Chapter 10

Fourier Series

10.1 Periodic Functions and Orthogonality Relations

The differential equation

𝑦 ′′ + 𝛽 2 𝑦 = 𝐹 cos 𝜔𝑡

models a mass-spring system with natural frequency 𝛽 with a pure cosine

forcing function of frequency 𝜔. If 𝛽 2 ∕= 𝜔 2 a particular solution is easily
found by undetermined coefficients (or by using Laplace transforms) to be

𝐹
𝑦𝑝 = cos 𝜔𝑡.
𝛽2 − 𝜔2
If the forcing function is a linear combination of simple cosine functions, so
that the differential equation is
𝑁
∑
𝑦 ′′ + 𝛽 2 𝑦 = 𝐹𝑛 cos 𝜔𝑛 𝑡
𝑛=1

where 𝛽 2 ∕= 𝜔𝑛2 for any 𝑛, then, by linearity, a particular solution is obtained

as a sum
𝑁
∑ 𝐹𝑛
𝑦𝑝 (𝑡) = 2 − 𝜔2
cos 𝜔𝑛 𝑡.
𝑛=1
𝛽 𝑛

This simple procedure can be extended to any function that can be repre-
sented as a sum of cosine (and sine) functions, even if that summation is not
a finite sum. It turns out that the functions that can be represented as sums
in this form are very general, and include most of the periodic functions that
are usually encountered in applications.

723
724 10 Fourier Series

Periodic Functions

A function 𝑓 is said to be periodic with period 𝑝 > 0 if

𝑓 (𝑡 + 𝑝) = 𝑓 (𝑡)

for all 𝑡 in the domain of 𝑓 . This means that the graph of 𝑓 repeats in
successive intervals of length 𝑝, as can be seen in the graph in Figure 10.1.

𝑝 2𝑝 3𝑝 4𝑝 5𝑝

Fig. 10.1 An example of a periodic function with period 𝑝. Notice how the graph repeats
on each interval of length 𝑝.

The functions sin 𝑡 and cos 𝑡 are periodic with period 2𝜋, while tan 𝑡 is
periodic with period 𝜋 since

sin(𝑡 + 𝜋) − sin 𝑡
tan(𝑡 + 𝜋) = = = tan 𝑡.
cos(𝑡 + 𝜋) − cos 𝑡

The constant function 𝑓 (𝑡) = 𝑐 is periodic with period 𝑝 where 𝑝 is any

positive number since
𝑓 (𝑡 + 𝑝) = 𝑐 = 𝑓 (𝑡).
Other examples of periodic functions are the square wave and triangular wave
whose graphs are shown in Figure 10.2. Both are periodic with period 2.

𝑦 𝑦

1 1

𝑡 𝑡
−1 0 1 2 3 4 −2 −1 0 1 2 3 4 5
(a) Square wave sw(𝑡) (b) Triangular wave tw(𝑡)

Fig. 10.2
10.1 Periodic Functions and Orthogonality Relations 725

Since a periodic function of period 𝑝 repeats over any interval of length 𝑝,

it is possible to define a periodic function by giving the formula for 𝑓 on an
interval of length 𝑝, and repeating this in subsequent intervals of length 𝑝.
For example, the square wave sw(𝑡) and triangular wave tw(𝑡) from Figure
10.2 are described by
{
0 if −1 ≤ 𝑡 < 0
sw(𝑡) = ; sw(𝑡 + 2) = sw(𝑡).
1 if 0 ≤ 𝑡 < 1
{
−𝑡 if −1 ≤ 𝑡 < 0
tw(𝑡) = ; tw(𝑡 + 2) = tw(𝑡).
𝑡 if 0 ≤ 𝑡 < 1

There is not a unique period for a periodic function. Note that if 𝑝 is a

period of 𝑓 (𝑡), then 2𝑝 is also a period because

𝑓 (𝑡 + 2𝑝) = 𝑓 ((𝑡 + 𝑝) + 𝑝) = 𝑓 (𝑡 + 𝑝) = 𝑓 (𝑡)

for all 𝑡. In fact, a similar argument shows that 𝑛𝑝 is also a period for any
positive integer 𝑛. Thus 2𝑛𝜋 is a period for sin 𝑡 and cos 𝑡 for all positive
integers 𝑛.
If 𝑃 > 0 is a period of 𝑓 and there is no smaller period then we say 𝑃 is the
fundamental period of 𝑓 although we will usually just say the period. Not
all periodic functions have a smallest period. The constant function 𝑓 (𝑡) = 𝑐
is an example of such a function since any positive 𝑝 is a period. The funda-
mental period of the sine and cosine functions is 2𝜋, while the fundamental
period of the square wave and triangular wave from Figure 10.2 is 2.

Periodic functions under scaling

If 𝑓 (𝑡) is periodic of period 𝑝 and 𝑎 is any positive number let 𝑔(𝑡) = 𝑓 (𝑎𝑡).
Then for all 𝑡
( 𝑝) ( ( 𝑝 ))
𝑔 𝑡+ = 𝑓 𝑎 𝑡+ = 𝑓 (𝑎𝑡 + 𝑝) = 𝑓 (𝑎𝑡) = 𝑔(𝑡).
𝑎 𝑎
Thus
𝑝
If 𝑓 (𝑡) is periodic with period 𝑝, then 𝑓 (𝑎𝑡) is periodic with period .
𝑎

It is also true that if 𝑃 is the fundamental period for the periodic function
𝑓 (𝑡), then the fundamental period of 𝑔(𝑡) = 𝑓 (𝑎𝑡) is 𝑃/𝑎. To see this it is only
necessary to verify that any period 𝑟 of 𝑔(𝑡) is at least as large as 𝑃/𝑎, which
is already a period as observed above. But if 𝑟 is a period for 𝑔(𝑡), then 𝑟𝑎 is
a period for 𝑔(𝑡/𝑎) = 𝑓 (𝑡) so that 𝑟𝑎 ≥ 𝑃 , or 𝑟 ≥ 𝑃/𝑎.
726 10 Fourier Series

Applying these observations to the functions sin 𝑡 and cos 𝑡 with funda-
mental period 2𝜋 gives the following facts.
Theorem 1. For any 𝑎 > 0 the functions cos 𝑎𝑡 and sin 𝑎𝑡 are periodic with
period 2𝜋/𝑎. In particular, if 𝐿 > 0 then the functions
𝑛𝜋 𝑛𝜋
cos 𝑡 and sin 𝑡, 𝑛 = 1, 2, 3, . . . .
𝐿 𝐿
are periodic with fundamental period 𝑃 = 2𝐿/𝑛.
Note that since the fundamental period of the functions cos 𝑛𝜋 𝑛𝜋
𝐿 𝑡 and sin 𝐿 𝑡
is 𝑃 = 2𝐿/𝑛, it follows that 2𝐿 = 𝑛𝑃 is also a period for each of these
functions. Thus, a sum
∞
∑ ( 𝑛𝜋 𝑛𝜋 )
𝑎𝑛 cos 𝑡 + 𝑏𝑛 sin 𝑡
𝑛=1
𝐿 𝐿

will be periodic of period 2𝐿.

Notice that if 𝑛 is a positive integer, then cos 𝑛𝑡 and sin 𝑛𝑡 are periodic
with period 2𝜋/𝑛. Thus, each period of cos 𝑡 or sin 𝑡 contains 𝑛 periods of
cos 𝑛𝑡 and sin 𝑛𝑡. This means that the functions cos 𝑛𝑡 and sin 𝑛𝑡 oscillate
more rapidly as 𝑛 increases, as can be seen in Figure 10.3 for 𝑛 = 3.

1 1

−𝜋 𝜋 −𝜋 𝜋

−1 −1

(a) The graphs of cos 𝑡 and cos 3𝑡. (b) The graphs of sin 𝑡 and sin 3𝑡.

Fig. 10.3

Example 2. Find the fundamental period of each of the following periodic

functions.
1. cos 2𝑡
2. sin 32 (𝑡 − 𝜋)
3. 1 + cos 𝑡 + cos 2𝑡
4. sin 2𝜋𝑡 + sin 3𝜋𝑡

▶ Solution. 1. 𝑃 = 2𝜋/2 = 𝜋.
2. sin 32 (𝑡 − 𝜋) = sin( 32 𝑡 − 23 𝜋) = sin 32 𝑡 cos 32 𝜋 − cos 32 𝑡 sin 32 𝜋 = cos 32 𝑡. Thus,
𝑃 = 2𝜋/(3/2) = 4𝜋/3.
10.1 Periodic Functions and Orthogonality Relations 727

3. The constant function 1 is periodic with any period 𝑝, the fundamental

period of cos 𝑡 is 2𝜋 and all the periods are of the form 2𝑛𝜋 for a positive
integer 𝑛, and the fundamental period of cos 2𝑡 is 𝜋 with 𝑚𝜋 being all
the possible periods. Thus, the smallest number that works as a period
for all the functions is 2𝜋 and this is also the smallest period for the sum.
Hence 𝑃 = 2𝜋.
4. The fundamental period of sin 2𝜋𝑡 is 2𝜋/2𝜋 = 1 so that all of the periods
have the form 𝑛 for 𝑛 a positive integer, and the fundamental period of
sin 3𝜋𝑡 is 2𝜋/3𝜋 = 2/3 so that all of the periods have the form 2𝑚/3 for
𝑚 a positive integer. Thus, the smallest number that works as a period
for both functions is 2 and thus 𝑃 = 2.
◀

Orthogonality Relations for Sine and Cosine

The family of linearly independent functions

{ }
𝜋 2𝜋 3𝜋 𝜋 2𝜋 3𝜋
1, cos 𝑡, cos 𝑡, cos 𝑡, . . . , sin 𝑡, sin 𝑡, sin 𝑡, . . .
𝐿 𝐿 𝐿 𝐿 𝐿 𝐿

form what is called a mutually orthogonal set of functions on the interval

[−𝐿, 𝐿], analogous to a mutually perpendicular set of vectors. Two functions
𝑓 and 𝑔 defined on an interval 𝑎 ≤ 𝑡 ≤ 𝑏 are said to be orthogonal on the
interval [𝑎, 𝑏] if
∫ 𝑏
𝑓 (𝑡)𝑔(𝑡) 𝑑𝑡 = 0.
𝑎

A family of functions is mutually orthogonal on the interval [𝑎, 𝑏] if any

two distinct functions are orthogonal. The mutual orthogonality of the family
of cosine and sine functions on the interval [−𝐿, 𝐿] is a consequence of the
following identities.
Proposition 3 (Orthogonality Relations). Let 𝑚 and 𝑛 be positive in-
tegers, and let 𝐿 > 0. Then
∫ 𝐿 ∫ 𝐿
𝑛𝜋 𝑛𝜋
cos 𝑡 𝑑𝑡 = sin 𝑡 𝑑𝑡 = 0 (1)
−𝐿 𝐿 −𝐿 𝐿
∫ 𝐿
𝑛𝜋 𝑚𝜋
cos 𝑡 sin 𝑡 𝑑𝑡 = 0 (2)
−𝐿 𝐿 𝐿
∫ 𝐿 {
𝑛𝜋 𝑚𝜋 𝐿, if 𝑛 = 𝑚,
cos 𝑡 cos 𝑡 𝑑𝑡 = (3)
−𝐿 𝐿 𝐿 0 if 𝑛 =
∕ 𝑚.
∫ 𝐿 {
𝑛𝜋 𝑚𝜋 𝐿, if 𝑛 = 𝑚,
sin 𝑡 sin 𝑡 𝑑𝑡 = (4)
−𝐿 𝐿 𝐿 0 if 𝑛 =
∕ 𝑚.
728 10 Fourier Series

Proof. For (1):

∫ 𝐿 𝐿
𝑛𝜋 𝐿 𝑛𝜋  𝐿
cos 𝑡 𝑑𝑡 = sin 𝑡 = (sin 𝑛𝜋 − sin(−𝑛𝜋)) = 0.
−𝐿 𝐿 𝑛𝜋 𝐿 −𝐿 𝑛𝜋

For (3) with 𝑛 ∕= 𝑚, use the identity

1
cos 𝐴 cos 𝐵 = (cos(𝐴 + 𝐵) + cos(𝐴 − 𝐵)),
2
to get
∫ 𝐿 ∫ 𝐿 ( )
𝑛𝜋 𝑚𝜋 1 (𝑛 + 𝑚)𝜋 (𝑛 − 𝑚)𝜋
cos 𝑡 cos 𝑡 𝑑𝑡 = cos 𝑡 + cos 𝑡 𝑑𝑡
−𝐿 𝐿 𝐿 −𝐿 2 𝐿 𝐿
( )𝐿
𝐿 (𝑛 + 𝑚)𝜋 𝐿 (𝑛 − 𝑚)𝜋 
= sin 𝑡+ sin 𝑡  = 0.
2(𝑛 + 𝑚)𝜋 𝐿 2(𝑛 − 𝑚)𝜋 𝐿 −𝐿

For (3) with 𝑛 = 𝑚, use the identity cos2 𝐴 = (1 + cos 2𝐴)/2 to get
∫ 𝐿 ∫ 𝐿(
𝑛𝜋 𝑚𝜋 𝑛𝜋 )2
cos 𝑡 cos 𝑡 𝑑𝑡 = cos 𝑡 𝑑𝑡 =
−𝐿 𝐿 𝐿 −𝐿 𝐿
∫ 𝐿 ( ) ( )𝐿
1 2𝑛𝜋 1 𝐿 2𝑛𝜋 
= 1 + cos 𝑡 𝑑𝑡 = 𝑡+ sin 𝑡  = 𝐿.
−𝐿 2 𝐿 2 2𝑛𝜋 𝐿 −𝐿

The proof of (4) is similar, making use of the identities sin2 𝐴 = (1−cos 2𝐴)/2
in case 𝑛 = 𝑚 and
1
sin 𝐴 sin 𝐵 = (cos(𝐴 − 𝐵) − cos(𝐴 + 𝐵))
2
in case 𝑛 ∕= 𝑚. The proof of (2) is left as an exercise.

Even and Odd Functions

A function 𝑓 defined on a symmetric interval [−𝐿, 𝐿] is even if 𝑓 (−𝑡) = 𝑓 (𝑡)

for −𝐿 ≤ 𝑡 ≤ 𝐿, and 𝑓 is odd if 𝑓 (−𝑡) = −𝑓 (𝑡) for −𝐿 ≤ 𝑡 ≤ 𝐿.

Example 4. Determine whether each of the following functions is even,

odd, or neither.
1. 𝑓 (𝑡) = 3𝑡2 + cos 5𝑡
2. 𝑔(𝑡) = 3𝑡 − 𝑡2 sin 2𝑡
3. ℎ(𝑡) = 𝑡2 + 𝑡 + 1

▶ Solution. 1. Since 𝑓 (−𝑡) = 3(−𝑡)2 + cos 5(−𝑡) = 3𝑡2 + cos 5𝑡 = 𝑓 (𝑡) for
all 𝑡, it follows that 𝑓 is an even function.
10.1 Periodic Functions and Orthogonality Relations 729

2. Since 𝑔(−𝑡) = 3(−𝑡) − (−𝑡)2 sin 3(−𝑡) = −3𝑡 + 𝑡2 sin 3𝑡 = −𝑔(𝑡) for all 𝑡,
it follows that 𝑔 is an odd function.
3. Since ℎ(−𝑡) = (−𝑡)2 + (−𝑡) + 1 = 𝑡2 − 𝑡 + 1 = 𝑡2 + 𝑡 + 1 = ℎ(𝑡) ⇐⇒
−𝑡 = 𝑡 ⇐⇒ 𝑡 = 0, we conclude that ℎ is not even. Similarly, ℎ(−𝑡) =
−ℎ(𝑡) ⇐⇒ 𝑡2 −𝑡+1 = −𝑡2 −𝑡−1 ⇐⇒ 𝑡2 +1 = (−𝑡2 +1) ⇐⇒ 𝑡2 +1 = 0,
which is not true for any 𝑡. Thus ℎ is not odd, and hence it is neither
even or odd.
◀

The graph of an even function is symmetric with respect to the 𝑦-axis, while
the graph of an odd function is symmetric with respect to the origin, as
illustrated in Figure 10.4.

𝑓 (𝑡)
𝑓 (−𝑡) 𝑓 (𝑡)
−𝑡
−𝑡 𝑡 𝑡

𝑓 (−𝑡) = −𝑓 (𝑡)

(a) The graph of an even function. (b) The graph of an odd function.

Fig. 10.4

Here is a list of basic properties of even and odd functions that are useful
in applications to Fourier series. All of them follow easily from the definitions,
and the verifications will be left to the exercises.
Proposition 5. Suppose that 𝑓 and 𝑔 are functions defined on the interval
−𝐿 ≤ 𝑡 ≤ 𝐿.
1. If both 𝑓 and 𝑔 are even then 𝑓 + 𝑔 and 𝑓 𝑔 are even.
2. If both 𝑓 and 𝑔 are odd, then 𝑓 + 𝑔 is odd and 𝑓 𝑔 is even.
3. If 𝑓 is even and 𝑔 is odd, then 𝑓 𝑔 is odd.
4. If 𝑓 is even, then
∫ 𝐿 ∫ 𝐿
𝑓 (𝑡) 𝑑𝑡 = 2 𝑓 (𝑡) 𝑑𝑡.
−𝐿 0

5. If 𝑓 is odd, then
∫ 𝐿
𝑓 (𝑡) 𝑑𝑡 = 0.
−𝐿

Since the integral of 𝑓 computes the signed area under the graph of 𝑓 , the
integral equations can be seen from the graphs of even and odd functions in
Figure 10.4.
730 10 Fourier Series

Exercises

1–9. Graph each of the following periodic functions. Graph at least 3 periods.
{
3 if 0 < 𝑡 < 3
1. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 6) = 𝑓 (𝑡).
−3 if −3 < 𝑡 < 0
⎧
⎨−3 if −2 ≤ 𝑡 < −1

2. 𝑓 (𝑡) = 0 if −1 ≤ 𝑡 ≤ 1 ; 𝑓 (𝑡 + 4) = 𝑓 (𝑡).

⎩
3 if 1 < 𝑡 < 2

3. 𝑓 (𝑡) = 𝑡, 0 < 𝑡 ≤ 2; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

4. 𝑓 (𝑡) = 𝑡, −1 < 𝑡 ≤ 1; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

5. 𝑓 (𝑡) = sin 𝑡, 0 < 𝑡 ≤ 𝜋; 𝑓 (𝑡 + 𝜋) = 𝑓 (𝑡).

{
0 if −𝜋 ≤ 𝑡 < 0
6. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).
sin 𝑡 if 0 < 𝑡 ≤ 𝜋
{
−𝑡 if −1 ≤ 𝑡 < 0
7. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).
1 if 0 ≤ 𝑡 < 1

8. 𝑓 (𝑡) = 𝑡2 , −1 < 𝑡 ≤ 1; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

9. 𝑓 (𝑡) = 𝑡2 , 0 < 𝑡 ≤ 2; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

10–17. Determine if the given function is periodic. If it is periodic find the

fundamental period.
10. 1

11. sin 2𝑡

12. 1 + cos 3𝜋𝑡

13. cos 2𝑡 + sin 3𝑡

14. 𝑡 + sin 2𝑡

15. sin2 𝑡

16. cos 𝑡 + cos 𝜋𝑡

17. sin 𝑡 + sin 2𝑡 + sin 3𝑡

18–26. Determine if the given function is even, odd, or neither.

10.2 Fourier Series 731

18. 𝑓 (𝑡) = ∣𝑡∣

19. 𝑓 (𝑡) = 𝑡 ∣𝑡∣

20. 𝑓 (𝑡) = sin2 𝑡

21. 𝑓 (𝑡) = cos2 𝑡

22. 𝑓 (𝑡) = sin 𝑡 sin 3𝑡

23. 𝑓 (𝑡) = 𝑡2 + sin 𝑡

24. 𝑓 (𝑡) = 𝑡 + ∣𝑡∣

25. 𝑓 (𝑡) = ln ∣cos 𝑡∣

26. 𝑓 (𝑡) = 5𝑡 + 𝑡2 sin 3𝑡

27. Verify the orthogonality property (2) from Proposition 3:

∫ 𝐿
𝑛𝜋 𝑚𝜋
cos 𝑡 sin 𝑡 𝑑𝑡 = 0
−𝐿 𝐿 𝐿

28. Use the properties of even and odd functions (Proposition 10.4) to eval-
uate the following integrals.

∫ 1 ∫ 1
(a) 𝑡 𝑑𝑡 (b) 𝑡4 𝑑𝑡
−1 −1
∫ 𝜋 ∫ 𝜋
(c) 𝑡 sin 𝑡 𝑑𝑡 (d) 𝑡 cos 𝑡 𝑑𝑡
−𝜋 −𝜋
∫ 𝜋 ∫ 𝜋
𝑛𝜋 𝑚𝜋
(e) cos 𝑡 sin 𝑡 𝑑𝑡 (f) 𝑡2 sin 𝑡 𝑑𝑡
−𝜋 𝐿 𝐿 −𝜋

10.2 Fourier Series

We start by considering the possibility of representing a function 𝑓 as a sum

of a series of the form
∞ ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓 (𝑡) = + 𝑎𝑛 cos + 𝑏𝑛 sin (1)
2 𝑛=1
𝐿 𝐿
732 10 Fourier Series

where the coefficients 𝑎0 , 𝑎1 , . . ., 𝑏1 , 𝑏2 , . . ., are to be determined. Since the

individual terms in the series (1) are periodic with periods 2𝐿, 2𝐿/2, 2𝐿/3,
. . ., the function 𝑓 (𝑡) determined by the sum of the series, where it converges,
must be periodic with period 2𝐿. This means that only periodic functions of
period 2𝐿 can be represented by a series of the form (1). Our first problem is
to find the coefficients 𝑎𝑛 and 𝑏𝑛 in the series (1). The first term of the series
is written 𝑎0 /2, rather than simply as 𝑎0 , to make the formula to be derived
below the same for all 𝑎𝑛 , rather than a special case for 𝑎0 .
The coefficients 𝑎𝑛 and 𝑏𝑛 can be found from the orthogonality relations
of the family of functions cos(𝑛𝜋𝑡/𝐿) and sin(𝑛𝜋𝑡/𝐿) on the interval [−𝐿, 𝐿]
given in Proposition 3 of Sect. 10.1. To compute the coefficient 𝑎𝑛 for 𝑛 = 1, 2,
3, . . ., multiply both sides of the series (1) by cos(𝑚𝜋𝑡/𝐿), with 𝑚 a positive
integer and then integrate from −𝐿 to 𝐿. For the moment we will assume
that the integrals exist and that it is justified to integrate term by term. Then
using (1), (2), and (3) from Sect. 10.1, we get
∫ 𝐿 ∫ 𝐿
𝑚𝜋 𝑎0 𝑚𝜋
𝑓 (𝑡) cos 𝑡 𝑑𝑡 = cos 𝑡 𝑑𝑡
−𝐿 𝐿 2 −𝐿 𝐿
| {z }
=0
∞ [ ∫ 𝐿 ∫ 𝐿 ]
∑ 𝑛𝜋 𝑚𝜋 𝑛𝜋 𝑚𝜋
+ 𝑎𝑛 cos 𝑡 cos 𝑡 𝑑𝑡 +𝑏𝑛 sin 𝑡 cos 𝑡 𝑑𝑡 = 𝑎𝑚 𝐿.
𝑛=1 −𝐿 𝐿 𝐿 −𝐿 𝐿 𝐿
| {z } | {z }
=0
⎧
⎨0

if 𝑛 ∕= 𝑚
=
⎩𝐿

if 𝑛 = 𝑚

Thus,
∫ 𝐿
1 𝑚𝜋
𝑎𝑚 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡, 𝑚 = 1, 2, 3, . . .
𝐿 −𝐿 𝐿
or, replacing the index 𝑚 by 𝑛,
∫ 𝐿
1 𝑛𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡, 𝑛 = 1, 2, 3, . . . (2)
𝐿 −𝐿 𝐿

To compute 𝑎0 , integrate both sides of (1) from −𝐿 to 𝐿 to get

∫ 𝐿 ∫ 𝐿 ∞ [ ∫ 𝐿 ∫ 𝐿 ]
𝑎0 ∑ 𝑛𝜋 𝑛𝜋
𝑓 (𝑡) 𝑑𝑡 = 𝑑𝑡 + 𝑎𝑛 cos 𝑡 𝑑𝑡 +𝑏𝑛 sin 𝑡 𝑑𝑡
−𝐿 2 −𝐿 −𝐿 𝐿 −𝐿 𝐿
| {z } 𝑛=1 | {z } | {z }
=2𝐿 =0 =0
= 𝑎0 𝐿.

Thus,
∫ 𝐿
1
𝑎0 = 𝑓 (𝑡) 𝑑𝑡. (3)
𝐿 −𝐿
10.2 Fourier Series 733

Hence, 𝑎0 is two times the average value of the function 𝑓 (𝑡) over the interval
−𝐿 ≤ 𝑡 ≤ 𝐿. Observe that the value of 𝑎0 is obtained from (2) by setting
𝑛 = 0. Of course, if the constant 𝑎0 in (1) were not divided by 2, we would
need a separate formula for 𝑎0 . It is for this reason that the constant term
in (1) is labeled 𝑎0 /2. Thus, for all 𝑛 ≥ 0, the coefficients 𝑎𝑛 are given by a
single formula
∫ 𝐿
1 𝑛𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡, 𝑛 = 0, 1, 2, . . . (4)
𝐿 −𝐿 𝐿

To compute 𝑏𝑛 for 𝑛 = 1, 2, 3, . . ., multiply both sides of the series (1)

by sin(𝑚𝜋𝑡/𝐿), with 𝑚 a positive integer and then integrate from −𝐿 to 𝐿.
Then using (1), (2), and (4) from Sect. 10.1, we get
∫ 𝐿 ∫ 𝐿
𝑚𝜋 𝑎0 𝑚𝜋
𝑓 (𝑡) sin 𝑡 𝑑𝑡 = sin 𝑡 𝑑𝑡
−𝐿 𝐿 2 −𝐿 𝐿
| {z }
=0
∞ [ ∫ 𝐿 ∫ 𝐿 ]
∑ 𝑛𝜋 𝑚𝜋 𝑛𝜋 𝑚𝜋
+ 𝑎𝑛 cos 𝑡 sin 𝑡 𝑑𝑡 +𝑏𝑛 sin 𝑡 sin 𝑡 𝑑𝑡 = 𝑏𝑚 𝐿.
𝐿 𝐿 𝐿 𝐿
𝑛=1 | −𝐿 {z } | −𝐿 ⎧ {z }
=0
⎨0

if 𝑛 ∕= 𝑚
=
⎩𝐿

if 𝑛 = 𝑚

Thus, replacing the index 𝑚 by 𝑛, we find that

∫ 𝐿
1 𝑛𝜋
𝑏𝑛 = 𝑓 (𝑡) sin 𝑡 𝑑𝑡, 𝑛 = 1, 2, 3, . . . (5)
𝐿 −𝐿 𝐿

We have arrived at what are known as the Euler Formulas for a function
𝑓 (𝑡) that is the sum of a trigonometric series as in (1):

∫ 𝐿
1
𝑎0 = 𝑓 (𝑡) 𝑑𝑡 (6)
𝐿 −𝐿
∫ 𝐿
1 𝑛𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡, 𝑛 = 1, 2, 3, . . . (7)
𝐿 −𝐿 𝐿
∫ 𝐿
1 𝑛𝜋
𝑏𝑛 = 𝑓 (𝑡) sin 𝑡 𝑑𝑡, 𝑛 = 1, 2, 3, . . . (8)
𝐿 −𝐿 𝐿

The numbers 𝑎𝑛 and 𝑏𝑛 are known as the Fourier coefficients of the func-
tion 𝑓 . Note that while we started with a periodic function of period 2𝐿, the
formulas for 𝑎𝑛 and 𝑏𝑛 only use the values of 𝑓 (𝑡) on the interval [−𝐿, 𝐿].
734 10 Fourier Series

We can then reverse the process, and start with any function 𝑓 (𝑡) defined on
the symmetric interval [−𝐿, 𝐿] and use the Euler formulas to determine a
trigonometric series. We will write
∞ ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓 (𝑡) ∼ + 𝑎𝑛 cos + 𝑏𝑛 sin , (9)
2 𝑛=1
𝐿 𝐿

where the 𝑎𝑛 , 𝑏𝑛 are defined by (6), (7), and (8), to indicate that the right
hand side of (9) is the Fourier series of the function 𝑓 (𝑡) defined on [−𝐿, 𝐿].
Note that the symbol ∼ indicates that the trigonometric series on the right
of (9) is associated with the function 𝑓 (𝑡); it does not imply that the series
converges to 𝑓 (𝑡) for any value of 𝑡. In fact, there are functions whose Fourier
series do not converge to the function. Of course, we will be interested in the
conditions under which the Fourier series of 𝑓 (𝑡) converges to 𝑓 (𝑡), in which
case ∼ can be replaced by =; but for now we associate a specific series using
Equations (6), (7), and (8) with 𝑓 (𝑡) and call it the Fourier series. The mild
conditions under which the Fourier series of 𝑓 (𝑡) converges to 𝑓 (𝑡) will be
considered in the next section.
Remark 1. If an integrable function 𝑓 (𝑡) is periodic with period 𝑝, then the
integral of 𝑓 (𝑡) over any interval of length 𝑝 is the same; that is
∫ 𝑐+𝑝 ∫ 𝑝
𝑓 (𝑡) 𝑑𝑡 = 𝑓 (𝑡) 𝑑𝑡 (10)
𝑐 0

for any choice of 𝑐. To see this, first observe that for any 𝛼 and 𝛽, if we use
the change of variables 𝑡 = 𝑥 − 𝑝, then

∫ 𝛽 ∫ 𝛽+𝑝 ∫ 𝛽+𝑝 ∫ 𝛽+𝑝

𝑓 (𝑡) 𝑑𝑡 = 𝑓 (𝑥 − 𝑝) 𝑑𝑥 = 𝑓 (𝑥) 𝑑𝑥 = 𝑓 (𝑡) 𝑑𝑡.
𝛼 𝛼+𝑝 𝛼+𝑝 𝛼+𝑝

Letting 𝛼 = 𝑐 and 𝛽 = 0 gives

∫ 0 ∫ 𝑝
𝑓 (𝑡) 𝑑𝑡 = 𝑓 (𝑡) 𝑑𝑡
𝑐 𝑐+𝑝

so that
∫ 𝑐+𝑝 ∫ 0 ∫ 𝑐+𝑝
𝑓 (𝑡) 𝑑𝑡 = 𝑓 (𝑡) 𝑑𝑡 + 𝑓 (𝑡) 𝑑𝑡
𝑐 𝑐 0
∫ 𝑝 ∫ 𝑐+𝑝 ∫ 𝑝
= 𝑓 (𝑡) 𝑑𝑡 + 𝑓 (𝑡) 𝑑𝑡 = 𝑓 (𝑡) 𝑑𝑡,
𝑐+𝑝 0 0

which is (10). This formula means that when computing the Fourier coeffi-
cients, the integrals can be computed over any convenient interval of length
2𝐿. For example,
10.2 Fourier Series 735
∫ 𝐿 ∫ 2𝐿
1 𝑛𝜋 1 𝑛𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡.
𝐿 −𝐿 𝐿 𝐿 0 𝐿

We now consider some examples of the calculation of Fourier series.

Example 2. Compute the Fourier series of the odd square wave function
of period 2𝐿 and amplitude 1 given by
{
−1 −𝐿 ≤ 𝑡 < 0,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝐿) = 𝑓 (𝑡).
1 0 ≤ 𝑡 < 𝐿,

See Figure 10.5 for the graph of 𝑓 (𝑡).

𝑦
1

𝑡
−3𝐿 −2𝐿 −𝐿 𝐿 2𝐿 3𝐿
−1

Fig. 10.5 The odd square wave of period 2𝐿

▶ Solution. Use the Euler formulas for 𝑎𝑛 (Equations (6) and (7)) to con-
clude ∫
1 𝐿 𝑛𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡 = 0
𝐿 −𝐿 𝐿
for all 𝑛 ≥ 0. This is because the function 𝑓 (𝑡) cos 𝑛𝜋
𝐿 𝑡 is the product of an
odd and even function, and hence is odd, which implies by Proposition 5
(Part 5) of Section 10.1 that the integral is 0. It remains to compute the
coefficients 𝑏𝑛 from (8).
∫ 𝐿 ∫ 0 ∫ 𝐿
1 𝑛𝜋 1 𝑛𝜋 1 𝑛𝜋
𝑏𝑛 = 𝑓 (𝑡) sin 𝑡 𝑑𝑡 = 𝑓 (𝑡) sin 𝑡 𝑑𝑡 + 𝑓 (𝑡) sin 𝑡 𝑑𝑡
𝐿 −𝐿 𝐿 𝐿 −𝐿 𝐿 𝐿 0 𝐿
∫ 0 ∫ 𝐿
1 𝑛𝜋 1 𝑛𝜋
= (−1) sin
𝑡 𝑑𝑡 + (+1) sin 𝑡 𝑑𝑡
𝐿−𝐿 𝐿 𝐿 0 𝐿
{[ ]0 [ ]𝐿 }
1 𝐿 𝑛𝜋 𝐿 𝑛𝜋
= cos 𝑡 − cos 𝑡
𝐿 𝑛𝜋 𝐿 −𝐿 𝑛𝜋 𝐿 0
1
= [(1 − cos(−𝑛𝜋)) + (1 − cos(𝑛𝜋))]
𝑛𝜋
2 2
= (1 − cos 𝑛𝜋) = (1 − (−1)𝑛 ).
𝑛𝜋 𝑛𝜋
Therefore,
736 10 Fourier Series
{
0 if 𝑛 is even,
𝑏𝑛 = 4
𝑛𝜋 if 𝑛 is odd,
and the Fourier series is
( )
4 𝜋 1 3𝜋 1 5𝜋 1 7𝜋
𝑓 (𝑡) ∼ sin 𝑡 + sin 𝑡 + sin 𝑡 + sin 𝑡+ ⋅⋅⋅ . (11)
𝜋 𝐿 3 𝐿 5 𝐿 7 𝐿
◀

Example 3. Compute the Fourier series of the even square wave function
of period 2𝐿 and amplitude 1 given by
⎧
⎨−1 −𝐿 ≤ 𝑡 < −𝐿/2,

𝑓 (𝑡) = 1 −𝐿/2 ≤ 𝑡 < 𝐿/2, ; 𝑓 (𝑡 + 2𝐿) = 𝑓 (𝑡).

⎩
−1 𝐿/2 ≤ 𝑡 < 𝐿,

See Figure 10.6 for the graph of 𝑓 (𝑡).

1𝑦

𝑡
− 5𝐿
2
− 3𝐿
2
−𝐿
2
𝐿
2
3𝐿
2
5𝐿
2
−1

Fig. 10.6 The even square wave of period 2𝐿

▶ Solution. Use the Euler formulas for 𝑏𝑛 (Equation (8)) to conclude

∫ 𝐿
1 𝑛𝜋
𝑏𝑛 = 𝑓 (𝑡) sin 𝑡 𝑑𝑡 = 0,
𝐿 −𝐿 𝐿

for all 𝑛 ≥ 1. As in the previous example, this is because the function

𝑓 (𝑡) sin 𝑛𝜋
𝐿 𝑡 is the product of an even and odd function, and hence is odd,
which implies by Proposition 5 (Part 5) of Section 10.1 that the integral is
0. It remains to compute the coefficients 𝑎𝑛 from (6) and (7).
For 𝑛 = 0, 𝑎0 is twice the average of 𝑓 (𝑡) over the period [−𝐿, 𝐿], which
is easily seen to be 0 from the graph of 𝑓 (𝑡). For 𝑛 ≥ 1,
∫ 𝐿
1 𝑛𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡
𝐿 −𝐿 𝐿
∫−𝐿/2 ∫ 𝐿/2 ∫ 𝐿
1 𝑛𝜋 1 𝑛𝜋 1 𝑛𝜋
= 𝑓 (𝑡) cos 𝑡 𝑑𝑡 + 𝑓 (𝑡) cos 𝑡 𝑑𝑡 + 𝑓 (𝑡) cos 𝑡 𝑑𝑡
𝐿 −𝐿 𝐿 𝐿 −𝐿/2 𝐿 𝐿 𝐿/2 𝐿
10.2 Fourier Series 737
∫ ∫ ∫
1 −𝐿/2
𝑛𝜋 1 𝐿/2 𝑛𝜋 1 𝐿 𝑛𝜋
= (−1) cos𝑡 𝑑𝑡 + (+1) cos 𝑡 𝑑𝑡 + (−1) cos 𝑡 𝑑𝑡
𝐿 −𝐿 𝐿 𝐿 −𝐿/2 𝐿 𝐿 𝐿/2 𝐿
{ [ ]−𝐿/2 [ ]𝐿/2 [ ]𝐿 }
1 𝐿 𝑛𝜋 𝐿 𝑛𝜋 𝐿 𝑛𝜋
= − sin 𝑡 + sin 𝑡 − sin 𝑡
𝐿 𝑛𝜋 𝐿 −𝐿 𝑛𝜋 𝐿 −𝐿/2 𝑛𝜋 𝐿 𝐿/2
1 [ ( 𝑛𝜋 ) ( 𝑛𝜋 ) ( 𝑛𝜋 ) ( 𝑛𝜋 )]
= − sin − + sin(−𝑛𝜋) + sin − sin − − sin 𝑛𝜋 + sin
𝑛𝜋 2 2 2 2
4 𝑛𝜋
= sin .
𝑛𝜋 2
Therefore, ⎧
⎨0
 if 𝑛 is even,
𝑎𝑛 = 4
if 𝑛 = 4𝑚 + 1
 𝑛𝜋
⎩ 4
− 𝑛𝜋 if 𝑛 = 4𝑚 + 3
and the Fourier series is
( )
4 𝜋 1 3𝜋 1 5𝜋 1 7𝜋
𝑓 (𝑡) ∼ cos 𝑡 − cos 𝑡 + cos 𝑡 − cos 𝑡 + ⋅⋅⋅
𝜋 𝐿 3 𝐿 5 𝐿 7 𝐿
∞ 𝑘
4 ∑ (−1) (2𝑘 + 1)𝜋
= cos 𝑡.
𝜋 2𝑘 + 1 𝐿
𝑘=0

Example 4. Compute the Fourier series of the even triangular wave func-
tion of period 2𝜋 given by
{
−𝑡 −𝜋 ≤ 𝑡 < 0,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).
𝑡 0 ≤ 𝑡 < 𝜋,

See Figure 10.7 for the graph of 𝑓 (𝑡).

−3𝜋 −2𝜋 −𝜋 0 𝜋 2𝜋 3𝜋

Fig. 10.7 The even triangular wave of period 2𝜋.

▶ Solution. The period is 2𝜋 = 2𝐿 so 𝐿 = 𝜋. Again, since the function

𝑓 (𝑡) is even, the coefficients 𝑏𝑛 = 0. It remains to compute the coefficients 𝑎𝑛
from the Euler formulas (6) and (7).
For 𝑛 = 0, using the fact that 𝑓 (𝑡) is even,
738 10 Fourier Series
∫ 𝜋 ∫ 𝜋
1 2
𝑎0 = 𝑓 (𝑡) 𝑑𝑡 =
𝑓 (𝑡) 𝑑𝑡
𝜋
−𝜋 𝜋 0
∫ 𝜋
2 𝜋 2 𝑡2 
= 𝑡 𝑑𝑡 = = 𝜋.
𝜋 0 𝜋 2 0

For 𝑛 ≥ 1, using the fact that 𝑓 (𝑡) is even, and taking advantage of the
integration by parts formula
∫
𝑥 cos 𝑥 𝑑𝑥 = 𝑥 sin 𝑥 + cos 𝑥 + 𝐶,

∫ ∫
1 𝜋 2 𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑛𝑡 𝑑𝑡 = 𝑓 (𝑡) cos 𝑛𝑡 𝑑𝑡
𝜋 −𝜋 𝜋 0
∫ ( )
2 𝜋 𝑥 𝑑𝑥
= 𝑡 cos 𝑛𝑡 𝑑𝑡 let 𝑥 = 𝑛𝑡 so 𝑡 = and 𝑑𝑡 =
𝜋 0 𝑛 𝑛
∫
2 𝑛𝜋 𝑥 𝑑𝑥 2 𝑥=𝑛𝜋
= cos 𝑥 = 2 [𝑥 sin 𝑥 + cos 𝑥]𝑥=0
𝜋 0 𝑛 𝑛 𝑛 𝜋
2 2
= 2 [cos 𝑛𝜋 − 1] = 2 [(−1)𝑛 − 1]
𝑛 𝜋 𝑛 𝜋

Therefore, {
0 if 𝑛 is even,
𝑎𝑛 =
− 𝑛42 𝜋 if 𝑛 is odd
and the Fourier series is
( )
𝜋 4cos 𝑡 cos 3𝑡 cos 5𝑡 cos 7𝑡
𝑓 (𝑡) ∼ − + + + + ⋅⋅⋅
2 𝜋 12 32 52 72
∞
𝜋 4 ∑ cos(2𝑘 + 1)𝑡
= − .
2 𝜋 (2𝑘 + 1)2
𝑘=0

Example 5. Compute the Fourier series of the sawtooth wave function of

period 2𝐿 given by

𝑓 (𝑡) = 𝑡 for −𝐿 ≤ 𝑡 < 𝐿; 𝑓 (𝑡 + 2𝐿) = 𝑓 (𝑡).

See Figure 10.8 for the graph of 𝑓 (𝑡).

▶ Solution. As in Example 2, the function 𝑓 (𝑡) is odd, so the cosine terms

𝑎𝑛 are all 0. Now compute the coefficients 𝑏𝑛 from (8). Using the integration
by parts formula
10.2 Fourier Series 739

𝑡
−3𝐿 −2𝐿 −𝐿 𝐿 2𝐿 3𝐿

−𝐿

Fig. 10.8 The sawtooth wave of period 2𝐿

∫
𝑥 sin 𝑥 𝑑𝑥 = sin 𝑥 − 𝑥 cos 𝑥 + 𝐶,

∫ 𝐿
1 𝑛𝜋
𝑏𝑛 = 𝑓 (𝑡) sin 𝑡 𝑑𝑡
𝐿 −𝐿 𝐿
∫ ( )
2 𝐿 𝑛𝜋 𝑛𝜋 𝐿 𝐿
= 𝑡 sin 𝑡 𝑑𝑡 let 𝑥 = 𝑡 so 𝑡 = 𝑥 and 𝑑𝑡 = 𝑑𝑥
𝐿 0 𝐿 𝐿 𝑛𝜋 𝑛𝜋
∫ 𝑛𝜋 ∫ 𝑛𝜋
2 𝐿 𝐿 𝑑𝑥 2𝐿
= 𝑥 sin 𝑥 = 2 2 𝑥 sin 𝑥 𝑑𝑥
𝐿 0 𝑛𝜋 𝑛𝜋 𝑛 𝜋 0
2𝐿 𝑥=𝑛𝜋
= 2 2 [sin 𝑥 − 𝑥 cos 𝑥]𝑥=0
𝑛 𝜋
2𝐿 2𝐿
= − 2 2 (𝑛𝜋 cos 𝑛𝜋) = − (−1)𝑛 .
𝑛 𝜋 𝑛𝜋
Therefore, the Fourier series is
( )
2𝐿 𝜋 1 2𝜋 1 3𝜋 1 4𝜋
𝑓 (𝑡) ∼ sin 𝑡 − sin 𝑡 + sin 𝑡 − sin 𝑡+ ⋅⋅⋅
𝜋 𝐿 2 𝐿 3 𝐿 4 𝐿
∞ 𝑛+1
2𝐿 ∑ (−1) 𝑛𝜋
= sin 𝑡.
𝜋 𝑛=1 𝑛 𝐿

All of the examples so far have been of functions that are either even or
odd. If a function 𝑓 (𝑡) is even, the resulting Fourier series will only have
cosine terms, as in the case of Examples 3 and 4, while if 𝑓 (𝑡) is odd, the
resulting Fourier series will only have sine terms, as in Examples 2 and 5.
Here are some examples where both sine an cosine terms appear.

Example 6. Compute the Fourier series of the function of period 4 given

by
740 10 Fourier Series
{
0 −2 ≤ 𝑡 < 0,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 4) = 𝑓 (𝑡).
𝑡 0 ≤ 𝑡 < 2,
See Figure 10.9 for the graph of 𝑓 (𝑡).

𝑦
2

𝑡
−6 −4 −2 2 4 6

Fig. 10.9 A half sawtooth wave of period 4

▶ Solution. This function is neither even nor odd, so we expect both sine
and cosine terms to be present. The period is 4 = 2𝐿 so 𝐿 = 2. Because
𝑓 (𝑡) = 0 on the interval (−2, 0), each of the integrals in the Euler formulas,
which should be an integral from 𝑡 = −2 to 𝑡 = 2, can be replaced with an
integral from 𝑡 = 0 to 𝑡 = 2. Thus, the Euler formulas give

∫ 2 [ ]2
1 1 𝑡2
𝑎0 = 𝑡 𝑑𝑡 = = 1;
2 0 2 2 0
∫ ( )
1 2 𝑛𝜋 𝑛𝜋 2𝑥 2𝑑𝑥
𝑎𝑛 = 𝑡 cos 𝑡 𝑑𝑡 let 𝑥 = 𝑡 so 𝑡 = and 𝑑𝑡 =
2 0 2 2 𝑛𝜋 𝑛𝜋
∫ 𝑛𝜋 ∫ 𝑛𝜋
1 2𝑥 2𝑑𝑥 2
= cos 𝑥 = 2 2 𝑥 cos 𝑥 𝑑𝑥
2 0 𝑛𝜋 𝑛𝜋 𝑛 𝜋 0
2 𝑥=𝑛𝜋
= [cos 𝑥 + 𝑥 sin 𝑥]𝑥=0
𝑛 𝜋2
2
2 2
= 2 2
(cos 𝑛𝜋 − 1) = 2 2 ((−1)𝑛 − 1).
𝑛 𝜋 𝑛 𝜋
Therefore, ⎧
⎨1
 if 𝑛 = 0,
𝑎𝑛 = 0 if 𝑛 is even and 𝑛 ≥ 2,

− 𝑛24𝜋2
⎩
if 𝑛 is odd.
Now compute 𝑏𝑛 :
∫ 2 ( )
1 𝑛𝜋 𝑛𝜋 2𝑥 2𝑑𝑥
𝑏𝑛 = 𝑡 sin 𝑡 𝑑𝑡 let 𝑥 = 𝑡 so 𝑡 = and 𝑑𝑡 =
2 0 2 2 𝑛𝜋 𝑛𝜋
10.2 Fourier Series 741
∫ 𝑛𝜋 ∫ 𝑛𝜋
1 2𝑥 2𝑑𝑥 2
= sin 𝑥 = 2 2 𝑥 sin 𝑥 𝑑𝑥
2 0 𝑛𝜋 𝑛𝜋 𝑛 𝜋 0
2
= [sin 𝑥 − 𝑥 cos 𝑥]𝑥=𝑛𝜋
𝑥=0
𝑛2 𝜋 2
2 2
= 2 2 (−𝑛𝜋 cos 𝑛𝜋) = − (−1)𝑛 .
𝑛 𝜋 𝑛𝜋
Thus,
2(−1)𝑛+1
𝑏𝑛 = for all 𝑛 ≥ 1.
𝑛𝜋
Therefore, the Fourier series is
∞ ∞
1 4 ∑ 1 (2𝑘 + 1)𝜋 2 ∑ (−1)𝑛+1 𝑛𝜋
𝑓 (𝑡) ∼ − 2 cos 𝑡 + sin 𝑡.
2 𝜋 (2𝑘 + 1)2 2 𝜋 𝑛=1 𝑛 2
𝑘=0

Example 7. Compute the Fourier series of the square pulse wave function
of period 2𝜋 given by
{
1 0 ≤ 𝑡 < ℎ,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).
0 ℎ ≤ 𝑡 < 2𝜋,

See Figure 10.10 for the graph of 𝑓 (𝑡).

−4𝜋 −2𝜋 0 ℎ 2𝜋 4𝜋

Fig. 10.10 A square pulse wave of period 2𝜋.

▶ Solution. For this function, it is more convenient to compute the 𝑎𝑛 and

𝑏𝑛 using integration over the interval [0, 2𝜋] rather than the interval [−𝜋, 𝜋].
Thus,
∫ 2𝜋 ∫ ℎ
1 1 ℎ
𝑎0 = 𝑓 (𝑡) 𝑑𝑡 = 1 𝑑𝑡 = ,
𝜋 0 𝜋 0 𝜋
∫ ℎ
1 sin 𝑛ℎ
𝑎𝑛 = cos 𝑛𝑡 𝑑𝑡 = , and
𝜋 0 𝜋𝑛
∫ ℎ
1 1 − cos 𝑛ℎ
𝑏𝑛 = sin 𝑛𝑡 𝑑𝑡 = ,
𝜋 0 𝜋𝑛
742 10 Fourier Series

and the Fourier series is

∞ ( )
ℎ 1 ∑ sin 𝑛ℎ 1 − cos 𝑛ℎ
𝑓 (𝑡) ∼ + cos 𝑛𝑡 + sin 𝑛𝑡 .
2𝜋 𝜋 𝑛=1 𝑛 𝑛

◀
If the square pulse wave 𝑓 (𝑡) is divided by ℎ, then one obtains a function
whose graph is a series of tall thin rectangles of height 1/ℎ and base ℎ, so
that each of the rectangles with the bases starting at 2𝑛𝜋 has area 1, as
in Figure 10.11. Now consider the limiting case where ℎ approaches 0. The

1
ℎ

−4𝜋 −2𝜋 0 ℎ 2𝜋 4𝜋

Fig. 10.11 A square unit pulse wave of period 2𝜋.

graph becomes a series of infinite height spikes of width 0 as in Figure 10.12.

This looks like an infinite sum of Dirac delta functions, which is the regular
delta function extended to be periodic of period 2𝜋. That is,
∞
𝑓 (𝑡) ∑
lim = 𝛿(𝑡 − 2𝑛𝜋).
ℎ→0 ℎ
𝑛=−∞

Now compute the Fourier coefficients 𝑎𝑛 /ℎ and 𝑏𝑛 /ℎ as ℎ approaches 0.

sin 𝑛ℎ 1
𝑎𝑛 = → and
𝜋𝑛ℎ 𝜋
1 − cos 𝑛ℎ
𝑏𝑛 = → 0 as ℎ → 0
𝜋𝑛ℎ
Also, 𝑎0 /ℎ = 1/𝜋. Thus, the 2𝜋-periodic delta function has a Fourier series
∞ ∞
∑ 1 1∑
𝛿(𝑡 − 2𝑛𝜋) ∼ + cos 𝑛𝑡. (12)
𝑛=−∞
2𝜋 𝜋 𝑛=1

It is worth pointing out that this is one series that definitely does not converge
for all 𝑡. In fact it does not converge for 𝑡 = 0. However, there is a sense of
convergence in which convergence makes sense. We will discuss this in the
next section.
10.2 Fourier Series 743

−4𝜋 −2𝜋 0 2𝜋 4𝜋

Fig. 10.12 The limiting square unit pulse wave of period 2𝜋 as ℎ → 0.

Example 8. Compute the Fourier series of the function 𝑓 (𝑡) = sin 2𝑡 +

5 cos 3𝑡.

▶ Solution. Since 𝑓 (𝑡) is periodic of period 2𝜋 and is already given as a

sum of sines and cosines, no further work is needed. The coefficients can be
read off without any integration as 𝑎𝑛 = 0 for 𝑛 ∕= 3 and 𝑎3 = 5 while 𝑏𝑛 = 0
for 𝑛 ∕= 2 and 𝑏2 = 1. The same results would be obtained by integration
using Euler’s formulas.

Example 9. Compute the Fourier series of the function 𝑓 (𝑡) = sin3 𝑡

▶ Solution. This can be handled by trig identities to reduce it to a finite

sum of terms of the form sin 𝑛𝑡.
1
sin3 𝑡 = sin 𝑡 sin2 𝑡 =
sin 𝑡(1 − cos 2𝑡)
2
1 1 1 1(1 )
= sin 𝑡 − sin 𝑡 cos 2𝑡 = sin 𝑡 − (sin 3𝑡 + sin(−𝑡))
2 2 2 2 2
3 1
= sin 𝑡 − sin 3𝑡
4 4
The right hand side is the Fourier series of sin3 𝑡.

Exercises

1–17. Find the Fourier series of the given periodic function.

{
0 if −5 ≤ 𝑡 < 0
1. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 10) = 𝑓 (𝑡).
3 if 0 ≤ 𝑡 < 5
{
2 if −𝜋 ≤ 𝑡 < 0
2. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).
−2 if 0 ≤ 𝑡 < 𝜋
744 10 Fourier Series
{
4 if −𝜋 ≤ 𝑡 < 0
3. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).
−1 if 0 ≤ 𝑡 < 𝜋

4. 𝑓 (𝑡) = 𝑡, 0 ≤ 𝑡 < 2; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

Hint: It may be useful to take advantage of Equation (10) in applying
the Euler formulas.

5. 𝑓 (𝑡) = 𝑡, −𝜋 ≤ 𝑡 < 𝜋; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).

⎧
⎨0 0 ≤ 𝑡 < 1,

6. 𝑓 (𝑡) = 1 1 ≤ 𝑡 < 2, 𝑓 (𝑡 + 3) = 𝑓 (𝑡).

⎩
0 2 ≤ 𝑡 < 3,

7. 𝑓 (𝑡) = 𝑡2 , −2 ≤ 𝑡 < 2; 𝑓 (𝑡 + 4) = 𝑓 (𝑡).

{
0 if −1 ≤ 𝑡 < 0
8. 𝑓 (𝑡) = 2 ; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).
𝑡 if 0 ≤ 𝑡 < 1

9. 𝑓 (𝑡) = sin 𝑡, 0 ≤ 𝑡 < 𝜋; 𝑓 (𝑡 + 𝜋) = 𝑓 (𝑡).

{
0 if −𝜋 ≤ 𝑡 < 0
10. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).
sin 𝑡 if 0 ≤ 𝑡 < 𝜋
{
1 + 𝑡 if −1 ≤ 𝑡 < 0
11. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).
1 − 𝑡 if 0 ≤ 𝑡 < 1
{
1+𝑡 if −1 ≤ 𝑡 < 0
12. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).
−1 + 𝑡 if 0 ≤ 𝑡 < 1
{
−𝑡(𝜋 + 𝑡) if −𝜋 ≤ 𝑡 < 0
13. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).
𝑡(𝜋 − 𝑡) if 0 ≤ 𝑡 < 𝜋

14. 𝑓 (𝑡) = 3 cos2 𝑡, −𝜋 ≤ 𝑡 < 𝜋; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡)

15. 𝑓 (𝑡) = sin 2𝑡 , −𝜋 ≤ 𝑡 < 𝜋; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡)

16. 𝑓 (𝑡) = sin 𝑝 𝑡, −𝜋 ≤ 𝑡 < 𝜋; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡) (where 𝑝 is not an integer)

17. 𝑓 (𝑡) = 𝑒𝑡 ; −1 ≤ 𝑡 < 1; 𝑓 (𝑡 + 2) = 𝑓 (𝑡)

Then,
∫ 1
𝑎𝑛 = 𝑒𝑡 cos 𝑛𝜋𝑡 𝑑𝑡
−1
10.3 Convergence of Fourier Series 745
1
1 𝑡

= 𝑒 [cos 𝑛𝜋𝑡 + 𝑛𝜋 sin 𝑛𝜋𝑡] 
1 + 𝑛2 𝜋 2 
−1
1
= [𝑒1 cos 𝑛𝜋 − 𝑒−1 cos(−𝑛𝜋)]
1 + 𝑛2 𝜋 2
(𝑒1 − 𝑒−1 )(−1)𝑛 2(−1)𝑛 sinh(1)
= 2 2
= ,
1+𝑛 𝜋 1 + 𝑛2 𝜋 2
and,
∫ 1
𝑏𝑛 = 𝑒𝑡 sin 𝑛𝜋𝑡 𝑑𝑡
−1
1
1 𝑡

= 2 2
𝑒 [sin 𝑛𝜋𝑡 − 𝑛𝜋 cos 𝑛𝜋𝑡] 
1+𝑛 𝜋 
−1
1 1 −1
= [𝑒 (−𝑛𝜋 cos 𝑛𝜋) − 𝑒 (−𝑛𝜋 cos(−𝑛𝜋))]
1 + 𝑛2 𝜋 2
(𝑒1 − 𝑒−1 )(−𝑛𝜋)(−1)𝑛 2(−1)𝑛 (−𝑛𝜋) sinh(1)
= = .
1 + 𝑛2 𝜋 2 1 + 𝑛2 𝜋 2
Therefore, the Fourier series is
∞
∑ (−1)𝑛 (cos 𝑛𝜋𝑡 − 𝑛𝜋 sin 𝑛𝜋𝑡)
𝑓 (𝑡) ∼ sinh(1) + 2 sinh(1) .
𝑛=1
1 + 𝑛2 𝜋 2

10.3 Convergence of Fourier Series

If 𝑓 (𝑡) is a periodic function of period 2𝐿, the Fourier series of 𝑓 (𝑡) is

∞ ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓 (𝑡) ∼ + 𝑎𝑛 cos + 𝑏𝑛 sin , (1)
2 𝑛=1
𝐿 𝐿

where the coefficients 𝑎𝑛 and 𝑏𝑛 are given by the Euler formulas

∫ 𝐿
1 𝑛𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑡 𝑑𝑡, 𝑛 = 0, 1, 2, 3, . . . (2)
𝐿 −𝐿 𝐿
∫ 𝐿
1 𝑛𝜋
𝑏𝑛 = 𝑓 (𝑡) sin 𝑡 𝑑𝑡, 𝑛 = 1, 2, 3, . . . (3)
𝐿 −𝐿 𝐿

The questions that we wish to consider are (1) under what conditions on
the function 𝑓 (𝑡) is it true that the series converges, and (2) when does it
converge to the original function 𝑓 (𝑡)? ∑∞
As with any infinite series,
∑𝑚 to say that 𝑛=1 𝑐𝑛 = 𝐶 means that the
sequence of partial sums 𝑛=1 𝑐𝑛 = 𝑆𝑚 converges to 𝐶, that is,
746 10 Fourier Series

lim 𝑆𝑚 = 𝐶.
𝑚→∞

The partial sums of the Fourier series (1) are

𝑚 ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑆𝑚 (𝑡) = + 𝑎𝑛 cos + 𝑏𝑛 sin . (4)
2 𝑛=1
𝐿 𝐿

The partial sum 𝑆𝑚 (𝑡) is a finite linear combination of the trigonometric

functions cos 𝑛𝜋𝑡 𝑛𝜋𝑡
𝐿 and sin 𝐿 , each of which is periodic of period 2𝐿. Thus,
𝑆𝑚 (𝑡) is also periodic of period 2𝐿, and if the partial sums converge to a
function 𝑔(𝑡), then 𝑔(𝑡) must also be periodic of period 2𝐿.

Example 1. Consider the odd square wave function of period 2 and am-
plitude 1 from Example 2 of Section 10.2
{
−1 −1 ≤ 𝑡 < 0,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 2) = 𝑓 (𝑡). (5)
1 0 ≤ 𝑡 < 1,

The Fourier series of this function was found to be (letting 𝐿 = 1)

( )
4 1 1 1
𝑓 (𝑡) ∼ sin 𝜋𝑡 + sin 3𝜋𝑡 + sin 5𝜋𝑡 + sin 7𝜋𝑡 + ⋅ ⋅ ⋅ (6)
𝜋 3 5 7
∞
4 ∑ 1
= sin(2𝑘 + 1)𝜋𝑡.
𝜋 2𝑘 + 1
𝑘=0

Letting 𝑡 = 0 in this series gives

∞ ∞
4∑ 1 4∑
sin(2𝑘 + 1)𝜋0 = 0 = 0.
𝜋 2𝑘 + 1 𝜋
𝑘=0 𝑘=0

Since 𝑓 (0) = 1, it follows that the Fourier series does not converge to 𝑓 (𝑡)
for all 𝑡. Several partial sums 𝑆𝑚 are graphed along with the graph of 𝑓 (𝑡) in
Figure 10.13. It can be seen that the graph of 𝑆15 (𝑡) is very close to the graph
of 𝑓 (𝑡), except at the points of discontinuity, which are 0, ±1, ±2, . . .. This
suggests that the Fourier series of 𝑓 (𝑡) converges to 𝑓 (𝑡) at all points except
for where 𝑓 (𝑡) has a discontinuity, which occurs whenever 𝑡 is an integer 𝑛.
For each integer 𝑛,
[2]
𝑚
[2] 𝑚
4∑ 1 4∑
𝑆𝑚 (𝑛) = sin(2𝑘 + 1)𝜋𝑛 = 0 = 0.
𝜋 2𝑘 + 1 𝜋
𝑘=0 𝑘=0

Hence the Fourier series converges to 𝑓 (𝑡) whenever 𝑡 is not an integer and
it converges to 0, which is the midpoint of the jump of 𝑓 (𝑡) at each integer.
10.3 Convergence of Fourier Series 747

𝑡
−3 −2 −1 1 2 3

−1

4
𝑆1 (𝑡) = 𝜋
sin 𝜋𝑡

𝑡
−3 −2 −1 1 2 3

−1

4 1 1
( )
𝑆5 (𝑡) = 𝜋
sin 𝜋𝑡 + 3
sin 3𝜋𝑡 + 5
sin 5𝜋𝑡

𝑡
−3 −2 −1 1 2 3

−1

4
∑7 1
𝑆15 (𝑡) = 𝜋 𝑘=0 2𝑘+1 sin(2𝑘 + 1)𝜋𝑡

Fig. 10.13 Fourier series approximations 𝑆𝑚 to the odd square wave of period 2 for
𝑚 = 1, 5, and 15.

Similar results are true for a broad range of functions. We will describe the
types of functions to which the convergence theorem applies and then state
the convergence theorem.
Recall that a function 𝑓 (𝑡) defined on an interval 𝐼 has a jump discon-
tinuity at a point 𝑡 = 𝑎 if the left hand limit 𝑓 (𝑎− ) = lim𝑡→𝑎− 𝑓 (𝑡) and the
right hand limit 𝑓 (𝑎+ ) = lim𝑡→𝑎+ 𝑓 (𝑡) both exist (as real numbers, not ±∞)
and
𝑓 (𝑎+ ) ∕= 𝑓 (𝑎− ).
748 10 Fourier Series

The difference 𝑓 (𝑎+ ) − 𝑓 (𝑎− ) is frequently referred to as the jump in 𝑓 (𝑡)

at 𝑡 = 𝑎. For example the odd square wave function of Example 10.2.2 has
jump discontinuities at the points ±𝑛𝐿 for 𝑛 = 0, 1, 2, . . . and the sawtooth
wave function of Example 10.2.5 has jump discontinuities at ±𝑛 for 𝑛 = 0,
1, 2, . . .. On the other hand, the function 𝑔(𝑡) = tan 𝑡 is a periodic function
of period 𝜋 that is discontinuous at the points 𝜋2 + 𝑘𝜋 for all integers 𝑘, but
the discontinuity is not a jump discontinuity since, for example

lim tan 𝑡 = +∞, and lim

𝜋+
tan 𝑡 = −∞.
𝑡→ 𝜋
2
− 𝑡→ 2

A function 𝑓 (𝑡) is piecewise continuous on a closed interval [𝑎, 𝑏] if 𝑓 (𝑡)

is continuous except for possibly finitely many jump discontinuities in the in-
terval [𝑎, 𝑏]. A function 𝑓 (𝑡) is piecewise continuous for all 𝑡 if it is piecewise
continuous on every bounded interval. In particular, this means that there
can be only finitely many discontinuities in any given bounded interval, and
each of those must be a jump discontinuity. For convenience it will not be
required that 𝑓 (𝑡) be defined at the jump discontinuities, or 𝑓 (𝑡) can be de-
fined arbitrarily at the jump discontinuities. The square waves and sawtooth
waves from Section 10.2 are typical examples of piecewise continuous periodic
functions, while tan 𝑡 is not piecewise continuous since the discontinuities are
not jump discontinuities.
A function 𝑓 (𝑡) is piecewise smooth if it is piecewise continuous and
the derivative 𝑓 ′ (𝑡) is also piecewise continuous. As with the convention on
piecewise continuous, 𝑓 ′ (𝑡) may not exist at finitely many points in any closed
interval. All of the examples of periodic functions from Section 10.2 are piece-
wise smooth. The property of 𝑓 (𝑡) being piecewise smooth and periodic is
sufficient to guarantee that the Fourier series of 𝑓 (𝑡) converges, and the sum
can be computed. This is the content of the following theorem.

Theorem 2. (Convergence of Fourier Series) Suppose that 𝑓 (𝑡) is a

periodic piecewise smooth function of period 2𝐿. Then the Fourier series (1)
converges
(a) to the value 𝑓 (𝑡) at each point 𝑡 where 𝑓 is continuous, and
(b) to the value 12 [𝑓 (𝑡+ ) + 𝑓 (𝑡− )] at each point where 𝑓 is not continuous.

Note that 12 [𝑓 (𝑡+ ) + 𝑓 (𝑡− )] is the average of the left-hand and right-hand
limits of 𝑓 at the point 𝑡. If 𝑓 is continuous at 𝑡, then 𝑓 (𝑡+ ) = 𝑓 (𝑡) = 𝑓 (𝑡− ),
so that
𝑓 (𝑡+ ) + 𝑓 (𝑡− )
𝑓 (𝑡) = , (7)
2
for any 𝑡 where 𝑓 is continuous. Hence Theorem 2 can be rephrased as follows:
The Fourier series of a piecewise smooth periodic function 𝑓 converges for
every 𝑡 to the average of the left-hand and right-hand limits of 𝑓 .
10.3 Convergence of Fourier Series 749

Example 3. Continuing with the odd square wave function of Example 1,

it is clear from Equation (5) or from Figure 10.13 that if 𝑛 is an even integer,
then
lim 𝑓 (𝑡) = +1 and lim 𝑓 (𝑡) = −1,
𝑡→𝑛+ 𝑡→𝑛−

while if 𝑛 is an odd integer, then

lim 𝑓 (𝑡) = −1 and lim 𝑓 (𝑡) = +1.

𝑡→𝑛+ 𝑡→𝑛−

Therefore, whenever 𝑛 is an integer,

𝑓 (𝑛+ ) + 𝑓 (𝑛− )
= 0.
2
This is consistent with Theorem 2 since the Fourier series (6) converges to 0
whenever 𝑡 is an integer 𝑛 (because sin(2𝑘 + 1)𝜋𝑛 = 0).

For any point 𝑡 other than an integer, the function 𝑓 (𝑡) is continuous, so
Theorem 2 gives an equality
( )
4 1 1 1
𝑓 (𝑡) = sin 𝜋𝑡 + sin 3𝜋𝑡 + sin 5𝜋𝑡 + sin 7𝜋𝑡 + ⋅ ⋅ ⋅ .
𝜋 3 5 7

By redefining 𝑓 (𝑡) at integers to be 𝑓 (𝑛) = 0 for integers 𝑛, then this equality

holds for all 𝑡. Figure 10.14 shows the graph of the redefined function. Putting

𝑡
−3 −2 −1 1 2 3

−1

Fig. 10.14 The odd square wave of period 2 redefined so that the Fourier series converges
to 𝑓 (𝑡) for all 𝑡.

in the value of 𝑓 (𝑡) for some values of 𝑡 leads to a trigonometric identity

( )
4 1 1 1
sin 𝜋𝑡 + sin 3𝜋𝑡 + sin 5𝜋𝑡 + sin 7𝜋𝑡 + ⋅ ⋅ ⋅ = 𝑓 (𝑡) = 1 if 0 < 𝑡 < 1.
𝜋 3 5 7

Letting 𝑡 = 1/2 then gives the series

750 10 Fourier Series

1 1 1 𝜋
1− + − + ⋅⋅⋅ = .
3 5 7 4
Example 4. Let 𝑓 be periodic of period 2 and defined on the interval 0 ≤
𝑡 < 2 by 𝑓 (𝑡) = 𝑡2 . Without computing the Fourier coefficients, give an
explicit description of the sum of the Fourier series of 𝑓 for all 𝑡.

▶ Solution. The function is defined in only one period so the periodic

extension is defined by

𝑓 (𝑡) = 𝑡2 for 0 ≤ 𝑡 < 2; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

Both 𝑓 (𝑡) and 𝑓 ′ (𝑡) are continuous everywhere except at the even integers.
Thus, the Fourier series converges to 𝑓 (𝑡) everywhere except at the even
integers. At the even integer 2𝑚 the left and right limits are

𝑓 (2𝑚− ) = lim 𝑓 (𝑡) = lim 𝑓 (𝑡) = lim (𝑡 + 2)2 = 4, and

𝑡→2𝑚− 𝑡→0− 𝑡→0−

𝑓 (2𝑚+ ) = lim + 𝑓 (𝑡) = lim+ 𝑓 (𝑡) = lim+ 𝑡2 = 0.

𝑡→2𝑚 𝑡→0 𝑡→0

Therefore the sum of the Fourier series at 𝑡 = 2𝑚 is

𝑓 (2𝑚− ) + 𝑓 (2𝑚+ ) 4+0
= = 2.
2 2
The graph of the sum is shown in Figure 10.15. ◀

𝑦
4

b b b b b b b

𝑡
−6 −4 −2 2 4 6

Fig. 10.15 The graph of sum of the Fourier series of the period 2 function 𝑓 (𝑡) for Example
4.

∑∞
Example 5. The 2𝜋-periodic delta function 𝑛=−∞ 𝛿(𝑡 − 2𝑛𝜋) has a
Fourier series ∞ ∞
∑ 1 1∑
𝛿(𝑡 − 2𝑛𝜋) ∼ + cos 𝑛𝑡 (8)
𝑛=−∞
2𝜋 𝜋 𝑛=1
10.3 Convergence of Fourier Series 751

that
∑∞ was computed in Section 10.2. See Figure 10.12 for the graph of
𝑛=−∞ 𝛿(𝑡 − 2𝑛𝜋). The 2𝜋-periodic delta function does not satisfy the con-
ditions of the convergence theorem. Moreover, it is clear that the series does
not converge since the individual terms do not approach 0. If 𝑡 = 0 the series
becomes
1 1
+ [1 + 1 + 1 + ⋅ ⋅ ⋅ ],
2𝜋 𝜋
while, if 𝑡 = 𝜋 the series is
1 1
+ [−1 + 1 − 1 + ⋅ ⋅ ⋅ ].
2𝜋 𝜋
Neither of these series are convergent.
∑∞ However, there is a sense of∑convergence
∞
in which the Fourier series of 𝑛=−∞ 𝛿(𝑡−2𝑛𝜋) "converges" to 𝑛=−∞ 𝛿(𝑡−
2𝑛𝜋). To explore this phenomenon, we will explicitly compute the 𝑛𝑡ℎ partial
sum 𝛥𝑛 (𝑡) of the series (8):
1 1
𝛥𝑛 (𝑡) = + [cos 𝑡 + cos 2𝑡 + ⋅ ⋅ ⋅ + cos 𝑛𝑡]
2𝜋 𝜋
1
= [1 + 2 cos 𝑡 + 2 cos 2𝑡 + ⋅ ⋅ ⋅ + 2 cos 𝑛𝑡]
2𝜋
Now use the Euler identity 2 cos 𝜃 = 𝑒𝑖𝜃 + 𝑒−𝑖𝜃 to get
1 [ ]
𝛥𝑛 (𝑡) = 1 + 𝑒𝑖𝑡 + 𝑒−𝑖𝑡 + 𝑒𝑖2𝑡 + 𝑒−𝑖2𝑡 + ⋅ ⋅ ⋅ + 𝑒𝑖𝑛𝑡 + 𝑒−𝑖𝑛𝑡 .
2𝜋
Letting 𝑧 = 𝑒𝑖𝑡 , this is seen to be a geometric series of ratio 𝑧 that begins
with the term 𝑧 −𝑛 and ends with the term 𝑧 𝑛 . Since the sum of a geometric
series is known: 𝑎 + 𝑎𝑟 + 𝑎𝑟2 + ⋅ ⋅ ⋅ + 𝑎𝑟𝑚 = 𝑎(1 − 𝑟𝑚+1 )/(1 − 𝑟) it follows that
1 [ ]
𝛥𝑛 (𝑡) = 1 + 𝑒𝑖𝑡 + 𝑒−𝑖𝑡 + 𝑒𝑖2𝑡 + 𝑒−𝑖2𝑡 + ⋅ ⋅ ⋅ + 𝑒𝑖𝑛𝑡 + 𝑒−𝑖𝑛𝑡
2𝜋
1 [ −𝑛 ] 1 −𝑛 [ ]
= 𝑧 + 𝑧 −𝑛+1 + ⋅ ⋅ ⋅ + 𝑧 𝑛 = 𝑧 1 + 𝑧 + ⋅ ⋅ ⋅ + 𝑧 2𝑛
2𝜋 2𝜋
1 −𝑛 1 − 𝑧 2𝑛+1 1 𝑧 −𝑛 − 𝑧 𝑛+1
= 𝑧 =
2𝜋 1−𝑧 2𝜋 1−𝑧
1 𝑒−𝑖𝑛𝑡 − 𝑒𝑖(𝑛+1)𝑡 1 𝑒𝑖(𝑛+1)𝑡 − 𝑒−𝑖𝑛𝑡 𝑒−𝑖𝑡/2
= =
2𝜋 1 − 𝑒𝑖𝑡 2𝜋 𝑒𝑖𝑡 − 1 𝑒−𝑖𝑡/2
𝑖(𝑛+ 12 )𝑡 −𝑖(𝑛+ 12 )𝑡
1 𝑒 −𝑒
= 𝑖𝑡/2
2𝜋 𝑒 − 𝑒−𝑖𝑡/2
( 1
)
1 sin 𝑛 + 2 𝑡
= .
2𝜋 sin 12 𝑡
752 10 Fourier Series

For those values of 𝑡 where sin 12 𝑡 = 0, the value of 𝛥𝑛 (𝑡) is the limiting value.
The graphs of 𝛥𝑛 (𝑡) for 𝑛 = 5 and 𝑛 = 15 over 1 period −𝜋 ≤ 𝑡 ≤ 𝜋] are
shown in Figure 10.16. As 𝑛 increases the central hump increases in height
and becomes narrower, while away from 0 the graph is a rapid oscillation of
small amplitude. Also, the area enclosed by 𝛥𝑛 (𝑡) in one period −𝜋 ≤ 𝑡 ≤ 𝜋
is always 1. This is easily seen from the expression of 𝛥𝑛 (𝑡) as a sum of cosine
terms:
∫ 𝜋 ∫ 𝜋
1
𝛥𝑛 (𝑡) 𝑑𝑡 = [1 + 2 cos 𝑡 + 2 cos 2𝑡 + ⋅ ⋅ ⋅ + 2 cos 𝑛𝑡] 𝑑𝑡 = 1
−𝜋 −𝜋 2𝜋
∫𝜋
since −𝜋 cos 𝑚𝑡 𝑑𝑡 = 0 for all nonzero integers 𝑚. Most of the area is con-
centrated in the first hump since the rapid oscillation cancels out the contri-
butions above and below the 𝑡 axis away from this hump.

3 𝛿15 (𝑡)

1 𝛿5 (𝑡)

−𝜋 𝜋

−1

Fig. 10.16 The graphs of the partial sums 𝛿𝑛 (𝑡) for 𝑛 = 5 and 𝑛 = 15.

∑∞
To see the sense in which 𝛥𝑛 (𝑡) converges to 𝑛=−∞ 𝛿(𝑡 − 2𝑛𝜋), recall
that ∫𝛿(𝑡) is described via integration by the property that 𝛿(𝑡) = 0 if 𝑡 ∕= 0,
and 𝑓 (𝑡)𝛿(𝑡) 𝑑𝑡 = 𝑓 (0) for all test functions 𝑓 (𝑡). Since we are working
with periodic functions, consider the integrals over 1 period, in our case
[−𝜋, 𝜋].∑Thus let 𝑓 (𝑡) be a smooth function and expand it in a cosine series
𝑓 (𝑡) = 𝑎𝑚 cos 𝑚𝑡. Multiply by 𝛥𝑛 (𝑡) and integrate, using the orthogonality
conditions to get ∫ 𝜋
𝛥𝑛 (𝑡)𝑓 (𝑡) 𝑑𝑡 = 𝑎0 + ⋅ ⋅ ⋅ + 𝑎𝑛 . (9)
−𝜋

As 𝑛 → ∞ this sum converges to 𝑓 (0). Thus

10.3 Convergence of Fourier Series 753
∫ 𝜋 ∫ 𝜋
lim 𝛥𝑛 (𝑡)𝑓 (𝑡) 𝑑𝑡 = 𝑓 (0) = 𝛿(𝑡)𝑓 (𝑡) 𝑑𝑡.
𝑛→∞ −𝜋 −𝜋
∑∞
It is in this weak sense that 𝛥𝑛 (𝑡) → 𝑛=−∞ 𝛿(𝑡 − 2𝑛𝜋). Namely, the effect
under integration is the same.

Exercises

1–9. In each exercise, (a) Sketch the graph of 𝑓 (𝑡) over at least 3 periods.
(b) Determine the points 𝑡 at which the Fourier series of 𝑓 (𝑡) converges to
𝑓 (𝑡). (c) At each point 𝑡 of discontinuity, give the value of 𝑓 (𝑡) and the value
to which the Fourier series converges.
{
3 if 0 ≤ 𝑡 < 2
1. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 4) = 𝑓 (𝑡).
−1 if 2 ≤ 𝑡 < 4

2. 𝑓 (𝑡) = 𝑡, −1 ≤ 𝑡 ≤ 1, 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

3. 𝑓 (𝑡) = 𝑡, 0 ≤ 𝑡 ≤ 2, 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

4. 𝑓 (𝑡) = ∣𝑡∣, −1 ≤ 𝑡 ≤ 1, 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

5. 𝑓 (𝑡) = 𝑡2 , −2 ≤ 𝑡 ≤ 2, 𝑓 (𝑡 + 4) = 𝑓 (𝑡).

6. 𝑓 (𝑡) = 2𝑡 − 1, −1 ≤ 𝑡 < 1, 𝑓 (𝑡 + 2) = 𝑓 (𝑡).

{
2 if −2 ≤ 𝑡 < 0
7. 𝑓 (𝑡) = ; 𝑓 (𝑡 + 4) = 𝑓 (𝑡).
𝑡 if 0 ≤ 𝑡 < 2

8. 𝑓 (𝑡) = cos(𝑡/2), 0 ≤ 𝑡 < 𝜋, 𝑓 (𝑡 + 𝜋) = 𝑓 (𝑡).

9. 𝑓 (𝑡) = ∣cos 𝜋𝑡∣, 0 ≤ 𝑡 < 1, 𝑓 (𝑡 + 1) = 𝑓 (𝑡).

10. Verify that the following function is not piecewise smooth.

{√
𝑡 if 0 ≤ 𝑡 < 1
𝑓 (𝑡) =
0 if −1 ≤ 𝑡 < 0.

11. Verify that

1 1 1 𝑡
sin 𝑡 − sin 2𝑡 + sin 3𝑡 − sin 4𝑡 + ⋅ ⋅ ⋅ = , for −𝜋 < 𝑡 < 𝜋.
2 3 4 2
Explain how this gives the summation
754 10 Fourier Series

1 1 1 𝜋
1− + − + ⋅⋅⋅ = .
3 5 7 4

12. Use the Fourier series of the function 𝑓 (𝑡) = ∣𝑡∣, −𝜋 < 𝑡 < 𝜋 (see Example
4 of Section 10.2) to compute the sum of the series
1 1 1
1+ 2
+ 2 + 2 + ⋅⋅⋅ .
3 5 7

13. Verify that

∞
1 4 ∑ (−1)𝑛
+ cos 𝑛𝜋𝑡 = 𝑡2 , for −1 ≤ 𝑡 ≤ 1.
3 𝜋 2 𝑛=1 𝑛2

14. Using the Fourier series in problem 13, find the sum of the following
series ∞ ∞
∑ 1 ∑ (−1)𝑛+1
and .
𝑛=1
𝑛2 𝑛=1
𝑛2

15. Compute the Fourier series for the function 𝑓 (𝑡) = 𝑡4 , −𝜋 ≤ 𝑡 ≤ 𝜋;

𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡) for all 𝑡. Using this result, verify the summations
∞ ∞
∑ 1 𝜋4 ∑ (−1)𝑛+1 7𝜋 4
4
= and 4
= .
𝑛=1
𝑛 90 𝑛=1
𝑛 720

10.4 Fourier Sine Series and Fourier Cosine Series

In Section 10.2 we observed (see the discussion following Example 5) that the
Fourier series of an even function will only have cosine terms, while the Fourier
series of an odd function will only have sine terms. We will take advantage
of this feature to obtain new types of Fourier series representations that are
valid for functions defined on some interval 0 ≤ 𝑡 ≤ 𝐿. In practice, we will
frequently want to represent 𝑓 (𝑡) by a Fourier series that involves only cosine
terms or only sine terms. To do this, we will first extend 𝑓 (𝑡) to the interval
−𝐿 ≤ 𝑡 ≤ 0. We may then extend 𝑓 (𝑡) to all of the real line by assuming that
𝑓 (𝑡) is 2𝐿 periodic. That is, assume 𝑓 (𝑡 + 2𝐿) = 𝑓 (𝑡). We will extend 𝑓 (𝑡) in
such a way as to obtain either an even function or an odd function.
If 𝑓 (𝑡) is defined only for 0 ≤ 𝑡 ≤ 𝐿, then the even 2𝐿-periodic exten-
sion of 𝑓 (𝑡) is the function 𝑓𝑒 (𝑡) defined by
10.4 Fourier Sine Series and Fourier Cosine Series 755
{
𝑓 (𝑡) if 0 ≤ 𝑡 ≤ 𝐿,
𝑓𝑒 (𝑡) = ; 𝑓𝑒 (𝑡 + 2𝐿) = 𝑓𝑒 (𝑡). (1)
𝑓 (−𝑡) if −𝐿 ≤ 𝑡 < 0,

The odd 2𝐿-periodic extension of 𝑓 (𝑡) is the function 𝑓𝑜 (𝑡) defined by

{
𝑓 (𝑡) if 0 ≤ 𝑡 ≤ 𝐿,
𝑓𝑒 (𝑡) = ; 𝑓𝑜 (𝑡 + 2𝐿) = 𝑓𝑜 (𝑡). (2)
−𝑓 (−𝑡) if −𝐿 ≤ 𝑡 < 0,

These two extensions are illustrated for a particular function 𝑓 (𝑡) in Figure
10.17. Note that both 𝑓𝑒 (𝑡) and 𝑓𝑜 (𝑡) agree with 𝑓 (𝑡) on the interval 0 < 𝑡 <
𝐿.

𝑡
𝐿

The graph of 𝑓 (𝑡) on the interval 0 ≤ 𝑡 ≤ 𝐿.

𝑦 𝑦

𝑡 𝑡
−3𝐿 −2𝐿 −𝐿 𝐿 2𝐿 3𝐿 −3𝐿 −2𝐿 −𝐿 𝐿 2𝐿 3𝐿

The graph of 𝑓𝑒 (𝑡), the even The graph of 𝑓𝑜 (𝑡), the odd
2𝐿-periodic extension of 𝑓 (𝑡). 2𝐿-periodic extension of 𝑓 (𝑡).

Fig. 10.17

Since 𝑓𝑒 (𝑡) is an even 2𝐿-periodic function, then 𝑓𝑒 (𝑡) has a Fourier series
expansion
∞ ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓𝑒 (𝑡) ∼ + 𝑎𝑛 cos + 𝑏𝑛 sin
2 𝑛=1
𝐿 𝐿

where the coefficients are given by Euler’s formulas (Equations (6)-(8) of

Section 10.2)
∫ 𝐿
1
𝑎0 = 𝑓𝑒 (𝑡) 𝑑𝑡
𝐿 −𝐿
756 10 Fourier Series
∫ 𝐿
1 𝑛𝜋
𝑎𝑛 = 𝑓𝑒 (𝑡) cos 𝑡 𝑑𝑡, 𝑛 = 1, 2, 3, . . .
𝐿 −𝐿 𝐿
∫ 𝐿
1 𝑛𝜋
𝑏𝑛 = 𝑓𝑒 (𝑡) sin 𝑡 𝑑𝑡, 𝑛 = 1, 2, 3, . . . .
𝐿 −𝐿 𝐿

Since 𝑓𝑒 (𝑡) is an even function, then 𝑓𝑒 (𝑡) sin 𝑛𝜋

𝐿 𝑡 is odd, so the integral defin-
ing 𝑏𝑛 is 0 (see Proposition 5 of Section 10.1). Thus, 𝑏𝑛 = 0 and we get the
Fourier series expansion
∞
𝑎0 ∑ 𝑛𝜋𝑡
𝑓𝑒 (𝑡) ∼ + 𝑎𝑛 cos ,
2 𝑛=1
𝐿

where ∫ 𝐿
1 𝑛𝜋𝑡
𝑎𝑛 = 𝑓𝑒 (𝑡) cos 𝑑𝑡, for 𝑛 ≥ 0.
𝐿 −𝐿 𝐿
Since the integrand is even and 𝑓𝑒 (𝑡) = 𝑓 (𝑡) for 0 ≤ 𝑡 ≤ 𝐿, it follows that
∫ 𝐿
2 𝑛𝜋𝑡
𝑎𝑛 = 𝑓 (𝑡) cos 𝑑𝑡, for 𝑛 ≥ 0. (3)
𝐿 0 𝐿

If we assume that 𝑓 (𝑡) is piecewise smooth on the interval 0 ≤ 𝑡 ≤ 𝐿, then

the even 2𝐿-periodic extension 𝑓𝑒 (𝑡) is also piecewise smooth, and the con-
vergence theorem (Theorem 2 of Section 10.3) shows that the Fourier series
of 𝑓𝑒 (𝑡) converges to the average of the left-hand and right-hand limits of
𝑓𝑒 at each 𝑡. By restricting to the interval 0 ≤ 𝑡 ≤ 𝐿, we obtain the series
expansion
∞
𝑎0 ∑ 𝑛𝜋𝑡
𝑓 (𝑡) ∼ + 𝑎𝑛 cos , for 0 ≤ 𝑡 ≤ 𝐿, (4)
2 𝑛=1
𝐿

where the coefficients are given by (3) and this series converges to the average
of the left-hand and right-hand limits of 𝑓 at each 𝑡 in the interval [0, 𝐿]. The
series in (4) is called the Fourier cosine series for 𝑓 on the interval [0, 𝐿].
Similarly, by using the Fourier series expansion of the odd 2𝐿-periodic
extension of 𝑓 (𝑡), and restricting to the interval 0 ≤ 𝑡 ≤ 𝐿, we obtain the
series expansion
∞
∑ 𝑛𝜋𝑡
𝑓 (𝑡) ∼ 𝑏𝑛 sin , for 0 ≤ 𝑡 ≤ 𝐿, (5)
𝑛=1
𝐿

where the coefficients are given by

∫ 𝐿
2 𝑛𝜋𝑡
𝑏𝑛 = 𝑓 (𝑡) sin 𝑑𝑡, for 𝑛 ≥ 1, (6)
𝐿 0 𝐿
10.4 Fourier Sine Series and Fourier Cosine Series 757

and this series converges to the average of the left-hand and right-hand limits
of 𝑓 at each 𝑡 in the interval [0, 𝐿]. The series in (5) is called the Fourier
sine series for 𝑓 on the interval [0, 𝐿].
Example 1. Let 𝑓 (𝑡) = 𝜋 − 𝑡 for 0 ≤ 𝑡 ≤ 𝜋. Compute both the Fourier
cosine series and the Fourier sine series for the function 𝑓 (𝑡). Sketch the graph
of the function to which the Fourier cosine series converges and the graph of
the function to which to Fourier sine series converges.
▶ Solution. In this case, the length of the interval is 𝐿 = 𝜋 and from (4),
the Fourier cosine series of 𝑓 (𝑡) is
∞
𝑎0 ∑
𝑓 (𝑡) ∼ + 𝑎𝑛 cos 𝑛𝑡,
2 𝑛=1

with ∫ 𝜋 ∫ 𝜋
2 2
𝑎0 = 𝑓 (𝑡) 𝑑𝑡 = (𝜋 − 𝑡) 𝑑𝑡 = 𝜋,
𝜋 0 𝜋 0
and for 𝑛 ≥ 1,
∫ ∫
2 𝜋 2 𝜋
𝑎𝑛 = 𝑓 (𝑡) cos 𝑛𝑡 𝑑𝑡 = (𝜋 − 𝑡) cos 𝑛𝑡 𝑑𝑡
𝜋 0 𝜋 0
𝜋 ∫ 𝜋
2  2
= sin 𝑛𝑡 − 𝑡 cos 𝑛𝑡 𝑑𝑡
𝑛 0 𝜋 0
[ 𝜋 ∫ ]
2 𝑡  1 𝜋
=− sin 𝑛𝑡 −
 sin 𝑛𝑡 𝑑𝑡
𝜋 𝑛 0 𝑛 0
𝜋
2  2
= 2
(− cos 𝑛𝑡) = (− cos 𝑛𝜋 + 1)
𝜋𝑛 0 𝜋𝑛2
2
= 2
(1 − (−1)𝑛 )
𝜋𝑛
{
0 if 𝑛 is even,
= 4
𝜋𝑛2 if 𝑛 is odd.

Thus, the Fourier cosine series of 𝑓 (𝑡) is

( )
𝜋 4 1 1
𝑓 (𝑡) = 𝜋 − 𝑡 = + cos 𝑡 + 2 cos 3𝑡 + ⋅ ⋅ ⋅ cos(2𝑘 − 1)𝑡 + ⋅ ⋅ ⋅ .
2 𝜋 3 (2𝑘 − 1)2

This series converges to 𝑓 (𝑡) for 0 ≤ 𝑡 ≤ 𝜋 and to the even 2𝜋-periodic

extension 𝑓𝑒 (𝑡) otherwise, since 𝑓𝑒 (𝑡) is continuous, as can be seen from the
graph of 𝑓𝑒 (𝑡) which is given in Figure 10.18.
Similarly, from (5), the Fourier sine series of 𝑓 (𝑡) is
∞
∑
𝑓 (𝑡) ∼ 𝑏𝑛 sin 𝑛𝑡
𝑛=1
758 10 Fourier Series

−3𝜋 −2𝜋 −𝜋 𝜋 2𝜋 3𝜋

Fig. 10.18 The even 2𝜋-periodic extension 𝑓𝑒 (𝑡) of the function 𝑓 (𝑡) = 𝜋 − 𝑡 defined on
0 ≤ 𝑡 ≤ 𝜋.

with (for 𝑛 ≥ 1)
∫ 𝜋 ∫
2 2 𝜋
𝑏𝑛 = 𝑓 (𝑡) sin 𝑛𝑡 𝑑𝑡 = (𝜋 − 𝑡) sin 𝑛𝑡 𝑑𝑡
𝜋0 𝜋 0
𝜋 ∫
2  2 𝜋
= − cos 𝑛𝑡 − 𝑡 sin 𝑛𝑡 𝑑𝑡
𝑛 0 𝜋 0
𝜋 [ 𝜋 ∫ ]
2  2 𝑡  1 𝜋
= − cos 𝑛𝑡 −
 − cos 𝑛𝑡 + cos 𝑛𝑡 𝑑𝑡
𝑛 0 𝜋 𝑛 0 𝑛 0
2
=
𝑛
Thus, the Fourier sine series of 𝑓 (𝑡) is
∞
∑ sin 𝑛𝑡
𝑓 (𝑡) = 𝜋 − 𝑡 = 2 .
𝑛=1
𝑛

This series converges to 𝑓 (𝑡) for 0 < 𝑡 < 𝜋 and to the odd 2𝜋-periodic
extension 𝑓𝑜 (𝑡) otherwise, except for the jump discontinuities of 𝑓𝑜 (𝑡), which
occur at the multiples of 𝑛𝜋, as can be seen from the graph of 𝑓𝑜 (𝑡) which is
given in Figure 10.19. At 𝑡 = 𝑛𝜋, the series converges to 0. ◀

−3𝜋 −2𝜋 −𝜋 𝜋 2𝜋 3𝜋

−𝜋

Fig. 10.19 The odd 2𝜋-periodic extension 𝑓𝑒 (𝑡) of the function 𝑓 (𝑡) = 𝜋 − 𝑡 defined on
0 ≤ 𝑡 ≤ 𝜋.
10.5 Operations on Fourier Series 759

Exercises

1–11. A function 𝑓 (𝑡) is defined on an interval 0 < 𝑡 < 𝐿. Find the Fourier
cosine and sine series of 𝑓 and sketch the graphs of the two extensions of 𝑓
to which these two series converge.
1. 𝑓 (𝑡) = 1, 0 < 𝑡 < 𝐿

2. 𝑓 (𝑡) = 𝑡, 0 < 𝑡 < 1

3. 𝑓 (𝑡) = 𝑡, 0 < 𝑡 < 2

4. 𝑓 (𝑡) = 𝑡2 , 0 < 𝑡 < 1

{
1 if 0 < 𝑡 < 𝜋/2
5. 𝑓 (𝑡) =
0 if 𝜋/2 < 𝑡 < 𝜋
{
𝑡 if 0 < 𝑡 ≤ 1,
6. 𝑓 (𝑡) =
0 if 1 < 𝑡 < 2

7. 𝑓 (𝑡) = 𝑡 − 𝑡2 , 0 < 𝑡 < 1

8. 𝑓 (𝑡) = sin 𝑡, 0 < 𝑡 < 𝜋

9. 𝑓 (𝑡) = cos 𝑡, 0 < 𝑡 < 𝜋

10. 𝑓 (𝑡) = 𝑒𝑡 , 0 < 𝑡 < 1

11. 𝑓 (𝑡) = 1 − (2/𝐿)𝑡, 0 < 𝑡 < 𝐿

10.5 Operations on Fourier Series

In applications it is convenient to know how to compute the Fourier series of

function obtained from other functions by standard operations such addition,
scalar multiplication, differentiation, and integration. We will consider con-
ditions under which these operations can be used to facilitate Fourier series
calculations. First we consider linearity.

Theorem 1. If 𝑓 and 𝑔 are 2𝐿-periodic functions with Fourier series

∞ ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓 (𝑡) ∼ + 𝑎𝑛 cos + 𝑏𝑛 sin
2 𝑛=1
𝐿 𝐿

and
760 10 Fourier Series

∞ ( )
𝑐0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑔(𝑡) ∼ + 𝑐𝑛 cos + 𝑑𝑛 sin ,
2 𝑛=1
𝐿 𝐿

then for any constants 𝛼 and 𝛽 the function defined by ℎ(𝑡) = 𝛼𝑓 (𝑡) + 𝛽𝑔(𝑡)
is 2𝐿-periodic with Fourier series
∞ ( )
𝛼𝑎0 + 𝛽𝑐0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
ℎ(𝑡) ∼ + (𝛼𝑎𝑛 + 𝛽𝑐𝑛 ) cos + (𝛼𝑏𝑛 + 𝛽𝑑𝑛 ) sin . (1)
2 𝑛=1
𝐿 𝐿

This result follows immediately from the Euler formulas for the Fourier
coefficients and from the linearity of the definite integral.

Example 2. Compute the Fourier series of the following periodic functions.

1. ℎ1 (𝑡) = 𝑡 + ∣𝑡∣, −𝜋 ≤ 𝑡 < 𝜋, ℎ1 (𝑡 + 2𝜋) = ℎ1 (𝑡).
2. ℎ2 (𝑡) = 1 + 2𝑡, −1 ≤ 𝑡 < 1, ℎ2 (𝑡 + 2) = ℎ2 (𝑡).

▶ Solution. The following Fourier series were calculated in Examples 4 and

5 of Section 10.2:
∞
𝜋 4 ∑ cos(2𝑘 + 1)𝑡
∣𝑡∣ = −
2 𝜋 (2𝑘 + 1)2
𝑘=0

and ∞
2𝐿 ∑ (−1)𝑛+1 𝑛𝜋
𝑡= sin 𝑡.
𝜋 𝑛=1 𝑛 𝐿
Since ℎ1 is 2𝜋-periodic, we take 𝐿 = 𝜋 in the Fourier series for 𝑡 over the
interval 𝐿 ≤ 𝑡 ≤ 𝐿. By the linearity theorem we then obtain
∞ ∞
𝜋 4 ∑ cos(2𝑘 + 1)𝑡 ∑ (−1)𝑛+1
ℎ1 (𝑡) = 𝑡 + ∣𝑡∣ = − + 2 sin 𝑛𝑡.
2 𝜋 (2𝑘 + 1)2 𝑛=1
𝑛
𝑘=0

Since ℎ1 (𝑡) is piecewise smooth, the equality is valid here, with the convention
that the function is redefined to be (ℎ1 (𝑡+ ) + ℎ1 (𝑡− ))/2 at each point 𝑡 where
the function is discontinuous. The graph of ℎ1 (𝑡) is similar to that of the half
sawtooth wave from Figure 10.9, where the only difference is that the period
of ℎ1 (𝑡) is 2𝜋, rather than 4.
The function ℎ2 (𝑡) = 1 + 2𝑡 is a sum of an even function 1 and an odd
function 2𝑡. Since the constant function 1 is it’s own Fourier series (that is,
𝑎0 = 2, 𝑎𝑛 = 0 = 𝑏𝑛 for 𝑛 ≥ 1, then letting 𝐿 = 1 in the Fourier series for 𝑡,
we get
∞
4 ∑ (−1)𝑛+1
ℎ2 (𝑡) = 1 + 2𝑡 = 1 + sin 𝑛𝜋𝑡.
𝜋 𝑛=1 𝑛
◀
10.5 Operations on Fourier Series 761

Differentiation of Fourier Series

We next consider the term-by-term differentiation of a Fourier series. This

will be needed for the solution of some differential equations by substituting
the Fourier series of an unknown function into the differential equation. Some
care is needed since the term-by-term differentiation of a series of functions
is not always valid. The following result gives sufficient conditions for term-
by-term differentiation of Fourier series. Note first that if 𝑓 (𝑡) is a periodic
function of period 𝑝, then the derivative 𝑓 ′ (𝑡) (where the derivative exists)
is also periodic with period 𝑝. This follows immediately from the chain rule:
Since 𝑓 (𝑡 + 𝑝) = 𝑓 (𝑡) for all 𝑡,

𝑑 𝑑
𝑓 ′ (𝑡 + 𝑝) = 𝑓 (𝑡 + 𝑝) = 𝑓 (𝑡) = 𝑓 ′ (𝑡).
𝑑𝑡 𝑑𝑡
Theorem 3. Suppose that 𝑓 is a 2𝐿-periodic function that is continuous for
all 𝑡, and suppose that the derivative 𝑓 ′ is piecewise smooth for all 𝑡. If the
Fourier series of 𝑓 is
∞ ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓 (𝑡) ∼ + 𝑎𝑛 cos + 𝑏𝑛 sin , (2)
2 𝑛=1
𝐿 𝐿

then the Fourier series of 𝑓 ′ is the series

∞ ( )
∑ 𝑛𝜋 𝑛𝜋𝑡 𝑛𝜋 𝑛𝜋𝑡
𝑓 ′ (𝑡) ∼ − 𝑎𝑛 sin + 𝑏𝑛 cos (3)
𝑛=1
𝐿 𝐿 𝐿 𝐿

obtained by term-by-term differentiation of (2). Moreover, the differentiated

series (3) converges to 𝑓 ′ (𝑡) for all 𝑡 for which 𝑓 ′ (𝑡) exists.

Proof. Since 𝑓 ′ is assumed to be 2𝐿-periodic and piecewise smooth, the con-

vergence theorem (Theorem 2 of Section 10.3) shows that the Fourier series
of 𝑓 ′ (𝑡) converges to the average of the left-hand and right-hand limits of 𝑓 ′
at each 𝑡. That is,
∞ ( )
′ 𝐴0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓 (𝑡) = + 𝐴𝑛 cos + 𝐵𝑛 sin , (4)
2 𝑛=1
𝐿 𝐿

where the coefficients 𝐴𝑛 and 𝐵𝑛 are given by the Euler formulas

∫ 𝐿 ∫ 𝐿
1 𝑛𝜋𝑡 1 𝑛𝜋𝑡
𝐴𝑛 = 𝑓 ′ (𝑡) cos 𝑑𝑡 and 𝐵𝑛 = 𝑓 ′ (𝑡) sin 𝑑𝑡.
𝐿 −𝐿 𝐿 𝐿 −𝐿 𝐿

To prove the theorem, it is sufficient to show that the series in Equations (3)
and (4) are the same. From the Euler formulas for the Fourier coefficients of
𝑓 ′,
762 10 Fourier Series
∫ 𝐿
1 1
𝐴0 = 𝑓 ′ (𝑡) 𝑑𝑡 = 𝑓 (𝑡)∣𝐿
−𝐿 = 𝑓 (𝐿) − 𝑓 (−𝐿) = 0
𝐿 −𝐿 𝐿
since 𝑓 is continuous and 2𝐿-periodic. For 𝑛 ≥ 1, the Euler formulas for both
𝑓 ′ and 𝑓 , and integration by parts give
∫ 𝐿
1 𝑛𝜋𝑡
𝐴𝑛 = 𝑓 ′ (𝑡) cos 𝑑𝑡
𝐿 −𝐿 𝐿
𝐿 ∫
1 𝑛𝜋𝑡  𝑛𝜋 1 𝐿 𝑛𝜋𝑡
= 𝑓 (𝑡) cos + ⋅ 𝑓 (𝑡) sin 𝑑𝑡
𝐿 𝐿 −𝐿 𝐿 𝐿 −𝐿 𝐿
𝑛𝜋
= 𝑏𝑛 ,
𝐿
since 𝑓 (−𝐿) = 𝑓 (𝐿), and
∫ 𝐿
1 𝑛𝜋𝑡
𝐵𝑛 = 𝑓 ′ (𝑡) sin 𝑑𝑡
𝐿 −𝐿 𝐿
𝐿 ∫
1 𝑛𝜋𝑡  𝑛𝜋 1 𝐿 𝑛𝜋𝑡
= 𝑓 (𝑡) sin − ⋅ 𝑓 (𝑡) cos 𝑑𝑡
𝐿 𝐿 −𝐿 𝐿 𝐿 −𝐿 𝐿
𝑛𝜋
= − 𝑎𝑛 .
𝐿
Therefore, the series in Equations (3) and (4) are the same. ⊔
⊓

Example 4. The even triangular wave function of period 2𝜋 given by

{
−𝑡 −𝜋 ≤ 𝑡 < 0,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡),
𝑡 0 ≤ 𝑡 < 𝜋,

whose graph is shown in Figure 10.7, is continuous for all 𝑡, and

{
−1 −𝜋 ≤ 𝑡 < 0,
𝑓 ′ (𝑡) = ; 𝑓 ′ (𝑡 + 2𝜋) = 𝑓 (𝑡),
1 0 < 𝑡 < 𝜋,

is piecewise smooth. In fact 𝑓 ′ is an odd square wave of period 2𝜋 as in Figure

10.5. Therefore, 𝑓 satisfies the hypotheses of Theorem 3. Thus, the Fourier
series of 𝑓 :
( )
𝜋 4 cos 𝑡 cos 3𝑡 cos 5𝑡 cos 7𝑡
𝑓 (𝑡) = − + + + + ⋅ ⋅ ⋅
2 𝜋 12 32 52 72

(computed in Example 4 of Section 10.2) can be differentiated term by term.

The result is
( )
′ 4 1 1 1
𝑓 (𝑡) = sin 𝑡 + sin 3𝑡 + sin 5𝑡 + sin 7𝑡 + ⋅ ⋅ ⋅ ,
𝜋 3 5 7
10.5 Operations on Fourier Series 763

which is the Fourier series of the odd square wave of period 2𝜋 computed in
Eq (11) of Section 10.2. ◀

Example 5. The sawtooth wave function of period 2𝐿 given by

𝑓 (𝑡) = 𝑡 for −𝐿 ≤ 𝑡 < 𝐿; 𝑓 (𝑡 + 2𝐿) = 𝑓 (𝑡).

(see Example 5 of Section 10.2) has Fourier series

( )
2𝐿 𝜋 1 2𝜋 1 3𝜋 1 4𝜋
𝑓 (𝑡) ∼ sin 𝑡 − sin 𝑡 + sin 𝑡 − sin 𝑡+ ⋅⋅⋅
𝜋 𝐿 2 𝐿 3 𝐿 4 𝐿
∞ 𝑛+1
2𝐿 ∑ (−1) 𝑛𝜋
= sin 𝑡.
𝜋 𝑛=1 𝑛 𝐿

Since the function 𝑡 is continuous, it might seem that this function satisfies the
hypotheses of Theorem 3. However, looking at the graph of the function (see
Figure 10.8) shows that the function is discontinuous at the points 𝑡 = ±𝐿,
𝑡 = ±3𝐿, . . .. Term-by-term differentiation of the Fourier series of 𝑓 gives
( )
𝜋 2𝜋 3𝜋 4𝜋
2 cos 𝑡 − cos 𝑡 + cos 𝑡 − cos 𝑡+ ⋅⋅⋅
𝐿 𝐿 𝐿 𝐿
∞
∑ 𝑛𝜋
=2 (−1)𝑛+1 cos 𝑡.
𝑛=1
𝐿

This series does not converge for any point 𝑡. However, it is possible to make
some sense out of this series using the Dirac delta function. ◀

Integration of Fourier Series

We now consider the termwise integration of the Fourier series of a piecewise

continuous function. Start with a piecewise continuous 2𝐿-periodic function
𝑓 and define an antiderivative of 𝑓 by
∫ 𝑡
𝑔(𝑡) = 𝑓 (𝑥) 𝑑𝑥. (5)
−𝐿

Then 𝑔 is a continuous function. It is not automatic that 𝑔 is periodic. How-

ever, if
∫ 𝐿
𝑔(𝐿) = 𝑓 (𝑥) 𝑑𝑥 = 0, (6)
−𝐿

it follows that
∫ 𝑡+2𝐿 ∫ 𝐿 ∫ 𝑡+2𝐿
𝑔(𝑡 + 2𝐿) = 𝑓 (𝑥) 𝑑𝑥 = 𝑓 (𝑥) 𝑑𝑥 + 𝑓 (𝑥) 𝑑𝑥
−𝐿 −𝐿 𝐿
764 10 Fourier Series
∫ 𝑡+2𝐿
= 𝑓 (𝑥 − 2𝐿) 𝑑𝑥 since 𝑓 is 2𝐿-periodic
𝐿
∫ 𝑡
= 𝑓 (𝑢) 𝑑𝑢 using the change of variables 𝑢 = 𝑥 − 2𝐿
−𝐿
= 𝑔(𝑡),

so that 𝑔 is periodic of period 2𝐿. Therefore we have verified the following

result.
Theorem 6. Suppose that 𝑓 is a piecewise continuous 2𝐿-periodic function.
Then the antiderivative 𝑔 of 𝑓 defined by (5) is a continuous piecewise smooth
2𝐿-periodic function provided that (6) holds.
Now we can state the main result on termwise integration of Fourier series.
Theorem 7. Suppose that 𝑓 is piecewise continuous and 2𝐿-periodic with
Fourier series
∞ ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓 (𝑡) ∼ + 𝑎𝑛 cos + 𝑏𝑛 sin . (7)
2 𝑛=1
𝐿 𝐿
∫𝑡 ∫𝐿
If 𝑔(𝑡) = −𝐿 𝑓 (𝑥) 𝑑𝑥 and 𝑔(𝐿) = −𝐿 𝑓 (𝑥) 𝑑𝑥 = 0, which is equivalent to
𝑎0 = 0, then the Fourier series (7) can be integrated term-by-term to give the
Fourier series
∫ 𝑡 ∞ ( )
𝐴0 ∑ 𝐿 𝑛𝜋𝑡 𝐿 𝑛𝜋𝑡
𝑔(𝑡) = 𝑓 (𝑥) 𝑑𝑥 ∼ + 𝑎𝑛 sin − 𝑏𝑛 cos (8)
−𝐿 2 𝑛=1
𝑛𝜋 𝐿 𝑛𝜋 𝐿

where ∫ 𝐿
1
𝐴0 = − 𝑡𝑓 (𝑡) 𝑑𝑡. (9)
𝐿 −𝐿
∫𝑡
The integrated series (8) converges to 𝑔(𝑡) = −𝐿
𝑓 (𝑥) 𝑑𝑥 for all 𝑡.
Proof. Since 𝑓 is assumed to be 2𝐿-periodic and piecewise continuous and
𝑎0 = 0, Theorem 6 and the convergence theorem shows that the Fourier series
of 𝑔(𝑡) converges to 𝑔(𝑡) at each 𝑡. That is,
∞ ( )
𝐴0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑔(𝑡) = + 𝐴𝑛 cos + 𝐵𝑛 sin , (10)
2 𝑛=1
𝐿 𝐿

where the coefficients 𝐴𝑛 and 𝐵𝑛 are given by the Euler formulas

∫ 𝐿 ∫ 𝐿
1 𝑛𝜋𝑡 1 𝑛𝜋𝑡
𝐴𝑛 = 𝑔(𝑡) cos 𝑑𝑡 and 𝐵𝑛 = 𝑔(𝑡) sin 𝑑𝑡.
𝐿 −𝐿 𝐿 𝐿 −𝐿 𝐿

For 𝑛 ≥ 1 apply integration by parts using 𝑢 = 𝑔(𝑡), 𝑑𝑣 = cos 𝑛𝜋

𝐿 𝑡 𝑑𝑡 so that
𝐿
𝑑𝑢 = 𝑔 ′ (𝑡) 𝑑𝑡 = 𝑓 (𝑡) 𝑑𝑡, 𝑣 = 𝑛𝜋 sin 𝑛𝜋
𝐿 𝑡. This gives
10.5 Operations on Fourier Series 765
∫ 𝐿
1 𝑛𝜋𝑡
𝐴𝑛 = 𝑔(𝑡) cos 𝑑𝑡
𝐿 −𝐿 𝐿
𝐿 ∫ 𝐿
1 𝑛𝜋𝑡  1 𝑛𝜋𝑡
= 𝑔(𝑡) sin − 𝑔 ′ (𝑡) sin 𝑑𝑡
𝑛𝜋 𝐿 −𝐿 𝑛𝜋 −𝐿
 𝐿
( ∫ )
𝐿 1 𝐿 𝑛𝜋𝑡
=− 𝑓 (𝑡) sin 𝑑𝑡 .
𝑛𝜋 𝐿 −𝐿 𝐿

𝐿
Hence, 𝐴𝑛 = − 𝑛𝜋 𝑏𝑛 where 𝑏𝑛 is the corresponding Fourier coefficient for 𝑓 .
𝐿
Similarly, 𝐵𝑛 = 𝑛𝜋 𝑎𝑛 and using integration by parts with 𝑢 = 𝑔(𝑡), 𝑑𝑣 = 𝑑𝑡
so that 𝑑𝑢 = 𝑔 ′ (𝑡) 𝑑𝑡 = 𝑓 (𝑡) 𝑑𝑡, 𝑣 = 𝑡,
∫ 𝐿 𝐿 ∫ ∫
1 1  1 𝐿 1 𝐿
𝐴0 = 𝑔(𝑡) 𝑑𝑡 = 𝑡𝑔(𝑡) − 𝑡𝑓 (𝑡) 𝑑𝑡 = − 𝑡𝑓 (𝑡) 𝑑𝑡.
𝐿 −𝐿 𝐿 −𝐿 𝐿 −𝐿 𝐿 −𝐿

⊔
⊓

Example 8. Let 𝑓 (𝑡) be the odd square wave of period 2𝐿:

{
−1 −𝐿 ≤ 𝑡 < 0,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝐿) = 𝑓 (𝑡),
1 0 ≤ 𝑡 < 𝐿,

which has Fourier series

( )
4 𝜋 1 3𝜋 1 5𝜋 1 7𝜋
𝑓 (𝑡) ∼ sin 𝑡 + sin 𝑡 + sin 𝑡 + sin 𝑡+ ⋅⋅⋅
𝜋 𝐿 3 𝐿 5 𝐿 7 𝐿

that was computed in Eq (11) of Section 10.2. For −𝐿 ≤ 𝑡 < 0, 𝑔(𝑡) =

∫𝑡 ∫𝑡 ∫𝑡
−𝐿 𝑓 (𝑥) 𝑑𝑥 = −𝐿 (−1) 𝑑𝑥 = −𝑡 − 𝐿 and for 0 ≤ 𝑡 < 𝐿, 𝑔(𝑡) = −𝐿 𝑓 (𝑥) 𝑑𝑥 =
∫0 ∫𝑡
−𝐿
(−1) 𝑑𝑥 + 0 1 𝑑𝑥 = 𝑡 − 𝐿. Thus, 𝑔(𝑡) is the even triangular wave function
of period 2𝐿 shifted down by 𝐿. That is, in the interval −𝐿 ≤ 𝑡 ≤ 𝐿
∫ 𝑡
𝑔(𝑡) = 𝑓 (𝑥) 𝑑𝑥 = ∣𝑡∣ − 𝐿
−𝐿

with 2𝐿 periodic extension to the rest of the real line. See Figure 10.20.
Theorem 7 then gives, for −𝐿 ≤ 𝑡 ≤ 𝐿,
( )
𝐴0 4 𝐿 𝜋 𝐿 3𝜋 𝐿 5𝜋
𝑔(𝑡) = ∣𝑡∣ − 𝐿 = + − cos 𝑡 − cos 𝑡− cos 𝑡 − ⋅ ⋅ ⋅
2 𝜋 𝜋 𝐿 9𝜋 𝐿 25𝜋 𝐿
( )
𝐴0 4𝐿 𝜋 1 3𝜋 1 5𝜋
= − 2 cos 𝑡 + cos 𝑡+ cos 𝑡+ ⋅⋅⋅
2 𝜋 𝐿 9 𝐿 25 𝐿
766 10 Fourier Series

𝑦 𝑦
1 𝐿
−3𝐿 −2𝐿 −𝐿 𝐿 2𝐿 3𝐿
𝑡 𝑡
−3𝐿 −2𝐿 −𝐿 𝐿 2𝐿 3𝐿
−1 −𝐿
{ ∫𝑡
−1 −𝐿 ≤ 𝑡 < 0, (b) 𝑔(𝑡) = 𝑓 (𝑥) 𝑑𝑥
(a) 𝑓 (𝑡) = −𝐿
1 0 ≤ 𝑡 < 𝐿.

Fig. 10.20

∞
𝐴0 4𝐿 ∑ 1 (2𝑘 − 1)𝜋
= − 2 2
cos 𝑡.
2 𝜋 (2𝑘 − 1) 𝐿
𝑘=1

This can also be written as

∞
( 𝐴0 ) 4𝐿 ∑ 1 (2𝑘 − 1)𝜋
∣𝑡∣ = 𝐿 + − 2 cos 𝑡.
2 𝜋 (2𝑘 − 1)2 𝐿
𝑘=1

The constant term 𝐿 + 𝐴0 /2 can be determined from Eq. (9):

∫ 𝐿 ∫ 𝐿
𝐴0 1 1 𝐿 𝐿
𝐿+ =𝐿− 𝑡𝑓 (𝑡) 𝑑𝑡 = 𝐿 − ∣𝑡∣ 𝑑𝑡 = 𝐿 − = .
2 2𝐿 −𝐿 2𝐿 −𝐿 2 2

The constant term 𝐿 + 𝐴0 /2 can also be determined as one-half of the zero

Fourier coefficient of ∣𝑡∣:
∫ 𝐿
𝐴0 1 𝐿
𝐿+ = ∣𝑡∣ 𝑑𝑡 = .
2 2𝐿 −𝐿 2
◀

Exercises

1. Let the 2𝜋-periodic function 𝑓1 (𝑡) and 𝑓2 (𝑡) be defined on −𝜋 ≤ 𝑡 < 𝜋 by

𝑓1 (𝑡) = 0, −𝜋 ≤ 𝑡 < 0; 𝑓1 (𝑡) = 1, 0 ≤ 𝑡 < 𝜋;

𝑓2 (𝑡) = 0, −𝜋 ≤ 𝑡 < 0; 𝑓2 (𝑡) = 𝑡, 0 ≤ 𝑡 < 𝜋.

Then the Fourier series of these functions are

1 2 ∑ sin 𝑛𝑡
𝑓1 (𝑡) ∼ + ,
2 𝜋 𝑛
𝑛=odd
10.5 Operations on Fourier Series 767

∞
𝜋 2 ∑ cos 𝑛𝑡 ∑ sin 𝑛𝑡
𝑓2 (𝑡) ∼ − + (−1)𝑛+1 .
4 𝜋 𝑛2 𝑛=1
𝑛
𝑛=odd

Without further integration, find the Fourier series for the following 2𝜋-
periodic functions:
(a) 𝑓3 (𝑡) = 1, −𝜋 ≤ 𝑡 < 0; 𝑓3 (𝑡) = 0, 0 ≤ 𝑡 < 𝜋;
(b) 𝑓4 (𝑡) = 𝑡, −𝜋 ≤ 𝑡 < 0; 𝑓4 (𝑡) = 0, 0 ≤ 𝑡 < 𝜋;
(c) 𝑓5 (𝑡) = 1, −𝜋 ≤ 𝑡 < 0; 𝑓5 (𝑡) = 𝑡, 0 ≤ 𝑡 < 𝜋;
(d) 𝑓6 (𝑡) = 2, −𝜋 ≤ 𝑡 < 0; 𝑓6 (𝑡) = 0, 0 ≤ 𝑡 < 𝜋;
(e) 𝑓7 (𝑡) = 2, −𝜋 ≤ 𝑡 < 0; 𝑓7 (𝑡) = 3, 0 ≤ 𝑡 < 𝜋;
(f) 𝑓8 (𝑡) = 1, −𝜋 ≤ 𝑡 < 0; 𝑓8 (𝑡) = 1 + 2𝑡, 0 ≤ 𝑡 < 𝜋;
(g) 𝑓9 (𝑡) = 𝑎 + 𝑏𝑡, −𝜋 ≤ 𝑡 < 0; 𝑓9 (𝑡) = 𝑐 + 𝑑𝑡, 0 ≤ 𝑡 < 𝜋.

2. Let 𝑓1 (𝑡) = 𝑡 for −𝜋 < 𝑡 < 𝜋 and 𝑓1 (𝑡 + 2𝜋) = 𝑓1 (𝑡) so that the Fourier
series expansion for 𝑓1 (𝑡) is
∞
∑ (−1)𝑛+1
𝑓1 (𝑡) = 2 sin 𝑛𝑡.
𝑛=1
𝑛

Using this expansion and Theorem 7 compute the Fourier series of each
of the following 2𝜋-periodic functions.
(a) 𝑓2 (𝑡) = 𝑡2 for −𝜋 < 𝑡 < 𝜋.
(b) 𝑓3 (𝑡) = 𝑡3 for −𝜋 < 𝑡 < 𝜋.
(c) 𝑓4 (𝑡) = 𝑡4 for −𝜋 < 𝑡 < 𝜋.

3. Given that
𝜋 4 ∑ 1
∣𝑡∣ = − cos 𝑛𝑡, for −𝜋 < 𝑡 < 𝜋
2 𝜋 𝑛2
𝑛=odd

compute the Fourier series for the function

{
2 𝑡2 if 0 < 𝑡 < 𝜋
𝑓 (𝑡) = 𝑡 sgn 𝑡 =
−𝑡2 if −𝜋 < 𝑡 < 0.

4. Compute the Fourier series for each of the following 2𝜋-periodic functions.
You should take advantage of Theorem 1 and the Fourier series already
computed or given in Exercises 2 and 3.
(a) 𝑔1 (𝑡) = 2𝑡 − ∣𝑡∣, −𝜋 < 𝑡 < 𝜋.
(b) 𝑔2 (𝑡) = 𝐴𝑡2 + 𝐵𝑡 + 𝐶, −𝜋 < 𝑡 < 𝜋, where 𝐴, 𝐵, and 𝐶 are constants.
(c) 𝑔3 (𝑡) = 𝑡(𝜋 − 𝑡)(𝜋 + 𝑡), −𝜋 < 𝑡 < 𝜋.

5. Let 𝑓 (𝑡) be the 4-periodic function given by

768 10 Fourier Series
{
−𝑡 if −2 < 𝑡 < 0
𝑓 (𝑡) = 𝑡2 2 𝑡
2 − 2 if 0 ≤ 𝑡 < 2.

The Fourier series of 𝑓 (𝑡) is

∞ ( )
1 4 ∑ (−1)𝑛 𝑛𝜋 (−1)𝑛 − 1 𝑛𝜋
𝑓 (𝑡) ∼ − + 2 cos 𝑡 + sin 𝑡 .
3 𝜋 𝑛=1 𝑛2 2 𝜋𝑛3 2

(a) Verify that the hypotheses of Theorem 3 are satisfied for 𝑓 (𝑡).
(b) Compute the Fourier series of the derivative 𝑓 ′ (𝑡).
(c) Verify that the hypotheses of Theorem 3 are not satisfied for 𝑓 ′ (𝑡).

10.6 Applications of Fourier Series

The applications of Fourier series that will be considered will be applications

to differential equations. In this section we will consider finding periodic solu-
tions to constant coefficient linear differential equations with periodic forcing
function. In the next chapter, Fourier series will be applied to solve certain
partial differential equations.

Periodically Forced Differential Equations

Example 1. Find all 2𝜋-periodic solutions of the differential equation

𝑦 ′ + 2𝑦 = 𝑓 (𝑡) (1)

where 𝑓 (𝑡) is a given piecewise smooth 2𝜋-periodic function.

▶ Solution. Since 𝑓 (𝑡) is piecewise smooth it can be expanded into a

Fourier series ∞
𝑎0 ∑
𝑓 (𝑡) = + (𝑎𝑛 cos 𝑛𝑡 + 𝑏𝑛 sin 𝑛𝑡). (2)
2 𝑛=1

If 𝑦(𝑡) is a 2𝜋-periodic solution of (1), then it can also be expanded into a

Fourier series
∞
𝐴0 ∑
𝑦(𝑡) = + (𝐴𝑛 cos 𝑛𝑡 + 𝐵𝑛 sin 𝑛𝑡).
2 𝑛=1

Since 𝑦(𝑡) can be differentiated termwise,

∞
∑
𝑦 ′ (𝑡) = (−𝑛𝐴𝑛 sin 𝑛𝑡 + 𝑛𝐵𝑛 cos 𝑛𝑡).
𝑛=1
10.6 Applications of Fourier Series 769

Now substitute 𝑦(𝑡) into the differential equation (1) to get

∞
∑
𝑦 ′ + 2𝑦 = (−𝑛𝐴𝑛 sin 𝑛𝑡 + 𝑛𝐵𝑛 cos 𝑛𝑡)
𝑛=1
( ∞
)
𝐴0 ∑
+2 + (𝐴𝑛 cos 𝑛𝑡 + 𝐵𝑛 sin 𝑛𝑡)
2 𝑛=1
∞
∑
= 𝐴0 + ((2𝐴𝑛 + 𝑛𝐵𝑛 ) cos 𝑛𝑡 + (2𝐵𝑛 − 𝑛𝐴𝑛 ) sin 𝑛𝑡)
𝑛=1
∞
𝑎0 ∑
= + (𝑎𝑛 cos 𝑛𝑡 + 𝑏𝑛 sin 𝑛𝑡).
2 𝑛=1

Comparing coefficients of cos 𝑛𝑡 and sin 𝑛𝑡 shows that the coefficients 𝐴𝑛 and
𝐵𝑛 must satisfy the following equations:
𝑎0
𝐴0 = ,
2
and for 𝑛 ≥ 1,
2𝐴𝑛 + 𝑛𝐵𝑛 = 𝑎𝑛
(3)
−𝑛𝐴𝑛 + 2𝐵𝑛 = 𝑏𝑛 ,
which can be solved to give
2𝑎𝑛 − 𝑛𝑏𝑛 𝑛𝑎𝑛 + 2𝑏𝑛
𝐴𝑛 = and 𝐵𝑛 = . (4)
𝑛2 + 4 𝑛2 + 4
Hence, the only 2𝜋-periodic solution of (1) has the Fourier series expansion
∞ ( )
𝑎0 ∑ 2𝑎𝑛 − 𝑛𝑏𝑛 𝑛𝑎𝑛 + 2𝑏𝑛
𝑦(𝑡) = + cos 𝑛𝑡 + sin 𝑛𝑡 .
4 𝑛=1
𝑛2 + 4 𝑛2 + 4

For a concrete example of this situation, let 𝑓 (𝑡) be the odd square wave of
period 2𝜋:
{
−1 −𝜋 ≤ 𝑡 < 0,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 2𝜋) = 𝑓 (𝑡).
1 0 ≤ 𝑡 < 𝜋,

which has Fourier series

( )
4 1 1 1
𝑓 (𝑡) = sin 𝑡 + sin 3𝑡 + sin 5𝑡 + sin 7𝑡 + ⋅ ⋅ ⋅
𝜋 3 5 7
4 ∑ sin 𝑛𝑡
= .
𝜋 𝑛
𝑛=odd
770 10 Fourier Series

Since all of the cosine terms and the even sine terms are 0, (4) gives 𝐴𝑛 =
𝐵𝑛 = 0 for 𝑛 even, and for 𝑛 odd
−4 8
𝐴𝑛 = and 𝐵𝑛 =
𝜋(𝑛2 + 4) 𝜋𝑛(𝑛2 + 4)

so that the periodic solution is

( )
4 ∑ −1 2
𝑦(𝑡) = cos 𝑛𝑡 + sin 𝑛𝑡 .
𝜋 𝑛2 + 4 𝑛(𝑛2 + 4)
𝑛=odd

Rewriting this in phase amplitude form gives

4 ∑
𝑦(𝑡) = 𝐶𝑛 cos(𝑛𝑡 − 𝜙𝑛 )
𝜋
𝑛=odd

where √
√ 4 4
𝐶𝑛 = 𝐴2𝑛 + 𝐵𝑛2 = 1+ .
𝜋(𝑛2 + 4) 𝑛2
Thus, computing numerical values gives

𝐶1 = 0.5985
𝐶3 = 0.1227
𝐶5 = 0.0493
𝐶7 = 0.0260
𝐶9 = 0.0160

The amplitudes are decreasing, but they are decreasing at a slow rate. This
is typical of first order equations. ◀

Now consider a second order linear constant coefficient differential equation

with 2𝐿-periodic forcing function

𝑎𝑦 ′′ + 𝑏𝑦 ′ + 𝑐𝑦 = 𝑓 (𝑡), (5)

where 𝑓 (𝑡) is a piecewise continuous 2𝐿-periodic with Fourier series

∞ ( )
𝑎0 ∑ 𝑛𝜋𝑡 𝑛𝜋𝑡
𝑓 (𝑡) ∼ + 𝑎𝑛 cos + 𝑏𝑛 sin . (6)
2 𝑛=1
𝐿 𝐿

To simplify notation, let 𝜔 = 𝜋/𝐿 denote the frequency. Then the Fourier
series is written as
∞
𝑎0 ∑
𝑓 (𝑡) ∼ + (𝑎𝑛 cos 𝑛𝜔𝑡 + 𝑏𝑛 sin 𝑛𝜔𝑡) . (7)
2 𝑛=1
10.6 Applications of Fourier Series 771

Assume that there is a 2𝐿-periodic solution to (5) that can be expressed in

Fourier series form
∞
𝐴0 ∑
𝑦(𝑡) = + (𝐴𝑛 cos 𝑛𝜔𝑡 + 𝐵𝑛 sin 𝑛𝜔𝑡) . (8)
2 𝑛=1

Since
∞
∑
𝑦 ′ (𝑡) = (−𝑛𝜔𝐴𝑛 sin 𝑛𝜔𝑡 + 𝑛𝜔𝐵𝑛 cos 𝑛𝜔𝑡)
𝑛=1
∑∞
( 2 2 )
𝑦 ′′ (𝑡) = −𝑛 𝜔 𝐴𝑛 cos 𝑛𝜔𝑡 − 𝑛2 𝜔 2 𝐵𝑛 sin 𝑛𝜔𝑡
𝑛=1
,

substituting into the differential equation (5) gives

𝑎𝑦 ′′ (𝑡) + 𝑏𝑦 ′ (𝑡) + 𝑐𝑦(𝑡) =

∞
𝐴0 ∑ [(( ) )
𝑐 + 𝑐 − 𝑎𝑛2 𝜔 2 𝐴𝑛 + 𝑏𝑛𝜔𝐵𝑛 cos 𝑛𝜔𝑡
2 𝑛=1
(( ) ) ]
+ 𝑐 − 𝑎𝑛2 𝜔 2 𝐵𝑛 − 𝑏𝑛𝜔𝐴𝑛 sin 𝑛𝜔𝑡
∞
𝑎0 ∑
= + (𝑎𝑛 cos 𝑛𝜔𝑡 + 𝑏𝑛 sin 𝑛𝜔𝑡) . (9)
2 𝑛=1

Equating coefficients of cos 𝑛𝜔𝑡 and sin 𝑛𝜔𝑡 leads to equations for 𝐴𝑛 and
𝐵𝑛 :
𝐴0 𝑎0
𝑐 =
2 2
and for 𝑛 ≥ 1, ( )
𝑐 − 𝑎𝑛2 𝜔 2 𝐴𝑛 + 𝑏𝑛𝜔𝐵𝑛 = 𝑎𝑛
( ) (10)
−𝑏𝑛𝜔𝐴𝑛 + 𝑐 − 𝑎𝑛2 𝜔 2 𝐵𝑛 = 𝑏𝑛 .
This system can be solved for 𝐴𝑛 and 𝐵𝑛 as long as the determinant of the
coefficient matrix is not zero. That is, if
 
𝑐 − 𝑎𝑛2 𝜔 2 𝑏𝑛𝜔  ( )2
 = 𝑐 − 𝑎𝑛2 𝜔 2 + (𝑏𝑛𝜔)2 ∕= 0.
 −𝑏𝑛𝜔 𝑐 − 𝑎𝑛2 𝜔 2 

If 𝑏 ∕= 0 this√expression is always nonzero, while if 𝑏 = 0 it is nonzero as

long as 𝑛𝜔 ∕= 𝑐/𝑎. These two conditions can be combined compactly into
one using the characteristic polynomial 𝑞(𝑠) = 𝑎𝑠2 + 𝑏𝑠 + 𝑐 of the differential
equation (5). Note that if 𝑠 = 𝑖𝑛𝜔 then

𝑞(𝑖𝑛𝜔) = 𝑎(𝑖𝑛𝜔)2 + 𝑏𝑖𝑛𝜔 + 𝑐 = −𝑎𝑛2 𝜔 2 + 𝑖𝑏𝑛𝜔 + 𝑐.

772 10 Fourier Series
√
Thus, 𝑞(𝑖𝑛𝜔) ∕= 0 if and only if 𝑏 ∕= 0 or 𝑏 = 0 and 𝑛𝜔 ∕= 𝑐/𝑎. In this case,
solving (10) for 𝐴𝑛 and 𝐵𝑛 (using Cramer’s rule) gives
( )
𝑎𝑛 𝑐 − 𝑎𝑛2 𝜔 2 − 𝑏𝑛𝜔𝑏𝑛
𝐴𝑛 = 2
(𝑐 − 𝑎𝑛2 𝜔 2 ) + (𝑏𝑛𝜔)2
( ) (11)
𝑏𝑛 𝑐 − 𝑎𝑛2 𝜔 2 + 𝑏𝑛𝜔𝑎𝑛
𝐵𝑛 = 2
(𝑐 − 𝑎𝑛2 𝜔 2 ) + (𝑏𝑛𝜔)2

Thus we have arrived at the following result.

Theorem 2. If 𝑓 (𝑡) is a 2𝐿-periodic piecewise continuous function, the dif-
ferential equation (5) with characteristic polynomial 𝑞(𝑠) has a unique 2𝐿-
periodic solution whose Fourier series (8) has coefficients computed from
those of 𝑓 (𝑡) by (11), provided that 𝑞(𝑖𝑛𝜔) ∕= 0 for all 𝑛 ≥ 1.

Example 3. Find all 2-periodic solutions of

𝑦 ′′ + 10𝑦 = 𝑓 (𝑡) (12)

where 𝑓 (𝑡) is the 2-periodic odd square wave function

{
−1 −1 ≤ 𝑡 < 0,
𝑓 (𝑡) = ; 𝑓 (𝑡 + 2) = 𝑓 (𝑡).
1 0 ≤ 𝑡 < 1,

▶ Solution. The Fourier series for 𝑓 (𝑡) is

4 ∑ sin 𝑛𝜋𝑡
𝑓 (𝑡) ∼ .
𝜋 𝑛
𝑛=odd
√
Since 𝑛𝜋 ∕= 10 for any 𝑛 ≥ 1, there is a unique 2-periodic solution 𝑦𝑝 (𝑡) of
(12) with Fourier series (whose coefficients are computed from (11))

4 ∑ 1
𝑦𝑝 (𝑡) = sin 𝑛𝜋𝑡.
𝜋 𝑛(10 − 𝑛2 𝜋 2 )
𝑛=odd

If we calculate the first few terms of this Fourier series we see

( )
4 1 1 1
𝑦𝑝 (𝑡) = sin 𝜋𝑡 + sin 3𝜋𝑡 + sin 5𝜋𝑡 + ⋅ ⋅ ⋅
𝜋 10 − 𝜋 2 3(10 − 9𝜋 2 ) 5(10 − 25𝜋 2 )
∼
= 9.76443 sin 𝜋𝑡 − 0.00538 sin 3𝜋𝑡 − 0.00108 sin 5𝜋𝑡 − 0.00038 sin 7𝜋𝑡 + ⋅ ⋅ ⋅
∼
= 9.76443 sin 𝜋𝑡 + (small terms).

Since all the terms in the Fourier series except for the first sine term are
extremely small, the effect of the square wave input (forcing function) is an
output that is essentially a single sinusoidal term
10.6 Applications of Fourier Series 773

𝑦𝑝 (𝑡) ∼
= 9.76443 sin 𝜋𝑡.

It is typical of second order differential equations to very rapidly reduce the

effect of the higher order frequencies of the periodic input (forcing) function.
◀

Example 4. Suppose a mass-spring-dashpot system has mass 𝑚 = 1 kg,

spring constant 𝑘 = 25 kg/sec2, damping constant 𝜇 = 0.02 kg/sec, and
2𝜋-periodic forcing function 𝑓 (𝑡) defined by
𝜋
𝑓 (𝑡) = − ∣𝑡∣ , −𝜋 < 𝑡 < 𝜋.
2
Thus, the equation of motion is

𝑦 ′′ + 0.02𝑦 ′ + 25𝑦 = 𝑓 (𝑡). (13)

Find the unique 2𝜋-periodic solution 𝑦(𝑡) for this problem.

▶ Solution. Start by expanding 𝑓 (𝑡) into a Fourier series (see Example 4

of Section (10.2)):
( )
4 cos 𝑡 cos 3𝑡 cos 5𝑡 cos 7𝑡
𝑓 (𝑡) = + + + + ⋅ ⋅ ⋅
𝜋 12 32 52 72
4 ∑ cos 𝑛𝑡 (14)
= .
𝜋 𝑛2
𝑛=odd

The unique 2𝜋-periodic solution is

∑
𝑦(𝑡) = 𝐴𝑛 cos 𝑛𝑡 + 𝐵𝑛 sin 𝑛𝑡
𝑛=odd

where where coefficients 𝐴𝑛 and 𝐵𝑛 are computed from (11):

4(25 − 𝑛2 ) 0.08
𝐴𝑛 = and 𝐵𝑛 = .
𝑛2 𝜋((25 − 𝑛2 )2 + (0.02𝑛)2 ) 𝑛𝜋((25 − 𝑛2 )2 + (0.02𝑛)2 )

Using the phase amplitude formula, each term 𝑦𝑛 (𝑡) = 𝐴𝑛 cos 𝑛𝑡 + 𝐵𝑛 sin 𝑛𝑡
can be represented in the form 𝐶𝑛 sin(𝑛𝑡 − 𝜃𝑛 ) where
√ 4
𝐶𝑛 = 𝐴2𝑛 + 𝐵𝑛2 = √
𝑛2 𝜋 (25 − 𝑛2 )2 + (0.02𝑛)2

and the solution 𝑦(𝑡) can be expressed as

∑
𝑦(𝑡) = 𝐶𝑛 sin(𝑛𝑡 − 𝜃𝑛 ).
𝑛=odd
774 10 Fourier Series

The first few numerical values of 𝐶𝑛 are

𝐶1 = 0.0530
𝐶3 = 0.0088
𝐶5 = 0.5100
𝐶7 = 0.0011
𝐶9 = 0.0003

Thus

𝑦(𝑡) ∼
= (0.0530) sin(𝑡 + 𝜃1 ) + (0.0088) sin(3𝑡 + 𝜃3 ) + (0.5100) sin(5𝑡 + 𝜃5 )
+ (0.0011) sin(7𝑡 + 𝜃7 ) + (0.0003) sin(9𝑡 + 𝜃9 ) + ⋅ ⋅ ⋅ .

The relatively large magnitude of 𝐶5 is due to the fact that the quantity
(25 − 𝑛2 )2 + (0.02𝑛)2 in the denominator is very small for 𝑛 = 5, making
(0.5100) sin(5𝑡 + 𝜃5 ) the dominate term in the Fourier series expansion of
𝑦(𝑡). Thus, the dominant motion of a spring represented by the differential
equation (13) almost a sinusoidal oscillation whose frequency is five times
that of the periodic forcing function. ◀

Exercises

1. Let 𝑓 (𝑡) be the 2-periodic function defined on −1 < 𝑡 < 1 by

{
0 if −1 < 𝑡 < 0
𝑓 (𝑡) = .
1 if 0 < 𝑡 < 1

Find all 2-periodic solutions to 𝑦 ′′ + 4𝑦 = 𝑓 (𝑡).

2. Find all periodic solutions of

∞
∑ 1
𝑦 ′′ + 𝑦 ′ + 𝑦 = cos 𝑛𝑡.
𝑛=1
𝑛3

3. Find the general solution of

∞
∑ 1
𝑦 ′′ + 𝑦 = 2
cos 𝑛𝑡
𝑛=1
𝑛

Hint: First
∑∞ solve 𝑦 ′′ + 𝑦 = cos 𝑡 by undetermined coefficients. Then let
𝑓 (𝑡) = 𝑛=2 𝑛 cos 𝑛𝑡 and solve 𝑦 ′′ + 𝑦 = 𝑓 (𝑡).
−2
10.6 Applications of Fourier Series 775

4–7. Find the unique periodic solution of the given differential equation.
4. 𝑦 ′′ + 3𝑦 = 𝑓 (𝑡) where 𝑓 (𝑡) is the 2𝜋-periodic function defined by 𝑓 (𝑡) = 5
if 0 < 𝑡 < 𝜋 and 𝑓 (𝑡) = −5 if −𝜋 < 𝑡 < 0.

5. 𝑦 ′′ + 10𝑦 = 𝑓 (𝑡) where 𝑓 (𝑡) is the 4-periodic even function defined by

𝑓 (𝑡) = 5 if 0 < 𝑡 < 1 and 𝑓 (𝑡) = 0 if 1 < 𝑡 < 2.

6. 𝑦 ′′ + 5𝑦 = 𝑓 (𝑡) where 𝑓 (𝑡) is the 2-periodic odd function defined by

𝑓 (𝑡) = 𝑡 if 0 < 𝑡 < 1.

7. 𝑦 ′′ + 5𝑦 = 𝑓 (𝑡) where 𝑓 (𝑡) is the 2-periodic even function defined by

𝑓 (𝑡) = 𝑡 if 0 < 𝑡 < 1.

8–10. The values of 𝑚, 𝜇, and 𝑘 for a mass-spring-dashpot system are given.

Find the 2𝐿-periodic motion expressed in phase amplitude form
∞
𝐴0 ∑ 𝑛𝜋𝑡
𝑦𝑝 (𝑡) = + 𝐶𝑛 sin( − 𝜃𝑛 )
2 𝑛=1
𝐿

of the mass under the 2𝐿-periodic forcing function 𝑓 (𝑡). Compute the first
three nonzero terms 𝐶𝑛 .
8. 𝑚 = 1, 𝜇 = 0.1, 𝑘 = 4; 𝑓 (𝑡) is the force in Problem 4.

9. 𝑚 = 2, 𝜇 = 0.1, 𝑘 = 18; 𝑓 (𝑡) is the 2𝜋-periodic odd function defined by

𝑓 (𝑡) = 2𝑡 if 0 < 𝑡 < 𝜋.

10. 𝑚 = 1, 𝜇 = 1, 𝑘 = 10; 𝑓 (𝑡) is the force in Problem 5.