MAA 102

Chapter 3: Power Series

SERIES

In the previous chapters, we concentrated on infinite series with constant terms. Of major
importance in applications are series whose terms contain variables. In particular, if x is a
variable, then a series of the form

a x

n=0

= a0 + a1 x + a2 x 2 + + an x n +

is called a power series in x.

The series

a ( x b)
n=0

= a0 + a1 ( x b ) + a2 ( x b ) + + an ( x b ) + is called a power
2

series in x b .

The main objective of this chapter is to determine all values of x for which a power series
converges.

THEOREM 3.1.

If

a x
n=0

is a power series, then precisely one of the following is true.

(i)

The series converges only if x = x0 .

(ii)

The series is absolutely convergent for all x.

(iii) There is a positive number r > 0 such that the series is absolutely convergent if
x < r and divergent if x > r .

3-1

NOTE.

r is called the radius of convergence. From Theorem 3.1,

radius of convergence = 0 .

(ii)

radius of convergence = .

(iii) radius of convergence = r .

require special consideration. To find the interval of convergence for

a x
n=0

, we often

employ the Ratio Test.

EXAMPLE 3.1.

Find the interval of convergence for the following power series.

(a)

(c)

xn

n =0 n !

n !( x 2 )
n=0

(b)

xn

n =1 2n

(d)

n =1

( x 2)

SOLUTION.

(a)

3-2

(b)

xn
If we let an =
, then
2n
lim

an +1
2n
x n +1
n
n
= lim
n = lim x
= x lim
= x
n 2 ( n + 1) x
n
n n + 1
an
( n + 1)

xn
is absolutely convergent if the following equivalent

n =1 2n

By the Ratio Test 2.4,

inequalities are true:

xn

2n = 2n = 2 + 4 + 6 +
n =1

n =1

which is divergent (see Example 2.9 Integral Test). If we let x = 1 , we obtain

( 1) = 1 + 1 1 +
xn
=

2 4 6
n =1 2n
n =1 2n
n

which is conditionally convergent because

while

n =1

( 1)
2n

( 1)
2n

converges (Alternating Test)

xn
diverges (Integral Test). Hence, by Definition 2.4,
is
n =1 2n

conditionally convergent at x = 1 .

Interval of convergence: 1 x < 1 or [-1, 1)

(c)

If we let an = n !( x 2 ) , then
n

( n + 1)!( x 2 )
a
lim n +1 = lim
n
n a
n
n !( x 2 )
n

n +1

By the Ratio Test 2.4,

= lim ( n + 1)( x 2 ) = x 2 lim ( n + 1)

n !( x 2 )

is absolutely convergent if the following

n=0

equivalent inequalities are true:

x 2 = 0 x 2 lim ( n + 1) = 0 < 1
n

By the Ratio Test 2.4,

n !( x 2 )

is absolutely convergent at x = 2 .

n=0

Interval of convergence: x = 2
Radius of convergence = 0

(d)

If we let an

( x 2)
=

, then

( x 2)
a
lim n +1 = lim
n a
n
n +1
n

n +1

( x 2)

By the Ratio Test 2.4,

= lim ( x 2 )
n

( x 2)

n =1

n
n
= x 2 lim
= x 2
n n + 1
n +1

is absolutely convergent if the following

equivalent inequalities are true:

The series diverges if x 2 > 1 , that is if x > 3 or x < 1 . If x 2 = 1 , the Ratio Test
gives no information and hence the numbers 1 and 3 require special consideration.
Substituting 1 for x in the given series, we obtain

n =1

( x 2)

( 1)

n =1

which is conditionally convergent (see (b)). If we let x = 3 , we obtain

n =1

( x 2)
n

1
n =1 n

xn
which is divergent (p-series, Theorem 2.4). Hence,
is absolutely convergent at
n =1 2n

at 1 < x < 3 , conditionally convergent at x = 1 and divergent at x = 3 .

Interval of convergence: 1 x < 3

3-4

3.2

POWER SERIES REPRESENTATIONS OF FUNCTIONS

A power series

a x

n=0

may be used to define a function f whose domain is the interval of

convergence of the series. Specifically, for each x in this interval, we let f ( x ) equal the sum
of the series, that is,
f ( x ) = a0 + a1 x + a2 x 2 + + an x n + .

If a function f is defined in this way, we say that

a x
n=0

is a power series representation

for f(x) (or of f(x)). We also use the phrase f is represented by the power series.

Suppose

THEOREM 3.2.

a x
n=0

(i)

= f ( x ) with x < where > 0 . Then

f is a differentiable function for x < and

d
d

n
a
x
a
an x n
=
+

0
n

dx n =0
n =1
dx

d

,
= ( an x n ) = nan x n 1
n =1
n =1 dx
= a1 + 2a2 x + 3a3 x 2 + + nan x n 1 +

f '( x) =

(ii)

x <

f is a integrable function in any interval in x < and

x
x

f ( t ) dt = ant n dt = ant n
0
0
n =0
n=0

a t n +1
x n +1
= n = an
n +1
n =0 n + 1 0
n=0

x <

1
1
1
= a0 x + a1 x 2 + a2 x 3 + +
an x n +1 +
n +1
2
3

3-5

NOTE.

The series in (i) and (ii) of Theorem 3.2 have the same radius of convergence as

a x
n=0

However, the nature of their convergence may differ at the end points of the interval of
convergence.

EXAMPLE 3.2.

Find the power series representation for f ( x ) =

(1 x )

SOLUTION.

3-6

EXAMPLE 3.3.

Find the power series representation for ln (1 + x ) .

SOLUTION.

3-7

EXAMPLE 3.4.

Find the power series representation for arctan x .

SOLUTION.

3-8

3.3

TAYLOR AND MACLAURIN SERIES

Let f be a function and n a positive integer such that the derivative f ( n ) ( x ) exists for every
x around a. Then f can be defined by the power series representation
f '' ( a )
f ( n) ( a )
2
n
f ( x ) = f ( a ) + f ' ( a )( x a ) +
( x a) + +
( x a) +
n!
2!
n
(
)

f (a)
n
=
( x a)
n!
n =0

This series is called the Taylor series for f around x = a .

If a = 0 , this series is known as the Maclaurin series, that is, f ( x ) =

EXAMPLE 3.5.

f (n) ( 0 ) n
x .
n!

Find the Maclaurin series for f ( x ) = e 2 x . Then evaluate e1 .

SOLUTION.

3-9

EXAMPLE 3.6.

Find the Maclaurin series for f ( x ) = sin x .

SOLUTION.

3-10

EXAMPLE 3.7.

Find the Maclaurin series for f ( x ) = cos x .

SOLUTION.

3-11

SOME IMPORTANT MACLAURIN SERIES

Maclaurin Series

Interval of Convergence

1
= x k = 1 + x + x 2 + x3 +
1 x k =0

1 < x < 1

1
k
=
( 1) x 2 k = 1 x 2 + x 4 x 6 +

2
1+ x
k =0

1 < x < 1

ex =
k =0

xk
x 2 x3 x 4
= 1+ x + + + +
k!
2! 3! 4!

sin x = ( 1)

k =0

x 2 k +1
x3 x5 x 7
= x + +
3! 5! 7!
( 2k + 1)!

x2k
x2 x 4 x6
cos x = ( 1)
= 1 + +
2! 4! 6!
( 2k ) !
k =0
k

tan 1 x = ( 1)
k =0

x 2 k +1
x3 x5 x 7
= x + +
3 5 7
( 2k + 1)

1 x 1

x 2 k +1
x3 x5 x 7
sinh x =
= x+ + + +
3! 5! 7!
k = 0 ( 2k + 1) !

x2k
x2 x 4 x6
= 1+ + + +
2! 4! 6!
k = 0 ( 2k ) !

cosh x =
m

= 1+
k =1

< x < +
1 < x 1

(1 + x )

< x < +

x k +1
x 2 x3 x 4
= x + +
k +1
2 3 4

ln (1 + x ) = ( 1)
k =0

< x < +

m ( m 1)

( m k + 1) x k

k!

< x < +
< x < +
1 < x < 1*
(m 0, 1, 2,)

*The behaviour at the endpoints depends on m: For m > 0 the series converges absolutely at both
endpoints; for m 1 the series diverges at both endpoints; and for 1 < m < 0 the series converges
conditionally at x = 1 and diverges at x = 1 .

3-12

EXAMPLE 3.8.

Find the Taylor series for f ( x ) = e x around a = 1 .

SOLUTION.

3-13

TAYLOR POLYNOMIAL

Polynomial functions are the best among all functions in the sense that they are easy to
calculate. The sum of the first n + 1 terms of the Taylor series is often denoted by Pn ( x ) and
called the nth-degree Taylor polynomial of f at a, that is,
f '(a)
f '' ( a )
f (n) ( a )
2
n
Pn ( x ) = f ( a ) +
( x a) +
( x a) ++
( x a)
1!
2!
n!
(k )
n
f (a)
k
=
( x a) .
k!
k =0

The following expression is called the remainder after n + 1 terms:

Thus we may write

It follows that if the numerical value of Rn ( x ) is small, then f ( x ) may be approximated by

Since f ( x ) Pn ( x ) = Rn ( x ) , the error involved in using this polynomial approximation

3-14

DEFINITION 3.1.

For the nth Taylor approximation Pn ( x ) at x = a of f ( x ) , the error at x,

Rn ( x ) = f ( x ) Pn ( x ) , satisfies
Rn ( x )

M
n +1
xa
( n + 1)!

where M is any number such that f ( n +1) ( z ) M for all z between a and x.

EXAMPLE 3.9.

Use the Taylor polynomial of degree 1 and 2 to approximate e x (x 0).

SOLUTION.

P1 ( x ) = f ( 0 ) + f ' ( 0 )( x 0 ) = e0 + e0 x = 1 + x
P2 ( x ) = f ( 0 ) + f ' ( 0 )( x 0 ) +

f '' ( 0 )
2
( x 0 ) = e0 + e0 x + e0 x 2 = 1 + x + x 2
2!

ex

P1(x)

P2(x)

-0.4

0.6703

0.6

0.680

-0.1

0.9048

0.9

0.905

0.4

1.4918

1.4

1.480

1.0

2.7183

2.0

2.500

1.5

4.4817

2.5

3.625

2.0

7.3891

3.0

5.000

3-15

Approximate

EXAMPLE 3.10.

0.1

e x dx with the magnitude of error E 0.000005

(accurate to five decimal places).

SOLUTION.

e x = ( 1)
2

n=0

0.1

x2

dx =

0.1

x2n
n!

x <

2n

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

( 1) ( 0.1)
n ! ( 2n + 1)
n =0
n

2 n +1

= 0.1

x 2 n +1

( 2n + 1)

0.1

0.001 0.00001 0.0000001

+
3
10
42

We need to find n such that

( 1) ( 0.1)
n ! ( 2n + 1)

2 n +1

( 0.1) < 0.000005 .

The smallest value of n that would fulfill the above requirement is n = 3 . Therefore, to obtain
the required accuracy, we need to use the terms when n = 0,1, 2 .

0.1

e x dx 0.1
2

0.001 0.00001
+
= 0.099667
3
10

3-16

EXAMPLE 3.10A.

Estimate ln ( 0.95 ) using a Taylor polynomial of degree 5.

i.

What is the maximum error in the approximation?

ii.

What is the actual error?

iii.

How many terms are needed to estimate ln ( 0.95 ) to an accuracy of 1015 ?

SOLUTION.

3-17

EXAMPLE 3.10B.

Estimate the maximum error in using Taylor polynomial P3 ( x ) about

a = 1 to approximate f ( x ) =

1
when 0.9 x 1.1 .
x+2

SOLUTION.

3-18

BINOMIAL SERIES

This series has the form

(1 + x )

EXAMPLE 3.11.

= 1 + nx +

n ( n 1) 2 n ( n 1)( n 2 ) 3
x +
x +
2!
3!

Find power series representation for

x < 1.

1
.
1+ x

SOLUTION.

3-19