An Instructor’s Solutions Manual to Accompany

ADVANCED ENGINEERING MATHEMATICS,

8TH EDITION, SI
PETER V. O’NEIL
SI EDITION PREPARED BY QABOOS IMRAN
© 2018, 2012 Cengage Learning ISBN: 978-1-337-27453-1
WCN: 01-100-101 Cengage Learning
20 Channel Center Street
ALL RIGHTS RESERVED. No part of this work covered by the Boston, MA 02210
copyright herein may be reproduced, transmitted, stored, or USA
used in any form or by any means graphic, electronic, or
mechanical, including but not limited to photocopying, Cengage Learning is a leading provider of customized
recording, scanning, digitizing, taping, Web distribution, learning solutions with office locations around the globe,
information networks, or information storage and retrieval including Singapore, the United Kingdom, Australia,
systems, except as permitted under Section 107 or 108 of the Mexico, Brazil, and Japan. Locate your local office at:
1976 United States Copyright Act, without the prior written international.cengage.com/region.
permission of the publisher except as may be permitted by the
license terms below. Cengage Learning products are represented in
Canada by Nelson Education, Ltd.

For your course and learning solutions, visit

For product information and technology assistance, contact us at
www.cengage.com/engineering.
Cengage Learning Academic Resource Center,
1-800-423-0563.
Purchase any of our products at your local college
For permission to use material from this text or product, submit
store or at our preferred online store
all requests online at www.cengage.com/permissions. www.cengagebrain.com.
Further permissions questions can be emailed to
permissionrequest@cengage.com.

NOTE: UNDER NO CIRCUMSTANCES MAY THIS MATERIAL OR ANY PORTION THEREOF BE SOLD, LICENSED, AUCTIONED,
OR OTHERWISE REDISTRIBUTED EXCEPT AS MAY BE PERMITTED BY THE LICENSE TERMS HEREIN.

READ IMPORTANT LICENSE INFORMATION

Dear Professor or Other Supplement Recipient: electronically must be through a password-protected site, with all
copy and download functionality disabled, and accessible solely by
Cengage Learning has provided you with this product (the your students who have purchased the associated textbook for the
“Supplement”) for your review and, to the extent that you adopt Course. You may not sell, license, auction, or otherwise redistribute
the associated textbook for use in connection with your course the Supplement in any form. We ask that you take reasonable
(the “Course”), you and your students who purchase the steps to protect the Supplement from unauthorized use,
textbook may use the Supplement as described below. reproduction, or distribution. Your use of the Supplement indicates
Cengage Learning has established these use limitations in your acceptance of the conditions set forth in this Agreement. If you
response to concerns raised by authors, professors, and other do not accept these conditions, you must return the Supplement
users regarding the pedagogical problems stemming from unused within 30 days of receipt.
unlimited distribution of Supplements.
All rights (including without limitation, copyrights, patents, and trade
Cengage Learning hereby grants you a nontransferable license secrets) in the Supplement are and will remain the sole and
to use the Supplement in connection with the Course, subject to exclusive property of Cengage Learning and/or its licensors. The
the following conditions. The Supplement is for your personal, Supplement is furnished by Cengage Learning on an “as is” basis
noncommercial use only and may not be reproduced, posted without any warranties, express or implied. This Agreement will be
electronically or distributed, except that portions of the governed by and construed pursuant to the laws of the State of
Supplement may be provided to your students IN PRINT FORM New York, without regard to such State’s conflict of law rules.
ONLY in connection with your instruction of the Course, so long
as such students are advised that they may not copy or Thank you for your assistance in helping to safeguard the integrity
distribute any portion of the Supplement to any third party. Test of the content contained in this Supplement. We trust you find the
banks and other testing materials may be made available in the Supplement a useful teaching tool.
classroom and collected at the end of each class session, or
posted electronically as described herein. Any material posted
 INSTRUCTOR'S SOLUTIONS MANUAL
TO ACCOMPANY

Advanced Engineering
Mathematics
8th EDITION, SI

PETER V. O’NEIL

SI Edition Prepared by
QABOOS IMRAN
Contents

1 First-Order Differential Equations 1

1.1 Terminology and Separable Equations 1
1.2 The Linear First-Order Equation 13
1.3 Exact Equations 19
1.4 Homogeneous, Bernoulli and Riccati Equations 29
2 Second-Order Differential Equations 39
2.1 The Linear Second-Order Equation 39
2.2 The Constant Coefficient Homogeneous Equation 43
2.3 Particular Solutions of the Nonhomogeneous Equation 48
2.4 The Euler Differential Equation 55
2.5 Series Solutions 60
3 The Laplace Transform 71
3.1 Definition and Notation 71
3.2 Solution of Initial Value Problems 74
3.3 The Heaviside Function and Shifting Theorems 79
3.4 Convolution 88
3.5 Impulses and the Dirac Delta Function 94
3.6 Systems of Linear Differential Equations 95

iii
iv CONTENTS

4 Sturm-Liouville Problems and Eigenfunction Expansions 103

4.1 Eigenvalues and Eigenfunctions and Sturm-Liouville Problems 103
4.2 Eigenfunction Expansions 109
4.3 Fourier Series 116
5 The Heat Equation 139
5.1 Diffusion Problems on a Bounded Medium 139
5.2 The Heat Equation With a Forcing Term F (x, t) 149
5.3 The Heat Equation on the Real Line 152
5.4 The Heat Equation on a Half-Line 155
5.5 The Two-Dimensional Heat Equation 157
6 The Wave Equation 159
6.1 Wave Motion on a Bounded Interval 159
6.2 Wave Motion in an Unbounded Medium 169
6.3 d’Alembert’s Solution and Characteristics 175
6.4 The Wave Equation With a Forcing Term K(x, t) 192
6.5 The Wave Equation in Higher Dimensions 194
7 Laplace’s Equation 199
7.1 The Dirichlet Problem for a Rectangle 199
7.2 The Dirichlet Problem for a Disk 204
7.3 The Poisson Integral Formula 207
7.4 The Dirichlet Problem for Unbounded Regions 207
7.5 A Dirichlet Problem in 3 Dimensions 210
7.6 The Neumann Problem 213
7.7 Poisson’s Equation 219
8 Special Functions and Applications 223
8.1 Legendre Polynomials 223
8.2 Bessel Functions 237
8.3 Some Applications of Bessel Functions 253
9 Transform Methods of Solution 265
9.1 Laplace Transform Methods 265
9.2 Fourier Transform Methods 270
9.3 Fourier Sine and Cosine Transforms 273
10 Vectors and the Vector Space Rn 277
10.1 Vectors in the Plane and 3− Space 277
10.2 The Dot Product 229
10.3 The Cross Product 280
10.4 n− Vectors and the Algebraic Structure of Rn 282
10.5 Orthogonal Sets and Orthogonalization 286
10.6 Orthogonal Complements and Projections 289
11 Matrices, Determinants and Linear Systems 293
11.1 Matrices and Matrix Algebra 293
11.2. Row Operations and Reduced Matrices 297
11.3 Solution of Homogeneous Linear Systems 301
11.4 Nonhomogeneous Systems 308
11.5 Matrix Inverses 315
11.6 Determinants 317
11.7 Cramer’s Rule 320
11.8 The Matrix Tree Theorem 322
 v

12 Eigenvalues, Diagonalization and Special Matrices 325

12.1 Eigenvalues and Eigenvectors 325
12.2 Diagonalization 329
12.3 Special Matrices and Their Eigenvalues and Eigenvectors 334
12.4 Quadratic Forms 338
13 Systems of Linear Differential Equations 341
13.1 Linear Systems 341
13.2 Solution of X0 = AX When A Is Constant 343
13.3 Exponential Matrix Solutions 350
13.4 Solution of X0 = AX + G for Constant A 352
13.5 Solution by Diagonalization 355
14 Nonlinear Systems and Qualitative Analysis 361
14.1 Nonlinear Systems and Phase Portraits 361
14.2 Critical Points and Stability 365
14.3 Almost Linear Systems 366
14.4 Linearization 371
15 Vector Differential Calculus 375
15.1 Vector Functions of One Variable 375
15.2 Velocity, Acceleration and Curvature 378
15.3 The Gradient Field 383
15.4 Divergence and Curl 387
15.5 Streamlines of a Vector Field 389
16 Vector Integral Calculus 393
16.1 Line Integrals 393
16.2 Green’s Theorem 395
16.3 Independence of Path and Potential Theory 400
16.4 Surface Integrals 407
16.5 Applications of Surface Integrals 410
16.6 Gauss’s Divergence Theorem 414
16.7 Stokes’s Theorem 416
17 Fourier Series 421
17.1 Fourier Series on [−L, L] 421
17.2 Sine and Cosine Series 425
17.3 Integration and Differentiation of Fourier Series 430
17.4 Properties of Fourier Coefficients 432
17.5 Phase Angle Form 434
17.6 Complex Fourier Series 437
17.7 Filtering of Signals 440
vi CONTENTS

18 Fourier Transforms 443

18.1 The Fourier Transform 443
18.2 Fourier sine and Cosine Transforms 450
19 Complex Numbers and Functions 453
19.1 Geometry and Arithmetic of Complex Numbers 453
19.2 Complex Functions 457
19.3 The Exponential and Trigonometric Functions 463
19.4 The Complex Logarithm 469
19.5 Powers 470
20 Complex Integration 475
20.1 The Integral of a Complex Function 475
20.2 Cauchy’s Theorem 479
20.3 Consequences of Cauchy’s Theorem 481
21 Series Representations of Functions 487
21.1 Power Series 487
21.2 The Laurent Expansion 494
22 Singularities and the Residue Theorem 499
22.1 Classification of Singularities 499
22.2 The Residue Theorem 501
22.3 Evaluation of Real Integrals 507
23 Conformal Mappings 517
23.1 The Idea of a Conformal Mapping 517
23.2 Construction of Conformal Mappings 535
 Chapter 1

First-Order Differential
Equations

1.1 Terminology and Separable Equations

1. The differential equation is separable because it can be written
dy
3y 2 = 4x,
dx
or, in differential form,
3y 2 dy = 4x dx.
Integrate to obtain
y 3 = 2x2 + k.
This implicitly defines a general solution, which can be written explicitly
as
y = (2x2 + k)1/3 ,
with k an arbitrary constant.
2. Write the differential equation as
dy
x = −y,
dx
which separates as
1 1
dy = − dx
y x
if x 6= 0 and y =
6 0. Integrate to get

ln |y| = − ln |x| + k.

Then ln |xy| = k, so
xy = c

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

with c constant (c = ek ). y = 0 is a singular solution, satisfying the

original differential equation.

3. If cos(y) 6= 0, the differential equation is

y sin(x + y)
=
dx cos(y)
sin(x) cos(y) + cos(x) sin(y)
=
cos(y)
= sin(x) + cos(x) tan(y).

There is no way to separate the variables in this equation, so the differen-

tial equation is not separable.

4. Write the differential equation as

dy
ex ey = 3x,
dx
which separates in differential form as

ey dy = 3xe−x dx.

Integrate to get
ey = −3e−x (x + 1) + c,
with c constant. This implicitly defines a general solution.

5. The differential equation can be written

dy
x = y 2 − y,
dx
or
1 1
dy = dx,
y(y − 1) x
and is therefore separable. Separating the variables assumes that y 6= 0
6 1. We can further write
and y =
 
1 1 1
− dy = dx.
y−1 y x

Integrate to obtain

ln |y − 1| − ln |y| = ln |x| + k.

Using properties of the logarithm, this is

y−1
ln = k.
xy

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.1. TERMINOLOGY AND SEPARABLE EQUATIONS 3

Then
y−1
= c,
xy
with c = ek constant. Solve this for y to obtain the general solution
1
y= .
1 − cx
y = 0 and y = 1 are singular solutions because these satisfy the differential
equation, but were excluded in the algebra of separating the variables.

6. The differential equation is not separable.

7. The equation is separable because it can be written in differential form as

sin(y) 1
dy = dx.
cos(y) x

This assumes that x 6= 0 and cos(y) 6= 0. Integrate this equation to obtain

− ln | cos(y)| = ln |x| + k.

This implicitly defines a general solution. From this we can also write

sec(y) = cx

with c constant.
The algebra of separating the variables required that cos(y) 6= 0. Now
cos(y) = 0 if y = (2n+1)π/2, with n any integer. Now y = (2n+1)π/2 also
satisfies the original differential equation, so these are singular solutions.

8. The differential equation itself requires that y 6= 0 and x 6= −1. Write the
equation as
x dy 2y 2 + 1
=
y dx x
and separate the variables to get
1 1
dy = dx.
y(2y 2 + 1) x(x + 1)

Use a partial fractions decomposition to write this as

   
1 2y 1 1
− dy = − dx.
y 2y 2 + 1 x x+1

Integrate to obtain
1
ln |y| − ln(1 + 2y 2 ) = ln |x| − ln |x + 1| + c
2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

with c constant. This implicitly defines a general solution. We can go a

step further by writing this equation as
!  
y x
ln p = ln +c
1 + 2y 2 x+1

and take the exponential of both sides to get

 
y x
p =k ,
1 + 2y 2 x+1

which also defines a general solution.

9. The differential equation is
dy
= ex − y + sin(y),
dx
and this is not separable. It is not possible to separate all terms involving
x on one side of the equation and all terms involving y on the other.
10. Substitute

sin(x − y) = sin(x) cos(y) − cos(x) sin(y),

cos(x + y) = cos(x) cos(y) − sin(x) sin(y),
and
cos(2x) = cos2 (x) − sin2 (x)

into the differential equation to get the separated differential form

(cos(y) − sin(y)) dy = (cos(x) − sin(x)) dx.

Integrate to obtain the implicitly defined general solution

cos(y) + sin(y) = cos(x) + sin(x) + c.

11. If y 6= −1 and x 6= 0, we obtain the separated equation

y2 1
dy = dx.
y+1 x
To make the integration easier, write this as
 
1 1
y−1+ dy = dx.
1+y x
Integrate to obtain
1 2
y − y + ln |1 + y| = ln |x| + c.
2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.1. TERMINOLOGY AND SEPARABLE EQUATIONS 5

This implicitly defines a general solution. The initial condition is y(3e2 ) =

2, so put y = 2 and x = 3e2 to obtain

2 − 2 + ln(3) = ln(3e2 ) + c.

Now
ln(3e2 ) = ln(3) + ln(e2 ) = ln(3) + 2,
so
ln(3) = ln(3) + 2 + c.
Then c = −2 and the solution of the initial value problem is implicitly
defined by
1 2
y − y + ln |1 + y| = ln |x| − 2.
2
12. Integrate
1
dy = 3x2 dx,
y+2
assuming that y 6= −2, to obtain

ln |2 + y| = x3 + c.

This implicitly defines a general solution. To have y(2) = 8, let x = 2 and

y = 8 to obtain
ln(10) = 8 + c.
The solution of the initial value problem is implicitly defined by

ln |2 + y| = x3 + ln(10) − 8.

We can take this a step further and write

 
2+y
ln = x3 − 8.
10

By taking the exponential of both sides of this equation we obtain the

explicit solution
3
y = 10ex −8 − 2.

13. With ln(y x ) = x ln(y), we obtain the separated equation

ln(y)
dy = 3x dx.
y
Integrate to obtain
(ln(y))2 = 3x2 + c.
For y(2) = e3 , we need

(ln(e3 ))2 = 3(4) + c,

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

or 9 = 12 + c. Then c = −3 and the solution of the initial value problem

is defined by
(ln(y))2 = 3x2 − 3.
Solve this to obtain the explicit solution
√ 2
y = e 3(x −1)

if |x| > 1.
2 2
14. Because ex−y = ex e−y , the variables can be separated to obtain
2
2yey dy = ex dx.

Integrate to get
2
ey = ex + c.
To satisfy y(4) = −2 we need

e4 = e4 + c

so c = 0 and the solution of the initial value problem is implicitly defined

by
2
ey = ex ,
which reduces to the simpler equation

x = y2 .
√
Because y(4) = −2, the explicit solution is y = − x for x > 0.
15. Separate the variables to obtain

y cos(3y) dy = 2x dx.

Integrate to get
1 1
y sin(3y) + cos(3y) = x2 + c,
3 9
which implicitly defines a general solution. For y(2/3) = π/3, we need
1π 1 4
sin(π) + cos(π) = + c.
33 9 9
This reduces to
1 4
−
= + c,
9 9
so c = −5/9 and the solution of the initial value problem is implicitly
defined by
1 1 5
y sin(3y) + cos(3y) = x2 − ,
3 9 9
or
3y sin(3y) + cos(3y) = 9x2 − 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.1. TERMINOLOGY AND SEPARABLE EQUATIONS 7

16. Let T (t) be the temperature function. By Newton’s law of cooling, T 0 (t) =
k(T − 15.5) for some constant k to be determined. This equation is sepa-
rable and is easily solved to obtain:

T (t) = 15.5 + 16.7ekt .

To determine k, use the fact that T (10) = 31.1:

T (10) = 15.5 + 16.7e10k = 31.1.

Then
31.1 − 15.5 15.6
e10k = = ,
16.7 16.7
so
1
ln(15.6/16.7).
k=
10
Now we know the temperature function completely:

T (t) = 15.5 + 16.7ekt

= 15.5 + 16.7eln(15.6/16.7)t/10 .

We want to know T (20), so compute

T (20) = 15.5 + 16.7e2 ln(15.6/16.7)

 2
15.6
= 15.5 + 16.7 ≈ 30.06.
16.7
To see how long it will take for the object to reach 18.3 degrees, solve for
t in
T (t) = 15.5 + 16.7eln(15.6/16.7)t/10 = 18.3.
Then
16.7eln(15.6/16.7)t/10 = 2.8,
so
2.8
eln(15.6/16.7)t/10 =
16.7
and then    
15.6 t 2.8
ln = ln .
16.7 10 16.7
Finally,  
ln(2.8/16.7)
t = 10 ≈ 262.08.
ln(15.6/16.7)
17. Suppose the thermometer was removed from the house at time t = 0, and
let T (t) be the temperature function. Let A be the ambient temperature
outside the house (assumed constant). By Newton’s law,

T 0 (t) = k(t − A).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

We are also given that T (0) = 21.1 and T (5) = 15.5. Further, fifteen
minutes after being removed from the house, the thermometer reads 10.2,
so
T (15) = 10.2.
We want to determine A, the outside temperature. From the differential
equation for T ,
1
dT = kdt.
T −A
Integrate this, as we have done before, to get

T (t) = A + cekt .

Now,
T (0) = 21.1 = A + c,
so c = 21.1 − A and

T (t) = A + (21.1 − A)ekt .

Now use the other two conditions:

T (5) = A + (21.1 − A)e5k = 15.5 and T (15) = A + (21.1 − A)e15k = 10.2.

From the equation for T (5), solve for e5k to get

15.5 − A
e5k = .
21.1 − A
Then  3
15.5 − A
e15k = (e5k )3 = .
21.1 − A
Substitute this into the equation T (15) = 10.2 to get
 3
15.5 − A
(21.1 − A) = 10.2 − A.
21.1 − A
Then
(15.5 − A)3 = (10.2 − A)(21.1 − A)2 .
This is a cubic equation for A, with distinct roots A ≈ 7.3 and A ≈ 18.9.
Certainly A must be less than 10.2, so A ≈ 7.3. This is the outside
temperature.
18. The amount A(t) of radioactive material at time t is modeled by

A0 (t) = kA, A(0) = e3

together with the given half-life of the material,

1 3
A(ln(2)) = e .
2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.1. TERMINOLOGY AND SEPARABLE EQUATIONS 9

Solve this (as in the text) to obtain

 t/ ln(2)
1
A(t) = e3 .
2
Then  3/ ln(2)
13
A(3) = e = 1 tonne.
2
19. The problem is like Problem 18, and we find that the amount of Uranium-
235 at time t is
 t/(4.5(109 ))
1
U (t) = 10 ,
2
with t in years. Then
 1/4.5
9 1
U (10 ) = 10 ≈ 8.57 kg.
2

20. At time t there will be A(t) = 12ekt grams, and A(4) = 12e4k = 9.1. Solve
this for k to get  
1 9.1
k = ln .
4 12
The half-life of this element is the time t∗ it will take for there to be 6
grams, so ∗
A(t∗ ) = 6 = 12eln(9.1/12)t /4 .
Solve this to get
4 ln(1/2)
t∗ = ≈ 10.02 minutes.
ln(9.1/12)

21. Let Z ∞
2
−(x/t)2
I(x) = e−t dt.
0
The integral we want is I(3). Compute
Z ∞
1 −t2 −(x/t)2
I 0 (x) = −2x 2
e dt.
0 t
Let u = x/t, so t = x/u and
x
dt = − du.
u2
Then
0
u2
 
−x
Z
2
−u2
I 0 (x) = −2x e−(x/u) du
∞ x2 u2
= −2I(x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 10 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

Then I(x) satisfies the separable differential equation I 0 = −2I, with

general solution of the form I(x) = ce−2x . Now observe that
Z ∞ √
−t2 π
I(0) = e dt = = c,
0 2
in which we used a standard integral that arises often in statistics. Then
√
π −2x
I(x) = e .
2
Finally, put x = 3 for the particular integral of interest:
Z ∞ √
2 2 π −6
I(3) = e−t −(9/t) dt = e .
0 2

22. Begin with the logistic equation

P 0 (t) = aP (t) − bP (t)2 ,

in which a and b are positive constants. Then

dP
= P (a − bP )
dt
so
1
dP = dt
P (a − bP )
and the variables are separated. To make the integration easier, write this
equation as  
11 b 1
+ dP = dt.
aP a a − bP
Integrate to obtain
1 b
ln(P ) − ln(a − bP ) = t + c.
a a
if P (t) > − and a − bP (t) > 0. Using properties of the logarithm, we can
write this equation as
 
P
ln = at + k,
a − bP
in which k = ac is still constant. Then
P
= eat+k = ek eat = Keat ,
a − bP
in which K = ek is a positive constant. Now suppose the initial population
(say at time zero) is p0 . Then P (0) = p0 and
p0
= K.
a − bp0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.1. TERMINOLOGY AND SEPARABLE EQUATIONS 11

We now have
P p0
= eat .
a − bP a − bp0
It is a straightforward algebraic manipulation to solve this for P (t):
ap0
P (t) = eat .
a − bp0 + bp0
This is the solution of the logistic equation with P (0) = p0 .
Because a − bp0 > 0 by assumption, then
bp0 eat < a − bp0 + bpeat ,
so
ap0 at a
P (t) < e = .
bp0 eat b
This means that this population function is bounded above. Further, by
multiplying the numerator and denominator of P (t) by e(−at, we have
ap0
lim P (t) = lim
t→∞ t→∞(a − bp0 )e−at + bp0
ap0 a
= lim = .
t→∞ bp0 b

23. With a and b as given, and p0 = 3, 929, 214 (the population in 1790), the
logistic population function for the United States is
123, 141.5668
P (t) = e0.03134t .
0.03071576577 + 0.0006242342283e0.03134t
If we attempt an exponential model Q(t) = Aekt , then take A = Q(0) =
3, 929, 214, the population in 1790. To find k, use the fact that
Q(10) = 5308483 = 3929214e10k
and we can solve for k to get
 
1 5308483
k= ln ≈ 0.03008667012.
10 3929214
The exponential model, using these two data points (1790 and 1800 pop-
ulations), is
Q(t) = 3929214e0.03008667012t .
Table 1.1 uses Q(t) and P (t) to predict later populations from these two
initial figures. The logistic model remains quite accurate until about 1960,
at which time it loses accuracy quickly. The exponential model becomes
quite inaccurate by 1870, after which the error becomes so large that it
is not worth computing further. Exponential models do not work well
over time with complex populations, such as fish in the ocean or countries
throughout the world.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 12 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

year population P (t) percent error Q(t) percent error

1790 3,929,213 3,929,214 0 3,929,214 0
1800 5,308,483 5,336,313 0.52 5,308,483 0
1810 7,239,881 7,228,171 -0.16 7,179,158 -0.94
1820 9,638,453 9,757,448 1.23 7,179,158 0.53
1830 12,886,020 13,110,174 1.90 13,000,754 1.75
1840 17,169,453 17,507,365 2.57 17,685,992 3.61
1850 23,191,876 23.193,639 0.008 23,894,292 3.03
1860 31,443,321 30,414,301 -3.27 32,281,888 2.67
1870 38,558,371 39,374,437 2.12 43,613,774 13.11
1880 50,189,209 50,180,383 -0.018 58,923,484 17.40
1890 62,979766 62,772,907 -0.33 79,073,491 26.40
1900 76,212,168 76,873,907 0.87 107,551,857 41.12
1910 92,228,496 91,976,297 -0.27 145,303,703 57.55
1920 106,021,537 107,398,941 1.30 196,312,254 83.16
1930 123,202,624 122,401,360 -0.65
1940 132,164,569 136,329,577 3.15
1950 151,325,798 148,679,224 -1.75
1960 179,323,175 150,231,097 -11.2
1970 203,302,031 167,943,428 -17.39
1980 226,547,042 174,940,040 -22.78

Table 1.1: Census data for Problem 23

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.2. THE LINEAR FIRST-ORDER EQUATION 13

1.2 The Linear First-Order Equation

1. With p(x) = −3/x, and integrating factor is
R
e (−3/x) dx
= e−3 ln(x) = x−3

for x > 0. Multiply the differential equation by x−3 to get

x−3 y 0 − 3x−4 = 2x−1 .

or
d −3 2
(x y) = .
dx x
Integrate to get
x−3 y = 2 ln(x) + c,
with c an arbitrary constant. For x > 0 we have a general solution

y = 2x3 ln(x) + cx3 .

In the last integration, we can allow x < 0 by replacing ln(x) with ln |x|
to derive the solution
y = 2x3 ln |x| + cx3
for x 6= 0.
R
2. e dx = ex is an integrating factor. Multiply the differential equation by
ex to get
1
y 0 ex + yex = (e2x − 1).
2
Then
1
(ex y)0 = (e2x − 1)
2
and an integration gives us
1 2x 1
ex y = e − x + c.
4 2
Then
1 x 1 −x
e − xe + ce−x
y=
4 2
is a general solution, with c an arbitrary constant.
R
3. e 2 dx = e2x is an integrating factor. Multiply the differential equation by
e2x :
y 0 e2x + 2ye2x = xe2x ,
or
(e2x y)0 = xe2x .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 14 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

Integrate to get
1 2x 1 2x
e2x y = xe − e + c.
2 4
giving us the general solution
1 1
y= x − + ce−2x .
2 4

4. For an integrating factor, compute

R
e sec(x) dx
= eln | sec(x)+tan(x)| = sec(x) + tan(x).

Multiply the differential equation by this integrating factor:

y 0 (sec(x) + tan(x)) + sec(x)(sec(x) + tan(x))y

= y 0 (sec(x) + tan(x)) + (sec(x) tan(x) + sec2 (x))y
= ((sec(x) + tan(x))y)0
= cos(x)(sec(x) + tan(x))
= 1 + sin(x).

We therefore have

((sec(x) + tan(x))y)0 = 1 + sin(x).

Integrate to get

y(sec(x) + tan(x)) = x − cos(x) + c.

Then
x − cos(x) + c
y= .
sec(x) + tan(x)
This is a general solution. If we wish, we can also observe that

1 cos(x)
=
sec(x) + tan(x) 1 + sin(x)

to obtain
 
cos(x)
y = (x − cos(x) + c)
1 + sin(x)
2
x cos(x) − cos (x) + c cos(x)
= .
1 + sin(x)

5. First determine the integrating factor

R
−2 dx
e = e−2x .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.2. THE LINEAR FIRST-ORDER EQUATION 15

Multiply the differential equation by e−2x to get

(e−2x y)0 = −8x2 e−2x .
Integrate to get
Z
e−2x y = −8x2 e−2x dx = 4x2 e−2x + 4xe−2x + 2e−2x + c.

This yields the general solution

y = 4x2 + 4x + 2 + ce2x .
R
6. e 3 dx = e3x is an integrating factor. Multiply the differential equation by
e3x to get
(e3x y)0 = 5e5x − 6e3x .
Integrate this equation:
e3x y = e5x − 2e3x + c.
Now we have a general solution
y = e2x − 2 + ce−3x .
We need
y(0) = 2 = 1 − 2 + c,
so c = 3. The unique solution of the initial value problem is
y = e2x + 3e−3x − 2.

7. x − 2 is an integrating factor for the differential equation because

R
(1/(x−2)) dx
e = eln(x−2) = x − 2.
Multiply the differential equation by x − 2 to get
((x − 2)y)0 = 3x(x − 2).
Integrate to get
(x − 2)y = x3 − 3x2 + c.
This gives us the general solution
1
y= (x3 − 3x2 + c).
x−2
Now we need
y(3) = 27 − 27 + c = 4,
so c = 4 and the solution of the initial value problem is
1
y= (x3 − 3x2 + 4).
x−2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS
R
8. e (−1) dx = e−x is an integrating factor. Multiply the differential equation
by e−x to get:
(ye−x )0 = 2e3x .
Integrate to get
2 3x
ye−x = e + c,
3
and we have the general solution
2 4x
y= e + cex .
3
We need
2
+ c = −3,
y(0) =
3
so c = −11/3 and the initial value problem has the solution

2 4x 11 x
y= e − e .
3 3

9. First derive the integrating factor

2
R
(2/(x+1)) dx
e = e2 ln(x+1) = eln((x+1) )
= (x + 1)2 .

Multiply the differential equation by (x + 1)2 to obtain

0
(x + 1)2 y = 3(x + 1)2 .

Integrate to obtain
(x + 1)2 y = (x + 1)3 + c.
Then
c
y =x+1+ .
(x + 1)2
Now
y(0) = 1 + c = 5
so c = 4 and the initial value problem has the solution
4
y =x+1+ .
(x + 1)2

10. An integrating factor is

5/9
R
(5/9x) dx
e = e(5/9) ln(x) = eln(x )
= x5/9 .

Multiply the differential equation by x5/9 to get

(yx5/9 )0 = 3x32/9 + x14/9 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.2. THE LINEAR FIRST-ORDER EQUATION 17

Integrate to get
27 41/9 9
yx5/9 = x + x23/9 + c.
41 23
Then
27 4 9
y= x + x2 + cx−5/9 .
41 23
Finally, we need
27 9
y(−1) = + − c = 4.
41 23
Then c = −2782/943, so the initial value problem has the solution

23 4 9 2782 −5/9
y= x + x2 − x .
41 23 943

11. Let (x, y) be a point on the curve. The tangent line at (x, y) must pass
through (0, 2x2 ), and so has slope

y − 2x2
y0 = .
x
This is the linear differential equation

1
y0 − y = −2x.
x
An integrating factor is
R 1
e− (1/x) dx
= e− ln(x) = eln(1/x) = ,
x

so multiply the differential equation by 1/x to get

1 0 1
y − 2 y = −2.
x x
This is
 0
1
y = −2.
x
Integrate to get
1
y = −2x + c.
x
Then
y = −2x2 + cx,

in which c can be any number.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 18 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

12. Let A(t) be the number of kilograms of salt in the tank at time t ≥ 0.
Then
dA
= rate salt is added − rate salt is removed
dt  
A(t)
=3−2 .
50 + t
We must solve this subject to the initial condition A(0) = 12.5. The
differential equation is
2
A0 + A = 3,
50 + t
which is linear with integrating factor
R
2/(50+t) dt
e = e2 ln(50+t) = (50 + t)2 .

Multiply the differential equation by (50 + t)2 to get

(50 + t)2 A0 + 2(50 + t)A = 3(50 + t)2 .

This is
0
(50 + t)2 A = 3(50 + t)2 .
Integrate this equation to get

(50 + t)2 A = (50 + t)3 + c,

which we will write as

c
A(t) = (50 + t) + .
(50 + t)2
We need c so that
c
A(0) = 50 +
= 12.5,
2500
so c = −93, 750. The number of kilograms of salt in the tank at time t is
93, 750
A(t) = (50 + t) − .
(50 + t)2

13. Let A1 (t) and A2 (t) be the number of kilograms of salt in tanks 1 and 2,
respectively, at time t. Then
5 20
A01 (t) = − A1 (t); A1 (0) = 10
4 400
and
20 20
A02 (t) = A1 (t) − A2 (t); A2 (0) = 45.
400 600
Solve the linear initial value problem for A1 (t) to get

A1 (t) = 25 − 15e−t/20 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.3. EXACT EQUATIONS 19

Substitute this into the differential equation for A2 (t) to get

1 5 3
A02 + A2 = − e−t/20 ; A2 (0) = 45.
30 4 4
Solve this linear problem to obtain
75 75
A2 (t) = + 45e−t/20 − e−t/30 .
2 2
Tank 2 has its minimum when A02 (t) = 0, and this occurs when
5 −t/30 9 −t/20
e − e = 0.
4 4
Then et/60 = 9/5, or t = 60 ln(9/5). This minimum value is
2725
A2 (t)min = A2 (60 ln(9/5)) =
81
pounds.

1.3 Exact Equations

In these problems it is assumed that the differential equation has the form
M (x, y) + N (x, y)y 0 = 0, or, in differential form, M (x, y) dx + N (x, y) dy = 0.

1. With M (x, y) = 2y 2 + yexy and N (x, y) = 4xy + xexy + 2y. Then

∂N ∂M
= 4y + exy + xyexy =
∂x ∂y

for all (x, y), so the differential equation is exact on the entire plane. A
potential function ϕ(x, y) must satisfy

∂ϕ
= M (x, y) = 2y 2 + yexy
∂x
and
∂ϕ
= N (x, y) = 4xy + xexy + 2y.
∂y
Choose one to integrate. If we begin with ∂ϕ/∂x = M , then integrate
with respect to x to get

ϕ(x, y) = 2xy 2 + exy + α(y),

with α(y) the “constant” of integration with respect to x. Then we must

have
∂ϕ
= 4xy + xexy + α0 (y) = 4xy + xexy + 2y.
∂y

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 20 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

This requires that α0 (y) = 2y, so we can choose α(y) = y 2 to obtain the
potential function
ϕ(x, y) = 2xy 2 + exy + y 2 .
The general solution is defined implicitly by the equation

2xy 2 + exy + y 2 = c, ,

with c an arbitrary constant.

2. ∂M/∂y = 4x = ∂N/∂x for all (x, y), so this equation is exact on the entire
plane. For a potential function, we can begin by integrating

∂ϕ
= 2x2 + 3y 2
∂y
to get
ϕ(x, y) = 2x2 y + y 3 + c(x).
Then
∂ϕ
= 4xy + 2x = 4xy + c0 (x).
∂x
Then c0 (x) = 2x so we can choose c(x) = x2 to obtain the potential
function
ϕ(x, y) = 2x2 y + y 3 + x2 .
The general solution is defined implicitly by

2x2 y + y 3 + x2 = k,

with k an arbitrary constant.

3. ∂M/∂y = 4x + 2x2 and ∂N/∂x = 4x, so this equation is not exact (on
any rectangle).

4.
∂M ∂N
= −2 sin(x + y) + 2x cos(x + y) = ,
∂y ∂x
for all (x, y), so this equation is exact on the entire plane. Integrate
∂ϕ/∂x = M or ∂ϕ/∂y = N to obtain the potential function

ϕ(x, y) = 2x cos(x + y).

The general solution is defined implicitly by

2x cos(x + y) = k

with k an arbitrary constant.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.3. EXACT EQUATIONS 21

5. ∂M/∂y = 1 = ∂N/∂x, for x 6= 0, so this equation is exact on the plane

except at points (0, y). Integrate ∂ϕ/∂x = M or ∂ϕ/∂y = N to find the
potential function
ϕ(x, y) = ln |x| + xy + y 3
for x 6= 0. The general solution is defined by an equation

ln |x| + xy + y 3 = k.

6. For the equation to be exact, we need

∂M ∂N
= αxy α−1 = = −2xy α−1 .
∂y ∂x
This will hold if α = −2. With this choice of α, the (exact) equation is

3x2 + xy −2 − x2 y −3 y 0 = 0.

Routine integrations produce a potential function

x2
ϕ(x, y) = x3 + .
2y 2
The general solution is defined by the equation

x2
x3 + = k,
2y 2
for y 6= 0.
7. For this equation to be exact, we need
∂M ∂N
= 6xy 2 − 3 = = −3 − 2αxy 2 .
∂y ∂x
This will be true if α = −3. By integrating, we find a potential function

ϕ(x, y) = x2 y 3 − 3xy − 3y 2

and a general solution is defined implicitly by

x2 y 3 − 3xy − 3y 2 = k.

8. We have
∂M ∂N
= 2 − 2y sec2 (xy 2 ) − 2xy 3 sec2 (xy 2 ) tan(xy 2 ) = ,
∂y ∂x
for all (x, y), so this equation is exact over the entire plane. By integrating
∂ϕ/∂x = 2y − y 2 sec2 (xy 2 ) with respect to x, we find that

ϕ(x, y) = 2xy − tan(xy 2 ) + c(y).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

Then
∂ϕ
= 2x − 2xy sec2 (xy 2 )
∂y
= 2x − 2xy sec2 (xy 2 ) + c0 (y).

Then c0 (y) = 0 and we can choose c(y) = 0 to obtain the potential function

ϕ(x, y) = 2xy − tan(xy 2 ).

A general solution is defined implicitly by

2xy − tan(xy 2 ) = k.

For the solution satisfying y(1) = 2, put x = 1 and y = 2 into this

implicitly defined solution to get

4 − tan(4) = k.

The solution of the initial value problem is defined implicitly by

2xy − tan(xy 2 ) = 4 − tan(4).

9. Because ∂M/∂y = 12y 2 = ∂N/∂x, this equation is exact for all (x, y).
Straightforward integrations yield the potential function

ϕ(x, y) = 3xy 4 − x.

A general solution is defined implicitly by

3xy 4 − x = k.

To satisfy the condition y(1) = 2, we must choose k so that

48 − 1 = k,

so k = 47 and the solution of the initial value problem is specified by the

equation
3xy 4 − x = 47.
In this case we can actually write this solution explicitly with y in terms
of x.

10. First,
∂M 1 1 y
= ey/x − ey/x − 2 ey/x
∂y x x x
y y/x ∂N
= − 2e = ,
x ∂x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.3. EXACT EQUATIONS 23

so the equation is exact for all (x, y) with x 6= 0. For a potential function,
we can begin with
∂ϕ
= ey/x
∂y
and integrate with respect to y to get
ϕ(x, y) = xey/x + c(x).
Then we need
∂ϕ y y
= 1 + ey/x − ey/x = ey/x − ey/x + c0 (x).
∂x x x
This requires that c0 (x) = 1 and we can choose c(x) = x. Then
ϕ(x, y) = xey/x + x.
The general solution of the differential equation is implicitly defined by
xey/x + x = k.
To have y(1) = −5, we must choose k so that
e−5 + 1 = k.
The solution of the initial value problem is given by
xey/x + x = 1 + e−5 .
This can be solve for y to obtain the explicit solution
1 + e−5
 
y = x ln
x+1
for x + 1 > 0.
11. First,
∂M ∂N
= −2x sin(2y − x) − 2 cos(2y − x) = ,
∂y ∂x
so the differential equation is exact for all (x, y). For a potential function,
integrate
∂ϕ
= −2x cos(2y − x)
∂y
with respect to y to get
ϕ(x, y) = −x sin(2y − x) + c(x).
Then we must have
∂ϕ
= x cos(2y − x) − sin(2y − x)
∂x
= x cos(2y − x) − sin(2y − x) + c0 (x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 24 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

Then c0 (x) = 0 and we can take c(x) to be any constant. Choosing c(x) = 0
yields
ϕ(x, y) = −x sin(2y − x).
The general solution is defined implicitly by
−x sin(2y − x) = k.
To satisfy y(π/12) = π/8, we need
π
− sin(π/6) = k,
12
so choose k = −π/24 to obtain the solution defined by
π
−x sin(2y − x) = −
24
which of course is the same as
π
x sin(2y − x) = .
24
We can also write
1  π 
y= x + arcsin
2 24x
for x 6= 0.
12.
∂M ∂N
= ey =
∂y ∂x
so the differential equation is exact. Integrate
∂ϕ
= ey
∂x
with respect to x to get
ϕ(x, y) = xey + c(y).
Then
∂ϕ
= xey + c0 (y) = xey − 1,
∂y
so c0 (y) = −1 and we can let c(y) = −y. This gives us the potential
function
ϕ(x, y) = xey − y.
The general solution is given by
xey − y = k.
For y(5) = 0 we need
5−0=k
so k = 5 and the solution of the initial value problem is given by
xey − y = 5.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.3. EXACT EQUATIONS 25

13. ϕ + c is also a potential function if ϕ is because

∂ϕ ∂(ϕ + c)
=
∂x ∂x
and
∂ϕ ∂(ϕ + c)
= .
∂y ∂y
The function defined implicitly by

ϕ(x, y) = k

is the same as that defined by

ϕ(x, y) + c = k

if k is arbitrary.
14. (a)
∂M ∂N
= 1 and = −1
∂y ∂x
so this equation is not exact over any rectangle in the plane.
(b) Multiply the differential equation by x−2 to obtain

yx−2 − x−1 y 0 = 0.

This is exact because

∂M ∗ ∂N ∗
= x−2 = .
∂y ∂x
This new equation has potential function ϕ(x, y) = −yx−1 and so has
general solution defined implicitly by
y
− = k.
x
This also defines a general solution of the original differential equation.
(c) Multiply the differential equation by y −2 to obtain

y −1 − xy −2 y 0 = 0.

This is exact on any region of the plane not containing y = 0, because

∂M ∗∗ ∂N ∗∗
= −y −2 = .
∂y ∂x

The new equation has potential function ϕ(x, y) = xy −1 , so its general

solution is defined implicitly by

xy −1 = k.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 26 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

It is easy to check that this also defines a solution of the original differential
equation.
(d) Multiply the differential equation by xy −2 to obtain

xy −2 − x2 y −3 y 0 = 0.

This is exact (on any region not containing y = 0) because

∂M ∗∗∗ ∂N ∗∗∗
= −2xy −3 = .
∂y ∂x

Integrate ∂ϕ/∂x = xy −2 with respect to x to obtain

1 2 −2
ϕ(x, y) = x y + c(y).
2
Then
∂ϕ
= −x−2 y −3 + c0 (y) = −x2 y −3 ,
∂y
so c0 = 0 and we can choose c(y) = 0. Then
1 2 −2
ϕ(x, y) = x y
2
and we can define a general solution of this differential equation as

x2 y −2 = k.

Here we absorbed the factor of 1/2 into the arbitrary constant c. This
again defines a solution of the original differential equation.
(e) The original differential equation can be written as the linear equation
1
y0 − y = 0.
x
This has integrating factor
−1
R
−(1/x) dx
e = e− ln(x) = eln(x )
= x−1 .

Multiply the differential equation by x−1 to write this equation as

(x−1 y)0 = 0,

so x−1 y = c implicitly defines the general solution.

(f) The methods of (b) through (e) yield the same general solution. For
example, in (b) we obtained −yx−1 = c, which we can write as y = −cx.
Because c is an arbitrary constant, this general solution can be written
y = kx. And in (d) we obtained x2 y −2 = c, and this gives the same
solutions as y 2 = cx2 , or y = kx.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.3. EXACT EQUATIONS 27

15. First,
∂M 3 ∂N
= x − y −5/2 and = 2x.
∂y 2 ∂x
and these are not equal on any rectangle in the plane.
In differential form, the differential equation is

(xy + y −3/2 ) dx + x2 dy = 0.

Multiply this equation by xa y b to get

(xa+1 y b+1 + xa y b−3/2 ) dx + xa+2 y b dy = 0 = M ∗ dx + N ∗ dy.

For this to be exact, we need

∂M ∗
 
a+1 b 3
= (b + 1)x y + (b − xa y b−5/2
∂y 2
∂N ∗
= = (a + 2)xa+1 y b .
∂x
Divide this equation by xa y b to get
 
3
(b + 1)x + b − y −5/2 = (a + 2)x.
2

This will hold for all x and y if we let b = 3/2 and then choose a and b so
that b + 1 = a + 2. Thus choose
1 3
a= and b =
2 2

to get the integrating factor µ(x, y) = x1/2 b3/2 . Multiply the original
differential equation by this to get

(x3/2 y 5/2 + x1/2 ) dx + x5/2 y 1/2 dy = 0.

To find a potential function, integrate

∂ϕ
= x5/2 y 3/2
∂y

with respect to y to get

2 5/2 5/2
ϕ(x, y) = x y + c(x).
5
Then we need
∂ϕ
= x3/2 y 5/2 + c0 (x) = x3/2 y 5/2 + x1/2 .
∂x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 28 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

Therefore c0 (x) = x1/2 , so c(x) = 2x3/2 /3 and

2 5/2 5/2 2 3/2
ϕ(x) =x y + x .
5 3
The general solution of the original differential equation is given implicitly
by
2 2
(xy)5/2 + x3/2 = k.
5 3
In this we must have x 6= 0 and y 6= 0 to ensure that the integrating factor
µ(x, y) 6= 0.
16. It is routine to verify that the differential equation is not exact. To find
an integrating factor, first multiply by xa y b to get
(2xa y b+2 − 9xa+1 y b+1 ) dx + (3xa+1 y b+1 − 6xa+2 y b ) dy = 0.
For this to be exact, we must have
∂M
= 2(b + 2)xa y b+1 − 9(b + 1)xa+1 y b
∂y
∂N
= = 3(a + 1)xa y b+1 − 6(a + 2)x+1 y b .
∂x
Divide by xa y b and rearrange terms to obtain
(2(b + 2) − 3(a + 1))y = (9(b + 1) − 6(a + 2))x.
Because x and y are independent, both coefficients must be zero:
2(b + 2) − 3(a + 1) = 0 and 9(b + 1) − 6(a + 2) = 0.
Solve these to get a = b = 1, so µ(x, y) = xy is an integrating factor.
Multiply the differential equation by xy to obtain, in differential form,
(2xy 3 − 9x2 y 2 ) dx + (3x2 y 2 − 6x3 y) dy = 0.
This equation is exact. For a potential function, integrate
∂ϕ
= 2xy 3 − 9x2 y 2
∂x
with respect to x to get
ϕ(x, y) = x2 y 3 − 3x3 y 2 + c(y).
Then
∂ϕ
= 3x2 y 2 − 6x3 y + c0 (y) = 3x2 y 2 − 6x3 y.
∂y
Then c0 (y) = 0 and we can choose c(y) = 0 got s potential function
ϕ(x, y) = x2 y 3 − 3x3 y 2 .
A general solution of this equation, and also the original equation, is given
by
x2 y 3 − 3x3 y 2 = k.
This requires that µ(x, y) 6= 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.4. HOMOGENEOUS, BERNOULLI AND RICCATI EQUATIONS 29

1.4 Homogeneous, Bernoulli and Riccati Equa-

tions
1. This is a Riccati equation and one solution (by inspection) is S(x) = x.
Let y = x + 1/z to obtain
 2  
1 0 1 1 1 1
2− 2
z = 2 x+ − x+ + 1.
z x z x z
This simplifies to
1 1
z = − 2,z0 +
x x
a linear equation with integrating factor
R
(1/x) dx
e = eln(x) = x.

The differential equation for z can therefore be written

1
(xz)0 = − .
x
Integrate to get
xz = − ln(x) + c,
so
ln(x) c c − ln(x)
z=− + = .
x x x
for x > 0. Then
1 x
y =x+ =x+
z c − ln(x)
for x > 0.
2. This is a Bernoulli equation with α = −4/3. Put v = y 7/3 , so y = v 7/3 .
Substitute this into the differential equation to get
3 −4/7 0 7 3/7 14
v v + v = 2.
7 3x 3x
This simplifies to the linear differential equation
7 14
v0 + v= 2
3x 3x
which has integrating factor
7/3
R
7/3x dx
e = e(7/3) ln(x) = eln(x )
= x7/3

for x > 0. Multiply the differential equation by x7/3 to get

14 1/3
(x7/3 v)0 = x .
3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 30 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

Integrate to get
7 4/3
vx7/3 = x + c.
2
Because v = y 7/3 , this gives us

2y 7/3 x7/3 − 7x4/3 = k,

in which k = 2c is an arbitrary constant. This equation implicitly defines

the general solution.
3. This is a Bernoulli equation with α = 2, so let v = y 1−α = y −1 for y 6= 0
and y = 1/v. Compute
dy dv 1
y0 = = − 2 xv 0 .
dv dx v
The differential equation becomes
1 0 x x
− 2
v + = 2.
v v v
This is
v 0 − xv = −x,
2
a linear equation with integrating factor e−x /2
. We can therefore write
2 2
(e−x /2
v)0 = −xe−x /2
.

Integrate to get
2 2
e−x /2
v = e−x /2
+ c,
so 2
v = 1 + ce−x /2
.
The original differential equation has the general solution
1 1
y= = ,
v 1 + ce−x2 /2
in which c is an arbitrary constant.
4. This equation is homogeneous. With y = ux we obtain
1
u + xu0 = u + .
u
Then
du 1
= , x
dx u
a separable equation. In differential form, this is
1
u du = dx.
x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.4. HOMOGENEOUS, BERNOULLI AND RICCATI EQUATIONS 31

Integrate to get
1 2
u = ln |x| + c.
2
Then
1 2
x = 2 ln |x| + k,
y2
where k = 2c is an arbitrary constant. This implicitly defines the general
solution.
5. This differential equation is homogeneous and setting y = ux gives us
u
u + xu0 = .
1+u
This is the separable equation
du u
x = −u
dx 1+u
which, in terms of x and y, is
 
1 1 1
+ du = − dx.
u2 u x
Integrate to get
1
+ ln |u| = − ln |x| + c.
u
With u = y/x this reduces to

−x + y ln |y| = cy,

with c an arbitrary constant.

6. This is a Riccati equation and one solution (by inspection) is S(x) = 4.
After some routine computation we obtain the general solution
6x3
y =4+ .
c − x3

7. The differential equation is exact, with general solution defined implicitly

by
xy − x2 − y 2 = c.

8. The differential equation is homogeneous, and y = ux yields the general

solution defined by y y
sec + tan = cx.
x x
9. The differential equation is of Bernoulli type with α = −3/4. The general
solution is defined by

5(xy)7/4 + 7x−5/4 = c.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 32 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

10. The differential equation is homogeneous and y = ux leads to the separable

differential equation
1 − u + u2 1
du = dx.
x
Integrate and set u = y/x to obtain the general solution implicitly defined
by  
2 2y − x
√ arctan √ = ln |x| + c.
3 3x

11. The equation is Bernoulli with α = 2 and the change of variables v = y −1

leads to the general solution
2
y =2+ .
cx2 −1

12. The equation is homogeneous and y = ux leads to the general solution

defined by
1 x2
= ln |x| + c.
2 y2

13. The differential equation is Riccati and one solution is S(x) = ex . A

general solution is given explicitly by
2ex
y= .
ce2x − 1

14. The equation is Bernoulli with α = 2 and a general solution is given by

2
y= .
3 + cx2

15. For the first part,

   
ax + by + c a + b(y/x)c/x y
F =F =f
dx + py + r d + p(y/x) + r/x x

if and only if c = r = 0.
Next, suppose x = X + h and y = Y + k. Then
 
dY a(X + h) + b(Y + k) + c
=F
dX d(x + h) + p(Y + k) + r
 
aX + bY + c + ah + bk + c
=F .
dX + pY + r + dh + pk + r

This equation is homogeneous exactly when h and k can be chosen so that

ah + bk = −c and dh + pk = −r.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.4. HOMOGENEOUS, BERNOULLI AND RICCATI EQUATIONS 33

This 2 × 2 system of algebraic equations has a solution exactly when the

determinant of the coefficients is nonzero, and this is the condition that
a b
= ap − bd 6= 0.
d p

16. Comparing this with problem 15, we have

a = 0, b = 1, c = −3, d = p = 1 and r = −1.

The system to solve for h and k is

k = 3, h + k = 1.

Then k = 3 and h = −2. Let X = x − 2, Y = y + 3 to obtain

dY Y
= .
dX X +Y
This is a homogeneous equation solved in problem 5. The general of the
current problem is defined by

(y − 3) ln |y − 3| − (x + 2) = c(y − 3),

with c an arbitrary constant.

17. Let X = x − 2, Y = y + 3 to get the homogeneous equation
dY 3X − Y
= .
dX X +Y
The general solution of the original equation (in terms of x and y) is
defined by
3(x − 2)2 − 2(x − 2)(y + 3) − (y + 3)2 = c,
with c an arbitrary constant.
18. Set X = x + 5, Y = y + 1 to obtain the implicitly defined general solution

(x + 5)2 + 4(x + 5)(y + 1) − (y + 1)2 = c.

19. Let X = x − 2, Y = y + 1 to obtain the general solution given by

(2x + y − 3)2 = c(y − x + 3).

20. Suppose at time t = 0 the dog is at the origin of an x, y− coordinate

system, and the person is at (A, 0). The person moves directly upward
and at time t is at (A, vt), while the dog is at (x, y) and runs toward the
person at a speed 2v. The tangent to the dog’s path joins these two points,
and so has slope
vt − y
y0 = .
A−x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 34 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

To find the equation of the dog’s path, we will first eliminate t from this
equation. In the time the person has moved vt units upward, the dog has
run 2vt units along its path of motion, so
Z x  2 !1/2
dy
2vt = 1+ dξ.
0 dξ

Then 2 !
Z x 
dy 1 dy
vt = y + (A − x) = 1+ dξ.
dx 2 0 dξ
Then 2 !
Z x 
0 dy
2(A − x)y = 1+ dξ − 2y.
0 dξ
Differentiate this equation to get
 2 !
00 0 dy
2(A − x)y − 2y = 1+ − 2y 0 ,
dx

so
1/2
2(A − x)y 00 = 1 + (y 0 )2 ,
0
together with the conditions y(0) = y (0) = 0.
Now let u = y 0 and rewrite the resulting equation to get
1 1
2 1/2
du = dx.
(1 + u ) 2(A − x)
This has the general solution
p 1
ln(u + 1 + u2 ) = − ln(A − x) + c.
2
Use the condition that y 0 (0) = u(0) = 0 to obtain
 1/2
p A
u+ 1+ u2 = .
A−x
In terms of y, we now have
√
0
p A
y + 1 + (y 0 )2 = √ ; y(0) = 0.
A−x
But p
1 + (y 0 )2 = 2(A − x)y 00 ,
so √
A
y 0 + 2(A − x)y 00 = √ .
A−x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.4. HOMOGENEOUS, BERNOULLI AND RICCATI EQUATIONS 35

Let w = y 0 to get
√
0 1 A
w + w= .
2(A − x) 2(A − x)3/2
√
This linear differential equation has integrating factor 1/ A − x, so
 0
w A
√ = .
A−x 2(A − x)2
Integrate this to get
√
A 1 √
w= √ + c A − x.
2 A−x
Use the fact that w(0) = 0 to get
√
A 1 1 √
w= √ − √ A − x = y0 .
2 A−x 2 A
Integrate this to get
√ √ 1
y = − A A − x + √ (A − x)3/2 + c.
3 A
Because y(0) = 0,
√ √ 1 2
y = − A A − x + √ (A − x)3/2 + A.
3 A 3
Now the dog catches the person at x = A, so they meet at (A, 2A/3). This
is also the point (A, vt), so vt = 2A/3 and they meet at time
2A
t= .
3v

21. It is convenient to use polar coordinates to formulate a model for this

problem. Put the origin at the submarine at the time of sighting, and
the polar axis the line from there to the destroyer at this time (the point
(9, 0)). Initially the destroyer should steam at speed 2v directly toward the
origin, until it reaches (3, 0). During this time the submarine, moving at
speed v, will have moved three units from the point where it was sighted.
Let θ = ϕ be the ray (half-line) along which the submarine is moving.
Upon reaching (3, 0), the destroyer should execute a search pattern along
a path r = f (θ). The object is to choose this path so that the sub and
the destroyer both reach (f (ϕ, ϕ) at the same time T after the sighting.
From sighting to interception, the destroyer travels a distance
Z ϕp
6+ (f (θ))2 + (f 0 (θ))2 dθ,
0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 36 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

so  Z ϕ 
1 p
T = 6+ (f (θ))2 + (f 0 (θ))2 dθ .
2v 0

For the submarine,

1
T =
f (ϕ).
v
Equate these two expressions for T and differentiate with respect to ϕ to
get
1p
(f (ϕ))2 + (f 0 (ϕ))2 = f 0 (ϕ).
2
Denote the variable as θ and rearrange the last equation to obtain

f 0 (θ) 1
= ±√ .
f (θ) 3
The positive sign here indicates that the destroyer should execute a star-
board (left) turn, while the negative sign is for a portside turn. Taking
the positive sign, solve for f (θ) to get
√
f (θ) = keθ/ 3
.

Now f (0) = k = 3, so the path of the destroyer is part of the graph of

√
f (θ) = 3seθ/ 3
.

After sailing directly to (3, 0), the destroyer should execute √this spiral
pattern. A similar conclusion follows if the negative sign of 1/ 3 is used.
This shows that the destroyer can carry out a maneuver that will take it
directly over the submarine at some time. However, there is no way to
solve for the specific time, so it is unknown when this will occur.

22. (a) Observe that each bug follows the same curve of pursuit relative to the
center from which it starts. Place a polar coordinate system as suggested √
and determine the pursuit curve for the bug starting at θ = 0, r = a/ 2.
At t > 0, the bug will be at (f (θ), θ) and its target is at (f (θ), θ + π/2).
Show that
dy dy/dθ f 0 (θ) sin(θ) + f (θ) cos(θ)
= = 0 .
dx dx/dθ f (θ)cos(θ) − f (θ) sin(θ)
At the same time, the direction of the tangent must be from the position
(f (θ), θ) to the target location (f (θ), θ + π/2), so we also have

dy f (θ) sin(θ + π/2) − f (θ) sin(θ)

=
dx f (θ) cos(θ + π/2) − f (θ) cos(θ)
cos(θ) − sin(θ) sin(θ) − cos(θ)
= = .
− sin(θ) − cos(θ) sin(θ) + cos(θ)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 1.4. HOMOGENEOUS, BERNOULLI AND RICCATI EQUATIONS 37

Equate these two expressions for dy/dx and rearrange terms to get

f 0 (θ) + f (θ) = 0.
√
Further, f (0) = a/ 2. This is a separable, and also linear, differential
equation, and the initial value problem has the solution
a
r = f (θ) = √ e−θ .
2
This is the pursuit curve (in polar coordinates).
(b) The distance traveled is
Z ∞p
r2 + (r0 )2 dθ
0
Z ∞
" 2  2 #1/2
a a
= √ e−θ + − √ e−θ dθ
0 2 2
Z ∞
=a e−θ dθ = a.
0

√
(c) Because r = ae−θ / 2 > 0 for all θ, no bug actually catches its quarry.
The actual distance between pursuer and quarry is ae−θ .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 38 CHAPTER 1. FIRST-ORDER DIFFERENTIAL EQUATIONS

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 2

Second-Order Differential
Equations

2.1 The Linear Second-Order Equation

1. It is a routine exercise in differentiation to show that y1 (x) and y2 (x) are
solutions of the homogeneous equation, while yp (x) is a solution of the
nonhomogeneous equation. The Wronskian of y1 (x) and y2 (x) is
sin(6x) cos(6x)
W (x) = = −6 sin2 (x) − 6 sin2 (x) = −6,
6 cos(6x) −6 sin(6x)
and this is nonzero for all x, so these solutions are linearly independent
on the real line. The general solution of the nonhomogeneous differential
equation is
1
y = c1 sin(6x) + c2 cos(6x) + (x − 1).
36
For the initial value problem, we need
1
y(0) = c2 − = −5
36
so c2 = −179/36. And
1
y 0 (0) = 2 = 6c1 +
36
so c1 = 71/216. The unique solution of the initial value problem is
71 179 1
y(x) = sin(6x) − cos(6x) + (x − 1).
216 36 36

2. The Wronskian of e4x and e−4x is

e4x e−4x
W (x) = = −8 6= 0
4e4x −4e−4x

39

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 40 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

so these solutions of the associated homogeneous equation are indepen-

dent. With the particular solution yp (x) of the nonhomogeneous equation,
this equation has general solution
1 1
y(x) = c1 e4x + c2 e−4x − x2 − .
4 32
From the initial conditions we obtain
1
y(0) = c1 + c2 − = 12
32
and
y 0 (0) = 4c1 − 4c2 = 3.
Solve these to obtain c1 = 409/64 and c2 = 361/64 to obtain the solution

409 4x 361 −4x 1 2 1

y(x) = e + e − x − .
64 64 4 32

3. The associated homogeneous equation has solutions e−2x and e−x . Their
Wronskian is
e−2x e−x
W (x) = −2x = e−3x
−2e −e−x
and this is nonzero for all x. The general solution of the nonhomogeneous
differential equation is
15
y(x) = c1 e−2x + c2 e−x + .
2
For the initial value problem, solve
15
y(0) = −3 = c1 + c2 +
2
and
y 0 (0) = −1 = −2c1 − c2
to get c1 = 23/2, c2 = −22. The initial value problem has solution

23 −2x 15
y(x) = e − 22e−x + .
2 2

4. The associated homogeneous equation has solutions

y1 (x) = e3x cos(2x), y2 (x) = e3x sin(2x).

The Wronskian of these solutions is

e3x cos(2x) e3x sin(2x)
W (x) = 3x = e6x 6= 0
3e cos(2x) − 2e3x sin(2x) 3e 3x
sin(2x) + 2e3x cos(2x)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.1. THE LINEAR SECOND-ORDER EQUATION 41

for all x. The general solution of the nonhomogeneous equation is

1
y(x) = c1 e3x cos(2x) + c2 e3x sin(2x) − ex .
8
To satisfy the initial conditions, it is required that
1
y(0) = −1 = c1 −
8
and
1
3c1 + 2c2 −
= 1.
8
Solve these to obtain c1 = −7/8 and c2 = 15/8. The solution of the initial
value problem is
7 15 1
y(x) = − e3x cos(2x) + e3x sin(2x) − ex .
8 8 8

5. The associated homogeneous equation has solutions

y1 (x) = ex cos(x), y2 (x) = ex sin(x).

These have Wronskian

ex cos(x) ex sin(x)
W (x) = = e2x 6= 0
ex cos(x) − ex sin(x) ex sin(x) + ex cos(x)

so these solutions are independent. The general solution of the nonhomo-

geneous differential equation is
5 5
y(x) = c1 ex cos(x) + c2 ex sin(x) − c2 − 5x − .
2 2
We need
5
y(0) = c1 − =6
2
and
y 0 (0) = 1 = c1 + c2 − 5.
Solve these to get c1 = 17/2 and c2 = −5/2 to get the solution
17 x 5 5 5
y(x) = e cos(x) − ex sin(x) − x2 − 5x − .
2 2 2 2

6. Suppose y1 and y2 are solutions of the homogeneous equation (2.2). Then

y100 + py10 + qy1 = 0

and
y200 + py20 + qy2 = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 42 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

Multiply the first equation by y2 and the second by −y1 and add the
resulting equations to obtain

y100 y2 − y200 y1 + p(y10 y2 − y20 y1 ) = 0.

We want to relate this equation to the Wronskian of these solutions, which

is
W = y1 y20 − y2 y10 .
Now
W 0 = y1 y200 − y2 y100 .
Then
W 0 + pW = 0.
This is a linear first-order differential equation for W . Multiply this equa-
tion by the integrating factor
R
p(x) dx
e

to obtain R R
p(x) dx p(x) dx
We + pW e = 0,
which we can write as  0
R
p(x) dx
We = 0.

Integrate this to obtain R

p(x) dx
We = k,
with k constant. Then
R
W (x) = ke− p(x) dx
.

This shows that W (x) = 0 for all x (if k = 0), and W (x) 6= 0 for all x (if
k 6= 0).
Now suppose that y1 and y2 are independent and observe that

y1 y20 − y2 y10
 
d y2 1
= = 2 W (x).
dx y1 y12 y1

If k = 0, then W (x) = 0 for all x and the quotient y2 /y1 has zero derivative
and so is constant:
y2
=c
y1
for some constant c. But then y2 (x) = cy1 (x), contradicting the assump-
tion that these solutions are linearly independent. Therefore k 6= 0 and
W (x) 6= − for all x, as was to be shown.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.2. THE CONSTANT COEFFICIENT HOMOGENEOUS EQUATION 43

7. The Wronskian of x2 and x3 is

x2 x3
W (x) = = x4 .
2x 3x2

Then W (0) = 0, while W (x) 6= 0 if x 6= 0. This is impossible if x2 and

x3 are solutions of equation (2.2) for some functions p(x) and q(x). We
conclude that these functions are not solutions of equation (2.2).
8. It is routine to verify that y1 (x) and y2 (x) are solutions of the differential
equation. Compute
x x2
W (x) = = x2 .
1 2x
Then W (0) = 0 but W (x) > 0 if x 6= 0. However, to write the differential
equation in the standard form of equation (2.2), we must divide by x2 to
obtain
2 2
y 00 − y 0 + 2 y = 0.
x x
This is undefined at x = 0, which is in the interval −1 << 1, so the
theorem does not apply.
9. If y2 (x) and y2 (x) both have a relative extremum (max or min) at some
x0 within (a, b), then
y 0 (x0 ) = y20 (x0 ) = 0.
But then the Wronskian of these functions vanishes at 0, and these solu-
tions must be independent.
10. By assumption, ϕ(x) is the unique solution of the initial value problem

y 00 + py 0 + qy = 0; y(x0 ) = 0.

But the function that is identically zero on I is also a solution of this initial
value problem. Therefore these solutions are the same, and ϕ(x) = 0 for
all x in I.
11. If y1 (x0 ) = y2 (x0 ) = 0, then the Wronskian of y1 (x) and y2 (x) is zero at
x0 , and these two functions must be linearly dependent.

2.2 The Constant Coefficient Homogeneous Equa-

tion
1. From the differential equation we read the characteristic equation

λ2 − λ − 6 = 0,

which has roots −2 and 3. The general solution is

y(x) = c1 e−2x + c2 e3x .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 44 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

2. The characteristic equation is

λ2 − 2λ + 10 = 0

with roots 1 ± 3i. We can write a general solution

y(x) = c1 ex cos(3x) + c2 ex sin(3x).

3. The characteristic equation is

λ2 + 6λ + 9 = 0

with repeated roots −3, −3. Then

y(x) = c1 e−3x + c2 xe−3x

is a general solution.

4. The characteristic equation is

λ2 − 3λ = 0

with roots 0, 3, and

y(x) = c1 + c2 e3x
is a general solution.

5. characteristic equation λ2 + 10λ + 26 = 0, with roots −5 ± i; general

solution
y(x) = c1 e−5x cos(x) + c2 e−5x sin(x).

6. characteristic equation λ2 +6λ−40 = 0, with roots 4, −10; general solution

y(x) = c1 e4x + c2 e−10x .

√
7. characteristic equation λ2 +3λ+18 = 0, with roots −3/2±3 7i/2; general
solution
√ ! √ !
−3x/2 3 7x −3x/2 3 7x
y(x) = c2 e cos + c2 e sin .
2 2

8. characteristic equation λ2 + 16λ + 64 = 0, with repeated roots −8, −8;

general solution
y(x) = e−8x (c1 + c2 x).

9. characteristic equation λ2 − 14λ + 49 = 0, with repeated roots 7, 7; general

solution
y(x) = e7x (c1 + c2 x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.2. THE CONSTANT COEFFICIENT HOMOGENEOUS EQUATION 45
√
10. characteristic equation λ2 −6λ+7 = 0, with roots 3± 2i; general solution
√ √
y(x) = c1 e3x cos( 2x) + c2 e3x sin( 2x).

In each of Problems 11–20 the solution is found by finding a general solution

of the differential equation and then using the initial conditions to find the
particular solution of the initial value problem.

11. The differential equation has characteristic equation λ2 + 3λ = 0, with

roots 0, −3. The general solution is

y(x) = c1 + c2 e−3x .

Choose c1 and c2 to satisfy:

y(0) = c1 + c2 = 3,
y 0 (0) = −3c2 = 6.

Then c2 = −2 and c1 = 5, so the unique solution of the initial value

problem is
y(x) = 5 − 2e−3x .

12. characteristic equation λ2 + 2λ − 3 = 0, with roots 1, −3; general solution

y(x) = c1 ex + c2 e−3x .

Solve
y(0) = c1 + c2 = 6, y 0 (0) = c1 − 3c2 = −2
to get c1 = 4 and c2 = 2. The solution is

y(x) = 4ex + 2e−3x .

13. The initial value problem has the solution y(x) = 0 for all x. This can
be seen by inspection or by finding the general solution of the differential
equation and then solving for the constants to satisfy the initial conditions.
14. y(x) = e2x (3 − x)
15. characteristic equation λ2 + λ − 12 = 0, with roots 3, −4. The general
solution is
y(x) = c1 e3x + c2 e−4x .
We need
y(2) = c1 e6 + c2 e−8 = 2
and
y 0 (2) = 3c1 e6 − 4c2 e−8 = −1.
Solve these to obtain
c1 = e−6 , c2 = e8 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 46 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

The solution of the initial value problem is

y(x) = e−6 e3x + e8 e−4x .
This can also be written
y(x) = e3(x−2) + e−4(x−2) .

16. √
6 x  √6x √ 
y(x) = e e − e− 6x
4
17. y(x) = ex−1 (29 − 17x)
18.
8 √ √
y(x) = √ sin( 23)e5x/2 cos( 23x/2)
e5 23
8 √ √
− √ cos( 23)e5x/2 sin( 23x/2)
5
e 23
19.
h √
y(x) = e(x+2)/2 cos( 15(x + 2)/2)
√

5
+ √ sin( 15(x + 2)/2)
15
20. √ √
y(x) = ae(−1 + 5)x/2 + be(−1− 5)x/2 ,
where
√ ! √ !
9+7 5 √ 7 5−9 √
a= √ e−2+ 5 and b = √ e−2− 5
2 5 2 5

21. (a) The characteristic equation is λ2 − 2αλ + α2 = 0, with α as a repeated

root. The general solution is
y(x) = (c1 + c2 x)eαx .

(b) The characteristic equation is λ2 − 2αλ + (α2 − 2 ) = 0, with roots

α + , α − . The general solution is
y (x) = c1 e(α+)x + c2 e(α−)x .
We can also write
y (x) = c1 ex + c2 e−x eαx .

In general,
lim y (x) = (c1 + c2 )eαx 6= y(x).
→0
Note, however, that the coefficients in the differential equations in (a) and
(b) can be made arbitrarily close by choosing  sufficiently small.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.2. THE CONSTANT COEFFICIENT HOMOGENEOUS EQUATION 47

22. With a2 = 4b, one solution is y1 (x) = e−ax/2 . Attempt a second solution
y2 (x) = u(x)e−ax/2 . Substitute this into the differential equation to get

a2
 
 a 
u − au + u + a u − u + bu e−ax/2 = 0.
00 0 0
4 2

Because a2 − 4b = 0, this reduces to

u00 (x) = 0.

Then u(x) = cx + d, with c and d arbitrary constants, and the functions

(cx + d)e−ax/2 are also solutions of the differential equation. If we choose
c = 1 and d = 0, we obtain y2 (x) = xe−ax/2 as a second solution. Further,
this solution is independent from y1 (x), because the Wronskian of these
solutions is
e−ax/2 xe−ax/2
W (x) = = e−ax ,
−(a/2)e−ax/2 e −ax/2
− (a/2)xe−ax/2

and this is nonzero.

23. The roots of the characteristic equation are
√ √
−a + a2 − 4b −a − a2 − 4b
λ1 = and λ2 = .
2 2
Because a2 − 4b < a2 by assumption, λ1 and λ2 are both negative (if
a2 − 4b ≥ 0), or complex conjugates (if a2 − 4b < 0). There are three
cases.

Case 1 - Suppose λ1 and λ2 are real and unequal. Then the general
solution is
y(x) = c1 eλ1 x + c2 eλ2 x
and this has limit zero as x → ∞ because λ1 and λ2 are negative.

Case 2 - Suppose λ1 = λ2 . Now the general solution is

y(x) = (c1 + c2 x)eλ1 x ,

and this also has limit zero as x → ∞.

Case 3 - Suppose λ1 and λ2 are complex. Now the general solution is

h p p i
y(x) = c1 cos( 4b − a2 x/2) + c2 sin( 4b − a2 x/2) e−ax/2 ,

and this has limit zero as x → ∞ because a > 0.

√
If, for example, a = 1 and b = −1, then one solution is e(−1+ 5)x/2
, and
this tends to ∞ as x → ∞.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 48 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

2.3 Particular Solutions of the Nonhomogeneous

Equation
1. Two independent solutions of y 00 + y = 0 are y1 (x) = cos(x) and y2 (x) =
sin(x), with Wronskian

cos(x) sin(x)
W (x) = = 1.
− sin(x) cos(x)

Let f (x) = tan(x) and use equations (2.7) and (2.8). First,
Z Z
y2 (x)f (x)
u1 (x) = − =− tan(x) sin(x) dx
W (x)
sin2 (x)
Z
=− dx
cos(x)
1 − cos2 (x)
Z
=− dx
cos(x)
Z Z
= cos(x) dx − sec(x) dx

= sin(x) − ln | sec(x) + tan(x)|.

Next,
Z Z
y1 (x)f (x)
u2 (x) = dx = cos(x) tan(x) dx
W (x)
Z
= sin(x) dx = − cos(x).

The general solution is

y(x) = c1 cos(x) + c2 sin(x) + u1 (x)y1 (x) + u2 (x)y2 (x)

= c1 cos(x) + c2 sin(x) − cos(x) ln | sec(x) + tan(x)|.

2. Two independent solutions of the associated homogeneous equation are

y1 (x) = e3x and y2 (x) = ex . Their Wronskian is W (x) = −2e4x . Compute

2ex cos(x + 3)
Z
u1 (x) = − dx
−2e4x
Z
= e−3x cos(x + 3) dx
3 −3x 1
=− e cos(x + 3) + e−3x sin(x + 3)
10 10

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.3. PARTICULAR SOLUTIONS OF THE NONHOMOGENEOUS EQUATION49

and
2e3x cos(x + 3)
Z
u2 (x) = dx
−2e4x
Z
= e−x cos(x + 3) dx
1 −x 1
= e cos(x + 3) − e−x cos(x + 3).
2 2
The general solution is
y(x) = c1 e3x + c2 ex
3 1
− cos(x + 3) + sin(x + 3)
10 10
1 1
+ cos(x + 3) − sin(x + 3).
2 2
More compactly, the general solution is
1 2
y(x) = c1 e3x + c2 ex + cos(x + 3) − sin(x + 3).
5 5
For Problems 3–6, some details of the calculations are omitted.
3. The associated homogeneous equation has independent solutions y1 (x) =
cos(3x) and y2 (x) = sin(3x), with Wronskian 3. The general solution is
4
y(x) = c1 cos(3x) + c2 sin(3x) + 4x sin(3x) + cos(3x) ln | cos(3x)|.
3

4. y1 (x) = e3x and y2 (x) = e−x , with W (x) = −4e−2x . With

f (x) = 2 sin2 (x) = 1 − cos(2x)
we find the general solution
1 7 4
y(x) = c1 e3x + c2 e−x − + cos(2x) + sin(2x).
3 65 65

5. y1 (x) = ex and y2 (x) = e2x , with Wronskian W (x) = e3x . With f (x) =
cos(e−x ), we find the general solution
y(x) = c1 ex + c2 e2x − e2x cos(e−x ).

6. y1 (x) = e3x and y2 (x) = e2x , with Wronskian W (x) = e−5x . Use the
identity
8 sin2 (4x) = 4 cos(8x) − 1
in determining u1 (x) and u2 (x) to write the general solution
2 58 40
y(x) = c1 e3x + c2 e2x + + cos(8x) + sin(8x).
3 1241 1241

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 50 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

In Problems 7–16 the method of undetermined coefficients is used to find

a particular solution of the nonhomogeneous equation. Details are included
for Problems 7 and 8, and solutions are outlined for the remainder of these
problems.

7. The associated homogeneous equation has independent solutions y1 (x) =

e2x and e−x . Because 2x2 +5 is a polynomial of degree 2, attempt a second
degree polynomial
yp (x) = Ax2 + Bx + C
for the nonhomogeneous equation. Substitute yp (x) into this nonhomoge-
neous equation to obtain

2A − (2Ax + B) − 2(Ax2 + Bx + C) = 2x2 + 5.

Equating coefficients of like powers of x on the left and right, we have the
equations

−2A = 2( coefficients of x2 )
−2A − 2B − 0( coefficients of x
2A − 2B − 2C = 5( constant term.)

Then A = −1, B = 1 and C = −4. Then

yp (x) = −x2 + x − 4

and a general solution of the (nonhomogeneous) equation is

y = c1 e2x + c2 e−x − x2 + x − 4.

8. y1 (x) = e3x and y2 (x) = e−2x are independent solutions of the associated
homogeneous equation. Because e2x is not a solution of the homogeneous
equation, attempt a particular solution yp (x) = Ae2x of the nonhomoge-
neous equation. Substitute this into the differential equation to get

4A − 2A − 6A = 8,

so A = −2 and a general solution is

y(x) = c1 e3x + c2 e−2x − 2e2x .

9. y1 (x) = ex cos(3x) and y2 (x) = ex sin(3x) are independent solutions of

the associated homogeneous equation. Try a particular solution yp (x) =
Ax2 + Bx + C to obtain the general solution

y(x) = c1 ex cos(3x) + c2 ex sin(3x) + 2x2 + x − 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.3. PARTICULAR SOLUTIONS OF THE NONHOMOGENEOUS EQUATION51

10. For the associated homogeneous equation, y1 (x) = e2x cos(x) and y2 (x) =
e2x sin(x). Try yp (x) = Ae2x to get A = 21 and obtain the general solution

y(x) = c1 e2x cos(x) + c2 e2x sin(x) + 21e2x .

11. For the associated homogeneous equation, y1 (x) = e2x and y2 (x) = e4x .
Because ex is not a solution of the homogeneous equation, attempt a
particular solution of the nonhomogeneous equation of the form yp (x) =
Aex . We get A = 1, so a general solution is

y(x) = c1 e2x + c2 e4x + ex .

12. y1 (x) = e−3x and y2 (x) = e−3x . Because f (x) = 9 cos(3x) (which is not
a solution of the associated homogeneous equation), attempt a particular
solution
yp (x) = A cos(3x) + B sin(3x).
This attempt includes both a sine and cosine term even though f (x) has
only a cosine term, because both terms may be needed to find a particular
solution. Substitute this into the nonhomogeneous equation to obtain
A = 0 and B = 1/2, so a general solution is
1
y(x) = (c1 + c2 x)e−3x + sin(3x).
2
In this case yp (x) contains only a sine term, although f (x) has only the
cosine term.
13. y1 (x) = ex and y2 (x) = e2x . Because f (x) = 10 sin(x), attempt

yp (x) = A cos(x) + B sin(x).

Substitute this into the (nonhomogeneous) equation to find that A = 3

and B = 1. A general solution is

y(x) = c1 ex + c2 e2x + 3 cos(x) + sin(x).

14. y1 (x) = 1 and y2 (x) = e−4x . Finding a particular solution yp (x) for this
problem requires some care. First, f (x) contains a polynomial term and
an exponential term, so we are tempted to try yp (x) as a second degree
polynomial Ax2 + Bx + C plus an exponential term De3x to account for
the exponential term in the equation. However, note that y1 (x) = 1, a
constant solution, is one term of the proposed polynomial part, so multiply
this part by x to try

yp (x) = Ax3 + Bx2 + Cx + De3x .

Substitute this into the nonhomogeneous differential equation to get

6Ax + 2B + 9De3x − 4(3Ax2 + 2Bx + C + 3De3x ) = 8x2 + 2e3x .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 52 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

Matching coefficients of like terms, we conclude that

2B − 4C = 0( from the constant terms)

6A − 8B = 0( from the x terms),
−12A = 8( from the x2 terms)
−3D = 2.

Then D = − 32 , A = − 23 , B = − 21 and C = − 41 . Then

2 1 1 2
yp (x) = − x3 − x2 − x − e3x
3 2 4 3
and a general solution of the nonhomogeneous equation is
2 1 1 2
y(x) = c1 + c2 e−4x − x3 − x2 − x − e3x .
3 2 4 3

15. y1 (x) = e2x cos(3x) and y2 (x) = e2x sin(3x). Try

yp x = Ae2x + Be3x .

This will work because neither e2x nor e3x is a solution of the associated
homogeneous equation. Substitute yp (x) into the differential equation and
obtain A = 1/3, B = −1/2. The differential equation has general solution
1 1
y(x) = [c1 cos(3x) + c2 sin(3x)]e2x + e2x − e3x .
3 2

16. y1 (x) = ex and y2 (x) = xex . Try

yp (x) = Ax + B + C cos(3x) + D sin(3x).

This leads to the general solution

3
y(x) = (c1 + c2 x)ex + 3x + 6 + cos(3x) − 2 sin(3x).
2

In Problems 17–24 the strategy is to first find a general solution of the dif-
ferential equation, then solve for the constants to find a solution satisfying the
initial conditions. Problems 17–22 are well suited to the use of undetermined co-
efficients, while Problems 23 and 24 can be solved fairly directly using variation
of parameters.

17. y1 (x) = e2x and y2 (x) = e−2x . Because e2x is a solution of the asso-
ciated homogeneous equation, use xe2x in the method of undetermined
coefficients, attempting

yp (x) = Axe2x + Bx + C.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.3. PARTICULAR SOLUTIONS OF THE NONHOMOGENEOUS EQUATION53

Substitute this into the nonhomogeneous differential equation to obtain

4Axe2x + 4Axe2x − 4Axe2x − 4Bx − 4C = −7e2x + x.

Then A = −7/4, B = −1/4 and C = 0, so the differential equation has

the general solution
7 1
y(x) = c1 e2x + c2 e−2x − xe2x − x.
4 4
We need
y(0) = c1 + c2 = 1
and
7 1
y 0 (0) = 2c1 − 2c2 − − = 3.
4 4
Then c1 = 7/4 and c2 = −3/4. The initial value problem has the unique
solution
7 3 7 1
y(x) = e2x − e−2x − xe2x − x.
4 4 4 4

18. y1 = 1 and y2 (x) = e−4x are independent solutions of the associated

homogeneous equation. For yp (x), try

yp (x) = Ax + B cos(x) + C sin(x),

with the term Ax because 1 is a solution of the homogeneous equation.

This leads to the general solution

y(x) = c1 + c2 e−4x + 2x − 2 cos(x) + 8 sin(x).

Now we need
y(0) = c1 + c2 − 2 = 3
and
y 0 (0) = −4c2 + 2 + 8 = 2.
Then c1 = 3 and c2 = 2, so the initial value problem has the solution

y(x) = 3 + 2e−4x + 2x − 2 cos(x) + 8 sin(x).

19. We find the general solution

1 7
y(x) = c1 e−2x + c2 e−6x + e−x + .
5 12
The solution of the initial value problem is
3 −2x 19 −6x 1 −x 7
y(x) = e − e + e + .
8 120 5 12

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 54 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

20. 1 and e3x are independent solutions of the associated homogeneous equa-
tion. Attempt a particular solution
yp (x) = Ae2x cos(x) + Be2x sin(x)
of the nonhomogeneous equation to find the general solution
1
y(x) = c1 + c2 e3x − e2x (cos(x) + 3 sin(x)).
5
The solution of the initial value problem is
1 1
y(x) = + e3x − (cos(x) + 3 sin(x)).
5 5
21. e4x and e−2x are independent solutions of the associated homogeneous
equation. The nonhomogeneous equation has general solution
y(x) = c1 e4x + c2 e−2x − 2e−x − e2x .
The solution of the initial value problem is
y(x) = 2e4x + 2e−2x − 2e−x − e2x .

22. The general solution of the differential equation is

" √ ! √ !#
x/2 3 3
y(x) = e c1 cos x + c2 sin x + 1.
2 2

It is easier to fit the initial conditions specified at x = 1 if we write this

general solution as
" √ ! √ !#
x/2 3 3
y(x) = e d1 cos (x − 1) + d2 sin (x − 1) + 1.
2 2

Now
√
1/2 1 1/2
0 3 1/2
y(1) = e d1 + 1 = 4 and y (1) = e d1 + e d2 = −2.
2 2
√
Solve these to obtain d1 = 3e−1/2 and d2 = −7e−1/2 / 3. The solution of
the initial value problem is
" √ ! √ !#
3 7 3
y(x) = e(x−1)/2 3 cos (x − 1) − √ sin (x − 1) + 1.
2 3 2

23. The differential equation has general solution

y(x) = c1 ex + c2 e−x − sin2 (x) − 2.
The solution of the initial value problem is
y(x) = 4e−x − sin2 (x) − 2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.4. THE EULER DIFFERENTIAL EQUATION 55

24. The general solution is

y(x) = c1 cos(x) + c2 sin(x) − cos(x) ln | sec(x) + tan(x)|,

and the solution of the initial value problem is

y(x) = 4 cos(x) + 4 sin(x) − cos(x) ln | sec(x) + tan(x)|.

2.4 The Euler Differential Equation

Details are included with solutions for Problems 1–2, while just the solutions
are given for Problems 3–10. These solutions are for x > 0.

1. Read from the differential equation that the characteristic equation is

r2 + r − 6 = 0

with roots 2, −3. The general solution is

y(x) = c1 x2 + c2 x−3 .

2. The characteristic equation is

r2 + 2r + 1 = 0

with repeated root −1, −1. The general solution is

y(x) = c1 x−1 + c2 x−1 ln(x).

We can also write

1
y(x) = (c1 + c2 ln(x))
x
for x > 0.
3.
y(x) = c1 cos(2 ln(x)) + c2 sin(2 ln(x))

4.
1
y(x) = c1 x2 + c1
x2
5.
1
y(x) = c1 x2 + c1
x4
6.
1
y(x) = (c2 cos(3 ln(x)) + c2 sin(3 ln(x)))
x2
7.
1 1
y(x) = c1 + c2 3
x2 x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 56 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

8.
y(x) = x2 (c1 cos(7 ln(x)) + c2 sin(7 ln(x)))
9.
1
y(x) = (c1 + c2 ln(x))
x12
10.
y(x) = c1 x7 + c2 x5
11. The general solution of the differential equation is
y(x) = c1 x3 + c2 x−7 .
From the initial conditions, we need
y(2) = 8c1 + 2−7 c2 = 1 and y 0 (2) = 3c1 22 − 7c2 2−8 = 0.
Solve for c1 and c2 to obtain the solution of the initial value problem
7  x 3 3  x −7
y(x) = + .
10 2 10 2
12. The initial value problem has the solution
y(x) = −3 + 2x2 .

13. y(x) = x2 (4 − 3 ln(x))

14. y(x) = −4x−12 (1 + 12 ln(x))
15. y(x) = 3x6 − 2x4
16.
11 2 17 −2
y(x) = x + x
4 4
17. With Y (t) = y(et ), use the chain rule to get
dY dt 1
y 0 (x) = = Y 0 (t)
dt dx x
and then
 
00 d 1 0
y (x) = Y (t)
dx x
1 1 d
= − 2 Y 0 (t) + (Y 0 (t))
x x dx
1 1 dY 0 dt
= − 2 Y 0 (t) +
x x dt dx
1 0 1 1 00
= − 2 Y (t) + Y (t)
x xx
1
= 2 (Y 00 (t) − Y 0 (t)).
x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.4. THE EULER DIFFERENTIAL EQUATION 57

Then
x2 y 00 (x) = Y 00 (t) − Y 0 (t).
Substitute these into Euler’s equation to get

Y 00 (t) + (A − 1)Y 0 (t) + BY (t) = 0.

This is a constant coefficient second-order homogeneous differential equa-

tion for Y (t), which we know how to solve.
18. If x < 0, let t = ln(−x) = ln |x|. We can also write x = −et . Note that

dt 1 1
= (−1) =
dx −x x
just as in the case x > 0. Now let y(x) = y(−et ) = Y (t) and proceed with
chain rule differentiations as in the solution of Problem 17. First,
dY dt 1
y 0 (x) = = Y 0 (t)
dt dx x
and
 
d 1 0
y 00 (x) = Y (t)
dx x
1 1 dt 00
= − 2 Y 0 (t) + Y (t)
x x dx
1 1
= − 2 Y 0 (t) + 2 Y 00 (t).
x x
Then
x2 y 00 (x) = Y 00 (t) − Y 0 (t)
just as we saw with x > 0. Now Euler’s equation transforms to

Y 00 + (A − 1)Y 0 + BY = 0.

We obtain the solution in all cases by solving this linear constant coefficient
second-order equation. Omitting all the details, we obtain the solution of
Euler’s equation for negative x by replacing x with |x| in the solution for
positive x. For example, suppose we want to solve

x2 y 00 + xy 0 + y = 0

for x < 0. Solve this for x > 0 to get

y(x) = c1 cos(ln(x)) + c2 sin(ln(x)).

The solution for x < 0 is

y(x) = c1 cos(ln |x|) + c2 sin(ln |x|).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 58 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

19. The problem to solve is

x2 y 00 − 5dxy 0 + 10y = 0; y(1) = 4, y 0 (1) = −6.

We know how to solve this problem. Here is an alternative method, us-

ing the transformation x = et , or t = ln(x) for x > 0 (since the initial
conditions are specified at x = 1). Euler’s equation transforms to

Y 00 − 6Y 0 + 10Y = 0.

However, also transform the initial conditions:

Y (0) = y(1) = 4, Y 0 (0) = (1)y 0 (1) = −6.

This differential equation for Y (t) has general solution

Y (t) = c1 e3t cos(t) + c2 e3t sin(t).

Now
Y (0) = c2 = 4
and
Y 0 (0) = 3c1 + c2 = −6,
so c2 = −18. The solution of the transformed initial value problem is

Y (t) = 4e3t cos(t) − 18e3t sin(t).

The original initial value problem therefore has the solution

y(x) = 4x3 cos(ln(x)) − 19x3 sin(ln(x))

for x > 0. The new twist here is that the entire initial value problem
(including initial conditions) was transformed in terms of t and solved for
Y (t), then this solution Y (t) in terms of t was transformed back to the
solution y(x) in terms of x.
20. Suppose
x2 y 00 + Axy 0 + By = 0
has repeated roots. Then the characteristic equation

r2 + (A − 1)r + B = 0

has (1 − A)/2 as a repeated root, and we have only one solution y1 (x) =
x(1−A)/2 so far. For another solution, independent from y1 , look for a
solution of the form y2 (x) = u(x)y1 (x). Then

y20 = u0 y1 + uy10

and
y200 = u00 y1 + 2u0 y10 + uy100 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.4. THE EULER DIFFERENTIAL EQUATION 59

Substitute y2 into the differential equation to get

x2 (u00 y1 + 2u0 y10 + uy100 ) + Ax(u0 y1 + uy10 ) + Buy1 = 0.

Three terms in this equation cancel, because

u(x2 y100 + Axy10 + By1 ) = 0

by virtue of y1 being a solution. This leaves

x2 u00 y1 + 2x2 u0 y10 + Axu0 y1 = 0.

Assuming that x > 0, divide by x to get

xu00 y1 + 2xu0 y10 + Au0 y1 = 0.

Substitute y1 (x) = x(1−A)/2 into this to obtain

 
00 (1−A)/2 0 1−A
xu x + 2xu x(−1−A)/2 + Au0 x(1−A)/2 = 0.
2

Divide this by x(1−A)/2 to get

xu00 + (1 − A)u0 + Au0 = 0,

and this reduces to

xu00 + u0 = 0.

Let z = u0 to obtain
xz 0 + z = 0,

or
(xz)0 = 0.

Then xz = c, constant, so
c
z = u0 = .
x
Then u(x) = c ln(x) + d. We only need one second solution, so let c = 1
and d = 0 to get u(x) = ln(x). A second solution, independent from y1 (x),
is
y2 (x) = y1 (x) ln(x),

as given without derivation in the chapter.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 60 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

2.5 Series Solutions

2.5.1 Power Series Solutions
P∞
1. Put y(x) = n=0 an xn into the differential equation to obtain
∞
X ∞
X
y 0 − xy = nan xn−1 − an xn+1
n=1 n=0
X∞ X∞
= nan xn−1 − an−2 xn−1
n=1 n=2
∞
X
= a1 + (2a2 − a0 )x + (nan − an−2 )xn−1
n=3
= 1 − x.

Then a0 is arbitrary, a1 = 1, 2a2 − a0 = −1, and

1
an = an−2 for n = 3, 4, · · · .
n
This is the recurrence relation. If we set a0 = c0 + 1, we obtain the
coefficients
1 1 1
a2 = c0 , a4 = c0 , a6 = c0 ,
2 2·4 2·4·6
and so on. Further,
1 1 1
a1 = 1, a3 = , a5 = , a7 =
3 3·5 3·5·7
and so on. The solution can be written
∞
X 1
y(x) = 1 + x2n+1
n=0
3 · 5 · · · 2n + 1
∞
!
X 1 2n
+ c0 1+ x .
n=1
2 · 4 · · · 2n

2. Write
∞
X ∞
X
y 0 − x3 y = nan xn−1 − an xn+3
n=1 n=0
∞
X
= a1 + 2a2 x + 3a3 x2 + (nan − an−4 )xn−1 = 4.
n=4

The recurrence relation is

1
an = an−4 for n = 4, 5, · · · ,
n

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.5. SERIES SOLUTIONS 61

with a0 arbitrary, a1 = 4 and a2 = a3 = 0. This yields the solution

∞
X 1
y(x) = 4 x4n+1
n=0
1 · 5 · 9 · · · (4n + 1)
∞
!
X 1 4n
+ a0 1 + x .
n=1
4 · 8 · 12 · · · 4n

3. Write
∞
X ∞
X ∞
X
y 0 + (1 − x2 )y = nan xn−1 + an xn − an xn+2
n=1 n=0 n=0
∞
X
= (a1 + a0 ) + (2a2 + a1 )x + (nan + an−1 − an−3 )xn−1
n=3
= x.

The recurrence relation is

nan + an−1 − an−3 = 0 for n = 3, 4, · · · .

Here a0 is arbitrary, a1 + a0 = 0 and 2a2 + a1 = 1. This gives us the

solution
 
1 1 7 19
y(x) = a0 1 − x + x2 + x3 − x4 + x5 + · · ·
2! 3! 4! 5!
1 2 1 3 1 4 11 5 31 6
+ x − x + x + x − x + ··· .
2! 3! 4! 5! 6!
4. Begin with
∞
X ∞
X ∞
X
y 00 + 2y 0 − xy = n(n − 1)an xn−2 + 2nan xn−1 + an xn+1
n=2 n=1 n=0
= (2a2 + 2a1 ) + (3 · 2a3 + 2 · 2a2 + a0 )x
X∞
+ (n(n − 1)an + 2(n − 1)an−1 + an−2 )xn−2 = 0.
n=1

The recurrence relation is

n(n − 1)an + 2(n − 1)an−1 + an−2 = 0 for n = 4, 5, · · · .

Further, a0 and a1 are arbitrary, a2 = −A1 and

6a3 + 4a2 + a0 = 0.

Taking a0 = 1, a1 = 0, we obtain the solution

1 1 1 1
y1 (x) = 1 − x3 + x4 − x5 + x6 + · · · .
6 12 30 60

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 62 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

With a0 = 0, a1 = 1 we get a second, linearly independent solution

2 5 7
y2 (x) = x − x2 + x3 − x4 + x5 + · · · .
3 12 60

5. Write
∞
X ∞
X ∞
X
00 0 n−2 n
y − xy + y = n(n − 1)n x − nan x + an xn
n=2 n=1 n=0
∞
X ∞
X ∞
X
= 2a2 + a0 + (n + 2)(n + 1)an+2 xn − nan xn + an xn = 3.
n=0 n=1 n=0

Here a0 and a1 are arbitrary and a2 = (3 − a0 )/2. The recurrence relation

is
n−1
an+2 = for n = 1, 2, · · · .
(n + 2)(n + 1)
This yields the general solution
3 − a0 2 3 − a0 4
y(x) = a0 + a1 x + x + x
2 4!
3(3 − a0 ) 6 3 · 5(3 − a0 ) 8
+ x + x + ··· .
6! 8!

6. Begin with
∞
X ∞
X ∞
X
y 00 + xy 0 + xy = n(n − 1)axn xn−2 + nan xn + an xn+1
n=2 n=1 n=0
∞
X
= 2a2 + (n(n − 1)an + (n − 2)an−2 + an−3 )xn−2 = 0.
n=3

Here a0 and a1 are arbitrary and a2 = 0. The recurrence relation is

(n − 2)an−2 − an−3
an = − for n = 3, 4, · · · .
n(n − 1)
With a0 = 1 and a1 = 0, we obtain one solution
2 3
y1 (x) = 1 − x3 + x5
3 2·3·4·5
1 3·5
+ x6 − x7 + · · · .
2·3·5·6 2·3·4·5·6·7
With a0 = 0 and a1 = 1, we obtain a second, linearly independent solution
1 3 1 4
y2 (x) = x − x − x
2·3 3·4
3 3·5
+ x5 + x6 + · · · .
2·3·4·5 2·3·5·6

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.5. SERIES SOLUTIONS 63

7. We have
∞
X
y 00 − x2 y 0 + 2y = n(n − 1)an xn−2
n=2
∞
X ∞
X
− nan xn+1 + 2an xn
n=1 n=0
= 2a2 + 2a0 + (6a3 + 2a1 )x
∞
X
+ (n(n − 1)an − (n − 3)an−3 + 2an−2 )xn−2 = x.
n=1

Then a0 and a1 are arbitrary, a2 = −a0 , and 6a3 +2a1 = 1. The recurrence
relation is
(n − 3)an−3 − 2an−2
an =
n(n − 1)
for n = 4, 5, · · · . The general solution has the form
 
2 1 4 1 5 1 6
y(x) = a0 1 − x + x − x − x + · · ·
6 10 90
 
1 3 1 4 1 5 7 6
+ a1 x − x + x + z − x + ···
3 12 30 180
1 1 1 1 7 1 8
+ x3 − x5 + x6 + x − x + ··· .
6 6 60 1260 480

Note that a0 = y(0) and a1 = y 0 (0). The third series represents the
solution obtained subject to y(0) = y 0 (0) = 0.

8. Using the Maclaurin expansion for cos(x), we have

∞
X ∞
X
y 0 + xy = nan xn − 1 + an xn+1
n=1 n=0
∞
X
= a1 + (2na2n + a2n−2 )x2n−1
n=0
∞
X
+ ((2n + 1)a2n+1 + a2n−1 )x2n
n=1
∞
X (−1)n 2n
= cos(x) = x .
n=0
(2n)!

a0 is arbitrary and a1 = 1. The recurrence relation is

1 −a2n−1 + (−1)n /((2n)!)

a2n = − a2n−2 and a2n+1 =
2n 2n + 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 64 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

for n = 1, 2, · · · . This yields the solution

1 4 2 · 4 · 66
 
1 2
y(x) = a0 1 − x + x − + ···
2 2·4 x
 
3 3 13 5 79 7 633 9
+ x− x + x − x + x + ··· .
3! 5! 7! 9!

9. We have
∞
X
00 0
y + (1 − x)y + 2y = n(n − 1)an xn−2
n=2
∞
X ∞
X ∞
X
+ nan xn−1 − nan xn + 2 an xn
n=1 n=1 n=0
∞
X
= (2a2 + a1 + 2a0 ) + (n(n − 1)an + (n − 1)an−1 − (n − 4)a2n−2 )xn−2
n=3
= 1 − x2 .

Then a0 and a1 are arbitrary, 2a2 + a1 + 2a0 = 1, 6a3 + 2a2 + a1 = 0, and

12a4 + 3a3 = −1. The recurrence relation is
−(n − 1)an−1 + an−4 an−2
an =
n(n − 1)
for n = 5, 6, · · · . The general solution is
 
2 1 3 1 4 1 5
y(x) = a0 1 − x + x − x + x − · · ·
3 12 30
 
1 1 1 1 1 6 1 7
+ a1 x − x2 + x2 − x3 − x4 − x + x + ··· .
2 2 6 24 360 2520
Here a0 = y(0) and a1 = y 0 (0).
10. Using the Maclaurin expansion of ex , we have
∞
X ∞
X
y 00 + xy 0 = n(n − 1)an xn−2 + nan xn
n=2 n=1
∞
X
= 2a2 + (n(n − 1)an + (n − 2)an−2 )xn−2
n=3
∞
X 1
=− xn−2 .
n=3
(n − 2)!

Then a0 and a1 are arbitrary, a2 = 0 and

−(n − 2)an−2 − 1/(n − 2)!
an =
n(n − 1)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.5. SERIES SOLUTIONS 65

for n = 3, 4, · · · . This leads to the solution

 
1 3 3 5 15 7 105 9
y(x) = a0 + a1 x − x + x − x + x + ···
3! 5! 7! 9!
 
1 3 1 4 2 5 3 6 11 7 19 8
+ − x − x + x + x − x + x + ··· .
3! 4! 5! 6! 7! 8!

Note that a0 = y(0) and a1 = y 0 (0).

2.5.2 Frobenius Solutions

P∞ n+r
1. Substitute y(x) = n=0 cn x into the differential equation to get

∞
X
xy 00 + (1 − x)y 0 + y = (n + r)(n + r − 1)cn xn+r−2
n=0
∞
X ∞
X ∞
X
+ (n + r)cn xn+r−1 − (n + r)cn xn+r + cn xn+r
n=0 n=0 n=0
∞
X
= r2 c0 xr−1 + ((n + r)2 cn − (n + r − 2)cn−1 )xn+r−1
n=1
= 0.

Because c0 is assumed to be nonzero, r must satisfy the indicial equation

r2 = 0, so r1 = r2 = 0. One solution has the form

∞
X
y1 (x) = cn xn ,
n=0

while a second solution has the form

∞
X
y2 (x) = y1 (x) ln(x) + c∗n xn .
n=0

For the first solution, choose the coefficients to satisfy c0 = 1 and

n−2
cn = cn−1 for n = 1, 2, · · · .
n2

This yields the solution y1 (x) = 1 − x. The second solution is therefore

∞
X
y2 (x) = (1 − x) ln(x) + c∗n xn .
n=0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 66 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

Substitute this into the differential equation to obtain

   
2 1−x 1−x
x − − + (1 − x) − ln(x) +
x x2 x
X∞ ∞
X
+ (1 − x) ln(x) + n(n − 1)c∗n xn−1 + (1 − x) c∗n xn−1
n=2 n=1
∞
X
+ c∗n xn
n=1
∞
X
= (−3 + c∗1 ) + (1 + 4c∗2 )x + (n2 c∗n − (n − 2)c∗n−1 )xn−2
n=3
= 0.

The coefficients are determined by c∗1 = 3, c∗2 = −1/4, and

n−2
c∗n = for n = 3, 4, · · · .
n2
A second solution is
∞
X 1
y2 (x) = (1 − x) ln(x) + 3x − xn .
n=2
n(n − 1)

2. Omitting some routine details, the indicial equation is r(r − 1) = 0, so

r1 = 1 and r2 = 0. There are solutions
∞
X ∞
X
y1 (x) = cn xn+1 and y2 (x) = ky1 (x) ln(x) c∗n xn .
n=0 n=0

For y1 , the recurrence relation is

2(n + r − 2)
cn = cn−1
(n + r)(n + r − 1)

for n = 1, 2, · · · . With r = 1 and c0 = 1, this yields

y1 (x) = x,

a solution that can be seen by inspection from the differential equation.

For the second solution, substitute y2 (x) into the differential equation to
get

(2c∗0 + k) + 2(c∗2 − k)x

X ∞
+ (n(n − 1)c∗n − 2(n − 2)c∗n−1 )xn−1 = 0.
n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.5. SERIES SOLUTIONS 67

Choose c∗0 = 1 to obtain k = −2. c∗1 is arbitrary, and we will take c∗1 = 0.
Finally, c∗2 = −2 and
2(n − 2) ∗
c∗n = c for n = 3, 4, · · · .
n(n − 1) n−1
This yields the second solution
∞
X 2n
y2 (x) = −2x ln(x) + 1 − xn .
n=2
n!(n − 1)

3. The indicial equation is r2 − 4r = 0, so r1 = 4 and r2 = 0. There are

solutions of the form
∞
X ∞
X
y1 (x) = cn x n+4
and y2 (x) = ky1 (x) ln(x) + c∗n xn .
n=0 n=0

With r = 4 the recurrence relation is

n+1
cn = cn−1 for n = 1, 2, · · · .
n
Then
y1 (x) = x4 (1 + 2x + 3x2 + 4x3 + · · · ).
Using the geometric series, we can observe that
d
y1 (x) = x4 (1 + x + x2 + x3 + · · · )
dx 
x4

d 1
= x4 = .
dx 1 − x (1 − x)2
This gives us the second solution
3 − 4x
y2 (x) = .
(1 − x)2

4. The indicial equation is 4r2 − 9 = 0, with roots r1 = 3/2 and r2 = −3/2.

There are solutions
∞
X ∞
X
y1 (x) = cn xn+3/2 and y2 (x) = ky1 (x) ln(x) + c∗n xn−3/2 .
n=0 n=0

Upon substituting these into the differential equation, we obtain

∞
" #
n
X (−1)
y1 (x) = x3/2 1 + n n!(5 · 7 · 9 · · · (2n + 3))
x2n
n=1
2

and
∞
" #
−3/2
X (−1)n+1 2n
y2 (x) = x 1+ x .
n=1
2n+1 n!(3) · · · (2n − 3)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 68 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

5. The indicial equation is 4r2 − 2r = 0, with roots r1 = 1/2 and r2 = 0.

There are solutions of the form
∞
X ∞
X
y1 (x) = cn xn+1/2 and y2 (x) = c∗n xn .
n=0 n=0

Substitute these into the differential equation to get

∞
" #
1/2
X (−1)n n
y1 (x) = x 1+ x
n=1
2n n!(3 · 5 · 7 · · · (2n + 1))
 
1 1 2 1 3 1
= x1/2 1 − x + x − x + x4 + · · ·
6 120 5040 362880
and
∞
X (−1)n
y2 (x) = 1 + n n!(1 · 3 · 5 · · · (2n − 1))
xn
n=1
2
1 1 1 3 1
= 1 − x + x2 − x + x4 + · · · .
2 24 720 40320

6. The indicial equation is 4r2 − 1 = 0, with roots r1 = 1/2 and r2 = −1/2.

There are solutions
∞
X ∞
X
y1 (x) = cn x n+1/2
and y2 (x) = ky1 (x) ln(x) + c∗n xn−1/2 .
n=0 n=0

After substituting these into the differential equation, we obtain the simple
solutions
y1 (x) = x1/2 and y2 (x) = x−1/2 .
These solutions are consistent with the observation that, upon division by
4, the differential equation is an Euler equation.
7. The indicial equation is r2 − 3r + 2 = 0, with roots r1 = 2 and r2 = 1.
There are solutions
∞
X ∞
X
y(x) = cn xn+2 and c∗n xn−2 .
n=0 n=0

Substitute these in turn into the differential equation to obtain the solu-
tions
1 1 1
y1 (x) = x2 + x4 + x6 + x8 + · · ·
3! 5! 7!
and
1 1 1
y2 (x) = x − x2 + x3 − x4 + x5 − · · · .
2! 3! 4!
We can recognize these series as
y1 (x) = x sinh(x) and y2 (x) = xe−x .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 2.5. SERIES SOLUTIONS 69

8. The indicial equation is r2 − 2r = 0, with roots r1 = 2, r2 = 0. There are

solutions
∞
X ∞
X
y1 (x) = cn xn+2 and y2 (x) = ky1 (x) ln(x) + c∗n xn .
n=0 n=0

The recurrence relation for the c0n s is

−2
cn = for n = 3, 4, · · ·
n(n − 2)

and we obtain, with c0 = 1,

∞
X (−1)n 2n+1 n+2
y1 (x) = x
n=0
n(n + 2)
2 1 1 1 6
= x2 − x3 + x4 − x5 + x + ··· .
3 6 45 540
For the second solution, substitute y2 (x) into the differential equation to
get
∞ 
(−1)n 2n k
X 
2c∗0 − c∗1 + n(n − 2)c∗n + c∗n−1 + 2
xn−1 = 0.
n=1
n((n − 2)!)

Setting c∗0 = 1 for simplicity, we obtain c∗1 = 2, k = −2, c∗2 arbitrary (we
take this to be zero), and the recurrence relation

(−1)n 2n+1
 
∗ 1 ∗
cn = − 2cn−1 +
n(n − 2) n((n − 2)!)2

for n = 3, 4, · · · . We obtain the second solution

16 3 25 4 157 6
y2 (x) = −2y1 ln(x) + 1 + 2x + x − x + x − ··· .
9 36 1350

9. The indicial equation is 2r2 = 0, with roots r1 = r2 = 0. There are

solutions
∞
X ∞
X
y1 (x) = cn xn and y2 (x) = y1 (x) ln(x) + c∗n xn .
n=0 n=1

Upon substituting these into the differential equation, we obtain the in-
dependent solutions
y1 (x) = 1 − x
and  
x
y2 (x) = (1 − x) ln − 2.
x−2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 70 CHAPTER 2. SECOND-ORDER DIFFERENTIAL EQUATIONS

10. The indicial equation is r2 − 1 = 0, with roots r1 = 1 and r2 = −1. There

are solutions
∞
X ∞
X
y1 (x) = cn xn+1 and y2 (x) = ky1 (x) ln(x) + c∗n xn−1 .
n=0 n=0

Substitute each of these into the differential equation to get

∞
" #
X (−1)n (1 · 4 · 7 · · · (3n − 2)) 3n
y1 (x) = x 1 + x
n=1
3n n!(5 · 8 · 11 · · · (3n + 2))

and
∞
" #
1 X (−1)n+1 (1 · 2 · 5 · · · (3n − 1)) 3n
y2 (x) = 1+ x .
x n=1
3n n!(4 · 7 · · · (3n − 2))

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 3

The Laplace Transform

3.1 Definition and Notation

1.
3(s2 − 4)
F (s) =
(s2 + 4)2

2.
8
G(s) =
(s + 4)2 + 64

3.
14 7
H(s) = − 2
s2 s + 49
4.
s s
W (s) = − 2
s2 + 9 s + 49
5.
10 3
K(s) = − + 2
(s + 4)2 s +9

6.
7
r(t) = sinh(3t)
3
7. q(t) = cos(8t)

8.
5 √ √
g(t) = √ sin( 12t) − 4 cos( 8t)
12
9.
1
p(t) = e−42t − t3 e−3t
6

71

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 72 CHAPTER 3. THE LAPLACE TRANSFORM

10.
5
f (t) = − t sin(t)
2
11. (a) From the definition,
Z R
F (s) = L[f ](s) = lim e−st f (t) dt.
R→∞ 0

For each R, let N be the largest integer such that (N + 1)T ≤ R to write
Z R N Z
X (n+1)T Z R
e−st f (t) dt = e−st f (t) dt + e−st f (t) dt.
0 n=0 nT (N +1)T

By choosing R sufficiently large, we can make the last integral on the right
as small as we like. Further, R → ∞ as N → ∞, so
Z ∞ X∞ Z (n+1)T
−st
e f (t) dt = e−st f (t) dt.
0 n=0 nT

(b) Use the periodicity of f (t) and the change of variables u = t − nT to

write
Z (n+1)T Z T
−st
e f (t) dt = e−s(u+nT ) f (u + nT ) dt
nT 0
Z T
= e−snT e−su f (u) du,
0

because f (u + nT ) = f (u).
(c) Use the results of (a) and (b) to write
∞ Z (n+1)T
X
L[f ](s) = e−st f (t) dt
n=0 nT

X∞ Z T
= e−snT f (t) dt
n=0 0
"∞ #Z
X T
= e−snT e−st f (t) dt,
n=0 0

because the summation is independent of t.

(d) For s > 0, 0 < e−st < 1 and we can use the geometric series to obtain
∞ ∞
X X
n 1
e−snt = e−sT = .
n=0 n=0
1 − e−sT

Therefore Z T
1
L[f ](s) = e−st f (t) dt.
1 − e−sT 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.1. DEFINITION AND NOTATION 73

12. f has period T = 6, and

Z 6 Z 3
5
e−st f (t) dx = 5e−st = (1 − e−3s ).
0 0 s
Then
5 1 − e−3s
L[F ](s) =
s 1 − e−6s
5 1 − e−3s
=
s (1 − e )(1 + e−3s )
−3s

5
= .
s(1 − e−3s )

13. f has period T = π/ω and

Z T Z π/ω
e−st f (t) dt = Ee−st sin(ωt) dt
0 0
Eω (1 + e−πs/ω )
= .
s2 + ω 2 1 − e−πs/ω

14. f has period T = 25 and, from the graph,


0 for 0 < t ≤ 5,

f (t) = 5 for 5 < t ≤ 10,

0 for 10 < t ≤ 25.


Now, Z 25 Z 10
5 −5s
e−st f (t) dt = 5e−st dt = e (1 − e−5s ).
0 5 s
Then
5e−5s (1 − e−5s )
L[f ](s) = .
s(1 − e−25s )

15. f has period 6 and f (t) = t/3 for 0 ≤ t < 6. Compute

Z T Z 6
1 −st 1
e−st f (t) dt = te dt = 2 (1 − 6se−6s − e−6s ).
0 0 3 3s
Then
1 1 − 6se−6s − e−6s
L[f ](s) = .
3s2 1 − e−6s
16. f is periodic with period 8, and
(
3t/2 for 0 < t < 2,
f (t) =
0 for 2 ≤ t ≤ 8.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 74 CHAPTER 3. THE LAPLACE TRANSFORM

Then
Z 8
1
L[f ](s) = e−st f (t) dt
1 − e−8s 0
3 1 − 2e−2s − e−2s
= 2 .
2s 1 − e−8s

17. Here (
h for 0 < t ≤ a,
f (t) =
0 for a < t ≤ 2a,
with period 2a. Then
Z 2a Z a
h
e−st f (t) dt = he−st dt = (1 − e−as ).
0 0 s
Then
h 1 − e−as
L[f ](s) = .
s 1 − e−2as
18. T = 2a and
(
ht/a for 0 ≤ t < a,
f (t) =
−h(t − 2a)/a for a < t ≤ 2a.

Then
h (1 − e−as )2
L[f ](s) =
as2 1 − e−2as
h 1 − e−as
= 2 .
as 1 + e−as
This can also be written in terms of the hyperbolic tangent function:
h  as 
L[f ](s) = 2 tanh .
as 2

3.2 Solution of Initial Value Problems

In many of these problems a partial fractions decomposition is used to find the
inverse transform of the transform of the solution (hence find the solution).
Partial fractions are reviewed in a web module.

1. Transform the differential equation, using the operational formula, to ob-

tain
1
sY (s) − y(0) + 4Y (s) = .
s
With y(0) = 3, this is
1
sY − 3 + 4Y = .
s

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.2. SOLUTION OF INITIAL VALUE PROBLEMS 75

Then  
1 1 1 − 3s
Y (s) = −3 = .
s+4 s s(s + 4)
Decompose this into a sum of simpler fractions:

A B
Y (s) = + .
s s+4
It these fractions are added, the numerator must equal the numerator of
the original fraction, 1 − 3s:

A(s + 4) + Bs = 1 − 3s.

Then
(A + B)s + 4A = 1 − 3s.
Matching coefficients of like powers of x, this requires that

A + B = −3 and 4A = 1.

Then A = 1/4 and B = −13/4. Now

1 1 13 1
Y (s) = − .
4s 4 s+4
Now we immediately read from Table 3.1 that
1 13 −4t
y(t) = L−1 [Y ](t) = − e .
4 4
This is the solution of the initial value problem.

2. The transform of the differential equation is

1
sY (s) − 5 − 9Y (s) = ,
s2
in which we have incorporated the initial condition y(0) = 5. Then

1 + 5s2
Y (s) = .
s2 (s − 9)

Use a partial fractions decomposition to write this as

A B C
Y (s) = + 2+ .
s s s−9

The numerator of the sum of these fractions must equal 1 + 5s2 , so

As(s − 9) + B(s − 9) + Cs2 = 1 + 5s2 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 76 CHAPTER 3. THE LAPLACE TRANSFORM

Equate coefficients of like powers of x on both sides of this equation to get

A + C = 5, −9A + B = 0, −9B = 1.
Then A = −1/81, B = −1/9 and C = 406/81, so
1 1 1 1 406 1
Y (s) = − − + .
81 s 9 s2 81 s − 9
From this we read the inverses to obtain
1 1 406 9t
y(t) = − − t + e .
81 9 81
3. Take the transform of the differential equation and insert the initial data
and solve for Y (s) to get
 
1 s
Y (s) = .
s + 4 s2 + 1
Use a partial fractions decomposition to obtain
4 1 1 4s + 1
Y (s) = − + .
17 s + 4 17 s2 + 1
The inverse of this is the solution:
4 4 1
y(t) = − e−4t + cos(t) + sin(t).
17 17 17
4. Take the transform of the differential equation to get
1
sY − 1 + 2Y = .
s+1
Then  
1 1 1
Y (s) = +1 = .
s+2 s+1 s+1
The solution is
y(t) = e−t .
5. The transform of the initial value problem is
1 1
sY − 4 − 2Y = − .
s s2
Then
 
1 1 1
Y (s) = − +4
s − 2 s s2
1 1 1 1 17 1
= − + .
2 s2 4s 4 s−2
The solution is
1 1 17
y(t) = t − + e2t .
2 4 4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.2. SOLUTION OF INITIAL VALUE PROBLEMS 77

6. Transform the differential equation, using the operational formula for the
second derivative to get

1
s2 Y (s) − sy(0) − y 0 (0) + Y (s) = .
s
Inserting the initial conditions, we have

1
s2 Y − 6s + Y = .
s
Then    
1 1 1 s
Y (s) = 2 + 6s = +5 2
.
s +1 s s s +1
The solution is the inverse of this expression:

y(t) = 1 + 5 cos(t).

7. Transform the differential equation to obtain

s
s2 Y − sy(0) − y 0 (0) − 4(sY − y(0)) + 4Y = .
s2 + 1
Insert the initial conditions to get
s
(s − 2)2 Y = + s − 5.
s2 + 1
Then
s s−5
Y (s) = + .
(s2 + 1)(s − 2)2 (s − 2)2
With a some manipulation, write this as

s3 − 5s2 + 2s − 5
Y (s) = .
(s2 + 1)(s − 2)2

Expand this in partial fractions:

As + B C D
Y (s) = + + .
s2 + 1 s − 2 (s − 2)2

To determine the coefficients, we need the numerator in the sum of these

fractions to equal the numerator in Y (s):

(As + B)(s − 2)2 + C(s − 2)(s2 + 1) + D(s2 + 1)

= (A + C)s3 + (−4A + B − 2C + D)s2 + (4A − 4B + C)s + (4B − 2C + D)
= s3 − 5s2 + 2s − 5.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 78 CHAPTER 3. THE LAPLACE TRANSFORM

Matching coefficients of powers of s, we obtain:

A + C = 1,
−4A + B − 2C + D = −5,
4A − 4B + C = 2,
4B − 2C + D = −5.
Then
3 4 22 13
A= ,B = − ,C = ,D = − .
25 25 25 5
Then
3 s 4 1 22 1 13 1
Y (s) = − + − .
25 s2 + 1 25 s2 + 1 25 s − 2 5 (s − 2)2
These terms are easily inverted to obtain the solution
3 4 22 13
y(t) = cos(t) − sin(t) + e2t − te2t .
25 25 25 5
8. Transform the differential equation to obtain
 
1 2
Y (s) = 2
s + 9 s3
2 1 2 1 2 s
= 3
− + 2
.
9s 81 s 81 s + 9
The solution is
1 2 2 2
y(t) = t − + cos(3t).
9 81 81
9. Upon transforming the differential equation and inserting the initial con-
ditions, we have
 
1 1 1
Y (s) = 2 + − 2s + 1
s + 16 s s2
1 1 1 1 33 s 15 1
= + − − .
16 s2 16 s 16 s2 + 16 64 s2 + 16
The solution is
1 33 15
y(t) = (1 + t) − cos(4t) + sin(4t).
16 16 64
10. Transforming the differential equation, we obtain
 
1 1
Y (s) = − 10
(s − 2)(s − 3) s + 1
29 1 39 1 1 1
= − + .
3 s−2 4 s − 3 12 s + 1
The solution is
29 2t 39 3t 1
y(t) = e − e + e−t .
3 4 12

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.3. THE HEAVISIDE FUNCTION AND SHIFTING THEOREMS 79

11. Begin with the definition of the Laplace transform and integrate by parts:
Z ∞
0
L[f ](s) = e−st f 0 (t) dt
0
Z ∞
b
= e−st f (t) a − −se−st f (t) dt
Z ∞0
= −f (0) + s e−st f (t) dt
0
= sF (s) − f (0).

12. Apply the definition of the transform to f 00 (t) and then use equation (3.1)
to obtain
Z ∞
L[f 00 ](s) = e−st f 00 (t) dt
0
Z ∞
−st ∞
−se−st f 0 (t) dt

=e 0
−
0
= −f 0 (0) + s(sF (s) − f (0))
= s2 F (s) − sf (0) − f 0 (0).

3.3 The Heaviside Function and Shifting Theo-

rems
In each of Problems 1–15, we will indicate shifting f (t) by a, replacing t with
t − a, by writing
[f (t)]t→t−a .
Similarly, if we replace s with s − a in the transform F (s) of f (t), we will write

[F (s)]s→s−a

or sometimes
L[f (t)]s→s−a .
This notation is sometimes useful in applying a shifting theorem or inverse
shifting theorem.

1. Apply the shifting theorem:

L[(t3 − 3t + 2)e−2t ](s) = L[t3 − 3t + 2]s→s+2

6 3 2
= 4
− 2
+ .
(s + 2) (s + 2) s+2

2. We know that
1 2
L[t](s) = 2
and L[−2](s) = − .
s s

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 80 CHAPTER 3. THE LAPLACE TRANSFORM

Use the shifting theorem to obtain

L[te−3t − 2e−3t ](s) = L[te−3t ](s) − 2L[−2e−3t ](s)

   
1 2
= −
s2 s→s+3 s s→s+3
1 2
= − .
(s + 3)2 s+3

3. First write

f (t) = [1 − H(t − 7)] + H(t − 7) cos(t)

= [1 − H(t − 7)] + H(t − 7) cos((t − 7) + 7)
= [1 − H(t − 7)] + cos(7)H(t − 7) cos(t − 7) − sin(7)H(t − 7) sin(t − 7).

Then
1 s 1
1 − e−7s + 2 cos(7)e−7s − 2 sin(7)e−7s .


L[f ](s) =
s s +1 s +1

4.
   
1 s
L[f ](s) = −
s2 s→s+4 s2 + 1 s→s+4
1 s+4
= 2
− .
(s + 4) (s + 4)2 + 1

5. First, write the function as

f (t) = t + (1 − 4t)H(t − 3)
= t + (1 − 4(t − 3) + 3)H(t − 3)
= t − 11H(t − 3) − 4(t − 3)H(t − 3).

Then
1 11 4
L[f ](s) = − e−3s − 2 e−3s .
s2 s s
6. First, write

f (t) = [2(t − π) + 2π + sin(t − π)][1 − H(t − π)].

Then
2 1 2 2π −πs s
L[f ](s) = − 2 − e−πs − e − 2 e−πs .
s2 s + 1 s2 s s +1

7. Replace s with s + 1 in the transform of 1 − t2 + sin(t) to get

1 2 1
L[f ](s) = − 2
+ .
s + 1 (s + 1) (s + 1)2 + 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.3. THE HEAVISIDE FUNCTION AND SHIFTING THEOREMS 81

8. In terms of the Heaviside function,

f (t) = t2 + (1 − t − 4t2 )H(t − 2)

= t2 + [1 − (t − 2) − 2 − 4((t − 2) + 2)2 ]H(t − 2)
= t2 − [17 + 17(t − 2) + 4(t − 2)2 ]H(t − 2).

Then  
2 17 17 8
L[f ](s) = − + 2 + 3 e−2s .
s3 s s s

9. First, write

f (t) = (1 − H(t − 2π)) cos(t) + H(t − 2π)(2 − sin(t)).

Then  
s 2 s 1
l[f ](s) = 2
+ − 2 − 2 e−2πs .
s +1 s s +1 s +1

10. First write

f (t) = −4(1 − H(t − 1)) + e−t H(t − 3)

= −4 + 4H(t − 1) + e−3 e−(t−3) H(t − 3).

Then
4 4 e−3 −3s
L[f ](s) = − + e−s + e .
s s s+1
11. Because
s2 − 9
L[t cos(t)](s) = ,
(s2 + 9)2
we obtain the transform of te−t cos(t) by replacing s with s + 1:

(s + 1)2 − 9
L[te−t cos(t)](s) = .
((s + 1)2 + 9)2

12. The transform of 1 − cosh(t) is

1 s
− 2 ,
s s −1
then
1 s−1
L[et (1 − cosh(t))](s) = − .
s − 1 (s − 1)2 − 1
This can be written
1 s−1
L[et (1 − cosh(t))](s) = − .
s − 1 s(s − 2)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 82 CHAPTER 3. THE LAPLACE TRANSFORM

13. First, put f (t) in terms of the Heaviside function as

f (t) = (1 − H(t − 16))(t − 2) − H(t − 16)

= t − 2 + (1 − t)H(t − 16).

Then  
1 2 1 1
L[f ](s) = − + − e−16s .
s2 s s s2

14. In terms of the Heaviside function,

f (t) = 1 − cos(2t) − (1 − cos(2t))H(t − 3π).

Then  
1 s
F (s) = − (1 − e−3πs ).
s s2 + 1

15. Replace s with s + 5 in the transform of t4 + 2t2 + 1 to obtain

24 4 1
F (s) = 5
+ 3
+ .
(s + 5) (s + 5) (s + 5)2

16. Notice that

1
F (s) = .
(s + 2)2 + 8
√ √
This is the transform of (1/2 2) sin(2 2t), with s replaced by s + 2.
Therefore,
1 √
f (t) = √ e−2t sin(2 2t).
2 2
17. Write
1
F (s) = .
(s − 2)2 + 1
This is the transform of sin(t) with s replaced by s − 2. Therefore

f (t) = e2t sin(t).

18.
e−5s
   
−1 −1 1
L 3
(t) = L
s s3 t→t−5
1
= (t − 5)2 H(t − 5).
2

19. Because 3/(s2 + 9) is the transform of sin(3t), then

1
f (t) = sin(3(t − 2))H(t − 2).
3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.3. THE HEAVISIDE FUNCTION AND SHIFTING THEOREMS 83

20. Because the transform of 3e−2t is 3/(s + 2), then

f (t) = e−2(t−4) H(−4).

21. We recognize that

1
F (s) =
(s + 3)2 − 2
so
1 √
f (t) = √ e−3t sinh( 2t).
2
22. Write
s−4
F (s) =
(s − 4)2 − 6
to obtain √
f (t) = e4t cosh( 6t).

23. Write
(s + 3) − 1
F (s) =
(s + 3)2 − 8
to obtain
√ 1 √
f (t) = e−3t cosh(2 2t) − √ e−3t sinh(2 2t).
2 2

24. Put a = 1 and F (s) = 1/(s − 5) in the second shifting theorem. Then
f (t) = e5t and
 
−1 1 −s
L e (t) = e5(t−1) H(t − 1).
s−5

25. First use a partial fractions decomposition to write

1 1 1 1 s
= − .
s(s2 + 16) 2
16 s 16 s + 16

From this,
1
f (t) = (1 − cos(4(t − 21)))H(t − 21).
16
26. By the first shifting theorem,
 Z t 
L e−2t e2w cos(3w) dw = F (s + 2),
0

where Z t 
F (s) = L e2w cos(3w) dw .
0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 84 CHAPTER 3. THE LAPLACE TRANSFORM

Now Z t
d
e2w cos(3w) dw = e2t cos(3t).
dt 0

By the operational rule for the Laplace transform, applied to the case of
a first derivative, we have

d t 2w
Z 
e cos(3w) , dw = L e2t cos(3t) (s)
 
L
dt 0
Z t 
=L e2w cos(3w) dw (s) = sF (s).
0

Then
1 1 s−2
F (s) = L[e2t cos(3t)](s) = .
s s (s − 2)2 + 9
Therefore
 Z t 
−2t 2w s
L e e cos(3w) dw = .
0 (s + 2)(s2 + 9)

27. The initial value problem is

y 00 + 4y = 3H(t − 4); y(0) = 1, y 0 (0) = 0.

This transforms to
 
3 1 s s
Y (s) = − 2 e−4s + 2 .
4 s s +4 s +4

Invert this to obtain the solution

3
y(t) = cos(2t) + (1 − cos(2(t − 4)))H(t − 4).
4

28. The problem is

y 00 − 2y 0 − 3y = 12H(t − 4); y(0) = 1, y 0 (0) = 0.

Transform this problem and solve for Y (s) to obtain

   
1 1 3 1 3 4 −4s
Y (s) = + + + − e .
4 s−3 s+1 s−3 s+1 s

Invert this to obtain the solution

1 3t
y(t) = (e + 3e−t ) + (e3(t−4) + 3e−(t−4) − 4)H(t − 4).
4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.3. THE HEAVISIDE FUNCTION AND SHIFTING THEOREMS 85

29. The problem is

y (3) − 8y 0 = 2H(t − 6); y(0) = y 0 (0) = 0.
The transform of the problem is
 
1 1 1 1 s
Y (s) = − + + e−6s .
4s 12 s − 2 6 s2 + 2s − 4
Invert this to obtain the solution
√
 
1 1 −2(t−6) 1 −(t−6)
y(t) = − + e + e cos( 3(t − 6)) H(t − 6).
4 12 6

30. The problem is

y 00 + 5y 0 + 6y = −2(1 − H(t − 3)); y(0) = y 0 (0) = 0.
Transform this problem to obtain
 
1 2 1 1
Y (s) = − − H(t − 3).
s + 2 3 s + 3 3s
The inverse of this is the solution
 
2 1 2
y(t) = e−2t − e−3t − − e−2(t−3) − e−3(t−3) H(t − 3).
3 3 3

31. The problem is

y (3) − y 00 + 4y 0 − 4y = 1 + H(t − 5); y(0) = y 0 (0) = 0, y 00 = 1.
Transform this to obtain
 
1 2 1 3 s 2 1
Y (s) = − + − − (1 − e−5s ).
4s 5 s − 1 20 s2 + 4 5 s2 + 4
Invert this for the solution
1 2 3 1
y(t) = − + et − cos(2t) − sin(2t)
 4 5 20 5 
1 2 t−5 3 1
− − + e − cos(2(t − 5)) − sin(2(t − 5)) H(t − 5).
4 5 20 5

32. The initial value problem is

y 00 − 4y 0 + 4y = t + 2H(t − 3); y(0) = −2, y 0 (0) = 1.
Take the transform of the problem to obtain
1 1 9 1 43 1
Y (s) = + 2− −
4s 4s 4s−2 4 (s − 2)2
 
1 1 1 1
+ − + e−3s .
2s 2 s − 2 (s − 2)2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 86 CHAPTER 3. THE LAPLACE TRANSFORM

The solution is
1 1 9 43
y(t) = + t − e2t − te2t
4 4 4 4 
1 1 2(t−3)
+ − e + (t − 3)e2(t−3) H(t − 3).
2 2

33. The current i(t) is modeled by

Li0 + Ri = k(1 − H(t − 5)); i(0) = 0.
Transform the problem and solve for I(s) to get
k
I(s) = (1 − e−5s )
Ls + R
 
k 1 1
= − (1 − e−5s ).
R s s + R/L
Invert this to obtain the solution for the current:
k k
i(t) = (1 − e−Rt/L ) − (1 − e−R(t−5)/L )H(t − 5).
R R
In Problems 34–38, the Heaviside formula is used to compute an inverse
transform. One way to use this formula efficiently is to begin with F (s), which
has the form
p(s)
F (s) = .
(s − a1 )(s − a2 ) · · · (s − an )
For the first term, cover up the s − a1 factor and evaluate the resulting quotient
at a1 to get the coefficient of ea1 t . Then cover up the s − a2 factor and evaluate
the resulting quotient at a2 for the coefficient of ea2 t . Continue this through the
n simple zeros of the denominator to obtain f (t).
34 . Write
s+1
F (s) = ,
(s − 1)(s − 2)
so a1 = 1 and a2 = 2. Then
2 t 3 2t
f (t) = e + e
−1 1
= −2et + 3e2t .

35. With
s2
F (s) =
(s − 1)(s − 2)(s + 5)
we have a1 = 1, a2 = 2 and a3 = −5. Then
12 22 2t 52
f (t) = et + e + e5t
(−1)(6) (1)(7) (−6)(−7)
1 4 25
= et + e2t + e5t .
6 7 42

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.3. THE HEAVISIDE FUNCTION AND SHIFTING THEOREMS 87

36. With
s+4
F (s) = ,
(s + 3i)(s − 3i)
then
4 + 3i 3it 4 − 3i −3it
f (t) = e + e .
6i −6i
This is correct, but we can write the inverse in terms of trigonometric
functions using Euler’s formula:
4 + 3i 4 − 3i
f (t) = (cos(3t) + i sin(3t)) + (cos(3t) − i sin(3t))
6i −6i
1
= [4 cos(3t) + 3i sin(3t) + 4i sin(3t) − 3 sin(3t) − 4 cos(3t) + 3i cos(3t) + 4i sin(3t) + 3 sin(3t)]
6i
4
= cos(3t) + sin(3t).
3

37. Here
s2 + 2s − 1
F (s) = .
(s − 3)(s − 5)(s + 8)
Then
14 34 5t 47
f (t) = e3t + e + e−3t
(−2)(11) (2)(3) (−11)(−13)
7 17 47 −3t
= − e3t + e5t + e .
11 13 143

38. Now
s
F (s) = .
(s − 2i)(s + 2i)(s − 3i)(s + 4i)
Using Heaviside’s formula and Euler’s formula, as in Problem 36, we obtain
1 1
f (t) = cos(2t) − cos(3t).
5 5

39. Write
p(s) p(s)
(s − aj ) =
q(s) (q(s) − q(aj )/(s − aj ))
and take the limit as s → aj . Finally, use the fact that

q(s) − q(aj )
lim = q 0 (aj ).
s→aj s − aj

40. We will find A1 . The process for each Aj is the same. First observe that

p(s) s − a1 s − a1
(s − a1 ) = A1 + A2 + · · · + An ,
q(s) s − a2 s − an

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 88 CHAPTER 3. THE LAPLACE TRANSFORM

in which we use the fact that the zeros of q(s) are simple, so a1 6= aj
for j = 2, · · · , n. In the limit as s → a1 , the right side of this equation
approaches A1 , because all other terms have limit zero. But in this limit,
(s − a1 )p(s)/q(s) is exactly the quotient of p(s) with the polynomial q1 (s)
obtained from q(s) by deleting the factor s − a1 . This yields Heaviside’s
formula.

3.4 Convolution
1. Let
1 1
F (s) = and G(s) = 2 .
s2 +4 s −4
Then
1 1
L−1 [F ](t) = sin(2t) and L−1 [G](t) = sinh(2t).
2 2
By the convolution theorem,

1 1
L−1 [F (s)G(s)](t) = sin(2t) ∗ sinh(2t)
2 2
1 t
Z
= sin(2(t − τ )) sinh(τ ) dτ
4 0
1 t
= [sin(2(t − τ )) cosh(2τ ) + cos(2(t − τ )) sinh(2τ )]0
16
1
= (sinh(2t) − sin(2t)).
16

2. Choose
s 1
F (s) = and G(s) = e−2s .
s2 + 16 s
Then

L[F (s)G(s)](t) = cos(4t) ∗ H(t − 2)

Z t
= cos(4(t − τ ))H(τ − 2) dτ
(0
0 for t < 2,
= Rt
2
cos(4(t − τ )) dτ for t ≥ 2.

Upon carrying out the last integration, we have

e−2s
 
1
L−1 2
(s) = sin(4(t − 2))H(t − 2).
s + 16 4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.4. CONVOLUTION 89

3. There are two cases. Suppose first that a2 6= b2 . Then

 
s 1 1
L−1 2 2 2 2
(t) = cos(at) ∗ sin(bt)
s +a s +b b
Z t
1
= cos(a(t − τ )) sin(bτ ) dτ
b 0
Z t
1
= [sin((b − a)τ + at) + sin((b + a)τ − at) dτ ] dτ
2b 0
 t
1 cos((b − a)τ + at) cos((b + a)τ − at)
= − −
2b b−a b+a
  0
1 cos(bt) cos(bt) cos(at) cos(at)
= − − + +
2b b−a b+a b−a b+a
cos(at) − cos(bt)
= .
(b − a)(b + a)
If a2 = b2 , then
 
s 1 1
L−1 2 (t) = cos(at) ∗ sin(at)
s + a2 s2 + b2 a
1 t
Z
= cos(a(t − τ )) sin(aτ ) dτ
a 0
Z t
1
= (sin(at) + sin(2aτ − at)) dτ
2a 0
 t
1 1
= τ sin(at) − cos(a(2τ − t))
2a 2a 0
1
= t sin(at).
2a
4. First write
s 1 s
= .
(s − 3)(s2 + 5) s − 3 s2 + 5
Then
√
 
−1 s 3t
L (t) = e ∗ cos( 5t)
(s − 3)(s2 + 5)
Z t √
= cos( 5τ )e3(t−τ ) dτ
0
Z t √
= e3t cos( 5τ )e−3τ dτ
0
t
√ √ √

1 −3τ
= e3t e − 3 cos( 5τ ) + 5 sin( 5τ )
14 0
√
3 3t 3 √ 5 √
= e − cos( 5t) + sin( 5t).
14 14 14

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 90 CHAPTER 3. THE LAPLACE TRANSFORM

5. First,
   
−1 1 1 −1 1 1
L (t) = 2 (1−cos(at)) and L (t) = t sin(at).
s(s2 + a2 ) a s2 + a2 a
Then
 
−1 1 1
L (t) = 2 [1 − cos(at)] ∗ sin(at)
s(s2 + a2 ) a
Z t
a
= 3 [1 − cos(a(t − τ ))] sin(aτ ) dτ
a 0
 t
1 1 1 1
= 3 − cos(aτ ) − τ sin(at) + cos(2aτ − at)
a a 2 4a 0
1 1
= 4 (1 − cos(at)) − 3 sin(at).
a 2a
6.
 
−1 1 1 1
L (t) = e5t ∗ t3
s4 s − 5 6
Z t
1
= e5(t−τ ) τ 3 dτ
6 0
Z t
1
= e5t τ 3 e−5τ dτ
6 0
1 5t 1 1 1 1
= e − t3 − t2 − t− .
625 30 50 125 625
7.
1 e−4s
 
−1
L (t) = e−2t ∗ H(t − 4)
s+2 s
Z t
= e−2(t−τ ) dτ
4
(
1 −2(t−4)
e if t > 4,
= 2
0 if t ≤ 4.
We can therefore write the inverse transform as
1 e−4s
 
1
L−1 (t) = (1 − e−2(t−4) )H(t − 4).
s+2 s 2
8.
  √
−1 2 1 2 sin( 5t)
L (t) = t ∗ √
s2 s2 + 5 5
1
Z t √
=√ τ 2 sin( 5(t − τ )) dτ
5 0
1 2 2 √
= t− + cos( 5t).
5 25 25

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.4. CONVOLUTION 91

9. Take the transform of the initial value problem and solve for Y (s) to get
 
F (s) 1 1
Y (s) = = − F (s).
s2 − 5s + 6 s−3 s−2

By the convolution theorem,

y(t) = e3t ∗ f (t) − e2t ∗ f (t).

10. Take the transform of the initial value problem to obtain

F (s) s
Y (s) = +
(s + 6)(s + 4) (s + 6)(s + 4)
   
1 1 1 3 1
= F (s) − + −2 .
2 s+4 s+6 s+6 s+4

Then
1 −4t 1
y(t) = e ∗ f (t) − e−6t ∗ f (t) + 3e−6t − 2e−4t .
2 2

For Problems 11–16 the solution is given, but the details (similar to those
of Problems 9 and 10) are omitted.

11.
1 6t 1
y(t) = e ∗ f (t) − e2t ∗ f (t) + 2e6t − 5e2t
4 4

12.
1 5t 1 1 3
y(t) = e ∗ f (t) − e−t ∗ f (t) + e5t + e−t
6 6 2 2

13.
1 1
y(t) = sin(3t) ∗ f (t) − cos(3t) + sin(3t)
3 3

14.
1 4
y(t) = sinh(kt) ∗ f (t) − 2 cosh(kt) − sinh(kt)
k k

15.
1 2t 1 1 1 1 4
y(t) = e ∗ f (t) + e−2t ∗ f (t) − et ∗ f (t) − e2t − e−2t + et
4 12 3 4 12 3

16.
√ √
1 3t 1 −3t 2 √2t 2 −√2t
y(t) = e ∗ f (t) − e ∗ f (t) − e ∗ f (t) − e ∗ f (t)
42 42 28 28

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 92 CHAPTER 3. THE LAPLACE TRANSFORM

17. The integral equation can be expressed as

f (t) = −1 + f (t) ∗ e−3t .

Take the transform of this equation to obtain

1 F (s)
F (s) = − + .
s s+3
Then
s+3 1 1 31
F (s) = − = − .
s(s + 2) 2s+2 2s
Invert to obtain the solution
1 −2t 3
f (t) = e − .
2 2

18. The equation is f (t) = −1 + f (t) ∗ sin(t). Take the transform of this
equation and solve for F (s) to obtain

(s2 + 1) 1 1
F (s) = − = − 2 − 4.
s4 s s
Then
1
f (t) = −t − t3 .
6

19. The equation is f (t) = e−t + f (t) ∗ 1. Take the transform and solve for
F (s) to get

s 1 1 1 1
F (s) = = + .
(s − 1)(s + 1) 2s+1 2s−1

Then
1 −t 1 t
f (t) = e + e = cosh(t).
2 2

20. The equation is f (t) = −1 + t − 2f (t) ∗ sin(t). From this we obtain

(1 − s)(s2 + 1)
F (s) =
s2 (s2 + 3)
   
1 1 11 2 s 2 1
= − − + .
3 s2 3 s 3 s2 + 3 3 s2 + 3

Invert this to get

√
1 1 2 √ 2 3 √
f (t) = t − − cos( 3t) + sin( 3t).
3 3 3 9

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.4. CONVOLUTION 93

21. The equation is f (t) = 3 + f (t) ∗ cos(2t). From this we obtain

3(s2 + 4) 3 3
F (t) = = + 2 .
s(s2 − s + 4) s s −s+4
The inverse of this is
√ √ !
2 15 t/2 15
f (t) = 3 + e sin t .
5 2

22. Write the equation as

Z t
f (t) = cos(t) + f (τ )e−2(t−τ ) dτ = cos(t) + f (t) ∗ e−2t .
0

Transform this and solve for F (s) to obtain

s(s + 2)
F (s) =
(s + 1)(s2 + 1)
   
1 3 s 1 1
=− + + .
2(s + 1) 2 s2 + 1 2 s2 + 1
Invert this to get the solution
1 −t 3 1
f (t) = e + cos(t) + sin(t).
2 2 2
23. Let F = L[f ] and G = L[g]. Then
Z ∞
F (s)G(s) = F (s) e−sτ g(τ ) dτ.
0

Now recall that

e−sτ F (s) = L[H(t − τ )f (t − τ )](s).
Substitute this into the expression for F (s)G(s) to get
Z ∞
F (s)G(s) = L[H(t − τ )f (t − τ )](s)g(τ ) dτ.
0

But, from the definition of the Laplace transform,

Z ∞
L[H(t − τ )f (t − τ )] = e−st H(t − τ )f (t − τ ) dt.
0

Then
Z ∞ Z ∞ 
F (s)G(s) = e−st H(t − τ )f (t − τ ) dt g(τ ) dτ
Z0 ∞ Z 0
∞
= e−st g(τ )H(t − τ )f (t − τ ) dτ dt.
0 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 94 CHAPTER 3. THE LAPLACE TRANSFORM

But H(t − τ ) = 0 if 0 ≤ t < τ , while H(t − τ ) = 1 if t ≥ τ . Therefore

Z ∞Z ∞
F (s)G(s) = e−st g(τ )f (t − τ ) dt dτ.
0 τ

The last integration is over the wedge in the t, τ − plane consisting of

points (t, τ ) with 0 ≤ τ ≤ t < ∞. Reverse the order of integration to
write
Z ∞Z t
F (s)G(s) = e−st g(τ )f (t − τ ) dτ dt
0 0
Z ∞ Z t 
−st
= e g(τ )f (t − τ ) dτ dt
Z0 ∞ 0

−st
= e (f ∗ g)(t) dt
0
= L[f ∗ g](s).

This is what we wanted to show.

3.5 Impulses and the Dirac Delta Function

In Problem 1 details of the solution are given. For Problems 2–5, the details
are similar and only the solution is given.

1. Transform the initial value problem to obtain

(s2 + 5s + 6)Y (s) = 3e−2s − 4e−5s .

Using a partial fractions decomposition, this gives us

   
1 1 −3s 1 1
Y (s) = 3 − e −4 − e−5s .
s+2 s+3 s+2 s+3

Invert this to obtain the solution

h i h i
y(t) = 3 e−2(t−2) − e−3(t−2) H(t − 2) − 4 e−2(t−5) − e−3(t−5) H(t − 5).

2.
4 2(t−3)
y(t) = e sin(3(t − 3))H(t − 3)
3
3.
y(t) = 6(e−2t − e−t + te−t )

4.
y(t) = 3 cos(4t) + 3 sin(4(t − 5π/8))H(t − 5π/8)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.6. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS 95

5.
y(t) = (B + 9)e−2t − (B + 6)e−3t

6. Begin with
Z ∞ Z ∞
1
f (t)δ (t − a) dt = [H(t − a) − H(t − a − )]f (t) dt
0 0 
Z a+
== f (t) dt.
a

By the mean value theorem for integrals, there is some t between t and
t +  such that
Z a+
1
f (t) dt = f (t ).
a 
Then Z ∞
f (t)δ (t − a) dt = f (t ).
0

Now let  → 0+. Then a +  → a, so t → a also. Because f (t) is assumed

to be continuous, f (t ) → f (a). Finally, we have
Z ∞ Z ∞
lim f (t)δ (t − a) dt = f (t) lim δ (t − a) dt
→0+ 0 →0+
Z0 ∞
= f (t)δ(t − a)
0
= lim f (t ) = f (a).
→0+

3.6 Systems of Linear Differential Equations

1. Take the transform of the system:

1
sX − 2sY = , sX − X + Y = 0.
s
Then
11 2 4
X(s) = − + , =−
s2 (2s
− 1) s 2 s 2s − 1
1−s 1 1 2
Y (s) = 2 =− 2 − + .
s (2s − 1) s s 2s − 1

Apply the inverse transform to get the solution

x(t) = −t − 2 + 2et/2 , y(t) = −t − 1 + et/2 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 96 CHAPTER 3. THE LAPLACE TRANSFORM

2. Take the transform of the system to obtain

1
2sX + (s − 3)Y = 0, sX + sY = .
s2
Then
s−3 2
X(s) = − , Y (s) = 2 .
s2 (s + 3) s (s + 3)
Using partial fractions, we can write
21 2 1 1 2 1
X(s) = − + 3− ,
9 s 3 s2 s 9s+3
21 2 1 2 1
Y (s) = − + + .
9 s 3 s2 9s+3
Inverting, we obtain the solution
2 2 1 2
x(t) = − t + t2 − e−3t ,
9 3 2 9
2 2 2 −3t
y(t) = − + t + e .
9 3 9

3. After transforming the system, we obtain

1
sX + (2s − 1)Y = , 2sX + Y = 0.
s
Then
1 4 16 1
X(s) = − = − + ,
s2 (4s − 3) 9s 9(4s − 3) 3s2
2 2 8
Y (s) = =− + .
s(4s − 3) 3s 3(4s − 3)
Invert these to obtain
4 1
x(t) = (1 − e3t/4 ) + t,
9 3
2
y(t) = (−1 + e3t/4 ).
3

4. The transformed system is

s
sX + sY − X = , sX + 2sY = 0.
s2 + 1
Then
2s 1 4s − 2 4
X(s) = =− 2 + ,
s3 − 2s2 + s − 2 5 s + 1 5(s − 2)
s 1 2s − 1 2
Y (s) = − 3 2
= 2
− .
s − 2s + s − 2 5 s + 1 5(s − 2)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.6. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS 97

The solution is
4 2 4
x(t) = − cos(t) + sin(t) + e2t ,
5 5 5
2 1 2 2t
y(t) = cos(t) − sin(t) − e .
5 5 5

5. The system transforms to

2
3sX − Y = sX + sY − Y = 0.
s2
Then
2(s − 1) 3 1 1 9
X(s) = = + + 3− ,
s2 (3s − 2) 4s 2s2 s 4(3s − 2)
2 3 1 2(3s − 2)
Y (s) = − 2 = + −9
s (3s − 2) 2s s2 .

Then
3 1 1 3
x(t) = + t + t2 − e2t/3 ,
4 2 2 4
3 3 2t/3
y(t) = + t − e .
2 2

6. Apply the transform to get

1
sX + 4sY − Y = 0sX + 2Y = .
s+1
Then
4s − 1 1 32 5
X(s) = = + −
s(4s2 + s − 3) 3s 21(4s − 3) 7(s + 1)
1 4 1
Y (s) = − 2 =− + .
4s + s − 3 7(4s − 3) 7(s + 1)

Then
1 8 5
x(t) = + e3t/4 − e−t ,
3 21 7
1 3t/4 1 −t
y(t) = − e + e .
7 7

7. The transform of the system is

2
sX + 2X − sY = 0, sX + Y + X = .
s3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 98 CHAPTER 3. THE LAPLACE TRANSFORM

Then
2 1 s+1 1
X(s) = = 2+ 2 − ,
s2 (s2 + 2s + 2) s s + 2s + 2 s
2(s + 2) 1 1 2
Y (s) = 3 2 =− 2 + 2 + .
s (s + 2s + 2) s s + 2s + 2 s3
The solution is

x(t) = t + e−t cos(t) − 1

y(t) = −t + e−t sin(t) + t2 .

In inverting X(s) and Y (s), the terms involving s2 + 2s + 2 can be treated

by using a shifting theorem, expressing these as functions of s + 1.
8. The system transforms to
1
sX + 4X − Y = 0, sX + sY = .
s2
Then
1 1 1 1 1
X(s) = = − − + ,
s2 (s + 5) 125s 25s2 125(s + 5) 5s2
s+4 1 1 1 4
Y (s) = 3 =− + + + .
s (s + 5) 125s 25s2 125(s + 5) 5s3
The solution is
1 1 1 −5t 1
x(t) = − t− e + t2
125 25 125 10
1 1 1 −3t 2 2
y(t) = − + t+ e + t .
125 25 125 5

9. First,
1
sX + sY + X − Y = 0, sX + 2sY + X = .
s
Then
1−s −2 1 1
X(s) = = − + ,
s(s + 1)2 s + 1)2 s+1 s
1 1 1
Y (s) = = − .
s(s + 1) s s+1
Then
x(t) = 1 − e−t (2t + 1), y(t) = 1 − e−t .

10. The transform of the system is

6
sX + 2sY − X = 0, 4sX + 3sY + Y = − .
s

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.6. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS 99

Then,
12 6(s − 1)
X(s) = − , Y (s) = .
5s2 + 2s + 1 s(5s2 + 2s + 1)
By using a shifting theorem, invert these to get

x(t) = −6e−t/5 sin(2t/5),

y(t) = −6 + 6e−t/5 (cos(2t/5) + sin(2t/5)) .

11. The system transforms to

1 1
sX − 2sY + 3X = 0, X − 4sY + 3sZ = , X − 2sY + 3sZ = − .
s2 s
Then
s+1 2 2 1
X(s) = = − − 2,
s2 (s
+ 3) 9s 9(s + 3) 3s
1s+1 −1 1
Y (s) = − 3
= 3 − 2,
2 s 2s 2s
2 s2 + 3s + 1 2 2 2 16
Z(s) = − =− − + − .
3 s3 (s + 3) 9s 81s 81(s + 3) 27s2
The solution is
2 1 2
x(t) = + t − e−3t ,
9 3 9
1
y(t) = − t(t + 2)
4
16 1 2 2
z(t) = − t − t2 − + e−3t .
27 9 81 81

12. The loop currents in the circuit satisfy

5i01 + 5i1 − 5i02 = 1 − H(t − 4) sin(2(t − 4)),

−5i01 + 5i02 + 5i2 = 0.

Apply the Laplace transform to these equations and solve for I1 (s) and
I2 (s) to get

2e−4s
 
s+1 1
I1 (s) = − 2
5(2s + 1) s s + 4
   
1 1 1 2 2 s 9
= − − − 2 + 2 e−4s
5 s 2s + 1 85 2s + 1 s + 4 s + 4
2e−4s
 
1
I2 (s) = 1− 2
2s + 1 s +4
 
1 2 2 s 8
= + − 2 − 2 e−4s .
5(2s + 1) 85 2s + 1 s + 4 s + 4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 100 CHAPTER 3. THE LAPLACE TRANSFORM

Apply the inverse Laplace transform to obtain the solution for the currents:
 
1 1 −t/2
i1 (t) = 1− e
5 2
 
2 −(t−4)/2 9
− e − cos(2(t − 4)) + sin(2(t − 4)) H(t − 4),
85 2
1 −t/2
i2 (t) = e
10
2 h −(t−4)/2 i
+ e − cos(2(t − 4)) − 4 sin(2(t − 4)) H(t − 4).
85

13. The equations for the loop currents are

20i01 + 10(i1 − i2 ) = E(t) = 5H(t − 5),

30i02 + 10i2 + 10(i2 − i1 ) = 0.

Initial conditions are

i1 (0) = i2 (0) = 0.
Transform the system to obtain

5(30s + 20)e−5s
I1 (s) =
s(600s2 + 700s + 100)
 
1 1 27 1
= − − e−5s ,
s 10(s + 1) 5 6s + 1
50e−5s
I2 (s) = 2
s(600s + 700s + 100)
 
1 10 18 1
= + − e−5s .
2s s + 1 5 6s + 1

Invert these to obtain the current functions:

 
1 9
i1 (t) = 1 − e−(t−5) − e−(t−5)/6 H(t − 5),
10 10
 
1 1 3
i2 (t) = + e−(t−5) − e−(t−5)/6 H(t − 5).
2 10 10

14. Let x1 (t) and x2 (t) be the amounts of salt (in kilograms) in tanks 1 and
2, respectively, at time t. Now

x01 (t) = rate of change of salt in tank 1

= (rate salt is added ) − ( rate salt is removed).

Then
1 3 5
x01 (t) = + x2 − x1 .
3 18 60

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 3.6. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS 101

Similarly,
5 5
x02 = x1 − x2 + 11(H(t − 4) − H(t − 6)).
60 18
Initial conditions are x1 (0) = 11, x2 (0) = 7. Transform this system to
obtain
4
(12s + 1)X1 − 2X2 = + 132,
s
396 −4s −6s
−3X1 + (36s + 10)X2 = (e − e ) + 252.
s
Then
4752s2 + 1968s + 40 + 792(e−4s − e−6s )
X1 (s) =
s(432s2 + 156s + 4)
 
10 6 108 99 27 3888
= − + +2 + − (e−4s − e−6s ),
s 3s + 1 36s + 1 s 3s + 1 36s + 1
3024s2 + 648s + 12 + 396(12s + 1)(e−4s − e−6s )
X2 (s) =
s(432s2 + 156s + 4)
 
3 9 36 99 81 2592
= + + + − − (e−4s − e−6s ).
s 3s + 1 36s + 1 s 3s + 1 36s + 1
Apply the inverse transform to obtain the solution:
x1 (t) = 10 − 2e−t/3 + 3e−t/36 + 2(99 + 9e−(t−4)/3 − 108e−(t−4)/6 )H(t − 4)
− 2(99 + 9e−(t−6)/3 − 108e−(t−6)/36 )H(t − 6),
x2 (t) = 3 + 3e−t/3 + e−t/36 + (99 − 27e−(t−4)/3 − 72e−(t−4)/36 )H(t − 4)
− (99 − 27e−(t−6)/3 − 72e−(t−6)/36 )H(t − 6).

15. Using the notation of the preceding problem, we can write

6 3
x01 = − x1 + x2 ,
200 100
4 4
x02 = x1 − x2 + 5H(t − 3).
200 200
Initial conditions are x1 (0) = 10, x2 (0) = 5. Apply the transform to this
initial value problem and rearrange terms to obtain
(100s + 3)X1 − 3X2 = 1000,
−2X1 + (100s + 4)X2 = 500 + 500e−3s .
Solve these to get
100000s + 5500 + 1500e−3s
X1 (s) =
10000s2 + 700s + 6
 
50 900 300 150
= + + − e−3s ,
50s + 3 100s + 1 100s + 1 50s + 3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 102 CHAPTER 3. THE LAPLACE TRANSFORM

and
50000s + 3500 + (50000s + 1500)e−3s
X2 (s) =
10000s2 + 700s + 6
 
50 600 150 200
=− + + + e−3s .
50s + 3 100s + 1 50s + 3 100s + 1

Apply the inverse transform to obtain the solution:

i1 (t) = e−3t/50 + 9e−t/100 + 3(e−(t−3)/100 − e−3(t−3)/50 )H(t − 3),

i2 (t) = −e−3t/50 + 6e−t/100 + (3e−3(t−3)/50 + 2e−(t−3)/100 )H(t − 3).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 4

Sturm-Liouville Problems
and Eigenfunction
Expansions

4.1 Eigenvalues, Eigenfunctions and Sturm-Liouville

Problems
For these problems, an eigenfunction is found for each eigenvalue, and it is
understood that nonzero constant multiples of eigenfunctions are also eigen-
functions.

1. The problem is regular on [0, L]. To find the eigenvalues and eigenfunc-
tions, take cases on λ.

Case 1. If λ = 0, the differential equation is y 00 = 0, with solutions

y = a + bx. Now y(0) = a = 0, so y = bx. But then y 0 (L) = b = 0 also,
so this case has only the trivial solution and 0 is not an eigenvalue of this
problem.

Case 2. If λ is negative, say λ = −α2 with α > 0, then the differential

equation is
y 00 − α2 y = 0
with general solution
y = c1 eαx + c2 e−αx .
Now
y(0) = c1 + c2 = 0,
so c2 = −c1 and

y(x) = c1 eαx − e−αx = 2c1 sinh(αx).

103

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 104CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

From the other boundary condition,

y 0 (L) = 2c1 α cosh(αL) = 0.

But cosh(αL) > 0 and α > 0, so we must have c1 = 0 and this case also
has only the trivial solution. This problem has no positive eigenvalue.

Case 3. Suppose λ is positive, write λ = α2 , with α > 0. Now the

differential equation is
y 00 + α2 y = 0,
with general solution

y(x) = c1 cos(αx) + c2 sin(αx).

Immediately y(0) = c1 = 0, so

y(x) = c2 sin(αx).

From the other boundary condition, we must have

y 0 (L) = c2 α cos(αL) = 0.

We need to be able to choose c2 6= 0 to have nontrivial solutions. This

requires that we α must be chosen to satisfy

cos(αL) = 0.

We know that the zeros of the cosine function have the form (2n − 1)π/2
for integer π, so let
(2n − 1)π
αL = ,
2
with n = 1, 2, · · · . Then acceptable values of α are

(2n − 1)π
α= .
2L
Because λ = α2 , the eigenvalues of this problem, indexed by n, are
 2
(2n − 1)π
λn =
2L

for n = 1, 2, · · · . Corresponding eigenfunctions are

 
(2n − 1)π
ϕn (x) = sin x ,
2L

or any nonzero constant multiple of this function.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.1. EIGENVALUES, EIGENFUNCTIONS AND STURM-LIOUVILLE PROBLEMS105

2. The problem is regular on [0, L]. The eigenvalues are

λ0 = 0, λn = n2 for n = 1, 2, · · · .

Corresponding eigenfunctions are

ϕn (x) = cos(nπx) for n = 0, 1, 2, · · · .

Any nonzero constant multiple of an eigenfunction is also an eigenfunction

(for the same eigenvalue). Notice in this example that 0 is an eigenvalue.
The number zero can be an eigenvalue, but the trivial function (identically
zero) cannot be an eigenfunction.
3. The problem is regular on [0, L];
  2
1 π
λn = n−
2 4
is an eigenvalue for n = 1, 2, · · · , with eigenfunctions
 
(2n − 1)π
ϕn (x) = cos .
8

4. The problem is periodic on [0, π]. Eigenvalues are λn = 4n2 for n =

0, 1, 2, · · · and eigenfunctions are

ϕn (x) = an cos(2nx) + bn sin(2nx)

with an and bn constant and not both zero.

5. The problem is periodic on [−3π, 3π]. Eigenvalues are

n2
λ0 = 0 and λn = for n = 1, 2, · · · .
9
Eigenfunctions are

ϕn (x) = an cos(nx/3) + bn sin(nx/3)

for n = 0, 1, 2, · · · , with an and bn constant and not both zero.

6. The problem is regular on [0, π]. To find the eigenvalues and eigenfunc-
tions, take cases on λ.

Case 1. If λ = 0, then y = ax + b for constants a and b. Because

y(0) = b = 0, then y = ax. But then

y(π) + 2y 0 (π) = 0 = aπ,

so a = 0 and this case has only the trivial solution. 0 is not an eigenvalue
of this problem.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 106CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Case 2. If λ = −α2 , with α > 0, then

y(x) = c1 eαx + c2 e−αx .

Now y(0) = c1 + c2 = 0, so c2 = −c1 and

y(x) = c1 (eαx − e−αx ).

From the boundary condition at π, we must have

y(π) + 2y 0 (π) = c1 (eαπ − e−απ ) = 0.

But e−απ < eαπ because the exponential function is strictly increasing.
Therefore c1 = 0 and this case admits only the trivial solution. The
problem has no negative eigenvalue.

Case 3. If λ = α2 with α > 0, then

y(x) = c1 cos(αx) + c2 sin(αx).

Immediately, y(0) = c1 = 0, so y(x) = c2 sin(αx). The second boundary

condition is

y(π) + 2y 0 (π)0 = c2 sin(απ) + 2c2 α cos(απ).

To look for nontrivial solutions, suppose c2 6= 0. Then this equation is

sin(απ) = −2α cos(απ)

or
tan(απ) = −2α.
This is a transcendental equation, which cannot be solved by algebraic
manipulations. There are infinitely many positive solutions, however, be-
cause the graphs of y = tan(απ) and y = −2α intersect infinitely often
in the right half-plane. Let the first coordinates of these points of inter-
section be α1 , α2 , · · · , in increasing order. Then the eigenvalues of this
problem are
λn = αn2 .
Using a numerical approximation program, the first four eigenvalues are
approximately

λ1 ≈ 0.48705, λ2 ≈ 2.54914, λ2 ≈ 6.56059, λ4 ≈ 12.56423.

Corresponding eigenfunctions are

p
ϕn (x) = sin( λn x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.1. EIGENVALUES, EIGENFUNCTIONS AND STURM-LIOUVILLE PROBLEMS107

7. The problem is regular on [0, 1]. The analysis to find eigenvalues and
eigenfunctions is similar to that done for Problem 6. Take cases on λ. It
is routine to check that λ = 0 or λ < 0 lead only to the trivial solution,
so the problem has no negative eigenvalue and zero is not an eigenvalue.
If λ = α2 for α > 0, then

y(x) = c1 cos(αx) + c2 sin(αx).

Using the first boundary condition,

y(0) − 2y 0 (0) = 0 = c1 − 2c2 α

so c1 = 2αc2 and

y(x) = 2αc2 cos(αx) + c2 sin(αx).

Now use the boundary condition at 1:

y 0 (1) = −2α2 c2 sin(α) + c2 α cos(α) = 0.

For a nontrivial solution we need to be able to choose c2 nonzero. This

requires that
−2α sin(α) + cos(α) = 0,
or
1
tan(α) =.
2α
Solutions of this equation must be numerically approximated. There are
infinitely many positive solutions α1 < α2 < · · · , and the eigenvalues are
λj = αj2 . The first four eigenvalues are

λ1 ≈ 0.42676, λ2 ≈ 10.8393, λ3 ≈ 40.4702, λ4 ≈ 89.8227.

Corresponding eigenfunctions are

√ p p
ϕn (x) = 2 λ cos( λn x) + sin( λn x).

8. The problem is regular on [0, 1]. The differential equation has character-
istic equation
r2 + 2r + (1 + λ) = 0,
√
with roots r = −1 ± λi. The general solution of the differential equation
is √ √
y(x) = c1 e−x cos( λx) + c2 e−x sin( λx).
Because y(0) = 0 = c1 , we have just
√
y(x) = c2 e−x sin( λx).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 108CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

The boundary condition at 1 requires that

√
y(1) = 0 = c2 e−1 sin( λ).

This can be satisfied with c2 6= 0 if

√
λ = nπ,

a positive zero of the sine function, for n = 1, 2, · · · . The eigenvalues are

therefore
λn = n2 π 2
and the eigenfunctions are

ϕn (x) = e−x sin(nπx).

9. The problem is regular on [0, π]. The differential equation can be written

y 00 + 2y 0 + λy = 0.

and the characteristic equation has roots

√
−1 ± 1 − λ.

Here it is convenient to take cases on 1 − λ.

Case 1. If 1 − λ = 0, then λ = 1 and the general solution is

y(x) = c1 e−x + c2 xe−x .

Now y(0) = c1 = 0, so y(x) = c2 xe−x . And

y(π) = 0 = c2 πe−π = 0

forces c2 = 0, so this case has only the trivial solution and 0 is not an
eigenvalue.

Case 2. If 1 − λ is positive, say 1 − λ = α2 with α > 0, then

y(x) = c1 e(−1+α)x + c2 e(−1−α)x .

Now
y(0) = c1 + c2 = 0
so c2 = −c1 and
 
y(x) = c1 e(−1+α)x − e(−1−α)x .

Then  
y(π) = c1 e(−1+α)π − e(−1−α)π .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.2. EIGENFUNCTION EXPANSIONS 109

If c1 6= 0, this requires that

eαπ = e−απ ,

which is impossible if α > 0. The problem has no negative eigenvalue.

Case 3. If 1 − λ is negative, write 1 − λ = −α2 . Now

y(x) = c1 e−x cos(αx) + c2 e−x sin(αx).

Immediately y(0) = c1 = 0. Next,

y(π) = c2 e−π sin(απ) = 0.

To have c2 6= 0, we must choose

√
α= λ − 1 = n,

any positive integer. Then λ = 1 + n2 , so the eigenvalues are

λn = 1 + n2 for n = 1, 2, · · · .

Eigenfunctions are
ϕn (x) = e−x sin(nx).

10. The problem is regular on [0, 8]. Details are similar to those of Problem 9
and we find eigenvalues
λn = 8 + n2 π 2
for n = 1, 2, · · · , and eigenfunctions

ϕn (x) = e3x sin(nπx).

11. If λn = 1 − 1/n, then the eigenvalues are listed in increasing order, and

lim λn = 0.
n→∞

This is impossible by Theorem 4.1.

4.2 Eigenfunction Expansions

In Problems 1–5, the weight function is p(x) = 1 (read from the differential
equation). In Problem 6 the differential equation must be put into standard
Sturm-Liouville form to read the weight function p(x) as the coefficient of λ.
In graphing partial sums of eigenfunctions and comparing them to the func-
tion, note the differences in the number of terms that must be taken to have the
partial sum fit reasonably close to the function. The convergence theorem does
not give any information about how fast an eigenfunction expansion converges
to the function.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 110CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

1. It is routine to find the eigenfunctions

 nπx 
ϕn (x) = sin
2
for this problem. The expansion has the form
∞
X
cn sin(nπx/2),
n=1

where R2
0
(1 − ξ) sin(nπξ/2) dξ
cn = R2 2 .
0
sin (nπξ/2) dξ
These integrals are Z 2
sin2 (nπξ/2) dξ = 1
0
and
2
2(1 + (−1)n )
Z
(1 − ξ) sin(nπξ/2) dξ = .
0 nπ
The eigenfunction expansion on [0, 2] is
∞
X 2(1 + (−1)n )
sin(nπx/2).
n=1
nπ

Figure 4.1 shows a graph of f (x) = 1 − x and the fortieth partial sum of
this expansion. By the convergence theorem, this expansion converges to
1 − x for 0 < x < 2. Clearly the expansion converges to 0 at both x = 0
and x = 2 because the eigenfunctions vanish there.
2. The problem has eigenfunctions

ϕn (x) = sin((2n − 1)x/2).

The eigenfunction expansion on [0, π] has the coefficients

8 (−1)n+1
cn =
π (2n − 1)2
and the expansion is
∞
X
cn sin((2n − 1)x/2).
n=1

Figure 4.2 shows a graph of the function and the fifth partial sum of this
eigenfunction expansion. Unlike Problem 1, this expansion converges very
rapidly to the function. By the convergence theorem, it converges to x for
0 < x < π. The graph suggests that it converges to x at the endpoints as
well, but this is not given by the theorem.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.2. EIGENFUNCTION EXPANSIONS 111

Figure 4.1: Comparison of 1 − x and the fortieth partial sum of its eigenfunction
expansion on [0, 2].

Figure 4.2: Comparison of x and the fifth partial sum of its eigenfunction ex-
pansion on [0, π].

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 112CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.3: Comparison of f (x) and the sixtieth partial sum of its eigenfunction
expansion on [0, 4].

3. The eigenfunctions are

ϕn (x) = cos((2n − 1)πx/8).
The coefficients in the expansion of f (x) on [0, 4] are
R2 R4
− cos((2n − 1)πx/8) dx + 2 cos((2n − 1)πx/8) dx
cn = 0 R4
0
cos2 ((2n − 1)πx/8) dx
4 h √ i
= (−1)n+1 + 2(cos(nπ/2) − sin(nπ/2)) .
(2n − 1)π
The expansion has the form
∞
X
cn cos((2n − 1)πx/8).
n=1

Figure 4.3 compares f (x) with the sixtieth partial sum of this eigenfunc-
tion expansion. The theorem tells us that the expansion converges to f (x)
on (0, 2) and on(2, 4), as well at to 0 at x = 0 (average of left and right
limits there).
4. The eigenfunctions are
ϕ0 (x) = 1, ϕn = cos(nx) for n = 1, 2, · · · .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.2. EIGENFUNCTION EXPANSIONS 113

Figure 4.4: Comparison of f (x) and the tenth partial sum of its eigenfunction
expansion in Problem 4.

The coefficients in the eigenfunction expansion of f (x) = sin(2x) on [0, π]

are
1 π
Z
c0 = sin(1ξ) dξ = 0,
π 0
2 π
Z
c2 = sin(2ξ) cos(2ξ) dξ = 0,
π 0
and, for n = 1, 3, 4, 5, · · · ,
2 π 4 (−1)n − 1
Z
cn = sin(2ξ) cos(nξ) dξ = .
π 0 π n2 − 4
The eigenfunction expansion is
∞
4 X 4 ((−1)n − 1)
cos(nx).
π π n2 − 4
n=1,n6=2

Figure 4.4 compares a graph of f (x) with the tenth partial sum of this
eigenfunction expansion. The theorem tells us that this expansion con-
verges to sin(2x) for 0 < x < π.
5. The eigenfunctions are

ϕ0 (x) = 1, ϕn (x) = an cos(nx/3) + bn sin(nx/3)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 114CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.5: Comparison of f (x) and the fifth partial sum of its eigenfunction
expansion in Problem 5.

for n = 1, 2, · · · . The coefficients in the eigenfunction expansion of x2 on

[−3π, 3π] are
Z 3π
1
a0 = ξ 2 dξ = 3π 2 ,
6π −3π
1 2 36
an = ξ cos(nξ/3) dξ = 2 (−1)n ,
3π n
Z 3π
1
bn = ξ 2 sin(nξ/3) dξ = 0.
3π −3π

The eigenfunction expansion is

∞
2
X (−1)n
3π + 36 cos(nx/3).
n=1
n2

Figure 4.5 shows f (x) and the fifth partial sum of this eigenfunction expan-
sion. By the theorem, the expansion converges to x2 for −3π < x < 3π.

6. The eigenfunctions are

ϕn (x) = e−x sin(nπx)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.2. EIGENFUNCTION EXPANSIONS 115

Figure 4.6: Comparison of f (x) and the eightieth partial sum of its eigenfunction
expansion in Problem 6.

for n = 1, 2, · · · . Notice that the differential equation of the problem can

written in standard Sturm-Liouville form as

0
e2x y 0 + (1 + λ)e2x y = 0,

with the coefficient of λ equal to e2x . Therefore, in the expansion of a

function in terms of these eigenfunctions, the weight function p(x) = e2x
must be used. With this in mind, the coefficients in this expansion are
R1
1/2
e2ξ e−ξ sin(nπξ) dξ
cn = R 1
0
e2ξ e−2ξ sin2 (nπξ) dξ
R1 ξ
1/2
e sin(nπξ) dξ
= R1 2
0
sin (nπξ) dξ
1/2
2e (nπ cos(nπ/2) − sin(nπ/2)) − 2enπ(−1)n
= .
1 + n2 π 2
Figure 4.6 shows a graph of the function and the eightieth partial sum of
this eigenfunction expansion.

7. Recall that the complex conjugate of z = a + ib is z = a − ib. Suppose λ is

an eigenvalue of a Sturm-Liouville problem, with eigenfunction ϕ(x). By

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 116CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

taking the complex conjugate of the Sturm-Liouville differential equation

and appropriate boundary conditions, it is routine to check that λ is also
an eigenvalue with eigenfunction ϕ(x). If λ is complex and not real, then
λ 6= λ, so the eigenfunctions must be orthogonal with respect to the weight
function p, and
Z b
p(x)ϕ(x)ϕ(x) dx = 0.
a
Now,
ϕ(x)ϕ(x) = |ϕ(x)|2 ,
so Z b
p(x)|ϕ(x)|2 | dx = 0.
a
This is impossible because p(x) > 0 on (a, b) and ϕ(x) is continuous and
not identically zero on the interval. This contradiction shows that λ = λ,
so λ is real.

4.3 Fourier Series

1. The Fourier series of f (x) = 4 on [−3, 3] has the form
∞
1 X
a0 + [an cos(nπx/3) + bn sin(nπx/3)].
2 n=1

All that is left is to compute the coefficients. First, because f (x) is an

even function, each bn = 0. Compute
2 3
Z
a0 = 4 dx = 8
3 −3
and, for n = 1, 2, · · · ,
Z 3
2
an = 4 cos(nπx/3) dξ = 0.
3 0

With each an = 0 for n = 1, 2, · · · , the Fourier series is

1
a0 = 4.
2
This series consists of a single term, namely the constant term (which
seems obvious by hindsight, if not noticed immediately). This one-term
series converges to 4 on [−3, 3].
2. Because f (x) is an odd function on [−1, 1], each an = 0. Compute the
Fourier coefficients
Z 1
2
bn = 2 −x sin(nπx) dx = (−1)n .
0 nπ

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 117

Figure 4.7: Comparison of f (x) and the thirtieth partial sum of the Fourier
series in Problem 2.

The Fourier series of −x on [−1, 1] is

∞
2 X (−1)n
sin(nπx).
π n=1 n

This series converges to −x for −1 < x < 1, and to 0 at x = −1 and at

x = 1. The latter is consistent with the convergence theorem because

lim (−x) + lim (−x) = 1 − 1 = 0.

x→−1+ x→1−

Figure 4.7 is a graph of f (x) and the thirtieth partial sum of the Fourier
series.

3. Because cosh(πx) is an even function, each bn = 0 in the Fourier series,

which will have the appearance
∞
1 X
a0 + an cos(nπx).
2 n=1

Compute
Z 1
2
a0 = 2 cosh(πx) dx sinh(π)
0 π

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 118CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.8: Comparison of f (x) and the eighth partial sum of the Fourier series
in Problem 3.

and, for n = 1, 2, · · · ,
Z 1
2 sinh(π) (−1)n
an = 2 cosh(πx) cos(nπx) dx = .
0 π 1 + n2
The Fourier series is
∞
1 X 2 sinh(π) (−1)n
sinh(π) + cos(nπx).
π n=1
π 1 + n2

This series converges to cosh(πx) for −1 ≤ x ≤ 1. Figure 4.7 shows a

graph of f (x) and the eighth partial sum of this Fourier series.
4. Omitting the routine integrations, the Fourier series of 1 − |x| on [−2, 2]
is
∞
8 X 1
cos((2n − 1)πx/2).
π 2 n=1 (2n − 1)2
This series converges to 1 − |x| for −2 ≤ x ≤ 2. Figure 4.9 shows a graph
of f (x) and the fifth partial sum of this Fourier series.
5. The Fourier series of f (x) on [−π, π] is
∞
16 X 1
sin((2n − 1)x).
π n=1 (2n − 1)2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 119

Figure 4.9: Comparison of f (x) and the fifth partial sum of the Fourier series
in Problem 4.

This series converges to


−4
 for −π < x < 0,
4 for 0 < x < π,

0 for x = 0, −π, π.


Figure 4.10 is a graph of f (x) and the twentieth partial sum of this Fourier
series.

6. Because f (x) = sin(2x) is periodic of period π, this function is its own

Fourier series on this [−π, π].

7. The Fourier series of f (x) on [−2, 2] is

∞  
13 X n 16 1
+ (−1) cos(nπx/2) + sin(nπx/2) .
3 n=1
(nπ)2 nπ

This series converges to

(
x2 − x + 3 for −2 < x < 2,
7 for x = ±2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 120CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.10: Comparison of f (x) and the twentieth partial sum of the Fourier
series in Problem 5.

At the endpoints,

1 1
(f (−2+) + f (2−)) = (9 + 5) = 7.
2 2

Figure 4.11 shows f (x) and the twentieth partial sum of this Fourier series.

8. The Fourier series of f (x) on [−5, 5] is

∞
1 X
a0 + [an cos(nπx/5) + bn sin(nπx/5)],
2 n=1

where
Z 5
1 71
a0 = f (x) dx = ,
5 −5 6

1 5
Z
an = f (x) cos(nπx/5) dx
5 −5
5((−1)n − 1) 50(−1)n
= +
n2 π 2 n2 π 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 121

Figure 4.11: Comparison of f (x) and the twentieth partial sum of the Fourier
series in Problem 7.

and

1 5
Z
bn = f (x) sin(nπx/5) dx
5 −5
5(−1)n 1
= − 3 3 (50 − n2 π 2 − 50(−1)n + 26n2 π 2 (−1)n ).
nπ n π
This series converges to


x for −5 < x < 0,
1 + x2

for 0 < x < 5,
1/2
 for x = 0,

31/2 at x = ±5.


Figure 4.12 shows f (x) and the fiftieth partial sum of this Fourier series
on [−5, 5].

9. The Fourier series of f (x) on [−π, π] is

∞
3 2X 1
+ sin((2n − 1)x).
2 π n−1 2n − 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 122CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.12: Comparison of f (x) and the fiftieth partial sum of the Fourier
series in Problem 8.

This series converges to


1
 for −π < x < 0,
2 for 0 < x < π,

3/2 for x = 0, −π, π,


Figure 4.13 shows the function and the thirtieth partial sum of the Fourier
series in Problem 9.

10. The Fourier series of f (x) on [−π, π] is

∞
2 4 X (−1)n
− cos(nx).
π π n=1 4n2 − 1

This converges to cos(x/2) − sin(x) for −π < x < π, and to 0 at x = −π

and at x = π. Figure 4.14 is a graph of f (x) and the fourth partial sum
of the Fourier series.

11. The Fourier series of cos(x) on [−3, 3] is

∞
1 X (−1)n+1
sin(3) + 6 sin(3) cos(nπx/3).
3 n=1
π 2 n2 − 9

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 123

Figure 4.13: Comparison of f (x) and the thirtieth partial sum of the Fourier
series in Problem 9.

Figure 4.14: Comparison of f (x) and the fourth partial sum of the Fourier series
in Problem 10.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 124CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.15: Comparison of f (x) and the fifth partial sum of the Fourier series
in Problem 11.

This converges to cos(x) on [−3, 3]. Figure 4.15 is a graph of f (x) and the
fifth partial sum of this Fourier expansion on [−3, 3].
It might seem at first that cos(x) should be its own Fourier expansion,
but this problem illustrates the importance of the interval. If you expand
cos(x) in a Fourier series on [−π, π], you obtain just cos(x). But this is
not the expansion on [−3, 3].

12. The Fourier expansion of f (x) on [−1, 1] is

∞ 
3 X 1 − (−1)n 1 − 2(−1)n

− cos(nπx) + sin(nπx) .
4 n=1 n2 π 2 nπ

This converges to


1−x for −1 < x < 0,

0 for 0 < x < 1,


1/2 at x = 0,
1 for x = ±1.


Figure 4.16 is a graph of f (x) and the thirtieth partial sum of this Fourier
series.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 125

Figure 4.16: Comparison of f (x) and the thirtieth partial sum of the Fourier
series in Problem 12.

13. The Fourier series of f (x) on [−3, 3] converges to



3/2 for x = ±3,

2x for −3 < x < −2,





−2 for x = −2,


0 for −2 < x < 1,
1/2 for x = 1,





 2
x for 1 < x < 3.

14. The Fourier series converges to




1 for x = ±1 and for 1/2 < x < 3/4,
0 for −1 < x < 1/2,



2 for 3/4 < x < 1,

1/2 for x = 1/2,





3/2 for x = 3/4.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 126CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

15. The Fourier series converges to



 −1 for x = ±4,

3/2 for x = −2,


 5/2 for x = 2,
f (x) for all other x in [−4, 4].


16. Suppose f (x) is an even function on [−L, L]. In the Fourier series for f (x)
on [−L, L], the sine terms have coefficients

1 L
Z
bn = f (x) sin(nπx/L) dx
L −L
Z 0 Z L !
1
= f (x) sin(nπx/L) dx + f (x) sin(nπx/L) dx .
L −L 0

In the next to last integral in this equation, let x = −ξ. Using the fact
that f (x) is an even function, and sin(−nπx/L) = − sin(nπx/L) then
Z 0
f (x) sin(nπx/L) dx
−L
Z 0
= f (−ξ) sin(−nπξ/L)(−1) dξ
L
Z L
=− f (ξ) sin(nπξ/L) d/xi.
0

Therefore each bn = 0.
For the cosine terms, the coefficients are

1 L
Z
an = f (x) cos(nπx/L) dx
L −L
Z 0 Z L !
1
= f (x) cos(nπx/L) dx + f (x) cos(nπx/L) dx .
L −L 0

In the integral from −L to 0, make the change of variables x = −ξ, except

now note that cos(nπx/L) = cos(−nπx/L) to show that

2 L
Z
an = f (x) cos(nπx/L) dx.
L 0

17. The argument is like that used in Problem 16, except now use the fact
that f (−x) = −f (x).
18. Define
(x) + f (−x)
fe (x) =
2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 127

and
f (x) − f (−x)
fo (x) = .
2
It is routine to check that fe (x) is an even function and fo (x) is odd on
[−L, L].

19. Suppose f (x) is both even and odd on [−L, L]. Then, for every x in this
interval,
f (x) = f (−x) = −f (x).
But then f (x) = 0, so f (x) is identically zero on the interval.

20. The Fourier cosine series for f (x) = 4 on [0, 3] is just the constant 4,
converging to 4 at each point of the interval.
The sine expansion of 4 on [0, 3] is
∞
X
Bn sin(nπx/3),
n=1

where Z 3
2 8
Bn = 4 sin(nπx/3) dx = (1 − (−1)n ).
3 0 nπ
Because (
2 n if n is odd,
1 − (−1) =
0 if n is even,
this sine series can also be written
16 X ∞ 1
n =1 sin((2n − 1)πx/3).
π 2n − 1
This sums over just the odd positive integers.
The sine expansion converges to
(
4 for 0 < x < 4,
0 for x = 0 and for x = 3.

21. The cosine series is

∞
4 X (−1)n
− cos((2n − 1)πx/2).
π n=1 2n − 1

This converges to 
1
 for 0 ≤ x < 1,
0 for x = 1,

−1 for 1 < x ≤ 2.


© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 128CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.17: Comparison of f (x) and the thirtieth partial sum of the Fourier
cosine series in Problem 21.

Figure 4.17 compares the function to the thirtieth partial sum of this
cosine series.
The sine series is
∞
2X
(1 + (−1)n − 2 cos(nπ/2)) sin(nπx/2),
π n=1

which converges to

1
 for 0 < x < 1,
0 for x = 0, 1, 2,

−1 for 1 < x < 2.


Figure 4.18 shows f (x) and the seventieth partial sum of this sine expan-
sion on [0, 2].

22. The cosine series is

∞
1 X 2n sin(nπ/2)
cos(x) + cos(nx/2).
2 π(n2 − 4)
n=1,n6=2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 129

Figure 4.18: Comparison of f (x) and the seventieth partial sum of the Fourier
sine series in Problem 21.

This converges to 

0 for 0 ≤ x < π,

−1/2 for x = π,


cos(x) for π < x < 2π,
1 for x = 2π,


Graphs of the function and the fifteenth partial sum of this cosine series
are shown in Figure 4.19.
The sine series is
∞
2 X 2n
− sin(x/2) − 2 − 4)π
((−1)n + cos(nπ/2)) sin(nx/2).
3π n=3
(n

This converges to 

0 for 0 ≤ x < π,

−1/2 for x = π,


cos(x) for π < x < 2π,
0 for x = 2π.


Figure 4.20 shows the function and the fortieth partial sum of its sine
expansion in Problem 22.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 130CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.19: Comparison of f (x) and the fifteenth partial sum of the cosine
series in Problem 22.

Figure 4.20: Comparison of f (x) and the fortieth partial sum of the sine expan-
sion in Problem 22.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 131

23. The cosine series is

∞
8 X 1
1− cos((2n − 1)πx),
π n=1 (2n − 1)2
2

converging to 2x for 0 ≤ x ≤ 1.
The sine series is
∞
4 X (−1)n
− sin(nπx),
π n=1 n
converging to 2x for 0 < x < 1 and to 0 for x = 0 and for x = 1.

24. The cosine series is

∞
4 10 X (−1)n
+ cos(nπx/2),
3 π 2 n=1 n2

converging to x2 for 0 ≤ x ≤ 2.
The sine expansion is
∞ 
8 X (−1)n 2(1 − (−1)n )

− + sin(nπx/2).
π n=1 n n3 π 2

This converges to x2 for 0 ≤ x < 2 and to 0 at x = 2.

25. The cosine series is

∞
X 1 − (−1)n e−1
−1 − e−1 + 2 cos(nπx),
n=1
1 + n2 π 2

converging to e−x for 0 ≤ x ≤ 1.

The sine series is
∞  
X n n −1
2π (1 − (−1) e ) sin(nπx),
n=1
1 + n2 π 2

converging to e−x for 0 < x < 1 and to 0 at x = 0 and at x = 1.

26. The cosine expansion of f (x) is

∞
1 X
+ An cos(nπx/3),
2 n=1

where
−6(1 + (−1)n ) + 4nπ sin(2nπ/3) + 12 cos(2nπ/3)
An = .
n2 π 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 132CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.21: f (x) and the fortieth partial sum of the cosine series in Problem
26.

This cosine series converges to


x
 for 0 ≤ x < 2,
1 for x = 2,

2−x for 2 < x ≤ 3.


This fortieth partial sum of this cosine expansion of f (x) is shown in

Figure 4.21.
The sine expansion is
∞
X
Bn sin(nπx/3),
n=1
where
1 + sin(2nπ/3) − 4nπ cos(2nπ/3) + 2nπ(−1)n
Bn = .
n2 π 2
This sine series converges to


 x for 0 ≤ x < 2,

1 for x = 2,


 2−x for 2 < x < 3,
0 for x = 3.


© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 133

Figure 4.22: f (x) and the thirtieth partial sum of the sine series in Problem 26.

Figure 4.22 shows the function and the thirtieth partial sum of its sine
expansion on [0, 3].

27. The cosine expansion is

∞
1 4X1
− + cos(nπ/5) sin(2nπ/5) cos(nπx/5),
5 π n=1 n

converging to



1 for 0 ≤ x < 1,
1/2 for x = 1,



0 for 1 < x < 3,

−1/2 for x = 3,





−1 for 3 < x < 5.

Figure 4.23 shows the function and the sixtieth partial sum of this cosine
expansion.
The sine series is
∞
4X 1
(1 + (−1)n − 2 cos(nπ/5) cos(2nπ/5)) sin(nπx/5),
π n=1 2n

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 134CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.23: f (x) and the sixtieth partial sum of the cosine series in Problem
27.

converging to



1 for 0 < x < 1,
1/2 for x = 1,



0 for 1 < x < 3, x = 0, or x = 5,

−1/2

 for x = 3,


−1 for 3 < x < 5.

Figure 4.24 shows f (x) and the one hundredth partial sum of its sine
expansion on [0, 5].
28. The cosine series is
∞  
5 16 X 1 4
+ cos(nπ/4) − 3 sin(nπ/4) cos(nπx/4),
6 π 2 n=1 n2 n π

converging to x2 for 0 ≤ x ≤ 1 and to 1 for 1 ≤ x ≤ 4.

Figure 4.25 shows the function and the tenth partial sum of this cosine
expansion.
The sine series is
∞ 
2(−1)n

X 16 64
sin(nπ/2) + 3 3 (cos(nπ/4) − 1) − sin(nπx/4),
n=1
n2 π 2 n π nπ

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 135

Figure 4.24: Comparison of f (x) and the hundredth partial sum of the sine
series in Problem 27.

Figure 4.25: f (x) and the tenth partial sum of the Fourier cosine series in
Problem 28.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 136CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

Figure 4.26: f (x) and the eightieth partial sum of the sine series in Problem 28.

and this converges to x2 for 0 ≤ x ≤ 1, to 1 if 1 ≤ x < 4, and to 0 at

x = 4.
Figure 4.26 shows f (x) and the eightieth partial sum of its sine expansion
on [0, 4].

29. The cosine series is

∞  
24 X 1 n 4 n
−1 − 2 2(−1) + 2 2 (1 − (−1) ) cos(nπx/3),
π n=1 n2 n π

converging to 1 − x2 for 0 ≤ x ≤ 2.
The sine series is
∞  
2X1 48
1 + 7(−1)n − 2 2 sin(nπx/2).
π n=1 n n π

This series converges to 1 − x2 for 0 < x < 2 and to 0 at x = 0 and x = 2.

30. It is routine to determine the Fourier sine expansion of sin(x) on [0, π]:
∞
2 4X 1
− cos(2nx).
π π n=1 4n2 − 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 4.3. FOURIER SERIES 137

This converges to sin(x) for 0 ≤ x ≤ π. Put x = π/2 in this expansion to

obtain
∞
2 4 X (−1)n
sin(π/2) = 1 = −
π π n=1 4n2 − 1

because cos(nπ) = (−1)n . Solve this for the series in question to obtain
∞
X (−1)n 1 π
2−1
= − .
n=1
4n 2 4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 138CHAPTER 4. STURM-LIOUVILLE PROBLEMS AND EIGENFUNCTION EXPANSIONS

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 5

The Heat Equation

5.1 Diffusion Problems on a Bounded Medium

For the first three problems, separate the variables by letting u(x, t) = X(x)T (t).
The boundary conditions u(0, t) = u(L, t) = 0 leads to eigenvalues

n2 π 2
λn =
L2
for the separation constant, and corresponding eigenfunctions

Xn (x) = sin(nπx/L).

Corresponding solutions for T are

2
π 2 kt/L2
Tn (t) = e−n .

Solutions of these problems therefore all have the form

∞
2
π 2 t/L2
X
u(x, t) = cn sin(nπx/L)e−kn ,
n=1

in which cn is determined by the initial condition u(x, 0) = f (x) by

Z L
2
cn = f (ξ) sin(nπξ/L) dξ.
L 0

Therefore, for these problems, all we need do is evaluate these integrals for the
coefficients.

1. With f (x) = x(L − x),

L
4L2
Z
2
cn = ξ(L − ξ) sin(nπξ/L) dξ = (1 − (−1)n ).
L 0 n3 π 3

139

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 140 CHAPTER 5. THE HEAT EQUATION

The solution is
∞
X 4L2 2 2 2
u(x, t) = 3 π3
(1 − (−1)n ) sin(nıx/L)e−kn π t/L .
n=1
n

We can also observe that

(
n 0 if n is even,
1 − (−1) =
2 if n is odd,

so the solution can also be written by summing over just the odd positive
integers. This can be done by replacing n with 2n − 1 in the summation:
∞
8L2 X 1 2 2 2
u(x, t) = 3 3
sin((2n − 1)πx/L)e−k(2n−1) π t/L .
π n=1 (2n − 1)

2. Now k = 4 and u(x, 0) = f (x) = x2 (L − x), so the coefficients are

L
4L3
Z
2
cn = ξ 2 (L − ξ) sin(nπξ/L) dξ = − (1 + 2(−1)n ).
L 0 n3 π 3
The solution is
∞ 
4L3 X 1 + 2(−1)n

2 2 2
u(x, t) = sin(nπx/L)e−4n π t/L .
π 3 n=1 n3

3. Here k = 3 and f (x) = L(1 − cos(2πx/L)). Compute

8((−1)n −1)
(
2 L if n 6= 2,
Z
nπ(n2 −4)
cn = L(1 − cos(2πξ/L)) sin(nπξ/L) dξ =
L 0 0 for n = 2,

Because (−1)n − 1 = 0 if n is even, and −2 if n is odd, we actually have

16L
cn = −
nπ(n2 − 4)

for n = 1, 3, 5, · · · . We can write the solution as

∞
16L X 1 2 2 2
u(x, t) = − sin((2n−1)πx/L)e−3(2n−1) π t/L .
π n=1 (2n − 1)((2n − 1)2 − 4)

Problems 4–7 have insulated boundary conditions, so separation of variables

by putting u(x, t) = X(x)T (t) leads to eigenvalues and eigenfunctions λ0 =
1, X1 = 1 and, for n = 1, 2, · · · ,

n2 π 2
λn = , Xn (x) = cos(nπx/L).
L2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.1. DIFFUSION PROBLEMS ON A BOUNDED MEDIUM 141

We also find that 2

π 2 t/L2
Tn (t) = e−kn
as in the case of boundary conditions u(0, t) = u(L, t) = 0. Now the solution
has the form
∞
1 X 2 2 2
u(x, t) = c0 + cn cos(nπx/L)e−kn π t/L ,
2 n=1

where
2
cn = f (ξ) cos(nπξ/L) dξ.
L
4. Now k = 1, f (x) = sin(x) and L = π. The coefficients in the series for the
solution are
2 4
c0 = sin(ξ) dξ =
π π
and, for n = 1, 2, · · · ,
( n
2 π − π2 1+(−1) if n 6= 1,
Z
n2 −1
cn = sin(ξ) cos(nξ) dξ =
π 0 0 if n = 1.
Because 1 + (−1)n = 0 if n is odd, each c2n−1 = 0 and we need sum over
only even n. For n even
4 1
cn = − .
π n2 − 1
The solution is
∞
2 4X 1 2
u(x, t) = − cos(2nx)e−4n t .
π π n=1 4n2 − 1
Here we let n = 1, 2, · · · in the summation, but replaced n with 2n in the
terms of the series to sum over all the even positive integers.
5. Now k = 4, L = 2π and f (x) = x(2π − x)2 . The coefficients are
1 2π
Z
4 3
c0 = ξ(2π − ξ)2 dξ = π
π 0 3
and, for n = 1, 2, · · · ,
Z 2π
1
cn = ξ(2π − ξ)2 cos(nξ/2) dξ
π 0
16 2 2
=− (n π − 6(1 − (−1)n )).
πn4
The solution is
2 3
u(x, t) = π
3
∞
X 16 n2 π 2 − 6(1 − (−1)n ) 2
− cos(nx/2)e−n t .
n=1
π n4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 142 CHAPTER 5. THE HEAT EQUATION

6. Now L = 3, k = 4 and f (x) = x sin(πx). Compute

2 3
Z
2
c0 = ξ sin(πξ) dξ = ,
3 0 π

2 3 18(−1)n
Z
cn = ξ sin(πξ) cos(nπξ/3) dξ = −
3 0 π(n2 − 9)
for n = 1, 2, 4, 5, · · · , and, for n = 3,
2 3
Z
1
c3 = ξ sin(πξ) cos(πξ) dξ = − .
3 0 2π
The solution is
1 1 2
u(x, t) = − cos(πx)e−4π t
π 2π
∞
X 18(−1)n+1 −4n2 π2 t/9
+ e .
π(n2 − 9)
n=1,n6=3

7. In this problem L = 6, k = 2 and f (x) = x cos(πx/4). The coefficients in

the solution are
1 6
Z
8 2 + 3π
c0 = ξ cos(πξ/4) dξ = − ,
3 0 3 π2
and, for n = 1, 2, · · · ,
1 6
Z
cn = ξ cos(πξ/4) cos(nπξ/6) dξ
6 0
24(−18 − 8n2 − 27π(−1)n + 12πn2 (−1)n )
= .
π 2 (2n − 3)2 (2n + 3)2
The solution is
4 2 + 3π
u(x, t) = −
3 π2
∞
X 2 2
+ cn cos(nπx/6)e−n π t/18 .
n=1

8. The initial-boundary value problem modeling this setting is

ut = kuxx for 0 < x < L, t > 0,
ux (0, t) = ux (L, t) = 0,
u(x, 0) = B.
The coefficients in the series solution are
2 L
Z
c0 = B dξ = 2B
L 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.1. DIFFUSION PROBLEMS ON A BOUNDED MEDIUM 143

and, for n = 1, 2, · · · ,
Z L
2
cn = B cos(nπξ/L) dξ = 0.
L 0

The solution is
u(x, t) = B.
This is consistent with intuition - with no energy loss, the bar maintains
a constant temperature.
9. The initial-boundary value problem for the temperature function is

ut = uxx for 0 < x < L, t > 0,

u(0, t) = ux (L, t) = 0,
Bx
u(x, 0) = .
L
Separate variables by putting u(x, t) = X(x)T (t) to obtain

X 00 + λX = 0; X(0) = X 0 (L) = 0

and
T 00 + λkT = 0.
By taking cases on λ, we find the eigenvalues and corresponding eigen-
functions:
 2
(2n − 1)π
λn = and Xn (x) = sin((2n − 1)πx/2L).
2L
Further,
2
π 2 t/4L2
Tn (x) = e−k(2n−1) .
The solution has the form
∞
2
π 2 t/4L2
X
u(x, t) = cn sin((2n − 1)πx/2L)e−k(2n−1) .
n=1

The coefficients are

Z L
2 B
cn = ξ sin((2n − 1)πξ/2L) dξ
L
0 L
−8B
= 2 (−1)n .
π (2n − 1)2
The solution is
∞
8B X (−1)n 2 2 2
u(x, t) = − sin((2n − 1)πx/2L)e−k(2n−1) π t/4L .
π 2 n=1 (2n − 1)2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 144 CHAPTER 5. THE HEAT EQUATION

10. The initial-boundary value problem for the temperature function is

ut = 9uxx for 0 < x < 2, t > 0,
u(0, t) = ux (2, t) = 0,
u(x, 0) = x2 .
A routine separation of variables leads to eigenvalues and eigenfunctions
n2 π 2
λn = , Xn (x) = sin((2n − 1)πx/4)
4
and solutions for the time-dependent parts as
2
π 2 t/4
Tn (t) = e−9n .
The solution has the form
∞
2
π 2 t/4
X
u(x, t) = cn sin((2n − 1)πx/4)e−9n ,
n=1

where
Z 2
cn = ξ 2 sin((2n − 1)πξ/4) dξ
0
64 2 + (2n − 1)π(−1)n
=− .
π3 (2n − 1)3

11. Make the transformation u(x, t) = eαx+βt v(x, t). Following the discussion
of the text, let α = −A/2 = −4/2 = −2 and β = k(B − A2/4) = −2 also,
so
u(x, t) = e−2x−2t v(x, t)
and v is the solution of the problem
vt = vxx for 0 < x < π, t > 0,
v(0, t) = v(π, t) = 0,
v(x, 0) = e2x u(x, 0) = xe2x (π − x).
This has the solution
∞
X t
v(x, t) = cn sin(nx)e−n t ,
n=1

where
Z π
2
cn = ξe2ξ (π − ξ) sin(nξ) dξ
π 0
4
24n − 2n3 + 16nπ + 4n3 π − 24ne2π (−1)n

=− 2 3
π(4 + n )
+2e2π n3 (−1)n + 16nπe2π (−1)n + 4n3 πe2π (−1)n .


© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.1. DIFFUSION PROBLEMS ON A BOUNDED MEDIUM 145

The solution of the original problem is

u(x, t) = e−2x−2t v(x, t).

12. Again following the discussion in the text, we have A = 6, B = 0, k =

1, L = 4 and u(x, 0) = f (x) = 1. Find that α = −3 and β = −9 to let

u(x, t) = e−3x−9t v(x, t).

The v(x, t) is the solution of the standard problem

vt = vxx for 0 < x < 4, t > 0,

v(0, t) = v(4, t) = 0,
v(x, 0) = e3x f (x) = e3x .

This problem has the solution

∞
2
π 2 t/16
X
v(x, t) = cn sin(nπx/4)e−n ,
n=1

where
Z 4
1
cn = e3ξ sin(nπξ/4) dξ
2 0
2nπ
= (1 − e12 (−1)n ).
144 + n2 π 2
The original problem has the solution
∞
2
π 2 t/16
X
u(x, t) = e−3x−9t v(x, t) = e−3x−9t cn sin(nπx/4)e−n .
n=1

13. Here we have A = 6, B = 0, k = 1, L = π and u(x, 0) = f (x) = x(π − x).

Let α = 3 and β = −9 and let

u(x, t) = e3x−9t v(x, t).

The v(x, t) satisfies

vt = vxx for 0 < x < π, t > 0,

v(0, t) = v(π, t) = 0,
v(x, 0) = e−3x f (x) = x(π − x)e−3x .

The solution of this problem is

∞
X 2
v(x, t) = cn sin(nx)e−n t ,
n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 146 CHAPTER 5. THE HEAT EQUATION

where
Z π
2
cn = e−3ξ ξ(π − ξ) sin(nξ) dξ
π 0
4n
= (1 − (−1)n e−3π )(3π(n2 + 9) + n2 − 27).
π(n2 + 9)3
The original problem has the solution
v(x, t) = e3x−9t v(x, t).

14. Because of the nonhomogeneous boundary conditions, we cannot solve this

problem by a standard separation of variables. Transform the problem by
letting
u(x, t) = v(x, t) + ψ(x),
where Lψ(x) must be chosen so that we have a problem for v(x, t) that
we know how to solve. Substitute u(x, t) into the initial-boundary value
problem to obtain
vt = 16vxx + 16ψ 00 (x) for 0 < x < 1, t > 0,
v(0, t) + ψ(0) = u(0, t) = 2, v(1, t) + ψ(1) = u(x, t) = 5,
v(x, 0) + ψ(x) = u(x, 0) = x(1 − x)2 .
First, simplify the differential equation for v by setting ψ 00 (x) = 0, so
ψ(x) = cx + d.
Now, we will have v(0, t) = 0 if ψ(0) = 2, and v(1, t) = 0 if ψ(1) = 5.
Therefore choose d = 2 and c = 3, so
ψ(x) = 3x + 2.
The problem for v is
vt = 16vxx for 0 < x < 1, t > 0,
v(0, t) = v(1, t) = 0,
v(x, 0) = u(x, 0) − ψ(x) = sin(πx) − 3x − 2.
This problem has the solution
∞
2
π2 t
X
v(x, t) = cn sin(nπx)e−16n ,
n=1

where
Z 1
cn = 2 (sin(πξ) − 3ξ − 2) sin(nπξ) dξ
0
−1 + 10(−1)n
=
nπ

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.1. DIFFUSION PROBLEMS ON A BOUNDED MEDIUM 147

if n = 2, 3, · · · , and
Z 1
−14 + π
c1 = 2 (sin(πξ) − 3ξ − 2) sin(πξ) dξ = .
0 π

This determines v(x, t), and u(x, t) = v(x, t) + 3x + 2.

15. Let u(x, t) = v(x, t) + ψ(x). To get a standard problem for v(x, t), choose
ψ(x) so that ψ 00 = 0 and

ψ(0) = T, ψ(L) = 0.

Then
T
ψ(x) = (L − x)
L
The problem for v is

vt = kvxx for 0 < x < L, t > 0,

v(0, t) = v(L, t) = 0,
T
v(x, 0) = u(x, 0) − ψ(x) = x(L − x)2 − (L − x).
L
This has the solution
∞
2
π 2 t/L2
X
v(x, t) = cn sin(nπx/L)e−kn ,
n=1

where
Z L 
2 T
cn = ξ(1 − ξ)2 − (L − ξ) sin(nπx/L) dξ
L 0 L
2  2 2
−n π T + 4L3 + 2L3 (−1)n .

= 3 3
n π

16. Here we have k = 4, L = 9 and u(x, 0) = x sin((9 − x)π). Let

v(x, t) = ekAt v(x, t)

and the problem for v(x, t) is

vt = 4vxx for 0 < x < 9, t > 0,

v(0, t) = v(9, t) = 0,
v(x, 0) = u(x, 0) = x sin((9 − x)π).

The solution for v(x, t) is

∞
2
π 2 t/81
X
v(x, t) = cn sin(nπx/9)e−4n ,
n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 148 CHAPTER 5. THE HEAT EQUATION

where
Z 9
2
cn = ξ sin((9 − ξ)π) sin(nπξ/9) dξ
9 0
324n(1 + (−1)n )
=−
π 2 (n − 9)2 (n + 9)2
for n 6= 9, and
Z 9
2 9
c9 = ξ sin((9 − ξ)π) sin(πξ) dξ = .
9 0 2
The solution of the original problem is
u(x, t) = ekAt v(x, t).

17. Let u(x, t) = v(x, t) + h(x) and substitute this into the initial-boundary
value problem to choose h(x) and obtain a standard problem for v(x, t).
We find that  x
h(x) = T 1 −
L
and the problem for v(x, t) is
vt = 9vxx for 0 < x < L, t > 0,
v(0, t) = v(L, t) = 0,
 x
v(x, 0) = −T 1 − .
L
This has the solution
∞
2
π 2 t/L2
X
v(x, t) = cn sin(nπx/L)e−9n ,
n=1

where Z L  
2 ξ 2T
cn = −T 1 − sin(nπξ/L) dξ = − .
L 0 L nπ
Then
∞
 x  2T X 1 2 2 2
u(x, t) = T 1 − − sin(nπx/L)e−9n π t/L .
L π n=1 n

18. The diffusion equation in this case is

dη
ut = duxx − ux .
L
Let u(x, t) = eαx+βt v(x, t) and solve for α and β to obtain a standard
problem for v(x, t). We find that
dη d3 η 2
α= and β = − 2 .
2L 4L

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.2. THE HEAT EQUATION WITH A FORCING TERM F (X, T ) 149

The problem for v(x, t) is

vt = dvxx for 0 < x < L, t > 0,

v(0, t) = v(L, t) = 0,
kL −dηx/2L  
v(x, 0) = e−dηx/2L u(x, 0) = e 1 − e−η(1−x/L) .
dη

The solution for v(x, t) is

∞
2
π 2 t/L2
X
v(x, t) = cn sin(nπx/L)e−dn ,
n=1

where
Z L
2 kL −dηξ/2L  
cn = e 1 − e−η(1−ξ/L) dξ
L0 dη
4Lk  
= 2 2 d − 2 − de−η + 2e−dη/2
d η (d − 2)
if d 6= 2. If d = 2, then
Lk −η
cn = − (e + 1 + η).
η2
This determines v(x, t) and therefore u(x, t).

5.2 The Heat Equation With a Forcing Term

F (x, t)
In Problems 1–5, notation of the text is used for Bn (t), bn and Tn (t). Note that
the second term in the solution for u(x, t) is the solution to the problem without
the forcing.

1. Here k = 4, L = π, f (x) = x(π − x) and F (x, t) = t. We need

2 π
Z
2t
Bn (t) = t sin(nξ) dξ = (1 − (−1)n ),
π 0 nπ
2 π
Z
4
bn = f (ξ) sin(nξ) dξ = (1 − (−1)n ),
π 0 nπ 3
and
Z t
2 2
Tn (t) = e−4n (t−τ )
Bn (τ ) dτ + bn e−4n t
0
1 2
= (1 − (−1)n )(−1 + 4n2 t + e−4n t ).
8πn5

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 150 CHAPTER 5. THE HEAT EQUATION

The solution is
∞
X 1 2
u(x, t) = 5
(1 − (−1)n )(−1 + 4n2 t + e−4n t ) sin(nx)
n=1
8πn
∞
X 4 2
+ 3
(1 − (−1)n ) sin(nx)e−4n t .
n=1
πn

2. Compute
Z 4
1 8
Bn (t) = ξ sin(t) sin(nπξ/4) dξ = (−1)n+1 sin(t)
2 0 nπ

and Z 4
1 2
bn = sin(nπξ/4) dξ = (1 − (−1)n )
2 0 nπ
and
∞
X 128(−1)n 2 2
Tn (t) = 4 4
(16 cos(t) − n2 π 2 sin(t) − 16e−n π t/16 ) sin(nπx/4)
n=1
nπ(n π + 256)
∞
X 2 2 2
+ (1 − (−1)n ) sin(nπx/4)e−n π t/16 .
n=1
nπ

3. First,
Z 5
2
Bn (t) = t cos(ξ) sin(nπξ/5) dξ
5 0
2t
= ((−1)n+1 (nπ + 5) + nπ),
n2 π 2 − 25
Z 5
2 500
bn = ξ 2 (5 − ξ) sin(nπξ/5) dξ = ((−1)n+1 − 1)
5 0 n3 π 3
and
50(1 − cos(5)(−1)n )  2 2 −n2 π 2 t/25

Tn (t) = n π t − 25 + 25e .
n3 π 3 (n2 π 2 − 25)

The solution is
∞
X 50(1 − cos(5)(−1)n ) 2 2 2 2
u(x, t) = 3 3 2 2
(n π t − 25 + 25e−n π t/25 ) sin(nπx/5)
n=1
n π (n π − 25)
∞
X 500 2 2
+ 3 π3
((−1)n+1 − 1) sin(nπx/5)e−n π t/25 .
n=1
n

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.2. THE HEAT EQUATION WITH A FORCING TERM F (X, T ) 151

Figure 5.1: Solution surface for Problem 3, without effects of the forcing term
included.

Figure 5.2: Solution surface for Problem 3, including effects of the forcing term.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 152 CHAPTER 5. THE HEAT EQUATION

Sometimes a graphic can display features of a solution. This is done for

Problem 3 (Section 5.2). Figure 5.1 shows part of a surface plot of the solution
without the forcing term, and Figure 5.2 shows the solution with the forcing
term. In Figure 5.1, the temperature decreases quickly to zero, without the
introduction of new energy, while in Figure 5.2 this does not occur.

4. First we need
Z 1
2K
Bn (t) = K sin(nπξ/2) dξ = (1 − cos(nπ/2)),
0 nπ

b1 = 1 and bn = 0 for n = 2, 3, · · · ,
and
2K 2 2
Tn (t) = 3 3
(1 − cos(nπ/2))(1 − e−n π t ).
n π
The solution is
∞
X 2K 2 2
u(x, t) = 3 π3
(1 − cos(nπ/2))(1 − e−n π t ) sin(nπx/2)
n=1
n
2
+ sin(πx/2)e−π t .

5. First compute
Z 3
2 6t
Bn (t) = ξt sin(nπξ/3) dξ = (−1)n+1 ,
3 0 nπ
Z 3
2 2K
bn = K sin(nπξ/3) dξ = (1 − (−1)n ),
3 0 nπ
and
27(−1)n+1 2 2
Tn (t) = (16n2 π 2 − 9 + 9e−16n π t/9 ).
128n5 π 5
The solution is
∞
X 27(−1)n+1 2 2
u(x, t) = 5 π5
(16n2 π 2 − 9 + 9e−16n π t/9 ) sin(nπx/3)
n=1
128n
∞
X 2K 2 2
+ (1 − (−1)n ) sin(nπx/5)e−16n π t/9 .
n=1
nπ

5.3 The Heat Equation on the Real Line

1. With f (x) = e−|x| , compute

1 ∞ −|ξ|
Z
8 1
aω = e cos(ωξ) dξ =
π −∞ π 16 + ω 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.3. THE HEAT EQUATION ON THE REAL LINE 153

and bω = 0 because f (x) is an even function on the real line. The solution
is
8 ∞
Z
1 2
u(x, t) = 2
cos(ωx)e−ω kt .
π 0 16 + ω

The solution can also be written in the form

Z ∞
1 2
u(x, t) = √ e−|ξ| e−(x−ξ) /4kt dξ.
2 πkt −∞

2. The coefficients in the Fourier integral solution are aω = 0 because f (x)

is an odd function, and

1 π
Z
2 sin(ωπ)
bω = sin(ξ) sin(ωξ) dξ = .
π −π π(ω 2 − 1)

The solution is
Z ∞
2 sin(ωπ) 2
u(x, t) = 2
sin(ωx)e−ω kt .
π 0 ω −1

Alternatively, we can write the solution as

Z π
1 2
u(x, t) = √ sin(ξ)e−(x−ξ) /4kt dξ.
2 πkt −π

3. The coefficients are

1 4
Z
1
aω = ξ cos(ωξ) dξ = (4ω sin(4ω) + cos(4ω) − 1)
π 0 πω 2

and Z 4
1 1
bω = ξ sin(ωξ) dξ = (sin(4ω) − 4ω cos(4ω)).
π 0 πω 2
The solution is
Z ∞
2
u(x, t) = (aω cos(ωx) + bω sin(ωx))e−ω kt
dω.
0

We can also write

Z 4
1 2
u(x, t) = √ ξe−(x−ξ) /4kt
dξ.
2 πkt 0

4. We need
Z 1
1 2 cos(ω) sinh(1) + ω sin(ω) cosh(1)
aω = e−ξ cos(ωξ) dξ =
π −1 π ω2 + 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 154 CHAPTER 5. THE HEAT EQUATION

and
Z 1
1 2
bω = e−ξ sin(ωξ) dξ = ω cos(ω) sinh() − sin(ω) cosh(1)ω 2 + 1.
π −1 π

The solution is
Z ∞
2 2
u(x, t) = (aω cos(ωx) + bω sin(ωx))e−ω kt
dω.
π 0

The solution can also be written

Z 1
1 2
u(x, t) = √ e−ξ e−(x−ξ) /4kt
dξ.
2 πkt −1

In each of Problems 5–8, the solution has the form

Z ∞
2
u(x, t) = [aω cos(ωx) + bω sin(ωx)]e−ω kt
0

and just aω and bω are given

5. aω = 0 because f (x) is an even function, and

4(1 − cos(ω))
bω = .
πω

6.
2
aω = − (2 sin(ω) + 3 sin(3ω) − 7 sin(9ω))
πω
and
2
bω = (2 cos(ω) + 3 cos(3ω) − 7 cos(9ω) + 2)
πω

7. Each bω = 0, while
2 cos(πω/2)
aω = .
π(1 − ω 2 )

8.
2
aω = (cos(ω) + 2ω sin(ω) − 2 + cos(2ω) + 3ω sin(2ω)),
πω 2
and
2
bω = (− sin(ω) + 2ω cos(ω) + sin(2ω) − 3ω cos(2ω))
πω 2

9. Let Z ∞
2
F (x) = e−ζ cos(xζ) dζ.
0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.4. THE HEAT EQUATION ON A HALF-LINE 155

By differentiating under the integral sign and then integrating by parts,

we obtain
Z ∞
2
F 0 (x) = −ζe−ζ sin(xζ) dζ
0
∞ Z ∞
1 −ζ 2 1 2
= ζe sin(xζ) − x e−ζ cos(xζ) dζ
2 0 2 0
Z ∞
1 −ζ 2
=− x e cos(xζ) dζ
2 0
1
= − xF (x).
2
The linear differential equation
1
F 0 (x) + xF (x) = 0
2
has the general solution
2
F (x) = ke−x /4
,
with k an arbitrary constant. However, we also know that
Z ∞
2 1√
F (0) = e−ζ dζ = π,
0 2
an integral that can be found in tables and is widely used in statistics.
Therefore Z ∞
2 1 √ −x2 /4
F (x) = e−ζ cos(xζ) dζ = πe .
0 2
Now let x = α/β to obtain
Z ∞
1 √ −α2 /4β 2
 
−ζ 2 αζ
e cos dζ = πe .
0 β 2
Finally, this integral is half the value of the integral of the same function
from −∞ to ∞, so
Z ∞
√
 
−ζ 2 αζ 2 2
e cos dζ = πe−α /4β .
−∞ β
This is equation (5.17).

5.4 The Heat Equation on a Half-Line

In each of Problems 1–4, the initial condition is u(x, 0) = f (x) and the solution
has the form Z ∞
2
u(x, t) = bω sin(ωx)e−kω t dω,
0
where Z ∞
2
bω = f (ξ) sin(ωξ) dξ.
π 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 156 CHAPTER 5. THE HEAT EQUATION

1. Compute Z ∞
2 2 ω
bω = e−αξ sin(ωξ) dξ = ,
π 0 π ω 2 + α2
so the solution is
Z ∞
2 ω 2
u(x, t) = sin(ωx)e−kω t dω.
π 0 ω2 +α 2

2. First, Z ∞
2 2
bω = ξe−αξ sin(ωξ) dξ = αω(α2 + ω 2 )2 ,
π 0 π
so the solution is
Z ∞
4 αω 2
u(x, t) = sin(ωx)e−kω t dω.
π 0 (α2 + ω 2 )2

3. The coefficients are

h
2 1 − cos(hω)
Z
2
bω = sin(ωξ) dξ = ,
π 0 π ω

and the solution is

∞
1 − cos(hω)
Z
2 2
u(x, t) = sin(ωx)e−kω t dω.
π 0 ω

4. The coefficients are

2
2 sin(2ω) − 2ω cos(2ω)
Z
2
bω = ξ sin(ωξ) dξ = ,
π 0 π ω2

so the solution is
∞  
sin(2ω) − 2ω cos(2ω)
Z
2 2
u(x, t) = sin(ωx)e−kω t dω.
π 0 ω2

In Problems 5–8, the heat equation is to be solved on the half-line x > 0,

but the initial condition is now the insulation condition ux (x, 0) = f (x). Now
the solution is Z ∞
2
u(x, t) = aω cos(ωx)e−ω kt
dω,
0

where Z ∞
2
aω = f (ξ) cos(ωξ) dξ.
π 0

We will just give aω for each problem.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 5.5. THE TWO-DIMENSIONAL HEAT EQUATION 157

5.
Z 4
2
aω = ξ(ξ + 1) cos(ωξ) dξ
π 0
2
= (20ω 2 sin(4ω) − ω − 2 sin(4ω) + 9ω cos(4ω))
πω 3

6.
Z π
2
aω = ξ 2 cos(ωξ) dξ
π 0
2
= (−2 sin(πω) + ω 2 π 2 sin(πω) + 2ωπ cos(πω)
πω 3

7.

2 9
Z
aω = 4 cos(ωξ) dξ
π 5
8(sin(9ω) − sin(5ω))
=
πω

8.
Z ∞
2
aω = e−ξ sin(ξ) cos(ωξ) dξ
π 0
4 − 2ω 2
=
π(2 + 2ω + ω 2 )(2 − 2ω + ω 2 )

5.5 The Two-Dimensional Heat Equation

With the condition of zero initial temperature on the sides of the rectangle, and
initial temperature u(x, y, 0) = f (x, y), the solution is
∞ X
X ∞
cnm sin(nπx/L) sin(mπy/K)e−αnm kt ,
n=1 n=1

where
n2 π 2 m2 π 2
αnm = +
L2 K2
and
Z L Z K
4
cnm = f (ξ, η) sin(nπξ/L) sin(mπη/K) dη dξ.
LK 0 0

In the problems we will give the values of αnm and cnm for the particular initial
temperature function.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 158 CHAPTER 5. THE HEAT EQUATION

1. Here k = 1, L and K are positive numbers, and

f (x, y) = x(L − x)y 2 (K − y).
Because f (x, y) is a product of a function of x and a function of y, we
have
Z L ! Z !
K
4 2
cnm = ξ(L − ξ) sin(nπξ/L) dξ η (K − η) sin(mπη/K) dη
LK 0 0
16 3 3 6
=− n m π (1 − (−1)n )(1 + 2(−1)m ))
L2 K 3
and
n2 π 2 m2 π 2
αnm =
+ .
L2 K2
2. Now k = 4, L = 2 and K = 3, and
f (x, y) = x2 (2 − x)(3 − y) sin(y).
Now
Z π  Z π 
4 2
cnm = ξ (2 − ξ) sin(nπξ/2) dξ (3 − η) sin(η) sin(mη/3) dη
π2 0 0
2 3
16L K
= 3 3 6 (−1 + (−1)n )(1 + 2(−1)m ).
n m π
And
n2 π 2 m2 π 2
αnm = + .
4 9
3. Now k = 1 and L = K = π, and
Z π  Z π 
4
cnm = 2 sin(ξ) sin(nξ) dξ η cos(η/2) cos(mη) dη .
π 0 0

Now, (
Z π
π/2 if n = 1,
sin(ξ) sin(nξ) dξ =
0 0 for n = 2, 3, · · · .
Therefore, in the double summation for u(x, y, t), we have only c1m terms
and the summation is for m = 1 to ∞. Completing the computation of
the integrations with respect to η, we obtain
32m(−1)m+1
c1,m = .
(4m2 − 1)2
Further,
α1m = 1 + m2 .
The solution is
∞
X 32m(−1)m+1 2
u(x, y, t) = 2 2
sin(x) sin(my)e−(1+m )t .
m=1
(4m − 1)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 6

The Wave Equation

6.1 Wave Motion on an Interval

For each of Problems 1–8, the problem involves the wave equation on [0, L], with
fixed ends, initial position y(x, 0) = f (x), and initial velocity yt (x, 0) = g(x).
The solution is
∞
X
y(x, t) = [an cos(nπct/L) + bn sin(nπct/L)] sin(nπx/L),
n=1

where Z L
2
an = f (ξ) sin(nπξ/L) dξ
L 0

and Z L
2
bn = g(ξ) sin(nπξ/L) dξ.
nπc 0

1. Here c = 1, L = 2, the initial position is given by f (x) = 0, and the initial

velocity is (
2x for 0 ≤ x ≤ 1,
g(x) =
0 for 1 < x ≤ 2.
In the general expression for the solution, then, we have an = 0 for n =
1, 2, · · · and
Z 2
2
bn = g(ξ) sin(nπξ/2)
nπ 0
Z 1
2
= 2ξ sin(nπξ/2) dξ
nπ 0
8
= 3 3 [2 sin(nπ/2) − nπ cos(nπ/2)].
n π

159

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 160 CHAPTER 6. THE WAVE EQUATION

The solution is
∞
8 X 1
y(x, t) = [2 sin(nπ/2) − nπ cos(nπ/2)] sin(nπx/2) sin(nπt/2).
π 3 n=1 n3

2. Now c = 3, L = 4, f (x) = 2 sin(πx) and g(x) = 0, so the string is lifted to

its initial position and released from rest. Now each bn = 0, and
(
1 4
Z
2 if n = 4,
an = 1 sin(πξ) sin(nπξ/4)dξ =
2 0 0 for n = 1, 2, 3, 5, 6, · · · .

The solution is
y(x, t) = 2 sin(πx) cos(3πt).

3. Each an = 0 and
Z 3
1 54
bb = ξ(3 − ξ) sin(nπx/3) dξ = (1 − (−1)n ).
nπ 0 n4 π 4
The solution is
∞
X 54
y(x, t) = 4 π4
(1 − (−1)n ) sin(nπx/3) sin(2nπt/3).
n=1
n

Because (1 − (−1)n ) is 2 if n is odd, and zero if n is even, we can also

write the solution by summing only over the odd positive integers. This
is achieved by replacing n with 2n − 1 in the expression being summed,
and replacing each (1 − (−1)n ) with 2:
∞
X 108
y(x, t) = sin((2n − 1)πx/3) sin(2(2n − 1)πt/3).
n=1
(2n − 1)4 π 4

4. After carrying out the integrations for the coefficients, we get

∞
4 X 1
y(x, t) = sin(x) cos(3t) + sin((2n − 1)x) sin(3(2n − 1)t).
3π n=1 (2n − 1)2

5. The solution is
∞
24 X (−1)n+1 √
y(x, t) = 2
sin((2n − 1)x/2) cos((2n − 1) 2t).
π n=1 (2n − 1)

6. The solution is
∞
5 X 1
y(x, t) = [5 sin(4nπ/5)+nπ cos(4nπ/5)] sin(nπx/5) sin(2nπt/5).
π 3 n=1 n3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.1. WAVE MOTION ON AN INTERVAL 161

7. The solution is
∞
32 X 1
y(x, t) = − sin((2n − 1)πx/2) cos((3(2n − 1)πt/2)
π 3 n=1 (2n − 1)3
∞
4 X 1
+ [cos(nπ/4) − cos(nπ/2)] sin(nπx/2) sin(3nπt/2).
π 2 n=1 n2

8. The solution is
∞
2X 1
y(x, t) = sin(2x) cos(10t) + sin(nx) sin(5nt).
5 n=1 n2

9. Let y(x, t) = Y (x, t) + ψ(x) and substitute into the wave equation

ytt = Ytt = 3yxx + 2x = 3Yxx + 3ψ 00 (x) + 2x.

Choose ψ(x) so that 3ψ 00 (x) + 2x = 0. This means that

1
ψ(x) = − x3 + cx + d.
9
Now,
y(0, t) = Y (0, t) + ψ(0) = 0
so let d = 0 to have ψ(0) = 0. Then Y (0, t) = 0.
Next
8
y(2, t) = Y (2, t) + ψ(2) = Y (2, t) −
+ 2c = 0.
9
We will have Y (2, t) = 0 if c = 4/9. This means that
1 4 1
ψ(x) = − x3 + x = x(4 − x2 ).
9 9 9
The problem for Y (x, t) is

Ytt = 3Yxx for 0 < x < 2, t > 0,

Y (0, t) = Y (2, t) = 0,
1
Y (x, 0) = y(x, 0) − ψ(x) = x(x2 − 4).
9
The solution for Y (x, t) is
∞
X 32 (−1)n √
Y (x, t) = 3 π3
sin(nπx/2) cos(nπ t/2).
n=1
3 n

The original problem has the solution

1
y(x, t) = Y (x, t) + x(4 − x2 ).
9

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 162 CHAPTER 6. THE WAVE EQUATION

10. Here L = 4 and c = 3. Let y(x, t) = Y (x, t) + ψ(x) and substitute this
into the wave equation to get

Ytt = 9(Yxx + ψ 00 (x)) + x2 .

Set 9ψ 00 (x) + x2 = 0. By two integrations,

1 4
ψ(x) = − x + cx + d.
108
Now, y(0, t) = Y (0, t) + ψ(0) = 0 will give us Y (0, t) = 0 if ψ(0) = 0. Thus
choose d = 0.
Next, y(4, t) = Y (4, t) + ψ(4) = 0 will give us Y (4, t) = 0 if ψ(4) = 0.
Then c = 16/27, so
1 4 16
ψ(x) = x + x.
108 27
The problem for Y (x, t) is

Ytt = 9Yxx for 0 < x < 4, t > 0,

Y (0, t) = Y (4, t) = 0,
x4 16
Y (x, 0) = y(x, 0) − ψ(x) = sin(πx) + − x,
108 27
Yt (x, 0) = 0.

This has a solution of the form

∞
X
Y (x, t) = an cos(3nπt/4) sin(nπx/4),
n=1

where
1 4 ξ4
Z  
16
an = sin(πξ) + − ξ sin(nπξ/4) dξ
2 0 108 27
8 1
128n2 − 128n1 (−1)n − 2048

=
9 n5 π 5 (n2 − 16)
+2048(−1)n + 64n4 π 2 (−1)n − 1024n2 π 2 (−1)n ,


if n 6= 4, and

ξ4
 
1 16
a4 = sin(πξ) + − ξ sin(πξ) dξ
2 108 27
3
1 8 + 9π
= .
9 π3
With these, the solution of the original problem is

y(x, t) = Y (x, t) + ψ(x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.1. WAVE MOTION ON AN INTERVAL 163

11. Let y(x, t) = Y (x, t) + ψ(x). Substitute this into the wave equation to get

ytt = Ytt = yxx = Yxx + ψ 00 (x) − cos(x).

This will give us Ytt = Yxx if ψ 00 (x) = cos(x), which means that

ψ(x) = − cos(x) + cx + d.

Now
y(0, t) = 0 = Y (0, t) + ψ(0) = −1 + d.
This will give us Y (0, t) = 0 if d = 1. Next,

y(2π, t) = 0 = Y (2π, t) − cos(2π) + 2πc + 1.

This will give us Y (2π, 0) = 0 if c = 0. Then

ψ(x) = − cos(x) + 1.

Finally,
y(x, 0) = Y (x, 0) − cos(x) + 1 = 0
implies that Y (x, 0) = cos(x) − 1. And

yt (x, 0) = Yt (x, 0) = x.

The problem for Y (x, t) is:

Ytt = Yxx for 0 < x < 2π, t > 0,

Y (0, t) = Y (2π, t) = 0,
Y (x, 0) = cos(x) − 1, Yt (x, 0) = x.

This has a solution of the form

∞
X
Y (x, t) = [an cos(nt/2) + bn sin(nt/2)] sin(nx/2),
n=1

where
Z 2π
1
an = (cos(ξ) − 1) sin(nξ/2) dξ
π 0
(
16
nπ(n2 −4) if n is odd,
=
0 if n is even,

and Z 2π
2 8
bn = ξ sin(nξ/2) dξ = (−1)n+1 .
π 0 n2
These coefficients determine Y (x, t), and then y(x, t) = Y (x, t)+1−cos(x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 164 CHAPTER 6. THE WAVE EQUATION

12. Let y(x, t) = Y (x, t) + ψ(x) and substitute into the wave equation to get

Ytt = 9(Yxx + ψ 00 (x)) + 5x3 .

Choose ψ(x) so that 9ψ 00 (x) = −5x3 . This requires that

1 5
ψ(x) = − x + cx + d.
36
To have Y (0, t) = Y (4, t) = 0, we need ψ(0) = ψ(4) = 0. We get ψ(0) = 0
by choosing d = 0. And
1 4
ψ(4) = − 4 + 4c = 0
36
requires that c = 64/9. Then
1 5 64
ψ(x) = − x + x.
36 9
Now Y satisfies:

Ytt = 9Yxx for 0 < x < 4, t > 0,

Y (0, t) = Y (4, t) = 0,
1 5 64
Y (x, 0) = f (x) − ψ(x) = 1 − cos(πx) + x − x, yt (x, 0) = 0.
36 9

This has the solution

∞
X
Y (x, t) = an cos(3nπt/4) sin(nπx/4),
n=1

where
Z 4  
1 1 5 64
an = 1 − cos(πξ) + ξ − x sin(nπξ/4) dξ.
2 0 36 9

We find that
20 8π 2 − 3
a4 =
9 π5
while, for n 6= 4,
32
−9n5 π 5 + 30720nπ(−1)n + 9n5 π 5 (−1)n

an =
9n6 π 6 (n2
− 16)
+320n5 π 3 (−1)n − 5120n3 π 3 (−1)n − 1920n3 π(−1)n .


The original problem has the solution

y(x, t) = Y (x, t) + ψ(x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.1. WAVE MOTION ON AN INTERVAL 165

13. Let y(x, t) = Y (x, t) + ψ(x) and substitute this into the wave equation of
the problem to choose ψ(x) so that

7ψ 00 (x) + e−x = 0, ψ(0) = ψ(2) = 0.

This leads to
1 1 1
ψ(x) = − e−x + (e−2 − 1)x + .
7 14 7
Now

Ytt = 7Yxx for 0 < x < 2, t > 0,

Y (0, t) = Y (2, t) = 0,
Y (x, 0) = −ψ(x), Yt (x, 0) = 5x.

This has the solution

∞ h
X √ √ i
Y (x, t) = an cos(nπ 7t/2) + bn sin(nπ 7t/2) sin(nπx/2),
n=1

where
Z 2  
1 −ξ 1 1
an = e − (e−2 − 1)ξ − sin(nπξ/2) dξ
0 7 14 7
2
= (−4 − n2 π 2 e−2 (−1)n + e−1 n2 π 2 (−1)n + 4e−1 (−1)n ),
7nπ(4 + n2 π 2 )
and
2
40(−1)n+1
Z
2
bn = √ 5ξ sin(nπξ/2) dξ = √ .
nπ 7 0 n2 π 2 7
These coefficients determine Y (x, t), and then y(x, t) = Y (x, t) + ψ(x).
14. Let y(x, t) = Y (x, t) + ψ(x). Solve

4ψ 00 + cos(πx) = 0; ψ(0) = ψ(4) = 0

to get
1
ψ(x) = (cos(πx) − 1).
4π 2
Then

Ytt = 4Yxx for 0 < x < 4, t > 0,

Y (0, t) = Y (4, t) = 0,
Y (x, 0) = x(π − x) − ψ(x), Yt (x, 0) = x2 .

Then
∞
X
Y (x, t) = [an cos(nπt/2) + bn sin(nπt/2)] sin(nπx/4) dx,
n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 166 CHAPTER 6. THE WAVE EQUATION

where
Z 4 
1 1
an = ξ(4 − ξ) − 2 (cos(πξ) − 1) sin(nπξ/4) dξ
20 4π
8
=− 3 3 2 (128 − 7n2 + 7n2 (−1)n − 128(−1)n ),
n π (n − 16)

and
Z 4
1
bn = ξ 2 sin(nπξ/4) dξ
πn 0
64
=− (2(1 − (−1n ) + n2 π 2 (−1)n ).
n4 π 4
Finally, y(x, t) = Y (x, t) + ψ(x).

15. (a) Substitute y(x, t) = X(x)T (t) into the fourth-order differential equa-
tion to get
X (4) − λX = 0, T 00 + λa4 λT = 0,
with λ the separation constant. Note - by rearranging terms differently,
we can reach different separated equations for X and T . For example, we
could have kept the a4 factor with the X terms.
(b) Consider cases on λ, noting that the boundary conditions are

X 00 (0) = X 00 (π) = X (3) (0) = X (3) (π) = 0.

Case 1 - Suppose λ = 0. Then X (4) (x) = 0 and four integrations give us

X(x) = A + Bx + Cx2 + DX 3 .

The boundary conditions force C = D = 0, while A and B are arbitrary.

Therefore 0 is an eigenvalue of this problem, with eigenfunctions X0 (x) =
A + Bx, with A and B not both zero. In this case solutions for T are
T (t) = α + βt.

Case 2 - Suppose λ < 0. The notation is simplified if we set λ = −4α4 ,

with α > 0. The differential equation for X is

X (4) + 4α4 X = 0,

with characteristic equation r4 + 4α4 = 0. This has roots

(1 + i)α, (1 − i)α, (−1 + i)α and (−1 − i)α.

In this case solutions are

X(x) = eαx (A cos(αx) + B sin(αx)) + e−αx (C cos(αx) + D sin(αx)).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.1. WAVE MOTION ON AN INTERVAL 167

Apply the boundary conditions to this general solution to obtain:

B − D = 0,
A − B − C − D = 0,
− Aeαπ sin(απ) + Beαπ cos(απ) + Ce−απ sin(απ) − De−απ cos(απ) = 0,
− Aeαπ (cos(απ) + sin(απ)) + Beαπ (cos(απ) − sin(απ))
− Ce−απ (cos(απ) − sin(απ)) − De−απ (cos(απ) + sin(απ)) = 0.
This is a 4 × 4 homogeneous system of linear algebraic equations. This
system has a nontrivial solution if and only if the determinant of the
coefficients is zero:
cosh(2απ) − cos(2απ) = 0.
But this equation is satisfied only by α = 0, and in this case α > 0.
Therefore the system has only the trivial solution A = B = C = D = 0,
and this problem has no negative eigenvalue.

Case 3 - Suppose λ > 0, say λ = α4 with α > 0. Now X (4) − α4 X = 0,

and the characteristic equation has roots
α, −α, αi, −αi.
The general solution is
X(x) = A cos(αx) + B sin(αx) + C cosh(αx) + D sinh(αx).
The boundary conditions give us four equations:
− A + C = 0,
− A cos(απ) − B sin(απ) + C cosh(απ) + D sinh(απ) = 0,
− B + D = 0,
A sin(απ) + B cos(απ) + C sinh(απ + D cosh(απ) = 0.
From the first and third equations, A = C and B = D. This reduces the
system to the second and fourth equations in two unknowns:
C(cosh(απ) − cos(απ)) + D(cosh(απ) − sin(απ)) = 0,
C(sinh(απ) + sin(απ)) + D(cosh(απ) − cos(απ)) = 0.
This 2 × 2 system has a nontrivial solution if and only if the determinant
of the system is nonzero. This requires that
cos(απ) cosh(απ) = 1.
It may not be obvious, but this equation has infinitely many positive
solutions for α (two are 2.499752670 and 0.000000207171091). If these
solutions for α are listed α1 < α2 < · · · , then λn = αn4 is an eigenvalue of
the problem. Eigenfunctions then have the form
Xn (x) = A cos(αn x) + B sin(αn x) + C cosh(αn x) + D sinh(αn x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 168 CHAPTER 6. THE WAVE EQUATION

16. Separation of variables yields

X 00 + λX = 0; X(0) = X(L) = 0

and
T 00 + AT 0 + (B + c2 λ)T = 0; T 0 (0) = 0.

The problem for X(x) is a familiar one, with eigenvalues and eigenfunc-
tions
n2 π 2
λn = , Xn (t) = sin(nπx/L).
L2
With these eigenvalues, the characteristic equation for the differential
equation for T is

r2 + Ar + (B + c2 n2 π 2 /L2 ) = 0

with roots s
c2 n 2 π 2
 
A 1
r=− ± A2 −4 B+ .
2 2 L2

The given condition that A2 L2 < 4(V L2 + c2 n2 π 2 ) ensures that these

roots for r are complex. Let

rn = 4(BL2 + c2 n2 π 2 ) − A2 L2 .

The roots are then

A rn
r=− ± i.
2 2L
Then
Tn (t) = e−At/2 [an cos(rn t/2L) + bn sin(rn t/2L)].

Now T 0 (0) = 0 gives us −Aan /2 + bn rn /2L = 0, so

AL
bn = an .
rn

By superposition,
∞  
X AL
u(x, t) = e−At/2 an cos(rn t/2L) + sin(rn t/2L) sin(nπx/L).
n=1
rn

To satisfy u(x, 0) = f (x), choose

Z L
2
an = f (ξ) sin(nπξ/L) dξ.
L 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.2. WAVE MOTION IN AN UNBOUNDED MEDIUM 169

6.2 Wave Motion in an Unbounded Medium

For the wave equation on a the real line,

ytt = c2 yxx for − ∞ < x < ∞, t > 0,

y(x, 0) = f (x), yt (x, 0) = 0,

specifying an initial position but zero initial velocity, a solution can be found
very much like the problem for an interval [0, L], with Fourier integrals replacing
the Fourier series seen in the bounded interval case. The solution is
Z ∞
y(x, t) = [aω cos(ωx) + bω sin(ωx)] cos(ωct) dω,
0

where Z ∞
1
aω = f (ξ) cos(ωξ) dξ
π −∞

and Z ∞
1
bω = f (ξ) sin(ωξ) dξ.
π −∞

If y(x, 0) = 0 and yt (x, 0) = g(x) (string released without initial displacement,

but with initial velocity g(x)), then the solution is
Z ∞
y(x, t) = [αω cos(ωx) + βω sin(ωx)] sin(ωct) dω,
0

where Z ∞
1
αω = g(ξ) cos(ωξ) dξ
πωc −∞

and Z ∞
1
βω = g(ξ) sin(ωξ) dξ.
πωc −∞

If the problem has f (x) and g(x) both nonzero, then the solution is the sum of
the solution with zero initial velocity, and the solution with no initial displace-
ment.

1. With c = 5, f (x) = e−5|x| and g(x) = 0, compute

1 ∞ −5|ξ|
Z
10
aω = e cos(ωξ) dξ =
π −∞ (25 + ω 2 )π

and bω = 0 because f (x) is an even function. The solution is

10 ∞
Z  
1
y(x, t) = cos(ωx) cos(12ωt) dω.
π 0 25 + ω 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 170 CHAPTER 6. THE WAVE EQUATION

2. The coefficients are

8
1 − cos(8ω)
Z
1
aω = (8 − ξ) cos(ωξ) dξ =
π 0 πω 2
and
8
8ω − sin(8ω)
Z
1
bω = (8 − ξ) sin(ωξ) dξ = .
π 0 πω 2
The solution is
∞  
1 − cos(8ω)
Z
y(x, t) = cos(ωx) cos(8ωt) dω
0 πω 2
Z ∞ 
8ω − sin(8ω)
+ sin(ωx) cos(8ωt) dω.
0 πω 2

3. Compute the coefficients to determine the solution

Z ∞ 
sin(πω)
y(x, t) = − sin(ωx) sin(4ωt) dω.
0 2πω(ω 2 − 1)

4. The solution is
Z ∞  
2
y(x, t) = (2 − cos(2ω)) cos(ωx) cos(ωt) dω.
0 πω 2

5. The solution is
Z ∞
y(x, t) = [αω cos(ωx) + βω sin(ωx)] sin(3ωt) dω,
0

where
∞
2 cos(ω) − ω sin(ω)
Z
1 1
αω = e−2ξ cos(ωξ) dξ =
3πω 1 3e2 πω 4 + ω2
and
Z ∞
1 1 2 sin(ω) + ω cos(ω)
βω = e−2ξ sin(ωξ) dξ = .
3πω 1 3e2 πω 4 + ω2

6. The solution is
∞  
1 − cos(2ω)
Z
y(x, t) = sin(ωx) sin(2ωt) dω.
0 πω 2

7. The solution for the problem with the given displacement and zero initial
velocity is
Z ∞
y1 (x, t) = [aω cos(ωx) + bω sin(ωx)] cos(7ωt) dω,
0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.2. WAVE MOTION IN AN UNBOUNDED MEDIUM 171

where
1 5
Z
aω = f (ξ) cos(ωξ) dξ
π −1
sin(ω) − 2 sin(2ω) + 3 sin(5ω)
=
πω
and
1 5
Z
bω = f (ξ) sin(ωξ) dξ
π −1
cos(ω) + 2 cos(2ω) − 3 cos(5ω)
= .
πω
The solution for the problem with the given velocity, but zero initial dis-
placement, is
Z ∞
y2 (x, t) = [αω cos(ωx) + βω sin(ωx)] sin(7ωt) dω,
0

where
Z 1
1
αω = e−|ξ| cos(ωξ) dξ
7πω −1
2
=− (e−1 cos(ω) − ωe−1 sin(ω) − 1)
7π(1 + ω 2 )
and Z 1
1
βω = e−|ξ| sin(ωξ) dξ = 0.
7πω −1

The solution of the problem with initial displacement f (x) and initial
velocity g(x) is
y(x, t) = y1 (x, t) + y2 (x, t).

8. Let y1 (x, t) be the solution of the problem with initial displacement f (x)
and zero initial velocity, and y2 (x, t) the solution with no initial displace-
ment, but initial velocity g(x). Then
y(x, t) = y1 (x, t) + y2 (x, t).
Now, Z ∞ √
y1 (x, t) = [aω cos(ωx) + bω sin(ωx)] cos( 7ωt).
0
where
Z 3π
1
aω = sin(ξ) cos(ωξ) dξ
π −π
4 cos(πω) sin2 (πω)
=
π(ω 2 − 1)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 172 CHAPTER 6. THE WAVE EQUATION

and
Z 3π
1
bω = sin(ξ) sin(ωξ) dξ
π −π
4 sin(πω) cos2 (πω)
= .
π(1 − ω 2 )

And ∞
Z √
y2 (x, t) = [αω cos(ωx) + βω sin(ωx)] sin( 7ωt),
0

where
Z ∞
1
αω = √ g(ξ) cos(ωξ) dξ
7πω −∞
2
= √ (8ω 2 sin(4ω) − sin(4ω) + 4ω cos(4ω))
2
πω 7
and
Z ∞
1
βω = √ g(ξ) sin(ωξ) dξ
7πω −∞
−2
=√ (1 + 8ω 2 cos(4ω) − cos(4ω) − 4ω sin(4ω)).
7πω 4
Then
y(x, t) = y1 (x, t) + y2 (x, t).

9. Following the notation of Problems 7 and 8, form y1 (x, t) with coefficients

Z 2
1
aω = ξ cos(ωξ) dξ = 0
π −2

and

1 2
Z
bω = ξ sin(ωξ) dξ
π −2
2 sin(2ω) − 4 cos(2ω)
= .
πω 2
And y2 (x, t) has coefficients
Z 3
4
αω = ξ 2 cos(ωξ) dξ
πω −3
8
= (9ω 2 sin(3ω) − 2 sin(3ω) + 6ω cos(3ω))
πω 4
and βω = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.2. WAVE MOTION IN AN UNBOUNDED MEDIUM 173

The solution is
Z ∞
y(x, t) = an cos(ωx) cos(ωt/4)
Z 0∞
+ βω sin(ωx) sin(ωt/4).
0

The the problem on a half-line [0, ∞), there is a boundary condition which
we will take to be
y(0, t) = 0
along with initial conditions

y(x, 0) = f (x), yt (x, 0) = g(x)

for x > 0. The solution has the form

Z ∞
y(x, t) = [Aω cos(ωct) + Bω sin(ωct)] sin(nωx) dω,
0

where Z ∞
2
Aω = f (ξ) sin(ωξ) dξ
π 0
and Z ∞
2
Bω = g(ξ) sin(ωξ) dξ.
π 0

10. With zero initial velocity, we need only compute

2 1
Z
Aω = ξ(1 − ξ) sin(ωξ) dξ
π 0
 
2 2 sin(ω)
= (1 − cos(ω)) − .
π ω3 ω2
The solution is
Z ∞  
2 2 sin(ω)
y(x, t) = (1 − cos(ω)) − sin(ωx) cos(3ωt) dω.
π 0 ω3 ω2

11. Here Aω = 0 and

Z 11
2
Bω = 2 sin(ωξ) dξ
3πω 4
4(cos(4ω) − cos(11ω))
= .
3πω 2
The solution is
∞
cos(4ω) − cos(11ω)
Z
4
y(x, t) = sin(ωx) sin(3ωt) dω.
3π 0 ω2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 174 CHAPTER 6. THE WAVE EQUATION

12. Aω = 0 and
Z 5π/2
2
Bω = cos(ξ) sin(ωξ) dξ
2πω π/2
sin(ωπ/2) − sin(5ωπ/2)
= .
πω(ω 2 − 1)
The solution is
∞
sin(ωπ/2) − sin(5ωπ/2)
Z
y(x, t) = sin(ωx) sin(2ωt) dω.
0 πω(ω 2 − 1)

13. With g(x) = 0, Bω = 0 and

Z ∞
2
Aω = −2e−ξ sin(ωξ) dξ
π 0
4ω
=− .
π(1 + ω 2 )
The solution is
Z ∞
4 ω
y(x, t) = − sin(ωx) cos(6ωt) dω.
π 0 1 + ω2

14. Now Aω = 0 and

Z 3
2
Bω = ξ 2 (3 − ξ) sin(ωξ) dξ
14πω 0
3
= (2 sin(3ω) − 4ω cos(3ω) − 3ω 2 sin(3ω) − 2ω).
7πω 5
The solution is
Z ∞
y(x, t) = Bω sin(ωx) sin(14ωt) dω.
0

15. Compute

2 1
Z
Aω = f (ξ) sin(ωξ) dξ
π 0
2 sin(ω)
= 2
π − ω2
and
Z 4
2
Bω = √ g(ξ) sin(ωξ) dξ
13πω 0
2
=√ (1 − 2 cos(ω) + cos(4ω)).
13πω 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 175

The solution is
Z ∞ √ √
y(x, t) = [Aω cos( 13ωt) + Bω sin( 13ωt)] sin(ωx) dω.
0

16. Compute the coefficients

2 ∞ −3ξ
Z
12ω
Aω = ξe sin(ωξ) dξ =
π 0 π(9 + ω 2 )2

and
2π
4 sin2 (πω)
Z
2
Bω = cos(ξ) sin(ωξ) dξ = .
5πω 0 (w2 − 1)π
The solution is
Z ∞
y(x, t) = [Aω cos(5ωt) + Bω sin(5ωt)] sin(ωx) dω.
0

6.3 d’Alembert’s Solution and Characteristics

1. The characteristics are the lines x − t = k1 and x + t = k2 , with k1 and
k2 arbitrary real numbers. d’Alembert’s solution of the problem is

1 x+t
Z
1 2 2
y(x, t) = [(x − t) + (x + t) ] + −ξ dξ
2 2 x−t
 x+t
1 1
= [x2 − 2xt + t2 + x2 + 2xt + t2 − ξ 2
2 2 x−t
= x2 − xt + t2 .

2. characteristics: x − 4t = k1 , x + 4t = k2 ;
1
(x − 4t)2 − 2(x − 4t) + (x + 4t)2 − 2(x + 4t)


y(x, t) =
2
1 x+4t
Z
+ (cos(ξ) dξ
8 x−4t
1
= x2 + 16t2 − 2x + cos(x) sin(4t)
4

3. characteristics: x − 7t = k1 , x + 7t = k2 ;
1 49
y(x, t) = [cos(π(x − 7t)) + cos(π(x + 7t))] + t − x2 t − t3
2 3
This solution can also be written
1 49
y(x, t) = cos(πx) cos(7πt) + t − x2 t − t3 .
2 3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 176 CHAPTER 6. THE WAVE EQUATION

4. characteristics: x − 5t = k1 , x + 5t = k2 ;
1
y(x, t) = [sin(2(x − 5t)) + sin(2(x + 5t))] + x3 t + 25xt3
2
= sin(2x) cos(10t) + x3 t + 25xt3

5. characteristics: x − 14t = k1 , x + 14t = k2 ;

1  x−14t
+ ex+14t + xt

y(x, t) = e
2
= ex cosh(14t) + xt.

6. characteristics: x − 12t = k1 , x + 12t = k2 ;

y(x, t) = x2 + 144t2 − 5x + 3t.

√ √
7. characteristics x − 3t = k1 , x + 3t = k2 ;
1 h −3|x−√3t| √ i
y(x, t) = e + e−3|x+ 3t|
2 "
√ ! √ !#
1 x + 3t x − 3t
+ √ sin − sin
3 2 2

8. characteristics: x − 9t = k1 , x + 9t = k2 ;
1
y(x, t) = (2 − cos(x − 9t) − cos(x + 9t))
2
1
+ e−x+9t (cos(−x + 9t) − sin(−x + 9t))
36
1
− e−x−9t (cos(x + 9t) + sin(x + 9t))
36

9. The solution with y(x, 0) = f (x) = sin(x) is

1
y(x, t) = (sin(x − t) + sin(x + t)).
x
With y(x, 0) = sin(x) + , the solution is

1
y (x, t) = sin(x − t) +  + sin(x + t) +  = y(x, t) + .
2

10. Let

y(x, t) = F (x − ct) + G(x + ct)

= (x − ct)2 − (x − ct) + (x + ct) cos(x + ct).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 177

Compute

yx (x, t) = 2(x − ct) − 1 + cos(x + ct) − (x + ct) sin(x + xt),

yxx (x, t) = 2 − sin(x + ct) − sin(x + ct) − (x + ct) cos(x + ct),

yt (x, t) = 2(x − ct)(−c) + c + c cos(x + ct) − c(x + ct) sin(x + ct),

and

ytt (x, t) = 2c2 − c2 sin(x + ct) − c2 sin(x + ct) − c2 (x + ct) cos(x + ct).

Now it is routine to check that

ytt = c2 yxx .

11. Now

y(x, t) = e−3(x−ct) + sin(4(x + ct)

= e−3x e3ct + sin(4(x + ct)).

Then

yx = −3e−3x e3ct + 4 cos(4(x + ct)),

yxx = 9e−3x e3ct − 16 sin(4(c + ct)),
yt = 3ce−3x e3ct + 4c cos(4(x + ct)),
ytt = 9c2 e−3x e3ct − 16c2 sin(4(x + ct)).

It is easy to check that ytt = c2 yxx .

In each of Problems 12–17, with c = 1 and g(x) = 0, the forward wave is

F (x, t) = 21 f (x − t) and the backward wave is B(x, t) = 12 f (x + t). The solution
is
y(x, t) = F (x, t) + B(x, t).

12. Figures 6.1–6.7 are graphs of y(x, t) for

1 3
t = 0, , , 2, 3, 4, 6
2 4
respectively.

13. Figures 6.8–6.13 show graphs of y(x, t) for times

1 3 7 3
t = 0, , , , 1, .
2 4 8 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 178 CHAPTER 6. THE WAVE EQUATION

Figure 6.1: y(x, 0) in Problem 12.

Figure 6.2: y(x, 1/2), Problem 12.

Figure 6.3: y(x, 3/4), Problem 12.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 179

Figure 6.4: y(x, 2), Problem 12.

Figure 6.5: y(x, 3), Problem 12.

Figure 6.6: y(x, 4), Problem 12.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 180 CHAPTER 6. THE WAVE EQUATION

Figure 6.7: y(x, 6), Problem 12.

Figure 6.8: y(x, 0), Problem 13.

Figure 6.9: y(x, 1/2), Problem 13.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 181

Figure 6.10: y(x, 3/4), Problem 13.

Figure 6.11: y(x, 7/8), Problem 13.

Figure 6.12: y(x, 1), Problem 13.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 182 CHAPTER 6. THE WAVE EQUATION

Figure 6.13: y(x, 3/2), Problem 13.

Figure 6.14: y(x, 0), Problem 14.

Figure 6.15: y(x, 1/4), Problem 14.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 183

Figure 6.16: y(x, 1/2), Problem 14.

Figure 6.17: y(x, 3/4), Problem 14.

Figure 6.18: y(x, 1), Problem 14.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 184 CHAPTER 6. THE WAVE EQUATION

Figure 6.19: y(x, 5/4), Problem 14.

14. Figures 6.14–6.21 show the graphs of y(x, t) for times

1 1 3 5 3
t = 0, , , , 1, , , 2.
4 2 4 4 2

15. Figures 6.22–6.27 show graphs of y(x, t) for

1 1 3 3
t = 0, , , , 1, .
4 2 4 2

16. Figures 6.28–6.34 show graphs of y(x, t) for

1 1 3 5 7
t = 0, , , , 1, , .
4 2 4 4 4

17. Figures 6.35–6.41 show graphs of y(x, t) for

1 1 3 5/4 7 5
t = 0, , , , , .
4 2 4 , 4 2

18. Derive d’Alembert’s solution as follows. Begin with the fact that any
solution of the wave equation on the line must look like

y(x, t) = F (x + ct) + G(x − ct),

for some twice differentiable functions F and G. To satisfy the initial

conditions, we need

y(x, 0) = F (x) + G(x) = f (x)

and, by chain rule differentiations,

yt (x, 0) = −cF 0 (x) + cG0 (x) = g(x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 185

Figure 6.20: y(x, 3/2), Problem 14.

Figure 6.21: y(x, 2), Problem 14.

Figure 6.22: y(x, 0), Problem 15.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 186 CHAPTER 6. THE WAVE EQUATION

Figure 6.23: y(x, 1/4), Problem 15.

Figure 6.24: y(x, 1/2), Problem 15.

Figure 6.25: y(x, 3/4), Problem 15.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 187

Figure 6.26: y(x, 1), Problem 15.

Integrate the last equation to obtain

1 x
Z
−F (x) + G(x) = g(ξ) dξ − F (0) + G(0).
c 0
Add this to the equation for y(x, 0) to obtain
1 x
Z
2G(x) = f (x) + g(ξ) dξ − F (0) + G(0).
c 0
Solve this for G(x) to obtain
Z x
1 1 1 1
G(x) = f (x) + g(ξ) dξ − F (0) + G(0).
2 2c 0 2 2
But then, from the equation for y(x, 0),
F (x) = f (x) − G(x)
Z x
1 1 1 1
= f (x) − g(ξ) dξ + F (0) − G(0).
2 2c 0 2 2
From the last two equations,
y(x, t) = F (x − ct) + G(x + ct)
Z x−ct
1 1 1
= f (x − ct) − g(ξ) dξ + (F (0) − G(0))
2 2c 0 2
Z x
1 1 1
+ f (x + ct) + g(ξ) dξ − (F (0) − G(0)).
2 2c 0 2
Upon combining the integrals and canceling terms where possible, this
yields d’Alembert’s formula.
19. We know that
Z x+ct
1 1
y(x, t) = (f (x − ct) + f (x + ct)) + g(ξ) dξ
2 2c x−ct

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 188 CHAPTER 6. THE WAVE EQUATION

Figure 6.27: y(x, 3/2), Problem 15.

Figure 6.28: y(x, 0), Problem 16.

Figure 6.29: y(x, 1/4), Problem 16.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 189

Figure 6.30: y(x, 1/2), Problem 16.

and
1 ˜ 1
ỹ(x, t) = (f (x − ct) + f˜(x + ct)) + g̃(ξ) dξ.
2 2c
Then

y(x, t) − ỹ(x, t)
Z x+ct
1 1
= (f (x − ct) + f (x + ct)) + g(ξ) dξ
2 2c x−ct
1 ˜  1 Z x+ct
− ˜
f (x − ct) + f (x + ct) − g̃(ξ) dξ
2 2 x−ct
1 1
= (f (x − ct) − f˜(x − ct)) + (f (x + ct) − f˜(x + ct))
2 2
Z x+ct
1
+ (g(ξ) − g̃(ξ)) dξ.
2c x−ct

Then

|y(x, t) − ỹ(x, t)|

1 1
≤ |f (x − ct) − f˜(x − ct)| + |f (x + ct) − f˜(x + ct)|
2 2
Z x+ct
1
+ |g(ξ) − g̃(ξ)| dξ
2c x−ct
Z x+ct
1 1 1
≤ 1 + 1 + 2 dξ
2 2 2c x−ct
1
≤ 1 + 2 ((x + ct) − (x − ct))
2c
= 1 + 2 t.

20. For the problem on the real line, with initial position fo (x) and initial

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 190 CHAPTER 6. THE WAVE EQUATION

Figure 6.31: y(x, 3/4), Problem 16.

Figure 6.32: y(x, 1), Problem 16.

Figure 6.33: y(x, 5/4), Problem 16.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.3. D’ALEMBERT’S SOLUTION AND CHARACTERISTICS 191

Figure 6.34: y(x, 7/4), Problem 16.

Figure 6.35: y(x, 0), Problem 17.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 192 CHAPTER 6. THE WAVE EQUATION

velocity go (x), the d’Alembert solution is

Z x+ct
1 1
y(x, t) = (fo (x − ct) + fo (x + ct)) + g(ξ) dξ.
2 2c x−ct

This function satisfies the wave equation for all x, and therefore also for
all x > 0. Further, if x ≥ 0,

1 1
y(x, 0) = (fo (x) + fo (x)) = (f (x) + f (x)) = f (x).
2 2

And,

1 1
yt (x, 0) = (−cfo0 (x) + cfo0 (x)) + (cgo (x) − (−c)go (x))
2 2c
= go (x) = g(x).

Therefore y(x, t) is also a solution of the problem on the half-line x ≥ 0.

6.4 The Wave Equation With a Forcing Term

K(x, t)
1. Here c = 4, f (x) = x, g(x) = e−x and K(x, t) = x + t. The solution is

1 x+4t −ξ
Z
1
y(x, t) = ((x − 4t) + (x + 4t)) + e dξ
2 8 x−4t
1 t x+4t−4T
Z Z
+ (X + T ) dX dT
8 0 x−4t+4T
1 −x+4t
− e−x−4t


=x+ e
8
Z t
+ (xt − xT + tT − T 2 ) dT
0
1
1 1
= x + e−x e4t − e−4t + t3 + xt2 .
8 6 2

We can also write this solution as

1 1 1
y(x, t) = x + e−x sinh(4t) + t3 + xt2 .
4 6 2

2. We are given c = 2, f (x) = sin(x), g(x) = 2x and K(x, t) = 2xt. The

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.4. THE WAVE EQUATION WITH A FORCING TERM K(X, T ) 193

solution is
1 x+2t
Z
1
y(x, t) = (sin(x − 2t) + sin(x + 2t)) + 2ξ dξ
2 4 x−2t
1 t x+2t−2T
Z Z
(X + T ) dX dT
8 0 x−2t+2T
1 1
(x + 2t)2 − (x − 2t)2


= (sin(x − 2t) + sin(x + 2t)) +
2 4
1 t
Z
T (x + 2t − 2T )2 − (x − 2t + 2T )2 dT.


+
4 0
After carrying out the last integration, this expression reduces to
1 1
y(x, t) = (sin(x − 2t) + sin(x + 2t)) + 2xt + xt3 .
2 3

3. The solution is
Z x+8t
1 1
y(x, t) = (f (x − 8t) + f (x + 8t)) + cos(2ξ) dξ
2 16 x−8t
Z t Z x+8t−8T
1
+ XT 2 dX dT
16 0 x−8t+8T
1
= x2 + 64t2 − x + (sin(−2x + 16t) + sin(2x + 16t))
32
Z t
+ −xT 2 (−t + T ) dT
0
1 1
= x2 + 64t2 − x + (sin(−2x + 16t) + sin(2x + 16t)) + xt4 .
32 32

4.

1 x+4t −ξ
Z
1
y(x, t) = (x − 4t)2 + (x + 4t)2 + ξe dξ
2 8 x−4t
1 t x+4t−4T
Z Z
+ X sin(T ) dX dT
8 0 x−4t+4T
1 1 1
= x2 + 16t2 + e−x+4t + xe−x+4t − te−x+4t
8 8 2
1 1 1
− e−x−4t − xe−x+4t − te−x−4t
8 8 2
Z t
+ x(t − T ) sin(T ) dT
0
1 1
= x2 + 16t2 + (1 + x)e−x+4t − t e−x+4t + e−x−4t


8 2
1
− (1 + x)e−x−4t + xt − x sin(t).
8

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 194 CHAPTER 6. THE WAVE EQUATION

5.
Z x+3t
1 1
y(x, t) = (cosh(x − 3t) + cosh(x + 3t)) + dξ
2 6 x−3t
Z t Z x+3t−3T
1
+ 3XT 3 dX dT
6 0 x−3t+3T
Z t
1
= (cosh(x − 3t) + cosh(x + 3t)) + t + −3xT 3 (T − t) dT
2 0
1 3
= (cosh(x − 3t) + cosh(x + 3t)) + t + xt5 .
2 20
6.
1
y(x, t) = ((1 + x − 7t) + (x + x + 7t))
2
Z t Z x+7t−7T
1
+ (X − cos(T )) dX dT
14 0 x−7t+7T
Z t
=1+x+ (xt − xT − t cos(T ) + T cos(T )) dT
0
1
= x + xt2 + cos(t).
2

6.5 The Wave Equation in Higher Dimensions

For Problems 1–3 the solution has the form
∞ X
X ∞
z(x, y, t) = anm sin(nπx/L) sin(mπy/K) cos(αnm πct),
n=1 m=1

where r
n2 m2
αnm = +
L2 K2
and Z L Z K
4
anm = f (ξ, η) sin(nπξ/L) sin(mπη/K) dη dξ.
LK 0 0

1. Because f (x, y) = x2 y is a product of a function of x and a function of y,

we can compute the coefficients as a product of integrals:
Z 2π Z K
1
anm = 2 ξ 2 sin(nξ/2) dξ η sin(mη/2) dη
π 0 0
32(−1)m
2(1 − (−1)n ) + n2 π 2 (−1)n .


= 3
mn π
Further,
1 p 2
αnm = n + m2 .
2π

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.5. THE WAVE EQUATION IN HIGHER DIMENSIONS 195

Because c = 1, the solution is

∞ X ∞
X 32(−1)m
z(x, y, t) = 3π
(2(1 − (−1)n )
n=1 m=1
mn
p
+ n2 π 2 (−1)n sin(nx/2) sin(my/2) cos( n2 + m2 t/2).

2. Now c = 3, L = 1 and K = 4, and f (x, y) = sin(πx) cos(πy/2). The

coefficients in the solution are
Z 1 Z 4
anm = sin(πξ) sin(nπξ) dξ cos(πη/2) sin(mπη/4) dη.
0 0

Now (
1
for n = 2, 3, · · · ,
Z
0
sin(πξ) sin(nπξ) dξ =
0 1/2 for n = 1.
And
(
4
0 if m = 2,
Z
cos(πη/2) sin(mπη/4) dη = m
0 − 4m(−1+(−1)
π(m2 −4)
)
6 2.
if m =

This means that n will only assume the value n = 1 and the solution is a
single summation over m, with m = 2 excluded. Further,
r
m2
α1m = 1 +
16
for m = 1, 3, 4, 5, · · · .
Then
∞
X 4m(−1 + (−1)m )
z(x, y, t) = − sin(nπx) sin(mπy/4) cos(3α1m πt).
2π(m2 − 4)
m=1,m6=2

3. We have c = 1, L = K = π and f (x, y) = xey . A routine integration yields

Z π Z π
4
anm = 2 ξ sin(nξ) dξ eη sin(mη) dη
π 0 0
4(−1)n+1 m
= (1 − e (−1)m ).
π
πn2 (m2 + 1)
Further,
1p 2
αnm = n + m2 .
π
The solution is
∞ X
X ∞ p
z(x, y, t) = anm sin(nx) sin(my) cos(2 n2 + m2 t).
n=1 m=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 196 CHAPTER 6. THE WAVE EQUATION

4. Now suppose that the membrane is initially unmoved and that an initial
velocity is given by zt (x, y, 0) = g(x, y). Separate variables in the two-
dimensional wave equation. Because the boundary conditions are the same
as in the zero initial velocity case, we obtain

n2 π 2
λn = , Xn (x) = sin(nπx/L)
L2
and
m2 π 2
µm = , Ym (y) = sin(mπy/K).
K2
The problem for T is

n2 π 2 m2 π 2
 
T 00 + c2 2
+ T = 0,
L K2

but now we have T (0) = 0, not T 0 (0) = 0. Then

Tnm (t) = cos(αnm πct),

where r
n2 m2
2
αm =
+ 2.
L K
We therefore are led to attempt a solution
∞ X
X ∞
z(x, y, t) = bnm sin(nπx/L) sin(mπy/K) cos(αnm πct).
n=1 m=1

We need
∞ X
X ∞
zt (x, y, 0) = αnm πc sin(nπx/L) sin(mπy/K) = g(x, y).
n=1 m=1

This is a Fourier double series expansion of g(x, y), and the coefficients
are
Z LZ K
4
bnm αnm πc = g(ξ, η) sin(nπξ/L) sin(mπη/K).
LK 0 0

Then
Z L Z K
4
bnm = g(ξ, η) sin(nπξ/L) sin(mπη/K).
LKαnm πc 0 0

5. Suppose c = 3, L = K = π, f (x, y) = 0 and g(x, y) = xy. Now

1p 2
αnm = n + m2 .
π

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 6.5. THE WAVE EQUATION IN HIGHER DIMENSIONS 197

Compute
4 (−1)n+m
bnm = √ .
3 n2 + m2 nm
The solution is
∞ X
X ∞ p
z(x, y, t) = bnm sin(nx) sin(mx) cos(3 n2 + m2 t).
n=1 m=1

1
√
6. With L = K = 2π, we have αnm = 2 n2 + m2 . Compute
Z 2π Z 2π
4
bnm = 1
√ sin(nξ) dξ sin(mη) dη
(2π)(2π) 2π n2 + m2 2π 0 0
1 4
= √ (1 − (−1)n )(1 − (−1)m ).
π n + m nm
2 2 2

The solution is
∞
X p
z(x, y, t) = sum∞ 2 2
m=1 bnm sin(nx/2) sin(my/2) cos( n + m t)
n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 198 CHAPTER 6. THE WAVE EQUATION

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 7

Laplace’s Equation

7.1 The Dirichlet Problem for a Rectangle

1. Substitute u(x, y) = X(x)Y (y) into Laplace’s equation to obtain

X 00 + λX = 0; X(0) = X(1) = 0

and
Y 00 − λY = 0; Y (π) = 0.
Solutions for X are

λ = n2 π 2 , Xn (x) = sin(nπx).

With these values of λ, the problem for Y (y) has solutions that are con-
stant multiples of sinh(nπ(π − y). To find a solution satisfying the bound-
ary condition u(x, 0) = sin(πx), use a superposition
∞
X
u(x, y) = an sin(nπx) sinh(nπ(π − y)).
n=1

We need
∞
X
u(x, 0) = an sin(πx) sinh(nπ 2 ) = sin(πx).
n=1

We can compute the Fourier coefficients of this sine expansion, or simply

observe that we can take an = 0 for n = 2, 3, · · · and
1
a1 = .
sinh(π 2 )
The solution is
1
u(x, y) = (sin(πx) sinh(π(π − y))).
sinh(π 2 )

199

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 200 CHAPTER 7. LAPLACE’S EQUATION

2. Separate variables and use the boundary conditions u(x, 0) = u(x, 2) = 0

to obtain the problem

Y 00 + λY = 0; Y (0) = Y (2) = 0,

with eigenvalues and eigenfunctions

n2 π 2
λn = , Yn (y) = sin(nπy/2).
L2
The problem for X is

n2 π 2
X 00 − X = 0; X(3) = 0.
L2
Solutions are constant multiples of sinh(nπ(3 − x)/2). Now attempt a
solution of the form
∞
X
u(x, y) = bn sinh(nπ(3 − x)/2) sin(nπy/2).
n=1

To satisfy u(3, y) = 0, we need

∞
X
u(0, y) = y(2 − y) = bn sinh(3nπ/2) sin(nπy/2).
n=1

This is the Fourier sine expansion of y(2 − y) on [0, 2], and the coefficients
are
Z 2
1
bn = η(2 − η) sin(nπη/2) dη
sinh(3nπ/2) 0
16 1 − (−1)n
= .
sinh(3nπ/2) n3 π 3
The solution is
∞
16 X 1 − (−1)n sinh(nπ(3 − x)/2)
u(x, y) = sin(nπy/2).
π 3 n=1 n3 sinh(3nπ/2)

3. After separating the variables and applying the boundary conditions, we

find that the solution has the form
∞
X sinh(nπy)
u(x, y) = an sin(nπx).
n=1
sinh(4nπ)

The coefficients must be determined so that

∞
X
u(x, 4) = an sin(nπx) = x cos(πx/2).
n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.1. THE DIRICHLET PROBLEM FOR A RECTANGLE 201

This is a Fourier sine expansion of x cos(πx/2) on [0, 1]. Choose the coef-
ficients as
Z 1
32n(−1)n+1
an = 2 ξ cos(πξ/2) sin(nπξ) dξ = 2 .
0 π (4n2 − 1)2

These determine the solution.

5. There are nonhomogeneous boundary conditions on two sides of the rect-

angle, so write
u(x, y) = v(x, y) + w(x, y)
where

∇2 v = 0; v(0, y) = v(π, y) = v(x, 0) = 0, v(x, π) = x sin(πx)

and

∇2 w = 0; w(x, 0) = w(x, π) = w(0, y) = 0, w(w, y) = sin(y).

These are defined on 0 < x < 2, 0 < y < π. Solve these problems inde-
pendently.
First, separate variables in the problem for w to find that it has a solution
of the form
∞
X sinh(nx)
w(x, y) = bn sin(ny) .
n=1
sinh(2n)

Observe that we can solve this problem for w by taking b1 = 1 and all
other bn = 0, so
sinh(x)
w(x, y) = sin(y) .
sinh(2)
The problem for v has a solution of the form
∞
X sinh(nπy/2)
v(x, y) = an sin(nπx/2) .
n=1
sinh(nπ 2 /2)

We need
∞
X
v(x, π) = x sin(πx) = an sin(nπx/2).
n=1

This is a Fourier sine expansion of x sin(πx) on [0, 2], so choose

Z 2
bn = ξ sin(nξ) sin(nπξ/2) dξ
(0
16n n
π 2 ((n2 −4)2 ) ((−1) − 1) for n = 1, 3, 4, 5, · · · ,
=
1 for n = 2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 202 CHAPTER 7. LAPLACE’S EQUATION

Then
sinh(πy)
v(x, y) = sin(πx)
sinh(π 2 )
∞
10 X n sinh(nπy/2)
+ ((−1)n − 1) sin(nπx/2) .
π2 (n2 − 4)2 sinh(nπ 2 /2)
n=1,n6=2

6. A routine separation of variables, together with the zero boundary condi-

tions on sides y = 0, y = b and x = 0, lead to a solution of the form
∞
X sinh((2n − 1)πx/2b)
u(x, y) = an sin((2n − 1)πy/2) .
n=1
sinh((2n − 1)πa/2b)

Now we need to choose the coefficients so that

∞
X
u(a, y) = g(y) = an sin((2n − 1)πy/2b).
n=1

Then Z b
2
an = g(η) sin((2n − 1)πη/2b) dη.
b 0

7. Separation of variables and the zero boundary conditions on x = 0, x = a

and y = 0 yield a general form of the solution:
∞
X sinh((2n − 1)πy/2a)
u(x, y) = an sin((2n − 1)πx/2a) .
n=1
sinh((2n − 1)πb/2a)

Now we need
∞
X
u(x, 0) = f (x) = an sin((2n − 1)πx/2a).
n=1

This leads us to choose

Z a
2
an = f (ξ) sin((2n − 1)πξ/2a) dξ.
a 0

8. The homogeneous conditions on the sides x = 0, x = a and y = 0 lead to

the form of the solution:
X sinh(nπy/a)
u(x, y) = ∞ sin(nπx/a) .
n=1
sinh(nπb/a)

Then
∞
X
2
u(x, a) = x(x − a) = an sin(nπx/a).
n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.1. THE DIRICHLET PROBLEM FOR A RECTANGLE 203

The coefficients in this eigenfunction expansion are

2 a
Z
an = ξ(ξ − a)2 sin(nπξ/a) dξ
a 0
4
= 3 3 (1 + 2nπ(−1)n ).
n π

9. Write the solution as u(x, y) = v(x, y) + w(x, y), where v is the solution
of the problem

∇2 v = 0, v(x, 0) = v(x, 1) = v(4, y) = 0, v(0, y) = sin(πy),

and w is the solution of

∇2 (w) = 0, w(x, 0) = w(x, 1) = w(0, y) = 0, w(4, y) = y(1 − y).

These problems are defined on 0 ≤ x ≤ 4, 0 ≤ y ≤ 1. A separation of

variables yields a general form of the solution of the problem for v:
∞
X sinh(nπ(4 − x))
v(x, y) = an sin(nπy) .
n=1
sinh(4nπ)

We need
∞
X
v(0, y) = sin(πy) = an sin(nπy),
n=1

so by observation we can let a1 = 1 and an = 0 for n = 2, 3, · · · . Then

sinh(π(4 − x))
v(x, y) = sin(πy) .
sinh(4π)

Another separation of variables leads to a general form of the solution for

w:
X∞
w(x, y) = bn sin(nπy) sinh(nπx).
n=1

Then
∞
X
w(4, y) = y(1 − y) = bn sinh(4nπ) sin(nπy),
n=1

so
Z 1
2
bn = ξ(1 − ξ) sin(nπξ) dξ
sinh(4nπ) 0
4(1 − (−1)n )
= 3 3 .
n π sinh(4nπ)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 204 CHAPTER 7. LAPLACE’S EQUATION

7.2 The Dirichlet Problem for a Disk

For each of Problems 1–8, a solution
∞  n
1 X r
u(r, θ) = a0 + [an cos(nθ) + bn sin(nθ)],
2 n=1
R

where Z π
1
an = f (ξ) cos(nξ) dξ
πRn −π

for n = 0, 1, 2, · · · and
Z π
1
bn = f (ξ) sin(nξ) dξ
πRn −π

for n = 1, 2, · · · .

1. We can see by observation that u(r, θ) = 1 is a solution. This can also be

obtained the long way by carrying out the integrations, obtaining a0 = 2
and an = bn = 0 for n = 1, 2, · · · .

2. As with Problem 1, we can obtain the solution by inspection by matching

coefficients of the proposed series solution to settle on only one nonzero
coefficient, namely a4 . If we carry out the integrations, we get
Z π
1
an = 8 cos(4ξ) cos(nξ) dξ = 0
π3n −π

for n = 0, 1, 2, 3, 5, 6, · · · , and
Z π
1
bn = 8 cos(4ξ) sin(nξ) dξ = 0
π3n −π

for n = 1, 2, · · · , while
Z π
1 8
a4 = 8 cos(4ξ) cos(4ξ) dξ = .
π34 −π 34

The solution therefore consists of a single term:

 r 4
u(r, θ) = 8 cos(4θ).
3

3. Calculate π
2π 2
Z
1
a0 = (ξ 2 − ξ) dξ = ,
π −π 3
π
4(−1)n
Z
1
an = (ξ 2 − ξ) cos(nξ) dξ =
2n π −π n2 2 n

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.2. THE DIRICHLET PROBLEM FOR A DISK 205

and π
2(−1)n
Z
1
bn = n (ξ 2 − ξ) sin(nξ) dξ = .
2 π −π n2n
The solution is
∞  n
π2 X r (−1)n
u(r, θ) = +2 [2 cos(nθ) + n sin(nθ)].
3 n=1
2 n2

4. After carrying out the integrations, the solution is

∞
1 X (−1)n n  r n
u(r, θ) = − r sin(θ) + 2 sin(nθ).
10 n=2
n2 − 1 5

5. The solution is
∞
sinh(π) 2 X (−1)n  r n
u(r, θ) = + sinh(π)[cos(nθ) + n sin(nθ)].
π π n=1 n2 + 1 4

6. If we write sin2 (θ) = (1 − cos(2θ))/2, we can identify the only nonzero

coefficients in the series solution as a0 = 1/2 and a2 = −1. The solution
is
1 r
u(r, θ) = − cos(2θ).
2 2
7. After the integrations, we obtain the solution
∞
1 X 4(−1)n+1  r n
u(r, θ) = 1 − π 2 + cos(nθ).
3 n=1
n2 8

8. After integrating to find the coefficients, the solution is

2 cosh(2π) − sinh(2π)
u(r, θ) =
4π
∞
2 X (−1)n  r n
+ [cosh(2π)(8π + 2nπ 2 ) + sinh(2π)(n2 + 4)] cos(nθ)
π n=1 (n2 + 4)2 4
∞
2 X (−1)n  r n
+ [cosh(2π)(4nπ + n3 π 3 ) − 4n sinh(2π)] sin(nθ).
π n=1 (n2 + 4)2 4

9. Let U (r, θ) = u(r cos(θ), r sin(θ). The problem given in rectangular coor-
dinates converts to the following problem in polar coordinates:

∇2 U (r, θ) = 0 for 0 ≤ r < 4, U (4, θ) = 16 cos2 (θ).

If we write 16 cos2 (θ) = 8(1 + cos(2θ), we can recognize by inspection that

1
a0 = 8, a2 (42 ) = 8,
2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 206 CHAPTER 7. LAPLACE’S EQUATION

and all other an = 0. The solution in polar coordinates is

 r 2
U (r, θ) = 8 + 8 cos(2θ).
4
Because the original problem was posed in rectangular coordinates, con-
vert this to rectangular coordinates by using x = r cos(θ), y = r sin(θ),
and the identity cos(2θ) = 2 cos2 (θ) − 1, to obtain
1
u(x, y) = 8 + (x2 − y 2 ).
2
10. In polar coordinates this Dirichlet problem is
∇2 U (r, θ) = 0 for 0 ≤ r < 3, U (3, θ) = 3(cos(θ) − sin(θ)).
In the series solution for this problem, identify 3a1 = 3 and 3b1 = 3, with
all other coefficients zero. Then
U (r, θ) = r(cos(θ) − sin(θ)).
In rectangular coordinates, this solution is
u(x, y) = x − y.
Perhaps by hindsight, this could have been seen with no computation
because x − y is harmonic on the entire plane.
11. In polar coordinates this problem is
∇2 U (r, θ) = 0 for 0 ≤ r < 2, U (2, θ) = 4(cos2 (θ) − sin2 (θ)) = 4 cos(2θ).
Identify a2 22 = 4, with all other coefficients zero, so
U (r, θ) = r2 cos(2θ).
In rectangular coordinates the solution is
u(x, y) = x2 − y 2 .

12. In polar coordinates the problem is

∇2 U (r, θ) = 0 for 0 ≤ r < 5, U (5, θ) = 25 sin(θ) cos(θ).
If we write
25
U (5, θ) = sin(2θ)
2
we can identify the coefficients in the series solution: a2 (52 ) = 25/2, with
all other coefficients zero. The solution is
1
U (r, θ) = r2 sin(2θ).
2
To convert this to rectangular coordinates, use
1 2
r sin(2θ) = r sin(θ)r cos(θ)
2
to obtain
u(x, y) = xy.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.3. THE POISSON INTEGRAL FORMULA 207

7.3 The Poisson Integral Formula

In Problems 1–4 the idea is to use the Poisson integral formula to write the re-
quested solution value as an integral which can be approximated by a numerical
technique. This assumes the availability of software that will do this.
For Problem 1, the Poisson integrals are given for each approximate value
calculated. For Problems 2, 3 and 4, we give just the approximate values.

1. With R = 1 and f (θ) = θ, the solution is

Z pi
1 1 − r2
U (r, θ) = 2
ξ dξ.
2π −π 1 + r − 2r cos(ξ − θ)
Then Z π
1 3ξ
U (1/2, π) = dξ = 0.
2π −π 5 − 4 cos(ξ − π)
Next,
Z π
1 7ξ
U (3/4, π/3) = dξ ≈ 0.8826128645.
2π −π 25 − 14 cos(ξ − π/3)
And
Z π
1 24ξ
U (1/5, π/4) = dξ ≈ 0.2465422.
2π −π 26 − 10 cos(ξ − π/4)

2. U (1, π/6) ≈ 0.0033829, U (3, 7π/2) ≈ 0, U (1, π/4) ≈ 0, U (2.5, π/12) ≈

0.132145.
3. U (4, π) ≈ −16.4654, U (12, π/6) ≈ 0.0694, U (8, π/4) ≈ 1.5281
4. U (5.5, 3π/5) ≈ 0.8230986392, U (4, 2π/7) ≈ 1.609925382, U (1, π) ≈ 3.986586385,
U (4, 9π/4) ≈ .461163821

7.4 The Dirichlet Problem for Unbounded Re-

gions
1. If we put f (ξ) = K in equation (7.7), we get the solution
Ky ∞
Z
1
u(x, y) = dξ
π −∞ y 2 + (ξ − x)2
Z L
Ky 1
= lim dξ
π L→∞ −L y + (ξ − x)2
2
   
K L−x −L − x
= lim arctan − arctan
π L→∞ y y
K π π
= + = K.
π 2 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 208 CHAPTER 7. LAPLACE’S EQUATION

Or, we can avoid this computation by observing that u(x, y) = K is har-

monic on the entire plane, and equals K on the real line (the boundary of
the upper half-plane). Therefore the solution is u(x, y) = K.
2. By equation (7.7), the solution of this problem for the upper half-plane is
y ∞
Z
1
u(x, y) = e−|ξ| dξ.
π −∞ y 2 + (ξ − x)2

3. By equation (7.7), the solution is

y ∞
Z
ξ
u(x, y) = dξ.
π −∞ y 2 + (ξ − x)2

4. The solution is
Z k
y 1
u(x, y) = 2 + (ξ − x)2
dξ
π −k y
 
1 ξ−x k
= arctan
π y −k
    
1 k−x −k − x
= arctan − arctan
π y y
    
1 k−x k+x
= arctan + arctan ,
π y y
in which the last line makes use of the fact that the arctangent is an odd
function.
5. Suppose u(x, y) is harmonic on the upper half-plane and u(x, 0) = f (x).
Then the function v(x, y) defined by v(x, y) = u(x, −y) on the lower half-
plane is harmonic, and v(x, 0) = f (x). But we know an integral formula
for u(x, y). Therefore the problem for the lower half-plane has the solution
y ∞
Z
f (ξ)
v(x, y) = u(x, −y) = − dξ.
π −∞ y 2 + (ξ − x)2
for all x and for y < 0.
6. By the result of Problem 5, the solution of this problem for the lower
half-plane is
y ∞
Z
f (ξ)
u(x, y) = − dξ
π −∞ y 2 + (ξ − x)2
y 0 −1 y 1
Z Z
1
=− 2 2
dξ − dξ
π −1 y + (ξ − x) π 0 y + (ξ − x)2
2
      
1 x x+1 x−1
=− 2 arctan − arctan − arctan .
π y y y

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.4. THE DIRICHLET PROBLEM FOR UNBOUNDED REGIONS 209

7. The boundary of the right quarter-plane consists of the nonnegative hor-

izontal and vertical axes. Define two Dirichlet problems, in each of which
boundary data is nonzero on just one part of the boundary:
Problem 1 ∇2 v = 0 for x > 0, y > 0 and v(x, 0) = f (x), v(0, y) = 0, and
Problem 2 ∇2 w = 0 for x > 0, y > 0 and w(x, 0) = 0, w(0, y) = g(y).
Both problems can be solved by separation of variables and Fourier inte-
grals, obtaining

2 ∞
Z Z ∞ 
v(x, y) = f (ξ) sin(ωξ) dξ sin(ωx)e−ωy dω
π 0 0

and Z ∞ Z ∞ 
2
w(x, y) = g(η) sin(ωη) dη sin(ωy)e−ωx dω.
π 0 0

The solution of the original problem is u(x, y) = v(x, y) + w(x, y).

8. Using the result of Problem 7, compute

Z ∞
ω
bω = e−3ξ sin(ωξ) dξ = .
0 9 + ω2

The solution is
Z ∞
2 ω
u(x, y) = sin(ωx)e−ωy dω.
π 0 9 + ω2

9.

Ay 8
Z
1
u(x, y) = dξ
π 4 y 2 + (ξ − x)2
    
A x−4 x−8
= arctan − arctan
π y y

10. The two-dimensional heat equation is

ut = k(uxx + uyy ) = k∇2 u.

In the steady-state case, ut = 0 and u satisfies Laplace’s equation. Use

the result of Problem 7. The solution is
2 ∞
Z
u(x, y) = Bω sin(ωy)e−ωx dω,
π 0

where Z ∞
2 ω
Bω = e−η sin(ωη) dη = .
π 0 1 + ω2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 210 CHAPTER 7. LAPLACE’S EQUATION

11. Using the results of Problem 7, the solution is

Z ∞
u(x, y) = [bω sin(ωx)e−ωy + Bω sin(ωy)e−ωx ] dω,
0

where
2 ∞
Z
bω = ξ sin(ωξ) dξ
π 0
2 sin(πω) − 2ωπ cos(πω)
=
πω 2
and
2 ∞ 2
Z
Bω = η sin(ωη) dη
π 0
4(cos(πω) − 1) − 2ω 2 π 2 cos(πω) + 4π sin(πω)
= .
πω 3

7.5 A Dirichlet Problem in 3 Dimensions

1. Let u(x, y, z) = X(x)Y (y)Z(z) to separate variables, obtaining first that

X 00 + λX = 0; X(0) = X(1) = 0

and
Y 00 + µY − 0; Y (0) = Y (1) = 0.
Then
λn = n2 π 2 , Xn (x) = sin(nπx)
and
µm = m2 π 2 , Ym (y) = sin(mπy).
Further,
Z 00 − (n2 + m2 )π 2 Z = 0; Z(0) = 0.
This leads to functions

unm (x, y, z) = cnm sin(nπx) sin(mπy) sinh(αnm πz),

√
where αnm = n + m2 . To satisfy the condition u(x, y, 1) = xy, use a
superposition
∞ X
X ∞
u(x, y, z) = cnm sin(nπx) sin(mπy) sinh(αnm πz).
n=1 m=1

We must choose the coefficients so that

∞ X
X ∞
u(x, y, 1) = xy = sin(nπx) sin(mπy) sinh(αnm π).
n=1 m=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.5. A DIRICHLET PROBLEM IN 3 DIMENSIONS 211

This is a double Fourier series on the square 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, and

we know from experience with the heat and wave equations that
Z 1 Z 1
4
cnm = ξ sinh(nπξ) dξ η sin(mπη) dη
sinh(αnm π) 0 0
4(−1)n+m
= .
nmπ 2 sinh(αnm π)

2. Again separate variables, except now exploit the zero boundary conditions
on sides z = 0, z = 1, y = 0, y = 2π and x = 0 to obtain Sturm-Liouville
problems for Y (y) and Z(z):

Y 00 + λY = 0; Y (0) = Y (2π) = 0

and
Z 00 + µZ = 0; Z(0) = Z(1) = 0.
Then
n2
λn = , Yn (y) = sin(ny/2)
4
and
µm = m2 π 2 , Zm (z) = sin(nπz).
Further
Xnm (x) = sinh(αnm x/2)
√
where αnm = n2 + 4m2 π 2 . Therefore use the superposition
∞ X
X ∞
u(x, y, z) = cnm sin(ny/2) sin(mπz) sinh(αnm x/2).
n=1 m=1

For the solution, choose the coefficients so that

u(2π, y, z) = z.

Thus choose
1 2π
Z Z 1
1
cnm = 2 sin(nη/2) dη 2 sin(mπτ ) dτ
sinh(αnm π) π 0 0
4
= (1 − (−1)n )(1 − (−1)m ).
sinh(αnm π)

3. The solution is the sum of the solutions of the following two problems:

∇2 w = 0,
w(0, y, z) = w(1, y, z) = w(x, 0, z) = w(x, 2π, z) = w(x, y, 0) = 0,
w(x, y, π) = 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 212 CHAPTER 7. LAPLACE’S EQUATION

and
∇2 v = 0,
v(0, y, z) = v(1, y, z) = v(x, y, 0) = v(x, y, π) = v(x, 0, z) = 0,
v(x, 2π, z) = xz 2 .
Each of these problems is solved by a separation of variables. For the first,
we obtain
X∞ X ∞ p
w(x, yz) = anm sin(nπx) sin(my/2) sinh( 4n2 π 2 + m2 z/2),
n=1 m=1

in which
Z 1 Z 2π
1 1
anm = √ 2 sin(nπξ) dξ sin(nπη) dη
2 2 2
sinh( 4n π + m π/2) 0 0 π
1 − (−1)n 1 − (−1)m
  
1
= √ .
sinh( n2 π 2 + m2 π/2) nπ mπ
For the second problem, obtain
X ∞
∞ X p
v(x, y, z) = bnm sin(nπx) sin(mz) sinh( n2 π 2 + m2 y),
n=1 m=1

in which
Z 1 Z π
4
bnm = √ ξ sin(nπξ) dξ τ 2 sin(mτ ) dτ )
π 2 sinh( n2 π 2 + m2 2π) 0 0
(−1)n+1 2 − 2(−1)m + m2 π 2 (−1)m
  
4
= √ .
π 2 sinh( n2 π 2 + m2 2π) nπ m3

The original problem has solution

u(x, y, z) = w(x, y, z) + v(x, y, z).

4. The solution u(x, y, z) is a sum of the solutions of the following two prob-
lems:
∇2 (v) = 0,
v(0, y, z) = 0, v(1, y, z) = sin(πy) sin(z),
v(x, 0, z) = v(x, 2, z) = 0,
v(x, y, 0) = v(x, y, π) = 0
and
∇2 w = 0,
w(0, y, z) = w(1, y, z) = 0,
w(x, 0, z) = w(x, 2, z) = 0,
w(x, y, 0) = x2 (1 − x)y(2 − y), w(x, y, π) = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.6. THE NEUMANN PROBLEM 213

Solve these problems by separation of variables. The first problem has a

solution of the form
∞ X
X ∞
v(x, y, z) = anm sinh(m2 + (mπ/2)2 ) sin(nπy/2) sin(mz),
n=1 m=1

while the second problem has a solution of the form

∞ X
X ∞ p
w(x, y, z) = bnm sin(nπx) ∼ (mπy/2) sinh( (nπ)2 + (mπ/2)2 (π−z)).
n=1 m=1
√
The coefficients anm can be obtained by inspection. Observe that a21 π 2 + 1 =
1 and all other anm = 0. The coefficients bnm require integrations:

bnm =
Z 1 Z 2
2 2
√ ξ (1 − ξ) sin(nπξ) dξ η(2 − η) sin(mπη/2) dη
sinh(π 4n2 + m2 /2) 0 0
64
= − 6 (1 + 2(−1)n )(1 − (−1)m ).
π
’

7.6 The Neumann Problem

1. First, Z 1
4 cos(πx) dx = 0,
0
so this problem may have a solution. (If this integral were nonzero, we
could conclude that the problem has no solution).
A separation of variables, making use of the three homogeneous boundary
conditions, leads to the problems

X 00 + λX = 0; X 0 (0) = X 0 (1) = 0

and
Y 00 − λY = 0; Y 0 (1) = 0.
These have solutions of the form

λn = n2 π 2 , Xn (x) = cos(nπy)

and
Yn (x) = cosh(nπ(1 − y)).
Thus attempt a solution of the Neumann problem of the form
∞
X
u(x, y) = c0 + cn cosh(nπ(1 − y)) cos(nπy).
n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 214 CHAPTER 7. LAPLACE’S EQUATION

The condition uy (0) = 4 cos(πx) requires that

∞
X
−cn nπ sinh(nπ) cos(nπx) = 4 cos(πx).
n=1

This is satisfied if we put cn = 0 for n = 2, 3, 4, · · · , and choose c1 so that

−c1 π sinh(1) cos(πx) = 4 cos(πx).

Therefore −c1 π sinh(π) = 4, and

4
c1 = − .
π sinh(π)
The solution is
4
u(x, y) = c0 − cosh(π(1 − y)) cos(πx).
π sinh(π)
Here c0 is an arbitrary constant, so this solution is not unique.
2. First check that Z π  π
y− dy = 0,
0 2
so it is worthwhile to look for a solution. From the zero boundary condi-
tions at y = 0 and y = π, separation of variables gives us a solution of the
form
∞
X
u(x, y) = c0 + [cn cosh(nx) + dn cosh(1 − x)] cos(ny).
n=1

The boundary conditions on the edges x = 0 and x = 1 give us

∞
∂u π X
(0, y) = y − = −ndn sinh(n) cos(ny)
∂x 2 n=1

and
∞
∂u X
(1, y) = cos(y) = ncn sinh(n) cos(ny).
∂x n=1

The coefficients are given by

Z π
1 2
cn = cos(y) cos(ny) dy
n sinh(n) π 0
(
0 if n 6= 1,
=
1/ sinh(1) if n = 1

and π
−1 2(1 − (−1)n )
Z  π
dn = y− dy =
n sinh(n) 0 2 πn2 sinh(n)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.6. THE NEUMANN PROBLEM 215

for n = 1, 2, · · · . The solution is

1
u(x, y) = c0 + cosh(x)
sinh(1)
∞
X 2((−1)n − 1)
+ cosh(n(1 − x)) cos(ny).
n=1
πn2 sinh(n)

Rπ
3. A solution may exist because 0 cos(3x) dx = 0. From the zero boundary
conditions on edges x = 0 and x = π, separation of variables yields a
solution of the form
∞
X
u(x, y) = c0 + [cn cosh(ny) + dn cosh(n(π − y))] cos(nx).
n=1

Now
∞
∂u X
(x, 0) = cos(3x) = −ndn sinh(nπ) cos(nx)
∂y n=1
so
1
d3 = −
3 sinh(3π)
and dn = 0 if n 6= 3. Next, the boundary condition at y = π gives us
∞
∂u X
(x, π) = 6x − 3π = ncn sinh(nπ) cos(nx).
∂u n=1

Then
2 π
Z
1
cn = (6x − 3π) cos(nx) dx
n sinh(nπ) π 0
1 12
= ((−1)n − 1).
n sinh(nπ) n2 π

The solution is
cosh(3(π − y))
u(x, y) = c0 − cos(3x)
3 sinh(3π)
∞
X 12((−1)n − 1)
+ cosh(ny) cos(nx).
n=1
n3 π sinh(nπ)

4. Let u(x, y) = X(x)Y (y) and use the boundary conditions to obtain the
problems
X 00 − λX = 0; X 0 (0) = X 0 (π) = 0
and
Y 00 + λY = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 216 CHAPTER 7. LAPLACE’S EQUATION

We find that X0 (x) = 1 and Xn (x) = cos(nx) for n = 1, 2, · · · , while

Yn (y) = cn cosh(ny) + dn cosh(n(π − y))

for n = 1, 2, · · · . We therefore seek a solution of the form

∞
X
u(x, y) = c0 + [cn cosh(ny) + dn cosh(n(π − y))] cos(nx).
n=1

At the edge y = π, we have

∞
X
u(x, y) = 0 = c0 + [cn cosh(nπ) + dn ] cos(nx).
n=1

Along the edge y = 0, we have

∞
X
u(x, 0) = f (x) = c0 + [cn + dn cosh(nπ)] cos(nx).
n=1

In order to have a solution, the coefficients must be chosen to satisfy the

equations

c0 = 0,
cn cosh(nπ) + dn = 0,
2 π
Z
cn + dn cosh(nπ) = f (ξ) cos(nξ) dξ
π 0

for n = 1, 2, · · · . For n ≥ 1, the determinant of the coefficients of this

system of equations is

cosh(nπ) 1
= cosh2 (nπ) − 1 = sinh2 (nπ) 6= 0.
1 cosh(nπ)

Therefore there is a unique solution of these algebraic equations for the

coefficients cn and dn for n = 1, 2, · · · . The problem therefore has a unique
solution (recalling that c0 = 0).

5. With u(x, y) = X(x)Y (y), we obtain:

X 00 − λX = 0

and
Y 00 + λY = 0; Y (0) = Y (1) = 0.
Then

Yn (y) = sin(nπy) and Xn (x) = cn cosh(nπx) + dn cosh(nπ(1 − x))

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.6. THE NEUMANN PROBLEM 217

for n = 1, 2, · · · . Look for a solution of the form

∞
X
u(x, y) = [cn cosh(nπx) + dn cosh(nπ(1 − x))] sin(nπy).
n=1

To solve for the constants, use the other two boundary conditions. First,
∞
∂u X
(1, y) = 0 = nπcn sinh(nπ) sin(nπy)
∂x n=1

so we each cn = 0. Next,
∞
∂u X
(0, y) = 3y 2 − 2y = −nπdn sinh(nπ) sin(nπy).
∂x n=1

Then
Z 1
−2
dn = (3η 2 − 2η) sin(nπη) dη
nπ sinh(nπ) 0
2
= 4 4 [n2 π 2 (−1)n + 6(1 − (−1)n )]
n π sinh(nπ)
for n = 1, 2, · · · . The solution is
u(x, y) =
∞
X 2
4 π 4 sinh(nπ)
[n2 π 2 (−1)n + 6(1 − (−1)n )] cosh(nπ(1 − x)) sin(nπy).
n=1
n
Rπ
6. Because −pi sin(3θ) dθ = 0, this problem may have a solution. Such a
solution will have the form
∞
1 X
u(r, θ) = a0 + [an rn cos(nθ) + bn rn sin(nθ)].
2 n=1

The boundary conditions give us

∂u
(R, θ) = sin(3θ)
∂r
X∞
= [nan Rn−1 cos(nθ) + nbn Rn−1 sin(nθ)].
n=1

By inspection, we see that we can choose each an = 0 and bn = 0 for

n 6= 3. Further, 3b3 R2 = 1, so
1
b3 =
3R2
and the solution is
1 R  r 3
u(r, θ) = a0 + sin(nθ).
2 3 R

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 218 CHAPTER 7. LAPLACE’S EQUATION

Rπ
6. First check that −π cos(2θ) dθ = 0, a necessary condition for a solution
to exist. A solution must have the form
∞
1 X
u(r, θ) = a0 + [an cos(nθ) + bn sin(nθ)].
2 n=1

From the boundary condition at r = R, we have

∂u
(R, θ) = cos(2θ)
∂r
X∞
= [nan Rn−1 cos(nθ) + nbn Rn−1 sin(nθ)].
n=1

As in the preceding problem, compare coefficients on both sides of this

equation to choose each bn = 0 and an = 0 except for n = 2. Further,
2a2 R = 1. The solution is
1 R  r 2
u(r, θ) = a0 + cos(2θ).
2 2 R
8. Because Z ∞
xe−|x| dx = 0
−∞
this problem may have a solution. Using the general solution developed
in Section 7.6.3, we can immediately write the solution
Z ∞
1
u(x, y) = ln(y 2 + (ξ − x)2 )ξe−|ξ| dξ + c.
2π −∞

9. Because Z ∞
e−|ξ| sin(ξ) dξ = 0,
−∞
a necessary condition for a solution to exist is satisfied. The solution is
Z
1
u(x, y) = −∞∞ ln(y 2 + (ξ − x)2 )e−|ξ| sin(ξ) dξ.
2π

10. A solution of the Dirichlet problem for the lowr half-plane was requested
in Problem 5, Section 7.4. Use this result and the line of argument used to
solve the Neumann problem for the upper half-plane to obtain the solution
Z ∞
1
u(x, y) = − ln(y 2 + (ξ − x)2 )f (ξ) dξ.
2π −∞

11. Problem 7, Section 7.4 requested a solution of the Dirichlet problem for
the right quarter-plane. Using this, we are led to attempt a solution for
the Neumann problem for the right quarter-plane of the form
Z ∞
u(x, y) = aω cos(ωx)e−ωy dω.
0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.7. POISSON’S EQUATION 219

Now Z ∞
∂u
(x, 0) = −ωaω cos(ωx) dω.
∂y 0

This tells us to choose

Z ∞
2
aω = − f (ξ) cos(ωξ) dξ.
πω 0

12. Attempt a solution of the form

Z ∞
u(x, y) = bω e−ωy sin(ωx) dω.
0

This is harmonic and has been chosen to satisfy u(0, y) = 0. Now

Z ∞
∂u
(x, 0) = f (x) = −ωbω sin(ωx) dω,
∂y 0

so choose Z ∞
2
bω = − f (ξ) sin(ωξ) dξ.
πω 0

7.7 Poisson’s Equation

1. Write u(x, y) = v(x, y) + w(x, y), where v is the solution of the Dirichlet
problem

∇2 v = 0 for 0 < x < 1, 0 < y < 1,

v(x, 0) = v(x, 1) = 0,
v(1, y) = 0,
v(0, y) = y

and w is the solution of the problem

∇2 w = 0 for 0 < x < 1, 0 < y < 1,

w(0, y) = w(1, y) = w(x, 0) = w(x, 1) = 0,
w(x, y) = xy for 0 < x < 1, 0 < y < 1.

For the first problem, for v(x, y), separate variables to obtain the solution:
∞
X
v(x, y) = an sin(nπ(1 − x)) sin(nπy),
n=1

where
2(−1)n+1
an = .
nπ sinh(nπ)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 220 CHAPTER 7. LAPLACE’S EQUATION

The problem for w(x, y) has the solution

∞ X
X ∞
w(x, y) = knm sin(nπx) sin(mπy),
n=1 m=1

where
Z 1 Z 1
−4
knm = 2 2 ξ sin(nπξ dξ η sin(mπη) dη
π (n + m2 ) 0 0
4(−1)n+m+1
= 2 .
(n + m2 )nmπ 4

2. Split the Poisson problem into two problems, as in Figure 7.6. The first
is a Dirichlet problem:

∇2 v = 0 for 0 < x < 1, 0 < y < 2,

v(0, y = v(1, y) = v(x, 2) = 0, v(x, 0) = x2 .

This has the solution

∞
X
v(x, y) = cn sin(nπx) sinh(nπ(2 − y)),
n=1

where
Z 1
2
cn = ξ 2 sin(nπξ) dξ
sinh(2nπ) 0
−2 + 2(−1)n − n2 π 2 (−1)n
 
2
= .
sinh(2nπ) n3 π 3
The second problem is

∇2 w = 0,
w(x, y) = 0 on the boundary of the rectangle,
w(x, y) = x sin(y) for 0 < x < 1, 0 < y < 2.

This problem has the solution

∞ X
X ∞
w(x, y) = knm sin(nπx) sin(mπy/2)
n=1 m=1

where
Z 1 Z 2
8
knm = − ξ sin(nπξ) dξ η 2 sin(mπη/2) dη
π 2 (4n2 + m2 ) 0 0
16 (−1)n+1 mπ sin(2)(−1)m
= 2 2 2
.
π (4n + m ) nπ −4 + m2 π 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 7.7. POISSON’S EQUATION 221

3. Split the problem into two problems, as we have been doing. However, the
first problem (see Figure 7.6) must itself be broken up into two problems,
in the first of which v(0, y) = 1 and v(π, y) = 0, and in the second of
which v(0, y) = 0 and v(π, y) = 0. Applying straightforward separation of
variables to these problems, we obtain
∞
X 2
v1 (x, y) = (1 − (−1)n ) sin(my) sinh(n(π − x))
n=1
nπ sinh(nπ)

for the Dirichlet problem with v(0, y) = 1 and v(π, y) = 0. If v(0, y) = 0

and v(π, y) = y, we obtain
∞
X 2
v2 (x, y) = (−1)n+1 sin(ny) sinh(nx).
n=1
n sinh(nπ)

For the problem for w in Figure 7.6, we have

∞ X
X ∞
w(x, y) = knm sin(nx) sin(my),
n=1 m=1

where

knm =
Z πZ π Z π
4
− 2 n2 π 2 + m2 π 2 ξ 2 sin(nξ) dξ η 2 sin(mη) dη
π 0 0 0
4
= 2 2 (2 − 2(−1)n + n2 π 2 (−1)n )(−2 + 2(−1)m − m2 π 2 (−1)m ).
π (n + m2 )

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 222 CHAPTER 7. LAPLACE’S EQUATION

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 8

Special Functions and

Applications

8.1 Legendre Polynomials

For Problems 1–4 and 6, graphs of the function and the sixth partial sum of
its Fourier-Legendre expansion on[−1, 1] appear nearly indistinguishable within
the scale of the graph. The “most“ functions many terms of this expansion are
needed to achieve a good fit between the partial sum and the function. This
is seen in Problem 5, where the sixth partial sum is a poor fit to the function,
while the fiftieth partial sum is much closer (though still a poor fit).

1. The coefficients are

Z 1
2n + 1
cn = sin(πx/2)Pn (x) dx.
2 −1

Carrying out these integrations, we obtain

12
c0 = c2 = c4 = 0, c1 = ,
π2
168(π 2 − 10) 660(π 4 − 112π 2 + 1008)
c3 = 4
, c5 = .
π π6
P5
Figure 1 shows a graph of f (x) and n=0 cn Pn (x).

2. The coefficients are

Z 1
2n + 1
cn = e−x Pn (x) dx.
2 −1

223

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 224 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.1: Graph of sin(πx/2) and the sixth partial sum of its Fourier-Legendre
expansion.

The first six coefficients are

1 3 5 35
c0 = (e − e−1 ), c1 = − , c2 = e − e−1 ,
2 e 2 2
35 259 −1 259 −1
c3 = e− e , c4 = 162e − e ,
2 2 2
3619 26741 −1
c5 = e− e .
2 2
Figure 2 is a graph of f (x) = e−x and the sixth partial sum of its Fourier-
Legendre expansion.
3. The coefficients are
Z 1
2n + 1
cn = sin2 (x)Pn (x) dx.
2 −1

The first six are

1 1
c0 = − sin(1) cos(1) + , c1 = c3 = c5 = 0,
2 2
5 15 15
c2 = − sin(1) cos(1) + − cos2 (1),
8 8 4
585 585 531
c4 = − + cos2 (1) + sin(1) cos(1).
32 16 32

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.1. LEGENDRE POLYNOMIALS 225

Figure 8.2: Graph of e−x and the sixth partial sum of its Fourier-Legendre
expansion.

Figure 8.3 shows the function and the sixth partial sum of this Fourier-
Legendre expansion.

4. The coefficients are

Z 1
2n + 1
cn = (cos(x) − sin(x))Pn (x) dx.
2 −1

The first six coefficients are

c0 = sin(1), c1 = −3 sin(1) + 3 cos(1), c2 = 15 cos(1) − 10 sin(1),

c4 = 549 sin(1) − 855 cos(1), c5 = −5940 sin(1) + 9250 cos(1).

Figure 8.4 is a graph of the function and the sixth partial sum.

5. The coefficients are

Z 1
2n + 1
cn = f (x)Pn (x) dx.
2 −1

The first six coefficients are

3 7 11
c0 = c2 = c4 = 0, c1 = , c3 = − , c5 = .
2 8 16

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 226 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.3: Graph of sin2 (x) and the sixth partial sum of its Fourier-Legendre
expansion.

Figure 8.4: Graph of cos(x) − sin(x) and the sixth partial sum of its Fourier-
Legendre expansion.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.1. LEGENDRE POLYNOMIALS 227

Figure 8.5: Graph of f (x) and the sixth partial sum of its Fourier-Legendre
expansion.

Figure 8.6: Graph of f (x) and the fiftieth partial sum of its Fourier-Legendre
expansion.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 228 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.7: Graph of (x + 1) cos(x) and the sixth partial sum of its Fourier-
Legendre expansion.

Figure 8.5 shows a graph of the function and this partial sum. For this
function the sixth partial sum does not fit the function well at all on [−1, 1].
Figure 8.6 shows the fiftieth partial sum, a better fit to the function.
6. The coefficients are
Z 1
2n + 1
(x + 1) cos(x)Pn (x) dx.
n −1

The first six coefficients are

c0 = sin(1), c1 = −3 sin(1) + 6 cos(1), c2 = 15 cos(1) − 10 sin(1)
c3 = 238 sin(1) − 378 cos(1), c4 = 549 sin(1) − 855 cos(1)
c5 = −34969 sin(1) + 54461 cos(1).
Figure 8.7 shows a graph of (x + 1) cos(x) and the sixth partial sum of the
Fourier-Legendre expansion.
7. For n = 7, [n/2] = [7/2] = 3, so
3
X (14 − 2k)!
P7 = (−1)k xn−2k
27 k!(7 − k)!(7 − 2k)!
k=0
429 7 693 5 315 3 35
= x − x + x − x.
16 16 16 16

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.1. LEGENDRE POLYNOMIALS 229

For n = 8, [n/2] = [4] = 4 and

4
X (16 − 2k)!
P8 (x) = (−1)k x8−2k
28 k!(8− k)!(8 − 2k)!
k=0
6435 8 3003 6 3465 4
= x − x + x
128 32 64
315 2 35
− x + .
33 128

For n = 9, [n/2] = [9/2] = 4 and

4
X (18 − 2k)!
P9 (x) = (−1)k x9−2k
29 k!(9− k)!(9 − 2k)!
k=0
12155 9 6435 7 9009 5
= x − x + x
128 32 64
1155 3 315
− x + x.
32 128

For n = 10, [n/2] = [5] = 5 and

5
X (20 − 2k)!
P10 (x) = (−1)k x10−2k
210 k!(10 − k)!(10 − 2k)
k=0
46189 10 109395 8 45045 6
= x − x + x
256 256 128
15015 4 3465 2 63
− x + x − .
128 256 256

8. Put n = 0 into the integral formula to obtain

Z π
1 1
dθ = (π) = 1 = P0 (x).
π 0 π

Next put n = 1 to get

Z π Z π
1  p 
x + x2 − 1 cos(θ) dθ
π 0 0
1h p iπ
= xθ + x2 − 1 sin(θ)
π 0
1
= (xπ) = x = P1 (x).
π

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 230 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

With n = 2, we have

1 π
Z p 2
x + x2 − 1 cos(θ) dθ
π 0
1 π 2
Z p 
= x + 2 x2 − 1 cos(θ) + (x2 − 1) cos2 (θ) dθ
π 0
1 π
Z  
p 1
= x2 + 2 x2 − 1 cos(θ) + (x2 − 1)(1 + cos(2θ)) dθ
π 0 2
 π
1 2 p
2
1 2
= x θ + 2 x − 1 sin(θ) + (x − 1)(θ + sin(2θ)/2)
π 2 0
 
1 2 1 2
= x π + (x − −1)π
π 2
3 1
= x2 − = P2 (x).
2 2

Finally, wit n = 3, we have

1 π
Z p 3
x + x2 − 1 cos(θ) dθ
π 0
1 π 3
Z p 
= x + 3x2 x2 − 1 cos(θ) − 3x3 cos2 (θ) dθ
π 0
1 π
Z p p 
+ −3x cos2 (θ) + x2 x2 − 1 cos3 (θ) − x2 − 1 cos3 (θ) dθ
π 0
 
1 3 3
= x3 π + πx3 − πx
π 2 2
5 3 3
= x − x = P3 (x).
2 2

9. Let Z π n
1  p
Qn (x) = x+ x2 − 1 cos(θ) dθ
π 0

for n = 0, 1, 2, · · · . The strategy is to show that Qn (x) satisfies the same

recurrence relation (8.7) that the Legendre polynomials do. From Prob-
lem 8, we also have that Q0 (x) = P0 (x) and Q1 (x) = P1 (x). Then the
recurrence relation will give us Q2 (x) = P2 (x), and then Q4 (x) = P4 (x),
and so on.
To show that Qn (x) satisfies equation (8.7), first substitute the integral
for Qn (x) into this equation and rearrange terms to obtain

1 π
Z p p 
−n(x2 − 1) sin2 (θ) + x2 − 1 cos(θ)[x + x2 − 1 cos(θ)]
π 0
 p n−1
× x + x2 − 1 cos(θ) dθ.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.1. LEGENDRE POLYNOMIALS 231

Now integrate this by parts, with

 p n
u = x + x2 − 1 cos(θ)

and
dv = cos(θ) dθ
to obtain
Z π n p
1  p
x+ x2 − 1 cos(θ) x2 − 1 cos(θ) dθ
π 0
1 p n−1
= x + x2 − 1 cos(θ) n(x2 − 1) sin2 (θ).
π
Use this in the substitution of Qn (x) into equation (8.7) to show that
Qn (x) satisfies this recurrence relation. This shows that Qn (x) = Pn (x).
10. We want to show that
Z 1
2
Pn2 (x) dx =
−1 2n + 1

for n = 0, 1, 2, · · · . Let An be the coefficient of xn in Pn (x). We have an

explicit formula for An in equation (8.8). For convenience, denote
Z 1
pn = Pn2 (x) dx.
−1

Consider the polynomial

An
q(x) = Pn (x) − xPn−1 (x).
An−1
Both Pn (x) and xPn−1 (x) are of degree n, and all terms involving xn in
q(x) cancel (by design), so q(x) has degree ≤ n − 1. Further,
An
Pn (x) = q(x) + xPn−1 (x).
An−1
Then
An
Pn2 (x) = q(x)Pn (x) + xPn−1 (x)Pn (x).
An−1
Integrate this equation, using Theorem 8.3, to get
Z 1
An
pn = xPn−1 (x)Pn (x) dx.
An−1 −1
Now use the recurrence relation to write
n+1 n
xPn (x) = Pn+1 (x) + Pn−1 (x).
2n + 1 2n + 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 232 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Then
n+1 n
xPn (x)Pn−1 (x) = Pn+1 (x)Pn−1 (x) + P 2 (x).
2n + 1 2n + 1 n−1
Finally, we have
Z 1
An
pn = xPn (x)Pn−1 (x) dx
An−1 −1
 Z 1 Z 1 
An n+1 n 2
= Pn+1 (x)Pn−1 (x) dx + Pn−1 (x) dx .
An−1 2n + 1 −1 2n + 1 −1

Connecting the first and last expressions in this equation, we have

An n
pn = pn−1 .
An−1 2n + 1
This is a recurrence relation for the numbers pn . Because we know An ,
this is
1 · 3 · 5 · · · (2n − 3)(2n − 1) (n − 1)! n
pn = pn−1
n! 1 · 3 · 5 · · · (2n − 3) 2n + 1
2n − 1
= pn .
2n + 1
From this we deduce pn :

1 n 1 1
Z Z
2
p1 = P0 (x)2 dx = dx = ,
3 −1 3 −1 3
3 32 2
p2 = p1 = = ,
5 53 5
5 2
p3 = p3 = ,
7 7
7 2
p4 = p3 = ,
9 9
and so on. A straightforward induction now yields
2
pn = .
2n + 1

11. Put x = t = 1/2 into the generating formula for the Legendre polynomials
to get
∞    n
1 X 1 1
p = Pn .
3/4 n=0 2 2
Then
∞  
2 X 1 1
√ = P
n n
.
3 n=0 2 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.1. LEGENDRE POLYNOMIALS 233

Then
∞  
X 1 1 1
P
n+1 n
=√ .
n=0
2 2 3

12. One way to derive these results is to use the recurrence relation. Another
way is to use the formula given in Problem 7:
[n/2]
X (2n − 2k)!
Pn (x) = (−1)k xn−2k .
2n k!(n − k)!(n − 2k)!
k=0

To do this, first observe that

   
2n + 1 1
= n+ = n,
2 2
so
n
X (4n + 2 − 2k)!
P2n+1 (x) = (−1)k x2n+1−2k .
22n+1 k!(2n + 1 − k)!(2n + 1 − 2k)!
k=0

Each term of this sum involves a positive power of x, so

P2n+1 (0) = 0.

Next,  
2n
= [n] = n,
2
so
n
X (4n − 2k)!
P2n (x) = (−1)k x2n−2k .
22n k!(2n − k)!(2n − 2k)!
k=0

The constant term in this polynomial occurs when k = n, so

(−1)n (2n)!
P2n (0) = .
22n (n!)2

13. Apply the law of cosines to the triangle in the diagram to get

R2 = r2 + d2 − 2rd cos(θ).

Then
R2 r r2
2
= 1 − 2 cos(θ) + 2 .
d d d
Then
1 1d 1 1
ϕ(x, y, z) = = = q .
R dR d 1 − 2 r cos(θ) + r2
d d2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 234 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

For the remainder of the problem, consider two cases on r/d. First, sup-
pose r/d < 1, so r < d. Put x = cos(θ) and t = r/d in the generating
function to obtain
∞  r n
1X
ϕ(r) = Pn (cos(θ)) ,
d n=0 d

or
∞
X 1
ϕ(r) = Pn (cos(θ))rn .
n=0
dn+1

If r/d > 1, so r > d, now write

R2 d d2
= 1 − 2 cos(θ) + .
r2 r r2
Then
r 1
=q .
R d
1 − 2 r cos(θ) + d2
r2

Again, comparing this with the generating function, we have

∞  n
1X d
ϕ(r) = Pn (cos(θ)) ,
r n=0 r

and this is equivalent to

∞
1X n
ϕ(r) = d Pn (cos(θ))e−n .
r n=0

14. With f (ϕ) = ϕ2 , the solution is

X  ρ n
u(ρ, ϕ) = ∞cn Pn (cos(ϕ)),
n=0
R

in which
Z 1
2n + 1
cn = f (arccos(ξ))Pn (ξ), dξ
2 −1
Z 1
2n + 1
= (arccos(ξ))2 Pn (ξ) dξ.
2 −1

Many of these coefficients can be explicitly computed. For example,

1 2 3 10
c0 = π − 2, c1 = − π 2 , c2 = ,
2 8 9
7 2 8 11 2
c3 = − π , c4 = , c5 = − π .
128 25 512

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.1. LEGENDRE POLYNOMIALS 235

With R = 1, use of the twenty-first partial sum of the solution series yields
the following approximations:

u(1, π/4) ≈ 0.615257, u(1, π/6) ≈ 0.272881, u(1, π/8) ≈ 0.154213.

These can be compared to f (ϕ) as a check on their accuracy.

15. With f (ϕ) = sin(ϕ), we have
Z 1
2n + 1
cn = sin(arccos(ξ))Pn (ξ) dξ
2 −1

and
∞
X  ρ n
u(ρ, ϕ) = cn Pn (cos(ϕ)).
n=0
R
In computing the coefficients, use can be made of the identity
p
sin(arccos(ξ)) = 1 − ξ 2 .

With R = 1, and using the twenty-first partial sum of the solution, we

obtain the approximations

u(1, π/4) ≈ 0.707274, u(1, π/6) ≈ 0.500761, u(1, π/8) ≈ 0.382683.

16. With f (ϕ) = ϕ3 , let

Z 1
2n + 1
cn = (arccos(ξ))3 Pn (ξ) dξ
2 −1

and
∞
X  ρ n
u(ρ, ϕ) = cn Pn (n, cos(ϕ)).
n=0
R
With R = 1, the twenty-first partial sum of this series yields the following
approximate values:

u(1, π/4) ≈ 0.476973, u(1, π/6) ≈ 0.143548, u(1, π/8) ≈ 0.060559.

17. With f (ϕ) = 2 − ϕ2 , let

Z 1
2n + 1
cn = (2 − arccos2 (ξ))Pn (ξ) dξ
2 −1

and
∞
X  ρ n
u(ρ, ϕ) = cn Pn (cos(ϕ)).
n=0
R
With R = 1, use the twenty-first partial sum to approximate:

u(1, π/4) ≈ 1.384743, u(1, π/6) ≈ 1.725844, u(1, π/8) ≈ 1.845787.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 236 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

18. Now f (ϕ) = ϕ − cos(ϕ). Let

Z 1
2n + 1
cn = (arccos(ξ) − ξ)Pn (ξ) dξ
2 −1

and
∞
X  ρ n
u(ρ, ϕ) = cn Pn (cos(ϕ)).
n=0
R
With R = 1, the twenty-first partial sum of this series gives us the following
approximate values:

u(1, π/4) ≈ 0.077951, u(1, π/6) ≈ −0.342074, u(1, π/8) ≈ −0.531180.

19. In spherical coordinates, the Dirichlet problem to be solved is:

2 cot(ϕ)
uρρ + uϕϕ + uϕ = 0, R1 < ρ < R2 , −π/2 ≤ ϕ ≤ π/2,
ρ ρ2
u(R1 , ϕ) = T, U (R2 , ϕ) = 0.

This problem can be solved by separation of variables. Let

u(ρ, ϕ) = F (ρ)Φ(ϕ).

This results in:

2 λ
F 00 + F 0 − 2 F = 0
ρ ρ
and
Φ00 + cot(ϕ)Φ0 + λΦ = 0.
The equation for Φ(ϕ) has the bounded solution

Φn (ϕ) = Pn (cos(ϕ)),

corresponding to an eigenvalue λn = n(n + 1) of Legendre’s equation. For

n = 0, 1, 2, · · · , solutions for F (ρ) are

Fn (ρ) = an ρn + bn ρ−n−1 .

Attempt a superposition
∞
X
u(ρ, ϕ) = (an ρn + bn ρ−n−1 )Pn (cos(ϕ)).
n=0

We require that
∞
X
u(R1 , ϕ) = T = (an R1n + bn R1−n−1 )Pn (cos(ϕ)).
n=0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.2. BESSEL FUNCTIONS 237

And the condition at ρ = R2 is that

∞
X
r(R2 , ϕ) = 0 = (an R2n + bn R2−n−1 )Pn (cos(ϕ)).
n=0

Recalling that P0 (cos(ϕ)) = 1, these equations are satisfied if we choose

the coefficients so that

a0 + b0 R1−1 = T, a0 + b0 R2−1 = 0

and, for n = 1, 2, · · · , let an = bb = 0. We should therefore let

T R1 T R1 R2
a0 = and b0 = − .
R1 − R2 R1 − R2

The solution is  
T R1 R2
u(ρ, ϕ) = −1 .
R1 − R2 ρ

8.2 Bessel Functions

1. With f (x) = e−x , the expansion in terms of zero-order Bessel functions is
∞
X
cn J0 (jn x),
n=1

where jn is the nth (in increasing order) positive zero of J0 (x) and
Z 1
2
cn = ξe−ξ J0 (jn ξ) dξ.
J12 (jn ) 0

Figure 8.8 shows the function and the tenth partial sum of this expansion,
and Figure 8.9 shows the twenty-fifth partial sum.

2. With f (x) = x, let jn be the nth positive zero of J2 (x). The Fourier-Bessel
expansion of f (x) on (0, 1) is
∞
X
cn J2 (jn x),
n=0

where Z 1
2
cn = ξ 2 J2 (jn ξ) dξ.
J32 (jn ) 0

Figure 8.10 shows the function and the fifteenth partial sum of this ex-
pansion, and Figure 8.11 has the thirtieth partial sum.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 238 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.8: Graph of e−x and the tenth partial sum of its Fourier-Bessel expan-
sion.

Figure 8.9: Graph of e−x and the twenty-fifth partial sum of its Fourier-Legendre
expansion.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.2. BESSEL FUNCTIONS 239

Figure 8.10: Graph of x and the fifteenth partial sum of its Fourier-Legendre
expansion.

Figure 8.11: Graph of x and the thirtieth partial sum of its Fourier-Legendre
expansion.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 240 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.12: Graph of x2 e−2x and the twentieth partial sum of its Fourier-
Legendre expansion.

3. Let jn be the nth positive zero of J1 (x) and

Z 1
2
cn = 2 ξ 3 e−2ξ J1 (jn ξ) dξ.
J2 (jn ) 0
The Fourier-Bessel expansion is
∞
X
cn J1 (jn x).
n=1

Figure 8.12 shows the function and the twentieth partial sum of this series,
while Figure 8.13 has the fortieth partial sum.
4. Let jn be the nth positive zero of J2 (x) and
Z 1
2
cn = 2 ξ 2 cos(2ξ)J2 (jn ξ) dξ.
J3 (jn ) 0
The Fourier-Bessel expansion is
∞
X
cn J2 (jn x).
n=0

Figure 8.14 is the tenth partial sum of this expansion, and Figure 8.15 the
thirtieth.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.2. BESSEL FUNCTIONS 241

Figure 8.13: Graph of x2 e−2x and the fortieth partial sum of its Fourier-
Legendre expansion.

Figure 8.14: Graph of x cos(x) and the tenth partial sum of its Fourier-Legendre
expansion.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 242 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.15: Graph of x cos(x) and the thirtieth partial sum of its Fourier-
Legendre expansion.

5. Let jn be the nth positive zero of J4 (x), and

Z 1
2
cn = 2 ξ sin(3ξ)J4 (jn ξ) dξ.
J5 (jn ) 0
The Fourier-Bessel expansion is
∞
X
cn J4 (jn x).
n=1

Figure 8.16 shows the function and the twentieth partial sum of this series,
while Figure 8.17 has the fortieth partial sum.
6. Let jn be the nth positive zero of J1 (x). Let
Z 1
2
cn = 2 ξ 3 cos(πξ)J1 (jn ξ) dξ.
J2 (jn ) 0
The Fourier-Bessel expansion is
∞
X
cn J1 (jn x).
n=1

Figures 8.18 and 8.19 compare graphs of the function with the twentieth
and fiftieth partial sums of this expansion, respectively.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.2. BESSEL FUNCTIONS 243

Figure 8.16: Graph of sin(3x) and the twentieth partial sum of its Fourier-
Legendre expansion.

Figure 8.17: Graph of sin(3x) and the fortieth partial sum of its Fourier-
Legendre expansion.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 244 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.18: Graph of x2 cos(πx) and the twentieth partial sum of its Fourier-
Legendre expansion.

Figure 8.19: Graph of x2 cos(πx) and the fiftieth partial sum of its Fourier-
Legendre expansion.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.2. BESSEL FUNCTIONS 245

7. A formal proof that r

2
J1/2 (x) = sin(x)
πx
can be made by using the Maclaurin expansion of sin(x) on the right side
and manipulating the coefficients to obtain the series defining J1/2 (x).
We will be less formal here and essentially carry out this type of argument
using a few terms of the series so that it is apparent what is happening.
Begin with
∞
X (−1)n
J1/2 (x) = x2n+1/2
n=0
22n+1/2 n!Γ(n+ 1/2 + 1)
∞
(−1)n
r
xX
= 2n
x2n
2 n=0 2 n!Γ(n + 1/2 + 1)
r 
x 1 1
= − x2
2 Γ(1/2 + 1) 22 Γ(1 + 1/2 + 1)

1 1 1
+ 4 − 6 x6 + 8 x8 + · · · .
2 2!Γ(2 + 1/2 + 1) 2 3!Γ(3 + 1/2 + 1) 2 4!Γ(4 + 1/2 + 1)

Now we need to know some values of the gamma function. First,

Z ∞
Γ(1/2) = t−1/2 e−t dt.
0

Letting t = u2 , this is
Z ∞
1 −u2
Γ(1/2) = e 2u du
0 u
Z ∞ √
2
=2 e−u du = π.
0

Here we have used the well-known result that

Z ∞
2 1√
e−u du = π,
0 2

which can be derived using double integrals or complex integration, and

is widely used in probability and statistics. Then, using the factorial
property of the gamma function, we can evaluate Γ(n+1/2+1) for various
values of n. In particular,
      √ √
1 1 1 3 π 3 π
Γ 1+ +1 = 1+ Γ 1+ = = ,
2 2 2 2 2 4
      √ √
1 1 1 53 π 3·5 π
Γ 2+ +1 = 2+ Γ 2+ = = ,
2 2 2 2 4 23

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 246 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS
      √
1 1 1 3·5·7 π
Γ 3+ +1 = 3+ Γ 3+ = ,
2 2 2 24
and so on. In general, if n is a positive integer, then
  √
1 3 · 5 · · · (2n + 1) π
Γ n+ +1 = .
2 2n+1
Now back to the series for J1/2 (x). We can now write
22
r 
x 2
J1/2 (x) = √ − 2 √ x2
2 π 2 ·3 π
3
2 24
+ 4 √ x4 − √ x6
2 2!3 · 5 π 6
2 3!3 · 57̇ π
25

8
+ 8 √ x −· .
2 4! · 3 · 5 · 7 · 9 π
This simplifies to
r 
x 2 1 1
J1/2 (x) = √ − √ x2 + √ x4
2 π 3 π 2 · 2 · 35̇ π

1 1
− 2 √ x6 + 3 √ x8 + · · ·
2 3!3 · 5 · 7 π 2 2·3·5·7·9 π
r 
x 1 1 1
= √ 2 − x2 + x4
2 π 3 2·2·3·5

1 1
− 2 x6 + 3 x8 + · · · .
2 3! · 3 · 5 · 7 2 ·2·3·4·3·5·7·9
Finally, write this as
r 
x 1 1 2 1
J1/2 (x) = √ 2 1− x + x4
2 π 2·3 2·2·2·3·5

1 1
− 3 x6 + 4 x8 · · ·
2 3!3 · 5 · 7 2 ·2·3·4·3·5·7·9
r 
2 1 3 1
= x− x + x5
πx 2·3 2·2·2·3·5

1 1
− 3 x7 + 4 x9 − · · ·
2 3! · 3 · 5 · 7 2 ·2·3·4·3·5·7·9
r ∞ n
2 X (−1)
= x2n+1
πx n=0 (2n + 1)!
r
2
= sin(x).
πx
A similar argument shows that
r
2
J−1/2 (x) = cos(x).
πx

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.2. BESSEL FUNCTIONS 247

8. We want to show that

r  
2 sin(x)
J3/2 (x) = f (x) = − cos(x) .
πx x
Using the Maclaurin expansions of sin(x) and cos(x), we obtain
r ∞
(−1)n+1 2n
  X
2 sin(x)
− cos(x) = x .
πx x n=1
(2n + 1)!

We also have
∞
 x 3/2 X (−1)k
J3/2 (x) = x2k .
2 22k k!Γ(k + 3/2 + 1)
k=0

Using the result from Problem 7 that

1 · 3 · · · (2n + 1) · (2n + 3)
Γ(n + 3/2 + 1) = ,
2n+1
it is routine to verify that these expressions for J3/ (x) and f (x) match up
term for term, using the n = k + 1 rule (so the n = 1 term equals the
k = 0 term, the n = 2 term equals the k = 1 term, and so on.
A similar argument shows that
r  
2 cos(x)
J−3/2 (x) = − sin(x) − .
πx x

9. First, recall that J00 (x) = −J1 (x). Then

Z α α
J1 (αs) ds == −J0 (x) = J0 (0) − J0 (α) = 1
0 0

because J0 (0) = 1 and J0 (α) = 0 by choice of α. Now change variables by

s = αx in the integral to get
Z α Z 1
J1 (s) ds = J1 (αx)α dx = 1
0 0

and this implies that Z 1

1
J1 (αx) dx = .
0 α
10. (a) Let u(x) = J0 (αx). Then
u0 (x) = αJ00 (αx) and u00 (x) = α2 J000 (αx)
so
xu00 + u0 + α2 xu = α2 J000 (αx) + αJ00 (αx) + α2 xJ0 (αx)
= α[αxJ000 (αx) + J00 (αx) + αxJ0 (αx)] = 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 248 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

because J0 (αx) satisfies Bessel’s equation of order ν = 0.

Similarly, if v(x) = J0 (βx), then

xv 00 + v 0 + β 2 xv = 0.

(b) Multiply the equation for u by v and the equation for v by u and
subtract the resulting equations to get

xuv 00 − xvu00 + uv 0 − vu0 + (β 2 − α2 )xuv = 0.

This can be written as

(β 2 − α2 )xuv = −[x(uv 0 − vu0 )].

(c) Integrate both sides of the equation derived in part (b) to obtain
Z
(β 2 − α2 ) xJ0 (αx)J0 (βx) dx = −x(βJ0 (αx)J00 (βx) − αJ0 (βx)J00 (αx)).

Upon multiplying through by the −1 coefficient on the right side side of

this equation, we obtain Lommel’s integral.
11. By equation (8.23),
(xn Jn (x))0 = xn Jn−1 (x).
Then Z
xn Jn−1 (x) dx = xn Jn (x).

In similar fashion, equation (8.24) immediately yields the second integral.

12. These results follow from the integrals given in Problem 11. We know
from the first conclusion of Problem 11 that
Z
xn Jn−1 (x) dx = xn Jn (x).

Let s = αx to obtain
Z
αn xn Jn−1 (αx) dx = α2 xn Jn (x).

The second conclusion follows from the second integral formula of Problem
11.
13. Define Z 1
In,k = (1 − x2 )k xn+1 Jn (αx) dx.
0
For part (a), begin with a result from Problem 11:
Z
sn Jn−1 (s) ds = xs Jn (s).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.2. BESSEL FUNCTIONS 249

Replacing n with n + 1, we have

Z
sn+1 Jn (s) ds = sn+1 Jn+1 (s).

Then Z α α
sn+1 Jn (s) ds = sn+1 Jn+1 (s) = αn+1 Jn+1 (α).
0 0

Now let s = αx to get

Z 1
αn+1 xn+1 Jn (αx) dx = αn+1 Jn+1 (α).
0

Then Z 1
1
xn+1 Jn (αx) dx = Jn+1 (α).
0 α
But,
Z 1
In,0 = xn+1 Jn (αx) dx.
0

Therefore
1
Jn+1 (α). In,0 =
α
Now use the integral of Problem 12, with n + 1 in place of n, to write
 
n+1 d 1 n+1
x Jn (αx) = x Jn+1 (αx) .
dx α

Upon substituting this into the definition of In,k , we have

Z 1  
d 1 n+1
In,k = (1 − x2 )k x Jn+1 (αx) dx.
0 dx α

This completes part (b). For part (c), apply integration by parts to the
integral of part (b):
Z 1  
d 1 n+1 2 k
In,k = (1 − x ) x Jn+1 (αx) dx
0 dx α
1 1
= (1 − x2 )k xn+1 Jn+1 (αx)
α 0
1 1 n+1
Z
− x Jn+1 (αx)k(1 − x2 )k−1 (−2x) dx
α 0
2k 1
Z
= (1 − x2 )k−1 xn+2 Jn+1 (αx) dx
α 0
2k
= In+1,k−1 .
α

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 250 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

This relates In,k to the value of this integral when n is increased by 1 and
k is decreased by 1. In particular, if we carry out k repetitions of this
operation, eventually increasing n to n + k, and decreasing k to k to 0, we
obtain
2k
In,k = In+1,k−1
α  
2k 2(k − 1)
= In+2,k−2
α α
2
2 k(k − 1)
= In−2,k−2
α2
22 k(k − 1) 2(k − 2)
 
= In+3,k−2
α2 α
23 k(k − 1)(k − 2)
= In+3,k−3
α3
2k k!
= · · · k In+k,0 .
α
Because k is a positive integer, we can write

k! = Γ(k + 1)

in this expression.
For part (e), combine the results of parts (a) and (d) to write
Z 1
2k Γ(k + 1)
(1 − x2 )k xn+1 Jn (αx) dx = Jn+k+1 (α).
0 αk+1
For part (f), write this equation as
1
αk+1
Z
Jn+k+1 (α) = (1 − x2 )k xn+1 Jn (αx) dx.
2k Γ(k + 1) 0

The rest is just notation to provide the appropriate perspective. In the

last equation, replace α with x and x with t to get
Z 1
xk+1
Jn+k+1 (x) = k tn+1 (1 − t2 )k Jn (xt) dt.
2 Γ(k + 1) 0
Finally, for part (g), let m − n = k + 1 to get
Z 1
2xm−n
Jm (x) = m−n tn+1 (1 − t2 )m−n−1 Jn (xt) dt.
2 Γ(m − n) 0
In these results, it is not necessary that k be an integer, because k! has
been replaced by Γ(k+1), which is defined if k+1 > 0. For the expressions
derived in this problem, it is enough to have n + −1, k > −1 and, in part
(g),m > n > −1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.2. BESSEL FUNCTIONS 251

14. Use a result of part (g) of Problem 13, with n = −1/2 and m > −1/2 to
write
Z 1 1/2
2xm+1/2

2
Jm (x) = m+1/2 t1/2 (1 − t2 )m−1/2 cos(xt) dt
2 Γ(m + 1/2) 0 πxt
Z 1
xm
= m−1 (1 − t2 )m−1/2 cos(xt) dt.
2 Γ(m + 1/2) 0
This is the requested result with m used in place of n.
15. Start with the following result from Problem 14:
Z 1
xm
Mm (x) = m−1 (1 − t2 )m−1/2 cos(xt) dt.
2 Γ(m + 1/2) 0
Make the change of variables t = sin(θ) to get
Z π/2
xm
Jm (x) = m−1 (cos2 (t))m−1/2 cos(x sin(θ)) dθ
2 Γ(m + 1/2) 0
Z π/2
xm
= m−1 cos2m (θ) cos(x sin(θ)) dθ.
2 Γ(m + 1/2) 0

16. Let y = xa Jν (bxc ) and compute the derivatives:

y 0 = axa−1 Jν (bxc ) + xa bcxc−1 Jν0 (bxc )
and
y 00 = a(a − 1)xa−2 Jν (bxc )
+ [2axa−1 bcxc−1 + xa bc(c − 1)xc−2 Jν0 (bxc )]
+ xa b2 c2 x2c−2 Jν00 (bxc ).

Substitute these into the differential equation and simplify to obtain

c2 xa−2 [(bxc )2 Jν00 + bxc Jν0 (bxc ) + ((bxc )2 − ν 2 )Jν (bxc )] = 0.

In Problems 17–24, the strategy is to match the given differential equation

to the differential equation of Problem 16 by choosing a, b, c and ν. This makes
it possible to write a general solution in terms of Bessel functions
17. The differential equation matches that of Problem 16 if
1 7
2a − 1 = − , 2c − 2 = 0, b2 c2 = 1, and a2 − ν 2 c2 = .
3 144
Then
1 1
a=
, b = c = 1, and ν = .
3 4
We can write a general solution
y(x) = c1 x1/3 J1/4 (x) + c2 x1/3 J−1/4 (x).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 252 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

18. We need
4
−(2a − 1) = 1, 2c − 2 = 2, b2 c2 = 4, and a2 − ν 2 c2 = − .
9

Then
1
a = 0, c = 2, b = 1 and ν = .
3
Then
y(x) = c1 J1/3 (x2 ) + c2 J−1/3 (x2 )

is a general solution.

19. Choose a = 3, c = 4, b = 2, ν = 1/2 to get

y(x) = c1 x3 J1/2 (2x4 ) + c2 J−1/2 (2x4 ).

20. Let a = −1, c = b = 2, ν = 3/4 to obtain

1 1
y(x) = c1 J3/4 (2x2 ) + c2 J−3/4 (2x2 ).
x x

21. Let a = 2, c = 3, b = 1, ν = 2/3 to get

y(x) = c1 x2 J2/3 (x3 ) + c2 x2 J−2/3 (x3 ).

22. Choose a = 4, c = 3, b = 2, ν = 3/4 for a general solution

y(x) = c1 x4 J3/4 (2x3 ) + c2 x4 J−3/4 (2x3 ).

23. Here we get a = b = 0, so this method produces only the trivial solution.
However, if the differential equation is multiplied by x2 , we obtain

1
x2 y 00 + xy 0 − y = 0,
16

which is an Euler equation with general solution

y(x) = c1 x1/4 + c2 x−1/4 .

24. With a = −2, c = b = 3 and ν = 1/2, the general solution is

y(x) = c1 x−2 J1/2 (3x3 ) + c2 x−2 J−1/2 (3x3 ).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.3. SOME APPLICATIONS OF BESSEL FUNCTIONS 253

Figure 8.20: First normal mode in Problem 1 at times t = 1/4, 1/2, 3/4.

8.3 Some Applications of Bessel Functions

1. With f (r) = r(1 − r) and g(r) = r2 , the coefficients in the solution are
Z 1
2
an = 2 Rs2 (1 − RS)J0 (jn s) ds
J1 (jn ) 0

and
Z 1
R 2
bn = R2 s3 J0 (jn s) ds.
jn c J12 (jn ) 0

The solution is
∞
X
z(r, t) = zn (r, t),
n=1

where
      
jn ct jn ct jn
zn (r, t) = an cos + bn sin J0 r .
R R R

Figures 8.20 through 8.23 show graphs of the first four normal modes of
the solution times t = 1/2, 1, 2 and 4, for R = 1, c = 2, f (r) = r(1 − r)
and g(r) = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 254 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.21: Second normal mode in Problem 1.

Figure 8.22: Third normal mode in Problem 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.3. SOME APPLICATIONS OF BESSEL FUNCTIONS 255

Figure 8.23: Fourth normal mode in Problem 1.

2. With f (r) = r cos(πr/2) and g(r) = e−r , the coefficients an and bn in the
nth normal mode are
Z 1
2
an = 2 Rs2 cos(πRs/2)J0 (Rs) ds
J1 (jn ) 0

and Z 1
R 2
bn = se−Rs J0 (jn s) ds.
jn c J12 (jn ) 0

With R = 1, c = 2, f (r) = r cos(πr/2 and g(r) = 0, Figures 8.24–8.27

show the first four normal modes for times t = 1/4, 1/2, 3/4.

3. With f (r) = r2 (1 − r) and g(r) = r + r2 , the coefficients are

Z 1
2
an = R2 s3 (1 − Rs)J0 (jn s) ds
J12 (jn ) 0

and Z 1
R 2
bn = 2 Rs2 (1 + Rs)J0 (jn s) ds.
jn c J1 (jn ) 0

With R = 1, c = 2, f (r) = r2 (1 − r) and g(r) = 0, Figures 8.28–8.31 show

the first four normal modes for times t = 1/4, 1/2, 3/4.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 256 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.24: First normal mode in Problem 2.

Figure 8.25: Second normal mode in Problem 2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.3. SOME APPLICATIONS OF BESSEL FUNCTIONS 257

Figure 8.26: Third normal mode in Problem 2.

Figure 8.27: Fourth partial sum in Problem 2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 258 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.28: First normal mode in Problem 3.

Figure 8.29: Second normal mode in Problem 3.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.3. SOME APPLICATIONS OF BESSEL FUNCTIONS 259

Figure 8.30: Third normal mode in Problem 3.

Figure 8.31: Fourth normal mode in Problem 3.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 260 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.32: First normal mode in Problem 4.

4. The coefficients in the solution are given by

Z 1
2
an = 2 s sin(πs)J0 (jn s) ds.
J1 (jn ) 0
Figures 8.32–8.35 show the first four normal modes, with R = 1, c = 2,
g(r) = 0 and f (r) = sin(πr).
In each of Problems 5–8, the solution has the form
∞
X 2
u(r, t) = an J0 (jn r)e−2jn t .
n=1

For each, the expression for the coefficients is given.

5. Z 1
2
an = ξ(1 + cos(πξ))J0 (jn ξ) dξ.
J12 (jn ) 0

6. Z 1
2
an = 2 ξ(1 − ξ 2 )J0 (jn ξ) dξ.
J1 (jn ) 0

7. Z 1
2
an = 2 ξ 2 cos(3πξ/2)J0 (jn ξ).
J1 (jn ) 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.3. SOME APPLICATIONS OF BESSEL FUNCTIONS 261

Figure 8.33: Second normal mode in Problem 4.

Figure 8.34: Third normal mode in Problem 4.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 262 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.35: Fourth normal mode in Problem 4.

8. Z 1
2
an = 2 ξ(2ξ 3 − ξ − 1)J0 (jn ξ) dξ.
J1 (jn ) 0

Figures 8.36–8.39 show graphs of u(r, t0 ) for t0 equal to 1/10, 1/5, 1/7
and 2/5.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 8.3. SOME APPLICATIONS OF BESSEL FUNCTIONS 263

Figure 8.36: Temperature profiles at specific times in Problem 5.

Figure 8.37: Temperature profiles in Problem 6.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 264 CHAPTER 8. SPECIAL FUNCTIONS AND APPLICATIONS

Figure 8.38: Temperature profiles in Problem 7.

Figure 8.39: Temperature profiles in Problem 8.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 9

Transform Methods of
Solution

9.1 Laplace Transform Methods

1. Apply the Laplace transform with respect to t to the partial differential
equation to get
K
s2 Y (x, s) = c2 Y 00 (x, s) + .
s
Here primes denote differentiation with respect to x and the initial condi-
tions have been inserted in using the operational rule for the derivatives.
Write this equation as
s2 K
Y 00 − 2 Y = − 2 .
c c s
Think of this as a second-order differential equation in x, with s carried
along as a parameter. The general solution is
K
Y (x, s) = c1 esx/c + c2 e−sx/c + .
s3
Here c1 and c2 are “constants”, but may involve s, because x is the variable
of the differential equation. Now
K
Y (0, s) = [y(0, t)](s) = F (s) = c1 + c2 + .
s3
We need limx→∞ y(0, t) = 0, so lims→∞ Y (x, s) = 0. Therefore c1 = 0
and
K
c2 = F (s) − 3 .
s
We now have  
K K
Y (x, s) = F (s) − 3 e−sx/c + 3 .
s s

265

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 266 CHAPTER 9. TRANSFORM METHODS OF SOLUTION

The solution is the inverse Laplace transform of Y (x, s). Recalling the
formula for the inverse Laplace transform of e−as F (s), we obtain
   
x K  x 2 x 1 2
y(x, t) = f t − − t− H t− + Kt ,
c 2 c c 2

in which H is the Heaviside function.

2. Apply the Laplace transform with respect to t to the partial differential
equation to get

9s2 Y (x, s) + Y 00 (x, s) − 6sY 0 (x, s) = 0.

Then
Y 00 − 6sY 0 + 9s2 Y = 0.
This second-order differential equation has characteristic equation

r2 − 6sr + 9s2 = 0,

with repeated root r = 3s. Then

Y (x, s) = c1 e3sx + c2 xe3sx .

Now,
L[y(0, t)](s) = Y (0, s) = 0 = c1 ,
so
Y (x, s) = c2 xe3xs .
Next,
L[y(2, t)](s) = F (s) = 2c2 e6s ,
so
1 −6s
c2 = e F (s).
2
Then
1 −6s 1
e F (s)xe3xs = xF (s)e(3x−6)s .
Y (x, s) =
2 2
The solution is obtained by inverting this to get
1
y(x, t) = xf (t − (6 − 3x))H(t − (6 − 3x)).
2

3. From the partial differential equation and the initial conditions,

A
s2 Y (x, s) = c2 Y 00 − .
s2
Then
s2 A
Y 00 − 2
= 2,
c s

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 9.1. LAPLACE TRANSFORM METHODS 267

with general solution

A
Y (x, s) = c1 esx/c + c2 e−sx/c − .
s4
Because limx→∞ y(x, t) = 0, we must have lims→∞ Y (x, s) = 0, so c1 = 0
and
A
Y (x, s) = c2 e−sx/c − 4 .
s
Next, y(0, t) = 0, so
A
Y (0, s) = c2 − 4
s
and then
A
c2 = 4 .
s
Then
A A
Y (x, s) = 4 e−sx/c − 4 .
s s
The solution is the inverse of this,
A x 3  x A 3
y(x, t) = t− H t− − t .
6 c c 6

5. Transform the partial differential equation to get

Ax
s2 Y (x, s) = c2 Y 00 (x, s) − .
s2
Then
s2 Ax
Y 00 − 2
Y = 2 2.
c c s
This has general solution
Ax
Y (x, s) = c1 esx/c + c2 e−sx/c − .
s4
The condition that limx→∞ y(x, t) = 0 forces lims→∞ Y (x, s) = 0, so
c1 = 0 and
Ax
Y (x, t) = c2 e−sx/c − 4 .
s
Next,
L[y(0, t)](s) = F (s) = Y (0, s) = c2 .
Then
Ax
Y (x, s) = F (s)e−sx/c − .
s4
Invert this for the solution
 x  x 1
y(x, t) = f t − H t− − Axt3 .
c c 6

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 268 CHAPTER 9. TRANSFORM METHODS OF SOLUTION

6. Take the Laplace transform (in t) in the heat equation and use the initial
conditions to get
s
U 00 − U = 0,
k
with general solution
√ √
U (x, s) = c1 e s/kx + c2 e− s/kx .

Because limx→∞ u(x, t) = 0, we have lims→∞ U (x, s) = 0, so c1 = 0 and

√
U (x, s) = c2 e− s/kx .

Next take the transform of u(0, t) = t2 to get

2
U (0, s) = = c2 ,
s3
so
2 −√s/kx
U (x, s) =
e .
s3
As preparation for using the convolution theorem, think of this as the
product
2 1 −√s/kx
 
U (x, s) = 2 e .
s s
By the convolution theorem, the inverse of this is the solution
 
x
u(x, t) = 2t ∗ erfc √ .
2 kt

7. Take the transform with respect to t of the heat equation to get

sU (x, s) − e−x = kU 00 (x, s),

or
s 1
U = − e−x .
U 00 −
k k
The associated homogeneous equation of this nonhomogeneous equation
has the general solution
√ √
Uh (x, s) = c1 e s/kx + c2 e− s/kx .

For a particular solution, use undetermined coefficients, trying

Up (x, s) = Ae−x .

Substitute this into the nonhomogeneous differential equation to get

s 1
A− A=− ,
k k

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 9.1. LAPLACE TRANSFORM METHODS 269

so
s−k
A=
.
Then,
√ √ 1 −x
U (x, s) = Uh (x, s) + Up (x, s) = c1 e s/kx
+ c2 e− s/kx + e .
s−k
Because limx→∞ u(x, t) = 0, choose c1 = 0, so
√ 1 −x
U (x, s) = c2 e− s/kx + e .
s−k
Take the transform of u(0, t) = 0 to get

1
U (0, s) = c2 + .
s−k
Then
1
c2 = −
s−k
so
1 −√s/kx 1 −x
U (x, s) = − e + e .
s−k s−k
Using the convolution theorem, write
h √ i
u(x, t) = −e−kt ∗ L−1 e− s/kx (t) + ekt e−x .

By consulting a table, we obtain

h √ i x 2
L−1 e−(x/ k)s (t) = √ e−x /4kt .
2 πkt 3

Then,
x 2
u(x, t) = −e−kt ∗ √ e−x /4kt + ekt−x .
2 πkt 3

8. Apply the transform (in t) to the heat equation to get

s
U 00 − U = 0,
k
with general solution
√ √
U (x, s) = c1 e s/kx
+ c2 e− x/kx .

Now u(x, 0) forces c1 = −c2 and the transformed solution has the form
p
U (x, s) = c sinh( s/kx).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 270 CHAPTER 9. TRANSFORM METHODS OF SOLUTION

With u(1, t) = f (t), we get U (1, s) = F (s), so

F (s)
c= p .
sinh( s/k)
Then p
sinh( s/kx)
U (x, s) = p F (s)
sinh( s/k)
and the solution is
u(x, t) = L−1 [U (x, s)](t).

9.2 Fourier Transform Methods

For the first four problems, the solution is given by
Z ∞
1 2
u(x, t) = √ f (ξ)e−(x−ξ) /4kt .
2 πkt −∞

1. With f (x) = e−4|x| , the solution is

Z ∞
1 2
√ e−4|ξ| e−(x−ξ) /4kt dξ.
2 πkt −∞

2. The solution is
Z π
1 2
u(x, t) = √ sin(ξ)e−(x−ξ) /4kt
dξ.
2 πkt −π

3. Z 4
1 2
u(x, t) = √ ξe−(x−ξ) /4kt
dξ.
2 πkt 0

4. Z 1
1 2
u(x, t) = √ e−ξ−(x−ξ) /4kt
dξ.
2 πkt −1

5. Take the Fourier transform of the wave equation with respect to x to get

yb00 = 144(iω)2 yb = −144ω 2 yb,

or
yb00 + 144ω 2 yb = 0.
Then
yb(ω, t) = c1 cos(12ωt) + c2 sin(12ωt),
in which primes denote differentiation with respect to t. Because yt (x, 0) =
0, then c1 = 0 and
yb(ω, t) = c1 cos(12ωt).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 9.2. FOURIER TRANSFORM METHODS 271

Next, y(x, 0) = f (x), so

yb(ω, 0) = fb(ω).

Then c1 = fb(ω), so
yb(ω, t) = fb(ω) cos(12ωt).
It is routine to compute (or use a software routine to find)

10
fb(ω) = .
25 + ω 2
Then
10
yb(ω, t) = cos(144ωt).
25 + ω 2
Finally, use the integral formula for the inverse Fourier transform to obtain
 Z ∞ 
1 10 iωx
y(x, t) = Re cos(12ωt)e dω .
2π −∞ 25 + ω 2

Of course y(x, t) is a real quantity, so in the last line we have taken the
real part of the integral. If we replace

eiωx = cos(ωx) + i sin(ωx)

in this integral, we can write more explicitly

Z ∞
1 10
y(x, t) = cos(ωx) cos(12ωt) dω.
2π −∞ 25 + ω 2

6. Apply the Fourier transform (in x) to the initial-boundary value problem

to get

yb00 + 64ω 2 yb = 0,
Z 8
1 − 8iω − e−8ωi
yb(ω, 0) = (8 − ξ)e−iωξ dξ = ,
0 ω2
yb0 (ω, 0) = 0.

This problem has the solution

1 − 8ωi − e−8ωi
yb(ω, t) = cos(8ωt).
ω2
Invert this to obtain
∞
1 − 8ωi − e−8ωi
 Z 
1 iωx
y(x, t) = Re cos(8ωt)e dω .
2π −∞ ω2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 272 CHAPTER 9. TRANSFORM METHODS OF SOLUTION

7. Apply the Fourier transform to the initial-boundary value problem to get

yb00 + 16ω 2 yb = 0,
yb(ω, 0) = 0,
Z π
2i sin(πω)
yb0 (ω, 0) = sin(ξ)e−iωξ dξ = .
−π ω2 − 1

The solution of the transformed problem is

2i sin(πω)
yb(ω, t) = sin(4ωt).
4ω(ω 2 − 1)

Invert this to obtain the solution

 Z ∞ 
1 i sin(πω) iωx
y(x, t) = Re sin(4ωt)e dω .
2π −∞ 2ω(ω 2 − 1)

8. The initial-boundary value problem transforms (in x) to

yb00 + ω 2 y = 0,
Z 2
2
yb(ω, 0) = (2 − |ξ|)e−iωξ dξ = 2 (1 − cos(2ω)),
−2 ω
yb0 (ω, 0) = 0.

This has solution

2
yb(ω, t) = (1 − cos(2ω)) cos(ωt)eiωx dω.
ω2

9. The problem transforms (in x) to

yb00 + 9ω 2 yb = 0,
yb(ω, 0) = 0,
(2 − iω)e−(2+iω)
yb0 (ω, 0) = F[e−2x H(x − 1)](ω) = .
4 + ω2

This problem has the solution

(2 − iω)e−(2+iω)
yb(ω, t) = .
3ω(4 + ω 2 )

Then
∞
(2 − iω)e−(2+iω)
Z 
1 iωx
y(x, t) = sin(3ωt)e dω .
2π −∞ 3ω(4 + ω 2 )

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 9.3. FOURIER SINE AND COSINE TRANSFORM METHODS 273

10. Transform the problem to get

yb00 + 4ω 2 yb = 0,
yb(ω, 0) = 0,
Z 2
2(1 − cos(2ω))
yb0 (ω, 0) = g(ξ)e−iωξ dξ = .
−2 ω
This has the solution
1 − cos(2ω)
yb(ω, t) = sin(2ωt).
ω2
Invert this to get
 ∞ 
1 − cos(2ω)
Z
1 iωx
y(x, t) = Re sin(2ωt)e dω .
2π −∞ ω2

11. This problem was solved by the Fourier transform in the text, so we can
use the result to immediately write
Z 0 Z 4 
y −1 1
u(x, y) = dξ + dξ
π −4 y 2 + (ξ − x)2 2
0 y + (ξ − x)
2
    
y x+4 x−4
=− arctan + arctan .
π y y

12. The solution is

y 1
Z
ξ
u(x, y) = dξ
π 0 y 2 + (ξ − x)2
1 
ln(x2 + y 2 ) − ln((x − 1)2 + y 2 ) − 2x arctan(x/y) + 2x arctan((x − 1)/y) .

−
2π

9.3 Fourier Sine and Cosine Transform Methods

1. Take a Fourier sine transform (in x) of the problem to get
ybS00 + 9ω 2 ybS = 0,
Z 1
2(1 − cos(ω) − ω sin(ω)
ybS (ω, 0) = ξ(1 − ξ) sin(ωξ) dξ = ,
0 ω3
ybS0 (ω, 0) = 0.
The solution of the transformed problem is
 
2(1 − cos(ω)) − ω sin(ω)
ybS (ω, t) = cos(3ωt).
ω3
Invert this to obtain the solution
2 ∞ 2(1 − cos(ω)) − ω sin(ω)
Z  
y(x, t) = sin(ωx) cos(3ωt) dω.
π 0 ω3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 274 CHAPTER 9. TRANSFORM METHODS OF SOLUTION

2. Apply the sine transform to the problem to get

ybS00 + 9ω 2 ybS = 0,
ybS (ω, 0) = 0,
Z 11
0 2(cos(4ω) − cos(11ω))
ybS = 2 sin(ωξ) dξ = .
4 ω
This problem has the solution
2(cos(4ω) − cos(11ω))
ybS (ω, t) = sin(3ωt).
3ω 2
Invert this to obtain the solution
Z ∞
4 cos(4ω) − cos(11ω)
y(x, t) = sin(ωx) sin(3ωt).
3π 0 ω2

3. The sine transformed problem is

ybS00 + 4ω 2 ybS = 0,
ybS (ω, 0) = 0,
Z 5π/2
sin(ωπ/2) − sin(5ωπ/2)
ybS0 (ω, 0) = cos(ξ) sin(ωξ) dξ = .
π/2 ω2 − 1

This has solution

sin(ωπ/2) − sin(5ωπ/2)
ybS (ω, t) = sin(2ωt).
2ω(ω 2 − 1)
Invert this to obtain the solution
2 ∞ sin(ωπ/2) − sin(5ωπ/2)
Z
y(x, t) = sin(ωx) sin(2ωt).
π 0 2ω(ω 2 − 1)

4. Upon taking the sine transform (in x) of the problem, we get

ybS00 + 36ω 2 ybS = 0,

Z ∞
ybS (ω, 0) = −2e−ξ sin(ωξ) dξ
0
ybS0 (ω, 0) = 0.

This has the solution

2ω
ybS (ω, t) = − cos(6ωt).
1 + ω2
Invert this to obtain
Z ∞
4 ω
y(x, t) = − sin(ωx) cos(6ωt).
π 0 1 + ω2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 9.3. FOURIER SINE AND COSINE TRANSFORM METHODS 275

5. After applying the sine transform to the initial-boundary value problem,

we obtain
ybS00 + 196ω 2 ybS = 0,
ybS (ω, 0) = 0,
Z 3
ybS0 (ω, 0) = ξ 2 (3 − ξ) sin(ωξ) dξ
0
3
= 4 (2 sin(3ω) − 4ω cos(3ω) − 3ω 2 sin(3ω) − 2ω).
ω
The transformed problem has the solution
ybS (ω, t) =
3
(2 sin(3ω) − 4ω cos(3ω) − 3ω 2 sin(3ω) − 2ω) sin(14ωt).
14ω 5
Then Z ∞
2
y(x, t) = ybS (ω, t) sin(ωx) dω.
π 0

6. Apply the Fourier sine transform with respect to x to the problem to get
b0S + ω 2 u
u bS + tb
uS = 0,
2ω
bS (ω, 0) = FS xe−x =
 
u .
(1 + ω 2 )2
This problem (involving a linear first-order differential equation) has the
solution
2ω 2 2
u
bS (ω, t) = e−(ω t−t /2) .
(1 + ω 2 )2
Invert this to solve the problem:
4 ∞
Z
ω 2 2
u(x, t) = 2 2
e−ω t−t /2 sin(ωx) dω.
π 0 (1 + ω )

7. Apply the Fourier cosine transform in x to the problem to get

b0C + (1 + ω 2 )b
u uC = −f (t); u
bC (ω, 0) = 0.
This has the solution
Z t
2 2
bC (ω, t) = e−(1+ω
u )t
f (τ )e(1+ω )τ
dτ
0
−(1+ω 2 )t
= −f (τ ) ∗ e .
Invert this to obtain
Z ∞
2 2
u(x, t) = − f (t) ∗ e−(1+ω )t
cos(ωx) dω.
π 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 276 CHAPTER 9. TRANSFORM METHODS OF SOLUTION

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 10

Vectors and the Vector

Space Rn

10.1 Vectors in the Plane and 3−Space

1.
√ √
F + G = (2 + 2)i + 3j, F − G = (2 − 2)i − 9j + 10k,
√ √
2F = 4i − 6j + 10k, 3G = 3 2i + 18j − 15k, k F k= 38

2.

F + G = i + 4j − 3k, F − G = i − 4j − 3k,
√
2F = 2i − 6k, 3G = 12j, k F k= 10

3.

F + G = 3i − k, F − G = i − 10j + k,
√
2F = 2i − 6k, 3G = 3i + 15j − 3k, k F k= 29

4.
√ √
F + G = ( 2 + 8)i + j − 4k, F − G = ( 2 − 9)i + j − 8k,
√
2F = 4i − 6j + 10k, 3G = 24i + 6k, k F k= 39

5.

F + G = 3i − j + 3k, F − G = −i + 3j − k,
√
2F = 2i + 2j + 2k, 3G = 6i − 6j + 6k, k F k= 3

277

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 278 CHAPTER 10. VECTORS AND THE VECTOR SPACE RN

6. The vector
F = −5i + j − 2k
is in the direction
√ from the first point to the second, but this vector has
length k F k= 30. The vector

5
√ F
30
has length 5 and extends from the first point to the second.

7.
9
√ (−5i − 4j + 2k)
45
8.
12
√ (10i − 3j − 4k)
125
9.
4
(−4i + 7j + 4k)
9
10. The vector M = i + j − 3k is oriented from (1, 0, 4) to (2, 1, 1). The vector

V = (x − 1)i + yj + (z − 4)k

is parallel to M exactly when (x, y, z) is on the line through the two given
points. Then, for some t, V = tM, so

(x − 1)i + yj + (z − 4)k = t(i + j − 3k).

Then
x − 1 = t, y = t, z − 4 = −3t.
Then
x = 1 + t, y = t, z = 4 − 3t
are parametric equations of the line through (1, 0, 4) and (2, 1, 1). Here t
varies over all real values. As a check, notice that these points are on the
line given by these parametric equations for t = 0 and t = 1 respectively.

11. x = 3 − 6t, y = 1 − t, z = 0

12. x = 2, y = 1, z = 1 − 3t This line is parallel to the z− axis, and also passes

through (2, 1, 0) when t = 0.

13. x = 0, y = 1 − t, z = 3 − 2t

14. x = 1 − 3t, y = −2t, z = −4 + 9t

15. x = 2 − 3t, y = −3 + 9t, z = 6 − 2t

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 10.2. THE DOT PRODUCT 279

10.2 The Dot Product

In Problems 1–6, F is the first vector given in the problem G is the second, and
θ is the angle between F and G.

1. F · G = 2 and
F·G 2
cos(θ) = =√ .
k F kk G k 14
F and G are not orthogonal.
√
2. F · G = 8, cos(θ) = 8/ 82 and the vectors are not orthogonal.
√ √
3. F · G = −23, cos(θ) = −23/ 29 41 and the vectors are not orthogonal.
√ √
4. F · G = −63, cos(θ) = −63/ 75 74 and the vectors are not orthogonal.
5. F · G = −18, cos(θ) = −9/10 and the vectors are not orthogonal.
6. F · G = 4, cos(θ) = 2/3 and the vectors are not orthogonal.

In Problems 7–12, if the given point is (x0 , y0 , z0 ) and the normal to the
proposed plane is N = ai + bj + ck, then the plane through the point and having
N as normal vector has equation

a(x − x0 ) + b(y − y0 ) + c(z − z0 ) = 0.

The constant terms are usually collected to write this equation in the form

ax + by + cz = k.

7. The plane has equation

3(x + 1) − (y − 1) + 4(z − 2) = 0,

or
3x − y + 4z = 4.

8. x − 2y = −1
9. 4x − 3y + 2z = 25
10. −3x + 2y = 1
11. 7x + 6y − 5z = −26
12. 4x + 3y + z = −6

In each of Problems 13–17, the projection of v onto u is the vector

u·v
u.
k u k2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 280 CHAPTER 10. VECTORS AND THE VECTOR SPACE RN

13.
9
proju v = − u
14

14.
11
− u
30

15.
1
u
62

16.
73
u
101

17.
15
u
53

10.3 The Cross Product

1.
i j k
F × G = −3 6 1 = 8i + 2j + 12k
−1 −2 1
and
i j k
G × F = −1 −2 1 = −8i − 2j − 12k = −F × G
−3 6 1

2.
F × G = i + 12j + 6k = −G × F

3.
F × G = −8i − 12j − 5k = −G × F

4.
F × G = 112k = −G × F

In Problems 5–9, the three given points are used to find two nonparallel
vectors in the plane that is sought (assuming that the points are not collinear).
The cross product of these vectors is a normal to the plane. We than have a
point on the plane and a normal to the plane, so we can find an equation of the
plane.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 10.3. THE CROSS PRODUCT 281

5. Vectors from the first point to the second and third points are F = 4i −
j − 6k and G = i − k. Compute

N = F × G = i − 2j + k.

Because this cross product is not the zero vector, the given points are not
collinear. The plane containing these points has equation

((x + 1)i + (y − 1)j + (z − 6)k) · N = 0.

This is
x + 1 − 2(y − 1) + z − 6 = 0,
or
x − 2y + z = 3.

6. The points are not collinear and the plane containing them has equation
x + 2y + 6z = 12.

7. 2x − 11y + z = 0

8. 5x + 16y − 2z = −1

9. 29x + 37y − 12z = 30

In Problems 10–12, recall that a plane ax + by + cz = k has normal vector

ai + bj + ck, or any nonzero multiple of this vector.

10. N = 8i − j + k, or any nonzero scalar multiple of this vector.

11. N = i − j + 2k

12. n = i − 3j + 2k

13. If two sides of a parallelogram meet at a point, with an angle θ between

the sides, then the area of the parallelogram is the product of the lengths
of these sides, times the cosine of θ. Now suppose F and G are vectors
drawn from a corner of the parallelogram, along two incident sides, and θ
the angle between these vectors. Then

k F kk G k cos(θ)

is the product of the lengths of incident sides, times the angle between
them, hence is the area of the parallelogram. But

k F kk G k cos(θ) =k F × G k

so the magnitude of this cross product is the area of the parallelogram.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 282 CHAPTER 10. VECTORS AND THE VECTOR SPACE RN

14. The vector N = G × H is normal to the base of the parallelopiped (box)

having sides along F, G, H (drawn as arrows from a corner of the box).
From Problem 13, the area of the base of this box is N. Then
(|G × F) · F = F · N =k N kk F k cos(θ)
is the magnitude of the volume of the box having incident sides F, G, H,
because
k F k cos(θ) = ± altitude of the box.

10.4 n− Vectors and the Algebraic Structure of

Rn
1. Let F =< −2, 1, −1, 4 >. Then O =< 0, 0, 0, 0 > is in S (the scalar
multiple of 0 with F). Further, if a and b are real numbers, then
aF + bF = (a + b)F,
so a sum of scalar multiples of F is again a scalar multiple of F, hence is
in S. Finally,
a(bF) = (ab)F
so a scalar multiple of a vector in S is in S. Therefore S is a subspace of
R4 .
2. < 0, 0, 0, 0, 0, 0 > is in S, having zero third and fifth components. Further,
if F and G have zero third and fifth components, so does any scalar
multiple of these vectors, as well as any sum of scalar multiples of these
vectors. Therefore S is a subspace of R6 .
3. S is not a subspace of R5 . The zero vector does not have fourth component
1, and a sum of vectors with fourth component 1 does not have fourth
component 1. S fails on scalar multiples as well.
4. S is not a subspace of R8 . The zero vector is in S, but a sum of vectors of
length less than 1 need not have length less than 1, and scalar multiples
could also fail to have length less than 1.
5. S is not a subspace of R4 . For example, < 1, 1, 1, 0 > and < 0, 1, 1, 1 >
are both in S, but their sum is not, having no zero component.
6. The zero vector is in S, and a linear combination of two vectors with equal
first and fourth components will have equal first and fourth components.
S is a subspace of R5 .
In Problems 7–16, keep in mind that a set of vectors is linearly dependent if
some linear combination of these vectors, with at least one nonzero coefficient,
is equal to the zero vector. And the vectors are linearly independent exactly
when the only linear combination of these vectors adding up to the zero vector
is the trivial one, with all zero coefficients.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 10.4. N − VECTORS AND THE ALGEBRAIC STRUCTURE OF RN 283

7. If
a(3i + 2j) + b(i − j) =< 0, 0 >,
then
3a + b = 0 and 2a − b = 0.
From the second equation, b = 2a, and then the first equation is 5a = 0,
so a = 0 and then b = 0. The only linear combination of these vectors
that equals the zero vector is the trivial combination, so the vectors are
linearly independent.
We could also observe that neither vector is a linear combination of the
other, which would be the case of these two vectors were linearly depen-
dent.

8. These vectors are linearly dependent because no one of them is a linear

combination of the others. It is also easy to check that the only linear
combination of these vectors that equals < 0, 0, 0, 0 > is the trivial linear
combination (all coefficients zero).

9. These two vectors are linearly independent because neither is a scalar mul-
tiple of the other (which would occur for two linearly dependent vectors).

10. These vectors are linearly dependent because

4 < 1, 0, 0, 0 > −6 < 0, 1, 1, 0 > + < −4, 6, 6, 0 >=< 0, 0, 0, 0 > .

11. The vectors are linearly dependent because

2 < 1, 2, −3, 1 > + < 4, 0, 0, 2 > − < 6, 4, −6, 4 >=< 0, 0, 0, 0 > .

12. Suppose that

α < 0, 1, 1, 1 > +β < −3, 2, 4, 4 > +γ < −2, 2, 34, 2 >

+ δ < 1, 1, −6, 2 >=< 0, 0, 0, 0 > .

Looking at the components of the sum on the left, we obtain

−3β − 2γ + δ = 0,
α + 2β + 2γ + δ = 0,
α + 4β + 34γ − 6δ = 0,
α + 4β + 2γ + 2δ = 0.

It is routine to solve these equations and show that

α = β = γ = δ = 0.

Therefore the vectors are linearly independent.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 284 CHAPTER 10. VECTORS AND THE VECTOR SPACE RN

13. The vectors are linearly dependent because

2 < 1, −2 > −2 < 4, 1 > + < 6, 6 >=< 0, 0 > .

14. The vectors are linearly independent (each has a component of 1 where
the other two have components of 0).
15. The vectors are linearly independent.
16. The vectors are linearly independent.

In each of Problems 17–21 it is routine to show that S is not empty, and

that a linear combination of vectors of S is in S, showing that S is a subspace
of the appropriate Rn . We will show how to find a basis for S.

17. Every vector in S has the form

x < 1, 0, 0, −1 > +y < 0, 1, −1, 0 > .

Therefore the two vectors < 1, 0, 0, −1 > and < 0, 1, −1, 0 > span S.
These vectors are also linearly independent, and so form a basis for S,
which has dimension 2.
18. Every vector in S is of the form

x < 1, 0, 2, 0 > +y < 0, 1, 0, 3 > .

The independent vectors < 1, 0, 2, 0 > and < 0, 1, 0, 3 > therefore span S
and form a basis. S has dimension 2.
19. S consists of all vectors in Rn of the form

< x1 , 0, x2 , · · · , xn−1 > .

The n − 1 independent vectors

< 1, 0, 0, · · · , 0 >, < 0, 0, 1, 0, · · · , 0 >, · · · , < 0, 0, 0, · · · , 0, 1 >

span S, so S has dimension n − 1.

20. S is spanned by the independent vectors

< 1, 1, 0, 0, 0, 0 >, < 0, 0, 1, 1, 0, 0 >, < 0, 0, 0, 0, 0, 1 >

and so has dimension 3.

21. Every vector in S is a scalar multiple of

0, 1, 0, 2, 0, 3, 0 >

so this vector forms a basis for S and S has dimension 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 10.4. N − VECTORS AND THE ALGEBRAIC STRUCTURE OF RN 285

22. The spanning vectors are independent, so they form a basis for the sub-
space S of R3 that they span. Further, by inspection,

< 3, 1, 0 >= 3 < 1, −1, 0 > +2 < 0, 1, 0 > .

23. The spanning vectors are independent and form a basis for the subspace
S of R3 that they span. Further, by inspection,

< −5, −3, −3 >= −5 < 1, 1, 1 > +2 < 0, 1, 1 > .

24. It is routine to check that the three spanning vectors are independent,
hence form a basis for the subspace S of R4 that they span. By inspection,

< 4, 5, 9, 4 >= 2 < 2, 1, 1, 0 > +3 < 0, 1, 1, 0 > +4 < 0, 0, 1, 1 > .

25. The spanning vectors are independent and so form a basis for the subspace
S of R4 that they span. This subspace has dimension 3. For X to be in
S, we need numbers a, b and c so that

a < 1, 0, −3, 2 > +b < 1, 0, −1, 1 >=< −4, 0, 10, −5 > .

This requires that

a + b = −4, −3a + b = 10, and 2a + b = −5.

Then a = −3, b = −1, so

< −4, 0, 10, −7 >= −3 < 1, 0, −3, 2 > − < 1, 0, −1, 1 >

and < −4, 0, 10, −4 > is in S.

26. All that is needed is to show that the dot product of the two vectors is
zero:

(X − Y) · (X + Y) = X · X + X · Y − Y · X − Y · Y
=k X k2 − k Y k2 = 0.

27. Because U is in S, which is spanned by V1 , · · · , Vk , we must have

U = c1 V1 + · · · + ck Vk ,

so V1 , · · · , Vk , U are linearly independent.

28. If U1 , · · · , Uk are linearly independent, then they form a basis for the
subspace of Rn spanned by S. If these vectors are not independent, then
at least one of the vectors must be a linear combination of the others.
Suppose Uk is a linear combination of U1 , · · · , Uk−1 . Then

U1 , · · · , Uk−1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 286 CHAPTER 10. VECTORS AND THE VECTOR SPACE RN

span S. If these vectors are linearly independent, they form a basis for
S. If not, one of these vectors, say Uk−1 is a linear combination of the
others. Now

U1 , · · · , Uk−2

span S. Continue in this way. If these vectors are linearly independent,

they form a basis for S. If not, one must be a linear combination of the
others, and then we obtain a set of k − 3 of the original set of vectors
that spans S. Continue in this way until a linearly independent set of the
original spanning vectors spans S. This set is a basis for S.

29. Suppose the set of vectors is V1 , · · · , Vk , O. Then

0V1 + 0V2 + · · · + 0Vk + 1O = O

is a linear combination of the vectors adding up to the zero vector, and

with a nonzero coefficient. This makes the original set of vectors (including
the zero vector) linearly dependent.

10.5 Orthogonal Sets and Orthogonalization

1.

k V1 + · · · + Vk k2 = (V1 + · · · + Vk ) · (V1 + · · · + Vk )
= V1 · (V1 + · · · + Vk ) + V2 · (V1 + · · · + Vk ) + · · ·
+ Vk · (V1 + · · · + Vk )
= V1 · V1 + V2 · V2 + · · · + Vk · Vk
=k V1 k2 + k V2 k2 + · · · k Vk k2 ,

in which we have used the fact that Vi · Vj = 0 if i 6= j.

2. Let

k
X
Y =X− (X · Vj )Vj .
j=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 10.5. ORTHOGONAL SETS AND ORTHOGONALIZATION 287

Compute
0 ≤k Y k2 = Y · Y
   
Xk k
X
= X − (X · Vj )Vj  · X − (X · Vj )Vj 
j=1 j=1
k
X
= X · X − 2X · (X · Vj )Vj
j=1
   
Xk k
X
+  (X · Vj )Vj  ·  (X · Vj )Vj 
j=1 j=1
k
X
= X · X − 2X (X · Vj )Vj
j=1
k X
X k
+ (X · Vj )(X · Vr )Vj · Vr .
j=1 r=1

Because Vj · Vr = 0 if j 6= r, and Vj · Vj = 1, the double sum collapses

to just terms in which j = r and we have
k
X k
X
0 ≤k X k2 −2 (X · Vj )2 + (X · Vj )2
j=1 j=1
k
X
=k X k2 − (X · Vj )2 .
j=1

Then
k
X
(X · Vj )2 ≤k X k2 .
j=1

3. Reason as in Problem 2, except now use the fact that

n
X
(X · Vj )Vj = X.
j=1

In each of Problem 4–11, the given vectors are denoted X1 , · · · , Xk in the

given order.
4. Let V1 = X1 and
X2 · X1
V2 = X2 − X1
X1 · X1
18
= X2 + X1
17
=< 52/17, −13/17, 0 > .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 288 CHAPTER 10. VECTORS AND THE VECTOR SPACE RN

5. Take V1 = X1 and
11
V2 = X2 + X1 =< 0, 4/5, 2/5, 0 > .
5
6. Let V1 = X1 . Next,
7
V2 = X2 + X1 =< 0, 4/3, 13/6, 29/6 > .
6
Finally,
3 43/2
V3 = X3 − V1 − V2
6 179/6
1 129
= X3 − V1 − V2
2 179
1
= < 0, 7, −11, 3 > .
179
7. Let V1 = X1 and
5
V2 = X2 + X1
26
1
= < 109, 0, −41, 58 > .
26
Finally, let
17 331/26
V3 = X3 − X1 − V1
26 651/26
17 331
= X3 − V1 − V2
26 651
1
= < −962, 0, −1406, 0, 814 > .
651
8. V1 = X1 ,
5
V2 = X2 − X1
7
1
= < 0, 0, −1, −19, 40 >,
9

2 17
V3 = X3 + V1 + V2
9 9
1
= < 0, 218, −341, 279, 62 >,
218
and
6 13 435
V4 = X1 + V2 − V3
9 3 1179
1
= < 0, 248, 88, −24, −32 > .
393

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 10.6. ORTHOGONAL COMPLEMENTS AND PROJECTIONS 289

9. V1 = X1 ,
1
V2 = X2 − X1
10
1
= < 21, −8, −60, −31, −18, 0 >,
10

3 163/10
V3 = X3 − X1 − V2
10 269/10
1
= < −423, −300, 489, −759, 132, 0 >,
269
and
15 13/2 4455/269
V4 = X4 + X1 − V2 − V4
10 260/10 4095/269
1
= < 337, −145, 250, 29, −9, 0 > .
91

10. V1 = X1 and
3 1
V2 = X2 + X1 = < 0, 0, −3, 3, 0, 0 > .
2 2

11. V1 = X1 and V2 = X2 because X2 and X1 are orthogonal. Finally,

4 4 1
V3 = X3 + V1 + V2 = < 0, −8, 0, −8, 0, 16 > .
12 2 3

10.6 Orthogonal Complements and Projections

1. Let V1 =< 1, −1, 0, 0 > and V2 =< 1, 1, 0, 0 >. These form an orthogonal
basis for S. Then
u · V1 u · V2
uS = V1 + V2
V1 · V1 V2 · V2
= −4V1 + 2V2 =< −2, 6, 0, 0 >

and
u⊥ = u − uS =< 0, 0, 1, 7 > .
The distance between u and S is
√
k u⊥ k= 50.

2. Let
V1 =< 1, 0, 0, 2, 0 > and V2 =< −2, 0, 0, 1, 0 > .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 290 CHAPTER 10. VECTORS AND THE VECTOR SPACE RN

We find that
2 1
uS = V1 + V2 =< 0, 0, 0, 1, 0 >
5 5
and
u⊥ =< 0, −4, −4, 0, 3 > .
The distance between u and S is
√
k u⊥ k= 41.

3. Let
V1 =< 1, −1, 0, 1, −1 >, V2 =< 1, 0, 0, −1, 0 >, V3 =< 0, −1, 0, 0, 1 > .
Then
7 1
uS = V1 + V2 − 3V3 = < 9, −1, 0, 5, −13 > .
2 2
and
1
u⊥ =
< −1, −1, 6, −1, −1 > .
2
The distance between u and S is
√
k u⊥ k= 10.

4. Let
V1 =< 1, −1, 0, 0 > and V2 =< 1, 1, 6, 1 > .
Then
31
uS = −3V1 + V2 =< −86/39, 148/39, 62/13, 31/39 >
39
and
1
u⊥ = < 203, 203, −230, −226 > .
309
The distance between u and S is
1 p
k u⊥ k= 186, 394.
309
5. Let
V1 =< 1, 0, 1, 0, 1, 0, 0 > and V2 =< 0, 1, 0, 1, 0, 0, 0 > .
We find that
1 1
uS = 3V1 + V2 = < 6, 1, 6, 1, 6, 0, 0 >
2 2
and
1
u⊥ =
< 10, 1, −4, −1, −6, −6, 8 > .
2
The distance between u and S is
1√
k u⊥ k= 254.
2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 10.6. ORTHOGONAL COMPLEMENTS AND PROJECTIONS 291

6. If every vector in S ⊥ is orthogonal to every vector in S, then, going the

other way, every vector in S is orthogonal to every vector in S ⊥ . Then

(S ⊥⊥ ) = S.

7. Let v1 , · · · , vk be an orthogonal basis for S, and u1 , · · · ur an orthogonal

basis for S ⊥ . Then the vectors

u1 , · · · , uk , v1 , · · · vr

span Rn because every n− vector is a sum of a vector in S (a linear

combination of v1 , · · · , vk ) and a vector in S ⊥ (a linear combination of
u1 , · · · , ur ). Further, the set of vectors

u1 , · · · , uk , v1 , · · · vr

are linearly independent, because no one of them is a linear combination

of others in this set. These k + r vectors therefore form a basis for Rn .
Because Rn has dimension n,

n = k + r.

That is,

dimension(S) + dimension (S ⊥ ) = dimension(Rn ).

8. Let u =< 1, −1, 3, −3 >. The idea is to use an orthogonal basis for S to
produce uS , which is the vector we seek. The given vectors

V1 =< 1, 0, 1, 0 > and V2 =< −2, 0, 2, 1 >

form an orthogonal basis for S. Then

1 1
uS = 2V1 + V2 = < 16, 0, 20, 1 > .
9 9
This is the vector in S closest to u.

9. Denote

V1 =< 2, 1, −1, 0, 0 >, V2 =< −1, 2, 0, 1, 0 >, V3 =< 0, 1, 1, −2, 0 > .

These form an orthogonal basis for S. With u =< 4, 3, −3, 4, 7 >, compute
7 4
uS = V1 + V2 − V3
3 3
1
= < 11, 9, −11, 11, 0 > .
3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 292 CHAPTER 10. VECTORS AND THE VECTOR SPACE RN

10. Denote

V1 =< 0, 1, 1, 0, 0, 1 >, V2 =< 0, 0, 3, 0, 0, −3 >, V3 =< 6, 0, 0, −2, 0, 0 > .

With u =< 0, 1, 1, −2, −2, 6 >, compute

8 5 1
uS = V1 − V2 + V3
3 6 10
=< 3/5, 8/3, 1/6, −1/5, 0, 31/6 > .

This is the vector in S closest to u.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 11

Matrices, Determinants and

Linear Systems

11.1 Matrices and Matrix Algebra

1.  
14 −2 6
2A − 3B =  10 −5 −6
−26 −43 −8

2.  
19 2
 6 −2
−5A + 3B = 
−28 38 


−27 35

3.

2 + 2x − x2 12x + (1 − x)(x + ex + 2 cos(x))

 
2
A + 2AB =
4 + 2x + 2ex + 2xex −22 − 2x + e2x + 2ex cos(x)

4. −3A − 4B = (18) This is a 1 × 1 matrix, which we can think of as just

the single matrix entry, in this case 18.

5.  
−36 0 68 196 20
4A + 8B =
128 −40 −36 −8 72

6.      
−17 18 −40 8 27 10
A2 − B2 = − =
6 1 −5 −39 11 40

293

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 294CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

7. BA is not defined;
 
−10 −34 −16 −30 −14
AB =  10 −2 −11 −8 −45
−5 1 15 61 −63

8.    
−16 0 12 −32
AB = ; BA =
17 28 −14 0

9. AB = (115);
 
3 −18 −6 −42 66
−2 12 4 28 −44 
 
BA = 
 −6 36 12 84 −132

0 0 0 0 0 
4 −24 −8 −56 88

10.  
48 1 1 −58  
 −96 2 2 220 
AB =   ; BA = 76 152
−288 −22 −22 −68 50 136
−16 6 6 184

11. AB is not defined;

 
410 36 −56 227
BA =
17 253 40 −1

12. BA is not defined;

 
−22 30 −10 −4
AB =
−42 45 30 6

13. AB is not defined;

BA = −16 −13 −5

14. Neither AB nor BA is defined.

15. BA is not defined;  
39 −84 21
AB =
−23 38 3

16. AB is not defined;

BA = 28 30

17. AB is 14 × 14, BA is 21 × 21.

18. Neither AB nor BA is defined.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.1. MATRICES AND MATRIX ALGEBRA 295

19. AB is not defined, BA is 4 × 2.

20. AB is 1 × 3, BA is not defined.
21. AB is not defined, BA is 7 × 6.
22. There are infinitely many examples. Here is one. Let
     
2 1 2 1 6 0
A= ,B = ,C = .
8 4 −1 1 −1 1
Then B 6= C, but  
12 6
BA = CA = .
6 3
23. The adjacency matrix of G is
 
0 1 1 0 0
1 1 0 1 1
 
A= 1 1 0 1 1
.
0 1 1 0 1
0 1 1 1 0
Compute
   
2 7 7 4 4 14 17 17 18 18
7 8 9 9 9 17 34 33 26 26
A3 =  4
   
7 9 8 9 9
 and A = 17 33 34 26 26
.

4 9 9 6 7 18 26 26 25 24
4 9 9 7 6 18 26 26 24 25
The number of v1 − −v4 walks of length 3 is (A)314 = 4 and the number of
v1 − −v4 walks of length 4 is (A4 )14 = 18. The number of v2 − −v3 walks
of length 3 is 9 and the number of v2 − −v4 walks of length 4 is 26.
24. The adjacency matrix of H is
 
0 1 1 0 1
1 0 1 0 1
 
A= 1 1 0 1 0
.
0 0 1 0 1
1 1 0 1 0
Then
   
3 2 1 2 1 19 18 11 14 11
2 3 1 2 1 18 19 11 11 11
A2 =  4
   
1 1 3 0 3
 and A = 11 11 20 4 20
.

2 2 0 2 0 14 14 4 12 4
1 1 3 0 3 11 11 20 4 20
The number of v1 − −v4 walks of length 4 is (A4 )14 = 14, the number of
v2 − −v3 walks of length 2 is (A2 )23 = 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 296CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

25. The adjacency matrix of K is

 
0 1 1 1 1
 1 0 1 1 0
 
A=
11 01 1 .

 1 1 1 0 1
1 0 1 1 0
Then
   
4 2 3 3 2 10 10 11 11 10
2 3 2 2 3 10 6 10 10 6
A2 = 
  3  
3 2 4 3  , A = 11
2 10 10 11 10


3 2 3 4 2 11 10 11 10 10
2 3 2 2 3 10 6 10 10 6
and  
42 32 41 41 32
32 30 32 32 30
A4 = 
 
41 32 42 41 32
.
41 32 31 42 32
32 30 32 32 30
The number of v4 − −v5 walks of length 2 is 2 and the number of v2 − −v3
walks of length 3 is 10. The number of v1 − −v2 walks of length 4 is 32
and the number of v4 − −v5 walks of length 4 is 32.
26. (a) The i, i element of A2 is the number of vi − −vi walks of length 2 in
the graph. Each such walk has the form vi − −vj − −vi for some j 6= i,
hence corresponds to a vertex vj adjacent to vi in the graph. Therefore
A2ii counts the number of vertices adjacent to vi in the graph, and this is
the degree of vi .
(b) The i − −i element of A3 is the number of vi − −vi walks of length 3 in
the graph. Any such walk has the form vi −−vj −−vk −−vi with j 6= k and
neither j nor k equal to i. These three vertices therefore form the vertices
of a triangle in the graph. However, each such triangle is counted twice in
the i, i element of A3 because this triangle actually represents two vI −−vi
walks of length 3, namely vi − −vj − −vk − −vi and vi − −vk − −vj − −vi .
Therefore
(A3 )ii = 2(number of triangles in G).

27. Let M be the set of all n × n real matrices. Each such matrix has nm
elements in its n rows and m columns (some possibly the same number, but
differentiated by location). If we string out the rows, with row 2 following
row 1, then row 3 following row 2, and so on, we have a string of nm
elements, which we can think of as a vector in Rnm . Each nm − −vector
corresponds to an n × m matrix, and each n × m matrix produces a unique
nm − − vector. In this way the n × m matrices correspond in a one-to-one
fashion with the vectors in Rnm .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.2. ROW OPERATIONS AND REDUCED MATRICES 297

11.2 Row Operations and Reduced Matrices

1. A is 3×4, so multiply A on the left by the matrix Ω formed by performing
this elementary row operation on I3 . Thus form
 
1 √0 0
Ω = 0 3 0 .
0 0 1

This choice of Ω can be checked by verifying that ΩA = B.

2.  
1 0 0 0
0 1 0 0
Ω=
0

6 1 0
0 0 0 1

3. In this case more than one elementary operation is performed, so Ω is a

product of elementary operations, the right-most factor performing the
first operation, then the next to right-most, the second operation, and so
on, with the left factor performing the last operation:
   √ 
5 0 0 0 1 0 1 0 13
Ω = 0 1 0 1 0 0 0 1 0 .
0 0 1 0 0 1 0 0 1

If these products are carried out, we get

 
0 5 √0
1 0 13
0 0 1

4.     
1 0 0 1 0 0 1 0 0
Ω = −1 1 0 0 0 1 = −1 0 1
0 0 1 0 1 0 0 1 0

5.    √   
0 1 1 0 1 3 0 √
15
Ω= =
1 0 0 15 0 1 1 3

6.     
1 0 0 1 0 0 √1 0 0 1 0 0
Ω = 0 1 0 0 1 0  3 1 0 0 1 0 .
0 1 1 0 0 4 0 0 1 1 0 1
Then  
√1 0 0
Ω= √3 1 0 .
4 + 31 4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 298CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

7.      
1 0 0 1 0 0 1 0 0 1 0 0
Ω = 0 0 1 14 1 0 0 1 0 =  0 0 4
0 1 0 0 0 1 0 0 4 14 1 0

8.     
1 0 0 0 0 1 1 0 0 1 0 0
Ω = 0 1 0 0 1 0 0 1 0 0 0 1 .
0 0 5 1 0 0 0 3 1 0 1 0
If these are multiplied out, we get
 
0 1 3
Ω = 0 0 1 .
5 0 0

In these and later problems, the delta notation is sometimes useful:

(
1 if i = j,
δij =
0 if i 6= j.

9. Let A = [aij ] be an n × m matrix. Now, B = [bij ] and E = [eij ] are

obtained, respectively, by interchanging rows s and t of A and In . Then,
for i 6= s and i 6= t,
bij = aij and eij = δij .
If i = s, bsj = atj and esj = δtj . And if i = t, bij = asj and eij = δsj .
Now consider the i, j − − element of EA. For i 6= s and i 6= t,
n
X
(EA)sj = eik akj = aij = bij .
k=1

For i = s,
n
X n
X
(EA)sj = esk akj = δtk akj = atj = bsj .
k=1 k=1

And for i = t,
n
X n
X
(EA)tj = eik akj = δsk akj = asj = btj
k=1 k=1

for j = 1, · · · , m. Therefore EA = B.

10. Let A be n × m. B and E are formed, respectively, by multiplying row

s of A and In by α. Then, for i 6= s, bij = aij and eij = δij . For i = s,
bsj = αasj and esj = αδsj .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.2. ROW OPERATIONS AND REDUCED MATRICES 299

Now consider the i, j − −element of EA. For i 6= s,

n
X n
X
(EA)ij = eik akj = αδik akj = aij = bij
k=1 k=1

and
n
X n
X
(EA)sj = esk akj = αδsk akj = bsj
k=1 k=1
for j = 1, 2, · · · , m. This shows that EA = B.
11. Let A be n × m. Now B and E are formed, respectively, from A and In
by adding α times row s to row t. For i 6= t, bij = aij and eij = δij , while
for i = t, btj = atj + αasj and etj = δtj + αδsj .
Now consider the i, j − − element of of EA. For i 6= t,
n
X n
X
(EA)ij = eik akj = δik akj = aij .
k=1 k=1

For i = t,
n
X n
X
(EA)tj = etk akj = (δtk + αakj )
k=1 k=1
= atj + αasj = bsj .
This shows that EA = B.
In Problems 12–23, keep in mind that a given matrix can be reduced by
different sequences of row operations, but the end result must be the same - a
matrix has only one reduced form. There may therefore be different matrices
Ω1 and Ω2 such that
Ω1 A = AR = Ω2 A.
We give one such matrix for each problem.
12. A is reduced by simply adding row two to row one:
   
1 1 0 1 0 5
Ω = 0 1 0 and AR = 0 1 2 .
0 0 1 0 0 0

13. We can reduce A by first subtracting row two from row one of I2 , then
multiplying row one of the resulting matrix by 1/3:
     
1 0 1 −1 1/3 −1/3
I2 = → → = Ω.
0 1 0 1 0 1
Then  
1 0 1/3 4/3
ΩA = AR = .
0 1 0 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 300CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

14. We can reduce A by first interchanging rows two and four of I4 , then (on
the resulting matrix), multiplying row one by −1, then adding row two to
row one:
   
1 0 0 0 1 0 0 0
0 1 0 0
 → 0 0 0 1
 
I4 = 
0 0 1 0 0 0 1 0
0 0 0 1 0 1 0 0
   
−1 0 0 0 −1 0 0 1
 0 0 0 1
 →  0 0 0 1 = Ω.
 
→  0 0 1 0  0 0 1 0
0 1 0 0 0 1 0 0
Then  
1 −4 −1 0
0 0 0 1
ΩA = 
0 0
 = AR .
0 0
0 0 0 0
15. A is in reduced form, so A = AR .
16. Starting with I4 , subtract row one from row four, subtract 3 times row
one from row three, interchange rows two and three, and then interchange
rows one and two:
   
1 0 0 0 1 0 0 0
 0 1 0 0  0 1 0 0
I4 →   0 0 1 0 → −3 0 1 0
  

−1 0 0 1 −1 0 0 1
   
1 0 0 0 −3 0 1 0
−3 0 0  1 0 0 1
→  0 1 0 0 →  − 1 0 0 .
  

−1 0 0 1 −1 0 0 1
Then  
1 0
0 1
ΩA = 
0
 = AR .
0
0 0
For Problems 17–23, just Ω and AR are given.
17.    
0 1 1 1
Ω= , AR = .
1 −2 0 0
18.    
−8 −2 38 1 0 0
1 
Ω= 37 43 −7 , AR = 0 1 0 = I3 .
270
19 −29 11 0 0 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.3. SOLUTION OF HOMOGENEOUS LINEAR SYSTEMS 301

19.    
−1/3 0 1 −4/3 −4/3
Ω= , AR =
0 1 0 0 0

20.    
0 1 1 0 0 0
Ω= , AR =
1/2 1/2 0 1 3/2 1/2

21.    
0 0 1 1 0 0 −3/4
1
Ω =  4 −4 −8 , AR = 0 1 0 3 
4
−4 8 8 0 0 1 0

22.    
0 1/2 −1 1 0 0
Ω= 0 0 1  , AR = 0 1 0
−1/7 2/7 −3/7 0 0 1

23.    
0 0 1 0 1
0 1 3 0
 , AR = 0
 

1 0 −6 0 0
0 0 −1 1 0

11.3 Solution of Homogeneous Linear Systems

In Problems 1–12, we use the facts that (1) the system AX = O has the same
solutions as the reduced system AR X = O, and (2) the solution of the reduced
system can be read by inspection from the reduced matrix AR .

1. The coefficient matrix and its reduced form are

   
1 2 −1 1 1 0 1 −1
A= , AR = .
0 1 −1 1 0 1 −1 1

Because AR has two nonzero rows and m = 4, the solution space has
dimension m − 2 = 2, which means that the general solution is in terms
of two of the unknowns, which can be given any values. Specifically, from
the reduced matrix,

x1 = −x3 + x4
x2 = x3 − x4 .

All solutions are given by

       
x1 −x3 + x4 −1 1
x2   x3 − x4  1 −1
X= x3  = 
   = x3   + x4   .
x3  1 0
x4 x4 0 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 302CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

It looks a little neater to write the general solution

   
−1 1
1 −1
X = α  1 +β 0 
  

0 1

in which α and β are arbitrary real numbers. The solution space of this
system is the subspace of R4 having basis vectors

< −1, 1, 1, 0 >, < 1, −1, 0, 1 > .

2. The coefficient matrix is

 
−3 1 −1 1 1
A= 0 1 1 0 4
0 0 −3 2 1

and  
1 0 0 1/9 11/9
AR = 0 1 0 2/3 13/3  .
0 0 1 −2/3 −1/3
With x4 = α and x5 = β, the general solution is
   
−1/9 −11/9
−2/3 −13/3
   
X = α  2/3  + β  1/3  .
  
 1   0 
0 1

The solution space is a subspace of R5 having dimension 2 and basis

< −1/9, −2/3, 2/3, 1, 0 >, < −11/9, −13/3, 1/3, 0, 1 > .

3. The coefficient matrix and its reduced form are

   
−2 1 2 1 0 0
A =  1 −1 0 , AR = 0 1 0 = I3 .
1 1 0 0 0 1

Here A has rank 3, the number of nonzero rows of A, and m− rank (A) =
3 − 3 = 0, so the solution space has dimension zero, consisting just of the
trivial solution  
0
X = 0 .
0
This is consistent with the reduced system being the system

x1 = 0, x2 = 0, x3 = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.3. SOLUTION OF HOMOGENEOUS LINEAR SYSTEMS 303

4. The coefficient matrix and its reduced form are

   
4 1 −3 1 1 0 −1/2 0
A= , AR = .
2 0 −1 0 0 1 −1 1

The general solution can be written

 
1/2
 1 

X = α 1  + β 0 − 101 .


The solution space is the subspace of R4 having basis vectors

< 1/2, 1, 1, 0 >, < 0, −1, 0, 1 > .

5. The coefficient matrix is

 
1 −1 3 −1 4
2 −2 1 1 0
A=
1 0 −2 0

1
0 0 1 1 −1

and  
1 0 0 0 9/4
0 1 0 0 7/4 
AR = 
0
.
0 1 0 5/8 
0 0 0 1 −13/8
The solution space has dimension 5 − 4 = 1 and the general solution is
 
−9/4
−7/4
 
X = α−5/8 .

 13/8 
1

The single vector < −9/4, −7/4, −5/8, 13/8, 1 > is a basis for the solution
space.
6. The coefficient matrix is
 
6 −1 1 0 0
A = 1 0 0 −1 2 
1 0 0 1 −2

and  
1 0 0 0 −2
AR = 0 1 −1 0 −12 .
0 0 0 1 −4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 304CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

The general solution is

   
0 2
1 12
   
1 + β  0  .
X = α   
0 4
0 1
The solution space is the subspace of R5 of dimension 2, having basis
< 0, 1, 1, 0, 0 >, < 2, 12, 0, 4, 1 > .

7. The coefficient matrix is

 
−10 −1 4 1 1 −1
 0 1 −1 3 0 0
A= 
 2 −1 0 0 1 0
0 1 0 −1 0 1
and  
1 0 0 05/6 5/9
0 1 0 0 2/3 10/9
AR = 0 0 1
.
0 8/3 13/9
0 0 0 1 2/3 1/9
From the reduced system read the general solution
   
−5/6 −5/9
−2/3 −10/9
   
−8/3 −13/9
X = α
 + β  −1/9  .
  
−2/3  
 1   0 
0 1
The solution space is a subspace of R6 having dimension 2, with basis
vectors
< −5/6, −2/3, −8/3, −2/3, 1, 0 >, < −5/9, −10/9, −13/9, −1/9, 0, 1 > .

8. The coefficient matrix is

 
1 0 0 0 7/6 −5/4
0 1 0 0 −20/3 9/2 
A= .
0 0 0 1 14/3 −11/2
0 0 0 1 −3 2
From the reduced matrix we can write the general solution
   
−7/6 5/4
 20/3  −9/2
   
−14/3
 + β  11/2  .
 
X = α  3   −2 
   
 1   0 
0 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.3. SOLUTION OF HOMOGENEOUS LINEAR SYSTEMS 305

The solution space is the subspace of R6 of dimension 2, having basis

vectors
< −7/6, 20/3, −14/3, 3, 1, 0 >, < 5/4, −9/2, 11/2, −2, 0, 1 > .

9. There is no x3 in the equations, so we actually have a system of three

equations in the four unknowns x1 , x2 , x4 and x5 . The coefficient matrix
is  
0 1 −3 1
A = 2 −1 1 0
2 −3 0 4
and  
1 0 0 −5/14 0 1 0 −11/17
AR = .
0 0 1 − 6/7
Now x1 , x2 and x4 depend on x5 and
 
5/14
11/7
 6/7  .
X = α 

The solution space is the subspace of R4 of dimension 1, having basis

vector < 5/14, 11/7, 6/7, 1 >.
10. The coefficient matrix is
 
4 −3 0 1 1 −3
0 2 0 1 −1 −6
A=
3 −2 0

0 4 −1
2 1 −3 4 0 0
and  
1 0 0 0 −41/6 −2/3
0 1 0 0 −49/6 −1/3
AR =  .
0 0 1 0 −13/6 −7/3
0 0 0 1 23/6 −4/3
The general solution is
   
41/6 2/3
 49/6  1/3
   
 13/6 
 + β 7/3 .
 
X = α
−23/6 4/3
   
 1   0 
0 1

The solution space is a subspace of R6 of dimension 2 and having basis

vectors
< 41/6, 49/6, 13/6, −23/6, 1, 0 >, < 2/3, 1/3, 7/3, 4/3, 0, 1 > .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 306CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

11. The coefficient matrix is

 
1 −2 0 0 1 −1 1
0 0 1 −1 1 −2 3
A= 
1 0 0 0 −1 2 0
2 0 0 −3 1 0 0
and  
1 0 0 0 −1 2 0
0 1 0 0 −1 3/2 −1/2
AR =  .
0 0 1 0 0 −2/3 3 
0 0 0 1 −1 4/3 0
The general solution is
     
1 −2 0
1 −3/2 1/2
     
0  2/3   −3 
     
X = α 1 + β −4/3 + γ  0  .
     
1  0   0 
     
0  1   0 
0 0 1
The solution space is the subspace of R7 of dimension 3, with basis vectors
< 1, 1, 0, 1, 1, 0, 0 >, < −2, −3/2, 2/3, −4/3, 0, 1, 0 >, < 0, 1/2, −3, 0, 0, 0, 1 > .

12. The coefficient matrix is

 
2 0 0 0 −4 0 1 1
0 2 0
 0 0 −1 1 −1

A= 0 0 1 −4 0 0 0 1
0 1 −1 1 0 0 0 0
0 1 0 0 − 1 1 −1 0
and  
1 0 0 0 0 −3 7/2 −1/2
0
 1 0 0 0 −1/2 1/2 −1/2

0
AR =  0 1 0 0 −2/3 2/3 −1 .
0 0 0 1 0 −1/6 1/6 −1/2
0 0 0 0 1 −3/2 3/2 −1/2
The general solution is

    
3 −7/2 1/2
1/2 −1/2 1/2
     
2/3 −2/3  1 
     
1/6 −1/6
 + γ 1/2 .
 
X = α +β
3/2 −3/2 1/2
     
 1   0   0 
     
 0   1   0 
0 0 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.3. SOLUTION OF HOMOGENEOUS LINEAR SYSTEMS 307

The solution space is the subspace of R8 having basis vectors

< 3, 1/2, 2/3, 1/6, 3/2, 1, 0, 0 >, < −7/2, −1/2, −2/3, −1/6, −3/2, 0, 1, 0 >,
< 1/2, 1/2, 1, 1/2, 1/2, 0, 0, 1 > .

13. The answer is yes. All that is required is that m − rank(A) > 0, so that
the solution space has positive dimension, hence non-zero vectors, which
are solutions of the system.
As a specific example, consider the system AX = O, where
 
1 0 3
A = 0 1 −1 .
3 0 9

This is a homogeneous system with three equations in three unknowns.

We find that  
1 0 3
AR = 0 1 −1 .
0 0 0
Then AR has two nonzero rows, so the solution space of AX = O has
dimension 3 − 2 = 1 > 0. In fact, the general solution is
 
3
X = α 1 .
1

14. Suppose A is n × m. Let the columns of A be C1 , · · · , Cm , written as

n × 1 column matrices. Now notice that, if
 
a1
 a2 
X= . 
 
 .. 
am

then AX = O is equivalent to

a1 C1 + · · · + am Cm = O,

the n × 1 zero matrix.

With these preliminaries, we can prove the proposition. If the columns of
A are linearly dependent, then there are numbersa1 , · · · , am not all zero
such that
a1 C1 + · · · + am Cm = O,
and then X, formed as a column with these coefficients as elements, is a
nontrivial solution of AX = O.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 308CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

Conversely, suppose the system has a nontrivial solution

 
a1
 a2 
X = X =  . .
 
 .. 
am

Then
a1 C1 + · · · + am Cm = O,
and at least one of these coefficients is nonzero, so the columns of A are
linearly dependent.
15. Let the rows of A be R1 , · · · , Rn . These are vectors in Rm . Let R be the
row space of A, which is the subspace of Rm spanned by the row vectors.
Now, X is in the solution space S(A) of the homogeneous system with
coefficient matrix A exactly when AX = O. This is true exactly when
the dot product Rj · X = O for j = 1, · · · , n, which is true exactly when
each row is orthogonal to X. But this is equivalent to X being orthogonal
to every linear combination of these rows, hence to every vector in the row
space R of A. This makes the solution space of the system the orthogonal
complement of the row space:

R⊥ = S(A).

Because the columns of At are the rows of A, similar reasoning shows that
the solution space S(At ) of the system At X = O has the column space
C of A as its orthogonal complement:

C ⊥ = S(At ).

11.4 Nonhomogeneous Systems

1. The augmented matrix is
..
 
3 −2 1
. 6
.. . 
[A.B] =  1 10 −1 .. 2

 
..
−3 −2 1 . 0

with reduced form

.
 
1 0 0 .. 1 
. .
[A..B]R = 0 0 .. .
 
 1 1/2

.
0 0 1 .. 4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.4. NONHOMOGENEOUS SYSTEMS 309

The reduced forms of the matrix of coefficients of the system and the
augmented matrix have the same number of nonzero rows, so both A
.
and [A..B] have the same rank (in this case 3). Therefore this system
is consistent. We can read the solution from the reduced form of the
augmented matrix:  
1
X = 1/2 .
4
In this case the solution is unique because m minus the rank of A is
3 − 3 = 0, so the associated homogeneous system has only the trivial
solution.

2. The augmented matrix is

..
 

.. 4 −2 3 10
.1 
[A.B] = 1 0 ..
,
 
 0 −3 .8 
..
2 −3 0 1 . 16

and
..
 

.. 1 0 0 −3 . 8 
[A.B]R = 0 1 0 −7/3 .
 .
. 
.
 0 
..
0 0 1 52/9 . −31/3
.
Because A and [A..B] have the same rank 3, this system is consistent.
Read from the reduced augmented matrix that
   
8 3
 0   7/3 
X= −31/3 + α −52/9 .
  

0 1

3. We have
..
 

. 2 −3 0 1 0 −1 . 0
[A..B] = 3 0 −2 0 . 
1 0 .. 1

 
..
0 1 0 −1 0 6 . 3
and
.
 
1 0 0 −1 0 17/2 .. 9/2 
. ..
[A..B]R = 0 1 0 −1 .
 
 0 6 . 3 
.
0 0 1 −3/2 −1/2 51/4 .. 25/4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 310CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

.
A and [A..B] have the same rank 3, so the system has solutions. Read
from the reduced augmented matrix that
       
9/2 1 0 −17/2
 3   1   0   −6 
       
 + α   + β   + γ −51/4 .
25/4 3/2 1/2  
X=  0   1   0   0 
       
 0   0   1   0 
0 0 0 1

4. From the system, we have

..
 

.  2 −3 . 1
[A..B] = −1 3 ... 0
 
 
..
1 −4 . 3

and
.
 
1 0 .. 0
.. .
[A.B]R = 0 1 .. .
 
 0

.
0 0 .. 1
The left two columns of this reduced augmented matrix form AR , and this
has two nonzero rows, while the reduced form of the augmented matrix
has three nonzero rows. Because the rank of A does not equal the rank
.
of [A..B], this system is inconsistent.
In this case the absence of solutions can be seen easily from the reduced
augmented matrix, which is the system

x1 = 0,
x2 = 0,
0x1 + 0x2 = 1.

Clearly the third of these equations can have no solution.

5. The augmented matrix is

..
 
0 3 0 −4 0 0 . 10 
.
.. 4 −1 .. 8 
 
1 −3 0 0
[A.B] = 
 ..


0 1
 1 −6 0 1 . −9 
..
1 −1 0 0 0 1 . 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.4. NONHOMOGENEOUS SYSTEMS 311

and its reduced form is

..
 
1 0 0 0 −2 2. −4 
.
. 1 ..
 
0 1 0 0 −2 −4 
[A..B]R =  .
.
9/2 .. −38 
 
0 0 1 0
 −7 
..
0 0 0 1 −3/2 3/4 . −11/2
.
Both A and [A..B] have the same rank 4 (number of nonzero rows in
their reduced forms), so this system has a solution. From the reduced
augmented matrix we read the general solution
     
−4 2 −2
 −4   2   −1 
     
 −38 
 + α   + β −9/2 .
 7   
X= −11/2 3/2 −3/4
     
 0   1   0 
0 0 1

6. We have
.
 
2 −3 0 1 .. 1
.
A = 0 3 1 −1 ..
 
 0

.
2 −3 10 0 .. 0
and
..
 

.. 1 0 0 1/20 . 11/20 
[A.B]R = 0 1 0 −9/30
 .. 
.
 . 1/30 

..
0 0 1 −1/10 . −1/10
Because A has rank 3, the same as the rank of the augmented matrix,
this system is consistent. Read the general solution from the reduced
augmented matrix:
   
11/20 −1/20
 1/30   9/30 
X= −1/10 + α  1/10  .
  

0 1

7. The augmented matrix is

.
 
8 −4 0 1/2 10 .. 1
. .
[A..B] = 0 1 1 −1 ..
 
 0 2

.
0 0 1 −3 2 .. 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 312CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

with reduced form

..
 

.. 1 0 0 1/2 3/4 . 9/8

.
[AR .C] = 0 −1 .. .
 
 1 0 1 2 

..
0 0 1 −3 2 . 0

The system is consistent (same number of zero rows in AR and the reduced
augmented matrix), and
     
9/8 −1/2 −3/4
 2   −1   1 
     
 −2  .
X =  0  + α 3  + β
    
 0   1   0 
0 0 1

8. The augmented matrix is

..
 
2 0 −3 . 1
 .. 
1 −1 . 1
 
.
2 −4 1 .. 2

with reduced form

..
 
1 0 0 . 3/4 
..
0 1 0 . −1/12 .
 
 
..
0 0 1 . 1/6
Both the matrix of coefficients and the augmented matrix have rank 3, so
the system is consistent. The system has the unique solution
 
3/4
X = −1/12 .
1/6

This unique solution could have been foreseen from the fact that the num-
ber of columns of A minus its rank is 3 − 3 = 0, so the associated homo-
geneous system has only the trivial solution.

9. The augmented matrix is

.
 
0 0 14 0 −3 0 1 .. 2 
.
1 1 1 −1 0 1 0 .. −4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.4. NONHOMOGENEOUS SYSTEMS 313

with reduced form

.
 
1 1 0 1 3/14 1 −1/14 .. −29/7
. .
0 0 1 0 −3/14 0 1/14 .. 1/7

The matrix of coefficients and its augmented matrix have rank 2, so the
system is consistent. The general solution is
         
−29/7 −1 1 −3/14 −1  
 0  1 0  0  0 1/14 0
 3/14   0  −1/14
           
 1/7  0 0
           0 
X=  0 +α  0 +β 1+γ  0 +δ  0 +  0  .
          
 0  0 0  1  0  
           0 
 0  0 0  0  1
1
0 0 0 0 0

10. The augmented matrix of coefficients is

..
 
3 −2 . −1
..
4 3 . 4

with reduced form

.
 
1 0 .. 5/17 
. .
0 1 .. 16/17
The rank of the coefficient matrix equals 2, the same as the rank of the
augmented matrix, so the system is consistent. The solution is unique:
 
5/17
X= .
16/17

11. The augmented matrix is

..
 
7 −3 4 0 . −7
.
4 .. 6 
 
2 1 −1
 
..
0 1 0 −3 . −5

and the reduced augmented matrix is

..
 
1 0 0 19/15 . 22/15 
 .. 
.
0 1 0
 −3 . −5  
..
0 0 1 −67/13 . −121/15

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 314CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

The rank of A and of the augmented matrix is 3, so the system is consis-

tent. The general solution is
   
22/15 −19/15
 −5   3 
X= −121/15 + α  67/15  .
  

0 1

12. The augmented matrix of coefficients is

..
 
−4 5 −6 . 2
 .. 
 2 −6 1 . −5
 
..
−6 16 −11 . 1
with reduced form
..
 
1 0 0 . −137/48
.
0 .. .
 
0 1 1/6 
 
..
0 0 1 . 41/24
The coefficient matrix and its augmented matrix have the same rank 3, so
the system is consistent. Because the number of unknowns is also 3, this
system has a unique solution
 
−137/48
X =  1/6  .
41/24

13. The augmented matrix is

..
 
 4 −1 4 . 1
.
1 −5 ..
 
1 0
 
.
−2 1 7 .. 4
and its reduced form is
.
 
1 0 0 .. 16/57
.
0 .. .
 
0 1 99/57
 
.
0 0 1 .. 23/57
A and the augmented matrix both have rank 3, which is also the number
of unknowns, so the system has a unique solution
 
16/57
X = 99/57 .
23/57

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.5. MATRIX INVERSES 315

14. The augmented matrix is

..
 
−6 2 −1 1 . 0
.
−1 .. −5
 
1 4 0
 
..
1 1 1 −7 . 0
and its reduced form is
..
 
1 0 0 21/23 . −15/23
 .. 
.
0 1 0 −11/23 . −25/23
 
..
0 0 1 −171/23 . 40/23
A and its augmented matrix have the same rank 3, so the system is con-
sistent. The solution is not unique because m − 3 = 4 − 3 = 1, so the
associated homogeneous system as a solution space of dimension 1. The
general solution is
   
−15/23 −21/23
−25/23  11/23 
X=  40/23  + α  171/23  .
  

0 1

15. Write 
α1
 α2 
X= . 
 
 .. 
αm
Let the columns of A be C1 , · · · , Cm . Then AX = B if and only if

α1 C1 + · · · + αm Cm = B.

This means that X can be a solution if and only if X is a linear combination

of the columns of A, hence is in the column space of A.

11.5 Matrix Inverses

Problem 1 shows all of the row operations used to reduce the matrix and find
the inverse, or show that the matrix is singular. For Problems 2–10, just the
inverse is given if the matrix is nonsingular.

1. Reduce
.. .
   
−1 2 . 1 0 −1 2 .. 1 0
. → add 2 times row one to row two  .
2 1 .. 0 1 0 5 .. 2 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 316CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

.
 
1 −2 .. −1 0
→ multiply row one by −1 
.
0 5 .. 2 1
..
 

→ multiply row two by 1/5 1 −2 . −1 0

..
0 1 .2/5 1/5
.
 
1 0 .. −1/5 2/5
→ add 2 times row two to row one . .
0 1 .. 2/5 1/5
Because I2 has appeared on the left, the given matrix is nonsingular and
the right two columns of this augmented matrix form the inverse matrix:
 
−1 1 −1 2
A = .
5 2 1

2. The matrix is singular (has no inverse) because we find that

 
1 1/4
AR = 6= I2 .
0 0

3.  
−1 1 −2 2
A =
12 1 5

4.  
1 4 0
A−1 =
4 −4 − 1

5.  
−1 1 3 −2
A =
12 −3 6

6.  
64 −4 49
1 
A−1 = −8 4 −7
56
0 0 14

7.  
−6 11 2
1 
A−1 = 3 10 −1
31
1 −7 10

8. The matrix is singular because

 
1 0 3
AR =  0 1 1 6= I3 .
0 0 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.6. DETERMINANTS 317

9.  
6 −6 0
1
A−1 = − −3 −9 2 
12
3 −3 −2
10. A has no inverse because
 
1 0 28/27
AR = 0 1 14/9  .
0 0 0

11.     
−1 −1 8 4 1 −23
−9 2 −5 14   2  = 1 −75
1 
X = A−1 B =
   
11  2 2 −5 3   0  11  −9 
3 3 −2 −1 −5 14
12.     
5 5 5 4 9
1  1  
X = A−1 B = −10 34 23 0 = 15
55 11
5 6 17 5 13
13.
X = A−1 B
    
11 12 9 −4 22
1  1
= 3 16 5  5  = 27
28 7
8 24 4 8 30
14.     
4 4 0 4 −1
1 1
X = A−1 B = 7 −6 39 −5 =  58 
52 52
1 11 13 0 −66
15.     
5 −15 −15 0 −21
1 1
X = A−1 B = − −10 15 10   0  =  14 
25 5
−5 10 0 −7 0

11.6 Determinants
In Problems 1–6 we provide a sequence of row and/or column operations leading
to a determinant that is easily evaluated. Other sequences of operations can
also be used.
1. Add 2 times row two to row one and then −7 times row two to row three
to obtain
−2 4 1 0 16 7
16 7
1 6 3 = 1 6 3 = (−1)3+1 (1) = −22.
−42 −17
7 0 4 0 −42 −17

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 318CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

2. Add 3 times row two to row one, then add row two to row three to obtain
2 −3 7 44 0 10
44 10
14 1 1 = 14 1 1 = (−1)2+2 (1) = 254.
1 6
−13 −1 5 1 0 6

3. Add column two to column one, then 2 times column two to column three:
−4 5 6 1 5 21
1 21
−2 3 5 = 1 3 14 = (−1)3+2 (−2) = −14.
1 14
2 −2 6 0 −2 0

4. Add 2 times row three to row one and 2 times row three to row two to
obtain
2 −5 8 28 −5 0
28 −5
4 3 8 = 30 3 0 = (−1)3+3 = −936.
30 3
13 0 −4 13 0 −4

5. Add 2 times column three to column one and then add column three to
column two to obtain
17 −2 5 27 3 5
27 3
1 12 0 = 1 12 0 = (−1)3+3 (−7) = −2, 247.
1 12
14 7 −7 0 0 −7

6. Add column one to column two, then 3 times column two to column three,
then 2 times column one to column four to obtain
−3 3 9 6 −3 0 0 0
1 −2 15 6 1 −1 18 8
=
7 1 1 5 7 8 22 19
2 1 −1 3 2 3 5 7
−1 18 8 −1 18 8
= (−1)1+1 (−3) 8 22 19 = −3 8 22 19 .
3 5 7 3 5 7
Now the problem of evaluating a 4 × 4 determinant has been reduced to
one of a 3 × 3 determinant. In this determinant, add 18 times column one
to column two and then 8 times column one to column three to obtain
−1 18 8 −1 0 0
166 83
8 22 19 = 8 166 83 = (−1) = −249.
59 31
3 5 7 3 59 31
Putting the pieces together,
−3 3 9 6
1 −2 15 6
= (−3)(−249) = 747.
7 1 1 5
2 1 −1 3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.6. DETERMINANTS 319

For Problems 7–10, we just give the value of the determinant.

7. −122

8. 293

9. 72

10. −2, 667

11. −15, 698

12. 1, 693

13. 3, 372

14.
1 a a2 1 a a2
1 b b2 = 0 b − a b − a2
2

1 c c2 0 c−a c2 − a2
1 a a2 1 a a2
= 0 b − a (b − a)(b + a) = (b − a)(c − a) 0 1 b + a
0 c − a (c − a)(c + a) 0 1 c+a
1 b+a
= (b − a)(c − a) = (b − a)(c − a)(c − b).
1 c+a

15. Add columns two, three and four to column one, then factor (a + b + c + d)
out of column one to obtain
a b c d 1 b c d
b c d a 1 c d a
= (a + b + c + d)
c d a b 1 d a b
d a b c 1 a b c

Now add
(−1)row two + row three − row four
to row one and factor out b − a + d − c from the new row one to obtain

1 b c d
1 c d a
(a + b + c + d)(b − a + d − c) .
1 d a b
1 a b c

16. Define the function

1 z y
L(x, y) = 1 x2 y2 = (y2 − y3 )x + (x3 − x2 )y + x2 y3 − x3 y2 .
1 x3 y3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 320CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

Notice that L(x, y) has the form ax+by +c, where a = y2 −y3 , b = x3 −x2
and c = x2 y3 − x3 y2 are given numbers. The graph of L(x, y) = 0 is a
straight line in the x, y− plane. Because

L(x2 , y2 ) = 0 and L(x3 , y3 ) = 0,

both of these points are on the line L(x, y) = 0.

Finally, L(x1 , y1 ) = 0 (x1 , y1 ) is also on this line, and this occurs only if

1 x1 y1
1 x2 y2 = 0.
1 x3 y3

17. Use induction to prove that, for n = 2, 3, · · · , the determinant of an n × n

upper triangular matrix is the product of the main diagonal elements.
For n = 2, this is obvious because

a11 a12
= a11 a22 .
0 a22

Now suppose the statement is true for some n ≥ 2. We want to prove that
it is true for n + 1. Let A be an n + 1 × n + 1 upper triangular matrix.
Then ai1 = 0 for i = 2, · · · , n + 1, so by expanding by column one, we
have
|A| = a11 |B|,
where B is the n × n upper triangular matrix obtained by deleting row
one and column one of A. By the inductive hypothesis,

|B| = a22 a33 · · · an+1,n+1 .

Then
|A| = a11 a22 · · · an+1,n+1 .
This completes the proof by induction.

11.7 Cramer’s Rule

1. |A| = 47 6= 0 so Cramer’s rule applies:

1 5 −4 11 1 15 5 100
x1 = = − , x2 = =− .
47 −4 1 47 47 8 −1 47

2. |A| = −3 and the solution is

1 3 4 1 1 3
x1 = − = −1, x2 = − = 1.
3 0 1 3 1 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.7. CRAMER’S RULE 321

3. |A| = 132 and the solution is

0 −4 3
1 66 1
x1 = −5 5 −1 = − =− ,
132 132 2
−4 6 1

8 0 3
1 114 19
x2 = 1 −5 −1 = − =− ,
132 132 22
−2 −4 1
and
8 −4 0
1 24 2
x3 = 1 5 −5 = = .
132 132 11
−2 6 −4

4. |A| = 108 and the solution is

63 7 165 55 −943 9
x1 = − = − , x2 = − = − , x3 = =− .
108 12 108 36 108 4

5. |A| = −6 and
5 10 5
x1 = , x2 = − , x3 = − .
6 3 6

6. |A| = −130 and

197 255 1260 42 173

x1 = , x2 = , x3 = , x4 = , x5 = .
130 130 130 130 130

7. |A| = 4 and the solution is

172 109 43 37
x1 = − = −86, x2 = − , x3 = − , x4 = .
2 2 2 2

8. |A| = 12 and
117 63 3 21
x1 = , , x3 = , x4 = .
12 12 2 12

9. |A| = 93 and

33 409 1 116
x1 = , x2 = − , x3 = − , x4 = .
93 93 93 93

10. |A| = 42,

69 162 24 54
x1 = , x2 = , x3 = , x4 = − .
21 21 21 21

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 322CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

11.8 The Matrix Tree Theorem

1. The tree matrix for this graph is
 
2 0 −1 0 −1
0
 2 −1 −1 0 

T= −1 −1 4 −1 −1
.
 0 −1 −1 3 −1
−1 0 −1 −1 3
Evaluate any 4 × 4 cofactor of T to obtain 21 as the number of spanning
trees in the labeled graph.
2.  
4 −1 −1 −1 0 −1
−1 3 −1 −1 0 0
 
−1 −1 2 0 0 0 
T=
 
−1 −1 0 4 −1 −1
0 0 0 −1 2 −1
−1 0 0 −1 −1 3
and each cofactor yields 55 spanning trees in G.
3.  
4 −1 0 −1 −1 −1
−1 2 −1 0 0 0
 
 0 −1 3 −1 −1 0 
T=
 
−1 0 −1 4 −1 −1

−1 0 −1 −1 3 0
−1 0 0 −1 0 2
and each cofactor gives 61 as the number of spanning trees
4.  
4 −1 −1 0 −1 −1
−1 3 −1 −1 0 0
 
−1 −1 3 −1 0 0
T=
 
 0 −1 −1 3 −1 0 

−1 0 0 −1 3 0
−1 0 0 0 −1 2
and each cofactor gives 64 for the number of spanning trees.
5.  
3 −1 0 0 −1 −1
−1 3 −1 0 −1 0 
 
 0 −1 4 −1 −1 −1
T=
 
0 0 −1 2 −1 0 
−1 −1 −1 −1 4 0
−1 0 −1 0 0 2
and the number of spanning trees is 61.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 11.8. THE MATRIX TREE THEOREM 323

6. The tree matrix for Kn is the n × n matrix

 
n−1 −1 −1 ··· −1
 −1 n −1 −1 ··· −1
 
 −1 −1 n−1 ··· −1
T= .

 .. .. .. ..

 . . . ··· .

−1 −1 −1 ··· n−1

We will compute −1)1+1 M11 , which is the n−1×n−1 determinant formed

by deleting row one and column one of T. In M11 , add the last n − 2 rows
to row one to obtain the new n − 1 × n − 1 determinant still equal to M11 :

1 1 1 ··· 1
−1 n − 1 −1 ··· −1
M11 = −1 −1 n−1 ··· −1 .
.. .. .. ..
. . . ··· .
−1 −1 −1 ··· n−1

Subtract column one of this determinant from each other column. This
does not change the value of the determinant, and we now have

1 0 0 0 ··· 0
−1 n 0 0 · · · 0
M11 = −1 −1 n 0 · · · 0 .
.. .. .. .. .. ..
. . . . . .
−1 0 0 0 · · · 0

This is a lower triangular n − 1 × n − 1 determinant, which is therefore

equal to the product of the diagonal elements, which is nn−2 . Therefore
the number of spanning trees in Kn is nn−2 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 324CHAPTER 11. MATRICES, DETERMINANTS AND LINEAR SYSTEMS

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 12

Eigenvalues and
Diagonalization

12.1 Eigenvalues and Eigenvectors

1.
pA (λ) = |λI2 − A| = λ2 − 2λ − 5
√ √
so the eigenvalues of A are 1 + 6 and 1 − 6, with eigenvectors, respec-
tively,
√   √ 
6 − 6
V1 = , V2 = .
2 1

The Gerschgorin circles are of radius 3 about (1, 0) and radius 2 about
(1, 0).

2. The characteristic equation of A is

λ2 − 2λ − 8 = 0.

Eigenvalues and eigenvectors are

   
0 6
λ1 = 4, V1 = , λ2 = −2, V2 = .
4 −1

Gerschgorin circles are of radius 1 about (4, 0) and radius 0 about −2.
The circle of radius 0 is a “degenerate” circle, consisting of just the center
point.

3. The characteristic equation is

λ2 + 3λ − 10 = 0

325

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 326 CHAPTER 12. EIGENVALUES AND DIAGONALIZATION

and eigenvalues and eigenvectors are

   
7 0
λ1 = −5, V1 = , λ2 = 2, V2 = .
−1 1

One Gerschgorin circle has radius 1 and center (2, 0), and the other is the
degenerate circle of radius 0 about (−5, 0).
4. pA (λ) = λ2 − 10λ + 18,
√ √
   
2√ 2√
λ1 = 5 + 7, V1 = , λ2 = 5 − 7, V2 =
1− 7 1+ 7
Gerschgorin circles have radius 2, center (6, 0), and radius 3, center (4, 0).
5. pA (λ) = λ2 − 3λ + 14,
√ 
√

−1 + 47i
λ1 = (3 + 47i)/2, V1 =
4
√ 
√

−1 − 47i
λ2 = (3 − 47i)/2, V2 =
4
The Gerschgorin circles have radius 6, center (1, 0), and radius 2, center
(2, 0).
6. pA (λ) = λ2 , and eigenvalues are λ1 = λ2 = 0. All engenvectors are
nonzero scalar multiples of
 
1
V= .
0
The lone Gerschgorin circle has radius 1 and center the origin.
7. pA (λ) = λ3 − 5λ2 + 6λ.
     
0 2 0
λ1 = 0, V1 = 1 , λ2 = 2, V2 = 1 , λ3 = 3, V3 = 2
0 0 3
The Gershgorin circle has radius 3, center (0, 0).
8. pA (λ) = (λ + 1)(λ2 − λ − 7)
   
0 √ 2√
λ1 = 1, V1 = 0 , λ2 = (1 + 29)/2, V2 = 5 + 29
1 0
and  
√ 2√
λ3 = (1 − 29)/2, V3 = 5 − 29
0
Gershgorin circles have radius 1, center (−2, 0), and radius 1, center (3, 0).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 12.1. EIGENVALUES AND EIGENVECTORS 327

9. pA (λ) = λ2 (λ + 3)
   
1 1
λ1 = −3, V1 = 0 , λ2 = λ3 = 0, V2 = 0
0 3

All eigenvectors associated with the double eigenvalue 0 are constant mul-
tiples of V2 . The Gershgorin circle has radius 2, center (−3, 0).
10. pA (λ) = λ(λ2 + 2)
     
0 √ 1 √ 1
λ1 = 0, V1 = 1 , λ2 = 2i, V2 =  −1
√
 , , λ3 = − 2i, V3 =  −1 
√
0 − 2i 2i

The Gershgorin circles have center (0, 0) and radii 1 and 2.

11. pA (λ) = (λ + 14)(λ − 2)2 ,
   
−16 0
λ1 = −14, V1 =  0  λ2 = λ3 = 2, V2 = 0 .
1 1

All eigenvectors associated with λ2 are constant multiples of V2 . The

Gershgorin circles have radius 1, center (−14, 0) and radius 3, center (2, 0).
12. pA (λ) = (λ − 3)(λ2 + λ − 42),
     
0 30 0
λ1 = 6, V1 =  1  , λ2 = 3, V2 = −2 , λ3 = −7, V3 = 8
−1 5 5

The Gerschgorin circles have radius 9, center (−2, 0) and radius 5, center
(1, 0).
13. pA (λ) = λ(λ2 − 8λ + 7),
     
14 6 0
λ1 = 0, V1 =  7  , λ2 = 1, V2 = 0 , λ3 = 7, V3 = 0
10 5 1

The Gershgorin circles have radius 2, center (1, 0), and radius 5, center
(7, 0).
14. pA (λ) = λ2 (λ2 + 2λ − 1),
   
1 0
2 0
λ1 = λ2 = 0, V1 = 
0 , V2 = 1
  

1 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 328 CHAPTER 12. EIGENVALUES AND DIAGONALIZATION

so the repeated eigenvalue 0 has two linearly independent eigenvectors

associated with it. For the other two eigenvalues,
   
1√ 1√
√ 1 + 2 √ 1 − 2
λ3 = −1 + 2, V3 =   0  , λ4 = −1 − 2, V4 =  0 
  

0 0

The Gershgorin circles have radius 1, center (−2, 0) and radius 2, center
(0, 0).
15. pA (λ) = (λ − 1)(λ − 2)(λ2 + λ − 13),
   
−2 0
−11 0
λ1 = 1, V1 = 
 0  , λ2 = 2, V2 = 1
  

1 0
√   √ 
√ 53 − 7 √ − 53 − 7
−1 + 53  0  , λ4 = −1 − 53 , V4 = 
  0 
λ3 = , V3 =  
2  0  2  0 
2 2
The Gershgorin circles have radius 2, center (−4, 0) and radius 1, center
(3, 0).
16. pA (λ) = λ2 (λ − 1)(λ − 5),
   
1 1
−4 0
λ1 = 1, V1 =  0  , λ2 = 5, V2 = 0
  

0 0
 
0
0
, λ3 = λ4 = 0, V3 = 
1


0
All eigenvectors associated with the double eigenvalue 0 are constant mul-
tiples of just the one eigenvector V3 . The Gershgorin circles have radius
10, center (5, 0), radius 9, center (1, 0), radius 9, center (0, 0).
17. We know that AE = λE. Then

A(AE) = A2 E = A(λE)
= λAE = λ(λE) = λ2 E.

This says that λ2 is an eigenvalue of A2 with eigenvector E. It is not a

routine inductive argument to show that λn is an eigenvalue of An with
eigenvector E, for any positive integer n.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 12.2. DIAGONALIZATION 329

18. The characteristic polynomial of A is

pA (λ) = |λIn − A|.

The
pA (0) = | − A|
is the constant term in this polynomial. But

| − A| = (−1n )|A|

because −A is A with each of the n rows multiplied by −1, and each

such multiplication introduces a factor of −1 into the determinant of the
resulting matrix.
Now, if A is singular, then |A| = 0. But then

pA (0) = (−1)n |A| = 0,

so 0 is a root of the characteristic polynomial, and is therefore an eigen-

value of A.

12.2 Diagonalization
1. The characteristic equation is λ2 − 3λ + 4 = 0, so the eigenvalues are
√ √
3 + 7i 3 − 7i
λ1 = and λ2 = .
2 2
Corresponding eigenvectors are
   
2√ 2√
V1 = and V2 = .
−3 + 7i −3 − 7i

The matrix  
2√ 2√
P=
−3 + 7i −3 − 7i
diagonalizes A, and
 √ 
−1 (3 + 7i)/2 0
√
P AP = .
0 (3 − 7i)/2

If we wrote the eigenvectors in different order in defining the columns of

P, then the columns of P−1 AP would be reversed.
2. The characteristic equation is λ2 − 8λ + 12 = 0 so the eigenvalues are
λ1 = 2 and λ2 = 6. Corresponding eigenvectors are
   
−1 3
V1 = and .
1 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 330 CHAPTER 12. EIGENVALUES AND DIAGONALIZATION

The matrix  
−1 3
P=
1 1
diagonalizes A, and  
−1 2 0
P AP = .
0 6

3. The characteristic equation is λ2 −2λ+1 = 0, with repeated root 1. Every

eigenvector is a scalar multiple of
 
0
V1 = .
1

Because A does not have two independent eigenvectors, A is not diago-

nalizable.
4. Eigenvalues and corresponding eigenvalues are
   
1 3
λ1 = −5, V1 = , λ2 = 9, V2 = .
0 14

The matrix  
1 3
P=
0 14
diagonalizes A, and  
−1 −5 0
A AP = .
0 9

5. The eigenvalues and corresponding eigenvectors are

     
0 5 0
λ1 = 0, V1 = 1 , λ2 = 5, V2 = 1 , λ3 = −2, V3 = −3 .
0 0 2

The matrix  
P=0 5 0
 1 1 −3
0 0 2
diagonalizes A and
 
0 0 0
P−1 AP = 0 5 0 .
0 0 −2

6. Eigenvalues and corresponding eigenvectors are

   
−2 √ 0
λ1 = 0, V1 = −3 , λ2 = (3 + 17)/2, V2 =  4√ 
1 3 + 17

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 12.2. DIAGONALIZATION 331

and  
√ 0
λ3 = (3 − 17)/2, V3 =  4√  .
3 − 17
The matrix  
−2 0 0
P = −3 4√ 4√ 
1 3 + 17 3 − 17
diagonalizes A, and
 
0 0
√ 0
P−1 AP = 0 (3 + 17)/2 √0 .
0 0 (3 − 17)/2

7. Eigenvalues and corresponding eigenvectors are

   
0 −3
λ1 = 1, V1 = 1 , λ2 = λ3 = −2, V2 =  1  .
0 0
All eigenvectors associated with the repeated eigenvalue −2 are scalar
multiples of V2 . Because A does not have three linearly independent
eigenvectors, A is not diagonalizable.
8. Eigenvalues and corresponding eigenvectors are
     
1 0 0
λ1 = 2, V1 = 0 , λ2 = 2 + i, V2 = −i , λ3 = 2 − i, V3 =  i  .
0 1 1
Let  
1 0 0
P = 0 −i i  .
0 1 −1
Then P diagonalizes A and
 
2 0 0
P−1 AP = 0 2+i 0 .
0 0 2−i

9. The characteristic equation is

(λ − 1)(λ − 4)(λ2 + 5λ + 5) = 0.
√ √
Eigenvalues are λ1 = 1, λ2 = 4, λ3 = (−5+ 5)/2, and λ4 = (−5− 5)/2).
Corresponding eigenvectors are
   
1 0
0 1
V1 = 
0 , V2 = 0 ,
  

0 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 332 CHAPTER 12. EIGENVALUES AND DIAGONALIZATION
   
0
√ 0
√
(2 − 3 5)/41 (2 + 3 5)/41
V3 =  √ √
 (−1 + 5)/2  , V4 =  (−1 − 5)/2  .
  

1 1
Let  
1 0 0
√ 0
√
0 1 (2 − 3 √5)/41 (2 + 3 √5)/41
P= .
0 0 (−1 + 5)/2 (−1 − 5)/2 
0 0 1 1
Then P diagonalizes A:
 
1 0 0 0 √
P−1 AP = 0 1 0 00√ 0 (−5 + 5)/2 0 .
0 0 0 (−5 − 5)/2

10. The characteristic equation is (λ + 2)4 = 0, so A has a repeated eigen-

value −2 of multiplicity 4. We find that A has only three independent
eigenvectors,      
0 0 0
1 0 0
  ,   , and   .
0 1 0
0 0 1
Because A does not have four independent eigenvectors, A is not diago-
nalizable.

11. Let  
λ1 0 0 ··· 0
0
 λ2 0 0 ··· 0 

D=0
 0 λ3 0 ··· 0 .

 .. .. .. .. .. 
. . . . ··· . 
0 0 0 0 ··· λn
Then
D = P−1 AP,
so
A = PDP−1 .
Then

Ak = (PDP−1 )k
= (PDP−1 )(PDP−1 · · · (PDP−1 )
= PDk P−1 ,

with the interior pairings of P and P−1 canceling.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 12.2. DIAGONALIZATION 333

Problems 12–15 can be solved using the idea of Problem 11, coupled with
the fact that the kth power of a diagonal matrix is the diagonal matrix formed
by raising each diagonal element to the kth power.

12. Eigenvalues of A are −1, −6. Make respective eigenvectors the columns
of a matrix  
−3 1
P= .
2 1
Then  
−1/5 2/5
P−1 = .
1/5 3/5
Then  
−1 −1 0
P AP = D = .
0 −6
Then  
5 5 −1 −3111 −4665
A = PD P = .
−3110 −4666

13. Eigenvalues of A are −1, −5, and the matrix of respective eigenvectors,
 
4 0
P=
1 1

diagonalizes A to  
−1 0
D= .
0 −5
Now,  
1/4 0
P−1 =
−1/4 1
so compute  
1 0
A6 = PD6 P−1 = .
−3906 15625
p √
14. Eigenvalues of A are −3 + (10), −3 − 10. Form P from respective
eigenvectors:  √ √ 
1 + 10 1 − 10
P= .
3 3
Then P diagonalizes A to
 √ 
−3 + 10 0√
D= .
0 −3 − 10

Then  
493 −694
A4 = PD4 P−1 = .
−684 949

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 334 CHAPTER 12. EIGENVALUES AND DIAGONALIZATION
√ √
15. Eigenvalues of A are 2, − 2. Form
√ √ 
2 − 2
P= .
1 1

Then √ 
P −1
= √2/4 1/2
.
− 2/4 1/2
Let √ 
2 0
√ .
B=
0 − 2
Then  
6 6 −1 8 0
A = PD P = .
0 8

16. Suppose A2 is an n × n diagonalizable matrix. Then A must have n lin-

early independent eigenvectors V1 , · · · , Vn , corresponding, respectively,
to eigenvalues λ1 , · · · , λn . We know that

pA (λj ) = (A2 − λj In )X = O

for j = 1, · · · , n. But then

p p
pA (λj ) = (A − λj In )(A + λj In )
p p
= pA ( λj )pA (− λj ) = 0.

Then p p
pA ( λj ) = 0 or pA (− λj ) = 0.
p p
Then λj or −λj is an eigenvalue of A with eigenvector Vj . This
means that A has n linearly independent eigenvectors, and is therefore
diagonalizable.

12.3 Special Matrices and Their Eigenvalues and

Eigenvectors
In Problems 1–12, find independent eigenvectors for the given matrix. Normalize
these by dividing each eigenvector by its length. These normalized eigenvectors
form columns of an orthogonal matrix that diagonalizes the given matrix.
It is routine to show that eigenvectors are orthogonal by taking their dot
product. We will omit the arithmetic of this verification.

1. We find eigenvectors
   
1 −2
V1 = , V2 = .
2 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 12.3. SPECIAL MATRICES AND THEIR EIGENVALUES AND EIGENVECTORS335

Divide each by its norm to get unit eigenvectors and form the orthogonal
matrix  √ √ 
1/√5 −2/√ 5
Q= .
2/ 5 1/ 5
This is an orthogonal matrix that diagonalizes A.
2.    
10
√ 10
√
V1 = , V2 = .
7 + 149 7 − 149
Normalize these to form columns of the orthogonal matrix
√ 10 √ √ 10 √
 
298+14 149 298−14 149
Q =  7+√149 √ .
√ √ √ 7− 149
√
298+14 149 298−14 149

3. Eigenvectors are
 √   √ 
1+ 2 1− 2
V1 = , V2 = .
1 1
Normalize these to form
 √ √ 
√1+
2
√ √1− 2
√
Q=  4+2 2 4−2 2  .
√ 1√ √ 1 √
4+2 2 4−2 2

√ √
4. Eigenvalues are 17 and − 17, with eigenvectors, respectively,
√   √ 
17 − 4 − 17 − 4
V1 = , V2 = .
1 1
Normalize these vectors to form an orthogonal matrix Q that diagonalizes
A: √ √ 
1 17 − 4 − 17 − 4
Q= p √ .
34 − 8 17 1 1
√ √
5. Eigenvalues of A are 3, 2 − 1 and − 2 − 1, with corresponding eigen-
vectors
     
0 √ 1 √1
V1 = 0 , V2 =  2 − 1 V3 = − 2 − 1 .
1 0 0
For an orthogonal matrix that diagonalizes A, let
√ 1 √ √ 1√
 
0
 4−2 2
√ 4+2
√ 2
√ 2−1√ − 2−1 
√
Q = 0 √ .

 4−2 2
√ 4+2
√ 2
√ 2−1√ − 2−1
√
1 √
4−2 2 4+2 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 336 CHAPTER 12. EIGENVALUES AND DIAGONALIZATION

6. Eigenvalues of A are
1 √ 1 √
1, (5 + 5), (5 − 5)
2 2
with eigenvectors, respectively,
     
1 0√ 0√
V1 = 0 , V2 = −1 + 5 , V3 = −1 − 5 .
0 2 2

Normalize these eigenvectors and use them as columns of an orthogonal

matrix that diagonalizes A:
 
1 0√ 0√
0 √−1+ 5 √−1− √5 
Q= √
10+2 5  .

 10−2 5
2 2
0 √ √ √ √
10−2 5 10+2 5

7. Eigenvalues are
1 √ 1 √
7, (5 + 41), (5 − 41)
2 2
with corresponding eigenvectors
   √   √ 
0 5 + 41 5 − 41
V1 = 1 , V2 =  0  , V3 =  0 .
0 4 4

The following orthogonal matrix diagonalizes A:

 √ √ 
1 √ 5+ 41√ √ 5− 41√
 82+10 41 82−10 41 
Q= 1 0 0 .

4 4
0 √ √ √ √
82+10 41 82−10 41

√ √
8. Eigenvalues are 0, 1 + 17, 1 − 17, with corresponding eigenvectors
   √   √ 
0 1 + 17 1 − 17
V1 = 0 , V2 =  −4  , V3 =  −4  .
1 6 6

For an orthogonal matrix that diagonalizes A, use the normalized eigen-

vectors as columns:
√ √
0 √1+ 17 √1− 17
 
√ √
70+2 17 70−2 17 
0 √ −4 √ √ −4 √  .

Q=  70−2 17 70+2 17 
1 √ 6 √ √ 6√
70−2 17 70+2 17

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 12.3. SPECIAL MATRICES AND THEIR EIGENVALUES AND EIGENVECTORS337

9. The matrix is not hermitian, skew-hermitian or unitary. Eigenvalues are

2, 2.
10. The matrix is not hermitian, skew-hermitian or unitary. Eigenvalues are
−1, −1.
√ √
11. The matrix is skew-hermitian because St = −S. Eigenvalues are 0, 3i, − 3i.
12. The matrix is unitary. Eigenvalues are
1 √ √ 1 √ √
1, (1 + i)( 2 + 6i), (−1 − i)(− 2 + 6i).
4 4

13. The matrix is not unitary, hermitian or skew-hermitian. Eigenvalues are

2, i, −i.
14. The matrix is hermitian, because Ht = H. Eigenvalues are
√ √
−1 − 41 −1 + 41
1, , .
2 2

15. Suppose H is hermitian. Then

H = Ht .

Then
HHt = HHt = HH = HH.

16. If H is hermitian, then Ht = H, so the diagonal elements satisfy hjj = hjj ,

so these diagonal elements must be real.
17. Suppose S is skew-hermitian. Then St = −S, so

sjj = −sjj for j = 1, 2, · · · , n.

Write sjj = ajj + ibjj . Then

sjj = ajj + ibjj = −ajj + ibjj = −ajj + ibjj .

But then each ajj = −ajj , so ajj = 0 for j = 1, 2, · · · , n. This makes each
diagonal element of S either pure imaginary (if bjj 6= 0) or zero.
18. Suppose U and V are unitary n × n matrices. Then
t t
U−1 = U and V−1 = V .

But then
t t t
(UV)−1 = V−1 U−1 = V U = ((U )(V ))t = UV ,

so UV satisfies the equation defining a matrix to be unitary.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 338 CHAPTER 12. EIGENVALUES AND DIAGONALIZATION

12.4 Quadratic Forms

1. The matrix of the quadratic form is
 
−5 2
A= .
2 3
√ √
This matrix has eigenvalues −1 + 2 5, −1 − 2 5 and the quadratic form
has standard form
√ √
(−1 + 2 5)y12 + (−1 − 2 5)y22 .

2. The matrix is  
4 −6
A=
−6 1
with eigenvalues
1 √ 1 √
(5 + 153) and (5 − 153).
2 2
The standard form of this quadratic form is

1 √ 1 √
(5 + 153)y12 + (5 − 153)y22 .
2 2

3. The matrix is  
−3 2
2 7
√
with eigenvalues 2 ± 29. The standard form is
√ √
(2 + 29)y12 + (2 − 29)y22 .

4. The matrix is  
4 −2
−2 1
√
with eigenvalues 3 ± 17)/2. The standard form is

1 √ 1 √
(3 + 17)y12 + (3 − 17)y22 .
2 2

5. The matrix is  
0 −3
−3 4
√
with eigenvalues 2 ± 13. The standard form is
√ √
(2 + 13)y12 + (2 − 13)y22 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 12.4. QUADRATIC FORMS 339

6. The matrix is  
5 2
2 2
with eigenvalues 1, 6. The standard form is

y12 + 6y22 .

7. The matrix is  
0 −1
−1 2
√
with eigenvalues 1 ± 2. the standard form is
√ √
(1 + 2)y12 + (1 − 2)y22 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 340 CHAPTER 12. EIGENVALUES AND DIAGONALIZATION

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 13

Systems of Linear
Differential Equations

13.1 Linear Systems

1. The two given solutions are linearly independent because neither is a con-
stant multiple of the other. Use them as columns of a fundamental matrix
 2t
3e6t

−e
Ω(t) = .
e2t e6t

Notice that Ω(0) has a nonzero determinant, which is also an indicator

that the columns are independent.
Now we have a general solution

X(t) = Ω(t)C,

in which C is a 2 × 1 matrix of constants. To satisfy the initial condition,

we need to choose C so that
 
0
Ω(0)C = .
4

This requires that

    
−1 3 c1 0
= .
1 1 c2 4

Then c1 = 3, c2 = 1, so the solution of the initial value problem is

 2t
3e6t −3e2t + 3e6t
   
−e 3
X(t) = = .
e2t e6t 1 3e2t + e6t

341

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 342 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

2. The two solutions are independent because Φ1 (0) and Φ2 (0) are indepen-
dent in R2 . We can form a fundamental matrix
 5t
e3t

e
Ω(t) = .
3e5t e3t
We now have a general solution
X(t) = Ω(t)C,
in which C is an arbitrary 2 × 1 matrix of real numbers. To solve the
initial value problem, we need
      
−2 1 1 c1 −2
X(0) = = Ω(0)C = = .
1 3 1 c2 1
Then c1 = 3/2 and c2 = −7/2, so the solution of the initial value problem
is
 5t
e3t 3e5t −7e3t
   
e 3/2 1
X(t) = Ω(t)C = = .
3e5t e3t −7/2 2 9e5t − 7e3t

3. Because Φ(0) and Φ(0) are independent in R2 , these solutions are inde-
pendent. Form the fundamental matrix
√ √ √ √ !
(2 + 2 3)e√(1+2 3t) (2 − 2 3)e√(1−2 3)t
Ω(t) = .
e(1+2 3)t e(1−2 3)t

Then X(t) = Ω(t)C is a general solution. To solve the initial value prob-
lem, we need  
2
Ω(0)C =
2
and we find that we must choose
1√ 1√
c1 = 1 − 3, c2 = 1 + 3.
6 6
4. From the given solutions we can write the fundamental matrix
Ω(t) =
 √ √ √ √ √ √ 
3t/2 − cos( 15t/2) −√ 15 sin( 15t/2) 15 cos( 15t/2)
√ − sin( 15t/2)
e
8 cos( 15t/2) 8 cos( 15t/2)

The general solution is Ω(t)C. For the solution of the initial value prob-
lem, solve for C so that  
0
Ω(0)C = .
7
We find that √ 
7 15
C= √ .
8(1 + 15) 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.2. SOLUTION OF X0 = AX WHEN A IS CONSTANT 343

5. Form the fundamental matrix

e−3t
 t
−et

e
Ω(t) = et 0 3e−3t  .
0 et e−3t

The general solution is X(t) = Ω(t)C. To solve the initial value problem,
we need C such that  
1
Ω(0)C = −3 .
5
Solve this equation to get
 
24
C =  14  .
−9

13.2 Solution of X0 = AX When A Is Constant

In these problems, different fundamental matrices may be found, depending on
the choice of eigenvectors used corresponding to eigenvalues of the coefficient
matrix.
In each problem, the general solution has the form X = Ω(t)C, where Ω(t)
is a fundamental matrix. We will give one choice for Ω(t).

1.  3t 
7e 0
Ω(t) =
5e3t e−4t
2.
2et e6t
 
Ω(t) =
−3et e6t
3.
1 e2t
 
Ω(t) =
−1 e2t
4.
et e−t e2t
 

Ω(t) = et e−t 2e2t 

et 2e−t e2t
5.
2e3t −e−4t
 
1
Ω(t) =  6 3e3t 2e−4t 
−13 −2e3t e−4t
6.
2et e−t
 
Ω(t) =
et e−t

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 344 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

7. A has eigenvalues and eigenvectors

 

−2i
2 + 2i, 2i 1 , 2 − 2i, .
1

Write      
2i 0 2
= +i
1 1 0
to form two independent solutions
 
−2 sin(2t)
Φ1 (t) = e2t
cos(2t)

and  
2 cos(2t)
2t
Φ2 (t) = e .
sin(2t)
Use these as columns of a fundamental matrix
 
−2 sin(2t) 2 cos(2t)
Ω(t) = .
cos(2t) sin(2t)
√
8. A has complex eigenvalues, with the eigenvalue 3+ 5i having eigenvector
 √    √ 
2 + 5i 2 5
= +i .
3 3 0

Two independent solutions are

 √ √ √ 
2t 2 cos( 5t) − √ 5 sin( 5t)
Φ1 (t) = e
3 cos( 5t)

and √ √ √ √ 
5 cos( 5t) −√ 5 sin( 5t)
2t
Φ2 (t) = e .
3 sin( 5t)
These form the columns of a fundamental matrix.
9. A has eigenvalues 2, 5, 5, with corresponding eigenvectors
   
1 −3
0 , −3 .
0 1

All eigenvectors associated with 5 are scalar multiples of this eigenvector.

Immediately we can write two independent solutions
 
1
Φ1 (t) = e2t 0
0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.2. SOLUTION OF X0 = AX WHEN A IS CONSTANT 345

and  
−3
Φ2 (t) = e5t −3 .
1
For a third solution, denote the eigenvector associated with 5 as E and let
Φ3 (t) = Ete5t + Ke5t .
Substitute this into X0 = AX and use the fact that AE = E to obtain
E + 5K = AK.
If we let  
a
K = b ,
c
we obtain
−3a + 5b + 6c = −3
3b + 9c = −3
−b − 3c = 1.
Then a = −2/3, b = −1, c = 0, so
 
−2/3
K=
−1 0
and we obtain the third solution
−3te5t − (2/3)e5t
 

Φ3 (t) =  −3te5t − e5t  .

te5t
The three solutions obtained are linearly independent and can be used to
form the columns of a fundamental matrix.
10. The coefficient matrix has eigenvalues 2, −2 − 3 and we can use the inde-
pendent eigenvectors to form the fundamental matrix
e−3t
 2t 
4e 0
Ω(t) = 3e2t e−t −3e−3t  .
3e2t e−2t 2e−3t

11. The coefficient matrix has eigenvalues 2+2i and 2−2i, with corresponding
eigenvectors    
2i −2i
, .
1 0
From these form two independent solutions which for the columns of the
fundamental matrix
−2e2t sin(2t) 2e2t cos(2t)
 
Ω(t) = .
e2t cos(2t) e2t sin(2t)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 346 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

12. The coefficient matrix has eigenvalues −1 ± 2i, and the eigenvector asso-
ciated with −1 + 2i is  
5
.
−1 + 2i
Form the fundamental matrix
 
5 cos(2t) 5 sin(2t)
Ω(t) = et .
− cos(2t) − 2 sin(2t) 2 cos(2t) − sin(2t)

13. The coefficient matrix has eigenvalues 1 ± i. An eigenvector associated

with 1 + i is  
2+i
.
1
Form the fundamental matrix
 
t 2 cos(t) − sin(t) cos(t) + 2 sin(t)
Ω(t) = e .
cos(t) sin(t)

14. Eigenvalues of the coefficient matrix are −1, −1, 2. Eigenvectors are
   
0 2
1 , 1 .
1 1

The second eigenvector is associated with 2, and every eigenvector asso-

ciated with −1 is a scalar multiple of the first. Immediately we have two
independent solutions  
0
Φ1 (t) = 1 e−t
1
and  
2
Φ2 (t) = 1 e2t .
1
Try
Φ3 (t) = Ete−t + Ke−t ,
where E is the eigenvector associated with the eigenvalue −1, and
 
a
K = b
c

is to be determined. Substitute Φ3 (t) into X0 = AX and use the fact that

AE = −E to get
E − K = AK.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.2. SOLUTION OF X0 = AX WHEN A IS CONSTANT 347

Solve the equations for a, b and c to get a = 1, b = 2, c = 0, in which c is

arbitrary and has been chosen to be zero. Then
     
0 1 1
Φ3 (t) = 1 te−t + 2 e−t = t + 2 e−t .
1 0 t
These solutions form the columns of a fundamental matrix for the system.
15. The coefficient matrix has eigenvalues −2, −1 + 2i, −1 − 2i. From −2 we
obtain the solution  
0
Φ(t) = e−2t 0 .
1
An eigenvalue for −1 + 2i is
   
1 0
1 + i 2 .
3 0
This gives us two more solutions:
 
cos(2t)
Φ2 (t) = e−t cos(2t) − 2 sin(2t)
3 cos(2t)
and  
sin(2t)
Φ3 (t) = e−t 2 cos(2t) + sin(2t) .
3 sin(2t)
These solutions form the columns of a fundamental matrix.
16. The coefficient matrix has eigenvalues 2, 2, and all eigenvectors are scalar
multiples of  
0
E= .
1
One solution is  
0
Φ1 (t) = .
e2t
For a second solution, try
Φ2 (t) = Ete2t + Ke2t .
Substitute this into the differential equation to find that we can take
 
1/5
K= .
0
This yields a second solution
e2t /5
 
Φ2 (t) = .
te2t

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 348 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

17. The coefficient matrix has eigenvalues 3, 3, and every eigenvector is a scalar
multiple of  
1
E= .
0
One solution is  3t 
e
Φ1 (t) = .
0
Attempt a second solution

Φ2 (t) = Ete3t + Ke3t .

Solve for K to get  

0
K= .
1/2
Then
te3t
 
Φ2 (t) = .
e3t /2
These solutions form columns of a fundamental matrix.

18. The eigenvalues of A are 1, 1, 1 and all eigenvectors are scalar multiples of
 
0
E = 0 .
1

One solution is  
0
Φ1 (t) =  0  .
et
For a second solution, let

Φ2 (t) = Etet + Ket

and obtain  
1/4
K =  0 .
0
Then
et /4
 
Φ2 (t) = .
tet
For a third solution, let

1 2
Φ3 (t) = Et + Ktet + Met .
2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.2. SOLUTION OF X0 = AX WHEN A IS CONSTANT 349

Substitute this into the system and solve for

 
1/2
M =  0 .
−2

This gives us 1
+ 14 t + 21
2

2t
Φ3 (t) = et  0 .
1 2
2t − 2
√ √
19. The coefficient matrix has eigenvalues 4 + 29i and 4 − 29i, with corre-
sponding eigenvectors
   √ 
−2 − 29
+i .
3 0

This gives us two independent solutions:

 √ √ √ 
4t −2 cos( 29t) +√ 29 sin( 29t)
Φ1 (t) = e
3 cos( 29t)

and  √ √ √ 
− 29 cos( 29t) √ − 2 sin( 29t) .
Φ2 (t) = e4t
3 sin( 29t)
√ √
20. The coefficient matrix has eigenvalues −1, −3 + 10, −3 − 10, with cor-
responding eigenvectors
   √   √ 
3 −4 − √10 −4 + √10
V1 = 1 , V2 = 10 + 3 10 , V3 = 10 − 3 10 .
6 4 4

The general solution is

√ √
X(t) = V1 e−t + V2 e(−3+ 10)t
+ V3 e(−3−10)t
.

These solutions can also be used as columns of a fundamental matrix.

21. The coefficient matrix has eigenvalues 1, 1, 3, 0, with corresponding eigen-
vectors
       
0 1 3 2
−2 0 2 0
V1 =   , V2 =   , V3 =   , V4 =  
       .
−2 0 2 1
1 0 0 0

We can write a general solution

X(t) = V1 et + V2 et + V3 e3t + V4 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 350 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

13.3 Exponential Matrix Solutions

In Problems 1–8, the exponential matrix can be obtained using the Putzer
algorithm or a software program.

1.
cos(2t) − 21 sin(2t) 1
 
At 2 sin(2t)
e =
− 52 sin(2t) cos(2t) + 21 sin(2t)

2.
−3t 1 −3t
2
+ 13 1

eAt = 3e 3 − 3e
2 2 −3t 1 −3t
3 − 3e 3e + 23

3.
 
cos(2t) + sin(2t) − sin(2t)
eAt = e3t
2 sin(2t) cos(2t) − sin(2t)

4.
√ √ √ !
cos( 7t/2) + 17 sin( 7t/2) − √27 sin( 7t/2)
e At
=e 3t/2 √ √ √
√4 sin( 7t/2) cos( 7t/2) − √17 sin( 7t/2)
7

5.
 
cos(2t) 2 sin(2t)
eAt = et
− 12 sin(2t) cos(2t)

6.
1 4 5t
− 25 e5t + 25
 
eAt = 5 + 5e
− 5 e + 52
2 5t 4
5 + 5e
1 5t

7.
√ √ √ !
−t/2
cos(3 3t/2) + √13 sin(3 3t/2) − √23 sin(3 3t/2)
e At
=e √ √ √
√2 sin(3 3t/2) cos(3 3t/2) − √13 sin(3 3t/2)
3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.3. EXPONENTIAL MATRIX SOLUTIONS 351

8. eAt is the 3 × 3 matrix [eij ], where

3 2t 2 1
e11 = e + cos(t) − sin(t),
5 5 5
2 1 1
e12 = sin(t) + cos(t) − e2t
5 5 5
1 1 1
e13 = sin(t) − cos(t) + e2t
5 5 5
3 3 2t 4
e21 = cos(t) − e − sin(t)
5 5 5
1 2t 4 3
e22 = e + cos(t) + sin(t)
5 5 5
7 1 1
e23 = sin(t) + cos(t) − e2t
5 5 5
3 3 2t 1
e31 = − cos(t) + e − sin(t)
5 5 5
1 1 2t 3
e32 = cos(t) − e − sin(t)
5 5 5
1 2t 4 2
e33 = e + cos(t) − sin(t)
5 5 5

9. First, because D is a diagonal matrix, Dn is a diagonal matrix for any

positive integer n, and the diagonal element of Dn is dnj . Then

∞
X 1 n n
eDt = D t
n=0
n!

and the diagonal element of eDt is

∞
X 1
(dj )n tn ,
n=0
n!

which is edj t .

10. Notice that

Bn = (P−1 AB)n
= (P−1 AP)(P−1 AP) · · · (P−1 AP)
= P−1 An P.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 352 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

Then
∞
Bt
X 1 −1
e = (P AP)n tn
n=0
n!
∞
X 1 −1 n n
= P A Pt
n=0
n!
∞
!
−1
X 1 n n
=P A t P
n=0
n!
= P−1 eAt P.

11. From the result of Problem 10,

eAt = PeDt P−1 ,

where P−1 AP = D, the diagonal matrix having the eigenvalues λ1 , · · · , λn

of A down its diagonal. But from the result of Problem 9, eDt is the di-
agonal matrix having diagonal elements eλj t .

13.4 Solution of X0 = AX + G for Constant A

1. The coefficient matrix A has the repeated eigenvalue 3, 3, and every eigen-
vector is a scalar multiple of
 
1
.
−1

One solution of the homogeneous system X0 = AX is

 
3t 1
e .
−1

Using methods from Section 13.2, find a second, independent solution of

this homogeneous system to write the fundamental matrix
 
1 + 2t 2t
Ω(t) = e3t .
−2t 1 − 2t

For a particular solution of the nonhomogeneous system, first compute

 
1 − 2t −2t
Ω−1 (t) = e−3t .
2t 1 + 2t

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.4. SOLUTION OF X0 = AX + G FOR CONSTANT A 353

Now use variation of parameters to compute a solution of the nonhomo-

geneous system. For this method, we need
Z
u(t) = Ω−1 (t)G(t) dt

−3et
Z   
1 − 2t −2t
= e3t dt
2t 1 + 2t e3t
Z  −2t
− 3e−2t − 2t −3te−2t − t2
  
6te
= dt = .
−6te−2t + 1 + 2t (3/2)(1 + 2t)e−2t + t + t2

The general solution is

X(t) = Ω(t)C + Ω(t)u(t)

  
3t 1 + 2t 2t c1
=e
−2t 1 − 2t c2
−3te−2t − t2
  
3t 1 + 2t 2t
+e
−2t 1 − 2t (3/2)(1 + 2t)e−2t + t + t2
e3t (c1 (1 + 2t) + 2c2 t) + t2 e3t
 
= 3t .
e (−2c1 t + c2 (1 − 2t)) + (t − t2 )e3t + 3et /2

2. A has eigenvalues 0, 0, and every eigenvector is a scalar multiple of

 
2
.
1

The associated homogeneous system has this eigenvector as a constant

solution. Apply the method of Section 13.2 to find a second, linearly
independent solution and produce the fundamental matrix
 
2 1 + 2t
Ω(t) = .
1 t

Use this fundamental matrix and variation of parameters to find the gen-
eral solution of the nonhomogeneous system

2c1 + c2 (1 + 2t) + t + t2 − 2t3

 
X(t) = .
c1 + c2 t + 2t2 − t3

3. A has eigenvalues 6, 6 and eigenvectors are scalar multiples of

 
1
.
1

We find the fundamental matrix

 
1 1+t
Ω(t) =
1 t

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 354 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

for the associated homogeneous system. Use this and variation of param-
eters to find the general solution of the nonhomogeneous system:
c + c2 (1 + t) + 2t + t2 − t3
 
X(t) = e6t 1 .
c1 + c2 t + 4t2 − t3

4. A has eigenvalues 2, 2, 2, and there are two linearly independent eigenvec-

tors    
1 0
0 , 1 .
0 1
We find the fundamental matrix
 
1 0 0
Ω(t) = e2t 0 1 −4t 
0 1 1 − 4t
of the associated homogeneous system. A general solution for the nonho-
mogeneous system is
c1 e2t
 

X(t) =  (c2 − 4c3 t)e2t + 1  .

(c2 + c3 (1 − 4t))e2t + 1

5. A has eigenvalues 1, 1, 3, 3. The eigenvalue 3 has two independent eigen-

vectors    
0 0
1 −9
 , ,
0  2 
1 0
and 1 has the eigenvector  
0
0
 ,
0
1
with all other eigenvectors associated with 1 scalar multiples of this one.
A fundamental matrix for the homogeneous system X0 = AX is
et
 
0 0 0
 0 −2et et −9e3t 
Ω(t) =  .
0 0 0 2e3t 
et −5tet e3t 0
The nonhomogeneous system has general solution
c2 et
 
 −2c2 et + (c3 − 9c4 )e3t + et 
X(t) =  .
 2c4 e3t 
t 3t t
(c1 − 5c2 t)e + c3 e + (1 + 3t)e

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.5. SOLUTION BY DIAGONALIZATION 355

For Problems 6–9, the idea is to find a general solution for the system, then
solve for the constants to obtain the solution satisfying the initial condition. For
these problems only the solution of the initial value problem is given.

6.
−1 + e2t
 
X(t) =
−5t + (3 + 5t)e2t

7.
(−1 − 14t)et
 
X(t) =
(3 − 14t)et

8.
13t − (8 + 12t + 3t2 )e2t
 

X(t) =  4et + (7 + 2t)e2t 

−et − e2t

9.
(6 + 12t + (1/)t2 )e−2t
 

X(t) =  (2 + 12t + (1/2)t2 )e−2t 

(3 + 38t + 66t2 + (13/6)t3 )e−2t

13.5 Solution by Diagonalization

1. The coefficient matrix A is diagonalized by
 
1 1
P=
1 4

and we find that  

1 4 −1
P−1 = .
3 −1 1
The system for Z is
    
−1 0 1 4 −1 0
Z0 = Z+ .
0 2 3 −1 1 10 cos(t)

This is the system

 0    
0 z1 −z1 1 −10 cos(t)
Z (t) = = + .
z20 2z2 3 10 cos(t)
Solve these two independent differential equations to get
 −t 
c1 e − (5/3) cos(t) − (5/3) sin(t)
Z(t) = .
c2 e2t − (4/3) cos(t) + (2/3) sin(t)

Then
c1 e−t + c2 e2t − 3 cos(t) − sin(t)
 
X(t) = PZ(t) = .
c1 e−t + 4c2 e2t − 7 cos(t) + sin(t)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 356 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

2. The coefficient matrix has eigenvalues 2, 6 and is diagonalized by

 
3 1
P= .
−1 1

We find that  
1 1 −1
P−1 = .
4 1 3
With X = PZ, the transformed system has the solution

c1 e2t − 1 − e3t
 
Z(t) = .
c2 e6t − 1/3 − et

The original system has the general solution

3c1 e2t + c2 e6t − 4e3t − 10/3

 
X(t) = .
−c1 e2t + c2 e6t + 2/3

3. The coefficient matrix has eigenvalues 0, 2 and is diagonalized by

 
1 1
P=
−1 1

which has inverse  

−1 1 1 −1
P = .
2 1 1
The system for Z(t) has the solution

c1 − 2t + e3t
 
Z(t) = .
c2 e2t − 1 + 3e3t

Then
c1 + 2c2 e2t − 1 − 2t + 4e3t
 
X(t) = .
−c1 + c2 e2t − 1 + 2t + 2e3t

4. A has eigenvalues 1, 7, and we find

   
1 5 −1 1 1 −5
P= ,P = .
−1 1 6 1 1

We obtain
c1 et + (1/15) cos(3t) − (3/15) sin(3t) + 20/3
 
Z(t) = .
c2 e7t + (7/87) cos(3t) − (2/58) sin(3t) − 4/21

Then
c1 et + 5c2 e7t + (68/145) cos(3t) − (54/145) sin(3t) + 40/7
 
X(t) = .
−c1 et + c2 e7t + (2/145) cos(3t) + (24/145) sin(3t) − 48/7

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.5. SOLUTION BY DIAGONALIZATION 357

5. A has eigenvalues 3i, −3i and is diagonalized by

 
1+i 1−i
P= .
3 3

We find that  
1 −3i 1 + i
P−1 = .
6 3i 1 − i
The transformed problem for Z has the solution

d1 e3it + ((2 − i)/6)e2t

 
Z(t) = .
d2 e−3it + ((2 + i)/6)e2t

If Euler’s formula is used on the complex exponential terms, we obtain

the real solution
c1 (cos(3t) − sin(3t)) − c2 (cos(3t) + sin(3t)) + e2t
 
X(t) = .
3c1 cos(3t)03c2 sin(3t) + 2e2t

6. The coefficient matrix is  

1 1
A=
1 1
with eigenvalues 0, 2. A is diagonalized by
 
1 1
P=
−1 1

and we find that  

1/2 −1/2
P−1 = .
1/2 1/2
The uncoupled system is
     2t 
0 0 0 1/2 −1/2 6e
Z = Z+ .
0 2 1/2 1/2 2e2t

The initial conditions convert to

 
−1 3
Z(0) = P X(0) = .
3

Solve for Z to obtain

2 + e2t
 
Z(t) = .
3e2t + 4te2t

The solution of the initial value problem for X is

2 + 4(1 + t)e2t
 
X(t) = PZ(t) = .
−2 + 2(1 + t)e2t

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 358 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

7. The coefficient matrix has eigenvalues 0, 3 and is diagonalized by

 
2 −1
P= .
1 1

Compute  
−1 1 1 1
P = .
3 −1 2
The uncoupled system is
    
0 0 0 1/2 1/3 2t
Z = Z+ .
0 3 −1/3 2/3 5

The initial condition is

 
−1 25/3
Z(0) = P X(0) = .
11/3

Then
(1/3)t2 + (5/3)t + 25/3
 
Z(t) = .
(127/27)e3t + (2/9)t − 28/27
The solution of the initial value problem for X is

−(127/27)e3t + (2/3)t2 + (28/9)t + 478/27

 
X(t) = PZ(t) = .
(127/27)e3t + (1/3)t2 + (17/9)t + 197/27

8. The coefficient matrix has eigenvalues i, −i and is diagonalized by

 
5 5
P= .
2−i 2+i

We find that  
−1 1 1 − 2i 5i
P = .
10 1 + 2i −5i
With X = PZ, the uncoupled system is
    
i 0 (1 − 2i)/10 i/2 5 sin(t)
Z0 = Z+ ,
0 −i (1 + 2i)/10 −i/2 0

with initial conditions

 
1 + i/2
Z(0) = P−1 X(0) = .
1 − i/2

Solve these uncoupled equations for z1 and z2 and then obtain

 
10 cos(t) − (5/2)t sin(t) − 5t cos(t)
X(t) = PZ(t) = .
5 cos(t) + (5/2) sin(t) − (5/2)t sin(t)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 13.5. SOLUTION BY DIAGONALIZATION 359

9. The coefficient matrix has eigenvalues 1, 1, −3, but there are two inde-
pendent eigenvectors associated with the repeated eigenvalue 1 and A is
diagonalized by  
1 −1 1
P = 1 0 3 .
0 1 1
We find that  
3 −2 3
P−1 =  1 −1 2  .
−1 1 −1
With X = PZ we obtain the uncoupled system

−3e−3t
    
1 0 0 3 =2 3
Z0 = 0 1 0  Z +  1 −1 2  t .
0 0 −3 −1 1 −1 0

The initial conditions are

 
11
Z(0) = P−1 X(0) =  6  .
−4

Solve this uncoupled system to obtain

(5/2)et − (8/3)e−3t + 3te−3t + (8/9) + (4/3)t

 

X(t) = PZ(t) =  (27/4)et − (113/12)e−3t + 9te−3t + (5/3) + 3t  .

(17/4)et − (113/36)e−3t + 3te−3t + (8/9) + (4/3)t

10. The coefficient matrix has eigenvalues 1, 2, 2 and is diagonalized by

 
1 1 1
P = 1 1 0
1 0 1

and we find that  

−1 1 1
P−1 = 1 0 −1 .
1 −1 0
The uncoupled system is
    
1 0 0 −1 1 1 −
Z0 =  0 2 0 Z +  1 0 −1  t  .
0 0 2 1 −1 0 2et

The initial conditions are

Z(0) = P−1 X(0).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 360 CHAPTER 13. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

Solve this initial value problem for Z(t) and then obtain

−(1/4)e2t + (2 + 2t)et − (3/4) − (1/2)t

 

X(t) = PZ(t) =  e2t + (2 + 2t)et − 1 − t .

(5/4)e2t + 2tet − (3/4) − (1/2)t

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 14

Nonlinear Systems and

Qualitative Analysis

14.1 Nonlinear Systems and Phase Portraits

1. The coefficient matrix is
 
3 −5
A= ,
5 −7

with eigenvalues −2, −2. Every eigenvector is a scalar multiple of

 
1
.
1

The origin is an improper nodal sink.

2. The coefficient matrix has eigenvalues −3, 4 and the origin is a saddle
point.

For Problems 3 − −16, only the eigenvalues of the coefficient matrix, and
the classification of the origin, are given. As typical cases, phase portraits are
drawn for the systems of Problems 3, 5, 6, 7 and 11.
Phase portraits are included for the systems of Problems 3, 4, 5 and 11.

3. eigenvalues 2i, −2i; center

4. 2, 3, nodal source
5. 4 + 5i, 4 − 5i, spiral point
6. −3, −5, nodal sink
7. 3, 3, and the coefficient matrix does not have two independent eigenvectors;
improper node

361

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 362CHAPTER 14. NONLINEAR SYSTEMS AND QUALITATIVE ANALYSIS

Figure 14.1: Center of Problem 3.

Figure 14.2: Nodal source of Problem 4.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 14.1. NONLINEAR SYSTEMS AND PHASE PORTRAITS 363

Figure 14.3: Spiral source of Problem 5.

√ √
8. 31i, − 31i, center
√ √
9. eigenvalues −2 + 3i, −2 − 3i, spiral sink

10. −13, −13, and the coefficient matrix does not have two independent eigen-
vectors; improper node
√ √
11. 5, − 5, saddle point
√ √
12. 3 + 5i, 3 − 5i, spiral source
√ √
13. −3 + 7, −3 − 7, both eigenvalues negative, nodal sink

14. 11, 11, and the coefficient matrix does not have two independent eigenvec-
tors, improper node
√ √
15. 2 + 3, 2 − 3, nodal source
√ √
16. 13i, − 13i, center

17. (a) First, write

dx 1
= x
dt t
as
1 1
dx = dt.
x t
Integrate to get
ln(x) = ln(t) + c

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 364CHAPTER 14. NONLINEAR SYSTEMS AND QUALITATIVE ANALYSIS

Figure 14.4: Saddle point of Problem 11.

for x > 0, t > 0. Then x = ct with c constant. Put this into the second
equation to get
1
y 0 = ct − y.
t
Write this as
1
y 0 + y = ct,
t
or
ty 0 + y = ct2 .
Then
(ty)0 = ct2 .
Integrate to get
c 3
ty = t + d.
3
Then
c 2 d
y= t + .
3 t
(b) Suppose x(t0 ) = 1 and y(t0 ) = 0. Then it is routine to solve for c and
d from part (a) to obtain
1 1 2 1
x(t) = t, y(t) = t − t20 .
t0 3t0 3t

(c) In part (b), we have trajectories through (1, 0) at any time t0 6= 0.

However, these translations are not trajectories of each other.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 14.2. CRITICAL POINTS AND STABILITY 365

14.2 Critical Points and Stability

In Problems 1–16, the stability type of the origin is given, based on information
in Problems 1–16 of Section 14.1.

1. The origin is an improper node that is both stable and asymptotically

stable

2. The origin is an unstable saddle point

3. stable but not asymptotically stable center

4. unstable nodal source

5. unstable spiral source

6. stable and asymptotically stable nodal sink

7. unstable improper node

8. stable but not asymptotically stable center

9. stable and asymptotically stable spiral sink

10. stable and asymptotically stable improper node

11. unstable saddle point

12. unstable spiral point

13. stable and asymptotically stable nodal sink

14. unstable improper node

15. unstable nodal source

16. stable but not asymptotically stable center

√ √
17. If  = 0, the eigenvalues are 5i, − 5i and the origin is a center, which
is stable but not asymptotically stable.
If  > 0, the eigenvalues are

1 1p 1 p
+ ( − 2)2 − 24,  ( − 2)2 − 24.
2 2 2
√
These have positive real part. If 0 <√ < 2(1 + 6), then the origin is an
unstable spiral point.
√ If  > 2(1 + 6), the origin is an unstable saddle
point. If  = 2(1 + 6), the origin is an unstable improper node.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 366CHAPTER 14. NONLINEAR SYSTEMS AND QUALITATIVE ANALYSIS

18. If  = 0, the eigenvalues are −3, −3 and every eigenvector is a scalar

multiple of  
1
.
−1
In this case the origin is a stable and asymptotically stable improper node.
If  > 0, the eigenvalues are
 − 6 1p  − 6 1p
+ ( + 10)2 − 100, − ( + 10)2 − 100.
2 2 2 2
These eigenvalues are real and distinct. Further, if 0 <  < 9/8, these
eigenvalues are both negative, so the origin is a stable and asymptotically
stable nodal sink. If  > 9/8, then the eigenvalues are of opposite sign
and the origin is an unstable saddle point. If  = 9/8, then the origin is
not an isolated singularity.

14.3 Almost Linear Systems

In Problem 1, the details of showing that the system is almost linear are in-
cluded. Problems 2–10 omit this demonstration and concentrate on analyzing
the critical point (0, 0).

1. To show that the system is almost linear, consider

x2 r2 cos2 (θ)
lim p = lim
(x,y)→(0,0) x2 + y 2 r→0 r
= lim r cos2 (θ) = 0.
r→0

The origin is a critical point and the matrix of coefficients of the linear
part is  
1 −1
,
1 2
which has eigenvalues
1 √ 1 √
(3 + 3i) and (3 − 3)i.
2 2
The origin is an unstable spiral point of the linear part, hence also of the
given system.

2. The matrix of coefficients of the linear part is

 
1 −1
1 1

with eigenvalues 1 + i, 1 − i. The origin is an unstable spiral source.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 14.3. ALMOST LINEAR SYSTEMS 367

3. The linear part has matrix

 
−2 2
A(0,0) =
1 4
√ √
with eigenvalues 1+ 11, 1− 11. These are of opposite sign, so the origin
is an unstable saddle point.
4. The matrix of the linear part is
 
−2 −3
1 4
√
with eigenvalues 1 ± 6, which are of opposite sign. The origin is an
unstable saddle point.
5. The linear part has matrix
 
3 12
,
−1 −3
√ √
with eigenvalues 3i, − 3i. The origin is a center for the linear part of
the system, so the nonlinear system could have a center or a spiral point
there.
6. The linear part has matrix
 
2 −4
,
1 1
√
with eigenvalues (3 ± 15i)/2. The origin is a spiral point and is unstable
because the real part of the eigenvalues is positive.
7. The linear part has matrix
 
−3 −4
.
1 1
This matrix has eigenvalues −1, −1, and all eigenvectors are scalar multi-
ples of  
−2
.
1
The origin is a stable improper nodal sink.
8. The linear part has matrix
 
−3 −4
−1 1
√
with eigenvalues −1 ± 2 2. Both eigenvalues are positive, so the origin is
an unstable nodal source.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 368CHAPTER 14. NONLINEAR SYSTEMS AND QUALITATIVE ANALYSIS

9. The linear part has matrix

 
−2 −1
−4 1

with eigenvalues 2, −3, so the origin is an unstable saddle point.

10. The linear part has matrix  

1 0
−2 1
with eigenvalues 1, 1. All eigenvectors associated with this eigenvalue are
scalar multiples of  
0
,
1
so the origin is an unstable (positive eigenvalue) improper node.

11. Refer to these as systems I and II, in the order given.

(a) For each system the linear part has coefficient matrix
 
0 1
A=
−1 0

with eigenvalues i, −i. Therefore the origin is a center for each system.
(b) For The first system, use polar coordinates, with

x = r cos(θ), y = r sin(θ)

and p
x2 + y 2 = r.
Then
p
−x x2 + y 2 −r2 cos(θ)
lim p = lim
(x,y)→(0,0) x2 + y 2 r→0 r
= lim −r cos(θ) = 0,
r→0

independent of θ. And, similarly

p
−y x2 + y 2
lim p = lim −r sin(θ) = 0.
(x,y)→(0,0) x2 + y 2 r→0

Therefore the first system is almost linear. The argument is essentially

the same for the second system.
(c) With r2 = x2 + y 2 , we have

rr0 = xx0 + yy 0 ,

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 14.3. ALMOST LINEAR SYSTEMS 369

where primes denote differentiation with respect to t. Now insert the

expressions for x0 and y 0 from the differential equations of system I to
obtain
p p
rr0 = x(y − x x2 + y 2 ) + y(−x − y x2 + y 2 )
p
= −(x2 + y 2 ) x2 + y 2
= −r3 .

Then
dr
r0 =
= −r2 for system I.
dt
Similarly, if we insert the expressions for x0 and y 0 from system II, we
obtain
p p
rr0 = x(y + x x2 + y 2 ) + y(−x + y x2 + y 2 )
p
= (x2 + y 2 ) x2 + y 2
= r3 .

Then
dr
= r2 for system II.
dt
(d) For system I,
dr
= −r2 .
dt
This is the separable equation
1
− dr = dt.
r2
This shows that, for system I, r0 (t) < 0, so the distance between the point
and the origin is decreasing with time.
Now integrate the differential equation for r(t) to get
1
= t + c.
r
To satisfy the initial condition r(t0 ) = r0 , we need
1
= t0 + c,
r0
so
1
c= − t0 .
r0
Then
1
r(t) =
t − t0 + 1/r0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 370CHAPTER 14. NONLINEAR SYSTEMS AND QUALITATIVE ANALYSIS

for system I. Then

r(t) → 0 as t → ∞.
The first system is asymptotically stable at the origin.
(e) By an entirely analogous derivation, for system II,

r0 = r2 > 0

so r(t) is increasing with time for system II. Solve this separable differential
equation subject to r(t0 ) = r0 to get

1
r(t) = .
1/r0 + t0 − t

Then for system II, r(t) → ∞ as t → t0 + 1/r0 from the left. We conclude
from this that the second system is unstable at the origin.
Parts (d) and (e) show that, in the case of a center at the origin, behavior
of the the linear part of an almost linear system at the origin does not
provide definitive information about the stability of the origin for the
nonlinear system.

12. The linear part of the system has coefficient matrix

 
α −1
A= .
1 −α
√ √
The eigenvalues are α2 − 1 and − α2 − 1. There are the following cases:
1. If 0 < |α < 1, the eigenvalues are pure imaginary. In this case the
origin is a center of the linear part and may be a center or spiral point of
the almost linear system.
2. If |α| > 1, then the eigenvalues are real and of opposite sign, so the
origin is an unstable saddle point of the linear part, hence also of the
nonlinear system.
3. If α = ±1, then A is a singular matrix and the theory developed for
almost linear systems does not apply.
The constants h and k will influence quantitative aspects of the phase
portrait of the system, but not the qualitative behavior.

13. Using results from Problem 11, we have

rr0 = xx0 + yy 0
= x(y + x(x2 + y 2 )) + y(−x + y(x2 + y 2 ))
= (x2 + y 2 )(x2 + y 2 )
= r4 .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 14.4. LINEARIZATION 371

Then
dr
= r3 .
dt
This is separable
1
dr =  dt.
r3
Integrate to get
1
− r−2 = t + c.
2
Then
1
r(t) = √ ,
k − 2t
where k = 2c is an arbitrary constant which is determined by specifying
a point that the trajectory is to pass through at some positive time.
If  < 0, then
1
r(t) = p →0
k + 2||t
as t → ∞. In this case trajectories approach the origin as t → ∞ and the
nonlinear system is asymptotically stable.
However, something different happens if  > 0. Suppose r(0) = ρ, so a
trajectory starts at a point ρ units from the origin at time zero. Then
k = 1/ρ2 and
1
r(t) = p .
(1/ρ)2 − 2t

Now, as t starts at zero and increases toward 1/(2ρ2 ), r(t) → ∞, so there

is a time close to which the point is arbitrarily far from the origin. This
makes the origin unstable.

14.4 Linearization
1. For critical points other than the origin, solve

x − y + x2 = 0, x + 2y = 0

to get (−3/2, 3/4). We find that

 
−2 −1
A(−3/2,3/4) = ,
1 2
√ √
which has eigenvalues 3 and − 3. This critical point is an unstable
saddle point of the linear part, and therefore of the given system.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 372CHAPTER 14. NONLINEAR SYSTEMS AND QUALITATIVE ANALYSIS

2. The critical points are

√ ! √ !
−1 + 17 −1 − 17
,4 , ,4 .
2 2

We have √ 
17 −1
√
A((−1+√17)/2,4) = ,
−7 2 − 17
with eigenvalues

√ √
q q
1 − 25 − 2 17, 1 + 25 − 2 17.

These are of opposite sign, being approximately equal to −3.093 · · · and

5.093 · · · . This critical point is an unstable saddle point.
Next,
 √ 
− 17 −1
√
A((−1−√17)/2,4)= ,
−7 2 + 17
p √ p √
with eigenvalues 1 − 25 + 2 17 and 1 + 25 + 2 17. These are also of
opposite sign, so this critical point is also an unstable saddle point.

3. The critical point other than the origin is (−5, −5). We find that
 
−2 2
A(−5,−5) = .
1 −6
√ √
This has eigenvalues −4 + 6, −4 − 6, which are unequal and both
negative. This critical point is a stable and asymptotically stable nodal
sink of the nonlinear system.

4. The only critical point other than the origin is (−20, 5). We find that
 
−2 −13
A(−20,5) = ,
1 4

having eigenvalues 1 + 2i, 1 − 2i. These are complex with positive real
part, so (−20, 5) is an unstable spiral source.

5. The system has one critical point other than the origin, (−1/2, 1/8). Cal-
culate  
3 12
A(−1/2,1/8) = ,
−1/4 −3
√ √
with eigenvalues are 6, − 6, so this critical point is an unstable saddle
point.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 14.4. LINEARIZATION 373

6. The system has two critical points other than the origin, (1, −2) and
(−2/3, 2/9). First,  
−4 −1
A(1,−2) =
3 1
√
and the eigenvalues are (−3± 13)/2. These have opposite sign, so (1, −2)
is an unstable saddle point.
 
8/3 −6
A(−2/3,2/9) = ,
−1/2 1
√
with eigenvalues (11 ± 97)/2. These are both positive, so (−2/3, 2/9) is
an unstable nodal source.

7. Aside from the origin, the system has critical points (1/2, −1/2). We have
 
−2 −3
A(1/2,−1/2) =
1 1
√
with eigenvalues (−1 ± 3i)/2, so (1/2, −1/2) is a spiral point, stable and
asymptotically stable because the real part of the eigenvalues is negative.

8. The critical points other than the origin are

r √ ! r √ !
7 21 7 21
,− and − , .
3 4 3 4

First,  
−3 −4
A(√7/3,−√21/4) =
−5/2 −4/3
√
with eigenvalues (−13 ± 385)/6, both positive, so this critical point is
an unstable nodal source.
Next,  
−3 −4
A(−√7/3,√21/4) = .
9/2 −4/3
√
This has eigenvalues (−13 ± 623i)/6. This critical point is a stable and
asymptotically stable spiral point.

9. The critical point other than the origin is (−3/8, −3/2). We find that
 
−2 −2
A(−3/8,−3/2) =
−4 1
√
with eigenvalues (−1 ± 23i)/2, so (−3/8, −3/2) is a stable and asymp-
totically stable spiral point.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 374CHAPTER 14. NONLINEAR SYSTEMS AND QUALITATIVE ANALYSIS

10. The system has two critical points other than the origin:
√ ! √ !
√ 2+ 2 √ 2− 2
3 + 2 2, , 3 − 2 2, .
2 2

First,  √ 
1√ −4 − 2√ 2
A(3+2√2,(2+√2)/2) =
(−1 + 2 2)/2 11 + 7 2
with eigenvalues
√ √
q q
7 1 7 1
6+ √ − 128 2 + 90, 6 + √ + 128 2 + 190.
2 2 2 2
These are approximately 1.3188 · · · and 20.5806 · · · , both positive, so this
critical point is an unstable proper nodal source,
 √ 
1√ −4 + 2√ 2
A(3−2√2,(2−√2)/2) =
(−1 − 2 2)/2 11 − 7 2
with eigenvalues
√ √
q q
7 1 7 1
7− √ − 190 − 128 2, 6 − √ + 190 − 128 2.
2 2 2 2
These are approximately −0.4481 · · · and 2.5486 · · · , having opposite signs,
so this critical point is an unstable saddle point.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 15

Vector Differential Calculus

15.1 Vector Functions of One Variable

In Problems 1 and 2 the details of the differentiation are carried out both ways.
For Problems 3–8 just the derivative is given.

1. First use the product rule:

(f (t)F(t))0 = f 0 (t)F(t) + f (t)F0 (t)

= (−12 sin(3t))F(t) + 4 cos(3t)(6tj + 2k)
= −12 sin(3t)i + (24t cos(3t) − 36t2 sin(3t))j
+ (8 cos(3t) − 24t sin(3t))k.

If we first carry out the product, we have

f (t)F(t) = 4 cos(3t)i + 12t2 cos(3t)j + 8t cos(3t)k,

so

(f (t)F(t))0 = −12 sin(3t)i

+ (24t cos(3t) − 36t2 sin(3t))j + (8 cos(3t) − 24t sin(3t))k.

2.

(F(t) · G(t))
= (i − 6tk) · (i + cos(t)k) + (ti − 3t2 k) · (− sin(t)k)
= 1 − 6t cos(t) + 3t2 sin(t)

375

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 376 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

3. Apply the product rule for cross products:

i j k i j k
(F(t) × G(t))0 = 1 0 0 + t 1 4
1 − cos(t) t 0 sin(t) 1
= −tj − cos(t)k + (1 − 4 sin(t))i − tj + t sin(t)k
= (1 − 4 sin(t))i − 2tj − (cos(t) − t sin(t))k.

4.

(F(t) × G(t))0 = (3t2 − 2t sinh(t) − t2 cosh(t))i

− 2tj − (sinh(t) + t cosh(t))k

5.

(f (t)F(t))0 = (1 − 8t2 )i
+ (6t2 cosh(t) − (1 − 2t3 ) sinh(t))j + (et − 6t2 et − 2t3 et )k

6.
(F(t) · G(t))0 = sin(t) + t cos(t) + 4 + 5t4

7.
(F(t) × G(t))0 = tet (2 + t)(j − k)

8.
(F(t) · G(t))0 = −16 cos2 (t) + 16 sin2 (t)

9.
F(t) = sin(t)i + cos(t)j + 45tk
is a position vector for the curve, and

F0 (t) = cos(t)i − sin(t)j + 45k

is a tangent vector. The distance function along C is

Z t Z t√ √
s(t) = k F0 (τ ) k dτ = 2026 dτ = 2026t.
0 0
√
Then t = s/ 2026 and we can write a position vector in terms of s:
   
s s 45s
G(s) = F(t(s)) = sin √ i + cos √ j+ √ k.
2026 2026 2026
Now we can write a tangent vector in terms of s:
     
1 s s
G0 (s) = √ cos √ i − sin √ j + 45k .
2026 2026 2026

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 15.1. VECTOR FUNCTIONS OF ONE VARIABLE 377

10. F(t) = t3 (i + j + k) is a position vector, and a tangent vector is

F0 (t) = 3t2 (i + j + k).

A distance function along C is given by

Z t Z t √
s(t) = k F0 (ξ) k dξ = 3 3(t3 + 1).
−1 −1

Then  1/3
s
t= √ −1 .
3
Let  
s
G(s) = F(t(s)) = √ − 1 (i + j + k).
3
This gives us the unit tangent vector
1
G0 (s) = √ (i + j + k).
3

11. F = t2 (2i + 3j + 4k) is a position vector for the curve, and

F0 (t) = 2t(2i + 3j + 4k)

is a tangent vector. The distance function along the curve is given by

Z t
0
√ Z t √
s(t) = k F (ξ) k dξ = 2 29 ξ dξ = 29(t2 − 1).
1 1

Then q √
t= 1 + s/ 29.
Let  
s
G(s) = F(t(s)) = √ + 1 (2i + 3j + 4k).
29
Then
1
G0 (s) = √ (2i + 3j + k)
29
is a unit tangent vector.
12. Because F(t) × F0 (t) = O, we must have

k F(t) kk F0 (t) k sin(θ) = 0

where θ is the angle between these two vectors. Then at least one of F(t)
and F0 (t) is the zero vector, or the angle between these vectors is zero (so
the vectors are parallel). Consider cases.
If F(t) = O then the particle simply remains at the origin for all time.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 378 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

If F0 (t) = O, then F(t) is a constant vector and the particle does not
move.
In the case that F(t) and F0 (t) are parallel, then the position and tangent
vectors are parallel, so the velocity vector is directed along the path of
motion and the motion is in a straight line.
We could also argue as follows in the last case. If the position and tangent
vectors are parallel, then for some number c,

F0 (t) = cF(t).

Then
x0 (t)i + y 0 (t)j + z 0 (t)k = c(x(t)i + y(t)j + z(t)k).
Then
x0 (t) = cx(t), y 0 (t) = cy(t), z 0 (t) = cz(t).
Then
x = x0 ect , y = y0 ect , z = z0 ect ,
where F(0) = x0 i + y0 j + z0 k. But these are parametric equations of a
straight line.

15.2 Velocity, Acceleration and Curvature

In Problems 1–10, we can compute

v = F0 (t), a(t) = F00 (t), v(t) =k v(t) k

by straightforward calculations. Next,

1 1
T(t) = v(t) = 0
F0 (t).
v(t) k F (t) k
Tangential and normal components of the acceleration can be obtained as
dv
q
aT = and aN = k a k2 −a2T .
dt
The unit normal is
1
N(t) = (a(t) − aT T(t)).
aN
In this way it is not necessary to compute s(t) and write vectors in terms of s,
which is often awkward. We can also compute
1 dT
N(t) = .
k dT/dt k dt
Curvature is often easily computed as
aN
κ= .
v2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 15.2. VELOCITY, ACCELERATION AND CURVATURE 379

We can also compute curvature as

k T0 (t) k
κ= .
k F0 (t) k

κ can also be obtained from a formula requested in Problem 13:

k F0 (t) × F00 (t) k

κ= .
k F0 (t) k3

1. The velocity is
v(t) = F0 (t) = 3i + 2tk
and the speed is p
v(t) =k v(t) k= 9 + 4t2 .
The acceleration is
a(t) = F00 (t) = 2k.
A unit tangent is
1
T(t) = √ (3i + 2tk).
9 + 4t2
The curvature is
k T0 (t) k 6
κ= = .
k F0 (t) k (9 + 4t2 )3/2
Finally,
dv 4t
aT = =√
dt 9 + 4t2
and
6
q
aN = k a k2 −a2T = √ .
9 + 4t2

2.
v(t) = (sin(t) + t cos(t))i + (cos(t) − t sin(t))j,

a(t) = (2 cos(t) − t sin(t))i − (2 sin(t) + t cos(t))j,

1
T(t) = √ v,
1 + t2
p
v(t) = 1 + t2 ,

t 2 + t2
aT = √ , aN = ,
1 + t2 1 + t2

2 + t2
κ=
(+t2 )3/2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 380 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

3.
v(t) = 2i − 2j + k, v = 3,
1
T= (2i − 2j + k),
3
aT = aN = κ = 0

4. √ t
v(t) = et (sin(t) + cos(t))i + et (cos(t) − sin(t))k, v(t) = 2e ,
a(t) = 2et (cos(t)i − sin(t)k),
1
T(t) = √ ((sin(t) + cos(t))i + (cos(t) − sin(t))k),
2
√ 1
aT = 2et = aN , κ = √ e−t
2
5.
v(t) = −3e−t (i + j − 2k), a(t) = 3e−t (i + j − 2k),
√ 1
v(t) = 3 6e−t , T(t) = √ (−i − j + 2k),
6
√ −t
aT = −3 6e , aN = 0, κ = 0

6. p
v(t) = −α sin(t)i + βj + α cos(t)k, v(t) = α2 + β 2 ,
a(t) = −α cos(t)i − α sin(t)k,
1
T(t) = p (−α sin(t)i + βj + α cos(t)k),
α2 + β2
α
aT = 0, aN = α, κ =
α2 + β2
7. p
v(t) = 2 cosh(t)j − 2 sinh(t)k, v(t) = 2 cosh(2t),
a(t) = 2 sinh(t)j − 2 cosh(t)k,
1
T(t) = p (cosh(t)j − sinh(t)k),
cosh(t)
2 sinh(2t) 2
a(t) = p , aN = p ,
cosh(2t) cosh(2t)
1
κ=
2(cosh(2t))3/2
Here we have used the identity

cosh(2t) = cosh2 (t) + sinh2 (t).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 15.2. VELOCITY, ACCELERATION AND CURVATURE 381

8.
1 1
v(t) = (i − j + 2k), a(t) = − 2 (i − j + 2k),
t t
√
6 1
v(t) = , T(t) = √ (i − j + 2k),
t 6
√
6
aT = − 2 , aN = 0, κ = 0
t
9.
v(t) = 2t(αi + βj + γk),
a(t) = 2(αi + βj + γk)
p
v(t) = 2|t| α2 + β 2 + γ 2 ,
1
T(t) = p (αi + βj + γk)
α2 + β2 + γ2
aN = 0, κ = 0,
and p
aN = 2(sgn(t)) α2 + β 2 + γ 2 ,
where (
1 if t > 0,
sgn(t) =
−1 if t < 0.

10.
v(t) = (3 cos(t) − 3t sin(t))j − (3 sin(t) + 3t cos(t))k,
p
v(t) = 3 1 + t2
a(t) = (−6 sin(t) − 3t cos(t))j − (6 cos(t) − 3t sin(t))k,
1
T(t) = √ ((cos(t) − t sin(t))j − (sin(t) + t cos(t))k,
1 + t2
3t (3t2 + 6)2
aT = √ , aN = √ ,
1 + t2 1 + t2
(3t2 + 6)2
κ=
9(1 + t2 )3/2

11. The position vector for a straight line has the form

F(t) = (a + bt)i + (c + dt)j + (p + ht)k.

The tangent vector is the constant vector

T(t) = bi + dj + hk.

Then T0 (t) = O, so κ = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 382 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

Conversely, suppose C is a smooth curve having zero curvature. Then

κ =k T0 (s) k=k F00 (s) k= 0.

If we write
F(s) = f (s)i + g(s)j + h(s)k,
this means that
f 00 (s) = g 00 (s) = h00 (s) = 0.
But then f (s) = a + bs, g(s) = c + ds, h(s) = p + hs for some constants
a, b, c, d, p, h. This makes F(s) the position vector of a straight line.
12. We may suppose that C is a circle of radius r about the origin in the
x, y−plane (translations and rotations will not affect the curvature). A
position vector for C in polar coordinates has the form

F(t) = r cos(t)i + r sin(t)j

with 0 ≤ t ≤ 2π. Then

F0 (t) == r sin(t)i + r cos(t)j

is a tangent vector, which has length r. A unit tangent is

T = − sin(t)i + cos(t)j.

Then
T0 (t) = − cos(t)i − sin(t)j.
Then
k T0 (t) k 1
κ= = .
k F0 (t) k r
13. First write
1 1 0
T(t) = f 0 (t) = F (t).
k F0 (t) k v(t)
This enables us to write
F0 = vT.
Now, F00 (t) is the acceleration a(t), and T × T = O, so

vT × F00 = vT(aT T + aN N)
= vaT T × T + vaN T × N
= vaN T × N
= v(v 2 κ)T × N.

But T and N are orthogonal unit vectors, so

k T × N k= 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 15.3. THE GRADIENT FIELD 383

Then
k F0 × F00 k= v 3 κ.
Finally,
v =k F0 k
so
3
k F0 (t) × F00 (t) k
κ= .
k F0 (t) k

15.3 The Gradient Field

1.
∂ ∂ ∂
∇ϕ(x, y, z) = (xyz)i + (xyz)j + (xyz)k = yzi + xzj + xyk,
∂x ∂y ∂z

∇ϕ(1, 1, 1) = i + j + k
The maximum value of Du (1, 1, 1) is
√
k ∇ϕ(1, 1, 1) k= 3.
√
The minimum value is − 3.

2.
∇ϕ(x, y, z) = (2xy − z cos(xz))i + x2 j − x cos(xz)k,
√ ! √
2π 2
∇ϕ(1, −1, π/4) = −2 − i+j− k.
8 2

The maximum value of Du (1, −1, π/4) is

q √
11/2 + π 2 /32 + π/ 2.

The minimum value is the negative of this maximum value.

3.
∇ϕ(x, y, z) = (2y + ez )i + 2xj + xez k,

∇ϕ(−2, 1, 6) = (2 + e6 )i − 4j − 2e6 k.
The maximum value of Du (−2, 1, 6) is
p
20 + 4e6 + 5e12

and the minimum value is the negative of this.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 384 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

4.
∇ϕ(x, y, z) = −yz sin(xyz)i − xz sin(xyz)j − xy sin(xyz)k,
π π
∇ϕ(−1, 1, π/2) = i − j − k.
2 2
The maximum value of Du (−1, 1, π/2) is
r
π2
1+ .
4
The minimum value is the negative of this radical.

5.
∇ϕ(x, y, z) = 2y sinh(2xy)i + 2x sinh(2xy)j − cosh(z)k,

∇ϕ(0, 1, 1) = − cosh(1)k.
The maximum value of Du (0, 1, 1) is cosh(1). The minimum value is
− cosh(1).

6.
1
∇ϕ(x, y, z) = p [xi + yj + zk]
x2 + y 2 + z 2

1
∇ϕ(2, 2, 2) = √ (i + j + k).
3
maxDu =k ∇ϕ(2, 2, 2) k= 1,

and the minimum is −1.

7.

Du ϕ(x, y, z) = ∇ϕ(x, y, z) · u
1
= ((8y 2 − z)i + 16xyj − xk) · √ (i + j + k)
3
1 2
= √ (8y − z + 16xy − x)
3

8.

Du ϕ(x, y, z)
1
= (− sin(x − y)i + sin(x − y)jez k) · √ (i − j + 2k)
6
1 z
= √ (−2 sin(x − y) + 2e )
6

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 15.3. THE GRADIENT FIELD 385

9.
Du (x, y, z)
1
= (2xyz 3 i + x2 z 3 j + 3x2 yz 2 k) · √ (2j + k)
5
1
= √ (2x2 z 3 + 3x2 yz 2 )
5
10.
Du (x, y, z)
1
= ((z + y)i + (z + x)j + (y + x)k) · √ (i − 4k)
17
1
= √ (z − 3y − 4x)
17

11. Let ϕ(x, y, z) = x2 + y 2 + z 2 so the level surface is given by ϕ(x, y, z) = 4.

The gradient provides a normal vector
N(x, y, z) = ∇ϕ(x, y, z) = 2xi + 2yj + 2zk.
Then √ √
N(1, 1, 2) = 2i + 2j + 2 2k
√
is normal to the surface at (1, 1, 2). The tangent plane at this point has
the equation
√ √
2(x − 1) + 2(y − 1) + 2 2(z − 2) = 0,
or √
x+y+ 2z = 4.
√
The normal line at (1, 1, 2) has parametric equations
√
x = y = 1 + 2t, z = 2(1 + 2t)
for all real t.
12. Write ϕ(x, y, z) = x2 + y − z, so the level surface is given by ϕ(x, y, z) = 0.
A normal vector at (−1, 1, 2) is given by
N(−1, 1, 2) = ∇ϕ(−1, 1, 2) = −2i + j − k.
The tangent plane at (−1, 1, 2) has the equation
−2x + y − z = 1.
The normal line at (−1, 1, 2) is given by
x = −1 − 2t, t = 1 + t, z = 2 − t
in which t takes on all real values.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 386 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

13. Let ϕ(x, y, z) = x2 − y 2 − z 2 . The normal vector at (1, 1, 0) is

N(1, 1, 0) = ∇ϕ(1, 1, 0) = 2i − 2j.

The tangent plane at (1, 1, 0) has equation

2x − 2y = 0

or x = y. The normal line at (1, 1, 0) has parametric equations

x = 1 + 2t, y = 1 − 2t, z = 0.

14. Let ϕ(x, y, z) = x2 − y 2 + z 2 . A normal vector at (1, 1, 0) is given by

N = 2i − 2j.

The tangent plane at (1, 1, 0) has equation y = x. The normal line at this
point has parametric equations

x = 1 + 2t, y = 1 − 2t, z = 0.

15. A normal vector is given by

N = ∇(2x − cos(xyz)) = 2i.

(1,π,1)

The tangent plane has equation x = 1 and the normal line at the point
has parametric equations

x = 1 + 2t, y = π, z = 1.

16. A normal vector at the point is

N = ∇(3x4 + 3y 4 + 6z 4 ) = 12(i + j + 2k).

(1,1,1)

The tangent plane at (1, 1, 1) has equation

x + y + 2z = 4

and the normal line has parametric equations

x = 1 + 12t, y = 1 + 12t, z = 1 + 24t.

17. Because ∇ϕ(x, y, z) = i + k, the normal to the surface ϕ(x, y, z) = c is the

constant vector
N(x, y, z) = i + k.
The surface must therefore be the plane x + z = c.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 15.4. DIVERGENCE AND CURL 387

15.4 Divergence and Curl

1.
∂ ∂ ∂
∇·F= (x) + (y) + (2z) = 4,
∂x ∂y ∂z
i j k
∇ × F = ∂/∂x ∂/∂y ∂/∂z = 0i + 0j + 0k = O,
x y 2z
∇ · (∇ × F) = 0.

2.

∇ · F = xz cosh(xyz),
∇ × F = −xy cosh(xyz)i + yz cosh(xyz)k
∂ ∂
∇ · (∇ × F) = (−xy cosh(xyz)) + (yz cosh(xyz))
∂x ∂z
= cosh(xyz)(y − y) + sinh(xyz)(−xy 2 z + xy 2 z) = 0.

3.
∇ · F = 2y + xey + 2,
∇ × F = (ey − 2x)k,
∂ y
∇ · (∇ × F) = (e − 2x) = 0.
∂x
4.
∇ · F = xzexyz + 3z 2 ,
∇ × F = −xyexyz i + (yexyz − 1)k,
∇ · (∇ × F) = −yexyz − xy 2 exyz + yexyz + xy 2 exyz = 0.

5.
∇ · F = cosh(x) + xz sinh(xyz) − 1,
∇ × F = (−1 − xy sinh(xyz))i − j + yz sinh(xyz)k,

∇ · (∇ × F) = (−y + y) sinh(xyz)
+ ((−xy 2 z + xy 2 z) cosh(xyz) = 0.

6.
∇ · F = cosh(x − z) + 2 + 1 = 3 + cosh(x − z),
∇ × F = −2yi − cosh(x − z)j,
∂ ∂
∇ · (∇ × F) = (2 y) + (− cosh(x − z)) = 0.
∂x ∂y

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 388 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

7.
∇ϕ = i − j + 4zk,

i j k
∇ × (∇ϕ) = ∂/∂x ∂/∂y ∂/∂z = 0.
1 −1 4z

8.
∇ϕ = (18yz + ex )i + 18xzj + 18xyk,

∇ × (∇ϕ) = (18x − 18x)i + (18y − 18y)j + (18z − 18z)k = O.

9.
∇ϕ = −6x2 yz 2 i − 2x3 z 2 j − 4x3 yzk,

∇ × (∇ϕ) = (−4x3 z + 4x3 z)i

+ (−12x2 yzi + 12x2 yz)j + (6x2 z 2 − 6x2 z 2 )k = O.

10.
∇ϕ = z cos(xz)i + x cos(xz)k,

i j k
∇ × (∇ϕ) = ∂/∂x ∂/∂y ∂/∂z
z cos(xz) 0 x cos(xz)
= 0i + (cos(xz) − xz sin(xz) − cos(xz) + xz sin(xz))j + 0k = O.

11.

∇ϕ = (cos(x + y + z) − x sin(x + y + z))i

− x sin(x + y + z)j − x sin(x + y + z)k,

∇ × (∇ϕ) = (−x cos(x + y + z) + x cos(x + y + z))i

+ (− sin(x + y + z) − x cos(x + y + z) + sin(x + y + z) + x cos(x + y + z))j
+ (− sin(x + y + z) − x cos(x + y + z) + sin(x + y + z) + x cos(x + y + z))k
= O.

12.
∇ϕ = ex+y+z (i + j + k),

∇ × (∇ϕ) = (ex+y+z − ex+h+z )(i + j + k) = O.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 15.5. STREAMLINES OF A VECTOR FIELD 389

13. Let F = f i + gj + hk. Then

∇ · (ϕF) = ∇ · (ϕf i + ϕgj + ϕhk)

∂ ∂ ∂
= (ϕf ) + (ϕg) + (ϕh)
∂x ∂y ∂z
= ϕx f + ϕy g + ϕz h
+ ϕ(fx + gy + hz )
= ∇ϕ · F + ϕ(∇ · F).

Next,

i j k
∇ × (ϕF) = ∂/∂x ∂/∂y ∂/∂z
ϕf ϕg ϕh
 
∂ ∂
= (ϕh) − (ϕg) i
∂y ∂z
 
∂ ∂
+ (ϕf ) − (ϕh) j
∂z ∂x
 
∂ ∂
+ (ϕg) − (ϕf k
∂x ∂y
     
∂ϕ ∂ϕ ∂ϕ ∂ϕ ∂ϕ ∂ϕ
= h− g i+ f− h j+ g− f k
∂y ∂z ∂z ∂x ∂x ∂y
     
∂h ∂g ∂f ∂h ∂g ∂f
+ϕ − i+ − j+ − k
∂y ∂z ∂z ∂x ∂x ∂y
= ∇ϕ × F + ϕ(∇ × F).

15.5 Streamlines of a Vector Field

1. The streamlines satisfy
dy dz
dx = − = .
y2 z
Integrate dx = −(1/y 2 ) dy to obtain
1
x= + c1 .
y
Next integrate dx = (1/z) dz to get

x = ln |z| + c2 .

Using x as the parameter, we can write equations of the streamline for

this vector field:
1
x = x, y = , z = ex−c2 .
x − c1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 390 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

For the streamline through (2, 1, 1), let x = 2. Then

1
1= and 1 = e2−c2 .
2 − c1
Thenc1 = 1 and c2 = 2, so this streamline has parametric equations
1
x = x, y = , z = ex−2 .
x−1
2. Streamlines satisfy
1
dx = − dy = dz
2
and two integrations give us
y = −2x + c1 , z = x + c2 .
For the streamline through (0, 1, 1), choose c1 = c2 = 1.
3. We have
dy dz
x dx = = .
ex −1
Integrate xex dx = dy to obtain
y = xex − ex + c1 .
Integrate x dx = −dz to get
x2 = −2z + c2 .
Using x as parameter, streamlines are given by
1
y = xex − ex + c1 , z = (c2 − x2 ).
2
For the streamline through (2, 0, 4), we need
1
e2 + c1 = 0 and 4 = (c2 − 4).
2
Then c1 = −e2 and c2 = 12, so this streamline has parametric equations
1
x = x, y = xex − ex − e2 , z = (12 − x2 ).
2
4. Streamlines satisfy
dx dy
= , dz = 0.
cos(y) sin(x)
Integrate sin(x) dx = cos(y) dy and dz = 0 to get
− cos(x) + c1 = sin(y) and z = c2 .
To pass through (π/2, 0, −4) we need c1 = 0 and c2 = −4. This streamline,
with x as parameter, is
x = x, y = arcsin(− cos(x)), z = −4.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 15.5. STREAMLINES OF A VECTOR FIELD 391

5. Streamlines satisfy
dy dz
z
=− .
2e cos(y)
This is the separable equation

cos(y) dy = −2ez dz.

Integrate this to get

sin(y) = c2 − 2ez .
We also have x
√= c1 . For the streamline through 3, π/4, 0), we need c1 = 3
and c2 = 2 + 2/2. With y as parameter, this streamline is given by
√ !
2 1
x = 3, y = y, z = ln + 1 − sin(y) .
4 2

6. Streamlines satisfy
dx dy dz
=− = 3.
3x2 y z
Integrate the equations
1 3 1 1
2
dx = − dy and dy = − 3 dz
x y y z
to obtain
1 1
= −3 ln |y| + c1 and 2 ln |y| + c2 = 2 .
x z
For the streamline passing through (2, 1, 6), we need c1 = 1/2 and c2 =
1/36. With y as parameter, this streamline has parametric equations
2 6
x= , y = y, z = p .
1 + 6 ln(y) 1 + 72 ln(y)

7. Circular streamlines about the origin in the x, y − −plane can be written

as x2 + y 2 = r2 , so
x dx + y dy = 0.
Then
dx dy
= − , dz = 0.
y x
A vector field having these streamlines is
1 1
F(x, y) = i − j.
x y

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 392 CHAPTER 15. VECTOR DIFFERENTIAL CALCULUS

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 16

Vector Integral Calculus

16.1 Line Integrals

1. On C, x = t, y = t and z = t3 , so
Z Z 1
x dx − dy + z dz = (t(1) − (1) + t3 (3t2 )) dt
C 0
Z 1
= (t − 1 + 3t5 ) dt = 0
0

2.
Z Z 1
2
−4x dx + y dy − yz dz = (−4(−t2 )(−2t)02 − 0) dt
C 0
Z 1
= −8t3 dt = −2
0

3.
Z Z 2 p
(x + y) ds = (2t 1 + 1 + 4t2 dt
C 0
Z 2
√
p 1 2 26 2
= 2t 2 + 4t2 dt = (2 + 4t2 )3/2 =
0 6 0 3
4. Parametric equations of C are
x = t, y = 1 + t, z = 1 − 2t for 0 ≤ t ≤ 1.
Then
Z Z 1 √
x2 z ds = t2 (1 − 2t) 6 dt
C 0
√ Z 1 2 1
= 6 (t − 2t3 ) dt = − √ .
0 6

393

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 394 CHAPTER 16. VECTOR INTEGRAL CALCULUS

5.
Z Z 3
F · dR = (cos(t)i + t2 j + tk) · (i − 2tj + 0k) dt
C 0
Z 3
81
= (cos(t) − 2t3 ) dt = sin(3) −
0 2

6. √
Z Z 2 √ 2 28 6
4xy ds = 4t 6 dt =
C 1 3
7. Parametrize C as x = 2 cos(t), y = 2 sin(t), z = 0 for 0 ≤ t ≤ 2π. Then
Z Z 2π
F · dR = (2 cos(t)i + 2 sin(t)j) · (−2 sin(t)i + 2 cos(t)j) dt
C 0
Z 2π
= (−4 cos(t) sin(t) + 4 cos(t) sin(t)) dt = 0.
0

8. Parametrize C by x = 1, y = t, z = t2 for 0 ≤ t ≤ 2 to get

Z Z 2 p
yz ds = t(t2 ) 1 + 4t2 dt
C 0
Z 2 p 1 √
= t3 1 + 4t2 dt = (391 17 + 1).
0 120
The last integration can be carried out by an integration by parts.
9.
Z Z 9 √
−xyz dz = −z z dz
C 4
2 9 422
= − z 5/2 =−
5 4 5
10. Z Z 3
xz dy = t(−4t2 ) dt = −80
C 1

11. Parametrize the line segment as

x = y = z = 1 + 3t for 0 ≤ t ≤ 1.

The work done is

Z Z 1
F · dR = ((1 + 3t)2 − 2(1 + 3t)2 + 1 + 3t)(3) dt
C 0
1
(1 + 3t)2 (1 + 3t)3

27
= − =− .
2 3 0 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.2. GREEN’S THEOREM 395

12. Parametrize the shape and location of the wire by x = y = z = t for

0 ≤ t ≤ 3. The mass M is
Z Z 3 √ √
27 3
M= δ(x, y, z) ds = 3t 3 dt = .
C 0 2

Because the density function and the position of the wire are symmetric
in the first octant, we will have x = y = z, so we need only compute
Z 3
2
x= √ xδ(x, y, z) ds
27 3 0
2
Z 3 √
= √ t(3t) 3 dt = 2.
27 3 0

The centroid is (2, 2, 2).

13. Take F(x) = f (x)i and R(t) = tj for a ≤ t ≤ b. The graph of the curve is
defined by this position vector is the interval [a, b], and
Z Z b
F · dR = f (x) dx.
C a

16.2 Green’s Theorem

1. The work done by F is
I ZZ  
∂ ∂
work = xy dx + x dy = (x) − (xy) dA
C Ω ∂x ∂y
Z 1 Z 6x Z 4 Z 8−2x
= (1 − x) dy dx + (1 − x) dy dx
0 0 1 0
Z 1 Z 4
= 6x(1 − x) dx + (8 − 2x)(1 − x) dx = −8
0 1

2.
I
work = F · dR
IC
= (ex − y + x cosh(x)) dx + (y 3/2 + x) dy
C
ZZ  
∂ 3/2 ∂ x
= (y + x) − (e − y + x cosh(x)) dA
∂x ∂y
ZZ D
= 2 dA = 2(area of D) = 2(62 )π = 72π
D

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 396 CHAPTER 16. VECTOR INTEGRAL CALCULUS

3.
I
work = (− cosh(4x4 ) + xy) dx + (e−y + x) dy
C
ZZ  
∂ −y ∂ 4
= (e + x) − (− cosh(4x ) + xy) dA
D ∂x ∂y
ZZ Z 3Z 7
= (1 − x) dA = (1 − x) dy dx
D 1 1
Z 3
= 6(1 − x) dx = −12
1

4.
I ZZ  
∂ ∂
F · dR = (−x) − (2y) dA
C ∂x ∂y
ZZ D
= (−3) dA = −3(area of D)
D
= −3(16π) = −48π

5.
I
˙
FdR
C
ZZ  
∂ ∂ 2
= (−2xy) − (x ) dA
D ∂x ∂y
ZZ Z 6 Z (22−2y)/5
= (−2y) dA = −2y dx dy
D 1 (y+4)/5
Z 6
2y
= (3y − 18) dy = −40
1 5
6.
Z ZZ  
∂ ∂
F · dR = (x − y) − (x + y) dA
C ∂x ∂y
ZZ D
= 0 dA = 0
D

7. I ZZ ZZ
∂ 2
F · dR = (8xy ) dA = 8y 2 dA.
C D ∂x D
To evaluate this integral, change to polar coordinates x = r cos(θ), y =
r sin(θ), with 0 ≤ θ ≤ 2π and 0 ≤ r ≤ 4. We get
ZZ Z 2π Z 4
8y 2 dA = 8r2 sin2 (θ)r dr dθ
D 0 0
Z 2π Z 4
= sin2 (θ) dθ 8r3 dr = 512π.
0 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.2. GREEN’S THEOREM 397

8.
I ZZ  
∂ ∂ 2
F · dR = (cos(2y) − e3y + 4x) − (x − y) dA
C ∂x ∂y
ZZ D
= 5 dA = 125
D

9.
I ZZ  
∂ x ∂ x
F · dR = (−e sin(y)) − (e cos(y)) dA
C ∂x ∂y
ZZ D
= (−ex sin(y) + ex sin(y)) dA = 0
D

10.
I ZZ  
∂ 2 ∂ 2
F · dR = (−xy ) − (x y) dA
C D ∂x ∂y
ZZ Z π/2 Z 2
= (−y 2 − x2 ) dA = (−r2 )r dr dθ
D 0 0
Z 2
π
= −r3 dr = −2π
2 0

11.
I ZZ  
˙ = ∂ ∂
FdR (xy 2 − ecos(y) ) − (xy) dA
C D ∂x ∂y
ZZ Z 3 Z 5−5x/3
= (y 2 − x) dA = (y 2 − x) dy dx
D 0 0
Z 3   Z 3  
1 5x 5x
= 5− dx − x 5− ], dx
0 3 3 0 3
95
=
4

12. (a) By Green’s theorem with F = −yi,

I ZZ   ZZ
∂ ∂
−y dx = (0) − (−y) dA = dA = area of D.
C D ∂x ∂y D

(b) Now apply Green’s theorem with F = xj to get

I ZZ   ZZ
∂ ∂
F · dR = (x) − (0) dA = dA = area of D.
C D ∂x ∂y D

(c) Add the results of parts (a) and (b).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 398 CHAPTER 16. VECTOR INTEGRAL CALCULUS

13. By Green’s theorem,

I ZZ     
∂ ∂ ∂ ∂u ∂ ∂u
− dx + dy = − − dA
C ∂y ∂x D ∂x ∂x ∂y ∂y
ZZ  2
∂ u ∂2u

= 2
+ 2 dA.
D ∂x ∂y
14. Assume that C is the join of two curves in two ways. First, C has an upper
piece, the graph of y = p(x), and a lower piece, the graph of y = q(x), for
a ≤ x ≤ b. Then D consists of the points (x, y) with
a ≤ x ≤ b, q(x) ≤ y ≤ p(x).
And C also has a left piece, the graph of x = α(y), and a right piece, the
graph of x = β(y), for c ≤ y ≤ d. Then D also consists of the points (x, y)
such that
c ≤ y ≤ d, α(y) ≤ x ≤ β(y).
Now use both of these descriptions of D as follows. Using the second
(looking at D from left to right),
I Z d Z c
g(x, y) dy = g(β(y), y) dy + g(α(y), y) dy.
C c d
Note that, on the right part of C, we take y varying from d to c to maintain
a counterclockwise orientation around C. Further,
ZZ Z d Z β(y)
∂g ∂g
dA = dx dy
D ∂x c α(y) ∂y
Z d
= (g(β(y), y) − g(α(y), y)) dy.
c

Then I ZZ
∂
g(x, y) dy = dA.
C D ∂x
This is “half” of the conclusion of Green’s theorem. For the rest, use the
first description of C. Now, looking from the lower to the upper piece of
C, and keeping in mind the counterclockwise orientation on C, we have
I Z a Z b
f (x, y) dx = f (x, p(x)) dx + f (x, q(x)) dx
C b a
Z b
=− (f (x, p(x)) − f (x, q(x))) dx
a
and
ZZ Z b Z p(x)
∂f ∂f
dA = dA
D ∂y a q(x) ∂y
Z b
= (f (x, p(x)) − f (x, q(x))) dx.
a

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.2. GREEN’S THEOREM 399

Then
I ZZ
∂f
f (x, y) dx = − dA.
C D ∂y

Upon adding these two results, we have

I ZZ  
∂g ∂f
f (x, y) dx + g(x, y) dy = − dA.
C D ∂x ∂y

15. If C does not enclose the origin, then Green’s theorem applies and
I
F · dR
C
ZZ     
∂ x ∂ −y 2
= − 2y − + x dA
∂x x2 + y 2 ∂y x2 + y 2
ZZ D
= 0 dA = 0.
D

If C does enclose the origin, let K be a circle about the origin of sufficiently
small radius r that K is in the region enclosed by C. Then, using the
extended Green’s theorem and polar coordinates, we have
I I
F · dR = F · dR
C K
Z 2π
  
−r sin(θ) 2 2
= + r cos (θ) (−r sin(θ)) dθ
0 r2
Z 2π   
r cos(θ)
+ − 2r sin(θ) (r cos(θ)) dθ
0 r2
Z 2π
= (1 − r2 cos2 (θ) sin(θ) − 2r2 sin(θ) cos(θ)) dθ
0
r3 2π
=θ+ cos2 (θ) − r2 sin2 (θ) = 2π.
3 0

16. If C does not enclose the origin, then

    
−y
I ZZ
∂ x ∂
F · dR = − y2 − + 3x dA = 0,
C D ∂x x2 + y 2 ∂y x2 + y 2

because all terms in the integrand cancel. If C does enclose the origin, let
K be a circle about the origin and entirely in the region enclosed by C.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 400 CHAPTER 16. VECTOR INTEGRAL CALCULUS

Then
I I
F · dR = F · dR
C K
2π 
−r sin(θ)
Z
= + 3r cos(θ) (−r sin(θ)) dθ
0 r2
Z 2π  
r cos(θ)
+ − r sin(θ) (r cos(θ)) dθ
0 r2
Z 2π
= (1 − 4r2 sin(θ) cos(θ)) dθ = 2π.
0

17. By a calculation like those of Problems 15 and 16, obtain

I
F · dR = 0
C

if C does not enclose the origin, and

I I
F · dR = F · dR = 0
C K

if C does enclose the origin, and K is a circle about the origin entirely in
the region enclosed by C.
18. If C does not enclose the origin, then by Green’s theorem we get
I
F · dR = 0.
C

If C encloses the origin, let K be a circle of radius r about the origin and
entirely in the region enclosed by C. Using polar coordinates for K, a
straightforward calculation yields
I I
F · dR = F · dR = 0.
C K

19. By arguing as in the last four problems, obtain

I
F · dR = 0
C

whether or not C encloses the origin.

16.3 Independence of Path and Potential The-

ory
1. First observe that
∂ 3 ∂
(y ) = (3xy 2 − 4)
∂y ∂x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.3. INDEPENDENCE OF PATH AND POTENTIAL THEORY 401

on the entire plane, so this vector function is conservative. To find a

potential function, we can begin with
∂ϕ
= y3
∂x
and integrate with respect to x to get

ϕ(x, y) = xy 3 + k(y).

Then we must have

∂ϕ
= 3xy 2 + k 0 (y) = 3xy 2 − 4
∂y
to conclude that k 0 (y) = −4, so we can choose k(y) = −4y. Then

ϕ(x, y) = xy 3 − 4y

is a potential function for F.

2. First,
∂ ∂
(6y + yexy ) = 6 + exy + xyexy = (6x + xexy )
∂y ∂x
for all (x, y). Then F has a potential function defined for all (x, y). To
find a potential function, we can begin with
∂ϕ
= 6y + yexy
∂x
and integrate with respect to x to get

ϕ(x, y) = 6xy + exy + k(y).

Then we must have

∂ϕ
= 6y + xexy + k 0 (y) = 6y + yexy .
∂y
Then k 0 (y) = 0 and we can choose k(y) to be any constant, say k(y) = 0
for convenience, to get

ϕ(x, y) = 6xy + exy .

3. F is conservative over the entire plane because

∂ ∂
(16x) = (2 − y 2 ) = 0
∂y ∂x
for all (x, y). To find a potential function, we can begin with
∂ϕ
= 16x
∂x

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 402 CHAPTER 16. VECTOR INTEGRAL CALCULUS

and integrate with respect to y to get

ϕ(x, y) = 8x2 + k(y).
Then
∂ϕ
= k 0 (y) = 2 − y 2
∂y
so k(y) = 2y − y 3 /3 and
1
ϕ(x, y) = 8x2 + 2y − y 3
3
is a potential function.
4. First,
∂ ∂
(2xy cos(x2 )) = 2x cos(x2 ) = (sin(x2 ))
∂y ∂x
for all (x, y). This vector field is conservative. For a potential function,
we can begin with
∂ϕ
= sin(x2 )
∂y
to conclude that
ϕ(x, y) = y sin(x2 ) + k(x).
Then we need
∂ϕ
= 2x cos(x2 ) = 2xy cos(x2 ) + k 0 (x).
∂x
Then k 0 = 0 and we can choose k(y) = 0 to get the potential function
ϕ(x, y) = y sin(x2 ).

5. First, if (x, y) 6= (0, 0), then

   
∂ 2x 4xy ∂ 2y
= − = .
∂y x2 + y 2 (x2 + y 2 )2 ∂y x2 + y 2
Then F is conservative on the plane with the origin removed. For a po-
tential function, we can begin with
∂ϕ 2x
= 2
∂x x + y2
and integrate with respect to x to get
ϕ(x, y) = ln(x2 + y 2 ) + k(y).
Then we need
∂ϕ 2y 2y
= 2 2
= 2 + k 0 (y).
∂y x +y x + y2
We can choose k(y) = 0 to obtain the potential function
ϕ(x, y) = ln(x2 + y 2 ).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.3. INDEPENDENCE OF PATH AND POTENTIAL THEORY 403

6. In this simple case, we can see a potential function by inspection:

ϕ(x, y, z) = x2 − y 2 + z 2

for all (x, y, z).

7. By inspection,
ϕ(x, y, z) = x − 2y + z
is a potential function for F, for all (x, y, z).
8. A routine calculation shows that ∇ × F = O for all (x, y, z), so F is
conservative. A potential function must satisfy
∂ϕ ∂ϕ ∂ϕ
= yz cos(x), = z sin(x) + 1, = y sin(x).
∂x ∂y ∂z
Integrate the first of these equations with respect to x to get

ϕ(x, y) = yz sin(x) + k(y, z).

Next, we need
∂ϕ ∂k
= z sin(x) + 1 = z sin(x) + .
∂y ∂y
Then
∂k
=1
∂y
Integrate this with respect to y to get

k(y, z) = y + c(z).

So far we have
ϕ(x, y, z) = yz sin(x) + y + c(z).
Finally, using the last equation, we have
∂ϕ
= y sin(x) + c0 (z).
∂z
Then c0 (z) = 0, so c(z) can be any constant. For convenience, choose
c(z) = 0. Then
ϕ(x, y, z) = yz sin(x) + y
is a potential function for F.
9. We find that ∇ × F 6= O, so this vector field is not conservative.
10. ∇ × F 6= O, so F is not conservative.

In Problems 11–20 we provide a potential function to use in evaluating the

line integral, but do not include the details of finding this potential function.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 404 CHAPTER 16. VECTOR INTEGRAL CALCULUS

11. By integrating, we find a potential function

ϕ(x, y) = x3 (y 2 − 4y).

Then Z
F · dR = ϕ(2, 3) − ϕ(1, 1) = −24 − 3 = −27.
C

12. ϕ(x, y) = ex cos(y) and

e2
Z
F · dR = ϕ(2, π/4) − ϕ(0, 0) = √ − 1.
C 2

13. In any region not containing points of the y− axis, we can use the potential
function
ϕ(x, y) = x2 y − ln |y|.
If C does not cross the x− axis, then
Z
F · dR = ϕ(2, 2) − ϕ(1, 3)
C
= 8 − ln(2) − 3 + ln(3) = 5 + ln(3/2).

14. With ϕ(x, y) = x + 3y 2 − cos(y), we have

Z
F · dR = ϕ(1, 3) − ϕ(0, 0) = 29 − cos(3).
C

15. ϕ(x, y) = x3 y 2 − 6xy 3 , so

Z
F · dR = ϕ(1, 1) − ϕ(0, 0) = −5.
C

16. ϕ(x, y, z) = xy cos(xz), so

Z
F · dR = ϕ(1, 1, 7) − ϕ(1, 0, π) = cos(7).
C

17. ϕ(x, y, z) = x − 3y 3 z, so
Z
F · dR = ϕ(0, 3, 5) − ϕ(1, 1, 1) = −403.
C

18. ϕ(x, y, z) = −8xy 2 − 4zy, so

Z
F · dR = ϕ(1, 3, 2) − ϕ(−2, 1, 1) = −108.
C

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.3. INDEPENDENCE OF PATH AND POTENTIAL THEORY 405

19. ϕ(x, y, z) = 2x3 eyz , so

Z
F · dR = ϕ(2, 1, −1) − ϕ(0, 0, 0) = 2e−2 .
C

20. ϕ(x, y, z) = xy − 2x2 z + z 3 , so

Z
F · dR = ϕ(3, 1, 4) − ϕ(1, 1, 1) = −5.
C

21. Let C be a smooth path of motion having position vector R(t) = x(t)i +
y(t)j + z(t)k. Let L(t) be the sum of the potential and kinetic energies.
Then
m
L(t) = k R0 (t) k −ϕ((xt), y(t), z(t))
2
m
= R0 (t) · R0 (t) − ϕ(x(t), y(t), z(t)).
2
Then
m ∂ϕ 0 ∂ϕ 0 ∂ϕ 0
L0 (t) = (2R00 (t) · R0 (t)) − x (t) − y (t) − z (t)
2 ∂x ∂y ∂z
= (mR00 (t) · R0 (t) − ∇ϕ · R0 (t)
= (mR00 (t) − ∇ϕ) · R0 (t).

Now, ∇ϕ is the force acting on the particle, so by Newton’s second law,

mR00 = ∇ϕ.

Therefore L0 (t) = 0.

22. We want to show that, in the plane, a potential function exists for F(x, y) =
f (x, y)i + g(x, y)j if
∂g ∂f
= .
∂x ∂y
We will use this condition to construct a potential function for F (on the
relevant region D). First observe that, if K is a closed path in D, then by
Green’s theorem,
I ZZ  
∂g ∂f
F · dR = − dA = 0,
K M ∂x ∂y
R
where M is the region enclosed by K. This means that C F · dR is
independent of path in D. Fix a point P0 : (a, b) in D. Then, for any P
in D, define Z
ϕ(x, y) = F · dR
C

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 406 CHAPTER 16. VECTOR INTEGRAL CALCULUS

in which C is any smooth path in D from P0 to P . We claim that ∇ϕ = F.

To show this, we will first show that
∂ϕ
= f (x, y).
∂x
Choose ∆x small enough that (x + ∆x, y) is also in D. Now

ϕ(x + ∆x, y) − ϕ(x, y)

Z (x+∆x,y) Z (x,y)
= F · dR − F · dR
P0 P0
Z (x,y) Z (x+∆x,y) Z (x,y)
= F · dR + F · dR − F · dR
P0 (x,y) P0
Z (x+∆x,y)
= F · dR
(x,y)
Z (x+∆x,y)
= f (ξ, η) dξ + g(ξ, η) dη.
(x,y)

The last line integral is over the horizontal segment from (x, y) to (x +
∆x, y), which can be parametrized by

ξ = x + t∆x, η = y for 0 ≤ t ≤ 1.

On this segment, dξ = (∆x) dt and dη = 0, because y is constant on this

segment. Then
Z 1
ϕ(x + ∆x, y) − ϕ(x, y) = ∆x f (x + t∆x, y) dt.
0

Then
1
ϕ(x + ∆x, y) − ϕ(x, y)
Z
= f (x + t∆x, y) dt.
∆x 0
By the mean value theorem for integrals, there is some t0 in (0, 1) such
that Z 1
f (x + t∆x, y) dt = f (x + t0 ∆x, y).
0
Then
ϕ(x + ∆x, y) − ϕ(x, y)
= f (x + t0 ∆x, y).
∆x
In the limit as ∆x → 0, we must have t0 → 0, so x + t0 ∆x → x and
f (x + t0 ∆x, y) → f (x, y). Then
∂ϕ ϕ(x + ∆x, y) − ϕ(x, y)
= lim
∂x ∆x→0 ∆x
= lim f (x + t0 ∆x, y) = f (x, y).
∆x→0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.4. SURFACE INTEGRALS 407

A similar argument, using a vertical segment from (x, y) to (x, y + ∆y),

shows that
∂ϕ
= g(x, y).
∂y

16.4 Surface Integrals

1. On the surface, z = 10 − x − 4y, so
p √
dσ = 1 + (∂z/∂x)2 + (∂z/∂y)2 dA = 3 2 dA.

Then
ZZ ZZ √
x dσ = 3 2x dA
Σ D
√ Z
√ Z 5/2 Z 10−4y 3 2 5/2
=3 2 x dx dy = (10 − 4y)2 dy
0 0 2 0
√
2 3
5/2 √
= (10 − 4y) = 125 2.
8 0

2. On the surface, z = x, so
p √
dσ = 1 + 12 + 02 dA = 2 dA

and
ZZ
2
ZZ √ 2
y dσ = 2y dA
Σ D
√
√ Z 2 Z 4
2 128 2
= 2 y dx dy = .
0 0 3

3. On Σ, p
dσ = 1 + 4x2 + 4y 2 dA,
and D is the annulus 2 ≤ x2 + y 2 ≤ 7. Then, using polar coordinates,
√
ZZ Z 2π Z 7 p
dσ = √ r 1 + 4r2 r dr dθ
Σ 0 2
 √7
1 π
= 2π (1 + 4r2 )3/2 √
= (293/2 − 27).
12 2 6

4. On the surface, z = (25 − 4x − 8y)/10, so

p 3
dσ = 1 + (2/5)2 + (4/5)2 dA = √ dA.
5

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 408 CHAPTER 16. VECTOR INTEGRAL CALCULUS

Then
ZZ ZZ
3
(x + y) dσ = (x + y) √ dA
Σ D 5
Z 1Z x
3
=√ (x + y) dy dx
5 0 0
3√
Z 1
3 3 2
=√ x dx = 5.
5 0 2 2

5. On the surface, z 2 = x2 + y 2 , so

∂z ∂z
2z = 2x and 2z = 2y.
∂x ∂y

Then
∂z x ∂z y
= and = .
∂x z ∂y z
Then r  x 2  y 2 √
dσ = 1+ + dA = 2 dA.
z z
Then
ZZ ZZ √ p
z dσ = 2 x2 + y 2 dA
Σ D
√ Z π/2 Z 4
28π √
= 2 r2 dr dθ = 2.
0 2 3

√
6. On Σ, dσ = 3 dA and z = x + y, so
ZZ √ ZZ
xyz dσ = 3 xy(x + y) dA
Σ D
√ Z 1Z 1
1
= 3 (x2 y + xy 2 ) dy dx = √ .
0 0 3

√
7. On the surface, dσ = 1 + 4x2 dA, so
Z ZZ p
y dσ = y 1 + 4x2 dA
Σ D
2Z 3
9 2p
Z p Z
= 1 + 4x2 dy dx =
y 1 + 4x2 dx
0 0 2 0
9 √ √
= ln(4 + 17 + 4 17).
8

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.4. SURFACE INTEGRALS 409
p
8. On the surface, dσ = 1 + 4(x2 + y 2 ) dA, so
ZZ ZZ p
x2 dσ = x2 1 + 4(x2 + y 2 ) dA
Σ D
Z 2π Z 2 p
= r2 cos2 (θ) 1 + 4r2 r dr dθ
0 0
Z 2π Z 2 p
= cos2 (θ) dθ r3 1 + 4r2 dr.
0 0

For the θ−integration, use the identity

1
cos2 (θ) = (1 + cos(2θ))
2
and for the r−integration, use the substitution

u = 1 + 4r2 .

These give us

1 2π
ZZ Z   Z 17
2 1
x dσ = (1 + cos(2θ) dθ (u3/2 − u1/2 ) du
Σ 2 0 32 1
π √
= (782 17 + 2).
240
√
9. On Σ, dσ = 3 dA and z = x − y, so
ZZ ZZ √
z dσ = 3(x − y) dA
Σ D
√ Z 1 Z 5 √
= 3 (x − y) dy dx = −10 3.
0 0

p
10. On Σ, dσ = 1 + 4y 2 dA and z = 1 + y 2 , so
ZZ ZZ p
xyz dσ = xy(1 + y 2 ) 1 + 4y 2 dA
Σ D
Z 1Z 1 p
= xy(1 + y 2 ) 1 + 4y 2 dy dx
0 0
Z 1
1 p
= y(1 + y 2 ) 1 + 4y 2 dy.
2 0

For this integral, let u = 1 + 4y 2 to get

5 √
ZZ Z  
1 1 1 3
xyz dσ = (u3/2 + 3u1/2 ) du = 5 5− .
Σ 2 32 1 16 5

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 410 CHAPTER 16. VECTOR INTEGRAL CALCULUS

16.5 Applications of Surface Integrals

1. The triangular shell is on the plane 6x + 2y + 3z = 6, which is the plane
through the three given points. The projection of Σ onto the x, y−plane
is the set D of points (x, y) such that 0 ≤ y ≤ 3 − 2x. On Σ,
2
z = 2 − y − 2x.
3
Then r
4
dσ = 1+ + 4 dA
9
and
ZZ
m= (xz + 1) dσ
ZZ Σ    
2
= x 2 − y − 2x + 1 dA
D 3
Z 1 Z 3−3x    
7 2 49
= x 2 − y − 2x + 1 dy dx = .
3 0 0 3 12
The first coordinate of the center of mass is
Z
12
x= x(xz + 1) dσ
49 Σ
ZZ    
12 7 2
= x x 2 − y − 2x + 1 dy dx
49 3 D 3
Z 1 Z 3−3x    
4 2
= x x 2 − y − 2x + 1 dx
7 0 0 3
Z 1 
24 2 12 4 12 12
= − x + x + x dx = .
0 7 7 7 35
The second coordinate is
ZZ
12
y= y(xz + 1) dσ
49 Σ
4 1 3−3x
Z Z    
2
= y y 2 − y − 2x + 1 dy dx
7 0 0 3
Z 1 
24 18 36 12 18
= − x − x2 + x3 − x4 + dx
0 7 7 7 7 7
33
= .
35
And, without all the details, the third coordinate is
ZZ
12 24
z= z(xz + 1) dσ = .
49 Σ 35

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.5. APPLICATIONS OF SURFACE INTEGRALS 411

2. On Σ,
p 3
dσ = 1 + (x/z)2 + (y/z)2 dA =
dA.
z
The part of the sphere above√the plane z = 1 projects onto the x, y−−plane
onto the disk D of radius 2 2 about the origin. The mass of the shell is
ZZ ZZ
3 1
m= K dA = 3K p dA.
Σ z D 9 − (x 2 + y2 )

Use polar coordinates:

√
Z 2π Z 2 2
r
m = 3K √
dr dθ
0 0 9 − r2
h p i2√2
= 6πK − 9 − r 2 = 12Kπ.
0

By symmetry, x = y = 0. Finally,
ZZ
1
z= Kz dσ
m
ZZ Σ  
1 3
= Kz dA
m D z
3K 1
= (area of D) = (24Kπ) = 2.
m m
The center of mass is (0, 0, 2).

3. On the surface,
p √
dσ = 1 + (x/z)2 + (y/z)2 dA = 2 dA.

Then
ZZ √ ZZ Z 2π Z 3 √
mass = K dσ = K 2 r dr dθ = 9πK 2.
Σ D 0 0

By symmetry, x = y = 0, and
ZZ
1
z= z dσ
m Σ
√ √
2K 2π 3 2
Z Z
18 2Kπ
= r dr dθ = = 2.
m 0 0 m

The center of mass is (0, 0, 2).

4. On the surface, p
dσ = 1 + 4(x2 + y 2 ) dA

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 412 CHAPTER 16. VECTOR INTEGRAL CALCULUS

and Σ projects onto the x, y − −plane to the quarter annulus D consisting

of points (x, y) with
x ≥ 0, y ≥ 0 and 1 ≤ x2 + y 2 ≤ 9.
The mass is
ZZ
xy
m= p dσ
Σ 1 + 4(x2 + y 2 )
ZZ Z π/2 Z 3
= xy dA = r3 sin(θ) cos(θ) dr dθ
D 0 0
  
1 π/2 1 43
2
= sin (θ) r
2 0 4 1
= 10.

Because the region is symmetric about the line y = x, and

1
δ(x, y, z) = δ(r, θ) = √
1 + 4r2
is independent of θ, we must have x = y. Compute
ZZ
1
x= xδ(x, y, z) dσ
m Σ
ZZ Z π/2 Z 3
1 2 1 121
= x y dA = r4 cos2 (θ) sin(θ) dr dθ = .
10 D 10 0 1 75
Finally,
ZZ ZZ
1 1
z= zδ(x, y, z) dσ = (16 − x2 − y 2 )xy dA
m Σ 10 D
Z π/2 Z 3
1 331
= (16 − r2 )r3 sin(θ) cos(θ) dr dθ = .
10 0 1 48
The center of mass is (121/75, 121/75, 331/48).
5. By symmetry of the surface and the density function, x = y = 0. Further,
p
dσ = 1 + 4x2 + 4y 2 dA.
Then
ZZ p
m= 1 + 4x2 + 4y 2 dσ
Σ
ZZ
= (1 − 4x2 + 4y 2 ) dA
D
√
Z 2π Z 6
= (1 + 4r2 )r dr dθ
0 0
= 2π(39) = 78π.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.5. APPLICATIONS OF SURFACE INTEGRALS 413

Finally,
ZZ
1
x= zδ(x, y, z) dσ
m Σ
ZZ
1
= (6 − x2 − y 2 )(1 + 4x2 + 4y 2 ) dA
m D
Z √6
1 2π
Z
= (6 − r2 )(1 + 4r2 )r dr dθ
m 0 0
162π 27
= = .
m 13

6. By symmetry, x = y = 0. On the surface,

p 1
dσ = 1 + (x/z)2 + (y/z)2 dA = dA.
z
The mass is
ZZ ZZ
K 1 1
m= Kz dσ = z(1/z) dA = area of D = .
Σ Kπ D π 4

7. A unit normal to the plane x + 2y + z = 8 is

1
n = √ (i + 2j + k).
6
Then
1
F · n = √ (x + 2y − z).
6
On Σ, z = 8 − x − 2y, so

F · n = 2x + 4y − 8.
√ √
Further, dσ = 1 + 4 + 1 dA = 6 dA, so the flux of F across the surface
is
ZZ ZZ Z 4 Z 8−2y
128
F·n dσ = (2x+4y −8) dA = (2x+4y −8) dx dy = .
Σ D 0 0 3

8. The sphere x2 + y 2 + z 2 = 4 has unit normal

1
n= (xi + yj + zk).
2
Then
1 2
F·n= (x z − yz).
2
Further, p
dσ = 1 + (−x/z)2 + (−y/z)2 dA.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 414 CHAPTER 16. VECTOR INTEGRAL CALCULUS

Therefore the flux of the force across Σ is

ZZ ZZ
F · n dσ = (x2 − y) dA.
Σ D
√
Use polar coordinates on D, which is the disk 0 ≤ r ≤ 3, 0 ≤ θ ≤ 2π.
Then the flux is
√
Z 2π Z 3 Z 2π
9 9π
(r2 cos2 (θ) − r sin(θ))r dr dθ = cos2 (θ) dθ = .
0 0 4 0 4

16.6 Gauss’s Divergence Theorem

1. ∇ · F = 1, so
ZZ ZZZ
F · n dσ = ∇ · F dV
Σ
M
4 3 256π
= volume of V = π4 = .
3 3

2. ∇ · F = 4 − 6 = −2, so
ZZZ
∇ · F dV = −2(volume of M ) = −2π(22 )(2) = −16π.
M

3. ∇ · F = 0, so ZZZ
∇ · F dV = 0.
M

4. ∇ · F = 3(x2 + y 2 + z 2 ), so
ZZZ ZZZ
∇ · F dV = 3 (x2 + y 2 + z 2 ) dV.
M M

Using spherical coordinates, we have

ZZZ Z 2π Z π Z 1
∇ · F dV = 3ρ4 sin(ϕ) dρ dϕ dθ
0 0 0
M
Z 2π Z π Z 1
= dθ
sin(ϕ) dϕ 3ρ4 dρ
0 0 0
 
3 12π
= (2π)(2) = .
5 5

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.6. GAUSS’S DIVERGENCE THEOREM 415

5. With ∇ · F = 4, compute
ZZZ
8π
∇ · F dV = 4(volume of )V = .
3
M

6. ∇ · F = 2 + x, so
ZZZ Z 3 Z 2 Z 4 Z 4
∇·F= (2 + x) dx dy dz = 6 (2 + x) dx = 96.
0 0 0 0
M

7. ∇ · F = 2(x + y + z), so, using cylindrical coordinates, we have

ZZZ Z 2π Z √2 Z √2
∇ · F dV = 2 (r cos(θ) + r sin(θ) + z)r dz dr dθ.
0 0 r
M

Do these integrations in turn. First,

Z √2 √ 1
(r2 (cos(θ)+sin(θ))+rz) dz = r2 (cos(θ)+sin(θ))( 2−r)+ r(2−r2 ).
r 2
Next,
Z √2  √

1 1 1
r2 (cos(θ) + sin(θ))( 2 − r) + r(2 − r2 ) dr = (cos(θ)+sin(θ))+ ,
0 2 3 2
and finally, Z 2π  
1 1
(cos(θ) + sin(θ)) + dθ = π.
0 3 2
Therefore ZZZ
∇ · F dV = 2π.
M

8. ∇ · F = 1 + 2x, so compute
ZZZ ZZZ
∇ · F dV = (1 + 2x) dV
M M
Z 2 ZZ
= dz (1 + 2x) dA,
0 D
√
where D is the plane region given in polar coordinates by 0 ≤ r ≤ 2,
0 ≤ θ ≤ 2π. Now compute
ZZZ ZZ
(1 + 2x) dV = 2 (1 + 2r cos(θ))r dr dθ
D
M
"√ Z #
4 2 2π
= 2 2π + cos(θ) dθ = 4π.
3 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 416 CHAPTER 16. VECTOR INTEGRAL CALCULUS

9. With the given conditions on F, Σ and M , we have

ZZ ZZZ
(∇ × F) · n dσ = ∇ · (∇ × F) dV = 0
Σ
M

because ∇ · (∇ × F) = 0.
10. Apply the divergence theorem to get
ZZ ZZZ
1 1
F · n dσ = (∇ · R) dV
3 Σ 3
M
ZZZ
1
= 3 dV = volume of M.
3
M

16.7 Stokes’s Theorem

1. The surface is a function of θ and ϕ, and (θ, ϕ) varies over its parameter
domain 0 ≤ θ ≤ 2π, 0 ≤ ϕ ≤ π. This is a rectangle in the θ, ϕ− plane,
with lower side L1 , upper right side L2 , top L3 and left side L4 . For
orientation, imagine (θ, ϕ) moves around this rectangle counterclockwise,
starting along L1 from the origin. We want to know what each side maps
to on the surface.
On L1 , the point (θ, 0) moves from (0, 0) to (2π, 0) as θ increases from 0
to 2π. The image point

Σ(θ, 0) = (R cos(θ), R sin(θ), 0)

moves from (R, 0, 0) along the circle x2 + y 2 = R2 in the plane z = 0, all

the way around to end at (R, 0, 0).
Then the point (2π, ϕ) moves up along L2 as ϕ increases from 0 to π.
Image points of points (2π, ϕ) on L2 are

Σ(2π, ϕ) = (R cos(ϕ), 0, R sin(ϕ))

which is a half-circle x2 + z 2 = R2 in the y = 0 plane, starting at (R, 0, 0)

and ending at (−R, 0, 0).
From (2π, π), (θ, ϕ) now moves left along L3 . The points are (θ, π), but
θ varies from 2π to 0 to maintain counterclockwise orientation on the
rectangle. The image points of L3 are

Σ(θ, π) = (−R cos(θ), −R sin(θ), 0)

asθ varies from2π to 0. The image of L3 on the surface consists of the

points
Σ(θ, π) = (−R cos(θ), −R sin(θ), 0),

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.7. STOKES’S THEOREM 417

and this point moves along the half-circle

x2 + y 2 = R 2

from (−R, 0, 0) to (−R, 0, 0) in the z = 0 plane.

Finally, on L4 , θ = 0 and ϕ varies from π to 0. Image points are

Σ(0, ϕ) = (R cos(ϕ), 0, R sin(ϕ))

from (−R, 0, 0) to (R, 0, 0).

Now trace out the image point on the surface as (θ, ϕ) moves over all
four sides of the rectangle. This curve on the graph of the surface is the
boundary of Σ.

2. The boundary curve C of the surface can be parametrized by

x = 2 cos(t), y = 2 sin(t), z = 0 for 0 ≤ t ≤ 2π.

On C, R(t) = 2 cos(t)i + 2 sin(t)j + 0k. Then

F · dR = (−16 cos2 (t) sin2 (t) − 16 cos2 (t) sin2 (t)) dt = −32 cos2 (t) sin2 (t).

Then I Z 2π
F dR = −32 cos2 (t) sin2 (t) dt = −8π.
C 0
RR
Now evaluate Σ
(∇ × F) · n dσ. First,

∇ × F = −(x2 + y 2 )k.

Further, we can use the gradient to find a normal to Σ:

∇(x2 + y 2 + z 2 ) = 2(xi + yj + zk).

This gives us the normal vector

1
n= (xi + yj + zk).
2
Then
p 2
dσ = 1 + (x/z)2 + (y/z)2 dA = dA.
z
Then
ZZ ZZ
(∇ × F) · n dσ = − (x2 + y 2 ) dA
Σ D
Z 2π Z 2
=− r3 dr dθ = −8π.
0 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 418 CHAPTER 16. VECTOR INTEGRAL CALCULUS

In Problems 3–8, one side of Stokes’s theorem is computed in detail, with the
choice being determined by which side appears to be the easiest computation.

3. The boundary curve C of the surface is the top of the parabolic bowl.
This is the circle of radius 3 about (0, 0, 9). Parametrize C by

x = 3 cos(t), y = 3 sin(t), z = 9 for 0 ≤ t ≤ 2π.

On C,
F(t) = 9 cos(t) sin(t)i + 27 sin(t)j + 27 cos(t)k.
Further,
dR = (−3 sin(t)i + 3 cos(t)j) dt.
Then
F · dR = (−27 cos(t) sin2 (t) + 81 cos(t) sin(t)) dt.
A routine integration gives
I Z 2π
F · dR = (−27 cos(r) sin2 (t) + 81 cos(t) sin(t)) dt = 0.
C 0
RR
Evaluation of Σ
(∇ × F) dσ involves considerably more labor.

4. Compute ∇ × F = i + j + k. A unit normal to Σ is

1
n= p (xi + yj + zk)
x + y2 + z2
2

This is
1
n= √ p (xi + yj − 2k).
2 x2 + y 2
Further, s
x2 y2 √
dσ = 1+ + dA = 2 dA.
x2 + y 2 x2 + y 2
Then
x+y−z
ZZ ZZ
(∇ · F) dσ = p dA
Σ D x2 + y 2
ZZ !
x+y
= p −1 dA.
D x2 + y 2
p
Here we used the fact that z = x2 + y 2 on σ. This integral is easily
evaluated using polar coordinates to get
ZZ
(∇ × F) dσ = −16π.
Σ

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 16.7. STOKES’S THEOREM 419

For the line integral around the boundary C of Σ, parametrize C by

x = 4 sin(t), y = 4 cos(t), z = 4 for 0 ≤ t ≤ 2π.

This orientation is determined by n. Notice that the normal in this prob-

lem has negative k component, unlike the normal in the preceding prob-
lem, where the surface was a parabolic bow. Parametrize C by
I Z 2π
F · dR = (−16 cos(t) sin(t) − 16 cos2 (t)) dt = −16π.
C 0

5. The boundary curve of Σ is the circle x2 + y 2 = 6 in the x, y− plane.

Parametrize C by
√ √
x = 6 cos(t), y = 6 sin(t), z = 0 for 0 ≤ t ≤ 2π.

Then I Z 2π
F · dR = 6 cos2 (t)6 sin2 (t) dt = 0.
C 0

6. The boundary C of the surface must be parametrized by three smooth

curves. This is not difficult, but it is tedious. This leads us to try the
surface integral side of Stokes’s theorem. First,

∇ × F = (x − y)i − yj − xk

and
1
n = √ (2i + 4j + k).
21
√
Finally, dσ = 21 dA, so
ZZ ZZ
(∇ × F) dσ = (x − 6y) dA
Σ D
Z 2 Z 4−2y
32
= (x − 6y) dx dy = − .
0 0 3
H
7. The circulation is C
F · dR. Take Σ to be the disk

0 ≤ x2 + y 2 ≤ 1

with boundary C parametrized by

x = cos(t), y = sin(t), z = 0 for 0 ≤ t ≤ 2π.

The proper unit normal to Σ (a disk in the x, y− plane) is n = k. Now,

∇ × F = −azj + (2xy + 1)k.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 420 CHAPTER 16. VECTOR INTEGRAL CALCULUS

Then
(∇ × F) · n = 2xy + 1.
Further, dσ = dA, so
I ZZ
F · dR = (∇ × F) · n dσ
C Σ
ZZ Z 2π Z 1
= (2xy + 1) dA = (2r3 cos(θ) sin(θ) + r) dr dθ = π.
D 0 0

8. First,
∇ × F = −i − j − k.
A normal to the surface is

n = i + 4j + k

so we have the unit normal

1
n = √ (i + 4j + k).
18
C is the boundary of the part of the plane x + 4y + z = 12 lying in the
first octant, and consists of three line segments: the line from (0, 0, 12)
to (12, 0, 0), then from (12, 0, 0) to (0, 3, 0), and then from (0, 3, 0) to
(0, 0, 12). We can think of this portion of the plane in the first octant as
having equation
z = 12 − x − 4y
with (x, y) varying over the triangle D bounded by the segment [0, 12] on
the x− axis, the segment [0, 3] on the y− axis, and the line x + 4y = 12.
D has area
1
area of D = (12)(3) = 18.
2
By Stokes’s theorem the circulation is
I ZZ
F · dT ds = (∇ × F) · n dσ.
C D

Now
6
(∇ × F) · n = − √ .
18
And √
dσ =k N k dx dy = 18.
Then
ZZ ZZ
6
(∇ × F) · n dσ = − √ dx dy
D D 18
= −6(area of D) = −6(18) = −108.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 17

Fourier Series

17.1 Fourier Series on [−L, L]

1. The Fourier coefficients are
1 3
Z
a0 = 4 dx = 8,
3 −3
1 3
Z
an = 4 cos(nπξ/3) dξ = 0,
3 −3
and Z 3
1
bn = 4 sin(nπξ/3) dξ = 0.
3 −3
The Fourier series of 4 on [−3, 3] is just
1
a0
2
or 4, as we might expect. This converges to 4 on [−3, 3].
2. The Fourier coefficients an are zero for n = 0, 1, 2, · · · because f is an odd
function on [−1, 1]. Next,
Z 1
2
bn = f (ξ) sin(nπξ) dξ = (−1)n .
−1 nπ
The Fourier series of f (x) on [−1, 1] is
∞
2X1
(−1)n sin(nπx).
π n=1 n

This series converges to

(
−x for −1 < x < 1,
0 for x = ±1.

421

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 422 CHAPTER 17. FOURIER SERIES

3. Because cosh(πx) is an even function on [−1, 1], each bn = 0. Compute

Z 1
2
a0 = cosh(πξ) dξ = sinh(π)
−1 π
and, for n = 1, 2, · · · ,
Z 1
2 sinh(π) (−1)n
an = cosh(πξ) cos(nπξ) dξ = .
−1 π n2 + 1
The Fourier series is
∞
1 X 2 sinh(π) (−1)n
sinh(π) + cos(nπx).
π n=1
π n2 + 1

This converges to cosh(πx) for −1 ≤ x ≤ 1.

For Problems 4–10, we give just the Fourier series and analyze its conver-
gence.

4. The series of f (x) on [−2, 2] is

∞
8 X 1
cos((2n − 1)πx/2).
π n=1 (2n − 1)2
2

This converges to 1 − |x| for −2 ≤ x ≤ 2.

5. The series of f (x) on [−π, π] is
∞
16 X 1
sin((2n − 1)x),
π n=1 2n − 1

converging to 
−4
 for −π < x < 0,
4 for 0 < x < π,

0 for x = π and for x = −π.


6. Because sin(2x) is odd and periodic of period π, then this is itself the
Fourier expansion of sin(2x) on [−π, π].
7. The Fourier series of f (x) on [−2, 2] is
∞  
13 X 16 4
+ (−1)n cos(nπx/2) + sin(nπx/2) .
3 n=1
(nπ)2 nπ

This converges to
(
f (x) for −2 < x < 2,
7 for x = 2 and for x = −2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.1. FOURIER SERIES ON [−L, L] 423

Convergence at the endpoints is determined by

1 1
(f (−2+) + f (2−)) = (9 + 5) = 7.
2 2

8. The Fourier series of f (x) on [−5, 5] is

∞
1 X
a0 + [an cos(nπx/5) + bn sin(nπx/5)] ,
2 n=1

where Z 5
1 71
a0 = f (ξ) dξ =
5 −5 6
and, for n = 1, 2, · · · ,
5
55(−1)n − 5
Z
1
an = f (ξ) cos(nπξ/5) dξ =
5 −5 (nπ)2
and
5
n2 π 2 − 50 + (50 − 21n2 π 2 )(−1)n
Z
1
bn = f (ξ) sin(nπξ/5) dξ = .
5 −5 n3 π 3
This Fourier series converges to


 −x for −5 < x < 0,
1 + x2 for 0 < x < 5,




 1/2 at x = 0,
31/2 for x = 5 and for x = −5.


9. The Fourier expansion of f (x) on [−π, π] is

∞
3 2X 1
+ sin((2n − 1)x)
2 π n=1 2n − 1

converging to

1
 for −π < x < 0,
2 for 0 < x < π,

3/2 at x = 0, x = π and at x = −π.


10. The Fourier series of f (x) on [−π, π] is

∞
2 4 X (−1)n
− sin(x) − cos(nx),
π π n=1 4n2 − 1

converging to f (x) for −π < x < π and to 0 at x = ±π.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 424 CHAPTER 17. FOURIER SERIES

11. The Fourier series is

∞
1 X (−1)n+1
sin(3) + 6 sin(3) cos(nπx/3),
3 n=1
n2 π 2 − 9

converging to cos(x) on [−3, 3].

12. The series is
∞ 
1 − (−1)n 1 − 2(−1)n

1 X
sin(3) − cos(nπx) + sin(nπx) ,
3 n=1
n2 π 2 nπ

converging to 

1−x for −1 < x < 0,

0 for 0 < x < 1,


1/2 for x = 0,
1 for x = ±1.


13. The Fourier series converges to



 3/2 for x = ±3,

2x for −3 < x < −2,





−2 for x = −2,
0

 for −2 < x < 1,
1/2 for x = 1,





 2
x for 1 < x < 3.

14. 

(1 − 2π)/2 for x = ±π,

3/2 for x = 1,


2x − 2 for −π < x < 1,
3 for 1 < x < π


15.  2
(2 + π )/2 for x = ±π,

x2

for −π < x < 0,
1
 for x = 0,

2 for 0 < x < π


16. 
(cos(2) + sin(2))/2 for x = ±2,


cos(x) for −2 < x < 0,
1/2
 for x = 0,

sin(x) for 0 < x < 2


© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.2. SINE AND COSINE SERIES 425

17. 
−1
 for −4 < x < 0,
0 for x = ±4 and for x = 0,

1 for 0 < x < 4


18. 


1 for x = ±1 and for 1/2 < x < 3/4,
0 for −1 < x < 1/2,



2 for 3/4 < x < 1,

1/2 for x = 1/2,





3/2 for x = 3/4
19. 

−4 for x = ±4,

3/2 for x = −2,
5/2
 for x = 2,

f (x) for all other x in [−4, 4]


17.2 Sine and Cosine Series

1. The cosine series is just 4, a single term. The sine series is
∞
16 X 1
sin((2n − 1)πx/3),
π n=1 2n − 1

converging to 0 for x = 0 or x = 3 and to 4 for 0 < x < 3.

2. The cosine series is
∞
4 X (−1)n
− cos((2n − 1)πx/2),
π n=1 2n − 1

converging to 
1
 for 0 ≤ x < 1,
0 for x = 1,

−1 for 1 < x ≤ 2.


The sine series is

∞
2X1
(1 + (−1)n − 2 cos(nπ/2)) sin(nπx/2),
π n=1 n

converging to 
1
 for 0 < x < 1,
0 for x = 0, 1, 2,

−1 for 1 < x < 2.


© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 426 CHAPTER 17. FOURIER SERIES

3. The cosine series is

∞
1 X 2n sin(nπ/2)
cos(x) + cos(nx/2)
2 π(n2 − 4)
n=1,n6=2

converging to


0 for 0 < x < π,

−1/2 for x = π,
cos(x) for π < x < 2π,


1 for x = 2π.


The sine series is

∞
2 X 2n
− sin(x/2) − 2 − 4)
((−1)n + cos(nπ/2)) sin(nx/2),
3π n=3
(n

converging to


0 for 0 < x < π,

−1/2 for x = π,


cos(x) for π < x < 2π,
0 for x = 2π.


4. The cosine series is

∞
8 X 1
1− cos((2n − 1)πx),
π n=1 (2n − 1)2
2

converging to 2x for 0 ≤ x ≤ 1. The sine series is

∞
4 X (−1)n
− sin(nπx),
π n=1 n

converging to 2x for 0 ≤ x < 1 and to 0 at x = 1.

5. The cosine series is

4 16 (−1)n
+ cos(nπx/2),
3 π 2 n2

converging to x2 for 0 ≤ x ≤ 2. The sine expansion is

∞ 
8 X (−1)n 2(1 − (−1)n )

− + sin(nπx/2),
π n=1 n n3 π 2

converging to x2 for 0 ≤ x < 2 and to 0 at x = 2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.2. SINE AND COSINE SERIES 427

6. The cosine series is

∞
1 X 1 − (−1)n e−1
−1 − +2 cos(nπx),
e n=1
1 + n2 π 2

converging to e−x for 0 ≤ x ≤ 1. The sine series is

∞  
X n n −1
2π (1 − (−1) e ) sin(nπx),
n=1
1 + n2 π 2

converging to e−x for 0 < x < 1 and to 0 at x = 0 and at x = 1.

7. The cosine expansion is

∞
1 X 1
+ [−6(1 + (−1)n ) + 12 cos(2nπ/3) + 4nπ sin(2nπ/3)] cos(nπx/3),
2 n=1 n2 π 2

converging to 
x
 for 0 ≤ x < 2,
1 for x = 2,

2−x for 2 < x ≤ 3.


The sine series is

∞
X 1
[12 sin(2nπ/3) − 4nπ cos(2nπ/3) + 12nπ(−1)n ] sin(nπx/3),
n=1
n2 π 2

converging to


x for 0 ≤ x < 2,

1 for x = 1,


2−x for 2 < x < 3,
0 for x = 3.


8. The cosine series is

∞
1 1X1
− + cos(nπ/5) sin(2nπ/5) cos(nπx/5),
5 π n=1 n

converging to 


1 for 0 ≤ x < 1,
1/2 for x = 1,



0 for 1 < x < 3,

−1/2 for x = 3,





−1 for 3 < x < 5.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 428 CHAPTER 17. FOURIER SERIES

The sine series is

∞
4X 1
(1 + (−1)n − 2 cos(nπ/5) cos(2nπ/5)) sin(nπx/5),
π n=1 2n

converging to



 1 for 0 < x < 1,
1/2 for x = 1,



0 for 1 < x < 3 and for x = 0 and x = 5,

−1/2

 for x = 3,


−1 for 3 < x < 5.

9. The cosine expansion is

∞  
5 16 X 1 4
+ 2 cos(nπ/4) − sin(nπ/4) cos(nπx/4),
6 π n=1 n2 n3 π

converging to x2 for 0 ≤ x < 1 and to 1 for 1 < x ≤ 4.

The sine series is
∞  
X 16 61 2
sin(nπ/4) + 3 3 (cos(nπ/4) − 1) − nπ sin(nπx/4),
n=1
n2 π 2 n π (−1)n

converging to x2 for 0 ≤ x < 1, to 1 if 1 < x < 4, and to 0 at x = 4.

10. The cosine series is
∞  
24 X 1 n 4 n
−1 − 2(−1) + (1 − (−1) ) cos(nπx/2),
π 2 n=1 n2 n2 π 2

converging to 1 − x2 for 0 ≤ x ≤ 2.
The sine series is
∞  
2X1 n 48 n
1 + 7(−1) − 2 2 (−1) sin(nπx/2),
π n=1 n n π

converging to 1 − x2 for 0 < x < 2 and t 0 at x = 0 and at x = 2.

11. The Fourier cosine expansion of sin(x) on [0, π] is
∞
2 4X 1
− 2
cos(2nx).
π π n=1 4n − 1

This converges to sin(x) for 0 ≤ x ≤ π. Put x = π/2 into this series to

obtain
∞
2 4X 1
cos(π/2) = 0 = − cos(nπ).
π π n=1 4n2 − 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.2. SINE AND COSINE SERIES 429

Upon putting cos(nπ) = (−1)n we obtain

∞
(−1)n
 
X π 2 1 π
2−1
= − 1 = − .
n=1
4n 4 π 2 4

12. Expand Go (x) in a Fourier series on [−L, L]. This series will be
∞
1 X
a0 + [an cos(nπx/L) + bn sin(nπx/L)],
2 n=1

in which an and bn are the Fourier coefficients for Go (x) on this interval.
Now example these coefficients. First,
1 L
Z
a0 = Go (x) dx = 0
L −L
because Go (x) is an odd function on [−L, L]. For n = 1, 2, · · · , Go (x) cos(nπx/L)
is also odd (product of an even and an odd function), so
1 L
Z
an = Go (x) cos(nπx/L) dx = 0.
L −L
Therefore the Fourier expansion of Go (x) contains only sine terms. Fur-
ther,
1 L
Z
bn = Go (x) sin(nπx/L) dx
L −L
2 L
Z
= Go (x) sin(nπx/L) dx
L 0
2 L
Z
= g(x) sin(nπx/L) dx,
L 0
because Go (x) sin(nπx/L) is an even function (product of two odd func-
tions), and Go (x) = g(x) for 0 ≤ x ≤ L. This is the Fourier sine coef-
ficient of g(x) on [0, L], and the Fourier series of Go (x) on [−L, L] yields
the Fourier sine expansion of g(x) on [0, L].
13. Use an argument similar to that made for Problem 12, except now the
Fourier coefficients of Ge (x) satisfy
−L L
Z
an = Ge (x) cos(nπx/L) dx
L −L
2 L
Z
= g(x) cos(nπx/L) dx
L 0
and
1 L
Z
bn = Ge (x) sin(nπx/L) dx = 0
L −L
because Ge (x) cos(nπx/L) is even and Ge (x) sin(nπx/L) is odd, and Ge (x) =
g(x) for 0 ≤ x ≤ L.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 430 CHAPTER 17. FOURIER SERIES

17.3 Integration and Differentiation of Fourier

Series
1. The Fourier expansion of f (x) on [−π, π] is
∞ 
(−1)n − 1 (−1)n

1 X
π+ cos(nx) + sin(nx) .
4 n=1
n2 π n
Because f is continuous and piecewise smooth on [−π, π], this series con-
verges to f (x) for −π < x < π.
f (x) satisfies the conditions of the theorem on term by term integration,
so this series can be integrated term by term to obtain
Z x
π
f (ξ) dξ = (x + π)
−π 4
∞ 
(−1)n

X 1 n 1
+ ((−1) − 1) sin(nx) + cos(nx) − 2 .
n=1
n3 π n2 n

2. The function is continuous and piecewise smooth on [−1, 1]. Further,

f 00 (x) exists on (−1, 1) and f (−1) = f (1) and f 00 (x) exists on (−1, 0) and
(0, 1). The Fourier series of f (x) on [−1, 1] is
∞
1 4 X 1
− cos((2n − 1)πx).
2 π 2 n=1 (2n − 1)2
Termwise differentiation yields the series
∞
4X 1
sin((2n − 1)πx).
π n=1 2n − 1
It is routine to check that this is the Fourier expansion, on [−1, 1], of
(
−1 for −1 < x < 0,
f 0 (x) =
1 for 0 < x < 1.

3. The Fourier expansion of f (x) on [−π, π] is

∞
1 X (−1)n
1− cos(x) − 2 cos(nx),
2 n=2
n2 − 1
converging to x sin(x) for −π ≤ x ≤ π. The function is continuous on
[−π, π], and f 0 (x) is piecewise continuous. Further, f (−π) = f (π) and
f 00 (x) exists on (−π, π). We can differentiate the Fourier series term by
term to obtain, for −π < x < π,
f 0 (x) = sin(x) + x cos(x)
∞
1 X n(−1)n
= cos(x) + 2 cos(nx).
22 n=2
n2 − 1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.3. INTEGRATION AND DIFFERENTIATION OF FOURIER SERIES431

4. The Fourier expansion of x2 on [−3, 3] is

∞
36 X (−1)n
3+ cos(nπx/3).
π 2 n=1 n2

This converges to x2 for −3 ≤ x ≤ 3. Further, f (3) = f (−3). f is

continuous and piecewise smooth, and f 00 exists on (−3, 3). Term by term
differentiation yields
∞
12 X 1
2x = − sin(nx)
π n=1 n

for −3 < x < 3. It is routine to verify this by expanding 2x in a Fourier

series on [−3, 3].
5. Let the Fourier coefficients of f on [−L, L]be an and bn , as usual. Let the
Fourier coefficients of f 0 (x) be An , Bn . Notice that
Z L
2
A0 = f 0 (ξ) dξ = f (L) − f (L ) = 0
L −L

because f (L) = f (−L).

Now we will develop some inequalities aimed at showing uniform conver-
gence of the Fourier series of f (x) on [−L, L].
Begin with an integration by parts:

1 L 0
Z
An = f (ξ) cos(nπξ/L) dξ
L −L
Z L
1 L n1
= [f (x) cos(nπx/L)]−L + f (ξ) sin(nπξ/L) dξ.
L πL −L

Now,
f (L) cos(nπ) − f (−L) cos(−nπ) = 0
for integer n, again because f (L) = f (L ). Then
Z L
nπ 1 nπ
An = f (ξ) cos(nπξ/L) dξ = an
L L −L L

for n = 1, 2, · · · . A similar integration by parts gives us

nπ
Bn = − an .
L
Observe that
 2
1 2 1
0≤ |An | − = A2n − |An | + 2
n n n

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 432 CHAPTER 17. FOURIER SERIES

and, similarly,
2 1
0 ≤ Bn2 − |Bn | + 2 .
n n
Add these two inequalities to get
2 2
(|An | + |Bn |) ≤ A2n + Bn2 + 2 .
n n
Multiply this by 1/2 to obtain
1 1 1
(|An | + |Bn |) ≤ (A2n + Bn2 ) + 2 .
n 2 n
On the left, insert

nπ|an | nπ|an |
|An | = and |Bn | =
L L
to obtain
L 2 L 1
|an | + |bn | ≤ (A + Bn2 ) + .
2π n π n2
Now,
∞
X ∞
X
A2n and Bn2
n=1 n+1

both converge, by Bessel’s inequality. Therefore, by the comparison test

for nonnegative series, we conclude that
∞
X
(|an | + |bn |)
n=1

converges. Finally, observe that, on [−L, L],

|an cos(nπx/L)| + bn sin(nπx/L)| ≤ |an | + |bn |.

By what is often known as the Weierstrass M - test (in this case with

Mn = |an | + |bn |

the Fourier series of f (x) on [−L, L] converges uniformly on this interval.

17.4 Properties of Fourier Coefficients

1. The argument is like that for sine series, but is notationally a little messier
because of the additional constant term in the cosine expansion. Let
N
1 X
SN = A0 + An cos(nπx/L).
2 n=1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.4. PROPERTIES OF FOURIER COEFFICIENTS 433

Now
Z L
2
0≤ (g(x) − SN (x)) dx
0
Z L
(g(x))2 dx − 2g(x)SN (x) + SN
2


= (x) dx
0
N
!
Z L Z L
2 1 X
= (g(x)) dx − 2 g(x) A0 + An cos(nπx/L) dx
0 0 2 n=1
Z L N
! N
!
1 X 1 X
+ A0 + An cos(nπx/2) A0 + An cos(nπx/L) dx
0 2 n=1
2 n=1
Z L Z L
= (g(x))2 dx − A0 g(x) dx
0 0
N Z
X L
−2 g(x)An cos(nπx/L) dx
n=1 0
N N N
!
Z L
1 2 XX X
+ A + An Am cos(nπx/L) cos(mπx/L) + A0 An cos(nπx/L) dx
0 4 0 n=1 m=1 n=1
Z L N
L 2 X
= (g(x))2 dx − A0 − L A2n
0 2 n=1
N
L 2 XL 2
+ A0 + A
4 n=1
2 n
Z L N
L LX 2
= (g(x))2 − A20 − A .
0 4 2 n=1 n

Here we have used the fact that

Z L
cos(nπx/L) dx = 0
0

and that
(
Z L
L/2 if n = m,
cos(nπx/L) cos(mπx/L) dx =
0 0 6 m,
if n =

Upon rearranging terms in the first and last lines (which are connected by
an inequality), we have

N
2 L
Z
1 2 X 2
A0 + An ≤ (g(x))2 dx.
2 n=1
L 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 434 CHAPTER 17. FOURIER SERIES

2. Use Bessel’s inequality to conclude that

N
X ∞
X
A2n and Bn2
n=1 n=1

converge. Then
lim A2n = lim Bn2 = 0.
n→∞ n→∞
Then
lim An = lim Bn = 0.
n→∞ n→∞
Upon inserting the integral expressions for An and Bn , we obtain Rie-
mann’s lemma.
Problems 3 and 4 obtained by adapting the argument of the text to the
notation cosine expansions on [0, L] and Fourier series on [−L, L], similar to the
solution of Problem 1.

17.5 Phase Angle Form

1.
(αf + βg)(x + p) = αf (x + p) + βg(x + p)
= αf (x) + βg(x) = (αf + βg)(x).

2.
g(x + p/α) = f (α(x + p/α)) = f (αx + p) = f (αx) = g(x)
and
h(x + αp) = f ((x + αp)/α) = f (x/α + p) = f (x/α) = h(x).

3.
f (x + p + h) − f (x + p)
f 0 (x + p) = lim
h→0 h
f (x + h) − f (x)
= lim = f 0 (x).
h→0 h
4. Expand f (x) in a Fourier series on [0, 2] to get
∞
2X1
1− sin(nπx).
π n=1 n

One way to obtain this expansion using the standard formulas for the
Fourier coefficients on an interval [−L, L] is write f (x), which has period
2, on the interval [−1, 1]:
(
2 + x for −1 ≤ x < 0,
f (x) =
x for 0 < x < 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.5. PHASE ANGLE FORM 435

One way to write the phase angle form is to use the identity
 π
sin(nπx) = cos nπx −
2
to obtain
∞
2X1  π
1− cos nπx − .
π n=1 n 2
The amplitude spectrum points are

(0, 1) and (nπ, −1/nπ) for n = 1, 2, · · · .

5. The Fourier series is

∞
1 2X 1
+ sin((2n − 1)πx),
2 π n=1 2n − 1

with phase angle form

∞
2X 1  π
1+ cos (2n − 1)πx − .
π n=1 2n − 1 2

Points of the amplitude spectrum are

(0, 1), (nπ, 1/((2n − 1)π)) for n = 1, 2, · · · .

6. The Fourier series is

∞  
48 X 1 π
16 + cos(nπx/2) − sin(nπx/2)
π 2 n=1 n2 n

and the phase angle form is

∞ √
48 X 1 + n2 π 2  nπx 
16 + 2 cos + arctan(nπ) .
π n=1 n2 2

Points of the amplitude spectrum are

√ !
nπ 24 1 + n2 π 2
(0, 16), , .
2 n2 π 2

7. The Fourier series is

∞
19 X 2
+ αn cos(nπx/2) + βn sin(nπx/2),
8 n=1
n2 π 2

where
αn = nπ sin(3nπ/2) + cos(3nπ/2)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 436 CHAPTER 17. FOURIER SERIES

and
nπ
βn = sin(3nπ/2) − − nπ cos(3nπ/2).
2
The phase angle form is
∞
19 1 X  nπx 
+ 2 dn cos + δN ,
8 π n=1 2

where
p
dn = 8 + 5n2 π 2 − 12nπ sin(3nπ/2) + 4(n2 π 2 − 2) cos(3nπ/2)

and  
nπ/2 + nπ cos(3nπ/2) − sin(3nπ/2)
δn = arctan .
nπ sin(3nπ/2) + cos(3nπ/2) − 1
8. The Fourier series is
∞
8X n
2
sin(2nπx)
π n=1 4n − 1

and the phase angle form is

∞
8X n  π
cos 2nπx − .
π n=1 4n2 − 1 2

9. We can write (
x for 0 ≤ x < 1,
f (x) =
x − 2 for 1 < x < 2
and f (x + 2)f (x), so f has period 2. The Fourier series is
∞
2 X (−1)n+1
sin(nπx).
π n=1 n

The phase angle form is

∞
2X1  π
cos nπx + (−1)n+1 .
π n=1 n 2

10. We can write (

−kx for −1 ≤ x < 0,
f (x) =
kx for 0 ≤ x < 1,
with f (x + 2) = f (x). The Fourier series is
∞
k X 2k
+ ((−1)n − 1) cos(nπx).
2 n=1 n2 π 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.6. COMPLEX FOURIER SERIES 437

The phase angle form is

∞
k 2k X
+ 2 cos((2n − 1)πx − π).
2 π n=1

11. We can write 

1 for 0 ≤ x < 1,

f (x) = 2 for 1 < x < 3,

1 for 3 < x < 4


with f (x + 4) = f (x). The Fourier series is

∞
3 X (−1)n
+ cos((2n − 1)πx/2).
2 n=1 2n − 1

The phase angle form is

∞
3 2X  πx π 
+ cos (2n − 1) + (1 − (−1)n ) .
2 π n=1 2 2

12. Write (
k for 0 ≤ x < 1,
f (x) =
0 for 1 < x < 2
with f (x + 2) = f (x). The Fourier series is
∞
k 2k X 1
+ sin((2n − 1)πx).
2 π n=1 2n − 1

The phase angle form is

∞
k 2k X 1  π
+ cos (2n − 1)πx − .
2 π n=1 2n − 1 2

17.6 Complex Fourier Series

1. Compute Z 3
1
d0 = 2t dt = 3
3 0
and Z 3
1 3
dn = 2xe2nπix/3 dx = i.
3 0 nπ
The complex Fourier series of f (x) is
∞ 2nπix/3
3i X 1/n
3+
π e
n=−∞,n6=0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 438 CHAPTER 17. FOURIER SERIES

converging to (
3 for x = 0 or x = 3,
2x for 0 < x < 3.
Points of the frequency spectrum are
 
2nπ 3
(0, 3), , ,
3 nπ

in which n is a nonzero integer.

2. The complex Fourier series of f is

∞  
4 X 2 2i
+ − enπix .
3 n2 π 2 nπ
n=−∞,n6=0

This converges to
(
2 for x = 0 and for x = 2,
x2 for 0 < x < 2.

Points of the frequency spectrum are

r !
1 1
(0, 4/3), nπ, 2 4 4
+ 2 2 .
n π n π

3. The complex Fourier series of the function is

∞
3 1 X 1
− (sin(nπ/2) + (cos(nπ/2) − 1)i)enπix/2 ,
4 2π n
n=−∞,n6=0

converging to 
1/2
 for x = 0, 1 or 4,
0 for 0 < x < 1,

1 for 1 < x < 4.


Points of the frequency spectrum are

 
nπ 1
q
2 2
(0, 3/4), , sin (nπ/2) + (cos(nπ/2) − 1) .
2 2nπ

4. The complex series is

∞
1 3i X 1 nπix/3
− − e ,
2 π n
n=−∞,n6=0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.6. COMPLEX FOURIER SERIES 439

converging to
(
−2 for x = 0 and for x = 6,
1 − x for 0 < x < 6.

Points of the frequency spectrum are

 
nπ 3
(0, 1/2), , .
2 nπ

5. The complex Fourier series is

1 3i X
+ e(n−1)πix/2 ,
2 π
n=−∞,n6=0

converging to

1/2 for x = 0, 2, 4,

−1 for 0 < x < 2,

2 for 2 < x < 4.


Points of the frequency spectrum are

 
nπ 3
(0, 1/2), , .
2 (2n − 1)π

6. The complex Fourier series is

∞
X 1 − e−5 2nπix/5
e ,
n=−∞
5 + 2nπi

converging to
2nπ 1 − e−5 p
 
, √ 2 2
25 + 4n π .
3 29

7. The complex Fourier series is

∞
1 2 X 1
− 2 e(2n−1)πix ,
2 π (2n − 1)2
n=−∞,n6=0

converging to f (x) for 0 ≤ x ≤ 2. Points of the frequency spectrum are

 
2 1
(0, 1/2), nπ, 2 .
π (2n − 1)2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 440 CHAPTER 17. FOURIER SERIES

17.7 Filtering of Signals

1. The complex Fourier coefficients of f are d0 = 0 and, for n 6= 0,
Z 0 Z 2 
1 −nπit/2 −nπit/2 i
dn = −e dt + e dt = ((−1)n − 1).
4 −2 0 nπ
The complex Fourier series is
∞
X i
((−1)n − 1)enπit/2 .
nπ
n=−∞,n6=0

After some routine calculation using Euler’s formula, we obtain the series
∞
4X 1
sin((2n − 1)πt/2).
π n=1 2n − 1

The N th partial sum is

N
4X 1
SN (t) = sin((2n − 1)πt/2).
π n=1 2n − 1

The N th Cesáro sum is formed by inserting factors 1 − |n|/N :

N  
4X 2n − 1 1
σN (t) = 1− sin((2n − 1)πt/2).
π n=1 N 2n − 1

2. The N th partial sum of the Fourier series of f is

N
13 X
SN (t) = + (an cos(nπt/2) + bn sin(nπt/2)),
8 n=1

where
2
an = − (−nπ sin(nπ/2) + cos(nπ/2)(−4 − n2 π 2 ) + 4(−1)n )
n3 π 3
and
2
bn = (sin(nπ/2)(4 − n2 π 2 ) + nπ cos(nπ/2) + 4(−1)n ).
n3 π 3
σN (t) is obtained by putting a factor of 1 − |n|/N into SN (t).
3. We obtain
N
X 2
SN (t) = (cos(nπ/2) − (−1)n ) sin(nπt)
n=1
nπ

and
N 
X n 2
σN (t) = 1− (cos(nπ/2) − (−1)n ) sin(nπt).
n=1
N nπ

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 17.7. FILTERING OF SIGNALS 441

4. The N th partial sums are

1
SN (t) = sin(3)
6
N
X −1
+ [3 sin(3)(−1)n cos(nπt/3) + nπ(−1 + (−1)n ) cos(3) sin(nπt/3)]
n=1
n2 π 2 − 9

σN (t) is obtained by putting a factor of 1 − n/N ) into the summation of

SN (t).
5. We find that
N 
17 X 1 − (−1)n 5 − 6(−1)n

SN (t) = + cos(nπt) + sin(nπt)
4 n=1
n2 π 2 nπ

and
N
n  1 − (−1)n 5 − 6(−1)n
 
17 X 
σN (t) = + 1− cos(nπt) + sin(nπt)
4 n=1
N n2 π 2 nπ

6. The partial sum of the Fourier series is

N
X 2
SN (t) = (1 − (−1)n ) sin(nπt/2).
n=1
nπ

The Cesáro, Hamming and Gaussian filtered partial sums are, respectively,
N
X 2  n
σN (t) = 1− (1 − (−1)n ) sin(nπt/2),
n=1
nπ N

N
X 2
HN (t) = (0.54 + 0.46 cos(nπ/N ))(1 − (−1)n ) sin(nπt/2),
n=1
nπ
N
2
π 2 /N 2
X
GN (t) = e−n (1 − (−1)n ) sin(nπt/2).
n=1

Here α = 1 has been used in the Gauss filter.

7. The partial sums are
N
X 2
SN (t) = 1 + (1 − 3(−1)n ) sin(nπt/2),
n=1
nπ

N
X 2  n
σN (t) = 1 + 1− (1 − 3(−1)n ) sin(nπt/2),
n=1
nπ N

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 442 CHAPTER 17. FOURIER SERIES

2
HN (t) = 1 + (0.54 + 0.46 cos(nπ/N ))(1 − 3(−1)n ) sin(nπt/2),
nπ
N
X 2 −n2 π2 /N 2
GN (t) = 1 + e (1 − 3(−1)n ) sin(nπt/2).
n=1
nπ

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 18

Fourier Transforms

18.1 The Fourier Transform

1. Z 0 Z 1
2i
fb(ω) = e−iωx dx + e−iωx dx = (cos(ω) − 1).
−1 0 ω
The amplitude spectrum is the graph of
2
|fb(ω)| = (cos(ω) − 1) .
ω

2. Write
f (x) = sin(x)[H(x + k) − H(x − k)]
with H the Heaviside function. Now use the modulation theorem to write
 
i 2 sin(k(ω + 1)) 2 sin(k(ω − 1))
fb(ω) = −
2 ω+1 ω−1
 
sin(k(ω + 1)) sin(k(ω − 1))
= − i.
ω+1 ω−1

The amplitude spectrum is a graph of

 
sin(k(ω + 1)) sin(k(ω − 1))
|fb(ω)| = − .
ω+1 ω−1

3. Write
f (x) = 5[H(x + 4 − 7) − H(x + 4 + 7)]
to obtain
 
2 sin(4ω) 10 −7iω
fb(ω) = 5e−7iω = e sin(4ω).
ω ω

443

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 444 CHAPTER 18. FOURIER TRANSFORMS

The amplitude spectrum is a graph of

10
|fb(ω)| = sin(4ω) .
ω
4. Use time shifting to write
r
π −ω2 /12 −5iω
f (ω) = 5
b e e .
3
The amplitude spectrum is a graph of
r
π −ω2 /12
|f (ω)| = 5
b e .
3

5.
Z ∞
fb(ω) = e−x/4 e−iωx dx
k
e−(iω+1/4)x ∞ 4e−(iω+1/4)k
= = .
−(iω + 1/4) k 1 + 4iω
The amplitude spectrum is a graph of
4e−k/4
|fb(ω)| = √ .
1 + 16ω 2

6.
2
fb(ω) = 3 (k 2 ω 2 sin(kω) + 2kω cos(kω) − 2 sin(kω)).
ω
The amplitude spectrum is a graph of |fb(ω)|.
7.
fb(ω) = πe−|ω| .
The amplitude spectrum is a graph of this function, which is nonnegative
and hence equals its own magnitude.
8. Write
f (x) = 3e−6 H(x − 2)e−3(x−2)
to obtain
e−2iω
 
fb(ω) = 3e−6 .
3 + iω
We can also write
1
fb(ω) = e−2(3+iω) .
3 + iω
The amplitude spectrum is a graph of
3e−6
|fb(ω)| = √ .
9 + ω2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 18.1. THE FOURIER TRANSFORM 445

9.
24
e2iω .
fb(ω) =
16 + ω 2
The amplitude spectrum is a graph of
24
|fb(ω)| = .
16 + ω 2

10. First write

f (x) = e−6 H(x − 3)e−2(x−3)
to obtain
e−3(2+iω)
fb(ω) = .
2 + iω
The amplitude spectrum is a graph of

e−6
|fb(ω)| = √ .
4 + ω2

11. r
2 −4ix −8x2
f (x) = 18 e e
π
12. Write
e−4(ω−5i)
fb(ω) =
3 + (ω − 5)i
to get
e−4iω
 
5ix b−1
f (x) = e f = e5ix H(x − 4)e−3(x−4) .
3 + iω
13. Write
e2(ω−3)i
fb(ω) =
5 + (ω − 3)i
to obtain
e2iω
 
3ix b−1
f (x) = e f
5 + iω
= e3ix H(x + 2)e−5(x+2) = H(x + 2)e−(10+(5−3i)x) .

14. Write
10 sin(3ω) 10 sin((ω + π))
fb(ω) = =− .
ω+π ω+π
Then
 
2 sin(3ω)
f (x) = −5e−πix fb−1
ω
= e−πix [H(x + 3) − H(x − 3)].

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 446 CHAPTER 18. FOURIER TRANSFORMS

15. Write
1 + iω 2 2
fb(ω) = = − .
(3 + iω)(2 + iω) 3 + iω 2 + iω
Then
f (x) = (2e−3x − e−2x )H(x).

16.
 
1
fb−1 = H(x)e−x ∗ H(x)e−2x
(1 + iω)(2 + iω)
Z ∞
= H(ξ)e−ξ H(x − ξ)e−2(x−ξ) dξ
−∞
Z x
= H(x)e−2x eξ dξ = H(x)e−2x (ex − 1)
0
= H(x)(e−x − e−2x ).

17.
 
1
fb−1 = H(x)e−x ∗ H(x)e−x
(1 + iω)2
Z ∞
= H(ξ)e−ξ H(x − ξ)e−(x−ξ) dξ
−∞
Z x
= H(x)e−x dξ = H(x)xe−x .
0

18.
 
sin(3ω) 1
fb−1 = [H(x + 3) − H(x − 3)] ∗ H(x)e−2x
(2 + iω)ω 2
1 ∞
Z
= (H(x + 3) − H(x − 3))H(x − ξ)e−2(x−ξ) dξ
2 −∞
 Z x Z x 
1 −2x −2ξ 2ξ
= e H(x + 3) e dξ − H(x − 3) e dξ
2 −3 3
1 1
= (1 − e−2(x+3) )H(x + 3) − (1 − e−2(x−3) H(x − 3).
4 4

19. Compute
Z ∞ Z ∞ Z ∞
1 1
|f (x)|2 dx = fb(ω)fb(ω) dω = |fb(ω)|2 dω.
−∞ 2π −∞ 2π −∞

20. One way to compute the energy is to start with

1
fb[H(x)e−2x ](ω) =
2 + iω

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 18.1. THE FOURIER TRANSFORM 447

and use Parseval’s theorem (Problem 19) to get

Z ∞ Z ∞ 2
1 1
|f (x)|2 dx = dω
−∞ 2π −∞ 2 + iω
Z ∞
1 1
= dω
2π −∞ 4 + ω 2
1 ω ∞
= arctan
4π 2 −∞
1
= .
4
We could also proceed directly in this example:
Z ∞ Z ∞
1
(H(x)e−2x )2 dx = e−4x dx = .
−∞ 0 4
Here the direct computation is easier, but sometimes it is useful to be
aware of the use of Parseval’s theorem.
21. Begin with
1 3 −iωx
  Z
1
F (H(x + 3) − H(x − 3)) (ω) = e dx
2 2 −3
e3iω − e−3iω
= .
2iω
Use the symmetry property of the transform to get
 
sin(3x)
F (ω) = π[H(−ω + 3) − H(−ω − 3)]
x
= π[H(ω + 3) − H(ω − 3)].
Now use Parseval’s identity to write
Z ∞ 2 Z 3
sin(3x) 1
dx = π 2 dω = 3π.
−∞ x 2π −3

22. Let yb(ω) = fb[y(x)](ω) and transform the differential equation to obtain

yb(−ω 2 + 6iω + 5) = fb[δ(t − 3)](ω) = e−3iω .

Then
e−3iω
yb(ω) =
−ω 2
+ 6iω + 5
e−3iω
=
(1 + iω)(5 + iω)
1 e−3iω e3 iω
 
= − .
4 1 + iω 5 + iω

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 448 CHAPTER 18. FOURIER TRANSFORMS

Invert this to obtain the solution

1
y(t) = (H(x − 3)e−(x−3) − H(x − 3)e−5(x−3) ).
4
23.
Z 5
fbwin (ω) = x2 e−iωx dx
−5
2
= 3 (25ω 2 sin(5ω) + 10ω cos(5ω) − 2 sin(5ω)).
ω
Because w(x) = 1 and the support of g is [−5, 5], then tC = 0. For the
RMS bandwidth of the window function, we have
R 5 2 !1/2
−5
x dx 10
wRMS = 2 R5 =√ .
−5
dx 3

24. Compute
Z 4π
fbwin (ω) = cos(ax)e−iωx dx
−4π
2
= (ω sin(4πω) cos(4aπ) − a cos(4πω) sin(4aπ)).
ω 2 − a2
Because w(x) is constant on on [−4π, 4π], tC = 0. We also have
R 4π 2 !1/2
−4π
x dx 8π
wRMS = 2 R 4π =√ .
−4π
dx 3

25. Compute
Z 1
1
fbwin (ω) = e−x e−iωx dx = (1 − e−4(1+iω) )
0 1 + iω
1
= (1 − e−4 (cos(4ω) − i sin(4ω))(1 − iω)
1 + ω2
1 − e−4 cos(4ω) + e−4 sin(4ω)
=
1 + ω2
 −4
e sin(4ω) + (e−4 cos(4ω) − 1)ω

+i .
1 + ω2
We also have R4
x dx
tC = R0 4 =2
0
dx
and !1/2
R4
0
(x − 2)2 dx 4
wRMS = 2 R4 =√ .
0
dx 3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 18.1. THE FOURIER TRANSFORM 449

26.
Z 1
fbwin (ω) = ex sin(πx)e−iωx dx
−1
Z 1
= sin(πx)e(1−iω)x dx
−1
π[2 sinh(1)(1 + π 2 ) − 2ω 2 cos(ω) + cosh(1)ω sin(ω)]
=
(1 + (π + ω)2 )(1 + (π − ω)2 )
π[sinh(1)(2ω 2 sin(ω) − (2 + 2π 2 ) sin(ω)) + cosh(1)ω cos(ω)]
+i .
(1 + (π + ω)2 )(1 + (π − ω)2 )
Finally, compute tC = 0 and
R1 !1/2
−1
x2 dx 2
wwin = 2 R1 =√ .
−1
dx 3

27.
Z 2
fbwin (ω) = (x + 2)2 e−iωx dx
−2
4
= ((4ω 2 − 1) sin(2ω) + 2ω cos(2ω))
ω3
8i
+ 2 (2ω cos(2ω) − sin(2ω)).
ω
With w(x) = 1 and support [−2, 2], we have tC = 0. Finally,
R2 !1/2
−2
x2 dx 4
wRMS = 2 R2 =√ .
−2
dx 3

28. We have
Z 5π
fb(ω) = e−iωx dx
3π
1 5πiω
=− (e − e3πiω )
iω
2eπiω e4πiω − e−4πiω
 
=−
ω 2i
= −2eπiω sin(4πω)
= −2 cos(πω) sin(4πω) − 2i sin(πω) sin(4πω).

Finally, R 5π
x dx
tC = R3π5π = 4π
3π
dx

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 450 CHAPTER 18. FOURIER TRANSFORMS

and !1/2
R 5π
3π
(x − 4π)2 dx 2π
wRMS = 2 R 5π =√ .
3π
dx 3

18.2 Fourier Sine and Cosine Transforms

In these problems the integrations are straightforward and details are omitted.

1.
Z ∞
1
fbC (ω) = e−x cos(ωx) dx = ,
0 1 + ω2
Z ∞
ω
fbS (ω) = e−x sin(ωx) dx =
0 1 + ω2

2.
a2 − ω 2
fbC (ω) = ,
(a2 + ω 2 )2
2aω
fbS (ω) = 2
(a + ω 2 )2

3.
 
1 sin(K(ω + 1)) sin(K(ω − 1))
fbC (ω) = + for ω 6= ±1,
2 ω+1 ω−1
K 1
fbC (1) = fbC (−1) = + sin(2K)
2 2
 
ω 1 cos((ω + 1)K) cos((ω − 1)K)
fbS (ω) = − + for ω 6= ±1,
ω2 − 1 2 ω+1 ω−1
1 1
fbS (1) = (1 − cos(2K)), fbS (−1) = − (1 − cos(2K))
4 4
4.
1
fbC (ω) = (2 sin(Kω) − sin(2Kω)),
ω
1
fbS (ω) = (1 − 2 cos(Kω) + cos(2Kω))
ω
5.
 
1 1 1
fbC (ω) = + ,
2 1 + (ω + 1)2 1 + (ω − 1)2
 
1 ω+1 ω−1
fbS (ω) = +
2 1 + (ω + 1)2 1 + (ω − 1)2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 18.2. FOURIER SINE AND COSINE TRANSFORMS 451

6.
1
fbC (ω) = (cosh(2K) cos(2Kω) − cosh(K) cos(Kω))
1 + ω2
1
+ (ω sinh(2K) sin(2Kω) − ω sinh(K) sin(Kω))
1 + ω2
1
fbS (ω) = (cosh(2K) cos(2Kω) − cosh(K) sin(Kω))
1 + ω2
1
+ (−ω sinh(2K) cos(2Kω) + ω sinh(K) cos(kω))
1 + ω2

7. Suppose, for each positive number L, f (4) (x) is piecewise continuous on

[0, L], f (3) (x) is continuous, and, as x → ∞, f (j) (x) → 0 for j = 1, 2, 3.
Then we can integrate by parts four times to obtain
Z ∞
FS [f (4) (x)](ω) = f (4) (x) sin(ωx) dx
0
h i∞
= f (3) (x) sin(ωx) − ωf 00 (x) cos(ωx) − ω 2 f 0 (x) sin(ωx) + ω 3 f (x) cos(ωx)
0
Z ∞
+ ω4 f (x) sin(ωx) dx
0
= ω 4 FS (ω) − ω 3 f (0) + ωf 00 (0).

8. Assuming the same conditions as in Problem 7, four integrations by parts

give us
Z ∞
FS [f (4) (x)](ω) = f (4) (x) cos(ωx) dx
0
h i∞
= f (x) cos(ωx) + ωf 00 (x) sin(ωx) − ω 2 f 0 (x) cos(ωx) − ω 3 f (x) sin(ωx)
(3)
0
Z ∞
+ ω4 f (x) cos(ωx) dx
0
= ω 4 {C (ω) + ω 2 f 0 (0) − d(3) (0).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 452 CHAPTER 18. FOURIER TRANSFORMS

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 19

Complex Numbers and

Functions

19.1 Geometry and Arithmetic of Complex Num-

bers
1.
(3 − 4i)(6 + 2i) = (18 + 8) + (−24 + 6)i = 26 − 18i

2. √
i(6 − 2i) + |1 − i| = 6i + 2 + 2

3.
2+i 2 + i 4 + 7i 1 + 18i
= =
4 − 7i 4 − 7i 4 + 7i 65
4.
(2 + i) − (3 − 4i) (−1 + 5i)(16 − 2i) 1
= = (−3 + 41i)
(5 − i)(3 + i) (16 + 2i)(16 − 2i) 130

5.
(17 − 6i)(−3 − 12i) = (17 − 6i)(−3 + 12i) = 4 + 228i

6.
3i 3
=√
−4 + 8i 80
7.
i3 − 4i2 + 2 = −i + 4 + 2 = 6 − i

8.
(3 + i)3 = 27 + 3(32 ) + 3(3)i2 + i2 = 18 + 26i

453

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 454 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

9.
 2  2
−6 + 2i (−6 + 2i)(1 + 8i)
=
1 − 8i (1 − 8i)(1 + 8i)
(−22 − 46i)2 1
= 2
= (−1632 + 2024i)
65 4225

10.
(−1 − 8i)(2i)(4 − i) = (−3 − 8i)(2 + 8i) = 58 − 40i

In each of Problems 11–16, n denotes an arbitrary integer.

11.
π
|3i| = 3, arg(3i) = + 2nπ
2
12.
√ 3π
| − 2 + 2i| = 2 2, arg(−2 + 2i) = + 2nπ
4
13. √
| − 3 + 2i| = 13, arg(−3 + 2i) = − arctan(2/3) + (2n + 1)π

14. √
|8 + i| = 65, arg(8 + i) = arctan(1/8) + 2nπ

15.
| − 4| = 4, arg(−4) = (2n + 1)π

16.
|3 + 4i| = 5, arg(3 + 4i) = − arctan(4/3) + 2nπ
√
17. Because | − 2 + 2i| = 2 2 and 3π/34 is an argument, the polar form of
−2 + 2i is √
−2 + 2i = 2 2e3iπ/4 .
Here we did not add the customary 2nπ to the argument because, first,
we need only one argument to write the polar form, and second, e2nπi = 1
for any integer n.

18. | − 7i| = 7 and an argument of −7i is 3π/2 (or −π/2 if you prefer), so the
polar form is
−7i = 7e3πi/2 .
We could also write
−7i = 7e−πi/2 .
√
19. |5 − 2i| = 29 and an argument of 5 − 2i is − arctan(2/5), so
√
5 − 2i = 29e− arctan(2/5)i .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.1. GEOMETRY AND ARITHMETIC OF COMPLEX NUMBERS 455
√
20. | − 4 − i| = 17 and an argument of −4 − i is π + arctan(π/4) (look at
the line from the origin to (−4, −1) in the third quadrant), so
√
−4 − i = 29e(π+arctan(π/4))i .

21. √
8+i= 65earctan(1/8)i .

22. √
−12 + 3i = 153e− arctan(1/8)i .

23. Because i2 = −1, we have

i4n = (i2 )2n = ((−1)2 )n = 1,

i4n+1 = i4n i = i,
i4n+2 = i4n i2 = i2 = −1,
i4n+3 = i4n i3 = i2 i = −i.

24. Because (a + bi)2 = a2 − b2 + 2abi, we have

Re((a + bi)2 ) = a2 − b2 and Im((a + bi)2 ) = 2ab.

25. Suppose first that z, w, u form vertices of a triangle, labeled in clockwise

order around the triangle. The sides of the triangle are vectors represented
by the complex numbers w−z, u−w, and z −u. This triangle is equilateral
if and only if the sides have the same length, or

|w − z| = |u − w| = |z − u|

and each of the vector sides can be rotated by θ = 2π/3 radians clockwise
to coincide with another side. This occurs exactly when

u − w = (w − z)e−2πi/3 and z − u = (u − w)e−2πi/3 .

Dividing these equations, we have

u−w w−z
= .
z−u u−w
Then
(u − w)(u − w) = (w − z)(z − u).
Then
u2 − 2uw + w2 = wz + zu − uw − z 2 .
Then
z 2 + w2 + u2 = zw + zu + wu.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 456 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

26. Let z = x + iy. Then

z 2 = (z)2 if and only if

(x + iy)2 = (x − iy)2 if and only if
x2 − y 2 + 2ixy = x2 − y 2 − 2ixy if and only if
2ixy = −2ixy if and only if 2ixy = 0.

Now, 2ixy = 0 can happen in two ways. If x = 0, then x + iy = iy is pure

imaginary. If y = 0, then x + iy = x is real.
27. Suppose first that |z| = 1. Then
p p
|z| = x2 + (−y)2 = x2 + y 2 = |z| = 1

also. Then
z−w z−w
=
1 − zw zz − zw
|z − w|
= = 1.
|z||z − w|
If |w| = 1, then
z−w z−w
=
1 − zw ww − zw
1 z−w
= =1
|w| z − w
because
|z − w| = |w − z| = |w − z|.

28. Compute

|z + w|2 + |z − w|2
= (z + w)(z + w)(z − w)
= zz + zw + wz + ww + zz + zw − wz − zw
= 2zz + 2ww
= 2 |z|2 + |w|2 .

29. M consists of all x + iy with y < 7. This is the half-plane lying below the
horizontal line y = 7. The boundary points are all points x + 7i on the
“edge” of M . M is open because it does no contain any of its boundary
points (all points of M are interior points).
30. S consists of all points outside the circle of radius 2 about the origin. The
boundary points of S are the points |z| = 2 on the circumference of the
circle. None of these boundary points are in S. Every point of S is an
interior point, and S is open.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.2. COMPLEX FUNCTIONS 457

31. U consists of all points in the vertical strip between the vertical lines x = 1
and x = 3, including points on the line x = 3, but none of the points on
the line x = 1. The boundary points of U are the points 1 + iy and 3 + iy
on these lines. U is not closed because there are boundary points of U
that do not belong to U . U is not open because U contains some of its
boundary points (so not every point of U is an interior point).

32. V consists of all points inside the rectangle having vertices (2, 1), (3, 1),
(2, −1) and (3, −1), and the points 3 + iy for −1 < y < 1 on the right side
of this rectangle. The boundary points are all the points on the four sides
of the rectangle. Some boundary points (on the right side) are in V , and
the other boundary points (on the other three sides) are not. Therefore
V is not closed, because it does not contain all of its boundary points. V
is not open, because it does contain some of its boundary points.

33. W consists of all x+iy with x > y 2 . These are the points “enclosed” by the
parabola x = y 2 , which opens to the right from the origin. The boundary
√
points are the points on the parabola, which are the points x + i x for
x ≥ 0. W does not contain any of its boundary points, and is open. W is
not closed.

34. One way to envision R is to think of certain points on vertical segments

of length 1. Begin with n = 1. The points

1
1+ i
m
are the points 1 + i, 1 + (1/2)i, 1 + (1/3)i, · · · , going down the line x = 1
and approaching arbitrarily lose to (but not reaching) 1 on the horizontal
axis. When n = 2, we have points

1 1
+ i,
2 m
moving down from x = 1/2 and approaching arbitrarily close to x = 1/2
as m is chosen larger. And so on - as n increases, these points remain on
parallel horizontal lines, but move down vertical segments closer to the
imaginary axis.
The boundary points are all points (1/m)i and all points 1/n for positive
integer n and m. None of these boundary points belong to R, so R is not
closed. But R is open, because every point of R is an interior point.

19.2 Complex Functions

1.
f (z) = z − i = x + iy − i = x + (y − 1)i

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 458 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

so u(x, y) = x and v(x, y) = y − 1. The Cauchy-Riemann equations for

this function are
∂u ∂v
=1=
∂x ∂y
and
∂u ∂v
=0=− .
∂y ∂x
Because u and v are continuous with continuous first partial derivatives,
and the Cauchy-Riemann equations are satisfied for all z, f (z) is differen-
tiable for all z.

2.

f (z) = z 2 − iz = x2 − y 2 + 2ixy − ix + y = x2 − y 2 + y + (2xy − x)i.

Let u(x, y) = x2 − y 2 + y and v(x, y) = 2xy − x. Then

∂u ∂v
= 2x =
∂x ∂y

and
∂v ∂u
= −2y + 1 =
∂x ∂y
so the Cauchy-Riemann equations are satisfied at every point. Because
u and v are continuous with continuous partial derivatives at all points,
f (z) is differentiable for all z.
p
3. f (z) = |x + iy| = x2 + y 2 , so
p
u(x, y) = x2 + y 2 and v(x, y) = 0.

If z 6= 0, then
∂u x ∂u y
=p , =p
∂x 2
x +y 2 ∂y x + y2
2

and
∂v ∂v
= = 0.
∂x ∂y
The Cauchy-Riemann equations are not satisfied at any nonzero z. To
check what happens at z = 0, compute

∂u u(h, 0) − u(0, 0)
(0, 0) = lim
∂x h→0 h
√
h2
= lim
h→0 h
|h|
= lim .
h→0 h

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.2. COMPLEX FUNCTIONS 459

This limit does not exist, because

(
|h| 1 if h > 0,
=
h −1 if h < 0.

Similarly, (∂u/∂y)(0, 0) does not exist. Therefore the Cauchy-Riemann

equations are not satisfied at any point, including the origin, and f (z) is
not differential for any z.

4. f (z) is defined for all nonzero z. For z 6= 0,

2z + 1 1
f (z) = =2+ .
z z
Then
1
f (z) = 2 +
x + iy
x − iy
=2+ 2
x + y2
x2 y
=2+ 2 − 2 i.
x + y2 x + y2
Then
x2 y
u(x, y) = 2 + and v(x, y) = − 2 .
x2 + y2 x + y2
For z 6= 0 compute the partial derivatives

∂u y 2 − x2 ∂u −2xy
= 2 2 2
, = 2 ,
∂x (x + y ) ∂y (x + y 2 )2

and
∂v 2xy ∂v y 2 − x2
= 2 , = .
∂x (x + y 2 )2 ∂y (x2 + y 2 )2
The Cauchy-Riemann equations hold for all nonzero z. Because u, v and
its partial derivatives are continuous for all nonzero z, f (z) is differentiable
for all z 6= 0.

5. f (z) = i|z|2 = (x2 + y 2 )i, so

u(x, y) = 0 and v(x, y) = x2 + y 2 .

Then
∂u ∂u
= =0
∂x ∂y
and
∂v ∂v
= 2x, = 2y.
∂x ∂y

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 460 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

The Cauchy-Riemann equations are satisfied only at z = 0, so f (z) is

certainly not differentiable at any nonzero z. To check at z = 0, fall back
on the definition of the derivative:
z(h) − f (0) i|h|2 ihh
lim = lim = lim = lim ih = 0.
h→0 h h→0 h h→0 h h→0

Therefore f 0 (0) = 0. 0 is the only point at which this function is differen-

tiable.
6. f (z) = x + iy + y = (x + y) + yi so let

u(x, y) = x + y and v(x, y) = y.

The Cauchy-Riemann equations fail to hold at any point, because

∂u ∂v
= 1 but = 0.
∂y ∂x

7. First,
x + iy y
f (z) = =1+ i
x x
for x 6= 0. This function is defined for all z except for points on the
imaginary axis. For x 6= 0, we can let
y
u(x, y) = 1, v(x, y) = .
x
Now
∂u ∂u
= =0
∂x ∂y
and
∂v y ∂v 1
= − 2, = .
∂x x ∂y x
These are not satisfied at any z at which the function is defined. Therefore
f (z) is not differentiable at any point at which it is defined.
8. Write

f (z) = (x + iy)3 − 8(x + iy) + 2

= x3 − 3xy 2 − 8x + 2 + (3x2 y − y 3 − 8y)i.

Let

u(x, y) = x3 − 3xy 2 − 8x + 2 and v(x, y) = 3x2 y − y 3 − 8y.

Then
∂u ∂v
= 3x2 − 3y 2 − 8 =
∂x ∂y

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.2. COMPLEX FUNCTIONS 461

and
∂u ∂v
= −6xy = − .
∂y ∂x
The Cauchy-Riemann equations are satisfied at every z. Further, u, v
and their first partial derivatives are continuous for all (x, y), so f (z) is
differentiable for all z.

9. First,
f (z) = (z)2 = (x − iy)2 = x2 − y 2 − 2xyi,

so let
u(x, y) = x2 − y 2 and v(x, y) = −2xy.

Then
∂u ∂v
= 2x but = −2x,
∂x ∂y
while
∂v ∂u
= −2y = .
∂x ∂y
The Cauchy-Riemann equations hold only a z = 0, so this is the only
point at which f (z) might have a derivative. To check this, look at

(h)2
 
f (h) − f (0) h
lim = lim = lim h=0
h→0 h h→0 h h→0 h

because h/h has magnitude 1 and h → 0 if h → 0.

Therefore f 0 (0) = 0, and 0 is the only point at which the function has a
derivative.

10. If z = x + iy, then

p
f (z) = iz + |z| = x2 + y 2 − y + xi.

Let
p
u(x, y) = x2 + y 2 − y and v(x, y) = x.

Then
∂u x ∂u y
=p , = −1 + p ,
∂x x + y ∂y
2 2 x + y2
2

while
∂v ∂v
= 1, = 0.
∂x ∂y
Using these, it is routine to check that f (z) is not differentiable at any z.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 462 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

11. For z 6= 0, write

1 1
f (z) = −4z + = −4x − 4iy +
z x + iy
x − iy
= −4x − 4yi + 2
x + y2
for (x, y) 6= (0, 0). Let
 
x y
u(x, y) = −4x + 2 and v(x, y) = −4y − .
x + y2 x2 + y 2
Then
∂u y 2 − x2 ∂u −2xy
= −4 + 2 , = 2 ,
∂x (x + y 2 )2 ∂y (x + y 2 )2
and
∂v 2xy ∂v y 2 − x2
= 2 , = −4 + .
∂x (x + y 2 )2 ∂y (x2 + y 2 )2
The Cauchy-Riemann equations are satisfied at each nonzero z. Because
u, v and the partial derivatives are continuous for (x, y) 6= (0, 0), f (z) is
differentiable for all nonzero z.
12. First,
z−i x + (y − 1)i
f (z) = =
z+i x + (y + 1)i
x2 + y 2 − 1 − 2xi
= .
x2 + (y + 1)2
Let
x2 + y 2 − 1 2x
u(x, y) = 2 2
and v(x, y) = − 2 .
x + (y + 1) x + (y + 1)2
for all (x, y) with x2 + (y + 1)2 6= 0. The partial derivatives are
∂u 4x(y + 1)
= 2 ,
∂x x + (y + 1)2
∂u 2(y + 1)2 − 2x2
= 2 ,
∂y x + (y + 1)2
∂v 2x2 − 2(y + 1)2 ∂v 4x(y + 1)
= 2 2
, = 2 .
∂x x + (y + 1) ∂y x + (y + 1)2
For (x, y) with x2 + (y − 1)2 6= 0, the Cauchy-Riemann equations are sat-
isfied. Further, u(x, y) and v(x, y) are continuous with continuous partial
derivatives. Therefore f (z) is differentiable at all z with z 6= i.
13. Let zn = xn + iyn and z0 = x0 + iy0 . Write f (z) = u(x, y) + iv(x, y).
Because u and v are continuous at (x0 , y0 ), then
f (zn ) = u(xn , yn ) + iv(xn , yn ) → u(x0 , y0 ) + iv(x0 , y0 ) = f (z0 ).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.3. THE EXPONENTIAL AND TRIGONOMETRIC FUNCTIONS 463

19.3 The Exponential and Trigonometric Func-

tions
1.
ei = e0+i = e0 (cos(1) + i sin(1)) = cos(1) + i sin(1)

2. There are several ways to proceed. One is to use

sin(x + iy) = sin(x) cosh(y) + i cos(x) sinh(y)

to get
sin(1 − 4i) = sin(1) cosh(4) − i cos(1) sinh(4).

We could also have begun with the definition of sin(z) and used Euler’s
formula.

3. Use the fact that

cos(x + iy) = cos(x) cosh(y) − i sin(x) sinh(y)

to get
cos(3 + 2i) = cos(3) cosh(2) − i sin(3) sinh(2).

4. From their definitions, the trigonometric and hyperbolic functions are re-
lated by
sin(z) = −i sinh(iz) and cos(z) = cosh(iz).

Then

sin(3i) −i sinh(−3)
tan(3i) = =
cos(3i) cosh(−3)
i sinh(3)
= .
cosh(3) = i tanh(3)

5.
e5+2i = e3 e2i = e3 cos(2) + ie3 sin(2).

6.

cos(1 − πi/4)
cot(1 − πi/4) =
sin(1 − πi/4)
cos(1) cosh(π/4) + i sin(1) sinh(π/4)
= .
sin(1) cosh(π/4) − i cos(1) sinh(π/4)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 464 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

To identify the real and imaginary parts of this quotient, multiply the
numerator and denominator by the conjugate of the denominator to obtain

cot(1 − iπ/4)
(cos(1) cosh(π/4) + i sin(1) sinh(π/4))(sin(1) cosh(π/4) + i cos(1) sinh(π/4))
=
sin2 (1) cosh2 (π/4) + cos2 (1) sinh2 (π/4)
sin(1)cos(1)(cosh2 (π/4) − sinh2 (π/4)) + i sinh(π/4) cosh(π/4)(sin2 + cos2 (1))
=
sin2 (1) cosh2 (π/4) + cos2 (1) sinh2 (π/4)
sin(1) cos(1) + i sinh(π/4) cosh(π/4)
= .
sin2 (1) cosh2 (π/4) + cos2 (1) sinh2 (π/4)

7.

1
sin2 (1 + i) = (1 − cos(2(1 + i)))
2
1
= [1 − cos(2) cosh(2) + i sin(2) sinh(2)].
2

8.

cos(2 − i) − sin(2 − i)
= cos(2) cosh(1) + i sin(2) sinh(1) − sin(2) cosh(1) − i cos(2) sinh(1)
= cosh(1)[cos(2) − sin(2)] − i sinh(1)[cos(2) − sin(2)]

9.
eiπ/2 = cos(π/2) + i sin(π/2) = i

10.

sin(ei ) = sin(cos(1) + i sin(1))

= sin(cos(1)) cosh(sin(1))i cos(cos(1)) sinh(sin(1))

11. Begin with

2 2
−y 2 +2ixy
ez = ex
2
−y 2
= ex [cos(2xy) + i sin(2xy)].

Then
2
−y 2 2
−y 2
u(x, y) = ex cos(2xy) and v(x, y) = ex sin(2xy).

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.3. THE EXPONENTIAL AND TRIGONOMETRIC FUNCTIONS 465

Now compute
∂u 2
−y 2
= ex [2x cos(2xy) − 2y sin(2xy)],
∂x
∂u 2
−y 2
= ex [−2y cos(2xy) − 2x sin(2xy)],
∂y
∂v 2
−y 2
= ex [2x sin(2xy) + 2y sin(2xy)],
∂x
∂v 2
−y 2
= ex [−2y sin(2x) + 2x cos(2xy)].
∂y

Then u and v satisfy the Cauchy-Riemann equations for all (x, y).
12. Begin by writing
1 1 x y
= = 2 − 2 i.
z x + iy x + y2 x + y2
Then
    
1/z x/(x2 +y 2 ) y y
e =e cos − i sin
x2 + y 2 x2 + y 2
= u(x, y) + iv(x, y).

Take the partial derivatives:

2 2
ex/(x +y )
    
∂u 2 2 y y
= 2 (y − x ) cos + 2xy sin ,
∂x (x + y 2 )2 x2 + y 2 x2 + y 2
2 2
ex/(x +y )
    
∂u y 2 2 y
= 2 −2xy cos − (x − y ) sin ,
∂y (x + y 2 )2 x2 + y 2 x2 + y 2
2 2
ex/(x +y )
    
∂v 2 2 y y
= 2 (x − y ) sin + 2xy cos ,
∂x (x + y 2 )2 x2 + y 2 x2 + y 2
2 2
ex/(x +y )
    
∂v y 2 2 y
= 2 2xy cos − (x − y ) cos .
∂y (x + y 2 )2 x2 + y 2 x2 + y 2
It is easy to verify that u and v satisfy the Cauchy-Riemann equations for
all (x, y).
13.

f (z) = zez = (x + iy)ex (cos(y) + i sin(y))

= xex cos(y) − yex sin(y) + (yex cos(y) + xex sin(y))i = u(x, y) + iv(x, y).

Then
∂u ∂v
= ex [cos(y) + x cos(y) − y sin(y)] =
∂x ∂y

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 466 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

and
∂u ∂v
= ex [−x sin(y) − sin(y) − y cos(y)] = − .
∂y ∂x
The Cauchy-Riemann equations are satisfied for all z.
14.
1
f (z) = cos2 (z) = (1 − cos(2z))
2
1 1
= − (cos(2x) cosh(2y) − i sin(2x) sinh(2y)).
2 2
Then
1 1 1
u(x, y) = − cos(2x) cosh(2y), v(x, y) = sin(2x) sinh(2y).
2 2 2
Now,
∂u
= sin(2x) cosh(2y),
∂x
∂u
= − cos(2x) sinh(2y),
∂y
∂v
= cos(2x) sinh(2y),
∂x
∂v
= sin(2) cosh(2y).
∂y
The Cauchy-Riemann equations are satisfied at every (x, y).
15. Suppose ez = 2i. With z = x + iy, then

ex cos(y) + iex sin(y) = 2i.

Then
ex cos(y) = 0 and ex sin(y) = 2.
Because ex 6= 0, cos(y) = 0, so

(2n + 1)π
y=
2
in which n can be any integer. Now we have
 
2n + 1
ex sin π = 2.
2

Now ex > 0 for real x, so sin((2n + 1)π/2) > 0. But

  (
2n + 1 1 if n is even,
sin π =
2 −1 if n is odd.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.3. THE EXPONENTIAL AND TRIGONOMETRIC FUNCTIONS 467

Therefore n must be even, say n = 2m. Now we have

4m + 1
y= π.
2
Now we are left with ex = 2, so x = ln(2). All the solutions of ez = 2i are
4m + 1
ln(2) + π
2
with m any integer.
16. For the first identity, write

sin(z) cos(w) + cos(z) sin(w)

1  iz
(e − e−iz )(eiw + e−iw ) + (eiz + e−iz )(eiw − e−iw )

=
4i
1  i(z+w) 
= e − e−i(z+w)
2i
= sin(z + w).

The second identity is proved by a similar manipulation.

17. Use the polar form of the given equation. If z = reiθ , the equation is

ez = er eiθ = −2.

Because θ is real, |eiθ | = 1, so

|ez | = er = | − 2| = 2.

Then r = ln(2). Next, we must also have

eiθ = −1 = cos(θ) + i sin(θ).

Then sin(θ) = 0, so θ = nπ, in which (so far) n can be any integer. But
cos(θ) = −1 means that n must be odd, so

θ = (2m + 1)π

in which m can be an y integer. Then

z = ln(2) + (2m + 1)π,

with m any integer, are all the solutions for z.

18. We want all z such that sin(z) = i. We need, for z = x + iy,

sin(x) cosh(y) + i cos(x) sinh(y) = i.

Then
sin(x) cosh(y) = 0 and cos(x) sinh(y) = 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 468 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

Because cosh(y) > 0 for y real, we must have sin 9x) = 0, so x = nπ, with
n as yet any integer.
The second equation now becomes

cos(nπ) sinh(y) = 1.

Then
1
sinh(y) = = (−1)n .
cos(nπ)
Then
ey − e−y = 2(−1)n .

Write this equation as

e2y − 2(−1)n ey − 1 = 0

and consider cases. First, if n is even, then (−1)n = 1 and we have a

quadratic equation for ey :

e2y − 2ey − 1 = 0

with roots
√
ey = 1 ±
2.
√
Because 1 − 2 < 0 and ey > 0, discard this root and set
√
ey = 1 + 2.
√
Then y = ln(1 + 2) and solutions for z are
√
z = 2mπ + ln(1 + 2)i,

with m any integer. In the case that n is odd, write n = 2m + 1. Now

(−1)n = −1 and we have

e2y + 2ey − 1 = 0,
√ √
with roots −1 ± 2. Again, we can only use the positive root −1 + 2,
so in this case
√
y = ln(−1 + 2)

and solutions for z are

√
z = (2m + 1)π + ln(−1 + 2)i,

with m any integer.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.4. THE COMPLEX LOGARITHM 469

19.4 The Complex Logarithm

1. In polar form,
z = −4i = 4e3nπi/2
so π 
log(−4i) = ln(4) + + 2nπ i.
2
√
2. First, 2 − 2i = 2 2e7πi/4 , so
√
 
7π
log(2 − 2i) = ln(2 2) + + 2nπ i.
4

3. −5 = 5eπi , so
log(−5) = ln(5) + (2n + 1)πi.
√ arctan(5)i
4. 1 + 5i = 26e , so
1
log(1 + 5i) = ln(26) + (arctan(5) + 2nπ)i.
2
√
5. −9 + 2i = 85e(arctan(−2/9)+π)i , so
1
log(−9 + 2i) = ln(85) + (− arctan(2/9) + (2n + 1)π)i.
2
6. 5 is its own polar form (argument zero), so
log(5) = ln(5) + 2nπi.

7. Note that log(zw), log(z) and log(w) all have infinitely many different
values, so we cannot expect to write the complex logarithm of the product
as the sum of the logarithms of the factors. What we can show is that
every value of log(zw) is the sum of a value of log(z) and a value of log(w).
Suppose that z and w are nonzero. Let θz be any argument of z, and θw
any argument of w. Then
z = |z|e(θz +2nπ)i and w = |w|e(θw +2mπ)i .
Then
zw = |z||w|e(θz +θw +2kπ)i ,
while
log(z) + log(w) = ln(|z|) + ln(|w|) + (θz + θw + 2(n + m)π)i.
This means that for any choice of n and m, we can choose k = n + m to
obtain a value of log(zw) that is equal to log(z) + log(w).
8. The argument is nearly identical to that used in Problem 7, except now
z |z| (θz −θw +2(n−m)π)i
= e .
w |w|

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 470 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

19.5 Powers
In these problems, n denotes an arbitrary integer.

1.

i1+i = e(1+i) log(i) = e(1+i)(π/2+2nπ)i

h π  π i
= e−(π/2+2nπ) cos + 2nπ + i sin + 2nπ
2 2
= ie−(π/2+2nπ) .

2.
√
(1 + i)2i = e2i log(1+i) = e2i(ln( 2)+i(π/4+2nπ))

= e−(π/2+4nπ) [cos(ln(2)) + i sin(ln(2))].

3.

ii = ei log(i) = ei(i(π/2+2nπ))
= e−π/2+2nπ .

This is consistent with Problem 1, because i1+i = iii .

4.

(1 + i)2−i = e(2−i) log(1+i)

√
= e(2−i)(ln( 2)+i(π/4+2nπ))
h π √  π √ i
= eln(2)+π/2+2nπ cos + 4nπ − ln( 2) + i sin + 4nπ − ln( 2)
√ 2 √ 2
= 2π/4+2nπ [sin(ln( 2)) + i cos(ln( 2))].

5.

(−1 + i)−3i = e−3i log(−1+i)

√
= e−3i ln( 2)+i(3π/4+2nπ)
√ √
= e9π/4+6nπ [cos(3 ln( 2)) + i sin(3 ln( 2))].

6.
√ 1/3
(1 − i)1/3 = 2e−i(π/4+2nπ)
= 21/6 e−i(π/12+2nπ) .

We obtain the three distinct cube roots by choosing n = 0, 1, 2. Other

choices of n repeat those obtained for these values.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.5. POWERS 471

7.
 1/4
i1/4 = ei(π/2+2nπ)
= ei(π/8+nπ/2) ,
the the four fourth roots obtained for n = 0, 1, 2, 3. Other choices of n
repeat these roots.
8.
161/4 = (16e2nπi )1/4 = 2enπi/2
= 2[cos(nπ/2) + i sin(nπ/2)]
with the distinct fourth roots obtained by using n = 0, 1, 2, 3.
9.
(−4)2−i = e(2−i) log(−4) = e(2−i)(ln(4)+i(π+2nπ))
= e2 ln(4)+π+2nπ [cos(ln(4)) − i sin(ln(4))].

10.
6−2−3i = e(−2−3i) log(6)
= e(−2−3i)(ln(6)+2nπi)
= e−2 ln(6)+6nπ e−(3 ln(6)+4nπ)i
1 6nπ
= e [cos(3 ln(6)) − i sin(3 ln(6))].
36
11.
 1/4
(−16)1/4 = 16ei(π+2nπ) = 2ei(π/4+nπ/2)
h  π nπ   π nπ i
= 2 cos + + i sin + .
4 2 4 2
We obtain the four fourth roots by taking n = 0, 1, 2, 3. These fourth
roots are √ √ √ √
2(1 + i), 2(−1 + i), 2(−1 − i), 2(1 − i).

12. First compute

1+i (1 + i)(1 + i)
= = i.
1−i (1 − i)(1 + i)
We therefore want i1/3 . These roots are
 1/3
i1/3 = ei(π/2+2nπ) = ei(π/6+2nπ/3)
   
π 2nπ π 2nπ
= cos + + i sin + .
6 3 6 3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 472 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

The three roots are obtained for n = 0, 1, 2 and are

1 √ 1 √
( 3 + i), (− 3 + i), −i.
2 2
13. These are the sixth roots of unity:

1/6
11/6 = e2nπi = enπi/3
= cos(nπ/3) + i sin(nπ/3).
These sixth roots are obtained for n = 0, 1, 2, 3, 4, 5, and are
1 √ 1 √ 1 √ 1 √
1, (1 + 3i), (−1 + 3i), −1, (−1 − 3i), (1 − 3i).
2 2 2 2
14.
(7i)3i = e3i log(7i) = e3i(ln(7)+i(π/2+2nπ))
= e−2(π/2+2nπ) [cos(3 ln(7)) + i sin(3 ln(7))].

15. Let ω be any nth root of 1 different from 1. The numbers ω j , for j =
0, 1, · · · , n − 1 are distinct, hence are all of the nth roots of 1. It is
therefore enough to show that
n−1
X
ω j = 0.
j=0

But this is a finite geometric series, whose sum is known:

n−1
X 1 − ωn
ωj = =0
j=0
1−ω
n
because ω = 1.
The conclusion can alsoPbe proved as follows. Let ω1 , · · · , ωn be the nth
n
roots of unity. Let S = j=1 ωj .
Now, one of the nth roots of unity is 1, but the other n − 1 roots are
different from 1. Pick one root that does not equal 1, say, possibly by
relabeling, ω1 6= 1. The numbers
ω1 ω1 , ω1 ω2 , · · · , ω1 ωn
are also nth roots of unity and are distinct, so this list includes all the nth
roots of unity. The sum of these numbers is therefore S:
n
X
S= ω1 ωj = ω1 S.
j=1

But then
S(1 − ω1 ) = 0.
Because ω1 6= 1, then S = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 19.5. POWERS 473

16. First recall that, if a 6= 1, then

n−1
X 1 − an
aj = .
j=0
1−a

Further, replacing a with −a,

n−1 n−1
X X 1 − (−1)n an
(−a)j = (−1)j aj = .
j=0 j=0
1+a

Apply this result with a = e2π/n , so an = e2πi = 1, to obtain

n−1
(
n
X 1 − (−1) 0 if n is even,
(−1)j e2πij/n = 2πi/n
= 2πi/n
j=0
1 + e 2/(1 + e ) if n is odd.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 474 CHAPTER 19. COMPLEX NUMBERS AND FUNCTIONS

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 20

Complex Integration

20.1 The Integral of a Complex Function

1. In this problem f (z) = 1 is differentiable for all z and we can write an
antiderivative F (z) = z. The curve has initial point γ(1) = 1 − i and
terminal point γ(3) = 9 − 3i, so
Z
f (x) dz = F (9 − 3i) − F (1 − i) = 9 − 3i − (1 − i) = 8 − 2i.
γ

We can also evaluate the integral by using the parametric equations of the
curve. On γ, z = γ(t) = t2 − it, so dz = 2t − i and
Z Z
f (z) dz = dz
γ γ
Z 3
= (2t − i) dt
1
3
= t2 − it = (9 − 3i) − (1 − i)
1
= 8 − 2i.

2. f (z) = z 2 − iz is differentiable for all z, with antiderivative

1 3 i 2
F (z) = z − z .
3 2
Because the curve extends from 2 to 2i, then
Z
8 4
f (z) dz = F (2i) − F (1) = − + i.
γ 3 3

If we want to use a parametrization of the curve, we can use polar coor-

475

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 476 CHAPTER 20. COMPLEX INTEGRATION

dinates and write γ(t) = 2eit for 0 ≤ t ≤ π/2. Then

Z Z π/2
f (z) dz = (4e2it − 2ieit )(2ieit dt
γ 0
Z π/2
= (8ie3it + 4e2it ) dt
0
 π/2
8 3it
− e − 2ie2it
3 0
8 4
= − + i.
3 3
3. f (z) = Re(z) is not differentiable, so there is no antiderivative. There are
many ways to parametrize the curve. One is by setting
γ(t) = 1 + (1 + i)t for 0 ≤ t ≤ 1.
On γ, f (z) = 1 + t and dz = (1 + i) dt, so
Z Z 1
f (z) dz = (1 + t)(1 + i) dt
γ 0
Z 1
= (1 + i + (1 + i)t) dt
0
1+i 2 1
= (1 + i)t + t
2 0
3
= (1 + i).
2
4. Parametrize γ(t) = 4eit for π/2 ≤ t ≤ 3π/2. Then
Z Z 3π/2
1 1
dz = it
4ieit dt = πi.
γ z π/2 4e

5. F (z) = (z − 1)2 /2 is an antiderivative of f (z), which is differentiable for

all z, so Z
13
f (z) dz = F (1 − 4i) − F (2i) = − + 2i.
γ 2
6. F (z) = iz 3 /3 is an antiderivative of f (z), which is differentiable for all z,
so Z
1
f (z) dz = F (3 + i) − F (1 + 2i) = (−28 + 29i).
γ 3
7. f (z) is differentiable for all z and has antiderivative F (z) = − cos(2z)/2,
so
Z
f (z) dz = F (−4i) − F (−i)
γ
1 1
= − (cos(−8i) − cos(−2i)) = − [cosh(8) − cosh(2)].
2 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 20.1. THE INTEGRAL OF A COMPLEX FUNCTION 477

8. f (z) is differentiable for all z and has antiderivative F (z) = z + z 3 /3, so

Z
f (z) dz = F (3i) − F (−3i) = −12i.
γ

9. f (z) is differentiable for all z and has antiderivative F (z) = −i sin(z), so

Z
f (z) dz = F (2 + i) − F (0) = −i sin(2 + i)
γ
= −i[sin(2) cosh(1) + i cos(2) sinh(1)]
= − cos(2) sinh(1) − i sin(2) cosh(1).

10. f (z) has no antiderivative, so parametrize the curve, say by γ(t) = −4 +

(4 + i)t for 0 ≤ t ≤ 1. On the curve,
z(t) = −4 + (4 + i)t and dz = (4 + i) dt
and
|z|2 = |4(t − 1) + it|2 = 16(t − 1)2 + t2
Then
Z Z 1
2
|z| dz = (16(t − 1)2 + t2 )(4 + i) dt
γ 0
4+i
= [16(t − 1)3 + t3 ] dt
3
17
= (4 + i).
3

11. Use the antiderivative F (z) = (z − i)4 /4 to get

Z
f (z) dz = F (2 − 4i) − F (0) = 10 + 210i.
γ

12. Use the antiderivative F (z) = −ieiz to get

Z
f (z) dz = F (−4 − i) − F (−2)
γ

= −i e1−4i − e−2i

= −e sin(4) + sin(2) + [cos(2) − e cos(1)]i.

13. f (z) has no antiderivative because this function is not differentiable. Parametrize
the curve by γ(t) = (−4 + 3i)t for 0 ≤ t ≤ 1. Then
Z Z 1
iz dz = −i(4t − 3ti)(−4 + 3i) dt
γ 0
 
3 25
= (−4 + 3i) − 2i = i.
2 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 478 CHAPTER 20. COMPLEX INTEGRATION

14. f (z) = Im(z) is not differentiable, so parametrize γ(t) = 4eit for 0 ≤ t ≤

2π. Then
Z Z 2π
Im(z) dz = 4 sin(t)4ieit dt
γ 0
Z 2π
= 16(− sin2 (t) + i cos(t) sin(t)) dt = −16π.
0

2
15. f (z) = |z| has no antiderivative, so write γ(t) = (1 + i)t − i for 0 ≤ t ≤ 1
to get Z Z 1
2
|z|2 dz = (t2 + (t − 1)2 )(1 + i) dt = (1 + i).
γ 0 3
16. Begin with Z
cos(z 2 ) dz ≤ 8πM,
γ

where 8π is the length of the circle γ, and M is a bound for cos(z 2 ) on γ,

so
| cos(z 2 )| ≤ M for z on γ.
There are of course many possible choices for M . Indeed, if we find a
bound, then any larger number is also a bound. To find one such number,
let z = x + iy so
z 2 = x2 − y 2 + 2ixy
and
| cos(z 2 )| = | cos(x2 − y 2 + 2ixy)|
= | cos(x2 − y 2 ) cosh(2xy) − i sin(x2 − y 2 ) sinh(2xy)|
≤ cosh(2xy) + | sinh(2xy)| = e2xy .
in which we have used the facts that the real hyperbolic cosine is positive,
and that the real sine and cosine functions are bounded by 1.
Now, for points on γ, x = 4 cos(t) and y = 4 sin(t) for 0 ≤ t ≤ 2π, so
e2xy = e2(4 cos(t))(4 sin(t)) = e32 sin(t) cos(t)
= e16 sin(2t) ≤ e16 .
We can choose M = e16 to obtain
Z
cos(z 2 ) dz ≤ 8πe16 .
γ

This is a massively huge bound for this integral, and is not claimed to be
an approximation of the integral in any sense. But this is what we get
with the quick and crude bounds made along the way. With more effort
we may get a smaller bound. However, in some applications, it is enough
to know that a certain term is bounded, and the size of the bound may
not matter.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 20.2. CAUCHY’S THEOREM 479
√
17. The length of γ is 5. Now we need a number M such that
1
≤ M on γ.
1+z
Notice that the point on γ closest to z = −1 is 2 + i, so for z on the curve,
√
|z + i| = |z − (−1)| ≥ |2 + i + i| = 10.

Then
1 1 1
= ≤√ .
1+z |1 + z| 10
√
We can therefore use M = 1/ 10 to get the bound
Z √
1 5 1
dz ≤ √ = √ .
γ 1+z 10 2

20.2 Cauchy’s Theorem

1. sin(z) is differentiable for all z, hence on an open set containing the curve
and all points enclosed by the curve. By Cauchy’s theorem,
I
sin(z) dz = 0.
γ

2. The circle encloses i, at which f (z) is not defined (hence not differen-
tiable), so Cauchy’s theorem does not apply. Evaluate the integral by
parametrizing γ(t) = 2i + 2eit for 0 ≤ t ≤ 2π. Then
Z 2π 
2i + 6eit it
I 
2i
dz = 3ie dt
γ z−i 0 3eit
Z 2π
= (−2 + 6ieit ) dt = −4π.
0

3. γ encloses 2i, at which f (z) is not defined. Parametrize

γ(t) = 2i + 2eit for 0 ≤ t ≤ 2π.

Then
I Z 2π
1 1
dt = 2ieit dt
γ (z − 2i)3 0 (2eit )3
Z 2π
i
= e−2it dt = 0.
4 0

This integral happens to be zero, but we could not conclude this from
Cauchy’s theorem, which does not apply here.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 480 CHAPTER 20. COMPLEX INTEGRATION

4. Because z 2 sin(z) is differentiable everywhere, and in particular on the

curve and throughout the points it encloses, then Cauchy’s theorem applies
and I
z 2 sin(z) dz = 0.
γ

5. f (z) = z is not differentiable. Write γ(t) = eit for 0 ≤ t ≤ 2π. Then

I Z 2π
z dz = e−it ieit dt = 2πi.
γ 0

6. f (z) = 1/z is differentiable except at the origin, which is enclosed by the

curve. Parametrize γ(t) = 5eit for 0 ≤ t ≤ 2π. Then
I Z 2π Z 2π
1 1
dz = −it
5ieit dt
γ z 0 0 5e
Z 2π
= ie2it dt = 0.
0

7. Because f (z) = zez is differentiable on the curve and throughout the

region it encloses, then by Cauchy’s theorem,
I
zez dz = 0.
γ

8. A polynomial is differentiable on the entire complex plane, so

I
(z 2 − 4z + i) dz = 0.
γ

9. f (z) = |z|2 is not differentiable at any point other than 0, Cauchy’s theo-
rem does not apply. Write γ(t) = 7eit for 0 ≤ t ≤ 2π to obtain
I Z 2π
|z|2 dz = 49(7ieit ) dt = 0.
γ 0

10. f (z) = sin(1/z) is not differentiable at 0, which is not on or enclosed by

γ, so Cauchy’s theorem applies and
I
sin(1/z) dz = 0.
γ

11. f (z) = Re(z) is not differentiable, so write γ(t) = 2eit for 0 ≤ t ≤ 2π.
Then
I Z 2π
Re(z) dz = 2 cos(t)(2ieit ) dt
γ 0
Z 2π
= [4i cos2 (t) − 4 cos(t) sin(t)] dt = 4πi.
0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 20.3. CONSEQUENCES OF CAUCHY’S THEOREM 481

12. f (z) = z 2 is differentiable for ll z, so by Cauchy’s theorem,

I
z 2 dz = 0
γ

for any simple closed path in the plane. We can therefore concentrate on
the integral of just Im(z) around the square. Let S1 be the left side, S2 the
lower side, S3 the right side, and S4 the top side, and orient the rectangle
counterclockwise. We can parametrize each side:
S1 : γ1 (t) = −2it,
S2 : γ2 (t) = 2t − 2i,
S3 : γ3 (t) = 2 − 2i(1 − t),
S4 : γ4 (t) = 2(1 − t).
For each side, t varies from 0 to 1. Then
I Z 1
Im(z) dz = (−2t)(−2i) dt
γ 0
Z 1 Z 1 Z 1
+ (−2)2 dt + (−2)(1 − t)(2i) dt + 0 dt
0 0 0
= 2i − 4 − 2i = −4.
In total, then, I
(z 2 + Im(z)) dz = −4.
γ

20.3 Consequences of Cauchy’s Theorem

1. Because 2i is the center of the circle γ, we can apply Cauchy’s integral
formula with f (z) = z 4 to obtain
z4
I
dz = 2πif (2i) = 2πi(2i)4 = 32πi.
γ z − 2i

2. By Cauchy’s integral formula with f (z) = sin(z 2 ),

sin(z 2 )
I
dz = 2πif (5) = 2πi sin(25).
γ z−5

3. Use Cauchy’s integral formula, with f (z) = z 2 − 4z + i, to obtain

I 2
z − 4z + i
dz = 2πif (1 − 2i)
γ z − 1 + 2i

= 2πi[(1 − 2i)2 − 4(1 − 2i) + i] = 2πi(−8 + 7i)

= −14π − 16πi.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 482 CHAPTER 20. COMPLEX INTEGRATION

4. Apply the Cauchy integral formula for derivatives with n = 1 and f (z) =
2z 3 to obtain
2z 3
I
2
dz = 2πif 0 (2) = 48πi.
γ (z − 2)

5. We can use the Cauchy integral formula for derivatives with n = 1 and
f (z) = iez :
iez
I
2
dz = 2πif 0 (2 − i)
γ (z − 2 + i)

= 2πi(ie2−i ) = −2πe2 [cos(1) − i sin(1)].

6. Apply Cauchy’s formula for derivatives with n = 2 and f (z) = cos(z − i)

to obtain
cos(z − i)
I
2πi 00
3
dz = f (−2i)
γ (z + 2i) 2
= −πi cos(−3i) = −πi cosh(3).

7. With f (z) = z sin(3z) and n = 2, Cauchy’s formula for derivatives gives

us
I
z sin(3z) 2πi 00
3
dz = f (−4)
γ (z + 4) 2
= πi[6 cos(12) − 36 sin(12)].

8. γ is not a closed curve, so parametrize γ(t) = 1 − t − it for 0 ≤ t ≤ 1. On

the curve,
p
f (γ(t)) = 2iγ(t)|γ(t)| = 2i[(1 − t) + it] 1 − 2t + 2t2

so
Z Z 1 p
2iz|z| dz = 2i[(1 − t) + it] 1 − 2t + 2t2 dt
γ 0
Z p 1 Z 1 p
=2 2
1 − 2t + 2t dt + 2i (2t − 1) 1 − 2t + 2t2 dt
0 0
√ !
√ 2+1
= 1 + 24 ln √ .
2−1

These definite integrals can be evaluated using

−1 + 4t p
Z p
1 − 2t + 2t2 dt = 1 − 2t + 4t2
8
√   
7 2 4 1
+ sinh−1 √ t−
32 7 4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 20.3. CONSEQUENCES OF CAUCHY’S THEOREM 483

and Z p 1
(2t − 1) 1 − 2t + 2t2 dt = (1 − 2t + 2t2 )3/2 .
3
9.
−(2 + i) sin(z 4 )
I
d
2
dz = −2πi(2 + i) (sin(z 4 ))
γ (z + 4) dz z=−4
 3 4

= 2πi(1 − 2i) 4z cos(z ) z=−4
= −512π(1 − 2i) cos(256).

10. γ is not a closed curve. An antiderivative of (z − i)2 is (z − i)3 /3, so

−i
Z
1
(z − i)2 dz = (z − i)3
γ 3 i
1 8
= (−2i)3 = i.
3 3
11. Parametrize γ(t) = 3 − t + (1 − 6t)i for 0 ≤ t ≤ 1. Then
Z Z 1
Re(z + 4) dz = (7 − t)(−1 − 6i) dt
γ 0
13 13
= (−1 − 6i) = − − 39i.
2 2
12.
3z 2 cosh(z)
I
d
2
dz = 2πi (3z 2 cosh(z))
γ (z + 2i) dz z=−2i
2
 
= 2πi 6 cosh(z) − 3z sinh(z) z=−2i
= 2πi[−12i cosh(2i) + 12 sinh(2i)]
= 24π[cos(2) − sin(2)].

13. First evaluate

ez
I
dz
γ z
by Cauchy’s integral formula to obtain
Z z
e
dz = 2πiez = 2πi.
γ z z=0

Now evaluate this integral by parametrizing γ(t) = eit for 0 ≤ t ≤ 2π:

I z Z 2π cos(t)+i sin(t)
e e
dz = ieit dt
γ z 0 eit
Z 2π Z 2π
=i ecos(t) cos(sin(t)) dt − ecos(t) sin(sin(t)) dt
0 0
= 2πi.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 484 CHAPTER 20. COMPLEX INTEGRATION

By equating the real parts of both sides of this equation, and then the
imaginary parts, we obtain
Z 2π
ecos(t) cos(sin(t)) dt = 2π
0

and Z 2π
ecos(t) sin(sin(t)) dt = 0.
0
The first integral is not obvious. The second could be done without com-
plex analysis by observing that the integral from 0 to π is the negative of
the integral from π to 2π.
14. First write
z − 4i z − 4i
f (z) = = .
z 3 + 4z z(z − 2i)(z + 2i)
Now let γ1 , γ2 and γ3 be nonintersecting circles in the region bounded by γ
and having centers, respectively, 0, 2i, −2i. By the extended deformation
theorem,
I 3 I
X
f (z) dz = f (z) dz.
γ n=1 γj

Consider each integral on the right. On and in the interior of γ1 , think of

(z − 4i)/(z − 2i)(z + 2i)
f (z) = .
z
Apply the Cauchy integral formula to obtain
 
z − 4i
I
f (z) dz = 2πi
γ1 (z − 2i)(z + 2i) z=0
 
−4i
= 2πi = 2π.
(−2i)(2i)
For the integral over γ2 , we can apply Cauchy’s integral formula to
(z − 4i)/z(z + 2i)
I I
f (z) dz =
γ2 γ2 z − 2i
 
−2i π
= 2πi =− .
(2i)(4i) 2
And, for the integral over γ3 , think of
(z − 4i)/z(z − 2i)
f (z) =
z + 2i
and apply Cauchy’s integral formula in this region to obtain
 
−6i
I
3π
f (z) dz = 2πi = .
γ3 (−2i)(−4i) 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 20.3. CONSEQUENCES OF CAUCHY’S THEOREM 485

Finally, we have I
π 3π
f (z) dz = 2π − − = 0.
γ 2 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 486 CHAPTER 20. COMPLEX INTEGRATION

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 21

Series Representations of
Functions

21.1 Power Series

In each of Problems 1–6, the strategy is to take the limit of the magnitude of
the ratio of successive terms of the series. The series converges when this limit
(if it exists) is less than 1.

1. Take the limit of the magnitude of successive terms:

(n + 2)/2n+1 1n+2
n
|z + 3i| = (z + 3i)
(n + 1)/2 2n+1
1
→ |z + 3i|.
2
The series converges (absolutely) if
1
|z + 3i| < 1
2
or
|z + 3i| < 2.
The power series has radius of convergence 2 and open disk of convergence
|z + 3i| < 2, the open disk of radius 2 about the center −3i.
2.
2
1/(2n + 3)2 (z − i)n+1

2n + 1
= |z − i|2
1/(2n + 1)2 (z − i)n 2n + 3
→ |(z − i)2 |.

if |z − i| < 1. This power series has radius of convergence 1, and open disk
of convergence |z − i| < 1, the open disk of radius 1 about the center i.

487

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 488 CHAPTER 21. SERIES REPRESENTATIONS OF FUNCTIONS

3.
(n + 1)n+1 /(n + 2)n+1
(z − 1 + 3i)
nn /(n + 1)n
(n + 1)2n+1
= n |z − 1 + 3i|
n (n + 2)n+1
 n  n+1
n+1 n+1
= |z − 1 + 3i|
n n+2
 n  n  
1 1 + 1/n 1 + 1/n
= 1+ |z − 1 + 3i|
n 1 + 2/n 1 + 2/n

and the limit of this quantity is less than 1 if |z − 1 + 3i| < 1. The radius
of convergence is 1 and the disk of convergence is the open disk of radius
1 centered at 1 − 3i.
In this limit, we have used (several times) the fact that
 x n
lim 1 + = ex .
n→∞ n
4.
(2i/(5 + i))n+1
|z + 3 − 4i|
(2i/(5 + i)n )n
2i
→ (z + 3 − 4i)
5+i
2
= √ |z + 3 − 4i|,
26
and this limit is less than 1 when
√
26
|z + 3 − 4i| < .
2
√
This power series has radius of convergence
√ 26/2, and the open disk of
convergence is the disk of radius 26/2 centered at −3 + 4i.
5.
in+1 /2n+2 1
(z + 8i) → |z + 8i|
in /2n 2
This ratio has limit < 1 if |z + 8i| < 2. The power series has radius of
convergence 2 and the open disk of convergence is the open disk of radius
2 centered at −8i.
6.
(1 − i)n+1 /(n + 3) n + 2√
n
(z − 3) = 2|z − 3|
(1 − i) /(n + 2) n+1
√
→ 2|z − 3|

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 21.1. POWER SERIES 489
√
√ than 1 if |z − 3| < 1/ 2. This power series has radius
and this limit is less
of convergence
√ 1/ 2 and the open disk of convergence is the open disk of
radius 1/ 2 centered at 3.
7. No. The power series has center 2i. If the series converges at 0, it must
also converge at the point i that is closer to the center 2i than 0 is.
8. The center of this power series is 4 − 2i. Now, 1 + i is closer to 4 − 2i than
i is, so if the series converges at i, it must also converge at 1 + i.

In each of Problems 9–14, we attempt to use known series to derive the

requested series.

9. Assuming that we know the series for cos(z):

∞
X (−1)n 2n
cos(z) = z .
n=0
(2n)!

This converges for all z. Replace z with 2z to obtain

∞
X (−1)n 22n 2n
cos(2z) = z .
n=0
(2n)!

In writing these series, be careful with factorials. For example, in general

(2n)! 6= 2n!.
10. We know that
∞
X 1 n
ez = z
n=0
n!
for all z. Then
∞ ∞
X 1 X (−1)n n
e−z = (−z)n = z .
n=0
n! n=0
n!

The expansion of e−z about −3i will have powers of z + 3i, so write

e−z = e−z−3i+3i = e3i e−(z+3i)

∞
3i
X (−1)n
=e (z + 3i)n ,
n=0
n!

converging for all z.

11. This is just a rearrangement of the given polynomial into powers of z−2+i.
This can be done algebraically, or we can write the Maclaurin series of this
polynomial about 2 − i. This series will be

f (z) = z 2 − 3z + i = c0 + c1 (z − 2 + i) + z2 (z − 2 + i)2 ,

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 490 CHAPTER 21. SERIES REPRESENTATIONS OF FUNCTIONS

where
c0 = f (2 − i) = −3, c1 = f 0 (2 − i) = 1 − 2i
and
1 00
c2 = f (2 − i) = 1.
2
The expansion of f (z) about 2 − i is

z 2 − 3z + i = −3 + (1 − 2i)(z − 2 + i) + (z − 2 + i)2 .

12. Use the series for ez and sin(z):

∞ ∞
z
X 1 n X (−1)n 2n+1
e = z and sin(z) = z
n=0
n! n=0
(2n + 1)!

to write
∞ ∞
X 1 n X (−1)n 2n+1
ez − i sin(z) = z −i z .
n=0
n! n=0
(2n + 1)!

13. Like Problem 11, this can be done as an algebraic rearrangement of terms
in f (z) = (z − 9)2 , or as a power series about 1 + i, which will be in powers
of z − i − i. Using the latter approach, compute the coefficients

c0 = f (1 + i) = 63 − 16i, c1 = f 0 (1 + i) = −16 + 2i

and
1 00
c2 = f (1 + i) = 1.
2
Then

(z − 9)2 = 63 − 16i + (−16 + 2i)(z − 1 − i) + (z − 1 − i)2 .

14. Replace z with z + i in the Maclaurin expansion of sin(z) to obtain

∞
X (−1)n
(z + i)2n+1 .
n=0
(2n + 1)!

This converges for all z.

15. We know that f (0) = 1, f 0 (0) = i, and f 00 (z) = 2f (z) + 1. Compute

f 00 (0) = 2f (0) + 1 = 3,
f (3) (0) = 2f 00 (0) = 2i,
f (4) (0) = 2f 00 (0) = 6,
f (5) (0) = 2f (3) (0) = 4i.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 21.1. POWER SERIES 491

Now use Taylor’s formula for the coefficients (in this case, about 0,
1 (n)
cn =f (0)
n!
to write the first six terms of the expansion:
3 2i 6 4i
1 + iz + z 2 + z 3 + z 4 + z 5 .
2 3! 4! 5!
In this problem it is not difficult to write the entire Maclaurin expansion,
because an inductive argument shows that
f (2n) (0) = 2n + 2n−1 and f (2n+1) (0) = 2n i.

16. With f (z) = sin2 (z), we will need

f 0 (z) = 2 sin(z) cos(z) = sin(2z), f 00 (z) = 2 cos(2z),
f (3) (z) = −4 sin(2z), f (4) (z) = −8 cos(2z),
f (5) (z) = 16 sin(2z), f (6) (z) = 32 cos(2z).

(a) Using these derivatives, we can find the first seven terms of the power
series expansion of f (z) about 0:
1 2
sin2 (z) = z 2 − z 4 + z 6 + · · · .
3 45
(b) Multiply
  
1 1 5 1 1 5
sin2 (z) = z − z3 + z + ··· z − z3 + z + ···
6 120 6 120
   
1 1 1 1 1
= z2 − + z4 + + + z6 + · · ·
6 6 120 36 120
1 2 6
= z2 − z4 = z + ··· ,
3 45
(c) Use the definition of sin(z) to write
1
2
sin2 (z) = − eiz − e−iz
4
1
= − e2iz + e−2iz − 2


4 
(2iz)2 (2iz)7

1 1
= − 1 + 2iz + + ··· + + ···
2 4 2! 7!
(2iz)2 (2iz)7
 
1
− 1 − 2iz + + ··· − + ···
4 2! 7!
 
1 1 4 8
= − 1 − 4z 2 + z 4 − z 6 + · · ·
2 4 3 45
1 2
= z2 − z4 + z6 + · · ·
3 45

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 492 CHAPTER 21. SERIES REPRESENTATIONS OF FUNCTIONS

(d) We can also write

1 1
sin2 (z) = − cos(2z)
2 2 
(2z)2 (2z)4 (2z)6

1 1
== − 1− + − + ···
2 2 2! 4! 6!
 
1 1 2 4
= − 1 − z2 + z4 − z6 + · · ·
2 2 3 45
1 2
= z2 − z4 + z6 + · · ·
3 45

17. Let z be a complex number and consider the integral

zn
I
1
ezw dw.
2πi γ n!wn+1

Here γ is the unit circle about the origin, oriented counterclockwise as

usual. Expand ezw in its Maclaurin series and parametrize γ(t) = eit for
0 ≤ t ≤ π to write

∞
zn z n X (zw)k
I I
1
ezw dw = dw
γ n!wn+1 2πi γ n!wn+1 k!
k=0
∞
z n+k wk−n−1
I X
1
= dw
2πi γ n!k!
k=0
Z 2π X∞ i(k−n−1)t
1 z n+k
= n!k!ieit dt
2πi 0 e
k=0
∞ Z 2π n+k
X 1 z
= ei(k−n)t dt.
2π 0 n!k!
k=0

Now,
(
2π
if k 6= n,
Z
i(k−n)t 0
e dt =
0 2π if k = n.

We therefore have

zn (z n )2
I
1
n+1
ezw dw = .
2πi γ n!w (n!)2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 21.1. POWER SERIES 493

Finally, we can write

∞ ∞
X 1 2n X 1 zn
2
z = n+1
ezw dw
n=0
(n!) n=0
2πi n!w
Z 2π X∞
1 zn it
= i(n+1)t
eze eit dt
2πi 0 n=0 n!e
∞
Z 2π " X #
1 (ze−it )n zeit
= e dt
2π 0 n=0
n!
Z 2π
1 −it it
= eze eze dt
2π 0
Z 2π
1 it −it
= ez(e +e ) dt
2π 0
Z 2π
1
= e2z cos(t) dt.
2π 0

18. f (z) has a zero of order 3 at 0 because z 3 has a zero of order 3 there and
cos(0) 6= 0.

19. f (z) has a zero of order 4 at 0 because z 2 has a zero of order 2 at 0 and
sin2 (z) also has a zero of 2 there (because sin(z) has a zero of order 1 at
0).

20. f (z) = (z − π/2)2 cos(z) has a zero of order 3 at π/2.

21. f (z) has a zero of order 3 at 3π/2 because cos(z) has a simple zero there.

22. f (z) has a zero of order 4 at 0 because cos(z) has a simple zero at π/2.

23. f (z) is not defined at z = 0, so we cannot really speak of it having a

zero there. However, notice something interesting. Using the Maclaurin
expansion of sin(z), with z 4 in place of z, and divided by z 2 , we can write
∞
1 4
X (−1)n 8n+2
sin(z ) = z .
z2 n=0
(2n + 1)!

This power series converges for all z, and has the value 0 at 0. We can
therefore extend f (z) by giving it the value 0 at z = 0, and obtain a
differentiable function. This extended function has a zero of order 2 at 0.

24. As in Problem 23, f (z) is not defined at π. However, (z − π)5 has a fifth
order zero at π, and sin2 (z) has a second order zero at π, and if we extend
f (z) by giving it the value 0 at π, this extended function has a third order
zero at π.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 494 CHAPTER 21. SERIES REPRESENTATIONS OF FUNCTIONS

5. Compute the kth derivative of f (z) at z0 using each series, obtaining

∞
X
f k (z0 ) = an (n)(n − 1) · · · (n − k + 1)(z − z0 )n−k
n=0
X∞
= bn (n)(n − 1) · · · (n − k + 1)(z − z0 )n−k .
n=0

Then
f k (z0 ) = k!ak = k!bk ,
so for each k = 0, 1, 2, · · · ,
1 (k)
ak = f (z0 ) = bk .
k!

21.2 The Laurent Expansion

Problems 1–10 are solved using manipulations of known series, such as geomet-
ric series and power series for exponential and trigonometric functions. It is
sometimes best, in seeking an expansion about z0 , to focus on getting an ex-
pression involving powers of z − z0 , using algebraic manipulations, or sometimes
integration and differentiation.
In particular, it is useful to know the geometric series
∞
1 X
= rn
1 − r n=0

and
∞
1 X
= (−1)n rn ,
1 + r n=0
for |r| < 1.

1. We want an expansion in powers of z − i. To this end, begin with

2z 1 1
2
= + .
1+z z−i z+i
The first term is already an expansion in powers of z − i (having only one
term). For the second term, write
1 1 1
= = z−i


z+i 2i + (z − i) 2i 1 − 2i
∞  n
1 X z−i
= (−)n
2i n=0 2i
∞
X (−1)n
= n+1
(z − i)n .
n=0
(2i)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 21.2. THE LAURENT EXPANSION 495

This expansion is valid for

z−i 1
= |z − i| < 1,
2i 2
or
|z − i| < 2.
The Laurent expansion of f (z) about i is therefore
∞
X (−1)n
1
+ (z − i)n .
z − i n=0 (2i)n+1

This represents f (z) in the annulus 0 < |z − i| < 2.

2. For z 6= 0, write
∞
sin(z) 1 X (−1)n 2n+1
= 2 z
z2 z n=0 (2n + 1)!
∞
X (−1)n 2n−1
= z .
n=0
(2n + 1)!

This series represents sin(z)/z 2 in the annulus 0 < |z < ∞, which is the
plane with the origin removed.
3. If z 6= 0, then
∞
" #
1 − cos(2z) 1 X (−1)n 2n
= 2 1− (2z)
z2 z n=0
(2n)!
∞
X (−1)n+1 4n 2n−2
= z .
n=1
(2n)!

4. Write
∞  2n
(−1)n i
 
2 i 2
X
z cos =z
z n=0
(2n)! z
∞
X 1 2−2n
= z ,
n=0
(2n)!

in which we have used the fact that i2n = (−1)n . This expansion repre-
sents the function in the annulus 0 < |z| < ∞.
5. The denominator is already in terms of z − 1, so concentrate on the nu-
merator:
z2 ((z − 1) + 1)2 1 + 2(z − 1) + (z − 1)2
= =−
1−z 1−z z−1
1
=− − 2 − (z − 1).
z−1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 496 CHAPTER 21. SERIES REPRESENTATIONS OF FUNCTIONS

This represents the function for 0 < |z − 1| < ∞, the complex plane with
1 removed.
6. As with Problem 5, algebraic manipulation will be enough here. Write
z2 + 1 1 z2 + 1 1 1 + [(z − 1/2) + 1/2]2
   
= =
2z − 1 2 z − 1/2 2 z − 1/2
 
5 1 1 1 1
= + + z−
8 z − 1/2 2 2 2

1
for 0 < z − 2 < ∞.

7. Use the exponential series to obtain

∞ ∞
1 z2 1 X 1 2n X 1 2n−2
e = z = z ,
z2 z 2 n=0 n! n=0
n!

for 0 < |z| < ∞.

8. Use the Maclaurin expansion of the sine function to get
∞ ∞
1 1 X (−1)n 2n+1 X (−1)n 2n−1
sin(4z) = z = z ,
z z n=0 (2n + 1)! n=0
(2n + 1)!

for 0 < |z| < ∞.

9. The denominator is already a power of z − i, so we can write
z+i 2i + (z − i) 2i
= =1+
z−i z−i z−i
for 0 < |z − i| < ∞.
10. Use the Maclaurin expansion of sinh(z) (or write sinh(z) = (ez − e−z )/2
and use the exponential series) to obtain
  X ∞  2n+1 X ∞
1 1 1 1
sinh 3
= 3
= z −6n−3
z n=0
(2n + 1)! z n=0
(2n + 1)!

for 0 < |z| < ∞.

11. By Cauchy’s integral formula, for any z enclosed by Γ1 ,
I
1 f (w)
f (z) = dw.
2πi Γ1 w − z

Because Γ2 does not enclose z,

I
1 f (w)
dw = 0
2πi Γ2 w − z

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 21.2. THE LAURENT EXPANSION 497

by Cauchy’s theorem. The factor of 1/2πi was included in the last equation
so we can add these two integrals to get
I I 
1 f (w) f (w)
f (z) = dw + dw .
2πi Γ2 w − z Γ1 w − z

Orientation on both curves is counterclockwise. In this sum of integrals,

L1 and L2 are traversed in both directions, so the integrals over these
segments are zero. The integrals in square brackets therefore give us the
integrals over γ1 and γ2 , but counterclockwise on γ1 and clockwise on γ2 .
Reversing this orientation on γ1 so that all integrals are over counterclock-
wise curves, we have
I I 
1 f (w) f (w)
f (z) = dw − dw .
2πi γ2 w − z γ1 w − z

Now manipulate the 1/(w − z) factor in each integral to derive the result
we want. For the integral over γ2 , write
1 1
=
w−z w − z0 − (z − z0 )
1 1
=
w − z0 1 − (z − z0 )/(w − w0 )
∞  n
1 X z − z0
=
w − z0 n=0 w − z0
∞
X 1
= n+1
(z − z0 )n .
n=0
(w − z0 )

This geometric expansion is valid because, for w on γ2 ,

w − z0
< 1.
z − z0
For the integral over γ1 , use the fact that, for w on this curve,
w − z0
< 1.
z − z0
Now we have
1 1
=
w−z w − z0 − (z − z0 )
−1 1
=
z − z0 1 − (w − z0 )/(z − z0 )
∞  
1 X w − z0
=−
z − z0 n=0 z − z0
∞
X 1
=− (w − z0 )n .
n=0
(z − z0 )n+1

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 498 CHAPTER 21. SERIES REPRESENTATIONS OF FUNCTIONS

Substitute these expressions into the sum of integrals representing f (z)

and interchange the integrals with the summation to obtain
∞
I !
1 X f (w)
f (z) = dw (z − z0 )n
2πi γ2 n=0 (w − z0 )n+1
∞
I ! n+1
1 X
n 1
+ f (w)(w − z0 ) dw
2πi γ1 n=0 z − z0
∞  I 
X 1 f (w)
= n+1
dw (z − z0 )n
n=0
2πi γ2 (w − z0 )
∞  I 
X 1 n 1
+ f (w)(w − z0 ) dw n+1
.
n=0
2πi γ1 (z − z 0)

Finally, use the deformation theorem to replace these integrals over γ1 and
γ2 with integrals over Γ, which is any simple closed path in the annulus
and enclosing z0 . This gives us
∞
X
f (z) = cn (z − z0 )n ,
n=−∞

with the integral expressions given for the coefficients cn .

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 22

Singularities and the

Residue Theorem

22.1 Singularities
1. cos(z)/z has one singularity, a double pole at z = 0.

2. f (z) has a double pole at −i and a simple pole at 1.

3. e1/z (z + 2i) has an essential singularity at 0.

4. Note that
sin(z) sin(z − π)
=−
z−π z−π
so
sin(z) sin(z − π)
lim = − lim
z→π z−π z→π z−π
sin(z)
= − lim = 1 6= 0.
z→0 z
Because this limit is nonzero, the function has a removable singularity at
z = π.

5. The function has a double pole at 1 and simple poles at i and −i.

6. The function has a double pole at −1.

7. Write
z−i z−i 1
= = ,
z2 + 1 (z + i)(z − i) z+i
so the function has a simple pole at −i.

499

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 500 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM

8. We need to know if sinh(z) has any zeros other than z = 0. For this, solve

1 z
e − e−z = 0.


sinh(z) =
2
For this to be true, we must have

e2z = 1

so 2z = 2nπi, or z = nπi, with n any integer. These are simple zeros of

sinh(z) because cosh(nπi) 6= 0 for any integer n.
For n = 0, observe that 0 is a simple zero of both sin(z) and sinh(z), so 0
is a removable singularity of the function.

9. The denominator has simple zeros at 1, −1, i, −i and these are simple
poles of the function because the numerator does not vanish at any of
these numbers.

10. tan(z) = sin(z)/ cos(z) has simple poles at the zeros of cos(z), which are
simple and occur at points z = (2n + 1)π/2, in which n any integer.

11. sec(z) = 1/ cos(z) has simple poles at the zeros of cos(z), which are the
simple zeros (2n + 1)π/2 with n any integer.

12. e1/z(z+1) has essential singularities at 0 and −1. One way to see this is to
write
1 1 1
= −
z+1 z z+1
so
e1/z(z+1) = e1/z e−1/(z+1) .

13. Suppose f is differentiable at z0 and f (z0 ) 6= 0, while g has a pole of order

m at z0 . We want to show that the product f g has a pole of order m at
z0 .
Because g has a pole of order m at z0 , the Laurent expansion in some
annulus about z0 has the form
∞
k X
g(z) = + cn (z − z0 )n .
(z − z0 )m n=−m+1

with k 6= 0. Then
∞
X
(z − z0 )m g(z) = k + cn (z − z0 )n+m .
n=−m+1

If we denote the power series on the right as h(z), then

(z − z0 )m g(z) = h(z),

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22.2. THE RESIDUE THEOREM 501

where h(z0 ) = k 6= 0. Further, in some annulus about z0 ,

f (z)h(z)
f (z)g(z) = .
(z − z0 )m
Because f (z0 )h(z0 ) 6= 0, f (z)g(z) has a pole of order m at z0 .

22.2 The Residue Theorem

1. The function has simple poles at 1 and −2i, both enclosed by γ. Keep in
mind that only singularities enclosed by the curve are relevant in evaluat-
ing the integral by the residue theorem.
Compute
d 1 + z2
 
Res(f, 1) = lim
z→1 dz z + 2i
(z + 2i)(2z) − (1 + z 2 )
= lim
z→1 (z + 2i)2
4i
= ,
−3 + 4i
and
1 + z2 −3
Res(f, −2i) = lim = .
z→−2i (z − 1)2 −3 + 4i
Then
1 + z2
I  
4i 3
dz = 2πi − = 2πi.
γ (z − 1)2 (z + 2i) −3 + 4i −3 + 4i

2. γ encloses i, which is a double pole and the only singularity of the function.
Then I
2z d
2
dz = 2πiRes(f, i) = lim (2z) = 4πi.
γ (z − i) z→i dz
3. The only singularity of ez /z is a simple pole at 0, and this is not enclosed
by γ, so I z
e
dz = 0
γ z
by Cauchy’s theorem.
4. f (z) has singularities at 2i and −2i. These are both simple poles, and
both are enclosed by γ. Then
I
cos(z)
dz = 2πiRes(f, 2i) + 2πiRes(f, −2i)
γ 4 + z2
cos(2i) cos(−2i)
= 2πi + 2πi
4i −4i
= 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 502 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM
√ √
5. The function has simple poles at 6i and − 6i, both enclosed by γ. Then
I
z+i h √ √ i
2
dz = 2πi Res(6, 6i) + Res(f, − 6i)
γ z +6
"√ √ #
6+1 6−1
= 2πi √ + √ = 2πi.
2 6 2 6

6. f (z) has a simple pole at −1/2. This is the only singularity, and it is
enclosed by γ. Then
 
z−i
I
1 1
dz = 2πiRes(f, 01/2) = − +i .
γ 2z + 1 2 2

7. z/ sinh2 (z) has a simple pole at 0 and double poles at nπi, for every
nonzero integer n. The only singularity enclosed by γ is 0, so
I
f (z) dz = 2πiRes(f, 0).
γ

Compute this residue as

z2
 
Res(f, 0) = lim zf (z) = lim
z→0 z→0 sinh2 (z)
2
z
= lim
z 2 + 61 z 4 + · · ·
z→0

1
= lim = 1.
z→0 1 + 1 z 2 + · · ·
6

Then I
z
dz = 2πi.
γ sinh2 (z)
8. cos(z)/zez has only one singularity, a simple pole at 0, and this is enclosed
by γ. Therefore
I
cos(z)
z
dz = 2πiRes(f, 0)
γ ze
cos(z)
= 2πi lim = 2πi.
z→0 ez

9. f (z) has simple poles at i, 3i and −3i. Only the pole at −3i is enclosed
by the curve, so
I
iz
2
dz = 2πiRes(f, −3i)
γ (z + 9)(z − i)
 
iz 1 πi
= 2πi lim = 2πi − =− .
z→−3i (z − 3i)(z − i) 8 4

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22.2. THE RESIDUE THEOREM 503

2
10. e3/z has an essential singularity at 0 and no other singularities. 0 also
lies in the region bounded by the curve. We need the residue of f (z) at 0.
Here we do not have a formula, but we can look at the Laurent expansion
about 0:
∞  n
2 X 1 3
e3/z = 2
.
n=0
n! z
The coefficient of 1/z in this expansion is zero, so the residue is zero and
I
2
e3/z dz = 0.
γ

11. f (z) has only one singularity, a simple pole at −4i, and this is outside the
region bounded by the curve. By Cauchy’s theorem,
8z − 4i + 1
I
dz = 0.
γ z + 4i

12. f (z) has a simple pole at 1 − 2i, which is enclosed by γ, so

z2
I
dz = 2πiRes(f, 1 − 2i)
γ z − 1 + 2i

= 2πi((−1 + 2i)2 ) = 2π(4 − 3i).

13. The singularities of coth(z) = cosh(z)/ sinh(h) are the zeros of sinh(z).
This means that coth(z) has simple poles at nπi, with n any integer.
Only the simple pole at 0 is enclosed by the curve, so
I
cosh(0)
coth(z) dz = Res(f, 0) = 2πi = 2πi.
γ cosh(0)

14. f (z) has simple poles at 2, 2e2πi/3 and 2e4πi/3 . Only 2 is enclosed by γ, so
(1 − z)2
I
3
dz = 2πiRes(f, 2)
γ z −8
(1 − z)2
= 2πi lim 2
  3z
z→2
1 πi
= 2πi = .
12 6

15. 0 and 4i are simple poles of f (z) and both are enclosed by γ, so
e2z
I
dz = 2πi [Res(f, 0) + Res(f, 4i)]
γ z(z − 4i)
e4i
 
1
= 2πi − +
4i 4i
π
= [cos(8) − 1 + i sin(8)].
2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 504 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM

16. The function has a double pole at 1 and γ is assumed to be a closed path
that enclosed 1, so
z2
I
2
dz = 2πiRes(f, 1)
γ (z − 1)
d 2
= 2πi lim (z ) = 4πi.
z→1 dz

17. z0 is a zero of order 2 of h(z), but g(z0 ) 6= 0. We want to show that

2g 0 (z0 ) 2 g(z0 )h(3) (z0 )

Res(g/h, z0 ) = − .
h00 (z0 ) 3 (h00 (z0 ))2
To do this, first write

h(z) = (z − z0 )2 ϕ(z),

with ϕ(z0 ) 6= 0. Then

 
d g(z)
Res(g/h, z0 ) = lim (z − z0 )2
z→z0 dz h(z)
 
d g(z)
= lim (z − z0 )2 ϕ(z)
z→z0 dz (z − z0 )2 ϕ(z)
 
d g(z)
= lim
z→z0 dz ϕ(z)
ϕ(z0 )g 0 (z0 ) − ϕ0 (z0 )g(z0 )
= .
(ϕ(z0 ))2
Now,
h0 (z) = 2(z − z0 )ϕ(z) + (z − z0 )2 ϕ0 (z).
,
h00 (z) = 2ϕ(z) + 4(z − z0 )ϕ0 (z) + (z − z0 )2 ϕ00 (z),
and
h(3) (z0 ) = 6ϕ0 (z) + 6(z − z0 )ϕ00 (z) + (z − z0 )2 ϕ(3) (z).
Then
1 00 1
ϕ(z0 ) =h (z0 ) and ϕ0 (z0 ) = h(3) (z0 ).
2 6
Substituting these into the expression for the residue, we have

2g 0 (z0 ) 2 g(z0 )h(3) (z0 )

Res(g/h, z0 ) = − .
h00 (z0 ) 3 (h00 (z0 ))2

18. Suppose first that f has a zero of order k at z0 in G. Consider the residue
of f 0 /f at z0 . First, write

f (z) = (z − z0 )k g(z),

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22.2. THE RESIDUE THEOREM 505

for some function g(z) that is differentiable on an open disk about z0 , and
g(z0 ) 6= 0. Then

f 0 (z) k(z − z0 )k−1 g(z) + (z − z0 )k g 0 (z)

=
f (z) (z − z0 )k g(z)
k g 0 (z)
= + .
z − z0 g(z)

Now g 0 /g is differentiable at z0 so the last equation implies that f 0 /f has

a simple pole at z0 , and

Res(f 0 /f, z0 ) = k.

Now see what happens if f has a pole of order m at z1 . In some annulus

about z1 ,
∞
X
f (z) = dn (z − z1 )n
n=−m

with d−m 6= 0. Then

∞
X ∞
X
(z − z1 )m f (z) = dn (z − z1 )n = dn−m (z − z1 )n = h(z),
n=−m n=0

with h(z) differentiable at z1 and h(z1 ) = d−m 6= 0. Then

1
f (z) = h(z).
(z − z1 )m

Now, in some disk about z1 ,

f 0 (z) −m(z − z1 )−m−1 h(z) + (z − z1 )−m h0 (z)

=
f (z) (z − z1 )−m h(z)
0
−m h (z)
= + .
z − z1 h(z)

This implies that

Res(f 0 /f, z1 ) = −m.
Therefore, the sum of the residues of f 0 /f at poles of f enclosed by γ in G
counts each zero of f enclosed by γ, according to its multiplicity, and each
pole of f enclosed by γ, according to the negative of its multiplicity. If Z
is the number of zeros of f enclosed by γ, counting multiplicities, and P
the number of poles of f enclosed by γ, also counting multiplicities, then
by the residue theorem,
I 0
f (z)
dz = 2πi(Z − P ).
γ f (z)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 506 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM

19. By the residue theorem, with g(z) = z/(2 + z 2 ),

I
z h √ √ i
2
dz = 2πi Res(g, 2i) + Res(g, − 2i)
γ 2+z
"√ √ #
2i − 2i
= 2πi √ + √
2 2i −2 2i
= 2πi.
To use the argument principle, write
f 0 (z) 1 2z
g(z) = = ,
f (z) 2 2 + z2
with f (z) = 2 + z 2 . Then f 0 /f = 2g. Now f (z) has two simple zeros
enclosed by γ, and no poles, so Z = 2, P = 0, and
I I
z 1 2z
2
dz = dz
γ 2 + z 2 γ 1 + z2
1
= (2πi)(Z − P ) = 2πi.
2
20. Let g(z) = tan(z). Then g has simple poles enclosed by γ at ±π/2. By
the residue theorem,
I
tan(z) dz = 2πi[Res(g, π/2) + Res(g, −π/2)]
γ
 
sin(π/2) sin(−π/2)
= 2πi +
− sin(π/2) − sin(−π/2)
= 2πi(−1 + (−1)] = −4πi.

To use the argument principle, write g(z) = −f 0 (z)/f (z) with f (z) =
cos(z). Now f (z) has no poles, and simple zeros at ±π/2 enclosed by γ.
Then Z = 2 and P = 0, so
I I 0
f (z)
g(z) dz = − = −2πi(Z − P ) = −4πi.
γ γ f (z)

√
21. g(z) = (z + 1)/(z 2 + 2z + 4) has simple poles at −1 ± 3i enclosed by γ.
By the residue theorem
I
z+1 h √ √ i
2
dz = 2πi Res(g, −1 − 3) + Res(g, −1 + 3)
γ z + 2z + 4
" √ √ #
1 − 1 − 3i −1 + 3i + 1
= 2πi √ + √
2(−1 − 3i) + 2 2(−1 + 3i) + 2
 
1 1
= 2πi + = 2πi.
2 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22.3. EVALUATION OF REAL INTEGRALS 507

To use the argument principle, write

z+1 1 f 0 (z)
=
z2 + 2z + 4 2 f (z)

where f (z) = z 2 + 2z + 4. f (z) has z = 2 zeros enclosed by γ and no poles

(P = 0), so I
2z + 2
2 + 2z + 4
dz = πi(Z − P ) = 2πi.
γ z

22. Because p(z) has exactly n simple zeros enclosed by γ, then p0 (z)/p(z) has
simple poles at z1 , · · · , zn and
p0 (zj )
Res(p0 /p, zj ) = = 1.
p0 (zj )
By the residue theorem,
I 0 n
p (z) X
dz = 2πi Res(p0 /p, zj ) = 2nπi.
γ p(z) j=1

To use the argument principle, note that p(z) has exactly n simple zeros
enclosed by γ, and a polynomial has no poles, so
I 0
p (z)
dz = 2πi(n − 0) = 2nπi.
γ p(z)

22.3 Evaluation of Real Integrals

1. With z = eiθ ,  
1 1 1
cos(θ) = z+ and dθ = dz
2 z iz
so
Z 2π I
1 1 1
dθ = dz 1
0 2 − cos(θ)
γ 2− + 1/z) iz 2 (z
I
1
= 2i 2
dz.
γ z − 4z + 1
√ √
The integrand has simple poles at z1 = 2 − 3 and z2 = 2 + 3. Only z1
is enclosed by γ, and
√ 1 1
Res(f, 2 − 3) = √ =− √ .
2(2 − 3) − 4 2 3
Then
2π
−1
Z
1 2π
dθ = 2i(2πi) √ = √ .
0 2 − cos(θ) 2 3 3

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 508 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM

2. f (z) = 1/(z 4 + 1) has two simple poles in the upper half-plane:

1 1
z1 = √ (1 + i) and z2 = √ (−1 + i).
2 2

We need the residues of f (z) at these points:

1 1
Res(f, z1 ) = and Res(f, z2 ) = − z2 .
4z12 4

Then Z ∞
1 2πi π
dx = − (z1 + z2 ) = √ .
−∞ x4 + 1 4 2

3. f (z) =√1/(1 + z 6 ) has simple √

poles in the upper half-plane at z1 = i,
z2 = ( 3 + i)/2, and z3 = (− 3 + i)/2. The residues of f (z) at these
poles are
1 1
Res(f, zj ) = 5 = − zj ,
6zj 6
so Z ∞  
1 1 2π
dx = 2πi (z1 + z2 + z3 ) = .
−∞ 1 + x6 6 3
Then Z ∞
1 π
dx = .
0 1 + x6 3

4.
Z 2π I
1 1 1
dθ = 1 dz
0 6 + sin(θ) γ 1 + 2 (z − 1/z) iz
I
1
=2 2 + 12iz + 1
dz.
γ z
√ √
The integrand has simple poles at z1 = (−6+ 37)i and z2 = (−6− 37)i.
Of these, only z1 is enclosed by γ. Further,
√ 1 1
Res(f, (−6 + 37)i) = = √ .
2z1 + 12i 2 37i
Then, recalling the factor of 2 in front of the previously written integral,
we have Z 2π  
1 1 2π
dθ = 2πi √ =√ .
0 6 + sin(θ) 37i 37

5. Let
ze2iz
f (z) = .
z 4 + 16

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22.3. EVALUATION OF REAL INTEGRALS 509
√
f has simple
√ poles in the upper half-plane at z1 = (1 + i)/ 2 and z2 =
(−1 + i)/ 2. Compute the residues:
√ √
e2 2i(−1+i)
e2 2
(−1 − i)
Res(f, z1 ) = and Res(f, z2 ) =
16i −16i

to obtain
" √ ! √ √ !#
∞
e−2 2
e2 2i
− e−2 2i
Z
x sin(2x)
dx = Im 2πi
−∞ x4 + 16 8 2i
√
πe−2 2 √
= sin(2 2).
4

2
6. f (z) = 1/(z√ − 2z + 6) has one simple pole in the upper half-plane, and it
is z1 = 1 + 5i. The residue there is
√ 1
Res(f, 1 + 5i) = √
2 5i
so Z ∞
1 π
dx = √ .
−∞ x2 − 2x + 6 5

7. First use the identity

1
cos2 (x) = (1 + cos(2x))
2
to write
∞ ∞
cos2 (x)
Z Z
1 1 + cos(2x)
2 2
dx = dx.
−∞ (x + 4) 2 −∞ (x2 + 4)2
Let
1 + e2iz
f (z) = .
(z 2 + 4)2
Then f has a pole of order 2 in the upper half-plane at 2i. Compute

d 1 + e2iz 1 + 5e−1
 
Res(f, 2i) = lim 2
= .
z→2i dx (z + 2i) 32i

Then
∞
cos2 (x) 1 + 5e−1
Z   
1
dx = Re 2πi
−∞ (x2 + 4)2 2 32i
π
= (1 + 5e−4 ).
32

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 510 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM

8. Complex methods will work with this integral, but it is easier to observe
that, with the change of variables θ = 2π − ϕ,
Z 2π Z π
sin(θ) sin(ϕ)
dθ = − dϕ.
π 2 + sin(θ) 0 2 + sin(ϕ)
Then Z 2π
sin(θ)
dθ = 0.
0 2 + sin(θ)
2 2 2
9. Let f (z) = z /(z + 4) . The only singularity of f in the upper half-plane
is 2i, which is a double pole. Compute
z2
 
d i
Res(f, 2i) = lim 2
=− .
z→2i dz (z + 2i) 8
Then ∞
x2
Z  
i π
dx = 2πi − = .
−∞ (x2 + 4)2 8 4

10. Let
eiβx
f (z) = .
(z 2 + α2 )2
The only singularity of f in the upper half-plane is a double pole at αi.
We need
eiαz
 
d
Res(f, αi) = lim
z→αi dz (z + αi)2
(αβ + 1)e−αβ
=− i.
4α3
Then
∞
(αβ + 1)e−αβ
Z  
cos(βx)
dx = 2πi − i
−∞ (x2 + α2 )2 4α3
(αβ + 1)eαβ π
= .
2α3

11. Let
eiαz
.
f (z) =
z2 + 1
The only singularity f has in the upper half-plane is a simple pole at i.
Compute
e−α
Res(f, i) = .
2i
Then Z ∞  −α 
cos(αx) e
2
dx = 2πi = πe−α .
−∞ x + 1 2i

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22.3. EVALUATION OF REAL INTEGRALS 511

12. Let
z 2 eiαz
f (z) = .
(z 2 + β 2 )2
This function has a double pole at βi in the upper half-plane. The residue
there is
d z 2 eiαz
Res(f, βi) = lim
z→βi dz (z + βi)2

= lim (2zeiαz (z + βi)−2 + iz 2 αz(z + βi)−2 − 2z 2 eiαz (z + βi)−3 )

z→βi

= 2βi(2βi)−2 + iα(−β)2 e−αβ (2βi)−2 − 2(−β 2 )e−αβ (2βi)−3

e−αβ
= i(αβ − 1).
4β
Then
∞  −αβ
x2 cos(αx)
Z 
e
dx = 2πi i(αβ − 1
−∞ (x2 + β 2 )2 4β
π −αβ
= e (1 − αβ).
αβ
13. Begin with
Z 2π
1
2 cos2 (θ) + β 2 sin2 (θ)
dθ
0 α
I
1 1
= 2 (z + 1/z)2 /4 − β 2 (z − 1/z)2 /4 iz
dz
γ α
I
4 z
= dz.
i γ (α2 − β 2 )z 4 + 2(α2 + β 2 )z 2 + (α2 − β 2 )
Singularities of the integrand satisfy
β−α β−α
z2 = or z 2 = .
β+α β+α
Because α and β are positive,
β−α β+α
< 1 and > 1.
β+α β−α
The simple poles enclosed by the unit circle are the square roots z1 and z2
of (β − α)/(β + α). The residue of the integrand at each of these poles can
be computed using Corollary 22.1. Omitting the arithmetic, we obtain
1
Res(f, zj ) =
8αβ
for j = 1, 2. Then
Z 2π
1 4 2 2π
2 dθ = (2πi) = .
0 α2 cos2 (θ) 2
+ β sin (θ) i 8αβ αβ

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 512 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM

14. Let
1
g(θ) = .
α + sin2 (θ)
Write
Z π/2 Z π/2 Z π
g(θ) dθ = g(θ) dθ + g(θ) dθ
0 0 π/2
Z 3π/2 Z 2π
+ g(θ) dθ + g(θ) dθ.
π 3π/2

In the second, third and fourth integrals on the right, put, respectively,

θ = π − u, θ = π + u, θ = 2π − u.

This leads to
π/2
1 2π
Z Z
1
g(θ) dθ = dθ
0 4 0 α + sin2 (θ)
I
1 1 1
= dz
4 γ α − (z − 1/z)2 /4 iz
I
z
=i 4 − (2 + 4α)z 2 + 1
dz.
γ z

The integrand has two simple poles bounded by the unit circle, and satis-
fying p
zj2 = (1 + 2α) − 2 α(α + 1).
Omitting the arithmetic, the residues at each of these poles is the same,
given by
−1
Res(f, zj ) = p .
8 α(α + 1)
Then
" #
π/2
−2
Z
1
dθ = i(2πi) p
0 α + sin2 (θ) 8 α(α + 1)
π
= p .
2 α(α + 1)

15. Let Γ be the given rectangular path. The four sides are:

Γ1 :z = t, −R ≤ t ≤ R,
Γ2 :z = R + it, 0 ≤ t ≤ β,
Γ3 :z = t + iβ, −R ≤ t ≤ R,
Γ4 :z = −R + it, 0 ≤ t ≤ β.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22.3. EVALUATION OF REAL INTEGRALS 513

These are, respectively, the lower side, right side, top side and left side of
the rectangle. In carrying out the integrations, limits of integration must
be consistent with counterclockwise orientation of Γ.
2
Because e−z is differentiable for all z, then by Cauchy’s theorem,
I 4 Z
2 X 2
e−z dz = 0 = e−z dz.
Γ j=1 Γj

Evaluate each of the four integrals in the sum on the right as follows.
Z Z R
−z 2 2
e dz = e−t dt,
Γ1 −R

Z Z β
−z 2 2
+2Rti−t2 )
e dz = e−(r i dt
Γ2 0
Z β
2 2
= ie−R et [cos(2Rt) − i sin(2Rt)] dt,
0

Z Z R
−z 2 2
+2βti−β 2 )
e dz = e−(t dt
Γ3 −R
Z R
2 2
= e−β e−t [cos(2βt) − i sin(2βt)] dt,
−R

and
Z Z 0
−z 2 2
−2Rti−t2 )
e dz = e−(R i dt
Γ4 β
Z β
2 2
= ie−R [et [− cos(2βt) − i sin(2βt)] dt.
0
2
Now let R → ∞. The terms having a factor of e−R go to zero in the
limit, and upon adding these integrals over the sides of the rectangle, we
obtain, using x as the variable of integration on the line,
Z ∞ Z ∞
2 2
e−x dx − e−β [cos(2βx) − i sin(2βx)] dx = 0.
−∞ −∞
2
Now, e−x sin(2βx) is an odd function on the real line, so
Z ∞
2
e−x sin(2βx) dx = 0.
−∞

We are therefore lift with

Z ∞ Z ∞
2 2
eβ cos(2βx) dx = e−x dx.
−∞ −∞

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 514 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM

Finally, use the known result that

Z ∞
2 √
e−x dx = π
−∞

to conclude that
Z ∞ √ −β 2
2
e−x cos(2βx) dx = πe .
−∞

2
Finally, because e−x cos(2βx) is an even function on the real line, then
Z ∞ √
−x2 π −β 2
e cos(2βx) dx = e .
0 2

16. Denote the given closed curve as Γ. By Cauchy’s theorem,

I
2
eiz dz = 0.
Γ

Γ consists of the segment Γ1 on the x− axis, the circular arc Γ2 , and the
segment Γ3 from the end of this arc to the origin.
On Γ1 , z = x and
Z Z R
iz 2
e dz = [cos(x2 ) + i sin(x2 )] dx.
Γ1 0

On Γ2 , z = Reiθ and
Z Z π/4
iz 2 2
e dz = eiR e2iθ dθ.
Γ2 0

On Γ3 , z = reiπ/4 and
Z Z 0
iz 2 2
e dz = e−r eiπ/4 dr,
Γ3 R

with the limits of integration set to maintain counterclockwise orientation

around Γ.
Now we want to take the limit as R → ∞. The integral over Γ3 has limit
2
zero because of the factor e−r . The integral over Γ1 only has R in its
upper limit of integration and becomes an integral over the half-line. The
integral over Γ2 is not as obvious. Make the change of variable u = 2θ to
get
1 π/2 iR2 cos(u)−R2 sin(u)
Z Z
2
eiz dz = e iReiu/2 du.
Γ2 2 0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 22.3. EVALUATION OF REAL INTEGRALS 515

Then
Z Z π/2
2 R 2 2
eiz dz ≤ eiR cos(u)
eiu/2 e−R sin(u)
du
Γ2 2 0

R π/2 −R2 sin(u)

Z
= e du
2 0
R π 
≤
2 2πR2
π
= →0
4R
as R → ∞. In the limit, we therefore obtain
Z ∞ Z ∞
2
2 2
[cos(x ) + i sin(x )] dx − e iπ/4
e−x dx = 0.
0 0

Because Z ∞ √
−x2 π
e dx = ,
0 2
and
1
eiπ/4 = √ (1 + i),
2
we finally have
Z ∞ Z ∞ √
2 2 π
cos(x ) dx + i sin(x ) dx = √ (1 + i).
0 0 2 2
From the real and imaginary parts of both sides of this equation, we have
Z ∞ Z ∞ r
1 π
cos(x2 ) dx = sin(x2 ) dx = .
0 0 2 2

17. First observe that, because the integrand is an even function,

Z ∞
1 ∞ x sin(αx)
Z
x sin(αx)
dx = dx.
0 x4 + β 4 2 −∞ x4 + β 4
Now,
zeiαz
f (z) =
z4 + β4
has simple poles in the upper half-plane at z1 = βeiπ/4 and z2 = βe3πi/4 .
Compute the residues of f at these poles:
 iαz 
ze eiαβzk
Res(f, zk ) = = .
4z 3 z=zk 4zk2
In particular,
1 iαβeiπ/4 1 3iπ/4
Res(f, z1 ) = e and Res(f, z2 ) = eiαβe .
4β 2 i −4β 2 i

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 516 CHAPTER 22. SINGULARITIES AND THE RESIDUE THEOREM

Then
∞
2πi  iαβ(1+i)/√2 √ 
Z  
x sin(αx) 1 iαβ(−1+i)/ 2 1
dx = Im e − e
0 x4 + β 4 2 4β 2 i
√
−αβ/ 2
 
πe αβ
= 2
sin √ .
2β 2

18. First write

Z 2π Z π Z 2π
1 1 1
dθ = dθ+ dθ.
0 (α + β cos(θ))2 0 (α + β cos(θ))2 π (α + β cos(θ))2

In the last integral on the right, put θ = 2π − u to show that the two
integrals on the right are equal. Therefore
Z π
1 2π
Z
1 1
2
dθ = dθ
0 (α + β cos(θ)) 2 0 (α + β cos(θ))2
I
1 1 1
= 2
dz
2 γ (α + β(z + 1/z)/2) iz
I
2 z
= dz.
i γ βz + 2αz + β)2
2

The function
z
f (z) =
(βz 2 + 2αz + β)2
has double poles at the zeros of βz 2 + 2αz + β, which are
p p
−α + α2 − β 2 −α − α2 − β 2
z1 = and z2 = .
β β
Because z2 lies outside the unit disk, we need only the residue at z1 :
 
d z
Res(f, z1 ) = lim
z→z1 dz β 2 (z − z1 )2
αβ 2
 
1
= 2
β 4(α2 − β 2 )3/2
α
= .
4(α − β 2 )3/2
2

Then
Z π
1 2 α
dθ = (2πi)
0 (α + β cos(θ))2 i 4(α2 − β 2 )3/2
πα
= 2 .
(α − β 2 )3/2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 Chapter 23

Conformal Mappings and

Applications

23.1 Conformal Mappings

For Problems 1–3, the image of the given rectangle is given as a graph for each
part of the problem.
1. (a) The rectangle defined by 0 ≤ x ≤ π, 0 ≤ y ≤ π maps to the sector
1 ≤ r ≤ eπ , 0 ≤ θ ≤ π.
See Figure 23.1.
(b) This rectangle maps to the sector
1 π π
≤ r ≤ e, − ≤ θ ≤ .
e 2 2
See Figure 23.2.
(c) The rectangle maps to the sector
π
1 ≤ r ≤ e, 0 ≤ θ ≤ .
4
See Figure 23.3.
(d) The rectangle maps to the sector
e ≤ r ≤ e2 , 0 ≤ θ ≤ π.
See Figure 23.4.
(e) The rectangle maps to the sector
1 π π
≤ r ≤ e2 , − ≤ θ ≤ .
e 2 2
See Figure 23.5.

517

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 518 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

Figure 23.1: Image of the rectangle 0 ≤ x ≤ π, 0 ≤ y ≤ π under w = ez .

Figure 23.2: Image of −1 ≤ x ≤ 1, −π/2 ≤ y ≤ π/2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.1. CONFORMAL MAPPINGS 519

Figure 23.3: Image of the rectangle 0 ≤ x ≤ 1, 0 ≤ y ≤ π/4.

Figure 23.4: Image of the rectangle 1 ≤ x ≤ 2, 0 ≤ y ≤ π.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 520 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

Figure 23.5: Image of the rectangle −1 ≤ x ≤ 2, −π/2 ≤ y ≤ π/2.

2. To determine the image of a rectangular area with sides parallel to the

axes, we need to know the images of horizontal and vertical lines under
this mapping. First write

w = cos(z) = cos(x + iy) = u + iv = cos(x) cosh(y) − i sin(x) sinh(y),

so
u = cos(x) cosh(y) and v = − sin(x) sinh(y).
This means that points x, y) map to points (cos(x) cosh(y), − sin(x) sinh(y))
in the w−plane.
Consider the images of vertical and horizontal lines, beginning with a
vertical line x = a. If cos(a) and sin(a) are nonzero, then

u2 v2
− = 1.
cos2 (a) sin2 (a)

This is the image of one branch of a hyperbola in the w−plane, because

cosh(y) > 0, forcing u > 0 if cos(a) > 0 and u < 0 if cos(a) < 0.
Next consider the case that sin(a) = 0 or cos(a) = 0.
If sin(a) = 0, then a = nπ, with n an integer, and cos(a) = (−1)n . If n
is even, say n = 2m, then cos(a) = 1. Because cosh(y) ≥ 1, the vertical
line x = 2mπ maps to the interval u ≥ 1, v = 0 on the real line in the

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.1. CONFORMAL MAPPINGS 521

Figure 23.6: Image of the rectangle 0 ≤ x ≤ 1, 1 ≤ y ≤ 2 under w = cos(z).

Figure 23.7: Image of the rectangle π/2 ≤ x ≤ π, 1 ≤ y ≤ 3.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 522 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

Figure 23.8: Image of the rectangle 0 ≤ x ≤ π, −π/2 ≤ y ≤ π.

w−plane. If n is odd, say n = 2m + 1, then cos(a) = −1 and a vertical

line x = (2m + 1)π maps to the interval u ≤ −1, v = 0 in the w−plane.
For the case cos(a) = 0, suppose a = (2n + 1)π/2, with n any integer.
The image of a point on the line x = a is (0, − sin((2n + 1)π/2) sinh(y),
varying over the entire imaginary axis in the w−plane as y varies over all
real values.
This takes care of images of horizontal lines. Now look at a horizontal line
y = b. The image of a point on this line is a point (u, v) with
u = cos(x) cosh(b), y = − sin(x) sinh(b).
If b 6= 0, these points lie on the ellipse
u2 v2
+ = 1.
cosh2 (b) sinh2 (b)
The image of a horizontal x = b 6= 0 is an ellipse.
If b = 0, then y = 0 is the real line. Now w = u = cos(x), and the image
of the real line is the interval −1 ≤ u ≤ 1 on the real line in the w−plane.
For the images of the rectangles of parts (a) through (e), see Figures 23.6–
23.10, respectively.
3. For the images of the given rectangles of parts (a) through (e), see Figures
23.11–23.15, respectively.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.1. CONFORMAL MAPPINGS 523

Figure 23.9: Image of the rectangle π ≤ x ≤ 2π, 1 ≤ y ≤ 2.

Figure 23.10: Image of the rectangle 0 ≤ x ≤ π/2, 0 ≤ y ≤ 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 524 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

Figure 23.11: Image of the rectangle 0 ≤ x ≤ π/2, 0 ≤ y ≤ π/2.

Figure 23.12: Image of the rectangle π/4 ≤ x ≤ π, 0 ≤ y ≤ π/2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.1. CONFORMAL MAPPINGS 525

Figure 23.13: Image of the rectangle 0 ≤ x ≤ 1, 0 ≤ y ≤ π/6.

Figure 23.14: Image of the rectangle π/2 ≤ x ≤ 3π, 0 ≤ y ≤ π/2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 526 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

Figure 23.15: Image of the rectangle 0 ≤ x ≤ π/2, 0 ≤ y ≤ 1.

Figure 23.16: Image of the rectangle 1 ≤ x ≤ 2, 1 ≤ y ≤ 2.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.1. CONFORMAL MAPPINGS 527

4. Write z = reiθ so w = z 2 = r2 e2iθ . If r varies from 0 to ∞, so does r2 .

And as θ varies from π/4 to 5π/4, 2θ varies from π/2 to 5π/2. This is an
interval of length 2π. Therefore the image of the sector π/4 ≤ θ ≤ 5π/4
is the entire plane.
5. The analysis proceeds like that of Problem 4. Let z = reiθ . If π/6 ≤
θ ≤ π/3, then π/2 ≤ 3θ ≤ π, so image points under this mapping lie in
the second quadrant. It is routine to verify that this mapping is onto the
second quadrant.
6. Let u = reiθ . Then
 
1 1
w = u + iv = reiθ + e−iθ .
2 e
Using Euler’s formula, we obtain
   
1 1 1 1
u= r+ cos(θ) and v = r− sin(θ).
2 r 2 r
It is routine to check that
 2  2
u v
1 + 1 =1
2 (r + 1/r) 2 (r − 1/r)

assuming that r 6= 1. This is an ellipse in the w−plane. Because r + 1/r >

r − 1/r, this ellipse has foci at (±c, 0), where
" 2  2 #
2 1 1 1
c = r+ − r− = 1.
4 r r

This means that a circle of radius r 6= 1 maps to an ellipse with foci (±1, 0)
in the w−plane.
If r = 1, then v = 0 and u = 2 cos(θ), so the image of the unit circle about
the origin in the x, y−plane is the interval [−2, 2] on the real axis in the
w−plane.
7. Using some of the analysis from Problem 6, a half-line θ = k maps to
points u + iv with
   
1 1 1 1
u= r+ cos(k), v = r− sin(k).
2 r 2 r
If sin(k) 6= 0 and cos(k) 6= 0, then a little algebraic manipulation shows
that
u2 v2
− = 1.
cos2 (k) sin2 (k)
This is the equation of a hyperbola with foci (±c, 0), where

c2 = cos2 (k) + sin(k) = 1.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 528 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

Figure 23.17: Image of the rectangle in Problem 8(a) with α = 2.

Finally, the cases cos(k) = 0 and sin(k) = 0 must be considered separately.

If cos(k) = 0, then k = (2n + 1)π/2. Now u = 0 and −∞ < v < ∞, so the
image is the imaginary axis in the x−plane.
If sin(k) = 0, then k = nπ with n an integer. Now v = 0 and
 
1 1
u= r+ (−1)n .
2 r
This is the half-interval u ≥ 1 on the real axis in the w−plane if n is even,
and u ≤ −1 if n is odd.
8. (a) First let w = cos(z). The analysis of Problem 2 tells us the images
of vertical and horizontal lines. Figure 23.16 shows the image of D for
α = 2. Different choices of α result in different images.
(b) For w = sin(z), use some of the analysis of Problem 3. Figure 23.17
shows the image of D for α = 2.
9. Write
w = 2z 2 = 2(x + iy)2 = 2(x2 − y 2 ) + 4ixy = u + iv.
Then vertical line x = 0 maps to u = −2y 2 , v = 0, so the image is the
negative u− axis. Other vertical lines x = a 6= 0 map onto parabolas
v2
u = 2a2 − .
2a2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.1. CONFORMAL MAPPINGS 529

Figure 23.18: Image of the rectangle in Problem 8b with α = 2.

The horizontal line y = 0 maps to u = 2y 2 ≥ 0, the positive u−axis.

Other horizontal lines y = b 6= 0 map onto the parabolas

v2
u= − 2b2 .
8b2
Figure 23.18 shows the image of the rectangle 0 ≤ x ≤ 3/2, −3/2 ≤ y ≤
3/2.

10. Let w = ez = ex+iy for all real x and for 0 ≤ y ≤ 2π. If we write

w = ex cos(y) + iex sin(y),

then y varies over an entire period of sin(y) and cos(y), so every point in
the w−plane, except the origin, is the image of a point in the z−plane
under this mapping.

11. If Re(z) = −4, then (z + z)/2 = −4, so z + z = −8. Now, w = 2i/z, so

z = 2i/w, so
2i 2i
z+z = − = −8.
w w
Multiply this by ww and rearrange terms to obtain

8ww − 2i(w − w) = 0.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 530 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

Figure 23.19: Image of the rectangle in Problem 9 with α = 2.

Now put w = u + iv to get

2(u2 + v 2 ) + v = 0,

or  2
2 1 1
u + v+ = .
4 4
This is the equation of a circle of radius 1/2 centered at (0, −1/4) in the
w− plane. This is the image of the line x = −4 under the given mapping.

12. If w = 2iz − 4, then

w+i
z= .
2i
Now Re(z) = (z + z)/2 becomes

w+4 w+4
− = 10.
2i 2i
Let w = u + iv to obtain
w−w
= Im(w) = v = 10.
2i
This is a horizontal line in the w− plane.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.1. CONFORMAL MAPPINGS 531

13. From the mapping, solve for z:

−1
z= .
w+i
Substitute this into the given line to obtain
   
1 −1 1 1 −1 1
− + + = 1.
2 w+i w−i 2i w + i w − i
Multiply this equation by 2i(w + i)(w − i) and rearrange terms, putting
w = u + iv to obtain
4(u2 + v 2 ) + 7v + u = 3.
Complete the square to write this as
 2  2
1 7 1
u+ + v+ = .
8 8 32
√
This is a circle of radius 1/2 2 and center (−1/8, −7/8), and is the image
of the given line.
14. Solve the mapping for z to obtain
−w − 1 + i
|z| = 4 = .
2w − 1
Then
|w + 1 − i| = 4|2w − 1|.
Put w = u + iv to get
(u + 1)2 + (v − 1)2 = 16(2u − 1)2 + 64v 2 .
Rearrange terms to write this as
 2  2
11 1 208
u− + v+ = .
21 63 3969
p
This is the equation of a circle of radius 208/3969 and center (11/21, −1/63).
15. Invert the mapping to obtain
5 + iw
z= .
2−w
Then
 
5 − v + iu
z − z = 2Re(z) = 2Re
2 − u − iv
2((5 − v)(2 − u) − uv)
=
(u − 2)2 + v 2
20 − 4v − 10u
= .
(u − 2)2 + v 2

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 532 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

Next,
1 (2 − u)u + (5 − v)v
(z − z) = Im(z) = .
2i (u − 2)2 + v 2
Substitute these into the equation of the given line and clear fractions to
obtain  2
19 377
(u − 1)2 + v + = .
4 16
√
This is the equation of a circle with radius 377/4 and having center
(1, −19/4).
16. From the mapping, obtain
−2
z= .
w − 3i − 1
Substitute this into |z − i| = 1 to get
−2
− i = 1.
w − 3i − 1
Then
|w − 3i − 1| = | − 2 − iw − 3 + i|.
Let w = u + iv in this expression to obtain

(u − 1)2 + (v − 3)2 = (u − 1)2 + (v − 5)2 .

From this, v = 4. The map is a translation followed by an inversion that

sends the given circle to the vertical line v = 4.
17. Substitute the given values into equation (23.1) to obtain

(1 − w)(1 + 2i)(−1)(3 − z) = (1 − z)(1)(1 + i)(1 − i − w).

Solve for w to obtain

(1 + 4i)z − (3 + 8i)
w= .
(2 + 3i)z − (4 + 7i)

18. Substitute the given values into equation (23.1) and solve for w to obtain
(1 + i)z − (2 + 2i)
w= .
(3 − i)z − 2

19. Here w3 = ∞ so use equation (23.2), obtaining

(1 + i − w)(1 − 2i)(4 − z) = (1 − z)(−2 + 2i)(4 − 2i).

Then
(33 + i)z − (48 + 16i)
w= .
5(z − 4)

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.1. CONFORMAL MAPPINGS 533

20. Using equation (23.2), we obtain

(9 − 7i) − (21 + 27i)
w= .
13(z + 1)

21. Using equation (23.1), we find that

(3 + 22i) + 4 − 75i
w= .
(2 + 3i) − (21 − 4i)

22. Let f be a conformal mapping from the z−plane to the w−plane. Let g
be a conformal mapping from the w−plane to the Z−plane. We want to
show that g ◦ f is conformal.
Certainly g ◦ f is differentiable. Let C1 and C2 be paths in D intersecting
at P at an angle θ (the angle between their tangents at P ). Because f
is conformal, f (C1 ) and f (C2 ) are paths in the w−plane intersecting at
angle θ at f (P ). Because g is conformal, g(f (C1 )) and g(f (C2 )) are paths
in the Z−plane intersecting at the same angle θ. Therefore g ◦ f preserves
angles.
Further, g ◦ f preserves orientation. If θ is the angle between C1 and C2 ,
going counterclockwise, then this sense of orientation is preserved by f ,
then by g, and is therefor also preserved by g ◦ f .
Therefore g ◦ f is conformal.
23. If we require that a conformal mapping be differentiable, then immediately
T (z) = z is disqualified. But it is also easy to see directly that this
mapping reverses orientation. For example, let C1 be the nonnegative real
axis and C2 the nonnegative imaginary axis. The sense of rotation from
C1 to C2 is counterclockwise. But T maps C1 to itself and C2 to the
negative imaginary axis, reversing the orientation to clockwise. Therefore
T is not conformal.
24. Suppose T is a bilinear transformation that is not the identity map and
not a translation. Write
az + b
T (z) = .
cz + d
If z is a fixed point of T , then
az + b
T (z) = z = .
cz + d
Then
cz 2 + (d − a)z − b = 0.
If c 6= 0, this is a quadratic equation with two roots, so T has two fixed
points (if these roots are distinct) and one fixed point (if the roots are
repeated). This argument fails if c = 0, in which case T is a translation
a b
T (z) = z+
d d

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 534 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

if b 6= 0. If T cannot be a translation, then b = 0 and then T (z) =

(a/d)z = kz. If a = d this is the identity mapping T (z) = z. If a 6= d,
then T is a magnification/rotation, which has only 0 as a fixed point.
25. Let
az + b
T (z) = .
cz + d
By the argument of Problem 24, if T is not a translation or the identity
mapping, then T can have at most two fixed points. Therefore, if T has
three fixed points, then T is either a translation or the identity map. But
a translation has no fixed point, so in this case T must be the identity
map.
26. First, observe that, if
az + b
T (z) =
cz + d
is a bilinear map, then ad−bc 6= 0, which guarantees that T has an inverse.
This inverse is
−dw + b
z = T −1 (w) =
cw − a
which is also bilinear. The composition T −1 ◦ T is the identity map I of
the z−plane to itself, I(z) = z.
Now suppose T (zj ) = S(zj ) for j = 1, 2, 3. Then

(S −1 ◦ T )(zj ) = S −1 (T (zj )) = S −1 (S(zj ))

= I(zj ) = zj .

Then S −1 ◦ T has three fixed points and is therefore the identity map:

S −1 ◦ T = I.

Then

S = S ◦ I = S ◦ (S −1 ◦ T )
= (S −1 ◦ S) ◦ T = I ◦ T = T.

27. Given z2 , z3 , z4 , let P be the unique bilinear transformation that maps

z2 → 1, z3 → 0, z4 → ∞.

Then by definition of the cross ratio,

P (z1 ) = [z1 , z2 , z3 , z4 ].

Now let T be any bilinear transformation. Then

[T (z1 ), T (z2 ), T (z3 ), T (z4 )] = R(T (z1 )),

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.2. CONSTRUCTION OF CONFORMAL MAPPINGS 535

where R is the unique bilinear transformation that maps

T (z2 ) = 1, T (z3 ) = 0, T (z4 ) = ∞.

Then R ◦ T = P . Then

[T (z1 ), T (z2 ), T (z3 ), T (z4 )] = R(T (z1 ))

= P (z1 ) = [z1 , z2 , z3 , z4 ].

28. Let
z3 − z4 z − z2
w = T (z) = 1 − .
z3 − z2 z − z4
A routine calculation shows that

w2 = T (z2 ) = 1, w3 = T (z3 ) = 0, and T (z4 ) = ∞.

Because three points and their images uniquely determine a bilinear trans-
formation, T is the unique bilinear transformation mapping z2 → 1,
z3 → 0 and z4 → ∞. Therefore

[z1 , z2 , z3 , z4 ] = T (z4 ).

29. In the definition of cross ratio, w2 , w3 and w4 all lie on an (extended)

line, the real axis. Because bilinear transformations map lines/circles to
lines/circles, then [z1 , z2 , z3 , z4 ] is real exactly when z1 , z2 , z3 , z4 ] all lie on
the same line or circle.

23.2 Construction of Conformal Mappings

If a conformal mapping is requested between two domains, there will in gen-
eral be many different possible mappings. In each of the solutions below, one
mapping is found, but other approaches may yield other suitable mappings.

1. Both domains are open disks, having radii 3 and 6, and different centers,
0 and 1 − i, respectively. We can map |z| < 3 ont |w − 1 + i| < 6 by using
a scaling factor of 2 and a translation to superimpose the center of the
initial domain onto the center of the image domain. Thus compose

z → 2z → 2z + 1 − i.

One mapping that does what we want is

w = 2z + 1 − i.

Now
|w − 1 + i| = 2|z| = 2(3) = 6
if |z| = 3.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 536 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

2. First perform an inversion (to map the interior of the unit disk to the
exterior), w1 = 1/z. Then expand by a factor of 18 by w2 = 18w1 (18
because (1/3)(18) = 6, the radius of the target disk, then, then translate
to the new center w3 = w2 + 1 − i. Putting these together, we have:

18 (1 − i)z + 18
w= +1−i= .
z z

3. We will need an inversion at some stage because we are mapping the

interior of a disk to the exterior of another disk. First translate by w1 =
z + 2i, so the image disk in the w−plane has the origin as its center. Next
invert by
1
w2 = .
z + 2i
Next scale by a factor of 2 to match the radii of the bounding circles:
2
w3 = 2w2 = .
z + 2i
Finally, translate to have center 2:
2 3z + 2 + 6i
w4 = w3 + 3 = +3= .
z + 2i z + 2i

4. A mapping of the half-plane Re(z) > 1 onto the half-plane Im(z) > −1
can be achieved by first rotating counterclockwise by π/2 (by w1 = iz),
and then shifting down two units by w2 = w1 − 2i. This is all done by the
mapping
w = iz − 2i = (z − 2)i.

5. We can map the line Re(z) = 0 to the circle by |w| = 4 by a bilinear

transformation. The domain Re(z) < 0 consists of numbers to the left of
the imaginary axis, which is the boundary of this domain. Choose three
points on this axis, ordered upward so the region Re(z) < 0 is on the
left as we walk up the line. Next choose three points on the image circle
|w| = 4, counterclockwise so the interior of the circle is on the left as we
walk counterclockwise around the circle. Convenient choices are

z1 = −i, z2 = 0, z3 = i and w1 = −4i, w2 = 4, w3 = −4i.

Using equation (23.1), we find the bilinear transformation mapping zj →

wj :  
1+z
w = T (z) = 4 .
1−z
As a check, z = −1, which has negative real part, maps to 0, interior to
the disk |w| < 4. Thus T maps Re(z) < 1 to |w| < 4, rather than to the
exterior |w| > 4.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.2. CONSTRUCTION OF CONFORMAL MAPPINGS 537

6. The domain Im(z) > −4 consists of all x + iy lying above the horizontal
line y = −4. The boundary of this domain is this line. We want to map
this domain onto |w − i| > 2, the exterior of the circle of radius 2 and
center i in the w−plane. The circle |w − i| = 2 is the boundary of this
domain. Choose three points on the line y = −4 in the z−plane, ordered
from left to right so the domain is on the left. Choose three points on
the circle in the w−plane, clockwise so as we walk around this circle, the
exterior is on the left. Convenient choices are
z1 = −1 − 4i, z2 = −4i, z3 = 1 − 4i and w1 = 3i, w2 = 2 + i, w3 = −i.
The bilinear transformation mapping zj → wj (use equation (23.1)) is
(−2 + i)z − (3 + 10i)
w= .
z + 3i
7. Because the boundary of the wedge in the w−plane is not a circle or line,
a bilinear transformation will not work here. However, wedges suggest
polar representations. Let z = reiθ for 0 < θ < π. These are points in the
upper half-plane. Let
w = z 1/3 = r1/3 eiθ/3 = ρeiϕ .
Here ρ > 0 and 0 ≤ ϕ ≤ π/3. This mapping is conformal because
dw 1
= z −2/3 6= 0
dz 3
for z in the upper half-plane, and the mapping takes the open upper half-
plane onto the open wedge 0 < θ < π/2.
8. Let z = x + iy = reiθ with y > 0. Then arg(z) = θ is unique (restricted
to 0 ≤ θ < 2π). Further,
w = ln(r) + iθ.
Because r can be any positive number, ln(r) varies over all the real num-
bers. Further,
Im(z) = θ in (0, π).
Therefore log(z) is in the strip 0 < Im(z) < π.
To show that this mapping is onto, choose any w = u + iv in this strip.
Let z = rew . Then
Im(z) = eu sin(v) > 0
and
log(z) = u + iv = w.
The mapping is therefore onto. Finally,
d 1
(log(z)) = 6= 0,
dz z
so the mapping is conformal.

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 538 CHAPTER 23. CONFORMAL MAPPINGS AND APPLICATIONS

9. The solution of this problem requires some familiarity with the gamma
and beta functions.
To show that f maps the upper half-plane onto the given rectangle, we
will evaluate the function at −1, 0, 1 and ∞ and then show that these are
the vertices of that rectangle.
First, it is obvious that f (0) = 0. Next,
Z 1
f (1) = 2i (ξ 2 − 1)−1/2 ξ −1/2 dξ
0
1
(1 − ξ 2 )−1/2 −1/2
Z
= 2i ξ dξ
0 i
Z 1
=2 (1 − ξ 2 )−1/2 ξ −1/2 dξ.
0

Let ξ = u1/2 to obtain

Z 1
f (1) = (1 − u)−1/2 u−3/4 du
0
Γ(1/4)Γ(1/2)
= B(1/4, 1/2) = = c,
Γ(3/4)

in which B(x, y) is the beta function and Γ(x) is the gamma function.
Next, write
Z 1
f (−1) = 2i (ξ 2 − 1)−1/2 ξ −1/2 dξ.
0

Let ξ = −u to obtain
Z 1
f (−1) = 2i (1 − u2 )−1/2 u−1/2 du
0
Γ(1/4)Γ(1/2)
= iB(1/4, 1/2) = i = ic.
Γ(3/4)

Finally,
Z ∞
f (∞) = 2i (ξ + 1)−1/2 (ξ − 1)−1/2 ξ −1/2 dξ
0
Z 1
= 2i (ξ + 1)−1/2 (ξ − 1)−1/2 ξ −1/2 dξ
Z 0∞
+ 2i (ξ + 1)−1/2 (ξ − 1)−1/2 ξ −1/2 dξ.
1

The first integral in the last line of the last equation is B(1/4, 1/2). In

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
 23.2. CONSTRUCTION OF CONFORMAL MAPPINGS 539

the second integral, put ξ = 1/u to get

0  −1/2  −1/2  
1−u
Z
1+u 1/2 1
f (∞) = c + 2i u du
1 u u u2
Z 1
= c + 2i (1 − u2 )−1/2 u−1/2 du = (1 + i)u.
0

© 2018 Cengage Learning. All Rights reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.