PART C

FOURIER ANALYSIS
AND PARTIAL
DIFFERENTIAL EQUATIONS
Chapter 10 Fourier Series, Fourier Integrals,
Fourier Transforms
Chapter 11 Partial Differenti.al Equations

Periodic phenomena occur quite frequently in physics and its engineering applica- .
tions, and it is an important practical problem to represent the corresponding periodic
functions in- terms of simple periodic functions such as sine and cosine. This leads
to Fourier series, whose terms are sine and cosine functions. Their introduction by
Fourier (after work by Euler and Daniel Bernoulli) was one of the most important
events in the development of applied mathematics. Chapter 10 is primarilycohcerned
with Fourier series. The corresponding ideas and techniques can also be extended
to nonperiodiC' phenomena. This leads to Fourier integrals and Fourier transforms
(Sees. 10.9-10.1 I), and a common name for the whole area is Fourier analysIs.
Chapter I J is concerned with the most important partial differential equations of
physicS and engineering. Jn this area, Fourier analysis has its most important ap-
plications, as a basic tool in solving boundary and initial value problems in mechanics,
heat flow, electrostatics and other fields .
.

581
 Chapter 10
Fourier Series,
Fourier Integrals,
Fourier Transforms
Fourier series! (Sec. lO.2) arc series of q)sine and sine term,s and arise
,in the importar.t practical task' 'of fe,prese,nting general periodic func-
tions. They con'stitute a very imPortant tool-in s'ol.ving problems that
involve ordinary and partial differential equations.
In the present chapte( we discuss basic concepts. facts and tech-
niques -in connection with Fourier series. Illustrative examples and '
some important engineering applications will be included. Further ap-
plicationswill be shown in the next chapter on partial differentia:) equa-
tions and initial and boundary value problems. ,
The theory of Fourier series is rather coniplicated~ but the application
of these series js simple. -Fourier series are, in a certain sense, more
universal than Taylor series, because many discontinuous periodic
functions of practical interest can be developed in Fourier series, but,
of course, do not have Taylor series representations.
The last three sections of this chapter concern Fourier integrals and
Fourier transformations, which extend the ideas and techniques of Fourier
series to nonperiodic functions defined for-all x. Corresponding appli-
cations to partia1_ differential equations wiU be- considered in the next
chapter (Sec. .1'1.14).
Prerequisite for this chapter: elementary integral calculus.
Sections that may be omitted in a shorter cour.'ie: 10.6---10.11.
Refe,.ences: Appendix 1. Part C.
Answers 10 problems: Appendix 2.

1JEAN.8AM'ISTE JOSEPH FOUR1ER (1768-1830). French physicist and mathematician.

lived and taught in I)aris. accompanied Napoleon to Egypt and was laier m;ldc prefect of
Grenobl~. He utilized Fourier series in his main work TI,eoTie "rfaiy'iqlle de la chaleuT (Allalytic
Theory o/Heat. Puris 1822) in which he de~eloped the theory ofhcat conduction (heat equation.
cr. Sec. 11.5). These new series became a most important tool in mathematical phy'sics and
also had COnsiderable inftucllCe on the further developmcnl of mathemalics itself; see Ref. l13}
in Appendix I.

,SEC. 10.1 PERIODIC FUNCTIONS. TRIGONOMETJlIC SERIES 583

Periodic Functions.
Trigonometric Series
A fum~tion f(x) is said to be periodic if it is defined for all real x and if there
,'is sO[lle positive -number p such that

{I) f(x + p) 'C' f(x) for all x.

This humber p i~ then called t;eriod 2

~t)j(x). The graph of such a function
is obtained by p~riodjc repetifion'.9f.J!s."graph in any interval of length p (Fig.
-, 237)., Periodic phenomena and functions occur in many applications.
Familiar examples of periodic functions are the sine and cosine functions,
and we note that the function f = c = canst is also a periodic fuhction in
the sense of the definition, because it satisfies, (I) for every positive p.
Examples of functions that are not periodic are x, x 2 , x 3 , eX and In x, to
mention just a few:,_
From (I) we hilVe f(x +,2p) ~ f[(x + p) + p] ~ f(x .+p)~ f(x), etc.,
and for any integer n,'

(2) 'f(x + np) ~ f(x) for all x.

Henc~ 2p. 3p, 4p,' .. are also periods of f(x).Furthermore, if f(x) and g(x)
have period P. then the function

"(x) ~ af(x) + bg(x) (0, b constant)

also has the period p.

OUf problem in the first few secti~ns of this chapter' will be the represen-
. talion of various functions of period p == 21T in tcrnis of the simple fUllctions

f{x}

'Fig. 237. PeriOdic function

21f <1- P.crlodic function I(x) has a smallest period p (> 0): thisisoften called the primlllvc
period of I(x),_ For: example, the primitive periods of sin x and sin 2xNe 211 and~,:rr. respectively.
Examples ofpcriodicTunctions withoufprimilive-period are f,~ cOflsl and f{x), '= 0 (x rational),
I(x) = I otherwise. . . _
 FOURIER SERIES.INTEGRAtS. TRANSFORMS CHAPjO.

(3) I. cosx, sinx. cos2x, sin2x~"., cos:'n:x'~" sin~.,,.,:",

which have the period 2,,(Fig. 238). The se.ries that will arise in this con
nection will be of the form '

(4) aD + a1 cos x + b~ sin x +.:a2-cos 2x + h2 ~in 2x + ...

whe're aD. 01' a 2 , hi' b2 are real constants. st;'ch a series is called
a trigonometric series; and the an and bn are calleq the ~fficients of the
series. Using the s~mmation sign,3 we may write this series

(4) 00 + 2:,(on cos 'nx + hn.,sin ni).

n=1

The set of functions"{i} from which we have,~p the series (4) is often
caned the trigonometrjc system, to have a shOrtname for it.
We see tbat each term of the series (4) has the- period 21T. Hence if the
series (4) converges. its slim will be ajllnction ojperiod 2.... .
Periodic functions that occur in engineering problem's are' often rather
complicated, and it is desirable 10 represent these..functions in terms of simple
periodic functions. We shall see that almost any periodic fllnction f(x) of
' \ period 21T that appears in applications-for example. in connection with
vibrations-can be represented by a trigonometric series, and in the next
section we shall derive formulas for the coefficients in (4) in terms of j(x)
such that (4) converges and has th~ sum f(x). In Sec. 10.3 we shall extend
\ our results to functions of arbitrary period; this extension will turn out to
\, be quite simple. ?

In the next section we define Fourier series, the most importam concept
in the Whole chapter. These are trigonometric series w.ith coefficients ob-
tained from a given function by integration [by the Euler formulas (6) in the
next section].

0. I.\_~ __DLlI'j
0.V J. J
0- ~~-~ 2~ o \j--\]2<r
cos x cos2x tos3x

n_."LLj'
-----'Cl- -.--V 0CWrj'
sin :t' sin2x'- sin 30t

Fig. 238. Cosine and sine functions having the period 2"

3And inserting parentneses, which y"ieJds from a conv~rgent series a c~nvergent sedes having
the same sum.
 '-C,,\'--'-"_ . ',-1" ')~~~,,- ......'>1-\ \ ~ -,.,''t i- " J;)

SEC. 10.1 ~Oe~dNCTlONS.-TIlIGONOMETRIC SERIES 585

(' 0:>5 {X~);; (iJ3~<,>,G<>,?%;'" ~f1)( hnll
'" + U:fi~/ "-, J

Problems for Sec. 10.1 'h~c..(:~~:t):;: 'W'I,<-(;.I),'t1i" -\ ~'s.)<----:'}.I'> ..ll

- ,~y
Find the smallest positive period p of the following functions. -
1. cos x, sin x, cos 2X; sin 2x, cos it\l-sin 7TX, cos 27TX, sin 27TX
!:l.."t ~ 21TX. 21TX " 27Tnx., 21Tnx
2. cos /lX, Sin nx, cos k' Sin k ' ~os -k-' Sin -k-
,
_ o,f-C'!;--b'1
r,'-
V
'3. If f(x) and g(x) have period p, show that h = af + bg (a, b constant) has the--
(Jl-) ~ (x--\i')..)-iJ<i?)period p. Thus all functions of period p form a vector space. '
-I'}\it
--.' '.~\-
r) ,!,ttf 4. If p is a period of f(x), show that np, n = 2,3, ... is a period of f(x).
;lj 5. Show that'the function f(x) = const is a periodic function of period p for every
"ft".}-.~ /~~: positive p.
<;;:G\(J:,tlSX ~F_'S1"6~If)f(x) is a periodic function of x of period p, show thatf(ax), a,p. 0, is a periodic
-,",," "',', ftl~ction of x of period pIa, and f(xlb), h = 0, is a periodic function of x of period
'0 d>.l-c.\s-..z "~ b. t-is:;~~~' Verify these results for f(x) = cos x, a = b = 2.

Graph the following functions I{x), which are assumed to be periodic of period '21T
and, for -7T < X < 1T, are given by the formulas
7. f(xl ~ x 8. f(xl ~ Ixl
9. f(xl ~ !xl' 10. I(xl ~ 2x - x'
!l. f(xl ~ 'sin x/2 12. !(xl ~ leos 2x1,
13. I(xl ~ 1 - e-I,I

14. I(xl ~ G if
if
-1T<X'<O
O<x<11"
15. I(xl ~ { 0
Sin X
if
if
-rr<x<O

O<x<11"
if -1T<X<O_ if -1T<X<O
16. [(xl .~ r+x 17. I(xl
,,-x if O<X<1T = {cos: lx if O<X<1T
Plot accurate graphs -of the following functions.
18. sin x, sin x + ! sin 3x, sin x + ! sin 3x + ksin 5x,
-7T/4 when -1T<X<O
I(xl ~
{ and f(x + 2.-1 = I(xl
1T14 when O<X<7T

19. sin 21TX, sin 21TX + ! sin 67TX, sin 27TX + Asin 61TX + ! sin lO7Tx
20. sin x, sin x - ! sin lx, sin x - ! sin 2x + sin 3x, !
f(x) "" x/2 when -11" < x < 'IT and f(x + 2rr) = f(x)
21. -cos x, -cos x + cos lx, -,cos x t + t cos 2x - ~ cos 3x,
f(x) = x 2/4 - 1T2/12 when -7T < x < 11" and f(x + 2'IT} = f(x)

Evaluate the following integrals where n = 0, 1,2, . "; .. (These are typical 'examples
of integrals that will be nt:;eded in our further work.)

fo," cos
22. /IX dx 23. J
- .. xcosllxdx 24. J-." .. x sin /IX dx

25. f'" sin IIX dx 26. f xsinllxdx 27. J" eX cos IIX dx

f:.
o 0 0

28. x sin IIX dx 29.

( eX sin,nx d;r 30. f'" x
0
2 ~os nx dx
),:J __ q,n "x- Jx:
>-"\1 -=- _~__ (;:,s'!)--f.
 FOURIER SERIES. INTEGRALS. TRANSFORMS CHAP. 10

Fourier Series
Fourier series arise from the task of representing a given periodic function
f(x) by a trigonometric series. The Fourier series of f(x) is a trigonometric
series whose coefficients are detenriined from Ifx) by certain formulas [",Eu-
ler formulas" (6), below] which we shan derive first. Later in this section
we take a look at the theory of Fourier series.
\
Euler Formulas for the Fourier Coefficients
Let us assume that f(x) is a periodic function of period 21T that can be
represented by a trigonometric series.,

(I) fIx) = aD + :E (a. cos ax + b. sin ax);

n=1

that is, we assume that this series converges and 'has f{x) as its sum. Given
such a function f(x), we wanl to determine the coefficients an a~d bn of the
corresponding series (1).
We first determine ao' Integrating Qn bQth sides Qf (1) from -7T..tQ :rf; y.re
have '
r . ..

r)(; fX ~ (.1:- + ,~, (lin cos /IX + bn sin /IX)} dx ..

If term-by:-term integ~i~n of the series is allowed. 4 then we Qbtain
f. f(x) dx = liD f. dx + n~' (an f. COS nx dx + bn f. sin /IX elf)' I
The first term Qn the right equals 27Ta o. All 'the Qtli~r'iritegrals"on.the-.rtght
are zero.. as can be readily seen by performing th~ in'tegratioJls~;Hence,our .. ,_
first result is

(2) aD =-
I
2'IT
f-.
f(x) dx.

We now determine al' a2 by a similar procedure. We mUltiply (I) by

cos mx. where m is any fixed positive integer, and integrate from -17' to. 17':

(3) f.f(X)~OSmXdx =
3,,'-
f.h + ,~, (anC6s'/lX+.~nSinll;)lcosm.xdx,
"

4This is jus'tifie~. fof instance. in the case of uniform convergenc;'lCC. Theorem 3 -in Sec';
14.8). . . .
5EC.l0.2 FOURIER SERIES 587

Integrating term by term, we see that the right-hand side becomes

The first' integr~l1. is zero. By applying (II) -in Appendix 3.1 we obtain

, = "2 J I'
+ m)x dx + 2" J~
If!
I
-~
cos nx cos mx dx
-n
cos (/1 cos (IT - m)x dx,

J ~innxcosmxdx=2J
'IT

-'IT
I W

-1r
sin(n+m)xdx+_
_ I
2 J sin(n-m)xdx.
-'IT
'IT

Integration shows that the four terms on the right are zero, except for the
last term in the first line, which equals 11 when n = m. Since in (3) this term
is multiplied by am' the right-hand side in (3) is equal <fod our second tcra:;,
result is ,- ~

(4) ICol~2
~ .
..

We finally determine b" b2 , in (I). lfwe m~ltiply (I) by sin mx, where
m is any fixed positive integer, and then int~gta{e from -7r to 17. we have

(5) . (f(X) sin mxdx ~ fJ o nt a + (an ~osnx + b n sin nXlJSin'mXdX.

Integrating term by term, we see that the right-hand side becomes
",: .

aoI;:inmxdx + n~1 [anI;" co~nxsin~xdx + bn I;:inllxsinmxdx l

~ ~

'The first integral is zero. The next integral is of the type considered b~rure.
and we know that it is zero for all n 1, 2, .... For the last in'tegral we
obtain '

. 1 'COS(Il-m)xdJ-:zf
, sinnxsin-,nxdx=2f 0 I '
I -7r -11" -_11"
cos(lI+m)xdx.

The last term 'is 'ler,o. The~first .term on t~e right is iero when It =F In a~d i~-
7r when. II ~ 1/1. Since in (5) this terl1):oismultipliedby b,)he right-hand
side in (5) is eqllal to bm 1T. and our 11st result is m~"
-~"
bm ~ -I
7f,
I" "":'11"
f(x) sin mxdx,
..
'L-'m = 1,2, " .
r" ,

',.,"-,-,_0-
588 FOURIER SERIES,INTEGRALS, TRANSFORMS CHAP. to

Writing n in place of m,~we ~together have the so-called Euler formulas 5

(8) >;0 :2'

/ 'IT
r -11'
f(x) dx
'

, f' f(x) cos nx dx n = 1,2,' .. ,.',

(6) (b) an = ;.
-.
(c) n 1, 2, ....

'These numbers given by (6) are called the Fourier coefficients of f(x). The
trigonometric series .

(7) ao + ~ (an cos nx + bn sin nx)

n=l ! ,~

with coefficients given by (6) is called the Fourier series of fIx) (regardless
of convergence-we discusS this below). '. ,

EXAMPLE 1. Square wave

Find the Fourier coefficients of the periodic fl.!RetioR f(x) in Filv239a. The analyti~ reprl!sen-
tation is '
if -1T<X<O
and f(x + 2'11") = lex}.
if o <x < 'IT
Functions of this type may occur as extemaUorces acting on m,ch"ni<'~I"y":em~;'~e<~~~'rive
forces in electric circuits,- etc.
Solution. From (6a) we obtain "0 = O. This can also be seeR witho"t i,~fegration.
under the curVe of f(x) between -wand 'IT is zero. From (6b),

an =;I f"_'lr!(x)cosnxdx=;~I [f'_tr'(-k)cosnxdx+

. J."0 kcosnx~ J
=-I [kSinnxl'
'IT'
- --
n
+ kSinnxl'
- - ] =0,
n . -'Ir 0

because sin nx ::: 0 at -1t', 0 and 1t' for ail~n '7' t. 2, .... Similarly, from (6c) we obtai~

bn "" .; f:. . !(x) sin nx dx,= ;. {f~1I" (-k) sin nx dx + ,~'Ir k sin nx dx]
='~ [k co: nx C-k I:]. . . co:nx

Since cos ( - a) "" co~ a and cos 0 = I. this Yields

bn "" .!.(cosO - cos (-n1t') - COsn1t' +'coS'Ol

n-zr ' ,-
=,~.2k (l4:,;e'~:'ii1t').
'n11' " .."

5Students familiar' ",ith Sec. 4'.7 may have noticed that in this d'erivation
used 'the t1hogonality of the trigonometric SY,stem (3), Sec. J.O.I~ ori an in~erVal
SEC. 10.2. FOURIER SERIES 589

~
-11"

y-k .0 '11"

L-..J
211"

(a) The given function f(x) (Periodic square wave)

. -~

\:~ sin3x

(b) The first three partial sums of the corresponding Fourier series

Fig. 239. Example 1

Now, cos 11" = -I. COS21T = I, cos 31r~7'- -I elc., ingeneral,_

_-I foroddn, 20 for odd n,

cos II1'/' = {
I for even II,
{
for even II.

Hence the Fourier coefficients bn of our function a'r;

4k
bs_= s;. .... ,

(8) 4k(
- I.S i
sinx'+ -sin:1;J.;+_'-sin5x +",' ")"
..
'1r 3 _,5
The partial sums are
 FOURIER SERIES. INTeGRALS. TRANSfORMS CHAP. \0
590

4k
Sl = -; sin x,
i )
.4k(sinx'+'3sin3x.
S2=-; etc.,

and their graphs in Fig. 239 seem to indicate that tl)c series is convergent and has the sum i(x}, .
the given function. We notice that at x = 0 and x = 1f, the points of discont,inuity of f(x). all
partial sums have the v~lue zero. the arithmetic mean of the values - k and k of our function.
Furthennore, assuming that f(x) is the sum of the series and' setting 'x = 1r12. we have

0'

This is a famous result by Leibniz (obtained about 1673 from geometrical considerations). It
illustrates that the values of various series with constant terms can be obtained by evaluating
. Fourier series at specific points: I

In this whole chapter we consider Fourier series from- a practical point of

view. We shall see that the application of these series is rather simple. In
contrast to. this. the theory of these series is complicated. and we--shaH 'nDt
go. into. any details o.f it. Acco.rdingly, we confine ourselves lo. a theorem Gn
the convergence and the sum of Fourier series. which we present next.

Convergence and Sum of.Fourier Series

Suppose that f(x) is any given periodic function of period 27T for which the
integrals in (6) exist; for instance, f(x) is continuous or merelY'_piecewise
continuous (continuous except for finitely many finiteJ~rrips.in the'inte~val_
-of integration). Then we can compute the Fourier coefficients (6) 'of f(x) and
use them to form the Fourier series (7) of f(x). It would be nice if the series
thus obtained converged and had the sum f(x). Most functions appearing in
applicatio.ns are such that this is true (except at jumps of f(x), which -we
discuss below). In this case, in which the Fourier series of f(x) does represent
f(x). we write

f(x) ao + L (an cos nx + bn sin nx)

n""l

with an equality sign. If the Fourier series of f(x) does not have the sum
.f(x) or does not converge. one still writes

f(x) - 0. + L
n=l
(on cos nx + bn sin nx)

with a tilde -_~ _\vhi~h indic~~es that the trigono~et~ic 'series o.n the right has
the Fourier coe-fficients of f(x) as its coefficients. so-:itds.the Fourier series-
of f ( x ) . " ," ".
The class of functions- tha.t- can be r~fesented by Fourier series'is shr-
prisingly large and .general.' Correst?O"rldibg sQfficieni -conditions co~enhg
almost any .conceivable engineering';application a~ as follows. .
 SEC. 10.2 FOURIER SERIES. 591

Theorem 1 (Representation by a Fourier series)

. If a periodic junction f(x) with period 21T is piecewise contim.lOlis fi ill the
interval -7T ~ x 2 1f and has a left-hand derivative alld right-hand derivative 7
at each point of that-interval. thell tlte Fourier series (7) of f(x) [with coeI-
is
, Jicients (6) is convergent. Its sum f(x), except at a point Xo at which f(x)
is discontlnuou~ Glid the sum of the series is the average of the le/t- and
right-hand limits 7 of f(x) at XO'

Proof of convergence in Theorem 1 for a continuous function !(xI

having continuous first and second derivatives. Integrating (6b) by
parts, we get
,
I fT/", f(x) sin l
an =- j(x) cos nx dx =
n7T
I1X
" .- n 7T
'__ f_'_" f'ex) sin I1X dx.
7T
-, .
1

:The first term on the right is zero. Another integration by parts gives

.
an::::
i'(x)
.
cos
2
n7T
nx I'
-'IT
- -2-
/T1T
I ""f'"f (x)
-'IT
COS,IlX dx.

The first term on the right is zero because of the periodicity and.continuity
of i'(x). Since f" is continuous in the Jnterval of integration, we have

1f"(x)1 < M

for an appropriate constant M. Furthermore, \cos I1X\ ;;: I. It follows that

1 . . Tr
<-f-, n 2
7T
Mdx
2M
--;;2.

6Definition in Sec. 5.1.

fix) 7The left-hand limit of f(x) at Xo is defined as the limit of
f(l - 0) . f(x) as x approaehesx o from the left and is frequently denoted
by f(.t o - 0). Thus

~ /(x o - 0) = lim /(x o - 11) as h_ Othrqugh positive values.

h_O

o x
The right-hand limit is denoted by /(x o + 0) and

Fig. 240. Left and f(xo + 0) = lim /(x o + l1)ash-Othroughpositivevalues.

righthand IImll' h-O

f(1 - 0) ~ 1. The left- and right-hand derivatives of [(xl at Xo are defined

as the limits of
f(l + 0) ~ ~
It-I'o -.It) - J(Xg- ~ 0) 1(\'0 + II) - I(xQ + 0)
of,the' function -I,
and
I,
X2 if x<1
re~peetiv_eJy. as h _ 0 through positive values. Of course if
f(x),:" { x/2 if x> 1 /(x) is.-contin\Jous at Xo the last term in both numerators is
simply /(x o).

r
 FOURIER SERIES, TRANSfORMS CHAP. 10

S_imilarly, Ibnl < 2 Mln 2 for all n. -Hence the absolute value of each term .of
the Fourier series .of f(x) is at most equal to the corresponding term .of the

'1 1 I I )
lal+2M
o ( 1+1+-+-+-+-.
22 22 32 32 + ...

which is convergent. Hence thai Fourier series converges and the proof is
complete. (Readers already familiar with uniform convergence will see that,
by the Weierstrass test in Sec. 14.8, under our present assumptions the
Fourier series converges uniformly. and our derivation of (6) by integrating
term by term is then justified by Theorem 3 of, Sec. 14.8.)
The proofs .of convergence in the case .of a piecewise continuous function
f(x) and of the last statement of Theorem 1 can be found in more adv_anced
texts, for example, in Ref. [ell) listed in Appendix 1. I
EXAMPU 2. Convergence at a lump as Indicated In, Theorem 1
The square wave in Example 1 has ajump at x == O. Its left-hand limit there is - k and its right-
hand limit is k (cf. Fig. 239). so that the average of these limits is o. The Fourier series (8) of
the square wave does indeed converge to this value when x "" 0 because then all its tenns are
O. Similarly for the other jumps. Thts',is in agr.eement with Theorem -I. I

Summary-. A Fourier series of a given function f(x) of period 2'11 is a series

of the form (7) with coefficients given by the Euler formulas (6). Theorem
1 gives conditions that are sufficient for this series to converge and at each
x to have-the value f(x). except at discontinuities of J(x), where the series
equals the arithmetic mean orthe left-hand and right-hand limits of f(x) at
that point.
Most periodic functions in applications will_have periods pother than just
21T, but in the next section we show that the transition from period p = 21T
to general p = 2L (a notation convenient for later applications) is quite ~ ....
simple and direct.

Problems for Sec. 10.2

Find the Fourier series of the function f(x), which is assumed to have the period
21T. and plot accurate graphs of'the first 'three partial sums8 where f(x) equals

'l~ ~
I.

-w -m2
I _ I 0 ..2 rr -
% ~------~~--~~~~~
-rr 0 rr/2
10
3. 4. {(x)

-rr o rr % -rr
-1

N
8That is, au + L (all cos nx + bn sin nx) for N == l, 2. 3..
n=l
 SEC. 10.3 FUNCTIONS OF ANY PERIOD P = 2L 593

\ /'5. f(x) ~
{
-k if --rrI2<x<ni2
..
, {-k
6. f(x) =
if O<x<7Tfl
,,i\ k if .,,/2 < i < 3,"/2 . 0 if 712 < X < 271
7. f(x) = x (-11" < X < 1T) 8. f(x) ~ " - 21xl (-" < x < ,,)
9. f(x) = x (0 < x < 2,,) 10. f(x) = x + Ixl (-" < x < ,,)
11. f(x) ,= x 2 ( - 7T < X < 1T) 12. f(x) = x 2 (0 < x < 211-)
X if. - ''1t12 < x < 1f/2 X if-n/2<x<'rIl2
13. f(x) = 14. f(x) = {
{
o if nl2 < x < 37T/2 7T - x if 7T/2 < x <' 37f12: '
_X 2 if - nl2 < x < 7t/2
~
-1T<X<O if
15. f(x) ~ 16. f(x) { x'
{
x 2 if 0 <x< 7r 7T 2 /4 if

17. Show that if f(x) has the Fourier coefficients an' hn and g(x) has the Fourier
coefficients an *, bn*, then kf(x) + 19(x) has the Fourier coefficients kan + Ian *._
kb n + Ib n *
18. Using Prob. 17, obtain the Fourier series of f(x) = Ixl ( - 7r < X < 7T) from Probs.
7 and 10.
19. Obtain the Fourier series in Prob. 3 from that in Prob. l.
20. Verify the last statement in Theorem I concerning the discontinuities for the
function in Prob. t.

Functions of Any Period p = 2L

Periodic' functions in applications rarely have period 27T but some other
period 9 p ~ 2L. If such a function f(x) has a Fourier series, we claim that
it is of the form

(1)

with the Fourier coefficients, of I(x) given by the Eule." rormulns

1
a o ~ 2L
fL f(x) dx
-L "

(2) (b) an~ L1 fL f(x) cos

n7rx
L dx n =: 1..2 ... ,
I --L '

I (c) bn ~ L1 fL f(x)
, . II'TrX
Sin - dx n=I,2.,;.
. -L L

I . 9This notaJion is practical since in ;ppli_cationSl; L will be the length of a vibrating string (Sec.

l
11.2), of a rod in heal conduction- (Sec. 11:5), etc. .
'r
594 FOURIER SERIES. INTEGRAlS. TRANSFORMS CHAP. to

Proof. Th~ idea is to derive this from Sec. 10.2 by a change of scale. We
set v = 1rxIL. J{en~e x = Lvl1T. Then x = L corresponds to v = ::!:: 11".
Thus f. regarded as' a function of v which we call g(v), -

fIx) = g(v),
has period 27T. Accordingly, by (7) and (6), Sec .. 10.2, with v instead of x,
this 27T-periodic function g{v) has the Fourier series

(3) g(v) = ao + L (an cos nv + bn sin nv)

n.;1

with coefficients

I
a o = 27T J-.
g(v) dv

I
J
11"

an = - g(v) cos IIV -dv

(4)
-.
7T .

= ;. f
. I'
bn sin nv dv.
-. g(v)

Since v = 7TxlL and g(v) = f(x), formula (3) gives (I). In (4) we introduce
x = LV/1T as variable of integration. The"n the limits of integration v = 1T
become x = L. Also. v = 1rxiL implies dv = '1J'dxlL. Thus dvl27f = dxl2L
in ao' Similarly, dvhr = dxlL in o'n and bn . Hence (4) gives (2).. I

The interval of integration in (2) may be replaced by any interval of length

p = 2L for example. by the interval ,0 ~ x ~ 2,.

~
"] , [U -2 -1 0 2
[ x

fig. 241. Example 1

EXAMPLE 1. PeriodIc square wave

Find the Fourier series of the function (d. Fig. 241)

o if -2<x<-1
f(x)= kif -I<x< I p = 2L = 4. L = 2.,
{
o if I <x< 2

Solution.. From (2a) and (2b) we obtain

a~'=
I
-4
J2 f(x) Jx "" -I J'I k
k Jx = - _",
-2 4, _I 2-
SEC. 10.3 FUNCTIONS OF ANY PERIOD P = 2[ 595
"{;.;o,

"'I
a "= -
11 2 -2
f(x)cos-dx
2.
2'
II'rrx = -
2 -1
.
lI1TX
kcos-dx = '2k. 1I1T
-SLI1-;--.
2 ~ '.1111'" 2
'II 1 , -

Thus an = _0 if n is even and

an = 2khl1f if ,~ = 1,5,9.' an = -2klll'rr if 11 = 3,7. II..

From (Ie) we find thai b n = 0 for" :: I. 2.... Hence the result is

[(xl =
t.: 2k(Tr I 37r I 511" ) ,
2" + -; cos'2x - J"cos"Tx + scosTx -+ ....

EXAMPLE 2. Halt-wave rectifier '

A sinusoidal voltage E sin wt, where t is time. is passed through a half-wave rectifier that clips
the negative portion of the wave (Fig. 242), Find the Fourier series of the resulting periodic
function

u(O =
, {O r if -L<1<0.
p=2L=~,L=2!:.
Esin"i;F)t if O<I<L w w
~---
Solution. Since u = 0 when - L < 1 < 0, we obtain from (2a), with ( instead of x.
E
ao =.~
~1T
f
0
wi..
E sin wI d( = -
7T

and from (2b). by using rormUlat)!J~ Appendix 3.1 with X"" WI and y = IIwl.

wI" ~"
sin wI os IIwl d1 "'" - wEI", + ,
+ sm (I
an = -
1f 0
E
. "',"""
~"
J
21T 0
Ism (I
-
lI)wl - lI)wr) dl.

If II := 1,- the integral on' the right is zero, and if n "" 2. 3, .. , . we readily obtain

a _ cos (I + lI)wl _ cos (I - lI)wl ',.

II
:= wE
27r [ (I + II)W (I - lI)w ] 0

= . (-cos (I + nhl' + + 1 -cos (I - 11)11" + I),

27r 1+11 r'--It

When It is odd. this is equal to zero. and for even 11 we have

E(2'
an = 2; ~ + ~
'2) '" - (11 -
2E'
1)(11 + 1)1T
(II = 2,4." .),

In a similnr fashion we find from (241hal "l := 12 nnd "n =: 0 for 11 := 2.3 ... , Consequently,

"2(1
u(t) = ;;
E
+ 2"E.Sin wI - --; ~ cos 2wl + 0I" cos 4wl +_. "
)
.
, ' '

The function in Example 1 is even, and its Fourier series has only cosine
terms. no sine terms. This-is .typi<;al. and in the next section we show how
in'
to save, work (and avoid errors) t-he case 'Of ~ven.or odd periodic fllncti?ns.

-1T/W o 1T/W

FIg. 242. Halfwave rectifier

 596 FOUR.IER SERIES, INTEGRALS, TRANSFORMS CHAP.10 .

Problems for Sec. 10.3

Find tbe' Fourier series of the periodic Junction f(x). of period p = 2L. an,ti plot
accurate ~raphs of f(x) and the first three partial :sums.
I,f(x)~ I (-I<x<I), f(x)~O (I <x<3); p~2L~4

2. f(x) ~ I (- I < x < 0), f(x) ~ - I (0.< x < I), p = 2L = 2

3. f(x) = 0 (- 2 < x < 0), f(x) = I (0 < i < 2), p = 2L = 4
4. [(x) ~ Ixl (- 2 < x < 2), p = 2L = 4
5, [(x) = x (-4 < x < 4), 'p = 2L = 8
6. f(x) = x (0 < x < I), [(x) = I - x (I < x < 2), p = 2L ~ 2
7. f(x) ~ In (-I < x < 0), f(x) = -x (0 < x <I), p = 2L = 2
8. f(x)~(lm+x (-1I2<x<0), f(x)~(1I2)-x (0<x<1I2), p=2L= I
9. f(x) = x (- ,,/8 < x < ,,/8), f(x) = (,,/4) - x (,,18 < x < 3.-18),
p=2L=1T12
10. f(x) = x 2 (0 < X < 2), p ~ 2L = 2
II. f(x) ~ x 2 (-I < x < I), p = 2L ~ 2
12. f(x) ~ x 3 ( - I < x < I), p ~ 2L ~ 2
Il. I(x) = sin'ITx (O<x< I), p = 2L = I
14. f(x) ~ I (-2 < x < 0), f(x) ~ e-' (0 < x <2), p ~ 2L ~ 4

rrs~~~ Find the Fourier series of the periodic function obtained by-passing the voltage
-- 'g(/) = 2 cos 10011'1 through a half~wave rectifier that clips the negative portion
of the wave.
16. Obtain the Fourier series in Prob . .3 directly from that in Example I.
17. Obtain the F.ourier series in Prob. 2 directly from that in Example I, Sec. 10.2.
18. Obtain;the Fourier series in Prob. 11 directly from that in Prob. ll~ Sec. 10.2.
19. Show that each term in (3), has the period p .::: 2L.
20. Show that in (2) the interval of integnltion may be replaced by any interval of
length p = 2L.

III Even and Odd Functions

Unnecessary work (and corresponding sources of errors) in determining
Fourier coefficients of a function can be avoided ,if the function is odd or
even.
We first remember Ihat a function y ~ g(x) is said to be.v~ if
/
/' g(~x) = g(x) for all x.

The graph of such a fUQction is symmetric with respect to the y~axis (Fig.
243). A function h(x) is said to be odd if

h( -x) ~ -h(x) for all x.

. , .
(Cr. Fig. 244). The fmiction -co~, ~ is even, while the fUnction sin nx is odd.
If g(x) is an even function, tli~n
sm 10.4 EVEN AND ODD fUNCTIONS 597

y
."

Fig. 243. Even function Fig. 244. Odd function

(I) J L

-L
g(x) dx ~ 2 f0
L
g(x) dx (g even).

ljh(x) is an oddjunction, then

L
(2) J
-L
h(x) dx' ~ 0 (h odd).

Formulas (0 and (2) are obvious from the graphs of g and h, an.d we leave
the formal proofs to the student.
The product q = gh of an even function g and an odd function II is odd,
because
q(-x) ~ g(-x)h(-x) ~ g(x)[-h(x)] ~ -q(x).

Hence if f(x) is even. then the integrand f sin (nTrxiL) in (2c), Sec. 10.3, is
odd, and b n ~ O. Similarly, if f(x) is odd, then f cos (nTrxiL) in (2b), Sec.
10.3, is 'odd, and au = O. From thi~and-(l) we obtain the following theorem.
,
Theorem 1. (Fourier series of even and odd f'cl,ncflons)
The Fourier series of all even junction [(x) of peri?d 2L is a "Fonrier cosine
series"

(3) (f even)

with coefficients

(4) 0 ~
.

ffo
L

f(x)dx, an = -2 J L
n7T'x
f(x)cos-.dx,

n ~ 1,2,,, '.
L 0 L.
\ '.

The Ppuri~r series of an odd junction I(x) of period. 2L is a "Fourier sine

series"

(5) , f(xl ~. 2: Pn sin ";x (f odd)

n=1

.;
 598 FOURIER SERIES. INTEGRALS. TRANSFORMS CHAP. 10.

with coefficients

(6) bn = l.2 JL fIx) Sin

. Ln1fX dx.
o

In particular. this theorem implies that the Fourier series of an even func- '
tion f(x) of period 2L ~ 27T is a Fourier cosine series

(3*) f(x) :::: Go + at cos x + 112 cos 2x + 113 cos 3x +

with coefficients

(4*) 00 = -I
7r
J' fIx) dx.
0
an:::: -
2
7r
J
0
'
f(x) cos IlX dx. n:::: 1,2,' ...

Similarly, the Fourier series of an odd function j(x) of period 21T is a Fourier
sine series

(5*) fIx) = b l sin x + b 2 sin 2x + ba sin 3x + ...

with coefficients

2
(6*) bn = -
7r
f0
or_
fIx) sin nx dx, n:::: t, 2,7

For instance, f(x) in Example I, Sec. 10.2, is odd and, therefore, iSTep-~
resented by a Fourier sine series. -
Further simplifications result from

Th,orem 2 (Sum ollunclIons)

The FOl!rier coefficients oIa sum II + 12 are the sums of the corresponding
. Fourier coefficients of II alld f2'
Tlte Fourier coefficients of cf arc c times the corresponding Fourier coef
fil:icnfSof f.
EXAMPLE 1. Rectangular pulse
The function f"'(x) in Fig. 245 is the sum of the function /(x) in Example I otSec. 10.2 and
the constant k. Hence. from that example and Theorem 2 we condude that

f"'{x) = k + ~ (Si~x + ~sjnJx + ~sin 5x + .. -).

Rg. 245.:Example 1
 SEC. 10.4 EVEN AND ODD FUNCTIONS 599

(a) The function [(x),

o
(b) Partial sums 8" (x).

Fig. 246. Example 2

EXAMPLE 2. Saw-toothed wave

Find the -Fourier series of the f~nction (Fig. 2~6a)

f(x) '" .x + 1f if :.- 1r < x.< 1f and I(x + 21T) '" [(x).

Solution. We may write

f:: II + /2 \Vhcrc It'" x and, /2 =. 1T.

The Fourier coefficients of /2 are zero, except for the first one (the constant term), which is
'fr. Hence. by Theorem 2. the Fourier coefficients an' b n are those of f I' except fOf Go. which
is ?T. Since II is odd, an '" 0 for II '" I. 2 .... , flod

. bn '" -
2 J' f . l(x) sin /IX dx = -
2 J' . x sin /IX dx.
1T 0 1T 0

Integrating by parts we obtain

2
b =-
n1/"
[-XCO,"X n
I'
ono
+-I J' cosllxdx ] 2
- - cos 111T

"
Hence b l = 2. b 2 := - 2/2, 113 '" 2/3, "4 =~ - 2/4 ... , and 'the .Fourier series of lex) is
I
/
f( x) =,1T + 2(smx
. I. . + 35.m
- '2sm2r I. J'x.-+ ) . . I

As an important application, In' the next section we ,show that a function

f(x) defined on an interval o";;';:.x:;:;:; L can be represented by a Fourier cosine
series as well 'as by a Fourier sine, series of period p == 2L; these are called
half~range expansions of f and will be, of basic interest in Chapf 11.

,--,,",: "='=f~~""";;'~';~"'~ '''''=''''===E;a;;;aE5=="",~'='~,;,;;";,,,;,,,;,;,,,;,;,,;,;;ocil''':iiIIl

 [OURIER'SERIES, INTEGRALS, TRANSFORMS CHAP, 10
600

Problems for Sec. 10.4

Are the following functions odd, eve~,or neither odd nor even?
J. jx31. x cos nx, x 2 cos nx,
cosh<x, sinh x, sin x + cos x, xlxl
2. x + x 2, lxi, eX, ex?-, sin 2 x, x
sin x, In x, x cos-x, e-
Ixl

Are the following functions J(x), which are assumed 'to be periodic, of period 21T,
even, odd or neither even nor odd?
-X 2 if -1f<x<O
4. f(x) ~ x
if -1T<X<O
, {O
3, f(x) ~ { if
x 2 if O<x<1f

5. f(x) ~ e if
if
-'t1'-<X<
-'TT/2<X<1T
<x-<-7r)
--7f/2
6. f(x) ~ r
8. f(x) =' Isin xl
0, if
if -''If/2 < x < 1T/2

(-1T<X<1T)
1r12 < x< 3'11-/2
7. f(x) = e 1xl (-7T
9. f(x) ~ xlxl (-1T<X<ll") 10.' f(x) ~. e- I" (-1f<x<1I")

(0 < x < 21T) 12. f(x) ~ x (0 < x < 27i)

11. f(x) ~ x'

Prove:
13. The sum of odd [uodions is odd.
14. The'product qftwoodd functions'is even.
15. The sum and the product of even functions are even.
16. If f(x) is even, then If(x)l. P(x), and P(x) are even.
17. If f(x) is odd, then If(x)1 and f'(xl,are even,
-18. If- g(x) is any function, defined for all x, then-p(x) = [g(x) + g( -x)]12 is even
and q(x) ~[g(x) - g( - x)]12 is odd. and g(i) ~ p(x) + q(x).
Represent the following functions as the sum of an even and'an odd function.
19. e' 20. e,-" '21. x/(I - x) 22.' (I + x)/(I - x)

23. Find all functions that are both even and odd.
24. Prove Theorem 2.
25. Show that the familiar id-entity sin 3 x. = i sin x - i sin 3x can be interpreted as
a Fourier series expansion. and the same holds for cos 3 x = i cos x + ! cos 3x.

Find the Fourier series of the following functions, whi.ch are assumed to have the
period 27T.. Hint. Use that some of these functions are even Of odd.
26.f(x)~x (-,,<x<,,) 27.f(x)~lxl (-,,<x<,,)
.,..."''7"",-
2
28. f(x) = x 3 ( - 'IT < X < 'IT) @jf(x) = x /4 (- 7T < X < 71')
30. f(x) ~ Isin xl (-" < x < ,,) 31. f(x) ~ lxi' (-" <x < ,,)
32. f(x) = x('lT 2 - x 2) (-1T < X <: 'IT)
1 if -7T/2 <.X < 'Tr12
34. f(x) ~
'{'X - 1r if O<X<7T
33. f(x) ~ { o if n12- <.. x < 37T/2 -x if 1T<x<21T'

if :""7T/2 < < 'lf/2 _X 2 if -'If<X<O

~{
x
35. f(x) x' 36,. f(x) ~
'lf2/4 if
{
x2 if o <-x < 'If

Show that
37 I _ , ,+51 _,_71+,'_
- .. , ~ "
-,
4 (Use Prob, 33.)
 601

Show that
. ".'
38. I + t + i + -h + is' + ... = 6 (Euler's famous result. D;se PrOb. 29.)

(Use Prob. 29.)

I 1 'I 1T2
40. I + J2 + 52 + P. + ... = 8 (Use Probs. 38. 39.)

III Half-Range Expansions

In various physical and engineering problems there is a practical need to
use Fourier series in connection with functions f(x) that are given merely
on some finite interval. Typical applications will arise in the next .chapter
(Secs. 11.3 and 11.5) in connection with partial differential equations. Then I(
f(x) will be defined on some interval 0 ;:;: x ~ L, and on .this interval we
want to represent f(x) by a Fourier series. We could chooseL as the interval
of periodicity and go ahead. We can do better by choosing the.period 2L
because then, since f(x) is given only on half of this rjInge (this'. interval), ;
we can get for f(x) a Fourier cosine series, representing '~he even extension
of f(x) to the full range -'L '" x "'. L (Fig. 247b), or we can get for f(x) a
Fourier sine series, representing the odd extension of f(x) to the full range
(Fig. 247c). These two series are called the two half-range expansions of the
function f(x), which is given o'nly on "half the range" (half the periodicity
interval of these series). The form of these series is given in Sec. lOA. The

{(x)~

L x
(a) The g.iven function {(x)

-L L x
(b) {(x) extended as an even periodic
function of period 2J,~

---1 d:::H r--1~

r -v-i P (c) ((x) extended as an odd per"looic
r x

function of period 2L.

FIg. 247.-(a) Function f(x) given on an inteNal 0 ~ x ~-L.

(b) its even extension to the full "range" (intefVal) - L ~ x ~ ((heavy cUfVe)
and t/:1e periodic extension of period 2L to the x-axis,
(e) its odd extension to -L ~ x ~ L (heavy curve) and the periodiC extension
of period 2L to the x-axis
 602 F.O~RIER .~E.RIES; INT~RALS, .TR;!li'lSFORMS CHAP.. IO

cosine hal.f-range expansion is [cf..-"(3), (4), -Sec.. 10.4]

(1)

where
-. .' _ L

(2)
2
an ~ L.o n7rX
f(x)cosT dx ,
f' n ~ 1,2, ....

The sine half-range expansion is [cf. (5), (6), Sec. IO.4J

(3) f(x) ~'i bn sinn,'x

=1 .-L
where
ri;'
(4) L fL
2
f(xl SIn
.T n1rX
dx, n t, 2, ....
o ,

k~-
o LI2 -L )x

Fig. 248. The given funetlon in Example 1

EXAMPLE 1. "Triangle" and Its half-range expansions

Find the two half-range expansions of the [unction (Fig. 248)
2k L
-x if O<X<"2
L
fIx) ~

{ 2k
L:(.. ~ x) L <x< L
if 1.

. Solution. (a) Even periodic extcn~ion. From (4), Sec. lOA, we obtain

ao "" ..!. -["'- fU2 x Jx + ~ fL (L - x) dX] = ~ ,

LLo LU2

on = L2 [2kiLn
T. n~ 2kfL.
x cos LX dx + L (L - x) cos
n~]
LX dx .
o Ln .

Now by integration by parts,

fo
U2 1111'
x cos -x dx "" -
I.x .
Sin
tiff
-x
ILn
- -
L J1 .J2 n1r
sin -x dx
L nu L b 1111 0 L

L2 . Inr' L2 ( _ n1T'- )
~ -2- sm -2 + 22 C05- - 1-;.
II1/" 11 1r '2
SEC.1D.5 HALFRANGEExPANSIONS 6D3
\
Similarly,
L
n~ L2 nnr- L2 ( n~)
f
U2
(L - x) cos-x dx = - 2un sin
L
"2 n2n2 cos n1T - cos 1: '.

By inserting these two results we obtain

4k ( n~ \.
an = n 2 u2 2 cos 2"" - cos nT' -

Thus,

and an = 0 if n # 2. 6, 10. 14, .... Hence the first half-range expansion of f(x) is

k 16k ( - I cos -
f(x) = - - -..
211', I 61T )
x + - 2 cos - x + ....
2 11'2 22 L 6 L

This series, represents the even periodic extension of tbe given function f(x) . of period 2L,
shown in Fig. 249a.
(b) Odd periodic eitensio~: Similarly. from (6), Sec-. lOA. we obtain

8k . mr '\,
(5) bn = n2w2 Sin "2 .
"'-
Hence the other half-range expansion offC;)i;--- ,.
8k (I
i(x) = -
.
-SIn-X -
11"2]2
1T
L 3
I,. 311"
--sm-x
2
+ -IS.t n511
L
-x-+
52 L
.. J
This series represents the odd periodic_extension of
. ';.~. ~
jfA of'''<'0:
period 2L, shown in Fig. 249b. I
'""'-.

(a) .Even extension

Fig. 249. PeriOdiC exten~6ns of i(x) in Example 1

Our main work so far in Sees. l0.2-10.5-has been concerned with finding
Fourier coefficients -by integratiQn. In,the:he_xt section we show that in most
practical cases, integration can be a-voided by appiyio.g what we call the
method of jumps, which is very powerful and of great practical interest.
 FC>URIE~,SERIES. INTEGRALS. TRANSFORMS CHAP. 10

Problems for Sec. 10.5

Represent the following ,functions j(x} -by '3, Fq~ri~; co~i~e series and graph the
corresponding periodi~,:ex~ension, Qf f(x).
1. f(x) ~ k (0 < x <L) 2. !(x) ~ x (0 < x < L)

3. f(x) ~
r I
if
if li2<x<:"!--
'Q.<x<-U2

ili
4. f(x)

6. f(x)
~
~
G
I -':!.
if
if JJ2<x<L

('O<x<L)
'O<x<IJ2

5. f(x) ~ x' (0 < x < L)

L
7. f(x) ~ x3 (0 < X < L) '8. f(x) ~ e" (O<x<L)
'C
, "X '. 'lTX
9. f(x) ~ Stn'L (O<x<L) 10. f(x) ~ sin 2L (0 < X < L)

Represent the following functions Ax) by a FOUl,-ier sine series and graph the cor-
responding periodic extension of f(x).
11. f(x) ~ k (0 < x < L) 12. f(x) ~ Ia,. (0 < x < L),
13. f(x) ~ L - x (0 < x < L) 14. f(x) ~ I -c (21Llx (0< x < L)

"2 if O<x<LI2 _{'I if 0< x< LI2

15. f(x) ~ { 312 if li2 <.x < L
.f/' 16. f(x) -
'0 if [J2<x<L
,

O<x~ , if O<x<L12
~
if
17. f(x) ~ { x 18; fix) { x
L-x if li2 < x <L U2 if U2<x<-L
19. f(x) ~ x2 (0 < x <L) 20. f(x) ~ x 3 (0 <x < L)

21. (Complex, form of the Fourier series, comple?, Fourier coefficients) Using the for-
mula e i6 = cos (I -+ i sin 8 (d. Sec. 2.4), show that
I, .j , I ,inX '
cos nx = 2" (eltU" + e- tRX ), sm nx = 2i (e - e- l 1lX)

and the Fourier series . i

f(x) = lao + n~l (an cos nx + &11 sin nx)
may be written in the for~

(6)

where Co = a o' c n =- (an - ib n)l2. kn = (an + ibn )/2, n = 1,2, .... Using (6),
Sec. 10.2, show that

c = -
n
I
21T_1f
I <
f(x)e-inx dx, k
n
~ -
I
21T
I <

-'Ir
!(x)e'= dx
_ '
n = 1,2,.

Introducing the notation kn ~ c_ n ; show that"(6) may be written

(7)
n=-'"
C
n
I,
= -211"- I-<
'If_~
J(i)e-: inx dx. n=O.I,2,.
 SEC. 10.6 fOURIER COEFFICIENTS WITHOUT INTEGRATION 605

This is the so-caned comple?, form of the FO,urier series, an~ the en are called the
complex Fourier coefficients of f(x).
22. Show'that the complex Fourier coefficients of an odd function are pure imaginary
. and those of an even function are real.
Find the complex form of the Fourier series of the following periodic" functions of
period ~1T. Obtain from it the corresponding real Fourier series and compare it with
previous results.
23. f(x) = x (-" < x < IT) (cf. Prob. 12).

" 24. f(x) = - J if - 17" < X < 0,

Sec. 10.2.)
f(x) = 1 if 0 < x <
.
1T.
(ef. Example I,

0 25. ind the complex fonn of the Fourier series of f(x) = eX

f(x + 2,,) = f(x). .
( - 7T

26. Obtain from the answer to Prob. 25 the corresponding real Fourier series .
< X < 1'1'),

Calculating Fourier CoeffiCients

Without Integration
(Method of Jumps)
In several of our previous examples, relatively complicated and lengthy
integrations-led to relatively simple expressions for the Fourier coefficients
an and bn - This raises the question of whether there might be a simpler way
of obtaining Fourier coefficients. There is, and we want to show that the
Fourier coefficients of a periodic function that is represented by polynomials
can be obtained in tenns of the jumps of the function and its derivatives.
Of course, this is a- big advantage. and the corresponding formulas are of
great practical importance. because by applying them we can avoid integra-
tions (~xcept for ao which must be determined as before).
By ajumpj ofa function g(x) at a point Xo we mean the difference between
the right-hand and left-hand limits (Sec. 10.2) of g(x) at xo; that is,

(I) .f ~ g(xo + 0) - g(xo - 0).

It follows that an upwaidjump is positive and a downward jump is negative;

see Fig. 250. .

k!~ (aJ
x
(hJ

Rg. 250. Jump of a function.

(a) Positive jump. (b) Negcrlive jump

7
606 FOUR.IER ~ERIE~. INT.EGRALS. TRANSFORMS CHAP. 10

CiJ;6
.:ro
(=-1r)
Xlr2V %4~
(=11") /'

Ag. 251. Example of a representati9n,

of the fonn (2) (with m ~ 4)

Let f(x) be a function that has period 2'lT and is represented by polynomials
PI' ... Pm in the interval -'IT' < X < 7f, say (Fig. 251),

.!Pl(X) if Xo < x <. xl" (xo = -1T)

(2)
f(x) ~ P2(x) if xl < X < x 2'

Pm(x) if x m _ 1 < X < xm (= 7T).

Then f may have jump's at x O' xl' .. ,' . x m." and the same is true for the
derivatives It, j", .... We choose the following notation. '

js = jump of [ at Xs'
(3) js' = jump of f' at is' (s 1,2, ... , m)

js" = jump of f" at xS ' etc.

Of course, if f is continuous at xs' then js = 0, and for the derivatives the

situation is similar, so that some of the numbers js' js', ... in (3) may be _
zero.
EXAMPLE 1. Jumps of a function and Its derivatives
Lei
o when --1T<X<O.
f(x) ~
{x2. when 0<x < 1t.

We plot graphs or f arid its derivatives (Fig. 252),'

f'" ,:" O.

We see that the jumps are

Jump at Jump at
Xl"" 0 x 2 == ~

~ 0 _~2
f 11 j2 ==
f' h' ~ 0 j2' ~ -h
I
f" i1" ~ 2 j2" ~ -2
~
Note that the jumps at,X = - 1t are,' ~Ol listed. because they are taken into account at x
the other end of the interval of periOdicity.
= 1t,
I I
~
,1

J
 )
, SEC.1O:6' FOURIER COEFFICIENTS WITHOUT INTEGRATION 607

ii
-1r 0 1f X

1
'J o x

Fig. 252. f(x) and derivatives in Example 1

To get the d~sired formula for the Fourier coefficients G 1 , G z' ... of the
function j. giv~n by (2), we start from the Euler formula (6b) in Sec. HJ.2:

(4) 1Tan =, f-x

x
f cos nx dx.

Since f is represented by (2), we'write the integral as the sum of m integrals:

where Xo = -1(' and xm = 7T. Integration by parts yields

(6)
x,
I f cos nx ,dx
f
= ';;. sin nx
IX, I IX~
- ;; f
,.
SID nx dx.
Xs-l Xs-l Xs-l

Now comes an important point: the, evaluation of the first expression on the
right. f(x) may be -discontinuous at Xs (Fig. 253), and we have to take the
left-hand limit I(xs - 0) of fat x" Similarly, at Xs_ 1 we have to take the
right-hand limit f(xs _ 1 + 0). Hence the first. expression ,on- the right-hand
side of (6) equals
. I
- [f(xs - 0) sin nxs - f(x S _ 1 + 0) sin nxs _ 1]'
n

((Xs_1+O)

XS - I x.<

Fig. 253. The formula (6)

608 FOURIER SERIES,INTEGRALS, TRANSFORMS CHAP. 10

Consequently,- by' inserting (6) into (5) and using the, short notations
So =. sin nxo' SI = sin nx1 , etc . we obtain

(7)

If we collect terms with the same r',' the expression in the-brackets becomes
.,~
- f(x o + O)So + [f(x1 - 0),.;- f(x 1 + O)JSl
(8):;'
+ [f(x2 - OJ'~ f(x 2 + 0)]S2 + ... + f(xm - O)Sm'

The expressions in the brackets. in (8) are the jumps of f, multiplied by - I.

Furthermore,_ because of periodicity, So = Sm and f(Xo) = f(xm ), so that
we may combine t~,e first and the last term in (8). Thus (8) is equal to

and from (7) we therefore havethe ,intermediate result

Im'f x
L ,
m
1
(9) 1Ta n = -;; _2: j~, sin nxs - ;; f' sin nx dx.
s=l. _, s=l Xs-t .

Applying the same procedure to,,"ihe 'Integrals on the right-hand side we find

(10)
mX.
2:
$",1
I'X . _I
1m'
f' sinnxdx =;; 2U/tosnxs +;; 2:
$=1
Imxs

$",1
JXs-i
~'~cosnxdx,

Continuing in this fashion, we obtain integrals involving higher and higher

derivatives of f. Since f is represenl~d by polynomials and the (r + l)th
derivative of a polynomial of degree r is identically zero, we shall reach the
point where no integrals are left, and this will happen after finitely many
steps. By inserting (10) and the analogous formulas obtained in the further
steps into (9) we obtain the desired formula

1 [ m 1 m
a n = -111T -:L j
$
sin nx$ - -Il 2: j
$
, cos IU
$
$=1 $=1

(118)

+2
1 ~.f!'
11
L.J J$ SIn
$= 1
I1X$ + 3n
.
1 ~.",
L.J
$"" 1
J, cos nx --++ .. J
 SEC. 10.6 FOURIER COEFFICIENTS WITHOUT INTEGRATION 609.

. where it ,~ 1,2, ....(anda.onfust be obtained by integration as before). In

precisely, the same fashion we obtain from (6c) in Sec. 10.2 the formula

(Ilb)
1m
+ -n3 L,J.s'" SIOm:
. ]
+--++ ....
. s
s",,1

To avoid errors, it is practical to graph f(x) and its derivatives and list
the jumps in a table as shown in Example 1.
EXAMPLE 2. PeriodIc square wave
Find the Fourier coefficients of the function (Fig. 254)

-k if -7T<X<O,
I(x) = { k
if 0< x < 7T.

Solution. We see that f' ... 0 and the jumps of fare

Jump at Jump at
xl = 0 X2 = ff

f h = 2k i2 = -2k

I is odd. Hence an = 0, and from (JIb),

I . I
b n = ;;; U1 cos nXl + i2 cos nx 2] = ;;; [2k cos 0 - 2k cos nul

2k {4klnU for odd n

=-(i-cosnu)= (cf. Example I. Sec. 10.2).
nu 0 for even n

EXAMPLE 3
Find the Fourier series of the function in Example 1 of the current section.
Solution. By integration,

a
o
=-1 x2dx=~
l'
2u 0
2
6
From (lla),

1 [
un ;;;;; u 2.
sm nu 2ff cos n7r + ;;Z
+ -; I(2.
Sin
.] = ~
2 cos n7r.
0 - 2 sm nu)

..... :~

o
-k

Fig. 254. Example 2

 610 FOURIER. SERIES, INTEGRALS, TRANSFORMS CijAP.10

hn = -
nv
1[, cos
~1T n7T +- h.
n
Sin
L
n7r - "2 (2 cos 0 - 2 cos mr)
n
. 1
n 2
- ;;,cos mT + n 31T (cos mr - 1).
Hence.

2'

and the Fourier series is

f(x) =' ~ - 2cosx\+ (11" -;) sin-x + ~cos ll-- si p 2x + .... I

Fourier series are very powerful in handling and solving models involving
differential equations. In the, next section we give a first illustration .of this
fact in terms .of a typical, vibrational problem. Chapter J 1 .on partial differ-.
ential equations will include numerous further such applications .of Fourier
series.

Problems for Sec. 10.6

1. Obtain the Fourier series ,in Exampl,el by the usual method and compare the
amount -of work with that in Example 3.
2. Find the Fourier series of f(x) = X4 ( - 7r < x- <'11), f(.i- + 211") = I(x) by the
use .of (6), Sec. 10.2, and (1), this section, and compare the amount of work.
3. Derive (lIb) from (6c), Sec. 10.2.
4. Show that in the case of a function lex) having period p 2L the formulas
corresponding to (11) are

(12a)
I fA I >II -

+ K 2 L js" sin Knxs + ](3 L js'" cos Knxs - - + +

ns~1 ns"'l

I [~.
bn = -n1T .J). cos
K x - -I ~ ., . K
.J J, Sin n~s
~ nSK
s~ 1 " s .. 1
(12b)

1
Using (11) or (12), find the -Fourier series of the functions! in:
5, Probs. 1-4, Sec. 10.2 6. Probs. 11-14, Sec. 10.2
7. Probs. 4-6, Sec. 10.3 8. Probs. 10-12, Sec. 10.3
Using (II) or (12), find the_Fourier sine series of
9. ! in Probs. 12 and 17-in Sec. 10.5 10. f in Probs. 14 and 18 in Sec. 10.5
11. f(x)~ x(,r - x') (0< x < ,,) 12. f(x)6 x' - x (0 < x < ,,)
 SEC. 10.7 FORCED OSCILLATIONS

Using' ('12), find the Fourier cosine series ~f

13. f in Prob~. 2 and 5)n Sec, 10.5 14. I(x) =::: x 3 (0 < x < L)
15. Can (II) be applied to find the Fourier coefficients of the function I(x) = eX
(0 < x < 2n), f(~ + 2'1T) = f(x)?

III Forced Oscillations

Fourier series have iJ!lportant applications in connection with differential
equations. Let us consider an important practiCal problem involving an or-
dinary differential equation. (Applications to partial differential equations
will be: c6nsid~red in the next chapter.) .
Froin Sec. 2.'13 we know that forced oscillations ofa body of mass,m on
a spri~g (cf. Fig. 255) are governed by the equalion

(1) my" + cy' + ky ~ r(t)

where k is the spring modulus, c-is the damping constarit;and t is time.

If the external force ret) is a sine or cosine function and the damping
constant c is not zero', the steady-state solution 'represents a harmonic os-
cillation having the frequency of the external for,:::e.
We shall now see that if r(t) is not a pure sine or cosine,function but any
other periodic function, then the steady-~tate solution will represent a su-
perposition of harmonic oscillations having the frequency of r(t) and mul-
tiples of this -frequency. If the frequency of one of these oscillations is close
to the re~onant frequency of the vibrating system (cf. Sec. 2.13), then that
oscillation may be the dominant part of the response of the system to the
external force. Of cpurse, "this is quite surprising to an observer not familiar
with the corresponding mathematical theory, which is highly important in
the study of vibrating systems and resonance. Let us illustrate the situation
by a typical example.

EXAMPLE 1. Forced oscillations under a nonslnusoldal periodic driving force

In (I), let m = I (gm), c = 0.02 (gmlsec), and k = 25 (gmlsec 2); so that q) becomes

Fig. 255. VibratJng system under consideration

612 FOURIER'SERIES; INTEGRALS. TRANSFORMS CHAP. 10

(2) y" + 0.02/ + ?5y = r(l)

wher~ ret) is measured in gm . em/sec 2. ~t (Fig. 256) rit)

~/2
~
t +- when < I <. 0, ,t,!
{
-1r

ret) = ~ ',(t + 21T) = ret).

-I +"2 when 0< 1-< 11', .~'"
~,

Fig. 256. Force in Example 1

Find the steady-state solution Y(I).

Solution. We represent ret) by a Fourier series; finding

(3) ret) =;.4( cos t :- ~I$ 'I'i)

~ cos 31 + 52 cos 51 + ......

Then we consider the differential equati~n

(4) y" + 0.02y' + 25y = n:1I" cos nt . (n "" 1,3" . -)

whose righthand side is a single term orthe series (3). From Sec. 2.13 we know that the steady-
state solution yn(O of (4) is of the form

(5)

'By ~~ubstituting this into (4) we find that

(6) B = 0.08 where

n - n'frD'

Since the differential equation (2) is linear. we may expect the steady-state solution to be

(7)

where Yn is given by (5) and (6). In fa~t, this follows readily by sUbstituting (7) into (2) -and
using the Fourier series of ret), provided that lcrmwise, differentiation o((7),},sJlermissible.
(Readers already familiar y.-ith th~ notion of uniform convergence (Sec. 14~8],ma:y' prove that
(7) may be differentiated tenn by term.) -
From (6) we find that the amplitude' of (5) is

Numerical values are

C 1 "" 0.0530

C3 ."" 0.0088
Cs "" 0.5100
C 7 "" 0 .. 0011

Cg "" 0.0003

For n "" 5 the quantity D is very small, the denominator of Cs is small, and Cs is so large that
Ys is the dominating term in (7). This implies that the steady-state motion is almost a harmonic
oscillation whose frequency equals five times that of the exciting force (Fig., 257). I

The apptication"of Fourier series to more general viprating systems, heat

conduction and other problems follows in Chap. 11. Another are~fbfp,factical .
"
 SEC. 10.7 FORCED OSCILLATIONS 613

Fig. 257. Input and steady-state output

In Example 1

interest in which Fourier series playa role is the approximation of functions

by simpler functions, known- as approximation theory. This will be explained
in the next section.

Problems for Sec. 10.7

1. What would happen to the amplitudes en in Example 1 (and, accordingly. to the
form of the vibrations) if we changed the spring constant to the value 91 If we
took a stiffer spring with k = 491 If we increased the damping?
Find a general solution of the differential equation y" + w2y = ret), where
2. r(t) = cos at + cos J3t (co 2 7'= (1'2, f32)
3, ret) = sin tt W = 0.5, 0.7,0.9, 1.1,1.5,2.0,10.0
4. ret) = sin t + i sin 31 + is
sin 5t, w = 0.5.0.9,1.1,2.2.9,3.1,4,4.9,5.1,
6,8
N
5. r(l) ~ Lb. sin nl, Iwl" I, 2, ... , N
,-1
N
6, r(t) = 2;' an cos nt, Iwl# 1,2, ... , N
11",1

7. r(/) ::: ~ Isin tl when -7T' < t < 7T' and ret + 27T') = r(t), Iwl ~. 0,2,4, ...
. {/+
8. ,(I) ~
17 if -7T'<t<O
and (,(I + 2".) ~ r(I), Iwl " 0, 1,3, ...
, -/+7T' if 0</<17
t if -IT/2-<t<7T'12
9. r(l) ~. { and ,(I + 2".) ~ r(I), Iwl " 1,3,5, ...
'1r-tif'1rl2<t<3'1r12
10. Some of the A(w) in Frob. 3 ate.positive and some negative. Is this phy;ically
understandable? Some are large in absolute value and some small. Is the reason
the same as in Example I of the text?
 'i
614 FOURIER'SERIES. INTEGRALS. TRANSFORMS CHAP. 10

Find the steady-state oscillation corresponding to y" + cy' ~ Y "'" r(I), where
c> 0 and
11. ret) = K sin t 12.- r(-I) = sin 31.
N
13. ret) = an COs nt 14. r(t) = L _bn sin._nl
n=l
",/4 if -1r12 <- t < w/2
IS. ret) = { - aqd ret + -21T) = ret)
n{ 7T - 1)14 if 7T/2 < t < 3'1(/2

16. r(l) = 121 (1T 2 - 12) if - 7T <" t < 1T and r(t + 21T) = r(/)

(RLC-circuit) Find the steady-state ~_urrent I(t) ,in the RLC-circuit in Fig. 258,
where R = 100 ohms, L = lO_henrys, C = 10- 2 farad,

IOO(1Tt (2) if + '_-'1['.-< -( < 0._ _

E(I) ~,' and E(I + 2,,) ~ E(I), '
{
IOO(1Tt - (2) if . 0 < t < 1f __

Proceed as follows. Develop E(I) in a Fourier series. I(t} will appear in the form
of a trigonometric series. Find the general formulas for the coefficients of this
series. Compute numerical values of the first few coefficients. Graph the sum'
of the first fewterms of that" series.
@same task as 'in Prob. 17 with R. ~L~' C as"before;and E(/) = 200/(11"2 - 12) volts
if -11" < I < 11" and E(I + 211") ='~(t). .."

.[], . E(t)

Fig. 258.RLCcircuit in Problems 17 and 18

Approximation
by Trigonometric Polynomials.
Square Error
Let f(x) be a given fu'nction of period 211" that can be represented by a Fourier
series. Then the Nth partial sum of this series is an approximation to f(x);
N
(I) f(x) = a o + L (an cos aX + bn sin ax),
n""l

It is natural to ask whether (I) is the "best" approximation to f by a trig-

onometric polynomial, that is, a function of the form
N
(2) F(x) = a o + 2.:. (an cos nx + f3 n sin nx).