabet, ε (epsilon), is
typically used in mathematics
to represent a small but strictly
positive quantity.
Figure 1.1 Graph of the frst
twenty terms of the sequence
a
n
n
{ } =
{ }
1
1
Ϫ .
6
Sequences, Limits and Improper Integrals
1
Applying one of the properties of absolute value inequalities (see Section
3.5 of the book), we have a
n
Ϫ1 < ε for all n N > .
Essentially we have come up with a more precise way of saying that as n
gets very large the terms of the sequence a
n
n
{ } =

1
1
Ϫ approach the fxed
number 1. This is expressed as follows:
Because for any given ε > 0, there always exists some positive integer N
such that all the terms of the sequence where the subscript is larger than N
will be less than a distance of ε away from the number 1, then the limit of
the value of the terms in the sequence is 1. That is, the sequence converges
to the limit of 1.
Using this idea, we now state the formal defnition for the limit of a
sequence.

Graphically, as illustrated in Figure 1.1 for the sequence 1
1
Ϫ
n

, this
ε – N defnition says that eventually the terms of a sequence that converges
to L will lie within the horizontal lines y L = + ε and y L = Ϫε.
Example 3 – Using the ε – N defnition to fnd the limit of a
sequence
In previous work in the core syllabus of this course (and just from
common sense) we have reasoned that the limit of the function f x
x
( ) =
1

as x → ∞ is zero, i.e. lim
x
x
→∞
=
1
0. The domain of f is x x x : , ∈ ≠ { }  0 .
If we change the domain to positive integers n n : ∈
{ }
+
 , we get the
sequence a
n
n
{ } =

1
.
Prove that the sequence
1
n

converges to zero, i.e. lim
n
n
→∞
=
1
0.
Solution
Let ε > 0. We begin by writing down the inequality from the ε – N
defnition with a
n
n
=
1
and L = 0. This gives
a L
n n
n
Ϫ Ϫ = = <
1
0
1
ε.
We now need to fnd a positive integer N such that
1
n
< ε for all n N > . It
is clear that
1
n
< ε for all n >
1
ε
. However, there is no reason to expect
1
ε

Defnition of the limit of a sequence (ε – N defnition)
The sequence a
n
{ } converges to the limit L if, for any ε > 0, there exists a positive
integer N such that a L
n
Ϫ < ε, for all n N > .
A sequence that has a limit
converges, whereas a
sequence that does not have a
limit diverges.
7
to be an integer; but this is easily addressed by choosing any integer N ജ
1
ε
.
Then every value of the subscript n greater than N will automatically be
greater than
1
ε
. It follows that
1
0
1
n n
Ϫ = < ε for all n N > . Therefore,
the sequence
1
n

converges to zero.
Example 4 – Using the ε – N defnition to fnd the limit of a
sequence
Prove that the limit of the sequence
n
n 2 1 +

is
1
2
.
Solution
We must show that for any ε > 0 there exists a positive integer N such that
if n is an integer and n N > then
n
n 2 1
1
2 +
< Ϫ ε.
Some algebra working produces the following:
n
n
n n
n n n
n
2 1
1
2
2 2 1
2 2 1
1
2 2 1
1
2 2 1
2
+
=
+ ( )
=
+ ( )
=
+ ( )
< ⇒ Ϫ
Ϫ Ϫ Ϫ
ε ++ > ⇒ > 1
1
2
1 2
4 ε
ε
ε
n
Ϫ
If we choose a positive integer N ജ
Ϫ 1 2
4
ε
ε
and n N > then it follows that
n
n 2 1
1
2 +
< Ϫ ε.
Thus the defnition of a limit of sequence is satisfed and the limit of the
sequence
n
n 2 1 +

is
1
2
.
Note: To see how the ε – N defnition works in this case, take, for
instance, that ε =
1
8
. Then N ജ
Ϫ 1 2
1
8
4
1
8
3
4
1
2
3
2

= = . If we then choose
N = 2
3
2
ജ , then all n > 2 must satisfy the inequality
n
n 2 1
1
2
1
8 +
< Ϫ . For
example, if n = 3 then
n
n 2 1
1
2
3
2 3 1
1
2
1
14 +
=
( ) +
= Ϫ Ϫ ; and since
1
14
1
8
<
then the condition a L
n
Ϫ < ε, for all n N > , has been met. The proof
shows that this condition is met for all n N > so long as N is chosen so
that N ജ
Ϫ 1 2
4
ε
ε
.
In Section 13.1 of the book, we used some algebraic techniques combined
with some informal reasoning to fnd limits of rational functions similar
to the rule for the sequence in Example 4. And in Example 3, we noted the
8
Sequences, Limits and Improper Integrals
1
function f x
x
( ) =
1
is associated with the sequence
1
n

. It seems reasonable
to conjecture that for a sequence a
n
{ } if the value of a
n
matches a function
f at every positive integer, and f x ( ) approaches a limit L as x → ∞, then
the sequence will converge to the same limit L.
Also, in Section 13.1 we presented a set of fve properties for limits of
functions. All of these can be translated into properties for limits of
sequences. We list here the set of fve corresponding properties of limits of
sequences and an additional important property on the limit of a rational
power of a sequence.
These six properties of limits of sequences can be stated in words as follows:
1. The limit of a constant is equal to the constant.
2. The limit of a constant times a sequence is the constant times the limit
of the sequence.
3. The limit of a sum/difference of sequences is the sum/difference of the
limits of the sequences.
4. The limit of a product of sequences is the product of the limits of the
sequences.
5. The limit of a quotient of sequences is the quotient of the limits of the
sequences (given that the limit of the sequence in the denominator is
not zero).
6. The limit of a rational power of a sequence is the rational power of the
limit of the sequence.
The converse of the limit of
a sequence theorem is not
true. That is, a convergent
sequence does not imply that
the associated real variable
function must also converge.
Limit of a sequence theorem
Suppose that f x ( ) is a function defned for all x ജk , k ∈
+
 , and a
n
{ } is a sequence
such that a f n n k
n
= ( ) when ജ . If lim
x
x
→∞
( ) = f L, then lim
n
n
a L
→∞
= .
Properties of limits of sequences
If a
n
{ } and b
n
{ } are convergent sequences such that lim
n
n
a L
→∞
= and lim
n
n
b K
→∞
= , and c is
any real number, then:
1. Constant sequence: lim
n
c c
→∞
=
2. Scalar multiple of a sequence:
lim
n
n
c a cL
→∞
⋅ ( ) =
3. Sum or diference of sequences: lim
n
n n
a b L K
→∞
± ( ) = ±
4. Product of sequences: lim
n
n n
a b LK
→∞
⋅ ( ) =
5. Quotient of sequences:
lim ,
n
n
n
a
b
L
K
→∞

=

K ≠ 0
6. Rational power of a sequence:
lim ,
n
n
p
p
a L p
→∞
( ) = ∈
9
In Chapter 13 of the book we reasoned informally that function values for
functions in the form f x
x
k
( ) =
1
, where k is a rational number, approach
zero as x goes to zero, i.e. lim ,
x
k
x
k
→∞
= ∈
1
0 . Thus, it makes sense that the
result from Example 3, lim
n
n
→∞

=
1
0, combined with property 6 for limits
of sequences above, leads to the following intuitive rule for the limit of
certain sequences.
Example 5
Determine whether the sequence
3 5 1
2 1
2
2
n n
n
+
+

Ϫ
is convergent or
divergent.
Solution
In Example 4, part d) of Section 13.1, we found lim
x
x x
x
→∞
+
+
3 5 1
2 1
2
2
Ϫ
to be
equal to
3
2
as follows:
lim lim
x x
x x
x
x
x
x
x x
x
x
→∞ →∞
+
+
=
+
+
3 5 1
2 1
3 5 1
2 1
2
2
2
2 2 2
2
2
Ϫ
Ϫ
xx
2

=
+
+
→∞
lim
x
x x
x
3
5 1
2
1
2
2
Ϫ

=
+
+
→∞ →∞ →∞
→∞ →∞
lim lim lim
lim lim
x x x
x x
x x
x
3
5 1
2
1
2
2
Ϫ

=
+
+
3 0 0
2 0
Ϫ
Hence, lim .
x
x x
x
→∞
+
+
=
3 5 1
2 1
3
2
2
2
Ϫ
Therefore, from the limit of a sequence theorem above, we can conclude
that the sequence
3 5 1
2 1
2
2
n n
n
+
+

Ϫ
is convergent and it converges to
3
2
.
In our discussion of the end behaviour of rational functions in Section 3.4
of the book, the following limit results were hinted at. We state them here
because by means of the limit of a sequence theorem they can also be
Dividing numerator and denominator
by largest power of x, i.e. x
2
.

Applying lim
lim
lim
x
x
x
x
x
x
x

( )
( )

=
( )
( )
a
a
a
f
g
f
g
and
lim lim lim .
x x x
x x x x
→ → →
( ) ± ( ) [ ] = ( ) ± ( )
a a a
f g f g
Applying lim , .
x
x
→∞
= ∈
1
0
k
k 
If r r > ∈ 0, , then lim
n
r
n
→∞
=
1
0 . Note: This rule is equivalent to lim
n
r
n
→∞
= 0 if r < 0.
10
Sequences, Limits and Improper Integrals
1
applied in fnding limits of sequences with rules that are rational functions,
such as the sequence in Example 5.
Another useful limit theorem for functions that can be rewritten for
sequences is the squeeze theorem from Section 13.2 where we used it to
prove that lim
sin
x
x
x

=
0
1.
Note that the terms of
sequence b
n
{ } do not need to
lie between a
n
{ } and c
n
{ } for all
values of n. The requirement is
that there must be some value
of n for which all of the terms
of b
n
{ } beyond this value must
lie between a
n
{ } and c
n
{ }. This
is illustrated in Example 7.
Limits of rational functions
Let R be the rational function given by
R
f
g
a a a a
b b
n
n
n
n
m
m
m
m
x
x
x
x x x
x x
( ) =
( )
( )
=
+ + + +
+
− 1
1
1 0
1
Ϫ
Ϫ

ϪϪ1
1 0
+ + +

b b x
1. If n < m, then lim
x
x
→∞
( ) = R 0.
2. If n = m, then lim
x
x
→∞
( ) = R
a
b
n
m
.
3. If n > m, then lim
x
x
→∞
( ) = ∞ R , i.e. does not exist.
Example 6 – Applying the squeeze theorem
Show that each of the sequences converges, and fnd its limit.
a)
1
2
n

b)
cosn
n

Solution
a) Because 2 0 2
n n
n > > and for all positive integers n, it follows that
0
1
2
1
ഛ ഛ
n
n
for all integers n ജ1. We proved in Example 3 that
lim
n
n
→∞
=
1
0. Therefore, lim
n
n
→∞
=
1
2
0 because lim lim
n n
n
→∞ →∞
= = 0
1
0 and the
sequence
1
2
n

converges to zero.

The squeeze theorem for sequences
If a b c
n n n
ഛ ഛ for all n such that n N ജ , N ∈
+
 , and lim lim
n
n
n
n
a c L
→∞ →∞
= = then lim
n
n
b L
→∞
= .
See Figure 1.2 below.
L
n
a
n
b
n
c
n
Figure 1.2 The sequences
a c
n n
{ } { } and , both with
limit of L, ‘squeezing’ the
sequence b
n
{ }.
11
b) Because Ϫ ഛ ഛ 1 1 cos x for all real numbers x, it follows that

Ϫ
ഛ ഛ
1 1
n
n
n n
cos
for all integers n ജ1. Therefore, lim
cos
n
n
n
→∞
= 0 because
lim lim
n n
n n
→∞ →∞

= = Ϫ
1 1
0 and the sequence
cosn
n

converges to zero.
Example 7 – Applying the squeeze theorem for an alternating
sequence
Consider the infnite sequence
Ϫ1 ( )

n
n!
.
a) Write out the frst six terms of the sequence.
b) Use the squeeze theorem to show that the sequence converges to 0.
Solution
a) The frst six terms of the sequence are Ϫ Ϫ Ϫ 1
1
2
1
6
1
24
1
120
1
720
, , , , , .
The sequence clearly alternates between positive and negative terms.
b) In order to apply the squeeze theorem, we need to fnd two convergent
sequences that converge to 0 for which all terms for n N ജ of the
sequence
Ϫ1 ( )

n
n!
will be between. Two sequences that will work in this
case are Ϫ
1
2
n

and
1
2
n

, both of which converge to 0.
The frst six terms of these two sequences, respectively, are
Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ
1
2
1
4
1
8
1
16
1
32
1
64
, , , , , and
1
2
1
4
1
8
1
16
1
32
1
64
, , , , , .
Observe that for n = 1 2 3 , and , the terms of
Ϫ1 ( )

n
n!
are not between
Ϫ
1
2
n

and
1
2
n

; however they are for n ജ4. That is,
Ϫ ഛ
Ϫ
ഛ ജ
1
2
1 1
2
4
n
n
n
n
n
( )
!
, .
Therefore, by the squeeze theorem it follows that the sequence
Ϫ1 ( )

n
n!
converges to zero.
The squeeze theorem can also be used to prove that the sequence of
absolute values for the sequence in Example 7,
Ϫ1 1 ( )

=

n
n n ! !
, also
converges to 0 since the inequality 0
1 1
2
ഛ ഛ
n
n
!
is true for all n ജ4. In
fact, there is a very useful theorem that states that if the absolute value
12
Sequences, Limits and Improper Integrals
1
sequence converges to 0, then the original sequence consisting of positive
and/or negative terms also converges to 0. It is often more effcient to
consider the sequence of absolute values and then apply the following
theorem to the original sequence.
Proof of the absolute value theorem is fairly straightforward. Consider
the two sequences a
n
{ } and Ϫa
n
{ }; one with all positive terms and one
with all negative terms. Because both of these sequences converge to 0 and
Ϫ ഛ ഛ a a a
n n n
we can conclude by means of the squeeze theorem that
a
n
{ } must also converge to 0.
The sequence
1
2
n

, equivalent to
1
2
j
(
,
\
,
(
¦
¦
¦
¦
¦
¦
¦
¦
¦
¦
n
, in Example 6 part a) is a
geometric sequence with a common ratio, r, equal to
1
2
. It was shown
to converge to zero. For what values of r, other than
1
2
, is the geometric
sequence r
n
{ }
convergent? Figure 1.3 shows the graphs of geometric
sequences, r
n
{ }
, for different positive values of r.
1
1 n
a
n
r = 1
r > 1
0 < r < 1
When r > 1 the sequence r
n
{ }
increases without bound, i.e. for r > 1,
lim
n
n
r
→∞
= ∞. Visually it appears that for 0 1 < < r , lim
n
n
r
→∞
= 0. In
Example 6, part a), we used the squeeze theorem to prove that lim
n
n
r
→∞
= 0
when r =
1
2
. We can use a similar argument to show that lim
n
n
r
→∞
= 0
for any value of r in the interval 0 1 < < r .
Thus, we have lim
n
n
r
r
r
→∞
=
∞ >
< <

if
if
1
0 0 1
What about negative values of r? Figure 1.4 (page 13) shows the graphs of
geometric sequences, r
n
{ }
, for different negative values of r. There is no
graph of r
n
{ }
for r = Ϫ1. In this case, the terms would oscillate infnitely
between 1 and Ϫ1, and clearly the sequence does not converge to any
number.
The converse of the absolute
value theorem is not true.
That is, if lim
n
n
a
→∞
= 0 it does
lim
n
n
a
→∞
= 0.
Note that the graph of a
real-valued function, e.g.
f x x
x
( ) = ∈ 2 , , is a
continuous smooth
curve; however, the
graph of a sequence, e.g.
a n
n
n
{ } = { } ∈
+
2 ,  , is discrete
points because the domain
consists of only positive
integers.
Figure 1.3 Graph of the sequence
r
n
{ } for diferent positive values of r.
Absolute value theorem
For the sequence a
n
{ }, if lim
n
n
a
→∞
= 0 then lim
n
n
a
→∞
= 0.
13
1
1
n
a
n
r < 1
1 < r < 0

Clearly, when r < Ϫ1 the sequence r
n
{ }
alternates between positive and
negative values that increase without bound. Thus, for r < Ϫ1, lim
n
n
r
→∞
does
not exist. Considering Ϫ1 0 < < r we can also write the inequality as
0 1 < < r . Additionally, lim lim
n
n
n
n
r r
→∞ →∞
= . Using the result above that
lim
n
n
r
→∞
= 0 for 0 1 < < r , and since 0 1 < < r , we can conclude that
lim
n
n
r
→∞
= 0. Therefore, by the absolute value theorem and the obvious
fact that lim
n
n
→∞
= 0 0 it is true that lim
n
n
r
→∞
= 0 for the interval Ϫ1 1 < < r ,
which is equivalent to r < 1. It is also obvious that lim
n
n
→∞
= 1 1. Thus the
sequence r
n
{ }
is convergent for the interval Ϫ ഛ 1 1 < r and divergent for
other values of r. This result is summarized as follows.
Example 8 – The factorial function and exponential functions
Show that the sequence
x
n
n
!

converges to 0 for any real number x.
Solution
If x < 0, then the terms of the sequence will be alternately positive and
negative. With the intention of applying the absolute value theorem, all
that needs to be shown is that lim
!
n
n
x
n
→∞
= 0. This takes a bit of work. We
start by choosing some positive integer N such that N x > . It follows
that the sequence
x
N
n
j
(
,
\
,
(
¦
¦
¦
¦
¦
¦
¦
¦
¦
¦
is geometric. Because N x > then
x
N
< 1
and it must follow that lim .
n
n
x
N
→∞

= 0 We now focus our attention on all
of the values of n such that n N > . For these values of n, we can write the
following:
Convergence of geometric sequences theorem
For r n ∈ ∈
+
» » and the geometric sequence r
n
{ } is convergent for Ϫ ഛ 1 1 < r such that
lim
n
n
r
r
r
→∞
=
<
=

0 1
1 1
if
Figure 1.4 Graph of the sequence
r
n
{ } for diferent negative values
of r.
14
Sequences, Limits and Improper Integrals
1
x
n
x
N N N n
n n
n N
!
=
× × × + ( ) + ( )
− ( )
1 2 3 1 2

factors

Ϫ
x
N N
x N
N N
N
N
x
N
n
n N
n
N
n
N
n
n
! ! !
= =

Hence, 0 ഛ ഛ
x
n
N
N
x
N
n
N
n
! !

. The expression
N
N
N
!
is a constant and will not
change as n changes. We know that lim
n
n
x
N
→∞

= 0, so applying the property
lim lim
n
n
n
n
c a c a
→∞ →∞
⋅ ( ) = we get lim
! !
lim
!
n
N
n
N
n
n
N
N
N
x
N
N
N
x
N
N
N
→∞ →∞

=

= ( ) 0 == 0.
Thus, 0 0 ഛ ഛ
x
n
N
N
x
N
n
N
n
! !

= and we can conclude that lim
!
n
n
x
n
→∞
= 0.
Therefore, by the absolute value theorem lim
!
,
n
n
x
n
→∞
= 0 the sequence
x
n
n
!

converges to 0 for any real value of x.
We have one more important theorem to consider that is an essential tool
for helping to determine the limits of certain functions, and consequently
the limits of certain sequences.
With limits of rational functions in Chapter 13 of the book, we were
sometimes confronted with an expression of indeterminate form,
commonly in the form
0
0
or

. We handled these by performing some
algebraic manipulations and applying limit theorems, as illustrated in
Example 5 of Chapter 13. Not all limits can be managed in such a way. The
following theorem specifcally addresses limits of rational expressions that
are of indeterminate form.
L’Hôpital’s rule states simply that, given the right conditions, the limit
of a quotient of functions is equal to the limit of the quotient of their
derivatives. It is important to frst verify the conditions regarding the limits
of f and g before applying l’Hôpital’s rule.
Because we have shown that
for any number x, lim
!
n
n
n
→∞
=
x
0,
we can conclude that the
factorial function increases
faster than any exponential
function.
When you are applying
l’Hôpital’s rule make sure that
you diferentiate the numerator
and denominator separately.
Do not use the quotient rule
for diferentiation.
L’Hôpital’s rule
Let f and g be functions whose derivative can be found at any value in an open
interval a b , ] [, except possibly at some value c where a c b < < . Assume that ′ ( ) ≠ g x 0,
except possibly at c. Suppose that lim
x
x

( ) =
c
f 0 and lim
x
x

( ) =
c
g 0; or
lim
x
x

( ) = ±∞
c
f and lim .
x
x

( ) = ±∞
c
g

(That is, the expression
f
g
x
x
( )
( )
is in indeterminate
form of
0
0
or

.)
Then lim lim
x x
x
x
x
x
→ →
( )
( )
=
′ ( )
′ ( )
c c
f
g
f
g

provided the limit on the right side exists (or is infnite).
15
Example 9 – Applying l’Hôpital’s rule
For each limit, use your GDC to conjecture a result, and then fnd the limit
using l’Hôpital’s rule.
a) lim
x
x
x
e
→0

b) lim
sec
tan
x
x
x

+
π
2
1
c) lim
x
x
x
e x

+
( )
0
1
Solution
a) To visualize lim
x
x
x
e
→0

we graph f x
x
e
x
( ) =

as shown in the GDC
images below.
WINDOW
Xmin=-2
Xmax=2
Xsc1=1
Ymin=-3
Ymax=1
Ysc1=1
Xres=1
Plot1
Y1= X/(1-e
ˆ
(X))
Plot2 Plot3
Y2=
Y3=
Y4=
Y5=
Y6=
Y7=

Although x = 0 is not in the domain of f, the graph appears to
pass through the point 0 1 ,Ϫ ( ) implying that lim
x
x
x
e

=
0
1
1
Ϫ
Ϫ . Since
lim
x
x

=
0
0 and lim
x
x
e

( )
=
0
1 0 Ϫ , lim
x
x
x
e
→0

is in the indeterminate
form
0
0
, and l’Hôpital’s rule applies. Differentiating the numerator
and denominator separately and evaluating the limit gives
lim lim .
x
x
x
x
x
e e
→ →
= = =
0 0
1
1 1
1
1
Ϫ Ϫ Ϫ
Ϫ
b) Instead of viewing a graph of f x
x
x
( ) =
+
sec
tan 1
to conjecture a value for
lim
sec
tan
x
x
x

+
π
2
1
, let’s use the GDC to construct a table of function values
near x = ≈
π
2
1 5708 . .
Y2=
Y3=
Y4=
Y5=
Y6=
TABLE SETUP
X Y1
X=1.6
1.54
1.55
1.56
1.57
1.58
1.59
1.6
.97057
.97984
.98938
.9992
1.0093
1.0198
1.0305
Tb1Start=1.5
∆Tbl=.01
Plot1
Y1= (1/cos(X))/(
1+tan(X))
Plot2 Plot3
The values in the table show that the function appears to be
approaching 1 from either direction.
The values of sec x vanish to +∞ when x →
π
2
from the left
i.e. x →

π
2
and vanish to −∞ when x →
+
π
2
.

16
Sequences, Limits and Improper Integrals
1
Similarly, lim tan
x
x

+ ( ) = +∞
π
2
1 and lim tan
x
x

+
+ ( ) = −∞
π
2
1 . So when
approaching
π
2
from the left we have
+∞
+∞
, and
−∞
−∞
when approaching
from the right. L’Hôpital’s rule also applies to one-sided limits. Applying
the rule to the right-hand limit gives
lim
sec
tan
lim
sec tan
sec
lim
x x x
x
x
x x
x
→ → →
+ +
+
= =
π π π
2 2
2
1
22
1
+
= sin . x
Note:
sec tan
sec
sin .
x x
x
x
2
simplifies to

The left-hand limit is also 1; therefore the two-sided limit is equal to 1,
i.e. lim
sec
tan
.
x
x
x

+
=
π
2
1
1
c) To visualize lim
x
x
x
e x

+
( )
0
1
we graph f x e x
x
x
( ) = +
( )
1
as shown in the
GDC images below.
If, after applying l’Hôpital’s rule,
the quotient of the derivatives
remains in indeterminate form
the rule can be applied more
than once.
WINDOW
Xmin=-.5
Xmax=.5
Xsc1=1
Ymin=-2
Ymax=14
Ysc1=1
Xres=1
Plot1
Y1= (e
ˆ
(X)+X)
ˆ
(1
Y1=(e
ˆ
(X)+X)^(1/X)
X=
-
.0106383Y=7.5094925
Plot2 Plot3
X)
Y2=
Y3=
Y4=
Y5=
Y6=
Y1=(e
ˆ
(X)+X)^(1/X)
X=.0106383 Y=7.2735696
X
X
Tracing on the graph indicates that as x →0 the function approaches
a value between 7.2735 and 7.5094. The exact value of the limit is not
clear.
We observe that lim
x
x
x
e x

+
( )
0
1
is in the indeterminate form 1

. However,
by taking the logarithm of both sides of f x e x
x
x
( ) = +
( )
1
and then
taking the limit we can change the indeterminate form to 0 0, to which
we can apply l’Hôpital’s rule.
ln ln ln
ln
f x e x
x
e x
e x
x
x
x
x
x
( ) [ ]
= +
( )

= +
( )
=
+
( ) 1 1
Thus, ln
ln
f x
e x
x
x
( ) [ ]
=
+
( )
, and taking the limit as x →0 of both sides
produces limln lim
ln
x x
x
f x
e x
x
→ →
( ) [ ]
=
+
( )
0 0
Right side in the form 0 0;

=
+
+
=
+
+
=

lim .
x
x
x
e
e x
e
e
0
0
0
1
1
1
0
2

apply l’Hôpital’s rule.
Hence, limln .
x
f x

( ) [ ]
=
0
2
Since f x e x
x
x
( ) = +
( )
1
, then lim lim
x
x
x
x
e x f x
→ →
+
( )
= ( )
0
1
0

=

( )
lim
ln
x
f x
e
0
Applying the rule e a
a ln
. =
=

lim .
x
e
0
2
Using result limln .
x
x

( ) [ ] =
0
2 f
17
Therefore, lim .
x
x
x
e x e

+
( )
=
0
1
2
e
2
7 389 ≈ . (to 4 s.f.), so the limit is within the range estimated from the
graph on the GDC.
Example 10 – Using l’Hôpital’s rule to determine convergence
of a sequence
Determine if the sequence a
n
n n
{ } =
+ 

2
1
3
converges. If it does, fnd its
limit.
Solution
Consider the function f x
x
x
x
( ) =
+

2
1
3
, , and its limit as x → ∞.
Since lim
x
x
x
→∞
+
2
1
3
is in indeterminate form of ∞ ∞, we can apply l’Hôpital’s
rule.

lim lim
ln
x
x
x
x
x x
→∞ →∞
+
=
( )
2
1
3
2
3 3
But this limit is still in indeterminate form of ∞ ∞, so we apply l’Hôpital’s
rule a second time.
lim lim
ln
lim
ln
x
x
x
x
x
x
x x
→∞ →∞ →∞
+
=
( )
=
( )
=
2
2
1
3
2
3 3
2
3 3
0
Because the value of a
n
matches the value of f x ( ) for every positive
integer, we can apply the limit of a sequence theorem and conclude that
lim .
n
n
n
→∞
+
=
2
1
3
0
Therefore, the sequence
n
n
2
1
3
+ 

converges to 0.
Example 11
Which sequence grows faster, lnn n { }
{ }
or ?
Solution
We can gain some insight into this question by graphing the real-valued
functions y x = ln and y x = . The graph (page 18) implies that the
sequence n
{ }
grows faster than lnn { }; that is, the infnite sequence
lnn
n

converges to 0. Using l’Hôpital’s rule to show that the limit of the function
f x
x
x
( ) =
ln
is 0 as x → ∞ will prove this result.
18
Sequences, Limits and Improper Integrals
1
5
4
3
2
1
5 10 15 20 25
y = ln x
y = x
x
y
lim
ln
lim lim
x x x
x
x
x
x
x
→∞ →∞ →∞
= = =
1
1
2
2
0 because lim ,
x
k
x
k
→∞
+
= ∈
1
0 
Therefore, the sequence
lnn
n

converges to 0, and we can conclude that
n
{ }
grows faster than lnn { }.
1.3 Improper integrals and infnite
limits of integration
Previously we have defned the defnite integral, f x dx
a
b
( )

, for a function
f that is continuous (i.e. no ‘gaps’ in the domain) for the fnite, bounded,
interval a x b ഛ ഛ . In this section, we will look at ways of evaluating
integrals where either one or both of the limits of integration (i.e. a and b)
are infnite, or the function f has an infnite discontinuity in the interval
a x b ഛ ഛ . An integral having either one of these characteristics is called an
improper integral.
Let’s look at an integral where one of the limits is infnite.
Example 12
Evaluate
1
2
1
x
dx

or show that it diverges.
Solution
We can replace the infnite limit of integration with a variable, say the
variable b, and then take the limit of the integral as b approaches infnity.
Taking the limit as b → ∞ gives
1 1 1
2
1
2
1
1
x
dx
x
dx
x
b
b
b
b

→∞ →∞
∫ ∫

j
(
,
\
,
(

,
¸
,
]
]
]
lim lim Ϫ +
j
(
,
\
,
(
+
→∞
lim .
b
b
Ϫ
1
1 0 1 1
1 1 1
2
1
2
1
1
x
dx
x
dx
x
b
b
b
b

→∞ →∞
∫ ∫

j
(
,
\
,
(

,
¸
,
]
]
]
lim lim Ϫ +
j
(
,
\
,
(
+
→∞
lim .
b
b
Ϫ
1
1 0 1 1
19
Therefore,
1
1
2
1
x
dx

= .
This result can be interpreted as indicating that the
area under the curve y
x
=
1
2
from one to infnity
is fnite and is exactly equal to 1 (see Figure 1.5 right).
Certainly, not all improper integrals converge to a fnite value.
Example 13
Evaluate
1
1
x
dx

or show that it diverges.
Solution
1 1
1 1
1
x
dx
x
dx x
b
b
b
b
b

→∞ →∞ →
∫ ∫
=

=
[ ]
= lim lim ln lim
∞∞ →∞
( ) = ( ) = ∞ ln ln lim ln b b
b
Ϫ 1
[or ‘limit does not exist’]
Therefore, the integral diverges. The area under the curve y
x
=
1
from 1 to
infnity is infnite.
The improper integral f x dx
a
b
( )

is called convergent if the corresponding
limit exists (as a fnite number as in Example 12), and is called divergent if
the limit does not exist (as in Example 13).
Example 14 – Using l’Hôpital’s rule to evaluate an improper
integral
Determine whether the integral
x
e
dx
x
1

converges or diverges; and if it
converges, fnd its value.
Solution
We can rewrite the integral as a limit,
x
e
dx
x
e
dx
x
b
x
b
1 1

→∞
∫ ∫
= lim ;
and now need to apply integration by parts to evaluate the integral.
Let
u x dv e dx
du dx v e
x
x
= =
= =
Ϫ
Ϫ
Ϫ
; then lim lim
b
x
b
b
x
b
x
b
x
e
dx xe e dx
→∞ →∞
∫ ∫
= 

+

1
1
1
Ϫ
Ϫ Ϫ
= 

→∞
lim
b
x x
b
xe e Ϫ Ϫ
Ϫ Ϫ
1
= + ( ) 

→∞
lim
b
x
b
x e Ϫ
Ϫ
1
1
=
+ ( )
+

→∞
lim
b
b
b
e e
Ϫ 1 2
3
2
1
1 2 3 4 5
x
y
1
y =
x
2
1
x
x x d =

ln , but note that
in Example 13 the absolute
value is omitted because the
integral is being evaluated
from 1 to ∞, i.e. only positive
numbers.
Figure 1.5 Area under the curve
y
x
=
1
2
from 1 to ∞.
20
Sequences, Limits and Improper Integrals
1
=
+ ( ) 

+
→∞
lim
b
b
b
e e
Ϫ 1 2
= + 0
2
e
Therefore,
x
e
dx
e
x
1
2
0 7357588823

= ≈ . (to ten signifcant fgures).
The GDC images below confrm our result. Note that even with an upper
limit of just x = 12 the defnite integral (computed on graph screen) agrees
to three decimal places with the value of the ‘improper’ integral with an
infnite upper limit; and when the upper limit is 20 (computed on home
screen) the values agree to six decimal places. The integral converges at a
fairly quick rate.
WINDOW
2/e
Xmin=0
.7357588823
fnInt(Y1,X,1,20)
.7357588391
Xmax=12
Xsc1=2
Ymin=
-
.05
Ymax=.4
Ysc1=.1
Xres=1
Plot1
Y1= X/e
ˆ
(X)
Plot2 Plot3
Y2=
Y3=
Y4=
Y5=
Y6=
Y7=
CALCULATE
1:value
2:zero
3:minimum
4:maximum
5:intersect
6:dy/dx
7:f(x)dx
f(x)dx=.73567901
What is an infnite discontinuity? A function f has an infnite discontinuity
at x c = if either lim
x c
f x

( ) = ∞ or lim
x c
f x

( ) = ∞ Ϫ such that x c →
from the right or left. For example, on the interval 0 1 ഛ ഛ x , the
function f x
x
( ) =
1
1
2
Ϫ
has an infnite discontinuity at x = 1 because
lim
x
x

= ∞
1
2
1

(note: x →1 from the left) which can be observed in
the graph in Figure 1.6.
5
4
3
2
1
1
1  x
2
1
x
y
y =
Figure 1.6
21
The region under the curve y
x
=
1
1
2
Ϫ
in the interval 0 1 ഛ ഛ x is
unbounded – and would, at frst thought, have an infnite area. However,
the unbounded region has a fnite area and we can fnd the exact area as
follows.
Example 15
Find the area, if possible (not possible if it’s infnite), under the curve
y
x
=
1
1
2
Ϫ
in the interval 0 1 ഛ ഛ x .
Solution
We can replace the limit of integration where the infnite discontinuity
occurs with a variable, say the variable b, and then take the limit of the
integral as b approaches the value of x where the discontinuity occurs
(approaching 1 from the left, in this case).
(Recall that the anti-derivative of
1
1
2
Ϫx
is arcsinx.)
Area = =

=
∫ ∫
→ →
− −
1
1
1
1
2 0
1
1
2 0
1
Ϫ Ϫ x
dx
x
dx
b
b
b
lim lim aarcsin lim arcsin arcsin x b
b
b
[ ]
= ( )

− 0
1
0 Ϫ
= ( ) = ( ) = ( ) =
→ →
− −
lim arcsin lim arcsin arcsin
b b
b b
1 1
0 1 Ϫ
ππ
2
Therefore, the unbounded region under the curve y
x
=
1
1
2
Ϫ
in the
interval 0 1 ഛ ഛ x has a fnite area of exactly
π
2
.
For questions 1–15, determine if the sequence converges or diverges. If it converges,
fnd the limit of the sequence.
1
7
3
n
{ }
2
2 1
1
2
2
n n
n
+ +
+

3
5 13
5
3
n
n n
Ϫ
+
{ }
4 cos nπ { } 5
Ϫ
Ϫ
1
2 1
1
( ) 

+ n
n
6 Ϫ
4
5
j
(
,
\
,
(
¦
¦
¦
¦
¦
¦
n
7
e
n
n
2

8
3 1
4 1
2
2 3
n
n
+

 Ϫ
9
2
1
n
n +

10 1
1
+
( ) 

Ϫ
n
n
11
n
n 1+
{ }
12 1
2
1
+
j
(
,
\
,
(
¦
¦
¦
¦
¦
¦
¦
¦
¦
¦
n
n
13
3
n
n!

14
ln
ln
2n
n
{ }
15
n
n
+ j
(
,
\
,
(
¦
¦
¦
¦
¦
¦
1
16 Use the squeeze theorem to show that lim
sin
n
n
n
→∞
=
2
0.
17 a Use the fact that lim
n
n
→∞
=
1
0
2
to prove that lim
n
n
n
→∞
+
=
2
2
1
1.
b Prove lim
n
n
n
→∞
+
=
2
2
1
1 by an ε – N proof.
Exercise 1.1
22
Sequences, Limits and Improper Integrals
1
18 Use the ε – N defnition for convergence of a sequence to prove that the
sequence 1 1 + ( )
{ }
Ϫ
n
is divergent.
For questions 19–21, use l’Hôpital’s rule to fnd the value of each limit.
19 lim
cos
x
x
x
→0
2

20 lim
x
x
x

+
1 2
1
3 2
Ϫ
Ϫ
21 lim
ln
x
x x

1
1 1
1
Ϫ
Ϫ
22 Determine whether the sequence n
n
sin
π
{ }
is convergent or divergent. If
convergent, fnd its limit. (Hint: Rewrite n
n
n
n
sin
sin
π
π
as
1
.)
In questions 23–28, evaluate the limit.
23 lim
x
x
x x
→1
2
2
1
4 3
Ϫ
Ϫ Ϫ
24 lim
x
x
x
→0
3
1 1 Ϫ Ϫ
25 lim
sin
x
x x
x
→0
3
Ϫ
26 lim
cot
x
x
x
x
→0
1
Ϫ
27 lim
ln
log
x
x
x
→∞
+ ( ) 1
2
28 lim , ,
x
x x
x

> >
0
0 0
a b
a b
Ϫ
29 Given f x x
x
( ) = + ( ) 1
1
, fnd lim
x
x
→∞
( ) f . (Hint: Start by taking the natural logarithm of
both sides, converting the right side to the indeterminate form
0
0
. Then you can
use l’Hôpital’s rule.)
In questions 30–37, evaluate, or identify as divergent, the given integral.
30
1
3
0
1
x
x

d 31
1
3
1
x
x

d
32
1
1
2
x
x
+
−∞

d 33
sinx
x
x
e
d
0

34 tanx x
0
2 π

d 35
1
1
0
+

e
d
x
x
36
1
1
0
1
Ϫx
x

d 37
x
x
x
k
d
k
2 2 0
Ϫ

38 a Find the area of the region bounded by x = 1
,
y
x
=
1
, and the x-axis.
b Find the volume of the solid generated by rotating this region about the
x-axis.
c Comment on the results for a and b.
23
2.1 Introduction
To start our study of infnite series in the option topic we consider using
the terms of a sequence a
n
{ } to form the sequence s
n
{ } of partial sums of
a
n
{ } as follows:
s a
1 1
=
s a a
2 1 2
= +
s a a a
3 1 2 3
= + +
We can use sigma notation to write the general expression for s
n
:
s a a a a a
n n i
i
n
= + + + + =
=

1 2 3
1
. . .
As pointed out in Section 4.4 (HL book) in our discussion on infnite
geometric series, the word ‘sum’ here is being used in a completely different
way from how it is normally used. Ordinary addition of real numbers is a
fnite process; hence, it does not make sense to fnd the ‘sum’ of infnitely
many terms. To be more precise, the ‘sum’ of an infnite series is a limit –
that is, the limit of the partial sums for the series. We can write the sum
as a a a a
n 1 2 3
+ + + + + . . . . . . but we must be careful not to assume that
the ‘+’ signs have the same properties to which we are accustomed. For
example, as we will see, a rearrangement of the terms of a convergent series
may change the value of its sum or even cause the series to diverge.
Example 1
a) Find the sum of the fnite series 21
1
1
6
( )
+
=

n
n
.
b) Consider the infnite series 2 2 2 2 1 1 1 1 1 1 1
1
1
( ) = + + +…
+
=

n
n
.

Determine if the series converges to a sum or diverges.
Series and
Convergence
2
Here we have used the letter
i as a subscript to indicate
the ith term of a sequence;
and have used the letter n
as a subscript to indicate the
nth partial sum. You need to
be comfortable with using
diferent letters for subscripts.
Defnition of the sum of an infnite series
If the sequence of partial sums s a
n i
i
n
{ } =

 =

1
converges, we say that its limit S is the
sum of the infnite series a a a
1 2 3
+ + + . . . and we write S a
i
i
=
=

1
. If the sequence s
n
{ }
diverges then we say that the infnite series a
i
i =

1
also diverges.
24
Series and Convergence
2
Solution
a) Clearly, 2 2 2 2 1 1 1 1 1 1 1 0
1
1
6
( ) = + + =
+
=

n
n
.
We can make the further observation that if the number of terms in this
fnite series was any even number, not just six, the sum is always 0; and,
if the number of terms is odd the sum is always 1. In either case, we can
‘pair up’ consecutive terms to get zero. For example,
2 2 2 2 1 1 1 1 1 1 1 0 0 0 0
1
1
6
( ) = ( ) + ( ) + ( ) = + + =
+
=

n
n
, or

2 2 2 2 1 1 1 1 1 1 1 1 0 0 0 1 1
1
1
7
( ) = ( ) + ( ) + ( ) + = + + + =
+
=

n
n
.
b) It is very tempting to use the same strategy of ‘pairing up’ consecutive
terms in this manner
2 2 2 2 1 1 1 1 1 1 1
1
1
( ) = ( ) + ( ) + ( ) +
+
=

n
n
. . .
to argue that the sum of this infnite series is 0. However, this is
erroneous. Consider that if we leave out the frst term and start ‘pairing
up’ from the second term we will obtain a different sum. The associative
property of addition is what allowed us to ‘pair up’ the numbers for the
fnite sum in part a). Although the associative property works for fnite
sums it is clear that it does not work for infnite sums. The sum of an
infnite series is defned to be the limit of the sequence of partial sums.
For the sequence s
n
i
i
n
{ } = ( )

+
=

21
1
1
, we have s s s s
1 2 3 4
1 0 1 0 = = = = , , , ,
etc. Clearly this sequence is not converging to a limit. Therefore, the
series has no sum and it diverges.
In studying infnite series, there are commonly two basic questions: Does a
particular series converge or does it diverge? If it does converge, what is its
sum?
Geometric series
There is one type of infnite series with which we are already familiar –
and for which we know how to answer questions regarding convergence/
divergence and computing sums; and this is infnite geometric series that
we encountered in Chapter 4 of the book.
If a
1
represents the frst term and r is the number that multiplies a term to
obtain the next term in the series, then an infnite geometric series can be
generalized as follows:
a a r a r a r a r a r
n n
n
1 1 1
2
1
3
1
1
1
1
1
+ + + + + + =
=

. . . . . .
2 2
, a
1
0 ≠
Let’s consider three cases: r r r = = ≠ ± 1 1 1 , , 2 and .
25
If r = 1, then the nth partial sum is s a a a a na
n
= + + + + =
1 1 1 1 1
. . . . Clearly
the sequence of partial sums, s
n
{ }, will increase without bound and the
geometric series diverges in this case.
If r = 21, then the nth partial sum is s a a a a
n
= + …
1 1 1 1
2 2 . The
sequence of partial sums, s
n
{ }, will behave in the same way as in Example
1 b) with s a s s a s
1 1 2 3 1 4
0 0 = = = = , , , . The sequence of partial sums is not
converging to a limit, so the geometric series also diverges for this case.
If r ≠ ±1, then s a a r a r a r a r
n
n
= + + + + +
1 1 1
2
1
3
1
1
. . . .
2
Multiplying through by r gives
rs a r a r a r a r
n
n n
= + + + +

1 1
2
1
1
1
. . . .
Subtracting the second equation from the frst produces
s rs a a r
n n
n
2 2 =
1 1
.
Factorizing yields s r a r
n
n
1 1
1
2 2 ( ) =
( )
.
Thus, the nth partial sum is
s
a r
r
n
n
=
( )
1
1
1
2
2
.
We know from the theorem for convergence of geometric sequences in
the previous section that if r < 1 then r
n
converges to 0 as n → ∞. We can
apply this fact and some properties of limits to give the following result:
lim lim lim
n
n
n
n
n
a r
r
a a r
r
a
r
→∞ →∞ →∞
( )
= =
1
1 1 1
1
1 1 1
2
2
2
2 2
2llim lim
n
n
n
n
a r
r
a
r
a
r
r
a
r
→∞ →∞
=

=
1 1 1 1
1 1 1 1 2 2
2
2 2
200
1
1
=
a
r 2
Therefore, if r < 1 then lim
n
n
s
a
r
→∞
=
1
12
.
This rigorously confrms a result that appeared in Chapter 4, and we state
it again here. In this chapter, when we refer to a geometric series, it can be
assumed that it is an infnite geometric series.
identifying the value of the common ratio, r, we can determine which
geometric series converge and which ones diverge; and for ones that
converge we can easily compute the sum with the formula S
a
r

=
1
12
.
For any geometric series, the interval r < 1, which can also be written as
21 1 < < r , is known as its interval of convergence.
Convergence of geometric series
The geometric series with common ratio r
a a r a r a r a r ar
n n
n
1 1 1
2
1
3
1
1 1
1
+ + + + + + =
=

∑  
2 2
converges to the sum
a
r
1
12
if r < 1, and diverges if r 1.
26
Series and Convergence
2
Partial fraction decomposition
The process by which a sum of two or more fractions is combined into
a single fraction by obtaining a common denominator is a basic skill of
elementary algebra. For example,

2
1
3
1
2
1
1
1
3
1
1
1
2 2
1
3 3
2
x x x
x
x x
x
x
x
x
x
+
+ =
+
⋅ + ⋅
+
+
= +
+
2
2
2 2
2
2 xx
x
x
2 2
1
5 1
1 2 2
=
+
.
The reverse process, in which a fraction with a factorable denominator is
split into the sum of two or more fractions (often called partial fractions)
whose denominators are the factors of the original denominator, is a very
useful tool. For the content of this course, the technique of decomposing
a fraction into partial fractions will be applied in two different situations:
(1) integrating a rational function, and (2) fnding the sum of a particular
type of series. Also, for this course, we will only be expected to apply this
technique to fractions that have denominators that factorize into linear
factors that do not repeat.
Example 2 – Partial fractions and integration
Evaluate
1
14 48
2
x x
dx
2 +

.
Solution
Attempting to decompose the integrand into partial fractions, we start by
factorizing the denominator,

1
14 48
1
8 6
2
x x x x 2 2 2 +
=
( )( )
, Note that the factors are linear and none
are repeated.
and then assume that there are constants A and B such that

1
8 6 8 6 x x
A
x
B
x 2 2 2 2 ( )( )
= + .
Multiplying both sides by x x 2 2 8 6 ( )( ) we obtain
1 6 8 = ( ) + ( ) A x B x 2 2 .
This equation is an identity and, thus, we wish to fnd the constants A and
B so that the equation will be true for all x. Therefore, we can substitute
any value in for x. Obvious choices are to substitute x = 6 and x = 8.
For x = 6: 1 0 2
1
2
= ( ) + ( ) ⇒ = A B B 2 2
For x = 8: 1 2 0
1
2
= ( ) + ( ) ⇒ = A B A
Thus,
1
14 48 8 6
1
2
1
8
1
2
2
1
2
1
2
x x
dx
x x
dx
x
dx
2 2
2
2 2
2
+
=

=
∫ ∫ ∫
11
6 x
dx
2

= +
1
2
8
1
2
6 ln ln x x C 2 2 2
= +
1
2
8
6
ln .
x
x
C
2
2
An obvious way to check
whether you correctly
decomposed a fraction is to
combine the partial fractions in
your result by means of getting
a common denominator,
and see if they add up to the
original fraction.
27
Example 3 – Partial fractions and series
a) Express
3
3 1 3 2 x x 2 ( ) + ( )
in partial fractions.
b) Find the sum of the series

3
2 5
3
5 8
3
8 11
3
3 1 3 2
3
3 1 3 2 ×
+
×
+
×
+…+
( ) + ( )
+…=
( ) + ( ) n n n n 2 2
nn=

1
.
Solution
a)
3
3 1 3 2 3 1 3 2 x x
A
x
B
x 2 2 ( ) + ( )
=
( )
+
+ ( )
⇒ = + ( ) + ( ) 3 3 2 3 1 A x B x 2
Let x =
1
3
: 3 3 0 1 = ( ) + ( ) ⇒ = A B A
Let x = 2
2
3
: 3 0 3 1 = ( ) + ( ) ⇒ = A B B 2 2
Therefore,
3
3 1 3 2
1
3 1
1
3 2 x x x x 2 2 ( ) + ( )
= −
+
.
b) Using the result from a) we can write

3
3 1 3 2
1
3 1
1
3 2
1 1
n n n n
n n
2 2 ( ) + ( )
= −
+

=

=

∑ ∑
.
This series telescopes as demonstrated in the following partial sums.

s
1
1
2
1
5
3
10
= = 2

s
2
1
2
1
5
1
5
1
8
1
2
1
8
3
8
=

+

= = 2 2 2

s
3
1
2
1
5
1
5
1
8
1
8
1
11
1
2
1
11
=

+

+

= = 2 2 2 2
99
22
And, in general:
s
n
n
=

+

+

+ …+
1
2
1
5
1
5
1
8
1
8
1
11
1
3 1
2 2 2
2
22
1
3 2 n +

s
n
n
=
+
1
2
1
3 2
2
Hence,
lim lim .
n
n
n
s
n
→∞ →∞
=
+

= =
1
2
1
3 2
1
2
0
1
2
2 2
Therefore, the series is convergent and
3
3 1 3 2
1
2
1
n n
n
2 ( ) + ( )
=
=

.
It is essential to understand that for any series a
n

there are two
important sequences for us to consider: the sequence s
n
{ } of its partial
sums and the sequence a
n
{ } of its terms.
Note in Example 3 that the
terms cancel in pairs. A series
that behaves in this manner
is called a telescoping series
because of these cancellations.
The sum collapses (in the way
that an old-style telescope
collapses) into just two terms.
It is also important to point
out that this ‘telescoping’
behaviour does not always
occur for a series whose rule
can be decomposed into
partial fractions. In fact, it
is a rare occurrence – and
only happened in Example 3
because the denominators
3n 2 1 and 3n + 2 are
equivalent when the value of
n in 3n 2 1 is one greater than
in 3n + 2.
28
Series and Convergence
2
Example 4
For each of the series, a
n
n=

1
, below
(i) write the frst four terms and fnd the limit (if it exists) of the sequence
of its terms, lim
n
n
a
→∞
; and
(ii) write the frst four terms of the sequence of its partial sums s
n
{ } and
fnd its limit (if it exists), i.e. the sum of the series.
a)
21
3
1
( )
=

n
n
n

b)
1
1
1
n n
n
+ ( )
=

c) 2 5
2 1
1
n n
n
2
=

Solution
a) (i)
2
2 2 2
1
3
1
3
1
9
1
27
1
81
1
( )
= + + …
=

n
n
n
The sequence of terms in the series is a
n
n
n
{ } =
( )

21
3
. This is a
geometric sequence with r = 2
1
3
and because 2 2 1
1
3
1 < < then it
follows that lim lim
n
n
n
n
n
a
→∞ →∞
=
( )
=
21
3
0.
(ii) The sequence of partial sums begins as follows:
s
1
1
3
= 2
s
2
1
3
1
9
2
9
= + = 2 2
s
3
1
3
1
9
1
27
7
27
= + = 2 2 2
s
4
1
3
1
9
1
27
1
81
20
81
= + + = 2 2 2
Because the series is geometric such that 21 1 < < r , then the
series converges to
lim .
n
n
s
a
r
→∞
= =

= =
1
1
1
3
1
1
3
1
3
4
3
1
4 2
2
2 2
2
2
Therefore, the sum of the series is 2
1
4
.
b) (i)
1
1
1
2
1
6
1
12
1
20
1
n n
n
+ ( )
= + + + +
=

. . .
This series is not geometric. From the limits of rational functions
in the previous section, we know that lim
x
x x
→∞
+
=
1
0
2
. Therefore,
for this sequence we have lim lim
n
n
n
a
n n
→∞ →∞
=
+ ( )
=
1
1
0.
29
(ii) To fnd the sum of the series
1
1
1
n n
n
+ ( )
=

we can use the technique
of partial fractions to write
1
1
1 1
1
1 1
n n n n
n n
+ ( )
=
+

=

=

∑ ∑
2 . Hence,
this is a telescoping series and the frst four partial sums are:
s
1
1
1
1
2
1
2
= = 2
s
2
1
1
1
2
1
2
1
3
1
1
3
2
3
=

+

= = 2 2 2
s
3
1
1
1
2
1
2
1
3
1
3
1
4
1
1
4
3
4
=

+

+

= = 2 2 2 2
s
4
1
1
1
2
1
2
1
3
1
3
1
4
1
4
1
5
=

+

+

+

2 2 2 2


= = 1
1
5
4
5
2
And, in general:
s
n
=

+

+

+

1
1
1
2
1
2
1
3
1
3
1
4
1
4
1
5
2 2 2 2


+ …+
+

1 1
1 n n
2

s
n
n
=
+
1
1
1
2
Hence,
lim lim
n
n
n
s
n
→∞ →∞
=
+

= = 1
1
1
1 0 1 2 2 .
Therefore, the series is convergent and

1
1
1
1
n n
n
+ ( )
=
=

.
c) (i) 2 5 4
16
5
64
25
256
125
2 1
1
n n
n
2
=

= + + + +…
The series appears to be geometric with r =
4
5
. We can confrm this
by simplifying the rule for the nth term:
2 5
2
5
4 4
5
4
4
5
2 1
2
1
1 1
1
1
n n
n
n
n
n
n
2
2
2
2
2
=
( )
= =

.
Hence, 2 5 4
4
5
2 1
1
1
1
n n
n
n
n
Ϫ
Ϫ
=

=

∑ ∑
=

and its clear that the series is
geometric with a
1
4 = and r =
4
5
. Because 21 1 < < r , then
lim lim lim
n
n
n
n
n
a
→∞ →∞ →∞
=

=

4
4
5
4
4
5
1 2


= ⋅ =
n21
4 0 0.
(ii) The sequence of partial sums begins as follows:
s
1
4 =
s
2
4
16
5
36
5
7 2 = + = = .
30
Series and Convergence
2
s
3
4
16
5
64
25
244
25
9 76 = + + = = .
s
4
4
16
5
64
25
256
125
1476
125
11 808 = + + + = = .
Because the series is geometric such that 21 1 < < r , then the
series converges to
lim .
n
n
s
a
r
→∞
= =

= =
1
1
4
1
4
5
4
1
5
20
2
2
Therefore, the sum of the series is 20.
It is obvious that any series whose sequence of terms does not converge to
zero, i.e. lim
n
n
a
→∞
≠ 0, will have a sequence of partial sums that diverges. In
such a case, the magnitude (positive or negative) of terms will increase,
causing the sequence of partial sums to increase without bound. We
established that all three series in Example 4 are convergent and also
that lim
n
n
a
→∞
= 0 for all three series. It seems reasonable to conjecture
that a necessary and suffcient condition for an infnite series
a a a a
n 1 2 3
+ + +…+ +… to converge to a fnite quantity is that the
sequence, a
n
{ }, of individual terms a
n
converges to zero. Is it possible for
the sequence of terms of a series to converge to zero but the series itself
does not converge, i.e. does not have a sum?
Example 5
Consider the series
1
1
1
2
1
3
1
4
1
n
n=

= + + + +…. Determine whether the
series converges or diverges.
Solution
Clearly, the sequence of terms converges to zero, i.e. lim
n
n
→∞
=
1
the question about convergence of the series we need to look at the
sequence of partial sums. Our analysis begins by bracketing the terms in
the following way:
s
n
= + + +

+ + …+

+ + …+

1
1
2
1
3
1
4
1
5
1
8
1
9
1
16

+ …+
+
+ …+

+ …
1
2 1
1
2
1 n n 2
so that the fnal term in each bracketed group is the reciprocal of a power
of two. Let’s consider the sum of the frst 2
n
terms,
s
n
2
1
1
2
1
3
1
4
1
5
1
6
1
7
1
8
1
9
1
10
= + + +

+ + + +

+ + …+
11
16
1
2 1
1
2
1

+ …+
+
+ …+

 n n 2

1
1
2
1
4
1
4
1
8
1
8
1
8
1
8
1
16
1
16
1
16
+ + +

+ + + +

+ + + ++ + + + +

+ …+
+
+ …+

1
16
1
16
1
16
1
16
1
16
1
2 1
1
2
1 n n


= + + + +…+ 1
1
2
1
2
1
2
1
2
= +

=
+
1
1
2
2
2
n
n

+
s
n
n
2
2
2

31
Clearly the sequence of these partial sums diverges, so s
n
2
diverges.
Hence, the series s
n
n
n
=
=

1
1
is greater than a series that diverges, so it must
also diverge.
Therefore, even though the sequence
1
0
n
n → → ∞ as , the series
1
1
n
n=

diverges. This series is called the harmonic series – and we will encounter
it often.
nth term divergence test
If lim
n
n
a
→∞
does not exist, or if lim
n
n
a
→∞
≠ 0, then the series a
n
n=

1
diverges.
The clever method used in Example 5 is attributed to a French scholar, Nicole Oresme
(1323–1382), who was the frst to mathematically prove that the harmonic series
diverges. Considering the state of mathematics in the 14th century, Oresme was well
ahead of his time by inventing a type of coordinate geometry and using the idea
of a fractional exponent – three centuries before Descartes developed coordinate
geometry and Newton frst invented our modern notation for fractional exponents.
With regard to his proof of the divergence of the harmonic series, Oresme’s ingenious
strategy involved replacing groups of fractions in the harmonic series with smaller
fractions that have a sum of
1

_

2
. The following shows the heart of his strategy:

1
1
2
1
2
1
2
2
2
+ > + =

1
1
2
1
3
1
4
1
1
4
1
4
3
2
+ + +

> + +

=
1
1
2
1
3
1
4
1
5
1
6
1
7
1
8
3
2
1
8
1
8
1
8
1
8
+ + + + + + +

> + + + +



=
4
2

1
1
2
1
8
1
9
1
10
1
16
4
2
1
16
1
16
+ + … + + + + … +

> + + … +



=
5
2
This process can be continued indefnitely, so that, in general, for any positive integer n
we have

1
1
2
1
3
1
2
1
2
+ + + … + >
+
.
n
n
For example, if n = 25 then

1
1
2
1
3
1
33554 432
1
1
2
1
3
1
2
25 1
2
25
+ + + + = + + + + >
+
= . . . . . . 113.
So Oresme’s strategy shows that by taking enough terms of the harmonic series,
we can guarantee that its sum will be greater than any fnite number. Therefore, the
series will diverge to infnity. It is interesting to note that although the harmonic series
diverges, it does so very slowly. The sum of the harmonic series does not get above 10
until we have added 12 367 terms of the series!
The fact that the harmonic series diverges (Example 5) serves as a
counterexample to our conjecture that lim
n
n
a
→∞
= 0 is both a necessary
and suffcient condition for the series a
n
n=

1
to converge. It is true that
convergence can only occur if lim
n
n
a
→∞
= 0 (i.e. a necessary condition), but
lim
n
n
a
→∞
= 0 is NOT suffcient to guarantee convergence (i.e. not a sufcient
condition). This leads to the following theorem.
32
Series and Convergence
2
Example 6 – Using the nth term divergence test
Determine, if possible, whether each of the following series converges or
diverges.
a)
n
n
n
2
2
1
1
1
2
4
5
9
10
16
17 +
= + + + +…
=

b) 3 1 3 3 3 3
1
1
2 2 2 ( ) = + +…
+
=

n
n
c)
2
3 1
1
2
1
5
1
14
1
41
1
n
n
+
= + + + +…
=

d)
n
n
n
!
! 3 1
1
4
2
7
6
19
24
73
1
+
= + + + +…
=

e)
n
n
n
2
1
1
1
2
2
5
3
10
4
17 +
= + + + +…
=

Solution
a) lim lim lim lim
n
n
n n n
a
n
n
n n
n n n
→∞ →∞ →∞ →∞
=
+
=
+
=
2
2
2 2
2 2 2
1 1
11
1 1
1
2
+
=
n
Therefore, by the nth term divergence test, the series is divergent.
b) lim lim
n
n
n
n
a
→∞ →∞
+
= ( )
( )
3 1
1
2 does not exist because the terms alternate
between +3 3 and 2 .
Therefore, by the nth term divergence test, the series is divergent.
c) Certainly, 3 1
n
+ → ∞ as n → ∞, so it follows that
1
3 1
0
n
+
→ as
n → ∞. Hence, lim lim lim
n
n
n
n
n
n
a
→∞ →∞ →∞
=
+
=
+
=
2
3 1
2
1
3 1
0. Since the limit
of the nth term is 0, the nth term divergence test does not apply and we
are not able to make a conclusion about convergence or divergence. We
can make an educated guess that it will probably converge because it is
very similar to the convergent geometric series
2
3
2
1
3
1 1
n
n
n
n =

=

∑ ∑
=

with
r =
1
3
. In the next section we will learn that it does in fact converge
and recognizing that it is similar to a convergent geometric series is
important.
d) lim lim
!
!
lim
! !
! ! !
n
n
n n
a
n
n
n n
n n n
→∞ →∞ →∞
=
+
=
+
=
+ 3 1 3 1
1
3 0
==
1
3
Therefore, by the nth term divergence test, the series is divergent.
e) lim lim lim
n
n
n n
a
n
n
n n
n n n
→∞ →∞ →∞
=
+
=
+
=
+
=
2
2
2 2 2
1 1
0
1 0
0
We cannot apply the nth term divergence test since the limit of the
nth term is 0. We will fnd in the next section that this series behaves
like the harmonic series, that is, even though the sequence of its terms
converges to 0 the series itself diverges.
33
Before moving onto the next section and investigating more thoroughly the
convergence of infnite series, we state below some important properties of
convergent series that are direct consequences of the properties of limits in
Section 1.2 of this publication.
Properties of convergent series
Given that a
n
n=

1
and b
n
n=

1
are both convergent series, and c is a constant, then the
following series are also convergent:
ca
n
n=

1
, a b
n n
n
+ ( )
=

1
and a b
n n
n
2 ( )
=

1
.
(i) ca c a
n
n
n
n =

=

∑ ∑
=
1 1
(ii) a b a b
n n
n
n
n
n
n
+ ( ) = +
=

=

=

∑ ∑ ∑
1 1 1
(iii) a b a b
n n
n
n
n
n
n
2 2 ( ) =
=

=

=

∑ ∑ ∑
1 1 1
2.2 Convergence of infnite series
In this section, we develop some more sophisticated tests for convergence.
These tests will allow us to effciently determine convergence for a wide
range of series. In Example 6 we were thwarted from determining whether
the series in parts c) and e) were convergent or divergent. In general, it
is not easy to fnd the exact sum of a series. We have been able to fnd
exact sums for certain geometric series and telescoping series because we
were able to obtain a formula for the sequence of partial sums, s
n
. In this
section, our purpose is to develop some tests that will let us determine
whether a series is convergent or divergent without the need for a formula
for the sequence of partial sums. Although in some cases the convergence
test being employed will help us to fnd the sum of a series (or at least an
approximation for the sum), in general, it is limited to fnding out about
convergence of a series without fnding the sum. We will study four useful
convergence tests that apply to series whose terms are non-negative and a
ffth test that will apply to alternating series.
Integral test
From our discussion about improper integrals in the previous section,
you may feel that there is a relationship between the convergence of an
improper integral and the convergence of a series. We can take the
formula for the nth term a
n
of a series a
n
n=

1
and replace n by x to write a
function f x ( ). The relationship between a
n
n=

1
and the improper integral
f x dx ( )

1
is explained in the following theorem.
34
Series and Convergence
2
For the harmonic series
1
1
n
n=

, the sequence of its terms
1
1
1
2
1
3
1
4
1
n n

= … … , , , , , , any number greater than or equal to 1 is an
upper bound, and any number that is less than or equal to zero is a lower
bound. For the sequence
1
n

we can call 1 the least upper bound and
0 the greatest lower bound. Another characteristic of the sequence
1
n

is that the terms are always decreasing and it is not surprising that the
sequence converges to its greatest lower bound. In our discussion of one-
to-one functions in Chapter 2 of the book, we used the word monotonic
to describe a function that is either always increasing or always decreasing.
Also for the harmonic series, we established that the sequence of its partial
sums, s
n
{ }, is divergent by essentially showing that s
n
{ } does not have an
upper bound, and hence is not bounded. It is sensible to conjecture that a
bounded monotonic sequence will be convergent.
Before we can conduct a formal proof of the integral test we need to
establish the defnition of two words for which we have had a common-
sense understanding up to now, and to state an important theorem.
Lower and upper bounds of a sequence
The number M is a lower bound of the sequence a
n
{ } if a M
n
 for all positive integers
n, and the number N is an upper bound of a
n
{ } if a N
n
 for all positive integers n. A
sequence a
n
{ } is bounded if and only if it has a lower bound and an upper bound.
Bounded sequence theorem
A monotonic sequence converges if and only if it is bounded.
Completeness postulate
In the real numbers, every non-empty set that has an upper bound has a least upper
bound.
The integral test for convergence
Let f be a function that is continuous, decreasing and positive for all x 1 and
a f n
n
= ( ), then the series a
n
n=

1
converges if and only if the improper integral
f d x x ( )

1
converges. In other words:
1) If f d x x ( )

1
converges, then a
n
n=

1
also converges.
2) If f d x x ( )

1
diverges, then a
n
n=

1
also diverges.
Before we conduct a formal proof of this theorem, we state an important
property of the real numbers with the following postulate.
35
Proof of the bounded sequence theorem
We prove the theorem for the case when the monotonic sequence, call
it a
n
{ }, is increasing. If it converges to some limit L then it is bounded
below by the frst term of the sequence a
1
and above by L and is therefore
bounded. Conversely, if a
n
{ } is bounded, then the completeness postulate
guarantees that a
n
{ } has a least upper bound L. We now need to show
that a
n
{ } must converge to L. Firstly, since L is an upper bound for a
n
{ }
then it follows that a L
n
 for all n. Also, since L is the least upper bound
then L 2ε is not an upper bound for any ε > 0. Hence, there exists an
integer N such that L a
N
2ε < . Because a
n
{ } is always increasing then
a a
N n
 whenever n N > . Therefore, L a L
n
Ϫ ഛ ε < and consequently
L a L
n
2ε ε < < + which is equivalent to 2 2 ε ε < < a L
n
and a L
n
2 < ε.
This satisfes the ε 2N definition for the limit of a sequence and completes
the proof for an increasing sequence a
n
{ }. A parallel argument can be
written to prove the theorem for a decreasing sequence a
n
{ }.
Proof of the integral test
The essential idea behind the proof is that the terms in a series a
n
n=

1
can
be assigned to represent the area of ever decreasing rectangles of constant
width and that the improper integral f x dx ( )

1
is approximated by the
sum of these rectangles. The total areas of the inscribed rectangles (Figure
2.1) and the circumscribed rectangles (Figure 2.2) are as follows:
a a a a a
i
i
n
n
=

= + + +…+
2
2 3 4
(Inscribed area, or lower sum.)
a a a a a
i
i
n
n
=

= + + +…+
1
1
1 2 3 1
2
2
(Circumscribed area, or upper sum.)
0
a
2
a
3
a
4
a
5
a
6
y
x
f
0
a
1
f
a
2
a
3
a
4
a
5
y
x
a
6
Figure 2.2 Circumscribed rectangles gives upper sum. Figure 2.1 Inscribed rectangles gives lower sum.

a f x dx
n
i
n
n
=

( )
2
1

f d a
n
n
i
n
x x ( )

=
1
1
1

Ϫ
The exact area under the graph of f from x = 1 to n, i.e. the defnite
integral f x dx
n
( )

1
, lies between the inscribed and circumscribed areas.
36
Series and Convergence
2
As Figures 2.1 and 2.2 illustrate,
a f x dx a
i
i
n
n
i
i
n
= =

( )
2
1
1
1
ഛ ഛ
Ϫ
.
Using the nth partial sum, s a a a a
n n
= + + +…+
1 2 3
, we can write the
inequality above as
s a f x dx s
n
n
n
Ϫ ഛ ഛ
Ϫ 1
1
1
( )

.
To prove part (1) we start by assuming f x dx
n
( )

1
converges to L. Then it
follows that for n 1
s a L
n
Ϫ ഛ
1
and consequently
s L a
n
 +
1
.
Hence, the sequence of partial sums s
n
{ } is bounded and monotonic and it
follows from the bounded sequence theorem that s
n
{ } converges, and
consequently the series a
n
n=

1
must also converge. For part (2) assume
that the improper integral f x dx
n
( )

1
diverges. Thus, f x dx
n
( )

1
goes
to infnity as n → ∞, and given the inequality s f x dx
n
n
Ϫ

1
1
( )

it must
n
{ } diverges which means that a
n
n=

1
also diverges.
Example 7 – Using the integral test
Determine the convergence or divergence of each series.
a)
1
2 n n n ln =

b)
n
e
n
n=

1
c)
1
2
1
5
1
10
1
17
1
26
+ + + + +…
d)
n
n
n
2
1
1 +
=

[Example 6 part e)]
Solution
Plot1
Y1= 1/(X (1n(X))
Plot2 Plot3
Y2=
Y3=
Y4=
Y5=
Y6=
a) Graphing the function f x
x x
( ) =
1
ln
on our GDC provides us with a
quick confrmation that f is continuous, decreasing and positive for all
x 2, , thereby satisfying the conditions for applying the integral test.
Recalling techniques for improper integrals from the frst section of this
37
chapter, we now need to evaluate
1
2
x x
dx
ln

to see if it converges to
a fnite number or diverges to infnity. For this integral we will also need to
apply the technique of u-substitution.

1 1
2
1
2
1
2
1
2
x x
dx x
x
dx u du u
ln
ln
∫ ∫ ∫
= ( )

= =
2
2
1 1
2 2
x x
dx
x x
dx
b
b
ln
lim
ln

→∞
∫ ∫
= Rewriting improper integral as a limit.
=

→∞
lim ln
b
b
x 2
2
Applying result from u-substitution.
=

→∞
lim ln ln
b
b 2 2 2 2
= ∞
Therefore, the integral
1
2
x x
dx
ln

diverges, and by the integral test
the series
1
2 n n n ln =

must also diverge.
b) For f x
x
e
x
( ) = , it is clear that f is continuous, decreasing and positive
for x 1 because e
x
> 0 and e
x
grows faster than x; so the integral test
applies. Using integration by parts:

x
e
dx xe dx
x
x
=
∫ ∫
2
Choose u du d = ⇒ = x x and dv e v e = ⇒ =
2 2
2
x x
.
=

2 2 2
2 2
xe e dx
x x
Substituting into formula udv uv v du
∫ ∫
= 2 .
= +

2
2 2
xe e dx
x x
= 2 2
2 2
xe e
x x

x
e
dx xe e
x
b
x x
b
1
1

→∞

= 

lim 2 2
2 2
Rewriting improper integral as a limit.
=
( ) ( )

→∞
lim
b
b b
be e e e 2 2 2 2 2
2 2 2 2 1 1
=
+
+

=
+

+
→∞ →∞ →
lim lim lim
b
b
b
b
b
b
e e
b
e
2 2
1 2 1
∞∞

2
e
Applying l’Hôpital’s rule to the frst limit gives lim lim .
b
b
b
b
b
e e
→∞ →∞
+

=

=
1 1
0
Therefore,
x
e
dx
e
x
1
2

= .
By the integral test, since the integral
x
e
dx
x
1

converges then the series
n
e
n
n=

1
must also converge.
c) We need to fnd a rule for the nth term for the series that starts

1
2
1
5
1
10
1
17
1
26
+ + + + +….
Using some inductive reasoning we determine that the series expressed
in summation notation is
1
1
1
2
1
5
1
10
1
17
1
26
1
1
2
1
2
n n
n
+
= + + + + +…+
+
+…
=

.
The function f x
x
( ) =
+
1
1
2
satisfes the conditions of the integral test.
Let u du d = = ln , x
x
x. then
1
As Example 7 a) illustrates, if
the summation index for an
infnite series starts at n k = > 1
rather than n = 1, the integral
test can still be applied. The
integral test can be modifed
as follows:
Let f be a function that is
continuous, decreasing and
positive for all x k such that
k > 1 and a f n
n
= ( ), then the
series a
n
n=

1
converges if and
only if the improper integral
f d
k
x x ( )

converges.
38
Series and Convergence
2
We need to recognize that the anti-derivative of
1
1
2
x +
is arctanx
(a ‘standard integral’ in the IB formula booklet).

1
1
1
1
2
1
2
1
x
dx
x
dx
b
b
+
=
+

→∞
∫ ∫
lim Rewriting improper integral as a limit.
=
[ ]
→∞
lim arctan
b
b
x
1
=
[ ]
→∞
lim arctan arctan
b
b 2 1
=
[ ]

→∞ →∞
lim arctan lim
b b
b 2
π
4
= −
π π
2 4
Therefore,
1
1 4
2
1
x
dx
+
=

π
.
By the integral test, since the integral
1
1
2
1
x
dx
+

converges then the
series
1
1
2
1
n
n
+
=

must also converge.
d)
n
n
n
2
1
1 +
=

was the series in Example 6 e) for which the nth term
divergence test was inconclusive because lim
n
n
n
→∞
+
=
2
1
0. The function
f x
x
x
( ) =
+
2
1
satisfes the conditions of the integral test. The method
of u-substitution will be useful to evaluate the integral
x
x
dx
2
1 +

.
Let u x = +
2
1 and it follows that du xdx du xdx = ⇒ = 2
1
2
.
Substituting gives

x
x
dx
u
du
u
du u x
2
2
1
1 1
2
1
2
1 1
2
1
2
1
+
= ⋅ = = = +
( )
∫ ∫ ∫
ln ln .
Using this result we have:

1
1
1
1
2
1
2
1
x
dx
x
dx
b
b
+
=
+

→∞
∫ ∫
lim Rewriting improper integral as a limit.
= +
( )

→∞
1
2
1
2
1
lim ln
b
b
x
= +
( )

→∞
1
2
1 2
2
lim ln ln
b
b 2
= ∞
By the integral test, since the integral
x
x
dx
2
1
1 +

diverges then the
series
n
n
n
2
1
1 +
=

must also diverge.
Estimating the sum of a series
As pointed out, if the integral used in the integral test converges, we know
that the associated series converges to a sum, but we are not any closer to
knowing the value of the sum. Recall that the sum of a series, s, is defned
It is very important to know
when using the integral
test that the value of the
improper integral is not
equal to the sum of the
series. The sum, expressed to
ten signifcant fgures, of the
frst 50 terms of the series
1
1
2
1
n
n
+
=

∑ , Example 7 d), is
approximately 1.056 875 301;
whereas
1
1 4
0 7853981634
2
1
x
x
+
= ≈

d
π
. .
Therefore, in general
a f d
n
n=

≠ ( )
1
1
x x.
39
to be the limit of the partial sums, i.e. lim
n
n
s s
→∞
= . For any specifc partial
sum, n will be some integer. What is the difference, or error, between the
partial sum s
n
and the exact sum s? Clearly this will be the sum of the
terms ‘beyond’ n (i.e. a a
n n + +
+ +
1 2 
). So when the partial sum s
n
is used
as an approximation for s, the error R
n
(sometimes called the remainder) is
given by
R s s a a a
n n n n n
= = + + +…
+ + +
2
1 2 3
.
We use some of the same ideas and notation that we used in our proof of
the integral test.
Assuming that the function f is continuous, decreasing and positive for all
x n  , we can see from Figure 2.3 that the area under f for x n  is an
upper bound for R
n
. That is, R a a a f x dx
n n n n
n
= + + + ( )
+ + +

∫ 1 2 3 
 .
Similarly, we can see from Figure 2.4 that the area under f for x n  +1 is a
lower bound for R
n
.
R a a a f x dx
n n n n
n
= + + + ( )
+ + +
+

∫ 1 2 3
1

a
n
a
n + 1
a
n + 2
a
n + 3
0
f
y
x n
0
f
y
x n
a
n +1
a
n + 3
a
n + 2
n + 1
0
f
y
x
n +1
0
f
y
x
a
n
a
n + 1
a
n + 2
a
n + 3
0
f
y
x n
0
f
y
x n
a
n +1
a
n + 3
a
n + 2
n + 1
0
f
y
x
n +1
0
f
y
x
Figure 2.4
Figure 2.3
40
Series and Convergence
2
Hence,

f x dx R f x dx
n
n
n
( ) ( )
+
∞ ∞
∫ ∫
1
  .
Since s s R
n n
n
to each side of the statement above gives
s f x dx s s f x dx
n
n
n
n
+ ( ) + ( )
+
∞ ∞
∫ ∫
1
  .
These lower and upper bounds can provide a very accurate approximation
for the sum of the series.
Example 8 – Two ways of estimating the sum of a series
Estimate the sum, s, of the series
1
1
1
4
1
9
1
16
1
2
1
2
n n
n=

= + + + + + +
 
by:
a) adding the frst 25 terms, and b) using the inequalities
s f x dx s s f x dx
n
n
n
n
+ ( ) + ( )
+
∞ ∞
∫ ∫
1
  with n = 10.
Solution
a) Using a GDC, s
25
is computed to be approximately 1.605 723 404 to ten
signifcant fgures.
b) We can compute the lower and upper bounds of our estimate from the
inequalities

s
x
dx s s
x
dx
10 2
11
10 2
10
1 1
+ +
∞ ∞
∫ ∫
  .
We need to fnd
1
2
x
dx
n

:
1 1 1 1 1
2
x
dx
x b n
n b
n
b
b

→∞ →∞

,
¸
,
]
]
]
+
j
(
,
\
,
(
lim lim 2 2
nn
Substituting gives: s s s
10 10
1
11
1
10
+ +  
Using s
10
1 549767731 ≈ . : 1 640676822 1 649767731 . .   s
If we estimate s using the midpoint of this interval, then the error for
our approximation can be no greater than half of the length of the
interval, which is approximately 0.004 55.
Therefore,
1
1 645222277
2
1
n
n

=

. with error < 0.004 55.
p-series
Before we move onto the next convergence test, we can use the integral test
to give us important results for any series that is in the form shown below,
known as a p-series.

1 1
1
1
2
1
3
1
1
n n
p p p p p

= + + + + +
 
where p is a constant.
Estimating sum of series for the integral test
Given that f n a
n
( ) = and f is continuous, positive and decreasing for x   n 1 and
a
n
n=

1
is convergent then s f d s s f d
n
n
n
n
+ ( ) + ( )
+
∞ ∞
∫ ∫
x x x x
1
  .
41
If p = 1, the p-series is the harmonic series which we know diverges. What
about series for other values of p? The following example will lead to a
simple test for the convergence of any p-series.
Example 9 – Convergence of p-series
For what values of p is the series
1
1
n
p

convergent?
Solution
Let’s consider when p p p < = > 0 0 0 , , and .
When p < 0, then
1
n
p
→ ∞ as n → ∞. For example, if p = 23 then
1
3
3
n
n
2
= ; and clearly n
3
increases without bound as n → ∞.
When p = 0, then lim
n
p
n n

= =
0
0
1 1
1.
In both of these cases, lim
n
p
n

0
1
0 so the p-series diverges by the nth term
divergence test.
When p > 0, the function f x
x
p
( ) =
1
is continuous, decreasing and
positive for x 1 so we can use the integral test. We know from Example 5
in the previous section that the harmonic series p = ( ) 1 diverges, so let’s
assume that p ≠ 1 and investigate the improper integral
1
1
x
dx
p

.

1
1 1
x
dx x dx
p
b
p
b ∞
→∞
∫ ∫
= lim
2
=
+

→∞
+
lim
b
p
b
x
p
2
2
1
1
1

j
(
,
\
,
(
,
¸
]
]
→∞
+
1
1
1
1
2
2
p
x
b
p
b
lim

j
(
,
\
,
(
,
¸
]
]
→∞
+
1
1
1
1
2
2
2
p
b
b
p
lim
If p > 1, then 2p + < 1 0 and consequently as b → ∞, b
p 2 +

1
0.
Hence, if p > 1 then
1 1
1
1
1
1
1
x
dx
p p
p

=

( ) =
2
2
2
. Therefore the
integral converges and the series
1
1
n
p

must also converge.
If p < 1, then 2p + > 1 0 and consequently as b → ∞, b
p 2 +
→ ∞
1
. Hence,
if p < 1 then the integral
1
1
x
dx
p

diverges and so does the series
1
1
n
p

.
The results from Example 9 are summarized below.
Convergence of p-series
The p-series
1 1
1
1
2
1
3
1
1
n n
p p p p p

= + + + + +
 
(i) converges if p > 1, and (ii) diverges if p < 1.
Note: When p = 1 this is the harmonic series.
42
Series and Convergence
2
Comparison test
The integral test compares a series consisting of all positive terms with
an integral as a means of testing the convergence of the series. It is
possible to use a second series in a similar way. If each term of a series of
positive terms is less than or equal to the corresponding term of a known
convergent series of positive terms, then the series is convergent. We will
call this the comparison test and can state it as follows.
Comparison test
Given 0 < a
n
< b
n
for all n > N for some integer N, it follows that
1 if b
n
n=

1
converges, then a
n
n=

1
also converges;
2 if a
n
n=

1
diverges, then b
n
n=

1
also diverges.
Note: The comparison test can also be applied for the series a
n
n=

1
and b
n
n=

1
whenever
there exists a positive constant c such that 0   a cb
n n
for all n N N  , ∈
+
 .
Positive series convergence
A series of positive terms is convergent if and only if its sequence of partial sums has an
upper bound.
In the statement of the
comparison test, n > N means
from some term onward. That is,
eventually for some term and
forever afterwards the terms of
the series b
n
n=

1
are always
greater than the corresponding
terms of the series a
n
n=

1
. This is
often expressed by saying
that b
n
n=

1
dominates a
n
n=

1
.
The comparison test
signifcantly expands our
ability to determine the
convergence of a series with
more complicated rules for
the nth term. We achieve this
by comparing a ‘complicated’
series to a ‘simpler’ series whose
convergence or divergence is
known.
Before proving both parts of the comparison test, we will fnd it helpful to
state a corollary to the bounded sequence theorem that we recall says the
following: A monotonic (always decreasing or always increasing) sequence
converges if and only if it is bounded. If all the terms of an infnite series
are positive, the sequence of partial sums is increasing. Therefore, the
following theorem follows directly from the bounded sequence theorem.
Proof of comparison test
Proof of 1: Let u v
n n
{ } { } and be sequences of the partial sums for the
series a
n
n=

1
and b
n
n=

1
, respectively. Because b
n
n=

1
is a series of positive
terms that is convergent, it follows from the positive series convergence
theorem that the sequence v
n
{ } has an upper bound – let’s call it B. Since
a b
n n
 for all n 1, we can conclude that a b B
n n
  for all n 1.
Thus, B is an upper bound of the sequence u
n
{ }. Because the terms of the
series a
n
n=

1
are all positive then it follows from the positive series
convergence theorem that a
n
n=

1
is convergent.
Proof of 2: If a
n
n=

1
is divergent, then since u
n
{ } is increasing u
n
→ ∞.
However, b a
n n
 , so v u
n n
 . It follows that v
n
→ ∞ and, therefore, b
n
n=

1

must also diverge.
43
Example 10 – Using the comparison test
Determine the convergence or divergence of each series.
a)
2
3 1
1
n
n
+
=

[Example 6 c)]
b)
1
3 1 + =

n n
c)
1
0
n
n
!
=

Solution
a) We can compare the given series

2
4
2
10
2
28
2
82
2
3 1
+ + + + +
+
+
 
n
with the nth term of the geometric series

2
3
2
9
2
27
2
81
2
3
+ + + + + +
 
n
which converges because its common ratio is between one and negative
one; r = <
1
3
1.
It is clear that each term in the given series is less than its corresponding
term in the geometric series. That is,
2
3 1
2
3
n n
+
< for all n 1.
Therefore, by the comparison test since the series
2
3
1
n
n=

converges the
series
2
3 1
1
n
n
+
=

must also converge.
Note that part 1 and part 2 of the comparison test require that 0   a b
n n
. You can
think of a
n ∑
as the ‘lower’ series and b
n ∑
as the ‘higher’ series (see Figure 2.5). Thus,
in a very informal sense the two parts of the comparison test say:
1. If the ‘higher’ series converges, then the ‘lower’ series must also converge.
2. If the ‘lower’ series diverges, then the ‘higher’ series must also diverge.
The ‘higher’ series dominates the ‘lower’ series.
n
a
n
b
n
Figure 2.5
44
Series and Convergence
2
b) The series
1
3 1 + =

n n
is similar to the p-series
1
1 2
1
n
n=

which diverges
because p =
1
2
1  . If we compare the given series to this p-series we
see that
1
3
1
+
<
n n
for all n 1. However, the comparison test
provides no conclusive result in this case where a series is dominated by
a divergent series. Suspecting that the given series does in fact diverge
we need to fnd a divergent series that the given series dominates. Let’s
compare it to the divergent harmonic series
1
1
n
n=

. Remember, to
satisfy the comparison test it is not necessary for a b
n n
 to be true for
all integers n 1 but for all integers n N  where N is some positive
integer.
Our GDC is a handy tool to quickly compare the terms of the given
series to the harmonic series. The screen images below show values for
the frst 14 terms of the two series in a table.
For the frst fve terms the terms in the harmonic series
1
1
n
n=

are
greater than those for
1
3 1 + =

n n
. However, it appears from the sixth
term onwards that this reverses, that is,

1 1
3
n
n
<
+
for n 6.
Therefore, by the comparison test the series
1
3 1 + =

n n
diverges.
c) Consider the frst few terms of the given series:

1
1 1
1
1 2
1
1 2 3
1
1 2 3 4
0
n
n
!
=

= + +

+
⋅ ⋅
+
⋅ ⋅ ⋅
+

Now consider the frst few terms of the convergent geometric series
with a
1
2 = and r =
1
2
.
2
1
2
2 1
1
2
1
4
1
8
0

= + + + + +
=

n
n

It appears that the terms of
1
0
n
n
!
=

are less than or equal to the
corresponding terms of the convergent geometric series for all n 1.
Recall that in Example 8 of this chapter we proved that lim
!
n
n
x
n
→∞
= 0 for
Y2=1/(3+ (X))
Y1= 1/X
X Y
1
Y
2
X=1
1
2
3
4
5
6
7
1
.5
.33333
.25
.2
.16667
.14286
.25
.22654
.21132
.2
.19098
.1835
.17712
X Y
1
Y
2
X=14
8
9
10
11
12
13
14
.125
.11111
.1
.09091
.08333
.07692
.07143
.17157
.16667
.16228
.15831
.1547
.15139
.14833
TABLE SETUP
TblStart=1
∆Tbl=1
Plot1 Plot2 Plot3
Y3=
Y4=
Y5=
Y6=
Y7=

How could we prove that the
inequality
1 1
3 n n
<
+
is true
for n 6? Try doing so by
proving the inequality
3 + < n n for n 6 by
mathematical induction.
In Example 10 c), we know that
the sum of the infnite
geometric series 2
1
2
0

=

n
n
is
S
a
r

= = =
1
1
2
1
2
1
4
2 2
. Thus the
sum
1
0
n
n
!
=

must be less than
4. In fact, we will learn in the
next section that this sum is
exactly the number e. That is,
e
n
= + + + + +
+ +
1 1
1
2
1
6
1
24
1

!
45
any real number x. From that we concluded that the factorial function
increases faster than any exponential function. Hence,
1
2
1
2 n
n
!

for
n 1. Therefore, by the comparison test the series
1
0
n
n
!
=

converges.
Limit comparison test
In order for the comparison test to provide us with a conclusive result
on the convergence or divergence of a series, the series being tested must
be dominated by (‘lower’ than) a convergent series, or it must dominate
(‘higher’ than) a divergent series. If these conditions are not met then the
comparison test (sometimes called the direct comparison test) cannot be
used. For example, consider the series
2
3 1
1
n
n
2
=

that is nearly identical
to the series
2
3 1
1
n
n
+
=

that we proved is convergent in Example 10 a).
We strongly expect
2
3 1
1
n
n
2
=

to also converge. However, the inequality
2
3 1
2
3
n n
2
> shows that the series dominates the convergent geometric
series
2
3
1
n
n=

so the comparison test does not apply. In a case like this
another form of the comparison test, known as the limit comparison test,
can be used. This test can be particularly useful in comparing a series to a
p-series or a geometric series.
Proof
1. Let k and m be positive numbers such that k L m < < . Since
lim
n
n
n
a
b
L
→∞
= then there is a positive integer N, where N n > , such that
k
a
b
m
n
n
< < .
It follows that
kb a mb
n n n
< < .
If the series b
n
n=

1
converges then from the properties of series, the series
If applying the limit
comparison test you get
lim
n
n
n
a
b
→∞
= 0 and b
n
n=

1

diverges, this does not imply
that the series a
n
n=

1
also
diverges.
Limit comparison test
Given a
n
> 0 and b
n
> 0 for all n N  for some integer N, it follows that:
1. If lim
n
n
n
a
b
L
→∞
= , where L is fnite and positive, then the two series a
n
n=

1
and b
n
n=

1

both converge or both diverge.
2. If lim
n
n
n
a
b
→∞
= 0 and b
n
n=

1
converges then a
n
n=

1
also converges.
3. If lim
n
n
n
a
b
→∞
= ∞ and b
n
n=

1
diverges then a
n
n=

1
also diverges.
46
Series and Convergence
2
mb
n
n=

1
must also converge. Since mb
n
n=

1
dominates a
n
n=

1
then by the
comparison test a
n
n=

1
must converge. Likewise, if the series b
n
n=

1
diverges
then the series kb
n
n=

1
must also diverge, and since a
n
n=

1
dominates kb
n
n=

1

then by the comparison test a
n
n=

1
must diverge.
The proofs of parts 2 and 3 are left as exercises.
Example 11 – Using the limit comparison test
Determine the convergence or divergence of each series.
a)
2
3 1
1
n
n
2
=

b)
n
n n
2
1
1 +
=

c)
n n
n n
n
2
6 3
1
7
3
+
=

2

d) sin
1
1
n
n

=

Solution
a) As mentioned above, this series resembles the convergent geometric
series
2
3
1
n
n=

. Thus, we evaluate the following limit.
lim lim
n
n
n
n
n
n
→∞ →∞
=
2
3 1
2
3
3
3 1
2
2

=
→∞
lim
n
n n
n n n
3 3
3 3 1 3 2
=
→∞
lim
n
n
1
1 1 3 2
= 1
Since the limit is fnite and positive and
2
3
1
n
n=

converges then by the
limit comparison test the series
2
3 1
1
n
n
2
=

must also converge.
b) The given series
n
n
n
3
1
1 +
=

is similar to
n
n
n
3
1 =

which is a p-series best
written as
1
2 3
1
n
n=

. Since p =
2
3
1  we know that
1
2 3
1
n
n=

diverges. We
then evaluate the following limit.
lim lim
n n
n
n
n
n n
n
→∞ →∞
+
=

+
3
2 3
1 3 2 3
1
1
1
47
=
+
→∞
lim
n
n
n 1
= 1
Since the limit is fnite and positive and
1
2 3
1
n
n=

diverges then by the
limit comparison test the series
n
n
n
3
1
1 +
=

must also diverge.
c) As we saw in part b), it is possible to fnd a suitable p-series for
comparison purposes by disregarding all but the highest powers of n in
the numerator and denominator. Hence, for the given series

n n
n n
n
2
6 3
1
7
3
+
=

2
we can compare the series to
n
n n
n n
2
6
1
4
1
1
=

=

∑ ∑
= which is a
convergent p-series.
lim lim
n n
n n
n n
n
n n n
n n
→∞ →∞
+
=
+
( )
2
6 3
4
4 2
6 3
7
3
1
7
3
2
2
=
+
→∞
lim
n
n n
n n
6 5
6 3
7
3 2

=
+
→∞
lim
n
n n n n
n n n n
6 6 5 6
6 6 3 6
7
3 2
=
+ 1 0
3 0 2
=
1
3
Since the limit is fnite and positive and
1
4
1
n
n=

converges then by the
limit comparison test the series
n n
n n
n
2
6 3
1
7
3
+
=

2
must also converge.
d) Remember that in Section 13.2 of the book we proved lim
sin
n
x
x
→∞
= 1 by
means of the squeeze theorem. So we can use the limit comparison
theorem and compare the given series sin
1
1
n
n

=

to the divergent
harmonic series
1
1
n
n=

.
Hence, lim
sin
.
n
n
n
→∞

=
1
1
1
Therefore, since
1
1
n
n=

diverges then sin
1
1
n
n

=

also diverges.
Ratio test
In a geometric series, the ratio of adjacent terms is constant. This can be
expressed as

a r
a r
r
n
n
1
1
1
+
= .
48
Series and Convergence
2
We know that a geometric series converges if and only if this ratio is
between 21 and 1. In other types of series, the ratio of adjacent terms
does not remain constant but it can still give us helpful information about
whether or not the series converges, as indicated in the following theorem.
Proof
1. For the case when L < 1, there must be a number r with 0 1 < < r such that

a
a
r
n
n
+ 1

for all n suffciently large. Suppose that there exists some integer N such
that
a
a
r
n
n
+ 1
 if n N  .
Then
a
a
r a r a
N
N
N N
+
+

1
1
ജ ഛ

a
a
r a r a r a
N
N
N N N
+
+
+ +

2
1
2 1
2
ജ ഛ ഛ
and so on. Thus,
a a a a r r
N N N N
+ + + + + +
+ + 1 2
2
1
 
 .
This shows that for n N  the series a
n
n=

1
is dominated by the geometric
series a r
N
n
n
21
1 =

. Because 0 1 < < r this geometric series converges and by
the comparison test a
n
n=

1
must also converge.
2. For the case when L > 1, it must be true that a a
n n +
>
1
for all n
suffciently large. Therefore, lim
n
n
a
→∞
≠ 0 and the series a
n
n=

1
must
diverge by the nth term divergence test.
3. Applying the ratio test to the general p-series
1
1
n
p
n=

gives
lim lim .
n
p
p
n
p
n
n
n
n
→∞ →∞
+ ( )
=
+

=
1
1
1
1
1
We know that a p-series converges if p > 1 and diverges if p 1.
Hence, this shows that if lim
n
n
n
a
a
→∞
+
=
1
1 then it is possible to have a
Ratio test
Let a
n
n=

1
be a series with non-zero terms, and with

lim .
n
n
n
a
a
L
→∞
+
=
1
Then 1 the series converges if L < 1
2 the series diverges if L > 1
3 the test is inconclusive if L = 1.
49
series that is either convergent or divergent. Therefore, the ratio test is
inconclusive if L = 1.
The ratio test is particularly useful for testing series involving exponential
expressions or expressions with factorials, as illustrated in the following
example.
Example 12 – Using the ratio test
Determine the convergence or divergence of each series.
a)
n
n
n
n
3 1
0
3
4
+
=

b)
n
n
n
n
!
=

1
Solution
a) All the terms of the given series are positive so we can do without the
absolute value signs.
lim lim
n
n
n
n
n
n
n
n
a
a
n
n
→∞
+
→∞
+
+
+
=
+ ( )
1
3
2
1
3 1
1 3
4
3
4
=
+ ( )
⋅ ⋅

→∞
+
+ +
lim
n
n
n
n
n
n
n
1 3
3
4
4
3
3
2
1 1
=
+ ( )
→∞
lim
n
n
n
3 1
4
3
3
= <
3
4
1
Therefore, by the ratio test the series
n
n
n
n
3 1
0
3
4
+
=

converges.
b) Again, the series has only positive terms so we can write the ratio test
without absolute signs.
lim lim
!
!
n
n
n
n
n
n
a
a
n
n
n
n
→∞
+
→∞
+
=
+ ( )
+ ( )
1
1
1
1
=
+ ( ) + ( )
+ ( )

→∞
lim
!
!
n
n
n
n n
n n
n
n
1 1
1
=
+ ( )
→∞
lim
n
n
n
n
n
1
=
+ 

→∞
lim
n
n
n
n
1
= +

→∞
lim
n
n
n
1
1
= > e 1
Therefore, by the ratio test the series
n
n
n
n
!
=

1
diverges.
50
Series and Convergence
2
As we will learn even further in the next section, the ratio test is useful in
Example 13
For what values of x will the series
2
1
n
n
n
nx
=

converge?
Solution
Applying the ratio test gives the following inequality to solve.
lim lim
n
n
n
n
n
n
n
n
a
a
n x
nx
→∞
+
→∞
+
+
=
+ ( )
<
1
1
1
2
1
2
1
lim
n
n
n
n
n
n x x
nx
→∞
+ ( )
⋅ <
2 2
1 2
1
lim
n
n
x n
→∞
+ ( )
<
2
1
1

1 2
1
1
x
n
n
n

+
<
→∞
lim

2
1
x
<
x > 2
Therefore, the series converges for any values of x such that
x x < > 22 2 or .
Note that when x < 22 the terms of
2
1
n
n
n
nx
=

will alternate between
positive and negative. The ratio test is very useful in analyzing series with
alternating terms that we will take a closer look at in the next section.
Although the ratio test worked in Example 12 part b) we could have used the nth
term divergence test to prove that the series diverges by considering the following:

n
n
n
n
!
=

= + + + +
1
2 3 4
1
2
2
3
6
4
24

and for the nth term a
n n n n
n
n
n
=
⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅ ⋅

1 2 3

Thus as n →∞ the terms do not approach 0 and the series diverges by the nth term
divergence test. It is often the case that we can determine whether or not a series
converges by more than one test. The summary at the end of this section gives some
tips on how to fnd the most efcient test to apply for a certain series.
When applying the ratio test to
series involving quotients
of expressions with factorials,
it is often necessary to perform
simplifcation steps similar
to those we did in Example 12:

n
n
n
n n n
!
!
!
! + ( )
=
+ ( )
=
+ 1 1
1
1
51
2.3 Alternating series and absolute
convergence
Although we have encountered series whose terms alternate between
positive and negative, the four convergence tests we have established – with
the exception of the ratio test – apply only to series with positive terms.
We have encountered series with some negative terms, for example these
alternating series:
2 2
2 1
3
1
9
1
27
1
81
1
3
+ + + +
( )
+
 
n
n
[Example 4 a)]
3 3 3 3 3 3 3 1
1
2 2 2 2 + + + + ( ) +
+
 
n
[Example 6 b)]
Series such as these, having terms that are alternately positive and negative,
are called alternating series. The frst series above is a geometric series
with r = 2
1
3
, so it converges to
2
2 2
2
1
3
1
1
3
1
4 

= , and the second series
diverges by the nth term divergence test. But not all alternating series will
be geometric nor satisfy the conditions of the nth term divergence test. The
following test can be used to determine the convergence of a wider range
of alternating series that satisfy certain conditions.
Proof
Consider the sequence of partial sums s
n
{ } for an alternating series
21
1
1
( )
+
=

n
n
n
a such that a a
n n + 1
 and lim
n
n
a
→∞
= 0. Figure 2.6 shows a graph
of a few terms of s
n
{ }.
0 s
2
s
4
+ a
1
a
2
+ a
3
a
4
+ a
5
a
6
s
6
s
5
s
3
s
1
s
Observe how the alternating signs of the terms of the series cause the
partial sums to be alternately larger and smaller. As n increases the points
corresponding to the nth partial sum ‘jump’ back and forth on either
side of their limit s, gradually closing in as the value of the terms go to
These examples show that
an alternating series can be
written in one of two forms:
21
1
1
( )
+
=

n
n
n
a where the
frst term is positive, or
21
1
( )
=

n
n
n
a where the frst
term is negative such that
a
n
is a positive number. The
alternating series test is stated
using the frst form. Since
2 2 2 1 1 1
1
( ) ( )

= ( )
+ n
n
n
n
a a

then the test also holds true for
the second form by applying
the property of convergent
series: if a
n
n=

1
converges, and
c is some real constant, then
ca
n
n=

1
also converges.
Alternating series test
The alternating series

2 2 2 1 0
1
1
1 2 3 4
( ) = + + > ( )
+
=

n
n
n
n
a a a a a a

converges if both of the following conditions are satisfed.
1. lim
n
n
a
→∞
= 0
2. a a
n n + 1
 for all n N  , for some positive integer N.
Figure 2.6 Convergence of the
partial sums of an alternating series
to their limit s.
52
Series and Convergence
2
zero. Also observe that as these ‘jumps’ get smaller and smaller (because
a a
n n + 1
 ), the odd-numbered terms in the sequence of partial sums,
s
n 2 1 +
{ }
, form a decreasing sequence and the even-numbered terms, s
n 2
{ },
form an increasing sequence. Furthermore, the decreasing sequence of
odd-numbered terms has s
2
as a lower bound and thus, by the bounded
sequence theorem, has a limit – call it L
1
. Similarly, the sequence of even-
numbered terms has s
1
as an upper bound and must also have a limit – call
it L
2
. If we can show that L L
1 2
= then the series converges to a unique limit
s L L = =
1 2
. With L
1
the limit of the sequence of odd-numbered terms in
the sequence of partial sums and L
2
the limit of the even-numbered terms,
it follows that
s L L s
n n 2 2 1 2 1
  
+
and
s s a
n n n 2 1 2 2 1 + +
= 2
Since
lim
n
n
a
→∞
+
=
2 1
0
L L
1 2
=
Therefore, the sequence of partial sums of the alternating series
21
1
1
( )
+
=

n
n
n
a converges.
When applying the alternating series test, it is best to verify condition (1)
frst. If condition (1) fails, that is, lim
n
n
a
→∞
≠ 0, then the series diverges by
the nth term divergence test. If lim
n
n
a
→∞
= 0 but condition (2) fails then the
alternating series test is inconclusive. There is another condition implied
that must be met – that the series is truly (eventually) alternating. If this is
not obvious by inspection then the easiest way to verify that a series
written in the form 21
1
1
( )
+
=

n
n
n
a or 21
1
( )
=

n
n
n
a is alternating is to show
that a
n
> 0 for all n N  , for some positive integer N.
Example 14 – Using the alternating series test
Determine the convergence or divergence of each series.
a) 21
1
1
1
2
( )
+
+
=

n
n
n
n

b)
2
2
1 2
3 1
1
( )
=

n
n
n
n

c) 21
1
1
( )
+
=

n
n
n
n
ln
Solution
a) The series 21
1
1
1
2
( )
+
+
=

n
n
n
n
is alternating because
n
n
2
1
0
+
> for n 1.
Condition (1) is easily verifed.

lim lim lim
n
n
n n
a
n
n
n n
n n n
→∞ →∞ →∞
=
+
=
+
=
+
=
2
2
2 2 2
1 1
0
1 0
0
53
Now, let’s attempt to satisfy condition (2) by proving the inequality
a
n + 1
< a
n
for a
n
n
n
=
+
2
1
.

n
n
n
n
+
+ ( ) +
+
1
1 1
1
2 2

n n n n + ( ) +
( )
+ ( ) +

1 1 1 1
2
2
 Cross-multiplying; both denominators
n n n n n n
3 2 3 2
1 2 2 + + + + + 

are positive.
1
2
n n +
n n + ( ) 1 1 
Since n 1, then the inequality n n + ( ) 1 1  is true. Hence, a a
n n + 1

and condition (2) is satisfed. Therefore, the series 21
1
1
1
2
( )
+
+
=

n
n
n
n

converges by the alternating series test.
b) The series
2
2
1 2
3 1
1
( )
=

n
n
n
n
is alternating since
2
3 1
0
n
n 2
> for all n 1, but
lim lim
n
n
n
a
n
n
→∞ →∞
= = ≠
2
3 1
2
3
0
2

so condition (1) is not satisfed.
Applying the nth term divergence test, we need to fnd the limit of the
nth term as n → ∞.
lim lim lim
n
n
n
n
n
n
n
n
n
→∞ →∞ →∞
( )
= ( ) ⋅
2
2
2
2
1 2
3 1
1
2
3 1

lim
n
n
n
→∞
=
2
3 1
2
3 2
but lim
n
n
→∞
( ) 21 does not exist (Example 1 a)), so

lim
n
n
n
n
→∞
( ) 2
2
1 2
3 1
does not exist. Therefore, the series diverges by the nth
term divergence test.
c) a
n
n
n
= >
ln
0 for all integers n 2, so the series is alternating.
Checking condition (1) we can evaluate the following limit using
l’Hôpital’s rule because it has the indeterminate form

.
lim lim
ln
lim
ln
lim
n
n
n n n
a
n
n
d
dx
n
d
dx
n
→∞ →∞ →∞ →∞
= =
( )
( )
=
11
1
0
n
=
Hence, condition (1) is satisfed.
For condition (2) we must show that the sequence given by a
n
n
n
=
ln

is decreasing. It is not obvious whether this is true so we consider the
derivative of the related function f x
x
x
( ) =
ln
.
′ ( ) =

= < f x
x
x
x
x
x
x
1
1
0
2 2
2
2
ln
ln
for all x e >
54
Series and Convergence
2
Hence, f is decreasing for x e > which means that f n f n + ( ) < ( ) 1 , so it
follows that a a
n n + 1
 for n 3.
Therefore, both conditions of the alternating series test have been
satisfed and the series 21
1
1
( )
+
=

n
n
n
n
ln
is convergent.
Take another look at Figure 2.6 that was used in the proof of the alternating
series test. Recalling that s is the limit of the partial sums, notice that
s s a 2
3 4
< , s s a 2
4 5
< , s s a 2
5 6
< , etc. Furthermore, note that s is always
between any two consecutive partial sums. This provides us with the means
to estimate the error when we use the partial sum s
n
to approximate the
sum of an alternating series.
Proof
As previously mentioned, the sum, s, of a convergent alternating series is
always between any two consecutive partial sums. That is,
s s s
n n
 
+ 1
, if n is even and s s s
n n + 1
  , if n is odd.
Whether n is even or odd, it follows that

s s s s
n n n
Ϫ ഛ Ϫ
+ 1
.
Given that
a s s
n n n + +
=
1 1
2
Remember 21
1
1
( )
+
=

n
n
n
a is an alternating series, so a
n +
>
1
0.
s s a
n n
Ϫ ഛ
+ 1
and therefore the proof is complete.
Example 15
Show that
21
4
1
( )
=

n
n
n
converges, and fnd the sum of the series with error
less than 0 0001 . .
Solution
Since lim
n
n
→∞
=
1
0
4
and
1
1
1
1
4 4
4
4
n
n
n n
+ ( )
⇒ + ( )   is true for all
n 1
;
the series satisfes both conditions of the alternating series test and
therefore converges.
Note that the alternating
series estimation theorem
does not give a formula for
the precise value of the error,
but rather a bound for the
error. Also note that this rule
for the bound of the error
when estimating s with s
n
only
applies to alternating series
that satisfy the condition of the
alternating series test.
Alternating series estimation theorem
Suppose that 21
1
1
( )
+
=

n
n
n
a is a convergent alternating series that satisfes both conditions
of the alternating series test and has an unknown sum of s. When estimating s with the
sum of the frst n terms, the absolute value of the remainder R
n
(i.e. the amount of error)
is less than or equal to the frst unused term. That is,
R s s a
n n n
=
+
Ϫ ഛ
1
.
In other words, the error generated in estimating the sum with the nth partial sum does
not exceed the value of the n + 1 term.
55
We know from the alternating series estimation theorem that the sum of the
frst nine terms will give an estimate for the sum with an error of at most
a
9 1 4
1
10
0 0001
+
= = . .
Our GDC computes the ninth partial sum to be
s
9 4 4 4 4 4 4 4 4
1
1
2
1
3
1
4
1
5
1
6
1
7
1
8
1
9
0 947092 = + + + + ≈ 2 2 2 2 2 2 . 55924.
This estimate of the sum of the series is accurate to three decimal places
because an error of less than 0.0001 does not affect the third decimal place.
Therefore, the sum of the series
21
4
1
( )
=

n
n
n
is s ≈ 20 947 . , correct to three
decimal places.
Example 16
Determine the convergence or divergence of the alternating harmonic
series
21
1
1
( )
+
=

n
n
n
.
Solution
Applying the alternating series test we have
1 lim lim
n
n
n
a
n
→∞ →∞
= =
1
0
and
2 a a
n n
n n
n n +

+
⇒ +
1
1
1
1
1    which is true for all n.
Therefore,
21
1
1
( )
+
=

n
n
n
converges by the alternating series test.
Absolute and conditional convergence
In the next section, we will learn that the alternating harmonic series
converges to exactly ln2.
ln2 1
1
2
1
3
1
4
1
5
1
6
1
1
= + + + +
( )
+
+
2 2 2
2
 
n
n
But more relevant to this section is that the result of Example 16 illustrates
an important point to investigate further. We know that the harmonic
series (a p-series with p = 1) diverges. However, if we take the harmonic
series and change the sign of alternate terms to get the alternating
harmonic series (Example 16), the positive and negative terms offset
one another to produce a series that converges even though the series
consisting of only positive terms diverges. The same situation is true of
the series
n
n
n
2
1
1 +
=

. You may recall that in Example 7 d) we used the
integral test to prove that this series diverges. However, in Example 14 c) of
this section we showed that the corresponding alternating series
21
1
1
1
2
( )
+
+
=

n
n
n
n
converges. In contrast, the alternating series
21
4
1
( )
=

n
n
n

-1+1/2^4-1/3^4+1
/4^4-1/5^4+1/6^4
-1/7^4+1/8^4-1/9
^4
-.947095924
56
Series and Convergence
2
(Example 15) converges and so does the corresponding series with positive
terms
1
4
1
n
n=

(a p-series with p = > 4 1). The difference between these two
situations requires us to defne two types of convergence when considering
the convergence of a series with positive and negative terms as occurs with
any alternating series.
We have seen then that
21
1
( )
=

n
n
n
(alternating harmonic series) and
21
1
1
1
2
( )
+
+
=

n
n
n
n
are both conditionally convergent because for each
the series composed of their terms all made positive diverges. Whereas
21
4
1
( )
=

n
n
n
is absolutely convergent because its corresponding series of
positive terms also converges. You may wonder if it is possible for a series
with positive terms, a
n
n=

1
, to converge, but for a related series with some
(or all) of the terms changed to negative, a
n
n=

1
, to diverge. The answer is
no, and we state the following theorem.
Proof
It is true that 0 2   a a a
n n n
+ because by the defnition of absolute
value a
n
is either a
n
or 2a
n
. A given condition for the theorem is that
a
n
n=

1
converges, so 2
1
a
n
n=

also converges. Therefore, by the comparison
test a a
n n
n
+ ( )
=

1
converges. Since a a a a
n n n n
= + ( ) 2 , it follows from
properties for convergent series that a a a a
n
n
n n
n
n
n =

=

=

∑ ∑ ∑
= + ( )
1 1 1
2 where
both series on the right converge. Therefore, a
n
n=

1
must converge. Q.E.D.
Absolute and conditional
convergence
Suppose a
n
n=

1
is a series with
positive and negative terms
that is convergent.
If a
n
n=

1
converges, then
a
n
n=

1
is said to be absolutely
convergent.
If a
n
n=

1
diverges, then a
n
n=

1

is said to be conditionally
convergent.
The absolute convergence
theorem essentially says
that it is not possible to take
a convergent series with only
positive terms and change
some of them to negative
to create a new series that is
divergent. However, as the
alternating harmonic series
demonstrates, it is possible to
take a convergent series with
positive and negative terms and
change them all to positive
to create a new series that is
divergent.
Absolute convergence theorem
If a
n
n=

1
converges, then a
n
n=

1
also converges, and therefore a
n
n=

1
is absolutely
convergent.
When trying to determine if an alternating series is absolutely convergent, conditionally
convergent, or divergent, it is most efective to frst check if the limit of the nth term is
zero. If it is not then the series diverges, and you are fnished. If the nth term divergence
test is inconclusive then check whether the related series of positive terms converges
(using any of the four tests for positive series). If it converges, then by the absolute
convergence theorem, the series is absolutely convergent and you are fnished. If it
diverges, then test the alternating series using the alternating series test. It is inefcient
to start by frst applying the alternating series test.
Example 17
Classify each series as absolutely convergent, conditionally convergent, or
57
divergent.
a)
21
1
( )
=

n
n
n!
b) 21
2 1
2
1
( )
+
=

n
n
n
n
c) 21
1 1
1
( )

+
=

n
n
n
sin
Solution
a) lim
!
n
n
n
→∞
( )
=
21
0, so result of nth term divergence test is inconclusive.
We next consider the corresponding series with only positive terms

1
1
n
n
!
=

. Recall that in Example 10 c), we used the comparison test
to show
1
1
n
n
!
=

converges. We now apply the alternating series test
to
21
1
( )
=

n
n
n!
. Knowing lim
n
n
→∞
=
1
0, and since 0
1 1
< <
n n !
for all
n 1 then lim
!
n
n
→∞
=
1
0. Thus lim
n
n
a
→∞
= 0. We now need to show that

1
1
1
1
n n
n n
+ ( )
⇒ + ( )
! !
! !   . Rewriting n + ( ) 1 ! as n n ! + ( ) 1 gives
n n n ! !  + ( ) 1 which is clearly true for all n 1. Thus a a
n n + 1
 ,
and we have satisfed both conditions of the alternating series
test. Hence,
21
1
( )
=

n
n
n!
converges and converges absolutely because

21 1
1 1
( )
=
=

=

∑ ∑
n
n n
n n ! !
also converges.
b) We can apply the nth term divergence test to show that 21
2 1
2
1
( )
+
=

n
n
n
n

diverges. Recall that the absolute value theorem stated that if lim
n
n
a
→∞
= 0
then lim
n
n
a
→∞
= 0. From this we can also say that if lim
n
n
a
→∞
≠ 0 then
lim
n
n
a
→∞
≠ 0. We apply l’Hôpital’s rule twice to prove that lim
n
n
n
→∞

2
0
2
.
lim lim
ln
lim
ln
n
n
n
n
n
n
n n
→∞ →∞ →∞
= =
( )
= ∞
2 2 2
2
2 2
2
2
2
(does not exist)
Therefore, by the nth term divergence test 21
2 1
2
1
( )
+
=

n
n
n
n
diverges.
c) In Example 11 d) we compared the series sin
1
1
n
n

=

to the divergent
harmonic series
1
1
n
n=

and using the limit comparison test showed
that sin
1
1
n
n

=

diverges. We now turn our attention to the given
Note: For Example 17 a), we
could have been more efcient
by applying the absolute
convergence theorem since
we have previously used the
comparison test to show that
( )
!
!
21
1
1 1
n
n n n
n
=
=

=

∑ ∑
converges.
58
Series and Convergence
2
series and frst need to confrm whether it is an alternating series. Since
sin
1
0
n

> for all n 1 then 21
1 1
1
( )

+
=

n
n
n
sin is an alternating series
and we can apply the test for alternating series.
The graph shown in Figure 2.7 provides confrmation that not only
limsin
n
n
→∞

=
1
0, but also that sin sin
1
1
1
n n +

<

for all n 1.
Thus the series satisfes the alternating series test and converges. Since
the corresponding series of positive terms, sin
1
1
n
n

=

, diverges,
21
1 1
1
( )

+
=

n
n
n
sin converges conditionally.
Rearrangements of conditionally convergent series
The distinction between absolute and conditional convergence is important in many
applications of infnite series. It seems perfectly logical that it is possible to rearrange
a fnite number of terms in an infnite series without afecting the sum. However, if we
rearrange an infnite number of terms in an infnite series, the sum is unchanged only
if the series is absolutely convergent. An extraordinary characteristic of series that are
conditionally convergent is that their terms can be rearranged to form a divergent
series, and even rearranged to form a series that converges to any predetermined sum.
This is a direct consequence of the fact that the sum of an infnite series is defned to be
the limit of the sequence of its partial sums. As mentioned previously, this means that
operations (such as the associative property) that are valid for fnite sums are not valid
for infnite sums.
We can illustrate this paradoxical behaviour with the alternating harmonic series that is
conditionally convergent. As stated earlier without explanation (next section), the sum
of the alternating harmonic series is ln 2.

1
1
2
1
3
1
4
1
5
1
6
1
7
1
8
1
9
1
10
2 2 2 2 2 2 + + + + + =

ln
(1)
Consider the following series:

1
1
3
1
2
1
5
1
7
1
4
1
9
1
11
1
6
+ + + + + + 2 2 2

(2)
(2) consists of a rearrangement of the same terms as in (1). It is plausible to expect that
the sum of the series in (2) is also ln 2.
Let’s continue by dividing (1) by 2, giving:

1
2
1
4
1
6
1
8
1
10
1
2
2 2 2 2 + + =

ln
(3)
Now we add (3) and (1):
1
1
2
1
3
1
4
1
5
1
6
1
7
1
8
1
9
1
10
1
11
2 1
1
2
1
4
2 2 2 2 2 2
2
+ + + + + = ( )

ln
++ + = ( )
1
6
1
8
1
10
1
2
2 3 2 2

ln

(1)
(3)
Figure 2.7
x
0 1
1
5
y
y = sin x
1
4
1
3
1
2
π
2
59
Test Converges Diverges Notes
nth term divergence test
a
n
n=

1
lim
n
n
a
→∞
≠ 0
Can only be used to show
divergence
Geometric series
a r
n
n
1
0 =

r < 1 r 1
S
a
r

=
1
12
p-series
1
1
n
p
n=

p > 1 p 1
Harmonic series when p = 1
Integral test
a
n
n=

1
;

a f n
n
= ( )
f is continuous, positive and decreasing
f d x x ( )

1

converges
f d x x ( )

1

diverges
s f d
n
n
+ ( )
+

x x
1
and

s f d
n
n
+ ( )

x x are bounds for
estimation of sum by s
n
Comparison test
a b
n
n
n
n =

=

∑ ∑
1 1
and
0 < a
n
< b
n
b
n
n=

1

converges

=

a
n
n 1

converges
a
n
n=

1

diverges

=

b
n
n 1
diverges
Useful for series similar to
p-series or geometric series
Limit comparison test
a b
n
n
n
n =

=

∑ ∑
1 1
and
a
n
> 0, b
n
> 0
lim
n
n
n
a
b
L
→∞
= ; if

0 < < ∞ ⇒ L

both behave the same
L = ⇒ 0

if b
n
converges then a
n
converges
L = ∞ ⇒

if b
n
diverges then a
n
diverges
Useful if not able to show
0 < a
n
< b
n
for direct
comparison
Ratio test
a
n
n=

1
lim
n
n
a
a
→∞
+
<
1
1
1 lim
n
n
a
a
→∞
+
>
1
1
1
Inconclusive if

lim
n
n
a
a
→∞
+
=
1
1
1
Alternating series test
21
1
1
( )
=

+

n
n
n
a
lim
n
n
a
→∞
= 0

and

0
1
< <
+
a a
n n
s
n
as estimate of sum
remainder:
R a
n n
<
+ 1
The result is
1
1
3
1
2
1
5
1
7
1
4
1
9
1
11
1
6
3
2
2 + + + + + + = 2 2 2

ln (4)
where the terms are arranged the same as in (2), but the sum is not what we
expected. So which is correct, (1) or (4)? The answer is that they are both correct.
Although both (1) and (4) are series containing the same terms, by rearranging the
terms we have manipulated how the partial sums are formed which afects the limit
of the partial sums and, consequently, afects the sum of the series.
Table 2.1 Tests for infnite series.
60
Series and Convergence
2
Guidelines for testing series for convergence
Important questions to consider:
1. Is lim
n
n
a
→∞
= 0? If not, the series diverges by the nth term divergence test.
2. Is the series geometric, or similar to a geometric series? If similar, apply one of the
comparison tests.
3. Is the series a p-series, or similar to a p-series? If similar, apply one of the
comparison tests.
4. Consider a
n
= f(n). Is f a continuous, positive, decreasing function and is it
possible to integrate f ? If so, try integral test.
5. Does a
n
involve n in a product or power, or has an expression with factorials? If
so, try the ratio test.
6. Is the series an alternating series? If so, try the alternating series test. Remember
that if a
n ∑
is convergent then a
n ∑
is absolutely convergent. Testing a
n ∑

makes more tests available.
1 Using properties of convergent series and geometric series, fnd the sum of each
of the series.
a
7
2
3
0
n
n
n=

b
1
2
2
3
0
n n
n
2

=

c
5 3 2
9
3
1
n n
n
n
+ ( ) 

=

2 Use partial fraction decomposition to fnd the sum of each of the following
telescoping series.
a
1
2
1
n n
n
+ ( )
=

b
1
1
2
2
n
n
2
=

c
1
1
1
1
n n
n
+ ( ) + + ( )
>
=

x x
x , 2

In questions 3–10, write the frst four terms of the infnite series and determine
whether the series is convergent or divergent. If the series is convergent, fnd its sum.
3
n
n n
2
1 1 + =

∑ 4
3
4
1
1
n
n
2
=

5
ln
1
1
n
n

=

6
21
3
2
1
1
( )
+
=

n
n
n
7
n
n
n
!
3
1 =

∑ 8
cos n
n
π ( )
=

1
9
2 3
5 6
1
n
n
n
+
+
=

∑ 10
e
n
n
2
=

1
11 a Find x x
x
e d
2

by using the method of integration by parts.
b Use the integral test to determine whether the series ne
n
n
2
=

1
is convergent
or divergent.
12 Use the integral test to determine whether the series is convergent or divergent.
a
1
2
n n
n
ln
=

b
1
4
2
1
n
n
+
=

13 Show that
n
n
n
2 3
2
1
+
=

∑ diverges by both of the following methods.
a Using the comparison test, compare the series to
1
2
1
n
n =

∑ .
b Using the limit comparison test, compare the series to
1
1
n
n =

.
Exercise 2.1
61
14 Show that
1
3
1
n
n
n =

converges by a the comparison test, and b the ratio test.
15 Give an example to show that the converse of the nth term divergence test is
false. That is, fnd a series that diverges even though lim
n
n
a
→∞
= 0.
16 Use the ratio test to show that
n
n
n
10
0
10
=

converges.
In questions 17–30, determine the convergence or divergence of the series.
17
1
1
2
2 n n n 2 =

18
n
n
n
+
=

1
2
1
19
lnn
n
n
2
1 =

20
2
3 1
0
n
n
n
+
=

21
n
n
n
!
! + ( )
=

2
0
22
2
1
1
n
n
n +
=

23 21
1
2 1
0
( )
+
+
=

n
n
n
n
24
n
n
n
3
1
2 ln ( )
=

25
n
n
n
n
!
=

1
26
n
n
n
n
+

=

1
1
27 21
2
1
2
( )
+
+
=

n
n
n
n n
28
21
1
( )
=

n
n
n
n
29
1
2 1
1
n
n
2
=

∑ 30
1
0
e
n
n =

31 Use the integral test to determine whether
1
2
n n
n
ln
=

converges or diverges.
32 Find the sum of the following infnite series.

5
1 2
5
2 3
5
3 5 ×
+
×
+
×
+

33 For each series, use the sum of the frst four terms to approximate the sum of the
series. State an upper bound for the error of the approximation.
a
2
2
1
1
2 1
1
2
1
( )
( )
+
=

n
n
n
b
21
1
1
( )
+
=

n
n
n
n
x
34 a Express n
2
+ 2n + 2 in the form (n + a)
2
+ b where a and b are integers.
b Use the integral test to determine whether
1
2 2
2
1
n n
n
+ +
=

∑ converges or
diverges.
35 Determine whether
arctann
n
n=

1
converges or diverges by comparing the series
to
1
1
n
n=

and applying the limit comparison test.
36 a Use the method of Example 8b) with n = 5 to estimate the sum of the series

1
3
1
n
n=

∑ .
b With this method, how many terms of the series are needed to make the
error of the estimate for
1
3
1
n
n=

less than 0.0005?
62
Series and Convergence
2
37 Use the alternating series estimation theorem to determine the minimum
number of terms of the series 1
1
2
1
3
1
4
4 4 4
2 2 + +

so that an approximation of
the sum has an error less than 0.000 05.
38 Give an example of a series that is conditionally convergent. That is, a series that
is convergent but not absolutely convergent.
In questions 39–44, determine whether each series converges absolutely, converges
conditionally, or diverges.
39
21
1
2
1
( )
+
=

n
n
n
40
2
2
1
2 1
1
1
( )
+
=

n
n
n
41
21
1
1
2
2
1
( )
+ ( )
+
=

n
n
n
n
42
21
2
( )
=

n
n
n ln
43
cos n
n
n
π ( )
=

2
1
44
n
n
n
n
2
2
3
4
1
1
( )
=

45 Describe how the terms of the alternating harmonic series 1
1
2
1
3
1
4
2 2 + +

can
be rearranged so that its sum is 1.
46 What is the minimum number of terms of the series
21
1
( )
=

n
n
n!
needed to
approximate the sum of the series correct to three decimal places?
47 Prove parts 2 and 3 of the limit comparison test.
63
3.1 Introduction
Have you ever considered how your calculator computes values for certain
functions? For functions such as f x x x g x
x x
x x
h x x ( ) = + ( ) =
+
+
( ) = 3 5 8
2
4 6
7 3
2
4
3 2
2
2 2
2 , , and

f x x x g x
x x
x x
h x x ( ) = + ( ) =
+
+
( ) = 3 5 8
2
4 6
7 3
2
4
3 2
2
2 2
2 , , and the method of evaluation is fairly straightforward
because these are algebraic functions. As explained in Chapter 3 of the book,
algebraic functions can be expressed as a fnite number of sums, differences,
multiples, quotients and radicals involving x
n
. Polynomial functions,
rational functions and functions involving radicals are examples of algebraic
functions. But how does your calculator compute values for a function such
as e
x
? This is an example of a transcendental function. A transcendental
function is non-algebraic, i.e. it cannot be expressed as a fnite number of
sums, differences, multiples, quotients and radicals involving x
n
. Other
familiar transcendental functions include the trigonometric and logarithmic
functions.
Let’s return to the primary question we wish to investigate. How does your
calculator compute the values of transcendental functions, such as e
x
? The
manufacturers of the calculator had to decide on a computational algorithm.
What computational method could be programmed into a calculator to
evaluate e
x
for a certain value of x ?
The answer lies in the fact that the calculator is summing up a type of series
with variable terms, called a power series, that is representing e
x
. In this
section we will see that the power series for the function f x e
x
( ) = is
x
n
x
x x x x
n
n
n
! ! ! ! !
=

= + + + + +…+ +…
0
2 3 4 5
1
2 3 4
. A calculator can only display a
fnite number of digits and thus it only sums enough terms to produce the
necessary degree of accuracy. For example, suppose we wanted to use this
series (we’ll investigate its derivation later) to evaluate e
2
to three signifcant
fgures.

e
2
2 3 4 5
1 2
2
2
2
3
2
4
2
5
= + + + + + +…
! ! ! !
Shown below are successively more accurate approximations for the value
of e
2
by summing the terms of the power series
x
n
n
n
k
!
=

0
for k = 2 3 9 , , ,  .
Power Series
3
Except for Example 13 in the
previous section, all the series
we have encountered thus far
contained terms consisting
of constants. A power series
is essentially a polynomial
function of infnite degree
expressed in terms of a single
variable (we will always use x).
A power series is a very useful
mathematical tool that can be
used to represent a range of
very important functions.
64
Power Series
3
e
2
1 2 3 ≈ + =
e
2
2
1 2
2
2
5 ≈ + + =
!
e
2
2 3
1
3
1 2
2
2
2
3
6 6 3 ≈ + + + = =
! !
.
e
2
2 3 4
1 2
2
2
2
3
2
4
7 ≈ + + + + =
! ! !
e
2
2 3 4 5
4
15
1 2
2
2
2
3
2
4
2
5
7 7 26 ≈ + + + + + = =
! ! ! !
.
e
2
2 3 4 5 6
16
45
1 2
2
2
2
3
2
4
2
5
2
6
7 7 35 ≈ + + + + + + = =
! ! ! ! !
.
e
2
2 3 4 5 6 7
8
21
1 2
2
2
2
3
2
4
2
5
2
6
2
7
7 7 380 ≈ + + + + + + + = =
! ! ! ! ! !
. 99523 7 38 ≈ .
e
2
2 3 4 5 6 7 8
122
3
1 2
2
2
2
3
2
4
2
5
2
6
2
7
2
8
7 ≈ + + + + + + + + =
! ! ! ! ! ! !
115
7 3873015 7 39 = ≈ . .
e
2
2 3 4 5 6 7 8 9
1 2
2
2
2
3
2
4
2
5
2
6
2
7
2
8
2
9
≈ + + + + + + + + + =
! ! ! ! ! ! ! !
77 7 38871252205 7 39
1102
2835
= ≈ . . 
Once we get past the ninth term in the series we are no longer adding
enough to change the frst three digits. Thus, the frst nine terms of the
series are suffcient to give an approximation of e
2
accurate to three
signifcant fgures.
A calculator (see screen image above) computes to an accuracy of ten
signifcant fgures the value of e
2
to be 7.389 056 099. It certainly appears
that the series
2
0
n
n
n!
=

is converging to e
2
. For any given value of x,
x
n
n
n
!
=

0

is an infnite series. This leads to an important question: For what values of
x does the power series converge?
Before addressing this question, let’s give a proper defnition for a power
series.
3.2 Power series
e^(2)
7.389056099
Performing such computations
entirely by hand would be
immensely tedious (and
prone to error). However, this
is not an impediment for an
electronic computing device
like a GDC. As we will see, the
more efcient by means of a
formula that determines the
number of terms required for
a power series to produce a
value to a given accuracy.
Defnition of power series
If x is a variable, then an infnite series of the form
a c a a c a c a c a
n
n
n
n
x x x x 2 2 2 2 ( ) = + ( ) + ( ) + ( ) + +
=

0
0 1 2
2
3
3

xx 2c
n
( ) +

is called a power series centred at c, where c is a constant and a
n
is the rule that
determines each of the coefcients a
0
, a
1
, a
2
, … . Note that we have (x 2 c)° = 1 even
when x = c.
For any power series
centred at c = 0, we have
a a a a
a a
n
n
n
n
n
x x x
x x
=

= + +
+ + + +
0
0 1 2
2
3
3
 
.
65
At each value of x, a power series becomes a series of constants. In
the previous section we gave a great deal of attention to such series,
investigating whether they converge or diverge. The issue of convergence
is very important for power series because for each value of x for which a
power series converges, the series represents the number that is the sum of
the series. Therefore, a power series defnes a function. The function
f x a x c
n
n
n
( ) = ( )
=

2
0
has as its domain all values of x for which the power
series converges. It is evident that every power series is convergent for
x c = . Some power series are only convergent at x c = (see Example 3). Far
more useful power series will converge for a fnite interval with the same
centre as the series (see Example 1), or converge for all x (see Example 2).
Example 1
For the general power series a x c
n
n
n
2 ( )
=

0
, if we let a
n
= 1 for all n and
‘centre’ the series at c = 0, we get the geometric series
x x x x x
n
n
n
=

= + + + + + +
0
2 3
1
 
having frst terms a
1
1 = , and
common ratio r x = .
The sum formula for geometric series assures us that this series converges
to
1
12x
when 21 1 < < x , and consequently diverges when x 1.
Therefore, we can write
1
1
1
1 1
2 3
+ + + + + + = < < x x x x
x
x
n
 
2
2 , .
The expression on the right side of this equation defnes a function whose
domain is x x ∈ ≠ , 1. The expression on the left side defnes a function
whose domain is the interval 21 1 < < x . The equation can only be true
where both sides are defned, so its domain is 21 1 < < x , equivalent to
x < 1. On this domain, the given power series is a polynomial
representation of the function f x
x
( ) =
1
12
(Figure 3.1, overleaf). A power
series is best regarded as an attempt to describe a function locally, near
where it is ‘centred’, i.e. near the value of c. To illustrate this point, partial
sums of the series x
n
n=

0
with 3, 6 and 9 terms have been graphed in
Figure 3.2. Figure 3.3 shows the same three partial sums along with
f x
x
( ) =
1
12
focused on the interval 21 1 < < x .
66
Power Series
3

Observe how in the interval 21 1 < < x the graph of a partial sum of
x
n
n=

0
gets closer to that of the graph of f x
x
( ) =
1
12
as the number of
terms increase, but are not close outside this interval.
Example 2
We’ve demonstrated that the power series
x
n
x
x x x x
n
n
n
! ! ! ! !
=

= + + + + +…+ +…
0
2 3 4 5
1
2 3 4
represents the function
f x e
x
( ) = . Find the values of x for which this power series converges.
4
3
2
1
1
2
3 2 1 1 2
x
y
10
8
6
4
2

2
1 1
x
y
4
3
2
1
1
2
3 2 1 1 2
x
y
Figure 3.1 Graph of y =
1
12x
.
Figure 3.2 Graphs of the partial sums 1 + x + x
2
,
1
2 3 4 5
+ + + + + x x x x x
and

1
2 3 4 5 6 7 8
+ + + + + + + + x x x x x x x x .
Figure 3.3 The partial sums 1 + x + x
2
, 1
2 3 4 5
+ + + + + x x x x x ,
1
2 3 4 5 6 7 8
+ + + + + + + + x x x x x x x x and
1
12x
(dashed).
67
Solution
Example 13 in the previous section illustrated that the ratio test is effective
for answering this kind of question. Applying the ratio test gives the
following inequality to solve.
lim lim
!
!
n
n
n
n
n
n
a
a
x
n
x
n
→∞
+
→∞
+
=
+ ( )
<
1
1
1
1
lim
!
!
n
n
n
x x
n n
n
x
→∞
+ ( )
⋅ <
1
1
lim
n
x
n
→∞
+
<
1
1
x
n
n

+
<
→∞
lim
1
1
1
x ⋅ < 0 1
0 1 <
Therefore, the power series
x
n
n
n
!
=

0
is convergent for all real values of x.
This means that as we compute partial sums of more and more terms
for
x
n
n
n
!
=

0
they will become ever more accurate representations of the
function f x e
x
( ) = over all real numbers, and not just constrained to a
fnite interval as occurred in the previous example. This is illustrated by
comparing graphs of f x e
x
( ) = (Figure 3.4) to graphs of partial sums of
x
n
n
n
!
=

0
with 4, 5 and 10 terms in Figure 3.5.

50
40
30
20
10
6 4 2 2 4
x
y
Figure 3.4 Graph of y = e
x
.
50
40
30
20
10
6 4 2 2 4
x
y
Figure 3.5 The partial sums 1 + x +
x
2

__

2!
+
x
3

__

3!
, 1
2 3 4
2 3 4
+ + + + x
x x x
! ! !

and

1
2 3 4 5 6 7 8 9
2 3 4 5 6 7 8 9
+ + + + + + + + + x
x x x x x x x x
! ! ! ! ! ! ! !
.
68
Power Series
3
Example 3
Find the values of x for which n x
n
n
!
=

0
is convergent.
Solution
Again, applying the ratio test gives:

lim lim
!
!
n
n
n
n
n
n
a
a
n x
n x
→∞
+
→∞
+
=
+ ( )
1
1
1
= + ( )
→∞
lim
n
x n 1
= + ( )
→∞
x n
n
lim 1
= ∞ ≠ , x 0
Hence, the series diverges for all values of x, x ≠ 0. We need to check the
series when x = 0.
n
n
n
!0 1 0 0 1
0 =

= + + + =

Therefore, the power series n x
n
n
!
=

0
converges only at its centre, at x = 0.
As illustrated in the preceding three examples, the domain of a power
series can be a single point, an interval of the real numbers centred at c, or
all real numbers. The following theorem (which we present without proof)
states that these are the only possibilities.
The set of all values of x for which a given power series is convergent is
called the interval of convergence of the power series. The number R of
possibility 3 in the theorem is called the radius of convergence of the
power series. If possibility 1 occurs, then R = 0; and if possibility 2 occurs,
then R = ∞. For possibility 3, there remains the question of what happens
at the endpoints of the interval. Each endpoint must be tested separately
to determine if the series converges or diverges for that value of x. Thus, if
0 < < ∞ R the interval of convergence can be one of four different kinds,
as illustrated in Figure 3.6.
Convergence of a power series theorem
For a given power series a x c
n
n
n
2 ( )
=

0
exactly one of the following is true:
1. The series converges only when x = c.
2. The series converges for all real values of x.
3. There is a positive number R such that the series converges for x 2c R < , and
diverges for x 2c R > . The series may or may not converge at either of the
endpoints x = c R 2 and x = + c R so we need to check for convergence at
each of the endpoints.
69
In Example 1 we showed that the interval of convergence for the power
series x
n
n=

0
is 21 1 < < x , so R = 1. But we did not check the
endpoints. At x = 21, x
n
n
n
n =

=

∑ ∑
= ( ) = + +
0 0
1 1 1 1 1 2 2 2

; and at x = 1,
x
n
n
n
n =

=

∑ ∑
= = + + +
0 0
1 1 1 1

. Both of these series diverge confrming that
the interval of convergence is, in fact, 21 1 < < x . In Example 2, R = ∞;
and, in Example 3, R = 0.
As we did in Examples 2 and 3, it is best to use the ratio test to determine
the radius of convergence for a power series. The ratio test will fail when
x is an endpoint of the interval of convergence, so you will need to check
endpoints with one of the other tests from the previous chapter.
Example 4
Find the radius of convergence and interval of convergence for the series

x
n
n
n
n
24
2 3
1
( )
( ) =

.
Solution
We apply the ratio test:
lim lim
n
n
n
n
n
n
x
n
n
x
→∞
+
+
→∞
( )
+ ( )( )

( )
( )
=
2
2
4
2 1 3
2 3
4
2
1
1
nn x
n
24
3 2 2
( )
+ ( )
=
+

→∞
lim
n
n
n
x
2
6 6
4 2
=
1
3
4 x 2
Hence, the series converges for
1
3
4 1 x 2 < or x 24 3 < .
The radius of convergence is R = 3. The series is centred at c = 4, so the
series converges for
4 3 4 3 2 < < + x or 1 7 < < x .
At x = 1, the series is
2 2 3
2 3
1
2
1 1
( )
( )
=
( )
=

=

∑ ∑
n
n
n
n
n
n n
, which converges by the
alternating series test.
At x = 7, the series is
3
2 3
1
2
1 1
( )
( )
=
=

=

∑ ∑
n
n
n n
n n
, which diverges by limit
comparison with the harmonic series. Therefore, the interval of
convergence is 1 7 x < , or x ∈
[ [
1 7 , .
x c
R
( )
]c

R, c + R[
x c
R
( ]
]c

R, c + R]
x c
R
[ )
[c

R, c + R[
x c
R
[ ]
[c

R, c + R]
Figure 3.6 Intervals of
convergence.
Why are we interested in
power series? One reason is
that power series share many
of the desirable properties
of polynomials. In particular,
polynomial functions are
generally much easier to
diferentiate and integrate. The
power series representation
of a particular function may
enable us to perform some
difcult operations on the
function that would otherwise
be quite difcult, for example
diferentiation and integration
of the function.
70
Power Series
3
3.3 Deriving power series – Taylor and
Maclaurin series
In the preceding part of this chapter we were given the power series
e
x
n
x
x
n
n
= ∈
=

!
,
0

without any indication about its origin. In this section we will devise a
general method for constructing a power series.
As we stated in our work for Example 1, ‘a power series is best regarded as
an attempt to describe a function locally, near where it is “centred”, i.e. near
the value of c ’, and that ‘a power series is essentially a polynomial function
of infnite degree’. With theses ideas in mind, we propose the following.
To fnd a polynomial function P that represents another function f, begin
by choosing a number c in the domain of f at which f and P have the same
value. This is the frst requirement – that P c f c ( ) ( ) = . Thus, the graphs
of f and P will pass through the same point, c f c , ( ) ( ). The approximating
polynomial P is said to be expanded about c or centred at c. Of course, there
will be many polynomial functions that will have the same value as f at
x c = . We can make the graph of P further resemble that of f near the point
they share by a second requirement: P has the same slope of f at x = c, that
is, ′ ( ) = ′ ( ) P c f c . We can continue to improve how well P mimics the
behaviour of f near c by additionally requiring that ′′ ( ) = ′′ ( ) P c f c ,
P c f c
3 3 ( ) ( )
( ) = ( ), and so on.
Deriving the power series for f(x) = e
x
centred at 0
We start by fnding a frst degree polynomial P x a a x
1 0 1
( ) = + whose value and slope
agree with the value and slope of f x e
x
( ) = at x = 0 , that is, P f 0 0 ( ) = ( ) and
′ ( ) = ′ ( ) P f 0 0 .
f e 0 1
0
( ) = = P a a a
1 0 1 0
0 0 1 1 ( ) = + ( ) = ⇒ =
′ ( ) = ⇒ ′ ( ) = = f x e f e
x
0 1
0
′( ) = ⇒ ′( ) = = ⇒ = P x a P a a
1 1 1 1 1
0 1 1
Therefore, P x x
1
1 ( ) = + .
Now to fnd the second degree polynomial approximation P x a a x a x
2 0 1 2
2
( ) = + +
we require that ′′ ( ) = ′′ ( ) P f 0 0 , knowing that a
0
1 = and a
1
1 = .
′′ ( ) = ⇒ ′′ ( ) = = f x e f e
x
0 1
0
P x a a x
2 1 2
2 ′ ( ) = +
P x a P a a
2 2 2 2 2
2 0 2 1
1
2
′′ ( ) = ⇒ ′′
( ) = = ⇒ =

Therefore, P x x x
2
2
1
1
2
( ) = + + .
We continue and fnd the third degree polynomial P x a a x a x a x
3 0 1 2
2
3
3
( ) = + + +
and require that P f
3 3
0 0
( ) ( )
( ) = ( ) (third derivatives are equal).
71
f x e f e
x 3 3 0
0 1
( ) ( )
( ) = ⇒ ( ) = = P x a a x a x
3 1 2 3
2
2 3 ′
( ) = + +
P x a a x
3 2 3
2 6
′′
( ) = +
P x a P a a
3
3
3 3
3
3 3
6 0 6 1
1
6
( ) ( )
( ) = ⇒ ( ) = = ⇒ =
Therefore, P x x x x
3
2 3
1
1
2
1
6
( ) = + + + .
x 21.0 20.2 20.1 0 0.1 0.2 1.0
f(x) = e
x
0.367 879 0.818 731 0.904 837 1 1.105 17 1.221 40 2.718 28
P
2
(x) 0.5 0.82 0.905 1 1.105 1.22 2.5
P
3
(x) 0.
_
3 0.818
_
6 0.904 8
_
3 1 1.105 1
_
6 1.221
_
3 2.
_
6
Table 3.1 Comparing P
2
2
1
1
2
x x x ( ) = + + , P
3
2 3
1
1
2
1
6
x x x x ( ) = + + + , and f e x
x
( ) = .
Let’s apply the procedure one more time to fnd
P x a a x a x a x a x
4 0 1 2
2
3
3
4
4
( ) = + + + + .
f x e f e
x 4 4 0
0 1
( ) ( )
( ) = ⇒ ( ) = = P x a a x a x a x
4 1 2 3
2
4
3
2 3 4 ′
( ) = + + +
P x a a x a x
4 2 3 4
2
2 6 12 ′′
( ) = + +
P x a a x
4
3
3 4
6 24
( )
( ) = +
P x a P a a
4
4
4 4
4
4 3
24 0 24 1
1
24
( ) ( )
( ) = ⇒ ( ) = = ⇒ =
Therefore, P x x x x x
4
2 3 4
1
1
2
1
6
1
24
( ) = + + + + .
From the function values for
f(x) = e
x
(six signifcant fgures),
P
2
(x) and P
3
(x) displayed
in Table 3.1, we can make
two observations: (1) the
accuracy of the approximating
polynomial improves as
x → 0; and we would expect,
in general, as x → c; and (2)
the higher the degree of the
approximating polynomial
(more terms of the partial sum
of the power series), the better
the polynomial represents the
function f.
By now we can see the pattern for P x
n
( ).
P x x
x x x x
n
n
n
( ) = + + + + + + 1
2 3 4
2 3 4
! ! ! !

We complete the derivation of the power series for f x e
x
( ) = by
determining its interval of convergence by means of the ratio test and
checking interval endpoints, if they exist. In this way we derive the result
that was presented to us at the start of this chapter. That is,
e
x
n
x
x
n
n
= ∈
=

!
,
0
.
General result
We now wish to generalize the method for fnding the power series
representation for a given function f(x). For expansions about an arbitrary
value of c, it is convenient to write the polynomial in the standard form
from the defnition.

P x a a x c a x c a x c a x c
n n
n
( ) = + ( ) + ( ) + ( ) + + ( )
0 1 2
2
3
3
2 2 2 2

where P x
n
( ) is the nth partial sum of the infnite series

a x c
n
n
n
2 ( )
=

0
.
We will streamline the procedure that we used for deriving the power series
for f x e
x
( ) = . Repeated differentiation of P
n
produces
72
Power Series
3
′ = + + + +

P a a c a c na c
P
n n
n
n
( ) ( ) ( ) ( )
(
x x x x
1 2 3
2 1
2 3 2 2 2
2

′ xx x x
x
) ( )( ) ( ) ( )
( )
( )
= + + +
=
2 2 3 1
2 3
2
3
a a c n n a c
P
n
n
n
2 2 2
2

22 3 1 2
1
3
3
( ) ( )( ) ( )
( ) ( )
( )
a n n n a c
P n n
n
n
n
n
+ +
=

2 2 2
2
2
x
x (( ) ( )( ) n a
n
22 2 1

Evaluating P
n
and its frst n derivatives at x c = , we obtain the following:
P c a P c a P c
n n n
( ) = ′( ) = ′′( ) =
0 1
2aa P c n a
n
n
n 2

( )
! ( ) =
Because the value of f and its frst n derivatives must agree with the value
of P
n
and its frst n derivatives at x = c, it follows that:
a f c a f c a
f c
0 1 2
= ( ) = ′ ( ) =
′′ ( )
22! !
( )

a
f c
n
n
n
=
( )
Consequently, we assert that if f has a power series representation centred
at x = c, that is, if

f x a x c x c R
n
n
n
( ) = ( ) <
=

2 2
0
,
then its coeffcients are given by the formula
a
f c
n
n
n
=
( )
( )
!
.
Substituting this formula for a
n
into a x c
n
n
n
2 ( )
=

0
we determine that if a
function f has a power series representation centred at x c = , then it will
take the form as defned below.
For the simplicity of notation
we agree to defne the ‘zeroth’
derivative, f
(0)
, to be the
function f itself.
Defnition of Taylor series and Maclaurin series
If a function f has derivatives of all orders at x = c, then the series
f
f c
n
c f c f c c
f c
n
n
n
x x x ( ) =
( )
( ) = ( ) + ′ ( )( ) +
′′ (
( )
=

!
0
2 2
))
( ) + +
( )
( ) +
2
2
! !
( )
x x 2 2 c
f c
n
c
n
n
 
is called the Taylor series of the function f centred at c. As often occurs, if c = 0, then
the series becomes

f
f
n
f f
f
n
n
n
x x x x ( ) =
( )
= ( ) + ′ ( ) +
′′ ( )
+ +
( )
=

0
0 0
0
2
0
2
! !

ff
n
n
n
( )
!
0 ( )
+ x

and is called the Maclaurin series for f.
Although the English mathematician Brook Taylor (1685–1731) did not invent the
process of using polynomial approximations for transcendental functions (others
such as Leibniz, Johann Bernoulli and Abraham de Moivre had already used series
that we would call Taylor series), Taylor was the frst to provide a comprehensive
study of their importance and applications in a textbook he wrote in 1715. Similarly,
the Scottish mathematician Colin Maclaurin (1698–1746) did not invent the series
named for him. He also wrote a textbook (in 1742) that greatly furthered applications
of power series, so much so that Taylor series centred at x = 0 became known as
Maclaurin series. Isaac Newton (1642–1727) was an early innovator with power
series, especially his work in writing binomial expressions as series. A contemporary
of Newton’s was another brilliant Scottish mathematician, James Gregory (1638–
1675). Gregory published power (Maclaurin) series for tan x, sec x, arctan x and
arcsec x ten years before Maclaurin was born. Apparently Taylor was not aware of
Gregory’s mathematical work when writing his 1715 textbook.
73
Example 5
Find the Maclaurin series for f x x ( ) = sin and its interval of convergence.
Solution
To use the defnition of a Maclaurin series we need to repeatedly
differentiate f x x ( ) = sin , and evaluate each derivative at x = 0, suffciently
to establish a pattern for the power series.
f x f x x
0 ( )
( ) = ( ) = sin f 0 0 ( ) =
′ ( ) = f x x cos ′ ( ) = f 0 1
′′ ( ) = f x x 2sin ′′ ( ) = f 0 0
f x x
3 ( )
( ) = 2cos f
3
0 1
( )
( ) = 2
f x x
4 ( )
( ) = sin f
4
0 0
( )
( ) =
f x x
5 ( )
( ) = cos f
5
0 1
( )
( ) =
The pattern that emerges repeats in blocks of four. Thus, the power series is

f
n
x f f x
f
x
f
n
n
n
n ( )
=

( )
= ( ) + ′ ( ) +
′′ ( )
+ +

0
0 0
0
2
0
0
2
! !
( )

(( )
+
n
x
n
!

= + ⋅ + + + + + +
( )
+ 0 1
0
2
1
3
0
4
1
5
0
2 3 4 5
x x x x x
f
n
x
n
n
! ! ! ! !
( )
2
 

= + + + ( )
+ ( )
+
+
x
x x x x
n
n
n
2 2 2
3 5 7
2 1
3 5 7
1
2 1 ! ! ! !
 
Therefore, the Maclaurin series for f x x ( ) = sin is 21
2 1
2 1
0
( )
+ ( )
+
=

n
n
n
x
n !
.
Using the ratio test, we have
lim
!
!
lim
n
n
n
n
x
n
n
x
x
→∞
+ ( ) +
+
→∞
+ ( ) + ( )

+ ( )
=
2 1 1
2 1
2 1 1
2 1
22 3
2 1
2 3
2 1
n
n
n
n
x
+
+
+ ( )

+ ( )
!
!
=
+ ( ) + ( )
→∞
lim
n
x
n n
2
2 3 2 2
=
+ ( ) + ( )
→∞
x
n n
n
2
1
2 3 2 2
lim
= ⋅ x
2
0
= < 0 1
Hence, the power series − ( )
+ ( )
+
=

1
2 1
2 1
0
n
n
n
x
n !
for sinx converges for all real
values of x.
74
Power Series
3
Differentiation and integration of power series
If R is the radius of convergence of the power series

f a c
n
n
n
x x ( ) = ( )
=

0
2
then

′ ( ) = ( )

= ( )
( )
=

=

∑ ∑
f
d
d
a c
d
d
a c
n
n
n
n
n
n
x
x
x
x
x
0 0
2 2 == ( )
=

na c
n
n
n
0
1
x 2
2

for

c R c R 2 < < + x
and

f d a c d a c
n
n
n
n
n
n
x x x x x ( ) = ( )

= ( )
( )

∫ ∫

=

=

0 0
2 2 dd
n
c
n
n
x x =
+
( )
=

+

1
1
0
1
2

for c R c R 2 < < + x .
Finding a Taylor (or Maclaurin) series by this fve-step process can prove
to be diffcult, primarily fnding a pattern for the values of f c
n ( )
( ). We now
establish some very useful properties of power series, giving us ‘short cuts’
for fnding Taylor series from a known Taylor series.
3.4 Operations with power series
The power and versatility of representing functions with power series is
due largely to the fact that they retain many of the properties of ‘fnite’
polynomials. Two of the properties that make polynomials particularly
useful in calculus are the ease with which they can be differentiated and
integrated term by term. It is natural to ask whether the same holds true
for power series. The answer is provided by the next theorem, which we
state without proof.
Finding a Taylor series centred at c for a function f(x)
Step 1: Compute several derivatives of f. The ‘zeroth’ derivative, f
(0)
(x), is f (x) itself.
Step 2: Evaluate the derivatives at x = c, and try to identify a pattern for the values of
f
(n)
(c).
Step 3: Knowing that the rule for coefcients of a Taylor series is a
f c
n
n
n
=
( )
( )
!
, write the
formula for the Taylor series, substituting into
f c
n
c
n
n
n
( )
=

( )
( )

!
0
x 2 .
Step 4: Using the ratio test determine the radius of convergence for the series.
Step 5: If 0 < R < ∞, use an appropriate convergence test to check for convergence/
divergence at both of the interval endpoints x = c 2 R and x = c + R. State the
interval of convergence.
of the series produced by
diferentiating or integrating
a power series is equivalent
to that of the original series.
However, the interval of
convergence may change due
to convergence/divergence
changing at one or both of the
endpoints.
The above theorem means that term-by-term differentiation or integration
of a power series produces a power series that converges, respectively, to
the derivative or integral of the function represented by the original series,
given that we are in the interval of convergence of the original series. This
75
gives us another way to derive power series representations from a known
power series, as the next example illustrates.
Example 6
Find the Maclaurin series for f x x ( ) = cos .
Solution
We could apply the formula for a Maclaurin series as we did in Example 5
but it will be more effcient to simply differentiate the power series that we
already established for f x x ( ) = sin .
cos sin
!
x
d
dx
x
d
dx
x
n
n
n
n
= ( ) =
( )
+ ( )

+
=

21
2 1
2 1
0
=
( )
+ ( )

=
∞ +

d
dx
x
n
n
n
n
0
2 1
1
2 1
2
!
= + ( )
( )
+ ( )

=

d
dx
n
x
n
n
n
n
0
2
2 1
1
2 1
2
!
=
( )
( )

=

d
dx
x
n
n
n
n
0
2
1
2
2
!
Note: 2 1 2 1 2 n n n + ( ) = + ( )( ) ! !
= + + 1
2 4 6
2 4 6
2 2
x x x
! ! !

The power series for sinx converges for all x, so the same is true for this
power series for cos x .
Therefore,
cos
! ! ! !
x
x x x x
n
n
n
= + + + ( )
( )
+ 1
2 4 6
1
2
2 4 6 2
2 2 2
 
for x ∈.
A further benefcial property of power series is that algebraic
manipulations valid for polynomials also work for power series. These are
described below.
It is important to note that the
operations for power series
stated here can change the
interval of convergence for the
resulting series. In general, the
interval of convergence of the
resulting series for properties
1 and 2 will be the same as the
original series; and for property
3 it will be the intersection of
the intervals of convergence of
the two original series.
For convenience, we state the
operations with power series
using series centred at 0, but
they also apply for series
centred at c, e.g. a c
n
n
n
=

( )
0
x 2
Properties of power series
Given the power series f a
n
n
n
x x ( ) =
=

0
and g b
n
n
n
x x ( ) =
=

0
and constant k, the following
hold true:
1 f k a k
n
n
n n
x x ( ) =
=

0

2 f a a
k
n
n
k
n
n
n
kn
x x x ( ) = ( ) =
=

=

∑ ∑
0 0

3 f g a b
n n
n
n
x x x ( ) ± ( ) = ± ( )
=

0
76
Power Series
3
Example 7
a) Use the fact that the Maclaurin series for f x
x
( ) =
1
12
is x
n
n=

0
for
21 1 < < x to fnd the Maclaurin series for g x
x
( ) =
+
1
1
.
b) Use the result from a) and integration of power series to fnd the
Maclaurin series for h x x ( ) = + ( ) ln 1 .
Solution
a) We can create the power series for g x
x
( ) =
+
1
1
by simply substituting
2x x for into the power series for f x
x
( ) =
1
12
. Thus,
g x f x x x
n
n
n
n
n
( ) = ( ) = ( ) = ( )
=

=

∑ ∑
2 2 2
0 0
1
and we can write this result as

1
1
1 1
2 3
+
= + + + ( ) +
x
x x x x
n
n
2 2 2
 
for 21 1 < < x .
Note: 21
0
( )
=

n
n
n
x diverges at both endpoints so the interval of
convergence is the same as for the original series x
n
n=

0
.
b) Knowing that
1
1
1
+
= + ( )

x
dx x ln , we can integrate the power series
for
1
1 + x
to obtain the power series for ln 1 + ( ) x .

ln 1
1
1
1
0
+ ( ) =
+
= ( )

 ∫

=

x
x
dx x dx
n
n
n
2
= ( )
( )

=

21
0
n
n
n
x dx
= ( )
+

+
=

21
1
1
1
0
n
n
n
n
x
= ( )
=

2
2
1
1
1
n
n
n
x
n
Starting the index at n = 1 rather than n = 0.
Note: At x = 21, 2
2
1
1
1
( )
=

n
n
n
x
n
becomes the divergent harmonic series
and at x = 1 it becomes the convergent alternating harmonic series, so
the interval of convergence changes to Ϫ ഛ 1 1 < x .
Therefore, we can write the result as
ln 1
2 3 4
1
2 3 4
1
+ ( ) = + + ( ) + x x
x x x x
n
n
n
2 2 2
2
 
for Ϫ ഛ 1 1 < x .
77
Knowing that
1
1
2
+
=

x
dx x arctan , we can use the same approach
used in Example 7 b) to show that the Maclaurin series for arctanx is
arctanx x
x x x x
n
n
n
= + + + ( )
+
+
+
2 2 2
3 5 7 2 1
3 5 7
1
2 1
 
for Ϫ ഛ ഛ 1 1 x .
We now have derived Maclaurin series for several important functions. We
list these below with their corresponding intervals of convergence. These
series can be used to construct other series.
The last fve of the Maclaurin series
listed here are the power series given
in the Mathematics HL Information
(formulae) booklet.
Selected Maclaurin series
Function Interval of convergence
1
1
1
0
2 3
2x
x x x x x = = + + + + + +
=

n
n
n
 

21 1 < < x
e
n
n
n
x
x
x
x x
= = + + + +
=

! ! !
0
2 3
1
2 3

2∞ < < ∞ x
ln 1 1
2 3 4
1
1
2 3 4
+ ( ) = ( ) = + +
=

x
x
x
x x x
2 2 2
2 n
n
n
n

Ϫ ഛ 1 1 < x
sin
! ! ! !
x
x
x
x x x
= ( )
+ ( )
= + +
+
=

2 2 2 1
2 1 3 5 7
2 1
0
3 5 7
n
n
n
n

2∞ < < ∞ x
cos
! ! ! !
x
x x x x
= ( )
( )
= + +
=

2 2 2 1
2
1
2 4 6
2
0
2 4 6
n
n
n
n

2∞ < < ∞ x
arctan x
x
x
x x x
= ( )
+
= + +
+
=

2 2 2 1
2 1 3 5 7
2 1
0
3 5 7
n
n
n
n

Ϫ ഛ ഛ 1 1 x
Example 8
a) Find the Maclaurin series for f x e
x
( ) =
2
.
b) Hence, fnd a series for e dx
x
2

.
c) Use the frst four terms of the series in b) to approximate the value
of e dx
x
2
0
1

.
Solution
a) We prefer not to derive a Taylor series (centred at 0, in this case) by the
direct method used to fnd the series for sinx in Example 5, if there
is an alternative approach. To fnd the Maclaurin series for e
x
2
we can
simply substitute x
2
in for x in the Maclaurin series for e
x
.
e
x
n
x
n
x
x x
x
n
n
n
n
2
2
0
2
0
2
4 6
1
2 3
=
( )
= = + + + +
=

=

∑ ∑
! ! ! !

for 2∞ < < ∞ x
This series converges for all real values of x because the original series
for e
x
did so.
78
Power Series
3
b) We can take the result from a) and integrate it term by term.
e dx x
x x x
n
dx
x
n
2
1
2 3
2
4 6 2
∫ ∫
= + + + + + +

! ! !
 
= + +

+

+ +
+ ( )
+
+
x
x x x x
n n
n 3 5 7 2 1
3 5 2 7 3 2 1 ! ! !
 
That is,
e dx
x
n n
x
n
n
2
2 1
0
2 1
=
+ ( )
+
=

!
and because the series for e
x
2
converged for all real values of x, then this
series does also.
c) The frst four terms of the series for e dx
x
2

are

x
n n
x
x x x
x
x x
n
n
2 1
0
3 3 5 7 3
2 1 3 5 2 7 3 3
+
=
+ ( )
= + +

+

= + +

! ! !
55 7
10 42
+
x
.
We use this to approximate the value of the defnite integral e dx
x
2
0
1

.
e dx x
x x x
x
2
0
1
3 5 7
0
1
3 10 42

≈ + + +

= + + + 1
1
3
1
10
1
42
= = ≈
51
35
1
16
35
1 457142857 . (to ten signifcant fgures)
Also, to ten signifcant fgures, a GDC computes the value of e dx
x
2
0
1

to
be 1.462 651 746.
The percentage difference between the value from the partial sum of the
series for e
x
2
and the calculator value is only about 0.377%.
Continuing with the idea of using power series for approximation
purposes, recall the discussion at the start of this chapter. As in the last
example, when a power series is used (by your GDC, for example) to
compute an approximate value for a function it does so with a suitable
partial sum of the power series that, by defnition, will be a polynomial.
Given the defnition of a Taylor series
f x
f c
n
x c
n
n
n
( ) =
( )
( )
( )
=

!
0
2
it follows that f x ( ) is the limit of the partial sums. In our derivation of
the power series for f x e
x
( ) = centred at 0 we found four partial sums that
were polynomials of degree 1, 2, 3 and 4.
We are unable to fnd a
function that is an anti-
derivative of e
x
2
with the
calculus techniques covered
in this course. Helping us to
integrate such functions is
a very useful application of
power series.
Taylor polynomials
In general the nth partial sum of a Taylor series is

P
f c
k
c
n
k
k
n
k
x x ( ) =
( )
( )
( )
=

!
0
2
which is a polynomial, typically called a Taylor polynomial, of degree n.
79
For instance, the result for Example 5 shows that the Taylor polynomials
for f x x ( ) = sin centred at 0 for n = 1 3 5 , and are:
P x x
1
( ) = P x x
x
3
3
3
( ) = 2
!
P x x
x x
5
3 5
3 5
( ) = + 2
! !
p p p p
x
1
1
y
2 2

In general, f x ( ) is the limit of the sequence of its Taylor polynomials. That
is,
f x P x
n
n
( ) = ( )
→∞
lim .
Because any particular Taylor polynomial, P x
n
( ), is an approximation to
f x ( ) we can write
f x P x R x
n n
( ) = ( ) + ( )
where R x
n
( ) is the remainder, also called the error term.
We are now in a position to present the following important theorem that
not only gives a formula for the Taylor polynomial for approximating a
function f but also two formulae for computing the error term.
Since the error term for a Taylor
polynomial results from ‘cutting
of’ the series after a certain
number of terms to a fnite
polynomial it is occasionally
referred to as the truncation
error.
Although the integral form is
capable of computing an exact
value for the error term R
n
(x)
it is often too cumbersome to
evaluate. The Lagrange form
cannot compute the error
exactly, however it is, generally,
more manageable. By fnding
the value of b between c and
x (inclusive) that maximizes
f
(n + 1)
(b), thereby maximizing
f b
n
c
n
n
+ ( )
+
( )
+ ( )
( )
1
1
1 !
x 2 , we can
compute the maximum error
from using an nth degree
Taylor polynomial centred
at c to approximate f(x) for a
specifc value of x.
Taylor’s theorem
If a function f has derivatives of all orders in an open interval I centred at c, then for each
positive integer n and for each x in I,

f P R
n n
x x x ( ) = ( ) + ( )
where

P
f c
k
c
n
k
k
n
k
x x ( ) =
( )
( )
( )
=

!
0
2
is the nth degree Taylor polynomial centred at x = c.
The error term, R
n
x ( ), can be computed with one of the following formulae:
R
f b
n
c
n
n
n
x x ( ) =
( )
+ ( )
( )
+ ( )
+
!
1
1
1
2 where b lies between c and x, inclusive (Lagrange form)
or
R
f t
n
t dt
n
n
n
a
x x
x
( ) =
( )
( )
+ ( )

1
!
2 (integral form).
Figure 3.7 The sine function
(in black) and the three Taylor
polynomials P
1
(x) = x,
P
3
3
3
x x
x
( ) = 2
!
and
P
5
3 5
3 5
x x
x x
( ) = + 2
! !
.
80
Power Series
3
Example 9
a) Find the ffth degree Taylor polynomial centred at x = 1 for f x x ( ) = ln .
b) Use this polynomial to approximate ln . 1 3 ( ).
c) Use the Lagrange error term to determine an upper bound to the error
in this approximation.
d) How many terms of the Taylor series centred at c = 1 are needed to
approximate ln . 1 3 ( ) so that the error is less than 0 000001 1 0 10
6
. . = ×
2
.
Solution
a) We can construct the Taylor series centred at x = 1 by substituting x 21
for x in the Maclaurin series for ln 1 + ( ) x .
Since ln 1 1
2 3 4
1
1
2 3 4
+ ( ) = ( ) = + +
=

x
x
n
x
x x x n
n
n
2 2 2
2

for Ϫ ഛ 1 1 < x ,
then
ln ln 1 1 1
1
1
1
1
1
+ ( ) ( ) = = ( )
( )
= ( )
(
=

x x
x
n
x
x
n
n
n
Ϫ Ϫ
Ϫ
Ϫ Ϫ
Ϫ
Ϫ
))
+
( ) ( )
+ <
2 3 4
2
1
3
1
4
0 2
x x
x
Ϫ
Ϫ
Ϫ

. for
Hence, the ffth degree Taylor polynomial centred at x = 1 is
x
x x x x
2 2
2 2
2
2 2
1
1
2
1
3
1
4
1
4
2 3 4 5
( )
( )
+
( ) ( )
+
( )
.
b) ln . .
. . . .
1 3 0 3
0 3
2
0 3
3
0 3
4
0 3
4
0
2 3 4 5
( ) ≈ ( )
( )
+
( ) ( )
+
( )
= 2 2 ..262461
c) For the Lagrange form of the error term we have x = 1 3 . , n = 5, c = 1. By
computing a few derivatives of f x x ( ) = ln
′ ( ) = f x
x
1
, ′′ ( ) = f x
x
2
1
2
, f x
x
3
3
2
( )
( ) = , f x
x
4
4
6
( )
( ) = 2
we can see that the nth derivative is f x
n
x
n
n
n
( )
+
( ) = ( )
( )
2
2
1
1
1
!
. Therefore,
the sixth derivative at x b = is f b
b
6
6
5
( )
( ) = 2
!
. We need to estimate its
largest magnitude, that is, 2
5 120
6 6
!
b b
= as b ranges from 1 to 1.3. Clearly
it has a maximum when b = 1. Therefore,
R
5
5 1
1 3
120
5 1
1 3 1
0 08748
720
0 000 121 .
!
.
.
. ( )
+ ( )
( ) = =
+
ഛ Ϫ 55.
The error is no larger than 0.000 1215. Given our approximation
ln . . 1 3 0 262461 ( ) ≈ , we can say with absolute certainty that the exact
value of ln . 1 3 ( ) is somewhere between 0.262 3395 and 0.262 5825.
d) The magnitude of the n + ( ) 1 st derivative is given by
f b
n
b
n
b
n
n
n n
+ ( )
+ +
( ) = ( ) =
1
1 1
1 2
! !
. As in the previous part, we have x = 1 3 .
81
Example 9
a) Find the ffth degree Taylor polynomial centred at for .
b) Use this polynomial to approximate .
c) Use the Lagrange error term to determine an upper bound to the error
in this approximation.
d) How many terms of the Taylor series centred at are needed to
approximate so that the error is less than .
Solution
a) We can construct the Taylor series centred at by substituting
for x in the Maclaurin series for .
Since for ,
then
Hence, the ffth degree Taylor polynomial centred at is

b)
c) For the Lagrange form of the error term we have , , . By
computing a few derivatives of
, , ,
we can see that the nth derivative is . Therefore,
the sixth derivative at is . We need to estimate its
largest magnitude, that is, as b ranges from 1 to 1.3. Clearly
it has a maximum when . Therefore,

The error is no larger than 0.000 1215. Given our approximation
, we can say with absolute certainty that the exact
value of is somewhere between 0.262 3395 and 0.262 5825.
d) The magnitude of the derivative is given by
. As in the previous part, we have
and c = 1, so 1 1 3   b . ; and the n + ( ) 1 st derivative will be largest
when b = 1. Therefore,
max R
n
n n
n
n
n
1 3
1
1 3 1
0 3
1
1
1
1
.
!
!
.
.
( )

=
+ ( )
( ) =
( )
+
<
+
+
2 .. . 0 10
6
×
2
By trial and error (see GDC images below), we determine that the
smallest value of n that satisfes this inequality is n = 9. So, the ninth
degree Taylor polynomial would give us an error of less than 1 0 10
6
. ×
2

when approximating ln . 1 3 ( ).
Example 10
a) Express the indefnite integral
sinx
x
dx

as an infnite series.
b) Evaluate the defnite integral
sinx
x
dx
0
1

accurate to three decimal
places.
Solution
a)
sin
!
x
x
dx dx
n
n
x
n
n
n
n
x
x
n
=
( )
= ( )
+
+ ( )

=

+
2
2
1
2 1
2 1
0
2
1
2 11
0
( )
=

!
n
dx
= ( )
( )
= ( )
( ) + +

∑ ∑
=

=
∞ +
2 2 1 1
2 2 1
2 1 2 1 2
0 0
n
n
n
n
x
n
x
n
dx
n n
! nn + ( ) 1 !
=

+
⋅ ⋅
+ x
x x x
2 2
3 5 7
3 3 5 5 7 7 ! ! !

b)
sin
!
x
x
dx
n
n
n
x
n n
0
1
0
0
1
1
2 1
2 1 2 1

= ( )
( )( )

=
∞ +
+ +
2
= ( )
( )( )
=

+ +
21
0
1
2 1 2 1
n
n
n n !
= + + 1
1
18
1
600
1
29400
2 2

We can use the alternating series estimation theorem that says that the
error generated in using the nth partial sum of an alternating series to
approximate its sum does not exceed the value of the (n + 1)st term.
Which is the frst term to have a magnitude so that adding/subtracting
it will not change the third decimal point of the sum?
a
3
1
600
0 0016 = = . , a
4
1
29400
0 000034 = ≈ . . Hence, the frst three terms
will give us an estimate with a maximum error of about 0.000 034,
suffcient to guarantee an estimate accurate to three decimal places.
Y1(7)
8.20125E-6
Y1(8)
2.187E-6
Y1(9)
5.9049E-7
Plot1
Y1= ((.3)^(X+1))
(X+1)
Plot2 Plot3
Y2=
Y3=
Y4=
Y5=
Y6=
82
Power Series
3
Therefore,

sin
.
x
x
dx
0
1
1
1
18
1
600
0 946

≈ + ≈ 2
A GDC confrms our result.
fnInt(sin(X)/X,X
,0,1)
.9460830704
In questions 1–14, determine the radius of convergence and interval of convergence
of the power series.
1
x
n
n
n +
=

1
0
2 n
n
n
x 22
1
( )
=

3
x 23
1
( )
=

n
n
n
4 2
2
2
1
2 1
2 1
1
( )
( )
=

n
n
n
n
x
!
5
x
n
n
n
2
0 =

6
x 22
2
1
( )
=

n
n
n
7 2 2 1 1
0
( ) ( )
=

n n
n
x 8
lnn
n
n
n
=

2
x
9 n
n
n
! x
=

0
10
3
4
1
n n
n
n
n
x
=

11
n
n
n
n
!
!
( )
( )
=

2
1
2
x 12
n
n
n
n
x
3
0 =

13
n
n
n
n
n
! x
=

1
14
2 2 1 4
3
0
( ) ( )
=

n n
n
n
n! x
15 Given k > 0, express the interval of convergence of k
n
n
x ( )
=

0
in terms of k.
16 a Find a power series for
1
1+ x
by replacing x with 2x in the geometric series
for
1
12x
(Example 1). Also determine the interval of convergence for the
series.
b Use the technique of partial fraction decomposition to fnd A and B so that

1
1 1 1
2
2 2 x x x
=
+
+
A B
.
c Hence, construct a power series for
1
1
2
2x
and determine its interval of
convergence.
17 a Find the Maclaurin series for e
2x
2
and determine its radius of convergence.
b Express e d

x
x
2
as a Maclaurin series and determine its radius of
convergence.
c Use the frst fve terms of the series from b to estimate e d

x
x
2
0
1
. Show that
the error for this estimate is less than 0.001.
18 Use multiplication or division of power series to fnd the frst three non-zero
terms in the Maclaurin series for each function.
a f x x x ( ) = sin
b g x x ( ) = tan Hint: tan
sin
cos
x
x
x
=

c f e x x
x
( ) = ( ) ln 12
19 Find a Maclaurin series for
1
1
2
2x ( )
by diferentiating the Maclaurin series for

1
12x
. State the interval of convergence for the series.
Exercise 3.1
83
20 a Find a power series for x
2
e
2x
.
b By diferentiating term by term the power series in a, show that

22
2
4
1
( )
+
=
+

n
n
n!
.
21 a Write down the seventh degree Taylor polynomial for sin x.
b Use this polynomial to estimate sin
π
12

.
c Use the Lagrange error term to determine an upper bound to the error in this
approximation.
22 Determine all the values of x for which the series
2
1
1
n n
n
n
x
ln + ( )
=

converges.
23 Find the frst four non-zero terms of the Taylor series centred at x 5 1 for the
function f(x) 5 (x 2 1)e
x
.
24 Find the Maclaurin series for ln
1
1
+

x
x 2
.
25 a Find the Maclaurin series for the function f x
x
( ) =
+
1
1
2
.
b Hence by integrating each term, show that

arctan . x x
x x x x
= + + + − ( )
+
+
+
2 2
3 5 7 2 1
3 5 7
1
2 1
 
n
n
n
c Show that this series converges for −1 1   x , and thus show that

21
2 1 4
0
( )
+
=
=

n
n
n
π
.
d Use the frst six terms of the series to estimate the value of p. Compute the
maximum error for this estimate.
26 The function f is defned by f
e e
x
x x
( ) =
+
2
2
.
a Obtain an expression for f
(n)
(x), the nth derivative of f(x) with respect to x.
b Hence, derive the Maclaurin series for f(x) up to and including the term in x
4
.
c Use the result to fnd a rational approximation to f
1
2

.
d Use the Lagrange error term to determine an upper bound to the error in this
approximation.
27 Estimate the range of values of x for which the Maclaurin approximation

sin
! !
x x
x x
≈ + 2
3 5
3 5
is accurate to within 0.005.
28 Find a power series for x e
x
and then integrate the resulting series term by term
from 0 to 1 to show that
1
2
1
2
1
n n
n
! + ( )
=
=

.
29 Find a power series for sec
2
x. (Hint: consider the derivative of tan x.)
30 Compute the Taylor series for each of the following functions f(x) centred at the
given point c.
a f e c x
x
( ) = = , 2

b f c x x ( ) = ( ) = sin ,
3
0
c f c x
x
( ) =
( )
=
1
1
0
3
2
,

d f c x x x ( ) = ( ) = 21 1
3
ln ,
31 Use series to evaluate each limit.
a lim
cos
x
x
x
x
→0
1
1
2
2 2 e

b lim
arctan
x
x x
x
→0
3
2
84
Power Series
3
1 Find the fourth degree Taylor polynomial centred at x = 0 for the function

f x x x ( ) = ( ) < ln cos , . 0
2

π
2 a Find the Maclaurin series for sin
2
x up to the term containing x
4
.
b Hence, write down a series for cos
2
x up to the term containing x
4
.
3 Find the frst three non-zero terms in the Maclaurin expansion of e
x
sin

x.
4 Find the frst four terms of the Maclaurin series for the function f(x) = e
3x
.
5 Find the Maclaurin series of sec

x up to the term containing x
4
.
6 a Write down the fourth degree Taylor polynomial centred at x = 0 for e
x
.
b Determine the fourth degree Taylor polynomial centred at x = 0 for e
x
2
.
c Hence, or otherwise, fnd the fourth degree Taylor polynomial centred at
x = 0 for e
x + x
2
.
7 Find the Maclaurin series for In(2 + 3x) and fnd an expression for the error
term R
n
(x).
8 a Find the fourth degree Taylor polynomial centred at x = 0 for the function

f x x ( ) = + 4 .
b Find an expression for the error term R
4
(x), and fnd an error bound when
x = 0.1.
9 Using the formula for the Lagrange form of the error term, determine how
many terms of the Maclaurin series for cos x are needed to approximate
cos(5°) correct to six decimal places. State this approximate value.
10 a Find the Maclaurin series for e
2x
2
.
b Use the frst three terms of this series to approximate e d
2x
x
2
0
1

.
c Find an upper bound for the error in this approximation.
11 a Find the Maclaurin series for
1
1
2
+ x
. Express the result as an equation in
the form
1
1
2
1
+
=
=

x
a
n
n
where a
n
is in terms of x and n.
b By integrating both sides of the equation found in a, show that

arctan x
x
= ( )
+

=

2
2
1
2 1
1
2 1
1
n
n
n
n
.
c Show that this series converges for Ϫ ഛ ഛ 1 1 x .
d Hence, fnd the exact value to which the series 1
1
3
1
5
1
7
2 2 + +

converges.
12 a Find the frst four terms in the Maclaurin series for both of the following
functions:
f x
x
( ) =
+
1
1
and g x
x
( ) =
1
12
b Rewrite the function h x
x
x x
( ) =
+
+
1
5 6
2
2
as the sum of two fractions.
c Hence, fnd the frst four terms in the Maclaurin series for the function
h x
x
x x
( ) =
+
+
1
5 6
2
2
.
Practice questions
Practice questions 1–70 cover
series work from Chapters 1–3
inclusive.
85
13 a Write down the function represented by the power series x
n
n
21
1 =

where
the interval of convergence is 21 1 < < x .
b Using the result from a and the identity
1 1
1 1 x x
=
+ ( )
2
2
, fnd the Taylor
series centred at x = 21 for the function f x
x
( ) =
1
. State the interval of
convergence.
14 Test the convergence or divergence of the following infnite series, indicating
the tests used to arrive at your conclusion.
a
k
k
k
=

+
1
1
3
b
k
k k
=

2
3
1
1 ( ) n
c ( ) 21
1
1
1
2
k
k
k
k
+
=

+
15 Express
3 4
2
x
x x
2
2
in partial fractions.
16 Discuss the convergence or divergence of the following series:
a
n
n
n
n
=
∞ +

0
2 1
2
3
b
n
k
k
n
n k
=

+
0
1 2
, k is a positive integer.
17 a Consider the function f (x) = 2
2
2x . . Show that

′′ = f ( )
( )
. x
x x
2
2 2
2 2
2 2
b The third and fourth derivatives of f (x) with respect to x are given by:
′′′ = f ( )
( )
x
x
x x
2
2 2
6
2 2
2 2 2
and f
( )
( )
( )
( )
4
2
2 3 2
12 2 1
2 2
x
x
x x
=
+
2 2
.
Use Maclaurin series to expand f (x), and to show that
2 2
1
4
2
2
4
2
2
x
x
x

+

 k
. Find the value of k.
c Hence, or otherwise, show that the value of
0
1
2 2
2

x x x 2 d is
approximately 0.394 38.
d Use the substitution x = 2 sin θ to show that the exact value of

0
1
2 2
2
8

x x x 2 d is
π
. Hence, fnd an approximation of p to four decimal
places.
18 a Find the Maclaurin series of the function g(x) = sin (x
2
) using the series
expansion of sin x, that is, sin ( )
( )!
. x
x
=
+
=
∞ +

21
2 1
0
2 1
n
n
n
n
b Using the Maclaurin series of g x x ( ) = ( ) sin
2
, evaluate the defnite integral

sin x x
2
0
1
( )

d
correct to four decimal places.
86
Power Series
3
19 a Use the ratio test to calculate the radius of convergence of the power
series
n
n
n
=

1
3
2
5 ( )
.
x 2
b Using your result from part a, determine all points x where the power
series given in a converges.
20 Test whether the following is a convergent series:

n
n
n
=

+

1
1
1
1
( )
!
. 2
If it is, then fnd an approximation for the sum to two decimal places; if it is not,
explain why this is so.
21 a Find a Maclaurin series expansion for f (x) = ln(1 + x), for 0 < x < 1.
b R
n
is the error term in approximating f (x) by taking the sum of the frst
(n + 1) terms of its Maclaurin series. Prove

R
n
n
 
1
1
0 1
+
< , ( ). x
22 Test the convergence or divergence of the following series:
a sin ;
1
1
n
n=

b ( )
cos
n
n
n
n
+

=

10
1
1 4
π
.
23 Consider the function f ( ) , x x
x
=
1
where x ∈ ℝ
+
.
a Show that the derivative ′ =

f f ( ) ( )
ln
. x x
x
x
1
2
2
b Sketch the function f (x), showing clearly the local maximum of the
function and its horizontal asymptote. You may use the fact that

lim
ln
.
x→∞
=
x
x
0
c Find the Taylor expansion of f (x) about x = e, up to the second degree term,
and show that this polynomial has the same maximum value as f (x) itself.
24 Determine whether the series
1
1
1
1
n
n
n
+
=

converges.
25 Consider the sequence of partial sums {S
n
} given by S
k
n
n
k
n
= = …
=

1
1 2
1
, , , .
a Show that for all positive integers n, S
2n
> S
n
+
1

__

2
.
b Hence, prove that the sequence {S
n
} is not convergent.
26 For positive integers k and n let u
k
S u
k
k
n k
k
n
=
+
+
=
=

1 2 1
1
2
1
2
( )
.
2
and
a Show that S
k
k k
n
k
n
2
1
4 1
2 2 1
=
+
=

2
( )
.
b Hence, or otherwise, determine whether the series u
k
k =

1
is convergent or
27 a Describe how the integral test is used to show that a series is convergent.
Clearly state all the necessary conditions.
b Test the series
n
e
n
n
2
1 =

for convergence.
87
28 a Find the frst four non-zero terms of the Maclaurin series for
i sin x ii e
x
2
.
b Hence, fnd the Maclaurin series for e
x
2
sin x, up to the term containing x
5
.
c Use the result of part b to fnd lim
sin
.
x
x
x x
x

0 3
2
e 2
29 Find the Maclaurin series of the function f (x) = ln(1 + sin x) up to and including
the term in x
4
.
30 Find the range of values of x for which the following series is convergent.

x
n
n
n +
=

1
0
31 a Show that the series sin
2
2
1
π
n
n=

is convergent.

Let sin . S
n
n
=
=

1
2

b Show that for positive integers n > 2,

1 1
1
1
2
n n n
<
2
2 .
c Hence, or otherwise, show that 1 < S < 2p.
32 Calculate
lim
sin
.
x
x x

0
1 1
2
33 Use the integral test to show that the series
1
1
n
p
n=

is convergent for p > 1.
34 a i Find the frst four derivatives with respect to x of y = ln(1 + sin x).
ii Hence, show that the Maclaurin series, up to the term in x
4
, for y is
y x x x x = + +… 2 2
1
2
1
6
1
12
2 3 4
.
b Deduce the Maclaurin series, up to and including the term in x
4
, for
i y = ln(1 2 sin x)
ii y = ln cos x
iii y = tan x.
c Hence, calculate lim
tan( )
ln cos
.
x
x
x

0
2
35 Find the radius of convergence of the series
( )!
!( )!
.
2 2
1
1
k
k k
k
k
2
2
x
=

36 a i Prove that the alternating series given by ( )
( )!
2
2
2
1
1
2 1
1
1
n
n
n
=

converges.
ii Approximate the series by fnding the fourth partial sum. Give your
iii What is the upper bound for the error in this approximation?
b i Find the frst four non-zero terms of the Maclaurin series for sin x.
ii Deduce the nth term of this series.
iii Use the ratio test to show that the series is convergent for all values of x.
iii Use your series for sin x to fnd the frst four non-zero terms of the
Maclaurin series for cos x.
88
Power Series
3
37 Find the value of
a lim
x
x
x

+
0
1 1 2
b lim ln .
x
x x
→0
38 a i Given that u
n
n=

1
is convergent, where u
n
> 0, prove that u
n
n
2
1 =

is also
convergent.
ii State, with a reason, whether or not the converse of this result is true.
b Use the integral test to determine the set of values of k for which the series

1
2
n n
k
n
(ln )
=

i is convergent
ii is divergent.
39 Consider the function f defned by
f (x) = arcsin x, for |x | < 1.
The derivatives of f (x) satisfy the equation
(1 2 x
2
) f
(n + 2)
(x) 2 (2n + 1) x f
(n + 1)
(x) 2 n
2
f
(n)
(x) = 0, for n > 1.
The coefcient of x
n
in the Maclaurin series for f (x) is denoted by a
n
. You may
assume that the series contains only odd powers of x.
a i Show that, for n > 1, (n + 1) (n + 2) a
n + 2
= n
2
a
n
.
ii Given that a
1
= 1, fnd an expression for a
n
in terms of n, valid for odd
n > 3.
b Find the radius of convergence of this Maclaurin series.
c Find an approximate value for p by putting x =
1

_

2
and summing the frst
three non-zero terms of this series. Give your answer to four signifcant
fgures.
40 The general term of a sequence is given by the formula a
n
=
n n
n
2
2
3
2 1
+
2
, n ∈ ℤ
+
.
a Given that lim
n
n
a L
→∞
= , where L ∈ ℝ, fnd the value of L.
b Find the smallest value of N ∈ ℤ
+
such that a L
n
2 < 10
2 3
for all n > N.
41 Estimate the range of values of x for which the Maclaurin approximation

sin
! !
x x
x x
≈ + 2
3 5
3 5
is accurate to within 0.005.
42 Determine whether the series ( ) n
n
n
+
=

1
0
2
is convergent or divergent.
43 Consider the series S =
1
1
n
n
!
.
=

a Use the ratio test to prove that this series is convergent.
b Use a comparison test to show that S < 2.
c Write down the exact value of S.
89
44 a Show that the polynomial approximation for ln x in the interval [0.5, 1.5]
obtained by taking the frst three non-zero terms of the Taylor series about
x = 1 is given by
ln . x
x x
x ≈ +
3 2
3
3
2
3
11
6
2 2
b Given ∫In x dx = x ln x 2 x + C, show by integrating the above series that
another approximation to ln x is given by
ln . x
x x x
x
≈ +
3 2
12 2
3
2
5
6
1
4
2 2 2
c Which is the better approximation when x = 1.5?
45 a i Find I
d
d
n
n
n
=

x x
x
x
+ 1
2
2
α
where α is a positive constant and n is a positive
integer.
ii Determine lim .
n
I
n
→∞
b Using l’Hôpital’s rule fnd

lim
tan tan
sin sin
x
x x
x x
→∞

β β
β β
2
2
where β is a non-zero constant diferent from ±1.
46 a Use l’Hôpital’s rule to fnd
i lim
ln
x
x
x
→1
2
1 2

ii lim
cos
.
tan
x
x
x
→0
2
12
b Giving a reason, state whether the following argument is correct or
incorrect.
“Using l’Hôpital’s rule, lim lim .
x x
x
x x
→ →
= =
1
2
3
3
3
1
2
1
6
2
2

47 Given that the Maclaurin series for e
sin x
is a + bx + cx
2
+ dx
3
+ … , fnd the
values of a, b, c, and d.
48 Consider the infnite series
1
2
1
n n
n
( )
.
+
=

a Show that the series is convergent.
b i Express
1
2 n n ( ) +
in partial fractions.
ii Hence, fnd
1
2
1
n n
n
( ) +
=

∑ .
49 Find the interval of convergence of the series sin .
π
n
n
n

=

x
1
50 Find:
a
lim
s
s s
s

+ 

4
3 4
4
2
2
b lim
cos
sin
x
x
x x

0
1 2
51 a Find constants A and B such that
1
16 8 3 4 1 4 3
2
r r
A
r
B
r +
= +
+ 2 2
.
b Find an expression for S
n
, the nth partial sum of
1
16 8 3
2
1
r r
r
+

=

2
.
c Hence, show that the series converges.
52 Find the values of p for which
1
1
x
p

dx converges.
90
Power Series
3
53 Calculate the following limits:
a lim
x
x
x
→0
2 1 2
b lim
ln( )
x
x
x x

+
( )
+
0
2
3
2
1
1
1
2
2
54 a The function f is defned by f (x) =
e e
x x
+
2
2
.
i Obtain an expression for f
(n)
(x), the nth derivative of f (x) with respect
to x.
ii Hence, derive the Maclaurin series for f (x) up to and including the term
in x
4
.
iii Use your result to fnd a rational approximation to f
1
2

.
iv Use the Lagrange error term to determine an upper bound to the error
in this approximation.
b Use the integral test to determine whether the series
ln n
n
n
2
1 =

is convergent
or divergent.
55 a i State the domain and range of the function f (x) = arcsin (x).
ii Determine the frst two non-zero terms in the Maclaurin series for f (x).
b Use the small angle approximation
cos( ) y
y y
≈ + 1
2 24
2 4
2
to fnd a series for cos (arcsin (x)) up to and including the term in x
4
.
c i Find the Maclaurin series for (p + qx
2
)
r
up to and including the term in
x
4
where p, q, r ∈ ℝ.
ii Find values of p, q and r such that your series in c i is identical to your
answer to b. Comment on this result.
56 a Determine lim
ln( )
ln( )
x
x
x

+ 

0
2 2
3
a
a2
, where a is a positive constant, not equal to 1.
b Calculate lim
ln( )
ln( )
.
x
x
x

+ 

0
2
2
1
12
c Show that lim
cos
cos
x
x x
x
x x

+
+

0
2
2 2
2
2
2
2
e e
= 1.
57 a i By considering the graph of y =
1

__

x
between x = a and x = a + 1, show
that

1
1
1 1
2
1 1
1 a
a
a a a +
<
+

< +
+

ln , for all a > 0.
ii Deduce that

1
2
2
3
4
< < ln .
iii Using i fnd p, q ∈ ℚ such that p < ln 3 < q.
b Given that H
n
is the nth partial sum of the harmonic series, show that
H n H
n
n n
2 2 2 1
1
2
1
2
< < ln ,

for n > 1.
c If y
n
= H
n
2 ln n, prove that the terms of the sequence {y
n
: n > 1} decrease
as n increases.
91
58 Determine whether the series
n
n
n
10
1
10
=

is convergent or divergent.
59
The diagram shows part of the graph of y =
1
3
x
together with line segments
parallel to the coordinate axes.
a Using the diagram, show that

1
4
1
5
1
6
1 1
3
1
4
1
5
3 3 3 3
3
3 3 3
+ + +…< < + + +…

x
x d .
b Hence, fnd upper and lower bounds for
1
3
1
n
n
.
=

60 The function f is defned by f x
x
( ) =

ln .
1
12
a Write down the value of the constant term in the Maclaurin series for f (x).
b Find the frst three derivatives of f (x) and hence show that the Maclaurin
series for f (x) up to and including the x
3
term is x
x x
+ +
2 3
2 3
.
c Use this series to fnd an approximate value for ln 2.
d Use the Lagrange form of the remainder to fnd an upper bound for the
error in this approximation.
e How good is this upper bound as an estimate for the actual error?
61 a Find the value of lim
ln
sin
.
x
x
x

1

b By using the series expansions for e
x
2
and cos x evaluate lim
cos
.
x
x
x

1
1
1
2
2
2
e
62 a Given that y = ln cos x, show that the frst two non-zero terms of the
Maclaurin series for y are 2 2
x x
2 4
2 12
.
b Use this series to fnd an approximation in terms of p for ln 2.
63 a Find the radius of convergence of the series
( )
( )
.
21
1 3
0
n n
n
n
n
x
+
=

b Determine whether the series n n
n
3 3
0
1 +
( )
=

∑ 2 is convergent or divergent.
64 The function f is defned by f (x) = In(1 + sin x).
a Show that ′′ =
+
f ( )
sin
. x
x
21
1
b Determine the Maclaurin series for f (x) as far as the term in x
4
.
c Deduce the Maclaurin series for ln(1 2 sin x) as far as the term in x
4
.
d By combining your two series, show that ln sec . x
x x
= + +…
2 4
2 12
e Hence, or otherwise, fnd lim
ln sec
.
x
x
x x
→0
0 3 4 5 6
y
x
92
Power Series
3
65 Find the coefcient of x
8
in the Maclaurin expansion of x x sin2 .
66 Write down the conditions for the integral test to be applied to the series
a
n
n =

1
and use it to determine if the series ne
n
n
2
2
1 =

converges.
67 Use Taylor’s series to express the polynomial p x x x x x ( ) = + + + +
4 3 2
3 2 1 in
terms of powers of x 21 ( ).
68 Find the limits, if they exist, of the following sequences u
n
{ } as n tends to
infnity.
a u
n
n
n
=
( ) sin
b u n
n
= ( ) arctan
c u
n
n
n
=
+
2 1
5 11
2
2
2
d u
e
n
n
n
=
!
e u ne
n
n n
= ( ) 2
2
1 f u n
n n
n

j
(
,
\
,
(
j
(
,
\
,
(
,
¸
,
]
]
]
2
1 1
2 cos sin 2 2
69 State, with reasons, if the following series converge or diverge.
a 1
1
1 1
1
1 21
1
1 331
1
1 4641
+ + + + +
. . . .

b e e e e + + + +
1
2
1
3
1
4

c
1
1
2
2
3
3
4
4
3 3 3 3
! ! ! !
+ + + +

d
1
1 3
2
3 5
3
5 7
4
7 9 ×
+
×
+
×
+
×
+

e
3
1
3
2
3
3
3
4
3 3 3 3
+ + + +

70 a Find the frst three terms of the Taylor series centred at x =
1

_

2
for the
function f x x ( ) = ( ) sin π .
b Hence, fnd an approximate value to sin
π π
2 8
+

, correct to three signifcant
fgures.
Questions 14–70 © International Baccalaureate Organization
93
Here are some examples:
1 x
dy
dx
y
dy
dx
y + = 2 0 frst order differential equation F x y
dy
dx
, ,

= 0
2
dy
dx
y y
x
+ =
2
2
0
2
frst order differential equation F x y
dy
dx
, ,

= 0
3
d y
dx
dy
dx
y
2
2
3 5 0 + = 2 second order differential equation F x y
dy
dx
d y
dx
, , ,
2
2
0

=
4
dy
dx
y x e
x
+ = sin
cos
2 0 frst order differential equation F x y
dy
dx
, ,

= 0
5 2 4 0
3
3
2
d y
dx
x
dy
dx
xy + ( )

+ = ln third order differential equation
F x y
dy
dx
d y
dx
d y
dx
, , , ,
2
2
3
3
0

=
4.1 Introduction
Equations involving an unknown function and its derivative(s) are called
differential equations and frequently occur in mathematical models of real-
life phenomena. Differential equations come in a great variety of forms, and
many different procedures – analytic, graphical and numerical – exist for
fnding their solutions. In the core syllabus (Section 16.8) we studied one
analytic solution method for a certain class of differential equation. We will
review that technique and some other important concepts from the core
syllabus. Then we will explore further into the topic by considering two
further classes of differential equations. Analytic methods are not always
successful in solving a differential equation, so we will also investigate a
graphical approach and a useful numerical method for approximating the
solution to a differential equation.
Differential
Equations
4
We were given an introduction
to diferential equations in the
core syllabus in Chapter 16.
You are strongly encouraged
again before working through
this chapter.
A diferential equation is an equation that relates an independent variable
(commonly x or t), a dependent variable (usually y), and one or more derivatives of an
unknown function y = f(x) [or y = f(t)]. The general form of a diferential equation (with
independent variable x) can be written as

F
d
d
d
d
d
d
k
k
x y
y
x
y
x
y
x
, , , , ,
2
2
0 

=
where the largest k for which
d
d
k
k
y
x
occurs in the equation is called the order of the
diferential equation.
94
Differential Equations
4
For this course, we only study frst order differential equations, such as
equations 1, 2 and 4 above. In a frst order differential equation, the frst
derivative,
dy
dx
, of the unknown function can be isolated on one side of
the equation. Hence, a simpler general form for frst order differential
equations is

dy
dx
F x y = ( ) ,
where
dy
dx
is expressed as a function in terms of x and y. Note that the frst
order differential equations 1, 2 and 4 can all be re-written in this form.
For example,
1. x
dy
dx
y
dy
dx
y
dy
dx
y
x y
+ = ⇒ =
+
2 0
The solution of a differential equation is the (initially unknown) function
y f x = ( ) whose derivative is
d
d
y
x
. Consider the differential equation

dy
dx x
x =
+

1
1
1 , . 2
Every solution of this equation is an anti-derivative of
1
1 x +
.
y
x
dx x C x =
+
= + + ≠

1
1
1 1 ln , 2
So the solution of the differential equation
dy
dx x
=
+
1
1
is the explicitly
defned function y x C = + + ln 1 where C is an arbitrary constant. This is
called a general solution because it is not a single function, but an infnite
‘family’ of functions dependent on the constant C. Figure 4.1 shows a few
members of this family.
A diferential equation may use
symbols for the independent
and dependent variables other
than x and y. For the sake of
simplicity, we will use x and
y while we are developing
theory and solution methods
for diferential equations. Also
note that we are using F (‘large
F’) to represent a two-variable
function that when set equal to
d
d
y
x
is the diferential equation,
and f (‘small f’) represents the
unknown function whose
slope at the point x y , ( ) is
d
d
y
x
.
10
8
6
4
2

2
4
6
8
10
12
6 4 2 2 4 6
x
y
C = 10
C = 5
C = 0
C = 5
C = 10
y = 1n | x + 1 | + C
Figure 4.1
95
In contrast, when we are given some initial conditions that allow us to
evaluate a particular value for C we obtain a single function that we call
a particular solution of the differential equation. For example, if we are
given the initial conditions that y = 5 when x = 0 then we can solve for C,
giving C = 5 and the particular solution of y x = + + ln 1 5.
Sometimes the solution of a differential equation will be expressed as an
implicitly defned function. For example, the general solution to equation
1 is
ln . y
x
y
C = +
It is an equation relating x and y and implies a function exists that defnes y
as a function of x.
To verify that this is a solution to 1, we differentiate – applying implicit
differentiation and the product rule:

d
dx
y
d
dx
x
y
C ln ( ) = +

1
1
y
dy
dx
d
dx
xy
d
dx
C =
( )
+ ( )
2

1
0
1 2
y
dy
dx
y x y
dy
dx
= + −

+
2 2

1 1
2
y
dy
dx y
x
y
dy
dx
= 2
y
y
dy
dx
y
y
x
y
dy
dx
2 2
2
1 1 

=

2
x
dy
dx
y
dy
dx
y + = 2 0
Therefore, for any real number C the function ln y
x
y
C = + is a solution,
in implicit form, to the differential equation x
dy
dx
y
dy
dx
y + = 2 0. This
means that the coordinates x and y of any point on the curve ln y
x
y
C = +
combined with the value of the derivative
dy
dx
at that point will solve the
equation x
dy
dx
y
dy
dx
y + = 2 0.
The only type of frst order differential equation covered in the core
syllabus of this course is a class of differential equations referred to as
separable equations. We solved these using a technique called separation
of variables. One of our key goals in this chapter is to develop an analytic
solution method for each of two further classes of frst order differential
equations. Before we delve into the details of these analytic methods, we
examine a useful graphical method for helping us to sketch the function, or
family of functions, that solves a differential equation.
In general, we wish to fnd
the explicit solution of a
diferential equation written
in the form y = f(x) where f is
a known function. However,
it is sometimes not possible
to solve for y. In such a case
we must settle for an implicit
solution written in the form
g(y) = f(x) where g and f are
known functions and
g(y) ≠ y.
96
Differential Equations
4
4.2 Slope felds
Often the primary objective when solving a frst order differential equation
is to fnd an explicit solution. However, many differential equations
used in mathematical models cannot be solved by means of an analytic
method. For such equations, we must resort to graphical and/or numerical
methods. Carried out by hand or by technology, a graphical method
provides us with rough qualitative information about the graph of a
solution to a differential equation.
A frst order differential equation in the form
dy
dx
F x y = ( ) , specifes the
slope of the solution curve y f x = ( ) at each point in the xy-plane where
F is defned. We can use this fact to draw a short line segment whose slope
is F x y , ( ) at any point x y , ( ) in the plane. A plot of these line segments
showing the slope (or direction) of the solution curve is a called a slope
feld (or direction feld) for the frst order differential equation. As a rule,
the segments are drawn at representative points evenly spaced in both
directions. Figure 4.2 shows a slope feld for the equation
dy
dx
x y = 2 .
As you can imagine, it can be quite tedious to draw a slope feld by hand. In
practice, slope felds are easily generated by suitable graphing technology.
However, there is a method that simplifes the process of doing it by hand.
Rather than compute
dy
dx
for a large number of x and y values, we look for
points where
dy
dx
has the same value. For some constant c, the graph of the
equation F x y c , ( ) = is a line, called an isocline, along which all the short
line segments of a slope feld have the same slope c. For the differential
equation
dy
dx
x y = 2 , the isoclines are x y c 2 = . Figure 4.3, shows (in red)
Although it looks fairly simple,
the diferential equation
d
d
y
x
x y = 2 is not easy
to solve. It can be solved
analytically with one of the
techniques that we develop
later in the chapter. It is an
example of a frst order
linear diferential equation,
and its general solution is
y x
x
= + Ce
2
21.
6
5
4
3
2
1

1
2
3
4
5
6
6 5 4 3 2 1 0 1 2 3 4 5 6
x
y
Figure 4.2 Slope feld for
d
d
y
x
x y. = 2
97
the isoclines for c and = 21 0 3 ,

and 3. By frst tracing in a few isoclines, we can
create a slope feld by easily drawing multiple line segments along it all
having the same slope.
Solutions to a differential equation can be sketched by drawing in curves
that are at each point tangent to the line segment at that point. Thus, a
family of solution curves can be produced. To use a slope feld to sketch a
particular solution all we need to know is one point (an initial condition)
that the solution curve passes through.
Example 1
a) Draw a slope feld for
dy
dx
x
y
= 2 on the xy-plane such that Ϫ ഛ ഛ 5 5 x
and Ϫ ഛ ഛ 5 5 y . Sketch some sample solution curves. What shape are
they?
b) Confrm that both y c x =
2 2
2 and y c x = 2 2
2 2
, where c is a
constant, are each a general solution of the equation.
Solution
a) Rather than evaluating
dy
dx
x
y
= 2 for a large number of x and y values,
we establish some isoclines by looking for points where 2
x
y
has a
constant value.
If
dy
dx
x
y
= = 2 0 then x = 0. Hence, the y-axis is an isocline where all
the line segments are horizontal.
6
5
4
3
2
1

1
2
3
4
5
6
6 5 4 3 2 1 0 1 2 3 4 5 6
x
y
c = 1
c = 0
c = 3
‘Isocline’ comes from ‘iso-’
meaning equal and ‘-cline’
meaning slope. Be aware that
isoclines themselves do not
give any direct information
diferential equation. They
serve to ease the process of
drawing a slope feld. It is
recommended that you draw
isoclines lightly in pencil, and
preferably dashed.
Isoclines are not always
straight lines. Isoclines are
analogous to contour lines
on a map indicating land of
equal elevation. Consider the
diferential equation
d
d
y
x
x y =
2
2 that has
isoclines that are parabolas
with equations of the form
y x =
2
2c. When isocline
curves are not lines, it is more
difcult to use them to sketch a
slope feld.
Figure 4.3 Slope feld and three
isoclines for
d
d
y
x
x y = 2 .
98
Differential Equations
4
If y = 0 (x-axis), then
dy
dx
is undefned. Hence, the x-axis is an isocline
where all the line segments are vertical (undefned slope).
If
dy
dx
x
y
= = 2 1 then y x = 2 is an isocline where all the line segments
have a slope of 1.
If
dy
dx
x
y
= = 2 21 then y x = is an isocline where all the line segments
have a slope of 21.
If necessary, we can continue in this manner and establish further
isoclines, such as:
y x = 2 is an isocline where all the line segments have a slope of 2
1
2
.
y x = 2
1
2
is an isocline where all the line segments have a slope of 2.
In fact, any line passing through the origin will be an isocline for the
slope feld for
dy
dx
x
y
= 2 .
The resulting slope feld – showing six lightly drawn isoclines – is
shown below in Figure 4.4.
Drawing curves parallel to the line segments gives a family of solution
curves that appear to be circles. Three members of the family are drawn
in Figure 4.5.
b) Checking that y c x =
2 2
2 is a solution, we compute
dy
dx
on the left
side and substitute c x
2 2
2 for y on the right side.

5
4
3
2
1

1
2
3
4
5
5 4 3 2 1 0 1 2 3 4 5
x
y
y = x
y = 
y = 2x y = 2x y = x
x
2
y =
x
2
5
4
3
2
1

1
2
3
4
5
5 4 3 2 1 1 2 3 4 5
x
y
Figure 4.4 Figure 4.5
99

dy
dx
x
y
= 2

d
dx
c x
x
c x
2 2
2 2
2 2
2
( )
=

1
2
2
2 2
1
2
2 2
c x x
x
c x
2 2
2
( ) ( ) = −

− =
x
c x
x
c x
2 2 2 2
2
2
2
Q.E.D.
Checking that y c x = 2 2
2 2
is a solution.

dy
dx
x
y
= 2

d
dx
c x
x
c x
2 2 2
2 2
2 2
2 2
( )
=

2 2 2
2
2 1
2
2
2 2
1
2
2 2
c x x
x
c x
( ) ( ) =

x
c x
x
c x
2 2 2 2
2 2
= Q.E.D.
The solution y c x =
2 2
2 is the family of curves consisting of the
upper half of each circle, and the solution y c x = 2 2
2 2
is the family
of curves consisting of the lower half of each circle.
Example 2
A model for the velocity v, in metres per second, at time t seconds of a
75 kg skydiver falling from an aeroplane is given by the equation

dv
dt
v
= 10
360
2
2 .
5 10 15 20
t
100
90
80
70
60
50
40
30
20
10
0
v
Figure 4.6
100
Differential Equations
4
a) From the direction feld shown in Figure 4.6, sketch the solution curves
with the following initial conditions:
(i) v 0 0 ( ) = , (ii) v 0 35 ( ) = , and (iii) v 0 90 ( ) = .
b) Explain why the value v = 60 is called the terminal velocity for this
situation.
Solution
a) Solutions to
dv
dt
v
= 10
360
2
2 satisfying v 0 0 ( ) = , v 0 35 ( ) = and
v 0 90 ( ) = are sketched in Figure 4.7 below.
b) From the slope feld it appears that all solutions have a limiting value
of 60 as t goes to infnity. Due to increasing air resistance the skydiver
reaches a maximum velocity, or terminal velocity, of 60 metres per
second.
5 10 15 20
t
100
90
80
70
60
50
40
30
20
10
0
v
Note that the scales on the axes for the slope
felds in Figures 4.2, 4.3, 4.4 and 4.5 are equal.
Thus, the short line segments accurately depict
the true slope for solution curves. The scales are
not equal on the axes in Figures 4.6 and 4.7, so
the line segments do not give a true indication
of the slope. However, this is not an error.
Sometimes, it is necessary to have unequal
scales in order to show an appropriate interval
of values for the independent and dependent
variables. Figure 4.8 shows a portion of the
same slope feld given in Figures 4.6 and 4.7 but
with equal scales on the axes.
0 10 20 30 40 50
t
80
70
60
50
40
v
Figure 4.7
Figure 4.8
101
Although there are two
integrals in the equation
1
q
d p d C
y
y x x
( )
= ( ) +
∫ ∫
,
only one constant of
integration is needed. We
could add a constant to both
sides but they could then be
combined into one constant.
4.3 Separable equations
A class of frst order differential equations with which we are already
familiar (Section 16.8 of the book) and which can be solved analytically
using integration is the class of separable equations. These are differential
equations
dy
dx
F x y = ( ) , that can be rewritten so that the variables x and y
(along with their differentials dx and dy) are on opposite sides of the
equation. For a frst order differential equation where this separation of
variables can be accomplished, the function F x y , ( ) can be factored into
a product of two functions – one involving only the independent variable
(e.g. x) and the other involving only the dependent variable (e.g. y). That is,

dy
dx
F x y p x q y = ( ) = ( ) ( ) , .
It is not always obvious whether or not a differential equation is separable.
Some algebraic manipulation is needed to confrm that the differential
equation can, in fact, be written in the form
dy
dx
p x q y = ( ) ( ). For example,
dy
dx xy
x
y
=
3
2
2 is separable because it can be written as
dy
dx y x
x =

1 3
2
2 ;
and
tan
ln
x
y
dy
dx y
=
2
is also separable because it can be written as
dy
dx
y
y
x =
2
ln
cot . However, the equations
dy
dx
x y = +
2 2
and
dy
dx
xy = + 1
are not separable.
Example 3
Find the general solution of the differential equation
x y
dy
dx
x x y
2
1 0 0 = + > > , , .
Solution
The equation is separable because algebraic rearrangements can be
performed to write the equation as

dy
dx y
x
x
=
+ 

1 1
2
Separable equation
A frst order diferential equation is considered separable if it can be written in the form

d
d
p q
y
x
x y = ( ) ( ).
The variables can then be separated by writing the equation in the form

1
q
d p d
y
y x x
( )
= ( )
and integrating both sides gives

1
q
d p d C
y
y x x
( )
= ( ) +
∫ ∫
which leads to a general solution.
102
Differential Equations
4
which is in the form
dy
dx
p x q y = ( ) ( ) with p x
x
x
( ) =
+1
2
and q x
y
( ) =
1
.
We now separate the variables and integrate, giving:
y dy
x
x
dx =
+1
2
y dy
x x
dx
∫ ∫
= +

1 1
2

1
2
1
2
y x
x
C = + ln 2
y x
x
C = + 2
2
ln 2
This is the general solution of x y
dy
dx
x x y
2
1 0 0 = + > > , , in explicit form.
With some thinking we can use our GDC to help confrm this result.
dy
dx y
x
x
=
+ 

1 1
2
is the rule that gives us the slope of the graph of the
function y x ( ) at any point x y , ( ). In the GDC screen images below we
enter the function y x
x
C = + 2
2
ln 2 ; choose a value for x (x = 4, for
example); use the GDC to fnd an approximate value for the corresponding
y-coordinate; use the rule for
dy
dx
to fnd the slope at that point; and then
check to see if the same value for
dy
dx
is given when evaluating it on the
graph screen.
The GDC can also draw the tangent line at x = 4 and display its equation,
confrming that the slope of the function at x = 4 is approximately
0.207 295 38, agreeing with the value computed by
dy
dx y
x
x
=
+ 

1 1
2
.
dy/dx=.20729538
Plot1
Y1= (21n(X)-2/X
Y1(4)(4)
Plot2 Plot3
Y2=
Y3=
Y4=
Y5=
Y6=
CALCULATE
1:value
2:zero
3:minimum
4:maximum
5:intersect
6:dy/dx
7:∫f(x)dx
X=4 Y=1.50751077
Y1= (21n(X)-2/X)
1.50751077
(1/Ans)((4+1)/4
2
) .2072953681
)
103
Here is an applied problem involving a separable differential equation.
Example 4
The rate of decay of a substance y at any time t is directly proportional to
the amount of y and also directly proportional to the amount of another
substance x. The constant of proportionality is 2
1

_

2
and the value of x at
any time t is given by x
t
=
+ ( )
4
1
2
.
a) Given the initial conditions that y = 10 when t = 0, fnd y as an explicit
function of t.
b) Determine the amount of the substance remaining as t becomes very
large.
Solution
a) The rate of decay of substance y is proportional to the product xy, and
with the constant of proportionality having a value of 2
1

_

2
and
x
t
=
+ ( )
4
1
2
, this gives:

dy
dt t
y =
+
( )

2
1
2
4
1
2

1 2
1
2
y
dy
t
dt =
+ ( )
2
Separating variables.

1
2
1
1
2
y
y d
t
dt
∫ ∫
=
+ ( )
2 Integrating both sides.
ln y
t
C =
+
+
2
1
y e
t
C
=
+
+
2
1
Exponentiating; using e as the base.
y e e
C t
=
+
2
1
y Ae
t
=
+
2
1
Let A = e
c
, a convenient form for the
arbitrary constant.
Solve for A knowing that initially y = 10 when t = 0:
10 10 10
2
1 0 2 2
= ⇒ = ⇒ =
+
Ae Ae A e
2
Substituting gives:

y e e y e y e
t t
t
t
= ⇒ = ⇒ =
+ + +
10 10 10
2
2
2
1
2
1
2
1 2
2
2
b) As t → ∞,
2
2
2
1
2
t
t +
→ ; thus, as t → ∞, y e → ≈ 10 1 36
2 2
.
104
Differential Equations
4
Example 5
Solve the differential equation x dx e y dy
x y
cos + =
+
0.
Solution
As it is the equation cannot be written in the variables separable form
dy
dx
p x q y = ( ) ( ). Since e e e
x y x y +
= we can make it so by multiplying both
sides of the equation by e
x 2
and doing some rearrangement.

xe dx e y dy e y dy xe dx
dy
dx
xe
e
x y y x x 2 2 2
2 2 + = ⇒ = ⇒ = cos cos 0
1
yy
y cos

Separating the variables and integrating both sides gives:
e y dy xe dx
y x
cos
∫ ∫
= 2
2

e
y y xe e C
y
x x
2
sin cos + ( ) = + +
2 2
Using integration by parts on both sides.
Therefore, the implicit function
e
y y e x C
y
x
2
1 sin cos + ( ) = + ( ) +
2
is the
general solution.
To fnish this section we will fnd an explicit solution by the method
of separation of variables for a relatively straightforward frst order
differential equation, but one whose solution will prove useful in
developing another solution method.
Example 6
Find the general solution to the differential equation
dy
dx
xy = 22 .
Solution

1
2
y
dy x dx = 2 Separating variables; note loss of solution where y = 0.

1
2
y
dy x dx
∫ ∫
= 2 Integrating both sides.
ln y x C = + 2
2
1

e e
y x C ln
=
+ 2
2
1
Exponentiate both sides to solve for y.

y e e
C x
=
1
2
2

y C e
x
=
2
2
2
e
c
1
is a positive constant; let e
c
1

= C
2
and C
2
> 0.
If y > 0, then y y = and the solution becomes

y C e
x
=
2
2
2
.
If y < 0, then y y = 2 and the solution becomes

y C e
x
= 2
2
2
2
.
105
We can include both of these solutions, and also the ‘lost’ solution y = 0,
by giving the general solution as

y Ce
x
=
2
2
with no restrictions on the constant C.
It is helpful for our review to recognize that the explicit solution
y Ce
x
=
2
2

for Example 6 defnes a ‘family’ of curves in the xy-plane. Some of these
curves, with the corresponding value of C, have been graphed in Figure 4.9.
In order to determine a specifc curve from this ‘family’ we must impose an
initial condition on the solution.
4.4 First order linear differential
equations – use of integrating
factor
As mentioned previously, a frst order differential equation is called such
because the frst derivative
dy
dx
appears in the equation. A differential
equation is called linear when both
dy
dx
and y appear only to the frst
power. The standard form for a frst order linear differential equation is

dy
dx
P x y Q x + ( ) = ( ).
We wish to develop a method to solve frst order linear differential equations
of this form (which could also be written as ′ + ( ) = ( ) y P x y Q x ).
2
1

1
2
2 1 1 2 x
y
C = 2
C = 1
C =

1
C =

2
Figure 4.9
106
Differential Equations
4
We start by considering a simple case when Q x ( ) = 0, so
dy
dx
P x y Q x + ( ) = ( ) becomes

dy
dx
P x y + ( ) = 0.
This equation is variables separable, giving us

1
y
dy
dx
P x = ( ) 2 .
This equation can be integrated in the same way as in Example 6 to give
ln y P x dx C = ( ) +

2
1
and following the same steps as in Example 6, we get

y Ce
P x dx
=

( ) 2
which is a general solution for the linear differential equation
dy
dx
P x y + ( ) = 0.
However, we wish to fnd a general solution to the more general frst
order linear differential equation
dy
dx
P x y Q x + ( ) = ( ) where Q x ( ) is not
necessarily zero. By applying the product rule and implicit differentiation
we observe that

d
dx
ye
dy
dx
e yP x e
P x dx P x dx P x dx ( ) ( ) ( )

( )
=

+ ( )

=

+ ( )

( )
e
dy
dx
P x y
P x dx
.
Thus, if we multiply both sides of
dy
dx
P x y Q x + ( ) = ( ) by the factor
e
P x dx ( )

(called an integrating factor), we get
e
dy
dx
P x y e Q x
P x dx P x dx ( ) ( )

+ ( )

=

( ).
From the working above, we can substitute
d
dx
ye
P x dx ( )

( )
for
e
dy
dx
P x y
P x dx ( )

+ ( )

, yielding

d
dx
ye e Q x
P x dx P x dx ( ) ( )

( )
=

( ).
Integrating both sides gives

ye e Q x dx C
P x dx P x dx ( ) ( )

=

( ) +

.
We can now solve for y, giving

y e e Q x dx C
P x dx P x dx
=
∫ ∫
( ) +

( ) ( )

2
.
107
Although the expression for the general solution given above looks quite
complicated, the basic steps for solving a frst order linear differential
equation by means of an integrating factor are relatively simple.
Step 1: Make sure the equation is in the standard form
dy
dx
P x y Q x + ( ) = ( ).
Step 2: Compute the integrating factor e
P x dx ( )

by fnding P x dx ( )

.
Step 3: Multiply both sides of the equation by the integrating factor.
Step 4: Integrate both sides of the equation. The left side will be

e
dy
dx
P x y
P x dx ( )

+ ( )

which is equivalent to
d
dx
ye
P x dx ( )

( )
and the
integral of this expression is ye
P x dx ( )

.
Step 5: Obtain an explicit solution for y by dividing both sides by the
integrating factor e
P x dx ( )

.
Let’s illustrate the fve basic solution steps with an example.
Example 7
Find the general solution of x
dy
dx
y x 22
2
= .
Solution
1.
x
x
dy
dx
y
x
x
x
2
2
2
= Divide both sides by x to get equation into standard form.

dy
dx x
y x 2
2 

= Standard form
d
d
P Q
y
x
x y x + ( ) = ( ) ; P x
x
( ) = 2
2
and Q x x ( ) = .
2. Integrating factor: e e e
e
x
x
dx
x x
x
2
2 2
2
2
2
2
2
1 1

= = = =
ln ln
ln
3.
1 2 1
2 2
x
y
x x
y
x
x
d
d
2
j
(
,
\
,
(
,
¸
,
]
]
]
( ) Multiply both sides by integrating factor.

1 2 1
2 3
x
dy
dx
y
x x
2 =
Solution to frst order linear differential equations
Given a frst order linear diferential equation in the form

d
d
P Q
y
x
x y x + ( ) = ( )
the general solution is

y x x
x x x x x x
= ( ) +
( ) ( ) ( )
∫ ∫ ∫

e e Q d Ce
P d P d P d 2 2
where C is an arbitrary constant.
When computing the
integrating factor

e
P d x x ( )

,
it

is standard practice to omit
the constant of integration
from the indefnite integral of
P x ( ).
108
Differential Equations
4
4.
1 2 1
2 3
x
dy
dx
y
x
dx
x
dx 2

=
∫ ∫
Integrate both sides with respect to x.

y
x
x C
1
2

= + ln

d
d
d
d x
y
x x
y
x
y
x
1 1 2
2 2 3
j
(
,
\
,
(
,
¸
,
]
]
]
2 ,
by product rule and implicit
diferentiation.
5. Therefore, y x x Cx = +
2 2
ln is the general solution.
Example 8
Find the particular solution of
x
dy
dx
xy x x
2 2
1 1 2 1 +
( )
+ = ( ) + 2
given that y = 2 when x = 1.
Solution
1.
x
x
dy
dx
xy
x
x x
x
2
2 2
2
2
1
1 1
1 2 1
1
+
+
+
+
=
( ) +
+
2

dy
dx
x
x
y
x
x
+
+

=
+
2
2
1
1 2
1
2
Standard form with P x
x
x
( ) =
+
2
1
, Q x
x
x
( ) =
+
1 2
1
2
2
.
2. Integrating factor:
P x dx
x
x
dx x x e x
x
( ) =
+
= +
( )
= + ⇒ = +
∫ ∫
+
2
2 2 1 2
1
1
2
1 1 1
2
ln ln
ln
3. x
dy
dx
x
x
x
y
x x
x
2 2
2
2
2
1 1
1
1 1 2
1
+ + +
+

=
+ ( )
+
2
Multiply both sides
by integrating factor.

x
dy
dx
x
x
y x
2
2
1
1
1 2 + +
+

= 2
4. x
dy
dx
x
x
y dx x dx
2
2
1
1
1 2 + +
+
j
(
,
\
,
(
,
¸
,
]
]
]
( )
∫ ∫
2 Integrate both sides.

y x x x C
2 2
1 + = + 2
5. y
x x C
x
=
+ +
+
2
2
2
1
Divide both sides by integrating factor.
To solve for C, we substitute y = 2 and x = 1.
2
1 1
1 1
2 2 =
+ +
+
⇒ =
2 C
C
Therefore, the particular solution is y
x x
x
=
+ +
+
2
2
2
2 2
1
.
It is appropriate to call the
diferential equation
d
d
P Q
y
x
x y x + ( ) = ( ) linear
because
d
d
P Q
y
x
x y x = ( ) + ( ) 2
is a linear function of y.
109
Once again, with a bit of effort, we can add some
confdence to our result for Example 8 by using our
GDC to graph the solution curve and then check to see
if the original differential equation accurately describes
its behaviour (shape).
Enter the solution curve for Y
1
and enter the
differential equation in the form
dy
dx
x
x
y
x
x
=
+

+
+
2
2
2
2
1
1 2
1
for Y
2
. Turn Y
2
‘off ’
(un-highlight) so that it is not graphed; only the
solution curve is graphed. Choose a value for x that
is in the graph window – say, x = 21; and evaluate
the corresponding y-value for a point on the solution
curve.
y =
( ) +
( ) +
=
+
=
2 2 2
2
2
2
1 1 2 2
1 1
2 2 2
2
2 2
2
2
; point
2 2 1 2 2 ,
( )
is on the solution curve. After setting y
equal to 2 2 2 , use Y
2
to fnd the value of
dy
dx
at 2 2 1 2 2 ,
( )
. Check
that this value for the slope of the curve at 2 2 1 2 2 ,
( )
, found to be
approximately 2.414 213 562, agrees with the value found on the graph
window. Both methods of fnding
dy
dx
at 2 2 1 2 2 ,
( )
, from the differential
equation and from the solution to the differential equation, give the same
value, thus supporting our particular solution to the differential equation.
Example 9
In the earlier section on slope felds, we displayed a slope feld (Figure 4.2)
for the differential equation
dy
dx
x y = 2 . Find the general solution to this
equation.
Solution
The equation frst appears that it may be separable, but it cannot be
expressed in the form
dy
dx
p x q y = ( ) ( ). It is a frst order linear differential
equation because it can be rearranged to
dy
dx
y x + = which puts it into the
standard form
dy
dx
P x y Q x + ( ) = ( ) such that P x ( ) = 1 and Q x x ( ) = . The
integrating factor is e e
dx
x ∫
= , and multiplying through by this gives
e
dy
dx
e y e x
x x x
+ =
2- (2)→Y4)
.5857864376
Y2(-1)
2.414213562
X=-1
)/ (X
2
+1)
Plot1
Y1=(-X
2
+X+2 (2)
Y1=(-X
2
+X+2 (2))/ (X
2
+1)
dy/dx=2.4142134
Plot2 Plot3
Y2=(-X/X
2
+1))Y
+(1-2X)/ (X
2
+1)
Y3=
Y4=
Y5=
CALCULATE
1:value
2:zero
3:minimum
4:maximum
5:intersect
6:dy/dx
7:∫f(x)dx
110
Differential Equations
4
and continuing with the steps for solving a frst order linear differential
equation yields
e
dy
dx
e y dx e x dx
x x x
+

=
∫ ∫
ye e x e C
x x x
= + 2 Using integration by parts on the right side.
Thus, the general solution is y x Ce
x
= + 2
2
1 . Figure 4.10 shows the same
slope feld displayed in Figure 4.2 for
dy
dx
x y = 2 along with the graphs
of three different solution curves generated from the general solution for
C = 1
1
10
4 , and 2 .
An analytic method for solving differential equations, such as those for
separable equations and frst order linear equations, demand fuency with
a range of integration techniques, and differentiation – as the next example
nicely illustrates.
Example 10
Find the particular solution to 1 1
4
+ ( ) = + ( ) sin cos sin x
dy
dx
y x x 2 given
y 0 1 ( ) = .
Solution
Dividing through by 1 + sinx, the equation becomes

dy
dx
x
x
y x 2
cos
sin
sin .
1
1
3
+

= + ( )
The integrating factor is
e e e
x
x
dx
x
x
2
2
2
cos
sin ln sin
ln sin
1 1
1
1
+ + ( )
+ ( )

= = =
11
1 + sin
.
x
6
5
4
3
2
1

1
2
3
4
5
6
6 5 4 3 2 1 0 1 2 3 4 5 6
x
y
C = 4
C = 1/10 C = 1
Figure 4.10
111
Multiplying both sides by the integrating factor gives

1
1
1
1
2
2
+
+ ( )

= + ( )
sin
cos
sin
sin .
x
dy
dx
x
x
y x 2
Our experience of differentiating functions and familiarity with the
solution pattern for frst order differential equations, informs us that the
left side is equal to the derivative of
y
x 1 + sin
.

d
dx
y
x
x x
1
2 1
2
+

= + +
sin
sin sin
We now integrate both sides. The integral of the left is simply
y
x 1 + sin
and
integrating each term on the right is straightforward except for sin
2
x. We
need to take the double-angle identity cos sin 2 1 2
2
x x = 2 and rearrange
it to give us sin cos
2
1
2
1
2
2 x x = 2 .

d
dx
y
x
dx x x dx
1
1
2
1
2
2 2 1
+

= + +

∫ ∫
sin
cos sin 2

y
x
x
x x x C
1 2
1
4
2 2
+
= + +
sin
sin cos 2 2
y x
x
x x C = + ( ) +

1
3
2
1
4
2 2 sin sin cos 2 2
Given y 0 1 ( ) = , it follows that 1 1 0 0 0 2 3 = + ( ) + ( ) ⇒ = 2 2 C C
Therefore, the particular solution is
y x x x x = + ( ) + ( )
1
4
1 6 2 8 12 sin sin cos . 2 2
4.5 Homogeneous differential
equations
When a frst order differential equation is not separable nor linear, it may
still be possible to transform it by an appropriate substitution into an
equation that we can solve analytically. One situation where this will always
work is when the frst order differential equation is homogeneous.
The function F can be written as a function of
y
x
if it can be expressed as
a quotient of two homogeneous functions of the same degree. In general,
a two-variable function is homogeneous of degree n if the sum of the
powers of x and y in each term is n. For example: g x y x xy y , ( ) = + 2 5
2 2
2
Homogeneous frst order differential equations
The diferential equation
d
d
F
y
x
x y = ( ) , is homogeneous if the right side can be
expressed as a function of the ratio
y
x
alone, that is,
d
d
F
y
x
y
x
=

 .
112
Differential Equations
4
is homogeneous of degree 2; and h x y y xy , ( ) = 3
3 2
2 is homogeneous of
degree 3. The function m x y x y x y , ( ) = 4
2 2 3 2
2 is not homogeneous.
Thus, if we solve for
dy
dx
and get it to be equal to a quotient in the form
M x y
N x y
,
,
,
( )
( )
where M and N are homogeneous functions of the same degree,
then the equation is a homogeneous differential equation. The equation
dy
dx
M x y
N x y
=
( )
( )
,
,
can be written as a function of
y
x
, i.e.
dy
dx
F
y
x
=

, by
dividing through both M x y , ( ) and N x y , ( ) by x
n
, where n is the degree of
M and N. Two examples are given below.
1.
dy
dx
xy
x y
=
6
2 2
2
is a homogeneous differential equation because both
the numerator, 6xy, and the denominator, x y
2 2
2 , are homogeneous
functions of degree 2. The right side can be expressed in terms of
y
x
by
dividing numerator and denominator by x
2
.

dy
dx
xy
x
x y
x
y
x
y
x
= =

6
6
1
2
2 2
2
2
2
2
2.
dy
dx
y xy
x x y xy
=
+
3
3 2
3 2 2
2
2
is a homogeneous differential equation because
both the numerator, 3
3 2
y xy 2 , and the denominator, x x y xy
3 2 2
+ 2 ,
are homogeneous functions of degree 3. We divide numerator and
denominator by x
3
to get

dy
dx
y
x
xy
x
x
x
x y
x
xy
x
y
x
y
=
+
=

3
4 2
3
3
3
2
3
3
3
2
3
2
3
3
2
2
2
xx
y
x
y
x

+

2
2
4 2 2
Once a homogeneous differential equation is written in the form
dy
dx
F
y
x
=

it can be solved analytically by making the substitution
y vx = or v
y
x
=

where v is a differentiable function of x. As we
will see, this substitution transforms the differential equation into a
separable equation for which we have a solution method.
Example 11
Find the particular solution for xy
dy
dx
x y
2 3 3
= + given y = 3 when x = 1.
113
Solution
Dividing both sides by xy
2
reveals that the differential equation is
homogeneous because both numerator and denominator on the right side
are homogeneous functions of degree 3.

dy
dx
x y
xy
=
+
3 3
2
Dividing both numerator and denominator by x
3
expresses the derivative
in terms of
y
x
.

dy
dx
x
x
y
x
xy
x
y
x
y
x
=
+
=
+

3
3
3
3
2
3
3
2
1
We now let y vx = which means that
dy
dx
v x
dv
dx
= + by means of the
product rule. Substituting v for
y
x
and v x
dv
dx
+ for
dy
dx
produces
v x
dv
dx
v
v
+ =
+ 1
3
2
which is a separable equation for the variables x and v because it can be
written in the form
dy
dx
p x q v = ( ) ( ), as shown below:
x
dv
dx
v
v
v x
dv
dx
v
v
v
v
x
dv
dx v
dv
dx
=
+
⇒ =
+
⇒ = ⇒
1 1 1
3
2
3
2
3
2 2
2 2 ==

1 1
2
x v
Separating the variables and integrating:
v dv
x
dx
2
1
=
v dv
x
dx
2
1
∫ ∫
=

1
3
3
v x C = + ln
If y = 3 when x = 1, then v
y
x
= = =
3
1
3, and substituting gives
9 1 9 = + ⇒ = ln C C . Thus,

1
3
9 3 27
3 3
v x v x = + ⇒ = + ln ln
Substituting
y
x
back in for v gives:

y
x
x

= +
3
3 27 ln
y x x
3 3
3 27 = + ( ) ln
Therefore, the particular solution is y x x = + ( ) 3 27
1
3
ln .
Using Example 11 as a guide we can outline the basic steps for solving a
frst order homogeneous differential equation.
Step 1: Confrm that, or rearrange it so that, the equation is in the form
114
Differential Equations
4

dy
dx
M x y
N x y
=
( )
( )
,
,
, where M and N are homogeneous functions of the
same degree.
Step 2: Divide both M x y , ( ) and N x y , ( ) by x
n
, where n is the degree of
M and N, so that the equation is in the form
dy
dx
F
y
x
=

.
Step 3: Let y vx = from which it follows that
dy
dx
v x
dv
dx
= + and
substitute v for
y
x
and v x
dv
dx
+

for
dy
dx
transforming the
equation into a separable equation in terms of v and x.
Step 4: By applying the technique of separation of variables, fnd a
solution in terms of v and x.
Step 5: Substitute
y
x
back in for v and write the solution in terms of y
and x.
Example 12
Consider the differential equation
dy
dx
x y
x y
=
+
2
where x y > > 0 0 , .
a) Use the substitution y vx = to show that x
dv
dx
v
v
=
+ 1
1
2
2
.
b) Hence, fnd the general solution of the differential equation, giving your
answer in the form C f x y = ( ) , .
Solution
a) 1. The equation is already in the form
dy
dx
M x y
N x y
=
( )
( )
,
,
where, in this
case, M and N are homogeneous of degree 1.
2. Divide numerator and denominator by x.

dy
dx
x
x
y
x
x
x
y
x
y
x
y
x
=
+
=
+
2 2
1
1
3. Letting y vx = , then
dy
dx
v x
dv
dx
= + . Substituting v for
y
x
and
v x
dv
dx
+ for
dy
dx
, gives:
v x
dv
dx
v
v
+ =
+ 1
12
x
dv
dx
v
v
v v
v
=
+ ( ) 1
1
1
1 2
2
2
2
x
dv
dx
v
v
=
+ 1
1
2
2
Q.E.D.
Do not forget to perform Step
5 – substituting
y
x
back in for
v – because you must express
your fnal solution in terms of
y and x.
115
b) 4. Separating the variables and integrating, yields

1
1
1
2
2v
v
dv
x
dx
+
=
∫ ∫
.
To integrate the left side we split up the fraction:

1
1 1
1
2 2
+ +
=
∫ ∫ ∫
v
dv
v
v
dv
x
dx 2
arctan ln ln v v x C 2
1
2
1
2
+
( )
= +
5. Substituting
y
x
back in for v gives
arctan ln ln .
y
x
y
x
x C

+

= + 2
1
2
1
2
2
Solving for C:
arctan ln ln
y
x
y
x
x C
j
(
,
\
,
(
+
j
(
,
\
,
(
+
,
¸
,
,
,
]
]
]
]
]
2 1
2
2
1
2
arctan ln
y
x
x
y
x
C

+

= 2 1
2
2
C
y
x
x y =

+ arctan ln 2
2 2
This is the general solution such that y is an implicit function of x.
Example 13
a) Show that
d
dx
x x
x
ln + +
( )

=
+
1
1
1
2
2
.
b) Show that the solution curve that satisfes the differential equation
x
dy
dx
y x y = + +
2 2
with initial conditions y 0 1 ( ) = 2 is the parabola
y
x
=
2
4
1 2 . [Hint: Use the result from a) to integrate the separable
equation that is in terms of v and x.]
Solution
a)
d
dx
x x
x x
d
dx
x x ln + +
( )
,
¸
]
]

+ +
+ +
( )
j
(
,
\
,
(
,
¸
,
1
1
1
1
2
2
2
1
2
]]
]
]
=
+ +
+ +
( ) ( )

1
1
1
1
2
1 2
2
2
1
2
x x
x x
2
=
+ +
+
+

1
1
1
1
2 2
x x
x
x
=
+ +
+
+
+
+

1
1
1
1 1
2
2
2 2
x x
x
x
x
x
116
Differential Equations
4
=
+ +
+ +
+

1
1
1
1
2
2
2
x x
x x
x
=
+
1
1
2
x
Q.E.D.
b) First, divide both sides by x to give
dy
dx
y x y
x
=
+ +
2 2
. The term
x y
2 2
+ has a degree of 1, so both numerator and denominator
are homogeneous functions of degree 1. Now divide numerator and
denominator on right side by x in order to write the equation in the
form
dy
dx
F
y
x
=

.

dy
dx
y
x
x y
x
x
x
y
x
y
x
=
+
+
= + +

2 2
2
2
1
Letting y vx = , then
dy
dx
v x
dv
dx
= + . Substituting v for
y
x
and
v x
dv
dx
+ for
dy
dx
, gives:
v x
dv
dx
v v x
dv
dx
v + = + + ⇒ = + 1 1
2 2
Separating the variables and integrating:

1
1
1
2
+
=
∫ ∫
v
dv
x
dx
From part a) we know that
d
dx
x x
x
ln + +
( )

=
+
1
1
1
2
2
. Therefore,
1
1
1
2
2
+
= + +
( )
+

v
dv v v C ln . Using this result gives:
ln ln ln v v x C + +
( )
= + 1
2
Setting arbitrary constant to ln C.
e e
v v
x C
ln
ln ln
+ +
+ ( )
=
1
2
Exponentiating both sides using base of e.
v v Cx + + = 1
2
1
2
2
2
+
( )
= ( ) v Cx v 2
1 2
2 2 2 2
+ = + v C x Cxv v 2
1 2
2 2
= C x Cxv 2
1 2
2 2
=

C x Cx
y
x
2 Substituting
y
x
back in for v.
2 1
2 2
Cy C x = 2
y Cx
C
=
1
2
1
2
2
2
117
Solve for C given the initial condition y(0) = 21:
2 2 1 0
1
2
1
2
= ⇒ =
C
C
Hence,
y x =

( )
1
2
1
2
1
2
2
1
2
2 .
Therefore, the particular solution curve is the parabola y x =
1
4
1
2
2 .
4.6 Euler’s method
We have established three analytic methods for solving different types of
frst order differential equations: separable equations, linear equations
(integrating factor) and homogeneous equations (substitution y = vx).
Also, earlier in this chapter we saw how a slope feld is an effective
graphical method that provides a rough idea about the behaviour of
solutions to a differential equation, especially for an equation that we
are not able to solve analytically. To roughly sketch a particular solution
to a differential equation using a slope feld, we need to know a point
(initial condition) that the solution curve passes through in order to
have a ‘starting point’ from which to sketch a curve that will be parallel to
the short line segments drawn at representative points that indicate the
slope of any solution. Several of the examples in this chapter have found
particular solutions to what is referred to as an initial-value problem that
is stated in the form

dy
dx
F x y y x y = ( ) ( ) = , , .
0 0
Consider the initial-value problem

dy
dx
x y y = + ( ) = , . 2 0
Figure 4.11 shows the slope feld for the differential equation
dy
dx
x y = + .
An approximation to the particular solution can be sketched by drawing a
smooth curve through the point 2 0 , ( ) that follows the slopes in the slope
feld, as shown in the fgure.
4
2
0
2
4 2 2 4
x
y
Figure 4.11 Slope feld for
d
d
y
x
x y = + and sketch of solution
passing through (2, 0).
118
Differential Equations
4
Let y(x) represent the solution curve. To approximate a value of y for a
specifc value of x, for example y when x = 3, we could make an educated
guess from the sketch of y made with the aid of the slope feld. But if we
want a more accurate approximation then we need to use a more refned
method. The simplest numerical method is called Euler’s method, after
the prolifc eighteenth-century mathematician who frst devised this
computational method to help him calculate the orbit of our Moon.
Euler’s method uses the basic idea behind the construction of slope felds
to fnd numerical approximations to solutions of differential equations.
Let’s illustrate the method with the initial-value problem that we have just
been considering, namely:

dy
dx
x y y = + ( ) = , 2 0
We know from the differential equation that the slope of the solution
curve is 2 at the point (2, 0) because
dy
dx
x y = + = + = 2 0 2. Hence,
the line tangent to the solution curve at (2, 0) has the equation:
y x y x 2 2 2 0 2 2 2 4 = ( ) ⇒ = . We can use this tangent line as a rough
approximation to the solution curve (see Figure 4.12). This approximation
clearly becomes less accurate as we move away from the point of tangency
(2, 0).
Euler’s method improves this approximation by moving a short horizontal
distance (the step size h) along this tangent line and then change direction
according to the slope feld. In this way we build an approximation to the
curve by attaching little line segments together, each having the slope of the
solution curve at its starting point.
In general, after being presented with an initial value problem:
dy
dx
F x y y x y = ( ) ( ) = , ,
0 0
we choose a step size h. Starting at the point
x y
0 0
, ( ), for the interval x x x h
0 0
  + , we approximate the solution
curve with the tangent line, i.e. the line with slope F x y
0 0
, ( ). This takes us
4
3
2
1
0
1
2
1
x
2 2 1 4
line tangent at (2, 0):
y = 2x  4
3
y
Figure 4.12
119
as far as the point x y
1 1
, ( ), whose coordinates are calculated as follows (see
Figure 4.13):

x x h
1 0
= + , y y hF x y
1 0 0 0
= + ( ) ,
Now we are at the starting point of the second line segment x y
1 1
, ( ). We
repeat the process, with the next line segment having slope F x y
1 1
, ( ).
This takes us to the next point x y
2 2
, ( ) on the Euler approximation where
x x h
2 1
= + and y y hF x y
2 1 1 1
= + ( ) , .
Repeating this process we get an approximation to the solution curve
consisting of line segments joining the points x y
0 0
, ( ), x y
1 1
, ( ), x y
2 2
, ( ), etc.
Each computed value y
n
is an estimate of the corresponding ‘true solution’
y at x x
n
= . The accuracy of the estimates depends on the choice of the step
size h and the overall number of steps (iterations). Decreasing the step size
while increasing the number of steps leads to increasingly more accurate
estimates for solution values.
slope = F(x
0
, y
0
)
(x
1
, y
1
)
h F(x
0
, y
0
)
(x
0
, y
0
)
y
x
h
Figure 4.13 Euler’s method starts at x y
0 0
, ( ) on the solution curve and moves along a
segment with slope F x y
0 0
, ( ) to defne a new point x y
1 1
, ( ) such that x x
1 0
= + h and
y y x y
1 0 0 0
= + ( ) hF , . The process is repeated with the new point.
Figure 4.14 Further iterations
of Euler’s method build an
approximation to the solution
curve.
solution
curve
Euler
approximation
(x
0
, y
0
)
(x
1
, y
1
)
(x
2
, y
2
)
(x
3
, y
3
)
(x
4
, y
4
)
(x
5
, y
5
)
120
Differential Equations
4
Let’s now apply Euler’s method to answer a question posed earlier for the
initial-value problem presented at the start of this section.
Example 14
For the differential equation
dy
dx
x y = + such that y 2 0 ( ) = , use Euler’s
method with a step value of 0.2 to fnd an approximate value of y when
Solution
We use Euler’s method to build an approximation to the ‘true’ solution
curve starting at x = 2 and fnishing at x = 3 by piecing together fve short
segments (Figure 4.15). We are given that h = 0 2 . , x
0
2 = , y
0
0 = and
F x y x y , ( ) = + . Using the appropriate formulae for x
n
and y
n
and iterating
fve times, we have:
x x h
1 0
2 0 2 2 2 = + = + = . .
y y hF x y
1 0 0 0
0 0 2 2 0 0 4 = + ( ) = + + ( ) = , . .
x x h
2 1
2 2 0 2 2 4 = + = + = . . .
y y hF x y
2 1 1 1
0 4 0 2 2 2 0 4 0 92 = + ( ) = + + ( ) = , . . . . .
x x h
3 2
2 4 0 2 2 6 = + = + = . . .
y y hF x y
3 2 2 2
0 92 0 2 2 4 0 92 1 584 = + ( ) = + + ( ) = , . . . . .
x x h
4 3
2 6 0 2 2 8 = + = + = . . .
y y hF x y
4 3 3 3
1 584 0 2 2 6 1 584 2 4208 = + ( ) = + + ( ) = , . . . . .
x x h
5 4
2 8 0 2 3 = + = + = . .
y y hF x y
5 4 4 4
2 4208 0 2 2 8 2 4208 3 46496 = + ( ) = + + ( ) = , . . . . .
Euler’s numerical method
For the diferential equation
d
d
F
y
x
x, y = ( ) with the initial condition y(x
0
) = y
0
, the
recursive formulae for generating the coordinates of the unknown (n + 1)st point
(x
n + 1
, y
n + 1
) from the known nth point (x
n
, y
n
) on the approximate solution curve (Euler
approximation) are:
x
n + 1
= x
n
+ h, y
n + 1
= y
n
+ hF (x
n
, y
n
) for n = 1, 2, 3, … , N
where h, the step size, is a constant; and N is the total number of steps (iterations).
solution
curve
Euler
approximation
2 2.2 2.4 3 2.8 2.6
x
4
3
2
1
0
y
Figure 4.15
121
This process leads to an approximate (three decimal places) value of
y ≈ 3 46 . when x = 3.
Because we will perform most of the calculations for each iteration on our
GDC, it is often suffcient to simply display relevant results for each
iteration in a table, as shown below.
n x
n
y
n
hF(x
n
, y
n
) x
n + 1
y
n + 1
0 2 0 0.4 2.2 0.4
1 2.2 0.4 0.52 2.4 0.92
2 2.4 0.92 0.664 2.6 1.584
3 2.6 1.584 0.8368 2.8 2.4208
4 2.8 2.4208 1.044 16 3.0 3.464 96
The frst order diferential equation in Example 14 is linear and hence can be solved
by means of an integrating factor. Given y(2) = 0 the particular solution is
y = 3e
x 2 2
2 x 2 1. To three signifcant fgures, the ‘true’ value of y(3) is approximately
4.15. Thus, our approximation of 3.46 has an error of approximately 16.6%. Using
a program on our GDC or a spreadsheet, we could easily decrease the step size
(and increasing the number of steps) in order to improve the accuracy of the
approximation. For example, if we used a step size of h = 0.01 (requiring 100
iterations) we would get an estimate of 4.11 (3 s.f.), reducing the error to less than 1%.
A numerical method like Euler’s is especially useful when applied to a
differential equation that cannot be solved by any known analytic methods,
as we will do in the next example.
Example 15
Given that
dy
dx
x
xy
=
+
+
1
2
and y = 1 when x = 0, use Euler’s method with
step size h = 0.25 to approximate the value of y when x = 1. Give the
approximation to three signifcant fgures.
Solution
We have that x y h
0 0
0 1 0 25 = = = , , . and F x y
x
xy
, ( ) =
+
+
1
2
. Thus the
recursive formula for y
n
is:
y y hF x y y
x
x y
y y
x
n n n
n
n n
n n
n
+ +
= + ( ) = + ( )
+
+
⇒ = +
+
1 1
0 25
1
2
, .
11
4 8 x y
n n
+
n = 0: x x h
1 0
0 0 25 0 25 = + = + = . .
y y
x
x y
1 0
0
0 0
1
4 8
1
0 1
4 0 1 8
9
8
1 125 = +
+
+
= +
+
+
= =
( )( )
.
n = 1: x x h
2 1
0 25 0 25 0 5 = + = + = . . .
122
Differential Equations
4
y y
x
x y
2 1
1
1 1
1
4 8
1 125
0 25 1
4 0 25 1 125 8
= +
+
+
= +
+
+
≈ .
.
( . )( . )
11 261986 .
n = 2: x x h
3 2
0 5 0 25 0 75 = + = + = . . .
y y
x
x y
3 2
2
2 2
1
4 8
1 261986
0 5 1
4 0 5 1 261986
= +
+
+
= +
+
.
.
( . )( . ))
.
+

8
1 404518
n = 3: x x h
4 3
0 75 0 25 1 = + = + = . .
y y
x
x y
4 3
3
3 3
1
4 8
1 404518
0 75 1
4 0 75 1 4045
= +
+
+
= +
+
.
.
( . )( . 118 8
1 547801
)
.
+

Therefore, the approximate value of y when x = 1 is y ≈ 1 55 . .
1 Match each slope feld with its diferential equation, listed below.
i
d
d
y
x
y = 22
ii
d
d
y
x
x y =
2
2
iii
d
d
y
x
y
x
= 2
iv
d
d
y
x
x y = + 2
2 Solve the following separable diferential equations.
a
d
d
y
x
x
y
=
2
b
d
d
y
x
y
x
=
2
2

c
x
y
x
y y
2 2
d
d
= 2
d x
y
x
y
d
d
= tan e
d
d
y
x
xy = f x
y
x
x
y
2
1 + =
d
d
g
d
d e
y
x
y
x
=
2
1 2
h lny
y
x
d
d
= 1
Exercise 4.1
x
y
x
y
x
y
x
y
1
0 1
1
0
1
1
0 1
1
0 1
a b
c d
123
3 Using the method of separation of variables, show that an implicit solution for
the diferential equation
d
d
y
x
xy y
xy x
=
+
+
is y x
y x
e A e = where A is an arbitrary
constant.
4 Find the general solution, in explicit form, to the diferential equation
y
y
x
x
d
d
= cos . Comment on the possible values of the constant C.
5 The equation for the rate of change of the population (in thousands), p, of a
certain species is given by

dp
dt
p p = 5 2
2
2 .
a Sketch the slope feld.
b If the initial population is 4000 (that is, p(0) = 4), then what appears to be the
limiting value of the population (that is, lim
t
p t
→∞
( ))?
c If p(0) = 0.5, what is lim
t
p t
→∞
( )?
d Comment on the long-term behaviour of the species’ population growth.
6 Solve the initial-value problem:

d
d
y
x
x x
y
y =
+
( ) =
2
2
0 5
2
sec
, 2
7 Consider the initial-value problem:

1 1 0 0 1
2 2
+ ( ) + + = ( ) = x
y
x
y y
d
d
, 2
a Show that the implicit solution can be expressed as arctan arctan y x + =
π
4
.
b Use the formula for tan(A + B) to fnd the explicit solution.
8 Match each slope feld with its diferential equation, listed below.
i
d
d
y
x x y
=
+
5
ii
d
d
y
x y x
=
5
2

iii
d
d
y
x
y x
= 2
10
iv
d
d
y
x
xy
=
10
a
b (a)
(c)
(b)
(d)
(a)
(c)
(b)
(d)
(a)
(c)
(b)
(d)
(a)
(c)
(b)
(d) c d
124
Differential Equations
4
9 Solve the initial-value problem:

1 1 2 3
2 2
+ ( ) = + ( ) = x
y
x
y y
d
d
,
Write the solution in explicit form, expressing y in terms of x.
10 a Use the method of partial fractions to express
1
2
2
x x 2 2
as the sum of two
fractions.
b Consider the diferential equation
d
d
y
x
y
x x
x = >
2
2
2
2
2 2
, such that y = 1
when x = 5. Show that the solution is 2
1
2
3 3
e
2
2
y
y
x
x
=
+
.
11 Consider the diferential equation 1 2 2
2
2x
y
x
xy x. ( ) + =
d
d
a Find the general solution in the form y x = ( ) f by the method of separation of
variables.
b Write the diferential equation in the standard form for a frst order linear
diferential equation,
d
d
P Q
y
x
y x x + ( ) = ( ), and fnd the general solution by
means of an integrating factor.
12 Solve each of the following frst order linear diferential equations.
a
d
d
y
x x
y x +

=
2
6
3
b
d
d
y
x
xy x 2 =
c
d
d
y
x
y
x
x 2 =
3
d
d
d
e
y
x
y x
x
+ = sin
cos
e
d
d
e
y
x
x y
x
23
2
3
= f x
y
x
x y
d
d
= +
13 Solve the frst order linear diferential equation
tan sec x
y
x
y x
d
d
+ =

14 Consider the initial-value problem:

d
d
y
x
xy
x
2
2 1
1
2
=
, y 0 1 ( ) =
a Show that the diferential equation is a frst order linear equation by writing it
in the form
d
d
P Q
y
x
y x x + ( ) = ( ) .
b Show that the integrating factor is 1
2
2x .
c By using the substitution x = sin u, show that

1
1
2 2
2
2
2
2
x x
x x x
d C = + +

arcsin
.
d Find the solution to the initial-value problem expressed in the form y = f(x).
15 a Show that tan ln cos x x x d

= 2 .
b Show that
d
d
y
x
y x = + 1 tan is a frst order linear diferential equation.
c Find the general solution of
d
d
y
x
y x x = + < < 1
2 2
tan , 2
π π
.
125
16 Find the particular solution to the diferential equation
d
d
y y
x
x x
x
=
2
ln 2
given
that y = 1 when x = 1.
17 Find the general solution, in explicit form, to the diferential equation

x
y
x
x xy
2 3
0
d
d
2 + = .
18 Find the general solution to the frst order homogenous diferential equation

d
d
y
x
y x
x y
=
3
3
2
2
.
Write the answer in the form C f = ( ) x y , .
19 Solve each of the following frst order homogeneous diferential equations.
a
d
d
y
x
y
x
=
+ 1
b
d
d
y
x
x y
x
=
+ 2
c x
y
x
x y
d
d
= + 2 3 d
d
d
y
x
x y
xy y
=
+
+
2
2
2 3
2 2
2
e xy
y
x
x y
d
d
=
2 2
2 f x y x
y
x
y x y 2 ( ) = + ( )
d
d
20 Consider the diferential equation
d
d
y
x
x y
y x
=
+ 2
3 2 2
, for x > 0.
a Use the substitution y = vx to show that v
dv
d
v
v
+ =
+
x
x
1 2
3 2 2
.
b Hence, fnd the solution of the diferential equation, given that y = 0
when x = 1.
21 Use the substitution y = vx to show that the general solution to the diferential
equation
y x xy
y
x
2 2
0 2 + =
d
d

is 2
2 2 4
x y x 2 = C, where C is a constant.
22 Consider the initial-value problem:
x
y
x
y x y y
d
d
= + ( ) =
2 2
1 1 2 ,
a Use the substitution y = vx to show that v
dv
d
v v + = + x
x
1
2
2 .
b Hence, show that the solution is arcsin ln
y
x
x

= +
π
2
.
23 Consider the diferential equation
d
d
y y y
x x
=
+
2
.
a Find the general solution.
b Given that y = 1 when x = 1, fnd a particular solution solved explicitly for y.
c Use Euler’s method with step size h = 0 2 . to approximate the solution at
x = 1 2 1 4 1 6 1 8 . , . , . . . and
d Compute the percentage error for each of the approximate solutions found
in c compared to the solution for the same value of x found using the explicit
solution found in b.
126
Differential Equations
4
24 Given that
d
d
y
x
xy =
2
and y = 1 at x = 0, use Euler’s method with 5 steps to
approximate the value of y at x = 1.
25 Use Euler’s method with step size h = 0.1 to approximate the value of y when
x = 1 for the diferential equation
d
d
e
y
x
xy
= given that the solution curve passes
through the point 0 1 , ( ).
26 Use the substitution y = vx to fnd the general solution to the diferential
equation

d
d
y
x
x y
xy
=
+
2 2
3
2
.
27 Given that
d
d
y
x
x y = and y = 4 when x = 1, use Euler’s method with step size
h = 0.1 to approximate the solution at x = 1.1, 1.2, 1.3, 1.4 and 1.5.
28 Consider the initial-value problem:

d
d
y
x
x y y = ( ) = 2 , 0 0
a Show that the solution is y x
x
= + e
2
21.
b Use Euler’s method with 5 steps to fnd an approximate value of y when
x = 1.
c Use Euler’s method with 10 steps to fnd another approximation for y(1).
d Compare the approximate values for y(1) found in b and c to the actual value
using the solution y = e
–x
+ x 2 1. Comment.
1 Consider the diferential equation x
y
x
y x
d
d
23
4
= .
a Find the general solution.
b Given that y = 2 when x = 1, fnd the particular solution in explicit form.
2 Given that y = 2 when x = 1, solve the following diferential equation explicitly
for y.

y
y
x
x x
d
d
23
4
=
3 Find the general solution of the diferential equation
d
d
y
x
y
x
x = ≠
2
0 , .
4 Solve
d
d
y
x x
y x x + = ≠
1
0 cos , , giving your answer in the form y x = ( ) f .
5 Consider the diferential equation
d
d
y
x
y
x
x + =
2
.
a Find the general solution.
b Given that y = 20 when x = 4, fnd an explicit solution for y in terms of x.
Practice questions
127
6 Match each of the diferential equations with its direction feld.
a
d
d
y
x
y y = ( ) 21 5 . b
d
d
y
x
xy = c
d
d
y
x
xy = 2
d
d
d
y
x
x
y
= e
d
d
y
x
x y = 2 f
d
d
y
x
y y = ( ) + ( ) 21 2
7 Find an equation for the curve that passes through the point
π
6
0 ,

and for which
the slope of the curve at any point (x, y) on the curve is
2 4 y
x
+
tan
.
8 For all positive values of x the slope of a curve at the point (x, y) is given by
y
x x
2
+
.
The point P(3, 6) lies on this curve. Find:
a the equation of the normal to the curve at P.
b the equation of the curve where y is expressed in terms of x.
9 Solve the diferential equation

d
d
y
x
= y tan x + 1, 0 < x <
π
2
,
if y = 1 when x = 0.
10 Consider the diferential equation
d
d
y
x
y x
xy
=
+ 3
2
2 2
, for x > 0.
a Use the substitution y = vx to show that v
dv
d
v
v
+ =
+
x
x
3 1
2
2
.
b Hence, fnd the solution of the diferential equation given that y = 2 when x = 1.
11 Use the substitution y = vx to show that the general solution to the diferential
equation
(x
2
+ y
2
) + 2xy
d
d
y
x
= 0, x > 0, is x
3
+ 3xy
2
=

k, where k is a constant.
2 3 1 2 1
3
2
1

1
2
3
2 3 1
3
2
1

1
2
3
2 3 1
3
2
1

1
2
3
3
(1)
2 3 1 2 1
3
2
1

1
2
3
2 3 1
3
2
1

1
2
3
2 3 1
3
2
1

1
2
3
3
(4)
2 1 3
(5)
2 1 3
(6)
2 1 3
(2)
2 1 3
(3)
128
Differential Equations
4
12 Consider the diferential equation
d
d
y
x
yx
x
+ =
4
1
2
2
, 1, where |x| < 2 and y = 1 when
x = 0.
a Use Euler’s method with h = 0.25 to fnd an approximate value of y when
b i By frst fnding an integrating factor, solve this diferential equation.
ii Calculate, correct to two decimal places, the value of y when x = 1.
c Sketch the graph of y = f (x) for 0 < x < 1. Use your sketch to explain why
your approximate value of y is greater than the true value of y.
13 a Given that
x
x x x
x
x
2
2 2
1 1 1 1 ( )( ) ( ) ( )
,
+ +

+
+
+
+
a b c
calculate the value of a, b and c.
b i Hence, fnd I =

+ +
x
x x
x
2
2
1 1 ( )( )
. d
ii If I =
π
4
when x = 1, calculate the value of the constant of integration
giving your answer in the form p + q ln r where p, q, r ∈ ℝ.
14 Given that
d
d
y
x
y
xy
=
+
+
2
1
and y = 1 when x = 0, use Euler’s method with interval
h = 0.5 to fnd an approximate value of y when x = 1.
15 a Show that tan lnsec , x x x d c = +

ln sec x + C, where C is a constant.
b Hence, fnd an integrating factor for solving the diferential equation

d
d
y
x
+ y tan x = sec x.
c Solve this diferential equation given that y = 2 when x = 0. Give your
answer in the form y = f (x).
16 Consider the diferential equation
d
d
y
x
x y
xy
=
+ 3
2 2
where x, y > 0.
a Show that the diferential equation is homogeneous.
b Find the general solution of the diferential equation, giving your answer in
the form y
2
= f (x).
c Solve the diferential equation given that y = 2 when x = 1.
17 Consider the diferential equation
d
d
y
x
y x 22 = sin with boundary condition
y = 1 when x = 0.
Use four steps of Euler’s method starting at x = 0, with interval h = 0.1, to fnd an
approximate value for y when x = 0.4.
18 a Show that
d
dx
x
x x
x ln , .
1
1
2
1
1
2
+

= <
2 2
b Find the solution to the homogeneous diferential equation
x
2

d
d
y
x
= x
2
+ xy 2 y
2
given that y =
1

__

2
when x = 1.
129
19 a Use integration by parts to show that

sin cos ( sin ) .
sin sin
x x x x
x x
e d e C
2 2
2 = + +

1
Consider the diferential equation
d
d
y
x
2 y cos x = sin x cos x.
b Find an integrating factor.
c Solve the diferential equation given that y = 22 when x = 0. Give your
answer in the form y = f (x).
20 a Sketch on graph paper the slope feld for the diferential equation
d
d
y
x
= x 2 y
at the points (x, y) where x ∈ {0, 1, 2, 3, 4} and y ∈ {0, 1, 2, 3, 4}. Use a scale of
2 cm for 1 unit on both axes.
b On the slope feld sketch the curve that passes through the point (0, 3).
c Solve the diferential equation to fnd the equation of this curve. Give your
answer in the form y = f (x).
21 Consider the diferential equation
d
d
y
x
+ (2x 2 1)y = 0, given that y = 2 when
x = 0.
a i Show that
d
d
d
d
d
d
4
4
3
3
2
2
1 2 6
y
x
x
y
x
y
x
= ( ) . 2 2
ii By fnding the values of successive derivatives at x = 0, obtain a
Maclaurin series for y up to and including the x
4
term.
b A local maximum value of y occurs when x = 0.5. Use your series to calculate
an approximation to this maximum value.
c Use Euler’s method with a step value of 0.1 to obtain a second
approximation for the maximum value of y. Set out your solution in tabular
form.
d How can each of the approximations found in b and c be made more
accurate?
22 Solve the diferential equation x
y
x
y x
d
d
+ = + 2 1
2

given that y = 1 when
x = 3.
23 a A homogeneous diferential equation has the form

d
d
F
y
x
y
x
=

.
Show that the substitution v =
y

__

x
leads to a diferential equation which can
be solved by separation of variables.
b Show that the linear change of variables X = x 2 1 and Y = y 2 2 transforms
the equation

d
d
y
x
x y
x y
=
+ 3 7
3 1
2
2 2
.

to a homogeneous form.
Hence, solve this equation.
24 Consider the diferential equation

x
y
x
y
x
x
d
d
22
1
3
2
=
+
.
a Find an integrating factor for this diferential equation.
b Solve the diferential equation given that y = 1 when x = 1, giving your
answer in the form y = f (x).
130
Differential Equations
4
25 Find the exact value of
dx
x x ( )( )
.
+ +

2 2 1
0
26 A curve that passes through the point (1, 2) is defned by the diferential
equation

d
d
y
y
x
x x = + 2 1
2
( ). 2
a i Use Euler’s method to get an approximate value of y when x = 1.3,
taking steps of 0.1. Show intermediate steps to four decimal places in a
table.
ii How can a more accurate answer be obtained using Euler’s method?
b Solve the diferential equation, giving your answer in the form y = f (x).
27 Solve the diferential equation
( )( ) . x x
y
x
y x + + + = + 1 2 1
d
d
28 Solve the diferential equation x
y
x
xy x x
2 2
2
1 2 2 1 + ( ) = + + ( )
d
d
.
29 In a certain circuit the electric current I satisfes the diferential equation
L
dI
dt
RI E + =
where L, R and E are positive constants. Find I in terms of the time t given that
l = 0 when t = 0. Deduce the value of I as t →∞.
30 Solve the frst order linear diferential equation cos sin sin x
y
x
y x x
d
d
+ = 2

for
those values of x for which cos x is greater than zero.
Questions 9–30 © International Baccalaureate Organization
131
Index
Page numbers in italics refer to information boxes and tables.
Greek
e (epsilon) 5
A __________________________________________________
absolute convergence theorem 58
absolute value theorem 12–14
absolutely convergent series 56, 57
algebraic functions 63
alternating harmonic series 55, 58, 76
alternating series 51–5, 56
alternating series estimation theorem 54, 55, 81–2
alternating series test 51– 4, 55, 59, 60
B __________________________________________________
bounded sequence theorem 34, 35
bounded sequences 34
bounds (see lower bounds, upper bounds)
C _________________________________________________
calculators 63, 64
comparison test 42–5, 48, 59
completeness postulate 34, 35
conditionally convergent series 56, 57–9
convergence
improper integrals 19
power series 65–9
sequences 4
D _________________________________________________
differential equations 93–5
Euler’s method 118–22
frst order linear equations 105–11
homogeneous equations 111–17
initial value problems 117
separable equations 101–5
slope felds 96–100
divergent integrals 19
divergent sequences 4
E __________________________________________________
error term, Taylor polynomials 79
Euler’s method 118–22
explicit formulas 4
explicitly defned functions 94
exponential function (e
x
) 70–1, 77
F __________________________________________________
Fibonacci sequence 4
fnite sequences 3
frst order differential equations 94, 96, 101
frst order linear differential equations 105–11
G _________________________________________________
general solutions (differential equations) 94
geometric sequences 13
geometric series 24–5, 28, 45, 47–8, 51, 59, 60
greatest lower bounds 34
Gregory, James 72
H _________________________________________________
harmonic series 31, 41, 76
homogeneous differential equations 111–17
homogeneous functions 111–12
I __________________________________________________
implicit formulas 4
implicitly defned functions 95
improper integrals 18–21, 33
infnite series (see series)
initial conditions 95
initial value problems 117, 118, 120
integral error term, Taylor polynomials 79
integral test 34, 35–9, 39, 59, 60
integrating factors 106, 107, 108, 109, 110–11, 121
intervals of convergence 68, 69, 77
isoclines 96–7, 98
L __________________________________________________
l’Hôpital’s rule 14–18, 19, 53, 57
Lagrange error term, Taylor polynomials 79
least upper bounds 34
limit comparison test 45–7, 57, 59
limit of a sequence theorem 8
limits 4–18
absolute value theorem 12–14
l’Hôpital’s rule 14–18
squeeze theorem 10–11
linear differential equations 96, 105, 108
lower bounds 34, 39 – 40
132
133
M _________________________________________________
Maclaurin, Colin 72
Maclaurin series 72, 73– 4, 75, 76–7
monotonic sequences 34, 35
N _________________________________________________
nth term divergence test 31–3, 51, 52, 53, 57, 59, 60
Newton, Isaac 72
Numerical methods 121
O _________________________________________________
order, differential equations 93
Oresme, Nicole 31
P __________________________________________________
p‑series 40–1, 45, 48, 59, 60
partial fractions 26–7, 29
partial sums 23, 27, 30
particular solutions, differential equations 95
algebraic operations 75
construction 70– 4
convergence 65–9
differentiation 74, 75
integration 74, 76
R __________________________________________________
ratio test 48–50, 51, 59, 60
power series 67, 68, 69
rational functions
limits 10, 14
recursive formulas 4
S __________________________________________________
separable equations 101–5, 109, 113, 115
separation of variables 104–5, 113, 115, 116
sequences 3– 4
bounded sequences 34
absolutely convergent series 56, 57
alternating series 51– 4
alternating series test 51– 4, 59, 60
comparison test 42–5, 59
conditionally convergent series 56, 57–9
estimation 38– 40, 54, 55
134
Index
geometric series 24–5, 59, 60
harmonic series 31
integral test 34, 35–9, 59, 60
limit comparison test 45–7, 59
nth term divergence test 31–3, 59, 60
p‑series 40–1, 59, 60
partial fractions 27–31
ratio test 48–50, 59, 60
slope felds 96–100, 117, 118
solution curves 96, 97, 118
squeeze theorem 10–11, 12
subscripts 23
T __________________________________________________
Taylor, Brook 72
Taylor polynomials 78–81
Taylor series 72, 77, 78
telescoping series 27
terminal velocities 100
transcendental functions 63
truncation error 79
U _________________________________________________
upper bounds 34, 39–40

Series and Differential Equations
Assessment statements
10.1 Infinite sequences of real numbers. Limit theorems as n approaches infinity. Limit of a sequence. ∞ Improper integrals of the type ∫ f (x ) dx .
a

The integral as a limit of a sum; lower sum and upper sum.

10.2 Convergence of infinite series. Partial fractions and telescoping series (method of differences). Tests for convergence: comparison test; limit comparison test; ratio test; integral test. 1 The p-series, ∑ p . n Use of integrals to estimate sums of series. 10.3 Series that converge absolutely. Series that converge conditionally. Alternating series. 10.4 Power series: radius of convergence and interval of convergence. Determination of the radius of convergence by the ratio test. 10.5 Taylor polynomials and series, including the error term. p Maclaurin series for e x , sin x , cos x , arctan x , ln (1 + x ) , (1 + x ) . Use of substitution to obtain other series. The evaluation of limits of the f (x ) using l’Hôpital’s rule and/or the Taylor series. form lim x →a g (x ) 10.6 First order differential equations: geometric interpretation using slope dy = f (x , y ) using Euler’s method. fields; numerical solution of dx dy  y = f   using the substitution Homogenous differential equation  x dx y = vx . Solution of y ′ + P (x ) y = Q (x ) , using the integrating factor.
2

1

Sequences, Limits and Improper Integrals

1.1

Introduction

Important concepts regarding sequences, series and limits were covered in previous textbook chapters on the core syllabus. It would be helpful to go back and read through the first four sections of Chapter 4, especially the material on infinite geometric series in Section 4.4. The first section in Chapter 13 includes an informal approach to limits of functions and also covers properties of limits. Central to any discussion about sequences, series and limits is the concept of a function. Thus, it may also prove worthwhile to review some of the fundamental ideas, terminology and notation for functions covered in the first section of Chapter 2. Arithmetic and geometric series, both finite and infinite, were discussed in Chapter 4. In this option topic we are primarily interested in characteristics and applications of infinite series. As you will see, infinite series are mathematically interesting and have very useful applications. Our treatment of series in this option topic will require a more formal approach than taken in Chapter 4. In order to develop a rigorous theory of infinite series, we must first consider infinite sequences of numbers.

1.2

Infinite sequences

Sequences and series are closely related, so you need to be careful to apply these words correctly. A sequence is an ordered list of numbers commonly written out with commas separating the numbers. A series is a sum of a sequence. The finite sequence 1 1 1 1, , , is an ordered list 2 4 8 whereas the closely related 1 1 1 finite series 1 + + + is a sum 2 4 8 that is precisely equal to the 15 number . 8

Sequences occur in many areas of mathematics. For example, the positive even numbers less than or equal to 10 form a sequence: 2, 4, 6, 8, 10. This sequence is finite because the list of numbers ends with a specific number, 10 in this case. If a sequence does not end, it is infinite. We will be focusing on infinite sequences, so from now on if we use the word ‘sequence’ it is understood that we are referring to an infinite sequence. From the definition it is understood that an infinite sequence is a rule that associates a number to each positive integer. The number associated with the integer n is called the nth term of the sequence. Instead of using the familiar function notation f (n) to represent the value (term)
3

Definition of a sequence A sequence of numbers is a discrete function whose domain is the set of + positive integers,  .

. . }. an . otherwise we say that {an } diverges. c) The terms of the sequence {c n } =   are 1. . The sequence {an } defined by a1 = a2 = 1. This highlights the fact that although it is often 3 3 15 helpful to view some of the initial terms in an infinite sequence. n. For example. such as an or un . It produces the famous Fibonacci sequence: 1. Our foremost concern with a sequence {an } is whether an has a limit L as n approaches infinity (n → ∞ ). … . we say that {an } converges to L. b) The terms of the sequence {bn } =   are 1. 8. . . The notation {an } represents all the terms of a sequence. an + 2 = an + 1 + an for n 1 is a sequence that we saw in Chapter 4 of the book. . If it does. 1 2 6 24 120 (n 1) !  It is not necessary for a sequence to be defined by an explicit formula. Although a bit complicated. the rules for sequences and series in this chapter will be explicit rather than recursive. n ∈ + . 5. A function f whose domain is the half-open interval [ 1. 2. … . it seems appropriate that in our investigation of limits of sequences. . See Chapter 4 of the book for discussion of explicit and recursive formulae for sequences. for the sequence formed by the reciprocals of the positive integers. 1. . .) Example 2 – A sequence defined by a recursive formula 1 2 3 4  1 a) The terms of the sequence {an } = 1  are 0. 2 3 n Example 1 – Listing the terms of a sequence The first six terms of the sequence {c n } can be simplified to 4 2 4 1. . 2 3 4 5  n     2n 1  2 4 8 16 32 . 2.  and an = .…. . the 4 . or more simply with the notation {an } . … . a2 . Limits and Improper Integrals of a sequence f for a certain positive integer n. knowing the explicit formula for the value of the nth term is even more useful. . It follows that an is an explicit formula (sometimes called a closed formula) that is a function whose domain. 2 3 4 5  n n +1 ( 1)  1 1 1 1   . Since a sequence is a type of function. a2 . 2.1 Sequences. is the set of positive integers and generates the value of the nth term of a sequence. an explicit formula exists for the nth term of the Fibonacci sequence. we can apply the same ideas from our work with limits of functions in Chapter 13 of the book. it is customary to use a subscripted letter. we can 1 1 1 write {an } = 1. . i. (Note: Evaluating the first term in the sequence {c n } required using the definition that 0 ! = 1. . Hence. we will denote a sequence by {a1 . . 13.  .…. as in Example 1. .e. not just a single term. The rule giving an + 2 in terms of an + 1 and an is an example of a recursion formula. ∞ [ can be converted into a sequence by restricting its domain to the integers in that interval. In general. 3.

What precisely does ‘arbitrarily close’ mean? Let’s try to answer this question by looking at a specific sequence. if it was established that lim f (x ) = L . . we stated an informal definition of a limit of a function and five properties of limits (Section 13. In our development of the derivative through a limit process.6 0. Let’s start by using the informal definition for the limit of a function (from Section 13.1 of the book) to develop a more precise definition for the limit of a sequence. ε The fifth letter in the Greek alphabet. . 1 2 2 1 1 1 1 < an < 1 + < an < 1 + 1 for all n > 10 . . given a sequence {an }. is typically used in mathematics to represent a small but strictly positive quantity. . Therefore. 1 ε < an < 1 + ε for all sufficiently large n. Indeed. n 5 10 15 20 n We can ensure that an is within an arbitrarily small distance of 1 for all 1 1 < an < 1 + for all n > 2. sufficiently large n. then the points representing an lie between these lines for all sufficiently large n. then the limit of f (x ) as x approaches c is L.8 0. part a) – 1 2 3 4  1 namely {an } = 1  which starts with 0. … . that is.1). The graph of {an } is shown in Figure 1. We have seen a few sequences already in this section that seem to have a limit that is not infinite. ε (epsilon). it is often possible to define a function f on [ 1. y 1. one at an arbitrarily small distance of ε below 1 and the other the same distance of ε above 1.1 Graph of the first twenty terms of the sequence {an} = { } 1 1 . The idea we are developing here is that no matter how small we choose ε to be ( ε > 0 ). If we choose N to be the ε 1 smallest integer such that N .1. 10 100 10 100 etc. ∞ [ such that f (n) = an for each integer n > 0.4 0. we 1 observe that 1 ε < an < 1 + ε for all n > . if we graph a pair of horizontal lines. then 1 ε < an < 1 + ε for all n > N . Conversely. results obtained in Chapter 13 of the book for limits of functions are available for our work with limits of sequences.2 + + + + 1+ 1 1 Figure 1. Our earlier informal definition of a limit of a function said that if f (x ) becomes arbitrarily close to a unique finite number L as x approaches c from either side. Thus. For example. 5 . 1 for all n > 100. It appears that 2 3 4 5  n n→∞ n n→∞ x→∞ lim a = 1. it would necessarily follow that lim an = L . One such limit is the one in Example 1.2 + + + + + + + + + + + + + + + + 1 0.positive integers + .

lim = 0. Example 3 – Using the sequence ε – N definition to find the limit of a In previous work in the core syllabus of this course (and just from 1 common sense) we have reasoned that the limit of the function f (x ) = x 1 as x → ∞ is zero. we have an 1 < ε for all n > N . We begin by writing down the inequality from the ε – N 1 definition with an = and L = 0. The domain of f is {x : x ∈ . whereas a sequence that does not have a limit diverges. there exists a positive integer N such that an L < ε . there always exists some positive integer N such that all the terms of the sequence where the subscript is larger than N will be less than a distance of ε away from the number 1. n→∞ n n  Solution Let ε > 0. i. i. for all n > N . This gives n 1 1 an L = 0 = < ε.1 Sequences. However. It n 1 1 1 is clear that < ε for all n > . we get the 1  sequence {an } =   .e. The sequence {an} converges to the limit L if. for any ε > 0. That is. we now state the formal definition for the limit of a sequence.e. then the limit of the value of the terms in the sequence is 1. x ≠ 0}. x→∞ x If we change the domain to positive integers {n : n ∈ + }. as illustrated in Figure 1. n  1 1  Prove that the sequence   converges to zero. Using this idea.  1 Graphically. the sequence converges to the limit of 1. lim = 0.1 for the sequence 1  . Limits and Improper Integrals Applying one of the properties of absolute value inequalities (see Section 3.5 of the book). Definition of the limit of a sequence (ε – N definition) A sequence that has a limit converges. this  n ε – N definition says that eventually the terms of a sequence that converges to L will lie within the horizontal lines y = L + ε and y = L ε . n n 1 We now need to find a positive integer N such that < ε for all n > N . This is expressed as follows: Because for any given ε > 0. Essentially we have come up with a more precise way of saying that as n  1 gets very large the terms of the sequence {an } = 1  approach the fixed  n number 1. there is no reason to expect n ε ε 6 .

Therefore.  2n + 1  2 Solution We must show that for any ε > 0 there exists a positive integer N such that n 1 < ε. n  Example 4 – Using the sequence 1 . for all n > N . 4ε In Section 13. if n = 3 then n 2n + 1 1 3 = 2 2 (3) + 1 1 1 1 1 < = . that ε = . take. n n ε 1  the sequence   converges to zero. and since 2 14 14 8 then the condition an L < ε . Then N = 1  1 2 8 4   8 2 3 n 1 1 . then all n > 2 must satisfy the inequality < . ε ε – N definition to find the limit of a  n  1 Prove that the limit of the sequence   is . has been met. we used some algebraic techniques combined with some informal reasoning to find limits of rational functions similar to the rule for the sequence in Example 4. For N =2 2 2n + 1 2 8 example. we noted the 7 . but this is easily addressed by choosing any integer N Then every value of the subscript n greater than N will automatically be 1 1 1 greater than .1 of the book. The proof shows that this condition is met for all n > N so long as N is chosen so 1 2ε that N . if n is an integer and n > N then 2n + 1 2 Some algebra working produces the following: n 2n + 1 1 2n 2n 1 1 1 = = = <ε 2 2 (2n + 1) 2 (2n + 1) 2 (2n + 1) ⇒ 2n + 1 > 1 2ε ⇒ n> 1 2ε 4ε 1 2ε If we choose a positive integer N and n > N then it follows that 4ε n 1 < ε. If we then choose instance.to be an integer. It follows that 0 = < ε for all n > N . And in Example 3.  2n + 1  2 Note: To see how the ε – N definition works in this case. 2n + 1 2 Thus the definition of a limit of sequence is satisfied and the limit of the  n  1 sequence   is . for  1 3 1 2   8 1 4 = 3.

Rational power of a sequence: n→ ∞ n→ ∞ n→ ∞ n→ ∞ lim c = c lim (c ⋅ an ) = cL lim (an ± bn ) = L ± K lim (an ⋅ bn ) = LK a  L lim  n  = .1 Sequences. If lim f (x ) = L. and f (x ) approaches a limit L as x → ∞. Limits and Improper Integrals 1 1  is associated with the sequence   . then the sequence will converge to the same limit L. 2. All of these can be translated into properties for limits of sequences. That is. 5. 4. and c is n→ ∞ n→ ∞ any real number. Scalar multiple of a sequence: 3. a convergent sequence does not imply that the associated real variable function must also converge. then: 1.1 we presented a set of five properties for limits of functions. p ∈  These six properties of limits of sequences can be stated in words as follows: 1. We list here the set of five corresponding properties of limits of sequences and an additional important property on the limit of a rational power of a sequence. 8 . The limit of a product of sequences is the product of the limits of the sequences. It seems reasonable x n  to conjecture that for a sequence {an } if the value of an matches a function function f (x ) = f at every positive integer.  bn  K p n→ ∞ K ≠0 n→ ∞ lim (an ) = Lp . Constant sequence: 2. The limit of a sum/difference of sequences is the sum/difference of the limits of the sequences. Sum or difference of sequences: 4. The limit of a constant is equal to the constant. The limit of a constant times a sequence is the constant times the limit of the sequence. 3. Also. The limit of a quotient of sequences is the quotient of the limits of the sequences (given that the limit of the sequence in the denominator is not zero). The limit of a rational power of a sequence is the rational power of the limit of the sequence. Product of sequences: 5. Quotient of sequences: 6. and {an} is a sequence such that an = f (n) when n k . then lim an = L . k ∈ + . 6. in Section 13. Limit of a sequence theorem Suppose that f (x ) is a function defined for all x k . x→∞ n→ ∞ The converse of the limit of a sequence theorem is not true. Properties of limits of sequences If {an} and {bn} are convergent sequences such that lim an = L and lim bn = K .

we can conclude 3n 2 + 5n 1  3 that the sequence   is convergent and it converges to . k ∈ .e. lim 3x 2 + 5x 1 3 = . Hence. x →∞ 2 2x 2 + 1 Therefore. approach x 1 zero as x goes to zero. combined with property 6 for limits n→∞ n   of sequences above. we found lim to be x→∞ 2x 2 + 1 3 equal to as follows: 2 Dividing numerator and denominator 3x 2 5x 1 by largest power of x. i.1. x2. Note: This rule is equivalent to lim nr = 0 if r < 0. lim   = 0. i. then lim n→ ∞ 1 = 0 .4 of the book. k ∈ . r ∈  . 2 2  2n + 1  In our discussion of the end behaviour of rational functions in Section 3. We state them here because by means of the limit of a sequence theorem they can also be 9 . + 2 2 2 3x 2 + 5x 1 x lim = lim x 2 x x→∞ x→∞ 2x 2 + 1 2x 1 + 2 x x2 = lim 3+ 5 x 1 x2 5 x x→∞ 1 x2 lim  f (x )  x → a f (x ) Applying lim   = lim g (x ) and x → a g (x )   x→a x→a x→a x→a x→∞ 2+ = x→∞ lim 3 + lim x→∞ lim 1 x2 lim [f (x ) ± g (x )] = lim f (x ) ± lim g (x ) . leads to the following intuitive rule for the limit of certain sequences. part d) of Section 13.In Chapter 13 of the book we reasoned informally that function values for 1 functions in the form f (x ) = k . 1 lim 2 + lim 2 x→∞ x→∞ x 3+0 0 = 2+0 Applying lim 1 x→∞ xk = 0 . it makes sense that the x→∞ x 1  result from Example 3. the following limit results were hinted at. Solution 3x 2 + 5x 1 In Example 4. lim k = 0.e. Thus. If r > 0. n→ ∞ nr Example 5 3n 2 + 5n 1  Determine whether the sequence   is convergent or 2  2n + 1  divergent. from the limit of a sequence theorem above. where k is a rational number.

If n > m. it follows that 1 1 for all integers n 1. N ∈  . then lim R (x ) = ∞ . + N . i. such as the sequence in Example 5. x →0 x The squeeze theorem for sequences Note that the terms of sequence {bn} do not need to lie between {an} and {cn} for all values of n. ‘squeezing’ the sequence {bn}. We proved in Example 3 that 0 n n 2 1 1 1 lim = 0. 1  a)  n  2  cosn  b)    n  Solution a) Because 2n > 0 and 2n > n for all positive integers n. both with limit of L. If n = m.2 where we used it to sin x prove that lim = 1. Limits and Improper Integrals applied in finding limits of sequences with rules that are rational functions. Therefore. The requirement is that there must be some value of n for which all of the terms of {bn} beyond this value must lie between {an} and {cn}. does not exist. and lim an = lim cn = L then lim bn = L. 2  10 . then lim R (x ) = x→∞ x→∞ x→∞ Another useful limit theorem for functions that can be rewritten for sequences is the squeeze theorem from Section 13.e. an . lim n = 0 because lim 0 = lim = 0 and the n→∞ n n→∞ 2 n→∞ n→∞ n 1  sequence  n  converges to zero. If an bn cn for all n such that n See Figure 1. This is illustrated in Example 7. and find its limit.1 Sequences. 2. then lim R (x ) = 0. bm 3. Limits of rational functions Let R be the rational function given by anx n + an − 1x n 1 +  + a1x + a0 f (x ) R (x ) = = g (x ) bm x m + bm 1x m 1 +  + b1x + b0 1.2 The sequences Example 6 – Applying the squeeze theorem Show that each of the sequences converges. If n < m. n→ ∞ n→ ∞ n→ ∞ an L bn cn n {an} and {cn}.2 below. Figure 1.

it follows that cos n 1 cos n 1 = 0 because for all integers n 1. there is a very useful theorem that states that if the absolute value 11 1 . n→∞  n n→∞ n  n  Example 7 – Applying the squeeze theorem for an alternating sequence ( 1)n    Consider the infinite sequence  . also n !  n !     1 1 is true for all n 4. 2 6 24 120 720 The sequence clearly alternates between positive and negative terms. In converges to 0 since the inequality 0 n ! 2n fact. 1. . . That is. . 64 2 4 8 16 32 64 n  ( 1)   Observe that for n = 1. both of which converge to 0. b) Use the squeeze theorem to show that the sequence converges to 0. we need to find two convergent sequences that converge to 0 for which all terms for n N of the ( 1)n    sequence   will be between. b) In order to apply the squeeze theorem. Two sequences that will work in this n!      1 1  case are  n  and  n  . Therefore. . . respectively. 8 1 . Solution a) The first six terms of the sequence are 1 1 1 1 1 . . 16 1 . 2 1 . . 4 1 . by the squeeze theorem it follows that the sequence    n!    converges to zero. The squeeze theorem can also be used to prove that the sequence of  ( 1)n   1    absolute values for the sequence in Example 7. . n!     a) Write out the first six terms of the sequence.b) Because 1 cos x 1 for all real numbers x.n 2n 4. are 1 1 1 1 1 1 1 and . 32 ( 1)n n! ( 1)n    Therefore. lim n→∞ n n n n 1 cosn   1 lim   = lim = 0 and the sequence   converges to zero. however they are for n 4. 2 and 3.  2  2  1 2n 1 .   =  .  2  2  The first six terms of these two sequences. . . the terms of   are not between    n!   1 1   n  and  n  .

we used the squeeze theorem to prove that lim r n = 0 n →∞ 1 when r = .  1  n    1  The sequence  n . for different positive values of r. one with all positive terms and one with all negative terms. for r > 1. if lim an = 0 it does n→ ∞ not necessarily follow that lim an = 0.g. n→ ∞ n→ ∞ The converse of the absolute value theorem is not true. n→ ∞ Proof of the absolute value theorem is fairly straightforward. f (x ) = 2x . is the geometric 2 sequence {r n } convergent? Figure 1. {r n }. {r n } . and clearly the sequence does not converge to any number. x ∈ .3 Graph of the sequence {r n} for different positive values of r. other than . It is often more efficient to consider the sequence of absolute values and then apply the following theorem to the original sequence. e. 12 . r.3 shows the graphs of geometric sequences. Visually it appears that for 0 < r < 1. in Example 6 part a) is a  2  2      1 geometric sequence with a common ratio. is discrete points because the domain consists of only positive integers. Consider the two sequences { an } and { an }. 1 r=1 0<r<1 1 n n →∞ When r > 1 the sequence {r n } increases without bound. There is no graph of {r n } for r = 1. if lim an = 0 then lim an = 0.1 Sequences. In n →∞ Example 6. the terms would oscillate infinitely between 1 and 1. Because both of these sequences converge to 0 and an an an we can conclude by means of the squeeze theorem that {an } must also converge to 0. i. for different negative values of r.  ∞ if r > 1  Thus.e. is a continuous smooth curve. equivalent to   . Absolute value theorem For the sequence {an}. lim r n = 0. an r>1 Figure 1. In this case. Limits and Improper Integrals sequence converges to 0. equal to . n ∈ + . part a).g. Note that the graph of a real-valued function. then the original sequence consisting of positive and/or negative terms also converges to 0. we have lim r n =  n →∞   0 if 0 < r < 1 What about negative values of r? Figure 1. That is. {an} = {2n} . lim r n = ∞ . We can use a similar argument to show that lim r n = 0 n →∞ 2 for any value of r in the interval 0 < r < 1. e. It was shown 2 1 to converge to zero. the graph of a sequence.4 (page 13) shows the graphs of geometric sequences. For what values of r. however.

we can write the following: 13 . and since 0 < r < 1. Thus. Therefore. With the intention of applying the absolute value theorem. for r < 1. We now focus our attention on all n→∞  N  of the values of n such that n > N . Because N > x then N    N   n x and it must follow that lim   = 0. This result is summarized as follows. For these values of n. by the absolute value theorem and the obvious n →∞ n →∞ fact that lim 0n = 0 it is true that lim r n = 0 for the interval n →∞ 1 < r < 1.an Figure 1. Thus the sequence {r n } is convergent for the interval 1 < r 1 and divergent for other values of r. lim r n does n →∞ not exist. Additionally. we can conclude that lim r n = 0. then the terms of the sequence will be alternately positive and negative. This takes a bit of work. 1 1 r < 1 1 < r < 0 n Clearly. Convergence of geometric sequences theorem  0 if r < 1  lim r n =  n→ ∞  1 r =1  For r ∈ » and n ∈ »+ the geometric sequence {r n} is convergent for 1 < r 1 such that Example 8 – The factorial function and exponential functions x n  Show that the sequence   converges to 0 for any real number x. all n x = 0. We that needs to be shown is that lim n→∞ n ! start by choosing some positive integer N such that N > x . lim r n = lim r . which is equivalent to r < 1. n !  Solution If x < 0.4 Graph of the sequence {r n} for different negative values of r. It is also obvious that lim 1n = 1. Using the result above that n →∞ n →∞ lim r n = 0 for 0 < r < 1. It follows  x  n  x   <1 that the sequence    is geometric. Considering 1 < r < 0 we can also write the inequality as n n →∞ n →∞ 0 < r < 1. when r < 1 the sequence {r n } alternates between positive and negative values that increase without bound.

Limits and Improper Integrals x = n! 1 × 2 × 3 x n! n n x N × ( N + 1) ( N + 2 ) (n − N ) factors n n n x N !N n n N n n x NN NN  x  = =   N!  N  N !N n n Hence. lim n n = 0. n→ ∞ x x NN  x  Thus.1 Sequences. we were sometimes confronted with an expression of indeterminate form. by the absolute value theorem lim n→∞ n ! x n    converges to 0 for any real value of x. ∞ 0 commonly in the form or . and consequently the limits of certain sequences. L’Hôpital’s rule Let f and g be functions whose derivative can be found at any value in an open interval ] a . L’Hôpital’s rule states simply that. the limit of a quotient of functions is equal to the limit of the quotient of their derivatives. n! we can conclude that the factorial function increases faster than any exponential function. b [ . Assume that g′ (x ) ≠ 0 . The expression   N! N! N n x change as n changes. as illustrated in Example 5 of Chapter 13. except possibly at c. n→∞ n→∞ n→∞ N !  N    N ! n→∞  N  N! Because we have shownn that for any number x. Suppose that lim f (x ) = 0 and lim g (x ) = 0 . Then lim x → c g (x ) x → c g′ (x ) n n n When you are applying l’Hôpital’s rule make sure that you differentiate the numerator and denominator separately. except possibly at some value c where a < c < b . It is important to first verify the conditions regarding the limits of f and g before applying l’Hôpital’s rule. the sequence Therefore. given the right conditions. or x→c x→c f (x ) lim f (x ) = ± ∞ and lim g (x ) = ± ∞. N!   xn = 0. Not all limits can be managed in such a way. The following theorem specifically addresses limits of rational expressions that are of indeterminate form. 14 . n !  x n! We have one more important theorem to consider that is an essential tool for helping to determine the limits of certain functions. 0 lim  N  = 0 and we can conclude that n → ∞ n ! = 0. With limits of rational functions in Chapter 13 of the book. so applying the property n→∞  N  NN  x  NN NN x lim (c ⋅ an ) = c lim an we get lim lim   = = ( 0) = 0. 0 NN NN  x  is a constant and will not .) 0 ∞ f (x ) f ′ (x ) = lim provided the limit on the right side exists (or is infinite). (That is. We know that lim   = 0. the expression is in indeterminate x→c x→c g (x ) 0 ∞ form of or . We handled these by performing some ∞ 0 algebraic manipulations and applying limit theorems. Do not use the quotient rule for differentiation.

01 X=1. x →  2  2   15 .Example 9 – Applying l’Hôpital’s rule For each limit. and then find the limit using l’Hôpital’s rule.5 ∆Tbl=.98938 . x a) lim x →0 1 ex sec x b) lim π 1 + tan x x→ c) lim (e x + x ) x x →0 2 1 Solution a) To visualize lim images below.56 1. use your GDC to conjecture a result.58 1. 1) implying that lim = 1.9992 1.97057 . and l’Hôpital’s rule applies.0093 1.54 1. lim is in the indeterminate x →0 x →0 1 x →0 ex 0 form .0305 Y1 The values in the table show that the function appears to be approaching 1 from either direction. Differentiating the numerator 0 and denominator separately and evaluating the limit gives 1 1 x lim = lim x = = 1. Since x →0 1 ex x lim x = 0 and lim (1 e x ) = 0 . 2 Plot1 Plot2 Plot3 Y1= (1/cos(X))/( 1+tan(X)) Y2= Y3= Y4= Y5= Y6= TABLE SETUP Indpnt: Auto Ask Depend: Auto Ask Tb1Start=1.57 1. x →0 x 1 e x we graph f (x ) = x 1 ex as shown in the GDC Plot1 Plot2 Plot3 Y1= X/(1-eˆ(X)) Y2= Y3= Y4= Y5= Y6= Y7= WINDOW Xmin=-2 Xmax=2 Xsc1=1 Ymin=-3 Ymax=1 Ysc1=1 Xres=1 Although x = 0 is not in the domain of f.6 X . let’s use the GDC to construct a table of function values lim π 1 + tan x x→ 2 π near x = ≈ 1. the graph appears to x pass through the point (0.55 1.0198 1.e. π The values of sec x vanish to + ∞ when x → from the left 2 − +  π  π and vanish to − ∞ when x → . i. x x →0 1 x →0 e 1 e sec x b) Instead of viewing a graph of f (x ) = to conjecture a value for 1 + tan x sec x .97984 .59 1.5708 .6 1.

the quotient of the derivatives remains in indeterminate form the rule can be applied more than once. ln [ f (x )] = x →0 ln (e x + x ) x .2735696 Tracing on the graph indicates that as x → 0 the function approaches a value between 7.2735 and 7. lim+ π 1 + tan x π π sec 2 x x→ x→ x→ 2 2 2 sec x tan x   simplifies to sin x .e. therefore the two-sided limit is equal to 1. 16 . by taking the logarithm of both sides of f (x ) = (e x + x ) and then taking the limit we can change the indeterminate form to 0 0. The exact value of the limit is not clear. apply l’Hôpital’s rule. then lim (e x + x ) = lim e x →0 x →0 1x = lim f (x ) x →0 ln f (x ) Applying the rule eln a = a. However. after applying l’Hôpital’s rule. ln e x + x ) (e x + x )1 x  = 1 ln (e x + x ) = ( ln [ f (x )] = ln   x x 1x We observe that lim (e x + x ) is in the indeterminate form 1∞ .0106383 Y=7.0106383 Y=7. Limits and Improper Integrals Similarly. 1x x →0 Thus. sec x = 1. and when approaching from the left we have +∞ −∞ 2 from the right. i.5 Xsc1=1 Ymin=-2 Ymax=14 Ysc1=1 Xres=1 Y1=(eˆ(X)+X)^(1/X) X Y1=(eˆ(X)+X)^(1/X) X X=-.5 Xmax=.5094925 X=. lim π 1 + tan x x→ 2 1x x →0 c) To visualize lim (e x + x ) GDC images below. produces lim ln [ f (x )] = lim x →0 ex + 1 x e0 + 1 = lim e + x = 0 = 2. x →0 = lim e 2 . Applying the rule to the right-hand limit gives sec x tan x sec x = lim+ = lim+ sin x = 1. lim− (1 + tan x ) = + ∞ and lim+ (1 + tan x ) = − ∞.1 Sequences.5094.  Note :   sec 2 x The left-hand limit is also 1. to which we can apply l’Hôpital’s rule. Plot1 Plot2 Plot3 we graph f (x ) = (e x + x ) 1x as shown in the Y1= (eˆ(X)+X)ˆ(1 X) Y2= Y3= Y4= Y5= Y6= WINDOW Xmin=-. and taking the limit as x → 0 of both sides ln (e x + x ) x Right side in the form 0 0. So when x→ π 2 x→ π 2 approaching π +∞ −∞ . x →0 1 e +0 Hence. L’Hôpital’s rule also applies to one-sided limits. x →0 1x x →0 Since f (x ) = (e x + x ) . lim ln [ f (x )] = 2. If. Using result lim ln[f (x )] = 2.

so we apply l’Hôpital’s rule a second time. the sequence  n  converges to 0. find its  3  limit. e 2 ≈ 7. so the limit is within the range estimated from the graph on the GDC. Solution x2 + 1 Consider the function f (x ) = . and its limit as x → ∞. lim (e x + x ) x →0 1x = e 2.  3  Example 11 Which sequence grows faster.).389 (to 4 s. n→∞ 3 n 2 + 1  Therefore. {lnn} or Solution { n }? We can gain some insight into this question by graphing the real-valued functions y = ln x and y = x . Using l’Hôpital’s rule to show that the limit of the function ln x f (x ) = is 0 as x → ∞ will prove this result. The graph (page 18) implies that the  lnn  sequence n grows faster than {lnn}. we can apply the limit of a sequence theorem and conclude that n2 + 1 lim n = 0. x2 + 1 2x 2 lim x = lim = lim =0 x x→∞ x → ∞ (ln 3) 3 x→∞ 3 (ln 3)2 3x Because the value of an matches the value of f (x ) for every positive integer. we can apply l’Hôpital’s x→∞ 3 rule. x ∈ . x2 + 1 2x lim x = lim x→∞ x → ∞ (ln 3) 3x 3 But this limit is still in indeterminate form of ∞ ∞. the infinite sequence    n converges to 0. Example 10 – Using l’Hôpital’s rule to determine convergence of a sequence n 2 + 1  Determine if the sequence {an } =  n  converges.f.Therefore. x { } 17 . that is. 3x x2 + 1 Since lim x is in indeterminate form of ∞ ∞. If it does.

ln x Improper integrals and infinite limits of integration ∫ f (x ) dx . →∞  →∞ →∞      18 . the sequence   converges to 0.e. Let’s look at an integral where one of the limits is infinite. we will look at ways of evaluating integrals where either one or both of the limits of integration (i. Limits and Improper Integrals y 5 4 3 2 1 y= x y = ln x 5 10 15 20 25 x { } 1. bounded. In this section. x2 We can replace the infinite limit of integration with a variable. for a function b a Previously we have defined the definite integral. or the function f has an infinite discontinuity in the interval a x b . a and b) are infinite. say the variable b. no ‘gaps’ in the domain) for the finite. k ∈  + x→∞ x→∞ x→∞ x x x→∞ 1 x 2 x  lnn  Therefore. and we can conclude that  n n grows faster than {lnn}.e. b ∞ 1  b 1   1   1 +1 Taking the limit as b → ∞ gives ∫ 2 dx = lim  ∫ 2 dx  = lim   = lim  1 x b→∞  1 x  b→∞  x 1 b→∞  b   b ∞ 1  b 1   1   1 ∫ 1 x 2 dx = blim  ∫ 1 x 2 dx = blim  x  1 = blim  b + 1 = 0 + 1 = 1. Example 12 Evaluate ∫ Solution ∞ 1 1 dx or show that it diverges. f that is continuous (i. and then take the limit of the integral as b approaches infinity.1 Sequences. interval a x b .3 1 2 1 lim = lim x = lim = 0 because lim k = 0. An integral having either one of these characteristics is called an improper integral.

x Solution 1 b  b1  dx = lim  ∫ dx  = lim [ln x ]1 = lim (ln b 1 x b→∞  1 x b→∞  b→∞ [or ‘limit does not exist’] ∫ ∞ ln 1) = lim (ln b) = ∞ b→∞ ∫ x d x = ln x . b→∞ e e 1 1 and now need to apply integration by parts to evaluate the integral. but note that in Example 13 the absolute value is omitted because the integral is being evaluated from 1 to ∞. not all improper integrals converge to a finite value. i. and is called divergent if the limit does not exist (as in Example 13). b Example 14 – Using l’Hôpital’s rule to evaluate an improper integral ∞ Determine whether the integral converges. Example 13 Evaluate ∫ ∞ 1 1 dx or show that it diverges.e.y 1 Therefore. 1 1 Therefore.5 Area under the curve 1 y = 2 from 1 to ∞. The improper integral ∫ f (x ) dx is called convergent if the corresponding a limit exists (as a finite number as in Example 12). ∫ x dx = lim ∫ x dx . Solution ∫e 1 x x dx converges or diverges. then lim ∫ x dx = lim  xe 1 + ∫ e Let  b→∞ e b→∞ du = dx v = e x  1 1 x ∞ b  dx   = lim  xe x e b→∞  = lim  (x + 1) e   = lim  b→∞  b→∞ x b 1 x b (b + 1) + 2 eb  e 19   1  . ∫ 2 dx = 1. the integral diverges. b b  x u = x dv = e x dx x b . and if it x x We can rewrite the integral as a limit. find its value. only positive numbers.5 right). 1 x This result can be interpreted as indicating that the 1 area under the curve y = 2 from one to infinity x is finite and is exactly equal to 1 (see Figure 1. x 1 2 3 4 5 x Certainly. ∞ 3 2 y= 1 x2 1 Figure 1. The area under the curve y = from 1 to x infinity is infinite.

Limits and Improper Integrals  (b + 1)  2 = lim  + b→∞   eb  e 2 = 0+ e ∞ x 2 Therefore.7357588823 (to ten significant figures). The integral converges at a fairly quick rate.6 y 5 4 3 2 1 y= 1 1  x2 1 x 20 . e e 1 The GDC images below confirm our result.05 Ymax=. Plot1 Plot2 Plot3 Y1= X/eˆ(X) Y2= Y3= Y4= Y5= Y6= Y7= WINDOW Xmin=0 Xmax=12 Xsc1=2 Ymin=-.1. Figure 1.73567901 What is an infinite discontinuity? A function f has an infinite discontinuity at x = c if either lim f (x ) = ∞ or lim f (x ) = ∞ such that x → c x →c x →c from the right or left. ∫ x dx = ≈ 0. the 1 function f (x ) = has an infinite discontinuity at x = 1 because 1 x2 1 lim = ∞ (note: x → 1 from the left) which can be observed in − x →1 1 x2 the graph in Figure 1.4 Ysc1=.X. For example.7357588391 f(x)dx=.20) . and when the upper limit is 20 (computed on home screen) the values agree to six decimal places.7357588823 fnInt(Y1.1 Sequences.6. Note that even with an upper limit of just x = 12 the definite integral (computed on graph screen) agrees to three decimal places with the value of the ‘improper’ integral with an infinite upper limit.1 Xres=1 CALCULATE 1:value 2:zero 3:minimum 4:maximum 5:intersect 6:dy/dx 7:f(x)dx 2/e . on the interval 0 x 1.

if possible (not possible if it’s infinite).) (Recall that the anti-derivative of 1 x2 1  b  1 1 b dx = lim  ∫ dx  = lim [arcsin x ]0 = lim (arcsin b Area = ∫ − 2 2 0 b →1  0 b → 1−  b →1− 1 x 1 x π = lim (arcsin b 0) = lim (arcsin b) = arcsin (1) = − − b →1 b →1 2 1 Therefore. 17 a Use the fact that lim 2 = 0 to prove that lim 2 n→ ∞ n n→ ∞ n + 1 2 n b Prove lim 2 = 1 by an ε – N proof. 16 Use the squeeze theorem to show that lim n→ ∞ n 1 n2 = 1. under the curve 1 in the interval 0 x 1. If it converges. 2 Exercise 1. find the limit of the sequence. y= 1 x2 Solution We can replace the limit of integration where the infinite discontinuity occurs with a variable. Example 15 The region under the curve y = 1 Find the area. However. in this case). n→ ∞ n + 1 13   3   n!  { } { } n 1+ n ln 2n ln n  3 n2 + 1     3 2  4 n 1   9   2n    n + 1    1 2 n  12 1 +       n    15   21 . 1 { } 3 7 n 2  2n2 + n + 1  2  n +1  ( 1)n + 1    2n 1  3 { 5n 13 n3 + 5n } 4 {cos nπ} 7  2 10 1 +  n 5  8  6      n 4    5  e  n  n  ( 1)n   n  11 14  n + 1    n  sin 2n = 0. the unbounded region has a finite area and we can find the exact area as follows. 1 is arcsin x.1 arcsin 0) For questions 1–15. at first thought. say the variable b. determine if the sequence converges or diverges. and then take the limit of the integral as b approaches the value of x where the discontinuity occurs (approaching 1 from the left. have an infinite area. the unbounded region under the curve y = in the 1 x2 π interval 0 x 1 has a finite area of exactly .in the interval 0 x 1 is 1 x2 unbounded – and would.

) convergent. (Hint: Rewrite n sin as 1 n n In questions 23–28.1 Sequences. 30 32 34 36 0 ∫ ∫ ∫ ∫ 1 0 1 x 3 dx 31 33 35 37 ∫ ∫ ∫ ∫ ∞ 1 ∞ 0 ∞ 0 k 0 1 x3 dx ∞ −∞ 1 dx x +1 2 sin x dx ex 1 dx 1+ ex π 2 0 1 0 tan x dx 1 1 x dx x k 2 x2 dx 1 38 a Find the area of the region bounded by x = 1. or identify as divergent.) In questions 30–37. Then you can 0 use l’Hôpital’s rule. converting the right side to the indeterminate form . evaluate the limit. y = . evaluate. 22 Determine whether the sequence n sin { } 3 x → 1  ln 1 x x 1 1   23 lim x2 1 x →1 x 4x 3 2 24 lim 1 x 1 x →0 1 x cot x lim 25 x → 0 x sin x x 3 lim 26 x → 0 x x bx 27 lim ln(x + 1) x → ∞ log x 2 x→∞ 28 lim ax x →0 x . 22 . use l’Hôpital’s rule to find the value of each limit. Limits and Improper Integrals 18 Use the ε – N definition for convergence of a sequence to prove that the n sequence {1 + ( 1) } is divergent. (Hint: Start by taking the natural logarithm of both sides. find lim f (x ). b > 0 29 Given f (x ) = (1 + x )1 x . If n π sin π n . c Comment on the results for a and b. x b Find the volume of the solid generated by rotating this region about the x-axis. 19 lim 1 cos x x →0 x2 20 lim x →1 x 1 x2 + 3 2 21 lim   π is convergent or divergent. and the x-axis. For questions 19–21. a > 0. the given integral. find its limit.

+ an + . .2 Series and Convergence 2. but we must be careful not to assume that the ‘+’ signs have the same properties to which we are accustomed.4 (HL book) in our discussion on infinite geometric series.1 Introduction To start our study of infinite series in the option topic we consider using the terms of a sequence {an } to form the sequence {sn } of partial sums of {an } as follows: s1 = a1 s 2 = a1 + a2 s3 = a1 + a2 + a3 We can use sigma notation to write the general expression for sn : sn = a1 + a2 + a3 + . . Ordinary addition of real numbers is a finite process. For example. . we say that its limit S is the    i =1  ∞ sum of the infinite series a1 + a2 + a3 + . To be more precise. a rearrangement of the terms of a convergent series may change the value of its sum or even cause the series to diverge. . b) Consider the infinite series ∑ (21) n =1 ∞ n +1 = 1 21 + 1 21 + 1 21 + … . and have used the letter n as a subscript to indicate the nth partial sum. . it does not make sense to find the ‘sum’ of infinitely many terms. As pointed out in Section 4. We can write the sum as a1 + a2 + a3 + . . the ‘sum’ of an infinite series is a limit – that is. Example 1 a) Find the sum of the finite series ∑ (21) n =1 6 n +1 . If the sequence {sn} diverges then we say that the infinite series ∑ a also diverges. and we write S = ∑ ai . hence. . + an = ∑a i =1 n i Definition of the sum of an infinite series  n   If the sequence of partial sums {sn} = ∑ ai  converges. the limit of the partial sums for the series. i =1 i ∞ i =1 Here we have used the letter i as a subscript to indicate the ith term of a sequence. 23 . . the word ‘sum’ here is being used in a completely different way from how it is normally used. Determine if the series converges to a sum or diverges. You need to be comfortable with using different letters for subscripts. as we will see.

= ∑a r n =1 1 ∞ n 21 . For example. to argue that the sum of this infinite series is 0. n i + 1 For the sequence {sn } = ∑ (21)  . what is its sum? Geometric series There is one type of infinite series with which we are already familiar – and for which we know how to answer questions regarding convergence/ divergence and computing sums. not just six. Although the associative property works for finite sums it is clear that it does not work for infinite sums. + a1r n 2 1 + . i =1    etc. then an infinite geometric series can be generalized as follows: a1 + a1r + a1r 2 + a1r 3 + . . ∑ (21) ∑ (21) n =1 ∞ n +1 b) It is very tempting to use the same strategy of ‘pairing up’ consecutive terms in this manner n +1 = (1 2 1) + (1 2 1) + (1 2 1) + . 24 . a1 ≠ 0 Let’s consider three cases: r = 1. ∑ (21) n =1 6 n +1 = 1 2 1 + 1 2 1 + 1 2 1 = 0. there are commonly two basic questions: Does a particular series converge or does it diverge? If it does converge. if the number of terms is odd the sum is always 1. and r ≠ ±1. Consider that if we leave out the first term and start ‘pairing up’ from the second term we will obtain a different sum. s 2 = 0. The sum of an infinite series is defined to be the limit of the sequence of partial sums. If a1 represents the first term and r is the number that multiplies a term to obtain the next term in the series.2 Series and Convergence Solution a) Clearly. the series has no sum and it diverges. Clearly this sequence is not converging to a limit. and. In studying infinite series. The associative property of addition is what allowed us to ‘pair up’ the numbers for the finite sum in part a). we can ‘pair up’ consecutive terms to get zero. this is erroneous. s 4 = 0. or = (1 2 1) + (1 2 1) + (1 2 1) + 1 = 0 + 0 + 0 + 1 = 1. However. r = 21. Therefore. we have s1 = 1. . the sum is always 0. In either case. . s3 = 1. . . We can make the further observation that if the number of terms in this finite series was any even number. ∑ (21) n =1 7 n =1 6 n +1 = (1 2 1) + (1 2 1) + (1 2 1) = 0 + 0 + 0 = 0. and this is infinite geometric series that we encountered in Chapter 4 of the book. .

it can be assumed that it is an infinite geometric series. then the nth partial sum is sn = a1 + a1 + a1 + . . In this chapter. If r = 21. n→∞ 12 r This rigorously confirms a result that appeared in Chapter 4. so the geometric series also diverges for this case. which can also be written as 21 < r < 1. the interval r < 1. Multiplying through by r gives rsn = a1r + a1r 2 + . the nth partial sum is . r. . . we can determine which geometric series converge and which ones diverge. By identifying the value of the common ratio. The sequence of partial sums is not converging to a limit.If r = 1. Clearly the sequence of partial sums. when we refer to a geometric series. If r ≠ ±1. will behave in the same way as in Example 1 b) with s1 = a1 . 1 2r We know from the theorem for convergence of geometric sequences in the previous section that if r < 1 then r n converges to 0 as n → ∞. and diverges if r  1. 25 . if r < 1 then lim sn = 1 . Thus. is known as its interval of convergence. s 2 = 0. The sequence of partial sums. and for ones that a converge we can easily compute the sum with the formula S∞ = 1 . s 4 = 0. lim = lim Convergence of geometric series The geometric series with common ratio r a1 + a1r + a1r 2 + a1r 3 +  + a1r n 2 1 +  = ∑ ar n 2 1 n =1 a1 if r < 1. + a1r n 2 1 . {sn }. sn = a1 (1 2 r n ) a1 (1 2 r n ) This result answers the two basic questions about geometric series. s3 = a1 . . 12 r For any geometric series. + a1r n − 1 + a1r n . and we state it again here. will increase without bound and the geometric series diverges in this case. then sn = a1 + a1r + a1r 2 + a1r 3 + . {sn }. . + a1 = na1. then the nth partial sum is sn = a1 2 a1 + a1 2 a1 … . Subtracting the second equation from the first produces Factorizing yields sn (1 2 r ) = a1 (1 2 r n ) . We can apply this fact and some properties of limits to give the following result: a rn a a a1 2 a1r n a a  a  = lim 1 2 lim 1 = 1 2  1  lim r n = 1 2 0 = 1 n→∞ n→∞ 1 2r n→∞ n→∞ 12r 1 2r  1 2r  n→∞ 1 2r 1 2r 1 2r 1 2r a Therefore. converges to the sum 12 r ∞ sn 2 rsn = a1 2 a1r n . .

Also. for this course. Therefore. x + 1 x 21 x + 1 x 21 x 21 x + 1 x 2 21 x 2 21 x 2 21 The reverse process. 2 3 2 x 21 3 x + 1 2x 2 2 3x + 3 5x + 1 + = ⋅ + ⋅ = + = . we wish to find the constants A and B so that the equation will be true for all x. and then assume that there are constants A and B such that 1 A B = + . (x 2 8)(x 2 6) x 2 8 x 2 6 An obvious way to check whether you correctly decomposed a fraction is to combine the partial fractions in your result by means of getting a common denominator. Example 2 – Partial fractions and integration Evaluate Solution ∫x 2 1 dx . in which a fraction with a factorable denominator is split into the sum of two or more fractions (often called partial fractions) whose denominators are the factors of the original denominator. 1 1 1 1 1 1  1  2 2 ∫ x 2 2 14x + 48 dx = ∫  x 2 8 2 x 2 6 dx = 2 ∫ x 2 8 dx 2 2 ∫ x 2 6 dx   1 1 = ln x 2 8 2 ln x 2 6 + C 2 2 1 x 28 = ln + C. 1 For x = 6: 1 = A (0) + B (22) ⇒ B = 2 2 1 For x = 8: 1 = A (2) + B (0) ⇒ A = 2 Thus. the technique of decomposing a fraction into partial fractions will be applied in two different situations: (1) integrating a rational function. 26 This equation is an identity and.2 Series and Convergence Partial fraction decomposition The process by which a sum of two or more fractions is combined into a single fraction by obtaining a common denominator is a basic skill of elementary algebra. we start by factorizing the denominator. we can substitute any value in for x. Note that the factors are linear and none = 2 x 2 14 x + 48 (x 2 8)(x 2 6) are repeated. 1 1 . For the content of this course. we will only be expected to apply this technique to fractions that have denominators that factorize into linear factors that do not repeat. and (2) finding the sum of a particular type of series. 2 14 x + 48 Attempting to decompose the integrand into partial fractions. For example. Multiplying both sides by (x 2 8)(x 2 6) we obtain 1 = A (x 2 6) + B (x 2 8) . thus. Obvious choices are to substitute x = 6 and x = 8. 2 x 26 . is a very useful tool. and see if they add up to the original fraction.

Note in Example 3 that the terms cancel in pairs. 2 n = 1 (3n 2 1) (3n + 2) It is essential to understand that for any series ∑ an there are two important sequences for us to consider: the sequence {sn } of its partial sums and the sequence {an } of its terms.Example 3 – Partial fractions and series a) Express 3 in partial fractions. in general: 1 1   1  1 1  1 1 1  sn =  2  +  2  +  2  + … +  2 8  8 11  2 5  5  3n 2 1 3n + 2   1 1 sn = 2 2 3n + 2 Hence. A series that behaves in this manner is called a telescoping series because of these cancellations. 27 . the series is convergent and ∑ = . Therefore. (3x 2 1) (3x + 2) 3x 2 1 3x + 2 b) Using the result from a) we can write ∞ ∞ 3 1   1 = ∑ − ∑ (3n 21) (3n + 2) . 3x 2 1) (3x + 2) ( b) Find the sum of the series 3 3 3 3 + + +…+ +…= 2 × 5 5 × 8 8 × 11 (3n 2 1) (3n + 2) Solution ∑ (3n 21) (3n + 2) . In fact.  3n + 2  n =1 n = 1 3n 2 1 This series telescopes as demonstrated in the following partial sums. 1 1  1 1 lim s = lim  2  = 20 = . n→∞ n n→∞  2 2 3n + 2  2 ∞ 3 1 Therefore. 1 1 3 s1 = 2 = 2 5 10  1 1   1 1 1 1 3 s2 =  2  +  2  = 2 =  2 5   5 8 2 8 8 1 1   1 1  1 1 1 1 9 s3 =  2  +  2  +  2  = 2 = 8   8 11 2 11 22 2 5  5 And. n =1 ∞ 3 a) (3x 2 1) (3x + 2) 3 = A B + ⇒ 3 = A (3x + 2) + B (3x 2 1) (3x 2 1) (3x + 2) 1 Let x = : 3 = A (3) + B (0) ⇒ A = 1 3 2 Let x = 2 : 3 = A (0) + B (23) ⇒ B = 21 3 3 1 1 = − . it is a rare occurrence – and only happened in Example 3 because the denominators 3n 2 1 and 3n + 2 are equivalent when the value of n in 3n 2 1 is one greater than in 3n + 2. It is also important to point out that this ‘telescoping’ behaviour does not always occur for a series whose rule can be decomposed into partial fractions. The sum collapses (in the way that an old-style telescope collapses) into just two terms.

Therefore. in the previous section. From the limits of rational functions 1 = 0.. n→∞ n→∞ 3 (ii) The sequence of partial sums begins as follows: 1 s1 = 2 3 1 1 2 s2 = 2 + = 2 3 9 9 1 1 1 7 =2 s3 = 2 + 2 3 9 27 27 1 1 1 1 20 =2 s4 = 2 + 2 + 3 9 27 81 81 Because the series is geometric such that 21 < r < 1. below (i) write the first four terms and find the limit (if it exists) of the sequence of its terms. i.. lim an . and (ii) write the first four terms of the sequence of its partial sums {sn } and find its limit (if it exists). the sum of the series is 2 . we know that lim 2 x→∞ x + x 1 for this sequence we have lim an = lim = 0. the sum of the series. This is a  3    1 1 geometric sequence with r = 2 and because 21 < 2 < 1 then it 3 3 n (21) follows that lim an = lim n = 0.e.2 Series and Convergence Example 4 For each of the series. ∑ an . ∞ ∞ ∞ 1 (21)n c) ∑ 22n51 2 n b) ∑ a) ∑ n n =1 n = 1 n (n + 1) n =1 3 Solution n→∞ n =1 ∞ a) (i) ∑ n =1 ∞ (21)n 3 n (21)n    The sequence of terms in the series is {an } =  n . 4 ∞ 1 1 1 1 1 b) (i) ∑ = + + + + . then the series converges to 1 1 2 2 a1 3 3 = 21 . lim sn = = = n→∞ 4  1 1 2r 4 1 2 2   3 3 1 Therefore. 2 6 12 20 n = 1 n (n + 1) This series is not geometric. n→∞ n → ∞ n (n + 1) 1 1 1 1 =2 + 2 + 2… 3 9 27 81 28 .

1   lim sn = lim 1 2  = 1 2 0 = 1. ∑ 22n51 n =1 = 4 geometric with a1 = 4 and r = .   n =1 ∞ 1 1  this is a telescoping series and the first four partial sums are: 1 1 1 s1 = 2 = 1 2 2  1 1   1 1 1 2 s2 =  2  +  2  = 1 2 = 3 3  1 2   2 3 1 1   1 1 3 1   1 1 s3 =  2  +  2  +  2  = 1 2 = 4 4 3   3 4 1 2   2 1 1   1 1 4 1  1 1   1 1 s4 =  2  +  2  +  2  +  2  = 1 2 = 5 5 3  3 4   4 5 1 2   2 And. We can confirm this 5 by simplifying the rule for the nth term: 2 5 2n 1 2 n (2 ) = ∞ 2 n 5n 2 1 414n 2 1  4 = n 21 = 4    5 5 n n 21 . Because 21 < r < 1. then 5 n2 1   4  n 2 1  4 lim an = lim  4    = 4 lim   = 4 ⋅ 0 = 0. Hence. n→∞  n→∞ n + 1 Therefore. n =1 ∞ 16 64 256 ∑ 22n51 2 n = 4 + 5 + 25 + 125 + … n =1 c) (i) 4 The series appears to be geometric with r = . in general: 1 1   1 1   1 1   1 1  1 1  sn =  2  +  2  +  2  +  2  + … +  2 1 2   2 3   3 4   4 5   n n + 1  1 sn = 1 2 n +1 Hence.2 5 5 29  4 ∑ 4  5   n =1 ∞ n 1 . n→∞ n→∞ n→∞  5   5  (ii) The sequence of partial sums begins as follows: s1 = 4 s2 = 4 + 16 36 = = 7.(ii) To find the sum of the series ∑ of partial fractions to write ∑ ∞ 1 we can use the technique n = 1 n (n + 1) ∞ 1 = n = 1 n (n + 1) ∑  n 2 n + 1 . and its clear that the series is Hence. the series is convergent and ∞ 1 ∑ n (n + 1) = 1.

Is it possible for the sequence of terms of a series to converge to zero but the series itself does not converge. We established that all three series in Example 4 are convergent and also that lim an = 0 for all three series. does not have a sum? Example 5 n→∞ Consider the series ∑ n = 1 + 2 + 3 + 4 + … .2 Series and Convergence s3 = 4 + 16 64 244 + = = 9. i.e.e. In such a case. causing the sequence of partial sums to increase without bound. n→∞ n 1  4 1 2r 12   5 5 Therefore. It is obvious that any series whose sequence of terms does not converge to zero. Solution 1 = 0. Let’s consider the sum of the first 2n terms. To answer n→∞ n the question about convergence of the series we need to look at the sequence of partial sums. then the series converges to a 4 4 lim s = 1 = = = 20.808 s4 = 4 + + 5 25 125 125 Because the series is geometric such that 21 < r < 1. the sequence of terms converges to zero. lim an ≠ 0. i. Determine whether the n =1 ∞ 1 1 1 1 series converges or diverges.e. It seems reasonable to conjecture n→∞ that a necessary and sufficient condition for an infinite series a1 + a2 + a3 + … + an + … to converge to a finite quantity is that the sequence. the sum of the series is 20. 1  1 1   1 1 1 1  1 1 1 1  1 +  +  +  + + +  +  + … +  + … +  n 21 +…+ n 2 2  3 4   5 6 7 8   9 10 16  2  +1 1  1 1   1 1 1 1  1 1 1 1 1 1 1 1 1  1 1+ +  +  +  + + +  +  + + + + + + +  +…+  n −1 +…+ n 2 2  4 4   8 8 8 8   16 16 16 16 16 16 16 16  2  +1 1 1 1 1 = 1+ + + +…+ 2 2 2 2 n+2  1 n + 2 = 1 + n  = ⇒ s 2n   2 2 2 s 2n = 1 + 30 . Our analysis begins by bracketing the terms in the following way: Clearly. lim sn = 1 + 1 1 + + 2 3 1 1  1 1  1 1  1 +…+ n +…  +  + … +  +  + … +  + … +  n 21 2 4  5 8  9 16  +1 2  so that the final term in each bracketed group is the reciprocal of a power of two. the magnitude (positive or negative) of terms will increase.76 5 25 25 16 64 256 1476 + = = 11. of individual terms an converges to zero. {an }. i. will have a sequence of partial sums that diverges.

even though the sequence The clever method used in Example 5 is attributed to a French scholar. It is true that n =1 ∞ convergence can only occur if lim an = 0 (i. Therefore. the series ∑ n n n =1 diverges. the series sn = ∑ is greater than a series that diverges. then the series ∑ an diverges. Oresme was well ahead of his time by inventing a type of coordinate geometry and using the idea of a fractional exponent – three centuries before Descartes developed coordinate geometry and Newton first invented our modern notation for fractional exponents. It is interesting to note that although the harmonic series diverges. if n = 25 then 1 1 1 1 1 1 25 + 1 1+ + + . ∞ 1 1 → 0 as n → ∞ . 2 3 33554 432 2 3 2 2 So Oresme’s strategy shows that by taking enough terms of the harmonic series. not a sufficient nth term divergence test n→ ∞ n→ ∞ condition). Nicole Oresme (1323–1382). This leads to the following theorem. 2 3 2 2 For example. The following shows the heart of his strategy: 2 1 1 1 2 > + = 2 2 2 2 1  1 1  1 1 3 1+ +  +  > 1+  +  =  4 4 2 2  3 4 1 1 1  1 1 1 1 3  1 1 1 1+ + + +  + + +  > +  + + + 2 3 4  5 6 7 8 2  8 8 8 1+ 1+ 1 4 = 8 2 1 1 1 1 1 4  1 1 5 + … + + + + … +  > + + … +  = 2 8  9 10 16  2  16 16  2 This process can be continued indefinitely. who was the first to mathematically prove that the harmonic series diverges. + = 1 + + + . This series is called the harmonic series – and we will encounter it often.e. a necessary condition). we can guarantee that its sum will be greater than any finite number. but n→∞ n→∞ lim an = 0 is NOT sufficient to guarantee convergence (i.Clearly the sequence of these partial sums diverges. so it must n n =1 also diverge. the series will diverge to infinity. Considering the state of mathematics in the 14th century. . With regard to his proof of the divergence of the harmonic series. ∞ 1 Hence. If lim an does not exist. n= 1 ∞ 31 . . . for any positive integer n we have 1 1 1 n+1 1+ + + … + n > . or if lim an ≠ 0.e. so s 2n diverges. + 25 > = 13. . it does so very slowly. Therefore. in general. The sum of the harmonic series does not get above 10 until we have added 12 367 terms of the series! The fact that the harmonic series diverges (Example 5) serves as a counterexample to our conjecture that lim an = 0 is both a necessary n→∞ and sufficient condition for the series ∑ an to converge. so that. Oresme’s ingenious strategy involved replacing groups of fractions in the harmonic series with smaller 1 fractions that have a sum of _.

the series is divergent. 3n + 1 → ∞ as n → ∞. b) lim an = lim 3 (21) n→∞ n→∞ ( n +1 ) does not exist because the terms alternate between +3 and 23. the series is divergent. Therefore.2 Series and Convergence Example 6 – Using the nth term divergence test Determine. by the nth term divergence test. 32 . by the nth term divergence test. Hence. n n n2 0 e) lim an = lim 2 = lim 2 2 = =0 2 n→∞ n→∞ n + 1 n→∞ n n + 1 n 1+ 0 We cannot apply the nth term divergence test since the limit of the nth term is 0. lim an = lim n = 2 lim n = 0. by the nth term divergence test. so it follows that n → 0 as 3 +1 2 1 n → ∞. even though the sequence of its terms converges to 0 the series itself diverges. n! n! n! 1 1 d) lim an = lim = lim = = n→∞ n → ∞ 3n ! + 1 n → ∞ 3n ! n ! + 1 n ! 3+0 3 Therefore. In the next section we will learn that it does in fact converge 3 and recognizing that it is similar to a convergent geometric series is important. the series is divergent. We can make an educated guess that it will probably converge because it is n ∞ ∞ 2  1 very similar to the convergent geometric series ∑ n = 2∑   with   n =1 3 n =1 3 1 r = . 1 c) Certainly. that is. whether each of the following series converges or diverges. if possible. the nth term divergence test does not apply and we are not able to make a conclusion about convergence or divergence. We will find in the next section that this series behaves like the harmonic series. Since the limit n→∞ n→∞ 3 + 1 n→∞ 3 + 1 of the nth term is 0. ∞ 1 4 9 16 n2 = + + + +… a) ∑ 2 2 5 10 17 n =1 n + 1 b) c) d) e) ∑ 3 (21) n =1 ∞ ∞ n +1 = 3 23 + 3 2 3 + … ∑3 n =1 ∞ n =1 ∞ n 2 1 1 1 1 = + + + +… + 1 2 5 14 41 n! 1 2 6 24 ∑ 3n ! + 1 = 4 + 7 + 19 + 73 + … ∑n n =1 2 1 2 3 4 n = + + + +… + 1 2 5 10 17 Solution a) lim an = lim n→∞ n→∞ n2 n2 n2 1 = lim 2 2 = lim =1 2 n→∞ n n + 1 n2 n→∞ 1 + 1 n2 n +1 Therefore.

in general. we develop some more sophisticated tests for convergence. you may feel that there is a relationship between the convergence of an improper integral and the convergence of a series. it is limited to finding out about convergence of a series without finding the sum. Integral test From our discussion about improper integrals in the previous section. Although in some cases the convergence test being employed will help us to find the sum of a series (or at least an approximation for the sum). sn . In this section.2 of this publication. In general.Before moving onto the next section and investigating more thoroughly the convergence of infinite series. We have been able to find exact sums for certain geometric series and telescoping series because we were able to obtain a formula for the sequence of partial sums. it is not easy to find the exact sum of a series. We will study four useful convergence tests that apply to series whose terms are non-negative and a fifth test that will apply to alternating series. We can take the formula for the nth term an of a series ∑ an and replace n by x to write a n =1 ∞ ∞ function f (x ). The relationship between ∑ an and the improper integral ∫ ∞ 1 f (x ) dx is explained in the following theorem. In Example 6 we were thwarted from determining whether the series in parts c) and e) were convergent or divergent. ∑ (a n n= 1 n =1 ∞ ∞ n + bn ) and ∑ (an 2 bn ). we state below some important properties of convergent series that are direct consequences of the properties of limits in Section 1. and c is a constant. 33 n =1 . These tests will allow us to efficiently determine convergence for a wide range of series. then the n= 1 n= 1 ∞ ∞ following series are also convergent: ∑ ca . our purpose is to develop some tests that will let us determine whether a series is convergent or divergent without the need for a formula for the sequence of partial sums. Properties of convergent series Given that ∑ an and ∑ bn are both convergent series.2 Convergence of infinite series In this section. n =1 ∞ ∞ (i) (ii) (iii) ∑ ca n =1 ∞ ∞ n = c ∑ an n =1 ∑ (a n =1 ∞ n + bn ) = ∑ an + ∑ bn n =1 ∞ n =1 ∞ ∞ ∞ ∑ (a n =1 n 2 bn ) = ∑ an 2 ∑ bn n =1 n =1 2.

It is sensible to conjecture that a bounded monotonic sequence will be convergent. For the harmonic series 1 1 1 1 1    = 1. … any number greater than or equal to 1 is an 2 3 4 n n  upper bound. {sn }. . A sequence {an} is bounded if and only if it has a lower bound and an upper bound. also diverges. then the series ∫ ∞ 1 ∑a n= 1 ∞ n converges if and only if the improper integral f (x ) dx converges. and hence is not bounded. Completeness postulate In the real numbers. we established that the sequence of its partial sums. . is divergent by essentially showing that {sn } does not have an upper bound. For the sequence   we can call 1 the least upper bound and n  1  0 the greatest lower bound. In our discussion of oneto-one functions in Chapter 2 of the book. Lower and upper bounds of a sequence The number M is a lower bound of the sequence {an} if an  M for all positive integers n. . every non-empty set that has an upper bound has a least upper bound. 34 . Also for the harmonic series. … .2 Series and Convergence The integral test for convergence Let f be a function that is continuous. we state an important property of the real numbers with the following postulate. . and any number that is less than or equal to zero is a lower 1  bound. we used the word monotonic to describe a function that is either always increasing or always decreasing. and the number N is an upper bound of {an} if an  N for all positive integers n. In other words: 1) If 2) If ∫ ∞ 1 ∞ 1 f (x ) dx converges. decreasing and positive for all x  1 and an = f (n) . Another characteristic of the sequence   n  is that the terms are always decreasing and it is not surprising that the sequence converges to its greatest lower bound. ∑ n . then f (x ) dx diverges. Bounded sequence theorem A monotonic sequence converges if and only if it is bounded. then ∫ ∑a n= 1 n= 1 ∞ ∑a n ∞ n also converges. and to state an important theorem. the sequence of its terms n =1 ∞ 1 Before we conduct a formal proof of this theorem. Before we can conduct a formal proof of the integral test we need to establish the definition of two words for which we have had a commonsense understanding up to now.

since L is an upper bound for {an } then it follows that an  L for all n. or upper sum. there exists an integer N such that L 2 ε < aN.) y y f f a2 0 a3 a4 a5 n a6 x 0 a1 a2 a3 a4 a5 a6 x ∑a i =2 n n  ∫ f (x ) dx 1 ∫ n 1 f (x ) dx ∑a i =1 n 1 n Figure 2.) ∑a i (Circumscribed area. Proof of the integral test ∞ The essential idea behind the proof is that the terms in a series ∑ an can be assigned to represent the area of ever decreasing rectangles of constant width and that the improper integral ∫ f (x ) dx is approximated by the 1 n ∞ n =1 sum of these rectangles. We now need to show that {an } must converge to L. Also.2 Circumscribed rectangles gives upper sum. or lower sum. n 1 35 .1 Inscribed rectangles gives lower sum.1) and the circumscribed rectangles (Figure 2. The exact area under the graph of f from x = 1 to n. the definite integral ∫ f (x ) dx . The total areas of the inscribed rectangles (Figure 2. This satisfies the ε 2 N definition for the limit of a sequence and completes the proof for an increasing sequence {an } .2) are as follows: ∑a i =2 n 21 i =1 i = a2 + a3 + a4 + … + an = a1 + a2 + a3 + … + an 2 1 (Inscribed area. if {an } is bounded. Hence. Therefore. is increasing. then the completeness postulate guarantees that {an } has a least upper bound L. Firstly. i. L ε < an L and consequently L 2 ε < an < L + ε which is equivalent to 2ε < an 2 L < ε and an 2 L < ε. A parallel argument can be written to prove the theorem for a decreasing sequence {an } . since L is the least upper bound then L 2 ε is not an upper bound for any ε > 0. call it {an }. Conversely. Figure 2.e. If it converges to some limit L then it is bounded below by the first term of the sequence a1 and above by L and is therefore bounded. lies between the inscribed and circumscribed areas.Proof of the bounded sequence theorem We prove the theorem for the case when the monotonic sequence. Because {an } is always increasing then aN  an whenever n > N .

2 Series and Convergence As Figures 2. Recalling techniques for improper integrals from the first section of this 36 . . Determine the convergence or divergence of each series. thereby satisfying the conditions for applying the integral test. ∞ 1 a) ∑ n = 2 n ln n ∞ n b) ∑ n n =1 e 1 1 1 1 1 + + + + +… c) 2 5 10 17 26 ∞ n d) ∑ 2 [Example 6 part e)] n =1 n + 1 Solution Plot1 Plot2 Plot3 Y1= 1/(X (1n(X)) Y2= Y3= Y4= Y5= Y6= 1 a) Graphing the function f (x ) = on our GDC provides us with a x ln x quick confirmation that f is continuous. and consequently the series sn  L + a1 . and given the inequality sn follow that {sn } diverges which means that Example 7 – Using the integral test 1 ∑a n =1 ∞ ∫ f (x ) dx it must n 1 n also diverges. we can write the inequality above as sn ∫ f (x ) dx n 1 sn 1 . ∫ f (x ) dx goes 1 to infinity as n → ∞. For part (2) assume n 1 that the improper integral ∫ f (x ) dx diverges. decreasing and positive for all x  2. i Using the nth partial sum. the sequence of partial sums {sn } is bounded and monotonic and it follows from the bounded sequence theorem that {sn } converges. sn = a1 + a2 + a3 + … + an.2 illustrate. Then it n 1 Hence. To prove part (1) we start by assuming follows that for n  1 sn a1 L and consequently ∫ f (x ) dx converges to L.1 and 2. Thus. ∑a i =2 n i ∫ a1 n 1 f (x ) dx n 1 i =1 ∑a . ∑a n =1 ∞ n n must also converge.

the integral the series ∑ n=2 ∞ ∑a n= 1 ∞ n converges if and only if the improper integral 1 n ln n x b) For f (x ) = x . if the summation index for an infinite series starts at n = k > 1 rather than n = 1. For this integral we will also need to apply the technique of u-substitution. x +1 b x dx = lim 2xe 2x 2 e 2x 1 x  b →∞  e Rewriting improper integral as a limit. ∫ ∞ 1 = lim (2be 2b 2 e 2b ) 2 (2e 21 2 e 21 )  b →∞  b + 1  2  b + 1 2 = lim 2 b +  = 2 lim  b  + lim   b →∞ b →∞ b→∞ e e  e     e 1 b + 1  Applying l’Hôpital’s rule to the first limit gives lim  b  = lim  b  = 0. 37 . Applying result from u-substitution. ∫ x dx = . Using integration by parts: x 2x Choose u = x ⇒ du = dx and dv = e2x ⇒ v = 2e2x . ∑ en n =1 c) We need to find a rule for the nth term for the series that starts 1 1 1 1 1 + + + + +…. 2 ∫ ∞ 1 ∫ ∞ k f (x ) dx converges.chapter. ∫ x ln x x x  2 ∫ ∞ 1 ∫ ∞ 2 1 x ln x dx = lim ∫ b→∞ b b 2 1 x ln x dx Rewriting improper integral as a limit. decreasing and positive for all x  k such that k > 1 and an = f (n) . we now need to evaluate dx to see if it converges to x ln x a finite number or diverges to infinity. 1 1 1 2 1 2 1  1 dx = ∫ (ln x ) 2  dx  = ∫ u 2 du = 2u 2 Let u = ln x . and by the integral test x ln x must also diverge. ∫ e x dx = ∫ xe dx = 2xe 2x 2 ∫ 2e 2x dx Substituting into formula ∫ u dv = uv 2 ∫ v du. the integral test can still be applied. 1 e e ∞ x By the integral test. As Example 7 a) illustrates. decreasing and positive e for x  1 because e x > 0 and e x grows faster than x. n =1 1 The function f (x ) = 2 satisfies the conditions of the integral test. since the integral ∫ x dx converges then the series 1 e ∞ n must also converge. it is clear that f is continuous. then the series = lim 2 ln x  2 b→∞  = lim 2 ln b 2 2 ln 2   b→∞  =∞ Therefore. so the integral test applies. 2 5 10 17 26 Using some inductive reasoning we determine that the series expressed in summation notation is ∞ 1 1 1 1 1 1 1 ∑ n 2 + 1 = 2 + 5 + 10 + 17 + 26 + … + n 2 + 1 + … . then du = dx. The integral test can be modified as follows: Let f be a function that is continuous. = 2xe 2x + ∫ e 2x dx = 2xe 2x 2 e 2x dx diverges. b →∞ b →∞ e    e  ∞ x 2 Therefore.

since the integral series ∞ 1 dx converges then the x +1 2 ∑a ≠ ∫ n n= 1 ∞ ∞ 1 f (x ) d x . ∫ ∞ 1 ∑n n =1 2 n must also diverge. ∫ 1 x 2 + 1 dx = blim ∫ 1 x 2 + 1 dx →∞ b = lim [arctan x ]1 = lim [arctan b 2 arctan 1] b→∞ π  = lim [arctan b] 2 lim   b→∞ b→∞ 4   π π = − 2 4 ∞ 1 π Therefore. The sum. expressed to ten significant figures. is n =1 approximately 1. The function divergence test was inconclusive because lim 2 n→∞ n + 1 x f (x ) = 2 satisfies the conditions of the integral test. ∞ b 1 1 Rewriting improper integral as a limit. is defined 38 . +1 x dx diverges then the x +1 2 Estimating the sum of a series As pointed out.2 Series and Convergence We need to recognize that the anti-derivative of (a ‘standard integral’ in the IB formula booklet). Therefore.7853981634. whereas ∞ 1 π ∫1 x 2 + 1d x = 4 ≈ 0. since the integral ∞ b→∞ 1 is arctan x x +1 2 ∫ ∞ 1 It is very important to know when using the integral test that the value of the improper integral is not equal to the sum of the series. we know that the associated series converges to a sum. The method x +1 x of u-substitution will be useful to evaluate the integral ∫ 2 dx . but we are not any closer to knowing the value of the sum. Recall that the sum of a series. 2 Substituting gives 1 1 1 1 1 1 x 2 ∫ x 2 + 1 dx = ∫ u ⋅ 2 du = 2 ∫ u du = 2 lnu = 2 ln (x + 1). Example 7 d). x +1 1 Let u = x 2 + 1 and it follows that du = 2xdx ⇒ du = xdx .056 875 301. series ∑ 2 n =1 n + 1 ∞ n was the series in Example 6 e) for which the nth term d) ∑ 2 n =1 n + 1 n = 0. Using this result we have: ∞ b 1 1 ∫ 1 x 2 + 1 dx = blim ∫ 1 x 2 + 1 dx →∞ b 1 = lim ln (x 2 + 1)1  b→∞  2 1 = lim ln (b 2 + 1) 2 ln 2  2 b→∞  =∞ By the integral test. 1 x +1 4 By the integral test. s. in general 1 must also converge. if the integral used in the integral test converges. ∫ 2 dx = . of the first 50 terms of the series ∞ 1 ∑ n2 + 1. Rewriting improper integral as a limit.

or error. R n = an + 1 + an + 2 + an + 3 +   ∫ f (x ) dx .3 that the area under f for x  n is an f upper bound for R n . we can see from Figure 2.e.4 Similarly. y y x f f y an an + 1 an + 2 an + 3 f x 0 n x 0 n y Figure 2.3 an +1 an + 2 an + 3 Assuming that the function f is continuous. For any specific partial sum. an an + 1 an + 2 an + 3 n→∞ n We use some of the same ideas and notation that we used0in our proof of the integral test. lim sn = s. an + 1 + an + 2 + ). decreasing and positive for all x  n . What is the difference. So when the partial sum sn is used as an approximation for s. n will be some integer. the error R n (sometimes called the remainder) is given by R n = s 2 sn = an + 1 + an + 2 + an + 3 + … .4 that the area under f for x  n + 1 is a f lower bound for R n . between the f partial sum sn and the exact sum s? Clearly this will be the sum of the terms ‘beyond’ n (i. n 0 ∞ n +1 x y y f 0 n x f an +1 an + 2 an + 3 0 n +1 x 0 n+1 x y Figure 2. i. That is.e. R n = an + 1 + an + 2 + an + 3 +   ∫ ∞ n +1 f (x ) dx 39 . we can see from Figure 2.y to be the limit of the partial sums.

549767 731: 1. ∞ 1 Therefore. s. Estimating sum of series for the integral test Given that f (n) = an and f is continuous. we can use the integral test to give us important results for any series that is in the form shown below. s 25 is computed to be approximately 1. ∑ 2 ≈ 1. which is approximately 0. 1 40 . n ∞ Solution a) Using a GDC. and b) using the inequalities ∑n ∞ 1 2 = 1+ n +1 f (x ) dx  s  sn + ∫ f (x ) dx with n = 10. of the series sn + ∫ ∞ 1 1 1 1 + + +  + 2 +  by: 4 9 16 n n =1 a) adding the first 25 terms. adding sn to each side of the statement above gives sn + ∫ f (x ) dx  s  sn + ∫ f (x ) dx . b) We can compute the lower and upper bounds of our estimate from the inequalities ∞ 1 ∞ 1 s10 + ∫ 2 dx  s  s10 + ∫ 2 dx . n ∞ n +1 ∞ Since s = sn + R n.004 55. n =1 n p-series Before we move onto the next convergence test. n ∞ These lower and upper bounds can provide a very accurate approximation for the sum of the series.640676822  s  1. Example 8 – Two ways of estimating the sum of a series Estimate the sum. then the error for our approximation can be no greater than half of the length of the interval.004 55. 11 x 10 x b ∞ 1 ∞ 1  1 1 1  1 We need to find ∫ 2 dx : ∫ 2 dx = lim 2  = lim 2 +  = n x n x b→∞ b→∞  b n n  x n 1 1 Substituting gives: s10 +  s  s10 + 11 10 Using s10 ≈ 1.649767 731 If we estimate s using the midpoint of this interval. positive and decreasing for x  n  1 and ∑a n= 1 ∞ n is convergent then sn + ∫ ∞ n +1 f (x ) d x  s  sn + ∫ ∞ n f (x ) d x .645 222 277 with error < 0.2 Series and Convergence Hence. ∫ ∞ n +1 f (x ) dx  Rn  ∫ f (x ) dx .605 723 404 to ten significant figures. ∞ 1 1 1 1 1 ∑ n p = 1p + 2 p + 3p +  + n p +  where p is a constant. known as a p-series.

For example. 1 When p < 0. then 2p + 1 < 0 and consequently as b → ∞. then lim p = 0 = 1.If p = 1. b2p + 1 → ∞. 1 When p > 0. 1 x ∞ 1 b 2p ∫ 1 x p dx = blim ∫ 1 x dx →∞  x 2p + 1  = lim  b → ∞ 2p + 1   1 b b  1  = lim  2p + 1 1  b → ∞ x  12 p  1  = lim b2p + 1 2 1  12 p  b→∞   If p > 1. n→0 n n 1 In both of these cases. What about series for other values of p? The following example will lead to a simple test for the convergence of any p-series. and p > 0 . Hence. Example 9 – Convergence of p-series ∞ 1 For what values of p is the series ∑ p convergent? 1 n Solution Let’s consider when p < 0. decreasing and x positive for x  1 so we can use the integral test. Convergence of p-series 1 1 1 1 + p + p ++ p + p n 1 2 3 1 (i) converges if p > 1. The p-series ∞ ∑n 1 p = Note: When p = 1 this is the harmonic series. and clearly n 3 increases without bound as n → ∞. then 2p + 1 > 0 and consequently as b → ∞. then p → ∞ as n → ∞. b2p + 1 → 0. 1 x n 1 The results from Example 9 are summarized below. if p = 23 then n 1 = n 3. Therefore the 1 x 12 p ∞ 1 integral converges and the series ∑ p must also converge. the p-series is the harmonic series which we know diverges. ∞ 1  1  1 Hence. n 1 If p < 1. We know from Example 5 in the previous section that the harmonic series ( p = 1) diverges. the function f (x ) = p is continuous. so let’s ∞ 1 assume that p ≠ 1 and investigate the improper integral ∫ p dx . p = 0. 41 . ∞ ∞ 1 1 if p < 1 then the integral ∫ p dx diverges and so does the series ∑ p . n23 1 1 When p = 0. if p > 1 then ∫ p dx =   (21) = p 2 1 . lim p ≠ 0 so the p-series diverges by the nth term n→0 n divergence test. and (ii) diverges if p < 1.

2

Series and Convergence

Comparison test
The integral test compares a series consisting of all positive terms with an integral as a means of testing the convergence of the series. It is possible to use a second series in a similar way. If each term of a series of positive terms is less than or equal to the corresponding term of a known convergent series of positive terms, then the series is convergent. We will call this the comparison test and can state it as follows.
In the statement of the comparison test, n > N means from some term onward. That is, eventually for some term and forever afterwards the terms of the series Comparison test Given 0 < an < bn for all n > N for some integer N, it follows that 1 2 if if

∑b
n= 1 ∞ n= 1

n

converges, then

∑a
n= 1 n

n

also converges;

∑b
n= 1

n

are always

∑a

n

diverges, then

∑b
n= 1

also diverges.

greater than the corresponding terms of the series that

∑ a . This is
n n= 1

Note: The comparison test can also be applied for the series

∑a
n= 1

n

and

∑b
n= 1

n

whenever

often expressed by saying

there exists a positive constant c such that 0  an  cbn for all n  N , N ∈  + .

∑b
n= 1

n

dominates

∑a .
n n= 1

The comparison test significantly expands our ability to determine the convergence of a series with more complicated rules for the nth term. We achieve this by comparing a ‘complicated’ series to a ‘simpler’ series whose convergence or divergence is known.

Before proving both parts of the comparison test, we will find it helpful to state a corollary to the bounded sequence theorem that we recall says the following: A monotonic (always decreasing or always increasing) sequence converges if and only if it is bounded. If all the terms of an infinite series are positive, the sequence of partial sums is increasing. Therefore, the following theorem follows directly from the bounded sequence theorem.
Positive series convergence A series of positive terms is convergent if and only if its sequence of partial sums has an upper bound.

Proof of comparison test

Proof of 1: Let {un } and {v n } be sequences of the partial sums for the series

∑ an and
n =1

∑ bn , respectively. Because
n =1

∑b
n =1

n

is a series of positive

terms that is convergent, it follows from the positive series convergence theorem that the sequence {v n } has an upper bound – let’s call it B. Since an  bn for all n  1, we can conclude that an  bn  B for all n  1. Thus, B is an upper bound of the sequence {un }. Because the terms of the series ∑ an are all positive then it follows from the positive series convergence theorem that ∑ an is convergent.
n =1 ∞ n =1 ∞ ∞

Proof of 2: If ∑ an is divergent, then since {un } is increasing un → ∞. However, bn  an, so v n  un . It follows that v n → ∞ and, therefore, must also diverge.
42
n =1

∑b
n =1

n

Example 10 – Using the comparison test

Determine the convergence or divergence of each series. ∞ 2 a) ∑ n [Example 6 c)] n =1 3 + 1 ∞ 1 b) ∑ n =1 3 + n ∞ 1 c) ∑ n=0 n !
Solution

a) We can compare the given series 2 2 2 2 2 + + + + ++ n 4 10 28 82 3 +1 with the nth term of the geometric series 2 2 2 2 2 + + + ++ n + 3 9 27 81 3 which converges because its common ratio is between one and negative 1 one; r = < 1. 3 It is clear that each term in the given series is less than its corresponding 2 2 term in the geometric series. That is, n < for all n  1. 3 + 1 3n ∞ 2 Therefore, by the comparison test since the series ∑ n converges the 3 n =1 ∞ 2 series ∑ n must also converge. n =1 3 + 1

bn

an

n

Figure 2.5 Note that part 1 and part 2 of the comparison test require that 0  an  bn. You can think of ∑ an as the ‘lower’ series and ∑ bn as the ‘higher’ series (see Figure 2.5). Thus, in a very informal sense the two parts of the comparison test say: 1. If the ‘higher’ series converges, then the ‘lower’ series must also converge. 2. If the ‘lower’ series diverges, then the ‘higher’ series must also diverge. The ‘higher’ series dominates the ‘lower’ series.
43

2

Series and Convergence

b) The series

∑ 3+
n =1

1 n

is similar to the p-series

∑n
n =1

1
12

which diverges

1  1. If we compare the given series to this p-series we 2 1 1 see that < for all n  1. However, the comparison test 3+ n n provides no conclusive result in this case where a series is dominated by a divergent series. Suspecting that the given series does in fact diverge we need to find a divergent series that the given series dominates. Let’s ∞ 1 compare it to the divergent harmonic series ∑ . Remember, to n =1 n satisfy the comparison test it is not necessary for an  bn to be true for all integers n  1 but for all integers n  N where N is some positive integer. because p = Our GDC is a handy tool to quickly compare the terms of the given series to the harmonic series. The screen images below show values for the first 14 terms of the two series in a table.
Plot1 Plot2 Plot3

Y1= 1/X Y2=1/(3+ (X)) Y3= Y4= Y5= Y6= Y7=

X=1

1 2 3 4 5 6 7

X

1 .5 .33333 .25 .2 .16667 .14286

Y1

.25 .22654 .21132 .2 .19098 .1835 .17712

Y2

X=14

8 9 10 11 12 13 14

X

.125 .11111 .1 .09091 .08333 .07692 .07143

Y1

.17157 .16667 .16228 .15831 .1547 .15139 .14833

Y2

How could we prove that the 1 1 is true inequality < n 3+ n for n  6? Try doing so by proving the inequality 3 + n < n for n  6 by mathematical induction. In Example 10 c), we know that the sum of the infinite n ∞  1 geometric series ∑ 2   is  2 n= 0 2 a1 S∞ = = = 4 . Thus the 1 12 r 12 2 ∞ 1 sum ∑ n ! must be less than n= 0 4. In fact, we will learn in the next section that this sum is exactly the number e. That is, 1 1 1 e = 1+ 1+ + + + 2 6 24  1 + + n!

For the first five terms the terms in the harmonic series greater than those for

. However, it appears from the sixth n term onwards that this reverses, that is, 1 1 < for n  6. n 3+ n ∞ 1 Therefore, by the comparison test the series ∑ diverges. n =1 3 + n c) Consider the first few terms of the given series: ∞ 1 1 1 1 ∑ n! = 1 + 1 + 1⋅ 2 + 1⋅ 2 ⋅ 3 + 1⋅ 2 ⋅ 3 ⋅ 4 +  n=0
n =1

∑ 3+

1

∑n
n =1

1

are

Now consider the first few terms of the convergent geometric series 1 with a1 = 2 and r = . 2 n ∞ 1 1 1  1 ∑ 2  2 = 2 + 1 + 2 + 4 + 8 +    n=0 ∞ 1 are less than or equal to the It appears that the terms of ∑ n=0 n ! corresponding terms of the convergent geometric series for all n  1. xn Recall that in Example 8 of this chapter we proved that lim = 0 for n→∞ n !

44

In a case like this n =1 3 another form of the comparison test. where N > n. Let k and m be positive numbers such that k < L < m . This test can be particularly useful in comparing a series to a p-series or a geometric series. this does not imply that the series ∑ an also diverges. can be used. ∞ ∞ a 2. where L is finite and positive. the series being tested must be dominated by (‘lower’ than) a convergent series. For example.  2 n! ∞ 1 n  1. or it must dominate (‘higher’ than) a divergent series. the series n =1 ∞ 45 . consider the series ∑ n n =1 3 21 ∞ 2 to the series ∑ n that we proved is convergent in Example 10 a). it follows that: 1. If lim an = L . by the comparison test the series ∑ converges. From that we concluded that the factorial function n 1  1  2   for increases faster than any exponential function. If these conditions are not met then the comparison test (sometimes called the direct comparison test) cannot be ∞ 2 that is nearly identical used. n =1 3 + 1 ∞ 2 to also converge. bn It follows that kbn < an < mbn . If the series ∑ bn converges then from the properties of series. n→ ∞ b n= 1 n= 1 n ∞ ∞ an = ∞ and ∑ bn diverges then ∑ an also diverges. Therefore. the inequality We strongly expect ∑ n n =1 3 21 2 2 > shows that the series dominates the convergent geometric 3n 2 1 ∞ 3n 2 series ∑ n so the comparison test does not apply. known as the limit comparison test. Since a lim n = L then there is a positive integer N. n=0 n ! Limit comparison test In order for the comparison test to provide us with a conclusive result on the convergence or divergence of a series. However. If lim n = 0 and ∑ bn converges then ∑ an also converges. Hence. then the two series bn both converge or both diverge. If lim n→ ∞ b n= 1 n= 1 n n→ ∞ ∑a n= 1 ∞ n and ∑b n= 1 ∞ n If applying the limit comparison test you get ∞ a lim n = 0 and ∑ bn n→ ∞ b n= 1 n diverges. n= 1 ∞ Proof 1.any real number x. such that n→∞ b n an k< < m. 3. Limit comparison test Given an > 0 and bn > 0 for all n  N for some integer N.

Thus. Likewise. this series resembles the convergent geometric ∞ 2 series ∑ n . Example 11 – Using the limit comparison test Determine the convergence or divergence of each series. we evaluate the following limit. n =1 3 2 n 3n lim 3 2 1 = lim n n → ∞ 3 21 n→∞ 2 n 3 3n 3n = lim n n n → ∞ 3 3 2 1 3n = lim =1 Since the limit is finite and positive and limit comparison test the series ∞ 3 1 n → ∞ 1 2 1 3n ∞ ∑3 n =1 2 n converges then by the ∑3 n =1 ∞ n 2 must also converge. n n1 3 ⋅ n 2 3 lim n + 1 = lim n→∞ n→∞ n + 1 1 23 n 46 3 . n =1 n =1 ∞ n =1 n =1 ∞ ∞ n =1 n =1 ∞ n =1 n =1 ∞ ∞ ∞ ∞ ∑ kb n =1 ∞ n The proofs of parts 2 and 3 are left as exercises. Since p =  1 we know that ∑ 2 3 diverges. and since ∑ an dominates then by the comparison test ∑ an must diverge. 21 ∞ 3 n n b) The given series ∑ is similar to ∑ which is a p-series best n =1 n + 1 n =1 n ∞ ∞ 2 1 1 written as ∑ 2 3 . We 3 n =1 n n =1 n then evaluate the following limit.2 Series and Convergence ∑ mbn must also converge. if the series ∑ bn diverges then the series ∑ kbn must also diverge. ∞ ∞ 2 n2 + 1 a) ∑ n b) ∑ n n =1 3 21 n =1 c) ∑ 3n n =1 ∞ n 2 + 7n 6 2n3 d) ∑ sin  n    n =1 ∞  1 Solution a) As mentioned above. Since ∑ mbn dominates ∑ an then by the comparison test ∑ an must converge.

This can be expressed as a1r n + 1 = r.= lim =1 n→∞ n n +1 Since the limit is finite and positive and ∞ 3 ∑n n =1 ∞ 1 23 diverges then by the limit comparison test the series ∑ n +1 n =1 n must also diverge. the ratio of adjacent terms is constant. Hence. since ∑ diverges then ∑ sin   also diverges. n 2 + 7n 6 3 n 4 (n 2 + 7n) lim 3n 2 n = lim n→∞ n→∞ 1 3n 6 2 n 3 4 n n 6 + 7n 5 = lim 6 n → ∞ 3n 2 n 3 n 6 n 6 + 7n 5 n 6 = lim 6 6 n → ∞ 3n n 2 n 3 n 6 1+ 0 = 320 1 = ∞ 3 1 Since the limit is finite and positive and ∑ 4 converges then by the n =1 n 2 ∞ n + 7n limit comparison test the series ∑ 6 must also converge. lim = 1. So we can use the limit comparison ∞  1 theorem and compare the given series ∑ sin   to the divergent  n ∞ n =1 1 harmonic series ∑ . n→∞ 1 n ∞ ∞ 1  1 Therefore.2 of the book we proved lim n→∞ x means of the squeeze theorem. a1r n 47 . for the given series ∞ ∞ ∞ n 2 + 7n n2 1 ∑ 3n 6 2 n 3 we can compare the series to ∑ n 6 = ∑ n 4 which is a n =1 n =1 n =1 convergent p-series. n =1 n  1 sin    n Hence. it is possible to find a suitable p-series for comparison purposes by disregarding all but the highest powers of n in the numerator and denominator. c) As we saw in part b). 3 n = 1 3n 2 n sin x = 1 by d) Remember that in Section 13.  n n =1 n n =1 Ratio test In a geometric series.

Applying the ratio test to the general p-series 1 n→∞ ∑n n =1 ∞ 1 p gives 48 1 np We know that a p-series converges if p > 1 and diverges if p  1.2 Series and Convergence We know that a geometric series converges if and only if this ratio is between 21 and 1. n =1 n =1 ∞ ∞ n =1 ∞ r ⇒ aN + 2 r aN + 1 r 2 aN 2. and with lim an + 1 = L. In other types of series. an + 1 Hence. n → ∞  n + 1 p . this shows that if lim = 1 then it is possible to have a n→∞ a n lim (n + 1) p  n  = lim   = 1. an Then 1 the series converges if L < 1 2 the series diverges if L > 1 3 the test is inconclusive if L = 1. it must be true that an + 1 > an for all n sufficiently large. n→ ∞ Proof 1. 3. as indicated in the following theorem. an aN + 1 Then r ⇒ aN + 1 r aN aN aN + 2 aN + 1 and so on. For the case when L < 1. Therefore. Thus. This shows that for n  N the series ∑ an is dominated by the geometric series aN ∑ r n 2 1. there must be a number r with 0 < r < 1 such that an + 1 an r for all n sufficiently large. For the case when L > 1. Suppose that there exists some integer N such an + 1 that  r if n  N . lim an ≠ 0 and the series n→∞ ∑a n =1 ∞ n must diverge by the nth term divergence test. Ratio test Let ∞ ∑a n= 1 n be a series with non-zero terms. the ratio of adjacent terms does not remain constant but it can still give us helpful information about whether or not the series converges. aN + aN + 1 + aN + 2 +   aN 1 + r + r 2 +  . Because 0 < r < 1 this geometric series converges and by the comparison test ∑ an must also converge.

by the ratio test the series ∑ n! n =1 ∞ nn diverges.series that is either convergent or divergent. Example 12 – Using the ratio test Determine the convergence or divergence of each series. ∞ ∞ n 3 3n + 1 nn b) ∑ a) ∑ 4n n=0 n =1 n ! Solution a) All the terms of the given series are positive so we can do without the absolute value signs. the ratio test is inconclusive if L = 1. The ratio test is particularly useful for testing series involving exponential expressions or expressions with factorials. the series has only positive terms so we can write the ratio test without absolute signs. Therefore. 49 . an + 1 an (n + 1)3 3n + 2 = lim n→∞ n→∞ lim 4n + 1 n 3 3n + 1 4n  (n + 1)3 3n + 2 4n  = lim  ⋅ n + 1 ⋅ n + 1 3 n→∞ 3 4   n 3 3 (n + 1) = lim n→∞ 4n 3 3 = <1 ∞ 4 n 3 3n + 1 Therefore. by the ratio test the series ∑ converges. 4n n=0 b) Again. an + 1 an = lim n→∞ lim (n + 1)n + 1 (n + 1) ! nn n! n→∞  (n + 1) (n + 1)n n !  = lim  ⋅ n n→∞ n   (n + 1) n ! n (n + 1) = lim n→∞ nn n  n + 1 = lim   n→∞  n  1  = lim 1 +  n→∞  n =e >1 n Therefore. as illustrated in the following example.

∑ ∞ As we will learn even further in the next section. the series converges for any values of x such that x < 22 or x > 2. It is often the case that we can determine whether or not a series converges by more than one test. Example 13 For what values of x will the series Solution ∑ nx n =1 ∞ 2n n converge? Applying the ratio test gives the following inequality to solve. an + 1 an 2n + 1 (n + 1) x n + 1 = lim n→∞ 2n nx n n→∞ n→∞ lim <1 lim lim 2n 2 nx n ⋅ n <1 (n + 1) x n x 2 2n <1 x (n + 1) n→∞ 1 2n ⋅ lim <1 n→∞ n + 1 x 2 <1 x x >2 Therefore. 2n ∑ nx n will alternate between n =1 positive and negative. The summary at the end of this section gives some tips on how to find the most efficient test to apply for a certain series. The ratio test is very useful in analyzing series with alternating terms that we will take a closer look at in the next section. Note that when x < 22 the terms of ∞ 50 .2 Series and Convergence When applying the ratio test to series involving quotients of expressions with factorials. as in the following example. the ratio test is useful in answering questions about convergence. it is often necessary to perform simplification steps similar to those we did in Example 12: 1 n! n! = = (n + 1)! (n + 1) n ! n + 1 Although the ratio test worked in Example 12 part b) we could have used the nth term divergence test to prove that the series diverges by considering the following: n ⋅ n ⋅ n ⋅ ⋅ n 22 33 4 4 nn n = 1+ + + + and for the nth term an = 1⋅ 2 ⋅ 3 ⋅  ⋅ n 2 6 24  n! n =1 Thus as n → ∞ the terms do not approach 0 and the series diverges by the nth term divergence test.

Figure 2. As n increases the points corresponding to the nth partial sum ‘jump’ back and forth on either side of their limit s. so it converges to = 2 . ∑ (21) n =1 n→ ∞ ∞ n+1 an = a1 2 a2 + a3 2 a4 +  (an > 0) converges if both of the following conditions are satisfied.2. having terms that are alternately positive and negative. We have encountered series with some negative terms. The first series above is a geometric series 1 2 1 1 3 with r = 2 . Since (21) (21)n an  = (21)n + 1 an   then the test also holds true for the second form by applying the property of convergent series: if ∑a n= 1 ∞ n converges. gradually closing in as the value of the terms go to 51 . for some positive integer N. The following test can be used to determine the convergence of a wider range of alternating series that satisfy certain conditions. lim an = 0 2. The alternating series test is stated using the first form.6 shows a graph n→∞ Figure 2. and the second series  1 4 3 1 2 2   3 diverges by the nth term divergence test. the four convergence tests we have established – with the exception of the ratio test – apply only to series with positive terms. But not all alternating series will be geometric nor satisfy the conditions of the nth term divergence test. or ∑ (21) n =1 n an where the first term is negative such that an is a positive number. then ∑ ca n= 1 ∞ n also converges. and c is some real constant.6 Convergence of the partial sums of an alternating series to their limit s. an + 1  an for all n  N . Alternating series test The alternating series These examples show that an alternating series can be written in one of two forms: ∑ (21) n =1 ∞ ∞ n+1 an where the first term is positive. + a1 + a3 a4 + a5 a6 0 s2 s4 s6 s a2 ∑ (21) n =1 ∞ n +1 an such that an + 1  an and lim an = 0. s5 s3 s1 Observe how the alternating signs of the terms of the series cause the partial sums to be alternately larger and smaller. for example these alternating series: 1 1 1 1 (21)n 2 + 2 + ++ n + [Example 4 a)] 3 9 27 81 3 n +1 3 2 3 + 3 2 3 + 3 2 3 +  + 3 (21) +  [Example 6 b)] Series such as these.3 Alternating series and absolute convergence Although we have encountered series whose terms alternate between positive and negative. 1. are called alternating series. Proof Consider the sequence of partial sums {sn } for an alternating series of a few terms of {sn } .

the sequence of evennumbered terms has s1 as an upper bound and must also have a limit – call it L2 . it follows that s 2n  L2  L1  s 2n + 1 and s 2n + 1 2 s 2n = a 2n + 1 Since n→∞ lim a2n + 1 = 0 then it must follow that L1 = L2 Therefore. When applying the alternating series test. {s }. If lim an = 0 but condition (2) fails then the n→∞ alternating series test is inconclusive. With L1 the limit of the sequence of odd-numbered terms in the sequence of partial sums and L2 the limit of the even-numbered terms. the odd-numbered terms in the sequence of partial sums. has a limit – call it L1. Similarly. Example 14 – Using the alternating series test Determine the convergence or divergence of each series. If we can show that L1 = L2 then the series converges to a unique limit s = L1 = L2 . 2n + 1 2n {s }. by the bounded sequence theorem. a) ∑ (21) n =1 ∞ n +1 n 2 n +1 ∞ b) ∑ n =1 ∞ (21)n 2n 3n 2 1 c) ∑ (21) n =1 ∞ n +1 ln n n Solution a) The series ∑ (21) n n > 0 for n  1. There is another condition implied that must be met – that the series is truly (eventually) alternating. for some positive integer N. that is. Also observe that as these ‘jumps’ get smaller and smaller (because an + 1  an).2 Series and Convergence zero. it is best to verify condition (1) first. n n n2 0 lim an = lim 2 = lim 2 2 = =0 n→∞ n→∞ n + 1 n→∞ n n + 1 n2 1+ 0 n +1 2 52 . the sequence of partial sums of the alternating series ∑ (21) n =1 ∞ n +1 an converges. If condition (1) fails. Furthermore. lim an ≠ 0. form a decreasing sequence and the even-numbered terms. If this is not obvious by inspection then the easiest way to verify that a series written in the form ∑ (21) n =1 ∞ n +1 an or ∑ (21) an is alternating is to show n n =1 ∞ that an > 0 for all n  N . then the series diverges by n→∞ the nth term divergence test. the decreasing sequence of odd-numbered terms has s 2 as a lower bound and thus. form an increasing sequence. is alternating because 2 n +1 n +1 n =1 Condition (1) is easily verified.

∞ d 1 (ln n) ln n lim a = lim = lim dx = lim n = 0 n→∞ n n→∞ n n→∞ n→∞ 1 d (n) dx Hence. let’s attempt to satisfy condition (2) by proving the inequality n . Therefore. n Checking condition (1) we can evaluate the following limit using ∞ l’Hôpital’s rule because it has the indeterminate form . Therefore. n→∞ n n → ∞ 3n 2 1 3 Applying the nth term divergence test. It is not obvious whether this is true so we consider the ln x derivative of the related function f (x ) = .Now. b) The series 3n 2 1 2n 2 lim a = lim = ≠ 0 so condition (1) is not satisfied. an + 1 < an for an = 2 n +1 (n + 1) + 1 (n + 1) (n 2 + 1)  n (n + 1)2 + 1   2 2 n +1  n n +1 3 n + n + n + 1  n + 2n + 2n 3 2 2 Cross-multiplying. condition (1) is satisfied. 1  n2 + n n (n + 1)  1 Since n  1. Hence. For condition (2) we must show that the sequence given by an = ln n n is decreasing. so the series is alternating. the series ∑ (21) 2 n +1 n =1 converges by the alternating series test. we need to find the limit of the nth term as n → ∞ . an + 1  an ∞ n n +1 and condition (2) is satisfied. then the inequality n (n + 1)  1 is true. both denominators are positive. the series diverges by the nth lim c) an = ln n > 0 for all integers n  2. n =1 n→∞ ∑ ∞ (21)n 2n is alternating since 2n > 0 for all n  1. x  1 x   2 ln x  x 1 2 ln x f ′ (x ) = = < 0 for all x > e 2 x x2 53 . n→∞ (21)n 2n does not exist. but 3n 2 1 (21)n 2n = lim (21)n ⋅ lim lim 3n 2 1 n→∞ n→∞ 2n 3n 2 1 2n 2 n = but lim (21) does not exist (Example 1 a)). so n → ∞ 3n 2 1 n→∞ 3 lim 3n 2 1 term divergence test.

the amount of error) is less than or equal to the first unused term. Rn = s sn an + 1. the absolute value of the remainder Rn (i. n =1 ∑ ∞ (21)n converges. note that s is always between any two consecutive partial sums. etc. sn  s  sn + 1. both conditions of the alternating series test have been ∞ n + 1 ln n satisfied and the series ∑ (21) is convergent. Since lim n→∞ 54 . Proof As previously mentioned. Recalling that s is the limit of the partial sums. Example 15 Show that n4  less than 0. and find the sum of the series with error Solution 1 1 1 4 = 0 and  4 ⇒ n 4  (n + 1) is true for all 4 4 n (n + 1) n n  1. of a convergent alternating series is always between any two consecutive partial sums. In other words. n =1 s sn an + 1 and therefore the proof is complete. the sum. notice that s 2 s3 < a4. the error generated in estimating the sum with the nth partial sum does not exceed the value of the n + 1 term. Note that the alternating series estimation theorem does not give a formula for the precise value of the error. s. if n is even and sn + 1  s  sn . Given that ∞ n+1 an + 1 = sn + 1 2 sn Remember ∑ (21) an is an alternating series. so an + 1 > 0. That is. so it follows that an + 1  an for n  3. it follows that s sn sn + 1 sn . s 2 s5 < a6 . When estimating s with the sum of the first n terms. n n =1 Take another look at Figure 2. Furthermore. Therefore. Alternating series estimation theorem Suppose that ∑ (21) n =1 ∞ n+1 an is a convergent alternating series that satisfies both conditions of the alternating series test and has an unknown sum of s. f is decreasing for x > e which means that f (n + 1) < f (n). Whether n is even or odd. That is. the series satisfies both conditions of the alternating series test and therefore converges. Also note that this rule for the bound of the error when estimating s with sn only applies to alternating series that satisfy the condition of the alternating series test. if n is odd. but rather a bound for the error.0001. This provides us with the means to estimate the error when we use the partial sum sn to approximate the sum of an alternating series.6 that was used in the proof of the alternating series test.2 Series and Convergence Hence.e. s 2 s 4 < a5 .

You may recall that in Example 7 d) we used the the series ∑ 2 n =1 n + 1 integral test to prove that this series diverges. the positive and negative terms offset one another to produce a series that converges even though the series consisting of only positive terms diverges.0001 does not affect the third decimal place. We know that the harmonic series (a p-series with p = 1) diverges. we will learn that the alternating harmonic series converges to exactly ln 2. However. 2 an + 1  an ⇒ n +1 n Therefore. ∞ (21)n is s ≈ 20.0001. the sum of the series ∑ 4 n =1 n decimal places. 10 Our GDC computes the ninth partial sum to be 1 1 1 1 1 1 1 1 s9 = 21 + 4 2 4 + 4 2 4 + 4 2 4 + 4 2 4 ≈ 20. in Example 14 c) of this section we showed that the corresponding alternating series ∞ ∞ n (21)n n +1 ∑ (21) n 2 + 1 converges. ∑ n =1 ∞ (21)n + 1 converges by the alternating series test.947 092 5924. the alternating series ∑ n 4 n =1 n =1 55 . The same situation is true of ∞ n . 1 1 1 1 1 (21)n + 1 + ln 2 = 1 2 + 2 + 2 +  +  2 3 4 5 6 n But more relevant to this section is that the result of Example 16 illustrates an important point to investigate further. series ∑ n n =1 Solution Applying the alternating series test we have 1 1 lim an = lim = 0 n→∞ n→∞ n and 1 1  ⇒ n  n + 1 which is true for all n. Example 16 -1+1/2^4-1/3^4+1 /4^4-1/5^4+1/6^4 -1/7^4+1/8^4-1/9 ^4 -. n Absolute and conditional convergence In the next section. if we take the harmonic series and change the sign of alternate terms to get the alternating harmonic series (Example 16). In contrast. 2 3 4 5 6 7 8 9 This estimate of the sum of the series is accurate to three decimal places because an error of less than 0.947.947095924 Determine the convergence or divergence of the alternating harmonic ∞ (21)n + 1. However.We know from the alternating series estimation theorem that the sum of the first nine terms will give an estimate for the sum with an error of at most 1 a9 + 1 = 4 = 0. correct to three Therefore.

conditionally convergent. and therefore ∑a n= 1 ∞ n is absolutely value an is either an or 2an . by the comparison test ∑ (an + an ) converges. You may wonder if it is possible for a series with positive terms. ∑ an . When trying to determine if an alternating series is absolutely convergent.2 Series and Convergence Absolute and conditional convergence Suppose ∑a n= 1 ∞ n is a series with positive and negative terms that is convergent. so ∞ ∑2a n =1 ∞ n also converges. it is possible to take a convergent series with positive and negative terms and change them all to positive to create a new series that is divergent. then by the absolute convergence theorem. to diverge. it is most effective to first check if the limit of the nth term is zero. It is inefficient to start by first applying the alternating series test. ∞ (21)n (alternating harmonic series) and We have seen then that ∑ n n =1 ∞ n n +1 ∑ (21) n 2 + 1 are both conditionally convergent because for each n =1 the series composed of their terms all made positive diverges. Example 17 Classify each series as absolutely convergent. but for a related series with some (or all) of the terms changed to negative. and we state the following theorem. ∑ ∞ an converges. The difference between these two n =1 n situations requires us to define two types of convergence when considering the convergence of a series with positive and negative terms as occurs with any alternating series.D. Therefore. then convergent. Q.E. then test the alternating series using the alternating series test. If it diverges. However. it follows from n =1 properties for convergent series that ∑ an = n =1 ∞ ∑ (a n =1 ∞ n =1 ∞ n + an ) 2 ∑ an where n =1 ∞ both series on the right converge. n= 1 ∑a n= 1 ∞ n also converges. If it converges. as the alternating harmonic series demonstrates. Therefore. conditionally convergent. The answer is n =1 n ∞ It is true that 0  an + an  2 an because by the definition of absolute Absolute convergence theorem If The absolute convergence theorem essentially says that it is not possible to take a convergent series with only positive terms and change some of them to negative to create a new series that is divergent. and you are finished. or 56 . A given condition for the theorem is that ∑ n =1 ∞ an converges. to converge. or divergent. the series is absolutely convergent and you are finished. Proof n =1 ∞ ∑ a . If ∑ ∞ ∞ an converges. then n n= 1 ∑a n= 1 is said to be absolutely ∞ convergent. n= 1 ∑a n= 1 n (Example 15) converges and so does the corresponding series with positive ∞ 1 terms ∑ 4 (a p-series with p = 4 > 1). Since an = (an + an ) 2 an . no. If ∑ ∞ an diverges. ∑ an must converge. then is said to be conditionally convergent. If the nth term divergence test is inconclusive then check whether the related series of positive terms converges (using any of the four tests for positive series). If it is not then the series diverges. Whereas ∞ (21)n is absolutely convergent because its corresponding series of ∑ n4 n =1 positive terms also converges.

∑ n! n =1 n =1 n ! ∞ b) We can apply the nth term divergence test to show that ∑ (21) 2n n2 n =1 diverges. so result of nth term divergence test is inconclusive. n n=1 n=1 1 1  ⇒ n !  (n + 1) ! . We apply l’Hôpital’s rule twice to prove that lim lim 2n ≠ 0. We now turn our attention to the given  n n =1 2 n→∞ lim an ≠ 0. Recall that the absolute value theorem stated that if lim an = 0 n +1 n→∞ then lim an = 0. We now apply the alternating series test n =1 n ! n→∞ (21)n lim to ∑ n =1 ∞ (21)n . n→∞ n2 57 . Thus lim an = 0. ∞ (21)n a) ∑ n =1 n ! b) c) n +1 ∑ (21) n =1 ∞ ∞ 2n n2  1 sin    n ∑ (21) n =1 n +1 Solution a) = 0 . From this we can also say that if lim an ≠ 0 then n→∞ n→∞ 2n (ln 2) 2n 2n ln 2 = lim = lim = ∞ (does not exist) n→∞ n2 n→∞ n→∞ 2n 2 n ∞ n +1 2 diverges. and we have satisfied both conditions of the alternating series ∞ (21)n converges and converges absolutely because test. Thus an + 1  an . Rewriting (n + 1) ! as n ! (n + 1) gives (n + 1) ! n ! n !  n ! (n + 1) which is clearly true for all n  1. Therefore. Knowing n! n→∞ n→∞ lim 1 1 1 = 0. ∑ n =1 n ! n ∞ ∞ 21) 1 ( = ∑ also converges. We now need to show that n→∞ n! Note: For Example 17 a). n! We next consider the corresponding series with only positive terms ∞ 1 ∑ n ! . we used the comparison test n =1 ∞ 1 to show ∑ converges.divergent. and since 0 < < for all n n! n n  1 then lim 1 = 0. by the nth term divergence test ∑ (21) n2 n =1 ∞  1 c) In Example 11 d) we compared the series ∑ sin   to the divergent  n ∞ n =1 1 harmonic series ∑ and using the limit comparison test showed n =1 n ∞ 1  that ∑ sin   diverges. Recall that in Example 10 c). Hence. we could have been more efficient by applying the absolute convergence theorem since we have previously used the comparison test to show that ∞ ∞ n 1 ∑ (21!) = ∑ n ! converges.

However.7 provides confirmation that not only  1   1  1 lim sin   = 0. y = sin x 0 11 1 1 54 3 2 1 π 2 x Figure 2. and even rearranged to form a series that converges to any predetermined sum. diverges. As stated earlier without explanation (next section). giving: 1 1 1 1 1 1 2 + 2 + 2 = ln 2 2 4 6 8 10  2 Now we add (3) and (1): 1 1 1 1 1 1 1 1 1 1 12 + 2 + 2 + 2 + 2 + 2  = ln 2 2 3 4 5 6 7 8 9 10 11 1 1 1 1 1 1 + + 2 2  = ln 2 2 6 8 10 2 2 4 (3) (1) (3) (1) (3) 58 . It is plausible to expect that the sum of the series in (2) is also ln 2. but also that sin  < sin   for all n  1.   n =1 Rearrangements of conditionally convergent series The distinction between absolute and conditional convergence is important in many applications of infinite series. the sum is unchanged only if the series is absolutely convergent.  n ∞ n =1 n +1  1 ∑ (21) sin  n  converges conditionally. this means that operations (such as the associative property) that are valid for finite sums are not valid for infinite sums. Let’s continue by dividing (1) by 2. the sum of the alternating harmonic series is ln 2. ∑ sin   .7 The graph shown in Figure 2. As mentioned previously.2 y Series and Convergence series and first need to confirm whether it is an alternating series. We can illustrate this paradoxical behaviour with the alternating harmonic series that is conditionally convergent. 1 1 1 1 1 1 1 1 1 12 + 2 + 2 + 2 + 2 +  = ln 2 2 3 4 5 6 7 8 9 10 Consider the following series: 1 1 1 1 1 1 1 1 1+ 2 + + 2 + + 2 +  3 2 5 7 4 9 11 6 (1) (2) (2) consists of a rearrangement of the same terms as in (1). It seems perfectly logical that it is possible to rearrange a finite number of terms in an infinite series without affecting the sum. This is a direct consequence of the fact that the sum of an infinite series is defined to be the limit of the sequence of its partial sums. Since ∞ n +1  1  1 sin   > 0 for all n  1 then ∑ (21) sin   is an alternating series  n  n n =1 and we can apply the test for alternating series. n→∞  n + 1  n   n Thus the series satisfies the alternating series test and converges. if we rearrange an infinite number of terms in an infinite series. Since ∞  1 the corresponding series of positive terms. An extraordinary characteristic of series that are conditionally convergent is that their terms can be rearranged to form a divergent series.

a n n= 1 ∞ Integral test n = f (n) ∫ ∞ ∞ 1 f (x ) dx ∫ ∞ ∞ 1 f (x ) dx sn + ∞ converges diverges f is continuous. affects the sum of the series.1 Tests for infinite series. Test nth term divergence test Converges Diverges n→ ∞ Notes Can only be used to show divergence S∞ = a1 12 r ∑a n= 1 ∞ ∞ lim an ≠ 0 n Geometric series ∑a r n= 0 1 r <1 r 1 n p-series ∞ 1 ∑ np n= 1 p >1 p 1 Harmonic series when p = 1 ∑a . positive and decreasing Comparison test sn + ∫ f (x ) d x are bounds for n estimation of sum by sn Useful for series similar to p-series or geometric series ∫ ∞ n+1 f (x ) d x and ∑a n =1 ∞ n and ∑ bn n =1 ∞ ∑b n= 1 n converges n ∑a n= 1 n diverges n 0 < an < bn Limit comparison test ⇒ an bn ∑a n =1 ∞ converges ⇒ ∑b n =1 ∞ diverges Useful if not able to show 0 < an < bn for direct comparison ∑a n =1 ∞ n and ∑ bn n =1 ∞ n→ ∞ lim = L . by rearranging the terms we have manipulated how the partial sums are formed which affects the limit of the partial sums and. bn > 0 Ratio test ∑a n= 1 ∞ n n→ ∞ lim an + 1 a1 <1 n→ ∞ lim an + 1 a1 >1 Inconclusive if lim an + 1 a1 n→ ∞ =1 Alternating series test ∑ (21) n =1 ∞ n+1 n→ ∞ lim an = 0 and 0 < an + 1 < an an sn as estimate of sum remainder: Rn < an + 1 59 . So which is correct. Although both (1) and (4) are series containing the same terms. Table 2. (1) or (4)? The answer is that they are both correct. if 0 < L < ∞ ⇒ both behave the same L = 0 ⇒ if bn converges then an converges L = ∞ ⇒ if bn diverges then an diverges an > 0.The result is 1 1 1 1 1 1 1 1 3 1 + 2 + + 2 + + 2 +  = ln 2 3 2 5 7 4 9 11 6 2 (4) where the terms are arranged the same as in (2). consequently. but the sum is not what we expected.

try integral test. try the alternating series test. Is the series an alternating series? If so. compare the series to ∑ . If the series is convergent. apply one of the comparison tests. +3 n =1 ∞ 1 a Using the comparison test. compare the series to ∑ . or similar to a geometric series? If similar. Is the series geometric. Testing ∑ an makes more tests available. b Use the integral test to determine whether the series or divergent. 3.1 1 Using properties of convergent series and geometric series. Consider an = f(n).2 Series and Convergence Guidelines for testing series for convergence Important questions to consider: 1. 6. 5. n =1 n ∑ 2n ∞ 2 60 . a ∑ n(n + 2) n =1 1 b n =2 ∑n ∞ 2 1 21 c ∑ (n + x )(n + x + 1) . 4. find the sum of each of the series. write the first four terms of the infinite series and determine whether the series is convergent or divergent. n = 1 2n ∞ 1 b Using the limit comparison test. Is f a continuous. or similar to a p-series? If similar. the series diverges by the nth term divergence test. Is the series a p-series. Remember that if ∑ an is convergent then ∑ an is absolutely convergent. decreasing function and is it possible to integrate f ? If so. 3 5 7 9 ∑ ∞ ∞ n n +1 2 4 6 8 10 2x n =1 ∑4 n =1 ∞ n= 1 ∞ ∞ 3 n 21   ∑ ln  n   n =1 ∞ 1 ∑ (21) n =1 ∞ n+1 3 2n ∑3 n =1 ∞ n =1 n! n ∑ cos (n π) ∑e n =1 2n ∑ 5n + 6 2n + 3 11 a Find ∫ xe dx by using the method of integration by parts. positive. try the ratio test. or has an expression with factorials? If so. Is lim an = 0? If not. n =2 ∑n ∞ 1 ln n b ∑n n =1 ∞ 2 1 +4 13 Show that n diverges by both of the following methods. a ∑ ne n =1 ∞ 2n is convergent 12 Use the integral test to determine whether the series is convergent or divergent. apply one of the comparison tests. n→ ∞ 2. find its sum. a ∑2 n= 0 ∞ ∞ 7n 3n b  ∑ 2  n= 0 ∞ 1 n 2 2  3n  c ∑  ∞  5n + 3 (23n )  9n  n =1 ∞ 2 Use partial fraction decomposition to find the sum of each of the following telescoping series. Exercise 2. Does an involve n in a product or power. n =1 1 x > 21 In questions 3–10.

1 b Use the integral test to determine whether ∑ 2 converges or n = 1 n + 2n + 2 diverges. a ∑ (21) n= 1 ∞ n+1 1 (2n 2 1)2 b ∑ (21) n =1 ∞ n+1 xn n ∞ 34 a Express n2 + 2n + 2 in the form (n + a)2 + b where a and b are integers. determine the convergence or divergence of the series. how many terms of the series are needed to make the error of the estimate for ∑n n= 1 1 3 less than 0.0005? 61 . and b the ratio test. use the sum of the first four terms to approximate the sum of the series. 16 Use the ratio test to show that n =0 ∑ 10 18 20 22 24 26 ∞ n10 n converges. That is.14 Show that ∑ n3 n =1 ∞ 1 n converges by a the comparison test. 15 Give an example to show that the converse of the nth term divergence test is lim false. 17 19 21 23 25 27 29 n =2 ∑n ∞ ∞ 1 n 21 2 ∑ ∞ ∞ n =1 n+1 n2 2n +1 ∑ ∞ lnn 2 n =1 n n! n= 0 ∞ ∑3 n= 1 ∞ n n= 0 ∞ ∑ (n + 2)! ∑ (21) ∞ n= 1 ∞ n ∑ n+1 ∑ (ln 2) n= 1 ∞ 2n n= 0 n+1 2n + 1 n3 n ∑ nn n! n+2 n2 + n  ∑  n= 1 ∞ n + 1  n  n n ∑ (21) n= 1 ∞ 28 30 ∑ ∞ (21)n n n= 1 n ∑2 n= 1 n 1 21 n =0 ∑e 1 n ∞ 31 Use the integral test to determine whether 32 Find the sum of the following infinite series. n 1 n =1 n 36 a Use the method of Example 8b) with n = 5 to estimate the sum of the series ∑n n= 1 ∞ 3 . 5 5 5 + + + 1× 2 2 × 3 3 × 5  n= 2 ∑ nln n converges or diverges. 1 33 For each series. ∞ arctann converges or diverges by comparing the series 35 Determine whether ∑ to ∑ ∞ n= 1 1 and applying the limit comparison test. In questions 17–30. ∞ b With this method. State an upper bound for the error of the approximation. find a series that diverges even though n → ∞ an = 0.

62 .2 Series and Convergence 37 Use the alternating series estimation theorem to determine the minimum number of terms of the series 12 the sum has an error less than 0. 1 1 1 + 4 2 4 +  so that an approximation of 4 2 3 4 38 Give an example of a series that is conditionally convergent. determine whether each series converges absolutely. 46 What is the minimum number of terms of the series 1 1 1 ∑ ∞ (21)n needed to n! n =1 approximate the sum of the series correct to three decimal places? 47 Prove parts 2 and 3 of the limit comparison test. That is. 39 41 43 ∑ ∞ ∞ (21)n + 1 n 2 n+1 40 n (n + 1)2 2 n =1 ∑ ∑ n= 2 ∞ ∞ ∞ (21)n + 1 2n 2 1 n =1 ∑ (21) n =1 ∞ 42 44 (21)n ln n n (23) 4n 21 n ∑ n =1 cos (nπ) n2 ∑ n =1 45 Describe how the terms of the alternating harmonic series 12 + 2 +  can 2 3 4 be rearranged so that its sum is 1. a series that is convergent but not absolutely convergent. In questions 39–44. converges conditionally.000 05. or diverges.

multiples. Polynomial functions. A power series is essentially a polynomial function of infinite degree expressed in terms of a single variable (we will always use x). 3. that is representing e x . such as e x ? The manufacturers of the calculator had to decide on a computational algorithm.e. In this section we will see that the power series for the function f (x ) = e x is ∞ x2 x3 x4 x5 xn = 1+ x + + + +…+ + … . A transcendental function is non-algebraic.1 Introduction Have you ever considered how your calculator computes values for certain x 4 + 2x functions? For functions such as f (x ) = 3x 2 2 5x + 8. differences. A calculator can only display a ∑ n! 2 ! 3! 4 ! n=0 n ! finite number of digits and thus it only sums enough terms to produce the necessary degree of accuracy. What computational method could be programmed into a calculator to evaluate e x for a certain value of x ? The answer lies in the fact that the calculator is summing up a type of series with variable terms. As explained in Chapter 3 of the book.  . How does your calculator compute the values of transcendental functions. all the series we have encountered thus far contained terms consisting of constants. it cannot be expressed as a finite number of sums. algebraic functions can be expressed as a finite number of sums. and h (x ) = 7 x 2 3 24 x 3 + x 2 2 6 4 + 2x . called a power series. Except for Example 13 in the Let’s return to the primary question we wish to investigate. g (x ) = .3 Power Series 3. Shown below are successively more accurate approximations for the value k xn of e 2 by summing the terms of the power series ∑ for k = 2. But how does your calculator compute values for a function such as e x ? This is an example of a transcendental function. A power series is a very useful mathematical tool that can be used to represent a range of very important functions. e2 = 1 + 2 + 2 2 2 2 + + + +… 2 ! 3! 4 ! 5 ! 2 3 4 5 previous section. n=0 n ! 63 . 9. differences. suppose we wanted to use this series (we’ll investigate its derivation later) to evaluate e 2 to three significant figures. Other familiar transcendental functions include the trigonometric and logarithmic functions. and h (x ) = 7 x 2 3 the method of evaluation is fairly straightforward 3 + x 2 2 6 because these are algebraic functions. For example. i. multiples. rational functions and functions involving radicals are examples of algebraic functions. quotients and radicals involving x n . quotients and radicals involving x n .

39 2835 7 ! 8! 9! A calculator (see screen image above) computes to an accuracy of ten significant figures the value of e 2 to be 7.38871252205  ≈ 7. This leads to an important question: For what values of x does the power series converge? Before addressing this question. Thus. … . is called a power series centred at c.380 9523 ≈ 7. then an infinite series of the form ∑a x n= 0 n ∞ n = a0 + a1x + a2 x 2 ∑ a (x 2 c) n= 0 n ∞ n = a0 + a1 (x 2 c) + a2 (x 2 c) + a3 (x 2 c) +  + an (x 2 c) +  2 3 n + a3x 3 +  + anx n +  .387 3015 ≈ 7. a2.3 Power Series Once we get past the ninth term in the series we are no longer adding enough to change the first three digits. the computation process is made more efficient by means of a formula that determines the number of terms required for a power series to produce a value to a given accuracy. this is not an impediment for an electronic computing device like a GDC.3 3 + + + + + + 24 =7 4! 2 4 25 + 4 ! 5! 2 4 25 + 4 ! 5! 2 4 25 + 4 ! 5! 2 4 25 + 4 ! 5! 2 4 25 + 4 ! 5! 4 = 7 15 = 7. It certainly appears ∞ ∞ 2n xn that the series ∑ is converging to e 2. let’s give a proper definition for a power series.389056099 =5 + + + + + + + 23 3! 23 3! 23 3! 23 3! 23 3! 23 3! 23 3! = 6 1 = 6. where c is a constant and an is the rule that determines each of the coefficients a0.38 7! 27 28 + + = 7 122 = 7. 3. e2 ≈ 1 + 2 = 3 e2 ≈ 1 + 2 + e2 ≈ 1 + 2 + e2 ≈ 1 + 2 + e2 ≈ 1 + 2 + e2 ≈ 1 + 2 + e2 ≈ 1 + 2 + e2 ≈ 1 + 2 + e2 ≈ 1 + 2 + Performing such computations entirely by hand would be immensely tedious (and prone to error). 64 . 22 2! 22 2! 22 2! 22 2! 22 2! 22 2! 22 2! 22 2! e^(2) 7. a1.35 45 + 27 8 = 7 21 = 7. ∑ n=0 n ! n=0 n ! is an infinite series.39 315 7 ! 8! 27 28 29 + + + = 7 1102 = 7. we have Power series Definition of power series If x is a variable. As we will see.2 For any power series centred at c = 0. For any given value of x.26 26 6! 26 + 6! 26 + 6! 26 + 6! + = 7 16 = 7. the first nine terms of the series are sufficient to give an approximation of e 2 accurate to three significant figures. However.389 056 099. Note that we have (x 2 c)° = 1 even when x = c.

The expression on the left side defines a function whose domain is the interval 21 < x < 1 . and consequently diverges when x  1. so its domain is 21 < x < 1. It is evident that every power series is convergent for x = c . a power series defines a function. Far more useful power series will converge for a finite interval with the same centre as the series (see Example 1). investigating whether they converge or diverge. the given power series is a polynomial 1 (Figure 3.Radius of convergence At each value of x. equivalent to x < 1. if we let an = 1 for all n and n ∞ ‘centre’ the series at c = 0. Some power series are only convergent at x = c (see Example 3). i. 1 + x + x 2 + x 3 +  + xn +  = 12 x The expression on the right side of this equation defines a function whose domain is x ∈ . The equation can only be true where both sides are defined. f (x ) = 12 x n=0 ∞ n=0 n=0 ∑x ∞ n = 1 + x + x 2 + x 3 +  + x n +  having first terms a1 = 1. Figure 3. a power series becomes a series of constants.e. In the previous section we gave a great deal of attention to such series. The sum formula for geometric series assures us that this series converges 1 when 21 < x < 1. x ≠ 1. overleaf). The function f (x ) = n=0 ∑ a (x 2 c ) n ∞ n has as its domain all values of x for which the power series converges. the series represents the number that is the sum of the series. and 65 . to 12 x Therefore. partial sums of the series ∑ x n with 3. A power representation of the function f (x ) = 12 x series is best regarded as an attempt to describe a function locally.3 shows the same three partial sums along with 1 focused on the interval 21 < x < 1. The issue of convergence is very important for power series because for each value of x for which a power series converges. On this domain.1. near where it is ‘centred’.2. we can write 1 . 6 and 9 terms have been graphed in Figure 3. near the value of c. we get the geometric series common ratio r = x . Example 1 For the general power series ∑ an (x 2 c ) . or converge for all x (see Example 2). 2 1 < x < 1. To illustrate this point. Therefore.

Example 2 We’ve demonstrated that the power series ∞ x2 x3 x4 x5 xn ∑ n ! = 1 + x + 2 ! + 3! + 4 ! + …+ n ! + … represents the function n=0 f (x ) = e x .2 Graphs of the partial sums 1 + x + x 2.3 Power Series y 4 3 2 1 y 4 3 2 1 3 2 1 1 2 1 2 x 3 2 1 1 2 1 2 x Figure 3.1 Graph of y = 1 . 12 x 10 8 6 4 2 1 2 1 x Observe how in the interval 21 < x < 1 the graph of a partial sum of ∞ 1 ∑ x n gets closer to that of the graph of f (x ) = 12 x as the number of n=0 terms increase. 1 + x + x 2 + x 3 + x 4 + x 5 . y Figure 3. 1 + x + x 2 + x 3 + x 4 + x 5 and 1+ x + x 2 + x 3 + x 4 + x 5 + x 6 + x 7 + x 8 .3 The partial sums 1 + x + x 2. but are not close outside this interval. Find the values of x for which this power series converges. 12 x Figure 3. 66 . 1 1 + x + x 2 + x 3 + x 4 + x 5 + x 6 + x 7 + x 8 and (dashed).

and 1 + x + y 50 40 30 20 10 + + Figure 3.4 Graph of y = e x. 1 + x + 2! 3! 2 ! 3! 4 ! x2 x3 + x2 x3 x4 x2 2! + x3 3! + x4 4! + x5 5! + x6 6! x7 7! + x8 8! + x9 . 9! y 50 40 30 20 10 6 4 2 2 4 x6 4 2 2 4 x 67 .Solution Example 13 in the previous section illustrated that the ratio test is effective for answering this kind of question. Applying the ratio test gives the following inequality to solve.4) to graphs of partial sums of ∞ xn ∑ with 4. This is illustrated by comparing graphs of f (x ) = e x (Figure 3. an + 1 an xn + 1 (n + 1) ! < 1 = lim n→∞ xn n! lim xnx n! ⋅ n <1 (n + 1) n ! x x <1 n +1 n→∞ n→∞ lim n→∞ n→∞ lim x ⋅ lim 1 <1 n +1 x ⋅0 < 1 0 <1 Therefore. the power series n=0 ∑ n! ∞ xn is convergent for all real values of x. n=0 n ! Figure 3. 5 and 10 terms in Figure 3.5. and not just constrained to a finite interval as occurred in the previous example.5 The partial sums 1 + x + __ + __. This means that as we compute partial sums of more and more terms ∞ xn for ∑ they will become ever more accurate representations of the n=0 n ! function f (x ) = e x over all real numbers.

The number R of possibility 3 in the theorem is called the radius of convergence of the power series. then R = ∞. x ≠ 0. an interval of the real numbers centred at c. the series diverges for all values of x. The following theorem (which we present without proof) states that these are the only possibilities. and if possibility 2 occurs. then R = 0. 2. at x = 0. as illustrated in Figure 3. If possibility 1 occurs. The set of all values of x for which a given power series is convergent is called the interval of convergence of the power series. x≠0 Hence. We need to check the series when x = 0. The series may or may not converge at either of the endpoints x = c 2 R and x = c + R so we need to check for convergence at each of the endpoints. For possibility 3. there remains the question of what happens at the endpoints of the interval. and diverges for x 2 c > R . applying the ratio test gives: n→∞ lim an + 1 an = lim n→∞ (n + 1) ! x n + 1 n !xn n→∞ = lim x (n + 1) = x lim (n + 1) n→∞ = ∞. n=0 ∑ n !0 ∞ n = 1+ 0 + 0 + = 1 ∞ n=0 Therefore. As illustrated in the preceding three examples.6. 3. Thus. or all real numbers. Convergence of a power series theorem For a given power series n=0 ∑ a (x 2 c ) n ∞ n exactly one of the following is true: 1. Again. the power series ∑ n ! x n converges only at its centre.3 Power Series Example 3 Find the values of x for which Solution n=0 ∑ n !x ∞ n is convergent. if 0 < R < ∞ the interval of convergence can be one of four different kinds. Each endpoint must be tested separately to determine if the series converges or diverges for that value of x. the domain of a power series can be a single point. There is a positive number R such that the series converges for x 2 c < R . The series converges only when x = c. The series converges for all real values of x. 68 .

]c  R. the series is ∑ 2n (3 ) ∑ n =1 ∞ n (23)n (3)n = ∞ (21)n . which diverges by limit n =1 n = n =1 comparison with the harmonic series. In Example 2. for example differentiation and integration of the function. and. Example 4 n=0 ∑1 n=0 n=0 ∑ (21) ∞ n = 1 2 1 + 1 2 1 +  . c + R[ [c  R. The series is centred at c = 4. Both of these series diverge confirming that Find the radius of convergence and interval of convergence for the series ∞ (x 2 4)n . But we did not check the ∞ endpoints. in fact. the interval of convergence is 1  x < 7 . 7 [ .R R R R ( c ) x ( c ] x [ c ) x [ c ] x Figure 3. R = ∞. R = 0. which converges by the 2n 1 n =1 ∞ ∑ 2n (3 ) ∑ 2n . 69 . At x = 7. c + R] In Example 1 we showed that the interval of convergence for the power n=0 ∑x ∞ ∞ n is 21 < x < 1. polynomial functions are generally much easier to differentiate and integrate. 3 The radius of convergence is R = 3. As we did in Examples 2 and 3. so the series converges for Hence. Why are we interested in power series? One reason is that power series share many of the desirable properties of polynomials. c + R] [c  R. in Example 3. c + R[ series ]c  R. n = 1 + 1 + 1 +  . The power series representation of a particular function may enable us to perform some difficult operations on the function that would otherwise be quite difficult. In particular. so you will need to check endpoints with one of the other tests from the previous chapter. so R = 1. At x = 21. 21 < x < 1. the series is ∞  2n  = lim   x 24 n → ∞  6n + 6  1 = x 24 3 alternating series test. Therefore. The ratio test will fail when x is an endpoint of the interval of convergence. the series converges for 4 2 3 < x < 4 + 3 or 1 < x < 7. and at x = 1. or x ∈ [ 1. ∑ x n = n=0 ∑x ∞ n = the interval of convergence is.6 Intervals of convergence. At x = 1. ∑ 2n (3n ) n =1 Solution We apply the ratio test: lim n→∞ (x 2 4)n + 1 ⋅ 2n (3n ) 2 (n + 1) (3n + 1 ) (x 2 4)n = lim n→∞ 2n (x 2 4) 3 (2n + 2) 1 x 2 4 < 1 or x 2 4 < 3. it is best to use the ratio test to determine the radius of convergence for a power series.

i. To find a polynomial function P that represents another function f. begin by choosing a number c in the domain of f at which f and P have the same value. we propose the following. near the value of c ’. near where it is “centred”. there will be many polynomial functions that will have the same value as f at x = c . x ∈ n=0 n ! without any indication about its origin. that is. We can make the graph of P further resemble that of f near the point they share by a second requirement: P has the same slope of f at x = c. f ′′ (x ) = e x ⇒ f ′′ (0) = e 0 = 1 P2′ (x ) = a1 + 2a2 x 1 Therefore. Deriving the power series for f(x) = ex centred at 0 We start by finding a first degree polynomial P1 (x ) = a0 + a1x whose value and slope agree with the value and slope of f (x ) = e x at x = 0 .3 Power Series 3. (c . Thus. ‘a power series is best regarded as an attempt to describe a function locally. As we stated in our work for Example 1. In this section we will devise a general method for constructing a power series. We can continue to improve how well P mimics the behaviour of f near c by additionally requiring that P ′′ (c ) = f ′′ (c ). that is.3 Deriving power series – Taylor and Maclaurin series In the preceding part of this chapter we were given the power series ∞ xn ex = ∑ . P1 (x ) = 1 + x . f (0) = e 0 = 1 f ′ (x ) = e x P1 (0) = a0 + a1 (0) = 1 ⇒ a0 = 1 0 ⇒ f ′ (0) = e = 1 P1′(x ) = a1 ⇒ P1′(0) = a1 = 1 ⇒ a1 = 1 Therefore. P ′ (c ) = f ′ (c ). P2′′ (x ) = 2a2 ⇒ P2′′ (0) = 2a2 = 1 ⇒ a2 = 1 2 70 . With theses ideas in mind. P (3) (c ) = f (3) (c ). P2 (x ) = 1 + x + x 2 . Of course. This is the first requirement – that P (c ) = f (c ). 2 We continue and find the third degree polynomial P3 (x ) = a0 + a1x + a2 x 2 + a3x 3 and require that P (3) (0) = f (3) (0) (third derivatives are equal). and so on. the graphs of f and P will pass through the same point. Now to find the second degree polynomial approximation P2 (x ) = a0 + a1x + a2 x 2 we require that P ′′ (0) = f ′′ (0). and that ‘a power series is essentially a polynomial function of infinite degree’. P (0) = f (0) and P ′ (0) = f ′ (0) .e. f (c )). The approximating polynomial P is said to be expanded about c or centred at c. knowing that a0 = 1 and a1 = 1.

367 879 0. ∞ xn e x = ∑ .1 0. In this way we derive the result that was presented to us at the start of this chapter. 2 6 24 By now we can see the pattern for Pn (x ) .718 28 2. it is convenient to write the polynomial in the standard form from the definition. P2(x) and P3(x) displayed in Table 3. where Pn (x ) is the nth partial sum of the infinite series n=0 Pn (x ) = a0 + a1 (x 2 c ) + a2 (x 2 c ) + a3 (x 2 c ) +  + an (x 2 c ) 2 3 n ∑ a (x 2 c ) . P4(4) (x ) = 24a4 ⇒ P4(4) (0) = 24a4 = 1 ⇒ a3 = 1 24 We now wish to generalize the method for finding the power series representation for a given function f(x). if they exist.904 83 _ 0 1 1 1 0.1. x2 x3 x4 xn Pn (x ) = 1 + x + + + ++ n! 2 ! 3! 4 ! We complete the derivation of the power series for f (x ) = e x by determining its interval of convergence by means of the ratio test and checking interval endpoints. in general.3 _ 20.f (3) (x ) = e x ⇒ f (3) (0) = e 0 = 1 P3′ (x ) = a1 + 2a2 x + 3a3x 2 P3′′ (x ) = 2a2 + 6a3x 1 1 Therefore.905 0.1 Comparing P2 (x ) = 1 + x + x 2 . we can make two observations: (1) the accuracy of the approximating polynomial improves as x → 0. P3 (x ) = 1 + x + x 2 + x 3 . as x → c.2213 _ 1.2 0.0 2.221 40 1. For expansions about an arbitrary value of c.2 1.105 16 _ 0. Repeated differentiation of Pn produces 71 .818 731 0. That is.5 2.82 0. 2 6 2 Let’s apply the procedure one more time to find P4 (x ) = a0 + a1x + a2 x 2 + a3x 3 + a4 x 4 .105 1.105 17 1.5 0. P3 (x ) = 1 + x + x 2 + x 3 . f (4) (x ) = e x ⇒ f (4) (0) = e 0 = 1 P4′ (x ) = a1 + 2a2 x + 3a3x 2 + 4a4 x 3 P4′′ (x ) = 2a2 + 6a3x + 12a4 x 2 P4(3) (x ) = 6a3 + 24a4 x 1 1 1 Therefore. and f (x ) = e x .0 0. x ∈ .8186 _ 20. P4 (x ) = 1 + x + x 2 + x 3 + x 4 . and (2) the higher the degree of the approximating polynomial (more terms of the partial sum of the power series). and we would expect.1 1.6 _ 1 1 1 Table 3. the better the polynomial represents the function f. n=0 n ! General result From the function values for f(x) = ex (six significant figures). n n ∞ We will streamline the procedure that we used for deriving the power series for f (x ) = e x .22 1.904 837 0. 2 6 x f(x) = e P2(x) P3(x) x P3(3) (x ) = 6a3 ⇒ P3(3) (0) = 6a3 = 1 ⇒ a3 = 1 6 21.

f (x ) = ∑ ∞ Although the English mathematician Brook Taylor (1685–1731) did not invent the process of using polynomial approximations for transcendental functions (others such as Leibniz. Gregory published power (Maclaurin) series for tan x. an = ∞ n! n Substituting this formula for an into ∑ an (x 2 c ) we determine that if a function f has a power series representation centred at x = c . it follows that: f (n ) (c ) f ′′ (c ) a0 = f (c ) a1 = f ′ (c ) a2 =  an = n! 2! Consequently. Johann Bernoulli and Abraham de Moivre had already used series that we would call Taylor series). we assert that if f has a power series representation centred at x = c. 72 . arctan x and arcsec x ten years before Maclaurin was born. if f (x ) = n=0 ∑ a (x 2 c ) . Definition of Taylor series and Maclaurin series If a function f has derivatives of all orders at x = c. As often occurs. the Scottish mathematician Colin Maclaurin (1698–1746) did not invent the series named for him. Isaac Newton (1642–1727) was an early innovator with power series. A contemporary of Newton’s was another brilliant Scottish mathematician. sec x. Taylor was the first to provide a comprehensive study of their importance and applications in a textbook he wrote in 1715. if c = 0. then the series f (x ) = ∑ f (n) (c) f ′′ (c) f ( n ) (c) (x 2 c)n = f (c) + f ′ (c)(x 2 c) + (x 2 c)2 +  + (x 2 c)n +  n! 2! n! n= 0 ∞ n=0 is called the Taylor series of the function f centred at c.3 Power Series Pn′(x ) = a1 + 2a2 (x 2 c ) + 3a3 (x 2 c )2 + Pn′′(x ) = 2a2 + 2(3a3 )(x 2 c ) + Pn(3) (x ) = 2(3a3 ) + + nan (x 2 c )n 2 1 + n(n 2 1)an (x 2 c )n 2 2 + n(n 2 1)(n 2 2)an (x 2 c )n 2 3 Pn(n )(x ) = n(n 2 1)(n 2 2) (2)(1)an Evaluating Pn and its first n derivatives at x = c . James Gregory (1638– 1675). to be the function f itself. He also wrote a textbook (in 1742) that greatly furthered applications of power series. we obtain the following: Pn (c ) = a0 Pn′ (c ) = a1 Pn′′ (c ) = 2a2 a  Pn(n ) (c ) = n !an Because the value of f and its first n derivatives must agree with the value of Pn and its first n derivatives at x = c. then the series becomes f (n) (0) n f ′′ (0) 2 f ( n ) (0) n x = f (0) + f ′ (0) x + x ++ x + n! 2! n! n= 0 and is called the Maclaurin series for f. that is. then it will take the form as defined below. especially his work in writing binomial expressions as series. n n ∞ x 2c < R then its coefficients are given by the formula f (n ) (c ) . f (0). Similarly. so much so that Taylor series centred at x = 0 became known as Maclaurin series. For the simplicity of notation we agree to define the ‘zeroth’ derivative. Apparently Taylor was not aware of Gregory’s mathematical work when writing his 1715 textbook.

the power series values of x. we have n →∞ n =0 ∑ (21) ∞ n x 2n + 1 . the power series is f ′′ (0) 2 f (n ) (0) n f (n) (0) n x = f (0) + f ′ (0) x + x ++ x + n! 2! n! n =0 f (n ) (0) n 0 21 3 0 4 1 5 x + = 0 + 1⋅ x + x2 + x + x + x ++ 2! 3! 4! 5! n! 2n + 1 x3 x5 x7 n x = x 2 + 2 +  + (21) + 3! 5 ! 7 ! (2n + 1) !  ∑ ∞ Therefore. Thus. (2n + 1) ! lim x 2(n + 1) + 1 x 2n + 3 (2n + 1) ! (2n + 1) ! ⋅ 2n + 1 = lim ⋅ 2n + 1 n → ∞ (2n + 3) ! x (2 (n + 1) + 1) ! x = lim x2 n → ∞ (2n + 3) (2n + 2) 1 = x 2 lim n → ∞ (2n + 3) (2n + 2) = x2 ⋅ 0 = 0 <1 Hence. and evaluate each derivative at x = 0. Solution To use the definition of a Maclaurin series we need to repeatedly differentiate f (x ) = sin x .Example 5 Find the Maclaurin series for f (x ) = sin x and its interval of convergence. n =0 ∑ (−1) (2n + 1) ! for sin x converges for all real n ∞ x 2n + 1 73 . f (0) (x ) = f (x ) = sin x f ′ (x ) = cos x f ′′ (x ) = 2sin x f (0) = 0 f ′ (0) = 1 f ′′ (0) = 0 f (3) (0) = 21 f (4) (0) = 0 f (5) (0) = 1 (x ) = 2cos x 4) f ( (x ) = sin x f (5) (x ) = cos x f (3) The pattern that emerges repeats in blocks of four. the Maclaurin series for f (x ) = sin x is Using the ratio test. sufficiently to establish a pattern for the power series.

This 74 . giving us ‘short cuts’ for finding Taylor series from a known Taylor series. State the interval of convergence.3 Power Series Finding a Taylor series centred at c for a function f(x) Step 1: Step 2: Step 3: Compute several derivatives of f. use an appropriate convergence test to check for convergence/ divergence at both of the interval endpoints x = c 2 R and x = c + R. write the n! (n) ∞ f (c) (x 2 c)n. The above theorem means that term-by-term differentiation or integration of a power series produces a power series that converges. Differentiation and integration of power series If R is the radius of convergence of the power series f (x ) = then f ′ (x ) = n =0 ∑ a (x 2 c) n ∞ n ∞ ∞ d  ∞ d n n n 21  ∑ an (x 2 c)  = ∑ dx an (x 2 c) = ∑ nan (x 2 c) dx  n = 0  n= 0 n= 0 ( ) for c 2 R < x < c + R The radius of convergence of the series produced by differentiating or integrating a power series is equivalent to that of the original series. If 0 < R < ∞. which we state without proof. and ∫ f (x ) dx = ∫  ∑ a (x 2 c)  dx = ∑ (∫ a (x 2 c) ) dx = ∑ n + 1(x 2 c)   ∞ n ∞ n ∞   1 n+1 n =0 n n= 0 n n= 0 for c 2 R < x < c + R. The answer is provided by the next theorem. The ‘zeroth’ derivative. primarily finding a pattern for the values of f (n) (c ). f ( n ) (c) . and try to identify a pattern for the values of f (n)(c). formula for the Taylor series. respectively. 3. We now establish some very useful properties of power series. Two of the properties that make polynomials particularly useful in calculus are the ease with which they can be differentiated and integrated term by term. is f (x) itself. substituting into ∑ n! n =0 Using the ratio test determine the radius of convergence for the series. given that we are in the interval of convergence of the original series. f (0)(x). Evaluate the derivatives at x = c. to the derivative or integral of the function represented by the original series. Knowing that the rule for coefficients of a Taylor series is an = Step 4: Step 5: Finding a Taylor (or Maclaurin) series by this five-step process can prove to be difficult. the interval of convergence may change due to convergence/divergence changing at one or both of the endpoints.4 Operations with power series The power and versatility of representing functions with power series is due largely to the fact that they retain many of the properties of ‘finite’ polynomials. However. It is natural to ask whether the same holds true for power series.

Example 6 Find the Maclaurin series for f (x ) = cos x . Properties of power series ∞ ∞ Given the power series f (x ) = ∑ anx n and g (x ) = ∑ bnx n and constant k. In general. the following n= 0 n= 0 hold true: 1 2 3 f (kx ) = ∑ ank nx n n= 0 ∞ ∞ f (x k ) = ∑ an (x k ) = ∑ anx kn n n= 0 n= 0 ∞ f (x ) ± g (x ) = ∑ (an ± bn ) x n n= 0 ∞ It is important to note that the operations for power series stated here can change the interval of convergence for the resulting series. 2! 4 ! 6! (2n) !  For convenience. the interval of convergence of the resulting series for properties 1 and 2 will be the same as the original series.g. Solution We could apply the formula for a Maclaurin series as we did in Example 5 but it will be more efficient to simply differentiate the power series that we already established for f (x ) = sin x . 75 .gives us another way to derive power series representations from a known power series. so the same is true for this power series for cos x . and for property 3 it will be the intersection of the intervals of convergence of the two original series. n d  ∞ (21) x 2n + 1  d cos x = (sin x ) =  ∑ dx  n = 0 (2n + 1) !  dx  d  (21) x 2n + 1  =∑   n = 0 dx  (2n + 1) !  ∞ n d  (21)n x 2n  = ∑  (2n + 1) (2n + 1) !   n = 0 dx  n 2n ∞ d  (21) x  =∑  Note: (2n + 1) ! = (2n + 1)(2n) ! (2n) !   n = 0 dx  x2 x4 x6 = 12 + 2 +  2! 4! 6! ∞ The power series for sin x converges for all x. as the next example illustrates. 2n x2 x4 x6 n x cos x = 1 2 + 2 +  + (21) + for x ∈ . These are described below. but they also apply for series centred at c. e. ∑ a (x 2 c) n= 0 n ∞ n A further beneficial property of power series is that algebraic manipulations valid for polynomials also work for power series. we state the operations with power series using series centred at 0. Therefore.

∑ (21) n 21 Therefore.3 Power Series Example 7 ∞ 1 is ∑ x n for 12 x n = 0 1 21 < x < 1 to find the Maclaurin series for g (x ) = . so the interval of convergence changes to 1 < x 1. 2x for x into the power series for f (x ) = 12 x g (x ) = f (2x ) = n=0 ∑ (2x ) = n ∞ n=0 ∑ (21) ∞ n xn and we can write this result as 1 n = 1 2 x + x 2 2 x 3 +  + (21) x n +  for 21 < x < 1. we can integrate the power series ∑ ∫ ((21) ∞ n x dx n ) n=0 ∑ (21) ∞ n  1  n +1 x   n + 1  xn n Starting the index at n = 1 rather than n = 0. n 21 ∑ (21) Note: At x = 21. 1+ x b) Use the result from a) and integration of power series to find the Maclaurin series for h (x ) = ln (1 + x ). Thus. 1+ x Note: n=0 ∑ (21) ∞ n x n diverges at both endpoints so the interval of n=0 convergence is the same as for the original series b) Knowing that for ∑x . n ∞ 1 to obtain the power series for ln (1 + x ) . 1+ x  ∞  1 n ln (1 + x ) = ∫ dx = ∫  ∑ (21) x n  dx 1+ x   = = = n=0 ∞ n=0 ∞ n =1 1 ∫ 1 + x dx = ln (1 + x ). 76 . we can write the result as n x2 x3 x4 n 21 x ln (1 + x ) = x 2 + 2  + (21) +  for n 2 3 4 1<x 1. xn becomes the divergent harmonic series n n =1 and at x = 1 it becomes the convergent alternating harmonic series. a) Use the fact that the Maclaurin series for f (x ) = Solution a) We can create the power series for g (x ) = 1 by simply substituting 1+ x 1 .

ex = ∑ n= 0 ∞ xn n! = 1+ x + ∞ n 21 x2 2! + x3 3! + ln(1 + x ) = ∑ (21) n =1 xn n = x2 = x2 x2 2 x3 3 2 2 x4 4 + + 1< x 1 sin x = ∑ (21) n= 0 ∞ ∞ n (2n + 1)! (2n)! n x 2n + 1 x3 3! + x5 5! 2 x7 7! 2∞ < x < ∞ 2∞ < x < ∞ cos x = ∑ (21) n= 0 ∞ n x 2n = 12 x2 2! x4 4! x6 6! 5 + arctan x = ∑ (21) n= 0 x 2n + 1 2n + 1 = x2 x3 3 + x5 2 x7 7 + 1 x 1 Example 8 a) Find the Maclaurin series for f (x ) = e x . 77 . We list these below with their corresponding intervals of convergence. Selected Maclaurin series Function ∞ 1 = ∑ x n = 1+ x + x 2 + x 3 +  + x n +  12 x n = 0 Interval of convergence 21 < x < 1 2∞ < x < ∞ + + The last five of the Maclaurin series listed here are the power series given in the Mathematics HL Information (formulae) booklet. To find the Maclaurin series for e x we can simply substitute x 2 in for x in the Maclaurin series for e x . we can use the same approach 1 + x2 used in Example 7 b) to show that the Maclaurin series for arctan x is 2n + 1 x3 x5 x7 n x arctan x = x 2 + 2 +  + (21) +  for 1 x 1. 3 5 7 2n + 1 Knowing that ∫ We now have derived Maclaurin series for several important functions. These series can be used to construct other series.1 dx = arctan x . 2 c) Use the first four terms of the series in b) to approximate the value of ∫ 1 0 e x dx . 2 Solution a) We prefer not to derive a Taylor series (centred at 0. e x2 =∑ n=0 ∞ (x ) n! 2 n = x4 x6 x 2n = 1 + x2 + + +  for 2∞ < x < ∞ ∑ 2 ! 3! n=0 n ! ∞ This series converges for all real values of x because the original series for e x did so. find a series for ∫ e x dx . in this case) by the direct method used to find the series for sin x in Example 5. if there 2 is an alternative approach. 2 b) Hence.

+ + =x+ + 3 5 ⋅ 2 ! 7 ⋅ 3! 3 10 42 1 We use this to approximate the value of the definite integral We are unable to find a function that is an anti2 derivative of ex with the calculus techniques covered in this course. 78 . Helping us to integrate such functions is a very useful application of power series. when a power series is used (by your GDC. by definition. Taylor polynomials In general the nth partial sum of a Taylor series is Pn (x ) = ∑ f (k) (c) (x 2 c)k k! k =0 n which is a polynomial. 2   x4 x6 x 2n e x dx = ∫ 1 + x 2 + + ++ +  dx ∫ n! 2 ! 3!   =x+ That is.377%. In our derivation of the power series for f (x ) = e x centred at 0 we found four partial sums that were polynomials of degree 1. 2 ∑ ∞ it follows that f (x ) is the limit of the partial sums. then this series does also. c) The first four terms of the series for ∫ e x dx are 2 n=0 ∑ (2n + 1) n ! = x + 1 x2 3 x 2n + 1 x3 x5 x7 x3 x5 x7 + . typically called a Taylor polynomial. a GDC computes the value of ∫ e x dx to 0 be 1.  x3 x5 x7  e dx ≈ x + + + ∫0 3 10 42 0   1 1 1 = 1+ + + 3 10 42 16 51 = = 1 ≈ 1. 3 and 4.462 651 746. recall the discussion at the start of this chapter. for example) to compute an approximate value for a function it does so with a suitable partial sum of the power series that. of degree n. Given the definition of a Taylor series f (x ) = f (n) (c ) (x 2 c )n n! n=0 ∫ 1 0 e x dx . will be a polynomial. x3 x5 x7 x 2n + 1 + + ++ + 3 5 ⋅ 2 ! 7 ⋅ 3! (2n + 1) n !  ∫e x2 dx = ∑ x 2n + 1 n = 0 (2n + 1) n ! ∞ 2 and because the series for e x converged for all real values of x.457142857 (to ten significant figures) 35 35 1 2 Also. Continuing with the idea of using power series for approximation purposes. The percentage difference between the value from the partial sum of the 2 series for e x and the calculator value is only about 0.3 Power Series b) We can take the result from a) and integrate it term by term. As in the last example. 2. to ten significant figures.

The Lagrange form cannot compute the error exactly. then for each positive integer n and for each x in I. 3 and 5 are: x3 x3 x5 P1 (x ) = x P3 (x ) = x 2 P5 (x ) = x 2 + 3! 3! 5 ! y 1 Figure 3. more manageable. can be computed with one of the following formulae: n+ 1 f ( ) (b) Rn (x ) = (x 2 c)n + 1 where b lies between c and x. P5 (x ) = x 2 + 3! 5! p  p 2 p 2 p x 1 In general. x3 P3 (x ) = x 2 and 3! x3 x5 .For instance. however it is.7 The sine function (in black) and the three Taylor polynomials P1(x) = x. f (x ) = lim Pn (x ) . also called the error term. By finding the value of b between c and x (inclusive) that maximizes f (n + 1)(b). the result for Example 5 shows that the Taylor polynomials for f (x ) = sin x centred at 0 for n = 1. generally. we can (n + 1)! compute the maximum error from using an nth degree Taylor polynomial centred at c to approximate f(x) for a specific value of x. is an approximation to f (x ) we can write f (x ) = Pn (x ) + Rn (x ) where Rn (x ) is the remainder. Taylor’s theorem If a function f has derivatives of all orders in an open interval I centred at c. inclusive (Lagrange form) (n + 1)! or Rn (x ) = n→∞ Since the error term for a Taylor polynomial results from ‘cutting off’ the series after a certain number of terms to a finite polynomial it is occasionally referred to as the truncation error. Rn (x ) . That is. k! k =0 The error term. We are now in a position to present the following important theorem that not only gives a formula for the Taylor polynomial for approximating a function f but also two formulae for computing the error term. thereby maximizing n+1 f ( ) (b) (x 2 c)n + 1. ∫ x a f( n+1 ) (t ) n! (x 2 t)n dt (integral form). Because any particular Taylor polynomial. f (x ) = Pn (x ) + Rn (x ) where n f (k) (c) Pn (x ) = ∑ (x 2 c)k is the nth degree Taylor polynomial centred at x = c. Although the integral form is capable of computing an exact value for the error term Rn(x) it is often too cumbersome to evaluate. Pn (x ) . 79 . f (x ) is the limit of the sequence of its Taylor polynomials.

Therefore. f (3) (x ) = 3 . n ∞ x2 x3 x4 n21 x Since ln (1 + x ) = ∑ (21) = x 2 + 2 +  for 1 < x 1.3) is somewhere between 0. (x 2 1) 2 2 3 4 4 b) ln (1. Clearly b b it has a maximum when b = 1. By computing a few derivatives of f (x ) = ln x 1 2 6 1 f ′ (x ) = . As in the previous part. c = 1.262 461 c) For the Lagrange form of the error term we have x = 1.0 × 1026.3) 4 4 + (0.3) = 0. (1. Given our approximation ln (1. Therefore.3) ≈ 0.3) so that the error is less than 0.3. We need to estimate its b 5 ! 120 largest magnitude. Hence.3 Power Series Example 9 a) Find the fifth degree Taylor polynomial centred at x = 1 for f (x ) = ln x . b) Use this polynomial to approximate ln (1.000001 = 1.3 1)5 + 1 = 720 (5 + 1) ! The error is no larger than 0.3) 3 3 2 (0. c) Use the Lagrange error term to determine an upper bound to the error in this approximation.000 1215. 0.262 5825.3) 2 (0. that is.262 461.3) 2 2 + (0. we can see that the nth derivative is f (n) (x ) = (21) xn 5! the sixth derivative at x = b is f (6) (b) = 2 6 . f (4) (x ) = 2 4 x x x x n + 1 (n 2 1) ! .262 3395 and 0.3) ≈ (0.087 48 120 R 5 (1. 2 6 = 6 as b ranges from 1 to 1.3) 4 5 = 0. n 2 3 4 n =1 then ln (1 + (x 1)) = ln x = = (x ∑ ( 1) n =1 ∞ n 1 (x n 1) n (x 1) 1) 2 2 (x + 1) 3 3 (x 4 1) 4 +  for 0 < x 2.3. the fifth degree Taylor polynomial centred at x = 1 is (x 2 1)2 + (x 2 1)3 2 (x 2 1)4 + (x 2 1)5 .3). we can say with absolute certainty that the exact value of ln (1. Solution a) We can construct the Taylor series centred at x = 1 by substituting x 21 for x in the Maclaurin series for ln (1 + x ). n = 5. d) The magnitude of the (n + 1) st derivative is given by n! n n! f (n + 1) (b) = (21) n + 1 = n + 1 . we have x = 1. d) How many terms of the Taylor series centred at c = 1 are needed to approximate ln (1. f ′′ (x ) = 2 2 .000 1215.3 b b 80 .

20125E-6 (X+1) Y1(8) Y2= 2. the first three terms a3 = 600 29400 will give us an estimate with a maximum error of about 0.3).0 × 1026. so 1  b  1. Hence.000 034. Therefore.0016 . 81 .3 .3 2 1) n +1 By trial and error (see GDC images below). Plot1 Plot2 Plot3 Y1= ((.3)^(X+1)) 8.9049E-7 Y5= Y6= Y1(7) Example 10 a) Express the indefinite integral ∫ sin x dx as an infinite series. n! (0. So. n +1 max R n (1.3) = =   (n + 1) ! (1. and the (n + 1) st derivative will be largest when b = 1.and c = 1. a4 = ≈ 0. we determine that the smallest value of n that satisfies this inequality is n = 9. the ninth degree Taylor polynomial would give us an error of less than 1. sufficient to guarantee an estimate accurate to three decimal places. Which is the first term to have a magnitude so that adding/subtracting it will not change the third decimal point of the sum? 1 1 = 0.187E-6 Y3= Y1(9) Y4= 5.3)n + 1 < 1. Solution a) sin x ∫ x dx = = ∫ ∑ (21) (xn + 1)! 2 n=0 x n ∞ n 2n + 1 dx = n ∫ ∑ (21) n=0 ∞ n ∞ x 2n (2n + 1) ! x dx ∑ (21) ∫ (2n + 1) ! dx = ∑ (21) (2n + 1)(2n + 1) ! n=0 n=0 x ∞ 2n 2n + 1 = x2 b) x3 x5 x7 + 2 + 3 ⋅ 3! 5 ⋅ 5 ! 7 ⋅ 7 ! 1 ∫ 1 0 2n + 1  ∞ sin x x n dx =  ∑ (21) x (2n + 1)(2n + 1) !  0  n = 0 ∞ 1 n = ∑ (21) n=0 (2n + 1)(2n + 1) ! = 12 1 1 1 2 + + 18 600 29400 We can use the alternating series estimation theorem that says that the error generated in using the nth partial sum of an alternating series to approximate its sum does not exceed the value of the (n + 1)st term. x 1 sin x dx accurate to three decimal b) Evaluate the definite integral ∫ 0 x places.000034.0 × 1026 when approximating ln (1.

3 Power Series fnInt(sin(X)/X. 1 (Example 1).9460830704 Therefore. Exercise 3. 2 c Use the first five terms of the series from b to estimate the error for this estimate is less than 0.946 A GDC confirms our result. 1 sin x 1 1 ∫ 0 x dx ≈ 1 2 18 + 600 ≈ 0. determine the radius of convergence and interval of convergence of the power series. 2 A B 1 = + . 12 x 2 1 + x 12 x 1 b Express ∫ e − x dx as a Maclaurin series and determine its radius of convergence. State the interval of convergence for the series. ∑ (kx ) n= 0 ∞ n in terms of k. 12 x 82 (12 x )2 by differentiating the Maclaurin series for .1) .001. Show that 2 18 Use multiplication or division of power series to find the first three non-zero terms in the Maclaurin series for each function. ∫ 1 0 e − x d x . Also determine the interval of convergence for the 12 x 1 by replacing x with 2x in the geometric series 1+ x b Use the technique of partial fraction decomposition to find A and B so that and determine its interval of c Hence. 17 a Find the Maclaurin series for e2x and determine its radius of convergence. express the interval of convergence of 16 a Find a power series for for series. construct a power series for 12 x 2 convergence.0.1 In questions 1–14. a b c f (x ) = x sin x g (x ) = tan x f (x ) = e x ln(12 x ) 1 Hint: tan x = sin x   cos x    19 Find a Maclaurin series for 1 . 1 3 5 7 9 n =0 ∑ n+1 ∞ ∞ xn 2 4 6 ∑ n(x 2 2) n =1 ∞ ∞ n ∑ ∞ n= 0 ∞ (x 2 3)n n n =1 ∑ (21) n =1 ∞ n (2n 2 1)! x 2n 2 1 ∑2 n= 0 ∞ x n n ∑ ∑ n= 2 ∞ ∞ (x 2 2)n n2 lnn n x n 3n x n n4 n nx n 3n n =1 ∑ (21) ∑ n! x n= 0 ∞ n (x 2 1)n 8 10 12 14 n ∑ ∑ n= 0 ∞ ∞ n =1 (n !)2 n x 11 ∑ n = 1 (2n !) 13 ∑ ∞ n =1 n! x n nn ∑ n= 0 (21)n n! (x 2 4)n 3n 15 Given k > 0.X .

d Use the Lagrange error term to determine an upper bound to the error in this approximation. c=0 31 Use series to evaluate each limit.     2 x3 3! + x 5 is accurate to within 0. and thus show that ∞ (21)n π .) 30 Compute the Taylor series for each of the following functions f(x) centred at the given point c. show that arctan x = x 2 x3 3 + x5 5 2 x7 7 +  + (−1) n x 2n + 1 2n + 1 + . b By differentiating term by term the power series in a.   c Use the Lagrange error term to determine an upper bound to the error in this approximation. n! 21 a Write down the seventh degree Taylor polynomial for sin x. the nth derivative of f(x) with respect to x. c = 0 3 d f (x ) = (x 21) ln x . c Show that this series converges for −1  x  1. derive the Maclaurin series for f(x) up to and including the term in x 4. a c f (x ) = e x . n =1 ∞ 1 1 29 Find a power series for sec2 x. ∑ 2n + 1 = 4 n= 0 d Use the first six terms of the series to estimate the value of p.005.  24 Find the Maclaurin series for ln  1+ x  .  12 x   1 . (Hint: consider the derivative of tan x. π b Use this polynomial to estimate sin   . Compute the maximum error for this estimate.20 a Find a power series for x 2e2x. 1+ x 2 25 a Find the Maclaurin series for the function f (x ) = b Hence by integrating each term. n =1 ∞ 2n x n 23 Find the first four non-zero terms of the Taylor series centred at x 5 1 for the function f(x) 5 (x 2 1)ex. c = 2 f (x ) = b f (x ) = sin (x 3 ) . 27 Estimate the range of values of x for which the Maclaurin approximation sin x ≈ x 2 1 c Use the result to find a rational approximation to f   . b Hence. c = 1 (12 x )3 1 . a x →0 lim 12 cos x e x 2 12 x b lim x →0 x 2 arctan x x3 83 . 26 The function f is defined by f (x ) = e x + e2x . 2 a Obtain an expression for f (n)(x). 5! 28 Find a power series for x e x and then integrate the resulting series term by term from 0 to 1 to show that ∑ n !(n + 2) = 2 . 22 Determine all the values of x for which the series  12 ∑ ln(n + 1) converges. show that ∑ (22) n+1 n+2 = 4.

3 Find the first three non-zero terms in the Maclaurin expansion of e x sin x. h (x ) = x 2 2 5x + 6 84 . State this approximate value. 2 c Find an upper bound for the error in this approximation. write down a series for cos2 x up to the term containing x 4. or otherwise. 0  x < 2 . 1 3 1 5 1 7 d Hence. Practice questions 1 Find the fourth degree Taylor polynomial centred at x = 0 for the function π f (x ) = ln(cos x ) . 2 b Find an expression for the error term R4(x).1. 1 x 2n − 1 c Show that this series converges for x 1. find the first four terms in the Maclaurin series for the function x +1 . 12 a Find the first four terms in the Maclaurin series for both of the following functions: f (x ) = 1 1 and g (x ) = 1+ x 12 x b Rewrite the function h (x ) = x +1 as the sum of two fractions. and find an error bound when x = 0. 2 1+ x n=1 1 . find the exact value to which the series 12 + 2 +  converges. b Determine the fourth degree Taylor polynomial centred at x = 0 for e x . x 2 5x + 6 2 c Hence. b Use the first three terms of this series to approximate 2 ∫ 1 0 e2x dx . show that arctan x = ∑ (21) n=1 ∞ n+1 2n 2 1. 8 a Find the fourth degree Taylor polynomial centred at x = 0 for the function f (x ) = 4+x. find the fourth degree Taylor polynomial centred at 2 x = 0 for e x + x . b Hence. determine how many terms of the Maclaurin series for cos x are needed to approximate cos(5°) correct to six decimal places. 10 a Find the Maclaurin series for e 2x . 7 Find the Maclaurin series for In(2 + 3x) and find an expression for the error term Rn(x). 2 a Find the Maclaurin series for sin2 x up to the term containing x 4. 4 Find the first four terms of the Maclaurin series for the function f(x) = e 3x. 5 Find the Maclaurin series of sec x up to the term containing x 4.3 Power Series Practice questions 1–70 cover series work from Chapters 1–3 inclusive. 6 a Write down the fourth degree Taylor polynomial centred at x = 0 for ex. 11 a Find the Maclaurin series for the form ∞ 1 = ∑ an where an is in terms of x and n. 9 Using the formula for the Lagrange form of the error term. c Hence. Express the result as an equation in 1+ x 2 b By integrating both sides of the equation found in a.

State the interval of x = convergence. that is. sin x = ∑ (21)n n= 0 ( 2n + 1)! b Using the Maclaurin series of g (x ) = sin (x 2 ). ( x 2 2 )3 2 2 x 2 ( x 2 2 )2 2 2 x 2 Use Maclaurin series to expand f (x). show that the value of approximately 0. k +k 2 2 x 2 . a b c ∑ k =1 ∞ k =2 ∞ ∞ k +1 3k 1 3 ∑ k(1n k ) ∑ (21) k =1 k +1 k k2 + 1 15 Express ∞ 3x 2 4 in partial fractions. Find the value of k.394 38. evaluate the definite integral ∫ 1 0 sin (x 2 ) dx correct to four decimal places. b The third and fourth derivatives of f (x) with respect to x are given by: 12( 2x 2 + 1) 26x f ′′′( x ) = 2 and f ( 4 ) ( x ) = 2 . x2 2 x 16 Discuss the convergence or divergence of the following series: a b ∑ n= 0 ∞ n= 0 n2 2 n + 1 3n ∑n nk 2 1 . and to show that x2   2 2 x 2 ≈ 2 12 + kx 4  . find the Taylor 12 (x + 1) 1 series centred at x = 21 for the function f (x ) = .   4 c Hence. or otherwise. ∫x 0 1 2 2 2 x 2 dx is d Use the substitution x = 2 sin θ to show that the exact value of 1 π 2 2 ∫ x 2 2 x dx is . b Using the result from a and the identity 1 ∑x n=1 ∞ n 21 where 21 . 85 . indicating the tests used to arrive at your conclusion.13 a Write down the function represented by the power series the interval of convergence is 21 < x < 1. Hence. find an approximation of p to four decimal places. Show that 17 a Consider the function f (x) = f ′′( x ) = 2 ( x 2 2) 2 2 x 2 2 . expansion of sin x. k is a positive integer. x 14 Test the convergence or divergence of the following infinite series. 0 8 18 a Find the Maclaurin series of the function g(x) = sin (x 2) using the series ∞ x 2n + 1 .

or otherwise. and show that this polynomial has the same maximum value as f (x) itself. 12 ln x  . then find an approximation for the sum to two decimal places. 2k ( 2k + 1) k =1 n 1 k k =1 n 2n 1 + 2(21)k and S2 n = ∑ uk . 21 a Find a Maclaurin series expansion for f (x) = ln(1 + x). explain why this is so. 2 b Hence. showing clearly the local maximum of the function and its horizontal asymptote. n n=1 n ∞ 25 Consider the sequence of partial sums {Sn} given by Sn = ∑ . justifying your answer. x→∞ 1 x x c Find the Taylor expansion of f (x) about x = e. n = 1. b Rn is the error term in approximating f (x) by taking the sum of the first (n + 1) terms of its Maclaurin series. n+1 22 Test the convergence or divergence of the following series: ∞ ∞ cos nπ 1 a ∑ sin . prove that the sequence {Sn} is not convergent.3 Power Series 19 a Use the ratio test to calculate the radius of convergence of the power ∞ ( x 2 5) n . …. 26 For positive integers k and n let uk = a Show that S2 n = ∑ 4k 2 1 . up to the second degree term. ( 0  x < 1). 2. k +1 k =1 b Hence. b ∑ ( n + 10 )  1. You may use the fact that ln x lim = 0. 24 Determine whether the series ∑    1  1 1 +  converges. 20 Test whether the following is a convergent series: ∑ (21) n =1 ∞ n+1 1 . ∑u k =1 ∞ k is convergent or 27 a Describe how the integral test is used to show that a series is convergent. 86 . a Show that the derivative f ′( x ) = f ( x )   x2  b Sketch the function f (x). 1 a Show that for all positive integers n. S2n > Sn + __ . where x ∈ ℝ+ .4    n= 1 n n =1 n 23 Consider the function f ( x ) = x . for 0 < x < 1. Prove 1 Rn  . determine all points x where the power series given in a converges. series ∑ 3 n =1 2 n b Using your result from part a. b Test the series ∑e n =1 ∞ n n2 for convergence. if it is not. n! If it is. Clearly state all the necessary conditions. determine whether the series not.

up to and including the term in x 4. calculate lim  35 Find the radius of convergence of the series 36 a i ii ∑ k !( k 2 1)! x k =1 ∞ ( 2k 2 2 )! k . or otherwise. n2 ∑ sin n n= 1 ∞ 2π 2 is convergent. 1 1  32 Calculate lim  2 . 30 Find the range of values of x for which the following series is convergent. find the Maclaurin series for ex sin x. n2 n 2 1 n c Hence. x → 0  ln cos x   iii y = tan x. Deduce the nth term of this series. Find the first four derivatives with respect to x of y = ln(1 + sin x). for y is 1 1 1 y = x 2 x 2 + x 3 2 x 4 + …. for i ii y = ln(1 2 sin x) y = ln cos x  tan( x 2 )  . iii What is the upper bound for the error in this approximation? b i ii iii Use the ratio test to show that the series is convergent for all values of x. Give your answer to six decimal places. c Hence. show that the Maclaurin series. show that 1 < S < 2p. 87 . ( 2n 2 1)! Prove that the alternating series given by ∑ (21) n =1 ∞ n21 Approximate the series by finding the fourth partial sum. ∞ xn ∑n+1 n= 0 31 a Show that the series Let S = ∑ sin n =1 ∞ 2π . x →0   x2 b Hence. Hence. x 29 Find the Maclaurin series of the function f (x) = ln(1 + sin x) up to and including the term in x 4. 2 6 12 b Deduce the Maclaurin series. up to the term in x 4. 2 2   x c Use the result of part b to find lim  e sin 3 2 x  . Find the first four non-zero terms of the Maclaurin series for sin x. 1 converges.28 a Find the first four non-zero terms of the Maclaurin series for i sin x ii ex . x →0  x   sin x  33 Use the integral test to show that the series 34 a i ii ∑n n= 1 ∞ 1 p is convergent for p > 1. iii Use your series for sin x to find the first four non-zero terms of the Maclaurin series for cos x. b Show that for positive integers n > 2. 1 1 1 < 2 . up to the term containing x 5.

1 c Find an approximate value for p by putting x = _ and summing the first 2 three non-zero terms of this series. 42 Determine whether the series 43 Consider the series S = ∑ 1 . where L ∈ ℝ. 39 Consider the function f defined by f (x) = arcsin x. 40 The general term of a sequence is given by the formula an = a Given that lim an = L . n→ ∞ n2 + 3n . where un > 0. b Use the integral test to determine the set of values of k for which the series ∞ 1 ∑ n(ln n)k n= 2 i ii is convergent is divergent. valid for odd n > 3.3 Power Series 37 Find the value of a lim 1+ x 21 x →0 x b lim x ln x . n! n =1 ∞ ∑ ( n + 1) n= 0 ∞ 2n is convergent or divergent. b Use a comparison test to show that S < 2. n ∈ ℤ+ . for n > 1. 88 . find an expression for an in terms of n. b Find the radius of convergence of this Maclaurin series. c Write down the exact value of S. Given that a1 = 1. (n + 1) (n + 2) an + 2 = n 2an. The derivatives of f (x) satisfy the equation (1 2 x 2) f (n + 2) (x) 2 (2n + 1) x f (n + 1) (x) 2 n 2 f (n) (x) = 0. You may assume that the series contains only odd powers of x. with a reason. find the value of L. 2 n2 2 1 b Find the smallest value of N ∈ ℤ+ such that an 2 L < 102 3 for all n > N. The coefficient of xn in the Maclaurin series for f (x) is denoted by an. a Use the ratio test to prove that this series is convergent. for |x | < 1. 41 Estimate the range of values of x for which the Maclaurin approximation sin x ≈ x 2 x3 3! + x5 5! is accurate to within 0. Give your answer to four significant figures.005. x →0 38 a i ii Given that ∑ un is convergent. for n > 1. n= 1 n =1 ∞ ∞ State. whether or not the converse of this result is true. prove that ∑ un2 is also convergent. a i ii Show that.

48 Consider the infinite series ∑ n( n + 2) . 1. x 23 1 1 = lim = . find the values of a. find ∑ n( n + 2) . the nth partial sum of c Hence. n =1 n=1 ∞ 1 49 Find the interval of convergence of the series 50 Find:  s 2 3s + 4   a lim  s→ 4   42s ∑ sin  n  x   ∞ π n . 89 .” “Using l’Hôpital’s rule. and d.  cos x 2 1 b lim   x → 0  x sin x  51 a Find constants A and B such that A B 1 = + . b i Express n( n + 2) ii Hence. b Given ∫In x dx = x ln x 2 x + C. show that the series converges. c. 3 2 6 ∫2n 1+ x n→ ∞ αn x dx 2 d x where α is a positive constant and n is a positive Determine lim In .  tan βx 2 β tan x   sin βx 2 β sin x  b Using l’Hôpital’s rule find x→∞   lim where β is a non-zero constant different from ±1.44 a Show that the polynomial approximation for ln x in the interval [0. n =1 ∞ 1 a Show that the series is convergent. state whether the following argument is correct or incorrect. + 8 r 2 3 ∫ ∞ 1 1 xp dx converges. show by integrating the above series that another approximation to ln x is given by x 3 x 2 3x 5 1 2 2 2 ln x ≈ + . 16r 2 + 8r 2 3 4 r 2 1 4 r + 3 b Find an expression for Sn.5? 45 a i ii Find In = integer. b.5. x →1 x 2 1 x → 0 12 cos x b Giving a reason. 1 in partial fractions. 52 Find the values of p for which ∑  16r  r =1 ∞ 2 1 . lim 2 x →1 x 2 3 x → 3 2x 6 47 Given that the Maclaurin series for esin x is a + bx + cx 2 + dx 3 + … . 12 2 2 6 4x c Which is the better approximation when x = 1.5] obtained by taking the first three non-zero terms of the Taylor series about x = 1 is given by 11 x 3 3x 2 2 ln x ≈ + 3x 2 . 46 a Use l’Hôpital’s rule to find ln x 2 tan2 x ii lim i lim .

q and r such that your series in c i is identical to your answer to b.  2 iv Use the Lagrange error term to determine an upper bound to the error in this approximation.  a  2  a + a + 1 a +1 ii Deduce that 1 3 < ln 2 < . 1 iii Use your result to find a rational approximation to f   . 2 Hence. for all a > 0. b Given that Hn is the nth partial sum of the harmonic series. the nth derivative of f (x) with respect to x. 2 4 iii Using i find p. show that 1 1 Hn 2 1 < ln n < Hn 2 2 . Comment on this result. 90 . for n > 1. b Use the integral test to determine whether the series or divergent. derive the Maclaurin series for f (x) up to and including the term in x 4. q. ln( a 2 x 3 )  2 24 56 a Determine lim  x →0  b Calculate lim   ln(1 + x 2 )  . x → 0  ln(12 x 2 )   2 + x 2 2 2 cos x  c Show that lim  x = 1. prove that the terms of the sequence {yn : n > 1} decrease as n increases. 2 2n c If yn = Hn 2 ln n. q ∈ ℚ such that p < ln 3 < q. 55 a i ii ∑ ln n 2 is convergent n=1 n ∞ State the domain and range of the function f (x) = arcsin (x).  ln( a 2 + x 2 )  . r ∈ ℝ. show x that 1 a + 1 1  1 1  < ln  < . c i ii Find the Maclaurin series for (p + qx 2)r up to and including the term in x 4 where p. where a is a positive constant. Find values of p.3 Power Series 53 Calculate the following limits: a lim 2x 2 1 x 3 (1 + x 2 ) 2 2 1 b lim x → 0 ln(1 + x ) 2 x x →0 54 a The function f is defined by f (x) = i ii Obtain an expression for f (n) (x). not equal to 1. e x + e2x . x → 0  e + e2x 2 2 cos x   1 57 a i By considering the graph of y = __ between x = a and x = a + 1. b Use the small angle approximation y2 y4 cos( y ) ≈ 12 + to find a series for cos (arcsin (x)) up to and including the term in x 4. Determine the first two non-zero terms in the Maclaurin series for f (x).

63 a Find the radius of convergence of the series b Determine whether the series ∑( n= 0 ∞ ∑ ( n + 1)3 . show that ln sec x = e Hence. b Find the first three derivatives of f (x) and hence show that the Maclaurin x2 x3 series for f (x) up to and including the x 3 term is x + + . or otherwise. 1 + sin x x2 2 + x4 12 + …. 0 3 4 5 6 x The diagram shows part of the graph of y = 3 together with line segments x parallel to the coordinate axes. find upper and lower bounds for ∑ 3 . 2 x → 1  12 cos x 62 a Given that y = ln cos x. find lim ln sec x x →0 21 . e How good is this upper bound as an estimate for the actual error?  ln x  . d By combining your two series. a Using the diagram. c Use this series to find an approximate value for ln 2. show that 1 1 1 + + + …. 91 .  12 x   d Use the Lagrange form of the remainder to find an upper bound for the error in this approximation. 64 The function f is defined by f (x) = In(1 + sin x). show that the first two non-zero terms of the Maclaurin series for y are 2 x2 2 2 x4 12 . 33 4 3 53 ∞ 1 b Hence. ∞ b Use this series to find an approximation in terms of p for ln 2.58 Determine whether the series 59 y ∑ 10 n =1 ∞ n10 n is convergent or divergent. n= 0 n (21)n x n 3 n3 + 1 2 n) is convergent or divergent. a Show that f ′′( x ) = b Determine the Maclaurin series for f (x) as far as the term in x 4. n =1 n 1 1 1 + + +…< 4 3 53 63 1 ∫ ∞ 1 3 x3 dx < 60 The function f is defined by f (x ) = ln   a Write down the value of the constant term in the Maclaurin series for f (x). x x . c Deduce the Maclaurin series for ln(1 2 sin x) as far as the term in x 4. 2 3 1  . 61 a Find the value of lim1 x →  sin 2π x    12 e x  2 b By using the series expansions for ex and cos x evaluate lim  .

21 1. if they exist.1 1. 68 Find the limits. 1 1 1 1 + + + + 1. 66 Write down the conditions for the integral test to be applied to the series ∑a n=1 ∞ n and use it to determine if the series ∑ n e2n n=1 ∞ 2 converges. of the following sequences {un} as n tends to infinity. with reasons. find an approximate value to sin  +  . 67 Use Taylor’s series to express the polynomial p (x ) = x 4 + 3x 3 + x 2 + 2x + 1 in terms of powers of (x 21) . correct to three significant   2 8  figures. Questions 14–70 © International Baccalaureate Organization 92 .3 Power Series 65 Find the coefficient of x 8 in the Maclaurin expansion of x sin2x . if the following series converge or diverge.4641 13 23 33 4 3 + + + + c 1! 2 ! 3 ! 4 ! a 1+ b e + e + e + e + 4 1 2 1 3 1 d 1 2 3 4 + + + + 1× 3 3 × 5 5 × 7 7 × 9 e 3 3 3 3 + + + + 13 23 33 4 3 1 70 a Find the first three terms of the Taylor series centred at x = _ for the 2 function f (x ) = sin(π x ) . a un = c un = sin(n) n b un = arctan(n) d un = en n! 1 1 2 n2 2 1 5n2 + 11 e un = (21)n n e2n  f un = n 2 cos   2 sin   2 2       n  n   69 State.331 1. π π b Hence.

Differential equations come in a great variety of forms. a dependent variable (usually y).4 Differential Equations 4. y. y .  = 0  dx   dy d 2 y  second order differential equation F  x . and one or more derivatives of an unknown function y = f(x) [or y = f(t)]. y . Analytic methods are not always successful in solving a differential equation. 2 . A differential equation is an equation that relates an independent variable (commonly x or t). 2 .8 again before working through this chapter. 2  = 0 dx dx   dy   first order differential equation F  x . so we will also investigate a graphical approach and a useful numerical method for approximating the solution to a differential equation.  = 0  dx  dy   first order differential equation F  x . y . Here are some examples: dy dy + y 2y = 0 1 x dx dx 2 3 4 dy y 2 2 y + =0 dx x2 d 2y dy + 3 2 5y = 0 2 dx dx dy + y sin x 2 e cos x = 0 dx 2 dy   first order differential equation F  x . The general form of a differential equation (with independent variable x) can be written as dy d 2 y dk y   F x.  = 0  dx  third order differential equation  dy d 2 y d 3 y  F x. We were given an introduction to differential equations in the core syllabus in Chapter 16. We will review that technique and some other important concepts from the core syllabus. y .8) we studied one analytic solution method for a certain class of differential equation. . You are strongly encouraged to read through Section 16. y. Then we will explore further into the topic by considering two further classes of differential equations. 3  = 0 dx dx dx   93 d 3y  dy  5 2 3 + (ln x )   + 4 xy = 0  dx  dx . graphical and numerical – exist for finding their solutions. .1 Introduction Equations involving an unknown function and its derivative(s) are called differential equations and frequently occur in mathematical models of reallife phenomena. In the core syllabus (Section 16. and many different procedures – analytic. . k = 0  dx dx dx  dk y where the largest k for which occurs in the equation is called the order of the dx k differential equation..

x +1 1 y=∫ dx = ln x + 1 + C . . of the unknown function can be isolated on one side of dx the equation. dx and f (‘small f’) represents the unknown function whose dy .1 shows a few members of this family. Hence. x dx x + y dx dx The solution of a differential equation is the (initially unknown) function dy y = f (x ) whose derivative is . x ≠ 21 x +1 dy 1 = is the explicitly So the solution of the differential equation dx x + 1 defined function y = ln x + 1 + C where C is an arbitrary constant. For the sake of simplicity. y ) is dx For this course. = dx x + 1 1 Every solution of this equation is an anti-derivative of . 2 and 4 above.1 y 10 8 6 4 2 6 4 2 2 2 C = 10 C=5 C=0 4 6 x C = 5 4 6 8 10 12 y = 1n | x + 1 | + C C = 10 94 . In a first order differential equation. the first dy derivative. This is called a general solution because it is not a single function. Consider the differential equation dx 1 dy . Figure 4.4 Differential Equations A differential equation may use symbols for the independent and dependent variables other than x and y. but an infinite ‘family’ of functions dependent on the constant C. x ≠ 21. y ) dx dy is expressed as a function in terms of x and y. Figure 4. we only study first order differential equations. Note that the first where dx order differential equations 1. such as equations 1. For example. we will use x and y while we are developing theory and solution methods for differential equations. Also note that we are using F (‘large F’) to represent a two-variable function that when set equal to dy is the differential equation. a simpler general form for first order differential equations is dy = F (x . slope at the point (x . 2 and 4 can all be re-written in this form. dy y dy dy + y 2y = 0 ⇒ = 1.

Sometimes the solution of a differential equation will be expressed as an implicitly defined function. However. for any real number C the function ln y = + C is a solution. we examine a useful graphical method for helping us to sketch the function. y It is an equation relating x and y and implies a function exists that defines y as a function of x. it is sometimes not possible to solve for y. x In general. we differentiate – applying implicit differentiation and the product rule:  d x d (ln y ) =  + C   dx  y dx d 1 dy d = (xy 21) + dx (C ) y dx dx dy  1 dy  = y 21 + x  − y 22  + 0  dx  y dx 1 dy 1 x dy = 2 2 y dx y y dx 1  1 dy  x dy  y2  = y2  2 2   y dx    y y dx  dy dy + y 2y = 0 dx dx x Therefore. Before we delve into the details of these analytic methods. to the differential equation x dx dx x means that the coordinates x and y of any point on the curve ln y = + C y dy combined with the value of the derivative at that point will solve the dx dy dy + y 2 y = 0.In contrast. when we are given some initial conditions that allow us to evaluate a particular value for C we obtain a single function that we call a particular solution of the differential equation. giving C = 5 and the particular solution of y = ln x + 1 + 5. equation x dx dx The only type of first order differential equation covered in the core syllabus of this course is a class of differential equations referred to as separable equations. y dy dy + y 2 y = 0. For example. To verify that this is a solution to 1. or family of functions. This in implicit form. that solves a differential equation. the general solution to equation 1 is x ln y = + C . We solved these using a technique called separation of variables. One of our key goals in this chapter is to develop an analytic solution method for each of two further classes of first order differential equations. 95 . For example. we wish to find the explicit solution of a differential equation written in the form y = f(x) where f is a known function. if we are given the initial conditions that y = 5 when x = 0 then we can solve for C. In such a case we must settle for an implicit solution written in the form g(y) = f(x) where g and f are known functions and g(y) ≠ y.

It is an example of a first order linear differential equation. y 6 5 4 3 2 1 6 5 4 3 2 1 0 1 2 3 4 1 2 3 4 5 6 x Figure 4. dx 5 6 As you can imagine.4 Differential Equations 4. A plot of these line segments showing the slope (or direction) of the solution curve is a called a slope field (or direction field) for the first order differential equation. y ) specifies the A first order differential equation in the form dx slope of the solution curve y = f (x ) at each point in the xy-plane where F is defined. dy for a large number of x and y values. It can be solved analytically with one of the techniques that we develop later in the chapter. slope fields are easily generated by suitable graphing technology. Carried out by hand or by technology. y ) in the plane. In practice. the differential equation dy = x 2 y is not easy dx to solve. the graph of the dx equation F (x . it can be quite tedious to draw a slope field by hand. As a rule. directions.3. and its general solution is y = Ce2x + x 21.2 Slope fields Often the primary objective when solving a first order differential equation is to find an explicit solution. there is a method that simplifies the process of doing it by hand. we must resort to graphical and/or numerical methods. the segments are drawn at representative points evenly spaced in both dy = x 2 y. called an isocline. For such equations. the isoclines are x 2 y = c . many differential equations used in mathematical models cannot be solved by means of an analytic method. y ) = c is a line. For some constant c. However. we look for Rather than compute dx dy points where has the same value. a graphical method provides us with rough qualitative information about the graph of a solution to a differential equation. shows (in red) equation dx 96 . dy = F (x . For the differential dy = x 2 y .2 shows a slope field for the equation dx Although it looks fairly simple.2 Slope field for dy = x 2 y. Figure 4. We can use this fact to draw a short line segment whose slope is F (x . However. Figure 4. y ) at any point (x . along which all the short line segments of a slope field have the same slope c.

Solution Isoclines are not always straight lines. To use a slope field to sketch a particular solution all we need to know is one point (an initial condition) that the solution curve passes through. dx ‘Isocline’ comes from ‘iso-’ meaning equal and ‘-cline’ meaning slope. we can 3 create a slope field by easily drawing multiple line segments along it all having the same slope. are each a general solution of the equation. a family of solution curves can be produced. the y-axis is an isocline where all dx y the line segments are horizontal. Isoclines are analogous to contour lines on a map indicating land of equal elevation. it is more difficult to use them to sketch a slope field. When isocline curves are not lines. 97 . a) Rather than evaluating dy x = 2 for a large number of x and y values. Consider the differential equation dy = x 2 2 y that has dx isoclines that are parabolas with equations of the form y = x 2 2 c.3 Slope field and three dy isoclines for = x 2y. It is recommended that you draw isoclines lightly in pencil. They serve to ease the process of drawing a slope field. Example 1 a) Draw a slope field for and 5 they? y dy x = 2 on the xy-plane such that 5 x 5 dx y 5 . Sketch some sample solution curves. Solutions to a differential equation can be sketched by drawing in curves that are at each point tangent to the line segment at that point. y 6 5 4 3 2 1 6 5 4 3 2 1 0 1 2 3 4 5 6 1 2 3 4 5 6 x c=3 c = 1 c=0 Figure 4. 0 and 3. where c is a constant. and preferably dashed. dx y x we establish some isoclines by looking for points where 2 has a y constant value. dy x If = 2 = 0 then x = 0. Thus.the isoclines for c = 21. What shape are b) Confirm that both y = c 2 2 x 2 and y = 2 c 2 2 x 2 . Be aware that isoclines themselves do not give any direct information about solution curves for the differential equation. By first tracing in a few isoclines. Hence.

dy x = 2 = 21 then y = x is an isocline where all the line segments If dx y have a slope of 21. 2 In fact. Hence. Figure 4. Three members of the family are drawn in Figure 4. If necessary. dy on the left dx 98 . such as: 1 y = 2x is an isocline where all the line segments have a slope of 2 .4. Drawing curves parallel to the line segments gives a family of solution curves that appear to be circles. dy x If = 2 = 1 then y = 2x is an isocline where all the line segments dx y have a slope of 1. dx y The resulting slope field – showing six lightly drawn isoclines – is shown below in Figure 4. we compute side and substitute c 2 2 x 2 for y on the right side. we can continue in this manner and establish further isoclines. any line passing through the origin will be an isocline for the dy x slope field for =2 .4 Differential Equations If y = 0 (x-axis). the x-axis is an isocline dx where all the line segments are vertical (undefined slope). then dy is undefined.5.4 y = x y = 2x y 5 4 y= x 2 3 2 1 5 4 3 2 1 0 1 2 3 4 5 x 5 4 3 2 1 1 2 3 4 5 y= x 2 y = 2x y=x y 5 4 3 2 1 1 2 3 4 5 x Figure 4.5 1 2 3 4 5 b) Checking that y = c 2 2 x 2 is a solution. 2 1 y = 2 x is an isocline where all the line segments have a slope of 2.

Checking that y = 2 c 2 2 x 2 is a solution.D. and the solution y = 2 c 2 2 x 2 is the family of curves consisting of the lower half of each circle.E. = 10 2 dt 360 v 100 90 80 70 60 50 40 30 20 10 0 5 10 15 20 t 99 Figure 4. in metres per second. The solution y = c 2 2 x 2 is the family of curves consisting of the upper half of each circle. at time t seconds of a 75 kg skydiver falling from an aeroplane is given by the equation dv v2 .dy x =2 dx y d dx ( c 2x ) = 2 2 2 x c 2x2 x 2 1 − 1 2 (c 2 x 2 ) 2 (22x ) = − 2 2 2 c 2x x x − =2 c2 2x2 c2 2x2 Q.D. Example 2 A model for the velocity v. dy x =2 dx y   x d 2 c 2 2 x 2 = 2  2 2 dx 2 c 2 x  1 2 x 1 2 (c 2 2 x 2 ) 2 (22x ) = 2 2 c 2x2 x x = 2 2 2 c 2x c 2x2 ( ) Q.6 .E.

Due to increasing air resistance the skydiver reaches a maximum velocity.4 Differential Equations a) From the direction field shown in Figure 4. and (iii) v (0) = 90.7 but with equal scales on the axes. Thus.6.7.7 v 100 90 80 70 60 50 40 30 20 10 0 dv v2 satisfying v (0) = 0.4 and 4. Solution a) Solutions to Figure 4. (ii) v (0) = 35. this is not an error.6 and 4. 4.7 below.3. Figure 4. b) Explain why the value v = 60 is called the terminal velocity for this situation.8 0 10 20 30 40 50 t 100 . Note that the scales on the axes for the slope fields in Figures 4. v 80 70 60 50 40 Figure 4. The scales are not equal on the axes in Figures 4. sketch the solution curves with the following initial conditions: (i) v (0) = 0.5 are equal. or terminal velocity. Sometimes. 5 10 15 20 t b) From the slope field it appears that all solutions have a limiting value of 60 as t goes to infinity.8 shows a portion of the same slope field given in Figures 4.6 and 4.2. so the line segments do not give a true indication of the slope. the short line segments accurately depict the true slope for solution curves. it is necessary to have unequal scales in order to show an appropriate interval of values for the independent and dependent variables. However. v (0) = 35 and = 10 2 dt 360 v (0) = 90 are sketched in Figure 4. of 60 metres per second. 4.

x > 0. Example 3 Find the general solution of the differential equation dy x 2y = x + 1. be written in the form dx 2 3 dy x dy 1  3  = =  2 x 2 . Some algebraic manipulation is needed to confirm that the differential dy = p (x ) q ( y ). dx Separable equation A first order differential equation is considered separable if it can be written in the form dy = p (x ) q ( y ) . y > 0. For example. the equations = = x 2 + y 2 and = 1 + xy dx ln y dx dx are not separable. For a first order differential equation where this separation of variables can be accomplished.4.g. in fact. However. It is not always obvious whether or not a differential equation is separable. y).8 of the book) and which can be solved analytically using integration is the class of separable equations. y ) can be factored into a product of two functions – one involving only the independent variable (e. dx The variables can then be separated by writing the equation in the form 1 dy = p (x ) dx q (y ) and integrating both sides gives 1 ∫ q (y ) d y = ∫ p (x ) d x + C which leads to a general solution. dy = F (x . equation can. These are differential dy = F (x . the function F (x . Although there are two integrals in the equation 1 ∫ q (y ) d y = ∫ p (x ) d x + C . That is. 2 is separable because it can be written as  dx xy y dx y  x 2 tan x dy = is also separable because it can be written as and y dx ln y 2y dy dy dy cot x .3 Separable equations A class of first order differential equations with which we are already familiar (Section 16. x) and the other involving only the dependent variable (e. y ) that can be rewritten so that the variables x and y equations dx (along with their differentials dx and dy) are on opposite sides of the equation. only one constant of integration is needed.g. We could add a constant to both sides but they could then be combined into one constant. y ) = p (x ) q ( y ) . dx Solution The equation is separable because algebraic rearrangements can be performed to write the equation as dy 1  x + 1 =   dx y  x 2  101 .

agreeing with the value computed by =  . y > 0 in explicit form. confirming that the slope of the function at x = 4 is approximately dy 1  x + 1 0. use the rule for dx dy is given when evaluating it on the check to see if the same value for dx graph screen.207 295 38. choose a value for x (x = 4.4 Differential Equations x +1 1 dy = p (x ) q ( y ) with p (x ) = 2 and q (x ) = . y ). dx y  x 2  Y1= (21n(X)-2/X ) Y2= Y3= Y4= Y5= Y6= CALCULATE 1:value 2:zero 3:minimum 4:maximum 5:intersect 6:dy/dx 7:∫f(x)dx Y1= (21n(X)-2/X) Plot1 Plot2 Plot3 Y1(4)(4) ) 1. The GDC can also draw the tangent line at x = 4 and display its equation.50751077 (1/Ans)((4+1)/42 . dy 1  x + 1 =   is the rule that gives us the slope of the graph of the dx y  x 2  function y (x ) at any point (x . dx With some thinking we can use our GDC to help confirm this result. x > 0. and then y-coordinate. In the GDC screen images below we enter the function y = 2 ln x 2 2 + C . use the GDC to find an approximate value for the corresponding dy to find the slope at that point.2072953681 X = 4 Y = 1.50751077 dy/dx=. for x example).20729538 102 . giving: x +1 y dy = 2 dx x 1 1  ∫ y dy = ∫  x + x 2  dx   which is in the form 1 2 1 y = ln x 2 + C 2 x y = 2 ln x 2 2 +C x This is the general solution of x 2 y dy = x + 1. dx y x We now separate the variables and integrate.

Here is an applied problem involving a separable differential equation.
Example 4

The rate of decay of a substance y at any time t is directly proportional to the amount of y and also directly proportional to the amount of another 1 substance x. The constant of proportionality is 2_ and the value of x at 2 4 any time t is given by x = . (1 + t )2 a) Given the initial conditions that y = 10 when t = 0, find y as an explicit function of t. b) Determine the amount of the substance remaining as t becomes very large.
Solution

a) The rate of decay of substance y is proportional to the product xy, and 1 with the constant of proportionality having a value of 2_ and 2 4 x= , this gives: (1 + t )2 dy 1 4  =2  y dt 2  (1 + t 2 )  1 22 dy = dt y (1 + t )2
Separating variables.

∫ y dy = 22 ∫ (1 + t )
ln y = 2 +C 1+t
2 +C

1

1

2

dt

Integrating both sides.

y = e 1+t y =e e
C

Exponentiating; using e as the base.

2 1+t

2

y = Ae 1 + t

Let A = e c, a convenient form for the arbitrary constant.

Solve for A knowing that initially y = 10 when t = 0: 10 = Ae 1 + 0
2
2

10 = Ae 2

2
22

A = 10e 22
22t

Substituting gives:

y = 10e 22e 1 + t ⇒ y = 10e 1 + t ⇒ y = 10e 1 + t 22t b) As t → ∞, → 22 ; thus, as t → ∞, y → 10e 22 ≈ 1.36 1+t

103

4

Differential Equations

Example 5

Solve the differential equation x dx + e x + y cos y dy = 0.
Solution

As it is the equation cannot be written in the variables separable form dy = p (x ) q ( y ). Since e x + y = e xe y we can make it so by multiplying both dx sides of the equation by e 2x and doing some rearrangement. xe 2x dx + e y cos y dy = 0 ⇒ e y cos y dy = 2xe 2x dx ⇒  1  dy = 2xe 2x  y  e cos y   dx

Separating the variables and integrating both sides gives:

ey (sin y + cos y ) = xe 2x + e 2x + C Using integration by parts on both sides. 2 ey Therefore, the implicit function (sin y + cos y ) = e 2x (x + 1) + C is the 2 general solution. To finish this section we will find an explicit solution by the method of separation of variables for a relatively straightforward first order differential equation, but one whose solution will prove useful in developing another solution method.
Example 6

∫e

y

cos y dy = 2∫ xe 2x dx

Find the general solution to the differential equation
Solution

dy = 22xy . dx

1 dy = 22x dx y 1 ∫ y dy = 2∫ 2x dx ln y = 2x 2 + C1 e ln y = e 2x
2

Separating variables; note loss of solution where y = 0. Integrating both sides.

+ C1

Exponentiate both sides to solve for y.

y = e C1e 2x y = C 2e 2x

2

2

c c e 1 is a positive constant; let e 1 = C2 and C2 > 0.

If y > 0, then y = y and the solution becomes y = C 2e 2x . If y < 0, then y = 2y and the solution becomes y = 2C 2e 2x .
104
2 2

We can include both of these solutions, and also the ‘lost’ solution y = 0, by giving the general solution as y = Ce 2x
2

with no restrictions on the constant C.
2x It is helpful for our review to recognize that the explicit solution y = Ce for Example 6 defines a ‘family’ of curves in the xy-plane. Some of these curves, with the corresponding value of C, have been graphed in Figure 4.9. In order to determine a specific curve from this ‘family’ we must impose an initial condition on the solution.
2

y

Figure 4.9

2

1

C=2 C=1
2 1 1 2 x

C = 1
1

C = 2

2

4.4

First order linear differential equations – use of integrating factor

As mentioned previously, a first order differential equation is called such dy appears in the equation. A differential because the first derivative dx dy equation is called linear when both and y appear only to the first dx power. The standard form for a first order linear differential equation is dy + P (x ) y = Q (x ) . dx We wish to develop a method to solve first order linear differential equations of this form (which could also be written as y ′ + P (x ) y = Q (x )).

105

we wish to find a general solution to the more general first dy + P (x ) y = Q (x ) where Q (x ) is not order linear differential equation dx necessarily zero. we get 2 P (x ) dx y = Ce ∫ which is a general solution for the linear differential equation dy + P (x ) y = 0. giving 2 P (x ) dx  P (x ) dx y =e ∫ e∫ Q (x ) dx + C  . we can substitute dx P (x ) dx  dy  e∫  + P (x ) y  . We can now solve for y.4 Differential Equations We start by considering a simple case when Q (x ) = 0. dx This equation is variables separable. y dx This equation can be integrated in the same way as in Example 6 to give ln y = 2∫ P (x ) dx + C1 and following the same steps as in Example 6. ∫    106 . if we multiply both sides of dx P (x ) dx (called an integrating factor). dx However. = e∫ dx  dy  ∫ P (x) dxQ (x ) . yielding  dx  P (x ) dx P (x ) dx d ye ∫ Q (x ) . we get e∫ e∫ P (x ) dx P (x ) dx d ye ∫ for From the working above. so dy + P (x ) y = Q (x ) becomes dx dy + P (x ) y = 0.  + P (x ) y  = e  dx  ( ) ( ) Integrating both sides gives ye ∫ P (x ) dx = ∫e∫ P (x ) dx Q (x ) dx + C .  dx  dy + P (x ) y = Q (x ) by the factor Thus. By applying the product rule and implicit differentiation we observe that P (x ) dx P (x ) dx dy P (x) dx d ye ∫ = e∫ + yP (x ) e ∫ dx dx ( ) = e∫ P (x ) dx  dy   + P (x ) y  . giving us 1 dy = 2P (x ) .

2  2   y  = 2 (x ) Multiply both sides by integrating factor. integral of this expression is ye ∫ When computing the . Although the expression for the general solution given above looks quite complicated. dy + P (x ) y = Q (x ). x dy 2 y x 2 2 = x dx x x dy  2  2  y = x dx  x  Divide both sides by x to get equation into standard form. Step 1: Make sure the equation is in the standard form dx P (x ) dx by finding P (x ) dx . P (x ) d x ( ) Step 5: Obtain an explicit solution for y by dividing both sides by the P (x ) dx integrating factor e ∫ .Solution to first order linear differential equations Given a first order linear differential equation in the form dy + P (x ) y = Q (x ) dx the general solution is 2 P (x ) d x ∫ P (x ) d x Q (x ) d x + Ce2∫ P (x ) d x y=e ∫ ∫e where C is an arbitrary constant. Step 2: Compute the integrating factor e ∫ ∫ Step 3: Multiply both sides of the equation by the integrating factor. x dx  x   x 1 dy 2 y 1 2 = x 2 dx x 3 x 107 . Step 4: Integrate both sides of the equation. Standard form 2 dy + P (x ) y = Q (x ) . Example 7 Find the general solution of x Solution dy 2 2 y = x 2. integrating factor e ∫ it is standard practice to omit the constant of integration from the indefinite integral of P (x ) . x dx 2 2 dx 2 1 1 2. Integrating factor: e ∫ x = e 22 ln x = e 2ln x = ln x 2 = 2 x e 1  dy  2   1 3. the basic steps for solving a first order linear differential equation by means of an integrating factor are relatively simple. Let’s illustrate the five basic solution steps with an example. The left side will be e∫ P (x ) dx P (x ) dx d  dy  ye ∫ and the  + P (x ) y  which is equivalent to  dx  dx P (x ) dx . P (x ) = 2 and Q (x ) = x . dx 1.

Therefore. 2 + = x + 1 dx x 2 + 1 x2 + 1 1 2 2x dy  x  + 2 y = dx  x + 1 x2 + 1 2. 4. Example 8 Find the particular solution of (x 2 + 1) dy + xy = (1 2 2x ) x 2 + 1 dx given that y = 2 when x = 1.   differentiation. we substitute y = 2 and x = 1. Integrating factor: Standard form with P (x ) = 12 2x x . Divide both sides by integrating factor. y = x 2 ln x + Cx 2 is the general solution. Solution x 2 + 1 dy xy (1 2 2x ) x 2 + 1 1. x2 + 1 2 x 1 dx = ln (x 2 + 1) = ln x 2 + 1 ⇒ e ln 2 +1 x 2 + 1 (1 2 2x ) x2 + 1 x2 + 1 = x2 + 1 dy  x  + x2 + 1 2  y =  x + 1 dx dy  x  + y = 1 2 2x   dx  x 2 + 1  dy  x   + y dx =    dx  x 2 + 1   Multiply both sides by integrating factor. x2 + 1 x2 + 1 ∫ P (x ) dx = ∫ x 3. x2 + 1 4.4 Differential Equations It is appropriate to call the differential equation dy + P (x ) y = Q (x ) linear dx dy because = 2P (x ) y + Q (x ) dx is a linear function of y. ∫   x2 + 1 ∫ (1 2 2x ) dx Integrate both sides. ∫x   1 dy 2 y  2  dx = 2 dx x 3  ∫ x dx d dx 1 Integrate both sides with respect to x. To solve for C. Q (x ) = . 2= Therefore. the particular solution is y = 108 . 5.  1 y  2  = ln x + C x  1 dy 2 y   1  by product rule and implicit    y  x 2   = x 2 dx 2 x 3 . y = 2x 2 + x + C x2 + 1 21 + 1 + C 1+1 ⇒ C=2 2 2x 2 + x + 2 2 x2 + 1 . y x2 + 1 = x 2x2 + C 5.

found to be 2 X=-1 dy/dx=2.2) dy = x 2 y .5857864376 2. 2 2 2 ) . with a bit of effort. 2 2 2) is on the solution curve. Enter the solution curve for Y1 and enter the differential equation in the form 1 2 2x dy  x  for Y2. Turn Y2 ‘off ’ = 2 2  y +  x + 1 dx x2 + 1 (un-highlight) so that it is not graphed. Choose a value for x that is in the graph window – say. and evaluate the corresponding y-value for a point on the solution curve. we displayed a slope field (Figure 4. Check equal to 2 2 2 . but it cannot be dy = p (x ) q ( y ). from the differential window. 2 2 2 ) . 2 2 2 . The dx dx integrating factor is e ∫ = e x . only the solution curve is graphed. we can add some confidence to our result for Example 8 by using our GDC to graph the solution curve and then check to see if the original differential equation accurately describes its behaviour (shape). 2 2(21) 2 1 + 2 2 22 + 2 2 y= = = 2 2 2 . use Y to find the value of dx that this value for the slope of the curve at (21.414 213 562.4142134 approximately 2. and multiplying through by this gives dy ex + ex y = exx dx 109 . agrees with the value found on the graph dy at 21. point 2 (21)2 + 1 Y1=(-X2+X+2 (2) )/ (X2+1)2 Y2=(-X/X +1))Y +(1-2X)/ (X2+1) Plot1 Plot2 Plot3 Y3= Y4= Y5= 2. After setting y dy at (21. Both methods of finding dx equation and from the solution to the differential equation. x = 21. Solution The equation first appears that it may be separable. It is a first order linear differential expressed in the form dx dy + y = x which puts it into the equation because it can be rearranged to dx dy standard form + P (x ) y = Q (x ) such that P (x ) = 1 and Q (x ) = x . give the same value. Find the general solution to this for the differential equation dx equation. ( ) Example 9 In the earlier section on slope fields.414213562 1:value 2:zero 3:minimum 4:maximum 5:intersect 6:dy/dx 7:∫f(x)dx Y1=(-X2+X+2 (2))/ (X2+1) (21.Once again. thus supporting our particular solution to the differential equation.(2)→Y4) Y2(-1) CALCULATE .

demand fluency with a range of integration techniques. Figure 4. such as those for separable equations and first order linear equations. and differentiation – as the next example nicely illustrates. Example 10 dy 4 Find the particular solution to (1 + sin x ) 2 y cos x = (1 + sin x ) given dx y (0) = 1. the general solution is y = x 2 1 + Ce 2x . the equation becomes dy  cos x  3 2  y = (1 + sin x ) . dx  1 + sin x  21  dx  The integrating factor is e ∫ 1 + sin x = e 2ln(1 + sin x) = e ln (1 + sin x)  = 2cos x 1 .2 for dx of three different solution curves generated from the general solution for 1 and 2 4 .4 Differential Equations and continuing with the steps for solving a first order linear differential equation yields dy  + e x y  dx = ∫ e x x dx  dx x x x ye = e x 2 e + C ∫ e   x Using integration by parts on the right side. Solution Dividing through by 1 + sin x. 10 y C = 1/10 C=1 Figure 4.10 shows the same dy = x 2 y along with the graphs slope field displayed in Figure 4. 1 + sin x 110 . Thus.10 6 5 4 3 2 1 6 5 4 3 2 1 0 1 2 3 4 5 6 C = 4 1 2 3 4 5 6 x An analytic method for solving differential equations. C = 1.

y ) = 2x 2 + xy 2 5 y 2 111 . Homogeneous first order differential equations dy = F (x .Multiplying both sides by the integrating factor gives 1 dy  cos x  2 2 y = (1 + sin x ) . 4 4. expressed as a function of the ratio  x dx x The function F can be written as a function of y if it can be expressed as x a quotient of two homogeneous functions of the same degree. a two-variable function is homogeneous of degree n if the sum of the powers of x and y in each term is n. that is. In general. left side is equal to the derivative of 1 + sin x d  y  2   = sin x + 2 sin x + 1 dx  1 + sin x  y and We now integrate both sides. The integral of the left is simply 1 + sin x integrating each term on the right is straightforward except for sin 2 x . 2 2 d  y  1 1  ∫ dx  1 + sin x  dx = ∫  2 2 2 cos 2x + 2 sin x + 1 dx     y x 1 = 2 sin 2x 2 2 cos x + x + C 1 + sin x 2 4  3x 1  y = (1 + sin x )  2 sin 2x 2 2 cos x + C   2  4 Given y (0) = 1. One situation where this will always work is when the first order differential equation is homogeneous. For example: g (x . y ) is homogeneous if the right side can be The differential equation dx dy y  y = F  . We need to take the double-angle identity cos 2x = 1 2 2 sin 2 x and rearrange 1 1 it to give us sin 2 x = 2 cos 2x .5 Homogeneous differential equations When a first order differential equation is not separable nor linear. informs us that the y . the particular solution is 1 y = (1 + sin x ) (6x 2 sin 2x 2 8 cos x + 12) . alone. it follows that 1 = (1 + 0) (0 2 0 2 2 + C ) ⇒ C = 3 Therefore. 2 1 + sin x dx  (1 + sin x )  Our experience of differentiating functions and familiarity with the solution pattern for first order differential equations. it may still be possible to transform it by an appropriate substitution into an equation that we can solve analytically.

The right side can be expressed in terms of x 2 dividing numerator and denominator by x . The function m (x . where M and N are homogeneous functions of the same degree. y ) then the equation is a homogeneous differential equation. N (x . dx 112 . =  x dx N (x . x 3 + x 2 y 2 xy 2 . As we  x will see. dy 3 y 3 2 xy 2 = 3 is a homogeneous differential equation because dx x + x 2 y 2 xy 2 both the numerator. this substitution transforms the differential equation into a separable equation for which we have a solution method. and h (x . 6xy . Two examples are given below. 3 y 3 2 xy 2 . y ) = 3 y 3 2 xy 2 is homogeneous of degree 3. if we solve for dx M (x . y ) and N (x . y ) y dy  y = F   . are homogeneous y by functions of degree 2. and the denominator. dy 6xy is a homogeneous differential equation because both = 2 dx x 2 y 2 the numerator. by can be written as a function of . where n is the degree of M and N. 1. y ) x dx dividing through both M (x . y ) = 4 x 2 y 2 2 x 3 y 2 is not homogeneous. We divide numerator and denominator by x 3 to get  y y 3 y 3 xy 2 3  2   2 3 3  x  x dy x x = = 2 dx 4 x 3 x 2 y 2xy 2 y  y + 3 2 3 4 + 2 2  x3 x x  x x Once a homogeneous differential equation is written in the form dy  y = F   it can be solved analytically by making the substitution  x dx y  y = vx  or v =  where v is a differentiable function of x. x 2 2 y 2 .  y 6xy 6  2  x dy = 2x 2 = 2 dx x 2 y  y 12  2 x  x 2. The equation dy M (x . Example 11 3 2 Find the particular solution for xy 2 dy = x 3 + y 3 given y = 3 when x = 1. i. y ) .4 Differential Equations is homogeneous of degree 2. y ) by x n . are homogeneous functions of degree 3.e. and the denominator. dy and get it to be equal to a quotient in the form Thus.

the equation is in the form 113 y 3 = x 3 (3 ln x + 27) 1 . or rearrange it so that. Using Example 11 as a guide we can outline the basic steps for solving a first order homogeneous differential equation. the particular solution is y = x (3 ln x + 27) 3 . Substituting v for and v + x for produces x dx dx 3 dv 1 + v v+x = dx v2 which is a separable equation for the variables x and v because it can be dy = p (x ) q (v ). and substituting gives x 1 9 = ln 1 + C ⇒ C = 9 . 1 3 v = ln x + 9 ⇒ v 3 = 3 ln x + 27 3 y back in for v gives: Substituting x 3  y   = 3 ln x + 27  x Therefore. x 3 x3 y3  y + 3 1+   dy x 3 x x = = 2 dx xy 2  y  x x3 dv dy =v+x by means of the We now let y = vx which means that dx dx dv y dy product rule. Thus. dy x 3 + y 3 = dx xy 2 Dividing both numerator and denominator by x 3 expresses the derivative y in terms of . as shown below: written in the form dx 1 dv 1 + v 3 dv 1 + v 3 v 3 dv dv  1   1  = = = 2 ⇒ 2v ⇒ x 2 2 ⇒ x =    2 2 dx dx v dx  x   v 2  dx v v v Separating the variables and integrating: x v 2dv = 1 dx x 1 2 ∫ v dv = ∫ x dx 1 3 v = ln x + C 3 If y = 3 when x = 1. Step 1: Confirm that.Solution Dividing both sides by xy 2 reveals that the differential equation is homogeneous because both numerator and denominator on the right side are homogeneous functions of degree 3. then v = y 3 = = 3.

Step 4: By applying the technique of separation of variables. so that the equation is in the form  x dx dv dy =v+x and Step 3: Let y = vx from which it follows that dx dx y dv dy substitute v for and v + x for transforming the x dx dx equation into a separable equation in terms of v and x. dy x + y where x > 0. a) Use the substitution y = vx to show that x dx 1 2 v b) Hence. M and N are homogeneous of degree 1. then = v + x . y ) case. y > 0 . gives: v+x dx dx v+x x x 1+v dv = 12 v dx v (1 2 v ) dv 1+v = 2 dx 12v 12v 1 + v2 dv = 12v dx Q. 114 . Consider the differential equation Solution a) 1. where n is the degree of dy  y = F  . y ) . Letting y = vx . M and N. find the general solution of the differential equation. y ) same degree. 2. y ) by x n . where M and N are homogeneous functions of the = dx N (x . Substituting v for and dx dx x dv dy for .4 Differential Equations dy M (x .E. find a solution in terms of v and x. y ). Step 2: Divide both M (x . y ) where. y Step 5: Substitute back in for v and write the solution in terms of y x and x. = dx x 2 y dv 1 + v 2 = . Divide numerator and denominator by x. in this = dx N (x . y x y + 1+ dy x = x x = y x y dx 2 12 x x x dv dy y 3.D. Example 12 Do not forget to perform Step y 5 – substituting back in for x v – because you must express your final solution in terms of y and x. giving your answer in the form C = f (x . The equation is already in the form dy M (x . y ) and N (x .

yields 1 2v 1 ∫ 1 + v 2 dv = ∫ x dx .b) 4. Separating the variables and integrating. Example 13 d  1 ln x + 1 + x 2  = .  x 2  x  Solving for C: 1   y2 2  y   arctan   2 ln 1 + 2  + ln x  = C   x   x       y2   y arctan   2 ln  x 1 + 2  = C  x x    y C = arctan   2 ln x 2 + y 2  x This is the general solution such that y is an implicit function of x. [Hint: Use the result from a) to integrate the separable y= 4 equation that is in terms of v and x.] a) Show that Solution ( ) a) 1 d  ln x + 1 + x 2  =  x + 1 + x2 dx  = = = 1 ( ) 1  d  x + (1 + x 2 ) 2   dx     1 2  1  1 + (1 + x 2 ) 2 (2x ) 2  2  x + 1+ x  x   1 +  x + 1+ x  1 + x2  1 2  1 + x2 x  +   1 + x2  x + 1 + x2  1 + x2   1 115 . To integrate the left side we split up the fraction: v 1 1 ∫ 1 + v 2 dv 2 ∫ 1 + v 2 dv = ∫ x dx 1 arctan v 2 ln (1 + v 2 ) = ln x + C 2 y back in for v gives 5. Substituting x y2  y 1  arctan   2 ln 1 + 2  = ln x + C .  dx  1 + x2 b) Show that the solution curve that satisfies the differential equation dy = y + x 2 + y 2 with initial conditions y (0) = 21 is the parabola x dx x2 2 1.

form  x dx dy = dx y + x x2 + y2 x2 x x = y  y + 1+    x x 2 Letting y = vx . The term dx x 2 2 x + y has a degree of 1. b) First.4 Differential Equations = =  x + 1 + x2    x + 1 + x2  1 + x2  1 1 1 + x2 Q. then v+x dv dy for . Therefore. so both numerator and denominator are homogeneous functions of degree 1.D. ) = e ln x + lnC 2 Exponentiating both sides using base of e. v + 1 + v 2 = Cx 1 + v 2 = C 2 x 2 2 2Cxv + v 2 1 = C 2 x 2 2 2Cxv  y 1 = C 2 x 2 2 2Cx    x 2Cy = C 2 x 2 2 1 1 1 y = Cx 2 2 2 2C 116 ( 1+v ) 2 2 = (Cx 2 v ) Substituting y back in for v. x . gives: dx dx dv dy y =v+x . divide both sides by x to give dy y + x 2 + y 2 = .E. Now divide numerator and denominator on right side by x in order to write the equation in the dy  y = F  . Using this result gives: v+x ( ) ( ) ln v + 1 + v 2 = ln x + ln C e ln v + 1 + v 2 ( ( ) Setting arbitrary constant to ln C. From part a) we know that   dx 1 + x2 1 2 ∫ 1 + v 2 dv = ln v + 1 + v + C . Substituting v for and dx dx x dv dv = v + 1 + v2 ⇒ x = 1 + v2 dx dx Separating the variables and integrating: 1 1 ∫ 1 + v 2 dv = ∫ x dx d  1 ln x + 1 + x 2  = .

1  1 1 y =   x2 2 1 . dx Consider the initial-value problem dy = x + y . the particular solution curve is the parabola y = 1 2 x 2 1. 4 2 0 2 4 x 2 117 . Also. especially for an equation that we are not able to solve analytically. 0).11 shows the slope field for the differential equation dx An approximation to the particular solution can be sketched by drawing a smooth curve through the point (2. earlier in this chapter we saw how a slope field is an effective graphical method that provides a rough idea about the behaviour of solutions to a differential equation.6 Euler’s method We have established three analytic methods for solving different types of first order differential equations: separable equations. linear equations (integrating factor) and homogeneous equations (substitution y = vx). Several of the examples in this chapter have found particular solutions to what is referred to as an initial-value problem that is stated in the form dy = F (x . y ) . 4 4. 0) that follows the slopes in the slope field. dx dy = x + y. To roughly sketch a particular solution to a differential equation using a slope field. y 4 2 Figure 4. y (x 0 ) = y 0 . Figure 4. we need to know a point (initial condition) that the solution curve passes through in order to have a ‘starting point’ from which to sketch a curve that will be parallel to the short line segments drawn at representative points that indicate the slope of any solution.11 Slope field for dy = x + y and sketch of solution dx passing through (2.Solve for C given the initial condition y(0) = 21: 1 1 ⇒ C= 21 = 0 2 2C 2 Hence. y (2) = 0. as shown in the figure. 2  2 2(2) Therefore.

12). i. This approximation clearly becomes less accurate as we move away from the point of tangency (2. 0). for the interval x 0  x  x 0 + h . y (2) = 0 dx We know from the differential equation that the slope of the solution dy = x + y = 2 + 0 = 2. In general. y 0 ). y (x 0 ) = y 0 we choose a step size h. To approximate a value of y for a specific value of x. Starting at the point dx (x 0 . We can use this tangent line as a rough approximation to the solution curve (see Figure 4. the line with slope F (x 0 . 0) has the equation: y 2 0 = 2 (x 2 2) ⇒ y = 2x 2 4.4 Differential Equations Let y(x) represent the solution curve. curve is 2 at the point (2. 0) because dx the line tangent to the solution curve at (2. This takes us 118 . after being presented with an initial value problem: dy = F (x . we could make an educated guess from the sketch of y made with the aid of the slope field. namely: dy = x + y .12 4 3 2 1 0 1 2 x line tangent at (2. each having the slope of the solution curve at its starting point. In this way we build an approximation to the curve by attaching little line segments together. Let’s illustrate the method with the initial-value problem that we have just been considering. Hence. 0): y = 2x  4 2 1 1 2 3 4 Euler’s method improves this approximation by moving a short horizontal distance (the step size h) along this tangent line and then change direction according to the slope field. Euler’s method uses the basic idea behind the construction of slope fields to find numerical approximations to solutions of differential equations. after the prolific eighteenth-century mathematician who first devised this computational method to help him calculate the orbit of our Moon. y Figure 4. But if we want a more accurate approximation then we need to use a more refined method. we approximate the solution curve with the tangent line.e. y 0 ). The simplest numerical method is called Euler’s method. y ) . for example y when x = 3.

y 0 ) Now we are at the starting point of the second line segment (x1 . y1 ). y 0 ) on the solution curve and moves along a segment with slope F (x 0 . etc. y1) h F(x0. The process is repeated with the new point. The accuracy of the estimates depends on the choice of the step size h and the overall number of steps (iterations). y0) (x0.13): x1 = x 0 + h. y 0 ). y1 ). y 0 ) to define a new point (x1. (x1 . solution curve (x5. y1 ) .as far as the point (x1 . y0) 119 . (x 2 . y1) Euler approximation (x0. y0) h x Figure 4. whose coordinates are calculated as follows (see Figure 4. y4) Figure 4. Decreasing the step size while increasing the number of steps leads to increasingly more accurate estimates for solution values. y slope = F(x0. y5) (x4. y1 ). y3) (x2. with the next line segment having slope F (x1 .14 Further iterations of Euler’s method build an approximation to the solution curve. y 0 ). Repeating this process we get an approximation to the solution curve consisting of line segments joining the points (x 0 . y1 ).13 Euler’s method starts at (x 0 . (x3. Each computed value y n is an estimate of the corresponding ‘true solution’ y at x = xn. y1) such that x1 = x 0 + h and y1 = y 0 + hF (x 0 . y2) (x1. y0) (x1. We repeat the process. y 2 ). y1 = y 0 + h F (x 0 . y 2 ) on the Euler approximation where x 2 = x1 + h and y 2 = y1 + h F (x1 . This takes us to the next point (x 2 .

yn) on the approximate solution curve (Euler approximation) are: xn + 1 = xn + h.584 + 0. We are given that h = 0.4 y 2 = y1 + h F (x1 . Example 14 dy = x + y such that y (2) = 0. Using the appropriate formulae for xn and y n and iterating five times.4 + 0.2 + 0.2 = 2.15).92 + 0.464 96 120 .4208) = 3.2 = 2. … .2 (2. and N is the total number of steps (iterations).4 Differential Equations Euler’s numerical method dy = F (x. the step size.584 x 4 = x 3 + h = 2.2 = 2.2 (2.6 y 3 = y 2 + h F (x 2 .2 (2.4) = 0.6 + 0.92) = 1. giving your answer to two decimal places.8 y 4 = y 3 + h F (x 3 . y ) with the initial condition y(x0) = y0. y ) = x + y . use Euler’s dx method with a step value of 0.4 + 0. y 4 ) = 2.6 2. y 2 ) = 0.4208 x 5 = x 4 + h = 2.2 y1 = y 0 + h F (x 0 .8 + 0. we have: x1 = x 0 + h = 2 + 0.8 3 x We use Euler’s method to build an approximation to the ‘true’ solution curve starting at x = 2 and finishing at x = 3 by piecing together five short segments (Figure 4. y 0 ) = 0 + 0. Let’s now apply Euler’s method to answer a question posed earlier for the initial-value problem presented at the start of this section.2 2. yn + 1) from the known nth point (xn. N where h.15 4 3 2 1 0 solution curve Euler approximation 2 2. the For the differential equation dx recursive formulae for generating the coordinates of the unknown (n + 1)st point (xn + 1. 2.6 + 1. yn + 1 = yn + hF (xn.2 . yn) for n = 1.8 + 2. is a constant. y 3 ) = 1.4 + 0.2 + 0. x 0 = 2. 3.4 x 2 = x1 + h = 2.2 (2.92 x 3 = x 2 + h = 2. y1 ) = 0.2 to find an approximate value of y when x = 3.4208 + 0. For the differential equation Solution y Figure 4. y 0 = 0 and F (x .2 = 2.2 (2 + 0) = 0.4 2.584) = 2.2 = 3 y 5 = y 4 + h F (x 4 .

To three significant figures.92 1.4 0. if we used a step size of h = 0.15.25 y1 = y 0 + x0 + 1 0 +1 9 = = 1. yn) 0. Thus.25 and F (x .584 2.25 = 0. it is often sufficient to simply display relevant results for each iteration in a table.2 2. Using a program on our GDC or a spreadsheet.0 yn + 1 0. we could easily decrease the step size (and increasing the number of steps) in order to improve the accuracy of the approximation. y 0 = 1.4 0.8 yn 0 0.584 2.25 + 0. For example. y ) = xy + 2 recursive formula for y n is: y n + 1 = y n + h F (x . Because we will perform most of the calculations for each iteration on our GDC.8 3.6%. A numerical method like Euler’s is especially useful when applied to a differential equation that cannot be solved by any known analytic methods. n 0 1 2 3 4 xn 2 2.4208 3. use Euler’s method with = dx xy + 2 step size h = 0.01 (requiring 100 iterations) we would get an estimate of 4. Example 15 Given that dy x +1 and y = 1 when x = 0. the ‘true’ value of y(3) is approximately 4. Give the approximation to three significant figures. reducing the error to less than 1%.f.46 has an error of approximately 16.4 0.25 to approximate the value of y when x = 1.5 121 . as shown below. Thus the We have that x 0 = 0.92 1. Solution x +1 .46 when x = 3.4 2. our approximation of 3.4208 hF(xn . as we will do in the next example.).2 2.This process leads to an approximate (three decimal places) value of y ≈ 3. h = 0.6 2.6 2.25 = 0.11 (3 s.464 96 The first order differential equation in Example 14 is linear and hence can be solved by means of an integrating factor.8368 1. Given y(2) = 0 the particular solution is y = 3ex 2 2 2 x 2 1.25) n = 0: xn + 1 xn + 1 ⇒ yn + 1 = yn + xn yn + 2 4xn yn + 8 x1 = x 0 + h = 0 + 0.52 0.044 16 xn + 1 2.4 2.664 0.125 = 1+ 4(0)(1) + 8 8 4x 0 y 0 + 8 n = 1: x 2 = x1 + h = 0. y ) = y n + (0.

4 Differential Equations y 2 = y1 + n = 2: x1 + 1 0.1 1 Match each slope field with its differential equation.25 = 1 y4 = y3 + Therefore.261986 + ≈ 1.125 + 4(0. a dy 2 x = y dx dy = tan y dx b e h dy y 2 = dx x 2 dy = xy dx ln y dy =1 dx c f x2 dy = y2 2 y dx dy x = dx y d x g x2 + 1 dy y 2 2 1 = dx ex 122 .75 + 0.55.261986) + 8 4x 2 y 2 + 8 x3 + 1 0.404518 4(0. the approximate value of y when x = 1 is y ≈ 1. Exercise 4.25 + 1 ≈ 1. y y a b 1 1 0 1 x 0 1 x c y d y 1 1 0 1 x 0 1 x i iii dy = 22 y dx dy y =2 x dx ii iv dy = x2 2 y dx dy = 2x + y dx 2 Solve the following separable differential equations.404518 + ≈ 1.5 + 0.5 + 1 = 1.404518) + 8 4x 3 y 3 + 8 x 3 = x 2 + h = 0.75 + 1 = 1.25 = 0.75 y3 = y2 + n = 3: x 4 = x 3 + h = 0.75)(1.547 801 4(0.25)(1.125) + 8 4 x1 y1 + 8 x2 + 1 0.5)(1. listed below.261986 = 1.

(a) (a) 8 Match each slope field with its differential equation. what is lim p (t ) ? t →∞ t →∞ d Comment on the long-term behaviour of the species’ population growth. y (0) = 21 π . show that an implicit solution for is ye y = Axe x where A is an arbitrary the differential equation = dx xy + x constant. dt a Sketch the slope field.3 Using the method of separation of variables. (a) (a) a (b) b (b) (b) (b) (c) (c) (c) c (c) (d) d (d) (d) (d) i iii dy 5 = dx x + y dy xy =2 10 dx ii iv dy = dx 5 x2y dy xy = dx 10 123 . 4 a Show that the implicit solution can be expressed as arctan y + arctan x = b Use the formula for tan(A + B) to find the explicit solution. in explicit form. Comment on the possible values of the constant C. 6 Solve the initial-value problem: dy 2x + sec 2 x = .5. to the differential equation dy = cos x. p. of a certain species is given by dp = 5p 2 2 p2 . dx dy xy + y y 5 The equation for the rate of change of the population (in thousands). 4 Find the general solution. listed below. p(0) = 4). lim p (t ) )? c If p(0) = 0. then what appears to be the limiting value of the population (that is. b If the initial population is 4000 (that is. 2y dx dy (1 + x ) dx + 1 + y 2 2 y (0) = 25 7 Consider the initial-value problem: = 0.

4

Differential Equations

9 Solve the initial-value problem:
dy (1 + x ) dx = 1 + y
2 2

,

y (2) = 3

Write the solution in explicit form, expressing y in terms of x. as the sum of two 10 a Use the method of partial fractions to express 2 x 2x 22 fractions. b Consider the differential equation
dy y2 , x > 2 such that y = 1 = 2 dx x 2 x 2 2
32 3y

1

when x = 5. Show that the solution is 2e 11 Consider the differential equation (12 x 2 ) variables.

y

=

x +1 . x 22

dy + 2xy = 2x. dx a Find the general solution in the form y = f (x ) by the method of separation of

b Write the differential equation in the standard form for a first order linear differential equation, means of an integrating factor. a c e
dy  2  +   y = 6x 3 dx  x  dy y 2 = x3 dx x
3 dy 2 3x 2 y = e x dx

dy + yP (x ) = Q (x ), and find the general solution by dx

12 Solve each of the following first order linear differential equations. b d f
dy 2 xy = x dx dy + y sin x = e cos x dx

x

dy =x+y dx

13 Solve the first order linear differential equation
tan x dy + y = sec x giving your answer in the form y = f(x). dx

14 Consider the initial-value problem:
dy xy 2 = 1, y (0) = 1 dx 12 x 2

a Show that the differential equation is a first order linear equation by writing it in the form b Show that the integrating factor is 12 x 2 . c By using the substitution x = sin u, show that
dy + yP (x ) = Q (x ) . dx

12 x 2 d x =

x 12 x 2
2

+

arcsin x + C. 2

d Find the solution to the initial-value problem expressed in the form y = f(x). 15 a Show that ∫ tan x d x = 2ln cos x . b Show that c
dy = 1 + y tan x is a first order linear differential equation. dx dy π π = 1 + y tan x , 2 < x < . Find the general solution of 2 2 dx

124

16 Find the particular solution to the differential equation = dx that y = 1 when x = 1.
dy 2 x 3 + xy = 0. dx

dy

x 2 ln x 2 y given x

17 Find the general solution, in explicit form, to the differential equation

x2

18 Find the general solution to the first order homogenous differential equation
dy 3 y 2 x . = dx 3x 2 y

Write the answer in the form C = f (x , y ). 19 Solve each of the following first order homogeneous differential equations. a c e
dy y = dx x + 1 dy x + 2 y = x dx 2x 2 + y 2 dy =2 2xy + 3 y 2 dx

b d f

x

dy = 2x + 3 y dx dy = x2 2 y2 dx

xy

x (y 2 x )

dy = y (x + y ) dx

20 Consider the differential equation

dy x + 2y = , for x > 0 . dx 3 y 2 2 x dv 1 + 2v = . a Use the substitution y = vx to show that v + x dx 3v 2 2

b Hence, find the solution of the differential equation, given that y = 0 when x = 1.

21 Use the substitution y = vx to show that the general solution to the differential equation

y 2 2 x 2 + xy

dy = 0 is 2x 2 y 2 2 x 4 = C , where C is a constant. dx

22 Consider the initial-value problem:

x

dy = y + x2 2 y2 , dx

y (1) = 1

a Use the substitution y = vx to show that v + x

dv = v + 12 v 2 . dx π  y b Hence, show that the solution is arcsin   = ln x + .  x 2 dy y 2 + y . = x dx

23 Consider the differential equation a Find the general solution.

b Given that y = 1 when x = 1, find a particular solution solved explicitly for y. c Use Euler’s method with step size h = 0.2 to approximate the solution at

x = 1.2, 1.4 , 1.6 and 1.8. d Compute the percentage error for each of the approximate solutions found in c compared to the solution for the same value of x found using the explicit solution found in b.

125

4

Differential Equations

24 Given that

approximate the value of y at x = 1.

dy = xy 2 and y = 1 at x = 0, use Euler’s method with 5 steps to dx

25 Use Euler’s method with step size h = 0.1 to approximate the value of y when

x = 1 for the differential equation
through the point (0, 1) .

dy xy = e given that the solution curve passes dx

26 Use the substitution y = vx to find the general solution to the differential equation
dy x 2 + 3 y 2 = . dx 2xy

27 Given that

h = 0.1 to approximate the solution at x = 1.1, 1.2, 1.3, 1.4 and 1.5. 28 Consider the initial-value problem:
dy = x 2 y, dx

dy = x y and y = 4 when x = 1, use Euler’s method with step size dx

y (0) = 0

a Show that the solution is y = e2x + x 21. b Use Euler’s method with 5 steps to find an approximate value of y when x = 1. c Use Euler’s method with 10 steps to find another approximation for y(1). d Compare the approximate values for y(1) found in b and c to the actual value using the solution y = e–x + x 2 1. Comment.

Practice questions

1 Consider the differential equation x a Find the general solution.

dy 2 3y = x 4 . dx

b Given that y = 2 when x = 1, find the particular solution in explicit form. 2 Given that y = 2 when x = 1, solve the following differential equation explicitly for y.

y

dy 2 3x = x 4 dx dy y = , x ≠ 0. dx x 2

3 Find the general solution of the differential equation 4 Solve

dy 1 + y = cos x , x ≠ 0 , giving your answer in the form y = f (x ) . dx x dy y + = x2. 5 Consider the differential equation dx x

a Find the general solution.

b Given that y = 20 when x = 4, find an explicit solution for y in terms of x.

126

6 Match each of the differential equations with its direction field. 9 Solve the differential equation π dy = y tan x + 1. dx 2v y b Hence. (x 2 + y 2) + 2xy dx 127 . y) is given by 2 . 6) lies on this curve. 10 Consider the differential equation dy 3 y 2 + x 2 = . 0 < x < . Find: a the equation of the normal to the curve at P. is x 3 + 3xy 2 = k. 2 dx if y = 1 when x = 0. where k is a constant. find the solution of the differential equation given that y = 2 when x = 1. 11 Use the substitution y = vx to show that the general solution to the differential equation dy = 0. y) on the curve is tan x a dy = y ( y 21. x > 0. the slope of the curve at any point (x. b the equation of the curve where y is expressed in terms of x. 0 and for which  6   2y + 4 . x +x The point P(3. for x > 0.5) dx dy x = dx y b dy = xy dx dy e = x2y dx c 8 For all positive values of x the slope of a curve at the point (x. (1) 3 2 1 (2) 3 2 1 (3) 3 2 1 3 2 1 1 2 3 (4) 3 2 1 1 2 3 3 2 1 1 2 3 (5) 3 2 1 1 2 3 3 2 1 1 2 3 (6) 3 2 1 1 2 3 3 2 1 1 2 3 1 2 3 3  2 1 1 2 3 1 2 3 3 2 1 1 2 3 1 2 3 dy = 2xy dx dy d f = ( y 21) ( y + 2) dx π 7 Find an equation for the curve that passes through the point  . 2xy dx dv 3v 2 + 1 a Use the substitution y = vx to show that v + x = .