c06.qxd
10/28/10
6:33 PM
Page 203
CHAPTER
6
Laplace Transforms
Laplace transforms are invaluable for any engineer’s mathematical toolbox as they make
solving linear ODEs and related initial value problems, as well as systems of linear ODEs,
much easier. Applications abound: electrical networks, springs, mixing problems, signal
processing, and other areas of engineering and physics.
The process of solving an ODE using the Laplace transform method consists of three
steps, shown schematically in Fig. 113:
Step 1. The given ODE is transformed into an algebraic equation, called the subsidiary
equation.
Step 2. The subsidiary equation is solved by purely algebraic manipulations.
Step 3. The solution in Step 2 is transformed back, resulting in the solution of the given
problem.
IVP
Initial Value
Problem
1
AP
Algebraic
Problem
2
Solving
AP
by Algebra
3
Solution
of the
IVP
Fig. 113. Solving an IVP by Laplace transforms
The key motivation for learning about Laplace transforms is that the process of solving
an ODE is simplified to an algebraic problem (and transformations). This type of
mathematics that converts problems of calculus to algebraic problems is known as
operational calculus. The Laplace transform method has two main advantages over the
methods discussed in Chaps. 1–4:
I. Problems are solved more directly: Initial value problems are solved without first
determining a general solution. Nonhomogenous ODEs are solved without first solving
the corresponding homogeneous ODE.
II. More importantly, the use of the unit step function (Heaviside function in Sec. 6.3)
and Dirac’s delta (in Sec. 6.4) make the method particularly powerful for problems with
inputs (driving forces) that have discontinuities or represent short impulses or complicated
periodic functions.
203
c06.qxd
10/28/10
204
6:33 PM
Page 204
CHAP. 6 Laplace Transforms
The following chart shows where to find information on the Laplace transform in this
book.
Topic
Where to find it
ODEs, engineering applications and Laplace transforms
PDEs, engineering applications and Laplace transforms
List of general formulas of Laplace transforms
List of Laplace transforms and inverses
Chapter 6
Section 12.11
Section 6.8
Section 6.9
Note: Your CAS can handle most Laplace transforms.
Prerequisite: Chap. 2
Sections that may be omitted in a shorter course: 6.5, 6.7
References and Answers to Problems: App. 1 Part A, App. 2.
6.1
Laplace Transform. Linearity.
First Shifting Theorem (s-Shifting)
In this section, we learn about Laplace transforms and some of their properties. Because
Laplace transforms are of basic importance to the engineer, the student should pay close
attention to the material. Applications to ODEs follow in the next section.
Roughly speaking, the Laplace transform, when applied to a function, changes that
function into a new function by using a process that involves integration. Details are as
follows.
If f (t) is a function defined for all t ⭌ 0, its Laplace transform1 is the integral of f (t)
times eⴚst from t ⫽ 0 to ⬁ . It is a function of s, say, F(s), and is denoted by l( f ); thus
(1)
F(s) ⫽ l( f ) ⫽
˛
冮
ⴥ
eⴚstf (t) dt.
0
Here we must assume that f (t) is such that the integral exists (that is, has some finite
value). This assumption is usually satisfied in applications—we shall discuss this near the
end of the section.
1
PIERRE SIMON MARQUIS DE LAPLACE (1749–1827), great French mathematician, was a professor in
Paris. He developed the foundation of potential theory and made important contributions to celestial mechanics,
astronomy in general, special functions, and probability theory. Napoléon Bonaparte was his student for a year.
For Laplace’s interesting political involvements, see Ref. [GenRef2], listed in App. 1.
The powerful practical Laplace transform techniques were developed over a century later by the English
electrical engineer OLIVER HEAVISIDE (1850–1925) and were often called “Heaviside calculus.”
We shall drop variables when this simplifies formulas without causing confusion. For instance, in (1) we
wrote l( f ) instead of l( f )(s) and in (1*) lⴚ1(F) instead of lⴚ1 (F)(t).
c06.qxd
10/28/10
6:33 PM
Page 205
SEC. 6.1 Laplace Transform. Linearity. First Shifting Theorem (s-Shifting)
205
Not only is the result F(s) called the Laplace transform, but the operation just described,
which yields F(s) from a given f (t), is also called the Laplace transform. It is an “integral
transform”
冮
F(s) ⫽
ⴥ
k(s, t) f (t) dt
0
with “kernel” k(s, t) ⫽ eⴚst.
Note that the Laplace transform is called an integral transform because it transforms
(changes) a function in one space to a function in another space by a process of integration
that involves a kernel. The kernel or kernel function is a function of the variables in the
two spaces and defines the integral transform.
Furthermore, the given function f (t) in (1) is called the inverse transform of F(s) and
is denoted by lⴚ1(F ); that is, we shall write
˛
f (t) ⫽ lⴚ1(F ).
(1*)
Note that (1) and (1*) together imply l⫺1(l( f )) ⫽ f and l(l⫺1(F )) ⫽ F.
Notation
Original functions depend on t and their transforms on s—keep this in mind! Original
functions are denoted by lowercase letters and their transforms by the same letters in capital,
so that F(s) denotes the transform of f (t), and Y(s) denotes the transform of y(t), and so on.
EXAMPLE 1
Laplace Transform
Let f (t) ⫽ 1 when t ⭌ 0. Find F(s).
Solution.
From (1) we obtain by integration
l( f ) ⫽ l(1) ⫽
冮
ⴥ
0
1
e⫺st dt ⫽ ⫺ e⫺st `
s
ⴥ
⫽
0
1
s
(s ⬎ 0).
Such an integral is called an improper integral and, by definition, is evaluated according to the rule
冮
T
ⴥ
eⴚstf (t) dt ⫽ lim
T:⬁
0
冮e
ⴚst
f (t) dt.
0
Hence our convenient notation means
冮
eⴚst dt ⫽ lim
0
1
1
1
1
lim c ⫺ eⴚsT ⫹ e0 d ⫽
c ⫺ s eⴚst d ⫽ T:⬁
s
s
s
T
ⴥ
T:⬁
䊏
We shall use this notation throughout this chapter.
EXAMPLE 2
(s ⬎ 0).
0
Laplace Transform l (eat) of the Exponential Function eat
Let f (t) ⫽ eat when t ⭌ 0, where a is a constant. Find l( f ).
Solution.
Again by (1),
l(eat) ⫽
冮
ⴥ
ⴥ
eⴚsteat dt ⫽
0
1
eⴚ(sⴚa)t 2 ;
a⫺s
0
hence, when s ⫺ a ⬎ 0,
l(eat) ⫽
1
.
s⫺a
䊏
c06.qxd
10/28/10
6:33 PM
206
Page 206
CHAP. 6 Laplace Transforms
Must we go on in this fashion and obtain the transform of one function after another
directly from the definition? No! We can obtain new transforms from known ones by the
use of the many general properties of the Laplace transform. Above all, the Laplace
transform is a “linear operation,” just as are differentiation and integration. By this we
mean the following.
THEOREM 1
Linearity of the Laplace Transform
The Laplace transform is a linear operation; that is, for any functions f (t) and g(t)
whose transforms exist and any constants a and b the transform of af (t) ⫹ bg(t)
exists, and
l{af (t) ⫹ bg(t)} ⫽ al{f (t)} ⫹ bl{g(t)}.
PROOF
This is true because integration is a linear operation so that (1) gives
l{af (t) ⫹ bg(t)} ⫽
冮
ⴥ
eⴚst3af (t) ⫹ bg(t)4 dt
0
⫽a
冮
ⴥ
eⴚstf (t) dt ⫹ b
0
EXAMPLE 3
冮
ⴥ
eⴚstg(t) dt ⫽ al{f (t)} ⫹ bl{g(t)}. 䊏
0
Application of Theorem 1: Hyperbolic Functions
Find the transforms of cosh at and sinh at.
Solution.
Since cosh at ⫽ 12 (eat ⫹ eⴚat) and sinh at ⫽ 12 (eat ⫺ eⴚat), we obtain from Example 2 and
Theorem 1
l(cosh at) ⫽
l(sinh at) ⫽
EXAMPLE 4
1
2
1
2
(l(eat) ⫹ l(eⴚat)) ⫽
(l(eat) ⫺ l(eⴚat)) ⫽
1
2
a
1
s⫺a
⫹
1
s⫹a
b⫽
s
s2 ⫺ a2
1
1
1
a
⫺
.
a
b⫽ 2
2 s⫺a
s⫹a
s ⫺ a2
䊏
Cosine and Sine
Derive the formulas
l(cos vt) ⫽
s
s ⫹v
2
2
l(sin vt) ⫽
,
v
s ⫹ v2
2
.
We write L c ⫽ l(cos vt) and L s ⫽ l(sin vt). Integrating by parts and noting that the integralfree parts give no contribution from the upper limit ⬁ , we obtain
Solution.
Lc ⫽
冮
ⴥ
冮
ⴥ
eⴚst cos vt dt ⫽
0
Ls ⫽
0
ⴥ
eⴚst
v
cos vt 2 ⫺
⫺s
s
0
eⴚst sin vt dt ⫽
ⴥ
eⴚst
v
sin vt 2 ⫹
⫺s
s
0
冮
ⴥ
eⴚst sin vt dt ⫽
1
v
⫺ L s,
s
s
eⴚst cos vt dt ⫽
v
L .
s c
0
冮
ⴥ
0
c06.qxd
10/28/10
7:44 PM
Page 207
SEC. 6.1 Laplace Transform. Linearity. First Shifting Theorem (s-Shifting)
207
By substituting L s into the formula for L c on the right and then by substituting L c into the formula for L s on
the right, we obtain
Lc ⫽
1
v v
⫺ a Lcb ,
s
s s
L c a1 ⫹
v2
1
b⫽ ,
s
s2
Lc ⫽
s
,
s 2 ⫹ v2
Ls ⫽
v 1
v
a ⫺ Lsb ,
s s
s
L s a1 ⫹
v2
v
b ⫽ 2,
s2
s
Ls ⫽
v
.
s 2 ⫹ v2
䊏
Basic transforms are listed in Table 6.1. We shall see that from these almost all the others
can be obtained by the use of the general properties of the Laplace transform. Formulas
1–3 are special cases of formula 4, which is proved by induction. Indeed, it is true for
n ⫽ 0 because of Example 1 and 0! ⫽ 1. We make the induction hypothesis that it holds
for any integer n ⭌ 0 and then get it for n ⫹ 1 directly from (1). Indeed, integration by
parts first gives
l(t n⫹1) ⫽
冮
⬁
ⴥ
1
n⫹1
eⴚstt n⫹1 dt ⫽ ⫺ s eⴚstt n⫹1 2 ⫹ s
0
0
冮
ⴥ
eⴚstt n dt.
0
Now the integral-free part is zero and the last part is (n ⫹ 1)>s times l(t n). From this
and the induction hypothesis,
l(t n⫹1) ⫽
n⫹1
n ⫹ 1 # n!
(n ⫹ 1)!
l(t n) ⫽
⫽ n⫹2 .
s n⫹1
s
s
s
This proves formula 4.
Table 6.1 Some Functions ƒ(t) and Their Laplace Transforms ᏸ( ƒ)
ƒ(t)
ᏸ(ƒ)
1
1
1>s
7
cos ␻ t
2
t
1>s 2
8
sin ␻ t
3
t2
2!>s 3
9
cosh at
4
tn
(n ⫽ 0, 1, • • •)
10
sinh at
11
eat cos ␻ t
12
eat sin ␻ t
5
6
n!
s
n⫹1
ta
(a positive)
⌫(a ⫹ 1)
eat
1
s⫺a
s
a⫹1
ƒ(t)
ᏸ(ƒ)
s
s 2 ⫹ v2
v
s ⫹ v2
2
s
s ⫺ a2
2
a
s ⫺ a2
2
s⫺a
(s ⫺ a) 2 ⫹ v2
v
(s ⫺ a) 2 ⫹ v2
c06.qxd
10/28/10
6:33 PM
208
Page 208
CHAP. 6 Laplace Transforms
⌫(a ⫹ 1) in formula 5 is the so-called gamma function [(15) in Sec. 5.5 or (24) in
App. A3.1]. We get formula 5 from (1), setting st ⫽ x:
l(t a) ⫽
ⴥ
冮
eⴚstta dt ⫽
0
冮
a
ⴥ
x dx
1
eⴚx a b
⫽ a⫹1
s
s
s
0
冮
ⴥ
eⴚxx a dx
0
where s ⬎ 0. The last integral is precisely that defining ⌫(a ⫹ 1), so we have
⌫(a ⫹ 1)>s a⫹1, as claimed. (CAUTION! ⌫(a ⫹ 1) has x a in the integral, not x a⫹1.)
Note the formula 4 also follows from 5 because ⌫(n ⫹ 1) ⫽ n! for integer n ⭌ 0.
Formulas 6–10 were proved in Examples 2–4. Formulas 11 and 12 will follow from 7
and 8 by “shifting,” to which we turn next.
s-Shifting: Replacing s by s ⫺ a in the Transform
The Laplace transform has the very useful property that, if we know the transform of f (t),
we can immediately get that of eatf (t), as follows.
THEOREM 2
First Shifting Theorem, s-Shifting
If f (t) has the transform F(s) (where s ⬎ k for some k), then eatf (t) has the transform
F(s ⫺ a) (where s ⫺ a ⬎ k). In formulas,
l{eatf (t)} ⫽ F(s ⫺ a)
or, if we take the inverse on both sides,
eatf (t) ⫽ lⴚ1{F(s ⫺ a)}.
PROOF
We obtain F(s ⫺ a) by replacing s with s ⫺ a in the integral in (1), so that
F(s ⫺ a) ⫽
冮
ⴥ
eⴚ(sⴚa)tf (t) dt ⫽
0
冮
ⴥ
0
eⴚst3eatf (t)4 dt ⫽ l{eatf (t)}.
If F(s) exists (i.e., is finite) for s greater than some k, then our first integral exists for
s ⫺ a ⬎ k. Now take the inverse on both sides of this formula to obtain the second formula
in the theorem. (CAUTION! ⫺a in F(s ⫺ a) but ⫹a in eatf (t).)
䊏
EXAMPLE 5
s-Shifting: Damped Vibrations. Completing the Square
From Example 4 and the first shifting theorem we immediately obtain formulas 11 and 12 in Table 6.1,
l{eat cos vt} ⫽
s⫺a
(s ⫺ a) ⫹ v
2
2
,
l{eat sin vt} ⫽
For instance, use these formulas to find the inverse of the transform
l( f ) ⫽
3s ⫺ 137
s ⫹ 2s ⫹ 401
2
.
v
(s ⫺ a)2 ⫹ v2
.
c06.qxd
10/28/10
6:33 PM
Page 209
SEC. 6.1 Laplace Transform. Linearity. First Shifting Theorem (s-Shifting)
Solution.
209
Applying the inverse transform, using its linearity (Prob. 24), and completing the square, we obtain
f ⫽ lⴚ1b
3(s ⫹ 1) ⫺ 140
(s ⫹ 1) ⫹ 400
2
r ⫽ 3lⴚ1b
s⫹1
(s ⫹ 1) ⫹ 20
2
2
r ⫺ 7lⴚ1b
20
(s ⫹ 1)2 ⫹ 202
r.
We now see that the inverse of the right side is the damped vibration (Fig. 114)
䊏
f (t) ⫽ eⴚt(3 cos 20t ⫺ 7 sin 20t).
6
4
2
0
0.5
1.0
1.5
2.0
2.5
t
3.0
–2
–4
–6
Fig. 114. Vibrations in Example 5
Existence and Uniqueness of Laplace Transforms
This is not a big practical problem because in most cases we can check the solution of
an ODE without too much trouble. Nevertheless we should be aware of some basic facts.
A function f (t) has a Laplace transform if it does not grow too fast, say, if for all t ⭌ 0
and some constants M and k it satisfies the “growth restriction”
ƒ f (t) ƒ ⬉ Mekt.
(2)
(The growth restriction (2) is sometimes called “growth of exponential order,” which may
be misleading since it hides that the exponent must be kt, not kt 2 or similar.)
f (t) need not be continuous, but it should not be too bad. The technical term (generally
used in mathematics) is piecewise continuity. f (t) is piecewise continuous on a finite
interval a ⬉ t ⬉ b where f is defined, if this interval can be divided into finitely many
subintervals in each of which f is continuous and has finite limits as t approaches either
endpoint of such a subinterval from the interior. This then gives finite jumps as in
Fig. 115 as the only possible discontinuities, but this suffices in most applications, and
so does the following theorem.
a
b
t
Fig. 115. Example of a piecewise continuous function f (t).
(The dots mark the function values at the jumps.)
c06.qxd
10/28/10
6:33 PM
210
Page 210
CHAP. 6 Laplace Transforms
THEOREM 3
Existence Theorem for Laplace Transforms
If f (t) is defined and piecewise continuous on every finite interval on the semi-axis
t ⭌ 0 and satisfies (2) for all t ⭌ 0 and some constants M and k, then the Laplace
transform l( f ) exists for all s ⬎ k.
PROOF
Since f (t) is piecewise continuous, eⴚstf (t) is integrable over any finite interval on the
t-axis. From (2), assuming that s ⬎ k (to be needed for the existence of the last of the
following integrals), we obtain the proof of the existence of l( f ) from
ƒ l( f ) ƒ ⫽ `
冮
ⴥ
0
eⴚstf (t) dt ` ⬉
冮
ⴥ
ƒ f (t) ƒ eⴚst dt ⬉
0
冮
ⴥ
Mekteⴚst dt ⫽
0
M
.
s⫺k
䊏
Note that (2) can be readily checked. For instance, cosh t ⬍ et, t n ⬍ n!et (because t n>n!
is a single term of the 2Maclaurin series), and so on. A function that does not satisfy (2)
for any M and k is et (take logarithms to see it). We mention that the conditions in
Theorem 3 are sufficient rather than necessary (see Prob. 22).
Uniqueness. If the Laplace transform of a given function exists, it is uniquely
determined. Conversely, it can be shown that if two functions (both defined on the positive
real axis) have the same transform, these functions cannot differ over an interval of positive
length, although they may differ at isolated points (see Ref. [A14] in App. 1). Hence we
may say that the inverse of a given transform is essentially unique. In particular, if two
continuous functions have the same transform, they are completely identical.
PROBLEM SET 6.1
1–16
LAPLACE TRANSFORMS
15.
Find the transform. Show the details of your work. Assume
that a, b, v, u are constants.
1. 3t ⫹ 12
2. (a ⫺ bt)2
3. cos pt
4. cos2 vt
2t
5. e sinh t
6. eⴚt sinh 4t
7. sin (vt ⫹ u)
8. 1.5 sin (3t ⫺ p>2)
9.
10.
k
1
c
1
11.
12.
b
1
1
b
13.
14.
1
2
k
2
a
–1
16.
1
b
1
0.5
1
17–24
1
2
SOME THEORY
17. Table 6.1. Convert this table to a table for finding
inverse transforms (with obvious changes, e.g.,
lⴚ1(1>s n) ⫽ t nⴚ1>(n ⫺ 1), etc).
18. Using l( f ) in Prob. 10, find l( f1), where f1(t) ⫽ 0
if t ⬉ 2 and f1(t) ⫽ 1 if t ⬎ 2.
19. Table 6.1. Derive formula 6 from formulas 9 and 10.
2
20. Nonexistence. Show that et does not satisfy a
condition of the form (2).
21. Nonexistence. Give simple examples of functions
(defined for all t ⭌ 0) that have no Laplace
transform.
22. Existence. Show that l(1> 1t) ⫽ 1p>s. [Use (30)
⌫(12) ⫽ 1p in App. 3.1.] Conclude from this that the
conditions in Theorem 3 are sufficient but not
necessary for the existence of a Laplace transform.
c06.qxd
10/28/10
6:33 PM
Page 211
SEC. 6.2 Transforms of Derivatives and Integrals. ODEs
23. Change of scale. If l( f (t)) ⫽ F(s) and c is any
positive constant, show that l( f (ct)) ⫽ F(s>c)>c (Hint:
Use (1).) Use this to obtain l(cos vt) from l(cos t).
24. Inverse transform. Prove that lⴚ1 is linear. Hint:
Use the fact that l is linear.
INVERSE LAPLACE TRANSFORMS
25–32
Given F(s) ⫽ l( f ), find f (t). a, b, L, n are constants. Show
the details of your work.
25.
27.
29.
31.
0.2s ⫹ 1.8
26.
s 2 ⫹ 3.24
s
L s ⫹n p
2 2
12
s
4
⫺
2
2
228
s
6
s ⫹ 10
s2 ⫺ s ⫺ 2
6.2
28.
30.
32.
5s ⫹ 1
211
33–45
41.
s 2 ⫺ 25
1
(s ⫹ 12)(s ⫺ 13)
4s ⫹ 32
s ⫺ 16
2
1
(s ⫹ a)(s ⫹ b)
APPLICATION OF s-SHIFTING
In Probs. 33–36 find the transform. In Probs. 37–45 find
the inverse transform. Show the details of your work.
33. t 2eⴚ3t
34. keⴚat cos vt
ⴚ4.5t
35. 0.5e
36. sinh t cos t
sin 2pt
p
6
37.
38.
2
(s ⫹ p)
(s ⫹ 1)3
4
21
39.
40. 2
4
s ⫺ 2s ⫺ 3
(s ⫹ 22)
p
s ⫹ 10ps ⫹ 24p2
a0
a2
a1
42.
⫹
2 ⫹
(s ⫹ 1)
(s ⫹ 1)3
s⫹1
43.
45.
2
2s ⫺ 1
s ⫺ 6s ⫹ 18
k 0 (s ⫹ a) ⫹ k 1
2
44.
a (s ⫹ k) ⫹ bp
(s ⫹ k)2 ⫹ p2
(s ⫹ a)2
Transforms of Derivatives and Integrals.
ODEs
The Laplace transform is a method of solving ODEs and initial value problems. The crucial
idea is that operations of calculus on functions are replaced by operations of algebra
on transforms. Roughly, differentiation of f (t) will correspond to multiplication of l( f )
by s (see Theorems 1 and 2) and integration of f (t) to division of l( f ) by s. To solve
ODEs, we must first consider the Laplace transform of derivatives. You have encountered
such an idea in your study of logarithms. Under the application of the natural logarithm,
a product of numbers becomes a sum of their logarithms, a division of numbers becomes
their difference of logarithms (see Appendix 3, formulas (2), (3)). To simplify calculations
was one of the main reasons that logarithms were invented in pre-computer times.
THEOREM 1
Laplace Transform of Derivatives
The transforms of the first and second derivatives of f (t) satisfy
(1)
l( f r ) ⫽ sl( f ) ⫺ f (0)
(2)
l( f s ) ⫽ s 2l( f ) ⫺ sf (0) ⫺ f r (0).
Formula (1) holds if f (t) is continuous for all t ⭌ 0 and satisfies the growth
restriction (2) in Sec. 6.1 and f r (t) is piecewise continuous on every finite interval
on the semi-axis t ⭌ 0. Similarly, (2) holds if f and f r are continuous for all t ⭌ 0
and satisfy the growth restriction and f s is piecewise continuous on every finite
interval on the semi-axis t ⭌ 0.
c06.qxd
10/28/10
6:33 PM
212
Page 212
CHAP. 6 Laplace Transforms
PROOF
We prove (1) first under the additional assumption that f r is continuous. Then, by the
definition and integration by parts,
l( f r ) ⫽
冮
ⴥ
0
e
f r (t) dt ⫽ 3e
ⴚst
f (t)4 `
ⴥ
ⴚst
0
⫹s
冮
ⴥ
eⴚstf (t) dt.
0
Since f satisfies (2) in Sec. 6.1, the integrated part on the right is zero at the upper limit
when s ⬎ k, and at the lower limit it contributes ⫺f (0). The last integral is l( f ). It exists
for s ⬎ k because of Theorem 3 in Sec. 6.1. Hence l( f r ) exists when s ⬎ k and (1) holds.
If f r is merely piecewise continuous, the proof is similar. In this case the interval of
integration of f r must be broken up into parts such that f r is continuous in each such part.
The proof of (2) now follows by applying (1) to f s and then substituting (1), that is
l( f s ) ⫽ sl( f r ) ⫺ f r (0) ⫽ s3sl( f ) ⫺ f (0)4 ⫽ s 2l( f ) ⫺ sf (0) ⫺ f r (0).
䊏
Continuing by substitution as in the proof of (2) and using induction, we obtain the
following extension of Theorem 1.
THEOREM 2
Laplace Transform of the Derivative f (n) of Any Order
Let f, f r , Á , f (nⴚ1) be continuous for all t ⭌ 0 and satisfy the growth restriction
(2) in Sec. 6.1. Furthermore, let f (n) be piecewise continuous on every finite interval
on the semi-axis t ⭌ 0. Then the transform of f (n) satisfies
(3)
EXAMPLE 1
l( f (n)) ⫽ s nl( f ) ⫺ s nⴚ1f (0) ⫺ s nⴚ2f r (0) ⫺ Á ⫺ f (nⴚ1)(0).
Transform of a Resonance Term (Sec. 2.8)
Let f (t) ⫽ t sin vt. Then f (0) ⫽ 0, f r (t) ⫽ sin vt ⫹ vt cos vt, f r (0) ⫽ 0, f s ⫽ 2v cos vt ⫺ v2t sin vt. Hence
by (2),
l( f s ) ⫽ 2v
EXAMPLE 2
s
s ⫹v
2
2
⫺ v2l( f ) ⫽ s 2l( f ),
thus
l( f ) ⫽ l(t sin vt) ⫽
2vs
(s ⫹ v2)2
2
.
䊏
Formulas 7 and 8 in Table 6.1, Sec. 6.1
This is a third derivation of l(cos vt) and l(sin vt); cf. Example 4 in Sec. 6.1. Let f (t) ⫽ cos vt. Then
f (0) ⫽ 1, f r (0) ⫽ 0, f s (t) ⫽ ⫺v2 cos vt. From this and (2) we obtain
l( f s ) ⫽ s 2l( f ) ⫺ s ⫽ ⫺v2l( f ).
By algebra,
l(cos vt) ⫽
s
s 2 ⫹ v2
.
Similarly, let g ⫽ sin vt. Then g(0) ⫽ 0, g r ⫽ v cos vt. From this and (1) we obtain
l(g r ) ⫽ sl(g) ⫽ vl(cos vt).
Hence,
l(sin vt) ⫽
v
v
.
l(cos vt) ⫽ 2
s
s ⫹ v2
䊏
Laplace Transform of the Integral of a Function
Differentiation and integration are inverse operations, and so are multiplication and division.
Since differentiation of a function f (t) (roughly) corresponds to multiplication of its transform
l( f ) by s, we expect integration of f (t) to correspond to division of l( f ) by s:
c06.qxd
10/28/10
6:33 PM
Page 213
SEC. 6.2 Transforms of Derivatives and Integrals. ODEs
THEOREM 3
213
Laplace Transform of Integral
Let F(s) denote the transform of a function f (t) which is piecewise continuous for t ⭌ 0
and satisfies a growth restriction (2), Sec. 6.1. Then, for s ⬎ 0, s ⬎ k, and t ⬎ 0,
le
(4)
PROOF
冮
t
0
1
f (t) dt f ⫽ s F(s),
t
冮 f (t) dt ⫽ l
thus
ⴚ1
0
1
e s F(s) f .
Denote the integral in (4) by g(t). Since f (t) is piecewise continuous, g(t) is continuous,
and (2), Sec. 6.1, gives
ƒ g(t) ƒ ⫽ `
冮
t
0
f (t) dt ` ⬉
冮
t
t
ƒ f (t) ƒ dt ⬉ M
0
冮e
kt
M kt
M
(e ⫺ 1) ⬉ ekt
k
k
dt ⫽
0
(k ⬎ 0).
This shows that g(t) also satisfies a growth restriction. Also, g r (t) ⫽ f (t), except at points
at which f (t) is discontinuous. Hence g r (t) is piecewise continuous on each finite interval
and, by Theorem 1, since g(0) ⫽ 0 (the integral from 0 to 0 is zero)
l{f (t)} ⫽ l{g r (t)} ⫽ sl{g(t)} ⫺ g(0) ⫽ sl{g(t)}.
Division by s and interchange of the left and right sides gives the first formula in (4),
from which the second follows by taking the inverse transform on both sides.
䊏
EXAMPLE 3
Application of Theorem 3: Formulas 19 and 20 in the Table of Sec. 6.9
Using Theorem 3, find the inverse of
Solution.
1
s(s 2 ⫹ v2)
and
1
s 2(s 2 ⫹ v2)
.
From Table 6.1 in Sec. 6.1 and the integration in (4) (second formula with the sides interchanged)
we obtain
lⴚ1 b
1
sin vt
,
r⫽
s 2 ⫹ v2
v
lⴚ1 b
1
r⫽
s(s 2 ⫹ v2)
冮
t
0
sin vt
1
dt ⫽ 2 (1 ⫺ cos vt).
v
v
This is formula 19 in Sec. 6.9. Integrating this result again and using (4) as before, we obtain formula 20
in Sec. 6.9:
lⴚ1 b
1
s 2(s 2 ⫹ v2)
r⫽
1
v2
冮 (1 ⫺ cos vt) dt ⫽ c v
t
0
t
2
⫺
sin vt
v3
d
t
⫽
0
t
v2
⫺
sin vt
v3
.
It is typical that results such as these can be found in several ways. In this example, try partial fraction
reduction.
䊏
Differential Equations, Initial Value Problems
Let us now discuss how the Laplace transform method solves ODEs and initial value
problems. We consider an initial value problem
(5)
y s ⫹ ay r ⫹ by ⫽ r(t),
y(0) ⫽ K 0,
y r (0) ⫽ K 1
c06.qxd
10/28/10
6:33 PM
214
Page 214
CHAP. 6 Laplace Transforms
where a and b are constant. Here r(t) is the given input (driving force) applied to the
mechanical or electrical system and y(t) is the output (response to the input) to be obtained.
In Laplace’s method we do three steps:
Step 1. Setting up the subsidiary equation. This is an algebraic equation for the transform
Y ⫽ l(y) obtained by transforming (5) by means of (1) and (2), namely,
3s 2Y ⫺ sy(0) ⫺ y r (0)4 ⫹ a3sY ⫺ y(0)4 ⫹ bY ⫽ R(s)
where R(s) ⫽ l(r). Collecting the Y-terms, we have the subsidiary equation
(s 2 ⫹ as ⫹ b)Y ⫽ (s ⫹ a)y(0) ⫹ y r (0) ⫹ R(s).
Step 2. Solution of the subsidiary equation by algebra. We divide by s 2 ⫹ as ⫹ b and
use the so-called transfer function
(6)
Q(s) ⫽
1
s 2 ⫹ as ⫹ b
⫽
1
(s ⫹ 12 a)2 ⫹ b ⫺ 14 a 2
.
(Q is often denoted by H, but we need H much more frequently for other purposes.) This
gives the solution
(7)
Y(s) ⫽ 3(s ⫹ a)y(0) ⫹ y r (0)4Q(s) ⫹ R(s)Q(s).
If y(0) ⫽ y r (0) ⫽ 0, this is simply Y ⫽ RQ; hence
Q⫽
l(output)
Y
⫽
R
l(input)
and this explains the name of Q. Note that Q depends neither on r(t) nor on the initial
conditions (but only on a and b).
Step 3. Inversion of Y to obtain y ⴝ lⴚ1(Y ). We reduce (7) (usually by partial fractions
as in calculus) to a sum of terms whose inverses can be found from the tables (e.g., in
Sec. 6.1 or Sec. 6.9) or by a CAS, so that we obtain the solution y(t) ⫽ lⴚ1(Y ) of (5).
EXAMPLE 4
Initial Value Problem: The Basic Laplace Steps
Solve
y s ⫺ y ⫽ t,
Solution.
y(0) ⫽ 1,
y r (0) ⫽ 1.
Step 1. From (2) and Table 6.1 we get the subsidiary equation 3with Y ⫽ l(y)4
s 2Y ⫺ sy(0) ⫺ y r (0) ⫺ Y ⫽ 1>s 2,
thus
(s 2 ⫺ 1)Y ⫽ s ⫹ 1 ⫹ 1>s 2.
Step 2. The transfer function is Q ⫽ 1>(s 2 ⫺ 1), and (7) becomes
Y ⫽ (s ⫹ 1)Q ⫹
1
s
2
Q⫽
s⫹1
s ⫺1
2
⫹
1
s (s ⫺ 1)
2
2
Simplification of the first fraction and an expansion of the last fraction gives
Y⫽
1
1
1
⫹
⫺ 2b.
s ⫺ 1 a s2 ⫺ 1
s
.
c06.qxd
10/28/10
6:33 PM
Page 215
SEC. 6.2 Transforms of Derivatives and Integrals. ODEs
215
Step 3. From this expression for Y and Table 6.1 we obtain the solution
y(t) ⫽ lⴚ1(Y ) ⫽ lⴚ1 e
1
1
1
⫹ lⴚ1 e 2
⫺ lⴚ1 e 2 f ⫽ et ⫹ sinh t ⫺ t.
s ⫺ 1f
s ⫺ 1f
s
䊏
The diagram in Fig. 116 summarizes our approach.
t-space
s-space
Given problem
y" – y = t
y(0) = 1
y'(0) =1
(s2 – 1)Y = s + 1 + 1/s2
Solution of given problem
Solution of subsidiary equation
Subsidiary equation
y(t) = et + sinh t – t
Y=
1
1 – 1
+
s – 1 s2 – 1 s2
Fig. 116. Steps of the Laplace transform method
EXAMPLE 5
Comparison with the Usual Method
Solve the initial value problem
y s ⫹ y r ⫹ 9y ⫽ 0.
Solution.
y(0) ⫽ 0.16,
y r (0) ⫽ 0.
From (1) and (2) we see that the subsidiary equation is
s 2Y ⫺ 0.16s ⫹ sY ⫺ 0.16 ⫹ 9Y ⫽ 0,
thus
(s 2 ⫹ s ⫹ 9)Y ⫽ 0.16(s ⫹ 1).
The solution is
Y⫽
0.16(s ⫹ 1)
s2 ⫹ s ⫹ 9
⫽
0.16(s ⫹ 12 ) ⫹ 0.08
(s ⫹ 12 )2 ⫹ 35
4
.
Hence by the first shifting theorem and the formulas for cos and sin in Table 6.1 we obtain
y(t) ⫽ lⴚ1(Y ) ⫽ eⴚt>2 a0.16 cos
35
0.08
35
t⫹1
sin
tb
B4
B4
35
22
⫽ eⴚ0.5t(0.16 cos 2.96t ⫹ 0.027 sin 2.96t).
This agrees with Example 2, Case (III) in Sec. 2.4. The work was less.
Advantages of the Laplace Method
1. Solving a nonhomogeneous ODE does not require first solving the
homogeneous ODE. See Example 4.
2. Initial values are automatically taken care of. See Examples 4 and 5.
3. Complicated inputs r(t) (right sides of linear ODEs) can be handled very
efficiently, as we show in the next sections.
䊏
c06.qxd
10/28/10
6:33 PM
216
Page 216
CHAP. 6 Laplace Transforms
EXAMPLE 6
Shifted Data Problems
This means initial value problems with initial conditions given at some t ⫽ t 0 ⬎ 0 instead of t ⫽ 0. For such a
~
~
problem set t ⫽ t ⫹ t 0, so that t ⫽ t 0 gives t ⫽ 0 and the Laplace transform can be applied. For instance, solve
y(14 p) ⫽ 12 p,
y s ⫹ y ⫽ 2t,
Solution.
y r (14 p) ⫽ 2 ⫺ 12.
~
We have t 0 ⫽ 14 p and we set t ⫽ t ⫹ 14 p. Then the problem is
~y s ⫹ ~y ⫽ 2(~t ⫹ 1 p),
4
~y (0) ⫽ 1 p,
2
~y r (0) ⫽ 2 ⫺ 12
~
~
where ~y ( t ) ⫽ y(t). Using (2) and Table 6.1 and denoting the transform of ~y by Y , we see that the subsidiary
equation of the “shifted” initial value problem is
1
2
2p
~
~
s 2Y ⫺ s # 12 p ⫺ (2 ⫺ 12) ⫹ Y ⫽ 2 ⫹
,
s
s
1
2
2p
1
~
(s 2 ⫹ 1)Y ⫽ 2 ⫹
⫹ ps ⫹ 2 ⫺ 12.
s
s
2
thus
~
Solving this algebraically for Y , we obtain
~
Y⫽
2
(s ⫹ 1)s
2
2
⫹
1
2
p
⫹
(s ⫹ 1)s
2
1
2
ps
s ⫹1
2
⫹
2 ⫺ 12
s2 ⫹ 1
.
The inverse of the first two terms can be seen from Example 3 (with v ⫽ 1), and the last two terms give cos
and sin,
~
~
~
~
~
~
~
y ⫽ lⴚ1( Y ) ⫽ 2( t ⫺ sin t ) ⫹ 12 p(1 ⫺ cos t ) ⫹ 12 p cos t ⫹ (2 ⫺ 12) sin t
~
~
⫽ 2t ⫹ 12 p ⫺ 12 sin t .
1
~
~
Now t ⫽ t ⫺ 14 p, sin t ⫽
(sin t ⫺ cos t), so that the answer (the solution) is
12
䊏
y ⫽ 2t ⫺ sin t ⫹ cos t.
PROBLEM SET 6.2
1–11
INITIAL VALUE PROBLEMS (IVPS)
Solve the IVPs by the Laplace transform. If necessary, use
partial fraction expansion as in Example 4 of the text. Show
all details.
1. y r ⫹ 5.2y ⫽ 19.4 sin 2t, y(0) ⫽ 0
2. y r ⫹ 2y ⫽ 0, y(0) ⫽ 1.5
3. y s ⫺ y r ⫺ 6y ⫽ 0, y(0) ⫽ 11, y r (0) ⫽ 28
4. y s ⫹ 9y ⫽ 10eⴚt, y(0) ⫽ 0, y r (0) ⫽ 0
5. y s ⫺ 14 y ⫽ 0, y(0) ⫽ 12, y r (0) ⫽ 0
6. y s ⫺ 6y r ⫹ 5y ⫽ 29 cos 2t, y(0) ⫽ 3.2,
y r (0) ⫽ 6.2
7. y s ⫹ 7y r ⫹ 12y ⫽ 21e3t, y(0) ⫽ 3.5,
y r (0) ⫽ ⫺10
8. y s ⫺ 4y r ⫹ 4y ⫽ 0, y(0) ⫽ 8.1, y r (0) ⫽ 3.9
9. y s ⫺ 4y r ⫹ 3y ⫽ 6t ⫺ 8, y(0) ⫽ 0, y r (0) ⫽ 0
10. y s ⫹ 0.04y ⫽ 0.02t 2, y(0) ⫽ ⫺25, y r (0) ⫽ 0
11. y s ⫹ 3y r ⫹ 2.25y ⫽ 9t 3 ⫹ 64, y(0) ⫽ 1,
y r (0) ⫽ 31.5
12–15
SHIFTED DATA PROBLEMS
Solve the shifted data IVPs by the Laplace transform. Show
the details.
12. y s ⫺ 2y r ⫺ 3y ⫽ 0, y(4) ⫽ ⫺3,
y r (4) ⫽ ⫺17
13. y r ⫺ 6y ⫽ 0, y(⫺1) ⫽ 4
14. y s ⫹ 2y r ⫹ 5y ⫽ 50t ⫺ 100, y(2) ⫽ ⫺4,
y r (2) ⫽ 14
15. y s ⫹ 3y r ⫺ 4y ⫽ 6e2tⴚ3,
y r (1.5) ⫽ 5
16–21
y(1.5) ⫽ 4,
OBTAINING TRANSFORMS
BY DIFFERENTIATION
Using (1) or (2), find l( f ) if f (t) equals:
16. t cos 4t
17. teⴚat
18. cos2 2t
19. sin2 vt
20. sin4 t. Use Prob. 19.
21. cosh2 t
c06.qxd
10/28/10
6:33 PM
Page 217
SEC. 6.3 Unit Step Function (Heaviside Function). Second Shifting Theorem (t-Shifting)
22. PROJECT. Further Results by Differentiation.
Proceeding as in Example 1, obtain
(a)
l(t cos vt) ⫽
s 2 ⫺ v2
(s 2 ⫹ v2)2
and from this and Example 1: (b) formula 21, (c) 22,
(d) 23 in Sec. 6.9,
(e) l(t cosh at) ⫽
(f ) l(t sinh at) ⫽
23–29
s2 ⫹ a2
(s 2 ⫺ a 2)2
,
2as
.
(s 2 ⫺ a 2)2
INVERSE TRANSFORMS
BY INTEGRATION
Using Theorem 3, find f (t) if l(F ) equals:
20
3
23. 2
24. 3
s ⫹ s>4
s ⫺ 2ps 2
1
1
25.
26. 4
s(s 2 ⫹ v2)
s ⫺ s2
s⫹1
3s ⫹ 4
27. 4
28. 4
s ⫹ 9s 2
s ⫹ k 2s 2
1
29. 3
s ⫹ as 2
6.3
217
30. PROJECT. Comments on Sec. 6.2. (a) Give reasons
why Theorems 1 and 2 are more important than
Theorem 3.
(b) Extend Theorem 1 by showing that if f (t) is
continuous, except for an ordinary discontinuity (finite
jump) at some t ⫽ a (⬎0), the other conditions remaining
as in Theorem 1, then (see Fig. 117)
(1*) l( f r ) ⫽ sl( f ) ⫺ f (0) ⫺ 3 f (a ⫹ 0) ⫺ f (a ⫺ 0)4eⴚas.
(c) Verify (1*) for f (t) ⫽ eⴚt if 0 ⬍ t ⬍ 1 and 0 if
t ⬎ 1.
(d) Compare the Laplace transform of solving ODEs
with the method in Chap. 2. Give examples of your
own to illustrate the advantages of the present method
(to the extent we have seen them so far).
f (t)
f (a – 0)
f (a + 0)
0
a
t
Fig. 117. Formula (1*)
Unit Step Function (Heaviside Function).
Second Shifting Theorem (t-Shifting)
This section and the next one are extremely important because we shall now reach the
point where the Laplace transform method shows its real power in applications and its
superiority over the classical approach of Chap. 2. The reason is that we shall introduce
two auxiliary functions, the unit step function or Heaviside function u(t ⫺ a) (below) and
Dirac’s delta d(t ⫺ a) (in Sec. 6.4). These functions are suitable for solving ODEs with
complicated right sides of considerable engineering interest, such as single waves, inputs
(driving forces) that are discontinuous or act for some time only, periodic inputs more
general than just cosine and sine, or impulsive forces acting for an instant (hammerblows,
for example).
Unit Step Function (Heaviside Function) u(t ⫺ a)
The unit step function or Heaviside function u(t ⫺ a) is 0 for t ⬍ a, has a jump of size
1 at t ⫽ a (where we can leave it undefined), and is 1 for t ⬎ a, in a formula:
(1)
u(t ⫺ a) ⫽ b
0
if t ⬍ a
1
if t ⬎ a
(a ⭌ 0).
c06.qxd
10/28/10
218
6:33 PM
Page 218
CHAP. 6 Laplace Transforms
u(t – a)
u(t)
1
1
0
t
0
a
t
Fig. 119. Unit step function u(t ⫺ a)
Fig. 118. Unit step function u(t)
Figure 118 shows the special case u(t), which has its jump at zero, and Fig. 119 the general
case u(t ⫺ a) for an arbitrary positive a. (For Heaviside, see Sec. 6.1.)
The transform of u(t ⫺ a) follows directly from the defining integral in Sec. 6.1,
l{u(t ⫺ a)} ⫽
冮
ⴥ
e
ⴚst
u(t ⫺ a) dt ⫽
0
冮
ⴥ
e
ⴚst
0
ⴚst ⴥ
# 1 dt ⫽ ⫺ e `
s
;
t⫽a
here the integration begins at t ⫽ a (⭌ 0) because u(t ⫺ a) is 0 for t ⬍ a. Hence
l{u(t ⫺ a)} ⫽
(2)
eⴚas
s
(s ⬎ 0).
The unit step function is a typical “engineering function” made to measure for engineering
applications, which often involve functions (mechanical or electrical driving forces) that
are either “off ” or “on.” Multiplying functions f (t) with u(t ⫺ a), we can produce all sorts
of effects. The simple basic idea is illustrated in Figs. 120 and 121. In Fig. 120 the given
function is shown in (A). In (B) it is switched off between t ⫽ 0 and t ⫽ 2 (because
u(t ⫺ 2) ⫽ 0 when t ⬍ 2) and is switched on beginning at t ⫽ 2. In (C) it is shifted to the
right by 2 units, say, for instance, by 2 sec, so that it begins 2 sec later in the same fashion
as before. More generally we have the following.
Let f (t) ⫽ 0 for all negative t. Then f (t ⫺ a)u(t ⫺ a) with a ⬎ 0 is f (t) shifted
(translated) to the right by the amount a.
Figure 121 shows the effect of many unit step functions, three of them in (A) and
infinitely many in (B) when continued periodically to the right; this is the effect of a
rectifier that clips off the negative half-waves of a sinuosidal voltage. CAUTION! Make
sure that you fully understand these figures, in particular the difference between parts (B)
and (C) of Fig. 120. Figure 120(C) will be applied next.
f (t)
5
0
5
π 2π
t
0
5
2 π 2π
–5
–5
(A) f (t) = 5 sin t
(B) f (t)u(t – 2)
t
0
2 π +2 2π +2
t
–5
(C) f (t – 2)u(t – 2)
Fig. 120. Effects of the unit step function: (A) Given function.
(B) Switching off and on. (C) Shift.
c06.qxd
10/28/10
6:33 PM
Page 219
SEC. 6.3 Unit Step Function (Heaviside Function). Second Shifting Theorem (t-Shifting)
219
4
k
1
4
t
6
–k
0
2
4
6
8
10
t
(B) 4 sin (12_ π t)[u(t) – u(t – 2) + u(t – 4) – + ⋅⋅⋅]
(A) k[u(t – 1) – 2u(t – 4) + u(t – 6)]
Fig. 121. Use of many unit step functions.
Time Shifting (t-Shifting): Replacing t by t ⫺ a in f (t)
The first shifting theorem (“s-shifting”) in Sec. 6.1 concerned transforms F(s) ⫽ l{f (t)}
and F(s ⫺ a) ⫽ l{eatf (t)}. The second shifting theorem will concern functions f (t) and
f (t ⫺ a). Unit step functions are just tools, and the theorem will be needed to apply them
in connection with any other functions.
THEOREM 1
Second Shifting Theorem; Time Shifting
If f (t) has the transform F(s), then the “shifted function”
(3)
~
f (t) ⫽ f (t ⫺ a)u(t ⫺ a) ⫽ b
0
if t ⬍ a
f (t ⫺ a)
if t ⬎ a
has the transform eⴚasF(s). That is, if l{f (t)} ⫽ F(s), then
(4)
l{f (t ⫺ a)u(t ⫺ a)} ⫽ eⴚasF(s).
Or, if we take the inverse on both sides, we can write
(4*)
f (t ⫺ a)u(t ⫺ a) ⫽ lⴚ1{eⴚasF(s)}.
Practically speaking, if we know F(s), we can obtain the transform of (3) by multiplying
F(s) by eⴚas. In Fig. 120, the transform of 5 sin t is F(s) ⫽ 5>(s 2 ⫹ 1), hence the shifted
function 5 sin (t ⫺ 2)u(t ⫺ 2) shown in Fig. 120(C) has the transform
eⴚ2sF(s) ⫽ 5eⴚ2s>(s 2 ⫹ 1).
PROOF
We prove Theorem 1. In (4), on the right, we use the definition of the Laplace transform,
writing t for t (to have t available later). Then, taking eⴚas inside the integral, we have
eⴚasF(s) ⫽ eⴚas
冮
ⴥ
eⴚstf (t) dt ⫽
0
冮
ⴥ
eⴚs(t⫹a)f (t) dt.
0
Substituting t ⫹ a ⫽ t, thus t ⫽ t ⫺ a, dt ⫽ dt in the integral (CAUTION, the lower
limit changes!), we obtain
eⴚasF(s) ⫽
冮
ⴥ
a
eⴚstf (t ⫺ a) dt.
c06.qxd
10/28/10
6:33 PM
220
Page 220
CHAP. 6 Laplace Transforms
To make the right side into a Laplace transform, we must have an integral from 0 to ⬁ ,
not from a to ⬁ . But this is easy. We multiply the integrand by u(t ⫺ a). Then for t from
~
0 to a the integrand is 0, and we can write, with f as in (3),
eⴚasF(s) ⫽
冮
ⴥ
eⴚstf (t ⫺ a)u(t ⫺ a) dt ⫽
0
冮
ⴥ
~
eⴚstf (t) dt.
0
(Do you now see why u(t ⫺ a) appears?) This integral is the left side of (4), the Laplace
~
䊏
transform of f (t) in (3). This completes the proof.
EXAMPLE 1
Application of Theorem 1. Use of Unit Step Functions
Write the following function using unit step functions and find its transform.
if 0 ⬍ t ⬍ 1
2
f (t) ⫽
d 12 t 2
if 1 ⬍ t ⬍ 12 p
cos t
Solution.
(Fig. 122)
1
2
t ⬎ p.
if
Step 1. In terms of unit step functions,
f (t) ⫽ 2(1 ⫺ u(t ⫺ 1)) ⫹ 12 t 2(u(t ⫺ 1) ⫺ u(t ⫺ 12 p)) ⫹ (cos t)u(t ⫺ 12 p).
Indeed, 2(1 ⫺ u(t ⫺ 1)) gives f (t) for 0 ⬍ t ⬍ 1, and so on.
Step 2. To apply Theorem 1, we must write each term in f (t) in the form f (t ⫺ a)u(t ⫺ a). Thus, 2(1 ⫺ u(t ⫺ 1))
remains as it is and gives the transform 2(1 ⫺ eⴚs)>s. Then
1
1
1
1
1
1
l e t 2u(t ⫺ 1) f ⫽ l a (t ⫺ 1)2 ⫹ (t ⫺ 1) ⫹ b u(t ⫺ 1) f ⫽ a 3 ⫹ 2 ⫹ b eⴚs
2
2
2
2s
s
s
2
1
1
1
1
p
1
p2
1
l e t 2u at ⫺ p b f ⫽ l e at ⫺ p b ⫹ at ⫺ p b ⫹
b u at ⫺ p b f
2
2
2
2
2
2
8
2
⫽a
l e (cos t) u at ⫺
1
2
1
p
p2 ⴚps>2
be
3 ⫹
2 ⫹
8s
s
2s
p b f ⫽ l e ⫺asin at ⫺
1
2
p bb u at ⫺
1
2
pb f ⫽ ⫺
1
eⴚps>2.
s2 ⫹ 1
Together,
l( f ) ⫽
2
2
1
1
1
1
p
p2 ⴚps>2
1
⫺ eⴚs ⫹ a 3 ⫹ 2 ⫹ b eⴚs ⫺ a 3 ⫹ 2 ⫹
⫺ 2
eⴚps>2.
be
s
s
2s
8s
s
s
s
2s
s ⫹1
If the conversion of f (t) to f (t ⫺ a) is inconvenient, replace it by
l{ f (t)u(t ⫺ a)} ⫽ eⴚasl{ f (t ⫹ a)}.
(4**)
(4**) follows from (4) by writing f (t ⫺ a) ⫽ g(t), hence f (t) ⫽ g(t ⫹ a) and then again writing f for g. Thus,
1
1
1
1
1
1
1
l e t 2u(t ⫺ 1) f ⫽ eⴚsl e (t ⫹ 1)2 f ⫽ eⴚsl e t 2 ⫹ t ⫹ f ⫽ eⴚs a 3 ⫹ 2 ⫹ b
2
2
2
2
2s
s
s
as before. Similarly for l{ 12 t 2u(t ⫺ 12 p)}. Finally, by (4**),
l e cos t u at ⫺
1
1
1
p b f ⫽ eⴚps>2l e cos at ⫹ p b f ⫽ eⴚps>2l{⫺sin t} ⫽ ⫺eⴚps>2 2
.
2
2
s ⫹1
䊏
c06.qxd
10/28/10
6:33 PM
Page 221
SEC. 6.3 Unit Step Function (Heaviside Function). Second Shifting Theorem (t-Shifting)
221
f (t)
2
1
0
␲
1
2␲
t
4␲
–1
Fig. 122. ƒ(t) in Example 1
EXAMPLE 2
Application of Both Shifting Theorems. Inverse Transform
Find the inverse transform f (t) of
F(s) ⫽
eⴚs
s 2 ⫹ p2
⫹
eⴚ2s
s 2 ⫹ p2
⫹
eⴚ3s
(s ⫹ 2)2
.
Solution.
Without the exponential functions in the numerator the three terms of F(s) would have the inverses
(sin pt)> p, (sin pt)> p, and teⴚ2t because 1>s 2 has the inverse t, so that 1>(s ⫹ 2)2 has the inverse teⴚ2t by the
first shifting theorem in Sec. 6.1. Hence by the second shifting theorem (t-shifting),
f (t) ⫽
1
1
p sin (p(t ⫺ 1)) u(t ⫺ 1) ⫹ p sin (p(t ⫺ 2)) u(t ⫺ 2) ⫹ (t ⫺ 3)e
ⴚ2(t⫺3)
u(t ⫺ 3).
Now sin (pt ⫺ p) ⫽ ⫺sin pt and sin (pt ⫺ 2p) ⫽ sin pt, so that the first and second terms cancel each other
when t ⬎ 2. Hence we obtain f (t) ⫽ 0 if 0 ⬍ t ⬍ 1, ⫺(sin pt)> p if 1 ⬍ t ⬍ 2, 0 if 2 ⬍ t ⬍ 3, and
(t ⫺ 3)eⴚ2(tⴚ3) if t ⬎ 3. See Fig. 123.
䊏
0.3
0.2
0.1
0
0
1
2
3
4
5
t
6
Fig. 123. ƒ(t) in Example 2
EXAMPLE 3
Response of an RC-Circuit to a Single Rectangular Wave
Find the current i(t) in the RC-circuit in Fig. 124 if a single rectangular wave with voltage V0 is applied. The
circuit is assumed to be quiescent before the wave is applied.
The input is V03u(t ⫺ a) ⫺ u(t ⫺ b)4. Hence the circuit is modeled by the integro-differential
equation (see Sec. 2.9 and Fig. 124)
Solution.
Ri(t) ⫹
C
v(t)
R
q(t)
C
⫽ Ri(t) ⫹
1
C
t
冮 i(t) dt ⫽ v(t) ⫽ V 3u(t ⫺ a) ⫺ u(t ⫺ b)4.
0
0
v(t)
i(t)
V0
V0/R
0
a
b
t
0
a
b
Fig. 124. RC-circuit, electromotive force v(t), and current in Example 3
t
c06.qxd
10/28/10
6:33 PM
222
Page 222
CHAP. 6 Laplace Transforms
Using Theorem 3 in Sec. 6.2 and formula (1) in this section, we obtain the subsidiary equation
RI(s) ⫹
I(s)
sC
⫽
V0
s
3eⴚas ⫺ eⴚbs4.
Solving this equation algebraically for I(s), we get
I(s) ⫽ F(s)(eⴚas ⫺ eⴚbs)
where
F(s) ⫽
V0IR
s ⫹ 1>(RC)
lⴚ1(F) ⫽
and
V0
R
eⴚt>(RC),
the last expression being obtained from Table 6.1 in Sec. 6.1. Hence Theorem 1 yields the solution (Fig. 124)
i(t) ⫽ lⴚ1(I) ⫽ lⴚ1{eⴚasF(s)} ⫺ lⴚ1{eⴚbsF(s)} ⫽
V0
R
3eⴚ(tⴚa)>(RC)u(t ⫺ a) ⫺ eⴚ(tⴚb)>(RC)u(t ⫺ b)4;
that is, i(t) ⫽ 0 if t ⬍ a, and
i(t) ⫽ c
K 1eⴚt>(RC)
if a ⬍ t ⬍ b
(K 1 ⫺ K 2)e
ⴚt>(RC)
if a ⬎ b
where K 1 ⫽ V0ea>(RC)>R and K 2 ⫽ V0eb>(RC)>R.
EXAMPLE 4
䊏
Response of an RLC-Circuit to a Sinusoidal Input Acting Over a Time Interval
Find the response (the current) of the RLC-circuit in Fig. 125, where E(t) is sinusoidal, acting for a short time
interval only, say,
E(t) ⫽ 100 sin 400t if 0 ⬍ t ⬍ 2p
and
E(t) ⫽ 0 if t ⬎ 2p
and current and charge are initially zero.
The electromotive force E(t) can be represented by (100 sin 400t)(1 ⫺ u(t ⫺ 2p)). Hence the
model for the current i(t) in the circuit is the integro-differential equation (see Sec. 2.9)
Solution.
t
0.1i r ⫹ 11i ⫹ 100
冮 i(t) dt ⫽ (100 sin 400t)(1 ⫺ u(t ⫺ 2p)).
i(0) ⫽ 0,
i r (0) ⫽ 0.
0
From Theorems 2 and 3 in Sec. 6.2 we obtain the subsidiary equation for I(s) ⫽ l(i)
0.1sI ⫹ 11I ⫹ 100
100 # 400s 1
eⴚ2ps
I
⫽ 2
a ⫺
b.
s
s
s ⫹ 4002 s
Solving it algebraically and noting that s 2 ⫹ 110s ⫹ 1000 ⫽ (s ⫹ 10)(s ⫹ 100), we obtain
l(s) ⫽
s
seⴚ2ps
1000 # 400
⫺ 2
a
b.
(s ⫹ 10)(s ⫹ 100) s 2 ⫹ 4002
s ⫹ 4002
For the first term in the parentheses ( Á ) times the factor in front of them we use the partial fraction
expansion
400,000s
(s ⫹ 10)(s ⫹ 100)(s 2 ⫹ 4002)
⫽
B
Ds ⫹ K
A
⫹
⫹ 2
.
s ⫹ 10
s ⫹ 100
s ⫹ 4002
Now determine A, B, D, K by your favorite method or by a CAS or as follows. Multiplication by the common
denominator gives
400,000s ⫽ A(s ⫹ 100)(s 2 ⫹ 4002) ⫹ B(s ⫹ 10)(s 2 ⫹ 4002) ⫹ (Ds ⫹ K)(s ⫹ 10)(s ⫹ 100).
c06.qxd
10/28/10
6:33 PM
Page 223
SEC. 6.3 Unit Step Function (Heaviside Function). Second Shifting Theorem (t-Shifting)
223
We set s ⫽ ⫺10 and ⫺100 and then equate the sums of the s 3 and s 2 terms to zero, obtaining (all values rounded)
(s ⫽ ⫺10)
⫺4,000,000 ⫽ 90(102 ⫹ 4002)A,
(s ⫽ ⫺100)
A ⫽ ⫺0.27760
⫺40,000,000 ⫽ ⫺90(1002 ⫹ 4002)B,
B ⫽ 2.6144
(s 3-terms)
0 ⫽ A ⫹ B ⫹ D,
D ⫽ ⫺2.3368
(s 2-terms)
0 ⫽ 100A ⫹ 10B ⫹ 110D ⫹ K,
K ⫽ 258.66.
Since K ⫽ 258.66 ⫽ 0.6467 # 400, we thus obtain for the first term I1 in I ⫽ I1 ⫺ I2
I1 ⫽ ⫺
0.2776
2.6144
2.3368s
0.6467 # 400
.
⫹
⫺ 2
2 ⫹
s ⫹ 10
s ⫹ 100
s ⫹ 400
s 2 ⫹ 4002
From Table 6.1 in Sec. 6.1 we see that its inverse is
i 1(t) ⫽ ⫺0.2776eⴚ10t ⫹ 2.6144eⴚ100t ⫺ 2.3368 cos 400t ⫹ 0.6467 sin 400t.
This is the current i(t) when 0 ⬍ t ⬍ 2p. It agrees for 0 ⬍ t ⬍ 2p with that in Example 1 of Sec. 2.9 (except
for notation), which concerned the same RLC-circuit. Its graph in Fig. 63 in Sec. 2.9 shows that the exponential
terms decrease very rapidly. Note that the present amount of work was substantially less.
The second term I1 of I differs from the first term by the factor eⴚ2ps. Since cos 400(t ⫺ 2p) ⫽ cos 400t
and sin 400(t ⫺ 2p) ⫽ sin 400t, the second shifting theorem (Theorem 1) gives the inverse i 2(t) ⫽ 0 if
0 ⬍ t ⬍ 2p, and for ⬎ 2p it gives
i 2(t) ⫽ ⫺0.2776eⴚ10(tⴚ2p) ⫹ 2.6144eⴚ100(tⴚ2p) ⫺ 2.3368 cos 400t ⫹ 0.6467 sin 400t.
Hence in i(t) the cosine and sine terms cancel, and the current for t ⬎ 2p is
i(t) ⫽ ⫺0.2776(eⴚ10t ⫺ eⴚ10(tⴚ2p)) ⫹ 2.6144(eⴚ100t ⫺ eⴚ100(tⴚ2p)).
䊏
It goes to zero very rapidly, practically within 0.5 sec.
C = 10 –2 F
R = 11 Ω
L = 0.1 H
E(t)
Fig. 125. RLC-circuit in Example 4
PROBLEM SET 6.3
1. Report on Shifting Theorems. Explain and compare
the different roles of the two shifting theorems, using your
own formulations and simple examples. Give no proofs.
2–11
SECOND SHIFTING THEOREM,
UNIT STEP FUNCTION
Sketch or graph the given function, which is assumed to be
zero outside the given interval. Represent it, using unit step
functions. Find its transform. Show the details of your work.
2. t (0 ⬍ t ⬍ 2)
4. cos 4t (0 ⬍ t ⬍ p)
3. t ⫺ 2 (t ⬎ 2)
5. et (0 ⬍ t ⬍ p>2)
6. sin pt (2 ⬍ t ⬍ 4)
8. t 2 (1 ⬍ t ⬍ 2)
10. sinh t (0 ⬍ t ⬍ 2)
12–17
7. eⴚpt (2 ⬍ t ⬍ 4)
9. t 2 (t ⬎ 32)
11. sin t (p>2 ⬍ t ⬍ p)
INVERSE TRANSFORMS BY THE
2ND SHIFTING THEOREM
Find and sketch or graph f (t) if l( f ) equals
12. eⴚ3s>(s ⫺ 1) 3
13. 6(1 ⫺ eⴚps)>(s 2 ⫹ 9)
ⴚ2s
ⴚ5s
14. 4(e
15. eⴚ3s>s 4
⫺ 2e )>s
ⴚs
ⴚ3s
2
16. 2(e ⫺ e )>(s ⫺ 4)
17. (1 ⫹ eⴚ2p(s⫹1))(s ⫹ 1)>((s ⫹ 1) 2 ⫹ 1)
c06.qxd
10/28/10
6:33 PM
224
18–27
Page 224
CHAP. 6 Laplace Transforms
IVPs, SOME WITH DISCONTINUOUS
INPUT
Using the Laplace transform and showing the details, solve
18. 9y s ⫺ 6y r ⫹ y ⫽ 0, y(0) ⫽ 3, y r (0) ⫽ 1
19. y s ⫹ 6y r ⫹ 8y ⫽ eⴚ3t ⫺ eⴚ5t, y(0) ⫽ 0, y r (0) ⫽ 0
20. y s ⫹ 10y r ⫹ 24y ⫽ 144t 2, y(0) ⫽ 19>12,
y r (0) ⫽ ⫺5
21. y s ⫹ 9y ⫽ 8 sin t if 0 ⬍ t ⬍ p and 0 if t ⬎ p;
y(0) ⫽ 0, y r (0) ⫽ 4
22. y s ⫹ 3y r ⫹ 2y ⫽ 4t if 0 ⬍ t ⬍ 1 and 8 if t ⬎ 1;
y(0) ⫽ 0, y r (0) ⫽ 0
23. y s ⫹ y r ⫺ 2y ⫽ 3 sin t ⫺ cos t if 0 ⬍ t ⬍ 2p and
3 sin 2t ⫺ cos 2t if t ⬎ 2p; y(0) ⫽ 1, y r (0) ⫽ 0
24. y s ⫹ 3y r ⫹ 2y ⫽ 1 if 0 ⬍ t ⬍ 1 and 0 if t ⬎ 1;
y(0) ⫽ 0, y r (0) ⫽ 0
25. y s ⫹ y ⫽ t if 0 ⬍ t ⬍ 1 and 0 if t ⬎ 1; y(0) ⫽ 0,
y r (0) ⫽ 0
26. Shifted data. y s ⫹ 2y r ⫹ 5y ⫽ 10 sin t if 0 ⬍ t ⬍ 2p
and 0 if t ⬎ 2p; y(p) ⫽ 1, y r (p) ⫽ 2eⴚp ⫺ 2
27. Shifted data. y s ⫹ 4y ⫽ 8t 2 if 0 ⬍ t ⬍ 5 and 0 if
t ⬎ 5; y(1) ⫽ 1 ⫹ cos 2, y r (1) ⫽ 4 ⫺ 2 sin 2
28–40
MODELS OF ELECTRIC CIRCUITS
28–30
RL-CIRCUIT
31. Discharge in RC-circuit. Using the Laplace transform,
find the charge q(t) on the capacitor of capacitance C
in Fig. 127 if the capacitor is charged so that its potential
is V0 and the switch is closed at t ⫽ 0.
32–34
Using the Laplace transform and showing the details, find
the current i(t) in the circuit in Fig. 128 with R ⫽ 10 ⍀ and
C ⫽ 10ⴚ2 F, where the current at t ⫽ 0 is assumed to be
zero, and:
32. v ⫽ 0 if t ⬍ 4 and 14 # 106eⴚ3t V if t ⬎ 4
33. v ⫽ 0 if t ⬍ 2 and 100(t ⫺ 2) V if t ⬎ 2
34. v(t) ⫽ 100 V if 0.5 ⬍ t ⬍ 0.6 and 0 otherwise. Why
does i(t) have jumps?
C
R
R
v(t)
Fig. 128. Problems 32–34
35–37
Using the Laplace transform and showing the details, find
the current i(t) in the circuit in Fig. 126, assuming i(0) ⫽ 0
and:
28. R ⫽ 1 k⍀ (⫽1000 ⍀), L ⫽ 1 H, v ⫽ 0 if 0 ⬍ t ⬍ p,
and 40 sin t V if t ⬎ p
29. R ⫽ 25 ⍀, L ⫽ 0.1 H, v ⫽ 490 eⴚ5t V if 0 ⬍ t ⬍ 1
and 0 if t ⬎ 1
30. R ⫽ 10 ⍀, L ⫽ 0.5 H, v ⫽ 200t V if 0 ⬍ t ⬍ 2 and 0
if t ⬎ 2
RC-CIRCUIT
LC-CIRCUIT
Using the Laplace transform and showing the details, find
the current i(t) in the circuit in Fig. 129, assuming zero
initial current and charge on the capacitor and:
35. L ⫽ 1 H, C ⫽ 10ⴚ2 F, v ⫽ ⫺9900 cos t V if
p ⬍ t ⬍ 3p and 0 otherwise
36. L ⫽ 1 H, C ⫽ 0.25 F, v ⫽ 200 (t ⫺ 13 t 3) V if
0 ⬍ t ⬍ 1 and 0 if t ⬎ 1
37. L ⫽ 0.5 H, C ⫽ 0.05 F, v ⫽ 78 sin t V if 0 ⬍ t ⬍ p
and 0 if t ⬎ p
L
C
L
v(t)
v(t)
Fig. 126. Problems 28–30
Fig. 129. Problems 35–37
38–40
C
R
Fig. 127. Problem 31
RLC-CIRCUIT
Using the Laplace transform and showing the details, find
the current i(t) in the circuit in Fig. 130, assuming zero
initial current and charge and:
38. R ⫽ 4 ⍀, L ⫽ 1 H, C ⫽ 0.05 F, v ⫽ 34eⴚt V if
0 ⬍ t ⬍ 4 and 0 if t ⬎ 4
c06.qxd
10/28/10
6:33 PM
Page 225
SEC. 6.4 Short Impulses. Dirac’s Delta Function. Partial Fractions
39. R ⫽ 2 ⍀, L ⫽ 1 H, C ⫽ 0.5 F, v(t) ⫽ 1 kV if
0 ⬍ t ⬍ 2 and 0 if t ⬎ 2
225
40. R ⫽ 2 ⍀, L ⫽ 1 H, C ⫽ 0.1 F, v ⫽ 255 sin t V
if 0 ⬍ t ⬍ 2p and 0 if t ⬎ 2p
30
C
20
10
R
0
L
2
4
6
8
10
12
t
–10
–20
v(t)
Fig. 131. Current in Problem 40
Fig. 130. Problems 38–40
6.4
Short Impulses. Dirac’s Delta Function.
Partial Fractions
An airplane making a “hard” landing, a mechanical system being hit by a hammerblow,
a ship being hit by a single high wave, a tennis ball being hit by a racket, and many other
similar examples appear in everyday life. They are phenomena of an impulsive nature
where actions of forces—mechanical, electrical, etc.—are applied over short intervals
of time.
We can model such phenomena and problems by “Dirac’s delta function,” and solve
them very effecively by the Laplace transform.
To model situations of that type, we consider the function
(1)
fk(t ⫺ a) ⫽ b
1>k
if a ⬉ t ⬉ a ⫹ k
0
otherwise
(Fig. 132)
(and later its limit as k : 0). This function represents, for instance, a force of magnitude
1>k acting from t ⫽ a to t ⫽ a ⫹ k, where k is positive and small. In mechanics, the
integral of a force acting over a time interval a ⬉ t ⬉ a ⫹ k is called the impulse of
the force; similarly for electromotive forces E(t) acting on circuits. Since the blue rectangle
in Fig. 132 has area 1, the impulse of fk in (1) is
(2)
Ik ⫽
冮
ⴥ
fk(t ⫺ a) dt ⫽
冮
a⫹k
a
0
1
dt ⫽ 1.
k
Area = 1
1/k
a a+k
t
Fig. 132. The function ƒk(t ⫺ a) in (1)
c06.qxd
10/28/10
226
6:33 PM
Page 226
CHAP. 6 Laplace Transforms
To find out what will happen if k becomes smaller and smaller, we take the limit of fk
as k : 0 (k ⬎ 0). This limit is denoted by d(t ⫺ a), that is,
d(t ⫺ a) ⫽ lim fk(t ⫺ a).
k:0
d(t ⫺ a) is called the Dirac delta function2 or the unit impulse function.
d(t ⫺ a) is not a function in the ordinary sense as used in calculus, but a so-called
generalized function.2 To see this, we note that the impulse Ik of fk is 1, so that from (1)
and (2) by taking the limit as k : 0 we obtain
(3)
d(t ⫺ a) ⫽ b
⬁
if t ⫽ a
0
otherwise
and
冮
ⴥ
d(t ⫺ a) dt ⫽ 1,
0
but from calculus we know that a function which is everywhere 0 except at a single point
must have the integral equal to 0. Nevertheless, in impulse problems, it is convenient to
operate on d(t ⫺ a) as though it were an ordinary function. In particular, for a continuous
function g(t) one uses the property [often called the sifting property of d(t ⫺ a), not to
be confused with shifting]
冮
(4)
ⴥ
g(t)d(t ⫺ a) dt ⫽ g(a)
0
which is plausible by (2).
To obtain the Laplace transform of d(t ⫺ a), we write
fk(t ⫺ a) ⫽
1
3u(t ⫺ a) ⫺ u(t ⫺ (a ⫹ k))4
k
and take the transform [see (2)]
l{fk(t ⫺ a)} ⫽
1 ⴚas
1 ⫺ eⴚks
3e
⫺ eⴚ(a⫹k)s4 ⫽ eⴚas
.
ks
ks
We now take the limit as k : 0. By l’Hôpital’s rule the quotient on the right has the limit
1 (differentiate the numerator and the denominator separately with respect to k, obtaining
seⴚks and s, respectively, and use seⴚks>s : 1 as k : 0). Hence the right side has the
limit eⴚas. This suggests defining the transform of d(t ⫺ a) by this limit, that is,
(5)
l{d(t ⫺ a)} ⫽ eⴚas.
The unit step and unit impulse functions can now be used on the right side of ODEs
modeling mechanical or electrical systems, as we illustrate next.
2
PAUL DIRAC (1902–1984), English physicist, was awarded the Nobel Prize [jointly with the Austrian
ERWIN SCHRÖDINGER (1887–1961)] in 1933 for his work in quantum mechanics.
Generalized functions are also called distributions. Their theory was created in 1936 by the Russian
mathematician SERGEI L’VOVICH SOBOLEV (1908–1989), and in 1945, under wider aspects, by the French
mathematician LAURENT SCHWARTZ (1915–2002).
c06.qxd
10/28/10
6:33 PM
Page 227
SEC. 6.4 Short Impulses. Dirac’s Delta Function. Partial Fractions
EXAMPLE 1
227
Mass–Spring System Under a Square Wave
Determine the response of the damped mass–spring system (see Sec. 2.8) under a square wave, modeled by
(see Fig. 133)
y s ⫹ 3y r ⫹ 2y ⫽ r(t) ⫽ u(t ⫺ 1) ⫺ u(t ⫺ 2),
Solution.
y(0) ⫽ 0,
y r (0) ⫽ 0.
From (1) and (2) in Sec. 6.2 and (2) and (4) in this section we obtain the subsidiary equation
s 2Y ⫹ 3sY ⫹ 2Y ⫽
1 ⴚs
(e ⫺ eⴚ2s).
s
Y(s) ⫽
Solution
1
(eⴚs ⫺ eⴚ2s).
s(s 2 ⫹ 3s ⫹ 2)
Using the notation F(s) and partial fractions, we obtain
F(s) ⫽
1
s(s ⫹ 3s ⫹ 2)
2
⫽
1
s(s ⫹ 1)(s ⫹ 2)
⫽
1
2
s
⫺
1
s⫹1
⫹
1
2
s⫹2
.
From Table 6.1 in Sec. 6.1, we see that the inverse is
f (t) ⫽ lⴚ1(F) ⫽ 12 ⫺ eⴚt ⫹ 12 eⴚ2t.
Therefore, by Theorem 1 in Sec. 6.3 (t-shifting) we obtain the square-wave response shown in Fig. 133,
y ⫽ lⴚ1(F(s)eⴚs ⫺ F(s)eⴚ2s)
⫽ f (t ⫺ 1)u(t ⫺ 1) ⫺ f (t ⫺ 2)u(t ⫺ 2)
(0 ⬍ t ⬍ 1)
0
1
2
⫽d ⫺e
⫺e
ⴚ(tⴚ1)
ⴚ(tⴚ1)
⫹
⫹e
1 ⴚ2(tⴚ1)
2e
ⴚ(tⴚ2)
⫹
(1 ⬍ t ⬍ 2)
1 ⴚ2(tⴚ1)
2e
⫺
1 ⴚ2(tⴚ2)
2e
(t ⬎ 2).
䊏
y(t)
1
0.5
0
0
1
2
3
4
t
Fig. 133. Square wave and response in Example 1
EXAMPLE 2
Hammerblow Response of a Mass–Spring System
Find the response of the system in Example 1 with the square wave replaced by a unit impulse at time t ⫽ 1.
Solution.
We now have the ODE and the subsidiary equation
y s ⫹ 3y r ⫹ 2y ⫽ d(t ⫺ 1),
(s 2 ⫹ 3s ⫹ 2)Y ⫽ eⴚs.
and
Solving algebraically gives
Y(s) ⫽
eⴚs
(s ⫹ 1)(s ⫹ 2)
⫽a
1
s⫹1
⫺
1
s⫹2
b eⴚs.
By Theorem 1 the inverse is
y(t) ⫽ lⴚ1(Y) ⫽ c
0
eⴚ(tⴚ1) ⫺ eⴚ2(tⴚ1)
if 0 ⬍ t ⬍ 1
if
t ⬎ 1.
c06.qxd
10/28/10
6:33 PM
228
Page 228
CHAP. 6 Laplace Transforms
y(t) is shown in Fig. 134. Can you imagine how Fig. 133 approaches Fig. 134 as the wave becomes shorter and
shorter, the area of the rectangle remaining 1?
䊏
y(t)
0.2
0.1
0
0
1
3
t
5
Fig. 134. Response to a hammerblow in Example 2
EXAMPLE 3
Four-Terminal RLC-Network
Find the output voltage response in Fig. 135 if R ⫽ 20 ⍀, L ⫽ 1 H, C ⫽ 10ⴚ4 F, the input is d(t) (a unit impulse
at time t ⫽ 0), and current and charge are zero at time t ⫽ 0.
Solution.
To understand what is going on, note that the network is an RLC-circuit to which two wires at A
and B are attached for recording the voltage v(t) on the capacitor. Recalling from Sec. 2.9 that current i(t) and
charge q(t) are related by i ⫽ q r ⫽ dq>dt, we obtain the model
Li r ⫹ Ri ⫹
q
C
⫽ Lq s ⫹ Rq r ⫹
q
C
⫽ q s ⫹ 20q r ⫹ 10,000q ⫽ d(t).
From (1) and (2) in Sec. 6.2 and (5) in this section we obtain the subsidiary equation for Q(s) ⫽ l(q)
(s 2 ⫹ 20s ⫹ 10,000)Q ⫽ 1.
Solution
Q⫽
1
(s ⫹ 10)2 ⫹ 9900
.
By the first shifting theorem in Sec. 6.1 we obtain from Q damped oscillations for q and v; rounding 9900 ⬇ 99.502,
we get (Fig. 135)
q ⫽ lⴚ1(Q) ⫽
1
99.50
␦(t)
eⴚ10t sin 99.50t
and
v⫽
q
C
䊏
⫽ 100.5eⴚ10t sin 99.50t.
v
80
R
L
C
40
0
A
B
0.05
0.1
0.15
0.2
0.25
0.3
t
–40
v(t) = ?
–80
Network
Voltage on the capacitor
Fig. 135. Network and output voltage in Example 3
More on Partial Fractions
We have seen that the solution Y of a subsidiary equation usually appears as a quotient
of polynomials Y(s) ⫽ F(s)>G(s), so that a partial fraction representation leads to a sum
of expressions whose inverses we can obtain from a table, aided by the first shifting
theorem (Sec. 6.1). These representations are sometimes called Heaviside expansions.
c06.qxd
10/28/10
6:33 PM
Page 229
SEC. 6.4 Short Impulses. Dirac’s Delta Function. Partial Fractions
229
An unrepeated factor s ⫺ a in G(s) requires a single partial fraction A>(s ⫺ a).
See Examples 1 and 2. Repeated real factors (s ⫺ a)2, (s ⫺ a)3, etc., require partial
fractions
A2
(s ⫺ a)
2
⫹
A1
s⫺a
A3
,
(s ⫺ a)
3
⫹
A2
(s ⫺ a)
2
⫹
A1
s⫺a
,
etc.,
The inverses are (A2t ⫹ A1)eat, (12A3t 2 ⫹ A2t ⫹ A1)eat, etc.
Unrepeated complex factors (s ⫺ a)(s ⫺ a), a ⫽ a ⫹ ib, a ⫽ a ⫺ ib, require a partial
fraction (As ⫹ B)>3(s ⫺ a)2 ⫹ b24. For an application, see Example 4 in Sec. 6.3.
A further one is the following.
EXAMPLE 4
Unrepeated Complex Factors. Damped Forced Vibrations
Solve the initial value problem for a damped mass–spring system acted upon by a sinusoidal force for some
time interval (Fig. 136),
y s ⫹ 2y r ⫹ 2y ⫽ r(t), r(t) ⫽ 10 sin 2t if 0 ⬍ t ⬍ p and 0 if t ⬎ p;
y(0) ⫽ 1,
y r (0) ⫽ ⫺5.
Solution.
From Table 6.1, (1), (2) in Sec. 6.2, and the second shifting theorem in Sec. 6.3, we obtain the
subsidiary equation
(s 2Y ⫺ s ⫹ 5) ⫹ 2(sY ⫺ 1) ⫹ 2Y ⫽ 10
2
s ⫹4
2
(1 ⫺ eⴚps).
We collect the Y-terms, (s 2 ⫹ 2s ⫹ 2)Y, take ⫺s ⫹ 5 ⫺ 2 ⫽ ⫺s ⫹ 3 to the right, and solve,
Y⫽
(6)
20
(s ⫹ 4)(s ⫹ 2s ⫹ 2)
2
2
⫺
20eⴚps
(s ⫹ 4)(s ⫹ 2s ⫹ 2)
2
2
⫹
s⫺3
s ⫹ 2s ⫹ 2
2
.
For the last fraction we get from Table 6.1 and the first shifting theorem
lⴚ1 b
(7)
s⫹1⫺4
(s ⫹ 1)2 ⫹ 1
ⴚt
r ⫽ e (cos t ⫺ 4 sin t).
In the first fraction in (6) we have unrepeated complex roots, hence a partial fraction representation
20
(s 2 ⫹ 4)(s 2 ⫹ 2s ⫹ 2)
⫽
As ⫹ B
s2 ⫹ 4
⫹
Ms ⫹ N
s 2 ⫹ 2s ⫹ 2
.
Multiplication by the common denominator gives
20 ⫽ (As ⫹ B)(s 2 ⫹ 2s ⫹ 2) ⫹ (Ms ⫹ N)(s 2 ⫹ 4).
We determine A, B, M, N. Equating the coefficients of each power of s on both sides gives the four equations
(a) 3s 34 :
0⫽A⫹M
(b)
(c)
0 ⫽ 2A ⫹ 2B ⫹ 4M
(d)
3s4 :
3s 24 :
0 ⫽ 2A ⫹ B ⫹ N
3s 04 : 20 ⫽ 2B ⫹ 4N.
We can solve this, for instance, obtaining M ⫽ ⫺A from (a), then A ⫽ B from (c), then N ⫽ ⫺3A from (b),
and finally A ⫽ ⫺2 from (d). Hence A ⫽ ⫺2, B ⫽ ⫺2, M ⫽ 2, N ⫽ 6, and the first fraction in (6) has the
representation
(8)
⫺2s ⫺ 2
s2 ⫹ 4
⫹
2(s ⫹ 1) ⫹ 6 ⫺ 2
(s ⫹ 1)2 ⫹ 1
.
Inverse transform:
⫺2 cos 2t ⫺ sin 2t ⫹ eⴚt(2 cos t ⫹ 4 sin t).
c06.qxd
10/28/10
230
6:33 PM
Page 230
CHAP. 6 Laplace Transforms
The sum of this inverse and (7) is the solution of the problem for 0 ⬍ t ⬍ p, namely (the sines cancel),
y(t) ⫽ 3eⴚt cos t ⫺ 2 cos 2t ⫺ sin 2t
(9)
if 0 ⬍ t ⬍ p.
In the second fraction in (6), taken with the minus sign, we have the factor eⴚps, so that from (8) and the second
shifting theorem (Sec. 6.3) we get the inverse transform of this fraction for t ⬎ 0 in the form
⫹2 cos (2t ⫺ 2p) ⫹ sin (2t ⫺ 2p) ⫺ eⴚ(tⴚp) 32 cos (t ⫺ p) ⫹ 4 sin (t ⫺ p)4
⫽ 2 cos 2t ⫹ sin 2t ⫹ eⴚ(tⴚp) (2 cos t ⫹ 4 sin t).
The sum of this and (9) is the solution for t ⬎ p,
y(t) ⫽ eⴚt3(3 ⫹ 2ep) cos t ⫹ 4ep sin t4
(10)
if t ⬎ p.
Figure 136 shows (9) (for 0 ⬍ t ⬍ p) and (10) (for t ⬎ p), a beginning vibration, which goes to zero rapidly
because of the damping and the absence of a driving force after t ⫽ p.
䊏
y(t)
2
1
y = 0 (Equilibrium
position)
y
0
π
2π
3π
4π
t
–1
Driving force
Dashpot (damping)
–2
Mechanical system
Output (solution)
Fig. 136. Example 4
The case of repeated complex factors 3(s ⫺ a)(s ⫺ a )42, which is important in connection
with resonance, will be handled by “convolution” in the next section.
PROBLEM SET 6.4
1. CAS PROJECT. Effect of Damping. Consider a
vibrating system of your choice modeled by
y s ⫹ cy r ⫹ ky ⫽ d(t).
(a) Using graphs of the solution, describe the effect of
continuously decreasing the damping to 0, keeping k
constant.
(b) What happens if c is kept constant and k is
continuously increased, starting from 0?
(c) Extend your results to a system with two
d-functions on the right, acting at different times.
2. CAS EXPERIMENT. Limit of a Rectangular Wave.
Effects of Impulse.
(a) In Example 1 in the text, take a rectangular wave
of area 1 from 1 to 1 ⫹ k. Graph the responses for a
sequence of values of k approaching zero, illustrating
that for smaller and smaller k those curves approach
the curve shown in Fig. 134. Hint: If your CAS gives
no solution for the differential equation, involving k,
take specific k’s from the beginning.
(b) Experiment on the response of the ODE in Example
1 (or of another ODE of your choice) to an impulse
d(t ⫺ a) for various systematically chosen a (⬎ 0);
choose initial conditions y(0) ⫽ 0, y r (0) ⫽ 0. Also consider the solution if no impulse is applied. Is there a
dependence of the response on a? On b if you choose
bd(t ⫺ a)? Would ⫺d(t ⫺ a苲) with a苲 ⬎ a annihilate the
effect of d(t ⫺ a)? Can you think of other questions that
one could consider experimentally by inspecting graphs?
3–12
EFFECT OF DELTA (IMPULSE)
ON VIBRATING SYSTEMS
Find and graph or sketch the solution of the IVP. Show the
details.
3. y s ⫹ 4y ⫽ d(t ⫺ p), y(0) ⫽ 8, y r (0) ⫽ 0
c06.qxd
10/28/10
6:33 PM
Page 231
SEC. 6.4 Short Impulses. Dirac’s Delta Function. Partial Fractions
4. y s ⫹ 16y ⫽ 4d(t ⫺ 3p), y(0) ⫽ 2, y r (0) ⫽ 0
5. y s ⫹ y ⫽ d(t ⫺ p) ⫺ d(t ⫺ 2p),
y(0) ⫽ 0, y r (0) ⫽ 1
6. y s ⫹ 4y r ⫹ 5y ⫽ d(t ⫺ 1), y(0) ⫽ 0, y r (0) ⫽ 3
7. 4y s ⫹ 24y r ⫹ 37y ⫽ 17e⫺t ⫹ d(t ⫺ 12),
y(0) ⫽ 1, y r (0) ⫽ 1
8. y s ⫹ 3y r ⫹ 2y ⫽ 10(sin t ⫹ d(t ⫺ 1)), y(0) ⫽ 1,
y r (0) ⫽ ⫺1
9. y s ⫹ 4y r ⫹ 5y ⫽ 31 ⫺ u(t ⫺ 10)4et ⫺ e10d(t ⫺ 10),
y(0) ⫽ 0, y r (0) ⫽ 1
10. y s ⫹ 5y r ⫹ 6y ⫽ d(t ⫺ 12p) ⫹ u(t ⫺ p) cos t,
y(0) ⫽ 0, y r (0) ⫽ 0
11. y s ⫹ 5y r ⫹ 6y ⫽ u(t ⫺ 1) ⫹ d(t ⫺ 2),
y(0) ⫽ 0, y r (0) ⫽ 1
12. y s ⫹ 2y r ⫹ 5y ⫽ 25t ⫺ 100d(t ⫺ p), y(0) ⫽ ⫺2,
y r (0) ⫽ 5
13. PROJECT. Heaviside Formulas. (a) Show that for
a simple root a and fraction A>(s ⫺ a) in F(s)>G(s) we
have the Heaviside formula
A ⫽ lim
(s ⫺ a)F(s)
G(s)
s:a
231
Set t ⫽ (n ⫺ 1)p in the nth integral. Take out eⴚ(nⴚ1)p
from under the integral sign. Use the sum formula for
the geometric series.
(b) Half-wave rectifier. Using (11), show that the
half-wave rectification of sin vt in Fig. 137 has the
Laplace transform
(s 2 ⫹ v2)(1 ⫺ eⴚ2ps>v)
v
⫽
.
2
2
(s ⫹ v )(1 ⫺ eⴚps>v)
(A half-wave rectifier clips the negative portions of the
curve. A full-wave rectifier converts them to positive;
see Fig. 138.)
(c) Full-wave rectifier. Show that the Laplace transform of the full-wave rectification of sin vt is
v
F(s)
⫽
Am
ps
2v
.
f (t)
1
Amⴚ1
⫹
2
.
(s ⫺ a)
(s ⫺ a)mⴚ1
A1
⫹ s ⫺ a ⫹ further fractions
m
coth
s ⫹v
2
0
(b) Similarly, show that for a root a of order m and
fractions in
G(s)
v(1 ⫹ eⴚps>v)
l( f ) ⫽
π /ω
2π /ω
3π /ω
t
Fig. 137. Half-wave rectification
f (t)
⫹ Á
1
0
π /ω
2π /ω
3π /ω
t
Fig. 138. Full-wave rectification
we have the Heaviside formulas for the first coefficient
Am ⫽ lim
(d) Saw-tooth wave. Find the Laplace transform of the
saw-tooth wave in Fig. 139.
(s ⫺ a)mF(s)
s:a
G(s)
f (t)
and for the other coefficients
k
m
d mⴚk (s ⫺ a) F(s)
1
lim
Ak ⫽
d,
c
(m ⫺ k)! s:a ds mⴚk
G(s)
k ⫽ 1, Á , m ⫺ 1.
0
p
2p
t
3p
Fig. 139. Saw-tooth wave
14. TEAM PROJECT. Laplace Transform of Periodic
Functions
(a) Theorem. The Laplace transform of a piecewise
continuous function f (t) with period p is
15. Staircase function. Find the Laplace transform of the
staircase function in Fig. 140 by noting that it is the
difference of kt>p and the function in 14(d).
f (t)
(11)
l( f ) ⫽
1
1 ⫺ eⴚps
p
冮e
ⴚst
f (t) dt
(s ⬎ 0).
k
0
0
Prove this theorem. Hint: Write 兰0⬁ ⫽ 兰0p ⫹ 兰p2p ⫹ Á .
p
2p
3p
Fig. 140. Staircase function
t
c06.qxd
10/28/10
6:33 PM
232
6.5
Page 232
CHAP. 6 Laplace Transforms
Convolution. Integral Equations
Convolution has to do with the multiplication of transforms. The situation is as follows.
Addition of transforms provides no problem; we know that l( f ⫹ g) ⫽ l( f ) ⫹ l(g).
Now multiplication of transforms occurs frequently in connection with ODEs, integral
equations, and elsewhere. Then we usually know l( f ) and l(g) and would like to know
the function whose transform is the product l( f )l(g). We might perhaps guess that it
is fg, but this is false. The transform of a product is generally different from the product
of the transforms of the factors,
l( fg) ⫽ l( f )l(g)
in general.
To see this take f ⫽ et and g ⫽ 1. Then fg ⫽ et, l( fg) ⫽ 1>(s ⫺ 1), but l( f ) ⫽ 1>(s ⫺ 1)
and l(1) ⫽ 1>s give l( f )l(g) ⫽ 1>(s 2 ⫺ s).
According to the next theorem, the correct answer is that l( f )l(g) is the transform of
the convolution of f and g, denoted by the standard notation f * g and defined by the integral
t
h(t) ⫽ ( f * g)(t) ⫽
(1)
˛
冮 f (t)g(t ⫺ t) dt.
˛
0
THEOREM 1
Convolution Theorem
If two functions f and g satisfy the assumption in the existence theorem in Sec. 6.1,
so that their transforms F and G exist, the product H ⫽ FG is the transform of h
given by (1). (Proof after Example 2.)
EXAMPLE 1
Convolution
Let H(s) ⫽ 1>[(s ⫺ a)s]. Find h(t).
1>(s ⫺ a) has the inverse f (t) ⫽ eat, and 1>s has the inverse g(t) ⫽ 1. With f (t) ⫽ eat and
g(t ⫺ t) ⬅ 1 we thus obtain from (1) the answer
Solution.
t
h(t) ⫽ eat * 1 ⫽
冮e
at
0
# 1 dt ⫽ 1 (eat ⫺ 1).
a
To check, calculate
H(s) ⫽ l(h)(s) ⫽
EXAMPLE 2
1
1
1
a
1
1 # 1
a
⫺ b⫽ # 2
⫽
⫽ l(eat)l(1).
s
s⫺a s
a s⫺a
a s ⫺ as
䊏
Convolution
Let H(s) ⫽ 1>(s 2 ⫹ v2)2. Find h(t).
Solution.
The inverse of 1>(s 2 ⫹ v2) is (sin vt)>v. Hence from (1) and the first formula in (11) in App. 3.1
we obtain
h(t) ⫽
t
冮 sin vt sin v(t ⫺ t) dt
1
sin vt sin vt
*
⫽ 2
v
v
v
⫽
0
t
1
2
2v
冮 [⫺cos vt ⫹ cos (2vt ⫺ vt)] dt
0
c06.qxd
10/28/10
6:33 PM
Page 233
SEC. 6.5 Convolution. Integral Equations
233
⫽
⫽
1
2
2v
1
2v2
c ⫺t cos vt ⫹
sin vt t
v d t⫽0
c ⫺t cos vt ⫹
sin vt
v d
䊏
in agreement with formula 21 in the table in Sec. 6.9.
PROOF
We prove the Convolution Theorem 1. CAUTION! Note which ones are the variables
of integration! We can denote them as we want, for instance, by t and p, and write
冮
F(s) ⫽
ⴥ
eⴚstf (t) dt
and
冮
G(s) ⫽
0
ⴥ
eⴚspg( p) dp.
0
We now set t ⫽ p ⫹ t, where t is at first constant. Then p ⫽ t ⫺ t, and t varies from
t to ⬁ . Thus
G(s) ⫽
冮
ⴥ
eⴚs(tⴚt)g(t ⫺ t) dt ⫽ est
t
冮
ⴥ
eⴚstg(t ⫺ t) dt.
t
t in F and t in G vary independently. Hence we can insert the G-integral into the
F-integral. Cancellation of eⴚst and est then gives
F(s)G(s) ⫽
冮
ⴥ
eⴚstf (t)est
冮
ⴥ
冮
eⴚstg(t ⫺ t) dt dt ⫽
t
0
ⴥ
f (t)
0
冮
ⴥ
eⴚstg(t ⫺ t) dt dt.
t
Here we integrate for fixed t over t from t to ⬁ and then over t from 0 to ⬁ . This is the
blue region in Fig. 141. Under the assumption on f and g the order of integration can be
reversed (see Ref. [A5] for a proof using uniform convergence). We then integrate first
over t from 0 to t and then over t from 0 to ⬁ , that is,
F(s)G(s) ⫽
冮
ⴥ
eⴚst
0
冮
t
f (t)g(t ⫺ t) dt dt ⫽
0
冮
ⴥ
eⴚsth(t) dt ⫽ l(h) ⫽ H(s).
0
䊏
This completes the proof.
τ
t
Fig. 141. Region of integration in the
t␶-plane in the proof of Theorem 1
c06.qxd
10/28/10
6:33 PM
234
Page 234
CHAP. 6 Laplace Transforms
From the definition it follows almost immediately that convolution has the properties
f *g ⫽ g* f
(commutative law)
f * (g1 ⫹ g2) ⫽ f * g1 ⫹ f * g2
(distributive law)
( f * g) * v ⫽ f * (g * v)
(associative law)
f *0⫽0*f⫽0
similar to those of the multiplication of numbers. However, there are differences of which
you should be aware.
EXAMPLE 3
Unusual Properties of Convolution
f * 1 ⫽ f in general. For instance,
t*1⫽
冮
t
0
1
t # 1 dt ⫽ t 2 ⫽ t.
2
( f * f )(t) ⭌ 0 may not hold. For instance, Example 2 with v ⫽ 1 gives
sin t * sin t ⫽ ⫺12 t cos t ⫹ 12 sin t
(Fig. 142).
䊏
4
2
0
2 4 6 8 10
t
–2
–4
Fig. 142. Example 3
We shall now take up the case of a complex double root (left aside in the last section in
connection with partial fractions) and find the solution (the inverse transform) directly by
convolution.
EXAMPLE 4
Repeated Complex Factors. Resonance
In an undamped mass–spring system, resonance occurs if the frequency of the driving force equals the natural
frequency of the system. Then the model is (see Sec. 2.8)
y s ⫹ v 02 y ⫽ K sin v 0 t
where v20 ⫽ k>m, k is the spring constant, and m is the mass of the body attached to the spring. We assume
y(0) ⫽ 0 and y r (0) ⫽ 0, for simplicity. Then the subsidiary equation is
s 2Y ⫹ v 02Y ⫽
Kv 0
s 2 ⫹ v 02
.
Its solution is
Y⫽
Kv 0
(s 2 ⫹ v 02) 2
.
c06.qxd
10/28/10
6:33 PM
Page 235
SEC. 6.5 Convolution. Integral Equations
235
This is a transform as in Example 2 with v ⫽ v0 and multiplied by Kv0. Hence from Example 2 we can see
directly that the solution of our problem is
y(t) ⫽
K
Kv 0
sin v 0 t
a⫺t cos v 0 t ⫹
b⫽
(⫺v 0 t cos v 0 t ⫹ sin v 0 t).
2v 02
2v 02
v0
We see that the first term grows without bound. Clearly, in the case of resonance such a term must occur. (See
䊏
also a similar kind of solution in Fig. 55 in Sec. 2.8.)
Application to Nonhomogeneous Linear ODEs
Nonhomogeneous linear ODEs can now be solved by a general method based on
convolution by which the solution is obtained in the form of an integral. To see this, recall
from Sec. 6.2 that the subsidiary equation of the ODE
y s ⫹ ay r ⫹ by ⫽ r(t)
(2)
(a, b constant)
has the solution [(7) in Sec. 6.2]
Y(s) ⫽ [(s ⫹ a)y(0) ⫹ y r (0)]Q(s) ⫹ R(s)Q(s)
with R(s) ⫽ l(r) and Q(s) ⫽ 1>(s 2 ⫹ as ⫹ b) the transfer function. Inversion of the first
term 3 Á 4 provides no difficulty; depending on whether 14a 2 ⫺ b is positive, zero, or
negative, its inverse will be a linear combination of two exponential functions, or of the
form (c1 ⫹ c2t)eⴚat>2, or a damped oscillation, respectively. The interesting term is
R(s)Q(s) because r(t) can have various forms of practical importance, as we shall see. If
y(0) ⫽ 0 and y r (0) ⫽ 0, then Y ⫽ RQ, and the convolution theorem gives the solution
t
冮 q(t ⫺ t)r(t) dt.
y(t) ⫽
(3)
0
EXAMPLE 5
Response of a Damped Vibrating System to a Single Square Wave
Using convolution, determine the response of the damped mass–spring system modeled by
y s ⫹ 3y r ⫹ 2y ⫽ r(t),
r(t) ⫽ 1 if 1 ⬍ t ⬍ 2 and 0 otherwise,
y(0) ⫽ y r (0) ⫽ 0.
This system with an input (a driving force) that acts for some time only (Fig. 143) has been solved by partial
fraction reduction in Sec. 6.4 (Example 1).
Solution by Convolution.
Q(s) ⫽
1
s 2 ⫹ 3s ⫹ 2
⫽
The transfer function and its inverse are
1
(s ⫹ 1)(s ⫹ 2)
⫽
1
s⫹1
⫺
1
s⫹2
,
q(t) ⫽ eⴚt ⫺ eⴚ2t.
hence
Hence the convolution integral (3) is (except for the limits of integration)
y(t) ⫽
冮 q(t ⫺ t) # 1 dt ⫽ 冮 3e
ⴚ(tⴚt)
⫺ eⴚ2(tⴚt)4 dt ⫽ eⴚ(tⴚt) ⫺
1
2
eⴚ2(tⴚt).
Now comes an important point in handling convolution. r(t) ⫽ 1 if 1 ⬍ t ⬍ 2 only. Hence if t ⬍ 1, the integral
is zero. If 1 ⬍ t ⬍ 2, we have to integrate from t ⫽ 1 (not 0) to t. This gives (with the first two terms from the
upper limit)
y(t) ⫽ eⴚ0 ⫺ 12 eⴚ0 ⫺ (eⴚ(tⴚ1) ⫺ 12 eⴚ2(tⴚ1)) ⫽ 12 ⫺ eⴚ(tⴚ1) ⫹ 12 eⴚ2(tⴚ1).
c06.qxd
11/4/10
12:22 PM
236
Page 236
CHAP. 6 Laplace Transforms
If t ⬎ 2, we have to integrate from t ⫽ 1 to 2 (not to t). This gives
y(t) ⫽ eⴚ(tⴚ2) ⫺ 12 eⴚ2(tⴚ2) ⫺ (eⴚ(tⴚ1) ⫺ 12 eⴚ2(tⴚ1)).
Figure 143 shows the input (the square wave) and the interesting output, which is zero from 0 to 1, then increases,
reaches a maximum (near 2.6) after the input has become zero (why?), and finally decreases to zero in a monotone
fashion.
䊏
y(t)
1
Output (response)
0.5
0
0
1
2
3
4
t
Fig. 143. Square wave and response in Example 5
Integral Equations
Convolution also helps in solving certain integral equations, that is, equations in which the
unknown function y(t) appears in an integral (and perhaps also outside of it). This concerns
equations with an integral of the form of a convolution. Hence these are special and it suffices
to explain the idea in terms of two examples and add a few problems in the problem set.
EXAMPLE 6
A Volterra Integral Equation of the Second Kind
Solve the Volterra integral equation of the second kind3
y(t) ⫺
冮
t
y(t) sin (t ⫺ t) dt ⫽ t.
0
Solution. From (1) we see that the given equation can be written as a convolution, y ⫺ y * sin t ⫽ t. Writing
Y ⫽ l(y) and applying the convolution theorem, we obtain
Y(s) ⫺ Y(s)
1
s2 ⫹ 1
⫽ Y(s)
s2
s2 ⫹ 1
⫽
1
s2
.
The solution is
Y(s) ⫽
s2 ⫹ 1
s
4
⫽
1
s
2
⫹
1
s
4
and gives the answer
y(t) ⫽ t ⫹
t3
6
.
Check the result by a CAS or by substitution and repeated integration by parts (which will need patience).
EXAMPLE 7
䊏
Another Volterra Integral Equation of the Second Kind
Solve the Volterra integral equation
y(t) ⫺
冮
t
(1 ⫹ t) y(t ⫺ t) dt ⫽ 1 ⫺ sinh t.
0
3
If the upper limit of integration is variable, the equation is named after the Italian mathematician VITO
VOLTERRA (1860–1940), and if that limit is constant, the equation is named after the Swedish mathematician
ERIK IVAR FREDHOLM (1866–1927). “Of the second kind (first kind)” indicates that y occurs (does not
occur) outside of the integral.
c06.qxd
10/28/10
6:33 PM
Page 237
SEC. 6.5 Convolution. Integral Equations
237
By (1) we can write y ⫺ (1 ⫹ t) * y ⫽ 1 ⫺ sinh t. Writing Y ⫽ l(y), we obtain by using the
convolution theorem and then taking common denominators
Solution.
1
1
1
1
,
Y(s) c 1 ⫺ a ⫹ 2 b d ⫽ ⫺ 2
s
s
s
s ⫺1
s2 ⫺ s ⫺ 1
s2 ⫺ 1 ⫺ s
Y(s) #
⫽
.
2
s
s(s 2 ⫺ 1)
hence
(s 2 ⫺ s ⫺ 1)>s cancels on both sides, so that solving for Y simply gives
Y(s) ⫽
s
s2 ⫺ 1
and the solution is
䊏
y(t) ⫽ cosh t.
PROBLEM SET 6.5
1–7
CONVOLUTIONS BY INTEGRATION
Find:
1. 1 * 1
2. 1 * sin vt
t
ⴚt
3. e * e
4. (cos vt) * (cos vt)
5. (sin vt) * (cos vt)
6. eat * ebt (a ⫽ b)
t
7. t * e
8–14
INTEGRAL EQUATIONS
Solve by the Laplace transform, showing the details:
8. y(t) ⫹ 4
冮
16. TEAM PROJECT. Properties of Convolution. Prove:
(a) Commutativity, f * g ⫽ g * f
(b) Associativity, ( f * g) * v ⫽ f * (g * v)
(c) Distributivity, f * (g1 ⫹ g2) ⫽ f * g1 ⫹ f * g2
(d) Dirac’s delta. Derive the sifting formula (4) in Sec.
6.4 by using fk with a ⫽ 0 [(1), Sec. 6.4] and applying
the mean value theorem for integrals.
(e) Unspecified driving force. Show that forced
vibrations governed by
t
y s ⫹ v2y ⫽ r(t), y(0) ⫽ K 1,
y(t)(t ⫺ t) dt ⫽ 2t
0
9. y(t) ⫺
冮
t
冮
t
冮
t
冮
t
with v ⫽ 0 and an unspecified driving force r(t)
can be written in convolution form,
y(t) dt ⫽ 1
0
10. y(t) ⫺
y r (0) ⫽ K 2
y⫽
y(t) sin 2(t ⫺ t) dt ⫽ sin 2t
K2
1
sin vt * r(t) ⫹ K 1 cos vt ⫹
sin vt.
v
v
0
11. y(t) ⫹
17–26
(t ⫺ t)y(t) dt ⫽ 1
0
12. y(t) ⫹
y(t) cosh (t ⫺ t) dt ⫽ t ⫹ e
t
0
13. y(t) ⫹ 2et
冮
t
y(t)eⴚt dt ⫽ tet
0
14. y(t) ⫺
冮
t
0
1
y(t)(t ⫺ t) dt ⫽ 2 ⫺ t 2
2
15. CAS EXPERIMENT. Variation of a Parameter.
(a) Replace 2 in Prob. 13 by a parameter k and
investigate graphically how the solution curve changes
if you vary k, in particular near k ⫽ ⫺2.
(b) Make similar experiments with an integral equation
of your choice whose solution is oscillating.
INVERSE TRANSFORMS
BY CONVOLUTION
Showing details, find f (t) if l( f )
5.5
17.
18.
(s ⫹ 1.5)(s ⫺ 4)
2ps
19. 2
20.
(s ⫹ p2)2
v
21. 2 2
22.
s (s ⫹ v2)
40.5
23.
24.
s(s 2 ⫺ 9)
25.
equals:
1
(s ⫺ a)2
9
s(s ⫹ 3)
eⴚas
s(s ⫺ 2)
240
(s 2 ⫹ 1)(s 2 ⫹ 25)
18s
(s 2 ⫹ 36)2
26. Partial Fractions. Solve Probs. 17, 21, and 23 by
partial fraction reduction.
c06.qxd
10/28/10
6:33 PM
238
6.6
Page 238
CHAP. 6 Laplace Transforms
Differentiation and Integration of Transforms.
ODEs with Variable Coefficients
The variety of methods for obtaining transforms and inverse transforms and their
application in solving ODEs is surprisingly large. We have seen that they include direct
integration, the use of linearity (Sec. 6.1), shifting (Secs. 6.1, 6.3), convolution (Sec. 6.5),
and differentiation and integration of functions f (t) (Sec. 6.2). In this section, we shall
consider operations of somewhat lesser importance. They are the differentiation and
integration of transforms F(s) and corresponding operations for functions f (t). We show
how they are applied to ODEs with variable coefficients.
Differentiation of Transforms
It can be shown that, if a function f(t) satisfies the conditions of the existence theorem in
Sec. 6.1, then the derivative F r (s) ⫽ dF>ds of the transform F(s) ⫽ l( f ) can be obtained
by differentiating F(s) under the integral sign with respect to s (proof in Ref. [GenRef4]
listed in App. 1). Thus, if
F(s) ⫽
冮
ⴥ
eⴚstf (t) dt,
F r(s) ⫽ ⫺
then
0
冮
ⴥ
eⴚstt f (t) dt.
0
Consequently, if l( f ) ⫽ F(s), then
(1)
l{tf (t)} ⫽ ⫺F r (s),
lⴚ1{F r (s)} ⫽ ⫺tf (t)
hence
where the second formula is obtained by applying lⴚ1 on both sides of the first formula.
In this way, differentiation of the transform of a function corresponds to the multiplication
of the function by ⫺t.
EXAMPLE 1
Differentiation of Transforms. Formulas 21–23 in Sec. 6.9
We shall derive the following three formulas.
l( f )
(2)
(3)
(4)
1
1
(s ⫹ b )
s
2
2 2
(s 2 ⫹ b2)2
s2
(s ⫹ b )
2
f (t)
2 2
(sin bt ⫺ bt cos bt)
2b3
1
sin bt
2b
1
(sin bt ⫹ bt cos bt)
2b
From (1) and formula 8 (with v ⫽ b) in Table 6.1 of Sec. 6.1 we obtain by differentiation
(CAUTION! Chain rule!)
Solution.
l(t sin bt) ⫽
2bs
(s ⫹ b2)2
2
.
c06.qxd
10/30/10
12:06 AM
Page 239
SEC. 6.6 Differentiation and Integration of Transforms. ODEs with Variable Coefficients
239
Dividing by 2b and using the linearity of l, we obtain (3).
Formulas (2) and (4) are obtained as follows. From (1) and formula 7 (with v ⫽ b) in Table 6.1 we find
l(t cos bt) ⫽ ⫺
(5)
(s 2 ⫹ b2) ⫺ 2s 2
(s ⫹ b )
2
2 2
s 2 ⫺ b2
⫽
(s 2 ⫹ b2)2
.
From this and formula 8 (with v ⫽ b) in Table 6.1 we have
l at cos bt ⫾
1
b
sin btb ⫽
s 2 ⫺ b2
⫾
(s 2 ⫹ b2)2
1
˛
s 2 ⫹ b2
.
On the right we now take the common denominator. Then we see that for the plus sign the numerator becomes
s 2 ⫺ b2 ⫹ s 2 ⫹ b2 ⫽ 2s 2, so that (4) follows by division by 2. Similarly, for the minus sign the numerator
takes the form s 2 ⫺ b2 ⫺ s 2 ⫺ b2 ⫽ ⫺2b2, and we obtain (2). This agrees with Example 2 in Sec. 6.5. 䊏
Integration of Transforms
Similarly, if f (t) satisfies the conditions of the existence theorem in Sec. 6.1 and the limit
of f (t)>t, as t approaches 0 from the right, exists, then for s ⬎ k,
(6)
le
f (t)
f ⫽
t
冮
ⴥ
F(s苲) ds苲
lⴚ1 e
hence
s
冮
ⴥ
F(s苲 ) ds苲 f ⫽
s
f (t)
.
t
In this way, integration of the transform of a function f (t) corresponds to the division of
f (t) by t.
We indicate how (6) is obtained. From the definition it follows that
冮
ⴥ
ⴥ
苲
冮 c冮
苲
F(s ) ds ⫽
s
s
ⴥ
0
eⴚs tf (t) dt d ds苲,
~
and it can be shown (see Ref. [GenRef4] in App. 1) that under the above assumptions we
may reverse the order of integration, that is,
冮
ⴥ
ⴥ
F(s苲) ds苲 ⫽
s
ⴥ
冮 c冮
0
s
~
苲t
ⴚs
Integration of e
with respect to 苲s gives e
ⴚst
equals e >t. Therefore,
冮
ⴥ
苲
苲
F(s ) ds ⫽
s
EXAMPLE 2
冮
eⴚstf (t) ds苲 d dt ⫽
冮
苲t
ⴚs
ⴥ
eⴚst
0
ⴥ
0
f (t) c
冮
s
ⴥ
eⴚst ds苲 d dt.
~
>(⫺t). Here the integral over 苲s on the right
f (t)
f (t)
dt ⫽ l e
f
t
t
(s ⬎ k). 䊏
Differentiation and Integration of Transforms
Find the inverse transform of ln a1 ⫹
Solution.
v2
s2
b ⫽ ln
s 2 ⫹ v2
s2
.
Denote the given transform by F(s). Its derivative is
F r (s) ⫽
d
ds
(ln (s 2 ⫹ v2) ⫺ ln s 2) ⫽
2s
s 2 ⫹ v2
⫺
2s
s2
.
c06.qxd
10/28/10
6:33 PM
240
Page 240
CHAP. 6 Laplace Transforms
Taking the inverse transform and using (1), we obtain
lⴚ{F r (s)} ⫽ lⴚ1 e
2s
2
⫺ f ⫽ 2 cos vt ⫺ 2 ⫽ ⫺tf (t2.
s 2 ⫹ v2
s
Hence the inverse f (t) of F(s) is f (t) ⫽ 2(1 ⫺ cos vt)>t. This agrees with formula 42 in Sec. 6.9.
Alternatively, if we let
G(s) ⫽
2s
2
⫺ ,
s
s 2 ⫹ v2
g(t) ⫽ lⴚ1(G) ⫺ 2(cos vt ⫺ 1).
then
From this and (6) we get, in agreement with the answer just obtained,
lⴚ1 e ln
s 2 ⫹ v2
f ⫽ lⴚ1 e
s2
冮
ⴥ
s
G(s) ds f ⫽ ⫺
g(t)
t
⫽
2
(1 ⫺ cos vt2,
t
the minus occurring since s is the lower limit of integration.
In a similar way we obtain formula 43 in Sec. 6.9,
lⴚ1 e ln a1 ⫺
a2
2
b f ⫽ (1 ⫺ cosh at2.
t
s2
䊏
Special Linear ODEs with Variable Coefficients
Formula (1) can be used to solve certain ODEs with variable coefficients. The idea is this.
Let l(y) ⫽ Y. Then l(y r ) ⫽ sY ⫺ y(0) (see Sec. 6.2). Hence by (1),
l(ty r ) ⫽ ⫺
(7)
d
dY
[sY ⫺ y(0)] ⫽ ⫺Y ⫺ s .
ds
ds
Similarly, l(y s ) ⫽ s 2Y ⫺ sy(0) ⫺ y r (0) and by (1)
(8)
l(ty s ) ⫽ ⫺
d 2
dY
[s Y ⫺ sy(0) ⫺ y r (0)] ⫽ ⫺2sY ⫺ s 2
⫹ y(0).
ds
ds
Hence if an ODE has coefficients such as at ⫹ b, the subsidiary equation is a first-order
ODE for Y, which is sometimes simpler than the given second-order ODE. But if the latter
has coefficients at 2 ⫹ bt ⫹ c, then two applications of (1) would give a second-order
ODE for Y, and this shows that the present method works well only for rather special
ODEs with variable coefficients. An important ODE for which the method is advantageous
is the following.
EXAMPLE 3
Laguerre’s Equation. Laguerre Polynomials
Laguerre’s ODE is
ty s ⫹ (1 ⫺ t)y r ⫹ ny ⫽ 0.
(9)
We determine a solution of (9) with n ⫽ 0, 1, 2, Á . From (7)–(9) we get the subsidiary equation
2
c ⫺2sY ⫺ s
dY
ds
⫹ y(0) d ⫹ sY ⫺ y(0) ⫺ a⫺Y ⫺ s
dY
ds
b ⫹ nY ⫽ 0.
c06.qxd
10/28/10
6:33 PM
Page 241
SEC. 6.6 Differentiation and Integration of Transforms. ODEs with Variable Coefficients
241
Simplification gives
(s ⫺ s 2)
dY
ds
⫹ (n ⫹ 1 ⫺ s)Y ⫽ 0.
Separating variables, using partial fractions, integrating (with the constant of integration taken to be zero), and
taking exponentials, we get
(10*)
n
dY
n⫹1⫺s
n⫹1
ds ⫽ a
b ds
⫽⫺
⫺
s
Y
s⫺1
s ⫺ s2
Y⫽
and
(s ⫺ 1)n
s n⫹1
.
We write l n ⫽ lⴚ1(Y) and prove Rodrigues’s formula
l 0 ⫽ 1,
(10)
l n(t) ⫽
et d n
n! dt n
(t neⴚt),
n ⫽ 1, 2, Á .
These are polynomials because the exponential terms cancel if we perform the indicated differentiations. They
are called Laguerre polynomials and are usually denoted by L n (see Problem Set 5.7, but we continue to reserve
capital letters for transforms). We prove (10). By Table 6.1 and the first shifting theorem (s-shifting),
l(t neⴚt) ⫽
n!
(s ⫹ 1)
n⫹1
,
le
hence by (3) in Sec. 6.2
dn
dt
n
(t neⴚt) f ⫽
n!s n
(s ⫹ 1)n⫹1
because the derivatives up to the order n ⫺ 1 are zero at 0. Now make another shift and divide by n! to get
[see (10) and then (10*)]
l(l n) ⫽
(s ⫺ 1)n
s n⫹1
䊏
⫽ Y.
PROBLEM SET 6.6
1. REVIEW REPORT. Differentiation and Integration
of Functions and Transforms. Make a draft of these
four operations from memory. Then compare your draft
with the text and write a 2- to 3-page report on these
operations and their significance in applications.
2–11
TRANSFORMS BY DIFFERENTIATION
Showing the details of your work, find l( f ) if f (t) equals:
2. 3t sinh 4t
3. 12 teⴚ3t
4. teⴚt cos t
5. t cos vt
6. t 2 sin 3t
7. t 2 cosh 2t
8. teⴚkt sin t
9. 12t 2 sin pt
10. t nekt
11. 4t cos 12 pt
12. CAS PROJECT. Laguerre Polynomials. (a) Write a
CAS program for finding l n(t) in explicit form from (10).
Apply it to calculate l 0, Á , l 10. Verify that l 0, Á , l 10
satisfy Laguerre’s differential equation (9).
(b) Show that
(⫺1)m n m
a bt
m
m⫽0 m!
n
l n(t) ⫽ a
and calculate l 0, Á , l 10 from this formula.
(c) Calculate l 0, Á , l 10 recursively from l 0 ⫽ 1, l 1 ⫽
1 ⫺ t by
(n ⫹ 1)l n⫹1 ⫽ (2n ⫹ 1 ⫺ t)l n ⫺ nl nⴚ1.
(d) A generating function (definition in Problem Set
5.2) for the Laguerre polynomials is
ⴥ
n
ⴚ1 tx>(xⴚ1)
.
a l n(t)x ⫽ (1 ⫺ x) e
n⫽0
Obtain l 0, Á , l 10 from the corresponding partial sum
of this power series in x and compare the l n with those
in (a), (b), or (c).
13. CAS EXPERIMENT. Laguerre Polynomials. Experiment with the graphs of l 0, Á , l 10, finding out
empirically how the first maximum, first minimum, Á
is moving with respect to its location as a function of
n. Write a short report on this.
c06.qxd
10/28/10
6:33 PM
242
Page 242
CHAP. 6 Laplace Transforms
14–20
INVERSE TRANSFORMS
Using differentiation, integration, s-shifting, or convolution,
and showing the details, find f (t) if l( f ) equals:
s
14. 2
(s ⫹ 16)2
s
15. 2
(s ⫺ 9)2
6.7
2s ⫹ 6
16.
(s ⫹ 6s ⫹ 10)2
s
17. ln
s⫺1
2
19. ln
s2 ⫹ 1
(s ⫺ 1)
2
s
18. arccot p
20. ln
s⫹a
s⫹b
Systems of ODEs
The Laplace transform method may also be used for solving systems of ODEs, as we shall
explain in terms of typical applications. We consider a first-order linear system with
constant coefficients (as discussed in Sec. 4.1)
y1r ⫽ a11y1 ⫹ a12y2 ⫹ g1(t)
(1)
y2r ⫽ a21y1 ⫹ a22y2 ⫹ g2(t).
Writing Y1 ⫽ l( y1), Y2 ⫽ l( y2), G1 ⫽ l(g1), G2 ⫽ l(g2), we obtain from (1) in Sec. 6.2
the subsidiary system
˛
˛˛
sY1 ⫺ y1(0) ⫽ a11Y1 ⫹ a12Y2 ⫹ G1(s)
sY2 ⫺ y2(0) ⫽ a21Y1 ⫹ a22Y2 ⫹ G2(s).
By collecting the Y1- and Y2-terms we have
(2)
(a11 ⫺ s)Y1 ⫹
a21Y1
a12Y2
⫽ ⫺y1(0) ⫺ G1(s)
⫹ (a22 ⫺ s)Y2 ⫽ ⫺y2(0) ⫺ G2(s).
By solving this system algebraically for Y1(s),Y2(s) and taking the inverse transform we
obtain the solution y1 ⫽ lⴚ1(Y1), y2 ⫽ lⴚ1(Y2) of the given system (1).
Note that (1) and (2) may be written in vector form (and similarly for the systems in
the examples); thus, setting y ⫽ 3y1 y24T, A ⫽ 3ajk4, g ⫽ 3g1 g24T, Y ⫽ 3Y1 Y24T,
G ⫽ 3G1 G24T we have
y r ⫽ Ay ⫹ g
EXAMPLE 1
and
(A ⫺ sI)Y ⫽ ⫺y(0) ⫺ G.
Mixing Problem Involving Two Tanks
Tank T1 in Fig. 144 initially contains 100 gal of pure water. Tank T2 initially contains 100 gal of water in which
150 lb of salt are dissolved. The inflow into T1 is 2 gal>min from T2 and 6 gal>min containing 6 lb of salt from
the outside. The inflow into T2 is 8 gal/min from T1. The outflow from T2 is 2 ⫹ 6 ⫽ 8 gal>min, as shown in
the figure. The mixtures are kept uniform by stirring. Find and plot the salt contents y1(t) and y2(t) in T1 and
T2, respectively.
c06.qxd
10/30/10
1:52 AM
Page 243
SEC. 6.7 Systems of ODEs
Solution.
243
The model is obtained in the form of two equations
Time rate of change ⫽ Inflow>min ⫺ Outflow>min
for the two tanks (see Sec. 4.1). Thus,
8
2
y1r ⫽ ⫺ 100
y1 ⫹ 100
y2 ⫹ 6.
8
8
y2r ⫽ 100
y1 ⫺ 100
y2.
The initial conditions are y1(0) ⫽ 0, y2(0) ⫽ 150. From this we see that the subsidiary system (2) is
(⫺0.08 ⫺ s)Y1 ⫹
0.08Y1
⫽⫺
0.02Y2
6
s
⫹ (⫺0.08 ⫺ s)Y2 ⫽ ⫺150.
We solve this algebraically for Y1 and Y2 by elimination (or by Cramer’s rule in Sec. 7.7), and we write the
solutions in terms of partial fractions,
Y1 ⫽
Y2 ⫽
9s ⫹ 0.48
s(s ⫹ 0.12)(s ⫹ 0.04)
150s 2 ⫹ 12s ⫹ 0.48
s(s ⫹ 0.12)(s ⫹ 0.04)
⫽
100
⫽
100
s
s
62.5
⫺
⫹
⫺
s ⫹ 0.12
125
⫺
s ⫹ 0.12
37.5
s ⫹ 0.04
75
s ⫹ 0.04
.
By taking the inverse transform we arrive at the solution
y1 ⫽ 100 ⫺ 62.5eⴚ0.12t ⫺ 37.5eⴚ0.04t
y2 ⫽ 100 ⫹ 125eⴚ0.12t ⫺ 75eⴚ0.04t.
Figure 144 shows the interesting plot of these functions. Can you give physical explanations for their main
features? Why do they have the limit 100? Why is y2 not monotone, whereas y1 is? Why is y1 from some time
on suddenly larger than y2? Etc.
䊏
6 gal/min
y(t)
150
2 gal/min
Salt content in T2
100
T1
8 gal/min
T2
50
6 gal/min
Salt content in T1
50
100
150
200
t
Fig. 144. Mixing problem in Example 1
Other systems of ODEs of practical importance can be solved by the Laplace transform
method in a similar way, and eigenvalues and eigenvectors, as we had to determine them
in Chap. 4, will come out automatically, as we have seen in Example 1.
EXAMPLE 2
Electrical Network
Find the currents i 1(t) and i 2(t) in the network in Fig. 145 with L and R measured in terms of the usual units
(see Sec. 2.9), v(t) ⫽ 100 volts if 0 ⬉ t ⬉ 0.5 sec and 0 thereafter, and i(0) ⫽ 0, i r (0) ⫽ 0.
Solution.
The model of the network is obtained from Kirchhoff’s Voltage Law as in Sec. 2.9. For the lower
circuit we obtain
0.8i 1r ⫹ 1(i 1 ⫺ i 2) ⫹ 1.4i 1 ⫽ 100[1 ⫺ u(t ⫺ 12 )]
c06.qxd
10/28/10
244
6:33 PM
Page 244
CHAP. 6 Laplace Transforms
L2 = 1 H
i2
i(t)
30
R1 = 1 Ω
i1
L1 = 0.8 H
i1(t)
20
i2(t)
10
R2 = 1.4 Ω
0
0
v(t)
0.5
1
1.5 2
Currents
2.5
3
t
Network
Fig. 145. Electrical network in Example 2
and for the upper
1 # i 2r ⫹ 1(i 2 ⫺ i 1)
⫽ 0.
Division by 0.8 and ordering gives for the lower circuit
i 1r ⫹ 3i 1 ⫺ 1.25i 2 ⫽ 125[1 ⫺ u(t ⫺ 12 )]
and for the upper
i 2r ⫺ i 1 ⫹
i 2 ⫽ 0.
With i 1(0) ⫽ 0, i 2(0) ⫽ 0 we obtain from (1) in Sec. 6.2 and the second shifting theorem the subsidiary
system
1
eⴚs>2
(s ⫹ 3)I1 ⫺ 1.25I2 ⫽ 125 a ⫺
b
s
s
⫺I1 ⫹ (s ⫹ 1)I2 ⫽ 0.
Solving algebraically for I1 and I2 gives
I1 ⫽
I2 ⫽
125(s ⫹ 1)
s(s ⫹ 12 )(s ⫹ 72 )
125
s(s ⫹ 12 )(s ⫹ 72 )
(1 ⫺ eⴚs>2),
(1 ⫺ eⴚs>2).
The right sides, without the factor 1 ⫺ eⴚs>2, have the partial fraction expansions
500
7s
⫺
125
3(s ⫹
1
2)
⫺
625
21(s ⫹ 72 )
and
500
7s
⫺
250
3(s ⫹
1
2)
⫹
250
21(s ⫹ 72 )
,
respectively. The inverse transform of this gives the solution for 0 ⬉ t ⬉ 12 ,
ⴚt>2
ⴚ7t>2
i 1(t) ⫽ ⫺ 125
⫺ 625
⫹ 500
3 e
21 e
7
ⴚt>2
ⴚ7t>2
i 2(t) ⫽ ⫺ 250
⫹ 250
⫹ 500
3 e
21 e
7
(0 ⬉ t ⬉ 12 ).
c06.qxd
10/28/10
6:33 PM
Page 245
SEC. 6.7 Systems of ODEs
245
According to the second shifting theorem the solution for t ⬎
1
2
is i 1(t) ⫺ i 1(t ⫺ 12 ) and i 2(t) ⫺ i 2(t ⫺ 12 ), that is,
1>4 ⴚt>2
7>4 ⴚ7t>2
i 1(t) ⫽ ⫺ 125
)e
⫺ 625
)e
3 (1 ⫺ e
21 (1 ⫺ e
1>4 ⴚt>2
7>4 ⴚ7t>2
i 2(t) ⫽ ⫺ 250
)e
⫹ 250
)e
3 (1 ⫺ e
21 (1 ⫺ e
(t ⬎ 12 ).
Can you explain physically why both currents eventually go to zero, and why i 1(t) has a sharp cusp whereas
i 2(t) has a continuous tangent direction at t ⫽ 12?
䊏
Systems of ODEs of higher order can be solved by the Laplace transform method in a
similar fashion. As an important application, typical of many similar mechanical systems,
we consider coupled vibrating masses on springs.
k
m1 = 1
0
y1
k
m2 = 1
0
y2
k
Fig. 146. Example 3
EXAMPLE 3
Model of Two Masses on Springs (Fig. 146)
The mechanical system in Fig. 146 consists of two bodies of mass 1 on three springs of the same spring constant
k and of negligibly small masses of the springs. Also damping is assumed to be practically zero. Then the model
of the physical system is the system of ODEs
y s1 ⫽ ⫺ky1 ⫹ k(y2 ⫺ y1)
(3)
y s2 ⫽ ⫺k(y2 ⫺ y1) ⫺ ky2.
Here y1 and y2 are the displacements of the bodies from their positions of static equilibrium. These ODEs
follow from Newton’s second law, Mass ⫻ Acceleration ⫽ Force, as in Sec. 2.4 for a single body. We again
regard downward forces as positive and upward as negative. On the upper body, ⫺ky1 is the force of the
upper spring and k(y2 ⫺ y1) that of the middle spring, y2 ⫺ y1 being the net change in spring length—think
this over before going on. On the lower body, ⫺k(y2 ⫺ y1) is the force of the middle spring and ⫺ky2 that
of the lower spring.
We shall determine the solution corresponding to the initial conditions y1(0) ⫽ 1, y2(0) ⫽ 1, y1r (0) ⫽ 23k,
y r2(0) ⫽ ⫺ 23k. Let Y1 ⫽ l(y1) and Y2 ⫽ l(y2). Then from (2) in Sec. 6.2 and the initial conditions we obtain
the subsidiary system
s 2Y1 ⫺ s ⫺ 23k ⫽ ⫺kY1 ⫹ k(Y2 ⫺ Y1)
s 2Y2 ⫺ s ⫹ 23k ⫽ ⫺k(Y2 ⫺ Y1) ⫺ kY2.
This system of linear algebraic equations in the unknowns Y1 and Y2 may be written
(s 2 ⫹ 2k)Y1 ⫺
⫺ky1
kY2
⫽ s ⫹ 23k
⫹ (s ⫹ 2k)Y2 ⫽ s ⫺ 23k.
2
c06.qxd
10/28/10
246
6:33 PM
Page 246
CHAP. 6 Laplace Transforms
Elimination (or Cramer’s rule in Sec. 7.7) yields the solution, which we can expand in terms of partial fractions,
Y1 ⫽
(s ⫹ 23k)(s 2 ⫹ 2k) ⫹ k(s ⫺ 23k)
(s ⫹ 2k) ⫺ k
2
2
2
(s ⫹ 2k)(s ⫺ 23k) ⫹ k(s ⫹ 23k)
⫽
s
s ⫹k
2
2
Y2 ⫽
(s 2 ⫹ 2k) 2 ⫺ k 2
⫽
s
s2 ⫹ k
⫹
⫺
23k
s ⫹ 3k
2
23k
s 2 ⫹ 3k
.
Hence the solution of our initial value problem is (Fig. 147)
y1(t) ⫽ lⴚ1(Y1) ⫽ cos 2kt ⫹ sin 23kt
y2(t) ⫽ lⴚ1(Y2) ⫽ cos 2kt ⫺ sin 23kt.
We see that the motion of each mass is harmonic (the system is undamped!), being the superposition of a “slow”
oscillation and a “rapid” oscillation.
䊏
2
y1(t)
y2(t)
1
2π
0
4π
t
–1
–2
Fig. 147. Solutions in Example 3
PROBLEM SET 6.7
1. TEAM PROJECT. Comparison of Methods for
Linear Systems of ODEs
(a) Models. Solve the models in Examples 1 and 2 of
Sec. 4.1 by Laplace transforms and compare the amount
of work with that in Sec. 4.1. Show the details of your
work.
(b) Homogeneous Systems. Solve the systems (8),
(11)–(13) in Sec. 4.3 by Laplace transforms. Show the
details.
(c) Nonhomogeneous System. Solve the system (3) in
Sec. 4.6 by Laplace transforms. Show the details.
2–15
SYSTEMS OF ODES
Using the Laplace transform and showing the details of
your work, solve the IVP:
2. y1r ⫹ y2 ⫽ 0, y1 ⫹ y2r ⫽ 2 cos t,
y1(0) ⫽ 1, y2(0) ⫽ 0
3. y1r ⫽ ⫺y1 ⫹ 4y2, y2r ⫽ 3y1 ⫺ 2y2,
y1(0) ⫽ 3, y2(0) ⫽ 4
4. y1r ⫽ 4y2 ⫺ 8 cos 4t, y2r ⫽ ⫺3y1 ⫺ 9 sin 4t,
y1(0) ⫽ 0, y2(0) ⫽ 3
5. y1r ⫽ y2 ⫹ 1 ⫺ u(t ⫺ 1), y2r ⫽ ⫺y1 ⫹ 1 ⫺ u(t ⫺ 1),
y1(0) ⫽ 0, y2(0) ⫽ 0
6. y1r ⫽ 5y1 ⫹ y2, y2r ⫽ y1 ⫹ 5y2,
y1(0) ⫽ 1, y2(0) ⫽ ⫺3
7. y1r ⫽ 2y1 ⫺ 4y2 ⫹ u(t ⫺ 1)et,
y2r ⫽ y1 ⫺ 3y2 ⫹ u(t ⫺ 1)et, y1(0) ⫽ 3, y2(0) ⫽ 0
8. y1r ⫽ ⫺2y1 ⫹ 3y2, y2r ⫽ 4y1 ⫺ y2,
y1(0) ⫽ 4, y2(0) ⫽ 3
9. y1r ⫽ 4y1 ⫹ y2,
y2(0) ⫽ 1
y2r ⫽ ⫺y1 ⫹ 2y2, y1(0) ⫽ 3,
10. y1r ⫽ ⫺y2, y2r ⫽ ⫺y1 ⫹ 2[1 ⫺ u(t ⫺ 2p)] cos t,
y1(0) ⫽ 1, y2(0) ⫽ 0
11. y1s ⫽ y1 ⫹ 3y2, y2s ⫽ 4y1 ⫺ 4et,
y1(0) ⫽ 2, y1r (0) ⫽ 3, y2(0) ⫽ 1,
y2r (0) ⫽ 2
12. y1s ⫽ ⫺2y1 ⫹ 2y2, y2s ⫽ 2y1 ⫺ 5y2,
y1(0) ⫽ 1, y1r (0) ⫽ 0, y2(0) ⫽ 3, y2r (0) ⫽ 0
13. y1s ⫹ y2 ⫽ ⫺101 sin 10t, y2s ⫹ y1 ⫽ 101 sin 10t,
y1(0) ⫽ 0, y1r (0) ⫽ 6, y2(0) ⫽ 8, y2r (0) ⫽ ⫺6
c06.qxd
10/28/10
6:33 PM
Page 247
SEC. 6.7 Systems of ODEs
14. 4y1r ⫹ y2r ⫺ 2y3r ⫽ 0, ⫺2y1r ⫹ y3r ⫽ 1,
2y2r ⫺ 4y3r ⫽ ⫺16t
y1(0) ⫽ 2, y2(0) ⫽ 0, y3(0) ⫽ 0
15. y1r ⫹ y2r ⫽ 2 sinh t, y2r ⫹ y3r ⫽ et,
y3r ⫹ y1r ⫽ 2et ⫹ eⴚt, y1(0) ⫽ 1, y2(0) ⫽ 1,
y3(0) ⫽ 0
247
will the currents practically reach their steady state?
4Ω
8Ω
i1
i2
8Ω
v(t)
FURTHER APPLICATIONS
16. Forced vibrations of two masses. Solve the model in
Example 3 with k ⫽ 4 and initial conditions y1(0) ⫽ 1,
y1r (0) ⫽ 1, y2(0) ⫽ 1, y2r ⫽ ⫺1 under the assumption
that the force 11 sin t is acting on the first body and the
force ⫺11 sin t on the second. Graph the two curves
on common axes and explain the motion physically.
17. CAS Experiment. Effect of Initial Conditions. In
Prob. 16, vary the initial conditions systematically,
describe and explain the graphs physically. The great
variety of curves will surprise you. Are they always
periodic? Can you find empirical laws for the changes
in terms of continuous changes of those conditions?
18. Mixing problem. What will happen in Example 1 if
you double all flows (in particular, an increase to
12 gal>min containing 12 lb of salt from the outside),
leaving the size of the tanks and the initial conditions
as before? First guess, then calculate. Can you relate
the new solution to the old one?
19. Electrical network. Using Laplace transforms,
find the currents i 1(t) and i 2(t) in Fig. 148, where
v(t) ⫽ 390 cos t and i 1(0) ⫽ 0, i 2(0) ⫽ 0. How soon
2H
4H
Network
i(t)
40
i1(t)
20
i2(t)
0
2
4
6
8
10
t
–20
–40
Currents
Fig. 148. Electrical network and
currents in Problem 19
20. Single cosine wave. Solve Prob. 19 when the EMF
(electromotive force) is acting from 0 to 2p only. Can
you do this just by looking at Prob. 19, practically
without calculation?
c06.qxd
10/28/10
248
6.8
6:33 PM
Page 248
CHAP. 6 Laplace Transforms
Laplace Transform: General Formulas
Formula
冮
F(s) ⫽ l{ f (t)} ⫽
Name, Comments
Sec.
ⴥ
eⴚstf (t) dt
Definition of Transform
0
6.1
f (t) ⫽ lⴚ1{F(s)}
Inverse Transform
l{af (t) ⫹ bg(t)} ⫽ al{ f (t)} ⫹ bl{g(t)}
Linearity
6.1
s-Shifting
(First Shifting Theorem)
6.1
l{eatf (t)} ⫽ F(s ⫺ a)
lⴚ1{F(s ⫺ a)} ⫽ eatf (t)
l( f r ) ⫽ sl( f ) ⫺ f (0)
l( f s ) ⫽ s 2l( f ) ⫺ sf (0) ⫺ f r (0)
Differentiation of Function
l( f (n)) ⫽ s nl( f ) ⫺ s (nⴚ1)f (0) ⫺ Á
Á ⫺f
le
6.2
(nⴚ1)
(0)
t
冮 f (t) dtf ⫽ 1s l( f )
Integration of Function
0
t
( f * g)(t) ⫽
冮 f (t)g(t ⫺ t) dt
0
t
⫽
冮 f (t ⫺ t)g(t) dt
Convolution
6.5
t-Shifting
(Second Shifting Theorem)
6.3
0
l( f * g) ⫽ l( f )l(g)
l{ f (t ⫺ a) u(t ⫺ a)} ⫽ eⴚasF(s)
˛
ⴚ1
l
{eⴚasF (s)} ⫽ f (t ⫺ a) u(t ⫺ a)
l{tf (t)} ⫽ ⫺F r (s)
le
f (t)
l( f ) ⫽
t
f ⫽
冮
Differentiation of Transform
ⴥ
F( 苲
s ) d苲
s
6.6
Integration of Transform
s
1
1 ⫺ eⴚps
冮
0
p
eⴚstf (t) dt
f Periodic with Period p
6.4
Project
16
c06.qxd
10/28/10
6:33 PM
Page 249
SEC. 6.9 Table of Laplace Transforms
6.9
249
Table of Laplace Transforms
For more extensive tables, see Ref. [A9] in Appendix 1.
F (s) ⫽ l{ f (t)}
f (t)
˛
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
1>s
1>s 2
1>s n
1> 1s
1>s 3>2
1>s a
(n ⫽ 1, 2, Á )
(a ⬎ 0)
1
s⫺a
1
teat
(s ⫺ a)
n
1
(s ⫺ a)
k
(n ⫽ 1, 2, Á )
1
t nⴚ1eat
(n ⫺ 1)!
(k ⬎ 0)
1 kⴚ1 at
t
e
⌫(k)
1
(s ⫺ a)(s ⫺ b)
s
(s ⫺ a)(s ⫺ b)
1
s ⫹v
s
(a ⫽ b)
1
(eat ⫺ ebt)
a⫺b
1
(aeat ⫺ bebt)
a⫺b
cos vt
s 2 ⫹ v2
1
s ⫺a
s
(a ⫽ b)
1
sinh at
a
2
cosh at
s2 ⫺ a2
1
(s ⫺ a)2 ⫹ v2
s⫺a
(s ⫺ a) ⫹ v
2
2
eat cos vt
s(s ⫹ v )
v2
1
1
s 2(s 2 ⫹ v2)
v3
2
t 6.1
1 at
e sinh vt
v
1
1
2
t 6.1
1
sin vt
v
2
2
t 6.1
eat
(s ⫺ a)2
1
2
1
t
t nⴚ1>(n ⫺ 1)!
1> 1pt
2 1t> p
t aⴚ1>⌫(a)
Sec.
(1 ⫺ cos vt)
x 6.2
(vt ⫺ sin vt)
(continued )
c06.qxd
10/28/10
250
6:33 PM
Page 250
CHAP. 6 Laplace Transforms
Table of Laplace Transforms (continued )
F (s) ⫽ l{ f (t)}
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
1
(sin vt ⫺ vt cos vt)
2v3
t
sin vt
2v
(s ⫹ v )
s
2 2
(s 2 ⫹ v2) 2
s2
2 2
(s 2 ⫹ a 2)(s 2 ⫹ b 2)
(a 2 ⫽ b 2)
1
1
b 2 ⫺ a2
1
s ⫹ 4k
s
4
4
4k 3
1
s 4 ⫹ 4k 4
1
2k 2
1
s4 ⫺ k 4
s
2k 3
1
s4 ⫺ k 4
2k 2
1s ⫺ a ⫺ 1s ⫺ b
1
1s ⫹ a 1s ⫹ b
1
s
(k ⬎ 0)
1 ⴚk>s
e
s
1 ⴚk>s
e
1s
1
(sinh kt ⫺ sin kt)
(cosh kt ⫺ cos kt)
1
22pt 3
(ebt ⫺ eat)
eⴚ(a⫹b)t>2I0 a
a⫺b
tb
2
eat(1 ⫹ 2at)
kⴚ1>2
Ikⴚ1>2(at)
I 5.5
u(t ⫺ a)
d(t ⫺ a)
6.3
6.4
J0(2 1kt)
J 5.4
1pt
1
1pk
(k ⬎ 0)
I 5.5
J 5.4
1p t
a b
⌫(k) 2a
1
ek>s
eⴚk1s
sin kt sinh kt
1pt
eⴚas>s
eⴚas
3>2
(sin kt cos kt ⫺ cos kt sinh kt)
1
3>2
(s 2 ⫺ a 2)k
(cos at ⫺ cos bt)
J0(at)
2s ⫹ a 2
2
(s ⫺ a)
1
t 6.6
1
(sin vt ⫹ vt cos vt)
2v
(s ⫹ v )
s
2
s
39
1
2
38
Sec.
f (t)
˛
cos 2 1kt
sinh 2 1kt
k
22pt
eⴚk
>4t
2
3
(continued )
c06.qxd
10/28/10
6:33 PM
Page 251
Chapter 6 Review Questions and Problems
251
Table of Laplace Transforms (continued )
F (s) ⫽ l{ f (t)}
f (t)
˛
40
1
ln s
s
41
ln
42
ln
43
ln
Sec.
⫺ln t ⫺ g (g ⬇ 0.5772)
s⫺a
s⫺b
1 bt
(e ⫺ eat)
t
s 2 ⫹ v2
2
(1 ⫺ cos vt)
t
s2
s2 ⫺ a2
s
2
2
(1 ⫺ cosh at)
t
v
s
1
sin vt
t
44
arctan
45
1
arccot s
s
g 5.5
6.6
App.
A3.1
Si(t)
CHAPTER 6 REVIEW QUESTIONS AND PROBLEMS
1. State the Laplace transforms of a few simple functions
from memory.
2. What are the steps of solving an ODE by the Laplace
transform?
3. In what cases of solving ODEs is the present method
preferable to that in Chap. 2?
4. What property of the Laplace transform is crucial in
solving ODEs?
5. Is l{ f (t) ⫹ g(t)} ⫽ l{ f (t)} ⫹ l{g(t)}?
l{ f (t)g(t)} ⫽ l{ f (t)}l{g(t)}? Explain.
6. When and how do you use the unit step function and
Dirac’s delta?
7. If you know f (t) ⫽ lⴚ1{F(s)}, how would you find
lⴚ1{F(s)>s 2 } ?
8. Explain the use of the two shifting theorems from memory.
9. Can a discontinuous function have a Laplace transform?
Give reason.
10. If two different continuous functions have transforms,
the latter are different. Why is this practically important?
11–19
LAPLACE TRANSFORMS
Find the transform, indicating the method used and showing
the details.
11. 5 cosh 2t ⫺ 3 sinh t
12. eⴚt(cos 4t ⫺ 2 sin 4t)
1
13. sin2 (2pt)
14. 16t 2u(t ⫺ 14)
15. et>2u(t ⫺ 3)
17. t cos t ⫹ sin t
19. 12t * eⴚ3t
16. u(t ⫺ 2p) sin t
18. (sin vt) * (cos vt)
20–28
INVERSE LAPLACE TRANSFORM
Find the inverse transform, indicating the method used and
showing the details:
7.5
s ⫹ 1 ⴚs
20. 2
21.
e
s ⫺ 2s ⫺ 8
s2
22.
24.
1
16
1
2
s ⫹s⫹
s 2 ⫺ 6.25
2
(s 2 ⫹ 6.25)2
2s ⫺ 10 ⴚ5s
26.
e
s3
3s
28. 2
s ⫺ 2s ⫹ 2
23.
25.
27.
v cos u ⫹ s sin u
s 2 ⫹ v2
6(s ⫹ 1)
s4
3s ⫹ 4
s 2 ⫹ 4s ⫹ 5
29–37
ODEs AND SYSTEMS
Solve by the Laplace transform, showing the details and
graphing the solution:
29. y s ⫹ 4y r ⫹ 5y ⫽ 50t, y(0) ⫽ 5, y r (0) ⫽ ⫺5
30. y s ⫹ 16y ⫽ 4d(t ⫺ p), y(0) ⫽ ⫺1, y r (0) ⫽ 0
c11-a.qxd
10/30/10
1:24 PM
Page 474
CHAPTER
11
Fourier Analysis
This chapter on Fourier analysis covers three broad areas: Fourier series in Secs. 11.1–11.4,
more general orthonormal series called Sturm–Liouville expansions in Secs. 11.5 and 11.6
and Fourier integrals and transforms in Secs. 11.7–11.9.
The central starting point of Fourier analysis is Fourier series. They are infinite series
designed to represent general periodic functions in terms of simple ones, namely, cosines
and sines. This trigonometric system is orthogonal, allowing the computation of the
coefficients of the Fourier series by use of the well-known Euler formulas, as shown in
Sec. 11.1. Fourier series are very important to the engineer and physicist because they
allow the solution of ODEs in connection with forced oscillations (Sec. 11.3) and the
approximation of periodic functions (Sec. 11.4). Moreover, applications of Fourier analysis
to PDEs are given in Chap. 12. Fourier series are, in a certain sense, more universal than
the familiar Taylor series in calculus because many discontinuous periodic functions that
come up in applications can be developed in Fourier series but do not have Taylor series
expansions.
The underlying idea of the Fourier series can be extended in two important ways. We
can replace the trigonometric system by other families of orthogonal functions, e.g., Bessel
functions and obtain the Sturm–Liouville expansions. Note that related Secs. 11.5 and
11.6 used to be part of Chap. 5 but, for greater readability and logical coherence, are now
part of Chap. 11. The second expansion is applying Fourier series to nonperiodic
phenomena and obtaining Fourier integrals and Fourier transforms. Both extensions have
important applications to solving PDEs as will be shown in Chap. 12.
In a digital age, the discrete Fourier transform plays an important role. Signals, such
as voice or music, are sampled and analyzed for frequencies. An important algorithm, in
this context, is the fast Fourier transform. This is discussed in Sec. 11.9.
Note that the two extensions of Fourier series are independent of each other and may
be studied in the order suggested in this chapter or by studying Fourier integrals and
transforms first and then Sturm–Liouville expansions.
Prerequisite: Elementary integral calculus (needed for Fourier coefficients).
Sections that may be omitted in a shorter course: 11.4–11.9.
References and Answers to Problems: App. 1 Part C, App. 2.
11.1
Fourier Series
Fourier series are infinite series that represent periodic functions in terms of cosines and
sines. As such, Fourier series are of greatest importance to the engineer and applied
mathematician. To define Fourier series, we first need some background material.
A function f (x) is called a periodic function if f ( x) is defined for all real x, except
474
c11-a.qxd
10/30/10
1:24 PM
Page 475
SEC. 11.1 Fourier Series
475
f (x)
x
p
Fig. 258. Periodic function of period p
possibly at some points, and if there is some positive number p, called a period of f (x),
such that
f (x ⫹ p) ⫽ f (x)
(1)
for all x.
(The function f (x) ⫽ tan x is a periodic function that is not defined for all real x but
undefined for some points (more precisely, countably many points), that is x ⫽ ⫾p>2,
⫾3p>2, Á .)
The graph of a periodic function has the characteristic that it can be obtained by periodic
repetition of its graph in any interval of length p (Fig. 258).
The smallest positive period is often called the fundamental period. (See Probs. 2–4.)
Familiar periodic functions are the cosine, sine, tangent, and cotangent. Examples of
functions that are not periodic are x, x 2, x 3, ex, cosh x, and ln x, to mention just a few.
If f (x) has period p, it also has the period 2p because (1) implies f (x ⫹ 2p) ⫽
f ([x ⫹ p] ⫹ p) ⫽ f (x ⫹ p) ⫽ f (x), etc.; thus for any integer n ⫽ 1, 2, 3, Á ,
f (x ⫹ np) ⫽ f (x)
(2)
for all x.
Furthermore if f (x) and g (x) have period p, then af (x) ⫹ bg (x) with any constants a and
b also has the period p.
Our problem in the first few sections of this chapter will be the representation of various
functions f (x) of period 2p in terms of the simple functions
(3)
1,
cos x,
sin x,
cos 2x,
sin 2x, Á ,
cos nx,
sin nx, Á .
All these functions have the period 2p. They form the so-called trigonometric system.
Figure 259 shows the first few of them (except for the constant 1, which is periodic with
any period).
0
π
2π
π
0
cos x
0
π
sin x
π
2π
π
0
cos 2x
2π
π
0
π
sin 2x
π
2π
π
cos 3x
2π
π
0
π
2π
π
sin 3x
Fig. 259. Cosine and sine functions having the period 2p (the first few members of the
trigonometric system (3), except for the constant 1)
c11-a.qxd
10/30/10
476
1:24 PM
Page 476
CHAP. 11 Fourier Analysis
The series to be obtained will be a trigonometric series, that is, a series of the form
a0 ⫹ a1 cos x ⫹ b1 sin x ⫹ a2 cos 2x ⫹ b2 sin 2x ⫹ Á
ⴥ
(4)
⫽ a0 ⫹ a (an cos nx ⫹ bn sin nx).
n⫽1
a0, a1, b1, a2, b2, Á are constants, called the coefficients of the series. We see that each
term has the period 2p. Hence if the coefficients are such that the series converges, its
sum will be a function of period 2p.
Expressions such as (4) will occur frequently in Fourier analysis. To compare the
expression on the right with that on the left, simply write the terms in the summation.
Convergence of one side implies convergence of the other and the sums will be the
same.
Now suppose that f (x) is a given function of period 2p and is such that it can be
represented by a series (4), that is, (4) converges and, moreover, has the sum f (x). Then,
using the equality sign, we write
ⴥ
(5)
f (x) ⫽ a0 ⫹ a (an cos nx ⫹ bn sin nx)
n⫽1
and call (5) the Fourier series of f (x). We shall prove that in this case the coefficients
of (5) are the so-called Fourier coefficients of f (x), given by the Euler formulas
(0)
(6)
(a)
(b)
a0 ⫽
1
2p
1
an ⫽ p
1
bn ⫽ p
冮
p
f (x) dx
ⴚp
冮
p
冮
p
f (x) cos nx dx
n ⫽ 1, 2, Á
f (x) sin nx dx
n ⫽ 1, 2, Á .
ⴚp
ⴚp
The name “Fourier series” is sometimes also used in the exceptional case that (5) with
coefficients (6) does not converge or does not have the sum f (x)—this may happen but
is merely of theoretical interest. (For Euler see footnote 4 in Sec. 2.5.)
A Basic Example
Before we derive the Euler formulas (6), let us consider how (5) and (6) are applied in
this important basic example. Be fully alert, as the way we approach and solve this
example will be the technique you will use for other functions. Note that the integration
is a little bit different from what you are familiar with in calculus because of the n. Do
not just routinely use your software but try to get a good understanding and make
observations: How are continuous functions (cosines and sines) able to represent a given
discontinuous function? How does the quality of the approximation increase if you take
more and more terms of the series? Why are the approximating functions, called the
c11-a.qxd
10/30/10
1:24 PM
Page 477
SEC. 11.1 Fourier Series
477
partial sums of the series, in this example always zero at 0 and p? Why is the factor
1>n (obtained in the integration) important?
EXAMPLE 1
Periodic Rectangular Wave (Fig. 260)
Find the Fourier coefficients of the periodic function f (x) in Fig. 260. The formula is
(7)
⫺k
if
⫺p ⬍ x ⬍ 0
k
if
0⬍x⬍p
f (x) ⫽ b
and
f (x ⫹ 2p) ⫽ f (x).
Functions of this kind occur as external forces acting on mechanical systems, electromotive forces in electric
circuits, etc. (The value of f (x) at a single point does not affect the integral; hence we can leave f (x) undefined
at x ⫽ 0 and x ⫽ ⫾p.)
From (6.0) we obtain a0 ⫽ 0. This can also be seen without integration, since the area under the
curve of f (x) between ⫺p and p (taken with a minus sign where f (x) is negative) is zero. From (6a) we obtain
the coefficients a1, a2, Á of the cosine terms. Since f ( x) is given by two expressions, the integrals from ⫺p
to p split into two integrals:
Solution.
an ⫽
p冮
1
p
f (x) cos nx dx ⫽
ⴚp
⫽
1
p c
冮
0
(⫺k) cos nx dx ⫹
ⴚp
1
p c ⫺k
冮
p
k cos nx dx d
0
p
sin nx 0
sin nx
` ⫹k
` d ⫽0
n
n
ⴚp
0
because sin nx ⫽ 0 at ⫺p, 0, and p for all n ⫽ 1, 2, Á . We see that all these cosine coefficients are zero. That
is, the Fourier series of (7) has no cosine terms, just sine terms, it is a Fourier sine series with coefficients
b1, b2, Á obtained from (6b);
bn ⫽
1
p
冮
p
f (x) sin nx dx ⫽
ⴚp
⫽
1
pc
1
冮
0
(⫺k) sin nx dx ⫹
ⴚp
p ck
冮
p
0
k sin nx dx d
cos nx 0
cos nx p
` ⫺k
` d.
n
n
ⴚp
0
Since cos (⫺a) ⫽ cos a and cos 0 ⫽ 1, this yields
bn ⫽
k
2k
[cos 0 ⫺ cos (⫺np) ⫺ cos np ⫹ cos 0] ⫽
(1 ⫺ cos np).
np
np
Now, cos p ⫽ ⫺1, cos 2p ⫽ 1, cos 3p ⫽ ⫺1, etc.; in general,
⫺1
cos np ⫽ b
1
for odd n,
and thus
for even n,
1 ⫺ cos np ⫽ b
2
for odd n,
0
for even n.
Hence the Fourier coefficients bn of our function are
b1 ⫽
4k
p
,
b2 ⫽ 0,
b3 ⫽
4k
3p
,
b4 ⫽ 0,
b5 ⫽
4k Á
, .
5p
Fig. 260. Given function f (x) (Periodic reactangular wave)
c11-a.qxd
10/30/10
478
1:24 PM
Page 478
CHAP. 11 Fourier Analysis
Since the an are zero, the Fourier series of f (x) is
4k
(8)
p
(sin x ⫹ 13 sin 3x ⫹ 15 sin 5x ⫹ Á ).
The partial sums are
S1 ⫽
4k
p
S2 ⫽
sin x,
4k
p
asin x ⫹
1
sin 3xb .
3
etc.
Their graphs in Fig. 261 seem to indicate that the series is convergent and has the sum f (x), the given function.
We notice that at x ⫽ 0 and x ⫽ p, the points of discontinuity of f (x), all partial sums have the value zero, the
arithmetic mean of the limits ⫺k and k of our function, at these points. This is typical.
Furthermore, assuming that f (x) is the sum of the series and setting x ⫽ p>2, we have
p
4k
1
1
Á
fa b⫽k⫽
2
p a1 ⫺ 3 ⫹ 5 ⫺ ⫹ b .
Thus
1⫺
1
3
⫹
1
5
⫺
1
7
p
⫹⫺ Á ⫽ .
4
This is a famous result obtained by Leibniz in 1673 from geometric considerations. It illustrates that the values
of various series with constant terms can be obtained by evaluating Fourier series at specific points.
䊏
Fig. 261. First three partial sums of the corresponding Fourier series
c11-a.qxd
10/30/10
1:24 PM
Page 479
SEC. 11.1 Fourier Series
479
Derivation of the Euler Formulas (6)
The key to the Euler formulas (6) is the orthogonality of (3), a concept of basic importance,
as follows. Here we generalize the concept of inner product (Sec. 9.3) to functions.
THEOREM 1
Orthogonality of the Trigonometric System (3)
The trigonometric system (3) is orthogonal on the interval ⫺p ⬉ x ⬉ p (hence
also on 0 ⬉ x ⬉ 2p or any other interval of length 2p because of periodicity); that
is, the integral of the product of any two functions in (3) over that interval is 0, so
that for any integers n and m,
(a)
冮
p
冮
p
cos nx cos mx dx ⫽ 0
(n ⫽ m)
ⴚp
(9)
(b)
sin nx sin mx dx ⫽ 0
(n ⫽ m)
ⴚp
(c)
冮
p
sin nx cos mx dx ⫽ 0
(n ⫽ m or n ⫽ m).
ⴚp
PROOF
This follows simply by transforming the integrands trigonometrically from products into
sums. In (9a) and (9b), by (11) in App. A3.1,
冮
p
冮
p
1
2
cos nx cos mx dx ⫽
ⴚp
sin nx sin mx dx ⫽
ⴚp
1
2
冮
p
1
2
cos (n ⫹ m)x dx ⫹
ⴚp
冮
p
cos (n ⫺ m)x dx ⫺
ⴚp
1
2
冮
p
cos (n ⫺ m)x dx
ⴚp
冮
p
cos (n ⫹ m)x dx.
ⴚp
Since m ⫽ n (integer!), the integrals on the right are all 0. Similarly, in (9c), for all integer
m and n (without exception; do you see why?)
冮
p
1
sin nx cos mx dx ⫽
2
ⴚp
冮
p
1
sin (n ⫹ m)x dx ⫹
2
ⴚp
冮
p
sin (n ⫺ m)x dx ⫽ 0 ⫹ 0.
䊏
ⴚp
Application of Theorem 1 to the Fourier Series (5)
We prove (6.0). Integrating on both sides of (5) from ⫺p to p, we get
冮
p
f (x) dx ⫽
ⴚp
冮
p
ⴚp
ⴥ
c a0 ⫹ a (an cos nx ⫹ bn sin nx) d dx.
n⫽1
We now assume that termwise integration is allowed. (We shall say in the proof of
Theorem 2 when this is true.) Then we obtain
冮
p
ⴚp
f (x) dx ⫽ a0
冮
p
ⴥ
dx ⫹ a aan
ⴚp
n⫽1
冮
p
cos nx dx ⫹ bn
ⴚp
冮
p
sin nx dxb .
ⴚp
c11-a.qxd
10/30/10
1:24 PM
480
Page 480
CHAP. 11 Fourier Analysis
The first term on the right equals 2pa0. Integration shows that all the other integrals are 0.
Hence division by 2p gives (6.0).
We prove (6a). Multiplying (5) on both sides by cos mx with any fixed positive integer
m and integrating from ⫺p to p, we have
(10)
冮
p
f (x) cos mx dx ⫽
ⴚp
冮
p
ⴚp
ⴥ
c a0 ⫹ a (an cos nx ⫹ bn sin nx) d cos mx dx.
n⫽1
We now integrate term by term. Then on the right we obtain an integral of a0 cos mx,
which is 0; an integral of an cos nx cos mx , which is amp for n ⫽ m and 0 for n ⫽ m by
(9a); and an integral of bn sin nx cos mx, which is 0 for all n and m by (9c). Hence the
right side of (10) equals amp. Division by p gives (6a) (with m instead of n).
We finally prove (6b). Multiplying (5) on both sides by sin mx with any fixed positive
integer m and integrating from ⫺p to p, we get
(11)
冮
p
ⴚp
f (x) sin mx dx ⫽
冮
p
ⴚp
ⴥ
c a0 ⫹ a (an cos nx ⫹ bn sin nx) d sin mx dx.
n⫽1
Integrating term by term, we obtain on the right an integral of a0 sin mx, which is 0; an
integral of an cos nx sin mx, which is 0 by (9c); and an integral of bn sin nx sin mx, which
is bmp if n ⫽ m and 0 if n ⫽ m, by (9b). This implies (6b) (with n denoted by m). This
completes the proof of the Euler formulas (6) for the Fourier coefficients.
䊏
Convergence and Sum of a Fourier Series
The class of functions that can be represented by Fourier series is surprisingly large and
general. Sufficient conditions valid in most applications are as follows.
THEOREM 2
Representation by a Fourier Series
Let f (x) be periodic with period 2p and piecewise continuous (see Sec. 6.1) in the
interval ⫺p ⬉ x ⬉ p. Furthermore, let f (x) have a left-hand derivative and a righthand derivative at each point of that interval. Then the Fourier series (5) of f (x)
[with coefficients (6)] converges. Its sum is f (x), except at points x0 where f (x) is
discontinuous. There the sum of the series is the average of the left- and right-hand
limits2 of f (x) at x 0.
f (x)
f (1 – 0)
2
The left-hand limit of f (x) at x 0 is defined as the limit of f (x) as x approaches x0 from the left
and is commonly denoted by f (x 0 ⫺ 0). Thus
1
f (1 + 0)
0
x
1
Fig. 262. Left- and
right-hand limits
ƒ(1 ⫺ 0) ⫽ 1,
ƒ(1 ⫹ 0) ⫽ _1
2
of the function
f (x) ⫽ b
x
2
x>2
if x ⬍ 1
if x ⭌ 1
ƒ(x0 ⫺ 0) ⫽ lim ƒ( x0 ⫺ h) as h * 0 through positive values.
h*0
The right-hand limit is denoted by ƒ(x0 ⫹ 0) and
ƒ(x0 ⫹ 0) ⫽ lim ƒ( x0 ⫹ h) as h * 0 through positive values.
h*0
The left- and right-hand derivatives of ƒ(x) at x0 are defined as the limits of
f (x 0 ⫺ h) ⫺ f (x 0 ⫺ 0)
⫺h
and
f (x 0 ⫹ h) ⫺ f (x 0 ⫹ 0)
⫺h
,
respectively, as h * 0 through positive values. Of course if ƒ(x) is continuous at x0, the last term in
both numerators is simply ƒ(x0).
c11-a.qxd
10/30/10
1:24 PM
Page 481
SEC. 11.1 Fourier Series
PROOF
481
We prove convergence, but only for a continuous function f (x) having continuous first
and second derivatives. And we do not prove that the sum of the series is f (x) because
these proofs are much more advanced; see, for instance, Ref. 3C124 listed in App. 1.
Integrating (6a) by parts, we obtain
1
an ⫽ p
冮
p
f (x) sin nx
2
np
f (x) cos nx dx ⫽
ⴚp
p
ⴚp
冮
1
⫺ np
p
f r (x) sin nx dx.
ⴚp
The first term on the right is zero. Another integration by parts gives
an ⫽
f r (x) cos nx
n 2p
p
2
⫺
ⴚp
1
n 2p
冮
p
f s (x) cos nx dx.
ⴚp
The first term on the right is zero because of the periodicity and continuity of f r (x). Since
f s is continuous in the interval of integration, we have
ƒ f s (x) ƒ ⬍ M
for an appropriate constant M. Furthermore, ƒ cos nx ƒ ⬉ 1. It follows that
1
冮
p
1
ƒ an ƒ ⫽ 2
f s (x) cos nx dx 2 ⬍ 2
n p ⴚp
n p
2
冮
p
M dx ⫽
ⴚp
2M
n2
.
Similarly, ƒ bn ƒ ⬍ 2 M>n 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 series
ƒ a0 ƒ ⫹ 2M a1 ⫹ 1 ⫹
1
2
2
⫹
1
2
2
⫹
1
3
2
⫹
1
32
⫹ Áb
which is convergent. Hence that Fourier series converges and the proof is complete.
(Readers already familiar with uniform convergence will see that, by the Weierstrass
test in Sec. 15.5, 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
䊏
Sec. 15.5.)
EXAMPLE 2
Convergence at a Jump as Indicated in Theorem 2
The rectangular wave in Example 1 has a jump at x ⫽ 0. Its left-hand limit there is ⫺k and its right-hand limit
is k (Fig. 261). Hence the average of these limits is 0. The Fourier series (8) of the wave does indeed converge
to this value when x ⫽ 0 because then all its terms are 0. Similarly for the other jumps. This is in agreement
䊏
with Theorem 2.
Summary. A Fourier series of a given function f (x) of period 2p is a series of the form
(5) with coefficients given by the Euler formulas (6). Theorem 2 gives conditions that are
sufficient for this series to converge and at each x to have the value f (x), except at
discontinuities of f (x), where the series equals the arithmetic mean of the left-hand and
right-hand limits of f (x) at that point.
c11-a.qxd
10/30/10
1:24 PM
482
Page 482
CHAP. 11 Fourier Analysis
PROBLEM SET 11.1
1–5
PERIOD, FUNDAMENTAL PERIOD
The fundamental period is the smallest positive period. Find
it for
1. cos x, sin x, cos 2x, sin 2x, cos px, sin px,
cos 2px, sin 2px
2px
2px
2pnx
2. cos nx, sin nx, cos
, sin
, cos
,
k
k
k
2pnx
sin
k
3. If f ( x) and g (x) have period p, show that h (x) ⫽
af (x) ⫹ bg(x) (a, b, constant) has the period p. Thus
all functions of period p form a vector space.
4. Change of scale. If f (x) has period p, show that
f (ax), a ⫽ 0, and f (x>b), b ⫽ 0, are periodic functions
of x of periods p>a and bp, respectively. Give examples.
5. Show that f ⫽ const is periodic with any period but has
no fundamental period.
17.
–π
10. f (x) ⫽ b
FOURIER SERIES
Find the Fourier series of the given function f (x), which is
assumed to have the period 2p. Show the details of your
work. Sketch or graph the partial sums up to that including
cos 5x and sin 5x.
12. f (x) in Prob. 6
13. f (x) in Prob. 9
14. f (x) ⫽ x 2 (⫺p ⬍ x ⬍ p)
15. f (x) ⫽ x 2 (0 ⬍ x ⬍ 2p)
16.
1
π
2
–π
0
1
π
2
π
π
0
19.
π
–π
π
0
20.
1
π
2
–π
– 1π
2
21.
1
π
2
0
– 1π
2
π
π
–π
π
⫺cos2 x if ⫺p ⬍ x ⬍ 0
cos2 x if
0⬍x⬍p
11. Calculus review. Review integration techniques for
integrals as they are likely to arise from the Euler
formulas, for instance, definite integrals of x cos nx,
x 2 sin nx, eⴚ2x cos nx, etc.
12–21
1
–π
GRAPHS OF 2p–PERIODIC FUNCTIONS
Sketch or graph f (x) which for ⫺p ⬍ x ⬍ p is given as
π
0
18.
6–10
follows.
6. f (x) ⫽ ƒ x ƒ
7. f (x) ⫽ ƒ sin x ƒ , f (x) ⫽ sin ƒ x ƒ
8. f (x) ⫽ eⴚƒ x ƒ, f (x) ⫽ ƒ eⴚx ƒ
x
if ⫺p ⬍ x ⬍ 0
9. f (x) ⫽ b
p ⫺ x if
0⬍x⬍p
π
–π
22. CAS EXPERIMENT. Graphing. Write a program for
graphing partial sums of the following series. Guess
from the graph what f (x) the series may represent.
Confirm or disprove your guess by using the Euler
formulas.
(a) 2(sin x ⫹ 13 sin 3x ⫹ 15 sin 5x ⫹ Á)
⫺ 2( 12 sin 2x ⫹ 14 sin 4x ⫹ 16 sin 6x Á)
(b)
1
4
1
1
⫹ 2 acos x ⫹ cos 3x ⫹
cos 5x ⫹ Á b
p
2
9
25
(c)
2
3
1
p2 ⫹ 4(cos x ⫺ 14 cos 2x ⫹ 19 cos 3x ⫺ 16
cos 4x
⫹ ⫺ Á)
23. Discontinuities. Verify the last statement in Theorem
2 for the discontinuities of f (x) in Prob. 21.
24. CAS EXPERIMENT. Orthogonality. Integrate and
graph the integral of the product cos mx cos nx (with
various integer m and n of your choice) from ⫺a to a
as a function of a and conclude orthogonality of cos mx
c11-a.qxd
10/30/10
1:24 PM
Page 483
SEC. 11.2 Arbitrary Period. Even and Odd Functions. Half-Range Expansions
and cos nx (m ⫽ n) for a ⫽ p from the graph. For what
m and n will you get orthogonality for a ⫽ p>2, p>3,
p>4? Other a? Extend the experiment to cos mx sin nx
and sin mx sin nx.
25. CAS EXPERIMENT. Order of Fourier Coefficients.
The order seems to be 1>n if f is discontinous, and 1>n 2
11.2
483
if f is continuous but f r ⫽ df>dx is discontinuous, 1>n 3
if f and f r are continuous but f s is discontinuous, etc.
Try to verify this for examples. Try to prove it by
integrating the Euler formulas by parts. What is the
practical significance of this?
Arbitrary Period. Even and Odd Functions.
Half-Range Expansions
We now expand our initial basic discussion of Fourier series.
Orientation. This section concerns three topics:
1. Transition from period 2p to any period 2L, for the function f, simply by a
transformation of scale on the x-axis.
2. Simplifications. Only cosine terms if f is even (“Fourier cosine series”). Only sine
terms if f is odd (“Fourier sine series”).
3. Expansion of f given for 0 ⬉ x ⬉ L in two Fourier series, one having only cosine
terms and the other only sine terms (“half-range expansions”).
1. From Period 2p to Any Period p ⫽ 2L
Clearly, periodic functions in applications may have any period, not just 2p as in the last
section (chosen to have simple formulas). The notation p ⫽ 2L for the period is practical
because L will be a length of a violin string in Sec. 12.2, of a rod in heat conduction in
Sec. 12.5, and so on.
The transition from period 2p to be period p ⫽ 2L is effected by a suitable change of
scale, as follows. Let f (x) have period p ⫽ 2L. Then we can introduce a new variable v
such that f (x), as a function of v, has period 2p. If we set
(1)
(a) x ⫽
p
2p
v,
2p
p
(b) v ⫽ p x ⫽ x
L
so that
then v ⫽ ⫾p corresponds to x ⫽ ⫾L. This means that f, as a function of v, has period
2p and, therefore, a Fourier series of the form
(2)
ⴥ
L
f (x) ⫽ f a p vb ⫽ a0 ⫹ a (an cos nv ⫹ bn sin nv)
n⫽1
with coefficients obtained from (6) in the last section
1
a0 ⫽ 2p
(3)
冮
p
ⴚp
fa
L
p
vb dv,
1
bn ⫽ p
冮
p
ⴚp
1
an ⫽ p
冮
p
fa
ⴚp
L
p
L
f a p vb sin nv dv.
vb cos nv dv,
c11-a.qxd
10/30/10
1:24 PM
484
Page 484
CHAP. 11 Fourier Analysis
We could use these formulas directly, but the change to x simplifies calculations. Since
v⫽
(4)
p
L
x,
dv ⫽
we have
p
dx
L
and we integrate over x from ⫺L to L. Consequently, we obtain for a function f (x) of
period 2L the Fourier series
(5)
ⴥ
np
np
f (x) ⫽ a0 ⫹ a aan cos
x ⫹ bn sin
xb
L
L
n⫽1
with the Fourier coefficients of f (x) given by the Euler formulas (p>L in dx cancels
1> p in (3))
(0)
(6)
(a)
(b)
冮
1
2L
a0 ⫽
1
L
an ⫽
1
bn ⫽
L
L
f (x) dx
ⴚL
冮
L
冮
L
f (x) cos
ⴚL
f (x) sin
ⴚL
npx
dx
L
n ⫽ 1, 2, Á
n px
dx
L
n ⫽ 1, 2, Á .
Just as in Sec. 11.1, we continue to call (5) with any coefficients a trigonometric series.
And we can integrate from 0 to 2L or over any other interval of length p ⫽ 2L.
EXAMPLE 1
Periodic Rectangular Wave
Find the Fourier series of the function (Fig. 263)
Solution.
0
if
⫺2 ⬍ x ⬍ ⫺1
f (x) ⫽ d k
if
⫺1 ⬍ x ⬍
1
0
if
1⬍x⬍
2
p ⫽ 2L ⫽ 4, L ⫽ 2.
From (6.0) we obtain a0 ⫽ k>2 (verify!). From (6a) we obtain
an ⫽
1
2
冮
2
f (x) cos
ⴚ2
npx
2
dx ⫽
1
2
冮
1
k cos
npx
ⴚ1
2
dx ⫽
2k
np
sin
np
2
.
Thus an ⫽ 0 if n is even and
an ⫽ 2k>np if
n ⫽ 1, 5, 9, Á ,
an ⫽ ⫺2k>np if n ⫽ 3, 7, 11, Á .
From (6b) we find that bn ⫽ 0 for n ⫽ 1, 2, Á . Hence the Fourier series is a Fourier cosine series (that is, it
has no sine terms)
f (x) ⫽
k
2
⫹
2k
p
acos
p
2
x⫺
1
3
cos
3p
2
x⫹
1
5
cos
5p
2
x ⫺ ⫹ Áb .
䊏
c11-a.qxd
10/30/10
1:24 PM
Page 485
SEC. 11.2 Arbitrary Period. Even and Odd Functions. Half-Range Expansions
485
f(x)
k
f(x)
–2
k
2
x
–k
–2
–1
0
1
x
2
Fig. 263. Example 1
EXAMPLE 2
Fig. 264. Example 2
Periodic Rectangular Wave. Change of Scale
Find the Fourier series of the function (Fig. 264)
⫺k
if
⫺2 ⬍ x ⬍ 0
k
if
0⬍x⬍2
f (x) ⫽ c
Solution.
p ⫽ 2L ⫽ 4,
L ⫽ 2.
Since L ⫽ 2, we have in (3) v ⫽ px>2 and obtain from (8) in Sec. 11.1 with v instead of x, that is,
g(v) ⫽
4k
p
asin v ⫹
1
3
sin 3v ⫹
1
5
sin 5v ⫹ Á b
the present Fourier series
f (x) ⫽
4k
p
asin
p
2
x⫹
1
3
sin
3p
2
1
x⫹
5
sin
x ⫹ Áb .
5p
2
䊏
Confirm this by using (6) and integrating.
EXAMPLE 3
Half-Wave Rectifier
A sinusoidal voltage E sin vt, where t is time, is passed through a half-wave rectifier that clips the negative
portion of the wave (Fig. 265). Find the Fourier series of the resulting periodic function
0
u(t) ⫽ c
E sin vt
Solution.
if
⫺L ⬍ t ⬍ 0,
if
0⬍t⬍L
p ⫽ 2L ⫽
2p
,
v
L⫽
p
v
.
Since u ⫽ 0 when ⫺L ⬍ t ⬍ 0, we obtain from (6.0), with t instead of x,
a0 ⫽
2p 冮
v
p>v
E sin vt dt ⫽
0
E
p
and from (6a), by using formula (11) in App. A3.1 with x ⫽ vt and y ⫽ nvt,
an ⫽
p冮
v
p>v
E sin vt cos nvt dt ⫽
0
2p 冮
vE
p>v
[sin (1 ⫹ n) vt ⫹ sin (1 ⫺ n) vt] dt.
0
If n ⫽ 1, the integral on the right is zero, and if n ⫽ 2, 3, Á , we readily obtain
an ⫽
⫽
vE
2p
E
2p
c⫺
a
cos (1 ⫹ n) vt
(1 ⫹ n) v
⫺
⫺cos (1 ⫹ n)p ⫹ 1
1⫹n
cos (1 ⫺ n) vt
(1 ⫺ n) v
⫹
d
p>v
0
⫺cos (1 ⫺ n)p ⫹ 1
1⫺n
b.
If n is odd, this is equal to zero, and for even n we have
an ⫽
E
2
2
2E
a
⫹
b⫽⫺
2p 1 ⫹ n
1⫺n
(n ⫺ 1)(n ⫹ 1)p
(n ⫽ 2, 4, Á ).
c11-a.qxd
10/30/10
1:24 PM
486
Page 486
CHAP. 11 Fourier Analysis
In a similar fashion we find from (6b) that b1 ⫽ E>2 and bn ⫽ 0 for n ⫽ 2, 3, Á . Consequently,
u(t) ⫽
E
p
E
⫹
2
sin vt ⫺
2E
p
a
1
1
cos 2vt ⫹
cos 4vt ⫹ Á b .
1#3
3#5
䊏
u(t)
– π /ω
π /ω
0
t
Fig. 265. Half-wave rectifier
2. Simplifications: Even and Odd Functions
If f (x) is an even function, that is, f (⫺x) ⫽ f (x) (see Fig. 266), its Fourier series (5)
reduces to a Fourier cosine series
x
ⴥ
np
f (x) ⫽ a0 ⫹ a an cos
x
L
n⫽1
(5*)
Fig. 266.
Even function
( f even)
with coefficients (note: integration from 0 to L only!)
x
Fig. 267.
Odd function
(6*)
a0 ⫽
1
L
冮
L
f (x) dx,
an ⫽
0
冮
2
L
L
f (x) cos
0
npx
dx,
L
n ⫽ 1, 2, Á .
If f (x) is an odd function, that is, f (⫺x) ⫽ ⫺f (x) (see Fig. 267), its Fourier series (5)
reduces to a Fourier sine series
(5**)
ⴥ
np
f (x) ⫽ a bn sin
x
L
n⫽1
( f odd)
with coefficients
bn ⫽
(6**)
2
L
冮
L
f (x) sin
0
npx
dx.
L
These formulas follow from (5) and (6) by remembering from calculus that the definite
integral gives the net area (⫽ area above the axis minus area below the axis) under the
curve of a function between the limits of integration. This implies
(a)
冮
L
冮
L
g (x) dx ⫽ 2
ⴚL
(7)
(b)
冮
L
g (x) dx
for even g
0
h (x) dx ⫽ 0
for odd h
ⴚL
Formula (7b) implies the reduction to the cosine series (even f makes f (x) sin (npx>L) odd
since sin is odd) and to the sine series (odd f makes f (x) cos (npx>L) odd since cos is even).
Similarly, (7a) reduces the integrals in (6*) and (6**) to integrals from 0 to L. These reductions
are obvious from the graphs of an even and an odd function. (Give a formal proof.)
c11-a.qxd
10/30/10
1:25 PM
Page 487
SEC. 11.2 Arbitrary Period. Even and Odd Functions. Half-Range Expansions
487
Summary
Even Function of Period 2␲. If f is even and L ⫽ p, then
ⴥ
f (x) ⫽ a0 ⫹ a an cos nx
n⫽1
with coefficients
1
a0 ⫽ p
冮
p
2
an ⫽ p
f (x) dx,
0
冮
p
f (x) cos nx dx, n ⫽ 1, 2, Á
0
Odd Function of Period 2p. If f is odd and L ⫽ p, then
ⴥ
f (x) ⫽ a bn sin nx
n⫽1
with coefficients
2
bn ⫽ p
冮
p
n ⫽ 1, 2, Á .
f (x) sin nx dx,
0
EXAMPLE 4
Fourier Cosine and Sine Series
The rectangular wave in Example 1 is even. Hence it follows without calculation that its Fourier series is a
Fourier cosine series, the bn are all zero. Similarly, it follows that the Fourier series of the odd function in
Example 2 is a Fourier sine series.
In Example 3 you can see that the Fourier cosine series represents u(t) ⫺ E> p ⫺ 12 E sin vt. Can you prove
that this is an even function?
䊏
Further simplifications result from the following property, whose very simple proof is left
to the student.
THEOREM 1
Sum and Scalar Multiple
The Fourier coefficients of a sum f1 ⫹ f2 are the sums of the corresponding Fourier
coefficients of f1 and f2.
The Fourier coefficients of cf are c times the corresponding Fourier coefficients of f.
EXAMPLE 5
Sawtooth Wave
Find the Fourier series of the function (Fig. 268)
f (x) ⫽ x ⫹ p if ⫺p ⬍ x ⬍ p
and
f (x ⫹ 2p) ⫽ f (x).
f (x)
–π
π
x
Fig. 268. The function f(x). Sawtooth wave
c11-a.qxd
10/30/10
488
1:25 PM
Page 488
CHAP. 11 Fourier Analysis
y
S20
S3
5
S2
S1
y
–π
π
0
x
Fig. 269. Partial sums S1, S2, S3, S20 in Example 5
We have f ⫽ f1 ⫹ f2, where f1 ⫽ x and f2 ⫽ p. The Fourier coefficients of f2 are zero, except for
the first one (the constant term), which is p. Hence, by Theorem 1, the Fourier coefficients an, bn are those of
f1, except for a0, which is p. Since f1 is odd, an ⫽ 0 for n ⫽ 1, 2, Á , and
Solution.
bn ⫽
p冮
2
p
f1 (x) sin nx dx ⫽
0
p冮
2
p
x sin nx dx.
0
Integrating by parts, we obtain
bn ⫽
2 ⫺x cos nx
2
pc
n
p
⫹
0
1
n
冮
p
0
2
cos nx dx d ⫽ ⫺ cos np.
n
Hence b1 ⫽ 2, b2 ⫽ ⫺ 22 , b3 ⫽ 23 , b4 ⫽ ⫺ 24 , Á , and the Fourier series of f (x) is
f (x) ⫽ p ⫹ 2 asin x ⫺
1
2
sin 2x ⫹
1
3
sin 3x ⫺ ⫹ Á b .
(Fig. 269)
䊏
3. Half-Range Expansions
Half-range expansions are Fourier series. The idea is simple and useful. Figure 270
explains it. We want to represent f (x) in Fig. 270.0 by a Fourier series, where f (x)
may be the shape of a distorted violin string or the temperature in a metal bar of length
L, for example. (Corresponding problems will be discussed in Chap. 12.) Now comes
the idea.
We could extend f (x) as a function of period L and develop the extended function into
a Fourier series. But this series would, in general, contain both cosine and sine terms. We
can do better and get simpler series. Indeed, for our given f we can calculate Fourier
coefficients from (6*) or from (6**). And we have a choice and can take what seems
more practical. If we use (6*), we get (5*). This is the even periodic extension f1 of f
in Fig. 270a. If we choose (6**) instead, we get (5**), the odd periodic extension f2 of
f in Fig. 270b.
Both extensions have period 2L. This motivates the name half-range expansions: f is
given (and of physical interest) only on half the range, that is, on half the interval of
periodicity of length 2L.
Let us illustrate these ideas with an example that we shall also need in Chap. 12.
c11-a.qxd
10/30/10
1:25 PM
Page 489
SEC. 11.2 Arbitrary Period. Even and Odd Functions. Half-Range Expansions
489
f (x)
x
L
(0) The given function f (x)
f1(x)
–L
x
L
(a) f (x) continued as an even periodic function of period 2L
f2(x)
–L
x
L
(b) f (x) continued as an odd periodic function of period 2L
Fig. 270. Even and odd extensions of period 2L
EXAMPLE 6
“Triangle” and Its Half-Range Expansions
Find the two half-range expansions of the function (Fig. 271)
k
0
L/2
L
2k
x
L
x
if
(L ⫺ x)
if
L
0⬍x⬍
2
f(x) ⫽ e
Fig. 271. The given
function in Example 6
2k
L
Solution.
L
⬍ x ⬍ L.
2
(a) Even periodic extension. From (6*) we obtain
a0 ⫽
an ⫽
1 2k
c
L L
冮
L>2
2 2k
c
L L
冮
L>2
x dx ⫹
L 冮
2k
0
L
L>2
x cos
L
np
L
k
(L ⫺ x) dx d ⫽
L 冮
2k
x dx ⫹
2
,
L
(L ⫺ x) cos
np
L
L>2
x dx d .
We consider an. For the first integral we obtain by integration by parts
冮
L>2
x cos
np
L
x dx ⫽
Lx
sin
np
np
L
L>2
x2
0
0
⫽
L2
2np
sin
⫺
np
2
冮
np
L>2
L
sin
np
⫹
L2
2
n p2
acos
np
2
x dx
L
0
⫺ 1b .
Similarly, for the second integral we obtain
冮
L
(L ⫺ x) cos
L>2
np
L
x dx ⫽
L
np
(L ⫺ x) sin
⫽ a0 ⫺
L
np
aL ⫺
np
L
L
2
L
x2
⫹
L>2
b sin
np
2
np 冮
L
L
sin
L>2
b⫺
L2
2
n p2
np
L
x dx
acos np ⫺ cos
np
2
b.
c11-a.qxd
10/30/10
490
1:25 PM
Page 490
CHAP. 11 Fourier Analysis
We insert these two results into the formula for an. The sine terms cancel and so does a factor L2. This gives
an ⫽
a2 cos
4k
n 2p2
np
2
⫺ cos np ⫺ 1b .
Thus,
a2 ⫽ ⫺16k>(22p2),
a6 ⫽ ⫺16k>(62p2),
a10 ⫽ ⫺16k>(102p2), Á
and an ⫽ 0 if n ⫽ 2, 6, 10, 14, Á . Hence the first half-range expansion of f (x) is (Fig. 272a)
k
f (x) ⫽
2
⫺
16k
a
1
p2 22
cos
2p
x⫹
L
1
cos
62
6p
L
x ⫹ Áb .
This Fourier cosine series represents the even periodic extension of the given function f (x), of period 2L.
(b) Odd periodic extension. Similarly, from (6**) we obtain
8k
bn ⫽
(5)
2
sin
2
n p
np
2
.
Hence the other half-range expansion of f (x) is (Fig. 272b)
f (x) ⫽
8k
2
p
a
1
2
sin
1
p
1
x⫺
L
2
sin
3
3p
1
x⫹
2
L
5
sin
5p
L
x ⫺ ⫹ Á b.
The series represents the odd periodic extension of f (x), of period 2L.
Basic applications of these results will be shown in Secs. 12.3 and 12.5.
–L
䊏
x
L
0
(a) Even extension
–L
x
L
0
(b) Odd extension
Fig. 272. Periodic extensions of f(x) in Example 6
PROBLEM SET 11.2
1–7
EVEN AND ODD FUNCTIONS
Are the following functions even or odd or neither even nor
odd?
1. ex, eⴚƒ x ƒ, x 3 cos nx, x 2 tan px, sinh x ⫺ cosh x
2. sin2 x, sin (x 2), ln x, x>(x 2 ⫹ 1), x cot x
3. Sums and products of even functions
4. Sums and products of odd functions
5. Absolute values of odd functions
6. Product of an odd times an even function
7. Find all functions that are both even and odd.
8–17
FOURIER SERIES FOR PERIOD p = 2L
Is the given function even or odd or neither even nor
odd? Find its Fourier series. Show details of your
work.
8.
1
–1
0
1
c11-a.qxd
10/30/10
1:25 PM
Page 491
SEC. 11.2 Arbitrary Period. Even and Odd Functions. Half-Range Expansions
9.
(b) Apply the program to Probs. 8–11, graphing the first
few partial sums of each of the four series on common
axes. Choose the first five or more partial sums until
they approximate the given function reasonably well.
Compare and comment.
1
2
–2
–1
10.
22. Obtain the Fourier series in Prob. 8 from that in
Prob. 17.
4
23–29
–4
4
11. f (x) ⫽ x
HALF-RANGE EXPANSIONS
Find (a) the Fourier cosine series, (b) the Fourier sine series.
Sketch f (x) and its two periodic extensions. Show the
details.
23.
–4
2
491
1
(⫺1 ⬍ x ⬍ 1), p ⫽ 2
12. f (x) ⫽ 1 ⫺ x 2>4 (⫺2 ⬍ x ⬍ 2), p ⫽ 4
13.
4
1
2
24. 1
1
2
–1
2
14. f ( x) ⫽ cos px (⫺ 12 ⬍ x ⬍ 12), p ⫽ 1
15.
π
2
4
25. π
–
2
–π
π
π
– π–
2
16. f ( x) ⫽ x ƒ x ƒ
17.
(⫺1 ⬍ x ⬍ 1), p ⫽ 2
26.
π–
2
1
–1
1
27.
20. Numeric Values. Using Prob. 11, show that 1 ⫹ 14 ⫹
1
1
Á ⫽ 16 p2.
9 ⫹ 16 ⫹
21. CAS PROJECT. Fourier Series of 2L-Periodic
Functions. (a) Write a program for obtaining partial
sums of a Fourier series (5).
28.
π
π–
2
π
π–
2
18. Rectifier. Find the Fourier series of the function
obtained by passing the voltage v(t) ⫽ V0 cos 100 pt
through a half-wave rectifier that clips the negative
half-waves.
19. Trigonometric Identities. Show that the familiar
identities cos3 x ⫽ 34 cos x ⫹ 14 cos 3x and sin3 x ⫽ 34
sin x ⫺ 14 sin 3x can be interpreted as Fourier series
expansions. Develop cos 4 x.
π–
2
L
L
29. f (x) ⫽ sin x (0 ⬍ x ⬍ p)
30. Obtain the solution to Prob. 26 from that of
Prob. 27.
c11-a.qxd
10/30/10
1:25 PM
492
11.3
Page 492
CHAP. 11 Fourier Analysis
Forced Oscillations
Fourier series have important applications for both ODEs and PDEs. In this section we
shall focus on ODEs and cover similar applications for PDEs in Chap. 12. All these
applications will show our indebtedness to Euler’s and Fourier’s ingenious idea of splitting
up periodic functions into the simplest ones possible.
From Sec. 2.8 we know that forced oscillations of a body of mass m on a spring of
modulus k are governed by the ODE
my s ⫹ cy r ⫹ ky ⫽ r (t)
(1)
where y ⫽ y (t) is the displacement from rest, c the damping constant, k the spring constant
(spring modulus), and r (t) the external force depending on time t. Figure 274 shows the
model and Fig. 275 its electrical analog, an RLC-circuit governed by
LI s ⫹ RI r ⫹
(1*)
1
I ⫽ E r (t)
C
(Sec. 2.9).
We consider (1). If r (t) is a sine or cosine function and if there is damping (c ⬎ 0),
then the steady-state solution is a harmonic oscillation with frequency equal to that of r (t).
However, if r (t) is not a pure sine or cosine function but is any other periodic function,
then the steady-state solution will be a superposition of harmonic oscillations with
frequencies equal to that of r (t) and integer multiples of these frequencies. And if one of
these frequencies is close to the (practical) resonant frequency of the vibrating system (see
Sec. 2.8), then the corresponding oscillation may be the dominant part of the response of
the system to the external force. This is what the use of Fourier series will show us. Of
course, this is quite surprising to an observer unfamiliar with Fourier series, which are
highly important in the study of vibrating systems and resonance. Let us discuss the entire
situation in terms of a typical example.
C
Spring
R
External
force r (t)
L
Mass m
Dashpot
E(t)
Fig. 274. Vibrating system
under consideration
EXAMPLE 1
Fig. 275. Electrical analog of the system
in Fig. 274 (RLC-circuit)
Forced Oscillations under a Nonsinusoidal Periodic Driving Force
In (1), let m ⫽ 1 (g), c ⫽ 0.05 (g>sec), and k ⫽ 25 (g>sec2), so that (1) becomes
(2)
y s ⫹ 0.05y r ⫹ 25y ⫽ r (t)
c11-a.qxd
10/30/10
1:25 PM
Page 493
SEC. 11.3 Forced Oscillations
493
r(t)
ππ/2
–π
π
π
–π
ππ/2
t
Fig. 276. Force in Example 1
where r (t) is measured in g ⴢ cm>sec2. Let (Fig. 276)
p
t⫹
2
⫺p ⬍ t ⬍ 0,
if
r (t) ⫽ e
r (t ⫹ 2p) ⫽ r (t).
p
⫺t ⫹
2
0 ⬍ t ⬍ p,
if
Find the steady-state solution y(t).
Solution.
We represent r (t) by a Fourier series, finding
r (t) ⫽
(3)
4
p
acos t ⫹
1
32
cos 3t ⫹
1
52
cos 5t ⫹ Á b .
Then we consider the ODE
y s ⫹ 0.05y r ⫹ 25y ⫽
(4)
4
2
n p
(n ⫽ 1, 3, Á )
cos nt
whose right side is a single term of the series (3). From Sec. 2.8 we know that the steady-state solution yn (t)
of (4) is of the form
yn ⫽ An cos nt ⫹ Bn sin nt.
(5)
By substituting this into (4) we find that
(6)
An ⫽
4(25 ⫺ n 2)
n 2pDn
,
Bn ⫽
0.2
npDn
,
Dn ⫽ (25 ⫺ n 2)2 ⫹ (0.05n)2.
where
Since the ODE (2) is linear, we may expect the steady-state solution to be
y ⫽ y1 ⫹ y3 ⫹ y5 ⫹ Á
(7)
where yn is given by (5) and (6). In fact, this follows readily by substituting (7) into (2) and using the Fourier
series of r (t), provided that termwise differentiation of (7) is permissible. (Readers already familiar with the
notion of uniform convergence [Sec. 15.5] may prove that (7) may be differentiated term by term.)
From (6) we find that the amplitude of (5) is (a factor 1Dn cancels out)
Cn ⫽ 2A2n ⫹ B 2n ⫽
4
n p 2Dn
2
.
Values of the first few amplitudes are
C1 ⫽ 0.0531 C3 ⫽ 0.0088 C5 ⫽ 0.2037 C7 ⫽ 0.0011 C9 ⫽ 0.0003.
Figure 277 shows the input (multiplied by 0.1) and the output. For n ⫽ 5 the quantity Dn is very small, the
denominator of C5 is small, and C5 is so large that y5 is the dominating term in (7). Hence the output is almost
a harmonic oscillation of five times the frequency of the driving force, a little distorted due to the term y1, whose
amplitude is about 25% of that of y5. You could make the situation still more extreme by decreasing the damping
constant c. Try it.
䊏
c11-a.qxd
10/30/10
494
1:25 PM
Page 494
CHAP. 11 Fourier Analysis
y
0.3
Output
0.2
0.1
–3
–2
–1
0
1
2
t
3
–0.1
Input
–0.2
Fig. 277. Input and steady-state output in Example 1
PROBLEM SET 11.3
1. Coefficients Cn. Derive the formula for Cn from An
and Bn.
2. Change of spring and damping. In Example 1, what
happens to the amplitudes Cn if we take a stiffer spring,
say, of k ⫽ 49? If we increase the damping?
3. Phase shift. Explain the role of the Bn’s. What happens
if we let c : 0?
4. Differentiation of input. In Example 1, what happens
if we replace r (t) with its derivative, the rectangular wave?
What is the ratio of the new Cn to the old ones?
5. Sign of coefficients. Some of the An in Example 1 are
positive, some negative. All Bn are positive. Is this
physically understandable?
6–11
GENERAL SOLUTION
Find a general solution of the ODE y s ⫹ v2y ⫽ r (t) with
r (t) as given. Show the details of your work.
6. r (t) ⫽ sin at ⫹ sin bt, v2 ⫽ a2, b2
7. r (t) ⫽ sin t, v ⫽ 0.5, 0.9, 1.1, 1.5, 10
8. Rectifier. r (t) ⫽ p/4 ƒ cos t ƒ if ⫺p ⬍ t ⬍ p and
r (t ⫹ 2p) ⫽ r (t), ƒ v ƒ ⫽ 0, 2, 4, Á
9. What kind of solution is excluded in Prob. 8 by
ƒ v ƒ ⫽ 0, 2, 4, Á ?
10. Rectifier. r (t) ⫽ p/4 ƒ sin t ƒ if 0 ⬍ t ⬍ 2p and
r (t ⫹ 2p) ⫽ r (t), ƒ v ƒ ⫽ 0, 2, 4, Á
⫺1 if ⫺p ⬍ t ⬍ 0
11. r (t) ⫽ b
ƒ v ƒ ⫽ 1, 3, 5, Á
1 if
0 ⬍ t ⬍ p,
12. CAS Program. Write a program for solving the ODE
just considered and for jointly graphing input and output
of an initial value problem involving that ODE. Apply
the program to Probs. 7 and 11 with initial values of your
choice.
13–16
STEADY-STATE DAMPED OSCILLATIONS
Find the steady-state oscillations of y s ⫹ cy r ⫹ y ⫽ r (t)
with c ⬎ 0 and r (t) as given. Note that the spring constant
is k ⫽ 1. Show the details. In Probs. 14–16 sketch r (t).
N
13. r (t) ⫽ a (an cos nt ⫹ bn sin nt)
n⫽1
14. r (t) ⫽ b
⫺1 if ⫺p ⬍ t ⬍ 0
1 if
15. r (t) ⫽ t (p2 ⫺ t 2)
r (t ⫹ 2p) ⫽ r (t)
16. r (t) ⫽
t if ⫺p>2
e
p ⫺ t if p>2
17–19
and r (t ⫹ 2p) ⫽ r (t)
0⬍t⬍p
if ⫺p ⬍ t ⬍ p and
⬍ t ⬍ p>2
⬍ t ⬍ 3 p>2
and r (t ⫹ 2 p) ⫽ r (t)
RLC-CIRCUIT
Find the steady-state current I (t) in the RLC-circuit in
Fig. 275, where R ⫽ 10 ⍀, L ⫽ 1 H, C ⫽ 10 ⴚ1 F and with
E (t) V as follows and periodic with period 2p. Graph or
sketch the first four partial sums. Note that the coefficients
of the solution decrease rapidly. Hint. Remember that the
ODE contains E r(t), not E (t), cf. Sec. 2.9.
17. E (t) ⫽ b
⫺50t 2
if
⫺p ⬍ t ⬍ 0
2
if
0⬍t⬍p
50t
c11-a.qxd
10/30/10
1:25 PM
Page 495
SEC. 11.4 Approximation by Trigonometric Polynomials
18. E (t) ⫽ b
100 (t ⫺ t 2)
if
⫺p ⬍ t ⬍ 0
100 (t ⫹ t 2)
if
0⬍t⬍p
19. E (t) ⫽ 200t (p2 ⫺ t 2) (⫺p ⬍ t ⬍ p)
11.4
495
20. CAS EXPERIMENT. Maximum Output Term.
Graph and discuss outputs of y s ⫹ cy r ⫹ ky ⫽ r (t) with
r (t) as in Example 1 for various c and k with emphasis on
the maximum Cn and its ratio to the second largest ƒ Cn ƒ .
Approximation
by Trigonometric Polynomials
Fourier series play a prominent role not only in differential equations but also in
approximation theory, an area that is concerned with approximating functions by
other functions—usually simpler functions. Here is how Fourier series come into the
picture.
Let f (x) be a function on the interval ⫺p ⬉ x ⬉ p that can be represented on this
interval by a Fourier series. Then the Nth partial sum of the Fourier series
N
(1)
f (x) ⬇ a0 ⫹ a (an cos nx ⫹ bn sin nx)
n⫽1
is an approximation of the given f (x). In (1) we choose an arbitrary N and keep it fixed.
Then we ask whether (1) is the “best” approximation of f by a trigonometric polynomial
of the same degree N, that is, by a function of the form
N
(2)
F (x) ⫽ A0 ⫹ a (An cos nx ⫹ Bn sin nx)
(N fixed).
n⫽1
Here, “best” means that the “error” of the approximation is as small as possible.
Of course we must first define what we mean by the error of such an approximation.
We could choose the maximum of ƒ f (x) ⫺ F (x) ƒ . But in connection with Fourier series
it is better to choose a definition of error that measures the goodness of agreement between
f and F on the whole interval ⫺p ⬉ x ⬉ p. This is preferable since the sum f of a Fourier
series may have jumps: F in Fig. 278 is a good overall approximation of f, but the maximum
of ƒ f (x) ⫺ F (x) ƒ (more precisely, the supremum) is large. We choose
(3)
E⫽
冮
p
( f ⫺ F)2 dx.
ⴚp
f
F
x0
Fig. 278. Error of approximation
x
c11-a.qxd
10/30/10
496
1:25 PM
Page 496
CHAP. 11 Fourier Analysis
This is called the square error of F relative to the function f on the interval ⫺p ⬉ x ⬉ p.
Clearly, E ⭌ 0.
N being fixed, we want to determine the coefficients in (2) such that E is minimum.
Since ( f ⫺ F)2 ⫽ f 2 ⫺ 2fF ⫹ F 2, we have
E⫽
(4)
冮
p
f dx ⫺ 2
2
ⴚp
冮
p
f F dx ⫹
ⴚp
冮
p
F 2 dx.
ⴚp
We square (2), insert it into the last integral in (4), and evaluate the occurring integrals.
This gives integrals of cos2 nx and sin2 nx (n ⭌ 1), which equal p, and integrals of
cos nx, sin nx, and (cos nx)(sin mx), which are zero (just as in Sec. 11.1). Thus
冮
p
F dx ⫽
2
ⴚp
冮
p
N
ⴚp
2
c A0 ⫹ a (An cos nx ⫹ Bn sin nx) d dx
n⫽1
2
2
⫽ p(2A02 ⫹ A12 ⫹ Á ⫹ AN
⫹ B12 ⫹ Á ⫹ BN
).
We now insert (2) into the integral of f F in (4). This gives integrals of f cos nx as well
as f sin nx, just as in Euler’s formulas, Sec. 11.1, for an and bn (each multiplied by An or
Bn). Hence
冮
p
f F dx ⫽ p(2A0a0 ⫹ A1a1 ⫹ Á ⫹ ANaN ⫹ B1b1 ⫹ Á ⫹ BNbN).
ⴚp
With these expressions, (4) becomes
E⫽
(5)
冮
p
ⴚp
N
f 2 dx ⫺ 2p c 2A0 a0 ⫹ a (An an ⫹ Bn bn) d
n⫽1
N
⫹ p c 2A02 ⫹ a (An2 ⫹ Bn2) d .
n⫽1
We now take An ⫽ an and Bn ⫽ bn in (2). Then in (5) the second line cancels half of the
integral-free expression in the first line. Hence for this choice of the coefficients of F the
square error, call it E*, is
(6)
E* ⫽
冮
p
ⴚp
N
f 2 dx ⫺ p c 2a02 ⫹ a (an2 ⫹ bn2) d .
n⫽1
We finally subtract (6) from (5). Then the integrals drop out and we get terms
A2n ⫺ 2Anan ⫹ a 2n ⫽ (An ⫺ an)2 and similar terms (Bn ⫺ bn)2:
N
E ⫺ E* ⫽ p e 2(A0 ⫺ a0)2 ⫹ a [(An ⫺ an)2 ⫹ (Bn ⫺ bn)2] f .
n⫽1
Since the sum of squares of real numbers on the right cannot be negative,
E ⫺ E* ⭌ 0,
thus
E ⭌ E*,
and E ⫽ E* if and only if A0 ⫽ a0, Á , BN ⫽ bN. This proves the following fundamental
minimum property of the partial sums of Fourier series.
c11-a.qxd
10/30/10
1:25 PM
Page 497
SEC. 11.4 Approximation by Trigonometric Polynomials
THEOREM 1
497
Minimum Square Error
The square error of F in (2) (with fixed N) relative to f on the interval ⫺p ⬉ x ⬉ p
is minimum if and only if the coefficients of F in (2) are the Fourier coefficients of f.
This minimum value E* is given by (6).
From (6) we see that E* cannot increase as N increases, but may decrease. Hence with
increasing N the partial sums of the Fourier series of f yield better and better approximations to f, considered from the viewpoint of the square error.
Since E* ⭌ 0 and (6) holds for every N, we obtain from (6) the important Bessel’s
inequality
2a02
(7)
ⴥ
⫹ a
(an2
⫹
1
⬉p
bn2)
n⫽1
冮
p
f (x)2 dx
ⴚp
for the Fourier coefficients of any function f for which integral on the right exists. (For
F. W. Bessel see Sec. 5.5.)
It can be shown (see [C12] in App. 1) that for such a function f, Parseval’s theorem holds;
that is, formula (7) holds with the equality sign, so that it becomes Parseval’s identity3
2a02
(8)
ⴥ
⫹ a
(an2
⫹
n⫽1
EXAMPLE 1
bn2)
1
⫽p
冮
p
f (x)2 dx.
ⴚp
Minimum Square Error for the Sawtooth Wave
Compute the minimum square error E* of F (x) with N ⫽ 1, 2, Á , 10, 20, Á , 100 and 1000 relative to
f (x) ⫽ x ⫹ p
(⫺p ⬍ x ⬍ p)
on the interval ⫺p ⬉ x ⬉ p.
Solution.
F (x) ⫽ p ⫹ 2 (sin x ⫺
Sec. 11.3. From this and (6),
E* ⫽
1
2
冮
sin 2x ⫹
p
ⴚp
1
3
sin 3x ⫺ ⫹ Á ⫹
(⫺1)N⫹1
N
N
(x ⫹ p)2 dx ⫺ p a2p2 ⫹ 4 a
n⫽1
1
n2
sin Nx) by Example 3 in
b.
Numeric values are:
2π
π
–π
0
π x
Fig. 279. F with
N ⫽ 20 in Example 1
N
E*
N
E*
N
E*
N
E*
1
2
3
4
5
8.1045
4.9629
3.5666
2.7812
2.2786
6
7
8
9
10
1.9295
1.6730
1.4767
1.3216
1.1959
20
30
40
50
60
0.6129
0.4120
0.3103
0.2488
0.2077
70
80
90
100
1000
0.1782
0.1561
0.1389
0.1250
0.0126
3
MARC ANTOINE PARSEVAL (1755–1836), French mathematician. A physical interpretation of the identity
follows in the next section.
c11-a.qxd
10/30/10
498
1:25 PM
Page 498
CHAP. 11 Fourier Analysis
F ⫽ S1, S2, S3 are shown in Fig. 269 in Sec. 11.2, and F ⫽ S20 is shown in Fig. 279. Although ƒ f (x) ⫺ F (x) ƒ
is large at ⫾p (how large?), where f is discontinuous, F approximates f quite well on the whole interval, except
near ⫾p, where “waves” remain owing to the “Gibbs phenomenon,” which we shall discuss in the next section.
Can you think of functions f for which E* decreases more quickly with increasing N?
䊏
PROBLEM SET 11.4
factors on which the decrease of E* with N depends.
For each function considered find the smallest N such
that E* ⬍ 0.1.
1. CAS Problem. Do the numeric and graphic work in
Example 1 in the text.
2–5
MINIMUM SQUARE ERROR
Find the trigonometric polynomial F (x) of the form (2) for
which the square error with respect to the given f (x) on the
interval ⫺p ⬍ x ⬍ p is minimum. Compute the minimum
value for N ⫽ 1, 2, Á , 5 (or also for larger values if you
have a CAS).
2. f (x) ⫽ x (⫺p ⬍ x ⬍ p)
3. f (x) ⫽ ƒ x ƒ (⫺p ⬍ x ⬍ p)
4. f (x) ⫽ x 2 (⫺p ⬍ x ⬍ p)
⫺1 if ⫺p ⬍ x ⬍ 0
5. f (x) ⫽ b
1 if
0⬍x⬍p
6. Why are the square errors in Prob. 5 substantially larger
than in Prob. 3?
7. f (x) ⫽ x 3 (⫺p ⬍ x ⬍ p)
8. f (x) ⫽ ƒ sin x ƒ (⫺p ⬍ x ⬍ p), full-wave rectifier
9. Monotonicity. Show that the minimum square error
(6) is a monotone decreasing function of N. How can
you use this in practice?
10. CAS EXPERIMENT. Size and Decrease of E*.
Compare the size of the minimum square error E* for
functions of your choice. Find experimentally the
11.5
PARSEVALS’S IDENTITY
11–15
Using (8), prove that the series has the indicated sum.
Compute the first few partial sums to see that the convergence
is rapid.
2
Á ⫽ p ⫽ 1.233700550
⫹
32
52
8
Use Example 1 in Sec. 11.1.
11. 1 ⫹
1
⫹
1
p4
⫹Á⫽
⫽ 1.082323234
2
3
90
Use Prob. 14 in Sec. 11.1.
12. 1 ⫹
1
4
⫹
1
4
4
Á ⫽ p ⫽ 1.014678032
⫹
96
34
54
74
Use Prob. 17 in Sec. 11.1.
13. 1 ⫹
14.
冮
p
冮
p
1
⫹
1
⫹
cos4 x dx ⫽
3p
4
cos6 x dx ⫽
5p
8
ⴚp
15.
1
ⴚp
Sturm–Liouville Problems.
Orthogonal Functions
The idea of the Fourier series was to represent general periodic functions in terms of
cosines and sines. The latter formed a trigonometric system. This trigonometric system
has the desirable property of orthogonality which allows us to compute the coefficient of
the Fourier series by the Euler formulas.
The question then arises, can this approach be generalized? That is, can we replace the
trigonometric system of Sec. 11.1 by other orthogonal systems (sets of other orthogonal
functions)? The answer is “yes” and will lead to generalized Fourier series, including the
Fourier–Legendre series and the Fourier–Bessel series in Sec. 11.6.
To prepare for this generalization, we first have to introduce the concept of a Sturm–
Liouville Problem. (The motivation for this approach will become clear as you read on.)
Consider a second-order ODE of the form
c11-a.qxd
10/30/10
1:25 PM
Page 499
SEC. 11.5 Sturm–Liouville Problems. Orthogonal Functions
(1)
499
[ p (x)y r ] r ⫹ [ q (x) ⫹ lr (x)]y ⫽ 0
on some interval a ⬉ x ⬉ b, satisfying conditions of the form
(2)
(a)
k 1 y ⫹ k 2 y r ⫽ 0 at x ⫽ a
(b)
l 1 y ⫹ l 2 y r ⫽ 0 at x ⫽ b.
Here l is a parameter, and k 1, k 2, l 1, l 2 are given real constants. Furthermore, at least one
of each constant in each condition (2) must be different from zero. (We will see in Example
1 that, if p(x) ⫽ r(x) ⫽ 1 and q(x) ⫽ 0, then sin 1lx and cos 1lx satisfy (1) and constants
can be found to satisfy (2).) Equation (1) is known as a Sturm–Liouville equation.4
Together with conditions 2(a), 2(b) it is know as the Sturm–Liouville problem. It is an
example of a boundary value problem.
A boundary value problem consists of an ODE and given boundary conditions
referring to the two boundary points (endpoints) x ⫽ a and x ⫽ b of a given interval
a ⬉ x ⬉ b.
The goal is to solve these type of problems. To do so, we have to consider
Eigenvalues, Eigenfunctions
Clearly, y ⬅ 0 is a solution—the “trivial solution”—of the problem (1), (2) for any l
because (1) is homogeneous and (2) has zeros on the right. This is of no interest. We want
to find eigenfunctions y (x), that is, solutions of (1) satisfying (2) without being identically
zero. We call a number l for which an eigenfunction exists an eigenvalue of the Sturm–
Liouville problem (1), (2).
Many important ODEs in engineering can be written as Sturm–Liouville equations. The
following example serves as a case in point.
EXAMPLE 1
Trigonometric Functions as Eigenfunctions. Vibrating String
Find the eigenvalues and eigenfunctions of the Sturm–Liouville problem
(3)
y s ⫹ ly ⫽ 0,
y (0) ⫽ 0, y(p) ⫽ 0.
This problem arises, for instance, if an elastic string (a violin string, for example) is stretched a little and fixed
at its ends x ⫽ 0 and x ⫽ p and then allowed to vibrate. Then y (x) is the “space function” of the deflection
u (x, t) of the string, assumed in the form u (x, t) ⫽ y (x)w (t), where t is time. (This model will be discussed in
great detail in Secs, 12.2–12.4.)
From (1) nad (2) we see that p ⫽ 1, q ⫽ 0, r ⫽ 1 in (1), and a ⫽ 0, b ⫽ p, k 1 ⫽ l 1 ⫽ 1,
k 2 ⫽ l 2 ⫽ 0 in (2). For negative l ⫽ ⫺␯2 a general solution of the ODE in (3) is y (x) ⫽ c1e␯x ⫹ c2eⴚ␯x. From
the boundary conditions we obtain c1 ⫽ c2 ⫽ 0, so that y ⬅ 0, which is not an eigenfunction. For l ⫽ 0 the
situation is similar. For positive l ⫽ ␯2 a general solution is
Solution.
y(x) ⫽ A cos ␯x ⫹ B sin ␯x.
4
JACQUES CHARLES FRANÇOIS STURM (1803–1855) was born and studied in Switzerland and then
moved to Paris, where he later became the successor of Poisson in the chair of mechanics at the Sorbonne (the
University of Paris).
JOSEPH LIOUVILLE (1809–1882), French mathematician and professor in Paris, contributed to various
fields in mathematics and is particularly known by his important work in complex analysis (Liouville’s theorem;
Sec. 14.4), special functions, differential geometry, and number theory.
c11-a.qxd
10/30/10
500
1:25 PM
Page 500
CHAP. 11 Fourier Analysis
From the first boundary condition we obtain y (0) ⫽ A ⫽ 0. The second boundary condition then yields
y (p) ⫽ B sin ␯p ⫽ 0,
␯ ⫽ 0, ⫾ 1, ⫾ 2, Á .
thus
For ␯ ⫽ 0 we have y ⬅ 0. For l ⫽ ␯2 ⫽ 1, 4, 9, 16, Á , taking B ⫽ 1, we obtain
y (x) ⫽ sin ␯x
(␯ ⫽ 2l ⫽ 1, 2, Á ).
Hence the eigenvalues of the problem are l ⫽ ␯2, where ␯ ⫽ 1, 2, Á , and corresponding eigenfunctions are
y(x) ⫽ sin ␯x, where ␯ ⫽ 1, 2 Á .
䊏
Note that the solution to this problem is precisely the trigonometric system of the Fourier
series considered earlier. It can be shown that, under rather general conditions on the
functions p, q, r in (1), the Sturm–Liouville problem (1), (2) has infinitely many eigenvalues.
The corresponding rather complicated theory can be found in Ref. [All] listed in App. 1.
Furthermore, if p, q, r, and p r in (1) are real-valued and continuous on the interval
a ⬉ x ⬉ b and r is positive throughout that interval (or negative throughout that interval),
then all the eigenvalues of the Sturm–Liouville problem (1), (2) are real. (Proof in App. 4.)
This is what the engineer would expect since eigenvalues are often related to frequencies,
energies, or other physical quantities that must be real.
The most remarkable and important property of eigenfunctions of Sturm–Liouville
problems is their orthogonality, which will be crucial in series developments in terms of
eigenfunctions, as we shall see in the next section. This suggests that we should next
consider orthogonal functions.
Orthogonal Functions
Functions y1(x), y2 (x), Á defined on some interval a ⬉ x ⬉ b are called orthogonal on this
interval with respect to the weight function r (x) ⬎ 0 if for all m and all n different from m,
b
(4)
(ym, yn) ⫽
冮 r (x) y
m (x) yn (x)
dx ⫽ 0
(m ⫽ n).
a
(ym, yn) is a standard notation for this integral. The norm 储ym储 of ym is defined by
b
(5)
储 ym 储 ⫽ 2(ym, ym) ⫽
冮 r (x) y
G
2
m (x)
dx.
a
Note that this is the square root of the integral in (4) with n ⫽ m.
The functions y1, y2, Á are called orthonormal on a ⬉ x ⬉ b if they are orthogonal
on this interval and all have norm 1. Then we can write (4), (5) jointly by using the
Kronecker symbol5 dmn, namely,
b
( ym , yn ) ⫽
冮 r (x) y
m (x) yn (x)
a
5
dx ⫽ dmn ⫽ e
0
if
m⫽n
1
if
m ⫽ n.
LEOPOLD KRONECKER (1823–1891). German mathematician at Berlin University, who made important
contributions to algebra, group theory, and number theory.
c11-a.qxd
10/30/10
1:25 PM
Page 501
SEC. 11.5 Sturm–Liouville Problems. Orthogonal Functions
501
If r (x) ⫽ 1, we more briefly call the functions orthogonal instead of orthogonal with
respect to r (x) ⫽ 1; similarly for orthognormality. Then
b
b
(ym, yn) ⫽
冮y
m (x) yn (x)
dx ⫽ 0 (m ⫽ n),
储ym储 ⫽ 2(ym, yn) ⫽
a
冮y
G
2
m(x)
dx.
a
The next example serves as an illustration of the material on orthogonal functions just
discussed.
EXAMPLE 2
Orthogonal Functions. Orthonormal Functions. Notation
The functions ym (x) ⫽ sin mx, m ⫽ 1, 2, Á form an orthogonal set on the interval ⫺p ⬉ x ⬉ p, because for
m ⫽ n we obtain by integration [see (11) in App. A3.1]
( ym, yn ) ⫽
冮
p
ⴚp
sin mx sin nx dx ⫽
2冮
1
p
cos (m ⫺ n)x dx ⫺
ⴚp
2冮
1
p
ⴚp
cos (m ⫹ n)x dx ⫽ 0, (m ⫽ n).
The norm 储 ym 储 ⫽ 1( ym, ym) equals 1p because
冮
储 ym 储2 ⫽ ( ym, ym ) ⫽
p
sin2 mx dx ⫽ p
(m ⫽ 1, 2, Á )
ⴚp
Hence the corresponding orthonormal set, obtained by division by the norm, is
sin x
1p
,
sin 2x
1p
sin 3x
,
1p
,
䊏
Á.
Theorem 1 shows that for any Sturm–Liouville problem, the eigenfunctions associated with
these problems are orthogonal. This means, in practice, if we can formulate a problem as a
Sturm–Liouville problem, then by this theorem we are guaranteed orthogonality.
THEOREM 1
Orthogonality of Eigenfunctions of Sturm–Liouville Problems
Suppose that the functions p, q, r, and p r in the Sturm–Liouville equation (1) are
real-valued and continuous and r (x) ⬎ 0 on the interval a ⬉ x ⬉ b. Let ym (x) and
yn (x) be eigenfunctions of the Sturm–Liouville problem (1), (2) that correspond to
different eigenvalues lm and ln , respectively. Then ym, yn are orthogonal on that
interval with respect to the weight function r, that is,
b
(6)
(ym, yn) ⫽
冮 r (x)y
m (x)yn (x)
dx ⫽ 0
(m ⫽ n).
a
If p (a) ⫽ 0, then (2a) can be dropped from the problem. If p(b) ⫽ 0, then (2b)
can be dropped. [It is then required that y and y r remain bounded at such a point,
and the problem is called singular, as opposed to a regular problem in which (2)
is used.]
If p(a) ⫽ p(b), then (2) can be replaced by the “periodic boundary conditions”
(7)
y(a) ⫽ y(b),
y r (a) ⫽ y r (b).
The boundary value problem consisting of the Sturm–Liouville equation (1) and the periodic
boundary conditions (7) is called a periodic Sturm–Liouville problem.
c11-a.qxd
10/30/10
1:25 PM
502
Page 502
CHAP. 11 Fourier Analysis
PROOF
By assumption, ym and yn satisfy the Sturm–Liouville equations
r ) r ⫹ (q ⫹ lmr) ym ⫽ 0
( pym
(py nr ) r ⫹ (q ⫹ lnr)yn ⫽ 0
respectively. We multiply the first equation by yn, the second by ⫺ym, and add,
(lm ⫺ ln)rym yn ⫽ ym( pynr ) r ⫺ yn( py rm) r ⫽ [( py nr ) ym ⫺ [( py rm) yn] r
where the last equality can be readily verified by performing the indicated differentiation
of the last expression in brackets. This expression is continuous on a ⬉ x ⬉ b since p and
p r are continuous by assumption and ym, yn are solutions of (1). Integrating over x from
a to b, we thus obtain
b
(8)
(lm ⫺ ln)
冮 ry
m yn
a
b
dx ⫽ [ p(y rn ym ⫺ y rm yn)]a
(a ⬍ b).
The expression on the right equals the sum of the subsequent Lines 1 and 2,
(9)
p(b)[ynr (b) ym(b) ⫺ y rm (b) yn(b)]
(Line 1)
⫺p (a)[ y nr (a)ym (a) ⫺ y m
r (a)yn (a)]
(Line 2).
Hence if (9) is zero, (8) with lm ⫺ ln ⫽ 0 implies the orthogonality (6). Accordingly,
we have to show that (9) is zero, using the boundary conditions (2) as needed.
Case 1. p (a) ⴝ p (b) ⴝ 0. Clearly, (9) is zero, and (2) is not needed.
Case 2. p (a) ⴝ 0, p (b) ⴝ 0. Line 1 of (9) is zero. Consider Line 2. From (2a) we have
k1 yn(a) ⫹ k 2 ynr (a) ⫽ 0,
k1 ym(a) ⫹ k 2 y m
r (a) ⫽ 0.
Let k 2 ⫽ 0. We multiply the first equation by ym (a), the last by ⫺yn (a) and add,
k 2[ynr (a)ym(a) ⫺ y m
r (a)yn(a)] ⫽ 0.
This is k 2 times Line 2 of (9), which thus is zero since k 2 ⫽ 0. If k 2 ⫽ 0, then k 1 ⫽ 0
by assumption, and the argument of proof is similar.
Case 3. p(a) ⴝ 0, p(b) ⴝ 0. Line 2 of (9) is zero. From (2b) it follows that Line 1 of (9)
is zero; this is similar to Case 2.
Case 4. p(a) ⴝ 0, p(b) ⴝ 0. We use both (2a) and (2b) and proceed as in Cases 2 and 3.
Case 5. p(a) ⴝ p(b). Then (9) becomes
p(b)[ ynr (b)ym(b) ⫺ ym
r (b)yn(b) ⫺ ynr (a)ym (a) ⫹ ym
r (a)yn(a)].
The expression in brackets [ Á ] is zero, either by (2) used as before, or more directly by
(7). Hence in this case, (7) can be used instead of (2), as claimed. This completes the
䊏
proof of Theorem 1.
EXAMPLE 3
Application of Theorem 1. Vibrating String
The ODE in Example 1 is a Sturm–Liouville equation with p ⫽ 1, q ⫽ 0, and r ⫽ 1. From Theorem 1 it follows
䊏
that the eigenfunctions ym ⫽ sin mx (m ⫽ 1, 2, Á ) are orthogonal on the interval 0 ⬉ x ⬉ p.
c11-a.qxd
10/30/10
1:25 PM
Page 503
SEC. 11.5 Sturm–Liouville Problems. Orthogonal Functions
503
Example 3 confirms, from this new perspective, that the trigonometric system underlying
the Fourier series is orthogonal, as we knew from Sec. 11.1.
EXAMPLE 4
Application of Theorem 1. Orthogonlity of the Legendre Polynomials
Legendre’s equation (1 ⫺ x 2) y s ⫺ 2xy r ⫹ n (n ⫹ 1) y ⫽ 0 may be written
[(1 ⫺ x 2) y r ] r ⫹ ly ⫽ 0
l ⫽ n (n ⫹ 1).
Hence, this is a Sturm–Liouville equation (1) with p ⫽ 1 ⫺ x 2, q ⫽ 0, and r ⫽ 1. Since p (⫺1) ⫽ p (1) ⫽ 0, we
need no boundary conditions, but have a “singular” Sturm–Liouville problem on the interval ⫺1 ⬉ x ⬉ 1. We
know that for n ⫽ 0, 1, Á , hence l ⫽ 0, 1 # 2, 2 # 3, Á , the Legendre polynomials Pn (x) are solutions of the
problem. Hence these are the eigenfunctions. From Theorem 1 it follows that they are orthogonal on that interval,
that is,
(10)
冮
1
Pm (x)Pn (x) dx ⫽ 0
(m ⫽ n).
䊏
⫺1
What we have seen is that the trigonometric system, underlying the Fourier series, is
a solution to a Sturm–Liouville problem, as shown in Example 1, and that this
trigonometric system is orthogonal, which we knew from Sec. 11.1 and confirmed in
Example 3.
PROBLEM SET 11.5
1. Proof of Theorem 1. Carry out the details in Cases 3
and 4.
2–6
set p ⫽ exp ( 兰 f dx), q ⫽ pg, r ⫽ hp. Why would you
do such a transformation?
ORTHOGONALITY
2. Normalization of eigenfunctions ym of (1), (2) means
that we multiply ym by a nonzero constant cm such that
cmym has norm 1. Show that z m ⫽ cym with any c ⫽ 0
is an eigenfunction for the eigenvalue corresponding
to ym.
3. Change of x. Show that if the functions y0 (x), y1 (x), Á
form an orthogonal set on an interval a ⬉ x ⬉ b (with
r (x) ⫽ 1), then the functions y0 (ct ⫹ k), y1 (ct ⫹ k),
Á , c ⬎ 0, form an orthogonal set on the interval
(a ⫺ k)>c ⬉ t ⬉ (b ⫺ k)>c.
4. Change of x. Using Prob. 3, derive the orthogonality
of 1, cos px, sin px, cos 2px, sin 2px, Á on
⫺1 ⬉ x ⬉ 1 (r (x) ⫽ 1) from that of 1, cos x, sin x,
cos 2x, sin 2x, Á on ⫺p ⬉ x ⬉ p.
5. Legendre polynomials. Show that the functions
Pn(cos u), n ⫽ 0, 1, Á , from an orthogonal set on the
interval 0 ⬉ u ⬉ p with respect to the weight function
sin u.
6. Tranformation to Sturm–Liouville form. Show that
y s ⫹ fy r ⫹ (g ⫹ lh) y ⫽ 0 takes the form (1) if you
7–15
STURM–LIOUVILLE PROBLEMS
Find the eigenvalues and eigenfunctions. Verify orthogonality. Start by writing the ODE in the form (1), using
Prob. 6. Show details of your work.
7. y s ⫹ ly ⫽ 0, y (0) ⫽ 0,
y (10) ⫽ 0
8. y s ⫹ ly ⫽ 0, y (0) ⫽ 0,
y (L) ⫽ 0
9. y s ⫹ ly ⫽ 0,
10. y s ⫹ ly ⫽ 0,
y (0) ⫽ 0,
y r(L) ⫽ 0
y (0) ⫽ y (1),
y r(0) ⫽ y r(1)
11. ( y r >x) r ⫹ (l ⫹ 1)y>x ⫽ 0, y (1) ⫽ 0, y (ep) ⫽ 0.
(Set x ⫽ et.)
3
12. y s ⫺ 2y r ⫹ (l ⫹ 1) y ⫽ 0, y (0) ⫽ 0, y (1) ⫽ 0
13. y s ⫹ 8y r ⫹ (l ⫹ 16) y ⫽ 0, y (0) ⫽ 0, y (p) ⫽ 0
14. TEAM PROJECT. Special Functions. Orthogonal
polynomials play a great role in applications. For
this reason, Legendre polynomials and various other
orthogonal polynomials have been studied extensively;
see Refs. [GenRef1], [GenRef10] in App. 1. Consider
some of the most important ones as follows.
c11-a.qxd
10/30/10
1:25 PM
504
Page 504
CHAP. 11 Fourier Analysis
(a) Chebyshev polynomials6 of the first and second
kind are defined by
that Tn (x), n ⫽ 0, 1, 2, 3, satisfy the Chebyshev
equation
Tn (x) ⫽ cos (n arccos x)
(1 ⫺ x 2)y s ⫺ xy r ⫹ n 2y ⫽ 0.
Un (x) ⫽
sin [(n ⫹ 1) arccos x]
21 ⫺ x
(b) Orthogonality on an infinite interval: Laguerre
polynomials7 are defined by L 0 ⫽ 1, and
2
L n(x) ⫽
respectively, where n ⫽ 0, 1, Á . Show that
T0 ⫽ 1,
T1(x) ⫽ x,
T2(x) ⫽ 2x 2 ⫺ 1.
T3(x) ⫽ 4x 3 ⫺ 3x,
U0 ⫽ 1,
U1(x) ⫽ 2x,
U2(x) ⫽ 4x 2 ⫺ 1,
n
n ⴚx
ex d (x e )
,
n!
dx n
n ⫽ 1, 2, Á .
Show that
L n(x) ⫽ 1 ⫺ x,
L 2 (x) ⫽ 1 ⫺ 2x ⫹ x 2>2,
U3(x) ⫽ 8x 3 ⫺ 4x.
L 3 (x) ⫽ 1 ⫺ 3x ⫹ 3x 2>2 ⫺ x 3>6.
Show that the Chebyshev polynomials Tn(x) are
orthogonal on the interval ⫺1 ⬉ x ⬉ 1 with respect
to the weight function r (x) ⫽ 1> 21 ⫺ x 2 . (Hint.
To evaluate the integral, set arccos x ⫽ u.) Verify
Prove that the Laguerre polynomials are orthogonal on
the positive axis 0 ⬉ x ⬍ ⬁ with respect to the weight
function r (x) ⫽ eⴚx. Hint. Since the highest power in
L m is x m, it suffices to show that 兰 eⴚxx kL n dx ⫽ 0
for k ⬍ n. Do this by k integrations by parts.
11.6
Orthogonal Series.
Generalized Fourier Series
Fourier series are made up of the trigonometric system (Sec. 11.1), which is orthogonal,
and orthogonality was essential in obtaining the Euler formulas for the Fourier coefficients.
Orthogonality will also give us coefficient formulas for the desired generalized Fourier
series, including the Fourier–Legendre series and the Fourier–Bessel series. This generalization is as follows.
Let y0, y1, y2, Á be orthogonal with respect to a weight function r (x) on an interval
a ⬉ x ⬉ b, and let f (x) be a function that can be represented by a convergent series
⬁
(1)
f (x) ⫽ a am ym (x) ⫽ a0 y0 (x) ⫹ a1 y1 (x) ⫹ Á .
m⫽0
This is called an orthogonal series, orthogonal expansion, or generalized Fourier series.
If the ym are the eigenfunctions of a Sturm–Liouville problem, we call (1) an eigenfunction
expansion. In (1) we use again m for summation since n will be used as a fixed order of
Bessel functions.
Given f (x), we have to determine the coefficients in (1), called the Fourier constants
of f (x) with respect to y0, y1, Á . Because of the orthogonality, this is simple. Similarly
to Sec. 11.1, we multiply both sides of (1) by r (x)yn (x) (n fixed ) and then integrate on
6
PAFNUTI CHEBYSHEV (1821–1894), Russian mathematician, is known for his work in approximation
theory and the theory of numbers. Another transliteration of the name is TCHEBICHEF.
7
EDMOND LAGUERRE (1834–1886), French mathematician, who did research work in geometry and in
the theory of infinite series.
c11-a.qxd
10/30/10
1:25 PM
Page 505
SEC. 11.6 Orthogonal Series. Generalized Fourier Series
505
both sides from a to b. We assume that term-by-term integration is permissible. (This is
justified, for instance, in the case of “uniform convergence,” as is shown in Sec. 15.5.)
Then we obtain
( f, yn ) ⫽
冮
b
r fyn dx ⫽
a
冮
b
a
⬁
⬁
m⫽ 0
m⫽ 0
冮
r a a am ym b yn dx ⫽ a am
b
⬁
rym yn dx ⫽ a am (ym, yn).
m⫽ 0
a
Because of the orthogonality all the integrals on the right are zero, except when m ⫽ n.
Hence the whole infinite series reduces to the single term
a n (yn, yn ) ⫽ an 储 y n 储 2.
( f, yn ) ⫽ an 储 yn 储2.
Thus
Assuming that all the functions yn have nonzero norm, we can divide by 储yn储2; writing again
m for n, to be in agreement with (1), we get the desired formula for the Fourier constants
am ⫽
(2)
( f, ym)
储 ym 储 2
b
冮 r (x) f (x)y
1
⫽
m (x)
储 ym 储 2
dx
(n ⫽ 0, 1, Á ).
a
This formula generalizes the Euler formulas (6) in Sec. 11.1 as well as the principle of
their derivation, namely, by orthogonality.
EXAMPLE 1
Fourier–Legendre Series
A Fourier–Legendre series is an eigenfunction expansion
ⴥ
f (x) ⫽ a amPm (x) ⫽ a0P0 ⫹ a1P1 (x) ⫹ a2P2 (x) ⫹ Á ⫽ a0 ⫹ a1x ⫹ a2 ( 32 x 2 ⫺ 12 ) ⫹ Á
m⫽0
in terms of Legendre polynomials (Sec. 5.3). The latter are the eigenfunctions of the Sturm–Liouville problem
in Example 4 of Sec. 11.5 on the interval ⫺1 ⬉ x ⬉ 1. We have r (x) ⫽ 1 for Legendre’s equation, and (2)
gives
am ⫽
(3)
2m ⫹ 1
2
冮
1
m ⫽ 0, 1, Á
f (x)Pm (x) dx,
⫺1
because the norm is
储 Pm 储 ⫽
(4)
G
冮
1
Pm (x)2 dx ⫽
ⴚ1
2
(m ⫽ 0, 1, Á )
B 2m ⫹ 1
as we state without proof. The proof of (4) is tricky; it uses Rodrigues’s formula in Problem Set 5.2 and a
reduction of the resulting integral to a quotient of gamma functions.
For instance, let f (x) ⫽ sin px. Then we obtain the coefficients
am ⫽
2m ⫹ 1
2
1
冮 (sin px)P
m (x)
ⴚ1
dx,
thus
a1 ⫽
2 冮
3
1
x sin px dx ⫽
ⴚ1
3
,
p ⫽ 0.95493
etc.
c11-a.qxd
11/1/10
10:39 PM
506
Page 506
CHAP. 11 Fourier Analysis
Hence the Fourier–Legendre series of sin px is
sin px ⫽ 0.95493P1 (x) ⫺ 1.15824P3 (x) ⫹ 0.21929P5 (x) ⫺ 0.01664P7 (x) ⫹ 0.00068P9 (x)
⫺ 0.00002P11 (x) ⫹ Á .
The coefficient of P13 is about 3 # 10ⴚ7. The sum of the first three nonzero terms gives a curve that practically
coincides with the sine curve. Can you see why the even-numbered coefficients are zero? Why a3 is the absolutely
biggest coefficient?
䊏
EXAMPLE 2
Fourier–Bessel Series
These series model vibrating membranes (Sec. 12.9) and other physical systems of circular symmetry. We derive
these series in three steps.
Step 1. Bessel’s equation as a Sturm–Liouville equation. The Bessel function Jn (x) with fixed integer n ⭌ 0
satisfies Bessel’s equation (Sec. 5.5)
##
#
⬃
⬃
⬃2
2
⬃
x⬃2J n ( x⬃) ⫹ xJ
n ( x ) ⫹ ( x ⫺ n )Jn( x ) ⫽ 0
⬃
2
⬃2
⬃
⬃
where Jn ⫽ dJn>d
by the chain rule, Jn ⫽ dJn>d x⬃ ⫽
## x and J n2 ⫽ d Jn>d x . We set x ⫽ kx. Then x ⫽ x >k and
2
(dJn>dx)/k and J n ⫽ Jns >k . In the first two terms of Bessel’s equation, k and k drop out and we obtain
#
##
#
x 2Jns (kx) ⫹ xJ nr (kx) ⫹ (k 2x 2 ⫺ n 2)Jn(kx) ⫽ 0.
Dividing by x and using (xJnr (kx)) r ⫽ xJ ns (kx) ⫹ Jnr (kx) gives the Sturm–Liouville equation
[xJnr (kx)] r ⫹ a⫺
(5)
n2
⫹ lxb Jn(kx) ⫽ 0
x
l ⫽ k2
with p (x) ⫽ x, q (x) ⫽ ⫺n 2>x, r (x) ⫽ x, and parameter l ⫽ k 2. Since p (0) ⫽ 0, Theorem 1 in Sec. 11.5
implies orthogonality on an interval 0 ⬉ x ⬉ R (R given, fixed) of those solutions Jn(kx) that are zero at
x ⫽ R, that is,
Jn(kR) ⫽ 0
(6)
(n fixed).
Note that q (x) ⫽ ⫺n 2>x is discontinuous at 0, but this does not affect the proof of Theorem 1.
Step 2. Orthogonality. It can be shown (see Ref. [A13]) that Jn( ⬃x ) has infinitely many zeros, say,
苲
x ⫽ an,1 ⬍ an,2 ⬍ Á (see Fig. 110 in Sec. 5.4 for n ⫽ 0 and 1). Hence we must have
kR ⫽ an,m
(7)
thus
k n,m ⫽ an,m>R
(m ⫽ 1, 2, Á ).
This proves the following orthogonality property.
THEOREM 1
Orthogonality of Bessel Functions
For each fixed nonnegative integer n the sequence of Bessel functions of the first
kind Jn(k n,1x), Jn(k n,2x), Á with k n,m as in (7) forms an orthogonal set on the
interval 0 ⬉ x ⬉ R with respect to the weight function r (x) ⫽ x, that is,
(8)
冮
R
xJn (k n,mx)Jn(k n, jx) dx ⫽ 0
( j ⫽ m, n fixed).
0
Hence we have obtained infinitely many orthogonal sets of Bessel functions, one for each of J0, J1, J2, Á .
Each set is orthogonal on an interval 0 ⬉ x ⬉ R with a fixed positive R of our choice and with respect to
the weight x. The orthogonal set for Jn is Jn(k n,1x), Jn(k n,2x), Jn(k n,3x), Á , where n is fixed and k n,m is
given by (7).
c11-a.qxd
10/30/10
1:25 PM
Page 507
SEC. 11.6 Orthogonal Series. Generalized Fourier Series
507
Step 3. Fourier–Bessel series. The Fourier–Bessel series corresponding to Jn (n fixed) is
ⴥ
f (x) ⫽ a amJn(k n,mx) ⫽ a1Jn(k n,1x) ⫹ a2Jn(k n,2x) ⫹ a3Jn(k n,3x) ⫹ Á
(9)
(n fixed).
m⫽1
The coefficients are (with an,m ⫽ k n,mR)
am ⫽
(10)
冮
2
R
R2J 2n⫹1(an,m) 0
m ⫽ 1, 2, Á
x f (x) Jn(k n,mx) dx,
because the square of the norm is
储 Jn(k n,mx) 储 2 ⫽
(11)
R
冮
xJn2 (k n,mx) dx ⫽
R2
0
2
J 2n⫹1(k n,mR)
䊏
as we state without proof (which is tricky; see the discussion beginning on p. 576 of [A13]).
EXAMPLE 3
Special Fourier–Bessel Series
For instance, let us consider f (x) ⫽ 1 ⫺ x 2 and take R ⫽ 1 and n ⫽ 0 in the series (9), simply writing l for
a0,m. Then k n,m ⫽ a0,m ⫽ l ⫽ 2.405, 5.520, 8.654, 11.792, etc. (use a CAS or Table A1 in App. 5). Next we
calculate the coefficients am by (10)
am ⫽
冮
J (l)
2
2
1
1
x(1 ⫺ x 2)J0(lx) dx.
0
This can be integrated by a CAS or by formulas as follows. First use [xJ1(lx)] r ⫽ lxJ0(lx) from Theorem 1 in
Sec. 5.4 and then integration by parts,
am ⫽
冮
J (l)
2
2
1
1
0
x(1 ⫺ x 2)J0(lx) dx ⫽
1
1
1
(1 ⫺ x 2)xJ1(lx) ` ⫺
2
l
l
0
J1 (l)
2
c
冮
1
0
xJ1(lx)(⫺2x) dx d .
The integral-free part is zero. The remaining integral can be evaluated by [x 2J2(lx)] r ⫽ lx 2J1(lx) from Theorem 1
in Sec. 5.4. This gives
am ⫽
4J2 (l)
l2J12 (l)
(l ⫽ a0,m).
Numeric values can be obtained from a CAS (or from the table on p. 409 of Ref. [GenRef1] in App. 1, together
with the formula J2 ⫽ 2x ⴚ1J1 ⫺ J0 in Theorem 1 of Sec. 5.4). This gives the eigenfunction expansion of 1 ⫺ x 2
in terms of Bessel functions J0, that is,
1 ⫺ x 2 ⫽ 1.1081J0(2.405x) ⫺ 0.1398J0(5.520x) ⫹ 0.0455J0(8.654x) ⫺ 0.0210J0(11.792x) ⫹ Á.
A graph would show that the curve of 1 ⫺ x 2 and that of the sum of first three terms practically coincide.
䊏
Mean Square Convergence. Completeness
Ideas on approximation in the last section generalize from Fourier series to orthogonal series
(1) that are made up of an orthonormal set that is “complete,” that is, consists of “sufficiently
many” functions so that (1) can represent large classes of other functions (definition below).
In this connection, convergence is convergence in the norm, also called mean-square
convergence; that is, a sequence of functions f k is called convergent with the limit f if
(12*)
lim 储 f ⫺ f 储 ⫽ 0;
k :⬁
k
c11-a.qxd
10/30/10
508
1:25 PM
Page 508
CHAP. 11 Fourier Analysis
written out by (5) in Sec. 11.5 (where we can drop the square root, as this does not affect
the limit)
b
(12)
冮 r (x)[ f (x) ⫺ f (x)]
lim
2
k
k :⬁
dx ⫽ 0.
a
Accordingly, the series (1) converges and represents f if
b
(13)
lim
k :⬁
冮 r (x)[s (x) ⫺ f (x)]
2
k
dx ⫽ 0
a
where sk is the kth partial sum of (1).
k
sk(x) ⫽ a am ym(x).
(14)
m⫽0
Note that the integral in (13) generalizes (3) in Sec. 11.4.
We now define completeness. An orthonormal set y0, y1, Á on an interval a ⬉ x ⬉ b
is complete in a set of functions S defined on a ⬉ x ⬉ b if we can approximate every
f belonging to S arbitrarily closely in the norm by a linear combination a0y0 ⫹
a1y1 ⫹ Á ⫹ akyk, that is, technically, if for every P ⬎ 0 we can find constants a0, Á , ak
(with k large enough) such that
储 f ⫺ (a0y0 ⫹ Á ⫹ akyk)储 ⬍ P.
(15)
Ref. [GenRef7] in App. 1 uses the more modern term total for complete.
We can now extend the ideas in Sec. 11.4 that guided us from (3) in Sec. 11.4 to Bessel’s
and Parseval’s formulas (7) and (8) in that section. Performing the square in (13) and
using (14), we first have (analog of (4) in Sec. 11.4)
冮
b
r (x)[sk (x) ⫺ f (x)]2 dx ⫽
a
冮
b
冮
b
rsk2 dx ⫺ 2
a
⫽
a
冮
b
b
rfsk dx ⫹
a
冮 rf
2
dx
a
k
k
2
r c a am ym d dx ⫺ 2 a am
m⫽0
m⫽0
冮
b
a
rfym dx ⫹
冮
b
rf 2 dx.
a
The first integral on the right equals ga 2m because 兰 rymyl dx ⫽ 0 for m ⫽ l, and
2
2
兰 rym dx ⫽ 1. In the second sum on the right, the integral equals am, by (2) with 储 ym 储 ⫽ 1.
Hence the first term on the right cancels half of the second term, so that the right side
reduces to (analog of (6) in Sec. 11.4)
k
2
⫺ a am
⫹
m⫽0
冮
b
rf 2 dx.
a
This is nonnegative because in the previous formula the integrand on the left is nonnegative
(recall that the weight r (x) is positive!) and so is the integral on the left. This proves the
important Bessel’s inequality (analog of (7) in Sec. 11.4)
k
(16)
2
2
a am ⬉ 储 f 储 ⫽
m⫽0
冮
b
a
r (x) f (x)2 dx
(k ⫽ 1, 2, Á ),
c11-a.qxd
10/30/10
1:25 PM
Page 509
SEC. 11.6 Orthogonal Series. Generalized Fourier Series
509
Here we can let k : ⬁ , because the left sides form a monotone increasing sequence that
is bounded by the right side, so that we have convergence by the familiar Theorem 1 in
App. A.3.3 Hence
ⴥ
2
2
a am ⬉ 储 f 储 .
(17)
m⫽0
Furthermore, if y0, y1, Á is complete in a set of functions S, then (13) holds for every f
belonging to S. By (13) this implies equality in (16) with k : ⬁. Hence in the case of
completeness every f in S saisfies the so-called Parseval equality (analog of (8) in Sec. 11.4)
ⴥ
(18)
a
2
am
⫽
储f储 ⫽
2
m⫽0
冮
b
r (x) f (x)2 dx.
a
As a consequence of (18) we prove that in the case of completeness there is no function
orthogonal to every function of the orthonormal set, with the trivial exception of a function
of zero norm:
THEOREM 2
Completeness
Let y0, y1, Á be a complete orthonormal set on a ⬉ x ⬉ b in a set of functions S.
Then if a function f belongs to S and is orthogonal to every ym , it must have norm
zero. In particular, if f is continuous, then f must be identically zero.
PROOF
Since f is orthogonal to every ym, the left side of (18) must be zero. If f is continuous,
then 储 f 储 ⫽ 0 implies f (x) ⬅ 0, as can be seen directly from (5) in Sec. 11.5 with f instead
of ym because r (x) ⬎ 0 by assumption.
䊏
PROBLEM SET 11.6
1–7
FOURIER–LEGENDRE SERIES
Showing the details, develop
63x 5 ⫺ 90x 3 ⫹ 35x
(x ⫹ 1)2
1 ⫺ x4
1, x, x 2, x 3, x 4
Prove that if f (x) is even (is odd, respectively), its
Fourier–Legendre series contains only Pm (x) with even
m (only Pm (x) with odd m, respectively). Give examples.
6. What can you say about the coefficients of the Fourier–
Legendre series of f (x) if the Maclaurin series of f (x)
contains only powers x 4m (m ⫽ 0, 1, 2, Á )?
7. What happens to the Fourier–Legendre series of a
polynomial f (x) if you change a coefficient of f (x)?
Experiment. Try to prove your answer.
1.
2.
3.
4.
5.
8–13
CAS EXPERIMENT
FOURIER–LEGENDRE SERIES. Find and graph (on
common axes) the partial sums up to Sm0 whose graph
practically coincides with that of f (x) within graphical
accuracy. State m 0. On what does the size of m 0 seem to
depend?
8. f ( x) ⫽ sin px
9. f ( x) ⫽ sin 2px
10. f ( x) ⫽ eⴚx
2
11. f ( x) ⫽ (1 ⫹ x 2)ⴚ1
12. f ( x) ⫽ J0(a0,1 x), a0,1 ⫽ the first positive zero
of J0( x)
13. f (x) ⫽ J0(a0,2 x), a0,2 ⫽ the second positive zero
of J0(x)
c11-a.qxd
10/30/10
1:25 PM
510
Page 510
CHAP. 11 Fourier Analysis
14. TEAM PROJECT. Orthogonality on the Entire Real
Axis. Hermite Polynomials.8 These orthogonal polynomials are defined by He0 (1) ⫽ 1 and
(19)
>2
2
Hen (x) ⫽ (⫺1)nex
dn
(eⴚx >2),
2
dx n
n ⫽ 1, 2, Á .
(22)
REMARK. As is true for many special functions, the
literature contains more than one notation, and one sometimes defines as Hermite polynomials the functions
d neⴚx
2
H n*(x) ⫽ (⫺1)nex
dx n
.
This differs from our definition, which is preferred in
applications.
(a) Small Values of n. Show that
He1 (x) ⫽ x,
He3 (x) ⫽ x 3 ⫺ 3x,
He2 (x) ⫽ x 2 ⫺ 1,
He4 (x) ⫽ x 4 ⫺ 6x 2 ⫹ 3.
(b) Generating Function. A generating function of the
Hermite polynomials is
(20)
>2
2
etxⴚt
ⴥ
⫽ a an (x) t n
n⫽0
because Hen (x) ⫽ n! an(x). Prove this. Hint: Use the
formula for the coefficients of a Maclaurin series and
note that tx ⫺ 12 t 2 ⫽ 12 x 2 ⫺ 12 (x ⫺ t)2.
(c) Derivative. Differentiating the generating function with respect to x, show that
(21)
Henr (x) ⫽ nHen⫺1 (x).
(d) Orthogonality on the x-Axis needs a weight function
that goes to zero sufficiently fast as x : ⫾⬁, (Why?)
11.7
Henr (x) ⫽ xHen(x) ⫺ Hen⫹1 (x).
Using this with n ⫺ 1 instead of n and (21), show that
y ⫽ Hen(x) satisfies the ODE
(23)
2
H 0* ⫽ 1,
Show that the Hermite polynomials are orthogonal on
⫺⬁ ⬍ x ⬍2 ⬁ with respect to the weight function
r (x) ⫽ eⴚx >2. Hint. Use integration by parts and (21).
(e) ODEs. Show that
y s ⫽ xy r ⫹ ny ⫽ 0.
Show that w ⫽ e ⴚx
equation
>4
2
y is a solution of Weber’s
(24) w s ⫹ (n ⫹ 12 ⫺ 14 x 2) w ⫽ 0
(n ⫽ 0, 1, Á ).
15. CAS EXPERIMENT. Fourier–Bessel Series. Use
Example 2 and R ⫽ 1, so that you get the series
(25)
f (x) ⫽ a1J0 (a0,1x) ⫹ a2J0 (a0,2x)
⫹ a3J0 (a0,3x) ⫹ Á
With the zeros a0,1 a0,2, Á from your CAS (see also
Table A1 in App. 5).
(a) Graph the terms J0 (a0,1x), Á , J0 (a0,10 x) for
0 ⬉ x ⬉ 1 on common axes.
(b) Write a program for calculating partial sums of (25).
Find out for what f (x) your CAS can evaluate the
integrals. Take two such f (x) and comment empirically
on the speed of convergence by observing the decrease
of the coefficients.
(c) Take f (x) ⫽ 1 in (25) and evaluate the integrals
for the coefficients analytically by (21a), Sec. 5.4, with
v ⫽ 1. Graph the first few partial sums on common
axes.
Fourier Integral
Fourier series are powerful tools for problems involving functions that are periodic or are of
interest on a finite interval only. Sections 11.2 and 11.3 first illustrated this, and various further
applications follow in Chap. 12. Since, of course, many problems involve functions that are
nonperiodic and are of interest on the whole x-axis, we ask what can be done to extend the
method of Fourier series to such functions. This idea will lead to “Fourier integrals.”
In Example 1 we start from a special function fL of period 2L and see what happens to
its Fourier series if we let L : ⬁. Then we do the same for an arbitrary function fL of
period 2L. This will motivate and suggest the main result of this section, which is an
integral representation given in Theorem 1 below.
8
CHARLES HERMITE (1822–1901), French mathematician, is known for his work in algebra and number
theory. The great HENRI POINCARÉ (1854–1912) was one of his students.
c11-a.qxd
10/30/10
1:25 PM
Page 511
SEC. 11.7 Fourier Integral
EXAMPLE 1
511
Rectangular Wave
Consider the periodic rectangular wave fL (x) of period 2L ⬎ 2 given by
0
if
⫺L ⬍ x ⬍ ⫺1
fL (x) ⫽ d 1
if
⫺1 ⬍ x ⬍
0
if
1
1 ⬍ x ⬍ L.
The left part of Fig. 280 shows this function for 2L ⫽ 4, 8, 16 as well as the nonperiodic function f(x), which
we obtain from fL if we let L : ⬁,
f (x) ⫽ lim fL (x) ⫽ e
L:ⴥ
1
if ⫺1 ⬍ x ⬍ 1
0
otherwise.
We now explore what happens to the Fourier coefficients of fL as L increases. Since fL is even, bn ⫽ 0 for
all n. For an the Euler formulas (6), Sec. 11.2, give
a0 ⫽
2L 冮
1
1
dx ⫽
ⴚ1
1
L
,
an ⫽
L冮
1
1
cos
ⴚ1
npx
L
dx ⫽
1
cos
L冮
2
0
npx
L
dx ⫽
2 sin (np>L)
L
np>L
.
This sequence of Fourier coefficients is called the amplitude spectrum of fL because ƒ an ƒ is the maximum
amplitude of the wave an cos (npx>L). Figure 280 shows this spectrum for the periods 2L ⫽ 4, 8, 16. We see
that for increasing L these amplitudes become more and more dense on the positive wn-axis, where wn ⫽ np>L.
Indeed, for 2L ⫽ 4, 8, 16 we have 1, 3, 7 amplitudes per “half-wave” of the function (2 sin wn)>(Lwn) (dashed
in the figure). Hence for 2L ⫽ 2k we have 2kⴚ1 ⫺ 1 amplitudes per half-wave, so that these amplitudes will
eventually be everywhere dense on the positive wn-axis (and will decrease to zero).
The outcome of this example gives an intuitive impression of what about to expect if we turn from our special
function to an arbitrary one, as we shall do next.
䊏
Waveform fL(x)
1
Amplitude spectrum an(wn)
n=1
fL(x)
–2
n=5
0
x
2
wn
n=7
n=3
2L = 4
1
_
2
n=2
fL(x)
–4
n = 10
0
x
4
n=6
2L = 8
fL(x)
–8
wn = nπ/L
π
1
_
4
0
8
wn
n=4
n = 20
x
n = 12
2L = 16
f(x)
–1 0 1
n = 14
x
Fig. 280. Waveforms and amplitude spectra in Example 1
n = 28
wn
c11-a.qxd
10/30/10
512
1:25 PM
Page 512
CHAP. 11 Fourier Analysis
From Fourier Series to Fourier Integral
We now consider any periodic function fL (x) of period 2L that can be represented by a
Fourier series
ⴥ
fL (x) ⫽ a0 ⫹ a (an cos wnx ⫹ bn sin wnx),
wn ⫽
n⫽1
np
L
and find out what happens if we let L : ⬁. Together with Example 1 the present
calculation will suggest that we should expect an integral (instead of a series) involving
cos wx and sin wx with w no longer restricted to integer multiples w ⫽ wn ⫽ np>L
of p>L but taking all values. We shall also see what form such an integral might
have.
If we insert an and bn from the Euler formulas (6), Sec. 11.2, and denote the variable
of integration by v, the Fourier series of fL (x) becomes
1
2L
fL (x) ⫽
冮
L
1 ⴥ
a c cos wnx
L n⫽1
fL (v) dv ⫹
ⴚL
冮
L
fL (v) cos wnv dv
ⴚL
⫹ sin wnx
冮
L
fL (v) sin wnv dv d .
ⴚL
We now set
¢w ⫽ wn⫹1 ⫺ wn ⫽
(n ⫹ 1)p
L
⫺
np
p
⫽ .
L
L
Then 1>L ⫽ ¢w> p, and we may write the Fourier series in the form
(1)
fL (x) ⫽
1
2L
冮
L
ⴥ
fL (v) dv ⫹ 1 a c (cos wn x) ¢w
ⴚL
p n⫽1
⫹ (sin wnx)¢w
冮
L
冮
L
fL (v) cos wnv dv
ⴚL
fL (v) sin wnv dv d .
ⴚL
This representation is valid for any fixed L, arbitrarily large, but finite.
We now let L : ⬁ and assume that the resulting nonperiodic function
f (x) ⫽ lim fL (x)
L :⬁
is absolutely integrable on the x-axis; that is, the following (finite!) limits exist:
(2)
lim
a :⫺⬁
冮
0
a
ƒ f (x) ƒ dx ⫹ lim
b :⬁
冮
b
0
ƒ f (x) ƒ dx awritten
冮
ⴥ
ƒ f (x) ƒ dxb.
ⴚⴥ
Then 1>L : 0, and the value of the first term on the right side of (1) approaches zero.
Also ¢w ⫽ p>L : 0 and it seems plausible that the infinite series in (1) becomes an
c11-a.qxd
10/30/10
1:25 PM
Page 513
SEC. 11.7 Fourier Integral
513
integral from 0 to ⬁, which represents f(x), namely,
(3)
1
f (x) ⫽ p
冮
ⴥ
0
c cos wx
ⴥ
冮
f (v) cos wv dv ⫹ sin wx
ⴚⴥ
冮
ⴥ
f (v) sin wv dv d dw.
ⴚⴥ
If we introduce the notations
(4)
1
A (w) ⫽ p
冮
ⴥ
1
B (w) ⫽ p
f (v) cos wv dv,
ⴚⴥ
冮
ⴥ
f (v) sin wv dv
ⴚⴥ
we can write this in the form
f (x) ⫽
(5)
冮
ⴥ
[A (w) cos wx ⫹ B (w) sin wx] dw.
0
This is called a representation of f (x) by a Fourier integral.
It is clear that our naive approach merely suggests the representation (5), but by no
means establishes it; in fact, the limit of the series in (1) as ¢w approaches zero is not
the definition of the integral (3). Sufficient conditions for the validity of (5) are as follows.
THEOREM 1
Fourier Integral
If f (x) is piecewise continuous (see Sec. 6.1) in every finite interval and has a righthand derivative and a left-hand derivative at every point (see Sec 11.1) and if the
integral (2) exists, then f (x) can be represented by a Fourier integral (5) with A and
B given by (4). At a point where f (x) is discontinuous the value of the Fourier integral
equals the average of the left- and right-hand limits of f (x) at that point (see Sec. 11.1).
(Proof in Ref. [C12]; see App. 1.)
Applications of Fourier Integrals
The main application of Fourier integrals is in solving ODEs and PDEs, as we shall see
for PDEs in Sec. 12.6. However, we can also use Fourier integrals in integration and in
discussing functions defined by integrals, as the next example.
EXAMPLE 2
Single Pulse, Sine Integral. Dirichlet’s Discontinuous Factor. Gibbs Phenomenon
Find the Fourier integral representation of the function
f (x) ⫽ e
1
if
ƒxƒ ⬍ 1
0
if
ƒxƒ ⬎ 1
(Fig. 281)
f(x)
1
–1
0
1
Fig. 281. Example 2
x
c11-a.qxd
10/30/10
514
1:25 PM
Page 514
CHAP. 11 Fourier Analysis
Solution.
From (4) we obtain
A (w) ⫽
p 冮
1
⬁
f (v) cos wv dv ⫽
ⴚ⬁
p 冮
1
冮
1
1
cos wv dv ⫽
ⴚ1
B (w) ⫽
1
p
sin wv
`
pw
1
⫽
ⴚ1
2 sin w
pw
sin wv dv ⫽ 0
ⴚ1
and (5) gives the answer
f (x) ⫽
(6)
p 冮
2
⬁
cos wx sin w
dw.
w
0
The average of the left- and right-hand limits of f (x) at x ⫽ 1 is equal to (1 ⫹ 0)>2, that is, 12.
Furthermore, from (6) and Theorem 1 we obtain (multiply by p>2)
冮
(7)
⬁
0
p>2
cos wx sin w
dw ⫽ dp>4
w
0
if
0 ⬉ x ⬍ 1,
if
x ⫽ 1,
if
x ⬎ 1.
We mention that this integral is called Dirichlet’s discontinous factor. (For P. L. Dirichlet see Sec. 10.8.)
The case x ⫽ 0 is of particular interest. If x ⫽ 0, then (7) gives
冮
(8*)
⬁
sin w
0
p
dw ⫽
w
2
.
We see that this integral is the limit of the so-called sine integral
冮
Si(u) ⫽
(8)
u
0
sin w
dw
w
as u : ⬁ . The graphs of Si(u) and of the integrand are shown in Fig. 282.
In the case of a Fourier series the graphs of the partial sums are approximation curves of the curve of the
periodic function represented by the series. Similarly, in the case of the Fourier integral (5), approximations are
obtained by replacing ⬁ by numbers a. Hence the integral
2
(9)
p
冮
a
0
cos wx sin w
dw
w
approximates the right side in (6) and therefore f (x).
y
Integrand
Si(u)
π–
2
1
0.5
–4π
–3π
–2π
–1π 0
–0.5
1π
2π
3π
4π u
–1
– π–
2
Fig. 282. Sine integral Si(u) and integrand
c11-a.qxd
10/30/10
1:25 PM
Page 515
SEC. 11.7 Fourier Integral
515
y
y
y
a = 16
a=8
–2 –1 0
1
–2 –1 0
2x
1
a = 32
2x
–2 –1 0
1
2x
Fig. 283. The integral (9) for a ⴝ 8, 16, and 32, illustrating
the development of the Gibbs phenomenon
Figure 283 shows oscillations near the points of discontinuity of f (x). We might expect that these oscillations
disappear as a approaches infinity. But this is not true; with increasing a, they are shifted closer to the points
x ⫽ ⫾1. This unexpected behavior, which also occurs in connection with Fourier series (see Sec. 11.2), is known
as the Gibbs phenomenon. We can explain it by representing (9) in terms of sine integrals as follows. Using
(11) in App. A3.1, we have
2
p
冮
a
0
cos wx sin w
1
dw ⫽
w
p
冮
a
0
sin (w ⫹ wx)
1
dw ⫹
w
p
冮
a
0
sin (w ⫺ wx)
dw.
w
In the first integral on the right we set w ⫹ wx ⫽ t. Then dw>w ⫽ dt>t, and 0 ⬉ w ⬉ a corresponds to
0 ⬉ t ⬉ (x ⫹ 1) a. In the last integral we set w ⫺ wx ⫽ ⫺t. Then dw>w ⫽ dt>t, and 0 ⬉ w ⬉ a corresponds to
0 ⬉ t ⬉ (x ⫺ 1) a. Since sin (⫺t) ⫽ ⫺sin t, we thus obtain
p 冮
2
a
0
cos wx sin w
1
dw ⫽
w
p
冮
(x⫹1) a
0
sin t
1
dt ⫺
t
p
冮
(xⴚ1) a
0
sin t
dt.
t
From this and (8) we see that our integral (9) equals
1
1
p Si(a[x ⫹ 1]) ⫺ p Si(a[x ⫺ 1])
and the oscillations in Fig. 283 result from those in Fig. 282. The increase of a amounts to a transformation
of the scale on the axis and causes the shift of the oscillations (the waves) toward the points of discontinuity
䊏
⫺1 and 1.
Fourier Cosine Integral and Fourier Sine Integral
Just as Fourier series simplify if a function is even or odd (see Sec. 11.2), so do Fourier
integrals, and you can save work. Indeed, if f has a Fourier integral representation and is
even, then B (w) ⫽ 0 in (4). This holds because the integrand of B (w) is odd. Then (5)
reduces to a Fourier cosine integral
(10)
f (x) ⫽
冮
⬁
A (w) cos wx dw
where
2
A (w) ⫽ p
0
冮
⬁
f (v) cos wv dv.
0
Note the change in A (w): for even f the integrand is even, hence the integral from ⫺⬁ to
⬁ equals twice the integral from 0 to ⬁ , just as in (7a) of Sec. 11.2.
Similarly, if f has a Fourier integral representation and is odd, then A (w) ⫽ 0 in (4). This
is true because the integrand of A (w) is odd. Then (5) becomes a Fourier sine integral
(11)
f (x) ⫽
冮
⬁
0
B (w) sin wx dw
where
2
B (w) ⫽ p
冮
⬁
0
f (v) sin wv dv.
c11-a.qxd
10/30/10
1:25 PM
516
Page 516
CHAP. 11 Fourier Analysis
Note the change of B (w) to an integral from 0 to ⬁ because B (w) is even (odd times odd
is even).
Earlier in this section we pointed out that the main application of the Fourier integral
representation is in differential equations. However, these representations also help in
evaluating integrals, as the following example shows for integrals from 0 to ⬁ .
EXAMPLE 3
1
Laplace Integrals
We shall derive the Fourier cosine and Fourier sine integrals of f (x) ⫽ eⴚkx, where x ⬎ 0 and k ⬎ 0 (Fig. 284).
The result will be used to evaluate the so-called Laplace integrals.
Solution.
(a) From (10) we have A (w) ⫽
0
冮e
Fig. 284. f(x)
in Example 3
ⴚkv
2
p
cos wv dv ⫽ ⫺
冮
⬁
eⴚkv cos wv dv. Now, by integration by parts,
0
eⴚkv a⫺
k
k2 ⫹ w2
w
k
sin wv ⫹ cos wvb .
If v ⫽ 0, the expression on the right equals ⫺k>(k 2 ⫹ w 2). If v approaches infinity, that expression approaches
zero because of the exponential factor. Thus 2> p times the integral from 0 to ⬁ gives
A (w) ⫽
(12)
2k> p
k ⫹ w2
2
.
By substituting this into the first integral in (10) we thus obtain the Fourier cosine integral representation
2k
f (x) ⫽ eⴚkx ⫽
p
冮
⬁
0
cos wx
dw
k2 ⫹ w2
(x ⬎ 0, k ⬎ 0).
From this representation we see that
冮
(13)
⬁
cos wx
k ⫹w
2
0
(b) Similarly, from (11) we have B (w) ⫽
冮e
ⴚkv
2
p
sin wv dv ⫽ ⫺
冮
⬁
dw ⫽
2
p
eⴚkx
2k
(x ⬎ 0, k ⬎ 0).
eⴚkv sin wv dv. By integration by parts,
0
eⴚkv a
w
k2 ⫹ w2
k
w
sin wv ⫹ cos wvb .
This equals ⫺w>(k 2 ⫹ w 2) if v ⫽ 0, and approaches 0 as v : ⬁ . Thus
B (w) ⫽
(14)
2w> p
k ⫹ w2
2
.
From (14) we thus obtain the Fourier sine integral representation
f (x) ⫽ eⴚkx ⫽
2
p
冮
⬁
0
w sin wx
k2 ⫹ w2
dw.
From this we see that
(15)
冮
⬁
0
w sin wx
k2 ⫹ w2
dw ⫽
The integrals (13) and (15) are called the Laplace integrals.
p
2
eⴚkx
(x ⬎ 0, k ⬎ 0).
䊏
c11-a.qxd
10/30/10
1:25 PM
Page 517
SEC. 11.7 Fourier Integral
517
PROBLEM SET 11.7
EVALUATION OF INTEGRALS
1–6
Show that the integral represents the indicated function.
Hint. Use (5), (10), or (11); the integral tells you which one,
and its value tells you what function to consider. Show your
work in detail.
0
if x ⬍ 0
⬁
cos xw ⫹ w sin xw
1.
dx ⫽ d p/2
if x ⫽ 0
1 ⫹ w2
0
ⴚx
pe
if x ⬎ 0
冮
2.
冮
⬁
sin pw sin xw
1 ⫺ w2
0
3.
冮
⬁
0
p
dw ⫽ b
2
sin x
if
0⬉x⬉p
0
if
x⬎ p
1
1 ⫺ cos pw
2p
sin
xw
dw
⫽
b
w
0
functions of x. Graph approximations obtained by
replacing ⬁ with finite upper limits of your choice.
Compare the quality of the approximations. Write a
short report on your empirical results and observations.
14. PROJECT. Properties of Fourier Integrals
(a) Fourier cosine integral. Show that (10) implies
4.
冮
5.
冮
cos pw
1⫺w
0
⬁
1
2
冮
cos xw dw ⫽ b
sin w ⫺ w cos w
w2
0
6.
2
ⴥ
0
7–12
w 3 sin xw
w4 ⫹ 4
1
2
0
ƒxƒ ⭌ p
if
1
2
px if 0 ⬍ x ⬍ 1
1
4
if
x⫽1
0
if
x⬎1
sin xw dw ⫽ d p
dw ⫽ 12 peⴚx cos x
if x ⬎ 0
Represent f (x) as an integral (10).
8. f ( x) ⫽ b
1
0
x
if
0⬍x⬍1
if
x⬎1
2
0⬍x⬍1
if
B* ⫽ ⫺
x 2 f ( x) ⫽
11. f ( x) ⫽ b
12. f ( x) ⫽ b
0
eⴚx
0
if
d A
.
dw 2
(b) Solve Prob. 8 by applying (a3) to the result of Prob. 7.
(c) Verify (a2) for f (x) ⫽ 1 if 0 ⬍ x ⬍ a and
f (x) ⫽ 0 if x ⬎ a.
(d) Fourier sine integral. Find formulas for the Fourier
sine integral similar to those in (a).
15. CAS EXPERIMENT. Sine Integral. Plot Si(u) for
positive u. Does the sequence of the maximum and
minimum values give the impression that it converges
and has the limit p>2? Investigate the Gibbs phenomenon
graphically.
A* ⫽ ⫺
16–20
FOURIER SINE INTEGRAL
REPRESENTATIONS
Represent f(x) as an integral (11).
x⬎a
0⬍x⬍p
if
x⬎p
if 0 ⬍ x ⬍ a
if
18. f ( x) ⫽ b
19. f ( x) ⫽ b
x⬎a
13. CAS EXPERIMENT. Approximate Fourier Cosine
Integrals. Graph the integrals in Prob. 7, 9, and 11 as
冮 A*(w) cos xw dw,
0
17. f ( x) ⫽ b
sin x
A as in (10)
2
a 2 ⫺ x 2 if 0 ⬍ x ⬍ a
if
dA
,
dw
⬁
16. f ( x) ⫽ b
0
*
0
0
if
x⬎1
9. f (x) ⫽ 1>(1 ⫹ x 2) [x ⬎ 0 . Hint. See (13).]
10. f ( x) ⫽ b
冮 B (w) sin xw dw,
p cos x if 0 ⬍ ƒ x ƒ ⬍ 12p
FOURIER COSINE INTEGRAL
REPRESENTATIONS
7. f ( x) ⫽ b
(Scale change)
xf (x) ⫽
(a2)
x⬎p
1
2
w
A a a b cos xw dw
⬁
(a3)
⬁
ⴥ
0
(a ⬎ 0)
if 0 ⬍ x ⬍ p
if
冮
1
f (ax) ⫽ a
(a1)
20. f ( x) ⫽ b
if 0 ⬍ x ⬍ a
x
x⬎a
0 if
1
if
0⬍x⬍1
0
if
x⬎1
cos x
if
0⬍x⬍p
0
if
x⬎p
e
x
0
e
ⴚx
0
if 0 ⬍ x ⬍ 1
x⬎1
if
if 0 ⬍ x ⬍ 1
if
x⬎1
c11-b.qxd
10/30/10
518
11.8
1:31 PM
Page 518
CHAP. 11 Fourier Analysis
Fourier Cosine and Sine Transforms
An integral transform is a transformation in the form of an integral that produces from
given functions new functions depending on a different variable. One is mainly interested
in these transforms because they can be used as tools in solving ODEs, PDEs, and integral
equations and can often be of help in handling and applying special functions. The Laplace
transform of Chap. 6 serves as an example and is by far the most important integral
transform in engineering.
Next in order of importance are Fourier transforms. They can be obtained from the
Fourier integral in Sec. 11.7 in a straightforward way. In this section we derive two such
transforms that are real, and in Sec. 11.9 a complex one.
Fourier Cosine Transform
The Fourier cosine transform concerns even functions f (x). We obtain it from the Fourier
cosine integral [(10) in Sec. 10.7]
f (x) ⫽
冮
ⴥ
A(w) cos wx dw,
where
2
A (w) ⫽ p
0
冮
ⴥ
f (v) cos wv dv.
0
Namely, we set A(w) ⫽ 22> p fˆc (w), where c suggests “cosine.” Then, writing v ⫽ x in
the formula for A(w), we have
(1a)
fˆc(w) ⫽
冮
2
Bp
ⴥ
f (x) cos wx dx
0
and
(1b)
f (x) ⫽
2
Bp
冮
ⴥ
fˆc (w) cos wx dw.
0
Formula (1a) gives from f (x) a new function fˆc(w), called the Fourier cosine transform
of f (x). Formula (1b) gives us back f (x) from fˆc(w), and we therefore call f (x) the inverse
Fourier cosine transform of fˆc(w).
The process of obtaining the transform fˆc from a given f is also called the Fourier
cosine transform or the Fourier cosine transform method.
Fourier Sine Transform
Similarly, in (11), Sec. 11.7, we set B (w) ⫽ 22> p fˆs(w), where s suggests “sine.” Then,
writing v ⫽ x, we have from (11), Sec. 11.7, the Fourier sine transform, of f (x) given by
(2a)
fˆs(w) ⫽
2
Bp
冮
ⴥ
0
f(x) sin wx dx,
c11-b.qxd
10/30/10
1:31 PM
Page 519
SEC. 11.8 Fourier Cosine and Sine Transforms
519
and the inverse Fourier sine transform of fˆs (w), given by
2
f (x) ⫽
(2b)
Bp
冮
ⴥ
fˆs (w) sin wx dw.
0
The process of obtaining fs (w) from f (x) is also called the Fourier sine transform or
the Fourier sine transform method.
Other notations are
fc ( f ) ⫽ fˆc,
fs ( f ) ⫽ fˆs
and fcⴚ1 and fsⴚ1 for the inverses of fc and fs, respectively.
EXAMPLE 1
Fourier Cosine and Fourier Sine Transforms
Find the Fourier cosine and Fourier sine transforms of the function
k
f (x) ⫽ b
a
x
Fig. 285. ƒ(x) in
Example 1
Solution.
k if
0⬍x⬍a
0 if
x⬎a
(Fig. 285).
From the definitions (1a) and (2a) we obtain by integration
fˆc (w) ⫽
fˆs (w) ⫽
2
Bp
2
Bp
a
k
冮 cos wx dx ⫽ B p k a
2
0
a
k
冮 sin wx dx ⫽ B p k a
2
0
sin aw
b
w
1 ⫺ cos aw
b.
w
This agrees with formulas 1 in the first two tables in Sec. 11.10 (where k ⫽ 1).
Note that for f (x) ⫽ k ⫽ const (0 ⬍ x ⬍ ⬁), these transforms do not exist. (Why?)
EXAMPLE 2
䊏
Fourier Cosine Transform of the Exponential Function
Find fc(eⴚx).
Solution.
By integration by parts and recursion,
fc(eⴚx ) ⫽
Bp 冮
2
ⴥ
0
eⴚx cos wx dx ⫽
ⴥ
22> p
eⴚx
⫽
.
2 (⫺cos wx ⫹ w sin wx) `
1 ⫹ w2
Bp 1 ⫹ w
0
2
This agrees with formula 3 in Table I, Sec. 11.10, with a ⫽ 1. See also the next example.
䊏
What did we do to introduce the two integral transforms under consideration? Actually
not much: We changed the notations A and B to get a “symmetric” distribution of the
constant 2> p in the original formulas (1) and (2). This redistribution is a standard convenience, but it is not essential. One could do without it.
What have we gained? We show next that these transforms have operational properties
that permit them to convert differentiations into algebraic operations (just as the Laplace
transform does). This is the key to their application in solving differential equations.
c11-b.qxd
10/30/10
1:31 PM
520
Page 520
CHAP. 11 Fourier Analysis
Linearity, Transforms of Derivatives
If f (x) is absolutely integrable (see Sec. 11.7) on the positive x-axis and piecewise
continuous (see Sec. 6.1) on every finite interval, then the Fourier cosine and sine
transforms of f exist.
Furthermore, if f and g have Fourier cosine and sine transforms, so does af ⫹ bg for
any constants a and b, and by (1a)
fc (af ⫹ bg) ⫽
冮
2
Bp
⫽a
ⴥ
[af (x) ⫹ bg (x)] cos wx dx
0
2
Bp
冮
ⴥ
2
f (x) cos wx dx ⫹ b
Bp
0
冮
ⴥ
g (x) cos wx dx.
0
The right side is afc( f ) ⫹ bfc(g). Similarly for fs, by (2). This shows that the Fourier
cosine and sine transforms are linear operations,
(3)
THEOREM 1
(a)
fc(af ⫹ bg) ⫽ afc( f ) ⫹ bfc(g),
(b)
fs(af ⫹ bg) ⫽ afs( f ) ⫹ bfs(g).
Cosine and Sine Transforms of Derivatives
Let f (x) be continuous and absolutely integrable on the x-axis, let f r (x) be piecewise
continuous on every finite interval, and let f (x) : 0 as x : ⬁. Then
(a)
2
fc{ f r(x)} ⫽ w fs{f (x)} ⫺
f (0),
Bp
(4)
fs{f r (x)} ⫽ ⫺wfc{f (x)}.
(b)
PROOF
This follows from the definitions and by using integration by parts, namely,
fc{f r (x)} ⫽
⫽
2
Bp
2
Bp
⫽⫺
ⴥ
冮
f r (x) cos wx dx
0
c f (x) cos wx `
ⴥ
⫹w
0
冮
ⴥ
f (x) sin wx dx d
0
2
f (0) ⫹ w fs{f (x)};
Bp
˛
and similarly,
fs{f r (x)} ⫽
⫽
2
Bp
冮
ⴥ
f r (x) sin wx dx
0
ⴥ
c f (x) sin wx ` ⫺ w
0
Bp
2
⫽ 0 ⫺ wfc{f(x)}.
冮
ⴥ
0
f (x) cos wx dx d
䊏
c11-b.qxd
10/30/10
1:31 PM
Page 521
SEC. 11.8 Fourier Cosine and Sine Transforms
521
Formula (4a) with f r instead of f gives (when f r , f s satisfy the respective assumptions
for f, f r in Theorem 1)
fc{f s (x)} ⫽ w fs{f r (x)} ⫺
˛
2
f r (0);
Bp
hence by (4b)
(5a)
2
fc{f s (x)} ⫽ ⫺w 2 fc{f (x)} ⫺
f r (0).
Bp
fs{f s (x)} ⫽ ⫺w 2 fs{f (x)} ⫹
wf (0).
Bp
Similarly,
(5b)
2
A basic application of (5) to PDEs will be given in Sec. 12.7. For the time being we
show how (5) can be used for deriving transforms.
EXAMPLE 3
An Application of the Operational Formula (5)
fc(eⴚax) of f (x) ⫽ eⴚax, where a ⬎ 0.
Solution. By differentiation, (eⴚax) s ⫽ a 2eⴚax; thus
Find the Fourier cosine transform
a 2f (x) ⫽ f s (x).
From this, (5a), and the linearity (3a),
a 2 fc( f ) ⫽ fc( f s )
⫽ ⫺w 2 fc( f ) ⫺
2
Bp
⫽ ⫺w 2 fc( f ) ⫹ a
f r (0)
2
Bp
.
Hence
(a 2 ⫹ w 2)fc( f ) ⫽ a22> p.
The answer is (see Table I, Sec. 11.10)
fc(eⴚax) ⫽
2
a
a
b
B p a2 ⫹ w 2
(a ⬎ 0).
Tables of Fourier cosine and sine transforms are included in Sec. 11.10.
䊏
c11-b.qxd
10/30/10
522
1:31 PM
Page 522
CHAP. 11 Fourier Analysis
PROBLEM SET 11.8
1–8
FOURIER COSINE TRANSFORM
9–15
9. Find fs(eⴚax), a ⬎ 0, by integration.
1. Find the cosine transform fˆc(w) of f (x) ⫽ 1 if
0 ⬍ x ⬍ 1, f (x) ⫽ ⫺1 if 1 ⬍ x ⬍ 2, f (x) ⫽ 0 if
x ⬎ 2.
2. Find f in Prob. 1 from the answer fˆc.
3. Find fˆc(w) for f (x) ⫽ x if 0 ⬍ x ⬍ 2, f (x) ⫽ 0 if
x ⬎ 2.
4. Derive formula 3 in Table I of Sec. 11.10 by integration.
5. Find fˆc(w) for f (x) ⫽ x 2 if 0 ⬍ x ⬍ 1, f (x) ⫽ 0 if x ⬎ 1.
6. Continuity assumptions. Find ĝc(w) for g (x) ⫽ 2 if
0 ⬍ x ⬍ 1, g (x) ⫽ 0 if x ⬎ 1. Try to obtain from it
fˆc(w) for f (x) in Prob. 5 by using (5a).
7. Existence? Does the Fourier cosine transform of
x ⴚ1 sin x (0 ⬍ x ⬍ ⬁) exist? Of x ⴚ1 cos x? Give
reasons.
8. Existence? Does the Fourier cosine transform of
f (x) ⫽ k ⫽ const (0 ⬍ x ⬍ ⬁) exist? The Fourier sine
transform?
11.9
FOURIER SINE TRANSFORM
10. Obtain the answer to Prob. 9 from (5b).
11. Find fs (w) for f (x) ⫽ x 2 if 0 ⬍ x ⬍ 1, f (x) ⫽ 0 if
x ⬎ 1.
12. Find fs(xeⴚx >2) from (4b) and a suitable formula in
Table I of Sec. 11.10.
2
13. Find fs(eⴚx) from (4a) and formula 3 of Table I in
Sec. 11.10.
14. Gamma function. Using formulas 2 and 4 in Table II
of Sec. 11.10, prove ⌫(12) ⫽ 1p [(30) in App. A3.1],
a value needed for Bessel functions and other
applications.
15. WRITING PROJECT. Finding Fourier Cosine and
Sine Transforms. Write a short report on ways of
obtaining these transforms, with illustrations by
examples of your own.
Fourier Transform.
Discrete and Fast Fourier Transforms
In Sec. 11.8 we derived two real transforms. Now we want to derive a complex transform
that is called the Fourier transform. It will be obtained from the complex Fourier integral,
which will be discussed next.
Complex Form of the Fourier Integral
The (real) Fourier integral is [see (4), (5), Sec. 11.7]
f (x) ⫽
冮
ⴥ
[A(w) cos wx ⫹ B(w) sin wx] dw
0
where
1
A(w) ⫽ p
冮
ⴥ
f (v) cos wv dv,
ⴚⴥ
1
B(w) ⫽ p
冮
ⴥ
f (v) sin wv dv.
ⴚⴥ
Substituting A and B into the integral for f, we have
1
f (x) ⫽ p
ⴥ
冮 冮
0
ⴥ
ⴚⴥ
f (v)[cos wv cos wx ⫹ sin wv sin wx] dv dw.
c11-b.qxd
10/30/10
1:31 PM
Page 523
SEC. 11.9 Fourier Transform. Discrete and Fast Fourier Transforms
523
By the addition formula for the cosine [(6) in App. A3.1] the expression in the brackets
[ Á ] equals cos (wv ⫺ wx) or, since the cosine is even, cos (wx ⫺ wv). We thus obtain
(1*)
1
f (x) ⫽ p
ⴥ
冮 冮
ⴥ
B
f (v) cos (wx ⫺ wv)dvR dw.
ⴚⴥ
0
The integral in brackets is an even function of w, call it F (w), because cos (wx ⫺ wv) is
an even function of w, the function f does not depend on w, and we integrate with respect
to v (not w). Hence the integral of F (w) from w ⫽ 0 to ⬁ is 12 times the integral of F (w)
from ⫺⬁ to ⬁ . Thus (note the change of the integration limit!)
(1)
f (x) ⫽
1
2p
ⴥ
冮 冮
ⴥ
B
ⴚⴥ
f (v) cos (wx ⫺ wv) dvR dw.
ⴚⴥ
We claim that the integral of the form (1) with sin instead of cos is zero:
(2)
1
2p
ⴥ
冮 冮
ⴥ
B
ⴚⴥ
f (v) sin (wx ⫺ wv) dvR dw ⫽ 0.
ⴚⴥ
This is true since sin (wx ⫺ wv) is an odd function of w, which makes the integral in
brackets an odd function of w, call it G (w). Hence the integral of G (w) from ⫺⬁ to ⬁
is zero, as claimed.
We now take the integrand of (1) plus i (⫽ 1⫺1) times the integrand of (2) and use
the Euler formula [(11) in Sec. 2.2]
eix ⫽ cos x ⫹ i sin x.
(3)
Taking wx ⫺ wv instead of x in (3) and multiplying by f (v) gives
f (v) cos (wx ⫺ wv) ⫹ if (v) sin (wx ⫺ wv) ⫽ f (v)ei(wxⴚwv).
Hence the result of adding (1) plus i times (2), called the complex Fourier integral, is
f (x) ⫽
(4)
1
2p
ⴥ
冮 冮
ⴚⴥ
ⴥ
f (v)eiw(xⴚv) dv dw
(i ⫽ 1⫺1).
ⴚⴥ
To obtain the desired Fourier transform will take only a very short step from here.
Fourier Transform and Its Inverse
Writing the exponential function in (4) as a product of exponential functions, we have
(5)
f (x) ⫽
冮
22p
1
ⴥ
ⴚⴥ
B
冮
22p
1
ⴥ
f (v)eⴚiwv dvR eiwx dw.
ⴚⴥ
The expression in brackets is a function of w, is denoted by fˆ(w), and is called the Fourier
transform of f ; writing v ⫽ x, we have
(6)
fˆ(w) ⫽
冮
22p
1
ⴥ
ⴚⴥ
f (x)eⴚiwx dx.
c11-b.qxd
10/30/10
1:31 PM
524
Page 524
CHAP. 11 Fourier Analysis
With this, (5) becomes
f (x) ⫽
(7)
1
22p
冮
ⴥ
fˆ(w)eiwx dw
ⴚⴥ
and is called the inverse Fourier transform of fˆ(w).
Another notation for the Fourier transform is
fˆ ⫽ f( f ),
so that
f ⫽ fⴚ1( fˆ).
The process of obtaining the Fourier transform f( f ) ⫽ fˆ from a given f is also called
the Fourier transform or the Fourier transform method.
Using concepts defined in Secs. 6.1 and 11.7 we now state (without proof) conditions
that are sufficient for the existence of the Fourier transform.
THEOREM 1
Existence of the Fourier Transform
If f (x) is absolutely integrable on the x-axis and piecewise continuous on every finite
interval, then the Fourier transform fˆ(w) of f (x) given by (6) exists.
EXAMPLE 1
Fourier Transform
Find the Fourier transform of f (x) ⫽ 1 if ƒ x ƒ ⬍ 1 and f (x) ⫽ 0 otherwise.
Solution.
Using (6) and integrating, we obtain
fˆ(w) ⫽
冮
12p
1
1
eⴚiwx dx ⫽
ⴚ1
1
12p
ⴚiwx 1
#e
` ⫽
⫺iw
ⴚ1
1
⫺iw 12p
(e
ⴚiw
⫺ eiw).
As in (3) we have eiw ⫽ cos w ⫹ i sin w, eⴚiw ⫽ cos w ⫺ i sin w, and by subtraction
eiw ⫺ eⴚiw ⫽ 2i sin w.
Substituting this in the previous formula on the right, we see that i drops out and we obtain the answer
fˆ(w) ⫽
EXAMPLE 2
p sin w
B2
w
䊏
.
Fourier Transform
Find the Fourier transform f (eⴚax) of f (x) ⫽ eⴚax if x ⬎ 0 and f (x) ⫽ 0 if x ⬍ 0; here a ⬎ 0.
Solution.
From the definition (6) we obtain by integration
f (eⴚax) ⫽
冮
12p
1
ⴥ
eⴚaxeⴚiwx dx
0
⫽
1
eⴚ(a⫹iw)x
22p ⫺(a ⫹ iw)
This proves formula 5 of Table III in Sec. 11.10.
`
ⴥ
x⫽0
⫽
1
12p(a ⫹ iw)
.
䊏
c11-b.qxd
10/30/10
1:31 PM
Page 525
SEC. 11.9 Fourier Transform. Discrete and Fast Fourier Transforms
525
Physical Interpretation: Spectrum
The nature of the representation (7) of f (x) becomes clear if we think of it as a superposition
of sinusoidal oscillations of all possible frequencies, called a spectral representation.
This name is suggested by optics, where light is such a superposition of colors
(frequencies). In (7), the “spectral density” fˆ(w) measures the intensity of f (x) in the
frequency interval between w and w ⫹ ¢w ( ¢w small, fixed). We claim that, in connection
with vibrations, the integral
冮
ⴥ
ƒ fˆ(w) ƒ 2 dw
ⴚⴥ
can be interpreted as the total energy of the physical system. Hence an integral of ƒ fˆ(w) ƒ 2
from a to b gives the contribution of the frequencies w between a and b to the total energy.
To make this plausible, we begin with a mechanical system giving a single frequency,
namely, the harmonic oscillator (mass on a spring, Sec. 2.4)
my s ⫹ ky ⫽ 0.
Here we denote time t by x. Multiplication by y r gives my r y s ⫹ ky r y ⫽ 0. By integration,
1
2
2 mv
⫹ 12 ky 2 ⫽ E 0 ⫽ const
where v ⫽ y r is the velocity. The first term is the kinetic energy, the second the potential
energy, and E 0 the total energy of the system. Now a general solution is (use (3) in
Sec. 11.4 with t ⫽ x)
y ⫽ a1 cos w0 x ⫹ b1 sin w0 x ⫽ c1eiw0x ⫹ cⴚ1eⴚiw0x,
w 20 ⫽ k>m
where c1 ⫽ (a1 ⫺ ib1)>2, cⴚ1 ⫽ c1 ⫽ (a1 ⫹ ib1)>2. We write simply A ⫽ c1eiw0x,
B ⫽ cⴚ1eⴚiw0x. Then y ⫽ A ⫹ B. By differentiation, v ⫽ y r ⫽ A r ⫹ B r ⫽ iw0 (A ⫺ B).
Substitution of v and y on the left side of the equation for E 0 gives
E 0 ⫽ 12 mv2 ⫹ 12 ky 2 ⫽ 12 m(iw0)2(A ⫺ B)2 ⫹ 12 k(A ⫹ B)2.
Here w 20 ⫽ k>m, as just stated; hence mw 20 ⫽ k. Also i 2 ⫽ ⫺1, so that
E 0 ⫽ 12 k[⫺(A ⫺ B)2 ⫹ (A ⫹ B)2] ⫽ 2kAB ⫽ 2kc1eiw0xcⴚ1eⴚiw0x ⫽ 2kc1cⴚ1 ⫽ 2k ƒ c1 ƒ 2.
Hence the energy is proportional to the square of the amplitude ƒ c1 ƒ .
As the next step, if a more complicated system leads to a periodic solution y ⫽ f (x)
that can be represented by a Fourier series, then instead of the single energy term ƒ c1 ƒ 2
we get a series of squares ƒ cn ƒ 2 of Fourier coefficients cn given by (6), Sec. 11.4. In this
case we have a “discrete spectrum” (or “point spectrum”) consisting of countably many
isolated frequencies (infinitely many, in general), the corresponding ƒ cn ƒ 2 being the
contributions to the total energy.
Finally, a system whose solution can be represented by an integral (7) leads to the above
integral for the energy, as is plausible from the cases just discussed.
c11-b.qxd
10/30/10
1:31 PM
526
Page 526
CHAP. 11 Fourier Analysis
Linearity. Fourier Transform of Derivatives
New transforms can be obtained from given ones by using
THEOREM 2
Linearity of the Fourier Transform
The Fourier transform is a linear operation; that is, for any functions f (x) and g(x)
whose Fourier transforms exist and any constants a and b, the Fourier transform
of af ⫹ bg exists, and
f(af ⫹ bg) ⫽ af ( f ) ⫹ bf (g).
(8)
PROOF
This is true because integration is a linear operation, so that (6) gives
f{af (x) ⫹ bg (x)} ⫽
1
12p
⫽a
冮
ⴥ
[af (x) ⫹ bg (x)] eⴚiwx dx
ⴚⴥ
1
12p
冮
ⴥ
1
12p
f (x)eⴚiwx dx ⫹ b
ⴚⴥ
冮
ⴥ
g (x)eⴚiwx dx
ⴚⴥ
⫽ af{f (x)} ⫹ bf{g (x)}.
䊏
In applying the Fourier transform to differential equations, the key property is that
differentiation of functions corresponds to multiplication of transforms by iw:
THEOREM 3
Fourier Transform of the Derivative of f (x)
Let f (x) be continuous on the x-axis and f (x) : 0 as ƒ x ƒ : ⬁ . Furthermore, let f r (x)
be absolutely integrable on the x-axis. Then
f {f r (x)} ⫽ iwf {f (x)}.
(9)
PROOF
From the definition of the Fourier transform we have
f{f r (x)} ⫽
1
12p
冮
ⴥ
f r (x)eⴚiwx dx.
ⴚⴥ
Integrating by parts, we obtain
ⴥ
f{f r (x)} ⫽
1
Bf (x)eⴚiwx `
⫺ (⫺iw)
12p
ⴚⴥ
冮
ⴥ
f (x)eⴚiwx dxR .
ⴚⴥ
Since f (x) : 0 as ƒ x ƒ : ⬁, the desired result follows, namely,
f{f r (x)} ⫽ 0 ⫹ iw f{f (x)}.
䊏
c11-b.qxd
10/30/10
3:46 PM
Page 527
SEC. 11.9 Fourier Transform. Discrete and Fast Fourier Transforms
527
Two successive applications of (9) give
f ( f s ) ⫽ iwf ( f r ) ⫽ (iw)2f ( f ).
Since (iw)2 ⫽ ⫺w 2, we have for the transform of the second derivative of f
f{f s (x)} ⫽ ⫺w 2f{f (x)}.
(10)
Similarly for higher derivatives.
An application of (10) to differential equations will be given in Sec. 12.6. For the time
being we show how (9) can be used to derive transforms.
EXAMPLE 3
Application of the Operational Formula (9)
Find the Fourier transform of xeⴚx from Table III, Sec 11.10.
2
Solution.
We use (9). By formula 9 in Table III
f (xeⴚx ) ⫽ f{⫺ 12 (eⴚx ) r }
2
2
⫽ ⫺ 12 f{(eⴚx ) r }
2
⫽ ⫺ 12 iwf(eⴚx )
2
1 ⴚw2>4
1
⫽ ⫺ iw
e
2
12
⫽⫺
iw
2 12
eⴚw
>4
2
䊏
.
Convolution
The convolution f * g of functions f and g is defined by
(11)
h (x) ⫽ ( f * g) (x) ⫽
冮
ⴥ
f (p) g (x ⫺ p) dp ⫽
ⴚⴥ
冮
ⴥ
f (x ⫺ p)g (p) dp.
ⴚⴥ
The purpose is the same as in the case of Laplace transforms (Sec. 6.5): taking the
convolution of two functions and then taking the transform of the convolution is the same
as multiplying the transforms of these functions (and multiplying them by 12p):
THEOREM 4
Convolution Theorem
Suppose that f (x) and g(x) are piecewise continuous, bounded, and absolutely
integrable on the x-axis. Then
(12)
f ( f * g) ⫽ 12p f ( f ) f (g).
c11-b.qxd
10/30/10
1:31 PM
528
Page 528
CHAP. 11 Fourier Analysis
PROOF
By the definition,
1
f ( f * g) ⫽
12p
ⴥ
冮 冮
ⴥ
f (p) g (x ⫺ p) dp eⴚiwx dx.
ⴚⴥ ⴚⴥ
An interchange of the order of integration gives
f ( f * g) ⫽
1
12p
ⴥ
ⴥ
冮 冮
f (p) g (x ⫺ p) eⴚiwx dx dp.
ⴚⴥ ⴚⴥ
Instead of x we now take x ⫺ p ⫽ q as a new variable of integration. Then x ⫽ p ⫹ q
and
f ( f * g) ⫽
1
12p
ⴥ
冮 冮
ⴥ
f (p) g (q) eⴚiw (p⫹q) dq dp.
ⴚⴥ ⴚⴥ
This double integral can be written as a product of two integrals and gives the desired
result
1
f ( f * g) ⫽
12p
⫽
冮
ⴥ
f (p)e
ⴚiwp
ⴚⴥ
dp
冮
ⴥ
g (q) eⴚiwq dq
ⴚⴥ
1
[12p f ( f )][12p f (g)] ⫽ 12p f ( f ) f (g).
12p
䊏
By taking the inverse Fourier transform on both sides of (12), writing fˆ ⫽ f ( f ) and
ĝ ⫽ f (g) as before, and noting that 12p and 1> 12p in (12) and (7) cancel each other,
we obtain
(13)
( f * g) (x) ⫽
冮
ⴥ
fˆ(w)ĝ (w)eiwx dw,
ⴚⴥ
a formula that will help us in solving partial differential equations (Sec. 12.6).
Discrete Fourier Transform (DFT),
Fast Fourier Transform (FFT)
In using Fourier series, Fourier transforms, and trigonometric approximations (Sec. 11.6)
we have to assume that a function f (x), to be developed or transformed, is given on some
interval, over which we integrate in the Euler formulas, etc. Now very often a function f (x)
is given only in terms of values at finitely many points, and one is interested in extending
Fourier analysis to this case. The main application of such a “discrete Fourier analysis”
concerns large amounts of equally spaced data, as they occur in telecommunication, time
series analysis, and various simulation problems. In these situations, dealing with sampled
values rather than with functions, we can replace the Fourier transform by the so-called
discrete Fourier transform (DFT) as follows.
c11-b.qxd
10/30/10
1:31 PM
Page 529
SEC. 11.9 Fourier Transform. Discrete and Fast Fourier Transforms
529
Let f (x) be periodic, for simplicity of period 2p. We assume that N measurements of
f (x) are taken over the interval 0 ⬉ x ⬉ 2p at regularly spaced points
xk ⫽
(14)
2pk
,
N
k ⫽ 0, 1, Á , N ⫺ 1.
We also say that f (x) is being sampled at these points. We now want to determine a
complex trigonometric polynomial
N⫺1
q (x) ⫽ a cneinxk
(15)
n⫽0
that interpolates f (x) at the nodes (14), that is, q (x k) ⫽ f (x k), written out, with fk denoting
f (x k),
N⫺1
fk ⫽ f (x k) ⫽ q (x k) ⫽ a cneinxk,
(16)
k ⫽ 0, 1, Á , N ⫺ 1.
n⫽0
Hence we must determine the coefficients c0, Á , cNⴚ1 such that (16) holds. We do this
by an idea similar to that in Sec. 11.1 for deriving the Fourier coefficients by using the
orthogonality of the trigonometric system. Instead of integrals we now take sums. Namely,
we multiply (16) by eⴚimxk (note the minus!) and sum over k from 0 to N ⫺ 1. Then we
interchange the order of the two summations and insert x k from (14). This gives
(17)
N⫺1
N⫺1 N⫺1
N⫺1
N⫺1
k⫽0
k⫽0 n⫽0
n⫽0
k⫽0
ⴚimxk
⫽ a a cnei(nⴚm)xk ⫽ a cn a ei (nⴚm) 2pk>N.
a fke
Now
ei (nⴚm)2pk>N ⫽ [ei (nⴚm)2p>N]k.
We donote [ Á ] by r. For n ⫽ m we have r ⫽ e0 ⫽ 1. The sum of these terms over k
equals N, the number of these terms. For n ⫽ m we have r ⫽ 1 and by the formula for a
geometric sum [(6) in Sec. 15.1 with q ⫽ r and n ⫽ N ⫺ 1]
1 ⫺ rN
k
ar ⫽ 1⫺r ⫽0
k⫽0
N⫺1
because r N ⫽ 1; indeed, since k, m, and n are integers,
r N ⫽ ei(nⴚm)2pk ⫽ cos 2pk(n ⫺ m) ⫹ i sin 2pk(n ⫺ m) ⫽ 1 ⫹ 0 ⫽ 1.
This shows that the right side of (17) equals cmN. Writing n for m and dividing by N, we
thus obtain the desired coefficient formula
(18*)
cn ⫽
1 N⫺1
ⴚinxk
a fke
N k⫽0
fk ⫽ f (x k),
n ⫽ 0, 1, Á , N ⫺ 1.
Since computation of the cn (by the fast Fourier transform, below) involves successive
halfing of the problem size N, it is practical to drop the factor 1>N from cn and define the
c11-b.qxd
10/30/10
1:31 PM
530
Page 530
CHAP. 11 Fourier Analysis
discrete Fourier transform of the given signal f ⫽ [ f0
fˆ ⫽ [ fˆ0 Á fˆNⴚ1] with components
fNⴚ1]T to be the vector
Á
N⫺1
fˆn ⫽ Ncn ⫽ a fkeⴚinxk,
(18)
fk ⫽ f (x k),
n ⫽ 0, Á , N ⫺ 1.
k⫽0
This is the frequency spectrum of the signal.
In vector notation, f̂ ⫽ FNf, where the N ⫻ N Fourier matrix FN ⫽ [enk] has the
entries [given in (18)]
(19)
enk ⫽ eⴚinxk ⫽ eⴚ2pink>N ⫽ w nk,
w ⫽ wN ⫽ eⴚ2pi>N,
where n, k ⫽ 0, Á , N ⫺ 1.
EXAMPLE 4
Discrete Fourier Transform (DFT). Sample of N ⴝ 4 Values
Let N ⫽ 4 measurements (sample values) be given. Then w ⫽ eⴚ2pi>N ⫽ eⴚpi>2 ⫽ ⫺i and thus w nk ⫽ (⫺i)nk.
Let the sample values be, say f ⫽ [0 1 4 9]T. Then by (18) and (19),
(20)
w0
w0
w0
w
0
w
1
w
2
w
3
w
0
w
2
w
4
w
6
f̂ ⫽ F4 f ⫽ E
w0
w0
w3
w6
w9
1
1
1
1
⫺i
⫺1
1
⫺1
1
i
U f⫽E
1
0
14
i
1
⫺4 ⫹ 8i
1
⫺1
4
⫺6
⫺1
⫺i
9
⫺4 ⫺ 8i
U E U⫽E
U.
From the first matrix in (20) it is easy to infer what FN looks like for arbitrary N, which in practice may be
1000 or more, for reasons given below.
䊏
From the DFT (the frequency spectrum) f̂ ⫽ FNf we can recreate the given signal
1 nk
f̂ ⫽ F ⴚ1
[w ]
N f, as we shall now prove. Here FN and its complex conjugate FN ⫽
N
satisfy
(21a)
FNFN ⫽ FNFN ⫽ NI
where I is the N ⫻ N unit matrix; hence FN has the inverse
(21b)
PROOF
F ⴚ1
N ⫽
1
FN.
N
We prove (21). By the multiplication rule (row times column) the product matrix
GN ⫽ FNFN ⫽ [gjk] in (21a) has the entries gjk ⫽ Row j of FN times Column k of FN.
That is, writing W ⫽ w jw k, we prove that
gjk ⫽ (w jw k)0 ⫹ (w j wk )1 ⫹ Á ⫹ (w j w k )Nⴚ1
⫽ W 0 ⫹ W 1 ⫹ Á ⫹W Nⴚ1 ⫽ b
0 if j ⫽ k
N if j ⫽ k.
c11-b.qxd
10/30/10
1:31 PM
Page 531
SEC. 11.9 Fourier Transform. Discrete and Fast Fourier Transforms
531
Indeed, when j ⫽ k, then w kw k ⫽ (ww)k ⫽ (e2pi>Neⴚ2pi>N)k ⫽ 1k ⫽ 1, so that the sum
of these N terms equals N; these are the diagonal entries of GN. Also, when j ⫽ k, then
W ⫽ 1 and we have a geometric sum (whose value is given by (6) in Sec. 15.1 with q ⫽ W
and n ⫽ N ⫺1)
1 ⫺ WN
W 0 ⫹ W 1 ⫹ Á ⫹W Nⴚ1 ⫽
⫽0
1⫺W
because W N ⫽ (w jw k)N ⫽ (e2pi)j(eⴚ2pi)k ⫽ 1j # 1k ⫽ 1.
䊏
We have seen that fˆ is the frequency spectrum of the signal f (x). Thus the components
fˆn of fˆ give a resolution of the 2p-periodic function f (x) into simple (complex) harmonics.
Here one should use only n’s that are much smaller than N>2, to avoid aliasing. By this
we mean the effect caused by sampling at too few (equally spaced) points, so that, for
instance, in a motion picture, rotating wheels appear as rotating too slowly or even in the
wrong sense. Hence in applications, N is usually large. But this poses a problem. Eq. (18)
requires O (N) operations for any particular n, hence O (N 2) operations for, say, all
n ⬍ N>2. Thus, already for 1000 sample points the straightforward calculation would
involve millions of operations. However, this difficulty can be overcome by the so-called
fast Fourier transform (FFT), for which codes are readily available (e.g., in Maple). The
FFT is a computational method for the DFT that needs only O (N) log 2 N operations
instead of O (N 2). It makes the DFT a practical tool for large N. Here one chooses N ⫽ 2p
( p integer) and uses the special form of the Fourier matrix to break down the given problem
into smaller problems. For instance, when N ⫽ 1000, those operations are reduced by a
factor 1000>log 2 1000 ⬇ 100.
The breakdown produces two problems of size M ⫽ N>2. This breakdown is possible
because for N ⫽ 2M we have in (19)
2
wN2 ⫽ w 2M
⫽ (eⴚ2pi>N)2 ⫽ eⴚ4pi>(2M) ⫽ eⴚ2pi>(M) ⫽ wM.
The given vector f ⫽ [ f0 Á fNⴚ1]T is split into two vectors with M components each,
namely, f ev ⫽ [ f0 f2 Á fNⴚ2]T containing the even components of f, and f od ⫽
[ f1 f3 Á fNⴚ1]T containing the odd components of f. For f ev and f od we determine
the DFTs
fˆev ⫽ [ fˆev,0 fˆev,2
Á
fˆev,Nⴚ2]T ⫽ FM f ev
fˆod ⫽ [ fˆod,1 fˆod,3
Á
fˆod,Nⴚ1]T ⫽ FM f od
and
involving the same M ⫻ M matrix FM. From these vectors we obtain the components of
the DFT of the given vector f by the formulas
(22)
(a)
ˆ
fˆn ⫽ fˆev,n ⫹ w n
N fod,n
n ⫽ 0, Á , M ⫺ 1
(b)
ˆ
fˆn⫹M ⫽ fˆev,n ⫺ w n
N fod,n
n ⫽ 0, Á , M ⫺ 1.
c11-b.qxd
10/30/10
1:31 PM
532
Page 532
CHAP. 11 Fourier Analysis
For N ⫽ 2p this breakdown can be repeated p ⫺ 1 times in order to finally arrive at N>2
problems of size 2 each, so that the number of multiplications is reduced as indicated
above.
We show the reduction from N ⫽ 4 to M ⫽ N>2 ⫽ 2 and then prove (22).
EXAMPLE 5
Fast Fourier Transform (FFT). Sample of N ⴝ 4 Values
When N ⫽ 4, then w ⫽ wN ⫽ ⫺i as in Example 4 and M ⫽ N>2 ⫽ 2, hence w ⫽ wM ⫽ eⴚ2pi>2 ⫽ eⴚpi ⫽ ⫺1.
Consequently,
fˆ0
fˆev ⫽
cˆ d
f̂ od ⫽
cˆ d
⫽ F2f ev ⫽
c
1
1
1
⫺1
⫽ F2 f od ⫽
c
1
1
1
⫺1
f2
fˆ1
f3
dc d
⫽
c
f0 ⫹ f2
dc d
⫽
c
f1 ⫹ f3
f0
f2
f1
f3
f0 ⫺ f2
f1 ⫺ f3
d
d.
From this and (22a) we obtain
fˆ0 ⫽ fˆev,0 ⫹ w 0N fˆod,0 ⫽ ( f0 ⫹ f2) ⫹ ( f1 ⫹ f3) ⫽ f0 ⫹ f1 ⫹ f2 ⫹ f3
fˆ1 ⫽ fˆev,1 ⫹ w 1N fˆod,1 ⫽ ( f0 ⫺ f2) ⫺ i( f1 ⫹ f3) ⫽ f0 ⫺ if1 ⫺ f2 ⫹ if3.
Similarly, by (22b),
fˆ2 ⫽ fˆev,0 ⫺ w 0N fˆod,0 ⫽ ( f0 ⫹ f2) ⫺ ( f1 ⫹ f3) ⫽ f0 ⫺ f1 ⫹ f2 ⫺ f3
fˆ3 ⫽ fˆev,1 ⫺ w 1N fˆod,1 ⫽ ( f0 ⫺ f2) ⫺ (⫺i)( f1 ⫺ f3) ⫽ f0 ⫹ if1 ⫺ f2 ⫺ if3.
This agrees with Example 4, as can be seen by replacing 0, 1, 4, 9 with f0, f1, f2, f3.
䊏
We prove (22). From (18) and (19) we have for the components of the DFT
Nⴚ1
kn
fk.
fˆn ⫽ a w N
k⫽0
Splitting into two sums of M ⫽ N>2 terms each gives
Mⴚ1
Mⴚ1
2kn
(2k⫹1)n
fˆn ⫽ a w N
f2k ⫹ a w N
f2k⫹1.
k⫽0
k⫽0
We now use wN2 ⫽ wM and pull out w n
N from under the second sum, obtaining
Mⴚ1
(23)
Mⴚ1
kn
n
kn
fˆn ⫽ a w M
fev,k ⫹ w N
a w M fod,k.
k⫽0
k⫽0
The two sums are fev,n and fod,n, the components of the “half-size” transforms Ff ev and
Ff od.
Formula (22a) is the same as (23). In (22b) we have n ⫹ M instead of n. This causes
a sign changes in (23), namely ⫺w n
N before the second sum because
ⴚ2piM>N
wM
⫽ eⴚ2pi>2 ⫽ eⴚpi ⫽ ⫺1.
N ⫽ e
This gives the minus in (22b) and completes the proof.
䊏
c11-b.qxd
10/30/10
1:31 PM
Page 533
SEC. 11.9 Fourier Transform. Discrete and Fast Fourier Transforms
533
PROBLEM SET 11.9
1. Review in complex. Show that 1>i ⫽ ⫺i, eⴚix ⫽
cos x ⫺ i sin x, eix ⫹ eⴚix ⫽ 2 cos x, eix ⫺ eⴚix ⫽
2i sin x, eikx ⫽ cos kx ⫹ i sin kx.
2–11
FOURIER TRANSFORMS BY
INTEGRATION
Find the Fourier transform of f (x) (without using Table
III in Sec. 11.10). Show details.
2. f (x) ⫽ e
3. f (x) ⫽ e
4. f (x) ⫽ e
5. f (x) ⫽ e
e2ix if ⫺1 ⬍ x ⬍ 1
otherwise
0
1 if a ⬍ x ⬍ b
0 otherwise
ekx if x ⬍ 0 (k ⬎ 0)
if x ⬎ 0
0
ex if ⫺a ⬍ x ⬍ a
otherwise
0
6. f (x) ⫽ eⴚƒ x ƒ
7. f (x) ⫽ e
(⫺⬁ ⬍ x ⬍ ⬁)
x if 0 ⬍ x ⬍ a
0 otherwise
8. f (x) ⫽ e
xeⴚx if ⫺1 ⬍ x ⬍ 0
9. f (x) ⫽ e
ƒxƒ
if ⫺1 ⬍ x ⬍ 1
0
otherwise
10. f (x) ⫽ e
otherwise
0
x if ⫺1 ⬍ x ⬍ 1
0 otherwise
⫺1 if ⫺1 ⬍ x ⬍ 0
11. f (x) ⫽ μ
1 if
0⬍x⬍1
0 otherwise
USE OF TABLE III IN SEC. 11.10.
12–17
OTHER METHODS
12. Find f ( f (x)) for f (x) ⫽ xeⴚx if x ⬎ 0, f (x) ⫽ 0 if
x ⬍ 0, by (9) in the text and formula 5 in Table III
(with a ⫽ 1). Hint. Consider xeⴚx and eⴚx.
2
13. Obtain f(eⴚx >2) from Table III.
14. In Table III obtain formula 7 from formula 8.
15. In Table III obtain formula 1 from formula 2.
16. TEAM PROJECT. Shifting (a) Show that if f (x)
has a Fourier transform, so does f (x ⫺ a), and
f{ f (x ⫺ a)} ⫽ eⴚiwaf{ f (x)}.
(b) Using (a), obtain formula 1 in Table III, Sec. 11.10,
from formula 2.
(c) Shifting on the w-Axis. Show that if fˆ (w) is the
Fourier transform of f (x), then fˆ (w ⫺ a) is the Fourier
transform of eiaxf (x).
(d) Using (c), obtain formula 7 in Table III from 1 and
formula 8 from 2.
17. What could give you the idea to solve Prob. 11 by using
the solution of Prob. 9 and formula (9) in the text?
Would this work?
18–25
DISCRETE FOURIER TRANSFORM
18. Verify the calculations in Example 4 of the text.
19. Find the transform of a general signal
f ⫽ [ f1 f2 f3 f4]T of four values.
20. Find the inverse matrix in Example 4 of the text and
use it to recover the given signal.
21. Find the transform (the frequency spectrum) of a
general signal of two values [ f1 f2]T.
22. Recreate the given signal in Prob. 21 from the
frequency spectrum obtained.
23. Show that for a signal of eight sample values,
w ⫽ eⴚi>4 ⫽ (1 ⫺ i)> 12. Check by squaring.
24. Write the Fourier matrix F for a sample of eight values
explicitly.
25. CAS Problem. Calculate the inverse of the 8 ⫻ 8
Fourier matrix. Transform a general sample of eight
values and transform it back to the given data.
c11-b.qxd
10/30/10
1:31 PM
534
11.10
Page 534
CHAP. 11 Fourier Analysis
Tables of Transforms
Table I.
Fourier Cosine Transforms
See (2) in Sec. 11.8.
fˆc (w) ⫽ fc ( f )
f (x)
if 0 ⬍ x ⬍ a
1
e
2
x aⴚ1 (0 ⬍ a ⬍ 1)
ap
2 ⌫ (a)
a cos
2
Bp w
3
eⴚax (a ⬎ 0)
a
2b
Bp a ⫹ w
4
eⴚx
5
eⴚax
6
x neⴚax (a ⬎ 0)
7
e
8
cos (ax 2) (a ⬎ 0)
1
w2
p
⫺ b
cos a
4a
4
12a
9
sin (ax 2) (a ⬎ 0)
1
w2
p
⫹ b
cos a
4a
4
12a
1
0 otherwise
>2
a
2
eⴚw
2
2
2 sin aw
w
Bp
(⌫(a) see App. A3.1.)
2
>2
2
(a ⬎ 0)
cos x
0
if 0 ⬍ x ⬍ a
otherwise
10
sin ax
x
11
eⴚx sin x
x
12
J0(ax) (a ⬎ 0)
(a ⬎ 0)
1
>(4a)
eⴚw
2
12a
2
n!
2
2 n⫹1
B p (a ⫹ w )
Re (a ⫹ iw)n⫹1
Re ⫽
Real part
sin a(1 ⫹ w)
sin a(1 ⫺ w)
1
⫹
d
c
1⫺w
1⫹w
12p
p
B2
(1 ⫺ u(w ⫺ a))
1
12p
2
arctan
2
w2
1
B p 2a ⫺ w 2
2
(See Sec. 6.3.)
(1 ⫺ u(w ⫺ a)) (See Secs. 5.5, 6.3.)
c11-b.qxd
10/30/10
1:31 PM
Page 535
SEC. 11.10 Tables of Transforms
535
Table II.
Fourier Sine Transforms
See (5) in Sec. 11.8.
fˆs (w) ⫽ fs ( f )
f (x)
1
e
1 if 0 ⬍ x ⬍ a
0 otherwise
2
1> 1x
1> 1w
3
1>x 3>2
21w
4
x aⴚ1 (0 ⬍ a ⬍ 1)
5
eⴚax (a ⬎ 0)
6
eⴚax
x
7
x neⴚax (a ⬎ 0)
8
xeⴚx
9
xeⴚax
(a ⬎ 0)
>2
2
e
11
cos ax
x
12
arctan
0
sin
ap
2
(⌫(a) see App. A3.1.)
2
w
a 2
2b
Bp a ⫹ w
2
Bp
w
a
arctan
2
n!
2
2 n⫹1
B p (a ⫹ w )
Im (a ⫹ iw)n⫹1
Im ⫽
Imaginary part
>2
2
(a ⬎ 0)
sin x if 0 ⬍ x ⬍ a
10
2 ⌫ (a)
a
Bp w
weⴚw
2
1 ⫺ cos aw
d
w
c
2
Bp
otherwise
(a ⬎ 0)
2a
x
(a ⬎ 0)
w
(2a)
eⴚw
>4a
2
3>2
sin a(1 ⫹ w)
sin a(1 ⫺ w)
1
⫺
d
c
1⫺w
1⫹w
22p
p
B2
u (w ⫺ a)
12p
sin aw ⴚaw
e
w
(See Sec. 6.3.)
c11-b.qxd
10/30/10
536
1:31 PM
Page 536
CHAP. 11 Fourier Analysis
Table III. Fourier Transforms
See (6) in Sec. 11.9.
fˆ(w) ⫽ f( f )
f (x)
1
e
2
e
3
1 if ⫺b ⬍ x ⬍ b
0
1 if b ⬍ x ⬍ c
eⴚibw ⫺ eⴚicw
iw12p
0 otherwise
1
x 2 ⫹ a2
if b ⬍ x ⬍ 2b
e
eⴚax if x ⬎ 0
6
e
eax
7
e
eiax
8
e
eiax
0
otherwise
(a ⬎ 0)
if b ⬍ x ⬍ c
otherwise
0
0
if ⫺b ⬍ x ⬍ b
⫺1 ⫹ 2eibw ⫺ e ⴚ2ibw
12pw 2
1
12p(a ⫹ iw)
e(aⴚiw)c ⫺ e(aⴚiw)b
12p(a ⫺ iw)
2 sin b(w ⫺ a)
w⫺a
otherwise
Bp
if b ⬍ x ⬍ c
i eib(aⴚw) ⫺ eic(aⴚw)
a⫺w
22p
otherwise
9
eⴚax
(a ⬎ 0)
10
sin ax
x
(a ⬎ 0)
2
a
otherwise
5
0
B2
if 0 ⬍ x ⬍ b
μ 2x ⫺ b
0
p eⴚaƒwƒ
(a ⬎ 0)
x
4
2 sin bw
w
Bp
otherwise
1
12a
p
B2
eⴚw
>4a
2
if ƒ w ƒ ⬍ a; 0 if ƒ w ƒ ⬎ a
c11-b.qxd
10/30/10
1:31 PM
Page 537
Chapter 11 Review Questions and Problems
537
CHAPTER 11 REVIEW QUESTIONS AND PROBLEMS
1. What is a Fourier series? A Fourier cosine series? A
half-range expansion? Answer from memory.
2. What are the Euler formulas? By what very important
idea did we obtain them?
3. How did we proceed from 2p-periodic to generalperiodic functions?
4. Can a discontinuous function have a Fourier series? A
Taylor series? Why are such functions of interest to the
engineer?
5. What do you know about convergence of a Fourier
series? About the Gibbs phenomenon?
6. The output of an ODE can oscillate several times as
fast as the input. How come?
7. What is approximation by trigonometric polynomials?
What is the minimum square error?
8. What is a Fourier integral? A Fourier sine integral?
Give simple examples.
9. What is the Fourier transform? The discrete Fourier
transform?
10. What are Sturm–Liouville problems? By what idea are
they related to Fourier series?
11–20
FOURIER SERIES. In Probs. 11, 13, 16, 20 find
the Fourier series of f (x) as given over one period and
sketch f (x) and partial sums. In Probs. 12, 14, 15, 17–19
give answers, with reasons. Show your work detail.
11. f (x) ⫽ e
0 if ⫺2 ⬍ x ⬍ 0
2 if
0⬍x⬍2
12. Why does the series in Prob. 11 have no cosine terms?
13. f (x) ⫽ e
0 if ⫺1 ⬍ x ⬍ 0
x if
0⬍x⬍1
14. What function does the series of the cosine terms in
Prob. 13 represent? The series of the sine terms?
15. What function do the series of the cosine terms and the
series of the sine terms in the Fourier series of
ex (⫺5 ⬍ x ⬍ 5) represent?
16. f (x) ⫽ ƒ x ƒ (⫺p ⬍ x ⬍ p)
17. Find a Fourier series from which you can conclude that
1 ⫺ 1/3 ⫹ 1/5 ⫺ 1/7 ⫹ ⫺ Á ⫽ p/4.
18. What function and series do you obtain in Prob. 16 by
(termwise) differentiation?
19. Find the half-range expansions of f (x) ⫽ x
(0 ⬍ x ⬍ 1).
20. f (x) ⫽ 3x 2 (⫺p ⬍ x ⬍ p)
21–22
GENERAL SOLUTION
Solve, y s ⫹ v2y ⫽ r (t), where ƒ v ƒ ⫽ 0, 1, 2, Á , r (t) is
2p-periodic and
21. r (t) ⫽ 3t 2 (⫺p ⬍ t ⬍ p)
22. r (t) ⫽ ƒ t ƒ (⫺p ⬍ t ⬍ p)
23–25
MINIMUM SQUARE ERROR
23. Compute the minimum square error for f (x) ⫽ x> p
(⫺p ⬍ x ⬍ p) and trigonometric polynomials of
degree N ⫽ 1, Á , 5.
24. How does the minimum square error change if you
multiply f (x) by a constant k?
25. Same task as in Prob. 23, for f (x) ⫽ ƒ x ƒ > p
(⫺p ⬍ x ⬍ p). Why is E* now much smaller (by a
factor 100, approximately!)?
26–30
FOURIER INTEGRALS AND TRANSFORMS
Sketch the given function and represent it as indicated. If you
have a CAS, graph approximate curves obtained by replacing
⬁ with finite limits; also look for Gibbs phenomena.
26. f (x) ⫽ x ⫹ 1 if 0 ⬍ x ⬍ 1 and 0 otherwise; by the
Fourier sine transform
27. f (x) ⫽ x if 0 ⬍ x ⬍ 1 and 0 otherwise; by the Fourier
integral
28. f (x) ⫽ kx if a ⬍ x ⬍ b and 0 otherwise; by the Fourier
transform
29. f (x) ⫽ x if 1 ⬍ x ⬍ a and 0 otherwise; by the Fourier
cosine transform
30. f (x) ⫽ eⴚ2x if x ⬎ 0 and 0 otherwise; by the Fourier
transform
c11-b.qxd
10/30/10
538
1:31 PM
Page 538
CHAP. 11 Fourier Analysis
11
SUMMARY OF CHAPTER
Fourier Analysis. Partial Differential Equations (PDEs)
Fourier series concern periodic functions f (x) of period p ⫽ 2L, that is, by
definition f (x ⫹ p) ⫽ f (x) for all x and some fixed p ⬎ 0; thus, f (x ⫹ np) ⫽ f (x)
for any integer n. These series are of the form
ⴥ
np
np
f (x) ⫽ a0 ⫹ a aan cos
x ⫹ bn sin
xb
L
L
n⫽1
(1)
(Sec. 11.2)
with coefficients, called the Fourier coefficients of f (x), given by the Euler formulas
(Sec. 11.2)
a0 ⫽
(2)
1
2L
冮
L
ⴚL
bn ⫽
冮
1
L
冮
1
L
an ⫽
f (x) dx,
L
f (x) sin
ⴚL
L
f (x) cos
ⴚL
npx
dx
L
npx
dx
L
where n ⫽ 1, 2, Á . For period 2p we simply have (Sec. 11.1)
ⴥ
f (x) ⫽ a0 ⫹ a (an cos nx ⫹ bn sin nx)
(1*)
n⫽1
with the Fourier coefficients of f (x) (Sec. 11.1)
1
a0 ⫽
2p
冮
p
f (x) dx, an ⫽
ⴚp
1
p
冮
p
f (x) cos nx dx, bn ⫽
ⴚp
1
p
冮
p
f (x) sin nx dx.
ⴚp
Fourier series are fundamental in connection with periodic phenomena, particularly
in models involving differential equations (Sec. 11.3, Chap, 12). If f (x) is even
[ f (⫺x) ⫽ f (x)] or odd [ f (⫺x) ⫽ ⫺f (x)], they reduce to Fourier cosine or Fourier
sine series, respectively (Sec. 11.2). If f (x) is given for 0 ⬉ x ⬉ L only, it has two
half-range expansions of period 2L, namely, a cosine and a sine series (Sec. 11.2).
The set of cosine and sine functions in (1) is called the trigonometric system.
Its most basic property is its orthogonality on an interval of length 2L; that is, for
all integers m and n ⫽ m we have
冮
L
cos
ⴚL
mpx
npx
cos
dx ⫽ 0,
L
L
冮
L
sin
ⴚL
mpx
npx
sin
dx ⫽ 0
L
L
and for all integers m and n,
冮
L
cos
ⴚL
mpx
npx
sin
dx ⫽ 0.
L
L
This orthogonality was crucial in deriving the Euler formulas (2).
c05.qxd
10/28/10
1:33 PM
Page 190
190
CHAP. 5 Series Solutions of ODEs. Special Functions
1
J0
0.5
J1
0
5
10
x
Fig. 110. Bessel functions of the first kind J0 and J1
Formula (14) is surprisingly accurate even for smaller x (⬎0). For instance, it will give you good starting
values in a computer program for the basic task of computing zeros. For example, for the first three zeros of J0
you obtain the values 2.356 (2.405 exact to 3 decimals, error 0.049), 5.498 (5.520, error 0.022), 8.639 (8.654,
error 0.015), etc.
䊏
Bessel Functions J␯(x) for any ␯ ⭌ 0. Gamma Function
We now proceed from integer ␯ ⫽ n to any ␯ ⭌ 0. We had a0 ⫽ 1>(2nn!) in (9). So we
have to extend the factorial function n! to any ␯ ⭌ 0. For this we choose
a0 ⫽
(15)
1
2 ⌫(␯ ⫹ 1)
␯
with the gamma function ⌫(␯ ⫹ 1) defined by
⌫(␯ ⫹ 1) ⫽
(16)
⬁
冮e
ⴚt ␯
t dt
(␯ ⬎ ⫺1).
0
(CAUTION! Note the convention ␯ ⫹ 1 on the left but ␯ in the integral.) Integration
by parts gives
⬁
⌫(␯ ⫹ 1) ⫽ ⫺eⴚtt ␯ ` ⫹ ␯
0
⬁
冮e
ⴚt ␯ⴚ1
t
dt ⫽ 0 ⫹ ␯⌫(␯).
0
This is the basic functional relation of the gamma function
⌫(␯ ⫹ 1) ⫽ ␯⌫(␯).
(17)
Now from (16) with ␯ ⫽ 0 and then by (17) we obtain
⌫(1) ⫽
冮
⬁
0
⬁
eⴚt dt ⫽ ⫺eⴚt ` ⫽ 0 ⫺ (⫺1) ⫽ 1
0
and then ⌫(2) ⫽ 1 # ⌫(1) ⫽ 1!, ⌫(3) ⫽ 2⌫(1) ⫽ 2! and in general
(18)
⌫(n ⫹ 1) ⫽ n!
(n ⫽ 0, 1, Á ).
c05.qxd
10/28/10
1:33 PM
Page 191
SEC. 5.4 Bessel’s Equation. Bessel Functions J␯ (x)
191
Hence the gamma function generalizes the factorial function to arbitrary positive ␯.
Thus (15) with ␯ ⫽ n agrees with (9).
Furthermore, from (7) with a0 given by (15) we first have
a2m ⫽
(⫺1)m
22mm! (␯ ⫹ 1)(␯ ⫹ 2) Á (␯ ⫹ m)2␯⌫(␯ ⫹ 1)
.
Now (17) gives (␯ ⫹ 1)⌫(␯ ⫹ 1) ⫽ ⌫(␯ ⫹ 2), (␯ ⫹ 2)⌫(␯ ⫹ 2) ⫽ ⌫(␯ ⫹ 3) and so on,
so that
(␯ ⫹ 1)(␯ ⫹ 2) Á (␯ ⫹ m)⌫(␯ ⫹ 1) ⫽ ⌫(␯ ⫹ m ⫹ 1).
Hence because of our (standard!) choice (15) of a0 the coefficients (7) are simply
a2m ⫽
(19)
(⫺1)m
22m⫹␯m! ⌫(␯ ⫹ m ⫹ 1)
.
With these coefficients and r ⫽ r1 ⫽ ␯ we get from (2) a particular solution of (1), denoted
by J␯(x) and given by
(20)
ⴥ
(⫺1)mx 2m
m⫽0
22m⫹␯m! ⌫(␯ ⫹ m ⫹ 1)
J␯(x) ⫽ x ␯ a
.
J␯(x) is called the Bessel function of the first kind of order ␯. The series (20) converges
for all x, as one can verify by the ratio test.
Discovery of Properties from Series
Bessel functions are a model case for showing how to discover properties and relations of
functions from series by which they are defined. Bessel functions satisfy an incredibly large
number of relationships—look at Ref. [A13] in App. 1; also, find out what your CAS knows.
In Theorem 3 we shall discuss four formulas that are backbones in applications and theory.
THEOREM 1
Derivatives, Recursions
The derivative of J␯(x) with respect to x can be expressed by J␯ⴚ1(x) or J␯ⴙ1(x) by
the formulas
(21)
(a)
[x ␯J␯(x)] r ⫽ x ␯J␯ⴚ1(x)
(b) [x ⴚ␯J␯(x)] r ⫽ ⫺x ⴚ␯J␯⫹1(x).
Furthermore, J␯(x) and its derivative satisfy the recurrence relations
(21)
2␯
(c) J␯ⴚ1(x) ⫹ J␯⫹1(x) ⫽ x J␯(x)
(d) J␯ⴚ1(x) ⫺ J␯⫹1(x) ⫽ 2J␯r(x).
bapp03.qxd
11/3/10
A66
8:27 PM
Page A66
APP. 3 Auxiliary Material
(22)
(23)
{
sinh (x ⫾ y) ⫽ sinh x cosh y ⫾ cosh x sinh y
cosh (x ⫾ y) ⫽ cosh x cosh y ⫾ sinh x sinh y
tanh x ⫾ tanh y
tanh (x ⫾ y) ⫽ ᎏᎏ
1 ⫾ tanh x tanh y
Gamma function (Fig. 553 and Table A2 in App. 5). The gamma function ⌫(␣) is defined
by the integral
⌫(␣) ⫽
(24)
冕e
⬁
ⴚt ␣ⴚ1
t
(␣ ⬎ 0),
dt
0
which is meaningful only if ␣ ⬎ 0 (or, if we consider complex ␣, for those ␣ whose real
part is positive). Integration by parts gives the important functional relation of the gamma
function,
⌫(␣ ⫹ 1) ⫽ ␣⌫(␣).
(25)
From (24) we readily have ⌫(1) ⫽ 1; hence if ␣ is a positive integer, say k, then by
repeated application of (25) we obtain
⌫(k ⫹ 1) ⫽ k!
(26)
(k ⫽ 0, 1, • • •).
This shows that the gamma function can be regarded as a generalization of the elementary
factorial function. [Sometimes the notation (␣ ⫺ 1)! is used for ⌫(␣), even for noninteger
values of ␣, and the gamma function is also known as the factorial function.]
By repeated application of (25) we obtain
⌫(␣ ⫹ 2)
⌫(␣ ⫹ k ⫹ 1)
⌫(␣ ⫹ 1)
⌫(␣) ⫽ ᎏᎏ ⫽ ᎏᎏ ⫽ • • • ⫽ ᎏᎏᎏᎏ
␣(␣ ⫹ 1)
␣(␣ ⫹ 1)(␣ ⫹ 2) • • • (␣ ⫹ k)
␣
Γ(α )
5
–4
–2
2
4
–2
–4
Fig. 553. Gamma function
α
bapp03.qxd
11/3/10
8:27 PM
Page A67
SEC. A3.1 Formulas for Special Functions
A67
and we may use this relation
⌫(␣ ⫹ k ⫹ 1)
⌫(␣) ⫽ ᎏᎏᎏ
␣(␣ ⫹ 1) • • • (␣ ⫹ k)
(27)
(␣ ⫽ 0, ⫺1, ⫺2, • • •),
for defining the gamma function for negative ␣ (⫽ ⫺1, ⫺2, • • •), choosing for k the
smallest integer such that ␣ ⫹ k ⫹ 1 ⬎ 0. Together with (24), this then gives a definition
of ⌫(␣) for all ␣ not equal to zero or a negative integer (Fig. 553).
It can be shown that the gamma function may also be represented as the limit of a
product, namely, by the formula
(28)
n! n␣
⌫(␣) ⫽ lim ᎏᎏᎏᎏ
n*⬁ ␣(␣ ⫹ 1)(␣ ⫹ 2) • • • (␣ ⫹ n)
(␣ ⫽ 0, ⫺1, • • •).
From (27) or (28) we see that, for complex ␣, the gamma function ⌫(␣) is a meromorphic
function with simple poles at ␣ ⫽ 0, ⫺1, ⫺2, • • • .
An approximation of the gamma function for large positive ␣ is given by the Stirling
formula
␣ ␣
⌫(␣ ⫹ 1) ⬇ 兹2苶
␲␣ ( ᎏ )
e
(29)
where e is the base of the natural logarithm. We finally mention the special value
⌫(_12) ⫽ 兹␲
苶.
(30)
Incomplete gamma functions
冕e
x
(31)
P(␣, x) ⫽
ⴚt ␣ⴚ1
t
Q(␣, x) ⫽
dt,
ⴚt ␣ⴚ1
t
dt
(␣ ⬎ 0)
x
0
(32)
⬁
冕e
⌫(␣) ⫽ P(␣, x) ⫹ Q(␣, x)
Beta function
冕t
1
(33)
B(x, y) ⫽
(1 ⫺ t)yⴚ1 dt
xⴚ1
0
Representation in terms of gamma functions:
(34)
⌫(x)⌫(y)
B(x, y) ⫽ ᎏ
⌫(x ⫹ y)
Error function (Fig. 554 and Table A4 in App. 5)
(35)
(36)
2
erf x ⫽ ᎏ
兹␲
苶
冕e
x
ⴚt2
dt
0
2
x3
x5
x7
erf x ⫽ ᎏ (x ⫺ ᎏ ⫹ ᎏ ⫺ ᎏ ⫹ ⫺ • • •)
1!3
2!5
3!7
兹␲
苶
(x ⬎ 0, y ⬎ 0)
bapp03.qxd
11/3/10
A68
8:27 PM
Page A68
APP. 3 Auxiliary Material
erf x
1
0.5
–2
–1
1
2
x
–0.5
–1
Fig. 554. Error function
erf (⬁) ⫽ 1, complementary error function
2
erfc x ⫽ 1 ⫺ erf x ⫽ ᎏ
兹␲
苶
(37)
⬁
冕e
ⴚt2
dt
x
Fresnel integrals1 (Fig. 555)
冕 cos (t ) dt,
x
C(x) ⫽
(38)
冕 sin (t ) dt
x
S(x) ⫽
2
0
2
0
C(⬁) ⫽ 兹苶,
␲/8 S(⬁) ⫽ 兹苶,
␲/8 complementary functions
(39)
s(x) ⫽
⬁
␲
ᎏ ⫺ C(x) ⫽
8
冕
␲
ᎏ ⫺ S(x) ⫽
8
冕
冪莦
c(x) ⫽
冪莦
cos (t 2 ) dt
x
⬁
sin (t 2 ) dt
x
Sine integral (Fig. 556 and Table A4 in App. 5)
sin t
冕ᎏ
dt
t
x
Si(x) ⫽
(40)
0
y
1
C (x)
0.5
S(x)
0
1
2
3
4
x
Fig. 555. Fresnel integrals
1
AUGUSTIN FRESNEL (1788–1827), French physicist and mathematician. For tables see Ref. [GenRef1].
bapp03.qxd
11/3/10
8:27 PM
Page A69
SEC. A3.2 Partial Derivatives
A69
Si(x)
2
1
0
5
10
x
Fig. 556. Sine integral
Si(⬁) ⫽ ␲ /2, complementary function
(41)
␲
si(x) ⫽ ᎏ ⫺ Si(x) ⫽
2
⬁
冕
x
sin t
ᎏ dt
t
Cosine integral (Table A4 in App. 5)
ci(x) ⫽
(42)
⬁
冕
x
cos t
ᎏ dt
t
(x ⬎ 0)
Exponential integral
Ei(x) ⫽
(43)
⬁
冕
x
eⴚt
ᎏ dt
t
(x ⬎ 0)
Logarithmic integral
li(x) ⫽
(44)
冕
x
0
A3.2
dt
ᎏ
ln t
Partial Derivatives
For differentiation formulas, see inside of front cover.
Let z ⫽ ƒ(x, y) be a real function of two independent real variables, x and y. If we keep
y constant, say, y ⫽ y1, and think of x as a variable, then ƒ(x, y1) depends on x alone. If
the derivative of ƒ(x, y1) with respect to x for a value x ⫽ x1 exists, then the value of this
derivative is called the partial derivative of ƒ(x, y) with respect to x at the point (x1, y1)
and is denoted by
⭸ƒ
ᎏj
⭸x
or by
(x1,y1)
⭸z
ᎏj
⭸x
.
(x1,y1)
Other notations are
ƒx (x1, y1)
and
zx (x1, y1);
these may be used when subscripts are not used for another purpose and there is no danger
of confusion.