Elementary methods, diﬀerentiation, complex numbers

PART I: Elementary methods, differentiation, complex numbers
Chapter 1: Standard functions and techniques, 1

Chapter 2: Differentiation, 15
Chapter 3: Further techniques for differentiation, 23
Chapter 4: Applications of differentiation, 32
Chapter 5: Taylor series an approximations, 46
Chapter 6: Complex numbers, 58
Chapter 1: Standard functions and techniques
1.1. (a) y = x4 for −1.5 ≤ x ≤ 1:

y
5
4
1
x
-1.5 -1 -0.5 0.5 1
Figure 1: Problem 1.1a
(b) y = x(1 − x) for −1 ≤ x ≤ 2:

y
x
-1 -0.5 0.5 1 1.5 2
-0.5
-1
-1.5
-2
Figure 2: Problem 1.1b
(c) y = 1 + x + x2 for |x − 1| ≤ 2:
y
12
10
8
6
4
2
x
-1 1 2 3
Figure 3: Problem 1.1c
(d) y = |x − 1| for −3 ≤ x ≤ 3:
y
4
x
-3 -2 -1 1 2 3
Figure 4: Problem 1.1d
1
y
10
x
-3 -2 -1 1 2 3
Figure 5: Problem 1.1e

y
1
0.8
0.6
0.4
0.2
x
-2 -1 1 2
Figure 6: Problem 1.1f
(e) y = |x| + |x − 3| + |x + 2| for −3 ≤ x ≤ 4:

(f) y = ||x|
√ − 1| for −2 ≤ x ≤ 2:
(g) y = (x2 + 1) for |x| ≤ 2:
y
2
1.5
1
0.5
x
-2 -1 1 2
Figure 7: Problem 1.1g
1.2. (a) y = −2x + 3; (b) y = 1; (c) y =√32 x − 13 . The√intersections occur at A : (2, 1), B : ( 54 , 12 ),
C : (1, 1). The side lengths are: AB = 41 13, BC = 41 5, CA = 1.
y
H2,1L
1 H1,1L
0.8
0.6
0.4 H54,12L
0.2
x
0.5 1 1.5 2
Figure 8: Problem 1.2
1.3. (a) Slope is 1, and the line cuts the axes at (0, −1) and (1, 0).
(b) Slope is 13 , and the line cuts the axes at (0, − 23 ) and (2, 0).
(c) Slope is − 25 , and the line cuts the axes at (0, − 45 ) and (2, 0).
1.4. (a) y = x + 1; (b) y = −2x − 4; (c) y = 0.5x − 0.5; (d) y = 3x − 1; (e) the slope of the line
must be − 14 : y = − 14 x + 11
4 .
1.5. The products of the slopes in each case must be −1. The slopes are: (a) − 32 and 23 ; (b) 2 and
− 21 ; (c) 2 and − 12 ; (d) 1 and −1.
1.6. At the point of intersection, x + y + 1 = 0 and 2x − 3y − 2 = 0, so the line
(x + y + 1) + α(2x − 3y − 2) = 0
2
must pass through this point, which has coordinates (− 51 , − 45 ). The straight line joining this point
to (1, 1) is
2y = 3x − 1,
with α = 12 .
√
1.7. (a) Centre at (0, 0), radius √ 3; (b) centre at (1, 0), radius 2; (c) centre at (1, 1), radius 23;
(d) centre at ( 21 , − 12 ), radius 12 11.
1.8. (x − 1)2 + (y + 2)2 = 9.
1.9. Eliminate one of the variables in each case and solve the resulting quadratic equation.
(a) (2, 2) and (2, −2);
(b) Eliminate y, so that x satisfies the equation
x2 + (2x + 1)2 − 2x + 2(2x + 1) − 4 = 0, or 5x2 + 6x − 1 = 0.
The points of intersection are

√ √ √ √
( 15 (−3 − 14, −1 − 2 14) and ( 51 (−3 + 14, −1 + 2 14)
√ √
(c) ( 12 2, 12 2), one point only since the line is tangential to the circle.
1.10. To three decimal places, the distances of the points from the origin are
1.060, 0.993, 1.011, 0.896, 1.124.
The average value of these distances is r = 1.017. The equation of the circle is
x2 + y 2 = r2 = 1.034.
1.11. (a) x = H(t + 1) − H(t − 1).
x
1
0.8
0.6
0.4
0.2
t
-2 -1 1 2
(b) x = sgn(1 + t) + sgn(1 − t).
x
2
1.5
1
0.5
t
-2 -1 1 2
(c) x = tH(t − 1).

(d) (t2 − 1)[sgn(t + 1) + sgn(1 − t)].
1.12. (a) f (t) = H(2 − t) + H(t + 1) − 1; (b) f (t) = 2tH(t);
(d) f (t) = (3 − t)H(3 − t) + (t − 2)H(2 − t) + (t − 1)H[1 − t) − tH(−t).
3
x
2
1.5
0.5
t
0.5 1 1.5 2

x
t
-2 -1 1 2
-0.5
-1
-1.5
-2
1.13. (a) 16 π radians; (b) 23 π radians.

√ √ √ √ √
1.14. √(a) 1/ 2; (b) 1; (c) 0; (d) −1/ 2; (e) 3/2; (f) − 3/2; (g) − 3/2;
(h) − 3/2.
1.15. (a) Using the identity cos2 B = 12 (1 + cos 2B)
1 1
cos4 A = (1 + cos 2A)2 = (1 + 2 cos 2A + cos2 2A)
4 4
1 1 1
= (1 + 2 cos 2A + (1 + cos 4A)) = (3 + 4 cos 2A + cos 4A).
4 2 8
(b) Use the identities sin2 B = 12 (1 − cos 2B) and cos2 2B = 12 (1 + cos 4B):
1 3 1 1
sin4 A = (1 − 2 cos 2A + cos2 2A) = − cos 2A + cos 4A.
4 8 2 8
1.16. (a) cos(x + 12 π) = cos x cos 12 π − sin x sin 21 π = sin x; (b) cos x; (c) − cos x; (d) − cos x (for
both); (e) sin x (for both).
1.17. (a)
1 1 1 1
cos x + cos y = cos[ (x + y) + (x − y)] + cos[ (x + y) − (x − y)]
2 2 2 2
1 1 1 1
= cos[ (x + y)] cos[ (x − y)] − sin[ (x + y)] sin[ (x − y)] +
2 2 2 2
1 1 1 1
cos[ (x + y)] cos[ (x − y)] + sin[ (x + y)] sin[ (x − y)]
2 2 2 2
1 1
= 2 cos[ (x + y)] cos[ (x − y)]
2 2
(b) sin x − sin y = 2 sin[ 21 (x − y)] cos[ 12 (x + y)].
(c) cos x − cos y = −2 sin[ 21 (x + y)] sin[ 12 (x − y)].
1.18. (a) x = nπ, (n = 0, ±1, ±, 2, . . .);
(b) x = 12 (2n + 1)π, (n = 0, ±1, ±, 2, . . .);
(c) x = 2nπ, (n = 0, ±1, ±, 2, . . .);
(d) x = 16 (2n + 1), (n = 0, ±1, ±, 2, . . .);
(e) x = 12 nπ, (n = 0, ±1, ±, 2, . . .);
(f) x = 2n, (n = 0, ±1, ±, 2, . . .).
4
1.19.
amplitude angular frequency period phase
(a) 2 0.2 10π 3.2
(b) 1.5 0.2 10π −0.48
(c) 3.87 0.2 10π −0.135
(d) 1 1 2π π
√
1.20. (a) F (x) = 2 −x; (b) F (x) = 2 (x − 3); (c) F (x) = 21 arcsin x;
1 1
1
(d) F (x) = arcsin(£12 x); (e) F¤(x) = [arccos x] 2 ;
(f) F (x) = arccos π2 arcsin x ; √
(g) F (x) = x−4 ; (h) F (x) = − 12 + (x + 14 ).
1.21. The graph shows y = x3 − x + 1 (the dashed curve) and its inverse.
y
2
1.5
0.5
x
-2 -1.5 -1 -0.5 0.5 1 1.5 2
-0.5
-1
-1.5
-2
1.22. (a) x = 12 ln 3; (b) x = 13 e2 ; (c) x = e−3 ; (d) x = − 31 ln 3;

(e) the equation is the same as (ex√− 1)2 = 0: hence √ x = 0;
(f) x = 2; (g) x = 2/17; (h) x = ± 2; (i) x = ± (1 + ee ); √
(j) x = (ln 3)/(ln 2); (k) x = −(ln 2)/(2 ln 3); (l) x = 12 ln[4 + 17];
(m) no solutions.
1.23. 2x = ex ln 2 .
1.24. Consider two values of x, say x1 and x2 , where x1 > x2 . Then if 10x1 = 2 × 10x2 , it follows
that
ln 2
10x1 −x2 = 2, or x1 − x2 = ,
ln 10
an interval which is independent of x1 and x2 .
1.25. (a) (x − 1)2 + y 2 ≤ 9.
y
3
x
-2 -1 1 2 3 4
-1
-2
-3
(b) x ≥ 0, y ≥ 0, and x + y ≤ 1.
(c) (x2 /4) + (y 2 /9) ≤ 1.
(d) x2 + y 2 ≤ 1 and x ≥ 0.
(e) |x| + |y| ≤ 1.
1.26. Let
sinh y ey − e−y
x = tanh y = = y ,
cosh y e + e−y
5
y
1
0.8
0.6
0.4
0.2
x
0.2 0.4 0.6 0.8 1

y
3
x
-2 -1 1 2
-1
-2
-3
where −1 < x < 1. Hence

· ¸
2y 1 1+x
(1 − x)e = 1 + x so that y = ln
2 1−x
as required.
1.27. From triangle ABC

√
AC = AB sin θ + (BC 2 − AB 2 cos2 θ)
√
= 2.5[sin ωt + (4 − cos2 ωt)] cm,
where θ = ωt. The angular frequency ω = 400π/3.
1.28. x = 5 cos(ωt − 0.927). The amplitude c = 5 and the phase angle is φ = −0.927.
1.29. f (0) = 2 implies C = 2 and f (1) = 0.5 implies 0.5 = Ce−α = 2e−α . Hence α = ln 4. Also
f (2) = 18 .
1.30. The tidal period is 2π/0.5 = 12.57 hours. We require the times when the depth is 2m in one
period, which are given by solutions of
2 = 5 + 4.5 sin 0.5t so that sin 0.5t = − 23 .
y
1
0.5
0.20.40.60.8 1 x
-0.5
-1
6
y
1
0.5
x
-1 -0.5 0.5 1
-0.5
-1
Two consecutive times are 11.11 hours and 7.74 hours. Hence the yacht can float free for 9.20
hours in each tidal period. The yacht floats when sin 0.5t > − 23 . It is helpful to sketch y = sin 0.5t
and y = − 32 and plot their intersections.
1.31. (a) The cardioid r = 0.5(1 + cos θ):
0.5
0.5 1 1.5 2
-0.5
-1
(b) The folium r = (4 sin2 θ − 1) cos θ
0.75
0.5
0.25
-1 -0.75 -0.5 -0.25 0.25 0.5
-0.25
-0.5
-0.75
(c) r = sin 2θ:

0.75
0.5
0.25
-0.75 -0.5 -0.25 0.25 0.5 0.75
-0.25
-0.5
-0.75
(d) The Archimedean spiral r = 0.04θ:

]
(e) The equiangular spiral r = 0.1e0.1θ :
1.32. (a) sgn (sin x):
(b) sgn cos 2x:
7
0.4
0.2
-0.6 -0.4 -0.2 0.2 0.4 0.6
-0.2
-0.4
-0.6

0.4
0.2
-0.4 -0.2 0.2 0.4 0.6
-0.2
-0.4
(c) H(x) sin x:

(d) sin2 x:
(e) | sin x|:
(f) sin |x|:
(g) H(x − π) sin x:
1.33. Let the points be A : (−7, 3), B : (1, −3) and C : (4, 1). The slope of AB is − 34 and the
d is a right angle. Let D
slope of BC is 43 ; the product of the slopes is −1 which means that ABC
be the fourth vertex. Then the equations of the lines AD and DC are
y − 3 = 34 (x + 7) and y − 1 = − 34 (x − 4),
or
3y − 4x − 37 = 0 and 4y + 3x − 16 = 0.
These lines intersect at the point D : (−4, 7)
There is a general formula buried here, if you notice that the coordinates of D are (−7 + 4 −
1, 3 + 1 − (−3)).
1.34. (a) periodic, period 12 π; (b) periodic, period 2π; (c) periodic, period 2π; (d) not periodic;
(e) periodic, period 2π; (f) periodic, period π; (g) not periodic; (h) periodic, period π; (i) periodic,
period π; (j) periodic, period 8π; (k) periodic, period 23 , since sin 3t has period 32 π and cos 9t has
period 29 π but has the period of sin 3t; (l) not periodic.
1.35. (a) neither odd nor even; (b) even; (c) odd since sin x is odd; (d) odd since product of odd
and even functions; (e) even; (f) even; (g) neither odd nor even.
1 1 1 2 1 1
1.36. (a) 5(x−2) − 5(x+3) ; (b) − x+1 + x+2 ; (c) x + x−1 ;
y
1
0.5
x
-6 -4 -2 2 4 6
-0.5
-1
8
y
1
0.5
x
-2 -1 1 2
-0.5
-1

y
1
0.5
x
-2 2 4 6
-0.5
-1
1 1 1 1
(d) 2x − x+1 + 2(x+2) ; (e) 2(x−1) − x1 + 2(x+1)
1
;
1 1 1 1 4
(f) 4x − 2(x+2)2 − 4(x+2) ; (g) x+1 = (x+2)2 ;
1 1
(h) 2(x−3) + 2(x+1) ; (i) x1 − x−1
1 1
+ (x−1)2;
1 1 1
(j) x2 − x + x+1 .
1 x+1 1 1−x 1 x+6
1.37. (a) x − x2 +x+1 ; (b) 2(x−1) + 2(x2 +1) ; (c) − 5(x+1) + 5(x2 +2x+6) .
1.38. (a) x12 − x21+1 ; (b) x − 3 − 1

x+1 + 8
x+2 ;
9 1 9
(c) 1 + 8(x−3) + 8(x+1) − 4(x+3) .
1 1 1 1
1.39. (a) 4 + 8 + 16 + 32; (b) 1 + 2 + 5 + 10 + 17 ;
(c) x + 2x2 + 3x3 + 4x4 .
1.40. For (a), (b), (c), (e) and (f) proceed as in Example 1.17.
(a) 2[1 − ( 12 )8 ] = 255 128 .
(b) 12 · 13 [1 − ( 13 )5 ] = 121
729 .
(c) (1 − e−12 )/(1 − e2 ).
(d) The sum is 642. More generally, let
x + 2x2 + · · · + nxn = T.
Then
T − xT = (1 − x)T = xS − nxn+1 ,
where S is the sum of the geometric series
1 + x + · · · + xn−1 .
y
1
0.8
0.6
0.4
0.2
x
-6 -4 -2 2 4 6
9
y
1
0.8
0.6
0.4
0.2
x
-6 -4 -2 2 4 6

y
1
0.5
x
-6 -4 -2 2 4 6
-0.5
-1
For the given problem x = 2.

(e) − 12 32 [1 − (− 12 )10 = − 1024
341
.
7 7
(f) 2[ 1−(0.5)
0.5 ] + 3[( 1−(0.6)
0.4 ] = 11.258 . . ..
1.41. The series can be expressed as
10
X
x + x5 + x9 + · · · + x41 = x (x4 )n .
n=0
Using (1.33), the sum of the series is

1 − x44
x .
1 − x4
1.42. Let D be the foot of the perpendicular on to the side AC. Then
c2 = DB 2 + DA2 = DB 2 + (AC − DC)2 .
But DB = a sin C and DC = a cos C. Therefore
c2 = a2 sin2 C + (b − a cos C)2

= a2 sin2 C + a2 cos2 C + b2 − 2ab cos C
= a2 + b2 − 2ab cos C
1.43. The ratio of any pair of successive terms is
f (t0 + (n + 1)T ) Aec(t0 +(n+1)T )

= = ecT ,
f (t0 + nT ) Aec(t0 +nT )
y
1
0.5
x
2 4 6 8 10 12
-0.5
-1
10
which is independent of n. The common ratio is ecT .
1 1 1
1.44. (a) 1.111 . . . = 1 + 10 + 100 + · · · is an infinite geometric series with common ratio 10 and
1
sum 1/(1 − 10 ) = 11/9. (b) The common ratio is 1/10, and as a fraction the sum is 1; (c) the
common ratio is 1/100, and as a fraction the sum is 1/99; (d) the common ration is 1/100, and
as a fraction the sum is 1/11; (e) the common ration is 1/10 and as a fraction the sum is 2/3; (f)
the notation means 2.7̇2̇ = 2.727272 . . .: the common ratio is 1/100 and the sum represents the
fraction 30/11.
1.45. The sum of the infinite geometric series is
∞
X xm
xn = , |x| < 1.
n=m
1−x
(a) 2; (b) 10/9; (c) e/(e − 1); (d) 32 ; (e) 3/5.

1.46. (a) 24, 720, 5040; (b) 12; (c) 720; (d) 220; (e) 120; (f) 1, 3, 3, 1.
1.47. (a) (i) n(n − 1); (ii) (n + 1)n; (b) (i) 2m m!; (ii) (2m + 1)!/(2m m!).
¡ ¢
1.48. (a) (i) 120; (ii) 504; (iii) 120; (iv) 35; (v) 35; (vi) 252; (vii) 4950; (viii) 10
7 =10 C7 = 120.
n!
(b) n Pn = (n−n)! = n!0! = n! Also P
n n−1 = n!
(n−n+1)! = n!
1! = n!. Consider a collection of n
different letters. The number of different words of length n which can be made without repetition
is n Pn = n!. Suppose that the letters are A, B, C, . . . . Suppose that the first letter of an n
character word is A. Then the remaining n − 1 letters can be chosen in (n − 1)! ways. Repeat
the procedure for words with first letters B, C and so on. We obtain all words with n characters
again, and there are n(n − 1)! = n! of them.
1.49. (a) 4 P1 = 4!; (b) 4 C3 = 4; (c) 44 = 256; (d) 20; (e) 4 P4 +4 P3 +4 P2 +4 P1 = 4+12+24+24 = 64.
(f) Without repetitions the number of combinations is
4 C1 +4 C2 +4 C3 +4 C4 = 4 + 6 + 4 + 1 = 15.
With 2 letters the same there are 4 + 12 + 12 = 28 possibilities, and with 3 letters the same there
are 4 + 12 = 16. Hence the total number of combinations is 15 + 28 + 16 = 59
1.50. (a) With no E’s there are 4 P3 = 24 words, with 1 E there are 3 ×4 P2 = 36, and with 2 E’s
there are 3 ×4 P1 = 12. Hence there are 24 + 36 + 12 = 72 words.
(b) Label six letters A, B, C, D, E1 , E2 . Then the number of words treating E1 and E2 as distinct
is 6! = 720. The letters E1 and E2 can be interchanged in 2! = 2 ways. Hence the number of
six-letter words is 720/2 = 360.
1.51. (a) There are 5 P4 = 5!/1! = 120 distinct four-digit numbers.
(b) To be divisible by 5, the last digit must be 5. The preceding 3 digits can be chosen in 4 P3 = 24
ways. Hence there are 24 numbers divisible by 5.
(c) To be divisible by 2 the final digit must be 2 or 4. As in (b) the number of numbers is 24 P3 = 48.
(d) The numbers contain either 1, 2, 3 or 4 digits. There are 4 one-digit numbers (excluding
zero). For two-digit numbers we must exclude those starting with zero since they are the same
as the one-digit numbers. Hence there are 16 distinct two-digit numbers. Similarly there are
44 P2 = 48 three-digit numbers and 44 P3 = 96 four-digit numbers. Hence the total number is
4 + 16 + 48 + 96 = 164 words.
1.52. (a) Without restriction, the number of distinct combinations of personnel (no distinction
being made as to which particular post is assigned to each person) is 7 C4 = 7!/(4!3!) = 35.
(b) There is one selection with 4 females, 12 with 3 females and one male, 18 with 2 females and
2 males and 4 with one female and three males. (b) The posts can be filled in the following ways:
4 C4 = 1 with 4 females; 4 C3 3 C1 = 12 with 3 females and one male; 4 C2 3 C2 = 18 with 2 females
and 2 males; 4 C13 C3 = 4 with one female and 3 males. This confirms the 35 combinations of
personnel.
11
1.53. (a) We may model the problem by thinking of an ordered line of N pool balls, of various
colours (types) denoted by A, B, . . ., the number of each colour being NA , NB , . . . . The number
of possible orders (permutations) for the individual balls is N !, but we cannot distinguish visually
between balls having the same colour, so many of the N ! orders will look identical.
Suppose that the number of distinguishable arrangements is M . Each one of of these corresponds
to a possible NA !NB ! . . . permutations within the separate colours, so that
N!
N ! = M [NA !NB ! . . .], or M = .
NA !NB ! . . .
(b) We require the total number of different combinations, involving every number 1, 2, . . . , N of
balls. Consider any one of these: it contains 0 or 1 or 2. . . or NA (that is, (1 + NA ) possibilities)
of type A; 0 or 1 or 2. . . or NB of type B; and so on. The number of possible combinations is
therefore
(1 + NA )(1 + NB ) . . . − 1,
in which the term −1 is introduced to exclude the case of an all-zero ‘combination’.
1.54. (a) The national groups may be ordered (permuted) in 4! ways. By allowing for 5! permu-
tations possible within each group we obtain
5!5!5!5!4! = 5!4 4! = 2880
distinct line-ups.
(b) The number of distinct orderings of the 4 groups around a circular table is (4 − 1)! = 3! (see
Example 1.23). All possible permutations within the groups are then to be allowed for, so the total
number of arrangements is 5!3! = 720.
1.55. (a) (Prizes identical) The number of combinations of 3 distinct prizewinners out of 10
eligibles is 10 C3 = 120.
(b) (Prizes different) Call the Prizes P1 , P2 , P3 . P1 may go to any of 10 people; with each allocation
P2 may go to any of the remaining 9; then P3 to any of the remaining 8; all of these distributions
being distinct. The total number of possibilities is 10 × 9 × 8 = 720.
(c) (Prizes equal, distribution arbitrary) There are 3 types of distribution which can occur:
(i) One person gets all the prizes: there are 10 possibilities.
(ii) There are 10 persons who might get 2 prizes. With each of these there are 9 persons eligible
for the other prize. There are therefore 9 × 10 = 90 possibilities.
(iii) Three different people get prizes. Part (a) gives the number: there are 120 possibilities.
Therefore the total number of possibilities is
10 + 90 + 120 = 220
(d) P1 may go to any of the 10; similarly with P2 and P3 . Therefore the total number is 10×10×10 =
1000.
1.56. (a) The table shows the permissible numbers in the 3 categories. The number of combinations
possible within each category are given in brackets.
Accountants Lawyers Doctors Committees

- 1(2) 3(1) 2
- 2(1) 2(3 C2 ) 3
1(2) - 3(1) 2
1(2) 1(2) 2(3 C2 ) 12
1(2) 2(1) 1(3) 6
2(1) - 2(3 C2 ) 3
2(1) 1(2) 1(3) 6
2(1) 2(1) - 1
Check: 7 C4 = 35
12
Committees with exactly 1 accountant: 2 + 12 + 6 = 20.
Committees with exactly 1 doctor: 6 + 6 = 12.
(b) To locate the fallacy consider combinations of the 7 letters A, B, C, D, E, F, G, and take n = 4
and r = 3. Take, say, the r = 3-fold combination ABC and supplement it by, say, the unused letter
D, to form the combination ABCD. In the fallacious construction this will be counted several
times; for example, the same combination is counted again when it arises from supplementing
BCD by A.
The result is shown to be false by simply substituting the given numbers: only one contradiction
is sufficient to dispose of it.
1.57. (a) Refer back to (1.44).
(b) (1 − x)6 = 1 − 6x + 15x2 − 20x3 + 15x4 − 6x5 + x6 .
(c)
(x + x−1 )5 = x5 + 5x4 x−1 + 10x3 x−2 +

10x2 x−3 + 5xx−4 + x−5
= x5 + 5x3 + 10x + 10x−1 + 5x−3 + x−5
(x − x−1 )5 = x5 + 5x4 (−x)−1 + 10x3 (−x)−2 + 10x2 (−)x−3 +

5x(−x)−4 + (−x)−5
= x5 − 5x3 + 10x − 10x−1 + 5x−3 − x−5
1.58.
(1.01)10 = (1 + 0.01)10
= 1 + 10 × (0.01) + 45 × (0.01)2 + 120 × (0.01)3 + · · ·
= 1 + 0.1 + 0.0045 + 0.00012 + · · · = 1.105
to three decimal places.

Similarly
(0.99)8 = (1 − 0.01)8
= 1 − 8 × (0.01) + 28 × (0.01)2 − 56 × (0.01)3 + · · ·
= 1 − 0.08 + 0.0028 − 0.00056 + · · · = 0.923
to 3 decimal places.
1.59. Use the binomial theorem in the form
(1 + x)n = 1 +n C1 x +n C2 + · · · +n Cn xn .
(a) Put x = 2, so that

3n = 1 + 2n C1 + 22 n C2 + · · · + 2n n Cn xn .
For the second result put x = −1:
0 = 1 −n C1 +n C2 − · · · + (−1)n n Cn xn .
(b) Obtain two series with x = 1 and x = −1. Then add and subtract the series.
1.60. F (n, k) is defined for n = 0, 1, 2, . . ., and k = 0, 1, 2, . . . , n by
F (n, k) = n C0 + n+1 C1 + n+2 C2 + · · · + n+k Ck . (i)
A certain formula, namely

F (n, k) = n+k+1 Ck (ii)
13
is suggested for the sum in (i), and its truth for small values of k can be confirmed by calculation;
for example, from (i)
n!
F (n, 0) = n C0 = = 1 ≡ n+1 C0 , (iii)
0!n!
(n + 1)!
F (n, 1) = n C0 + n+1 C1 = 1 + = n + 2 ≡ n+2 C1 ;
1!n!
and so on.
To prove the truth of (ii) for all values of 0 ≤ k ≤ n, recast (i) into the form
F (n, k + 1) ≡ F (n, k) + n+k+1 Ck+1 , (iv)
a ‘recurrence formula’ enabling us to advance one step at a time in k, starting, for example, with
F (n, 0) and finding F (n, 1), F (n, 2), . . . , successively.
Now suppose we have verified the formula (ii) for any one particular value of k, say for k = K;
that is, we know somehow that
(n + K + 1)!
F (n, K) = n+K+1 CK = (v)
K!(n + 1)!
(for all n). Then from (iv)
F (n, K + 1) = F (n, K) + n+K+1 CK+1

= n+K+1 CK + n+K+1 CK+1 from (v)
(n + K + 1)! (n + K + 1)!
= +
K!(n + 1)! (K + 1)!n!
µ ¶
(n + K + 1)! 1 1
= +
K!n! n+1 K +1
(n + K + 2)!
= = n+K+2 CK+1
(K + 1)!(n + 1)!
We have proved that if (iv) is true for k = K, it is true for k = K + 1, where K may take any
value in 0 ≤ K < n.
But we verified in (iii) that (iv) holds good when k = K = 0. Therefore by (vi) it is true when
k = K + 1 = 1, so by (vi) again it is true when k = K + 2 = 2, and so on. It is therefore true for
all k.
1.61. Using partial fractions
1 1 1 1
= = − .
x2 + 3x + 2 (1 + x)(2 + x) 1+x 2+x
Write as
1 1 1
= (1 + x)−1 − (1 + x)−1 ,
x2 + 3x + 2 2 2
and expand both terms using (1.37) for infinite geometric series. Hence
1 1 x x2 x3
= (1 − x + x2 − x3 + · · ·) − (1 − + 2 − 3 + · · ·)
x2 + 3x + 2 2 2 2 2
1 1 1 1
= − (1 − )x + (1 − )x2 − (1 − )x3 + · · · .
2 2.2 2.22 2.23
1.62. V1 = A(1 + R), V2 = V1 (1 + R) = A(1 + R)2 , etc. In pounds: for 1000 @ 3% p.a.;
V5 = 1000(1 + 0.03)5 = 1159.27, V10 = 1343.92; V15 = 1557.97.
14
(b) Let the period start at an arbitrary time T0 . Then
VT0 +T (1 + R)t0 +T
= = (1 + R)T .
VT 0 (1 + R)T0
(c) Let T2 be the doubling period, so that from (b)
ln 2
(1 + R)T2 = 2 and T2 = .
ln(1 + R)
If R = 3%, T2 = 23.4 yr; if R = 6%, T2 = 11.9 yr; if R = 9%, T2 = 8.0 yr.

For the ten-times period, T10 = ln 10/ ln(1 + R). If R = 6%, then T10 + ln 10/ ln 1.06 = 39.6 yr.
1.63. If the income is withdrawn annually, it has been allowed to accrue to the fund through the
previous year with interest at the going rate R annually, or r monthly, the relation being
A(1 + R) = A(1 + r)12
where A is the value of the fund at the start of that year. By the binomial theorem,
(1 + r)12 = 1 + 12r + · · · · · · > 1 + 12r,
so R > 12r.
1.64. If the interest is payable monthly at the rate of rM per month, the interest on a fixed debt
D over any 12-month period is D(1 + rM )12 . This is equal to D(1 + R) where R is the annual
equivalent rate (AER). Therefore R = (1 + rM )12 − 1. If rM = 1%, R = 1.0112 − 1 = 0.126 (12.6%).
If rM = 3%, R = 0.425 (42.5%).
1.65. (a) After N complete years the initial payment A has drawn interest for N yrs, the second
payment for N − 1 yrs, and so on, and the (N − 1)th payment for 1 yr. The value VN of the fund
is then given by the geometric series
A(1 + R)N + A(1 + R)N −1 + · · · + A(1 + R)

= A(1 + R){1 + (1 + R) + · · · + (1 + R)N −1 }
= A(1 + R){(1 + R)N − 1}/R.
(b) N = 10, R = 5%. We obtain
100(1.05)(1.0510 − 1)
V10 = = £1320.68,
0.05
equivalent to a gain of 32% on the total investment of £1000.
(c) M investments of 2A, at 2-year intervals. Formula (a), with the fund value increasing by a
factor (1 + R)2 in each interval, becomes
(2A)(1 + R)2 ){[(1 + R)2 ]M − 1}

V2M = .
{(1 + R)2 − 1}
Using the data in (b) we obtain
200(1.05)2 (1.0510 − 1)
V10 = = £1352.88.
1.052 − 1
Chapter 2: Differentiation
2.1. Below are some sample values for three values of x on either side of the point where the
tangent is required. (The exact value of the slope is also given here.)
15
(a) y = x3 at (1, 1).
x 0.94 0.96 0.98 1.02 1.04 1.06
chord slope 2.82 2.88 2.94 3.06 3.12 3.18
The slope is 3.
√
(b) y = x at (1, 1).
x 0.85 0.90 0.95 1.025 1.10 1.15
chord slope 0.520 0.513 0.506 0.494 0.488 0.483
The slope is 0.5
√
(c) y = cos x at ( 14 π, 1/ 2).
x − 14 π −0.09 −0.06 −0.03 +0.03 0.06 0.09
chord slope 0.674 0.685 0.696 0.718 0.728 0.738
√
The slope is 1/ 2 = 0.707
(d) y = ex at (0, 1).
x −0.15 −0.10 −0.05 0.05 0.10 0.15
chord slope 0.929 0.952 0.975 1.025 1.052 1.079
The slope is 1.
(e) y = e2x at (0, 1).
x −0.15 −0.10 −0.05 0.05 0.10 0.15
chord slope 1.728 1.813 1.903 2.103 2.214 2.332
The slope is 2.
1
(f) y = x3 + x 2 at (1, 2).
x 0.94 0.96 0.98 1.02 1.04 1.06
chord slope 3.33 3.38 3.44 3.56 3.62 3.68
The slope is 3.5, the sum of the slopes in (a) and (b).
(g) y = ln x at (1, 0).
x 0.94 0.96 0.98 1.02 1.04 1.06
chord slope 1.031 1.0206 1.010 0.990 0.981 0.971
The slope is 1
2.2. (a) For y = 3x at (2, 6),
· ¸
dy δy 3(2 + δx) − 6
= lim = lim = 3.
dx δx→0 δx δx→0 δx
(b) For y = 3 − 2x at (1, 1),

· ¸
dy δy [3 − 2(1 + δx) − (3 − 2)
= lim = lim = −2.
dx δx→0 δx δx→0 δx
(c) For y = 3x2 at (1, 3),

· ¸
dy 3(1 + δx)2 − 3(1)2
= lim
dx δx→0 δx
= lim [6 + 3δx] = 6.
δx→0
(d) For y = x3 at (1, 1),

· ¸
dy (1 + δx)3 − 13
= lim
dx δx→0 δx
· ¸
3δx + 3(δx)2 + (δx)3
= lim
δx→0 δx
= lim [3 + 3δx + (δx)2 ] = 3
δx→0
16
(e) For y = 1/x at (2, 12 ),
· ¸
dy 1 1 1
= lim −
dx δx→0 δx 2 + δx 2
· ¸
−1 1
= lim =− .
δx→0 2(2 + δx) 4
(f) For y = 3x + 2x2 at (1, 5),
· ¸
dy 3(1 + δx) + 2(1 + δx)2 − 3 − 2
= lim δx → 0
dx δx
= lim [3 + 4 + 2δx] = 7.
δx→0
(g) For y = (1 + 2x) = 1 + 4x + 4x2 at (−1, 1),

2
dy 1
= lim [1 + 4(−1 + δx) + 4(−1 + δx)2 − 1 + 4 − 4]
dx δx→0 δx
= lim [4 + 4(−2 + (δx)2 ] = −4.
δx→0
2.3. (a) y = 3x2

dy 1
= lim [3(x + δx)2 − 3x2 ]
dx δx→0 δx
= lim [6x + 3δx]
δx→0
= 6x.
(b) y = x3 .
dy 1
= lim[(x + δx)3 − x3 ]
dx δx→0 δx
= lim [3x2 + 3xδx + (δx)2 ]
δx→0
= 3x2 .
(c) y = 1/x.
· ¸
dy 1 1 1
= lim −
dx δx→0 δx x + δx x
· ¸
−1
= lim
δx→0 x(x + δx)
= −1/x2 .
(d) y = x + 12 .
dy 1 1 1
= lim [(x + δx + ) − (x + )]
dx δx→0 δx 2 2
= 1.
(e) y = x + (1/x).
·µ ¶ µ ¶¸
dy 1 1 1
= lim x + δx + − x+
dx δx→0 δx x + δx x
· ¸
1
= lim 1 − 2
δx→0 x + xδx
1
= 1 − 2.
x
17
(f) y = 2x2 − 3.
dy 1
= lim [(2(x + δx)2 − 3) − (2x2 − 3)]
dx δx→0 δx
= lim [4x + 2δx]
δx→0
= 4x.
2.4. Let x = f (t) be the displacement function in each case. The average velocity over the interval
t to t + δt equals
[f (t + δt) − f (t)]/δt.
(a) x = f (t) = 3t. When t = 1
Interval δt 0.5 0.25 0.1 0.01
f (1 + δt) 4.5 3.75 3.3 3.03
Average velocity 3 3 3 3
(Since f (t) is linear in t the velocity 3 at all t.)
(b) x = f (t) = 5t2 . When t = 3.
Interval δt 0.5 0.25 0.1 0.01
f (3 + δt) 61.25 52.81 48.05 45.35
Average velocity 32.5 31.25 30.5 30.05
The values are approaching the limit 30.
(c) x = f (t) = 2t − 5t2 . When t = 1.
Interval δt 0.5 0.25 0.1 0.01
f (1 + δt) −8.25 3.75 −3.85 −3.08
Average velocity −10.5 −9.25 −8.5 −8.05
The limit is −8.
(d) x = 2t − 5t2 . When t = 0.2.
Interval δt 0.5 0.25 0.1 0.01
f (0.2 + δt) −1.25 −0.3125 −0.05 −0.0005
Average velocity −25 −1.25 −0.5 −0.05
In the limit the velocity is zero.
2.5. (a) dy/dx = 1 for all x; (b) dy/dx = 3x2 so that dy/dx = 27 at x = 3;
(c) dy/dx = 4x3 so that dy/dx = 32 at x = 2 and −32 at x = −2.
2.6. (a) y = x; dy/dx = 1: the graph is a straight line at 45◦ to the x axis.
y
2
x
-2 -1 1 2
-1
-2
(b) y = x2 ; dy/dx = 2x: the slope is negative for x < 0 and positive for x > 0, and increases from
−∞ to +∞: the curve is a parabola.
18
y
1
0.8
0.6
0.4
0.2
x
-1 -0.5 0.5 1

y
1
0.5
x
-1 -0.5 0.5 1
-0.5
-1
(c) y = x3 ; dy/dx = 3x2 : the slope is positive except at x = 0 where it is zero.

(d) y = x4 ; dy/dx = 4x3 .
y
1
0.8
0.6
0.4
0.2
x
-1 -0.5 0.5 1
(e) y = x5 ; dy/dx = 5x4 .

y
1
0.5
x
-1 -0.5 0.5 1
-0.5
-1
2.7. For the displacement x = t3 , the velocity of the point is dx/dt = 3t2 and its acceleration is
d2 x/dt2 = 6t. The graph of acceleration against time is a straight line.
2.8. (a) If V = 43 πr3 then dV /dr = 4πr2 .
(b) If S = πd2 then dS/dd = 2πd.
(c) If E = kT 4 then dE/dt = 4kT 3 .
(d) If I = V /R then dI/dV = 1/R.
(e) If H = RI 2 then dH/dI = 2RI.
(f) If V = RT /P then dV /dT = R/P .
2.9.
d d d d
(a) (3x2 − 2x + 1) = 3 (x2 ) − 2 (x) + (1) = 6x − 2.
dx dx dx dx
19
d 7
(b) (x − 3x6 + x + 1) = 7x6 − 18x5 + 1.
dx
d
(c) (x + C) = 1.
dx
d d 2
(d) [x(x − 1)] = (x − x) = 2x − 1.
dx dx
d 2 2 d 4
(e) [x (x + 1) − 1] = [x + x2 + 1] = 4x3 + 2x.
dx dx
d
(f) (ax2 + bx + c) = 2ax + b.
dx
d d 2
(g). [(x − 1)2 ] = (x − 2x + 1) = 2x − 2.
dx dx
2.10. Let m1 and m2 be the slopes of the curves at the point of intersection, and check that
m1 m2 = −1. Then
(a) m1 = (d/dx)(1 + x − x2 ) = 1 − 2x = −1 at x = 1,
m2 = (d/dx)(1 − x + x2 ) = −1 + 2x = 1 at x = 1. Hence m1 m2 = −1 as required.
(b) m1 = −x = −1, m2 = 1 at x = 1.
(c) m1 = −x = −1, m2 = x = 1 at x = 1.
2.11. (a) The curves y = x2 and y = 1 − x2 intersect where x2 = 1 − x2 or where x2 = 21 . Hence
the points of intersection occur at A : ( √12 , 12 ) and B; (− √12 , 12 ).
The slopes of the curves at A are
√ √ √ √
m1 = 2x = 2/ 2 = 2 and m2 = −2x = −2/ 2 = − 2.
Let
√ 1 √ 1
tan α1 = 2 (0 < α1 < π) and tan α2 = − 2 (− π < α1 < 0).
2 2
Using the identity from (1.17a):
√ √
tan α1 − tan α2 2+ 2
tan(α1 − α2 ) = = √ √
1 + tan α1 tan α2 1− 2 2
√
= −2 2
We choose a positive value √for the angle (a sketch of the intersection of the curves is helpful).
Hence α1 − α2 = arctan(−2 2) = 109.47◦ .
The slopes of the curves at B are
√ √ √ √
n1 = −2x = −2/ 2 = − 2 and n2 = 2x = 2/ 2 = 2.
The two slopes at B are interchanged but otherwise the same. Hence the angle between the
tangents will also be 109.47◦ . (Note that in both these cases you might obtain the alternative
angles (180 − 109.47)◦ .)
(b) The curves y = 13 x3 and y = x2 − 2x + 43 intersect where
x3 = 3x2 − 6x + 4 or where (x − 1)(x2 − 2x + 4) = 0.
The only real root is x = 1. Hence the point of intersection is at (1, 31 ). The slopes of the curves at
this point are m1 = 1 and m2 = 0. Let tan α1 = 1 and tan α2 = 0. Then we can choose α1 = 14 π
and α2 = 0. The required angle is 41 π.
20
2.12. Use the limits in Section 2.6. Note that, in all the following, ε is never zero, so cancellation
is legitimate.
ε
(a) lim = lim 1 = 1.
ε→0 ε ε→0
ε 1 1
(b) lim = lim = .
ε→0 2ε ε→0 2 2
ε2
(c) lim = lim ε = 0.
ε→0 ε ε→0
e2ε − 1 eµ − 1
(d) lim = lim = 1, (µ = 2ε) (from (2.11)).
ε→0 2ε µ→0 µ
e2ε − 1 eµ − 1
(e) lim = lim 2 = 2, (µ = 2ε).
ε→0 ε µ→0 µ
sin 2ε sin µ
(f) lim = lim = 1, (µ = 2ε) (from (2.13).
ε→0 2ε µ→0 µ
sin 2ε sin µ
(g) lim = lim 2 = 2, (µ = 2ε).
ε→0 ε µ→0 µ
ln(1 + ε2 ) ln(1 + µ)
(h) lim 2
= lim 2 = 1, (µ = ε2 ) (from (2.14)).
ε→0 ε µ→0 µ
(i) Note that (2.13) is only true if ε is measured in radians. Therefore replace ε degrees by 180ε/π
radians. Hence
sin ε sin(πε/180) π sin µ
lim = lim = lim (µ = πε/180)
ε→0 ε ε→0 ε µ→0 180µ
= π/180 (from (2.13)).

(j)
tan ε sin ε 1 sin ε 1
lim = lim = lim lim
ε→0 ε ε→0 ε cos ε ε→0 ε ε→0 cos ε
= 1×1=1
(k)
sinh ε eε − e−ε e2²−1
lim = lim = lim lim e−ε
ε→0 ε ε→0 2ε ε→0 2ε ε→0
= 1×1=1
e−ε−1 eε − 1
(l) lim = lim lim [−e−ε ] = −1.
ε→0 ε ε→0 ε ε→0
2.13.
· ¸
d cos(x + δx) − cos x
(cos x) = lim
dx δx→0 δx
· ¸
−2 sin 12 (2x + δx) sin 12 (δx)
= lim
δx→0 δx
· ¸
1 sin( 12 δx)
= lim − sin(x + δx) 1
δx→0 2 2 δx
= − sin x
21
2.14.
· 2(x+ε) ¸ · 2ε ¸
d 2x e − e2x e −1
(a) (e ) = lim = 2e2x lim = 2e2x ,
dx ε→0 ε ε→0 2ε
(from (2.11)).
(b)
· ¸
d sin[2(x + ε)] − sin 2x
(sin 2x) = lim
dx ε→0 ε
· ¸
2 sin ε 1
= lim cos (4x + 2ε)
ε→0 ε 2
= 2 cos 2x
· −(x+ε) ¸ · −ε ¸
d −x e − e−x e −1
(c). (e ) = lim = −e−x lim = −e−x .
dx ε→0 ε ε→0 −ε
Thus
d 1 d x 1
(sinh x) = (e − e−x ) = (ex + e−x ) = cosh x.
dx 2 dx 2
d 1 d x 1
(cosh x) = (e + e−x ) = (ex − e−x ) = sinh x.
dx 2 dx 2
2.15.
d d d
(a) (2 sin x − 3 cos x) = 2 (sin x) − 3 (sin x) = 2 cos x + 3 sin x.
dx dx dx
d d 1
(b) (ln 3x) = (ln 3 + ln x) = .
dx dx x
d d 3
(c) (ln x3 ) = (3 ln x) = .
dx dx x
d
(d) (sin x − x) = cos x − 1.
dx
d x 1
(e) (e − 1 − x − x2 ) = ex − 1 − x.
dx 2
2.16. The required tangent lines are

(a) y = 3x − 2; (b) y = 24x − 39; (c) y = −x + 21 π; (d) y = x/e; (e) y = 1; (f) y = −x + 3.
2.17.
dy d2 y d3 y
(a) y = x6 , = 6x5 , = 30x4 , = 120x3 .
dx dx2 dx3
dy d2 y d3 y
(b) y = 3x2 − 2x + 2, = 6x − 2, = 6, = 0.
dx dx2 dx3
dy d2 y d3 y
(c) y = x6 − x2 , = 6x5 − 2x, = 30x4 − 2, = 120x3 .
dx dx2 dx3
22
dy
(d) y = 2 sin x − 3 cos x, = 2 cos x + 3 sin x,
dx
d2 y d3 y
= −2 sin x + 3 cos x, = −2 cos x − 3 sin x.
dx2 dx3
1 dy d2 d3 y
(e) y = ex − 1 − x − x2 , = ex − 1 − x, = ex − 1, = ex .
2 dx dx2 dx3
2.18. To prove that (dxN /dxN )(xN ) = N ! for all integers N ≥ 1!. We can confirm the formula
for the case N = 1:
d
(x) = 1 = 1! (i)
dx
as a starting-point in a step-by-step argument.
Suppose that we have somehow established that the result is true for any one particular value
of N , say for N = K, so that
dN (xN )
= K! when N = K. (ii)
dxN
Next, consider the transition to N = K + 1:
· K ¸ · ¸
dK+1 K+1 d d K+1 dK d K+1
(x ) = (x ) = (x )
dxK+1 dx dxK dxK dx
dK
= {(K + 1)xK } (by (2.9))
dxK
= (K + 1)K! (by (ii))
= (K + 1)!
In other words, if (ii) is true for some integer N = K, it follows that it is also true for
N = K + 1. Since we now know it is true for N = K + 1, the same argument implies that it is
true for N = (K + 1) + 1 = K + 2; and so on for all subsequent values of N .
But we have verified its truth in the case N = 1 (in equation (i)); therefore (ii) is true for all
values of N . This is proof by induction.
2.19. If y = x2 (x2 − 1), then
dy d2 y
= 4x3 − 6x, = 12x2 − 6.
dx dx2
√ √
(a) The slope of √
the curve is positive
√ where dy/dx > 0 or where x(2x + 3)(2x − 3) > 0. This
occurs where − 21 3 < x < 0 and 12 3 < x. √ √ √
of the curve is negative where x(2x + 3)(2x − 3) < 0, that is where x < − 12 3
(b) The slope √
and 0 < x < 12 3. √ √ √
second derivative is positive where 12(x + 12 2)(x − 12 2) > 0, that is where x > 21 2 and
(c) The √
x < − 12 2. √ √
(d) The second derivative is negative where − 12 2 < x < 12 2.
2.20. At x = x0 the tangent has slope m0 = 2x0 . Hence the slope of the normal is −1/m0 =
−1/(2ax0 ). The equation of the normal is therefore
1
y − ax20 = − (x − x0 ).
2ax0
Chapter 3: Further techniques for differentiation
23
3.1.
d d d
(a) (xex ) = x (ex ) + (x)ex = xex + ex .
dx dx dx
d
(b) (x sin x) = x cos x + sin x.
dx
d
(c) (x cos x) = −x sin x + cos x
dx
d x
(d) (e sin x) = ex cos x + ex sin x.
dx
d
(e) (x ln x) = 1 + ln x.
dx
d 2
(f) (x ln x) = x2 x−1 + 2x ln x = x + 2x ln x.
dx
d x ex
(g) (e ln x) = + ex ln x.
dx x
d 2 x
(h) (x e ) = x2 ex + 2xex .
dx
d
(i) (sin x cos x) = − sin2 x + cos2 x = cos 2x.
dx
d 2 3 d 5
(j) (x x ) = x2 .3x2 + 2x.x3 = 5x4 = (x ).
dx dx
3.2. All these problems illustrate the reciprocal and quotient rules given in (3.2).
(a)
d d ³ cos x ´ sin x(− sin x) − cos x(cos x)
(cot x) = =
dx dx sin x sin2 x
1
= − 2 = −cosec 2 x
sin x
µ ¶ µ ¶
d x 1 d d 1
(b) = (x + 1) (x) − x (x + 1) = .
dx x+1 (x + 1)2 dx dx (x + 1)2
µ ¶
d sin x 1
(c) = (x cos x − sin x).
dx x x2
µ ¶
d ex ex
(d) = (x − 1).
dx x x2
µ ¶
d x2 − 1 1 4x
(e) = [(x2 + 1)(2x) − (x2 − 1)(2x)] = 2 .
dx x2 + 1 (x2 + 1)2 (x + 1)2
24
µ ¶
d tan x 1 2 1
(f) = (x sec2 x − 2x tan x) = 3 (x sec2 x − 2 tan x).
dx x2 x4 x
(g)
µ ¶
d sin x + cos x
dx sin x − cos x
(sin x − cos x)(cos x − sin x) − (sin x + cos x)(cos x + sin x)
=
(sin x − cos x)2
2
= − .
(sin x − cos x)2
µ ¶
d d 1 sin x
(h) (sec x) = = = sec x tan x.
dx dx cos x cos2 x
µ ¶
d d 1 cos x
(i) (cosec x) = =− = − cot xcosec x.
dx dx sin x sin2 x
µ ¶
d x −2 − 3x2
(j) 2
= .
dx 3x − 2 (−2 + 3x2 )2
µ ¶
d 1 −1 − 4x3
(k) = .
dx x(x3 + 1) x2 (x3 + 1)
µ ¶
d 1 1
(l) =− .
dx ln x x(ln x)2
µ ¶
d n d 1
(m) (x ) = = nxn−1 (by the quotient rule).
dx dx x−n
µ ¶
d 1 1
(n) =− .
dx x+1 (x + 1)2
µ ¶
d −x d 1 1 d x
(o) (e ) = =− (e ) = −e−x .
dx dx ex e2x dx
µ ¶
d d 1
(p) (cot x) = −cosec 2 x (as in (a)).
=
dx dxtan x
µ ¶ µ ¶
d −2 d ln x 1 x2 1 − 2 ln x
(x ln x) = = − 2x ln x = .
dx dx x2 x4 x x3
3.3.
· ¸ · ¸
d 1 1 d2 1 2
(a) = 2
; 2
= ,
dx 1 − x (1 − x) dx (1 − x) (1 − x)2
· ¸
d3 1 6
= .
dx3 (1 − x) (1 − x)4
25
d d2
(b) (x sin x) = x cos x + sin x; (x sin x) = 2 cos x − x sin x;
dx dx2
d3
(x sin x) = −x cos x − 3 sin x.
dx3
µ ¶ µ ¶
d x 1 d2 x 2
(c) =− ; = ;
dx x−1 (x − 1)2 dx2 x − 1 (x − 1)3
µ ¶
d3 x 6
=− .
dx3 x − 1 (x − 1)4
(d) Let y = f (x)g(x). Then
dy df dg
=g +f ,
dx dx dx
d2 y df dg d2 f d2 g
= 2 + g + f ,
dx2 dx dx dx2 dx2
d3 y dg d2 f df d2 g d3 f d3 g
= 3 + 3 + g + f .
dx3 dx dx dx dx2 dx3 dx3
3.4. These problems use the chain rule (3.3) in the form
dy dy du
= .
dx du dx
(a) Let u = sin x. Then y = u2 , and
dy dy du
= = 2u cos x = 2 sin x cos x = sin 2x.
dx du dx
(b) Let u = cos x, y = u2 . Then (d/dx)(cos2 x) = −2 sin x cos x = − sin 2x.
(c) Let u = x2 , y = sin u. Then (d/dx)(sin x2 ) = 2x cos x2 .
(d) Let u = x2 , y = cos u. Then (d/dx)(cos x2 ) = −2x sin x2 .
(e) Let u = tan x, y = u2 . Then d/dx)(tan2 x) = 2 sec2 x tan x.
(f) Let u = x2 , y = tan u. Then (d/dx)(tan x2 ) = 2x sec2 x2 .
(g) Let u = 1/x, y = cos u. Then d/dx)[cos(1/x)] = 2 sin(1/x)/x2 .
(h) Let u = −x, y = eu . Then (d/dx)(e−x ) = −e−x .
(i) Let u = 1/(x + 1), y = u5 . Then (d/dx)(1/(x + 1)5 ) = −5/(x + 1)6 .
(j) Let u = x3 + 1, y = u4 . Then (d/dx)[(x3 + 1)4 ] = 12x2 (x3 + 1)3 .
(k) Let u = 3x, y = sin u. Then (d/dx)(sin 3x) = 3 cos 3x.
(l) Let u = 12 x, cos u. Then (d/dx)(cos 12 x) = − 12 sin 12 x.
(m) Let u = 12 x, y = tan u. Then (d/dx)(tan 12 x) = 12 sec2 x.
(n) Let u = −3x, y = eu . Then (d/dx)(e−3x ) = −3e−3x .
(o) Let u = 2x + 1, y = sin u. Then (d/dx)[sin(2x + 1)] = 2 cos(2x + 1).
(p) Let u = 3x − 2, y = cos u. Then (d/dx)[cos(3x − 2)]) = −3 sin(3x − 2).
(q) Let u = 1 − 2x, y = tan u. Then (d/dx)[tan(1 − 2x)] = −2 sec2 (1 − 2x).
(r) Let u = 1/x, y = eu . Then (d/dx)(e1/x ) = −e1/x /x2 .
3.5. All these problems use the result that (d/dx)xα = αxα−1 .
d −2
((a)) (x ) = −2x−3 .
dx
d −1 1
(b) (x ) = − 2 .
dx x
26
d 1 1 2
(c) x 3 = x− 3 .
dx 3
d −1 1 4
(d) (x 3 ) = − x− 3 .
dx 3
d 3 3 1
(e) (x 2 ) = x 2 .
dx 2
d √ d 1 1 1
(f) ( x) = (x 2 ) = x− 2 .
dx dx 2
d √ 3 d 3 3 1
(g) x = (x 2 ) = x 2 .
dx dx 2
µ ¶
d 1 d −1 1
(h) = (x ) = − 2 .
dx x dx x
· ¸
d 1 d −1 −1
(i) √ = (x 2 ) = 3 .
dx x dx 2x 2
3.6. (a)
d 1 1 sin x
(x 2 sin x) = x 2 cos x + 1 .
dx 2x 2
d 1 cos x
(b) (sin 3 x) = 2 .
dx 3 sin 3 x
d 1 x
(c) [(x2 + 1)− 2 ] = − 3 .
dx 2
(x + 1) 2
d
(d) [sin2 (3t + 1)] = 6 cos(3t + 1) sin(3t + 1).
dt
d −t
(e) (e cos t) = −e−t (cos t + sin t).
dt
d −t
(f) (e sin t) = e−t (cos t − sin t).
dt
d −2t
(g) (e cos 3t) = −e−2t (2 cos 3t + 3 sin 3t).
dt
d −3t
(h) (e cos 2t) = −e−3t (3 cos 2t + 2 sin 2t).
dt
d
(i) (sin x cos2 x) = cos3 x − 2 cos x sin2 x.
dx
27
d
(j) (sin2 x cos x) = 2 cos2 x sin x − sin3 x.
dx
"µ ¶2 #
d sin x 2 cos x sin x 2 sin2 x
(k) = − .
dx x x2 x3
d £ ¤
(l) x(sin3 x) = 3x cos x sin2 x + sin3 x.
dx
d £ ¤
(m) x(cos3 x) = −3x cos2 x sin x + cos3 x.
dx
3.7. (a)
d d 1
(cos2 x) = (1 + cos 2x) = − sin 2x.
dx dx 2
d d 1
(sin2 x) = (1 − cos 2x) = sin 2x.
dx dx 2
(b)
d d
(cos2 x) = (cos x cos x) = − cos x sin x − sin x cos x = − sin 2x.
dx dx
d d
(sin2 x) = (sin x sin x) = sin x cos x + cos x sin x = − sin 2x.
dx dx
(c) To apply the chain rule let u = cos x, y = u2 . Then
d d 2 d
(cos2 x) = (u ) (cos x) = −2u sin x = −2 cos x sin x = − sin 2x.
dx du dx
Let u = sin x Then
d d 2 d
(sin2 x) = (u ) (sin x) = 2u cos x = 2 sin x cos x = sin 2x.
dx du dx
3.8.
d2 x
(a) + 4x = (−4A cos 2t − 4B sin 2t) + 4(A cos 2t + B sin 2t) = 0
dt2
d2 x
(b) + n2 x = (−n2 A cos nt − n2 B sin nt) + n2 (A cos nt + B sin nt) = 0.
dt2
d2 x
(c) − 9x = (9Aent + 9Be−nt ) − 9(Aent + Be−nt ) = 0.
dt2
d2 x
(d) − n2 x = (n2 Aent + n2 Be−nt ) − n2 (Aent + Be−nt ) = 0.
dt2
dx
(e) = (−A + B)e−t cos t − (A + B)e−t sin t.
dt
d2 x
= −2Be−t cos t + 2Ae−t sin t.
dt2
28
Hence
d2 x dx
2
+2 + 2x =
dt dt
[−2Be−t cos t + 2Ae−t sin t] +
2[(−A + B)e−t cos t − (A + B)e−t sin t] + 2[Ae−t cos t + Be−t sin t] = 0.
(f) The fourth derivative of each term in y returns the same function in each case. Hence
d4 y
− y = 0.
dx4
3.9. Use the chain rule in the form

dy dy dv du
= .
dx dv du dx
The intermediate variables are defined in each problem.
(a) Let u = cos x, v = u2 , so that y = ev . Hence
dy 2
= ev .2u.(− sin x) = −2ecos x cos x sin x.
dx
(b) Let u = x2 , v = cos u, so that y = e−v . Hence
dy 2
= (−e−v ).(− sin u).2x = 2x sin(x2 )e− cos x .
dx
(c) Let u = x2 , v = cos u, so that y = ln v. Hence
dy 1
= .(− sin u).2x = −2x tan(x2 ).
dx v
(d) Let u = x2 , v = eu − 1, so that y = v 4 . Hence
dy 2 2
= 4v 3 .eu .2x = 8xex (ex − 1)3 .
dx
3.10. Use the result (3.7) which states that if y = u(x)v(x)w(x) then ln y = ln u + ln v + ln w
which when differentiated gives
dy du dv dw
= vw + wu + uv .
dx dx dx dx
(a) Let u = x, v = ex , w = sin x. Then
dy
= ex . sin x.1 + sin x.x.ex + x.ex . cos x = ex [sin x + x sin x + x cos x].
dx
(b) Different variables are used. Let x = tet cos t, and let u = t, v = et , w = cos t. Hence
dx
= et . cos t.1 + cos t.t.et + t.et .(− sin t) = et [cos t + t cos t − t sin t].
dt
1 1
(c) Let u = x 2 , v = e2x , w = sin 2 3x. Then
dy 1 1 1 1 1 1 3 3
= e2x . sin 2 3x.( x− 2 ) + sin 2 3x.x 2 .2e2x + x 2 .e2x . cos 3x sin− 2 3x
dx 2 2
e2x
= 1 1 [3x cos 3x + (4x + 1) sin 3x].
2x 2 sin 2 x
29
This function and its derivative will only be real for restricted values of x—for example, for 0 ≤
x ≤ 13 π.
3.11. (a) Treating y as a function of x, and using the chain rule for y(x),
d 2 dy
(x + y 2 ) = 0, or 2x + 2y = 0.
dx dx
Hence dy/dx = −x/y as required. Solving the equation x2 + y 2 = 4 for y, it follows that y =
1
±(4 − x2 ) 2 . Differentiating with respect to x, we have
dy 1 x
= ∓x(4 − x2 )− 2 = − ,
dx y
after substitution back in terms of y. This agrees with the answer obtained by the implicit method.
Note that will always be two points on a circle which have the same slope. The tangent is always
perpendicular to the radius to the point which has slope m = y/x.
(b) In this case for x > 0, y ≥ 0,
d 1 1 1 1 1 1 dy
(x 2 + y 2 ) = 0, or x− 2 + y − 2 = 0.
dx 2 2 dx
Hence r
dy y
=− .
dx x
(c) In this case
d 3 dy dy
(x + xy − y 3 ) = 0 or 3x2 + y + x − 3y 2 .
dx dx dx
Hence
dy y + 3x2
=− .
dx x − 3y 2
(d) In this case
d dy dy
(x sin y − y sin x) = 0 or sin y + x cos y − sin x − y cos x = 0.
dx dx dx
Hence
dy y cos x − sin y
= .
dx x cos y − sin x
3.12. The expression for dy/dx obtained from the implicit relation f (x, y) = c does not depend
on c. For example for x2 + y 2 = c, we always have dy/dx = −x/y. However, the value of dy/dx
will depend indirectly on c since x and y must always satisfy x2 + y 2 = c.
3.13. If xy 2 − x2 y = 1, then
dy dy
y 2 + 2xy − 2xy − x2 = 0. (i)
dx dx
Hence
dy 2xy − y 2
= . (ii)
dx 2xy − x2
Differentiate (i) again with respect to x:
µ ¶2
dy dy dy d2 y dy dy d2 y
2y + 2y + 2x + 2xy 2 − 2y − 2x − 2x − x2 2 = 0. (iii)
dx dx dx dx dx dx dx
Eliminate dy/dx between (ii) and (iii): the answer is
d2 y 6xy(−x3 + 2x2 y − 2xy 2 + y 3 )
= .
dx2 (2xy − x2 )3
30
3.14. (a) Let y = arcsin x. Then x = sin y. Differentiating with respect to y,
dx √ √
= cos y = (1 − sin2 y) = (1 − x2 ).
dy
Hence µ ¶−1
dy dx 1
= =√ .
dx dy (1 − x2 )
(b) Let y = arccos x so that x = cos y. Then
µ ¶−1
dy dx 1 1
= =− = −√ .
dx dy sin y (1 − x2 )
(c) Let y = arctan x so that x = tan y. Then

µ ¶−1
dy dx 1 1
= = 2
= cos2 y = .
dx dy sec y 1 + x2
(d) Let y = sinh−1 x so that x = sinh y. Then

µ ¶−1
dy dx 1 1 1
= = =√ =√ .
dx dy cosh y (1 + sinh2 y) (1 + x2 )
(e) Let y = cosh−1 x so that x = cosh y. Then

µ ¶−1
dy dx 1 1 1
= = =√ 2 =√ 2 .
dx dy sinh y (cosh −1) (x − 1)
(f) Let y = tanh−1 x so that x = tanh y. Then

µ ¶−1
dy dx 1 1 1
= = = = .
dx dy sech 2 y 1 − tanh2 y 1 − x2
3.15. (a) If r = sin 21 θ, then the (x, y) coordinates are
1 1
x = r cos θ = sin θ cos θ, y = r sin θ = sin θ sin θ.
2 2
Using parametric differentiation,
Á 1
dy dy dx 2 cos 12 θ sin θ + sin 12 θ cos θ
= = 1 .
dx dθ dθ 2 cos 12 θ cos θ − sin 12 θ sin θ
At θ = 12 π,
1
2 2 +0
√
dy 1
= =− .
dx 0 − √12 2
(b) If r = 1 + sin2 θ, then
x = r cos θ = (1 + sin2 θ) cos θ, y = r sin θ = (1 + sin2 θ) sin θ.
Hence
dx 1
= − sin θ − sin3 θ + 2 cos2 θ sin θ = − √ at θ = 14 π,
dθ 2 2
and
dy 1 3 5 1
= cos θ + 3 sin2 θ cos θ = √ + √ = √ at θ = π.
dθ 2 2 2 2 2 4
31
Hence Á
dy dy dx
= = −5
dx dθ dθ
.
3.16. (a) For x = t3 and y = t2 ,

Á
dy dy dx 2t 2 2
= = 2 = = 1 .
dx dt dt 3t 3t 3x 3
(b) For x = 2 cos t and y = 2 sin t,
Á
dy dy dx 2 cos t x
= = =± √ ,
dx dt dt −2 sin t 2 (4 − x2 )
assuming 0 ≤ t ≤ 12 π.
3.17. Elimination of t between x and y, using the identity cos2 t + sin2 t = 1, gives the ellipse
x2 y2
+ = 1.
a2 b2
The derivative is given by
Á
dy dy dx b cos t b
= = = − cot t.
dx dt dt −a sin t a
The speed of the point is

"µ ¶2 µ ¶2 # 12
dx dy √
+ = [a2 sin2 t + b2 cos2 t].
dt dt
3.18. In exponential form ax = ex ln a . Hence

d x
(a ) = (ln a)ex ln a = xa ln a.
dx
Chapter 4: Applications of differentiation
4.1. f (u) = u2 , so f 0 (u) = 2u and f 00 (u) = 2 for all arguments u.

(a) Let u = t. Then f 0 (t) = 2t.
(b) Let u = t2 . Then f 0 (t2 ) = 2u = 2t2 .
(c) (d/dt)f (t2 ) = (d/dt)t4 = 4t3 .
1 1 1
(d) Let u = t 2 . Then f 0 (t 2 ) = 2u = 2t 2 .
1
(e) (d/dt)f (t 2 ) = (d/dt)t = 1.
1 1
(f) Let u = t 2 . Then f 00 (t 2 ) = 2.
4.2. Denote the function in each case by f (x). The stationary points are given by f 0 (x) = 0. If
A : x = a is a stationary point, then A is a minimum if f 00 (a) > 0, or a maximum if f 00 (a) < 0. If
f 00 (a) = 0, then the stationary point can be a minimum, maximum or point of inflection depending
on the sign of f 0 (x) on either side of x = a.
(a) Since f (x) = x2 − x, then
f 0 (x) = 2x − 1, f 00 (x) = 2.
Stationary point: x = 12 .
32
Test: f 00 ( 12 ) = 2 > 0 so x = 12 is a minimum.
(b) Since f (x) = x2 − 2x − 3, then
f 0 (x) = 2x − 2, f 00 (x) = 2.
Stationary point: x = 1.
Test: f 00 (1) = 2 > 0 so x = 1 is a minimum.
(c) Since f (x) = x ln x, then
1
f 0 (x) = 1 + ln x, f 00 (x) = .
x
Stationary point: x = e−1 .
Test: f 00 (e−1 ) = e > 0 so x = e−1 is a minimum.
(d) Since f (x) = xe−x , then
f 0 (x) = (1 − x)e−x , f 00 (x) = (−2 + x)e−x .
Test: f 00 (1) = −e−1 < 0 so x = 1 is a maximum.
(e) Since f (x) = 1/(x2 + 1), then
−2x 2(3x2 − 1)
f 0 (x) = , f 00 (x) = .
(x + 1)2
2 (x2 + 1)3
Test: f 00 (0) = −2 < 0 so x = 0 is a maximum.
(f) Since f (x) = x2 − 3x + 2, then
f 0 (x) = 2x − 3, f 00 (x) = 2.
Test: f 00 ( 32 ) = 2 > 0 so x = 32 is a minimum.
(g) Since f (x) = ex + e−x , then
f 0 (x) = ex − e−x , f 00 (x) = ex + e−x .
Test: f 00 (0) = 2 > 0 so x = 0 is a minimum.
(h) Since f (x) = x2 + 4x + 2, then
f 0 (x) = 2x + 4, f 00 (x) = 2.
Stationary point: x = −2.

Test: f 00 − 2) = 2 > 0 so x = −2 is a minimum.
(i) Since f (x) = x − x3 , then
f 0 (x) = 1 − 3x2 , f 00 (x) = −6x.

√
Stationary points:
√ x = ±1/
√ 3. √
00
Tests: f√ (1/ 3) √
= −6/ 3 < 0, so x√= 1/ 3 is a maximum.
f 00 (−1/ 3) = 6/ 3 > 0 so x = −1/ 3 is a minimum.
(j) Since f (x) = x2 (x − 1), then
f 0 (x) = 3x2 − 2x, f 00 (x) = 6x − 2.
Stationary points: x = 0 and x = 23 .
33
Tests: f 00 (0) = −2 < 0 so x = 0 is a maximum.
f 00 ( 23 ) = 2 > 0 so x = 23 is a minimum.
(k) Since f (x) = sin x − cos x, then
f 0 (x) = cos x + sin x, f 00 (x) = − sin x + cos x.
Stationary points: √ x = 43 π and 74 π for 0 < x < 2π.

00 3 3
Tests: f (√ 4 π) = − 2 < 0 so x = 4 π is a maximum;
00 7 7
f ( 4 π) = 2 > 0 so x = 4 π is a minimum.
(l) Since f (x) = sin x cos x = 12 sin 2x, then
f 0 (x) = cos 2x, f 00 (x) = −2 sin 2x.
Stationary points: x = − 43 π, x = − 14 π, x = 14 π, x = 43 π for −π < x < π.

Tests: f 00 (− 43 π) = −2 < 0 so x = − 34 π is a maximum;
f 00 (− 14 π) = 2 > 0 so x = − 14 π is a minimum;
f 00 ( 14 π) = −2 < 0 so x = 14 π is a maximum;
f 00 ( 34 π) = 2 > 0 so x = 34 π is a minimum.
(m) Since f (x) = e−x sin x, then
f 0 (x) = e−x (− sin x + cos x), f 00 (x) = −2e−x cos x.
Stationary points occur where tan x = 1, at x = (n + 14 )π, (n = 0, ±1 ± 2, . . .).

√ 1
Tests: f 00 [(n + 14 )π] = − 2(−1)n e−(n+ 4 )π < 0 or > 0 according as n is even or odd. Hence the
stationary point is a maximum x = (n + 14 )π is a maximum if n is even, and a minimum if n is
odd.
1
(n) Since f (x) = e− 3 x sin 2x, then
1 1
f 0 (x) = 13 e− 3 x (− sin 2x + 6 cos 2x), f 00 (x) = 19 e− 3 x (−12 cos 2x − 35 sin 2x).
Stationary points occur where tan 2x = 6 at x = α+ 12 nπ, (n = 0, ±1±2, . . .), where α = 21 arctan 6.
1 1 √
Tests: f 00 (α + 12 nπ) = −222e− 3 (α+ 2 nπ) (−1)n /[9 37] < 0 or > 0 according as n is even or odd.
Hence the stationary point x = α + 21 nπ is a maximum if n is even, and a minimum if n is odd.
(o) Since f (x) = x − cos x, then
f 0 (x) = 1 + sin x, f 00 (x) = cos x.
Stationary points occur where sin x = −1 at x = (2n − 12 )π, (n = 0, ±1, ±2, . . .).
Tests: f 00 [(2n − 12 )π] = 0 for all n. Hence the test fails. But f 0 (x) = 1 + sin x ≥ 0 for all x. Hence
all the stationary points must be points of inflection.
(p) Since f (x) = 2ex − 12 e2x , then
f 0 (x) = 2ex − e2x , f 00 (x) = 2ex − 2e2x .
Staionary point occurs where ex = 2 at x = ln 2.

Test: f 00 (ln 2) = −4 < 0 so x = ln 2 is a maximum.
(q) Since f (x) = x2 e−x , then
f 0 (x) = xe−x (2 − x), f 00 (x) = e−x (2 − 4x + x2 ).
Stationary points at x = 0 and x = 2.

Tests: f 00 (0) = 2 > 0 so that x = 0 is a minimum; f 00 (2) = −2e−2 < 0 so that x = 2 is a maximum.
(r) Since f (x) = (ln x)/x, then
1 1
f 0 (x) = (1 − ln x), f 00 (x) = (−3 + 2 ln x).
x2 x3
34
Stationary point where ln x = 1 at x = e.
Test: f 00 (e) = −e−3 < 0 so x = e is a maximum.
(s) Since f (x) = (1 − x)3 , then
f 0 (x) = −3(1 − x)3 , f 00 (x) = 6(1 − x)2 .
Test: f 00 (1) = 0 and the test fails. However, f 0 (x) ≤ 0 for all x. Hence x = 1 is a point of inflection.
(t) Since f (x) = sin3 x, then
f 0 (x) = 3 sin2 x cos x, f 00 (x) = 3 sin x(2 − 3 sin2 x).
Stationary points occur where sin x = 0, at x = nπ, and where cos x = 0, at x = (n + 21 )π for
n = 0, ±1, ±2, . . ..
Tests: f 00 (nπ) = 0 so that the test fails. However, if n is even, then f 0 (x) is positive in a small
interval including x = nπ, and if n is odd, then f 0 (x) is negative in a small interval including
x = nπ. In both cases, therefore, the stationary point is a point of inflection.
f 00 [(n + 12 )π] = −3(−1)n so that x = (n + 12 )π is a maximum if n is even, and a minimum if n
is odd.
2
(u) Since f (x) = e−x , then
2 2
f 0 (x) = −2xe−x , f 00 (x) = 2e−x (2x2 − 1).
Test: f 00 (0) = −2 < 0 so x = 0 is a maximum.
2
(v) Since f (x) = ex −x
, then
2 2
f 0 (x) = (2x − 1)ex −x
, f 00 (x) = (4x2 − 4x + 3)ex −x
.
1
Test: f 00 ( 12 ) = 2e− 4 > 0 so x = 1
2 is a minimum.
(w) Since f (x) = x + x−1 , then
1 2
f 0 (x) = 1 − , f 00 (x) = .
x2 x3
Stationary points: x = ±1.
Tests: f 00 (1) = 2 > 0 so x = 1 is a minimum; and f 00 (−1) = −2 < 0 so x = −1 is a maximum.
(x) Since f (x) = x3 e−x , then
f 0 (x) = x2 e−x (3 − x), f 00 (x) = xe−x (6 − 6x + x2 ).
Stationary points: x = 0 and x = 3.

Tests: f 00 (0) = 0 and the test fails. In a small interval which includes the origin f 0 (x) > 0 which
means that x = 0 is a point of inflection.
f 00 (3) = −9e−3 < 0 so x = 3 is a maximum.
4.3. If y = f [u(x)], then by the chain rule
dy df du
= = f 0 (u)u0 (x),
dx du dx
and
d2 y d 0 d 0
= [f (u)u0 (x)] = [f (u)]u0 (x) + f 0 (u)u00 (x) (product rule)
dx2 dx dx
= f 00 (u)[u0 (x)]2 + f 0 (u)u00 (x)
35
Since f 0 (u) > 0 for all u, then dy/dx can only be zero if u0 (x) = 0 (by the chain rule (3.3)). Hence
f [u(x)] and u(x) have stationary points only at the same values of x.
In 4.2(v), f (u) = eu and u = x2 − x.
4.4. If the sides have lengths x > 0 and y > 0, then the given area A = xy. The length of the
perimeter is P = 2x + 2y. Eliminate y so that
2A
P = 2x + .
x
The first and second derivatives of P are
dP 2A d2 P 4A
=2− 2, 2
= 3.
dx x dx x
√
Hence the perimeter length
√ is stationary where dP/dx = 0: at x = ± A. Since x > 0, choose the
stationary value x = A. For this value
d2 P 4A
= 3 > 0,
dx2 A2
√
so that the perimeter is a minimum when x = y = A. The piece of ground must be a square.
4.5. Let the the base of the cross-section be x > 0 which will also be the diameter of the semicircle,
and let the height of the rectangle be y > 0. The given area A of the tunnel cross-section is
A = xy + 18 πx2 .
The length of the perimeter is P = x + 2y + 12 πx. Eliminate y, so that

µ ¶
2 1 2 1 2A 1
P = A − πx + x(1 + π) = + (1 + π)x.
x 8 2 x 4
This is stationary where

dP 2A 1
= − 2 + 1 + π = 0,
dx x 4
q
8A
which occurs at x = 4+π (choosing the positive root). The perimeter is a minimum since
d2 P 4A
= 3 >0
dx2 x
at the stationary point.
4.6. Let r be the radius of the base and h the height of the drum. The volume V of the drum is
given by V = πr2 h and its prescribed surface area by A = 2πr2 + 2πrh. We are given that A is a
constant, so eliminate h in the expression for V :
1
V = [Ar − 2πr3 ].
2
Differentiating
dV 1 d2 V
= [A − 6πr2 ], = −6πr.
dr 2 dr2
The volume is stationary where r
dV A
= 0, at r = ,
dr 6π
choosing the positive root. Obviously d2 V 2
√/dr < 0 which proves that this radius gives a minimum
volume. The height of this drum is h = [2A/(3π)] which is equal to its diameter.
36
4.7. Similar to 4.6: the volume is given by the same formula but the prescribed A is different:
V = πr2 h, A = πr2 + 2πrh.
Elimination of h leaves
1
V = r(A − πr2 ).
2
Differentiating
dV 1 d2 V
= [A − 3πr2 ], = 3πr.
dr 2 dr2
Hence the radius and height of the drum of minimum volume are
r
A
r=h= .
3π
4.8. (a) y = 1/(x2 + 1):
y
1
0.8
0.6
0.4
0.2
x
-4 -2 2 4

2
(b) y = ex :
y
0.8
0.6
0.4
0.2
x
-2 -1 1 2
(c) y = x/(x − 1):

y
4
x
-3 -2 -1 1 2 3 4
-1
-2
-3
-4
(d) y = xe−x :
(e) y = x2 e−x :
(f) y = x3 e−x :
(g) y = e2x − 4ex :
(h) y = (ln x)/x for x > 0:
(i) [ln(−x)]/x for x < 0:
(j) y = x ln x − x for x > 0:
37
y
0.5
x
-1 -0.5 0.5 1 1.5 2
-0.5
-1
-1.5

y
1
0.8
0.6
0.4
0.2
x
-1 -0.5 0.5 1 1.5 2
-0.2
(k) y = sin(1/x):
(l) y = (x2 − 1)2 :
(m) y = x(x2 − 1)2 :
(n) y = (sin x)/x:
4.9. (a) y = 1/(x2 − 1):
(b) y = x/(x2 − 1):
(c) 1/[x(x − 2)]:
(d) y = x3 /(1 − x):
(e) y = (x + 2)/(x − 1):
(f) y = 1/(x + 1) + 1/(x + 2):
4.10. The incremental formula given by (4.4) is
δy ≈ f 0 (a)δx at x = a.
The exact value is given by

δy = f (a + δx) − f (a).
(a) f (x) = x3 : δy ≈ 3x2 δx. With x = 2 and δx = 0.1, the approximate and exact values are given
by
δy ≈ 1.200, δy = (2.1)3 − 23 = 1.157 . . . .
1
(b) f (x) = x sin x: δy ≈ (sin x + x cos x)δx. With x = 2π and δx = −0.2 the approximate and
exact values are given by
1 1 1
δy ≈ (sin π + π cos π)(−0.2) = −0.2,
2 2 2
1 1 1 1
δy = ( π − 0.2) sin( π − 0.2) − π sin π = −0.227 . . . .
2 2 2 2
y
1
0.75
0.5
0.25
x
-1 -0.5 0.5 1
-0.25
-0.5
-0.75
-1
38
y
10
7.5
5
2.5
x
-1 -0.5 0.5 1 1.5 2
-2.5
-5
-7.5

y
x
0.5 1 1.5 2
-2
-4
-6
-8
Figure 43: Problem 4.8h
(c) f (x) = cos x: δy ≈ − sin xδx. With x = 14 π and δx = 0.1 the approximate and exact values
are given by
1
δy ≈ (− sin π)(0.1) = −0.0707 . . . ,
4
1 1
δy = cos( π + 0.1) − cos( π) = −0.0741 . . . .
4 4
2
(d) f (x) = (1 + x)/(1 − x): δy = 2/(1 − x) δx. With x = 2 and δx = −0.2 the approximate and
2
δy ≈ (−0.2) = −0.4, δy = −0.5.
(1 − 2)2
(e) y = tan x: δy ≈ sec2 xδx. With x = 14 π and δx = 0.1 the approximate and exact values are
given by
1 1 1
δy ≈ (sec2 π)(0.1) = 0.2, δy = tan( π + 0.1) − tan π = 0.223 . . . .
4 4 4
2 0 2 2
(f) f (x) = 1/(1 − x ): f (x) = 2x/(1 − x ) . With x = 0.5 and δx = ±0.1 the approximate and
1
δy ≈ (±0.1) = ±0.177 . . . ,
(1 − 0.52 )2
1 1
δy = 2
− = 0.229 . . . or − 0.142 . . . .
1 − (0.05 ± 0.1) 1 − (0.05)2
4.11. (a) With f fixed,

dv f2
=− .
du (u − f )2
y
6
5
4
3
2
1
x
-3 -2.5 -2 -1.5 -1 -0.5
-1
Figure 44: Problem 4.8i
39
y
0.5
x
1 2 3 4
-0.5
-1
Figure 45: Problem 4.8j
y
1
0.5
x
0.1 0.2 0.3 0.4 0.5 0.6 0.7
-0.5
-1
Figure 46: Problem 4.8k
Hence, with f = 0.75, u = 1.25 and δu = 0.05,
f 2 δu −(0.75)2 (0.05)
δv ≈ − 2
= = −0.112 . . . .
(u − f ) (1.25 − 0.75)2
(b) The voltage is given by

E(R1 R4 − R2 R3 )
v= .
(R1 + R2 )(R3 + R4 )
Its derivative with respect to R1 is
dv ER2
= .
dR1 (R1 + R2 )2
Hence
ER2 5
δv ≈ = δR1 .
(R1 + R2 )2 18
(c) With b and A constant in a = b sin A/(sin B),
da −b sin A cos B
= = −a cot B.
dB sin2 B
Hence
δa ≈ −a cot B δB.
(d) In terms of a, b, c,
1√
A= [(a + b + c)(−a + b + c)(a − b + c)(a + b − c)].
4
y
1.2
1
0.8
0.6
0.4
0.2
x
-1.5 -1 -0.5 0.5 1 1.5
-0.2
Figure 47: Problem 4.8l
40
y
1
0.8
0.6
0.4
0.2
x
-1.5 -1 -0.5 0.5 1 1.5
-0.2
-0.4
Figure 48: Problem 4.8m

y
1
0.8
0.6
0.4
0.2
x
-10 -5 5 10
-0.2
-0.4
Figure 49: Problem 4.8n
Logarithmic differentiation (see equation (3.7)) gives

· ¸
1 dA 1 1 1 1 1
= + + + .
A dc 2 a + b + c −a + b + c a − b + c a + b − c
The incremental formula for δA becomes, at a = 2, b = 4, c = 5

dA 25
δA ≈ δc = − √ (0.1) = 0.0822 . . . .
dc 2 231
4.12. Given C = P (1 + r)n .

(a) With n and P fixed,
dC
= P n(1 + r)n−1 , so that δC ≈ P n(1 + r)n−1 δr.
dr
(b) With r and P fixed,
dC d n ln(1+r)
=P e = P (1 + r)n ln(1 + r) see Problem 3.18.
dn dn
Hence
δC ≈ P (1 + r)n ln(1 + r)δn.
(c) Suppose that P = £100, r = 0.05 (5% ) and n = 10 years. The tables below show comparisons
between the approximate increments δC for decreasing values of δr (n fixed) and δn (r fixed).
y
12
10
8
6
4
2
x
-2 -1 1 2
41
y
10
x
-2 -1 1 2
-5
-10

y
10
x
-1 1 2 3
-5
-10
δr approximate increment in C exact increment in C

P n(1 + r)n−1 δr P (1 + r + δr)n − P (1 + r)n
0.01 15.513 . . . 16.195 . . .
0.005 7.756 . . . 7.925 . . .
0.001 1.551 . . . 1.557 . . .
0.0001 0.155 . . . 0.155 . . .
δn approximate increment in C exact increment in C

P (1 + r)n ln(1 + r)δn P (1 + r)n+δn − P (1 + r)n
1 7.947 . . . 8.144 . . .
0.1 0.795 . . . 0.797 . . .
0.01 0.080 . . . 0.080 . . .
4.13. The iterations in Newton’s method are

f (xn )
xn+1 = xn − , n = 0, 1, . . . ,
f 0 (xn )
for a given initial value x0 . (a) Let f (x) = x4 + 2x2 − x − 1. Then
f 0 (x) = 4x3 + 4 − 1.
For example, if we start with x0 = 0.75, we obtain

f (0.75)
x1 = 0.75 − = 0.833 . . . .
f 0 (0.75)
y
5
2.5
x
-2 -1 1 2 3
-2.5
-5
-7.5
-10
-12.5
-15
42
y
10
7.5
2.5
x
-4 -3 -2 -1 1 2 3
-2.5
-5
-7.5
-10

y
10
7.5
2.5
x
-4 -3 -2 -1 1
-2.5
-5
-7.5
-10
Repeat the process starting with x1 to obtain x2 , and so on. The solution is x = 0.825 . . ..
1
(b) Let f (x) = x4 + x 3 − 1. Then
1 2
f 0 (x) = 4x3 + x− 3 .
3
The solution is x = 0.619 . . ..
(c) Let f (x) = x ln x + 0.3. Then
f 0 (x) = 1 + ln x.
(d) Let f (x) = ex − 4x3 . Then
f 0 (x) = ex − 12x2 .
(e) Let f (x) = tan x − 2x. Then
f 0 (x) = sec2 x − 2.
By Newton’s method the solution is x = 1.165 . . . 4
(f) Let f (x) = ex sin x/(1 + x). Then
f (x) (1 + x)[2(1 + x)e−x − sin x]

x− = x + .
f 0 (x) (1 + x) cos x + x sin x
Let x0 = 1.85. Then

x1 = 1.663, x2 = 1.689, x3 = 1.690
to three decimal places.
4.14. Since f (x) = xe−x + 1 then f 0 (x) = (1 − x)e−x . The function has its only stationary point
at x = 1 which is a maximum. The slope of y = f (x) in the neighbourhood of the solution of
f (x) = 0 is therefore positive whilst that for any value of x greater than 1 will be negative. By
the geometrical construction of Newton’s method illustrated in Figure 4.15, any tangent which
starts for x > 1 will produce iterations which diverge from the required solution. The graph of
y = xe−x − 1 is shown in the figure.
4.15. (a) The graph shows a continuous function in which f (a) and f (b) have opposite signs.
(b) Let g(x) = ex − 3x. The table gives a sequence of values for g(x) at intervals 0.25.
x 0 0.25 0.5 0.75 1.0 1.25 1.5
g(x) 1 0.534 0.149 -0.133 -0.282 -0.260 -0.018
43
y
x
-2 -1 1 2
-1
-2
-3
-4

y
a fHbL
x
b
fHaL
x 1.75 2.0 2.25 2.5

g(x) 0.505 1.389 2.738 4.682
Evidently the solutions of the equation lie between x = 0.5 and x = 0.75, and between x = 1.5 and
x = 1.75. Note also that the function has a minimum value at x = ln 3 = 1.098 . . ., which means,
for example, that any initial value for the smaller solution must start at a value of x < ln 3 for the
reasons outlined in Problem 4.14. Similar conditions apply to the other solution.
The solutions are x = 0.6190 . . ., and x = 1.5121 . . ..
4.16. (a) Calculate f (a + nE) for n = 1, 2, . . .. Stop the program at n = N , when f (a + N E) and
f (a + (N − 1)E) have different signs.
(b) In the following table the interval is bisected four times with E = 0.125 and N = 8.
x 0 0.125 0.250 0.375 0.500 0.625

f (a + N E) -1 -0.983 -0.929 -0.829 -0.676 -0.457
x 0.750 0.875 1
f (a + N E) -0.162 0.224 0.718
The solution of the equation lies between x = 0.75 and x = 0.875. The computed solution is
x = 0.806 . . ..
(c) Four decimal accuracy is obtained after 10 iterations using the bisection method, whilst New-
ton’s method achieve the same accuracy after just 4 iterations
4.17. The slope of the normal at x = x0 is −1/f 0 (x0 ) and at x = x0 + δx0 is −1/f 0 (x0 + δx0 ).
Hence their equations are
1
y − f (x0 ) = − 0 (x − x0 ),
f (x0 )
1
y − f (x0 + δx0 ) = − (x − x0 − δx0 ).
f 0 (x0 + δx0 )
Solving these equations for x and y:
f 0 (x0 )[f 0 (x0 + δx0 ){f (x0 + δx0 ) − f (x0 )} + δx0 ]

x = x0 − ,
f 0 (x0 + δx0 ) − f 0 (x0 )
44
[f 0 (x0 + δx0 ){f (x0 + δx0 ) − f (x0 )} + δx0 ]
y = f (x0 ) + ,
f 0 (x0 + δx0 ) − f 0 (x0 )
Divide the numerators and denominators by δx0 and let the increment tend to zero, so that the
centre of curvature (xc , yc ) is located at
µ ¶
f 0 (x0 )[1 + f 0 (x0 )2 ] [1 + f 0 (x0 )2 ]
x0 − , f (x0 ) + .
f 00 (x0 ) f 00 (x0 )
The radius of curvature
3
√ 2 2 [1 + f 0 (x0 )2 ] 2
R= [(xc − x0 ) + (yc − y0 ) )] = .
f 00 (x0 )
For the parabola y = x2 ,
f (x) = x2 , f 0 (x) = 2x, f 00 (x) = 2.
Hence the centre of curvature of the point (x0 , x20 ) is located at

1
[x0 − x0 (1 + 4x20 ), x20 + (1 + 4x20 )],
2
3
and its radius of curvature is R = 21 (1 + 4x20 ) 2 .
4.18. We shall prove Leibniz’s formula by induction. For n = 1, the formula is true since
(f g)(1) = f (1) g + f g (1)
by the product rule: note that 1 C1 = 1. Assume that the given formula is true for n = k and all
x. Then
(f g)(k) = f (k) g + k C1 f (k−1) g (1) + k C2 f (k−2) g (2) + · · · + k Ck f g (k) .
Differentiate both sides with respect to x:
(f g)(k+1) =
(f (k+1) g + f (k) g (1) ) + (k C1 f (k) g (1) + k C1 f (k−1) g (2) )
+(k C2 f (k−1) g (2) + k C2 f (k−2) g (2) ) + · · · + (k Ck f (1) g (k) + k Ck f g (k+1) )
= f (k+1) g + (1 + k C1 )f (k) g (1) + (k C1 + k C2 )f (k−1) g (2 + · · · + k Ck f g (k+1) .
The coefficients can be written as

k!
1 + k C1 = 1 + = k + 1 = k+1 C1 ,
1!(k − 1)!
k! k! k(k − 1) k(k + 1)
k−1 C1 + k C2 = + =k+ = = k+1 C2 ,
1!(k − 1)! 2!(k + 2)! 2! 2!
and, in general,
k! k!
k Cr + k Cr+1 = +
r!(k − r)! (r + 1)!(k − r − 1)!
· ¸
k! 1 1
= +
r!(k − r − 1)! k − r r + 1
k!(k + 1) (k + 1)!
= =
r!(k − r − 1)!(k − r)(r + 1) (r + 1)!(k − r)!
= k+1 Cr+1
Hence
(f g)(k+1) = f (k+1) g + k+1 C1 f (k) g (1) + k+1 C2 f (k−1) g (2) + · · · + k+1 Ck+1 f g (k+1) .
45
Hence if the result is true for n = k, then it is true for n = k + 1. We have shown that it is true
for n = 1 (the product rule); therefore it is true for n = 2, 3, . . ..
Chapter 5: Taylor series and approximations

1
5.1. (a) For f (x) = e 2 x ,
1 1x 1 1x 1 1x
f 0 (x) = e2 , f 00 (x) = e2 , e2 ,
2 4 8
so that
1 1 1
f (0) = 1, f 0 (0) =
, f 00 (0) = , f 00 (0) = .
2 4 8
The Taylor polynomial approximation to four terms becomes
1 11 2 11 3
f (x) ≈ 1 + x + x + x .
2 2! 4 3! 8
We can estimate that the three term approximation will be accurate to two decimal places if for
the fourth term ¯ ¯
¯ 1 3¯
¯ ¯
¯ 8.3! x ¯ < 0.005, or |x| < 0.621 . . . .
1
(b) For f (x) = (1 + x) 2 , the Taylor approximation is
1 1 1 1
(1 + x) 2 ≈ 1 + x − x2 + x3 .
2 8 16
The three-term approximation will be accurate to two decimal places if
¯ 3¯
¯x ¯
¯ ¯ < 0.005 or |x| < 0.432 . . . .
¯ 16 ¯
1
(c) For f (x) = (1 + x)− 3 , the four-term Taylor polynomial is
1 ¡ ¢ ¡ ¢¡ ¢³ 2´ ¡ ¢¡ ¢¡ ¢³ 3´
(1 + x)− 3 ≈ 1 + − 13 x + − 13 − 43 x2! + − 13 − 34 − 73 x3!
1 2 14
≈ 1 − x + x2 − x3
3 9 81
14 3
|x| < 0.005 or |x| < 0.306 . . .
81
.
(d) The Taylor approximation to four terms for sin 2x can be obtained form the series for sin y
where y = 2x (use (5.4c)):
1 1 1
sin 2x ≈ (2x) − (2x)3 + (2x)5 − (2x)7
3! 5! 7!
4 3 4 5 8 7
≈ 2x − x + x − x
3 15 315
8 7
|x| < 0.005 or |x| < 0.196 . . . .
15
(e) Using the expansion for cos z, where z = 12 x (see (5.4d)):

1 1 1 4 1
cos x ≈ 1 − x2 + x + x6 .
2 8 384 46080
46
The three-term polynomial will be accurate to two decimal places if
1 6
46080 x < 0.005 or |x| < 2.475 . . . .
(f) The four-term expansion is (see(5.4e)):
ln(1 + x) ≈ x − 12 x2 + 13 x3 − 14 x4 .

1 4
x < 0.005 or |x| < 0.376 . . . .
4
1
(g) Let f (x) = (1 + x2 ) 2 . Put u = x2 . Then, as in (b),
1 1 1 1 1
(1 + x2 ) 2 = (1 + u) 2 ≈ 1 + u − u2 + u3
2 8 16
1 1 1
≈ 1 + x2 − x4 + x6 .
2 8 16
1 6
x < 0.005 or |x| < 0.656 . . . .
16
(h) The four-term Taylor polynomial for ln(1 + 3x) is (put u = 3x, etc. ),
1 1 1
ln(1 + 3x) ≈ (3x) − (3x)2 + (3x)3 − (3x)4
2 3 4
9 2 81
≈ 3x − x + 9x3 − x4 .
2 4
81 4
x < 0.005 or |x| < 0.125 . . . .
4
5.2. The Taylor expansion for f (x) about x = 0 is

1 00 1
f (x) = f (0) + f 0 (0) + f (0)x2 + f 000 (0)x3 + · · · .
2! 3!
(a) Let f (x) = ex . Then f 0 (x) = f 00 (x) = · · · = ex . Hence
f (0) = f 0 (0) = f 00 (0) = · · · = 1.
(b) Let f (x) = sin x. Then
f 0 (x) = cos x, f 00 (x) = − sin x, f 000 (x) = − cos x, f (4) (x) = sin x, etc.
so that
f (0) = 0, f 0 (0) = 1, f 00 (0) = 0, f 000 (0) = −1, f (4) (0) = 0, . . . ,
the Taylor coefficients being
1 1
1, , , ... .
1! 2!
(c) Let f (x) = cos x. Then
f 0 (x) = − sin x, f 00 (x) = − cos x, f 000 (x) = sin x, f (4) (x) = cos x, . . . ,
47
so that
f (0) = 1, f 0 (0) = 0 f 00 (0) = −1 f 000 (0) = 0, f (4) (0) = 1, . . . .
(d) Let f (x) = (1 + x)α . Then
f 0 (x) = α(1 + x)α−1 , f 00 (x) = α(α − 1)(1 − x)α−2 , . . . .
so that
f (0) = 1, f 0 (0) = α, f 00 (0) = α(α − 1), . . . .
(e) Let f (x) = ln(1 + x). Then

1 1
f 0 (x) = , f 00 (x) = − ,
1+x (1 + x)2
2×1 3×2×1
f 000 (x) = , f (4) (x) = , ....
(1 + x)3 (1 + x)4 )
so that
f (0) = 0, f 0 (0) = 1, f 00 (0) = −1, f 000 (0) = 2!, f (4) (0) = 3!, . . . .
Therefore the coefficient of xn for n ≥ 1 is
(n − 1)! (−1)n−1
(−1)n−1 = .
n! n
5.3. (a) The general term for ex is xn /n!. Hence, for four-decimal point accuracy we require n
such that, for |x| = 2,
xn 2n
= < 0.00005.
n! n!
For n = 11, 2n /n! = 0.000051 . . . and for n = 12, 2n /n! = 0.0000085 . . . < 0.00005. Hence terms
up to x11 are required.
(b) The general term for sin x is (−1)n x2n−1 /(2n − 1)!. Hence for four-decimal accuracy we require
the smallest n such that, for |x| = 2,
¯ 2n−1 ¯
¯ x ¯ 2n−1
¯ ¯= 2
¯ (2n − 1)! ¯ (2n − 1)! < 0.00005.
For n = 6, 22n−1 /(2n − 1)! = 0.000051 . . . and for n = 7, 22n−1 /(2n − 1)! = 0.0000031 . . .. Hence
terms up to and including x11 are required.
(c) The general term for cos x is (−1)n x2n /(2n)!. Hence for four decimal accuracy we require the
smallest n such that, for |x| = 2,
¯ 2n ¯
¯ x ¯ 2n
¯ ¯= 2
¯ (2n)! ¯ (2n)! < 0.00005.
For n = 5, 22n /(2n)! = 0.00028 . . . and for n = 6, 22n /(2n)! = 0.0000085 . . .. Hence terms up to
and including x10 are required.
1
(d) For (1 + x) 2 the general term is
α(α − 1)(α − 2) . . . (α − n + 1) n
x ,
n!
where α = − 12 . For four-decimal accuracy we require the smallest n such that, for |x| = 0.5,
¯ ¯
¯ α(α − 1)(α − 2) . . . (α − n + 1) ¯
¯ (0.5) n¯
¯ n! ¯ < 0.00005.
48
For n = 8 the last term has magnitude 0.000051 . . ., and for n = 9 the magnitude is 0.000021 . . . .
(e) For ln(1 + x), the general term in its Taylor series is (−1)n+1 xn /n. For four-decimal accuracy
we require the smallest n such that, for |x| = 0.5,
¯ ¯
¯ n¯ n
¯(−1)n+1 x ¯ = 0.5 < 0.00005.
¯ n¯ n
n
0.5n
For n = 10, 0.5
n = 0.000097 . . ., whilst for n = 11, n = 0.000044 . . .. Hence terms up to and
including x10 are required.
5.4. (a) Let f (x) = arcsin x. Then
1 x 1 + 2x2
f 0 (x) = √ , f 00 (x) = 3 , f 000 (x) = 5
(1 − x2 ) (1 − x2 ) 2 (1 − x2 ) 2
so that f (0) = 0, f 0 (0) = 1, f 00 (0) = 0, f 000 (0) = 1. Hence

1
arcsin x = x + x3 + · · · .
6
(b) Let f (x) = arccos x. Then

1 x
f 0 (x) = − √ , f 00 (x) = − 3
(1 − x2 ) (1 − x2 ) 2
so that f (0) = 12 π, f 0 (0) = −1, f 00 (0) = 0. The Taylor series starts with
1
arccos x = π − x + ···.
2
(c) Let f (x) = arctan x. Then
1 2x −2 + 6x2
f 0 (x) = , f 00 (x) = − , f 000 (x) = .
1 + x2 (1 + x2 )2 (1 + x2 )3
Hence f (0) = 0, f 0 (0) = 1, f 00 (0) = 0, f 000 (0) = −2. The Taylor series starts with
1
arctan x = x − x3 + · · · .
3
(d) Let f (x) = e−x sin x. Then
f 0 (x) = e−x (cos x − sin x), f 00 (x) = −2e−x cos x.
Hence f (0) = 0, f 0 (0) = 1, f 00 (0) = −2. The Taylor series starts with
e−x sin x = x − x2 + · · · .
(e) Let f (x) = e−x cos x. Then
f 0 (x) = −e−x (cos x + sin x).
Hence f (0) = 1, f 0 (0) = −1 so that the Taylor series starts with
e−x cos x = 1 − x.
5.5. (a) Let f (x) = 1/(1 + 3x). Then f 0 (x) = −3/(1 + 3x)2 , f 00 (x) = 18/(1 + 3x)3 so that
f (0) = 1, f 00 (0) = −3, f 00 (0) = 18.
49
The first three terms of its Taylor series are
1
= 1 − 3x + 9x2 + · · · .
1 + 3x
Alternatively, the binomial expansion (5.4f) can be used. Also from (5.4f) the expansion will be
valid for
1 1
−1 < 3x < 1 or − < x < .
3 3
(b) Adapting (5.4f),
h
³ x í−1 · ³ x ´ (−1)(−2) ³ x ´2 ¸
1 −1 1
= 2 1+ − = 1 + (−1) − + − + ···
2−x 2 2 2 2! 2
1 1 1
= + x + x2 + · · ·
2 4 8
the series is valid for −1 < 12 x < 1 or −2 < x < 2.
(c) Using (5.4f) again
1 1 1 £ ¤
(3 − x) 3 = 3 3 (1 − 13 x) = 3 3 1 − 91 x − 1 2
81 x − ··· .
The series is valid for −3 < x < 3.

(d) Using (5.4f) again
1 1
h ³ x í 13
(x − 3) 3 = (−3) 3 1 + −
· 3 ¸
1 1 1
= (3) 3 −1 + x + x2 + · · · .
9 81

(e) Adapting (5.4e),
· µ ¶¸ µ ¶
1 1 1 1 2
ln(9 − x) = ln 9 1 − x = ln 9 + ln 1 − x = 2 ln 3 − x − x − ···.
9 9 9 162

(f) From (5.4d),
1 1 ³ x ´2 1 ³ x ´4
cos( x) = 1− + − ···
2 2! 2 4! 2
1 1 4
= 1 − x2 + x − ···.
8 384
The series is valid for all x.
1
(g) Put u = x 2 for x > 0, and use (5.4c):
1 1 1 5 1 1 3 1 5
sin(x 2 ) = sin u = u − u3 + u − · · · = x2 − x2 + x2 − · · · .
6 120 6 120
1
Since x 2 is not real for x < 0 the series will be valid only for x ≥ 0.
1
(h) Put u = x 2 for x > 0, and use (5.4d):
1
cos(x 2 ) = 1 − 12 x + 1 2
24 x + ···.
The series is valid for all x ≥ 0.
50
5.6. Multiply standard expansions for
1 1
e−x = 1 − x + x2 + · · · and = (1 + x)−1 = 1 − x + x2 + · · · .
2 1+x
Hence
µ ¶
e−x 1 2 ¡ ¢
= 1 − x + x + · · · 1 − x + x2 + · · · .
1+x 2
5
= 1 − 2x + x2 + · · ·
2
(b) As in (a)
µ ¶µ ¶
1 1 1 1
(1 − x) 2 ex = 1 − x − x2 − · · · 1 + x + x2 + · · ·
2 8 2
1 1 2
= 1 + x − x + ···
2 8
(c) This time square the series for ln(1 − x):

· ¸2
1 1 1 2 1 3
[ln(1 − x)]2 = −x − x − x − · · ·
x2 x2 2 3
· ¸
1 2 3 11 4
= x + x + x + ···
x2 12
11
= 1 + x + x2 + · · ·
12
5.7. Start with the Taylor series

1
1 + ln(1 + x) = 1 + x − x2 + · · · .
2
Then, assuming that the expansion takes the form b0 + b1 x + b2 x2 + · · ·,
1 1
= = b0 + b1 x + b2 x2 + · · · .
[1 + ln(1 + x)] 1 + x − 21 x2 + · · ·
We now equate coefficients of powers of x in the identity

µ ¶
1 ¡ ¢
1 = 1 + x − x2 + · · · b0 + b1 x + b2 x2 + · · · ,
2
1
= b0 + (b1 + b0 )x + (b2 − b1 − b0 )x2 + · · ·
2
Hence b0 = 1, b1 = −b0 = −1, b2 = b1 + 12 b0 = 3
2 and the Taylor series starts with
1 3
= 1 − x + x2 + · · · .
[1 + ln(1 + x)] 2
(b) Write tan x = sin x/ cos x and use the series for sin x and cos x. Thus
x − 16 x3 + 120
1
x5 + · · ·
tan x = 1 1 3
1 − 2x + 6x + · · ·
= b1 x + b3 x3 + b5 x5 + · · · .
51
Note that the series will contain only odd powers of x. By cross-multiplying
µ ¶
1 3 1 5 1 1 3 ¡ ¢
x− x + x + · · · = 1 − x + x + · · · b1 x + b3 x3 + b5 x5 + · · · .
6 120 2 6
By matching the coefficients of x, x2 , . . . on either side we obtain

1 2
b1 = 1, b3 = , b5 = ,
3 15
so that
1 2
tan x = x + x3 + x5 + · · · .
3 15
(c) From (5.4b)
1 1
1 + ex = 2 + x + x2 + x3 · · · .
2 6
Assume that
1
= b0 + b1 x + b2 x2 + · · · :
1 + ex
then
µ ¶
1 2 1 3 ¡ ¢
1 = 2 + x + x + x · · · b0 + b1 x + b2 x2 + b3 x3 · · ·
2 6
µ ¶
1
= 2b0 + (b0 + 2b1 ) x + b0 + b1 + 2b2 + · · ·
2
Solving for b0 , b1 , b2 and b3 ,
1 1 1 1
x
= − x + x3 + · · · .
1+e 2 4 48
(d) Use the definition
sinh x ex − e−x
tanh x = = x ,
cosh x e + ex
where
1 1 5 1 1
sinh x = x + x3 + x + · · · , cosh x = 1 + x2 + x4 + · · · .
6 120 2 24
Hence, if the required series is b1 x + b3 x3 + b5 x5 + · · · (it must be an odd function), then
µ ¶
1 1 5 1 1 ¡ ¢
x + x3 + x + ··· = 1 + x2 + x4 + · · · b1 x + b3 x3 + b5 x5 + · · ·
6 120 2 24
1 1 1
= x + (b1 x + (b3 + b1 )x3 + (b5 + b3 + )x5 + · · ·
2 2 24
Comparing powers of x, it follows that b1 = 1, b3 = − 13 , b5 = 2
15 so that
1 2
tanh x = x − x3 + x5 + · · · .
3 15
(e) Since x/ sin x is an even function and sin x = x − 16 x3 + 120

1
x5 , then
µ ¶
1 1 5 ¡ ¢
x = x − x3 + x b0 + b2 x2 + b4 x4 + · · ·
6 120
µ ¶ µ ¶
1 1 1
= b0 x + b2 − b0 + b4 − b2 + b0 x5 + · · ·
6 6 120
1 7
Hence b0 = 1, b2 = 6 and b4 = 360 so that
x 1 7 4
= 1 + x2 + x + ···.
sin x 6 360
52
5.8. The following series provide approximations for large values of x.
(a) Let u = 1/x. Then
µ ¶1
1 2 1 1 1
1− = (1 − u) 2 = 1 − u − u2 + · · · (binomial series)
x 2 8
1 1
= 1− − + ···
2x 8x2
which will be valid for |u| < 1, or equivalently, |x| > 1.
1
(b) Let x > 0 and u = 1/x 2 > 0. Then
µ ¶
1 1 1
ln 1 + 1 = ln(1 + u) = u − u2 + u3 + · · ·
x2 2 3
1 1 1
= 1 − + + ···.
x2 x 3x 32
This series will be valid for 0 < u ≤ 1, or x ≥ 1.
(c) Let x > 0 and u = 1/x. Then
1
x2 1
1 = 1 = (1 + u)−1
(1 + x) 2 (1 + x1 ) 2
1 3 2
= 1− u+ u + · · · (binomial expansion)
2 8
1 3
= 1− + + ···
2x 8x2
The series is valid for 0 < u < 1 or x > 1.
(d) Let u = (1/x) + (1/x)2 . Then, using (5.4e),
1
ln(1 + x + x2 ) ln(x2 ) + ln(1 + u) = ln(x2 ) + u − u2 + · · ·
=
2
µ ¶ µ ¶2
1 1 1 1 1
= ln(x2 ) + + − + + ···
x x2 2 x x2
1 1
= ln(x2 ) + + 2 + · · ·
x 2x
√ √
The series is valid for −1 < (1/x)+(1/x2 ) < 1, that is, when x < − 12 ( 5−1) or when x > 21 (1+ 5).
(e) Let u = 1/x. Then
1 1 1
= = 1 3 1 5
sin(1/x) sin u u − 3! u + 5! u + ···
1 1
= · 1 2 1 4
u 1 − 3! u + 5! u ···
" µ ¶ µ ¶2 #
1 1 2 1 4 1 2 1 4
= 1+ u + u + u + u + ···
u 6 120 6 120
· ¸
1 1 7 4
= 1 + u2 + u + ···
u 6 360
1 7
= x+ + + ···
6x 360x3
5.9. (a) Using the two-term Taylor expansion for sin x,

1 1 1 1
≈ ≈
sin x x − 16 x3 x 1 − 61 x2
53
µ ¶
1 1 2
≈ 1+ x (using (5.4f))
x 6
1 1
≈ + x,
x 6
for small x. Note that for x = 0.5, the error is
¯ µ ¶¯
¯ ¯
¯sin x − 1 + 1 x ¯ = 0.0024 . . . ,
¯ x 6 ¯
that is, about 0.1%.

(b) Write as
µ ¶1 µ ¶
1 1 1 2 1 1
(1 + x) 2 = x2 1+ ≈ x2 1 +
x 2x
1 1
= x2 + 1 ,
2x 2
for large x.
(c) Using the result from (b)
1 1 1 1 1 1
(2 + x) 2 − (1 + x) 2 = 2 2 (1 + x) 2 − (1 + x) 2
·³ ´2 1 ¸ · ¸
1 x 2 1 1 1
≈ 22 + 1/(2x) 2 − x 2 + 1
2 2x 2
1
≈ 1 ,
2x 2
for large x.
(d) Using the three-term Taylor expansion for cos x,
· ¸− 1
1 x2 x4 2
1 ≈ −
(1 − cos x) 2 2 24
1 · ¸
22 1 2
≈ 1 + x (binomial expansion)
x 24
1 1
22 22 x
≈ + ,
x 24
for small x.
5.10. (a) Expanding about x = 1,
1 1
ln x = ln[1 + (x − 1)] = (x − 1) − (x − 1)2 + (x − 1)3 + · · · (using (5.4e)).
2 3
The series is valid for −1 < x − 1 < 1 or 0 < x < 2.
(b) For an expansion about x = 21 π, write
1 1 1
cos x = cos[ π + (x − π)] = − sin(x − π).
2 2 2
Now use the Taylor expansion for the sine:
1 1 1 1 1 1
cos x = − sin(x − π) = −(x − π) + (x − π)3 − (x − π)5 + · · ·
2 2 3! 2 5! 2
The series is valid for all x − 12 π, which means for all x.
54
(c) For a Taylor series centred at x = 1, write
1 1 1 1 1
(1 + x) 2 = [2 + (x − 1)] 2 = 2 2 [1 + (x − 1)] 2 .
2
Now expand the right-hand side using the binomial series (5.4f):
1 √ 1 1
(1 + x) 2 = 2 + √ (x − 1) − √ (x − 1)2 + · · · .
2 2 16 2
The series is valid for −1 < 12 (x − 1) < 1, that is for −1 < x < 3.
5.11. (a) Since f (x) has a stationary point at x = c, f 0 (c) = 0 and its Taylor series about x = c
will be
1 1
f (x) = f (c) + f 00 (c)x2 + f 000 (c)x3 + · · · .
2! 3!
Approximately
1
f (x) ≈ f (c) + f 00 (c)x2 ,
2
00
for |x − c| small. Hence if f (c) > 0, then f (x) > f (c) close to x = c excluding x = c. The
conclusion is that x = c is a minimum. Similarly if f 00 (c) < 0, then x = c is a maximum.
(b) If f 00 (c) = 0, then we must look at the signs of higher derivatives. Suppose that f (N ) (c) 6= 0 is
the first non-zero derivative (that is, f (r) (c) = 0 for r = 1, 2, . . . , N − 1). Hence, approximately,
1 N
f (x) ≈ f (c) + f (c)(x − c)N .
N!
If N is even and f (N ) (c) > 0 then x = c is a minimum, whilst if f (N ) (c) < 0 then x = c is a
maximum. If N is odd then the stationary point will be a point of inflection.
5.12. (Compare eqn (2.15).) Put (ex − 1)/x = f (x) for x 6= 0, and use the Taylor series (5.4b) for
ex :
1 1
f (x) = [(1 + x + x2 + · · ·) − 1]/x = 1 + x + · · · ,
2! 2!
for x 6= 0. Therefore limx→0 f (x) = 1, which is the ‘missing value’ at x = 0.
(b) Put (1 − cos x)/x2 = f (x) for x 6= 0. From (5.4d)
1 1 1
f (x) = [1 − (1 − + x4 − · · ·)]
x2 2! 4!
1 1 2 1
= ( x − x4 + · · ·)
x2 2! 4!
1 1
= − x2 + · · · (for x 6= 0)
2! 4!
Therefore limx→0 f (x) = 21 .
(c) Put [ln(1 + x) − x]/ sin x = f (x), x 6= 0. From (5.4c,e),
(x − 21 x2 + · · ·) − x − 12 x2 + · · ·
f (x) = 1 3 = 1 3
x − 3! x + ··· x − 3! x + ···
x(− 12 + · · ·
= 1 3 (for x 6= 0).
1 − 3! x + ···
Therefore limx→0 f (x) = 0.

(Alternatively, rewrite f (x) in the form
ln(1 + x) − x x
x sin x
55
and use the limits (2.13) and (2.14).)
(d) Put sin x/(1 − cos x) = f (x), x 6= 0:
1 1 2
x − 3! + ··· x(1 − 3! x + · · ·)
f (x) = 1 2 = 1 2 1 2
1 − (1 − 2! x + · · ·) 2 x (1 − 4! x + · · ·)
1 2(1 − · · ·)
= .
x (1 − · · ·)
This does not tend to a limit; it approaches ∞ as x → 0. Therefore this function does not possess
a fill-in value at x = 0 which would make it continuous.
5.13. (a)
(1 − x)12 − 1 −12x + higher powers 12
lim = lim = ,
x→0 (1 − x)10 − 1 x→0 −10x + higher powers 10
where the binomial theorem (4.7) was used to expand the powers of (1 − x).
(b)
1 3
sin x − x x − 3! x + · · ·) − x
lim = lim 1 3 1 2
x→0 sin x − x cos x (x − 3! x+ · · ·) − x(1 − 2! x + · · ·)
x→0
1
− 3! + higher powers
= lim 1
3 + higher powers
x→0
1
− 3! + higher powers
= lim
1
x→0
3 + higher powers
1
= − .
2
(c) Put x = π + u. Then
cos x + 1 cos(π + u) + 1
lim = lim
x→π x−π u→0 u
cos π cos u − sin π sin u + 1
= lim (from (1.17a))
u→0 u
1 2
− cos u + 1 −(1 − 2! u + · · ·) + 1
= lim = lim
u→0 u u→0 u
1
= lim u( + · · ·) = 0.
u→0 2
(d) Put x = u + 12 π: then
sin x − 1 sin( 12 π + u) − 1
lim1 = lim
x→ 2 π cos 5x u→0 cos( 5 π + 5u)
2
sin 12 π cos u + cos 12 π sin u − 1
= lim (from (1.17a))
u→0 cos 5 π cos 5u − sin 5 π sin 5u
2 2
1 2
cos u − 1 u − ···
= lim = lim 2
u→0 − sin 5u u→0 −5u + · · ·
= 0
5.14. Let
ex − 1 (ex − 1)/x
f (x) = = .
ex − 1 − x [({ex − 1)/x} − 1]
Since limx→0 [(ex − 1)/x] = 1 (see eqn (2.15) or Problem 5.12a), f (x) approaches infinity as x → 0
56
To determine the question of signs, return to the original form and take the following steps:
(i) ex − 1 is negative when x < 0 and positive when x > 0.
(ii) (d/dx)(ex − 1 − x) = ex , which is greater than zero for all x. Therefore ex − 1 − x is a steadily
increasing function for all x. Since also it is zero at x = 0, it must be negative when x < 0 and
positive for x > 0.
(iii) From (i) and (ii), f (x) is negative when x < 0 and positive when x > 0. Therefore f (x) → −∞
as x → 0 from the left, and f (x) → +∞ as x → 0 from the right.
5.15. (a)
· ¸
sin3 3x (3x + · · ·)3
lim = lim
x→0 1 − cos x x→0 1 − (1 − 1 x2 + · · ·)
2!
27x3 + · · ·
= lim
x→0 1 x2 + · · ·
2
27x(1 + · · ·)
= lim = 0.
x→0 1 (1 + · · ·)
2
1
(b) Let [(ex − 1)/x] 2 = g(x). We know from eqn (2.15) that
· x ¸
2 e −1
lim [g(x)] = lim = 1.
x→0 x→0 x
But also
lim [g(x)]2 = lim [g(x)g(x)] = lim g(x) lim g(x) = [ lim g(x)]2 .
x→0 x→0 x→0 x→0 x→0
Therefore √
lim g(x) = 1 = 1,
x→0
the positive square root being taken because g(x) is never negative.
(c)
· ¸ · ¸ · ¸
(2 + tan x) sin x 2 + tan x sin x
lim = lim lim
x→0 x(3 − tan2 x) x→0 3 − tan2 x x→0 x
2+0 2
= · 1 = (where we refer to eqn (2.13))
3−0 3
5.16. Let f (x) = 3x − sin x and g(x) = x. Then

¯
f (x) f 0 (x) 3 − cos x ¯¯
lim = lim 0 = ¯ = 2.
x→0 g(x) x→0 g (x) 1 x=0
5.17. In the following, S represents the required sum.

(a)
X∞ ∞
1 n X
S = x − (−x)n
n=0
n! n=0
1
= ex − (1 − x + x2 − · · ·) = ex − ,
1+x
the second term being a geometric series with common ratio (−x): see (5.4a).
(b) S = x3 + 21 x4 + 13 x5 + · · · = x2 (x + 12 x2 + 13 x3 + · · ·). From (5.4e),
1 1
ln(1 − x) = −x − x2 − x3 − · · · .
2 3
Therefore S = −x2 ln(1 − x).
57
1 2 1 4
(c) S = 1 + 2! x + 4! x + · · ·. But from (5.4b),
1 2 1 1 1
ex = 1 + x + x + x3 + · · · and e−x = 1 − x + x2 − x3 + · · · .
2! 3! 2! 3!
Therefore ex + e−x = 2S, and
1 x
S= (e + e−x ) = cosh x
2
from (1.26).
5.18. In the following, S represents the required sum.
(a) From (5.4b)
2 3
e2 = 1 + 2 + 22! + 23! + · · ·,
and
22 23
e−2 = 1 − 2 + 2! − 3! + · · ·,
so that ³ ´
23 25
e2 − e−2 = 2 2 + 3! + 5! + ··· ,
or
S = 12 (e2 − e−2 ).
1
(b) From (5.4b), S = e 2 .
(c) S is geometric with common ratio (− 14 ). Therefore, by (5.4a),
1 4
S= 1 = .
1+ 4
5
Chapter 6: Complex numbers
6.1. (a) x = −1 ± i; (b) x = 3 ± i; (c) x = i or −3i.

6.2. The equation is a quadratic equation in x2 . Hence x2 = −4 or 1. Taking square roots
x = ±1 or ± 2i.
6.3. The standard form of a complex number is a + ib, where a and b are real numbers. Thus the
answers are (a) 4 + 3i; (b) 3 − 5i; (c) −11 + 15i; (d) 9 + 3i; (e) 21 + 12 i; (f) 1 + 6i; (g) −3 − 4i: (h)
− 78 96
25 − 25 i; (i) −4 − 4gm.
6.4. The boundary between real and complex roots in the (p, q) plane is the parabola p2 = 4q:
the roots are real if p2 ≥ 4q and complex if p2 < 4q. The roots are both real and negative in the
quadrant p > 0, q < 0.
1
6.5. (a) 4 + i; (b) 5 + 5i; (c) 5 − 75 i; (d) − 13
25 −
9
25 i.
6.6. (a) −4i; (b) −7 + 4i; (c) − 15 + 85 i; (d) − 15 − 85 i.

6.7. (a) 1 − i; (b) 2i; (c) −2i; (d) 12 (1 + gm); (e) i.
6.8. Numerically to 3 decimal places the answers are: (a) 4.482 + 2.218i; (b) 16.233 − 0.167i; (c)
−1.248 + 2.728i; (d) 88.669; (e) 266.050 + 0.512i.
√
6.9. (a) |z1 | = 2 2, Arg z1 = 34 π; (b) |z2 | = 8, Arg z2 = − 13 π; (c) |z3 | = 5, Arg z3 = − 21 π; (d)
|z4 | = 3, Arg z4 = π; (e) |z5 | = 5, Arg z5 = arctan( 34 ).
6.10. The curves are: (a) the circle x2 + y 2 = 1; (b) the straight line y = 2; (c) The circle
(x − a1 )2 + (y − a2 )2 = 1 where a1 = Re(a) and a2 = Im(a); (d) the parabola y 2 = 4x; (e) the
ellipse 3x2 + 4y 2 = 12 (need to square twice to remove the square roots); the complex formula
58
expresses the well-known property of ellipses that the sum of the distances from any point on an
ellipse to the foci is a constant; (f) the straight line y = x for x ≥ 0; (g) the archimedean spiral
r = θ.
√
6.11. (a) 2 exp( 34 iπ); √ (b) 2 exp(iπ); (c) 3 exp(− 21 iπ);√ √
1
(d) 14√
exp(− 3 iπ); (e) 2 2 exp(iθ) where θ = arctan[( 3 − 1)/( 3 + 1)];
2
(f) 1+√ exp(− 41 iπ); (g) e2 exp(i);
√ 3
(h) √ 2 exp(iθ) where θ = arctan[(cos 2 + sin 2)/(cos 2 − sin 2)]; (i) 512 exp(iπ);
(j) 2 exp( 43 iπ).
6.12. Use the identity
ei(θ1 +θ2 ) = eiθ1 eiθ2 .
Hence
cos(θ1 + θ2 ) + i sin(θ1 + θ2 )
= (cos θ1 + i sin θ1 )(cos θ2 + i sin θ2 )
= cos θ1 cos θ2 − sin θ1 sin θ2 + i(sin θ1 cos θ2 + cos θ1 sin θ2 ).
Equating real and imaginary parts it follows that
cos(θ1 + θ2 ) = cos θ1 cos θ2 − sin θ1 sin θ2 ,
and
sin(θ1 + θ2 ) = sin θ1 cos θ2 + cos θ1 sin θ2 .
For the other identities use
ei(θ1 −θ2 ) = eiθ1 e−iθ2 .
6.13.
-
z1 -z2 z1 +z2
-
-z2
z2
-
z1 z1 +z2
-
z1 -z2 z1
-
-z2 z2
- -
z1 +z2
Figure 58: Problem: 6.13
6.14 For the general case with f (θ) = a cos θ + b sin θ,
f 0 (θ) = −a sin θ + b cos θ,
and
f 00 (θ) = −a cos θ − b sin θ = −f (θ).
The first case can be obtained by putting a = 1 and b = i.
6.15. Using exponential forms for cos and sin, it follows that
sin ia 2[exp(ai2 ) − exp(−ai2 )]
tan a = =
cos ia 2i[exp(ai2 ) + exp(−ai2 )]
1 exp(−a) − exp(a) exp(a) − exp(−a)
= · =i
i exp(−a) + exp(a) exp(a) + exp(−a)
= i tanh a.
59
6.16. (a) The equation cosh z = 1 implies
1 z
(e + e−z ) = 1 ⇒ e2z − 2ez + 1 = 0 ⇒ (ez − 1)2 = 0.
2
Hence ez = 1. If z = a + bi (a, b real), then ea eib = 1 = e2nπi , (n = 0, ±1, ±2, . . .). Thus a = 0 and
b = 2nπ. The roots are given by z = 2nπi, (n = 0, ±1, ±2, . . .).
(b) Similarly sinh z = 1 implies
e2z − 2ez − 1 = 0.
√
Hence ez = 1 ± 2. If z = a + bi, then
√
ea+bi = ea (cos b + i sin b) = 1 ± 2.
It follows that sin b = 0 so that b = nπ, n = 0, ±1, ±2, . . .). Hence

√
ea cos b = ea cos(nπ) = (−1)n ea = 1 ± 2,
so that √ √
a = ln[ 2 − 1], (n odd) a = ln[ 2 + 1] (n even) .
The complex roots are
√ √
z = ln[ 2 − 1] + inπ, (n odd) z = ln[ 2 + 1] + inπ (n even).
(c) ez = −1 = e2n+1 πi , (n = . . . − 2, −1, 0, 1, 2, . . .). It follows that the roots are
z = (2n + 1)πi (n = . . . − 2, −1, 0, 1, 2, . . .).

√
(d) cos z = 2 implies that
1 iz √ √
(e + e−iz ) = 2 ⇒ e2iz − 2 2eiz + 1 = 0.
2
Hence √ √
eiz = 2 ± 1 = ( 2 ± 1)e2nπi , (n = 0, ±1, ±2, . . .).
If z = a + ib, then √
e−b = ( 2 ± 1), and a = 2nπ.
Hence the roots are √
z = 2nπ − i ln[ 2 ± 1].
√ √ √
6.17. (a) Log(1 + i 3) = ln( (1 + 3)) + iArg (1 + i 3) = ln 2 + 13 iπ.
(b) We can write log z = Log|z| + i(Arg z + 2kπ) where k is an integer. Hence if log z = πi, then
|z| = 1 and Argz + 2kπ = π so that k = 0 and z = πi is the only solution.
(c) Log(ei) = ln(e) + 12 πi = 1 + 12 πi.
(d) elog z = eln r+iθ+2kπi = eln r eiθ = reiθ = z. Therefore log z defines the set of functions inverse to
ez , as is suggested by the notation.
6.18. (a) 2i = ei ln 2 = cos(ln 2) + i sin(ln 2).
(b)
1 1
ii = ei ln i = exp[i ln(e 2 πi )] = e− 2 π .
This number is real: hence Arg(ii ) = 0.
(c) The equation becomes
z i = ei log z ei[Log|z|+i(Arg z+2kπ0] = e−(Arg z+2kπ) eLog|z| = −1 = e(2n+1)πi ,
where n and k are integers. Hence
Log |z| = (2n + 1)π, Arg z = 0, k = 0.
60
Therefore, the solutions are given by z = e(2n+1)π , where n is any integer.
6.19. Write the equation as
z 5 = −1 = e(2n+1)πi , n, any integer.
The solutions are given by

· ¸
1
z = exp (2n + 1)πi , (n = 1, 2, 3, 4, 5).
5
Other values of n merely repeat these solutions. On the Argand diagram the solutions all lie on
the unit circle centred at the origin at polar angles 15 π, 35 π, π, 75 π, 95 π.
y
1
0.5
x
-1 -0.5 0.5 1
-0.5
-1
Figure 59: Problem: 6.19
6.20. Denote the complex number by z in each case.

(a) z = 2e3+2i = 2e3 [cos 2 + i sin 2]. Hence
|z| = 2e3 , Arg z = 2, Re z = 2e3 cos 2, Im z = 2e3 sin 2.
(b) z = 4ei = 4[cos 1 + i sin 1]. Hence
|z| = 4, Arg z = 1, Re z = 4 cos 1, Im z = 4 sin 1.

√ √ √
(c) z = 5 exp[cos( 14 π) + i sin( 14 π)] = 5 exp(1/ 2)[cos(1/ 2) + i sin(1/ 2)]. Hence
√ √
|z| = 5 exp(1/ 2), Arg z = 1/ 2,
√ √ √ √
Re z = 5 exp(1/ 2) cos(1/ 2, Im z = 5 exp(1/ 2 sin(1/ 2).
(d) z = e1+i = e(cos 1 + i sin 1). Hence
|z| = e, Arg z = 1, Re z = e cos 1, Im z = e sin 1.
6.21. Let z = ceα+iβ = ceα [cos β + i sin β]. Comparing with
x = 0.04e−0.01t sin 12t,
we can identify
1
c = 0.04, α = −0.01t, β = 12t + π.
2
6.22. We have to express the sine as a cosine. Hence
1
i(t) = ce−0.05t sin(0.4t + 0.5) = ce−0.05t cos(0.4t + 0.5 − π),
2
1
= Re[ce−0.05t ei(0.4t+0.5− 2 π) ],
1
= Re[ce−0.05t+i(0.4t+0.5− 2 π) ]
61
6.23. (a) z 2 = (x + iy)2 = (x2 − y 2 ) + i2xy. Hence
Re(z 2 ) = x2 − y 2 , Im(z 2 ) = 2xy.
(b) First
z 3 = (x + iy)(x2 − y 2 + i2xy) = (x3 − 3xy 2 ) + i(3x2 y − y 3 ).
Thus
z + 2z 2 + 3z 3 = (x + 2x2 − 2y 2 + 3x3 − 9xy 2 ) + i(y + 4xy + 9x2 y − 3y 3 ).
Hence
Re(z + 2z 2 + 3z 3 ) = x + 2x2 − 2y 2 + 3x3 − 9xy 2 ,
Im(z + 2z 2 + 3z 3 ) = y + 4xy + 9x2 y − 3y 3 .
(c)
1 i(x+iy)
sin z = sin(x + iy) = [e − e−i(x+iy) ]
2i
1 −y ix
= [e e − ey e−ix ]
2i
1 −y
= [e (cos x + i sin x) − ey (cos x − i sin x)]
2i
= e−y sin x.
Hence
Re(sin z) = e−y sin x, Im(sin z) = 0.
(d)
1 i(x+iy) 1
cos z = [e + e−i(x+iy) ] = [eix e−y + e−ix ey ]
2 2
= cos x cosh y − i sin x sinh y.
Hence
Re(cos z) = cos x cosh y, Im(cos z) = − sin x sinh y.
(e) Using (d)
1 x+iy
ez cos z = e [cos x cosh y − i sin x sinh y]
2
1 x
= e (cos y + i sin y)[cos x cosh y − i sin x sinh y]
2
1 x
= e [(cos y cos x cosh y + sin y sin x sinh y) +
2
(sin y cos x cosh y − cos y sin x sinh y)].
(f) exp(z 2 ) = exp(x2 − y 2 ) exp(2xyi) = exp(x2 − y 2 )[cos 2xy + i sin 2xy]. Hence
Re[exp(z 2 )] = exp(x2 − y 2 cos 2xy, Im[exp(z 2 )] = exp(x2 − y 2 sin 2xy.
6.24. w = u + iv = f (z) = z 2 = (x + iy)2 = x2 − y 2 + 2xyi. Hence, equating real and imaginary

parts
u = x2 − y 2 , v = 2xy.
The hyperbolas map into the straight lines u = 1 and v = 2 respectively in the w plane.
6.25. Substituting for z it follows that
c c(x − iy)
w = z + = x + iy + 2
z x + y2
µ ¶ µ ¶
cx cy
= x+ 2 +i y− 2 .
x + y2 x + y2
62
For the circle |z| = 1, x2 + y 2 = 1, so that
w = x(1 + c) + iy((1 − c).
Hence u = x(1 + c) and v = y(1 − c). Thus, on the circle |z| = 1
u2 v2
x2 + y 2 = 1 = + ,
(1 + c)2 (1 − c)2
which is the equation of an ellipse.
6.26. The derivation of the formula for cos6 θ is given in Example 6.20. For sin6 θ use the identity
µ ¶
1 1
sin nθ = zn − n ,
2i z
where z = cos θ + i sin θ. Then
µ ¶6
6 1
(2 sin θ) = − z−
z
µ ¶ µ ¶ µ ¶
6 1 4 1 2 1
= − z + 6 + 6 z + 4 − 15 z + 2 + 20
z z z
= 2 (− cos 6θ + 6 cos 4θ − 15 cos 2θ + 20) .
Finally
sin6 θ = 1
32 (− cos 6θ + 6 cos 4θ − 15 cos 2θ + 20).
6.27. The displacement is given by

x = Re z = e−0.2t cos 0.5t.
Hence x = 0 where cos 0.5t = 0. The required zeros of x are given by
1
0.5t = (2n + 1)π, for integer n .
2
Hence t = (2n + 1)π, (n = 0, ±1, ±2, . . .).
The velocity is given by
dx d £ −0.2t ¤
= e cos 0.5t = −e−0.2t [0.2 cos 0.5t + 0.5 sin 0.5t],
dt dx
or, alternatively, by
· ¸
dz d (−0.2+0.5i)t
Re = Re e
dt dt
£ ¤
= Re (−0.2 + 0.5i)e−0.2t (cos 0.5t + i sin 0.5t)
= −e−0.2t [0.2 cos 0.5t + 0.5 sin 0.5t].
6.28. If z = 2 + i is a solution then so is its conjugate 2 − i since the coefficients of the polynomial
are real. Therefore (z − 2 − i)(z − 2 + i) = z 2 − 4z + 5 is a factor. Hence
z 4 − 2z 3 − z 2 + 2z + 10 = (z 2 − 4z + 5)(z 2 + 2z + 2).
The other solutions are given by z 2 + 2z + 2 = 0, that is, z = −1 ± i.
6.29. (a)
1 1
S = 1 − sin θ + sin 2θ − sin θ + · · ·
"∞ 2! # 3! " ∞ #
X (−1)n eniθ X zn
= Im = Im ,
n=0
n! n=0
n!
63
where z = −eiθ . The infinite series is the Taylor series for the exponential function ez (see Section
5.4). Hence
S = 1 + Im[ez ] = 1 + Im[exp(−eiθ )]
= 1 + Im[exp(− cos θ − i sin θ)]
= 1 − e− cos θ sin(sin θ).
(b) In this case
22 23
T = 1 + 2 cos θ + cos 2θ + cos 3θ + · · ·
2! " 3! #
X∞ X∞
2n 2n niθ
= cos nθ = Re e
n=0
n! n=0
n!
Using the Taylor series for an exponential function;

X∞ X∞
2n niθ (2eiθ )n
e = = exp[2eiθ ] = exp[2 cos θ + 2i sin θ]
n=0
n! n=0
n!
= e2 cos θ [cos(2 sin θ) + i sin(2 sin θ)]
Hence
T = e2 cos θ cos(2 sin θ).
64
PART IV: Transforms and Fourier Series
Chapter 24: The Laplace transform, 1
Chapter 25: Laplace and z transforms: applications, 14
Chapter 26: Fourier series, 30
Chapter 27: Fourier transforms, 47
Chapter 24: The Laplace transform
24.1. The Laplace transform of the function f (t) is defined by

Z ∞
L{f (t)} = F (s) = e−st f (t)dt.
0
Use (24.2), (24.3) and (24.4).

Z ∞
1
(a) L{et } = et e−st dt = .
0 s−1
4
(b) L{4e−t } = .
1+s
3 1
(c) L{3et − e−t } = L{3et } − L{e−t } = − .
s−1 s+1
6 1
(d) L{3t2 − 1} = − .
s3 s
(e) L{ 12 t3 + 2t2 − 3} = 3
s4 + 4
s3 − 3s .
3 48
(f) L{3 + 2t4 } = + 5.
s s
3 s 3−s
(g) L{3 sin t − cos t} = L{3 sin t} − L{cos t} = − = .
1 + s2 1 + s2 1 + s2
2s 2
(h) L{2(cos t − sin t)} = L{2 cos t} − L{2 sin t} = − .
1 + s2 1 + s2
½ ¾
t t2 tn
(i) L 1 + + + ··· + = s + s2 + s3 + · · · + sn+1 .
1! 2! n!
24.2. The scale rule (24.5) states that, if L{f (t)} = F (s) and k > 0, then
1 ³s´
L{f (kt)} = F .
k k
(a) Since L{et } = 1/(s − 1), then
1 1 1
L{e3t } = = .
3 ( 3s − 1) s−3
1
1 2 2−s
(b) L{1 − 2e−2t } = − = .
s s+2 s(s + 2)
ω
(c) L{sin ωt} = .
s2 + ω 2
s
(d) L{cos ωt} = .
s2 + ω 2
3s 4 3s − 4
(e) L{3 cos 2t − 2 sin 2t} = 2
− 2
= .
4+s 4+s 4 + s2
(f) Use the identity cos2 t = 12 (1 + cos 2t). Then
2 + s2
L{cos2 t} = L{ 12 (1 + cos 2t)} = .
s(4 + s2 )
(g) Use the identity sin2 t = 12 (1 − cos 2t), so that
2
L{sin2 t} = .
s(4 + s2 )
24.3. To evaluate these Laplace transforms use the scale, shift and other rules in Section 24.3.
(a) By (24.2), L{t2 } = 2!/s3 . Then by the shift rule (24.7)
2!
L{t2 et } = .
(s − 1)3
(b) In a way similar to (a), L{t} = 1/s2 , so that, by the shift rule
1
L{te−2t } = .
(s + 2)2
(c) Using the shift rule again

2!
L{t2 e−t } = .
(s + 1)2
(d) Since, by (24.4), L{cos t} = s/(s2 + 1), the shift rule (24.7) implies
s−2 s−2
L{e2t cos t} = 2
= 2 .
(s − 2) + 1 s − 4s + 5
(e) Since, by (24.4), L{sin t} = 1/(s2 + 1)}, the shift rule (24.7) implies
1 1
L{e−t sin t} = = 2 .
(s + 1)2 + 1 s + 2s + 2
(f) Using the scale rule (24.6) and the shift rule (24.7)
3
L{et sin 3t} = .
s2 − 2s + 10
(g) Using the scale rule (24.6) and the shift rule (24.7)
3
L{e−2t sin 3t} = .
s2 + 4s + 13
2
(h) Using the scale rule (24.6) and the shift rule (24.7) again
3+s
L{e−3t cos 2t} = .
s2 + 6s + 13
(i) Let F (s) = L{cos 3t} = s/(s2 + 9). Using the multiplication rule
µ ¶
d s s2 − 9
L{t cos 3t} = − = .
ds s2 + 9 (s2 + 9)2
(j) Let F (s) = L{sin 3t} = 3/(s2 + 9). Now use the multiplication rule (24.8):
µ ¶
dF (s) d 3 6s
L{t sin 3t} = − =− 2
= 2 .
ds ds s + 9 (s + 9)2
1
(k) Let F (s) = L{sin t} = s2 +1 . By the multiplication rule (24.8):
µ ¶
2 d2 F (s) d2 1
L{t sin t} = = 2
ds2 ds s2 + 1
µ ¶
d 2s
= −
ds (s2 + 1)2
6s2 − 2
= .
(s2 + 1)3
(l) Let f (t) = t4 e−t .

(i) Since L{t4 } = 4!/s5 , the shift rule (24.7) implies
4! 24
L{f (t)} = = .
(s + 1)5 (s + 1)5
(ii) Since L{e−t } = 1/(s + 1) = F (s), say, the multiplication rule (24.8) gives
4
· ¸
4 −t 4 d F (s) d4 1
L{t e } = (−1) = 4
ds4 ds (s + 1)
24
=
(s + 1)5
(iii) The direct method requires repeated integration by parts:

Z ∞ Z ∞
L{f (t)} = t4 e−t dt = [−t4 e−t ]∞0 + 4t3 e−t dt
0 0
Z ∞
= 0 + [−4t3 e−t ]∞ 0 + 12t2 e−t dt
0
Z ∞
24
= 12 t2 e−t dt = .
0 (s + 1)5
24.4. From the definition (24.1):

Z ∞
s
e−st cos ktdt = .
0 s2 + k2
Differentiate both sides of this equation with respect to k. Then

Z ∞
2ks
− te−st sin ktdt = − 2 ,
0 (s + k 2 )2
3
so that
2ks
L{t sin kt} = .
(s2 + k 2 )2
24.5. In these problems we have to find the function f (t) in the transform
Z ∞
F (s) = e−st f (t)dt,
0
when F (s) is given. We attempt this by using the tables in Section 24.4 and in Appendix F, and
the various rules. Note that all the time functions are defined to be zero for t < 0.
(a) F (s) = 1/s2 is the transform of f (t) = t by (24.2). We also write this as
1
↔ t (see Section 24.4).
s2
1
(b) ↔ 1 (see 24.2).
s
3 3
(c) ↔ .
2s 2
3 t4
(d) ↔ .
s5 8
(e) Use the shift rule (24.7) L{ekt f (t)} = F (s − k). Since L{1} = 1/s,
1
↔ e3t .
s−3
(f) Using the shift rule again
1
↔ e−4t .
s+4
3 3 1
(g) ↔ e 2 t.
2s − 1 2
2 2 2
(h) ↔ − e 3 t.
2 − 3s 3
(i) Using partial fractions
1 1 1
= − ↔ et − 1.
s(s − 1) s−1 s
√ √
(j) Since s2 + s − 1 = (s + 12 − 12 5)(s + 12 + 12 5), the partial fractions expansion is
· ¸
1 1 1 1
= √ √ − √
s2 + s − 1 5 s + 12 − 12 5 s + 21 + 21 5
1 h 1 √ 1 √ i
↔ √ e− 2 (1− 5)t − e− 2 (1+ 5t)
5
(k) Using partial fractions

s 1 1 1
= + ↔ [e−t + et ].
s2 − 1 2(s + 1) 2(s − 1) 2
4
(l) Using partial fractions
2s − 1 1 3 1 3
2
= + ↔ et + e−t .
s −1 2(s − 1) 2(s + 1) 2 2
(m) By (24.4)
s
↔ cos t.
s2 + 1
(n) By (24.6c),
1 1
↔ sin 2t.
s2 + 4 2
(o) By (24.6),
2s − 1 2s 1 1
2
= 2 − 2 ↔ 2 cos 2t − sin 2t.
s +4 s +4 s +4 2
(p) Using partial fractions
2s − 1 1 1
= + ↔ 1 + et .
s(s − 1) s s−1
(q) Using partial fractions
s2 − 1 1 1 2
= + −
s(s − 1)(s + 2)(s + 3) 6s 2(s + 2) 3(s + 3)
1 1 −2t 2 −3t
↔ + e − e
6 2 3
(r) Using partial fractions and (24.6)
s 1 s 1
= − +
(s − 1)(s2 + 1) 2(s − 1) 2(1 + s ) 2(1 + s2 )
2
1 t 1 1
↔ e − cos t + sin t
2 2 2
(s) Since L{t2 } = 2/s3 , the shift rule (24.7) applied to this transform implies
1 1
↔ et t2 .
(s − 1)3 2
(t) Since s2 − 2s + 2 = (s − 1)2 + 1, we can write the transform as
2s + 1 2(s − 1) + 3
= .
s2 − 2s + 1 (s − 1)2 + 1
Now use the shift rule (24.7) and (24.6b,c):

2s + 1 1
↔ et [cos t + 6 sin t].
s2 − 2s + 1 2
(u) Using partial fractions
s s s 1
= − ↔ [cos t − cos 2t].
(s2 + 1)(s2 + 4) 3(s2 + 1) 3(s2 + 4) 3
24.6. These problems use the results on the transforms of derivatives given in (24.12):
L{ẋ(t)} = sX(s) − x(0), L{ẍ(t)} = s2 X(s) − sx(0) − ẋ(0).
(a) L{ẋ(t)} = sX(s) − x(0) = sX(s) − 6.
5
(b) L{ẋ(t)} = sX(s) − x(0) = sX(s).
(c) L{ẍ(t)} = s2 X(s) − sx(0) − ẋ(0) = s2 X(s) − 3s − 5.
(d) L{ẍ(t)} = s2 X(s) − sx(0) − ẋ(0) = s2 X(s).
(e)
L{2ẍ(t) + 3ẋ(t) − 2x(t)} = 2s2 X(s) − 2sx(0) − 2ẋ(0) + 3sX(s) − 3x(0) − 2X(s)
= (2s2 + 3s − 2)X(s) − (3 + 2s)x(0) − 2ẋ(0)
= (2s2 + 3s − 2)X(s) − 10s − 9
1
(f) L{3ẍ(t) − 5ẋ(t) + x(t) − 1} = (3s2 − 5s + 1)X(s) − .
s
24.7. (a) Take the Laplace transform of the differential equation:
L{ẍ + 3ẋ + 2x} = s2 X(s) − sx(0) − ẋ(0) + 3sX(s) − 3x(0) + 2X(s)

= (s2 + 3s + 2)X(s) − 1 = 0.
Therefore
1 1 1
X(s) = = − .
(s + 1)(s + 2) s+1 s+2
Taking the inverse transform x(t) = e−t − e−2t .
(b) Take the Laplace transform of the equation:
L{ẍ + ẋ − 2x} = s2 X(s) − sx(0) − ẋ(0) + sX(s) − x(0) − 2X(s)

= (s2 + s − 2)X(s) − 3(s + 1) = 0.
Hence the transform of the solution is

3(s + 1) 3s 2 1
X(s) = 2
= = + .
s +s−2 (s − 1)(s + 2) s−1 s+2
Hence the inverse transform is x(t) = 2et + e−2t .

(c) Take the Laplace transform of the equation:
L{ẍ + 4ẋ} = s2 X(s) − sx0 − y0 + 4sX(s) − 4x0

= (s2 + 4s)X(s) − (s + 4)x0 − y0 = 0
Therefore the transform of the solution is

(s + 4)x0 + y0 4x0 + y0 y0
X(s) = = − .
s(s + 4) 4s 4(s + 4)
Hence the solution is
x = 41 (4x0 + y0 ) − 14 y0 e−4t .
(d) Take the Laplace transform of the equation:
L{ẍ + ω 2 x} = s2 X(s) − sx(0) − ẋ(0) + ω 2 X(s) = (s2 + ω 2 )X(s) − sc = 0.
Therefore the transform of the solution is

sc
X(s) =
s2 + ω 2
6
giving the solution x = c cos ωt.
(e) Taking the Laplace transform of the equation:
L{ẍ + 2ẋ + 5x} = s2 X(s) − 3s + 3 + 2sX(s) − 6 + 5X(s)
= (s2 + 2s + 5)X(s) − 3s − 3 = 0.
Therefore
3s + 3 3(s + 1)
X(s) = = .
s2 + 2s + 5 (s + 1)2 + 4
Using the shift rule
x = 3e−t cos 2t.
(f) For this fourth-order equation
s4 Y (s) − s3 y(0) − s2 y 0 (0) − sy 00 (0) − y 000 (0) − Y (s) = (s4 − 1)Y (s) − s3 = 0.
Therefore
s3 1 1 s
Y (s) = 4
= + + 2
.
s −1 4(s − 1) 4(s + 1) 2(s + 1)
The solution can now be constructed using (24.6):
y = 14 ex + 14 e−x + 1
2 cos x.
24.8. Take the Laplace transform of the equation in each case.

(a) ẍ = 1 + t + et , x(0) = ẋ(0) = 0. Then
1 1 1
s2 X(s) = + + .
s s2 s−1
Hence
1 1 1 1 1 1 1 1
X(s) = 3
+ 4+ 2 = 3+ 4− − 2+ .
s s s (s − 1) s s s s s−1
Hence, using the table of Laplace transforms, the solution is
1 1
x = −1 − t + t2 + t3 + et .
2 6
(b) ẍ + x = 3, x(0) = 0, ẋ(0) = 1. The Laplace transform of this equation is
3
s2 X(s) − 1 + X(s) = .
s
Hence
1 3 1 3 3s
X(s) = + = 2 + − .
s2 + 1 s(s2 + 1) s + 1 s s2 + 1
Using (24.6) for the inversion:
x = sin t + 3 − 3 cos t.
(c) ẍ + 2ẋ + 2x = 3, x(0) = 1, ẋ(0) = 0. The Laplace transform of the system is
3
s2 X(s) − s + 2sX(s) − 2 + 2X(s) = .
s
Hence
s+2 3
X(s) = +
(s + 1)2 + 1 s[(s + 1)2 + 1]
s+2 3 3(s + 2)
= + −
(s + 1)2 + 1 2s 2[(s + 1)2 + 1]
s+1 1 3
= − 2
− 2
+ .
2[(s + 1) + 1] 2[(s + 1) + 1] 2s
7
Hence using the inverse table and the shift rule
1 3
x = − e−t (cos t + sin t) + .
2 2
(d) ẍ − x = e2t , x(0) = 0, ẋ(0) = 1. The Laplace transform of the system is
1
ss X(s) − 1 − X(s) = .
s−2
Hence
s−1 1 1 1
X(s) = = = − ,
(s − 2)(s2 − 1) (s − 2)(s + 1) 3(s − 2) 3(s + 1)
using partial fractions. By inverting, we have
1
x= (−e−t + e2t ).
3
(e) ẍ − x = tet , x(0) = 1, ẋ(0) = 1. The Laplace transform of the system is
1
s2 X(s) − s − 1 − X(s) = .
(s − 1)2
Hence
1 1 9 1 1 1
X(s) = + = − − + .
s − 1 (s − 1)2 (s2 − 1) 8(s − 1) 8(s + 1) 4(s − 1)2 2(s − 1)3
The solution is
x = 89 et − 18 e−t − 14 tet + 14 t2 et .
(f) ẍ − 4x = 1 − e2t , x(0) = 1, ẋ(0) = −1. The Laplace transform of the system is
1 1
s2 X(s) − s + 1 − 4X(s) = − .
s s−2
Hence
1 7 1 13
X(s) = − 2
+ − + .
4(s − 2) 16(s − 2) 4s 16(s + 2)
Inverting
1 7 1 13
x = − te2t + e2t − + e−2t .
4 16 4 16
2t −2t
(g) ẍ − 4x = e + e , x(0) = 0, ẋ(0) = 0. The Laplace transform of the system is
1 1
s2 X(s) − 4X(s) = + .
s−2 s+2
Hence
1 1 1 1
X(s) = + = − .
(s2 − 4)(s − 2) (s2 − 4)(s + 2) 4(s − 2)2 4(s + 2)2
Using the shift rule
t 2t
x= (e − e−2t ).
4
(h) ẍ + ω 2 x = C cos ωt, x(0) = x0 , ẋ(0) = y0 . The Laplace transform of the equation is
Cs
s2 X(s) − sx0 − y0 + ω 2 X(s) = .
s2 + ω2
Hence
x0 s + y0 Cs
X(s) = + 2 .
(s + ω ) (s + ω 2 )2
2 2
8
Note that
ω
L{sin ωt} = = F (s),
s2 + ω 2
say, so that by (24.8)
· ¸
dX(s) d ω 2sω
− =− 2 2
= 2 = L{t sin ωt}.
ds ds s + ω (s + ω 2 )2
Use this result in the inversion the transform above:

y0 Ct
x = x0 cos ωt + sin ωt + sin ωt.
ω 2ω
···
(i) x − 2ẍ − ẋ + 2x = e−2t , x(0) = ẋ(0) = 0, ẍ(0) = 2. The Laplace transform of the equation is
1
s3 X(s) − 2 − 2s2 X(s) − sX(s) + 2X(s) = ,
s+2
or
1
(s3 − 2s2 − s + 2)X(s) = (s − 1)(s + 1)(s − 2)X(s) = 2 + .
s+2
Hence, applying partial fractions,
2 1
X(s) = +
(s − 1)(s + 1)(s − 2) (s − 1)(s + 1)(s − 2)(s + 2)
3 7 1 1
= − − − .
4(s − 2) 6(s − 1) 2(s + 1) 12(s + 2)
Inverting
3 2t 7 t 1 −t 1
x= e − e − e − e−2t .
4 6 2 12
24.9. (a) ẋ = x − y, ẏ = x + y, x(0) = 1, y(0) = 0. Let L{x} = X(s) and L{Y } = Y (s). Taking
Laplace transforms of the differential equations:
sX(s) − x(0) = sX(s) − 1 = X(s) + Y (s), sY (s) − y(0) = sY (s) = X(s) + Y (s),
which are simultaneous linear equations in X(s) and Y (s), namely
(s − 1)X(s) + Y (s) = 1, X(s) + (s − 1)Y (s) = 0.
Solving them
s−1 s−1
X(s) = , Y (s) = .
(s − 1)2 + 1 (s − 1)2 + 1
Finally invert the transforms using the shift rule (24.7) applied to (24.4) to give the solutions
x = et cos t, y = et sin t.
(b) ẋ = 2x + 4y + e4t , ẏ = x + 2y, x(0) = 1, y(0) = 0. Taking Laplace transforms of the differential
equations:
1
sX(s) − 1 = 2X(s) + 4Y (s) + , sY (s) = X(s) + 2Y (s),
s−4
or
1
(s − 2)X(s) − 4Y (s) = 1 + , X(s) + (s − 2)Y (s) = 0.
s−4
The solutions are
s−2 s−2 1 1
X(s) = + , Y (s) = + .
s(s − 4) s(s − 4)2 s(s − 4) s(s − 4)2
9
Using partial fractions
3 5 1 3 3 1
X(s) = + + , Y (s) = − + + .
8s 8(s − 4) 2(s − 4)2 16s 16(s − 4) 4(s − 4)2
Finally, using (24.6a) and the rules (24.7) and (24.8):
3 5 4t 1 4t 3 3 1
x= + e + te , y=− + e4t + te4t .
8 8 2 16 16 4
(c) ẋ = x − 4y, ẏ = x + 2y, x(0) = 2, y(0) = 1. Taking Laplace transforms of the differential
equations:
sX(s) − 2 = X(s) − 4Y (s), sY (s) − 1 = X(s) + 2Y (s),
or
(s − 1)X(s) + 4Y (s) = 2, X(s) − (s − 2)Y (s) = −1.
Solving for X(s) and
2s − 8 2s − 8 s+1 s+1
X(s) = = , Y (s) = = .
s2 − 3s + 6 (s − 32 )2 + 15
4
s4 − 3s2 + 6 (s − 32 )2 + 15
4
These transforms can be inverted using the shift rule and (24.6b,c):
" Ã√ ! Ã √ !#
2 3t 15t √ 15t
x = − e 2 −3 cos + 15 sin ,
3 2 2
" Ã√ ! Ã √ !#
1 3t 15t √ 15t
y = e 2 3 cos + 15 sin .
3 2 2
24.10. (a) ẍ + x = et , x(0) = A, ẋ(0) = B. Take the Laplace transform of the equation:
1
s2 X(s) − As − B = .
s−1
Hence
As + B 1 (A − 12 )s + (B − 12 ) 1
X(s) = + = + .
s2 + 1 (s − 1)(s2 + 1) s2 + 1 2(s − 1)
Inverting using (24.6)
x = (A − − 12 ) cos t + (B − 21 ) sin t + 21 et .
(b) ẍ − x = 3, x(0) = A, ẋ(0) = B. The Laplace transform of the equation is:
3
s2 X(s) − As − B − X(s) = .
s
Hence
As + B 3 3+A+B 3 3+A−B
X(s) = + = − + .
s2 − 1 s(s2 − 1) 2(s − 1) s 2(s + 1)
using partial fractions. Inversion gives the solution
x = −3 + 12 (3 + A − B)e−t + 21 (3 + A + B)et .
(c) ẍ − 2ẋ + x = et , x(0) = A, ẋ(0) = B. The Laplace transform of the equation is:
1
s2 X(s) − As − B − 2sX(s) − A + X(s) = .
s−1
Hence
As + B − 2A 1 A B−A 1
X(s) = 2
+ 3
= + 2
+ .
(s − 1) (s − 1) s − 1 (s − 1) (s − 1)3
10
Inversion gives the solution
x = Aet + (B − A)tet + 12 t2 et .
24.11. d4 y/dx4 − y = ex , y(0) = A, y 0 (0) = B, y 00 (0) = C, y 000 (0) = D. The Laplace transform of
the equation is (with Ly(x) = Y (s))
1
s4 Y (s) − As3 − Bs2 − Cs − D − Y (s) = .
s−1
Therefore, after a lengthy partial fraction expansion,
As3 + Bs2 + Cs + D 1
Y (s) = +
(s4 − 1) (s − 1)(s4 − 1)
1 −3 + 2A + 2B + 2C + 2D 1 + 2A − 2B + 2C − 2D
= 2
+ +
4(s − 1) 8(s − 1) 8(s + 1)
1 + 2B − 2D + (1 + 2A − 2C)s
+ .
4(s2 + 1)
The inversion gives

1 x 1 1
y = xe + (2A + 2B + 2C + 2D − 3)ex + (1 + 2A − 2B + 2C − 2D)e−x
4 8 8
1 1
+ (1 + 2A − 2C) cos x + (1 + 2B − 2D) sin x.
4 4
24.12. The equation for x0 (t) has a different form from the subsequent equations; its transform
using x0 = 1 is
sX0 (s) − 1 + βX0 (s) = 0,
so that
1
X0 (s) = , and x0 (t) = e−βt . (i)
s+β
For all r ≥ 1, the form of the equation is
xr + βxr = βxr−1 ,
with xr (0) = 0. Therefore the transform is
sXr (s) + βXr (s) = βXr−1 (s),
so that
β
Xr (s) = Xr−1 (s). (ii)
s+β
Starting with the case r = 1, and using (i), we obtain the sequence
β β β β2
X1 (s) = , X2 (s) = = ,
(s + β)2 s + β (s + β)2 (s + β)3
β β2 β3
X3 (s) = = ,...,
s + β (s + β)3 (s + β)4
and, in general, for r ≥ 1,
βr
Xr (s) = . (iii)
(s + β)r+1
From (24.2),
1 1
↔ .
sr+1 r!
11
From the shift rule (24.7), with k = −β, we obtain the inverse of (iii),
β r tr −βt
xr (t) = e , r ≥ 1.
r!
Together with (i), this provides the required solution.
24.13. By the multiplication rule (24.8),
µ ¶
d s s2 − 1
t cos tH(t) ↔ − = .
ds s2 + 1 (s2 + 1)2
By the delay rule (24.15), with c = 2,
e−2s (s2 − 1)
(t − 2) cos(t − 2)H(t − 2) ↔ .
(s2 + 1)2
By the shift rule (24.7), with k = −1,
e−2(s+1) [(s + 1)2 − 1] e−2(s+1) s(s + 2)

e−t (t − 2) cos(t − 2)H(t − 2) ↔ = .
[(s + 1)2 + 1] (s2 + 2s + 2)2
24.14. (a) G(s) = e−2s /(s + 3). By (24.6a),

1
↔ e−3t .
s+3
Then by (24.15)
e−2s
= e−3(t−2) H(t − 2).
s+3
(b) G(s) = (1 − se−s )/(s2 + 1). G(s) is the sum of two transforms. For the first
1
↔ sin t.
s2 +1
For the second start with
s
↔ cos t.
s2 +1
Using (24.15),
se−2s
↔ cos(t − 1)H(t − 1).
s2 + 1
Finally, taking the difference
G(s) = (1 − se−s )/(s2 + 1) ↔ sin t − cos(t − 1)H(t − 1).
(c) e−2s /(s − 4) ↔ e4t−8 H(t − 2).
(d) G(s) = se−s /[(s + 1)(s + 2)]. The e−s term indicates that the delay rule will apply. Using
partial fractions
s 1 2
=− + ↔ −e−t + 2e−2t .
(s + 1)(s + 2) s+1 s+2
The delay rule (24.15) gives the required function:
(−e1−t + 2e2−2t )H(t − 1).
e−s
(e) ↔ et−1 [cos(t + 1)].
(s − 1)(s2 − 2s + 2)
12
24.15. All problems have the same initial conditions x(0) = ẋ(0) = 0.
(a) ẍ + x = f (t), where ½
1 for 0 < t ≤ 1,
f (t) = H(1 − t) = .
0 for t > 1.
Take the Laplace transform of the equation with the initial conditions x(0) = ẋ(0) = 0:
Z ∞ Z 1
1 − e−s
s2 X(s) + X(s) = e−st f (t)dt = e−st dt = .
0 0 s
Hence · ¸
1 − e−s −s 1 s
X(s) = = (1 − e ) − .
s(s2 + 1) s s2 + 1
Inverting using the delay rule:
x = 1 − cos t − [1 − cos(1 − t)]H(t − 1).
(b) ẍ − 4x = f (t), where

½
1 for 0 < t ≤ 1,
f (t) = = H(1 − t).
0 for t > 1.
Taking Laplace transforms:

1 − e−s
s2 X(s) − 4X(s) = ,
s
as in (a). Hence, using partial fractions,
· ¸ · ¸
1 − e−s 1 1 1 −s 1 1 1
X(s) = = + − − e + − .
s(s2 − 4) 8(s − 2) 8(s + 2) 4s 8(s − 2) 8(s + 2) 4s
Inverting · ¸
1 2t 1 −2t 1 1 1 1
x= e + e − − e2t−2 + e−(2t−2) + H(t − 1).
8 8 4 8 8 4
(c) ẍ − 4x = f (t), where 
 t for 0 < t ≤ 1,
f (t) = 2−t for 1 < t ≤ 2,

0 for t > 2.
In terms of step functions
f (t) = (2 − t)H(2 − t) − (2 − 2t)H(1 − t).
Taking Laplace transforms

Z 2 Z 1
2 st
s X(s) − 4X(s) = L{f (t)} = (2 − t)e dt − (2 − 2t)e−st dt
0 0
· −2s ¸ · −s ¸
e 1 − 2s 2e 2(1 − s)
= − − −
s2 s2 s2 s2
−s −2s
1 − 2e + e
=
s2
Hence
· ¸
1 − 2e−s + e−2s −s −2s 1 1 1
X(s) = = (1 − 2e + e ) − + − .
s2 (s2 − 4) 4s2 16(s − 2) 16(s + 2)
13
Inverting this transform using the delay rule (24.15),
· ¸
1 1 t 1 1 1 t
x = − e−2t + e2t − + − e4−2t + + e2t−4 − H(t − 2)
16 16 4 16 2 16 4
· ¸
1 1 1 t
+ e2−2t − − e2t−2 + H(t − 1)
8 2 8 2
(d) ẍ + x = f (t), where ½

cos t for 0 < t ≤ π,
f (t) =
0 for t > π.
so that f (t) = cos t[H(t) − H(t − π)]. Take the Laplace transform of the equation:
(1 + e−πs )s
s2 X(s) + X(s) = .
s2 + 1
Hence
(1 + e−πs )s
X(s) = .
s2 + 1)2
The inversion gives the solution
1 1
x = − t sin t − (t − π)H(t − π) sin t.
2 2
Chapter 25: Laplace and z transforms: applications

Rt
25.1. Use the division rule (25.1) which states that, if G(s) ↔ g(t), then G(s)/s ↔ 0
g(τ )dτ .
Use also Table (24.10) of inverse transforms
(a) Since
1
↔ sin t,
s2 + 1
then Z t
1
↔ sin τ dτ = [− cos τ ]t0 = 1 − cos t.
s(s2 + 1) 0
(b) The division rule is applied to the result from (a) as follows:
Z t
1
↔ τ (1 − cos τ )dτ = [τ − sin τ ]t0 = t − sin t.
s2 (s2 + 1) 0
(c) Apply the division rule again to the result in (b):

Z τ
1
↔ (τ − sin τ )dτ = [ 12 τ 2 + cos τ ]t0 = 12 t2 + cos t − 1
s3 (s2 + 1) 0
25.2. The RLC circuit has the equation

Z t
di 1
L + Ri + i(τ )dτ = v(t).
dt C 0
Take the Laplace transform of the equation and use the division rule (25.1):
1
L[sI(s) − I(0)] + RI(s) + I(s) = V (s),
Cs
where i(t) ↔ I(s) and v(t) ↔ V (s). Solving for I(s), we obtain
Cs[V (s) + Li(0)]
I(s) = ,
CLs2 + CRs + 1
14
1
since I(0) = 0. Given the data L = 2, R = 3, C = 3 and v(t) = 3 cos t ↔ 3s/(s2 + 1):
3s2 3(3s − 1) 9(2s − 1)

I(s) = = −
(s2 + 1)(2s2 + 3s + 3) 10(s2 + 1) 10(2s2 + 3s + 3)
3(3s − 1) 9(2s − 1)
= −
10(s2 + 1) 20[(s + 34 )2 + 15
16 ]
Using table (24.10) and the shift rule (24.7), the inverse is
√ Ã√ ! Ã√ !
9 3 3 15 − 3 t 15 9 −3t 15
i(t) = cos t − sin t + e 4 sin t − e 4 cos t .
10 10 10 4 10 4
(b) If v(t) = 0 and the capacitor has initial charge q0 , the equation for the current is (see Sec-
tion 25.1) µZ t ¶
di 1
L + Ri + i(τ )dτ + q0 = 0.
dt C 0
Taking the Laplace transform of the equation, noting that i(0) = 0,

1 q0
LsI(s) + RI(s) + I(s) + = 0.
Cs Cs
Hence
q0 3q0 3q0
I(s) = − =− 2 =− 15 .
CLs2 + CRs + 1 2s + 3s + 3 2[(s + 34 )2 + 16 ]
The inverse of this transform is
r Ã√ !
3 −3t 15
i(t) = −2 q0 e 4 sin t .
5 4
(c) We can represent the applied voltage by v(t) = 300 × 0.01δ(t − t0 ) = 3δ(t − t0 ). Hence the
circuit equation for the current is
Z
di 1 t
L + Ri + i(τ )dτ = 3δ(t − t0 ).
dt C 0
Hence
3se−st0 3se−st0
I(s) = = 15 .
2s2 + 3s + 3 2[(s + 34 )2 + 16 ]
Using the shift rule (24.7) and the delay rule (24.15), the inverse of the Laplace transform I(s) is
" Ã√ ! √ Ã√ !#
− 34 (t−t0 ) 3 15 3 15 15
i(t) = e − cos (t − t0 ) + sin (t − t0 ) H(t − t0 ).
2 4 10 4
25.3. The equation of motion is

ẍ + 2k ẋ + ω 2 x = f (t).
The impulse can be represented by a delta function so that f (t) = Iδ(t − t0 ). Take the Laplace
transform of the equation, noting that x(0) = 1 and ẋ(0) = 1,
s2 X(s) − s − 1 + 2ksX(s) − 2k + ω 2 X(s) = L{Iδ(t − t0 )} = Ie−t0 s .
Hence
s + 1 + 2k + Ie−t0 s s + 1 + 2k + Ie−t0 s
X(s) = 2 2
= , β 2 = k2 − ω2 .
s + 2ks + ω (s + k)2 − β 2
15
The term containing e−t0 s arising from the impulse will lead to the term with a step function on
inversion (see the delay rule (24.15)). The full solution is
1
x = [(−k − 1 + β)e−t(k+β) + (k + 1 + β)e−t(k−β) ]
2β
I
+ [e−(t−t0 )(k−β) − e−(t−t0 )(k+β) ]H(t − t0 ).
2β
for t > 0 and t0 > 0.
25.4. The displacement u(x) of the plank satisfies
d4 u
K = f (x).
dx4
The mountaineer standing at the centre of the plank is treated as a point load which can be
represented by the delta function M gδ(x − 12 l), so that
d4 u 1
K 4
= M gδ(x − l).
dx 2
Let A = u0 (0) and B = u000 (0) since only u(0) = 0 and u00 (0) = 0 are given: the constants A and
B will be found from the conditions at x = l when we have solved the equation. The Laplace
transform of the equation is
1
Ks4 U (s) − As2 − B = L{M gδ(x − 12 l)} = M ge− 2 ls .
Therefore " #
1
1 A B e− 2 ls
U (s) = + 4 + Mg 4 .
K s2 s s
Inversion of this transform using table (24.10) and delay rule (24.15) gives
1 £ ¤
u(x) = Ax + 61 Bx3 + 16 M g(x − 21 l)3 H(x − 21 l) .
K
The conditions at x = l are u(l) = 0 and u00 (l) = 0. Hence
1 £ ¤ 1 £ ¤
u(l) = Al + 16 Bl3 + 16 M g(l − 12 l)3 H( 12 l) = Al + 16 Bl3 + 1
48 M gl
3
= 0,
K K
and
1 d2 £ ¤ 1 £ ¤
u00 (l) = Ax + 16 Bx3 + 16 M g(x − 12 l)3 x=l = Bl + 12 M gl = 0.
K dx2 K
Solving these equations
B = − 12 M g, A = 16 1
M gl2 .
25.5. Use the impedance rules listed in (25.8) and (25.9).

(a) For the resistor r and inductor in parallel the impedance Z1 is given by
1 1 1
= + .
Z1 R Ls
Therefore
RLs
Z1 = .
R + Ls
The impedance Z1 is in series with the capacitor C. If Z is the impedance of the whole circuit,
then
1 RLs 1
Z = Z1 + = + .
Cs R + lsL Cs
16
Applying the given data
6s 2
Z= + .
2 + 3s s
(b) These components are all in parallel. Hence the impedance Z is given by
1 1 1 1 Ls + R + CRLs2
= + + = .
Z R Ls 1/(Cs) RLs
Hence
RLs 2s
Z= = 2
CRLs2 + Ls + R 6s + s + 1
for the given data.
(c) For the parallel resistor R = 1 and the inductor L = 1 the impedance Z1 is given by
1 1 s
= 1 + so that Z1 = .
Z1 s s+1
The impedance is in series with the resistor R = 2 which has the impedance
s 3s + 2
Z2 = 2 + Z1 = 2 + = .
s+1 s+1
Finally Z2 is in parallel with the capacitor C = 2 giving the impedance
1 1 1 6s2 + 5s + 1
= + = .
Z Z2 2/s 2(3s + 2)
Therefore
3s + 2
Z= .
6s2 + 5s + 1
25.6. (a) Let I1 (s) be the current in the s domain through the capacitor C = 2, I2 (s) be the
current through the resistor R = 3 and the inductor L = 1 and I3 (s) through the resistor R = 5 .
Then by Kirchhoff’s laws
I(s) − I1 (s) − I2 (s) = 0, I1 (s) + I2 (s) − I3 (s) = 0,
and
V1 (s) = I2 (s)(3 + s) + 5I3 (s),
I1 (s) I1 (s)
− I2 (s)(s + 3) = 0, V2 (s) = .
2s 2s
It follows from the first two equation that I2 (s) = I(s) − I1 (s) and that (not surprisingly) that
I3 (s) = I1 (s) + I2 (s) = I(s). Hence
V1 (s) = (I(s) − I1 (s))(s + 3) + 5I(s),
and
I1 (s)
− (I(s) − I1 (s))(s + 3) = 0.
2s
Eliminate I(s) between these equations so that
V1 (s) = 2I1 (s)(2s2 + 5s − 2)).
Therefore
V2 (s) 1
= .
V1 (s) 2s(2s2 + 5s − 2)
and
V2 (s) 1
= 2
.
I(s) 2s(2s + 6s + 1)
17
(b) Let I1 (s) be the current in the s domain through the inductor L = 2. Apply Kirchhoff’s law
in the s domain to the three subcircuits in Figure 25.22(b). Then
V1 (s) = 2(I1 (s) + I(s)) + sI(s) = 2I1 (s) + 5I(s), (i)

µ ¶
1
2s + I1 (s) − 3I(s) = 0, (ii)
2s
I1 (s)
V2 (s) = . (iii)
2s
From (ii) I(s) = I1 (s)(4s2 + 1)/(6s). Using this equation eliminate I(s) in (i) so that
(20s2 + 12s + 5)I1 (s)

V1 = .
6s
Finally combining this equation with (iii):
V2 (s) 3s
= .
V1 (s) 20s2 + 12s + 5
and
V2 (s) I1 (s) 6s 3
= 2
= 2 .
I(s) 2s I1 (s)(4s + 1) 4s + 1
25.7. The convolution theorem (25.11) states that if

Z t Z t
f (t) = g(τ )h(t − τ )dτ = h(τ )g(t − τ )dτ, G(s) ↔ g(t), H(s) ↔ h(t),
0 0
then
F (s) = L{f (t)} = G(s)H(s).
The convolution integrals are either integrated directly or by using the convolution theorem.
(a) g(t) = et , h(t) = 1. By direct integration
Z t Z t
f (t) = g(τ )h(t − τ )dτ = eτ dτ = [eτ ]t0 = et − 1.
0 0
Alternatively, using the convolution theorem and g(t) = et ↔ 1/(s − 1) = G(s), h(t) = 1 ↔ 1/s =
H(s), we obtain
1 1 1
F (s) = G(s)H(s) = = − ↔ et − 1 = f (t).
s(s − 1) s−1 s
(b) g(t) = 1, h(t) = 1. By direct integration

Z t Z t
f (t) = g(τ )h(t − τ )dτ = dτ = t.
0 0
(c) g(t) = et , h(t) = et . By direct integration

Z t Z t Z t
τ t−τ t
f (t) = g(τ )h(t − τ )dτ = e e dτ = e dτ = tet .
0 0 0
(d) g(t) = e−t , h(t) = t. Since e−t ↔ 1/(s + 1) and t ↔ 1/s2 , the convolution theorem gives
1 1 1 1
F (s) = G(s)H(s) = =− + 2 + ↔ −1 + t + e−t = f (t).
s2 (s + 1) s s s+1
18
(e) g(t) = t, h(t) = sin t. Using the tables of transforms, t ↔ 1/s2 and sin t ↔ 1/(s2 + 1), the
convolution of g(t) and h(t) has the transform
1 1 1
F (s) = G(s)H(s) = = − 2 ↔ t − sin t = f (t).
s2 (s2 + 1) s 2 s +1
(f) g(t) = cos t, h(t) = t. Using tables of transforms, t ↔ 1/s2 and cos t ↔ s/(s2 + 1), the
convolution of the transform is
1 1 s
F (s) = G(s)H(s) = = − 2 ↔ 1 − cos t = f (t).
s(s2 + 1) s s +1
(g) g(t) = sin 3t, h(t) = e−2t . The transforms of g(t) and h(t) are sin 3t ↔ 3/(s2 + 9) and
e−2t ↔ 1/(s + 2). Hence by the convolution theorem
3
F (s) = G(s)H(s) =
(s + 2)(s2 + 9)
3 6 3s
= + −
13(s + 2) 13(s2 + 9) 13(s2 + 9)
3 −2t 2 3
↔ e + sin 3t − cos 3t = f (t)
13 13 13
(h) g(t) = h(t) = sin t. Since sin t ↔ 1/(s2 + 1), the convolution theorem gives
· ¸
1 1 s2 − 1 1
F (s) = G(s)H(s) = 2 = − 2 + 2
(s + 1)2 2 (s + 1)2 s +1
1 1
↔ − t cos t + sin t = f (t)
2 2
by (24.9).
(i) g(t) = t4 , h(t) = sin t. Since t4 ↔ 24/s5 and sin t ↔ 1/(s2 + 1), the convolution theorem gives
· ¸
24 1 1 1 s
F (s) = G(s)H(s) = 5 2 = 24 5 − 3 + − 2
s (s + 1) s s s s +1
↔ t4 − 12t2 + 24 − 24 cos t = f (t)
(j) g(t) = tn , h(t) = tm . Since tn ↔ n!/sn+1 and tm ↔ m!/sm+1 (assuming that n and m are
positive integers), the convolution theorem gives
n!m! n!m!tn+m+1
F (s) = G(s)H(s) = ↔ = f (t).
sn+m+2 (n + m + 1)!
25.8. (a) The Laplace transform of
d2 x
+ ω 2 x = f (t)
dt2
is
s2 X(s) − sx(0) − x0 (0) + ω 2 X(s) = F (s),
where X(s) = L{x(t)} and F (s) = L{f (t)}. Choose a particular solution such that x(0) = 0 and
x0 (0) = 0. Then
F (s)
X(s) = 2 .
s + ω2
From (24.6)
1
↔ sin t.
s2 + ω 2
19
Hence by the convolution theorem, a particular solution is
Z t
x(t) = f (τ ) sin(t − τ )dτ.
0
(b) The Laplace transform of

d2 x
− ω 2 x = f (t)
dt2
is.
s2 X(s) − sx(0) − x0 (0) − ω 2 X(s) = F (s).
Choose a particular solution such that x(0) = 0 and x0 (0) = 0. Then
· ¸
F (s) 1 1 1 1 ωt
X(s) = 2 = − ↔ [e − e−ωt ].
s − ω2 2ω s − ω s + ω 2ω
Hence by the convolution theorem, a particular solution is

Z t
1
x(t) = f (τ )[eω(t−τ ) − e−ω(t−τ ) ]dτ.
2ω 0
Rt
25.9. (a) 0
x(τ )(t − τ )dτ = t4 . Let X(s) = L{x(t)}. The transform of the equation is
1 4!
X(s)L{t} = L{t4 }, or X(s) 2
= 5.
s s
Hence
24
X(s) = = 12t2 .
s3
Therefore the solution of the integral equation is x(t) = 12t2 .
Rt
(b) x(t) = 1 + 0 x(τ )(t − τ )dτ . Take the Laplace transform of the equation using the convolution
theorem: · ¸
1 X(s) s 1 1 1
X(s) = + 2 so that X(s) = 2 = + ,
s s s −1 2 s−1 s+1
using partial fractions. Inversion gives the solution
x(t) = 12 [e−t + et ] = cosh t.

Rt
(c) x(t) = sin t + 0
x(τ ) cos(t − τ )dτ . The Laplace transform of this integral equation is
1 s
X(s) = + X(s) 2 ,
s2 + 1 s +1
using the convolution theorem. Hence
1 1
X(s) = = .
s2 −s+1 (s − 2 )2 + 43
1
We can invert this transform using (24.6c) and the shift rule (24.7) resulting in the solution
1 √
x(t) = √23 e 2 t sin[ 12 3t].
25.10. The input f (t) = H(t) has the transform F (s) = 1/s. As in Section 25.6, the response
x∗∗ (t) has the transform
G(s)
X ∗∗ (s) = F (s)G(s) = so that G(s) = sX ∗∗ (s),
s
20
where G(s) is the transfer function between input and output. The transform of the output x(t) is
X(s) = G(s)F (s) = X ∗∗ (s)sF (s).
Using the convolution theorem, the output

Z t · ¸
∗∗ d
x(t) = x (t − τ ) (f (τ )) + f (0) dτ,
0 dτ
by the derivative rule (24.12). Let
Z t
X ∗∗ (s)F (s) ↔ q(t) = x∗∗ (τ )f (t − τ )dτ.
0
By the rule on the differentiation of integrals (15.20), the right-hand side can be expressed as
·Z t ¸
d ∗∗
x(t) = x (τ )f (t − τ )dτ .
dt 0
In the example
1 1 1 1
X ∗∗ (s) = = − ↔ (et − e−2t ),
(s − 1)(s + 2) 3(s − 1) 3(s + 2) 3
and f (t) = H(t) sin ωt.
25.11. To model this problem, approximate by assuming that the learning/ forgetting process
takes place continuously through whatever time-range is involved. Note that the learning data
refers to new words (forgotten words are not revised). The words learned at a time t = τ through
a small period δτ is equal to 50δτ . at any later time t, the number of newly-learned words that
are remembered is 50δτ e−0.01(t−τ ) , (that is, the elapsed time α is t − τ for this event). The total
number of words recalled at time t is the limit of the sum of the contributions between times 0
and t:
X t Z t
N (t) = lim 50e−0.01(t−τ ) δτ = 50e−0.01(t−τ ) dτ (i)
δτ →0 0
τ =0
(which is the convolution 50 ∗ e−0.01t ). Therefore

· 0.01τ ¸t
−0.01t e 1 − e−0.01t
N (t) = 50e = 50 .
0.01 0 0.01
(After 30 days N (t) becomes 50 × 25.9 in place of 50 × 30 attempted: a loss of 14%.)

(b) If the student aims at learning 50 + 0.1t words per day (thus increasing the input with time)
we obtain the convolution integral
Z t Z t
N (t) = (50 + 0.1τ )e−0.01(t−τ ) dτ = e−0.01t (50 + 0.1τ )e0.01τ dτ (ii)
0 0
(by 25.11). The integration by parts formula (17.8) with u = τ , dv/dτ = eAτ gives
Z t · µ ¶ ¸t · ¸
1 1 1 1
τ eAτ dτ = τ− eAτ = teAt + (1 − eAt ) ,
0 A A 0 A A
and applying this to (ii) with A = 0.01,

· µ ¶¸
−0.01t 1 − e0.01t 0.1 0.01t 1 − e0.01t
N (t) = e −50 + te +
0.01 0.01 0.01
· 0.01t
¸
1−e
= e−0.01t −40 + 10te0.01t .
0.01
21
25.12. The original population p0 declines as p0 e−γt . In time δτ , the number born is bp(τ )δτ , but
these individuals die out at the rate e−β(t−τ ) after the elapsed time t − τ . At time t the balance
between these births and deaths together with the decline of p0 gives the population p(t) at time
t: Z t
p(t) = p0 e−γt + b p(τ )e−β(t−τ ) dτ.
0
Taking the Laplace transform of this equation, which includes a convolution on the right-hand
side, we obtain
p0 bP (s)
P (s) = + ,
s+γ s+β
where P (s) = L{p(t)}. Therefore
· ¸
p0 (s + β) p0 b γ−β
P (s) = = + .
(s + γ)(s + β − b) b+γ−β s+β−b s+γ
This transform can now be inverted using (24.6a) giving the solution
p0 h i
p(t) = be(b−β)t + (γ − β)e−γt .
b+γ−β
25.13. The equation of motion is
mẍ + kx = F0 [H(t) − H(t − t0 )],
where the difference of the unit functions on the right ensures that the forcing is zero for t > t0 .
The initial conditions are x(0) = 0 and ẋ(0) = 0. The Laplace transform of this equation is
· ¸
2 1 e t0 s
ms X(s) + kX(s) = F0 − .
s s
Therefore · ¸
F0 (1 − e−t0 s ) F0 1 − e−t0 s s(1 − e−t0 s )
X(s) = = −
ms(s2 + ω 2 ) mω 2 s s2 + ω 2
where ω 2 = k/m. Inversion gives
F0
x(t) = [(1 − cos ωt) − (1 − cos ω(t − t0 ))H(t − t0 )].
k
Hence for t < t0 , the solution is
F0
x(t) = (1 − cos ωt),
k
whilst for t > t0 the solution is
F0
x(t) = [cos ω(t − t0 ) − cos ωt].
k
25.14. The differential equation is

dx(t)
= x(t − 1) + t,
dt
and x(t) = 0 for t ≤ 0. Take the Laplace transform of the equation:
Z ∞
1
sX(s) − x(0) = 2 + x(t − 1)e−st dt,
s 0
where X(s) = L{x(t)}. Since x(t) = 0 for t ≤ 0, and by using the second shift rule (24.15),
Z ∞ Z ∞
1 −st 1 1
sX(s) = 2 + x(t − 1)e dt = 2 + x(τ )e−s(u+1) du = 2 + e−s X(s).
s 1 s 0 s
22
Hence
1 1
X(s) = = 3 .
s2 (s −s
−e ) s (1 − e−s /s)
Using the binomial expansion
µ ¶−1
1 e−s 1 e−s e−2s
X(s) = 3 1− = + + + ···.
s s s3 s4 s5
The general term in this series is e−ns /sn+3 . The function of which this the Laplace transform is
½
(t − n)n+2 (t − n)n+2 /(n + 2)! n ≤ t
H(t − n) =
(n + 2)! 0 n>0
We sum the series as far as n where bt − 1c < n ≤ btc and btc is the largest integer less than or
equal to t. Therefore
btc
X (t − n)n+2
x(t) = .
n=0
(n + 2)!
25.15. The Laplace transform of the integral equation

Z t
2 cos(t − u)x(u)du = x(t) − t
0
is
s 1
2 X(s) = X(s) − 2 .
s2 + 1 s
Hence
s2 + 1 2 2 1 2
X(s) = = − + +
s2 (s− 1)2 (s − 1)2 s − 1 s2 s
using partial fractions. Finally, inverting this transform, the required solution is
x(t) = 2(t − 1)et + t + 2.
25.16. The differential equation is

d2 x dx
+t − x = 0,
dt2 dt
where x(0) = 0 and x0 (0) = 1. The derivative dx/dt has a variable coefficient t. From (24.8)
½ ¾
dx d dX(s)
L t = − [sX(s) − x(0)] = X(s) + s .
dt ds ds
Hence the Laplace transform of the full equation is
dX(s)
s2 X(s) − 1 − X(s) − s − X(s) = 0.
ds
Therefore X(s) satisfies
dX(s)
−s + (s2 − 2)X(s) = 1.
ds
Let X(s) = 1/s2 . Then
dX(s) 2 1
−s + (s2 − 2)X(s) = 2 + (s2 − 2) 2 = 1.
ds s s
Hence X(s) = 1/s2 satisfies the equation. The corresponding time-solution is x(t) = t which we
can confirm satisfies the initial conditions.
23
25.17. These equations have some coefficients which are not constant, and their transforms are
obtained by using the multiplication rule (24.8).
(a) tx00 (t) + (1 − t)x0 (t) − x(t) = 0, x(0) = x0 (0) = 1. Take the Laplace transform of the equation:
d 2 d
− [s X(s) − sx(0) − x0 (0)] + sX(s) − x(0) + [sX(s) − x(0)] − X(s) = 0,
ds ds
or
−2sX(s) − s2 X 0 (s) + 1 + sX(s) − 1 + X(s) + sX 0 (s) − X(s) = 0.
Hence
(s − 1)X 0 (s) + X(s) = 0.
This a first-order separable equation with general solution X(s) = C/(s−1), where C is a constant.
By the Table of Laplace transforms (Appendix F), inversion gives x(t) = Cet . The initial condition
x(0) = 1 means that C = 1. Hence the required solution is x(t) = et .
(b) x00 (t) + tx0 (t) − 2x(t) = 2, x(0) = x0 (0) = 0. Take the Laplace transform of the equation:
d 2
s2 X(s) − sx(0) − x0 (0) − [sX(s) − x(0)] − 2X(s) = ,
ds s
or
2
s2 X(s) − X(s) − sX 0 (s) − 2X(s) = .
s
Hence µ ¶
0 3 2
X (s) − s − X(s) = − 2 .
s s
This is a first-order differential equation of integrating-factor type (see Section 19.5). The inte-
grating factor is R
1 2 1 2
e (−s+3/s)ds = e(− 2 s +3 ln s) = s3 e− 2 s .
The equation can be expressed in the form
d 1 2 1 2
(X(s)s3 e− 2 s ) = −2se− 2 s .
ds
Integrating Z
1 2 1 2 1 2
X(s)s3 e− 2 s = − 2se− 2 s ds + C = 2e− 2 s + C,
where C is a constant. Hence the transform of the solution is

1 2
2 e2s
X(s) = 3
+C 3 .
s s
Do not attempt to invert the second term on the right. The inverse of the first term is t2 . This
term alone satisfies the initial conditions x(0) = x0 (0) = 0. We conclude that C = 0 which means
that the required solution is x(t) = t2 .
(c) tx00 (t) − x0 (t) + tx(t) = sin t, x(0) = 1, x0 (0) = 0. The Laplace transform of the equation is
d dX(s) 1
− [s2 X(s) − sx(0) − x0 (0)] − sX(s) + x(0) − = 2 ,
d ds s +1
or
3s 2 1
X 0 (s) + X(s) = 2 − .
s2 + 1 s + 1 (s2 + 1)2
This is an equation of integrating-factor type as in (b). The integrating factor is
R 2 3 2 3
e 3s/(s +1)ds = e 2 ln(s +1) = (s2 + 1) 2 ,
24
so that
d 3 1 1
[X(s)(s2 + 1) 2 ] = 2(s2 + 1) 2 − (s2 + 1)− 2 .
ds
Integrating:
Z Z
3 1 1
2 2
X(s)(s + 1) 2 = 2 (s + 1) ds − 2 (s2 + 1)− 2 ds + C
Z
d 1
= [s(s2 + 1) 2 ]ds + C
ds
1
= s(s2 + 1) 2 + C
Hence
s C
X(s) = + .
s2 + 1 (s + 1) 32
2
The inverse of the first term is cos t which alone satisfies the initial conditions. Therefore C = 0
and the required solution is x(t) = cos t.
25.18. (a) {1, 2, 1, 0, 0, 0, . . .}, 1+ z2 + z1 , and the Laplace transfrom 1+2e−T s +e−2T s are equivalent.
They represent the sum of the impulses: δ(t) + 2δ(t − T ) + δ(t − 2T ).
(b) The sequence {0, 1, 2, 3, . . .}; the z transform
1 2 3
+ 2 + 3 + ···,
z z z
and the Laplace transform
e−T s + 2e−2T s + 3e−3T s + · · ·
are equivalent, representing the time function
δ(t − T ) + 2δ(t − 2T ) + 3δ(t − 3T ) + · · · .
Note also that

1 2 3 z
+ 2 + 3 + ··· =
z z z (1 − z)2
for large z (compare Example 25.18), which is the z transform in finite terms.
(c) The notation {3} means here the sequence {3, 0, 0, 0, . . .}, standing for the z transform equal
to 3/z 0 = 3. The Laplace transform therefore equals 3, corresponding to the time function 3δ(t).
2 3
(d) {(−2)n }, {1, −2, 22 , −23 , . . .}, 1 − z2 + z22 − z23 + · · ·, and
1 − 2e−T s + 22 e−2T s − · · · are all equivalent, and correspond to the time function δ(t) − 2δ(t − T ) +
22 δ(t − 2T ) − · · ·. In finite terms: the (geometric) series for the z transform has the sum
1 z
=
1 + 2/z z+2
for large z.
(e) {0, 0, 3}, or 3/z 2 , is the z transform, 3e−2T s the Laplace transform and 3δ(t − 2T ) the time
function.
25.19. Note. The periods are denoted by T . Inputs and outputs are related through their z
transforms by Y(z) = G(z)X (z) (see (25.26)).
µ ¶µ ¶
1 1 2 1
(a) Y(z) = 1 + 1+ = 1+ + 2.
z z z z
The inverse is
y(t) = δ(t) + 2δ(t − T ) + δ(t − 2T ).
25
1 1 1
(b) G(z) = 1 + + 2 2 + · · · , and X (z) = 1 + .
2z 2 z z
Therefore
µ¶ µ ¶ µ ¶
1 1 1 1 1 1 1 1
Y(z) = 1+ +1 + + + + 2 + ···
2 z 22 2 z2 23 2 z3
µ ¶
3 1 1 1
= 1+ + 2 + 2 3 + ··· .
2 z 2z 2 z
The inverse is given by
· ¸
3 1 1
Y(t) = δ(t) + δ(t − T ) + δ(t − 2T ) + 2 δ(t − 3T ) + · · · .
2 2 2
(c)
µ ¶µ ¶ µ ¶µ ¶
1 1 2 2 2 1 1 1
Y(z) = 1 − + 2 − ··· + 2 = 1+ 1 − + 2 − ···
z z z z z z z z
µ ¶
2 1 1 2
= 1+ · = .
z z 1 + z1 z
Therefore y(t) = 2δ(t − T ).

25.20. G(s) = 1/(1 − 31 e−T s ), X(s) = e−T s + 2e−2T s . Put eT s = z (see (25.22)); then the
corresponding z transforms are given by
1 1 2
G= , X (z) = + 2.
1 − 13 z1 z z
Expand G(z) as a geometric series (see (5.4a)) in powers of 1/z:

11 1 1
G(z) = 1 + + 2 2 + ···.
3z 3 z
Since Y(z) = G(z)X (z) (see (25.26)), we have
µ ¶ µ ¶
11 1 1 1 2
Y(z) = 1+ + 2 2 + ··· 1+
3z 3 z z z
µ µ ¶ µ ¶ µ ¶ ¶
1 1 1 1 1 1 1 1 1
= 1+ +2 + +2 + + 2 + · · ·
z 3 z 3 3 z2 32 3 z3
µ ¶
1 7 1 1 1
= + 2
+ 3 + 2 4 + ···
z 3 z 3z 3 z
The output y(t) is therefore given by
y(t) = δ(t − T ) + 73 [δ(t − 2T ) + 31 δ(t − 3T ) + 1
32 δ(t − 4T ) + · · ·].
25.21. (a)
1 2 1
X (z) = + 2 − 3.
z z z
(b)
1 1 1 1 z
X (z) = 1 − + 2 − 3 + ··· = = (from (5.4a)) .
z z z 1 + 1/z z+1
(c)
1 1 1 1
X (z) = {1, , 2 , . . .} = 1 + + 2 2 + ···
2 2 2z 2 z
1 2z
= = .
1 − (1/2z) 2z − 1
26
from (5.4a).
(d) Put z = eT s as in (25.22):
1 1 z
X (z) = = 2 .
z 1 − (1/z 2 ) z −1
25.22. (a) By sampling at times t = nT , n = 0, 1, 2, . . ., we obtain x(t) = {nT }. Then
T 2T 3T
X (t) = + 2 + 3 + ···.
z z z
This series can be summed by using the process in Example 25.18 to give
Tz
X (z) = .
(z − 1)2
(b) As in (a) we obtain
e−T e−2T e−T 1

X (z) = + 2 + ··· =
z z z 1 − (e−T /z)
by summing the geometric series.

[Note: (c) and (d) have been deleted from the 2003 reprint.)
25.23. (a) Y = GX in general, so
µ ¶ µ ¶
1 1
1− = 1+ G(z)
z z
which gives G(z) in the closed form

µ ¶Áµ ¶
1 1
G(z) = 1 − 1+ .
z z
To obtain g(t), first expand G(z) in powers of 1/z for large z. We have (using (5.4a) for geometric
series) µ ¶µ ¶
1 1 1 2 2 2
G(z) = 1 − 1 − + 2 − ··· = 1 − + 2 − 3 + ···.
z z z z z z
Therefore
g(t) = δ(t) − 2δ(t − T ) + 2δ(t − 2T ) − · · · .
(b) As in (a) µ ¶ µ ¶
1 3
1+ = 1 + 3 G(z),
z z
so µ ¶Áµ ¶
1 3
G(z) = 1+ 1+ 3 .
z z
µ ¶ µ ¶
1 1
(c) 1+ = 1− G(z),
z z
so that µ ¶Áµ ¶
1 1
G(z) = 1 + 1− .
z z
1 1 1
(d) Y(z) = 1 − 2
+ 4 − ··· = ,
z z 1 + (1/z 2 )
27
1 1 1
X (z) = 1 + + 4 + ··· = .
z2 z 1 − (1/z 2 )
Since G(z) = Y(z)/X (z),
1 − (1/z) z(z − 1)
G(z) = = 2 .
1 + (1/z 2 z +1
(e) As in (d), we find
1 1
Y(z) = , X (z) = ,
1 + (1/z 2 ) 1 − (1/z 2 )
so that
z2 + 1
G(z) = .
z2 − 1
25.24. The transform of y(t), defined by the discrete values
{x0 , x1 e−CT , x2 e−2CT , . . .}
is Y (z) given by
x1 e−CT x2 e−2CT
Y (z) = x0 + + + ···
z z2
µ −CT ¶ µ −CT ¶2
e e
= x0 + x1 + x2 + ···
z z
= X (e−CT /z),
where X (z) = {x0 , x1 , x2 , . . .}.

25.25.
x1 x2
(a) X (z) = x0 + + 2 + ···
z z
is the transform of x(t). The transform of (0, x0 , x1 , . . .) is
x0 x1
+ 2 + ···,
z z
and this is equal to z1 X (z).
(b) The rule (a) applies to z1 X (z) as it did to X (z): the introduction of a further zero to move
the row along causes the transform to be multiplied by a further factor z1 , giving z12 X (z). Perform
this procedure N times to introduce N zeros; the transform becomes z1N X (z).
25.26. From the definition
x1 x2
{x0 , x1 , x2 , . . .} = x0 + + 2 + · · · = X (z).
z z
Put
xN +1
{xN , xN +1 , . . .} = xN + + · · · = X (N ) (z) (say).
z
Obviously
z N X (z) = z N x0 + z N −1 x1 + · · · + xN −1 + X (N ) (z).
Therefore
X (N ) (z) = z N X (z) − z N x0 − z N −1 x1 − · · · − zxN −1 .
25.27. Stability: the system is stable only if the poles of G(z) all have modulus less than 1.
(a) G(z) = (z + 1)/(z 2 − 4). In the argand diagram the poles are at z = ±2 = c1 , c2 . Since |c1 | (or
|c2 |) > 1, the system is unstable.
28
(b) G(z) = (z 2 − z)/(4z 2 − 1). There are poles at z = c1 = 12 and z = c2 = − 21 . Their moduli are
both < 1, so the system is stable.
(c) G(z) = 1/(4z 2 + 1). The poles are at c1 = 14 i and c2 = − 14 i. |c1 | and |c2 | are both < 1, so the
system is stable.
(d) G(z) = (z 3 + 1)/(2z 4 + 5z 2 + 2). The poles occur where 2z 4 + 5z 2 + 2 = 0. Solving this √ as a
quadratic equation for z 2 , we obtain z 2 = − 12 and −2, so the poles are at z = ± √12 i and ± 2i.
√ √ √
The poles at ± 2i have | ± 2i| = 2 > 1, so the system is unstable.
Growth rate of transients. From the results leading up to eqn (25.40) the overall growth (or decay)
of responses of the system is determined by the behaviour of a factor |c|N , where c is the pole
associated with a particular mode or transient, and N = t/T , where t is time and T the period.
For the given functions G(z), we have :
(a) The factor is 2N for both poles.
(b) The factor is ( 12 )N for both poles.
(c) The factor is ( 14 )N for both poles.
√ 1 1
(d) The factors are ( 2)N = 2 2 N for two poles, and 2− 2 N for the other two.
25.28. Notes. For the solution method see Example 25.23, or use the general result from Prob-
lem 25.26. The notation ‘→’ stands for the words “has the z transform equal to”. The counter n
runs through n = 0, 1, 2, . . ..
(a) 4yn+2 − yn = xn ; y0 = 1, y1 = 2. (i)
Put {xn } = {x0 , x1 , . . .} → X (z), and {yn } = {y0 .y1 , . . .} → Y(z). Then
{yn+2 } → z 2 Y − z 2 y0 − zy1 = z 2 Y − (z 2 + 2z)
(by Problem 25.26). The transform of (i) is therefore given by
4[z 2 Y − (z 2 + 2z)] − Y = X ,
so that
4(z 2 + 2z) + X (z) 1 + 8z + X (z)
Y(z) = 2
=1+ . (ii).
4z − 1 4z 2 − 1
The poles occur at z = ± 12 , for which |z| < 1. Inspection of (ii) shows that there are no growth or
non-diminishing terms arising from an impulsive input, so the system is stable.
(b) yn+2 − 3yn+1 + 2yn = 2xn; y0 = 0, y1 = 1. (ii)
Proceding as in (a), (ii) becomes
(z 2 Y − z) + 3zY + 2Y = 2X ,
so that
2X (z) + z
Y(z) = .
(z + 1)(z + 2)
Since one of the poles has a modulus > 1, the system is unstable.
(c) 2yn+2 + yn+1 + yn = xn+1 − xn ; y0 = 0, y1 = 1. (iii)
Proceeding as in (a), (iii) becomes
2(z 2 Y − z) + zY + Y = zX − zx0 − X ,
so that
(z − 1)X (z) − zx0 + 2z
Y(z) = .
2z 2 + z + 1
√ √
The poles are at z = 14 (−1 ± i 7), whose moduli are equal to 1/ 2 < 1. The argument in
Section 25.10 leads us to expect stability; that is, no terms arising from an impulsive input whose
effect does not decrease to zero.
29
(d) 2yn+2 + 3yn+1 − yn = xn ; y0 = 1, y1 = 1. (iv Proceeding as in (a), (iv) becomes
(2z 2 + 3z − 1)Y − (2z 2 + 2z + 3z) = X ,
or
X (z) + 2z 2 + 5z
Y(z) = .
2z 2 + 3z − 1
√
There are poles at 14 [−3 ± 17] with modulus > 1, so the system is unstable.
Chapter 26: Fourier series
26.1. These functions are all odd 2π-periodic, which means that a0 = a1 = a2 = . . . = 0, whilst
Z
2 π
bn = f (t) sin ntdt.
π 0
The figures show the graph of the function over the interval −π < t < π.
(a) ½
−1 (−π < t < 0)
f (t) =
1 (0 ≤ t ≤ π)
x
1
t
-Π Π
-1
Therefore Z π ½
2 4/(nπ) n odd
bn = sin ntdt =
π 0 0 n even
The Fourier series for f (t) is
4 4 4
sin t + sin 3t + sin 5t + · · · .
π 3π 5π
(b) f (t) = t. Then Z π
2 2(−1)n
bn = t sin ntdt = − .
π 0 n
x
3
2
1
t
-Π Π
-1
-2
-3
30
x
t
-Π Π
-4
-8
(c) ½
−t2 (−π < t < 0)
f (t) =
t2 (0 ≤ t ≤ π)
Therefore Z π
2 2
bn = t2 sin ntdt = {2 + (−1)n [n2 π 2 − 2]}.
π 0 n3 π
(d) ½
e−t − 1 (−π < t < 0)
f (t) =
−(et − 1) (0 ≤ t ≤ π)
x
20
10
t
-Π Π
-10
-20
Therefore
Z π
2 2[−1 + (−1)n + (−1)n (1 − eπ )n2 ]
bn = − (et − 1) sin ntdt = − .
π 0 n(1 + n2 )π
(e) 

 1 (−π < t ≤ − 12 π)

−1 (− 21 π < t ≤ 0
f (t) =

 1 (0 < t ≤ 12 π)

−1 ( 12 π < t ≤ π)
Therefore
Z 1
2π
Z π
2 2
bn = sin ntdt − sin ntdt
π 0 π 1
2π
(· ¸ 12 π · ¸π )
2 cos nt cos nt
= − − −
π n 0 n 1
2π
2 1
= [1 − 2 cos πn + (−1)n ].
πn 2
31
x
1
t
-Π Π Π Π
-
2 2
-1
The sequence of coefficients is

4 4 4
b1 = 0, b2 = , b3 = b4 = b5 = 0, b6 = , b7 = b8 = b9 = 0, b10 = ,....
π 3π 5π
26.2. For even functions the coefficients bn are all zero, whilst for 2π-periodic functions
Z
2 π
an = f (t) cos ntdt, (n = 0, 1, 2, . . .).
π 0
The diagrams show the functions over one period −π < t < π.
(a) 
 −1 (−π < t ≤ 12 π)
f (t) = 1 (− 12 π < t ≤ 12 π)

−1 ( 12 π < t ≤ π)
x
1
t
-Π Π Π Π
-
2 2
-1
Then Z π
2 4 1
an = f (t) cos ntdt = sin nπ.
π 0 nπ 2
The even coefficients a0 , a2 , a4 , . . . are zero, whilst
4 4 4
a1 = , a3 = − , a5 = ,....
π 3π 5π
(b) f (t) = t2 , (−π < t ≤ π)..
In this case Z π
2 4(−1)n
an = t2 cos ntdt = , (n ≥ 1)
π 0 n2
integrating by parts. For n = 0, a0 = 2π 2 /3.
(c) f (t) = cos 12 t, (−π < t ≤ π).
Then
Z
2 π 1 4 cos nπ 4(−1)n
an = cos t cos ntdt = − 2
=− , (n = 0, 1, 2, . . .).
π 0 2 (4n − 1)π (4n2 − 1)
32
x
t
-Π Π
-4
-8
x
1
t
-Π Π
26.3. The Fourier coefficients for a 2π-periodic function f (t) are

Z
2 π
an = f (t) cos ntdt, (n = 0, 1, 2, . . .).
π 0
Z
2 π
bn = f (t) sin ntdt. (n = 1, 2, . . .).
π 0
(a) ½
0 (−π < t < 0)
f (t) = .
t (0 ≤ t ≤ π)
x
3
t
-Π Π
The Fourier coefficients are

Z Z
1 π π 1 π
(−1)n − 1
a0 = tdt = an = t cos ntdt = , (n = 1, 2, . . .)
π 0 , π 0 n2 π
Z π
1 (−1)n
bn = t sin ntdt = − , (n = 1, 2, . . .).
π 0 n
(b) ½
t+π (−π < t < 0)
f (t) = .
t (0 < t ≤ π)
33
x
3
t
-Π Π

Z Z
1 π 1 π
a0 = f (t)dt = π, an = f (t) cos ntdt = 0, (n = 1, 2, . . .),
π −π π −π
Z π
1 1 + (−1)n
bn = f (t) sin ntdt = − .
π −π n
26.4. The half-rectified sine wave is generated by the 2π-periodic function

½
0 (−π < t ≤ 0)
f (t) = .
sin t (0 < t ≤ π)

2 1 + (−1)n
a0 = , a1 = 0, an = − (n = 2, 3, 4, . . .),
π (n2 − 1)π
1
b1 = , an = 0, (n = 2, 3, 4, . . .).
2
The Fourier series is
1 1 2 2
f (t) = + sin t − 2 cos 2t − 2 cos 4t − · · · .
π 2 (2 − 1)π (4 − 1)π
26.5. The Fourier coefficients of the 2π-periodic function

½
0 (−π < t ≤ 0)
f (t) =
1 (0 < t ≤ π)
are Z π
1
a0 = 1, an = cos ntdt = 0 (n = 1, 2, 3, . . .),
π 0
Z π
1 1 − (−1)n
bn = sin ntdt = (n = 1, 2, 3, . . .).
π 0 n
Hence b1 = 2/π, b2 = 0, b3 = 2/(3π), b4 = 0, b5 = 2/(5π), and so on, which generates the Fourier
series µ ¶
1 2 1 1
+ sin t + sin 3t + sin 5t + · · · .
2 π 3 5
At t = 0, the Fourier series takes the value 12 , in agreement with the definition f (0) = 0. There
is a discontinuity in f (t) at t = 0. By (26.12), the Fourier series takes the average value of the left
and right hand values of the function at the discontinuity.
Put t = 12 π. Then sin 12 π = 1, sin 3π 5π
2 = −1, sin 2 = 1 and so on. Since the Fourier series
equals the function on continuous parts of the curve,
¡ ¢
f ( 12 π) = 1 = 12 + π2 1 − 13 + 15 − 71 + · · · .
34
Therefore
1 1 1 π
1− + − + ··· = .
3 5 7 4
26.6. The fully rectified sine wave of period 2π is f (t) = F | sin t|. This is an even function so that
bn = 0. The other Fourier coefficients are
Z
1 π
an = | sin t| cos ntdt
π −π
Z π
2 2(1 + (−1)n
= sin t cos ntdt = − , (n = 0, 1, 2, . . .).
π 0 (n2 − 1)π
Hence
2 4 4
−
| sin t| = cos 2t − cos 4t − · · · .
π 3π 15π
The amplitude of the first non-zero harmonic (period π) is 4/(3π).
26.7. The Fourier series is
∞
X n+a
3 + an + 3
sin nt.
n=1
n
The first two harmonics have amplitudes (a + 1)/(a + 3) and (a + 2)/(2a + 11). These are in the
ratio 2 : 1 if
a + 1 2a + 11
= 2.
a+3 a+2
Therefore
(a + 1)(2a + 11) = 2(a + 3)(a + 2), or 2a2 + 13a + 11 = 2(a2 + 5a + 6).
Hence a = 13 .
The next harmonic (n = 3) has amplitude
3+a 3 + 13 10
= = .
27 + 3a + 3 27 + 1 + 3 93
26.8. (See Section 26.9) Note that both functions are odd with period 2π. From Figure 26.18(a)
the straight line is given by x = F t/π. Its Fourier series is given essentially by Problem 26.1(b),
with coefficients (suitably scaled)
2F (−1)n
an = 0, (n = 0, 1, 2, . . .), bn = − , (n = 1, 2, . . .).
nπ
The second step function can be obtained from Problem 26.1(a) by inserting a minus sign. Its
Fourier coefficients are
½
−4/(nπ) n odd
cn = 0, (n = 0, 1, 2, . . .), dn = .
0 n even
The Fourier series is therefore

∞
X 2F − 4 F 2F − 4 F
(bn + dn ) sin nt = sin t − sin 2t + sin 3t − sin 4t + · · · .
n=1
π π 3π 2π
The leading harmonic has zero amplitude if F = 2.

26.9. The T -periodic function is Q(t) = 14 T 2 − t2 for − 12 T ≤ t ≤ 12 T , and it is even. Hence all
bn = 0. The other Fourier coefficients are
Z 12 T
2 T2
a0 = ( 14 T 2 − t2 )dt = .
T − 12 T 6
35
Z 1
2T
2 (−1)n T 2
an = ( 14 T 2 − t2 ) cos(2πnt/T )dt = − , (n = 1, 2, 3, . . .).
T − 12 T n2 π 2
The approximation given by the first four terms of the Fourier series is
1
Q4 (t) = = a0 + a1 cos(2πt/T ) + a2 cos(4πt/T ) + a3 cos(6πt/T )
2· ¸
1 1 1 1
= T2 + 2 cos(2πt/T ) − 2 cos(4πt/T ) + 2 cos(6πt/T ) .
6 π 4π 9π
(a) At t = 0, Q(0) = 14 T 2 = 0.25T 2 and Q4 (0) = ( 16 + 36π 31

2 )T
2
= 0.2539 . . . T 2 .
(b) At t = 14 T , Q( 14 T ) = 16
3 2
T = 0.1875T 2 and Q4 ( 14 T ) = ( 61 + 4π1 2 )T 2 = 0.1919 . . . T 2 .
26.10. The 2π-periodic function
½
βt(π − t) (0 < t ≤ π)
f (t) = .
βt(π + t) (−π < t ≤ 0)
is odd, so that all Fourier coefficients an are zero. The other coefficients are
Z
2 π 4(1 − (−1)n )β
bn = βt(π − t) sin ntdt = ,
π 0 n3 π
using integration by parts. The Fourier series is
X∞
4(1 − (−1)n )β
f (t) = sin nt
n=1
n3 π
8β 8β 8β
= sin t + sin 3t + sin 5t + · · · .
π 27π 125π
The ratio of the first and third harmonics is
b1 8β 27π
= = 27.
b3 π 8β
Let
8β 8β
f3 (t) = sin t + sin 3t.
π 27π
Then comparison of f (t) and f3 (t) at t = 12 π gives
π2 β 8β 8β
f ( 12 π) = 4 = 2.467 . . . β, f4 ( 12 π) = π − 27π = 2.452 . . . β.
26.11. The function f (t) = t(π 2 − t2 ) is an odd function which means that all coefficients an = 0.
The sine coefficients are given by
Z
1 π 12(−1)n
bn = t(π 2 − t2 ) sin ntdt = − .
π −π n3
The Fourier series of f (t) is

∞
X X∞
12(−1)n
f (t) = bn sin nt = − sin nt. (i)
0 n=0
n3
The derivative of f (t) is f 0 (t) = π 2 − 3t2 . As expected f 0 (t) is an even function so that the
coefficients bn = 0. The cosine coefficients are
Z
1 π 2 12(−1)n
a0 = 0, an = (π − 3t2 ) cos ntdt = − .
π −π n2
36
The Fourier series of f 0 (t) is
X∞
12(−1)n
− cos nt.
n=1
n2
It can be seen that this Fourier series is the derivative of the Fourier series in (i), which confirms
that, in this case, the derivative of the Fourier series of f (t) is the Fourier series of the derivative
of f (t).
The function g(t) = t3 is odd so that all the Fourier coefficients an are zero. The sine coefficients
are given by
2(−1)n 2 2
bn = − (n π − 6), (n = 1, 2, . . .)
n3
Hence
X∞
2(−1)n (6 − n2 π 2 )
g(t) = sin nt.
n=1
n3
The derivative g 0 (t) = 3t2 which is an even function. Hence all bn = 0, and
12(−1)n
a0 = 2π 2 , an = , (n = 1, 2, . . .)
12n2
Therefore
X∞
12(−1)n
g 0 (t) = 3t2 = 2π 2 + cos nt.
n=1
n2
Clearly the derivative of the Fourier series of g(t) obtained by term-by-term differentiation, namely,
X∞
2(−1)n (6 − n2 π 2 )
cos nt
n=1
n2
P∞
is not the Fourier series of g 0 (t). A problem arises because the series n=1 cos nt which occurs
among the terms in this expression does not converge. This series does not have a sum.
26.12. A 2π-periodic rectified sine wave is defined by
½
0 (−π ≤ t ≤ 0)
x = P (t) =
| sin 2t| (0 < t ≤ π)
(see Fig. 11).
x
1
t
-2 Π -Π Π 2Π
The Fourier coefficients are given by

Z π
4 4
a0 = , an = | sin 2t| cos ntdt = − 2 [−(n + 1) − 4 cos 21 nπ],
π 0 (n − 4)π
Z ∞
8
bn = | sin 2t| sin ntdt = − 2 sin 12 nπ.
0 (n − 4)
37
26.13. One particular representation of the function f (t) = t on −π < t ≤ π is obtained by
considering its periodic extension (see Section 26.7). The Fourier coefficients are an = 0, (n =
0, 1, 2, . . .), and Z
1 π 2(−1)n
bn = t sin ntdt = − .
π −π n
Hence
X∞
(−1)n−1
t=2 sin nt,
n=1
n
on −π < t ≤ π, as required.
Integrate this series term-by-term from t = 0 to t = x:
Z x Z x Xx
1 2 (−1)n−1
tdt = x = 2 sin ntdt
0 2 0 n=1
n
∞
X (−1)n−1
= −2 2
[cos nt]xt=0
n=1
n
X∞ X∞
(−1)n−1 (−1)n−1
= −2 2
cos nx + 2 .
n=1
n n=1
n2
Therefore, for −π < x ≤ π,

X∞ X∞
(−1)n−1 (−1)n−1
x2 = 4 − 4 cos nx. (i)
n=1
n2 n=1
n2
1
Equation (26.10) states that the average value of the function is equal to 2 a0 . The average
value of x2 over a period is Z π
1 1
x2 dx = π 2 .
2π −π 3
Therefore the constant term in (i) is given by
X∞
(−1)n−1 1
4 2
= π2 .
n=1
n 3
This inverse method determines the sum of the series on the left.
26.14. The Fourier series for t2 can be obtained by referring back to Problem 26.13. Quoting the
result
X∞
1 (−1)n
t2 = π 2 + 4 cos nt.
3 n=1
n2
Integrate both sides of this equation:
Z x Z x X∞ Z
1 3 1 2 (−1)n x
t2 dt = x = π dt + 4 cos ntdt
0 3 3 0 n=1
n2 0
X∞
1 2 (−1)n
= π x+4 3
[sin nt]x0
3 n=1
n
X∞
1 2 (−1)n
= π x+4 sin nx
3 n=1
n3
Finally
X∞
(−1)n
x3 − π 2 x = 12 sin nx.
n=1
n3
38
26.15. The T -period function
½
−2t (− 12 T ≤ t < 0)
P (t) =
2t (0 ≤ t < 21 T )
is even, so that the Fourier coefficients bn are all zero. The cosine coefficients are given by
Z 1
2T
4 2((−1)n − 1)T
a0 = T, an = 2t cos(2πnt/T )dt = (n = 1, 2, . . .).
T 0 n2 π 2
Hence
∞
1 2T X (−1)n − 1
P (t) = T+ 2 cos(2πnt/T )
2 π n=1 n2
∞
1 4T X 1
= T− 2 cos[2π(2m + 1)t/T ].
2 π m=0 (2m + 1)2
The spectral components are | 21 a0 |, |a1 |, |a3 |, |a5 | . . ., that is
T 4T 4T 4T
, , , ,...,
2 π 2 9π 2 25π 2
at n = 0, 1, 3, 5, . . .. The spectrum is shown in Fig. 12.
amplitudeT
0.5
0.4
0.3
0.2
0.1
n
1 2 3 4 5 6 7
26.16. Use formula (26.15) with f (t) = 1 and t0 = 1.

(a) For the half-range sine series, the coefficients are
Z 1
2[1 − (−1)n ]
bn = 2 sin(nπt)dt = .
0 nπ
Therefore
X∞ ∞
2[1 − (−1)n ] 4 X sin(2r − 1)πt
1= sin(nπt) = .
n=1
nπ π r=1 2r − 1
for 0 < t < 1. Note that this series represents an odd function which takes the value −1 for
−1 < t < 0, and (by (26.12)) zero at t = 0. It is shown in Fig. 13.
(b) For the half-range cosine series, the coefficients are
Z 1
2
a0 = 2, an = 2 cos(nπt)dt = [sin(nπt)]10 = 0.
0 nπ
This represents an even function which is 1 for all t. In other words the function is its own
half-range cosine series
26.17. Use formula (26.15) with f (t) = t and t0 = 1
39
x
1
t
-2 -1 1 2
-1
x
1
t
-2 -1 1 2
(a) For the half-range cosine series, the coefficients are

Z 1 Z 1
2((−1)n − 1)
a0 = 2 tdt = 1, an = 2 t cos(nπt)dt = .
0 0 n2 π 2
Hence
∞ ∞
1 2 X [(−1)n − 1] 1 4 X cos[(2r − 1)πt]
t= + 2 cos(nπt) = − .
2 π n=1 n2 2 π 2 r=1 (2r − 1)2
(b) For the half-range sine series, the coefficients are

Z 1
2(−1)n+1
bn = 2 t sin(nπt)dt = .
0 nπ
Therefore the half-range sine series is

∞
2 X 2(−1)n+1
t= sin(nπt),
π n=1 n
for 0 < t < 1. For −1 < t < 0 the sum of the series is −t.
26.18. Use formula (26.15) with f (t) = sin ωt and t0 = π/ω. The coefficients for the half-range
cosine series are
4
a0 = , a1 = 0,
π
Z Z
2ω π/ω 2ω π/ω
an = f (t) cos(nωt)dt = sin(ωt) cos(nωt)dt
π 0 π 0
2(1 + (−1)n
= − , (n = 2, 3, 4, . . .)
π(n2 − 1)
Hence
∞ ∞
2 2 X (1 + (−1)n ) 2 4 X cos(2rωt)
sin ωt = − cos(nωt) = − ,
π π n=2 n2 − 1 π π r=1 4r2 − 1
40
x
1
t
2Π Π Π 2Π
-
Ω Ω Ω Ω
for 0 < t < π/ω. The series represents an even function with sum | sin ωt|, which is shown in Fig.
15.
26.19. Use formula (26.15) with f (t) = cos ωt and t0 = π/ω. The coefficients for the half-range
sine series are Z
2ω π/ω 2n(1 + (−1)n )
b1 = 0, bn = cos(ωt) sin(nωt)dt = .
π 0 π(n2 − 1)
Hence
∞ ∞
2 X n(1 + (−1)n ) 8X r
cos ωt = sin(nωt) = sin(2rωt),
π n=2 n2 − 1 π r=1 4r2 − 1
for 0 < t < π/ω. The sum of the series is odd, so that for −π/ω < t < 0 the sum is − cos ωt except
at t = nπ/ω, where its sum is zero.
x
1
t
2Π Π Π 2Π
- -
Ω Ω Ω Ω
-1
26.20. Use formula (26.15) with f (t) = cos t and t0 = 2π. The coefficients for the half-range sine
series are Z
1 2π 2n[(−1)n − 1]
bn = cos t sin(nt/2)dt = − ,
π 0 π(n2 − 1)
provided n 6= 2. For n = 2, b2 = 0. Hence
∞
4 X sin[ 12 (2r − 1)t]
cos t = ,
π r=1 (2r − 1)2 − 4
for 0 < t < 2π.

26.21. Use formula (26.15) with f (t) = cos t and t0 = 2π.The coefficients for the half-range cosine
series are given by Z Z
1 2π 1 2π
a0 = cos tdt = 0, a2 = cos2 tdt = 1,
π 0 π 0
Z
1 2π
an = cos t cos(nt/2)dt = 0, (n = 1, 3, 4, . . .).
π 0
The function f (t) = cos t is its own half-range cosine series.
41
26.22. The function f (t) is given by
½
1 (0 ≤ t < 12 π)
f (t) =
0 ( 12 π ≤ t ≤ π)
In (26.15), t0 = π. (a) For the half-range sine series the coefficients are
Z
2 π
bn = f (t) sin ntdt
π 0
Z 1
2 2π 2
= sin ntdt = [1 − cos( 12 nπ)].
π 0 nπ
Hence the half-range sine series is

∞
2 X [1 − cos 12 nπ]
f (t) = sin nt.
π n=1 n
The terms for which n = 4, 8, 12, . . . are all zero.

(b) For the half-range cosine series the coefficients are a0 = 1 and
Z
2 π
an = f (t) cos ntdt
π 0
Z 1
2 2π 2 1
= cos ntdt = sin nπ.
π 0 nπ 2
Hence
∞ ∞
1 2 X sin 12 nπ 2 X (−1)r+1
f (t) = + cos nt = cos(2r − 1)t.
2 π n=1 n π r=1 2r − 1
26.23. (See Section 26.9) The 2π-periodic function specified by

½
−t (−π ≤ t ≤ 0)
P (t) =
t (0 ≤ t ≤ π)
has the Fourier series

µ ¶
1 4 cos t cos 3t cos 5t
P (t) = π − + + + ··· .
2 π 12 32 52
(a) Rescale t by putting t = πτ /2 so that when t = π, τ = 2. Hence

µ ¶
1 4 cos(πτ /2) cos(3πτ /2) cos(5πτ /2)
P (πτ /2) = π − + + + ··· .
2 π 12 32 52
Finally, rescaling P ,
½
6 −3τ (−2 ≤ τ ≤ 0)
Q(τ ) = P (πτ /2) =
π 3τ (0 ≤ t ≤ 2)
µ ¶
24 cos(πτ /2) cos(3πτ /2) cos(5πτ /2)
= 3− 2 + + + · · · .
π 12 32 52
The symbol τ can be replaced by t in the answer.

(b) First rescale t by putting t = πτ , so that when t = π, τ = 1. Therefore
µ ¶
1 4 cos πτ cos 3πτ cos 5πτ
P (πτ ) = π − + + + ··· .
2 π 12 32 52
42
Then µ ¶
1 1 4 cos πτ cos 3πτ cos 5πτ
R(τ ) = 1 − P (πτ ) = + + + + · · · .
π 2 π 12 32 52
As before τ can be replaced by t.
(c) P (t) has the spectral components
π/2, 4/π, 4/(9π), 4/(25π), . . . at n = 0, 1, 3, 5, . . . .
The functions Q(t) and R(t) have spectral components at the same values of n but scaled in
magnitude.
26.24. (See Section 26.9) The period-2 function
½
−1 (−1 ≤ t < 0)
P (t) =
1 (0 ≤ t < 1)
has the Fourier series

µ ¶
2 1 1
P (t) = sin πt + sin 3πt + sin 5πt + · · · .
π 3 5
Let t = 2τ /T so that when t = 1, τ = 12 T . Hence

µ ¶
2 1 1
P (2τ /T ) = sin(2πτ /T ) + sin(6πτ /T ) + sin(10πτ /T ) + · · · .
π 3 5
Multiply both sides by a and change the sign. Then
½
a ( 12 T ≤ t < 12 T )
Q(τ ) = −aP (2τ /T ) =
−a (0 ≤ 0)
µ ¶
2a 1 1
= sin(2πτ /T ) + sin(6πτ /T ) + sin(10πτ /T ) + · · ·
π 3 5
This function is periodic with period 2a, which means that Q(τ ) = a for 12 T ≤ t < T .
26.25. The 2π-periodic function f (t) = t, (−π ≤ t ≤ π) is odd so that the coefficients an are all
zero. The sine coefficients are given by
Z
1 π 2(−1)n
bn = t sin ntdt = − , (n = 1, 2, 3, . . .).
π −π n
Hence
X∞
(−1)n+1
t=2 sin nt.
n=1
n
To determine a particular solution of
d2 x
+ Ω2 x = K sin ωt,
dt2
try a solution x = A cos ωt. Then
d2 x
+ Ω2 x − K sin ωt = −Aω 2 cos ωt + Ω2 A cos ωt − K cos ωt
dt2
= [A(Ω2 − ω 2 ) − K] cos ωt = 0
for all t, if A = K/(Ω2 − ω 2 ). The forced solution is therefore

K cos ωt
x= , (Ω2 6= ω 2 )
Ω2 − ω 2
43
Consider the general term in the Fourier series for f (t), namely
(−1)n+1
2 sin nt.
n
Comparison with the particular solution just found thus generates a forcing term (with K =
2(−1)n+1 /n, ω = n),
2(−1)n+1 cos nt
,
Ω2 − n2
provided Ω2 6= n2 . We now sum these terms over n to give the forced solution
X∞
(−1)n+1
x=2 cos nt.
n=1
Ω2 − n2
The system will resonate if Ω is close to any positive integer n.

26.26. Multiply both sides of
X∞
1
f (t) = a0 + (an cos ωt + bn sin ωt)
2 n=1
by f (t) and integrate term-by-term over the interval (− 12 T, 12 T ):

Z 1
2T
f (t)2 dt
− 21 T
Z 1 ∞
Ã Z 1 Z 1
!
1 2T X 2T 2T
= a0 f (t)dt + an f (t) cos ωtdt + bn f (t) sin ωtdt
2 − 12 T n=0
1
2T
1
2T
∞
1 2 X 2
= a0 + (an + b2n ),
2 n=1
using (26.9) for the Fourier coefficients.

(a) With T = π and the odd function
½
−1 (− 12 π < t ≤ 0)
f (t) = ,
1 (0 < t ≤ 12 π)
the Fourier coefficients an = 0 and
Z 1
2π
4 2(1 − (−1)n )
bn = sin 2ntdt = .
π − 12 π nπ
Hence
4 4 4
b1 =
, b2 = 0, b3 = , b4 = 0, b5 = ,....
π 3π 5π
Hence, using Parseval’s identity,
∞ ∞
1 2 X 2 16 X 1
a0 + (an + b2n ) =
2 n=1
π 2 m=1 (2m + 1)2
Z 1 Z 1
2 2π 2 2π
= f (t)2 dt = dt
π − 12 π π − 12 π
= 2.
Therefore
∞
X 1 π2
= .
m=1
(2m + 1)2 8
44
(b) The Fourier coefficients of f (t) = t, (−π < t ≤ π) are
2(−1)n+1
an = 0, bn =
n
(see Problem 26.1b). Using Parseval’s identity
∞
X X∞ Z
1 1 π 2 1 3π 2π 2
(a2n + b2n ) = 4 2
= t dt = [t ]−π = .
n=1 n=1
n π −π 3π 3
Hence
X∞
1 π2
2
= .
n=1
n 6
26.27. As far as the Laplace transform is concerned, the transform is that of the function f (t)H(t),
which is zero for t < 0. The Laplace transforms of cos nωt and sin nωt are
s nω
L{cos nωt} = , L{sin nωt} = .
s2 + n 2 ω 2 s2 + n 2 ω 2
Hence
∞
a0 X an s + bn nω
F (s) = L{f (t)} = + .
2s n=1 s2 + n2 ω 2
The Fourier coefficients of the function
½
−t2 (−π < t < 0)
f (t) =
t2 (0 ≤ t ≤ π)
are
2
an = 0, {2 + (−1)n [n2 π 2 − 2]}.
bn =
n3 π
Hence the Fourier transform of the function f (t)H(t) is
∞
2 X 2 + (−1)n [n2 π 2 − 2]
F (s) = .
π n=1 n2 (s2 + n2 )
26.28. A radio wave is defined by
x(t) = a cos ωt cos ω0 t,
where ω0 is very much greater than ω. Using the product formula in Appendix B(d),
x(t) = 12 a[cos(ω0 − ω)t + cos(ω0 + ω)t].
The figure shows the wave x(t) = cos t cos(10t) for −20 < t < 20.
x
1
0.5
t
-20 -10 10 20
-0.5
-1
45
(a) With ω = 500 and ω0 = 100001, the sums and differences are
ω0 + ω = 100501, ω0 − ω = 99501.
The greatest common divisor of these numbers is 1: hence the period of x(t) is 2π.
(b) If ω = p/q and ω0 = r/s, then
x(t) = a cos(pt/q) cos(rt/s).
Then x(t) is periodic with period T if T is the smallest values for which x(t + T ) = x(t) for all t.
For the given function
µ ¶ µ ¶
p(t + T ) r(t + T )
x(t + T ) = a cos cos .
q s
This equals x(t) if pT /q and rT /s are integer multiples of 2π. Since q and s have only the common
divisor 1 the smallest value of T is 2πqs, which is the period of x(t).
As the sum of two waves
µ ¶ µ ¶ · µ ¶ µ ¶ ¸
pt rt a r p r p
x(t) = a cos cos x = cos − t + cos + t .
q s 2 s q s q
This is the Fourier cosine series of x(t) over the period 2πsq.
√
(c) Given x1 (t) = cos t cos 2t, then
√ √
x(t + T ) = cos(t + T ) cos( 2t + 2T ).
√
In this case
√ we require T and 2T to be integer multiples of 2π which is impossible since by
definition 2 can never be equal to the ratio of two integers.
26.29. (a) If m = n then
Z 1
2T
Z 1
2T
i2πnf0 t −i2πmf0 t
e e dt = dt = T.
− 21 T − 12 T
If m 6= n and f0 = 1/T , then

Z 1 Z 1 · ¸ 12 T
2T 2T T e2πit(n−m)/T
i2πnf0 t −i2πmf0 t i2π(n−m)f0 t
e e dt = e dt =
− 12 T − 12 T 2πi(n − m) − 12 T
T sin 2π(n − m)
= =0
π(n − m)
(b) From (26.18a),

∞
X
xP (t) = Xn ei2πnf0 t .
n=−∞
Multiply both sides by e−i2πN f0 t and integrate over one period, from t = − 12 T to 12 T :
Z 1
2T
∞
X
xP (t)ei2πN t/T dt = Xn ei2πnt/T e−i2πN t/T dt = XN T = XN /f0 .
− 12 T n=−∞
Result (26.18) follows.

26.30. For xP (t) = t/T , the coefficients of the two-sided Fourier series are, from (26.18), given by
Z T
f0 1
X0 = tdt = T,
T 0 2
46
Z T
f0 i
Xn = te−i2πnf0 t dt = .
T 0 2nπ
Hence
∞
t 1 i X 1 i2πnt/T
= + e .
T 2 2π n=−∞ n
Chapter 26: Fourier transforms
27.1 The real and imaginary parts of the integral

Z ∞
2 e−t e2πif t dt
0
give the Fourier cosine and sine transforms Xc (f ) and Xs (f ) respectively of e−t :
Z ∞
2
Xc + iXs = 2 et(−1+2πif ) dt = [et(−1+2πif ) ]∞
0
0 −1 + 2πif
2 2(1 + 2πif ) 2(1 + 2πif )
= = =
1 − 2πif (1 − 2πif )(1 + 2πif ) 1 + (2πf )2
2 + 4πif
=
1 + 4π 2 f 2
Therefore
2 4πf
Xc = , Xs = .
1 + 4π 2 f 2 1 + 4π 2 f 2
The inverse is Z ∞
4πf
I(t) = 2 sin(2πf t)df.
0 1 + 4π 2 f 2
I(t) is zero at t = 0. This is connected with (27.9). I(t) is an odd function on −∞ < t < ∞;
I(t) = e−t for t > 0 and I(t) = −e−t for t < 0, so it jumps in value at t0 = 0, the mean of the
values at either side of t = 0 being zero.
27.2. The cosine transform of
½
1 − t, 0≤t≤1
x(t) =
0, t>1
is Z Z
∞ 1
Xc (f ) = 2 x(t) cos(2πf t)dt = 2 (1 − t) cos(2πf t)dt.
0 0
By integrating by parts we obtain
2 4 sin2 (πf ) sin2 (πf )

Xc = [1 − cos(2πf )] = = = sinc 2 f.
(2πf )2 (2πf )2 π2 f 2
The inverse of Xc (f ) is
Z ∞
sin2 (πf )
2 cos(2πf t)df.
0 π2 f 2
This is an even function of t; it delivers (1 − t) for t > 0, and (1 + t) for t < 0, and is continuous
at t = 0, taking the value 1. Therefore put t = 0 into the inverse integral; we obtain
Z ∞
sin2 (πf )
2 df = 1.
0 π2 f 2
47
Now put πf = u; then the equality becomes
Z ∞ Z ∞
sin2 u du sin2 u 1
2 2
= 1, or 2
du = π.
0 u π 0 u 2
2
27.3. Here, x(t) = e−t , and we require the cosine transform Xc (f ), where
Z ∞
2
Xc (f ) = 2 e−t cos(2πf t)dt.
0
By differentiating with respect to f under the integral sign

Z ∞
dXc 2
= −4π (te−t ) sin(2πf t)dt. (i)
df 0
2 2
Integrate by parts, putting into eqn (17.8) u = sin(2πf t), and dv/dt = te−t so that v = − 12 e−t :
½h ³ í∞ Z ∞ ¾
dXc 2 2
= −4π sin(2πf t) − 12 e−t − (− 12 e−t )[2πf cos(2πf t)]dt
df 0 0
Z ∞
2
= −4π 2 f e−t cos(2πf t)dt = −2π 2 f Xc .
0
The differential equation

dXc
= −2π 2 f Xc (ii)
df
is separable: Z Z
dXc
= −2π 2 f df,
Xc
leading to the general solution
2
f2
Xc (f ) = Ke−π
where K is an arbitrary constant. The inverse integral is therefore
Z ∞
2 2
x(t) = 2K e−π f cos(2πf t)df (iii)
0
for some value of K.

2
To find K, put t = 0 into (iii). We have x(t) = e−t , so
Z ∞ Z √
2 2 2K ∞ −u2 2K π K
x(0) = 1 = 2K e−π f df = e du = =√ ,
0 π 0 π 2 π
√
from the standard integral given. Therefore K = π, and
2 √ 2 2
e−t ↔ πe−π f .
27.4. (a) x(t) is an even function, that is x(t) = x(−t). The Fourier transform is given by
Z ∞ Z ∞ Z 0
−2πif t −2πif t
F[x(t)] = x(t)e dt = x(t)e dt + x(t)e−2πif t dt.
−∞ 0 −∞
Change the variable in the second integral by putting t = −t , and put x(−t0 ) = x(t0 ) (evenness
0
property):
Z ∞ Z ∞
0
F[x(t)] = x(t)e−2πif t dt + x(t0 )e2πif t dt0
0 0
Z ∞
e2πif t + e−2πif t
= 2 x(t) dt
2
Z0 ∞
= 2 x(t) cos(2πf t).dt (i)
0
48
The right-hand side of (i) is the Fourier cosine transform Xc (f ).
2
(b) The function e−t is even, so eqn (i) applies. We then have
2 √ 2 2
F[e−t ] ≡ Xc (f ) = πe−π f .
By the time-scaling rule (27.18b),

µ ¶ r
2 1 2 1 f π −π2 f 2 α
F[e−αt ] ≡ F[e−(α 2 t) ] = 1 Xc 1 = e .
α2 α 2 α
2 2
(Notice that if α = π, the result becomes symmetrical: e−πt ↔ e−πf .
27.5. With x(t) an odd, real function
Z ∞ Z ∞ Z 0
−2πif t −2πif t
X(f ) = F[x(t)] = x(t)e dt = x(t)e dt + x(t)e−2πif t dt.
−∞ 0 −∞
Change the variable in the second integral to t0 = −t, and use the oddness property, x(−t0 ) =
−x(t0 ):
Z ∞ Z ∞
0
X(f ) = x(t)e−2πif t dt − x(t0 )e2πif t dt0
Z0 ∞ 0
= x(t){e−2πif t − e2πif t }dt

0
Z ∞
= −2i x(t) sin(2πf t)dt = −iXs (f )
0
where Xs (f ) is the sine transform of x(t).

27.6. See the answer to Problem (27.4b), in the special case of α = π.
27.7. To derive a form of Fourier transform pair that is an alternative to that in eqns (27.8).
Change the frequency variable f in eqns (27.8) to a new variable ω through the relation f =
ω/(2π) (ω then takes the meaning of circular, or angular frequency ). Then (27.8a) becomes
Z ∞ ³ ´
ω dω
x(t) = X eiωt , (i)
−∞ 2π 2π
and (27.8b) becomes

³ω´ Z ∞
X = x(t)e−iωt dt. (ii)
2π −∞
Now put
1 ³ω´
√ X = X1 (ω).
(2π) 2π
Then (i) and (ii) become Z ∞
1
x(t) = √ X1 (ω)eiωt dω,
(2π) −∞
and Z ∞
1
X1 (ω) = √ x(t)e−iωt dt,
(2π) −∞
as required
27.8. (This is an alternative approach from that in Problem 27.4.) In (27.8b), change the variable
from t to (−t) and use the evenness property of x(t). We then have two alternative forms of the
equation: Z Z
∞ ∞
X(f ) = x(t)e−2πif t dt and X(f ) = x(t)e2πif t dt.
−∞ −∞
49
Add the two versions and divide by 2:
Z Z ∞
1 ∞
X(f ) = x(t)(e2πif t + e−2πif t )dt = x(t) cos(2πf t)dt
2 ∞ −∞
Z ∞
= 2 x(t) cos(2πf t)dt, (i)
0
since the integrand is even in t. Also X(f ) is even (in f ), so to obtain the inverse, start with
(27.8a), and carry out a similar process of changing the variable from f to (−f ), obtaining
Z ∞
x(t) = 2 X(f ) cos(2πf t)df. (ii)
0
Equations (i) and (ii) take a real form (though x(t) may be complex), recognizable as a Fourier
cosine transform and its inverse.
27.9. The function x(t) is odd (that is, x(−t) = −x(t)). In (27.8b) change the variable from t to
(−t) and use the oddness property. We now have two alternative forms for X(f ):
Z ∞ Z ∞
−2πif t
X(f ) = x(t)e dt and X(f ) = − x(t)e2πif t dt.
−∞ −∞
Add the two versions and divide by 2; we obtain

Z Z ∞
1 ∞
X(f ) = x(t)(e−2πif t − e2πif t )dt = −i x(t) sin(2πf t)dt
2 −∞ −∞
Z ∞
= −2i x(t) sin(2πf t)dt, (i)
0
since the integrand is an even function of t.

Notice that X(f ) is an odd function of f (that is, X(−f ) = −X(f )), and consider the in-
verse integral (27.8a), changing the variable (as before) from f to (−f ) so as to obtain a second
representation of x(t):
Z ∞ Z ∞
2πif t
x(t) = X(f )e df and x(t) = − X(f )e−2πif t df.
−∞ −∞
Add the two versions and divide by 2:

Z Z ∞
1 ∞
x(t) = X(f )(e2πif t − e−2πif t )dt = i X(f ) sin(2πf t)df
2 −∞ −∞
Z ∞
= 2i X(f ) sin(2πf t)dt (ii)
0
Now define a function X1 (f ) by writing iX(f ) = X1 (f ). Then (i) and (ii) become the pair
Z ∞
X1 (f ) = 2 x(t) sin(2πf t)dt,
0
and Z ∞
x(t) = 2 X1 (f ) sin(2πf t)df.
0
This is the sine transform pair (eqn (27.5)), and is a real form (though x and X may be complex).
27.10. Time-scaling rule (27.18b). Let x(t) ↔ X(f ), and consider the transform of x(At):
Z ∞
F[x(At)] = x(At)e−2πif t dt.
−∞
50
Change the variable of integration by putting At = t0 . If A > 0, we obtain
Z µ ¶
1 ∞ 0 1 f
x(t0 )e−2πif t /A dt0 = X .
A −∞ A A
If A < 0, we obtain (note the inversion of the limits in the integral):

Z µ ¶
1 −∞ 0 −2πif t0 /A 0 1 f
x(t )e dt = − X .
A ∞ A A
Both results can be combined into one by writing

µ ¶
1 f
F[x(At)] = X .
|A| A
Time-delay rule (27.18c).

Z ∞
F[x(t − B)] = x(t − B)e−2πif t dt.
−∞
Put t − B = t0 . The integral becomes

Z ∞
0
x(t0 )e−2πif (t +B) dt0 = e−2πif B X(f ).
−∞
27.11. (a) By the time-delay rule (27.18c), with B = 12 ,
F[Π(t − 21 )] = e−2πi(1/2)f sinc f = e−πif sinc f.
(b) To confirm this directly: Π(t − 12 ) is zero when t − 1

2 < − 12 and t − 1
2 > 12 , so
Z 1
1 −2πif t 1
F[Π(t − 21 )] = (1)e−2πif t dt = −[e ]0
0 2πf
e−2πif − 1 (eπif − e−πif )e−πif sin πf
= − = = e−πif
2πif 2πif πf
−πif
= e sinc f.
(c) The graph of x(t) is shown in the figure. It is obtained by moving a centrally-placed rectangle
of width c given by Π(t/c) a distance b > 21 c to the left, and changing its sign to produce A, and
a distance b to the right to produce B.
1
c c B
-b- -b -b+
2 2 t
c b c
b- b+
A 2 2
-1
Therefore µ ¶ µ ¶
t+b t−b
x(t) = −Π +Π .
c c
51
By the time-delay rule (28.18c), with B = ±b,
· µ ¶¸
t
F[x(t)] = (−e2πibf + e−2πibf )F Π
c
= (−e2πibf + e−2πibf )[c sinc (cf )]
(by the time-scaling rule (28.18b) with B = 1/c)
= −2ic sin(2πbf )sinc (cf ).
27.12. (a) Given that Λ(t) ↔ sinc 2 f , we obtain from the time-scaling rule (27.18b) with A = 2:
Λ(2t) ↔ 12 sinc 2 ( 21 f ).
(b) In general, suppose that x(t) ↔ X(f ). To obtain F[x(2t − 3)] we can use the time-scaling rule
(27.18b), then the time-delay rule (27.18c), as follows:
x(2t) ↔ 12 X( 12 f ) (time-scaling).
and
x(2t − 3) = x(2[t − 23 ]) ↔ e−2πi(3/2)f { 12 X( 21 f )} = 12 e−3πif X( 21 f ).
For the particular case of x(t) = Λ(t), X(f ) = sinc 2 f , so that we have
Λ(2t − 3) ↔ 21 e−3πif sinc 2 ( 12 f ).
27.13. (a) Frequency shift rule (27.18e). From the definition of F

Z ∞
F[x(t)e2πiDt ] = x(t)e2πiDt e−2πif t dt
−∞
Z ∞
= x(t)e−2πi(f −D)t dt = X(f − D)
−∞
where X(f ) = F[x(t)].
(b) F[x(t)e±2πif0 t ] = X(f ∓ f0 ), (from (a))
(c) From (b)

1
F[x(t) cos(2πf0 t)] = F[x(t){e2πif0 t + e−2πif0 t }]
2
1
= {X((f − f0 ) + X(f + f0 )}, .
2
1
F[x(t) sin(2πf0 t)] = F[x(t){e2πif0 t − e−2πif0 t }]
2i
1
= {X((f − f0 ) − X(f + f0 )} .
2i
(d) F[Π(t)] = sinc f . By the time-scaling rule (27.18b) with A = 12 ,
F[Π( 21 t)] = 2sinc 2f.
Apply the results in (c); then
F[Π( 12 t) cos(2πf0 t)] = sinc 2(f − f0 ) + sinc 2(f + f0 ),
and
F[Π( 12 t) sin 2πf0 t)] = −i{sinc 2(f − f0 ) − sinc 2(f + f0 )}.
52
27.14. (a) To obtain F[sinc 2 t], given that F[Λ(t)] = sinc 2 f . In the duality rule (27.18b) put
X(t) = sinc 2 t and x(−f ) = Λ(−f ) = Λ(f ) since Λ is an even function. We obtain F[sinc 2 t] =
Λ(f ).
This result may alternatively be obtained without using (27.18b). From the basic transform
pair, starting with the inverse relation (27.8a), we know that
Z ∞
e2πif t sinc 2 f df = Λ(t).
−∞
This equation is an identity, so if we interchange the letters f and t in it, the result remains true:
Z ∞
e2πif t dt = Λ(f )
−∞
for all values of the parameter f . Therefore

Z ∞
e2πi(−f )t sinc 2 tdt = Λ(−f ),
−∞
or Z ∞
e−2πif t sinc 2 dt = Λ(−f ) ≡ Λ(f )
−∞
since Λ is even. Therefore F[sinc 2 t] = Λ(f ).

(b) Given that F[sinc 2 t] = Λ(f ), the time-scaling rule (27.18b) with A = a gives
µ ¶
2 1 f
F[sinc (at)] = Λ .
a a
Now write sinc 2 (at + b) ≡ sinc 2 [a(t + (b/a))], and use the time-delay rule (27.18c) with B = −b/a:
µ ¶
2 1 2πibf /a f
F[sinc (at + b)] = e Λ .
a a
27.15. (a) To prove the differentiation rule (27.18h). From (27.8a),

Z ∞
x(t) = e2πif t X(f )df.
−∞
By differentiating with respect to t under the integral sign:

Z ∞
dx(t)
= e2πif t {2πif X(f )}df,
dt −∞
that is
dx(t)
↔ (2πif )X(f ).
dt
As each subsequent differentiation introduces another factor 2πif , we have
dn x(t)
↔ (2πif )n X(f ).
dtn
(This process can only be carried out so long as the functions on the right continue to have valid
Fourier integrals.)
(b) Given e−|t| ↔ 2/(1 + 4π 2 f 2 ), to find the inverse of if /(1 + 4π 2 f 2 ). The differentiation rule in
(a), with x(t) = e−|t| , gives
d −|t| 4πif
(e ) =↔ . (i)
dt 1 + 4π 2 f 2
53
Also ½
d −|t| et for t < 0
(e ) = ,
dt −e−t for t > 0
that is,
d −|t|
(e ) = −e−|t| sgn (t),
dt
where sgn (t) stands for the sign of t (see Section 1.4). At t = 0 the derivative does not exist.
Therefore, from (i), the inverse of if /(1 + 4π 2 f 2 ) is
1 −|t|
− e sgn (t).
4π
27.16. Given (eqn (27.17)) that 1/(α + 2πif ) is the transform of the function h(t) defined by
½
0, t<0
h(t) =
e−αt , t > 0
to obtain F[e−α|t| ]. It is easy to show that

e−α|t| = h(t) + h(−t),
by sketching diagrams of f (t) and f (−t). From the time-reversal rule (27.18b),
F[e−α|t| ] = F[h(t)] + F[h(−t)]
1 1 2α
= + = 2 .
α + 2πf α + 2π(−f ) α + 4π 2 f 2
27.17. (a) H(t) is the Heaviside unit function, eqn (1.13).

1
F[e−αt H(t)] = X(f ) = ,
α + 2πif
from eqn (27.17). From the modulation rules (27.18f), with K = β/(2π),
½ µ ¶ µ ¶¾
1 β β
F[e−αt cos(βt)H(t)] = X f− +X f +
2 2π 2π
( )
1 1 1
= β
+ β
(i)
2 α + 2πi(f − 2π ) α + 2πi(f + 2π )
and ( )
−αt 1 1 1
F[e sin(βt)H(t)] = β
− β
(ii)
2i α + 2πi(f − 2π ) α + 2πi(f + 2π )
(b) In the case of F[e−αt cos(2πf0 t + φ)H(t)], expand the cosine term, obtaining
F[e−αt cos(2πf0 t + φ)H(t)] =
cos φF[e−αt cos(2πf0 t)H(t)] − sin φF[e−αt sin(2πf0 t)H(t)].
The results (i) and (ii) above, with β = 2πf0 , apply to the two terms, giving
½ ¾
1 1 1
cos φ + −
2 α + 2πi(f − f0 ) α + 2πi(f + f0 )
½ ¾
1 1 1
sin φ −
2i α + 2πi(f − f0 ) α + 2πi(f + f0 )
½ ¾
1 eiφ e−iφ
= +
2 α + 2πi(f − f0 ) α + 2πi(f + f0 )
(α cos φ − 2πf0 sin φ) + 2πif cos φ
= .
α2 − 4π 2 (f 2 − f02 ) + 4πiαf
54
27.18. By (27.27), the notation x1 (t) ∗ x2 (t) means
Z ∞
x1 (τ )x2 (t − τ )du.
−∞
Put x2 = H(t) and x1 (t) = x(t)H(t). Then x2 (τ ) is zero for τ < 0 and x(t − τ ) is zero for t − τ < 0,
or τ > t. Therefore Z t
x1(t) ∗ x2 (t) = x(τ )dτ.
0
27.19. (a) Z ∞
−t −t
e H(t) ∗ e H(t) = e−u e−(t−u) H(u)du = I(t), say.
∞
If t > 0, H(u)H(t − u) is zero for u < 0 and u > t. If t < 0, then H(u)H(t − u) is zero for all u.
Therefore ½ Rt Rt
e−u e−(t−u) du = e−t 0 du = te−t for t > 0
I(t) = 0
zero for t < 0.
(b) From (27.17), F[e−t H(t)] = 1/(1 + 2πif ). From (a)
F[te−t H(t)] = F[e−t H(t) ∗ e−t H(t)].
Therefore, by the convolution theorem (27.28),

1
F[te−t H(t)] = F[e−t H(t)]F[e−t H(t)] = . (i)
1 + 2πif )2
(c) H(αt) = H(t) for α > 0, so the time-scaling rule (27.18b) applied to (i) gives
1 1 α
F[αte−αt H(t)] = = .
α (1 + 2πif /α)2 (α + 2πif )2
Therefore, dividing through by α,
1
F[te−αt H(t)] = . (ii)
(α + 2πif )2
(d) From (27.17), F[e−αt H(t)] = 1/(α + 2πif ) for α > 0

Z ∞
df
e2πif t = e−αt H(t).
−∞ α + 2πif
Differentiate both sides with respect to α; we obtain
Z ∞
df
− e2πif = −te−αt H(t).
−∞ (α + 2πif )2
Therefore
1
te−αt H(t) ↔ ,
(α + 2πif )2
as already found in (ii)
27.20. By (27.15), F[Π(t)] = sinc f . From the time-delay rule (27.18c):
F[t ± 12 ] = e±πif sinc f.
Therefore by the convolution theorem
F[Π(t − 12 ) ∗ Π(t + 12 )] = F[Π(t − 12 )]F[Π(t + 12 )]

= (e−iπf sinc y)(eiπf sinc y) = sinc 2 f
55
(b) By proceeding as in (a), we obtain
F[Π(t − a)] ∗ Π(t − b)] = (e−2πiaf sinc f )(e−2πibf sinc f )

= e−2πi(a+b)f sinc 2 f
(c) In the definition (27.27a) put x1 (t) = Π(t), x2 (t) = Π( 21 t), and use the second form in (27.27a)
for simplicity. Then put
Z ∞
Π(t) ∗ Π( 12 t) = Π(t − u)Π( 21 u)du = (t). (i)
−∞
Π(t − u) is zero for t − u < − 21 and for t − u > 12 , and Π( 12 u) is zerofor u < −1 and for u > 1. By
sketching a diagram as in Example 27.10 we can establish the effective limits of the integral in (i)
in terms of t.
u
1
1

2
t
3 -1 1 1 3
-
2 2 1 2 2
-
2
-1
Also F (t) is an even function, so we only have to calculate for t ≥ 0 and the values for t < 0
will echo those values. The diagram has boundaries u = ±1 and u = t ± 12 . The values of F (t) are
given in the following table.
Range of t F (t)
t < − 32 zero
R t+ 12
− 32 ≤ t ≤ − 12 du = t + 3
2
R−1
t+ 12
− 12 ≤ t ≤ 1
2 1 du = 1
1 3
Rt−
1
2
2 <t≤ 2 t− 12
du = 1 − (t − 12 ) = −t + 3
2
t > 12 zero
27.21. By Section 27.9, the energy E is given by

Z ∞ Z ∞
−αt 2
E = {e H(t)} dt = e−2αt H(t)dt, (since [H(t)]2 = H(t))
−∞ −∞
Z ∞
1
= e−2αt dt = .
0 2α
The transform (frequency distribution) of e−αt is 1/(α + 2πf ) = X(f ), say, by (27.17). The
energy associated with the frequency range −f0 ≤ f ≤ f0 in the expression (27.33) is
Z f0 Z f0 Z f0
2 df df
|X(f )| df = 2
= 2 + 4π 2 f 2
−f0 −f0 |α + 2πif | f0 α
Z 2πf0 /α
α 1 du
= (after putting f = αu/(2π))
2π α2 −2πf0 /α 1 + u2
µ ¶
1 2πf0 /α 1 2πf0
= [arctan u]−2πf0 /α = arctan .
2πα πα α
56
27.22. The problem asks you to show that if x(t) is zero for t < − 12 T and t > 21 T , then q
qT (t)∗x(t)
is periodic with period T . By the convolution theorem
F[qqT (t) ∗ x(t)] = F[qqT (t)]F[x(t)]. (i)
Here Z 1
2T
F[x(t)] = e−2πif t x(t)dt,
− 12 T
and from (27.32), since f0 = 1/T ,

∞
1 1 X
F[qqT (t)] = qq T1 (f ) = δ(f − (n/T )).
T T n=−∞
Therefore (i) becomes

∞
Ã Z 1
!
X 1 2T
−2πif t
F[qqT (t)] = e x(t)dt δ(f − (n/T ))
n=−∞
T − 12 T
which, by (27.21), is the Fourier transform of the T -periodic extension of x(t) to all values of t
(remember that 1/T = f0 in (27.21)).
27.23. Take the Fourier transform of the differential equation
d2 x 1
−x= ,
dt2 1 + t2
using the differentiation rule (27.18h) to convert the derivative:
· 2 ¸ · ¸
d x 2 1
F − x = (2πif ) X(f ) − X(f ) = F .
dt2 1 + t2
Therefore · ¸ · ¸
1 1 1 −|t| 1
X(f ) = − F = − F[e ]F
1 + 4π 2 f 2 1 + t2 2 1 + t2
(by Example 27.3). This is the product of two transforms, so by the convolution theorem the
inverse x(t) of X(f ) can be written as the convolution integral
Z ∞
1 e−|t−u|
x(t) = − du.
2 −∞ 1 + u2
(There are, of course, two linearly independent solutions but only one has a Fourier transform.)
27.24. Π(t) is continuous at t = 0 taking the value 1. Therefore we may put t = 0 into the defining
Fourier integral Z ∞
e−2πif t sinc t dt = Λ(t).
−∞
We obtain Z ∞ Z ∞
sinc tdt = 1 = 2 sinc t dt,
−∞ 0
R∞
since sinc t is even. Therefore 0
sinc t dt = 12 .
Z ∞
2 2
(b) e−2πif t e−πt dt = e−πf , (see Problem 27.6)
−∞
57
Differentiate both sides of the equation with respect to f :
Z ∞
2 2
−2πi e−2πif t te−πt dt = −2πf e−πf ,
−∞
2 2
so F[te−πt ] = −if e−πf .
27.25. By the convolution theorem and the result F[sinc t] = Π(f ),
F[sinc t ∗ sinc t] = F[sinc t]F[sinc t] = Π2 (f ) = Π(f ).
Similarly, and using the time-delay rule (27.18c),
F[sinc (t − a) ∗ sinc (t + a)] = (e−2πiaf Π(f ))(e2πiaf Π(f )) = Π2 (f ) = Π(f ).
27.26. (a) Moving average of g(t). To show that

Z t+ 12 τ
1
gτ (t) = g(u)du
τ t− 12 τ
can be written as µ ¶
1 t
Π ∗ g(t).
τ τ
The convolution is defined by (27.27):
µ ¶ Z µ ¶
1 t 1 ∞ t−u
Π ∗ g(t) = Π g(u)du. (i)
τ τ τ −∞ τ
Π[(t − u)/τ ] is zero when [(t − u)/τ ] < − 12 or [(t − u)/τ ] > 12 ; that is, when u > t + 12 τ or u < t − 12 τ .
Therefore the integral is zero for these ranges, and equals g(u) elsewhere, and (i) becomes
Z t+ 12 τ
1
g(u)du = gτ (t).
τ t− 12 τ
(b) The moving average of Π(t) over intervals of length τ is given by Fτ (t), where
Z ³u´
1 ∞
Fτ (t) = Π(t − u)Π du, (i)
τ −∞ τ
(using (a) with the convolution form (27.26b), with Π(t) for x1 (t) and x2 (t) = Π(t/τ ), the averaging
function). Although this looks more complicated than the direct definition of gτ (t), it permits
evaluation on the lines of Example 27.10 and Fig. 27.14, which allows extensions to other functions
also.
The boundary of the region where the integrand is nonzero (it takes the value 1) is shown by
the parallelogram in the figure. The formulae giving the effective limits of integration in (i) change
as t passes over the points A0 , B 0 , B, A. The figure shows a case where τ > 1, together with
the values of t at these critical points. When τ < 1 the construction is the same and the formula
is the same, but B will be on the negative side of the u axis and B 0 on the positive side. Both
possibilities are covered by expressing the results in the following way:
R t+ 12
− 12 (τ + 1) ≤ t ≤ − 21 |τ − 1| Fτ = 1
τ − 12 τ
du = τ1 {(t + 12 ) − (− 12 τ )} = τ1 (t + 12 τ + 21 );
R 1
2τ
− 12 |τ − 1| ≤ t ≤ 12 |τ − 1| Fτ (t) = 1
τ − 12 τ du = 1;
1 1 1
R 12 τ 1 1 1
2 |τ − 1| ≤ t ≤ 2 (τ + 1) Fτ (t) = τ t− 12 du = τ (−t + 2 τ + 2 );
(note that, predictably, this is an even function of t.) Fτ (t) = 0 elsewhere.
58
Τ
u=
1 2
Τ
2
1

2
1
2
t+
u=
A’ B’ B A
O
1 1 1 1 1 1
- HΤ+1L - - HΤ-1L HΤ-1L HΤ+1L
2 2 2 2 2 2
1
2
t-
u=
1
-
2
1
- Τ
Τ 2
u=-
2
27.27. The definition of the convolution is

Z ∞ µ ¶
w+b
F (t) = g(w + b)Π qqT (t − w − a)dw
−∞ τ
with τ ≤ T . Put w − a = u:
Z ∞ µ ¶
u+a−b
F (t) = g(u + a − b)Π qqT (t − u)du
−∞ τ
X∞ Z ∞ ½ µ ¶¾
u+a−b
= g(u + a − b)Π δ(t − u − nT )du (i)
n=−∞ −∞
τ
This resembles the form in Example 27.11, with the function x(t) displaced along the u axis by a
distance a − b to the left, as in the figure.
A B
1 -a+b 1
-a+b+ T -a+b- T
2 2
T
The condition τ ≤ T ensures that the nonzero segment of the function does not extend beyond
the range AB: −a + b − 12 T ≤ u ≤ −a + b + 12 T . The argument in Example 27.11 can be adapted
to fit this case. For a given t, there is always exactly one value of n, say N , such that a single
impulse in (i) lies in the interval AB of length T . But if n is any integer
(t − nT ) + (N + n)T ≡ t − N T,
59
so (from the sifting property) the value picked out is reproduced at t − nT , for all n. Therefore
the convolution represents the periodic extension, period T , of g(u + a − b) as depicted.
27.28. The function Z ∞
h(t) = Π(t) ∗ Λ(t) = Π(u)Λ(t − u)du
−∞
from (27.27). Π(u) is zero for u < 21 and u > 12 , and Λ(t − u) is zero for t − u < −1 and t − u > 1.
To obtain the effective limits of integration, follow the procedure leading to Fig. 27.14 in the book.
The boundaries of the region of the (t, u) plane on which the integrand is nonzero are u = ± 12 and
t − u = ±1, shown in Fig. 22. (The definition of Λ(u) is given in Example 27.10.) We have, for
u
1
u=
2
0.4
1
t+
0.2
u=
-1.5 -1 -0.5 0.5 1 1.5

-0.2
1
t-
u=
-0.4
1
u=-
2
t ≤ 32 h(t) = 0;
R t+1
− 32 ≤ t ≤ − 12 h(t) = − 1 (1 + u)du = 12 t2 + 2t + 15 8 ;
R 12 2
− 12 ≤ t ≤ 0 h(t) = − 1 (1 + u)du = 1;
1
R 12
0≤t≤ 2 h(t) = −2 1 (1 − u)du = 1;
1 3
R 12 2
2 ≤t≤ 2 h(t) = t−1 (1 − u)du = 12 t2 + 2t + 15
8 ;
t ≥ 32 h(t) = 0.
By the convolution Theorem (27.28),
F[h(t)] = F[Π(t)]F[Λ(t)] = sinc f sinc 2 f = sinc 3 f
(by Example 27.10). Therefore, since h(t) is continuous (no jumps) the inverse of F[h(t)] for all t
is given by Z ∞
e2πif t sinc 3 f df = h(t).
−∞
Put t = 0: Z ∞
sinc 3 f df = h(0) = 1.
−∞
27.29. (a)
Z ∞
x(t) ∗ {Ay(t) + Bz(t)} = x(u){Ay(t − u) + Bz(t − u)}du
−∞
Z ∞ Z ∞
= A x(u)y(t − u)du + B x(u)z(t − u)du
−∞ −∞
= Ax(t) ∗ y(t) + Bx(t) ∗ z(t).
(b) Quoted in (27.26) in the text: here we check it directly:

Z ∞
x(t) ∗ y(t) = x(u)y(t − u)du.
−∞
60
Change the variable to v = t − u:
Z −∞
x(t) ∗ y(t) = − x(t − v)y(v)dv
∞
Z ∞
= y(v)x(t − v)dv = y(t) ∗ x(t)
−∞
(c) We shall consider the Fourier integrals of the terms, so as to avoid double integrals (Chapter 32).
By the convolution theorem (27.28), used twice,
F[x(t) ∗ {y(t) ∗ z(t)}] = F[x(t)]F[y(t) ∗ z(t)]

= F[x(t)]F[y(t)]F[z(t)].
The same product appears if the group is bracketed differently (or placed in any order).
61

Elementary methods, diﬀerentiation, complex numbers

Uploaded by

Copyright:

Available Formats

You might also like

Elementary methods, diﬀerentiation, complex numbers

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Elementary methods, diﬀerentiation, complex numbers

Uploaded by

Copyright:

Available Formats

PART I: Elementary methods, differentiation, complex numbers

Chapter 1: Standard functions and techniques, 1

Chapter 1: Standard functions and techniques

1.1. (a) y = x4 for −1.5 ≤ x ≤ 1:

Figure 1: Problem 1.1a

(b) y = x(1 − x) for −1 ≤ x ≤ 2:

Figure 2: Problem 1.1b

Figure 3: Problem 1.1c

Figure 4: Problem 1.1d

Figure 5: Problem 1.1e

Figure 6: Problem 1.1f

(e) y = |x| + |x − 3| + |x + 2| for −3 ≤ x ≤ 4:

Figure 7: Problem 1.1g

Figure 8: Problem 1.2

x2 + (2x + 1)2 − 2x + 2(2x + 1) − 4 = 0, or 5x2 + 6x − 1 = 0.

The points of intersection are

1.060, 0.993, 1.011, 0.896, 1.124.

1.11. (a) x = H(t + 1) − H(t − 1).

Figure 9: Problem 1.11a

(b) x = sgn(1 + t) + sgn(1 − t).

Figure 10: Problem 1.11b

(c) x = tH(t − 1).

Figure 11: Problem 1.11c

Figure 12: Problem 1.11d

1.13. (a) 16 π radians; (b) 23 π radians.

Figure 13: Problem 1.21

1.22. (a) x = 12 ln 3; (b) x = 13 e2 ; (c) x = e−3 ; (d) x = − 31 ln 3;

Figure 14: Problem 1.25a

Figure 15: Problem 1.25b

Figure 16: Problem 1.25c

where −1 < x < 1. Hence

1.27. From triangle ABC

Figure 17: Problem 1.25d

Figure 18: Problem 1.25e

Figure 19: Problem 1.31a

(b) The folium r = (4 sin2 θ − 1) cos θ

-1 -0.75 -0.5 -0.25 0.25 0.5

Figure 20: Problem 1.31b

(c) r = sin 2θ:

-0.75 -0.5 -0.25 0.25 0.5 0.75

Figure 21: Problem 1.31c

(d) The Archimedean spiral r = 0.04θ:

-0.6 -0.4 -0.2 0.2 0.4 0.6

Figure 22: Problem 1.31d

-0.4 -0.2 0.2 0.4 0.6

Figure 23: Problem 1.31e

(c) H(x) sin x:

Figure 24: Problem 1.32a

Figure 25: Problem 1.32b

Figure 26: Problem 1.32c

1.38. (a) x12 − x21+1 ; (b) x − 3 − 1

Figure 27: Problem 1.32d

Figure 28: Problem 1.32e

Figure 29: Problem 1.32f

For the given problem x = 2.

Using (1.33), the sum of the series is

c2 = DB 2 + DA2 = DB 2 + (AC − DC)2 .

But DB = a sin C and DC = a cos C. Therefore

c2 = a2 sin2 C + (b − a cos C)2