4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 833

11.11 Fourier Series 833

11.11 Fourier Series

We have seen how Taylor series can be used to approximate a function ƒ by polynomials.
HISTORICAL BIOGRAPHY The Taylor polynomials give a close fit to ƒ near a particular point x = a, but the error in
Jean-Baptiste Joseph Fourier the approximation can be large at points that are far away. There is another method that
often gives good approximations on wide intervals, and often works with discontinuous
(1766–1830) functions for which Taylor polynomials fail. Introduced by Joseph Fourier, this method ap-
proximates functions with sums of sine and cosine functions. It is well suited for analyzing
periodic functions, such as radio signals and alternating currents, for solving heat transfer
problems, and for many other problems in science and engineering.

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 834

834 Chapter 11: Infinite Sequences and Series

Suppose we wish to approximate a function ƒ on the interval [0, 2p] by a sum of sine
and cosine functions,
ƒnsxd = a0 + sa1 cos x + b1 sin xd + sa2 cos 2x + b2 sin 2xd + Á
+ san cos nx + bn sin nxd
or, in sigma notation,
n
ƒnsxd = a0 + a sak cos kx + bk sin kxd. (1)
k=1

We would like to choose values for the constants a0 , a1, a2 , Á an and b1, b2 , Á , bn that
make ƒnsxd a “best possible” approximation to ƒ(x). The notion of “best possible” is
defined as follows:
1. ƒnsxd and ƒ(x) give the same value when integrated from 0 to 2p.
2. ƒnsxd cos kx and ƒ(x) cos kx give the same value when integrated from 0 to
2p sk = 1, Á , nd.
3. ƒnsxd sin kx and ƒ(x) sin kx give the same value when integrated from 0 to
2p sk = 1, Á , nd.
Altogether we impose 2n + 1 conditions on ƒn :
2p 2p
ƒnsxd dx = ƒsxd dx,
L0 L0
2p 2p
ƒnsxd cos kx dx = ƒsxd cos kx dx, k = 1, Á , n,
L0 L0
2p 2p
ƒnsxd sin kx dx = ƒsxd sin kx dx, k = 1, Á , n.
L0 L0
It is possible to choose a0 , a1, a2 , Á an and b1, b2 , Á , bn so that all these conditions are
satisfied, by proceeding as follows. Integrating both sides of Equation (1) from 0 to 2p
gives
2p
ƒnsxd dx = 2pa0
L0
since the integral over [0, 2p] of cos kx equals zero when k Ú 1, as does the integral of
sin kx. Only the constant term a 0 contributes to the integral of ƒn over [0, 2p]. A similar
calculation applies with each of the other terms. If we multiply both sides of Equation (1)
by cos x and integrate from 0 to 2p then we obtain
2p
ƒnsxd cos x dx = pa1 .
L0
This follows from the fact that
2p
cos px cos px dx = p
L0
and
2p 2p 2p
cos px cos qx dx = cos px sin mx dx = sin px sin qx dx = 0
L0 L0 L0

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 835

11.11 Fourier Series 835

whenever p, q and m are integers and p is not equal to q (Exercises 9–13). If we multiply
Equation (1) by sin x and integrate from 0 to 2p we obtain
2p
ƒnsxd sin x dx = pb1 .
L0

Proceeding in a similar fashion with

cos 2x, sin 2x, Á , cos nx, sin nx

we obtain only one nonzero term each time, the term with a sine-squared or cosine-
squared term. To summarize,

2p
ƒnsxd dx = 2pa0
L0
2p
ƒnsxd cos kx dx = pak, k = 1, Á , n
L0
2p
ƒnsxd sin kx dx = pbk, k = 1, Á , n
L0

We chose ƒn so that the integrals on the left remain the same when ƒn is replaced by ƒ, so
we can use these equations to find a0 , a1, a2 , Á an and b1, b2 , Á , bn from ƒ:
2p
1
a0 = ƒsxd dx (2)
2p L0

2p
1
ak = p ƒsxd cos kx dx, k = 1, Á , n (3)
L0

2p
1
bk = p ƒsxd sin kx dx, k = 1, Á , n (4)
L0

The only condition needed to find these coefficients is that the integrals above must exist.
If we let n : q and use these rules to get the coefficients of an infinite series, then the re-
sulting sum is called the Fourier series for ƒ(x),
q
a0 + a sak cos kx + bk sin kxd. (5)
k=1

EXAMPLE 1 Finding a Fourier Series Expansion

Fourier series can be used to represent some functions that cannot be represented by Taylor
series; for example, the step function ƒ shown in Figure 11.16a.

ƒsxd = e
1, if 0 … x … p
2, if p 6 x … 2p.

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 836

836 Chapter 11: Infinite Sequences and Series

y y

2 2

1 1

x x
0 ␲ 2␲ –2␲ –␲ 0 ␲ 2␲ 3␲ 4␲

(a) (b)

FIGURE 11.16 (a) The step function

ƒsxd = e
1, 0 … x … p
2, p 6 x … 2p

(b) The graph of the Fourier series for ƒ is periodic and has the value 3>2 at each point of
discontinuity (Example 1).

The coefficients of the Fourier series of ƒ are computed using Equations (2), (3), and (4).
2p
1
a0 = ƒsxd dx
2p L0
p 2p
a
1 3
= 1 dx + 2 dxb =
2p L0 Lp 2

2p
1
ak = p ƒsxd cos kx dx
L0
p 2p
= p a
1
cos kx dx + 2 cos kx dxb
L0 Lp
p 2p
= p ac d + c d b = 0,
1 sin kx 2 sin kx
k Ú 1
k 0 k p

2p
1
bk = p ƒsxd sin kx dx
L0
p 2p
= p a
1
sin kx dx + 2 sin kx dxb
L0 Lp
p 2p
= p a c- d + c- d b
1 cos kx 2 cos kx
k 0 k p

cos kp - 1 s -1dk - 1
= = .
kp kp
So
3
a0 = , a1 = a2 = Á = 0,
2

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 837

11.11 Fourier Series 837

and
2 2 2
b1 = - p , b2 = 0, b3 = - , b4 = 0, b5 = - , b6 = 0, Á
3p 5p
The Fourier series is

- p asin x + + Áb .
3 2 sin 3x sin 5x
+
2 3 5
Notice that at x = p, where the function ƒ(x) jumps from 1 to 2, all the sine terms vanish,
leaving 3> 2 as the value of the series. This is not the value of ƒ at p, since ƒspd = 1. The
Fourier series also sums to 3> 2 at x = 0 and x = 2p. In fact, all terms in the Fourier se-
ries are periodic, of period 2p, and the value of the series at x + 2p is the same as its
value at x. The series we obtained represents the periodic function graphed in Figure
11.16b, with domain the entire real line and a pattern that repeats over every interval of
width 2p. The function jumps discontinuously at x = np, n = 0, ;1, ;2, Á and at
these points has value 3> 2, the average value of the one-sided limits from each side. The
convergence of the Fourier series of ƒ is indicated in Figure 11.17.

y y y

2 f 2 f 2 f

1.5 f1 1.5 f3 1.5 f5

1 1 1

x x x
0 ␲ 2␲ 0 ␲ 2␲ 0 ␲ 2␲
(a) (b) (c)

y y

2 f 2 f

1.5 f9 1.5 f15

1 1

x x
0 ␲ 2␲ 0 ␲ 2␲
(d) (e)

The Fourier approximation functions ƒ1, ƒ3, ƒ5, ƒ9 , and ƒ15 of the function ƒsxd = e
1, 0 … x … p
FIGURE 11.17 in Example 1.
2, p 6 x … 2p

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 838

838 Chapter 11: Infinite Sequences and Series

Convergence of Fourier Series

Taylor series are computed from the value of a function and its derivatives at a single point
x = a, and cannot reflect the behavior of a discontinuous function such as ƒ in Example 1
past a discontinuity. The reason that a Fourier series can be used to represent such functions
is that the Fourier series of a function depends on the existence of certain integrals, whereas
the Taylor series depends on derivatives of a function near a single point. A function can be
fairly “rough,” even discontinuous, and still be integrable.
The coefficients used to construct Fourier series are precisely those one should choose
to minimize the integral of the square of the error in approximating ƒ by ƒn . That is,
2p
[ƒsxd - ƒnsxd]2 dx
L0
is minimized by choosing a0 , a1, a2 , Á an and b1, b2 , Á , bn as we did. While Taylor series
are useful to approximate a function and its derivatives near a point, Fourier series mini-
mize an error which is distributed over an interval.
We state without proof a result concerning the convergence of Fourier series. A func-
tion is piecewise continuous over an interval I if it has finitely many discontinuities on the
interval, and at these discontinuities one-sided limits exist from each side. (See Chapter 5,
Additional Exercises 11–18.)

THEOREM 24 Let ƒ(x) be a function such that ƒ and ƒ¿ are piecewise contin-
uous on the interval [0, 2p]. Then ƒ is equal to its Fourier series at all points
where ƒ is continuous. At a point c where ƒ has a discontinuity, the Fourier series
converges to
ƒsc + d + ƒsc - d
2
where ƒsc + d and ƒsc - d are the right- and left-hand limits of ƒ at c.

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 838

838 Chapter 11: Infinite Sequences and Series

EXERCISES 11.11

e x,
6. ƒsxd = e
Finding Fourier Series 0 … x … p
In Exercises 1–8, find the Fourier series associated with the given 0, p 6 x … 2p

7. ƒsxd = e
functions. Sketch each function. cos x, 0 … x … p
1. ƒsxd = 1 0 … x … 2p . 0, p 6 x … 2p

2. ƒsxd = e 8. ƒsxd = e
1, 0 … x … p 2, 0 … x … p
-1, p 6 x … 2p -x, p 6 x … 2p

3. ƒsxd = e
x, 0 … x … p
x - 2p, p 6 x … 2p
Theory and Examples
Establish the results in Exercises 9–13, where p and q are positive
x 2,
4. ƒsxd = e
0 … x … p
integers.
0, p 6 x … 2p 2p
5. ƒsxd = e x 0 … x … 2p . 9. cos px dx = 0 for all p .
L0

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 839

11.11 Fourier Series 839

2p 14. Fourier series of sums of functions If ƒ and g both satisfy the

10. sin px dx = 0 for all p . conditions of Theorem 24, is the Fourier series of ƒ + g on
L0
2p [0, 2p] the sum of the Fourier series of ƒ and the Fourier series of
cos px cos qx dx = e
0, if p Z q
11. . g? Give reasons for your answer.
L0 p, if p = q
15. Term-by-term differentiation
sHint: cos A cos B = s1>2d[cossA + Bd + cossA - Bd].d
2p
a. Use Theorem 24 to verify that the Fourier series for ƒsxd in
sin px sin qx dx = e
0, if p Z q Exercise 3 converges to ƒsxd for 0 6 x 6 2p .
12. .
L0 p, if p = q
b. Although ƒ¿sxd = 1 , show that the series obtained by term-
sHint: sin A sin B = s1>2d[cos sA - Bd - cos sA + Bd].d by-term differentiation of the Fourier series in part (a)
2p diverges.
13. sin px cos qx dx = 0 for all p and q .
L0 16. Use Theorem 24 to find the Value of the Fourier series determined
p2
q
sHint: sin A cos B = s1>2d[sin sA + Bd + sin sA - Bd].d 1
in Exercise 4 and show that = a 2.
6 n=1 n

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 839

Chapter 11 Questions to Guide Your Review 839

Chapter 11 Questions to Guide Your Review

1. What is an infinite sequence? What does it mean for such a se- 15. When do p-series converge? Diverge? How do you know? Give
quence to converge? To diverge? Give examples. examples of convergent and divergent p-series.
2. What is a nondecreasing sequence? Under what circumstances 16. What are the Direct Comparison Test and the Limit Comparison
does such a sequence have a limit? Give examples. Test? What is the reasoning behind these tests? Give examples of
3. What theorems are available for calculating limits of sequences? their use.
Give examples. 17. What are the Ratio and Root Tests? Do they always give you the
4. What theorem sometimes enables us to use l’Hôpital’s Rule to information you need to determine convergence or divergence?
calculate the limit of a sequence? Give an example. Give examples.
5. What six sequence limits are likely to arise when you work with 18. What is an alternating series? What theorem is available for deter-
sequences and series? mining the convergence of such a series?
6. What is an infinite series? What does it mean for such a series to 19. How can you estimate the error involved in approximating the
converge? To diverge? Give examples. sum of an alternating series with one of the series’ partial sums?
What is the reasoning behind the estimate?
7. What is a geometric series? When does such a series converge?
Diverge? When it does converge, what is its sum? Give examples. 20. What is absolute convergence? Conditional convergence? How
are the two related?
8. Besides geometric series, what other convergent and divergent se-
ries do you know? 21. What do you know about rearranging the terms of an absolutely
convergent series? Of a conditionally convergent series? Give
9. What is the nth-Term Test for Divergence? What is the idea be-
examples.
hind the test?
22. What is a power series? How do you test a power series for con-
10. What can be said about term-by-term sums and differences of
vergence? What are the possible outcomes?
convergent series? About constant multiples of convergent and di-
vergent series? 23. What are the basic facts about
11. What happens if you add a finite number of terms to a convergent a. term-by-term differentiation of power series?
series? A divergent series? What happens if you delete a finite b. term-by-term integration of power series?
number of terms from a convergent series? A divergent series? c. multiplication of power series?
12. How do you reindex a series? Why might you want to do this? Give examples.
13. Under what circumstances will an infinite series of nonnegative 24. What is the Taylor series generated by a function ƒ(x) at a point
terms converge? Diverge? Why study series of nonnegative x = a ? What information do you need about ƒ to construct the
terms? series? Give an example.
14. What is the Integral Test? What is the reasoning behind it? Give 25. What is a Maclaurin series?
an example of its use.

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch11p746-847 8/25/04 2:41 PM Page 840

840 Chapter 11: Infinite Sequences and Series

26. Does a Taylor series always converge to its generating function? 31. How can you sometimes use power series to estimate the values
Explain. of nonelementary definite integrals?
27. What are Taylor polynomials? Of what use are they? 32. What are the Taylor series for 1>s1 - xd, 1>s1 + xd, e x, sin x,
28. What is Taylor’s formula? What does it say about the errors in- cos x, ln s1 + xd, ln [s1 + xd>s1 - xd] , and tan-1 x ? How do
volved in using Taylor polynomials to approximate functions? In you estimate the errors involved in replacing these series with
particular, what does Taylor’s formula say about the error in a lin- their partial sums?
earization? A quadratic approximation? 33. What is a Fourier series? How do you calculate the Fourier coeffi-
29. What is the binomial series? On what interval does it converge? cients a0 , a1, a2 , Á and b1, b2 , Á for a function ƒ(x) defined on
How is it used? the interval [0, 2p] ?
30. How can you sometimes use power series to solve initial value 34. State the theorem on convergence of the Fourier series for ƒ(x)
problems? when ƒ and ƒ¿ are piecewise continuous on [0, 2p] .

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley