Operations on Power Series Related to Taylor Series

In this problem, we perform elementary operations on Taylor series term by term dieren
tiation and integration to obtain new examples of power series for which we know their sum.
Suppose that a function f has a power series representation of the form:


f (x) = a0 + a1 (x c) + a2 (x c)2 + = an (x c)n
n=0

convergent on the interval (c R, c + R) for some R. The results we use in this example are:

(Dierentiation) Given f as above, f (x) has a power series expansion obtained by by dier
entiating each term in the expansion of f (x):

f (x) = a1 + a2 (x c) + 2a3 (x c) + =

nan (x c)n1
n=1

(Integration) Given f as above, f (x) dx has a power series expansion obtained by by inte
grating each term in the expansion of f (x):

a1 a2 an

f (x) dx = C + a0 (x c) + (x c)2 + (x c)3 + = C + (x c)n+1
2 3 n=0
n + 1

for some constant C depending on the choice of antiderivative of f .

Questions:

1. Find a power series representation for the function f (x) = arctan(5x). (Note: arctan x is the
inverse function to tan x.)

2. Use power series to approximate 1

(Note: sin(x2 ) is a function whose antiderivative is not an elementary function.)

Solution:
1
For question (1), we know that arctan x has a simple derivative: , which then has a power
1 + x2
1
series representation similar to that of , where we subsitute x2 for x. Hence:
1x

d 5
2 n

arctan(5x) = = 5 (25x ) = (1)n 52n+1 x2n ,
dx 1 + 25x2 n=0 n=0

1
where the second equality above follows from the familiar geometric series representation for .
1x
The last equality presents a cleaner nal form after straightforward algebraic simplication. Thus
to obtain a power series expression for arctan x we may integrate this power series expression term
by term. This gives:

53 3 52n+1 2n+1
arctan(5x) = C + 5x
x
+
= C + (1)n x ,

3
n=0
2n + 1

and we may solve for C by comparing both sides of the equality for any value of x. Choosing x = 0,
we see that arctan(x) = 0 and all non-constant terms of the power series are 0, hence C = 0 as
well.

For question (2), we have seen that sin(x) has a power series expansion:

x
3 x
5 x
2n+1
sin(x) = x
+ = (1)n .

3!
5! n=0
(2n + 1)!

Using a change of variable (replacing x by x2 in the power series above), we have the power series
expansion

x
6 x
10 x
4n+2
sin(x2 ) = x2 + =
(1)n .

3! 5! n=0
(2n + 1)!

Now taking the indenite integral of both sides, we obtain a power series representation for the
antiderivative of sin(x2 ):

1 1
x
7 1 x
10 (1)n x
4n+3

sin(x
2 ) dx = x
3 + =
.

3
7 3! 11 5! n=0
4n + 3 (2n + 1)!

The power series expression is valid for any real number x since the power series for sin(x), and
hence sin(x2 ) converged for all x.
To approximate the denite integral, we may use as many terms of the series as we like. For
example, using only the rst non-zero term would give:

1
x=1 1

1
sin(x
) dx x
3
2
=
.

0 3
x=0 3

The rst two non-zero terms gives:

x=1
1 3 1 x7
2
1 1
13

2
sin(x ) dx x

=
=
.

1 3
7 3!

x=0
3 42 42

Using a numerical integration on a computer-algebra system, we nd that the answer is approxi

18.01SC Single Variable Calculus

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.