Advanced Modern Engineering Mathematics Glyn James Solutions Manual 4th Edition
User Manual:
Open the PDF directly: View PDF 
.
Page Count: 688
| Download | |
| Open PDF In Browser | View PDF |
Solutions Manual
Advanced Modern
Engineering Mathematics
fourth edition
Glyn James
Solutions Manual
Advanced Modern
Engineering Mathematics
th
4 edition
Glyn James
ISBN 978-0-273-71925-0
c Pearson Education Limited 2011
Lecturers adopting the main text are permitted to download the manual as
required.
c Pearson Education Limited 2011
i
Pearson Education Limited
Edinburgh Gate
Harlow
Essex CM20 2JE
England
and Associated Companies throughout the world
Visit us on the World Wide Web at:
www.pearsoned.co.uk
This edition published 2011
c Pearson Education Limited 2011
The rights of Glyn James, David Burley, Dick Clements, Phil Dyke, John Searl,
Nigel Steele and Jerry Wright to be identified as authors of this work have been
asserted by them in accordance with the Copyright, Designs and Patents Act 1988.
ISBN: 978-0-273-71925-0
All rights reserved. Permission is hereby given for the material in this
publication to be reproduced for OHP transparencies and student handouts,
without express permission of the Publishers, for educational purposes only.
In all other cases, no part of this publication may be reproduced, stored
in a retrieval system, or transmitted in any form or by any means, electronic,
mechanical, photocopying, recording, or otherwise without either the prior
written permission of the Publishers or a licence permitting restricted copying
in the United Kingdom issued by the Copyright Licensing Agency Ltd,
90 Tottenham Court Road, London W1T 4LP. This book may not be lent,
resold, hired out or otherwise disposed of by way of trade in any form
of binding or cover other than that in which it is published, without the
prior consent of the Publishers.
ii
TABLE OF CONTENTS
Page
Chapter 1. Matrix Analysis
1
Chapter 2. Numerical Solution of Ordinary Differential Equations
86
Chapter 3. Vector Calculus
126
Chapter 4. Functions of a Complex Variable
194
Chapter 5. Laplace Transforms
270
Chapter 6. The z Transform
369
Chapter 7. Fourier Series
413
Chapter 8. The Fourier Transform
489
Chapter 9. Partial Differential Equations
512
Chapter 10. Optimization
573
Chapter 11. Applied Probability and Statistics
639
iii
1
Matrix Analysis
Exercises 1.3.3
1(a) Yes, as the three vectors are linearly independent and span threedimensional space.
1(b) No, since they are linearly dependent
⎡ ⎤ ⎡ ⎤
⎡ ⎤
1
1
3
⎣2⎦ − 2⎣0⎦ = ⎣2⎦
3
1
5
1(c)
No, do not span three-dimensional space. Note, they are also linearly
dependent.
2
Transformation matrix is
⎡
1
1
1 ⎣
1 −1
A= √
2 0
0
⎤ ⎡ 1
⎤⎡
√
0
1 0 0
2
1
⎦
⎣
⎦
⎣
0
0
1
0
=
√
√
2
0 0 1
2
0
√1
2
− √12
0
⎤
0
0⎦
1
Rotates the (e1 , e2 ) plane through π/4 radians about the e3 axis.
3
By checking axioms (a)–(h) on p. 10 it is readily shown that all cubics
ax3 + bx2 + cx + d form a vector space. Note that the space is four dimensional.
3(a)
All cubics can be written in the form
ax3 + bx2 + cx + d
and {1, x, x2 , x3 } are a linearly independent set spanning four-dimensional space.
Thus, it is an appropriate basis.
c Pearson Education Limited 2011
2
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
3(b)
No, does not span the required four-dimensional space. Thus a general
cubic cannot be written as a linear combination of
(1 − x), (1 + x), (1 − x3 ), (1 + x3 )
as no term in x2 is present.
3(c) Yes as linearly independent set spanning the four-dimensional space
a(1 − x) + b(1 + x) + c(x2 − x3 ) + d(x2 + x3 )
= (a + b) + (b − a)x + (c + a)x2 + (d − c)x3
≡ α + βx + γx2 + δx3
3(d) Yes as a linear independent set spanning the four-dimensional space
a(x − x2 ) + b(x + x2 ) + c(1 − x3 ) + d(1 + x3 )
= (a + b) + (b − a)x + (c + d)x2 + (d − c)x3
≡ α + βx + γx2 + δx3
3(e)
No not linearly independent set as
(4x3 + 1) = (3x2 + 4x3 ) − (3x2 + 2x) + (1 + 2x)
4
x + 2x3 , 2x − 3x5 , x + x3 form a linearly independent set and form a basis
for all polynomials of the form α + βx3 + γx5 . Thus, S is the space of all odd
quadratic polynomials. It has dimension 3.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 1.4.3
5(a) Characteristic polynomial is λ3 − p1 λ2 − p2 λ − p3 with
p1 = trace A = 12
⎤
⎡
−9
2
1
B1 = A − 12I = ⎣ 4 −7 −1 ⎦
2
3 −8
⎤
⎡
−17 −5 −7
7 ⎦
A2 = A B1 = ⎣ −18 −30
2
−5 −33
p2 =
1
trace A2 = −40
2
⎤
23 −5 −7
7⎦
B2 = A2 + 40I = ⎣ −18 10
2 −5
7
⎤
⎡
35 0
0
A3 = A B2 = ⎣ 0 35 0 ⎦
0 0 35
⎡
p3 =
1
trace A3 = 35
3
Thus, characteristic polynomial is
λ3 − 12λ2 + 40λ − 35
Note that B3 = A3 − 35I = 0 confirming check.
5(b)
Characteristic polynomial is λ4 − p1 λ3 − p2 λ2 − p3 λ − p4 with
p1 = trace A = 4
⎡
−2 −1
⎢ 0 −3
B1 = A − 4I = ⎣
−1
1
1
1
⎡
−3
4
⎢ −1 −2
A2 = A B1 = ⎣
2
0
−3 −3
⎤
1
2
1
0⎥
⎦
−3
1
1 −4
⎤
0 −3
1
−2
1⎥
⎦ ⇒ p2 = trace A2 = −2
−2 −5
2
−1
3
c Pearson Education Limited 2011
3
4
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎤
−1
4
0 −3
0 −2
1⎥
⎢ −1
= A2 + 2I = ⎣
⎦
2
0
0 −5
−3 −3 −1
5
⎤
⎡
−5
2
0 −2
1
0 −2 −4 ⎥
⎢ 1
= A B2 = ⎣
⎦ ⇒ p3 = trace A3 = −5
−1 −7 −3
4
3
0
4 −2 −7
⎤
⎡
0
0
0 −2
5 −2 −4 ⎥
⎢ 1
= A3 + 5I = ⎣
⎦
−1 −8
2
4
0
4 −2 −2
⎤
⎡
−2
0
0
0
1
0
0⎥
⎢ 0 −2
= A B3 = ⎣
⎦ ⇒ p4 = trace A4 = −2
0
0 −2
0
4
0
0
0 −2
⎡
B2
A3
B3
A4
Thus, characteristic polynomial is λ4 − 4λ3 + 2λ2 + 5λ + 2
Note that B4 = A4 + 2I = 0 as required by check.
6(a)
1
2
Eigenvalues given by 1−λ
1 1−λ = λ − 2λ = λ(λ − 2) = 0
so eigenvectors are λ1 = 2, λ2 = 0
Eigenvectors given by corresponding solutions of
(1 − λi )ei1 + ei2 = 0
ei1 + (1 − λi )ei2 = 0
Taking i = 1, 2 gives the eigenvectors as
e1 = [1 1]T , e2 = [1 − 1]T
6(b)
2
2
Eigenvalues given by 1−λ
3 2−λ = λ − 3λ − 4 = (λ + 1)(λ − 4) = 0
so eigenvectors are λ1 = 4, λ2 = −1
Eigenvectors given by corresponding solutions of
(l − λi )ei1 + 2ei2 = 0
3ei1 + (2 − λi )ei2 = 0
c Pearson Education Limited 2011
(1)
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Taking i = 1, 2 gives the eigenvectors as
e1 = [2 3]T , e2 = [1 − 1]T
6(c) Eigenvalues given by
1 − λ
0
−4
0
5−λ
4 = λ3 + 9λ2 + 9λ − 81 = (λ − 9)(λ − 3)(λ + 3) = 0
−4
4
3 − λ
So the eigenvalues are λ1 = 9, λ2 = 3, λ3 = −3.
The eigenvectors are given by the corresponding solutions of
(1 − λi )ei1 + 0ei2 − 4ei3 = 0
0ei1 + (5 − λi )ei2 + 4ei3 = 0
−4ei1 + 4ei2 + (3 − λi )ei3 = 0
Taking i = 1, λi = 9 solution is
e12
e13
e11
=−
=
= β1
8
16
−16
⇒ e1 = [−1 2 2]T
Taking i = 2, λi = 3 solution is
e21
e22
e23
=−
=
= β2
−16
16
8
⇒ e2 = [2 2 − 1]T
Taking i = 3, λi = −3 solution is
e32
e33
e31
=−
=
= β3
32
16
32
6(d)
Eigenvalues given by
1 − λ
0
−1
⇒ e3 = [2 − 1 2]T
1
2
2−λ
2 = 0
1
3 − λ
Adding column 1 to column 2 gives
1 − λ 2 − λ
2
0
2−λ
2 = (2 − λ)
−1
0
3 − λ
1 − λ 0
0
R1 −R2 (2 − λ) 0
1
2
−1
0 3 − λ
1− λ 1
2
0
1
2
−1
0 3 − λ
= (2 − λ)(1 − λ)(3 − λ)
c Pearson Education Limited 2011
5
6
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so the eigenvalues are λ1 = 3, λ2 = 2, λ3 = 1.
Eigenvectors are the corresponding solutions of (A − λi I)ei = 0
When λ = λ1 = 3 we have
⎡
−2 1
⎣ 0 −1
−1 1
⎤
2
2⎦
0
⎡
⎤
e11
⎣ e12 ⎦ = 0
e13
leading to the solution
e12
e13
e11
=−
=
= β1
−2
2
−1
so the eigenvector corresponding to λ2 = 3 is e1 = β1 [2 2 1]T , β1 constant.
When λ = λ2 = 2 we have
⎡
−1
⎣ 0
−1
1
0
1
⎤
2
2⎦
1
⎡
⎤
e21
⎣ e22 ⎦ = 0
e23
leading to the solution
e21
e22
e23
=−
=
= β3
−2
2
0
so the eigenvector corresponding to λ2 = 2 is e2 = β2 [1 1 0]T , β2 constant.
When λ = λ3 = 1 we have
⎡
0
⎣ 0
−1
1
1
1
⎤
2
2⎦
2
⎡
⎤
e31
⎣ e32 ⎦ = 0
e33
leading to the solution
e32
e33
e31
=−
=
= β1
0
2
1
so the eigenvector corresponding to λ3 = 1 is e3 = β3 [0 − 2 1]T , β3 constant.
6(e)
Eigenvalues given by
5 − λ
0
6
= λ3 − 14λ2 − 23λ − 686 = (λ − 14)(λ − 7)(λ + 7) = 0
0
11
−
λ
6
6
6
−2 − λ
so eigenvalues are λ1 = 14, λ2 = 7, λ3 = −7
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Eigenvectors are given by the corresponding solutions of
(5 − λi )ei1 + 0ei2 + 6ei3 = 0
0ei1 + (11 − λi )ei2 + 6ei3 = 0
6ei1 + 6ei2 + (−2 − λi )ei3 = 0
When i = 1, λ1 = 14 solution is
−e12
e13
e11
=
=
= β1 ⇒ e1 = [2 6 3]T
12
−36
18
When i = 2, λ2 = 7 solution is
−e22
e23
e21
=
=
= β2 ⇒ e2 = [6 − 3 2]T
−72
−36
−24
When i = 3, λ3 = −7 solution is
e31
−e32
e33
=
=
= β3 ⇒ e3 = [3 2 − 6]T
54
−36
−108
6(f)
Eigenvalues given by
−1 − λ
0
−1 − λ
−1
0
1
2−λ
1
2−λ
1 R1 +R2
−2
1
−1 − λ
1
−1 − λ
−1
0
0
0 = 0, i.e. (1 + λ)(2 − λ)(1 − λ) = 0
= (1 + λ) 1 2 − λ
−2
1
1 − λ
1 − λ
1
−2
so eigenvalues are λ1 = 2, λ2 = 1, λ3 = −1
Eigenvectors are given by the corresponding solutions of
(1 − λi )ei1 − ei2 + 0ei3 = 0
ei1 + (2 − λi )ei2 + ei3 = 0
−2ei1 + ei2 − (1 + λi )ei3 = 0
Taking i = 1, 2, 3 gives the eigenvectors as
e1 = [−1 1 1]T , e2 = [1 0 − 1]T , e3 = [1 2 − 7]T
c Pearson Education Limited 2011
7
8
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
6(g)
Eigenvalues given by
5− λ 5 − λ 5 − λ
1
1
5−λ
4
5 − λ 4 R1 + (R2 + R3 ) 2
−1
−1
−λ
−1
−λ
1
0
0
2 = (5 − λ)(3 − λ)(1 − λ) = 0
= (5 − λ) 2 3 − λ
−1
0
1− λ
4− λ
2
−1
so eigenvalues are λ1 = 5, λ2 = 3, λ3 = 1
Eigenvectors are given by the corresponding solutions of
(4 − λi )ei1 + ei2 + ei3 = 0
2ei1 + (5 − λi )ei2 + 4ei3 = 0
−ei1 − ei2 − λi ei3 = 0
Taking i = 1, 2, 3 and solving gives the eigenvectors as
e1 = [2 3 − 1]T , e2 = [1 − 1 0]T , e3 = [0 − 1 1]T
6(h)
Eigenvalues given by
1 − λ 2 − 2λ
0
−4
−2
3−λ
1
3−λ
1 R1 +2R2 0
1
2
4 − λ
2
4 − λ
1
0
0
1 = (1 − λ)(3 − λ)(4 − λ) = 0
= (1 − λ) 0 3 − λ
1
0
4 − λ
1− λ
0
1
so eigenvalues are λ1 = 4, λ2 = 3, λ3 = 1
Eigenvectors are given by the corresponding solutions of
(1 − λi )ei1 − 4ei2 − 2ei3 = 0
2ei1 + (3 − λi )ei2 + ei3 = 0
ei1 + 2ei2 + (4 − λi )ei3 = 0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
9
Taking i = 1, 2, 3 and solving gives the eigenvectors as
e1 = [2 − 1 − 1]T , e2 = [2 − 1 0]T , e3 = [4 − 1 − 2]T
Exercises 1.4.5
7(a)
Eigenvalues given by
1 − λ −1 + λ
2
1
0
3−λ
1 R1 −R2 0
3−λ
1
1
2
2 − λ
2
2− λ
1
0
0
1 = (1 − λ)[λ2 − 6λ + 5] = (1 − λ)(λ − 1)(λ − 5) = 0
= (1 − λ) 1 4 − λ
1
3
2 − λ
2 − λ
1
1
so eigenvalues are λ1 = 5, λ2 = λ3 = 1
The eigenvectors are the corresponding solutions of
(2 − λi )ei1 + 2ei2 + ei3 = 0
ei1 + (3 − λi )ei2 + ei3 = 0
ei1 + 2ei2 + (2 − λi )ei3 = 0
When i = 1, λ1 = 5 and solution is
−e12
e13
e11
=
=
= β1 ⇒ e1 = [1 1 1]T
4
−4
4
When λ2 = λ3 = 1 solution is given by the single equation
e21 + 2e22 + e23 = 0
Following the procedure of Example 1.6 we can obtain two linearly independent
solutions. A possible pair are
e2 = [0 1 2]T , e3 = [1 0 − 1]T
c Pearson Education Limited 2011
10
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
7(b)
Eigenvalues given by
−λ
−1
−1
−2
1−λ
−1
−2
2 = −λ3 + 3λ2 − 4 = −(λ + 1)(λ − 2)2 = 0
2 − λ
so eigenvalues are λ1 = λ2 = 2, λ3 = −1
The eigenvectors are the corresponding solutions of
−λi ei1 − 2ei2 − 2ei3 = 0
−ei1 + (1 − λi )ei2 + 2ei3 = 0
−ei1 − ei2 + (2 − λi )ei3 = 0
When i = 3, λ3 = −1 corresponding solution is
e31
−e32
e33
=
=
= β3 ⇒ e3 = [8 1 3]T
8
−1
3
When λ1 = λ2 = 2 solution is given by
−2e21 − 2e22 − 2e23 = 0
(1)
−e21 − e22 + 2e23 = 0
(2)
−e21 − e22 = 0
(3)
From (1) and (2) e23 = 0 and it follows from (3) that e21 = −e22 . We deduce that
there is only one linearly independent eigenvector corresponding to the repeated
eigenvalues λ = 2. A possible eigenvector is
e2 = [1 − 1 0]T
7(c)
Eigenvalues given by
1 − λ
6
6
3−λ
2 R1 −3R3 1
−1
−5
−2 − λ
1
−3
0
2 = (1 − λ)
= (1 − λ) 1 3 − λ
−1
−5
−2 − λ
4− λ
1
−1
−3 + 3λ
0
3−λ
2
−5
−2 − λ
1
0
0
1 6 − λ
2
1
−8
−2 − λ
= (1 − λ)(λ2 + λ + 4) = (1 − λ)(λ − 2)2 = 0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
11
so eigenvalues are λ1 = λ2 = 2, λ3 = 1.
The eigenvectors are the corresponding solutions of
(4 − λi )ei1 + 6ei2 + 6ei3 = 0
ei1 + (3 − λi )ei2 + 2ei3 = 0
−ei1 − 5ei2 − (2 + λi )ei3 = 0
When i = 3, λ3 = 1 corresponding solution is
−e32
e33
e31
=
=
= β3 ⇒ e3 = [4 1 − 3]T
4
−1
−3
When λ1 = λ2 = 2 solution is given by
2e21 + 6e22 + 6e23 = 0
e21 + e22 + 2e23 = 0
−e21 − 5e22 − 4e23 = 0
so that
e21
−e22
e23
=
=
= β2
6
−2
−4
leading to only one linearly eigenvector corresponding to the eigenvector λ = 2. A
possible eigenvector is
e2 = [3 1 − 2]T
7(d)
Eigenvalues given by
1− λ
7 − λ −2
−4
R1 −2R2 3
3
−λ
−2
6
6
−2 −3 − λ
1 −2
0
−2 = (1 − λ)
= (1 − λ) 3 −λ
6 −2 −3 − λ
−2 + 2λ
0
−λ
−2
−2
−3 − λ
1
0
0
3 6− λ
−2
6
10
−3 − λ
= (1 − λ)(λ − 2)(λ − 1) = 0
so eigenvalues are λ1 = 2, λ2 = λ3 = 1.
The eigenvectors are the corresponding solutions of
(7 − λi )ei1 − 2ei2 − 4ei3 = 0
3ei1 − λi ei2 − 2ei3 = 0
6ei1 − 2ei2 − (3 + λi )ei3 = 0
c Pearson Education Limited 2011
12
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
When i = 1, λ2 = 2 and solution is
e11
−e12
e13
=
=
= β1 ⇒ e1 = [2 1 2]T
6
−3
6
When λ2 = λ3 = 1 the solution is given by the single equation
3e21 − e22 − 2e23 = 0
Following the procedures of Example 1.6 we can obtain two linearly independent
solutions. A possible pair are
e2 = [0 2 − 1]T , e3 = [2 0 3]T
8
⎡
−4
(A − I) = ⎣ 2
1
⎤
−7 −5
3
3⎦
2
1
Performing
⎤a series of row and column operators this may be reduced to the form
⎡
0 0 0
⎣ 0 0 1 ⎦ indicating that (A − I) is of rank 2. Thus, the nullity q = 3 − 2 = 1
1 0 0
confirming that there is only one linearly independent eigenvector associated with
the eigenvalue λ = 1. The eigenvector is given by the solution of
−4e11 − 7e12 − 5e13 = 0
2e11 + 3e12 + 3e13 = 0
e11 + 2e12 + e13 = 0
giving
9
e11
−e12
e13
=
=
= β1 ⇒ e1 = [−3 1 1]T
−3
−1
1
⎡
1
(A − I) = ⎣ −1
−1
⎤
1 −1
−1
1⎦
−1
1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
13
Performing
⎤a series of row and column operators this may be reduced to the form
⎡
1 0 0
⎣ 0 0 0 ⎦ indicating that (A − I) is of rank 1. Then, the nullity of q = 3 − 1 = 2
0 0 0
confirming that there are two linearly independent eigenvectors associated with the
eigenvalue λ = 1. The eigenvectors are given by the single equation
e11 + e12 − e13 = 0
and two possible linearly independent eigenvectors are
e1 = [1 0 1]T and e2 = [0 1 1]T
Exercises 1.4.8
10
These are standard results.
11(a) (i)
(ii)
Trace A = 4 + 5 + 0 = 9 = sum eigenvalues;
det A = 15 = 5 × 3 × 1 = product eigenvalues;
⎡
(iii)
A−1
4
1 ⎣
−4
=
15
3
⎤
−1 −1
1 −14 ⎦ . Eigenvalues given by
3
18
−1
−1
1 − 15λ
−14 C3 −C2
3
18 − 15λ
4 − 15λ
−1
0
1 − 15λ −1
= (15 − 15λ) −4
3
3
1
4 − 15λ
−4
3
4 − 15λ
−4
3
−1
1 − 15λ
3
0
−15 + 15λ
15 − 15λ
= (15 − 15λ)(15λ − 5)(15λ − 3) = 0
confirming eigenvalues as 1, 13 , 15 .
c Pearson Education Limited 2011
14
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎤
2 −1
5 −1 ⎦ having eigenvalues given by
4 0
⎡
(iv)
4
AT = ⎣ 1
1
4 − λ
1
1
2
−1
5 − λ −1 = (λ − 5)(λ − 3)(λ − 1) = 0
4
−λ
that is, eigenvalue as for A .
⎡
11(b) (i)
8
2A = ⎣ 4
−2
2
10
−2
⎤
2
8 ⎦ having eigenvalues given by
0
8− λ
2
2
4
10 − λ 8 C1 −C2
−2
−2
−λ
1
2
2
= (6 − λ) −1 10 − λ 8
0
−2
−λ
2
2
10 − λ 8
−2
−λ
1
2
2
= (6 − λ) 0 12 − λ 10
0
−2
−λ
6−λ
−6 + λ
0
= (6 − λ)(λ − 10)(λ − 2) = 0
Thus eigenvalues are 2 times those of A; namely 6, 10 and 2.
⎡
(ii)
⎤
6
1 1
A + 2I = ⎣ 2
7 4 ⎦ having eigenvalues given by
−1 −1 2
6 − λ
1
2
7−λ
−1
−1
1
4 = −λ3 + 15λ2 − 71λ + 105 = −(λ − 7)(λ − 5)(λ − 3) = 0
2 − λ
confirming the eigenvalues as 5 + 2, 3 + 2, 1 + 2.
Likewise for A − 2I
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎡
(iii)
17
A2 = ⎣ 14
−6
8
23
−6
⎤
8
22 ⎦ having eigenvalues given by
−5
25 − λ 25 − λ 25 − λ
8
8
23 − λ 22 − λ R1 + (R2 ) + R3 ) 14
23 − λ
22
−6
−6
−5 − λ
−6
−5 − λ
1
0
0
8 = (25 − λ)(9 − λ)(1 − λ) = 0
= (25 − λ) 14 9 − λ
−6
0
1 − λ
17 − λ
14
−6
that is, eigenvalues A2 are 25, 9, 1 which are those of A squared.
12
Eigenvalues of A given by
−3
−3
1−λ
−1 R3 +R2
−1
1 − λ
−3 − λ
−3
−3
1 − λ −1
= (λ − 2) −3
0
1
−1
−3 − λ
−3
−3
−3
−3
1−λ
−1
−2 + λ 2 − λ
−3 − λ
−3
C3 +C2 (λ − 2) −3
(1 − λ)
0
1
−3 − λ
−3
0
= −(λ − 2)(λ + 6)(λ − 3) = 0
so eigenvalues are λ1 = −6, λ2 = 3, λ3 = 2
Eigenvectors are given by corresponding solutions of
(−3 − λi )ei1 − 3ei2 − 3ei3 = 0
−3ei1 + (1 − λi )ei2 − ei3 = 0
−3ei1 − ei2 + (1 − λi )ei3 = 0
Taking i = 1, 2, 3 gives the eigenvectors as
e1 = [2 1 1]T , e2 = [−1 1 1]T , e3 = [0 1 − 1]T
It is readily shown that
eT1 e2 = eT1 e3 = eT2 e3 = 0
so that the eigenvectors are mutually orthogonal.
c Pearson Education Limited 2011
−6
−λ
0
15
16
13
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Let the eigenvector be e = [a b c]T then since the three vectors are mutually
orthogonal
a + b − 2c = 0
a+b−c=0
giving c = 0 and a = −b so an eigenvector corresponding to λ = 2 is e = [1 −1 0]T .
Exercises 1.5.3
14
Taking x(0) = [1 1 1]T iterations may then be tabulated as follows:
Iteration k
x(k)
A x(k)
λ
0
1
1
1
9
10
5
10
1
0.9
1
0.5
7.6
8.7
4.3
8.7
2
0.874
1
0.494
7.484
8.61
4.242
8.61
3
0.869
1
0.493
7.461
8.592
4.231
8.592
4
0.868
1
0.492
7.457
8.589
4.228
8.589
Thus, estimate of dominant eigenvalue is λ 8.59 and corresponding eigenvector
x [0.869 1 0.493]T or x [0.61 0.71 0.35]T in normalised form.
15(a)
Taking x(0) = [1 1 1]T iterations may then be tabulated as follows:
Iteration k
x(k)
A x(k)
λ
0
1
1
1
3
4
4
4
1
0.75
1
1
2.5
3.75
3.75
3.75
2
0.667
1
1
2.334
3.667
3.667
3.667
3
0.636
1
1
2.272
3.636
3.636
3.636
4
0.625
1
1
2.250
3.625
3.625
3.625
5
0.620
1
1
2.240
3.620
3.620
3.620
6
0.619
1
1
Thus, correct to two decimal places dominant eigenvalue is 3.62 having
corresponding eigenvectors [0.62 1 1]T .
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
17
Taking x(0) = [1 1 1]T iterations may be tabulated as follows:
15(b)
Iteration k
x(k)
A x(k)
λ
0
1
1
1
4
6
11
11
1
0.364
0.545
1
2.092
3.818
7.546
7.546
2
0.277
0.506
1
1.831
3.566
7.12
7.12
3
0.257
0.501
1
1.771
3.561
7.03
7.03
4
0.252
0.493
1
1.756
3.49
6.994
6.994
5
0.251
0.499
1
Thus, correct to two decimal places dominant eigenvalue is 7 having corresponding
eigenvector [0.25 0.5 1]T .
15(c)
Taking x(0) = [1 1 1 1]T iterations may then be tabulated as follows:
Iteration k
x(k)
A x(k)
λ
0
1
1
1
1
1
0
0
1
1
1
1
0
1
1
2
−1
−1
2
2
2
1
−0.5
− 0.5
1
2.5
−1.5
−1.5
2.5
2.5
3
1
−0.6
−0.6
1
2.6
−1.6
−1.6
2.6
2.6
4
1
−0.615
−0.615
1
2.615
−1.615
−1.615
2.615
2.615
5
1
−0.618
− 0.618
1
2.618
−1.618
−1.618
2.618
2.618
6
1
− 0.618
− 0.618
1
Thus, correct to two decimal places dominant eigenvalue is 2.62 having
corresponding eigenvector [1 − 0.62 − 0.62
16
1]T .
The eigenvalue λ1 corresponding to the dominant eigenvector e1 = [1 1 2]T
is such that A e1 = λ1 e1 so
⎤⎡ ⎤
⎡ ⎤
1
3 1 1
1
⎣ 1 3 1 ⎦ ⎣ 1 ⎦ = λ1 ⎣ 1 ⎦
2
1 1 5
2
⎡
so λ1 = 6.
c Pearson Education Limited 2011
18
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Then
1 1 2
A1 = A − 6ê1 êT1 where ê1 = √ √ √
6 6 6
so
⎤
⎡
1
3 1 1
⎦
⎣
⎣
A1 = 1 3 1 − 1
2
1 1 5
⎡
1
1
2
⎤
⎡
2
2
⎦
⎣
0
2 =
−1
4
T
⎤
0 −1
2 −1 ⎦
−1
1
Applying the power method with x(0) = [1 1 1]T
⎤
1
= ⎣ 1 ⎦ = x(1)
−1
⎤
⎤
⎡
⎡
1
3
= ⎣ 3⎦ =3 ⎣ 1 ⎦
−1
−3
⎡
y(1) = A1 x(0)
y(2) = A1 x(1)
1
Clearly, λ2 = 3 and ê2 = √ [1 1 − 1]T .
3
Repeating the process
⎡
2
T
⎣
0
A2 = A1 − λ2 ê2 ê2 =
−1
0
2
−1
⎤
⎡
1
−1
⎦
⎣
1
−1 −
−1
1
1
1
−1
⎤
⎡
1
−1
⎦
⎣
−1 = −1
0
1
Taking x(0) = [1 − 1 0]T the power method applied to A2 gives
⎤
⎤
⎡
1
2
= ⎣ −2 ⎦ = 2 ⎣ −1 ⎦
0
0
⎡
y(1) = A2 x(0)
1
and clearly, λ3 = 2 with ê3 = √ [1 − 1 0]T .
2
17
The three Gerschgorin circles are
| λ − 5 |= 2, | λ |= 2, | λ + 5 |= 2
c Pearson Education Limited 2011
⎤
−1 0
1 0⎦
0 0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
19
which are three non-intersecting circles. Since the given matrix A is symmetric its
three eigenvalues are real and it follows from Theorem 1.2 that
3 < λ1 < 7 , −2 < λ2 < 2 , −7 < λ3 < 7
(i.e., an eigenvalue lies within each of the three circles).
18
The characteristic equation of the matrix A is
10 − λ
−1
0
that is
−1
2−λ
2
0
2 = 0
3 − λ
(10 − λ)[(2 − λ)(3 − λ) − 4] − (3 − λ) = 0
or
f(λ) = λ3 − 15λ2 + 51λ − 17 = 0
Taking λ0 = 10 as the starting value the Newton–Raphson iterative process
produces the following table:
f(λi )
f (λi )
i
λi
f(λi )
f (λi )
−
0
10
7
−51.00
0.13725
1
10.13725
− 0.28490
−55.1740
− 0.00516
2
10.13209
− 0.00041
−55.0149
−0.000007
Thus to three decimal places the largest eigenvalue is λ = 10.132
Using Properties 1.1 and 1.2 of section 1.4.6 we have
3
3
λi =| A |= 17
λi = trace A = 15 and
i=1
i=1
c Pearson Education Limited 2011
20
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus,
λ2 + λ3 = 15 − 10.132 = 4.868
λ2 λ3 = 1.67785
λ2 (4.868 − λ2 ) = 1.67785
so
λ22 − 4.868λ2 + 1.67785 = 0
λ2 = 2.434 ± 2.0607
that is
19(a)
λ2 = 4.491 and
λ3 = 0.373
If e1 , e2 , . . . , en are the corresponding eigenvectors to λ1 , λ2 , . . . , λn then
(KI − A)ei = (K − λi )ei so that A and (KI − A) have the same eigenvectors and
eigenvalues differ by K .
Taking x(o) =
n
αr ei then
i=1
n
x
(p)
= (KI − A)x
(p−1)
= (KI − A) x
2 (p−2)
αr (K − λr )p er
= ... =
r=1
Now K − λn > K − λn−1 > . . . > K − λ1 and
n
x
(p)
= αn (K − λn ) en +
αr (K − λr )p er
p
r=1
n−1
= (K − λn )p [αn en +
αr
r=1
K − λr
K − λn
p
er ]
→ (K − λn ) αn en = Ken as p → ∞
p
Also
(p+1)
xi
(p)
xi
→
(K − λn )p+1 αn en
= K − λn
(K − λn )p αn en
Hence, we can find λn
19(b)
Since A is a symmetric matrix its eigenvalues are real. By Gerschgorin’s
theorem the eigenvalues lie in the union of the intervals
| λ − 2 |≤ 1, | λ − 2 |≤ 2, | λ − 2 |≤ 1
i.e.
| λ − 2 |≤ 2 or 0 ≤ λ ≤ 4.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
21
Taking K = 4 in (a)
⎡
2
KI − A = 4I − A = ⎣ 1
0
1
2
1
⎤
0
1⎦
2
Taking x(0) = [1 1 1]T iterations using the power method are tabulated as follows:
Iteration k
x(k)
A x(k)
λ
0
1
1
1
3
4
3
4
1
0.75
1
0.75
2.5
3.5
2.5
3.5
2
0.714
1
0.714
2.428
3.428
2.428
3.428
3
0.708
1
0.708
2.416
3.416
2.416
3.416
4
0.707
1
0.707
2.414
3.414
2.414
3.414
5
0.707
1
0.707
Thus λ3 = 4 − 3.41 = 0.59 correct to two decimal places.
Exercises 1.6.3
20
Eigenvalues given by
−1 − λ
Δ = 0
0
−12
30 = 0
20 − λ
6
−13 − λ
−9
−13 − λ
30
= (−1 − λ)(λ2 − 7λ + 10)
Now Δ = (−1 − λ)
−9
20 − λ
= (−1 − λ)(λ − 5)(λ − 2) so Δ = 0 gives λ1 = 5, λ2 = 2, λ3 = −1
Corresponding eigenvectors are given by the solutions of
(A − λi I)ei = 0
When λ = λ1 = 5 we have
⎡
−6
6
⎣ 0 −18
0
−9
⎤
−12
30 ⎦
15
⎡
⎤
e11
⎣ e12 ⎦ = 0
e13
c Pearson Education Limited 2011
22
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
leading to the solution
e11
−e12
e13
=
=
= β1
−36
−180
108
so the eigenvector corresponding to λ1 = 5 is e1 = β1 [1 − 5 − 3]T
When λ = λ2 = 2, we have
⎡
−3
6
⎣ 0 −15
0
−9
leading to the solution
⎤
−12
30 ⎦
18
⎡
⎤
e21
⎣ e22 ⎦ = 0
e23
−e22
e23
e21
=
=
= β2
0
−90
45
so the eigenvector corresponding to λ2 = 2 is e2 = β2 [0 2 1]T
When λ = λ3 = −1, we have
⎡
0
6
⎣ 0 −12
0 −9
leading to the solution
⎤
−12
30 ⎦
21
⎡
⎤
e31
⎣ e32 ⎦ = 0
e33
−e32
e33
e31
=
=
= β3
18
0
0
so the eigenvector corresponding to λ3 = −1 is e3 = β3 [1 0 0]T
A modal matrix M
⎡
1 0
⎣
M = −5 2
−3 1
⎡
0
1
−1
⎣
M = 0
3
1 −1
21
and spectral matrix Λ are
⎤
⎤
⎡
5 0
0
1
0⎦
0⎦
Λ = ⎣0 2
0 0 −1
0
⎤
−2
−5 ⎦ and matrix multiplication confirms M−1 A M = Λ
2
From Example 1.9 the eigenvalues and corresponding normalised eigenvectors
of A are
λ1 = 6, λ2 = 3, λ3 = 1
1
1
ê1 = √ [1 2 0]T , ê2 = [0 0 1]T , ê3 = √ [−2 1 0]T ,
5
5
⎡
⎤
1 0 −2
1 ⎣
2 0
1⎦
M̂ = √
5 0 √5
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎡
M̂T A M =
1
5
1
=
5
=
22
1
5
1 2
⎣ 0 0
−2 1
⎡
6 12
⎣ 0
0
−2 1
⎡
30 0
⎣ 0 15
0
0
⎤⎡
2 2
√0
⎦
⎣
2 5
5
0 0
0
⎤⎡
1
0
√
⎦
⎣
2
3 5
0
0
⎤
⎡
6
0
⎦
⎣
0 = 0
0
5
⎤⎡
1
0
⎦
⎣
2
0
0
3
0
√0
5
⎤
23
⎤
−2
1⎦
0
0 −2
1 ⎦
√0
5 0
⎤
0 0
3 0⎦ = Λ
0 1
The eigenvalues of A are given by
5 − λ
10
10
2−λ
8
−2
8
−2 = −(λ3 −18λ2 −81λ+1458) = −(λ−9)(λ+9)(λ−18) = 0
11 − λ
so eigenvalues are λ1 = 18, λ2 = 9, λ3 = −9
The eigenvectors are given by the corresponding solutions of
(5 − λi )ei1 + 10ei2 + 8ei3 = 0
10ei1 + (2 − λi )ei2 − 2ei3 = 0
8ei1 − 2ei2 + (11 − λi )ei3 = 0
Taking i = 1, 2, 3 and solving gives the eigenvectors as
e1 = [2 1 2]T , e2 = [1 2 − 2]T , e3 = [−2 2 1]T
Corresponding normalised eigenvectors are
ê1 =
1
1
1
[2 1 2]T , ê2 = [1 2 − 2]T , ê3 = [−2 2 1]T
3
3
3
⎡
2
1 ⎣
1
M̂ =
3
2
⎤
1 −2
2
2 ⎦,
−2
1
⎡
2 1
1 ⎣
T
M̂ =
1 2
3
−2 2
c Pearson Education Limited 2011
⎤
2
−2 ⎦
1
24
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎡
M̂T A M =
=
1
9
1
9
⎡
2
⎣ 1
−2
⎡
36
⎣ 9
18
4
2
⎣
2
= 1
2 −2
⎡
18 0
⎣
= 0 9
0 0
⎤⎡
⎤⎡
2
5
10
8
2
⎦
⎣
⎦
⎣
1
10
2 −2
−2
2
8 −2 11
1
⎤
⎤⎡
2
1 −2
18
36
⎦
⎣
1
2
2⎦
18 −18
2 −2
1
−18 −9
⎤
⎤⎡
2
1 −2
4
⎦
⎣
1
2
2⎦
−2
2 −2
1
−1
⎤
0
0⎦ =Λ
−9
1
2
2
⎡
23
1 1
⎣
A = −1 2
0 1
1
2
−2
⎤
−2
2⎦
1
⎤
−2
1⎦
−1
Eigenvalues given by
1 − λ
0 = −1
0
1
2−λ
1
−2
1 = −(λ3 − 2λ2 − λ + 2) = −(λ − 1)(λ − 2)(λ + 1) = 0
−1 − λ
so eigenvalues are λ1 = 2, λ2 = 1, λ3 = −1.
The eigenvectors are given by the corresponding solutions of
(1 − λi )ei1 + ei2 − 2ei3 = 0
−ei1 + (2 − λi )ei2 + ei3 = 0
0ei1 + ei2 − (1 + λi )ei3 = 0
Taking i = 1, 2, 3 and solving gives the eigenvectors as
e1 = [1 3 1]T , e2 = [3 2 1]T , e3 = [1 0 1]T
⎤
⎡
2
1
⎦
⎣
0 , Λ= 0
0
1
⎤
2 −2 −2
1 ⎣
−3
0 −3 ⎦
=−
6
1
2 −7
⎡
1
⎣
M= 3
1
M−1
3
2
1
⎡
0
1
0
c Pearson Education Limited 2011
⎤
0
0 ⎦
−1
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
25
Matrix multiplication then confirms
M−1 A M = Λ
and A = M Λ M−1
24 Eigenvalues given by
3 − λ
−2
4
−2
−2 − λ
6 = −λ3 + 63λ − 162 = −(λ + 9)(λ − 6)(λ − 3) = 0
4
6
−1 − λ
so the eigenvalues are λ1 = −9, λ2 = 6, λ3 = 3.
The eigenvectors are the
corresponding solutions of
(3 − λi )ei1 − 2ei2 + 44ei3 = 0
−2ei1 − (2 + λi )ei2 + 6ei3 = 0
4ei1 + 6ei2 − (1 + λi )ei3 = 0
Taking i = 1, 2, 3 and solving gives the eigenvectors as
e1 = [1 2 − 2]T , e2 = [2 1 2]T , e3 = [−2 2 1]T
Since eT1 e2 = eT1 e3 = eT2 e3 = 0 the eigenvectors are orthogonal
⎡
1
1 ⎣
2
L = [ê1 ê2 ê3 ] =
3
−2
⎡
25
L̂ A L =
1
9
=
1
9
=
1
9
1
⎣ 2
−2
⎡
−9
⎣ 12
−6
⎡
−81
⎣ 0
0
2
1
2
⎤
−2
2⎦
1
⎤⎡
⎤⎡
1 2
3 −2
4
−2
6 ⎦⎣ 2 1
2 ⎦ ⎣ −2 −2
−2 2
4
6 −1
1
⎤
⎤⎡
1 2 −2
−18 18
2⎦
6 12 ⎦ ⎣ 2 1
−2 2
1
6
3
⎤
⎤
⎡
−9 0 0
0
0
54 0 ⎦ = ⎣ 0 6 0 ⎦ = Λ
0 0 3
0 27
2
1
2
Since the matrix A is symmetric the eigenvectors
e1 = [1 2 0]T , e2 = [−2 1 0]T , e3 = [e31 e32 e33 ]T
c Pearson Education Limited 2011
⎤
−2
2⎦
1
26
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
are orthogonal. Hence,
eT1 e3 = e31 + 2e32 = 0 and eT2 e3 = −2e31 + e32 = 0
Thus, e31 = e32 = 0 and e33 arbitrary
⎡
6
Using A = M̂ Λ M̂T where Λ = ⎣ 0
0
⎡
A= ⎣
⎡
− √25
√1
5
√2
5
2
⎣
= 2
0
√1
5
0
2
5
0
0
⎤
0
0⎦
3
so a possible
eigenvector is e3 = [0 0 1]T .
⎤
0 0
1 0 ⎦ gives
0 3
⎤⎡
0
6
⎦
⎣
0
0
0
1
0
1
0
⎤⎡ 1
√
0
5
0 ⎦ ⎣ − √2
5
3
0
√2
5
√1
5
0
⎤
0
0⎦
1
⎤
⎤
⎡
0
0 0
−4 −7 −5
3
3 ⎦ ∼ ⎣ 0 −1 0 ⎦ is of rank 2
26 A − I = ⎣ 2
1
0 0
1
2
1
Nullity (A − I) = 3 − 2 = 1 so there is only one linearly independent vector
⎡
corresponding to the eigenvalue 1. The corresponding eigenvector e1 is given by
the solution of (A − I)e1 = 0 or
−4e11 − 7e12 − 5e13 = 0
2e11 + 3e12 + 3e13 = 0
e11 + 2e12 + 212 = 0
that is, e1 = [−3 1 1]T . To obtain the generalised eigenvector e∗1 we solve
(A − I)e∗1 = e1 or
⎤
⎤⎡ ∗ ⎤
⎡
⎡
e11
−3
−4 −7 −5
⎣ 2
3
3 ⎦ ⎣ e∗12 ⎦ = ⎣ 1 ⎦
1
1
2
1
e∗13
giving e∗1 = [−1 1 0]T . To obtain the second generalised eigenvector e∗∗
1 we solve
∗
(A − I)e∗∗
1 = e1 or
⎤
⎤ ⎡ ∗∗ ⎤
⎡
⎡
e11
−1
−4 −7 −5
⎦ = ⎣ 1⎦
⎣ 2
3
3 ⎦ ⎣ e∗∗
12
∗∗
0
1
2
1
e13
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
27
T
giving e∗∗
1 = [2 − 1 0] .
⎤
−3 −1
2
⎣ 1
1 −1 ⎦
M = [e1 e∗1 e∗∗
1 ]=
1
0
0
⎤
⎡
⎡
0
0
0 −1
−1
⎦
⎣
⎣
det M = −1 and M = − −1 −2 −1 = 1
1
−1 −1 −2
⎡
0
2
1
⎤
1
1⎦
2
Matrix multiplication then confirms
⎤
1 1 0
A M = ⎣0 1 1⎦
0 0 1
⎡
M−1
27
Eigenvalues are given by
| A − λI |= 0
that is, λ4 − 4λ3 − 12λ2 + 32λ + 64 = (λ + 2)2 (λ − 4)2 = 0 so the eigenvalues are
− 2, − 2, 4 and 4 as required.
Corresponding to the repeated eigenvalue λ1 , λ2 = −2
⎡
3
⎢ 0
(A + 2I) = ⎣
−0.5
−3
0
3
−3
0
0
−3
3
0
⎤
⎡
1 0
−3
0⎥
⎢0 1
⎦ ∼ ⎣
0 0
0.5
0 0
3
0
0
0
0
⎤
0
0⎥
⎦ is of rank 2
0
0
Thus, nullity (A+2I) is 4−2 = 2 so there are two linearly independent eigenvectors
corresponding to λ = −2.
Corresponding to the repeated eigenvalues λ3 , λ4 = 4
⎡
−3
0
⎢
(A − 4I) = ⎣
−0.5
−3
0
−3
−3
0
0
−3
−3
0
⎤
⎤
⎡
1 0 0 0
−3
0⎥
⎢0 1 0 0⎥
⎦ is of rank 3
⎦ ∼ ⎣
0 0 0 0
0.5
0 0 0 1
−3
Thus, nullity (A − 4I) is 4 − 3 = 1 so there is only one linearly independent
eigenvector corresponding to λ = 4.
c Pearson Education Limited 2011
28
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
When λ = λ1 = λ2 = −2 the eigenvalues are given by the solution of (A+2I)e = 0
giving e1 = [0 1 1 0]T , e2 = [1 0 0 1]T as two linearly independent solutions. When
λ = λ3 = λ4 = 8 the eigenvectors are given by the solution of
(A − 4I)e = 0
giving the unique solution e3 = [0 1 − 1 0]T . The generalised eigenvector e∗3 is
obtained by solving
(A − 4I)e∗3 = e3
giving e∗3 = (6 − 1
0 − 6]T . The Jordan canonical form is
⎡
⎤
⎢ −2 0
⎢
⎢ 0 −2
⎢
J= ⎢
⎢
⎢ 0
0
⎣
0
0
0
0
4
0
0⎥
⎥
0⎥
⎥
⎥
⎥
1⎥
⎦
4
Exercises 1.6.5
28
[ x1
The quadratic form may be written in the form V = xT Ax where x =
T
x2 x3 ] and
⎤
⎡
2 2 1
A = ⎣2 5 2⎦
1 2 2
The eigenvalues of A are given by
2− λ
2
1
=0
2
5
−
λ
2
1
2
2 − λ
⇒ (2 − λ)(λ2 − 7λ + 6) + 4(λ − 1) + (λ − 1) = 0
⇒ (λ − 1)(λ2 − 8λ + 7) = 0 ⇒ (λ − 1)2 (λ − 7) = 0
giving the eigenvalues as λ1 = 7, λ2 = λ3 = 1
Normalized eigenvector corresponding to λ1 = 7 is
ê1 =
√1
6
√2
6
√1
6
T
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
29
and two orthogonal linearly independent eigenvectors corresponding to λ − 1 are
√1
2
ê2 =
− √12
0
ê3 = − √13
√1
3
T
− √13
T
Note that ê2 and ê3 are automatically orthogonal to ê1 . The normalized
orthogonal modal matrix M̂ and spectral matrix Λ are
⎡
⎢
M̂ = ⎣
√1
6
√2
6
√1
6
√1
2
− √13
0
√1
3
− √13
− √12
⎤
⎡
7
⎥
⎣
⎦,Λ = 0
0
0
1
0
⎤
0
0⎦
1
such that M̂T AM̂ = Λ.
Under the orthogonal transformation x = M̂y the quadratic form V reduces to
V = yT M̂T AM̂y = yT Λy
⎤⎡ ⎤
⎡
y1
7 0 0
= [ y1 y 2 y 3 ] ⎣ 0 1 0 ⎦ ⎣ y 2 ⎦
0 0 1
y3
= 7y21 + y22 + y23
⎤
1 −1
2
2 −1 ⎦ and its leading
29(a) The matrix of the quadratic form is A = ⎣ −1
2 −1
7
principal minors are
1 −1
= 1, det A = 2
1,
−1
2
⎡
Thus, by Sylvester’s condition (a) the quadratic form is positive definite.
⎡
29(b)
1 −1
⎣
2
Matrix A = −1
2 −1
1
1,
−1
⎤
2
−1 ⎦ and its leading principal minors are
5
−1
= 1, det A = 0
2
Thus, by Sylvester’s condition (c) the quadratic form is positive semidefinite.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎤
⎡
1 −1
2
2 −1 ⎦ and its leading principal minors are
29(c) Matrix A = ⎣ −1
2 −1
4
30
1
1,
−1
−1
= 1, det A = −1.
2
Thus, none of Sylvester’s conditions are satisfied and the quadratic form is
indefinite.
a −b
and its leading
30(a) The matrix of the quadratic form is A =
−b c
principal minors are a and ac − b2 . By Sylvester’s condition (a) in the text the
quadratic form is positive definite if and only if
a > 0 and ac − b2 > 0
that is, a > 0 and ac > b2
⎤
2 −1 0
30(b) The matrix of the quadratic form is A = ⎣ −1 a b ⎦ having principal
0 b 3
2
minors 2, 2a − 1 and det A = 6a − 2b − 3. Thus, by Sylvester’s condition (a) in
⎡
the text the quadratic form is positive definite if and only if
2a − 1 > 0 and 6a − 2b2 − 3 > 0
or 2a > 1 and 2b2 < 6a − 3
31
The eigenvalues of the matrix A are given by
2 − λ
3 − λ 3 − λ
1
−1
0
2−λ
1 R1 +R3 1
2−λ
1
0 = 1
−1
−1
1
2− λ
1
2 − λ
1
1
0
1
= (3 − λ) 1 2 − λ
−1
1
2− λ
1
0
0
1 = (3 − λ)(λ2 − 3λ)
= (3 − λ) 1 1 − λ
−1
2
2− λ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
31
so the eigenvalues are 3, 3, 0 indicating that the matrix is positive semidefinite.
The principal minors of A are
2
2,
1
1
= 3, det A = 0
2
confirming, by Sylvester’s condition (a), that the matrix is positive semidefinite.
⎤
K 1
1
32 The matrix of the quadratic form is A = ⎣ 1 K −1 ⎦ having principal
1 −1 1
minors
K 1
= K2 − 1 and det A = K2 − K − 3
K,
1 K
⎡
Thus, by Sylvester’s condition (a) the quadratic form is positive definite if and only
if
K2 − 1 = (K − 1)(K + 1) > 0 and K2 − K − 3 = (K − 2)(K + 1) > 0
i.e. K > 2 or K < −1.
If K = 2 then det A = 0 and the quadratic form is positive semidefinite.
33
Principal minors of the matrix are
3+ a
3 + a,
1
1
= a2 + 3a − 1, det A = a3 + 3a2 − 6a − 8
a
Thus, by Sylvester’s condition (a) the quadratic form is positive definite if and only
if
3 + a > 0, a2 + 3a − 1 > 0 and a3 + 3a2 − 6a − 8 > 0
or (a + 1)(a + 4)(a − 2) > 0
3 + a > 0 ⇒ a > −3
a2 + 3a − 1 > 0 ⇒ a < −3.3 or a > 0.3
(a + 1)(a + 4)(a − 2) > 0 ⇒ a > 2 or − 4 < a < −1
Thus, minimum value of a for which the quadratic form is positive definite is
a = 2.
c Pearson Education Limited 2011
32
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎤
1
2 −2
λ −3 ⎦
34 A = ⎣ 2
−2 −3
λ
Principal minors are
⎡
2
= λ − 4, det A = λ2 − 8λ + 15 = 0
λ
1
1,
2
Thus, by Sylvester’s condition (a) the quadratic form is positive definite if and only
if
λ−4>0 ⇒ λ>4
and (λ − 5)(λ − 3) > 0 ⇒ λ < 3 or λ > 5
Thus, it is positive definite if and only if λ > 5.
Exercises 1.7.1
35
The characteristic equation of A is
5− λ
6
= λ2 − 8λ + 3 = 0
2
3 − λ
2
Now A =
A − 8A + 3I =
2
37
16
5
2
48
21
6
3
5
2
−
6
3
40
16
=
48
24
27
16
+
48
21
3
0
so that A satisfies its own characteristic equation.
36
The characteristic equation of A is
1 − λ
2
= λ2 − 2λ − 1 = 0
1
1 − λ
c Pearson Education Limited 2011
so
0
3
=
0
0
0
0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
By Cayley–Hamilton theorem
A2 − 2A − I = 0
36(a)
2
Follows that A = 2A + I =
2
36(b)
6
A = 2A + A =
4
36(c)
14
A = 2A + A =
10
37(a)
The characteristic equation of A is
3
2
4
3
2
8
6
+
20
14
that is,
1
1
+
+
2
1
2 − λ
1
4
2
2
1 0
0 1
=
3
2
4
3
7
5
=
10
7
=
3
2
4
3
17
12
24
17
1
=0
2 − λ
λ2 − 4λ + 3 = 0
Thus, by the Cayley–Hamilton theorem
so that
37(b)
A2 − 4A + 3I = 0
1
I = [4A − A2 ]
3
1
A−1 = [4I − A]
3
1
2
4 0
−
=
1
0 4
3
1
2
1
=
3
The characteristic equation of A is
1 − λ
1
3
1
−
λ
2
3
that is,
2
1 = 0
1 − λ
λ3 − 3λ2 − 7λ − 11 = 0
c Pearson Education Limited 2011
2 −1
−1 2
33
34
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎡
1
2
⎣
A = 3
2
1
1
3
⎤ ⎡
1
2
⎦
⎣
3
1
2
1
1
1
3
⎤
⎡
8
2
⎦
⎣
8
1 =
13
1
8
7
8
⎤
5
8⎦
8
Using (1.44)
1
(A2 − 3A − 7I)
11 ⎡
⎤
−2 5 −1
1 ⎣
−1 −3 5 ⎦
=
11
7 −1 −2
A−1 =
⎤ ⎡
2 3
2 3 1
38 A2 = ⎣ 3 1 2 ⎦ ⎣ 3 1
1 2
1 2 3
The characteristic equation of A
⎡
⎤
⎡
14
1
2 ⎦ = ⎣ 11
11
3
is
11
14
11
⎤
11
11 ⎦
14
λ2 − 6λ2 − 3λ + 18 = 0
so by the Cayley–Hamilton theorem
A3 = 6A2 + 3A − 18I
giving
A4 = 6(6A2 + 3A − 18I) + 3A2 − 18A = 39A2 − 108I
A5 = 39(6A2 + 3A − 18I) + 108A = 234A2 + 9A − 702I
A6 = 234(6A2 + 3A − 18I) + 9A2 − 702A = 1413A2 − 4212I
A7 = 1413(6A2 + 3A − 18I) + 4212A = 8478A2 + 27A − 25434I
Thus,
A7 − 3A6 + A4 + 3A3 − 2A2 + 3I = 4294A2 + 36A − 12957I
⎤
⎡
47231 47342 47270
= ⎣ 47342 47195 47306 ⎦
47270 47306 47267
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
39(a)
Eigenvalues A are λ = 1 (repeated). Thus,
eAt = α0 I + α1 A with
= α0 + α1
⇒ α1 = tet , α0 = (1 − t)et
= α1
t
e
At
t
t
so e = (1 − t)e I + te A =
tet
et
tet
39(b)
Eigenvalues A are λ = 1 and λ = 2. Thus,
et
e2t
40
0
et
eAt = α0 I + α1 A with
= α0 + α1
⇒ α0 = 2et − e2t , α1 = e2t − et
= α0 + 2α1
At
t
2t
2t
t
so e = (2e − e )I + (e − e )A =
Eigenvalues of A are λ1 = π, λ2 =
Thus,
et
2t
e − et
π
π
, λ3 = .
2
2
sin A = α0 A + α1 A + α2 A2 with
sin π = 0 = α0 + α1 π + α2 π2
π
π2
π
= 1 = α0 + α1 + α2
2
2
4
π
cos = 0 = α1 + πα2
2
4
4
Solving gives α0 = 0, α1 = , α2 = − 2 so that
π
π
⎤
⎡
0 0 0
4
4
sin A = A − 2 A2 = ⎣ 0 1 0 ⎦
π
π
0 0 1
sin
41(a)
dA
=
dt
d 2
dt (t + 1)
d
dt (5 − t)
d
dt (2t − 3)
d 2
dt (t − t + 3)
=
2t
−1
c Pearson Education Limited 2011
2
2t − 1
0
e2t
35
36
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
41(b)
1
2
2 2
(t + 1)dt
Adt = 12
(5 − t)dt
1
42
2
A =
t2 + 1
5
2
(2t − 3)dt
21 2
(t − t + 3)dt
1
t−1
0
t2 + 1
5
⎡ 10
= ⎣
t−1
0
0
7
2
23
6
⎤
⎦
t3 − t2 + t − 1
=
5t − 5
3
d
4t + 4t + 5 3t2 − 2t + 1
2
(A ) =
10t
5
dt
3
2
dA
4t + 4t 2t + 1
2A
=
20t
0
dt
Thus,
t4 + 2t2 + 5t − 4
5t2 + 5
3
d
dA
(A2 ) = 2A
.
dt
dt
Exercises 1.8.4
43(a)
row rank
⎡
⎤
⎡
⎤
3
4
row2 − 3row1 1 2
1 2 3 4
⎣ 0 −2 −2 −2 ⎦
→
A = ⎣ 3 4 7 10 ⎦
row3 − 2row1 0 −3 −1 −1
2 1 5 7
⎤
⎤
⎡
⎡
1 2 3 4
1 2
4
4
1
− 2 row2 ⎣
row3 + 3row2 ⎣
0 1 1 1⎦
0 1
1
1 ⎦
→
→
0 0 2 2
0 −3 −1 −1
echelon form, row rank 3
column rank
⎡
⎤
0 0 col3 − col2 1
⎣0
→
−2 2 ⎦
2 0 col4 − col2 2
⎤
⎡
1 0 0 0
col4 − col3 ⎣
3 −2 0 0 ⎦
→
2 −3 2 0
col2 − 2col1 ⎡
1
→
⎣
3
A
col3 − 3col1
2
col4 − 4col1
0
−2
−3
echelon form, column rank3
Thus row rank(A) = column rank(A) = 3
c Pearson Education Limited 2011
⎤
0 0 0
−2 0 0 ⎦
−3 2 2
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
37
(b) A is of full rank since rank(A)=min( m, n) =min(3,4) = 3
⎤
4
8
37 9
333 81
4 11 14 ⎣
T
⎦
=9
11 7
=
44(a) AA =
9 13
81 117
8 7 −2
14 −2
The eigenvalues λi of AAT are given by the solutions of the equations
8
333 − λ
T
= 0 ⇒ λ2 − 450λ + 32400 = 0
AA − λI =
81
117 − λ
⎡
⇒ (λ − 360)(λ − 90) = 0
giving the eigenvalues as λ1 = 360, λ2 = 90. Solving the equations.
(AAT − λi I)ui = 0
gives the corresponding eigenvectors as
u1 = [ 3
T
T
1 ] , u2 = [ 1 −2 ]
with the corresponding normalized eigenvectors being
√1
10
√3
10
û1 =
T
√1
10
, û2 =
− √310
T
leading to the orthogonal matrix
Û =
⎡
4
T
⎣
A A = 11
14
⎤
8
4
7 ⎦
8
−2
√3
10
√1
10
11
7
√1
10
− √310
⎡
80
14
⎣
= 100
−2
40
T
80 − μ
Solving A A − μI = 100
40
100
170 − μ
140
⎤
40
140 ⎦
200
40
140 = 0
200 − μ
100
170
140
gives the eigenvalues μ1 = 360, μ2 = 90, μ3 = 0 with corresponding normalized
eigenvectors
v̂1 = [ 13
2
3
2 T
3 ]
, v̂2 = [ − 32
− 13
2 T
3 ]
, v̂3 = [ 23
c Pearson Education Limited 2011
− 23
1 T
3 ]
38
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
leading to the orthogonal matrix
⎡1
V̂ = ⎣
The singular values of A are σ1 =
√
− 23
− 13
3
2
3
2
3
2
3
− 23
1
3
2
3
⎤
⎦
√
√
√
360 = 6 10 and σ2 = 90 = 3 10 giving
√
6 10
Σ=
0
√0
3 10
0
0
Thus, the SVD form of A is
A = ÛΣV̂T =
√3
10
√1
10
√1
10
− √310
√
6 10
0
4
(Direct multiplication confirms A =
8
√0
3 10
0
0
⎡
2
3
− 13
− 23
1
3
⎣ −2
3
2
3
2
3
2
3
1
3
⎤
⎦
14
)
−2
11
7
(b) Using (1.55) the pseudo inverse of A is
⎡
√1
6 10
A† = V̂Σ∗ Û, Σ∗ = ⎣ 0
0
3
†
AA =
1
180
4
8
0
⎤
⎡1
⎦⇒⎣
− 23
− 13
3
2
3
2
3
2
3
− 23
1
3
⎤⎡
√1
6 10
⎦⎣ 0
2
0
0
3
⎤
⎡
−1 13
√
√1
†
1 ⎣
10
10
4
8 ⎦
⇒ A = 180
√1
√3
−
10
10
10 −10
⎤
⎡
−1 13
14 ⎣
180
0
1
4
8 ⎦ = 180
=I
−2
0 180
10 −10
11
7
√2
3 10
0
√1
3 10
⎤
⎦
0
(c) Rank( A) = 2 so A is of full rank. Since number of rows is less than the number
of columns A† may be determined using (1.58b) as
⎡
A† = AT (AAT )−1
4
⎣
= 11
14
⎤
8
333
7 ⎦
81
−2
81
117
−1
which confirms with the value determined in (b).
c Pearson Education Limited 2011
⎤
−1 13
1 ⎣
4
8 ⎦
= 180
10 −10
⎡
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
39
⎤
⎡
⎤
⎤
⎡
1 1
1 1
1
row3 + row2
row2 − 3row1
0⎥
⎢ 0 −3 ⎥
⎢ 0 −3 ⎥
⎥ row4 + 23 row2 ⎢
⎥
⎥ row3 + 2row1 ⎢
1⎥
⎢0 3 ⎥
⎢0 0 ⎥
→
→
⎦
⎣
⎦
⎦
⎣
0 2
0 0
2
row5 + row1
row5 + row2
0 3
0 0
2
echelon form so row rank = 2 = column rank
⎡
1
⎢ 3
⎢
45 A = ⎢ −2
⎣
0
−1
Thus, rank A = 2 =min(5,2) and so A is of full rank.
Since A is of full rank and number of rows is greater than number of columns we
can determine the pseudo inverse using result (1.58a)
†
T
A = (A A)
−1
−1
1 3 −2 0
15 −3
A =
1 0 1 2
−3 10
1 3 −2 0
10 3
1
= 141
1 0 1 2
3 15
13 30 −17 6 −4
1
= 141
18 9
9
30 27
T
†
A A=
1
141
13
18
30
9
−17
9
6
30
1
T
⎣
(b) AA = −2
2
The eigenvalues λi
⎤
−1
1 2
2 ⎦
−1 2
−2
are given by
2 − λ
−4
−4
2−λ
4
−8
−1
2
⎡
1
⎢ 3
−4 ⎢
⎢ −2
27 ⎣
0
−1
⎤
1
0⎥
⎥
1⎥ =
⎦
2
2
1
141
141
0
0
=I
141
⎤
−1
0 ⎦
0
⎡
⎤
1 −1 row2 + 2row1 1
⎣0
→
46(a) A = ⎣ −2 2 ⎦
2 −2 row3 − 2row1 0
Thus, rank A = 1and is not of full rank
⎡
⎡
−1
2
⎡
2
2
⎣
= −4
−2
4
−4
8
−8
⎤
4
−8 ⎦
8
4
−8 = 0 ⇒ λ2 (−λ + 18) = 0
8− λ
giving the eigenvalues as λ1 = 18, λ2 = 0, λ3 = 0. The corresponding eigenvectors
and normalized eigenvectors are
c Pearson Education Limited 2011
40
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
T
u1 = [ 1 −2
2 ] ⇒ û1 = [ 13
u2 = [ 0
1
T
1 ] ⇒ û2 = 0
u3 = [ 2
1
0 ] ⇒ û3 =
T
− 32
√1
2
√2
5
√1
5
2 T
3 ]
T
√1
2
T
0
leading to the orthogonal matrix
⎡
Û =
AT A =
1
−1
1
3
⎢ −2
⎣ 3
2
3
√2
5
√1
5
0
√1
2
√1
2
⎡
1
−2 2 ⎣
−2
2 −2
2
⎤
⎥
⎦
0
⎤
−1
1
−1
2 ⎦=9
−1 1
−2
having eigenvalues μ1 = 18 and μ2 = 0 and corresponding eigenvectors
T
√1
2
v1 = [ 1 −1 ] ⇒ v̂1 =
T
√1
2
√1
2
v2 = [ 1 1 ] ⇒ v̂2 =
− √12
T
T
leading to the orthogonal matrix
V̂ =
√1
2
√1
2
√1
2
− √12
√
√
single
⎤ (equal to its rank) singular value σ1 = 18 = 3 2 so that
0
0 ⎦ and the SVD form of A is
0
⎡ 1
⎤⎡ √
⎤
√2
0
3
3 2 0
5
√1
√1
−
⎢
⎥
1
1
2
2
A = ÛΣV̂T = ⎣ − 3 √2 √5 ⎦ ⎣ 0
0 ⎦ 12
√
√1
2
2
2
0
0
√1
0
3
2
⎤
⎡
1 −1
Direct multiplication confirms that A = ⎣ −2 2 ⎦
2 −2
A has
√
⎡ the
3 2
Σ=⎣ 0
0
(c) Pseudo inverse is given by
†
∗
T
A = V̂Σ Û =
√1
2
− √12
√1
2
√1
2
1
√
3 2
0
0
0
⎡
1
3
0 ⎣ 0
0
2
√
5
− 23
√1
2
√1
5
c Pearson Education Limited 2011
2
3
√1
2
0
⎤
⎦=
1
18
1 −2
−1 2
2
−2
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
41
Direct multiplication confirms AA† A = AandA† AA† = A†
(d) Equations may be written as
⎡
1
⎣ −2
2
⎡ ⎤
⎤
1
−1
x
⎣
⎦
= 2 ⎦ ≡ Ax = b
2
y
3
−2
x
†
The least squares solution is x = A b ⇒
=
y
1
6
giving x = 16 and y = − 16
− 61
1
18
1
−1
−2
2
⎡ ⎤
1
2 ⎣ ⎦
2 =
−2
3
(e) Minimize L = (x − y − 1)2 + (−2x + 2y − 2)2 + (2x − 2y − 3)2
∂L
= 0 ⇒ 2(x − y − 1) − 4(−2x + 2y − 2) + 4(2x − 2y − 3) = 18x − 18y − 6 = 0
∂x
⇒ 3x − 3y − 1 = 0
∂L
= 0 ⇒ −2(x − y − 1) + 4(−2x + 2y − 2) − 4(2x − 2y − 3) = −18x + 18y + 6 = 0
∂y
⇒ −3x + 3y + 1 = 0
Solving the two simultaneous equations gives the least squares solution x =
1
6,
y = − 16 confirming the answer in (d)
47(a) Equations may be written as
⎡
3
⎣1
1
⎡ ⎤
⎤
1
−1
x
⎣
⎦
= 2 ⎦ ≡ Ax = b
3
y
3
1
Using the pseudo inverse obtained in Example 1.39, the least squares solution is
x
†
x=A b⇒
=
y
giving x = y =
1
60
17
−7
4
16
⎡ ⎤
1
2
5 ⎣ ⎦
2 = 32
5
3
3
2
3
c Pearson Education Limited 2011
42
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
(b) Minimize L = (3x − y − 1)2 + (x + 3y − 2)2 + (x + y − 3)2
∂L
= 0 ⇒ 6(3x − y − 1) + 2(x + 3y − 2) + 2(x + y − 3) = 0
∂x
⇒ 11x + y − 8 = 0
∂L
= 0 ⇒ −2(3x − y − 1) + 6(x + 3y − 2) + 2(x + y − 3) = 0
∂y
⇒ x + 11y − 8 = 0
Solving the two simultaneous equations gives the least squares solution x = y =
2
3
confirming the answer in (a)
48(a)
⎡
1
⎢ 0
A=⎣
−1
2
0
1
1
−1
⎡
⎤
1
−2
row3 + row1
−1 ⎥
⎢0
→
⎣
⎦
0
1
row4 − 2row1
0
2
0
1
1
−1
⎡
⎤
1
−2
row3 − row2
−1 ⎥
⎢0
→
⎣
⎦
0
−1
row4 + row2
0
6
0
1
0
0
⎤
−2
−1 ⎥
⎦
0
5
Thus, A is of rank 3 and is of full rank as 3=min(4,3)
(b) Since A is of full rank
⎤−1 ⎡
6 −3 1
1
†
T
−1 T
⎦
⎣
⎣
A = (A A) A = −3 3 −2
0
1 −2 10
−2
⎤
⎤⎡
⎡
1
0 −1 2
26 28 3
1 ⎣
1
1 −1 ⎦ =
⇒ A† = 75
28 59 9 ⎦ ⎣ 0
−2 −1 1
2
3 9 9
⎡
⎤
0 −1 2
1
1 −1 ⎦
−1 1
2
⎤
⎡
4
5 1 6
1 ⎣
2 10 8 3 ⎦
15
−3 0 3 3
(c) Direct multiplication confirms that A† satisfies the conditions
AAT and AT A are symmetric, AA† A = A and A† AA† = A†
⎡
⎤
1
2 ⎦ is of full rank 2 so pseudo inverse is
1
0.6364 −0.3636 0.0909
†
T
−1 T
A = (A A) A =
−0.3636 0.6364 0.0909
2
49(a) A = ⎣ 1
1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
43
Equations (i) are consistent with unique solution
⎡ ⎤
3
x
†⎣ ⎦
=A 3 ⇒x=y=1
y
2
Equations (ii) are inconsistent with least squares solution
⎡ ⎤
3
x
†⎣ ⎦
= A 3 ⇒ x = 1.0909, y = 1.0909
y
3
⎤
2 1
0.5072
†
⎦
⎣
with pseudo inverse A =
(b) A = 1 2
−0.4928
10 10
Equations (i) are consistent with unique solution
⎡
−0.4928
0.5072
0.0478
0.0478
⎡ ⎤
3
x
= A† ⎣ 3 ⎦ ⇒ x = y = 1
y
20
Equations (ii) are inconsistent and have least squares solution
⎡ ⎤
3
x
= A† ⎣ 3 ⎦ ⇒ x = y = 1.4785
y
30
⎤
2
1
0.5001
†
2 ⎦ with pseudo inverse A =
(c) A = ⎣ 1
−o.4999
100 100
Equations (i) are consistent with unique solution
⎡
⎤
⎡
3
x
= A† ⎣ 3 ⎦ ⇒ x = y = 1
y
200
Equations (ii) are inconsistent with least squares solution
⎤
⎡
3
x
= A ⎣ 3 ⎦ ⇒ x = y = 1.4998
y
300
c Pearson Education Limited 2011
−0.4999
0.5001
0.0050
0.0050
44
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Since the sets of equations (i) are consistent weighting the last equation has no
effect on the least squares solution which is unique. However, since the sets of
equations (ii) are inconsistent the solution given is not unique but is the best in
the least squares sense. Clearly as the weighting of the third equation increases
from (a) to (b) to (c) the better is the matching to the third equation, and the last
case (c) does not bother too much with the first two equations.
50 Data may be represented in the matrix form
⎡ ⎤
⎤
1
1
1⎥
⎢1⎥
⎢ ⎥
⎥ m
= ⎢2⎥
1⎥
c
⎣ ⎦
⎦
2
1
3
1
⎡
0
⎢1
⎢
⎢2
⎣
3
4
Az = Y
MATLAB gives the pseudo inverse
−0.2
A =
0.8
†
−0.1
0.4
0
0.2
0.1
0
0.2
−0.2
and, the least squares solution
m
0.5
†
=A y=
c
0.8
leads to the linear model
y = 0.5x + 0.8
Exercises 1.9.3
51(a)
Taking x1 = y
dy
dt
d2 y
ẋ2 = x3 = 2
dt
d3 y
ẋ3 = 3 = u(t) − 4x1 − 5x2 − 4x3
dt
ẋ1 = x2 =
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, state space form is
⎡ ⎤
⎡
ẋ1
0
1
0
ẋ = ⎣ ẋ2 ⎦ = ⎣ 0
−4 −5
ẋ3
45
⎤ ⎡ ⎤
⎡ ⎤
x1
0
0
1 ⎦ ⎣ x2 ⎦ + ⎣ 0 ⎦ u(t)
−4
1
x3
y = x1 = [1 0 0] [x1 x2 x3 ]T
51(b)
x1 = y
dy
dt
d2 y
x3 = ẋ2 = 2
dt
d3 y
x4 = ẋ3 = 3
dt
4
d y
ẋ4 = 4 = −4x2 − 2x3 + 5u(t)
dt
x2 = ẋ1 =
Thus, state space form is
⎡ ⎤
⎡
ẋ1
0
⎢ ẋ ⎥
⎢0
ẋ = ⎣ 2 ⎦ = ⎣
0
ẋ3
0
ẋ4
⎤ ⎡ ⎤
⎡ ⎤
x1
1
0 0
0
0
1 0 ⎥ ⎢ x2 ⎥
⎢0⎥
⎦ ⎣ ⎦ + ⎣ ⎦ u(t)
0
0 1
0
x3
−4 −2 0
5
x4
y = x1 = [1 0 0 0] [x1 x2 x3 x4 ]T
52(a)
Taking A to be the companion matrix of the LHS
⎤
⎡
0
1
0
0
1 ⎦
A=⎣ 0
−7 −5 −6
and taking b = [ 0
0
T
1]
and then using (1.67) in the text c = [ 5 3
Then from (1.84) the state-space form of the dynamic model is
ẋ = Ax + bu, y =cx
(b) Taking A to be the companion matrix of
⎡
0 1
⎣
A= 0 0
0 −3
the LHS
⎤
0
1 ⎦
−4
c Pearson Education Limited 2011
1 ].
46
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
and taking b = [ 0
0
T
1 ] then using (1.67) in the text c = [ 2
3
1 ]. Then
from (1.84) the state-space form of the dynamic model is
ẋ = Ax + bu, y =cx
53
Applying Kirchhoff’s second law to the individual loops gives
di1
1
, v̇c = (i1 + i2 )
dt
C
di2
+ R2 i2
e = R1 (i1 + i2 ) + vc + L2
dt
e = R1 (i1 + i2 ) + vc + L1
so that,
di1
R1
vc
e
R1
i2 −
+
= − i1 −
dt
L1
L1
L1
L1
di2
R1
(R1 + R2 )
vc
e
= − i1 −
i2 −
+
dt
L2
L2
L2
L2
dvc
1
= (i1 + i2 )
dt
C
Taking x1 = i1 , x2 = i2 , x3 = vc , u = e(t) gives the state equation as
⎡
⎤
⎡ R1
− L1
ẋ1
⎣ ẋ2 ⎦ = ⎣ − R1
L2
1
ẋ3
C
2
−R
L1
2)
− (R1L+R
2
1
C
⎤
⎡ 1 ⎤
x1
L1
1 ⎦ ⎣ x ⎦ + ⎣ 1 ⎦ u(t)
− L2
2
L2
x
0
3
0
− L11
⎤ ⎡
(1)
The output y = voltage drop across R2 = R2 i2 = R2 x2 so that
y = [0 R2 0] [x1 x2 x3 ]T
(2)
Equations (1) and (2) are then in the required form
ẋ = A x + bu , y = cT x
54
The equations of motion, using Newton’s second law, may be written down
for the body mass and axle/wheel mass from which a state-space model can be
deduced. Alternatively a block diagram for the system, which is more informative
for modelling purposes, may be drawn up as follows
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
47
where s denotes the Laplace ‘s’ and upper case variables X, Y, Y1 denote the
corresponding Laplace transforms of the corresponding lower case time domain
variables x(t), y(t), y1 (t); y1 (t) is the vertical displacement of the axle/wheel mass.
Using basic block diagram rules this block diagram may be reduced to the
input/output transfer function model
X
−→
(M1
s2
K1 (K + Bs)
+ K1 )(Ms2 + Bs + K) + Ms2 (K + Bs)
Y
−→
or the time domain differential equation model
d4 y
d3 y
d2 y
M1 M 4 + B(M1 + M) 3 + (K1 M + KM1 + KM) 2
dt
dt
dt
dy
dx
+ K1 B + K1 Ky = K1 K2 x + K1 B
dt
dt
A possible state space model is
⎡
ż1
⎤
⎡
−B(M1 + M)
⎢
⎢ ⎥
⎢ −(K1 M +KM1 +KM )
⎢ ⎥
ż
⎢
⎢ 2⎥
M M1
⎢
⎢ ⎥
⎢ ⎥ = ⎢
⎢
⎢ ⎥
−K1 B
⎢
⎢ ż3 ⎥
M1 M
⎣
⎣ ⎦
ż4
−K1 K
M1 M
1
0 0
⎤ ⎡
⎥
⎥
0 1 0⎥
⎥
⎥
⎥
0 0 1⎥
⎦
0 0 0
z1
⎤
⎡
⎢ ⎥
⎢
⎢ ⎥
⎢
z
⎢ 2⎥
⎢
⎢ ⎥
⎢
⎢ ⎥ + ⎢
⎢ ⎥
⎢
⎢ z3 ⎥
⎢
⎣ ⎦
⎣
z4
0
0
K1 B
M1 M
⎤
⎥
⎥
⎥
⎥
⎥ x(t)
⎥
⎥
⎦
K1 K 2
M M1
y = [1 0 0 0]z(t), z = [z1 z2 z3 z4 ]T .
Clearly alternative forms may be written down, such as, for example, the
companion form of equation (1.66) in the text. Disadvantage is that its output
y is not one of the state variables.
c Pearson Education Limited 2011
48
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
55
Applying Kirchhoff’s second law to the first loop gives
x1 + R3 (i − i1 ) + R1 i = u
that is, (R1 + R3 )i − R3 i1 + x1 = u
Applying it to the outer loop gives
x2 + (R4 + R2 )i1 + R1 i = u
Taking α = R1 R3 + (R1 + R3 )(R4 + R2 ) then gives
αi = (R2 + R3 + R4 )u − (R2 + R4 )x1 − R3 x2
and αi1 = R3 u + R1 x1 − (R1 + R3 )x2
Thus,
α(i − i1 ) = (R4 + R2 )u − (R1 + R2 + R4 )x1 + R1 x2
1
(i − i1 )
C1
1
[−(R1 + R2 + R4 )x1 + R1 x2 + (R4 + R2 )u](1)
=
αC1
Voltage drop across C1 : ẋ1 =
1
i1
C2
1
=
[R1 x1 − (R1 + R3 )x2 + R3 u]
αC2
Voltage drop across C2 : ẋ2 =
(R1 + R3 )
R1
R3
x−
x2 +
u
α
α
α
(R4 + R2 )
R3
R3 R1
y2 = R2 (i − i1 ) = − (R1 + R2 + R4 )x1 +
x2 + R3
u
α
α
α
y1 = i1 =
(2)
(3)
(4)
Equations (1)–(4) give the required state space model.
Substituting the given values for R1 , R2 , R3 , R4 , C1 and C2 gives the state matrix
A as
⎡
A= ⎣
−9
35.10−3
1
35.10−3
1
35.10−3
−4
35.10−3
⎤
3
⎦ = 10
35
−9
1
c Pearson Education Limited 2011
1
−4
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
103
then eigenvalues are solutions of
Let β =
35
−9β − λ
β
= λ2 + 13βλ + 35β2 = 0
β
−4β − λ
giving
−13 ±
λ=
2
√
29
β −2.6 × 102 or − 1.1 × 102
Exercises 1.10.4
1 0
56
Φ (t) = e where A =
1 1
Eigenvalues of A are λ = 1, λ = 1 so
At
eAt = α0 (t)I + α1 (t)A
where α0 , α1 satisfy
eλt = α0 + α1 λ,
λ=1
teλt = α1
giving α1 = tet , α0 = et − tet
Thus,
Φ (t) = e
At
56(a)
Φ (0) =
et − tet
=
0
1
0
0
1
0
t
e − tet
+
tet
tet
0
tet
=
et
tet
0
et
=I
56(b)
t
e1
0
0
et2 e−t1
Φ(t1 ) =
Φ(t2 − t1 )Φ
(t2 − t1 )et2 e−t1 et2 e−t1
t1 et1 et1
t
e2
0
0
et2
=
= Φ(t2 )
=
(t2 − t1 )et2 + t1 et2 et2
t2 et2 et2
c Pearson Education Limited 2011
49
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
t
−t
1
e
e
0
0
−1
=
= Φ (−t)
56(c)
Φ = 2t
−tet et
−te−t e−t
e
50
dy
d2 y
, ẋ2 = 2 = −x1 − 2x2 so in vector–matrix
57 Take x1 = y, x2 = ẋ1 =
dt
dt
form the differential equation is
0
1
x, y = [1 0]A
ẋ =
−1 −2
0
Taking A =
−1
1
−2
its eigenvalues are λ = −1, λ = −1
eAt = α0 I + α1 A where α0 , α1 satisfy
eλt = α0 + α1 λ, λ = −1
teλt = α1
giving α0 = e−t + te−t , α1 = te−t . Thus,
−t
e + te−t
At
e =
−te−t
te−t
−t
e − te−t
Thus, solution of differential equation is
x(t) = eAt x(0), x(0) = [1 1]T
−t
e + 2te−t
=
e−t − 2te−t
giving y(t) = x1 (t) = e−t + 2te−t
The differential equation may be solved directly using the techniques of Chapter 10
of the companion text Modern Engineering Mathematics or using Laplace
transforms. Both approaches confirm the solution
y = (1 + 2t)e−2t
58
Taking A =
1 0
1 1
then from Exercise 56
e
At
et
=
tet
0
et
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
and the required solution is
x(t) = e
At
et
x(0) =
tet
0
et
et
1
=
1
(1 + t)et
0
1
its eigenvalues are λ1 = −3, λ2 = −2.
59 Taking A =
−6 −5
Thus, eAt = α0 I + α1 A where α0 , α1 satisfy
e−3t = α0 − 3α1 , e−2t = α0 − 2α1
α0 = 3e−2t − 2e−3t , α1 = e−2t − e−3t
so
e
At
3e−2t − 2e−3t
=
6e−3t − te−2t
e−2t − e−3t
3e−3t − 2e−2t
Thus, the first term in (6.73) becomes
e
At
x(0) = e
At
[1 − 1] =
T
2e−2t − e−3t
3e−3t − 4e−2t
and the second term is
e
0
6e−2(t−τ ) − 6e−3(t−τ )
dτ
bu(τ)dτ =
2
18e−3(t−τ ) − 12e−2(t−τ )
0
−2(t−τ )
t
3e
− 2e−3(t−τ )
=2
6e−3(t−τ ) − 6e−2(t−τ ) 0
1 − 3e−2t + 2e−3t
=2
6e−2t − 6e−3t
t
A(t−τ )
t
Thus, required solution is
2e−2t − e−3t + 2 − 6e−3t + 4e−3t
x(t) =
3e−3t − 4e−2t + 12e−2t − 12e−3t
2 − 4e−2t + 3e−3t
=
8e−2t − 9e−3t
that is, x1 = 2 − 4e−2t + 3e−3t , x2 = 8e−2t − 9e−3t
c Pearson Education Limited 2011
51
52
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
60
In state space form,
2
0
1
u(t), u(t) = e−t , x(0) = [0 1]T
x +
ẋ =
0
−2 −3
0
1
its eigenvalues are λ1 = −2, λ2 = −1 so
Taking A =
−2 −3
eAt = α0 I + α1 A where α0 , α1 satisfy
e−2t = α0 − 2α1 , e−t = α0 − α1 ⇒ α0 = 2e−t − e−2t , α1 = e−t − e−2t
Thus,
2e−t − e−2t
e =
−2e−t + 2e−2t
−t
e − e−2t
At
and e x(0) =
−e−t + 2e−2t
At
t
(t−τ )
A
0
e−t − e−2t
−e−t + 2e−2t
4e−(t−τ ) − 2e−2(t−τ )
bu(τ)dτ =
−4e−(t−τ ) + 4e−2(t−τ )
0
t
4e−t − 2e−2t eτ
dτ
=
−4e−t + 4e−2t eτ
0
t
4τe−t − 2e−2t eτ
=
−4τe−t + 4e−2t eτ 0
4te−t − 2e−t + 2e−2t
=
−4te−t + 4e−t − 4e−2t
t
e−τ dτ
We therefore have the solution
t
At
eA(t−τ ) bu(τ)dτ
x(t) = e x(0) +
0
−t
4te + e−2t − e−t
=
−4te−t + 3e−t − 2e−2t
that is,
x1 = 4te−t + e−2t − e−t , x2 = −4te−t + 3e−t − 2e−2t
61
3
Taking A =
2
4
1
its eigenvalues are λ1 = 5, λ2 = −1.
eAt = α0 I + α1 A where α0 , α1 satisfy
e5t = α0 + 5α1 , e−t = α0 − α1 ⇒ α0 =
1 5t 5 −t
1
1
e + e , α1 = e5t + e−t
6
6
6
6
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, transition matrix is
1
e
At
=
−t
+ 23 e5t
3e
1 5t
1 −t
3e − 3e
2 5t
3e
1 5t
3e
− 23 e−t
+ 23 e−t
5t
−t
−
e
2e
and eAt x(0) = eAt [1 2]T =
e5t + e−t
t
t
4
0 1
A(t−τ )
A(t−τ )
dτ
e
Bu(τ)dτ =
e
3
1 1
0
0
t
3
t−τ
=
dτ
A
7
0
t 20 5(t−τ ) 11 −(t−τ )
− 3e
3 e
=
10 5(t−τ )
−(t−τ ) dτ
+ 11
0
3 e
3 e
4 5(t−τ ) 11 −(t−τ ) t
− 3e
−3e
=
2 5(t−τ )
− e
+ 11 e−(t−τ ) 0
3 11 −t 3 4 5t
−5 + 3 e + 3 e
=
−t
3 − 11
+ 23 e5t
3 e
Thus, solution is
t
At
eA(t−τ ) Bu(t)dτ
x(t) = e x(0) +
0
8 −t
5t
−5 + 3 e + 10
e
3
=
3 − 83 e−t + 53 e5t
Exercises 1.10.7
62
Eigenvalues of matrix A =
− 23
1
3
4
− 52
are given by
| A − λI |= λ2 + 4λ + 3 = (λ + 3)(λ + 1) = 0
that is, λ1 = −1, λ2 = −3
having corresponding eigenvectors e1 = [3 2]T, e2 = [1 − 2]T.
Denoting the reciprocal basis vectors by
r1 = [r11 r12 ]T , r2 = [r21 r22 ]T
c Pearson Education Limited 2011
53
54
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
and using the relationships rTi ej = δij (i, j = 1, 2) we have
3r11 + 2r12 = 1
r11 − 2r12 = 0
3r21 + 2r22 = 0
r21 − 2r22 = 1
r1 = [ 14
1 T
8]
r2 = [ 14 − 38 ]T
Thus,
rT1 x(0) =
1 1
1 3
+ = 1, rT2 x(0) = − = −1
2 2
2 2
so the spectral form of solution is
x(t) = e−t e1 − e−3t e2
The trajectory is readily drawn showing that it approaches the origin along the
eigenvector e1 since e−3t → 0 faster than e−t . See Figure 1.9 in the text.
−2
2
eigenvalues are λ1 = −6, λ2 = −1 having
63
Taking A =
2 −5
corresponding eigenvectors e1 = [1 − 2]T , e2 = [2 1]T .
Denoting the reciprocal basis vectors by
r1 = [r11 r12 ]T, r2 = [r21 r22 ]T
and using the relationships rTi ej = δij (i, j = 1, 2) we have
r11 − 2r12 = 1
2r11 + r12 = 0
r21 − 2r22 = 0
2r21 + r22 = 1
Thus,
⇒ r11 = 15 , r12 = − 25 ⇒ r1 = 15 [1 − 2]T
⇒ r21 = 25 , r22 = − 15 ⇒ r2 = 15 [2 1]T
4
1
2
=−
= [1 − 2]
3
5
5
7
1
2
=
rT2 x(0) = [2 1]
3
5
5
rT1 x(0)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
55
then response is
2
rTi x(0)eλit ei
x(t) =
i=1
4
= − e−6t
5
7 −t 2
1 −4e−6t + 14e−t
1
+ e
=
−2
1
8e−6t + 7e−t
5
5
Again, following Figure 1.9 in the text, the trajectory is readily drawn and showing
that it approaches the origin along the eigenvector e2 since e−6t → 0 faster than
e−t .
0 −4
eigenvalues are λ1 = −2 + j2, λ2 = −2 − j2 having
64 Taking A =
2 −4
corresponding eigenvectors e1 = [2 1 − j]T , e2 = [2 1 + j]T .
Let r1 = r1 + jr1 be reciprocal base vector to e1 then
rT1 e1 = 1 = [r + jr1 ]T [e1 + je1 ]T where e1 = e1 + je1
rT1 e2 = 0 = [r1 + jr1 ]T [e1 − je1 ]T since e2 = conjugate e1
Thus,
[(r1 )T e1 − (r1 )T e1 ] + j[(r1 )T e1 + (r1 )T e1 ] = 1
and
[(r1 )T e1 − (r1 )T e1 ] + j[(r1 )T e11 − (r1 )T e1 ] = 0
c Pearson Education Limited 2011
56
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
giving
(r1 )T e1 =
1
1
, (r1 )T e1 = , (r1 )T e1 = (r1 )T e1 = 0
2
2
Now e1 = [2 1]T , e1 = [0 − 1]T
Let r1 = [a b]T and r1 = [c d]T then from above
1
1
2a + b = , −b = 0 and −d = − , 2c + d = 0
2
2
1
1
1
giving a = , b = 0, c = − , d = so that
4
4
2
r1 = r1 + jr1 =
1
[1 − j 2j]T
4
Since r2 is the complex conjugate of r1
r2 =
1
[1 + j − 2j]T
4
so the solution is given by
x(t) = rT1 x(0)eλ1 t e1 + rT2 x(0)eλ2 t e2
and since rT1 x(0) =
x(t) = e
−2t
−2t
=e
=e
−2t
1
1
(1 + j), rT2 x(0) = (1 − j)
2
2
1
(1 + j)e2jt
2
2
1−j
1
+ (1 − j)e−2jt
2
2
1+j
0
2
− (cos 2t + sin 2t)
(cos 2t − sin 2t)
−1
1
(cos 2t −
sin 2t)e1
− (cos 2t +
sin 2t)e1
where e1 = e1 + je1
To plot the trajectory, first plot e1 , e1 in the plane and then using these as a frame
of reference plot the trajectory. A sketch is as follows
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
65
57
Following section 1.10.6 if the equations are representative of
ẋ = A x + bu , y = cT x
then making the substitution x = M ξ , where M is the modal matrix of A,
reduces the system to the canonical form
ξ̇ξ = Λ ξ + (M−1 b)u , y = (cT M)ξξ
where Λ is the spectral matrix of A.
Eigenvalues of A are given by
1 − λ
−1
0
1
−2
2−λ
1 = λ3 − 2λ2 − λ + 2 = (λ − 1)(λ + 2)(λ + 1) = 0
1
−1 − λ
so the eigenvalues are λ1 = 2, λ2 = 1, λ3 = −1. The corresponding eigenvectors
are readily determined as
e1 = [1 3 1]T , e2 = [3 2 1]T , e3 = [1 0 1]T
⎡
1
⎣
Thus, M = 3
1
3
2
1
⎤
1
0⎦
1
⎡
2
⎣
and Λ = 0
0
⎤
0 0
1 0 ⎦
0 −1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎤
⎡
2 −2 −2
1
1
adj M = − ⎣ −3
0
3 ⎦ so required canonical form is
M−1 =
det M
6
1
2 −7
58
⎡
⎤
⎡
ξ̇1
2
⎣ ξ̇2 ⎦ = ⎣ 0
0
ξ̇3
0
1
0
⎤ ⎡ ⎤
⎡ 1 ⎤
ξ1
0
3
0 ⎦ ⎣ ξ2 ⎦ + ⎣ 0 ⎦ u
−1
− 43
ξ3
y = [1 − 4 − 2] [ξ1 ξ2 ξ3 ]T
66
Let r1 = [r11 r12 r13 ]T , r2 = [r21 r22 r23 ]T , r3 = [r31 r32 r33 ]T be the
reciprocal base vectors to e1 = [1 1 0]T , e2 = [0 1 1]T , e3 = [1 2 3]T .
⎫
rT1 e1 = r11 + r12 = 1
⎬
rT1 e2 = r11 + r13 = 0
⎭
rT1 e3 = r11 + 2r12 + 3r13 = 0
⎫
rT2 e1 = r21 + r22 = 0
⎬
rT2 e2 = r22 + r23 = 1
⎭
rT2 e3 = r21 + 2r22 + 3r23 = 0
⎫
rT3 e1 = r31 + r32 = 0
⎬
rT3 e2 = r32 + r33 = 0
⎭
rT3 e3 = r31 + 2r32 + 3r33 = 1
⇒ r1 =
1
[1 1 − 1]T
2
⇒ r2 =
1
[−3 3 1]T
2
⇒ r3 =
1
[1 − 1 1]T
2
Then using the fact that x(0) = [1 1 1]T
α0 = rT1 x(0) = − 12 , α1 = rT2 x(0) =
67
1
2
, α3 = rT3 x(0) =
1
2
The eigenvectors of A are given by
5− λ
4
= (λ − 6)(λ − 1) = 0
1
2 − λ
so the eigenvalues are λ1 = 6, λ2 = 1. The corresponding eigenvectors are readily
determined as e1 = [4 1]T , e2 = [1 − 1]T .
4
1
then substituting x = Mξξ
Taking M to be the modal matrix M =
1 −1
into ẋ = Ax(t) reduces it to the canonical form
ξ̇ξ = Λ ξ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
6 0
. Thus, the decoupled canonical form is
where Λ =
0 1
ξ̇1
ξ̇2
=
6
0
0
1
ξ1
ξ2
59
or ξ̇1 = 6ξ1 and ξ̇2 = ξ2
which may be individually solved to give
ξ1 = αe6t and ξ1 = βet
1 −1 −1
1
1
−1
=
Now ξ (0) = M x(0) = −
−3
4
4
5 −1
so ξ1 (0) = 1 = α and ξ2 (0) = −3 = β
giving the solution of the uncoupled system as
6t
e
ξ=
−3et
The solution for x(t) as
x=M ξ =
4
1
1 −1
e6t
−3et
=
4e6t − 3et
e6t + 3et
3 4
its eigenvalues are λ1 = 5, λ2 = −1 having
68
Taking A =
2 1
corresponding
e1 = [2 1]T , e2 = [1 − 1]T .
eigenvectors
2
1
be the modal matrix of A, then ẋ = M ξ reduces the
Let M =
1 −1
equation to
5 0
0 1
−1
ξ +M
ξ̇ξ (t) =
u(t)
0 −1
1 1
1
1 1
1
−1
we have,
Since M =
adj M =
det M
3 1 −2
ξ̇ξ (t) =
5
0
0
−1
1
ξ+
3
1
−2
2
−1
With u(t) = [4 3]T the decoupled equations are
10
3
11
ξ̇2 = −ξ2 −
3
ξ̇1 = 5ξ1 +
c Pearson Education Limited 2011
u(t)
60
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
which can be solved independently to give
2
11
, ξ2 = βe−t −
3
3
1 1
1
1
1
−1
so
=
We have that ξ (0) = M x(0) =
−1
2
3 1 −2
ξ1 = αe5t −
2
5
⇒ α=
3
3
8
11
⇒ β=
−1 = β −
3
3
1=α−
giving
ξ=
5 5t
3e −
8 −t
−
3e
2
3
11
3
5 5t 2
−5 + 83 e−t + 10
e −3
e5t
2 1
3
3
=
and x = M ξ =
8 −t
1 −1
− 11
3 − 83 e−t + 53 e5t
3e
3
which confirms Exercises 57 and 58.
Exercises 1.11.1 (Lyapunov)
69
Take tentative Lyapunov function V(x) = xT Px giving
V̇(x) = xT (AT P + PA)x = −xT Qx where
AT P + PA = −Q
(i)
Take Q = I so that V̇(x) = −(x21 + x22 ) which is negative definite. Substituting in
(i) gives
−4 3
2 −2
p11
p12
p
p12
+ 11
p22
p12
p12
p22
−1
−4 2
=
0
3 −2
0
−1
Equating elements gives
−8p11 + 6p12 = −1, 4p12 − 4p22 = −1, 2p11 − 6p12 + 3p22 = 0
5 2
5
2
11
3
Principal minors
Solving gives p11 = 8 , p12 = 3 , p22 = 12 so that, P = 82 11
of P are:
5
8
3
12
55
> 0 and det P = ( 96
− 49 ) > 0 so P is positive definite and the system
is asymptotically stable
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
5 2
4
11 2
8 x1 + 3 x1 x2 + 12 x2
4
11
2
2
3 x1 ẋ2 + 6 x2 ẋ2 = −x1 − x2
Note that, in this case, we have V(x) =
definite and V̇(x) =
5
4
4 x1 ẋ1 + 3 ẋ1 x2 +
61
which is positive
which is negative
definite.
70
Take tentative Lyapunov function V(x) = xT Px giving
V̇(x) = xT (AT P + PA)x = −xT Qx where
(i)
AT P + PA = −Q
Take Q = I so that V̇(x) = −(x21 + x22 ) which is negative definite. Substituting in
(i) gives
−3 −1
2 −1
p
p12
+ 11
p22
p12
p11
p12
p12
p22
−1
−3 2
=
0
−1 −1
0
−1
Equating elements gives
−6p11 − 2p12 = −1, 4p12 − 2p22 = −1, 2p11 − 4p12 − p22 = 0
Solving gives p11 =
7
40 , p12
Principal minors of P are:
=
1
− 40
, p22
7
40
=
18
40
so that P =
> 0 and det P =
5
64
7
40
1
− 40
1
− 40
18
40
> 0 so P is positive definite
and the system is asymptotically stable.
71
Take tentative Lyapunov function V(x) = xT Px giving
V̇(x) = xT (AT P + PA)x = −xT Qx where
AT P + PA = −Q
(i)
Take Q = I so that V̇(x) = −(x21 + x22 ) which is negative definite. Substituting in
(i) gives
0 −a
1 −b
p11
p12
p
p12
+ 11
p22
p12
p12
p22
0
−a
−1
1
=
0
−b
0
−1
Equating elements gives
−8p12 = −1, 2p12 − 2bp22 = −1, p11 − bp12 − ap22 = 0
c Pearson Education Limited 2011
62
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Solving gives p12 =
1
2a , p22
=
a+1
2ab , p11
=
b2 +a2 +a
2ab
b2 +a2 +a
so that, P =
2ab
1
2a
1
2a
a+1
2ab
For asymptotic stability the principal minors of P must be positive. Thus,
b2 + a2 + a
>0
2ab
(ii)
and (b2 + a2 + a)(a + 1) > b2
(iii)
Case 1 ab > 0
(ii) ⇒ a2 + b2 + a > 0 so (iii) ⇒ a + 1 >
b2
b2 + a2 + a
⇒ a[a2 + (a + 1)2 ] > 0 ⇒ a > 0.
Since ab > 0 ⇒ b > 0 it follows that (ii) and (iii) are satisfied if a, b > 0
Case 2 ab < 0 No solution to (ii) and (iii) in this case.
Thus, system is asymptotically stable when both a > 0 and b > 0.
Note: This example illustrates the difficulty in interpretating results when using
the Lyapunov approach. It is a simple task to confirm this result using the Routh–
Hurwitz criterion developed in Section 5.6.2.
72(a)
ẋ1 = x2
(i)
ẋ2 = −2x2 + x3
(ii)
ẋ3 = −kx1 − x3
(iii)
If V̇(x) is identically zero then x3 is identically zero ⇒ x1 is identically zero from
(iii)
⇒ x2 is identically zero from (i)
Hence V̇(x) is identically zero only at the origin.
(b) AT P + PA = −Q ⇒
⎡
0
⎣1
0
0
−2
1
⎤⎡
p11
−k
⎦
⎣
0
p12
−1
p13
p12
p22
p23
⎤ ⎡
p11
p13
⎦
⎣
p23 + p12
p33
p13
p12
p22
p23
⎤⎡
0
1
p13
⎦
⎣
0 −2
p23
−k 0
p33
c Pearson Education Limited 2011
⎤ ⎡
0
0
⎦
⎣
1 = 0
0
−1
0
0
0
⎤
0
0 ⎦
−1
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
63
Equating elements and solving for the elements of P gives the matrix
⎡ k2 +12k
P=⎣
6k
12−2k
3k
12−2k
k
12−2k
12−2k
6k
12−2k
0
0
k
12−2k
6
12−2k
⎤
⎦
(c) Principal minors of Pare:
k2 + 12k
> 0 if k > 0and(12 − 2k) > 0 ⇒ 0 < k < 6
122− 2k
36k2
3k
k + 12k
3k3
−
Δ2 =
> 0 if k > 0
=
12 − 2k
12 − 2k
12 − 2k
(12 − 2k)2
(k2 + 12k)(8k − k2 )
216k2
−
> 0if (6k3 − k4 ) > 0 ⇒ 0 < k < 6
Δ3 =
(12 − 2k)3
(12 − 2k)3
Δ1 =
Thus system asymptotically stable for 0 < k < 6.
73
State-space form is
ẋ1
ẋ =
ẋ2
0
=
−k
1
−a
x1
x2
(i)
Take V(x) = kx21 + (x2 + ax1 )2 then
V̇(x) = 2kx1 ẋ1 + 2(x2 + ax1 )(ẋ2 + aẋ1 )
= 2kx1 (x2 ) + 2(x2 + ax1 )(−kx1 − ax2 + ax1 )using (i)
= −2kax21
Since k>0 and a>0 then V̇(x) is negative semidefinite but is not identically zero
along any trajectory of (i). Consequently, this choice of Lyapunov function assures
asymptotic stability.
Review Exercises 1.13
1(a)
Eigenvalues given by
−1 − λ
6
0
−13 − λ
0
−9
12
30 = (1 + λ)[(−13 − λ)(20 − λ) + 270] = 0
20 − λ
c Pearson Education Limited 2011
64
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
that is, (1 + λ)(λ − 5)(λ − 2) = 0
so eigenvalues are λ1 = 5, λ2 = 2, λ3 = −1
Eigenvectors are given by corresponding solutions of
⎤⎡
⎤
⎡
ei1
6
12
−1 − λi
⎣
0
−13 − λi
30 ⎦ ⎣ ei2 ⎦ = 0
0
−9
20 − λi
ei3
When i = 1, λi = 5 and solution given by
e11
−e12
e13
=
=
= β1
198
−90
54
so e1 = [11 5 3]T
When i = 2, λi = 2 and solution given by
e21
−e22
e23
=
=
= β2
216
−54
27
so e2 = [8 2 1]T
When i = 3, λi = −1 and solution given by
e31
−e32
e33
=
=
= β3
1
0
0
so e3 = [1 0 0]T
1(b)
Eigenvalues given by
2 − λ
0
1
4 − λ −1
+ −1
−1
4−λ
−1 =
−1
2
−λ
−1
2
0 − λ
4 − λ
=0
2
that is, 0 = (2 − λ)[(4 − λ)(−λ) + 2] + [−2 + (4 − λ)]
= (2 − λ)(λ2 − 4λ + 3) = (2 − λ)(λ − 3)(λ − 1) = 0
so eigenvalues are
λ1 = 3, λ2 = 2, λ3 = 1
Eigenvectors are given by the corresponding solutions of
(2 − λi )ei1 + 0ei2 + ei3 = 0
−ei1 + (4 − λi )ei2 − ei3 = 0
−ei1 + 2ei2 − λi ei3 = 0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
65
Taking i = 1, 2, 3 gives the eigenvectors as
e1 = [1 2 1]T , e2 = [2 1 0]T , e3 = [1 0 − 1]T
1(c)
Eigenvalues given by
1 − λ
−1
0
−1
2−λ
−1
−1
−1
that is, λ −1 2 − λ
0
−1
0
−1 R1 + (R2 + R3 )
1 − λ
−1
−1 = λ
1− λ
−λ
−1
0
−1
0
−1 3 − λ
0
−1
−λ
2−λ
−1
−λ
−1 = 0
1 − λ
0
0 = λ(3 − λ)(1 − λ) = 0
1 − λ
so eigenvalues are λ1 = 3, λ2 = 1, λ3 = 0
Eigenvalues are given by the corresponding solutions of
(1 − λi )ei1 − ei2 − 0ei3 = 0
−ei1 + (2 − λi )ei2 − ei3 = 0
0ei1 − ei2 + (1 − λi )ei3 = 0
Taking i = 1, 2, 3 gives the eigenvectors as
e1 = [1 − 2 1]T , e = [1 0 − 1]T , e3 = [1 1 1]T
2
Principal stress values (eigenvalues) given by
6 − λ
3− λ
2
1
2
2
R
3
−
λ
1
+
(R
+
R
)
1
2
3
1
1
1
4− λ
1
1
1
1 = 0
= (6 − λ) 2 3 − λ
1
1
4 − λ
1
that is, (6 − λ) 2
1
6 − λ 6 − λ
3−λ
1
1
4 − λ
0
0
1−λ
−1 = (6 − λ)(1 − λ)(3 − λ) = 0
0
3 − λ
so the principal stress values are λ1 = 6, λ2 = 3, λ3 = 1.
c Pearson Education Limited 2011
66
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Corresponding principal stress direction e1 , e2 and e3 are given by the solutions
of
(3 − λi )ei1 + 2ei2 + ei3 = 0
2ei1 + (3 − λi )ei2 + ei3 = 0
ei1 + ei2 + (4 − λi )ei3 = 0
Taking i = 1, 2, 3 gives the principal stress direction as
e1 = [1 1 1]T, e2 = [1 1 − 2]T, e3 = [1 − 1 0]T
It is readily shown that eT1 e2 = eT1 e3 = eT2 e3 = 0 so that the principal stress
directions are mutually orthogonal.
3
Since [1 0 1]T is an eigenvector of A
⎡
2
⎣ −1
0
−1
3
b
⎡ ⎤
⎤ ⎡ ⎤
1
1
0
b⎦ ⎣0⎦ = λ ⎣0⎦
1
1
c
so 2 = λ, −1 + b = 0, c = λ
giving b = 1 and c = 2.
Taking these values A has eigenvalues given
2 − λ
−1
0
−1
3−λ
1 = (2 − λ)
0
1
2− λ
by
3− λ
1
1
− (2 − λ)
2 − λ
= (2 − λ)(λ − 1)(λ − 4) = 0
that is, eigenvalues are λ1 = 4, λ2 = 2, λ3 = 1
Corresponding eigenvalues are given by the solutions of
(2 − λi )ei1 − ei2 + 0ei3 = 0
−ei1 + (3 − λi )ei2 + ei3 = 0
0ei1 + ei2 + (2 − λi )ei3 = 0
Taking i = 1, 2, 3 gives the eigenvectors as
e1 = [1 − 2 − 1]T , e2 = [1 0 1]T , e3 = [1 1 − 1]T
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
4
67
The three Gerschgorin circles are
| λ − 4 | =| −1 | + | 0 |= 1
| λ − 4 | =| −1 | + | −1 |= 2
|λ−4|=1
Thus, | λ − 4 |≤ 1 and | λ − 4 |≤ 2 so | λ − 4 |≤ 2 or 2 ≤ λ ≤ 6.
Taking x(o) = [−1 1 − 1]T iterations using the power method may be tabulated
as follows
Iteration k
x(k)
A x(k)
λ
0
−1
1
−1
−5
6
−5
6
1
−0.833
1
−0.833
−4.332
5.666
−4.332
5.666
2
−0.765
1
−0.765
−4.060
5.530
−4.060
5.530
3
−0.734
1
−0.734
−3.936
5.468
−3.936
5.468
4
−0.720
1
−0.720
−3.88
5.44
3.88
5.44
5
−0.713
1
−0.713
−3.852
5.426
−3.852
5.426
6
−0.710
1
−0.710
Thus, correct to one decimal place the dominant eigenvalue is λ = 5.4
5(a)
Taking x (o) = [1 1 1]7 iterations may be tabulated as follows
Iteration k
x(k)
A x(k)
λ
0
1
1
1
4
4.5
5
5
1
0.800
0.900
1
3.500
4.050
4.700
4.700
2
0.745
0.862
1
3.352
3.900
4.607
4.607
3
0.728
0.847
1
3.303
3.846
4.575
4.575
4
0.722
0.841
1
3.285
3.825
4.563
4.563
5
0.720
0.838
1
3.278
3.815
4.558
4.558
6
0.719
0.837
1
3.275
3.812
4.556
4.556
7
0.719
0.837
1
Thus, estimate of dominant eigenvalues is λ 4.56 with associated eigenvector
x = [0.72 0.84 1]T
5(b)
3
i=1
λi = trace A ⇒ 7.5 = 4.56 + 1.19 + λ3 ⇒ λ3 = 1.75
c Pearson Education Limited 2011
68
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
5(c) (i)
det A =
3
λi = 9.50 so A−1 exists and has eigenvalues
i=1
1
1
1
,
,
1.19 1.75 4.56
so power method will generate the eigenvalue 1.19 corresponding to A.
(ii)
A − 3I has eigenvalues
1.19 − 3, 1.75 − 3, 4.56 − 3
that is, −1.91, −1.25, 1.56
so applying the power method on A − 3I generates the eigenvalues corresponding
to 1.75 of A.
6
ẋ = αλeλt , ẏ = βλeλt , ż = γλeλt so the differential equations become
αλeλt = 4αeλt + βeλt + γeλt
βλeλt = 2αeλt + 5βeλt + 4γeλt
γλeλt = −αeλt − βeλt
Provided eλt = 0 (i.e. non-trivial solution) we have the eigenvalue problem
⎡
4
⎣ 2
−1
Eigenvalues given
4 − λ
1
2
5
−
λ
−1
−1
⎡ ⎤
⎤ ⎡ ⎤
α
α
1 1
⎦
⎣
⎦
⎣
β = λ β⎦
5 4
γ
γ
−1 0
by
4− λ
1
4 C2 −C3 2
−1
0
0
1−λ
λ−1
4 − λ
1
4 = (λ − 1) 2
−1
−λ
0
−1
1
1
4
−λ
= −(λ − 1)(λ − 5)(λ − 3)
so its eigenvalues are 5, 3 and 1.
When λ = 1 the corresponding eigenvector is given by
3e11 + e12 + e13 = 0
2e11 + 4e12 + 4e13 = 0
−e11 − e12 − e13 = 0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
−e12
e13
e11
=
=
= β1
0
2
2
Thus, corresponding eigenvector is β[0 − 1 1]T
having solution
7
Eigenvalues are given by
8− λ
−8
−2
−3 − λ
−2 = 0
| A − λI | = 4
3
−4
1 − λ
Row 1 − (Row 2 + Row 3) gives
1 − λ
| A − λI | = 4
3
1
−1
−1
−1 + λ
−2
−2 = (1 − λ) 4 −3 − λ
3
−4
1 − λ
1−λ
0
0
1−λ
2 = (1 − λ)[(1 − λ)(4 − λ) + 2]
−1
4 − λ
−1 + λ
−3 − λ
−4
1
= (1 − λ) 4
3
= (1 − λ)(λ − 2)(λ − 3)
Thus, eigenvalues are λ1 = 3, λ2 = 2, λ3 = 1.
Corresponding eigenvectors are given by
(8 − λ)ei1 − 8ei2 − 2ei3 = 0
4ei1 − (3 + λ)ei2 − 2ei3 = 0
3ei1 − 4ei2 + (1 − λ)ei3 = 0
When i = 1, λi = λ1 = 3 and solution given by
e11
−e12
e13
=
=
= β1
4
−2
2
so a corresponding eigenvector is e1 = [2 1 1]T .
When i = 2, λi = λ2 = 2 and solution given by
e21
−e22
e23
=
=
= β2
−3
2
−1
so a corresponding eigenvector is e2 = [3 2 1]T .
When i = 3, λi = λ3 = 1 and solution given by
e31
−e32
e33
=
=
= β3
−8
6
−4
c Pearson Education Limited 2011
69
70
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so a corresponding eigenvector is e3 = [4 3 2]T .
Corresponding modal and spectral matrices are
⎤
⎤
⎡
⎡
3 0 0
2 3 4
M = ⎣ 1 2 3 ⎦ and Λ = ⎣ 0 2 0 ⎦
1 0 1
1 1 2
⎤
⎡
1 −2 1
M−1 = ⎣ 1
0 −2 ⎦ and matrix multiplication confirms M−1 A M = Λ
−1 1
1
8
Eigenvectors of A are given by
1 − λ
0
−4
=0
0
5
−
λ
4
−4
4
3 − λ
that is, λ3 − 9λ2 − 9λ + 81 = (λ − 9)(λ − 3)(λ + 3) = 0
so the eigenvalues are λ1 = 9, λ2 = 3 and λ3 = −3.
The eigenvectors are given by the corresponding solutions of
(1 − λi )ei1 + 0ei2 − 4ei3 = 0
0ei1 + (5 − λi )ei2 + 4ei3 = 0
−4ei1 + 4ei2 + (3 − λi )ei3 = 0
Taking i = 1, 2, 3 the normalized eigenvectors are given by
ê1 = [ 13
−2 −2 T
3
3 ] , ê2
The normalised modal matrix
= [ 23
⎡
1
1 ⎣
−2
M̂ =
3
−2
so
2 −1 T
3 3 ] , ê3
2
2
−1
= [ 23
−1 2 T
3 3]
⎤
2
−1 ⎦
2
⎤ ⎡
⎤ ⎡
1
1 0 −4
1 −2 −2
1 ⎣
M̂T A M̂ =
2 2 −1 ⎦ ⎣ 0 5 4 ⎦ ⎣ −2
9
−2
−4 4 3
2 −1 2
⎤
⎡
9 0 0
= ⎣0 3 0 ⎦ = Λ
0 0 −3
⎡
c Pearson Education Limited 2011
2
2
−1
⎤
2
−1 ⎦
2
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
71
⎤
−6
0
0 0
0 0⎥
⎢ 6 −4
9 Ṅ = ⎣
⎦ N, N = [N1 N2 N3 N4 ]T
0
4 −2 0
0
0
2 0
Since the matrix A is a triangular matrix its eigenvalues are the diagonal elements.
⎡
Thus, the eigenvalues are
λ1 = −6, λ2 = −4, λ3 = −2, λ4 = 0
The eigenvectors are the corresponding solutions of
(−6 − λi )ei1 + 0ei2 + 0ei3 + 0ei4 = 0
6ei1 + (−4 − λi )ei2 + 0ei3 + 0ei4 = 0
0ei1 + 4ei2 + (−2 − λi )ei3 + 0ei4 = 0
0ei1 + 0ei2 + 2ei3 − λi ei4 = 0
Taking i = 1, 2, 3, 4 and solving gives the eigenvectors as
e1 = [1 − 3 3 − 1]T , e2 = [0 1 − 2 1]T
e3 = [0 0 1 − 1]T , e4 = [0 0 0 1]T
Thus, spectral form of solution to the equation is
N = αe−6t e1 + βe−4t e2 + γe−2t e3 + δe4
Using the given initial conditions at t = 0 we have
⎤
⎡ ⎤
⎤
⎡
⎤
⎡
⎤
⎡
0
0
0
1
C
⎢0⎥
⎢ 0⎥
⎢ 1⎥
⎢ −3 ⎥
⎢0⎥
⎦ +δ ⎣ ⎦
⎦ +γ ⎣
⎦ +β ⎣
⎣ ⎦ =α ⎣
0
1
−2
3
0
1
−1
1
−1
0
⎡
so C = α, 0 = −3α + β, 0 = 3α − 2β + γ, 0 = −α + β − γ + δ
which may be solved for α, β, γ and δ to give
α = C, β = 3C, γ = 3C, δ = C
Hence,
N4 = −αe−6t + βe−4t − γe−2t + δ
= −Ce−6t + 3Ce−4t − 3Ce−2t + C
c Pearson Education Limited 2011
72
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
10(a)
(i)
Characteristic equation of A is λ2 − 3λ + 2 = 0 so by the Cayley–Hamilton
theorem
4
A = 3A − 2I =
3
0
1
2
8 0
A = 3(3A − 2I) − 2A = 7A − 6I =
7 1
16 0
4
A = 7(3A − 2I) − 6A = 15A − 14I =
15 1
32 0
5
A = 15(3A − 2I|) − 14A = 31A − 30I =
31 1
64 0
6
A = 31(3A − 2I) − 30A = 63A − 62I =
63 1
128 0
7
A = 63(3A − 2I) − 62A = 127A − 126I =
127 1
−29 0
7
6
4
3
2
Thus, A − 3A + A + 3A − 2A + 3I =
−32 3
3
(ii)
Eigenvalues of A are λ1 = 2, λ2 = 1. Thus,
Ak = α0 I + α1 A where α0 and α1 satisfy
2k = α0 + 2α1 , 1 = α0 + α1
α1 = 2k − 1, α0 = 2 − 2k
Thus, Ak =
10(b)
α0 + 2α1
α1
0
α0 + α1
=
2k
k
2 −1
0
1
Eigenvalues of A are λ1 = −2, λ2 = 0. Thus,
Thus, eAt
eAt = α0 I + α1 A where α0 and α1 satisfy
1
e−2t = α0 − 2α1 , 1 = α0 ⇒ α0 = 1, α1 = (1 − e−2t )
2
1 12 (1 − e−2t )
α1
α0
=
=
0
e−2t
0 α0 − 2α1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
73
⎤
1 2 3
11 The matrix A = ⎣ 0 1 4 ⎦ has the single eigenvalue λ = 1 (multiplicity 3)
0 0 1
⎤
⎤
⎡
⎡
0 1 0
0 2 3
(A − I) = ⎣ 0 0 4 ⎦ ∼ ⎣ 0 0 1 ⎦ is of rank 2 so has nullity 3 − 2 = 1
0 0 0
0 0 0
indicating that there is only one eigenvector corresponding to λ = 1.
⎡
This is readily determined as
e1 = [1 0 0]T
The corresponding Jordan canonical form comprises a single block so
⎤
1 1 0
J = ⎣0 1 1⎦
0 0 1
⎡
Taking T = A − I the triad of vectors (including generalized
eigenvectors)
has
⎤
⎡
0 0 8
the form {T2 ω, T ω, ω} with T2 ω = e1 . Since T2 = ⎣ 0 0 0 ⎦ , we may take
0 0 0
1 T
2 1
T
ω = [0 0 8 ] . Then, T ω = [ 8 8 0] . Thus, the triad of vectors is
e1 = [1 0 0]T , e∗1 = [ 38
1
2
1 T
0]T , e∗∗
1 = [0 0 8 ]
The corresponding modal matrix is
⎡
1
M = ⎣0
0
⎡
1
16
M−1 = 16 ⎣ 0
0
⎡ 1
16
M−1 A M = 16 ⎣ 0
0
⎡
1 1
= ⎣0 1
0 0
3
− 64
1
8
0
3
− 64
1
8
0
⎤
3
8
1
2
0
⎤
0
0⎦
1
8
⎤
0
0 ⎦ and by matrix multiplication
1
2
⎤ ⎡
1
0
⎦
⎣
0
0
1
0
2
2
1
0
⎤ ⎡
1
3
⎦
⎣
4
0
1
0
0
1⎦ = J
1
c Pearson Education Limited 2011
3
8
1
2
0
⎤
0
0⎦
1
8
74
12
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Substituting x = X cos ωt, y = Y cos ωt, z = Z cos ωt gives
−ω2 X = −2X + Y
−ω2 Y = X − 2Y + Z
−ω2 Z = Y − 2Z
or taking λ = ω2
(λ − 2)X + Y = 0
X + (λ − 2)Y + Z = 0
Y + (λ − 2)Z = 0
For non-trivial solution
λ − 2
1
0
1
λ−2
1
0
1 = 0
λ − 2
that is, (λ − 2)[(λ − 2)2 − 1] − (λ − 2) = 0
(λ − 2)(λ2 − 4λ + 2) = 0
√
so λ = 2 or λ = 2 ± 2
When λ = 2 , Y = 0 and X = −Z so X : Y : Z = 1 : 0 : −1
√
√
√
When λ = 2 + 2 , X = Z and Y = − 2X so X : Y : Z = 1 : − 2 : 1
√
√
√
When λ = 2 − 2 , X = Z and Y = 2X so X : Y : Z = 1 : 2 : 1
13
In each section A denotes the matrix of the quadratic form.
⎤
2 −1
0
2 −1
= 1 and
1 −1 ⎦ has principal minors of 2,
13(a) A = ⎣ −1
−1 1
0 −1
2
det A = 0
⎡
so by Sylvester’s condition (c) the quadratic form is positive-semidefinite.
⎡
3
13(b) A = ⎣ −2
−2
det A = 6
−2
7
0
⎤
−2
3
0 ⎦ has principal minors of 3,
−2
2
−2
= 17 and
7
so by Sylvester’s condition (a) the quadratic form is positive-definite.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎡
16
13(c) A = ⎣ 16
16
det A = −704
16
36
8
⎤
16
16
⎦
8
has principal minors of 16,
16
17
75
16
= 320 and
36
so none of Sylvester’s conditions are satisfied and the quadratic form is indefinite.
⎡
−21
13(d) A = ⎣ 15
−6
and det A = 0
15
−11
4
⎤
−6
−21
⎦
4 has principal minors of −21,
15
−2
15
=6
−11
so by Sylvester’s condition (d) the quadratic form is negative-semidefinite.
⎤
−1
1
1
−1 1
= 2 and
1 ⎦ has principal minors of −1,
13(e) A = ⎣ 1 −3
1 −3
1
1 −5
det A = −4 so by Sylvester’s condition (b) the quadratic form is negative-definite.
⎡
⎡
14
7
2
A e1 = ⎣ 4
− 32
− 21
−1
3
2
⎡ ⎤
⎤ ⎡ ⎤
1
1
− 12
⎦
⎣
⎦
⎣
2 = 2⎦
0
1
3
3
2
Hence, e1 = [1 2 3]T is an eigenvector with λ1 = 1 the corresponding eigenvalue.
Eigenvalues are given by
7
− − λ
− 12
2
4
−1 − λ
0 =
3
−3
2
2
− 12
0 = −λ3 + 3λ2 + λ − 3
1
2 −λ
= (λ − 1)(λ2 + 2λ + 3)
= −(λ − 1)(λ − 3)(λ + 1)
so the other two eigenvalues are λ2 = 3, λ3 = −1.
Corresponding eigenvectors are the solutions of
(− 27 − λi )ei1 − 12 ei2 − 12 ei3 = 0
4ei1 − (1 + λi )ei2 + 0ei3 = 0
− 23 ei1 + 32 ei2 + ( 21 − λi )ei3 = 0
c Pearson Education Limited 2011
76
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Taking i = 2, 3 gives the eigenvectors as
e2 = [1 1 0]T , e3 = [0 − 1 1]T
The differential equations can be written in the vector–matrix form
ẋ = A x , x = [x y z]T
so, in special form, the general solution is
x = αeλ1 t e1 + βeλ2 t e2 + γeλ3 t e3
⎤
⎡ ⎤
⎡ ⎤
⎡
1
1
0
= αet ⎣ 2 ⎦ + βe3t ⎣ 1 ⎦ + γe−t ⎣ −1 ⎦
3
0
1
With x(0) = 2, y(0) = 4, z(0) = 6 we have
α = 2, β = 0, γ = 0
⎡ ⎤
1
x = 2et ⎣ 2 ⎦
3
so
that is, x = 2et , y = 4et , z = 6et .
15(a)
1.2
AA =
1.6
T
⎤
1.2 1.6
18.25
−4 ⎣
0.9 1.2 ⎦ =
−9
3
−4 3
0.9
1.2
⎡
−9
13
Eigenvalues λi given by
(18.25 − λ)(13 − λ) − 81 = 0 ⇒ (λ − 25)(λ − 6.25) = 0
⇒ λ1 = 25, λ2 = 6.25
having corresponding eigenvectors
u1 = [ −4
u2 = [ 3
T
3 T
5 ]
3 ] ⇒ û1 = [ − 45
T
4 ] ⇒ û2 = [ 35
4
5
]
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
77
leading to the orthogonal matrix
Û =
3
5
⎤
1.6
1.2
1.2 ⎦
1.6
3
⎡
1.2
AT A = ⎣ 0.9
−4
− 54
0.9
1.2
3
5
4
5
⎡
4
−4
= ⎣3
3
0
3
2.25
0
⎤
0
0 ⎦
25
Eigenvalues μi given by
(25 − μ) [(4 − μ)(2.25 − μ) − 9] = 0 ⇒ (25 − μ)μ(μ − 6.25) = 0
⇒ μ1 = 25, μ2 = 6.25, μ3 = 0
with corresponding eigenvalues
v1 = v̂1 = [ 0
v2 = [ 4
1]
T
T
0 ] ⇒ v̂2 = [ 45
3
v3 = [ −3
0
4
3
5
0]
T
0 ] ⇒ v̂3 = [ − 35
4
5
T
T
0]
leading to the orthogonal matrix
⎡
0
⎣
V̂ = 0
1
4
5
3
5
− 35
0
0
4
5
⎤
⎦
√
√
=
25
=
5
and
σ
=
6.25 = 2.5 so that
The singular values
of
A
are
σ
1
2
5 0 0
giving the SVD form of A as
Σ=
0 2.5 0
T
A = ÛΣV̂ =
−0.8
0.6
0.6
0.8
5
0
0
2.5
1.2 0.9
(Direct multiplication confirms A =
1.6 1.2
⎡
0
0 ⎣
0.8
0
−0.6
−4
)
3
c Pearson Education Limited 2011
0
0.6
0.8
⎤
1
0⎦
0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
78
⎡
⎤⎡ 1
4
3
0
−
5
5
5
∗
4 ⎦⎣
0
(b) A† = V̂Σ ÛT = ⎣ 0 35
5
1 0
0
0
⎤
⎡
0.192 0.256
= ⎣ 0.144 0.192 ⎦
−0.16 0.12
0
2
5
⎤
⎦
0
− 54
3
5
3
5
4
5
⎡
24
1 ⎣
18
= 125
−20
⎤
32
24 ⎦
15
AA† = I
CHECK
LHS =
1
125
1.2 0.9
1.6 1.2
⎤
24 32
−4 ⎣
18 24 ⎦ =
1
−24 15
⎡
1
125
125
0
0
= I = RHS
125
(c) Since A is of full rank 2 and there are more columns than rows
⎤
⎤
⎡
−1
1.2 1.6
1.2 1.6
18.25 −9
1
⎣ 0.9 1.2 ⎦ 13
= ⎣ 0.9 1.2 ⎦
= 156.25
−9
13
9
−4 ⎡3
−4
⎤ 3
⎤ ⎡
0.192 0.256
30
40
1
⎣ 22.5
= 156.25
30 ⎦ = ⎣ 0.144 0.192 ⎦
−0.16 0.12
−25 18.25
⎡
†
T
T −1
A = A (AA )
9
18.25
which checks with the answer in (b).
16 (a)
Using partitioned matrix multiplication the SVD form of A may be
expressed in the
form
T
A = ÛΣV̂ = [ Ûr
S
Ûm−r ]
0
0
0
V̂Tr
V̂Tn−r
T
= Ûr SV̂r
(b) Since the diagonal elements in S are non-zero the pseudo inverse may be expressed
in the form
∗
A† = V̂Σ ÛT = V̂r S−1 ÛTr
⎤
1 −1
(c) From the solution to Q46, exercises 1.8.4, the matrix A = ⎣ −2 2 ⎦ has a single
2 −2
√
√
singularity σ1 = 18 so r = 1 and S is a scalar 18; Ûr = Û1 = û1 = [ 13 − 23 23 ]T
⎡
and
V̂r = V̂1 = v̂1 =
√1
2
− √12
T
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
79
The SVD form of A is
⎡
A = û1 Sv̂T1 =
1
3
⎣ −2
3
2
3
⎤
√
⎦ 18
⎡
1
⎣
with direct multiplication confirming A = −2
2
Thus, the pseudo inverse is
A† = v̂1 S−1 ûT1 =
√1
2
√
− 12
=
√1
2
⎤
−1
2 ⎦
−2
√1
18
1
18
− √12
[
1
−1
1
3
− 23
−2
2
2
3
2
−2
]=
1
6
− 16
[ 13
− 32
2
3
]
which agrees with the answer obtained in Q46, Exercises 1.8.4
17
ẋ = A x + bu , y = cT x
Let λi , ei , i = 1, 2, . . . , n, be the eigenvalues and corresponding eigenvectors of A.
Let M = [e1 , e2 , . . . , en ] then since λi ’s are distinct the ei ’s are linearly
independent and M−1 exists. Substituting x = M ξ gives
M ξ̇ξ = A M ξ + bu
Premultiplying by M−1 gives
ξ̇ξ = M−1 A M ξ + M−1 bu = Λ ξ + b1 u
where Λ = M−1 A M = (λi δij ), i, j = 1, 2, . . . , n, and b1 = M−1 b
Also, y = cT x ⇒ y = cT Mξξ = cT1 ξ , cT1 = cT M. Thus, we have the desired
canonical form.
If the vector b1 contains a zero element then the corresponding mode is
uncontrollable and consequently (A1 b1 c) is uncontrollable. If the matrix cT
has a zero element then the system is unobservable.
The eigenvalues of A are λ1 = 2, λ2 = 1, λ3 = −1 having corresponding
eigenvectors e1 = [1 3 1]T , e2 = [3 2 1]T and e3 = [1 0 1]T .
c Pearson Education Limited 2011
80
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The modal matrix
⎡
1
⎣
M = [e1 e2 e3 ] = 3
1
3
2
1
so canonical form is
⎡
⎤
⎡
ξ̇1
2
⎣ ξ̇2 ⎦ = ⎣ 0
0
ξ̇3
⎤
⎡
1
2
1 ⎣
−1
⎦
0 with M = −
−3
6
1
1
0
1
0
⎤
−2 −2
0
3 ⎦
2 −7
⎤ ⎡ ⎤
⎡ 1 ⎤
ξ1
0
3
0 ⎦ ⎣ ξ2 ⎦ + ⎣ 0 ⎦ u
−1
− 43
ξ3
y = [1 − 4 − 2][ξ1 ξ2 ξ3 ]T
We observe that the system is uncontrollable but observable. Since the system
matrix A has positive eigenvalues the system is unstable. Using Kelman matrices
⎤
⎡ ⎤
⎡ ⎤
2
0 1 1
0
(i) A2 = ⎣ −3 4 3 ⎦ , A b = ⎣ 2 ⎦ , A2 b = ⎣ 4 ⎦
2
−1 1 2
0
⎤
⎤
⎡
⎡
1 0 0
−1 2 0
Thus, [b A b A2 b] = ⎣ 1 2 4 ⎦ ∼ ⎣ 0 1 0 ⎦ and is of rank 2
0 0 0
−1 2 0
so the system is uncontrollable.
⎡
⎡
−2
(ii) [c AT c (AT )2 c] = ⎣ 1
0
so the system is observable.
18
−3
0
5
⎤
⎡
0
−3
2 ⎦ ∼ ⎣1
0
1
0
0
1
⎤
1
0 ⎦ and is of full rank 3
0
Model is of form ẋ = Ax + Bu and making the transformation x = Mz gives
Mż = AMz + Bu ⇒ ż = M−1 AMz + M−1 Bu ⇒ ż = Λz + M−1 Bu
where M and Λ are respectively the modal and spectral matrices ofA.
The eigenvalues of A are given by
−2 − λ
0
0
−2
0
−λ
1 = 0 ⇒ −(2 − λ)(4λ + λ2 + 3) = 0
−3 −4 − λ
⇒ (λ + 2)(λ + 1)(λ + 3) = 0
⇒ λ1 − 1, λ2 = −2, λ3 = −3
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
81
with corresponding eigenvectors
e1 = [ −2
T
1 −2 ] , e2 = [ 1
0
T
0 ] and e3 = [ −2
−1
3]
T
Thus, the modal and spectral matrices are
⎤
⎡
−1
1 −2
⎦
⎣
0 −4 andΛ = 0
0
0 −1
⎡
−2
⎣
M= 1
−1
⎡
and det M = −2 ⇒ M−1
⎡1
2
=⎣3
1
2
⎤
0
⎣
= 1
0
1
2
3
2
0
−2
0
⎡
⎤
0
−1
⎣
⎦
2 ⇒M B= 1
1
0
2
4
1
2
2
6 ⎦ leading to the canonical form
1
⎤⎡ ⎤ ⎡ 1
⎡ ⎤ ⎡
z1
−1 0
0
ż1
2
⎦
⎣
⎦
⎣
⎣
z2 ⎦ + ⎣ 3
ż = ż2 = 0 −2 0
1
0
0 −3
ż3
z3
2
⎤
0
0 ⎦
−3
1
2
3
2
⎤⎡
1
⎦
⎣
0
2
1
1
2
4
1
2
⎤
0
1⎦
1
⎤
2
u
6⎦ 1
u2
1
From (1.99a) the solution is given by
⎡
⎤ ⎡ −t
z1
e
⎣ z2 ⎦ = ⎣ 0
0
z3
⎤
⎤⎡
⎡
t e−(t−τ )
z1 (0)
0
⎣ 0
0 ⎦ ⎣ z2 (0) ⎦ +
0
e−3t
z3 (0)
0
0
e−2t
0
⎤
2
2
⎣ 3 6 ⎦ τ dτ
1
1
1
2
⎤⎡ ⎤
1
3
10
2
2
⎦
⎣
5 ⎦ = [ 17
4 2
2
1
1
2
2
2
0
e−2(t−τ )
0
e−3(t−τ )
⎡1
⎡
−1
with z(0) = M
⎡
0
⎣
x(0) = 1
0
17 t
2 e
−2t
z = ⎣ 34e
7 −3t
2e
⎡
+⎣
⎤
34
7 T
2 ]
. Thus,
⎡
⎤
⎡ 17 t ⎤
1
−(t−τ )
τ)e
(2
+
2
2 e
⎦ + t ⎣ (6 + 3τ)e−2(t−τ ) ⎦ dτ ⇒ z = ⎣ 34e−2t ⎦
0
7 −3t
(1 + 12 τ)e−3(t−τ )
2e
1
3
3 −t
2t + 2 − 2e
9
9 −2t
3
2t + 4 − 4e
5
5 −3t
1
6 t + 18 − 18 e
⎤
⎡
⎤
+ 32 + 7e−t
⎦ ⇒ z = ⎣ 3 t + 9 − 127 e−2t ⎦
2
4
4
5
29 −3t
1
6 t + 18 + 9 e
1
2t
c Pearson Education Limited 2011
0
0
⎤
⎦
82
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎤⎡ 1
⎤
3
−t
−2 1 −2
t
+
+
7e
2
2
−2t ⎦
giving x = Mz = ⎣ 1 0 −1 ⎦ ⎣ 32 t + 94 − 127
4 e
5
29 −3t
1
−1 0 3
6 t + 18 + 9 e
⎡
⎤
58 −3t
1
47
−2t
−14e−t + 127
e
−
e
+
t
−
4
9
6
36
−3t
⎦
7e−t − 29
+ 13 t + 11
⇒ x(t) = ⎣
9 e
9
29 −3t
2
−t
−7e + 3 e
−3
⎡
19(a)
Eigenvalues of the matrix given by
−1
−9 C1 −C2
1 − λ
3−λ
2
−1
−3 + λ 6 − λ
−9
0
1
1− λ
1
2
−1
= (3 − λ) 0 8 − λ −10
0
1
1 − λ
5 − λ
2
6−λ
0 = 3
1
1
= (3 − λ)(λ2 − 9λ + 18) = (3 − λ)(λ − 3)(λ − 6)
λ3 = 3⎡
so the eigenvalues are λ⎡1 = 6, λ2 = ⎤
⎤
0 0 1
2 2 −1
When λ = 3, A − 3I = ⎣ 3 3 −9 ⎦ ∼ ⎣ 1 0 0 ⎦ is of rank 2
0 0 0
1 1 −2
so there is only 3 − 2 = 1 corresponding eigenvectors.
The eigenvector corresponding to λ1 = 6 is readily determined as e1 = [3 2 1]T .
Likewise the single eigenvector corresponding to λ2 = 6 is determined as
e2 = [1 −1 0]T
The generalized eigenvector e∗2 determined by
(A − 2I)e∗2 = e2
or
3e∗21 + 2e∗22 − e∗23 = 1
3e∗21 + 3e∗22 − 9e∗23 = −1
e∗21 + e∗22 − 2e∗23 = 0
giving e∗2 = [
1 1 1 T
3 3 3]
.
For convenience, we can take the two eigenvectors corresponding to λ = 3 as
e2 = [3 − 3 0]T, e∗2 = [1 1 1]T
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎡
6
The corresponding Jordan canonical form being J = ⎣ 0
0
19(b)
83
⎤
0
1⎦
3
0
3
0
The generalised modal matrix is then
⎡
3 −3
⎣
M = 2 −3
1 0
⎡
5
⎣
AM= 3
1
⎡
3
M J = ⎣2
1
⎤
−1
−9 ⎦
1
⎤
3 1
−3 1 ⎦
0 1
2
6
1
⎡
3
⎣2
1
⎡
6
⎣0
0
−3
−3
0
0
3
0
⎤
1
1⎦
1
⎤
⎤
⎡
18
9 6
1
1 ⎦ = ⎣ 12 −9 0 ⎦
6
0 3
1
⎤
⎤
⎡
13
9 6
0
1 ⎦ = ⎣ 12 −9 0 ⎦
6
0 3
3
so A M = M J
⎡
19(c)
M−1
−3 −3
1 ⎣
−1
2
=−
9
3
3
so
⎤
⎡ 6t
e
6
Jt
⎦
⎣
0
−1 , e =
−15
0
⎤ ⎡ 6t
e
3
3 1
1
x(t) = − ⎣ 2 −3 1 ⎦ ⎣ 0
9
1
0 1
0
⎤
⎡ 6t
3t
9e − 9(1 + t)e
1 ⎣ 6t
=
6e + (3 + 9t)e3t ⎦
9
3e6t − 3e3t
⎡
20
0
e3t
0
0
e3t
0
⎤
0
te3t ⎦
e3t
⎤ ⎡
−3 −3
0
2
te3t ⎦ ⎣ −1
3t
3
3
e
⎤ ⎡ ⎤
0
6
−1 ⎦ ⎣ 1 ⎦
0
−15
Substituting x = eλt u, where u is a constant vector, in x = A x gives
λ2 u = A u or (A − λ2 I)u = 0
(1)
so that there is a non-trivial solution provided
| A − λ2 I |= 0
c Pearson Education Limited 2011
(2)
84
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
If λ21 , λ22 , . . . , λ2n are the solutions of (2) and u1 , u2 , . . . , un the corresponding
solutions of (1) define
M = [u1 u2 . . . un ] and S = diag (λ21 λ22 . . . λ2n )
Applying the transformation x = M q , q = [q1 q2 . . . qn ] gives
M q̈ = A M q
giving
q̈ = M−1 A M q provided u1 , u2 , . . . , un are linearly independent
so that q̈ = S q since M−1 A M = S
This represents n differential equations of the form
q̈i = λ2i qi , i = 1, 2, . . . , n
When λ2i < 0 this has the solution of the form
qi = Ci sin(ωi t + αi )
where Ci and αi are arbitrary constants and λi = jωi
The given differential equations may be written in the vector–matrix form
x1
ẍ1
−3
2
ẋ =
=
1 −2
ẍ2
x2
which is of the above form
ẍ = A x
0 =| A − λ2 I | gives (λ2 )2 + 5(λ2 ) + 4 = 0 or λ21 = −1, λ22 = −4.
Solving the corresponding equation
(A − λ2i I) ui = 0
we have that u1 = [1 1]T and u2 = [2 − 1]T . Thus, we take
−1
0
1 2
and S =
M=
0 −4
1 −1
The normal modes of the system are given by
q̈1
q1
−1
0
=
0 −4
q̈2
q2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
giving
85
q1 (t) = C1 sin(t + α1 ) ≡ γ1 sin t + β1 cos t
q2 (t) = C2 sin(2t + α2 ) ≡ γ2 sin 2t + β2 cos 2t
5
1 −1 −2
1
−1
3
=
Since x = M q we have that q(0) = M x(0) = −
2
− 13
3 −1 1
also q̇(0) = M−1 ẋ(0) so that q̇1 (0) = 2 and q̇2 (0) = 0
Using these initial conditions we can determine γ1 , β1 , γ2 and β2 to give
5
cos t + 2 sin t
3
1
q2 (t) = − cos 2t
3
q1 (t) =
The general displacements x1 (t) and x2 (t) are then given by x = M q so
2
5
cos t + 2 sin t − cos 2t
3
3
1
5
x2 = q1 − q2 = cos t + 2 sin t − cos 2t
3
3
x1 = q1 + 2q2 =
c Pearson Education Limited 2011
2
Numerical Solution of Ordinary
Differential Equations
Exercises 2.3.4
1
Euler’s method for the solution of the differential equation
dx
= f(t, x) is
dt
Xn+1 = Xn + hFn = Xn + hf(tn , Xn )
dx
Applying this to the equation
= − 21 xt with x(0) = 1 and a step size of h = 0.1
dt
yields
x0 = x(0) = 1
X1 = x0 + hf(t0 , x0 ) = x0 + h (− 12 x0 t0 )
= 1 − 0.1 ×
1
2
× 1 × 0 = 1.0000
X2 = X1 + hf(t1 , X1 ) = X1 + h (− 12 X1 t1 )
= 1.0000 − 0.1 ×
1
2
× 1.0000 × 0.1 = 0.9950
X3 = X2 + hf(t2 , X2 ) = X2 + h (− 12 X2 t2 )
= 0.9950 − 0.1 ×
1
2
× 0.9950 × 0.2 × 0.98505
Hence Euler’s method with step size h = 0.1 gives the estimate X(0.3) = 0.98505.
2
Euler’s method for the solution of the differential equation
Xn+1 = Xn + hFn = Xn + hf(tn , Xn )
c Pearson Education Limited 2011
dx
= f(t, x) is
dt
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Applying this to the equation
h = 0.025 yields
87
dx
= − 12 xt with x(1) = 0.1 and a step size of
dt
x0 = x(1) = 0.1
X1 = x0 + hf(t0 , x0 ) = x0 + h (− 12 x0 t0 )
= 0.1 − 0.025 ×
1
2
× 0.1 × 1 = 0.09875
X2 = X1 + hf(t1 , X1 ) = X1 + h (− 12 X1 t1 )
= 0.09875 − 0.025 ×
1
2
× 0.09875 × 1.025 = 0.09748
X3 = X2 + hf(t2 , X2 ) = X2 + h (− 12 X2 t2 )
1
= 0.09748 − 0.025 × × 0.09748 × 1.050 = 0.09621
2
X4 = X3 + hf(t3 , X3 ) = X3 + h (− 12 X3 t3 )
= 0.09621 − 0.025 ×
1
2
× 0.09621 × 1.075 = 0.09491
Hence Euler’s method with step size h = 0.1 gives the estimate X(1.1) = 0.09491.
3
Euler’s method for the solution of the differential equation
dx
= f(t, x) is
dt
Xn+1 = Xn + hFn = Xn + hf(tn , Xn )
Applying this to the equation
x
dx
=
with x(0.5) = 1 and a step of h = 0.1
dt
2(t + 1)
yields
x0 = x(0.5) = 1
1
x0
= 1 + 0.1
= 1.0333
2(t0 + 1)
2(0.5 + 1)
1.0333
X1
X2 = X1 + hf(t1 , X1 ) = X1 + h
= 1.0333 + 0.1
= 1.0656
2(t1 + 1)
2(0.6 + 1)
X1 = x0 + hf(t0 , x0 ) = x0 + h
(Note that tn = t0 + nh = 0.5 + 0.1n.) X3 , X4 and X5 may be computed in similar
fashion. It is usually easier to set out numerical solutions in a systematic tabular
form such as the following:
n
0
1
tn
0.5
0.6
Xn
1.0000
1.0333
f(tn , Xn )
0.3333
0.3229
Xn + hf(tn , Xn )
1.0333
1.0656
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
88
2
3
4
5
0.7
0.8
0.9
1.0
1.0656
1.0969
1.1274
1.1571
0.3134
0.3047
0.2967
1.0969
1.1274
1.1571
Hence Euler’s method with step size h = 0.1 gives the estimate X(1) = 1.1571.
4
Euler’s method for the solution of the differential equation
dx
= f(t, x) is
dt
Xn+1 = Xn + hFn = Xn + hf(tn , Xn )
Applying this to the equation
yields
4−t
dx
=
with x(0) = 1 and a step size of h = 0.05
dt
t+x
x0 = x(0.0) = 1
4 − t0
4−0
= 1.2000
= 1 + 0.05
t0 + x0
0+1
4 − t1
4 − 0.05
= 1.2000 + 0.05
X2 = X1 + hf(t1 , X1 ) = X1 + h
= 1.3580
t1 + X1
0.05 + 1.2000
X1 = x0 + hf(t0 , x0 ) = x0 + h
(Note that tn = t0 + nh = 0.0 + 0.05n.) X3 , X4 , . . . , X10 may be computed in
similar fashion.
It is usually easier to set out numerical solutions in a systematic tabular form such
as the following:
tn
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
Xn
1.0000
1.2000
1.3580
1.4917
1.6090
1.7140
1.8095
1.8972
1.9784
2.0541
2.1250
f(tn , Xn )
4.0000
3.1600
2.6749
2.3451
2.1006
1.9093
1.7540
1.6242
1.5136
1.4177
Xn + hf(tn , Xn )
1.2000
1.3580
1.4917
1.6090
1.7140
1.8095
1.8972
1.9784
2.0541
2.1250
Hence Euler’s method with step size h = 0.05 gives the estimate X(0.5) = 2.1250.
5
Figure 2.1 shows a suitable pseudocode program for computing the estimates
Xa (2) and Xb (2) . Figure 2.2 shows a Pascal implementation of the pseudocode
program.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
89
procedure deriv (t, x → f)
f ← x∗ t/(t∗ t + 2)
endprocedure
t start ← 1
x start ← 2
t end ← 2
write (vdu, "Enter step size")
read (keyboard, h)
write (printer, t start, x start )
t ← t start
x ← x start
repeat
deriv (t, x → f)
t←t+h
x ← x + h∗ f
write (printer, t,x)
until t >= t end
Figure 2.1: Pseudocode algorithm for Exercise 5
Using this program the results Xa (2) = 2.811489 and Xb (2) = 2.819944 were
obtained. Using the method described in Section 2.3.6, the error in Xb (2) will
be approximately equal to Xb (2) − Xa (2) = −0.008455 and so the best estimate
of X(2) is 2.819944 + 0.008455 = 2.828399. The desired error bound is 0.1% of
this value, 0.0028 approximately. Since Euler’s method is a first-order method, the
error in the estimate of X(2) varies like h ; so, to achieve an error of 0.0028, a step
size of no more than (0.0028/0.008455) × 0.05 = 0.0166 is required. We will choose
a sensible step size which is less than this, say h = 0.0125. This yields an estimate
X(2) = 2.826304.
The exact solution of the differential equation may be obtained by separation:
xt
dx
t dt
dx
2
1
= 2
⇒
=
⇒
ln
x
=
ln(t
+
2)
+
C
⇒
x
=
±D
t2 + 2
2
dt
t +2
x
t2 + 2
√
t2 + 2
x(1) = 2 ⇒ 2 = ± 3D ⇒ x = 2
3
√
Hence x(2) = 2 2 = 2.828427 and the true errors in Xa (2), Xb (2) and the final
estimate of X(2) are 0.016938, 0.008483 and 0.002123 respectively. The estimate,
X(2), derived using the step size h = 0.0125 is comfortably within the 0.1% error
requirement.
c Pearson Education Limited 2011
90
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
var t start, x start, t end, h,x,t,f:real;
procedure deriv (t,x:real;var f:real);
begin
f := x∗ t/(t∗ t + 2)
end;
begin
t start := 1;
x start := 2;
t end := 2;
write ( Enter step size ==> );
readln (h);
writeln (t start : 5 : 2, x start : 10 : 6) ;
t := t start ;
x := x start ;
repeat
deriv(t,x,f);
t : = t + h;
x : = x + h∗ f ;
writeln (t:5:2, x:10:6);
until t > = t end ;
end.
Figure 2.2: Pascal program for Exercise 5
6
The programs shown in Figures 2.1 and 2.2 may readily be modified to solve
this problem. Estimates Xa (2) = 1.573065 and Xb (2) = 1.558541 should be
obtained. Using the method described in Section 2.3.6, the error in Xb (2) will
be approximately equal to Xb (2) − Xa (2) = −0.014524 and so the best estimate
of X(2) is 1.558541 − 0.014524 = 1.544017. The desired error bound is 0.2% of
this value, 0.0031 approximately. Since Euler’s method is a first-order method, the
error in the estimate of X(2) varies like h so, to achieve an error of 0.0031, a step
size of no more than (0.0031/0.014524) × 0.05 = 0.0107 is required. We will choose
a sensible step size which is less than this, say h = 0.01. This yields an estimate
X(2) = 1.547462.
The exact solution of the differential equation may be obtained by separation:
dt
1
dx
=
⇒ xdx =
⇒ 12 x2 = ln t + C ⇒ x = ± 2(ln t + C)
dt
xt
t
√
x(1) = 1 ⇒ 1 = 2C ⇒ x(t) = 2 ln t + 1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Hence x(2) =
√
91
2 ln 2 + 1 = 1.544764 and the true errors in Xa (2), Xb (2) and
the final estimate of X(2) are −0.028301, −0.013777 and − 0.002698 respectively.
The estimate, X(2), derived using the step size h = 0.01 is comfortably within the
0.2% error requirement.
7
The programs shown in Figures 2.1 and 2.2 may readily be modified to solve
this problem. Estimates Xa (1.5) = 2.241257 and Xb (1.5) = 2.206232 should be
obtained. Using the method described in Section 2.3.6, the error in Xb (1.5) will be
approximately equal to Xb (1.5) − Xa (1.5) = −0.035025 and so the best estimate
of X(1.5) is 2.206232 − 0.035025 = 2.171207. The desired error bound is 0.25%
of this value, 0.0054 approximately. Since Euler’s method is a first-order method,
the error in the estimate of X(1.5) varies like h; so, to achieve an error of 0.0054,
a step size of no more than (0.0054/0.035025) × 0.025 = 0.0039 is required. If we
choose h = 0.04, this yields an estimate X(1.5) = 2.183610.
The exact solution of the differential equation may be obtained by separation:
1
dx
=
⇒ ln xdx = dt ⇒ x ln x − x = t + C
ct
ln x
x(1) = 1.2 ⇒ 1.2 ln 1.2 − 1.2 = 1 + C ⇒
C = −1.981214 ⇒ x ln x − x = t − 1.981214
Hence, by any non-linear equation solving method (e.g. Newton–Raphson), we
may obtain x(1.5) = 2.179817 and the true errors in Xa (1.5), Xb (1.5) and the
final estimate of X(1.5) are 0.061440, 0.026415 and 0.003793 respectively. The
estimate, X(1.5), derived using the step size h = 0.04 is comfortably within the
0.25% error requirement.
c Pearson Education Limited 2011
92
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 2.3.9
8
The starting process, using the second-order predictor–corrector method, is
X̂1 = x0 + hf(t0 , x0 )
X1 = x0 + 12 h f(t0 , x0 ) + f(t1 , X̂1 )
and the second-order Adams–Bashforth method is
Xn+1 = Xn + 12 h (3f(tn , Xn ) − f(tn−1 , Xn−1 ))
8(a)
Applying this method to the problem
h = 0.1, we have
dx
= x2 sin t − x, x(0) = 0.2 with
dt
X̂1 = x0 + hf(t0 , x0 ) = 0.2 + 0.1 × (0.22 sin 0 − 0.2) = 0.1800
X1 = x0 + 12 h f(t0 , x0 ) + f(t1 , X̂1 )
= 0.2 + 12 0.1 × (0.22 sin 0 − 0.2 + 0.182 sin 0.1 − 0.18) = 0.1812
X2 = X1 + 12 h (3f(t1 , X1 ) − f(t0 , x0 ))
= 0.1812 + 12 0.1 × 3(0.18122 sin 0.1 − 0.1812) − (0.22 sin 0 − 0.2) = 0.1645
X3 , X4 and X5 are obtained as X2 . The computation is most efficiently set out
as a table.
n
0
1
2
3
4
5
tn
0.0
0.1
0.2
0.3
0.4
0.5
Xn
0.2000
0.1812
0.1645
0.1495
0.1360
0.1238
f(tn , Xn )
−0.2000
−0.1779
−0.1591
−0.1429
−0.1288
− f(tn−1 , Xn−1 ))
(use predictor–corrector)
−0.016685
−0.014970
−0.013480
−0.012175
1
2 h(3f(tn , Xn )
Hence X(0.5) = 0.1238.
c Pearson Education Limited 2011
Xn+1
0.1812
0.1645
0.1495
0.1360
0.1238
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
8(b)
Applying this method to the problem
0.1, we have
93
dx
= x2 etx , x(0.5) = 0.5 with h =
dt
X̂1 = x0 + hf(t0 , x0 ) = 0.5 + 0.1 × 0.52 e0.5×0.5 = 0.5321
X1 = x0 + 12 h f(t0 , x0 ) + f(t1 , X̂1 )
= 0.5 + 12 0.1 × (0.52 e0.5×0.5 + 0.53212 e0.6×0.5321 ) = 0.5355
X2 = X1 + 12 h (3f(t1 , X1 ) − f(t0 , x0 ))
= 0.5355 + 12 0.1 × 3 × 0.53552 e0.6×0.5355 − 0.52 e0.5×0.5 = 0.5788
X3 , X4 , X5 , X6 and X7 are obtained as X2 . The computation is most efficiently
set out as a table.
n
0
1
2
3
4
5
6
7
tn
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
Xn
0.5000
0.5355
0.5788
0.6344
0.7095
0.8191
0.9998
1.3740
f(tn , Xn )
0.3210
0.3955
0.5024
0.6685
0.9534
1.5221
3.0021
− f(tn−1 , Xn−1 ))
(use predictor–corrector)
0.043275
0.055585
0.075155
0.109585
0.180645
0.374210
1
2 h(3f(tn , Xn )
Xn+1
0.5355
0.5788
0.6344
0.7095
0.8191
0.9998
1.3740
Hence X(1.2) = 1.3740.
9
The starting process, using the second-order predictor–corrector method, is
X̂1 = x0 + hf(t0 , x0 )
X1 = x0 + 12 h f(t0 , x0 ) + f(t1 , X̂1 )
X̂2 = X1 + hf(t1 , X1 )
1
X2 = X1 + 2 h f(t1 , X1 ) + f(t2 , X̂2 )
and the third-order Adams–Bashforth method is
Xn+1 = Xn +
1
12
h (23f(tn , Xn ) − 16f(tn−1 , Xn−1 ) + 5f(tn−2 , Xn−2 ))
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
94
Applying this method to the problem
have
dx 2
= x + 2t, x(0) = 1 with h = 0.1, we
dt
X̂1 = x0 + hf(t0 , x0 ) = 1.0 + 0.1 × 12 + 2 × 0 = 1.100
X1 = x0 + 12 h f(t0 , x0 ) + f(t1 , X̂1 )
= 1.0 + 12 0.1 ×
12 + 2 × 0 + 1.12 + 2 × 0.1 = 1.1094
X̂2 = X1 + hf(t1 , X1 ) = 1.1094 + 0.1 × 1.10942 + 2 × 0.1 = 1.2290
X2 = X1 + 12 h f(t1 , X1 ) + f(t2 , X̂2 )
= 1.1094 + 12 0.1 ×
1.10942 + 2 × 0.1 + 1.22902 + 2 × 0.2 = 1.2383
h (23f(t2 , X2 ) − 16f(t1 , X1 ) + 5f(t0 , x0 ))
1
= 1.2383 + 12 0.1 × 23 1.23832 + 2 × 0.2 − 16 1.10942 + 2 × 0.1
+ 5 1.02 + 2 × 0 = 1.3870
X3 = X2 +
1
12
X4 and X5 are obtained as X3 . The computation is most efficiently set out as a
table.
n
tn
Xn
f(tn , Xn )
h(23f(tn , Xn ) − 16f(tn−1 , Xn−1 )
Xn+1
1.1094
0
0.0
1.0000
1.0000
+ 5f(tn−2 , Xn−2 ))/12
(use predictor–corrector)
0
0.1
1.1094
1.1961
(use predictor–corrector)
1.2383
2
3
0.2
0.3
1.2383
1.3870
1.3905
1.5886
0.1487
0.1689
1.3870
1.5559
4
0.4
1.5559
1.7947
0.1901
1.7460
5
0.5
1.7460
Hence X(0.5) = 1.7460.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
10
95
The second-order predictor–corrector method is
X̂n+1 = Xn + hf(tn , Xn )
Xn+1 = Xn + 12 h(f(tn , Xn ) + f(tn+1 , X̂n+1 ))
10(a)
Applying this method to the problem
h = 0.05, we have
dx
= (2t + x) sin 2t, x(0) = 0.5 with
dt
X̂1 = x0 + hf(t0 , x0 ) = 0.5 + 0.05 × (2 × 0 + 0.5) sin 0 = 0.5
X1 = x0 + 12 h(f(t0 , x0 ) + f(t1 , X1 ))
= 0.5 + 12 0.05 × ((2 × 0 + 0.5) sin 0 + (2 × 0.05 + 0.5) sin(2 × 0.05)) = 0.5015
X2 to X10 are obtained as X1 . The computation is most efficiently set out as a
table.
n
0
1
2
3
4
5
6
7
8
9
10
tn
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
Xn
0.5000
0.5015
0.5065
0.5160
0.5311
0.5527
0.5820
0.6198
0.6673
0.7254
0.7948
f(tn , Xn )
0.0000
0.0150
0.0497
0.1034
0.1752
0.2637
0.3670
0.4832
0.6098
0.7442
X̂n+1
0.5000
0.5045
0.5135
0.5281
0.5492
0.5780
0.6153
0.6623
0.7199
0.7890
f(tn+1 , X̂n+1 )
0.0599
0.1400
0.2404
0.3614
0.5030
0.6651
0.8474
1.0490
1.2689
1.5054
Xn+1
0.5015
0.5065
0.5160
0.5311
0.5527
0.5820
0.6198
0.6673
0.7254
0.7948
Hence X(0.5) = 0.7948.
10(b)
Applying this method to the problem
1+x
dx
=−
, x(0) = −2 with
dt
sin(t + 1)
h = 0.1, we have
1−2
= −1.8812
X̂1 = x0 + hf(t0 , x0 ) = −2 + 0.1 × −
sin(0 + 1)
X1 = x0 + 12 h f(t0 , x0 ) + f(t1 , X̂1 )
= −2 + 12 0.1 × −
1 − 1.8812
1−2
−
sin(0 + 1) sin(0.1 + 1)
= −1.8911
X2 to X10 are obtained as X1 . The computation is most efficiently set out as a
table.
c Pearson Education Limited 2011
96
n
0
1
2
3
4
5
6
7
8
9
10
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
tn
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Xn
−2.0000
−1.8911
−1.7987
−1.7189
−1.6489
−1.5867
−1.5309
−1.4803
−1.4339
−1.3910
−1.3511
f(tn , Xn )
−1.3072
−1.2343
−1.1802
−1.1416
−1.1162
−1.1028
−1.1005
−1.1092
−1.1295
−1.1624
X̂n+1
−1.8812
−1.7912
−1.7130
−1.6443
−1.5830
−1.5279
−1.4778
−1.4318
−1.3893
1.3497
f(tn+1 , X̂n+1 )
0.9887
0.8488
0.7400
0.6538
0.5845
0.5281
0.4818
0.4434
0.4114
0.3846
Hence X(1.0) = −1.3511.
11
Taylor’s theorem states that
df
h3 d 3 f
h4 d 4 f
h2 d 2 f
(t)
+
(t)
+
(t) + K
(t) +
dt
2! dt2
3! dt3
4! dt4
dx
dx
(t − h) and
(t − 2h) yields
Applying this to
dt
dt
f(t + h) = f(t) + h
dx
dx
d2 x
h2 d 3 x
(t − h) =
(t) − h 2 (t) +
(t) + O(h3 )
dt
dt
dt
2! dt3
dx
dx
d2 x
4h2 d3 x
(t − 2h) =
(t) − 2h 2 (t) +
(t) + O(h3 )
dt
dt
dt
2! dt3
Multiplying the first equation by 2 and subtracting the second yields
dx
dx
(t − h) −
(t − 2h) =
dt
dt
d3 x
dx
that is, h2 3 (t) = −2 (t − h) +
dt
dt
2
d3 x
dx
(t) − h2 3 (t) + O(h3 )
dt
dt
dx
dx
(t − 2h) +
(t) + O(h3 )
dt
dt
Multiplying the first equation by 4 and subtracting the second yields
dx
dx
d2 x
dx
(t − h) −
(t − 2h) = 3 (t) − 2h 2 (t) + O(h3 )
dt
dt
dt
dt
dx
dx
dx
d2 x
(t − 2h) + 3 (t) + O(h3 )
that is, 2h 2 (t) = −4 (t − h) +
dt
dt
dt
dt
4
Now Taylor’s theorem yields
h2 d 2 x h3 d 3 x
dx
+
(t) + O(h3 )
x(t + h) = x(t) + h (t) +
2
3
dt
2! dt
3! dt
c Pearson Education Limited 2011
Xn+1
−1.8911
−1.7987
−1.7189
−1.6489
−1.5867
−1.5309
−1.4803
−1.4339
−1.3910
1.3511
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Hence, substituting for h
3
d2 x
2d x
(t)
and
h
(t) yields
dt2
dt3
h
dx
dx
dx
dx
(t) +
−4 (t − h) +
(t − 2h) + 3 (t) + O(h3 )
dt
4
dt
dt
dt
dx
dx
dx
h
−2 (t − h) +
(t − 2h) +
(t) + O(h3 ) + O(h3 )
+
6
dt
dt
dt
dx
dx
dx
h
23 (t) − 16 (t − h) + 5 (t − 2h) + O(h3 )
= x(t) +
12
dt
dt
dt
x(t + h) = x(t) + h
12
Taylor’s theorem states that
f(t + h) = f(t) + h
Applying this to
h3 d 3 f
h4 d 4 f
df
h2 d 2 f
(t)
+
(t)
+
(t) + K
(t) +
dt
2! dt2
3! dt3
4! dt4
dx
dx
(t + h) and
(t − h) yields
dt
dt
dx
(t + h) =
dt
dx
(t − h) =
dt
dx
d2 x
(t) + h 2 (t) +
dt
dt
dx
d2 x
(t) − h 2 (t) +
dt
dt
h2 d 3 x
(t) + O(h3 )
2! dt3
h2 d 3 x
(t) + O(h3 )
2! dt3
Summing the two equations yields
d3 x
dx
dx
dx
(t + h) +
(t − h) = 2 (t) + h2 3 (t) + O(h3 )
dt
dt
dt
dt
3
d
x
dx
dx
dx
(t + h) +
(t − h) − 2 (t) + O(h3 )
that is, h2 3 (t) =
dt
dt
dt
dt
Subtracting the second from the first yields
dx
d2 x
dx
(t + h) −
(t − h) = 2h 2 (t) + O(h3 )
dt
dt
dt
2
dx
d x
dx
that is, 2h 2 (t) =
(t + h) −
(t − h) + O(h3 )
dt
dt
dt
Now Taylor’s theorem gives
h2 d 2 x h3 d 3 x
dx
+
(t) + O(h4 )
x(t + h) = x(t) + h (t) +
2
3
dt
2! dt
3! dt
c Pearson Education Limited 2011
97
98
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Hence, substituting for h
3
d2 x
2d x
(t)
and
h
(t), we have
dt2
dt3
dx
h dx
dx
(t) +
(t + h) +
(t − h) + O(h3 )
dt
4 dt
dt
dx
dx
h dx
(t + h) +
(t − h) − 2 (t) + O(h3 ) + O(h4 )
+
6 dt
dt
dt
dx
dx
dx
h
5 (t + h) + 8 (t) −
(t − h) + O(h4 )
= x(t) +
12
dt
dt
dt
dx
h
dx
dx
5
+ O(h4 )
= xn +
+8
−
12
dtn+1
dtn
dtn−1
x(t + h) = x(t) + h
that is, xn+1
13
Taylor’s theorem states that
f(t + h) = f(t) + h
h2 d 2 f
df
h3 d 3 f
h4 d 4 f
(t) +
(t)
+
(t)
+
(t) + K
dt
2! dt2
3! dt3
4! dt4
Applying this to x(t − h) and
dx
(t − h) yields
dt
h2 d 2 x
dx
h3 d 3 x
(t) +
(t)
−
(t) + O(h4 )
dt
2! dt2
3! dt3
dx
dx
d2 x
h2 d 3 x
(t − h) =
(t) − h 2 (t) +
(t) + O(h3 )
dt
dt
dt
2! dt3
x(t − h) = x(t) − h
Multiplying the first equation by 2, the second equation by h and adding yields
dx
dx
h3 d3 x
(t) + O(h4 )
(t − h) = 2x(t) − h (t) +
dt
dt
6 dt3
dx
dx
h3 d 3 x
(t) = 2x(t − h) + h (t − h) − 2x(t) + h (t) + O(h4 )
that is,
3
6 dt
dt
dt
2x(t − h) + h
Multiplying the first equation by 3, the second equation by h and adding yields
dx
h2 d2 x
dx
(t − h) = 3x(t) − 2h (t) +
(t) + O(h4 )
dt
dt
2 dt2
dx
h2 d 2 x
dx
(t) = 3x(t − h) + h (t − h) − 3x(t) + 2h (t) + O(h4 )
that is,
2
2 dt
dt
dt
3x(t − h) + h
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
99
Now Taylor’s theorem gives
x(t + h) = x(t) + h
Hence, substituting for h2
dx
h2 d 2 x h3 d 3 x
+
(t) + O(h4 )
(t) +
dt
2! dt2
3! dt3
3
d2 x
3d x
(t)
and
h
(t), we have
dt2
dt3
dx
dx
dx
(t) + 3x(t − h) + h (t − h) − 3x(t) + 2h (t) + O(h4 )
dt
dt
dt
dx
dx
+ 2x(t − h) + h (t − h) − 2x(t) + h (t) + O(h4 ) + O(h4 )
dt
dt
dx
dx
= −4x(t) + 5x(t − h) + 4h (t) + 2h (t − h) + O(h4 )
dt
dt
x(t + h) = x(t) + h
that is, xn+1 = −4xn + 5xn−1 + 2h 2
dx
dx
+
dtn
dtn−1
+ O(h4 )
This gives rise to the approximate scheme Xn+1 = −4Xn +5Xn−1 +2h(2Fn +Fn−1 )
which may equally be written as Xn+1 = 5[Xn−1 +2hFn ]−4[Xn + 12 h(3Fn −Fn−1 )];
in other words, this scheme gives an approximation for Xn+1 which is 5 times the
central difference approximation minus 4 times the second-order Adams–Bashforth
approximation. Because of the inclusion of the central difference approximation,
this scheme will be unstable whenever the central difference scheme is.
14
The predictor–corrector scheme specified is
X̂n+1 = Xn + 12 h(3Fn − Fn−1 ) = Xn + 12 h(3f(tn , Xn ) − f(tn−1 , Xn−1 ))
Xn+1 = Xn +
1
12
h(5Fn+1 + 8Fn − Fn−1 ) = Xn +
1
12
h(5f(tn+1 , X̂n+1 )
+ 8f(tn , Xn ) − f(tn−1 , Xn−1 ))
This is obviously not self-starting and requires that one initial step is taken using a
dx
self-starting method. For the problem
= x2 + t2 , x(0.3) = 0.1 with a step size
dt
c Pearson Education Limited 2011
100
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
of h = 0.05, using the fourth-order Runge–Kutta method for the initial step we
obtain
c1 = hf(t0 , x0 ) = h(x20 + t20 ) = 0.05 × (0.12 + 0.32 ) = 0.0050
c2 = hf(t0 + 12 h, x0 + 12 c1 ) = 0.05 × (0.1 + 12 0.0050)2 + (0.3 + 12 0.05)2
= 0.0058
c3 = hf(t0 + 12 h, x0 + 12 c2 ) = 0.05 × (0.1 + 12 0.0058)2 + (0.3 + 12 0.05)2
= 0.0058
c4 = hf(t0 + h, x0 + c3 ) = 0.05 × (0.1 + 0.0058)2 + (0.3 + 0.05)2
= 0.0067
X1 = x0 + 16 (c1 + 2c2 + 2c3 + c4 ) = 0.1 + 16 (0.0050 + 2 × 0.0058
+ 2 × 0.0058 + 0.0067) = 0.1058
Now we can say
X̂2 = X1 + 12 h (3f(t1 , X1 ) − f(t0 , x0 ))
= 0.1058 + 12 0.05 × 3(0.10582 + 0.352 ) − (0.12 + 0.32 ) = 0.1133
1
X2 = X1 + 12
h 5f(t2 , X̂2 ) + 8f(t1 , X1 ) − f(t0 , x0 )
1
= 0.1058 + 12
0.05 × 5(0.11332 + 0.42 ) + 8(0.10582 + 0.352 ) − (0.12 + 0.32 )
= 0.1134
Computing X3 and X4 in a similar manner and setting the computation out in
tabular fashion we obtain:
n
tn
Xn
f(tn , Xn )
X̂n+1
f(tn+1 , X̂n+1 )
0
0.30
0.1000
0.1000
1
0.35
0.1058
0.1337
0.1133
0.1728
0.1134
2
0.40
0.1134
0.1729
0.1230
0.2176
0.1231
3
0.45
0.1231
0.2177
0.1351
0.2683
0.1352
4
0.50
0.1352
(done by Runge–Kutta method)
Hence X(0.5) = 0.1352.
c Pearson Education Limited 2011
Xn+1
0.1058
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
101
15
The fourth-order Runge–Kutta method for the solution of the differential
dx
= f(t, x) using a stepsize of h is given by
equation
dt
c1 = hf(tn , Xn )
c2 = hf(tn + 12 h, Xn + 12 c1 )
c3 = hf(tn + 12 h, Xn + 12 c2 )
c4 = hf(tn + h, Xn + c3 )
Xn+1 = Xn + 16 (c1 + 2c2 + 2c3 + c4 )
15(a)
To solve the equation
h = 0.15, we write
dx
= x + t + xt, x(0) = 1 , using a stepsize of
dt
c1 = hf(t0 , x0 ) = 0.15 × (1 + 0 + 1 × 0) = 0.1500
c2 = hf(t0 + 12 h, x0 + 12 c1 )
= 0.15 × ((1 + 12 0.1500) + (0 + 12 0.15) + (1 + 12 0.1500) × (0 + 12 0.15))
= 0.1846
c3 = hf(t0 + 12 h, x0 + 12 c2 )
= 0.15 × ((1 + 12 0.1846) + (0 + 12 0.15) + (1 + 12 0.1846) × (0 + 12 0.15))
= 0.1874
c4 = hf(t0 + h, x0 + c3 ) = 0.15 × ((1 + 0.1874) + (0 + 0.15) + (1 + 0.1874)
× (0 + 0.15)) = 0.2273
X1 = x0 + 16 (c1 + 2c2 + 2c3 + c4 ) = 1 + 16 (0.1500 + 2 × 0.1846
+ 2 × 0.1874 + 0.2273) = 1.1869
X2 , X3 , X4 and X5 may be computed in a similar manner.
computation out in tabular fashion we obtain:
c Pearson Education Limited 2011
Setting the
102
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
n
0
1
2
3
4
5
tn
0.00
0.15
0.30
0.45
0.60
0.75
Xn
1.0000
1.1869
1.4630
1.8603
2.4266
3.2345
c1
0.1500
0.2272
0.3303
0.4721
0.6724
c2
0.1846
0.2727
0.3921
0.5583
0.7954
c3
0.1874
0.2769
0.3984
0.5681
0.8109
c4
0.2273
0.3304
0.4724
0.6728
0.9623
Xn+1
1.1869
1.4630
1.8603
2.4266
3.2345
Hence X(0.75) = 3.2345.
15(b)
To solve the equation
h = 0.1 we write
c1 = hf(t0 , x0 ) = 0.1 ×
dx
1
=
, x(1) = 2 , using a stepsize of
dt
x+t
1
= 0.0333
2+1
c2 = hf(t0 + 12 h, x0 + 12 c1 ) = 0.1 ×
c3 = hf(t0 + 12 h, x0 + 12 c2 ) = 0.1 ×
c4 = hf(t0 + h, x0 + c3 ) = 0.1 ×
1
(2 +
1
2 0.0333)
(2 +
1
2 0.0326)
+ (1 + 12 0.1)
1
+ (1 + 12 0.1)
= 0.0326
= 0.0326
1
= 0.0319
(2 + 0.0326) + (1 + 0.1)
X1 = x0 + 16 (c1 + 2c2 + 2c3 + c4 ) = 2 + 16 (0.0333 + 2 × 0.0326
+ 2 × 0.0326 + 0.0319) = 2.0326
X2 , X3 , . . . , X10 may be computed in a similar manner. Setting the computation
out in tabular fashion we obtain:
n
0
1
2
3
4
5
6
7
8
9
10
tn
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
Xn
2.0000
2.0326
2.0639
2.0939
2.1228
2.1507
2.1777
2.2037
2.2289
2.2534
2.2771
c1
0.0333
0.0319
0.0306
0.0295
0.0284
0.0274
0.0265
0.0256
0.0248
0.0241
c2
0.0326
0.0313
0.0300
0.0289
0.0279
0.0269
0.0260
0.0252
0.0244
0.0237
c3
0.0326
0.0313
0.0300
0.0289
0.0279
0.0269
0.0260
0.0252
0.0244
0.0237
Hence X(2) = 2.2771.
c Pearson Education Limited 2011
c4
0.0319
0.0306
0.0295
0.0284
0.0274
0.0265
0.0256
0.0248
0.0241
0.0234
Xn+1
2.0326
2.0639
2.0939
2.1228
2.1507
2.1777
2.2037
2.2289
2.2534
2.2771
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
16
In this exercise the differential equation problem
is solved using a variety of methods.
16(a)
103
3
dx
= x2 + t 2 , x(0) = −1 ,
dt
Figure 2.3 shows a pseudocode algorithm for solving the equation
using the second-order Adams–Bashforth method with a second-order predictor–
corrector starting step and Figure 2.4 shows a Pascal program derived from it.
procedure deriv (t, x → f)
f ← x∗ x+sqrt (t)∗ t
endprocedure
t start ← 0
x start ← −1
t end ← 2
write (vdu, "Enter step size")
read (keyboard, h)
write (printer, t start , x start)
deriv (t start, x start → f)
x hat ← x start + h∗ f
deriv(t start + h, x hat → f hat)
t ← t start + h
x ← x start + h∗ (f + f hat)/2
write (printer, t,x)
f n minus one ← f
repeat
deriv (t, x → f)
t←t+h
x ← x + h∗ (3∗ f − f n minus one)/2
f n minus one ← f
write (printer, t,x)
until t >= t end
Figure 2.3: Pseudocode algorithm for Exercise 16(a)
c Pearson Education Limited 2011
104
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
var x start, x hat, x,t start, t end; t:real;
h,f,f hat, f n minus one:real;
procedure deriv (t,x:real;var f:real);
begin
f := x*x+sqrt(t)*t
end;
begin
t start := 0;
x start := -1;
t end := 2;
write( Enter step size ==> );
readln(h);
writeln(t start:10:3, x start:10:6);
deriv(t start, x start, f);
x hat := x start + h*f;
deriv(t start + h, x hat, f hat);
t :=t start + h;
x :=x start + h *(f+f hat)/2;
writeln(t:10:3, x:10:6);
f n minus one := f;
repeat
deriv(t,x,f);
t := t + h;
x := x + h*(3*f-f n minus one)/2;
f n minus one := f;
writeln(t:10:3, x:10:6);
until t >= t end ;
end.
Figure 2.4: Pascal program for Exercise 16(a)
Using this program with h = 0.2 gives X(2) = 2.242408 and, with h = 0.1, X(2) =
2.613104. The method of Richardson extrapolation given in Section 2.3.6 gives the
estimated error in the second of these as (2.613104 − 2.242408)/3 = 0.123565.
For 3 decimal place accuracy in the final estimate we need error ≤ 0.0005; in
other words, the error must be reduced by a factor of 0.123565/0.0005 = 247.13.
Since Adams–Bashforth is a second-order method, the required step length will
√
be 0.1/ 247.13 = 0.0064. Rounding this down to a suitable size suggests that
a step size of h = 0.005 will give a solution accurate to 3 decimal places. In
fact the program of Figure 2.4 yields, with h = 0.005, X(2) = 2.897195. With
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
105
h = 0.0025 it gives X(2) = 2.898175. Richardson extrapolation predicts the error
in the h = 0.0025 solution as 0.000327 and therefore that in the h = 0.005 as
0.001308. The required accuracy was therefore not achieved using h = 0.005 but
was achieved with h = 0.0025.
16(b) Figure 2.5 shows a pseudocode algorithm for solving the equation using the
second-order predictor–corrector method and Figure 2.6 shows a Pascal program
derived from it.
procedure deriv (t, x → f)
f ← x∗ x+sqrt (t)∗ t
endprocedure
t start ← 0
x start ← −1
t end ← 2
write (vdu, "Enter step size")
read (keyboard, h)
write (printer, t start , x start)
t ← t start
x ← x start
repeat
deriv (t, x → f)
x hat ← x + h∗ f
derive (t + h, x hat → f hat)
t←t+h
x ← x + h∗ (f + f hat)/2
write (printer, t,x)
until t >= t end
Figure 2.5: Pseudcode algorithm for Exercise 2.16(b)
Using this program with h = 0.2 gives X(2) = 2.788158 and, with h = 0.1, X(2) =
2.863456. The method of Richardson extrapolation given in Section 2.3.6 gives the
estimated error in the second of these as (2.863456 − 2.788158)/3 = 0.025099. For
3 decimal place accuracy in the final estimate we need error ≤ 0.0005; in other
words, the error must be reduced by a factor of 0.025099/0.0005 = 50.20. Since
the second-order predictor–corrector method is being used, the required step size
√
will be 0.1/ 50.20 = 0.014. Rounding this down to a suitable size suggests that
a step size of h = 0.0125 will give a solution accurate to 3 decimal places. In
fact the program of Figure 2.6 yields, with h = 0.0125, X(2) = 2.897876. With
c Pearson Education Limited 2011
106
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
h = 0.00625 it gives X(2) = 2.898349. Richardson extrapolation predicts the error
in the h = 0.00625 solution as 0.000158 and therefore that in the h = 0.0125 as
0.000632. The required accuracy was therefore not quite achieved using h = 0.0125
but was achieved with h = 0.00625.
var x start,x hat,x,t start,t end, t:real;
h,f,f hat:real;
procedure deriv(t,x:real;var f:real);
begin
f := x*x+sqrt(t)*t
end;
begin
t start := 0;
x start := -1;
t end := 2;
write(‘Enter step size ==> ’);
readln(h);
writeln(t start:10:3,x start:10:6);
t := t start;
x := x start;
repeat
deriv(t,x,f);
x hat := x + h*f;
deriv(t + h,x hat,f hat);
t := t + h;
x := x + h*(f + f hat)/2;
writeln (t:10:3,x:10:6);
until t >= t end;
end.
Figure 2.6: Pascal program for Exercise 16(b)
16(c)
Figure 2.7 shows a pseudocode algorithm for solving the equation using
the fourth-order Runge–Kutta method and Figure 2.8 shows a Pascal program
derived from it. Using this program with h = 0.4 gives X(2) = 2.884046 and,
with h = 0.2, X(2) = 2.897402. The method of Richardson extrapolation given
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
107
in Section 2.3.6 gives the estimated error in the second of these as (2.897402 −
2.884046)/15 = 0.000890. For 5 decimal place accuracy in the final estimate we
need error ≤ 0.000005; in other words, the error must be reduced by a factor
of 0.000890/0.000005 = 178. Since Runge–Kutta is a fourth-order method the
required step size will be 0.2/(178)1/4 = 0.0547. Rounding this down to a suitable
size suggests that a step size of h = 0.05 will give a solution accurate to 5 decimal
places. In fact the program of Figure 2.8 yields, with h = 0.05, X(2) = 2.89850975.
With h = 0.025 it gives X(2) = 2.89850824. Richardson extrapolation predicts
the error
procedure deriv (t,x → f)
f → x∗ x + sqrt(t)∗ t
endprocedure
t start ← 0
x start ← −1
t end ← 2
write(vdu, "Enter step size")
read(keyboard, h)
write (printer,t start,x start)
t ← t start
x ← x start
repeat
deriv(t, x, → f)
c1 ← h∗ f
deriv(t + h/2, x + c1/2 → f)
c2 ← h∗ f
deriv(t + h/2, x + c2/2 → f)
c3 ← h∗ f
deriv(t + h, x + c3 → f)
c4 ← h∗ f
t←t+h
x ← x + (c1 + 2∗ c2 + 2∗ c3 + c4)/6
write(printer,t,x)
until t >= t end
Figure 2.7: Pseudocode algorithm for Exercise 16(c)
c Pearson Education Limited 2011
108
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
{Program for exercise 2.16c}
var x start,x,t start,t end,t:real;
h,f,c1,c2,c3,c4:real;
procedure deriv (t,x:real;var f:real);
begin
f := x ∗ x+sqrt(t) ∗ t
end;
begin
t start := 0;
x start := -1;
t end := 2;
write(‘Enter step size ==> ’);
readln(h);
writeln(t start:10:3,x start:10:6);
t := t start;
x := x start;
repeat
deriv(t,x,f);
c1 := h*f;
deriv(t + h/2,x + c1/2, f);
c2 := h*f;
deriv(t + h/2,x + c2/2, f);
c3 := h*f;
deriv(t + h,x + c3, f);
c4 := h*f;
t := t + h;
x := x + (c1 + 2*c2 + 2*c3 + c4)/6;
writeln(t:10:3,x:10:6);
until t >= t end;
end.
Figure 2.8: Pascal program for Exercise 16(c)
in the h = 0.025 solution as 0.000000101 and therefore that in the h = 0.05 as
0.00000161. The required accuracy was therefore achieved using h = 0.05.
17
The pseudocode algorithm shown in Figure 2.7 and the Pascal program in
Figure 2.8 may easily be modified to solve this problem. With a step size of h = 0.5
the estimate X(3) = 1.466489 is obtained, whilst with a step size of h = 0.25, the
estimate is X(3) = 1.466476. Richardson extrapolation suggests that the step
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
109
size of h = 0.25 gives an error of 0.00000087 which is comfortably within the
0.000005 range, which 5 decimal place accuracy requires. Hence, X(3) = 1.46648
to 5 decimal places. In fact, of course, the analytic solution of the problem is
e
and so x(3) = 1.466474.
1−t
e
+e−1
Exercises 2.4.3
18
In each part of this question, the technique is to introduce new variables to
represent each derivative of the dependent variable up to one less than the order
of the equation. This can be done by inspection.
18(a)
dx
= v,
dt
dv
= 4xt − 6(x2 − t)v,
dt
x(0) = 1
v(0) = 2
18(b)
dx
= v,
dt
dv
= −4(x2 − t2 ),
dt
x(1) = 2
v(1) = 0.5
18(c)
dx
= v,
dt
dv
= − sin v − 4x,
dt
x(0) = 0
v(0) = 0
18(d)
dx
= v,
dt
dv
= w,
dt
dw
= e2t + x2 t − 6et v − tw,
dt
x(0) = 1
v(0) = 2
w(0) = 0
18(e)
dx
= v,
dt
dv
= w,
dt
dw
= sin t − tw − x2 ,
dt
c Pearson Education Limited 2011
x(1) = 1
v(1) = 0
w(1) = −2
110
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
18(f)
dx
= v,
dt
dv
= w,
dt
dw
= (tw + t2 x2 )2 ,
dt
x(2) = 0
v(2) = 0
w(2) = 2
18(g)
dx
dt
dv
dt
dw
dt
du
dt
= v,
x(0) = 0
= w,
v(0) = 0
= u,
w(0) = 4
= ln t − x2 − xw,
u(0) = −3
18(h)
dx
dt
dv
dt
dw
dt
du
dt
19
= v,
x(0) = a
= w,
v(0) = 0
= u,
w(0) = b
= t2 + 4t − 5 +
√
xt − v − (v − 1)tu,
u(0) = 0
First, we recast the equation as a pair of coupled first-order equations in the
form
dx
= f1 (t, x, v),
dt
dv
= f2 (t, x, v),
dt
x(0) = x0
v(0) = v0
This yields
dx
= v,
dt
dv
= sin t − x − x2 v,
dt
c Pearson Education Limited 2011
x(0) = 0
v(0) = 1
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
111
Now, applying Euler’s method to the two equations we have
X1 = x0 + h1 f(t0 , x0 , v0 )
that is, X1 = 0 + 0.1 × 1 = 0.10000
X2 = X1 + hf1 (t1 , X1 , V1 )
that is, X2 = 0.10000 + 0.1 × 1.00000
V1 = v0 + hf2 (t0 , x0 , v0 )
V1 = 1.00000 = 1 + 0.1 × (sin 0 − 0 − 02 × 1)
V2 = V1 + hf2 (t1 , X1 , V1 )
V2 = 1 + 0.1 × (sin 0.10000
− 0.10000 − 0.100002 × 1.00000)
= 0.20000
= 0.99898
X3 = X2 + hf1 (t2 , X2 , V2 )
that is, X2 = 0.20000 + 0.1 × 0.99898
= 0.29990
20
The second-order Adams–Bashforth method applied to a pair of coupled
equations is
Xn+1 = Xn + 12 h (3f1 (tn , Xn , Yn ) − f1 (tn−1 , Xn−1 , Yn−1 ))
Yn+1 = Yn + 12 h (3f2 (tn , Xn , Yn ) − f2 (tn−1 , Xn−1 , Yn−1 ))
First, we recast the differential equation as a pair of coupled first-order equations.
dx
= v,
x(0) = 0
dt
dv
v(0) = 1
= sin t − x − x2 v,
dt
Now, since the Adams–Bashforth method is a two-step process, we need to start the
computation with another method. We use the second-order predictor–corrector.
This has the form
X̂n+1 = Xn + hf1 (tn , Xn , Yn )Ŷn+1 = Yn + hf2 (tn , Xn , Yn )
1
Xn+1 = Xn + h f1 (tn+1 , X̂n+1 , Ŷn+1 ) + f1 (tn , Xn , Yn )
2
1
Yn+1 = Yn + h f2 (tn+1 , X̂n+1 , Ŷn+1 ) + f2 (tn , Xn , Yn )
2
Hence we have
X̂1 = 0 + 0.1 × 1 = 0.10000V̂1 = 1 + 0.1 × (0 − 0 − 02 × 1)1 = 1.00000
X1 = 0 + 0.05 × (1 + 1) = 0.10000V1 = 1 + 0.05 × (−0.01017 + 0) = 0.99949
c Pearson Education Limited 2011
112
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Now, continuing using the Adams–Bashforth method we have
X2 = X1 + 12 h (3f1 (t1 , X1 , Y1 ) − f1 (t0 , x0 , v0 ))
V2 = V1 + 12 h (3f2 (t1 , X1 , V1 ) − f2 (t0 , x0 , v0 ))
that is,
X2 = 0.10000 + 0.05 × (3 × 0.99949 − 1) = 0.19992
V2 = 0.99949 + 0.05 × (3 × −0.01016 − 0) = 0.99797
X3 = X2 + 12 h (3f1 (t2 , X2 , V2 ) − f1 (t1 , X1 , V1 ))
that is,
21
X3 = 0.19992 + 0.05 × (3 × 0.99797 − 0.99949) = 0.29964
First, we formulate the problem as a set of 3 coupled first-order differential
equations
dx
= u,
x(0.5) = −1
dt
du
= v,
u(0.5) = 1
dt
dv
= x2 − v(x − t) − u2 ,
v(0.5) = 2
dt
We can then solve these by the predictor–corrector method. Notice that we need to
compute the predicted values for all three variables before computing the corrected
values for any of them.
X̂1 = x0 + hf1 (t0 , x0 , u0 , v0 ) = −1 + 0.05(1) = −0.95000
Û1 = u0 + hf2 (t0 , x0 , u0 , v0 ) = 1 + 0.05(2) = 1.10000
V̂1 = v0 + hf3 (t0 , x0 , u0 , v0 ) = 2 + 0.05((−1)2 − 2(−1 − 0.5) − 1) = 2.15000
X1 = x0 + h f1 t1 , X̂1 , Û1 , V̂1 + f1 (t0 , x0 , u0 , v0 )
1
2
= −1 + 0.025 × (1.10000 + 1.00000) = −0.94750
U1 = u0 + 12 h f2 t1 , X̂1 , Û1 , V̂1 + f2 (t0 , x0 , u0 , v0 )
= 1 + 0.025 × (2.15000 + 2.00000) = 1.10375
1
V1 = v0 + 2 h f3 t1 , X̂1 , Û1 , V̂1 + f3 (t0 , x0 , u0 , v0 )
= 2 + 0.025 × (2.91750 + 3.00000) = 2.14794
Continuing the process we obtain
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
113
X̂2 = −0.94750 + 0.05 × 1.10375 = −0.89231
Û2 = 1.10375 + 0.05 × 2.14794 = 1.21115
V̂2 = 2.14794 + 0.05 × 2.89603 = 2.29274
X2 = −0.94750 + 0.025 × (1.21115 + 1.10375) = −0.88963
U2 = 1.10375 + 0.025 × (2.29274 + 2.14794) = 1.21477
V2 = 2.14794 + 0.025 × (2.75083 + 2.89603) = 2.28911
X̂3 = −0.88963 + 0.05 × 1.21477 = −0.82889
Û3 = 1.21477 + 0.05 × 2.28911 = 1.32923
V̂3 = 2.28911 + 0.05 × 2.72570 = 2.42539
X3 = −0.88963 + 0.025 × (1.32923 + 1.21477) = −0.82603
22
The first step in solving this problem is to convert the problem to a pair of
coupled first-order differential equations
dx
= v,
dt
dv
= sin t − x2 v − x,
dt
x(0) = 0
v(0) = 1
A pseudocode algorithm to compute the value of X(1.6) is shown in Figure 2.9.
Using a program derived from this algorithm with a step size h = 0.4 gives
X(1.6) = 1.220254 and, with a step size h = 0.2, gives X(1.6) = 1.220055. The
method of Richardson extrapolation, given in Section 2.3.6, is equally applicable
to problems such as this one involving coupled equations. Since the Runge–
Kutta method has a local error of O(h)5 the global error will be O(h4 ) . The
method therefore gives the estimated error in the second value of X(1.6) as
(1.220254 − 1.220055)/15 = 0.000013. For 6 decimal place accuracy in the final
estimate we need error ≤ 0.0000005; in other words, the error must be reduced
by a factor of 0.000013/0.0000005 = 26. Since Runge–Kutta is a fourth-order
√
method the required step length will be 0.2/4 26 = 0.088. Rounding this down to
a suitable size suggests that a step size of h = 0.08 will give a solution accurate to
6 decimal places. In fact the program yields, with h = 0.08, X(1.6) = 1.2200394.
With h = 0.04 it gives X(1.6) = 1.2200390. Richardson extrapolation predicts the
c Pearson Education Limited 2011
114
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
error in the h = 0.04 solution as 0.00000003 and therefore that in the h = 0.08 as
0.0000005. The required accuracy was therefore just achieved using h = 0.08.
{program solves a pair of ordinary differential equations by the 4th
order Runge-Kutta Method}
procedure f1(t, x, v → f1)
f1 ← v
endprocedure
procedure f2(t, x, v → f2)
f2 ← sin(t) − x∗ x∗ v − x
endprocedure
{procedure computes values of x and v at the next time step}
procedure rk4(t, x, v, h → xn, vn)
c11 ← h∗ f1(t, x, v)
c21 ← h∗ f2(t, x, v)
c12 ← h∗ f1(t + h/2, x + c11/2, v + c21/2)
c22 ← h∗ f2(t + h/2, x + c11/2, v + c21/2)
c13 ← h∗ f1(t + h/2, x + c12/2, v + c22/2)
c23 ← h∗ f2(t + h/2, x + c12/2, v + c22/2)
c14 ← h∗ f1(t + h, x + c13, v + c23)
c24 ← h∗ f2(t + h, x + c13, v + c23)
xn ← x + (c11 + 2∗ (c12 + c13) + c14)/6
vn ← v + (c21 + 2∗ (c22 + c23) + c24)/6
endprocedure
t start ← 0
t end ← 1.6
x0 ← 0
v0 ← 1
write(vdu, "Enter step size")
read(keyboard, h)
write(printer,t start,x0)
t ← t start
x ← x0
v ← v0
repeat
rk4(t, x, v, h → xn, vn)
x ← xn
v ← vn
t←t+h
write(printer,t,x)
until t >= t end
Figure 2.9: Pseudocode algorithm for Exercise 22
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
23
115
The first step in solving this problem is to convert the problem to a set of
coupled first-order differential equations
dx
= v,
dt
dv
= w,
dt
dw
= x2 − v2 − (x − t)w,
dt
x(0.5) = −1
v(0.5) = 1
w(0.5) = 2
A pseudocode algorithm to compute the value of X(2.2) is shown in Figure 2.11.
The procedures used in the algorithm are defined in Figure 2.10.
{procedures for pseudocode algorithm in figure 2.11}
procedure f1(t, x, v, w → f1)
f1 ← v
endprocedure
procedure f2(t, x, v, w → f2)
f2 ← w
endprocedure
procedure f3(t, x, v, w → f3)
f3 ← x∗ x − v∗ v − (x − t)∗ w
endprocedure
{procedure computes values of x, v and w at the next time step using
the 4th order Runge-Kutta procedure}
procedure rk4(t, x, v, w, h → xn, vn, wn)
c11 ← h∗ f1(t, x, v, w)
c21 ← h∗ f2(t, x, v, w)
c31 ← h∗ f3(t, x, v, w)
c12 ← h∗ f1(t + h/2, x + c11/2, v + c21/2, w + c31/2)
c22 ← h∗ f2(t + h/2, x + c11/2, v + c21/2, w + c31/2)
c32 ← h∗ f3(t + h/2, x + c11/2, v + c21/2, w + c31/2)
c13 ← h∗ f1(t + h/2, x + c12/2, v + c22/2, w + c32/2)
c23 ← h∗ f2(t + h/2, x + c12/2, v + c22/2, w + c32/2)
c33 ← h∗ f3(t + h/2, x + c12/2, v + c22/2, w + c32/2)
c14 ← h∗ f1(t + h, x + c13, v + c23, w + c33)
c24 ← h∗ f2(t + h, x + c13, v + c23, w + c33)
c34 ← h∗ f3(t + h, x + c13, v + c23, w + c33)
(Continued)
c Pearson Education Limited 2011
116
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
xn ← x + (c11 + 2∗ (c12 + c13) + c14)/6
vn ← v + (c21 + 2∗ (c22 + c23) + c24)/6
wn ← w + (c31 + 2∗ (c32 + c33) + c34)/6
endprocedure
{procedure computes values of x, v and w at the next time step using
the 3rd order predictor-corrector procedure}
procedure pc3(t, xo, vo, wo, x, v, w, h → xn, vn, wn)
xp ← x + h∗ (3∗ f1(t, x, v, w) − f1(t − h, xo, vo, wo))/2
vp ← v + h∗ (3∗ f2(t, x, v, w) − f2(t − h, xo, vo, wo))/2
wp ← w + h∗ (3∗ f3(t, x, v, w) − f3(t − h, xo, vo, wo))/2
xn ← x + h∗ (5∗ f1(t + h, xp, vp, wp) + 8∗ f1(t, x, v, w)
-f1(t-h,xo,vo,wo))/12
∗ ∗
vn ← v + h (5 f2(t + h, xp, vp, wp) + 8∗ f2(t, x, v, w)
-f2(t-h,xo,vo,wo))/12
∗ ∗
wn ← w + h (5 f3(t + h, xp, vp, wp) + 8∗ f3(t, x, v, w)
-f3(t-h,xo,vo,wo))/12
endprocedure
Figure 2.10: Pseudocode procedures for algorithm for Exercise 23
Using a program derived from this algorithm with a step size h = 0.1 gives
X(2.2) = 2.923350 and, with a step size h = 0.05, gives X(2.2) = 2.925418. The
method of Richardson extrapolation given in Section 2.3.6, is equally applicable
to problems such as this one involving coupled equations. Since the third-order
predictor–corrector method used has a local error of O(h4 ) , the global error will
be O(h3 ) . The method therefore gives the estimated error in the second value of
X(2.2) as 2.923350−2.925418)/7 = −0.000295. For 6 decimal place accuracy in the
final estimate we need error ≤ 0.0000005; in other words, the error must be reduced
by a factor of 0.000295/0.0000005 = 590. Since we are using a third-order method,
√
the required step length will be 0.05/3 590 = 0.00596. Rounding this down to a
suitable size suggests that a step size of h = 0.005 will give a solution accurate to
6 decimal places. In fact the program yields, with h = 0.005, X(2.2) = 2.92575057.
With h = 0.0025 it gives X(2.2) = 2.92575089. Richardson extrapolation predicts
the error in the h = 0.0025 solution as 0.000000046 and therefore that in the
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
h = 0.005 as 0.00000037.
117
The required accuracy was therefore comfortably
achieved using h = 0.005.
{program solves three ordinary differential equations by the
3rd order predictor-corrector method}
t start ← 0.5
t end ← 2.2
x start ← −1
v start ← 1
w start ← 2
write(vdu, "Enter step size")
read(keyboard, h)
write(printer, t start, x start)
t ← t start
xo ← x start
vo ← v start
wo ← w start
rk4(t, xo, vo, wo, h → x, v, w)
t←t+h
write(printer,t,x)
repeat
pc3(t, xo, vo, wo, x, v, w, h → xn, vn, wn)
xo ← x
vo ← v
wo ← w
x ← xn
v ← vn
w ← wn
t←t+h
write (printer, t, x)
until t >= t end
Figure 2.11: Pseudocode algorithm for Exercise 23
Review exercises 2.7
1
Euler’s method for the solution of the differential equation
Xn+1 = Xn + hFn = Xn + hf(tn , Xn )
c Pearson Education Limited 2011
dx
= f(t, x) is
dt
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
118
Applying this to the equation
yields
dx √
= x with x(0) = 1 and a step size of h = 0.1
dt
x0 = x(0) = 1
√
√
X1 = x0 + hf(t0 , x0 ) = x0 + h x0 = 1 + 0.1 × 1 = 1.1000
√
X2 = X1 + hf(t1 , X1 ) = X1 + h X1 = 1.1000 + 0.1 1.1000 = 1.2049
√
X3 = X2 + hf(t2 , X2 ) = X2 + h X2 = 1.2049 + 0.1 1.2049 = 1.3146
√
X4 = X3 + hf(t3 , X3 ) = X3 + h X3 = 1.3146 + 0.1 1.3146 = 1.4293
√
X5 = X4 + hf(t4 , X4 ) = X4 + h X4 = 1.4293 + 0.1 1.4293 = 1.5489
Hence Euler’s method with step size h = 0.1 gives the estimate X(0.5) = 1.5489.
2
Euler’s method for the solution of the differential equation
dx
= f(t, x) is
dt
Xn+1 = Xn + hFn = Xn + hf(tn , Xn )
Applying this to the equation
yields
dx
= −ext with x(1) = 1 and a step size of h = 0.05
dt
x0 = x(1) = 1
X1 = x0 + hf(t0 , x0 ) = x0 + h(−ex0 t0 ) = 1 − 0.05 exp(1 × 1) = 0.86409
X2 = X1 + hf(t1 , X1 ) = X1 + h(−eX1 t1 )
= 0.86409 − 0.05 exp(0.86409 × 1.05) = 0.74021
X3 = X2 + hf(t2 , X2 ) = X2 + h(−eX2 t2 )
= 0.74021 − 0.05 exp(0.74021 × 1.10) = 0.62733
X4 = X3 + hf(t3 , X3 ) = X3 + h(−eX3 t3 )
= 0.62733 − 0.05 exp(0.62733 × 1.15) = 0.52447
Hence Euler’s method with step size h = 0.05 gives the estimate X(1.1) = 0.52447.
3
This question could be solved by hand computation or using a computer
program based on a simple modification of the pseudocode algorithm given in
Figure 2.1. With a step size of h = 0.1 it is found that X(0.4) = 1.125584
and, with a step size of h = 0.05, X(0.4) = 1.142763. Using the Richardson
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
119
extrapolation method, because Euler’s method is a first-order method and the
global error is therefore of O(h), the error in the estimate of X(0.4) is approximately
1.142763 − 1.125584 = 0.017179. To obtain X(0.4) accurate to 2 decimal places
we need error ≤ 0.005. To achieve this we would need to reduced step size by a
factor of 0.017179/0.005 = 3.44. This suggests a step size of 0.05/3.44 = 0.0145.
Rounding this down to a sensible figure suggests trying a step size of 0.0125 or 0.01.
4
This question could be solved by hand computation or using a computer
program based on a simple modification of the pseudocode algorithm given in
Figure 2.1. With a step size of h = 0.05 it is found that X(0.25) = 2.003749
and, with a step size of h = 0.025, X(0.25) = 2.004452. Using the Richardson extrapolation method, because Euler’s method is a first-order method and the global
error is therefore of O(h) , the error in the estimate of X(0.25) is approximately
2.004452−2.003749 = 0.000703. To obtain X(0.25) accurate to 3 decimal places we
need error ≤ 0.0005. To achieve this we would need to reduce step size by a factor
of 0.0007/0.0005 = 1.4. This suggests a step size of 0.025/1.4 = 0.0179. Rounding
this down to a sensible figure suggests trying a step size of 0.0166667 or 0.0125.
5
This question could be solved by hand computation or using a computer
program based on a simple modification of the pseudocode algorithm given in
Figure 2.5. By either method it is found that X1 (1.2) = 2.374037, X2 (1.2) =
2.374148 and X3 (1.2) = 2.374176. The local error of the second-order predictor–
corrector is O(h3 ) so the global error is O(h2 ) . Hence it is expected
|X − x| ∝ h2
that is,
therefore, |X1 − X2 | ≈ x − αh −
2
and |X2 − X3 | ≈ x − α
Hence
h
2
x ≈ X + αh2
x−α
2
−
|X1 − X2 |
≈
|X2 − X3 |
2
h
2
x−α
h
4
3
2
4 ah
3
2
16 ah
≈4
c Pearson Education Limited 2011
=
2
=
3 2
αh
4
3
αh2
16
120
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
In fact we find that
|2.374037 − 2.374148|
| − 0.000111|
|X1 − X2 |
=
=
= 3.97 ≈ 4
|X2 − X3 |
|2.374148 − 2.374176|
| − 0.000028|
6
This question is best solved using a computer program based on a simple
modification of the pseudocode given Figure 2.7. Let X1 denote the solution using
a step size of h = 0.2, X2 that using h = 0.1 and X3 that using h = 0.05.
By either method it is found that X1 (2) = 5.19436687, X2 (2) = 5.19432575 and
X3 (2) = 5.19432313. The local error of the fourth-order Runge–Kutta method is
O(h5 ) so the global error is O(h4 ) . Hence, by Richardson extrapolation, we may
expect
4
x(2) = X2 (2) + αh = X3 (2) + α
h
2
4
therefore αh4 ≈ 16(x(2) − X3 (2))
therefore x(2) = X2 (2) + 16(x(2) − X3 (2))
16 × 5.19432313 − 5.19432575
16X3 (2) − X2 (2)
=
= 5.19432296
15
15
and the most accurate estimate of x(2) is 5.19432296.
Hence x(2) =
7
The boundary conditions for this problem are p(r0 ) = p0 and p(r1 ) = 0.
Hence we have
p+r
dp
= 2a − p
dr
⇒
dp 2p
2a
+
=
dr
r
r
This is a linear differential equation, so we first find the integrating factor.
2
2lnr
= r2
r dr = 2 ln r so the integrating factor is e
therefore, r2
dp
+ 2rp = 2ar
dr
that is, r2 p = ar2 + C
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Now p(r0 ) = p0
p(r1 ) = 0
⇒
⇒
r20 p0 = ar20 + C
⇒
121
C = r20 (p0 − a)
0 = ar21 + C = ar21 + r20 (p0 − a)
⇒
a=
r20 p0
r20 − r21
r20
r21
r20 p0
r20 p0
r20 p0
Hence p(r) = 2
+ p0 − 2
= 2
−1
r0 − r21
r0 − r21 r2
r1 − r20 r2
22
12 1
7
− 1 = 13 ( 16
and p(1.5) = 2
9 − 1) = 27
2 − 12 1.52
2(3p + 1)
dp
=−
, p(1) = 1 . This may easily
The problem to solve numerically is
dr
3r
be solved using a modification of the pseudocode algorithm of Figure 2.7 or
of the program of Figure 2.8.
We find that, using a step size of h = 0.05,
p(1.5) = 0.25925946.
8
The first step is to recast the problem as a set of three coupled (linked)
first-order ordinary differential equations
dx
= v,
x(1) = 0.2
dt
dv
= w,
v(1) = 1
dt
dw
w(1) = 0
= sin(t) + xt − 4v2 − w2 ,
dt
Figure 2.12 shows a pseudocode algorithm for the solution of these three equations
by Euler’s Method.
{program solves three ordinary differential equations by the Euler
method}
procedure f1(t, x, v, w → f1)
f1 ← v
endprocedure
procedure f2(t, x, v, w → f2)
f2 ← w
endprocedure
procedure f3(t, x, v, w → f3)
f3 ← sin(t) + x∗ t − 4∗ v∗ v − w∗ w
endprocedure
(Continued)
c Pearson Education Limited 2011
122
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
{procedure computes values of x, v and w at the next time step using
the Euler procedure}
procedure euler(t, x, v, w, h → xn, vn, wn)
xn ← x + h∗ f1(t, x, v, w)
vn ← v + h∗ f2(t, x, v, w)
wn ← w + h∗ f3(t, x, v, w)
endprocedure
t start ← 1.0
t end ← 2.0
x start ← 0.2
v start ← 1
w start ← 0
write(vdu, "Enter step size")
read(keyboard, h)
write(printer, t start, x start)
t ← t start
x ← x start
v ← v start
w ← w start
repeat
euler(t, x, v, w, h → xn, vn, wn)
x ← xn
v ← vn
w ← wn
t←t+h
write(printer, t,x)
until t >= t end
Figure 2.12: Pseudocode algorithm for Review Exercise 8
Using a program derived from this algorithm with a step size h = 0.025 gives
X(2) = 0.847035 and, with a step size h = 0.0125, gives X(2) = 0.844067. The
method of Richardson extrapolation, given in Section 2.3.6, is equally applicable to
problems such as this one involving coupled equations. Since Euler’s method has a
local error of O(h2 ) the global error will be O(h) . The method therefore gives the
estimated error in the second value of X(2) as (0.847035−0.844067)/1 = 0.002968.
This is less than 5 in the third decimal place, so we have two significant figures of
accuracy. The best estimate we can make is that x(2) = 0.84 to 2dp.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
123
9 The first step is to recast the problem as a set of two coupled (linked) first-order
ordinary differential equations
dx
= v,
dt
dv
= (1 − x2 )v − 40x,
dt
x(0) = 0.02
v(0) = 0
Figure 2.13 shows a pseudocode algorithm for the solution of these equations by
the second-order predictor–corrector method. Using a program derived from this
algorithm with a stepsize h = 0.02 gives X(4) = 0.147123 and, with a step size
h = 0.01, gives X(4) = 0.146075. The method of Richardson extrapolation,
given in Section 2.3.6, is applicable to problems such as this one involving coupled
equations. Since the second-order predictor–corrector method has a local error of
O(h3 ) , the global error will be O(h2 ) . The method therefore gives the estimated
error in the second value of X(4) as (0.147123 − 146075)/3 = 0.001048. For 4
decimal place accuracy in the final estimate we need error ≤ 0.00005; in other
words, the error must be reduced by a factor of 0.001048/0.00005 = 20.96. Since
this predictor–corrector is a second-order method the required step length will be
√
0.01/ 20.96 = 0.0022. Rounding this down to a suitable size suggests that a
step size of h = 0.002 will give a solution accurate to 4 decimal places. In fact
the program yields, with h = 0.002, X(4) = 0.145813. With h = 0.001 it gives
X(4) = 0.145807. Richardson extrapolation predicts the error in the h = 0.001
solution as 0.000002 and therefore that in the h = 0.002 as 0.000008. The required
accuracy was therefore comfortably achieved using h = 0.002. We can be confident
that x(4) = 0.1458 to 4dp.
{program solves two ordinary differential equations by the second
order predictor-corrector method}
procedure f1(t, x, v → f1)
f1 ← v
endprocedure
procedure f2(t, x, v → f2)
f2 ← (1 − x∗ x)∗ v − 40∗ x
endprocedure
(Continued)
c Pearson Education Limited 2011
124
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
{procedure computes values of x, v and w at the next time step using
the second order predictor-corrector procedure}
procedure pc2(t, x, v, h → xn, vn)
xp ← x + h∗ f1(t, x, v)
vp ← v + h∗ f2(t, x, v)
xn ← x + h∗ (f1(t + h, xp, vp) + f1(t, x, v))/2
vn ← v + h∗ (f2(t + h, xp, vp) + f2(t, x, v))/2
endprocedure
t start ← 0.0
t end ← 4.0
x start ← 0.02
v start ← 0
write(vdu, "Enter step size")
read(keyboard, h)
writeprinter, t start, x start
t ← t start
x ← x start
v ← v start
repeat
pc2(t, x, v, h → xn, vn)
x ← xn
v ← vn
t←t+h
write(printer,t,x)
until t >= t end
Figure 2.13: Pseudcode algorithm for Review Exercise 9
10
The first step is to recast the problem as a set of three coupled (linked)
first-order ordinary differential equations.
dx
= v,
x(1) = −1
dt
dv
= w,
v(1) = 1
dt
dw
= sin(t) + xt − 4v3 − |w|,
w(1) = 2
dt
A minor modification of the pseudocode algorithm shown in Figure 2.9 provides
an algorithm for the solution of this problem. Using a program derived from
this algorithm with a step size h = 0.1 gives X(2.5) = −0.651076 and,
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
with a step size h = 0.05, gives X(2.5) = −0.653798.
of Richardson extrapolation, given in Section 2.3.6.
125
We use the method
Since the Runge–Kutta
5
method has a local error of O(h ) the global error will be O(h4 ) . The method
therefore gives the estimated error in the second value of X(2.5) as (0.651076 −
0.653798)/15 = −0.000181. For 4 decimal place accuracy in the final estimate we
need error ≤ 0.00005; in other words, the error must be reduced by a factor of
0.000181/0.00005 = 3.63. Since the Runge–Kutta is a fourth-order method the
√
required step length will be 0.054 3.63 = 0.036. Rounding this down to a suitable
size suggests that a stepsize of h = 0.025 will give a solution accurate to 4 decimal
places. In fact the program yields, with h = 0.025, X(2.5) = −0.653232. With
h = 0.0125 it gives X(2.5) = −0.653217. Richardson extrapolation predicts the
error in the h = 0.0125 solution as 0.0000009 and therefore that in the h = 0.025
as 0.000015. The required accuracy is therefore easily achieved using h = 0.025.
We can be confident that x(2.5) = −0.6532 to 4dp.
c Pearson Education Limited 2011
3
Vector Calculus
Exercises 3.1.2
1(a)
f(x, y) = c ⇒ x2 + y2 = 1 + ec
Contours are a family of concentric circles, centre (0,0) and radius > 1.
1(b)
f(x, y) = c ⇒ y = (1 + x) tan c
Contours are a family of straight lines whose y intercept equals their slope and
pass through (-1,0).
2(a)
Flow lines are given by
dx
dt
= y and
dy
dt
= 6x2 − 4x .
Thus,
6x2 − 4x
dy
=
dx
y
y dy = (6x2 − 4x) dx + c
1 2
y = 2x3 − 2x2 + c
2
y2 = 4x2 (x − 1) + C
2(b) Flow lines are given by
dx
dt
= y and
dy
dt
= 16 x3 − x .
Thus,
−x
y
1
y dy = ( x3 − x) dx + c
6
1 2
1 4 1 2
y =
x − x +c
2
24
2
1 2 2
2
x (x − 12) + C
y =
12
dy
=
dx
1 3
6x
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
3(a)
127
Level surfaces are given by f(r) = c ⇒ z = c + xy .
3(b)
Level surfaces are given by f(r) = c ⇒ z = c − e−xy .
4(a)
Field lines are given by
dx
dt
= xy ,
dy
dt
= y2 + 1,
dz
dt
= z.
dz
= z ⇒ z = Aet
dt
dy
= 1 + y2 ⇒ y = tan(t + α)
dt
dx
= xy ⇒ ln x = C − ln(cos(t + α))
dt
B
cos(t + α)
x 2
is a hyperbola, the curve is on a
Since 1 + tan2 θ = sec2 θ ⇒ 1 + y2 = B
x=
hyperbolic cylinder.
4(b) Field lines are given by
dx
dt
= yz ,
dy
dt
dz
dt
= zx ,
= xy .
Hence,
dy
x
= ⇒ y 2 = x2 − c
dx
y
dz
x
= ⇒ z 2 = x2 + k
dx
z
The curve is the intersection of these hyperbolic cylinders.
5(a)
∂f
∂x
∂2 f
∂x2
∂2 f
∂x∂y
∂2 f
∂z∂x
= yz − 2x
= −2
=z
=y
∂f
∂y
∂2 f
∂y2
∂2 f
∂y∂x
∂2 f
∂y∂z
= xz + 1
=0
=z
=x
∂f
= xy − 1
∂z
∂2 f
=0
∂z2
∂2 f
=y
∂x∂z
∂2 f
=x
∂z∂y
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
128
5(b)
∂f
∂x
∂2 f
∂x2
∂2 f
∂x∂y
∂2 f
∂z∂x
= 2xyz3
= 2yz3
= 2xz3
= 6xyz2
∂f
∂y
∂2 f
∂y2
∂2 f
∂y∂x
∂2 f
∂y∂z
∂f
= 3x2 yz2
∂z
∂2 f
= 6x2 yz
2
∂z
∂2 f
= 6xyz2
∂x∂z
∂2 f
= 3x2 z2
∂z∂y
= x2 z 3
=0
= 2xz3
= 3x2 z2
5(c)
−y
z
∂f
−yz
=
=
y 2
∂x
1 + (x)
x2
x2 + y 2
∂f
z
=
∂y
1 + ( xy )2
xz
1
= 2
x
x + y2
∂2 f
−2xyz
= 2
2
∂y
(x + y2 )2
∂2 f
2xyz
= 2
2
∂x
(x + y2 )2
∂f
−1 y
= tan
∂z
x
2
∂ f
=0
∂z2
∂2 f
z
2x2 z
z(x2 + y2 ) − 2x2 z
(y2 − x2 )z
= 2
−
=
=
∂x∂y
x + y2
(x2 + y2 )2
(x2 + y2 )2
(x2 + y2 )2
∂2 f
2y2 z
2y2 z − z(x2 + y2 )
(y2 − x2 )z
−z
+
=
=
= 2
∂y∂x
x + y2
(x2 + y2 )2
(x2 + y2 )2
(x2 + y2 )2
−y
∂2 f
∂2 f
−y
1
−y
=
=
=
y 2
∂x∂z
1 + (x)
x2
x2 + y 2
∂z∂x
x2 + y 2
∂2 f
∂2 f
x
1
1
x
=
=
= 2
y 2
2
2
∂y∂z
1 + (x)
x
x +y
∂z∂y
x + y2
6(a)
∂f
∂x
= 2x ,
∂f
∂y
= 2y ,
∂f
∂z
= −1,
dx
dt
= 3t2 ,
df
= 2(t3 − 1)(3t2 ) + 2(2t)(2) + (−1)
dt
dy
dt
= 2,
dz
dt
−1
(t − 1)2
=
−1
(t−1)2
= {2t(3t4 − 3t + 4)(t − 1)2 + 1}/(t − 1)2
= {2t(3t6 − 6t5 + 3t4 − 3t3 + 10t2 − 11t + 4) + 1}/(t − 1)2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
129
6(b)
∂f
∂f
∂f
dx
= yz,
= xz,
= xy,
= e−t (cos t − sin t),
∂x
∂y
∂z
dt
dy
dz
= −e−t (cos t + sin t),
=1
dt
dt
df
= te−t cos t.e−t (cos t − sin t) − te−t sin t.e−t (cos t + sin t) + e−2t sin t cos t.(1)
dt
= te−2t (cos2 t − sin2 t − 2 sin t cos t) + e−2t sin t cos t
1
= te−2t (cos 2t − sin 2t) + e−2t sin 2t
2
7
r2 = x2 + y2 + z2 , tan φ = xy , tan θ =
(x2 +y 2 )1/2
z
∂r
1
y
∂φ
1
x
cos φ
= = sin θ sin φ,
=
= 2
=
y 2
2
∂y
r
∂y
1 + (x) x
x +y
r sin θ
∂θ
∂
yz
(x2 + y2 )1/2
sin φ cos θ
=
{tan−1
}= 2
=
2
2
2
2
1/2
∂y
∂y
z
r
(x + y + z )(x + y )
∂f
∂f
cos φ ∂f
sin φ cos θ ∂f
= sin θ sin φ +
+
∂y
∂r r sin θ ∂φ
r
∂θ
z
∂φ
∂r
= = cos θ,
=0
∂z
r
∂z
∂θ
1
(x2 + y2 )1/2
−(x2 + y2 )1/2
− sin θ
=
−
=
=
2
2
2
2
2
2
x
+y
∂z
z
x +y +z
r
1 + z2
∂f
∂f sin θ ∂f
= cos θ −
∂z
∂r
r ∂θ
8
∂u
∂x
=
df ∂r
∂r
dr ∂x , ∂x
=
x
r
⇒
∂u
∂x
=
x df
r dr
∂2 u
∂ x df x ∂ df
+
=
∂x2
∂x r dr
r ∂x dr
x
r − x( r ) df x d2 f x
+
=
r2
dr
r dr2 r
2
2
2 2
y + z df x d f
+ 2 2
=
r.r2 dr
r dr
c Pearson Education Limited 2011
130
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Similarly (by symmetry),
∂2u
∂y 2
=
x2 +y 2 df
r.r 2 dr
+
y2 ∂ 2 f
d2 u
r2 ∂r2 , dz 2
=
x2 +z 2 df
r.r 2 dr
+
2
z2 d f
r2 dr2
∂2 u ∂2 u ∂2 u
2(x2 + y2 + z2 ) df x2 + y2 + z2 d2 f
+
⇒ 2+ 2 + 2 =
∂x
∂y
∂z
r.r2
dr
r2
dr2
2 df d2 f
+ 2
=
r dr
dr
Hence, the result.
9
V(x, y, z) =
∂V
∂x
∂2 V
∂x2
∂V
∂y
∂2 V
∂y2
∂V
∂z
1
z
2
+y
exp − x 4z
2
1
x2 + y2 −x
= exp −
z
4z
2z
2
2
x +y
−x 2
1
1
x2 + y 2
= exp −
− 2 exp −
z
4z
2z
2z
4z
1
x2 + y2 −y
= exp −
z
4z
2z
2
2
−y 2
1
1
x2 + y 2
x +y
= exp −
− 2 exp −
z
4z
2z
2z
4z
1
x2 + y 2
x2 + y 2
1
x2 + y 2
+ exp −
= − 2 exp −
z
4z
z
4z
4z2
⇒
10
∂V
∂2 V ∂ 2 V
+ 2 =
2
∂x
∂y
∂z
V = sin 3x cos 4y cosh 5z
∂2 V
= −9 sin 3x cos 4y cosh 5z
∂x2
∂2 V
= −16 sin 3x cos 4y cosh 5z
∂y2
∂2 V
= 25 sin 3x cos 4y cosh 5x
∂z2
⇒
∂2 V ∂2 V ∂2 V
+ 2 + 2 =0
∂x2
∂y
∂z
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 3.1.4
11
x + y = u,
y = uv
∂x ∂y
+
= 1,
∂u ∂u
∂y
∂x
=v⇒
=1−v
∂u
∂u
∂y
∂x
∂x ∂y
+
= 0,
=u⇒
= −u
∂v
∂v
∂v
∂v
∂(x, y) 1 − v v
= u − uv − (−uv) = u
=
∂(u, v) −u u
12
x + y + z = u, y + z = uv, z = uvw
∂x ∂y
∂z
∂y
∂z
∂z
+
+
= 1,
+
= v,
= vw
∂u ∂u ∂u
∂u ∂u
∂u
⇒
∂y
∂x
∂z
= vw,
= v(1 − w),
=1−v
∂u
∂u
∂u
∂x ∂y ∂z
∂y ∂z
∂z
+
+
= 0,
+
= u,
= uw
∂v
∂v ∂v
∂v ∂v
∂v
∂z
∂y
∂x
= uw,
= u − uw,
= −u
∂v
∂v
∂v
∂x
∂y
∂z
∂y
∂z
∂z
+
+
= 0,
+
= 0,
= uv
∂w ∂w ∂w
∂w ∂w
∂w
∂y
∂x
∂z
= uv,
= −uv,
=0
∂w
∂w
∂w
1 − v v − vw vw
∂(x, y, z)
⇒
= −u u − uw uw
∂(u, v, w)
0
−uv
uv
1 − v v vw
1 − v v
= −u u uw = uv
−u
u
0
0 uv
2 1− v v
= u v
= u2 v
−1 1
⇒
c Pearson Education Limited 2011
131
132
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
13
x = eu cos v, y = eu sin v
∂x
= eu cos v,
∂u
∂x
= −eu sin v,
∂v
∂(x, y) eu cos v
=
⇒
∂(u, v) −eu sin v
∂y
= eu sin v
∂u
∂y
= eu cos v
∂v
eu sin v
= e2u (cos2 v + sin2 v) = e2u
eu cos v
1
ln(x2 + y2 )
2
y
y
= tan v ⇒ v = tan−1
x
x
x2 + y2 = e2u ⇒ u =
∂u
∂v
x
y
y
x
∂u
∂v
= 2
= 2
=− 2
= 2
,
,
2
2
2
∂x
x +y
∂y
x + y ∂x
x + y ∂y
x + y2
x
y
2
2
− x2 +y
∂(u, v) x2 +y
1
1
2
2
= x +y =
= y
= 2u
x
2
2
2
2
2
∂(x, y)
(x + y )
x +y
e
x2 +y 2
x2 +y 2
Hence, the result.
14
∂(u, v) (− sin x cos y − λ cos x sin y) (cos x cos y − λ sin x sin y)
=
∂(x, y) (− cos x sin y − λ sin x cos y) (− sin x sin y + λ cos x cos y)
= − (sin x cos y + λ cos x sin y)(− sin x sin y + λ cos x cos y)
+ (cos x sin y + λ sin x cos y)(cos x cos y − λ sin x sin y)
= − [− sin2 x sin y cos y + λ sin x cos x cos2 y − λ sin x cos x sin2 y
+ λ2 cos2 x sin y cos y] + [cos2 x sin y cos y − λ sin x cos x sin2 y
+ λ sin x cos x cos2 y − λ2 sin2 x sin y cos y]
= sin y cos y − λ2 sin y cos y
∂(u, v)
= 0 ⇒ λ2 = 1 ⇒ λ = −1 or
∂(x, y)
c Pearson Education Limited 2011
1
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
15
133
2Kx 3 (3z + 6y)
∂(u, v, w)
= 8y 2 (2z + 6x)
∂(x, y, z)
2z 1 (2y + 3x)
Kx − 3z 0 (3z − 9x)
= 2 4y − 2z 0 (2z − 4y)
z
1 (2y + 3x)
Kx − 3z 3z − 9x
= 4(z − 2y)(−Kx + 9x)
= −2
4y − 2z 2z − 4y
∂(u, v, w)
=0⇒K=9
∂(x, y, z)
u = 9x2 + 4y2 + z2
v2 = 9x2 + 4y2 + z2 + 12xy + 6xz + 4yz
2w = 12xy + 6xz + 4yz
u = v2 − 2w
16
∂u ∂x ∂u ∂y
+
(differentiating u = g(x, y) with respect to u)
∂x ∂u ∂y ∂u
∂v ∂x ∂v ∂y
0=
+
(differentiating v = h(x, y) with respect to u)
∂x ∂u ∂y ∂u
1 ∂u
∂y
0 ∂v
∂v
∂x
=
⇒
= ∂u ∂y
/J
∂u
∂x ∂y
∂u
∂y
∂v ∂v
1=
∂x
∂y
∂y
∂v
= − /J
∂u
∂x
Similarly, differentiating u = g(x, y) and v = h(x, y) with respect to v obtains the
other two expressions.
17
u = ex cos y, v = e−x sin y
∂u
∂v
= ex cos y = u,
= −e−x sin y = −v
∂x
∂x
∂u
∂v
= −ex sin y,
= e−x cos y
∂y
∂y
c Pearson Education Limited 2011
134
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
∂x
∂u
∂x
∂v
∂y
∂u
∂y
∂v
e−x cos y
cos2 y − sin2 y
ex sin y
=
cos2 y − sin2 y
e−x sin y
=
cos2 y − sin2 y
ex cos y
=
cos2 y − sin2 y
=
Since 2uv = 2 sin y cos y = sin 2y , it is possible to express these results in terms of
u and v.
√
1
(1 + 1 − 4u2 v2 )
2
√
1
cos y = (1 − 1 − 4u2 v2 )
2
√
1
(1 + 1 − 4u2 v2 )
ex =
2u
sin y =
Exercises 3.1.6
18(a)
∂ 2
(y + 2xy + 1) = 2y + 2x
∂y
∂
(2xy + x2 ) = 2y + 2x
∂x
Therefore, it is an exact differential.
Let
∂f
∂x
= y2 + 2xy + 1, then f(x, y) = xy2 + x2 y + x + c(y)
and
But,
∂f
∂y
∂f
dc
= 2xy + x2 +
∂y
dy
= 2xy + x2 from the question. Hence,
dc
dy
= 0; so c is independent of x
and y ⇒ f(x, y) = x y + y x + x + c.
2
2
18(b)
∂
(2xy2 + 3y cos 3x) = 4xy + 3 cos 3x
∂y
∂
(2x2 y + sin 3x) = 4xy + 3 cos 3x
∂x
Therefore, it is an exact differential.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Let
∂f
∂x
135
= 2xy2 + 3y cos 3x , then f(x, y) = x2 y2 + y sin 3x + c(y)
and
Hence,
dc
dy
∂f
dc
= 2x2 y + y sin 3x +
∂y
dy
= 0 and c is a constant with respect to both x and y
⇒ f(x, y) = x2 y2 + y sin 3x + c
18(c)
∂
(6xy − y2 ) = 6x − 2y
∂y
∂
(2xey − x2 ) = 2ey − 2x
∂x
Not equal, so not an exact differential.
18(d)
∂ 3
(z − 3y) = −3
∂y
∂
(12y2 − 3x) = −3
∂x
Hence, exact.
∂f
∂y
= −3x +
Let
∂c
∂y
∂
(12y2 − 3x) = 0
∂z
∂
(3xz2 ) = 0
∂y
∂ 3
(z − 3y) = 3z2
∂z
∂
(3xz2 ) = 3z2
∂x
= z3 − 3y , then f(x, y, z) = z3 x − 3xy + c(y, z) and
∂f
∂x
. This inturn implies that
∂c
∂y
= 12y2 and c(y, z) = 4y3 + k(z) .
∂f
∂c
dk
= 3z2 x +
= 3z2 x +
.
∂z
∂z
dz
This inturn implies that
19
dk
dz
= 0 and so f(x, y, z) = z3 x − 3xy + 4y3 + K.
∂
(y cos x + λ cos y) = cos x − λ sin y
∂y
∂
(x sin y + sin x + y) = sin y + cos x
∂x
Equal, if λ = −1.
Let
∂f
∂y
∂f
∂x
=
y cos x − cos y , then f(x, y)
=
y sin x − x cos y + c(y) and
= sin x + x sin y + c (y) so that c (y) = y and c(y) = 12 y2 + k .
c Pearson Education Limited 2011
136
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Hence, f(x, y) = y sin x − x cos y + 12 y2 + k .
f(0, 1) = 0 ⇒ 0 = 0 + 0 +
1
1
+k⇒k=−
2
2
and f(x, y) = y sin x − x cos y + 12 (y2 − 1) .
20
∂
(10x2 + 6xy + 6y2 ) = 6x + 12y
∂y
∂
(9x2 + 4xy + 15y2 ) = 18x + 4y
∂x
Hence, not exact.
∂
[(2x + 3y)m (10x2 + 6xy + 6y2 )] = 3m(2x + 3y)m−1 (10x2 + 6xy + 6y2 )
∂y
+ (2x + 3y)m (6x + 12y)
∂
[(2x + 3y)m (9x2 + 4xy + 15y2 )] = 2m(2x + 3y)m−1 (9x2 + 4xy + 15y2 )
∂x
+ (2x + 3y)m (18x + 4y)
Hence, exact if
3m(10x2 +6xy+6y2 )+(2x+3y)(6x+12y) = 2m(9x2 +4xy+15y2 )+(2x+3y)(18x+4y)
Comparing coefficients of x2 gives m = 2. Let
∂f
= (2x + 3y)2 (10x2 + 6xy + 6y2 ) = 40x4 + 144x3 y + 186x2 y2 + 126xy3 + 54y4
∂x
⇒ f(x, y) = 8x5 + 36x4 y + 62x3 y2 + 63x2 y3 + 54xy4 + c(y)
∂f
= 36x4 + 124x3 y + 99x2 y2 + 216xy3 + c (y)
∂y
⇒ c (y) = 9y2 × 15y2 ⇒ c(y) = 27y5 + k
Hence, f(x, y) = 8x5 + 36x4 y + 62x3 y2 + 63x2 y3 + 5xy4 + 27y5 + k .
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
137
Exercises 3.2.2
21
grad f = (2xyz2 , x2 z2 , 2x2 yz) .
At (1,2,3), grad f = (36, 9, 12) = 3(12, 3, 4) .
21(a)
Unit vector in direction of (−2, 3, −6) is √(−2,3,−6) = 17 (−2, 3, −6) .
(4+9+36)
Directional derivative of f in direction of (−2, 3, −6) at (1,2,3) is
3(12, 3, 4) · (−2, 3, −6)/7 = −117/7
21(b)
Maximum rate of change is |grad f| = 3 (144 + 9 + 16) = 39 and is in
the direction of grad f, that is, (12,3,4)/13.
22(a)
∇(x2 + y2 − z) = (2x, 2y, −1)
22(b)
∇ z tan
−1
y
x
=
−zy
zx
−1 y
,
,
tan
x2 + y 2 x2 + y 2
x
22(c)
∇
e−x−y+z
x3 + y 2
−e−x−y+z
1 3x2 e−x−y+z −e−x−y+z
−
,
2 (x3 + y2 )3/2
x3 + y 2
x3 + y 2
1 2ye−x−y+z e−x−y+z
−
,
2 (x3 + y2 )3/2
x3 + y 2
e−x−y+z
3 2
3
2
3
2
3
2
−x − y − x , −x − y − y, x + y
= 3
2
(x + y2 )3/2
=
22(d)
∇(xyz sin π(x + y + z)) =(yz sin π(x + y + z) + πxyz cos π(x + y + z),
xz sin π(x + y + z) + πxyz cos π(x + y + z),
xy sin π(x + y + z) + πxyz cos π(x + y + z))
c Pearson Education Limited 2011
138
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
23
grad (x2 + y2 − z) = (2x, 2y, −1) .
At (1,1,2), grad f = (2, 2, −1) .
1
3 (2, 2, −1) .
1
1
3 (2, 2, −1) = 3 (4 + 4
Unit vector in the direction of (4, 4, −2) is
Directional derivative is (2, 2, −1) ·
24
+ 1) = 3.
∇(xy2 − 3xz + 5) = (y2 − 3z, 2xy, −3x) .
At (1, −2, 3) , grad f = (−5, −4, −3) .
√
Unit vector in the direction of grad f is (−5, −4, −3)/ 50 .
√
Unit normal to surface xy2 − 3xz + 5 = 0 at (1, −2, 3) is (5, 4, 3)/ 50 .
25(a)
r=
x2 + y 2 + z 2
∇r =
x
y
z
,
,
x2 + y 2 + z 2
x2 + y 2 + z 2
x2 + y 2 + z 2
1
(x, y, z)
r
r
=
r
=
25(b)
−x
1
−y
−z
=
∇
,
,
r
(x2 + y2 + z2 )3/2 (x2 + y2 + z2 )3/2 (x2 + y2 + z2 )3/2
r
=− 3
r
26
∂φ
= 2xy + z2 ⇒ φ(x, y, z) = x2 y + xz2 + f(y, z)
∂x
∂φ
∂f
= x2 + z ⇒ x2 + z = x2 +
⇒ f(y, z) = zy + g(z)
∂y
∂y
∂φ
dg
= y + 2xz ⇒ y + 2xz = 2xz + y +
∂z
dz
Hence,
dg
dz
= 0 ⇒ g(z) = c, a constant.
Hence, φ(x, y, z) = x2 y + xz2 + zy + c.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
27
φ(x, y, z) = x2 y − 3xyz + z3
grad φ = (2xy − 3yz, x2 − 3xz, −3xy + 3z2 )
At (3,1,2), grad φ = (0, −9, 3) .
√
Unit vector in direction of (3, −2, 6) = (3, −2, 6)/ 49 .
Directional derivative at (3,1,2) in the direction of (3, −2, 6) is
(0, −9, 3) · (3, −2, 6)/7 = 36/7
28
∇(x2 + y2 + z2 − 9) = (2x, 2y, 2z) .
At (2, −1, 2) , grad (x2 + y2 + z2 − 9) = (4, −2, 4) .
Unit normal to surface at (2, −1, 2) is (2, −1, 2)/3.
∇(x2 + y2 − z − 3) = (2x, 2y, −1).
At (2, −1, 2) , grad (x2 + y2 − z − 3) = (4, −2, −1) .
√
Unit normal to surface at (2, −1, 2) is (4, −2, −1)/ 21 .
Let angle between normals be θ, then
cos θ =
(2, −1, 2) (4, −2, −1)
√
·
3
21
8
⇒ cos θ = √ , hence, θ = 54.41◦
3 21
29(a)
∇(x2 + 2y2 + 3z2 − 6) = (2x, 4y, 6z) .
At (1,1,1), grad f = (2, 4, 6) , so tangent plane at (1,1,1) is
(1, 2, 3) · (x − 1, y − 1, z − 1) = 0 i.e. x + 2y + 3z = 6
and normal line is
29(b)
y−1
z−1
x−1
=
=
1
2
3
∇(2x2 + y2 − z2 + 3) = (4x, 2y, −2z)
At (1,2,3), grad f = (4, 4, −6) , so the tangent plane at (1,2,3) is
(2, 2, −3) · (x − 1, y − 2, z − 3) = 0 i.e. 2x + 2y − 3z = −3
and the normal line is
y−2
3−z
x−1
=
=
2
2
3
c Pearson Education Limited 2011
139
140
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
29(c)
∇(x2 + y2 − z − 1) = (2x, 2y, −1)
At (1,2,4), grad f = (2, 4, −1) , so that the tangent plane is
(2, 4, −1) · (x − 1, y − 2, z − 4) = 0 i.e. 2x + 4y − z = 6
and the normal line is
x−1
y−2
z−4
=
=
2
4
−1
30
The change Δr in the vector r can be resolved into the three directions ur ,
uθ , uφ . Thus,
Δr = Δrur + rΔθuθ + r sin θΔφuφ
Hence,
f(r + Δr) − f(r)
Δr→0
|Δr|
1 ∂f
1 ∂f
∂f
ur +
uθ +
uφ
=
∂r
r ∂θ
r sin θ ∂φ
grad f = lim
Exercises 3.3.2
31(a)
div (3x2 y, z, x2 ) = 6xy + 0 + 0 = 6xy
31(b)
div (3x + y, 2z + x, z − 2y) = 3 + 0 + 1 = 4
32
div F = 2y2 − 2yz3 + 2yz − 3xz2 .
At (−1, 2, 3) , div F = −61.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
33
141
∇(a · r) = ∇(a1 x + a2 y + a3 z)
= (a1 , a2 , a3 ) = a
∂
∂
∂
(x, y, z)
(a · ∇)r = a1 , a2 , a3
∂x
∂y
∂z
= (a1 , a2 , a3 ) = a
a(∇ · r) = a(1 + 1 + 1) = 3a
34
∇·v=
∂
since
∂x
1 x2
− 3
r
r
x
x2 + y 2 + z 2
+
=
1 y2
− 3
r
r
+
1
x2 + y 2 + z 2
1 z2
−
r r3
−
x.(2x)
1
2
2 (x + y2 + z2 )3/2
3 x2 + y 2 + z 2
2
Hence, ∇.v = −
=
r
r3
r
− 12 .(2x)
− 12 (2y)
− 12 (2z)
2
=2
,
,
∇
r
(x2 + y2 + z2 )3/2 (x2 + y2 + z2 )3/2 (x2 + y2 + z2 )3/2
2
2r
= − 3 (x, y, z) = − 3
r
r
35
div F = 4xy2 + 9xy2 + λxy2 = (4 + 9 + λ)xy2
div F = 0 ⇒ λ = −13
36
In spherical polar coordinates, an element of volume has side Δr in the ur
direction, rΔθ in the uθ direction and r sin θΔφ in the uφ direction.
The total flow out of the elementary volume is
∂
∂
∂
(v · ur r2 sin θΔθΔφ)Δr + (v · uθ r sin θΔφΔr)Δθ +
(v · uφ rΔθΔr)Δφ
∂r
∂θ
∂φ
+ terms of order |Δr|2
c Pearson Education Limited 2011
142
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Dividing by the volume of the element, r2 sin θΔθΔφΔr , and proceeding to the
limit, we obtain
div v =
∂
1 ∂
1
1 ∂ 2
(r
(r
sin
θv
(vφ )
v
)
+
)
+
r
θ
r2 ∂r
r sin θ ∂θ
r sin θ ∂φ
37
div
x y z
r
=
div
, ,
r3
r3 r3 r3
3x2
1
3y2
1
3z2
1
= 3− 5 + 3− 5 + 3− 5
r
r
r
r
r
r
3
(x2 + y2 + z2 )
= 3 −3
=0
r
r5
Exercises 3.3.4
38
39
40
i
∂
curl v = ∂x
3xz2
i
∂
curl v = ∂x
yz
j
∂
∂y
xz
i
∂
curl v = ∂x
2x + yz
j
k
∂
∂y
∂
∂z
−yz
x + 2z
= (y, 6xz − 1, 0)
k
∂
∂z = (x − x, y − y, z − z) = 0
xy
∂
∂
= (0, 0, 0) = 0
∂y
∂z
2y + zx 2z + xy
∂f ∂f ∂f
∂f
grad f =
,
,
⇒
= 2x + yz
∂x ∂u ∂z
∂x
∂f
= 2x + yz ⇒ f(x, y, z) = x2 + xyz + g(y, z)
∂x
∂f
∂f
∂g
= 2y + zx and
= xz +
⇒ g(y, z) = y2 + h(z)
∂y
∂y
∂y
∂f
∂f
dh
= 2z + xy and
= xy +
⇒ h(z) = z2 + c
∂z
∂z
dz
j
k
Hence, f(x, y, z) = x2 + y2 + z2 + xyz + C.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
41
i
j
k
∂
∂
∂
∇ × (fv) =
= (x, 5x3 − 2y, z)
∂x
∂y
∂z
zx3 − zy 0 −x4 + xy
i
j
k
∂
∂
∂
f(∇ × v) = (x3 − y) ∂x
∂y
∂z
z
0 −x
= (x3 − y)(0, 2, 0) = (0, 2x3 − 2y, 0)
(∇f) × v = (3x2 , −1, 0) × v
i
j
k
2
= 3x −1 0 = (x, 3x3 , z)
z
0 −x
42
i
∂
∇×F=
∂x
4xy + az3
j
∂
∂y
∂
= (c − 3, 3az2 − 6z2 , 2bx − 4x)
∂z
6xz2 + cy
k
bx2 + 3z
∇ × F = 0 ⇒ c = 3, a = 2, b = 2
∂φ ∂φ ∂φ
,
,
grad φ =
∂x ∂y ∂z
∂φ
= 4xy + 2z3 ⇒ φ(x, y, z) = 2x2 y + 2xz3 + f(y, z)
∂x
∂φ
∂φ
∂f
∂f
= 2x2 + 3z and
= 2x2 +
⇒
= 3z
∂y
∂y
∂y
∂y
Hence, f(y, z) = 3yz + g(z) .
dg
∂φ
dg
∂φ
= 6xz2 + 3y and
= 6xz2 + 3y +
⇒
=0
∂z
∂z
dz
dz
Hence, φ(x, y, z) = 2x2 y + 2xz3 + 3yz + C.
c Pearson Education Limited 2011
143
144
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
i
1
1 ∂
ω = curl u = ∂x
2
2
−y
43
=
j
∂
∂y
x
k
∂
∂z
xyz
1
(xz, −yz, 2)
2
At (1,3,2), ω = 12 (2, −6, 2) = (1, −3, 1)
⇒
ω| =
|ω
√
11
44
div v = a + d
i
j
∂
∂
curl v = ∂x
∂y
ax + by cx + dy
div v = 0 ⇒ a = −d
k
∂
∂z = (0, 0, c − b)
0
curl v = 0 ⇒ c = b
v = (ax + by)i + (bx − ay)j
= grad φ
∂φ
∂φ
= ax + by and
= bx − ay
∂x
∂y
1
⇒ φ(x, y) = ax2 + bxy + f(y)
2
⇒
∂φ
1
= bx + f (y) ⇒ f (y) = −ay ⇒ f(y) = − ay2 + K
∂y
2
Hence, φ(x, y) =
45
1 2
1
ax + bxy − ay2 + K.
2
2
In spherical polar coordinates, an element of volume has side Δr in the ur
direction, rΔθ in the uθ direction and r sin θΔφ in the uφ direction.
Setting v · ur = vr , v · uθ = vθ and v · uφ = vφ , we see that the circulation around
the ur direction is
∂
∂
(vφ r sin θΔφ)Δθ −
(vrΔθ)Δφ + terms of order Δθ2 etc.
∂θ
∂φ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
145
The area around which this circulation takes place is r2 sin θΔθΔφ , so, proceeding
to the limit we have
(curl v) · ur =
Similarly
(curl v) · uθ =
and
(curl v) · uφ =
∂
∂
(vφ r sin θ) −
(vθ r) /(r2 sin θ)
∂θ
∂φ
∂
∂
(vr ) − (r sin θvφ ) /(r2 sin θ)
∂φ
∂r
∂
∂
(rvθ ) − (vr ) /r
∂r
∂θ
Hence, the result.
Exercises 3.3.6
46
∂g ∂g ∂g
, ,
grad g =
∂x ∂y ∂z
dg ∂r dg ∂r dg ∂r
,
,
=
dr ∂x dr ∂y dr ∂z
dg x y z
=
, ,
since r2 = x2 + y2 + z2
dr r r r
1 dg
=
r
r dr
div [(u × r)g] = (u × r) · grad g + g∇ · (u × r)
∇ · (u × r) = r · (∇ × u) − u · (∇ × r)
∇×r=0
⇒
But (u × r) is perpendicular to grad g =
(3.19d)
(3.19f)
∇ · (u × r) = r · curl u
1 dg
r, so
r dr
(u × r) · grad g = 0
Hence, div ((u × r)g) = r · curl u.
47
φ(x, y, z) = x2 y2 z3 ,
F (x, y, z) = (x2 y, xy2 z, −yz2 )
c Pearson Education Limited 2011
146
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
47(a)
∇2 φ = 2y2 z3 + 2x2 z3 + 6x2 y2 z
47(b)
grad div F = grad (2xy + 2xyz − 2yz)
= (2y + 2yz, 2x + 2xz − 2z, 2xy − 2y)
47(c)
i
∂
curl F = ∂x
x2 y
j
∂
∂y
xy2 z
k
∂
∂z
−yz2
= i(−z2 − xy2 ) + j(0) + k(y2 z − x2 )
i
j
k
∂
∂
∂
curl (curl F) =
∂x
∂y
∂z
−z2 − xy2 0 y2 z − x2
= i(2yz) + j(2x − 2z) + k(2xy)
48
grad [(r · r)(a · r)] = [grad (r · r)](a · r) + (r · r)grad (a · r)
(3.19b)
= 2r(a · r) + (r · r)a
div {grad [(r · r)(a · r)]} = 2div [r(a · r)] + div [(r · r)a]
= 2{[div r](a · r) + r · grad (a · r)}
+ [grad (r · r)] · a + (r · r)div a
= 2{3(a · r) + r · a} + 2r · a + 0
= 10(r · a)
c Pearson Education Limited 2011
(3.19d)
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
49
v = x3 yi + x2 y2 j + x2 yzk
∇2 v = 6xyi + 2(x2 + y2 )j + 2yzk
grad div v = grad (3x2 y + 2x2 y + x2 y) = grad (6x2 y) = 12xyi + 6x2 j
i
j
k
∂
∂
∂
2
2
3
curl v = ∂x
∂y
∂z = x zi − 2xyzj + (2xy − x )k
x3 y x2 y2 x2 yz
i
j
k
∂
∂
∂
curl(curl v) = ∂x
∂y
∂z
x2 z −2xyz 2xy2 − x3
= (4xy + 2xy)i + (x2 − 2y2 + 3x2 )j + (−2yz)k
grad div v − curl curl v = 6xyi + 2(x2 + y2 )j + 2yzk as required.
50
i
u × v = 0
xy
j
xy
0
k
xz = xy2 zi + x2 yzj − x2 y2 k
yz
div (u × v) = y2 z + x2 z = (x2 + y2 )z
i
j
k
∂
∂
∂
curl u = ∂x
∂y
∂z = 0i − zj + yk
0 xy xz
v · curl u = y2 z
i
∂
curl v = ∂x
xy
j
∂
∂y
0
k
∂
∂z = zi + 0j − xk
yz
u · curl v = −x2 z
⇒ v · curl u − u · curl v = (x2 + y2 )z
c Pearson Education Limited 2011
147
148
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
i
∂
curl (u × v) = ∂x
xy2 z
j
∂
∂y
x2 yz
∂
= −3x2 yi + 3xy2 j + 0k
∂z
−x2 y2
k
udiv v = (xyj + xzk)(y + y) = 2xy2 j + 2xyzk
vdiv u = (xyi + yzk)(x + x) = 2x2 yi + 2xyzk
∂
∂
+ yz
(xyj + xzk) = xy2 j + 2xyzk
(v · ∇)u = xy
∂x
∂z
∂
∂
(xyi + yzk) = x2 yi + 2xyzk
(u · ∇)v = xy + xz
∂y
∂z
2
2
⇒ udiv v − vdiv u + (v · ∇)u − (u · ∇)v = −3x yi + 3xy j + 0k
51(a)
r
1
=− 3
grad
r
r
div
r
1
1
1
= −div 3 = − 3 div r − r · grad
grad
r
r
r
r3
3
=− 3 −r·
r
−3r
r5
=−
3
3r2
+
=0
r3
r5
51(b)
curl
−k × r
1
1
= curl
= curl (yi − xj) 3
k × grad
3
r
r
r
1
= [curl (yi − xj)] 3 + grad
r
i
∂
= ∂x
y
j
∂
∂y
−x
1
r3
× (yi − xj)
k
3r
∂ 1
∂z r3 − r5 × (yi − xj)
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
149
i
j
k
3
1
z
= (0i + 0j − 2k) 3 − 5 x y
r
r
y −x 0
2k
3
= − 3 + 5 (−xzi − yzj + (x2 + y2 )k)
r r
−z
k·r
1
= grad − 3
= grad
grad k · grad
r
r
r3
1
1
− (grad z) 3
= −zgrad
3
r
r
1
3r
= −z − 5 − 3 k
r
r
curl
1
3k
1
3
+ grad k · grad
= − 3 + 5 (−xzi − yzj
k × grad
r
r
r
r
+ (x2 + y2 )k)
3
+ 5 (xzi + yzj + z2 k)
r
=0
52(a)
grad
A·r
r3
= grad
A
r3
·r
A
A
A
A
+ (r · ∇) 3 +
· ∇ r (3.19c)
= 3 × curl r + r × curl
r
r3
r
r3
1
A
1
× A + A(r · ∇) 3 + 3
= 0 + r × grad
3
r
r
r
A
3r
3r2
=r× − 5 ×A +A − 5 + 3
r
r
r
Now, a × (b × c) = (a · c)b − (a · b)c
so
3r
r× A× 5
r
Hence
grad
=
A·r
r3
3r
r· 5
r
=
A − (A · r)
A
(A · r)
− 3r
3
r
r5
c Pearson Education Limited 2011
3r
r5
150
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
52(b)
curl
A×r
r3
A
A
A
A
= (r · ∇) 3 − r ∇ · 3 −
· ∇ r + 3 (∇ · r)
3
r
r
r
r
A
A·r
A
3A
− 3 +3 3
= − 3 − r −3 5
r
r
r
r
=
3
A
(A · r)r − 3
5
r
r
(A × r) × r = (A · r)r − (r · r)A
(A · r)r = (A × r) × r + Ar2
3
A×r
3Ar2
A
= 5 [(A × r) × r] + 5 − 3
curl
3
r
r
r
r
=2
53(a)
A
3
+ 5 (A × r) × r
3
r
r
i
∂
Δ × r = curl r = ∂x
x
53(b)
(a · ∇)r =
k
∂
∂z = 0
z
j
∂
∂y
y
∂
∂
∂
+ a2
+ a3
a1
∂x
∂y
∂z
(xi + yj + zk)
= a1 i + a2 j + a3 k = a
53(c)
∇ × [(a · r)b − (b · r)a] = ∇ × [(a × b) × r]
= (a × b)(∇ · r) − [(a × b) · ∇]r
= 3(a × b) − a × b
= 2a × b
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
151
53(d)
∇ · [(a · r)b − (b · r)a] = ∇ · [(a × b) × r]
= (a × b) · (∇ × r)
= (a × b) · (0) = 0
54
1 ∂f
1 ∂f
∂f
ur +
uθ +
uφ
(Exercise 30)
∂r
r ∂θ
r sin θ ∂φ
1 ∂ sin θ ∂f
1
∂
1 ∂f
1 ∂
2 ∂f
r
+
+
∇ · (∇f) = 2
r ∂r
∂r
r sin θ ∂θ
r ∂θ
r sin θ ∂φ r sin θ ∂φ
∇f =
(using Exercise 36)
1
∂
∂f
∂2 f
1 ∂
1
2 ∂f
r
+ 2 2
sin θ
= 2
+ 2
r ∂r
∂r
r sin θ ∂θ
∂θ
r sin θ ∂φ2
55
1
div H =
c
div
∂Z
curl
∂t
=0
(3.22)
div E = div (curl curl Z) = 0
1 ∂E
curl H =
becomes
c ∂t
∂Z
1
curl H = curl curl
c
∂t
1 ∂E
1 ∂
1
∂Z
=
(curl curl Z) = curl curl
c ∂t
c ∂t
c
∂t
curl E = curl curl curl Z
1
∂2 Z
1 ∂H
= curl
c ∂t
c
∂t2
1
∂2 Z
1 ∂H
⇒ curl curl curl Z = − curl
curl E = −
c ∂t
c
∂t2
1 ∂2 Z
⇒ curl curl Z = −
c ∂t2
1 ∂2 Z
⇒ grad (div Z) − ∇2 Z = −
c ∂t2
Hence, ∇2 Z =
1 ∂2 Z
when div Z = 0
c ∂t2
c Pearson Education Limited 2011
(3.22)
152
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 3.4.2
56
B
24
√
(2 x) 1 + 1/x dx
24
√
4
(x + 1)3/2
2 x + 1 dx =
3
y ds =
A
3
=
3
=
57
24
3
4
[125 − 8] = 156
3
B
[2xy dx + (x2 − y2 ) dy] = I
∂S
A
x2 + y2 = 1 ⇒ x dx = −y dy
y=1
[−2y2 + (1 − 2y2 )] dy
I=
y=0
1
4
(1 − 4y ) dy = y − y3
3
1
=−
2
=
0
0
1
3
58
r = (t3 , t2 , t)
dr = (3t2 , 2t, 1) dt
1
[(2yz + 3x2 )(3t2 ) + (y2 + 4xz)2t + (2z2 + 6xy)1] dt
V· dr =
0
C
1
[6t5 + 9t8 + 2t5 + 8t5 + 2t2 + 6t5 ] dt
=
0
1
(22t5 + 9t8 + 2t2 ) dt =
=
0
2
16
11
+1+ =
3
3
3
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
59
153
A = (2y + 3, xz, yz − x) .
A· dr where r = (2t2 , t, t3 ) and dr = (4t, 1, 3t2 ) dt
59(a)
C
1
[(2t + 3)4t + (2t5 )1 + (t4 − 2t2 )3t2 ] dt
A· dr =
C
0
1
8 6 1 3
− + +
3 5 3 7
(12t + 8t2 − 6t4 + 2t5 + 3t6 ) dt = 6 +
=
0
288
=
35
59(b)
Q
A · dr +
A· dr =
P
C
R
S
A · dr +
Q
A · dr
R
where P = (0, 0, 0) , Q = (0, 0, 1) , R = (0, 1, 1) , S = (2, 1, 1)
(using straight lines)
On PQ
A = 3i
(x = y = 0)
On QR
A = (2y + 3)i + yk
(x = 0, z = 1) r = yj + k
On RS
A = 5i + xj + (1 − x)k (y = 1, z = 1) r = xi + j + k
3i · k dz +
A· dr =
C
1
0
1
2
[(2y + 3)i + yk] · j dy +
0
r = zk
[5i + xj + (1 − x)k] · i dx
0
= 10
since i · k = 0 etc.
59(c)
S
A · dr
A· dr =
C
P
where C is a straight line, P = (0, 0, 0) and S = (2, 1, 1) .
Parametrically, straight line is r = (2, 1, 1)t , so
1
[(2t + 3)i + 2t2 j + (t2 − 2t)k] · (2i + j + k) dt
A· dr =
0
C
1
1
[4t + 6 + 2t + t − 2t] dt =
2
=
0
2
(2t + 6 + 3t2 ) dt
0
= [1 + 6 + 1] = 8
c Pearson Education Limited 2011
154
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
60
F is conservative if there exists a φ such that
∂φ ∂φ ∂φ
,
,
F = (y cos x + z , 2y sin x − 4, 3xz + z) = −
∂x ∂y ∂z
2
3
2
Such a φ is readily determined giving
1 2
2
3
F = −grad 4y − y sin x − xz − 2 z
Hence, work done in moving an object is
1
F· dr = − 4y − y sin x − xz − z2
2
C
2
1 2 3 3
4t , 8t
2
F· dr =
0
C
, so that dr =
61(a) Curve is r = t,
1 2 3 3
2 3 4
F = 3t , 4 t − 4 t , 8 t
(0,1,−1)
1
− [−5 − 4π − 2] = 4π + 10.5
2
= 4−
(π/2,−1,2)
3
1,
1 9 2
2 t, 8 t
dt and
3
1
27
3t2 + t5 − t3 + t5 dt
8
8
64
3
1
9 6
t
= t + t6 − t4 +
48
32
128
2
3
=8+4−
61(b)
0
1 9
+ = 16
2 2
Curve is r = (2t, t, 3t), 0 ≤ t ≤ 1, so that
dr = (2, 1, 3) dt and F = (12t2 , 12t2 − t, 3t) .
1
(24t2 + 12t2 − t + 9t) dt
F· dr =
0
C
1
(36t2 + 8t) dt = 12 + 4 = 16
=
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
61(c) No.
If F is conservative, there is a function U(x, y, z) such that
F = −grad U
Test for existence of U: F · dr has to be an exact differential
∂
∂
(3x2 ) =
(2xz − y)
∂y
∂x
Hence, not exact and F is not conservative.
62
F = (3x2 − y, 2yz2 − x, 2y2 z)
∂
∂
(3x2 − y) = −1 =
(2yz2 − 1)
∂y
∂x
∂
∂
(3x2 − y) = 0 =
(2y2 z)
∂z
∂x
∂
∂
(2y2 z) = 4yz =
(2yz2 − x)
∂y
∂z
Hence, conservative and F = −grad U where U = −x3 + xy − y2 z2 .
div F = 6x + 2z2 + 2y2 = 0, hence not solenoidal.
C
63
155
(1,2,3)
F· dr = x3 − xy + y2 z2 (0,0,0) = 1 − 2 + 36 = 35
F = (2t3 , −t3 , t4 ) , r = (t2 , 2t, t3 ) , dr = (2t, 2, 3t2 ) dt
i
F × dr = 2t3
2t
j
−t3
2
k
4
t dt
2
3t
= [(−3t5 − 2t4 )i + (−4t5 )j + (4t3 + 2t4 )k] dt
1
[(−3t5 − 2t4 )i − 4t5 j + (4t3 + 2t4 )k] dt
F× dr =
C
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
156
=
=−
4
2
1 2
i− j+ 1+
k
− −
2 5
6
5
2
7
9
i− j+ k
10
3
5
64
i
A × B = 3x + y
2
j
k
−x y − z = (3y − 3z − x)i + (y − 2z − 3x)j + (−3y − 7x)k
−3
1
r = (2 cos θ, 2 sin θ, 0)
dr = (−2 sin θ, 2 cos θ, 0) dθ
On circle, z = 0 and
A × B = (6 sin θ − 2 cos θ)i + (2 sin θ − 6 cos θ)j − (6 sin θ + 14 cos θ)k
i
(A × B) × dr = 6 sin θ − 2 cos θ
−2 sin θ
k
−6 sin θ − 14 cos θ dθ
0
2π
(A × B) × dr =
C
j
2 sin θ − 6 cos θ
2 cos θ
{(6 sin θ + 14 cos θ)2 cos θi + (6 sin θ + 14 cos θ)2 sin θj
0
+ [(6 sin θ − 2 cos θ)(2 cos θ) + (2 sin θ − 6 cos θ)(2 sin θ)]} dθ
2π
2π
sin θ cos θ dθ =
0
0
2π
2π
=0
0
2π
2
cos2 θ dθ = π
sin θ dθ =
0
1
1
sin 2θ dθ =
cos 2θ
2
4
0
(A × B) × dr = 28πi + 12πj + 0k
C
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 3.4.4
65(a)
3
2
3
1 2 2 1 3
x y + xy
2
3
xy(x + y) dy dx =
0
1
0
3
2
dx
1
1 2
1
x (4 − 1) + x(8 − 1) dx
2
3
=
0
3
1 3 7 2
x + x
=
2
6
0
27 21
+
= 24
=
2
2
65(b)
3
5
5
2
x y dy dx =
2
3
2
x dx
1
2
y dy
1
3
5
1 2
1 3
=
x
y
3
2 2
1
1
1
= (27 − 8) (25 − 1)
3
2
= 76
65(c)
1
2
−1
66
2
1
1
2x y + y3
(2x + y ) dy dx =
dx
3
−2
−1
−2
1
16 32
16
2
dx =
+
= 16
=
8x +
3
3
3
−1
2
2
1
2
2
0
2
2
1
dy
x2 dx
y
1
0
8
8
= ln 2
= (ln 2)
3
3
x2
dx dy =
y
2
c Pearson Education Limited 2011
157
158
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
67
1
1−x
2
1
2
(x + y ) dy dx =
0
0
1
x y + y3
3
0
1
=
0
68(a)
2
2x
dx
1
2
1
x2
x
1 4
1
x − (1 − x)4
4
12
1
1
− −
=
12
6
dy
+ y2
2x
y
1
=
dx
tan−1
x
x x
1
2
1
=
(tan−1 2 − tan−1 1) dx
1 x
= (tan−1 2 − tan−1 1)[ln x]21
1
−1
ln 2
= tan
3
68(b)
dx
0
1−x
(x + y) dy
0
1
=
0
1
=
0
1
xy + y2
2
1−x
dx
0
1
x(1 − x) + (1 − x)2
2
x3
1
x2
−
− (1 − x)3
2
3
6
1
1 1 1
− +
=
=
2 3 6
3
dx
0
1
x2 (1 − x) + (1 − x)3 dx
3
1 3
=
x −
3
1 1
=
−
3 4
1−x
2
dx
1
= −
0
c Pearson Education Limited 2011
1
0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1
68(c)
dx
0
√
√
1−x2
x−x2
159
y
1
1 − x2 − y2
√
1
dy
1
Part circle
x 2 + y2 = 1
1−x2
y
sin−1 √
dx
Part circle
1 − x2 √x−x2
0
x2 − x + y 2 = 0
1
2
x−x
sin−1 1 − sin−1
dx
=
1 − x2
0
1
0
1/2
1
x
x
−1
−1
=
dx
sin 1 − sin
1+x
0
1
x
π
−1
= − x sin
2
1+x 0
1 √
x
1
dx
+
2 0 (1 + x)
√
√ 1
1
π
= − sin−1 √ + x − tan−1 x 0 , using substitution x = tan2 θ
2
2
π
π
=1
= + 1−
4
4
=
69
y
2
(1,2)
(2,1)
1
0
1
2
x
1
sin π(x + y) dx dy
2
2
1
x
3−x
1
1
=
dx
sin π(x + y) dy +
dx
sin π(x + y) dy
2
2
0
x/2
1
x/2
1
2
x
3−x
1
1
2
2
− cos π(x + y)
− cos π(x + y)
=
dx +
dx
π
2
π
2
0
1
x/2
x/2
2
1
3
2
3
2
3π
2
2
cos πx − cos πx dx +
cos πx − cos
dx
=
π
4
π
π
4
π
2
0
1
c Pearson Education Limited 2011
160
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1
2
3
8
3
8
sin πx − 2 sin πx +
sin πx
=
2
2
3π
4
π
3π
4
0
8
3
3π
3π
8
8
= 2 sin π + 2 sin
− 2 sin
3π
4
3π
2
3π
4
8
=− 2
3π
70(a)
1
1
dx
x
π/2
70(b)
0 π/2
=
cos 2y
1
0
1 − k2 sin2 x dx
0 π/2
dx
0
y
dy
1
xy
dy
1 + y4
y
1
xy
dy
dx
=
1 + y4
0
0
1 1 2 y
x y
2
=
dy
1 + y4 0
0
1 1 3
y
1 2
2
.
dy =
1 + y4
=
8 1
1 + y4
0
1 √
= ( 2 − 1)
4
0
2
cos 2y
1 − k2 sin2 x dy
x
1
=
dx
sin 2y 1 − k2 sin2 x
2
0
x
π/2
=
− sin x cos x 1 − k2 sin2 x dx
0 −1
1
1 + k2 t dt
=
2
0
(Let t = − sin2 x , then
π/2
π/2
dt = −2 sin x cos x dx )
−1
1
2 3/2
(1 + k t)
=
3k2
0
1
2 3/2
= 2 ((1 − k ) − 1)
3k
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1
71
dy
√
0
1
y
1
x2
dx
=
1
=
√
0
0
dx
y(1 + x2 )
√
2
1
√
[2 y ]x0 dx
1 + x2
√
2x
dx
1 + x2
0
1
=
1
1
√ dy
2
y
1+x
0
√
= [2 1 + x2 ]10 = 2( 2 − 1)
72
1
√
x−x2
dx
0
0
x
dy
x2 + y 2
Equation of circle in polar coordinates is r = cos θ
√
1
dx
0
0
x−x2
x
x2
+
y2
π/2
cos θ
dy =
0
0
π/2
r cos θ
r2 cos2 θ + r2 sin2 θ
r dr dθ
cos θ
=
(cos θ)r dr dθ
0
0
π/2
=
0
π/2
=
0
=
1
cos θ r2
2
cos θ
dθ
0
1
1 2
cos3 θ dθ = .
2
2 3
1
3
c Pearson Education Limited 2011
π/2
cos θ dθ
0
161
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
162
73
√
1
1−x2
dx
0
0
x+y
dy
x2 + y 2
Change to polar coordinates
1
√
1−x2
dx
0
0
x+y
dy =
x2 + y 2
π/2
1
r cos θ + r sin θ
r dr dθ
r
0
0
π/2
1
=
(cos θ + sin θ) dθ
r dr
0
π/2
= [sin θ − cos θ]0
1 2
r
2
0
1
0
=1
x+y
dx dy
+ y2 + a2
over first quadrant of circle
74
x2
using polar coordinates
π/2
a
r cos θ + r sin θ
r dr dθ
r2 + a2
0
0
π/2
a
r2
=
(cos θ + sin θ) dθ
dr
2
2
0
0 r +a
a
a2
1− 2
dr
=2
r + a2
0
r a
π
= 2a 1 −
= 2 r − a tan−1
a 0
4
x+y
dx dy =
2
x + y2 + a2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
75
Using polar coordinates, parabola becomes
r2 sin2 θ = 4 − 4r cos θ
r2 = 4 − 4r cos θ + r2 cos2 θ
= (2 − r cos θ)2
r = (2 − r cos θ),
2
=
1 + cos θ
positive root because r = 2 at θ =
x2 − y 2
dx dy =
x2 + y 2
π/2
2
1+cos θ
(cos2 θ − sin2 θ)r dr dθ
0
0
π/2
2(cos2 θ − sin2 θ)
=
0
π/2
=
0
1
dθ
(1 + cos θ)2
2(2 cos2 θ − 1)
dθ
(1 + cos θ)2
= (6π − 20)/3 = −0.3835
(use the substitution t = tan 12 θ).
76
Circles are r = a cos θ, r = b sin θ and intersect at θ = tan−1 ab .
(x2 + y2 )2
dx dy =
(xy)2
tan−1
a
b
π/2
1
r dr dθ
sin θ cos2 θ
2
0
0
b sin θ
a cos θ
+
tan−1
tan−1
a
b
tan−1
a
b
=
0
=
0
0
π/2
1 b2 sin2 θ
1 a2 cos2 θ
dθ
+
dθ
2 sin2 θ cos2 θ
2 sin2 θ cos2 θ
tan−1 a
b
π/2
1 2
1 2
2
b sec θ dθ +
a cosec2 θ dθ
a 2
2
−1
tan
b
1 2
=
b tan θ
2
=
a
b
1
r dr dθ
sin θ cos2 θ
2
tan−1
0
a
b
1
+ − a2 cot θ
2
π/2
tan−1
a
b
1
1
ab + ab = ab
2
2
c Pearson Education Limited 2011
π
2
163
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
164
Exercises 3.4.6
77
[sin y dx + (x − cos y) dy]
C1
=
1
1
1
sin πx + (x − cos πx) π dx
2
2
2
0
0
1
+
sin π dx
2
1 0
+
[− cos y] dy
π/2
1
1
1
2
1
= − cos πx − πx2 − sin πx
π
2
4
2
0
0
1
+ x sin π + [− sin y]0π/2
2 1
2 1
= −1 + + π − 1 + 1
π 4
2 π
= −1 + +
π 4
[1 − cos y] dx dy
[sin y dx + (x − cos y) dy] =
C
A
1
π/2
dx
=
0
1
2 πx
(1 − cos y) dy
1
π π
− x − 1 + sin πx dx
=
2
2
2
0
2
π
2
π π
= − −1+ = −1+
2
4
π
4
π
1
78
[(x2 y − y) dx + (x + y2 ) dy]
(1 − x2 + 1) dx dy
=
A
2
(2 − x2 ) dy
dx
=
0 2
x
0
1
=
(2x − x ) dx = x − x4
4
0
=4−4=0
3
2
2
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
165
79
(xy dx + x dy)
C
1
(x3 + 2x2 ) dx
0
0
1 1/2
3/2
x + x
dx
+
2
1
11
2 1
1 2
+
=
= + −
4 3
5 3
60
=
(1 − x) dx dy
(xy dx + x dy) =
C
A
1
√
x
(1 − x) dy
dx
=
x2
0
1
√
(1 − x)( x − x2 ) dx
1
√
=
0
=
x − x3/2 − x2 + x3 dx
0
=
11
2 2 1 1
− − + =
3 5 3 4
60
80
(e − 3y ) dx + (e + 4x ) dy =
x
2
y
2
c
(8x + 6y) dx dy
A
2π
=
2
dθ
0
(8r cos θ + 6r sin θ)r dr dθ
0
2π
2
(8 cos θ + 6 sin θ) dθ
=
0
r dr
0
= 0(2) = 0
81
a
2a−x
dx
0
x
y−x
dy = I
4a2 + (y + x)2
u=x+y
v=x−y
⇒
x = (u + v)/2
y = (u − v)/2
and
∂(x, y)
∂(u, v)
=
c Pearson Education Limited 2011
1
2
1
2
1
2
− 12
=
−
1
2
166
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
y=x
⇒
u−v=u+v
⇒
v=0
y = 2a − x
⇒
u − v = 4a − u − v
⇒
u = 2a
x=0
⇒
u = −v
⇒
(0, 0)
(0, 0), (a, a)
2a
0
du
I=
−u
0
⇒
−v
2
4a + u2
(2a, 0), (0, 2a)
⇒
(2a, −2a)
2a
1
u2
− dv = 1
du
2
4 0 4a2 + u2
2a
1
4a2
1− 2
du
4 0
4a + u2
2a
π
a
1
−1 u
u − 2a tan
1−
=
=
4
2a 0
2
4
=
82
1
dy
0
y
2−y
x + y x+y
e
dx
x2
u=x+y
v = y/x
⇒
∂(u, v)
∂(x, y)
=
1
1
− xy2
1
x
=
1
(x + y)
x2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
y=x
⇒
v=1
y=0
⇒
v=0
y=2−x
⇒
u=2
1
2−y
dy
0
y
x + y x+y
e
dx =
x2
2
1
eu dv
du
0
2
0
=e −1
Exercises 3.4.8
83
Surface area =
1+
∂z
∂x
2
∂z
∂y
+
2
dx dy
A
where A is the domain x2 + y2 ≤ 2, z = 0.
z = 2 − x2 − y2 ⇒
∂z
∂z
= −2x,
= −2y
∂x
∂y
1 + 4x2 + 4y2 dx dy
Surface area =
A
Set x = r cos θ, y = r sin θ, then
Surface area =
√
2π
dθ
0
2
1 + 4r2 r dr
0
1
(1 + 4r2 )3/2
= 2π
12
1
(27 − 1)
= 2π
12
√
2
0
= 52π/12 = 13π/3
84(a)
Direction cosines of normal to S are ( 32 , 13 , 23 ) , so that
2
2
(x2 + y2 )
(x + y ) dS =
S
dx dy
2/3
A
c Pearson Education Limited 2011
167
168
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
where A is the area between y = 6−2x and y = 2−2x lying between y = 0 and
y = 3.
2
Thus
3
2
(x + y ) dS =
dy
2−y
2
0
S
6−y
2
3 2
(x + y2 ) dx
2
6−y
3
2
3 x3
+ y2 x
dy
=
2−y
3
0 2
2
3
15 2
183
13 − 6y + y
dy =
=
4
4
0
z dS =
84(b)
S
1
dx
√
+ x−x2
z
√
− x−x2
0
x
∂z
=− ,
∂x
z
∂z
y
=−
∂y
z
1+
dx
0
S
1
=2
2
+
∂z
∂y
2
dy
and x2 + y2 + z2 = 1
1
z dS =
∂z
∂x
√
√
+ x−x2
√
− x−x2
1 dy
x − x2 dx
0
π
=
4
(Use the substitution x =
radius
1
2
1
2
+
1
2
sin t . Alternatively, recognize area of circle of
.)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
85(a)
⇒
169
v = (xy, −x2 , x + z)
2 2 1
dS
dS = n dS =
, ,
3 3 3
2
2 2 1
v · dS =
xy − x + (x + z) dS
3
3
3
S
S
3
3−x
[xy − 2x2 + (x + 6 − 2x − 2y)] dy
dx
=
0
0
0
3
=
3
=
0
3−x
1 2
xy − 2x2 y − xy + 6y − y2
2
dx
0
x
(3 − x)2 − 2x2 (3 − x) − x(3 − x) + 6(3 − x)
2
dx
= 27/4
85(b)
Use cylindrical polar coordinates, then
2
3
dS = (i cos φ + j sin φ) dφ dz on
2
cylinder and v = (3y, 2x , z ) = (3 sin φ, 2 cos φ, z3 )
v · dS
S
2π
=
0
1
(3 sin φ cos φ + 2 cos2 φ sin φ) dz
dφ
0
3
2
sin2 φ − cos3 φ
=
2
3
2
z dS =
86
S
2π
=0
0
z2 / (1 − x2 − y2 ) dx dy
A
where x2 + y2 + z2 = 1 and A is the interior of the circle x2 + y2 = 1,
2
z dS =
S
1 − x2 − y2 dx dy
A
2π
=
1
dθ
0
√
1 − r2 r dr
0
1
= 2π − (1 − r2 )3/2
3
1
=
0
c Pearson Education Limited 2011
2π
3
z=0
170
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1 dS =
87(a)
S
1 1 + 4x2 + 4y2 dx dy
A
where A is the interior of the circle x2 + y2 = 2
dS =
√
2π
2
dθ
0
S
1 + 4r2 r dr
0
√
1
(1 + 4r2 )3/2
= 2π
12
= 13π/3
87(b)
2
(x + y ) dS =
√
2
r2
dθ
0
S
=
0
2π
(27 − 1)
12
Surface Area
2π
2
2
1 + 4r2 r dr
0
√
2π
=
16
π
=
8
2
1
1
(1 + 4r2 )5/2 − (1 + 4r)3/2
5
3
0
1
1
(243 − 1) − (27 − 1) = 149π/30
5
3
2nd moment of surface area about z -axis.
87(c)
z dS =
2
(2 − r2 )
dθ
0
S
√
2π
1 + 4r2 r dr
0
=2
√
2π
2
dθ
0
1 + 4r2 r dr −
0
√
2
2π
r2
dθ
0
0
111π
37π
26π 149π
−
=
=
=
3
30
30
10
88
Direction cosines of normal to S are ( 23 , 13 , 23 ) so that
dS =
S
dx dy
2/3
A
c Pearson Education Limited 2011
1 + 4r2 dr
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
where A is the interior of x2 + y2 = 64 lying in the first quadrant.
π/2
dS =
0
S
89
8
dθ
0
3
π 3 1 2
r dr = ×
r
2
2 2 2
8
= 24π
0
x 2 y 2
dS =
1+
+
dx dy
2
2
S
A
where A is the annulus between x2 + y2 = 4 and x2 + y2 = 12
dS =
12
dθ
0
S
√
2π
2
1
1 + r2 r dr
4
√
3/2 12
4
1
8π 3/2
= 2π
1 + r2
[4 − 23/2 ]
=
3
4
3
2
16π
(4 − 21/2 )
=
3
90
Using cylindrical polar coordinates,
and
Thus
and
dS = (4i cos φ + 4j sin φ) dφ dz
V = zi + 2 cos φj − 12 sin2 φzk
V · dS = (4z cos φ + 8 sin φ cos φ) dφ dz
π/2
5
V · dS =
dφ
(4z cos φ + 8 sin φ cos φ) dz
0
S
0
π/2
(50 cos φ + 40 sin φ cos φ) dφ
=
0
π/2
= [50 sin φ + 40 sin2 φ]0
91
= 90
F = yi + (x − 2xz)j − xyk
i
∂
curl F = ∂x
y
j
k
∂
∂y
∂
∂z
x − 2xz
−xy
= (x, y, −2z)
On sphere, x = a sin θ cos φ , y = a sin θ sin φ , z = a cos θ
c Pearson Education Limited 2011
171
172
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
and
dS = a2 (sin θ cos φ, sin θ sin φ, cos θ) sin θ dθ dφ
curl F · dS = a3 (sin2 θ cos2 φ + sin2 θ sin2 φ − 2 cos2 θ) sin θ dθ dφ
= a3 (sin2 θ − 2 cos2 θ) sin θ dθ dφ
2π
π
curl F · dS =
dφ
a3 [sin3 θ − 2 cos2 θ sin θ] dθ
0
S
0
2π
3
=
π
1
3
sin θ − sin 3θ − 2 cos2 θ sin θ
4
4
0
π
1
2
3
3
cos 3θ + cos θ = 0
− cos θ +
4
12
3
0
a dφ
0
2π
= a3 φ 0
dθ
Exercises 3.4.10
92(a)
1
dx
0
2
3
dy
2
3
xyz dz dy dx =
1
2
2
2
0
z dz
0
1
1
8
1 1
. (4) (32 − 1) =
3 2
2
3
4
3
y dy
0
=
92(b)
x dx
1
2
2
x yz dz =
0
1
2
3
dx
0
1
2
2
3
x dx
=
0
4
xyz2 dz
dy
z2 dz
y dy
1
4
2
1
1
1
(4) (32 − 1) (43 − 23 )
2
2
3
448
=
3
=
93
1
dz
−1
z
dx
0
x+z
(x + y + z) dy =
z
{2z(x + z) + 2xz} dx
dz
−1
1
x−z
1
0
3z3 dz = 0
=
−1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
173
94
π
sin(x + y + z) dx dy dz =
dx
0
π
=
0
π−x
[− cos(x + y + z)]π−x−y
dy
0
0
π
π−x
dx
=
sin(x + y + z) dz
0
dx
π−x−y
dy
0
V
π−x
0
[1 + cos(x + y)] dy
0
π
[y + sin(x + y)]π−x
dx
0
=
0 π
(π − x + sin π − sin x) dx
=
0
x2
+ cos x
= πx −
2
π
=
0
1 2
π −2
2
95
1
xyz dx dy dz =
dx
0
V
dy
0
x dx
0
0
1
=
0
1−x−y
xyz dz
0
1
=
1−x
1−x
1
y(1 − x − y)2 dy
2
1
1
2
1
x (1 − x)2 y2 − (1 − x)y3 + y4
2
2
3
4
1
1
x(1 − x)4 dx
24
0
1
1
(x − 4x2 + 6x3 − 4x4 + x5 ) dx
=
24 0
1
=
720
=
c Pearson Education Limited 2011
1−x
dx
0
174
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
96
V=
dV =
V
1
√
dx dy dz
V
dx
=
1
=
dz
0
x
(2 − x − y) dy
dx
2−x−y
dy
x2
√
0
x
x2
0
√
1
1
2 x − x3/2 − x − 2x2 + x3 + x4
=
2
2
0
1
11
4 2 1 2 1
− − − + +
=
=
3 5 4 3 4 10
30
1
dx
97
2
2
π/2
2
(x + y + z )x dx dy dz =
dφ
1
r2 r sin θ cos φ sin θr2 dr
dθ
0
V
π/2
0
0
π/2
0
r5 dr
sin θ dθ
0
=1×
1
2
cos φ dφ
=
π/2
0
1 π 1
π
× × =
2 2 6
24
x2 y2 z2 (x + y + z) dx dy dz
98
V
3 2 2
x y z dx dy dz +
=
V
1
3
=
2
x dx
0
0
z3 dz
+
0
1
x3 dx
=3
0
1−z
0
y2 dy
0
V
1−y
3
y dy
0
2
0
y2 dy
0
1−x−y
z2 dz from symmetry
0
c Pearson Education Limited 2011
1−z−y
x2 dx
z dz
1−z−x
x2 dx
1−x
1
z dz +
0
2
y dy
1
x2 y3 z2 dx dy dz
x y z dx dy dz +
V
1−x−y
1−x
2 2 3
0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1−x
x dx
0
1
0
1−x
{y2 [(1 − x)3 − 3(1 − x)2 y + 3(1 − x)y2 − y3 ]} dy
x3 dx
=
1 2
y (1 − x − y)3 dy
3
3
=3
0
0
1
1
3
3
1
(1 − x)6 − (1 − x)6 + (1 − x)6 − (1 − x)6
3
4
5
6
x3
=
0
1
=
60
=
1
1
60
1
x3 (1 − x)6 dx
0
1
x6 (1 − x)3 dx
0
1
1
x6 (1 − 3x + 3x2 − x3 ) dx
=
60 0
1
1 1 3 3
− + −
=
60 7 8 9 10
1
=
50400
99
⎫
u = x + y + z⎬
uv = y + z
⎭
uvw = z
x = u − uv
y = uv − uvw
z = uvw
1 − v v − vw vw 1 − v
∂(x, y, z)
= −u u − uw uw = −u
∂(u, v, w)
0
−uv
uv 0
1 v vw
= 0 u uw = u2 v
0 0 uv
v vw
u uw
0 uv
exp(−(x + y + z)3 ) dx dy dz
I=
V
1
=
dx
0
1−x
1−x−y
exp(−(x + y + z)3 ) dz
dy
0
0
x+y+z=1
x=0
⇒
⇒
u=y+z
uv = y + z
u=1
⇒
c Pearson Education Limited 2011
v=1
dx
175
176
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
⎫
u = x + z⎬
uv = z
⎭
uvw = z
⇒
y=0
⎫
u = x + y⎬
uv = y
⎭
uvw = 0
⇒
z=0
1
I=
1
du
0
2 −u3
u e
=
1
0
1
dv
0
1
du
0
⇒
w=1
⇒
w=0
e−u u2 v dv
3
0
1
1
v dv
3
1 2
1
v
= − e−u
3
0 2
1
= [1 − e−1 ]
6
dw
0
1
[w]10
0
100
1
yz dx dy dz =
dx
dy
0
V
0
1
=
0
1
0
0
0
0
yz dz
1−x
1
y(2 − x − y)2 dy
2
1−x
1
[y(2 − x)2 − 2y2 (2 − x) + y3 ] dy
2
0
dx
=
2−x−y
0
dx
1−x
1
1 1
2
1
(1 − x)2 (2 − x)2 − (1 − x)3 (2 − x) + (1 − x)4 dx
2 2
3
4
1
1 1 2
2
1
t (1 + t)2 − t3 (1 + t) + t4
2 2
3
4
=
=
dt
1
1 1 2
1
2
2
1
t + t3 + t 4 − t 3 − t 4 + t 4
2
3
3
4
0 2 2
1
1
2
1 1
+
+
=
=
2 6 12 60
15
=
c Pearson Education Limited 2011
dt
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
177
Volume of prism =
dx dy dz
V
1
=
dx
0
1
2−x−y
dy
dz
0
0
1−x
(2 − x − y) dy
dx
=
1−x
0
0
1
1
(2 − x)(1 − x) − (1 − x)2 dx
2
0
1
1 2
1
2
2 − 3x + x − + x − x
dx
=
2
2
0
1
2
3
= −1+ =
2
6
3
=
Let the coordinates of the centroid be ( x, y , z), the taking moments about
appropriate axes, we have
2
x=
3
2
x dx dy dz, y =
3
y dx dy dz,
V
2
z=
3
V
z dx dy dz
V
From symmetry y = x .
3
x=
2
3
=
2
1
dx
0
0
x dz
0
1−x
x(2 − x − y)dy
dx
0
2−x−y
dy
1
1−x
0
1
3 13
x − 2x2 + x3 dx
=
2 0 2
2
5
3 3 2 1
=
=
− +
2 4 3 8
16
c Pearson Education Limited 2011
178
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
3
z=
2
3
=
2
=
3
2
3
=
2
1
dx
0
1−x
0
1
(2 − x − y)2 dy
2
y=1−x
−1
(2 − x − y)3
6
1
0
z dz
0
dx
0
2−x−y
dy
0
1
1−x
1
−
0
dx
y=0
1
1 − (2 − x)3 dx
6
1
1
3
− x − (2 − x)4
2
6
24
11
3 11
=
= ×
2 24
16
1
=
0
101
2π
z dx dy dz =
π/2
dφ
0
V
r3 cos θ sin θ dr
π/4
2π
0
π/2
1
sin 2θ
2
dφ
=
1
dθ
0
π/4
1
= [2π] − cos 2θ
4
π/2
1
r3 dr
0
π
1
=
4
8
π/4
102
x dx dy dz =
π/2
dφ
0
V
π/2
0
r sin2 θ.r cos φ dr
dθ
0
0
π/2
π/2
2
cos φ dφ
=
a
r3 dr
sin θ dθ
0
π 1
a4 = πa4 /16
= [1]
4 4
c Pearson Education Limited 2011
a
0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
179
Exercises 3.4.13
F · dS
103
F = (4xz, −y2 , yz)
S
S has six faces and the integral can be evaluated as the sum of six integrals.
F · dS =
F · dS +
on x=0
S
F · dS +
on x=1
F · dS +
on y=1
1
F · dS
on y=0
+
F · dS +
on z=0
F · dS
on z=1
1
(0, −y2 , yz) · (−i dy dz)
=
0
0
1 1
(4z, −y2 , yz) · (i dy dz)
+
0
0
0
1
0
0
0
1
(4xz, 0, 0) · (−j dx dz)
+
1
1
(4xz, −1, yz) · (j dx dz)
+
1
1
(0, −y2 , 0) · (−k dx dy)
+
0
1
0
1
(4x, −y2 , y) · (k dx dy)
+
0
0
1
1
0
=2 − 1 +
0
1
4z dy dz + 0 +
=0 +
1
0
S
div F dV
V
(z + z + 2z) dx dy dz
=
V
2π
0
π/2
dφ
=
1
y dx dy
0
1
3
=
2
2
F · dS =
(−1) dx dz + 0 +
0
104
1
2
4r cos θ.r2 sin θ dr
dθ
0
0
π/2
= [2π] [− cos 2θ]0
1 4
r
4
2
= 16π
0
c Pearson Education Limited 2011
0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
180
dS = −k dx dy,
105 On z = 0,
On z = 3,
F · dS = 0
F · dS = 9 dx dy
F = (4x, −2y , 9),
2
dS = k dx dy,
On x2 + y2 = 4,
F = (4x, −2y2 , 0),
F = (8 cos φ, −8 sin φ, z2 )
dS = (i cos φ + j sin φ)2 dφ dz,
and F · dS = 16(cos2 φ − sin3 φ) dφ dz
F · dS =
2π
9 dx dy +
0
S
(z=3)
3
16(cos2 φ − sin3 φ) dz
dφ
0
2π
cos2 φ − sin3 φ dφ
= 36π + 48
0
= 84π
div F dV =
(4 − 4y + 2z) dx dy dz
V
V
2π
=
2
dφ
0
0
2π
(21 − 12r sin φ)r dr
0
(4 − 4r sin φ + 2z)r dz
0
2
dφ
=
3
dr
0
2π
(42 − 32 sin φ) dφ = 84π
=
0
106
div (F × grad φ) = grad φ · curl F − F · curl (grad φ) and curl (grad φ) ≡
0 for all φ.
grad φ · curl F dV =
Hence
V
107
F · dS =
S
(F × grad φ) · dS
S
F · dS +
on x=0
F · dS +
on y=0
F · dS +
on z=0
F · dS
on z=1
F · dS
+
on x2 +y 2 =4
On x = 0, F = (y2 , 0, 0), dS = −i dy dz,
F · dS =
on x=0
2
0
dy
1
0
−y2 dz = − 83
On y = 0, F = (0, 0, 0) , so contribution is zero
On z = 0, F = (xy + y2 , x2 y, 0) ,
2
2
On z = 1, F = (xy + y , x y, 0) ,
dS = −k dx dy , so contribution is zero
dS = k dx dy , so contribution is zero
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
On x2 + y2 = 4, F = (4(sin φ cos φ + sin2 φ), 8 cos2 φ sin φ, 0) ,
181
dS = 2(cos φi +
sin φj) dφ dz
π/2
F · dS =
0
on x2 +y 2 =4
1
8 sin φ cos2 φ + 8 sin2 φ cos φ + 16(cos2 φ sin2 φ) dz
dφ
0
8
1
8
= − cos3 φ + sin3 φ + 2φ − sin 4φ
3
3
2
16
=
+π
3
V
π/2
0
0
π/2
=
0
0
2
(r2 sin φ + r3 cos2 φ) dr
dφ
0
1
(r sin φ + r2 cos2 φ)r dz
dr
π/2
=
2
dφ
=
=
0
(y + x2 ) dx dy dz
div F dV =
V
π/2
0
8
sin φ + 4 cos2 φ
3
dφ
8
+π
3
Hence, result
108
i
∂
curl F =
∂x
36xz + 6y cos x
j
∂
∂y
∂
∂z
2
18x − cos y
k
3 + 6 sin x + z sin y
= i(sin y − sin y) + j(36x − 36x) + k(6 cos x − 6 cos x)
=0
Hence, there is a function φ(x, y, z) , such that F = grad φ
c Pearson Education Limited 2011
182
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
109
curl A · dS
A · dr =
C
S
i
j
k
∂
∂
∂
curl A = ∂x
∂y
∂z
−y x
0
2k · dS
A · dr =
C
Let S be the ellipse
Then,
2
x
a2
+
= 2k
S
2
y
b2
= 1, z = 0
dS = k · dx dy , and
A · dr = 2
dx dy = 2πab
S
110
i
∂
curl F = ∂x
2x − y
=k
∂
∂y
−yz2
2π
curl F · dS = 16
π/2
k · (sin θ cos φi + sin θ sin φj + cos θk) sin θ dθ
dφ
0
S
k
∂
∂z = i(−2yz + 2yz) + j(0) + k(1)
−y2 z
j
0
2π
π/2
dφ
= 16
0
sin θ cos θ dθ = 16[2π]
0
1
sin2 θ
2
π/2
0
= 16π
F · dr
C
On circle x2 + y2 = 16, z = 0, x = 4 cos φ , y = 4 sin φ , r = 4(cos φ, sin φ, 0)
F = (8 cos φ − 4 sin φ, 0, 0),
2π
F · dr =
C
dr = 4(− sin φ, cos φ, 0) dφ
(−32 cos φ sin φ + 16 sin2 φ + 0) dφ
0
= 16π
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
111
curl (af(r)) = −a × grad f(r)
−(a × grad f(r)) · n dS =
af(r)· dr
C
S
−(grad f(r) × n) · a dS =
⇒
a·
(n × grad f(r)) dS = a ·
S
n × grad f(r) dS =
f(r) dr
C
S
⇒
f(r) = 3xy2
1
f(r) dr
C
⇒
af(r)· dr
C
S
⇒
183
grad f(r) = (3y2 , 6xy, 0)
n × grad f(r) = k × grad f(r) = (−6xy, 3y2 , 0)
2
1
2
dx
(−6xy, 3y , 0) dy =
(−12x, 8, 0) dx
0
0
0
= (−6, 8, 0)
1
f(r) dr =
0.i dx +
0
C
2
3y j dy +
0
0
2
0
12xi dx +
1
0.j dy
2
= (−6, 8, 0)
curl F · dS =
112
F · dr
C
S
i
∂
curl F = ∂x
2y + z
j
∂
∂y
x−z
∂
∂z = (2, 2, −1)
y − x
k
curl F · dS =
S
(2, 2, −1) · (sin θ cos φ, sin θ sin φ, cos θ) sin θ dφ dθ
S
π/2
(2 sin θ cos φ + 2 sin θ sin φ − cos θ) sin θ dθ
dφ
=
0
=
0
π/2
π/2
0
1
π sin φ −
2
dφ =
3π
4
c Pearson Education Limited 2011
184
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Here, C has three portions:
⇒
On z = 0, r = (cos φ, sin φ, 0)
dr = (− sin φ, cos φ, 0) dφ
and F = (2 sin φ, cos φ, sin φ − cos φ)
π/2
(−2 sin2 φ + cos2 φ) dφ = −
F · dr =
0
On y = 0, r = (sin θ, 0, cos θ)
⇒
π
4
dr = (cos θ, 0, − sin θ) dθ
and F = (cos θ, sin θ − cos θ, − sin θ)
π/2
F · dr =
(cos2 θ + sin2 θ) dθ =
0
⇒
On x = 0, r = (0, sin θ, cos θ)
π
2
dr = (0, cos θ, − sin θ) dθ
and F = (2 sin θ + cos θ, − cos θ, sin θ)
0
(− cos2 θ − sin2 θ) dθ =
F · dr =
π/2
π
2
Review Exercises 3.7
1
−y
∂u
= nxn−1 f(t) + xn f (t)
∂x
x2
∂u
1
= xn f (t)
∂y
x
x
∂u
∂u
+y
= nxn f(t) = nu
∂x
∂y
(1)
Differentiate (1) w.r.t. x
∂u
∂2 u
∂u
∂2 u
+x 2 +y
=n
∂x
∂x
∂x∂y
∂x
(2)
Differentiate (1) w.r.t. y
x
∂u
∂2 u
∂2 u
∂u
+
+y 2 =n
∂x∂y ∂y
∂y
∂y
c Pearson Education Limited 2011
(3)
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
x × (2) + y × (3)
185
⇒
2
2
∂u
∂u
∂u
∂2 u
∂u
2∂ u
2∂ u
+y
+x
+y
+y
x
+ 2xy
=n x
∂x
∂y
∂x2
∂x∂y
∂y2
∂x
∂y
⇒
x2
2
∂2 u
∂2 u
2∂ u
+
2xy
= n(n − 1)u
+
y
∂x2
∂x∂y
∂y2
u(x, y) = x4 + y4 + 16x2 y2
∂u
= x(4x3 + 32xy2 )
x
∂x
∂u
= y(4y3 + 32x2 y2 )
y
∂y
x
x2
∂u
∂u
+y
= 4(x4 + 16x2 y2 + y4 )
∂x
∂y
2
∂2 u
∂2 u
2∂ u
+
y
+
2xy
= x2 (12x2 + 32y2 ) + 2xy(64xy) + y2 (12y2 + 32x2 )
∂x2
∂x∂y
∂y2
= 12(x4 + y4 + 16x2 y2 )
2
∂f
∂f ∂f ∂2 f
∂2 f
∂2 f
∂2 f
=
+ , 2 =
+
+
2
∂x
∂u ∂v ∂x
∂u2
∂u∂v ∂v2
∂2 f
∂2 f
∂2 f
∂2 f
∂2 f
=a 2 +b
+
a
b+
∂x∂y
∂u
∂u∂v ∂v2
∂v∂u
∂f
∂2 f 2
∂2 f
∂f
∂f ∂2 f
∂2 f 2
a
+
2
b
=
a + b, 2 =
ab
+
∂y
∂u
∂v ∂y
∂u2
∂u∂v
∂v2
9
⇒
2
2
∂2 f
∂2 f
∂2 f
2 ∂ f
2 ∂ f
+
2
−
9
=(9
−
9a
+
2a
)
+
(9
−
9b
+
2b
)
∂x2
∂x∂y
∂y2
∂u2
∂v2
2
∂ f
9
+ 2 9 − (a + b) + 2ab
2
∂u∂v
⎫
9 − 9a + 2a2 = 0 ⎪
⎪
⎬
2
9 − 9b + 2b = 0
⎪
⎪
⎭
9
9 − 2 (a + b) + 2ab = 0
⇒
a = b
⇒
∂2 f
=0 ⇒
f = F(u) + G(v)
∂u∂v
i.e. f(x, y) = F(x + 3y) + G(x + 3y/2)
c Pearson Education Limited 2011
a = 3,
b=
3
2
186
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
f(x, 0) = F(x) + G(x) = sin x
∂f
3
(x, 0) = 3F (x) + G (x) = 3 cos x
∂y
2
1
F(x) + G(x) = sin x + k
2
⇒
1
G(x) = −k and F(x) = sin x + 2k
2
⇒
f(x, y) = sin(x + 3y)
⇒
3
i
j
k
∂
∂
∂
∇ × (P, Q, R) = ∂x ∂y ∂z
∂f ∂f ∂f
∂x
∂y
∂z
2
2
2
∂ f
∂ f
∂ f
∂2 f
∂2 f
∂2 f
=i
+j
+k
−
−
−
∂y∂z ∂z∂y
∂x∂z ∂z∂x
∂x∂y ∂y∂x
=0
⇒
4(a)
∇ × (∇f) ≡ 0
xy = c, hyperbolas
grad f = (y, x) =
y=
That is, tangent in direction of
t · grad f =
4(b)
x2
x
=c
+ y2
c
x
c
x
⇒
, x on hyperbola
dy
c
=− 2
dx
x
1, − xc2 = t
c
c
− =0
x x
that is, orthogonal
circles, centres on x -axis, through (0,0)
grad f =
−2y
y2 − x2
, 2
2
2
2
(x + y ) (x + y2 )2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
5(a)
ω · r) = ω × (∇ × r) + r × (∇ × ω) + (ω
ω · ∇)r + (r · ∇)ω
ω
grad(ω
=ω×0+0+ω+0
=ω
5(b)
ω × r) = −(ω
ω · ∇)r + ω(∇ · r)(+r · ∇)ω
ω − r(∇ · ω)
curl (ω
ω + 3ω
ω+0+0
= −ω
ω
= 2ω
6(a)
See problem 3 above.
6(b)
div v = div {grad [zf(r)] + αf(r)k}
= div {kf(r) + z grad f(r)} + αk · ∇f(r)
= k · ∇f(r) + k · grad f(r) + z∇2 f(r) + αk · ∇f(r)
∂f
= (2 + α)
∂z
2
∇ v = ∇(∇ · v) − ∇ × (∇ × v)
∂f
− ∇ × (∇ × (∇(zf) + αfk))
= (2 + α)∇
∂z
∇ × ∇(zf) ≡ 0
∇ × (αfk) = α∇f × k
∇ × (∇ × αfk) = α(k · ∇)∇f − αk(∇2 f)
∂f
∂
= α (∇f) = α∇
∂z
∂z
∂f
⇒
∇2 v = 2∇
∂z
7
F = (x2 − y2 + x)i − (2xy + y)j
i
j
∂
∂
∇×F=
∂x
∂y
x2 − y2 + x −2xy − y
k
∂
∂z = (0, 0, −2y + 2y) = 0
0
c Pearson Education Limited 2011
187
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
188
∇f =
∂f ∂f ∂f
, ,
∂x ∂y ∂z
⇒
= (x2 − y2 + x, −2xy − y, 0)
f(x, y, z) =
(2,1)
(2,1)
F · dr =
(1,2)
x3
x2 − y 2
− y2 x +
+c
3
2
(2,1)
grad f · dr = [f](1,2) =
(1,2)
22
3
dr = i dx + j dy = (i − j) dx
as on y = 3 − x ,
dy = − dx
2
2
(x − y + x + 2xy + y) dx =
2
(x2 − (3 − x)2 + x + 2(3 − x) + 3 − x) dx
2
1
1
22
=
3
8
W=
F· dr
C
r = (1 − cos θ)i + sin θj
dr = (sin θi + cos θj) dθ
8(a)
F = 2 sin 12 θi
θ
θ
cos dθ
2
2
0
π
θ
8
8
sin3
=
=
3
2 0
3
F· dr =
C
8(b)
π
4 sin2
F = 2 sin θ2 n̂ = 2 sin θ2 (sin θi + cos θj)
π
F· dr =
C
0
θ
θ
2 sin dθ = 4 − cos
2
2
c Pearson Education Limited 2011
π
=4
0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
9
0≤t≤1
r = (i + j + k)t
dr = (i + j + k) dt
1
F · dr
W=
0
F = (xy, −y, 1) = (t2 , −t, 1)
1
1 1
5
(t2 − t + 1) dt = − + 1 =
W=
3 2
6
0
⇒
10
dr × B
F=I
C
θ
r = sin θi + cos θj + sin k
2
θ
1
dr = cos θi − sin θj + cos k dθ
2
2
B = sin θi − cos θj + k
2π
θ
1
θ
1
cos cos θ − sin θ + j
cos sin θ − cos θ
i
F=I
2
2
2
2
0
4
+ k(sin2 θ − cos2 θ) dθ = Ij
3
11
v · dr
Circulation =
C
−1
=
1
−y dx
+
x2 + y 2
−1
1
x dy
+
2
x + y2
1
−1
y dx
+
2
x + y2
1
−1
x dy
+ y2
x2
on y = 1
on x = −1
on y = −1
on x = 1
−1
1
1
1
1
dy
dx
dy
dx +
−
+
=
2
2
2
2
1+y
−1 1 + x
1
−1 1 + x
−1 1 + y
=0
12
ρ(x2 + y2 ) dA,
Iz =
A
c
c
(x2 + y2 )kxy dy
dx
=
0
where density ρ = kxy
x2 /c
c Pearson Education Limited 2011
189
190
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
c
c
1
kx(x2 + y2 )2
=
dx
4
0
x2 /c
c
4 2
1
x
1
=
− kx x2 + 2
+ kx(x2 + c2 )2 dx
4
c
4
0
c 9
1
x
2x7
3 2
4
dx
= k
−
+ 2 − 2x c − xc
4 0
c4
c
13 6
=
kc
80
13
Equation of cone is x2 + y2 =
V=2
dx
c
z dy
0
a
√
a2 −x2
h 2
h+
x + y2
a
dx
=2
0
c
− h)2
√
a2 −x2
a
a2
h2 (z
dy
√
a −x
hx2
hy 2
−1 y
2
=2
hy +
x +y
dx
sinh
−
2a
x 2a
c
0
√
a √
2
2 − x2
a
h 2
hx
sinh−1
dx
a − x2 +
=2
2
2a
x
c
√
hc √
3
2 − c2
a
2ha2 π
hc
−1
−1 c
− sin
−
tanh
=
a2 − c2 −
3
2
a
3
3a
a
14
a
Volume is 8
2
2
dV
x y z≥0
x2 + y2 = a2 is a cylinder with z -axis as axis of symmetry and radius a.
z2 + y2 = a2 is a cylinder with x -axis as axis of symmetry and radius a.
⇒
a
√a2 −y2
√a2 −y2
dy
dx
dz
0
0
a √
√ 2 2
a2 −y 2
a −y
=8
[x]0
[z]0
dy
0
a
a
1
16a3
(a2 − y2 ) dy = 8 a2 y − y3 =
=8
3
3
0
0
V=8
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
15
Elastic energy of ΔV is q2 ΔV/(2EI) where q = q0 ρ/r and ρ is the distance
from the centre and r is radius of cylinder.
2π
Total energy =
dφ
0
=
On x = 0,
dS = −i dy dz
On y = 0,
v · dS = 0
On z = 1,
v · dS = 0
⇒
On
0
2 2
πq0 r l
0
l
q20 ρ3
dz
2EIr2
q20 ρ3 dρ
2EIr2
4EI
and
v · dS = −3x2 y dy dz ≡ 0
x + y = 1, dS =
1
dx
0
dρ
z = 0, v · dS = 0
On
⊂⊃ v · dS =
r
0
r
= 2πl
16
0
1
√1 (i
2
+ j) dS
√
3 2
1
2
√ x (1 − x) + √ x(1 − x)
2 dz
2
2
1
(2x2 + x)(1 − x) dx
=
0
1
=
3
17
191
⊂⊃ v · dS
S
dS = (i sin θ cos φ + j sin θ sin φ + k cos θ)a2 sin θ dθ dφ
On S,
and v = i2a sin2 θ cos φ sin φ − ja2 sin2 θ sin2 φ + k(a sin θ cos φ + a sin θ sin φ)
⊂⊃ v · dS =
2π
{2a3 sin4 θ cos2 φ sin φ − a4 sin4 θ sin3 φ
dθ
0
S
π
0
+ a3 sin2 θ cos θ cos φ + a3 cos θ sin2 θ sin φ} dφ
=0
F · dr
18
C
C is the circle x2 + y2 = 16, z = 0, so that, on the circle
F = x2 + y − 4, 3xy, 0 ,
r = 4 (cos θ, sin θ, 0)
c Pearson Education Limited 2011
192
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
and d r = (−4 sin θ, 4 cos θ, 0) dθ
2π
θ=0
2π
−16 4 cos2 θ + sin θ − 1 sin θ + 192 cos θ sin2 θ dθ
F · dr =
C
−16 sin2 θ dθ
=
(from symmetries)
0
= −16π
i
∂
curl F =
∂x
x2 + y − 4
∂
= (0, −2z, 3y − 1)
∂z
2
2xz + z
j
k
∂
∂y
3xy
On the hemisphere
r = 4 (sin θ cos φ, sin θ sin φ, cos θ)
dS = 16 (sin θ cos φ, sin θ sin φ, cos θ) sin θ dφ dθ
curl F . dS =
S
π/2
2π
16(−8 sin2 θ cos θ sin φ + 12 sin2 θ cos θ cos φ − cos θ sin θ)dφ
dθ
0
0
π/2
−16 cos θ sin θ [2π] dθ = −16π
=
0
19
a · dS =
div a dV −
S
V
a · dS
S1
where V is the hemisphere x2 + y2 + z2 = a2 (different a from the vector a), S1
is the circle x2 + y2 = a2 , z = 0. div a = 0 and d S = −k dx dy on S1
a · dS =
S1
Hence
(xi + yj) · (−k dx dy) = 0
S
a · dS = 0
S
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
20
1
1−x
xyz dV =
V
2−x
xyz dz dy dx
0
0
1
0
0
1−x
=
0
1
2
xy (2 − x) dy dx
2
1
1
2
2
x (1 − x) (2 − x) dx
0 4
1
1 5
x − 6x4 + 13x3 − 12x2 + 4x dx
=
0 4
1
3
13
1
13
=
−
+
−1+ =
24 10 16
2
240
=
21
−u
−v
y + ∆y
v
x
u
y
x + ∆x
Net circulation (anti clockwise) is
−u(x, y)Δx − v(x + Δx, y)Δy + u(x, y + Δy)Δx + v(x, y)Δy
If net circulation is zero then, dividing by ΔxΔy ,
u(x, y + Δy) − u(x, y) v(x + Δx, y) − v(x, y)
−
=0
Δy
Δx
Δx, Δy → 0 gives
∂u ∂v
−
= 0.
∂y
∂x
Since u = − ∂ψ
∂y and v =
∂ψ
∂x
we obtain
∂2 ψ ∂2 ψ
+ 2 =0
∂x2
∂y
Laplace equation.
c Pearson Education Limited 2011
193
4
Functions of a Complex Variable
Exercises 4.2.2
1(a)
If | z − 2 + j |=| z − j + 3 |, so that
| x + jy − 2 + j |=| x + jy − j + 3 |
or
(x − 2)2 + (y + 1)2 = (x + 3)2 + (y − 1)2
x2 − 4x + 4 + y2 + 2y + 1 = x2 + 6x + 9 + y2 − 2y + 1
Then cancelling the squared terms and tidying up,
y=
1(b)
5
5
x+
2
4
z + z∗ + 4j(z − z∗ ) = 6
Using, z + z∗ = 2x, z − z∗ = 2jy gives
2x + 4j2jy = 6
3
1
y= x−
4
4
2
The straight lines are
| z − 1 − j | =| z − 3 + j |
| z − 1 + j | =| z − 3 − j |
which, in Cartesian form, are
(x − 1)2 + (y − 1)2 = (x − 3)2 + (y + 1)2
that is,
x2 − 2x + 1 + y2 − 2y + 1 = x2 − 6x + 9 + y2 + 2y + 1
y=x−2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
195
and (x − 1)2 + (y + 1)2 = (x − 3)2 + (y − 1)2
i.e.
x2 − 2x + 1 + y2 + 2y + 1 = x2 − 6x + 9 + y2 − 2y + 1
y = −x + 2
These two lines intersect at π/2 (the products of their gradients is −1) and
y = 0, x = 2 at their intersection, that is, z = 2 + j0.
3
w = jz + 4 − 3j can be written as
w = ejπ/2 z + 4 − 3j (since j = cos
π
π
+ j sin = ejπ/2 )
2
2
which is broken down as follows:
z
−→
−→
ejπ/2 z
rotate
translation
anticlockwise
by 12 π
(0, 0) → (4, −3)
ejπ/2 z + 4 − 3j = w
Let w = u + jv and z = x + jy
so that,
u + jv = j(x + jy) + 4 − 3j
= jx − y + 4 − 3j
that is,
u = −y + 4
(1)
v=x−3
(2)
Taking 6 times equation (2) minus equation (1) gives,
6v − u = 6x + y − 22
so that, if 6x + y = 22, we must have 6v − u = 0 so that, u = 6v is the image of
the line
6x + y = 22
c Pearson Education Limited 2011
196
4
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Splitting the mapping w = (1 − j)z into real and imaginary parts gives
u + jv = (1 − j)(x + jy)
= x + y + j(y − x)
that is,
u=x+y
v=y−x
so that,
u + v = 2y
Therefore y > 1 corresponds to u + v > 2.
5
Since w = jz + j
x = v − 1, y = −u
so that x > 0 corresponds to v > 1.
6
Since w = jz + 1
v=x
u = −y + 1
so that x > 0 ⇒ v > 0
and 0 < y < 2 ⇒ −1 < u < 1 or | u |< 1.
This is illustrated below
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
7 Given w = (j +
parts,
√
197
√
3)z + j 3 − 1, we obtain, on equating real and imaginary
√
√
√
u = x 3 − y − 1, v = x + y 3 + 3
√
√
or v 3 − u = 4y + 4, and v + u 3 = 4x
on rearranging.
Thus 7(a)
√
√
y = 0 corresponds to v 3 − u = 4 or u = v 3 − 4
7(b)
√
√
x = 0 corresponds to v + u 3 = 0 or v = −u 3
7(c)
√
√
√
Since u + 1 = x 3 − y and v − 3 = x + y 3 squaring and adding gives
(u + 1)2 + (v −
√ 2
√
√
3) = (x 3 − y)2 + (x + y 3)2
= 4x2 + 4y2
Thus, x2 + y2 = 1 ⇒ (u + 1)2 + (v −
7(d)
√ 2
3) = 4
√
√
Since v 3 − u = 4y + 4 and v + u 3 = 4x , squaring and adding gives
4v2 + 4u2 = 16(y + 1)2 + 16x2
or
u2 + v2 = 4(x2 + y2 + 2y + 1)
Thus, x2 + y2 + 2y = 1 corresponds to u2 + v2 = 8
c Pearson Education Limited 2011
198
8(a)
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
w = αz + β
Inserting z = 1 + j, w = j and z = −1, w = 1 + j gives the following two equations
for α and β
j = α(1 + j) + β
1 + j = −α + β
or
from which, by subtraction,
−1 = (2 + j)α or
so that, β = 1 + j + α =
8(b)
gives
α=
1
(−2 + j)
5
1
(3 + 6j) gives 5w = (−2 + j)z + 3 + 6j
5
Writing w = u + jv, z = x + jy and equating real and imaginary parts
5u = −2x − y + 3
5v = x − 2y + 6
Eliminating y yields
5v − 10u = 5x
or
v − 2u = x
5u + 10v = −5y + 15
or
u + 2v = −y + 3
Eliminating x yields
so that y > 0 corresponds to u + 2v < 3
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
8(c)
From part (b),
x = v − 2u
y = 3 − u − 2v
Squaring and adding gives
x2 + y2 = (v − 2u)2 + (3 − u − 2v)2
= 5(u2 + v2 ) − 6u − 12v + 9
| z |< 2 ⇒ x2 + y2 < 4
so that, 5(u2 + v2 ) − 6u − 12v + 5 < 0
or (5u − 3)2 + (5v − 6)2 < 20
4 2 √ 2
6 2
20
3 2
= =
+ v−
<
5
that is, v −
5
5
25
5
5
8(d)
The fixed point(s) are given by
5z = (−2 + j)z + 3 + 6j
so that,
3 + 6j
(3 + 6j)(7 + j)
=
7−j
50
3
(1 + 3j)
=
10
z=
Exercises 4.2.5
9
Writing w =
1
1
, z=
z
w
u − jv
1
= 2
u + jv
u + v2
v
y=− 2
u + v2
⇒ x + jy =
so that
c Pearson Education Limited 2011
199
200
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
v
>c
+ v2
v
or rearranging u2 + v2 + < 0
c
If y > c then, −
u2
1
u + v+
2c
2
2
<
1
2c
2
v
>c⇒v<0
+ v2
v
If c < 0, put c = −d and − 2
> −d
u + v2
1 2 1 2
>
or, on rearranging, u2 + v −
2d
2d
If c = 0, −
u2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1
3
7
3
1
in z + + j =
gives + + j =
10
Putting z =
w
4
4
w 4
7
3
1 +
+ j w = | w | which, writing w = u + jv and expanding, gives
4
4
2
2
3
3
49 2
1+ u−v +
v+u =
(u + v2 )
4
4
16
201
7
or
4
or, on rearranging gives
2
4
u2 + v2 − u + v − = 0
3
3
2
2 2
1
2
7
u−
+ v+
=
2
3
6
a circle centre
1 2
7
,−
and radius .
2 3
6
1
1
in | z − a |= a gives | 1 − aw |= a | w | from which u =
w
2a
can be obtained (on writing w = u2 + v2 ).
1
1
1
under w = · · · =
; that is, the half
Hence | z − a |= a maps to Re{w} =
2a
z
2a
1
.
plane Re{w} >
2a
1
. The
Moreover, the interior of | z − a |= a maps to right of the line Re{w} =
2a
2
1
point z = a mapping to w = confirms this.
2
a
11
Putting z =
12
The general bilinear mapping is
w=
az + b
cz + d
with z = 0, w = j ⇒ b = jd ,
c Pearson Education Limited 2011
202
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
z = −j, w = 1 ⇒ d − jc = b − ja
and z = −1, w = 0 ⇒ a = b
Hence b = a, d = −ja and c = ja
and the mapping is
w=
z+1
j(z − 1)
Making z the subject of this formula, we obtain
z=
jw + 1
jw − 1
Writing z = x + jy, w = u + jv and equating real and imaginary parts
x=
−2u
u2 + v2 − 1
, y= 2
2
2
u + (v + 1)
u + (v + 1)2
Lines x = constant = k , say, transform to
or
k[u2 + (v + 1)2 ] = u2 + v2 − 1
1
2k
v=−
u2 + v2 +
k−1
k−1
This can be rewritten as
k
u + v+
k−1
2
2
=
1
k
1
=
−
2
(k − 1)
k−1
(k − 1)2
which are circles (except k = 1 which is v = −1).
Lines y = constant = l, say, transform to
or
2u
u2 + (v + 1)2 +
=0
l
2
1
1
+ (v + 1)2 = 2
u+
l
l
which are circles (except l = 0 which is u = 0).
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The fixed points are given by
z+1
jz − j
2
jz − (j + 1)z − 1 = 0
z=
or
(j + 1) ±
(j + 1)2 + 4j
2j
√
(j + 1) ± 6j
=
2j
√
1
= (j − 1)(−1 ± 3)
2
z=
√
√
6j = ±(1 + j) 3 ).
(since
13
w=
1+j
z
13(a)
z=1⇒w=1+j
1+j
z=1−j⇒w=
=j
1−j
z=0⇒w=∞
c Pearson Education Limited 2011
203
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
√
2
|1+j|
13(b) | w |=
=
|z|
√ |z|
√
2
< 1 ⇒| w |> 2
so that | z |=
|w|
that is, interior of the unit circle maps to the exterior of the circle, centre as the
√
origin and radius 2.
204
13(c)
z=
1+j
w
(1 + j)
(u − jv) so that
u2 + v2
u+v
u−v
x= 2
, y= 2
2
u +v
u + v2
⇒ x + jy =
Therefore x = y corresponds to v = 0 (the real axis) and x + y = 1 corresponds
2u
to 2
= 1 that is (u − 1)2 + v2 = 1 a circle, centre (1.0) and radius 1.
u + v2
The fixed point of the mapping is given by z2 = 1 + j . Using the polar
√
form 1 + j = 2eπj/4 , so z = ±21/4 eπj/8
13(d)
14
The bilinear transformation
w=
z+1
z−1
Writing z = x + jy, w = u + jv and equating real and imaginary parts gives
x2 + y2 − 1
2y
u=
, v=
2
2
(1 + x) + y
(1 + x)2 + y2
Hence, all points on the circle x2 + y2 = 1 correspond to u = 0.
From the point (0, −1) to the point (0, 1) on the circle x2 + y2 = 1 we use
the Parameterization x = cos θ, y = sin θ, π/2 ≤ θ ≤ 3π/2. Using v =
2y
2y
y
=
we note that v =
on x2 + y2 = 1, so that
2
2
2
2
(1 + x) + y
1 + x + y + 2x
1+x
2 sin 12 θ cos 12 θ
1
sin θ
π
3π
1
=
tan
v=
=
θ
and
between
θ
=
and
θ
=
,
tan
θ
1
1 + cos θ
2
2
2
2
2 cos2 2 θ
ranges from 1 to ∞ and from −∞ to −1 hence | v |≥ 1.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
205
15(a)
With w = u + jv and z = x + jy
3w + j
z+j
implies z =
from which we deduce that
The transformation w =
z−3
w−1
3(u2 + v2 ) − 3u + v
, y=
(u − 1)2 + v2
x2 + y2 − 3x + y
u=
, v=
(x − 3)2 + y2
x=
and
u − 3v − 1
(u − 1)2 + v2
x − 3y − 3
(x − 3)2 + y2
The line y = 0 corresponds to the line u − 3v − 1 = 0 in the w plane. The line
x = y corresponds to the curve
that is,
3(u2 + v2 ) − 3u + v = u − 3v − 1
2
2
2
5
2
+ v+
=
u−
3
3
9
(1)
2 2
1√
,−
and radius
5 in the w plane.
3 3
3
The origin in the z plane (the intersection of the line y = 0 and x = y ) corresponds
1
to the point w = −j in the w plane. The point at infinity in the z plane (the
3
other ‘intersection’) corresponds to the point w = 1 in the w plane.
a circle centre
The origin (in the w plane) lies outside the circle (1), and is also outside the wedge
shaped region in the z plane (z = −j3 is its image).
So, the following figure can be drawn:
2
lies inside the shaded region in the w plane, and corresponds to
3
3· 2 + j
= −3(2 + j) = −6 − 3j inside the shaded region of the z
the point z = 2 3
3 −1
plane. (This is a useful check.)
The point w =
c Pearson Education Limited 2011
206
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
15(b)
The fact that w = 1 does not correspond to any finite value of z has
already been established.
z+j
.
z−3
Taking the modulus of both sides gives
Consider the equation w =
z + j
| w |=
z−3
If
| w |= 1 ⇒| z + j | =| z − 3 |
or
x2 + (y + 1)2 = (x − 3)2 + y2
x2 + y2 + 2y + 1 = x2 − 6x + 9 + y2
so that,
or
2y = −6x + 8
y + 3x = 4
Hence the unit circle in the w plane, | w |= 1, corresponds to the line y + 3x = 4.
z−j
z+j
−wj − j
then z =
w−1
w + 1
.
so that | z |=
w−1
So if | z |= 2, | w + 1 |= 2 | w − 1 |
16
If w =
or (u + 1)2 + v2 = 4(u − 1)2 + 4v2
which simplifies to
5
u−
3
2
16
, a circle centre
+v =
9
2
5
4
, 0 and radius
3
3
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
17
If w = ejθ0
taking modulus
207
z − z0
then,
z − z∗0
z − z0
since | ejθ0 |= 1
| w |=
z − z∗0
If z is real (i.e. z is on the real axis) then
| z − z0 |=| z − z∗0 |= (x − x0 )2 + y20
1/2
and z0 = x0 + jy0
Hence | w |= 1. Thus the real axis in the z plane corresponds to the unit circle
| w |= 1 in the w plane. Making z the subject of the transformation gives
z=
wz∗0 − ejθ0 z0
w − ejθ0
Hence the origin in the w plane maps to z = z0 .
Thus the inside of the unit circle in the w plane corresponds to the upper half of
the z plane provided
Im{z0 } > 0
Since w = 0 maps to z = z0 and z0 = j and z = ∞ maps to w = ejθ0 = −1 it
gives θ0 = π.
18
For the transformation
w=
2jz
z+j
the fixed points are given by
2jz
z+j
2
z + jz = 2jz
z=
or
z(z − j) = 0, z = 0 or j
Hence circular arcs or straight lines through z = 0, j are transformed to circular
arcs or straight lines through w = 0, j by the properties of bilinear transformation
(section 4.2.4).
The inverse transformation is
z=
jw
2j − w
c Pearson Education Limited 2011
208
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
jw
1
1
1
1
|<
becomes
− <
which simplifies to | w − 1 |< 1 (use
2
2
2j − w 2
2
w = u + jv and split into real and imaginary parts).
1
1
4
2
Similarly, | z − |< becomes | w − |>
2
2
3
3
| z−
19
The general bilinear mapping is
w=
az + b
cz + d
if w = 0 corresponds to z = z0 then
w=
(z − z0 )ejθ0
cz + d
z − z0
= 1 and the inverse
If, additionally, | w |= 1 is mapped to | z |= 1 then
cz + d
of z0 is also mapped to the inverse of w = 0 that is, w = ∞.
Hence cz + d can be replaced by z∗0 z − 1 giving the mapping as
z − z0
jθ0
w=e
z∗0 z − 1
where θ0 is any real number.
Exercises 4.2.7
20
Under the mapping w = z2 , u = x2 − y2 , v = 2xy
It is not possible to achieve formulae of the type x = φ(u, v), y = ψ(u, v) , however
we can use u = x2 − y2 , v = 2xy to determine images. Points (0 + j0), (2 + j0)
and (0 + j2) transform to (0 + j0), (4 + j0) and (−4 + j0) respectively.
The positive real axis y = 0, x ≥ 0 transforms to the (positive) real axis
v = 0, u = x2 .
The positive imaginary axis x = 0, y ≥ 0 transforms to the (negative) real axis
v = 0, u = −y2 .
The line joining the point 2 + j0 to the point 0 + j2 has equation x + y = 2.
By using the equations u = x2 − y2 and v = 2xy we obtain
u = 4(1 − y), v = 2y(2 − y)
from which, eliminating y we get
8v = 16 − u2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
209
Hence we deduce the following picture:
21
Under the transformation w = z2 , u = x2 − y2 and v = 2xy .
Hence the line y = x transforms to u = 0, v > 0
and the line y = −x transforms to u = 0, v < 0.
2m
u.
The line y = mx transforms to v =
1 − m2
2m
= tan 2θ0 .
Putting m = tan θ0 ,
1 − m2
Hence y = x tan θ0 transforms to v = u tan 2θ.
Thus lines through the origin of slope θ0 in the z plane transform to lines through
the origin of slope 2θ0 in the w plane. Hence the angle between the lines through
the origin in the z plane is doubled by the transformation w = z2 .
22
w = zn
Writing z = rejθ , w = rn enjθ
22(a)
Circles | z |= r are transformed to circles | w |= rn
22(b)
Straight lines passing through the origin intersecting with angle θ0 are
θ = k and θ = k + θ0 . These are transformed to w = rn enjk and w = rn enj(k+θ0 )
that is, lines φ = nk and φ = nk + nθ0 as required.
1 + z2
1
z∗
=z+ =z+
z
z
| z |2
x
y
then u = x + 2
and v = y − 2
2
x +y
x + y2
23
If w =
c Pearson Education Limited 2011
210
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1
1
If | z |= r , then u = x 1 + 2 and v = y 1 − 2
r
r
x=
r2 u
r2 v
,
y
=
r2 + 1
r2 − 1
r2 u 2 r2 v 2
+ 2
= r2 (r = 1)
(I) .
Squaring and adding gives 2
r +1
r −1
If r = 1 and v = 0, | x |≤ 1 (because x2 = 1 − y2 ) and u = 2x . Hence the image
of the unit circle | z |= 1, that is, −2 ≤ u ≤ 2, v = 0 and the portion of the real
axis in the w plane is between −2 and +2.
| r2 − 1 |
1 + r2
and
minor
axis
if r is
The curves ( I) are ellipses, major axis
r2
r2
very large, and both of these quantities tend to 1. Hence the image curve I tends
to a circle u2 + v2 = r2 .
Exercises 4.3.3
24(a)
zez = (x + jy)ex+jy
= ex (x + jy)(cos y + j sin y)
= ex (x cos y − y sin y) + jex (y cos y + x sin y)
so u = (x cos y − y sin y)ex , v = (y cos y + x sin y)ex
We need to check the Cauchy–Riemann equations
∂u
∂x
∂u
∂y
∂v
∂x
∂v
∂y
= (x cos y − y sin y + cos y)ex
= (−x sin y − y cos y − sin y)ex
= (y cos y + x sin y + sin y)ex
= (−y sin y + cos y + x cos y)ex
∂v
∂u
∂v
∂u
=
and
=−
∂x
∂y
∂y
∂x
Thus the Cauchy–Riemann equations are valid and
Hence
d
(zez ) = (z + 1)ez
dz
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
24(b)
211
Following the same procedure as in (a), we deduce that sin 4z is analytic
with derivative 4 cos 4z .
This time, zz∗ = x2 + y2 which is real.
∂u
∂v
Obviously, therefore,
(= 2x) =
(= 0) .
∂x
∂y
Thus zz∗ is not analytic.
24(c)
24(d)
25
Similarly to part (a), cos 2z is analytic with derivative −2 sin 2z .
w = x2 + ay2 − 2xy + j(bx2 − y2 + 2xy) = u + jy
∂u
∂u
= 2x − 2y ,
= 2ay − 2x
∂x
∂y
∂v
∂v
= 2bx + 2y ,
= −2y + 2x
∂x
∂y
∂u ∂v
∂u
∂v
=− ,
=
.
∂x
∂y ∂y
∂x
The second is satisfied and the first only holds if a = −1, b = 1. Since w(z) =
w(x+jy) we simply put y = 0 which gives w(x) = x2 +jx2 and hence w(z) = z2 +jz2
dw
= 2(1 + j)z
and
dz
The Cauchy–Riemann equations are
26
With u = 2x(1 − y) = 2x − 2xy
∂u
∂u
= 2 − 2y ,
= −2x
∂x
∂y
The Cauchy–Riemann equations demand
∂v
∂v
= 2 − 2y ,
= 2x
∂y
∂x
Integrating and comparing, these give
v = x2 − y2 + 2y + C (take C = 0)
Form u + jv = 2x − 2xy + j(x2 − y2 + 2y) = w(z) .
c Pearson Education Limited 2011
212
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Since z = x + jy , if we put y = 0, we can find w(x) which will give the functional
form of w. Thus
w(x) = 2x + jx2
Hence w(z) = 2z + jz2
27
hence
φ(x, y) = ex (x cos y − y sin y)
∂φ
= ex (x cos y − y sin y + cos y)
∂x
∂φ
= ex (−x sin y − y cos y − sin y)
∂y
∂2 φ
= ex (x cos y − y sin y + 2 cos y)
∂x2
∂2 φ
= ex (−x cos y + y sin y − 2 cos y)
∂y2
∂2φ
∂x2
+
∂2φ
∂y 2
= 0 and φ is harmonic.
Writing z = φ(x, y) + jψ(x, y) , the Cauchy–Riemann equations demand
∂φ
∂ψ
=−
= ex (x sin y + y cos y + sin y)
∂x
∂y
∂ψ
∂φ
=
= ex (x cos y − y sin y + cos y)
∂y
∂x
Integrating
∂ψ
∂x
x
with respect to x (using integration by parts for the first term)
gives ψ = e (x sin y + y cos y) + f(y) . Examining φ(x, y) demands that f(y) = 0
because all terms will be multiplied by ex .
Hence w(z) = φ(x, y) + jψ(x, y) = ex (x cos y − y sin y) + jex (x sin y + y cos y) and
w(x + j0) = w(x) = xex . Hence w(z) = zez .
28
Here we have u(x, y) = sin x cosh y
so that
hence
so that ∇2 u =
∂2u
∂x2
+
∂u
∂u
= cos x cosh y and
= sin x sinh y
∂x
∂y
∂2 u
∂2 u
=
−
sin
x
cosh
y
and
= sin x cosh y
∂x2
∂y2
∂2u
∂y 2
= 0 and u is harmonic.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
213
Using the Cauchy–Riemann equations gives v = cos x sinh y so that u + jv =
w(z) = sin x cosh y + j cos x sinh y . Putting y = 0 gives w(x + j0) = sin x so that
w(z) = sin z .
29
The orthogonal trajectories of a family of curves φ(x, y) = α are ψ(x, y) = β
where φ and ψ are conjugate functions: that is, φ(x, y) + jψ(x, y) = w(z) which
is an analytic function.
Proceeding as in the previous examples.
1 4
3
(x + y4 ) − x2 y2
4
2
29(a)
If φ(x, y) = x3 y − xy3 then ψ(x, y) =
29(b)
1
If φ(x, y) = e−x cos y + xy then ψ(x, y) = e−x sin y + (x2 − y2 ) .
2
Hence the orthogonal trajectories are,
29(a)
x4 − 6x2 y2 + y4 = β , a constant
29(b)
2e−x sin y + x2 − y2 = β , a constant.
30(a)
z2 e2z
= (x + jy)2 e2(x+jy)
= (x2 − y2 + 2jxy)(e2x (cos 2y + j sin 2y))
= e2x ((x2 − y2 ) cos 2y − 2xy sin 2y) + je2x ((x2 − y2 ) sin 2y + 2xy cos 2y)
30(b)
sin 2z
= sin(2x + j2y)
= sin 2x cosh 2y + j cos 2x sinh 2y
Straightforward calculus reveals that both functions obey the Cauchy–Riemann
equations and are thus analytic. Their derivatives are (a) (2z2 + 2z)e2z and (b)
2 cos 2z respectively.
31
Writing w = sin−1 z we can say that
z = sin w = sin(u + jv) = sin u cosh v + j cos u sinh v
c Pearson Education Limited 2011
214
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so that, equating real and imaginary parts,
x = sin u cosh v
and y = cos u sinh v
Squaring and adding gives
x2 + y2 = sin2 u cosh2 v + cos2 u sinh2 v
= sin2 u cosh2 v + (1 − sin2 u)(cosh2 v − 1)
= sin2 u + cosh2 v − 1
from which
x2 + y2 + 1 = sin2 u +
x2
sin2 u
(I)
Solving for sin2 u gives
sin2 u =
1
1
(1 + x2 + y2 )2 − 4x2
(1 + x2 + y2 ) −
2
2
where the minus sign is taken, since with u = π/2 (i.e.
inconsistencies result otherwise. From cosh2 v =
cosh2 v =
2
x
sin2 u
we obtain
1
1
(1 + x2 + y2 ) +
(1 + x2 + y2 )2 − 4x2
2
2
(This is most easily found by solving equation (I) for
x2
sin2 u
x = cosh v, y = 0)
1
sin2 u
then using cosh2 v =
.)
Square rooting and inverting give u and v in terms of x and y . It can be shown
that the expression under the square root sign is positive, for 1 + x2 + y2 − 2x =
(x − 1)2 + y2 ≥ 0 for all real x and y thus (1 + x2 + y2 )2 ≥ 4x2 . Hence w = sin−1 z
is an analytic function with derivative
32
√ 1
1−z 2
.
| sin z |2 =| sin x cosh y + j cos x sinh y |2
= sin2 x cosh2 y + cos2 x sinh2 y
= cosh2 y − cos2 x = sinh2 y + sin2 x
The result follows immediately from the last two expressions.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
215
Exercises 4.3.5
33
Mappings are not conformal at the points where
dw
=0
dz
dw
= 2z = 0 when z = 0. z = 0 is the only point where the mapping
dz
fails to be conformal.
33(a)
dw
= 6z2 − 42z + 72 = 0 when z2 = 7z + 12 = 0 that is, non-conformal
dz
points are z = 4, z = 3 (both real).
33(b)
33(c)
points.
34
√
1
1 1±j 3
dw
1
3
= 8 − 3 = 0 when z =
giving ,
as non-conformal
dz
z
8
2
4
Proceeding as in Example 4.13, the mapping
w=z−
1
z
has a fixed point at z = ∞, is analytic everywhere except at z = 0 and conformal
dw
=0
except where
dz
1
that is, 1 + 2 = 0, z = ±j
z
Since both of these occur on the imaginary axis, consideration of this axis is
adequate to completely analyse this mapping.
The image of z = j is w = 2j , and the image of z = −j is w = −2j . Writing
z = j + jε, ε real, we find that
1
j + jε
= j[1 + ε − (1 + ε)−1 ]
w = jε −
= j[1 + ε + 1 − ε + ε2 + . . .]
j[2 + ε2 ]
So, no matter whether ε > 0 or ε < 0, the image point of z = j + jε is above
w = j2 on the imaginary axis.
c Pearson Education Limited 2011
216
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
that is, points Q and P in the z plane both map to R in the w plane in a manner to
Example 4.13, the non-conformality of z = ±j is confirmed and as the imaginary
axis (in the z plane) is traversed from −jz to 0, the imaginary axis (in similar the
w plane) is traversed from −jz to −j2 and back to −j∞ (when z = −j, w reaches
−j2). Similarly, as the imaginary axis (in the z plane) is traversed from +j∞ to
0, the imaginary axis (in the w plane) is traversed from +j∞ to +j2 and back to
+j∞ again.
Finally, points on the imaginary axis in the w plane such that w = aj, −2 < a < 2,
do not arise from any points on the imaginary axis in the z plane. This point is
obvious once one solves
aj = z −
to obtain
z=
35
1
z
1√
1
aj ±
4 − a2
2
2
If w = ez
then u = ex cos y and v = ex sin y
Hence the expressions u2 + v2 = e2x and v = u tan y can be derived.
35(a)
0 ≤ x < ∞ is mapped to the exterior of the unit circle u2 + v2 = 1
35(b)
0 ≤ x ≤ 1 is mapped to the annulus 1 ≤ u2 + v2 ≤ e2
0 ≤ y ≤ 1 is mapped to the region between the radiating lines v = 0 and
v = u tan 1.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
217
1
35(c)
2 π ≤ y ≤ π is mapped to the region between u = 0(v > 0) and
v = 0(u < 0)
that is,
Thus if 0 ≤ x < ∞ then the image region in the w plane is in the shaded quadrant,
but outside the unit circle.
36
If w = sin z then
dw
dz
= cos z .
Since cos z = 0 when z = (2n + 1)π/2 these are the points where the mapping is
not conformal
w = sin z ⇒ u + jv = sin x cosh y + j cos x sinh y
c Pearson Education Limited 2011
218
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Hence v = sin x cosh y, v = cos x sinh y
2
2
thus lines x = k transform to sinu k − cosv k = 1 (hyperbolae)
u 2 v 2
+ sinh l = 1 (ellipses)
and lines y = k transform to cosh
l
a2
jθ
ζ and ζ = Re
2
then z = Rejθ + aR e−jθ
2
so that x = R + aR cos θ and y
37
If z = ζ +
= R−
a2
R
sin θ.
If R = a, x = 2a cos θ and y = 0 then the real line between ±2a is traversed.
Length of line segment = 4a.
For a circle of radius b,
a2
a2
cos θ, y = b −
sin θ
x= b+
b
b
Hence the image in the z plane is an ellipse of the form
b2 x2
b2 y2
+ 2
=1
(a2 + b2 )2
(b − a2 )2
Exercises 4.4.2
38(a)
38(b)
1
z −1
z z 2
= (z − j)−1 = j 1 −
=j 1+ +
+ ...
z−j
j
j
j
2
= j + z − jz − z3 + jz4 . . .
j −1
j j 2
1
1
1
= 1+ +
+ ...
= 1−
z−j
z
z
z
z
z
j
1
j
1
= + 2 − 3 − 4 + ...
z z
z
z
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
38(c)
In order that | z − 1 − j |<
219
√
2 we write
1
1
=
= (1 + z − 1 + j)−1
z−j
z−1−j+1
= 1 − (z − 1 − j) + (z − 1 − j)2 − (z − 1 − j)3 + . . .
Which is valid inside | z − 1 − j |<| 1 − j |=
√
2.
39
1
= (z2 + 1)−1 = 1 − z2 + z4 − z6 + . . .
z2 + 1
where | z |< 1.
Using the fact that we can differentiate power series term-by-term and the radius
of convergence remains unaltered
−
(z2
2z
= −2z + 4z3 − 6z5 + . . .
2
+ 1)
so
39(a)
1
= 1 − 2z2 + 3z4 − 4z6 + 5z8 + . . .
2
2
(z + 1)
| z |< 1
Operating on
(z2
1
in a similar fashion gives
+ 1)2
−
4z
= −4z + 12z3 − 24z5 + 40z7 . . .
(z2 + 1)3
so
39(b)
(z2
1
= 1 − 3z2 + 6z4 − 10z6 + 15z8 + . . .
+ 1)3
| z |< 1
c Pearson Education Limited 2011
220
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 4.4.4
40
Taylor’s theorem is
f(z) = f(a) + (z − a)f (a) +
(z − a)2
f (a) + . . .
2!
We thus compute f(z) and its first few derivatives then evaluate them at z = a.
40(a)
f(z) =
1
1
2
6
, f (z) = −
,
f
(z)
=
and
f
(z)
=
−
1+z
(1 + z)2
(1 + z)3
(1 + z)4
Hence
f(1) =
1
1
1
2
3
, f (1) = − , f (1) = = and f (1) = −
2
4
8
4
8
thus
1 1
1
1
1
= − (z − 1) + (z − 1)2 − (z − 1)3 + . . .
1+z
2 4
8
16
The radius of convergence is the distance between the nearest singularity of f(z)
to the point about which the expansion is made. The point z = −1 is the only
singularity and the distance between this and z = 1 is 2 (along the real axis).
40(b)
f(z) =
f (z) =
f (z) =
f (z) =
fiv (z) =
fv (z) =
j 1
1
1
=
−
using partial fractions
z(z − 4j)
4 z z − 4j
1
j 1
1
− 2+
; f (1) = 0
f(1)
=
4 z
(z − 4j)2
4
j 2
2
1
− 3+
; f (1) = 0
f
(1)
=
−
4 z
(z − 4j)3
8
j 6
6
3 v
iv
− 4+
f
(1)
=
+
; f (1) = 0
4 z
(z − 4j)4
8
j 24
24
45
vi
− 4 +
f
(1)
=
−
4
z
(z − 4j)5
16
j 120
120
− 6 +
etc.
4
z
(z − 4j)6
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
221
Thus
1
1
1
1
1
= − (z − 2j)2 + (z − 2j)4 −
(z − 2j)6 + . . .
z(z − 4j)
4 16
64
256
The radius of convergence is 2 since z = 2j is 2 from the singularities at z = 0 and
z = 4j.
40(c)
f(z) =
1
z2
gives f (z) = − z23 , f (z) =
6
z4
and f (z) = − 24
z5 .
Putting z = 1 + j gives
j
2
1+j
f(1 + j) = − , f (1 + j) = −
=
3
2
(1 + j)
2
6
3
24
f (1 + j) =
= and f (1 + j) = −
= −3(1 − j)
4
(1 + j)
2
(1 + j)5
Hence
1
j
1
1
3
= − + (1 + j)(z − 1 − j) + (z − 1 − j)2 − (1 − j)(z − 1 − j)3 + . . .
2
z
2 2
4
2
The radius of convergence is the distance between the origin (a double pole) and
√
1 + j that is, 2.
1
1 + z + z2
we could use the binomial expansion
41
With f(z) =
f(z) = (1 + z + z2 )−1 gathering terms to O(z3 )
This is certainly more efficient than using the derivatives of f(z) . However, the
best way is to use the fact that (z3 − 1) = (z − 1)(z2 + z + 1) . That is
1
1−z
z−1
=
= 3
2
1+z+z
z −1
1 − z3
= (1 − z)(1 − z3 )−1
c Pearson Education Limited 2011
222
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
= 1 − z + z3 . . .
to order z3
valid in the region | z |< 1.
42
If f(z) =
1
z 4 −1
the singularities are at the points where z4 = 1 that is,
z = 1, −1, −j and j. The radii of convergence are the minimum distances of the
points z = 0, 1 + j and 2 + j2 from these singularities.
z = 0 is equidistant (1) from each radius of convergence = 1.
z = 1 + j is distance (1) from z = 1 and z = j with radius of convergence = 1.
z = z + j2 is a distance | 2 + j2 − 1 | from 1 and a distance | 2 + j2 − j | from j.
√
Both of these distances = [22 + (2 − 1)2 ]1/2 = 5.
43
If f(z) = tan z then f (z) = sec2 (z) and f = 2 sec2 z tan z but subsequent
derivatives get cumbersome to compute (except by using a Computer Algebra
sin z
, we can use the series for sin z and cos z as follows:
package). Since tan z =
cos z
tan z =
z3
z5
6 + 120
2
4
− z2 + z24
z−
1
−1
z2
z2
z4
z4
1−
=z 1−
+
−
6
120
2
24
2
z4
z2
z4 z2
z4
z2
+
1+
−
+
−
+ ...
=z 1−
6
120
2
24
2
24
1
1
1 5
1
1 3
−
−
+
z + ...
=z+ z +
3
120 12 24 4
2
1
tan z = z + z3 + z5 + · · ·
3
15
Since z = π/2 is the closest singularity, the radius of convergence is
π
2
.
Exercises 4.4.6
1
has a simple pole at z = 0 and a double pole at
z(z − 1)2
z = 1. In order to find the Laurent expansions, we simply find the following
44
The function
binomial expansions
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
223
1
1
(1 − z)−2 = (1 + 2z + 3z2 + 4z3 + . . .)
z
z
1
= + 2 + 3z + 4z2 + . . . about z = 0
z
valid for 0 <| z |< 1
and
1
(1−(1 − z))−1
(z − 1)2
1
=
[1 + (1 − z) + (1 − z)2 + (1 − z)3 + . . .]
(z − 1)2
1
1
+ 1 + (1 − z) + (1 − z)2 + . . .
+
=
2
(1 − z)
1−z
valid for 0 <| 1 − z |< 1
With f(z) = z2 sin z1 , there is a singularity at z = 0 and another at
z = ∞. Expanding sin z1 as a power series in z1 we find
1
1
1
2
2 1
=z
+
− ···
z sin
−
z
z 3!z3
5!z5
1
1
+
− ···
=z−
3!z 5!z3
1
11
= ··· +
−
+ z.
3
5!z
3! z
45(a)
Since the principal part is infinite, there must be an essential singularity at z = 0.
(b)
in order to investigate z = ∞ we obtain
1
1
w5
w3
1
2
= 2 sin w = 2 w −
+
− ···
z sin
z
w
w
3!
5!
1
w w3
= − +
− ···
w 3!
5!
1
1
+ 3 − ···
=z−
z3! z 5!
Writing z =
1
w
which implies a simple pole at z = ∞. (The expansion is the same as that about
z = 0, but re-interpreted.)
At any other point z2 sin z1 is regular and has a Taylor series of the
form f(z) = a0 + a1 z + a2 z2 + . . . . specifically, about z = a, z2 sin z1 =
(c)
a2 sin a1 + a1 z + a2 z2 + . . . where a1 = f (a), a2 = f (a) , etc.
c Pearson Education Limited 2011
224
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
46
With f(z) =
z
(z−1)(2−z)
there are simple poles at z = 1 and z = 2.
46(a)
Inside the unit circle | z |= 1, therefore there is a Taylor series
z(1 − z)−1 (2 − z)−1
z −1
z
= (1 − z)−1 1 −
2
2
z
z z 2 z 3
2
3
= (1 + z + z + z + . . .) 1 + +
+
+ ···
2
2
2
2
z 3 1 3 1 3 1 3
z 3 2 z 2 1 2 1 2
z + z + z + ... +
z + z + z + z + ...
= + z +
2 4
2
2
4
2
2
4
8
3
7
15
1
= z + z2 + z3 + z4 + · · · | z |< 1
2
4
8
16
46(b)
In the annulus 1 <| z |< 2 we rearrange f(z) to obtain a Laurent series
as follows
−1
−1
1
1
z
2
1
z
1−
−
=
−
=− 1−
(z − 1)(z − 2)
z−2 z−1
2
z
z
2
1
1
1
z z
+ ··· −
1 + + 2 + ···
=− 1+ +
2
4
z
z z
1
1
1
z z2
= ... − 3 − 2 − − 1 − −
− ···
z
z
z
2
4
46(c)
For | z |> 2 we rearrange as follows
z
2
1
=
−
(z − 1)(z − 2)
z−2 z−1
2 −1 1
1 −1
2
= 1−
− 1−
z
z
z
z
2
4
1
1
2
8
1
1
1 + + 2 + 3 + ··· −
1 + + 2 + 3 + ···
=
z
z z
z
z
z z
z
3
7
15
1
= + 2 + 3 + 4 + ···
z z
z
z
46(d)
If w =
For | z − 1 |> 1 we write
1
z−1
then wz − w = 1 or z
1
z−1 =
= 1+w
w
w and find a Taylor’s series in w.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so that
225
z
1+w
=w
(z − 1)(z − 2)
1−w
= w(1 + w)(1 + w + w2 + w3 + . . .)
= w + 2w2 + 2w3 + . . .
2
2
1
+
+
+ ···
=
2
z − 1 (z − 1)
(z − 1)3
46(e)
For 0 <| z − 2 |< 1 we write w = z − 2 hence
z
w+2
2
=
= 1+
(1 + w)−1
(z − 1)(z − 2)
w(w + 1)
w
2
(1 − w + w2 − w3 + . . .)
= 1+
w
2
= − 1 + w − w2 + w3 − . . .
w
2
− 1 + (z − 2) − (z − 2)2 + (z − 2)3 . . .
=
(z − 2)
Exercises 4.5.2
47
The point at infinity is ignored in this question. Most if not all can be found
immediately by inspection.
47(a)
47(b)
cos z
:
z2
double pole at z = 0, zeros whenever cos z = 0 that is,
z = 12 (2n + 1)π, n = integer.
1
: has a double pole at z = −j , a simple pole at z = j
(z + j)2 (z − j)
and no zeros in the finite z plane.
47(c)
z
: simple poles at z4 = 1 that is, z = 1, −1, j, −j and a zero at
4
z −1
z = 0.
47(d)
cosh z : since coth z =
cosh z
this has simple poles at those points where
sinh z
z = jnπ and zero at those points where z = 12 j(2n + 1)π, n =
integer.
c Pearson Education Limited 2011
226
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
47(e)
sin z
: simple poles at z = ±jπ and zeros at z = nπ, n = integer.
z2 + π2
47(f)
ez/(1−z) : this has an essential singularity at z = 1 and no zeros.
47(g)
47(h)
47(i)
z−1
: this has simple poles at z = ±j and a zero at z = 1.
z2 + 1
z+j
: this has a triple pole at z = −2, a simple pole at z = 3
(z + 2)3 (z − 3)
and a zero at z = −j.
z2 (z2
1
: this has simple poles at z2 − 4z + 5 = 0, that is,
− 4z + 5)
z = 5, −1 and a double pole at z = 0.
1 − cos z
. In order to investigate this, we expand cos z . Only z = 0 is a
z2
possible (finite) singularity
48(a)
1− 1−
z2
2!
+
z2
z4
4!
− ···
=
z2
1
−
+ ···
2!
4!
The RHS is a power series, thus the singularity at z = 0 is removable.
2
48(b)
ez
2
. Using the power series for ez gives the expansion
3
z
2
ez
1
z4
2
+ ···
= 3 1+z +
z3
z
2!
z
1
1
= 3 + + + ···
z
z 2!
z = 0 is thus a pole of order 3.
48(c)
1
z 1+
1
1
z cosh z .
1
1
z 2 2! + z 4 4! +
Obviously the point z = 0 is a problem.
· · · is the Laurent series which indicates that z = 0 is an
essential singularity.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
48(d)
227
tan−1 (z2 + 2z + 2) . For this problem the easiest way to proceed is to find
the Maclaurin series from first principles. At z = 0, tan−1 (z2 + 2z + 2) = tan−1 2
which is finite. This means that z = 0 is a regular point, hence it is not actually
necessary to find the Laurent series (in this case Maclaurin series) for the function.
In fact
6
2
tan−1 (z2 + 2z + 2) = tan−1 2 + z − z2 + · · ·
5
25
49
If f(z) =
p(z)
q(z)
where p(z) and q(z) are polynomials, then the only singularities
of f(z) are the algebraic zeros of q(z) . These zeros are either distinct or multiple.
The distinct zeros give rise to simple poles of f(z) whereas the multiple zeros give
rise to poles of higher order. f(z) can only have these kinds of singularity, although
it may have none if q divides p, so that f(z) is polynomial. f(z) therefore cannot
have an essential singularity.
Exercises 4.5.4
2z + 1
2z + 1
=
hence the singularities are simple poles at
− z − 2)
(z − 2)(z + 1)
z = 2, z = −1.
50(a)
(z2
Using the formula residue = lim [(z − z0 )f(z)] the residues are
z→z0
at z = −1.
5
3
at z = 2 and
1
has a simple pole at z = 1 and a double pole at z = 0.
− z)
1
The residue at z = 1 is lim − 2 = −1
z→1
z
50(b)
z2 (1
1
1
1
1
= 2 (1 + z + z2 + . . .) = 2 + + 1 + · · ·
− z)
z
z
z
z2 (1
Hence the residue at z = 0 is 1.
c Pearson Education Limited 2011
1
3
228
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
50(c)
3z2 + 2
3z2 + 2
=
(z − 1)(z2 + 9)
(z − 1)(z + j3)(z − j3)
Hence there are simple poles at z = 1, j3, −j3
5
1
3+2
=
=
(1 + j3)(1 − j3)
10
2
−3 × 9 + 2
−25
5
at z = j3, residue =
=
=
(3 − j)
(j3 − 1)j6
6(−3 − j)
12
5
(3 + j) by symmetry.
at z = −j3, residue =
12
At z = 1, residue =
z3 − z2 + z − 1
z3 − z2 + z − 1
(z − 1)(z2 + 4)
=
=
z3 + 4z
z(z + j2)(z − j2)
z(z + j2)(z − j2)
which has simple poles at z = 0, j2, −j2.
50(d)
At z = 0, residue = −
1
4
(z − 1)(z2 + 4)
3
= (−1 + 2j)
z→j2
(z + j2)
8
3
at z = −j2, residue = (−1 − 2j) similarly.
8
at z = j2, residue = lim
z6 + 4z4 + z3 + 1
has a pole of order 5 at z = 1.
(z − 1)5
The formula for calculating residues is convenient for this problem.
50(e)
d4
1
lim 4 (z6 + 4z4 + z3 + 1)
4! z→1 dz
1
(6.5.4.3 + 4.4.3.2)
=
24
456
= 19
=
24
Residue =
50(f)
z + 1 2
has a double pole at z = 1.
z−1
d
(z + 1)2 = 4
z→1 dz
Residue = lim
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
50(g)
229
z+1
has a simple pole at z = −3 and a double pole at z = 1.
(z − 1)2 (z + 3)
1
8
1
d z + 1
=
Residue at z = 1 is
z=1
dz z + 3
8
Residue at z = −3 is −
3 + 4z
3 + 4z
=
has simple poles at z = −2, −1 and 0.
2
+ 3z + 2z
z(z + 1)(z + 2)
5
3
Residues are, respectively, − , 1 and
following the same procedure as part (c).
2
2
50(h)
z3
cos z
at z = 0 is simple, thus the residue is cos(0) = 1.
z
51(a)
The pole of
51(b)
The poles of
sin z
are all simple, and the residue at z = eπj/3 is
+ z2 + 1
z − eπj/3
sin z
πj/3
πj/3
= sin e
lim (z − e
) 4
lim
z + z2 + 1
z4 + z2 + 1
z→eπj/3
z→eπj/3
z4
Using L’Hôpital’s rule, the limit is
1
+ 2eπj/3
√
1
1
1
√ =
√ =
(−3 − j 3)
=
1
12
−3 + j 3
−4 + 2 2 + j 23
4eπj
√
√
1
(3 + j 3) sin 12 (1 + j 3)
giving the residue − 12
51(c)
The pole of
part. The residue is
z4 − 1
at z = eπj/4 is simple and we proceed as in the last
z4 + 1
z − eπj/4
− 2 lim
z4 + 1
z→eπj/4
1
j
1
1 1
1
√ +√ .
=
= −2 3πj/4 = −
2 − √1 + j √1
2
4e
2
2
2
2
√
Hence the residue is
2
4 (1
+ j) .
c Pearson Education Limited 2011
230
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
51(d)
z
has a simple pole at z = π
sin z
Residue = π lim
z→π
51(e)
(z2
z − π
= −π.
sin z
1
has a double pole at z = j
+ 1)2
d
1
(z − j)2
2
z→j dz
(z − j) (z + j)2
2
= lim −
z→j
(z + j)3
2
j
=− 3 =− .
8j
4
Residue = lim
52(a)
cos z
has a triple pole at z = 0.
z3
1
cos z
1
1
+ z − ···
= 3−
3
z
z
2z 24
52(b)
residue = −
1
2
z2 − 2z
has a double pole at z = −1.
(z + 1)2 (z2 + 4)
d (z + 1)2 (z2 − 2z)
at z = −1
dz (z + 1)2 (z2 + 4)
(2z − 2)(z2 + 4) − 2z(z2 − 2z)
14
=
=−
2
2
z=−1
(z + 4)
25
Residue =
52(c)
The function
ez
has a double pole wherever
sin2 z
sin z = 0 that is at z = nπ, n = an integer
In order to find the residue, we need to compute
lim
z→nπ
d (z − nπ)2 ez
dz
sin2 z
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Now
d (z − nπ)2 ez
dz
sin2 z
= ez
231
d (z − nπ)2
(z − nπ)2 z
+
e
dz
sin2 z
sin2 z
(I)
and
d (z − nπ)2
z − nπ d z − nπ
=2
·
dz
sin z dz sin z
sin2 z
2(z − nπ) sin z − (z − nπ) cos z
=
·
sin z
sin2 z
z − nπ
→ 1 and
As z → nπ,
sin z
cos z − cos z + (z − nπ) sin z
sin z − (z − nπ) cos z
(using L’Hôpital’s rule)
→
2
2 sin z cos z
sin z
→ 0 as z → nπ
Hence the RHS of equation (I) → enπ as z → nπ.
Thus the residue is enπ .
Exercises 4.6.3
53
(z2 + 3z)dz with z = x + jy and dz = dx + jdy
C
hence (z2 + 3z)dz = (x2 − y2 + j2xy + 3x + j3y)(dx + jdy)
= (x2 − y2 + 3x)dx − (2xy + 3y)dy
+ j[(x2 − y2 + 3x)dy + (2xy + 3y)dx]
53(a)
The straight line joining 2 + j0 to 0 + j2 has equation x + y = 2 in
Cartesian coordinates. This has parametric equation x = t and y = 2 − t from
which dx = dt and dy = −dt , and using the above expression for (z2 + 3z)dz
(z2 + 3z)dz = (t2 − (2 − t)2 + 3t)dt + (2t(2 − t) + 3(2 − t))dt
+ j[−(t2 − (2 − t)2 + 3t)dt + (2t(2 − t) + 3(2 − t))dt]
and the range of integration is from t = 2 to t = 0.
c Pearson Education Limited 2011
232
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Hence
0
(8t − 2t2 + 2)dt
2
(z + 3z)dz =
C
2
0
(−6t − 2t2 + 10)dt
+j
2
2
= 4t2 − t3 + 2t
3
8
44
so
(z2 + 3z)dz = − − j
3
3
C
53(b)
2
+ j −3t2 − t3 + 10t
3
2
0
0
2
On the straight line from 2 + j0 to 2 + j2, x = 2 and y goes from 0 to
2, so that dx = 0.
Therefore
2
−(4t + 3t)dt
2
(z + 3z)dz =
C1
0
2
(4 − t2 + 6)dt
+j
0
2
1
+ j 10t − t3
3
0
52
= −14 + j
3
7
= − t2
2
2
0
On the straight line from 2 + j2 to 0 + j2, y = 2 and x goes from 2 to 0, so that
dy = 0.
Therefore
0
(t2 − 4 + 3t)dt
2
(z + 3z)dz =
C2
2
0
+j
(4t + 6)dt
2
1 3
3
t − 4t + t2
3
2
2
= − − j14
3
0
=
Thus
C
(z2 + 3z)dz =
C1
(z2 + 3z)dz +
+ j 2t2 + 3t
2
(z2 + 3z)dz = −
C2
c Pearson Education Limited 2011
0
2
8
44
−j .
3
3
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
53(c)
233
For this part, we use z = 2ejθ on | z |= 2 and θ varies between 0 and π/2
on the quarter circle joining 2+j0 to 0+j2. Thus (z2 +3z)dz = (4e2jθ +6ejθ )2jejθ dθ
so that
π/2
2
8je3jθ + 12je2jθ dθ
(z + 3z)dz =
0
C
π/2
8 3jθ
e + 6e2jθ
=
3
0
8
8
+6
= − −6 −
3
3
8j 44
=− −
3
3
Hence the integrals are all the same.
54(a)
On | z |= 1, z = ejθ , 0 ≤ θ ≤ 2π
so that (5z4 − z3 + 2)dz
2π
(5e4jθ − e3jθ + 2)jejθ dθ
=
0
=
5 5jθ 1 4jθ
e − e + 2ejθ
5
4
2π
=0
0
hence e2πj = e0 = 1
54(b)
Integrating around the square in the order 0 + j0, 1 + j0, 1 + j1 and
0 + j1 gives the answers
j
11
11
4 , 3 + 4, − 4
and − 3 − 4j . Adding these together gives
0.
54(c)
On the parabola y = x2 , x = t and y = t2 so that z = t + jt2 and
dz = (1 + 2jt)dt .
On the parabola y2 = x, x = t2 and y = t so that z = t2 + jt and dz = (2t + j)dt .
The computation of (5z4 −z3 +2)dz is extremely long winded but straightforward
C
and gives the answer 0.
55
In order to evaluate
C
dz
(z − z0 )n
c Pearson Education Limited 2011
234
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
we surround the point z = z0 with a circle of radius ε on which z = z0 + εejθ ,
0 ≤ θ < 2π.
Using equation (4.45) the integral around C is the same as the integral around the
circle on which z = z0 + εejθ dθ. Thus
C
dz
=
(z − z0 )n
2π
0
jεejθ
dθ
εn enjθ
If n = 1, then the integral integrates to
jε
(1−n)
e(1−n)jθ
(1 − n)j
2π
=0
0
as in Example 4.30.
If n = 1,
C
56(a)
dz
=
(z − z0 )n
C
jdθ = 2πj
0
dz
=0
z−4
If z = 4 is inside C, by problem 55
C
57
2π
If z = 4 is outside C, by Cauchy’s theorem,
56(b)
dz
= 2πj
z−4
In order to use Cauchy’s integral theorem, we split into partial fractions
2/5
4/5
2z
=
+
(2z − 1)(z + 2)
2z − 1 z + 2
57(a)
If C is the circle | z |= 1
C
2zdz
dz
dz
2
4
=
+
(2z − 1)(z + 2)
5 C 2z − 1 5 C z + 2
2
4
= · 2πj + · 0
5
5
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
since z =
=
1
2
is inside | z |= 1 whereas z = −2 is outside. Hence
4
πj
5
57(b)
C
2zdz
(2z − 1)(z + 2)
If C is the circle | z |= 3, both singularities (poles) are inside C, and
hence
2zdz
2
4
= 2πj + · 2πj
(2z − 1)(z + 2)
5
5
12
πj
=
5
C
58
This follows a pattern similar to Exercise 57.
Using partial fractions gives
5z
=
(z + 1)(z − 2)(z + 4j)
58(a)
1
− 4j)
(1 − 2j)
+ 3
+
z+1
z−2
5
51 (−1
2
17 (−2
+ 9j)
z + 4j
Only the first two poles (z = −1, z = 2) are inside | z |= 3 hence
C
58(b)
5zdz
5
1
= 2πj
(−1 − 4j) + (1 − 2j)
(z + 1)(z − 2)(z + 4j)
51
3
4π
(9 + 2j)
=
17
All three poles are inside | z |= 5 hence
C
5zdz
5
1
= 2πj
(−1 − 4j) + (1 − 2j)
(z + 1)(z − 2)(z + 4j)
51
3
2
+ (−2 + 9j)
17
=0
59
235
Equation (4.48) gives the general form of Cauchy’s integral theorem
C
f(z)
2πj (n)
f (z0 )
dz =
n+1
(z − z0 )
n!
where C is a contour enclosing the point z = z0 .
c Pearson Education Limited 2011
236
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
59(a)
C
z3 + z
1
dz
=
(2z + 1)3
8
z3 + z
1 3 dz
C (z + 2 )
1 2πj d2
=
(z3 + z)(z = − 12 is inside | z |= 1)
8 2 dz2 1
−2
1
3πj
π
j 6(− ) = −
8
2
8
=
59(b)
First of all we need to separate the integrand using partial fractions
4
4
8
4z
9
9
3
−
+
=
(z − 1)(z + 2)2
z − 1 z + 2 (z + 2)2
Hence
C
4zdz
4
4
= 2πj − 2πj + 0 = 0
2
(z − 1)(z + 2)
9
9
using Cauchy’s integral theorem (the derivative of
8
3
is of course zero). All poles
of the integrand are inside the circle | z |= 3.
Exercises 4.6.6
60
z2
z
has poles at z = ±j
+1
z
1
is regular inside | z |=
+1
2
zdz
1
= 0 if C is the circle | z |=
2
2
C z +1
60(a)
Since
60(b)
The residues of
z2
z
at z = ±j (both inside | z |= 2) are
z2 + 1
lim
(z − j)z
z
1
= lim
=
(z + j)(z − j) z→+j z + j
2
lim
(z + j)z
z
1
= lim
=
(z + j)(z − j) z→−j z − j
2
z→+j
and
z→−j
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
237
Hence, using the residue theorem
C
zdz
1 1
= 2πj
= 2πj
+
z2 + 1
2 2
z2 + 3jz − 2
are at z3 + 9z = 0, that is, z = 0, 3j, −3j
z3 + 9z
(all simple poles). Only z = 0 is inside | z |= 1 but all three are inside | z |= 4.
61
The singularities of
Hence we shall find all the residues.
z2 + 3jz − 2
2
=−
At z = 0, residue is lim
2
z→0
z +9
9
At z = 3j, the residue is
(z − 3j)(z2 + 3jz − 2)
z→3j
z(z − 3j)(z + 3j)
−9 − 9 − 2
10
(3j)2 + 3j3j − 2
=
=
=
3j(3j + 3j)
−18
9
lim
At z = −3j , the residue is
(z + 3j)(z2 + 3jz − 2)
z→−3j
z(z − 3j)(z + 3j)
−9 + 9 − 2
1
(−3j)2 + (3j)(−3j) − 2
=−
=
=
(−3j)(−3j − 3j)
−18
9
lim
61(a)
For this part, since only the residue at z = 0 is inside C(| z |= 1)
∴
C
61(b)
2
z2 + 3jz − 2
4πj
dz = 2πj − = −
3
z + 9z
9
9
For this part, all residues need to be taken into account since all the poles
of f(z) are inside C(| z |= 4)
∴
C
2 10 1
z2 + 3jz − 2
= 2πj
dz
=
2πj
− +
+
z3 + 9z
9
9
9
Note that in this case, all the zeros of the denominator were obviously poles. In
general, we would need to check if they were not removable by factorizing the
numerator.
c Pearson Education Limited 2011
238
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
62
f(z) =
√
(z2 + 2)(z2 + 4)
has
poles
at
z
=
±j,
z
=
±j
6.
(z2 + 1)(z2 + 6)
Residue at z = j is
lim
z→j
(z − j)(z2 + 2)(z2 + 4)
(−1 + 2)(−1 + 4)
3j
=
=−
2
(z − j)(z + j)(z + 6)
2j(−1 + 6)
10
Residue at z = −j is
lim
z→−j
(z + j)(z2 + 2)(z2 + 4)
(−1 + 2)(−1 + 4)
3j
=
=
2
(z + j)(z − j)(z + 6)
(−2j)(−1 + 6)
10
√
Residue at z = j 6 is
√
(z − j 6)(z2 + 2)(z2 + 4)
(−6 + 2)(−6 + 4)
8
√
√
√
lim√
=
= √
2j 6(−6 + 1)
2j 6(−5)
z→j 6 (z − j 6)(z + j 6)(z2 + 1)
√
2
=
j 6
15
√
2 √
Residue at z = −j 6 is thus = − j 6
15
The circle | z |= 2 contains the poles at z = ±j but not those at
√
3j
3j
= 0.
z = ±j 6 . The sum of the residues inside C = − +
10 10
Hence the integral = 0.
62(a)
62(b)
Hence
The circle | z − j |= 1 contains the residue only at z = j.
C
62(c)
3j 3π
(z2 + 2)(z2 + 4)
dz
=
2πj
−
=
(z2 + 1)(z2 + 6)
10
5
The circle | z |= 4 contains all the poles. Since the sum of the residues is
zero, so is the integral.
63
The function
1
has double poles at z = 0 and z = ±j .
z2 (1 + z2 )2
Residue at z = 0 is
d
1
4z
=−
=0
dz (1 + z2 )2 z=0
(1 + z2 )3 z=0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
239
Residue at z = j is
d
1
= − 2(2z + j) = − 3 j
2
2
dz z (z + j) z=j
(z2 + jz)3
4
Residue at z = −j is
63(a)
3
j
4
If C is the circle | z |=
C
63(b)
1
, only the residue at z = 0 is in C. Thus
2
dz
= 2πj(0) = 0
z2 (1 + z2 )2
All the singularities are inside | z |= 2, but since they sum to 0,
C
z2 (1
dz
=0
+ z2 ) 2
3z2 + 2
are at z = 1, ±2j .
64(a) The singularities of
(z − 1)(z2 + 4)
They are all simple poles. Using the formula (4.37) the residues are :at z = 1 : 1
1
(2 − j)
8
1
at z = −2j : (2 + j)
8
at z = 2j :
(i) If C is | z − 2 |= 2 only the residue at z = 1 is included, hence
C
3z2 + 2
dz = 2πj
(z − 1)(z2 + 4)
(ii) If C is | z |= 4, all the residues are included, hence
C
3z2 + 2
5
dz = πj
2
(z − 1)(z + 4)
2
z2 − 2z
are at z = −1, ±2j . A double pole
(z + 1)2 (z2 + 4)
is at z = −1 and simple poles are at z = ±2j.
64(b)
The singularities of
c Pearson Education Limited 2011
240
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Residues are:-
14
25
at z = 2j : −1 + j
at z = −1 : −
at z = −2j : −1 − j
(i) If C is | z |= 3, all singularities are inside C.
Hence
C
z2 − 2z
14
128
dz = 2πj − − 2 = −
πj
2
(z + 1) (z + 4)
25
25
(ii) If C is | z + j |= 2 then z = −1 and z = −2j are inside C, but z = 2j is
not.
Hence
C
64(c)
z2 − 2z
14
2π
dz = 2πj − − 1 − j =
(25 − j39)
2
2
(z + 1) (z + 4)
25
25
The function
(z +
1)3 (z
1
− 1)(z − 2)
Simple poles at z = 1, 2, triple poles at z = −1.
Residues :-
1
z=1 : −
8
1
z=2 :
27
1
19
1
− =−
z = −1 :
27 8
216
1
(i) The circle | z |= 2 contains none of the singularities and therefore
C
dz
=0
(z + 1)3 (z − 1)(z − 2)
(ii) The circle | z + 1 |= 1 contains the singularity z = −1 and therefore
19
dz
19πj
=
2πj
−
=−
3
216
108
C (z + 1) (z − 1)(z − 2)
(iii) The rectangle and vertices ±j, 3 ± j, contains the singularities at z = 1, z = 2
and therefore
C
dz
19πj
=
−
(z + 1)3 (z − 1)(z − 2)
108
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
64(d)
The function
poles at z = ±2.
Residue at z = 2 is
(z2
1
324
Residue at z = −2 is −
241
z−1
has a pole of order 4 at z = −1 and simple
− 4)(z + 1)4
3
4
1 d3 z − 1
3! dz3 z2 − 4
This residue is calculated by using partial fractions
Residue at z = −1 is given by
1
3
z−1
4
4
=
+
z2 − 4
z−2 z+2
whence
1 d3 z − 1
1 d 3 1 1 d3 1
=
+
3! dz3 z2 − 4
24 dz3 z − 2
8 dz3 z + 2
1
3
=−
−
4
4(z − 2)
4(z + 2)4
putting z = −1 gives the residue −
(i) The circle | z |=
1
2
contains none of the singularities hence the integral
(z2
C
(ii) The circle | z +
3
2
61
1
3
− =−
4
4.3
4
81
(z − 1)
dz = 0
− 4)(z + 1)4
|= 2 contains the singularities at z = −1 and z = −2 but
not that at z = 2
Hence
C
3 61
(z − 1)
487πj
=−
dz
=
2πj
− −
2
4
(z − 4)(z + 1)
4 81
162
3
3
(iii) The triangle with vertices − + j, − − j, 3 + j0 contains all the singularities,
2
2
hence
C
z−1
1
3
3
1
dz = 2πj
− −
−
(z2 − 4)(z + 1)4
4.34
4 4.34
4
= −3πj
c Pearson Education Limited 2011
242
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
∞
dx
+x+1
−∞
Since the integrand satisfies the condition for Type 1 infinite real integrals, given
65(a)
x2
in section 4.6.5, we consider
C
z2
dz
where C is a semicircle with radius
+z+1
R and centre the origin in the upper half z -plane
By the residue theorem
C
z2
dz
= 2πj {sum of residues of poles of integrand inside C}
+z+1
z +z+1=0⇒z=
2
−1 ±
√
√
1−4
= − 12 ± j 21 3
2
Only one of these simple poles lies inside C (the one with positive imaginary part)
Residue there = lim (z − z0 )
z→z0
That is, residue = limz→z0
Thus
C
√
1
1
1
=
−
+
j
3
,
z
0
2
2
z2 + z + 1
1
1
using L’Hôpital’s rule (for simplicity) = √
2z + 1
j 3
dz
1
2π
√
√
=
2πj
·
=
z2 + z + 1
j 3
3
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Now,
=
and, as R → ∞,
On
R
C
243
R
+
−R
Γ
→0
Γ
, z = x = real.
−R
Thus, letting R → ∞ we find that
∞
−∞
65(b)
x2
dx
2π
=√
+x+1
3
This integral is done in precisely the same way as that of part (a). This
time, the poles are at ±j but they are both double.
−
(z2
C
dz
1
π
= 2πj × residue at j = 2πj·
=
2
+ 1)
4j
2
Thus
∞
−∞
65(c)
To evaluate
∞
0
(x2
dx
π
=
(x2 + 1)2
2
dx
we use the same semicircular contour,
+ 1)(x2 + 4)2
except that we note
∞
0
dx
1
=
(x2 + 1)(x2 + 4)2
2
∞
−∞
dx
(x2 + 1)(x2 + 4)2
1
has two poles inside C this time, the simple
+ 1)(z2 + 4)2
1
and residue
pole at z = j and the double pole at z = 2j . Residue at z = j is
18j
11
at z = 2j is −
.
288j
Thus
∞
1
dx
11
5π
1
−
=
= 2πj
2
2
2
(x + 1)(x + 4)
2
18j 288j
288
0
plus the fact that
(z2
c Pearson Education Limited 2011
244
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
2π
cos 3θ
dθ
0 5 − 4 cos θ
we follow Example 4.39 and put z = ejθ so that
65(d)
In order to evaluate
cos 3θ =
1
1 3
(z + z−3 ) and cos θ = (z + z−1 )
2
2
With dz = jejθ dθ. Hence we consider
+ z−3 )
dz
jz(5 − 2(z + z−1 ))
1 3
2 (z
C
The function under the integral can be written
1
1 + z6
1 + z6
=
2j 5z4 − 2z5 − 2z3
2j(z − 2)(1 − 2z)z3
The poles inside | z |= 1 are a triple pole at z = 0 and a simple pole at z =
1
.
2
65
1
gives
. Using the Laurent expansion
2
48j
1
21
. The sum is
. Hence
about z = 0 yields the residue −
16j
24j
Using the formula for the residue at z =
0
2π
cos 3θ
1
dθ =
5 − 4 cos θ
2j
C
1 + z6
1
2πj
dz
=
(z − 2)(1 − 2z)z3
24j
π
=
12
65(e)
2π
4dθ
5 + 4 sin θ
0
This follows in the same way as part (d).
Putting z = ejθ yields dz = jejθ dθ, that is, dθ =
1
dz
and sin θ = (z − z−1 )
jz
2j
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus
2π
0
4dθ
=
5 + 4 sin θ
245
4dz
C
jz(5 +
4
2j (z
− z−1 ))
Thus
2π
0
4dθ
=
5 + 4 sin θ
C
4dz
4
8π
= 2πj· =
(2z + j)(z + 2j)
3j
3
(C is the unit circle | z |= 1)
65(f)
∞
−∞
x2 dx
(x2 + 1)(x2 + 2x + 2)
This follows along similar lines to parts (a) and (b). Consider the semicircular
contour and centre the origin radius R on the upper half plane and labelled C
C
z2 dz
= 2πj{sum of residues inside C}
(z2 + 1)(z2 + 2z + 2)
Double pole is at z = j, simple pole is at z = −1 + j
1
3
(−4 + j3) ; residue at −1 + j is (3 − j4)
Residue at z = j is
20
5
7
7
π
Sum = − j giving the integral as
20
10
65(g)
2π
0
dθ
3 − 2 cos θ + sin θ
Once again let z = ejθ and consider the integral around the unit circle
2π
0
dθ
=
3 − 2 cos θ + sin θ
2 1
C z (2
dz
− j) + 3jz − j −
1
2
j
5
and −
. Only the first is inside | z |= 1.
1 − 2j
1 − 2j
1
1
and so the value of the integral is 2πj = π.
The residue is
2j
2j
The poles are at −
c Pearson Education Limited 2011
246
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
65(h)
∞
dx
x4 + 1
0
We have a choice here, let us choose a quarter circle contour as shown below.
1+j
Only the root of z4 + 1 = 0 in the positive quadrant, that is z = √ , needs to
2
be taken into account.
1
−1 − j
Residue at this point is
| 1+j
= √
2
√
4z z= 2
4 2
Hence
dz
π
jπ
(−1 − j)
√
= √ − √
= 2πj
4
4 2
2 2 2 2
C 1+z
0 R
dz
π
π
=
+
+
= √ −j √
4
2 2
2 2
C z +1
Γ
jR
0
on the imaginary axis, z = jy , and on the real axis z = x . Therefore
0
R
dz
jdy
dx
dz
=
+
+
4
4
4
4
C z +1
Γ z +1
R y +1
0 x +1
where we have used (jy)4 = y4 . Letting R → ∞, the last integral → 0. Thus
∞
0
jdy
dx
π
π
√
√
+
=
−
j
4
x4 + 1
2 2
2 2
∞ y +1
0
Equating real (or indeed imaginary) parts gives
∞
dx
π
= √
4
x +1
2 2
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
247
65(i)
∞
(x2
−∞
dx
+ 4x + 5)2
The semicircular contour is used and the poles of
(z2
1
are at
+ 4z + 5)2
z = −2 + j, −2 − j both double. Only the residue at z = −2 + j , which is
1
,
4j
needs to be taken into account.
Hence
∞
−∞
(x2
dx
π
1
= 2πj =
2
+ 4x + 5)
4j
2
65(j)
2π
cos θdθ
3 + 2 cos θ
0
z2 + 1
Again we use the unit circle on which z = e . The integrand is
2j(z3 + 3z2 + z)
√
√
3
3
1
1
with simple poles at z = 0, − +
5, − −
5. Only the first two are inside
2
2
2
2
3
3
C and residues are 1 and − √ . Hence the integral has the value π 1 − √ .
5
5
jθ
Exercises 4.8.3
1
1
x − jy
, u + jv =
= 2
z
x + jy
x + y2
x
1
, x2 + y2 = 2ax .
Thus u = 2
if u =
2
x +y
2a
For the two wires shown in Figure 4.41 potentials are centred at V0 or −V0
66
Since w =
and are tangent to the imaginary (y) axis. They are thus circles of the form
x2 + y2 = 2aV0 x . The equipotential curves are shown
c Pearson Education Limited 2011
248
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
67(a)
j+1
=j
1−j
1
49 + j7
(7 + j)(1 + j7)
z=
(24 + j7) ⇒ w =
=7
= j7
25
1 − j7
1 + 49
3
7
+1
3
z = − ⇒ w = 4 3 = 41 = 7
4
1− 4
4
z = −1 ⇒ w − 0, z = j ⇒ w =
giving images as (0,0),(0,1),(0,7),(7,0).
67(b)
If w =
w−1
z+1
then z =
1−z
w+1
(u − 1 + jv)(u + 1 − jv)
u − 1 + jv
=
u + 1 + jv
(u + 1)2 + v2
(u2 − 1) + v2 + j(vu + v − uv + v)
or x + yj =
(u + 1)2 + v2
u2 + v2 − 1 + 2vj
x + yj =
(u + 1)2 + v2
so that x + yj =
Hence y = 0 corresponds to v = 0.
If x2 + y2 = 1 ⇒| z |= 1
w − 1
w−1
=1
Since z =
this means that
w+1
w+1
or (u − 1)2 + v2 = (u + 1)2 + v2 from which u = 0
67(c)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
249
In order to progress, note that the image of the semicircular conductor in the
w plane is the positive quadrant u ≥ 0, v ≥ 0. Instead of temperature T o C we
πT
π
consider
since this function has the value on u = 0 (where T = 100o C ). The
200
2
mapping w = ez (Example 4.14) provides a means of eliminating the singularity at
w = 0. The complex variable z is already defined, therefore write w = eζ (complex
variable ζ). The imaginary part of ζ is identified with the (scaled) temperature
πT
.
200
v
πT
2y
Thus
= tan−1
= tan−1
as required.
2
2
200
u
1−x −y
68(a)
G(x, y) = 2x − 2xy ; thus
∂G
∂H
=
= 2 − 2y
∂y
∂x
∂H
∂G
=−
= 2x
∂x
∂y
Integrating this gives H = x2 − y2 + 2y
and
Hence W = G + jH = 2z − jz2
68(b)
If w = lnz , then z = ew
Given H(z) = 2z + jz2 = 2ew + je2w
equating real parts gives
G(x, y) = 2eu cos v − e2u sin 2v as required
68(c)
If w = f(z) then the real and imaginary parts of f(z) are harmonic
functions. Hence if ζ = g(w) , then the real and imaginary parts of g are harmonic.
So ζ = g(w) = g(f(z)) implies that harmonic functions (the real and imaginary
c Pearson Education Limited 2011
250
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
parts of f(z) ) transform to harmonic functions (the real and imaginary parts of
g(w) ).
69
If w =
z+3
then | w |= k transforms to
z−3
z + 3
=k
z−3
that is, (x + 3)2 + y2 = k2 (x − 3)2 + k2 y2
1 + k2
or x2 + y2 + 6
x + 9 = 0 as required.
1 − k2
If the centre of the circle is to be (−5, 0) , then
1 + k2
= −10 or k = 2
6
1 − k2
We thus (following section 4.8) require the potential V to be a harmonic function
which has a constant value on a circle u2 + v2 = 4. Hence V has the general form
V = A ln(u2 + v2 )
V0
on u2 + v2 = 4, V = V0 hence A =
ln 4
V0
so that V =
ln(u2 + v2 )
ln 4
z + 3 2 (x + 3)2 + y2
2
2
2
=
Now u + v =| w | =
z−3
(x − 3)2 + y2
Thus
V0
{ln[(x + 3)2 + y2 ] − ln[(x − 3)2 + y2 ]}
V=
ln 4
70
This problem follows a similar pattern to Exercise 67.
The points A, x = 1, y = 0; B, x = 0, y = 1; C, x = 0, y = −1 under the
j(1 − z)
transform to
mapping of w =
1+z
70(a)
w = 0 that is (0, 0), w =
and w =
j(1 − j)
= 1 that is (1, 0)
1+j
j(1 + j)
= −1 that is (−1, 0) respectively
1−j
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
70(b)
w=
251
j(1 − z)
1+z
1−x
j(1 − x)
is purely imaginary. Since
If z = x = real (i.e. y = 0) then w =
1+x
1+x
can take all real values, points on y = 0 correspond to points on u = 0, the
imaginary axis.
j−w
j(1 − z)
then z =
1+z
j+w
So that | z |= 1 ⇒| j − w |=| j + w | or v = 0
70(c)
If w =
For the last part we note the following property of the mapping that is, w =
j(1 − z)
(from (a), (b), and (c))
1+z
πT
which is (in the z plane),
100
π on the negative real axis, and 0 on the positive real axis. The mapping of w = eζ
πT
as imaginary part).
( ζ - complex variable which has the values of
100
Thus
v
1 − x2 − y2
πT
= tan−1
= tan−1
100
u
2y
In a way similar to Exercise 67, identify the function
(using w =
71
j(1 − z)
to find u and v). This gives the result.
1+z
This problem is similar to the last part of Exercise 69. The successive
mappings are
z1 =
z + j4
w = ln z
z − j4
c Pearson Education Limited 2011
252
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
In order for the circle centre 5j to be mapped to a circle centred at the origin in
the z plane we require
| z1 |2 =|
z + j4 2
| = k2 for some constant k
z − j4
2
1+k
that is, x2 + y2 + 8 1−k
2 y + 16 = 0 needs a centre at 5j
Therefore, 8(1 + k2 ) = 10(1 − k)2 ) or h =
1
3
In the w plane, | w |=| ln z1 |=| ln(x21 + y21 ) |= 2 ln 3 on the boundary of the circle
on which T = 100.
Thus writing T = A | ln(x21 + y21 ) | gives T = 100 when
x21 + y21 =
Thus
1
9
if
100 = A· 2 ln 3 or A =
50
ln 3
50
ln(x21 + y21 )
ln 3
x2 + (4 + y)2
50
ln
as required
=
ln 3
x2 + (4 − y)2
T=
Note that T = 0 corresponds to x21 + y21 = 1 or y = 0 as is also required
(| z + j4 |2 =| z − j4 |2 is y = 0).
1
was studied in Example 4.13. Writing, as usual,
z
w = u + jv and z = x + jv leads to
72
The mapping w = z +
u=x+
x
y
and
v
=
y
−
x2 + y 2
x2 + y 2
Hence the unit circle x2 + y2 = 1 in the z plane corresponds to v = 0 (the real
axis) in the w plane.
Points ejπ/3 and e2jπ/3 ( P and Q of this problem) correspond to u = 2 cos π3 and
2 cos 2π
3 respectively. The arc PQ thus corresponds to −1 ≤ u ≤ 1 in the w plane.
w+1
takes this portion of the real axis (−1 ≤ u ≤ 1)
The further mapping ζ =
w−1
to −∞ ≤ Re{ζ} ≤ 0 . This negative real axis corresponds to T = 100. Hence in
a similar fashion to Exercise 70, we identify the variable
c Pearson Education Limited 2011
πT
100
(which = π on this
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
253
line) with the imaginary part of ln ζ which is argζ. We cannot use tan−1 here
because the argument of the logarithm function is quadratic and so
πT
w+1
z + 1/z + 1
= argζ = arg
= arg
100
w−1
z + 1/z − 1
100
arg(z2 + z + 1) − arg(z2 − z + 1)
that is, T =
π
as required.
Review Exercises 4.9
1(a)
z = 1 + j, w = (1 + j)z + j = (1 + j)2 + j = 1 + 2j − 1 + j = 3j
1(b)
z = 1 − j2, w = j3z + j + 1 = j(1 − j2)3 + j + 1 = 3j + 6 + j + 1 = 4j + 7
1(c)
z = 1, w = 12 (1 − j)z + 12 (1 + j) = 1
1(d)
z = j2, w = 12 (1 − j)z + 12 (1 + j) = (1 − j)j + 12 (1 + j) = 32 (1 + j)
2(a)
y = 2x
(b)
x+y=1
For the mapping w = (1 + j)z + j
u = x − y, v = x + y + 1
so y = 2x ⇒ v + 3u = 1
and x + y = 1 ⇒ v = 2
For the mapping w = j3z + j + 1
u = −3y + 1, v = 3x + 1
so y = 2x ⇒ u + 2v = 3
and x + y = 1 ⇒ v − u = 3
For the mapping w = 12 (1 − j)z + 12 (1 + j)
u=
1
1
(x + y + 1), v = (x − y + 1)
2
2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
254
so y = 2x ⇒ 3v − u = 1
and x + y = 1 ⇒ u = 1
3
w = αz + β
when z = 2 − j, w = 1, and when z = 0, w = 3 + j
3(a)
Solving the simultaneous equations gives α = − 15 (3 + j4), β = 3 + j .
3(b)
Since
1
w = − (3 + 4j)z + 3 + j
5
1
z = (3 + j − w)(3 − j4)
5
so x = 13 − 3u − 4v
and Re{z} ≤ 0 corresponds to 3u + 4v ≥ 13
3(c)
| z |=
1
| 3 + j − w |5 ≤ 1
5
⇒| w − 3 − j |≤
3(d)
Fixed point is given by
z = αz + β or z =
4
1
5
w=
j
z
⇒x=
v
u2 +v 2 ,
4(a)
x=y+1⇒
4(b)
y = 3x ⇒ u = 3v
y=
v
u2 +v 2
=
1
β
= (7 − j)
1−α
4
u
u2 +v 2
u
u2 +v 2
+ 1 or u2 + v2 + u − v = 0
Line joining A(1 + j) to B(z + j3) or (1, 1) to (2, 3) is y = 2x − 1 which
u
2v
transforms to 2
=
− 1 or u2 + v2 + u − 2v = 0
u + v2
u2 + v2
4(c)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
4(d)
255
y = 4 ⇒ 4(u2 + v2 ) = u
The following Argand diagram shows all these curves.
5
w=
w+1
z+1
⇒z=
z−1
w−1
from which
u2 + v2 − 1
−2v
x=
, y=
2
2
(u − 1) + v
(u − 1)2 + v2
lines x = k and y = l map to circles
u2 + v2 −
k+1
2k
2v
u+
= 0 and u2 + v2 − 2u +
+1=0
k−1
k−1
l
Fixed points are z =
z+1
z−1
in z2 − 2z − 1 = 0, that is, z = 1 +
fixed points.
c Pearson Education Limited 2011
√
√
2, 1 − 2 are the
256
6
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
w=
1 − z2
z
Fixed points occur at z =
1 − z2
2
or
z 2 = 1 − z2
z2 = 1/2
√
Hence z = ± 2/2
z∗
1
−z=
− z ( z∗ = complex conjugate).
z
| z |2
1
x
−y
Whence u = 2
−
x
,
v
=
−
−
y
and
v
=
−y
+
1
x + y2
x2 + y 2
r2
2
2
r u
−r v
so 2
= x, 2
=y
r −1
r +1
Squaring and adding gives
Writing w =
ur2
r2 − 1
2
+
vr2
r2 + 1
2
= x2 + y 2 = 1
the required ellipses.
Since u =
7
x
x2 +y 2
− x , if x2 + y2 = 1 then u = 0 (imaginary axis in the w plane).
w = z3
= (x + jy)3 = x3 + 3jx2 y + 3j2 xy + j3 y3
so u = x3 − 3xy2
v = 3x2 y − y3 are the real and imaginary parts.
∂u
= 3x2 − 3y2 ,
∂x
∂v
= 3x2 − 3y2 ,
∂y
∂u
= −6xy
∂y
∂v
= 6xy
∂x
hence verifying the Cauchy–Riemann equations:
∂u
∂v ∂u
∂v
=
,
=−
∂x
∂y ∂y
∂x
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
8
257
u(x, y) = x sin x cosh y − y cos x sinh y
hence
and
∂u
= sin x cosh y + x cos x cosh y + sin x sinh y
∂x
∂u
= x sin x sinh y − y cos x cosh y − cos x sinh y
∂y
By the Cauchy–Riemann equations,
∂u
∂v
∂u
∂v
=
and
=−
∂x
∂y
∂y
∂x
hence, integrating
∂u
with respect to y gives:
∂x
v = sin x sinh y + x cos x sinh y + y sin x cosh y − sin x sinh y + f1 (y)
Integrating −
∂u
with respect to x gives:
∂y
v = (x cos x − sin x) sinh y + y sin x cosh y + sin x sinh y + f2 (x)
where f1 and f2 are arbitrary functions.
Comparing these gives v = y sin x cosh y + x cos x sinh y (ignoring the additive
constant).
Thus
w = u + jv =x sin x cosh y − y cos x sinh y
+ j(y sin x cosh y + x cos x sinh y)
Since this is f(z) , we put y = 0 to find f(x) which will give the functional form
of f, namely
f(x) = x sin x. Thus f(z) = z sin z.
az + b
(the general bilinear mapping) since z = 0 ⇒ w = ∞ we
cz + d
αz + β
must have d = 0, hence (relabelling the constants) w =
.
z
Writing this as wz = αz + β and inserting the pairs of values z = j, w = −j and
9
Writing w =
z = 12 (1 + j), w = 1 − j gives
1 = αj + β, 1 = 12 (1 + j)α + β
from which α = 0, β = 1. Hence w =
1
.
z
c Pearson Education Limited 2011
258
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
x
y
and v = − 2
2
+y
x + y2
the real axis (in the z plane) maps to the real axis (in the w plane).
9(a)
Since u =
x2
y
If y > 0 then v = − x2 +y
2 < 0 and vice versa.
Thus the lower half of the z plane maps to the upper half of the w plane.
9(b) The circle | z − 12 j |=
y
or v = −1 v = − x2 +y
2
If | z − 12 j |<
1
2
is x2 + y2 − y = 0
1
2
then x2 + y2 − y < 0 or v < −1, that is, the interior of | z − 12 j |=
1
2
maps to Im(w) < −1 as required.
a2
10 The mapping z = ζ +
4ζ
maps R = constant (where ζ = Rejθ ) to curves
z = Rejθ +
a2 −jθ
e
4R
which describe ellipses in the z plane as can be seen by writing
a2
a2
cos θ, y = R −
sin θ
x= R+
4R
4R
whence
x2
(R +
when R =
1
2 a,
2
a 2
4R )
+
y2
(R −
a2 2
4R )
=1
y = 0 (real axis). This mapping is used together with bilinear
mappings to map an aerofoil shape onto the unit circle.
aeronautical engineering.
1
= (1 + z3 )−1
1 + z3
using the binomial expansion gives
11
1
= 1 − z3 + z 6 − z 9 + · · ·
3
1+z
Similarly
1
= 1 − 2z3 + 3z6 − 4z9 + · · ·
(1 + z3 )2
and both are valid in the disc | z |< 1.
c Pearson Education Limited 2011
This is useful in
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
259
12(a)
2−1−z
2
1−z
=
=
−1
1+z
1+z
1+z
Using the binomial series again gives
1−z
= 1 − 2z + 2z2 − 2z3 + . . .
1+z
since the nearest singularity of the function to z = 0 is at z = −1, the radius of
convergence is 1.
12(b)
f(z) =
z2
use Taylor’s series
1
. This time we need to expand about the point z = 1. We
+1
2z
2
8z2
1
,
f
(z)
=
−
,
f
(z)
=
−
+
z2 + 1
(z2 + 1)2
(z2 + 1)2
(z2 + 1)3
24z
48z3
24
288z2
384z4
iv
f (z) = 2
−
,
f
(z)
=
−
+
(z + 1)3
(z2 + 1)4
(z2 + 1)3
(z2 + 1)4
(z2 + 1)5
f(z) =
At z = 1 these have values
1
1 1
2 , − 2 , 2 , 0, −3
giving the expansion
z2
1 1
1
1
1
= − (z − 1) + (z − 1)2 − (z − 1)4 + . . .
+1
2 2
4
6
√
1
are at z = ±j which are a distance 2 from z = 1,
2
1+z
√
hence the radius of convergence is 2 .
The singularities of
12(c)
1
z
=1−
= f(z)
1+z
1+z
1
2
6
f (z) =
, f = −
, f (z) = +
2
3
(1 + z)
(1 + z)
(1 + z)4
1
1
1
z
1
= (1 + j) + j(z − j) − (1 + j)(z − j)2 − (z − j)3 + . . .
1+z
2
2
8
√4
The radius of convergence is again 2.
Thus
1
has singularities at z = 0, j, −j . The radius of
+ 1)
convergence is the distance of the centre of the point of expansion from the nearest
13
The function
z(z2
c Pearson Education Limited 2011
260
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
of these singularities. These are found straightforwardly either by inspection or by
using Pythagoras’ theorem
z = 1; 1; z = −1; 1, z = 1 + j; 1, z = 1 + j 12 ;
14
f(z) =
14(a)
(z2
1
2
√
√
5, z = 2 + j3; 2 2 .
1
+ 1)z
The Laurent expansion is
1
(1 + z2 )−1
z
1
= (1 − z2 + z4 − z6 + . . .)
z
1
= − z + z3 − z 5 + . . .
z
valid for 0 <| z |< 1
14(b)
Since f(z) is regular at z = 1, f(z) has a Taylor expansion :
1
(z − 1)2
=
f(1)
+
(z
−
1)f
f (1) + . . .
(1)
+
z(1 + z2 )
2
1 + 3z2
1
, f (1) = −1
f(1) = , f (z) = −
2
(z + z3 )2
6z
2(1 + 3z2 )2
+
f (z) = −
(z + z3 )2
(z + z3 )3
5
4 2 × 16
=
so f (1) = − +
4
8
2
so that
1
1
5
=
−(z
−
1)
+
(z − 1)2 + . . .
2
z(1 + z )
2
4
valid for | z − 1 |< 1
15
f(z) = ez sin
15(a)
1
1−z
At z = 0, f(z) is regular. Thus the principal part is zero and f(0) =
sin 1, f(z) = sin 1 + q1 z + q2 z2 + . . . | z |< 1, Taylor series.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
261
15(b)
At z = 1 , f(z) has an essential singularity.
15(c)
At z = ∞, ez has an essential singularity. Hence for parts (b) and (c) the
principal part has infinitely many terms.
16(a)
ez sinh z =
=
=
Real part =
Imaginary part =
1 z z
1
e (e + e−z ) = (e2z + 1)
2
2
1
(1 + e2x (e2jy ))
2
1
(1 + e2x cos 2y + je2x sin 2y)
2
1
(1 + e2x cos 2y)
2
1 2x
e sin 2y
2
16(b)
cos 2z = cos(2x + j2y)
= cos 2x cosh 2y − j sin 2x sinh 2y
16(c)
sin z
x − jy
= 2
sin(x + jy)
z
x + y2
(x − jy)(sin x cosh y + j cos x sinh y)
=
x2 + y 2
x sin x cosh y + y cos x sinh y + j(x cos x sinh y − y sin x cosh y)
=
x2 + y 2
16(d)
tan x + tan jy
1 − tan x tan jy
tan x + j tanh y 1 + j tan x tanh y
·
=
1 − j tan x tanh y 1 + j tan x tanh y
tan z = tan(x + jy) =
from which tan z =
tan x(1 − tanh2 y) + j tanh y(1 + tan2 x)
1 + tan2 x tanh2 y
c Pearson Education Limited 2011
262
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
17(a)
Since
2
dw
= − 3 = 0, this mapping is conformal.
dz
z
17(b)
dw
= 6z2 + 6z + 6(1 − j)
dz
= 0 when z2 + z + 1 − j = 0
or (z − j)(z + j + 1) = 0
so z = j, −1 − j
are the points where the mapping fails to be conformal.
17(c)
w = 64z +
1
z3
dw
3
= 64 − 4 = 0
dz
z
3
where z4 =
64
so z = 0.465, −0.465, j0.465, −j0.465
are the points where the mapping fails to be conformal.
18
w = cos z ,
dw
= sin z = 0 when z = nπ, n = an integer.
dz
u + jv = cos(x + jy) = cos x cosh y − j sin x sinh y
u = cos x cosh y
v = − sin x sinh y
Lines x = k will thus transform to
u2
v2
−
= 1 − hyperpolae
cos2 k sin2 k
Lines y = l will thus transform to
u2
v2
+
= 1 − ellipses
cosh2 l sinh2 l
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
19(a)
at z = 0.
19(b)
263
sin z
sin z
sin z 1
· . Since
→ 1 is z → 0 this function has a simple role
=
2
z
z z
z
(z3
1
has double poles when z3 = 8, that is, at 2, 2e2πj/3 , 2e4πj/3 .
− 8)2
19(c)
z+1
1
z+1
= 2
=
4
2
z −1
(z − 1)(z + 1)
(z − 1)(z2 + 1)
The singularity at z = −1 was removable, those at z = 1, ±j are simple poles.
1
which has simple poles wherever cosh z = 0 that is,
cosh z
z = 12 jπ(2n + 1), n = 0, ±1, ±2, . . .
19(d)
sech z =
19(e)
sinh z is entire (no singularities in the finite plane).
19(f)
Essential singularity at z = 0.
19(g)
zz = ez ln z
which has an essential singularity at z = 0.
20(a)
e2z
double pole at z = −1 residue given by
(1 + z)2
d 2z
(e ) z=−1
dz
= 2e−2
c Pearson Education Limited 2011
264
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
20(b)
cos z
simple pole at z = π/2 residue is
2z − π
lim (z − π/2)
z→π/2
cos z
1
π
= cos = 0
(2z − π)
2
2
sin z
1
tan z
=
which has a double pole at z = π
2z − π
cos z(2z − π)
2
Writing w = z − π/2
20(c)
1 − 12 w2 + . . .
tan z
cos w
cos w
=−
=−
=−
2z − π
2w
2w sin w
2w(w − 16 w3 + . . .)
1
1
= − 2 + + ...
2w
4
Hence residue is 0.
z
has a triple pole at z = −8
(z + 8)3
z
1
1
8
Writing w = z + 8 and z = w − 8 gives
= 3 (w − 8) = 2 − 3
3
(z + 8)
w
w
w
Hence the residue is 0.
20(d)
21
f(z) =
(z2 − 1)(z2 + 3z + 5)
z(z4 + 1)
Zeros are at z = ±1 and at the roots of
z2 + 3z + 5 = 0
z=−
3±
√
9 − 20
3
1√
=− ±j
11
2
2
2
Simple poles are at z = 0 and where z4 = −1
√
that is, z = (±1 ± j)/ 2
The residue at z = 0 is given by
(z2 − 1)(z2 + 3z + 5)
= −5
z→0
z4 + 1
lim
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
265
The residue at z = z0 (where z40 + 1 = 0) is
(z − z0 )(z2 − 1)(z2 + 3z + 5)
lim
z→z0
z(z4 + 1)
z − z0
(z20 − 1)(z20 + 3z0 + 5)
=
lim
z→z0
z0
z4 + 1
(z2 − 1)(z20 + 3z0 + 5) 1
1
= 0
= − (z20 − 1)(z20 + 3z0 + 5)
3
z0
4z0
4
(using z40 = −1)
√ , − 1+j
√ , − 1−j
√ , 1−j
√
in turn gives the residues
Putting z0 = 1+j
2
2
2
2
√
√
√
3
3
3
3
3
3
2 − 4 2 + j, 2 − 4 2 − j and 2 + 4 2 + j respectively.
22
f(z) =
3
2
+
3
4
√
2 − j,
z7 + 6z5 − 30z4
has a triple pole at z = 1 + j
(z − 1 − j)3
1 d2 7
(z + 6z5 − 30z4 ) is evaluated at z = 1 + j
2! dz2
= 21z5 + 60z3 − 180z2 z=1+j
Residue =
= 21(−4 − 4j) + 60(−2 + 2j) − 360j
= −204 − j324
z
has poles at z = −6 and z = −1. Only the
z2 + 7z + 6
z
1
second is inside C. Residue =
|z=−1 = −
z+6
5
2πj
Integral = −
5
23(a)
The integrand
23(b)
The integrand
(z2 + 1)(z2 + 3)
has four simple poles ±2j, ±3j all
(z2 + 9)(z2 + 4)
inside C.
3
3 8
8
j, j, j and − j the sum of which is 0.
20 20 5
5
The integral is thus 0.
Residues are −
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
266
23(c) The integrand
1
has double poles at z = 0, 1 and −1.
z2 (1 − z2 )2
Residues are, at z = 0,
1
d
2z
=+
=0
2
2
2
3
z=0
dz (1 − z )
(1 − z ) z=0
1
2(2z + 1)
3
d
− 2
=
=
At z = 1,
2
2
3
dz
z (1 + z)
(z + z ) z=1
4
1
−2(1 − 2z)
d
3
=
At z = −1,
=
2
2
2
3
dz z (1 − z)
(z − z ) z=−1
4
(i) If C is | z |=
1
2
then only z = 0 is inside C, so integral = 0
(ii) If C is | z |= 2 then all poles are inside C, so integral = 3πj.
1
3j
has simple poles at
and −j .
(2z − 3j)(z + j)
2
1
1
3j
and −j are, respectively, − j and j.
Residues at
2
5
5
(i) Both poles are inside | z |= 2, but since their sum is zero so is the integral
1
(ii) Inside | z − 1 |= 1 the function
is regular. By Cauchy’s
(2z − 3j)(z + j)
theorem the integral = 0.
23(d)
The integrand
z3
has simple poles at
23(e) The integrand 2
(z + 1)(z2 + z + 1)
√
z = ±j, − 12 ± j 12 3 .
C is the circle | z − j |= 12 which contains the pole at z = j but not the other
three poles.
Residue at z = j is
1
2j
hence the integral = −π.
z−1
has simple poles at z = 0, 3 and a double pole at z = 2.
z(z − 2)2 (z − 3)
1
and − 34 .
Residues at z = 0, 2 are respectively 12
23(f)
(i) If C is | z |= 1 only the residue at z = 0 is considered : integral = πj
1 63
(ii) If C is | z |= 52 , residues at 0 at 2 are summed; integral = πj 12
−4 =
−4πj/3.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
267
∞
x2 dx
2
2 2
−∞ (x + 1) (x + 2x + 2)
we use the semicircular contour C on the upper half plane (see Exercise 65). The
24(a)
To evaluate
integral along the curved portion → 0 as the radius of the semicircle → ∞. The
residues in the upper half plane are at the (double) pole at j and the (simple) pole
at −1 + j.
d
z2
z2
at z = j is
which is
Residue of
(z2 + 1)2 (z2 + 2z + 2)
dz (z + j)2 (z2 + 2z + 2)
9j − 12
.
evaluated at z = j. This is (after some algebra)
100
2
z
evaluated at z = −1 + j which is
Residue at z = −1 + j is 2
(z + 1)(z + 1 + j)
3 − 4j
7j
. Sum of residues is −
25
100
∞
x2 dx
z2 dz
−7j
7π
=
= 2πj
=
2
2
2
2
2
2
100
50
C (z + 1) (z + 2z + 2)
−∞ (x + 1) (x + 2x + 2)
∞ x2 dx
one can either use a quarter circle contour (as in
4
0 x + 16
∞ x2 dx
∞ x2 dx
1
=
and use
Exercise 65(h)) or note that, by symmetry,
2
4
4
−∞ x + 16
0 x + 16
the same semicircular contour as above.
24(b)
To evaluate
The disadvantage of doing this is that there are two poles inside the semicircular
contour, but only one in the quarter circle. However, this is compensated by the
easier manipulation of the integral. We shall thus use the semicircle.
The poles inside C, both simple, are at
z=
√
2(−1 + j) and z =
√
2(1 + j)
The way to avoid unnecessary arithmetic/algebra is to determine the residue at
z = z0 where z0 is one of the above poles. This is given by
lim
z→z0
(z − z0 )z2
z4 + 16
Since z = z0 is a root of z4 + 16, we can use L’Hôpital’s rule to obtain
1
3z2 − 2zz0
=
Residue =
3
4z
4z0
z=z0
c Pearson Education Limited 2011
268
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Sum of residues is thus
1
1
j
√
+ √
=− √
4 2(−1 + j) 4 2(1 + j)
4 2
Thus
∞
−∞
x2 dx
=
x4 + 16
0
−j
z2 dz
π
π√
√
√
=
2πj
=
=
2
z4 + 16
4
4 2
2 2
Hence
∞
0
24(c)
x2 dx
π√
=
2
x4 + 16
8
To evaluate
2π
sin2 θdθ
5 + 4 cos θ
0
we follow Exercise 65(d) and put z = ejθ so that
z2 − z4 − 1
(2z3 + 5z2 + 2z)4
Hence
0
2π
sin2 θdθ
1
=
5 + 4 cos θ
4j
C
sin2 θ
dz
= dθ and
=
jz
5 + 4 cos θ
z2 − z4 − 1
dz
z2 (2z + 1)(z + 2)
where C is the unit circle. Residues at z = 0 and z = − 12 (not that at z = −2)
are summed.
5
d z 2 − z4 − 1
[evaluated
at
z
=
0]
is
. Residue at
dz 2z2 + 5z + 2
4
z2 − z4 − 1
1
13
)
z = − is ( 2
=−
2
z (z + 2) z=− 1
6
2
Hence
2π
sin2 θdθ
1
5 13
11π
= 2πj
−
=−
5 + 4 cos θ
4j
4
6
24
0
Residue at z = 0 is
24(d)
The integral
2π
0
cos 2θ
dθ
5 − 4 cos θ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
269
is evaluated similarly
2π
0
cos 2θ
1
dθ =
5 − 4 cos θ
2j
C
z4 + 1
17 5
19π
1
−
=
dz = 2πj
3
4
2
5z − 2z − 2z
2j
6
4
12
(In part (c) the negative sign arises from the choice of direction of the line integral.
sin2 θ
is always positive it can be ignored.)
Since the integrand
5 + 4 cos θ
c Pearson Education Limited 2011
5
Laplace Transforms
Exercises 5.2.6
1(a)
1
s
1
1 1
+
= 2
, Re(s) > 2
L{cosh 2t} = L{ (e2t + e−2t ) =
2
2 s−2 s+2
s −4
1(b)
L{t2 } =
1(c)
L{3 + t} =
1
3
3s + 1
+ 2 =
, Re(s) > 0
s s
s2
1(d)
L{te−t } =
1
, Re(s) > −1
(s + 1)2
2(a)
5
(g) 0
2
, Re(s) > 0
s3
(b) -3
(c) 0
(h) 0
(i) 2
(d) 3
(e) 2
(f) 0
(j) 3
5
5s − 3
3
, Re(s) > 0
− 2 =
s s
s2
3(a)
L{5 − 3t} =
3(b)
L{7t3 − 2 sin 3t} = 7.
3(c)
L{3 − 2t + 4 cos 2t} =
3(d)
L{cosh 3t} =
3(e)
L{sinh 2t} =
s2
s2
42
6
3
6
= 4 − 2
, Re(s) > 0
− 2. 2
4
s
s +9
s
s +9
3
s
4s
2
3s − 2
+ 2
− 2 + 4. 2
=
, Re(s) > 0
2
s s
s +4
s
s +4
s
, Re(s) > 3
−9
2
, Re(s) > 2
−4
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
3(f)
L{5e−2t + 3 − 2 cos 2t} =
3(g)
L{4te−2t } =
3(h)
L{2e−3t sin 2t} =
3(i)
L{t2 e−4t } =
5
3
s
+ − 2. 2
, Re(s) > 0
s+2 s
s +4
4
, Re(s) > −2
(s + 2)2
4
4
=
, Re(s) > −3
(s + 3)2 + 4
s2 + 6s + 13
2
, Re(s) > −4
(s + 4)3
3(j)
36
6
4
2
− 3+ 2−
4
s
s
s
s
36 − 6s + 4s2 − 2s3
=
, Re(s) > 0
s4
L{6t3 − 3t2 + 4t − 2} =
3(k)
3(l)
L{2 cos 3t + 5 sin 3t} = 2.
3
2s + 15
s
+
5.
=
, Re(s) > 0
s2 + 9
s2 + 9
s2 + 9
s
+4
s2 − 4
d s
=
L{t cos 2t} = −
, Re(s) > 0
ds s2 + 4
(s2 + 4)2
L{cos 2t} =
s2
3(m)
6s
d 3
= 2
2
ds s + 9
(s + 9)2
(s2 + 9)2 6 − 6s(s2 + 9)2 4s
6s
d
=
−
L{t2 sin 3t} = −
ds (s2 + 9)2
(s2 + 9)4
18s2 − 54
, Re(s) > 0
= 2
(s + 9)3
L{t sin 3t} = −
3(n)
L{t2 − 3 cos 4t} =
2
3s
, Re(s) > 0
− 2
3
s
s + 16
c Pearson Education Limited 2011
271
272
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
3(o)
3
2
(s + 1)
+
+
3
2
(s + 2)
(s + 1) + 4 s
3
2
s+1
+ , Re(s) > 0
=
+ 2
3
(s + 2)
s + 2s + 5 s
L{t2 e−2t − e−t cos 2t + 3} =
Exercises 5.2.10
4(a)
L−1
4(b)
L−1
4(c)
L−1
4(d)
L−1
1
14
1
1
= L−1
− 4
= [e−3t − e−7t ]
(s + 3)(s + 7)
s+3 s+7
4
−1
s+5
2
= L−1
= −e−t + 2e3t
+
(s + 1)(s − 3)
s+1 s−3
1
4
s−1
4
4 1
4
3
9
−1 9
=
L
−
= − t − e−3t
−
2
2
s (s + 3)
s
s
s+3
9 3
9
2s + 6
s
2
= L−1 2. 2
= 2 cos 2t + 3 sin 2t
+ 3. 2
2
2
s +4
s +2
s + 22
4(e)
−1
L
4(f)
L−1
1
1
1
−1 0
16
16
=
L
+
−
s2 (s2 + 16)
s
s2
s2 + 16
1
1
1
t−
sin 4t =
[4t − sin 4t]
=
16
64
64
s+8
−1 (s + 2) + 6
=
L
= e−2t [cos t + 6 sin t]
s2 + 4s + 5
(s + 2)2 + 1
4(g)
L−1
1
− 18 s + 12
s+1
−1 8
=
L
+
s2 (s2 + 4s + 8)
s
(s + 2)2 + 22
1 1 1 (s + 2) − 3(2)
= L−1 . −
8 s 8 (s + 2)2 + 22
1
= [1 − e−2t cos 2t + 3e−2t sin 2t]
8
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
4(h)
L−1
1
1
2
4s
−1
−
+
=
L
(s − 1)(s + 1)2
s − 1 (s + 1) (s + 1)2
= et − e−t + 2tet
4(i)
L−1
s+7
−1 (s + 1) + 3(2)
=
L
= e−t [cos 2t + 3 sin 2t]
s2 + 2s + 5
(s + 1)2 + 22
4(j)
−1
L
4(k)
L−1
1
12
3
s2 − 7s + 5
−1
=L
−
+ 2
(s − 1)(s − 2)(s − 3)
s−1 s−2 s−3
1
11
= et − 3e2t + e3t
2
2
−2
2s − 1
5s − 7
−1
=
L
+
(s + 3)(s2 + 2)
s + 3 s2 + 2
√
√
1
= −2e−3t + 2 cos 2t − √ sin 2t
2
4(l)
−1
L
4(m)
L−1
15
1
s−2
s
−1
=
L
−
(s − 1)(s2 + 2s + 2)
s − 1 5 s2 + 2s + 2
15
1 (s + 1) − 3
−
= L−1
s − 1 5 (s + 1)2 + 1
1
1
= et − e−t (cos t − 3 sin t)
5
5
(s + 1) − 2
s−1
−1
=
L
= e−t (cos 2t − sin 2t)
s2 + 2s + 5
(s + 1)2 + 22
4(n)
L−1
3
12
2
s−1
= L−1
−
+ 2
(s − 2)(s − 3)(s − 4)
s−2 s−3 s−4
1
3
= e2t − 2e3t + e−4t
2
2
c Pearson Education Limited 2011
273
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
274
4(o)
3s
3s
−1
=
L
(s − 1)(s2 − 4)
(s − 1)(s − 2)(s + 2)
3
1
−1
+ 2 − 2
= L−1
s−1 s−2 s+2
3
1
= −et + e2t − e−2t
2
2
9
1
36
−1 4
2s
2s
=
L
−
+
2
2
2
2
s(s + 1)(s + 9)
s s +1 s +9
1
9
= 4 − cos t + cos 3t
2
2
L−1
4(p)
L−1
4(q)
L−1
9
7s + 9
2s2 + 4s + 9
= L−1
−
3
2
2
(s + 2)(s + 3s + 3)
s + 2 (s + 2 ) + 3/4
√ √
9
7(s + 32 ) − 3. 3/2
−1
√
−
=L
s+2
(s + 32 )2 + ( 3/2)2
√
√
√
3
3
3
t − 3 sin
t
= 9e−2t − e− 2 t 7 cos
2
2
4(r)
L−1
1
1
1
19
1
−1
10
90 s + 9
=
L
−
−
2
2
(s + 1)(s + 2)(s + 2s + 10)
s + 1 s + 2 s + 2s + 10
1
1
1
s + 10
−1
9
10
−
−
=L
s + 1 s + 2 90 (s + 1)2 + 32
1
1
1 (s + 1) + 3(3)
−1
9
10
−
−
=L
s + 1 s + 2 90 (s + 1)2 + 32
1
1
1
= e−t − e−2t − e−t (cos 3t + 3 sin 3t)
9
10
90
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 5.3.5
5(a)
2s + 5
1
=
s+2
s+2
2s + 5
1
1
X(s) =
=
+
(s + 2)(s + 3)
s+2 s+3
(s + 3)X(s) = 2 +
x(t) = L−1 {X(s)} = e−2t + e−3t
5(b)
2
s2 + 6
=
s2 + 4
s2 + 4
35
3
4
s2 + 6
26
26 s + 26
=
−
X(s) =
(3s − 4)(s2 + 4)
3s − 4
s2 + 4
35 4 t
3
2
x(t) = L−1 {X(s)} =
e 3 − (cos 2t + sin 2t)
78
26
3
(3s − 4)X(s) = 1 +
5(c)
(s2 + 2s + 5)X(s) =
1
s
1
1
s+2
1
5
=
−
·
X(s) =
s(s2 + 2s + 5)
s
5 s2 + 2s + 5
1
1 (s + 1) + 12 (2)
= 5 −
s
5 (s + 1)2 + 22
1
1
x(t) = L−1 {X(s)} = (1 − e−t cos 2t − e−t sin 2t)
5
2
5(d)
4s
2s2 + 4s + 8
=
s2 + 4
s2 + 4
2s2 + 4s + 8
X(s) =
(s + 1)2 (s2 + 4)
12
6
1 12s − 32
25
5
=
−
+
(s + 1) (s + 1)2
25 s2 + 4
16
12 −t 6 −t 12
e + te −
cos 2t +
sin 2t
x(t) = L−1 {X(s)} =
25
5
25
25
(s2 + 2s + 1)X(s) = 2 +
c Pearson Education Limited 2011
275
276
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
5(e)
(s2 − 3s + 2)X(s) = 1 +
X(s) =
x(t) = L−1 {X(s)} =
2
s+6
=
s+4
s+4
1
7
4
s+6
= 15 − 5 + 3
(s + 4)(s − 1)(s − 2)
s+4 s−1 s−2
1 −4t 7 t 4 2t
e
− e + e
15
5
3
5(f)
(s2 + 4s + 5)X(s) = (4s − 7) + 16 +
X(s) =
3
s+2
3
(s + 2) + 1
4s2 + 17s + 21
=
+
2
(s + 2)(s + 4s + 5)
s + 2 (s + 2)2 + 1
x(t) = L−1 {X(s)} = 3e−2t + e−2t cos t + e−2t sin t
5(g)
(s2 + s − 2)X(s) = s + 1 +
5(2)
(s + 1)2 + 4
s3 + 3s2 + 7s + 15
X(s) =
(s + 2)(s − 1)(s2 + 2s + 5)
13
1
5
− 31
12
4s − 4
=
+
+
s + 2 s − 1 s2 + 2s + 5
13
− 31
1 (s + 1) − 3(2)
=
+ 12 +
s + 2 s − 1 4 (s + 1)2 + 22
1
13
1
3
x(t) = L−1 {X(s)} = − e−2t + et + e−t cos 2t − e−t sin 2t
3
12
4
4
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
5(h)
(s2 + 2s + 3)Y(s) = 1 +
3
s2
2
4
− 32
1
s2 + 3
3s + 3
=
+ 2+ 2
Y(s) = 2 2
s (s + 2s + 3)
s
s
s + 2s + 3
√
1
− 23
2 (s + 1) − √2 ( 2)
1
√
=
+ 2+
s
s
3 (s + 1)2 + ( 2)2
√
√
2
1
2
y(t) = L−1 {Y(s)} = − + t + e−t (cos 2t + √ sin 2t)
3
3
2
5(i)
(s2 + 4s + 4)X(s) =
X(s) =
=
x(t) = L−1 {X(s)} =
1
1
2
s+2+ 3 +
2
s
s+2
s5 + 6s4 + 10s3 + 4s + 8
2s3 (s + 2)3
3
8
s
−
1
2
s2
+
1
2
s3
+
1
8
s+2
+
3
4
(s +
2)2
+
1
(s + 2)3
3 1
1
3
1
1
− t + t2 + e−2t + te−2t + t2 e−2t
8 2
4
8
4
2
5(j)
(9s2 + 12s + 5)X(s) =
X(s) =
=
x(t) = L−1 {X(s)} =
1
s
1
1
4
1
5
5 s + 15
=
−
s
9s(s2 + 43 s + 59 )
(s + 23 )2 +
1
5
s
−
1 [(s + 23 ) + 23 ]
5 (s + 23 )2 + ( 31 )2
1
1 1 −2t
1
− e 3 (cos t + 2 sin t)
5 5
3
3
c Pearson Education Limited 2011
1
9
277
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
278
5(k)
1
4
−s3 − 6s2 − 16s + 32
(s2 + 8s + 16)X(s) = − s + 1 − 4 + 16· 2
=
2
s + 16
2(s2 + 16)
−s3 − 6s2 − 16s + 32
X(s) =
2(s + 4)2 (s2 + 16)
1
1
0
2s
+
=
−
2
2
s + 4 (s + 4)
s + 16
1
x(t) = L−1 {X(s)} = te−4t − cos 4t
2
5(l)
(9s2 + 12s + 4)Y(s) = 9(s + 1) + 12 +
1
s+1
9s2 + 30s + 22
Y(s) =
(3s + 2)2 (s + 1)
0
18
1
+
+
=
s + 1 3s + 2 (3s + 2)2
2
y(t) = L−1 {Y(s)} = e−t + 2te− 3 t
5(m)
1
2
+ 2
s s
3
s − 2s2 + 2s + 1
X(s) = 2
s (s − 1)(s − 2)(s + 1)
(s3 − 2s2 − s + 2)X(s) = s − 2 +
5
4
1
2
s2
5
2
1
12
+
− 3
= +
−
s
s−1 s−2 s+1
5 1
5
2
x(t) = L−1 {X(s)} = + t − et + e2t − e−t
4 2
12
3
5(n)
s
+9
9
3
2
s + 2s + 10s + 18
1 7s − 25
1 s+9
20
X(s) = 2
=
−
−
(s + 9)(s + 1)(s2 + 1)
s + 1 16 s2 + 1
80 s2 + 9
25
1
3
9 −t
7
e −
cos t +
sin t −
cos 3t −
sin 3t
x(t) = L−1 {X(s)} =
20
16
16
80
80
(s3 + s2 + s + 1) = (s + 1) + 1 +
s2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
279
6(a)
1
1
2sX(s) − (2s + 9)Y(s) = − +
2 s+2
(2s + 4)X(s) + (4s − 37)Y(s) = 1
Eliminating X(s)
1
1
)(2s + 4) − 2s = −3s
[−(2s + 9)(2s + 4) − 2s(4s − 37)]Y(s) = (− +
2 s+2
3s
1
s
Y(s) =
= ·
2
12s − 48s + 36
4 (s − 3)(s − 1)
3
1
1 2
− 2
=
4 s−3 s−1
1 3 3t 1 t 3 3t 1 t
y(t) = L−1 {Y(s)} =
e − e = e − e
4 2
2
8
8
Eliminating
dx
from the two equations
dt
dy
+ 4x − 28y = −e−2t
dt
1
dy 1 −2t 21 3t 7 t 27 3t 3 t
= −e
x(t) = −e−2t + 28y − 6
+ e − e − e + e
4
dt
4
4
2
3
4
1 15 3t 11 t
1
e − e − e−2t , y(t) = (3e3t − et )
i.e. x(t) =
4 4
4
8
6
6(b)
5
s2 + 1
1
(2s + 1)X(s) + (3s − 1)Y(s) =
s−1
(s + 1)X(s) + (2s − 1)Y(s) =
Eliminating X(s)
s+1
5
(2s + 1) −
+1
s−1
s+1
10s + 5
−
Y(s) =
2
s(s + 1)(s − 2) s(s − 1)(s − 2)
5
3
− 25
5 12
2
+ 2 − 2
−
−
+ 2
=
s
s−2 s +1
s
s−1 s−2
5 5
1
3
y(t) = L−1 {Y(s)} = − + e2t − 5 sin t − + 2et − e2t
2 2
2
2
2t
t
= −3 + e + 2e − 5 sin t
[(2s − 1)(2s + 1) − (3s − 1)(s + 1)]Y(s) =
s2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
280
Eliminating
dx
dt
from the original equations
dy
+ x − y = 10 sin t − et
dt
x(t) = 10 sin t − et − 3 + e2t + 2et − 5 sin t − 2e2t − 2et + 5 cos t
= 5 sin t + 5 cos t − 3 − et − e2t
6(c)
(s + 2)X(s) + (s + 1)Y(s) = 3 +
1
3s + 10
=
s+3
s+3
5X(s) + (s + 3)Y(s) = 4 +
4s + 13
5
=
s+2
s+2
Eliminating X(s)
[5(s + 1) − (s + 2)(s + 3)]Y(s) =
15s + 50
−4s2 − 10s + 11
− (4s + 13) =
s+3
s+3
9
7
− 12
4s2 + 10s − 11
2s − 2
=
+ 2
Y(s) =
(s + 3)(s2 + 1)
s+3
s +1
7
1
9
y(t) = L−1 {Y(s)} = − e−3t + cos t − sin t
2
2
2
From the second differential equation
21
3
3
27
cos t +
sin t − e−3t
5x = 5e−2t + e−3t −
2
2
2
2
+
7
9
sin t + cos t
2
2
x(t) = 3 sin t − 2 cos t + e−2t
6(d)
(3s − 2)X(s) + 3sY(s) = 6 +
sX(s) + (2s − 1)Y(s) = 3 +
1
6s − 5
=
s−1
s−1
1
3s + 1
=
s
s
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Eliminating X(s)
[3s2 − (3s − 2)(2s − 1)]Y(s) =
Y(s) =
s(6s − 5) (3s − 2)(3s + 1)
−
s−1
s
6s2 − 5s
9s2 − 3s − 2
−
s(3s − 1)(s − 2) (s − 1)(3s − 1)(s − 2)
18
14
1
5
= − +
+ 5
s 3s − 1 s − 2
9
14
− 12
10
−
−
+ 5
s − 1 3s − 1 s − 2
1
1
=− + 2
s s−1
1
y(t) = L−1 {Y(s)} = −1 + et +
2
Eliminating
+
9
2
3s − 1
3 t
e3
2
dx
from the original equations
dt
1
3
9 t
3
3 t
3 − et − 3 + e t + e 3 − e t − e 3
2
2
2
2
2
t
3
1
= e 3 − et
2
2
x(t) =
6(e)
(3s − 2)X(s) + sY(s) = −1 +
3
5s
−s2 + 5s + 2
+
=
s2 + 1 s2 + 1
s2 + 1
2sX(s) + (s + 1)Y(s) = −1 +
s
−s2 + s
1
+
=
s2 + 1 s2 + 1
s2 + 1
Eliminating Y(s)
[(3s − 2)(s + 1) − 2s2 ]X(s) =
X(s) =
=
s2
1
[(−s2 + 5s + 2)(s + 1) − (−s2 + s)s]
+1
3s2 + 7s + 2
3s + 1
=
2
(s + 2)(s − 1)(s + 1)
(s − 1)(s2 + 1)
2
2s − 1
− 2
s−1 s +1
x(t) = L−1 {X(s)} = 2et − 2 cos t + sin t
c Pearson Education Limited 2011
281
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
282
Eliminating
dy
from the original equation
dt
dx
dt
= −2 sin t − 4 cos t − 4et + 4 cos t − 2 sin t + 2et + 2 sin t + cos t
y(t) = −2 sin t − 4 cos t − 2x +
that is, y(t) = −2et − 2 sin t + cos t, x(t) = 2et − 2 cos t + sin t
6(f)
1
s2 + 1
=
s2
s2
s+1
1
(s + 1)X(s) + 4sY(s) = 1 + =
s
s
sX(s) + (s + 1)Y(s) = 1 +
Eliminating Y(s)
s2 + 1
3s2 − 2s + 3
(s + 1)2
=
−
[4s2 − (s + 1)2 ]X(s) = 4s
s2
s
s
2
−3
1
9
3s − 2s + 3
=
−
+
X(s) =
s(s − 1)(3s + 1)
s
s − 1 3s + 1
t
x(t) = L−1 {X(s)} = −3 + et + 3e− 3
Eliminating
dy
from the original equation
dt
dx
1
4t − 1 + x + 3
4
dt
t
t
1
= 4t − 1 − 3 + et + 3e− 3 − 3et + 3e− 3
4
t
1
3 t
that is, y(t) = t − 1 − et + e− 3 , x(t) = −3 + et + 3e− 3
2
2
y=
6(g)
14 + 7s
12 7
+ =
2
s
s
s2
14 14
14 − 14s
(5s + 4)X(s) − (3s − 6)Y(s) = 2 −
=
s
s
s2
(2s + 7)X(s) + 3sY(s) =
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
283
Eliminating Y(s)
1
[(3s − 6)(14 + 7s) + 3s(14 − 14s)]
s2
21
21(s2 + s − 2)X(s) = 21(s + 2)(s − 1)X(s) = 2 (−s2 + 2s − 4)
s
2
−s + 2s − 4
X(s) = 2
s (s + 2)(s − 1)
−1
1
0
2
=
+
+ + 2
s−1 s+2 s s
−1
x(t) = L {X(s)} = −et + e−2t + 2t
[(2s + 7)(3s − 6) + (5s + 4)(3s)]X(s) =
Eliminating
dy
from the original equations
dt
dx
dt
t
= 28t − 7 + 7e + 14e−2t − 14 + 11et − 11e−2t − 22t
1
7
giving y(t) = t − + 3et + e−2t , x(t) = −et + e−2t + 2t.
2
2
6y = 28t − 7 − 11x − 7
6(h)
(s2 + 2)X(s) − Y(s) = 4s
−X(s) + (s2 + 2)Y(s) = 2s
Eliminating Y(s)
[(s2 + 2)2 − 1]X(s) = 4s(s2 + 2) + 2s
(s4 + 4s2 + 3)X(s) = 4s3 + 10s
3s
s
4s3 + 10s
= 2
+ 2
2
2
(s + 1)(s + 3)
s +1 s +3
√
−1
x(t) = L {X(s)} = 3 cos t + cos 3t
X(s) =
From the first of the given equations
√
√
d2 x
= 6 cos t + 2 cos 3t − 3 cos t − 3 cos 3t
2
dt
√
√
that is, y(t) = 3 cos t − cos 3t, x(t) = 3 cos t − cos 3t
y(t) = 2x +
c Pearson Education Limited 2011
284
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
6(i)
35
+ 12 =
(5s2 + 6)X(s) + 12s2 Y = s
4
35
2
2
5s X(s) + (16s + 6)Y(s) = s
+ 16 =
4
83
s
4
99
s
4
Eliminating X(s)
s
[83(5s2 ) − 99(5s2 + 6)]
4
s
4
2
[−20s − 126s − 36]Y(s) = [−80s2 − 594]
4
[60s4 − (5s2 + 6)(16s2 + 6)]Y(s) =
s(40s2 + 297)
4(s2 + 6)(10s2 + 3)
25
− 14 s
2 s
+
= 2
s + 6 10s2 + 3
√
1
5
y(t) = L−1 {Y(s)} = − cos 6t + cos
4
4
Y(s) =
Eliminating
3
t
10
d2 x
from the original equations
dt2
√
3
3
d2 y 15 3
−
cos
t
+
−
+
3
cos
=
6t
dt2
4
4
10
4
√
√
3
3
5
3
1
t + cos 6t, x(t) = cos
t − cos 6t.
10
4
4
10
4
3x = 3y + 3
i.e. x(t) = cos
6(j)
(2s2 − s + 9)X(s) − (s2 + s + 3)Y(s) = 2(s + 1) − 1 = 2s + 1
(2s2 + s + 7)X(s) − (s2 − s + 5)Y(s) = 2(s + 1) + 1 = 2s + 3
Subtract
(−2s + 2)X(s) − (2s − 2)Y(s) = −2 ⇒ X(s) + Y(s) =
1
s−1
⇒ x(t) + y(t) = et
(i)
Add
(4s2 + 16)X(s) − (2s + 8)Y(s) = 4(s + 1)
2(s + 1)
⇒ 2x(t) − y(t)
2X(s) − Y(s) = 2
s +4
= 2 cos 2t + sin 2t
c Pearson Education Limited 2011
(ii)
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
285
Then from (i) and (ii)
x(t) =
1 t 2
1
2
1
2
e + cos 2t + sin 2t, y(t) = et − cos 2t − sin 2t
3
3
3
3
3
3
Exercises 5.4.3
7
1μF = 10−6 F so 50μ = 5.105 F
Applying Kirchhoff’s second law to the left hand loop
di1
di2
= E. sin 100t
i1 dt + 2
−
dt
dt
1
5.105
Taking Laplace transforms
2.104
100
I1 (s) + 2s[I1 (s) − I2 (s)] = E. 2
s
s + 104
50s
(104 + s2 )I1 (s) − s2 I2 (s) = E. 2
s + 104
(i)
Applying Kirchhoff’s law to the right hand loop
di1
di2
−
=0
100i2 (t) − 2
dt
dt
which on taking Laplace transforms gives
sI1 (s) = (50 + s)I2 (s)
Substituting in (i)
50s2
s2 + 104
Es2
2
4
(s + 200s + 10 )I2 (s) = 2
s + 104
s2
I2 (s) = E 2
(s + 104 )(s + 100)2
s(50 + s)
then from (ii) I1 (s) = E 2
(s + 104 )(s + 100)2
(104 + s2 )(50 + s)I2 (s) − s2 I2 (s) = E.
c Pearson Education Limited 2011
(ii)
286
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Expanding in partial functions
1
1
1
− 200
2
200 s
+
I2 (s) = E
+
s + 100 (s + 100)2
s2 + 104
1 −100t 1 −100t
1
e
cos 100t
i2 (t) = L−1 {I2 (s)} = E −
+ te
+
200
2
200
8
Applying Kirchhoff’s second law to the primary and secondary circuits
respectively gives
2i1 +
di2
di1
+1
= 10 sin t
dt
dt
2i2 + 2
di1
di2
+
=0
dt
dt
Taking Laplace transforms
(s + 2)I1 (s) + sI2 (s) =
10
+1
s2
sI1 (s) + 2(s + 1)I2 (s) = 0
Eliminating I1 (s)
[s2 − 2(s + 1)(s + 2)]I2 (s) =
10s
s2 + 1
I2 (s) = −
(s2
10s
10s
=− 2
2
+ 1)(s + 7s + 6)
(s + 1)(s + 6)(s + 1)
12
25
−1
s + 35
+ 37 + 37 2 37
s+1 s+6
s +1
=−
i2 (t) = L−1 {I2 (s)} = e−t −
9
35
12 −6t 25
e
cos t −
sin t
−
37
37
37
Applying Kirchhoff’s law to the left and right hand loops gives
d
(i1 + i2 ) + 1 i1 dt = E0 = 10
dt
di2
− 1 i1 dt = 0
i2 +
dt
(i1 + i2 ) +
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
287
Applying Laplace transforms
1
10
(s + 1)I1 (s) + (s + 1)I2 (s) + I1 (s) =
s
s
1
(s + 1)I2 (s) − I1 (s) = 0 ⇒ I1 (s) = s(s + 1)I2 (s)
s
Substituting back in the first equation
s(s + 1)2 I2 (s) + (s + 1)I2 (s) + (s + 1)I2 (s) =
(s2 + s + 2)I2 (s) =
I2 (s) =
10
s
10
s(s + 1)
10
s(s + 1)(s2 + s + 2)
Then from (i)
10
10
=
1
+s+2
(s + 2 )2 +
√
1
7
20
t
i1 (t) = L−1 {I1 (s)} = √ e− 2 t sin
2
7
I1 (s) =
10
s2
Applying Newton’s law to the motion of each mass
ẍ1 = 3(x2 − x1 ) − x1 = 3x2 − 4x1
ẍ2 = −9x2 − 3(x2 − x1 ) = −12x2 + 3x1
giving
ẍ1 + 4x1 − 3x2 = 0, x1 (0) = −1, x2 (0) = 2
ẍ2 + 12x2 − 3x1 = 0
Taking Laplace transforms
(s2 + 4)X1 (s) − 3X2 (s) = −s
−3X1 (s) + (s2 + 12)X2 (s) = 2s
c Pearson Education Limited 2011
7
4
(i)
288
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Eliminating X2 (s)
[(s2 + 4)(s2 + 12) − 9]X1 (s) = −s(s2 + 12) + 6s
(s2 + 13)(s2 + 3)X1 (s) = −s3 − 6s
3
7
− 10
s
−s3 − 6s
10 s
=
−
(s2 + 13)(s2 + 3)
s2 + 3 s2 + 13
√
√
3
7
cos 13t
x1 (t) = L−1 {X1 (s)} = − cos 3t −
10
10
X1 (s) =
From the first differential equation
3x2 = 4x1 + ẍ1
√
√
√
√
6
14
9
91
= − cos 3t −
cos 13t +
cos 3t +
cos 13t
5
5
10
10
√
√
1
[21 cos 13t − cos 3t]
x2 (t) =
10
√
√
√
√
1
1
(3 cos 3t + 7 cos 13t), x2 (t) =
[21 cos 13t − cos 3t]
10
10
√
√
Natural frequencies are 13 and 3 .
Thus, x1 (t) = −
11
The equation of motion is
Mẍ + bẋ + Kx = Mg ; x(0) = 0 , ẋ(0) =
2gh
The problem is then an investigative one where students are required to investigate
for different h values either analytically or by simulation.
12 By Newton’s second law of motion
M2 ẍ2 = −K2 x2 − B1 (ẋ2 − ẋ1 ) + u2
M1 ẍ1 = B1 (ẋ2 − ẋ1 ) − K1 x1 + u1
Taking Laplace transforms and assuming quiescent initial state
(M2 s2 + B1 s + K2 )X2 (s) − B1 sX1 (s) = U2 (s)
−B1 sX2 (s) + (M1 s2 + B1 s + K1 )X1 (s) = U1 (s)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
289
Eliminating X1 (s)
[(M1 s2 + B1 s + K1 )(M2 s2 + B1 s + K2 ) − B21 s2 ]X2 (s)
= (M1 s2 + B1 s + K1 )U2 (s) + B1 sU1 (s)
B1 s
(M1 s2 + B1 s + K1 )
U1 (s) +
U2 (s)
Δ
Δ
(M1 s2 + B1 s + K1 )
−1
−1 B1 s
and x2 (t) = L {X2 (s)} = L
U1 (s) +
U2 (s)
Δ
Δ
i.e. X2 (s) =
Likewise eliminating X2 (t) from the original equation gives
x1 (t) = L−1 {X1 (s)} = L−1
(M1 s + B1 s + K2 )
B1 s
U1 (s) +
U2 (s)
Δ
Δ
Exercises 5.5.7
13
f(t) = tH(t) − tH(t − 1)
= tH(t) − (t − 1)H(t − 1) − 1H(t − 1)
Thus, using theorem 2.4
L{f(t)} =
1
1
1
1
− e−s 2 − e−s = 2 (1 − e−s ) − e−s
2
s
s
s
s
14(a)
f(t) = 3t2 H(t) − (3t2 − 2t + 3)H(t − 4) − (2t − 8)H(t − 6)
= 3t2 H(t) − [3(t − 4)2 + 22(t − 4) + 43]H(t − 4) − [2(t − 6) + 4]H(t − 6)
Thus,
6
− e−4s L[3t2 + 22t + 43] − e−6s L[2t + 4]
s3
6
6
22 43 −4s 2
4
e
= 3− 3+ 2 +
− 2 + e−6s
s
s
s
s
s
s
L{f(t)} =
c Pearson Education Limited 2011
290
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
14(b)
f(t) = tH(t) + (2 − 2t)H(t − 1) − (2 − t)H(t − 2)
= tH(t) − 2(t − 1)H(t − 1) − (t − 2)H(t − 2)
Thus,
1
− 2e−s L{t} + e−2s L{t}
s2
1
= 2 [1 − 2e−s + e−2s ]
s
L{f(t)} =
e−5s
1
−1 −5s
=
L
{e
F(s)}
where
F(s)
=
and by the first
(s − 2)4
(s − 2)4
1
shift theorem f(t) = L−1 {F(s)} = t3 e2t .
6
Thus, by the second shift theorem
15(a)
L−1
L−1
15(b)
L−1
e−5s
= f(t − 5)H(t − 5)
(s − 2)4
1
= (t − 5)3 e2(t−5) H(t − 5)
6
3e−2s
= L−1 {e−2s F(s)} where
(s + 3)(s + 1)
3
− 32
3
=
+ 2
(s + 3)(s + 1)
s+3 s+1
3
3
f(t) = L−1 {F(s)} = e−t − e−3t
2
2
−2s
3e
= f(t − 2)H(t − 2)
L−1
(s + 3)(s + 1)
3
= [e−(t−2) − e−3(t−2) ]H(t − 2)
2
F(s) =
so
15(c)
L−1
s+1
e−s = L−1 {e−s F(s)} where
+ 1)
s2 (s2
F(s) =
1
1
s+1
s+1
= + 2− 2
+ 1)
s s
s +1
s2 (s2
f(t) = L−1 {F(s)} = 1 + t − cos t − sin t
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
291
so
L−1
s+1
e−s = f(t − 1)H(t − 1)
+ 1)
s2 (s2
= [1 + (t − 1) − cos(t − 1) − sin(t − 1)]H(t − 1)
= [t − cos(t − 1) − sin(t − 1)]H(t − 1)
15(d)
L−1
s+1
−πs
= L−1 {e−πs F(s)} where
e
s2 + s + 1
√
√1 ( 3 )
3 2
√
1 2
3 2
)
+
(
2
2 )
(s + 12 ) +
s+1
=
+ s + 1)
(s +
√
√
1
3
3
1
−2t
t + √ sin
t
cos
f(t) = e
2
2
3
F(s) =
(s2
so
L−1
√
√
√
1
3
3
1
s+1
(t−π)
−
−πs
2
√
e
(t
−
π)
+
sin
(t
−
π)
.H(t − π)
e
=
3
cos
s2 + s + 1
2
2
3
15(e)
L−1
s
−4πs/5
= L−1 {e−4πs/5 F(s)} where
e
s2 + 25
F(s) =
so
L−1
15(f)
L−1
s2
s
⇒ f(t) = L−1 {F(s)} = cos 5t
s2 + 25
4π
s
4π
H t−
e−4πs/5 = f t −
+ 25
5
5
4π
= cos(5t − 4π)H t −
5
4π
= cos 5t H t −
5
e−s (1 − e−s )
= L−1 {(e−s − e−2s )F(s)} where
s2 (s2 + 1)
F(s) =
−1
f(t) = L
1
s2 (s2
+ 1)
=
1
1
− 2
2
s
s +1
{F(s)} = t − sin t
c Pearson Education Limited 2011
292
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so
L−1 {(e−s − e−2s )F(s)} =f(t − 1)H(t − 1) − f(t − 2)H(t − 2)
=[(t − 1) − sin(t − 1)]H(t − 1)
− [(t − 2) − sin(t − 2)]H(t − 2)
16
dx
1
+ x = f(t), L{f(t)} = 2 (1 − e−s − se−s )
dt
s
Taking Laplace transforms with x(0) = 0
(1 + s)
1
− e−s
2
s
s2
1
1
X(s) = 2
− e−s 2
s (s + 1)
s
1
1
1
− e−s L{t}
=− + 2 +
s s
s+1
(s + 1)X(s) =
Taking inverse transforms
x(t) = −1 + e−t + t − (t − 1)H(t − 1)
= e−t + (t − 1)[1 − H(t − 1)]
or x(t) = e−t + (t − 1) for t ≤ 1
x(t) = e−t for t ≥ 1
Sketch of response is
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
17
293
d2 x dx
+ x = g(t), x(0) = 1, ẋ(0) = 0
+
dt2
dt
with L{g(t)} =
1
(1 − 2e−s + e−2s )
s2
Taking Laplace transforms
(s2 + s + 1)X(s) = s + 1 +
1
(1 − 2e−s + e−2s )
2
s
1
s+1
+
(1 − 2e−s + e−2s )
(s2 + s + 1) s2 (s2 + s + 1)
1
(s + 1)
s
1
[1 − 2e−s + e−2s ]
= 2
+ − + 2+ 2
(s + s + 1)
s s
s +s+1
√
√
(s + 12 ) + √13 ( 23 )
(s + 12 ) − √13 ( 23 )
1
1
√
√
=
[1 − 2e−s + e−2s ]
+ − + 2+
s s
3 2
3 2
)
)
(s + 12 )2 + (
(s + 12 )2 + (
2
2
√
√
1
3
3
1
−2t
−1
cos
x(t) = L {X(s)} = e
t + √ sin
t
2
2
3
√
√
1
3
3
1
−2t
t − √ sin
t
cos
+t−1+e
2
2
3
√
1
3
− 2 (t−1)
(t − 1)
− 2H(t − 1) t − 2 + e
cos
2
√
3
1
− √ sin
(t − 1)
2
3
√
1
3
− 2 (t−2)
(t − 2)
cos
+ H(t − 2) t − 3 + e
2
√
3
1
− √ sin
(t − 2)
2
3
X(s) =
that is,
1
x(t) = 2e− 2 t cos
√
3
2 t
+t−1
√
3
3
1
(t − 1) − √ sin
(t − 1)
cos
− 2H(t − 1) t − 2 + e
2
2
3
√
√
1
3
3
1
− 2 (t−2)
+ H(t − 2) t − 3 + e
cos
(t − 2) − √ sin
(t − 2)
2
2
3
1
− 2 (t−1)
√
c Pearson Education Limited 2011
294
18
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
π
π
π
= cos t − H t −
f(t) = sin tH t −
2
2
2
π
since cos t −
= sin t.
2
Taking Laplace transforms with x(0) = 1, ẋ(0) = −1
π
π
(s2 + 3s + 2)X(s) = s + 2 + L cos t − H t −
2
2
π
= s + 2 + e− 2 s L{cos t}
π
s
= s + 2 + e− 2 s · 2
s +1
π
s
1
X(s) =
+ e− 2 s
s+1
(s + 1)(s + 2)(s2 + 1)
2
π −1
1
1 s+3
2
=
+ e− 2 s
+ 5 + · 2
s+1
s + 1 s + 2 10 s + 1
π
1
1
2
1
=
+ e− 2 s L − e−t + e−2t + (cos t + 3 sin t)
s+1
2
5
10
π
π
π
1
2
1
so x(t) = L−1 {X(s)} = e−t + − e−(t− 2 ) + e−2(t− 2 ) + (cos t −
2
5
10
2
π
π
+ 3 sin t − ) H t −
2
2
π
π
1
sin t − 3 cos t + 4eπ e−2t − 5e 2 e−t H t −
= e−t +
10
2
19
f(t) = 3H(t) − (8 − 2t)H(t − 4)
= 3H(t) + 2(t − 4)H(t − 4)
2
3
3
L{f(t)} = + 2e−4s L{t} = + 2 e−4s
s
s s
Taking Laplace transforms with x(0) = 1, ẋ(0) = 0
2
3
+ 2 e−4s
s s
s
3
2
X(s) = 2
+
+ 2 2
e−4s
2
s + 1 s(s + 1) s (s + 1)
1
3
3
s
1 −4s
+ − 2
+2 2 − 2
e
= 2
s +1 5 s +1
s
s +1
2
3
+ 2e−4s L{t − sin t}
= − 2
5 s +1
(s2 + 1)X(s) = s +
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
295
Thus, taking inverse transforms
x(t) = 3 − 2 cos t + 2(t − 4 − sin(t − 4))H(t − 4)
20
θ̈0 + 6θ̇0 + 10θ0 = θi
θi (t) = 3H(t) − 3H(t − a)
3 3
3
so L{θi } = − e−as = (1 − e−as )
s s
s
Taking Laplace transforms in (1) with θ0 = θ̇0 = 0 at t = 0
3
(1 − e−as )
s
Φ0 (s) = 3(1 − e−as )
(s2 + 6s + 10)Φ0 (s) =
1
s(s2 + 6s + 10)
1
3
(s + 3) + 3
=
(1 − e−as ) −
10
s (s + 3)2 + 1
3
(1 − e−as )L 1 − e−3t cos t − 3e−3t sin t
=
10
Thus, taking inverse transforms
θ0 (t) =
3
[1 − e−3t cos t − 3e−3t sin t]H(t)
10
3
− [1 − e−3(t−a) cos(t − a) − 3e−3(t−a) sin(t − a)]H(t − a)
10
If T > a then H(T) = 1, H(T − a) = 1 giving
3 −3T
cos T − e−3(T −a) cos(T − a)]
[e
10
3
− [3e−3T sin T − 3e−3(T −a) sin(T − a)]
10
3 −3T
{cos T + 3 sin T − e3a [cos(T − a) + 3 sin(T − a)]}
=− e
10
θ0 (T) = −
21
θi (t) = f(t) = (1 − t)H(t) − (1 − t)H(t − 1)
= (1 − t)H(t) + (t − 1)H(t − 1)
c Pearson Education Limited 2011
(1)
296
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so
1
1
− 2 + e−s L{t}
s s
1
1
1
s−1
1
= − 2 + 2 e−s =
+ 2 e−s
2
s s
s
s
s
L{θi (t)} =
Then taking Laplace transforms, using θ0 (0) = θ̇0 (0) = 0
(s2 + 8s + 16)Φ0 (s) =
s−1
1 −s
+
e
s2
s2
1
s−1
−s
+
e
s2 (s + 4)2
s2 (s + 4)2
2
10 e−s 3
2
2
3
1
1 3
− 2−
−
+
+
+
+
= 2
s s s
s + 4 (s + 4)2
32 s s2
s + 4 (s + 4)2
Φ0 (s) =
which on taking inverse transforms gives
1
[3 − 2t − 3e−4t − 10te−4t ]
32
1
+ [−1 + 2(t − 1) + e−4(t−1) + 2(t − 1)e−4(t−1) ]H(t − 1)
32
1
[3 − 2t − 3e−4t − 10te−4t ]
=
32
1
+ [2t − 3 + (2t − 1)e−4(t−1) ]H(t − 1)
32
θ0 (t) = L−1 {Φ0 (s)} =
22
e(t) = e0 H(t − t1 ) − e0 H(t − t2 )
e0
L{e(t)} = (e−st1 − e−st2 )
s
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
297
By Kirchhoff’s second law current in the circuit is given by
Ri +
1
C
idt = e
which on taking Laplace transforms
RI(s) +
I(s) =
1
e0
I(s) = (e−st1 − e−st2 )
Cs
s
e0 C
(e−st1 − e−st2 )
RCs + 1
=
e0 /R −st1
− e−st2 )
1 (e
s + RC
=
e0 /R −st2
e0 /R −st1
−
1 e
1 e
s + RC
s + RC
then
i(t) = L−1 {I(s)}
e0 −(t−t1 )/RC
e
H(t − t1 ) − e−(t−t2 )/RC H(t − t2 )
=
R
23
Sketch over one period as shown and
readily extended to 0 ≤ t < 12.
c Pearson Education Limited 2011
298
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
f1 (t) = 3tH(t) − (3t − 6)H(t − 2) − 6H(t − 4)
= 3tH(t) − 3(t − 2)H(t − 2) − 6H(t − 4)
3
3
6
L{f1 (t)} = F1 (s) = 2 − 2 e−2s − e−4s
s
s
s
Then by theorem 5.5
1
F1 (s)
1 − e−4s
1
= 2
(3 − 3e−2s − 6se−4s )
s (1 − e−4s )
L{f(t)} = F(s) =
24
Take
K
t,
T
= 0,
f1 (t) =
then f1 (t) =
0T
Kt
K
K
K
tH(t) −
H(t − T) = tH(t) − (t − T)H(t − T) − KH(t − T)
T
T
T
T
L{f1 (t)} = F1 (s) =
K
K
K
K
K
= 2 (1 − e−sT ) − e−sT
− e−sT 2 − e−sT
2
Ts
Ts
s
Ts
s
Then by theorem 5.5
L{f(t)} = F(s) =
K
K e−sT
1
F
(s)
=
−
1
1 − e−sT
Ts2
s 1 − e−sT
Exercises 5.5.12
25(a)
2s2 + 1
10s + 11
9
19
=2−
=2+
−
(s + 2)(s + 3)
(s + 2)(s + 3)
s+2 s+3
2
2s + 1
= 2δ(t) + 9e−2t − 19e−3t
L−1
(s + 2)(s + 3)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
25(b)
5
s2 − 1
=1− 2
2
s +4
s +4
2
s − 1
5
= δ(t) − sin 2t
L−1 2
s +4
2
25(c)
2s + 3
s2 + 2
=1− 2
2
s + 2s + 5
s + 2s + 5
2(s + 1) + 12 (2)
=1−
(s + 1)2 + s2
s2 + 2
1
= δ(t) − e−t 2 cos 2t + sin 2t
L−1 2
s + 2s + 5
2
26(a) (s2 + 7s + 12)X(s) =
2
+ e−2s
s
−2s
1
2
+
e
s(s + 4)(s + 3)
(s + 4)(s + 3)
1
2
1
1
1 −2s
6
3
e
= −
+ 2 +
−
s
s+3 s+4
s+3 s+4
1 2 −3t 1 −4t
− e
+ e−3(t−2) − e−4(t−2) H(t − 2)
x(t) = L−1 {X(s)} =
+ e
6 3
2
X(s) =
26(b)
(s2 + 6s + 13)X(s) = e−2πs
1
e−2πs
(s + 3)2 + 22
1
= e−2πs L e−3t sin 2t
2
1
so x(t) = L−1 {X(s)} = e−3(t−2π) sin 2(t − 2π).H(t − 2π)
2
1 6π −3t
sin 2t.H(t − 2π)
= e e
2
X(s) =
c Pearson Education Limited 2011
299
300
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
26(c)
(s2 + 7s + 12)X(s) = s + 8 + e−3s
X(s) =
=
−3s
1
s+8
+
e
(s + 4)(s + 3)
(s + 4)(s + 3)
5
4 1
1 −3s
−
+
−
e
s+3 s+4
s+3 s+4
x(t) = L−1 {X(s)} = 5e−3t − 4e−4t + [e−3(t−3) − e−4(t−3) ]H(t − 3)
27(a)
Generalized derivative is
f (t) = g (t) − 43δ(t − 4) − 4δ(t − 6)
where
⎧
6t,
⎪
⎨
g (t) =
2,
⎪
⎩
0,
0≤t<4
4≤t<6
t≥6
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
27(b)
⎧
⎨ 1,
f (t) = g (t) = −1,
⎩
0,
0≤t<1
1≤t<2
t≥2
27(c)
f (t) = g (t) + 5δ(t) − 6δ(t − 2) + 15δ(t − 4)
where
⎧
⎨
2,
−3,
g (t) =
⎩
2t − 1,
0≤t<2
2≤t<4
t≥4
c Pearson Education Limited 2011
301
302
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
28
(s2 + 7s + 10)X(s) = 2 + (3s + 2)U(s)
= 2 + (3s + 2)
X(s) =
=
x(t) = L−1 {X(s)} =
29 f(t) =
∞
5s + 6
1
=
s+2
s+2
5s + 6
(s + 2)2 (s + 5)
19
9
s+2
−
4
3
(s + 2)2
−
19
9
(s + 5)
19 −2t 4 −2t 19 −5t
e
− te
− e
9
3
9
δ(t − nT)
n=0
Thus,
F(s) = L{f(t)} =
∞
L{δ(t − nT)} =
n=0
∞
e−snT
n=0
This is an infinite GP with first term 1 and common ratio e−sT and therefore
having sum (1 − e−sT )−1 . Hence,
F(s) =
1
1 − e−sT
Assuming zero initial conditions and taking Laplace transforms the response of the
harmonic oscillator is given by
1
1 − e−sT
∞
−snT
1
X(s) =
e
2
s + w2
n=0
(s2 + w2 )X(s) = F(s) =
1
sin wt
= [1 + e−sT + e−2sT + . . .]L
w
giving x(t) = L−1 {X(s)} =
1
[sin wt + H(t − T). sin w(t − T) + H(t − 2T).
w
sin w(t − 2T) + . . .]
∞
1
or x(t) =
H(t − nT) sin w(t − nT) .
w n=0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
303
29(a)
π
;
T=
w
∞
1
nπ
x(t) =
sin(wt − nπ)
H t−
w n=0
w
π
2π
1
sin wt − sin wt. H t −
+ sin wt. H t −
+ ...
=
w
w
w
and a sketch of the response is as follows
29(b)
T=
2π
;
w
x(t) =
∞
2πn
1
sin(wt − 2πn)
H t−
w n=0
w
2π
4π
1
sin wt + sin wt.H t −
+ sin wt.H t −
+ ...
=
w
w
w
and the sketch of the response is as follows
c Pearson Education Limited 2011
304
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
30
The charge q on the LCR circuit is determined by
L
d2 q
dq
1
+
R
+
q = e(t)
dt2
dt
C
where e(t) = Eδ(t), q(0) = q̇(0) = 0.
Taking Laplace transforms
2
1
Q(s) = L{Eδ(t)} = E
Ls + Rs +
C
Q(s) =
=
s2
E/L
+R
Ls +
1
LC
E/L
=
(s +
R 2
2L )
1
+ ( LC
−
E/L
R
,η =
,μ =
2
2
(s + μ) + η
2L
R2
4L2 )
1
R2
−
LC 4L2
E −μt
e
sin ηt
Lη
E −μt
e (η cos ηt − μ sin ηt)
and current i(t) = q̇(t) =
Lη
Thus, q(t) =
Exercises 5.5.14
31
M
1
H(x) + Wδ x −
− R1 δ(x) , where R1 = (M + W)
2
2
so the force function is
Load W(x) =
M + W
M
H(x) + Wδ x −
−
δ(x)
2
2
W(x) =
having Laplace transform
W(s) =
M
(M + W)
+ We−s/2 −
s
2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
305
Since the beam is freely supported at both ends
y(0) = y2 (0) = y() = y2 () = 0
and the transformed equation (2.64) of the text becomes
y1 (0) y3 (0)
1 M
W −s/2 M + W 1
+
Y(s) =
+
e
−
+ 4
EI s5
s4
2
s4
s2
s
Taking inverse transforms gives
1
1 1 M 4 1
x + W(x − )3 · H x −
− (M + W)x3
y(x) =
EI 24
6
2
2
12
1
+ y1 (0)x + y3 (0)x3
6
for x >
2
1 1 M 4 1
x + W x−
y(x) =
EI 24
6
2
3
1
1
3
− (M + W)x + y1 (0)x + y3 (0)x3
12
6
1
1 1M 2
x +W x−
− (M + W)x + y3 (0)x
y2 (x) =
EI 2
2
2
y2 () = 0 then gives y3 (0) = 0 and y() = 0 gives
W3
1
1 M3
1 3
3
+
− M − W + y1 (0)
0=
EI 24
24
12
2
1 1
1
M2 + W2
y1 (0) =
EI 24
16
2
1
Mx4 + 8W(x − )3 H x −
− 4(M + W)x3 + (2M + 3W)2 x
so y(x) =
48EI
2
2
c Pearson Education Limited 2011
306
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
32
Load W(x) = w(H(x − x1 ) − H(x − x2 )) − R1 δ(x), R1 = w(x2 − x1 )
so the force function is
W(x) = w(H(x − x1 ) − H(x − x2 )) − w(x2 − x1 )δ(x)
having Laplace transform
1
1
W(s) = w e−x1 s − e−x2 s − w(x2 − x1 )
s
s
with corresponding boundary conditions
y(0) = y1 (0) = 0, y2 () = y3 () = 0
The transformed equation (2.64) of the text becomes
y2 (0) y3 (0)
1 −x2 s (x2 − x1 )
w 1 −x1 s
+ 3 + 4
e
− 5e
−
Y(s) =
5
4
EI s
s
s
s
s
which on taking inverse transforms gives
y(x) =
w1
1
(x − x1 )4 H(x − x1 ) − (x − x2 )4 H(x − x2 )
EI 24
24
x2
x3
1
3
− (x2 − x1 )x + y2 (0) + y3 (0)
6
2
6
For x > x2
w1
1
1
x2
x3
(x − x1 )4 − (x − x2 )4 − (x2 − x1 )x3 + y2 (0) + y3 (0)
EI 24
24
6
2
6
w1
1
y2 (x) =
(x − x1 )2 − (x − x2 )2 − (x2 − x1 )x + y2 (0) + y3 (0)x
EI 24
2
w
(x − x1 ) − (x − x2 ) − (x2 − x1 ) + y3 (0) ⇒ y3 (0) = 0
y3 (x) =
EI
y(x) =
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The boundary condition y2 () = 0 then gives
w 1 2
1
( − 2x1 + x21 ) − (2 − 2x2 + x22 ) − x2 + x1 + y2 (0)
EI 2
2
w
2
2
(x − x1 )
⇒ y2 (0) =
2EI 2
w
(x − x1 )4 H(x − x1 ) − (x − x2 )4 H(x − x2 ) − 4(x2 − x1 )x3
y(x) =
24EI
+ 6(x22 − x21 )x2
0=
When x1 = 0, x2 = , max deflection at x =
ymax =
w
w4
{4 − 44 + 64 } =
24EI
8EI
33
Load W(x) = Wδ(x − b) − R1 δ(x), R1 = W so the force function is
W(x) = Wδ(x − b) − Wδ(x)
having Laplace transform
W(s) = We−bs − W
with corresponding boundary conditions
y(0) = y1 (0) = 0, y2 () = y3 () = 0
The transformed equation (2.64) of the text becomes
Y(s) = −
1 W −bs W y2 (0) y3 (0)
e
− 4 + 3 + 4
EI s4
s
s
s
which on taking inverse transforms gives
1 3
x2
x3
W 1
3
y(x) = −
(x − b) H(x − b) − x + y2 (0) + y3 (0)
EI 6
6
2
6
c Pearson Education Limited 2011
307
308
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
For x > b
y(x) = −
1
x2
x3
W 1
(x − b)3 − x3 + y2 (0) + y3 (0)
EI 6
6
2
6
y2 (x) = −
W
(x − b) − x + y2 (0) + y3 (0)x
EI
y3 (x) = −
W
1 − 1 + y3 (0) ⇒ y3 (0) = 0
EI
Using the boundary condition y2 () = 0
0=−
Wb
W
(−h) + y2 (0) ⇒ y2 (0) = −
EI
EI
giving
(x − b)3
bx2
W x3
−
H(x − b) −
EI 6
6
2
⎧
2
Wx
⎪
⎨−
(3b − x), 0 < x ≤ b
6EI
=
2
⎪
⎩ − Wb (3x − b), b < x ≤
6EI
y(x) =
Exercises 5.6.5
34(a)
gives
Assuming all the initial conditions are zero taking Laplace transforms
(s2 + 2s + 5)X(s) = (3s + 2)U(s)
so that the system transfer function is given by
G(s) =
3s + 2
X(s)
= 2
U(s)
s + 2s + 5
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
34(b)
309
The characteristic equation of the system is
s2 + 2s + 5 = 0
and the system is of order 2.
34(c)
The transfer function poles are the roots of the characteristic equation
s2 + 2s + 5 = 0
which are s = −1 ± j. That is, the transfer function has single poles at s = −1 + j
and s = −1 − j.
The transfer function zeros are determined by equating the numerator polynomial
2
to zero; that is, a single zero at s = − .
3
35
Following the same procedure as for Exercise 34
35(a)
The transfer function characterizing the system is
G(s) =
s3 + 5s + 6
s3 + 5s2 + 17s + 13
c Pearson Education Limited 2011
310
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
35(b)
The characteristic equation of the system is
s3 + 5s2 + 17s + 13 = 0
and the system is of order 3.
35(c)
The transfer function poles are given by
s3 + 5s2 + 17s + 13 = 0
that is, (s + 1)(s2 + 4s + 13) = 0
That is, the transfer function has simple poles at
s = −1, s = −2 + j3, s = −2 − j3
The transfer function zeros are given by
s2 + 5s + 6 = 0
(s + 3)(s + 2) = 0
that is, zeros at s = −3 and s = −2.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
36(a)
311
Poles at (s + 2)(s2 + 4) = 0; that is, s = −2, s = +2j, s = −j.
Since we have poles on the imaginary axis in the s-plane, system is marginally
stable.
Poles at (s + 1)(s − 1)(s + 4) = 0; that is, s = −1, s = 1, s = −4.
36(b)
Since we have the pole s = 1 in the right hand half of the s-plane, the system is
unstable.
36(c)
Poles at (s + 2)(s + 4) = 0; that is, s = −2, s = −4.
Both the poles are in the left hand half of the plane so the system is stable.
2
2
Poles
√ at (s + s + 1)(s + 1) = 0; that is, s = −1 (repeated),
3
1
.
s=− ±j
2
2
Since all the poles are in the left hand half of the s-plane the system is stable.
36(d)
√
39
1
36(e) Poles at (s + 5)(s − s + 10) = 0; that is, s = −5, s = ± j
.
2
2
Since both the complex poles are in the right hand half of the s-plane the system
2
is unstable.
37(a)
s2 − 4s + 13 = 0 ⇒ s = 2 ± j3.
Thus, the poles are in the right hand half s-plane and the system is unstable.
37(b)
5s3 + 13s2 + 31s + 15 = 0
a1
a0
a3
a2
Routh–Hurwitz (R-H) determinants are:
13
Δ1 = 13 > 0, Δ2 =
15
5
> 0, Δ3 = 15Δ2 > 0
31
so the system is stable.
c Pearson Education Limited 2011
312
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
37(c)
s3 + s2 + s + 1 = 0
R–H determinants are
1
Δ1 = 1 > 0, Δ2 =
1
1
= 0, Δ3 = 1Δ2 = 0
1
Thus, system is marginally stable. This is readily confirmed since the poles are at
s = −1, s = ±j
37(d)
24s4 + 11s3 + 26s2 + 45s + 36 = 0
R–H determinants are
11
Δ1 = 11 > 0, Δ2 =
45
24
<0
26
so the system is unstable.
37(e)
s3 + 2s2 + 2s + 1 = 0
R–H determinants are
2
Δ1 = 2 > 0, Δ2 =
1
3
= 1 > 0, Δ3 = 1Δ2 > 0
2
√
3
1
confirming the
and the system is stable. The poles are at s = −1, s = − ± j
2
2
result.
d3 x
d2 x
dx
+ Krx = 0; m, K, r, c > 0
38 m 3 + c 2 + K
dt
dt
dt
R–H determinants are
Δ1 = c > 0
c m
= cK − mKr > 0 provided r < c
Δ2 =
Kr K
m
Δ3 = KrΔ2 > 0 provided Δ2 > 0
Thus, system stable provided r <
c
m
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
39
s4 + 2s2 + (K + 2)s2 + 7s + K = 0
a2
a1 a0
a3
R–H determinants are
Δ1 =| a3 |= 9 > 0
2
a3 a4
=
Δ2 =
7
a1 a2
a3 a4 0
Δ3 = a1 a2 a3 =
0 a0 a1
3
1
= 2K − 3 > 0 provided K >
K+2
2
2
1
0
7 K + 2 2 = 10K − 21 > 0 provided K > 2
0
K
7
Δ4 = KΔ3 > 0 provided Δ3 > 0
Thus, the system is stable provided K > 2.1.
40
s2 + 15Ks2 + (2K − 1)s + 5K = 0, K > 0
R–H determinants are
Δ1 = 15K > 0
15K
1
= 30K2 − 20K
Δ2 =
5K (2K − 1)
Δ3 = 5KΔ2 > 0 provided Δ2 > 0
Thus, system stable provided K(3K − 2) > 0 that is K >
41(a)
2
, since K > 0.
3
Impulse response h(t) is given by the solution of
d2 h
dh
+ 15
+ 56h = 3δ(t)
2
dt
dt
with zero initial conditions. Taking Laplace transforms
(s2 + 15s + 56)H(s) = 3
H(s) =
3
3
3
=
−
(s + 7)(s + 8)
s+7 s+8
so h(t) = L−1 {H(s)} = 3e−7t − 3e−8t
Since h(t) → 0 as t → ∞ the system is stable.
c Pearson Education Limited 2011
313
314
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
41(b)
Following (a) impulse response is given by
(s2 + 8s + 25)H(s) = 1
1
(s + 4)2 + 32
1
so h(t) = L−1 {H(s)} = e−4t sin 3t
3
H(s) =
Since h(t) → 0 as t → ∞ the system is stable.
41(c)
Following (a) impulse response is given by
(s2 − 2s − 8)H(s) = 4
4
2 1
2 1
=
−
(s − 4)(s + 2)
3s−4 3s+2
2
so h(t) = L−1 {H(s)} = (e4t − e−2t )
3
H(s) =
Since h(t) → ∞ as t → ∞ system is unstable.
41(d)
Following (a) impulse response is given by
(s2 − 4s + 13)H(s) = 1
H(s) =
so h(t) = L−1 {H(s)} =
s2
1
1
=
− 4s + 13
(s − 2)2 + 32
1 2t
e sin 3t
3
Since h(t) → ∞ as t → ∞ system is unstable.
7
dx
2
= e−t − 3e−2t + e−4t
42 Impulse response h(t) =
dt
3
3
System transfer function G(s) = L{h(t)}; that is,
3
2
7
−
+
3(s + 1) s + 2 3(s + 4)
s+8
=
(s + 1)(s + 2)(s + 4)
G(s) =
The original unit step response can be reconstructed by evaluating
Note:
1
L−1 G(s) .
s
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
f(t) = 2 − 3 cos t , F(s) =
43(a)
315
3s
2
− 2
s s +1
sF(s) = 2 −
3
3s2
=
2
−
s2 + 1
1 + s12
Thus, lim (2 − 3 cos t) = 2 − 3 = −1
t→0+
and lim sF(s) = 2 −
s→∞
3
= −1 so confirming the i.v. theorem.
1
43(b)
f(t) = (3t − 1)2 = 9t2 − 6t + 1, lim f(t) = 1
t→0+
18 6
18
6
1
+
1
=1
F(s) = 3 − 2 + so lim sF(s) = lim
−
s→∞
s→∞ s2
s
s
s
s
thus, confirming the i.v. theorem.
43(c)
f(t) = t + 3 sin 2t , lim = 0
t→0+
1
6
1
6
so lim sF(s) = lim
+
F(s) = 2 − 2
=0
s→∞
s→∞ s
s
s +4
s + 4s
thus, confirming the i.v. theorem.
44(a)
f(t) = 1 + 3e−t sin 2t , lim f(t) = 1
t→∞
F(s) =
6
1
6s
+
and
lim
=1
1
+
sF(s)
=
lim
s→0
s→0
s (s + 1)2 + 4
(s + 1)2 + 4
thus confirming the f.v. theorem. Note that, sF(s) has its poles in the left half of
the s-plane so the theorem is applicable.
c Pearson Education Limited 2011
316
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
44(b)
f(t) = t2 3e−2t , lim f(t) = 0
t→∞
2s
2
F(s) =
=0
and
lim
sF(s)
=
lim
s→0
s→0 (s + 2)3
(s + 2)3
thus confirming the f.v. theorem. Again note that sF(s) has its poles in the left
half of the s-plane.
44(c)
f(t) = 3 − 2e−3t + e−t cos 2t , lim f(t) = 3
t→∞
2
(s + 1)
3
+
F(s) = −
s s + 3 (s + 1)2 + 4
2s
s(s + 1)
=3
lim sF(s) = lim 3 −
+
s→0
s→0
s + 3 (s + 1)2 + 4
confirming the f.v. theorem. Again sF(s) has its poles in the left half of the
s-plane.
45
For the circuit of Example 5.28
I2 (s) =
1.22
4.86
3.64
+
−
s
s + 59.1 s + 14.9
Then by the f.v. theorem
4.86s
1.22s
−
lim i2 (t) = lim sI2 (s) = lim 3.64 +
t→∞
s→0
s→0
s + 59.1 s + 14.9
= 3.64
which confirms the answer obtained in Example 5.28. Note that, sI2 (s) has all its
poles in the left half of the s-plane.
46
For the circuit of Example 5.29
sI2 (s) =
28s2
(3s + 10)(s + 1)(s2 + 4)
and since it has poles at s = ±j2 not in the left hand half of the s-plane the
f.v. theorem is not applicable.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
47
Assuming quiescent initial state taking Laplace transforms gives
1
4
+
+2
s s+3
4
1
2
Y(s) =
+
+
s(7s + 5) (s + 3)(7s + 5) 7s + 5
s
2s
4
+
+
sY(s) =
7s + 5 (s + 3)(7s + 5) 7s + 5
(7s + 5)Y(s) =
By the f.v. theorem,
lim y(t) = lim sF(s) = lim
t→∞
s→0
s→0
=
s
2s
4
+
+
7s + 5 (s + 3)(7s + 5) 7s + 5
4
5
By the i.v. theorem,
lim y(t) = y(0+) = lim sF(s) = lim
s→∞
t→0+
s→∞
=
s
2
4
+
+
7s + 5 (1 + 3s )(7s + 5) 7 + 5s
2
7
2
Thus, jump at t = 0 = y(0+) − y(0−) = 1 .
7
Exercises 5.6.8
48(a)
t
f ∗ g(t) =
τ cos(3t − 3τ)dτ
0
t
1
1
= − τ sin(3t − 3τ) + cos(3t − 3τ)
3
9
0
1
= (1 − cos 3t)
9
t
(t − τ) cos 3τdτ
g ∗ f(t) =
0
=
t
t
1
τ
1
sin 3τ − sin 3τ − cos 3τ = (1 − cos 3t)
3
3
9
9
0
c Pearson Education Limited 2011
317
318
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
48(b)
t
f ∗ g(t) =
(τ + 1)e−2(t−τ ) dτ
0
t
1
1
= (τ + 1)e−2(t−τ ) − e−2(t−τ )
2
4
0
1
1 1
= t + − e−2t
2
4 4
t
g ∗ f(t) =
(t − τ + 1)e−2τ dτ
0
t
1
1
= − (t − τ + 1)e−2τ + e−2τ
2
4
0
1 1 −2t
1
= t+ − e
2
4 4
48(c)
Integration by parts gives
t
t
τ2 sin 2(t − τ)dτ =
(t − τ)2 sin 2τdτ
0
0
1
1
1
= cos 2t + t2 −
4
2
4
48(d)
Integration by parts gives
t
t
e−τ sin(t − τ)dτ =
0
e−(t−τ ) sin τdτ
0
1
= (sin t − cos t + e−t )
2
49(a)
Since L−1
L−1
1
1 2 −3t
1
= 1 = f(t) and L−1
= t e
s
(s + 3)3
2
1
1
·
=
3
s (s + 3)
t
f(t − τ)g(τ)dτ
0
t
1
1. τ2 e−3τ dτ
2
0
t
2
2
1
= −τ2 e−3τ − τe−3τ − e−3τ
4
3
9
0
1
=
[2 − e−3t (9t2 + 6t + 2)]
54
=
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
319
Directly
L−1
L−1
49(b)
L−1
2
1
1
18
6
2
−1 1
·
−
=
L
−
−
s (s + 3)3
54 s (s + 3)3
(s + 3)2
(s + 3)
1
[2 − e−3t (9t2 + 6t + 2)]
=
54
1
1
= te2t = f(t), L−1
= te−3t = g(t)
2
2
(s − 2)
(s + 3)
1
1
=
·
2
2
(s − 2) (s + 3)
=e
t
(t − τ)e2(t−τ ) .τe−3τ dτ
0
t
−2t
(tτ − τ2 )e−5τ dτ
0
1
1
2 −5τ t
e
= e2t − (tτ − τ2 )e−5τ − (t − 2τ)e−5τ +
5
25
125
0
t
2
t
2
= e2t
e−5t +
e−5t +
−
25
125
25 125
1 2t
e (5t − 2) + e−3t (5t + 2)
=
125
Directly
−2
1
2
1
1
125
25
125
25
+
+
=
+
(s − 2)2 (s + 3)2
s − 2 (s − 2)2
(s + 3) (s + 3)2
1
−2 2t
1
2 −3t
1
e + te2t +
e
=
∴ L−1
+ te−3t
2
2
(s − 2) (s + 3)
125
25
125
25
1 2t
[e (5t − 2) + e−3t (5t + 2)]
=
125
49(c)
L−1
1
1
= e−4t = g(t)
= t = f(t), L−1
2
s
(s + 4)
−1
L
1
1
=
·
s2 s + 4
t
(t − τ)e−4t dτ
0
t
1
1
= − (t − τ)e−4τ + e−4τ
4
16
0
1 −4t 1
1
=
+ t−
e
16
4
16
c Pearson Education Limited 2011
320
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Directly
L−1
50
1
1 1 1 1
1
−1 1
=
L
·
−
· + ·
s2 (s + 4)
16 s + 4 16 s 4 s2
1
1
1 −4t
e
+ t
−
=
16
16 4
Let f(λ) = λ and g(λ) = e−λ so
F(s) =
1
1
and
G(s)
=
s2
s+1
Considering the integral equation
t
y(t) =
λe−(t−λ) dλ
0
By (5.80) in the text
−1
L
t
{F(s)G(s)} =
f(λ)g(t − λ)dλ
0
t
=
λe−(t−λ) dλ = y(t)
0
so
1
+ 1)
1
1
1
= L−1 − + 2 +
s s
s+1
−t
= (t − 1) + e
y(t) = L−1 {F(s)G(s)} = L−1
51
s2 (s
Impulse response h(t) is given by the solution of
d2 h 7dh
+ 12h = δ(t)
+
dt2
dt
subject to zero initial conditions. Taking Laplace transforms
(s2 + 7s + 12)H(s) = 1
H(s) =
1
1
1
=
−
(s + 3)(s + 4)
s+3 s+4
giving h(t) = L−1 {H(s)} = e−3t − e−4t
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
321
Response to pulse input is
x(t) = A
t
[e−3(t−τ ) − e−4(t−τ ) ]dτ H(t)
0
t −3(t−τ )
[e
− e−4(t−τ ) ]dτ H(t − T)
−A
T
1 1 1 −3t 1 −4t
− − e
=A
H(t)
+ e
3 4 3
4
1 1 1 −3(t−T ) 1 −4(t−T )
− − − e
H(t − T)
− e
3 4 3
4
1
=
A 1 − 4e−3t + 3e−4t − (1 − 4e−3(t−T ) + 3e−4(t−T ) )H(t − T)
12
or directly
u(t) = A[H(t) − H(t − T)] so U(s) = L{u(t)} =
A
[1 − e−sT ]
s
Thus, taking Laplace transforms with initial quiescent state
A
[1 − e−sT ]
s
1 1 1
1
1
1
(1 − e−sT )
X(s) = A
· − ·
+ ·
12 s 3 s + 3 4 s + 4
A
[1 − 4e−3t + 3e−4t − (1 − 4e−3(t−T ) + 3e−4(t−T ) )H(t − T)]
x(t) = L−1 {X(s)} =
12
(s2 + 7s + 12)X(s) =
52
Impulse response h(t) is the solution of
d2 h
dh
+ 5h = δ(t), h(0) = ḣ(0) = 0
+4
2
dt
dt
Taking Laplace transforms
(s2 + 4s + 5)H(s) = 1
H(s) =
s2
1
1
=
+ 4s + 5
(s + 2)2 + 1
so h(t) = L−1 {H(s)} = e−2t sin t.
c Pearson Education Limited 2011
322
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
By the convolution integral response to unit step is
t
θ0 (t) =
e−2(t−τ ) sin(t − τ).1dτ
0
t
= e−2t
e2τ sin(t − τ)dτ
0
which using integration by parts gives
t
e−2t 2τ
e [2 sin(t − τ) + cos(t − τ)] 0
5
1 1 −2t
= − e (2 sin t + cos t)
5 5
θ0 (t) =
Check
Solving
d2 θ0
dθ0
+ 5θ0 = 1 , θ̇0 (0) = θ0 (0) = 0
+4
2
dt
dt
gives
1
s
1
1
s+4
1
=
− ·
Φ0 (s) =
2
s(s + 4s + 5)
5s 5 (s + 2)2 + 1
1 1
so θ0 (t) = L−1 {Φ0 (s)} = − [cos t + 2 sin t]e−2t .
5 5
(s2 + 4s + 5)Φ0 (s) =
Exercises 5.7.2
53 State–space form of model is
−5
=
3
x
y = cT x ⇒ y = [ 1 2 ] 1
x2
ẋ1
ẋ = Ax + bu ⇒
ẋ2
−1
−1
x1
2
u
+
5
x2
System transfer function is
G(s) = c (sI − A)
T
−1
b = [1
s+5
1
2]
−3 s + 1
c Pearson Education Limited 2011
−1
2
5
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
s+5
1
= Δ = (s + 5)(s + 1) + 3 = (s + 2)(s + 4),
det
−3 s + 1
−1
1 s + 1 −1
s+5
1
=
−3 s + 1
3
s+5
Δ
2
s + 1 −1
1
1
=Δ
Thus, G(s) = Δ [ 1 2 ]
(12s + 59)
5
3
s+5
so the system transfer function is
323
12s + 59
(s + 2)(s + 4)
G(s) =
54 In this case, the denominator can be factorized
G(s) =
s+1
Y(s)
=
U(s)
(s + 1)(s + 6)
and care must be taken not to cancel the common factor, to avoid the system being
mistaken for a first order system. To proceed it is best to model the system by the
differential equation
ÿ + 7ẏ + 6y = u̇ + u
from which
ẋ1 = −7x1 + x2 + u
ẋ2 = −6x1 + u
y = x1
so that a state–space model is
−7
=
−6
x1
T
y = c x ⇒ y = [1 0]
x2
ẋ1
ẋ = Ax + bu ⇒
ẋ2
1
0
x1
1
u
+
1
x2
Check
G(s) = cT (sI − A)−1 b
s + 7
det(sI − A) = Δ =
6
G(s) =
1
[1
Δ
−1
= s2 + 7s + 6
s
s+1
s+1
1
s
1
=
0]
= 2
1
−6 s + 7
Δ
s + 7s + 6
c Pearson Education Limited 2011
324
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
55(a)
Taking A to be the companion matrix
⎡
0
1
0
A= ⎣ 0
−7 −5
⎤
0
1⎦
−6
then b = [0 0 1]T
c = [5 3 1]T
Then from equation (5.84) in the text the state space form of the dynamic model
is
ẋ = A x + bu
y = cT x
55(b)
Taking A to be the companion matrix
⎤
⎡
0
1
0
0
1⎦
A = ⎣0
0 −3 −4
then b = [0 0 1]T
c = [2 3 1]T
And state space model is
ẋ = A x + bu, y = cT x
56
We are required to express the transfer function in the state space form
ẋ = A x + bu
y = cT x
⎡
⎤
0
1
0
0
1 ⎦ and y = [1 0 0]x . To
where A is the companion matrix A = ⎣ 0
−6 −11 −6
determine b, we divide the denominator into the numerator as follows
5s−1 − 29s−2 + 120s−3
s3 + 6s2 + 11s + 6
5s2 + s + 1
5s2 + 30s + 55 |
neglect these terms
−29s − 54 |
−29s − 174 |
120
|
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
325
giving b = [5 − 29 − 120]T . Thus, state space form is
⎡
⎤
⎡
ẋ1
0
⎣
⎦
⎣
0
ẋ(t) = ẋ2 =
−6
ẋ3
1
0
−11
⎤
⎤
⎡
5
0
1 ⎦ x(t) + ⎣ −29 ⎦ u(t)
120
−6
y = [1 0 0]x(t)
It is readily checked that this is a true representation of the given transfer function.
57
s−3
−2
[sI − A] =
−4
, det[sI − A] = (s − 5)(s + 1)
s−1
Thus,
2/3
2/3
1/3
+ s+1
−
4
s−5
s−5
=
[sI − A]
1/3
1/3
1/3
s−3
− s−5 − s+1 s−5 +
2 5t
e + 1 e−t 2 e5t − 2 e−t
∴ L−1 [sI − A]−1 = 31 5t 31 −t 31 5t 32 −t = Φ1
3e − 3e
3e + 3e
4
1
0 1
s−1
4
−1
s
[sI − A] B U(s) =
7
1 1
2
5−3
(s − 5)(s + 1)
s
11/3
4/3
5
−
+
+
1
3s + 25
s
s+1
s−5
=
=
11/3
2/3
3
7s
−
15
s(s − 5)(s + 1)
s − s+1 + s−5
−1
−1
∴L
1
=
(s − 5)(s + 1)
−1
{[sI − A]
s−1
2
−t
+ 43 e5t
−5 + 11
3 e
B U(s)} =
−t
3 − 11
+ 23 e5t
3 e
= Φ2
Thus, solution is
x(t) = Φ1 x(0) + Φ2 =
5t
−5 + 83 e−t + 10
3 e
3 − 83 e−t + 53 e5t
which confirms the answer obtained in Exercise 61 in Chapter 1.
c Pearson Education Limited 2011
2/3
s+1
2/3
s+1
326
58
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
s−1
3
1 −3
, sI − A =
A=
−2 s + 4
2 −4
| sI − A | = Δ = (s − 1)(s + 4) + 6 = (s + 2)(s + 1)
1 s + 4 −3
−1
[sI − A] =
2
s−1
Δ
3
2
3
3
− s+2 − s+1
+ s+2
s+1
=
2
2
2
3
− s+1
+ s+2
s+1 − s+2
3e−t − 2e−2t −3e−t + 3e−2t
−1
−1
giving L [sI − A] =
2e−t − 2e−2t −2e−t + 3e−2t
e2t
−1
−1
so L [sI − A] x(0) =
e−2t
1
1 s + 4 −3
1
−1
[sI − A] bU(s) =
1 s+3
2
s−1
Δ
1
1
1
− s+3
(s+2)(s+3)
s+2
=
=
1
1
1
(s+2)(s+3)
s+2 − s+3
e−2t − e−3t
so L−1 {[sI − A]−1 bU(s)} =
e−2t − e−3t
Thus,
x(t) =
e−2t
e−2t
+
e−2t − e−3t
e−2t − e−3t
giving
x1 (t) = x2 (t) = 2e−2t − e−3t
2
0
1
, u(t) = e−t H(t), x0 = [1 0]T
, b=
59 A =
0
−2 −3
with X(s) and the solution x(t) given by equations (5.88) and (5.89) in the text
1
s+3 1
s −1
−1
giving (sI − A) =
(sI − A) =
−2 5
2 s+3
(s + 1)(s + 2)
so that,
(sI − A)
−1
1
x0 =
(s + 1)(s + 2)
s+3
−2
=
2
1
s+1 − s+2
2
2
− s+1
+ s+2
and
(sI − A)
−1
1
bU(s) =
(s + 1)(s + 2)
2(s + 3)
−4
1
=
s+1
c Pearson Education Limited 2011
2
s+2
4
s+1
+
−
4
(s+1)2
4
(s+1)2
−
−
2
s+1
4
s+2
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, X(s) =
giving x(t) =
60
4
(s+1)2
+
327
1
(s+2)
2
4
− (s+2)
− (s+1)
2
−t
−2t
4te + e
2e−t − 2e−2t − 4te−t
2
(s+1)
Taking A as the companion matrix following the procedure of Example 5.61
we have
⎡
0
⎣
0
A=
−6
1
0
−11
⎤
0
1 ⎦, b = [0 0 1]T , c = [1 2 3]T
−6
Eigenvalue of A given by
−λ
1
0
−λ
−6 −11
0
1 = −(λ3 + 6λ2 + 11λ + 6) = −(λ + 1)(λ + 2)(λ + 3) = 0
−6 − λ
so the eigenvalues are λ1 = −3, λ2 = −2, λ3 = −1. The eigenvectors are given by
the corresponding solutions of
−λi ei1 + ei2 + 0ei3 = 0
0ei1 − λi ei2 + ei3 = 0
−6ei1 − 11ei2 − (6 + λi )ei3 = 0
Taking i = 1, 2, 3 and solving gives the eigenvectors as
e1 = [1 − 3 9]T , e2 = [1 − 2 4]T , e3 = [1 − 1 1]T
⎤
1
1
1
Taking M to be the modal matrix M = ⎣ −3 −2 −1 ⎦ then the transformation
9
4
1
X = M ξ will reduce the system to the canonical form
⎡
ξ̇ξ = Λ ξ + M−1 bu , y = cT M ξ
⎡
M−1
2
1 ⎣
−6
=
2
6
⎤
⎤
⎡
3
1
1
1 ⎣
−8 −2 ⎦ , M−1 b =
−2 ⎦ , cT M = [22 9 2]
2
5
1
1
c Pearson Education Limited 2011
328
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, canonical form is
⎤
⎡
−3
ξ̇1
⎣ ξ̇2 ⎦ = ⎣ 0
0
ξ̇3
⎡
0
−2
0
⎤ ⎡ ⎤
⎡ 1 ⎤
ξ1
0
2
0 ⎦ ⎣ ξ2 ⎦ + ⎣ −1 ⎦ u
1
−1
ξ3
2
y = [22 9 2] [ξ1 ξ2 ξ3 ]T
Since the eigenvalues of A are negative the system is stable. Since vector M−1 b
has no zero elements the system is controllable and since cT M has no zero elements
the system is also observable.
b = [0 0 1]T, A b = [0 1 − 6]T, A2 b = [1 6 25]T
⎡
0
[b A b A2 b] = ⎣ 0
1
0
1
−6
⎤
⎤
⎡
0 0 1
1
6 ⎦ ∼ ⎣ 0 1 0 ⎦ which is of full rank 3
1 0 0
25
so the system is controllable.
c = [1 2 3]T, AT c = [−18 − 32 − 16],
⎤
⎡
⎡
1
1 −18 96
T
T 2
⎦
⎣
⎣
[c A c (A ) c] = 2 −32 158 ∼ 0
0
3 −16 63
(AT )2 c = [96 158 63]T
⎤
0 0
1 0 ⎦ which is of full rank 3
0 1
so the system is observable.
⎤
0
1
0
0
1 ⎦, b = [0 0 1]T, c = [5 3 1]T
61 A = ⎣ 0
0 −5 −6
Eigenvalues of A given by
⎡
−λ
1
0 = 0 −λ
0 −5
0
1 = −λ(λ + 5)(λ + 1) = 0
−6 − λ
so eigenvalues are λ1 = −5, λ2 = −1, λ3 = 0. The corresponding eigenvectors are
determined as
e1 = [1 − 5 25]T , e2 = [1 − 1 1] , e3 = [1 0 0]T
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition 329
⎤
⎡
1
1 1
Take M to be the modal matrix M = ⎣ −5 −1 0 ⎦ then the transformation
25
1 0
x = Mξξ will reduce the system to the canonical form
M−1
ξ̇ξ = Λ ξ + M−1 bu , y = cT M ξ
⎤
⎤
⎡
⎡
0
1
1
1
1 ⎣
1 ⎣
0 −25 −5 ⎦, M−1 b =
−5 ⎦, cT M = [15 3 5]
=
20
20
20
24
4
4
Thus, canonical form is
⎡ ⎤
⎡
ξ̇1
−5
⎣ ξ̇2 ⎦ = ⎣ 0
0
ξ̇3
⎤ ⎡ ⎤
⎡ 1 ⎤
ξ1
0 0
20
−1 0 ⎦ ⎣ ξ2 ⎦ + ⎣ − 41 ⎦ u
1
0 0
ξ3
5
y = [15 3 5] [ξ1 ξ2 ξ3 ]T
Since A has zero eigenvalues the system is marginally stable, since M−1 b has no
zero elements the system is controllable and since cT M has no zero elements the
system is observable. Again as in Exercise 60 these results can be confirmed by
using the Kalman matrices.
Exercises 5.7.4
1
1 −1
0 −2
0
, x0 =
, u=
, B=
62 A =
t
1
1
1 −3
1
The solution x(t) is given by equation (5.97) in the text.
(sI − A) =
s
2
−1 s + 3
giving (sI − A)
−1
1
=
(s + 1)(s + 2)
s+3
1
−2
s
so that,
(sI − A)
−1
1
x0 =
(s + 1)(s + 2)
−2
s
=
1
s + 3 −2
1
s
(s + 1)(s + 2)
(s+5)
1
−
2
s (s+1)(s+2)
= s(s+2)
s−1
1
+
s(s+2)
s2 (s+1)(s+2)
(sI − A)−1 B U(s) =
c Pearson Education Limited 2011
2
s+2
2
s+2
−
−
1
1
2
s+1
1
s+1
−1
1
⎡
⎣
1
s
1
s2
⎤
⎦
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
330
Thus,
2
2
1
(s + 5)
−
+
− 2
s + 2 s + 1 s(s + 2) s (s + 1)(s + 2)
6
15/4 5/2
9/4
−
+
− 2
=
s+2 s+1
s
s
X1 (s) = L{x1 (t)} =
giving x1 (t) =
63
9 −2t
15 5
e
− t
− 6e−t +
4
4
2
ÿ1 + ẏ1 − ẏ2 + y1 = u1
(i)
ÿ2 + ẏ2 − ẏ1 + y2 = u2
(ii)
Let x = [ x1
x2
T
x3
x4 ] = [ y 1
ẏ1
y2
T
ẏ2 ] then
ẋ1 = ẏ1 = x2
(i) ⇒ ÿ1 = ẋ2 = −ẏ1 + ẏ2 − y1 + u1 ⇒ ẋ2 = −x2 + x4 − x1 + u1
ẋ3 = ẏ2 = x4
(ii) ⇒ ÿ2 = ẋ4 = xx − ẏ2 + ẏ1 − y2 + u2 ⇒ ẋ4 = −x4 + x2 − x3 + u2
giving the state–space representation
⎤⎡ ⎤ ⎡
x1
1
0
0
0
−1 0
1 ⎥ ⎢ x2 ⎥ ⎢ 1
⎦⎣ ⎦ + ⎣
0
0
1
0
x3
1 −1 −1
0
x4
⎡ ⎤
x1
1 0 0 0 ⎢ x2 ⎥
y
y = Cx ⇒ 1 =
⎣ ⎦
0 0 1 0
y2
x3
x4
⎤ ⎡
0
ẋ1
⎢ ẋ ⎥ ⎢ −1
ẋ = Ax + Bu ⇒ ⎣ 2 ⎦ = ⎣
0
ẋ3
0
ẋ4
⎡
Transfer function = G(s) =
⎤
0
0 ⎥ u1
⎦
0
u2
1
Cadj(sI−A)B
det(sI−A)
0
s −1 0
1 s + 1 0 −1
det(sI − A) = Δ =
= s4 + 2s3 + 2s2 + 2s + 1
0
0
s
−1
0 −1 1 s + 1
⇒ Δ = (s + 1)2 (s2 + 1)
so that
1 s2 + s + 1
G(s) =
s
Δ
1
s2 + s + 1
s
=
2
s +s+1
s
(s + 1)2 (s2 + 1)
c Pearson Education Limited 2011
s
2
s +s+1
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
331
Poles given by Δ = (s + 1)2 (s2 + 1) = 0
Eigenvalues associated matrix A given by det(λI − A) = (λ + 1)2 (λ2 + 1) = 0
Thus, poles and eigenvalues are identical.
64 (a)
u2 = R1 (u1 + x1 + x2 + x3 ) + L1 ẋ1 ; R1 = 1, L1 = 1
⇒ ẋ1 = −x1 − x2 − x3 − u1 + u2
u2 = R1 (u1 + x1 + x2 + x3 ) + R2 (x2 + x3 ) + L2 ẋ2 ; R2 = 2, L2 = 1
⇒ ẋ2 = −x1 − 3x2 − 3x3 − u1 + u2
u2 = R1 (u1 + x1 + x2 + x3 ) + R2 (x2 + x3 ) + R3 x3 + L3 ẋ3 ; R3 = 3, L3 = 1
⇒ ẋ3 = −x1 − 3x2 − 6x3 − u1 + u2
giving the state–space model
⎤⎡ ⎤ ⎡
⎤ ⎡
x1
ẋ1
−1 −1 −1
−1
⎣ ẋ2 ⎦ = ⎣ −1 −3 −3 ⎦ ⎣ x2 ⎦ + ⎣ −1
−1 −3 −6
−1
ẋ3
x3
⎡
y1
y2
(b) Transfer matrix = G(s) =
0
=
0
Y(s)
U(s)
2
0
2
1
x1
x2
⎤
1
u
1
1⎦
u2
1
= C(sI − A)−1 B = C adj(sI−A)
det(sI−A) B
s + 1
1
1
s+3
3 = s3 + 10s2 + 16s + 6
det(sI − A) = Δ = 1
1
3
s + 6
⎡ 2
⎤
s + 9s + 9
−(s + 3)
−s
s2 + 7s + 5 −(3s + 2) ⎦
adj(sI − A) = ⎣ −(s + 3)
−s
−(3s + 2) s2 + 4s + 2
⎡
⎤⎡
s2 + 9s + 9
−1
−(s + 3)
−s
1 0 2 2 ⎣
2
⎦
⎣
−1
−(s + 3)
s + 7s + 5 −(3s + 2)
G(s) =
Δ 0 0 1
−1
−s
−(3s + 2) s2 + 4s + 2
1 −2s(2s + 3) 2s(2s + 3)
=
with Δ = s3 + 10s2 + 16s + 6
−s2
s2
Δ
c Pearson Education Limited 2011
⎤
1
1⎦
1
332
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
(c) Y(s) = G(s)U(s) where U(s) =
1 −2s(2s + 3)
=
−s2
Δ
2
1 −4s −2s+6
s
=
Δ
−s + 1
1
s
1
s2
2s(2s + 3)
s2
1
s
1
s2
To obtain response express in partial fractions and take inverse Laplace transforms
Factorizing Δ gives Δ = (s + 8.12)(s + 0.56)(s + 1.32) so
Y1 (s) =
1
0.578
1.824
0.246
−4s2 − 2s + 6
= +
−
+
s(s + 8.12)(s + 0.56)(s + 1.32)
s s + 8.12 s + 0.56 s + 1.32
⇒ y1 (t) = 1 + 0.578e−8.12t − 1.824e−0.56t + 0.246e−1.32t
0.177
0.272
0.449
−s + 1
=
+
−
Y2 (s) =
(s + 8.12)(s + 0.56)(s + 1.32)
s + 8.12 s + 0.56 s + 1.32
⇒ y2 (t) = 0.177e−8.12t + 0.272e−0.56t − 0.449e−1.32t
Exercises 5.9.3
65
dy
then
dt
1
0
u(t), y(t) = [1 0]x(t)
x(t) +
1
1
2
Choose x1 (t) = y(t), x2 (t) = ẋ1 (t) =
ẋ(t) =
0
1
2
Taking u(t) = K1 x1 (t) + K2 x2 (t) + uext (t)
0
ẋ(t) =
K1 +
1
2
1
K2 +
1
2
x(t) +
0
u
1 ext
The eigenvalues of the matrix are given by
0−λ
K1 + 1
2
1
=0
1
K2 + 2 − λ
or λ2 − (K2 + 12 )λ − (K1 + 12 ) = 0
If the poles are to be at λ = −4 then we require the characteristic equation to be
λ2 + 8λ + 16 = 0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
By comparison, we have −K2 − 12 = 8 and −K1 −
33 17
x + uext
so
u(t)
=
− −
K2 = − 17
2
2
2
66
ẋ(t) =
0
1
− 14
− 54
1
2
333
= 18 giving K1 = − 33
2 ,
0
u(t) , y(t) = [0 2]x(t)
1
x(t) +
Setting u = KT x + uext , K = [K1 K2 ]T gives the system matrix
A=
0
K1 −
1
4
1
K2 −
5
4
whose eigenvalues are given by λ2 − (K2 − 54 )λ − (K1 − 54 ) = 0. Comparing with
the desired characteristic equation
(λ + 5)2 = λ2 + 10λ + 25 = 0
35
gives K1 = − 99
4 , K2 = − 4 . Thus,
99 35
x(t) + uext
u(t) = − −
4
4
67
With u1 = [K1 K2 ]x(t) the system matrix A becomes
1 + K2
K1
A=
6 + K1 1 + K2
having characteristic equation
λ2 − λ(1 + K1 + K2 ) − 6(1 − K2 ) = 0
which on comparing with
λ2 + 10λ + 25 = 0
35 31
35
31
so that u1 (t) = − −
x(t)
gives K1 = − , K2 = −
6
6
6
6
0
1
Using u2 (t) the matrix A becomes
where KT = [−31 − 11]
6 + K1 1 + K2
68
See p. 472 in the text.
c Pearson Education Limited 2011
334
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
69
For the matrix of Exercise 68 the Kalman matrix is
1 0
2 −2
∼
M=
0 0
1 −1
which is of rank 1. Thus, the system is uncontrollable.
For the matrix of Exercise 65 the Kalman matrix is
0 1
0 1
M=
∼
1 12
1 0
which is of full rank 2. Thus, the system is controllable.
70
M = (b A b) =
8
35
v A = [4 − 1]
T
1 0
0
−1
, vT = [4 − 1]
, M =
4 −1
−1
4 −1
−2
1
, T−1 =
= [−3 1] so T =
−3 1
−9
3
1
4
1
4
Taking z(t) = Tx or x = T−1 z(t) then equation reduces to
1
T z(t) +
u
T ż(t) =
4
8 −2
1
−1
T z(t) + T
u
or ż(t) = I
35 −9
4
4 −1
1 1
8 −2
4 −1
z(t) +
=
−3
1
3 4
35 −9
−3 1
0
0 1
u
z(t) +
=
1
2 −1
−1
8
35
−2
−9
−1
1
4
u
Clearly both system matrices have the same eigenvalues λ = −2, λ = 1. This will
always be so since we have carried out a singularity transformation.
Review Exercises 5.10
1(a)
d2 x
dx
+ 5x = 8 cos t, x(0) = ẋ(0) = 0 Taking Laplace transforms
+4
2
dt
dt
(s2 + 4s + 5)X(s) =
8s
+1
s2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
335
8s
(s2 + 1)(s2 + 4s + 5)
s+1
s+5
= 2
− 2
s + 1 s + 4s + 5
1
(s + 2) + 3
s
+ 2
−
= 2
s + 1 s + 1 (s + 2)2 + 1
X(s) =
giving x(t) = L−1 {X(s)} = cos t + sin t − e−2t [cos t + 3 sin t]
d2 x
dx
− 2x = 6, x(0) = ẋ(0) = 1
−3
2
dt
dt
Taking Laplace transforms
1(b) 5
5s2 + 2s + 6
6
(5s − 3s − 2)X(s) = 5(s + 1) − 3(1) + =
s
s
5s2 + 2s + 6
X(s) =
5s(s + 25 )(s + 1)
2
13
15
3
=− + 7 + 7 2
5 s−1 s+ 5
giving x(t) = L−1 {X(s)} = −3 +
13 t 15 − 2 t
e + e 5
7
7
2(a)
Thus , L−1
2(b)
1
1
1
1
1
s+2
=
− ·
− · 2
2
(s + 1)(s + 2)(s + 2s + 2)
s + 1 2 s + 2 2 s + 2s + 2
1
1
1 (s + 1) + 1
1
− ·
− ·
=
s + 1 2 s + 2 2 (s + 1)2 + 1
1
1
1
= e−t − e−2t − e−t (cos t + sin t)
2
(s + 1)(s + 2)(s + 2s + 2)
2
2
From equation (5.26) in the text the equation is readily deduced.
Taking Laplace transforms
(s2 + 3s + 2)I(s) = s + 2 + 3 + V.
1
(s + 1)2 + 1
1
s+5
+V
(s + 2)(s + 1)
(s + 2)(s + 1)(s2 + 2s + 2)
3
4
−
+ V extended as in (a)
=
s+1 s+2
I(s) =
c Pearson Education Limited 2011
336
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, using the result of (a) above
1
1
i(t) = L−1 {I(s)} = 4e−t − 3e−2t + V e−t − e−2t − e−t (cos t + sin t)
2
2
3
Taking Laplace transforms
1
s2
2
−2sX(s) + (s2 − 4)Y(s) = −
s
(s2 − 1)X(s) + 5sY(s) =
Eliminating Y(s)
[(s2 − 1)(s2 − 4) + 2s(5s)]X(s) =
s2 − 4
11s2 − 4
+
10
=
s2
s2
11s2 − 4
s2 (s2 + 1)(s2 + 4)
1
5
4
=− 2 + 2
− 2
s
s +1 s +4
X(s) =
giving x(t) = L−1 {X(s)} = −t + 5 sin t − 2 sin 2t
From the first differential equation
1
d2 x
dy
= t+x− 2
dt
5
dt
1
= [t − t + 5 sin t − 2 sin 2t + 5 sin t − 8 sin 2t]
5
= (2 sin t − 2 sin 2t)
then y = −2 cos t + cos 2t + const.
and since y(0) = 0, constant = 1 giving
y(t) = 1 − 2 cos t + cos 2t
x(t) = −t + 5 sin t − 2 sin 2t
4
Taking Laplace transforms
(s2 + 2s + 2)X(s) = sx0 + x1 + 2x0 +
c Pearson Education Limited 2011
s2
s
+1
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
337
s
sx0 + x1 + 2x0
+ 2
2
2
s + 2s + 2
(s + 1)(s + 2s + 2)
1
s+4
x0 (s + 1) + (x1 + x0 ) 1 s + 2
+ · 2
− ·
=
2
(s + 1) + 1
5 s + 1 5 (s + 1)2 + 1
X(s) =
giving
x(t) = L−1 {X(s)}
1
= e−t (x0 cos t + (x1 + x0 ) sin t) + (cos t + 2 sin t)
5
1
− e−t (cos t + 3 sin t)
5
that is,
x(t) =
1
1
3
(cos t + 2 sin t) + e−t (x0 − ) cos t + (x1 + x0 − ) sin t
5
5
5
↑
↑
steady state
transient
2
1
Steady state solution is xs (t) = cos t + sin t ≡ A cos(t − α)
5
5
1
1 2
2 2
having amplitude A = ( 5 ) + ( 5 ) = √
5
−1
◦
and phase lag α = tan 2 = 63.4 .
5
Denoting the currents in the primary and secondary circuits by i1 (t) and i2 (t)
respectively Kirchoff’s second law gives
di1
di2
+
= 100
dt
dt
di1
di2
+
=0
20i2 + 3
dt
dt
5i1 + 2
Taking Laplace transforms
100
s
sI1 (s) + (3s + 20)I2 (s) = 0
(5 + 2s)I1 (s) + sI2 (s) =
Eliminating I1 (s)
[s2 − (3s + 20)(2s + 5)]I2 (s) = 100
c Pearson Education Limited 2011
338
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
20
−100
=
−
5s2 + 55s + 100
s2 + 11s + 20
1
20
20
√
=
−
=−
11 2
41
11
(s + 2 ) − 4
41 (s + 2 −
I2 (s) =
√
41
2 )
−
1
(s +
11
2
+
√
41
2 )
giving the current i2 (t) in the secondary loop as
√
20 −(11+√41)t/2
e
− e−(11− 41)t/2
i2 (t) = L−1 {I2 (s)} = √
41
6(a)
(i)
L{cos(wt + φ)} = L{cos φ cos wt − sin φ sin wt}
s
w
= cos φ 2
− sin φ 2
2
s +w
s + w2
= (s cos φ − w sin φ)/(s2 + w2 )
(ii)
L{e−wt sin(wt + φ)} = L{e−wt sin wt cos φ + e−wt cos wt sin φ}
s+w
w
+ sin φ
= cos φ
2
2
(s + w) + w
(s + w)2 + w2
= [sin φ + w(cos φ + sin φ)]/(s2 + 2sw + 2w2 )
6(b)
Taking Laplace transforms
s
+4
3
2
2s + 9s + 9s + 36
= 2
(s + 4)(s2 + 4s + 8)
1 39s + 172
1 s+4
· 2
+ · 2
=
20 s + 4 20 s + 4s + 8
1 s+4
1 39(s + 2) + 47(2)
=
· 2
+ ·
20 s + 4 20 (s + 2)2 + (2)2
(s2 + 4s + 8)X(s) = (2s + 1) + 8 +
giving x(t) = L−1 {X(s)} =
7(a)
L−1
s2
1
1
(cos 2t + 2 sin 2t) + e−2t (39 cos 2t + 47 sin 2t) .
20
20
s−4
−1 (s + 2) − 2(3)
=
L
s2 + 4s + 13
(s + 2)2 + 32
= e−2t [cos 3t − 2 sin 3t]
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
7(b)
Taking Laplace transforms
(s + 2)Y(s) = −3 +
4
2s
4
+ 2
+ 2
s s +1 s +1
Y(s) =
−3s3 + 6s2 + s + 4
s(s + 2)(s2 + 1)
=
5
2
2
−
+ 2
s s+2 s +1
Therefore, y(t) = L−1 {Y(s)} = 2 − 5e−2t + 2 sin t
8
Taking Laplace transforms
(s + 5)X(s) + 3Y(s) = 1 +
5X(s) + (s + 3)Y(s) =
2s
s2 − 2s + 6
5
−
=
s2 + 1 s2 + 1
s2 + 1
6
3s
6 − 3s
−
=
s2 + 1 s2 + 1
s2 + 1
Eliminating Y(s)
[(s + 5)(s + 3) − 15]X(s) =
(s2 + 8s)X(s) =
X(s) =
(s + 3)(s2 − 2s + 6) 3(6 − 3s)
− 2
s2 + 1
s +1
s3 + s2 + 9s
s2 + 1
1
1
s2 + s + 9
=
+ 2
2
(s + 8)(s + 1)
s+8 s +1
so x(t) = L−1 {X(s)} = e−8t + sin t
From the first differential equation
3y = 5 sin t − 2 cos t − 5x −
dx
= 3e−8t − 3 cos t
dt
Thus, x(t) = e−8t + sin t, y(t) = e−8t − cos t .
c Pearson Education Limited 2011
339
340
9
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Taking Laplace transforms
100
+ 104
2
(s + 100)(s + 200)Q(s) = 104 · 2
s + 104
2.104
Q(s) =
(s + 100)(s + 200)(s2 + 104 )
1
2
1
1 3s − 100
1
·
−
·
−
·
=
100 s + 100 500 s + 200 500 s2 + 104
(s2 + 300s + 2 × 104 )Q(s) = 200·
giving q(t) = L−1 {Q(s)} =
that is,
q(t) =
s2
1 −100t
2 −200t
1
e
e
(3 cos 100t − sin 100t)
−
−
100
500
500
1
1
[5e−100t − 2e−200t ] −
[3 cos 100t − sin 100t]
500
500
↑
↑
transient
steady state
1
3
sin 100t + cos 100t ≡ A sin(100t + α)
5o
5
18 12 .
Steady state current =
where α = tan−1
1
5
o
Hence, the current leads the applied emf by about 18 12 .
10
dx
+ 6x + y = 2 sin 2t
dt
d2 x
dy
= 3e−2t
+x−
2
dt
dt
4
dx
= −2 when t = 0 so from (i) y = −4 when t = 0.
dt
Taking Laplace transforms
Given x = 2 and
8s2 + 36
4
=
s2 + 4
s2 + 4
2s2 + 6s + 7
3
=
(s2 + 1)X(s) − sY(s) = 2s − 2 + 4 +
s+2
s+2
(4s + 6)X(s) + Y(s) = 8 +
Eliminating Y(s)
[s(4s + 6) + (s2 + 1)]X(s) =
8s2 + 36 2s2 + 6s + 7
+
s2 + 4
s+2
c Pearson Education Limited 2011
(i)
(ii)
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
X(s) =
=
=
341
2s2 + 6s + 7
8s2 + 36
+
s(s2 + 4)(s + 1)(s + 15 ) 5(s + 2)(s + 1)(s + 15 )
11
5
s+1
29
20
s+1
−
+
227
505
s+
1
5
1
3
s+2
1
3
49
1 76s − 96
3
4
· 2
+
−
+ 60 1
−
505 s + 4
s+2 s+1 s+ 5
+
445
1212
s + 15
−
1 76s − 96
505 s2 + 4
giving
x(t) = L−1 {X(s)} =
11(a)
29 −t 1 −2t
445 − 1 t
1
e + e
e 5 −
(76 cos 2t − 48 sin 2t)
+
20
3
1212
505
Taking Laplace transforms
(s2 + 8s + 16)Φ(s) =
s2
2
+4
Φ(s) =
2
4)2 (s2
(s +
+ 4)
1
1 2s − 3
1
1
1
=
− · 2
·
+ ·
2
25 s + 4 10 (s + 4)
50 s + 4
1 −4t
1
1
e
(4 cos 2t − 3 sin 2t)
so θ(t) = L−1 {Φ(s)} =
+ · te−4t −
25
10
100
1
that is, θ(t) =
(4e−4t + 10te−4t − 4 cos 2t + 3 sin 2t)
100
11(b)
Taking Laplace transforms
(s + 2)I1 (s) + 6I2 (s) = 1
I1 (s) + (s − 3)I2 (s) = 0
Eliminating I2 (s)
[(s + 2)(s − 3) − 6]I1 (s) = s − 3
I1 (s) =
giving i1 (t) = L−1 {I1 (s)} =
1
6
s−3
= 7 + 7
(s − 4)(s + 3)
s−4 s+3
1 4t
(e + 6e−3t )
7
c Pearson Education Limited 2011
342
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Then from the first differential equation
6
di1
6
= − e4t + e−3t
dt
7
7
1 −3t
1
giving i2 (t) = (e
− e4t ), i1 (t) = (e4t + 6e−3t ).
7
7
6i2 = −2i1 −
12
The differential equation
di
d2 i
LCR 2 + L + Ri = V
dt
dt
follows using Kirchhoff’s second law.
Substituting V = E and L = 2R2 C gives
2R3 C2
which on substituting CR =
d2 i
2 di
+
2R
C + Ri = E
dt2
dt
1
leads to
2n
1 d2 i
1 di
E
+
+i=
2
2
2n dt
n dt
R
and it follows that
d2 i
E
di
+ 2n + 2n2 i = 2n2
2
dt
dt
R
Taking Laplace transforms
(s2 + 2ns + 2n2 )I(s) =
2n2 E 1
·
R s
2n2
E
2
2
R s(s + 2ns + 2n )
E1
s + 2n
=
−
2
2
R s (s + n) + n
I(s) =
so that
i(t) =
E
[1 − e−nt (cos nt + sin nt)]
R
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
13
343
The equations are readily deduced by applying Kirchhoff’s second law to the
left- and right-hand circuits.
Note that from the given initial conditions we deduce that i2 (0) = 0.
Taking Laplace transforms then gives
E
s
−RI1 (s) + (sL + 2R)I2 (s) = 0
(sL + 2R)I1 (s) − RI2 (s) =
Eliminating I2 (s)
E
(sL + 2R)
s
E
(sL + 3R)(sL + R)I1 (s) = (sL + 2R)
s
[(sL + 2R)2 − R2 ]I1 (s) =
I1 (s) =
s + 2R
E
L
L s(s + R
)(s
+
L
3R
L )
1
6
1
E 23
− 2R −
R s
s+ L
s + 3R
L
R
3R
E
1
4 − 3e− L t − e− L t
giving i1 (t) = L−1 {I1 (s)} =
6R
=
For large t , the exponential terms are approximately zero and
i1 (t)
2E
3R
From the first differential equation
Ri2 = 2Ri1 + L
di1
−E
dt
Ignoring the exponential terms we have that for large t
i2
4E E
1E
−
=
3R R
3R
c Pearson Education Limited 2011
344
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
14
Taking Laplace transforms
(s2 + 2)X1 (s) − X2 (s) =
s2
2
+4
−X1 (s) + (s2 + 2)X2 (s) = 0
Eliminating X1 (s)
[(s2 + 2)2 − 1]X2 (s) =
X2 (s) =
=
so x2 (t) = L−1 {X2 (s)} =
s2
2
+4
2
(s2 + 4)(s2 + 1)(s2 + 3)
2
3
s2
+4
+
1
3
s2
+1
−
s2
1
+3
√
1
1
1
sin 2t + sin t − √ sin 3t
3
3
3
Then from the second differential equation
√
√
√
2
2
1
d 2 x2
2
4
√
sin
2t
+
sin
t
−
sin
2t
−
sin
t
+
sin
=
3t
−
3
sin
3t
dt2
3
3
3
3
3
√
1
1
2
or x1 (t) = − sin 2t + sin t + √ sin 3t
3
3
3
x1 (t) = 2x2 +
15(a)
(i)
L−1
s2
(s + 1) + 3
s+4
= L−1
+ 2s + 10
(s + 1)2 + 32
= e−t (cos 3t + sin 3t)
(ii)
L−1
1
2
s−3
1
= L−1
+
−
2
2
(s − 1) (s − 2)
(s − 1) (s − 1)
s−2
= et + 2tet − e2t
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
15(b)
Taking Laplace transforms
(s2 + 2s + 1)Y(s) = 4s + 2 + 8 + L{3te−t }
3
(s + 1)2 Y(s) = 4s + 10 +
(s + 1)2
4s + 10
3
Y(s) =
+
2
(s + 1)
(s + 1)4
4
3
6
=
+
+
2
s + 1 (s + 1)
(s + 1)4
1
giving y(t) = L−1 {Y(s)} = 4e−t + 6te−t + t3 e−t
2
1 −t
that is, y(t) = e (8 + 12t + t3 )
2
16(a)
5
2
5
= ·
− 14s + 53
2 (s − 7)2 + 22
5
Therefore, f(t) = L−1 {F(s)} = e7t sin 2t
2
F(s) =
16(b)
s2
d2 θ
dθ
n2 i
2
+
n
, θ(0) = θ̇(0) = 0, i const.
+
2K
θ
=
dt2
dt
K
Taking Laplace transforms
(s2 + 2Ks + n2 )Φ(s) =
Therefore, Φ(s) =
n2 i
Ks
n2 i
Ks(s2 + 2Ks + n2 )
For the case of critical damping n = K giving
1
1
K12
Ki
K2
K
=
Ki
−
−
Φ(s) =
s(s + K)2
s
s + K (s + K)2
c Pearson Education Limited 2011
345
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
346
Thus,
θ(t) = L−1 {Φ(s)} =
i
[1 − e−Kt − Kte−Kt ]
K
17(a)
(i)
L{sin tH(t − α)} = L{sin[(t − α) + α]H(t − α)}
= L{[sin(t − α) cos α + cos(t − α) sin α]H(t − α)}
cos α + s sin α −αs
.e
=
s2 + 1
(ii)
L−1
17(b)
se−αs
(s + 1) − 1
= L−1 e−αs
2
s + 2s + 5
(s + 1)2 + 4
1
= L−1 eαs L[e−t (cos 2t − sin 2t)]
2
1
= e−(t−α) [cos 2(t − α) − sin 2(t − α)]H(t − α)
2
Taking Laplace transforms
−e−sπ
1
−
by (i) above in part (a)
s2 + 1
s2 + 1
1 + e−πs
= 2
s +1
−πs
1 s−2
1
s
1+e
=
−
+
(1 + e−πs )
Y(s) = 2
(s + 1)(s2 + 2s + 5)
10 s2 + 1 10 s2 + 2s + 5
(s2 + 2s + 5)Y(s) =
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
347
giving
1
[2 sin t − cos t + e−(t−π) [2 sin(t − π) − cos(t − π)]H(t − π)]
10
1
+ e−t (cos 2t − sin 2t) + e−(t−π) [cos 2(t − z)
2
1
− sin 2(t − π)]H(t − π)]
2
1 −t
1
e (cos 2t − sin 2t) + 2 sin t − cos t
=
10
2
1
+ [e−(t−π) (cos 2t − sin 2t) + cos t − 2 sin t]H(t − π)
2
y(t) = L−1 {Y(s)} =
18
By theorem 5.5
L{v(t)} = V(s) =
T
1
1 − e−sT
1
=
1 − e−sT
0
T /2
e
0
e−st v(t)dt
−st
T
dt −
e
−st
dt
T /2
1 −st T /2 1 −st T
1
− e
− − e
=
1 − e−sT
s
s
0
T /2
1
1
= ·
(e−sT − e−sT /2 − e−sT /2 + 1)
s 1 − e−sT
(1 − e−sT /2 )2
1 1 − e−sT /2
1
=
=
s (1 − e−sT /2 )(1 + e−sT /2 )
s 1 + e−sT /2
Equation for current flowing is
250i +
1
(q0 +
C
t
i(τ)dτ) = v(t),
q0 = 0
0
Taking Laplace transforms
250I(s) +
1 1 − e−sT /2
1 1
·
I(s)
=
V(s)
=
·
10−4 s
s 1 + e−sT /2
1 1 − e−sT /2
(s + 40)I(s) =
250 1 + e−sT /2
1 − e−sT /2
1
·
or I(s) =
250(s + 40) 1 + e−sT /2
c Pearson Education Limited 2011
348
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
3
1
(1 − e−sT /2 )(1 − e−sT /2 + e−sT − e− 2 sT + e−2sT . . .)
250(s + 40)
3
1
=
[1 − 2e−sT /2 + 2e−sT − 2e− 2 sT + 2e−2sT . . .]
250(s + 40)
1 −40t
1
=
e
Since L−1
using the second shift theorem gives
250(s + 40)
250
I(s) =
i(t) =
T −40(t−T /2)
1
e−40t − 2H t −
e
+ 2H(t − T)e−40(t−T )
250
2
3T −40(t−3T /2)
e
−2H t −
+ ...
2
If T = 10−3 s then the first few terms give a good representation of the steady state
1
of the circuit is large compared to the period T.
since the time constant
4
19
The impulse response h(t) is the solution of
d2 h 2dh
+ 2h = δ(t)
+
dt2
dt
subject to the initial conditions h(0) = ḣ(0) = 0. Taking Laplace transforms
(s2 + 2s + s)H(s) = L{δ(t)} = 1
1
H(s) =
(s + 1)2 + 1
that is, h(t) = L−1 {H(s)} = e−t sin t.
Using the convolution integral the step response xs (t) is given by
t
h(τ)u(t − τ)dτ
xs (t) =
0
with u(t) = 1H(t) ; that is,
t
xs (t) =
1.e−τ sin τdτ
0
1
= − [e−τ cos τ + e−τ sin τ]t0
2
1
that is, xs (t) = [1 − e−t (cos t + sin t)].
2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Solving
d 2 xs
2dxs
+ 2xs = 1 directly we have taking Laplace transforms
+
2
dt
dt
(s2 + 2s + 2)Xs (s) =
1
s
1
+ 2s + 2)
s+2
1 1 1
= · −
2
2 s 2 (s + 1) + 1
Xs (s) =
s(s2
giving as before
xs (t) =
1 1 −t
− e (cos t + sin t)
2 2
20
d4 y
= 12 + 12H(x − 4) − Rδ(x − 4)
dx4
y(0) = y (0) = 0, y(4) = 0, y (5) = y (5) = 0
EI
With y (0) = A, y (0) = B taking Laplace transforms
12 12 −4s
+ e
− Re−4s
s
s
A
B
12 1
12 1 −4s
R 1 −4s
· 5+
· 5e
· e
Y(s) = 3 + 4 +
−
s
s
EI s
EI s
EI s4
EIs4 Y(s) = EI(sA + B) +
giving
A
B
1 4
1
x +
(x − 4)4 H(x − 4)
y(x) = L−1 {Y(s)} = x2 + x3 +
2
6
2EI
2EI
R
−
(x − 4)3 H(x − 4)
6EI
c Pearson Education Limited 2011
349
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
350
or
EIy(x) =
1
1
1
1
R
A1 x2 + B1 x3 + x4 + (x − 4)4 H(x − 4) − (x − 4)3 H(x − 4)
2
6
2
2
6
32
B1 + 128 ⇒ 3A1 + 4B1 = −48
3
y (5) = 0 ⇒ 0 = A1 + 5B1 + 6(25) + 6 − R ⇒ A1 + 5B1 − R = −156
y(4) = 0 ⇒ 0 = 8A1 +
y (5) = 0 ⇒ 0 = B1 + 12(5) + 12 − R ⇒ B1 − R = −72
which solve to give A1 = 18, B1 = −25.5, R = 46.5
Thus,
⎧
⎨
1 4
x − 4.25x3 + 9x2 , 0 ≤ x ≤ 4
2
y(x) = 1
⎩ x4 − 4.25x3 + 9x2 + 1 (x − 4)4 − 7.75(x − 4)3 , 4 ≤ x ≤ 5
2
2
R0 = −EIy (0) = 25.5kN, M0 = EIy (0) = 18kN.m
Check :
R0 + R = 72kN , Total load = 12 × 4 + 24 = 72kN
Moment about x = 0 is
12 × 4 × 2 + 24 × 4.5 − 4R = 18 = M0
21(a)
f(t) = H(t − 1) − H(t − 2)
e−s
e−2s
and L{f(t)} = F(s) =
−
s
s
c Pearson Education Limited 2011
√
√
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
351
Taking Laplace transforms throughout the differential equation
1 −s
(e − e−2s )
s
1
(e−s − e−2s )
X(s) =
s(s + 1)
1
1 −s 1
1 −2s
−
e −
−
e
=
s s+1
s s+1
(s + 1)X(s) =
giving x(t) = L−1 {X(s)} = [1 − e−(t−1) ]H(t − 1) − [1 − e−(t−2) ]H(t − 2)
21(b) I(s) =
E
s[Ls + R/(1 + Cs)]
(i) By the initial value theorem
E
=0
s→∞ Ls + R/(1 + Cs)
lim i(t) = lim sI(s) = lim
t→0
s→∞
(ii) Since sI(s) has all its poles in the left half of the s-plane the conditions of
the final value theorem hold so
lim i(t) = lim sI(s) =
t→∞
22
s→0
E
R
We have that for a periodic function f(t) of period T
L{f(t)} =
1
1 − e−sT
T
e−sT f(t)dt
0
Thus, the Laplace transform of the half-rectified sine wave is
π
1
e−sT sin tdt
L{v(t)} =
−2πs
1−e
0
π
1
(j−s)t
= Im
e
dt
1 − e−2πs 0
π
e(j−s)t
1
= Im
1 − e−2πs j − s 0
(−e−πs − 1)(−j − s)
1 + e−πs
1
=
= Im
1 − e−2πs
(j − s)(−j − s)
(1 − e−2πs )(1 + s2 )
1
that is, L{v(t)} =
2
(1 + s )(1 − e−πs )
c Pearson Education Limited 2011
352
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Applying Kirchoff’s law to the circuit the current is determined by
di
+ i = v(t)
dt
which on taking Laplace transforms gives
(s + 1)I(s) =
1
(1 +
− e−πs )
1
s+1 1
1
−
·
I(s) =
1 − e−πs s + 1 s2 + 1 2
s + 1
1 1
− 2
1 + e−πs + e−2πs + . . .
=
2 s+1 s +1
Since L−1
s2 )(1
1 1
s + 1 1
− 2
= (sin t − cos t + e−t )H(t) = f(t)
2 s+1 s +1
2
we have by the second shift theorem that
i(t) = f(t) + f(t − π) + f(t − 2π) + . . . =
∞
f(t − nπ)
n=0
The graph may be plotted by computer and should take the form
1
1
, L{te−t } =
2
s
(s + 1)2
−t
taking f(t) = t and g(t) = te in the convolution theorem
23(a)
Since L{t} =
L−1 [F(s)G(s)] = f ∗ g(t)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
353
gives
L−1
1
1
=
·
2
2
s (s + 1)
t
f(t − τ)g(τ)dτ
0
t
=
(t − τ)τe−τ
0
= −(t − τ)τe−τ − (t − 2τ)e−τ + 2e−τ
1
1
= t − 2 + 2e−t + te−t .
i.e. L−1 2 ·
2
s (s + 2)
23(b)
y(t) = t + 2
t
0
y(u) cos(t − u)du
t
0
s
Taking f(t) = y(t), g(t) = cos t ⇒ F(s) = Y(s), G(s) = 2
giving on taking
s +1
transforms
s
1
+ 2Y(s) 2
2
s
s +1
2
s +1
(s2 + 1 − 2s)Y(s) =
s2
2
1
2
s +1
2
2
+
+
or Y(s) = 2
=
−
s (s − 1)2
s s2
s − 1 (s − 1)2
Y(s) =
and y(t) = L−1 {Y(s)} = 2 + t − 2et + 2tet .
Taking transforms
(s2 Y(s) − sy(0) − y (0))(sY(s) − y(0)) = Y(s)
or (s2 Y(s) − y1 )(sY(s)) = Y(s)
1
y1
giving Y(s) = 0 or Y(s) = 2 + 3
s
s
which on inversion gives
y(t) = 0 or y(t) =
1 2
t + ty1
2
In the second of these solutions the condition on y (0) is arbitrary.
c Pearson Education Limited 2011
354
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
24
Equation for displacement is
EI
d4 y
= −Wδ(x − )
dx4
with y(0) = 0, y(3) = 0, y (0) = y (3) = 0
with y (0) = A, y (0) = B then taking Laplace transforms gives
EIs4 Y(s) = EI(sA + B) − We−s
−W −s A
B
Y(s) =
e
+ 3+ 4
4
EIs
s
s
−W
A
B
giving y(x) =
(x − )3 .H(x − ) + x2 + x3
6EI
2
6
−3W
B
(x − )2 + Ax + x2
6EI
2
so y (3) = 0 and y(3) = 0 gives
For x > , y (x) =
2
2W2
+ 3A + 9B
EI
2
3
9
4W
9
+ A2 + B3
0=−
3EI
2
2
20 W
4W
and B =
giving A = −
9EI
27 EI
0=−
Thus, deflection y(x) is
y(x) = −
2 W 2 10 W 3
W
(x − )3 H(x − ) −
x +
x
6EI
9 EI
81 EI
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
355
With the added uniform load the differential equation governing the deflection is
EI
25(a)
d4 y
= −Wδ(x − ) − w[H(x) − H(x − )]
dx4
Taking Laplace transforms
(s2 − 3s + 3)X(s) =
1 −as
e
s
1
1
−as
16
1
−as
6s − 2
·
e
−
·e
=
s(s2 − 3s + 3)
s
s2 − 3s + 3
√ √
1 1 (s − 32 ) − 3( 23 ) −as
√
−
e
=
6 s
(s − 3 )2 + ( 3 )2
X(s) =
2
2
√
√
√
3
3
3
e−as
L 1 − e− 2 t cos
t − 3 sin
t
=
6
2
2
giving
−1
x(t) = L
√
√
√
3
3
3
1
− 2 (t−a)
1−e
(t − a) − 3 sin
(t − a) H(t − a)
cos
{X(s)} =
6
2
2
25(b)
X(s) = G(s)L{sin wt} = G(s)
=
s2
w
+ w2
w
G(s)
(s + jw)(s − jw)
Since the system is stable all the poles of G(s) have negative real part. Expanding
in partial fractions and inverting gives
x(t) = 2Re
F(jw)w jwt
+ terms from G(s) with negative exponentials
·e
2jw
Thus, as t → ∞ the added terms tend to zero and x(t) → xs (t) with
xs (t) = Re
ejwt F(jw)
j
c Pearson Education Limited 2011
356
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
26(a)
In the absence of feedback the system has poles at
s = −3 and s = 1
and is therefore unstable.
26(b) G1 (s) =
26(c)
1
1
G(s)
=
= 2
1 + KG(s)
(s − 1)(s + 3) + K
s + 2s + (K − 3)
Poles G1 (s) given by s = −1 ±
√
4 − K.
These may be plotted in the s-plane for different values of K. Plot should be as
in the figure
26(d)
Clearly from the plot in (c) all the poles are in the left half plane when
K > 3. Thus system stable for K > 3.
a1
a0
a2
1s2 +
2s + (K − 3) = 0
Routh–Hurwitz determinants are
26(e)
Δ1 = 2 > 0
2
a1 a2
=
Δ2 =
0
0 a0
1
= 2(K − 3) > 0 if K > 3
K − 3
thus, confirming the result in (d).
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
27(a)
Closed loop transfer function is
G1 (s) =
2
G(s)
= 2
1 + G(s)
s + αs + 5
2
= h(t) = 2e−2t sin t
Thus L−1 2
s + αs + 5
α
2
−2t
−1
i.e. L
sin (5 −
= 2e
2
(s + α2 )2 + (5 − α4 )
giving α = 4
27(b)
357
α2
4 )t
= 2e−2t sin t
Closed loop transfer function is
G(s) =
1
10
s(s−1)
− (1+Ks)10
s(s−1)
=
s2
10
+ (10K − 1)s + 10
Poles of the system are given by
s2 + (10K − 1)s + 10 = 0
which are both in the negative half plane of the s-plane provided (10K − 1) > 0;
that is, K >
is K =
1
10
1
10
. Thus the critical value of K for stability of the closed loop system
.
28(a)
c Pearson Education Limited 2011
358
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
28(b)
L{e
At
1
} = [sI − A] =
(s + 2)(s + 3)
6
3
2
6
− s+3 s+2
− s+3
s+2
=
1
−2
1
1
s+3 − s+2
s+2 − s+3
−1
s+5
−1
6
8
Taking inverse transforms gives
e
28(c)
At
3e−2t − 2e−3t
e−3t − e−2t
=
6e−2t − 6e−3t
3e−3t − 2e−2t
Taking Laplace transforms
[sI − A]X(s) = x(0) + bU(s) ; Y(s) = cT X(s)
With x(0) = 0 and U(s) = 1 the transform Xδ (s) of the impulse response is
Xδ (s) = [sI − A]−1 b , Yδ (s) = cT [sI − A]−1 b
Inverting then gives the impulse response as
yδ (t) = [1 1]
6e−2t − 6e−3t
3e−3t − 2e−2t
With x(0) = [1 0]T and X(s) =
Y(s) = [1 1]
= [1 1]
so y(t) = [1 1]
= 4e−2t − 3e−3t , t ≥ 0
1
s
1
0 1
−1
+ [sI − A]
[sI − A]
0
1 s
1/6
1/2
1/3
3
2
s+2 − s+3 + 6 s − s+2 + s+3
1
1
1
1
s+3 − s+2 + s+2 − s+3
3e−2t − 2e−3t + 1 − 3e−2t + 2e−3t
e−3t − e−2t + e−2t − e−3t
−1
that is, y(t) = 1, t ≥ 0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
29
L{e
At
−1
} = [sI − A]
s+2
−2
=
1
s
−1
−1
(s+1)2 +1
s+2
(s+1)2 +1
s
(s+1)2 +1
2
(s+1)2 +1
=
−e−t sin t
e−t (cos t − sin t)
Thus e =
−t
−t
2e sin t
e (cos t + sin t)
At
and e x(0) = 0 since x(0) = 0
1
With U(s) = L{u(t)} = we have
s
359
At
−1
[sI − A]
bU(s) =
=
−1
so L
{(sI − A)
−1
Thus
bU(s)} =
s
s2 +2s+2
2
s2 +2s+2
−1
s2 +2s+2
s+2
s2 +2s+2
1
(s+1)2 +1
1
s+2
s − (s+1)2 +1
1
0
1
=
s
1
s2 +2s+2
2
s(s2 +2s+2)
e−t sin t
−t
1 − e (cos t + sin t)
x(t) = eAt x(0) + L−1 {(sI − A)−1 bU(s)}
e−t sin t
=
1 − e−t (cos t + sin t)
For the transfer function, we have,
Y(s) = c X(s) and when x(0) = 0
Y(s) = c[s I − A]−1 bU(s) = HU(s)
where H = c[s I − A]−1 b
For this system, we have H(s) = [1 1]
s
s2 +2s+1
2
s2 +2s+1
=
s+2
(s + 1)2 + 1
When u(t) = δ(t), U(s) = 1 and so the impulse response yδ (t) is given by
L{yδ (t)} = Yδ (s) =
s+1
1
s+2
=
+
2
2
(s + 1) + 1
(s + 1) + 1 (s + 1)2 + 1
⇒ yδ (t) = e−t (cos t + sin t)
c Pearson Education Limited 2011
360
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
30
The controllability question can be answered by either reducing to canonical
form as in section 6.7.8 of the text or by using the Kalman matrix criterion given
in Exercise 61 of the text. Adopting the Kalman matrix approach
⎤
⎡
1
2
0
A = ⎣ 0 −1 0 ⎦ and
−3 −3 −2
⎤
⎡
⎡ ⎤
⎡ ⎤
2
0
0
2
⎦
⎦
⎣
⎣
⎣
b = 1 , A b = −1 , A b = 1 ⎦
−3
9
0
so the controllability Kalman matrix is
⎡
0
[b A b A2 b] = ⎣ 1
0
⎤
2 0
−1 1 ⎦ = C
−3 9
Since det C = 0, rank C = 3 so the system is controllable.
The eigenvalues of A are given by
1 − λ
2
0
0
−1 − λ
0 = (1 − λ)
−3
−3
−2 − λ
−(1 + λ)
0
−3
−(2 + λ)
= (1 − λ)(1 − λ)(2 + λ) = 0
so that the eigenvalues are λ1 = −2, λ2 = −1, λ3 = 1. The system is therefore
unstable with λ3 = 1 corresponding to the unstable mode. The corresponding
eigenvectors of A are given by
(A − λi I)ei = 0
and are readily determined as
e1 = [0 0 1]T
e2 = [1 − 1 0]T
e3 = [1 0 − 1]T
To determine the control law to relocate λ3 = 1 at −5 we need to determine the
eigenvector v3 of AT corresponding to λ3 = 1. This is readily obtained as
v3 = [1 1 0]T
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, the required control law is
u(t) = KvT3 x(t) = K[1 1 0]T x(t)
6
(−5) − 1
p3 − λ3
⎡ ⎤ = = −6
=
where K =
T
1
v3 b
0
[1 1 0] ⎣ 1 ⎦
0
So u(t) = −6(x1 (t) + x2 (t))
31(a)
Let x [ x1
T
x2 ] then state − space model is
−2
=
0
x
y = cT x ⇒ y = [ 1 0 ] 1
x2
ẋ1
ẋ = Ax + bu ⇒
ẋ2
(b) G(s) =
Y (s)
U (s)
4
1
x1
1
u
+
1
x2
= cT (sI − A)−1 b
s + 2 −4
= (s + 2)(s − 1)
det(sI − A) = Δ =
0
s − 1
s−1
4
adj(sI − A) =
0
s+2
s+3
1
1
s−1
4
=
G(s) =
[1 0]
1
0
s+2
Δ
(s + 2)(s − 1)
System has positive pole s = 1 and is therefore is unstable.
(c) u(t) = r(t) − ky(t) ⇒ ẋ = Ax + b(r − kcT x) = (A − kbcT )x + br
−2 − k
1
4
[1 0] =
−k
−k
1
1
λ + 2 + k
4
=0⇒
⇒ Eigenvalues given by
−k
λ − 1
−2
(A − kbc ) =
0
4
1
T
λ2 + (k + 1)λ + 3k − 2 = 0
so system is stable if and only if 3k − 2 > 0 or k >
c Pearson Education Limited 2011
2
3
361
362
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
(d) r(t) = H(t) ⇒ R(s) =
Since k >
2
3
1
s
⇒ Y(s) = cT (sI − A + kbcT )−1 b 1s
, system stable, the final value theorem gives
lim y(t) = lim sY(s) = lim [cT (sI − A + kbcT )−1 b] = −cT (A − kbcT )−1 b
s→0
s→0
−1
1
1
−4
−2 − k 4
1
= −[1 0]
= −[1 0]
−k
1
1
3k − 2 k −2 − k
3
=
3k − 2
t→∞
Thus, lim y(t) = 1 if and only if
t→∞
32(a)
=1⇒k=
5
3
Overall closed loop transfer function is
G(s) =
32(b)
3
3k−2
1+
K
s(s+1)
K
s(s+1) (1 +
K1 s)
=
K
s2 + s(1 + KK1 ) + K
Assuming zero initial conditions step response x(t) is given by
X(s) = G(s)L{1.H(t)} =
K
s[s2 + s(1 + KK1 ) + K]
wn
+ 2ξwn s + w2n ]
1
s + 2ξwn
= − 2
s s + 2ξwn s + w2n
(s + ξwn ) + ξwn
1
= −
s
(s + ξwn )2 + [w2n (1 − ξ2 )]
(s + ξwn ) + ξwn
1
= −
s
(s + ξwn )2 + w2d
ξ
sin wd t , t ≥ 0.
giving x(t) = L−1 {X(s)} = 1 − e−ξwn t cos wd t +
1 − ξ2
=
32(c)
s[s2
The peak time tp is given by the solution of
dx
= e−ξwn t ξwn −
dt
= e−ξwn t
wn
1 − ξ2
ξwd
1 − ξ2
dx
=0
dt t=tp
cos wd t
ξ2 wn
1 − ξ2
sin wd t
c Pearson Education Limited 2011
+ wd sin wd t
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, tp given by the solution of
e−ξwn tp
wn
1 − ξ2
sin wd tp = 0
i.e. sin wd tp = 0
Since the peak time corresponds to the first peak overshoot
wd tp = π or tp =
π
wd
The maximum overshoot Mp occurs at the peak time tp . Thus
Mp = x(tp ) − 1 = e
=
ξw π
− wn
d
cos π
ξwn π
−
e wd
=e
ξ
+
−ξπ/
1 − ξ2
sin π
√
1−ξ 2
π
We wish Mp to be 0.2 and tp to be 1s, thus
e−ξπ/
√
1−ξ 2
= 0.2 giving ξ = 0.456
and
π
= 1 giving wd = 3.14
wd
wd
= 3.53 from which we deduce that
1 − ξ2
tp =
Then it follows that wn =
K = w2n = 12.5
and K1 =
32(d)
2wn ξ − 1
= 0.178.
K
The rise time tr is given by the solution of
x(tr ) = 1 = 1 − e−ξwn tr cos wd tr +
ξ
1 − ξ2
Since e−ξwn tr = 0
cos wd tr +
ξ
1 − ξ2
sin wd tr = 0
c Pearson Education Limited 2011
sin wd tr
363
364
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
giving tan wd tr = −
or
1 − ξ2
ξ
tr =
1
tan−1 −
wd
1 − ξ2
π − 1.10
=
= 0.65s.
ξ
wd
The response x(t) in (b) may be written as
x(t) = 1 −
so the curves 1 ±
e−ξwn t
1 − ξ2
e−ξwn t
1 − ξ2
sin wα t + tan−1
1 − ξ2
ξ
are the envelope curves of the transient response to
1
. The settling time ts may
ξwn
be measured in terms of T. Using the 2% criterion ts is approximately 4 times
a unit step input and have a time constant T =
the time constant and for the 5% criterion it is approximately 3 times the time
constant. Thus,
4
= 2.48s
ξwn
3
= 1.86s
5% criterion : ts = 3T =
ξwn
2% criterion : ts = 4T =
Footnote :
This is intended to be an extended exercise with students being
encouraged to carry out simulation studies in order to develop a better
understanding of how the transient response characteristics can be used in system
design.
33
As for Exercise 32 this is intended to be an extended problem supported by
simulation studies. The following is simply an outline of a possible solution.
Figure 5.67(a) is simply a mass-spring damper system represented by the
differential equation
d2 x
dx
+ K1 x = sin wt
M1 2 + B
dt
dt
Assuming that it is initially in a quiescent state taking Laplace transforms
X(s) =
M1
s2
w
1
· 2
+ Bs + K1 s + w2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
365
The steady state response will be due to the forcing term and determined by the
αs + β
term in the partial fractions expansion of X(s) . Thus, the steady state
s2 + w2
response will be of the form A sin(wt + δ) ; that is, a sinusoid having the same
frequency as the forcing term but with a phase shift δ and amplitude scaling A.
In the situation of Figure 5.67(b) the equations of motion are
d2 x
dx
+ K2 (y − x) + sin wt
= −K1 x − B
2
dt
dt
d2 y
M2 2 = −K2 (y − x)
dt
M1
Assuming an initial quiescent state taking Laplace transforms gives
[M1 s2 + Bs + (K1 + K2 )]X(s) − K2 Y(s) = w/(s2 + w2 )
−K2 X(s) + (s2 M2 + K2 )Y(s) = 0
Eliminating Y(s) gives
X(s) =
w(s2 M2 + K2 )
(s2 + w2 )p(s)
where p(s) = (M1 s2 + Bs + K1 + K2 )(s2 M2 + K2 ) .
Because of the term (s2 + w2 ) in the denominator x(t) will contain terms in
sin wt and cos wt . However, if (s2 M2 + K2 ) exactly cancels (s2 + w2 ) this will
be avoided. Thus choose K2 = M2 w2 . This does make practical sense for if the
natural frequency of the secondary system is equal to the frequency of the applied
force then it may resonate and therefore damp out the steady state vibration of
M1 .
It is also required to show that the polynomial p(s) does not give rise to any
undamped oscillations. That is, it is necessary to show that p(s) does not possess
purely imaginary roots of the form jθ, θ real, and that it has no roots with a positive
real part. This can be checked using the Routh–Hurwitz criterion.
To examine the motion of the secondary mass M2 solve for Y(s) giving
Y(s) =
(s2
K2 w
+ w2 )p(s)
Clearly due to the term (s2 + w2 ) in the denominator the mass M2 possesses an
undamped oscillation. Thus, in some sense the secondary system has absorbed the
energy produced by the applied sinusoidal force sin wt .
c Pearson Education Limited 2011
366
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
34
Again this is intended to be an extended problem requiring wider exploration
by the students. The following is an outline of the solution.
34(a)
Students should be encouraged to plot the Bode plots using the steps
used in Example 5.65 of the text and using a software package. Sketches of the
magnitude and phase Bode plots are given in the figures below.
34(b)
With unity feedback the amplifier is unstable. Since the −180◦ crossover
gain is greater than 0dB (from the plot it is +92dB) .
34(c)
Due to the assumption that the amplifier is ideal it follows that for
1
must be 92dB (that is, the plot is effectively
marginal stability the value of
β
lowered by 92dB) . Thus
20 log
34(d)
1
= 92
β
92
1
⇒β
= antilog
β
20
2.5 × 10−5
From the amplitude plot the effective 0dB axis is now drawn through
the 100dB point. Comparing this to the line drawn through the 92dB point,
corresponding to marginal stability, it follows that
Gain margin = −8dB
and Phase margin = 24◦ .
34(e)
G(s) =
K
(1 + sτ1 )(1 − sτ2 )(1 + sτ3 )
Given low frequency gain K = 120dB so
20 log K = 120 ⇒ K = 106
Ti =
1
where fi is the oscillating frequency in cycles per second of the pole.
fi
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
367
Since 1MHz = 10 cycles per second
1
1
= 6 since f1 = 1MHz
f1
10
1
1
=
since f2 = 10MHz
τ2 =
f2
10.106
1
1
=
since f3 = 25MHz
τ3 =
f3
25.106
τ1 =
Thus,
G(s) =
=
106
(1 +
s
106 )(1
(s +
106 )(s
+
s
10.106 )(1
24
+
s
25.106 )
250.10
+ 107 )(s + 52 .107 )
The closed loop transfer function G1 (s) is
G(s) =
34(f)
G(s)
1 + βG(s)
The characteristic equation for the closed loop system is
(s + 106 )(s + 107 )(s + 52 .107 ) + β25.1025 = 0
or
s3 + 36(106 )s2 + (285)1012 s + 1019 (25 + 25β106 ) = 0
↓
↓
↓
A1
A2
A3
By Routh–Hurwitz criterion system stable provided A1 > 0 and A1 A2 > A3 . If
β = 1 then A1 A2 < A3 and the system is unstable as determined in (b). For
marginal stability A1 A2 = A3 giving β = 1.40−5 (compared with β = 2.5.10−5
using the Bode plot).
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
368
Magnitude vs Frequency Plot
120
Data margin
– 8 dB
100
80
60
Gain dB
1/β = 92 dB
40
20
0
Corresponds to 180° phase lag
To phase plot
1
10
25
Log freq. MHz
Phase vs Frequency Plot
0°
– 90°
– 180°
Phase margin 24°
16 MHz
– 270°
1
10
c Pearson Education Limited 2011
25
ln freq. MHz
6
The
Transform
Z
Exercises 6.2.3
1(a)
F(z) =
∞
(1/4)k
k=0
1(b)
F(z) =
zk
∞
3k
k=0
1(c)
F(z) =
∞
(−2)k
k=0
1(d)
F(z) =
zk
zk
∞
−(2)k
k=0
zk
=
4z
1
=
1 − 1/4z
4z − 1
=
z
1
=
1 − 3/z
z−3
=
z
1
=
1 − (−2)/z
z+2
if | z |> 2
z
1
=−
1 − 2/z
z−2
if | z |> 2
=−
if | z |> 1/4
if | z |> 3
1(e)
Z{k} =
z
(z − 1)2
if | z |> 1
from (6.6) whence
Z{3k} = 3
2
z
(z − 1)2
if | z |> 1
k
uk = e−2ωkT = e−2ωT
whence
U(Z) =
z
z − e−2ωT
c Pearson Education Limited 2011
370
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 6.3.6
3
Z{sin kωT} =
=
4
z
z
1
1
−
jωT
2j z − e
2j z − e−jωT
z2
z sin ωT
− 2z cos ωT + 1
k
2z
1
}=
Z{
2
2z − 1
so
2
1
2z
= 2
×
3
z
2z − 1
z (2z − 1)
Z{yk } =
Proceeding directly
Z{yk } =
∞
xk−3
zk
k=3
5(a)
=
1
Z −
5
=
∞
xr
2
1
}
{x
=
=
×
Z
k
zr+3
z3
z2 (2z − 1)
r=0
r
∞
−1
r=0
5z
=
5z
5z + 1
| z |>
5(b)
{cos kπ} = (−1)k
so
Z {cos kπ} =
6
k
1
Z
2
z
z+1
=
By (3.5)
Z (ak ) =
so
Z (kak−1 ) =
| z |> 1
2z
2z − 1
z
z−a
z
(z − a)2
c Pearson Education Limited 2011
1
5
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
thus
Z (kak ) =
whence
k
1
Z k
2
az
(z − a)2
=
2z
(2z − 1)2
7(a)
sinh kα =
so
1
Z {sinh kα} =
2
1 α k 1 −α k
(e ) − (e )
2
2
z
z sinh α
z
= 2
−
α
−α
z−e
z−e
z − 2z cosh α + 1
7(b)
cosh kα =
1 α k 1 −α k
(e ) + (e )
2
2
then proceed as above.
8(a)
k
uk = e−4kT = e−4T ;
8(b)
Z {uk } =
z
z − e−4T
1 j kT
e
− e−j kT
2j
z
z sin T
z
= 2
−
j
T
−j
T
z−e
z−e
z − 2z cos T + 1
uk =
Z {uk } =
1
2j
8(c)
uk =
1 j 2kT
e
+ e−j 2kT
2
then proceed as in 8(b) to give
Z{uk } =
9
z2
z(z − cos 2T)
− 2z cos 2T + 1
Initial value theorem: obvious from definition.
c Pearson Education Limited 2011
371
372
9
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Final value theorem
(1 − z−1 )X(z) =
∞
xr − xr−1
zr
r=0
= x0 +
As z → 1 and if
lim
r→∞
x2 − x1
x1 − x0
xr − xr−1
+
+
.
.
.
+
+ ...
z
z2
zr
xr exists, then
lim (1 − z−1 )X(z) = lim xr
r→∞
z→1
10
Multiplication property (6.19): Let Z {xk } =
∞
ak xk
Z a xk =
k
k=0
10
zk
∞ xk
k=0 k = X(z) then
z
= X(z/a)
Multiplication property (6.20)
−z
∞
∞
k=0
k=0
kxk
d
d xk
=
= Z {kxk }
X(z) = −z
dz
dz
zk
zk
The general result follows by induction.
Exercises 6.4.2
11(a)
11(b)
11(c)
z
;
z−1
from tables uk = 1
z
z
=
;
z+1
z − (−1)
z
;
z − 1/2
from tables uk = (−1)k
from tables uk = (1/2)k
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
11(d)
1
z
1
z
=
←→ (−1/3)k
3z + 1
3 z + 1/3
3
11(e)
z
;
z−j
11(f)
from tables uk = (j )k
√
z
z
√ =
√ ←→ (−j 2)k
z+j 2
z − (−j 2)
11(g)
1
1 z
=
←→
z−1
zz−1
0;
1;
k=0
k>0
using first shift property.
11(h)
z+2
1 z
1;
=1+
←→
k−1
(−1)
;
z+1
zz+1
1;
k=0
=
k+1
(−1)
; k>0
k=0
k>0
12(a)
Y(z)/z =
so
Y(z) =
12(b)
1
Y(z) =
7
1 1
1 1
−
3z−1 3z+2
1 z
1
1 z
−
←→
1 − (−2)k
3z−1 3z+2
3
z
z
−
z − 3 z + 1/2
←→
1 k
(3) − (−1/2)k
7
12(c)
Y(z) =
1 z
1
z
1 1
+
←→ + (−1/2)k
3 z − 1 6 z + 1/2
3 6
c Pearson Education Limited 2011
373
374
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
12(d)
2
z
2 z
2
2
−
←→ (1/2)k − (−1)k
3 z − 1/2 3 z + 1
3
3
Y(z) =
=
12(e)
z
z
−
z−j
z − (−j )
z
z
−
z − ej π/2
z − e−j π/2
1
(ej π/2 )k − (e−j π/2 )k = sin kπ/2
←→
2j
1
Y(z) =
2j
1
=
2j
12(f)
2
2
(1/2)k + (−1)k+1
3
3
z
√
√
Y(z) =
z − ( 3 + j) z − ( 3 − j)
z
z
1
√
√
−
=
2j z − ( 3 + j ) z − ( 3 − j )
z
z
1
−
=
2j z − 2ej π/6
z − 2e−j π/6
1
2k ej kπ/6 − 2k e−j kπ/6 = 2k sin kπ/6
←→
2j
12(g)
Y(z) =
z
1 z
5
1 z
−
+
2
2 (z − 1)
4z−1 4z−3
←→
5
1
1 − 3k
k+
2
4
12(h)
Y(z)/z =
so
Y(z) =
z
(z −
1)2 (z2
− z + 1)
=
1
1
− 2
2
(z − 1)
z −z+1
⎛
⎞
z
1 ⎜
⎜
√
−
(z − 1)2
3j ⎝
⎟
z
z
√
√ ⎟
−
1 + 3j
1 − 3j ⎠
z−
z−
2
2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
=
1
z
−√
2
(z − 1)
3j
375
z
z
−
z − ej π/3
z − e−j π/3
2
2
←→ k − √ sin kπ/3 = k + √ cos(kπ/3 − 3π/2)
3
3
13(a)
Y(z) =
∞
xk
k=0
zk
=
2
1
+ 7
z z
whence x0 = 0, x1 = 1, x2 = x3 = . . . = x6 = 0, x7 = 2 and xk = 0, k > 7;
giving Y(z) ↔ {0, 1, 0, 0, 0, 0, 0, 2, · · ·}
13(b)
Proceed as in Exercise 13(a) to give
Y(z) ↔ {1, 0, 3, 0, 0, 0, 0, 0, 0, −2, · · ·}
13(c) Observe that
3z + z2 + 5z5
1
3
=5+ 3 + 4
5
z
z
z
and proceed as in Exercise 13(a) to give Y(z) ↔ {5, 0, 0, 1, 3, · · ·}
13(d)
Y(z) =
1
1
z
+
+
z2
z3
z + 1/3
←→ {0, 0, 1, 1} + {(−1/3)k }
13(e)
1
1/2
3
+ 2−
z z
z + 1/2
1 0, k = 0
←→ {1, 3, 1} −
2 (−1/2)k , k ≥ 1
⎧
⎧
1, k = 0
1, k = 0
⎪
⎪
⎪
⎪
⎪
⎪
⎨ 5/2, k = 1
⎨ 5/2, k = 1
= 5/4, k = 2
= 5/4, k = 2
⎪
⎪
⎪
⎪
⎪
⎪
⎩ − 1 (−1/2)k−3 , k ≥ 3
⎩ − 1 (−1/2)k−1 , k ≥ 3
2
8
Y(z) = 1 +
c Pearson Education Limited 2011
376
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
13(f)
Y(z) =
2
1
1
−
+
z − 1 (z − 1)2
z−2
0, k = 0
1 − 2(k − 1) + 2k−1 , k ≥ 1
0, k = 0
=
3 − 2k + 2k−1 , k ≥ 1
←→
13(g)
Y(z) =
←→
1
2
−
z−1 z−2
0, k = 0
2 − 2k−1 , k ≥ 1
Exercises 6.5.3
14(a)
If the signal going into the left D-block is wk and that going into the right
D-block is vk , we have
yk+1 = vk ,
1
vk+1 = wk = xk − vk
2
so
1
yk+2 = vk+1 = xk − vk
2
1
1
= xk − vk = xk − yk+1
2
2
that is,
1
yk+2 + yk+1 = xk
2
14(b)
Using the same notation
yk+1 = vk ,
1
1
vk+1 = wk = xk − vk − yk
4
5
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Then
1
1
yk+2 = xk − yk+1 − yk
4
5
or
1
1
yk+2 + yk+1 + yk = xk
4
5
15(a)
z2 Y(z) − z2 y0 − zy1 − 2(zY(z) − zy0 ) + Y(z) = 0
with y0 = 0, y1 = 1
Y(z) =
z
(z − 1)2
so yk = k, k ≥ 0.
15(b)
Transforming and substituting for y0 and y1 ,
Y(z)/z =
2z − 15
(z − 9)(z + 1)
so
Y(z) =
3 z
17 z
−
10 z − 9 10 z + 1
thus
yk =
3 k 17
9 − (−1)k , k ≥ 0
10
10
15(c) Transforming and substituting for y0 and y1 ,
Y(z) =
1
=
4j
thus
yk =
z
(z − 2j )(z + 2j )
z
z
−
z − 2ej π/2
z − 2e−j π/2
1 k j kπ/2
2 e
− e−j kπ/2 = 2k−1 sin kπ/2, k ≥ 0
4j
c Pearson Education Limited 2011
377
378
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
15(d)
Transforming, substituting for y0 and y1 , and rearranging
Y(z)/z =
so
Y(z) = 2
6z − 11
(2z + 1)(z − 3)
z
z
+
z + 1/2 z − 3
thus
yk = 2(−1/2)k + 3k , k ≥ 0
16(a)
6yk+2 + yk+1 − yk = 3,
y0 = y 1 = 0
Transforming with y0 = y1 = 0,
(6z2 + z − 1)Y(z) =
so
Y(z)/z =
and
Y(z) =
3z
z−1
3
(z − 1)(3z − 1)(2z + 1)
9
z
2
z
1 z
−
+
2 z − 1 10 z − 1/3 5 z + 1/2
Inverting
yk =
16(b)
9
1
2
− (1/3)k + (−1/2)k
2 10
5
Transforming with y0 = 0, y1 = 1,
(z2 − 5z + 6)Y(z) = z + 5
whence
Y(z) =
z
z−1
7 z
z
5 z
+
−6
2z−1 2z−3
z−2
so
yk =
5 7 k
+ (3) − 6 (2)k
2 2
16(c) Transforming with y0 = y1 = 0,
(z2 − 5z + 6)Y(z) =
z
z − 1/2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so
Y(z) =
z
2 z
2 z
4
−
+
15 z − 1/2 3 z − 2 5 z − 3
whence
yn =
16(d)
4
2
2
(1/2)k − (2)k + (3)k
15
3
5
Transforming with y0 = 1, y1 = 0,
(z2 − 3z + 3)Y(z) = z2 − 3z +
so
so
z
z−1
z
z
− 2
Y(z) =
z − 1 z − 3z + 3
⎧
⎫
⎪
⎪
⎪
⎪
⎬
z
1 ⎨
z
z
√ −
√
−√
=
z−1
3j ⎪
3 − 3j ⎪
⎪
⎪
⎩ z − 3 + 3j
⎭
z−
2
2
z
z
z
1
√
√
=
−
−√
z−1
3j z − 3ejπ/6
z − 3e−jπ/6
√
2 √ ejnπ/6 − e−jnπ/6
= 1 − 2( 3)n−1 sin nπ/6
yn = 1 − √ ( 3)k
2j
3
16(e) Transforming with y0 = 1, y1 = 2,
(2z2 − 3z − 2)Y(z) = 2z2 + z + 6
so
z
+z
Y(z) =
z−2
=
z
z
+
2
(z − 1)
z−1
z+5
(z − 1)2 (2z + 1)(z − 2)
12 z
2
z
z
z
−
−
−2
5 z − 2 5 z + 1/2 z − 1
(z − 1)2
so
yn =
12 n 2
(2) − (−1/2)n − 1 − 2n
5
5
16(f) Transforming with y0 = y1 = 0,
(z2 − 4)Y(z) = 3
z
z
−5
2
(z − 1)
z−1
c Pearson Education Limited 2011
379
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
380
so
z
1 z
1 z
z
−
−
−
2
z − 1 (z − 1)
2z−2 2z+2
Y(z) =
and
1
1
yn = 1 − n − (2)n − (−2)n
2
2
17(a)
Write the transformed equations in the form
z − 3/2
1
c(z)
zC0
=
zE0
−0.21
z − 1/2
e(z)
Then
c(z)
e(z)
1
= 2
z − 2z + 0.96
Solve for c(z) as
c(z) = 1200
z − 1/2 −1
0.21 z − 3/2
zC0
zE0
z
z
+ 4800
z − 1.2
z − 0.8
and
Ck = 1200(1.2)k + 4800(0.8)k
This shows the 20% growth in Ck in the long term as required.
(b) Then
Ek = 1.5Ck − Ck+1
= 1800(1.2)k + 7200(0.8)k − 1200(1.2)k+1 − 4800(0.8)k+1
Differentiate wrt k and set to zero giving
0.6 log(1.2) + 5.6x log(0.8) = 0 where x = (0.8/1.2)
k
Solving, x = 0.0875 and so
k=
log 0.0875
= 6.007
log(0.8/1.2)
The nearest integer is k = 6, corresponding to the seventh year in view of the
labelling, and C6 = 4841 approximately.
18
Transforming and rearranging
Y(z)/z =
1
z−4
+
(z − 2)(z − 3) (z − 1)(z − 2)(z − 3)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so
Y(z) =
z
1 z
1 z
+
−
2z−1 z−2 2z−3
thus
yk =
19
1
1
+ 2k − 3k
2
2
Ik = Ck + Pk + Gk
= aIk−1 + b(Ck − Ck−1 ) + Gk
= aIk−1 + ba(Ik−1 − Ik−2 ) + Gk
so
Ik+2 − a(1 + b)Ik+1 + abIk = Gk+2
Thus substituting
1
Ik+2 − Ik+1 + Ik = G
2
Using lower case for the z transform, we obtain
1
z
(z2 − z + )i(z) = (2z2 + z)G + G
2
z−1
⎡
whence
⎢
i(z)/z = G ⎣
⎡
⎤
1
1
z2 − z +
2
+
2 ⎥
⎦
z−1
⎤
⎢ 2
+
= G⎣
z−1
so
Thus
1
⎥
1+j
1−j ⎦
(z −
)(z −
)
2
2
⎫⎤
⎧
⎡
⎪
⎪
⎪
⎪
⎬⎥
⎨
⎢ z
2
z
z
⎥
⎢
+
−
i(z) = G ⎣2
⎦
1
1
⎪
z − 1 2j ⎪
⎪
z − √ e−j π/4 ⎪
⎭
⎩ z − √ ej π/4
2
2
#
&
%
2 1 k $ j kπ/4
(√ ) e
− e−j kπ/4
Ik = G 2 +
2j
2
(
'
k
1
sin kπ/4
= 2G 1 + √
2
c Pearson Education Limited 2011
381
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
382
20
Elementary rearrangement leads to
in+2 − 2 cosh α in+1 + in = 0
with cosh α = 1 + R1 /2R2 . Transforming and solving for I(z)/z gives
I(z)/z =
zi0 + (i1 − 2i0 cosh α)
(z − eα )(z − e−α )
# α
&
i0 e + (i1 − 2i0 cosh α) i0 e−α + (i1 − 2i0 cosh α)
1
=
−
2 sinh α
z − eα
z − e−α
Thus
ik =
(i0 eα + (i1 − 2i0 cosh α))enα − (i0 e−α + (i1 − 2i0 cosh α))e−nα
2 sinh α
=
1
{i1 sinh nα − i0 sinh(n − 1)α}
sinh α
Exercises 6.6.5
21
Transforming in the quiescent state and writing as Y(z) = H(z)U(z) , then
21(a)
H(z) =
z2
1
− 3z + 2
z2
z−1
− 3z + 2
21(b)
H(z) =
21(c)
H(z) =
22
z3
1 + 1/z
− z2 + 2z + 1
For the first system, transforming from a quiescent state, we have
(z2 + 0.5z + 0.25)Y(z) = U(z)
The diagram for this is the standard one for a second-order system and is shown
in Figure 6.1 and where Y(z) = P(z) , that is yk = pk .
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
383
Figure 6.1: The block diagram for the basic system of Exercise 22.
Transforming the second system in the quiescent state, we obtain
(z2 + 0.5z + 0.25)Y(z) = (1 − 0.6)U(z)
Clearly
(z2 + 0.5z + 0.25)(1 − 0.6z)P(z) = (1 − 0.6z)U(z)
indicating that we should now set Y(z) = P(z) − 0.6zP(z) and this is shown in
Figure 6.2.
Figure 6.2: The block diagram for the second system of Exercise 22.
c Pearson Education Limited 2011
384
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
23(a)
Yδ (z)/z =
so
z
1
z
1
−
2 z + 1/4 2 z + 1/2
Yδ (z) =
yk =
23(b)
1
(4z + 1)(2z + 1)
1
1
(−1/4)k − (1/2)k
2
2
Yδ (z)/z =
z2
z
− 3z + 3
whence
√
3 + 3j
√
Yδ (z) =
2 3j
so
√
z
z
3 − 3j
√
√
− √
2 3j
(3 + 3j )
(3 − 3j )
z−
z−
2
2
√
√
3 + 3j √ k j kπ/6 3 − 3j √ k −j kπ/6
√
( 3) e
( 3) e
− √
yk =
2 3j
2 3j
(
'√
√ k
3
1
sin kπ/6 + cos kπ/6
= 2( 3)
2
2
√
= 2( 3)k sin(k + 1)π/6
23(c)
Yδ (z)/z =
so
Yδ (z) =
then
yk =
z
(z − 0.4)(z + 0.2)
1 z
2 z
+
3 z − 0.4 3 z + 0.2
2
1
(0.4)k + (−0.2)k
3
3
23(d)
Yδ (z)/z =
so
Yδ (z) =
5z − 12
(z − 2)(z − 4)
z
z
+4
z−2
z−4
and
yk = (2)k + (4)k+1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
385
24(a)
Yδ (z) =
=
yk =
z2
1
− 3z + 2
1
1
−
z−2 z−1
0, k = 0
2k−1 − 1, k ≥ 1
24(b)
Yδ (z) =
so
yk =
25
0,
k−1
2
1
z−2
k=0
, k≥1
Examining the poles of the systems, we find
25(a)
Poles at z = −1/3 and z = −2/3, both inside | z |= 1 so the system is
stable.
25(b)
Poles at z = −1/3 and z = 2/3, both inside | z |= 1 so the system is
stable.
√
25(c) Poles at z = 1/2 ± 1/2j , | z |= 1/ 2 , so both inside | z |= 1 and the
system is stable.
25(d) Poles at z = −3/4 ±
system is unstable.
√
17/4, one of which is outside | z |= 1 and so the
25(e) Poles at z = −1/4 and z = 1 thus one pole is on | z |= 1 and the other is
inside and the system is marginally stable.
c Pearson Education Limited 2011
386
26
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
To use the convolution result, calculate the impulse response as yδ,k − (1/2)k .
Then the step response is
yk =
k
1 × (1/2)
k−j
= (1/2)
k
j=0
k
1 × (2)j = (1/2)k
j=0
1 − (2)k+1
1−2
= (1/2)k (2k+1 − 1) = 2 − (1/2)k
Directly,
Y(z)/z =
2
1
z
=
−
(z − 1/2)(z − 1)
z − 1 z − 1/2
so
yk = 2 − (1/2)k
27
t
k
k
k
= −
⇒ f(t) = k − ke− τ
1
s(sτ + 1)
s s+ τ
&
#
T
z
kz(1 − e− τ )
k
−
=
⇒ G(z) = k
T
z − 1 z − e− Tτ
(z − 1)(z − e− τ )
G(s) =
Characteristic equation is 1 + G(z) = 0
⇒ (z − 1)(z − e− τ ) + kz(1 − e− τ ) = 0
T
T
T
τ
= 0, where K = k(1 − e−a ) − (1 + e−a )
⇒ z2 + [k(1 − e−a ) − (1 + e−a )]z + e−a = 0, where a =
⇒ z2 + Kz + e−a
Using Jury’s procedure:
F(z) = z2 + Kz + e−a
F(1) = 1 + K + e−a = k(1 − e−a ) > 0 since k > 0, (1 − e−a ) > 0
(−1)2 F(−1) = 1 − Kz + e−a > 0 ⇒ 2(1 + e−a ) − k(1 − e−a ) > 0
T
2(1 + e−α )
a
= 2coth
⇒k<
= 2coth
−a
(1 − e )
2
2τ
)
)
) 1
e−a ))
= 1 − e−2a > 0
Δ1 = )) −a
e
1 )
T
Thus system is stable if and only if 0 < k < 2coth 2τ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
28
Substituting
yn+1 − yn + Kyn−1 = K/2n
or
yn+2 − yn+1 + Kyn = K/2n+1
Taking z transforms from the quiescent state, the characteristic equation is
z2 − z + K = 0
with roots
1 1√
1 1√
+
1 − 4K and z2 = −
1 − 4K
2 2
2 2
For stability, both roots must be inside | z |= 1 so if K < 1/4 then
z1 =
z1 < 1 ⇒
and
z2 > −1 ⇒
1 1√
+
1 − 4K < 1 ⇒ K > 0
2 2
1 1√
1 − 4K > −1 ⇒ k > −2
−
2 2
If K > 1/4 then
1
1√
+j
4K − 1 |2 < 1 ⇒ K < 1
2
2
The system is then stable for 0 < K < 1.
|
When k = 2/9, we have
2
1
yn+2 − yn+1 + yn =
9
9
Transforming with a quiescent initial state,
1
z
2
(z2 − z + )Y(z) =
9
9 z − 1/2
so
#
&
1
1
Y(z) = z
9 (z − 1/2)(z − 1/3)(z − 2/3)
=2
z
z
z
+2
−4
z − 1/3
z − 2/3
z − 1/2
which inverts to
yn = 2(1/3)n + 2(2/3)n − 4(1/2)n
c Pearson Education Limited 2011
387
388
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
29
z2 + 2z + 2 = (z − (−1 + j ))(z − (−1 + j ))
establishing the pole locations. Then
Yδ (z) =
So since (−1 ± j ) =
z
1
z
1
−
2j z − (−1 + j ) 2j z − (−1 − j )
√ ±3j π/4
2e
etc.,
√
yk = ( 2)k sin 3kπ/4
Exercises 6.8.3
30(a)
#
zI − A =
(zI − A)−1
k
−1
A =Z
{z(zI − A)
−1
z −1
−4 z
&
1
=
(z − 2)(z + 2)
−1
}=Z
#1
1/2
2 z−2
z
z−2
#
1 k 2
2
=
4
4
30(b)
#
zI − A =
#
z
4
1
z
'
&
=
1/2
z−2
1
z−2
+
−
1/2
z+2
1
z+2
&
1/2
1 z
1 z
+ 12 z+2
4 z−2 − 4 z+2
z
z
1 z
− z+2
2 z−2 + z+2
&
&
#
1
2 −1
k
+ (−2)
2
−4 2
z + 1 −3
−3 z + 1
1/4
z−2
1/2
z−2
−
+
&
&
#
1
z+1
3
(zI − A) =
3
z+1
(z + 4)(z − 2)
'
(
1/2
1/2
1/2
−1/2
+
+
z+4
z−2
z+4
z−2
=
1/2
1/2
1/2
1/2
− z+4
+ z−2
+
z−2
z+2
# 1 z
z
z
z &
+ 12 z−2
− 21 z+4
+ 12 z−2
k
−1
−1
−1
2
z+4
A = Z {z(zI − A) } = Z
z
z
1 z
1 z
− 12 z+4
+ 12 z−2
2 z−2 + 2 z−2
&
&
#
#
1
1 −1
1 1
k
k
+ (2)
(−4)
=
−1
1
1 1
2
−1
c Pearson Education Limited 2011
1/4
z+2
1/2
z+2
(
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
&
z + 1 −1
zI − A =
0
z+1
&
# 1
#
&
1
1
z+1
1
−1
z+1
(z+1)2
=
(zI − A) =
1
0
z+1
(z + 1)2
0
z+1
&
# z
&
#
z
1 −k
2
= (−1)k
Ak = Z−1 {z(zI − A)−1 } = Z−1 z+1 (z+1)
z
0 1
0
z+1
#
30(c)
31
Taking x1 = x and x2 = y we can express equations in the form
#
x(k + 1) =
&
4
x(k) with x(0) = [1 2]
1
−7
−8
The solution is given by
#
k
x(k + 1) = A x(0), A =
−7
−8
4
1
&
where Ak = α1 I + α1 A . That is, the solution is
#
x(k) =
α0 − 7α1
−8α1
4α1
α0 + α1
#
& # &
&
α0 + α1
1
=
2
2α0 − 6α1
The eigenvalues of A are given by
λ2 − 6λ + 25 = 0 so λ = 3 ± j4
or, in polar form, λ1 = 5ejθ, λ2 = 5e−jθ where θ = cos−1 (− 53 ) .
Thus, α0 and α1 are given by
5k ejkθ = α0 + α1 5ejθ, 5k e−jkθ = α0 + α1 5e−jθ
which are readily solved to give
α0 = −
sin kθ
(5)k sin(k − 1)θ
1
, α1 = (5)k
sin θ
5
sin θ
c Pearson Education Limited 2011
389
390
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
#
&
(5)k 1
sin kθ − sin(k − 1)θ
α0 + α1 =
sin θ 5
&
#
3
4
5 k 1
sin kθ + sin kθ + cos kθ
= (5)
4
5
5
5
Then,
= (5)k [sin kθ + cos kθ]
#
&
6
(5)k
−2(sin kθ cos θ − cos kθ sin θ) − sin kθ
2α0 − 6α1 =
sin θ
5
= (5)k [2 cos(kθ)]
Thus, solution is
x(k) = (5)k [sin kθ + cos kθ]
y(k) = (5)k [2 cos(kθ)]
We have
x(1) = 1 , y(1) = −6 , x(2) = −31 , y(2) = −14
#
32
A=
0
−0.16
1
−1
&
#
zI − A =
'
(zI − A)−1 =
#
−1
A =Z
{z(zI − A)
−1
z+1
(z+0.2)(z+0.8)
−0.16
(z+0.2)(z+0.8)
#
}=
A x(0) = A [1 − 1] =
k
1
(z+0.2)(z+0.8)
z
(z+0.2)(z+0.8)
T
(zI − A)
(
5
1
5
1
3 · z+0.2 − 3 · z+0.8
1
1
− 13 z+0.2
+ 43 · z+0.8
1
bU(z) =
(z + 0.2)(z + 0.8)
#
z+1
−0.16
c Pearson Education Limited 2011
&
5
5
k
k
3 (−0.2) − 3 (−0.8)
− 31 (−0.2)k + 43 (−0.8)k
− 31 (−0.2)k + 43 (−0.8)k
0.2
3.2
k
k
3 (−0.2) − 3 (−0.8)
U(z) = Z{u(k)} = z/(z − 1)
−1
&
4
1
k
k
3 (−0.2) − 3 (−0.8)
−0.8
0.8
k
k
3 (−0.2) + 3 (−0.8)
#
k
−1
z+1
4
1
1
1
3 · z+0.2 − 3 · z+0.8
−0.8
1
1
0.8
3 z+0.2 + 3 z+0.8
=
k
z
0.16
1
z
&
& # &
z
1
1 z−1
&
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
391
&
#
z
z+2
=
(z + 0.2)(z + 0.8)(z − 1) z − 0.16
' 5 z
z
25 z (
− 2 z+0.2 + 10
9 z+0.8 + 18 z−1
=
8
7 z
z
z
1
2 z+0.2 − 9 z+0.8 + 18 z−1
# 5
25 &
k
− 2 (−0.2)k + 10
−1
−1
9 (−0.8) + 18
Z {(zI − A) bU(z)} =
1
8
7
k
k
2 (−0.2) − 9 (−0.8) + 18
Thus, solution is
x(k) = Ak x(0) + Z−1 {(zI − A)−1 bU(t)}
(
' 17
25
k
− 6 (−0.2)k + 22
9 (−0.8) + 18
=
17.6
7
3.4
k
k
6 (−0.2) − 9 (−0.8) + 18
33
Let x1 (k) = y(k), x2 (k) = x1 (k + 1) = y(k + 1) then the difference equation
may be written
#
x(k + 1) =
x1 (k + 1)
x2 (k + 1)
&
#
=
0
1
1
1
& #
&
x1 (k)
, x(0) = [0 1]T
x2 (k)
√
√
&
1+ 5
1 5
0 1
its eigenvalues are λ1 =
, λ2 =
Taking A =
1 1
2
2
Ak = α1 I + α1 A where α0 and α1 satisfy
#
√
√ k
√
√ k
1 + 5
1 − 5
1 − 5
1 + 5
α1 ,
α1
= α0 +
= α0 +
2
2
2
2
giving
√ k
√ k&
#
1 − 5
1 1 + 5
−
α1 = √
2
2
5
√ k
√ k √
√ &
#
1 1 + 5 5 − 1 1 − 5 1 + 5
+
α0 = √
2
2
2
2
5
Solution to the difference equation is
&
#
#
α0
y(k)
=
x(k) =
α1
y(k + 1)
α1
α0 + α1
#
& # &
&
α1
0
=
1
α0 + α1
√ k
√ k&
#
1 − 5
1 1 + 5
so y(k) = α1 = √
−
2
2
5
c Pearson Education Limited 2011
392
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
√ k
√ k+1 &
#
1 − 5
1 1 + 5
using above
[Note that, y(k + 1) = α0 + α1 = √
−
2
2
5
values.]
√
√ k
1 − 5
( 1+2 5 )k+1
1 √
√
As k → ∞,
→ 0 and y(k + 1)/y(k) →
= ( 5 + 1)
2
2
( 1+ 5 )k
2
Exercises 6.9.3
&
# &
0
0 1
B=
A=
1
0 −2
#
34
&
#
#
&
⎡
1
1
s −1
s
+
2
1
⎢
⇒ (sI − A)−1 =
sI − A =
=⎣s
0 s+2
0
s
s(s + 2)
0
&
#
1 12 (1 − e−2T )
⇒ G = e−AT = L−1 {(sI − A)−1 } =
0
e−2T
&T # & # 1
#
* T
1 −2T
0
−
t 12 t + 14 e−2t
At
2T + 4e
=
e Bdt =
H=
1 −2t
1 −2T
1
1
0
−2e
0
2 − 2e
0
1
2
s
1
4
−
1
2
s+2 ⎥
⎦
1
s+2
&
Discretized form is:
#
& #
x1 [(k + 1)T]
1
=
x2 [(k + 1)T]
0
!
2 (1
− e−2T )
e−2T
&#
& #1
&
x1 (kT)
T + 14 (e−2T − 1)
2
+
u(kT)
1
−2T
x2 (kT)
)
2 (1 − e
In the particular case, when sampling period is T = 1, this reduces to
#
& #
x1 (k + 1)
1
=
0
x2 (k + 1)
0.432
0.135
&#
&
& #
x1 (k)
0.284
u(k)
+
0.432
x2 (k)
In MATLAB the commands:
A = [0, 1; 0, −2]; B = [0; 1];
[G, H] = c2d(A, B, 1)
return
1.0000
0
0.2838
H=
0.4324
G=
0.4323
0.1353
which check with the answer.
c Pearson Education Limited 2011
⎤
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
#
35
(a)
0
A=
−1
&
# &
0
1
B=
1
−1
Using (6.89),
#
1
T
G1 = (TA + I) =
−T 1 − T
# &
0
H1 = TB =
T
&
giving the Euler discretized form of state-space model as
& #
1
x1 [(k + 1)T]
=
x[(k + 1)T] =
−T
x2 [(k + 1)T]
#
y(kT) = [ 1
(b)
T
1−T
&#
& # &
x1 (kT)
0
u(kT)
+
T
x2 (kT)
0 ] x(kT)
&
&
#
1
−1
s+1 1
−1
⇒ (sI − A) = 2
s+1
s + s + 1 −1 s
⎡
⎤
1
1
(s + 2 ) + 2
1
√
√
⎢
1 2
3 2
1 2
3 2 ⎥
(s
+
)
+
(
)
(s
+
)
+
(
⎢
2
2
2
2 ) ⎥
⇒ (sI − A)−1 = ⎢
⎥
(s + 12 ) − 12
⎣
⎦
−1
#
s
sI − A =
1
(s + 12 )2 + (
√
3 2
2 )
⇒ G = L−1 {(sI − A)−1 }
√
√
'
3
√1 sin( 3 T)
cos(
T)
+
T
2
2
3√
= e− 2
2
3
√
− 3 sin( 2 T)
(s + 12 )2 + (
√
3 2
2 )
√
(
√2 sin( 3 T)
2
√ 3
√
3
1
√
cos( 2 T) − 3 sin( 23 T)
Since det(A) = 0 the matrix H is best determined using (6.95),
−1
#
−1
B = (G − I)
0
&
H = (G − I)A
√
√
(
'
T
T
1 − e− 2 cos( 23 T) − √13 e− 2 sin( 23 T)
√
⇒H=
T
√2 e− 2 sin( 3 T)
2
3
giving the step-invariant discretized form of the state-space model as
x[(k + 1)] = Gx(kT) + Hu(kT)
y(kT) = [ 1
0 ]x(kT)
c Pearson Education Limited 2011
393
394
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
36(a)
The eigenvalues of the matrix A are given by
)
)
)
)λ + 1
−1
)=0
det(λI − A) = 0 ⇒ ))
1
λ + 2)
√
√
3
3
⇒ λ2 + 3λ + 1 = 0 ⇒ λ1 = − + j 3, λ2 = − − j 3
2
2
Both eigenvalues have negative real parts so the matrix A represents a stable
system.
(b) From (6.89), the corresponding Euler discretized state matrix Ad is given by
#
1−T
T
Ad = G1 = I + TA =
−T 1 − 2T
(c)
&
The eigenvalues of the matrix Ad are given by
)
)λ − 1 + T
)
)
T
)
)
−T
)=0
λ − 1 + 2T )
⇒ λ2 + (3T − 2)λ + (3T2 − 3T + 1) = 0
Let F(λ) = λ2 + (3T − 2)λ + (3T2 − 3T + 1) then
F(1) = 1 + (3T − 2) + (1 − 3T + 3T2 ) = 3T2 > 0 if T > 0 (since T is non-negative
by definition).
(−1)2 F(−1) = 1 − ((3T − 2) + 1 − 3T + 3T2 = 3T2 − 6T + 4 > 0 all T.
Taking a1 = (3T − 2) and a0 = (1 − 3T + 3T2 ) gives
F(λ) = λ2 + a1 λ + a0
leading to Jury table:
1
a0
Δ1 = 1 − a20
a1 (1 − a0 )
a1
a1
a0
1
a1 (1 − a0 )
1 − a20
Δ2 = (1 − a20 )2 − a21 (1 − a0 )2 = (1 − a0 )2 [(1 + a0 )2 − a21 ]
with Δ1 > 0 if 1 − a20 > 0 ⇒ |a0 | < 1
and Δ2 > 0 if [(1 + a0 )2 − a21 ] > 0 ⇒ 1 + a0 > |a1 | or a0 > −1 + a1 and
a0 > −1 − a1 .
In terms of T these conditions become:
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
395
Δ1 > 0 ⇒ 3T2 − 3T + 1 < 1 ⇒ T(T − 1) < 0 ⇒ T < 1
Δ2 > 0 ⇒ 3T2 − 3T + 1 > 3T − 3 ⇒ 3T2 − 6T + 4 > 0 all T (as above);
and 3T2 − 3T + 1 > −3T − 1 ⇒ T2 > 0
Thus discrete system stable provided 0 < T < 1.
37(a)
#
−1
A=
1
&
#
k
0
B= 1
0
0
0
−1
&
⎡
⎤
1
0
1
s+1 0
s
0
⎢
⎥
⇒ (sI − A)−1 =
= ⎣ 1 s + 11
sI − A =
⎦
1
−1 s
s(s + 1) 1 s + 1
−
s s+1 s
&
#
−T
0
e
⇒ G = eAT = L−1 {(sI − A)−1 } =
−T
(1 − e ) 1
& #
&
&
#
* T
T #
k1 (1 − e−T )
k1 0
0
−e−t c
At
=
H=
e B dt =
t + e−t t 0 0 −1
k1 (e−T + T − 1) −T
0
#
&
#
&
Thus discrete form of model is
&
#
k1 (1 − e−T )
0
x(kT) +
k1 (e−T + T − 1)
1
#
e−T
x[(k + 1)T] =
(1 − e−T )
&
0
u(kT)
−T
In the particular case T = 1, the model becomes
#
0.368
x(k + 1) =
0.632
&
#
0.632 k1
0
x(k) +
1
0.368k1
&
0
u(k)
−1
(b) Taking sampling period T = 1 and feedback control policy,
u1 (k) = kc − x2 (k)
#
0.368
x(k + 1) =
0.632
#
0.368
⇒ x(k + 1) =
0.632
&#
& #
&#
&
x1 (k)
0.632k1 0
kc − x2 (k)
+
x2 (k)
0.368k1 −1
u2 (k)
&#
& #
&#
&
x1 (k)
0.632k1 0
kc
−0.632k1
+
x2 (k)
0.368k1 −1
1
u2 (k)
0
1
Given u2 = 1.1x1 (0) and k1 =
3
16
, the discrete-time state-equation becomes
c Pearson Education Limited 2011
396
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
#
0.368
x(k + 1) =
0.632
−0.1185
1
&#
& #
x1 (k)
0.1185
+
0.069
x2 (k)
0
−1
&#
kc
1.1x1 (0)
&
(c) Adopting the feedback control policy
u1 (t) = kc − x2 (t)
the given continuous-time state model becomes
&
#
k
−1 k1
x+ 1
ẋ =
0
1
0
#
Taking k1 =
3
16
0
−1
&#
kc
u2
&
and u2 = 1.1x1 (0) this reduces to
&#
&
# 3
&
3
k
0
−1 − 16
c
x + 16
ẋ =
0 −1
1
0
1.1x1 (0)
&
&
#
#
3
3
1
s − 16
s + 1 16
−1
⇒ (sI − Ac ) = 2
(sI − Ac ) =
3
−1
s
1 s+1
s + s + 16
⎡ −1
⎤
3
3
3
−8
2
2
8
+
+
⎢ s+ 1
s + 34 s + 14
s + 34 ⎥
−1
4
⎢
⎥
⇒ (sI − Ac ) = ⎣
3
1
⎦
2
2
2
2
−
−
s + 14
s + 34 s + 14
s + 34
#
giving
e
Ac t
−1
=L
{(sI − Ac )
−1
#
− 12 e− 4 t + 32 e− 4 t
}=
1
3
2e− 4 t − 2e− 4 t
1
− 38 e− 4 t + 38 e− 4 t
3
3 − 14 t
− 12 e− 4 t
2e
3
1
The response of the continuous feedback system is
#
x(t) = e
Ac t
&
& * t
#
x1 (0)
kc
A(t−τ )
+
e
dτB
kc
1.1x1 (0)
0
Carrying out the integration and simplifying gives the response
x1 (t) = x1 (0)[1.1 − 2.15e− 4 t + 2.05e− 4 t ]
1
3
x2 (t) = kc + x1 (0)[−5.867 + 8.6e− 4 t − 2.714e− 4 t ]
1
c Pearson Education Limited 2011
3
3
&
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
397
Exercises 6.11.6
38
H(s) =
Replace s with
s2
1
+ 3s + 2
2 z−1
to give
Δ z+1
H̃(z) =
Δ2 (z + 1)2
4(z − 1)2 + 6Δ(z2 − 1) + 2Δ2 (z + 1)2
Δ2 (z + 1)2
=
(4 + 6Δ + 2Δ2 )z2 + (4Δ2 − 8)z + (4 − 6Δ + 2Δ2 )
This corresponds to the difference equation
(Aq2 + Bq + C)yk = Δ2 (q2 + 2q + 1)uk
where
A = 4 + 6Δ + 2Δ2
B = 4Δ2 − 8
C = 4 − 6Δ + 2Δ2
Now put q = 1 + Δδ to get
(AΔ2 δ2 + (2A + B)Δδ + A + B + C)yk
= Δ2 (Δ2 δ2 + 4Δδ + 4)uk
With t = 0.01 in the q form the system poles are at z = 0.9048 and z = 0.8182,
inside | z |= 1. When t = 0.01 these move to z = 0.9900 and z = 0.9802,
closer to the stability boundary. Using the δ form with t = 0.1, the poles are at
ν = −1.8182 and ν = −0.9522, inside the circle centre (−10, 0) in the ν-plane with
radius 10. When t = 0.01 these move to ν = −1.9802 and ν = −0.9950, within
the circle centre (−100, 0) with radius 100, and the closest pole to the boundary
has moved slightly further from it.
39
The transfer function is
H(s) =
s3
+
1
+ 2s + 1
2s2
c Pearson Education Limited 2011
398
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
To discretize using the bi-linear form use s →
2z−1
to give
Tz+1
T3 (z + 1)3
H̃(z) =
Az3 + Bz2 + Cz + D
and thus the discrete-time form
(Aq3 + Bq2 + Cq + D)yk = T3 (q3 + 3q2 + 3q + 1)uk
where
A = T3 + 4T2 + 8T + 8,
C = 3T3 − 4T2 − 8T + 3,
To obtain the δ form use s →
B = 3T3 + 4T2 − 8T − 3,
D = T3 − 4T2 + 8T − 1
2δ
giving the δ transfer function as
2 + Δδ
(2 + Δδ)3
Aδ3 + Bδ2 + Cδ + D
This corresponds to the discrete-time system
(Aδ3 + Bδ2 + Cδ + D)yk = (Δ3 δ3 + 2Δ2 δ2 + 4Δδ + 8)uk
where
A = Δ3 + 4Δ2 + 8Δ + 8,
C = 12Δ + 16,
B = 6Δ2 + 16Δ + 16,
D=8
40 Making the given substitution and writing the result in vector–matrix form,
we obtain
#
&
# &
0
1
0
ẋ(t) =
x(t) +
u(t)
−2 −3
1
and
y(t) = [1, 0]x(t)
This is in the general form
ẋ(t) = Ax(t) + bu(t)
y = cT x(t) + d u(t)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
399
The Euler discretization scheme gives at once
x((k + 1)Δ) = x(k Δ) + Δ [Ax(k Δ) + bu(k Δ)]
Using the notation of Exercise 38 write the simplified δ form equation as
&
#
,
8
12 + 8Δ
1 + 2 2
2
δ+
yk =
Δ δ + 4Δδ + 4 uk
δ +
A
A
A
Now, as usual, consider the related system
&
#
12 + 8Δ
8
2
pk = uk
δ +
δ+
A
A
and introduce the state variables x1 (k) = pk , x2 (k) = δpk together with the
redundant variable x3 (k) = δ2 pk . This leads to the representation
⎡
0
δx(k) = ⎣ 8
−
A
#
yk =
8Δ2
4
− 2
A
A
⎤
# &
1
0
u(k)
12 + 8Δ ⎦ x(k) +
1
−
A
&
4Δ (12 + 8Δ)Δ2
Δ2
−
u(k)
,
x(k)
+
A
A2
A
or
x(k + 1) = x(k) + Δ [A(Δ)x(k) + bu(k)]
yk = cT (Δ)x(k) + d(Δ)uk
Since A(0) = 4 it follows that using A(0) , c(0) and d(0) generates the Euler
Scheme when x(k) = x(kΔ) etc.
41(a)
In the z form substitution leads directly to
H(z) =
12(z2 − z)
(12 + 5Δ)z2 + (8Δ − 12)z − Δ
When Δ = 0.1, this gives
H(z) =
12(z2 − z)
12.5z2 + −11.2z − 0.1
c Pearson Education Limited 2011
400
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
(b)
The γ form is given by replacing z by 1 + Δγ.
rearrangement gives
H̃(γ) =
γ2 Δ(12
12γ(1 + Δγ)
+ 5Δ) + γ(8Δ − 12) + 12
when Δ = 0.1, this gives
H̃(γ) =
12γ(1 + 0.1γ)
1.25γ2 − 11.2γ + 12
Review exercises 6.12
1
Z {f(kT)} = Z {kT} = TZ {k} = T
2
Z a sin kω = Z
k
z
(z − 1)2
ak (ej kω − e−j kω )
2j
1
Z (aej ω )k − (ae−j ω )k
2j
z
z
1
−
=
2j z − aej ω
z − ae−j ω
az sin ω
= 2
z − 2az cos ω + a2
=
3 Recall that
Z ak =
z
(z − a)2
Differentiate twice wrt a, then put a = 1 to get the pairs
k ←→
z
(z − 1)2
k(k − 1) ←→
2z
(z − 1)3
then
Z k2 =
2z
z
z(z + 1)
+
=
(z − 1)3
(z − 1)2
(z − 1)3
c Pearson Education Limited 2011
Substitution and
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
4
H(z) =
2z
3z
+
z − 1 (z − 1)2
so inverting, the impulse response is
{3 + 2k}
5
z
(z + 1)(z + 2)(z − 1)
1 z
1 z
1 z
+
+
=−
2z+1 3z+2 6z−1
YSTEP (z) =
Thus
1
1
1
ySTEP,k = − (−1)k + (−2)k +
2
3
6
6
F(s) =
1
1
1
= −
s+1
s s+1
which inverts to
f(t) = (1 − e−t )ζ(t)
where ζ(t) is the Heaviside step function, and so
F̃(z) = Z {f(kT)} =
z
z
−
z − 1 z − e−T
Then
e−sT F(s) ←→ f((t − T))
which when sampled becomes f((k − 1)T) and
Z {f((k − 1)T)} =
∞
f((k − 1)T)
k=0
That is
e−sT F(s) →
zk
=
1
F̃(z)
z
1
F̃(z)
z
So the overall transfer function is
z−1
z
1 − e−T
z
z
−
=
z − 1 z − e−T
z − e−T
c Pearson Education Limited 2011
401
402
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
7
H(s) =
2
1
s+1
=
−
(s + 2)(s + 3)
s+3 s+2
yδ (t) = 2e−3t − e2t −→ {2e−3kT − e2kT }
so
H̃(z) = 2
8(a)
z
z
−
−3T
z−e
z − e−2T
Simple poles at z = a and z = b. The residue at z = a is
lim (z − a)zn−1 X(z) = lim (z − a)
z→a
z→a
The residue at z = b is similarly
of these, that is
zn
an
=
(z − a)(z − b)
a−b
bn
and the inverse transform is the sum
b−a
n
a − bn
a−b
8(b)
(i) There is a only double pole at z = 3 and the residue is
d
zn
(z − 3)2
= n3n−1
z→3 dz
(z − 3)2
√
3
1
j . The individual residues are
(ii) There are now simple poles at z = ±
2
2
thus given by
n
√
1
3
±
j
2
2
√
±
lim√
3j
z→(1/2± 3/2j )
lim
Adding these and simplifying in the usual way gives the inverse transform
as
2
√ sin nπ/3
3
9
H(z) =
z
z
−
z+1 z−2
so
YSTEP (z) =
z
z
−
z+1 z−2
z
z−1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
403
3z
.z
(z − 1)(z + 1)(z − 2)
3 z
1 z
z
=
+
−2
2z−1 2z+1
z−2
=−
so
ySTEP,k =
3 1
+ (−1)k − 2k+1
2 2
1
z
z2
× 1−
=
Y(z) =
(z + 1)(z − 1)
z
z+1
10
so
yk = (−1)k
α + β αβ
z2
× 1−
+ 2 =1
Y(z) =
(z − α)(z − β)
z
z
11
so
yk = {δk } = {1, 0, 0, . . .}
z
is clearly given by
The response of the system with H(z) =
(z − α)(z − β)
Y(z) = 1/z , which transforms to
yk = {δk−1 } = {0, 1, 0, 0, . . .}
12
From H(s) =
s
the impulse response is calculated as
(s + 1)(s + 2)
yδ (t) = (2e−2t − e−t ) t ≥ 0
Sampling gives
{yδ (nT)} = 2e−2nT − enT t
with z transform
Z {yδ (nT)} = 2
z
z
−
= D(z)
−2T
z−e
z − e−T
Setting Y(z) = TD(z)X(z) gives
.
z
z
X(z)
−
Y(z) = T 2
z − e−2T
z − e−T
-
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
404
Substituting for T and simplifying gives
#
&
z − 0.8452
1
X(z)
Y(z) = z 2
2 z − 0.9744z + 0.2231
so
(z2 − 0.9744z + 0.2231)Y(z) = (0.5z2 − 0.4226z)X(x)
leading to the difference equation
yn+2 − 0.9744yn+1 + 0.2231yn = 0.5xn+2 − 0.4226xn+1
As usual (see Exercise 22), draw the block diagram for
pn+2 − 0.9744pn+1 + 0.2231pn = xn
then taking yn = 0.5pn+2 − 0.4226pn+1
yn+2 − 0.9744yn+1 + 0.2231yn = 0.5pn+4 − 0.4226pn+3
−0.9774(0.5pn+3 − 0.4226pn+2 ) + 0.2231(0.5pn+2 − 0.4226pn+1 )
= 0.5xn+2 − 0.4226xn+1
13
yn+1 = yn + avn
vn+1 = vn + bun
= vn + b(k1 (xn − yn ) − k2 vn )
= bk1 (xn − yn ) + (1 − bk2 )vn
so
yn+2 = yn+1 + a[bk1 (xn − yn ) + (1 − bk2 )vn ]
(a) Substituting the values for k1 and k2 , we get
1
yn+2 = yn+1 + (xn − yn )
4
or
1
1
yn+2 − yn+1 + yn = xn
4
4
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
405
Transforming with relaxed initial conditions gives
Y(z) =
(b) When X(z) =
1
X(z)
(2z − 1)2
A
,
z−1
#
&
z
z
z
A
4
−4
−2
Y(z) =
4
z−1
z − 1/2
(z − 1/2)2
then
yn =
14
,
A+
4 − 4(1/2)n − 2n(1/2)n−1
4
Substitution leads directly to
yk − 2yk−1 + yk−2
yk − yk−1
+ 2yk = 1
+
3
T2
T
Take the z transform under the assumption of a relaxed system to get
[(1 + 3Tz + 2T2 )z2 − (2 + 3T)z + 1]Y(z) = T2
z3
z−1
The characteristic equation is thus
(1 + 3Tz + 2T2 )z2 − (2 + 3T)z + 1 = 0
with roots (the poles)
1
1
, z=
1+T
1 + 2T
The general solution of the difference equation is a linear combination of these
z=
together with a particular solution. That is
yk = α
1
1+T
k
+β
1
1 + 2T
k
+γ
This can be checked by substitution which also shows that γ = 1/2. The
yk − yk−1
, y (0) = 0
condition y(0) = 0 gives y0 = 0 and since y (t) →
T
implies yk−1 = 0. Using these we have
1
=0
2
1
α(1 + T) + β(1 + 2T) + = 0
2
α+β+
c Pearson Education Limited 2011
406
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
with solution α = −1, β = 1/2 so
yk = −
1
1+T
k
1
+
2
1
1 + 2T
k
+
1
2
The differential equation is simply solved by inverting the Laplace transform
to give
y(t) =
1 −2t
(e
− 2e−t + 1), t ≥ 0
2
T = 0.1
Figure 6.3: Response of continuous and discrete systems in Review Exercise 14 over
10 seconds when T = 0.1.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
407
T = 0.05
Figure 6.4: Response of continuous and discrete systems in Review Exercise 14 over
10 seconds when T = 0.05.
15 Substitution for s and simplifying gives
[(4 + 6T + 2T2 )z2 + (4T2 − 8)z + (4 − 6T + 2T2 )]Y(z)
= T2 (z + 1)2 X(x)
The characteristic equation is
(4 + 6T + 2T2 )z2 + (4T2 − 8)z + (4 − 6T + 2T2 ) = 0
with roots
z=
That is,
z=
8 − 4T2 ± 4T
2(4 + 6T + 2T2 )
2−T
1−T
and z =
1+T
2+T
c Pearson Education Limited 2011
408
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The general solution of the difference equation is then
yk = α
1−T
1+T
k
+β
2−T
2+T
k
+γ
This can be checked by substitution which also shows that γ = 1/2. The
yk − yk−1
, y (0) = 0
condition y(0) = 0 gives y0 = 0 and since y (t) →
T
implies yk−1 = 0. Using these we have
α+β+
α
2+T 1
1+T
+β
+ =0
1−T
2−T 2
with solution
α=
Thus,
1−T
yk =
2
1
=0
2
1−T
2
1−T
1+T
k
β=−
2−T
2
2−T
+−
2
2−T
2+T
k
+
1
2
T = 0.05
Figure 6.5: Response of continuous and discrete systems in Review Exercise 15 over
10 seconds when T = 0.1.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
409
T = 0.05
Figure 6.6: Response of continuous and discrete systems in Review Exercise 14
over 10 seconds when T = 0.05.
16
f(t) = t2 ,
{f(kΔ)} = k2 Δ2 , k ≥ 0
Now
Z{k2 } = −z
z
d
z(z + 1)
=
2
dz (z − 1)
(z − 1)3
So
Z{k2 Δ2 } =
z(z + 1)Δ2
(z − 1)3
To get D -transform, put z = 1 + Δγ to give
FΔ (γ) =
(1 + Δγ)(2 + Δγ)Δ2
Δ3 γ3
Then the D -transform is
FΔ (γ) = ΔFΔ (γ) =
(1 + Δγ)(2 + Δγ)
γ3
c Pearson Education Limited 2011
410
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
17
Eigenvalues are given by
)
)1− λ
1
)
)
2−λ
0 = ) −1
) 0
1
)
−2 ))
1 )) = (1 − λ)(λ2 − λ − 3) + (1 − λ)
−1 − λ )
= (1 − λ)(λ − 2)(λ + 1)
so eigenvalues are λ1 = 2, λ2 = 1, λ3 = −1
Corresponding eigenvectors are the solutions of
(1 − λi )ei1 + ei2 − 2ei3 = 0
−ei1 + (2 − λi )ei2 + ei3 = 0
0ei1 + ei2 − (1 + λi )ei3 = 0
Taking i = 1, 2, 3 the eigenvectors are
e1 = [1 3 1]T , e2 = [3 2 1]T , e3 = [1 0 1]T
The modal matrix M and spectral matrix Λ are
⎡
1
M = ⎣3
1
⎡
2
M Λ = ⎣6
2
⎡
2
= ⎣6
2
⎤
⎡
2
1
0 ⎦, Λ = ⎣ 0
0
1
⎤
3 −1
2
0 ⎦, A M =
1 −1
⎤
3 −1
2
0⎦ =M Λ
1 −1
3
2
1
⎤
0 0
1 0 ⎦
0 −1
⎤
⎤ ⎡
⎡
1 3 1
1 1 −2
⎣ −1 2
1 ⎦ ⎣3 2 0⎦
1 1 1
0 1 −1
Substituting x = M y in x(k + 1) = A x(k) gives
M y(k + 1) = A M y(k)
or y(k + 1) = M−1 A M y(k) = Λ y(k)
so y(1) = Λ y(0), y(2) = Λ y(1) = Λ2 y(0)
⇒ y(k) = Λk y(0)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus,
411
⎤ ⎡
⎤ ⎡ ⎤
2k 0
0
1 3 1
α
⎦
⎣
⎣
⎦
⎣
0 1
0
β ⎦ say
x(k) = 3 2 0
k
0 0 (−1)
1 1 1
γ
⎤ ⎡
⎡
⎤
k
α2
1 3 1
⎦
⎣
⎣
= 3 2 0
β ⎦
1 1 1
γ(−1)k
⎤
⎡ k
α2 + 3β + γ(−1)k
⎦
3α2k + 2β
= ⎣
k
k
α2 + β + γ(−1)
⎡
⎤
⎡
⎡ ⎤
α + 3β + γ
1
⎣ 0 ⎦ = ⎣ 3α + 2β ⎦
α+β+γ
0
When k = 0
1
1
1
which gives α = − , β = , γ =
3
2
6
so that
⎡ 1 k 3 1
⎤
− 3 (2) + 2 + 6 (−1)k
⎦
x(k) = ⎣
−(2)k + 1
− 31 (2)k + 12 + 16 (−1)k
18
x1 (k + 1) = u(k) − 3x1 (k) − 4x2 (k)
x2 (k + 1) = −2x1 (k) − x2 (k)
y(k) = x1 (k) + x2 (k)
or in vector-matrix form
#
−3
x(k + 1) =
−2
−1
D(x) = c[zI − A]
#
1
0
−3
−2
−4
−1
&
# &
1
u(k), y(k) = [1 − 1]x(k)
x(k) +
0
1
[1 − 1]
b= 2
z + 4z − 5
z+3
= 2
z + 4z − 5
#
z + 1 −4
−2 z + 3
&
(i)
Mc =
(ii)
det Mc = −2 = 0 so Mc is of rank 2
c Pearson Education Limited 2011
& # &
1
0
(1)
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
&
#
#
&
1 − 32
1 −2 3
−1
=
(iii) Mc = −
2 0 1
0 − 12
1
T
so v = [0 − 2 ]
412
#
(iv)
T=
0
1
− 21
&
1
2
#
(v)
−1
det T = 0 and T
=2
1
2
−1
1
2
&
0
#
=
1 1
−2 0
&
Substituting z(k) = T x(k) in (1) gives
T−1 z(k + 1) = A T−1 z(k) + bu(k)
or z(k + 1) = T A
#
0
=
1
#
0
thatis, z(k + 1) =
5
T−1 z(k) + T bu(k)
& #
& #
1
−3 −4
− 12
1
−2
−2 −1
2
&
# &
0
1
u(k)
z(k) +
1
−4
1
0
&
#
z(k) +
0
1
− 12
1
2
& # &
1
u(t)
0
Thus C and bc are of the required form with α = −5, β = 4 which are coefficients
in the characteristic polynomial of D(z) .
c Pearson Education Limited 2011
7
Fourier series
Exercises 7.2.6
1(a)
1
a0 =
π
0
−π
−πdt +
π
tdt
0
2 π
t
π
1
π2
1
0
2
(−πt)−π +
−π +
=−
=
=
π
2 0
π
2
2
0
π
1
an =
−π cos ntdt +
t cos ntdt
π −π
0
π
0
t
1
1 π
− sin nt
sin nt + 2 cos nt
+
=
π
n
n
n
0
−π
2
1
− 2 , n odd
(cos
nπ
−
1)
=
=
πn
πn2
0,
n even
0
π
1
bn =
−π sin ntdt +
t sin ntdt
π −π
0
π
0
t
1
1 π
cos nt
+ − cos nt + 2 sin nt
=
π n
n
n
0
−π
⎧
3
⎪
⎨ ,
n odd
1
= (1 − 2 cos nπ) = n 1
⎪
n
⎩ − , n even
n
Thus, the Fourier expansion of f(t) is
2
1
3
π
+
sin nt −
sin nt
− 2 cos nt +
4
πn
n
n
n even
n odd
n odd
∞
∞
∞
sin(2n − 1)t
π 2 cos(2n − 1)t
sin 2nt
f(t) = − −
−
+3
2
4 π n=1 (2π − 1)
(2n − 1)
2n
n=1
n=1
f(t) = −
i.e.
c Pearson Education Limited 2011
414
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1(b)
0
1 t2
π
+ πt
(t + π)dt =
=
π 2
2
−π
−π
0
0
sin nt cos nt
1
1
(t + π)
+
an =
(t + π) cos ntdt =
π −π
π
n
n2 −π
0,
n even
1
2
=
(1
−
cos
nπ)
=
, n odd
πn2
πn2
0
cos nt sin nt
1 0
1
1
−(t + π)
+
bn =
(t + π) sin ntdt =
=
−
π −π
π
n
n2 −π
n
1
a0 =
π
0
Thus, the Fourier expansion of f(t) is
i.e.
1(c)
∞
2
1
π
sin nt
cos nt −
f(t) = +
2
4
πn
n
n=1
n odd
∞
∞
π 2 cos(2n − 1)t sin nt
f(t) = +
−
4 π n=1 (2n − 1)2
n
n=1
From its graph we see that f(t) is an odd function; so it has Fourier
expansion
f(t) =
∞
bn sin nt
n=1
with
t
2 π
2 π
sin ntdt
1−
f(t) sin nt =
bn =
π 0
π 0
π
π
1
t
1
2
2
− 1−
cos nt −
sin nt =
=
2
π
n
π
πn
nπ
0
Thus, the Fourier expansion of f(t) is
∞
2 sin nt
f(t) =
π n=1 n
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1(d)
From its graph f(t) is seen to be an even function; so its Fourier
expansion is
∞
a0
+
an cos nt
f(t) =
2
n=1
with
2
a0 =
π
2
an =
π
2
=
π
π
0
0
π
2
f(t)dt =
π
π/2
2 cos tdt =
0
2
f(t) cos ntdt =
π
2
4
π/2
[2 sin t]0 =
π
π
π/2
2 cos t cos ntdt
0
π/2
[cos(n + 1)t + cos(n − 1)t]dt
0
π/2
2 sin(n + 1)t sin(n − 1)t
+
=
π
(n + 1)
(n − 1) 0
1
2
π
1
π
=
sin(n + 1) +
sin(n − 1)
π (n + 1)
2 (n − 1)
2
⎧
0,
n odd
⎪
⎪
⎪
1
4
⎨−
, n = 4, 8, 12, . . .
=
π (n2 − 1)
⎪
⎪
4
1
⎪
⎩
, n = 2, 6, 10, . . .
2
π (n − 1)
Thus, the Fourier expansion of f(t) is
f(t) =
∞
4 (−1)n+1 cos 2nt
2
+
π π n=1
4n2 − 1
1(e)
π
1
t
t
4
2 sin
cos dt =
=
2
π
2 −π
π
−π
π
π
1
1
1
1
t
an =
cos(n + )t + cos(n − )t dt
cos cos ntdt =
π −π
2
2π −π
2
2
2
1
2
1
2
sin(n + )π +
sin(n − )π
=
2π (2n + 1)
2
(2n − 1)
2
⎧
4
⎪
, n = 1, 3, 5, . . .
⎨
π(4n2 − 1)
=
4
⎪
⎩−
, n = 2, 4, 6, . . .
2
π(4n − 1)
bn = 0
1
a0 =
π
π
c Pearson Education Limited 2011
415
416
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, the Fourier expansion of f(t) is
∞
4 (−1)n+1 cos nt
2
f(t) = +
π π n=1 (4n2 − 1)
1(f)
Since f(t) is an even function, it has Fourier expansion
∞
a0
an cos nt
f(t) =
+
2
n=1
2 π
2 π
| t | dt =
tdt = π
a0 =
π 0
π 0
π
1
2 π
2 t
sin nt + 2 cos nt
t cos ntdt =
an =
π 0
π n
n
0
0,
n even
2
4
=
(cos
nπ
−
1)
=
− 2 , n odd
πn2
πn
Thus, the Fourier expansion of f(t) is
with
π 4 1
−
cos nt
2 π
n2
n odd
∞
π 4 cos(2n − 1)t
that is, f(t) = −
2 π n=1 (2n − 1)2
f(t) =
1(g)
1 π
1 2
π
t − πt 0 = 0
a0 =
(2t − π)dt =
π 0
π
π
π
2
1
1 (2t − π)
sin nt + 2 cos nt
(2t − π) cos ntdt =
an =
π 0
π
n
n
0
4
2
=
(cos nπ − 1) = − πn2 , n odd
πn2
0,
n even
π
π
(2t − π)
2
1
1
−
cos nt + 2 sin nt
bn =
(2t − π) sin ntdt =
π 0
π
n
n
0
0,
n odd
1
2
= − (cos nπ + 1) =
− , n even
n
n
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
417
Thus, the Fourier expansion of f(t) is
f(t) =
−
4
2
cos
nt
+
− sin nt
2
πn
n
n even
odd
∞
∞
4 cos(2n − 1)t sin 2nt
f(t) = −
−
π n=1 (2n − 1)2
n
n=1
n
that is,
1(h)
1
a0 =
π
=
=
an =
=
=
=
1
bn =
π
0
t
(−t + e )dt +
−π
π
t
(t + e )dt
0
0
π
t2
1 t2
t
t
− +e
+
+e
π
2
2
−π
0
1 2
2
π + (eπ − e−π ) = π + sinh π
π
π
0
π
1
t
t
(−t + e ) cos ntdt +
(t + e ) cos ntdt
π −π
0
0
t
t
1
1
1
− sin nt + 2 cos nt
ne sin nt + et cos nt
+ 2
π
n
n
(n + 1)
−π
π
t
t
1
1
π
ne sin nt + et cos nt 0
+
sin nt + 2 cos nt + 2
n
n
(n + 1)
0
2 cos nπ eπ − e−π
2
(−1
+
cos
nπ)
+
πn2
π(n2 + 1)
2
2 (cos π − 1)
cos nπ
sinh π , cos nπ = (−1)n
+ 2
2
π
n
(n + 1)
0
π
t
(t + e ) sin ntdt
t
(−t + e ) sin ntdt +
−π
0
−π
0
0
π
t
1
1
1 t
cos nt − 2 sin nt
+ − cos nt + 2 sin nt
=
π n
n
n
n
−π
0
t
π
2
t
e cos nt e sin nt
n
−
+
+ 2
π +1
n
n2
−π
2n
n
cos nπ(eπ − e−π ) = −
cos nπ sinh π, cos nπ = (−1)n
=−
2
π(n + 1)
π(n2 + 1)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
418
Thus, the Fourier expansion of f(t) is
∞
2
π 1
(−1)n − 1 (−1)n sinh π
+ sinh π +
cos nt
+
f(t) =
2 π
π n=1
n2
n2 + 1
−
2
∞
2 n(−1)n
sinh π sin nt
π n=1 n2 + 1
Since the periodic function f(t) is an even function, its Fourier expansion is
∞
a0
+
an cos nt
f(t) =
2
n=1
with
π
1
2
2
3
− (π − t)
(π − t) dt =
= π2
π
3
3
0
0
π
π
2(π − t)
2
2 (π − t)2
2
2
sin nt −
an =
(π − t) cos ntdt =
cos nt − 3 sin nt
π 0
π
n
n2
n
0
4
= 2
n
2
a0 =
π
π
2
Thus, the Fourier expansion of f(t) is
∞
1
π2
+4
cos nt
f(t) =
3
n2
n=1
Taking t = π gives
0=
∞
1
π2
+4
(−1)n
2
3
n
n=1
so that
∞
1 2 (−1)n+1
π =
12
n2
n=1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
3
Since q(t) is an even function, its Fourier expansion is
∞
q(t) =
with
a0
+
an cos nt
2
n=1
2 π
a0 =
π 0
2 π
an =
π 0
Qt
dt = Q
π
π
Qt
2Q t
1
cos ntdt = 2
sin nt + 2 cos nt
π
π n
n
0
0,
n
even
2Q
4Q
= 2 2 (cos nπ − 1) =
− 2 2 , n odd
π n
π n
Thus, the Fourier expansion of q(t) is
∞
4 cos(2n − 1)t
1
q(t) = Q − 2
2 π n=1 (2n − 1)2
4
1 π
1
10
5 sin tdt = [−5 cos t]π0 =
a0 =
π 0
π
π
π
π
5
5
sin t cos ntdt =
[sin(n + 1)t − sin(n − 1)t]dt
an =
π 0
2π 0
π
cos(n + 1)t cos(n − 1)t
5
−
+
, n = 1
=
2π
(n + 1)
(n − 1) 0
1
5 cos nπ
cos nπ
1
− −
=
−
+
2π
n+1
(n − 1)
n+1 n−1
0,
n odd, n = 1
5
10
=−
(cos
nπ
+
1)
=
, n even
−
π(n2 − 1)
π(n2 − 1)
Note that in this case we need to evaluate a1 separately as
π
1 π
5
5 sin t cos tdt =
sin 2tdt = 0
a1 =
π 0
2π 0
π
5 π
5
bn =
sin t sin ntdt = −
[cos(n + 1)t − cos(n − 1)t]dt
π 0
2π 0
π
5 sin(n + 1)t sin(n − 1)t
−
, n = 1
=−
2π
(n + 1)
(n − 1) 0
= 0 , n = 1
c Pearson Education Limited 2011
419
420
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Evaluating b1 separately,
π
5 π
5
sin t sin tdt =
(1 − cos 2t)dt
b1 =
π 0
2π 0
π
1
5
5
t − sin 2t =
=
2π
2
2
0
Thus, the Fourier expansion of f(t) is
∞
10 cos 2nt
5 5
+ sin t −
π 2
π n=1 4n2 − 1
f(t) =
5
a0 =
=
an =
=
=
1
bn =
π
0
π
1
2
2
π dt +
(t − π) dt
π −π
0
π
1
4
1 2 0
3
π t −π + (t − π)
= π2
π
3
3
0
0
π
1
2
2
π cos ntdt +
(t − π) cos ntdt
π −π
0
0
(t − π)2
2(t − π)
2
1 π2
sin nt
sin nt +
+
cos nt − 3 sin nt
2
π
n
n
n
n
−π
2
n2
0
2
0
(t − π) sin ntdt
2
π sin ntdt +
−π
π
π
0
0
(t − π)2
(t − π)
2
1 π2
− cos nt
cos nt + 2
+ −
sin
nt
+
cos nt
=
π
n
n
n2
n3
−π
π2
1 π2
− + (−1)n
=
π
n
n
2
π
[1 − (−1)n ]
= (−1)n −
n
πn3
c Pearson Education Limited 2011
π
0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, the Fourier expansion of f(t) is
∞
∞
2 2 2
(−1)n
4 sin(2n − 1)t
f(t) = π +
cos nt +
π sin nt −
3
n2
n
π n=1 (2n − 1)3
n=1
5(a)
Taking t = 0 gives
∞
2
π2 + π2
2
= π2 +
2
3
n2
n=1
and hence the required result
5(b)
∞
1
1
= π2
2
n
6
n=1
Taking t = π gives
∞
2
π2 + 0
2
= π2 +
(−1)n
2
2
3
n
n=1
and hence the required result
∞
(−1)n+1
1 2
π
=
2
n
12
n=1
6(a)
c Pearson Education Limited 2011
421
422
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
6(b)
The Fourier expansion of the even function (a) is given by
∞
f(t) =
a0
+
an cos nt
2
n=1
with
2
a0 =
π
π/2
(π − t)dt
tdt +
0
π
π/2
π
π
2 1 2 π/2 1
t
+ − (π − t)2
=
=
π 2 0
2
2
π/2
π/2
π
2
t cos ntdt +
(π − t) cos ntdt
an =
π 0
π/2
π/2
π − t
1
1
2 t
sin nt + 2 cos nt
sin nt − 2 cos nt
+
=
π n
n
n
n
0
1
nπ
2 2
− 2 (1 + (−1)n )
cos
=
2
π n
2
n
⎧
n odd
⎪
⎨ 0,
8
= − 2 , n = 2, 6, 10, . . .
⎪
⎩ πn
0,
n = 4, 8, 12, . . .
Thus, the Fourier expansion of f(t) is
∞
π 2 cos(4n − 2)t
f(t) = −
4 π n=1 (2n − 1)2
Taking t = 0 where f(t) = 0 gives the required result.
c Pearson Education Limited 2011
π
π/2
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
423
7
1
a0 =
π
π
0
t
(2 − )dt +
π
2π
t/πdt
π
π
t2 2π
t2
1
2t −
=3
+
=
π
2π 0
2π π
π
2π
t
1
t
cos ntdt
an =
(2 − ) cos ntdt +
π 0
π
π
π
π
2π
t
t
1
1
1 2
sin nt −
sin nt −
sin
nt
+
=
cos
nt
+
cos
nt
π n
πn
πn2
πn
πn2
0
π
2
= 2 2 [1 − (−1)n ]
π n
0,
n even
4
=
, n odd
π2 n2
π
2π
t
1
t
bn =
(2 − ) sin ntdt +
sin ntdt
π 0
π
π
π
π
t
t
1
1
1 2
− cos nt +
sin
nt
+
−
sin nt
cos nt −
cos nt +
=
2
π
n
πn
πn
πn
πn2
0
=0
Thus, the Fourier expansion of f(t) is
∞
3
4 cos(2n − 1)t
f(t) = + 2
2 π n=1 (2n − 1)2
Replacing t by t − 12 π gives
∞
1
3
4 cos(2n − 1)(t − π)
f(t − π) = + 2
2
2 π n=1
(2n − 1)2
c Pearson Education Limited 2011
2π
π
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
424
Since
1
π
π
cos(2n − 1)(t − π) = cos(2n − 1)t cos(2n − 1) + sin(2n − 1)t sin(2n − 1)
2
2
2
n+1
sin(2n − 1)t
= (−1)
∞
3
4 (−1)n+1 sin(2n − 1)t
1
f(t − π) − = 2
2
2
π n=1
(2n − 1)2
The corresponding odd function is readily recognized from the graph of f(t) .
Exercises 7.2.8
8
Since f(t) is an odd function, the Fourier expansion is
f(t) =
∞
bn sin
n=1
with
2
bn =
0
nπt
t
nπt
nπt
2
nπt 2
−
t sin
sin
dt =
cos
+
nπ
nπ
0
2
cos nπ
=−
nπ
Thus, the Fourier expansion of f(t) is
∞
2 (−1)n+1
nπt
f(t) =
sin
π n=1
n
9
Since f(t) is an odd function (readily seen from a sketch of its graph) its
Fourier expansion is
f(t) =
∞
bn sin
n=1
with
2
bn =
0
nπt
K
nπt
( − t) sin
tdt
K
nπt Kt
nπt
K
nπt
2
−
cos
+
cos
−
sin
=
nπ
nπ
(nπ)2
0
2K
=
nπ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Thus, the Fourier expansion of f(t) is
∞
nπt
2K 1
sin
f(t) =
π n=1 n
10
1
a0 =
5
5
3dt = 3
0
5
nπt
1 15
nπt
3 cos
=0
dt =
sin
5
5 nπ
5 0
0
5
1
15
nπt
1 5
nπt
dt =
−
cos
bn =
3 sin
5 0
5
5
nπ
5 0
6
3
=
[1 − (−1)n ] = nπ , n odd
nπ
0,
n even
1
an =
5
5
Thus, the Fourier expansion of f(t) is
∞
3 6
1
(2n − 1)
f(t) = +
sin
πt
2 π n=1 (2n − 1)
5
11
π/ω
A
ω
2A
− cos ωt
A sin ωtdt =
=
π
ω
π
0
0
π/ω
π/ω
Aω
Aω
sin ωt cos nωtdt =
[sin(n + 1)ωt − sin(n − 1)ωt]dt
an =
π 0
2π 0
π/ω
cos(n + 1)ωt cos(n − 1)ωt
Aω
−
+
=
, n = 1
2π
(n + 1)ω
(n − 1)ω 0
A
A 2(−1)n+1
2
=
=
− 2
[(−1)n+1 − 1]
2
2
2π n − 1
n −1
π(n − 1)
⎧
n odd , n = 1
⎨ 0,
2A
= −
, n even
⎩
π(n2 − 1)
2ω
a0 =
2π
π/ω
Evaluating a1 separately,
Aω
a1 =
π
π/ω
0
A
sin ωt cos ωtdt =
2π
π/ω
sin 2ωtdt = 0
0
c Pearson Education Limited 2011
425
426
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Aω
bn =
π
π/ω
0
Aω
sin ωt sin nωtdt = −
2π
π/ω
[cos(n + 1)ωt − cos(n − 1)ωt]dt
0
π/ω
Aω sin(n + 1)ωt sin(n − 1)ωt
−
=−
, n = 1
2π
(n + 1)ω
(n − 1)ω 0
= 0, n = 1
Aω π/ω 2
Aω π/ω
A
b1 =
sin ωtdt =
(1 − cos 2ωt)dt =
π 0
2π 0
2
Thus, the Fourier expansion of f(t) is
∞
cos 2nωt
π
A
1 + sin ωt − 2
f(t) =
π
2
4n2 − 1
n=1
12
Since f(t) is an even function, its Fourier expansion is
∞
f(t) =
nπt
a0
+
an cos
2
T
n=1
with
2
a0 =
T
an =
=
2
T
T
T
2 1 3
2
t
t dt =
= T2
T 3 0
3
2
T
2 Tt
nπt
2tT2
2T3
nπt
nπt
nπt
2
dt =
sin
+
−
t cos
cos
sin
T
T nπ
T
(nπ)2
T
(nπ)3
T 0
2
0
T
0
2
4T
(−1)n
(nπ)2
Thus, the Fourier series expansion of f(t) is
f(t) =
∞
4T2 (−1)n
nπt
T2
+ 2
cos
3
π n=1 n2
T
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
13
2
a0 =
T
2
an =
T
T
0
T
0
T
E
2E 1 2
tdt = 2
t
=E
T
T 2 0
E
2πnt
t cos
dt
T
T
T
2E tT
2πnt
2πnt T 2
= 2
cos
=0
sin
+
T 2πn
T
2πn
T 0
2E T
2πnt
bn = 2
t sin
dt
T 0
T
T
2πnt T 2
tT
2πnt
E
2E
cos
+
sin
=−
= 2 −
T
2πn
T
2πn
T 0
πn
Thus, the Fourier expansion of e(t) is
∞
2πnt
E E1
sin
e(t) = −
2
π n=1 n
T
Exercises 7.3.3
14
Half range Fourier sine series expansion is given by
f(t) =
∞
bn sin nt
n=1
with
2
bn =
π
π
0
π
1
2
− cos nt
1 sin ntdt =
π
n
0
2
[(−1)n − 1]
nπ
0,
n even
4
=
, n odd
nπ
Thus, the half range Fourier sine series expansion of f(t) is
=−
∞
4 sin(2n − 1)t
f(t) =
π n=1 (2n − 1)
Plotting the graphs should cause no problems.
c Pearson Education Limited 2011
427
428
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
15
Half range Fourier cosine series expansion is given by
∞
a0
+
an cos nπt
f(t) =
2
n=1
with
2
a0 =
1
1
(2t − 1)dt = 0
0
1
(2t − 1) cos nπtdt
an = 2
0
1
2
(2t − 1)
sin nπt +
=2
cos nπt
nπ
(nπ)2
0
4
=
[(−1)n − 1]
(nπ)2
0,
n even
8
= −
, n odd
(nπ)2
Thus, the half range Fourier cosine series expansion of f(t) is
∞
1
8
cos(2n − 1)πt
f(t) = − 2
π n=1 (2n − 1)2
Again plotting the graph should cause no problems.
16(a)
1
a0 = 2
0
1
(1 − t2 )dt = 2 t − t3
3
1
=
0
4
3
1
(1 − t2 ) cos 2nπtdt
an = 2
0
1
(1 − t2 )
2
2t
=2
cos 2nπt +
sin 2nπt
sin 2nπt −
2nπ
(2nπ)2
(2nπ)3
0
1
=−
(nπ)2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1
(1 − t2 ) sin 2nπtdt
bn = 2
0
1
2t
(1 − t2 )
2
cos 2nπt −
=2 −
sin 2nπt −
cos 2nπt
2nπ
(2nπ)2
(2nπ)3
0
1
=
nπ
Thus, the full-range Fourier series expansion for f(t) is
f(t) = f1 (t) =
16(b)
∞
∞
1 1
2
11
− 2
sin 2nπt
cos
2nπt
+
3 π n=1 n2
π n=1 n
Half-range sine series expansion is
f2 (t) =
∞
bn sin nπt
n=1
with
1
(1 − t2 ) sin nπtdt
bn = 2
0
(1 − t2 )
=2 −
cos nπt −
nπ
2
1
=2 −
(−1)n +
3
(nπ)
nπ
⎧
2
⎪
⎨
,
n
= nπ
1
4
⎪
⎩2
+
, n
nπ (nπ)3
1
2
2t
sin nπt −
cos nπt
(nπ)2
(nπ)3
0
2
+
(nπ)3
even
odd
Thus, half-range sine series expansion is
∞
∞
2
4
1
11
f2 (t) =
sin 2nπt +
+
sin(2n − 1)πt
π n=1 n
π n=1 (2n − 1) π2 (2n − 1)3
16(c)
Half-range cosine series expansion is
∞
a0
f3 (t) =
an cos nπt
+
2
n=1
c Pearson Education Limited 2011
429
430
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
with
1
(1 − t2 )dt =
a0 = 2
0
4
3
1
(1 − t2 ) cos nπtdt
an = 2
0
1
2t
(1 − t2 )
2
sin nπt −
=2
cos nπt +
sin nπt
nπ
(nπ)2
(nπ)3
0
−4(−1)n
=
(nπ)2
Thus, half-range cosine series expansion is
∞
2
4 (−1)n+1
cos nπt
f3 (t) = + 2
3 π n=1
n2
Graphs of the functions f1 (t), f2 (t), f3 (t) for −4 < t < 4 are as follows:
17
Fourier cosine series expansion is
∞
a0
an cos nt
f1 (t) =
+
2
n=1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
with
2 π
1
(πt − t2 )dt = π2
a0 =
π 0
3
π
2
an =
(πt − t2 ) cos ntdt
π 0
π
(π − 2t)
2
2 (πt − t2 )
sin nt +
cos nt + 3 sin nt
=
π
n
n2
n
0
2
= − 2 [1 + (−1)n ]
n
0,
n odd
4
=
− 2 , n even
n
Thus, the Fourier cosine series expansion is
∞
1 2 1
f1 (t) = π −
cos 2nt
2
6
n
n=1
Fourier sine series expansion is
f2 (t) =
∞
bn sin nt
n=1
with
bn =
=
=
=
2 π
(πt − t2 ) sin ntdt
π 0
π
(πt − t2 )
(π − 2t)
2
2
−
cos nt +
sin nt − 3 cos nt
π
n
n2
n
0
4
[1 − (−1)n ]
πn3
0,
n even
8
, n odd
πn3
Thus, the Fourier sine series expansion is
∞
1
8
sin(2n − 1)t
f2 (t) =
π n=1 (2n − 1)3
c Pearson Education Limited 2011
431
432
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Graphs of the functions f1 (t) and f2 (t) for −2π < t < 2π are
18
2a
x, 0 n, then m + n > 2n which implies that
Dn+m (t2 − 1)n = 0
so that Im,n = 0
If m = n then
Im,n = In,n = (−1)
1
n
−1
= (2n)!(−1)
(t2 − 1)n D2n (t2 − 1)n dt
1
n
−1
(t2 − 1)n dt
1
(1 − t2 )n dt
= 2(2n)!
0
Making the substitution t = sin θ then gives
π/2
cos2n+1 θdθ = 2(2n)!
In,n = 2(2n)!
0
2
2
...
2n + 1
3
22n+1
(n!)2
=
2n + 1
and the result follows.
41(c)
f(t) = c0 P0 (t) + c1 P1 (t) + c2 P2 (t) + . . .
Multiplying by P0 (t)
1
−1
giving
f(t)P0 (t)dt = c0
1
(−1)1dt +
−1
1
−1
P20 (t)dt = 2c0
1
(1)1dt = 0 = 2c0 so that c0 = 0
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
461
Multiplying by P1 (t) ,
1
−1
giving
f(t)P1 (t)dt = c1
0
−1
P21 (t)dt =
1
(−1)tdt +
−1
1
2
3
c1 , so that c1 =
3
2
(1)tdt = 1 =
0
Likewise,
1
−1
f(t)P2 (t)dt = c2
1
−1
2
a1
3
P22 (t)dt =
2
c2
5
giving
1
2
0
1
(−1)(3t − 1)dt +
2
−1
0
(1)(3t2 − 1)dt = 0 =
2
and
1
1
−1
f(t)P3 (t)dt = c3
1
−1
P23 (t)dt =
2
c2 , so that c2 = 0
5
2
c3
7
giving
1
2
42
0
1
(−1)(5t − 3t)dt +
2
−1
1
(1)(5t3 − 3t)dt = −
3
0
1
7
2
= c3 , so that c3 = −
4
7
8
Taking
f(x) = c0 P0 (x) + c1 P1 (x) + c2 P2 (x) + c3 P3 (x) + . . .
and adopting the same approach as in 41(c) gives
−1
giving
1
f(x)P0 (x)dx = c0
1
−1
P20 (x)dx = 2c0
1
1
1
= 2c0 , so that c0 =
2
4
0
1
1
2
f(x)P1 (x)dx = c1
P21 (x)dx = c1
3
−1
−1
xdx =
c Pearson Education Limited 2011
462
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
giving
1
1
1
2
= c1 , so that c1 =
3
3
2
0
1
1
2
f(x)P2 (x)dx = c2
P22 (x)dx = c2
5
−1
−1
x2 dx =
giving
1
2
1
2
1
5
= c2 , so that c2 =
8
5
16
0
1
1
2
f(x)P3 (x)dx = c3
P23 (x)dx = c3
7
−1
−1
x(3x2 − 1)dx =
giving
1
2
1
x(5x3 − 3x)dx = 0 =
0
43(a)
2
c3 , so that c3 = 0
7
L0 (t) = et (t0 e−t ) = 1
L1 (t) = et (−te−t + e−t ) = 1 − t
Using the recurrence relation,
L2 (t) = (3 − t)L1 (t) − L0 (t) = t2 − 4t + 2
L3 (t) = (5 − t)L2 (t) − 4L1 (t)
= (5 − t)(t2 − 4t + 2) − 4(1 − t)
= 6 − 18t + 9t2 − t3
43(b)
This involves evaluating the integral
combinations of m and n.
43(c)
∞
If f(t) =
∞
0
e−t Lm (t)Ln (t)dt for the 10
cr Lr (t) to determine cn , multiply throughout by e−t Ln (t)
r=0
and integrate over (0, ∞)
∞
−t
∞
∞
e Ln (t)f(t)dt =
0
0
cr e−t Lr (t)Ln (t)dt
r=0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
463
Using the orthogonality property then gives
∞
−t
∞
e Ln (t)f(t)dt = cn
0
0
giving cn =
44(a)
e−t Ln (t)Ln (t)dt
1
(n!)2
= cn (n!)2
∞
e−t Ln (t)f(t)dt, n = 0, 1, 2, . . .
0
By direct use of formula,
2
H0 (t) = (−1)0 et
/2 −t2 /2
e
=1
2
2
d
H1 (t) = (−1)et /2 e−t /2 = t
dt
Using recurrence relation,
Hn (t) = tHn−1 (t) − (n − 1)Hn−2 (t)
H2 (t) = t.t − 1.1 = t2 − 1
H3 (t) = t(t2 − 1) − 2(t) = t3 − 3t
H4 (t) = t(t3 − 3t) − 3(t2 − 1) = t4 − 6t2 + 3
44(b) This involves evaluating the integral
combinations of n and m .
44(c)
If f(t) =
∞
∞
e−t
2
−∞
/2
Hn (t)Hm (t)dt for the 10
cr Hr (t) to determine cn , multiply throughout by
r=0
e−t
2
/2
Hn (t) and integrate over (−∞, ∞) giving
∞
e
−∞
−t2 /2
Hn (t)f(t)dt =
∞
∞
−∞ r=0
∞
= cn
= cn
cr e−t
2
e−t
−∞
2
/2
/2
Hn (t)Hr (t)dt
Hn (t)Hn (t)dt
(2π)n!
c Pearson Education Limited 2011
464
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so that
1
cn =
n! (2π)
45(a)
∞
e−t
2
/2
−∞
f(t)Hn (t)dt
Directly from the formula,
T0 (t) = cos 0 = 1
T1 (t) = cos(cos−1 t) = t
then from the recurrence relationship
T2 (t) = 2t(t) − 1 = 2t2 − 1
T3 (t) = 2t(2t2 − 1) − t = 4t3 − 3t
T4 (t) = 2t(4t3 − 3t) − (2t2 − 1) = 8t4 − 8t2 + 1
T5 (t) = 2t(8t4 − 8t2 + 1) − (4t3 − 3t) = 16t5 − 20t3 + 5t
45(b)
Evaluate the integral
1 Tn (t)Tm (t)
dt for the 10 combinations of n
−1
(1 − t2 )
and m .
∞
If f(t) =
cr Tr (t) to obtain cn ,
r=0
cn Tn (t)/ (1 − t2 ) and integrate over (−1, 1) giving
45(c)
1
−1
multiply throughout by
∞
cr Tn (t)Tr (t)
dt
(1 − t)2
−1 r=0
1
Tn (t)Tn (t)
= cn
dt Tn = 0, 1, 2, 3, . . .
(1 − t2 )
−1
T (t)f(t)
n
dt =
(1 − t2 )
1
=
c0 π,
cn π2 ,
n=0
n = 0
Hence the required results.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
465
46(a)
To show that they are orthonormal on (0, T) , evaluate the integral
T
0
Wn (t)
Wm (t)dt for the 10 combinations of n and m . For example,
T
T
W0 (t)W0 (t)dt =
0
0
and it is readily seen that this extends to
T
W1 (t)W2 (t)dt =
0
0
T /4
1
dt +
T
T /2
T /4
T
0
1
at = 1
T
W2n (t)dt = 1
(−1)
dt +
T
3T /4
T /2
c Pearson Education Limited 2011
1
dt +
T
T
3T /4
(−1)
dt = 0
T
466
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
46(b) f(t) = c0 W0 (t) + c1 W1 (t) + c2 W2 (t) + . . . where f(t) is the square wave
of Exercise 40. In this case T = 2π. Multiplying throughout by the appropriate
Walsh function and integrating over (0, 2π) gives
2π
2π
W0 (t)f(t)dt = c0
0
0
1
W20 (t)dt = c0 , W0 (t) = √
2π
giving
2π
1 π
1f(t)dt = √
dt −
dt = 0
2π 0
0
π
2π
2π
√1 ,
2
2π
W1 (t)f(t)dt = c1
W1 (t)dt = c1 , W1 (t) =
√1 ,
−
0
0
2π
1
c0 = √
2π
2π
0
T1 T2
4MK
·
π
T1 + T2
c Pearson Education Limited 2011
8
The Fourier Transform
Exercises 8.2.4
1
0
at −jωt
e e
F(jω) =
∞
dt +
−∞
e−at e−jωt dt
0
2a
= 2
a + ω2
2
0
Ae
F(jω) =
−jωt
−T
T
=
T
dt +
−Ae−jωt dt
0
2jA sin ωt dt
0
2jA
(1 − cos ωT)
ω
4jA
ωT
=
sin2
ω
2
ωT
= jωAT2 sinc2
2
=
3
T
At
At
−jωt
+A e
+ A e−jωt dt
F(jω) =
−
dt +
T
T
−T
0
T
At
+ A cos ωt dt
−
=2
T
0
ωT
2
= AT sinc
2
0
Exercise 2 is T × derivative of Exercise 3, so result 2 follows as (jω × T)
× result 3.
Sketch is readily drawn.
c Pearson Education Limited 2011
490
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
4
2
2Ke−jωt dt = 8K sinc(2ω)
F(jω) =
−2
1
Ke−jωt dt = 2K sinc(ω)
G(jω) =
−1
H(jω) = F(jω) − G(jω) = 2K(4 sinc(2ω) − sinc(ω))
5
−1
e
F(jω) =
−jωt
1
dt +
−2
e
−jωt
2
dt +
−1
−e−jωt dt
1
1 jω
2(e − e−jω ) − (e2jω − e−2jω )
=
jω
= 4 sinc(ω) − 2 sinc(2ω)
6
1
F(jω) =
2j
1
f̄(a) =
2j
=
π
a
π
−a
π
a
π
−a
sin ω πa
(ejat − e−jat )e−jωt dt
e
jat −jωt
e
1
dt =
2j
π
a
π
−a
ej(a−ω)t dt
j(a − ω)
F(jω) = f̄(a) + f̄(−a) =
7
F(jω) =
∞
π
2jω
sin
ω
ω2 − a2
a
e−at . sin ω0 t.e−jωt dt
0
= f̄(ω0 ) − f̄(−ω0 )
1 ∞ (−a+j(ω0 −ω)t)
where f̄(ω0 ) =
e
dt
2j 0
1
1
1
1
=
=
2j a − j(ω0 − ω)
2j (a + jω) − jω0
ω0
∴ F(jω) =
(a + jω)2 + ω20
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
8
1
Fc (x) =
4
define g(x, b) =
a
(ejt + e−jt )(ejxt + e−jxt ) dt
0
a
ej(b+x)t dt
0
1
[ej(b+x)a − 1]
j(b + x)
1
Fc (x) = [g(x, 1) + g(x, −1) + g(−x, 1) + g(−x, −1)]
4
1 sin(1 + x)a sin(1 − x)a
=
+
2
1+x
1−x
=
9
Consider F(x) =
a
0
1.ejxt dt
−j
(cos ax + j sin ax − 1)
x
sin ax
Fc (x) = Re F(x) =
x
1 − cos ax
Fs (x) = Im F(x) =
x
=
10
Consider F(x) =
∞ −at jxt
e e
0
=
dt
a + jx
a2 + x2
a
+ x2
x
Fs (x) = Im F(x) = 2
a + x2
Fc (x) = Re F(x) =
a2
Exercises 8.3.6
11
Obvious
12
(jω)2 Y(jω) + 3jωY(jω) + Y(jω) = U(jω)
1
U(jω)
(1 − ω2 ) + 3jω
1
H(jω) =
(1 − ω2 ) + 3jω
Y(jω) =
c Pearson Education Limited 2011
491
492
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
13
→ sinc
ω
2
→ e−iω3/2 + eiω3/2 sinc
ω
2
2
(sin(2ω) − sin(ω))
ω
= 4 sinc(2ω) − 2 sinc(ω)
=
14
F(jω) =
T
2
T
−2
cos(ω0 t)e−iωt dt
1
1
T
T
sin(ω0 − ω) +
sin(ω0 + ω)
ω0 − ω
2
ω0 + ω
2
T
T
sin(ω0 + ω) 2
T sin(ω0 − ω) 2
=
+
T
2 (ω0 − ω) 2
(ω0 + ω) T2
=
ω = ±ω0
Evaluating at ω = ±ω0 ⇒
T
T
T
sinc(ω0 − ω) + sinc(ω0 + ω)
F(jω) =
2
2
2
15
F(jω) =
T
cos ω0 t.e−jωt dt
0
1
= [f̄(ω0 ) + f̄(−ω0 )]
2
where
T
f̄(ω0 ) =
ej(ω0 −ω)t dt
0
=
1
[ej(ω0 −ω)T − 1]
j(ω0 − ω)
c Pearson Education Limited 2011
ω = ω0
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1
1
(ej(ω0 −ω)T − 1)
F(jω) =
2 j(ω0 − ω)
1
−j(ω0 +ω)T
(e
−
− 1)
j(ω0 − ω)
jω0 T /2
T
−jωT /2 e
sin(ω0 − ω)
=e
ω0 − ω
2
−jω0 T /2
e
T
sin(ω0 + ω)
+
ω0 + ω
2
ω = ω0
ω = ±ω0
Checking at ω = ±ω0 gives
T −jωT /2 jω0 T /2
T
T
−jω0 T /2
F(jω) = e
e
sinc(ω0 − ω) + e
sinc(ω0 + ω)
2
2
2
16
1
F(jω) =
sin 2t.e−jωt dt
−1
1 1 −j(ω−2)t
=
e
− e−j(ω+2)t dt
2j −1
1
f̄(a) =
e−j(ω−a)t dt = 2 sinc(ω − a)
−1
1
1
F(jω) = f̄(a) − f̄(−a), a = 2
2j
2j
= j[sinc(ω + 2) − sinc(ω − 2)]
Exercises 8.4.3
17
I H(s) =
s2
1
+ 3s + 2
∞
H(jω) =
h(t) = (e−t − e−2t )ξ(t)
(e−t − e−2t )e−jωt dt =
0
=
2−
1
as required.
+ 3jω
ω2
s+2
II H(s) = 2
s +s+1
h(t) = e−1/2t
1
1
−
1 + jω 2 + jω
√
√
√
3
3
t + 3 sin
t ξ(t)
cos
2
2
c Pearson Education Limited 2011
493
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
∞
e−(1/2tjω−jω0 )t dt
Consider G(ω0 ) =
494
0
=
1
2
1
+ j(ω − ω0 )
√
√
3
3
1
1
(G(ω0 ) − G(−ω0 )), ω0 =
H(jω) = G(ω0 ) + G(−ω0 ) +
2
2
2j
2
2 + 4jω
6
So H(jω) =
+
2
4 + 4jω − 4ω
4 + 4jω − 4ω2
2 + jω
=
1 − ω2 + jω
18
P(jω) = 2ATsinc ωT
So
F(jω) = (e−jωτ + eiωτ )P(jω)
= 4AT cos ωτ sinc ωT
19
G(s ) =
(s )2
√
(s )2 + 2s + 1
G(jω) =
=
−ω2
√
1 − ω2 + 2jω
1
ω2
1
√
− 1 + 2 ωj
Thus, | G(jω) |→ 0 as ω → 0
and | G(jω) |→ 1 as ω → ∞
High-pass filter.
20
g(t) = e−a|t| −→ G(jω) =
f(jt) =
a2
2a
+ ω2
1
G(jt) −→ πg(−ω) = πe−a|ω|
2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
21
F{f(t) cos ω0 t} =
1
1
F(j(ω − ω0 )) + F(j(ω + ω0 ))
2
2
F(jω) = 2T sinc ωT
∴ F{PT (t) cos ω0 t}
= T sinc(ω − ω0 )T + sinc(ω + ω0 )T
Exercises 8.5.3
22
1
2π
∞
−∞
πδ(ω − ω0 )e
jωt
1
dω +
2π
∞
−∞
πδ(ω + ω0 )ejωt dω
1 jω0 t
(e
+ e−jω0 t )
2
= cos ω0 t
=
23
F{e±jω0 t } = 2πδ(ω ∓ ω0 )
1
∴ F{sin ω0 t} = {2πδ(ω − ω0 ) − 2πδ(ω + ω0 )}
2j
= jπ[δ(ω + ω0 ) − δ(ω − ω0 )]
∞
1
jπ[δ(ω + ω0 ) − δ(ω − ω0 )]ejωt dω
2π −∞
j
= [e−jω0 t − e+jω0 t ] = sin ω0 t
2
∞
24
G(jω) =
g(t)e−jωt dt; G(jt) =
−∞
So
∞
−∞
g(ω)e−jωt dω
∞
f(t)G(jt) dt
∞
∞
−jωt
f(t)
g(ω)e
dω dt
=
−∞
−∞
∞
∞
−jωt
g(ω)
f(t)e
dt dω
=
−∞
−∞
∞
∞
g(ω)F(jω)dω =
g(t)F(jt) dt
=
−∞
−∞
−∞
c Pearson Education Limited 2011
495
496
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
25
Write result 24 as
∞
∞
f(ω)F{g(t)}dω =
−∞
∞
−∞
∞
f(ω)F{G(jt)}dω =
so
−∞
−∞
F{f(t)}g(ω)dω
F{f(t)}G(jω)dω
⎫
g(t)
→ G(jω) ⎬
Now
G(jt) → 2πg(−ω)
symmetry
⎭
G(−jt) → 2πg(ω)
∞
∞
f(ω).2πg(ω)dω =
F(jω)G(−jω)dω
Thus,
−∞
−∞
∞
∞
1
f(t)g(t) dt =
F(jω)G(−jω)dω
or
2π −∞
−∞
26
F{H(t) sin ω0 t}
∞
1
1
du
=
πj δ(ω − u + ω0 ) − δ(ω − u − ω0 ) πδ(u) +
2π −∞
ju
1
1
j
1
= πδ(ω + ω0 ) − πδ(ω − ω0 ) +
−
2
2 ω + ω0
ω − ω0
πj
ω0
δ(ω + ω0 ) − δ(ω − ω0 ) − 2
=
2
ω − ω20
27
A
an =
T
d/2
e−jnω0 t dt =
−d/2
f(t) =
nω0 d
Ad
sinc
,
T
2
ω0 = 2π/T
∞
Ad
nω0 d jnω0 t
e
sinc
,
T n=−∞
2
∞
nω0 d
2πAd
δ(ω − nω0 )
sinc
F(jω) =
T n=−∞
2
Exercises 8.6.6
28
T = 1,
N = 4,
Δω = 2π/(4 × 1) =
c Pearson Education Limited 2011
π
2
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
G0 =
G1 =
G2 =
3
n=0
3
n=0
3
gn e−j ×n×0×π/2 = 2
gn e−j ×n×1×π/2 = 0
gn e−j ×n×2×π/2 = 2
n=0
G3 =
3
gn e−j ×n×3×π/2 = 0
n=0
G = {2, 0, 2, 0}
29
N = 4, Wn = e−j nπ/2
⎡
1
0
⎢
gn = ⎣
1
0
0
1
0
1
⎡
⎤ ⎡
G00
1
⎢G ⎥ ⎢1
G = ⎣ 10 ⎦ = ⎣
1
G01
0
G11
Bit reversal gives
1
0
−1
0
1
−1
0
0
⎤⎡ ⎤ ⎡ ⎤
2
1
0
1⎥⎢0⎥ ⎢0⎥
⎦⎣ ⎦ = ⎣ ⎦
0
1
0
0
0
−1
0
0
1
1
⎤⎡ ⎤ ⎡ ⎤
2
2
0
0⎥⎢0⎥ ⎢2⎥
⎦⎣ ⎦ = ⎣ ⎦
0
0
−j
0
0
j
⎡ ⎤
2
⎢0⎥
G=⎣ ⎦
2
0
30 Computer experiment.
31 Follows by direct substitution.
c Pearson Education Limited 2011
497
498
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Exercises 8.9.3
32 We have θc =
π
2
so
jθ
D(e ) =
1,
|θ| ≤
0,
|θ| >
π
2
π
2
The filter coefficients are given by
1
hd (n) =
2π
=
1
2π
π
D(ejθ )ejnθ dθ
−π
π
2
π
−2
ejnθ dθ
nπ
1
, n = 0
=
sin
nπ
2
1
nπ
= sinc
2
2
Hence,
h±5 = 0.06366, h±4 = 0, h±3 = −0.10610
h±2 = 0, h±1 = 0.31831, h0 = 0.5
Thus, the non-causal transfer function is
D̃(z) = 0.06366z−5 − 0.1066z−3 + 0.31832z−1 + 0.5
+ 0.31831z − 0.10610z3 + 0.06366z5
and the causal version is
D(z) = 0.06366 − 0.10660z−2 + 0.31831z−4 + 0.5z−5
+ 0.31831z−6 − 0.010660z−8 + 0.06366z−10
33 The Hamming window coefficients are given by
πk
, |k| ≤ 5
wH (k) = 0.54 + 0.46 cos
5
Note that wH (±4)andwH (±2)are not needed.
Now wH (±5) = 0.08000, wH (±3) = 0.39785, wH (±1) = 0.91215, wH (0) = 1 and
the causal transfer function is found by multiplying the filter coefficients by the
appropriate wH giving
D(z) = 0.00509(1 + z−10 ) − 0.04221(z−2 + z−8 ) + 0.29035(z−4 + z−6 ) + 0.5z−5
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
499
Plots of the frequency responses for both Exercises 32 and 33 are given in the
following figure.
1.4
Rectangular Window
Hamming Window
1.2
Frequency response
1
0.8
0.6
0.4
0.2
0
−3pi
−2pi
−pi
0
pi
2pi
q in radians
Review exercises 8.10
1
FS (x) =
2
t sin xt dt +
0
2
1
sin xt dt =
1
sin x cos 2x
−
x2
x
πt π
π
+ H(t − 2)
f(t) = − H(−t − 2) + (H(t + 2) − H(t − 2))
2
4
2
1
+ πδ(ω)
jω
1
−2jω
F{H(t − 2)} = e
+ πδ(ω)
jω
F{H(t)} =
c Pearson Education Limited 2011
3pi
500
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
−1
2jω −1
+ πδ(−ω) = e
+ πδ(ω)
F{H(−t − 2)} = e
jω
jω
π
π 2 −jωt
π
H(t − 2)
te
dt + F
F{f(t)} = F{− H(−t − 2)} +
2
4 −2
2
2jω
=
−πj
sinc 2ω
ω
3
F {H(t + T/2) − H(t − T/2)} = T sinc
ωT
2
F {cos ω0 t} = π [δ(ω + ω0 ) + δ(ω − ω0 )]
Using convolution,
π
F {f(t)} =
2π
∞
T
T sinc (ω − u) (δ(u + ω0 ) + δ(u − ω0 ))du
2
−∞
T
T
T
sinc(ω − ω0 ) + sinc(ω + ω0 )
=
2
2
2
4
1
1
[πδ(ω − ω0 ) + πδ(ω + ω0 )] ∗ πδ(ω) +
F {cos ω0 tH(t)} =
2π
jω
1
=
2π
1
{πδ(ω − u − ω0 ) + πδ(ω − u + ω0 )} πδ(u) +
du
ju
−∞
∞
=
π
jω
[δ(ω − ω0 ) + δ(ω + ω0 )] + 2
2
ω 0 − ω2
5
F{f(t) cos ωc t cos ωc t}
=
F(jω + jωc ) + F(jω − jωc )
∗ π [δ(ω − ωc ) + δ(ω + ωc )]
2
1 ∞
[F(j(u + ωc )) + F(j(u − ωc ))] ×
=
4 −∞
[δ(ω − u − ωc ) + δ(ω − u + ωc )] du
1
1
= F(jω) + [F(jω + 2jωc ) + F(jω − 2jωc )]
2
4
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
501
Or write as
1
f(t) (1 + cos 2ωc t)
2
etc.
6
H(t + 1) − H(t − 1) ↔ 2 sinc ω
By symmetry,
2 sinc t ↔ 2π(H(−ω + 1) − H(−ω − 1)) = 2π(H(ω + 1) − H(ω − 1))
7(a) Simple poles at s = a and s = b. Residue at s = a is eat /(a − b) , at s = b
it is ebt /(b − a) , thus
1
eat − ebt H(t)
a−b
f(t) =
7(b) Double pole at s = 2, residue is
d
lim
s→2 ds
es t
(s − 2)
(s − 2)2
2
= te2t
So f(t) = te2t H(t)
7(c) Simple pole at s = 1, residue e−t , double pole at s = 0, residue
d
lim
s→0 ds
es t
s+1
= (t − 1)H(t)
Thus, f(t) = (t − 1 + e−t )H(t)
8(a)
∞
y(t) =
−∞
Thus,
− sin ω0 t =
∞
−∞
h(t − τ)u(τ) dτ
h(t − τ) cos ω0 τdτ = f(t), say
c Pearson Education Limited 2011
502
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
If u(τ) = cos ω0 (τ + π/4)
∞
y(t) =
−∞
∞
=
−∞
h(t − τ) cos ω0 (τ + π/4) dτ
h(t − (τ − π/4)) cos ω0 τ dτ = f(t + π/4)
= − sin ω0 (t + π/4)
8(b) Since sin ω0 t = cos ω0 (t − π/2ω0 )
∞
y(t) =
−∞
∞
=
−∞
∞
=
−∞
h(t − τ) sin ω0 t dτ
h(t − τ) cos ω0 (τ − π/2ω0 ) dτ
h(t − (τ + π/2ω0 )) cos ω0 τ dτ
= f(t − π/2ω0 ) = − sin(ω0 t − π/2) = cos ω0 t
8(c)
ejω0 t = cos ω0 t + j sin ω0 t
This is transformed from above to
− sin ω0 t + j cos ω0 t = j ejω0 t
8(d) Proceed as above using
e−jω0 t = cos ω0 t − j sin ω0 t
9
F(sgn(t)) = F(f(t)) = F(jω) =
2
, obvious
jω
Symmetry,
F(jt) =
2
↔ 2/πf(−ω) = 2πsgn(−ω)
jt
That is,
1
↔ −πsgn(ω)
jt
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
or
−
10
1
1
g(t) = − ∗ f(t) = −
πt
π
so
1
πt
∞
↔ jsgn(ω)
f(τ)
1
dτ =
t−τ
π
−∞
1
g(x) =
π
∞
−∞
∞
−∞
f(τ)
dτ = FHi (t)
τ−t
f(t)
dt = FHi (x)
t−x
So from Review Exercise 9
FHi (jω) = jsgn(ω) × F(jω)
so
|FHi (jω)| = |jsgn(ω)| |F(jω)| = |F(jω)|
and
arg(FHi (jω)) = arg(F(jω)) + π/2, ω ≥ 0
Similarly
arg(FHi (jω)) = arg(F(jω)) − π/2, ω < 0
11 First part, elementary algebra.
1
FHi (x) =
π
1
1
=
π x2 + a2
∞
−∞
12(a)
∞
−∞
t
(t2
+
a2 )(t
− x)
dt
xt
a2
t
−
dt
+
t2 + a2
t − x t2 + a2
a
= 2
x + a2
∞
f(t)
dt = FHi (x)
−∞ t − x
1 ∞ f(a + t)
dt
H{f(a + t)} =
π −∞ t − x
f(t)
1 ∞
dt = FHi (a + x)
=
π −∞ t − (a + x)
1
H{f(t)} =
π
c Pearson Education Limited 2011
503
504
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
12(b)
1 ∞ f(at)
H{f(at)} =
dt
π −∞ t − x
1 ∞ f(t)
dt = FHi (ax), a > 0
=
π −∞ t − ax
12(c)
1 ∞ f(−at)
dt
H{f(−at)} =
π −∞ t − x
1 ∞ f(t)
=−
dt = −FHi (−ax), a > 0
π −∞ t + ax
12(d)
H
1
=
π
df
dt
1
=
π
∞
−∞
∞
f(t)
+
t − x
−∞
∞
−∞
f (t)
dt
t−x
f(t)
dt
(t − x)2
Provided lim f(t)/t = 0, then
|t|→∞
df
dt
H
1
=
π
∞
−∞
f(t)
1 d
dt =
2
(t − x)
π dx
x
π
∞
−∞
f(t)
1
dt +
t−x
π
∞
−∞
∞
1
f(t) dt =
π
−∞
∞
−∞
= H{tf(t)}
13
f(t)
dt
t−x
d
FHi (x)
dx
=
12(e)
From Review Exercise 10
FHi (t) = −
1
∗ f(t)
πt
So from Review Exercise 9,
F{FHi (t)} = jsgn (ω) × F(j ω)
c Pearson Education Limited 2011
tf(t)
dt
t−x
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
so
F(jω) = −jsgn (ω) × F{FHi (t)}
Thus,
∞
f(t) =
−∞
1
1
FHi (τ)dτ = −
π(t − τ)
π
∞
−∞
1
FHi (x)dx
(x − τ)
14
fa (t) = f(t) − jFHi (t)
F{fa (t)} = F(jω) − j(jsgn (ω))F(jω) = F(jω) + sgn (ω)F(jω)
=
2F(jω), ω > 0
0,
ω<0
15
F{H(t)} =
1
+ πδ(ω) = F(jw)
jω
Symmetry,
F(j t) =
1
+ πδ(t) ↔ 2πH(−ω) = 2π[1 − H(ω)]
jt
= 2π[Fδ(t) − H(ω)]
or
H(ω) ↔
1
j
+ δ(t)
2πt 2
Thus,
F−1 {H(ω)} =
Then
j
1
+ δ(t)
2πt 2
j
1
1
∗ f(t) = f(t) − j −
∗ f(t)
f̂(t) = 2 δ(t) +
2
2πt
πt
= f(t) − jFHi (t)
When f(t) = cos ω0 t, ω0 > 0, then
F(jω) = π[δ(ω − ω0 ) + δ(ω + ω0 )]
so
F{f̂(t)} = 2πδ(ω − ω0 )
c Pearson Education Limited 2011
505
506
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
whence
f̂(t) = f(t) − jFHi (t) = ejω0 t = cos ω0 t + j sin ω0 t
and so
FHi (t) = − sin ω0 t
When g(t) = sin ω0 t, ω0 > 0
G(jω) = jπ[δ(ω + ω0 ) − δ(ω − ω0 )]
and thus
ĝ(t) = −jejω0 t = −j(cos ω0 t + j sin ω0 t)
so
H{sin ω0 t} = cos ω0 t
16 If h̄(t) = 0, t < 0, then when t < 0
h̄e (t) =
1
1
h̄(−t), and h̄o (t) = − h̄(−t)
2
2
that is, h̄o (t) = −h̄e (t)
When t > 0, then
h̄e (t) =
1
1
h̄(t), and h̄o (t) = h̄(t)
2
2
that is, h̄o (t) = h̄e (t)
That is,
h̄o (t) = sgn (t)h̄e (t) ∀t
Thus,
h̄(t) = h̄e (t) + sgn (t)h̄e (t)
When h(t) = sin t H(t) ,
⎧
⎪
⎨
1
sin t, t > 0
2
h̄e (t) =
⎪
⎩ − 1 sin t, t < 0
2
and since
sgn (t) h̄e (t) =
1
sin t ∀t
2
the result is confirmed.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Then taking the FT of the result,
#
$
H̄(jω) = H̄e (jω) + F sgn (t)h̄e (t)
2
∗ H̄e (jω)
jω
#
$
= H̄e (jω) + jH H̄e (jω)
1
= H̄e (jω) +
2π
When
∞
a
ω
−
j
a2 + ω2
a2 + ω2
e−at e−jwt dt =
H̄(jω) =
−∞
then
H
a
2
a + ω2
or
H
a
2
a + t2
=−
a2
ω
+ ω2
a2
x
+ x2
=−
Finally,
H
at
a2 + t2
x
1
= −x 2
+
a + x2
π
So
H
t
2
a + t2
∞
−∞
=
a2
a
a2
dt
=
a2 + t2
a2 + x2
a
+ x2
17(a)
∞
FH (s) =
e−at (cos 2πst + sin 2πst) dt =
0
17(b)
FH (s) =
18
T
(cos 2πst + sin 2πst) dt =
−T
∞
−∞
E(s) − j O(s) =
1
sin 2πst
πs
∞
f(t) cos 2πst dt O(s) =
E(s) =
a + 2πs
+ 4π2 s2
a2
f(t) sin 2πst dt
−∞
∞
f(t)e−j2πst dt = F(js)
−∞
c Pearson Education Limited 2011
507
508
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
From Review Exercise 17(a)
FH (s) =
1 + πs
2 + 2π2 s2
1
,
2 + 2π2 s2
O(s) =
F(j s) =
1 − jπs
2 + 2π2 s2
whence
E(s) =
so
πs
2 + 2π2 s2
agreeing with the direct calculation,
∞
F(js) =
e−2t e−j2πst dt =
0
19
H{f(t − T)} =
∞
−∞
1 − jπs
2 + 2π2 s2
f(t − T) cas 2πst dt
∞
=
f(τ) [cos 2πsτ(cos 2πsT + sin 2πsT)+
−∞
sin 2πsτ(cos 2πsT − sin 2πsT)] dt
= cos 2πsTFH (s) + sin 2πsTFH (−s)
20 The Hartley transform follows at once since
FH (s) = {F(js)} − {F(js)} =
1
1
δ(s) +
2
sπ
From time shifting,
1
1
1
1
+ cos πs δ(s) +
FH (s) = sin πs δ(−s) −
2
sπ
2
sπ
=
cos πs − sin πs
1
δ(s) +
2
πs
21
∞
δ(t) cas 2πst dt = 1
H{δ(t)} =
−∞
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
509
From Review Exercise 18, it follows that the inversion integral for the Hartley
transform is
∞
f(t) =
−∞
FH (s) cas 2πstds
and so the symmetry property is simply
f(t) ↔ FH (s) =⇒ FH (t) ↔ f(s)
Thus,
H{1} = δ(s)
At once,
H{δ(t − t0 )}
∞
−∞
δ(t − t0 ) cas 2πst dt = cas 2πst0
By symmetry,
H{cas 2πs0 t} = δ(s − s0 )
22
1
=
2
1
1
FH (s − s0 ) + FH (s + s0 )
2
2
∞
−∞
f(t) {cos 2π(s − s0 )t + sin 2π(s − s0 )t
+ cos 2π(s + s0 )t + sin 2π(s + s0 )t} dt
∞
f(t) cos 2πs0 t [cos 2πst + sin 2πst] dt
=
−∞
= H{f(t) cos 2πs0 t}
From Review Exercise 21, setting f(t) = 1
H{cos 2πs0 t} =
1
(δ(s − s0 ) + δ(s + s0 ))
2
also
H{sin 2πs0 t} = H{cas 2πs0 t} − H{cos 2πs0 t}
1
1
= δ(s − s0 ) − (δ(s − s0 ) + δ(s + s0 )) = (δ(s − s0 ) − δ(s + s0 ))
2
2
c Pearson Education Limited 2011
510
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
23
t
(1 + τ2 )−1 dτ = tan−1 t +
−∞
Thus,
F{tan
−1
t} = F
t
2 −1
(1 + τ )
π
2
dτ − F
π
2
∞
π
2 −1
=F
(1 + τ ) H(t − τ)dτ − F
2
−∞
π
1
=F
∗
H(t)
−
F
1 + t2
2
1
1
π
=F
×
+ πδ(ω) − × 2πδ(ω)
2
1+t
jω
2
−∞
But from Review Exercise 1
F e−|t| =
and so by symmetry,
F
whence
#
F tan
and so
−1
$
t = πe
1
1 + t2
−|ω|
×
2
1 + ω2
= πe−|ω|
π
1
+ πδ(ω) − × 2πδ(ω)
jω
2
#
$ πe−|ω|
F tan−1 t =
jω
24
1
1
[1 + cos ω0 t] ↔ [2πδ(ω) + πδ(ω − ω0 ) + πδ(ω + ω0 )]
2
2
and
H(t + T/2) − H(t − T/2) ↔ 2T sinc ω
so
F {x(t)} =
∞
−∞
2T sinc (ω − u)
1
× πδ(u) + (δ(ω − ω0 ) + δ(u + ω0 )) du
2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1
1
= T sinc ω + sinc (ω − ω0 ) + sinc (ω + ω0 )
2
2
25
511
1
2πνr
H(ν) =
f(r) cas
4 r=0
4
3
1
[f(0) + f(1) + f(2) + f(3)]
4
1
H(1) = [f(0) + f(1) − f(2) − f(3)]
4
1
H(0) = [f(0) − f(1) + f(2) − f(3)]
4
1
H(0) = [f(0) − f(1) − f(2) + f(3)]
4
H(0) =
so
⎡
1
1 ⎢1
T= ⎣
4 1
1
1
1
−1
−1
1
−1
1
−1
⎤
1
−1 ⎥
⎦
−1
1
By elementary calculation, T2 = 1/4T and if T−1 exists, T−1 = 4T. Since
T−1 T = I , it does. Then
⎡
1
⎢1
T−1 H = ⎣
1
1
1
1
−1
−1
⎤
⎤ ⎡
⎤⎡
f(0)
H(0)
1
1
−1 −1 ⎥ ⎢ H(1) ⎥ ⎢ f(1) ⎥
⎦
⎦=⎣
⎦⎣
f(2)
H(2)
1 −1
f(3)
H(3)
−1
1
c Pearson Education Limited 2011
9
Partial Differential Equations
Exercises 9.2.6
1
Differentiating
∂2 u
∂2 u
2
= −a cos at sin bx and 2 = −b2 cos at sin bx
2
∂t
∂x
and hence a2 = c2 b2
2 Since the function is a function of a single variable only, on differentiating
∂2 u
∂2 u
2
=
α
f
and
= f and hence α2 = c2 .
2
2
∂t
∂x
3
Verified by differentiation.
4
Differentiating
1
1
cos(r − ct) − sin(r − ct)
2
r
r
2
2
1
Zrr = 3 cos(r − ct) + 2 sin(r − ct) − cos(r − ct)
r
r
r
2
c
Ztt = − cos(r − ct)
r
Zr = −
and it can be checked that the equation is satisfied.
5
Applying the given expression into the equation gives
α αt
α
e V = eαt V or V = V
κ
κ
and the solution clearly depends on the sign of α.
α = 0 ⇒ V = 0
and hence V = A + Bx
α > 0 ⇒ V = a2 V and hence V = A sinh ax + B cosh ax
where a2 =
α
κ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
513
α < 0 ⇒ V = −b2 V and hence V = A cos bx + B sin bx
α
where b2 = −
κ
6
Substituting the expression into the LHS of the equation,
∂V
∂
= nrn−1 (3 cos2 θ − 1) and
∂r
∂r
2 ∂V
r
= n(n + 1)rn (3 cos2 θ − 1)
∂r
and in the RHS,
∂V
∂
= −rn 6 cos θ sin θ and
∂θ
∂θ
∂V
sin θ
∂θ
= −rn 6(− sin3 θ + 2 cos2 θ sin θ)
Applying these expressions into the equation,
n(n + 1)rn (3 cos2 θ − 1) − rn 6(− sin2 θ + 2 cos2 θ) = 0
or n(n + 1)rn (3 cos2 θ − 1) − rn 6(−1 + 3 cos2 θ) = 0
and hence n(n + 1) − 6 = 0 with roots −3 and 2.
7
Now,
c2
∂2 u
= −c2 m2 e−kt cos mx cos nt
∂x2
∂u
= −ke−kt cos mx cos nt − ne−kt cos mx sin nt
∂t
and
∂2 u
= k2 e−kt cos mx cos nt + 2kne−kt cos mx sin nt − n2 e−kt cos mx cos nt
∂t2
Thus,
∂2 u
∂u
+ 2k = [k2 −n2 +2k(−k)]e−kt cos mx cos nt + [2kn + 2k(−n)]e−kt cos mx sin nt
2
∂t
∂t
and comparing with the LHS gives k2 + n2 = c2 m2
8
Differentiating
Vx = 3x2 + ay2 and Vy = 2axy
c Pearson Education Limited 2011
514
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
and evaluating
x
∂V
∂V
+y
= 3x3 + axy2 + 2axy2 = 3(x3 + axy2 ) = 3V
∂x
∂y
gives the required result.
Now Vxx + Vyy = 6x + 2ax ⇒ rhs = 0 if a = −3
Putting r2 = x2 + y2 , first note that
2r
∂r
∂r
= 2x and 2r
= 2y
∂x
∂y
so
x
u = r3 V ⇒ ux = r3 Vx + 3r2 V = r3 Vx + 3rxV
r
and differentiating again
x
x2
uxx = r3 Vxx + 3r2 Vx + 3rxVx + 3rV + 3 V
r
r
Similarly for uyy and adding the two expressions and using the two previous results
uxx + uyy = r3 (Vxx + Vyy ) + 6r(xVx + yVy ) + 6rV + 3
(x2 + y2 )
V
r
the quoted answer is proved.
9
Differentiating φxx = Φxx e−kt/2 and
φ t = Φt e
−kt/2
k
− Φe−kt/2
2
φtt =
k2
Φtt − kΦt + Φ e−kt/2
4
and substituting gives
1
1 −kt/2 2
k2
k2
c Φxx − Φtt + kΦt − Φ − kΦt + Φ
0 = φxx − 2 (φtt + kφt ) = 2 e
c
c
4
2
Neglecting terms in k2 , the RHS is just the wave equation for Φ .
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
10(a)
515
With r = g = 0, the equations become
− Ix = cvt
− vx = LIt
⇒ −Ixx = cvxt = c(−LIt )t = −cLItt
and hence satisfy the wave equation.
10(b)
When L = 0,
− Ix = gv + cvt
− vx = rI
⇒ vxx = r(gv + cvt ) = rgv + rcvt
and the result is a heat conduction equation with an additional forcing term rgv.
Applying W = vegt/c it may be noted that
g
Wxx = vxx egt/c and Wt = vt + v egt/c
c
and hence comparing with the previous equation,
Wxx = (rc)Wt
which satisfies the usual heat conduction equation. The exponential damps the
solution to zero over a long time.
10(c) First eliminate I
−vxx = rIx + LIxt = r(−gv − cvt ) + L(−gv − cvt )t
vxx = Lcvtt + (rc + Lg)vt + rgv
Apply in the expression for a
1
rg
vxx = vtt + 2avt +
v
Lc
Lc
and substitute v = we−at
1
rg −at
wxx e−at = (wtt − 2awt + a2 w)e−at + 2a(wt − aw)e−at +
we
Lc
Lc
1
rg
2
w
wxx = wtt + −a +
Lc
Lc
c Pearson Education Limited 2011
516
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
But
rg
1
rg
=
−
−a +
Lc
Lc 4
2
g2
r2
rg
+
+
2
L2
Lc c2
1
=−
4
g2
r2
rg
+
−
2
L2
Lc c2
=0
from the condition rc = gL and hence the variable w satisfies the wave equation.
Such a transmission line is called a balanced line and transmits the signal exactly
in shape, though damped by the exponential.
11
Applying the expression into the equation
−a2 f sin(ay + b) = (f − 2af ) sin(ay + b)
so f must satisfy
f − 2af + a2 f = 0
which is a second-order constant coefficient equation with equal roots a. Thus,
f = (A + Bx)eax and agrees with the given result.
12
The given formula can be checked by differentiation.
The method of Section 9.2.5 solves the equations
dy
df
dx
=
= 2 2
x
y
4x y
dx
dy yields
=
−→ x = Ay
x
y
df
dy
= 2 2
y
4x y
yields
yields
−→ df = 4A2 y3 dy −→ f = A2 y4 + B
The arbitrary constants A and B can be isolated as
A=
x
and B = f − x2 y2
y
x
When x = x(t), y = y(t) are given on a curve with f = f(t) then A(t) =
y
x
for some function F. Putting this into B(t) gives
and hence t = F
y
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition 517
x
x
=g
for some function g and thus f is of the required form
B F
y
y
x
f(x, y) = x y + g
y
2 2
The MAPLE code produces this solution also.
Given that x = 1 − t, y = t, f = t2
x
and t = (1 − t) t + g
y
x
1−t
=
y
t
2
2 2
y3 (y + 2x)
x
=
Eliminating t gives g
.
y
(x + y)4
Use MAPLE to solve, as follows:
with (PDEtools):
Q12:=x ∗ diff(u(x,y),x)+y ∗ diff(u(x,y),y)-4 ∗ xˆ2 ∗ yˆ2;
sol:=pdsolve(Q12,u(x,y));
# this instruction gives the solution
sol:= u(x,y) = x 2 y 2 + F1(y/x)
eval (sol,{x=1-t,y=t,u(x,y)=t^2});
simplify (eval(%,t=1/(1+z)));
solve(%,_F1(1/z));
# gives the solution (1 + 2z)/(1 + z)4
∂u
∂ ∂u
+u =0⇒
+ u = f(y)
13 Write as
∂x ∂y
∂y
where f is an arbitrary function. Using an integrating factor ey , this partial
differential equation can now be written as
∂
(uey ) = ey f(y)
∂y
which can be integrated to give
u = e−y [H(x) + G(y)]
where H(x) and G(y) are arbitrary functions.
c Pearson Education Limited 2011
518
14
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The method of Section 9.2.5 solves the equations
dx
dy
du
= 2 =
2
x
y
(x + y)u
These equations yield
dx
dy
= 2
2
x
y
du
=
u
2
A
+
y 1 − Ay
yields
−→
1
1
= −A
x
y
yields
dy −→ u =
By2
= Bxy
1 − Ay
Hence from any starting curve, with parameter s,
u
x−y
and B(s) =
xy
xy
x−y
, where F is an arbitrary function
Eliminating s gives u = xyF
xy
determined by the conditions on the starting curve.
A(s) =
MAPLE gives this general solution.
Putting in the data x = s, y = 1, f = s2 , the arbitrary function becomes
1
and the given result for u follows.
F(z) =
1−z
Exercises 9.3.4
15
From the separated solutions (9.25) choose
u = sin(λx) cos(λct)
Clearly, both initial conditions (a) and (b) are satisfied for λ = 1.
The d’Alembert solution is obtained from equation (9.19) as
1
u = [sin(x + ct) + sin(x − ct)]
2
which gives the same result when the sines are expanded.
16 First note that
sin x(1 + cos x) = sin x + 12 sin 2x
The two initial conditions imply that the solution is of the form
u = A sin x sin ct + B sin 2x sin 2ct
and matching the conditions gives A = 1/c and B = 1/4c.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
17
519
The MAPLE implementation is as follows:
f:=(x-c ∗ t)/(1+(x-c ∗ t)ˆ2)+(x+c ∗ t)/(1-(x+c ∗ t)ˆ2);
simplify (f);
# gives the simplification - nearly
simplify(diff(f,x,x)-diff(f,t,t)/c^2);
18
Let
# gives zero as required
denote differentiation with respect to (ct − r) and ‘dot’ with respect to
(ct + r) ; then the terms of the spherically symmetric wave equation are
1
1
utt = [f (ct − r) + g̈(ct + r)]
2
c
r
and
1
1
[f(ct − r) + g(ct + r)] + [−f (ct − r) + ġ(ct + r)]
2
r
r
2
2
= 3 [f(ct − r) + g(ct + r)] − 2 [−f (ct − r) + ġ(ct + r)]
r
r
1
+ [f (ct − r) + g̈(ct + r)]
r
ur = −
urr
Collecting terms together
1
2
1
utt − urr − ur = 3 r2 (f + g̈) − 2(f + g) − 2r(f − ġ )
2
c
r
r
−r2 (f + g̈) + 2(f + g) + 2r(f − ġ) = 0
so the equation is satisfied for any functions f and g. The two terms represent an
outward spherical wave emanating from the origin and an inward wave converging
into the origin. Note the singular behaviour at r = 0.
19
The equation (9.28) is split by the trigonometric formula into two parts
2π2 u
π
π
1
3π
1
3π
= sin (x − ct) + sin (x + ct) − sin (x − ct) − sin (x + ct) . . .
(4l)
l
l
9
l
9
l
1
3π
1
5π
π
sin (x − ct) + . . .]
= [sin (x − ct) − sin (x − ct) +
l
9
l
25
l
π
1
3π
1
5π
+ [sin (x + ct) − sin (x + ct) +
sin (x + ct) + . . .]
l
9
l
25
l
c Pearson Education Limited 2011
520
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The two terms depend on (x−ct) and (x+ct) respectively and represent travelling
waves in the +x and −x directions.
20
The d’Alembert solution is obtained from equation (9.19) as
1
u=
2c
x+ct
x exp(−x2 )dx
x−ct
which on integration gives the quoted result.
21
Again the d’Alembert solution is obtained from equation (9.19) as
u = [F(x − ct) + F(x + ct)]/2
where F is the function given in the exercise.
22
Try a solution of the form u = f(x + ky) , so the equation gives
3f + 6kf + k2 f = 0 ⇒ 3 + 6k + k2 = 0
√
which has solutions k = −3 ± 6 and hence the characteristics are
x + (−3 +
x + (−3 −
23
√
√
6)y = const
6)y = const
Substituting
6f − λf − λ2 f = 0 ⇒ λ = 2, −3
and hence a solution of the form
u = f(x + 2t) + g(x − 3t)
The initial conditions give
x2 − 1 = f(x) + g(x) and 2x = 2f (x) − 3g (x)
Integrating and solving for f and g produces the solution
u=
1
[4(x + 2t)2 + (x − 3t)2 − 5]
5
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
24
521
Differentiating
g
∂u
g
= − 2 cos ωt + cos ωt
∂r
r
r
and
∂2 u
=
∂r2
2g 2g
g
−
+
r3
r2
r
cos ωt
Applying these expressions into the equation gives
2g 2g
2
g
g
g
+
−
−
+
+
r3
r2
r
r
r2
r
cos ωt = −
ω2 g
cos ωt
c2 r
and cancelling produces the equation for g as
ω2
g=0
c2
This simple harmonic equation has sine and cosine solutions which are written in
g +
the form
ω
ω
(b − r) + B sin (b − r)
c
c
The second boundary condition is now satisfied by applying A = 0 and the first
condition gives
ω
B
u(a, t) = β cos ωt = sin (b − a) cos ωt
a
c
and hence B is known and the required solution is
g = A cos
u(r, t) =
25
aβ cos ωt sin ωc (b − r)
r
sin ωc (b − a)
This question is similar to Example 9.11 but initially the velocity is given and
the displacement is zero.
On the initial line t = 0 the solution
u = f(x + t) + g(x − t)
satisfies condition (a) only if f = −g and therefore the condition (b) gives
ut (x, 0) = 2f (x) = exp(−|x|)
Integrate to obtain f
c Pearson Education Limited 2011
522
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
1 x
2e
for x < 0
1−
for x > 0
where it has been arranged that the function goes to zero at infinity and matches
f=
1 −x
2e
at x = 0.
The numerical solution can now be computed from the values on the initial line
given by f. The values at subsequent times t = 0.5, 1.0, 1.5, 2.0, . . . can be computed
easily from
u(x, 0.5) = f(x + 0.5) − f(x − 0.5)
u(x, 1) = f(x + 1) − f(x − 1)
u(x, 1.5) = f(x + 1.5) − f(x − 1.5)
etc.
to give the quoted solution. On a spreadsheet, applying f (x) into column B
corresponding to values of x = −3, −2.5, . . . , 2.5, 3, then a typical entry, which
can be copied onto the other entries in the column,
in column D, D7 reads
+B8 − B6
in column E, E7 reads
+B9 − B5
in column F, F7 reads
+B10 − B4
etc.
It is instructive to derive the exact solution and then compare with the numerical
solution.
u(x, t) =
26
for x < −t
ex sinh t
1 − e−t cosh x for − t < x < t
e−x sinh t
for x > t
From the possible separated solutions, the conditions (a) and (b) imply that
u = cos λct sin λx
is the only one that satisfies these conditions. The condition (c) gives sin λπ =
0 ⇒ λ = N which is an integer. Thus, a superposition of these solutions gives
∞
aN cos Nct sin Nx
u(x, t) =
N =1
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
523
and the condition (d) gives the standard Fourier problem of evaluating the
coefficients in
∞
πx − x =
2
aN sin Nx
N =1
The coefficients are obtained from the usual integral and the result follows by two
integrations by parts
aN =
27
4
(1 − cos Nπ)
πN3
Taking the Laplace transform with respect to t, in equation (9.33) both u(x,0)
and ut (x, 0) are zero from conditions (a) and (b); so the equation is
2d
2
U
= s2 U with solution U = Aesxlc + Be−sxlc
2
dx
From condition (d), the constant A = 0 since the solution must be bounded for all
c
x > 0. The condition (c) is transformed to
U(0, s) =
s2
aω
+ ω2
and hence the solution for U takes the form
aω
e−sxlc
2
+ω
and the exponential just shifts the solution as
U(x, s) =
s2
x
x
H ω t−
u(x, t) = a sin ω t −
c
c
It is easily checked that all the conditions are satisfied by the function. For x > ct
the wave has not reached this value of x, so u = 0 beyond this point.
Exercises 9.3.6
28
The problem is best solved by using MATLAB.
Explicit
n=5;L=0.25;x=[0:1/(n-1):1];z=zeros(1,5);
zz=.25 ∗ [0 .25 .5 .25 0]
c Pearson Education Limited 2011
524
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
zzz=[0,2 ∗ zz([2:n-1])-z([2:n-1])+L ∗ (zz([1:n-2])
-2 ∗ zz([2:n-1])+zz([3:n])),0]
% gives
0
0.1250 0.2188
0.1250
0
z=zz;zz=zzz;
zzz=[0,2 ∗ zz([2:n-1])-z([2:n-1])+L ∗ (zz([1:n-2])
-2 ∗ zz([2:n-1])+zz([3:n])),0]
% gives
0
0.1797
0.2656
0.1797
0
Implicit
n=5;L=0.25;
a=[-L 2 ∗ (1+L)-L];A=eye(n);for i=2:n-1,A(i,i-1:i+1)=a;end
b=[L/2 1-L L/2];C=eye(n);for i=2:n-1,C(i,i-1:i+1)=b;end
u=[0 0 0 0 0]’;v=C ∗ u+.25 ∗ [0 .25 .5 .25 0]’;
B=inv(A);
w=4 ∗ B ∗ v-u;w’
% gives 0 0.1224 0.2245 0.1224 0
u=v;v=w;w=4 ∗ B ∗ v-u;w’
% gives 0 0.1741 0.2815 0.1741 0
29
Again MATLAB is a convenient method for the explicit calculation.
n=6;L=0.01;delt=0.02;
format long
z=eye(1,6);zz=[sin(delt/2 ∗ pi),0 0 0 0 0]
% gives 0.0314107 0 0 0 0 0
zzz=[sin(2 ∗ delt/2 ∗ pi),2 ∗ zz([2:n-1])-z([2:n-1])
+L ∗ (zz([1:n-2])-2 ∗ zz([2:n-1])+zz([3:n])),0]
% gives 0.062790 0.000314 0 0 0 0
z=zz;zz=zzz;
zzz=[sin(3 ∗ delt/2 ∗ pi),2 ∗ zz([2:n-1])-z([2:n-1])
+L ∗ (zz([1:n-2])-2 ∗ zz([2:n-1])+zz([3:n])),0]
% gives 0.094108 0.001249 0.000003 0 0 0
30
Care must be taken to include the ‘+2’ term but the MATLAB
implementation is quite straightforward.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
525
Explicit
n=6;L=0.25;delt=0.2;x=[0:0.2:1];z=x. ∗ (1-x)
zz=[0,(1-L) ∗ z([2:n-1])+L ∗ (z([1:n-2])
+z([3:n]))/2,0]+[0,deltˆ2 ∗ ones(1,4),0]
% gives 0 0.1900
0.2700
0.2700
0.1900
0
zzz=[0,2 ∗ zz([2:n-1])-z([2:n-1])+L ∗ (zz([1:n-2])-2 ∗ zz([2:n-1])
+zz([3:n])),0] +[0,2 ∗ deltˆ2 ∗ ones(1,4),0]
% gives 0 0.2725
z=zz;zz=zzz;
0.3600
0.3600
0.2725
0
zzz=[0,2 ∗ zz([2:n-1])-z([2:n-1])+L ∗ (zz([l:n-2])-2 ∗ zz([2:n-1])
+zz([3:n])),0]+[0,2 ∗ deltˆ2 ∗ ones(1,4),0]
% gives 0 0.3888 0.5081 0.5081 0.3888 0
Implicit
n=6;L=0.25;delt=0.2;x=[0:0.2:1];
a=[-L2 ∗ (1+L)-L];A=eye(n);for i=2:n-1,A(i,i-1:i+1)=a;end
b=[L/2 1-L L/2];C=eye(n);for i=2:n-1,C(i,i-1:i+1)=b;end
u=(x. ∗ (1-x))’;v=C ∗ u+[0;deltˆ2 ∗ ones(4,1);0]
% gives
0 0.1900 0.2700
0.2700 0.1900 0
B=inv(A);
w=B ∗ (4 ∗ v+[0;2 ∗ deltˆ2 ∗ ones(4,1);0])-u
%gives
0
0.2319 0.3191
0.3191 0.2319 0
u=v;v=w;w=B ∗ (4 ∗ v+[0;2 ∗ deltˆ2 ∗ ones(4,1);0])-u
%gives
31
0
0.2785
0.3849
0.3849
0.2785
0
The problem is now more difficult since there is an infinite region. The
simplest way to cope with this difficulty for small times is to impose boundaries
some distance from the region of interest. Hopefully, the effect of any sensible
boundary condition would only affect the solution marginally. For longer times, an
alternative strategy must be sought. In the current problem, the region x = −1
to 2 is chosen with the solution quoted in the region x = 0 to 1.
Explicit
n=16;L=0.25;delt=0.2;
x=[-1:0.2:2];z=x. ∗ (1-x);
c Pearson Education Limited 2011
526
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
zz=[-2,(1-L) ∗ z([2:n-1])+L ∗ (z([l:n-2])
+z([3:n]))/2,-2]+deltˆ2 ∗ ones(1,16)
% gives 0.0300 0.1900 0.2700 0.2700
0.1900
0.0300
zzz=[-2,2 ∗ zz([2:n-1])-z([2:n-1])+L ∗ (zz([1:n-2])-2 ∗ zz([2:n-1])
+zz([3:n])),-2]+2 ∗ deltˆ2 ∗ ones(1,16)
% gives 0.1200 0.2800 0.3600 0.3600
0.2800
0.1200
-0.1200
z=zz;zz=zzz;
zzz=[-2,2 ∗ zz([2:n-1])-z([2:n-1])+L ∗ (zz([1:n-2])-2 ∗ zz([2:n-1])
+zz([3:n])),-2]+2 ∗ deltˆ2 ∗ ones(1,16)
% gives 0.2700 0.4300 0.5100 0.5100 0.4300
0.2700
Implicit
n=16;L=0.25;delt=0.2;
x=[-1:0.2:2];
a=[-L 2 ∗ (1+L)-L];A=eye(n);for i=2:n-1,A(i,i-1:i+1)=a;end
b=[L/2 1-L L/2];C=eye(n);for i=2:n-1,C(i,i-1:i+1)=b;end
u=(x. ∗ (1-x))’;v=C ∗ u+deltˆ2 ∗ ones(16,1)
% gives 0.0300 0.1900 0.2700 0.2700 0.1900 0.0300
B=inv(A);
w=B ∗ (4 ∗ v+2 ∗ deltˆ2 ∗ ones(16,1))-u
%
gives
0.0800 0.2400 0.3200 0.3200
u=v;v=w;w=B ∗ (4 ∗ v+2 ∗ deltˆ2 ∗ ones(16,1))-u
%
gives
0.1495 0.3099 0.3900 0.3900
0.2400
0.0800
0.3099
0.1495
Exercises 9.4.3
32
From the set of separated solutions in equation (9.40), the only ones that
satisfy condition (a) are u = e−αt cos λx and the second condition (b) implies
cos λ = 0 ⇒ λ = n + 12 π where n is an integer.
The third condition (c) can be rewritten as
a
cos
u=
2
3πx
2
+ cos
πx
2
for 0 ≤ x ≤ 1 when t = 0
Thus, the complete solution is
u=
a
[exp(−κπ2 t/4) cos(πx/2) + exp(−9κπ2 t/4) cos(3πx/2)]
2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
33
527
If v = ru , then differentiating produces
vr = u + rur
vrr = 2ur + rurr
and hence urr + 2 urr = 1r vrr .
Applying these expressions into the spherically symmetric heat conduction equation
gives
1
1
vrr =
vt
r
κr
Cancelling out the r, it is seen that v satisfies the standard heat conduction
equation. If v remains bounded it may be noted that u → 0 as r → ∞.
Since u = v/r , the boundary conditions for v are
yields
u(a, t) = T0 −→ v(a, t) = aT0 for t > 0
yields
u(b, t) = 0 −→ v(b, t) = 0 for t > 0
yields
u(r, 0) = 0 −→ v(r, 0) = 0 for a < r < b
The first two of these conditions are satisfied by the given expression
v(r, t) = aT0
b−r
−
b−a
∞
AN e−κλ t sin
2
r − a
N =1
b−a
Nπ
and the third gives the Fourier problem
b−r
=
b−a
∞
AN sin
r − a
N =1
b−a
Nπ
The coefficients can be obtained from integration or from standard tables of Fourier
series as AN = 2/πN .
34
Substituting into the partial differential equation gives the ODE
4ηF + (2 + η)F − αF = 0
and applying F = exp(κη) gives the equation
c Pearson Education Limited 2011
528
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
4ηκ2 + (2 + η)κ − α = 0
which is clearly satisfied by κ = − 14 and α = − 21 and produces the classic similarity
solution.
35
Differentiating
∂u
∂u
= −βf(x) cos(x − βt) and
= f (x) sin(x − βt) + f(x) cos(x − βt)
∂t
∂x
and
∂2 u
= f (x) sin(x − βt) + 2f (x) cos(x − βt) − f(x) sin(x − βt)
∂x2
Putting these expressions into the heat conduction equation and equating the sine
and the cosine terms gives
−βf = 2f and f − f = 0
Both equations can be satisfied only if β = 2 and f = Ae−x ; the solution is then
u = Ae−x sin(x − 2t)
Physically, the slab of material is given an initial temperature of Ae−x sin x , the
temperature is zero at infinity and at the end x = 0 the temperature is periodic
taking the form u(0, t) = −A sin 2t and hence A = −u0 .
36
The suggested substitution gives, on differentiation,
θ − θ0 = ue−ht
θt = (ut − hu)e−ht
θxx = uxx e−ht
Putting the expressions into the given equation and cancelling the exponential gives
ut − hu = κuxx − hu ⇒ ut = κuxx
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
529
and produces the standard equation for u. The term h(θ − θ0 ) is a heat loss term
proportional to the excess temperature over an ambient temperature θ0 ; this is
the usual Newton cooling through a surface.
37
First it is clear that the final steady solution is U = 0. The general separated
solution in equation (9.40) is
u = e−αt (A sin λx + B cos λx) where
λ2 = α/κ
Condition (a) can only be satisfied at x = 0 if A = 0. Condition (b) then implies
that
cos λl = 0 so that λl =
1
n+
2
π
and hence the solution takes the form
2 2
∞
π t
1 πx
1
cos n +
u(x, t) =
an exp −κ n +
2
l2
2
l
n=0
The initial condition given in (c) leads to the Fourier problem of evaluating the
coefficients in the expression
u0
1 x
−
2
l
∞
=
an cos
n=0
1
n+
2
πx
l
These can be evaluated by standard integration or using the standard series
∞
(−1)n
cos
2n + 1
n=0
and
∞
1
cos
2
(2n
+
1)
n=0
1
n+
2
πx
π
= for − l < x < l
l
4
π2
x
1 πx
=
1−
for 0 < x < 2l
n+
2
l
8
l
Thus, the coefficients can be calculated as a combination of the last two expressions
as
an = u0
8
2 (−1)n
−
π2 (2n + 1)2
π 2n + 1
c Pearson Education Limited 2011
530
38
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
At any time t , the sine term ensures that the sum is zero at x = 0 and L so
only the first term survives at the end points, and therefore v = v0 at x = 0 and
v = 0 at x = L. From the basic solutions obtained in the text, or by inspection, it
is clear that the heat equation is also satisfied. The additional condition at t = 0
leads to the Fourier series problem
nπx
x
+
1−
cn sin
L
L
n=1
0 = v0
∞
with the coefficients evaluated from
L
1−
0 = v0
0
nπx
L
x
sin
dx + cn
L
L
2
0
An integration by parts gives cn = − 2v
nπ as required.
39
At the ends of the bar the conditions are
(a) u = 0 at x = 0 for t > 0, (b) u = 0 at x = l for t > 0 and the initial
condition is (c) u = 10 for 0 < x < l at t = 0. From the set of separated solutions
in equation (9.40), the only ones that satisfy condition (a) are u = e−αt sin λx .
The condition (b) then gives sin λl = 0 ⇒ λ =
nx
l
where n is an integer. The
solution is therefore of the form
∞
an exp
u(x, t) =
n=1
−κn2 π2 t
l2
sin
nπx
l
The third condition (c) reduces the problem to a Fourier series, namely
∞
an sin
10 =
nπx
n=1
l
Integrating in the usual way over the interval
l
10 sin
nπx
l
dx =
l
an
2
0
gives an =
20
nπ (1
− cos nπ) and the required result.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
40
531
Taking Laplace transforms of the equation with respect to t leads to
sφ̄ − φ(x, 0) = aφ̄ +
b
s
so using condition (b) the required equation is
φ̄ =
s
b
φ̄ −
a
as
This equation has the obvious particular integral φ̄ = b/s2 and it is convenient
to write the complementary function in terms of sinh and cosh functions. The
solution is thus
b
φ̄ = 2 + A sinh
s
s
s
x + B cosh
x
a
a
The boundary conditions in (a) transform to φ̄(±h, s) = 0 and hence
s
s
h + B cosh
h
a
a
s
s
b
h + B cosh
h
0 = 2 − A sinh
s
a
a
b
0 = 2 + A sinh
s
Clearly, A = 0 and B is easily calculated to give
s
cosh
x
b
as
φ̄ = 2 1 −
s
cosh
ah
To transform back to the real plane needs either some tricky integrations or the
use of advanced tables of Laplace transform pairs. Tables give the solution,
φ
1 2
16h2
2
=
(h − x ) +
b
2a
aπ3
∞
(−1)n
(2n − 1)πx
(2n − 1)2 π2 at
cos
exp −
3
2
(2n − 1)
4h
2h
n=1
Exercises 9.4.5
41 In the explicit formulation equation (9.48), the MATLAB implementation can
be written using the ‘colon’ notation to great effect.
n=6;L=0.5;x=0:0.2:1,u=x.^2
c Pearson Education Limited 2011
532
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
v=[0,L ∗ (u([1:n-2])+u([3:n]))+(1-2 ∗ L) ∗ u([2:n-1]),1]
% gives
0
0.0800
0.2000
0.4000
0.6800
1.0000
u=v;v=[0,L ∗ (u([l:n-2])+u([3:n]))+(1-2 ∗ L) ∗ u([2:n-1]),1]
% gives 0
0.1000
0.2400
0.4400
0.7000
1.0000
Repeating the last line gives successive time steps.
42
Again a MATLAB formulation solves the problem very quickly; lamda (L in
the program) is chosen to be 0.4 and time step 0.05.
Explicit
n=6;L=0.4;u=[0 0 0 0 0 1];
v=[0,L ∗ (u([1:n-2])+u([3:n]))+(1-2 ∗ L) ∗ u([2:n-1]),exp(-0.05)]
% gives 0
0
0
0
0.4000
0.9512
for p=2:20,u=v;v=[0,L ∗ (u([1:n-2])+u([3:n]))+(1-2 ∗ L)
∗
u([2:n-1]),exp(-p ∗ 0.05)];end
v
% gives the values at t = 1
as 0 0.1094 0.2104 0.2939 0.3497 0.3679
Repeating the last two lines produces the solution at successive times.
Implicit
There are some slight differences in the solution depending on how the right hand
boundary is treated. Equation (9.49) is constructed in MATLAB again using the
‘colon’ notation
L=0.4;M=2 ∗ (1+L);N=2 ∗ (1-L);
n=6;u=zeros(n,1);u(n)=1;
p=[-L M -L];A=eye(n);for i=2:n-1,A(i,i-1:i+1)=p;end
q=[L N L];B=eye(n);for i=2:n-1,B(i,i-1:i+1)=q;end
DD=inv(A) ∗ B;
v=DD ∗ u;v(n)=exp(-0.05); % for first step
for p=2:20,u=v;v=DD ∗ u;v(n)=exp(-p ∗ 0.05);end
% gives for t =1
0 0.1082 0.2096 0.2955 0.3551 0.3679
Repeat the last line of code for further time steps.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
43
533
The equations are easily produced in MATLAB. Because of the derivative
boundary condition, the region is extended to x = −0.2 and u(−0.2, t) is obtained
from
u(0.2, t) − u(−0.2, t) = 0.4
L=0.5;M=2 ∗ (1+L);N=2 ∗ (1-L);n=7;
p=[-L M -L];A=eye(n);for i=2:n-1,A(i,i-1:i+1)=p;end
A(1,3)=1;A(1,1)=-1 % gives LHS matrix
q=[L N L];B=eye(n);for i=2:n-1,B(i,i-1:i+1)=q;end
B(1,1)=0
% gives RHS matrix
rhs=[0.4 0 0 0 0 0 0]’;
% gives vector from derivative condition at x=0
AA=inv(A);
x=0:0.2:1,u=[-0.24,x. ∗ (1-x)]’ % starting data
v=AA ∗ (B ∗ u+rhs)
% produces next time step
-0.2800 -0.0400 0.1200 0.2002 0.2012 0.1269 0
u=v;v=AA ∗ (B ∗ u+rhs)
% produces next time step
-0.3197 -0.0799 0.0803 0.1613 0.1657 0.1034 0
Repeating the last line produces further time steps.
Exercises 9.5.2
44
From equation (9.52), the only separated solution to satisfy u → 0 as y → ∞
is
u = (A sin μx + B cos μx)(Ceμy + De−μy ) with C = 0
Thus
u = (a sin μx + b cos μx)e−μy
To satisfy the boundary conditions,
u = 0 at x = 0 ⇒ b = 0
u = 0 at x = 1 ⇒ sin μ = 0 ⇒ μ = nπ where n is an integer.
The condition at y = 0 can be satisfied by a sum of terms over n.
c Pearson Education Limited 2011
534
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
On y = 0, u =
∞
an sin nπx =
n=1
1
16 (10 sin πx
− 5 sin 3πx + sin 5πx)
and the an can be obtained by inspection to give the required solution.
45
The four boundary conditions are satisfied by inspection and the Laplace
equation is satisfied by straightforward differentiation.
46 It can easily be checked that the function x2 y satisfies the given Poisson
equation. The boundary conditions on u(x, y) become
u(x, 0) = 0 for 0 ≤ x ≤ 1
u(x, 1) = sin πx for 0 ≤ x ≤ 1
u(0, y) = 0 for 0 ≤ y ≤ 1
u(1, y) = 0 for 0 ≤ y ≤ 1
The only solution in equation (9.52d) that satisfies these conditions is
u = sin πx
sinh πy
sinh π
and hence the final result
φ = x2 y + sin πx
47
sinh πy
sinh π
Differentiating
ur = Bnrn−1 sin nθ
urr = Bn(n − 1)rn−2 sin nθ
uee = −Bn2 rn sin nθ
and substitution gives
LHS = B sin nθrn−2 [n(n − 1) + n − n2 ] = 0 = RHS
and hence the Laplace equation in plane polars is satisfied. To be periodic in θ
the constant n must be an integer. A solution of the equation is a sum of the
expressions given, so that
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
535
∞
Bn rn sin nθ
u(r, θ) =
n=1
Putting the condition on the rim, r = a, gives the Fourier problem to calculate
Bn as B1 = 3/4a and B3 = −1/4a2 and otherwise zero. Thus,
1 r 3
3 r
sin θ −
sin 3θ
u(r, θ) =
4 a
4 a
48
Let D = x2 + y2 + 2x + 1; then the derivatives can be computed as
ux =
2y(2x + 2)
4y
4y(2x + 2)2
and
u
=
−
xx
D2
D2
D3
−2 2y2y
4y
8y
4y2 4y
and
u
=
+
−
+
yy
D
D2
D2
D2
D3
Adding the two second derivatives gives
uy =
1
2
2
2
3
16y(x
+
y
+
2x
+
1)
−
16y(x
+
1)
−
16y
D3
The RHS can easily be checked to be zero and hence the Laplace equation is
∇2 u =
satisfied. A similar process shows that v also satisfies the Laplace equation.
The u and v come from the complex variable expression
u + jv =
j(x − 1 + jy)(x + 1 − jy)
j(x − 1 + jy)
=
x + 1 + jy
(x + 1 + jy)(x + 1 − jy)
Multiplying out
u + jv =
−2y + j(x2 + y2 − 1)
x2 + y2 + 2x + 1
gives the expressions quoted in the question.
A check can be made by using MAPLE.
u:=2 ∗ y/(xˆ2+yˆ2+2 ∗ x+1);
simplify(diff(u,x,x)+diff(u,y,y));
# gives zero as required -v follows similarly
h:=I ∗ (x+I ∗ y-1)/(x+I ∗ y+1);
simplify(evalc(h));
# gives the u and v of the question
c Pearson Education Limited 2011
536
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
For fixed u and v the two expressions can be rearranged as
2
v 2
1
1
1
2
2
x+
+y =
and (x + 1) + y +
= 2
2
v−1
(v − 1)
u
u
which are circles with radii
1
v−1
and
1
u
−v
, centres ( v−1
, 0) and (−1, −1
u ) respectively.
Note that all the circles pass through the point ( −1,0).
49
This is an important example that illustrates that sensible solutions can only
be obtained if correct boundary conditions are set. First, it is a matter of simple
differentiation to verify that the given function satisfies the Laplace equation.
Again, since the sinh function is zero at x = 0 the first condition is satisfied.
Differentiating with respect to x,
1
∂u
1
∂u
= cosh nx sin ny, so at x = 0,
= sin ny
∂x
n
∂x
n
The solution therefore satisfies all the conditions of the problem. It is known that
the solution is unique.
For any given n, however large, sinh nx can be made as large as required and even
when divided by n2 it is still large; for instance, n = 10, x = 5 and y = π/200
gives u = 4.1 × 1018 .
The ‘neighbouring’ problem has a boundary condition ux = 0 and solution u
identically zero. For the values chosen for illustration, the maximum change at
the boundary is 0.1; yet the solution changes by 1018 . Such behaviour is very
unstable; these boundary conditions give a unique solution; yet small changes in
the boundary produce huge changes in the solution. Figure 9.59 should be referred
for a summary of the ‘correctness’ of boundary conditions.
50
It is useful in solution by separation to try to modify the problem so that the
function is zero on two opposite boundaries. Apply u = x + f(x, y) ; then f satisfies
the Laplace equation and the four boundary conditions become
f(0, y) = 0
f(1, y) = 0
for 0 < y < 1
f(x, 0) = −x f(x, 1) = 1 − x for 0 < x < 1
The solution given in equation (9.52d) is the appropriate one and the cosine can
be omitted since it cannot satisfy the first of the four conditions. Thus,
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
537
f = sin μx(a cosh μy + sinh μy)
The second condition now gives sin μ = 0 ⇒ μ = Nπ where N is an integer. The
solution therefore takes the form
∞
sin Nπx(aN cosh Nπy + bN sinh Nπy)
f=
N =1
and the coefficients are derived as Fourier series from the other two sides of the
boundary as
∞
−x =
aN sin Nπx
N =1
∞
1−x=
sin Nπx(aN cosh Nπ + bN sinh Nπ)
N =1
Straightforward integration gives
aN =
2 cos Nπ
2
and (aN cosh Nπ + bN sinh Nπ) =
Nπ
Nπ
The final solution is obtained by substituting back
∞
1
2
sinh Nπy
u=x+
sin Nπx cos Nπ cosh Nπy + (1 − cos Nπ cosh Nπ)
π
N
sinh Nπ
N =1
which can be tidied up to the given solution.
51
and
The boundary conditions on the four sides are
u(0, y) = 0
u(a, y) = 0
u(x, 0) = 0
u(x, a) = u0
for
0 0 where the equation
is elliptic. (b) y = 0 where the equation is parabolic and (c) y < 0 where the
equation is hyperbolic.
From equation (9.83), the characteristics are obtained from the solution of the
equation
√
−y
dy
=±
dx
y
The equation only makes sense in the hyperbolic region where y < 0. Substitute
z = −y and the differential equation becomes
√
z
1
dz
=±
= ±√
dx
z
z
which is easily integrated to
3
3
3
3
z 2 = ± x + K ⇒ (−y) 2 ± x = K
2
2
as the characteristics.
72
Differentiating
2B
B
2
3
fx = 3Ax − 3 y(1 − y ) and fy = Ax + 2 (1 − 3y2 )
x
x
B
B
2
3
fxx = 6 Ax + 4 y(1 − y ) and fyy = Ax + 2 (−6y)
x
x
2
The given equation can be checked by substitution.
Now ‘AC − B2 ’ = x2 (1 − y2 ) so
⎧
⎨ elliptic if |y| < 1
parabolic if x = 0 or y = ±1
⎩
hyperbolic if |y| > 1
c Pearson Education Limited 2011
558
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
73
Calculating ‘AC − B2 ’ = −4p2 +4q2 , it can be deduced that the equations are
(a) p > q or p < −q then the equations are hyperbolic
(b) p = q then the equations are parabolic
(c) −q < p < q then the equations are elliptic
Using the substitution −p2 +q2 = − 41 (x4 −2x2 y2 +y4 −x4 −2x2 y2 −y4 ) = x2 y2 > 0,
hence leads to the elliptic region when it is expected that the equation will look
like the Laplace equation.
vxx = vp + vq + x2 (vpp + 2vpq + vqq )
vx = xvp + xvq
vy = −yvp + yvq
and
vyy = −vp + vq + y2 (vpp − 2vpq + vqq )
vxy = xy(−vpp + vqq )
It may be noted that
vxx + vyy = 2vq + (x2 + y2 )vpp + 2(x2 − y2 )vpq + (x2 + y2 )vqq
from which the required transformation to the Laplace follows immediately.
74
In this case, ‘AC − B2 ’ = −(xy)2 so the equation is hyperbolic away from
the axes. The characteristics are computed from equation (9.83) as
dy
=±
dx
x2 y 2
y
=±
2
x
x
and the solution is obtained by integration as
ln y = ± ln x + K ⇒ y = ax and yx = b
so one set of characteristics are straight lines through the origin and the other are
rectangular hyperbolas. The domain of dependence and the range of influence can
now be sketched.
Review exercises 9.11
1
The boundary and initial conditions on y(x, t) are
y=0
String initially at rest
Initial displacement
⇒
∂y
∂t
=0
y = f(x) =
at
x=0
at
t=0
for 0 ≤ x ≤ b
at t = 0
for b ≤ x ≤ a
εx
b
ε(a−x)
a−b
and a
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
559
Out of the possible separated solutions (9.25) that satisfy the wave equation, the
only one that satisfies the conditions at x = 0 and y = 0 is y = sin λx cos λct .
The condition at x = a gives λa = nπ where n is an integer. Thus, the solution is
a sum of such terms
∞
An sin
y=
nπx
n=1
a
cos
nπct
a
and the final condition produces the Fourier series
∞
f(x) =
An sin
nπx
a
n=1
The coefficients are evaluated from the integral
1
aAn =
2
b
nπx
εx
sin
dx +
b
a
0
a
nπx
ε(a − x)
sin
dx
a−b
a
b
which, after some careful integration by parts, gives the required coefficient
2εa2
An = 2 2
sin
n π b(a − b)
nπb
a
It is interesting to look at the solution for various values of b since the solution gives
strengths of the harmonics for different musical instruments. It is these values that
give the characteristic sound of the instrument. For example, a violin has b = a/7;
it is seen that A7 = 0 and sevenths do not occur for this instrument.
2 Taking the Laplace transform of the equation and the boundary conditions
gives
φ̄ = s2 φ̄ − sx2
2l
s
The most convenient form for the complementary function is
and
φ̄(0, s) = 0, φ̄ (l, s) =
φ̄ = A cosh s(x − l) + B sinh s(x − l)
The particular integral is a quadratic in x which when substituted into the equation
gives
c Pearson Education Limited 2011
560
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
particular integral =
x2
2
+ 3
s
s
The complete solution is therefore
φ̄ = A cosh s(x − l) + B sinh s(x − l) +
2
x2
+ 3
s
s
Putting the two conditions on φ̄ into the solution
φ̄ (0, s) = 0 ⇒ 0 = A cosh(−sl) + B sinh(−sl) +
2
s3
2l
2l
2l
⇒
= sB +
⇒B=0
s
s
s
The Laplace transform of the solution is
φ̄ (l, s) =
2
x2
2 cosh s(x − l)
+ 3− 3
s
s
s
cosh sl
The solution in real space can be obtained from advanced tables of transform pairs.
φ̄ =
3
Take the separated solutions that are quoted and first note that the conditions
y(0) = y(l) = 0 are satisfied. Secondly, substitute into equation (9.92)
1
c2
Tn
1
+ Tn
τ
= −Tn
nπ 2
l
This equation can then be solved in the standard way by looking for solutions of
the form Tn = eαt which produces the quadratic equation
cnπ 2
1
α2 + α +
=0
τ
l
The equation has roots
⎞
⎛
#
#
2
2
cnπ 2
cnπ
1
1⎝ 1
l
1
⎠=− ±j
α=
−4
1−
− ±
2
τ
τ
l
2τ
l
2τcnπ
and hence the solution is
t
Tn (t) = exp −
2τ
(an cos ωn t + bn sin ωn t)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
and general solution
∞
y=
Tn (t) sin
n=1
The condition
∂y
∂t (x, 0)
561
nπx
l
= 0 implies, on differentiation and substituting t = 0,
1
an + bn ωn
2τ
From the other condition it is seen that the only term to survive is when n = 3.
0=−
Thus,
4 sin
3πx
l
0
= an exp −
2τ
1
3πx
cos ω3 0 +
sin ω3 0 sin
2τω3
l
and hence the solution satisfying all the conditions is
t
y(x, t) = 4exp −
2τ
1
3πx
cos ω3 t +
sin ω3 t sin
2τω3
l
and ω3 given in the question.
4
This problem is the extension of the wave equation to beams.
Simply
substituting the given form into the beam equation gives the equation (9.93) for V
and the end conditions follow immediately. Again simply substituting sin, sinh, cos
and cosh into the equation shows they are solutions of (9.93). A linear combination
of the functions is also a solution and since it contains four arbitrary constants, it
is the general solution.
To satisfy the end conditions
V(0) = 0 ⇒ A + B = 0
V (0) = 0 ⇒ C + D = 0
V(l) = 0
⇒
A cosh αl + B cos αl + C sinh αl + D sin αl = 0
V (l) = 0
⇒
A sinh αl − B sin αl + C cosh αl + D cos αl = 0
Thus,
A(cosh αl − cos αl) = D(sinh αl − sin αl)
A(sinh αl + sin αl) = D(− cos αl + cosh αl)
c Pearson Education Limited 2011
562
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Dividing these two equations, and using the trigonometric and hyperbolic identities,
produces the equation
cos αl cosh αl = 1
from which the natural frequencies of vibration of the beam can be calculated.
5 The separated solutions in equation (9.40) that satisfy the condition at x = 0
are θ = e−αt cos λx . To satisfy the condition at x = l requires that
λl =
1
n+
2
π where n is an integer
Thus, the solution takes the form
2
(2n + 1)π
(2n + 1)πx
θ(x, t) =
exp −
A2n+1 cos
t
2l
2al
n=0
∞
and the initial condition now produces the usual Fourier series problem with
1
A2n+1 =
2
l
f(x) cos
(2n + 1)πx
dx
2l
0
Integration by parts is required to obtain the coefficient for the given function
f(x) = θ0 (l − x) as
A2n+1 =
8θ0 l
+ 1)2
π2 (2n
and the temperature can be obtained from the series solution.
6 Evaluating the partial derivatives
∂φ
∂φ
∂2 φ
1
1
1 x
= f
= √ f ,
= − 3/2 f and
2
∂x
∂t
2t
∂x
t
t
Substituting the derivatives back into the equation
1 x
1
1
κ f = − 3/2 f ⇒ f = − zf
t
2t
2κ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
563
the problem has been reduced to an ordinary differential equation. Rearranging
the equation
df
1
z2
=
−
=
−
zdz
which
integrates
to
ln
f
+C
f
2κ
4κ
2
z
which on further integration gives
or f = A exp − 4κ
z
f=A
2
v
dv + B
exp −
4κ
0
Substitute u =
z
√
2 κ
f=a
&
v
√
2 κ
then the equation reduces to
e−u2 du + b where a and b are new arbitrary constants.
0
The solution is as required
f = a erf
z
√
2 κ
+b
For the particular problem,
at t = 0
φ(x, 0) = 0
for all x > 0
at x = 0
φ(0, t) = φ0
for t > 0
In the expression containing the error function, these conditions give respectively
0 = a erf(∞) + b and φ0 = a erf(0) + b
Since erf(0) = 0 and erf (∞) = 1, the solution can be constructed as
T(x, t) = T0 + φ0 1 − erf
7
x
√
2 κt
The problem is standard except for the treatment of the derivative boundary
conditions. Note the way that they are handled in the MATLAB implementation.
c Pearson Education Limited 2011
564
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Explicit
u=[1 1 1 1 1 1];
u=0.1 ∗ [u([2:6]),u(5)-0.4 ∗ u(6)]+0.8 ∗ u([1:6])
+.1 ∗ [u(2)-0.4 ∗ u(1),u([1:5])]
% gives 0.9600 1.0000 1.0000 1.0000 1.0000 0.9600
u=0.1 ∗ [u([2:6]),u(5)-0.4 ∗ u(6)]
+0.8 ∗ u([1:6])+.1 ∗ [u(2)-0.4 ∗ u(1),u([1:5])]
% gives 0.9296 0.9960 1.0000 1.0000 0.9960 0.9296
u=0.1 ∗ [u([2:6]),u(5)-0.4 ∗ u(6)]+
0.8 ∗ u([1:6])+.1 ∗ [u(2)-0.4 ∗ u(1),u([1:5])]
% gives 0.9057 0.9898 0.9996 0.9996 0.9898 0.9057
Repeating the last line of code produces subsequent time steps.
Implicit
A=[-2.4 2 0 0 0 0;1 -2 1 0 0 0;0 1 -2 1 0 0;0 0 1 -2 1 0;
0 0 0 1 -2 1;0 0 0 0 2 -2.4]
u=[1;1;1;1;1;1];
B=2 ∗ eye(6)-0.1 ∗ A
C=2 ∗ eye(6)+0.1 ∗ A
E=inv(B) ∗ C
u=E ∗ u
% gives 0.9641 0.9984 0.9999 0.9999 0.9984 0.9641
u=E ∗ u
% gives 0.9354 0.9941 0.9996 0.9996 0.9941 0.9354
u=E ∗ u
% gives 0.9120 0.9881 0.9988 0.9988 0.9881 0.9120
Repeating the last line of code produces subsequent time steps.
8 From the possible solutions in equation (9.52), one that can be chosen to satisfy
the conditions on x = 0 and y = a is
u = sin μx sinh μ(a − y)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
565
On x = a, u = 0 ⇒ sin μa = 0 ⇒ μa = nπ where n is an integer. The final
condition on y = 0 can be satisfied by taking a sum of such solutions
∞
x(a − x) =
bn sin
nπx
n=1
a
sinh nπ
The problem is the usual evaluation of the Fourier coefficients
a
bn sinh nπ =
2
a
(a − x) sin
nπx
a
dx
0
and after two integrations by parts gives
a 3
a
bn = 2
(1 − cos nπ)
2
nπ
All the even terms go to zero and putting n = 2r + 1, the expression quoted is
recovered.
9 This exercise is a harder one since the regular mesh points do not lie on the
boundary.
y
y2 = x
c
7
b
a
4
5
6
1
2
3
x
At a point such as node 4 the lengths of the mesh are not uniform, so the second
derivative needs to be approximated as
c Pearson Education Limited 2011
566
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
f =
f3 − f0
Δx
−
f0 − f1
Δx
2
Δx + Δx
for a typical configuration
x
x′
1
0
3
The curve does not pass through the mesh points and the lengths are calculated
√
√
as a = 0.25, b = 0.5 − 0.5 and c = 1.5 − 1. Working through the equations one
at a time, using this formula where appropriate
1 + u2 + 2u4 − 4u1 = 0
u1 + u3 + 2u5 − 4u2 = 0
2
0.5 + a
u2 + 1 + 2u6 − 4u3 = 0
u5 − u4
1 − u4
2
1 − u4
u1 − u4
+
+
+
= 0.5(0.5)2
0.5
a
0.5 + b
b
0.5
u4 + u2 + 1 + u6 − 4u5 = 0.0625
u5 + u3 + u7 + 1 − 4u6 = 0.09375
1 − u7
u6 − u7
1 + 1 − 2u7
2
+
= 1.5
+
0.52
c + 0.5
c
0.5
The equations can be transformed to matrix form as
⎡
−4
⎢ 1
⎢
⎢ 0
⎢
a = ⎢ 5.6569
⎢
⎢ 0
⎣
0
0
bT = [ −1
1
−4
1
0
1
0
0
0
0
1
−4
0
0
1
0
−1
2
0
0
−35.3137
1
0
0
−24.1985
0
2
0
5.3333
−4
1
0
−0.9375
0
0
2
0
1
−4
5.5192
−0.9063
⎤
0
0
⎥
⎥
0
⎥
⎥
0
⎥
⎥
0
⎥
⎦
1
−25.7980
−18.7788 ]
and the solution of Ax = b can be obtained from any package, for example
MATLAB, as
xT = [ 0.9850
0.9648
0.9602
0.9876
0.9570
c Pearson Education Limited 2011
0.9380
0.9286 ]
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
567
10 Clearly, at z = 0 the velocity u = U cos ωt so the given solution has the correct
velocity on the wall. It also satisfies the equation of motion since substituting the
derivatives
ut = −Uωe−αz sin(ωt − αz)
uz = −αUe−αz cos(ωt − αz) − αUe−αz sin(ωt − αz)
uzz = α2 Ue−αz cos(ωt − αz) − 2α2 Ue−αz sin(ωt − αz)
−α2 Ue−αz cos(ωt − αz)
into the equation gives −Uω = −ν2α2 U which agrees precisely with the definition
of α.
11
Differentiating
2
−1
−r2
r
+t
exp −
Ut = kt
2
4
t
4t
2
r
−2r
k
exp
Ur = t
4t
4t
2
∂ 2
r
−2r
−tk−1
2
3
(r Ur ) =
3r + r
exp
∂r
2
4t
4t
k−1
r2
exp −
4t
k
and applying into the spherically symmetric heat equation
tk−1
−
2
r2
3−
2t
2
2
r
r
r2
k−1
exp −
=t
exp −
k+
4t
4t
4t
gives the relation k = −3/2.
12 The equation is the same as Example 9.7 and can be dealt with in the same way,
see also equation (9.14). However, the MAPLE solution is very straightforward.
with (PDEtools):
rev 12:=diff(z(x,y),x) + diff(z(x,y),y);
sol:=pdesolve (rev12,z(x,y));
# gives the solution
sol:=z(x,y)=_F1(y-x)
c Pearson Education Limited 2011
568
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The boundary conditions x = s, y = −s, z = 2s for s > 0 give the solution
z(x, y) = x − y
for
x>y
and not defined otherwise.
13
The separated solutions in equation (9.52) that tend to zero for large y must
be of the form
φ = (A sin μx + B cos μx)e−μy
The condition φ(0, y) = 0 ⇒ B = 0 and φ(π, y) = 0 ⇒ sin μπ = 0; thus, the
required solution is
∞
φ=
cn e−ny sin nx
n=1
Yet again, the final condition requires the evaluation of the coefficients by Fourier
analysis. Integration by parts gives
π
cn =
2
π
x(π − x) sin nxdx =
2
(1 − cos nπ)
n3
0
When n is even, the coefficient is zero and the odd values give the result in the
exercise.
14
First observe that the function χ = a2 − x2 satisfies ∇2 χ = −2 and second
that a separated solution of the Laplace equation derived in equation (9.52) is
cosh μy cos μx and the overall solution is a combination of terms of these types.
The conditions that χ = 0 on x = ±a ⇒ cos μa = 0 ⇒ μa = n + 12 π where n is
an integer. Thus, an appropriate solution is
∞
χ(x, y) = a − x +
2
2
A2n+1 cosh
n=0
1
n+
2
y
π cos
a
1
n+
2
π
x
a
The function is even, so only the condition at y = b needs to be considered since
the other boundary is satisfied by symmetry. Therefore,
∞
0=a −x +
2
2
A2n+1 cosh
n=0
1
n+
2
b
π cos
a
c Pearson Education Limited 2011
1
n+
2
π
x
a
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
569
and the coefficients are derived by Fourier analysis. From tables of Fourier series
it may be noted that
(−1)n
π3
πx
=
cos
(2n
+
1)
(a2 − x2 ) for − a < x < a
3
2
(2n
+
1)
2a
32a
n=0
∞
and thus the coefficient can be identified as
A2n+1 =
(−1)n+1
32a2
π3 (2n + 1)3 cosh (2n+1)πb
2a
15
The possible separated solutions of the wave equation are given in equation
(9.25) and to satisfy conditions (a) and (b), the solution sin λx cos λct must be
chosen. The condition at x = 1 implies that sin λ = 0 ⇒ λ = nπ where n is an
integer. Thus, the solution takes the form
∞
an sin nπx cos nπt
u(x, t) =
n=1
and condition (c) is substituted to give the Fourier series
∞
1−x=
an sin nπx
n=1
The coefficients are obtained from
1
an =
2
1
(1 − x) sin nπxdx
0
which can be integrated by parts to give an =
2
nx
and agrees with the quoted
answer.
16
Complete drainage at top and bottom implies u = 0 at z = 0 and z = h .
Since there are no sources the pressure tends to zero as time becomes infinite.
Thus the solution that is relevant to this problem is u = e−cα t sin αz . It is readily
2
checked that this function is a solution of the consolidation equation. The only
c Pearson Education Limited 2011
570
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
boundary condition not satisfied is at z = h ; to do this choose αh = mπ where m
is an integer. Therefore,
∞
an sin
u(z, t) =
mπz
h
m=1
cm2 π2 t
exp −
h2
and the initial uniform pressure leads to the Fourier series problem of evaluating
the coefficients in
∞
A=
am sin
mπz
h
m=1
so
h
am =
2
h
A sin
mπz
h
dz =
Ah
(1 − cos mπ)
mπ
0
For m, even the coefficient is zero and substituting m = 2n+1 provides the solution
quoted in the exercise.
17
Substituting φ = X(x)T(t) into the equation
1 T̈ + KṪ
X
= 2
= −λ2
X
c
T
where the separation constant has been chosen to be negative, to ensure periodic
solutions for X . The variable X satisfies
X + λ2 X = 0 with solution
X = P cos λx + Q sin λx
From condition (a) the sine term must be zero so Q = 0. For the T equation,
T̈ + KṪ + c2 λ2 T = 0
so trying a solution of the form T = exp(at) gives the quadratic
1
2
2
−K ± K − 4(cλ)
a + Ka + (cλ) = 0 with solution a =
2
2
2
The constant λ can be identified as p, so the condition that (cp)2 > 14 K2 gives an
overall solution of the type
Kt
φ = cos px exp −
2
(M sin bt + N cos bt)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
571
where b2 = (cp)2 − 14 K2 .
Condition (a) requires that N = A
Condition (b) requires the derivative
K
Kt
∂φ
= − cos px exp −
(M sin bt + N cos bt)
∂t
2
2
Kt
(M cos bt − N sin bt)
+ b cos px exp −
2
At x = t = 0
1
1
− AK = − NK + bM
2
2
and since N = A then M = 0. The required solution is therefore
Kt
φ(x, t) = A cos px exp −
2
1
cos bt where b2 = (cp)2 − K2
4
Results can be checked easily in MAPLE.
f:=cos(p ∗ x) ∗ exp(-k ∗ t/2) ∗ cos(b ∗ t);
simplify(diff(f,x,x)-(diff(f,t,t)+k ∗ diff(f,t))/cˆ2);
gives
−4p2 c2 + k2 + 4b2
1
cos(px) exp − kt cos(bt)
4c2
2
18 Substituting the expressions for vr and vθ into the continuity equation, it is
satisfied for any stream function
∂
LHS =
∂r
∂ψ
∂θ
∂
+
∂θ
∂ψ
−
= 0 = RHS
∂r
since the cross partial derivatives are equal for all differentiable functions. For the
given stream function,
a2
a2
1
= U cos θ 1 − 2
vr = U cos θ r −
r
r
r
a2
vθ = −U sin θ 1 + 2
r
c Pearson Education Limited 2011
572
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
On the circle r = a, the radial velocity vr = 0 and there is no flow into the circle.
As r gets very large vr → U cos θ and vθ → −U sin θ
The velocities parallel and perpendicular to the axes are
V = vr cos θ − vθ sin θ
y
→U
and
vθ
W
vr
W = vr sin θ + vθ cos θ
→0
r
V
θ
x
and hence the flow far downstream is a
steady flow in the x direction with
velocity U. Overall the flow represents
inviscid, irrotational flow past a circular
obstacle in uniform flow.
19, 20, and 21 are intended to be open exercises extending the work of the chapter
to more investigative work. Consequently, no advice is offered for these problems.
For Exercises 19 and 20 the text quotes sources for the work and for Exercise 21
many books on heat transfer will have a version of this problem.
c Pearson Education Limited 2011
10
Optimization
Exercises 10.2.4
1
The required region is shaded on the graph and the lines of constant cost are all
parallel to the lines labelled f = 9 and f = 12. The point where f is a maximum
in the region is at the point x = 1 and y = 1 giving a maximum cost of f = 9.
3.0
y
2.0
f = 12
f=9
1.0
3x + 7y = 10
2x + y = 3
0.5
2
1.0
1.5
x
A graphical or a tabular solution is possible but the MAPLE solution is given
here.
with(simplex):
con2:={2 ∗ x-y<=6,x+2 ∗ y<=8,3 ∗ x+2 ∗ y<=18,y<=3};
obj2:=x+y;
maximize(obj2,con2,NONNEGATIVE);
# gives the solution {y=2, x=4}
3
Let x be number of type 1 and y the number of type 2. The profit from these
numbers is
f = 24x + 12y
c Pearson Education Limited 2011
574
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
T2
5x + 2y = 200
f = 1320
80
60
4x + 5y = 400
40
5x + 3y = 250
f = 840
20
f = 1080
10
20
30
T1
40
T2
f = 1320
80
60
4x + 5y = 400
min at (5,75)
40
5x + 3y = 250
5x + 2y = 175
20
f = 1020
10
20
30
T1
c Pearson Education Limited 2011
40
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
575
and the constraints are, in the appropriate units,
chipboard
4x + 5y ≤ 400
veneer
5x + 2y ≤ 200
labour
5x + 3y ≤ 250
and the obvious constraints x ≥ 0 and y ≥ 0. The first figure shows the feasible
region bounded by the axes and the three lines 4x + 5y = 400, 5x + 2y = 200 and
5x + 3y = 250. These lines intersect at x = 20, y = 50 and give the optimum profit
of £1080.
Reducing the available amount of oak veneer to 175 m , the diagram changes as in
the second figure. It can be seen that the same two constraints are active. They
give the solution x = 5, y = 75 and a reduced optimum profit of £1020.
4
Let n and s be the number of kg of nails and screws respectively. The profit
made is therefore z = 2n + 3s
The constraints are
labour
3n + 6s ≤ 24
material
2n + s ≤ 10
The figure shows the feasible region. The point of intersection of the two constraints
gives the maximum profit of 14p making 4 kg of nails and 2 kg of screws.
screws(Kg)
5
profit = 14
2n + s = 10
3
profit = 6
3n + 6s = 24
1
2
4
6
nails(Kg)
c Pearson Education Limited 2011
576
5
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Let C1 and C2 be the number of cylinders CYL1 and CYL2 produced. The
profit is 4C1+3C2 and the constraints given by the availability of the materials are:
M1
C1 + 5C2 ≤ 45
M2
C1 + 2C2 ≤ 21
M3
2C1 + C2 ≤ 24
It is clear from the figure that the optimum is z = 54 with C1 = 9 and C2 = 6.
The constraints M2 and M3 are active so all these materials are used up. The
constraint M1 has some slack, it may be checked that 6 units remain unused.
C2
2C1 + C2 = 24
12
profit = 36
10
8
C1 + 5C2 = 45
6
C1 + 2C2 = 21
4
profit = 54
2
2
6
4
6
8
10
12
C1
Let y be the number of Yorks and w the number of Wetherbys, then the profit
made is
z = 25y + 30w
The constraints are
cloth
3y + 4w ≤ 400
labour
3y + 2w ≤ 300
and the problem is a straightforward LP problem that can be solved graphically.
It can be seen from the figure that the optimum is at y = 66.67 and w = 50
with a profit of £3166.67. Note that the solution must be integral to make sense,
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
577
two-thirds of a jacket is not much good to anyone. However, the solution is an
approximate one and, to proceed correctly, it is necessary to undertake the problem
as an Integer Programming problem. This is much harder and can be found in any
advanced book on Mathematical Programming.
If the amount of cloth is increased then the line 3y + 4w = C is moved from its
present C = 400 parallel and upwards. The intersection point moves upwards and
the profit is increased. This situation continues until the solution is y = 0 and
w = 150 and all the labour is used on the Wetherbys. The amount of cloth required
would be 600 m and the profit in this case is £4500. This is the maximum possible
profit even if unlimited cloth is available because of the labour constraint.
160
140
120
100
3y + 2w = 300
f = 4500
f = 3000
3y + 4w = 400
w
80
60
40
20
f = 1500
0
−20
−40
7
0
10
20
30
40
50
y
60
70
80
90
100
The initial tableau is
z
x3
x4
x1
−k
1
3
x2
−20
2
1
x3
0
1
0
x4
0
0
1
Soln
0
20
25
x3
10
0.5
−0.5
x4
0
0
1
Soln
200
10
15
Eliminating the x2 column first
z
x2
x4
x1
10−k
0.5
2.5
x2
0
1
0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
578
If k < 10 then the tableau is optimal and the solution is x1 = 0, x2 = 10 and
the value of z is 200. If, however, k > 10 then the method continues with the x1
column cleared.
x1
0
0
1
z
x2
x1
x2
0
1
0
x3
(60 − k)/5
0.6
−0.2
x4
2(k − 10)/5
0.2
0.4
Soln
140 + 6k
7
6
If 10 < k < 60 then the solution is optimal and the solution is x1 = 6, x2 = 7 and
z takes the value 140 + 6k . If, however, k > 60 then the solution is not optimal
and a further tableau needs to be formed using the x3 column.
x1
0
0
1
z
x3
x1
x2
−(60 − k)/5
5/3
1/3
x3
0
1
0
x4
(7k − 120)/15
1/3
7/15
Soln
25k/3
35/3
25/3
For k > 60, the z row is positive and therefore optimal with x1 = 25/3, x2 = 0
and z = 25k/3.
x2
3x1 + x2 = 25
10
A
optimum for k = 90
8
optimum for k = 5
B
6
x1 + 2x2 = 20
4
optimum for k = 30
2
C
2
4
6
8
10
12
x1
The situation for this problem is illustrated in the figure. The feasible region
is shown together with three different cost functions corresponding to the three
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
579
different cases. Note that for small k the optimum is at the corner A; as the cost
steepens, the corner B is optimum; and finally as the cost steepens further with
the highest values of k, the point C is the optimum.
8
This problem has four variables and so it has to be solved using the simplex
algorithm.
There are no difficulties with this problem and the tableaux are
presented, to two decimal places, without comment.
z
x5
x6
x7
x1
−2
2
1
0
x2
−1
0
0
4
z
x5
x3
x7
x1
−0.67
1.67
0.33
−0.33
x2
−1
0
0
4
z
x5
x3
x2
x1
−0.75
1.67
0.33
−0.08
x2
0
0
0
1
z
x1
x3
x2
x1
0
1
0
0
x2
0
0
0
1
x3
−4
1
3
1
x3
0
0
1
0
x3
0
0
1
0
x3
0
0
1
0
x4
−1
0
1
1
x4
0.33
−0.33
0.33
0.67
x4
0.5
−0.33
0.33
0.17
x4
0.35
−0.2
0.4
0.15
x5
0
1
0
0
x6
0
0
1
0
x5
0
1
0
0
x5
0
1
0
0
x5
0.45
0.6
− 0.2
0.05
x6
1.33
−0.33
0.33
−0.33
x6
1.25
−0.33
0.33
−0.08
x6
1.1
− 0.2
0.4
− 0.1
x7
0
0
0
1
Soln
0
3
4
3
x7
0
0
0
1
Soln
5.33
1.67
1.33
1.67
x7
0.25
0
0
0.25
Soln
5.75
1.67
1.33
0.42
x7
0.25
0
0
0.25
Soln
6.5
1
1
0.5
The solution is read off as x1 = 1, x2 = 0.5, x3 = 1 and x4 = 0; the maximum is
at z = 6.5.
The MAPLE instructions
con8:= {2 ∗ x1+x3<=3,x1+3 ∗ x3+x4<=4,4 ∗ x2+x3+x4<=3};
obj8:=2 ∗ x1+x2+4 ∗ x3+x4;
maximize(obj8,con8,NONNEGATIVE);
and the MATLAB instructions
c Pearson Education Limited 2011
580
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
f=[-2;-1;-4;-1];A=[2,0,1,0;1,0,3,1;0,4,1,1];b=[3;4;3];
options=optimset(‘LargeScale’,‘off’,‘Simplex’,‘on’);
[x,fval]=linprog(f,A,b,[],[],zeros(4,1),[],[],options)
also produce the solution {x4=0, x1=1, x3=1, x2=1/2}.
9
If b1,b2,b3 are the respective number (×1000) of books printed then the profit
will be
z = 900b1 + 800b2 + 300b3
and the constraints are
sales restriction
b1 + b2 ≤ 15
paper
3b1 + 2b2 + b3 ≤ 60
The first tableau is easily set up and the other tableaux follow by the usual rules.
z
b4
b5
b1
−900
1
3
z
b1
b3
b1
0
1
0
b2
100
1
−1
z
b1
b3
b1
0
1
0
b2
−200
1
−1
z
b2
b3
b2
−800
1
2
b1
200
1
1
b3
−300
0
1
b3
−300
0
1
b3
0
0
1
b2
0
1
0
b3
0
0
1
b4
0
1
0
b4
900
1
−3
b4
0
1
−3
b4
200
1
−2
b5
0
0
1
Soln
0
15
60
b5
0
0
1
Soln
13,500
15
15
b5
300
0
1
Soln
18,000
15
15
b5
300
0
1
Soln
21,000
15
30
Thus the optimal profit is made if none of book 1, 15,000 of book 2 and 30,000 of
book 3 are printed giving a profit of £21,000.
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
581
10 Let L, M, S be the respective number of long, medium and short range aircraft
purchased. The profit is
z = 0.4L + 0.3M + 0.15S
and the constraints are
money available
4L + 2M + S ≤ 60
pilots
L + M + S ≤ 25
maintenance
2L + 1.5M + S ≤ 30
The initial tableau is
z
P
Q
R
L
−0.4
4
1
2
M
−0.3
2
1
1.5
S
−0.15
1
1
1
P
0
1
0
0
Q
0
0
1
0
R
0
0
0
1
Soln
0
60
25
30
The first column is chosen, since it has the most negative entry, and when the
ratios are calculated, the P and R rows produce the same value of 15. From the
two, the P row is chosen arbitrarily.
z
L
Q
R
L
0
1
0
0
M
−0.1
0.5
0.5
0.5
S
−0.05
0.25
0.75
0.5
P
0.1
0.25
−0.25
−0.5
Q
0
0
1
0
R
0
0
0
1
Soln
6
15
10
0
In this tableau the M, R element is the pivot and performing the elimination gives
z
L
Q
M
L
0
1
0
0
M
0
0
0
1
S
0.05
−0.25
0.25
1
P
0
0.75
0.25
−1
Q
0
0
1
0
R
0.2
−1
−1
2
Soln
6
15
10
0
Note that the final column has not changed between the last two tableaux because
of the zero in the very last entry. The optimal solution is to purchase 15 long range
aircraft and no medium or short range aircraft; the estimated profit is £6 m.
c Pearson Education Limited 2011
582
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
In the problem, it would seem that some operational constraints have been omitted.
In many optimisation problems it takes several steps to achieve a sensible cost and
set of constraints.
11
The first tableau is set up from the data. Choosing the pivot in the usual way,
x1 and x5 are interchanged and the elimination follows the basic rules. Again, the
pivot is found and x3 and x6 are interchanged; the elimination produces a top row
that is all positive so the solution is optimal. The solution is read from the tableau
as x1 = 1.5, x2 = 0, x3 = 2.5, x4 = 0 and f = 14.
z
x5
x6
x7
z
x1
x6
x7
z
x1
x3
x7
x1
−6
2
1
1
x1
0
1
0
0
x1
0
1
0
0
x2
−1
1
0
1
x3
−2
0
1
3
x4
−4
1
1
2
x2
2
0.5
−0.5
0.5
x3
−2
0
1
3
x4
−1
0.5
0.5
1.5
x2
1
0.5
−0.5
2
x3
0
0
1
0
x4
0
0.5
0.5
0
x5
0
1
0
0
x5
3
0.5
−0.5
−0.5
x5
2
0.5
− 0.5
1
x6
0
0
1
0
x7
0
0
0
1
x6
0
0
1
0
x6
2
0
1
−3
Soln
0
3
4
10
Ratio
1.5
4
10
x7
0
0
0
1
Soln
9
1.5
2.5
8.5
Ratio
x7
0
0
0
1
Soln
14
1.5
2.5
1
Ratio
2.5
2.83
The MAPLE implementation just gives the ‘answer’. Some of the detail can be
extracted from MAPLE and the various tableau can be identified.
con11:={2 ∗ x1+x2+x4<=3,x1+x3+x4<=4,x1+x2+3 ∗ x3+2 ∗ x4<=10};
obj11:=6 ∗ x1+x2+2 ∗ x3+4 ∗ x4;
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
583
maximize(obj11,con11,NONNEGATIVE);
# gives the same solution {x4=0, x2=0, x1=3/2, x3=5/2}
# for the detail
z:=setup(con11);
z:={_SL1=3-2 x1-x2-x4,_SL2=4-x1-x3-x4,
_SL3=10-x1-x2-3 x3-2 x4}
piv:=pivoteqn(z,x1);
piv:=[_SL1=3-2 x1-x2-x4]
z:=pivot(z,x1,piv);
z:={x1=3/2-1/2_SL1-1/2 x2-1/2 x4,
_SL2=5/2+1/2_SL1+1/2 x2-1/2 x4-x3,
_SL3=17/2+1/2 _SL1-1/2 x2-3/2 x4-3 x3}
obj11:=eval(obj11,z);
obj11:= 9-3_SL1 - 2x2 + x4 + 2 x3
# compare with second tableau
piv:= pivoteqn(z,x3);
piv:=[_SL2=5/2 + 1/2_SL1 + 1/2x2 - 1/2x4-x3]
z:=pivot(z,x3,piv);
z:= {_SL3=-_SL1 + 1 + 3_SL2 - 2 x2,
x1 = 3/2 - 1/2_SL1 - 1/2 x2 - 1/2 x4,
x3=1/2_SL1 + 5/2 -_SL2 + 1/2 x2
- 1/2 x4}
obj11:=eval(obj11,z);
obj11:=-2_SL2 + 14 - 2_SL1-x2
# compare with the final tableau
Note that the entry in position (z, x4 ) is zero so we expect many solutions – this
is easily identified from the tableau but not from the MAPLE results. Continuing
with MAPLE code
piv:=pivoteqn(z,x4);
piv:=[x1=3/2 - 1/2_SL1 - 1/2 x2 - 1/2 x4]
z:=pivot(z,x4,piv);
c Pearson Education Limited 2011
584
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
z:={_SL3 = -_SL1 + 1 + 3_SL2 - 2 x2,
x4 = -2 x1+3 -_SL1 - x2,
x3 =_SL1 + 1 -_SL2 + x2 + x1}
obj11:=eval(obj11,z);
obj11:=-2_SL2+14-2_SL1-x2
The new solutions are x1 = x2 = 0, x3 = 1, x4 = 3 and f = 14, giving, as expected,
the same function value.
The MATLAB instructions give the same result but no detail
f=[-6;-1;-2;-4];A=[2,1,0,1;1,0,1,1;1,1,3,2];b=[3;4;10];
options=optimset(‘LargeScale’,‘off’,‘Simplex’,‘on’);
[x,fval]=linprog(f,A,b,[],[],zeros(4,1),[],[],options)
Exercises 10.2.6
12 The problem can be easily plotted and it is clear that x = 1, y = 4 gives the
solution f = x + 2y = 9.
6
5
4
y 3
2
1
0
1
2
3
x
4
5
c Pearson Education Limited 2011
6
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
585
The MAPLE check is as follows:
with (simplex):
con12:={y>=1,y<=4,x+y<=5};
obj12:=x+2 ∗ y;
maximize(obj12,con12,NONNEGATIVE);
# gives
{y=4,x=1}
The graph can be plotted using the instructions
with(plots):
F:=inequal(con12,x=0..6,y=0..6):
G:=plot([(-x+7)/2,(-x+9)/2,(-x+11)/2],x=0..6,
thickness=3,labels=[‘x’,‘y’],color=yellow):
display(F,G);
13
The problem has ‘greater than’ constraints, so the two-phase method is
required. Note that a surplus variable x4 and an artificial variable x7 need to
be introduced. At the end of phase 1 the artificial variable x7 will be eliminated
from the tableau. The cost function in phase 1 is −x7 , but recall that this cost
has to be modified to ensure that the tableau is in standard form.
Phase 1
z
x3
x7
x5
x6
z
x3
x7
x1
x6
x1
−2
4
2
2
−2
x1
0
0
0
1
0
x2
−1
1
1
−1
1
x2
−2
3
2
−0.5
0
x3
0
1
0
0
0
x3
0
1
0
0
0
x4
1
0
−1
0
0
x4
1
0
−1
0
0
x5
0
0
0
1
0
x6
0
0
0
0
1
x5
1
−2
−1
0.5
1
x6
0
0
0
0
1
c Pearson Education Limited 2011
x7
0
0
1
0
0
x7
0
0
1
0
0
Soln
−12
32
12
4
8
Soln
−8
24
8
2
12
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
586
x1
0
0
0
1
0
z
x3
x2
x1
x6
x2
0
0
1
0
0
x3
0
1
0
0
0
x4
0
1.5
− 0.5
− 0.25
0
x5
0
− 0.5
−0.5
0.25
1
x6
0
0
0
0
1
x7
1
− 1.5
0.5
0.25
0
Soln
0
12
4
4
12
The artificial cost has been driven to zero so the artificial variable x7 can be
eiiminated and the original cost function reinstated.
Phase 2
Note that the new cost is negative because the problem is a minimisation problem.
x1
0
0
0
1
0
z
x3
x2
x1
x6
x2
0
0
1
0
0
x3
0
1
0
0
0
x4
3
1.5
−0.5
−0.25
0
x5
−2
−0.5
−0.5
0.25
1
x6
0
0
0
0
1
x2
f = 40
30
f = 30
– 2x1 – x2 = 8
20
f = 20
2x1 – x2 = 4
10
4x1 + x2 = 32
2x1 + x2 = 12
2
4
6
8
c Pearson Education Limited 2011
x1
Soln
−44
12
4
4
12
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
x1
0
0
0
1
0
z
x3
x2
x1
x5
x2
0
0
1
0
0
x3
0
1
0
0
0
x4
3
1.5
− 0.5
− 0.25
0
x5
0
0
0
0
1
x6
2
0.5
0.5
− 0.25
1
587
Soln
− 20
18
10
1
12
The solution is read off as x1 = 1, x2 = 10 with a minimal cost of 20.
14
Let S be the number of shoes produced and B the number of boots, then the
problem is to maximize the profit
z = 8B + 5S
production
2B + S ≤ 250
sales
B + S ≤ 200
customer
B ≥ 25
The tableaux are constructed in the usual way.
Phase 1
z
P
Q
T
B
−1
2
1
1
S
0
1
1
0
B
0
0
0
1
z
P
Q
B
P
0
1
0
0
S
0
1
1
0
Q
0
0
1
0
P
0
1
0
0
Q
0
0
1
0
R
1
0
0
−1
R
0
2
1
−1
T
0
0
0
1
T
1
−2
−1
1
Soln
−25
250
200
25
Soln
0
200
175
25
Phase 2
z
P
Q
B
B
0
0
0
1
S
−5
1
1
0
P
0
1
0
0
Q
0
0
1
0
c Pearson Education Limited 2011
R
−8
2
1
−1
Soln
200
200
175
25
588
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
B
0
0
0
1
z
R
Q
B
S
−1
0.5
0.5
0.5
B
0
0
0
1
z
R
S
B
P
4
0.5
−0.5
0.5
S
0
0
1
0
P
3
1
−1
1
Q
0
0
1
0
Q
2
−1
2
−1
R
0
1
0
0
R
0
1
0
0
Soln
1000
100
75
125
Soln
1150
25
150
50
Thus, the manufacturer should make 50 pairs of boots and 150 pairs of shoes and
the maximum profit is £1150.
A graphical solution is possible since the problem has only two variables. It is of
interest to follow the progress of the simplex solution on the graph.
boots
150
profit = 1150
2B + S = 250
100
B + S = 200
profit = 600
50
B = 25
30
15
60
90
120
150
180
shoes
There are techniques for adding a constraint to an existing solution and thence
determining the solution. However, here the whole problem will be recomputed.
From Exercise 9 the initial tableau can be constructed, but the additional constraint
is a ‘greater than’ constraint so the procedure for entering phase 1 must be followed.
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
z
b4
b5
b7
b1
−1
1
3
1
b2
−1
1
2
1
b3
−1
0
1
1
b4
0
1
0
0
b5
0
0
1
0
b6
1
0
0
−1
b7
0
0
0
1
589
Soln
−50
15
60
50
Note that the last row gives the new constraint and that the z row involves the
artificial variable b7, which has been eliminated to bring the tableau to standard
form.
Phase I
z
b1
b5
b7
b1
0
1
0
0
b2
0
1
−1
0
z
b1
b3
b7
b1
0
1
0
0
b2
−1
1
−1
1
b1
0
1
0
0
z
b1
b3
b4
b3
−1
0
1
1
b3
0
0
1
0
b2
0
0.5
0.5
0.5
b3
0
0
1
0
b4
1
1
−3
−1
b4
−2
1
−3
2
b5
0
0
1
0
b6
1
0
0
−1
b7
0
0
0
1
Soln
−35
15
15
35
b5
1
0
1
−1
b6
1
0
0
−1
b7
0
0
0
1
Soln
−20
15
15
20
b4
0
0
0
1
b5
0
0.5
− 0.5
− 0.5
b3
0
0
1
0
b4
0
0
0
1
b6
0
0.5
−1.5
−0.5
b7
1
−0.5
1.5
0.5
b5
300
0.5
−0.5
−0.5
b6
0
0.5
−1.5
−0.5
Soln
0
5
45
10
Phase 2
z
b1
b3
b4
b1
0
1
0
0
b2
−200
0.5
0.5
0.5
c Pearson Education Limited 2011
Soln
18,000
5
45
10
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
590
b1
400
2
−1
1
z
b1
b3
b4
b2
0
1
0
0
b3
0
0
1
0
b4
0
0
0
1
b5
500
1
−1
−1
b6
200
1
−2
−1
Soln
20,000
10
40
5
The solution is optimal with the production schedule as none for book 1, 10,000
of book 2 and 40,000 of book 3. The profit is down to £20,000 because of the
additional constraint.
16
The MAPLE solution is presented.
with (simplex):
con16:={x>=1,x+2*y<=3,y+3*z<=4};
obj16:=x+y+z;
maximize(obj16,con16,NONNEGATIVE);
# gives the solution {y=0,z=4/3,x=3}, f=x+y+z=13/3
Note that there is no detail of the two-phase method and just the ‘answer’ is given.
It is not easy to rewrite phase 1 into MAPLE so that the detail can be extracted.
The same is true of MATLAB, the following instructions produce the correct result
f=[-1;-1;-1];A=[1,2,0;0,1,3;-1,0,0];b=[3;4;-1];
options=optimset(‘LargeScale’,‘off,’Simplex’,‘on’);
[x,fval]=linprog(f,A,b,[],[],zeros(3,1),[],[],options)
17
This is a standard two-phase problem with surplus variables x5 , x6 and
artificial variables x7 , x8 . With the cost constructed from the artificial variables
x7 , x8 , and the tableau reduced to standard form, the sequence of tableaux is
presented.
Phase 1
z
x7
x9
x1
−1
1
0
x2
−1
0
1
x3
0
−1
1
x4
1
−1
0
x5
1
−1
0
x6
1
0
−1
c Pearson Education Limited 2011
x7
0
1
0
x8
0
0
1
Soln
−2
0
2
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
591
There is an arbitrary choice of the columns since two columns have the value −1;
the second one is chosen.
x1
−1
1
0
z
x7
x2
x1
0
1
0
z
x1
x2
x2
0
0
1
x3
1
−1
1
x4
1
−1
0
x5
1
−1
0
x6
0
0
−1
x7
0
1
0
x8
1
0
1
x2
0
0
1
x3
0
−1
1
x4
0
−1
0
x5
0
−1
0
x6
0
0
−1
x7
1
1
0
x8
1
0
1
Soln
0
0
2
Soln
0
0
2
The solution is optimal so phase 1 ends and the tableau is reconstituted for phase 2.
Phase 2
z
x1
x2
x1
0
1
0
x2
0
0
1
x1
0
1
0
z
x1
x3
x3
−1
−1
1
x2
1
1
1
x3
0
0
1
x4
7
−1
0
x5
2
−1
0
x4
7
−1
0
x5
2
−1
0
x6
7
0
−1
x6
6
−1
−1
Soln
−14
0
2
Soln
−12
2
2
The solution is now optimal with x1 = 2, x2 = 0, x3 = 2, x4 = 0 and the cost
function being 12.
18
Let the company buy a,b,c litres of the products A,B,C, then the cost of the
materials will be
Cost
1.8a + 0.9b + 1.5c
For a total of 100 litres
a + b + c = 100
Glycol
0.65a + 0.25b + 0.8c ≥ 50
Additive
0.1a + 0.03b ≥ 5
There are two ‘greater than’ constraints and an equality constraint. Recall that an
equality constraint is dealt with via an artificial variable. Thus, phase 1 is entered
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
592
with two surplus variables and three artificial variables; the cost row is manipulated
until the tableau has the standard form.
Phase 1
a
−1.75
1
0.65
0.1
z
r
s
t
b
−1.28
1
0.25
0.03
a
z −0.29
r
0.19
0.81
c
0.1
t
z
p
c
t
a
−0.1
0.15
1
0.1
z
p
c
a
a
0
0
0
1
c
−1.8
1
0.8
0
b
−0.72
0.69
0.31
0.03
c
0
0
1
0
b
−0.03
0.55
1
0.03
b
0
0.51
0.7
0.3
c
0
0
1
0
p
−1.25
1.25
−1.25
0
c
0
0
1
0
p
0
1
0
0
p
1
0
−1
0
p
0
1
0
0
q
0
1.5
10
− 10
q
1
0
0
−1
q
1
0
0
−1
r
0
1
0
0
r
0
1
0
0
q
1
0
0
−1
s
2.25
−1.25
1.25
0
r
1
0.8
1
0
r
1
0.8
1
0
s
0
0
1
0
t
0
0
0
1
Soln
−155
100
50
5
t
0
0
0
1
Soln
−42.5
37.5
62.5
5
s
1
−1
0
0
s
1
−1
0
0
t
1
− 1.5
− 10
10
t
0
0
0
1
Soln
−5
30
100
5
Soln
0
22.5
50
50
The artificial cost has been driven to zero so phase 1 is complete and the three
artificial columns are deleted and the actual cost introduced. Recall that this is
a minimisation problem so the phase 2 tableau can now be extracted from the
tableau above.
Phase 2
z
p
c
a
a
0
0
0
1
b
−0.69
0.51
0.7
0.3
c
0
0
1
0
p
0
1
0
0
q
3
1.5
10
−10
c Pearson Education Limited 2011
Soln
−165
22.5
50
50
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
a
0
0
0
1
z
b
c
a
b
0
1
0
0
c
0
0
1
0
p
1.37
1.98
− 1.39
−0.59
q
5.05
2.97
7.92
− 10.89
593
Soln
−134.26
44.55
18.81
36.63
The tableau is optimal so the solution can be read off as a = 36.63%,
b = 44.55%, c = 18.81% and a minimum cost of £134.26.
MATLAB solves the problem as follows
f=[1.8;0.9;1.5];A=[-0.65,-0.25,-0.8;-0.1,-0.03,0];b=[-50;-5];
options=optimset(‘LargeScale’,‘off,‘Simplex’,‘on’);
[x,fval]=linprog(f,A,b,[1,1,1],[100],zeros(3,1),[],[],options)
% Note how MATLAB deals with the equality constraint
19
Let s1 , s2 , s3 be the number of houses of the three styles the builder decides
to construct. His profit (× £100) is
10s1 + 15s2 + 25s3
and the constraints are
plots
s1 + 2s2 + 2s3 ≤ 40
facing stone
s1 + 2s2 + 5s3 ≤ 58
weather boarding
3s1 + 2s2 + s3 ≤ 72
local authority
− s1 + s2 ≥ 5
The solution requires the two-phase method. The tableaux are listed, to two
decimal places, without comment.
Phase 1
z
s4
s5
s6
s8
s1
1
1
1
3
−1
s2
−1
2
2
2
1
s3
0
2
5
1
0
s4
0
1
0
0
0
s5
0
0
1
0
0
s6
0
0
0
1
0
c Pearson Education Limited 2011
s7
1
0
0
0
−1
s8
0
0
0
0
1
Soln
−5
40
58
72
5
594
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
s1
0
3
3
5
−1
z
s4
s5
s6
s2
s2
0
0
0
0
1
s3
0
2
5
1
0
s4
0
1
0
0
0
s5
0
0
1
0
0
s6
0
0
0
1
0
s7
0
2
2
2
−1
s8
1
−2
−2
−2
1
Soln
0
30
48
62
5
s5
0
0
1
0
0
s6
0
0
0
1
0
s7
−15
2
2
2
−1
Soln
75
30
48
62
5
s5
0
0
1
0
0
s6
0
0
0
1
0
s7
1.67
0.67
0
−1.33
−0.33
Soln
325
10
18
12
15
Phase 2
z
s4
s5
s6
s2
z
s1
s5
s6
s2
z
s1
s3
s6
s2
s1
−25
3
3
5
−1
s1
0
1
0
0
0
s2
0
0
0
0
1
s2
0
0
0
0
1
s1
0
1
0
0
0
s3
−25
2
5
1
0
s3
−8.33
0.67
3
−2.33
0.67
s2
0
0
0
0
1
s3
0
0
1
0
0
s4
0
1
0
0
0
s4
8.33
0.33
−1
−1.67
0.33
s4
5.56
0.56
− 0.33
− 2.44
0.56
s5
2.78
− 0.22
0.33
0.78
− 0.72
s6
0
0
0
1
0
s7
1.67
0.67
0
− 1.33
− 0.33
Soln
375
6
6
26
11
The tableau is optimal, so the solution should build 6,11 and 6 houses of styles 1,2
and 3 respectively and make a profit of £37,500.
20 Let x1 x2 , x3 be the amounts (×1000 m2 ) of carpet of type C1,C2,C3 produced,
then the maximum of
2x1 + 3x2
is required subject to the constraints
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
M1
x1 + x2 + x3 ≤ 5
M2
x1 + x2 ≤ 4
policy1
x1 ≥ 1
policy2
x1 − x2 + x3 ≥ 2
595
Phase 1
z
x4
x5
x6
x9
x1
2
1
1
1
1
x2
1
1
1
0
−1
x3
−1
1
0
0
1
z
x4
x5
x1
x9
x1
0
0
0
1
0
x2
1
1
1
0
−1
x3
1
1
0
0
1
z
x4
x5
x1
x3
x1
0
0
0
1
0
x2
0
2
1
0
−1
x4
0
1
0
0
0
x4
0
1
0
0
0
x3
0
0
0
0
1
x4
0
1
0
0
0
x5
0
0
1
0
0
x5
0
0
1
0
0
x5
0
0
1
0
0
x6
1
0
0
−1
0
x6
−1
1
1
−1
1
x6
0
0
1
−1
1
x7
1
0
0
0
−1
x7
1
0
0
0
−1
x7
0
1
0
0
−1
x8
0
0
0
1
0
x9
0
0
0
0
1
Soln
−3
5
4
1
2
x8
2
−1
−1
1
−1
x9
0
0
0
0
1
Soln
−1
4
3
1
1
x9
1
−1
0
0
1
Soln
0
3
3
1
1
x8
1
0
−1
1
−1
Phase 2
z
x4
x5
x1
x3
x1
0
0
0
1
0
x2
3
8
1
0
1
x3
0
0
0
0
1
x4
0
1
0
0
0
x5
0
0
1
0
0
z
x2
x4
x1
x3
x1
0
0
0
1
0
x2
0
1
0
0
0
x3
0
0
0
0
1
x4
1.5
0.5
0.5
0
0.5
x5
0
0
1
0
0
x6
−2
0
1
−1
1
x6
−2
0
1
0
0
c Pearson Education Limited 2011
x7
0
1
0
0
−1
Soln
2
9
3
1
1
x7
1.5
0.5
0.5
0
−0.5
Soln
6.5
1.5
1.5
1.5
2.5
596
z
x2
x6
x1
x3
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
x1
0
0
0
1
0
x2
0
1
0
0
0
x3
0
0
0
0
1
x4
0.5
0.5
−0.5
−0.5
1
x5
2
0
1
1
−1
x6
0
0
1
0
0
x7
0.5
0.5
− 0.5
− 0.5
0
Soln
9.5
1.5
1.5
2.5
1
Hence, the optimum is 2500 m2 of C1, 1500 m2 of C2 and 1000 m2 of C3 giving a
profit of £9500.
Exercises 10.3.3
21
Introducing a Lagrange multiplier, the minimum is obtained from
2x + y + λ = 0
x + 2y + λ = 0
x+y = 1
with solution x = y = 1/2. To prove that it is a minimum, put
x=
1
2
+ ε and y =
1
2
−ε
and substitute into f to give
f=
3
4
+ ε2
and hence a minimum.
22
The answer to this exercise is geometrically obvious since it is just the length
of the minor axis which in this case, is a. It is required to find the minimum of
D2 = x2 + y2 subject to
x2
y2
+
=1
a2
b2
The Lagrange equations are
0 = 2x +
2λy
2λx
and
0
=
2y
+
a2
b2
Since λ cannot equal −a2 and −b2 at the same time then either x = 0 or y = 0.
If a < b the it is clear that y = 0 and x = ± a gives the minimum
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
23
The area of the rectangle is
597
y
A = 4xy
(x,y)
and since the point must lie
on the ellipse
x
x2
y2
+
=1
a2
b2
and hence a Lagrange multiplier problem is produced. The two equations are
2λx
2λy
and 0 = 4x + 2
2
a
b
√
√
It is soon checked that the required solution is x = a/ 2, y = b/ 2, A = 2ab.
0 = 4y +
24
The necessary conditions for an optimum are
∂f
∂g
+λ
= 0 = y2 z + λ
∂x
∂x
∂g
∂f
+λ
= 0 = 2xyz + 2λ
∂y
∂y
∂f
∂g
+λ
= 0 = xy2 + 3λ
∂z
∂z
(1)
(2)
(3)
together with the given constraint
x + 2y + 3z = 6
(4)
Dividing equation (2) by two and then subtracting the first two equations gives
yz(x − y) = 0
Thus, there are three cases
y = 0 Note that all the equations except the constraint are satisfied since λ = 0.
Thus, a possible solution is (6 − 3α, 0, α) for any α.
z = 0 Again, this implies that λ = 0 so from equation (3) x = 0 and hence from
(4) y = 3. The case when y = 0 in equation (3) has been covered in the first case.
x = y Equations (2) and (3) give
1 3
3x
= x2 z so there are two cases
either x = 0 and hence y = 0 and z = 2
c Pearson Education Limited 2011
598
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
or x = 3z and hence from (4) z = 1/2 and x = 3/2, y = 3/2
Equations (1) and (2) have been used to eliminate λ; so a similar exercise must be
undertaken with (2) and (3) to check whether any new solutions arise. They give
the equation
xy(y − 3z) = 0
and again three cases must be studied.
x = 0 In this case λ = 0 so it reduces to one of the above.
y = 0 Again, it reduces to the first case above.
y = 3z From equations (1) and (2) it can be seen that the same cases arise and
correspond to the solutions already obtained.
Note that in the solutions quoted, (6,0,0) is included in the solution (6 − 3α, 0, α)
with α = 0.
25
Using λ and μ as the Lagrange multipliers, the equations that give the
optimum are as quoted
0 = 2x + λ + (2z − 2y)μ
0 = 2y + λ + (z − 2x)μ
0 = 2z − λ + (y + 2x)μ
together with the two given constraints.
Adding the last two equations
2(y + z) + (y + z)μ = 0 ⇒ either y = −z or μ = −2
Adding the first and last equations
2(x + z) + (2z + 2x − y)μ = 0
So, λ has been eliminated.
In the first constraint x = −y + z = 2z . Thus, (2t, −t, t) gives, in
√
the second constraint, 7t2 = 1 so t = ±1/ 7 and two of the quoted solutions are
obtained.
y+z=0
μ = −2
From above 2(x + z) + (2z + 2x − y)(−2) = 0 and so this expression
and the second constraint give a pair of equations
0 = −x + y − z
0=x+y−z
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
599
and hence, x = 0 and y = z . Putting these values into the second constraint gives
y = ±1 and possible optimum points at (0, 1, 1) and (0,− 1,− 1).
26
The volume of the figure is
c
V = abc
and the surface area is
A = 2ab + 2ac + bc
a
The problem is a Lagrange multiplier
problem and the three equations are
b
0 = bc + λ(2b + 2c)
0 = ac + λ(2a + c)
0 = ab + λ(2a + b)
The last two equations give
a(c − b) + λ(c − b) = 0 ⇒ either c = b or λ = −a
c = b Putting back into the first equation gives
b2 + 4bλ = 0 so either b = 0 or λ = −b/4
The case b = 0 can be dismissed as geometrically uninteresting. In the third
equation this value of λ gives 0 = ab − 14 b(2a + b) ⇒ b = 2a since again the case
b = 0 is uninteresting.
Calculating a gives the possible maximum as
c = b = 2a with a =
A
and V = 4a3
12
λ = −a Putting this value back into the third of the Lagrange equations gives
a = 0 so gives zero volume and is geometrically uninteresting again.
27
This is a very tough problem but more typical of realistic problems in
engineering rather than most of the illustrative exercises in the book.
c Pearson Education Limited 2011
600
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The exercise will be solved using MAPLE. Note the efficient way it performs the
algebra and integrations for the simplest approximation
z:=A ∗ cos(Pi ∗ x/2)ˆ2 ∗ cos(Pi ∗ y/2)ˆ2;
I1:int(int(z^2,x=-1..1),y=-1..1);
# gives
I1:=9/16 A 2
f:=diff(z,x,x)+diff(z,y,y);
I2:=int(int(f^2,x=-1..1),y=-1..1);
# gives
I2:=1/2 A 2 Pi 4
freq:=I2/I1;
# gives
freq:=8/9 Pi 4 = 86.58 as required.
The first approximation could be done by hand but the next approximation needs
a package like MAPLE.
zz:=cos(Pi ∗ x/2)ˆ2 ∗ cos(Pi ∗ y/2)ˆ2
∗
(A+B ∗ cos(Pi ∗ x/2) ∗ cos(Pi ∗ y/2));
I3:=evalf(int(int(zz^2,x=-1..1),y=-1..1));
# gives
I3:=.5624999999 A 2 + .9222479291 A B + .3906249999 B 2
ff:=diff(zz,x,x)+diff(zz,y,y);
I4:=evalf(int(int(ff^2,x=-1..1),y=-1..1));
# gives
I4:=58.21715209 B 2 + 48.70454554 A 2 + 92.64268668 A B
dLbydA:=diff(I4+lam ∗ I3,A);
dLbydB:=diff(I4+lam ∗ I3,B);
fsolve({dLbydA,dLbydB,I3=1}, {A,B,lam});
# gives the solution
{lam=-81.38454660, A=-1.829151961, B=.6086242001}
The frequency is given by −lam which should be compared with the first
approximation.
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
28
601
The problem asks for a minimum, which is equivalent to the maximum of
−2x21 − x22 − 2x1 x2
The Kuhn–Tucker conditions for this problem are
−4x1 − 2x2 − μ = 0
−2x1 − 2x2 + μ = 0
x1 − x2 − α ≤ 0
μ(x1 − x2 − α) = 0
μ≥0
There are two cases
μ=0
The first two equations give x1 = x2 = 0 and these can only give the
optimum if the inequality is satisfied, namely α ≥ 0.
x1 − x2 = α
Adding the first two equations gives the solution in terms of the
simultaneous linear equations
3x1 + 2x2 = 0
x1 − x2 = α
which clearly have the solution x1 = 2α/5 and x2 = −3α/5. Calculation from the
first or second equation gives μ = −2α/5 which is only optimal if α ≤ 0.
Exercises 10.4.2
29(a)
The MATLAB version of this problem (note it deals with the maximum)
is as follows:
a=0.1;h=0.2;nmax=10;n=0;
zold=q29(a);a=a+h;z=q29(a);c=[z];
while(z(2)>zold(2))&(nb(2)
if xstar0 an=c;bn=b; else bn=c;an=a;end
% an and bn contain the new bracket values
c Pearson Education Limited 2011
603
604
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
30 If the derivatives are not used, then the bracket is 0 < x < 3, whilst using
the derivatives gives the much narrower range of 0 < x < 1 but at the expense of
twice the number of function evaluations.
30(a) The calculation is identical to Exercise 29 with q29 replaced by q30 in qapp
p=[q30(0),q30(1),q30(3)]; u=p(1,:),v=p(2,:)
% gives
u= 0
1
3
v= 0
0.4207 0.0141
[u,v]=qapp(u,v)
% gives
u= 0
1.0000 1.5113
v= 0 0.4207 0.3040
[u,v]=qapp(u,v)
% gives
u= 0
0.9898 1.0000
v= 0
0.4222 0.4207
where the function q30 is in the M-file
function a=q30(x)
a=[x;sin(x)/(1+xˆ2);cos(x)/(1+xˆ2)-2ˆ∗ xˆ∗ sin(x)/(1+xˆ2)ˆ2];
30(b) Again, as in Exercise 29, with q29 replaced by q30 in cufit
p=q30(0);q=q30(1);p ,q
% gives
x=0
f=0
f = 1
% and
x=1.0000
f=0.4207
f =-0.1506
[p,q]=cufit(p,q);p ,q
% gives
x=0
f=0
f = 1
% and
x=0.8667
f=0.4352
f =-0.0612
[p,q]=cufit(p,q);p ,q
% gives
x=0
f=0
f = 1
% and
x=0.8242
f=0.4371
f =-0.0247
The built in MATLAB procedure fminbnd gives x = 0.7980, f = 0.4374 in 9
iterations
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
605
f=@(x)-sin(x)/(1+x^2);
options=optimset(‘display’,‘iter’);
x=fminbnd(f,0,1,options)
31
The problem is a standard exercise to illustrate the use of the two basic
algorithms for a single variable search.
31(a) An adaptation of Exercise 29 with the appropriate function used gives
p=[q31(1),q31(5/3),q31(3)];u=p(1,:),v=p(2,:)
% gives
u=1.0000
1.6667
3.0000
v=0.2325
0.2553
0.1419
u=1.000
1.6200
1.6667
v=0.2325
0.2571
0.2553
1.4784
0.2602
1.6200
0.2571
[u,v]=qapp(u,v)
% gives
[u,v]=qapp(u,v)
% gives
u=1.0000
v=0.2325
where q31 is the M-file
function a=q31(x)
a=[x;x ∗ (exp(-x)-exp(-2 ∗ x));(exp(-x)-exp(-2 ∗ x))
-x ∗ (exp(-x)-2 ∗ exp(-2 ∗ x))];
31(b)
Use the same bracket and adapt the algorithm in Exercise 29.
p=q31(1);q=q31(3);p ,q
% gives
x=1.0000
f=0.2325
f = 0.1353
% and
x=3.0000
f=0.1419
f = -0.0872
[p,q]=cufit(p,q);p ,q
% gives
x=1.0000
f=0.2325
f = 0.1353
% and
x=1.5077
f=0.2599
f = -0.0136
[p,q]=cufit(p,q);p ,q
c Pearson Education Limited 2011
606
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
% gives
x=1.0000
f=0.2325
f = 0.1353
% and
x=1.4462
f=0.2603
f = -0.0001
The next iteration gives the three-figure accuracy required in (c).
x = 1.4456 f = 0.2603 f = 0.0000
32 The difficulty in this problem is that the eigenvalues must be computed at
each function evaluation. With a package such as MATLAB this is less of a problem
since the instruction eig(A) will give them almost instantly.
The M-file q32 performs this computation. Note the -max(eig(A)) since qapp
looks for a maximum
function a=q32(x)
A=[x -1 0;-1 0 -1;0 -1 x^2];
a=[x;-max(eig(A))];
The calculations are performed as in Exercise 29
p=[q32(-1),q32(0),q32(1)];u=p(1,:),v=p(2,:)
% gives
u=-1
0
v=-1.7321
1
-1.4142
-2.0000
[u,v]=qapp(u,v)
% gives
u=-1.0000
-0.1483
0
v=-1.7321
-1.3854
-1.4142
[u,v]=qapp(u,v)
% gives
u=-1.0000
-0.2356
-0.1483
v=-1.7321
-1.3769
-1.3854
See the text for the MATLAB fminbnd code for this problem.
33
To establish the formula is a matter of simple substitution into (10.10).
To find when f (x) = 0, from the Newton method, requires the iteration of the
formula
xnew = x −
f (x)
f (x)
If the function is known at x − h, x, x + h with values f1 , f2 , f3 respectively, then
the derivatives can be replaced by their approximations
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
f3 − f1
f1 − 2f2 + f3
and f (x) ≈
2h
h2
and the formula follows immediately.
f (x) ≈
34(a)
To obtain the Golden Section ratio, the figure gives
A
l
C
B
D
l
L
AC
AD
=
AD
AB
L−l
l
=
L−l
L
yields
−→
Solve for a = l/L to obtain a = 12 (3 −
√
5) .
The Golden Section algorithm can be written in MATLAB as
alpha = (3-sqrt(5))/2;a=[0;1;1;2.5];
f=@ (x)x ∗ sin(x);
a=[0;(1-alpha) ∗ a(1)+alpha ∗ a(4);alpha ∗ a(1)+(1-alpha) ∗ a(4);2.5];
F=[f(a(1));f(a(2));f(a(3));f(a(4))];aa=[a];FF=[F];
for i=1:5
if F(2)>F(3)a(4)=a(3);a(3)=a(2);F(4)=F(3);F(3)=F(2);
a(2)=(1-alpha) ∗ a(1)+alpha ∗ a(4);F(2)=f(a(2));
else a(1)=a(2);a(2)=a(3);F(1)=F(2);F(2)=F(3);
a(3)=alpha ∗ a(1)+(1-alpha) ∗ a(4);F(3)=f(a(3));
end, aa=[aa,a];FF=[FF,F];
end
The instructions aa, FF prints out x and f at five successive iterations
aa =
0
0.9549
1.5451
1.9098
1.9098
1.9098
0.9549
1.5451
1.9098
2.1353
2.0492
1.9959
1.5451
1.9098
2.1353
2.2746
2.1353
2.0492
2.5000
2.5000
2.5000
2.5000
2.2746
2.1353
c Pearson Education Limited 2011
607
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
608
FF =
0
0.7795
1.5446
1.8011
1.8011
1.8011
0.7795
1.5446
1.8011
1.8040
1.8191
1.8183
1.5446
1.8011
1.8040
1.7341
1.8040
1.8191
1.4962
1.4962
1.4962
1.4962
1.7341
1.8040
with the best value at x = 2.0492 and f = 1.8191.
34(b)
The algorithm is the same as part (a) except the function is now
f=@ (x)(log(x)+2 ∗(1-x)/(1+x))/(1-x)∧ 2;
The five iterations give
aa =
1.5000
1.8820
1.8820
2.0279
2.1180
2.1180
1.8820
2.1180
2.0279
2.1180
2.1738
2.1525
2.1180
2.2639
2.1180
2.1738
2.2082
2.1738
2.5000
2.5000
2.2639
2.2639
2.2639
2.2082
FF =
0.0218604
0.0260436
0.0260436
0.0265463
0.0266783
0.0266783
0.0260436
0.0266783
0.0265463
0.0266783
0.0267059
0.0266998
0.0266783
0.0262879
0.0266780
0.0262879
0.0266783
0.0266780
0.0267059
0.0266780
0.0267051
0.0266780
0.0267059
0.0267051
Note that the function is so flat that it has been quoted to six figures; the best
result is x = 2.1738, f = 0.0267059.
Exercises 10.4.4
35
The gradient is given by
1
1
+
G=
−1
0
−1
2
x1
x2
so, at the first step
1
1
, f = 1.5, G =
a=
1
1
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
609
The first search is for min [f(1 − μ, 1 − μ)] and a simple calculation gives μ = 2.
μ
The second iteration can be started as
1
−1
, f = −0.5, G =
a=
−1
−1
Note that the two gradients are perpendicular. The search min [f(−1 − μ, −1 + μ)]
μ
is easily performed (exactly in this problem) to give μ = 0.4 and the problem is
ready for the next iteration.
1/5
−7/5
, f = −0.9, G =
a=
1/5
−3/5
36
For this function the gradient is
6x + 2y + 3
G=
2x + 6y + 2
and it easily checked that x = −7/16, y = −3/16 gives zero gradient and hence
the minimum at f = −0.84375.
The steepest descent method is easily written in MATLAB. For more complicated
functions two M-files are needed, one to set up the function and its gradient and
the second to set up the line search routine. Anonymous functions can be used for
more .straightforward functions. The following lines of code solve the problem.
f=@ (x)2 ∗ (x(1)+x(2))ˆ2+(x(1)-x(2))ˆ2+3 ∗ x(1)+2 ∗ x(2);
g=@(x)[4∗(x(1)+x(2))+2∗(x(1)-x(2))+3;4∗(x(1)+x(2))-2∗(x(1)-x(2))+2];
fm=@ (t,xx,G)f(xx-t ∗ G);
% The function, its derivative and the line search are now set
xx=[0,0];ff=f(xx);G=g(xx);XX=[xx];FF=[ff];GG=[G];% Start values
[t,fval]=fminbnd@(t)fm(t,xx,G),0,2);% Note how the minimisation is done
xx=xx-t ∗ G;G=g(xx);ff=f(xx);XX=[XX,xx];GG=[GG,G];FF=[FF,ff];
%Next point
%Repeat the last two lines to iterate
c Pearson Education Limited 2011
610
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The first five values are produced in XX,FF and GG as
XX=
0
-0.3824
-0.4296
-0.4365
-0.4374
0
-0.2549
-0.1841
-0.1887
-0.1874
FF=
0
-0.8284
-0.8435
-0.8437
-0.8437
GG=
3.0000
2.0000
0.1961
-0.2941
0.0545
0.0363
0.0036
-0.0053
0.0010
0.0007
Because the function is a quadratic, the Newton method, which is based on the
assumption that the function is approximated by a quadratic, must converge in
one iteration.
37
The MATLAB implementation is similar to Exercise 36.
The program is identical except for the instructions that generate the functions,
which are now
f=@ (x)(x(1)-x(2)+x(3)) ∧ 2+(2 ∗ x(1)+x(3)-2) ∧ 2+(x(3) ∧ 2-1) ∧ 2;
g=@ (x)[2 ∗ (x(1)-x(2)+x(3))+4 ∗ (2 ∗ x(1)+x(3)-2);2 ∗ (x(1)-x(2)+x(3));...
2 ∗ (x(1)-x(2)+x(3))+2 ∗ (2 ∗ x(1)+x(3)-2)+4 ∗ x(3) ∗ (x(3) ∧ 2-1)];
The first few iterations are
XX =
2.0000
1.1518
0.4985
0.6175
0.4941
0.5178
2.0000
2.1696
1.8171
1.7505
1.6616
1.6358
2.0000
0.4733
0.7984
0.9653
1.0178
1.0300
FF =
29.0000
1.5023
0.4440
0.0729
0.0237
0.0158
GG =
20.0000
2.0184
-1.8592
0.4657
-0.2758
0.0859
-4.0000
1.0891
1.0405
0.3354
0.2995
0.1762
36.0000
-1.0044
-2.6078
-0.1980
-0.1414
0.2054
To use the built in routine fminunc, it is easier to use an M file.
function [f,g] = q37(x) %Question 37
f=(x(1)-x(2)+x(3)) ∧ 2+(2 ∗ x(1)+x(3)-2) ∧ 2+(x(3) ∧ 2-1) ∧ 2;
g=[2 ∗ (x(1)-x(2)+x(3))+4 ∗ (2 ∗ x(1)+x(3)-2);2 ∗ (x(1)-x(2)+x(3));...
2 ∗ (x(1)-x(2)+x(3))+2 ∗ (2 ∗ x(1)+x(3)-2)+4 ∗ x(3) ∗ (x(3) ∧ 2-1)];
end
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
611
The main progam is
x=[2;2;2];
options=optimset(‘GradObj’,‘on’,‘display’,‘iter’,‘HessUpdate’,
‘steepdesc’,‘LargeScale’,‘off’);
[x,fval]=fminunc(@q37,x,options)
The first three and last three iterations are
Iteration
0
Func-count
1
f(x)
29
1
2
3.67901
2
4
2.35628
with starting point x=(2,2,2)
. . . . . . . . . . . . .
38
59
119
8.6485e-010
60
121
6.17926e-010
61
123
4.41502e-010
with final point x=(0.5,1.5,1.0)
The function, gradient and Hessian matrix are computed in the M-file
newton38
function [a,agrad,ajac]=newton38(z)
t1=z(1)-z(2)+z(3);t2=2 ∗ z(1)+z(3)-2;
a=t1ˆ2+t2ˆ2+(z(3)ˆ2-1)ˆ2;
agrad(1)=2 ∗ t1+4 ∗ t2;
agrad(2)=-2 ∗ t1;
agrad(3)=2 ∗ t1+2 ∗ t2+4 ∗ z(3)ˆ3-4 ∗ z(3);
ajac(1,1)=10;ajac(1,2)=-2;ajac(1,3)=6;
ajac(2,1)=-2;ajac(2,2)=2;ajac(2,3)=-2;
ajac(3,1)=6;ajac(3,2)=-2;ajac(3,3)=12 ∗ z(3)ˆ2;
and the Newton iteration proceeds as
a=[2;2;2];E=[0;0;0;0];
for i=1:5
[f,G,J]=newton38(a);E=[E,[f;a]];
c Pearson Education Limited 2011
612
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
a=a-J\G ;
end
% gives
39
f
29.0000
1.2448
0.1056
0.0026
0.0000
x
2.0000
0.2727
0.4245
0.4873
0.4995
y
2.0000
1.7273
1.5755
1.5127
1.5005
z
2.0000
1.4545
1.1510
1.0253
1.0009
The figure illustrates the cost of the road from (0,0) to (1,a) to ( b,11-2b) is
1/2
C = 2(1 + a2 )1/2 + (b − 1)2 + (11 − 2b − a)2
and any of the minimization methods give a = 0.2294, b = 4.5083 and c = 5.9743,
giving equations of lines as y = 0.2294x, y = 0.5x − 0.2706 and cost = 5.974.
y
(b,11– 2b)
(1,a)
y = 11 – 2x
x
(0,0)
Exercises 10.4.7
40(a)
The gradient can be calculated as G =
10x − 2y − 8
−2x + 2y
choice of H is the unit matrix.
c Pearson Education Limited 2011
and the initial
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
613
Iteration 1
The computation commences
2
8
1 0
f = 4 G1 =
H1 =
a1 =
2 1
0
0 1
2 − 8λ
. The minimum
and the minimization takes place in the direction a =
2
in this direction may be obtained as λ = 0.1. The new values are
1.2
0
a2 =
f2 = 0.8 G2 =
2
1.6
and the values of h1 and y1 are calculated as
−0.8
h1 = a2 − a1 =
0
−8
and y1 = G2 − G1 =
1.6
Finally, the H is updated as
1 −0.8
1
−8
1 0
[−8 1.6] +
[−0.8 0]
−
H2 =
0
0 1
66.56 1.6
6.4
0.1 0
0.9615 −0.1923
1 0
+
−
=
0 0
−0.1923 0.0385
0 1
0.1385 0.1923
=
0.1923 0.9615
Iteration 2
The iteration starts with the variables computed from iteration 1
1.2
0
0.1385
f2 = 0.8 G2 =
H2 =
a2 =
2
1.6
0.1923
0.1923
0.9615
The method follows the same pattern as iteration 1 and so the computations are
not written down in the same detail.
1
0
0.1250
f = 0 G3 =
H3 =
a3 =
1 3
0
0.1250
0.1250
0.6250
The minimum has been achieved; this is expected since the function is quadratic
and it is known that the method converges in n steps for an n -dimensional
quadratic, provided the minimizations are performed exactly.
c Pearson Education Limited 2011
614
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
40(b)
The problem has three variables and is not quadratic. The one-dimensional
minimizations need a numerical procedure; so it is better to use a package such as
MATLAB. The M-file required for the function is
function [f,G]=q40b(z)
t1=z(1)-z(2)+z(3);t2=2 ∗ z(1)+z(3)-2;
f=t1ˆ2+t2ˆ2+z(3)ˆ4;
G(1)=2 ∗ t1+4 ∗ t2;
G(2)=-2 ∗ t1;
G(3)=2 ∗ t1+2 ∗ t2+4 ∗ z(3)ˆ3;
The main DFP segment is
1.
2.
a=[0;0;0];H=eye(3);[f,G]=q40b(a);
fm=@(t,xx,GG)q40b(xx-t ∗ GG);
3.
aa=[a];ff=[f];gg=[G];HH=[H];
4.
for n=1:4
5.
D=H ∗ G;[t,fval]=fminbnd(@ (t)fm(t,a,D),0,2);
6.
aold=a;Gold=G;a=aold-t ∗ D;[f,G]=q40b(a);
7.
h=a-aold;y=G-Gold;
8.
H=H-H ∗ y ∗ y’ ∗ H/(y’ ∗ H ∗ y)+h ∗ h’/(h’ ∗ y);
9.
aa=[aa,a];ff=[ff,f];gg=[gg,G];HH=[HH,H];
10. end
The instructions give the results
aa=
0
0.5853
1.0190
1.0186
1.0184
0
0
0.9813
0.9814
0.9815
0
0.2926
-0.0372
-0.0372
-0.0369
ff=
4.0000
1.0662
0.0000
0.0000
0.0000
gg=
-8.0000
-0.3916
0.0047
-0.0000
-0.0000
0
-4.0000
-1.7558
0.7823
-0.0012
0.0027
-0.0001
-0.0002
-0.0000
-0.0002
The convergence looks good but slows because H tends to a singular matrix.
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
41
615
Evaluating H i+1 y i it is readily shown that equation (10.18) is satisfied.
The update is one that is quite effective but suffers from the problem that the
denominator can become zero when line searches are not exact. A great deal of
remedial action must be taken to ensure that the problem is overcome.
41(a)
The function is quadratic so the solution is obtained in two steps. The
program in Exercise 39, suitably adapted, was used to compute the solution.
Iteration 1
1
2
1
H=
f=9G=
a=
0
8
2
0
1
Iteration 2
0.9948
0.9697
0.4848
H=
f = 0.2424 G =
a=
−0.0619
−0.2424
−0.0606
41(b)
−0.0619
0.2577
The exercise is similar to Exercise 40(b) but with a different function and
updating methods.
function [f,G]=q41b(z)
t1=z(1);t2=z(1)-z(2)+1;
f=t1ˆ2+t2ˆ2+z(2)ˆ2 ∗ z(3)ˆ2;
G(1)=2 ∗ t1+2 ∗ t2;
G(2)=-2 ∗ t2+2 ∗ z(2) ∗ z(3)ˆ2;
G(3)=2 ∗ z(3) ∗ z(2)ˆ2;
In the lines 1,2 and 6 of Exercise 40(b) the M file q41b replaces q40b. In line 8, the
Rank 1 and the BFGS updates replace the one in the program. Also, a different
start point is used. The results for Rank 1 update (i) are
aa=
0.5000
-0.0732
-0.1628
-0.00592
-0.0593
0.5000
0.8344
0.7747
0.9235
0.9234
0.5000
0.4522
0.0525
-0.0640
-0.0640
ff=
1.3125
0.1563
0.0321
0.0073
0.0073
gg=
3.0000
0.0386
-0.2006
-0.0839
-0.0839
-1.7500
0.1564
-0.1207
-0.0270
-0.0270
0.2500
0.6296
0.0630
-0.1091
-0.1092
c Pearson Education Limited 2011
616
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The results for BFGS update (ii) are
aa=
0.5000
-0.0732
0.0969
-0.0726
-0.0348
0.5000
0.8344
0.7384
0.9408
0.9344
0.5000
0.4522
0.0659
0.0141
0.0071
ff=
1.3125
0.1563
0.0389
0.0056
0.0022
gg=
3.0000
0.0386
0.1355
-0.1718
-0.0080
-1.7500
0.2500
0.1564
0.6296
-0.3229
0.0718
0.0271
0.0250
-0.0614
0.0124
Note that BFGS does better than Rank 1 and is the choice of method for the
built-in function fminunc.
42
It is easy to check that equation (10.18) is satisfied but note that the notation
has changed and y,h have been replaced by u,v respectively.
H u = Hu + vpT u − HuqT u
= Hu + v − Hu = v since pT u = qT u = 1
To match with the Davidon formula (10.19) first choose β = α = 0 and hence
pT u = αvT u = 1 and qT u = β uT Hu = 1
Substituting gives
H = H −
HuuT H vvT
+ T
uT Hu
v u
as required.
The formula in Exercise 41(i) is obtained by putting β = β = −α = −α . Thus
p = q = α(V − Hu) and hence pT u = α(v − Hu)T u = 1
Substituting gives the formula
H = H +
(v − Hu)(v − Hu)T
(v − Hu)T u
This whole class of solutions was devised by Huang. Many general results can be
proved for this class and many of the commonly used formulae are included in it.
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
43 (a)
The Fletcher-Reeves method is easily written as a MATLAB segment
global a b p
a=1;b=1;g=[3 ∗ a;b];p=-g;
lam=fminbnd(@ffr,0,2)
% gives
lam = 0.3571
∗
a=a+lam p(1),b=b+lam ∗ p(2),gold=g;g=[3 ∗ a;b]
% gives
∗
p=-g+p (g
% gives
a= -0.0714
∗
g)/(gold
∗
b= 0.6429
g= -0.2143
0.6429
gold)
p= 0.0765
-0.6888
lam=fminbnd(@ffr,0,2)
% gives
lam = 0.9333
a=a+lam ∗ p(1),b=b+lam ∗ p(2),gold=g;g=[3 ∗ a;b]
% gives
a=-4.0246e-016
b=-1.1102e-016
which is the minimum point. The M-file used is
function v=ffr(x)
global a b p
v=3 ∗(a+x ∗ p(1))ˆ2+(b+x ∗ p(2))ˆ2;
43(b)
617
The three-variable problem is handled in a similar manner.
global a b c p
a=0.5;b=0.5;c=0.5;
f=(a-b+1)ˆ2+aˆ2 ∗ bˆ2+(c-1)ˆ2;
g=[2 ∗ (a-b+1)+4 ∗ a ∗ bˆ2;-2 ∗ (a-b+1)+2 ∗ aˆ2 ∗ b;2 ∗ (c-1)];
p=-g;W=[f;a;b;c];
for i=1:5
gold =g;lam=fminbnd(@ffr2,-2,2);
a=a+lam ∗ p(1);b=b+lam ∗ p(2);c=c+lam ∗ p(3);
f=(a-b+1)ˆ2+aˆ2 ∗ bˆ2+(c-1)ˆ2;W=[W[f;a;b;c]];
g=[2 ∗ (a-b+1)+4 ∗ a ∗ bˆ2;-2 ∗ (a-b+1)+2 ∗ aˆ2 ∗ b;2 ∗ (c-1)];
p=-g+p ∗ (g ∗ g)/(gold ∗ gold);
end
c Pearson Education Limited 2011
618
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
W
% gives the sequence of iterations
f 1.3125
0.0764
0.0072
0.0007
0.0004
0.0000
a
0.5000
-0.0950
0.0057
-0.0251
-0.0079
-0.0032
b
0.5000
0.9165
0.9276
0.9633
0.9742
0.9978
c
0.5000
0.7380
0.9674
1.0009
1.0044
1.0014
Review Exercises 10.7
1
The successive tableaux are as follows:
z
x3
x4
x5
x1
−12
1
2
1
z
x3
x1
x5
x1
0
0
1
0
z
x2
x1
x5
x2
−8
1
1
3
x2
−2
0.5
0.5
2.5
x1
0
0
1
0
x3
0
1
0
0
x3
0
1
0
0
x2
0
1
0
0
x4
0
0
1
0
x4
6
−0.5
0.5
−0.5
x3
4
2
−1
−5
x4
4
−1
1
2
x5
0
0
0
1
Soln
0
350
600
900
x5
0
0
0
1
Soln
3600
50
300
600
x5
0
0
0
1
Soln
3800
100
250
350
Hence the solution is read from the table as x1 = 250, x2 = 100 and F = 3800 .
2
Let x1 , x2 , x3 be the numbers of sailboard constructed of types 1,2,3. The
profit is
10x1 + 15x2 + 25x3
and the constraints are
5x1 + 10x2 + 25x3 ≤ 290
3x1 + 2x2 + x3 ≤ 72
10x1 + 20x2 + 30x3 ≤ 400
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
619
The system only has ‘less than’ inequalities so the initial and subsequent tableaux
can be written down immediately.
z
x4
x5
x6
x1
−10
5
3
10
z
x3
x5
x6
x1
−5
0.2
2.8
4
z
x3
x5
x1
x1
0
0
0
1
z
x3
x4
x1
x1
0
0
0
1
x2
−15
10
2
20
x3
−25
25
1
30
x2
−5
0.4
1.6
8
x3
0
1
0
0
x2
5
0
−4
2
x3
0
1
0
0
x2
2.5
0.5
−5
0.5
x3
0
1
0
0
x4
0
1
0
0
x5
0
0
1
0
x4
1
0.04
−0.04
−1.2
x4
−0.5
0.1
0.8
−0.3
x4
0
0
1
0
x5
0
0
1
0
x5
0
0
1
0
x5
0.62
− 0.12
1.25
0.37
x6
0
0
0
1
x6
0
0
0
1
x6
1.25
−0.05
−0.7
0.25
x6
0.81
0.04
− 0.88
−0.01
Soln
0
290
72
400
Soln
290
11.6
60.4
52
Soln
355
9
24
13
Soln
370
6
30
22
The solution is x1 = 22, x2 = 0, x3 = 6 and maximum profit is £ 370.
3
Let x1 , x2 , x3 be the number of standard, super and deluxe cars respectively,
then the profit function is
100x1 + 300x2 + 400x3
and the constraints are
10x1 + 20x2 + 30x3 ≤ 1600
10x1 + 15x2 + 20x3 ≤ 1500
x2 + x3 ≤ 50
x1 + x2 + x3 ≥ 70
c Pearson Education Limited 2011
620
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The problem includes a ‘greater than’ inequality, so the two-phase approach is
needed. An artificial variable is introduced and the artificial cost is introduced in
phase 1. The z row of the tableau is manipulated to bring it to standard form.
z
x4
x5
x6
x8
x1
−1
10
10
0
1
x2
−1
20
15
1
1
x1
0
0
0
0
1
z
x4
x5
x6
x1
x2
0
10
5
1
1
x3
−1
30
20
1
1
x3
0
20
10
1
1
x4
0
1
0
0
0
x4
0
1
0
0
0
x5
0
0
1
0
0
x5
0
0
1
0
0
x6
0
0
0
1
0
x7
1
0
0
0
−1
x6
0
0
0
1
0
x7
0
10
10
0
−1
x8
0
0
0
0
1
x8
1
− 10
− 10
0
1
Soln
−70
1600
1500
50
70
Soln
0
900
800
50
70
A feasible solution has been obtained so the method moves to phase 2. The z row
is first re-calculated
z
x4
x5
x6
x1
x1
0
0
0
0
1
x2
−200
10
5
1
1
z
x3
x5
x6
x1
x1
0
0
0
0
1
x2
−50
0.5
0
0.5
0.5
z
x3
x5
x2
x1
x1
0
0
0
0
1
x2
0
0
0
1
0
x3
−300
20
10
1
1
x3
0
1
0
0
0
x3
0
1
0
0
0
x4
0
1
0
0
0
x5
0
0
1
0
0
x6
0
0
0
1
0
x7
−100
10
10
0
−1
Soln
7000
900
800
50
70
x4
15
0.05
−0.5
−0.05
−0.05
x5
0
0
1
0
0
x6
0
0
0
1
0
x7
50
0.5
5
−0.5
−1.5
Soln
20500
45
350
5
25
x4
10
0.1
−0.5
−0.01
0
x5
0
0
1
0
0
x6
100
−1
0
2
−1
c Pearson Education Limited 2011
x7
0
1
5
−1
−1
Soln
21,000
40
350
10
20
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
621
The solution is read off the table as x1 = 20, x2 = 10, x3 = 40 and the maximum
profit is £21,000.
Note that the (z, x7 ) entry is zero; so a non-unique solution is expected. Interchanging the x3 and x7 entries and completing one further tableau gives the
alternative solution x1 = 60, x2 = 50, x3 = 0 and profit is still £21,000.
The MAPLE solution
with(simplex):
cor3:={10 ∗ x+20 ∗ y+30 ∗ z<=1600,10 ∗ x+15 ∗ y+20 ∗ z<=1500,
y+z<=50,x+y+z>=70};
obr3:=100 ∗ x+300 ∗ y+400 ∗ z;
maximize(obr3,cor3,NONNEGATIVE);
# gives the solution {x=20, z=40,y=10}
MAPLE provides the same solution but does not identify the alternative solution.
Similarly, for the corresponding MATLAB code, which is
f=[-100,-300,-400];A=[10,20,30;10,15,20;0,1,1;-1,-1,-1];b=[1600;1500;50;-70];
options=optimset(‘LargeScale’,‘off’,‘Simplex’,‘on’);
[x,fval]=linprog(f,A,b,[],[],zeros(3,1),[],[],options)
4 Let the student buy x1 kg of bread and x2 kg of cheese, then the cost to be
minimized is
60x1 + 180x2
and the two constraints are
1000x1 + 2000x2 ≥ 3000
25x1 + 100x2 ≥ 100
There are two surplus and two artificial variables in the tableau and the artificial
cost function which has been processed to standard form.
Phase 1
z
x5
x6
x1
−1025
1000
25
x2
−2100
2000
100
x3
1
−1
0
x4
1
0
−1
x5
0
1
0
c Pearson Education Limited 2011
x6
0
0
1
Soln
−3100
3000
100
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
622
z
x5
x2
x1
−500
500
0.25
z
x1
x2
x1
0
1
0
x2
0
0
1
x2
0
0
1
x3
1
−1
0
x4
−20
20
−0.01
x3
0
−0.002
0.0002
x4
0
0.04
− 0.02
x5
0
1
0
x5
1
0.002
− 0.0002
x6
21
−20
0.01
x6
1
− 0.04
0.02
Soln
−1000
1000
1
Soln
0
2
0.5
Phase 1 is completed so the tableau is reconstituted as
z
x1
x2
x1
0
1
0
x2
0
0
1
x3
0.03
− 0.002
0.0002
x4
1.2
− 0.04
0.02
Soln
− 210
2
0.5
It may be noted that the z row entries are positive; so the tableau is optimal and
there is no need to enter phase 2. Thus the minimum cost of the diet is 210p and
is made up of 2 kg of bread and 0.5 kg of cheese.
5
The square of the distance from the origin to the point ( x, y ) is
f = x2 + y 2
so the problem is to optimize this function subject to the condition that it lies on
the curve
g = x2 − xy + y2 − 1 = 0
so
∂g
∂f
+λ
= 2x + λ(2x − y)
∂x
∂x
∂f
∂g
0=
+λ
= 2y + λ(−x + 2y)
∂y
∂y
0=
Subtracting these two equations gives
0 = 2(x − y) + λ(3x − 3y)
⇒
x=y
or
c Pearson Education Limited 2011
λ=−
2
3
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
623
Adding the two equations gives
⇒
0 = 2(x + y) + λ(x + y)
x = −y
or
λ = −2
The first possibility, x = y, gives the points, (1,1) and ( −1,−1) with f = 2 in
each case and the second possibility, x = −y , gives
1 −1
√ ,√
3
3
and
−1 1
√ ,√
3
3
with f =
2
3
2
and λ = −2 reproduce the identical solutions.
3
Although it is not proved, the first solution is the maximum and the second the
In this problem the cases λ = −
minimum.
6
The volume of the solid is
V = x3 + y 3
where the sides of the two cubes are x and y . The surface area has essentially the
area of two faces of the smaller cube removed so
S = 7 = 6x2 + 4y2
The Lagrange multiplier equations are
3x2 + 12xλ = 0
3y2 + 8yλ = 0
The solutions when x = y = 0 can be dismissed since the volume is zero. There
are three other cases
Case 1
x = 0,
−8
y=
λ=
3
Case 2
y = 0,
x = −4λ =
Case 3
x = −4λ,
y=
−8
λ,
3
32
7
7
,V =
4
4
32
7
7
,V =
6
6
⇒
7 = 96λ2 +
c Pearson Education Limited 2011
256 2
λ
9
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
624
3
2
35
±3
√
so the sides have lengths √ and √ and volume
3
(4 10)
10
10
(10) 2
The cases when either x = 0 or y = 0 imply that the problem has collapsed to a
and hence λ =
single cube so these solutions are omitted as geometrically uninteresting.
7
The problem is to maximize the distance
(x − 1)2 + y2 + z2
subject to
2x + y2 + z = 8
The Lagrange equations are
2(x − 1) + 2λ = 0 ⇒ x = 1 − λ
2y + 2yλ = 0 ⇒ y = 0 or λ = −1
−1
λ
2z + λ = 0 ⇒ z =
2
There are two cases
y=0
gives λ =
−12
5
so
x=
17
6
, y = 0, z =
5
5
λ = −1
1
gives x = 2, z =
2
and y = ±
7
2
The first of these possibilities gives the maximum distance.
8
The Lagrange equations for this problem are
1 + 2λx = 0
2 + 2λy = 0
3 + 2λz = 0
and putting back into the constraint gives 2λ = ±1. The local extrema are therefore
at (1,2,3) and (−1, −2, −3) with corresponding F = 14 and F = −14.
c Pearson Education Limited 2011
Glyn James: Advanced Modern Engineering Mathematics, 4th edition
625
To obtain the global extremum all the points on each of the boundaries must be
examined. However, in this case, the geometry is sufficiently simple to establish
√
the result. The region is the inside of the sphere of radius 14 in the region
where all the variables are positive. The cost function comprises a series of parallel
planes, so the global maximum in the region is the local maximum at (1,2,3) and
the global minimum in the region is when all the variables are zero, namely (0,0,0)
and F = 0.
9 (i) We are given that 2s = a + b + c, so maximizing A2 with respect to b and
c, together with this constraint, gives
∂(A2 )
∂(A2 )
= s(s − a)(s − c) + λ = 0,
= s(s − a)(s − b) + λ = 0
∂b
∂c
Clearly b = c and the triangle is isosceles.
(ii) Now, as a so to the above equations we add
∂(A2 )
= s(s − b)(s − c) + λ = 0
∂a
and we see that we must have a = b = c and the triangle is equilateral.
10
We need to consider the problem of maximizing
a 2
π 2
+
=b
r
h
Using the Lagrange multiplier approach we have the two equations
V = πr2 h subject to the constraint
∂V
2a2
= 2πrh − 3 λ = 0
∂r
r
∂V
2π2
= πr2 − 3 λ = 0
∂h
h
together with the constraint itself. Eliminating λ and solving gives
3π2
3a2
and r2 =
b
2b
Note that it has not been proved that these values give the maximum but this can
h2 =
be inferred from physical or geometric reasoning.
c Pearson Education Limited 2011
626
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
11
First notice that as k → ∞ then F → 0 and a careful Taylor expansion
shows that lim F = 0. Take any k > 1, then F takes a positive value so we know
k→1
there must be a maximum for k > 1. We start the bracket procedure with k = 1.1
and initial increment of 0.1.
k
F/A
1.1
0.00721
1.2
0.01258
1.4
0.01962
1.8
0.02556
2.6
0.02602
4.2
0.01995
The maximum has been bracketed by 1.8 ≤ k ≤ 4.2 and the quadratic algorithm
is given in terms of anonymous functions as
qr11=@(x)[x;(log(x)-2∗(x-1)/(x+1))/(x-1)∧2]; %x and the function value
zz=[qr11(1.8),qr11(2.6),qr11(4.2)] % start values
zz =
1.8000
2.6000
4.2000
0.0256
0.0260
0.0200
% repeat from here
p=polyfit(zz(1,:),zz(2,:),2);
xstar=-0.5 ∗ p(2)/p(1);zstar=qr11(xstar);
if zstar(2)>zz(2,2)
if zstar(1)zold(2))&(n μA is rejected.
9
From the data, respective sample averages X̄A = 7281 and X̄B = 6885,
standard deviations SA,n−1 = 419.1 and SB,n−1 = 402.6, and sizes nA = nB = 8,
the pooled estimate of standard deviation is
Sp =
7 × (419.12 + 402.62 )
= 410.9
14
Using t.05,14 = 1.761 and t.025,14 = 2.145, confidence intervals for μA − μB are
2
= (34, 758)
90% : 7281 − 6885 ± 1.761 × 410.9 ×
8
2
95% : 7281 − 6885 ± 2.145 × 410.9 ×
= (−45, 837)
8
The hypothesis that μA = μB is rejected at the 10% level but accepted at the 5%
level.
10
The sample proportion p̂ =
38
540
= 0.0704 and n = 540.
Using z.05 = 1.645 and z.025 = 1.96, confidence intervals for the true proportion
p are
0.0704 × (1 − 0.0704)
= (0.052, 0.089)
540
0.0704 × (1 − 0.0704)
= (0.049, 0.092)
95% : 0.0704 ± 1.96 ×
540
90% : 0.0704 ± 1.645 ×
The hypothesis that p < 0.05 is rejected at the 10% level but accepted at the 5%
level. Alternatively, the test statistic
Z=
0.0704 − 0.05
= 21.8
0.05 × (1 − 0.05)/540
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
642
leads to rejection at both 5% and 10% levels. The test statistic is more accurate
(the confidence interval is approximate).
11
Using z.05 = 1.645, the 90% confidence interval for proportion is
p̂ ± 1.645
p̂(1 − p̂)
n
Thus,
p̂(1 − p̂)
= 0.9
P | p − p̂ |≤ 1.645
n
so with probability 0.9 the maximum error is 1.645 p̂(1 − p̂)/n. Although p̂ is
unknown before the experiment, a figure in the region of 0.25 is expected, hence
we require
0.25 × 0.75
≤ 0.05
1.645
n
from which
n≥
0.25 × 0.75
= 203
(0.05/1.645)2
If in fact n = 200, the sample proportion is p̂ =
confidence interval for p is
0.275 ± 1.645
12
55
200
= 0.275, and the 90%
0.275 × 0.725
= (0.223, 0.327)
200
Using sample proportions p̂1 = 0.31 and p̂2 =
74
150
= 0.493, and z.05 = 1.645
and z.025 = 1.96, confidence intervals for p1 − p2 are
1/2
p̂1 (1 − p̂1 ) p̂2 (1 − p̂2 )
+
90% : p̂1 − p̂2 ± 1.645
= (−0.28, −0.08)
100
150
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
643
1/2
p̂1 (1 − p̂1 ) p̂2 (1 − p̂2 )
= (−0.30, −0.06)
95% : p̂2 − p̂1 ± 1.96
+
100
150
The hypothesis that p1 ≤ p2 − 0.08 is therefore accepted at the 10% level but
rejected at the 5% level.
13
Using sample proportions p̂1 =
30
180
= 0.1667 and p̂2 =
32
500−180
= 0.1, and
z.025 = 1.96, the 95% confidence interval for p1 − p2 is
1/2
p̂1 (1 − p̂1 ) p̂2 (1 − p̂2 )
+
= (0.003, 0.130)
p̂1 − p̂2 ± 1.96
180
320
The hypothesis that p1 > p2 is therefore accepted at the 5% level. Alternatively,
the test statistic
p̂1 − p̂2
1
p̂(1 − p̂) 180
+
Z=
(where p̂ =
30+32
500
1
320
1/2
= 2.17 > z.025
= 0.124) again leads to the hypothesis that p1 > p2 is accepted.
Exercises 11.4.7
14
Y
1
2
3
total
14(a)
1
0
0.20
0.14
0.34
X
2
0.17
0.11
0.25
0.53
3
0.08
0
0.05
0.13
total
0.25
0.31
0.44
1
Marginal distributions of X and Y (summing rows and columns) are
P(X = 1) = 0.34, P(X = 2) = 0.53, P(X = 3) = 0.13
P(Y = 1) = 0.25, P(Y = 2) = 0.31, P(Y = 3) = 0.44
c Pearson Education Limited 2011
644
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
14(b)
P(Y = 3 | X = 2) =
14(c)
P(X = 2 ∩ Y = 3)
0.25
=
= 0.472
P(X = 2)
0.53
The mean and variance of X are given by
E(X) = 1 × 0.34 + 2 × 0.53 + 3 × 0.13 = 1.79
E(X2 ) = 1 × 0.34 + 4 × 0.53 + 9 × 0.13 = 3.63
so σ2X = 3.63 − 1.792 = 0.426
Similarly the mean and variance of Y are given by
E(Y) = 1 × 0.25 + 2 × 0.31 + 3 × 0.44 = 2.19
E(Y2 ) = 1 × 0.25 + 4 × 0.31 + 9 × 0.44 = 5.45
so σ2X = 5.45 − 2.192 = 0.654
The expected value of the product XY is given by
E(XY) =1 × 0 + 2 × 0.17 + 3 × 0.08 + 2 × 0.2 + 4 × 0.11 + 9 × 0
+ 3 × 0.14 + 6 × 0.25 + 9 × 0.05 = 3.79
Hence the correlation coefficient is
3.79 − 1.79 × 2.19
= −0.246
ρX,Y = √
0.426 × 0.654
15
1/2
E(X) =
xdx = 0
−1/2
1/2
3
x3 dx = 0
E(X ) =
−1/2
2
Cov (X, X ) = E(X3 ) − E(X)E(X2 ) = 0
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
16
From cx ≤ y ≤ cx + 1 we infer
y−1
c
≤x≤
y
c
, so
fY (y) =
fX,Y (x, y)dx
⎧
y/c
y
1dx =
⎪
0
⎪
⎨
c
1
1dx
=
1
=
0
⎪
⎪
1
⎩1
1dx = 1 − (y − 1)
(y−1)/c
c
if 0 ≤ y ≤ c
if c ≤ y ≤ 1
if 1 ≤ y ≤ 1 + c
17(a)
1 ∞ ∞ −(x+y)/2
xe
dydx
8 1
1
1 ∞ −x/2
=
xe
[−2e−y/2 ]∞
1 dx
8 1
∞
1
= e−1/2
xe−x/2 dx
4
1
1 −1/2 −x/2 ∞ 1 −1/2
=− e
[xe
]1 + e
2
2
1
1
= e−1 + e−1/2 [−2e−x/2 ]∞
1
2
2
3
= 0.552
=
2e
P(X > 100, Y > 100) =
∞
e−x/2 dx
1
17(b)
fY (y) =
1
8
∞
xe−(x+y)/2 dx =
0
e−y/2
−xe−x/2 |∞
=
0 +
4
Hence
∞
P(Y > 2) =
2
1 −y/2
e
8
∞
∞
xe−x/2 dx
0
e−x/2 dx =
0
1 −y/2
e
2
1 −y/2
1
e
= 0.368
dy = e−y/2 |∞
2 =
2
e
c Pearson Education Limited 2011
645
646
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
18
From the data, if X denotes height and Y denotes weight, X̄ = 174.26,
SX = 7.184, Ȳ = 75.7, SY = 11.703, XY = 13270; so the sample correlation
coefficient is
r=
19
13270 − 174.26 × 75.7
= 0.934
7.184 × 11.703
From the data,
X̄ = 14.23, SX = 2.457, Ȳ = 16.68, SY = 3.4, XY = 243.47
so the sample correlation coefficient is
r=
20
243.47 − 14.23 × 16.68
= 0.732
2.457 × 3.4
Given a sample correlation coefficient r = 0.7 with n = 30, and using
z.025 = 1.96,
c = exp
2 × 1.96
√
27
= 2.126
and the 95% confidence interval for correlation is
21
1 + r − c(1 − r) 1 + r − (1 − r)/c
,
1 + r + c(1 − r) 1 + r + (1 − r)/c
= (0.45, 0.85)
From Exercise 18, r = 0.934 with n = 8. Using z.025 = 1.96,
c = exp
2 × 1.96
√
5
= 5.772
and the 95% confidence interval for correlation is
1 + r − c(1 − r) 1 + r − (1 − r)/c
,
1 + r + c(1 − r) 1 + r + (1 − r)/c
= (0.67, 0.99)
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
647
22 From the data, if X and Y denote mathematics and computer studies marks
respectively,
X̄ = 56.80, SX = 8.880, Ȳ = 54.40, SY = 12.05, XY = 3137.4
so the sample correlation coefficient is
r=
3137.4 − 56.8 × 54.4
= 0.444
8.88 × 12.05
Using z.05 = 1.645,
c = exp
2 × 1.645
√
17
= 2.221
and the 90% confidence interval is
1 + r − c(1 − r) 1 + r − (1 − r)/c
,
1 + r + c(1 − r) 1 + r + (1 − r)/c
= (0.08, 0.70)
Similarly using z.025 = 1.96,
c = exp
2 × 1.96
√
17
= 2.588
and the 95% confidence interval is (0.00, 0.74) . This suggests that the correlation
coefficient is significant at the 10% level but is marginal at the 5% level. The test
statistic
√
Z=
17 1 + 0.444
ln
= 1.968
2
1 − 0.444
leads to a similar conclusion. The ranks of the data are as follows:
Math.
Comp.
3.5
20
2
16.5
13
16.5
13
7
19
3.5
10.5
5
18
10.5
7
7
15
13
9
1
16.5
8
20
4
15
4
18.5
6.5
16.5
9
18.5
14
6.5
10
11.5
11.5
c Pearson Education Limited 2011
2
13
4
1
648
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The rank correlation is rs = 0.401, and the test statistic
√
Z = rs n − 1 = 1.748
This is significant at the 10% level. (The approximate formula for rank correlation
gives a value 0.405, from which Z = 1.767 with the same result.)
23(a)
1
1
1
1
c(1 − y)dydx
fX,Y (x, y)dydx =
0
x
0
x
1
2
c[y − y2 ]1x dx
0
1
1
1 2
c
=
− x + x dx
2
2
0
1
1
1 2 1 3
=c x− x + x
2
2
6
0
c
=
6
Since the joint density must integrate to unity, we must have c = 6.
=
23(b)
P(X <
1
3
,Y > ) =
4
2
1
3/4
fX,Y (x, y)dxdy
1/2
0
1
min{3/4,y}
c(1 − y)dxdy
=
1/2
0
3/4
y
1
3/4
c(1 − y)dxdy +
=
1/2
0
3/4
c(1 − y)dxdy
3/4
1
0
3
c(1 − y)dy
1/2
3/4 4
2
3/4
1
y
3
y3
y2
=c
+ c y−
−
2
3 1/2 4
2 3/4
27
1
1
9
1 3
9
9
−
− +
+ 1− − +
=6
32 192 8 24
2
2 4 32
=
c(1 − y)ydy +
= 0.484
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
23(c)
1
fX (x) =
x
y
1
x2
y2
1
−x+
for 0 ≤ x ≤ 1
c(1 − y)dy = 6 y −
=6
2 x
2
2
c(1 − y)dx = 6(1 − y)y for 0 ≤ y ≤ 1
fY (y) =
0
24
Individual density functions for X and Y are
2 for 29.8 < x < 30.3
0 otherwise
2 for 30.1 < y < 30.6
fY (y) =
0 otherwise
fX (x) =
By independence, the joint density function is therefore
fX,Y (x, y) =
4 for 29.8 < x < 30.3 and 30.1 < y < 30.6
0 otherwise
The required probability is therefore
30.3
min{30.6,x+0.6}
P(0 ≤ Y − X ≤ 0.6) =
4dydx
29.8
30.0
max{30.1,x}
x+0.6
29.8
30.1
30.1
30.6
=
4dydx
30.3
30.6
4dydx +
+
30.0
=4
30.1
30.0
4dydx
30.1
x
30.1
(x − 29.5)dx +
29.8
30.3
0.5dx
30.0
(30.6 − x)dx
+
30.1
2
x
− 29.5x
=4
2
30.0
x2
+ 0.5 × 0.1 + 30.6x −
2
29.8
= 0.84
c Pearson Education Limited 2011
30.3
30.1
649
650
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Alternatively, the shaded area
must be excluded from the
square. The two
parts of
shaded area together form
a square of side 0.2,
so
P(0 ≤ Y − X ≤ 0.6) =
0.5 × 0.5 − 0.2 × 0.2
= 0.84.
0.5 × 0.5
Exercises 11.5.5
25
From the data,
X̄ = 16.370, Sx = 6.789, Ȳ = 36.110, Sy = 14.576, XY = 689.343
Hence
689.343 − 16.37 × 36.11
= 2.13
6.7892
â = 36.11 − 2.13 × 16.37 = 1.22
b̂ =
26
From the data,
X̄ = 6.5, SX = 3.452, Ȳ = 101.5, SY = 50.74, XY = 834.25
Hence, the regression coefficients are
834.25 − 6.5 × 101.5
= 14.64
3.4522
â = 101.5 − 14.64 × 6.5 = 6.315
b̂ =
When load (X) is 15 kg , a deflection of 226 mm is predicted.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
27
651
From the results in Example 11.17 we find â = −2.294 and b̂ = 0.811, and
when 15 V is measured a tension of 9.88 kN is predicted.
Using t.025,12 = 2.179 and
SE =
S2Y − b̂2 S2X =
√
16.25 − 0.8112 × 34.51 = 0.360
(remember that X and Y are essentially reversed compared with Example 11.17),
the 95% confidence interval for tension when 15 V are measured is
1 + (15 − 12.07)2 /24.51
= (9.62, 10.14)
9.88 ± 2.179 × 0.36 ×
12
28(a)
From the data,
X̄ = 34.17, SX = 11.70, Ȳ = 453.8, SY = 59.34, XY = 15944
and the regression coefficients are
15944 − 34.17 × 453.8
= 3.221 (using unrounded figures)
11.72
â = 453.8 − 3.221 × 34.17 − 343.7
b̂ =
For advertising x = 6 (£ 6000), sales of £ 537,000 are predicted.
28(b)
SE =
√
59.342 − 3.2212 × 11.72 = 45.8 and using t.025,10 = 2.228 the
95% confidence interval for regression slope is
b̂ ± t.025,10
SE
√
SX 10
= (0.46, 5.98)
The hypothesis that b = 0 is rejected at the 5% level.
28(c)
The 95% confidence interval for mean sales when x = 60 is
537 ± 2.228 × 45.8 ×
1 + (60 − 34.17)2 /136.8
= (459, 615)
10
c Pearson Education Limited 2011
652
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
29
From the data,
X̄ = 11.5, SX = 2.291, Ȳ = 13.25, SY = 2.99, XY = 158.38
so the regression coefficients are
158.38 − 11.5 × 13.25
= 1.143
2.2912
â = 13.25 − 1.143 × 11.5 = 0.107
b̂ =
and Y = 16.1 is predicted when x = 14. Also using
SE = S2Y − b̂2 S2X = 1.442
and t.05,6 = 1.943, the 90% confidence interval for mean number of defectives per
hour when x = 14 is
16.1 ± 1.943 × 1.442 ×
30
1 + (14 − 11.5)2 /5.25
= (14.4, 17.8)
6
Given a model (with no constant) of form
Yi = bXi + Ei
and minimizing
Q=
n
[Yi − bXi ]2
i=1
we have
hence
n
dQ
= −2
Xi [Yi − b̂Xi ] = 0
db
i=1
n
Xi Y i
b̂ = i=1
n
2
i=1 Xi
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
31
If X denotes voltage and Y = X/R denotes current then from the data,
Σi Xi Yi = 5397, Σi X2i = 650
so that (using the result of the previous exercise)
b̂ = 8.30
The estimated resistance is R =
32
1000
= 120Ω .
8.3
Taking logs we have
ln Pi + λ ln Vi = ln C
This is of the form Yi = a + bXi with
Yi = ln Pi , a = ln C, b = −λ, Xi = ln Vi
From the data,
X̄ = 4.272, SX = 0.254, Ȳ = 3.423, XY = 14.452
so that
b̂ = −2.664 and â = 14.80
Hence
Ĉ = eâ = 2.69 × 106 and λ̂ = −b̂ = 2.66
When V = 80cm 3 , a pressure P = 22.9kg/cm 3 is predicted.
33
Taking logs we have
ln Yi = ln a + b ln Xi
or Yi = a + bXi
From the data,
X̄ = 6.183, SX = 1.515, Y¯ = 2.377, X Y = 12.266
c Pearson Education Limited 2011
653
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
654
so
b̂ = −1.059, â = 8.927 and â = 7533
For a lot size X = 300, a unit cost of 17.9 is predicted.
Exercises 11.6.3
Under the hypothesis, P(A) = 47 , P(B) = 27 , P(C) =
34
observed
63
22
15
A
B
C
Total
probability
4/7
2/7
1/7
expected
57.14
28.57
14.29
1
7
.
χ2 contribution
0.601
1.511
0.035
2.147
No parameters were estimated ( t = 0), so we compare χ2 = 2.147 with χ20.05,2 =
5.991. The hypothesis is accepted.
35
The total number of books is 640, so a uniform number would be 128 per
day. Hence
Obs. (fk ) :
Exp. (ek ) :
153
108
120
114
145
128
128
128
128
128
χ2
4.9
3.1
0.5
1.5
2.3
χ
contribution:
The total chi-square value is
χ2 = 12.3
and this is greater than χ2.05,4 = 9.49 (significant at 5% , but not significant at
1%).
36
The observed mean number of flaws per sample is (12 + 6 × 2)/50 or 0.48.
Setting λ = 0.48, the Poisson probabilities are given by λk e−λ /k! and hence
expected values are as follows:
number
of flaws
0
1
≥ 2
Obs. (fk )
32
12
6
probability
0.619
0.297
0.084
exp. (ek )
30.9
14.9
4.2
c Pearson Education Limited 2011
χ2 contribution
0.036
0.547
0.761
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
655
The total chi-square value is
χ2 = 1.35
which is very small, so the Poisson hypothesis is accepted.
37
The observed average number of α-particles per time interval (taking class
> 10 as 11 for this calculation) is
(203 + 2 × 383 + . . . + 11 × 6)/(57 + 203 + . . . + 6) = 3.87
Using this value for λ in the Poisson probabilities and proceeding as in the previous
exercise a total chi-square value
χ2 = 12.97
is obtained. One parameter has been estimated and used for predicting the expected values, so the comparison is with χ2.05,10 = 18.3, and the Poisson hypothesis
is accepted.
38
Using the measured average and standard deviation, probabilities can be
obtained from the normal table as follows:
X − 10
6.5 − 10
<
= Φ(−1.75) = 1 − Φ(1.75) = 0.0401
2
2
P(X < 7.5) = 1 − Φ(1.25) = 0.1056
P(X < 6.5) = P
P(X < 8.5) = 1 − Φ(0.75) = 0.2266
P(X < 9.5) = 1 − Φ(0.25) = 0.4013
P(X < 10.5) = Φ(0.25) = 0.5987
P(X < 11.5) = Φ(0.75) = 0.7734
P(X < 12.5) = Φ(1.25) = 0.8944
P(X < 13.5) = Φ(1.75) = 0.9599
P(X < 13.5) = 0.0401
c Pearson Education Limited 2011
656
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
The class probabilities can now be inferred by
P(6.5 < X < 7.5) = P(X < 7.5) − P(X < 6.5) = 0.0655
and so on; hence, the following table:
Class
< 6.5
6.5–7.5
7.5–8.5
8.5–9.5
9.5–10.5
10.5–11.5
11.5–12.5
12.5–13.5
> 13.5
Probability
0.0401
0.0655
0.1210
0.1747
0.1974
0.1747
0.1210
0.0655
0.0401
Expected
4.01
6.55
12.10
17.47
19.74
17.47
12.10
6.55
4.01
Sample 1
4
6
16
16
17
20
12
6
3
Sample 2
3
6
16
13
26
7
19
5
5
For sample 1, χ2 = 2.48 which is not significant.
For sample 2, χ2 = 15.51 which exceeds χ2.025,6 = 14.45 (significant at 2.5% level)
but does not exceed χ2.01,6 = 16.81 (not significant at 1% level). The second
subscript is m − t − 1 with m = 9 (classes) and t = 2 (parameters estimated).
39
A
B
C
Total
The contingency table is as follows:
Perfect
89 (89.04)
62 (58.88)
119 (122.07)
270
Intermediate
23 (21.44)
12 (14.18)
30 (29.39)
65
Unacceptable
12 (13.52)
8 (8.94)
21 (18.54)
41
The expected values are shown in brackets,
270 × 124
= 89.04
376
c Pearson Education Limited 2011
Total
124
82
170
376
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
657
and so on. Hence
(89 − 89.04)2
(21 − 18.54)2
+ ··· +
89.04
18.54
= 1.30
χ2 =
This is less than χ2.05,(3−1)(3−1) = 9.49, so there is no significant difference in
quality.
40
The contingency table (with the expected values in brackets) is as follows:
o.k.
442
536
544
397
593
442
434
195
438
594
585
541
5741
(441.6)
(539.8)
(539.8)
(392.5)
(588.8)
(441.6)
(441.6)
(196.3)
(441.6)
(588.8)
(588.8)
(539.8)
defective
8
14
6
3
7
8
16
5
12
6
15
9
109
(8.38)
(10.25)
(10.25)
(7.45)
(11.18)
(8.38)
(8.38)
(3.73)
(8.38)
(11.18)
(11.18)
(10.25)
total
450
550
550
400
600
450
450
200
450
600
600
550
5850
(442 − 441.6)2
(9 − 10.25)2
+ ··· +
= 20.56 which exceeds χ2.05,11 =
441.6 2
10.25
19.68 but is less than χ.025,11 = 21.92. The variation is significant at the 5%
Hence, χ2 =
level. For a 2 × c table of this form, effectively, a comparison of c proportions, it
is quicker (and equivalent) to compute
c
(fj − nj p̂)2
χ =
nj p̂(1 − p̂)
j=1
2
where fj is the number of defectives in column j (total nj ) and p̂ =
c
j=1
is the overall sample proportion of defectives.
c Pearson Education Limited 2011
fj /
c
j=1
nj
658
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
41
The contigency table (with expected values in brackets and adjusted residuals
underneath) is as follows:
Spending
level
Low
Medium
High
Total
Jacket
Shirt
Trousers
Shoes
Total
21(36)
−3.0
66(66)
0.0
58(43)
2.9
145
94(92)
0.4
157(169)
−1.5
120(110)
1.3
371
57(47)
1.8
94(87)
1.0
41(57)
−2.8
192
113(110)
0.4
209(204)
0.7
125(133)
− 1.1
447
285
526
344
1155
Chi-square = 20.7, d.f. = (4 − 1)(3 − 1) = 6, so compare with χ20.005,6 = 18.5 :
significant at 0.5% level. High-spending customers are tending to buy more of the
jacket and less of the trousers. For low-spending customers, it is the other way
round.
42
If p is the proportion requiring adjustments, the number of such sets in
a sample of size n is binomial with parameters n, p. With n = 4, to test the
hypothesis that p = 0.1 we have
P(0) = 0.94 = 0.6561
4
P(1) =
0.93 0.1 = 0.2916
1
4
P(2) =
0.92 0.12 = 0.0486
2
4
P(3) =
0.9 × 0.13 = 0.0036
3
Hence the following table:
k
0
1
2
3
fk
110
73
16
1
pk
0.6561
0.2916
0.0486
0.0036
ek = 200pk
131.22
58.32
9.72
0.72
χ2 contribution
3.43
3.70
4.06
0.11
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
659
The total chi-square value is χ2 = 11.30, which is significant at 5% (χ2.05,3 = 7.82)
but not quite significant at 1% (χ2.01,3 = 11.35) .
Using proportions, the total number of sets requiring adjustments is 73 + 2 × 16 +
3 × 1 = 108, so
108
= 0.135
p̂ =
800
Using z.025 = 1.96 and z.005 = 2.576, confidence intervals for the proportion p of
sets requiring adjustments are
p̂(1 − p̂)
= (0.111, 0.159)
800
p̂(1 − p̂)
= (0.104, 0.166)
p̂ ± 2.576
800
p̂ ± 1.96
95% :
99% :
This indicates that p > 0.1, significant (just) at the 1% level.
Exercises 11.7.4
43
We must have
∞
1=
∞
fX (x)dx = c
0
xe−2x dx
0
1
1
= − c[xe−2x ]∞
0 + c
2
2
∞
1
1
= c − e−2x
2
2
0
c
=
4
∞
e−2x dx
0
so c = 4. The m.g.f. is then
∞
MX (t) =
∞
tx
e fX (x)dx = 4
0
xe(t−2)x dx
0
4
4
[xe(t−2)x ]∞
0 −
t−2
t−2
4
provided t < 2
=
(t − 2)2
∞
=
0
c Pearson Education Limited 2011
e(t−2)x dx
660
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Hence
8
=−
=1
E(X) =
3
(t − 2) t=0
24
3
=
E(X2 ) = MX (0) =
4
(t − 2) t=0
2
1
Var(X) = E(X2 ) − [E(X)]2 =
2
MX (0)
44
If X1 , . . . , Xn are independent Poisson random variables with parameters
λ1 , . . . , λn and if Y = X1 + . . . + Xn then
MY (t) = MX1 (t) . . . MXn (t)
= exp[λ1 (et − 1)] . . . exp[λn (et − 1)]
= exp[(λ1 + . . . + λn )(et − 1)]
Hence, Y is another Poisson random variable with parameter λ = λ1 + . . . + λn .
45
Numbers of breakdowns in one hour are separately binomial with parameters
n1 = 30, p1 = 0.01 and n2 = 40, p2 = 0.005 respectively, and hence approximately
Poisson with parameters λ1 = 0.3 and λ2 = 0.2 respectively. The total number of
breakdowns per hour is therefore also approximately Poisson with λ = λ1 + λ2 =
0.5, and
46
λ2
= 0.014
P (three or more) 1 − e−λ 1 + λ +
2!
Let the proportion defective be p. By the Poisson approximation,
P (k defective in 100)
λk e−λ
k!
where λ = 100p. The requirement is
P (k ≤ 1) e−λ (1 + λ) > 0.9
from which λ < 0.531 (solving this as a nonlinear equation) and hence p <
0.531
100
0.0053. Therefore, at least 99.47% of servomechanisms must be satisfactory.
c Pearson Education Limited 2011
=
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
47
661
With
2
1
fZ (z) = √ e−z /2 , −∞ < z < ∞
2π
the m.g.f. is
1
MZ (t) = √
2π
1
=√
2π
2
= et
/2
∞
−∞
∞
−∞
∞
−∞
etz e−z
1
e− 2 (z
2
2
/2
dz
−2tz+t2 ) t2 /2
e
dz
1
2
1
√ e− 2 (z−t) dz
2π
The integrand is the p.d.f. of a normal random variable with mean t and standard
deviation equalling one, hence
2
MZ (t) = et
/2
Exercises 11.9.7
48
From the table (Figure 11.29), with n = 50, p = 0.1 so that np = 5, we read
off the Shewhart warning limit as 9.5 and the action limit as 13.5. For the given
data, the warning limit is exceeded at samples 3, 9 and 11, and the action limit is
exceeded at sample 12. The upper control limit is given by
UCL = np + 3 np(1 − p) = 11.4
and this is first exceeded at sample 9.
49
From Figure 11.29, with n = 100, p = 0.02 so that np = 2, we have Shewhart
warning and action limits 5.5 and 7.5 respectively. The warning limit is exceeded
three times (samples 20, 22 and 25) before the action limit is exceeded at sample
28. The upper control limit is given by
UCL = np + 3 np(1 − p) = 6.2
and this is first exceeded at sample 25.
c Pearson Education Limited 2011
662
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
50
Using σ = 3 and n = 10, the Shewhart warning and action limits are
√
cW = 1.96σ/ n = 1.86
√
cA = 3.09σ/ n = 2.93
(above and below the mean). Relative to the mean, the warning limit is exceeded
at samples 3 and 9, and the action limit at sample 12.
51
Using σ = 3 and n = 10, the Shewhart warning and action limits above the
design mean μ = 12 are
√
μ + 1.96σ/ n = 13.86
√
μ + 3.09σ/ n = 14.93
The warning limit is exceeded several times (at samples 6, 12, 15, 17 and 18) before
the action limit is crossed at sample 19.
52(a)
Using σ = 3 and n = 10, the cusum parameters are
σ
r = √ = 0.474 (relative to the mean)
2 n
σ
h = 5 √ = 4.74
n
The chart is built up in the following table:
value
cusum
-0.2 1.3 2.1
0 0.83 2.45
0.3 -0.8
2.28 1.00
1.7 1.3 0.6 2.5 1.4 1.6 3.0
2.23 3.05 3.18 5.21 6.13 7.26 9.78
The decision interval (h) is exceeded at sample 9.
52(b)
Using r = 0.3 and σ, n as above, the GMA action limits are
r
σ
√ = ±1.23
±3.09
2−r n
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
663
(relative to the mean). The chart is built up in the following table:
value
GMA
-0.2 1.3 2.1 0.3 -0.8 1.7 1.3 0.6 2.5 1.4 1.6 3.0
-0.06 0.35 0.87 0.70 0.25 0.69 0.87 0.79 1.30 1.33 1.41 1.89
The action limit is exceeded at sample 9.
53
For the cusum chart we have μ = 12, σ = 3 and n = 10, so
σ
r = μ + √ = 12.47
2 n
σ
h = 5 √ = 4.74
n
For the GMA chart with r = 0.3 we have action limits at
r
σ
√ = 10.77 and 13.23
μ ± 3.09
2−r n
The control charts are built up in the following table:
X̄m
cusum
GMA
X̄m
cusum
GMA
12.8 11.2 13.4 12.1 13.6 13.9 12.3 12.9 13.8 13.1
0.33
0 0.93 0.55 1.68 3.10 2.93
3.35 4.68 5.31
12.24 11.93 12.37 12.29 12.68 13.05 12.82 12.85 13.13 13.12
12.9 14.0 13.7 13.4 14.2 13.1 14.0 14.0 15.1 14.3
5.73 7.26 8.48 9.41 11.13 11.76 13.28 14.81 17.44 19.26
13.06 13.34 13.45 13.43 13.66 13.49 13.65 13.75 14.16 14.20
The cusum chart indicates action at sample 10, the GMA chart at sample 12.
54
Using n = 50, p = 0.1 so that np = 5, the cusum parameters from Figure
11.31 are r = 7, h = 8.5 (nearest values).
count
cusum
5 8 11 5 6 4 9 7 12 9 10
0 1 5 3 2 0 2 2 7 9 12
14
19
The chart indicates action at sample 10.
c Pearson Education Limited 2011
664
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
55
Using n = 100, p = 0.02 so that np = 2, the cusum parameters from Figure
11.31 are r = 3, h = 5.5 (nearest values).
count
cusum
count
cusum
3
0
3
6
3
0
4
7
5 3 5 0 3 1 3 5 4
2 2 4 1 1 0 0 2 3
5 6 5 6 4 4 7 5 4
9 12 14 17 18 19 23 25 26
2 4 3 5 4
2 3 3 5 6
8 5 6 6 7
31 33 36 39 43
The chart indicates action at sample 16.
56 For the Shewhart chart we have n = 12, σ = 1, and hence warning and action
limits given by
√
cW = 1.96σ/ n = 0.57
√
cA = 3.09σ/ n = 0.89
For the cusum chart we have
σ
r = √ = 0.144 (relative to the mean)
2 n
σ
h = 5 √ = 1.443
n
For the GMA chart with r = 0.2, the action limit is
r
σ
√ = 0.297
3.09
2−r n
X̄m
cusum
GMA
X̄m
cusum
GMA
X̄m
cusum
GMA
0.1
0
.020
0.1
.623
.197
0.5
3.535
.379
0.3
.156
.076
0.4
.878
.238
0.7
4.091
.443
-0.2
0
.021
0.6
1.334
.310
0.3
4.246
.415
0.4
.256
.097
0.3
1.490
.308
0.1
4.202
.352
0.1
.211
.097
0.4
1.745
.326
0.6
4.658
.401
0
.067
.078
0.3
1.901
.321
0.5
5.013
.421
0.2
.123
.102
0.6
2.357
.377
0.6
5.469
.457
-0.1
0
.062
0.5
2.712
.402
0.7
6.025
.505
0.2
.056
.089
0.4
2.968
.401
0.4
6.280
.484
0.4 0.5
.311 .667
.152 .221
0.2 0.3
3.024 3.179
.361 .349
0.5
6.636
.488
For the Shewhart chart there are several warnings but no action indicated. For the
cusum and GMA charts, action is indicated at samples 15 and 14 respectively.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
57
665
The GMA Sm is defined recursively by
S0 = μX
Sm = rX̄m + (1 − r)Sm−1 for m ≥ 1
Substituting for Sm−1 , then Sm−2 and so on gives
Sm = rX̄m + (1 − r)[rX̄m−1 + (1 − r)Sm−2 ]
= r[X̄m + (1 − r)X̄m−1 ] + (1 − r)2 [rXm−2 + (1 − r)Sm−3 ]
and eventually
Sm = r
m−1
(1 − r)i X̄m−i + (1 − r)m μX
i=0
But E(X̄m−i ) = μX and Var(X̄m−i ) =
Using the result
m−1
xi =
i=0
we have
E(Sm ) = rμX
σ2X
for all i, m .
n
1 − xm
for | x |< 1
1−x
m−1
(1 − r)i + (1 − r)m μX
i=0
1 − (1 − r)m
+ (1 − r)m μX
= rμX
r
= μX
m−1
σ2X 2
σ2 r2 1 − (1 − r)2m
Var(Sm ) =
[(1 − r)2 ]i = X
r
n
n 1 − (1 − r)2
i=0
σ2X
r
[1 − (1 − r)2m ]
n
2−r
2
r σX
as m → ∞
−→
2−r n
=
58
Let Xm = count of defectives for sample m , n = sample size, p = probability
of defective. Define
S0 = np
Sm = rXm + (1 − r)Sm−1 for m ≥ 1
c Pearson Education Limited 2011
666
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Substituting for Sm−1 , Sm−2 and so on (as in the previous exercise) leads to
Sm = r
m−1
(1 − r)i Xm−i + (1 − r)m np
i=0
From the mean and variance of the binomial distribution,
E(Xm−i ) = np
Var(Xm−i ) = np(1 − p)
hence
E(Sm ) = rnp
m−1
(1 − r)i + (1 − r)m np
i=0
= np[1 − (1 − r)m + (1 − r)m ] = np
Var(Sm ) = np(1 − p)r
2
m−1
[(1 − r)2 ]i
i=0
r
[1 − (1 − r)2m ]np(1 − p)
=
2−r
r
np(1 − p) as m → ∞
−→
2−r
The upper control limit for n = 50, p = 0.05, r = 0.2 is
r
np(1 − p) = 4.04
UCL = np + 3
2−r
Xm
Sm
Xm
Sm
3
2.6
7
4.36
5
3.08
4
4.29
2
2.86
5
4.43
2
2.69
5
4.55
1
2.35
8
5.24
6
3.08
6
5.39
4
3.27
5
5.31
4
3.41
9
6.05
2
3.13
7
6.24
6
3.70
8
6.59
The chart indicates action after 11 samples.
59 For the Shewhart chart we have n = 10, μ = 6, σ = 0.2 and hence warning
and action limits given by
√
cW = μ ± 1.96σ/ n = 5.88 and 6.12
√
cA = μ ± 3.09σ/ n = 5.80 and 6.20
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
667
For the cusum chart we have
σ
r = μ + √ = 6.032
2 n
σ
h = 5 √ = 0.316
n
For the GMA chart with r = 0.2, the action limits are
r
σ
√ = 5.935 and 6.065
μ ± 3.09
2−r n
X̄m
cusum
GMA
X̄m
cusum
GMA
X̄m
cusum
GMA
6.04
0.01
6.01
6.17
0.28
6.07
6.10
0.65
6.06
6.12
0.10
6.03
6.03
0.28
6.06
6.01
0.62
6.05
5.99
0.06
6.02
6.13
0.38
6.07
5.96
0.55
6.03
6.02
0.04
6.02
6.05
0.40
6.07
6.12
0.64
6.05
6.04
0.05
6.03
6.17
0.54
6.09
6.02
0.63
6.05
6.11
0.13
6.04
5.97
0.47
6.07
6.20
0.80
6.08
5.97
0.07
6.03
6.07
0.51
6.07
6.11
0.88
6.08
6.06
0.10
6.03
6.14
0.62
6.08
5.98
0.82
6.06
6.05
0.12
6.04
6.03
0.62
6.07
6.02
0.81
6.05
6.06
0.14
6.04
5.99
0.58
6.05
6.12
0.90
6.07
The Shewhart chart indicates action after 26 samples, the cusum chart after 13
samples and the GMA chart after 11 samples.
Exercises 11.10.6
60
If gales occur at rate
15
12
= 1.25 per month, and occur independently, then
the number of gales in any one month has a Poisson distribution, so
P (more than two in one month) = 1 − P(0) − P(1) − P(2)
λ2 T2
−λT
1 + λT +
=1−e
2!
= 0.132
(with λ = 1.25, T = 1).
c Pearson Education Limited 2011
668
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
61
If λ = 30 calls per hour (on average) arrive at the switchboard, and do so
independently, then the number of calls has a Poisson distribution. Hence
P (no calls in T = 3 min) = e−λT /60 = 0.223
P (more than five calls in T = 5 min)
−λT /60
=1−e
(λT)5
λT
+ ... + 5
1+
60
60 5!
= 0.042
62
The steady-state distribution for the number (N) in the system is
P(n in system) = pn = (1 − ρ)ρn , n ≥ 0
Now
∞
∞
d n n−1
d
(1 − ρ)−1 = (1 − ρ)−2
ρ =
nρ
=
dρ n=0
dρ
n=0
Hence, the mean number in the system is
NS =
∞
n(1 − ρ)ρ = ρ(1 − ρ)
n
n=0
∞
nρn−1
n=0
ρ(1 − ρ)
ρ
=
=
(1 − ρ)2
1−ρ
If there are n > 0 customers in the queue then there are n + 1 in the system, so
P(n > 0 in queue) = (1 − ρ)ρn+1 , n ≥ 1
(we do not need the probability for n = 0). Mean number in queue is
NQ = (1 − ρ)
∞
nρ
n+1
= ρ (1 − ρ)
n=1
∞
n=0
ρ (1 − ρ)
ρ
=
2
(1 − ρ)
1−ρ
2
=
2
2
c Pearson Education Limited 2011
nρn−1
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
63
For a single-channel queue with λ = 3 arrivals per hour and μ = 4 patients
treated per hour,
63(a)
P(0 in system) = 1 −
63(b)
NQ =
1
λ
=
μ
4
(λ/μ)2
9
=
1 − λ/μ
4
63(c)
P(> 3 in queue) = P(> 4 in system)
= 1 − P(0) − P(1) − P(2) − P(3) − P(4)
λ λ 2 λ 3 λ 4
λ
1+ +
+
+
=1− 1−
μ
μ μ
μ
μ
= 0.237
3
λ/μ
= hour.
μ−λ
4
63(d)
WQ =
63(e)
P (wait more than one hour) =
64
669
λ −(μ−λ)
= 0.276
e
μ
Mean number of aircraft on ground is
NS =
λ
μ−λ
so, total mean cost per hour (waiting time plus servicing) is
E [total cost per hour] =
c1 λ
+ c2 μ
μ−λ
c Pearson Education Limited 2011
670
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
Minimizing this with respect to μ:
d
c1 λ
E [total cost per hour] = −
+ c2 = 0
dμ
(μ − λ)2
from which
(μ − λ)2 =
so μ = λ +
65
c1 λ
c2
c1 λ/c2
Breakdown rate is λ = 3 per hour, and machine idle time is costed at £ 60
per hour per machine. For option A, service rate is μ = 4 per hour at £ 20 per
hour, so mean hourly cost is
60NS + 20 =
60λ
+ 20 = 200
μ−λ
For option B, service rate is μ = 5 at £ 40 per hour, so mean hourly cost is
60NS + 40 =
60λ
+ 40 = 130
μ−λ
Option B is preferred.
66
Ship arrival rate is λ =
1
3
per hour, and service rate per berth is μ =
1
12
, so
ρ = λ/μ = 4. Mean waiting time in the queue is
−1
c−1 n
ρ
ρc+1
ρc
1
+
WQ =
λ (c − 1)!(c − ρ)2 n=0 n!
(c − 1)!(c − ρ)
where c is the number of berths. For c = 5 berths we find WQ = 6.65 hours,
which exceeds the required minimum, so c = 6 berths are needed (WQ = 1.71) .
67
Arrival rate is λ = 2 per minute, and basic service rate per cashier is μ =
5
4
per minute. If this service rate is doubled (by providing a packer) then mean
queueing time is
4/5
ρ
=
= 1.6 min
WQ =
μ−λ
5/2 − 2
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
671
Alternatively, if an additional cash desk is provided then (using WQ as in the
previous exercise, and ρ = 8/5)
−1
ρ3
ρ2
1
1+ρ+
= 0.076 min
WQ =
λ (2 − ρ)2
2−ρ
Clearly, a second cash desk is preferable.
Exercises 11.11.3
68
We have
P(agent) = P(agent|option 1)P(option 1) + P(agent|option 2)P(option 2)
+ P(agent|option 3)P(option 3)
= 0.28 × 0.45 + 0.41 × 0.32 + 0.16 × 0.23
= 0.294
69
Total probability of explosion is
P(E) = P(E | (a))P((a)) + P(E | (b))P((b))
+ P(E | (c))P((c)) + P(E | (d))P((d))
= 0.25 × 0.2 + 0.2 × 0.4 + 0.4 × 0.25 + 0.75 × 0.15
= 0.3425
Hence, by Bayes’ Theorem,
P((a) | E) = P(E | (a))P((a))/P(E) = 0.146
P((b) | E) = 0.234
P((c) | E) = 0.292
P((d) | E) = 0.328
and sabotage is therefore the most likely cause of the explosion.
c Pearson Education Limited 2011
672
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
70
If two bullets hit the target then they could be fired by each possible pair of
marksmen, so
P (two hits) = P(A ∩ B ∩ C̄) + P(A ∩ B̄ ∩ C) + P(Ā ∩ B ∩ C)
(where A denotes ‘bullet from A hits target’, etc)
= 0.6 × 0.5 × 0.6 + 0.6 × 0.5 × 0.4 + 0.4 × 0.5 × 0.4
= 0.38
Hence
P(C | two hits) =
P(C ∩ two hits)
P (two hits)
P(A ∩ B̄ ∩ C) + P(Ā ∩ B ∩ C)
P ( two hits)
0.6 × 0.5 × 0.4 + 0.4 × 0.5 × 0.4
=
0.38
= 0.526
=
Thus, it is more probable than not that C hit the target.
71
Prior probabilities are P(A) = 13 , P(B) =
2
3
. Also P( Smith | A) = 0.1 and
P( Smith | B) = 0.05. Hence
P (Smith | A)P(A)
P (Smith | A)P(A) + P (Smith | B)P(B)
0.1 × 13
=
0.1 × 13 + 0.05 × 23
1
=
2
P(A | Smith) =
72
Let D denote ‘has disease’ and + denote ‘positive diagnosis’, so that P(D) =
0.08, P(+ | D) = 0.95 and P(+ | D̄) = 0.02
72(a)
P(+) = P(+ | D)P(D) + P(+ | D̄)P(D̄)
= 0.95 × 0.08 + 0.02 × 0.92 = 0.0944
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
72(b)
73
P(D | +) =
673
0.95 × 0.08
P(+ | D)P(D)
=
= 0.81
P(+)
0.0944
Let G denote ‘good stock’, B = Ḡ denote ‘bad stock’,
g denote ‘stockbroker says good’,
b denote ‘stockbroker says bad’,
so that P(G) = 0.5, P(g | G) = 0.6, P(b | B) = 0.8.
73(a)
P(g | G)P(G)
P(g | G)P(G) + P(g | B)P(B)
0.6 × 0.5
3
=
=
0.6 × 0.5 + 0.2 × 0.5
4
P(G | g) =
73(b)
Let E denote ‘k out of n stockbrokers say good’. Since the stockbrokers
are independent, by the binomial distribution
n
P(E | G) =
[P(g | G)]k [P(b | G)]n−k
k
n
P(E | B) =
[P(g | B)]k [P(b | B)]n−k
k
Hence
P(E | G)P(G)
P(E | G)P(G) + P(E | B)P(B)
n
k
G)]n−k P(G)
k [P(g | G)] [P(b|
= n
k
n−k P(G) + n [P(g | B)]k [P(b | B)]n−k P(B)
k [P(g | G)] [P(b | G)]
k
P(G | E) =
0.6k × 0.4n−k × 0.5
0.6k × 0.4n−k × 0.5 + 0.2k × 0.8n−k × 0.5
−1
1 k n−k
= 1+
2
3
=
c Pearson Education Limited 2011
674
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
74
Given that the probability of correct reception of a letter is 0.6, and the error
probabilities are 0.2 for the two alternatives, we have
P(ABCA received | AAAA transmitted) = 0.6 × 0.2 × 0.2 × 0.6 = 0.0144
P(ABCA received | BBBB transmitted) = 0.2 × 0.6 × 0.2 × 0.2 = 0.0048
P(ABCA received | CCCC transmitted) = 0.2 × 0.2 × 0.6 × 0.2 = 0.0048
Also P(AAAA transmitted) = 0.3 etc. Hence
P(ABCA received) = 0.0144 × 0.3 + 0.0048 × (0.4 + 0.3)
= 0.00768
and
0.0144 × 0.3
= 0.5625
0.00768
P(BBBB transmitted | ABCA received) = 0.25
P(AAAA transmitted | ABCA received) =
P(CCCC transmitted | ABCA received) = 0.1875
75
Average number of accidents per day =
1×12+2×4
100
=
1
5
First hypothesis (H1 ) is for a Poisson distribution, so set λ =
pi = P(i accidents in one day) =
1
5
and probabilities
λi e−λ
i!
Hence, p0 = 0.8187, p1 = 0.1637, p2 = 0.0164. Second hypothesis (H2 ) is for a
binomial distribution with n = 3, so set
np =
1
1
(hence p =
) and probabilities
5
15
3 i
p (1 − p)3−i
qi = P(i accidents in one day) =
i
Hence, q0 = 0.8130, q1 = 0.1742, q2 = 0.0124. If E denotes the evidence then the
odds are updated by
ln
P(E | H1 )
P(H1 )
P(H1 | E)
= ln
+ ln
P(H2 | E)
P(E | H2 )
P(H2 )
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
675
where
12 4
P(E | H1 ) = p84
0 p1 p2
12 4
P(E | H2 ) = q84
0 q1 q2
1
P(H1 )/P(H2 ) = (initial odds)
2
Hence
ln
p0
p1
p2
1
P(H1 | E)
= 84 ln + 12 ln + 4 ln + ln = 0.247
P(H2 | E)
q0
q1
q2
2
and the updated odds are therefore 1.28 to 1 in favour of the Poisson distribution.
76
If the probabilities of the evidence (E) under H1 and H2 are
n!
pn1 . . . pnk k
n1 ! . . . nk ! 1
n!
P(E | H2 ) =
qn1 1 . . . qnk k
n1 ! . . . nk !
P(E | H1 ) =
then the log-likelihood ratio becomes
k
p1 n1
pk nk
P(E | H1 )
pi
= ln
ln
=
···
ni ln
P(E | H2 )
q1
qk
qi
i=1
77
Under hypothesis H1 we have
p1 = 0.92, p2 = 0.05, p3 = 0.02, p4 = 0.01
and under H2
q1 = 1 − 0.05 − q3 − q4 , q2 = 0.05, q3 , q4 unknown
(where q3 = pB and q4 = PAB ). The likelihood of the evidence E under H2 is
P(E | H2 ) =
n!
(0.95 − q3 − q4 )n1 0.05n2 qn3 3 qn4 4
n1 ! . . . n4 !
c Pearson Education Limited 2011
676
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
where n1 = 912, n2 = 45, n3 = 27, n4 = 16. Thus
ln P(E | H2 ) = n1 ln(0.95 − q3 − q4 ) + n3 ln q3 + n4 ln q4 + constant
∂ ln P(E | H2 )
n1
n3
n3 (0.95 − q3 − q4 ) − n1 q3
=−
+
=
∂q3
0.95 − q3 − q4
q3
(0.95 − q3 − q4 )q3
= 0 if (n1 + n3 )q3 + n3 q4 = 0.95n3
∂ ln P(E | H2 )
n1
n4
n4 (0.95 − q3 − q4 ) − n1 q4
=−
+
=
∂q4
0.95 − q3 − q4
q4
(0.95 − q3 − q4 )q4
= 0 if n4 q3 + (n1 + n4 )q4 = 0.95n4
From the simultaneous equations
939q3 + 27q4 = 25.65
16q3 + 928q4 = 15.2
we find q3 = 0.0269, q4 = 0.0159 and therefore q1 = 0.9072.
It follows that (using the result of the previous exercise)
pi
P(E | H1 )
=
ni ln
ln
P(E | H2 )
qi
i=1
4
0.92
0.05
0.02
0.01
+ 45 ln
+ 27 ln
+ 16 ln
0.9072
0.05
0.0269
0.0159
= −2.645
= 912 ln
If initial odds are P(H1 )/P(H2 ) = 5 then updated odds are
P(H1 | E)
= e−2.645 × 5 = 0.355
P(H2 | E)
that is, 2.8 to 1 in favour of H2 .
78
Under hypothesis H1 we have separate estimates as follows:
24
= 4.0
6
36
λB = mean number of defects for B =
= 7.2
5
λA = mean number of defects for A =
Under hypothesis H2 we have a single estimate as follows:
λ = overall mean number of defects =
60
= 5.455
11
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
677
The evidence E can be expressed as follows:
E = {A : 2 with 3 defects, 2 with 4 defects, 2 with 5 defects;
B : 1 with 5 defects, 1 with 6 defects, 2 with 8 defects, 1 with 9 defects}
The log-likelihood ratio is
ln
λ3 e−λA
λ4 e−λA
λ5 e−λA
P(E | H1 )
= 2 ln A3 −λ + 2 ln A4 −λ + 2 ln A5 −λ
P(E | H2 )
λ e
λ e
λ e
+ ln
λ5B e−λB
λ6B e−λB
λ8B e−λB
λ9B e−λB
+
ln
+
2
ln
+
ln
λ5 e−λ
λ6 e−λ
λ8 e−λ
λ9 e−λ
λB
λA
+ (5 + 6 + 16 + 9)ln
λ
λ
+ (2 + 2 + 2)(λ − λA ) + (1 + 1 + 2 + 1)(λ − λB )
= (6 + 8 + 10)ln
= 2.551
and the updated odds (with no initial preference) are
P(H1 | E)
= e2.551
P(H2 | E)
or 12.8 to 1 in favour of H1 .
Review Exercises 11.12
1
For the standard corks, sample proportion oxidized is p̂1 =
for the plastic bungs, sample proportion oxidized is p̂2 =
proportion oxidized is p̂ =
9
96
= 0.0938. Test statistic
z=
p̂1 − p̂2
1
1
p̂(1 − p̂)( 60
) + ( 36
)
= 0.271
The hypothesis is accepted.
2
The model is
d = d0 e−λt
or equivalently
ln d = ln d0 − λt
c Pearson Education Limited 2011
3
36
6
60
= 0.1, whereas
= 0.0833. Overall
678
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
which is of the form
Y = a + bX
From the data,
X̄ = 5.746, SX = 3.036, Ȳ = −0.2811, SY = 0.721, XY = −3.775
so
b̂ =
XY − X̄Ȳ
= −0.234
S2X
â = Ȳ − b̂X̄ = 1.065
From these we infer
λ̂ = −b̂ = 0.234
d̂0 = eâ = 2.90
Also the error variance is (using unrounded results)
S2E = S2Y − b̂2 S2X = 0.01418
and the 95% confidence interval for λ is
λ̂ ± t.025,8
3
SE
√ = (0.202, 0.266)
SX 8
If position P and load X are related by
P = a + bX
then by linear regression on the data we find
â = 6.129
b̂ = 0.0624
But extension Y and load X are related by
b̂ =
L
Y
=
X
ÊA
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
679
where L = 101.4 and A = 1.62 × 10−5 . The estimate of Young’s modulus is
therefore
Ê =
L
= 1.003 × 108
b̂A
Also from the linear regression the error standard deviation is SE = 0.00624, so
the 95% confidence interval for b is
b̂ ± t.025,6
SE
√ = (0.0597, 0.0651)
SX 6
We infer the 95% confidence interval for E = L/bA as
(96.1 × 106 , 104.9 × 106 )
4 From the data, the mean time between arrivals is estimated as 9.422 hours,
and this is 1/λ for the exponential distribution. If we form a histogram of the
data, the expected probability of a class (a, b) under the exponential distribution
with parameter λ is
P(a < X < b) = FX (b) − FX (a) = 1 − e−λb − (1 − e−λa ) = e−λa − e−λb
Using class intervals of five hours we obtain the table as follows:
Class (k)
0-5
5-10
10-15
15-20
20-25
25-30
> 30
Observations (fk )
48
22
13
12
3
3
4
Expected (ek )
43.24
25.43
14.96
8.80
5.18
3.04
4.35
Probability
0.4118
0.2422
0.1425
0.0838
0.0493
0.0290
0.0414
The value of χ2 = 3.35 is less than χ2.05,5 = 11.07 (seven classes with one parameter
estimated) so the fit to the exponential distribution is good.
5
The maximum value is Xmax = 72 and the total is
y=
Xmax
= 0.0728
Σ i Xi
c Pearson Education Limited 2011
i
Xi = 989.3, so
680
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
and [1/y] = 13. Hence
P(Y ≤ 0.0728) =
13
(−1)
k=0
k
105
(1 − 0.0728k)104
k
= 0.9599
The probability of there occurring such a large value is therefore around 4% , so
the value 72 can be regarded as an outlier at the 5% significance level.
With the outlier included we have X̄ = 9.422, SX = 10.77 so the 95% confidence
interval for mean inter-arrival time is
SX
= (7.36, 11.48)
X̄ ± 1.96 √
105
With the outlier excluded we have X̄ = 8.820, SX = 8.90 so the confidence interval
is (7.11, 10.53).
6
The contingency table (with expected values in brackets and adjusted residuals
underneath) is as follows:
Grade
Very satisfied
Fairly satisfied
Neutral
Fairly dissatisfied
Very dissatisfied
Total
French
16 (15)
0.5
63 (50)
2.8
40 (42)
−0.5
10 (18)
−2.5
3 (7)
−1.8
132
German
6 (7)
−0.5
13 (24)
−3.2
27 (20)
1.9
13 (9)
1.6
5 (3)
64
Spanish
22 (22)
−0.1
76 (77)
−0.2
60 (64)
−1.0
32 (28)
1.2
12 (10)
0.8
202
Total
44
152
127
55
20
398
Chi-square = 20.0, d.f. = (5 − 1)(3 − 1) = 8, and so compare with χ20.025,8 = 17.54:
significant at 2.5% level. The French course scores highest, followed by Spanish
and then German.
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
7
681
Let D denote ‘has disease’ and O denote ‘operation performed’, then
1
2
1
P(survive | D ∩ Ō) =
20
4
P(survive | D̄ ∩ O) =
5
P(survive | D ∩ O) =
and we can assume that P(survive | D̄ ∩ Ō) = 1. If the operation is performed,
then using the hint we have
P(survive | O) = P(survive | D ∩ O)P(D) + P(survive | D̄ ∩ O)P(D̄)
4
4
3
1
= p + (1 − p) = − p
2
5
5 10
(where p = P(D) ). If the operation is not performed,
P(survive | Ō) = P(survive | D ∩ Ō)P(D) + P(survive | D̄ ∩ Ō)P(D̄)
19
1
p + (1 − p) = 1 − p
=
20
20
These probabilities are equal when
3
19
4
− p=1− p
5 10
20
from which p =
4
13
. The surgeon will operate if the assessment of P(D) exceeds
this value.
8
With 200 machines each becoming misaligned every 200 hours on average, the
rate at which machines become misaligned is λ0 =
200
200
= one per hour on average.
The total cost per hour for each option is the sum of three components: the fixed
cost per hour, the cost of correcting the output and the cost of lost production.
For option A, the fixed cost is £1 per hour per machine, hence £200 per hour.
The average run length ARL0 for a misaligned machine is 20 hours, and this
amount of output must be corrected, so the cost per hour of correcting the output
is λ0 × ARL0 × 10 = £200. Lost production occurs while a machine is in the queue
and being serviced, and this occurs whether the machine is actually misaligned or
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
682
not (false alarm). Actual misalignments occur at the rate λ0 and are detected by
the control chart. False alarms occur at the rate
λ1 =
200
= 0.2 per hour
ARL1
where ARL1 = 1000 is the mean time between false alarms for a well-adjusted
machine. These two kinds of action are independent, so the total rate of actions is
λ = λ0 + λ1 = 1.2. Also, the service rate μ = 2 per hour so ρ = λ/μ = 0.6. With
σs =
1
4
, the mean number of machines out of production is
NS = ρ +
(λσs )2 + ρ2
= 1.163
2(1 − ρ)
and the cost per hour of lost production is 200NS = £ 232.5. The total cost for
option A is therefore £ 200 + £ 200 + £ 232.5 = £ 632.5 per hour.
For option B, the fixed cost is £ 1.50 per hour per machine, hence £ 300 per hour.
With ARL0 = 4, the cost per hour of correcting the output is λ0 × ARL0 × 10 = £
40. With ARL1 = 750, false alarms occur at the rate λ1 = 200/750 = 0.267 per
hour, so machines are taken out of production at the total rate λ = λ0 +λ1 = 1.267,
hence ρ = λ/μ = 0.633. The mean number of machines out of production is
therefore NS = 1.317 at a cost per hour 200NS = £ 263.4. The total cost for
option B is therefore £ 300 + £ 40 + £ 263.4 = £ 603.4 per hour. This is less
than for option A.
9
For the source, P(in = 0) = α and P(in = 1) = 1 − α. For the channel,
P(out = 0 | in = 1) = P(out = 1 | in = 0) = p.
9(a)
P(out = 0) = P(out = 0 | in = 0)P(in = 0) + P(out = 0 | in = 1)P(in = 1)
= (1 − p)α + p(1 − α) = p̄α + pᾱ
c Pearson Education Limited 2011
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
683
(where p̄ = 1 − p, ᾱ = 1 − α). Hence
p̄α
P(out = 0 | in = 0)P(in = 0)
=
p̄α + pᾱ
P(out = 0)
pᾱ
P(in = 1 | out = 0) =
p̄α + pᾱ
pα
P(in = 0 | out = 1) =
pα + p̄ᾱ
p̄ᾱ
P(in = 1 | out = 1) =
pα + p̄ᾱ
P(in = 0 | out = 0) =
9(b)
P( in = 0 | out = 0) > P( in = 1 | out = 0) if
p̄α > pᾱ
from which
p̄α > p(1 − α)
hence
(p̄ + p)α = α > p
Similarly, P( in = 1 | out = 1) > P( in = 0 | out = 1) if
p̄ᾱ > pα
from which
p̄(1 − α) > pα
hence
(p + p̄)α = α < p̄
The source symbol is assumed to be the same as the received symbol if p < α < p̄.
10
For the binary symmetric channel, X = {0, 1} with P(X = 0) = α, and
Y = {0, 1} with P(Y = 0) = p̄α + pᾱ, P(Y = 1) = pα + p̄ᾱ(p̄ = 1 − p, ᾱ = 1 − α,
using the results of the previous exercise). Also
P(X = 0 ∩ Y = 0) = P(Y = 0 | X = 0)P(X = 0) = p̄α
P(X = 0 ∩ Y = 1) = P(Y = 1 | X = 0)P(X = 0) = pα
c Pearson Education Limited 2011
684
Glyn James, Advanced Modern Engineering Mathematics, 4th Edition
P(X = 1 ∩ Y = 0) = P(Y = 0 | X = 1)P(X = 1) = pᾱ
P(X = 1 ∩ Y = 1) = P(Y = 1 | X = 1)P(X = 1) = p̄ᾱ
The mutual information between X and Y is as follows:
I(X; Y) =
1
1
P(x, y) log2
x=0 y=0
P(x, y)
P(X = x)P(Y = y)
p̄α
pα
+ pα log2
α(p̄α + pᾱ)
α(pα + p̄ᾱ)
p̄ᾱ
pᾱ
+ p̄ᾱ log2
+ pᾱ log2
ᾱ(p̄α + pᾱ)
ᾱ(pα + p̄ᾱ)
= p̄α log2
= p̄(α + ᾱ) log2 p̄ + p(α + ᾱ) log2 p
− (p̄α + pᾱ) log(p̄α + pᾱ) − (pα + p̄ᾱ) log(pα + p̄ᾱ)
= H(p) − H(p̄α + pᾱ)
where H(t) = t log2 t + (1 − t) log2 (1 − t) is called the ‘entropy function’. In
particular, when α =
1
2
we have
p̄α + pᾱ =
and H( 21 ) =
1
2
log2
1
2
+
1
2
log2
1
2
1
1
(p̄ + p) =
2
2
= −1
so that
I(X; Y) = 1 + H(p) = 1 + p log2 p + (1 − p) log2 (1 − p)
When p = 12 , I(X; Y) = 1 +
1
2
log2
1
2
+
1
2
log2
1
2
= 0.
When p → 0, p log2 p → 0 and p̄ log2 p̄ → 0 so that
I(X; Y) → 1
and similarly when p → 1. Full information is transmitted through the channel
when either every bit is correct (p = 0) or every bit is inverted (p = 1) . No
information is transmitted when the bits are uniformly randomized (p = 12 ) .
c Pearson Education Limited 2011
Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.7 Linearized : Yes Author : dynstab2/ThePirateBay Create Date : 2011:08:26 12:38:18+02:00 Keywords : - Modify Date : 2011:09:19 06:21:14-04:00 Has XFA : No XMP Toolkit : Adobe XMP Core 5.2-c001 63.139439, 2010/09/27-13:37:26 Metadata Date : 2011:09:19 06:21:14-04:00 Creator Tool : - Format : application/pdf Title : - Creator : dynstab2/ThePirateBay Description : - Subject : - Document ID : uuid:806f6716-8e1d-4388-91ec-a3027283f743 Instance ID : uuid:2185403b-0f1e-427a-9c39-6987b86285d4 Producer : - Page Layout : SinglePage Page Mode : UseOutlines Page Count : 688EXIF Metadata provided by EXIF.tools