Fourier Integrals and Fourier Transforms

2
Fourier Integrals and

Fourier Transforms
2.1 Introduction
The concept of an infinite series dates back as far as the ancient Greeks
such as Archimedes (287-212 ø.c.), who summed a geometric series in
order to compute the area under a parabolic arc. In the eighteenth century,
power series expansions for functions like ex, sin x, and arctan x were
first published by the Scottish mathematician C. Maclaurin (1698-1746),
and British mathematician B. Taylor (1685-1731) generalized this work
by providing power series expansions about some point other than
x = 0.
By the middle of the eighteenth century it became important to study
the possibility of representing a given function by infinite series other
than power series. D. Bernoulli (1700-1783) showed that the mathematical
conditions imposed by physical considerations in solving the vibrating-
string problem were formally satisfied by functions represented as infinite
series involving sinusoidal functions. In the early 1800s, the French
physicist J. Fourier* came across similar representations and announced
* Jean Baptiste Joseph Fourier (1768-1830) is known mainly for his work on the
representation of functions by trigonometric series in his studies on the theory of heat
conduction. His basic papers, presented to the Academy of Sciences in Paris in 1807 and
1811, were criticized by the referees for a lack of rigor and consequently were not published
then. However, when publishing the classic Théorie analytique de la Chaleur in 1822, he
also incorporated his earlier work almost without change.
37
Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

38 • Chap. 2/Fourier Integrals and Fourier Transforms
in his work on heat conduction that an "arbitrary function" could be

expanded in a series of sinusoidal functions. Some of Fourier's work
lacked rigor, but nevertheless he provided the first real impetus to the
subject now bearing his name. The Fourier integral was also first introduced
by Fourier as an attempt to generalize his results from finite intervals
to infinite intervals. The Fourier transform, while appearing in some
early writings of A. L. Cauchy (1789-1857) and P. S. de Laplace (1749-
1827), also appears in the work of Fourier.
In this chapter we will discuss Fourier integral representations and
Fourier transforms, followed by a chapter on applications involving the
Fourier transform.
2.2 Fourier Integral Representations
An important problem in mathematical analysis is the determination of

various representations of a given function f. For example, a particular
representation may reveal information about the function that is not as
obvious by another representation. In the calculus we are taught that
certain functions have power series representations of the form
f(x) _ C n xn (2.1)
n=0
where
C„ = f (»(0)/n!, n = 0,1,2,...
Power series such as this are useful for numerical calculations in addition
to various other uses. If the function f is periodic with period 2p, it may
have a Fourier series representation*
f(x) = 1 a + n=i P
j (anCOS n + basin
P
—j
Pl
(2.2)
where
I IP nart
a n = p J f(t)cos P dt, n = 0,1,2,... (2.3)
P
and
1 f'p nart
b„p f(t)sin p dt, n = 1,2,3,... (2.4)
* For a general discussion of Fourier series, see L. C. Andrews, Elementary Partial

Differential Equations with Boundary Value Problems, Orlando: Academic Press, 1986.

2.2/Fourier Integral Representations • 39
The theory of Fourier series shows that a periodic function satisfying

certain minimal requirements can be represented by the infinite sum of
sinusoidal functions given in (2.2). The formal* limit of this representation
as the period tends to infinity can be used to introduce the notion of a
Fourier integral representation. In other words, while periodic functions
defined on the entire real axis have Fourier series representations, aperiodic
functions similarly defined have Fourier integral representations.
If f and f' are piecewise continuous functions on some interval [ – p, p],
we say that f is piecewise smooth. If f has this property and is periodic
with period 2p, it has the Fourier series representation (2.2)–(2.4). To
formally obtain the Fourier integral representation off from this series
as p –* oo, we begin by substituting the integral formulas for a o , a,,,
and b„ given by (2.3) and (2.4) into the Fourier series (2.2). This action
leads to
---
f(x) = 2
P P
j
p f(t)dt + 1 fP f(t)cos nart cos n-rx dt
n=t P P p p
P
+ p f f(t)sin —^
np t sin n px dt
- P
or
P
nlr(t – x)
f(x) = 2
p
fp f()dt + 1 J f(t) ^ cos dt (2.5)
p' n=1 P
where we have interchanged the order of summation and integration and

used the trigonometric identity
cos A cos B + sin A sin B = cos(A – B)
We now wish to examine what happens as we let p tend to infinity.
First, we must make the additional requirement that f is absolutely in-
tegrable, i.e.,
f If(t)dt <00 (2.6)
so that
lim 2 v
p P f(t)dt = 0
For the remaining infinite sum in (2.5), it is convenient to let As = Ir /p

(2.7)
and then consider the equivalent limit
* By "formal," we mean a procedure that is not mathematically rigorous.

f(x) = lim i
f v/As
f(t) cos[nis(t - x)]Os dt (2.8)

L s—oo 1r 1T /Os n1
(Observe that Os -> 0 as p -+ When As is a small positive number,

the points nLs are equally spaced along the s axis. In such a case we
may expect the series in (2.8) to approximate the integral
fo-
cos[s(t - x)]ds
in the limit as is - 0. While this does not mean that the limit of the
series in (2.8) is defined to be the above, we may take, under appropriate
conditions on f, that (2.8) tends to the integral form
f(x) =1 Jf(t)
10 cos[s(t - x)]ds dt
-
(2.9)
Upon switching the order of integration, we get the equivalent form
f(x) o J f(t)cos[s(t - x)]dt ds (2.10)

=
The purely formal procedure we just went through (since the passage
to the limit cannot be rigorously justified) has led us to an important
result known as Fourier's integral theorem.* We will state the theorem
here but not present its rather lengthy proof until Sec. 2.3.
Theorem 2.1 (Fourier Integral Theorem). If f and f' are piecewise con-
tinuous functions on every finite interval, and if
J If(x)Idx <00
then
f(x) _ -- f J_ f(t)cos[s(t - x)]dt ds
at points x where f is continuous. If x is a point of discontinuity of f,

the above integral converges to the average value 2[f(x) + f(x )] of -
the right-hand and left-hand limits,t
* For a rigorous discussion and a precise statement of the conditions under which
(2.10) holds, see E. C. Titchmarsh, Theory of Fourier Integrals, Oxford: Clarendon Press,
1937.
t Right-hand and left-hand limits are defined, respectively, by f(x+) = lim
E o+
f(x + e)
and f(x ) = lim f(x — s). At points of continuity it follows that f(x ) = f(x+) = f(x).
- E -

2.2 /Fourier Integral Representations • 41
The conditions listed in Theor. 2.1 are only sufficient conditions, not
necessary conditions. That is, there exist functions f that have valid
integral representations but which do not satisfy the conditions of this
theorem. Moreover, the conditions stated in Theor. 2.1 are not the most
general set of sufficient conditions that have been established over the
years. Nonetheless, these conditions are broad enough to cover most of
the functions commonly occurring in practice.
To emphasize the analogy between Fourier series and the Fourier
integral theorem, we rewrite (2.10) in the form
f(x) =ff o — f(t)(cos stcos sx + sin stsin sx)dt ds
or equivalently,
f(x) = f [A(s)cos sx + B(s)sin

o sx) ds (2.11)
where
A(s) = - f f(t)cos st dt (2.12)
and
B(s)
= -'-
I W f(t)sin st dt (2.13)
In this setting we refer to (2.11) as the Fourier integral representation

of the function f with coefficients defined by (2.12) and (2.13). The general
theory concerning such representations closely parallels that of Fourier
series.
Example 2.1: Find an integral representation of the form (2.11) for the
rectangle function f(x) = h(1 — lxi), where h is the Heaviside unit
function (see Fig. 2.1).
—1 1 x
Figure 2.1 Graph of f(x) = h(1 — jxl)

Solution: The coefficients A(s) and B(s) are given by
T
A(s) =—
1
I h(1 — IxD)cos sx dx = — 1 J ' 2cos sxsin

dx =
IT 1 ITS
s
and
B(s) = sin sx dx = 0
1
Thus the Fourier integral representation becomes

2 ° ( sins l
f(x) = JCossxds
J
?r s
Since x = 0 is a point of continuity of fin Exam. 2.1, we can use
the Fourier integral theorem to deduce that
2 (`° sin s
f (0) = 1 = — J ds
7r 0 s
which leads to the interesting result*
f o s's s ds = 2 (2.14)
Observe that at x = ± 1 there is a jump discontinuity in the function f

given above. At these points the Fourier integral converges to the average
value of the left-hand and right-hand limits. Hence, it follows that
1/2, x= —1
2 f fsins \1, —1<x<1
cos sx ds = ( 2.15)
ir o s 1/2, x=1
0, otherwise
Finally, it may be of interest to plot the "partial integral" of the
function f, defined by
2 / sins 1
Sµ(x) = - 1 s I cos sx ds
fo"
(2.16)
R
to see how it tends to f(x) as —p o. ecalling the identity
2 sin A cos B = sin(A + B) + sin(A — B)
we have
* This, of course, is a standard integral result that can also be derived by the use of
complex variable theory. It is an important result that we will refer to on several occasions.

2.2 /Fourier Integral Representations • 43
= 1 ( sin[s(1 + x)] ds + 1 f" sin[s(1 - x)] ds

S"(x) S
1J-Jo ITJO S
1 " ' + x) sint

(
1 "('-x) sin t
= -
7r o
—dt+-
t 7r o
—dt
t f
= 1 {Si[µ(1 + x)] + Si[µ(1 - x)]}
IT
(2.17)
where Si(z) is the sine integral defined by
Si(z) = o
z
r
sin t
dt (2.18)
Equation (2.17) is plotted in Fig. 2.2 for values µ = 4,16,128.
2.2.1 Cosine and Sine Integral Representations

If the function f is an even function, i.e., if f(-x) = f(x), it follows from
properties of integrals that
A(s) _f m
f(x)cos sx dx =
fO f(x)cos sx dx (2.19)
and
B(s) = --f(x)sin sx dx = 0 (2.20)

7r --
from which we deduce
A(s)cos f(x)
sx ds (2.21)
= fo-
x
-I
Figure 2.2 The partial integral of f(x) = h(1 - jxj)

We refer to (2.21) as a Fourier cosine integral representation. In a similar

manner, if f is an odd function, i.e., f(—x) = —f(x), we obtain the
Fourier sine integral representation
f(x) = fo B(s)sin sx ds (2.22)

—
where A(s) = 0 and
B(s) = 27r fo^f(x)sin sx dx (2.23)
Finally, if f should be a function defined only on the interval 0 <

x < —, we can represent it over this interval by either a Fourier cosine
integral or a Fourier sine integral, analogous to the half-range expansions
of Fourier series. Consider the following example.
Example 2.2: Find a Fourier cosine and Fourier sine integral repre-
sentation of the function (see Fig. 2.3)
.f(x)_ cos0, x, 0 <x < ir/2

x>Tr/2
Solution: For a cosine integral representation, we compute

7r/2
2 cos ns 2
A(s) = 2 f cos x cos sx dx =
Tr ir(1 —s)
and, therefore,
_ 2 (°° cos irs/2 1
sx ds
f (x) zr fo 1 —s2 ) cos
In the same manner

fw/2
2 2 l' s — sin irs/2 1
B(s) _ - cos x sin sx dx = - s2 — 1 /I
1 2 X
Figure 2.3 Graph of f(x) = h(or/2 — x) cos x, x> 0

2.2/Fourier Integral Representations • 45
from which we deduce the sine integral representation

f(x) = 2 f ( s — sin lrs/21 sin sx ds
^rJo s—1J
EXERCISES 2.2
1. By using the result of Eq. (2.14), show directly that
(a) f
°° sins cos s Ir
o
s ds = 4
(b) fo,sin axdx= ^,

a>0
2. If
f(x) = j0, x<0
l e x,
- x>0
(a) show that
1 cos sx + s sin sx ds
f (x)= —I
7r o s2+1
(b) Verify directly that the above integral representation converges
to the value 1/2 at x = 0.
3. Find an integral representation for
f(x) = p — x2 ' ^x^ > 1

and deduce the value of the integral `
I — (ó sin x z x cos x^
J 3 cos 2x dx
4.Use the result of Prob. 3 to deduce that

(°° ^ 1 — cos x 1 2 a
x J dx=2
(a))o
c sinzx dx =
oo
(b)
x 4
In Probs. 5-10, obtain the Fourier integral representation of the given
function.
—1, —1 <x<0
5.f(x)=e lxl, —oo<x<oo 6.f(x)= 1, - 0<x<1
0, otherwise

sin x, IxI < 7r sin x, 0 < x < ir

7. f(x) = 8. f(x) _
0, 1XI > -7r 0, otherwise
Cosx, 0<X<?r
9. f(x) = 10. f(x) = e 2 , — oc < x < oo
0, otherwise
11. Show that e -A, k> 0, has the half-range representations
2k cos sx
x>0
(a) e- _ zr Jo S z + kz ds,
2 s sin sx
(b) e = o s z + kz ds, x>0
12. Show that e - Xcos x has the half-range representations

S2 (
(a) e - xcos x= 2 f o s' + 4 cos sx ds, x> 0
2 s 3 sin sx
(b) e - Xcos x = jo s4 + 4 ds, x> 0
13. Using the results of Prob. 11, show that

_x _ 2x 6
= s sin sx
e — e = ir o (s 2 + 1)(s 2 + 4) ds,
x> 0
In Probs. 14-17, obtain the Fourier cosine and Fourier sine integral
representations of the given function.
0<x<p,K>0
14. f(x) = {K,
0, x > p
1 —x, 0<X<1
15. f(x) =
0, x>1
16. f(x)= 1 0sin x>Tr

x, 0 < x < Ir
0<x<1
17 f(x)—{' x>1
18. Determine whether or not the following functions are absolutely
integrable on the entire real axis.
lXi < 1 (b) f(x) = I 1, lxI < 1
(a) f(x) = f I1 + X^,
lxi > 1 1/xz, lxi > 1
s in
(c) f(x) = e lxl - (d) f(x) =
x
sin x l z
(e) f(x) = (
x J

2.3/Proof of the Fourier Integral Theorem • 47
2.3 Proof of the Fourier Integral Theorem

In order to provide a more rigorous justification of the Fourier integral
theorem (Theor. 2.1) we start with the following central result.
Lemma 2.1 (Riemann—Lebesgue). If f is piecewise continuous and ab-
solutely integrable on the entire real axis, then
lim f f(t)cos xt dt = lim J f(t)sin Xt dt = 0
or equivalently,
lim J f(t)e °" dt = 0
Proof• We will present the proof only for the case when f is continuous
and has a bounded derivative f' on the real axis. A slight modification
of the proof is required for the case whenf has some finite discontinuities.
Using integration by parts over the finite interval — p <_ t s p, we
get
f PP f(t)e ar dt _ f(l)^;ar ^P — I Pf,(t)e'u dt

- P
Clearly, f(t)e`Xt is bounded on all finite intervals for any X. Thus, the
first term on the right-hand side vanishes in the limit as A tends to infinity.
Also, because we assume that f' is bounded, it follows that
P f , (t)e ;^,t dt l < fP f'(t)Idt <
f P JJ P
for all A. Hence, we conclude that
P
lim f(t)e` ` dt = 0
)
A- - - P
and if we formally allow the limits of integration to become infinite in

extent, i.e., p -- 0, we get our intended result. n
Remark: The assumption that f' is bounded in the proof of Lemma
2.1 is not required for the validity of the theorem. However, by adding
this assumption, our proof is much simpler than would otherwise be
required.
Lemma 2.2. If f is piecewise smooth and absolutely integrable on the

entire real axis, and if x is a point of continuity off, then

1 sin ^t
lim - f(x + t) dt = f(x)
a — 00 ir -- t
Proof: We first note that

1 (P sin At1 fxp sin t
lim - dt = lim - — dt
A—000 IT P t X--- IT XP t
1 (- sit t d
t
=1
this last result following from (2.14). Hence, to prove the lemma we wish
to show that
P rf(x + t) - f(x)] sin At dt = 0
um-1-
;—,00 7r J P IL t
The function [f(x + t) - f(x)]/t is piecewise continuous for all
t * 0, and at t = 0, we find that
limf(x + t) - f(x) = f , (x)
t ->o t
which exists sincef is piecewise smooth. Thus, the conditions of Lemma

2.1 are satisfied by this function, and the above integral necessarily
vanishes, even in the limit as we formally allow p to become infinite.
n
2.3.1 Convergence at a Point of Continuity

We are now prepared to prove our main result, which is Theor. 2.1. We
will present the proof only for points of continuity of the function f,
leaving the proof for points of discontinuity to the exercises.
By changing the order of integration in the following iterated integral,
we obtain
1 f A J f(t)cos[s(t - x)]dt ds =
— o
1 f f(t) j cos[s(t - x)]ds dt
o
sin A(t - x)
=-1 J 00 ƒ(t) dt
t-x
sint At
_- J^f(x+t) dt
If we now allow A -* oo and invoke Lemma 2.2, we obtain our desired

result

2.4/Fourier Transform Pairs • 49
1 °°
1 o f(t)cos[s(t — x)] dt ds = f(x) (2.24)
where x is a point of continuity of the function f.
EXERCISES 2.3
1. Prove that
sin At
(a) lim P
Ä -- o J dt = 7 r
t 2
in At
(b) lim I 0 dt = 7 r
A-- f P t 2
2. Based on Lemma 2.1, show that if f and f' are piecewise continuous
on (0, p) and (— p,0), then
P
sin Xt
(a) lim— f(x + t) dt = f(x)
x .ô Tr o
- t
(b) lim ? I f(x + t)

^,-- —
o P
sin At
t
dt = f(x - )
3. Prove Theor. 2.1 for points of finite discontinuity of f, i.e., prove that
o J j ( f(t)cos[s(t — x)]dt ds = [f(x + ) + f(x)]

2
2.4 Fourier Transform Pairs

Fourier's integral theorem (Theor. 2.1) states that
f(x) = 1 JJ
— f(t)cos[s(t — x)] dt ds. (2.25)
Through the use of Euler's formula, cos x = 2(e" + e - 'x), we can express
(2.25) in terms of complex exponential functions. That is,
f(x) =
-7 JF -
f(t)cos[s(t — x)]dt ds
e-;str
f(t) [ e is(t_1) + x>]dt ds
2Tr Jo J
27r 1T 00 f -
f(t)eL«
dt ds
or

e-'Sx f - e ist f(t) dt ds (2.26)

.f (x) = - J
which is the exponential form of Fourier's integral theorem.
What we have established by the integral formula (2.26) is the pair
of transform formulas*
F(s) = 2^ J - e`s l f(t) dt (2.27)
and
f(t) _ J e - 'S` F(s) ds (2.28)

ç=
We define F(s) as the Fourier transform of f(t), also written as
F(s) = 3F{ f(t);s} (2.29)
and f(t) as the inverse Fourier transform of F(s), which may be written
as
f(t) = & - '{F(s);t} (2.30)
The location of the constant 1/21r in the definition of the transform pairs
is arbitrarily selected as long as (2.26) is satisfied. For reasons of symmetry
we have split the constant between the transform pairs, but in the literature
no universal agreement exists on the location of these constants. In some
texts, the constant 1/21r is positioned in front of one of the transform
pairs with no constant in front of the other. There is also some variation
as to which integral represents the transform and which one represents
the inverse transform. In practice, of course, these differences are of
little consequence but the user should be aware of them when consulting
different reference sources.
As an immediate consequence of (2.27), we observe that
IF(s)I f — If(t)I dt (2.31)
Hence, if f is absolutely integrable it follows that its transform function

F(s) is bounded. A similar argument applied to (2.28) shows that f(t)
is also bounded when F(s) is absolutely integrable. Furthermore, the
Riemann-Lebesgue Lemma (Lemma 2.1) shows that
lim F(s) = 0 (2.32)
^Sj--
* Unless stated otherwise, we will generally assume that both t and s are real variables.

Since the transform function F(s) associated with absolutely integrable

functions that are also piecewise smooth must satisfy this last relation,
it immediately rules out certain functions as possible transform functions.
For example, sines, cosines, and polynomials do not satisfy this relation.
Finally, it is a curious property that although the function f(t) may have
certain finite discontinuities, its transform F(s) can be shown to be a
continuous function. Because of this, the Fourier transform is sometimes
called a "smoothing process."
2.4.1 Fourier Cosine and Sine Transforms

In Sec. 2.2.1 we found that when the function f is even, the Fourier
integral representation of f(x) reduces to
f(x) = fo A(s)cos sx ds
cos sx fo f(t)cos st dt ds (2.33)

_ fO
Based on this relation we introduce the Fourier cosine transform
.c{f(t);s} =
J fa f(t)cos st dt = Fc(s), s > 0 (2.34)
and inverse cosine transform
.c'{Fc(s);t} = ^ F(s)cos st ds = f(t), t> 0 (2.35)

2 10
These results are interesting in that they imply the equivalence of the
operators Jw c and . C'. In other words, the cosine transform and its
inverse are exactly the same in functional form.
Similarly, whenf is an odd function its Fourier integral representation
becomes
f(x) = fo sin sx fo f(t)sin st dt ds (2.36)
which leads to the Fourier sine transform
Wis{f(t);s} _ f(t)sin st dt = Fa(s), s > 0 (2.37)

f0
and inverse sine transform
Wis' {Fs(s);t} = fo Fs (s)sin st ds = f(t), t>0

(2.38)
^— lr

Hence, we see that the Fourier sine transform and its inverse are also
exactly the same in functional form.
If the function f is neither even nor odd, but defined only for t> 0,
then it may have both a cosine transform and a sine transform. Moreover,
the even and odd extensions of f will then have exponential Fourier
transforms. To see the relations between these various transforms, let
us construct the even extension off by setting
fe(t) = .f(ItI), —00 < t < 00 (2.39)
The Fourier transform of fe (t) leads to
Lf,2(0;S} = _ I J_ oofe(t)e,st dt
= 1 cos st dt + i 1 I fe (t) st dt
a I fe(t) V21r
_ Jo f(t)cos st dt
{ f e (t);s} = O' c{ f(t);s}, — cc < s < 00 (2.40)
Based on (2.40), it is clear that the Fourier transform and cosine transform
of an even function give identical results. In particular, their transforms
are even functions of s (see Prob. 20 in Exer. 2.4). The odd extension
off is constructed by setting
f0(t) = f(t)sgn(t), —cc < t < cc (2.41)
where the signum function is defined by
sgn(t) = j 1, t> 0 (2.42)

In this case, we find
S{fo(t);s} = 1 J— f0 (t)e' dt
2ir °° 0
_ I Jf 0 % ss t dt + i 1 fo(t)sin st dt
2^r V2n
= i Jo f(t)sin st dt
Because the Fourier transform of an odd function is also an odd function

(see Prob. 20 in Exer. 2.4), we make the conclusion that the Fourier

transform and sine transform are related by

J1 {fo (t);s} = i.w s {f(t);jsI}sgn(s), —cc < s < cc (2.43)
The practical use of (2.40) is that if we want to evaluate the Fourier
transform of an even function, we can do so by simply calculating its
cosine transform. If the function we wish to transform is odd, we first
can find its sine transform and then use (2.43). These observations are
very useful when using tables to find transforms, since most of the known
transforms are either cosine or sine transforms. A short table of transforms
is presented in Appendix B.
2.4.2 Evaluating Transforms of Elementary Functions

As already pointed out, many elementary functions like sines, cosines,
polynomials, and in general any periodic function, do not have Fourier
transforms (at least in the usual sense) because they are not absolutely
integrable. A special class of elementary functions that do have Fourier
transforms and can be calculated by basic methods are those involving
exponential functions. Several such transforms are related to the integrals
I J
= o e - °`cos st dt = s2
a
az , a > 0 (2.44)
and
s
J = o e °` sin st dt =
- sZ + a2 9 a > 0 (2.45)
One way to verify these integral formulas is to use integration by parts

to obtain the relations
a a o
f
I = — s °° e °` sin st dt = 1 SJ
a a
- —
and
J = (s /a)I
Solving these last two equations simultaneously for I and J yields the
results given by (2.44) and (2.45).
Example 2.3: Find the Fourier transform of e - ° 1 ", a > 0.
Solution: Because the function is even, we can use (2.40) to write

{e 21 `1 ;s} = c{e - °`;s}
_ f
-
e - °`cos st dt

or, using (2.44), we see that
g{e - aJ tJ;s} =
J 2Tr
s +a
2
,
a 2 a>0
Because transform relations occur in pairs as given by Eqs. (2.27)

and (2.28), it follows that once we have established one transform or
integral relation, the other one is automatically known. For instance,
based on the result of Exam. 2.3, we have
9W{e-a^t1;s} = -L f e tste
J
- °lÌ dt =J 2 a z ,
s +a '
a >0 (2.46)
2
and thus it immediately follows that
+ a2 ;t)) } = j s 2e +a 2 dS = 1
2ay2
e % aIi a>
(2.47)
Moreover, by interchanging the roles of t and s in (2.47), and taking the
complex conjugate of the resulting expression (which is real in this example
since we are dealing with even functions), we now deduce the additional
Fourier transform
11'r
{ t2 + a2 ,s J = ^ _ / 2 e - °^ 5 ^, a > 0 (2.48)
In this fashion we see that the evaluation of a single Fourier transform
has the effect of giving us two transform relations from each integral.
It can be shown that both (2.44) and (2.45) are uniformly converging
integrals on any closed intervals for which a is positive and all closed
intervals involving s. Related integrals which can be formally derived by
differentiating or integrating (2.44) and (2.45) with respect to either pa-
rameter, a or s, can also be shown to converge uniformly. This means
that we can formally differentiate or integrate both sides of (2.44) and
(2.45) to produce new integral relations which can then be related to
other integral transforms. Consider the following examples.
Example 2.4: Find the Fourier sine and cosine transforms of te - a t ,
a > 0.
Solution: Formal differentiation of both sides of (2.45), first with

respect to a and then with respect to s, gives us, respectively,
—
J 0
te
- at
sin st dt = — tas /(s 2 + a2)2

and
fo te - °`cos st dt = (a' — s 2 )/(s 2 + a2)2
Thus, we deduce that
s{te
-°t
;s} = ^ + a ,
(s2 2 2) a>0
and
_ 2 a — s2
gC{te -at;s} 2 Z 2, a>0
7r(s +a)
Example 2.5: Find the Fourier sine transform of (1/t)e °`, a > 0. -
Solution: We begin by integrating both sides of (2.45) with respect

to the parameter a from a to o, which leads to
°`sin st dt = I° sZ +
fo t e
-
- az da
= - -tan 'a -
2 s
^s
=tan —
a
Thus, it follows that
jl 2 s
e - °`;s tan-` —, a>0
t zr a
If we allow a -p 0+ in the result of Exam. 2.5, we find
J's{1/t;s} = _ / 2 (2.49)
This result is only a formal result since neither 1/t nor ß/7r/2 satisfy the
conditions of the Fourier integral theorem. Nonetheless, it can be useful
to treat (2.49) as a limiting case of the transform relation given in Exam.
2.5. Using (2.43), we obtain the similar relation*
Jv{l/t;s} = i\/or/2 sgn(s) (2.50)
* Formal results like (2.49) and (2.50) are discussed in more detail in Sec. 2.8.

Finally, as a bonus we see that (2.50) provides a generalization of (2.14),

which is
(ó sin st
dt = 2 sgn(s) (2.51)
Example 2.6: Find the Fourier transform of e - °Z`Z , a > 0.
Solution: From definition, we have

eist - a212 dt
P{e _ Q212 ;s} = 1^^ f-
By writing
a 2 í — ist = (at — is/2a) 2 + s 2 /4a 2
we find
{e -azt . s} _
2 1 e -s2/aa2 e-cat-;s/za)2 dt
_m
= 1 e
a\/21T
s/ a
- 2 4 2
J —
e - '2 dx
—_ 1 e -S 2 "4" 2 r(1/2)
a 27r
where we have made the change of variable x = at — is/a and used

properties of the gamma function. Simplifying this last result leads
to
1 e_2/4a2
f{e-Q2'Z-s} = a>0
By setting a = 1/\h in the result of Exam. 2.6, we obtain the

interesting relation
f{e-`Z/Z,s} = e-x212 (2.52)
which says that the function e - `2/2 is self-reciprocal, i.e., it is its own
transform.
EXERCISES 2.4
1. Given the following functions, develop the even and odd extensions,
fe (t) and fo(t), respectively:
(a) f(t) = e a' - (b) f(t) = e ` sin t - 2
(c) f(t) = (1 + t)e °` - (d) f(t) = e ` + sinh t -

2. Determine the Fourier transform of the function f(t), given that
JIs{.f(t);s} = -ns — 1), b>0

V 2 (2e
In Probs. 3-10, determine the Fourier transform of each function.
t<0 0<t<1
3. f(t) _ e alp
t > 0, a > 0 4 ' f(t) 0 ,
= 1, otherwise
5. 1(t) =

1,
0,
1
Iti<b
otherwise
6.
1
f( t) =
sint,
0,
t
0<t<Tr
otherwise
7. 8. f(t) =
f(t) 5t 2 + 1 t2 + 7
t 1—t2
9. f(t) 10. f(t) = (t2 + 4) 2
= (? + 4)Z
11. Show that

sin mt 1zr
t ;s = J2 h(m — (si), m>0
12. Use the results of Exam. 2.4 to determine

(a) . {te - akl ;s}, a>0
(b) 9W{ItIe - ° 1 `ß;s}, a>0
13. Given the triangle- function f(t) = (1 — iti)h(1 — Its), show that
7 sin2(s/2)
(a) 3W{f(t);s} = V2 1T s2
(b) From (a), deduce that

sinx 2
f dx=7r
X
14. Letting f(t) = 1/Vt in the sine and cosine forms of Fourier's integral
theorem given by (2.33) and (2.36), respectively, show that
(a) i
cos t dt = f °°sin t dt = I-r
o \/t 112
S^Vt;sI - Î /t _ Vs
15. Use the result of Exam. 2.6 with a = (1 — i)/2 to derive
15.
(a) Jw{cos(t2/2);s} = 1= [cos(s2 /2) + sin(s2/2)]

(b) .{sin(tZ /2);s} = 1 [cos(z s 2) — sin(S 2 /2))
(c) sws { r cost/2);s} _ \/ [C(s/\/^) + S(s/\)l *
16. Show that

_ 2 aS Z + 2a 3
(a) JvC{e -a`COS at;s} 54 4 , a>0
a + 4a
2 2a 3 — as 2
(b) .we{e- at sin at;s} = 4 4 , a> 0
Tr s + 4a
17. From the results of Prob. 16, deduce that
' ± k4 ;t } = 2i e -kfrI/'f [coskt/V + sin(ktd/V2)], k> 0

18. Evaluate the sine transforms

(a) 9s{e - ° t ;s}, a>0
(b) .Fs{e °`cos at;s},
- a>0
(c) s{e - °`sin at;s}, a>0
19. From the results of Prob. 18, evaluate

t3 t
(a) s { 4 + k 4;S } (b) 9S {t4 + k 4;S }
(c) J'S { (t4 +k4)2 ;S } (d) 9S l (t4 + k4)2;S }
20. Prove the following properties of the Fourier transform for real func-
tions f(t):
(a) If f(t) is even, then F(s) is real and even.
(b) If f(t) is odd, then F(s) is imaginary and odd.
(c) If f(t) is neither even nor odd, then F(s) has an even real part
and an odd imaginary part.
2.5 Properties of the Fourier Transform

The calculation of integral transforms is often tedious and quite complex
in some instances. However, once we have derived the transforms of
some standard functions, we can deduce the transforms of many other
* C(x) and S(x) are the Fresnel integrals (see Sec. 1.3.2).

2.5/Properties of the Fourier Transform • 59
functions in a simple way through the use of certain operational properties

associated with the transform. These operational properties are basically
consequences of the properties of integrals.
If f(t) satisfies the conditions of the Fourier integral theorem, its
Fourier transform F(s) is uniquely determined by the integral
F(s) = T
1 00 e"'ƒ(t) dt (2.53)
Thus, there is only one transform function F(s) associated with each
function f(t). However, if f(t) and g(t) are two functions that are identical
everywhere except at certain isolated points, then both f(t) and g(t) will
have the same transform, say F(s). This means that the inverse transform
of F(s) can be either f(t) or g(t). Of course, the distinction between
functions that differ only at isolated points is mostly of academic interest
and has little effect in practical applications. If we agree to define a
function at a point of finite discontinuity as the average of its left-hand
and right-hand limits, then f(t) is uniquely related to F(s) by the inverse
transform relation
2 [f(t) + f(t)} = 2zr f e

-,s:
F(s) ds (2.54)
Another important property of the Fourier transform and inverse

Fourier transform is the linearity property.
Theorem 2.2 (Linearity property). If F(s) and G(s) are the Fourier trans-
forms, respectively, of f(t) and g(t), then for any constants C, and C 2 ,
it follows that
. {C,f(t) + C2g(t);s} = C,F(s) + C 2 G(s)
- '{C,F(s) + C2 G(s);t} = C,f(t) + C 2g(t)
Directly from the defining integral, we have

=f
Proof:
JW{C,f(t) + C2g(t);s} = e'S` [C,f(t) + C 2g(t)] dt
= C l 1= f(t) dt +
+
V2 f e`S` g(t) dt
= C,F(s) + C 2 G(s)
A similar argument proves the inverse transform linearity property. n
If a > 0, then

{f(at);s} _ J e"'f(at) dt
_ 1 eiucs/o) f(u) du
-- — aV2 ir
where we have set u = at. Thus, if F(s) is the Fourier transform of f(t),
we have just shown that
{f(at);s} = (1/a)F(s/a), a > 0 (2.55a)
Similarly, if a < 0 it follows that

.V{f(at);s} = —(1/a)F(s/a), a <0 (2.55b)
so that in general we have the scaling property
3w {f(at);s} = (1/lal)F(s/a), a * 0 (2.56)
Example 2.7: Give n th at the Fou rier transform of f(t) =

(sin b 1 + tZ )/\/ 1 + t Z is F(s) = \/Ir/2J0( b2 — s 2 ), I st < b and
F(s) = 0, Ist > b, determine the Fourier transform of
(sin b\/a 2 + t 2 )/ß/a 2 + ? Z b > 0. ,
Solution: We first observe that

sin b\/a 2 + t2 _ sin blaI\/1 + (t/a) Z
a 2 + tz JaIV1 + (t/a) Z
and then using (2.56), we see.that
{ a 2 + t 2 ;s / 1—1
sin b a 2 + t2 — ' / 2 Jo (\/b2a2 — a 2s 2), las < blat
0, last > blat
Ist < b
{ 0, Isl > b
2.5.1 Shift Properties
Multiplication of either f(t) or F(s) by a complex exponential causes a
shift in the transform variable upon completing the integration of the
transform or inverse transform. More precisely, we have the following
theorem.
Theorem 2.3 (Shifting property). If f(t) and F(s) are Fourier transform
pairs, then
(a) 9{e'° f(t);s} = F(s + a)
(b) {f(t — a);s} = eias F(s)

Proof.• From definition,
^2^ e'°` f(t) dt

,?{e'°f(t)•s} =
J e ist
_ l - f — e' ' 'f(t) dt°

= F(s +.l a)
In the same fashion,
3W - {eF(s);t} = r e -its e"-' F(s) ds
= I f e - j(t - a's F(s) ds
= fit — la)
JW{f(t — a);s} = e r°s F(s) M
Example 2.8: Find the Fourier inverse transform of 11(5 2 + ias + b),
b>0.
Solution: By completing the square, we have

1 _ 1
s 2 + ias + b [s + (ia /2)] 2 + [(a 2 /4) + b]
Then, using Theor. 2.3,*
9. t 1 t= e-ariz_-t
- 1 t
ts' + ias + b' sZ + [(a 2 /4) + b]'
exp[-2(at +ß/a 2 +4bftI)]
=V a2 + 4b
the last step of which follows from Eq. (2.47).
2.5.2 Transforms of Derivatives and Derivatives of Transforms

In applications involving differential equations it is important to know
how the Fourier transform behaves on derivatives of a function. If f is
continuous everywhere and f' is piecewise smooth, and both f and f'
are absolutely integrable, then
* Note that Theor. 2.3 implies that 9 '{F(s + a);t} = e'°`.# '{F(s);t}.
-

S7 {f'(t);s} = :;
J eis f'(t) dt
= 1 f(t)e,st
- f e;s' f(t) dt
2Sr _- ^/2^r J —
where we have employed an integration by parts. Now if f also satisfies
lim f(t) = 0
III-- o
we then obtain
f{ f'(t);s} = -isF(s) (2.57)
where F(s) is the Fourier transform of f(t). By repeated application of
(2.57), we can prove the following more general result.
Theorem 2.4 (Differentiation property). If f, f', ..., f ') are continuous
everywhere and absolutely integrable, f" is piecewise smooth and ab-
)
solutely integrable, and

lim f(t) = lim f'(t) = ... = lim f " " (t) = 0 ( -
ItI—=o frj— —
then
JW{f(")(t);s} = (-is)' F(s), n = 1,2,3,...
where F(s) is the Fourier transform of f(t).
Remark: If f(t) has a finite jump discontinuity at t = à, then f'(t)

contains an impulse at t = a (see Sec. 1.5.2). In this case the Fourier
transform of f'(t) must also contain the Fourier transform of the impulse
function. Such concepts, which involve the notion of generalized functions,
will be discussed in Sec. 2.8.
In the case of the cosine and sine transforms, the above results are
somewhat different. For example, in the case of the cosine transform
we use integration by parts to obtain
c{ f'(t);s} = 2/7r
10 f'(t)cos st dt
,
_ -\/2/^r f(0) + S\/7 Jo f(t)sin st dt

3W c{f'(t);s} = sFs(s) - \//f(0) (2.58)

Similarly, it can be shown that

3W s{f'(t);s} = —sFc(s) (2.59)
(see Prob. 9 in Exer. 2.5). For second derivatives, we are led to the
relations
Vic{ fri(t);s} _ — s 2 Fc(s) (2.60)
— \/2 /^ f '(0)
and
s{f"(t);s} = —s 2Fs (s) + ß/2/7r sf(0) (2.61)
the verification of which is left to the exercises (see Prob. 10 in Exer.
2.5). These last two formulas give us some indication of which transform
— cosine or sine — to use in a particular application. That is, in any
problem in which f(0) is known but f'(0) is not known, we should use
the Fourier sine transform of f"(t). In the same way, if f'(0) is known
rather than f(0), the Fourier cosine transform should be used.
If the transform of f(t) is F(s), then the transform of tmf(t), m =
1,2,3,..., can be found by repeated differentiation of F(s). To see this,
let us start with the Fourier integral
F(s) = 1 ~ f - e t f(t) dt (2.62)
and formally differentiate both sides with respect to s. This action yields
F'(s) = 1T00 e is t [itf(t)] dt
and thus we conclude that

5W{tf(t);s)} = —iF'(s) (2.63)
Of course, the validity of (2.63) requires that the transform of tf(t) exist.
Continued differentiation of (2.62) with respect to s leads to
F (m) () _ 1 r — eist
[(it) mf(t)] dt, m = 1,2,3,... (2.64)
which we now formulate as a theorem.
Theorem 2.5. If f is absolutely integrable and piecewise smooth, and

if t"f(t) has a Fourier transform, then
S'{tmf(t);s} = (—i)m F (m ) (s), m = 1,2,3,...
where F(s) is the Fourier transform of f(t).

By combining Theors. 2.4 and 2.5, we arrive at the result

fc^>(t);s} = (_ i)m+n dSm
{tm . [s"F(s)], m,n = 1,2,3,... (2.65)
Example 2.9: Find the Fourier transform of te` - `Z/Z
Solution: Recalling Eq. (2.52), we have

JW{e_ 12 ^ 2 ;s} = e -S2/2
Then, using (2.63), it immediately follows that

2{te - `z/2 ;s} = — i( — s)e -SZ /Z = is e -SZ/2
Finally, recalling Theor. 2.3, we deduce that
Ow{te t-t2/2 ;s} = i(s — i)e -(s-;2/2
= (1 + is)e" e-(s2-1WW2
EXERCISES 2.5
1. If f(t) is a real function with transform F(s), show that the complex
conjugate of F(s) satisfies
F(s) = JW{f(t);—s}
2. If f(t) is a complex function with transform F(s), show that the
complex conjugate of F(s) satisfies
F(s) = 9{ f(— t);s}
3. If F(s) is the Fourier transform of f(t), show that

9{e' b`/'°f(t/a);s} = aF(as + b), a>0
4. Show that
(a) 5 c{f(at);s} = (1/a)Fc(s/a), a>0
(b) .s{.f(at);s} = ( 1 /a)Fs(s/a), a > 0
5. Show that
(a) 3Wc{f(t)cos at;s} = 2[Fc (s + a) + F(s — a)]
(b) s{f(t)cos at;s} = 2[Fs (s + a) + Fs (s — a)]
6. Show that
(a) Fc{f(t)sin at;s} = fFS(s + a) — Fs (s — a)]
(b) 3rs{f(t)sin at;s} = 2[Fe(s — a) — F(s + a)]
7. Use the results of Probs. 5 and 6 to show that

2.6/Transforms of More Complicated Functions . 65
2 as z + 2a 3
(a) . F c{e - °`cos at;s}
=
, a>0
^ 7r S 4 + 4a 4
2 2a 3 - as 2
(b) c{e - a l sin at;s} = /; a>0
5 4 + 4a 4 '
8. Use the results of Probs. 5 and 6 to evaluate
(a) Jw s{e - °`cos at;s}, a>0
(b) 9s{e - a , sin at;s}, a>0
9. Verify that the Fourier sine transform satisfies the relation

.fi s{f'(t);s} = - sFc(s)
10. Derive the transform relations _

(a) Vic{ f"(t);s} = — s
2 Fc(s) — V2 /irf'( 0 )
(b) s{ f "(t);s} = — s 2 Fs (s) + \/2/Trsf(0)
In Probs. 11-15, evaluate the Fourier transform of the given function

using known transforms and appropriate properties of the transform.
11. f(t) = (1 - t)e - HÌ 12. f(t) = e b: - `2
13. f(t) = t2 e_ 12/2 14. f(t) = e - ' 2/2 cos 2t
15. f(t) = t cos t 2
In Probs. 16-18, evaluate the Fourier inverse transform of the given

function using known transform relations and appropriate properties of
the transform.
1 e us
16. F(s)= 17. F(s)=
sz +4s+7 s z +4s+7
18. F(s) = tan -' (s/a) sgn(s)
Hint: Examine F'(s).
19. Show that, under appropriate assumptions on f and its derivatives,
f'(0) — f(3)(0)]
3 W C{f (4)(t);s} = S 4Fc(s) + \/7 [SZ
20. Show that, under appropriate assumptions on f and its derivatives,

{f 4 (t) ;s} = s4Fs(s) — \/2/r (s 3f(0) — sf n(0)]
2.6 Transforms of More Complicated Functions

When the functions involved in a Fourier transform or inverse transform
are of a more complicated nature, we usually must resort to techniques

other than the standard integration methods of calculus. Sometimes it is

useful to represent part of the integrand in a power series and perform
termwise integration on the resulting expression.* In certain cases the
resulting integrated series can be summed to yield the transform we are
seeking. Other useful techniques are those involving the powerful methods
of complex variables.
To illustrate the power series method mentioned above, let us consider
the following example.
Example 2.10: Find the Fourier cosine transform of (a2 — t2)p-' n

x h(a — t), p > —1/2, a > 0, where h(t) is the Heaviside unit
function.
Solution: From definition of the cosine transfi >rm, we have
JvC{(a 2 — i2) p-1/2 h(a — t);s} _ t2) p -1/2 cos st dt

VVV Jo (a2 —
2 (— 1)k S2k
Vir k = o (2k)! J
0(a2 — t2)p-1/2 t2k dt
where we have replaced the cosine function with its power series
representation. The substitution t = a sin 0 in the above integral
leads to
l o (a2 _ t2) 1/2 t2k dt = a2p+2k
fo cos 2p 8 sin 2k0 dO
= a 2p+2k r(P + 2)r(k + 2)
2r(p +k+ 1)
by use of Eq. (1.11) (see Chap. 1). Next, employing the duplication
formula of the gamma function [see (G7) in Sec. 1.21
\/Irr(2z) = 22z-1 r(z)r(z + Z)
we find that
=
c{(a2 — t2)p-1/2h(a — t);s} a2pr(P + 2) (— 1 ) k r(k + I (a s)2k
V2r (2k)! r(k + p + 1)
k (as/2)2k+p
= 2 p 1/2 r(P + 4)(a/ s) E ( 1 1) + p + 1)
"
k=0 k.r(k
This last power series is recognized as the Bessel function J(as) (see
Sec. 1.4), and thus we have our result
* The power series in such cases must converge everywhere.

2.6/Transforms of More Complicated Functions • 67
Yc{(a 2 — t)P-112h(a — t);s}
= 2P - ' 12F(p + 2)(a/s)' J(as), p> —1/2, a > 0
Because the Fourier cosine transform and inverse cosine transform

are identical operations, we can use the result of Exam. 2.10 to deduce
the additional cosine transform relation
a2 P 1lpp 1(z
( h(a — s) , p >
JVc{t -" .J (at);s} = —
1/2, a > 0 (2.66)
2 Za I'(p + 2)l
Also, for p = 0 we obtain the special case

i2 1 .
1 9{J0 (at);s} c{Jo(at);s} = v
2.6.1 The Use of Residue Theory

0,
a2 — S2' Is t <a
Is^>a
(2.67)
The calculus of residues from complex variables is a powerful tool in

the calculation of many transform formulas. To deal with some of the
integrals that will arise, we will need to use the theorems from complex
variables provided in Appendix A.
To begin, we wish to derive the pair of transform formulas:
2 F(a)
^ c{t" ';s} =- s" cos(ira/2), s > 0, 0 < a < 1 (2.68a)
ÎT
s {t" - ';s} = _ /? (
F a) sin(ira/2), s >0,0< a < 1 (2.68b)
1T s"
—
Let us define the complex function f(z) = z" - ' e SZ and integrate it around -
the closed contour shown in Fig. 2.4. From Cauchy's integral theorem
(Theor. A.1 in Appendix A), it follows that
f c f (z)dz = 0 (2.69)
or
1CP f(z)dz +R
f f(x)dx + L' f(z)dz + f R .f(iy)d(iy) = 0 (2.70)
Along the imaginary axis we have set z = iy, so that

-^y"- Ie -isv
f(iy) = i"
If we now allow p —* 0 and R — oo, we have from Theors. A.3 and
A.4 in Appendix A that

Figure 2.4 Contour of integration
lim f f (z)dz = 0, lim f f (z)dz = 0 (2.71)

p-o c^- Roo CR
and hence (2.70) reduces to

0
fo— x"-'e-xdx + i" Jy" 'e sydy = 0 - -
Now setting i " = e - 2 , we can rewrite this last expression as

" le "Ydy
fo y
-
= e2 fo
— x"-'e-Sxdx
(2.72)
In order to evaluate the integral on the right, we make the change of

variable u = sx to get
f o -
x" è Sxdx = s-"
- -
f
o
e R u" ' du
- -
= s "I'(a)
-
Jo y
"-1 e - sydy =
, rsa e ir"n
) -;
(2.73)
Finally, multiplying both sides of (2.73) by the constant factor V/2/a

and equating real and imaginary parts, we are led to the desired formulas
given by (2.68a) and (2.68b).
If we let a = 1/2 in (2.68a) and (2.68b), we get the special cases
.We{1/^;s} = 1/\/s (2.74a)
and
.Vs{1/Vt;s} = 1 /V's (2.74b)

which shows that 1/\/t is self-reciprocal under cosine and sine

transformations.
The residue calculus is especially helpful in finding transforms of
rational functions. To illustrate the technique, let f(z) denote a complex
function with the following properties:
1. f(z) has a finite number of poles a,, a 2 , ..., an in the upper half-
plane.
2. f(z) is analytic along the real axis except at the .points b 1 , b 2 , ...,
b m , which are simple poles.
3. zf(z)e'SZ — 0 as z --* co, Im(z) > 0.
Suppose we integrate the complex function f(z)e`sz, s > 0, around the
contour shown in Fig. 2.5. By use of the residue theorem of complex
variables, we find that
m+l
j f(x)e3x
k=1 Lk
+ J f(z)e
k=1 Ck
`SZd
z +
CR
f(z)e'SZdz
= 21ri 1 Res {f(z)e SZ;a k}

1
k=1
where L,, L2 , ..., L,n + , are the straight line segments along the x axis
and C l , C2 , ..., Cm are small semicircles with centers at the simple poles
b 1 , b 2 , ..., bm . In the limit as R — and the radii of the small semicircles
tend to zero, we obtain
m n
f f(x)e "xdx — In 1 Res{f(z)e`sz;b k} = 21ri 1 Res {f(z)e'SZ;a k} ( 2.75)

k=1 k=1
where (see Theor. A.3 in Appendix A)
lim
R- — C1
JI zedz
Z =0
X
—R b, b2 bm R

Thus, we have developed the general transform formula for rational

functions (changing x to t in the integral)
n 1 m l
{ f(t);s} = i\/2^r E Res{ f(z)e";a k} + — E Res{ f(z)e'SZ;b k} (2.76)
k=1 2 k=1 J
for s > 0.
If f is either even or odd we can extend the result (2.76) to include
s < 0 by utilizing the relations (2.40) and (2.43). For more general f we
can set s = — cr < 0 and integrate the function f(z)e_"' around a contour
similar to that in Fig. 2.5, but in the lower half-plane. Equivalently, we
can replace z by — z and integrate the function f(— z)e - isZ around a
contour in the upper half-plane. The result of this latter approach is the
transform relation (see Prob. 9 in Exer. 2.6)
{ f(t);s} = i\ EN i
Res{ sz
f(— z)e ;a k}
k=1
+ Res{f(—z)e-isZ;ßk}
2k =1 J (2.77)
where s < 0. Here a,, a2 , ..., aN are the poles of f( — z) in the upper
half-plane and f3, ß 2 , ..., ß M are simple poles of f(— z) along the real
axis.
For calculating inverse Fourier transforms by this method, we simply
observe that the cases corresponding to positive and negative t are the
reverse of those for positive and negative s as a consequence of the fact
that the kernel of the inverse transform is the complex conjugate of the
kernel of the transform.
Example 2.11: Find the Fourier transform of 1/t(t2 + k 2), k> 0.
Solution: The complex function

f(z) = 1/z(z 2 + k 2)
has simple poles at z = 0 and z = ± ik. Hence, calculating residues
at z = 0 and z = ik, we find
Res{f(z)e 15z;O} = lim zf(z)e isz = 1/k 2
z-0
and
Res{f(z)e` z;ik} = lim (z —ik)f(z)e` z = — e -ks /2k 2
z —ik
From (2.76), we now have

1 1 e-ks
+ k2);s = i^/27r (2k
{
t(t 2 2 — 21cz
(1 — e-ks), s > 0
= ^2 /c 2
Because f(t) = 1/t(t2 + k2 ) is an odd function, we can use (2.43) to

deduce that, for all s,
—e klsl )sgn(s)
j t(t2 + k 2) ;s 1 = y 2 k2(1
Example 2.12: Find the inverse Fourier transform of F(s) =
1/(s 2 + ias + b), a > 0, b > 0.
Solution: By definition,
1 °°
{F(s);t} = e-"S F(s) ds
__ 1 J e"SF(—s) ds
where we have replaced s by —s. The complex function

1
F( — z) = z2 — iaz+b
has simple poles at z = z, and z = z 2 , where
z i =2(a+Va 2 +4b)
z 2 =2(a—\/a 2 +4b)
Clearly, z 1 lies in the upper half-plane while z 2 is in the lower half-
plane. Calculating the residue at z = z 1 leads to*
eitz I e—(a+ a)t/2
Res{F(— z)é `z ;z,} = I =
2z — ia Z=Z , i\/a2 + 4b
and thus we have
* Recall that if z = a is a simple pole of f(z) = P(z)/Q(z), then

(z — a)P(z)
Res{f(z);a} =im l Q(z) = P(a)/Q'(a).
—a

1
- ^{s 2 + ias + b' t
— 1 a2
2ir
For t < 0, we consider the complex function

4b 1
exp -2(a + Va t + 4b)t], t >0
1
F(z) =
z 2 + iaz + b
which has simple poles at z = z 3 and z = z4, where
Z3 = — (a + \/a 2 + 4b)
i
z4 = — (a — Va 2 + 4b)
The pole at z4 lies in the upper half-plane and

e-"n I e- (a - uz +ab)y2
Res{F(z)e - "z;z 4} = _
2z + ia Z _ z4 iVa 2 + 4b
Therefore,
1
s 2 +ias+b' t
exp — 2 (a — ß/a 2 + 4b)t t<0

— Ja 2 +^4b L ] ,
and by combining results, we deduce that for all t,*
2 1T
exp -2(at + Va 2 + 4bItI)]
{s2 + ias + b;t} ^^ 4b [
Example 2.13: Find the Fourier cosine transform of (cosh at/cosh t)

and Fourier sine transform of (sinh at/cosh t), jal < 1.
Solution: To start, let us integrate the complex function

e (n + ia)z
f(z) cosh z
around the rectangular contour shown in Fig. 2.6. The function f(z)
has simple poles at z = (n + 1/2)-rri, n = 0, ± 1, ±2, ..., but the
* Recall that we previously found this inverse transform relation through use of the
shifting property (see Exam. 2.8).

only pole enclosed by the rectangular contour C is ia/2. Hence, from

the residue theorem (Appendix A) it follows that
f c f(z)dz = 21ri Res{f(z);ia/2} = 21re ( 'a - s )n/ z
This can also be written as
f R
f(x)dx +
Ic_ ,
f(z)dz + ff(z)dz + J f(z)dz = 2ae (ia - s )7r/Z
C3 Co
Along the line segment C 2 we have z = R + iy, and thus

e
aR-sy e i(sR+ay)
f f(z)dz = I fo cosh(R + iy)dy
C
"
fir e-sy
^ eaR
dy
Jo Icosh(R + iy)j
It can be shown that the last integral vanishes in the limit as R -*
oc (see Prob. 20 in Exer. 2.6). The same in true of the integral along
the line segment C 4 . We have z = x + iir along the segment C 3 ,
which leads to
R e (a+is)(x+ivr)
ff(z)dz
c3 IR cosh(x + iTr) dx
-
e(a+is)x
= (^(ia-s)a
f R cosh x
dx
Therefore, as R -- we are left with

(a +is)x
(1 + e(ia-s)ar) e dx = 21re
(ia-s)n/2
-^ cosh x
or

e cu + 2lrel,u
- s,,r/2
1 cosh x dx — I + e l '° -' )IT

2ir
ecia—s)o/2 + e— (ja —s)ir/2
IT
(ia — s)Ir ]
coshr
L 2
By splitting this last expression into real and imaginary parts, we get
°° cosh ax sinh ax
sx dx + i — sin sx dx
-^ cosh x cos Cosh x
IT
cos-- cosh 2— i sin 2 sinh 2

air sir
ir cos
zr sin 2 Binh 2
= +i
2a?r 2 SIT 2 air 2 SIT 2air Zs?T lair 2sir
cos 2 cosh 2 + sin 2 sinh 2 cos 2 cosh -j + sin --- sinh 2
and by comparing real and imaginary parts, and simplifying the algebra,
we deduce that
_ cos--cosh--j
cosh
Sc
{t
cosh at
' s } — ^2^ cos a7r + cosh sir'
air sir
â^ < 1
sin j- sinh j
{
-
sinh at
cos air + cosh sir' Iai < 1
^S
cosh t ' s }
EXERCISES 2.6
1. Find the Fourier cosine transform of h(a — t), a > 0 and compare
this with Exam. 2.10 to deduce that
J 112(x) = 2 sin x
^ 7rx
2. Given that f(t) = (t2 — 1)h(1 — lti), n = 0,1,2,..., show that
i{f(t);s} = ( 1)n! (^)'"2 Jn+1/2(S)
3. Use the result of Prob. 2 to deduce that

in
f{Pn (t)h(1 — (tI);s} _ - Jn+112(s), n = 0,1,2,...
s

where P(t) is the nth Legendre polynomial defined by
P(t) n = 0,1,2,...
2 n! dt"[(t2 - 1)"1,
4. Show that
fs{t(a2 - t2 )P -312 h(a - t);s}
= 2P -3 / aPsPr(p - 1/2)JP (as),

2
a > 0, p> 1/2
5. Integrate both sides of Eq. (2.67) with respect to s to deduce the

Fourier sine transform of (1/t)Jo(at).
6. Show that
(a) 3w{Itj - «;s} _ / 2 r(I s j' a) sin 2 , 0<a<1
YY
2 I'(1 - «) ^r«
(b) Jw{frj'sgn(t);s} = i , _^ cos —, 0<a<I
a Isi 2
7. Given that r = ß/a 2 + s 2 and tan 0 = s/a, show that

2 I'(p)
{e "'tP ';s}
= ^-
ir r
P cos p8, a > 0, p > 0
^;2
31s{e "`t 1 ';s} =
- - F(p) sin p8, a > 0, p > 0
Hint: Consider the integral of f(z) = zP 'e - '°'z around the contour
-
in Fig. 2.4.
8. Provide the details leading to Eq. (2.77).
In Probs. 9-14, use residue theory to find the Fourier transform of the
given rational function.
9.f(t)= 10.f(í)=t22+1
t2 +9
1
11.f(t) =ta 12.f(t)=tb+l
+l
(t +
13. f(t) =
4)
2 1 2 14. f(t) = t - 12
t(t + 1)
In Probs. 15-17, use residue theory to find the inverse Fourier transform
of the given rational function.
15. F(s) = s2 + 1 16. F(s) =

2is
2
+l

17. F(s) =( s12 s+Z

4)2—
18. By expanding J0(at) in a power series, use residue theory to deduce

that
S c{(t2 + 1) - 'J0(at);s} = o(a)e '', s> 1

\2I
-
t'<
d2k eist
Hint: Observe that j = ? + 1 e is t dt = (-1)k dsZk tZ + 1 dt
J_=
19. Following the suggestion in Prob. 18, derive the Fourier sine transform
of t(? + 1) 'Jo(at).
-
20. Show that

(a) Icosh(R + iy) ? e R /4, R —t °°.
(b) Use (a) to show that the integral along C 2 in Exam. 2.12 satisfies
the inequality
1 C2
f(z)dz I 4ire'° - "R , R -- o0
and thus deduce that lim

R-.« Cz
J f(z)dz = 0.
sinh(z/2) ;,, 1 cosh(z/2) ; ,.z

21. By integrating the functions sinh z e and e around the
sin(z
contour shown in the accompanying figure, deduce that
sinh(t/2) 1
Binh t
cosh(t/2) 1
;s Jt
;s J
{ = J /;r2 sech irs
= J I7r2 tank its
s sinh t
—R —p p R X

22. Given the transform function F(s) = e'/\/,

(a) show that
Y-'{F(s);t} = 0, t<0
by integrating an appropriate complex function around a closed
contour in the upper half-plane.
(b) By using the contour shown in the accompanying figure, show
that
e- frde,
_ t {F(s);t} —1 t>0
\/IT 0 V e
and thus deduce that
eresla
I
1 h(t)
In Probs. 23-30, establish the given transform relation.
23.Jw c{(1 — t)h(l — t);s} = 1 2 Z(1 — cos s)

v rrs
24. Sc s f
sink at ;sI
tinh Ir
=
1
T
sin a
21T cosh s + cos a' a
cosh at 1 1 sink s
25. = IT
s sink irt' s 1/2ir cosh s + cos a' aC

e -a ^ t1 ^s2 + a 2 + a 1 /2
26. s = ^ z ) , a>0
(s
21r + a2 )
27. {sint 5 } 1 I
1g+
s \
=lo
1-s
'(e -b ` - e -o`) , s} = 1 log ^ssz + a 2

28. c{t - a>b
2 + 62 )
29. ^s{t-'(e-b` - e-°`);s} = - tan 1 1 (a +

L J ab,' a>b
30.
s n b\/a 2 + tZ
a2 + t2 ,s }
l
' / 2 Jo(aVb 2 - s 2)h(1 - Is /bi)
i
2.7 The Convolution Integrals of Fourier

One of the most important operational properties of the Fourier transform
is the convolution theorem. To derive this important property, let us
begin by defining the convolution of two integrable functions f and g by
the expression
(.f° g)(t) = I f(u)g(t - u) du (2.78)

2IT
The Fourier transform applied to this convolution integral leads to

{(f° g)(t);s}= J eist j f(u)g(t - u) du dt
2r
's` f(u)g(t - u)
2r L f- du
e
dt (2.79)
We can interpret (2.79) as an iterated integral for which the order of

integration can be interchanged. Hence, by changing the order of integration
and making the change of variable t = u + x, we find
31{(.f ° g)(t);s} = 2 J " f(u)g(x) dx du

( +x
-
- J e"
_ I r- ^ é suf(u)du , (^ ersxg(x)dx
V2r
and thus conclude that
{(f° g)(t);s} = F(s)G(s) (2.80)
where F(s) and G(s) are the Fourier transforms, respectively, of f(t) and

2.7/The Convolution Integrals of Fourier • 79
g(t). By applying the inverse Fourier transform to both sides of (2.80),

we obtain the alternate form
'{F(s)G(s);t} = (f ° g)(t) (2.81)
This shows that the inverse transform of a product of transform functions
can be found by convolving the inverse transforms of each product term.
Equation (2.81) is the Fourier convolution theorem, which sometimes
is expressed in the form
j
– e' F(s)G(s) ds = f f(u)g(t – u) du (2.82)
An interesting consequence of (2.82) follows by first setting t = 0, which

yields
f F(s)G(s) ds = Jf(u)g( – u) du (2.83)
For the special case where g(– u) = f(u), then*

G(s) _ S'{g(u);s} = 9{f(– u);s} = F(s)
(see Prob. 2 in Exer. 2.5) and we obtain
f F(s)F(s) ds = J f(u)f(u) du
or
f – (F(s) 2 ds = f — If(t)V Z dt (2.84)
Equation (2.84) is called Parseval's relation. In physical applications

the quantity on the right-hand side represents the total energy in a
waveform, such as a sound wave or electrical signal. Thus, Parseval's
relation states that the total energy is given by the area under the IF(s)V Z
curve. For this reason, the quantity JF(s)1 2 is called the energy spectrum
or energy spectral density function of f(t). Engineers are undoubtedly
familiar with the Fourier series counterpart of Parseval's relation, which
has the physical interpretation that the power associated with a periodic
function equals the sum of the powers associated with its harmonic
components.
Returning now to the convolution integral (2.78), we observe that it
satisfies certain formal properties of ordinary products. For example, if
C is a constant, it is immediately clear that
ƒ° (Cg) = (Cƒ) ° g = C(ƒ° g) (2.85)
* In most cases of interest to us the function f(t) is real and thus f(t) = f(►).

Similarly, it follows that the distributive law holds, i.e.,

f-(g+k) =fog +f-k (2.86)
By definition,
1
(g -ƒ)(t) = + f r g(u)f(t — u) du
and by making the change of variable v = t — u, we have
(g of)(t) = —
1 J y g(t — v)f(v) dv
1
= ^/2zr ^f(v)g(t — v) dv
from which we deduce the commutative law

fogg = g °.f (2.87)
Finally, the convolution integral (2.78) also satisfies the associative law
.f°(g°k) = (fog)°k (2.88)
but we leave the proof of this result to the exercises (see Prob. 8 in
Exer. 2.7).
Example 2.14: Use the convolution theorem to evaluate the inverse

Fourier transform
_ sin s
}
s(1 — is)' t
Solution: Let us define

sin s
F(s) = and G(s) = 1
1 is
which have inverse Fourier transforms
= l
.,,I h(l — Itj)
l
and
—is
;t} = V/2ae ` h(t)-
fill
Thus, using (2.81), we have

J - '{F(s)G(s);t} = 2 j e - ^` - "'h(t — u)h(1 — l ul) du

J
J 2 e ` ƒ e" du, ltl < 1
J2 e ` J- u l e"du, 1>1
0, t< — 1
or
^ 2 (1 — e ^r+n) Itl < 1
c_ 1 sin s
s(1 — is) ;t = t>1
0, t< — 1
Example 2.15: Evaluate the integral

dx
I=
-- (x 2 + a 2 )(x 2 + b 2 )
Solution: The given integral has the form

I = I F(x)G(x) dx
where F(x) = 1/(x 2 + a 2 ) and G(x) = 1/(x 2 + b 2 ). Thus, by the use
of Eq. (2.83), we immediately can relate this integral to
I = f f(t)g(— t) dt
where f(t) and g(t) are the inverse Fourier transforms of F(x) and
G(x), respectively. Recalling Exam. 2.3, we see that
1 hr e-"Irl
1(t) = a _/ 2 and g(t) = b J e -blij
and using the fact that these are even functions, we deduce that
I = 2fo f(t)g(t) dt
= ?r f e -'"+b)'
dt
ab o

or
°° dx _ ir
(x Z + a 2)(x 2 + b 2) ab(a + b)
2.7.1 Cosine and Sine Convolution Integrals

Other convolution integrals involving the cosine and sine transforms can
be derived similarly, but the resulting integrals are more complicated
than (2.82). For example, if f and g are functions defined for t> 0 which
have cosine transforms Fc(s) and Gc(s), respectively, then from previous
results we know that
9{f(ItI);s} = Fc(s)
and
.^{g(JtI);s} = Gc(s)
The substitution of these expressions into (2.82) yields
j - e - 'S` Fc(s)G c(s) ds = f — f(I uI)g(It — ui) du (2.89)

00
Since both Fc(s) and Gc(s) are necessarily even functions, we can write
(2.89) as
21 0
cos (st)Fc(s)Gc(s) ds = f °° f(u)g(It — uJ) du + I f(Iu)g(t —uI) du
= Io .f(u)g(I t — uI) du + Jo .f(u)g(t + u) du,

and thus we have derived the convolution integral (t> 0)
f cos(st)Fc(s)Gc(s) ds =
ff f(u) [g(It — uI) + g(t + u)] du (2.90)
Without providing the details, we simply state that the following
convolution integrals can be derived in a likewise manner (see Exer.
2.7):
cos(st)F5(s)G5 (s) ds = 2
fo - J f(u) [g(u + t) + g(u — t)] du (2.91)
j o
sin(st)F(s)G(s) ds 2 fo f(u) [g(lu — tI) — g(u + t)] du (2.92)
f sin(st)Fe(s)G5(s) ds = ff0 f(u) [g(u + t) — g(u — t)] du (2.93)

EXERCISES 2.7
1. Verify the convolution integral (2.82) for
(a) f(t) = g(t) = h(1 — Iti)
(b) f(t) = g(t) = e `2/'2 -
2. Use the convolution integral to find the inverse Fourier transform

of
1 1 1
F(s) =
(1— is) Z I — is I — is
3. Show that the convolution theorem can also be expressed in the
form
J — e'S` f(t)g(t) dt = f — F(u)G(s — u) du
^ t! sin t
and use this result to evaluate the Fourier transform of e
t
4. For a,b > 0, show that
J-= (x 2 +a 2 ( 2 +b 2 ) ß a+b
5. Use the Fourier transform relation
{(a 2 — t 2 ) ' 1Z h(a — l tI );s} = jJ o(as),

a>0
to show that
fo Jo (ax)Jo (bx) dx = bK(a/b), 0<a<b
where K(m) denotes the complete elliptic integral

^^z
K(m) = j (I — m 2 sin 2 6) '" 2 d6-
6. For 0 < a < b, show that

dx ora
o sin axsin bx XZ = 2
7. Verify Parseval's relation (2.84) for

(a) f(t) = e -°h `l , a>0
(b) f(t) = e_a 2t2 , a>0

8. Show that the Fourier convolution (2.78) satisfies the associative

property
.f°(g°k) = (fog)°k
9. For real g(t) = f(t), show that (2.90) leads to the Parseval relation
ds =f If(u)I 2 du
Jo IFc(s)1 2
10. Derive Eq. (2.91) by replacing Gs(s) with its defining transform integral
and interchanging the order of integration.
11. Derive Eq. (2.92).
Hint: See Prob. 10.
12. Derive Eq. (2.93).
Hint: See Prob. 10.
13. For real g(t) = f(t), show that (2.91) leads to the Parseval relation
du
10 IFs(s)1 2 ds = o If(u)1 2
14. By substituting the functions

f(t) = t -", 0<a<1
g(t) = (I – t 2)" - ' '2h(l – t), v > –1/2
into Eq. (2.90).

(a) show that
-^
sin 2 Jr(1 — a)r(v + 4)j o s" -°- 'J„(s) ds = J t - "(1 — t 2) 2 dt
V^
(b) Using properties of the beta function (see Prob. 21 in Exer. 1.2),
evaluate the integral on the right in (a) and deduce that
J o x
a--
vl
J(x) dx = r(v +
1 – a/2)'

0< <a 1,v > – 1/2
15. Using Eq. (2.90),

(a) show that
Jo(at) s 2_ e-S f ° cosh u
du, a>0
I t2 + I V r o ^%a z — u2

2.8/Transforms Involving Generalized Functions • 85
j ,o(at) 's t J = J2 e Slo(a), a > 0

c l t2 + l
,
where I0(a) is the modified Bessel function of order zero.

Hint: Express cosh u in a power series and use properties of the
beta function (see Prob. 21 in Exer. 1.2).
2.8 Transforms Involving Generalized Functions
In Sec. 2.4 we stated that the Fourier transform F(s) is a bounded function
provided the inverse transform f(t) is absolutely integrable, i.e., provided
J If(t)I dt < ( 2.94)
This is actually a sufficient condition of the Fourier integral theorem for

the existence of the transform of a given function f(t). Sinusoidal functions,
the Heaviside unit function, polynomials, and so forth, do not satisfy
the condition (2.94) and, therefore, do not have a Fourier transform
in the usual sense. However, it is possible to extend our definition of
Fourier transform to include such functions if we are willing to consider
the notion of generalized functions such as the impulse function 8(t) and
its derivatives S "'(t), n = 1,2,3,... .
(
The impulse function was first introduced in Sec. 1.5.2. It is a useful

concept in a wide variety of physical problems involving the ideas of
line spectrum, impulsive forces, or point sources. Because the rules of
manipulation of impulse functions do not follow naturally from the methods
of classical analysis, such functions are often referred to as generalized
functions. The general theory of such functions has been put on a solid
mathematical basis under the title of the theory of distributions.* None-
theless, our treatment of impulse functions and their derivatives will
continue to be based upon formal manipulations as developed in Sec.
1.5.2.
2.8.1 Impulse Function

By using the "sifting property" of the impulse function
* The most notable pioneering work on generalized functions is contained in L. Schwartz,

Théorie des Distributions, Tomes 1 and 2, Paris: Hermann and Cie, 1950, 1951. See also
A. H. Zemanian, Distribution Theory and Transform Analysis, New York: McGraw-Hill,
1965.

j _ S(t – a)f(t) dt = f(a) (2.95)
we immediately obtain the formal result
S{S(t);s} = 71 J^ e's` 6(t) dt = (2.96)

ßa2 -°° V 2^r
Hence, we say the Fourier transform of the impulse function 6(t) is the
constant 1/\/21x. On the other hand, we can use this transform relation
to deduce that
sw '{l;t} = 1 f e
- t (1) ds = V r 6(s) (2.97)
Of course, the integral in (2.97) has no meaning as an ordinary integral.

Its interpretation lies strictly in the fact that it is used to define the
generalized function S(s).
To better understand how the concept of an impulse function might
arise in a practical situation, let us briefly discuss the case where f(t)
represents a function in the time domain. For example, if f(t) is a waveform
— an electrical signal like a voltage or current, or an acoustic wave or
optical wave, etc. — the Fourier transform of f(t) describes the waveform
in the frequency domain. It is customary in this setting to let s = w,
where w is the angular frequency variable. Although the waveform phys-
ically exists in the time domain, we can say that it consists of those
components in its frequency domain description called its spectrum.
Mathematically, we write*
F(.o) ^ L f e ; »' f(t) dt = IF(w)Ie`«" (2.98))
where JF(W)I is the spectrum amplitude function and /((o) describes the
spectrum phase.
The sinusoidal functions f(t) = cos coo t, which represents one of the
simplest waveforms possible, does not satisfy the condition (2.94), and
hence, does not have a Fourier transform in the strict sense. Yet, from
a purely physical point of view we know this waveform has a single
frequency component, or line spectrum, at w = Wo . This apparent con-
tradiction can be explained by recognizing that the function cos Co ot
cannot exist for all time – cc < t < cc as we assume in the formal
definition of the Fourier transform. Like all waveforms, the cosine
waveform exists for only some finite interval of time, and as such will
* In much of the engineering literature, the spectrum function is defined by a multiple

of the complex conjugate of (2.98).

actually satisfy the condition of absolute integrability. For instance, let

us assume that f(t) = (cos u, o t)h(T — (ti), where 2T is the interval of
time that the waveform is present. In this case, the Fourier transform
of f(t) leads to
1 T
{ f(t);W} = J T e cos co o t dt
_
^ 2ir
i T
cos wt cos co o t dt
— ir 0
or
F(w) =
_ L r sin(w + w o)T + sin(w — Wa)T1 (2.99)
\/21r [.
W + CAD Cd — (00 ]
The graph of IF(co)i is shown in Fig. 2.7 for a fixed value of T. By allowing
T to become unbounded, it can be shown that the graph becomes more
and more narrow and peaked around W = m o and w = — w o . Hence, in
a formal sense we may consider the Fourier transform of cos coot for all
time as the limit
1 [ )
{cos Wot;u,} = lim
sin (wT
T—
+ coo
2ir W + á1p
+
sin()T
cw wo
—
W W
J (2.100)
provided this limit is meaningful.
In order to interpret the limit (2.100), let us recall that in the proof
of Theor. 2.1 (see Sec. 2.3) we established the limit relation
f(x) _ !J o f m f(t) cos[s(t — x)] dt ds

(2.101)
sin X
_ i f f(t) L lim ) (t z x 1 dt
Based on the sifting property (2.95), it becomes clear that the limit in
(2.101) leads to the formal definition
IF(w)I
— (o O cao
Figure 2.7 Spectrum of sinusoidal pulse

gg • Chap. 2/Fourier Integrals and Fourier Transforms
sin X(t — x)
I Jim = S(t — x) (2.102)
Ir ,ß_0o0 t—x
With this interpretation of the limit, we see that (2.100) gives the expected
result*
Jw{cos coot;cw} = \/?t/2 [S(w + wo) + S(co — Wo )] (2.103)
Hence, the notion of impulse functions is consistent with our previous
claim that the spectrum of cos w o t should be a single line at frequency
a, = wo (and from symmetry of the transform, a line also at w = — wo).
Example 2.16: Find the Fourier transform of sgn(t).
Solution: Recalling the transform relation [see Eq. (2.50)]
= i_ / 2 sgn(s)
it therefore follows from linverse transform relations that

21
f e' sgn(s) ds
^ .7r 'it
Interchanging the roles of t and s, and taking the complex conjugate

of the result, leads to
1 (^ e;st 21
sgn(t) dt = —
J is7r ^
and thus we deduce that
9{sgn(t);s} _
I 21i
'hIS
Example 2.17: Find the Fourier transform of the Heaviside unit function
h(t).
Solution: We first note that
h(t) = 2 [1 + sgn(t)]
Hence,
* We could also obtain (2.103) directly by writing cos m at = (e' °' + e - "'0')/2 and using
the formal relation (2.97).

S{sgn(t);s}
J{h(t);s} = 2 [S{l;s} +
J
2
= vI[&(s)+ 7rs
L]
where we are using (2.97)* and the result of Exam. 2.16.
Observe that, in Exam. 2.17, a superficial application of Eq. (2.57)
to S(t) = h'(t) would have resulted in
.3v{6(t);s} = — ism{ h (t);s}
leading to the incorrect transform relation
1
9{h(t);s} = —
is
That is, if sF(s) = sG(s), it does not follow that F(s) = G(s), but
rather that
F(s) = G(s) + k6(s)
where k is a constant. This is a consequence of the property s6(s) _
0 (recall Prob. 4a in Exer. 1.5).
2.8
1. Show formally that
(a) {6(t — a); s} _ ^^ eias
(b) 9W{e'°`;s} = V 8(s + a)

2. Find the Fourier transform of sin W ot.
3. Show that for any real function of time, the amplitude spectrum
IF(w) is necessarily an even function of W. Use this result to explain
why the Fourier transform of cos w ot leads to two impulse functions.
4. Using the property
{f " (t);s} _ (—is)"F(s),
( ) n = 1,2,3,...
show formally that
(a) 3W{S'(t);s} = —is.
(b) J{S " (t);s} = (—is)",
( ) n = 1,2,3,...
* Observe that {1;s} = 9 '{1;s} due to the even property of f(t)

- = 1.

5. Using the property

°{tnf(t);s} = (— i)nF (") ( s), n = 1,2,3,...
show formally that
(a) -F{t;s} = — iV/2zr S'(s)
(b) jV{tn;s} = (— i )n \/ S n)(s),
( n = 1,2,3,...
6. Based on the Fourier transform relation
i j t ;s t =i\2 sgn(s)
show that by writing
1 (- 1)n-' d o-1 1
t o (n — 1)! dt n-1
it can be formally deduced that
- .s l = i 2
nis) ni
1)! sgn(s),
(
n = 1,2,3,...
7. If J
Jf(t) dt = F(0) 0
where F(s) is the Fourier transform of f(t), show that
{l J f(x) dx;s}J = t F(s)

t
s
+ irF(0)S(s)
Hint: Write f f(x)dx = j — x)f(x) dx and use the convolution

— h(t
theorem.
8. Starting with the identity t6(t) = 0,
(a) show formally that
— i S(t)
(b) Use the result in (a) to deduce that

S(n)( t) = (-1)n (n! /t" ) 6(t), n = 1,2,3,...
9. If m and n are positive integers such that m < n, use Prob. 8 to
formally deduce that
t o -m S(n)(t) = (— l)n-m (n!/m!) S(m)(t)

2.9/Hilbert Transforms • 91
10. Based on the result of Prob. 9, formally show that

xt S(n)(t) _ j (n) to-k 8(k)(t)
(a) e -
k=0
(-1)kf k)(0)6(t)
(b) f(t)S (n) (t) = l 1
(s: )
In Probs. 11-15, verify the given Fourier transform relation.
11. {t'sgn(t);s} = 1 2 (—is)n+

n!
, , n = 0,1,2,...
n !
12. jw{tnh(t);s} = in S(n)(s) + ( Is
n = 0,1,2,...
l) n
13. 3{t «h(t);s} =

- r(1 — a) 151"-1 e'er('-«)S"(S), a> 1 (a nonintegral)
14. {ItI ";s} = ^

2 F(1 — a)IsI« - ` cos[Zlr(1 — a)],
7r
a > 1 (a non-
integral)
i_VÎ 2
15. S{^t^ - «sgn(t);s} = / 2 F(1 — a)IsI" ' sin[47r(1 — a)], - a>1
(a nonintegral)
2.9 Hilbert Transforms

In applications involving systems analysis to electrical networks, one
often finds the need to determine the frequency response function (i.e.,
the Fourier transform of a waveform) when only its real or imaginary
part is known. Basically the mathematical problem is to determine the
complex function
F(w) = R(o) + iX(oi) (2.104)
from only knowledge of the real functions R(a) or X(ca). If the function
f(t), which is the inverse Fourier transform of F(W), is a causal function
[i.e., f(t) = 0 for t < 0], then this mathematical problem can be solved
by means of Hilbert transforms, which fundamentally provide integral
relations between the real functions R(w) and X(w).
A related problem involving Hilbert transforms is to find a complex
representation of a real signal. For instance, if x(t) is a real signal and
x(t) is its Hilbert transform, then the analytic signal
u(t) = x(t) + i(t) (2.105)

has a Fourier transform that leads to a one-sided frequency spectrum,

i.e., the spectrum vanishes for negative frequencies. Analytic signal rep-
resentations are particularly useful in the analysis of narrowband
waveforms.
For consistency of notation with that normally used in the discussion
of Hilbert transforms, let us redefine the Fourier transform of a given
function f(t) by
{f(t);c^} = j ef(t) dt = F(w) (2.106)
The corresponding inversion formula then takes the form
.^ '{F(w),t} er°' F(w) dco = f(t) (2.107)

—1 J —
If f(t) is a real function, it follows that
R(a,) = J f(t)cos wt dt (2.108)

and
X(w) = – I f(t)sin wt dt (2.109)
where R(W) and X(w) are the real and imaginary parts of F(W) as given
in Eq. (2.104). We can immediately deduce from this that R(W) is an
even function and that X(w) is an odd function.
In the special case when f(t) is an even function, we see that
R(w) = 2fo f(t)cos ot dt (2.110)

and X(w) = 0, and through the inverse cosine transform relation, it can
be shown that
f(t) _ f R(w)cos wt dcw (2.111)
Similarly, when f(t) is an odd function we have R(w) = 0 and
X(w) _ – 2 f f(t)sin cut dt (2.112)
Thus,
f(t) _ – lo X(W)sin wt dW (2.113)

We see, therefore, that if f(t) is an even (odd) function, it can be recovered

entirely from the real (imaginary) part of its Fourier transform F(W).
Regardless of whether a function is even or odd, it can always be
expressed as the sum of an even and an odd function. That is, by writing
f(t) = 2[f(t) + f( - t)] + if f(t) - .f( - t)]
we have
f(t) = fe(t) + L(t) (2.114)

where
fe(t) = iff(t) + f(—t)] (2.115)
is an even function and
fa(t) = 2[f(t) — .f( — t)] (2.116)
is an odd function. From our discussion above, it now follows that R(W)
is the Fourier transform of fe (t) while X(co) is the Fourier transform of
fo(t). Hence, we have the transform pairs
R(W) = 21 fe (t)cos oit dt (2.117)
fe(t) = Jo R(a)cos Wt dw (2.118)
and
X(w) = —2 fo f0(t)sin cat dt (2.119)
f0(t) _ ô X(w)sin wt dw (2.120)

In the case when f(t) is causal, it happens that f( — t) = 0 for t > 0, and
therefore we deduce from (2.115) and (2.116) that
f(t) = 2fe(t) = 2fo (t), t > 0 (2.121)
Under this condition, Eqs. (2.118) and (2.120) can be written as
f(t) = 2 Io R(ca)cos wt dw (2.122)

and
f(t) = --- f X(w)sin wt dw (2.123)

Hence, a causal function f(t) can always be recovered from knowledge

of either R(w) or X(w), which suggests a possible relation between the
functions R(w) and X(w).
To obtain the relation between R(w) and X(w), we first note that fe (t)
and f (t) are related by
0
fe(t) = fo (t)sgn(t) (2.124)

f (t) = fe (t)sgn(t)
0 (2.125)
Hence, using (2.124) and properties of even and odd functions, we can
rewrite (2.117) as
R(w) =l e W -ißt fe (t) dt
=fe - "' fo (t)sgn(t) dt

t (2.126)
Treating this last expression as the Fourier transform of a product, we

can use the convolution integral to evaluate it. Recalling that
{ fo (t);w} = iX(w) (2.127)
and
J7{sgn(t);w} = 2 /kw (2.128)
we find that (2.126) is equivalent to
R(w) 2 -
J iX(Y)
[i(w 2 Y)) dy
or
R(w) = 1 ° X(Y) dy
(2.129)
7r -oo w—y
Similarly, since
iX(w) =fe - "" fo(t) dt

t
=fe - ""t fe(t)sgn(t) dt
we likewise deduce that

X(w) = — 1 f—
R(Y) dy
(2.130)
7rJ —w—y
Equations (2.129) and (2.130) form what we call a Hilbert transform pair.
They are also called Kramers —Krönig relations in electromagnetic theory

and are basically expressions of the causality condition, namely, the

effect must not precede in time the cause producing it.
What we have shown by (2.129) and (2.130) is that the real and
imaginary parts of the Fourier transform of a causal function satisfy
Hilbert transform relations. Conversely, it can also be shown (although
we do not give the proof here) that if the real and imaginary parts of
the Fourier transform F(w) satisfy Hilbert transform relations (2.129) and
(2.130), the inverse Fourier transform f(t) is a causal function.
Finally, there is an alternate form of the Hilbert transform pair that
is sometimes more convenient to use. We obtain this alternate form by
first recognizing that, since R(w) is even and X(w) is odd,
R(— w) _ — - f — w (+) dy = R(w) (2.131)

Y
Summing (2.129) and (2.131) leads to
2R(w) = 1 f X(Y) ( w 1 y w + y ) dY

R(w) = 2J° YX(Y )Z
0
dy (2.132)
IT w —y
In the same manner, we can show that
X(w) = — 2w — R(y) Z dy (2.133)
IT fo w- y
EXERCISES 2.9
In Probs. 1-4, verify that the given functions satisfy the Hilbert transform
relations (2.129) and (2.130).
a
1.R(w) = — w2
+ a 2 ' X(w) w2 + a 2
1 — cos w — w sin w 9 X(w) = sin w — ^2 cos w
2. R(w) =
a(a2 + w 2 + 1)
3. R(w) = 2)2
(a 2 + + 2(a 2 — w 2) + 1
w(1 — a 2 — w 2)
=
X(w) (a2 + w2) 2 + 2(a 2 — w 2) + 1,a>0
a cos co — a sines _ — wcosw — a sin w

4. R(w) = X(w) a>0
w2 + a2 w2 + a2

5. Prove that the Hilbert transform of a constant is zero, i.e., show

that dy
=0
J^(a—y
6. Given that R(w) _ — a /(WZ + a z ), determine the causal function f(t).
7. Given that X(w) _ — W/(w 2 + a 2 ), determine the causal function
f(t).
8. Show that
(a) X(w) = —
IT
i
f-
W - y
-
R(w) — R(y)
2w (°° R (w) — R(y)

dy
(b) X( )w _ Jo
dy
^,z — yz
(c) X(W) Jo
^r
R'(y)log w + yl dy
Hint: Integrate by parts in the result of (b).
9. Let F(w) be a complex function of w that is analytic in the upper-
half w-plane and tends to zero as w - > 0. [The condition that F(w)
has no poles in the upper-half w-plane is equivalent to the causality
condition that the inverse transform of F(w) vanishes for t < 0.]
Then show that (2.129) and (2.130) follow by applying Cauchy's
integral theorem to the integral
I(ce 1 ) =
L F(w) dw,
—(w 1 —is)
s —* 0 +
where w I is real, and the contour C is shown in the following figure

as the sum of C, along the real axis and C2 at infinity.
;(w)

2.10/Additional Topics • 97
10. If z(t) is the Hilbert transform of x(t), defined by
x(t) = 1IT J— x(T) dT

— °mot — T
show that x(t) and z(t) satisfy the orthogonality relation
I x(t)2(t) dt = 0
11. Let x(t) be a real waveform which has no constant component. Show
that x(t) and its Hilbert transform z(t) satisfy the Parseval relation
f Jx(t)J 2dt =f — Iz(t)I 2 dt
2.10 Additional Topics

In this final section on Fourier transforms we wish to touch briefly upon
some related topics.
2.10.1 Method of Stationary Phase

In the use of Fourier transforms or inverse transforms we often have to
deal with integrals whose evaluation by standardized methods is either
difficult or impossible. In such cases we may have to use a numerical
or approximation method to obtain the desired results. Sometimes we
may only need to know the asymptotic behavior of the integral for large
values of a parameter. In such instances Kelvin's method of stationary
phase can be very useful.
To illustrate the technique, let us consider the integral
I(t) =f a
b F(s)e"G(s ds
) (2.134)
whose asymptotic behavior for t —* ce is that which we seek. It is assumed

that a and b are real constants and that F(s) and G(s) are both twice
continuously differentiable real functions of s. Kelvin argued that the
integrand in (2.134), regarded as a function of s, oscillates with increasing
rapidity as t — cc so that the contributions to I(t) of adjacent portions
of the integrand cancel one another except in the immediate vicinity of
the end points of the interval and in the vicinity of those points at which
G(s) is stationary, i.e., points where G'(s) = 0. Moreover, in the first
approximation the contribution of stationary points is more significant
than that of the end points. Hence, if s,, is a point on the interval a <
s,, < b for which G'(s a) = 0 and G"(s o) # 0, then G(s) may be approximated
by the finite Taylor series

G(s) = G(s o) + z G"(so)(s — s o) Z , Is —s o d <e (2.135)

Similarly, we approximate F(s) by F(s o), the first term in its Taylor
expansion. Under the assumption that the major contribution to the
integral occurs in the neighborhood of s o , these approximations lead to
So +E
I(t) F(so)e"c(s ° ) r e
" c"csoxs-so)2ds (2.136)
J Sp—E
By formally extending the limits of integration over (— 00,00) and using
the result (see Prob. 11 in Exer. 2.10)
J e ''°Z"2 du = e'r'4 (2.137)
âl
we deduce that
1(t)— F(s o) 2ir/tl G"(s o)l exp[itG(so ) ± iir/4], t — o (2.138)
where the plus sign goes with G"(s o) > 0 and the minus sign with G"(s o)
< 0.
Example 2.18: Find the asymptotic behavior for t —> : of
I(t) = f 1 cos[t(s 2 — s)] ds

0
Solution: Here we first write
I(t) = Re{ J e'ns 2- s) ds

11 0
from which we identify F(s) = I and G(s) = s 2 — s. Thus, calculating

the derivatives
G'(s) = 2s — 1, G"(s) = 2
we find s o = 1/2, and from (2.138) it immediately follows that
I(t) — r cos 4(t — ir), t -* 00
For a more thorough investigation of the method of stationary phase,

the reader is advised to consult A. Erdelyi, Asymptotic Expansions, New
York: Dover, 1956.
2.9.2 Multiple Fourier Transforms

The Fourier transform pairs (2.27) and (2.28) may be extended to functions
of two or more variables. For example, if f(x,y) is a piecewise continuous
function such that

f j 1f(x,y)j dx dy <o (2.139)

we may consider the double Fourier transform pair
F(e, y) = —1 1-^ 1 f(x, Y) dx dy (2.140)
and
f(x,Y) = j e t ' F(e, r1)ded'rl (2.141)

1 J
We can interpret (2.140) as an iterated Fourier transform applied se-
quentially to the variable x and then to y. That is, we define
f* (g,Y) = 1 J e'^z f(x,Y) dx
followed by
F(, v) = + f - e"'y f* (^,Y) dy
The inverse transform (2.141) follows in a similar manner. The double

Fourier transform and inverse double Fourier transform are also denoted,
respectively, by the notation
(2){f(x,Y);x ê, Y -} = F(e,77) --
(2.142)
and
5Wcz>'{F(f,?1);e^x, ^1—y} = f(x,y) (2.143)
Operational properties of the Fourier transform developed in Sec. 2.5
carry over in a natural sort of way to the double Fourier transform
defined here. For example, the analogues of the shifting properties (Theor.
2.3) are
5'(2){e'cax+by f(x,Y);x - f, y - i} = F(f + a, 71 + b) (2.144)
and
cz>{.f(x - a, y - b);x-+e, y "I = e`c°C+ "' F(e,-1)
- (2.145)
The double Fourier transform of derivatives leads to the results
x(2)1 (x,Y);x ^^, Y^^1^ _

--
- i^F(^, 7I) (2.146)
(2) a?f (x,Y);x -> f, Y^^1^ _ - ^r1F(C,-1) (2.147)

` axay
and so on, and the convolution theorem assumes the form

J{F(, q)G( , );x-*, y-^ ?}
= 2_ f -^ f - f(x-u, y-v)g(u,v)du dv (2.148)
The verification of these properties is left to the reader.

Generalizations of the double Fourier transform and its properties to
n-dimensional Fourier transforms is obvious. For example, in three vari-
ables we define
F(^,-q,g) = (21r)-312

fJf e,ctx+,,y+^'
f(x,y,z)dx dy dz (2.149)
with a similar expression for the inverse transform. For reasons of the-
oretical tractability, one normally does not use Fourier transforms in
dimensions greater than two. Actually, even though physical systems all
have three-dimensions, we are often able to ignore one or two dimensions
in our analysis by using some reasonable simplifying assumption, such
as symmetry, and so on. The classical example where simplification to
one or two dimensions is not possible, however, is the problem of diffraction
of X-rays by crystals, which must be analyzed in three dimensions.
EXERCISES 2.10
In Probs. 1-3, verify the given asymptotic behavior of each integral.
1.
r' 1 Ilr
t e;,.a,
eus=ds--' t-+00
-
2. (
Jo
t
e
u ""-'
ds - E e' `
2t
(
4) t -> 00
3.fó cos (ts 2 - s) ds -- 1

^2t'
t-)00
4. Using properties of the Fr es nel integrals (Sec. 1.3.2), show that
(a)fo cos [t(s 2 - s)]ds =

J 2 cos C(1/2^) +sin 4 S I ^2^^
(b) As t - oo, show that the result on the right-hand side in (a)
approaches that given in Exam. 2.18.
5. If G'(s ) = G"(s o) = ... = G (m - '^(s ) = 0, G (m'(s ) 0, show that
o 0 0
Eq. (2.138) is generalized to
f F(s)e ^^ccs^
ds - r(lm ) F(so) itG(s o) ± 2m
.' [ tIG(m)(s.)]'/mexp

6.Given the integral representation

1
J(x) = -cos(x sin s - ns) ds, n = 0,1,2,...
T o
show that large-order Bessel functions have the asymptotic behavior
J(x) 00
'13' n
— 2 21 3 1 6irn
Hint: Use Prob. 5.

In Probs. 7-10, use the method of stationary phase to find an asymptotic
expression for the given integral as t -^ 00• In some cases it may be
necessary to use the result of Prob. 5.
7. J e"S 2 cosh s 2 ds 8. f co(ts)tan s ds

0 o
' 1
9. '
1 0
e ;`(s_S;° S) ds
10. fo sin[t(s + 6s 3 -
sinh s)]cos s ds
11. Show that
(a) w 1 f e _ia2 u 1
2
du = a
I l f Vu dv
2
o
cos v
i o Vu
sin v
v
dv
V
± 2U2 du = 1 e14
2%Í f e jai
12. Show that
3 (2){f(ax,by);x^C, y - ^ 71} = ^ I F(ela, i1/b)

13. Show that
JW(2){f(x - a, y - b);x -- , y-.-q} = e' c"f+b7, 'F(e,71)
14. Show that
3 (2){f(x,y)cos oix;x-+e, y—'q} = 2[F(e+w,71) + F(f -w,77)]
15. Verify the convolution formula
i1 *y} = (f * g)(x,y)
-
where
(f*g)(x,y) = 1 J^ f f(u,v)S(x u, y-v) du dv -
21r -

Fourier Integrals and Fourier Transforms

Uploaded by

Copyright:

Available Formats

Fourier Integrals and Fourier Transforms

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fourier Integrals and Fourier Transforms

Uploaded by

Copyright:

Available Formats

2

Fourier Integrals and

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

in his work on heat conduction that an "arbitrary function" could be

2.2 Fourier Integral Representations

An important problem in mathematical analysis is the determination of

* For a general discussion of Fourier series, see L. C. Andrews, Elementary Partial

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

The theory of Fourier series shows that a periodic function satisfying

where we have interchanged the order of summation and integration and

f If(t)dt <00 (2.6)

For the remaining infinite sum in (2.5), it is convenient to let As = Ir /p

and then consider the equivalent limit

* By "formal," we mean a procedure that is not mathematically rigorous.

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

f(t) cos[nis(t - x)]Os dt (2.8)

(Observe that Os -> 0 as p -+ When As is a small positive number,

Upon switching the order of integration, we get the equivalent form

f(x) o J f(t)cos[s(t - x)]dt ds (2.10)

f(x) _ -- f J_ f(t)cos[s(t - x)]dt ds

at points x where f is continuous. If x is a point of discontinuity of f,

the right-hand and left-hand limits,t

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

f(x) =ff o — f(t)(cos stcos sx + sin stsin sx)dt ds

f(x) = f [A(s)cos sx + B(s)sin

A(s) = - f f(t)cos st dt (2.12)

In this setting we refer to (2.11) as the Fourier integral representation

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

42 • Chap. 2/Fourier Integrals and Fourier Transforms

Solution: The coefficients A(s) and B(s) are given by

Thus the Fourier integral representation becomes

Observe that at x = ± 1 there is a jump discontinuity in the function f

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

= 1 ( sin[s(1 + x)] ds + 1 f" sin[s(1 - x)] ds

1 " ' + x) sint

where Si(z) is the sine integral defined by

Equation (2.17) is plotted in Fig. 2.2 for values µ = 4,16,128.

2.2.1 Cosine and Sine Integral Representations

B(s) = --f(x)sin sx dx = 0 (2.20)

Figure 2.2 The partial integral of f(x) = h(1 - jxj)

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

We refer to (2.21) as a Fourier cosine integral representation. In a similar

f(x) = fo B(s)sin sx ds (2.22)

where A(s) = 0 and

B(s) = 27r fo^f(x)sin sx dx (2.23)

Finally, if f should be a function defined only on the interval 0 <

.f(x)_ cos0, x, 0 <x < ir/2

Solution: For a cosine integral representation, we compute

In the same manner

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

from which we deduce the sine integral representation

(b) fo,sin axdx= ^,

f(x) = p — x2 ' ^x^ > 1

4.Use the result of Prob. 3 to deduce that

Downloaded From: https://www.spiedigitallibrary.org/ebooks on 2/2/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

sin x, IxI < 7r sin x, 0 < x < ir

12. Show that e - Xcos x has the half-range representations

(a) e - xcos x= 2 f o s' + 4 cos sx ds, x> 0

13. Using the results of Prob. 11, show that