Difference Equations Section 5.

7
to
Differential Equations Power Series

We are now in a position to pick up the story we left off in Section 5.2: the extension of
Taylor polynomials to Taylor series. We shall see that a Taylor series is a type of infinite
series whose nth partial sum is a Taylor polynomial. Such series are examples of power
series, objects that we will study in this section before considering Taylor series in Section
5.8.
Definition An infinite series of the form
∞
X
an (x − c)n = a0 + a1 (x − c) + a2 (x − c)2 + · · · (5.7.1)
n=0

is called a power series in x about c.

Example The infinite series
∞
X xn x2 x3
=1+x+ + + ···
n=0
n! 2! 3!

is a power series in x about 0. Note that if we let

 n
 x  |x|n
bn =   =
n! n!

for n = 0, 1, 2, . . ., then

|x|n+1
bn+1 (n + 1)! |x|
lim = lim n = lim =0
n→∞ bn n→∞ |x| n→∞ n + 1

n!
for any value of x. That is, by the ratio test, the series is absolutely convergent, and
hence convergent, for any value of x. Thus if we define a function, called the exponential
function, by
∞
X xn
exp(x) = (5.7.2)
n=0
n!
then this function is defined for all values of x. We shall have much more to say about this
function, which may be thought of as the simplest “infinite” polynomial which is defined
for all real numbers, in Chapter 6.

1 Copyright
c by Dan Sloughter 2000
2 Power Series Section 5.7

Notice that the convergence of (5.7.2) for all x implies, by the nth term test for diver-
gence, that
xn
lim =0 (5.7.3)
n→∞ n!

for any value of x. We have seen particular cases of this limit in the past, but this is the
first time we have had a simple proof that it is always 0.
Example Recall that the Taylor polynomial of order 2n + 1 for sin(x) at 0 is
n
X (−1)k x2k+1
P2n+1 (x) = .
(2k + 1)!
k=0

Hence P2n+1 (x) is a partial sum of the power series

∞
X (−1)k x2k+1
. (5.7.4)
(2k + 1)!
k=0

If, for k = 0, 1, 2, . . ., we let

 (−1)k x2k+1  2k+1

 
bk =   = |x| ,
(2k + 1)!  (2k + 1)!

then
|x|2k+3
bk+1 (2k + 3)! |x|2
lim = lim = lim =0
k→∞ bk k→∞ |x|2k+1 k→∞ (2k + 3)(2k + 2)

(2k + 1)!
for all values of x. Thus, by the ratio test, (5.7.4) is absolutely convergent, and hence
convergent, for all values of x. Moreover, from our work in Section 5.2, we know that

|x|2n+3
| sin(x) − P2n+1 (x)| ≤
(2n + 3)!

for all values of x. Since (using (5.7.3))

|x|2n+3
lim = 0,
n→∞ (2n + 3)!

it follows that
lim | sin(x) − P2n+1 (x)| = 0
n→∞

for all values of x. Hence

sin(x) = lim P2n+1 (x)
n→∞
Section 5.7 Power Series 3

for all values of x. That is, for any value of x,

∞
X (−1)k x2k+1
sin(x) = . (5.7.5)
(2k + 1)!
k=0

P∞
Example The series n=0 xn is a power series in x about 0. From our work on geometric
series, we know that this series will converge absolutely when −1 < x < 1 and will diverge
otherwise. In fact, we have seen that
∞
1 X
= xn
1 − x n=0

for −1 < x < 1.

These examples show that some functions may be expressed as power series. Such
functions are examples of analytic functions, which we now define.
Definition If f is a function for which there exists constants a0 , a1 , a2 , . . . such that
∞
X
f (x) = an (x − c)n (5.7.6)
n=0

for all values of x in some open interval about c, then we say f is analytic at c. If for some
h > 0 the equality (5.7.6) holds for all x in the interval I = (c − h, c + h), then we say f
is analytic on I and we call
X∞
an (x − c)n
n=0

a power series representation of f on I.

Example From the previous examples we see that

f (x) = exp(x)

and
g(x) = sin(x)
are analytic on (−∞, ∞) and
1
h(x) =
1−x
is analytic on (−1, 1).
Before we can work effectively with power series we need to consider their convergence
behavior. First note that the power series
∞
X
an (x − c)n (5.7.7)
n=0
4 Power Series Section 5.7

converges at x = c since in that case all terms after the first are P
0. Next, suppose the
∞
series converges at a point x = c + r, where r > 0. That is, suppose n=0 an rn converges.
Then, by the nth term test for divergence,

lim an rn = 0. (5.7.8)
n→∞

In particular, there exists an integer N such that

|an rn | < 1 (5.7.9)

for all n > N . Hence for any x and n > N ,

|an (x − c)n |
= |an rn | < 1. (5.7.10)
 x − c n
 
 
 r 

In particular, |an (x − c)n | is

 x − c n
  
O 
  .
r 
Now if c − r < x < c + r, then  
x − c
 r <1
 

and
∞ 
 x − c n
X 
 
 r 
n=0

is a convergent geometric series. Thus, by the limit comparison test,

∞
X
an (x − c)n
n=0

converges absolutely. In other words, we have shown that if (5.7.7) converges at c + r with
r > 0, then it converges absolutely for all x in (c − r, c + r). The same argument works to
show that if (5.7.7) converges at c − r with r > 0, then it converges absolutely for all x in
(c − r, c + r). Letting R be the largest real number such that (5.7.7) converges absolutely
for all x for which |x − c| < R, where we allow R = ∞ if (5.7.7) converges for all x, it
follows that
X∞
an (x − c)n
n=0

converges absolutely on (c − R, c + R) ((−∞, ∞) if R = ∞) and diverges for all x with

|x − c| > R.
Section 5.7 Power Series 5

Proposition For a power series

∞
X
an (x − c)n ,
n=0

there exists an R, with R = 0, R > 0, or R = ∞, such that the series converges absolutely
for all x satisfying |x − c| < R and diverges for all x satisfying |x − c| > R.
Definition With the notation of the previous proposition, the interval (c − R, c + R)
((−∞, ∞) if R = ∞) is called the interval of convergence and R is called the radius of
convergence of the power series.
Note that the proposition does not say anything about the behavior of the series at
x = c − R or x = c + R. In fact, any type of behavior is possible at the endpoints of the
interval of convergence; for a given series, these points must be checked individually for
convergence. Moreover, although the proposition does not provide a method for finding
the interval of convergence of a series, the next examples illustrate that the ratio test is
very useful in this regard.
Example Consider the power series
∞
X (−1)n+1 xn
. (5.7.11)
n=1
n2n

If we let
 (−1)n+1 xn  |x|n
 
bn = 
 =
 n2n ,
n2n
then
|x|n+1
 
bn+1 (n + 1)2n+1 n |x| |x|
lim = lim n = lim = .
n→∞ bn n→∞ |x| n→∞ n + 1 2 2
n2n
Hence, by the ratio test, (5.7.11) is absolutely convergent when

|x|
< 1,
2
that is, when −2 < x < 2. Thus the radius of convergence is R = 2 and the interval of
convergence is (−2, 2). Now at x = −2, (5.7.11) becomes
∞ ∞ ∞ ∞
X (−1)n+1 (−2)n X (−1)n+1 (−1)n 2n X (−1)2n+1 X 1
n
= n
= = − ,
n=1
n2 n=1
n2 n=1
n n=1
n

which is a multiple of the harmonic series and hence divergent. At x = 2, (5.7.11) becomes
∞ ∞
X (−1)n+1 2n X (−1)n+1
n
= ,
n=1
n2 n=1
n
6 Power Series Section 5.7

which is the alternating harmonic series and hence convergent, although not absolutely
convergent. Putting this together, we see that the power series
∞
X (−1)n+1 xn
n=1
n2n

converges absolutely for all x in (−2, 2), converges conditionally at x = 2, and diverges for
all other x.
Example In the first example of this section we used the ratio test to show that the
power series
∞
X xn
n=1
n!

converges absolutely for all values of x. Hence in this case the interval of convergence is
(−∞, ∞) and the radius of convergence is R = ∞.
P∞
Example Consider the power series n=1 n!xn . If we let

bn = |n!xn | = n!|x|n ,

then
(n + 1)!|x|n+1

bn+1 0, if x = 0,
lim = lim = lim (n + 1)|x| =
n→∞ bn n→∞ n!|x|n n→∞ ∞, if x 6= 0.
Hence, by the ratio test, this power series converges only when x = 0. Accordingly, the
radius of convergence is R = 0.
Example Consider the power series
∞
X (x − 1)n
. (5.7.12)
n=1
n

If we let
 (x − 1)n  |x − 1|n
 
bn =  = ,
n  n
then
|x − 1|n+1  
bn+1 n + 1 n
lim = lim = lim |x − 1| = |x − 1|.
n→∞ bn n→∞ |x − 1|n n→∞ n + 1

n
Using the ratio test, we see that (5.7.12) converges absolutely when |x − 1| < 1. Thus the
radius of convergence is R = 1 and the interval of convergence is (0, 2). At x = 0, (5.7.12)
becomes
∞ ∞
X (−1)n X (−1)n+1
=− ,
n=1
n n=1
n
Section 5.7 Power Series 7

which is a multiple of the alternating harmonic series and so converges conditionally. At

x = 2, (5.7.12) becomes
∞
X 1
,
n=1
n

which is the harmonic series and so diverges. Hence the power series
∞
X (x − 1)n
n=1
n

converges absolutely for x in the interval (0, 2), converges conditionally at x = 0, and
diverges for all other x.
A power series resembles a polynomial; in fact, often it is convenient to think of a power
series as a polynomial of infinite degree. Among the many nice properties of polynomials
is the ease with which they may be differentiated and integrated. Our next result states
that power series may be differentiated and integrated term by term in the same manner
as polynomials. Although we have the tools to provide justifications for these statements,
they are technical and perhaps best left to a more advanced text.
Differentiation and integration of power series Suppose the radius of convergence
of the power series
X∞
an (x − c)n
n=0

is R > 0 and let

∞
X
f (x) = an (x − c)n
n=0

for x in (c − R, c + R). Then

∞ ∞
0
X d n
X
f (x) = an (x − c) = nan (x − c)n−1 (5.7.13)
n=0
dx n=1

for all x in (c − R, c + R) and

∞ ∞
!
b b
an (b − c)n+1
Z X Z X
f (x)dx = an (x − c)n dx = (5.7.14)
c n=0 c n=0
n+1

for all b in (c − R, c + R).

Example Recall that the interval of convergence of
∞
X xn
exp(x) =
n=0
n!
8 Power Series Section 5.7

is (−∞, ∞). Hence

∞
d d X xn
exp(x) =
dx dx n=0 n!
∞
d xn
X  
=
n=0
dx n!
∞
X nxn−1
=
n=1
n!
∞
X xn−1
=
n=1
(n − 1)!
∞
X xn
=
n=0
n!
= exp(x)
for all x in (−∞, ∞). That is, the function exp(x) is its own derivative. We will have much
more to say about this interesting property of the exponential function in Chapter 6.
Example From our work above we know that
∞
X (−1)n x2n+1
sin(x) =
n=0
(2n + 1)!

for all x in (−∞, ∞). Now

Z x x
sin(t)dt = − cos(t) = − cos(x) + 1

0 0

for any x. However, we also know that

∞
!
x x
(−1)n t2n+1
Z Z X
sin(t)dt = dt
0 0 n=0
(2n + 1)!
∞ Z x
X (−1)n t2n+1
= dt
n=0 0 (2n + 1)!
∞ x
X (−1)n t2n+2 
=
n=0
(2n + 2)(2n + 1)! 0
∞
X (−1)n x2n+2
=
n=0
(2n + 2)!
∞
X (−1)n−1 x2n
= .
n=1
(2n)!
Section 5.7 Power Series 9

Hence Z x
cos(x) = 1 − sin(t)dt
0
∞
X (−1)n−1 x2n
=1−
n=1
(2n)!
∞
X (−1)n x2n
=1+
n=1
(2n)!
∞
X (−1)n x2n
=
n=0
(2n)!

for all x in (−∞, ∞). Thus we have found a power series representation of cos(x) on
(−∞, ∞). In particular, cos(x)is analytic on (−∞, ∞).
To close this section we note that a power series representation of a function about a
specific point c is unique. To see this, suppose
∞
X
f (x) = an (x − c)n (5.7.15)
n=0

on (c − R, c + R), where R > 0 is the radius of convergence of the power series. We need
to show that the coefficients an , n = 0, 1, 2, . . ., are uniquely determined by f . To start,
∞
X
f (c) = an (c − c)n
n=0
(5.7.16)
= a0 + a1 (c − c) + a2 (c − c)2 + a3 (c − c)3 + · · ·
= a0 ,

so
a0 = f (c). (5.7.17)
Next,
∞
X
0
f (c) = nan (c − c)n−1 = a1 , (5.7.18)
n=1
so
a1 = f 0 (c). (5.7.19)
For a2 we have
∞
X
00
f (c) = n(n − 1)an (c − c)n−2 = 2a2 , (5.7.20)
n=2
so
f 00 (c)
a2 = . (5.7.21)
2
10 Power Series Section 5.7

In general, for k = 0, 1, 2, . . .,

∞
X
(k)
f (c) = n(n − 1) · · · (n − k + 1)an (c − c)n−k = k!ak , (5.7.22)
n=k

from which it follows that

f (k) (c)
ak = . (5.7.23)
k!
As a consequence, the power series representation of f about c is uniquely determined by
the values of the derivatives of f at c.
Proposition Suppose
∞
X
f (x) = an (x − c)n (5.7.24)
n=0

on (c − R, c + R), where R > 0 is the radius of convergence of the power series. Then

f (n) (c)
an = (5.7.25)
n!

for n = 0, 1, 2, . . ..
Note that the coefficients an as given by (5.7.25) are the same as the coefficients used
in the definition of the Taylor polynomial of f at c. This observation leads immediately
to the question of extending Taylor polynomials to Taylor series, the topic of Section 5.8.
Our final example of this section illustrates how (5.7.25) may be used to find the
derivatives of f at c if we already know a power series representation for f about c.
Example In a previous example we saw that if

1
f (x) = ,
1−x

then
∞
X
f (x) = xn
n=0

for all x in (−1, 1). In this series, the coefficient of xn is 1 for all n, so, by the previous
proposition,
f (n) (0)
1=
n!
for n = 0, 1, 2, . . .. That is, f (n) (0) = n! for all n.
Section 5.7 Power Series 11

Problems

1. For each of the following power series, find the interval of convergence and determine
the behavior of the series at the endpoints of the interval. State clearly where the series
converges absolutely, where it converges conditionally, and where it diverges. Also, for
each series write out the first 5 nonzero terms.
∞ ∞
X xn X (−1)n+1 xn
(a) (b)
n=1
n n=1
n
∞ ∞
X
n
X x2n
(c) nx (d)
n=1 n=0
n!
∞ ∞
X (x − 2)n+1 X (−1)n xn
(e) (f)
n=0
n+1 n=0
3n
∞ ∞
X (x − 1)n X
(g) (h) n3 x2n
n=1
n2 n=1

2. For each of the following power series, find the interval of convergence and determine
the behavior of the series at the endpoints of the interval. State clearly where the series
converges absolutely, where it converges conditionally, and where it diverges. Also, for
each series write out the first 5 nonzero terms.
∞ ∞
X (x − 3)n X xn
(a) (b)
n=0
n! n=1
n5n
∞ ∞
X x2n X (−1)n+1 (x − 6)n
(c) (d)
n=0
2n n=1
n3n
∞ ∞
X x2n+1 X (−1)n (x − 1)2n
(e) (f)
n=1
n n=0
(2n)!
∞ ∞
X X (−1)n x2n+1
(g) 3n xn (h)
n=1 n=0
(2n + 1)(2n + 1)!

3. (a) Using the fact that

∞
1 X
= xn
1 − x n=0

for −1 < x < 1, find a power series representation about 0 for

1
f (x) =
1+x

on (−1, 1).
(b) Use your result from (a) to find f (35) (0).
12 Power Series Section 5.7

(c) Use your result from (a) to find a power series representation about 0 for
Z x
1
dt
0 1+t

on (−1, 1). Determine where the series converges absolutely, where it converges
conditionally, and where it diverges.
(d) Use your result from (c) to find an infinite series representation for
Z 1
1
dt.
0 1+t

Use this series to estimate the integral with an error of no more than 0.001 in
absolute value.
4. Use the power series representations of sin(x) and cos(x) about 0 to prove the following
identities.
(a) sin(−x) = − sin(x) (b) cos(−x) = cos(x)
d d
(c) sin(x) = cos(x) (d) cos(x) = − sin(x)
dx dx
5. Using the power series representation of cos(x) about 0, find an infinite series repre-
sentation of cos(1). Use the infinite series to estimate cos(1) with an error of no more
than 0.000001.
6. Use the fact that  
d 1 1
=
dx 1−x (1 − x)2
to find a power series representation about 0 for

1
g(x) = .
(1 − x)2

Find the interval of convergence for this power series and determine the behavior of
the series at the endpoints.
7. Use your result from Problem 6 to evaluate
∞
X n
.
n=1
2n−1

8. (a) A fair coin is tossed repeatedly. In Section 1.3 we saw that the probability that a
head appears for the first time on the nth toss is

1
Pn =
2n
Section 5.7 Power Series 13

for n = 1, 2, 3, . . .. The average number of tosses before the first head appears is
then given by
X∞
A= nPn .
n=1

Use Problem 7 to find A.

(b) A manufacturer of circuit boards tests every board as it comes off the assembly
line. If the probability that a board passes the test is p and the probability that
it fails is q = 1 − p, then the probability that n boards are tested before the first
defective board is encountered is Pn = pn−1 q. The average number of boards
tested before finding a defective one is then
∞
X
A= nPn .
n=1

Find A.