UNIVERSITY
PRESS
CI
ffl
a
F.
GERRISH
PURE MATHEMATICS
AN
c
m
>
VOLUME
II
ALGEBRA
TRIGONOMETRY
>
H
n
COORDINATE GEOMETRY
This twovolume textbook on Pure Mathematics has been designed to cover the whole of
London
B.Sc. General
of the
first
mathematics.
is
It
addressed
primarily
student.
also covers
It
matics required
Diploma
the
to
its
kind
General Degree
in
Association.
much
of
it
school
pupils
preparing
for
suitable for
is
by grammar
scholarships
in
However, though
specialists.
aimed to supply a
students at these
Mr
Gerrish has
need of examination
real
levels,
cram book
sense a
goes a
little
for examinations.
is in
no
Thus he
examination syllabus.
By a
conveniently into
falls
despite
cations
to
centres of gravity,
moments of
arclengths,
inertia,
and
coordinate geometry of
sions, in
when
illustrated
instructive.
Finally
the
book contains
many
sets
of
Answers are
iho provided.
15 s
8
sm
a my mi
3. SEP.
NAME
This book
is
last
date
shown above.
0 4^
MANGHE8TER.
PURE MATHEMATICS
VOLUME
II
Uniform with
VOLUME
I:
this
Volume
CALCULUS
PURE
MATHEMATICS
A UNIVERSITY AND
COLLEGE COURSE
BY
F.
GERRISH,
M.A.
VOLUME II
ALGEBRA, TRIGONOMETRY
COORDINATE GEOMETRY
CAMBRIDGE
AT THE UNIVERSITY PRESS
1960
PUBLISHED BY
THE SYNDICS OF THE CAMBRIDGE UNIVERSITY PRESS
Bentley House, 200 Euston Road, London, N.W, 1
American Branch: 32 East 57th Street, New York 22, N.Y.
CONTENTS
General Preface
p a g e xy ji
Preface to Volume
II
xi x
xxi
Chapter
363
10.1
10.11
363
Long
'equating coefficients
10.2
',
p.
366
Exercise 10(a)
ggg
370
Exercise 10(6)
10.3
373
coefficients
374
summary, p. 374. 10.32 Theory of enbic
10.33 Quartic equations, p. 377
10.31 Quadratics: a
equations, p. 375.
10.4
Exercise 10(c)
37 g
Elimination
378
10.42 Common root of
p. 378,
10.43 Repeated roots, p. 381
equations, p. 380.
10.5
two
Exercise 10 (d)
382
The
382
10.51
h.c.f, of
The
theorom,
two polynomials
p. 385.
10.53
10.52
An important algebraic
Theory of partial
fractions, p. 386
Exercise 10(e)
388
388
Two
391
391
CONTENTS
VI
11.2
page 393
Determinants
11.3
11.4
Exercise 11(a)
399
401
11.31 Definitions
and
alien cofactors, p.
403
notation, p. 401.
404
when A = 0:
11.43 Homogeneous
case
Exercise 11(6)
410
Factorisation of determinants
411
Exercise 11(c)
413
11.6
Derivative of a determinant
413
11.7
Determinants of order 4
414
Exercise 11 {d)
415
11.5
Miscellaneous Exercise
Chapter
12.1
12.
416
1 1 (e)
Series
419
419
12.2
Exercise 12(a)
422
Finite series
422
12.21 Notation
for
summing
binomial
and
definitions, p. 422.
coefficients, p. 424.
12.27
Some
Induction, p. 433
12.3
Exercise 12(5)
425
Exercise 12 (s)
432
Exercise 12(d)
435
Infinite series
436
12.31 Behaviour of
properties, p. 43*J
an
12.32 General
CONTENTS
12.4
vii
page 43^
test,
p. 443.
12.5
Exercise 12(e)
442
Exercise 12(f)
446
Exercise 12(g)
450
Series of positive
451
an
12.6
infinite series, p.
457
Exercise 12(A)
459
Maclaurin's series
460
Exercise 12
12.7
465
(i)
466
12.8
Exercise 12(j)
468
Exercise 12(A)
472
Exercise 12(1)
475
Scries
and approximations
476
Exercise 12 (m)
Miscellaneous Exercise 12
481
(w)
482
Chapter
13.1
Complex numbers
Extension of the real number system; ^/( 1), p. 486.
13.12 Tiv&b stage: formal development, p. 487,
13.13 Second stage: geometrical representation, p. 488,
13.11
486
4S6
CONTENTS
viii
page 493
Exercise 13 (a)
13.2
494
cube
493
Exercise 13(6)
13.3
490
p. 500.
500
Exercise 13(c)
13.4
507
3
with
13.5
'real' coefficients;
relations, p, 512.
Exercise 13
012
13.52 Unsymmetrical
13.53 Transformation of equations, p. 513
Symmetrical
relations, p. 512.
514
(d)
r
13.6
515
520
Exercise 13(e)
13.7
(further methods)
521
p.
523
Exercise 13 (/)
524
52o
Chapter
CONTENTS
de Moivre's Theorem
14.
ix
astd
some
Applications
14.1
p.
528
14.12
14.2
page 528
de Moivre's theorem
14.13 Examples,
p. 529,
530
Exercise 14(a)
532
534
14.21 oos m
Exercise 14(6)
14.3
14.4
538
Factorisation
14,31
xn l,
14,33
2n
538
p. 539.
14.32
2a;"costta+
1,
8*+l,p.
539.
p. 640.
Exercise 14(c)
544
Roots of equations
545
14.41 Construction of equations with roots given trigonometrically, p. 545. 14.42 Results obtained by using relations
coefficients, p.
546
Exercise 14(d)
14.5
547
and
summation by C + iS
548
14.52 Other
examples, p. 549
Exercise 14(e)
14.6
Infinite series of
550
single valued
550
rule, p.
558
Exercise 14(/)
556
Exercise 14 (g)
560
561
CONTENTS
Chapter 15. Survey or Elementary Coordinate
page
Geometry
15.1
15.11
15.2
565
Oblique axes
polar
15.14
15.16
565
Advantage
15.12 Cartesian and
coordinates, p. 565. 15.13 Distance formula, p. 566.
Section formulae, p. 566, 15.15 Gradient of a line, p. 567.
Area of a triangle, p. 568
of oblique axes, p, 565.
Forms
570
15.3
573
Further results
15.31 Sides of
a point from a
line, p.
line, p. 573.
575
575
Exercise 15(a)
15.4
Concurrence of straight
576
lines
15.5
Exercise 15(6)
578
Linepairs
579
15.52 The
15.51 Equations which facte rise linearly, p. 579.
locus ax 1 + 2hxy + fc^ 2 = 0, p. 580. 15.53 The general linepair
to the meets of the line
$ = 0, p. 581.
15,54 Linepair joining
Ix + my 1 and the locus s = 0, p. 584
15.6
Exercise 15 (c)
585
The
586
circle
15.7
Exercise 15(d)
592
Conies
594
15.74
K eduction
of 3
Exercise 15(e)
600
601
Chaptee
16.1
The
CONTENTS
The Parabola
16,
locus y 1
page 603
4ax
003
XI
p. 603.
16.12 Parametric
Exercise 16(a)
16.2
go 5
605
>
t,
Exercise 16(6)
16.3
610
Normal
p.
612
16.31
t,
p. 612.
612
Exercise 16(c)
16.4
614
Diameters
615
16.42 Diameters of
parabola, p. 615
Exercise lG(d)
616
616
Chapter
The Ellipse
17.
619
17.1
17.11
x*la*
&P = e.PM,
+ y*jb  l p.
2
x s fa?ij z fb 2
p. 621.
ellipse
17.2
p.
619.
620.
1, p. 620.
17.14 Second focus and directrix,
17.15 Further definitions, p. 621. 17.16 Form of the
a 2/a z + i/ 2/& 2
I,
p. 621.
17.17 Circle
and
ellipse, p.
622
ellipse
623
17.3
Parametric representation
17.31 Eccentric angle
Exercise 17(a)
g5,
p. 027.
627
17.32
The point
t,
p.
628
629
CONTENTS
Xii
17.4
page 630
17.43
Exercise 17 (b)
17.5
Normal
p.
636
17.51
17.6
634
p. 636.
17.52
Conormal
points,
636
Exercise 17(c)
637
The
637
distance quadratic
Exercise 17
C41
(d)
Chapter
18.1
18.
The Hyperbola
The hyperbola g fj
18.1
Form
18.2
The
1;
645
asymptotes
645
definition, p. 645.
18.14
642
bifocal property:
SPS'P =
2, p. 647
+
tir
18.3
18.5
Parametric representation
18.31 Hyperbolic functions, p. 649.
18.4
D2
648
649
t,
p. 650.
651
Exercise 18(a)
651
653
18.6
Exercise 18(5)
654
656
CONTENTS
18.7
Xlii
18.71
xy
18.73
2
,
p. 657.
page 6o7
6S9
Exercise 18(c)
659
Miscellaneous Exercise 18
Chapter
19.1
19,
The locus
19.
= Jcs'
664
664
Scheme of procedure,
P,P 2
19.13 Chord
19.2
661
(d)
of s
p. 664.
0, p.
666
666
p. 668.
19.26
p. 669.
19.3
670
19.4
671
Exercise 19(a)
672
19.5
Number
673
19.6
674
674,
19.7
two
conies, p. 679
= ks'
680
Exercise 19(&)
681
682
Chapter
20.1
19.65 Contact of
The
20.
684
684
straight line
20.2
The
686
circle
of r
2acoa0;
Exercise 20 (a)
688
CONTENTS
XiV
20.3
page 68S
Exercise 20(6)
696
698
Coordinates in space
700
21.2
21.12 Other
Fundamental formulae
21,21 Distance formula, p. 702.
Exercise 21
21.3
702
21,22 Section formulae, p. 703
704
(a)
Exercise 21
21.4
700
705
lines,
708
(&)
The plane
709
Cz+D
713
Exercise 21
21.5
The
713
(c)
714
line
Exercise 21
21.6
721
(d)
Planes in space
21.61 Planes through a
of three planes, p. 724
Exercise 21(e)
723
common
line, p. 723,
21.62 Incidence
726
CONTENTS
21.7
Skew
XV
lines
page 727
21,75 Standard form for the equations of two skew lines, p. 731
Exercise 21(f)
732
Miscellaneous Exercise 21
733
(g)
733
736
22.12
Some
22. 13
Tangent plane at
results
p.
definitions
and
Pu
Exercise 22(a)
22.2
= hs'
22.21
22.4
742
The
a given
22.3
740
circle, p.
742
Surfaces in general
744
Exercise 22(b)
745
The
747
spherical triangle
22.5
Triangle formulae
751
Exercise 22(c)
755
Miscellaneous Exercise 22
Answers to Volume
Index to Volume
756
(d)
II
(25)
II
xxiii
PM
II
xvii
GENERAL PREFACE
This twovolume textbook on Pure Mathematics has been designed
to cover completely the requirements of the revised regulations for the
B.So. General Degree (Part I) of the University of London. It presents
fill
The author
way
to further progress.
no other
book addressed primarily to the General Degree student which covers
the ground with the same selfcontained completeness and thoroughlevel.
On
the
tion.
As
it is
many
students
and inadequate
who
training,
begin
a com
is
included,
and
may
GENERAL PREFACE
XViii
much
and
it is
ability will
appended a Miscellaneous Exercise, both backward and forwardlooking in scope, for revision purposes. Answers are provided at the
end of each volume. It should be clear that, although practice in
solving problems is an important part of the student's training, in no
is
sense
is this
who
a crambook giving
drill
later reading.
a book of this size will be comfrom typographical errors, or the Answers from mathematical ones, despite numerous proof readings. I shall be grateful if
readers will bring to my notice any such corrections or other suggestions for possible improvements.
Finally, I thank the staff of the Cambridge University Press for
the way in which they have met my requirements, and for the expletely free
P.
GEBBISH
xix
PREFACE TO VOLUME
II
and mishandled by
beginners. It
is
subject
The
The
also
variable.
An anticipated criticism
of the book
is
the complex exponential have been introduced too late, with a con
PREFACE TO VOLUME
XX
II
Since
many
students
come
ellipse,
(real)
theory of equations
Owing
xxi
12.73 denotes
12.73(2) denotes
chapter (Ch.
12),
section,
subsection,
part.
in the
same
refers to
equation
refers to
worked example
(ii)
refers to
worked example
(ii)
in subsection 4.64.
refers to
problem number
wo
means
with respect
(ii)
ex.
(ii)
4.64, ex.
(ii)
In the
text,
(ii)
{other
'
omitted at a
first
reading
6 in Exercise 12
(b).
to.
section,
and may be
if time is short.
In an exercise
no. 6
refers to
a 'starred' problem
or
is
later chapter;
matter in
matter in
. .
. .
is
a hint
is
explanatory comment.
363
10
ALGEBRA OF POLYNOMIALS
10.1
10.11
Long
division; identities
The reader will be familiar with the process for dividing one polynomial by another of lower degree; we use it here to lead up to an
important algebraic theorem.
Example
Find the
quotient
x k )ax 2 + bx
ax
+c
ax + (b+ak)
akx
+ ak) x + c
(b + ak) x k{b + ak)
(6
ak 2 + bk + c
is
by
k.
remainder be
If
we now put x =
k in
this,
{xk)Q{x) + R.
we obtain
ak 2 + bk + c
since the
(i)
is
=R
(ii)
zero.
Remarks
may
relation
(i)
{ax 2 + bx + c)
26
 {ak 2 + bk + c) =
a{x 2  k2 )
+ b{x  k),
GFMII
ALGEBRA OF POLYNOMIALS
364
[10.12
so that the righthand side has the factor x k. If the other factor
is
which
= (xk) Q(x),
is equivalent to
with R given by
We may regard the 'long division as a process for constructing
(i)
(ii).
'
(ft)
The
n 1
or
is
of degree
n >
1,
R will be
a polynomial
less.)
(i),
in x (in contrast to
ax 2 + bx + c
= (x k) Q(x) + R.
(i)'
sign because
we may
ever,
(ii) is
= even when
we
How
the expressions
their identity.
10.12
p(x)
is
divided by x k.
p(x) p(k)
xn +p 1 xn
~1
+ ... +p n!X +p n
We have
=p
(xn
n
)
~
+p 1 (xn 1  k
71
1
)
+ ... +p n  1 (x  k).
for
(iii)
(When x
4= k,
a;
&
is
).
g.p. in
Which
will
(x
be a polynomial in
k) Q(x).
a;
of degree
(iv)
n 1
if
n>
1.
ALGEBRA OF POLYNOMIALS
10.12]
365
Examples
(i)
+nx 6.
Calling the polynomial p(x), we seek numbers k for which p(k) = 0. Only
those numbers k which are factors of the last term 6 need be tried, for any
other number could not be the constant term in a factor of the polynomial, f
p(l)
= 211 + 176
p{2)
= 1644 + 346 =
Find
a:
1 is
not a factor.
/.
a factor.
by
a and b
if 6x*
remainder) by 2x 2 + x
+ ax 2 + bx 2
(iii)
7,
Find
the
say
Ax + B.
= 1.
lfQ(x)
is
is divided
by (x a)(x b), a
4= 6,
where
2.
a:
p( a )
= Aa + B.
6:
p(b)
= Ab + B.
A
so the remainder
which give
a + &,
Put x =
a;
2xl
0,
is
p{x)
Put
1.
= ab8,
=
A and B, we find
_p(a)p(b)
ab
IS
ap(b)bp(a)
ab
is
{ p(a)
p(b)} x + ap(b)
 bp(a)
ab
t
is
inspection.
and so a
is
(a:2)(a;3)(2a?l)
factorisation
the values of
divisible, without
/.
= (x2)(2x*lx + 3),
=
(ii)
0,
p(x)
on completing the
#: 0,
13.62(1).
262
ALGEBRA OF POLYNOMIALS
366
[10.13
We
continue to write
p(x)
where p
,
. .
. ,
#= 0.
Theorem I.
av a 2
= p xn +p 1 xn ~ 1 + ... +p n
is zero
Ifp{x)
an then
,
P(x)
(v)
).
11
Since p{a 2 )
by
hypothesis,
we have
= ((hc^QnMand therefore
a x (also by hypothesis), hence Qn i(a 2 ) =
xa 2 is a factor of Q n ^ x {x). The other factor Q n _ 2 ( x )> obtained by
n  2 thus
division, is a polynomial whose first term is p x
As a 2
4=
Qnifr)
and so
p(x)
(x 
ax )
(x
 a 2 Q n  2 {x).
)
We
pQ
=(=
by
hypothesis.
n ~1
p xn +p 1 x + ...+p n is zero for more than n
distinct values of x, then p = p t = ... = p n = 0, and so p(x) = 0.
Proof. Either all of p ,Pi, ,Pn are z ^TO
or there is a first p which is not zero,' say pk (k < n).
Theorem
II.
//
>
In
p k Xn
+P k+1 Xn
k  1
+..+P n
(Pk*
0),
which
it
~k
is
ALGEBRA OF POLYNOMIALS
10.13]
367
given that it is zero for more than n values of x. Our second alternative
first
alternative
is
possible.
when p = p x =
Finally,
for all x,
i.e.
p(x)
...
= pn =
0,
is
zero
0.
Coeollary II (a). // p x n +p 1 xn ~1 +
Po=Pl =
For since the polynomial is
. .
+pn =
= 2> =
0,
then
0.
identically zero,
it is
zero for
more than
n distinct values of x.
Corollary II (6). //
~
p xn +p 1 xn 1 + ...+p n =
for more than
n distinct
Po
n
q x +
q^
+...+qn
Pi
qo,
Pn
$v
qn
>
For we have
1
n
(Poq )x + {p 1 q1 )x^ +... + (pn qn ) =
0,
is
principle*)"
cients' in
all
of 'equating
It asserts that if
values of x
(i.e. if
x),
two
the poly
then they
agree term
result it
by term
coeffi
(i.e.
same
for the
x,
Corollary II (c). //
~
p xn +p 1 xn 1 + ... +pn =
(where
then
1
m
q x + q^
+ ...+qm
= qm _ n _ 1
p = qm _ n
degree n.
Examples
(i)
Find
constants a,
n =
8
b, c,
dfor which
an(n +
1) (n
+ 2) + bn(n + 1) + cn + d.
p n = qm
ALGEBRA OF POLYNOMIALS
368
The righthand
[10.13
side
is
identically equal to
and only
if
3a + 6
1,
i.e.
(ii)
nz
2a + 6 + c
0,
1,
if
= 3,
1,
d=
0,
a*
0,
0.
= a(x s + Bx + 2) + b(x 2 + Ix + 1 2)
= (a + 6) x z + (3a + 76) x + (2a +126),
The expression
which
is
identical with
x%
if
and only
a + 6=l,
The
first.
if
3a + 76
2a+126 =
0,
0.
last
If a,b, c are
all distinct,
(x b)(x c)
(a 6)(a c)
and deduce
prove that
(x c)(x a)
(x
(6 a)
(c
(6
c)
a)(x 6) _
a)(c 6)
powers
of x.
When x =
a,
1.
(x c)(x a)
(b c)(b a)
(x b)(x c)
(a b){a c)
(x a)(x b)
(c
a)(cb)
is
(a6)(a c)
(a
6+c
b)(a c)
(a
6) (a c)
(6c)(6a)
(b
c+a
c)(b a)
(6
c)(6 a)
1
a)(c 6)
a+b
_
(c a)(c b)
ab
If ax s + bx % + cx + d contains (x+
l) 2
o,
(c
ca
be
(iv)
(c
a)(c 6)
a, b, c, d.
If (x + 1 ) 2
Hence
is
a factor, the other factor must be linear and of the form a(x + k)
ax* + bx* + cx+d
.*.
= a(x+l) 2 (x +
= oa;8 + a(A! + 2)a; 2 + a(2fc+l)a;lafc,
a(k + 2),
Jc)
a(2& +1),
= ak.
ALGEBRA OF POLYNOMIALS
10.13]
369
By eliminating h from the first and third, and from the second and third,
b = d + 2a, c=2d + a.
is
*(v) Given two polynomials f(x), g(x) of degrees m, n (m > n), prove that there
a unique pair of polynomials Q(x), B(x) such that E(x) has degree less than n and
f(x)=g(x)Q(x) + B(x).
(a)
of degree m n and
all
values of x
f(x){g(x)Q(x) + B(x)},
whose degree certainly does not exceed m, is zero for more than m values of x.
Hence by Theorem II it is identically zero, which proves (a). Compare 10.11,
Remark
(fi).
f(x)
less
than
we might have
= g(x)q(x)+r(x),
n. Subtraction of (b)
B(x)  r(x)
g(x) {q(x)
(b)
from
(a)
gives
 Q(x)}.
(c)
of (c) has degree < n, while (unless q(x) Q(x) = 0) the right
^ n (the degree of g{x)); and by Corollary II (c) this is impossible.
Hence q(x) Q(x) = 0, and therefore by (c), B(x)r(x) = 0. This proves the
The
left side
has degree
uniqueness.
Exercise 10(a)
Factorise
1
2x3 + Sx 2 l.
4 Solve x
4:X + x + 6 =
2
x* + 2x 3 + x 2  4.
0.
and
6 if
6a; 8
x+1.
6 The remainder when (xl)(x 2)
a and b.
divides x* + ax3 + b
is
a;+l.
2).
Find
Find
x2 x + 2, x 2 + x
respectively.
Find values
11
of a,
b, c,
d for which
n 8 + 6n = an(nl)(n2) + bn{nl) + cn + d.
12 x3
a(x+2)* + b(x+l) 2 + cx + d.
down
ALGEBRA OF POLYNOMIALS
370
2x + l
14 Express
If a,b,c are
^
16
15 Express
m the form
.
(xl)(x + 2)
b
H
D m the form
.
2x+l
.
x*+l
prove
+ =
+(a b)(a c)
{c a)(c b)
(b c)(b a)
+ b + x) (b + c + x)
(b c)(b a)
(a
x1 x + 2
a
bx + c
{2x+l)(x* +
all distinct,
[10.2
(b
+ c + x) (c + a + x)
(c a)(c b)
(c
x.
+ a + x)(a + b + x) _
~
{a b)(a c)
Verify that
(ii)
(xc){xa
(xb)(xc)
{xa)(xb)
B
c
c)(b a)
(b
(c a)(c b)
)
(a
b)(a c)
(i)
that this
is
quadratic.
*19 Write down the (unique) cubic polynomial in x which takes the values
A, B, C, D when x has the distinct values a, b, c, d respectively.
20 (i) If x3 +px + q contains a factor (x a) 2 prove 4p 3 + 21q % 0.
3
prove the converse of (i).
(ii) By writing q = 2a
,
21
(i)
x, y,
we may arrange
q in terms of a.
(ii)
10.2
(i).
is
%nPo(y) + x^Piiy) +
+p n (y)>
to be a polynomial in x whose coefficients are polypreceding theories can then be applied, f Similar
The
y.
considerations hold for more than two variables.
and consider
it
nomials in
Examples
(i)
P rove
If
a',
b',
c',
+ 2h'xy + b'y* +
+ Wv +
f=f, g = g', h = h'.
a'x*
a'x*
x,
<>'>
in y (or in
any
ALGEBRA OF POLYNOMIALS
10.22]
Equating
a
coefficients,
2(hy + g)
a',
371
2(h'y + g'),
by 2 + 2fy + c
b'y 2
+ 2f'y + c'.
(ii)
Since
2* 2 + 5xy  3y*
(x
linear factors,
try putting
2x 2 + 5xyZy 2 x+lly6
(x
+ Zy + a){2xy + b),
is
+ 5xy  Zy 2 + (2a + b) x +
36  a) y + ah.
This will be identical with the given polynomial if and only if a, 6 can be chosen
to satisfy
"
The
,
= \, 36a=ll,
ab = 6.
give a = 2, 6 = 3; and these values do
2a + b
two equations
Hence the given polynomial has linear
first
third.
satisfy the
factors
(x+Sy2)(2xy + Z).
Factorise
(iii)
Regarding
xy(x + y)
this as
nomials in y and
z,
0.
SO
xy(x + y)
0,
We obtain k =
1.
1,
1) in
10.22 Symmetric,
skew and
(e.g.
of x 2y).
y, z (e.g.
cyclic functions
variables
definitions.
if it is
unaltered
by the
ALGEBRA OF POLYNOMIALS
372
[10.22
Bemarks
(a)
constant multiple
change of x,
y,
then
Ix
i.e.
so
+ my + nz =
(l
+ mx + nz,
m)x+(m l)y =
= m. Similarly, we find I =
+ y + z).
l(x
ly
n,
and the
0,
k(x 2 + y 2 + z 2 )
(2)
consisting of
+ l(yz + zx + xy).
interchange of any two of them changes only the sign of the function.
For example, x y
a skew function of
is
(b
(x,
y)
and
c) (c a) {a b),
xy(x y)
are
of two symmetric
is
the product of
skew.
11.
Example
Factorise x3 + y z + z 8 3xyz. (Also see Ex. 10(/), no. 3.)
Regarding this as a polynomial in x, we find that when x = (y + z) the
expression is zero. Hence x + y + z is a factor. The other factor could be found
directly
in (x,y,z),
k{x2 + y % + z 2 )
+ l(yz + zx + xy),
and so
x* + y* + z z Zxyz
(x
ALGEBRA OF POLYNOMIALS
10.22]
Hence
by substituting numerical
0,
0,
give
k;
0,
y=l,
give
2{2k + l},
xP + y^+z 3 3xyz
:.
=  1.
Consider again
letters x, y, z are
of a circle as shown.
values:
xy(x  y)
373
When
the
first
term xy(x y)
is
Lxy{xy).
in (x, y, z).
similar definition can be given when there are
is cyclic
Exercise 10(6)
Prove that the following expressions have linear factors, and find them.
1
x*y i + 2xz14yz48z*.
ALGEBRA OF POLYNOMIALS
374
3
[10.3
linear
factors.
the
(x + y + z) 3
helpful, factorise
(x3 + y3 + z 3 ).
+ zx(z 2 x 2 +xy(x 2 y 2
x(yz) + y(zx) 3 + z(xy) 3
(x + y + z) 5 {x 6 + y 5 + z 5
6 yz(y 2
7
factors.
).
).
full
Zx 2
6, 7
in the Snotation.
(ii)
S6c(6c);
(hi)
(iv)
Sa 2 6c;
(v)
26c2
[In (v), cyclic interchange gives bc + ca + ab which is only half the number
of terms implied by S, meaning 'sum of all terms of the type "letter x another
letter squared'".]
2
11
Prove that
(i)
S(6c)
a(6
(iii)
12 Prove S(6  c) 3
=
2
0;
c
Xbc(bc)
(ii)
2
)
= Sa 2 (6c);
(bc)(ca)(ab).
3(6  c) (c  a) (a 
6).
Sa 8 (6 2 c 2 = (bc)(ca)(ab)(bc + ca + ab).
14 Prove that Sa n (6 c) contains the factor (6 c) (c a) (a 6)
positive integer n ^ 2.
13 Prove
*15
By taking x = 6 c, y = c a,
(i)
duce no.
12.
(ii)
16
(c
 a) 5 + (a  6) 5
[If/(a, 6, c) is
any
no. 8 to factorise (6  c) 5 +
for
is
also cyclic.
symmetric,
= f(b,a,c) =f(b,c,a);
f(b, a, c) =  { /(&, c, a)} = /(&, c, a).]
f(a,b,c)
if it is
10.3
skew,
f(a. 6, c)
coeffi
cients
10.31 Quadratics: a
If the equation ax
reader will
know
summary
+ bx + c =
that
has roots
a+/3 =
a,
fi
a/?
.
a and
ft
(such as
a 2 +/? 2
a//?
+ /?/a),
ALGEBRA OF POLYNOMIALS
10.32]
375
numbers
a,
/?,
satisfy
ax3 + bx 2 + cx + d
(i)
0,
ax3 + bx 2 + cx + d
numbers a,
Theorem xcc and xfi are
If the unequal
= a(xa){xfi)(xy).
ft
satisfy
(i),
'
Similarly,
(ii)
if a,
holds in
/?,
all
and by
is
By the statement
Thus,
(ii)
by
3
(ii) is a(x  a)
cases.
a are
'a, a,
side
(i),
by
then
direct
multiplication,
ax3 + bx 2 + cx + d
Equating
coefficients,
we find
a + p + Y = ^,
PY + Y + P =
a,
(xa)(xfl){xy) =
i.e.
in
^
ft,
a PY =
P)
y is
0,
0,
which
and
the coefficient of x3
is
the coefficient of x2
is
the coefficient of x
is
is
+1,
(sum of roots),
(sum of the products of the
roots taken in pairs),
(product of roots).
of the signs.
Observe the sequence
Remark. The relations (iii) do not help us to solve the cubic equation,
because ehmination of (say) and y from them leads to

act3
+ b<x 2 + ca + d =
0;
'a, /?
ALGEBRA OF POLYNOMIALS
376
[10.32
Examples
(i)
If a,
/?,
Sa 2
(a)
We have
S^,
(6)
Sa =
Sa 2 =
(a)
.Firs*
method. Since
S/?y
0,
express in terms of p, q, r:
0,
Sa 5
(a*)
(e)
S/? 2?
a^y = .
,
(Sa) 2 2S/?y
a
(c)
Sa 3
(c)
= 02? =
a/>y
/?,
+ q<x + r =
0;
y. Adding,
pSa 3 + gSa + 3r =
Sa 8 =
..
0,
Second method.
+/ff
0,
Sa =
since
0.
3r
Sa 3 = 3a#y =
..
P
(d)
pa.
Adding
this to the
0,
+ qu + ra =
0.
two
fi,
y,
and hence
0.
and so Sa 6 = 5qr/p 2
2
(e) Consider (Sa)(S/tfy). A term like y# y occurs only once, as the product
ft .[iy. The terms a/?y arises in three ways, from a./ffy, fi.ycc, y.afi. Hence
by using
(a)
and
(c),
(Sa)(S/?y)
:.
= S/?^
3r
,
p
Form the cubic whose roots are
S^ +
Sa/Jy.
and
Efi*y
Sr
= .
p
(a) ft + y, y + a, a
(b) fly/a, ya//?, ccfify,
fl
+x 2 24% 16 = 0.
We have
Sa = 1, S/?y = 24, a/?y = 16.
(a) /? + y = Saa = 1a. Similarly y + a = 1/?, a + /? = 1y. We
require the cubic whose roots are 1 a, 1 ($, 1 y.
(ii)
where a,
are
roofe o/ x 3
ALGEBRA OF POLYNOMIALS
10.33]
= I x, i.e. x = 1 y:
(l2/)
i.e.
377
+ (l2/)224(l2/)16 =
y* + 2y*  2Sy  8 = 0.
0,
(A)
The values of y which satisfy this are related to the values of a; which satisfy
the given equation by the formula y = 1 x. Since these values of a; are a, fi, y,
1 y; therefore (A) is the required
hence the values of y are 1 a, 1
cubic. (We could also say that
xs + 2x*2Zx 8
is
the required equation, because the letter used for the unknown
are given.)
(b) fiyfoc = afiyloc* = 16/a 2 Put y = 16/x*, i.e. x 2 = 16/y:
is
immaterial
\y
so
by
64cc a
squaring,
^3j =
*(?.)
y\y
in (a)
of 13.53.
shows that
16 2
^1^
=M
a
,
y)
y* 38t/ a +
which reduces to
4% 16 = 0.
By reasoning
as in 10.32
we
find that if a,
ax* Hfoe3 + cx % + dx + e
a + fi + y + 8 = ,
then
y,
0,
afi+a,y + a8+fiy+fiS+y8
,
= ,
i.e.
i.e.
aBy8 = 
Hafi
,
Ea^y = ,
Cb
and
Sa = ,
i.e.
/?,
Qi
signs.
and elsewhere.
16.26, ex.
(i);
(ii);
ALGEBRA OF POLYNOMIALS
378
[10.4
Exercise 10(c)
If a,
ft
a 2/? + a/? 2
(i)
(v) (a/?) 2 ;
2 If a,
roots are
(ii)
/?
(i)
(iii)
^ + ^;
(vi)
i/ a ,
calculate
(al)(/?~l);
a3 +
(vii)
+ 2* 4 =
3a; 2
a2
(ii)
0,
/?
(viii)
a 2 +/? 2
a4 +
/ff*.
a+
/?+ 1/a.
1//?,
to be positive.
0,
(iii)
/ff
(iv)
4 If a,
(a + l)(/?+l)(y+l);
(i)
(v)
Za 3
(vi)
(iii)
(ii)
2a 4
(i)
P+y, y + a,
a+fi;
(iv)
(ii)
a2
j^
,
/ff
2
,
L,
2
;
^;
(v)
(iv)
0y
S^;
(vii)
Sa 2
(iii)
a" 2
0,
a(/?
/?~ 2 ,
7~ 2
G 2 + 4H3 =
8 If the roots of x z +
9 If
a;
0.
[a + 7
whose
[Let the
to be in
2y#.]
(i)
a.p.
g.p.
(ii)
a 2 fiy = a 2 + a
J_
I
3
= a +r =
pct 2 q<x
= (pa + q).
,
~
_
are
10.4
Elimination
When a
is
cannot
all
ALGEBRA OF POLYNOMIALS
10.41]
379
There may be more than one such relation. Each is called an eliminant
of the system of equations. Owing to the importance of elimination,!
especially in coordinate geometry,
No general methods
(i)
Use of an
we
give
identity.
Eliminate x, y, z from
x + y+z
a,
b2
x3 + y 3 + z3
c3 ,
xyz
= d3
(10.22, ex.)
+ y + z 3xyz =
3
+ y + z)(x 2 + y 2 +z 2 yzzxxy)
c  3d 3 = a(b 2  Hyz)
= Ea: 2 + 22yz, so Xyz = %{a 2 b 2
2(c3  3d3 = a(36 2  a 2
(x
we have
(Sz) 2
and
Hence
(ii)
x2 + y 2 +z 2 =
).
).
Eliminate m,
= a2t 2 + b 2
+ 2xyt + (6 2  y 2 =
(ytx) 2
(a
i.e.
x
2
)
0.
y =
1, so
a2 x2
x 2 +y 2
i.e.
is
a2 + b 2
(iii)
x + ay + a 2z
a*,
x + by + b 2z
a, b, c
Hence
is
c*.
zero, the
sum
0.
of the roots
is
zero,
(a + b + c). Then
27
3 is
x + cy + c 2z
t*t2ztyx
remaining root
bi ,
GPMII
ALGEBRA OF POLYNOMIALS
380
Common root
10.42
of two equations
(1)
ax 2 + bx + c
have a
[10.42
0,
common root is
{ab' a'b)
b'c)
{be'
c'a) 2
{ca'
Necessary. If there
ating
first x,
(ab'
a'b) x 2 =
{ab' a'b)
be'
{be'
b'c,
b'c)
a'b) x =
{ab'
{ab'a'b) 2 x 2
ca'
c'a.
{ca'
c'a) 2
a'b
=(=
the condition
0,
(ca'c'a\ 2
be'
 a'b) A 2 =
be'
 b'c,
{ab'
a{bc'
(i)
b'c
{be'
 a'b) X =
{ab'
a'b) (aA 2 + bA + c) =
0;
and
ca' c'a.
c{ab' a'b)
Hence
can be written
and
on simplifying.
similarly
0.
The condition
(i) is
{b 2
(6
therefore equivalent to
{2ac'
+ 2a' c  bb'f.
(ii)
ALGEBRA OF POLYNOMIALS
10.43]
381
When the given equations are of degree higher than the second,
(2)
quadratics
/ = ax3 + bx 2 + cx + d =
and
have a common
root,
(
root
if
like
(i)
(iii)
satisfies
p / ax
.
0, i.e.
(iv)
0.
common
Conversely, a
common root
= px 2 + qx + r =
of g
and only
if (iii)
where g{a)
4= 0.
root
a,
r.
// p(x)
0,
(xct) r g{x)
(1
<
<
n),
not a factor
p'(x)
g(x).
x a of order
of order
suppose that
of the polynomial
factor
10.13,
r;
0.
has a root x
of order
=a
of order
r,
then
=a
is also
root of
rl.
Proof.
#'(a:)
+ (x a) r g'{x)
=# 0,
when
a.
Hence
a;
common
sufficient condition
0, p'{x)
have a
root.
272
ALGEBRA OF POLYNOMIALS
382
[10.5
Exercise 10(d)
Eliminate
from
tljt, y
).
[Solve the
first
6 Eliminate
(i)
y + nx
8
two equations
x, y, z
Eliminate
= b 2 zx = c 2
3
y + Ix = al + 2aZ,
third.]
a 2 yz
= d2
= aw3 + 2am, and
x2 + y 2 + z %
i/
+ ma;
Interpret geometrically.
*(ii)
Eliminate m, n from
m x my + a = 0,
n 2x ny + a
*(ii)
a3 b z using the
from
x,
= an 3 + 2an.
(i)
from xy
for
m n = c(l +mn).
0,
Interpret geometrically.
9 Eliminate
Z,
and Im =
b.
10 Eliminate A,
a2 + A
11
62
+A
a 2 +^
x + ay + a
=
a+b+c =
x+y+z =
12 Solve a + b + c
13 Solve
14 Solve
= a3
x
to
the necessary
x + cy + c 2z
c3
and
2px 2 + 2qx + r
and
is
x2 +px + q
and
Spqr + p r + q =
3
0.
+px + q =
0. (Cf.
20.)
0. (Cf.
21.)
b3
= 7, abc = 3.
2, a + b 2 + c 2 = 30, abc = 10.
3
3
3
2
2
2, x + y + z = 14, x + y + z = 20.
x*+px + q =
16 x
17
bc + ca + ab
5,
Find
x. y, z:
x + by + b 2z
6 2 +/t
for
axz + bx + c
have a double
0,
root.
px 3 + qx + r =
0.
[Eliminate x, x 3 in turn.]
*19 If b 2 # c and (x 2 + 2bx + c) r_1 is a factor of p(#) and of p\x), prove that
+ 2bx + c) r is a factor of p(x). Explain why the restriction b 2 4= c is needed.
(x 2
10.5
The
h.c.f.
10.51
The
h.c.f. process
The
of two polynomials
any) of p(x)
0,
we must
shows
find
all
factors
common
top(x) andp'(x).
ALGEBRA OF POLYNOMIALS
10.51]
common
highest
383
factor (h.c.f.) of
any
and so on.
m and deg g = n, where m ^ n. Then we can divide
obtaining the quotient q x (x) and remainder rx (x), wheref
Suppose deg/ =
f(x)
by
g(x),
f(x)
3W
getting quotient
(ii)
degr2 ^ n 2.
i.e.
rJx),
and hence
g(x),
g(x)
(i)
where
by
.
r x (x)
r 2 (x) gives
n~\
(m)
and deg rz ^ n 3.
Proceeding thus,
and
where
rs+1 (x)
applications:]:
First,
0.
we eventually
VaO*)
rs_ x (x)
The
rs (x)qs+1 (x)
stage
is
identity
(s)
+ rs+l (X),
(8+1)
n+l
f(x)
The
obtain
(i)
rx (x)
=fx (x)h(x),
g{x)
so that
gx (x)h(x).
shows that
 g x (x) qx (x)},
is
is
so that h(x)
a factor of r8 (x).
shows that
k(x) is also
(A)
1)
is
ALGEBRA OF POLYNOMIALS
384
and hence by
etc.
Thus from
(i) it
is
From
(B)
it
and g(x);
must also be a
is
a factor of
f(x)
we
(ii)
[10.51
g(x).
(B)
if rs (x) is
Examples
(i)
Find
the h.c.f. of
2a;
We begin by dividing
2(x6 x i + x 3 +
4x*2x + S)
by
2a;
2xl 4x3 + 5x 2 +
2x* + 2x 3  6a; 2 +
2x
18a;
2a;
6x 3 
6a;
+ 18a; 54
34a; + 51
2
2x+ 3 = r2
17a; 2
Remove
factor 17:
The K.c.v.
(ii)
is
a;
2a; 6
3.
is
common factor
is
any nonzero
ALGEBRA OF POLYNOMIALS
10.52]
H.C.F.
of / and /'
5(4a; 4
4a; 4
4a; 4
 14a;3 +
 2ar*+
 2a;3 +
The
it
h.c.f. is 2a; 3
4a; B
4a;4
that
+ 8a;
lx 2  4a; 8a; 2
7a: 2
385
r9
7a; 2 + 4a; + 4. By
we
find
From
10.43
it
follows that
(a;2) 2 (2a;+
in/,
of orders
1).
factors of orders 3, 2
3, 2.
5,
10.52
ri
ra
and so
on.
=
=
f+bo g,
where
(hf+brf,
where
ax
where 0,
\Jr
//
Theobem.
af+bg,
(C)
(C)
a<j>
+ bxjr,
hence
A,
= qx
b t = l + q1 qa
= 1,
= q 2
a, b
1.
(6)
= Af+Bg,
h = Afh + Bgh.
therefore gh = fk for some polynomial k, and so
h = (Ah + Bk)f,
.
and so
Since / is a factor of gh,
i.e.
/ is a factor of h.
t
The
ALGEBRA OF POLYNOMIALS
386
[10.53
Remarks
(a)
The polynomials
a, 6 in
a,
(where c
is
and
a\
0,
10.53
/ is divided by g.
Theorem. If p x p 2 are coprime polynomials,
f/(PiPz) can oe expressed uniquely in the form
+ r,
rx
3+
Pi
Pi
APi + BP2 =
_/_
J(APl + Bp = Bf + Af^
2)
PiPz
By
division,
where deg rx
<
Bf =
q x p x + rx
Pi'
PlPi
Pi
and Af =
q2p z + r2 ,
?i
+ ?2 + +
(i)
Pi
Pi
PiPi
where rx /p x r2 /p 2 are proper. They are also irreducible', for if (say) rx /p x reduces
to rx /p x then result (i) shows that//(p 1 p 2 ) is equal to a fraction with denominator p'xp % where deg (p[p 2 ) < deg (p x p 2 ); this contradicts the hypothesis
,
+ =
Pi
then
Pi
(i),
?'
suppose
if
+ +;
PiPi
(q  q') p x p 2 + (r x  r^) p 2 =
Pi
Pi
(r 2  r 2 p x
)
possible that
ALGEBRA OF POLYNOMIALS
10.53]
Since
p2
is
p2
side, therefore
387
a factor of
is
(r2
r2 )p x
Corollary. If p lt p 2
then the (irreducible)
in the form
q\
1
Pi
...
Pn
Pi
Pi(PiPn) to give
gi
then to p 2 (p 3
. . .
pn
first
to
+ +
J1
Pi
PzPn
to give
f
~
= ?2+
+;
p 2 P 3 ..Pn
^*2
/i
p 2 ...p n
and so
f =
q+
^ +S
X
{(x
(xa)
r
1*.
2
6)
(ii)
+ c 2 }*
where q and the rx r2 are polynomials and degrx < r. degr2 < 2s.
By division, any polynomial $(x) of degree m can be written in the form
,
<f>(x)
where
is
constant and
<f> x
(x)
on.
Combining
<f>(x)
all
a)$ 1 [x)+A a
has degree
$ x (x) =
and so
(x
Similarly
1.
(xa)<f> 2 (x)+A x ,
these results,
we havef
J^Lt
a) r
any polynomial
Similarly,
<fi(x)
where A Q
4>(x)
(a;
(4
a;
. . ,
a)
(j>
x {x),
+ + (^ 1 a; + 1 ){(a;&) 2 + c 2 } + ...
+ (A s _ x x + Bs X
etc.,
we obtain
This
is
the
Lemma in
6.41.
{(x
 6) 2 + c 2}**
ALGEBRA OF POLYNOMIALS
388
has degree 2s
can be decomposed as
if <j>(x)
x+B
{{xb) 2 + c 2 }'
By
(iii),
A ^x + B
x + B1
+ c2}
s.x
(xb) 2 + c 2
K
'
'
in proving the
unique.
(iv) is
combining
{{xb^ + c 2 }* 1
fraction rj{(x b) 2
or less.
[10.53
and
(ii), (iii)
(iv),
we completely prove
the statements
made
section).
Exercise 10(e)
Find
1
the h.c.f. of
x 3 + 3x z 8x24, x3 + 3x 2 3x9.
 a: 2 + 4a: + 15,
2a: 3
a;
14.
+12a;5.
a:
x*
 x 3 + 4a; 2  3x + 2.
and hence
+ 54
*9 Find polynomials A,
12a:
+ 4a:
45a:
solve completely
8 x6  5x 3 +
0.
of least degree
5a:
2 =
0.
such that
+ B(a: 2 + 2a:3) =
1.
Expand
{x
+ y) to
3
torise the
x3 + y3 + z 3 3xyz
4(a 3 + 6 3 + c 3 3abc).
4
4
(ii) Sa(6  c
+ 21uyz to prove
2
2 =
(i) S(6  c) 2 = 2E(a  6) (a  c)
(ii) Sa (6  c)
2S6c(a  6) (a  c).
1~
3
3
Eliminate t from x = t 2 + 1~ 2 y = t +
Eliminate x, y from x y = a, x 2 y 2 = b 2 x 3 y3 = c3
If pa: qy = 2 y 2 py + qx = 4xy, x2 + y 2 = 1, prove (p + g)* + (p g)* =
6 Factorise
(i)
S(6 3 + c 3 ) (bc);
(2a:) 2
).
Sa: 2
10
[Solve the
a;
first
two equations
2.
ALGEBRA OF POLYNOMIALS
389
when x has
the
values given
a1
a2
&l
&jj
63
C\
C2
Cg
"
0.
proportional.]
14
(i)
What can be said about the coefficients in an equation whose roots are
a,fi,ccfi;
(ii)
where 61 + 6 2 + 6a
is
(iii)
0,a,,fi,y;
an
a/b,b/c, c!a;
*(iv)
tan dv tan
a,
tan ds ,
= at?
ia.
has
16 If one root of x* + ax + b
prove that the roots are
is
= 0.
ft,
+px + q =
0,
express a +fi
i
+ yt
in terms
+ 49a;2 + 78x + 42 = 0,
20 EUminate
x2
A,
y*
fi,
0, find
from
x2
y3
z2
x2
y3
z2
+
+^+
=1,
+
= 1,
= 1
+ ?+
6+A c+A
a+/t e
b+/i c+/i
a+v b+v c+v
and
\/iv = abc.
s
3
=
If
21
x + Sax + Sbx + c
has a repeated root, prove that this root also
satisfies x 3 + 2ax + 6 = 0. Hence show that the repeated root is (c a6)/2(a 2 6).
22 Prove that x* +px + q = cannot have a repeated root of order 3.
a+A
'
23 Prove that
'
x3
x3
xn
24
rt
4=
and solve
if
n is odd.
Clearly p(
in each case.
=t=
390
*25 Determine
ALGEBRA OF POLYNOMIALS
the values of m and c for which the line y mx + c is
(i)
an
(ii)
y.]
*29 Cubic equations : Cardan's method^ of solution. Every cubic can be reduced
(see Ex. 13 {d), no. 15) to the standard form
x* +
shows that
BHx + G =
*/ y, z
(a)
0.
yz = H,
3
3
=
that
Prove
then x
y z will be a root of (a).
y z must be the roots of
3 =
2
0.
t GtH
z
y + z = G and
3
(b)
(c)
*30 (i) If O +
> 0, show that there are distinct numbers y, z satisfying (b),
so that x = y 2 is a root of (a). Use no. 4 (i) to show that there are no other
2
4Z?3
roots.
If
Put x
= kcosd
in (a),
be used.
4:
3. Show
that
then becomes cos 30 = Q/{2H*J( H)}. Verify that condition (iii) ensures
that 3d can be found from this, and hence that three values of cos 6 = x/k are
(a)
obtainable.
f Socalled, although discovered
by
Tartaglia.
391
11
11.11
Two
cx ,
a2 x + b 2 y
c2
a 2 b x )x =
cx b 2
c 2 b x
and
(a x b 2
a2 b x )y =
a x c 2 a 2 c x
(i)
If a x b 2 a 2 b x
When
arise.
(a)
If at least one of c x b 2 c 2 b x a x c 2 a 2 c x
,
contradiction.
is
not zero,
therefore have
no
(i)
gives a
solution,
and
If a x b 2 a 2 b x
cx b 2
c2 b x = ax c2 a2 cx =
0,
then
(i)
gives
no
ax
The
#= 0.
first
terms of y. When y
x be xn then
is
X~
ax
Cl
+ b 2 y  c2 = ^ a
CT
a2 x
let
Since
CT
2 Cl
+ b 2 y  c2
x
~ ai C2 +
ax
by hypothesis
(6),
any values of x, y satisfying the first equation also satisfy the second.
The solution is said to be indeterminate. (Roughly, hypothesis (6)
392
[11.12
ax b 2 a 2 b x
the lines are not parallel, and the above solution (i)
represents their unique point of intersection. In case (a), the lines are
4= 0,
to
same line.
unknowns
gives
(a x c 2
ax x + b x y + cx z
dx
a2 x + b 2 y + c2 z
d2
as x + b 3 y + c 3 z
d3
a 2 c x )x + (b x c 2 b 2 c x )y =
d x c 2 d 2 c x
d 2 c 3 d6 c 2
(b 2 c3
 63
2)
(d x c 2
3 c 2 )}
2 cx )
 (b x c 2  b 2 c x
coefficient of
(d 2 c 3
a x b 3 c 2 + a 2 b 3 c x a 2 b x c 3 + a3 b x c 2 a3 b 2 c x ),
similar expression.
expressions for y
3 c 2 ).
x is
We
and
is
also the
(ii)
denominator in the
z.
(ii).
There are six terms. In each term there is just one a, one b and
one c; and in each term the suffixes 1 2, 3 all occur, without repetition.
The signs preceding the terms are alternately +
(a)
11.2]
393
(/?) The suffixes of the letters a, b, c in each term form one of the six
permutations of the numbers 1, 2, 3. For example, consider a z b x c3 if
we interchange the suffixes in pairs until they are in natural order
:
we
1, 2, 3,
number of interchanges
is
is
is
is
evenf have
number of interchanges
is
either always
it
sign
while
can be proved
odd or
else
always
Determinants
1 1 .2
The symbol
A=
ax
bx
a9
b9
mean a x b 2 az b x The
.
The value of A
That is, if
(1)
is unaltered
A'
then A'
A.
b2
By use of this result, any property proved for rows extends at once
to columns,
will therefore
be stated for
rows only.
(2)
That
is,
a2
b2
ax
bx
= A.
A=
(3)
(4)
For example,
t
X
is
kb x
a9
ft,
A by
k.
= M.
reckoned as even.
minants
kax
0.
(11.22).
394
(5)
Addition
to
[11.22
alter
the value of A.
we obtain
a x + ka 2
+ kb a
bx
a9
first
row,
A.
b9
The symbol
ax
A=
a 2u
b9
defined to
is
&1
c3
a2
Co
as
c3
mean
+ CX
a%
b2
a3
b3
(i)
is
equal to
~ b i( a 2 c 3 ~ % c 2) + Ci(a 2 6 3  a3 b 2
= a 1 & 2 C3a 1 63C 2 + a 2 63C 1 a 2 6 1 c 3 + a3 6 1 c 2 a36 2 c 1
a i(h c 3 ~ h c z)
on rearranging. This
is
the expression
(ii)
of 11.12,
and
is
(ii)
sometimes
shortly as
The
^a
definition expressed
by
how
(i)
bj ck
ax
is
coefficient of
is
We may
(i)
are constructed
from A: the
11.23]
395
The diagonal running from top left to bottom right is called the
leading diagonal, and the product a x b 2 c z of the elements in it is called
the leading term in the expansion of A.
Examples
(i)
+3
= 4(842)5(421) + 3(1212)
= 51.
(ii)
h
+9
(ii)
= a 2
which
is
b3
+h
c3
ax
cx
ax
bx
as
c3
az
bz
(iii)
A=
az {\ c 2  6 2 c x )  b z {ax c 2 a 2 c x ) + c 3 (a x b % a z b x )
bx
cx
b2
c2
h
ax
cx
a2
c2
+ cz
ax
bx
a2
b2
(iv)
we may arrange
A=
=
is
(ii)
e.g.
a x (b 2 c z b z c 2 )a 2 (b x c z b z c x ) + az (b x c 2 b 2 c x )
a.
a*
first
+ 0
(v)
column.
GPMII
396
11.24 Properties of
We now
[11.24
first, all
subsequent properties
Proof of
(I).
A'
i a
az
bx
b2
b3
Ci
Ca
Co
&3
c3
C3
^2
+ 3
a.
a^&gCg  6 3 c 2 )
 a2
& i c3
bx
&2
Cl
C2
by
definition,
(v) in 11.23.
the determinant
Alternatively,
of the
first
is
we may
6
a
bx
e.g.
interchange
Cj
3 (&2 C l _ &1 C 2) 
i(& 3 c 2  b 2 c 3 )
Ma
2 Cl
 a l C t) + C 3
a2
 a i 6 2)
 h(a 3 c 2  2 c3 + c i( a 3 & 2  H h)
)
11.25]
row,
we
get
a x + ka 2
397
if
bx
+ kb 2
cx
+ kc 2
(vi)
The row
'not mentioned'
expand from
b t + kb 2
it,
cx
(iv)
to
+ kc2
60
is
+ kc
cx
ax + ka 2
a2
bx
+ kb 2
b2
+ c.
a9
b9
CL
C2
Remarks
There are
//
rt
^k 1 r1 + k2 r2 + k3 r3
A > kt A
then
(i
or 2 or
3).
11.25 Examples
(i)
By
By
Evaluate
rs > r3  ra ,
we
get
35
29
86
36
31
87
38
32
89
35
29
86
36
31
87
35
29
86
may be
r.
r2
followed
282
398
The
last
determinant
is
now
easily
29
35
[11.25
+ 86
= 35.329.0 + 86(3)
=  153.
(ii)
Prove
+c
By rx
>
l*x
+ r2 the determinant
is
a+b
+a
0.
equal to
(a + b
=
(iii)
+ c) a
Evaluate
Xs
z3
two rows
identical
Direct expansion would be easy, but the following method has the advantage
of giving the result in factorised form.
By c 2 c 2 c l5 followed by c 3 > c 3 c x the determinant is equal to
,
yx
z
z
x
xz
= (yx)(zx)
y 2 + xy + x 2
Xs
by removing the
factor y x
first
from column
2,
and
%
y + xy + x
2
x) (z x) (z 2 + xz y 2 xy)
(y x)(z x)(z y)(z + y + x)
= (y  z) (z  x) (x  y) (x + y + z)
=
=
x from column
3.
Ex
is
1
(yx) (zx)
+ xz + x 2
z2
z*
+ xz + x 2
(y
it is
on
on arranging
cyclically.
row
or column,
11.25]
(iv)
399
a2
ca
ab
c2
be
we have
be
6 in
1,
a2
a3
63
abc
ca
bz
abc
62
63
a6c
aa
a"
62
6s
abc
abc
abc
ab
first
ax
62
o2
62
63
C8
Ca
&3
Remarks
(a) In expanding a determinant from a given row (or column), the amount
of calculation is reduced if row and columnoperations can be used to introduce
one or more zeros into that row (or column), or indeed elsewhere also.
(ft) A determinant having a complete row or column of zeros has the value 0.
Exercise 11(a)
Evaluate the following determinants by direct expansion.
Ill
4
h
9
2
3
2
xs
Xi
a?
2/i
2/2
~f
7 Show that
1
cos(w l)x
coanx
cos(n+l)a;
sin(w
sin
nx
sin(n+l)a;
l)o;
(1
2acosa5 + a2 )sina?.
a2
62
c2
ca
ab
ax
by
cz
zx
xy
be
ax
by
cz
yz
abc
111
y
be
ca
ab
400
a+A
10 Verify that
+A
+A
= \* + (a + b + c)X 2 + (bc + ca + abpg*h*)\ + {abc + 2fghaf i bgi ch*).
Using
13
11
14
10
15
16
12
12
b c ca a b
14
15
101
19
102
20
12
cab
103
20
bc
ca
15
b+c
c+a
20 Prove
EE
a2
x1
Xs
{a
+ b + c)(a 2 + b2 + c*bccaab).
bx
19
a2
22 Prove
X*
16
111
18
cab
21
13
13
a b a+b
17
result
(6
6.
c) (c a) (a 6).
ax
Cj_
a2
o2
a3
+ la x
6 2 + la 2
6 3 + las
bx
c1 + mb 1 + na1
+ m& 2 + na 2
c 3 + m& 3 + na8
ct
(This shows that to each column we may add multiples of the preceding columns.)
What
a 1 + Xa 2 +/iai
(i)
bj^
a 2 + ra8
+ Xb^ + fiba
K + vb.
Cj^
+ Xc^+flCg
c2 + vcs
b,
+ A&i
&!
C!+/t&i
a 2 + A6 2
62
c 2 +/t& 2
&3
c 8 +/t6 3
aj
(ii)
a8 + A& 8
23 Solve
[c x >
3*
2a;
4jk
0.
11.3]
401
1
1
ca
ah
first
a+x
b+x
aa
62
+ c)
ca(c + a)
ab(a + b)
bc(b
+c
c+a
a+b
b
determinant.
+x
a+y b+y c+y =
25 Prove
abc
(bc) (ca){ab)(xy).
ca
26
relation
27 Prove
c^
+ a^
6i
+ 2/x ^ + 2!
a2
62
Cjj
a3
63
cs
2/i
+
a.
&8
a general property.
If each element of A consists of the sum of two terms, prove that A
28
(i)
equal to the
is
sum of 8
29
By
using no. 28
D=
(i),
show that
a1 + A6 1
fti+ztCj
c1
a 8 + A6 2
6 2 +/tc 8
c2
a3 + A68
&3 +/*c3
+ va1
+ m2
c3 + va z
=(l+\/iv)
ax
6X
cx
o2
62
c2
a8
6S
c3
We may be
not left the value of A unaltered. It is essential in applying Property (5) or its
extensions to leave at least one row or one column unaltered at each step.
11.3
11.31 Definitions
and notation
all
its
is
row or
[11.31
DETERMINANTS, LINEAR EQUATIONS
secondorder determinant obtained from A by deleting the row and
column which contain that element, and prefixed by the sign + or
In some cases the signs run + instead of the standard H
402
1
lies.
(b)
The
a2
c2
az
c3
>
element in the
cofactor of the
(_
is
c3
ith.
by (
is
is
a2
1)1+2
az
^3
that of c z
ax
=+
(1) 3+3
c3
a2
&2
Notation.
sions of A can
first,
2,
lf
A=
and
1?
for
a^A^
+ b^B^c^
(i)'
a 2 A 2 + b 2 B 2 + c 2 C2
(iii)
a 3 A s + b s Bz + c z C3
(iv)
'
'
columns we have
A=
a x A x + a 2 A 2 + az A z
bx
B + b 2 B2 + b z Bz
cx
Cx + c 2 C2 + cz Cz
(v)'
11.23.
11.32]
11.32 Expansion
by
403
alien cofactors
a2
b2
a9
b9
a3
^3
It has
=
=
c2
a2
b 2
c3
c2
a3
a 2 A x + b 2 Bx + c 2 C1
+ c2
az
A x Bx Cx are cofactors in
where
A=
u2
u2
a*
bo
'
of
1 1 .3 1
for expansion of
A by
'
true cofactors.
'
Examples
(i)
A=
B = acg
C = abh
A = bcf
F = ghaf, G = hfbg, H =fg ch.
2
different expansions
cofactors,
by true
404
[11.4
a x + ka 2
6X
+ Jcb z
CT<j
62
aZ
^3
cx
+ Jcc s
Cjj
C3
= (a 1 + ka 2 A x + (6 X + kb % Bx + (c t + fcc2 Gx
= (a 1 A 1 + b 1 B 1 + cx Cj) + Ha^ + 6 a B x + c 2 Cx
= A + k.O = A.
)
No. 27 of Ex. 11
11.4
(a)
(11.1),
ax x + bx y + cx z
a2 x + b 2 y + c 2 z
= d2
a3 % + b3 y + c3 z
dx
d3
ax
bx
^3
cx
A=
and add:
(a x
A x + a2 A + a3 A 3
2
x+
(b x A x + b 2 A 2 + b 3 A 3 y + (c x A x + c 2 A 2 + c3 A 3
= d x A x + d2 A 2 + d3 A 3
)
The
coefficient of
y, z are zero.
is
A, while
by
Hence
dx
X
bx
d2
d3
b3
cx
Co
because the
first
first
11.41]
405
sides
Az = A
(1)
where
(1)
denotes
Multiplying
^=
==
0,
A
= 
A
(1)
The
case
(3)
.
A=
will
A< 2>
'
A(2)
=A
Az
similarly
If A
is
=A
(3)
Z
'
be considered later
'
(11.42).
Example
5,
3x8y + 6z =
4,
Sx2y9z =
12.
2x + 4ySz
Solve
Here
A=
= 6x
2x 3x
1
4
1
3
by cx ^CjCg,
followed by c 2
>
c a + 2c
= 6x49;
and
A<
3
8
2
9
12
= 2x3x2x
2
1
12
12 x
2
12
1
= 12x7(6+1)
3
12 x 49.
=A
(1)
/A
2.
which
=A
(2)
,
Az = A <3) What we are now asserting is the converse of this, viz. that if there are
numbers x, y, z which satisfy these three derived equations, then they also satisfy
the given equations; see Ex. 11 (6), no. 16. The checking of solutions is thus not only
a practical precaution but a logical necessity.
.
406
Similarly
(2)
and
53
12
The
11.42
case
3 x 49,
so
h
2 x 49,
so
z
38
2
12
9
4
2
(3)
[11.42
when A = 0:
and indeterminacy
inconsistency
The proof of Cramer's rule shows that, if the given equations are satisfied by
(x,y,z),
then
Aa;
When A =
it
=A
Ay =
(1)
,
A< 2>,
Az
A< 3 >.
A<
A<
2>
A<
3>
we must have
0.
(i)
If one or more of the determinants A(1) A(2) , A (3) is nonzero, there can be no
common solution; the equations are inconsistent.
Assuming that the consistency requirements (i) are satisfied, two cases arise.
,
(1)
At
least
Suppose
C8
a^ + ^y =
a2 x + b 2 y
unknowns J x, y:
dx c x z,'
(")
d 2 c 2 z.
z is
+ b x y + c x z d1 =
+ c.2 z d 2 =
a3 x + b s y + c s z da =
a1 x
a z xQ + b y
and
fl
Suppose that
To show that
prove u = 0.
0.
u.
we have
to
Multiply the above equations by the cofactors of the elements in the last
column of
A =
(3)
dx
ax
6X
a2
62
<2 a
a3
6,
'3
dn
and add:
(a x Ci
Hence
+ a 2 C 2 + a 3 Gs x + (b G x + 6 2 C z + b 3 Gh
)
A<
>=C3 w,
+ Az A< 3 = C3 u.
>
11.43]
407
by
the coefficients of x and y being zero by 'alien cofactors', and A =
hypothesis. Also C3 4= by hypothesis, and A<3) = by the assumed consistency
of the equations. Therefore u = 0, which proves that the solution of the first
two equations for x, y in terms of z will be also satisfy the third.
Since the value of z is arbitrary, we say that the solution is indeterminate
with one 'degree of choice', or that there are oo 1 solutions.
(2)
All cofactors in
A are zero,
We may suppose a1
4= 0.
x in terms of y and z.
obtained for x be x then
for
Then the
When
y,
Lis nonzero.
Xq
and
so
a2 x
+ b2 y + c2 z d2 =
^ifri2/o c i go
,
a *(di  &i y
 Ci z + ai 6 2 Vo + ai c 2 2 o  a i d 2
a 2 d1 a 1 d 2
+ G3 y B3 z
a 2 dx a 1 d2
by
hypothesis. Similarly
a3 dj
will therefore
a1
dx
aa
d2
a1 d 3
be inconsistent unless
ax
dx
a3
d3
If both conditions are satisfied, the solution of the first equation (with y and z
arbitrary) will satisfy the second and third also. The solution is indeterminate
with two 'degrees of choice'; we may say there are oo 2 solutions.
The results just proved are intuitively evident; for when all cofactors in A
are zero, the coefficients in the lefthand sides of the given equations are
'proportional'. Clearly the equations will not be consistent unless the righthand sides are in the same proportion; and then the three equations are equivalent to only one.
In 21.62 we shall illustrate the results of this section geometrically.
11.43
(1)
Homogeneous
When
linear equations
Clearly this
is
always
a^ + bj^y + ^z =
0,
a 2 x + b^y + c 2 z
0,
a3 x + b a y + c 3 z
0.
satisfied
by x =
0,
0, z
0.
all zero,
we
408
Theorem
x, y, z
which
satisfies
is satisfied
[11.43
by values of
(x, y, z)
Ax =
Since x, y, z are not
0,
all zero,
Ay =
Az
0,
0.
we must have A = 0.
only solution is x =
Corollary 1(a). If A
=j=
0, the
ax x + b x y + c x
0,
a2 x + b z y + c2
0,
a3 x + b 3 y + c3
0.
0,
0, z
0.
solution, then
A=
0.
Corollary
1(c).
If
is
4= 0,
indicated in Ex. 11
the
(b),
no. 14.
inconsistent.
is (0, 0, 0),
system; for
be
Theorem
7/ A =
by values of x,
are associated
Theorem I we have
As a converse
II.
(c)
(15.42).
0,
11.43]
(1)
x = XA Z
= AB3
is
and
(ii)
a2 A z + b 2 Bz + c i Cz
(alien cofactors)
az A z + b z B z + c z Cz
(A
Suppose
all cofactors
of A is nonzero, say c x
x
Choose
Ac 1?
Let
=# 0.
in
=j=
0.
Then
= AC3
409
by
arbitrary.
hypothesis).
x,
= /ic x
(so
and
4=
/i =# 0).
(Aaj +/&!).
is
ai Xc 1 + biix& x c % Xa x c^pib x
X{a % c x a x c^+iL$ 2 c x b x c^
= XBz fiA z
=
by
is satisfied
by the above
values.
solution in ratios.
ax x + b x y + c x z
have a nonzero solution x
alien cofactors, or
a2 x + b 2 y + c 2 z
0,
= XA Z y =
,
by proceeding
written
AjB 3 z
,
= AC3
as in 11.11.
The
this is clear
by
solution can be
y_
B
i.e.
61
c2
c2
ax
ax
a2
a2
61
later, especially in
Ch. 21.
410
Exercise 11(6)
Solve the following systems of equations by
1
2xy + 5z =
x + lyIOz =
a? + f/ + z =
*4 If a,
6, c
2.
Zx + y 4z
2,
x2y3z =
Zx y + 5z =
1,
of determinants.
vise
xyz =
2,
5xy + Zz =
oj + y z =
1,
4.
13,
5,
3.
x+y+z
aa;
1,
+ 6t/ + cz =
= d2
a 2a; + b 2y + c 2z
(2,
[Solve the
the third.]
Show
first
two
for x,
x4y + 2z =
0,
y in terms of
z,
inconsistent:
is
4xZy + z =
9,
and
l.
xyz
3x + Zyz
0,
2x + yz
Q,
is
where A
7
is
1 + 2X, y= 1A,
two
3A,
equations for x,
y in terms of z.]
solution of
3x + y z
0,
can be written
x y + 2z
2A,
7A
=
;
4x 3y + z
0.
13A.
=
=
V2 + V22/ + (A2)z =
than x = y = z = 0. Find
\x + y + j2z
0,
x + Xy + j2z
0,
which
5,
Ax 3y
10 Solve
(if
4,
when A has
3x y
=l
these values.
= 4k,
= 2,
2x + y + z = 2
when (i) k * 0, k
1; (ii) k = 0; (iii) * = 1.
11 Eliminate x, y, z from # + &S/ + cz = 0, ax + y + cz = 0, ax + by + z = 0.
12 Solve Ex. 11 (a), no. 26 by regarding a = b cos C + c cos B, etc. as homox + y + kz
x + ky + z
=f=
a, b, c.
11.5]
13 If the quadratics ax 2 + bx + c
common root, show that
[If
is
the
ax + by + c
common
0,
px z + qx + r =
2
0, lx
411
+ mx + n =
have
0.
root,
px + qy + r =
0,
14 Find from
first principles
a 1 x + b 1 y + c1
a2 x + b 2 y + c 2
0,
b2
c2
b3
c3
a3
15
a3 x + b 3 y + c3
0,
[Assuming a 2 b3 a3 b a
y.
y
and
and
4= 0,
c3
first,
a2
62
a3
63
getting
A=
0.]
homogeneous system
a 1 x + b 1 y + c1 z
0,
a2 z + b2 y + c
ji
a3 x + b3 y + c3 z
0,
16
solve
when A
[The solution
#= 0,
is
Az
given
A=
0.
do
by
Ay = d 1 B 1 + d 2 B 2 + d3 B3
d 1 C1 + d z Cz + d3 C3
Hence
A(a 1 x + b^y + ^z)
d1 (a 1 J. 1 + 6 1 B1 + c 1 C 1 ) + c?2(a1 ^4 2 + 6 1 j52 + c 1 C 2 )
+ d3 {a1 A 3 + b 1 B3 + c 1 C3
= e^A
by true and alien
cofactors.]
Factorisation of determinants
11.5
In
11.25, ex.
determinant
it directly.
is
(iii)
and Ex.
11
(a),
no. 21
Factorise
A=
29
a2
62
c2
GPM II
412
[11.5
If A were expanded, we should obtain a polynomial in a, 6, c which is homogeneous of degree 3 (for each term in the expansion consists of a product of
factors taken one from each row and one from each column). We may regard
this polynomial as a quadratic in a whose coefficients are polynomials in 6 and c.
When a = 6, A = since it has two columns identical. Hence by applying
the Remainder Theorem to A regarded as a polynomial in a, we see that a b
is a factor.
Similarly 6 c, c a are factors. Hence
A=
k(b
c) (c a) (a b),
A and (6 c) (c a) (a b)
are both polyof total degree 3. To obtain the value of k, either compare
coefficients of a particular term, say 6c 2 or substitute particular values for
find k = + 1, and so
a, 6, c, say a = 0, 6 = 1, c =  1.
a, b, c
We
A = (bc)(ca)(a~b).
Considerations of symmetry and skewness.
(ii)
Factorise
A=
a2
62
example (i), the Remainder Theorem shows that b c,c a,a b are
The determinant is a homogeneous polynomial of degree 5 in a, 6, c,
while the product (6 c) (c a) (a 6) is homogeneous of degree 3; hence the
remaining factor P must be a homogeneous polynomial of degree 2 in a, 6, c.
Since A and (6 c) (c a) (a 6) are both skew functions of a, 6, c, hence P
must by symmetric in a, 6, c.
Hence by 10.22, Remark (/?), P must be of the form
As
in
factors.
A=
By comparing
a, 6, c,
l(bc
we find I = + 1. Hence
A=
(iii)
+ ca + ab) (6 c) (c a) (a 6).
(6
to
explicitly.
One example has already been given in 11.25, ex. (iii). As another
we factorise A in ex. (ii) above by this method.
The Remainder Theorem indicates the factors 6 a and c a. By c 2 c 2 cx
and removal of 6 a, followed by c8  c3 cx and removal of c a, we have
1
A=
(6
a) (ca)
(6
6+a
a2
3
+ 6a + a
6+a
a) (c a)
6
+ 6a + a
+a
+ ca + a 2
c
+a
+ ca + a 2
c
on expanding by the
first
row,
11.6]
=
=
(b
(6
+ c+a
+a
a)(c a) (b c)
+a
+ ca + a2
by c x
by
r 2 r i
= (ba)(ca)(bc)
b+a
c 2 and
removal of 6 c,
r2
a2
cx
413
c) (c a) (a 6) (6c + ca + a6)
cyclically.
Exercise 11(c)
Factoriae the following determinants.
1
a2
62
c2
6c
ca
a6
6+c
111
c+a a + 6
c
2
a3
63
c8
be
ca
ab
+c
2
(6 + c)
1
1
b+c
+c
+ a2
a2 + 6 2
6
+a
a+b
1
*8
a
(6
64
*9
c2
(c
+ a) 2
+ c) 2
(c
+a
a+6
(a
+ a) 2
(a
+ 6) 2
ab
ca
6c
+ 6) 2
c2
6
a*
6*
c*
x + ay + a z = az
x + by + b 2z = 6 3
* + cy + c 2z = c8
2
10
11
=
a +6 +c z =
a 3a; + b zy + c3z =
ax + by + cz
a;
?/
1,
1,
1.
*12
+ 6za; + ca^/ = 0,
?/z + zsc + a*/ = ajyz,
a2 /a; + b z /y + c 2/z = j9 2
a?/z
Derivative of a determinant
11.6
ax
cx
&2
^2
^2
da x db x dc 1
dx dx dx
d_
dx
a3 6 3
a*
ba
Ca
a,
da%
db 2
dc 2
dx
dx
dx
Q/<*
&q
Co
bx
c1
b2
c2
da 3 db z dc s
dx dx dx
292
414
Proof. It
is
[11.7
A = I a
l
^ = S
bj ck
Then
da,
.,
^ + Sa^
dbi
6
rfx
dc k
^
+ Sa' 6
Example
If three rows of a determinant whose elements are polynomials in x become
when x a, then (x a) 2 is a factor of A.
First solution. dAjdx sum of three determinants each having two rows
because each determinant in
the same as A. Therefore when x a, dA/dx =
the sum has two identical rows. Hence x a is a factor of dA/dx. Since also
identical
A=
11.7
dx
The symbol
d2
d3
d.
called
a,
fourthorder determinant,
&2
K
&4
This
is
a2
c2
a3
d3
c4
d,
is
defined to
d2
c2
d3
+ <a
d,
c4
mean
a2
&2
d2
a2
c2
a3
h
K
d3
d, a3
c3
d*
a4
a4
c4
11.22.f
t Determinants of order
'
n can be defined
n 1.
11.7]
415
The
proofs of these statements are not always easy, and for determinants of
order four or more it is better to approach the subject from a more advanced
point of view than we can consider here.
Exercise 11 00
Write down the derivative of
1
2x
4 If u'
3x 2
= du/dx,
etc.,
X3
log a;
x*
1/x
2x
ljx*
2
prove
w'
u'
w'
w"
u'"
dx
5 If u,
v,
3,
prove that
w"
is
than
[Derive
3.
A wo x four times,
6 Prove
a2
dxdy
X
7 If
A=
X
X2
by rows, regarding
[Derive
2(yx).
prove
3A
SA
8A
ox
oy
oz
z2
on the rows.]
8 If
8A
A=
find
8x
xs
y*
8A
8y
8A
8z'
zA
*9
*10
3
*11
1
*12 Prove
where the notation is that of 11.32, ex. (i). [Expand by the last column, then
by the last rows of the thirdorder determinants so obtained.]
416
*13 Solve
by determinants:
3x +
2/
5z
+ 2z + Zt
y + Bzx + 2y
Prove
(i)
(ii)
z2f
o,
o,
14.
r3
r2
r4
r3
r*
r5
r2
r4
r6
r3
r6
r9
r3
Factorise
= r* (r l) 3 (r 2
A=
x%
a2
ft
c2
a3
63
c8
a;
a?,
x 2 x in
,
is
c2
62
ca
62
a2
ah
be
c
a+6
Z
+m
p+q
abc
+c c+a
m+n n+l
r+p
q+r
A and
the
coeffi
417
Factorise
be
b+c
ca
ab
a2
62
c2
+a
a+b
1+a8
1 + 68
1+c8
62
+ c) 2 (c + a) 2 (a + 6)
& 2 c 2 + a 2d 2
bc + ad
1
2 2
2 2
c a + & a"
ca +
1
a 2 6 2 + c 2d 2 a& + ca" 1
(6
fta"
(6
factors of
+ c) 2
62
(c
ca
+ a)
c2
62
and hence
+ 6)
c2
a2
62
c2
and
(a
factorise it completely.
cos d
sin 20
cos 2
sin 6
sin 2(9
sin 2 6
sin
sin 2
cos 2 6
11
Express
ap + br
aq + bs
cp + dr
cq + ds
X
(This gives a rule for multiplying
Solve
12
x+1
2x
Sx2
2x
13
0.
X2
a3
62
as 6
(assuming that
a, b, c are
a8 c8
14 Prove that the vertices of the triangle formed
a 1 x + b 1 y + c1
0,
a 2 x + b 2 y + c2
0,
by the lines
az x + b 3 y + c3
418
15
x 2 + y 2 + 2gx + 2fy + c
is satisfied
by the values
Show that
16
 2),
of
(2, 4)
=
Interpret geometrically.
(x, y).
the equations
2x + 3y
= 1, Ax2y
4c 156c 439 > 0.
3x + Xy
4,
c satisfies
when
(i)
+ + 3z = 4,
2; (ii) a = 1,
+ 2?/ + 4z = 5. ay + az =
= 2; (iii) a = 6 = 1.
a;
2/
18 Find the conditions which A, /i must satisfy for the following system of
equations to have (i) a unique solution; (ii) no solution; (iii) an infinity of
solutions:
x + y + z=
x + 2y + 3z
e,
10,
x + 2y + Az
= p.
= 6,
= 12,
3x+12y + (X+10)z = k,
unless k is suitably chosen. When
(\ + 2)x + 4y + 3z
2x + (\ + 9)y + 6z
have no solution
is
20 If/,
g,
L ==f+g + 2h,
^=
8^ = 7/+8sr + 24fc,
2g + 2h,
Xx
Xx (where A is independent
find all solutions of the form/ = a eAa! , g = b e , h = ce
ratios of the
corresponding
and
the
of
A
values
of x), giving all the possible
constants a,
b, c.
21 If
W(x,y,z)=
where
of t and x
x, y, z are functions
W(x,y,z)
dxjdt, etc.,
prove that
x W(l.ylx,zfx).
22 Write
1
V1 =
a 38
b3
c3
x+b
x+a
and
(i)
(ii)
(iii)
Prove
f(x)
v3 =
a2
b2
c2
x+c
+ b) (x + c) 2
(x + c) 3
(x + b) 3
(x + a) 3
= (x + a)(x + b) (x + c) Vz
(x + a)
f(x)=
(x
Calculate/'fa:),/^*),/'"^) as determinants,
4.
Vr
(r
+ c)V3
=/(0)
1, 2, 3)
(Cf.
+ VyX + V2 x2 + Vs x3
is
15.)
419
12
SERIES
The binomial theorem
12.1
The reader should already be familiar with the work in this section;
a positive
integer,
+ a) n = xn + n Cx xn ~ xa + n C2 xn ~ 2a 2 +
number
+ n Cr xn^ar +...+an
. .
n(n 1)
...
(n r+ 1)
'
r\
...
in each bracket
(x + an ). To expand
by the terms in the
we obtain xn
we
multiply the aterm from one bracket and the a;terms from
all
possible ways,
+ a 2 +...+an )x
we obtain
n1
.
Next, multiplying the aterms from two brackets and the a>terms
from the rest, and doing this in all possible ways, we obtain
(a x a2
+ a x a3 +
+ a2 a3 +
..)xn
~2
,
~
where the coefficient of xn 2 consists of the products of different
terals
therefore nG2
from the
en
coefficient is
{a x a 2 az
+ a1 a 2 ai +
..)xn
arterms
~3
,
~
where the number of terms in the coefficient of xn z is n C3
In general, if we multiply aterms from r brackets and ajterms from
.
all
possible ways,
r
,
we get
series
420
brackets
ax a2
is
...
an
[12.12
all
nC a 2
2
na,
nC
az
n C ar
r
(1)
(2)
(3)
nC
nC
r\{nr)\
= n Cw ~r
by r =
(4) If
If n
is
(5)
is
even, there
is
This
is
a for a in
(x  a) n
= xn  n Cx ax
71
1
. .
+ xn
(i)
Find
+ n C2 a 2x n ~ 2
+  l) r nCr arxn ~r +... + ( l) n an
if
^_^
+ x 3a; 2
5
)
(2c)
18 '
18a'(
3
r213 "ral8 ~ ar
)
13 2r
13
Expand
Bjx) ls
is
i3Cr
(2
Examples
(ii)
%(n 1).
12.11,
12.13
result of 12. 1 1
an + nCx an xx + n Cz an 2x 2 +
useful in approximations
(6) If we write
\n.
(2
. .
series
12.13]
all
421
32 + 5. 16(x 3a; 2 ) +
8x 2 (l
On ignoring all
x*.
. .
expansion of (1
+ 3a;) 7
is
u1 + ui x + u3 x + ...+uB x7
2
7!
then
ur
= _ __
7!
r!(7r)!
(toy
/
v
'
(r
1)1
(8r)!
1
(3*)*v
'
= 8r 3x
r
.
if
l.
24 3r gi
i.e.
1,
1, i.e.
according as
4r,
i.e.
3f
4r
Hence
if r
ur+1 < ur
i.e.
r.
3, ur+1 > ur i.e. w4 > u a u a > u 2 and u.2 > u v If r > 4, then
m 5 < uit m 6 < w6 etc. These inequalities show that w4 is greater
,
(iv)
Cs(i) 3
n
Prove
is
is
35x(f)3.
Gr _Y + n Gr =
n+1 G
r.
+ x)(l+x) n = (l+x)+\
in which the coefficient of xT on the right is n+1 Cr On the left, terms involving
xr will arise from 1 x n Cr xr and x x "C^ xr~ x so the total coefficient is n Gr + n G x.
(l
The
result follows.
it, is
(l+x)
(1+x) 1
(l+x) 2
12
13
1
(1+*) 3
(l+x)*
(1+a;) 6
1
1
1
1
10
10
SERIES
422
[12.2
Exercise 12(a)
Write out the expansions of
2 (x+l/x) 5
(3*2) 4
Expand 1 + 2x x 2
(lx){l+x) i
6
)
term in xl
coefficient of x s in (2
coefficient of x~ in (x  1 /2x 2 1B
term independent of a; in (2a; 2 1 fx) 12
10
6thtermin(3c+l/c) n
20
2 6
6 coefficient of a; 10 in
a; )
11
if
the coefficient of xr in
+ x) 19
coefficient of x 3 in (2
12 Find r
1 \x
+ x) 20
(104) 5 ;
(i)
is
(ii)
+ x  3a; 2
(098) 5
7
)
~
twice the coefficient of xf x
(l
+ 2a; +
a;
)(l+a;) n
(l+a;)"+ 2
+ x) n = 1 + x) m+n to prove
+
m C n C + m C _ n C + ...+ m G n Cr = + n Cr (r^m,r^n).
r
r x
1
By considering the coefficients of x n in 1 + x) n 1+ x) n = 1 + x) 2n prove
17
x) m ( 1
where
cr
denotes
Cr
+ c + c +...+c =
2
^L,
18 Prove
(1
Show that
19
(2
...
+x n
coefficients of x
+ ^3) 5 + (2 JS) 5
is
rational,
and
20 Prove that
v
x + n(x + y)+
n(n
v
1)'
(x
n
+ 2y)+
x
n(nl)(n 2)
(x + 3y)
=
12.2
2 n ~\2x + ny).
Finite series
The reader
progressions,
work.
12.21 Notation
ur denotes the
and
rth.
definitions
term of a
series.
n
s
T, ur denotes the sum of the first n terms, and is called the sum
n~ r=l
%, u 2
.
,
un
Or sequence
is
considered in relation
series
12.22]
If ur
= ar xr
Later
we
then the
series is called
shall write s
423
a power
series in x.
call
7l>00
the sum to infinity of the infinite series T,ur For example, the
.
the
first
n terms
1,
x2
x,
x3
and if
<
\x\
1,
then xn >
n
= lx
t^;
summing
Given a
series,
our problem
terms.
The methods
first
of
(x 4= 1)
...
when n^ao
the
sum
is
a).
finite series
(l)(3) listed
below
will
sum
of
be illustrated
in this section.
known
Derivation or integration of a
value for
(2)
finite
x).
The
difference method. If
ur ^f{r+\)f{r),
then by taking r
n,
1,
n 2,
. .
.,
2, 1
uni =f(n)f(n
in turn
we have
1 )>
/(!)/( 2),
2 =/(3)/(2),
%=/(2)/(l).
and
By
adding,
IX =f(n+ l)f(l).
r=l
proving a stated
(3)
Mathematical Induction
(4)
applicable to
(for
result).
series
424
[12.23
(l+x) n
Used
thus, the
numbers
nf]
nfj
nf)
is
nfl
0'
C).
all,
(:>
G)
is
c 0>
Thus
(1
The
+ x) n =
c 2
Cn
'
+ qa; + c 2 a; 2 +
. . .
+ c n xn
(i)
binomial coefficients:
series as
(a)
identity based
Examples
(i)
Sum
Method
(a)
(c
= nx\\ + {nl)x +
{
(nl)(n2)
^
.
,)
1
xt+.^+x"
= nx(l+x) n K
Therefore the given sum = (1 + x) n + nx(l + x)" 1 =
Method
+ a;)"1 {1 + (n + 1) x}.
(b)
a;
+c
a;
+ c2 a; + ... +
8
(i)
c n a; B+1
by x:
= x(l+x) n
1:
wo x:
(iii)
not a power
series.,
2c 1
Sum
Method
+x) n + nx{l+x) n ~ 1
(1
result as before.
Sum
is
425
This
series
12.23]
(ii)
but
it
...
3c a
in ex.
+ ic x + ic 2 + ... +^ c.
n+
(b)
(i)
from
to
1
1
~
c a; n + 1 l = [j(l +*)+ 1
[ c x + fax* + $ct x* +... + !
n+ 1 n
[n + 1
Jo
Jo
L
+ cx + Jc2 +
i.e.
Method
=
=
cn
n(n 1)...2.1
(n+l)n
(n+l)n(nl)
^
h^r +...+
1.2.3
1.2...n(n+l)
(n+1
J_
n+l {(
1
1)
(c).
...
+ cn _1 cB
Consider
Coefficient of x n
"i "n 2
(2+ 1 l).
"I"
1^3
+ 1) +i_l }
0^ + 0^ + 0^ +
Sum
*o "n i
n+1
+
nTT\"l~ ~r2~"
n+1
(c
..
n(n
n
1.2
Method
(a)
_
The given sum =
(iv)
(i)
Coefficient of x n_1
+ c n _i c =
Cq Cj
on the right
a;
n
)
= (l+x) n (l + x) n
= (l+a;) 2 ".
is
+ Cj c 2 +
. . .
+ c B _j c B
because
cr
<?_,..
is
(2n)!
(nl)!(n+l)!'
(c).
Exercise 12(6)
1
Prove c
+ c 1 + c2 +...+c = 2 n
+ c 2 + c4 + ... = c 1 + c 3 + c5 +... =
2 Prove c
(i), then use no.
fl
1.]
2 B_1 .
[Put
a;
=1
in identity
^
series
426
3 Prove cx + 2c 2 x + 3c8 x 2
+ nc n x
. . .
n~x
[12.24
n( 1
+ 2c 2 + 3c3 +...+nc n
c1
. . .
7 Prove
c cr
+ CiCr+1 + c 2 c +2 +
. . .
What is the
value
if
9 Prove
cl
Sum
10
c c1
Find
11
when w
is
even,
(2nl)\
,
,
_a _
=
+ Scl+...+ncl
+ 2cl
{(n1)!} 2
,
+ 2c 1 c 2 + 3c 2 c3 + ...+nc n _ 1 c n
a;
oddl
is
x*
n
++ +
r2 Co+
F4 Ca
Ci
(w
'
Ci
K_
. ,
1U
..to(n+l)terms
,
3.4
2.3
1.2
*12 Prove
c2
<h
2+ 2 n3
(n+l)(n + 2)
+ _ i)ni! Cn = A + J+ + ... + I.
i.
...
+ c2 x +
. . .
[Write
*14 Let
(i)
ar
(l
+ a; +
2
t;
+ a1 x + a 2 x 2 + ...+a 2n x 2n
ljx, a; in
place of x,
and prove
a 2n r
(ii)
(iii)
ala + a\...+a n =
2
(iv)
(
Calculate a
v) Calculate
+ a3 + ...+ a 2n i
+ 2ax + 3a 2 +... +
2n +
1)
) (
an
a 2n
 x + x2
to prove that
or 2 or 3.
r=l
n
(1) 2 r.
r=l
This
is
an
a.p.,
427
series
12.24]
n
(2)
2>2
First method.
r=l
 (2r  l) 3 s
(2r + 1) 3
we have by putting r =
1 , 2,
. .
24r2 + 2,
w and adding
(2rc+l) 3 l 3
= 24f>2 + 2w,
r=l
24 S r 2
/.
r=l
2>2 =
and
+ 1) (2n + 1).
r=l
(r+l) 3 r3
and summing
for r
1, 2,
. .
.,
n,
(rc+l) 3 l 3
3r 2 + 3r+l
we have
= 32r 2 + 3r + rc.
r=l
r=l
From
(1) this
becomes
(3)
Sr 8 We
.
have
r=1
Summing
1, 2,
. .
.,
n,
we get
{n(n+l)} 2 0 2
= 42^,
r=l
a
S* = ia (+l)
so
r=l
is
the square of
r=l
also
series
(2).
this
(2).
difference
way.
GPMH
SERIES
428
[12.25
Sum
(i)
1.2 + 2. 3 + 3. 4 + ...+w(n+l).
1).
Consider
r(r+l)(r + 2)(rl)r(r+l)
This
is
r(r
3r(r+l)
for r
1, 2, ...,n,
r(r+l) = n(n+
i.e.
1)
(w +
2).
r=l
(ii)
The
Sumton
terms
+1
wr =
)
(r
r(r
+ 1) (r + 2).
....
Consider
+ 2) (r + 3  (r  1
)
r(r
+1
(r
+ 2)
4r(r+l)(r + 2)
4wr
Hence
r(r+
r=l
1) (r
; r ( r+ i)( r + 2)
i.e.
r=l
The same method can be used for any series in which the terms consist of
the same number of factors and the first factors in each term form an a.p.
having the same common difference as the successive factors in each term.
To obtain a difference function f(r +1), insert an extra factor at the end of the
rth term. Observe in exs. (i), (ii) how the sum can be written down from the
form of the general term.
Other series may be reducible to this type.
(iii)
Sum
...
to
n terms.
ur = r(r+l)(2r+l)
= r(r+l){2(r + 2)3}
=
/.
**r
r=l
22
r=l
2 x \n(n +
%n(n +
Alternatively,
2r(r+l)(r + 2)3r(r+l).
1)
1)
(n + 2) (n + 3)
(n + 2)
ur
3
x \n{n +
1)
(n + 2)
after factorisation.
2r8 + 3r 2 + r,
is
consistent with
series
12.26]
n
r=l
r=l
r=l
2 x n2 (n +
429
n
r=l
2
1)
by the results in 12.24, and this reduces to the expression just found.
The second method is inconvenient if uT is of degree higher than 3 in r, unless
n
we know formulae
for
2
r=l
4
*
etc.
ill
++
Examples.
o
tfwm
by
method
terms.
r(r+l)
r+1
"11
is
and adding,
y\ut =
T^ +
^ z^
+
ur =
n+1
r=i
Sum
tow
+...
o 4
Z o
^5
and
partial fractions,
difference
n+1
tontervm
r(r+l)(r + 2)
would lead to
three
11
(r+l)(r + 2)
r(r+l)
Hence
1.2
{r
(n + l)(n + 2)
is
+ 2)r
r(r+l)(r + 2)
n
*"
2 T\ u
f
2(n+l)(n + 2)
The terms of the series just considered are the reciprocals of those illustrated
The same method as in ex. (ii) here can be used to obtain a difference
function f(r), viz. write down ur and omit the last factor in the denominator.
This gives ur in the form/(r) f(r + 1 which, although not exactly the difference
considered in 12.22 (2), does enable the method to be used.
in 12.25.
Other
series
(r
r+l
+ 2)(r + 3)(r + 4)
(r
+ 3)(r + 4)
vr
w
r,
(r
(r + 2)l
+ 2) (r + 3) (r + 4)
1
(r
+ 2) (f + 3) (r + 4)
say.
302
'
series
430
1
1
_
rT3~r + 4 ~
Since
aSince
(r
After reduction,
we
r + 2)(r + 3)(r + 4)
(
r _l
3.4
(n + 3)(n + 4)
2Y.wr =
hence
2n + 5
riP'rl' r=l^"24.
(2)
= 2wr
find
series
~~4~n + l'
2
+ 3)(r + 4)
(r
"
Vr '
+ 2)(r + 3)
[12.26
summable by the
difference
'
fraction
less easily.
Example
(iv)
Sum
T^ + ^r5 + si^ +
t0ntermS
'
ABC
2r+l
= + T+
r
r+1 r + 3'
we find A = J, B =
C = f.

r(r+l)(r + 3)
(4.62)
6w r
n
and
62*. =
1\
n+l n + 2 n + 3/
+ i + l) + 3(i + i) + 
2(l
^r) + (* + * + Sr + S+i)
\r = 4 r
r=i r
+ *+*+
/
n+3
i
22 + 3S
5S
=2*"
*(
5
r+3
=ir+l5S
rir+3
n+1
r=l*'
r+1
n
22 + 32
rir
n 1
r=i
2

37
ZuUr
r=1
To sum this
series
37

36
1

3(n+l)
by the method of
6(n + 2)
12.25
6(n + 3)
we should write uT
as
(2r+l)(r + 2)
r(r+l)(r + 2) (r+3)
in order to
make
+ 5r + 2 =
2r(r+
1)
+ 3r + 2.
431
series
12.27]
Thus
2
r
and each
(r
+ 2)(r + 3)
(r+
1) (r
+ 2) (r + 3)
r(r+
1) (r
+ 2) (r + 3)'
Success of the 'partial fraction' method in this example depends on the fact
that the sum of the numerators of the three partial fractions is zero, so that the
major part
 f tne
sum
when the
r=4 r
degree of the numerator of ur is at least two lower than that of the denominator.
(3)
Sum
The
to infinity.
series in 12.24,
Some
12.27
(ii)
(iv) ft.
a,
(iii)
trigonometric series
of sines or cosines of angles in a.p.
(1) Series
C=
(i)
i,
by
then since
2 sin
by using one
the form/(r +
1) f(r),
S=
The
. .
0}
.,
n:
a product.
difference into
C = coa{* + $(nl)/3}
(ii)
1, 2,
this expression is of
2 sin \p G
by converting the
and
sinJnyff
"
sin/ff
result
S=
sin{a + (n
Sm
l)yff}
^f
sinf/J
(b)
(c)
(a)
2sin/ff for
finding
(iii)
(i)
wo a.
G'
S'
= smasin(a+
ff)
(i), (ii)
/?;
series
432
[12.27
(2)
= cot02cot20,
tan0
cosec 26
tan 6 sec 26
tan 26 tan 6
sum
to
suitable trigonometric
series.
Example
(xv)
Sum
tan  sec
wr = tansec
tan
2r_1
2=1u
The sum to
infinity is
that
tan r
2
:.
we see
tan 6 tan
2"
tan d.
Exercise 12(c)
23 +
3
... + (2n + l)
+
2
2
2
2 + 3 4 + ... to n
terms
if
l2
+ 3 2 + 5 2 + ... + (2wl) 2
even;
is (i)
(ii)
(n
r=l
5
Expand
(r
\2
=2*+2> 2
Z')
.]
r=l
Sr,
2r2 deduce
,
2 rS
r=l
6
(i)
(ii)
By writing r2 =
Writing r 8
r(r+
r(r 2
1)
r,
n
obtain
r2
r=l
by using
12.25.
n
obtain
f3
r=l
2/i
*"(*+
7 Calculate
r=+l
1).
1)
in the
form
ar(r + 1) (r + 2)
hence calculate
2 r r + 2) (2r
(
1).
r=l
9 Calculate
2 (nr) (r+1).
r=l
[ur
+ 6r(r + 1) + cr + d, and
series
12.28]
433
Sum the following series to n terms, and find the sum to infinity when it exists.
+ 2 2 .3 + 3 2 .4 + ....
10
l a .2
11
+ ....
[Method of
12.25.]
l +2 +3
l +2
I
+ ^ +
14
+ + ^^ +
1.4
4.7
7.10
16
2r+l
19
r ti r(r
to
7
h....
7.9.11
sum to infinity,
r +\
+ 2) (r + 3)
n terms.
20 cos
+....
1
5.7.9
S
"7~T~o\*
+ 2)
f r(r
3.5.7
to find the following
+ ^r^+,
ir^
3.5.7 5.7.9 7.9.11
17
15

r(r+l)(r + 2)
13
12 1.4.7
...
21 cos 2
. .
..
. . .
),
. . .
).
This
is
a general principle
which
Let
TC
always
is
divisible by 2
(n+l)\V.
by assuming the truth
If
(i.e. if
of
<j)
we can prove
the truth of
be true, then
we have
about
k+1
<j>
that
if
<$>
x is
known
to
1,
n is true for all positive integers n. For by taking h
= 2, 2 implies <j> z ; and we can
X implies <j> 2 taking 1c
\
The
<f>
n is
reached.
SERIES
434
[12.28
Examples
n
2 r% W n + 1) (2n +
Prove
(i)
1).
r=l
Suppose that,
some
for
integer k,
k
r =
2
i*(Jfc+l)(2Jfe+l>,
r=l
This
is
Then
fc
^
r=l
!> + (&+l)
2
r=l
=
+ l)(2fc+l) + (&+l) 2 by
= (fc+l){fc(2fc+l) + 6(fc+l)}
= (fc+l){2fc 2 + 7fc + 6}
= $(&+l)(fc + 2)(2A; + 3),
which
if
integers n.
(ii)
Prove
Write f(n)
that 3 2n
+ 7 is always divisible by 8.
7,
3 2n +
and consider
f(k+l)f(k)
Iff(k)
is
divisible
by
16,
Also/(l) = 3 2 + 7
by Induction.
(iii)
Prove
Then
3 2 *(3 2 l)
2 *+ 2
for
forr
+ 7)(3 2 * + 7)
n = k:
k  far +...+a h
G
x
+ r
= (x + a) (x + a) k the
particular value
x k + k C1 x k  1 a +
...
Cr ar + k Cr _ 1 ar =
k+1
which
The
is
a) k+1
a;*^
+ ^C^a +... +
k+1
when r =
Cr
x k r+1ar
when n =
1.
or
k+ 1. Hence
+...+ a k +\
k+1
for k.
follows
Induction.
(iv)
of
Cr ar
coefficient
1,2,...,^;)
k
(x
8.3 2 *.
by Induction.
some
afcr+iig
by
(3
a) k
32fc+2_ 32*
8, this
6.62.
by
series
12.28]
435
we assume that,
some
for
+^ _
^+^+
k,
...+U k =f(k),
u1 + u2 + ...+uk+1 =
Since
(u1 + u2 + ...+uk )
+ uk+1
=f(k) + uk+1
showing that
this is equivalent to
* =/(*+!)/(*).
so that/(r)
is
Exercise 12(d)
Prove
1
by Mathematical Induction.
+3+5+
..
+ (2nl) = n 2
+ 4 + 7 + ... + (3n2) =
2
12
16
8
+...
4
5
6
7
3.5
5.7
(i)
n(n +
(i)
9" 
a?
2 " 1
8 If 2mx
1)
to
7.9
(w + 2)
is
1 is divisible
%n(Gn2 3n
divisible
by
8;
by
(ii)
6;
(ii)
n + 2n is
3
+3
divisible
 8n  1 is divisible by
9n
3.
64.
2u 2
by
by x + y.
1),
and
1).
n_1
n terms = ^,3 + ( mx
1)
2ra
?/2ni jg divisible
= a + 6,
= (n+l)!l.
find limw.
prove
n>oo
9 Prove
cos
sec
+ cos 20 sec 2 +
. . .
+ cos n0 sec" =
s

"*"^
_j
sin 6/ cos"
Remark in
12.28.]
10 Prove
cosec a cosec (a + /?)
to
n terms
is
SERIES
436
12.3
[12.3
Infinite series
12.31 Behaviour of
an
infinite series
n
(1)
= 2 ur an(* um
Convergence. If sn
r=l
we
defined s to be the
exists
n><
sm to infinity
We also say that the series Hur converges (or is convergent) to s, and
write
00
= ur
y
r=l
or
It will be clear
from the
^ + ^2 + %+....
definition that s is not a 'sum' in the
limit of
results
may
arise.
>
series
1
One
is
tempted to regard
this as the
sum of two
l+x + x2 + x*+...
whose sums to
end of
X3
Xs
...,
12.21)
llX
and
infinite g.p.'s
,111
+ ++
and
XT
1/x
x\'
We
which add up to
zero.
thus appear to have a series of positive terms whose
to infinity is zero. The explanation is simply that the first g.p. converges
(i.e. possesses a sum to infinity) only when \x\ < 1, while the second converges
only when l/x\ < 1, i.e. \x\ > 1; there is no value of # for which both converge,
sum
series
has no
sum to infinity.
i++i+i+i+*+= (i+i++ ) + (i
from which
+
i+i+
i.e.
= 1+4 + *+...,
When
sn does
left is less
possibilities, illustrated
series
12.32J
(2)
The
If x
(i)
infinite g.p. 1
=f=
1,
if
then s n
1 < x <
=
1,
if
(iii)
if
then xn  oo
% is odd
a;
then xn >
x > 1,
divergent to +oo;
(ii)
437
+ x + x + x + ....
n )/(l x). Hence by 2.72
x
(1
2
TC
sn
even and xn +
is
 co when
oscillates infinitely;
. .
divergent to
+ oo.
oo,
We
finite oscillation
abbreviate 'Sr
and infinite
is
oscillation.
convergent' to
(i.e.
ty
Lur
c'; similarly,
'wr d'
not convergent)'.
0.
n><x>
oo;
result is false : if
and
un >
since
0,
not converge.
Example
2(l/r) is divergent.
We have
i+i
> i+i =
i++*+i>+i++i = h
i+^+..+^>i% = i,
etc.,
so that
> 1 + i + i = 1 + 2 x,
8 S > 1 + i+i + i = l + 3x h
s 16 > l + 4x,
etc.,
and
8&>>
s4
in general
Hence when p
> oo,
jj>
> oo.
+p x
un =
sn
sn _ v
series
438
[12.32
2 J' 1
S^pi
When n
00, i.e.
< *n ^ s
1/n >
Remark.
cx>
because Sg* 1
when n
is
+ 00), although
> 00.
> 00,
>
1)
(3)
(a)
(b)
This
uiiuu.
series
un
a*
is
unaltered by
These facts are clear from the properties of limits (2.3). Property
(a) is useful when the first few terms of a series behave irregularly,
and also shows that any test for convergence which we shall give need
apply only
definite
(4)
'for all
n ^
term of the
onwards.
to
converges to as + bt, a
For
series
and
sums
s, t
+ bvr
b being constants.
if
r=l
r=l
n
then
r=l
> as + bt
when
n > 00,
by
2.3,
(ii).
Similarly, if one of
2vr converges and the other diverges, then
H(aur + bvr ) diverges. However, if both diverge, T,(aur + bvr ) may
possibly converge, e.g. if ur = r + 2~r vr r, a = 1, b = 1; roughly
,
is
if
Swr
consists of positive
sn
+ un+x >
sn
Hence
a series of positive terms is either convergent or properly divergent to
+ 00.
439
series
12.4]
can be determined
directly because a formula for sn can be found; cf. 12.26 (3). When
no such formula is known, we resort to tests for convergence or
We recall that
its
statement 'limsn
(2.71) the
series converges,
they
sum to infinity.
s'
means that
if
any
>co
positive
(in general
depending on
e <
sn
<
s + e.
12.4
all
factor
1
12.41
Comparison
by
first
Series
removing the
tests
In these we compare the given series 2wr with a series T>vr whose
behaviour (c or d) is known. Roughly, a series which is less term by
term than a convergent series is also convergent, and one greater than
a divergent series term by term is divergent.
Test for convergence.
(1)
If 2iVr
either
or
and
c,
(a)
(6)
un ^
lim {ujvj
n ^ m, where
c is
a positive
constant,
= 1^0,
n>oo
then also lLur c.
00
Proof of (a). If 2 vr
t,
then since
all
for all n.
r=l
v1 + v2 +
By
...
+vn <
hypothesis,
Wm + Um +1++ Un <
<
<
Hence for
sn
all
C{vm
+ Vm+1 +...+Vn
n ^ m,
Therefore
by
12.32
(5),
Swr c.
say.
series
440
Proof of
Given
(b).
that
l
>
however
u
e <
< l+e
[12.41
small, there is a
for all
vn
number
m such
n ^ m.
un <
Taking
n ^ m.
for all
(a),
(2)
If
= l + e in
+ e) vn
(1
1>vr
and
i>,
either
(a)
or
(b)
un ^
n ^ m, where
c is
>
lim (ujvn )
(but
not
a positive constant,
0),f
n*<x>
then 1>ur d.
Proof of
(a).
vm + vm+1 +
By
+ vn
. . .
when
c(m
Swr
sn
.*.
^+
+ um +
. . .
>
(3)
(i)
m+1 +...+f;ft )
> oo.
+ un > oo when
n>co,
 oo.
d.
Proo/ of (b).
If
. . .
t;
when
> oo
take c
hypothesis,
fn+^m+i + + >
i.e.
(5))
oo
0,
then
= I e in
e >
(a),
(l
e)vn
for all
n ^ m.
we may
Standard comparison
series are
if
<x <
and d
if
x ^
(12.31 (2));
of
series,
(hi)
W
/
X(l/rp ) c if p
5
=
<
111
+ + +
i^
2^
3^
(^ ^)
1.
+^
+
(p
there being
++
^)
= 0;
e.g. S(l/r)
terms,
brackets;
+ '
there being
if
(12.32(2))
and
(l/ra 2 )/(l/n)
> 0,
series
12.41]
441
. . .
terms in sn have
Making
an upper estimate for the contents of each bracket, we have whenp > 0:
been replaced by brackets containing
*n
< Jt+
1
I?
1, 2, 4,
. . .
(p+p) + (&+  + p) + 4
positive terms.
( brackets)
+ p + 4^ + 8^ + 
(n terms)
3+
<
terms >
71
{l
 (^i)
/{*
}
~ 2^}
'
for all
' 1
>
if
1,
l(i) 2
in 12.32
(2).
s
n
If p
<
=
0,
if0<p<
11
h
IP
2p
1
h
...
nv
then clearly sn 
>
IjnP
If
(c)
1/n.
Hence
=
np
1
> nx
> 00
when
w > co.
00.
1.
Examples
(i)
Examine
the series
1.3.4
2.4.5
3.5.6
2n+3
Here
n(n + 2)(n + 3)
2n + 3
Sn
n(w + 2)(w + 3)
n.n.n
i.e.
test,
=
2).
if
n>
w n < 3v n
Hence 2w c also.
1/n2 , then
3,
for
n>
3,
and Zvc
series
442
[12.41
form of the
and taking v n
test,
l/w a as
before,
n^oo v n
\n(n + 2) (n + 3)/
nyoo
n2/
Hence Z B c
because Jjv n c.
Mixamme
(n)
x
large,
is
1/n,
w n = n/2n 2 =
then u n
>
\....
33
w8
,
>
2^
^=
hm
_oo w
frallM '
&i
TC
2n2
probably 2w d.]
l/2n, so that
%v n Therefore
^ftemotM*
and the same
+2+3
w3
tt 
We have
so if v,,
28
[When n
+2H
1
\
'
lim
2w n diverges with S n
/n+l/l\
n>oo \
2n 2
/)
n/
\,
result follows.
Exercise 12(e)
un
f+f+ H....
lf!l^ + l^?.
2
a;
a;
+> 0.
a;
V(r+1)Vr*
Using
(iii),
state
6 + !++*+....
Use
s
the
test to
2
32
42
^^
+
2
V( r 2r)
16 Prove that
? +
1+x
1
,
l+xz
1+x 3
3.4.6
K...
+l
6! \6/
if
<x<
1.
5/6
\1 2
6
1
stl.
r
3
14
JL+....
2.3.4
1.2.3
+j
11
h...
13
r+1
r!
1+!+^+....
3
1
2a
series converge.
i+i+^+^+^+..
10
comparison
+ ...C if >
a;
1,
r>
11(x r jr\)
o for
all
0.
<
18
11,
20
443
series
12.42]
<
Verify that
(i)
Deduce that
_\
therefore
>
2.
=2
r=2 (rl)r
1
1111,
<
Prove
*(ii)
v
(r l)r
<
r=lf
0,
(r+1) 2
'
<
r+1
<
2.
infinity a
when
"J
<
2.)
1,
r*
?*/
when
r2
sn
r(w\ ^
c.
S^
<1 + 25 +

11
^
(This shows that l/(n+ 1) < ss n < 1/n. Taking n = 100, we see that the first
hundred terms give s correct to only two decimal places. It can be proved that
8 = ^n 2 see 14.34, ex. (ii).)
*21 If Swr c, prove u n+1 + u n+2 +
+ u n+v > when n + 00, where p is any
:
. . .
The
an auxiliary
series
series itself.
1iUr is convergent if
either
(a)
or
(b)
for all
= <
n ^ m, where k is
constant,
1.
n*ao
either
(a')
or
(&')
un+1 jun ^
for all
n ^ m,
= 1>1.
n>ao
Proofs
(a)
By
hypothesis,
um+l ^ kum>
um+z < kum+1
and so
on.
s$
k 2um
line;
g.p.
GPM II
to
series
444
< k <
by
[12.42
hypothesis. Therefore
terms u 1 + u2 +
+um
...
do not
Swr
con
affect its
behaviour.
(b)
Given
>
0,
there
is
< l+e
le <
Since
<
= I+e
from
(a')
1,
we have
m such that
a number
+e <
for all
n ^ m.
Taking
(a).
By
hypothesis,
um+l ^ um>
e^ c
Hence for n ^
we have un ^ um so that i*n I* when n > oo,
proving nonconvergence of
The series.must actually be properly
divergent since it consists of positive terms (12.32 (5)).
(&')
Since
>
1,
we have Z e >
^1>1
and the
result follows
by
for all
(6)
The
then shows
n > m,
(a').
Remarks
(a)
if
is essential
the result
may be false
we merely have
^1<1
for all
n > m.
/i
n+l/ n
un
"
n+1
<1
foralln.
1)
and
k<l
as
we
please
by taking n
sufficiently large.
(/#)
The
tests give
1.
For example,
W>co
d and 2
S(l/r)
(l/r 2 ) c,
'
we must revert
series
12.42]
when the
form
445
is
inconclusive.
1+1 + i+ i+1 + 1+
2 42 2 3 4 4 2B 4
*"''
Jl
23'
so that
un+1 jun
_L
25
JL
93
27
'
29
oscillates infinitely.
it is
the
I
1 1 1
2"^42"^2^^"4*"^"""^
95
'
The
97
*"'
'
2 2n_1
i 2
q.p.'s):
aroyian
42
22
>f+T when
^+1 =
andj
Hence the
that for
n > co;
^ + 22^1^1^+
when
~>0
Observe also
and has sum to infinity
un does not steadily decrease to zero when n > 00
series converges,
tljis
series
Example
Examine the series 1 + 2x + Zx 2 + 4a;3 + for all positive x.
Ifx = 0, the series is 1 + + + ., which converges to the sum
~
If x 4= 0, then since u n = nx n x
. . .
. .
1.
w+i
un
The ratio
test
shows that
> 00.
if
< x <
the series
c,
and
if
>
the series d.
f The reader may feel that an 'irregular' series like this is not a fair example;
but in fact the terms are constructed according to a quite definite law: 'the rth term
is the rth power of 2 or 4, according as r is odd or even '. In this example it is easy to
write down a formula for ur viz. ur {3 + ( l) r} _r
% See the footnote on p. 451.
,
312
series
446
[12.43
. . .
Exercise 12(/)
Test the following series.
1
22
1+
Tl
3
If a
32
2i
and
42
3i
23
22
+ '
*+*+*+
if
<
and diverges
if
6.
1+
r.
a;
+ ii + Si + ""
a;
i2
x*
9
10
+ 
3!
a+l
a;
+
a:
+2
x+ %
x
*(ii)
+ ...
(aj+l).
'
By showing that
1.3.5... (2nl)
2.4.6... (2n)
374
....
1.3.5.7
1.3
1.3.5
(i)ix + x2 + x3 +
2.4.6.8
2.4
2.4.6
11
2!
series in
(i)
diverges
>
1.2.4... (2n2)
2.4.6... (2n)
when x =
_
~
2n'
1.
1x
1x
first
l+x + x 2 + x3 + ...
'
if sn
(\x\
<
gives a
1)
SERIES
12.44]
is
rapidly convergent
when
\x\ is
small, since
\xn
and x

447
xn
is
when x
is
just
less
than
1,(1 /r
2
)
1.
On
(ii)
shows that
converges slowly.
These notions are necessarily vague since the terms a good approximation', 'fairly small n' have not been specified; they are relative
and not absolute notions, and can become precise only after we have
selected some definite series by which to fix our standard of rapidity '.
However, they are useful in a descriptive way.
The essence of the comparison test for convergence (12.41 (1)) is
that if 1ur is term by term less than a convergent series cLvr then
1ur will converge. We may say that 1ur converges at least as rapidly
as Hvr
and if lim (ujvn ) = 0, that '1un converges faster than 1vn \
'
'
'
If Hur
is ineffective;
p> 1),
and
is
1.
no
finality
it
ineffective.
The
in
'
series
448
[12.44
(1)
If f{x)
oo
to zero
for x
1,
then
/*oo
2 /(**) and
r=l
Jl
Ver9*>
co
/co
/co
S/(r)<
Jl
f(t)dt+f(l).
Jl
r=l
for all
Proof. Since /(a;) steadily decreases to zero, we have f(x) >
n
fx
x ^ 1 so that fix)> f(t)dt are increasing functions of n,x respectively.
If r
/(r+1) </(*)</(#).
Integrating this form r to r +
1,
V+1
/(r+l)<
/(a
)&
</(r).
(i)
Summing
for r
1,2, ...,nl,
im<
r=2
n
\
Jl
f{t)dt
< S/(r).
(ii)
r=l
/oo
of n,
Suppose!
(a)
it
f(t) coexists.
Since
jj(t)dt<jf(t)dt.
From
the
left
of
(ii),
f(r
r2
< f7(0
Jl
series
12.44]
449
so that
r2
is
bounded. Hence
and
S/(r) ^ f"/W*+/(l).
Jl
r=l
It
of
(ii)
by letting n
> co that
r=l
(6)
V/(r)<
r=l
rl
r=l
Jl
so that
f(t) dt is
Z,
when
,> oo.
Sincef
/()
<ft
is
When X>co,
also
and so
oo,
/()
exists, it
X < w,
<
rn
f(t)dt<jJ(t)dt<jj(t)dL
(c)
rx
rni
<ft
>
Z,
/(f)
i.e.
<Zf
exists.
integral
other.
(2)
The function
$(n)= if(r)r=l
steadily decreases as
Proof.
By
f(t)dt
when
all n.
f(t)dt<f(l);
Jl
rx
between
rn
S/(r)r=l
is
Jl
(ii),
0<
f This step
{ f(t)dt
itself
j 1 f{t)dt>l when
cf.
f
I
n+ oo
{n being
Remark (fi)
in 2.71.
Thus
J
j cos nt dt
f
/
J 1
cos
ittdt
is 0;
but
cos ntdt
series
450
[12.44
rn+1
also
0(n+l)0(n) = /(n+l)
f(t)dt
<
Jn
by using
The
(i).
corollary
is
particularly interesting
example
integral diverge, as in
(i)
when both
series
and
when n
 oo.
following.
Examples
Taking/(a:)
(i)
= l + + +...+logn.
${n)
to
some
With/(a;)
then
limit
1 /r p )
if
if
p=
1,
test.
1,
=j=
f*/(f)*
while
= _I(l_XM*)
/(*)
<ft
logX.
p1
00
In fact y\
J >
1,
which
is
(cf.
12.41(3)).
p1
rP
r =l
>
2.
Exercise 12(g)
Prove
2^ diverges.
2 Discuss 2
rlog(2r)
If a
>
0,
a2
lies
between
4 If a
>
sum to
lie.
(a+l) 2
+ (a
+ 2) 2+
""
11
+
a2
must
'
infinity of
1/a
0,
r(logr) J
a2 +
l2
a 2 + 2 2+
'"'
to infinity of
series
12.5]
logxdx
<
Jl
451
n
^,logr
r=2
log x dx + log n.
<
Jl
nyoo
Series of positive
12.5
The simplest kind of series with terms of mixed sign is that in which
(i)
+ say
ui ~ u2 + % u* +
where each u
is positive,
un+1 < ur
(ii)
un steadily decreases to
i.e.
Km un =
(iii)
zero,
n>oo
Prf
By
(ii),
s 2n
Ku
each bracket
+ {u3  1*4 +
)
2)
is positive.
. . .
+ {u2n _ x  u2n ).
Hence
s 2n is
a steadily increasing
= Ul ~ U 2 ~ Ua) ~ (^4 M s)
(
each bracket
is
positive
by
(ii),
and
by
(iii),
I,
by (i). Hence
when n > oo.
s 2n
< uv
s 2n + u2n+l>
therefore
n>oo
lim s 2n+1
n>oo
Hence, whether n
is
lim s2n
n><x>
I.
oo, i.e.
the series
converges.
f The fact that 2n > I does not by itself show that the series converges. Thus for
the series 1 1 + 1 1 + ..., 2n =
and so lim s2n = 0; but the series does not converge since sa +1 = 1 it oscillates finitely.
:
series
452
Example
1 +
i+
[12.52
the theorem.
lies
%i = (l +  + ...+
(
(h),
no.
><*
S;)
8.
infinity
may not
con
+ *+i+i+,
(i), (iii)
are satisfied
by
but not
it,
(ii).
The
+ i+i+i + ;
22
32
42
ii
2a
'
i
32
4a
"
series
Examples
(i)
(ii)
1U )
^ + f 4+
1 i+
1
,111
 +
+
1
2
32
4l
is
not
A.c.
because
is a.c.
" lsA
because
C! 
because
+ + % + % + ... diverges.
1 ++
... converges.
1
,111
+_ + _ + _ +
meaning of
\x\.
...
converges.
SERIES
12.52]
453
series
we can
an absolutely convergent
convergent in the ordinary sense. More explicitly
Ei*r ,
7/2
series is also
this,
we
illustrate
%+
i*
+ i*4 +
+0 +W3 + O
series:
+u9 +
+0 +0 +
+1*5 + 1*6 +
+ i*u + i*ia +
+i*10 +
+0 + ....
+ 1*8 +
series
and
%+
which
with c
1*
series of positive
series
is
series converges.
Hence by
(1))
+ w2 + w3 +...,
where we define
and
(Thus relative to the given series Si*r Ei>r is the nonnegative part, and
Swr is the nonpositive part with its sign changed.)
\ur \. Comparing each of Svr 2t*?r (which
Clearly vr ^ i*r and wr
,
converges;
i.e.,
since ur =
vr wr for all r,
Hence
Si*r converges.
series of
series
454
[12.52
it is a.c.
(o.c.)
or semiconvergent.
Examples
*(iv)
The
'
equal to the
sum of
Slid:
vr tvr , \ur
vT + wr
Also, since
00
00
00
r=l
r=l
r=l
00
<
00
2 *V + r=l
Sw
r=l
K.
r=l
example can
this
\( f(x)dx\
a
\J
>
a and f(x)
if(x)
is
also
where b
<
continuous;
if
/(a)>0,
if
f(x)
=sj
( \f(x)\dx,
Ja
Ex.
cf.
7 (a), no. 5.
Define
f(x)
0,
if
f(x)<0,
if
f(x)
>
0.
and
f(x)
=A(x)Mx).
Also
f(x)dx\
I
Ja
f^dx
=
\J
rb
^1
fz (x)dx
Ja
rb
/i(aj)cfa;+
Ja
=
(v)
v
( {fi(x)+A(x)}dx
Ja
( \f(x)\dx.
Ja
1+ * +
/or aZZ values of x.
If a; = 0, the series
t See 1.14,
fi{x)dx,
Ja
I
is 1
(i).
+ + +
. .
x2
2!
x3,
3!
+
By
4.15(9).
series
12.52]
If x
>
0, all
455
n~X
U n+l _
I X
_X
n
w!/(w
un
1)!
Xn
^n+i
thus since
<
lim
or alternatively, since
n>
for ajj
2Xf
0,
3!
2!
positive y,
therefore converges for all negative x.
also c.
2.74)
(cf.
xn
it is
all x.
N.B.From
lim
and hence
for all
x independent of n.
noo n\
(vi)
x +
for
4 +
...
all values of x.
w2
First consider the series
2/
+ ~2
y*
"*"~3"*"~4
is
+ *""
+ + +
n+x ly n
u
n+i
^1
= y
/ = n
un
w+l/n n+1 y^V
so
is
<
if
<
The
A.c. for
. . .
when
0.
n>oo;
1 <x<
1,
and thus
<
<
1;
series in
. . .
since all the terms of this binomial expansion are positive. Therefore
2V
'
and sot+
x^
when
we
 oo.
case.
\x\ +
\x\ 2
f In fact
are considering
\x\ 3
. .
..
SERIES
456
[12.53
The modified
12.53
>
\.
The
ratio test
it
// either
or
then
Hur
(a)
(b)
u,n+l
^ k <
for all
=I<
lim
n ^ m, where k is
all positive.
constant,
1,
converges.
\ur c,
\
series.
Hence
lLur is a.c.,
and
therefore c.
If EITHER
(a')
or
(6')
then
(i.e.
n ^ m,
for all
'n+l
lim
Ymt is divergent
=I>
1,
not convergent).
first
sight all
this does
follows from
(a')
(&')
\
for all
is
Examples
(i)
ex. (vi)
in this way.
0.
a;
n+l
Since
when
00.
Hence if \x\ < 1 the series c; if \x\ > 1 the series n. Only the cases
remain to be investigated, and these are treated separately as before.
(ii)
l+mx\
m(m
1)
m(ml)(ra2)
#a + 
2!
If x
0,
the series
is 1
+ + +
. .
3!
x3 +
which converges to
\x\
SERIES
12.54]
If x
4= 0,
then from
457
m{m 1) ... (m n + 1)
n!
we
>
find
n+1
Hence */ 
The cases
1 </ie series o; */
\x\
1,
>
a;
when
\x\
oo.
d.
1 the series
it
if
if
1,
However,
and
<
re
m> 1
the series o if m Js
the series c
12.54 Regrouping
if
and
if ra
and
if
< 1;
m < 0.
infinite series
When dealing with finite sums it is permissible to group the terms in any
manner by means of brackets, or to rearrange the order of the terms, without
thereby altering the value of the sum. These properties
may
not extend to
infinite series, f
Example
(i)
The
series
+ + + ...,
i.e.
we
we
and
between
s B oscillates
1.
If we insert brackets as
obtain
+ + +
. .
which converges to 1.
This example shows that, in general, brackets cannot be removed or inserted
in an infinite series without altering its behaviour or its sum. However, the
following result is easily proved by using the definition of 'sum to infinity'.
Theorem I. // 2r converges to s, then it will still converge to s when brackets
are inserted in any way.
Proof. Let n x < n a < w8 < ... be any infinite sequence of positive integers,
and write
n%
fix
XM
r1
2 u
r=rii+l
r>
2U
^8 <
e.
for all
n ^ N,
(i)
r=l
f
They
but this
is
will hold if
we bracket
relatively trivial.
SERIES
458
Let n v be the
N. Then
to
first
(i)
[12.54
holds for
all
2s
is
n v and so
,
Tim
2u
s <
(ii)
r=l
for all
n +2
2u
r=l
(ii)
Tim
result
Hence
n P ).
2
i=l
becomes
<
for all
m ^ p;
i=l
i.e.
Y&i converges to
s.
The converse
This
illustrated
is
Example
If s n denotes the
(ii)
sum
W+W+i,
and
tn
denotes the
sum of n terms of
1 _l_l_l_i_l_lLl
1
8 '5
2 4=13
6
10
A_
1__L
12 '7
(formed from the first by taking one positive term followed by two negative ones)
sum to infinity of the second is half that of the first.
Consider
then the
= ** + **+...+
4n 2
4n
Wl =
hn+ *
therefore t n > %s
Example
series
(ii)
when n
*3n
s;
hence when n
4n
+ 2^T[*2 S
= hn +
2nTl ~ 4^T2
illustrates that, if
a convergent
It
new
series
series
12.54J
Theorem
II.
manner without
459
any
sum to infinity.
Proof. Let
ur ^
hence sm
<
function of n (because
s'
Thus
s.
and
all u' 3s
T
2u
r=l
'
Zwr Hence
.
for
a given n, there
is
0),
in
s'
n
Since
ur is a steadily increasing function of w
r=l
is also a steadily increasing
so ^ < s. Since
0), therefore by 2.77 s'n tends to a limit *' where
\u'
r
is
sum
to infinity.
Proof. If Hu'r
is
We use the notation vr w r oi 12.52, and its extension^, w'r for the rearranged
,
Hu'r Then clearly Hv'r , Ett/ are respectively the rearrangements of Svr ,
TiWr entailed by the rearrangement lLur of Zwr Since T,vr , Ewr are convergent
series
Theorem II
gives
00
00
00
00
2V
S < = r=l
r=l
Sw
r1
and
2 wf
r=l
'r
00
CO
00
00
00
00
CO
r=l
r=l
r=l
r=l
r=l
r=l
r=l
We shall not discuss the rearrangement of infinite series any further; but
enough has been said to caution the reader against treating infinite series
analogously to
finite ones.
Exercise 12(h)
Determine the behaviour of the following
1111
series.
i
_ i8 + i_
6
7 +
2
Z
1
1
1
1
1V2 + W3V4+
logi+log+logf+logf+log^+logi+.
+
1
2.3
1.2
3.4
4.5
+h
f+if+&H+....
xz
x6
*3 + 57 +
a;
2*
x2
10
'
<s <ux
all values of x.
x3
x7
prove ux u t
x6
x1
x*
+ 32 + 4^ + ""
f See the General Proof in 12.52.
32
GPM
II
series
460
[12.6
*12
pW = gWi
L w
*14 If ur
(n+l)aj
n!
S
1.3.5... (2n+l)
*13
L^
[The cases
\x\
may be omitted.]
nm
w<
then
<
*15 If Ewr
is
the converse of
<
i=l
~
r=l
I)
m+1
i=l
11.
r=i rj
for all r,
if
>wherea
Ifs
2 V =
Sw
r=l
Tiur ,t n
r=l
\ur \,
+oo;
r=l
~1
i(n*n)
r=l
\ + % \ + \ ...
is
rearranged as
(taking two positive terms followed by one negative term), prove that the
of the new series is f times that of the original. [Prove t3n = s in + s2 nO
sum
Maclaurin's series
12.6
We
tinuity
and
f(x)
where
power
series
derivability for
<
conditions of con
then
x,
. . .
+ ^_^/(i)(0) + Bn (x),
(see 6.54(1))
r
fx _
(*)
(Lagrange's remainder),
f(n)(0x)
x^ld)
(n1)
71
1
n
f( \dx)
(in
(Cauchy's remainder)
general not the same function in
< 6<
1.
by an
infinite series.
n,
however
large,
(n
\t) is
series
12.62]
O^t^x;
continuous for
then the
/(0)
We have
sn (x)
461
all
conditions
for
of
validity
=/(0) + */'(0) +
(i)
+ ^^/<i>(0),
...
= f(x)Rn (x)
by Maclaurin's theorem. If Rn {x) > when n+ao (perhaps only
some range of values of a;, say \x\ < c), then
when
sn (x) +f{x)
hence the
\x\
<
c.
infinite series
(i)
We write
/(z)=/(0) + */'(0) +
n > oo,
for
<
\x\
c;
(\x\
\x\
<
for
in
when
<c),
(ii)
x over the
c.
in detail in 12.7.
only
of
are
known
is
direct discussion
Bn (x) possible.
some elementary
12.62 Expansion of
(1)
The exponential
ex
series.
functions
ex for
Rn (x) >
e*=
(2)
If f(x)
mainder
therefore
(r)
sin a;,
(allx).
series.
sin.
(x
+ Inn), and
Lagrange's re
xn
is
LL
>
when
n > oo,
for all x.
322
series
462
Also fr\0)
[12.62
and /(^(O) = (  1 y 1
sin* =
^^T+^t
+ (
1 )r
for each
r,
~l
(
so
2 rl)!
(a//
'
Similarly
cos*=
T^e logarithmic
(3)
If
so
^Ji + ^T
/(a;)
in)
{0)
+( 1 ) r "
(2r _2)!
+ "'
(aW ^*
series.
log (1
vn
When x <
remainder
0,
{l
+ 0x) n
when
shows that
n>oo.
1 /
0.
1,
so that Lagrange's
\n
~n\l+6x)
might tend to
oo when n > oo
the behaviour of
(2.73); insufficient is
known about
as a function of n.
^
so that if
because
1 < x <
(1
\x\
0,
\Rn ( x )
Hence
Rn (x)
case
<x<
<
^
j^j
(see 1.14)
l\x\.
+ *) = *^+^... + (l)'+...
(1<*<1).
f)
log(l
r w / \ i
= < 1 )" 1 rTfe(rr^)
series
12.62]
463
f
If
is
n\x)
= m(m  1)
. . .
(ra
 w + 1) (1 + x) m ~ n
n = m+l, and
terms. If
when n
When
g
<
an
by
1,
1) n
= (m 1)... (mn+
^
x >
(n1)!
2.75. Also (1
0)/(l +#r)
is
+ ftc)"  1 <
(l
("(l
<
\Rn (x)\
m (1
Hence
+ *) m =
n > oo,
1
bl)11
if
if
m>
m<
1,
1.
when
[a;!) 1
n+co.
+ mx+
+
If
when
[(1IjcI)
(l
m(ml).(mr + l)^
(1
(M<J)
which are equal and opposite. Here we intend the positive value, since
sum of the series when x = is clearly + 1.
As in (3), Lagrange's remainder would cover only the range
^ x < 1. We omit consideration of the cases x = 1.
the
1
x.
(5) Gregory's series for tan
Iff(x) = tan 1 x, then
/<")()
(0)
Lagrange's remainder
is
so
(2r)
Rn(*)
so
by
2.73
\RJ&)\
and /^"(O) =
n (
n
l)
 \x\ n
"Ml +
+
 l)'" 1 (2r  2)
!.
when
 oo
if
as
<
1.
SERIES
464
8
Hence
tan
x?
ac
[12,63
8r1
= *__ + _... + (
<xr
(*
l).
12.63
<
\x\
by
%=
and
0,
while
1, all
=  n {n + l)an
coefficients
(yr ) x= o
with even
we find that
= ~ I)"" 1 (2n  2)
GWi
Assuming
that
(n
1, 2,
. .
.).
\x\
1 (see
to have y
tan 1 a; for
is
1.
It can be proved that, within its interval of convergence, every power series
the Maclaurin series of its sumfunction. In othqr words, if Har xr converges
to s(x) (for \x\ < c, say), then a n = d n)(0)/nl (n = 0, 1,2,...). Assuming this,
what we have shown by the above argument is that, if the series (i) converges
to s(x)
when
\x\ sS 1,
and/(c)
s(0)=/(0),
tan 1 #, then
*W(0)
'(0)=/'(0)
= /<'>(0),
(x
0),
C(0)
(ii)
....
1.
For
0,
can be shown that this continuous function G(x) has continuous derivatives of
(w = 1,2,...). Hence s(x) =f(x) + C(x)
all orders for \x\ < 1 and that C< n) (0) =
satisfies equations (ii), although C{x) $ 0. In short, the Maclaurin series of
f(x), even if convergent, may not have f(x) for its sumfunction. Further, it can be
shown by examples that even if f(x) and its derivatives of all orders are conit
SERIES
12.63]
465
tinuous in a certain interval, the Maclaurin series off(x) may not converge at
all for that interval.
Thus, to ensure the existence and validity of the Maclaurin expansion of
when n > oo, for
f(x) in 12.61, it is essential to prove somehow that
n (x) >
all x in the interval concerned. However, despite the cautionary nature of
these remarks, it is true that for 'ordinary' functions the formal expansions
obtained as above are valid whenever they converge; and with this assurance
the reader may accept them with confidence.
Exercise 12(i)
State the
sum to infinity of
ch x and sh x.
series for
Obtain Maclaurin series, stating the range of validity, for the following functions.
e^sina;.
[See 6.61,
dn
7 Prove
5 e~ a coax.
(vi).]
{e x coa *
e x 008 * cos
6 cosa;cha;.
(isma+ na),
dec
for all x,
*8
coa(x + h)
x3
x2
cosa;
prove that
in Taylor's theorem,
>
for all
10 If y
(lx*)yixy x y
and
ch(sin _1 a;)
If
a;
and
is
if
12 If 2/
an
even
(k
(1
aj 2 )ya xyi + n 2y =
0.
0,1,2,... ).
integer,
n is
0.
(n a + l)y n
+ a l x + a i x% \
+ x), prove
a
a
(l+a;)
n+8 + (2n+l)(l+a;)2/ n+1 + (n +l)2/ n =
cos log ( 1
2/
0.
OO
If y can be expanded as
o>r
^> prove
0,
series
466
12.7
[12.7
(1)
I ,
m(m
1)'^(m r+l)
...
The expression
is
The expansion of
(4)
(a + x) m
can be
as follows:
m
(x\
+ I =etc;
1
{a) if
(6)
,,
often abbreviatedf to
m
(a\
1+I = etc.
if
Examples
(i)
(1
Write down the general term and the first four terms of the expansion of
will
<
\2x\
1, i.e. if \x\
<
J,
if\
 _ i)r (r +
(
+ 2)
2rxr
1.2
=  l) r 2 y  1 (r + 1) (r + 2) xr
(
.Find
<7&e 9ft
(iii)
(putting r
/
1\
if
3 successively in
<
\.
= i(ii)"'.
i
\4seJ
\4xJ
8!
2.3...9/1X 10
8!
\x)
\4a:/
9
(4)
10
Calculate ^102 correct to four places of decimals by using the binomial series.
V102
and
is
(2)(3)...(9)
1
16a;
in this
w
0, 1, 2,
(\x\
i^) ! = (ii)"
(1
V{100(1
+ 002)} =
10(1
+ 002)*,
~*' ~
(002) 3
+ ....
m is a positive integer.
series
12.71]
467
the next term clearlyf being too small to influence the sixth place.
V102 =
(iv)
6
as far as the term in x*.
1 + x + x2 + 3
+ x + x 2 + x3 = 1  a^)/( 1  x), the given expression is
(la;) 5 (la:4 )~ 6 = (l5a;+10a; 2 10a;3 +...)(l + 5a;4 +...)
= l5*+10a; 2 10a;3 + ...
Expand
Since
v)
jc )
on neglecting
(
Hence
100995.
all
powers of x, stating
11
(2)(l + 3a0
~ 72C + 7 1 + 3*
= ^(i4^) 1 +f(i + 3)1
The
i.e.
a;
first
<
2;
(l)(2)...(r)
14
r!
=
(2)
IV
(l)(2)...(r)
2)
r\
k  3) = \
*<*>'
Summation of series
 (  sr *}
In the result
which
is
valid
3.4
when
\x\
<
4^(^)
3.4.5
1, let
"3
4
1
3" 5 ( a; 3+ "
72
us put x
\.
The
series
becomes
3
i+t+o<i) 2 +f^f(*> +.
t See 12.81, ex. (iii) for an estimate showing that the error in stopping here does
not affect the sixth place even when we allow for the succeeding infinite series of
terms.
series
468
which can be written
A
3.4
Our
A
K
3.4.5
result
(li)3
[12.71
8.
many
series of this
form
(in
Examples
(vi)
Find
sum to
the
infinity of
1
1.3.5
1.3
+ I78
4.8. 12
+ ""
The factors in the numerators form an a.p. with common difference 2, and
those in the denominators have common difference 4. We begin by dividing
each factor in the numerator of each term by 2, and each factor in the denominator by 4:
i a &
lis.
1_
(l)+
iT2
(l)2
~r2f3 (l)3+
"
1.3
t 1
,
(vu)
H
v
'
3 3.6
1.3.5
1
+ x) m with x =
= ^/f = \ ^6.
and
m = J.
1.3.5.7
1
3.6.9.12
3.6.9
12.
h...
lii
f(f) + ^(f)
+ ff(f) 8 +
JL
2.
& Z
^(l)* +
= {l+^() + Z^(f) 2 +
"i "f
1
...
= (!!)*! = V3L
Exercise 12(J)
down and simplify the general term and the first four terms of the expansions
of the following functions in ascending powers of x, stating when these expansions are valid.
Write
(l+x)*.
X
(1x)*
2 (lx)i.
X
2+x
+ 3*)*.
1+x2
(1
(1a:) 8
(4
jc )*.
SERIES
Find the named terms in
8
469
9 10th term in
(1
*11
the
is
(r
+ x) m
\x\
2a;)* if
<
1,
a;
< f?
show that
Deduce that when r becomes greater than m+ 1, then ur+1 /ur < according
as a; > 0; and hence that, after a certain stage in the expansion, the terms are
alternately positive and negative if x > 0, but all have the same sign ifx<0.
Find the numerically greatest term
12
+x)
(1
13 (1x) 10 whena;
Use
the
21 ' 3
binomial series
16
Expand
+ x + 2a;
V 98
(iii).]
when x =
15 (408)*.
17
in 12.13, ex.
14 (1 +x)~*
.
\.
decimals
2 )*
State the condition that the following can be expanded in ascending powers of x,
coefficient of
xr in
the expansion.
2 + 3a;
18
(la;)(l
a;
19
+ 2a;)
1
1
if r
sum
1
24
28
20
<I.
a;3 J
in the expansion of
[Writer
1.3.5
4.7
20T30
4.7.10
+ 20.30.40 +
when
\x\
^.]
calculate
,2
25
1.4
1.4.7
4 + 4T84XT2 + 
3.4.5
3.4
JJ_
+ ""*
+
+
2T4 2.4.6 2.4.6.8
< 1,
00
+ 3r3)a;r1 when
r=l
[r2
'
\_~
to infinity of
1.3
Verify that
Hence
+x + x
<m.
and of x2r+1
26
2r
/j=.
the
T _ lx l
{lx)*(2z)'
Find
6a;a; 2
'<1.
SERIES
470
*29
[12.72
By
shown
here,
12
13
1
1
(rwr/) (r)
4 6 4
5 10 10
of
e.
Taking x
111
e=1 +
From
H + 2! + 3! +  + H + ""
any
specified degree of
6.53).
1!
2!
n\
The
(n+l)\
+ (n
,
+ i)\{
(n+l)\{
{n
+ (n
+ 2)\
=
<
is
1+
1
+ 3)\
,
n+2
n~+l
(n + 2) (n + 3)
(n+1)! ll/(+l)
n+1 _
(n+l)\
+ ")
+ (n+l) 2+ ")
+ ...
by summing the
g.p.,
n\n'
11
'
1!
Multiplying by
q\,
q\
1!
we have
'
I<p.(q\)\<I +
q\
q\q
series
12.72]
where
is
an
and
integer,
so
q\
q\
p.{q
is
471
Our conclusion
1)!.
is
integers p, q;
i.e.
e is irrational.
ax chx, shx.
(2)
If a
and m =
>
log a, then a
= em and
mx
2
ax
o*
i.e.
From
e*
= l+mx +
3 3
the
chx =
definitions
ch._ 1+
%(ex
+ ... +
for all
H^r+...
(all x).
. . .
+ e~x ), shx =
^_^ +
a;,
%(ex
e~x
and
...,
(aU*).
8h ' = ,+
+ 5i +
3!
'"W")
The
write
of the given
first
step
is
always to
series,
it
into
a suitable form.
Examples
2
2!
3!
4!
I
2
3
+++
Sum to infinity
.
(l)
Here
The numerator
and so
r2
(r+ \)r r
u.r
(r+1)!
(rl)l
r\
(r+l)r (r+
1
(
oo
00
=
,5i^
if
1)
(r+1)!
2.
00
00
+ 1,
>
1.
,5iF^"r5iri\5i(^+i)i
= e 1.
Sometimes the decomposition does not hold for a few terms at the beginning.
series
472
Calculate
(ii)
Since
(r
+ 2) 2 =
r2
(rl)l
rl
+J}2
r=l
+ 2) 2
+ 4r + 4 = r(r l) + 5r + 4,
(r2)l
oo
(r
[12.73
1!
if
GO
r=2
= 9 + e + 5(el) + 4^el^
= 10e4.
Exercise 12(k)
Give the coefficient of xr in the expansion of
1
6 **+ 2 .
5 Write
.Find the
down the
6 3
34
3!
4!
H
,
2!
2.4
2.4.6
infinity is
1/e; (ii) e
(i)
2
;
(iii)
(e
+ 1/e) 2
42
46
44
1
3!
5!
f
....
7!
1+2
1+2+3
1 + 
+
+
....
1!
4 !!!f!.
2x
ex
3
1
whose sum to
....
1
2!
12
33
g, + te + 1
ea
series
sum to infinity
32
i^.
3!
2!
1!
=
....
3!
~'
r =i(r+l)!
r=l(r + 2)r!
log(l+*)
(a)
Replacing x by
* + +...
(6)
1 < x <
tJC^
(1
<x^
1).
x,
log(lz)
provided
{](^
1, i.e.
x2
provided
x3
1 ^x<
1.
valid if
1 <x<
1,
series
12.73]
1
. _
Xs X5
+X =
x+ + ~5 + 
^ l8 T^x
i.e.
In
(c)
The
(6)
write
(d)
In
(6)
put
1+x
=
1x
1+x p+l
l#
p
p + 1  log p =
side
<x<
the condition
The
series (b),
0,
{ +
and
1'
2^ +
'
+,
defined only
is
(c), (<?)
i.e.
'
)'
1
= y .
y+1
i.e.
y,
y
becomes y >
Then
log
1<x<1
(~
1 < x <
condition
473
under which
if
(6)
p>
0,
(2i>+l) 5
which corresponds to
certainly holds.
are
either the
(a)
separately;
(c) is
Example
(i)
Find
expansion of log ( 1
+ a;2
a;
if *
<
1.
l+x
= log= log(l+*3 )log(l+ar).
1+x
z
log(la; + a: a )
We
have
log(l+ar)
a?
 + ... + ( l) r_1
2
and
log(l+a:)
If r
is
If r
is
...
+ ( I)'"1
is
...
!
....
l) r_1 /r.
Examples
(11)
K
1.2
'
2.2*
3.2s
( .
. .
12
= log(li)
If the series
general term.
is
...
=log2.
series
474
(iii)
Sum to
[12.73
111
infinity
275
+ ""
377
r(2r+l)
1
2r+l
on using
2r
2r+l'
partial fractions,
^{+*>+ + (s5Tl)}
> 2{1
the
sum to
is
therefore 2 log 4.
infinity of
x
it
converges.
.
x3
x*
+ 273 + 374 +
l72
whenever
> oo,
= 1 + ;+....
log2
since
(iv)
log 2} when n
xr
xr
r{r+l)
) .
r+lj
\r
oo/i
oo
xr
r =l
xr+1
x r=1
r+l
provided that each series converges (12.32(4)), and when 1 ^ x < 1 this is
the case because the first is log ( 1 x) and the second is ( I fx) {log ( 1 x) + x).
Hence
oo
V u. =
r=l
When x =
=
r(r+l)
1
when
s$
<
r+1
sn
"
w+1
> 1
when n
Sum
> oo,
is 1.
(v)
1.
1,
u.r
/l
l)log(la;)+l
to infinity
111
2.3.4
4.5.6
6.7.8
+ ""
ABC
2r(2r+l)(2r + 2)
2r
2r+l
2r + 2'
1 <x^
1,
so
series
12.73]
2r
2r+l
2r + 2
= l/_
Sn=
Hence
\.
L_V
2r + 2/
/ 1
B = \, G
\,
2\2r+l
2r+lJ
2\2r
475
when
i(log2l) (log2l + )
> oo,
= flog2.
Exercise 12(/)
r
Find <Ae coefficient of x and give the first four terms in the expansions of the
following, together with the conditions of validity.
1
2 log(2 + *).
log(l4a;).
log(l+a;) 2
4 log(l + 5a; +
6a; ).
(1
6 log
a;)log(l +x).
+x+
.
7 log(la; + a; 2 a;8 ).
8
Expand log
9 If
>
a;
+ 2x + 3a; 2
expand log(l+a;)loga;
1,
term.
10 Taking x
of decimals.
11
Taking x
12 Use series
13
(d)
series (6)
and no
10. to
prove log 3
10,
prove log 10
23025851.
1098612.
2 log 10 a, where
=
=
...,
Ill
T72~2T2 2
x2
1.2
i
1.2.3
3T25
"
15
2.3
+ ....
19
3.4
5.6.7
173
_1_
+
+ 373s + _1_
573*
111
x*
X*
>_>
,g
9.10.11
+ *"'
r
Ur
[_
1.2.3
j
4=r3
3.4.5
x*
33
x*
4r2
4rl'
4a;
5.6.7
6a;
'
consider lim *

J
8a;
series
476
24 If y
2x*
1,
i^*2*
25 If
sin 2 a
[12.8
3a
3?/
52/
+ "*'
prove
=
x
26 Iff{x)
2 2 sinH<x
2 3 sinHa
7T +2
<
\x\
1,
prove/
2 n + 1 sin 2n
ia
+....
to
2x \
^^1 =
2/<<r).
we obtain
^=1* + **+...,
a relation discovered by Leibniz. This series converges very slowly (see Ex. 12(m)
no. 4) and is impracticable for calculating n.
better method, given by Machin, depends
4 tan" 1  tan 1 ^9
which
is
write 4a
tan
and a; =
Thus on putting x =
\ and
2 tan a 2 ~
_ 4 tana IV
~ ltan 2a/ _ ~ \ltan 2 aj
J
(
2a
result
Jar,
2 tan 2a
on the
1.
Hence 4a
tandzr + tan 1 *)
is
\ir; if
we
1 +x
= y
in Gregory's series
\,
II
izr
4J
gf
we have
1
 239
...J
3 23g3
can be shown that 6 terms from the first bracket and 2 from the second give
to eight decimal places; while 21 and 3 terms from the respective brackets
give ir to 16 places.
It
7t
12.8
Series
and approximations
Given a convergent
to infinity
=sn
infinite series,
by taking the
first
n terms.
we can approximate
to
its
sum
series
12.81]
by an
477
Examples
(i)
If
Ifwetake
we
= x  + ... + ( l) n
log(l+z)
is (see
12.62 (3))
When
>
0,
Whena;<0,
+ 6x >
\l
and so
i2 + i(^)<
2!
i2 n (a?)
<
(see
is
#
n /tt.!;
if
a;
>
+ ... +
E n (x) =
12.62(1))
but
x n+l
.
0, e 9x
(n+1 )( 1+a;)n+1
= 1 +x+
ex
Ifwetake
the error
<
and
(ii)
R+i(a;)
<
ex
\B n (x)\
X
(n1)!
(x n /nl)e 9x
and
x
<e
n!
When x <
we can assert
all
error is
n\
(n+l)!
e 9x
<
and
that
The
is
0,
e x.
(n + 2)\
+ "'
x*
_ xn / + x +
+ \
~n\ \ w+l (n+l)(n + 2) '"/'
,
this is less
xn /
wl\
if
than
n+l
approximation
is less
than
x2
g.p.
Thus
xn
n\
The argument
(n+l) 2+
_
=
'"J
if
>
xn
nllxl(;n+1)
error in the
n+1
n+ 1x'
is
===
sn (n
> m)
332
series
478
For when
n> m
error after
[12.81
n terms
,
& K+l>
is
+ un+2 + mtc+3 +
Kn+ll
'
is
etC
<
>
(cf.
t
Wl (l+* + *+...)
1
U,n+1
1&
==
steadily
sn (n
> m)
> m)
all
n > m, un >
is
The argument
x
 1 )
{un+1  un+2 + un+z r
after
}.
. .
.
En
converges to a limit
where
fixed,
the numerical
n terms
Example
(iii) Consider ex. (iii) in 12.71: we wish to show that all terms after the
fourth in the expansion of ( 1 + 0*02)* cannot influence the 6th place of decimals.
The general term is
ur+1 =
Hence for
and since
>
i ^
1}
ur+ 9 \ r
02
^: = rTT< =
1 (i.e.
(r;*l).
((H)2r
v\
2r
1

7?T<
01
<
r>1
>
2r 1
,
r+1
(001)
/
(
001
<
002,
479
series
12.82]
the terms steadily decrease numerically. Hence the error is numerically not
greater than
4 A
**.#
3.5
1
(002)4
(002) 4 = 625 x 10 9 .
2*4!
4!
S
Formal approximations
12.82
The
(i)
best
W
i
(a)
is
made.
x(l+ax)
ax
'
+ bx'
'
+ bx+cx2
ax
+ a;) =
1
when x is
small,
+ bx
we have
+ x)
(b)
i
correct to order
+X
\, i.e. b
.
Hence
2x
==
Ji
rrom
4=
x2
is
more convenient.
+ ax) = (l + bx + cx 2 )log(l+x),
2
z
l
2 =
2
x+ax
(1 + bx + cx {x%x +\x lx + ...)
2
= x + (b$)x + {c\b+\)x* + {\c + \b\)x l + ....
x(l
and
= 6,
6
1,
= c\b + \,
= \. Hence
= ic + i6i
right
(ii)
Prove
(
If 2/
(l
+ l/n),
1+ I)"_.(,_ +j iL +
then
log y
= wlog (l+^J
/l
)).
x5 on the
series
480
Writing exp (x)
[12.83
we have
*"{ L + t +0 $))
1
=e i+
[
{i + 3s
+o
+
(^)}
^ + 2^ + (^)]
x
(iii) If p is small, prove that a root of the equation e + x
mately \p  ^p 2
= 1+p
is
approxi
For a
first
+ 2x = 1 +p,
.'.
far as
\p.
x2
+ ip) =p,
x = \p{ 1 + 1^) 1 = ip  &p 2
2a*l
so
Example
.
ex
Find
lim
lx
0 log(l+aO
We have
ex
x*
= 1 + x + + a8
where
11
^ + ^+ +
4!
3!
/
....
5!
Id 2
\x\
<3
if
3!liH
1*1
<
4.
log(l
Similarly
where
\b\
+ a?) = x +bx z
< o+
T + iir+
<i(i +
=
f
a;

Vr
31a;
a;

+...)
if
N<111
linear in *.
series
12.83]
ex
Hence
lx
ix*
481
l(x2
i + ax >
,
when
+ 0.
Mm
x_
ex 1 x
log(l + a;)
lim
x>o
x
i
2i
+
= lim
x+0
1.
(6.9).
Exercise 12(m)
Prove
==
2/(2n+
1) is less
than
6n(n+l)(2ra+l)*
1+
Ii
lies
aj
<
1,
8i
+
and
nf n\
1+x
1x
i log^
is less
2i
^
(H n+l/w!
\
between
3 If
=?
'
xs
Z + + ... +
3
a:
2n+1 /{(2w
a;
"" 1
2nl
;
+ 1)(1 x2 )}.
Putting n
2,
**l*+*... + <D^
is
first
give
decimals.
4 lllx)
7 V(1+a!)
2n + (n +
l )a
^n*l
r
2n + (nl)*" .12*
af
'
SERIES
482
8 e x /*J(l
+ 2x) = 1+x 2 f# 3
9 log
10
11
ex
+ e x )} = \x + \x 2 correct to order x 3
= 1 + x + x 2 + ixz [Take logarithms.]
+ xy+ x
l
x*
r4
12
720.
= 1  +
[Method of
12.82, ex.
(i) (6).]
j
\
/
for 1/{(1
x) 6 (4 + #)*},
x)< /(l+x)},
s
x being small.
14 Find approximately the small positive value of x which satisfies
(1+x) 6
103(1
xf.
= p p 2 is an approximate solution of x e x p.
17 If p is small, prove that a root of logc+a; = 1+p is approximately
1 + \p, and find a closer approximation. [Put x = 1 + h, where h is small.]
16 If p
is
_
18
lirftits.
a;log(l+a;)
^
x
lim
x+0
'
trx
19
lim
a x l
>
(a
0).
a:>0
Prove
r= or!(n r)!
2n
.!
2 Prove
3
(c
. . .
. .
).
(i)
(l+x}
and use
*(ii)
~
~
to prove n Gr+1 =  1 Cr + n 2 Gr + n 3 Gr +
+ rCr for < r < n.
Conversely, assuming this relation, prove the binomial theorem.
it
For the
. . .
series (l)
. .
SERIES
Find
the
infinity of
2.3.4
tan 1 (2r+
+
+
T+2 l + 2 + 3 '"*
If r > 0, prove tan 1 (l/2r8 =
483
1)
9 If n straight lines are drawn in a plane so that no two of them are parallel
and no three concurrent, the number of regions into which the plane is divided
is denoted by f(n). Prove f(n+ 1) f(n) = n+l, and deduce the expression
for/(n).
10 Prove that (l
^
Deduc^thatt
,
when n
is
a positive integer or
+ %)x.
dx
sm$x
zero. What
Jo
sin.(n
when n
the value
is
is
a negative
integer?
11
Prove
(i)
sin no:
= 2cos(n
+ 2cos(n 3) a: + ... +
l)a;
suix
according as n
is
(1
(
[2 cos x
sin nx
dx
=n
or
!q 8W.X
Jo
according as
integer.
r n /sin na;\ 8
Prove
(ii)
Jo \sina;/
sin0
(i),
12 Prove that
2 4 + 3 4 ...
dn
14
dx n
(x* 1 e1 "")
(ii)
r=l
l4
Slog cos
to
to
=  l) n eW/af+ 1
(
1).
U =
+ +
+
nTi nT2  *n>
15 If
Wn=1 _ i +
then
6
6
m S ^tan.
tan ^5
S
^
z
z
z
z
r=l
r=l
+ ... + _i__ii
log 2.
i.e.
when x =
0;
SERIES
484
16 If
17 If s n denotes the
sum
for
w>
+a
then
1,
2x 2
x
1
8a
4a*
+ 1+a;
^ ^
t
1+* r
+ 1+a; 2 + 1+cc* +
1r.
18
Assuming that
+ a; 2 _
1x1
(l+a:)(l
19
+ 3i + 52 +  =
Find the
(i)
+ a + ^Z + ...  ~,
provethat
coefficient of a:4
(U)
,111
+ =
2 +
1
and of x*
3iii
12
in the expansion of
(l+x+x +x 3 + x i
st
(ii)
(a)
a+b+c
4;
(6)a + 6 + c
7?
20 In how many ways can a total of 10 be thrown with three dice of different
numbered 1, 2, 3, 4, 5, 6? [We require the number of
positive integral solutions not greater than 6 of the equation a + b + c 10.
This number is the coefficient of x 10 in the expansion of
colours, the faces being
(x
21 If a; 2 +px + q
+ x 3 + x* + x6 + a;6
has roots a,
log(l+px + qx
22
2
)
{<x
/?,
prove that
+ P)x + \{a?+p
3
]
(for
a range of x to be
tively.
is
infinity of
3.9
3.9.15
23 l + r + rrr + rrr^7+
8 8.16 8.16.24
_
b, c
are the lengths of the chords of the arc, off of the arc,
arc
stated),
25
3!
27
+ 6l +
7!
8
9!
+ '
26
~ + ^ +  +
30
2.3.5
3.4.7
9 15 21

'
16.24
16.24.32
4.5.9
,
915
r2
r(r+l)(r + 3)
^Tl irzi^'"
16
29
1.2.3
,234
+
28
r+l(rl)f
r+ 2
nA
24 Sr(r+ l)3r
31
3% 2
2 2a;
h
1!
4 2*3
h....
2!
3!
SERIES
32
S
33
(l^x^l).
r(r+l)(r + 2)
/v4
/yil2
/y8
JU
JU
/y*3
*/
1+35+51+15 + '
34 If p
>
1,
<">
3!
/yi7
*Mj
/yill
U/
+ 7l + lTl + 
prove
2
485
by
+l
K clog f = !og5.J
Pr the number of arrangements of n unlike things taken r
n
at a time, prove
no. 2.]
!).
486
13
Complex numbers
number system;
<J(
1)
=
2x =
2x + 3 =
2x
=
x =
2
x +l =
x2
(i)
6,
5,
(ii)
0,
(iii)
9,
(iv)
2,
(v)
0.
(vi)
If we
numbers which
If
we
(i)
that
of numbers
that
(i.e.
satisfy these).
and
is
(iv)
would remain
(ii) is
satisfied
by f The others
.
insoluble.
To enable us
was indicated in
be proved
1.11),
(see 1.11) to
'
'
that
is, ^/(
a solution
1)
is
existed.
'
COMPLEX ALGEBRA
13.12]
487
until
1) exist?
*J(
with.
leave ^( l) as a
further the concept of
we
generalise
still
we
by
Euler) for ^(
 1)
in 13.12, 13.13.
+ bi) + (c + di) =
+ c) + (b + d)i
(a
(i)
by
first
+ bi) (c + di) =
=
(ii)
= 1 and
grouping.
a + bi = c + di, then a c =
2
(a
c) = (d b) 2 which would
Also, if
gives
is
(d
b)i.
number
equal to a negative
zero,
i.e.
ac
if
a + bi
= c + di,
how
then
and
expressions involving
=
*
d.
(iii)
would behave
COMPLEX ALGEBRA
488
[13.13
when
was
contrast
desirable.
many advances
in fact lead to
it
may appear,
did
it
In the
by
we
i,
get iia
spirit of 13.12
we
obtain an imaginary
number
= la = a, which
successive multiplications by
real
number
a
number a is
multiplied
multiplying again
by
i,
we
is
convert
ia;
J.
^x
Fig. 125
one multiplication
rightangle.
by
is
as turning
OP
P" at distance
apply when a
a,
is
measured along Oy
(fig.
negative.
not as a number at
the length of
OP
all,
and so
We
OP
2a
as the
sense;
sense; the
in
on.)
negative) #axis
COMPLEX ALGEBRA
13.14]
489
representation of a + bi both have important applications, the significant thing at this stage is that
we
are led
P"
ia
\
\
Fig. 126
Fig. 127
in a plane,
(a, b)
is
from
i.e.
a pair of numbers
we can re
We
numberpairs as follows:
(a) Equal numberpairs.
[a, b]
[c,
d]
if
and only
if
a=
and
d.
COMPLEX ALGEBRA
490
Addition of numberpairs.
(b)
[a, b]
+ [c, d]
is
[a, b]
d]
[c,
Since subtraction
(6) this
gives
(a)
[x
so that
+ c, y + d] =
[a, b],
y+d
and
[a c, b
[c, d] =
r
[c,
xc yd
(a)
if [c, d]
n+
[a,6]
all
^]
tc >
[c,
b;
~&
d].
[a, b].
+ xd]
[a, b],
yc + xd
and
[x,
d]
b.
_bc ad7
V = ~c? d?'
+
ac + bd
+ '
Vac + bd bcad~]
fcflL ?Ta? ,^ T? J.
we can show
that complex
[0j
and so
d]
numbers obey
>
define
[0, 0],
r
By
we can
for x, y:
*
(2)
y]
[xc yd, yc
this gives
Hence
we can
such that
[a,b].
c,
y]
= a c, y = b d,
[a, b]
and so by
[x,
+ [c,d] =
(c)
bd, be + ad].
implies
hence
*/
+ c, b + d],
is
x+c
If
[ac
which by
By
[a
mean
to
is
[x,y]
By
mean
to
Multiplication of numberpairs.
(c)
[a, b]
[13.14
(ac  bd) 2
{a 2
eri
>
^] or
tc >
^] or
w*W
definition of 'multiplication'
and
+ {be + ad) 2 =
+ b 2 ){c 2 + d 2 ).
be
+ ad =
show that
0,
a 2c 2 + b 2d 2 + b 2c2 + a 2d 2
COMPLEX ALGEBRA
13.14]
Hence at
491
least
this implies
See also
Application of definitions
(3)
[a,0],[c,0]gives
[M
[a, 0]
+ [c,0] =
x
(b), (c)
0]
[c,
[a
+ e.0],
[ac, 0].
Apart from the presence of the brackets and the second part
in
each pair, these results are exactly like the laws of addition and
multiplication of the ordinary
of complex numbers
[a, 0]
and
numbers
real
a, c: the
laws of combination
the
same.f
we write
2 for +2,
In definition
agree to write
etc.,
positive signed
we
us
now choose a =
[0,
l]x[0,
[0,1?
i.e.
1]
0, c
algebraically.)
0, b
l,d
1.
Then
[1,0],
[1,0];
Thus
is
can be written
n9
2
[0, l]
1,
numberpairs
and
(6),
[a, 0]
+ [0,6]
[a, 0]
+ [b, 0] [0,
1]
= a + bi
etc.,
is
Let us abbreviate
for which i 2 = 1.
definition
[a, 6]
If
=  1.
equal to
By
r_
by
definition
by our
abbreviations.
replacing
and
all
numbers
in
[a, b]
by a + bi,
(c),
them
now
[a, 0]
and the
34
real
GPM II
COMPLEX ALGEBRA
492
Since
[a, b]
[b,
0][0, 1]
as a + ib instead of a + bi
V( 1)
as
easily verified,
is
when
for
convenient.
+ li, and bi
= 1,
(if it
we can
write
Our conventions
+ bi.
for
1 which do satisfy
so that this pair i has the property which *J( 1) would have
[13.15
i,
by applying the
'less
is
meaningless to write 3 + 2i
< 4 + 5i. In
be applied to
complex numbers.
13.15 Importance of complex algebra
This
by
Hence pairs of
results can be obtained from one calculation, thus making for economy.
(ii) From any general relation between complex numbers there can
be obtained a special case by taking the second parts of all numbers
equations in real algebra,
numbers
[x,0]
and
real
numbers x
(13.14(3)),
we deduce a
relation
in real algebra.
Since complex numbers arose from the attempt to solve all quadratic
0),
we may wonder
COMPLEX ALGEBRA
13.16]
493
Further,
(2)
. .
e.g.
Exercise 13(a)
Simplify and express in the form
1
4
7
9
10
12
X+Yi:
2 (2 + 7*)(3 + 4i).
3 (2 + 6t) + (53t).
+ 7i) + (3 + 4*).
6i)(5
5t)(2
3i).
3*).
6
5 (3 +
(2 +
(4 + 5i) (4 6).
+
+
(3 + 5i)j(2 + 3).
8 (6 + 5*)/(3 + 2*).
2
and hence 1 + i) 1 1 + i).
1 + i)
(l+*)(l + 2i)(l + 3i).
11 (cos0 + *sin0)(cos0 + isin^).
(cos0 + isin0)/(cos0 + *sin0).
Calculate (a + bi)/(c + di) by the process analogous to 'rationalising
(2
13
denominator' with surds:
a + bi
(a
c + di
(c
7, 8,
13. 14
12
+ bi)(c di)
+ di)(c di)
(ac + bd)
c2
+ (be ad) i
+ 2
<2
'
1)
by this method.
342
the
COMPLEX ALGEBRA
494
Z = X+ Yi
14 If
terms of x and
Z=
and
z1
and
z is
+ l)l(z\)
where
= x + yi,
express
X,
in
y.
Complex numbers z u
15
(z
[13.2
given by
z 2 are
= R + i(oL,
= R 2 g,
z2
Ill
=
denned by
for
which
z%
Zi
z is of the
form x + 0*.
X+ Yi.
= a 2 and
2xy
6.
Hence
18 Prove {(
 1 i V3)} 3 =
1.
numbers
19
[a, b]
20
[a, 6]
21
22
(a)(c)
plication).
+ [c, d]) x [e, /] = [a, 6] x [e, /] + [c, rf] x [e, /] (distributive law).
+ [a, 6] + tow terms = [a, 6] x [w, 0], each being equal to \na, nb].
25 By using the multiplication rule, verify that i = [0, 1] satisfies
x*+l = 0.
23
([a, 6]
24
[a, 6]
. . .
(positive, negative or
zero,
(a) {a, b)
(b)
{a, 6}
(c)
{a, 6}
{c,
d}
if
and only
if
{c,
d}
{c,
d}
{ac, bd}.
ad
be;
{a, 6} {c, d} and {a, 6} + {c, d}. Verify that this algebra
the same in structure as) the algebra of rational numbers
could define fractions abstractly in terms of integers in this
isomorphic to
(fractions).
(We
(i.e.
way.)
13.2
13.21
We
saw
represented uniquely
polar coordinates
by the point
then (see
(r, 6),
and so
complex number x + yi
rco&0,
x + yi
P(x, y)
1.62,
r(coa d
and
x + iy) can be
in the a;/plane. If P has
fig.
(or
25)
r8in6,
(i)
+ isind).
(ii)
COMPLEX ALGEBRA
13.21]
495
in direction
'
6\ a
number
'
as
is
tions
since r
is
restricted to
= +J{x* + y*)
(iii)
be positive; and
cos0:sin0:
x:y: + j(x2 + y2 ),
(iv)
which gives a unique value of 6 in the range it < 6 < + tt. Any value
differing from this by any positive or negative integral multiple of
2n would give the same point P and consequently is a possible value
of argz. Thus argz is a manyvalued function, and we define its
principal value to be that in the range ttkO^tt. (The single
equation tan#
it
gives two
Examples
(i)
+ i)(2U) _ 1313t _
*'
~ 4+9 ~
(2 + 3i)(23i)
We require r, such that rcos0 = 1, rsin0 = 1. Hence r = +<J2, and
given by cos0:sin0:l = 1: 1:^/2; the principal value of the argument
therefore \n. The required result is
{cos \n) + i sin In)}.
i sin in modulusargument form.
(ii) Express 1 cos
1 cos0 *sin0 = 2sin 8 0 2isin0cos0
= 2 sin 0 (sin 0 *cos0)
= 2sin0 {cos \d  \n) + * sin (0  \n)}.
This will be the required result if sin \Q > 0, i.e. if Inn < \Q < (2n + 1) n.
If (2n 1) n < 0 < 2nn, then sin \0 <
and we must write
=
1
cos0 isin0
2sin0{ cos(0 \n) isin(0 %n)}
= 2sin0 {cos (0 + \n) + * sin (0 + in)}.
5+i
2 + 3*
_
~
(5
is
is
t Sometimes 6
amp (x+yi).
is
x+yi and
is
COMPLEX ALGEBRA
496
[13.22
In numerical cases like ex. (i) the modulus and argument can usually
be written down easily after drawing a sketch.
We
OP. Also
see 13.32.
and properties
We agreed in
(1)
form x + Oi to x.
number z
of the
argz
is 'real',
If z
is
or
tt
It is
'
'
then
represented
it is
according as z
purely imaginary,
it
(i.e. a;) is
by a point on the
xaxis;
and
positive or negative.
= x+yi,
x is called the real part (or first part) and is sometimes written x = 8%{z)
and y is called the imaginary part (or second part), written y = J(z).
The process of deducing from the relation z x z 2 that xx = x% and
Vi = 2/2 i s called equating real and imaginary parts, respectively.
(2)
r(cosd + ismd),
then from
it
follows
s(cos0Msin^),
r(cos 6 + i sin 6)
s(cos
+ i sin
<p
<j>)
s cos
and
<j>
2
,
r sin 6
so that r
s sin
<j>.
s since
r,
s are positive.
cos
<f),
sin 6
sin
<j>,
<f>
differ
2ir.
If 6,
(f>
zt
z2 that
argz 2
is
\z x
z 2
is
called taking
This equation holds for principal values of 'arg'; but since 'arg'
is
many valued, we may write arg z x arg z 2 to emphasise that any value
of the lefthand side is also a value of the right, and conversely.
COMPLEX ALGEBRA
13.23]
(3)
the
497
number x yi
and
conjugate of z
i is
called the
is
= x + yi,
notation
imaginary.
(iv)
\z\
when
but
z;
and negative',
z is 'real
= argz, except
form x + Oi where x < 0.
(for
of the
i.e.
z2
z are
(v) If z
z,
then
z is 'real
'
if z
taking conjugates.
Finally, the definition of 'conjugate'
(z x z 2 )
^1 ^2
(z x \z 2 )
z x \z 2 ,
zx z2 ,
shows that
Z X Z 2^ Z l Zi)
z i z %\
Hence
\
Zl
r=

Z 2
by
Zl
Z 1 Z \' Z 2 Z 2

=

zx
+ z2
z.
zz in (3) it follows
that
Z l 2, l^l 2.
definition. Similarly, if z 2 #= 0,
Zl
\ Z 2/
+ zz =
z i z 2>
( )
\Z 2/
\z\
(z)
zx
Zl * 1
Z2 Z 2
\Z
Zl
Z z\
2>
IZ
z 2\
13.23
If z
Z 2
is
+ 1,
then z3
(zl)(z 2 + z+l)
hence either z 
or z 2 + z+
by
0, i.e.
0;
13.14
(2).
= f, z+ = iV 3 and 2 = Ki*V 3 )Since cos tt = \ and sin 7r = \ <J3, the first of these roots
modulusargument form cos \n + i sin tt. We denote this by cj.
(z+)
14.13, ex.
(ii).
has
COMPLEX ALGEBRA
498
Also,
cosfTr
=  and
^ow
(jfi
= \ ^3,
sin \n
[13.23
is
is
7r
+ 2i sin 7r
cos
7r
_
2
A
2
z2
relations
Example
Direct expansion shows that
= a3 + 6 s + c8 3abc.
The operation c x
is
also
as + b 3 + cs 3abc
+ c(o).
By comparing this result with
a s + b 3 + c 3 3abc = (a + b + c)(a 2 + b 2 + c 2 bccaab)
(see 10.22 (2),
(a
example; or Ex. 11
2
+b +c
2
+ b + c) (a + bo) + cco 2
(a),
no. 20),
bccaab =
(a
it
(a + ba> 2
follows that
+ bw + cio 2
(a
b(t)
+ co)).
Exercise 13(6)
Express in modulus argument form:
1
2 1*V3
l+iy/3.
(l
+ 3*V3)/(V3 + 2*).
x and
+cos# + *sin0.
10 If a + ib
(rK
+ M/)"
where
4 1i.
a, 6,
ar,
6+iJ3
V
4_2
2+i
(l+i)^.
^.
^3
are 'real',
and
COMPLEX ALGEBRA
13.3]
12 If
(z

 1 )/(z + 1
2,
^+^+^+1 =
ciTcle
13 If w
(x + 3) 2 + y 2
z/(z
+ 3)
1,
1
to
a)
2
2n
to"
to
0)
to
0.
w = u + iv and z = x + iy,
where
and
if z lies
on the
circle
14 Prove that
15 Prove
499
+ =
to
0,
to
w2
to
according as
3 or
to
n is
or
is
not a multiple of
3.
16 Express x3 + y3 as the product of three linear factors, and deduce the factors
of x 2 xy + y 2 [Method of 13.23, ex.]
(6  c) (c  a) (a  6). [Use
17 Prove (a + <ob + w 2c) 3  (a + to 2 6 + wc) 3 =  3i
.
factors of
x 3 y 3 .]
18 Prove that (a 2 + ah + 6 2 )
[Use tf + ab + b 2
(a;
+ a*/ +
/ )
form
tc
a:
13.3
is
represented
Also
\z\
J(x
+ y2
is
represented
(a?

by the
by P, then
a +
)
 y2
j{(x 1 x2 ) 2 + (y 1 y 2 ) 2 }
geometry.
Fig. 128
by OP.
COMPLEX ALGEBRA
500
If
is
cos 6 sin 6
x y
*J(x
it is
2
[13.32
given
+ y2
by
and is the angle of turn from the direction Ox towards the direction
OP, which can be written L {Ox, OP). Similarly, arg (z x z 2 ) is the angle
defined by
cos 6 sin 6
:
x x  x2 yx  y 2
<J{(xx
 x 2 2 + (yx  y2
)
If
2
},
P2 P
X,
written
loci)
and P
is the variable
point
z,
then
(i)
\z
ex,
is
<x.
P P
z x = \z z 2 is the perpen(iii) If
x,
2 are fixed points z x z 2 , the locus \z
dicular bisector of the line X 2 For
=
X
2 , and the result follows by pure
geometry. Alternatively, it can be proved algebraically from
,
PP
PP
PP
If P,
(iv)
radius
w
z,
2z + 3
2Z +
BU. Then
tt>
 3 + 4 =
w(34*)
since
\z\
r.
Therefore
(1)
sum and
Similarly,
a3 +
zx
j"
Since
2r.
difference of z lf z2
z2 ,
projection of
OP3 on Ox
projection of
OPx + projection of PX P3
m=
,
2r
by projecting on the
+ z2
\2z\
P2 OP P3
x3
z3
2z, so
gram
and
locus of Q.
r, find the
Write
i.e.
a
)
simplifying.
xi + xi) + %i + 2/2)
so that
OP2
yaxis
is
we
find y3
= x x + Wi) + x z + Wz),
(
= yx + y2 Hence
P P8
1
2.
COMPLEX ALGEBRA
501
Observe that the construction for P3 is the familiar parallelogram
law for compounding ('adding') two displacements OP v OP2 To
13.33J
the sum of two complex numbers corresponds the sum of the associated
vectors.
+ z 2 ), we first
which represents z2 The above
Since z x z 2
(2) Difference.
image P2 of P2 in 0,
tion performed with
zx
P P
x,
P4
will give
construct the
sumconstruc
corresponding to z x z %
y,
/
L
/
/
.Jp*
/
p*
Fig. 129
As OP
is
equal
13.33
The
displacement
l5
triangle inequalities
'the
OP3 ^OPx + Px P3
OP3 ^OPx + OP2
i.e.f
and hence
this
z,
we deduce
+ z2 ^
\
\z t \
+ \z2
\.
\z x
z 2 >
\\z x
z 1
+ z2 ^
^l
\\z 2 \\,
I^H'
so,
we
P2 P v
From
P2 P
results
can
all
we first observe
fig.
figure.
that
0t(z)
= **S
V(*
P P
lf
+ 2/ 2 = M
)
P3
129.
are collinear.
To
COMPLEX ALGEBRA
502
h+Z
'
is
and positive
real
By
\
f
^z^R
and
since z x z 2
KI
(Nil
+2M
+ W) 2
Therefore
<
ziZ 2 ,
^1 =
since
\z x
Z2 Z 2
z 2 *
of the form x + Oi
i.e.
13.22(3),
= (Zl + Z2)(2l+2 2
= %\Z x
Zg + Zj Z 2
= Iz^ +
2
2
[13.34
+z 2 <
ja^J
+ z 2
z 2
\z t
.
\
z 2 .
,
by
definition.
i.e.
zx
similarly.
modulusargument form.
P P
x,
and quotient of z z2
zx z2
=
=
=
+ i sin
(j>),
rs{cos (0 +
is
so that
<j>)
<j>
Therefore z x z 2
{rs,d +
<f>
represented
(j>)}
<j>)}.
(f>).
=
i 6J'
80
Also
a*5p2 + ^0Pi
Ordinary equality
may
(z x z 2 )
result
\z 1 z 2
arg z x + arg z 2
not hold
when
all
=
=
<p
+ e.
IzJ
z 2 ,
and
z2
cos fn + i sin
7r,
then
z x z2
also
COMPLEX ALGEBRA
13.34]
503
We
arg z 2
arg {z x z 2 )
fn,
=  n.
and multiplies
its
length
productconstruction in
We
by r.
If z 4
(2) Quotient.
OP^A
by
z 1 /z 2 ,
then
zx
(1),
z2
131).
(fig.
_ r cos 6 + i sin
<scos0 + isin0
_ r (cos# + isin#) (cos0 isin^)
s (cos + i sin $) (cos i sin ^)
<j>
r (cos
~s
<f>
cos
=  {cos (6 
<j>)
+ i(sin
2
<p + sin
sin 0)
+ i sin (0 
cos
cos
sin <j>)
(j>)}.
This calculation
verifies
that
COMPLEX ALGEBRA
504
[13.35
arg ^^
z3
arg {z x  z 2 )  arg (z 3  z 2 )
= L{Ox,P% P^L{Ox,P% Pz
^(PAP^),
i.e.
the angle
towards P^P^
Fig. 132
on the Argand
a;
(6)
(k 4=1).
By the Remark in
arg
(a)
(a)
representation
We shall
Px P2
respect to
13.34,
P P
lt
).
Fig. 133
Geometrical proof
z l z 2't
Divide
is
where A, B, represent
#= 1,
1;
an
PB~HB
<
(The figure
1.)
Then
PB~KB'
and by a theorem of pure geometry it follows that PH, PK are the internal and
A
A
external bisectors of APB. Hence HPK is a rightangle. Since H,
are fixed
points, P lies on the circle having HK for diameter.
Algebraical proof.
(xXi)* + {yVi)
i.e.
and
since k
4= 1
this represents
2

P Ps
lf
to
A,
respectively
is
made
for
COMPLEX ALGEBRA
13.35]
It is easy to
505
AH/HB = AK/KB,
follows that
it
OAOH OH OB
OA + OH~ OH + OB'
AHjAK = HB/KB,
20 _ 2 OH
and so
20H~ 20B
Fig. 134
We can now show that the circles in (a), (b) cut orthogonally (see 15.66(1)).
T is a common point, then OT 2 = OA OB. By the converse of the tangentchord theorem, OT touches the circle (a). Hence the radius GT of (a) is perpendicular to OT, and so touches (6) at T. Thus the tangents to (a), (6) at T
If
are perpendicular,
*(ii)
i.e.
Z4
22
Zs
f 8P 2^*1
and are
P4PJP 6 P6
= P 6P&P
Hence by the
'common
test
Conversely,
if
z1
0,
z2
1
By
rx
ri + z 2 rs followed
by
r 2 > ra + z 5 r3
Zl
Z2
22
Z4Z5
i.e.
we
obtain
ZZ
Z6
1
0.
 Z5
1
0.
COMPLEX ALGEBRA
506
Exercise 13(c)
Indicate in a diagram the points representing (zx + z 2 ), z x
2
between
If collinear points
P P
x,
2,
= 2P 2 P3
+ 2z 2
zx
2z 2
zx , z 2 , z3 .
in the ratio k
P if 2 <
6 Prove that
of
z
+ 1  2i\ <
least values of
+ l
if
3?
(i)
\z
4
if
\z\
1; (ii)
z
+ z 2 2 + ziz 2 2 = 2 { z i 2 + z s\ 2} (i) geometrically, (ii) algeuse the theorem of Apollonius on triangle OP3 P4 in fig. 129.]
z!
If
(i)
5.
braically. [For
PP
I.
z3
(i)
+ z2 =
zj.
z i~ z a
(ii)
If arg{(z 1 + z 2 )/(z 1
2 )}
differ
by
\n, prove
or
\it
Z]J
\ir
z 2 .
[Treat geometrically.]
be (1,0); produce PO to
9 Verify the, following construction for *Jz. Let
= 1. Through O draw a line parallel to
so that
to meet the circle
in Q and R; these are the required points.
PAB
AB
OB
11
is
any
chosen so that
12 If
bisects
point,
and on the
P OP =
PxOP 2 =
circle
<}>.
[Use Ex. 15
x (?.
no.
(a),
4z x
+ z 2 + z3 =
0,
P P
2,
are
prove that
1.]
point \{z x + z 2 + z3 + z 4 )
is its
centre.
Show
15
equilateral, with sides of length 1 and centroid at O. If A reprewhat do B, C represent? Show that the sum of these three numbers is
ABO is
17
sents
z,
always zero.
18 If A, B, O, D, E are the vertices of a regular pentagon inscribed in a circle
whose centre is O and radius is r, and if OA makes angle 6 with Ox, write down
the numbers which these vertices represent.
19 If 2/z x
relation
is (z x
[The
COMPLEX ALGEBRA
13.4]
507
On
*21
[By
13.35, ex.
(ii)
bc
drawn
are
so Hoc
ABC
triangles
BOX, CAY,
ABZ
XYZ coincide.
ratios,
b
ab
So;
Sa
'
Sa.]
*22 For any complex number z prove that the triangle with vertices zz x , zz 2
with vertices z lt z 2 , z 3
23 If P1P3 and P2 P4 are equal and perpendicular, the sense of rotation from
P 2 Pi being clockwise, prove z 1 iz 2 z z + izi = Q, and conversely.
X
S to
[z 4 z 2 = IzaZilandargftzsZiV^iSa)} = + \tt* Hence (z 3  Zi)/(z 4  ,) = <]
24 On the sides of a plane convex quadrilateral, squares are drawn externally.
Prove that the centres of these squares are the vertices of a quadrilateral whose
diagonals are equal and perpendicular. [Let the vertices in clockwise order be
PP
The centres of the squares are (z x + 2 + i(z 2  z x etc. Use no. 23.]
considering
the modulus of the lefthand side, prove that all the roots of
By
25
n
z sin na + z n_1 sin (n l)a + ... +zsina = 1
lie outside the circle z =
n
If a root z satisfies
[1 = z sinwa + ...+zsina < z+ ^l"^ ... + z.
n
z < \, then for this z, 1 < ()
+ (i)" 1 + + & = 1  {\) n a contradiction.]
Zi,
z 2 , zs , z4
),
J.
. . .
13.4
2.
This
is
In complex
p(z)
=p
zn
~
+p 1 zn 1 +
. . .
+p n =
(p Q
*0)
(i)
ONE TOOt.
known also as d'Alembert's or Gauss's theorem, has
not yet been proved by any strictly algebraical argument. We shall
has AT LEAST
This result,
later
GPMII
COMPLEX ALGEBRA
508
In complex
[13.42
has exactly
(i)
roots.
Proof.
The equation
n~
x , i.e.
therefore equivalent to
(i) is
Po(*
<x
?) +Pi(z
n~1  a?" 1
+p n  1 (z 
. . .
cc )
x
0;
i.e.
where fx (z)
is
0,
w1
a polynomial of degree
in
whose
z,
first
term
is
711
necessarily distinct
from
ol ),
x
and by
/i(z)
where
a polynomial in
Continuing thus, we find
2 (z) is
p(z)
is
degree
(z ai )(za 2 )
= a2
(not
+ k,
zn 2 .
n 2, beginning with p
in
or
= Po( z ~ ai) (2  a 2
. . .
z,
(z
cc n
has n roots a l5 a 2
z;
...,an
z; it
say. Therefore
 aJ.
(z
(za^J/^z),
...
a polynomial of degree
P(z)
has a root
z of
similar reasoning
(ii)
or equivalently, that
(which
may
or
may
not
all be distinct). Also, (ii) shows that these are the only values of z
which make^>(z) zero; for if /? is such a value, then
i>o(Aai)(Aa 2 )(/?aJ
and
since
cc2
f}
by
4=
where nx + n 2 +
is
. . .
4
0,
p(z)
(ii)
oc 2 ,
...,ct k
. . .
are
(z
a
fc
)"*,
all distinct.
where filt
(iii)
fi 2 ,
p(fir ) 4=
=p
(z
a.
(iii)
This decomposition
an
= p^z  ctj** (z  a 2 )
n k = n and ot lt
coincides with
i.e. /?
(z
 /?2 r
. . .
(z
if also
(iv)
. .
COMPLEX ALGEBRA
13.43]
509
witten
x,
p(z)
. . .
. . .
ft )
>
. . .
fc )
hand
side
is
not.
1 0. 1 3.
is
complex algebra.
Example
Solve the following three simultaneous equations in z x , z a z 3 , where ax , a % , a3 are
,
all different:
+ ^r+^T^rr
c^ + A, a^
+ Af a + A,
1'
2 ' 3 )
When
expanded,
it is
a polynomial in A of degree
3,
A3
when A = Alt A 2 A 8
=
we have
identity,
Putting A
ax in the
(a 2  ax (a3  a x z x =  1) 4 (ax + A x (a x + A a {ax + A 3
which gives z x Similarly the substitutions A = a 2 A = a3 give z a
).
The
),
z3 .
we
ss
zn
~
+p 1 z n 1 + ... +pn
denotes
lim
p(z + h)
_ p{z)
,
352
COMPLEX ALGEBRA
510
[13.45
From
it is
// the
(1)
and
if a
is also
coefficients
+ fii
(where
fi
p p ly
=f=
pn
0) is
root ofp(z)
0,
then
its
conjugate
a fii
root.
Proof.
fii
On expanding
to be a root
is
p(a + fii)
0, i.e.
1
+ ...+p n 1 (cc+j3i)+pn = 0.
and replacing i 2 by 1 whenever
the binomials
it
P + Qi =
where P,
0,
Hence also P Qi =
0* =
0,
Q=
0.
p(li)
0,
= (li)(l + i) = 2i #
The
p(z)
0,
where
p(z) is the
0.
is that a
fii is a root of
polynomial obtained from p(z) by changing
must
still
cients, p(z)
= p(z).
by
'taking conjugates'.
COMPLEX ALGEBRA
13.45]
For if p{z)
a root
oifii.
to the
Remove
same
order.
=a
511
(yff
4= 0),
then
it
also has
nomial fr (z), where a +fli is not a root of fr {z) = 0; hence afii cannot
be a root (otherwise its conjugate a +/3i would be a root). Thus a fti
is
also
an
0.
Example
Prove that the equation
has
ft*
62
6a
x ax
x a2
x an
. .
.,
the
4= 0.
Consider the corresponding equation in complex algebra, the a's, 6's and k
being real. Denote any rootf by a + fii; then a /?* is also a root, and hence
all
r=1
%
Since the
ct
+ Qi.
a+/3i ar)
\ocf}i ar
*
=1
(aar a +^ 2
)
0.
There
that in
is
(1).
a b^jc
is
Proof.
and if x
a root.
= a + b*Jc
yjc
root (where
^c
is
for
surd
The condition
simplified,
is
P=
and Q =
rational
0, for
otherwise
contradiction. Since
p(abjc) =
PQJc = 00 Vc =
a b <Jc
f That
is,
if
=f=
8.
= 0.
0,
the
is
COMPLEX ALGEBRA
512
[13.5
13.5
p
so that
zn
+p 1 zn
a lt a 2
~1
+pn s p
+...
(z
 ct^ (z  a 2
. . .
(z
 aJ,
 (Sa) z" 1 + (S ai a 2
{z n
zn
~ 2 ...
2a =
n~2
. . .
aj.
Sa, a 2
S ai a 2 a 3 = ^,
and
+ ( l) n
ai a 2 ...an
= (l)^;
in general,
has the
maximum
We remarked in
information that
if
'
cl^cl^, ...,an
some nonsymmetrical
given, then
it
relation
between some or
all
0'.
However,
of the roots
is
Example
Solve (in complex algebra) the equation
3x*
a+fi + y + d
(There
is
4.
We have
f,
afiyd
From a/? =
a,/H
0,
is
4.
not needed.)
COMPLEX ALGEBRA
13.53]
513
a/% + <S)+ytf(a+/?) =
+ (a+/?) =
4( r +
so that
f,
f.
<>)
relation can
first
be solved to give
7 + ^=1.
4
+
+^# = 0, viz. , 6; and
w2
Hence
a,
ft
roots of x % x + 1
a;
w,
0, viz.
y, 8 are the
cussed
by
(iii)
p(x)
(iii)
0.
The
known manner.
(i)
reciprocals of,
(ii)
k times,
process
(ii) is
known
by
&',
and
by k\
(iv)
Write y
(i)
nomial equation
If
ljx.
is
a root of p(x)
is satisfied
by y =
The required
Write y
p{yjk)
0.
Hence
kx.
When a
case k
p(x)
satisfies
0.
ka
0,
will
satisfy
P yn +Pikyn
The
Ijy,
poly
therefore
1 + ...+p 1 y+P Q =
(ii)
then since x
0,
1/a.
1
+p 2 k 2yn  2 +...+p n kn =
0.
of those of p(x)
(iii)
Write
p(y + k) = 0.
(iv) Write y
(i.e.
0,
(ii)
0.
below.
Examples
(i)
Transform
1,
by
k, i.e.
put y
3y* y* 
i.e.
The method
is
kx, so that
y/k.
Then
= 0,
+
4% +
=
0.
2k*y
+ ffc
f% +
s
5A: 4
COMPLEX ALGEBRA
514
[13.53
where y
3 (this
is
the smallest
then
42/s + 54y+135
0,
3x.
(ii)
is
(a)
0.
and put y =
2)=x* + 3,
(b)
yfy(3y + 2) = y + 3.
y(9y* + 12y + 4) = y* + 6y + 9,
9ya + 1 It/ 2  2y  9 = 0.
x*:
Squaring,
i.e.
(c)
Remark. Strictly, we should show that every root of equation (c) is the square
of a root of equation (a),f as follows. If y satisfies (c), then (by reversing the
working) it satisfies one of the equations
*Jy.(3y +
2)
= y+S.
Thus
either +*Jy or
event, y
and Ex.
is
10(c).
Exercise 13(d)
The algebra
1
is
complex.
z2
2i
z3
z"
+ 3T +  + ^=
2 Solve
3
z4
4z 2 + Sz 4 =
is
0,
given that
+ i is a root.
= 0,
a root of x9 2a; 2 7x + 2
completely.
One root of
x + x5 $x*10x 3 9x 2
3a; 6
+ 2a;4 + x3 6a; 2 5x 4 =
+x+l =
is *J2
+ J3. Find
0,
and
all
the
roots.
this
apparent paradox.
Verify that (z  a /?i) (z a (ii) = z 2 2az + a 2 + ft 2 , where a, /? are real,
Let g(z), az + b be the quotient and remainder when p(z) is divided by
2az + a 2 + /? 2 Ifp(z) has real coefficients, explain why a, b must be real.
(i)
(ii)
z2
satisfies
may
y 3 y
= 0,
COMPLEX ALGEBRA
13.6]
515
= 0, show by putting z = a
(z
2<xz + a 2 + /? 2 q{z) + az + b
p(z)
that act + 6 =
and aft = 0. If /# 4= 0, deduce that a = and
that z afiiia also a root of p(z) = 0.
If a
(iii)
is
in the identity f
a root ofp(z)
0,
and hence
0,
find Zafiyia+fi+y).
10
root of
a;
of the other
12 x 3
13
5a;
16* + 80 =
if
the
if
the
if
16 Find two substitutions of the form y = xk which remove the third term
from x4 + 4a;3 1 8a; 2 100a; 112 = 0, and use one of them to solve the equation.
17 Find a substitution y = kx which will transform the equation
8a;
term
is
+ 1.
are
1,
ax a 2
,
prove that
n _ 1?
. .
1,7
[y
+2 =
(fi*
1, 1 a
0,
n.
lf ....]
7* *fi
ay pa
7 p
+ y 2 + 2/?7)//?7 = a 2 /^ =  a 3 /q = pa\q + 1.]
13.6
13.61
(1)
We have proved
. . .
t Cf. 10.11,
Remark
(yff).
COMPLEX ALGEBRA
516
and
all
tinct,
constants
are real.
The
case
a, b, c
and may be
factors
I
[13.62
dis
0).
(2)
there
p{x)
Writing
is
p(x)
= p xn +p 1 xn ~1 + ... +pn
an identity
=p
{x
 a x (xa 2 )...(x ak
)
{{x
 btf + cf}
. . .
{(a?
 fy) 2 + cf}
(vii)
in real algebra corresponding to (vi), owing to the exact correspondence between real numbers x and complex numbers of the form x + Oi.
Since k + 21
n,
we see that
ifn is odd, k must also be odd, so that there is at least one factor
of the form x a in (vii), and certainly an odd number of such
(a)
factors;
(6) if
is
even, k
quadratic or an even
Remembering that
in real algebra
will
be
zero,
all
+ p^
coefficients,
11
1
is
well
Proof. Suppose
known and is
h\k
is
+ 1 and
ofPnThis result
term is
integer
which
is
a factor
COMPLEX ALGEBRA
13.62]
of generality
factor
we can assume
0. Then
that
h,
517
common
^k =  (pJi 11
so
+ p 2 hn ~ x k + ... +pn kn
~1
)
= an integer.
Hence hnjk is an integer,
an integer h. Since /.
and
so
is
so that k
,
=
,
1.
a factor of
x2 then
,
and
even or zero
if p{x^),
p(x 2 ) have
the
same
signs.
functions
follows.
By result
(vii)
in 13.61,
2>(a)
=p
(x
 a x {xa 2 )...(x ak
)
g(x),
where
g(x) is
event, g(x)
is
and (a?x ax
all
. .
COMPLEX ALGEBRA
518
[13.62
Examples
Let
(i)
ps
x n +p 1 x n
Ip^i + ..+p^x+Pnl ^
Then
a;
>
~1
+p n
+...
if
1
\p 1 \.\x\ +...
n i
\p s \{\x\
<
If
coefficient (or
in
+ \p^ l \.\z\ + \p n
+ ... + \x\ + l}
follows that
1 it
p(x)
xn
1
p{x)
i? g /{a;
is
x n (l
+ d)
\x\^K,
for all
where
<
\0\
1} ^
if
J.
we say
K
where K
is
arbitrarily large'.
(ii)
. . .
<
a n Con.
p(x)
(x  a x ) (x  a 2 )
+ b%(x  ax
. . .
. . .
(x  a n ) f(x)
(x  a n_!)
+ k(x  a
x)
(x
(xa n + ...
...
 a2
. . .
(x
 a).
In
this
==
0, it
aa
...
0"n\
an
(1)"" 1
(1)" 2
...
1
+1
k.
same sign as
k(l) n
negative.
ax
precisely n.
(3) Rollers
and
is
(2), (3)
COMPLEX ALGEBRA
13.62]
519
Examples
(iii)
locate
Writing
p(x)
then
p\x)
4a; 3
integers.
2
 6a; 2 + 12x  10 =
2(x 
2
1) (2a;
 x + 5),
the factor x
CO
+
X
Sign of p(x)
+00
Thus p(x)
(iv)
than
p(x)
has 4 distinct
less
= x*2Qx 2 + 4:Sxk =
roots.
p'{x)
8
4(a;
13a;+12)
4(a;l)(a;3)(a; + 4).
1, 3.
Value of p(x)
+oo
352A;
23A;
9k
+oo
In order that p(x) = shall have 4 distinct roots we require there to be four
changes of sign in the above sequence; the signs must therefore be
+  +
9
< k < 23.
This will be so if and only if
+.
The reader should sketch the graph of y = x*  26a;2 + 48a;, which has a maximum at x = 1 and a minimum at x = 3 (and at x = 4). The line y = k cuts the
curve in four distinct points if and only if this line lies between the levels of these
two turning points.
lie
between
COMPLEX ALGEBRA
520
Exercise 13(c)
The
algebra is real.
= 0.
3a; 4  4a; 3 + 6a: + 5 = 0.
3a^lla; 2 + 9a;2
2a: 4
0.
= 0.
2a: + a;10 = 0.
If a t < a 2 < a3 < a4 <
4 x* + 2x 2 + Sxl
6
7
xx 3 + x 2 2 =
0.
>
ft,
9 If ax a z ,
n
i
prove that the equation 2
r=l xa r
by considering changes of sign of
a n are
exactly
k.
n1 roots (i)
all different,
p(x)
three
has
(x a x ) ...(x a n )
S
r=\ x ar
(ii)
. . .
10
(i)
If p n
<
and n
is
p(x)
=p
x n +p 1 x n  1 +...+p n
pn <
is
odd
if
and only
0.
Determine the number of roots of the following equations by first finding the zeros
of the derived polynomial ; and locate each between consecutive integers.
11
xi + 4x 3 Sx 2 l
13
12
0.
8a;
5a4 40a; 8  50 =
0.
x* +
6a; 2
12a; 9
>
14 If a
1,
prove x5
5a*ar
+4 =
has 3 roots.
Find the range of kfor which the following equations have 4 distinct roots.
Illustrate
15 xl 14x z +
17 If p
>
24xk =
q and a,
ft
0.
16
2
3a; 4
= xz  3px 2 + Zqxr =
1?
0.
COMPLEX ALGEBRA
13.7]
521
has 3 roots if r lies between pq 2(p q)a and pq 2(p t q)fi. [By
division, f(x) = (xp) (x* 2px +q) + 2(q p 2 x + (pq r), so
2
long
18
/(a)
Prove that x* + ax + b
etc.]
<
0.
*19 Use the argument in ex. (i) of 13.62 to prove that p(x) = has every root
numerically less than 1 + \pg \. [If aisaroot, x n = p 1 x n ~ 1 + ... +p n ; by taking
moduli, \x\ n < \p t {\x\ l}l{\x\  1} if \x\ 4= 1. If \x\ > 1, the righthand side
< bM n /{Hl}> so 1 < p,/{a;  1}, i.e. \x\ < l + \pt \. Thus any root
numerically greater than 1 must be numerically less than 1 + \p,\.]
\
13.7
We have already explained Newton's method (6.73), which applies to equaThe following method also has this advantage, although it
usually does not give an approximation of a required order as quickly as
tions in general.
Newton's.
13.71
parts
will
'
'
=A>
0, f(b)
B<
0.
yA =
o
(b,B)
Fig. 135
(xa),
and
than either a or 6.
Newton's method
COMPLEX ALGEBRA
522
[13.72
Example
Find the
root of
By
we
trial
this root is 1
xz 2x 2
find/(l)
and
2.
If
J(l + h)f(l)
/(2)/(l)
'
21
+ h)l
and since /(l +h) = 0, this gives 5 = S/h, i.e. h = 06.
A better approximation is therefore x = 16. Since
/(l6) = ll, /(l7) = 0487, /(l8) =
the root actually
lies
between
17
(l
'
and
/(l8)/(l7)
1817
18.
/(l7
+ ft)/(l7
+ fc)l7
032
is
1768.
^
"
0009
0009.
13.72 Horner's
'
"
1,e
since /(1768)
^ /(l768 + ft)/(l768)
0241
"
+ h, then
0487
so that h
/(l8)/(l768)
is 17
(l7
'
0719
If the root
+0232,
18, c
1769 to three
1769 in Ex. 6(e), no. 15,
method
Suppose that the equation p{x) = has a root 276 .. and has no other root
whose integral part is 2. Thenp(2), p(3) have opposite signs; this fact, discovered
.
. . .
Example
Solve x 3 ~2x 2 =
correct to 3 places of decimals.
By trial, there is a root between 1 and 2. Diminish the roots
= x 1, i.e. x = y + 1 we get
;
y* + 3y 2 + y3
by
0.
10:
0.
by
by putting
COMPLEX ALGEBRA
13.73]
By trial, this has a root between 7 and 8. Diminish the roots by 7 by putting
= y 1, i.e. y = z + 7; we get
2 3 + 51z 2 + 6672 487 = 0.
by
10:
2s
523
by
t
and
7.
0.
6,
putting
0.
10:
0.
The process can be continued; but since the numbers in the last two terms are
large in comparison with those in the other terms, clearly a good approximation
is
given by
6
*
i.e. t
7,292,800*68,224,000
0,
9.
1769.
Remarks
(a) The method used in the last step gives a rough estimate for the trial root
when applied at the previous stage: 66,7002 487,000 = gives 2 = 7. Actually
23
we should
find that 0(7) and 0(8) have the same sign. This indicates that we
should calculate 0(6), which is found to have the opposite sign. Although this
estimate is rather rough, it is better than none, and often saves much futile
trial. Estimates thus made increase in accuracy the further we have got in the
process; in fact we may approximately double the number of significant figures
already obtained by using this 'division estimate '.
(/?) Horner's method will also give rational roots. Since these can always be
tested for exhaustively (13.62(1)), they should be removed from the equation
before applying the method. The same applies to repeated roots: see 10.51,
ex.
(ii).
To approximate to a negative root, first change the sign of all the roots (by
putting y = x), and then approximate to the corresponding positive root of
the new equation.
(S) Case of nearly equal roots. If two roots lie between consecutive integers
n and n + 1, then p(n) and p(n +1) have the same sign, and detection of these
roots is difficult; more refined methods than 'change of sign' are required.
Assuming that these roots have been approximately located, Horner's method
can still be used.
(e) Although the method applies in principle to equations in general, its
use is practicable only with polynomial equations.
(y)
13.73
36
all
It
GPMII
COMPLEX ALGEBRA
524
[13.73
*Jy{y+*>)
(cm/+c),
+ (6  2ac) y  c 2 = 0,
^f^ + By + C = 0,
=
A 26a 2 B = b 2 2ac, O = c2
and by squaring,
i.e.
where
+ (26  a 2
a 2"
Example
Solve x3 2x 2
Here a
0, 6
= in complex algebra.
= 2, c = 2. Proceeding
step
by
step,
we
construct the
following table.
A = 26a
4
8
a
a4
a8
a 16
2
 2ac G = c 2
4
4
16
16
B=
b2
256
96
9216
49,152
65,536
By trial the given equation is found to have a root a near 17; and this is the
only real root (e.g. see Ex. 13(e), no. 18). Since ctfiy = 2, it follows that
yffy = 2r 17 = l2,i.e.r 2 = 12 where/? = r(eos# + *sin0), y = r(cos0 isin#);
== 11. Thus a > r = /? = y.
From the theory just explained,
a 18 = J. = 9216, whence
hence r
and
16
(r
2 16
)
= G =
65,536,
so
a=
17692;
10632.
Referring now to the given equation and considering the sum of its roots,
a + 2rcos0 = O, from which cos 6 = 083215 and 6 = 146 19'. Hence
r(cos 6 i sin 6) can be found. The three roots are approximately
17692,
 08847 05896*.
Exercise 13(/)
The
algebra is real.
2 3a: 3 9a: + 2 =
x*2x5 = 0.
3
=
root].
5
[negative
x
2x
3
+
4 x3 + x 2 + x 100 = [two places of decimals].
[positive root, two places].
5 x 6 4a; 2 =
6 x3  %x + 1 =
[3 roots].
[Zwo roots lie between 1 and 2].
7 x3 7* + 7 =
1
0.
COMPLEX ALGEBRA
=
8 x
525
e~ x [three places].
*10 x x
2; write the
equation as
a;
a:,
decimals.
two places of
tha;, correct to
decimals.
* 1 3 If e is small, prove that 6 + sin 6 cos 6 = 2e cos 6 has a small root which is
approximately e es [Use Newton's method; assume the series for sin x, cos x.]
*14 Show (graphically or otherwise) that f(x) = 1 + x~ 2 tana; = has a root
near x = K, where
= ^n + nn and n is a large positive integer. Prove that
a better approximation is
+ A where f(K) + \f'(K) = 0, and that A = 1/2K2
Find an approximation correct to terms of order l/K*.
*15 Prove that large roots of seca;= 1 + 1/a; are given approximately by
x = 2rm where n is a large positive integer. Show that a better approximation is
x = 2nn \j^J{mr). [Assume the series for cos 6.]
.
If c
is real,
in the an/plane
2 If
c2
+ s2
3 If
+ s + ic)/( 1 + s ic)
2
+ ra 2 +n a =l and m + m =
1+iz
l + im
(l+l)z,
If zi,
<
z 2 , prove
6 The number z
and radius is 1.
(i)
(ii)
>
iz 2 ;
l^ + a,!
(ii)
>
circle
z 2 .
whose centre
is 1
+ 0*
(i)
If
z
z =
P P
right angle
Zl
af
in terms
is
(i)
a 2
\z
x,
z
and
(6)
arg(z
1)
arg(z 2 )
f arg(z
z).
PP
about
+ ^ZjjZx).
(ii) Two opposite vertices of a square are 1 + 2i, 3 5i. Find the numbers
(i)
If Jc
is
zx
+ k(z 2 z x
geometrically.
P1 OP
E,
2 meet PiP 2 at
and
is the midpoint of IE. If z x = cos \n + * sin in, z 2 = 2(cos \n + i sin \n),
show that I represents ( 1 + ^/3) ( 1 + i), and find the numbers of E and M.
(ii)
The
internal
9 The numbers
and external
bisectors of
z x , z 2 , z 8 are represented
P P
 z 2) a =
2,
by
the vertices
P P P3
x,
2,
of an
362
COMPLEX ALGEBRA
526
(i)
+ 3* 2 z3i 2 =
arg
(iii)
 =
12;
(ii)
*(iv)
Jtt;
z
+ K 2 +zH 2 =
zz x
zz 2
10fc 2
(Jfe
>
0);
1.
11
(i)
the ellipse
%1
(l+a 2
If z 2
(ii)
y2
2
(la 2
=
2
^"2
'
when 6
and
\tt
varies.
liu + iv =
12
is real,
constant, v
z
13
z
where a
ajz
correspond to u
of circles.
c,
then
direction.
n^d^O.
An ellipse has foci + ae, 0) and z x> z 2 correspond to the ends of conjugate
= a 2 e 2 [Using eccentric angles, if
z x a cos
+ ib sin ^ then z 2 = (a sin ib cos 0).]
2
If w = (z + az + 6) /(z 2 + cz + d) and a, 6, c, d are real, prove that the points
*15
(j)
* 16
z for
<f>
which u is real lie either on the real axis or on a circle whose centre is on the
and imaginary parts, and eliminate w.]
In
the following,
17 If x
a>
= a + b,
= a 3 + 6 3 Sa 2 = 3a2 Sa 3 =
3n
18 (i) If f{x) = Ear af, what
a*/z
1.
"Lx
3a,
ILyz
3a 2 ,
3(a3 + 6 3 ).
r=l
Calculate similarly f(x)
zn x
. . .
a, 6, c,
a"
21 If z
Sx* +
are real.
a,
2pxq =
(p, q real),
1,2 ^3, V2  1
prove a
real root of
coefficients
which has
is
0.
for roots.
if
2 <
<
2,
and one
real
root
3a;
+ 4pa;3 + s =
to have
no
COMPLEX ALGEBRA
25 Prove that for n odd, a + x + %x +
+
even there are
or 2 roots according as a <
which is to be determined.
2
. .
x n /n
By
= x+ Ijx,
527
6,
complex
algebra.
= 0.
= 0. [Remove the obvious root x = 1 first.]
28 2xi + 5xi + Sxs + ^xt + 5x + 2 = 0.
29 Sketch the graph of y sec x. Deduce that large roots of x cos x = 2 are
approximately (n + ^)n where n is a large positive integer. Find a closer
27 x6 5x* + 9x 3 9x2 + 5x
approximation.
30
Show graphically
positive integer,
(2w + J) n,
that sin a;
tha;
is
a large
show that
and that a
(2n + 1) n 2 e+*)#.
is
unaltered
when x
is
replaced
by
1/x,
528
14
14.1
14.11 If n is
an
d + i sin d) n
cos nd + i sin nd ;
Proof,
Let
(i)
d + i sin 0) (cos
(cos
<j>
+ i sin
<f>)
(cos
cos(0 + 0)
sin 0)
+ isin(0 + 0).
Similarly,
(cos
by two
this
d x + i sin dx ) (cos
{cos (dx + d 2 )
cos (0J + 6i +
+ i sin
(cos
2)
+ i sin (^ + d2 )} (cos
3)
+i
sin ((^
+ i sin
3
3)
+ i sin d3
S)
by step
way, we find
(cos d x +
Putting d 1
sin d x )
2
(cos
(ii)
.Lei
. . .
. . .
(cos
6n
6n + i sin 6n )
6 + ism d) n
0, this
cos (20)
becomes
say
+ i sin 0) w =
= m. Then
~~
(cos# + isin0) OT
1
cos md + i&mmd
f
(iii)
below.
+ i sin (20).
DE MOIVRE'S THEOREM
14.12]
by Case
(cos
so
(i),
since
*
.
529
cos m# i ammd
=
=
cos
cos 2 md + sin 2 md
ly
Hence again
zp .
Consider
pd
/
I
cos
+*sm
.
2/
by Case
(i),
(cos d + i sin 0) p
by Case
(i)
is
Hence by the
(ii)
a positive integer,
positive or negative.
pd
i
sin
14.12
or
is
pd
cos
since q
is
a value of
+ i sin 0) s =
sa (cos #0 + * sin q<j>) =
(cos d + i sin
s 3 (cos
i.e.
It
now follows
is
an integer or
is
cos pd + i smpd.
sa
. ,
pd
cos
2hn\
pd
+*sinH +
.
2kn\
(1)
=

kx k
,
<
q.
&2
DE MOIVRE'S THEOREM
530
No further
[14.13
0, 1 , 2,
. .
any
q  1 by some
integral multiple of q.
those in
on the unit
circle
\z\
1:
(Ox,OP1 )=pe/q,
and
arcs
PX P
2,
PqiPq
P^Pz,
sub
r(cos#Msin0),
Fi8 136
where
tt<6^7t,
cos
+ i sin
2kn
+
,)!
= 0, 1, 2,
q 1 and (as elsewhere in this book) rpli denotes
the positive qth root of rp They are represented by points equally
where k
circle
\z\
rplQ
to cis0.
14.13 Examples
(i)
l) n
z".
= cosO + isinO.
DE MOIVRE'S THEOREM
14.13]
531
By taking nth roots of both sides, and using the general result (i)
z+1
2kn\
cos0 +
2kn
cos
a kn
zl 2 sua 2
and
2%z sin
hn
^
^ ,
If k 4=0,
kn
n = cos n
1 /cot
kn\
1.
1)
kn
(k
...,n
0, 1, 2,
= ll + icot
(k
l,2,...,nl).
1 solutions.
stage
(a)
1).
kn\
1 /
\
1)
2^\
1,
h*sui
*sin
Jen
1,
kn\
kn
cos =
n
n
kn
cos
(
I
h2*sm
2kn\
l.e.y
1+* sin
A , ft
0,l,2,...,nl).
(k
2kn\
+*sin
in 14.12,
'
above by writing
cos 2kn + * sin 2kn',
z+1
2kn
cos
for
From
=
z
which takes
gives
1;
. .
.,
n 1;
1,
etc.
a,
a2
cos2&7r+*sin2&7r,
2kn
when k =
2=1. Write
0, 1, 2,
2n
cos
2kn
Msin
. .
.,
n 1.
2n
h*sin
then since
n +
2kn
cos
2kn
can be written
cos
a=
where k
0, 1, 2,
distinct values
The value k =
H*sin
2, 3, ...,n
1,
*sin
the
2kn
n roots
2n
(cos
\
are
1,
n
a,
h*sm
2n\
a",
a2
a n_1
=
1 =
l + a + o^+..+a"
0.
is
is
zero (13.51),
often useful.
we have
de moivre's theorem
532
[14.13
Compare the brevity of this work with the algebraical treatment of the equation
23
a n ~r =
[2{nr)7i\
 
cis
{2rn\
>
n
and hence a n_1 =
a,
a na = a 2
....
2rn
cos
2rn\
2rn
r
oc ,
*(iii)
the
ism =
n
cis27T
common factor
0, 1,2,
...,q
(ii)
1.
'.
cis
(d + 2k 1 ir)
then as in 13.22(2),
P (d
?
for
some integer m,
has a factor in
i.e.
common
Remarks
must be
the
the equation z 9
same as
'
3>
Exercise 14(a)
1
If x
cis d,
cis 0,
and m, n are
yn
Simplify
2
.
Prove
xm
(l
/l+sin0 + *cos0\"
(
\1
+sm0 *cos0/
integers,
+ cos0 + *sin0) 8
.. ,,
(cosf *sin/)*4
=
cis
prove
(inn no), n
bemg an mteger.
4 If a;
5
DE moivre's theorem
down
8 Write
+ i sin 0)*.
(cos
Observe that
cis {f(0
cos 3 03cos0sin 2
function.
first
By
cos30
533
and
sin 30
3 cos 8
sin 6  sin 8 0,
10 If z
cis 0,
express in terms of 0:
1
(i)
\;
z+ ;
(ii)
(iii)
z\;
z" + ^;
(iv)
z~.
(v)
11
= l.
13
(zl)"
15
z 2n
12 (z+l) n + (zl) n
= z.
14 (1+z)"
0.
(lz) B .
16
prove
cis 0,
\nd
21 n
Gr cis rd.
r=0
is
a positive
integer, prove
(l
+ a;)" =
c a + c4
and
cx
c3 + c 5
18 If
2.
cisf7r
and
\
1
+ *) n +
it\ n
4w/
i) n =
. . .
. . .
2* B sin Jtmt.
= a + a 2 +a 4 7 = a 3 + a 5 + a 8
/?
2n
fi,
7,
19
cos
If
4=n
87r
prove /ff+y
'
and
and a
= 1 and
+sm + sm = +JVv 7
sm
7
7
7
.
prove
/?y
2 n ~* I cos
2w
r=
17 If w
S n Cr cos =
Deduce that
0,
prove
that
[Put x
20
senting
all integral
powers of z should
points repre
by the sequence
....
*21 If n is any prime number and A denotes any complex root of z n = 1, prove
that the numbers 1, A, A 2
A" 1 are some arrangement of the numbers
2
a n_1 in 14.13, ex. (ii). [Use the argument in 14.13, ex. (iii).]
1, a, a ,
,
DE MOIVRE'S THEOREM
534
14.2
[14.2
This transformation
is
cos0 isin#,
2cos0 = z+,
so
Also zn
2*sin0
of
l/zn
(i)
zn
1

sin nd.
n =2i
+1 = 2 cos n6,
(ii)
By means
(i)
be expanded in
can be expressed in multiple angles.
Examples
(i)
Express cos8
From
(i),
(2cos0) 6
= (z+^y
=
=
=
..
(ii)
^z +
=
cos
sin4
(ii).
^ (z^j*
= (*+!)
(2cos0) 5 (2isin0) 4
.*.
cos
Express cos 6
+ .4 + z6
+ 6z4 +15z 2 + 20 +
z2
z
z
15
(*.4*. +8 i + I)
'4.) + (* ,+
?M* + ?)  *(* + ?) +6
(*
+ i)
= 2 cos 90 + 2 cos 70  4(2 cos 50)  4(2 cos 30) + 6(2 cos 0).
= jfa (cos 90 + cos 704 cos 504 cos 30 + 6 cos 0).
de moivre's theorem
14.22J
535
We
even,
and terms
like z
\\z
if
odd, therefore
is
14.22 cos
an
/18,
sin
#18,
is
can be
m is even or odd.
if
integer)
sinm
cn~Hs +
 (2) cn
Hence
"
cos nd
6)
cn_M 
cn 
c"3 (i5) 3 +
HI
~
cn 2s2
+ is) n
(c
say,
j^j
c^is)* +
cn ~ is 
(3
+ j^j cn~^ 
sin nO
and
+ i sin 6) n =
cn*(is)*
j^j
v+
(cos 6
14.21
. . . .
. . .
cn ~*s * +
. .
(hi)
(iv)
Taking the ratio of sinnfl to cosnd, and dividing top and bottom
of the righthand side
by
cn ,
t*nn0 =
where
we obtain
U/.
1t
W
WW
tan#.
";
(v)
[%'.
s2
+ c2 =
Examples
(i)
sin 66
c8
=
=
c8  10c8 ( 1
=
=
cs*
 c2 + 5c( 1  c2
16c8  20c8 + 5c.
c 8*
c8s 8
2
)
cs 8
sin 60
/.
c 8* 2
sine'
32c
c 2 + 6c(l c 2
)
 32c + 6c.
8
2
)
},
DE MOIVRE'S THEOREM
536
[14.22
(ii)
if (a)
(a) If
tan 50
0,
then 5t
lOt*
+ <B =
x 10x
6
so
0,
+ 5x 
tan 6
is satisfied
by tan
satisfies
0.
Now if tan 50 = 0, then 56 = m and 6 = \rir, so that tan 6 = tan Jnr, and the
values r = 0, 1, 2, 3, 4 give distinct values of tan Jttt. Therefore the roots of the
above equation are x = (corresponding to r = 0) and
tanf7r, tan7r, tan tt = tan f
tanf7T = tan^7T,
7r,
i.e.
(6)
If 50
then tan 50
\t\,
5t
hence
= tan
tan \tr.
tanf7r,
0,
and so
 10 + 1* = 1  10*2 + 5**;
s
satisfies
But if tan 50 =
1,
then 50
(4r+l
a:
The equation
tan
0, 1, 2, 3, 4.
("20~7
0.
= ^n+^rn, so that
fyr+rn and
'
2 ' 3 ' 4 )
by the method
in
Ifnis
sinn0
61 s + 6 8 s 8 +
...
+ b n sn
= (r 2 n 2
,
br ,
68 .
From formula
sinw0
=
(^j
coefficient of s n
= ~ I)*"""
(
 _ i)fcni )2 ni by
(
(j)
+  D*"
<
2.
DE MOIVRE'S THEOREM
14.23]
Also
&!
537
/n\
= II =
coefficient of a
n.
2& s
ncosn# =
twice
wo
6,
we obtain first
2 r&r*r_lc
r=l
 n 2 sin nd = 2 0(r  1)
and then
r=l
r2 c 2
 r&r *r1*]
r=l
r=l
hence
Equating
coefficients of r,
 n26r =
(r
cos(0 1 +
(cos 6X
+ 62 +
+...
+ 2) (r + 1
follows.
= (ln2 )n,
. . .
&r+a  r 26r ,
Taking r
68
i.e.
1,
we have
= Jn(lw2 ).
+ 6)
+ 0J + isin(0 1 +
+ i sin 6 X
(cos
+ ...+0J
+ * sm ^2)
cos @n + * sm ^)
as in 14.11, proof
cos 6X cos
. . .
cos
cos
cos
. .
cos Gn {l
where
TO (
+ #2) (1 + it2
. . .
1+ itn )
where
t
r
(i),
tan
r,
+ iS x + i 22 2 + i3S 3 +...),
taken r at a
time; this last step follows from the argument used in proving the
binomial theorem, 12.11. Equating real and imaginary parts,
of the products of tlt
cos(01 +
+...+0B ) =
sin(0 1 +
+...
2 , ...,tn
+ S 4 ...),
(vi)
and
By
The formulae
n)
(vii)
+...).
division,
tan(01 +
. . .
+ ...+0J =
(vi)(viii)
dn
6. All
include
^~^^ S~;;
5
(iii)(v)
down
when
any
would cease
easily for
to have a meaning.
()
>/
DE MOIVRE'S THEOREM
538
[14.3
Exercise 14(b)
Express the following in terms of multiple angles.
2 sin4 0.
cos*0.
sin 5
4 cos*
cos6
0,
sin6 0.
sin8 0.
Calculate
'in
6 Jsin7 0d0.
JW0sin6 0d0.
(ii)
sin 0.
11
in terms of sin 0.
Solve completely
13
1 6 sin 6
30+10 cos 6 =
14 Prove cos 70/co's = 1  2(2 cos 20)  (2 cos 20) 2 + (2 cos 20) 3
that the roots of x 3  x 2  2x + 1 = are x = 2 cos {%(2k + 1) n}, k
15 If x
2 cos 0, prove
\.
sin 50.
(x
Hence prove
0, 1, 2.
x ~2x l) 2
2
[Express
in halfangles.]
*16
Give the last terms in formulae (iii), (iv) when n is (a) even; (6) odd.
State the last terms in numerator and denominator of formula (v) when
even; (b) odd.
(i)
(ii)
n is
(a)
17 Write
tan0
1,
tan 7T.
a + /?
is
an
*20
for
odd.
14.3
We
Factorisation
now
Theorem
DE MOIVRE'S THEOREM
14.31]
539
to denote /(l)+/(2)
+ ... +f(n),
here
it is
con
venient to write
n/(r)/(l)/(2).../(n).
r=l
clear
xl
14.31
We
have xn
if
xn
cis2?rr, i.e. if
cis (2r7r/w).
For
. .
. . .
\n
to
2kn
a; (cos
and
. . .
their product
/
even,
x Icos
2kn
/'
*sm
2&7r\
1,
ft
2&7T
.
2kny / 2&7r\ 2
= #2 2#cos
+ sin
n
n J
n J
\
2kn\
is
x cos
h*sm
factors.
xn
(x
0,
hi.
1) (# + 1)
* n_1 /
# 2 2a; cos
2kn
+ 1\
l.
(i)
=
1 + 2{%{n
n factors altogether. Arguing as before, we find that
1)}
Case
():
*(nl)/
fornodd,
14.32
xn +
xn l
(xl)
[x*2xcos
2kn
+ 1).
(ii)
consecutive values of r.
37
GPMII
de moivre's theorem
540
Case
(i) :
As
even.
we wish
in 14.31,
[14.33
X CIS
.
we
2rl
71,
n
r'
+ (2r'l) =
1)
r=l,
pair
X CIS
2kl
X ICOS
+ *Sin
 1, 2,
is
2kl
n
 {%n 1).
X COS
2kl
xn +
in l
%
II \x
njn}
(ii)
r=l,
with
r'
factor x +
1.
Then
for
\(n +
odd,
factors
2kl
n
\
7T\.
+ 1 Hence
.
2xco&
2k
tt+
\
1
(iii)
1 .
odd.
The value
l&m
.
TT
&=i\
Case
The
altogether.
\
7T\,
x* 2x cos {(2k
even,
therefore
7T
=  (r 1);
0, i.e. if r'
This gives \n + {l
corresponding to r
71
3, ...,\n
2,
0,
2r'l
0,
3,...,i(nl)
2,
1, 2,
gives 2r
1)
as before
xn +l
%(n3).
n, to
real
we have
= (x+l)
x^lx" cos na + 1
The equation x 2n 2xn cos
Hnl)/
2k 1
\x 2 2x cos
H
k=i
n+l).
(iv)
14.33
(x n
+1 =
is
quadratic in x n
.'.
cos na
satisfy it are
2rn\
i sin na.
2rn\
cos
where
0, 1,2,
...,n
1;
na
is
zero or an
de moivre's theorem
14.33]
541
cos
2kn\
x*
rt
2k7T\)(
\
ni
x 2n 2xn coanoc+l
:.
2/br\
2kn\
2x cos a
to r
.
is
2kn\)
+ 1.
J
i
2
]J lx 2xcoa^a +
2kn\
\
+
l\.
(v)
Examples
Various deductions can be made from the result
(i) Putting a = 0, we obtain
(rc
and
if
n is even this
is
n if
2hn
&=ol
 l) 2 = T\W1xcos
(v).
+l},
)
equal to
in II
(xl) 2 (x+l) z
W~2xcos
fci
2kn
\*
+ 1}
)
(ii)
xB H
Putting x
2cosraa
*n
M
TT
n
+
2cosa
{a?H
*ol
2hn\\
1}.
J)
cis
is0,
8,
2cosn0 2cosna
fc=ol
ni
cos nd cos net
\
/
2" 1
2kn\\
(iii)
In
(v)
put x
and a =
sin 2 w/?
n 1
J] 2lcos(2y6f +
n1
2 2 " 2 II sin 2
o
)}.
J)
= 2 n  1 *n
o
sin
2kn\\
jl,
klt\
ft
\
1
sinn/?
2/?:
2(lcos2n/0
.'.
)},
})
(V+ \
nf
\
372
DE MOIVRE'S THEOREM
542
[14.34
To decide the sign of the righthand side, first suppose < /? < rt\n\ then
each factor on the right is positive, and so is sin nfi. Hence the sign + is appropriate for this range of /?. As /? increases, sinn/? changes sign whenever /? passes
through a value kn/n, and simultaneously one factor on the right changes.
Hence the sign is always +
:
nl
2"" 1
=
(iv)
By
w\
+
n sin(/?
/
fc=0
\
assuming that
nl
loglsinwytfl
fc=0
nl
wo
Deriving
(v) de
ncotnfi
S cot IAH
fe=0
\
jfc7j\
sin(^ +
/\
lvn\
nj
of the circle.
Also,
by the
6+
2m /n.
cosine rule,
Fig. 137
PA* = x 2 + a*2axcos[d
(" +
v)nll
PAl.PA\...PAl_ 1 =
=
by a
slight extension of
formula
(v).
2
II a 2aa:cos0 +
j+a
2rn\
x 2n  2a nx n cos nd + a 2n
).
\.
last
two
results are
is
as Cotes's properties.
odd
By ex.
is
known
(iii)
DE MOIVRE'S THEOREM
14.34]
Hence
for r
0, 1, 2, ...,n
1,
543
consequently
n _!
( sin 6 sin
&
,
sin n#
^
r=l\
w/
(vi)
I.
n distinct values of
+
r = 0, 1, 2,
1). The factors which corre k are sin 6 sin (kn/n), sin 6 + sin (kn/n), and have product
by
spond to r = k, r =
sin 2 aw? (kn/n). Hence
sin w#
!>/
6 n sin #
where
sin 6
(vii)
(vii)
I.
sin 2 (9
~ am? (kn/n))'
foA
^D/
A sin/
sin nd
sin 2
(sin 2
JJ
fc=l
J. is
and then
A=
.
sinn#
= n,
lim
d^O sin d
;
i(nl)
and so
sinw#
= nsin0
TT
1
fcL !
am 2$
rtt^ r~> \
I
am^ (kn/n) J
(viii)
Examples
Comparison of series and product.
(viii) with the expansion of sinw# obtained in
ex. (iii) of 14.22, we have for n odd:
(i)
By
i(nl)
ns
n
ft=l
 s 2 cosec 2 =
n/
]c7t\
ns + %n(l
w2 )s3 +
...
+& B s B
cosec 2
2j
ft=l
KD
i.e.
cosec 2
fc=l
n(
w = A(n
Jen
1)
n 2 ),
for
odd.
(ix)
CO
(ii)
If
Proof thm
<
<
J7r,
J
2  = tt 2
then sin 6
J
Put
<
<
<
cosec 2
tan 6 and so
1, 2,
\(n
1):
de moivre's theorem
544
Writing
(ix),
we have
7T
<
Since the series S(l//" 2 )
some
to
limit s
1)
sequence
for
nl
< ^ "J" + **
7r
oo in
odd
n.
so that s
22
W 2 1
known to converge by
is
when
11
[14.34
<
7T 2
s,
\n*.
Exercise 14(c)
s
6
1 Obtain real quadratic factors of x X +1.
x B and of x 3 prove that
By
equating coefficients of
and
s
2 Find the
is (x 7
for x and
2/3
nl
5 sinn/?
= 
6 cosn/?
= (l)*2 n ~ 1
2 n_1
l)* n
2~ 1 IIsinj^ + r=0
I
2w
n1
].
)
rjjX
JJ cos
i
>
if
even.
is
(2r+l)n\
]J cos{,ff+
if
is
nl
2 cosec
/ =
/
2
r7T \
/? H
r=0
9 Prove ch nx cos na
by
n
o;
ev
n_1 f
[Divide formula ( v)
/J
.]
[Derive
Prove that
1
x n + n
x
TT
A {* +
[Use formulae
12
wo /?.
2/7r\)
2" 1 TT { ch x cos ( a H
r=o I
\
even.
w 2 cosec 2 n/?.
^.
(2r+l)n\
4 cosw/?
or 2/3+n/n for
the following.
8 Prove
4a; 4
l)/{x 1).]
By substituting
+16.
quadratic factors of x6 + x5 + xi + x 3 + x i + x+l. [The
expression
2cos
(iii), (i)
By putting x =
2n
/2r+l
\)
n\),
xn
J)
xn
cis
nl/
(i)
cosn0
2" 1
sin nd
11,
x) rJi\
prove
II I cos 6 cos
r=0 \
n1
(ii)
"\
lVt1 a;
1
/
+ 2cos
II
= /x\
2r+l
n
2n
\
)
77A
nj
n
,
de moivre's theorem
14.4]
13
(i)
By
(ii)
Prove
ni
Vn = +2" 1
II sin
r=l
14
545
By deriving no.
(i)
12
logarithmically, prove
(i)
nt&nnd _ n ~ 1
2j
sin0
(ii)
> 0,
12
(ii)?
With the notation of 14.33, ex. (v), prove that if P lies on the circumference,
15
then
PA PA X ...PA^^
.
2a"
16
= i tan
sin w0
of
+ x) %n +
2rl
Deduce that
(l
+ a;) 2n + (la;) 2n =
" sec
 xf n =
where
n,
4n
1, 2, ...,n.
2 FT a; 2 + tan 2
r=i \
(2r
2
1)tt
4n
r=l
2n 2
are
(2rl)n\
cosn0 cosraa
2r7r\\
/
2 n_1 JJ <cos0 cos a + }>
r=ol
14.4
no. 20
and arguing as
/;
in 14.34.
Roots of equations
= %n,
equation
equation
(ii)
%Tt,
3
4a;
is
tt
all satisfy
3a; = is
satisfied
therefore
Form the
by
a;
+1 =
= cos 30.
40
It
3 cos 0,

the
The required
where
cos 30
is satisfied
cos f7T.
by
= 2mn 30,
i.e.
= \mn
by
(this includes
both solutions
1, 2, 3,
Since
tion
20 1 =
8a; 2 +l
4a;
3a;,
l) 1. The roots of
_ ^ cQS
cQg
CQg
4
8
8a; 4a; 8a; 2 + 3x+l = (x1) (8a; 3 + 4a; 2 4a; 1),
8a; 3 + 4a; 4a; 1 = 0.
since cos40
m = 0,
2cos
viz.
2(2x
by
^ ^ ^
CQS Q
de moivre's theorem
546
(iii)
Form
the equation
Writing
W=
tan 6,
becomes
it
we
0,
it
formula
is
tan
any
%tt.
integer or
(v) of 14.22)
(see
\tt,
+ tan^7T,
Putting x
by 6 = \rm, where n
is satisfied
<
has roots
[14.42
1, 2,
. .
.,
8, i.e.
tan^7r.
tanf7r,
tanf77,
follows that
has roots tan 2 7r, tan 2 f7<r, tan 2 \tt, tan 2 f7T.
N.B. Since tan 2 \tt = 3, we may remove the root x = 3 from the last
equation, obtaining x s B3x 2 + 21x3 = Q, which consequently has roots
tan 2 Aw, tan 2 7r, tan 2 7r.
14.42 Results
coefficients
Examples
(iv)
6,
6.
From ex.
(ii),
sec %tt
2/
=x
2
,
+ 4x 4x8 =
i.e.
2/
f7r,
= sec ftt,
= y 1,
i.e. 2/
 24t/
= z + 1,
23
4)
16(2/ 2)
becomes
21z 2 + 35z7 =
tan 2 f7r,
etc.
is
is
2
,
+ 80^/  64 =
this
sec far
0.
Writing
sec ftt
x
Putting
6.
0.
(after reduction)
0,
is
21.
hence
DE MOIVRE'S THEOREM
14.42]
(vi)
547
4x3 Zx cos 3a
is
by 6 =
satisfied
oc
+ f for
0,
where x = cos 6, and is satisfied by the distinct values x = cos a, cos (a + ftt)
and cos (a + ftt). By taking the product of the roots, the relation follows.
Exercise 14 (d)
1
Form
4 Show that x
2m
2 cos
yy
(r
1, 2,
. .
.,
5)
0.
2 at the end.]
Construct the equation whose roots are 2 sin f77, 2sin^7T, 2 sin f77
= 2^(1 2 ) and the result of ex. (ii) in 14.41.
3
2
(ii) Verify that 2 sinf7r, 2 sin f7r, 2 sin f rr are the roots of a; = + ^7 (a; 1).
[The equation in y obtained in (i) can be written
*5
(i)
by eliminating x from y
8
2/
7(2/2 1) 2 ,
i.e.
= Jl(y*l).
Since 2sinf7r > 1 and 2sinf7T > 1, and also 2sinf7T = 2sin$7r
and 1, these values of y give y 3 and y 2 1 the same sign.]
Use
lies
between
= f
= 33 and tan \tt tan f7r tan %tt = + V 3
8 sec 2 77 + sec 2 f7r + sec 2 $7r = 36 and sec 4 \tt + sec 4 %tt + sec 4 \tt = 1 104.
9 Calculate sin2 fzr + sin2 f7r + sin 2 ftt.
10 Prove sin 3a = 4 sin a sin (a + ftt) sin (a + f7r). [Consider sin 30 sin 3a.]
1 1 Prove tan a + tan (a + \tt) + tan (a + f7r) = 3 tan 3a.
[Consider tan 36 = tan 3a.]
*12 By considering tann0 = tanwa, prove that the equation
6 COS TT COS %TT + COS fTT COS TT + COS f71 COS %TT
7 tan2 \tt + tan2 ftt + tan 2 %tt
( l)
has roots x
X *~ 1 ~
a;n
~3
(3)
cot (a + nr/n), r
' "
2tt.
"~ 2
~~
fl5
(2)
w 1, where a
n/ = n
r=0
a;B
{
0, 1, 2,
n1
tegral multiple of
tan wa
'
"
is
not zero or an
rn\
" '
cotna.
in
DE MOIVRE'S THEOREM
548
14.5
[14.5
summation by C+iS
14.51 Cosines
and
(i)).
Write
C=
+ cos (a + 2fi) +
and
8 = sina + sin(a+/?) + sin(a + 2/?) +
Then
G + iS =
cis
a + cis a
cis
a{l
on summing the
multiple of 2n.
cis /?
. . .
...
+ cos {a + (n 1) /?}
+ sin{a + (n 1)/?}.
. . .
+ cis a(cis
provided
cis/? 4= 1, i.e. J3 is
not an integral
Now
l(cis^) n
1cisn/?
= 2* sin ^i/?(cos
= 2i sin
C + iS =
Hence
=
Equating
C=
If
/? is
real
cis
cis
2i sin
cis \nfl
2i sin /? cis
sin Aw/?
.
and imaginary
parts,
/?}.
we obtain
^^ cisla + lnl
sin \nB
.
+ i sin
sin InB
f
+ ^
1
\ C sin{a
sinf/?
.
, x
G = wcosa and S
(i), (ii).
DE MOIVRE'S THEOREM
14.52]
549
Sum 1 +
(i)
a;
cos 0+ a; 2 cos 20 +
Write
G
terms, where
x is real.
andf
S=
Then
C + iS = 1 +
=
ton
...
a;
sin
a;
cis
+
+
a;
a:
sin
cis
l+a;cis0 + (a;cis0) 2 +
+ (ccis0) n  1
...
n
= 1 (*cis0)^ on summing the
1 X CIS
.
x cis
g.p., if
4= 1
# n cisw0
x cis
_
~
(1
(la:cis0){la;cis(0)}
to
1
If
a;"cisn0 iccis(
l 2x cos
+ x2
0) are conjugate.
<
a;
1,
cos
a;
then x n
>
given by
lrm
Taking the
\x\
1,
there
is
no sum to
(ii)
103 008
w _kjo
If
_ 1
N.B.
make
2a; cos + x 2
infinity.
sin (a + /?)
^ sm
a+
=
Write
and
*S'
Then
cosa+^
C=
= sina+^
C + t =
cisa +
cos(a+/?)+Q^
sin(a+/ff)
j^j
^
^
cis(a+^)+
= cisa+^^ cisacis/?+^^
=
f
with
cisal
^ cis^+^
cos (a +
2/?)
sin(a + 2ytf)
cis (a
cis
+ sin (a + nf})
. . .
+ ^/?).
+ cos (a + n/?)
+ 2/?) +
a (cis /?) 2 +
(cis/?) 2
(2cos/ff) n sin(a
. . .
...
+ cis (a + w/?)
+ cisa(cis/?) n
+ ... + (cis/?)"J
de moivre's theorem
550
=
=
=
=
=
S=
.'.
+ cis/?) n
cisa(l + cos/? + *sin/?) n
cis a (2 cos 2 /? + 2i sin
[14.6
cisa(l
cis a (2
cos /?)"
n
(2 cos \fi) cis (a
(2 cos J/?)" sin (a
+ in/?).
+
on equating imaginary
parts.
Exercise 14(e)
2 Prove
Calculate Y.
to 2w terms.
S1X1 { Th
sin
X)
J~
cos"
sin 2r0.
rlW
4 Find
n
(2 cos 0)
(^j
1
(2 cos 0)" cos
2
2(9 f^j (2 cos0)" cos
...
to (n +
2n/n and
1)
terms.
5 Sum to w terms the series whose
w is a positive integer, prove that
rth term
is (2r
1) af1
If a
7 If
is
sin a + cos
8 If n
is
. . .
sin (a
a positive
+ nx cos +
where r
+ 6) + cos 2
26+
= + A/(l + 2cos0 +
+ cos 3
a:
c )
and
cos
x n cosnd
30+ ... +
cosa:sina
14.6
sum to infinity of
sin (a + 30)
. . .
prove that
integer,
x* cos
sin (a + 2/9)
If
(1
r n cosna,
+cos0)
:scsin0
:r.
(/), (g).
Infinite series
tions of a
14.61 Convergence
lim 5w
n> co
o"
+ *r.
'
for,
for all
n>N,
\<r n
<r\
<
\e
and
r n
r[
<
\e
DE MOIVRE'S THEOREM
14.62]
and so
\s n
(<r + iT)\
if this last
condition
Conversely,
\(<r n
o"no
we have
<r B >
(and
cr
For the
(2)
<
is satisfied
(oo")
<r n
+ tt <
<r]
for all
5s
N, then
2zr where
,
= 2
zr
n
zr
+ *(tt) <e,
a;,.
= 2 r + * 2 2/r =
sav
+ * T>
"
r=l
r=l
r=l
infinity of
2zr
is
the
when
sum
to
2zr is
e.
since
series
Sn
<r)
551
terms)
zr
converges,
j
we say that
It follows that if T,zr is a.c, then lLzr is also convergent in the sense
(2). For since \xr < J{x? + yf) = \zr \, and S \zr is convergent by
converges;
hypothesis, the comparison test of 12.41 shows that S
of
i.e.
ljyr
2#r
14.62
is a.c.
The
+ z + z2 + z3 + ...
sjz) = 1 + z + z 2 + ... + Z"1
infinite g.p. l
Write
by z and subtract
we find that, if z 1,
If we multiply
ratio),
g.p.
with a real
common
=i=
Putting z
zn
1z
lz
Ifr
(just as for
0,
we have
rn (cosnd + iainnd).
>
r
n cosw#
rTC
and
r
n sin?&0
limz
when
\z\
<
oo.
rn ,
Hence
1.
n>oo
Thus when
If r
>
1,
n>oo. If
z
< l,lims
TC
(z)
1/(1 z).
then r n > oo when w > oo, and so sn (z) has no Hmit when
r = 1 and z =j= 1, then zn = cosnd + isinnd; but cosnd,
DE MOIVRE'S THEOREM
552
smnd
oscillate
between
[14.63
= 1, clearly sn (z) = n.
2
The g.p. l + z + z + ... therefore converges only if \z\ <
to infinity is then 1/(1 z). We can write
tend to a
(1z) 1
The
l
g.p.
+ r + r + ...
<
+ z + z* + z* + ...
absolutely convergent
is
sn (z)
does not
limit. If z
This
converges.
l,i.e. \z\
<
1.
is
(\z\
<
1,
1).
the case
14.63
The exponential
series
Z2
Write
(1)
r(cos 6
converges for
all r,
23
+ l! + 2! + 3! + 
+ Ti + 2! + 3! + 
the series
z
z2
zz
+ T! + 2! + 3! + 
is a.c.
necessary,
we have
(z
+n+
1
1!
is
real series
whose sumfunction
in printing.
series
(i)
having
+  + n\)
2!
iTi
We
ex is
(ii)
many
z.
+ n + 2! + 3! + 
infinitely
for all
zn \
z2
sum
to infinity of
z,
and
it is
necessarily unique.
it is
Case
method of
more suggestive than exp (iy).
is
5.43,
de moivre's theorem
14.63J
(2)
>
By
now be shown
can
it
(ii),
553
that expz
= exp (z1 + z2
exp z x x exp z 2
just as
Let
exp x does in
zx
r lt
z a
real algebra.
and write
r2 ,
*W =
*n(Zl)*n(
nv u
a)
n \ 2 a/
+ Tl
z2
2!
2!
{1
nlf\
nl'
22
11
2!
ra!j
+ + S +...+ \
2*
2?
+ + ...+
2!
w!
1!
Z2
2X 22
2*Z a
g" za
nil!
1!
111!
211!
w2
a
2
w2
zl z2
^.2
z
z
l 2
2!
1!2I
212!
_J_2
+ i_L +
__
l!n!
Jg
2i
2g
rtw^fi
n\n\
' :
(s2)!2!
+ J1
. . .
2!n!
,ZLZl
n!2!
>2^n
2t*
w!
*i
f,
zn
2,
+ _2
"1
++
Zl
Zl
Z l\
= L+
++..+}
T
T
T
ii
^
s!
(iii)
),
(s
<
n)
iv )
2j
s!
1
= ^ (*! + *)
and consequently
2*2
*( z l)*( z 2)*n( z l+ z a)
where the
last
summation
2
p,qP '"'
is
diagonals.
We know
that,
when n
>
oo,
* n (r 1
From
+ r2 =
and
zero, so also y\
s n (r1 + r2 )
 >
v)
+ eri+r*.
0.
(iv),
12
^ N'W' ^gl!^
But when w
 oo, n (2 x )
(iii)
when
n>co.
follows for
DE MOIVRE'S THEOREM
554
14.64
By
[14.64
(iii),
exp{x + iy)
Now by putting z =
x in
(i),
expz= !+^+ +
=
Also
n (y)
ex
for all x.
1+^+^+ .+ n\
1!
2!
2
y ,y*
2! +
series in
4!rn
where the
2/
3
,
Since
series.
by
12.61 (2)
cost/
"3! + ^"'
y
+ ^...
=11
2/
y
y
amy = y __+__...
and
for all (real) values of y,
exp {iy)
lim sn {iy)
n> oo
=
expz =
Therefore
exp
(z
cosy + isiny.
is periodic,
(vi)
ex {cos (y
ex (cos
y + i sin y)
exp z.
From
(vi),
exp (iy)
cos y + i sin y,
exp
iy) =
cos / i sin
t Cf. the definition for real functions in 1.52(2). It can be shown from (vi) and
the fact that cos y and sin y have period 2n that 277* is the number p of smallest
modulus which satisfies exp (z+p) = exp z for all z, and that every such number p
is an integral multiple of 2ni.
DE MOIVRE'S THEOREM
14.66]
First adding,
cos y
and then
f {exp
{iy)
subtracting,
+ exp  iy)},
we find
sin y
555
(vii)
\{eiy + eiv ),
1/2, in
sin y
^ (eiv  eriy
),
are
14.66 Examples
(i)
infinity of
tan 0,
t
S=
and
(ii)
if
+ +
2 z
3 3
where
cis 0,
= exp {tz)
for all z
and
exp (sin
+ i sin
tan 0)
G=
....
all real
tan 0).
method of
the quadratic
a2 + 4p 2
that
..
trial
sin20 + sin30 +
sin0 +
O + iS = 1 +tz+
Then
46,
we
by a formal
m + am + 6 =
2
find that in
application of 5.32
(2), II,
with that
found in 5.33 (1), Case (iii), so that A', B' must be real constants; in fact A, B
in the above calculation are conjugate complex numbers.
(iii) Integration
notation)
38
= cos* +i sins.
GPMII
de moivre's theorem
556
and we
[14.66
u + iv to
= ie
dx
%x
le ix )
i.e.
More
by formula
generally,
(o+rt)!B
,\
e ax ( cos
{e(a+it)xj
dx
e ax
(vi)
fa
we have
(for real
e ax cog
constants
a, 6)
fa _j_ 1 g ax 8m faf
Thus the
stants m,
e(a+ib)xdx
where
c is
a+ib
e(o+t6)a!
real con
+ C)
This result permits rapid calculation of certain integrals involving combinations of exponential and circular functions. For example, put
then
e(.a+iU*dx
a + ib
a ib
a* + b
e ax (cos
e(a+ib)x
+C
bx + i sin bx) + c.
Z
e
+c
G
Ex.
Cf.
L,
S=
e + c2
cos,
sm
fadx;
35.
Exercise 14(f)
Simplify the following.
1
exp {(a
6 exp {(a +
7 Expand
3 exp(logr+i0).
+ i7ra).
exp {cis 6} + exp {cis 6)}.
2 exp(l
exp(?ri).
008
6.]
DE MOIVRE'S THEOREM
14.67]
557
10
11
5!
3!
r=lT"
i^.
12
r=l
u + iv = exp (x + iy), find the locus of the point (u, v) in the wplane
x = constant; (ii) y = constant. Verify that these two loci cut ortho
14 If
when
(i)
gonally.
15
argument.
dy
= sin x +
dx
1.
If y
*(cos# + isina;)
A e ix
Since y
%y\
when x =
0, therefore
cos x
therefore
A=
'
sin z
(vii),
we define
+ exP ~ * z )}>
{exp
(viii)
{exp
(iz)
exp iz)},
(
sinz/cosz, etc.
and
also to
cos z
cos z i sin z
= exp (iz),
z2
z4
26
2!
4!
o!
= exp iz),
(
= 1 ^7 + 77771 +
(all z).
Sm2=
z3
l!3!
(ix)
z5
+ 5!
it
can
now be shown
that the
cos0=l,
and
sin0
0,
cos
z) =
cos z,
sin(
z) = sinz,
de moivre's theorem
558
[14.68
+ z2 =
to translate each
(viii)
left.
We define
ch z
{exp
(z)
+ exp  z)},
sh z
{exp
(z)
 exp  z)},
(x)
to be shz/chz, etc.
These definitions are consistent with those of hyperbolic functions
of a real variable (4.44), and are equivalent to
chz=1 +
2!
+ 4! + 6! +
(xi)
(all z).
shz=
+
3!
li
Again
it
+ 5! +
all
+ ni)
ch(z + 7ra)
_ shz
chz
'
(x)
How
periodicity
sh. (z
th(z
+ 7n)
y
by using
2m
and the
1,
T)
rf
so that th
(z
+ ni) =
th z.
rule
We have
cos
(iz)
i{exp
(i
iz)
+ exp i
(
iz)}
= {exp(z) + exp(z)}
= chz
and
sin (iz)
^ {exp
(i
iz)
exp *
(
iz)}
by
definition
by
definition (x);
by
definition (viii)
by
definition (x).
(viii),
Hence
cos (iz)
= ch z,
sin (iz)
= i sh z.
(xii)
de moivre's theorem
14.68]
559
bolic sines. Since the formulae all hold in particular when the variables
are real,
ch (A
+ B) =
=
cos {iA
+ iB)
sin (iA)
sin (iB)
(i)
sin (x + iy)
and
\mi(x+%y)
=
=
=
=
cos a;
sin (x + iy)
cos (x + ry)
r
where we have multiplied top and bottom by the conjugate of the denominator
make the new denominator real. Using the formulae for products
in order to
into sums,
(ii)
tan (x + iy)
From
ex.
+ sin 22/
+ ^shzy
 = sinzaj
r
cos2a; + ch22/
cos 2a; + cos 2*2/
sin
2a;
(i),
cosz 2
= cos 2 x ch 2 y + sin 2 x sh 2 y
= (lsin 2 a;)ch 2 + sin 2 a;(ch 2 2/l)
= ch 2 y sin2 x
/
Similarly
Thus
x.
The reader should verify that the same inequalities hold for
(iii)
Find
= +
x iy
the equation of the curve described by the point (x, y) when z
sh. z
z is real. Sketch the part of the curve which lies between the
varies so that
lines
sinz
Ti.
8h(x + iy)
= ^ Bia(ix y) =
%
'
DE MOIVRE'S THEOREM
560
Clearly y
satisfies this
equation which,
chx =
smy
if
[14.68
is
The curve is symmetrical about Ox and Oy. When y > (through positive or
negative values), cha; > 1 and so x > (through positive or negative values).
When y > 7T chx > + oo and so \x\ oo; when y > ( n) + a; > oo. The
reader should verify that the branches have gradients + ^/3 at the origin.
,
vk
n
 TT
Fig. 138
Exercise 14(g)
1
By using results
cos z, sh (iz)
isinz.
(xii)
and
Use the
sin(2 1
Excess
+ 22 =
)
>
0,
and
a + ib.
5 sec (x + iy).
4 ch (x + iy).
7 If y
tan(a;
+ *t/) <
th.
(x
+ iy).
cotht/.
(i)
ez ;
(ii)
cosz are
real.
= 3i cos z.
Solve completely cosz = {e(li) + (1 +i)/e}/2 J2.
If ch {x + iy) cos (w + iv) = 1 and cos y cos u 4= 0, prove
tan u th v = th x tan y.
u + iv, prove
x v 2 cosec 2 x = 1
12 If cos (x + iy)
(i)
sec
(ii)
u 2 sech 2 y + v 2 eosech 2 y =
= constant,
= constant
DE moivre's theorem
13
Expand
561
sin 26
14
sin 40
sin 60
41
cos 96
cos 50
cos0H
15
h..
21
1....
\
61
9!
5!
If x
cis a
a;
2/
x+y
and y =
..
* tan
prove that
ciafi,
and
p)
+ y)(xyl)
(x
+(a
(x y)(xy+l)
sina + sin/?
sina sin/>
2 If the real and imaginary parts of (1 +ix) n are equal, n being a positive
integer and x real, prove that x = tan {(4r+ l)7r/4n} where r is any integer
or zero. [Put x
cot 6.]
3 Prove
{(cos a + i sin a)
where A
4
(i)
(ii)
1)
=}=
P P
is
. .
l) an
.,
+ (z l) 2n =
P2n and
is
the
. . .
7
is
a*
If a =
equal to
 3 + 4i sin
if
32(z
+ 1) B
are represented
is
 4 cos
2r7r\
cis (27r/n)
^ ^5
XI I 3 + 4isin
6
r=l \
Deduce that
l) 5
by points on a
(r
33
0, 1, 2, 3, 4).
2m\
4cosI.
]J[ (5
5 /
31 r=1 \
and n,
. .
a multiple of n, and
4= 1.]
COS0
.
sm <p
cos 6
= 1 + 22 tan r *0cosr0.
9 Prove that
^(^
dx n \x 2 +lj
where
a;
= (l) n n!cos(n+l)0sin n + 1 0,
v
'
(vii).]
az
DE moivre's theorem
562
10 If
to
1, o)
1,
=j=
prove
3
XZ
By taking x =
cis 20,
(OX
'
deduce that
3 cot 30
cot
3/9
and hence
and
between
inclusive.
12 If a x + a 2 +
. . .
sec
If
13
lt
2,
+ a 2n =
sec
2it,
. . .
sec
2n
= 1 E 2 + 2 4 ... + 1)"
(
and if A
tegral multiple of n,
4= 1 ,
prove
and 2n which
'
acos
+ 2/&cos0sm0 + &sm
+ 6S +
4)
2h/(a
2n .
in
(0
Express
11
2tt
Xl
satisfy
+ 2grcos0 + 2/sin0+c =
b).
[Put
0,
*.]
15 Factorise
(i)
(ii)
factor of
(iii)
(iv)
a;
2cos(a + 2/77/n) + 1
is
also a factor of
un
for
any
integer
16
un
Deduce that
r.
Solve
[u n
(z
is
+ 1) 8 z 8 =
0,
+ 1) 8 
(z
= ^(2z + 1)
+ 42 + cosec 2
II
r=l
16(cos 16 6  sin 16 6)
r=l
is a regular hexagon inscribed in the circle z = a, A being
17
the point {a, 0). If P, representing the number z, is any point on the circle, write
down the complex numbers represented by the six points obtained by drawing
lines from the origin equal to, parallel to, and in the same sense as the lines
FP, and prove that their product is z 6 d6 Hence prove that
AP, BP,
ABCDEF
6
.
18 Prove
sin 50
16 sin
sin (0 +
\tt)
sin (0 +
7r)
sin (0 +
sin
ftf)
\n sin
sin (0 + $n),
\ir
= + \ *J5.
0*O
19 If
w = u + iv,
<\it,
DE moivre's theorem
563
if a is real, the equation exp z = z + a has no purely imaginary+ iy (y + 0) is a solution, prove that x> 0. [Use sixxyjy < 1.]
21 If u + iv = (zl)exp(*a) + (zl) x exp(a) where z = x+iy and a is
consists
real, find u and v in terms of x, y, a. Prove that the locus of z when v =
20 Prove that
root. If x
of a circle with centre (1,0) and unit radius, and a straight line through the
centre of this circle.
22 If exp(w + *)
integer or zero.
*23 If x+iy
rcistf,
prove u + iv
is
any
(l+x) 2 +y 2
and
(ch + cosw) 2
(1
x) 2 +y 2
= (chvcosw) 2
iv
{1
if
<
0,
then
= 22e
2ra!
sin 2ry.
r=l
two
<A<
1.
<
if
real.
to infinity of
^ cos
2a + ^ cos3 3a +
. . .
...
(a real)
(x,
0real).
5!
ax
29 Expand eax cos bx as a power series in x. [Consider e cis bx
and put a + ib
30
(i)
*(ii)
r cis 0.]
+ .+?L + !_ + ... =
l,w
4= 1.
prove
considering
(a)
(b)
18.]
31 By putting z = x + iy, show that the two differential equations for the
functions x,yot (arising from the dynamics of Foucault's pendulum)
x 2ky + n 2x
are equivalent to
0,
+ 2ikz +
y + 2kx + n%y
2
0,
DE MOIVRE'S THEOREM
564
and solve the latter. Writing z' = ze m prove that x' = A cos [it,
where [i 2 = n 2 + k 2 and A, B are real arbitrary constants.
,
y'
= Bain /it,
32 s, ijr are the intrinsic coordinates of a point on a plane curve, and x, y are
the cartesian coordinates. Writing z = x + iy, prove that
dz
= e**,
.,
ds
[Assume equations
(iii)
K
(Cf.
Ex.
=
d 2z
ds 2
.,
ik e 1 ^.
~y\ds 2 )
\ds 2 )
~d^ds 2 ~dsds 2
'
33 Verify that the formulae for rotation of axes through angle 6 in 15.73 (3)
can be written z = z' e id , where z x + iy and z' = x' + iy' Deduce from this
the formulae for the reverse transformation.
.
565
15
SURVEY OF ELEMENTARY
COORDINATE GEOMETRY
15.1
Oblique axes
15.11
Although
it is
Fig. 139
The
same in form whether or
essential complication.
marked by the
we
sign
{&>}.
has
15.12 Cartesian
We
first
as pole
and Ox
as initial line
(fig.
tt
Hence
tan a
also
r2
tysinw
x + y cos a)
140).
Draw
566
[15.13
Example
If the line
(a; a
rsin#
6 with
Ox
(fig.
141),
then
 a^) + (y 2  yj cos w,
=
(y 2
Fig. 141
From
to triangle
PPQ
X
PX P\ = r =
2
(fig.
(x 2
141),
 Xj) 2 + (y 2  y x 2 + 2(x2  xx
(y2
 yx
cos o).
{to}
triangles
PP
PR
Since
PX Q = x xx and PR =
x=
from which
Similarly,
1\P
find
becomes
l+k
from
PP2
we
this
= PQ = y^yr
P2 R
tyi
l
y2 y
+ ky2
+k
in
fig.
I.
142,
15.15]
multiplied
(2)
k >
P P,
is
by that
of
as in
fig.
which corresponds
multiplied
is
Pv
I)
k,
567
142,
we
so that
fig.
still
PX PQ,
PP^R,
Pil^M
~
PR'
k
i.e.
=
~
from which
Similarly
'
lk
'
lk
Fig. 143
Fig. 142
Summary. In both
(3)
positive numbers.
cases above, k
However, the
=
l
which
ratio
tive
is
I.
when P
external,
Hence
divides
can be written
etc.,
(1) for
we agree that PX P PP2 shall be reckoned posiPX P2 internally and negative when the division is
if
=
l
The
xz
+
+ (k)
result of (2)
+k
viz.
= iyi+ty2
l + k
{to}
is
Since this
'
568
[15.16
that
if
m
+m m
tan 1
1
For
cc
= dx 62 where
,
tan 6 X
tan 6 2
x,
2,
and so
tan 6 X tan d 2
tana =
1
+ tan ^ tan # 2
m2
1
+ mx m 2
Fig. 144
'
Example
From the example in
tan0
15.16
(1)
(r 3 ,
3)
Area of a
One
;
fig.
145.
similarly for
+ (y2  y x
cos ft>
P3
Area of triangle
Then by
r 2 cos d 2
and
(x 2  x )
triangle
vertex at 0.
see
PxP2 is
(2/22/1) sin w
P2 P3
,
be
(r 2 ,# 2 ),
15.12,
= x2 + y 2 cos (o,
= y2 sin &>,
r2 sin d 2
r2 r3 sin(0 3 <9 2 )
i(^2
+ 2/2 cos w
(2/3
2 r3
cos
sin 6 2
sin w )
\ sin &>(a; 2 /3 x3 y2 )
1
2/2
after reduction
smw.
2/3
is
positive or negative
15.16]
(2)
569
ELEMENTARY COORDINATE GEOMETRY
Triangle PX P% PZ We may choose new axes through one of the
.
vertices,
say Px which are parallel to the original axes Ox, Oy. The new
,
coordinates of
and of P, are
are then
(fig.
146)
x%
x% ~ xi>
2/2
xz
= xz~ xi>
Vz
2/2
Vv
2/3
2/i
Fig. 145
Fig. 146
By
area of P1 P2 f3
>
(1),
ft
2/3
is
sinw
x\
XZ~ Xl
#2
2/22/1
2/22/1
""l
Ps
xi
X2
XZ
Vi
2/2
Vz
sinw
VzVi
sm<y
Vz 2/i
1
sin to
c 2 + /x c 3 the middle
;
570
[15.2
Example
The points Px P 2 P3 will be collinear if and only if the area of triangle P1 P 2 P3
and only if
,
is zero, i.e. */
15.2
x2
y2
#3
2/s
0.
straight line
m through (x
lt
Assuming the axes are rectangular, let P(x, y) be any point of the
line. Then the gradient is (y yJKx Xj), and hence
00
i.e.
~~~~
y y1
= m(x x1
).
line, is
The result has been used in previous chapters when finding equations
of tangents and normals to a curve. It can be written
y
where
= yx mx v
15.22 Gradient
There
is
= mx + c,
no simple analogue
form
1)
Rectangular axes.
= mx + c
For when x
locus; also
making intercept
on Oy.
on the
constant and
c
(0, c) lies
is
equal to m.
(2) Oblique axes. If the line makes angle 6 with Ox and intercept c on Oy, its
gradient is tan 6; and if P(x, y) is any point on the line, then by the example in
15.15 its gradient is
(y c) sin &>
.
x + (y c) cos &<
since (0,
c) lies
on the
line.
Hence
a
tana
,
from which
(y
(yc)sinw
c) sin (to 6)
x sin
(i)
15.23]
and
571
sin 6
= r. a\ x + e
sin (o)
6)
= mx + c is given, then m =
and
Hence
this
independent of
is
= mx + c,
but
now m
which
'
(x,y).
(y c)jx,
(0, c)
of gradient
1+m cos
clearly
it
makes ^/intercept
(o
c.
Example
Find an angle between
inclined at angle
the lines
x + cx y
,
m x+c
2,
a>.
sin o)
sin a)
l+m
1+miCosw
cosa)
By
is
15.15,
found to be
+ (m(m+ m m
,
tan 1
m =m
x
+ (* 1
If
Ax + By + C =
and
are parallel
is
B=
but
#= 0,
{to}
==
15.22
If
is
cos a) +
2.
by
sin to
mm
+ m )cosw + m m = 0,
2)
>
2)
line.
= C\A which
y
it
making
on Ox, Oy
intercepts a, b
{to}
y = mx + c.
nave
Hence
= ma + c and
c
c.
is
tv
^ y
 + =1.
a b
i.e.
39
GPMII
572
[15.25
CB
+ l/CB
Example
ABC is
is constant.
Prove that
CA,
AB
Fig. 147
Pv P2
For rectangular
(1)
{to}
axes, if P(x, y) is
line,
then
VVi = ViVz
ry*
*As
(2)
nix + c, then
is
is
/y*
/y
rv*
For
if
line.
is
is
obtained by division.
the equation
y
yi
0.
y%
is
linear in
x and y
(as is clear
the line
yt)
in direction 8
Let
Fig. 148
x x 1 = rco&6 and y y x =
r sin 6.
15.27]
573
Hence the coordinates of any point on the line are expressible in terms
of the parameter r as (xx + r cos 6,yx + r sin 6).
Later
we
no simple analogue
conies.
There
is
form
{to}
More generally, the equations x = at + b, y = ct + d (where a, b, c, d are constants of which a, c are not both zero, and t is a variable parameter) always
represent a straight line. For elimination of t shows that
c(x b)
which
is
linear
a(y d),
15.28 Perpendicular
line
(by 15.23).
form
line
have length p
foot
(essentially positive)
N
P/
\P(*. y)
A
>\
\ x
Fig. 149
p=
rcos(0 a)
= p.
OA = p sec a, OB p cosec a,
x cos a + y sin a
Alternatively, the intercepts
and the
15.3
result follows
from
15.24.
Further results
{co}
(fo/i
2 )/(^
+ ^)) ^ es on
+ kx 2 + b{lyx + ky 2 + c(l + k) =
)
tne une
if
0,
392
574
h
t
i.e. if
ax,+by1x + c
^r
ax 2 + by
2 +c
P P
If
lf
...
(*)
2 lie
internally
[15.31
and
so kjl
opposite signs.
Fig. 151
Fig. 150
Similarly
so ax1
if P\,
(fig.
and
Examples
(i)
P8 P
PPP
X,
u n = ax n + by n + c
we have by
(i)
{n
=
:
l lf
k2
1,2,3)
above:
&2
'
u9
&2 ^3
Multiplying,
1.
ij
(ii)
l3
1%
P Q P%Qi, P3Q3
P P P are conP P P3 Pi, P\Pi
1
X,
0,
Fig. 152
k
with similar equations for
PQ P
2
2,
a?2
&i
Q 3 we have
,
as in ex.
Vz
h ~ xx
(i)
that
etc.
PP
S
k3 h' Writing
3,
15.32]
On
multiplying,
575
Ki
/Cg
K3
Zj
Z2
Z3
d^{x% x f + {y y f.
Since P P is perpendicular to
2
(ii)
ax + by + c
0,
we have
= 1,
b{x 2 x x )a{y 2 y x )
i.e.
(hi)
0.
Fig. 153
Because
on ax + by + c
2 lies
0,
ax 2 + by 2 + c
and
this
(iii)
+ b(y 2 
and
(iv),
d 2 {a 2 + b*)
a
a
(iii))
in the
form
=  {ax x + by x + c).
x)
using
0;
(iv)
(ii):
{axx + by x + cf.
c
~ afti +2fyi +
2
V( + &
1
'
4=
is
line,
chosen to
then a x x + b x y + c
the
will
if
be positive when
lies
on the
origin side of
15.31.
Exercise 15(a)f
Prove that the point which divides in the ratio 2 1 the median Px Qx of
the triangle P1 P 2 P3 has coordinates {^{x 1 + x 2 + x 3 ), 1(2/1 + 2/2 + 2/3)) From the
symmetry of this result deduce that the medians of a triangle are concurrent.
2 Show that the equation of the line through (h, k) which is perpendicular
1
to
Ox is
x + y cos (0
= h + kcoso).
is
576
[15.4
is the point (h, k) referred to axes at angle (0. Prove that the line joining
3
the feet of the perpendiculars from P onto the axes is sin (o J(h 2 + k 2 + 2hk cos (o).
4
P,
that
BE CF = AB
AB,
prove that
From a
Perpendiculars
P if
MN = 2c [use no.
3].
PN
point
on one side of a fixed triangle, perpendiculars PM,
to the other sides. As P varies on this side, find the locus of the mid
are drawn
point of MN.
line
AD
If a straight line passes through a fixed point, find the locus of the midpoint of the part intercepted between two fixed intersecting lines.
11
*12 A line AB of constant length c slides between two given oblique lines
which meet at 0. Find the locus of the orthocentre of triangle OAB.
*13 Using similar triangles, give a proof of Menelaus's theorem (15.31, ex. (i))
by the methods of 'pure geometry.
*14 Using areas, prove Ceva's theorem (15.31, ex. (ii)) by 'pure' methods.
cut Ox, Oy at
15 Obtain the result of 15.32 as follows. Let ax + by + c =
L, M. Equate \p.LM to the area of triangle P X LM. Verify the result when
the line is parallel to Ox or Oy. (This method could be used when the axes are
'
oblique.)
15.4
Concurrence of straight
lines
x, y,
lines
{w}
L + kL' =
and U = 0.
satisfies both L =
= 0, U = intersect, the locus L + kL' = passes
through their common point.
is then
If k is constant, L + kU is linear in x, y, and so L + kU =
the equation of some straight fine through the meet of L = 0, L' = 0.
are parallel and k is constant, then L + kL' = is a
If L = 0, L' =
is satisfied
Hence
if
the lines L
is
clear
by expressing each
line in
'gradient form').
k. For example, the line
be required to pass through a given point, or to be parallel to
a given line.
may
577
ELEMENTARY COORDINATE GEOMETRY
Conversely, if the linear equation L =
can be written in the form
L l f lcL 2 = 0, where h is independent of x, y and L L 2 are linear with
constant coefficients, then L =
passes through a fixed point, viz.
the meet of the lines L x = 0, L 2 = 0.
15.42]
x,
Example
The midpoints of the diagonals of a complete quadrilateral are collinear.
If ABGD is any quadrilateral, let AD, BC be produced to meet at E; and
BA, CD to meet at O. The resulting figure is called a complete quadrilateral
Choose OAB for Ox and ODC for Oy. Call 4(2o, 0), 5(26, 0), (7(0, 2c), D(0, 2d),
and let
be the mid point of OE. Through
draw lines parallel to EA, EB;
Fig. 154
by the
V
a
Thus
ikf lies
on the
also
at the
be
lines
dx + ay
and therefore
Oy
ad
and
ca;
+ by =
6c,
on
(dx + ay ad)
(cx + by be) =
0,
i.e.
AC are
and these
clearly
lie
on
this last
line.
Let the
lines
{co}
be
a1 x + b 1 y + c1
Q,
a2 x + b 2 y + c2
0,
a3 x + b3 y + c3
0.
578
If
two of these
point (obtained
intersect,
by
solving for x, y)
a 2 b z a z b 2
This
lies
on the
'
[15.42
common
is
(i)
a 2 b z a z bj
first line if
& 2 C 3 & 3 C
i.e. if
>
61
A=
If no
^2
^3
0.
two of the lines intersect, then they are all parallel; this means
Cx = C% = Cz 0, and so
A=
Hence
CiCi + CgOa
+ CgOg =
0.
Conversely, if A
f b 2 cz
0,
the point
b z c 2 \
fc 2 a z
(i)
lies
c z a 2 \
on the
+c
A=
0.
first line
because then
a 2 b z a z b 2
0;
0.
az b 1 a1 b z
in
and
a 2 b z az b 2
0,
Hence when A =
the lines
is
(6), (c).
Exercise 15(6)
the origin.
5x2y =
3
4,
8xly + 5 =
line of gradient
f which
is
concurrent with
0.
3x ly =
2,
4x + 5y
9x+ 10y+ 15 = 0.
lines which make numerically
lines
to the line
0,
equal intercepts
2x y + 3 = 0.
15.5]
5
(i)
579
all
2x + 3yl
+ k(3xy +
4:)
3yl)c(3xy + 4)
=
=
0,
0.
3x 5y +
to the meet of
37
= 0, 2x3y+l =
= 0, 5x2y8 =
0.
=
=
2xyl  0, 1x5y+13 = 0.
xBy
+ 2 = 0, 3x + 8y6 = 0.
0,
(p + l)x + (pl)y+p = 0, (ql)x + (q+l)y + q = 0, x = y (p * g).
rc^ + aJ* = 0,xt 2 y + at* = 0, (^yi + flM^ = o(i + ) (*i * < 2 )are concurrent, find
If the lines x 2y + 3 = 0, 2x 3y = 0, ax + y+ 1 =
3x2y +
<t
8 5x + 2y4:
9
10
11
the value of
0,
a.
b 3 , 3cx + 2y
are concurrent,
13
3bx + 2y
c3
= a3
x sin 3a + y sin a
[If (h, k) is the
a,
a:
sin
3^ + y sin/?
<x,fi,y
a,
xsin 3y + ysm.y
15
0.
is
are
If a1 x + b 1 y + c 1 = 0, a 2 x + b 2 y + c 2 = 0, a 3 x + b 3 y + c3 =
and numbers I, m, n none of which is zero can be found such that
l{a 1 x
prove the
A=
0,
If
*(ii)
a,
(i)
lines,
a.
4h sin3
i.e.
distinct
(A)
0,
I,
m, n not
all
a^ + a^m + a^n =
and hence that
15.5
(A) holds.
0,
6 1 Z + 6 2 TO + &3 W
Deduce the
0,
converse of
c1 l
+ c 2 m + c3 n =
(i).
Linepairs
the pair
of lines y
so repre
580
[15.52
(iii)
and
x
0,
x+y
0,
aj
+ 2?/ =
0.
(a^ + b^ + Cj) (a z x + b 2 y + c 2 ) =
represents the pair of lines a 1 x + b 1 y + c 1
if (x
satisfies either
y )
and conversely,
one;
a i xo + bxVo + c x
(1)
if (x
or a 2 x
The
locus ax2
If b
15.52
satisfies
0,
a % x + \y + c 2
0.
For
+ b2 y + c2 =
or both.
+ 2hxy + by 2 =
the equation
0,
x(ax + 2hy)
is
0,
representing a line
pair.
If b
4= 0,
When (h/b) 2 > a\b, i.e. h 2 > ab, this quadratic has distinct roots
yjx = m x ra 2 and then y = m x x or y = m 2 x. Hence if h 2 > ab, the
,
When h % =
ab, the
equation
is
(y/x + h/b) 2
We
h2 <
0.
disposed
of at the start.
The work of
15.5 so far
is
we
(2)
have equations y
has roots
and
so
m m
x,
a.
x x,
2 x,
0.
Hence
(m 1 m 2 ) 2
= (m 1 + m 2 2 4m 1 m 2 = (h 2 ab).
)
15.53]
581
m,ma
teng=
2V(A*qfe)
b{l+ajb)
2<](h*ab)
"
provided a + b
If a + b
'
a+b
4= 0.
then
0,
Wjmg = 1 and
and
conversely.
(3) Bisectors of the angles
By expressing equality
m1 x = 0,y mi x = 0,
y) to the lines y
ym
V(!+i)
Hence the two
bisectors
ym x
V( 1 + W D*
2
have equation
{ym^f {ymz x) 2 =
l+m
1+raf
i.e.
(1
i.e.
If
4=
2,
by
satisfies
as of course
(2),
(1
*/ 2 )
it
in x,y.
0.
= 2(lm 1 m a )a;2/,
=
(ab)xy.
x2
xy
6=0.
1
linepair 5 =
is
'
can be written
The general
This
15.53
m m
h(x 2 y*)
This equation
of
2)
mjj) (z 2
(i),
tions. It
0,
this reduces to
K+
i.e.
0.
(i)
BA
582
[15.53
then
(i)
we
so
represents a linepair.
degree terms in
also
(ii)
(ii),
= A1A2
2h
= A X B2 +
= B X B2
X,
have
ax 2 + 2hxy + by 2
Therefore
*/
(i)
ax 2 + 2hxy + by 2
(iii)
This fact can be used to calculate the angle between the lines
which
is
When h 2 =
Cx C2
be equal to
(2)
If
Ax
(i)
=j=
(i),
(2).
to represent
a linepair.
(ii); and if
means that (i),
and
A% ^
(i.e. if
regarded as a quadratic in
Now
found in 15.52
A X :A 2 Bx
(i)
(iii),
(i),
x,
Bf
~A
#= 0,
since a
= AXA2
),
y ~T'
x=z
written as a quadratic in
Ba
Co
~T y ~T'
(lv)
x, is
this
0,
(v)
Hence (i) can be solved in the form (iv) if and only if the quadratic
y under the squareroot sign is a perfect square, i.e. if
in
(h 2
is
(which holds
gh af=
is
so if and only if
even when h ab =
2
0, for
0,
also), i.e. if
a(abc + 2fgh  af 2  bg 2  ch 2 )
j
0.
15.53]
Since
we
are assuming a
the condition
4= 0,
is
but b
quadratic in
If a
0, b
y,
4= 0,
then equation
(i)
but h
equation
4= 0,
v+
which,
j.
(vi)
0.
(vi) is
obtained.
(i) is
x+ l y + lL =
0,
if factorisable, is
H) H) f=
so that
2fghch*
i.e.
0,
becomes when a = b = 0.
These are the only cases that need be considered, since
and this
is
what condition
(vi)
a
then
(i)
(ii)
(i)
(vi).
be written
A=
(vii)
0.
if
By ex.
583
= b=l,
c=f = g =
or of equation
(i).
(i)
0,
whereas x 2 + y 2 has no linear factors. (A sufficient condition is indicated in Ex. 1 (/), no. 19. Other ways of obtaining the necessary
appear in Ex. 9 (/), no. 25 (ii); (3) below; and 15.73,
condition A =
ex.
(ii).)
584
in 10.21, ex.
(ii)).
(i),
[15.54
meet P, then
only one
Now when
root
is (see
(v)
1.31 (&))
so the coordinates of
P satisfy
Fig. 155
ax + hy + g
Similarly, writing
(i)
0.
(viii)
as a quadratic in y, viz.
P we have
at
Equations
see ex.
hx+f
hx + by+f
i.e.
(i)
Remark
0,
(ix)
0.
to determine
Since
(i)
Equations
0.
(x)
intersecting linepair.
Compare
0,
g%+fy+c =
of 11.43 gives
(2)
above, where
and Corollary
(vii)
1(6)
to represent an
was shown to be
(ii).
= 1 and
the locus j =
15.54]
is satisfied
(Ix
585
x, y:
0.
and of the locus, and hence it represents some other locus through these
homogeneous in x, y, it represents two
must be the equation of the required line
through O. Hence
lines
it
pair.
Exercise 15(c)
[Rectangular axes.]
lines
and
+ my =
Ix
 ab)/(am z  2hlm + bl 2 ).
Prove that the product of the lengths of the perpendiculars from
is <J(h 2
(xx , yx )
to the lines
ax 2 + 2hxy + by 2
is
lines
5 Find (i) the angle between, (ii) the equation of the bisectors of the angles
between, the lines 5x 2 + 2xy 4y 2 = 0.
ax 2 + 2hxy + by 2 =
anglebisectors, prove that h(a' b')
6 If the
lines
0,
h'(a b).
7 Show that any pair of lines which has the same angle bisectors as the pair
ax 2 + 2hxy + by 2 = can be written ax 2 + 2hxy + by 2 + \(x 2 + y 2 ) = 0. [Use no. 6.]
Find the equation of the pair of lines, one of which passes through (p, q),
and whose angle bisectors are x 2 y 2 0.
8
2x 2 + lxy + 3y 2 4xly + 2 =
15x 2 + xy  2y 2 + 3x  y = 0;
lines;
0;
xy3x + 2y6 = 0;
x 2 + 4xy2y 2 + 6x12yl5 = 0.
9 Write down the equations of those line pairs through the origin which are
(iii)
*(iv)
8.
[Use no.
4.]
to the
meet of x + 2y
3 with the
0, s'
in no. 11 possess
line in
common,
find its
586
The
15.6
circle
and radius
{x
which
the coefficients of x 2
Conversely,
and y 2 are
circle
with centre
_ hf + {y _ kf =
(j)
form
x 2 + y 2 + 2gx + 2fy + c
in
[15.6
equal,
(ii)
0,
and
there is
no xyterm.
<7)
(ii)
can be written
+ (2/+/) 2 =
+/ 2 c.
9'
ff>
/); and
if
+f 2 c <
0,
locus.
Example*
Referred to oblique axes at angle o), the circle with centre C(h, k) and radius r
has equation
(xh)* + (yk) 2 + 2(xh) (yk) cosw = r a ,
which
is
of the form
15.62 Circle
on diameter
0.
P P2
l
PX P = UzMl
and
circle;
gradient of
 xx
circle, is
no. 15.)
*^
Zz*
(x
i.e.
(x
PX P P2 P,
P2 P = U^
hence
then
is satisfied
_ lf
 x2 + (y yx
)
""""
(y
 y% =
)
0.
(iii)
15.63]
15.63 Tangent at
Let
Px (xx
yx )
the tangent at
circle
(ii)
is
xxx + yy x + gx +fy
i.e.
Since
Then since
and CPX has
in 15.61.
CPX
i.e.
587
is
x\ + y\ + gxx +fy x
0,
0,
Px can be written
= ~ 9xx ~fVi  c
i.e.
>
(iv)
0.
circle:
xx + yy + g(x+x)+f(y + y) + c
0;
and
is
the tangent at
Other applications
will
is
the equation of
Example
Find the condition for the line Ix + my = nto touch the circle (ii).
The line will be a tangent if and only if the perpendicular from the centre
g, f) to it is equal to the radius *J(g 2 +/ 2 c)
Ig mfn
i.e.,
on squaring,
15.64
If
(lg
+ mf+n) 2 =
,,
+/ 2 c) (Z a + m2 ).
,2
(fl
x is
P2 P3
,
Px if
P2 P3 is called the
drawn from
chord.
40
GPM II
588
Since
P2 P
at
centre
is
nas equation
xx z + yy 2
= a 2 whose
x2 + y2
circle
[15.64
= a2
#1^2 + 2/12/2
= x % y = y 2 satisfy the
xx x + yx y = a 2
equation
i.e.
that
2
.
Pi
Fig. 156
Similarly,
chord
P2 P3
2/2/i
= a2
circle
+ y + 2gx + 2fy + c =
2
Px has equation
= 0,
(x 1 + g)x + {y x +/) y + (gx x +fy x + c) = 0.
xx i + 2/2/i + 9(x + x i) +f(y + 2/i) + c
i.e.
when Px lies
outside,
it
Px
Similar
Examples
(i)
Find
(fig.
to
x 2 + y 2 + 2gx + 2fy + c
gx+fy + c =
0.
0.
15.65]
As
589
c(x 2 + y 2 )
which reduces to
(gx+fy) 2
It is satisfied by the points common to the circle and line, i.e. by the points
of contact of the tangents from O. It is homogeneous of degree 2 in x, y, and
therefore represents a linepair. Hence it represents the pair of tangents from 0.
Fig. 158
Fig. 157
(ii)
will divide it in
some ratio k
I,
and hence
/ lx x
i.e.
(x 2
lie
on the
+ kx 2
+k
will
ly x
'
have coordinates
+ ky 2 \
+ k )'
circle if
+ kx 2\ 2
l + k
)
+ ky 2 \ 2 _
l+ k
J
a 2 )kl + (x\ + y\ a 2 )l i =
/ lx x
( ly!
+ y\ a 2 )k 2 + 2{x x x i + y 1 y 2
0.
v)
This quadratic gives two values for the ratio k I, which correspond to the points
A, B where the line PX P2 cuts tne circle; the two values are PX A :AP2 and
P1 P:PP 2 (fig. 158). Equation (v) is known as JoachimsthaVs ratio quadratic
for the circle x 2 + y % = a 2
If P a lies on either tangent from Pt then the points A, B will coincide, and
hence (v) will have equal roots k I, so that
:
(x 2
+ y a
2
2
)
(x
2 (flJ 2
+y  a
2
2
)
Vz)
2
.
(xx x
+ yyx  a 2
2
)
x,
(i).
circle
x2 + y 2
402
[15.65
ELEMENTARY COORDINATE GEOMETRY
Let the tangents at the extremities A, B of the chord AB through P meet
at P Then by 15.64, AB has equation
590
2.
xx 2 + yy 2
Since
x lies
on
= a2
this line,
x x x 2 + yx y 2
lies
on the
line
a2 5
xx x+yx y
= a2
Fig. 159
If
the
lies inside
Fig. 160
which
(2)
is
lies
P P
Show that P
x,
divide
line,
which
lies
is
on
lies
xx x + yyx
the line
AP
PP
k'
the complete
2 is
Since by hypothesis
1 :P 1 B
so that A, B divide the line X
say
chosen so that
ally), f
the locus of
circle,
AP
2
.
Fig. 162
Fig. 161
Either point A,
B dividing P P
X
/ lx x
\
f
P x and Pj
in the ratio k
+ kx 2\
+ k )'
are said to
+ ky 2\
+ k )'
divide AB harmonically.
lyi
15.66]
and
lies
are
k'
on x
+y =
2
2
;
hence equation
xx x2 + y x y 2 a 2
Hence
Px
If
lies inside
the
which
lies
within the
line
0.
the whole
the locus of
2 is
which
line,
lies
com
2 is
circle.
xx x + yyx
a2
is
x is
lies outside,
(3) Definitions.
The whole
the locus of
circle,
holds,
(ii)
sum is zero,
V. Since their
591
= a 2 wo x 2 + y 2 = a 2
xx x + yy x
(2).
wo x 2 + y2 = a 2
Remarks
(a) If
from
(/?)
(y)
lies
outside the circle, the polar coincides with the chord of contact
x.
of
common point.
their tangents at a
(It is
one
circle is
Condition for
x 2 +y 2 + 2gx + 2fy + c
0,
Fig. 163
angled at P, and so
i.e.
(9
Conversely,
(9
+P c) + (g'
2gg' + 2ff'
when this
condition
g
is
cp2 +
 g'Y + (ff) 2 =
i.e.
CPC
The general
+9' 2
+f'
 C),
= c + c'.
is satisfied,
2
+f +f'
definition
then by adding
was given
in 5.72.
right
592
is
The condition
is
by the converse
obtained; hence
C'PC is
[15.66
rightangled at P.
and
gonality.
Example
Prove that in general
x 2 + y 2 + 2gr x + 2fr y + cr
we require to find g, f,
c for
(r
1,2,3),
which
2^i + 2//ic =
=
=
c
+
2^3 2//3
299i + 2fA c
and
cx ,
c 2
c3 .
If the determinant
A=
91
92
fi
9z
fz
nonzero, then this system of three equations for the three unknowns g, f, c
can be solved uniquely by Cramer's rule (11.41). The condition A 4= means
that the three centres are not collinear (15.16, ex.) and that no two centres
coincide; this is the situation 'in general'.
is
Exercise 15(d)
Write down the equation of the
Prove that the tangent at O is px + qy
1
circle
0.
is (p, q).
'
[CMCP
4
(i)
1 .]
(ii)
and
If Ix + my
Ix
7 (i) Find the condition for the chord of contact of tangents from Px to
x 2 + y 2 = a 2 to subtend a rightangle at O. (ii) What is the locus of the meet of
perpendicular tangents to x 2 + y 2 = a 2 ?
8 (i) Show geometrically that, when they exist, the direct common tangents
of two circles divide the line of centres externally in the ratio of the radii. What
the corresponding result for the transverse common tangents?
(ii) Use (i) to find the equations of the direct and transverse common
tangents to (x 15) 2 + y 2 = 64, (x 2) 2 + y 2 = 9. [If a tangent meets the line of
is
centres at
perpendicular to this line from either centre
is
593
that the
equal to the corresponding
radius.]
common tangents to
a x f + </ 2 = r\, (x  a 2 2 + y* =
(x
r\
have equation
r2 )x (a^ a 2 rx )} 2 = {(%
{("].
(rx
2
2)
2
2)
}y
2
.
+ 2gx + 2fy + c =
x*
0,
circle
on diameter
AB is
(gg')US')^+y 2 = (gf'rfHigrtxVDy}*1 1 A variable circle passes through the meet O of two given lines, and makes
intercepts OP, OQ such that m.OP + n.OQ = 1. Prove that this circle passes
)
x2 + y* + 2gx + 2fy + c
Prove that
its
centre
lies
on the
0.
line
fc
+ Fc.
and
x2 + y2
15
5,
x 2 + y 2 + 6x+l
x* + y* + 2gx + 2fy + c
(supposed to intersect)
is
COB 6
given
x i + y 2 <ix4ty + l
and
0.
circles
s'
0,
x 2 + y 2 + 2g'x + 2f y + c'
by
2gg' +
2V(fl'
+/
2ff'cc'
c)V(9''
+/ ,2 c')'
is also the
are the circles in no. 15, prove that s + As' =
16 If 8 = 0, ' =
equation of a circle for any constant A except A = 1 What is the interpretation
when A = 1? If s, *' intersect, explain why s + As' = passes through their
common points for all A. (As A varies we obtain a system or family of circles.)
.
to a 2 + y % + 2gx + 2fy + c
s s'
s'
0,
0.
Show that the locus of P is the whole line if s, s' do not intersect. When
is their common chord and that the
they do intersect, show that s s' =
(ii)
locus of
594
[15.7
20 Find the equation of the circle which passes through ( 3, 2) and the
points of intersection of the circles 3a; 2 + 3y 2 + 2x'7y6 = 0, x 2 + y 2 + y 2 = 0.
[Use no. 16.]
21 Find the equation of the circle whose diameter is the common chord of
the circles x 2 + y 2 2x + 2y + 3 = 0, 5x* + 5y 2 x + ly12 = 0.
Conies
15.7
15.71 Definitions
The
locus of a point
its
is
constant
is
called a conic.
The point 8 is called the focus, the line d the directrix, and the conby e) the eccentricity of the conic. According as e =j 1,
the conic is called an ellipse, parabola, or hyperbola; e is essentially
stant (denoted
positive.
If
is
SP = e.PM.
The name
'conic'
is
an abbreviation
the ellipse, parabola and hyperbola were obtained as the sections made
A plane parallel to
the axis of the cone will give two intersecting lines (the generators in
that plane); one touching the cone along a generator will give a single
line;
while a plane through the vertex but not cutting the cone else
where
a
will give
linepair,
It can
bolic, or
and
a single
line,
and a
circle,
elliptic,
para
8P = e.PM;
i.e.
We shall not
'
do so in
this book.
15.72]
15.72
the definition
SP =
2
(x
S
e
PM
and the
is (p, q)
2
is
fixed fine d
2
p) 2 + (y q) =
e2
(Ix
may
+ my + n =
0,
+ my + n) 2
2
2
l +m
we suppose 8
is Ix
becomes
When simplified,
particular, if
595
lies
on
d,
x, y.
we
dicular at
8P 2 =
e2
.PM 2
x2 + y 2
According as
is
expressed by
e2x 2 , i.e.
e
is
$ 1, this
(e
l)x 2
Fig. 164
.
as a degenerate ellipse
(e
<
1 for
(e
for both),
>
for both),
our remarks in 15.71 about particular 'conic sections'. The classification of a pair of parallel lines will be mentioned in Ex. 16 (e), no. 26 (ii);
for the circle, see 17.17.
by showing that
We
forms.
this equation
shall eventually
degree represents a conic (including the degenerate cases just mentioned), a parallel linepair,
15.73
(1)
we
circle,
or nothing.
(x, y)
of axes
596
changing the
(a)
(a translation
[15.73
by
axes unaltered
of axes)
(6)
(a rotation
of axes);
both together.
Remark. When a problem involves equations of two or more
(c)
in general
way;
it is
loci,
(ii).
O'
Fig. 165
Change of origin. If the new origin 0' is taken at the point (h, k),
then the point P whose coordinates were (x, y) now has coordinates
(2)
*',*,')
given by
Thus,
to
^ x_
h}
to the
y>
= yk
point
(h, k)
we
+ h for x
Oy isf(x, y) =
substitute x'
in 15.16(2).
Examples
(i)
= a\
x* + y* + 2x1 x + 2y1 y + (xl + yla 2 =
(x+x1 )* + (y + y 1
i.e.
0.
+ yla*)(x* +
y*)
= fax + ytf)*.
(i))
15.73]
597
+ y\ a 2
(x 2 + y\  a 2
(x 2
i.e.
 x x 2 + (y yx 2 } = {x x (x  x x + yx (y  y x )}\
{(x 2 + y 2  a 2 + (a; 2 + y\  a 2  2(xx x + yy x  o 2 )}
= {(xxx + yyx  a 2  (x 2 + y\ a 2 )} 2
{(x
{x\
i.e.
+ y\ a
2
)
(x
+ y a =
2
(xx x + yy x  a 2 ) 2
(ii)
Suppose the
becomes
lines
ax
i.e.
origin to
x,
0.
the equation
0,
0.
Since the new origin is the meet of the lines, this equation must involve only
terms in x 2 xy and y 2 (15.52 (1)). Hence we must have
,
axx + hyx + g
ax
and
= 0,
+ 2gx x + 2fyx + c = 0.
hx1 + by1 +f
0,
+ 2hx 1 yx + by
is
equivalent to gx x +fyx + c
15.53
(3),
the third
0.
Fig. 166
Rotation of axes through angle 6. Let the point P, whose coordinates wo Ox, Oy are (x,y), have coordinates (x',y') referred to
(3)
the
new axes
Ox', Oy',
where
xOx'
HOP =
r,
r cos (6
x'OP =
and
<f>.
OPN,
sin $
598
since
Thus,
for
r sin (0
to rotate the
<j>)
r sin
x' sin
cos <f>
f(x cos
rsin^. Similarly,
+ r cos
sin <j>
+ y' cos 0.
we substitute x' cos y' sin
= becomes
+ y' cos
x'
[15.74
0.
above equations;
or
by considering
x'
r cob {{d
+ <!>) 6} =
y'
= rsm{(d + <f>)6} =
....
scheme
y'
x'
cos 6
sin 6
sin 6
cos 6
n
7
n
^
7
f(x cos a y sin a + h, x sin a + y cos + k) = 0.
(4)
It follows that
a change of axes leaves the degree of any polynomial equation unaltered. For, instead of the linear function x, we have
and products of
linear
to standard forms
==
the equation
(i)
becomes
a(x cos
0,
(ii)
15.74]
in
which the
coefficient of
xy
is
4= 0,
this will
be zero when
tan 20
if b
a =
\ir
sin
satisfies
*
2ft
0, it will
value between +
.....
(m)
t\
0; i.e.
7r.
If A
be written
(a)
Ax * + By z + 20x + 2Fy + C =
cases
2?
X+
[
(iv)
0,
0,
=}=
a)
this
(V)
(v)
becomes
(vi)
a2
according as
can
pi
Q%
(iv)
+B y + B) =A + B[
(
4= 0,
is so known.
(ii)
now arise.
and
4=
the
Giving
The following
599
A /A,
/?
a2
'
I?/A are
/?
'
a2
negative.
// A
0, (vi)
a 2x2  fity2 =
0,
a 2x 2 + [S*y* =
according as A,
(b)
=}=
(_C_
\
2G
F*_
2BG'
Br
2
%
a
t The case cc /a + y \fl* = 1 can be brought to the second form
rotating the axes through angle $n.
shown by
600
By + 20x =
2
reduces to
(vii)
[15.74
0,
r = ~fi x
i.e.
(c)
If A
and B
={=
0,
equation
according as
F = BC.
IfB =
0, J. 4= 0,
(d)
0,
forming equation
(iv)
We
becomes
(iv)
By 2 + 2Fy + G =
which represents two
....
(vm)
cannot have
F+
0,
or */ j?
=^=
0,
A*
0,
then by transtt,
we obtain
(6), (c).
A =B=
degree,
general equation
(i)
^+^ = 1,
2/
ccx*+]3y*
0,
= aV(a 2 /?).
an intersecting
linepair,
a repeated
or a single point;
line,
and the
The
first
three
loci,
which
will
'
and
17.13.
Exercise 15(e)
1
By
a change of
1,3),
(4,
2) become
find a, ^.
when
the origin
601
6 Change the origin to (1,0) and then rotate the axes through the acute
angle whose tangent is , for the equation
43a; 2
0.
7 Show that the centroid of a lamina, as defined in 7.81, does not depend on
[Wo the new axes the centroid is given by
and
x'dA
Thus
= xcoad + yaind + h,
the point
(x, y)
referred to the
new axes.]
If
the points
M3 + M1 + M2 =
{at\,at\),
{at\,at\),
Conversely,
0*
are
(at\,at\)
collinear,
prove that
2 Find the condition that the four distinct points (ktr ,k/tr ), r = 1,2,3,4,
be concyclic. Show also that no three of these points can be collinear.
3 Show that a necessary condition for concurrence of the lines
shall
is equilateral.
If the parallelogram
OG and OA J_ BC]
represents three
straight lines equally inclined to one another. [In polar coordinates the equation
= tan 30. Writing = tan 3a, the three lines are 6 = a + \m, r 0, 1, 2.]
is
)
centroid; then
11
HxA =
4x3y + 2a =
12
2y^.]
whose
0.
x2 + y2 + 2yS =
line
xy
at the origin.
sides are
602
Show
13
{x
a) + y
r2
,
and
pair of parallel tangents is drawn to a given circle, and another pair perpendicular to these is drawn to a second circle of equal radius. Prove that each
diagonal of the square formed by the four tangents passes through a fixed point.
14 All members of a family of circles pass through two given points. Prove
that the common chords of these circles and a fixed circle not belonging to the
family are concurrent.
15 Write
parallel to
down
(i)
Oy;
and
which
is
(ii)
by these
line pairs,
and
x 2 + y 2 + 2g1 x + 2f1 y + c1
have centres A,
x 2 + y 2 + 2g2 x + 2f2 y + c 2
0,
2(x
+ y + 2( 9 1 + g 2
2
x + 2(/x +/2 ) y + (c x + c 2 )
on diameter
= r\\AB 2 where r
is
D is
0.
G,
0,
first circle.
17 Find the equation of the line pair joining O to the meets of the lines
2
4a;  15xy 4ty 2 + 39x + 65y  169 =
and x + 2y = 5. Show that the quadrilateral having the first pair and also the second pair as adjacent sides is cyclic,
and find the equation of its circumcircle.
18
to the meets A,
= 1 has equation
(a I 2 x 2 2lmxy + (6 m 2
B of the line Ix + my
If
0.
and the
ax 2 + 2hxy + by 2
If the lines ax
axes,
+ 2hxy + by =
2
1.
to the meets of Ix + my = 1
find the coordinates of the
1 and the lines
0.
varies
represents a locus passing through the vertices of the triangle formed by the
lines x 2 + 2hxy + y 2 = 0, x + y + 1 = 0. Deduce the equation of the circumcircle
of this triangle, and show that this circle is orthogonal to the circumcircle of
by the
lines
ax 2 + 2kxy + ay 2
0,
x y+1
0.
603
16
THE PARABOLA
The locus y2 = 4ax
16. 1
a) 2 + y 2 =
(x
(x
+ a) 2
PM
2
where 8 is the
any point P{x, y) on the locus is such that SP 2 =
line x + a = 0.
perpendicular
from
to
the
P
point (a, 0) and
is the
2 =
4ax is a parabola with
Hence by the definition in 15.71, the locus y
focus (a, 0) and directrix x + a = 0.
,
PM
It
is
the equation
is
y by y; Ox
unaltered
is
by replacing
parabola, and
loss
of
suppose a >
all
0;
a;
The parabola
= and hence y =
meets Oy where x
it
is
therefore touches
Oy at 0, and Oy is
Fig. 167
is
at right
(a,
2a), so
that
its
length
x=
is
a,
and
4a.
Remark. The equation y 2 = 4ax also shows that the square of the
distance of P(x, y) from Ox is proportional to its distance from the perpendicular line Oy. The locus of a point satisfying this condition is
therefore a parabola.
(x, y)
ofy 2
2x
so that y
41
2at
and x =
(0, 0),
at 2 ; (0, 0) is also
given thus
when
GFMII
0.
604
one value of
THE PARABOLA
= kax has coordinates
Also each value of
t.
[16.12
of the form
(at
2
,
2at),
The equations
at 2
2at
(i)
:2t
....
1.
(n)
the parameter of P.
is
9
2
x:y:a =
of the curve;
cf.
the
Remark in
16.21.
Examples
Find the condition for the chord joining the points tlf t2 to be a focal chord
to pass through the focus).
Any line through the focus (a, 0) has an equation of the form
(i)
(i.e.
l(x
for
l(at 2
a) + my =
0.
which
a)+m.2at =
0.
pt 2 + 2qtp
tx ,
lie
t2
on the
line
px + qy ap =
0,
and
this
A circlef
which
satisfy J
0,
0.
= 1,
0,
x* + y* + 2gx + 2fy + c
t
if
Since the
(a, 0).
<2
Since this equation is quartic in. t, a circle and parabola can intersect in at
most four points. In real algebra a quartic equation has four, two, or no roots
(some or all of which may coincide); hence (ignoring possible coincidences) the
number of intersections is four, two, or none.
we
are already taking the equation of the parabola in the simple standard
are not entitled to choose axes so that the equation of the circle
also takes a simple form (such as x 2 +y* = r 2 ). Simplification of the equation of the
circle would complicate that of the parabola.
J When discussing the intersections of two loci it is usually most convenient to
employ the cartesian equation of one and the parametric equations of the other,
as here.
f Since
form y 2
= 4<tx, we
THE PARABOLA
16.2]
605
WehaV6
quartic;
and
h + ** + ** + h =
t x , t 2 , ta , t
since the
v Then these
term in
is
absent,
0.
This necessary condition for concyclic points on the parabola is also sufficient.
In a general way this can be seen because three points of the curve determine
a unique circle, and one condition is required for this circle to pass through a
fourth point of the curve. In detail, given the four numbers tx t z t 3 tif where
Ht x = 0, consider the circle through tlt t2 t3 it cuts the curve again at a point
t' (since a quartic with three distinct roots must have a fourth), and so by the
t
above
t
t' = 0.
,
'
Hence
t1
t'
<4 ,
and so tv
t it t
z,
tt
+ t2 + 3 +
Exercise 16(a)
'
is
is
If
is
4ax.
t2 .
2 If the point
3
equal to $PQ.
4 P,Q,R are points on the parabola such that PQ passes through the focus
and PR is perpendicular to the axis. When P varies on the parabola, prove that
the midpoint of QR lies on the parabola y 2 = 2a(x + a).
5 The chord PQ subtends a rightangle at the vertex O. Prove that the midpoint of PQ lies on the parabola y 2 = 2a(x 4a).
6 A circle is drawn on a focal chord as diameter, and cuts the axes at
(xlt 0), (a> 2 0), (0, yx ), (0, y 2 ). Prove that xx x 2  y x y 2 =  3a2
cuts the parabola in at most two
7 (i) Prove that the line Ix + my + n =
.
when
points
(ii)
4= 0.
If there are
(iii)
If there
is
only one
+ t2 =
2m/l,
tx
and
2,
t1 t 2
prove that
n/al.
In,
and conversely.
circle cuts
PM
10
circle
[If
AB,
CD
A variable circle passes through a given point A and touches a given line
its
centre
is
directrix.
16.2
16.21
Chord
I.
P P2
t
THE PARABOLA
606
since y\
4axx and y\
yVi =
This equation
(x2
2 ).
is
To make it
x,).
Ix
(1)
of fractions:
2/(/i
By using the
[16.22
condition y\
ax x again,
this simplifies to
lax  (yx + y 2 )y + yx y2
0.
(ii)
Chord
16.22
tx t2
(1)
is
2at x 2at 2
at\
at\
'
x%(tx + t 2 )y + atx t 2 =
i.e.
0.
so
tx
is
+ t2
the chord
+ 2mt+l =
t t 2>
x
0.
this quadratic in
= 2m\l and t x t2 =
^
(iii)
0.
chord be Ix + my + a
and t 2
l~r t 2
and
\\l.
Tfb
Hence
^
t t2
x
tx t2
which
is
equivalent to
1 tx
tx t
y+a
0,
(iii).
a, since
the equation
THE PARABOLA
16.23]
the numbers
for
given by
16.23 Tangent at
(iv), i.e. t
the tangent at
x is
to substitute
because
tx t2
and
2t
iv )
it
the points
0,
x  Wi + h)y + at ih
to
{t x + t 2 )t + t x t 2
we may
2
x,
607
circle,
the tangent
is
defined as a limit:
is
4ax2y1 y + yl =
Since y\
0.
yyx
2a(x + x1 ).
(v)
(2) Alternatively,
calculating dy/dx
can be found by
2y
at
dy
dx
4a
dy
dx
and
4ax,
2a
=
y
Px has equation
2a.
yyi = (xxi)>
which
(v)
is
(i)
to
above.
Letting
Omitting the
>tx in equation
suffix,
(iii),
we
xty + at 2 =
0.
obtain x tx y + at\=0.
t
is
(vi)
THE PARABOLA
608
(2) Alternatively,
[16.25
dy
dx
_ dy jdx =
=
2a
dtj dt
2ai
l'
as
^(x at 2 ).
is
ijr
1/t.
is
a polynomial in
the simul
(x,y),
P P2 leads to
1
respectively, or to
 xx
(x
 x 2 ),
(y
 yx
(y
 y2
an equation in a parameter
containing
(**i)(*f.).t
the chord
P1 PZ
when
tends to
is
or an equation for
if
the
containing
(t
tx
(x
xx
2
)
(y
yt
it
2
)
2
)
am 2 =
More
In; cf.
generally, if
approach
two
loci
7.
cut at
P1
and
2,
and
is
made
to
along one of them, then the limit of the other will touch
the first at Px because they have a common tangent there. For example,
if t %
tx
in ex.
(ii)
at the point t x
in 6.72.
f This is not the case unless the equation of the curve is algebraic; e.g. the chord
of thecurvey = e* joining (0, 1) and (l,e) isy 1 = (e l)x, and leads toe* 1 = (e l)x
factors.
THE PARABOLA
16.26]
609
16.26 Examples
(i)
The tangents
at
t lt t 2
{at 1 t 2 , a(t
2 )}.
x hy + atl
x t 2 y + at\
0,
y
a
so that
ty+x =
d.
and
0,
+ tz and
tt
tx t 2 ,
(ii)
on
The
'
a parabola
lies
the directrix.
The tangents at tlt t 2 meet at (af1 2 .a(<i + *2 )) Tiie perpendicular from this
point to the tangent at tz has equation
hx + y
aih + t^ + t^t^,
(a)
(b)
2
the condition for y = mx + cto touch the parabola y
2
2
If this line also touches the circle x + y* = r prove that
Find
4ax.
m* + m 2
a2

r2
0,
and hence find the equations of the common tangents to the parabola and the
x 2 + y 2 = $a 2
2
(a) The line y = mx + c cuts the curve y = 4ax at points for which
circle
(mx + c) 2
4ax,
m x + 2(mc 2a)x + c = 0.
2
i.e.
The line will touch the curve if and only if this quadratic in x has equal roots, i.e.
(mc2a) 2
mc
2 2
,
i.e.
4= 0,
the line
mx y + c =
t)
= mx + a/m touches y 2 =
will
be the same
m=
Ijt
and
at
x ty + at2
(the
if
m 1 c
T ~ ~ at
i.e. if
4ax.
line as
'
= afm.
THE PARABOLA
610
[16.26
Similar methods will give the contact condition for the line and circle,
simpler to equate the radius r and the perpendicular from the centre
0) to the line mx y + ajm = 0:
(6)
but
(0,
it is
ajm
V(m 2 +l)'
m + m 2  0,
i
m=
so that
1.
x + a,
y
Tangents from
(iv)
(m 2 
i.e.
x ;
1)
(m 2 + 2)
0,
are thus
x a.
From a
x i ~ tyi + at 2
0,
4mxx this gives two values t = tx t2 i.e. there are then two points
the tangents at which pass through Px .f
The argument used in 15.64 will show that the chord of contact of tangents
fr mPli8
yy 1 = 2a{x + x1 ).
and
if y\
(v)
>
The condition
to
lie
on y 2
4tax
(yl
+ kx 2
+
Jc
ly x
'
+ JcyA
+ )
Jc
reduces to
 4aa 2
h 2 + 2{Vl y % 
<La{x x
+ x % )} U + (y 2  AaxJ I 2 =
has
Exercise 16(6)
1
Verify
that,
when
{yyi)(yy*)
is linear
Deduce
it is satisfied
PxP 2
(iii)
2
y *ax
by the points
P P
lt
on
the parabola.
in 16.22 that
t x t % is
a focal chord
if
and only
1.
4 Prove that the chord whose extremities are given by the roots of
= has equation ux + vy + aw = 0.
ut2 + 2vt + w
THE PARABOLA
5 Find the equation of each of the tangents from
611
(
2a, a)
to y
4ax.
PM
geometry.]
(ii)
light at
from a source of
[Angle of incidence
reflector: rays
angle of reflection.]
9 If the tangent at
is the tangent at P,
10 PQ is a variable focal chord;
to the axis. Find the locus of the midpoint oiPU.
11
(ii)
(iii)
tangents at
tangents at
12 If P,
triangle
13
PQ is a focal chord;
(i)
x, t
tx t %
M a
+4 =
t,
prove that
0.
common
*18 If tx #= t3 in ex. (ii) of 16.12, and t2 > tx and <4 + 3 the limiting circle
touches the parabola at < x and at ts (double contact). Prove that the chord of
contact tt t3 is perpendicular to the axis, and that the centre of the circle lies
on the axis and has abscissa ^ 2a.
19 Sketch the parabolas y 2 = 4ax, x 2 = 4by (b > 0), and find the equation of
,
their
common
tangent.
20 Write down the equation of the chord of contact from P(, f ) to y 2 = 4x.
By finding its intersections with the curve, deduce the equations of the tangents
from P.
THE PARABOLA
612
[16.3
21 Prove that the pair of tangents from T(xl9 y x ) cuts Oy at points A, B such
that the midpoint of
is M(Q,\y x ). [Use ex. (v) of 16.26.] If P, Q are the
points of contact of these tangents, prove
_[_ PQ. [Use the chord of contact
AB
SM
from T.]
22 If tangents are drawn to the parabola from points of a given
that the corresponding chords of contact are concurrent.
line,
prove
(v).]
*26 Use the ratio quadratic (16.26, ex. (v)) to obtain the equation of the
tangent at x [Since x lies on y 2 4ax, the ratio quadratic has one root k = 0;
the other root gives the remaining intersection of PxPa with the curve. If Px P a
is the tangent at
is a
this intersection must coincide with Px and k =
x
repeated root. Hence yx y% 2a(xt + x t ) = 0, which gives the locus of P2 .]
16.3
Normal
16.31
Normal
at the point
has gradient
has gradient
l/t,
i.e.
= t(x at 2
+y =
2at + ats
),
(i)
Example
Prove that
the
normal
at
so that
ts,
is
the point
gradient of normal at
s,
=
t
ts,
t =
i.e.
gradient of chord
t is
see Ex. 16
+s
(c),
no.
4.
Conormal points
Given a point
drawn from
it
(C
>2/o)>
we may
to the parabola.
fo>>yo)rf
tx
is
enquire
+y =
cubic in
t,
The normal
2at
+ t3
at
will pass
through
THE PARABOLA
16.32]
613
curve such that the normals from them pass through (x y ). Ignoring
coincidences, a cubic has either one or three roots; therefore either
one or three normals can be drawn from a given point. Also see
,
Ex.
).
is
are concurrent at
)ty =
at3 +(2ax
tx , t 2 , t3
0.
(ii)
absent,
fi
+ <a + 'a =
(iii)
Three points on the curve which are such that the normals at them
are concurrent are called conormal points. Hence (iii) is a necessary
condition for the points
currence
is
t
3
x , t2 ,
to be conormal.
given by
2axQ =
dLtx t2
atx t2 tz
(iv)
h + 2 + t3 =
t
we see that t lt
2 , t3
>
(ii),
If
(x
coincide. If t 2
ts ,
By
8.53, ex.
normals
(i.e.
(i),
the
(ii)
then
).
2a + Zat\,
2at\.
evolitte)
distinct
is
<2
is
t3 , i.e.
when each
0).
Examples
(i)
The
circle
the vertex.
If tx , t 2 ts are the feet, then <x + a + ts = 0. If the circle through these points
cuts the curve again at tt , then by ex. (ii) of 16.12 we have tl + t2 + t3 + ti = 0.
Hence tt = 0, which gives the vertex (0, 0).
,
THE PARABOLA
614
[16.32
Find
The
+ (2ah)tk =
0.
(v)
0, its
The
(i) t
cubics
The required
is
0.
(4a + 2a) * + 4/
0.
0,
(v), (vi)
circle is therefore
x 2 + y 2  (h + 2a) x  \ky
(iii)
0.
on
the
By
(a)
gradient
has
y
yyi =  a {x  x J'
(b)
(h, k) lies
(h, k)
to the curve.
Then since
2a(ky1 + y 1 (hx1 =
)
lies
0,
on the locus
2a{ky) + y{hx)
0.
Exercise 16(c)
PN
P cuts
NG
G respectively.
Prove
T8 =
mid point
P in
of a focal chord
V. Find the locus of V.
4 If the normal at t meets the curve again at s, find s in terms of t by sub= as 2 y = 2as in the equation of the normal and either (a) factorising
the cubic in t, or (6) observing that the resulting quadratic in * necessarily has
a root s = t, and that the sum of the roots is 2 It.
5 (i) Prove that the feet of normals which meet at the point * of the curve
stituting x
+ st + 2 =
If the normals at
t lf t z
0.
fx 2
2.
THE PARABOLA
16.4]
(iv)
615
Show that normals at the extremities of a focal chord can never meet on
the curve.
6 Prove that the normals at the extremities of all chords of the form
x + ky + 2a = meet on the curve. Show that all such chords pass through a
fixed point, and give its coordinates. [This line cuts the curve where
at 2
+ 2akt + 2a =
so that
0,
tx t 2
2.]
7 Find the equations of the three normals from (15a, 12a) to y 2 = 4ax.
2
8 Show that two of the three normals from (5a, 2a) to y = 4tax coincide.
9 Prove that the centroid of the triangle formed by conormal points lies
on the axis.
10 PQ is a chord of a parabola drawn in a fixed direction. Prove that the locus
of the meet of the normals at P and Q is a line which is itself normal to the
parabola. [If tlf
t2
+ h) = m,
i.e.
t1
+ t 2 + { 2/m) =
0;
tx , t
2
also concurrent.
a{t\
t lt t 2
+ tx t2 + t\ + 2),
t3 from S x =
=  atitjfa +
(a;,
where
2 ).
and equations
y),
(iv) in 16.32.]
P, Q
Find the
loci of
T, N,
M.
* 14 If C is the centre of the circle through the feet of normals from r prove
r
are Glf C2 , G3 , and conversely.
that when
3 are collinear, then so
ly
2
[Use ex. (ii) of 16.32.]
,
P P P
,
15 Show that y =
2
4ax meets the locus in 16.32, ex. (iii) (6) at points P(x,y)
which y3 + 4a(2a h)y 8a 2 k = 0. Deduce that at most three normals can
be drawn to the parabola from a given point (h, h).
for
16.4
Diameters
conic
In 17.63 the above definition will be reconciled with the usual meaning
of 'diameter'.
16.42 Diameters of a parabola
The chord
m = 2/(
all
i.e.
+ 2 ) The
midpoint
satisfy
2a/m, which
is
THE PARABOLA
616
[16.42
Thus
the diameters of
(d),
a parabola are
no.
6.
Example
Prove that
P of a diameter
bisects.
at
(which
is
and
is
= m,
Fig. 168
Exercise 16 (d)
1
2x 3y+ 1 = 0.
2 Write down the gradient
parallel to
the diameter 2y + 3
of the chords of y 2
0.
4 V is the midpoint of a chord QQ' of a parabola, and TQ, TQ' are tangents.
Prove that the parabola bisects TV.
*5 Prove that the circle
[Use no. 3.]
drawn on a
directrix.
A point is such that its distance from a fixed line is equal to the length of
the tangent from it to a fixed circle. Prove that its locus is a parabola. Locate
the focus and directrix when the given line touches the circle.
circle
A circle passes through a fixed point, and the chord cut off by it from a given
line is of constant length.
Two
senses.
is
a parabola.
that (provided the latera recta are unequal) the locus of the
another parabola.
mid point of PQ
is
4 Prove
that
its
vertex
and
the
focal chords
and D. Prove
THE PARABOLA
617
P,
lies
either
on a line
parallel to Oy, or
to the parabola y 2
it
2
13 Prove that three normals cannot be drawn from (h, k) to y = lax unless
h > 2a. [Prove f(t) = at3 + {2a h)t k is steadily increasing for all t if h S 2a,
so that f(t) =
has just one root.]
2
4ax so that two of its sides touch
Prove that the third side touches y % = lex, where (2a b) 2 c = ab 2
[Use the condition km 2 = In for tangency of Ix + my + n = and y 2 = Ikx.]
16 Given a line I and a point not on it, a variable line is drawn through S
to meet I at V. A line p is drawn through V perpendicular to SV. Prove that
p touches a parabola having S for focus and I for its tangent at the vertex.
[Choose I for 2/axis, and the perpendicular to it through S for a;axis; let these
meet at O and p meet SO at T. If p has equation Ix + my + n = 0, then
OV n/rn and OT = w/Z. By geometry OV 2 = TO. OS, so am 2 = nl where
a = OS.]
(15.26) cuts y 2 = 4ax at
17 Show that the line through Px in direction
points for which r is given by
*15
4bx.
r2 sin2
(This
is
from
0.
18 Prove that the chord of y 2 = 4ax whose midpoint is (xx , y) has equation
yyx 2ax = y\ 2axx [For the required chord the roots of the distance quadratic
are equal and opposite, so yx sin 6 2a cos 6 = 0, which detemiines cos 6 : sin 6.
.
{xx1 )jy1 =
19
Px is y 2 2ax = yy 2axx
m is the line y =
(ii)
becomes
(yyjfia.]
(i)
2a/m.
y =
2
46a;.
THE PARABOLA
618
22 Using no.
17,
Pv
[Two roots r
0.]
23 Obtain the equation of the pair of tangents from P x [Equal roots r.]
*24 Show that (3x + 4y) 2 = 54a; 53y + 29 represents a parabola as follows.
(i) Verify that the equation can be written
.
(3x + 4y + k) 2
(ii)
+ 29).
By choosing k so that the lines
3x + 4y + k = 0, (6k + 54:)x + (Sk53)y + (k 2 + 29) =
PN
If PM,
are the perpendiculars from P(x, y) to the (perpendicular)
2 = 2PN.
3x + 4=y + 1 = 0, 4x3y + 2 = 0, (a) shows
By the Remark
in 16.11, the curve is a parabola of latus rectum 2, having axis 3x + 4y + 1 =
and 4x 3y + 2 = for tangent at the vertex.
(iv) For points on the curve, (a) shows that 4x 3y + 2 ^ 0; hence the curve
is on the origin side of 4x 3y + 2 = 0, i.e. lies below this line. Sketch the curve.
*25 Sketch the parabola (x + 2y) 2 = 56x + 12y 184, giving the equation of
the axis and the coordinates of the vertex.
(iii)
PM
lines
*26
(i)
If ab
= h2
form a perfect square, say (px + qy) Prove by the method of no. 24 that, unless
g :f = p :q, the equation represents a parabola whose axis is parallel to px + qy = 0.
(ii) In the exceptional case show that the equation can be written
(px + qy + X) 2 = A 2 c where g = Xp and / = Xq. This represents a pair of parallel
lines, a repeated line, or nothing according as A 2
c. (Algebraically, a parallel
linepair is thus a degenerate parabola.)
*27 Show that the curve whose parametric equations are
2
x
is
at 2
+ 2bt, y =
ct 2
+ 2dt
(ad
4= 6c)
a parabola. [Eliminate
t,
28 A curve is
the chord joining the points
t,t
+ e has equation
f(t)
g(t)
h(t)
f(t
e)
g(t
e)
h(t
=0.
+ e)
[r 3 > (ra
r 2 )/e, and
f(t)
g(t)
h(t)
f'(t)
g'(t)
h'(t)
=0.
619
17
THE ELLIPSE
17.1
The
17.11
SP = e.PM
We now
loci
8P e.PM
by
(e
>
0, e 4= 1),
is
therefore
x 2 + y2
i.e.
2 (l
 e + 2e
2
e 2 (x
2
k) 2
kx + y
e2k2 .
Fig. 169
On
'
e2 k \
1e2 /
e4
e2k 2
(le 2 )
(le 2 )
Change the
by the formulae
x
x'
e 2k
le,2'
!
42
+ le
2
y'
GPM
II
11
THE ELLIPSE
620
[17.12
Ve*)x*+y* =
x2
(i)
^+p=
In the equation
~.
y2
a9
On
dividing these
we
<
hence
<
r.9
ft
'
T^
get
 e2 =
j
and
~52 ,
e2
i.e.
a2
<
1;
1.
a;
becomes
A;
_
Hence
e
times
<
is e,
<
(ii)
its
is
whose focus
_a
&
(ii)
is
is S(ae, 0),
is
where b 2 = a 2 (l e 2 ) and so
the equation of an ellipse whose eccentricity
and whose directrix is the line x = aje.
Therefore
1.
e 2&
aje,
The algebra of
17.12,
with 6 replaced by
x2
y2
b2
a
is
(in)
'
the locus of a point whose distance from 8(ae, 0) is e times its distance
from the
line
aje,
where b 2
a 2 (e 2
1)
and so
>
1.
Therefore
THE ELLIPSE
17.14]
(iii)
is the equation of
directrix
x=
a hyperbola of
621
eccentricity
e,
focus S(ae,
0),
and
a\e.
and
directrix
SP = e.PM is expressed by
(xae) 2 + y 2
x 2 2aex + a 2e 2 + y 2
i.e.
this
e2
(xY
= a2 2aex + e2x2
becomes
which shows that the distance of P(x,y) from the point S'( ae,0)
is e times its distance from the line x = aje: S'P = e.PM'.
Hence both curves have a second focus S'( ae, 0) and a corresponding directrix x = aje.
17.15 Further definitions
From the equations, both curves are clearly symmetrical about the
and about the /axis. The origin is the centre of each.
The ellipse meets Ox at the points A (a, 0) and A'( a, 0), and meets
Oy at B(0, b) and B'(0, b). AA', BB' are called the major and minor
axes of the ellipse; their lengths are 2a, 2b respectively. The points
caxis
called
a latus rectum.
We now
17.16
Form
of the
=1
ellipse
a
THE ELLIPSE
622
so that
[17.17
Similarly
b^y^b
OS =
OS'
ae,
OD =
OD' = 
(iv) f
B'
Fig. 170
17.17 Circle
When b =
and
a,
ellipse
the equation
(ii)
circle
with centre
and radius a; but equation (i) cannot represent a circle for any choice
of e and k. Hence a circle cannot be defined by the focusdirectrix
property, and therefore is not a 'conic' in the sense of the definition
in 15.71; in particular, a circle is not a special {or degenerate) ellipse.
We may
by
(ii)
= a (l e 2 shows
2
relation b
approach
O and
the
OP2 =
x 2 + y*
a 2 (l^)
= OS/OA,
+ y* = a 2 +
and indeed
(l
~) y
indicates to
>a\
foci
THE ELLIPSE
17.2]
623
Thus the circle with centre and radius a is the limit of the ellipse;
and the equation of this circle is certainly the limit when b > a of
the equation of the
17.2
ellipse.
ellipse
The
circle
ellipse. Its
on diameter
equation
is
A A'
is
x 2 \y2
_
a2
Fig. 171
and
circle.
Then P, Q
Since
PN = y = J{a?x*)
= ^(OQ*ON*) = ^QN,
i.e.
Conversely,
so that
PN QN = b:a.
an ordinate QN to a circle of radius a is divided at P
i.e.
x
1jJlLa*^b*
P is
the
THE ELLIPSE
624
[17.22
Example
The elliptic trammel.
Through P draw the
PR' = Q0 =
a.
line parallel to
QO
to
meet Ox, Oy at
jB,
R'; then
PR _ PN
QO ~ QN'
and hence PR = b.
If a rod PRR' has pegs on its under side at R, R', and these pegs are made to
slide in perpendicular grooves AOA', BOB', the point P will trace an ellipse
whose major and minor semiaxes are PR', PR respectively.
17.22 Focal distances
The lengths SP, S'P are the focal distances of the point
Since SP = e.PM and S'P = e.PM' (fig. 170),
P on the
ellipse.
the
sum
2a.
bifocal property).
SP + S'P
such that
constant is
S be
(c,
0)
and S' be
c, 0).
SP + S'P =
any point P(x, y) on the locus
V{(*  c)
i.e.
<J{(x
Squaring,
(x
2a,
satisfies
+ y 2} + <J{(x + c) 2 + y 2} =
2a,
+ c) 2 + y2 =
(x
 c) 2 + y 2 + 4a 2  4a *J{(x  cf + y 2},
4a J{(x  c) 2 + y 2}
i.e.
4a 2  4ca.
Again squaring,
ie
x2
1.
positive.
a2  c2
y2
^+~/ =
a* a z c2
From
is
^jx2 + y2 =
THE ELLIPSE
17.23]
625
=a
i.e.
2 2
the
by
17.12,
foci.
Example
Mechanical description of the ellipse: pinconstruction'
Fix pins S, S' at distance 2ae apart. Tie the ends of a string of length 2a at
S and S'. A pencil point P moving so that the string S'PS is kept taut will trace
the ellipse with foci S, S', eccentricity e, and major axis 2a.
'.
'
Given two planes a, a' which intersect in a line I, consider perpendiculars dropped from points of a onto a'. If the perpendicular
from P in a has foot P' in a', then P' is called the orthogonal projection
of P on a'. If P lies on a locus 2 in a, then the corresponding locus
of P' in a' is called the orthogonal projection of S on a'.
Fig. 172
We now
and
P'.
common line I for a;axis in both planes, and take any point
for origin. Draw lines Oy, Oy' in a, a' at rightangles to Ox.
Choose the
on
it
Wo
is
a;'
The
circle
+y =
2
and
a', i.e.
y'
y cos 0.
2
2
cos a
THE ELLIPSE
626
Choosing 6 so that cos 6
x' 2
Hence
<
bja (b
y' 2
a2
equation becomes
a), this
=
^h
b2
the ellipse
[17.24
o2
If
making
x'Oy',
lx'
Effect
1)
on angles.
+ m sec 6y' + n
0.
<j>,
<j>'
projection cut
its
Ox
at the
becomes
same point,
and
tan 0'
tan 6'
tan
cos 6.
msec
.
Hence parallel lines project into parallel lines, but in general angles are changed.
However, a line perpendicular or parallel to Ox projects into another such line.
tangent to a curve projects into the tangent to the projected curve at the
(2)
corresponding point. The tangent to y = f(x) at (xlf y^) is
yVi
and
projects into
(y'
=f
(^i)'(xx1 ),
(x'
x'j),
i.e.
to the curve
(x[, y[)
y'
 x[).
(x'
of y =f(x).
By
(1),
into
a normal.
on lengths. If PQ in a has length r and makes angle <j> with Ox, let
projection P'Q' on a' make angle 0' with Ox. Through O draw OPx equal and
parallel to PQ; then
is
1 is (rcos0, rsm<j>) and so its projection P'
t
(3) Effect
its
(r
Hence
and
also
Since
OP? =
=
OP[ r cos Px OP'^.
by (1) OP'x is parallel
+ sin 2
cos 2 0)
to P'Q', therefore
0P'x
is
(i)
THE ELLIPSE
17.3]
627
Hence, in general, lengths are changed by orthogonal projection; they are conserved only on lines parallel to Ox.
Formula (i) also shows that the ratio of lengths on parallel lines (or on the
same line) is unaltered; in particular, midpoints project into midpoints. On
intersecting lines the ratio of lengths is altered.
(4) Effect
on areas. Since
Example
PN is
the ordinate of
P;
the tangent at
P meets
the
major axis
at T.
Prove
ON.OT = OAK
In it, OP _L PT, and by
which can be written ON .OA = OA.OT
and involves the ratios of lengths on the same line.
First sketch the corresponding figure for the circle.
geometry
ON.OT = OP2 = OA 2
Fig. 173
If the figure is projected orthogonally onto a plane through the line OT, this
relation will hold for the new figure, and is the required property of an ellipse;
to the circle will project into an ordinate
for since
JL OT, the ordinate
PN
PN
to the ellipse.
For
N.B.
17.3
Parametric representation
Let
point
<f>
let
circle
corresponding to the
^ = NOQ
628
angle of P).
Then x
THE ELLIPSE
= ON = acos0; and since NQ =
[17.32
asin^, we have
y=
<f>
The point
17.32
By putting t
z
Hence
is
1 t2
= acos0 = aj^,
for each
t,
a
\
on the
ellipse
2/
2t
= 6sin0 =
6jj^.
the point
/
lies
<f>.
1t 2 2bt_\
i+t 2 >l+t 2)
x 2 \a 2 + y 2 \b 2
we refer to
it
as the point
t.
Each value oft determines just one point of the ellipse but the point
;
as the limit
when
although
t,
it
can be obtained
t>oo.
b2
t
therefore
lxla
y\
1^/557* '
Hen*,
(l):g:( 1+ ?)
and so
a b
(1
f 2 )
(i)
2*
(1
2
).
Example
Goncyclic points on the
The
circle
a2
i.e.
(a 2
ellipse.
x 2 + y 2 + 2gx + 2fy + c
/lt*\

+ 62
l

2t
\
)
+2ga
1t*

for
which
2t
+ 2/6 + c =
0,
0.
(ii)
THE ELLIPSE
17.32J
629
Since this is quartic in t, a circle and ellipse can cut in at most four points; the
(some of which may coincide),
actual number of intersections will be 4, 2, or
since a quartic has 4, 2 or no roots.
Suppose the above circle cuts the ellipse at the points tlf t 2 , t3 tt ; then these
numbers are the roots of (ii). Since the coefficients of J8 and t are equal, therefore
,
Now by
14.23,
(viii),
i^^i^.
which
is
by the above
zero
condition.
Thus
Hence S$0 X
*/ the points
20 x = 2mr.
This necessary condition for concyclic points
X,
= nn
a,
<Ui
some
for
integer
3,
concyclic, then
is
Ex. 17 (a),
no. 11.
Exercise 17(a)
'
The
ellipse''
means
the locus
x2 /a*+y 2 /b*
1.
to BB'.
3
If
foci
S, S' geometrically.
A rod AB moves with its ends on two fixed perpendicular lines; P is fixed
and BP = a, PA = 6. Find the locus of P.
in the rod,
6<j>
<!< a is
is
<f>).
).
centre
is
when
(i)
2a;
<f>
(ii).]
and cuts
13 If a circle touches an ellipse at
are equally inclined to each axis.
AC,
it
again at
AD
common chords
THE ELLIPSE
630
*15
[17.4
Find the point where the circle of curvature at ^ meets the ellipse again.
(i)
If the points <fi lt <fi 2 <p z , <j> t are concyclic, and the circles of curvature at
these points cut the ellipse again at <j>' <f>' <j>' <fi' prove these points are also
v 2 3 v
(ii)
concyclic.
*16 With coordinate axes as in 17.23, show that the orthogonal projection of
the parabola y 2 = 4nax is another parabola.
*17 Prove that the tangent to the circle x 2 + y 2 = a 2 at the point (a cos 0, asin0)
has equation #cos0 + 2/sin0 = a. By projection deduce the equation of the
tangent to the ellipse at the point <f>.
HK
OM
*18
is the midpoint of a chord
of an ellipse whose centre is O;
meets
the ellipse at P;
is the midpoint oiPH; and ON,
meet at Q. Prove that
bisects the chord which passes through
and Q. [In the circle figure, Q is
the orthocentre of triangle OPH.~\
HK
OH
prove
OM.OT = OB
with semiaxes
ellipse
a, b
from that of a
circle
of
radius a.
17.4
17.41
Chord P,P2
Since
and P2
lie
and
on the
(d).
P2 has equation
ellipse,
^+
 + pL
and
b2
so that
x\x\
a2
~ =
x 2 xx
a 2 y2
'
1.
yt
therefore becomes
b *
x2 T
+ x*1/
"
t
a2 yz + yi
I.e.
ie
^K + ^) +
t
I
a2
p(2/i
b2
+ 2/2 =
)
^+^+l
(i)
THE ELLIPSE
17.42]
631
{xxx ){xx% )
(yy1 ){yy2 )
b2
a2
a/
'
It is satisfied
(x v
(x2
yx ),
ellipse,
and so must
(a cos 0, b sin 0)
has equation
y2 ) on the
17.42
Chord
6<
(a cos
ybam.6
y
Y =
,
<j>,
b sin
and
<p)
&(sin0sin0)
j
 a cos g>)
Y
,
(a:
a(cos0cos0) v
_ 26cos(0 + 0)sin(00)
  2a
2arinlf/9nA)
sin^^)
sin
sin %{6 +
(0  <p)
<f>)
6cos(0 + 0)
a sin (0 + 0)
(#
acos0),
i.e.
cos
<j>)
+ 1 sin \{d +
<j>)
cos $ cos (0 + 0)
cos(0 + 0)+sin(0 + 0)
i.e.
17.43
Chord
t2
<j>)
(ii)
x t2
if
_ 1+ua
^+
t\* 2
which
0.
and then
Hence
for
2t
i.e.
sin (0 +
cos(00).
and
+ sin
ft fa
The line ux + vy + 1 =
and
2vb
ua,
THE ELLIPSE
632
[17.44
= tv t =
chord
x t2
(*i + y* +
:*:lfrom
(i)
(t
by
t 2 = 0, i.e.
)
(iv)
by
substituting for
in 17.32:
x and
given by
points
VJM
Tangent at
y,
(iv),
o.
2?h.2m =
a2
since
(t
)
(iv)
(ig(* 1+ y + M2 (i + 3 =
It is linear in
tx
0
(iii)
62
and y2
"^
"
a2
>
yx
in equation
9'
Letting 6 >
(J)
in equation
(ii),
we obtain
x
v
cos ^+^sin^
r
a
b
Letting
t2
>
(vi)'
1.
(a),
no. 17.
in equation
(iii),
we
suffix 1)
(l* 2 ) + 2*
a
o
(vii )
t.
THE ELLIPSE
17.47]
633
28.
17.47 Examples
Condition for
(i)
= mz+c to
x2
touch
y
+
=
1.
The line
x%
a2
(mx + c)* _
~h
62
i.e.
The
roots,
line will
touch the
ellipse if
i.e.
and only
if this
0.
ca
i.e.
lines
= mx V(a2m2 + 6 2
ellipse.
(ii)
y
{x*  a 2
i.e.
mx V(a 2m2 + 6 2
m2  2x y m + (y\  62 =
),
0.
Since this is quadratic in m, in general two tangents can be drawn from a given
point (x y ) to the ellipse. The roots lf 2 are the gradients of such tangents.
These tangents will be perpendicular if l 2 = 1, i^ if
m m
mm
2 a2
i.e. if
(x
lies
on
the locus
x*+y 2 = a 2 + 6 2
This
is
circle concentric
with the
ellipse,
(ii), (iii)).
15.64 will
THE ELLIPSE
634
Exercise 17(b)
1 Use equation (i) in 17.41 to prove that the midpoints of all chords of given
gradient m lie on the line y = b 2x\a 2m.
2 Find the condition for the chord 6<f> to pass through the centre of the
ellipse.
3 Prove that
Verify that this
6<}>
if
e cos (0 + 0) =
is
(e
l)/(e + 1)
or
+ l)/(e 1).
cos
(e
<j>).
4 Prove that the chord PP' of the ellipse and the chord joining the corresponding points Q, Q' of the auxiliary circle meet on the major axis. Deduce
a property of tangents at corresponding points P, Q.
5 If 6 +
conversely.
is
is
and
constant.
7 Find the envelope of the chords joining points of the ellipse whose
by a constant. [Let the angles be a, + a>; the
chord is (x/a) cos + (y/b) sin = cos a, which is a tangent to the ellipse
x 2 /a 2 + y 2 /b 2 = cos 2 a.]
eccentric angles differ
<f>
<f>
<j>
8 Prove
a 2 l 2 + b 2m 2
that
= n2
lx
+ my + n =
touches
(i);
x 2 /a 2 + y 2 /b 2
and only
if
if
2, 3)
to 4x 2 + 9y 2
36 are perpendicular.
for the
minor
PSZ is a rightangle.
1 3 (i) If Q is the point of contact of the other tangent from Z in no. 12, prove
geometrically that PSQ is a straight line.
12
directrix at Z.
Prove
(ii) Deduce that tangents at the extremities of a focal chord meet on the
corresponding directrix.
14 Prove that the focal radii SP, S'P are equally inclined
and also Ex. 17 (c), no. 3.)
to the
tangent at P.
15 If, p, p' are the lengths of the perpendiculars from the foci S, S' to
tangent, prove that pp' = b 2
any
16 Using the similar triangles SPY, S'PY', prove p/r = p'/r', where Y, Y'
are the feet of the perpendiculars in no. 15 and r SP, r' = S'P.
17 From nos. 15, 16 and the bifocal property r + r' = 2a deduce that the (p, r)
equation (8.2) of the ellipse wo the focus S as pole is b 2 /p 2 = 2a/r 1.
18 Prove that the points Y, Y' in no. 16
perpendicular from S(ae, 0) to
lie
x
sin
*
cos
a
+ Vsin o =
is
on the auxiliary
V
cos <b
a
i
ae
sin
b
<p
circle.
[The
THE ELLIPSE
635
19 The chord PQ subtends a rightangle at the vertex A(a, 0). Prove that PQ
passes through a fixed point on the major axis. [Use equation (iii) in 17.43.]
2
20 Prove that the chord joining the points given by the roots of ut + vt + w
=
0.
has equation (w u) x/a + vyjb + (w + u)
21
[Use equation
(vii)
tx , t2
x/a _ 1+xja _
in 17.46,
meet at
(x, y),
where
yjb
2
22 Write down the equation of the chord of contact from ( 1 2 ) to \x + $y 2
and deduce the equations of the tangents from this point.
,
2
2
23 The chord PQ of the ellipse x z /a 2 + y 2 /b 2 = 1 touches the circle x + y
Prove that the tangents at P, Q meet at a point T on the ellipse
x2
o*
y2 _
+ 6* ~
c2
1
'
c2
ellipse
touches the
circle;
use no.
8.]
PP
x
in the ratio
will
By
taking
on the
ellipse,
1 )''
ellipse).
Pv
Obtain from no. 25 the equation of the pair of tangents from Pv [See
(ii).] Deduce the equation of the director circle of the ellipse by using
the condition for these lines to be perpendicular (15.52 (2)).
27
15.64, ex.
so that
15.65(2).]
*29
(i)
a2
a;
/a 2
+ y 2 \b 2 =
b2
to be the line
L'
prove that if the polar of Px passes through P 2 then the polar of Pa passes
through Px
(ii) Prove that the polar of a focus is the corresponding directrix.
(iii) Verify that the centre has no polar.
,
43
GPM
II
THE ELLIPSE
636
17.5
[17.5
Normal
normal
xx 1
xx \a
ax sec
2atx  b( 1
17.52
<j>
to the ellipse at
0,
lt
y yi
==
yjb 2
by cosec
 t2 y =
{1)
'
a2 b 2
2(a 2  b 2 )
(ii)
(iii)
+1
Conormal points
The normal
at
2ath b{lt 2 )h
i.e.
bkt*
+ 2(ah + a 2  b 2
is
) t
2(a 2  b 2 )
(h, k) if
* (1
+ 2(ah a 2 + b 2 )tbk =
quartic in
t,
at
0.
(iv)
4, 2, or no such normals.
Suppose that the normals at t v t % t 3 i 4 meet at (h, k); then these
numbers must be the roots of (iv). Since the term in t 2 is absent, and
is
minus the
Y>t x t 2
coefficient of 4
and
we have
t^t^ = 1.
(v)
dent conditions are required in order that four lines shall pass through
the same point.
Example
From formula (iii) in 17.32 it follows that, when conditions (v) are satisfied,
= (n + )7r for some integer n (positive, negative, or zero). Hence
= (2n + 1 7T is a necessary condition for the points x 2 z i to be conormal.
then
<f>
, <j>
, <f>
Clearly this single condition cannot be sufficient. The example in 17.32 shows
that four distinct conormal points can never be concyclic.
THE ELLIPSE
17.6]
637
Exercise 17(c)
1 The normal at P meets the major axis at G, and PN is the ordinate at P.
Prove that 00 = e 2 ON. What is the corresponding result for the minor axis?
.
2 Prove that
SG =
e.SP.
Use no. 2 to prove SG/S'G = SP/S'P, and hence that PG bisects SPS'.
Deduce that the tangent at P is equally inclined to SP, S'P. (Cf. Ex. 17 (6),
3
no. 14.)
14,
and
find the
its foot.
6 If to + my + n
is
a normal to x 2 /a 2 +
a2
b2
/b 2
1,
prove that
(a 2 6 2 ) 2
7 PN, PN' are the perpendiculars to the axes from a point P of the ellipse
x2 /a 2 + y2 /b 2 = 1. Prove that NN' is always normal to a fixed concentric ellipse,
and give the equation of this curve.
*8 (i) If the perpendicular from the vertex A to the normal at the point
meets the ellipse again at the point <p', use Ex. 17 (6), no. 5 to prove <j)' = 20.
(ii) Deduce that the perpendiculars from a vertex to four concurrent
normals meet the ellipse again in four concyclic points.
*9 Prove that the circle through any three of four conormal points cuts the
<j>
ellipse
<f>'
t
are
= (2n + 2) n.~]
$x + 2 + 3 +
2
10 Prove that the feet of normals drawn from Px to x ja 2 + y 2 /b 2 = 1 lie on the
curve (a2 b^xy + b^x a 2xx y = 0. Verify that this locus passes through Px
and the centre of the ellipse. [Method of 16.32, ex. (iii) (6).]
*11 Prove that the curves a 2xx y = (a2 b 2 )xy + b 2yx x and b 2x 2 + a 2y 2 = a26 2
S0 X =
(2w +
1) n,
b (a x x y)
a:
therefore
which y
2
{(a
6
<j>
satisfies
)2/
+ 6Vi} 2 =
a 2 (b 2 y 2 ){(a 2 b 2 )y + b 2y x }\
The
17.61
distance quadratic
equations
^
y^) in direction
Q
Q
= x x + r cos a,
y = y x + r sin a.
It cuts the ellipse x 2 ja? + y 2 /b 2 = 1 at points for which r satisfies
.
b 2 (x x + r cos 0) 2 + a 2 (y x + r sin 6) 2
i.e.
(a 2 sin 2
a?b 2
+ b 2x x cos d) r
+ (b 2xl + a 2y\a 2b 2 =
)
0.
432
(i)
THE ELLIPSE
638
[17.62
on the
Example*
Newton's theorem.
y\
roots of
// cos
it
(i) is
fl
// \~a?~
sin 2 0\
~b*~)'
f
drawn
x Q'R
in direc
follows that
P Q.P R
P Q'.P R'
1
Pt
/ cos
2 6>'
Fig. 174
sin 0\
which is independent of
and
Hence if chords PX QR, P X Q'R' are drawn
through Px in fixed directions d, d', the ratio PX Q .PX R:PX Q' .PX R' is independent of Px
In particular, when Px is at the centre the ratio becomes OK 2 GIT 2 where
.
the radii
in directions 6, 6'.
P Q.P R _ OK*
P Q' ,P R' ~ 0K'
X
'
is
Px
17.62
P1
in direction 6
is
b^cosd + a^sind =
This determines the direction cos 6 sin
:
at P,
Its equation
x ~ xi
cos 6
bisected at
and
then the
0.
= yVx
sin
therefore becomes
b %x x {x  x x )
Pv
opposite, so that
+ a*yx {y  y x =
)
0,
THE ELLIPSE
17.63]
639
17.63 Diameters
If the chord
(ii)
m =  b^x^ahf^
m satisfy
Hence
62
a 2m
Ex. 17
no.
(b),
1.
Examples
Tangents at the extremities of a diameter are parallel
(i)
chords which
to the
Let P1 be a point where the ellipse is met by the diameter which bisects all
chords of gradient m; then y1 = bhc^ahn.
The tangent at x (17.44) has gradient ft^/a2^, ie  The result follows.
(ii)
bisects
that chord.
0i>2/i)
The sum
{xlf y x )
constant.
(^ + H =
ellipse
x 2 /a 2 +ys /b 2
has
1
7\~~
1
r
and
r
a*
62
2
+ iTT) = sin 2 ^
= cos^ + ^tt) sin^
2
2
1
1111
cos 2
1
62
_ + _ = _+_
Addmg,
Strictly, the
diameter
is
is
THE ELLIPSE
640
mm' = b
i.e.
ja
The symmetry of
all
[17.64
this relation
y = mx.
is
all
chords parallel
Hence y = mx, y
mm =
It
is
from the
clear
if and only if
~.
a2
Examples
//
(i)
<f>
6,
<[>
\tt.
is
tan 6 tan 6Y =
a2
<J>,
and the
^a
,
a2
<f>
an
Perpendicular diameters of a
conjugate diameters of
If
ellipse.
x* + y2
circle.
circle
17.23)
*(iii)
through
First
Then
/t
draw the
(fig.
175)
figure for
circle,
replacing 'conjugate'
by
line
'perpendicular'.
ratios of lengths
qy
qqs
PV.VP'
OP 2
"
THE ELLIPSE
17.64]
Choose OP,
641
OD for axes of x and y, and then let Q be the point (x, y).
Since
a\
a\
b\
This has the same form as the standard equation of the ellipse, which corresponds to the choice of the perpendicular conjugate diameters OA, OB for
coordinate axes.
Exercise 11(d)
2
2
1 Write down the equation of the diameter of 2x + Sy
the chords parallel to 2x
+ 3 = 0.
which bisects
In
the following,
are P', D'.
is the point
in fig. 176, show that the eccentric angles of D, P', D',
3 If
Q, Q' can be taken as <fi + %7T, <J) + tt, ^> \tt, <j> + a, $ a respectively. Write
out in full the coordinates of P, D, P', D'
4 If P(a cos <j>, b sin 0) and D( asin^, ftcos^) are extremities of equal
conjugate diameters, prove that tan a = 1. Deduce that the equations of
the equiconjugate diameters are y = bxja, and show that they lie along the
diagonals of the rectangle which circumscribes the ellipse. State the length
of OP.
5 Prove that
OP 2 + OD 2 = a + b\
2
6 (i) Prove that tangents at the extremities of a diameter are parallel to the
conjugate diameter. [Use ex. (i) of 17.63.]
(ii) If the tangents at P,
meet at T, pro ve OPTD is a parallelogram which
has area ab. (Cf. no. 15.)
(iii) If p is the distance of P from OD, prove that p OD = ab.
ellipse
wo
its
'
THE ELLIPSE
642
8 Prove PS. PS'
= OD
2
.
[Use Ex. 17
no.
(a),
3.]
9 The two chords joining any point of the ellipse to the extremities of a
diameter are called supplemental. Prove that supplemental chords are parallel
to conjugate diameters.
10 (i) Prove that the lines px 2 + 2rxy + qy 2 =
are conjugate diameters of
x2 /a 2 + y 2 /b 2 = 1 if and only if pa 2 + qb 2  0.
(ii) Find the condition for Ix + my + n =
to be the join of extremities of
two conjugate diameters of x 2 /a 2 + y 2 /b 2 = 1. [Use 15.54.]
(iii) Prove that the chord joining the extremities of conjugate diameters of
an ellipse touches another ellipse. [Use Ex. 17 (6), no. 8.]
11 If a tangent to the ellipse cuts the director circle at U, V, prove that O U,
diameters.
* 12
PR at T.
P bisects TQ.
P meets
the axes
PQ
*14
is
HK
a diameter and
at Y. Prove that
XY
is
is
PQ.
*15
at
constant area 4ab. (Cf. no. 6 (ii).)
Rt
^ p^, p^
OK' OK ~ OK' OK'
'
it
If parallel chords Px QR, P 2 Q'R' are drawn through the fixed points Px P a
prove that the ratio P^.P^.P^Q' .P2 R' is independent of the direction.
[Use the distance quadratic]
*17
xxl
yy 1

a2
b2
_
~
x2
y2
a2
b2
he on
'
quadratic in Ex.
PP PP
X
1 7 (6),
and PX A
AP
Z,
AP P
the roots k
2,
0, kjl
= .]
of
The
lines
P when
bx + aty ab
0, btx
ay + abt =
varies.
2 With a given point S and line d for focus and directrix, ellipses are drawn.
Prove that the locus of the ends of the minor axes is a parabola. [Choose S for
origin.]
THE ELLIPSE
643
3 Prove that the tangents at the ends of a latus rectum are concurrent with
the major axis and corresponding directrix.
*4 P,
at P,
circle is
1,
Prove
OP
6 Circles with centres S, S' pass through a point on the ellipse. Prove that
the common tangents to these circles touch the auxiliary circle, and that their
points of contact with this circle lie on the common chord of the original circles.
[Use pure geometry and the bifocal property.]
7
is
variable point
to S Y.
OS for #axis, let p meet Ox, Oy at T, T'. Explain why triTYO, TST' are similar, and prove OY.TT' = OT.ST'.
(ii) If S is (c, 0) and p has equation Ix + my + n = 0, prove that the equation
2
2 where
2 2
6 2 = a 2 c2
(i) reduces to a l + b^m = n
(i)
Choosing
angles
in
(iii)
ellipse
circle
when S
lies outside
the given
circle.
9 Find the
common tangents to x 2 +y 2 =
25 and
a;
/169 + y 2 /16
1.
10
The
11
(i)
and only if x
(ii)
+y
1 >
to x 2 /a 2 + y 2 /b 2
fa
From
/b
1 if
0.
if
then Px and Pa lie on opposite sides of the ellipse. (Accordingly, the set of points
for
>
is
ellipse.)
12 Using the contact condition ca = aha2 + b 2 for the line y = mx + c and the
ellipse x 2 /a 2 + y 2 /b 2 = 1, show that the equation of the pair of tangents from
P(*1 ,y1 ) can be written (yt x x y) 2 = a^y
+ b^x Xj) 2
Let these tangents cut Ox at E, F. Find the locus of (i) if PE, PF are perpendicular; (ii) if EF has fixed mid point (k, 0).
y^
TP, TP' are tangents to the ellipse. Find the equations of the linepairs
O parallel to TP, TP' and to ST, S'T. By showing that they have the
same anglebisectors, deduce that ST, S'T are equally inclined to TP, TP'
13
through
respectively.
THE ELLIPSE
644
ellipse
17 If the normal at
meets the major and minor axes at Q, Q', and if OF is
the perpendicular from the centre O to this normal, prove that PF.PO = b 2
and PF.PG' = a 2
1 8 If the normal at an extremity of a latus rectum passes through an extremity
of the minor axis, prove that e4 + e 2 1 = 0.
19 P and Q are variable points on x 2 /a 2 + y 2 /b 2 = 1 such that the mid point
of PQ lies on x2 /a 2 + y 2 /b 2 = k 2 Prove that the tangents at P, Q meet at a point
T on x 2 /a 2 + y 2 /b 2 = 1/k 2 [The chord of contact from 2 (ar1 ,y 1 ) is the same line
as the chord PQ having midpoint (x2 y 2 ), where x 2 /a 2 + y 2 /b 2 = k 2 .]
20 The tangents to x 2 /a 2 + y 2 /b 2  1 at P and Q meet at T{h, k). Obtain the
equation of the linepair OP, OQ. Find the locus of T when the diameters OP,
OQ are {i) perpendicular; (ii) conjugate. Show that these two loci meet at four
points which are the vertices of a rectangle.
21 OP, OD are conjugate semidiameters of the ellipse. The circles with
diameters OP, OD meet again at Q. Prove that Q lies on the curve
.
2(x 2 + y 2 ) 2
a 2x 2 + b 2 y 2
to
1.
that x lies on
tangents in no. 12.]
23
25
ellipse,
and find
and directrices.
&x + 2) 2 + \(yl) 2 = 1.
3(xy+l) 2 + 4:(x + yl) 2 =
24 8x2 + 9y 2  lQxQy
12.
63.
the centre,
645
18
THE HYPERBOLA
The hyperbola
18. 1
^p=
asymptotes
18.11
hyperbola
^2
in 17.1317.15, of the
oi2
5 =
i
7i
From
x2
^=
the equation,
so that
1,
y
6
is
y2
4p >
\x\ $s
between the
we see
that
when x ~> + cc
i.e.
lies
or x >
a2
oo,
'
therefore unbounded.
and D'
lies
between
and 4'.
1)
THE HYPERBOLA
646
2
x /a
ellipse
point
+y
/b*
^ a and
have
(2)
An
\yx
1
\
[18.13
P1
on
this curve
we
b.
asymptote of a curve
is
when
> oo along
the curve.
In
may mean
some fixed
The informal use of the term 'asymptote'
in 1.41,
Remark
(a), is
and so
M=7(M)The perpendicular
which by
(ii)
is
distance of
from the
<">
line xja
yjb =
is
equal to
When Px > oo along the curve, this expression tends to zero. Hence
the line xja yjb =
is an asymptote to the hyperbola (i). Similarly
the line x/a + y/b = Qis an asymptote.
The equation of the asymptotes, regarded as a linepair, is therefore
62
a2
Alternatively, the equation
y
If y
can be written
(i)
a a \*
/
1
^)
'
y
then
2/carve2/asnnPtote
series,
= +x[l
a 2 \*

when
"
+ (")" +
'
+CO.
THE HYPERBOLA
18.14]
and
this will
a2 \
= a X \ l x*)
647
and bja m =
=
= bxja
0. Hence
an asymptote to
is
i
'
Since
for
y\\
hence
lies entirely
the,
hyperbola
= bx/a which
18.14
If
The
Fig. 179
Fig. 178
Fig. 177
bifocal property:
lies
SPS'P= 2a
S'PSP = e.PM'e.PM
M'
= e.MM'
=()
=
If
S',
2a.
Fig. 180
then
2a.
'
THE HYPERBOLA
648
[18.2
The algebra of 1 7 .22 shows that, with the same choice of axes,
a2
a2 c 2
<
so that a
c.
= PS'PS <SS' =
b2
c2
i.e.
2c,
a 2 and
,
Example
Mechanical construction of a hyperbola.
A rod APB is movable about the fixed end A, and a string BPG passing
through a small ring P which slides along the rod is tied to the other end B and
to a fixed point G. If the string is kept taut,
prove that in general P moves on one branch
C
of a hyperbola with foci A, G.
Let I be the length of the string. Then
'
BP+PG = Since
AP = ABBP = AB(lPC)
l.
= (ABl)+PC,
AP PC is constant, viz. AB
AB
the locus of P is that branch
therefore
If
=#
I.
I,
of
Fig. 181
18.2
^+p= 1
Since the equations of the ellipse and hyperbola differ only in the
sign of 6 2
it
follows that
many
written down from the corresponding results for the ellipse by replacing
b2
by b 2
methods used
(i)
Chord
PX P2
is
a summary,
by the
THE HYPERBOLA
18.3J
(ii)
Tangent at
Px
(ui)
Normals
Pl
649
= L
^+^
0.
(see 18.13).
(v) If
bja <
m<
b\a,
the lines
= mx *J(a 2m 2 b 2
Director circle.
the
is
circle
x2 + y2
= a2 b 2
provided a >
(vii)
b2
a2
(viii)
Px
L_t_A
x l_yl_A 3
2
a2 b 2
J [a
(ix)
b2
xx i
a2
Px
= (xx2i_yyi_iY
2
[a
'
Wi _ x l
b2
a2
yl
'
b2
lie on
(x) Diameters. The midpoints of all chords of gradient
proved
for
stated
and
the
diameters
properties
Wxjahm.
The
of
y
ellipse in 17.63, exs. (i), (ii) remain true for the hyperbola.
(xi) Conjugate diameters, y = mx and y = m'x are conjugate if
and only
if
mm' =
18.3
^2
5.
a2
Parametric representation
The equation
(i)
in 18.11
all
0,
is satisfied
by x = ach0, y =
bsh<fr for
THE HYPERBOLA
650
[18.32
x
If t
now
is
1
t2
=a +
1
T
fi
oo.
(If
and +
18.32
b
2t
1 T
then
= thf^;
still
0);
denotes th \$,
1.)
The point
directly.
For since
IHH(H
= ~ =
x/a + 1
xjal
f
therefore
y/b
._
y/b
(H
Hence
::1 =
i.e.
! fi
a b
(l
+1
t,
say.
J
<
:1 '
+ t*):2t:(lt*).
Each value of t except t = + 1 gives just one point of the curve; and
each point of the curve except A'( a, 0) corresponds to just one
value of t. We refer to the point
/
a
\
as the point
18.33
t.
The point
The equation
t
<j>
also satisfied
(i) is
We refer to (a sec
If we put
1+t 2 2bt \
it*'lt 2 )
(j>,
by x = a sec $,y =
tan \<f>,
we obtain the
point
(i)
can be written
in 18.32.
THE HYPERBOLA
18.4]
we may put
x_
651
xyl
a
the curve;
t
p
Fig. 182
are:
1 <
AQ
PA
A'Q'
< 1
<
<
<
18.4
sec0 ^tan^ =
Tangent at 0:
$
t:
ax cos
2atx + b{l
<j)
+ by cot
+ t*)y =
<f>
= a2 + b 2
2(a 2
</>).
1.
t:
at
cos %($ +
kh
Normal
+ 62
1t 2
'
is
If
GPMII
THE HYPERBOLA
652
of P. [Let
AB;
to
u=
velocity of sound, v
velocity
parallel
then
AP
\n
= AB PB
1
(AB
An nn =u[
APPB
so that
n\ =
constant.]
Find the locus of the centre of a circle which touches externally two given
whose centres are A, B.
circles
(i)
+t
is
traced
by the point
2bt \
it 2 )
as
(ii)
(6)
oo to + oo.
What is the condition for the points tlt t2 to
increases from
(a) lie
be diametrically opposite?
AP
x 2 /a 2 y 2 /b 2
1
is
whose
vertices
the hyperbola
1.
6 If the points
prove that
6,
<j)
_e
or
/b 2
1,
e
.
Prove that
(ii)
focal
corresponding directrix.
of
9 Y is the foot of the perpendicular from S to the tangent at any point
the hyperbola. Prove that Y lies on the circle x 2 + y 2 = a 2 (the auxiliary circle).
lies
...
a; also
OS.OD =
a2
= OF 2
so
ODF
is
right
angle.]
With
13 the chord
then
S^tfjj
14 the tangent at
tx t 2 .
16 Prove that
if
and
x t % t z t
15 the normal at
t lt t 2 , ts , 4
t.
= 1.
17 Prove that the feet of four concurrent normals cannot all lie on the same
branch. [If they do, then by no. 4 (ii) (a) \tr are either all less than 1 or all greater
than 1, contradicting I^Ma^l = 1 in no. 16.]
\
where n
0 X = (2n+l)n
THE HYPERBOLA
Show that the feet of the normals from P to x
18.5]
19
20 If the points
18.5
t lt t , t ,
2
3
4
653
%
la
\b %
1 lie
on the
0.
x y
+fb
a
It meets the hyperbola at points
l+t 2
l.
(l
i.e.
i.e.
Since
=1
gives
by
section given
=
for
has equation
k.
which
2t
+ lt 2 ~
'
+ t) 2 = k{lt 2 ),
= l
1+* =
or
jfc(lJ).
no point of the curve, there is a unique inter(k l)/(k + l). A similar result holds for lines
0.
an
a
where
=1
b2
a2
of the tangent at
xx
e
xx
Px can be written
'
'
a2
x\'
The last equation shows that when Px > oo, then yx /xx
bja or b\a.
hmit
The
of the equation of the tangent at Px is thus either
#
a
x
or
+Jb =
0.
An
tangent at
when
442
THE HYPERBOLA
654
[18.53
Px > oo.
18.53 The family x 2 \a 2 y 2 \b 2
for all k.
kb 2
'
^_ll_
ka 2
which
is
kb 2
ie
'
tto
~
a2
b2
'
independent of k.
Example
In the quadratic which gives the meets of x 2 /a 2 y 2 /b 2 = k with the line
2
0, the coefficients of x and x are independent of k. Hence the
sum of the roots is independent of k. Taking k = 1, 0, it follows that the a;coordinate of the mid point of any chord of the hyperbola x 2 /a 2 y 2 jb % = 1 is the
same as the ascoordinate of the mid point of the same chord of the asymptotes
x 2 /a z y 2 /b 2 = 0. The same is true of the ^coordinates since Ix + my + n =
gives y uniquely in terms of x. Consequently, */ the line meets the hyperbola at
P, P' and the asymptotes at Q, Q', then PP' and QQ' have the same midpoint M.
Since
= MP' and
 MQ', therefore PQ = P'Q'. In particular,
when P' P this becomes PQ = PQ'; i.e. the part intercepted on a tangent by
Ix
+ my + n =
PM
QM
18.6
18.61 Definitions
THE HYPERBOLA
18.62]
655
IS
1.
b2
For, with
(ii)
Oy
for #axis and Ox' for yaxis, the equation of the conjugate hyperbola is x 2 jb 2  y 2 ja 2 = 1 On rotating these axes clockwise
.
viz.
y2
b2
(iii)
Remarks
Equation (iii) difiFers from (i) by the same constant that (ii)
from (iii). If we transform these equations by a change of axes
(15.73), this relationship will be preserved.
(/?) The equation of any hyperbola whose asymptotes are
(a)
differs
is
a2 x + b 2 y + c 2
A,
a1 x + b 1 y + c 1
0,
By
writing
+ b 1 y + c x ){a2 x\b % y + c z ) =
for A,
we
0.
hyperbola.
(y)
In particular, any hyperbola whose asymptotes are the co= 0, y = has equation xy = A.
ordinate axes x
wo
wo a hyperbola,
THE HYPERBOLA
656
[18.62
(ii)
The diameter y = mx
cuts hyperbola (i) if and only if \m\ < bja. The conjugate diameter y = b 2x/a 2 cuts (i) if and only if \b 2 /a 2m\ < b/a, i.e. \m\ > b/a.
Hence of two conjugate diameters, only one meets the hyperbola.
The other meets the conjugate hyperbola. For if y = mx cuts (i), then m < b/a;
in this case \b 2 /a 2m\ > b/a, so that y = b 2x/a 2
cuts hyperbola (ii).
(iii)
on each curve.
an extremity of one diameter, then m = b tan (j>/a sec
Hence the conjugate diameter has gradient m' = 6 a /a 2m = 6 sec la tan
and
its equation is y = m'x. It cuts the conjugate hyperbola where
the fact that only one pair can
If P(a sec
<j>,
lie
b tan <fi) is
<J>.
<j>,
<j>
<j>,
6tan0) and
Exercise 18(6)
1
If
e,
e'
1.
l/e a +l/e' 2
OP,
OD
and
its
conjugate, prove
1.
Prove
the following.
2
3
OP 2 ~OD =
2
a 2 b\
intersect
P by the
asymptotes
10 Prove
SP. S'P =
OD 2
[Use Ex. 18
(a),
no.
1.]
If
lies on the conjugate hyperbola, prove that its chord of contact to the
original hyperbola touches the conjugate at the point P' diametrically opposite
to P.
11
THE HYPERBOLA
18.7]
657
Remark
(yff).]
line pair.]
18.7
18.71
xy = c2
Fig. 184
+ y
b
\ai+ b 2 )
xja yjb
0,
h_yx
a
\a*
is
*_y\
_a?
62
a2
62
+ b*)
~
because x\/a2 yf/b 2
1.
Choosing the asymptotes for axes Ox', Oy', let Px have coordinates
x
( i>Vi) Then the lengths of the perpendiculars are x^sino), y'^iiid),
where o) is the angle between the asymptotes, and hence
THE HYPERBOLA
658
i.e.
Px satisfies
Since
o)
xy
a?b
sin o)
[18.72
2 tan 1 (b/a),
/
2tan^>
\l
+ tan
_
Sm 2 &>_
.
o>/
_ / 2b I a \ 2 _ 4a 2 6 2
~[l + b 2la 2) "(a 2 + 6 2
'
J(a
+ 6 2 ).
xy
where
c2
(a 2 + 6 2 ),
and
c2,
is
Example
xy = ax + by + c represents a hyperbola whose asympOx, Oy.
For the equation can be written (x b) (y a) = ab + c; i.e. x'y' = ab + c
after a change of origin.
Thus the locus in ex. (iii) (6) of 16.32 is a rectangular hyperbola, known as
the hyperbola ofApollonius of (h, Jc) wo the parabola y 2 = 4ax. See also Ex. 17(c),
no. 10, and Ex. 18 (a), no. 19.
Any equation
of the form
The equation xy =
c2
can be written
=
so that x
(ct, c/t),
ct
and y =
c/t.
say,
*,
t.
x:y:c
l)
:l:t.
The methods illustrated in 16.22 (2), (3) and 16.24 (1) do not appeal
and can be used in the present case of oblique axes to
to 'gradient',
t x t^
has equation
x + t1 t2 y
tangent at
c(^
+ #a
is
x + thj
2ct.
),
<
THE HYPERBOLA
18.73]
659
Example
the
asymp
Ox,
OQ =
2ct,
OR = 2c/t,
OQR is
^OQ.OR sin
(i)
2c 2 sin a>.
18.73
Fig. 185
When
o)
hyperbola xy
Pxty = C(^1).
Example
The
orthocentre of
triangle inscribed in
a rectangular hyperbola
lies
on
the
curve.
similarly,
and only
chord
if
ta tt
1/<3 4
has gradient
if
MaM* = !
If the given triangle has vertices tlt tt %, then the above work shows that
there is a point i4 on the curve (viz. that for which t 1 t% tz tl = 1 ) such that chord
*i* a _L chord 8 $4 The symmetry of the relation t^t^ = 1 shows that also
Ma J 'a '4 and hhLhh Hence this point tt = l/Ma^s ^s the orthocentre of
,
triangle Ma^s'
It is also clear
the ortho
*x
is
tt
on the curve
+ my + n =
{w}
The line
{to}
Tangents at
Ix
t lt t 2
is
meet at
a tangent
if
THE HYPERBOLA
660
{(t)}
{(o)
*6
{to}
*7
is
(x t y s c 2 )k 2
2
)
is
r sin 6 cos 6
10
11
12
13
3.]
2/
them.
1 5 {(i)} Prove that lines drawn from a variable point P on a hyperbola to any
two fixed points E, F on the curve intercept a constant length on either
asymptote.
16
{(t)}
*17
{(o}
If the tangents at P,
Q meet at T,
prove that
OT bisects PQ.
hyperbola.
18 Prove that no two tangents to a rectangular hyperbola can be perpendicular.
19 If P, Q,
are points on a rectangular hyperbola such that PQ subtends a
rightangle at R, prove that the tangent at
is perpendicular to PQ.
20 Concyclic points.
(i)
Show
hyperbola xy
+ 2gctz + dt 2 + 2fct + c 2 =
0.
and the
circle
21 Use no. 20 (ii) to prove that the common chords of a circle and rectangular
hyperbola are equally inclined in pairs to either axis of the hyperbola.
THE HYPERBOLA
661
A A',
BB',
CC, DD'
lie
ABC
27 (i) Prove that the normal at t to the rectangular hyperbola meets the
curve again at the point l/*8
*(ii) Verify from no. 23 (i) that this is diametrically opposite to the point
where the circle of curvature at t meets the curve again.
.
28 The normal at
the midpoint of PQ
lies
on 4x3y3 + c2 (x 2 y 2 ) 2
Prove that
0.
29 Conormal points.
(i) If the normal to the rectangular hyperbola xy = c 2 at the point
through (h, k), prove c*4 ht3 + ht c = 0.
(ii) If the normals at t lt t
=
t ta t4 are concurrent, prove that t l t 2 t3 t l
,
passes
1 and
0.
30 Prove that four conormal points are also orthocentric, but that four
conormal points can never be concyclic.
distinct
31
c
=}=
0,
variable line passes through a fixed point and meets two given intersecting lines at P, Q. Prove that the locus of a point dividing PQ in a given ratio
32
is
axes.]
its
<f>
NP:NT = b:a.
3 With the notation in 17.23 and cos# = b/a, what is the orthogonal projection of the rectangular hyperbola x 2 y 2 = a2 1 Show that the rectangular
hyperbola xy
4 .4(a,0) and A'( a,0) are fixed points. A variable circle through A and
A' cuts Oy at P, P'. If AP, A'P' meet at Q, prove that the locus of Q is the
rectangular hyperbola x 2 y 2 = a2
5 A variable circle of centre P touches each of two unequal intersecting
circles. Prove that the locus of P is an ellipse and one branch of a hyperbola.
.
THE HYPERBOLA
662
BP
AP
AQ
7 The tangent at
VQ,
= ab.
a hyperbola meet an asymptote at H, K. Prove
passes through the mid point of HK. [Use xy = c 2 oblique axes.]
The tangents
PQ
at P,
to
mh
12 Prove that the chord of contact from any point (xlf y x ) on the hyperbola
x 2 /a 2 y 2 /b 2 = 1 to the ellipse x 2 /a2 + y 2 /b 2 = 1 touches the hyperbola, and give
the coordinates of the point of contact.
variable chord of x 2 /a 2 y 2 /b 2 =
13
that the midpoint of this chord lies on
c2 .
Prove
16 If F, Y' are the feet of the perpendiculars from S, S' to the tangent at
to the hyperbola, prove that SY.S'Y' = b 2 and that triangles SPY, S'PY' are
similar. Hence prove that p/r = (b 2 /p)/S'P, where p =
andr = SP. Deduce
SY
from the
bifocal property that the (p,r) equation of the hyperbola wo the focus
as pole is b 2 /p 2 = 1 2a/r, where
corresponds to the branch enclosing S.
17 If an ellipse and a hyperbola have the same foci S, S' (i.e. are confocal),
prove that they cut orthogonally at each common point P. [The tangents at P
are respectively the external and internal bisectors of angle SPS'.]
18 (i) Prove that the equation x 2 /(a 2 + X)+ y 2 j(b 2 + X) = 1 (a > b) represents
an ellipse, a hyperbola, or nothing according as A > 6 2 a 2 < A < b 2 or
A < a 2 What happens when (a) A = b 2 (6) A a 2 ?
(ii) Prove that the foci of the above conic are independent of A. (When A
varies we obtain a family of confocal conies.)
* 1 9 Prove that through any given point P in the plane ( other than the common
foci) pass two conies of the family in no. 18, one of which is an ellipse and the
,
THE HYPERBOLA
x
663
= a b and y = 0),
2
> 0,
0, /(a
2
/(6 2 ) <
20 Write down the condition for the lines px2 + 2rxy + qy 2 = to be conjugate
a
2
2
2 =
a; /a y /b
1 (ii) the rectangular hyperbola xy = c 2
21 (i) Obtain the distance quadratic for the hyperbola x 2 /a 2 y 2 jb 2 = k.
(ii) Deduce that if chords P QR, P Q'R' of the hyperbola in (i) are drawn
X
1
through Px in given directions, then the ratio P1 Q P x R P1 Q' Px R' is independent of P x and k.
*(iii) Hence prove the analogue of Newton's theorem (17.61, ex.) for the pair
of conjugate hyperbolas x2/a 2 y2 /b 2 = 1.
22 Obtain the equation x 2 /a 2 y 2 /b 2 = 1 of the hyperbola referred to a pair
of conjugate diameters OP, OD as coordinate axes, where OP = a x and OD = 6 X
Explain why the new equation of the asymptotes is x 2 \a\ y 2 \b\ 0. [See
diameters of (i)
Give the centre, semiaxes, eccentricity, foci, directrices, and asymptotes of the
following hyperbolas.
23
V4z
25
36.
24 %{x+l) 2 \(y + 2) 2
I.
664
19
ks'
locus 5 =
19.1
The
19.11
Scheme of procedure
It has now been shown that the general equation of the second degree
s
represents either a conic (in the wide sense of 15.72), a circle, a pair
of parallel
chapter
even when
We
lines,
it will
or nothing;
cf.
methods
a parallel linepair.
begin the chapter by applying to the locus
it is
as a 'conic',
circle or
the general
'
Some
represents a linepair.
remain
when
is
independent of the
first,
and
is
19.12 Notation
To
(a)
We continue to put
(6)
We
8i}
we introduce the
write
Clearly siS
following notation.
sj{
+ xs +f(y4 + y} ) + c.
)
19.13]
The
(c)
by
fy;
denoted
thus
8i
This
665
is
by applying the
'rule of
For example,
sn
s 12
s22
sx
82
Notice that
and
st
Si
(fee,
is
n=
5 12 =
is
represents a line.
P
is the condition for P
or for P
st
s xx , s 12 is therefore as follows:
th condition for
this notation
make
to he on this conic;
to
line s 2
0,
to ue on the line s x
0.
lie
on the
it
provides a suggestive
19.13
first
do Ex. 19
(a),
nos. 13.
of s =
Chord
It
is
linear in
a;
and
y,
^'2
*12*
line. It is satisfied
666
by
Px
for sxx
since s 22
Hence
0.
P2
Letting
Px is s =
at
we
it is
2s
i.e.
0.
19.21
The
dividing
PX P2 in the ratio k
/lxx
\
on
if
s xx as
the limit of
+ kx2
+
Jc
lyx
'
has coordinates
+ ky2 \
+ k )'
and only if
{ax\ +
a(lx x
by
The point
i.e.
satisfied
.
0.
19.2
It will lie
obtain 2s x
Hence
PX P2
[19.2
0,
i.e.
P]P2 and
0.
0.
Its
Px A ^4P2 for
intersection A of the
the points of
Joachim
is
line
the conic.
Since the
Kg 186
no points.
When
negative.
PP
X
k and
I,
i.e.
= 0,
s lz
every point
and
if
Remark in
28
PP
X
= 0,
2 lies
19.24.
in
for
19.23]
19.23 Tangent at
= ks'
667
Px is a fixed point on s =
quadratic for k
is 0.
tangent at
Px has equation s =
(cf.
19.13).
Example*
Contact condition for Ix + my + n =
The tangent at Px has equation
0.
are proportional,
and
for
i.e. if
some A we have
=
=
gx 1 +fy 1 + cXn =
lx x + my 1 + n =
also
By eliminating xx yx
Corollary 1(6) in 11.43)
,
ax x + hyx + g\l
0,
hx1 + byx +/ Am
0,
0,
0.
(using
an obvious extension of
0.
from
IfP1 is a fixed point, the quadratic for k I has equal roots if and only
n s 22 = sf2 This means that P P2 meets the conic in only one point,
i.e. is a tangent from P
Hence s n s 22 = Si2 is the condition for P2 to
x
:
if s
denned by equation
(vii)
45
GPM II
668
= ks'
[19.25
19.25
1.
touch 8
at
2,
wo s =
the line s
(2)
and
on
0.
the line s 1
0.
it follows that A and B divide PiP 2 hi the same ratio
and one externally). Hence the ratio quadratic must have its
and opposite; this is so if and only if s 12 = 0, which shows that
the hypothesis
(one internally
roots k I equal
:
on *t
lies
AB
From
0.
2 is
P wo 8=0.
x
The diagrams in 15.65, with the circle replaced by any conic, show that
some positions ofP1 the locus of P2 in (1), (2) above will be only part of the
line x = 0. We unify these results by making the following definition.
The whole line s t = is called the polar of Pt wo s = 0.
(3)
for
Remarks
(a) If
P
P
(/?)
If
(y)
In
lies
lies
full,
*!
It does
(ax x
if
ax k + hy + g
x.
is
sx
is
P wo
there
linepair s
not exist
unique point
(3)
sx
(ax + hy + g)x 1
0.
Since
the polar of x always passes through the vertex of the linepair (15'53(3)).
(e) Fig. 187 shows the polar of
x as
(i) a chord of contact TT' (Remark (a))
(ii) the harmonic locus, (2);
(iii) the locus of the meets Q of tangents at the ends of chords through
Pi,
(1).
669
THE GENERAL CONIC; S = ks'
Reciprocal property. If the polar of P passes through Pa then the polar
()
(This is valid for a linepair provided P a is not the
of P a passes through P
= 0, and this passes through P if
is
vertex.) For the polar of P wo s =
s 12 = 0; this shows that P lies on s = 0, which is the polar of P a wo s = 0.
In fig. 187, P Q is the polar of P
19.27]
x.
2.
Fig. 188
Fig. 187
19.27
is
P,
2k8 li + U 11
0.
0,
(fig.
188)
The other
Then s 22 = 0,
by
root, given
ratio
2, this
root
is
is s t
= su
Example
Show
the curve s
sx
x lie
on
through
Pr
19.28 Diameters
Let
has equation
452
[19.29
ax x (x  x x ) + h{xx (y  yx ) + yx (x  x x )} + byx (y  y x )
+ g(xx 1 )+f(yyx =
0.
Hence
lies
on the
(ax + hy + g)
By
0,
line
+ m(hx + by+f) =
0.
chords of gradient m.
all
The diameter
m has gradient
a + hm
h + om
a + h(m + m') + bmm'
so that
The symmetry
gradient
in
0.
Two
such diameters
are conjugate.
19.3
The
The
distance quadratic
line
through
x
It
meets
y = y x + r sin 6.
is
(i)
given by
+ b(y x + r sin d) 2
+ 2g(xx + r cos 6) + 2f(yx + r sin d) + c =
i.e.
0,
r 2 (a cos 2 6
0.
(ii)
Px of the points
Px in direction 0.
which he on the
line
(i)
through
Examples
Chord whose midpoint is P y
= which passes through P1 is bisected there, then the roots
of equation (ii) must be equal and opposite, and hence
(i)
If a chord of *
0,
19.4]
= ks'
sm.6
671
(x,y)
is
any point on
'
(axx + hyx
so
(ii)
Segments of a chord.
P Q.Px R = rx r2 =
t
(a)
a cos2 d + 2h
eg
n e + b gin2
6, 6', it follows
that
'
x.
the segments of two chords drawn in given directions through the same point is
constant for all points.
(6) For chords P X QR, P Z Q'R' drawn through P lt P 2 in the same direction 6,
it also follows that
^ D
PX Q.P X R _
P%Q 'P%R
*22
directions.
19.4
Sometimes
it is
yj
0.
If
is
therefore
ax
o
2
+ 2hxy + by + 2fy
2
0.
Fi8 189
Example
Fregier's point.
is a fixed point on a nondegenerate conic s = 0, and PQ is a chord which
subtends a rightangle at O. Prove that PQ passes through a fixed point (the Fregier
point of O wo s) on the normal atO, or else is parallel to this normal.
672
= ks'
[19.4
'
by
15.54. Since
(i)
0,
15.52 (2)
a + (b + 2fm)
0.
If a + b
Exercise 19(a)
Write out in full the expressions for & xx s Xi when s denotes
,
^ + ^1
2 y 4ax.
xycK
(lx
show that
2iS
a^jx 2 + 2hxy+(b^jy =
2
Show
0.
2
_ 2{abh
if
a+b
r2
ax 2 + 2hxy + by 2
10 Writing
and
Use
11
no. 10.
x.
in nos. 1113.
by using
1.
u = qx + hy + g, v ~ hx + by+f,
x,
19.5]
12
{ux (x  xj)
+ vx {y  yx )} =
2
Using no.
(ii)
and hence
1
673
to s\
10,
= 8Sn
is
+ b(y y^} s u
s 11 2 = (s 2s1 +811 $n
.
(i)
that
= ks
at
mean;
<j>
/a
I
6 a cos
a 2 6 2
p,
~z
o sin <p
15 If equation (i) in 19.4 represents lines OP, OQ equally inclined to Oy, use
15.52 (3) to prove h +fl = 0. Deduce that if OP, OQ are equally inclined to the
normal to *
O, or
is
at O, then
PQ passes
19.5
If k
is
and
Ics
Jcax 2
represent the
same
locus; for if
a, b, c, f, g,
h are not
is 1.
then
satisfies one,
all zero,
it
also satisfies
The remaining
five coefficients
(i.e.
the five ratios a:b:c:f:g :h) can be chosen so as to satisfy five conditions,
conditions
number of
be
sets will
finite if
the
conditions,
conies
is finite.
Examples
(i)
(r=
In
drawn through
Pr
1,...,6).
For in general the five ratios a:b:c:f:g:h can be determined uniquely from
the five linear equations
axl + 2hxr yr + by\ + 2gxr + 2fyr + c
so that s
is
(r
1,
. .
.,
5),
674
= ks'
[19.6
Alternatively, a, b,
together with s
An
will
easier
way
0,
(i).
(ii)
conic,
I,
then
For no line can meet a nondegenerate conic in three points (see the footnote
on p. 666); and the only linepair through these five points is I and the line V
which joins the other two.
If four of the points are collinear on I, the conic consists of I and any line V
through the fifth point. If all five points lie on I, the conic consists of I and any
line in the plane. In these two cases of ex. (i) the conic is not unique.
19.6
The equation
19.61
Number
= ks'
s'
+ 2fy + c) = 0,
+ 2(h'y + g')x + (by + 2f'y + c') =
a'x
form
0.
By
no possible values of y
eliminating x 2 only,
we
see
the
number of
0.
possible intersections of
is 4, 2 or 0.
19.62 s = ks'
If s
0, s'
are conies
is
and k
when
and s' =
0, s'
repre
no. 16).
is constant, then s
(d),
ks' is also
// s,
s'
If
Px
is
on
cr,
o~
through
ks'.
ks'
passes through
lt
19.63]
viz.
= sn ls'u
(assuming that
is
not on
<r and s =
(19.5, ex.
= ks'
s'
675
the
0: see
points in
Remark
common,
(i)).
may
arise
when
Px need not
be collinear with any two of A, B, C, D. If A, B,C are collinear, then
s, s' must be of the forms txfi, ay where a =
is the line ABC and
are lines through D (19.5, ex. (ii)). Every conic through
fi = 0, y =
A,B,C,D then consists of a == and a line = Icy through D. Hence
points are collinear. Since
cr
is
x
an
arbitrary point of a,
a(fiky)
sJcs',
any) of s and
s'.
The results in
19.62 apply
s'
degenerate, and
kecfi.
a =
0, fi
with s
(fig.
Fig. 190
(ii)
ka, 2 .
When fi > a
touching s at each of
(if
any)
190).
Fig. 191
in
(i),
we have
a =
any) with a
(fig.
191).
0.
It
is
said
676
(iii)
JcotT,
where r
a conic meeting s at
it
a tangent
its intersections
at the contact of t
is
with s
to s.
(if
(fig.
= ks'
[19.63
any) with a
192).
kaft,
where a,
Fig. 196
ft meet on
s.
Fig. 197
contact with r
(vii) aft
(fig.
196).
(fig.
195).
= kyS is
ft,
y, 8
19.64]
= ky 2
677
(xi) s
&t
is
If S >
a conic touching
t.
19.64 Examples
The 's = ks" principle has already been used for conies in Ex. 15 (/),
(b), no. 1; and in the alternative method in 17.41.
(i) Find the equation of the conic through (3, 0), (0, 2), (5, 0),
The lines joining the first four points in pairs are as follows.
(3,0),
1 =
(0,2):
(0,2), (0,1):
0;
0;
(5,0), (0,1):
 + j,_l =
(3,0), (5,0):
(ii)
Find
is
and
Ix
+ my + n =
0;
0.
(vii))
The required
conic
formed by
the lines
0.
1 9 63, (iii)
its
provided
we choose p and q so
that
apl b qm and
2h
= mp + Iq.
By
and
q,
we have
mp + lq 2h =
eluninating p,
(lx
and
q,
0.
we obtain
+ my + n)x
(lx
+ my + n)y
ax 2 + 2hxy + by 2
ab
2h
is
(by 19.63,
6) if
therefore
0,
678
(iii)
= ks'
[19.64
l + yl _
The points
are
a2
b2
a=
0, /?
where
0,
 cos (0 X +
2)
+  sin (0 X +
2)
= cos(0 3 +
4)
+ sin(0 s +
3,
2,
l5
 cos ^(fa 
2 ),
)eos(0 3 0 4 ).
will
a2
o2
This requires that the coefficients ofx 2 and y 2 shall be equal, and the
xy zero. In general it is not possible for the single number k to
satisfy these two conditions; but the coefficient of xy is zero if
is
circle.
coefficient of
cos (0 X +
2)
sin (0 3
2)
0,
i.e.
<j)
some integer ,
The coefficients of x 2 and
sin
i.e.
for
^^ = k
c os i(0x
?/
2)
cos (0 3 +
4)
if
 ^ sin (0 X +
2)
sin (0 3
4)
( 0!
2)
+ ^ sin
J( 0!
2 J
cos nn ,
*(iv)
Normals
ellipse
The
a2
b2
conic
a2
b2
+ my =
1, 1'x
+ m'y =
of the
passes through the extremities of the chords, by 19.63, (i). If the normals at
these extremities are concurrent, say at (x^y^), then their feet lie on
(a 2 b 2 )xy
by Ex.
17
(c),
Comparing
x a 2x 1 y
coefficients of x 2
 Ml' = 0,
a2
Hence
+ b 2y
if the
of x 2 ja 2 + y 2 /b 2
no. 14.
b2
 kmm' =
0,
+ =
fc
Ix + my = 1, Vx + m'y = 1
= 1. Also see Ex. 19(6),
aHV =
b %mm'
0.
19.65]
= ks'
679
on s = and also
conic passes through four given points A, B, G,
*(v)
through the meet x of tangents at A, B. Prove that it passes through the meet 2
of tangents at C, D.
byP2
on a
If
1, 2, 3, 4, 5,
23, 56;
34, 61
are collinear.
=
s =
Z(34, 61).
The argument holds even when * = is a linepair provided that the sets
5 and 2, 4, 6 lie on different lines; the result is then known as Pappus's
1, 3,
theorem.
19.65 Contact of
two conies
points
AAGD
AACC
A A AD
AAAA
2point contact at
(figs.
AC
A
4point contact at A
3point contact at
192, 193),
(fig.
191),
(fig.
194),
(fig.
195).
when
and cutting
it
By
circle
at P,
we
see
contact at
k and a are chosen so that this conic is a circle, it will be the circle of
curvature at A. It is known (e.g. Ex. 16(6), no. 16 (i); Ex. 17(a),
no. 13) that
680
S = ks'
[19.7
Example
{at 2 ,
y 2at
t.
The required
 4ax =
The
coefficients of
= kt2 1, i.e.
x 2 + y2 _ 2a(3i 2 + 2) x + 4a%  3a 2<4 = 0.
centre of curvature at
is
0.
circle is
curvature
The
circle is
y
xyterm
that the
1/t.
= (x at2 ),
a = x + ty Sat 2
i.e.
has gradient
t is
therefore (2a + 3a 2 ,
2ats ); and
the radius of
8.
Remark. Iif(x) and g(x) are polynomials, then Remark (a) in 6.72
shows that when the curves y = f(x), y = g(x) have mthorder contact
at x = a, they have (m + l)point contact in the sense of the present
section, and conversely. For general curves the concepts are not
equivalent.
19.7
s= ks'
Iff = 0, g = represent any two curves, their intersections (if any) satisfy
every equation which can be deduced algebraically from / = and g = 0. The
new equation need not be of the form/ = kg; and even when it is, k may not be
constant.
do not hold
common points.
Examples
Circle through the feet of conormal points
(i)
The
where
(h, k) is
Hence they
4ax
and
also satisfy
+ (2a h)y 2 =
4a( a + y 2 )(2a + h) 4ax x 2 +y2 (2a+h)x =
x(4kax)
and
This
on a parabola.
is
the required
(ii).
2aky,
2aky,
iky.
0,
(iii) (6)).
19.7]
(ii)
Find
= ks'
681
the equation of conies through the meets, other than the origin, of
and
s'
0,
= y(by + 2f),
= y(b'y + 2f).
x(a'x + 2h y + 2g')
(0, 0)
which
satisfy
0.
This is therefore one conic through the meets of*, s', and does not pass through O.
Clearly
<r
+ ks + k's' =
(i)
Exercise 19(6)
Find
(2,3),
(1,1), (4,0),
(2,3).
2), (5,3).
3 Assuming that ax 2 + 2hxy + by 2 = c and a'x 2 + 2h'xy + b'y 2 =
in four points, find the condition for these points to be concyclic.
(3,
intersect
is
contact
is
parallel to either
or y
0.
2
0. Prove that the chord of
[Let the chord be Ix + my + n = 0.]
x2
circle
has equation s
at each
 + Vv =
[The
kr 2 where t
,
at(,0).
= x a.]
10 Show that a circle can have 4point contact with a parabola only at the
vertex.
682
= ks r
of x 2 ja 2 + y 2 jb 2 = 1,
1 1 Write down the equation of the tangent at the point
and find the equation of the chord through which has the same inclination to
Ox as this tangent. Hence show that the circle of curvature at is
0,
ca
at the point
and
t,
2y at
is
*14 Use example (iv) in 19.64 to prove that if normals at the ends of the chords
0i0 2 and 3 4 of x 2 /a 2 + y 2 /b 2 1 are concurrent, then
2)
COS (0 3 +
4)
sin i(0 x +
2)
sin (0 3 +
4)
COS (0 X
and
*15
(i)
1?
2,
2)
OS (0 3 
4)
2)
cos (0 3 
4)
0.
(2n +
03> 04 to
whose opposite
If p
is
*16
Show
x 2 + (2ah)x
(h,Jc)
to y 2
4ax
lie
on the parabola
(i).]
by 2 (ax 2 + by 2 ~
1)
ac 2 (by 2
l) 2 .
line is
drawn
to
meet a conic at
OQ 2
Prove that Q
lies
on a
OP
1
2
OP' 2
'
(at\,
2at
and
P'.
683
ks'
ba f =
ABGD
0.
AD
DC
is cyclic; AB,
produced meet at E; BG,
6
produced meet
at F. Prove that the bisectors of angles GEB, CFD are perpendicular. [With
for origin, let the pair AB,
bet = px 2 + 2rxy + qy 2 = 0, and BG,
be * = 0.
Express the condition for t = ks to be a circle, and use 15.52 (3).]
AD
DC
8 Find the equation of the rectangular hyperbola through the meets of the
= b/a,
x 2 /a 2 + y 2 jb 2 = 1 and the linepair y 2 = 2x 2 Show that when
this rectangular hyperbola cuts the ellipse orthogonally.
ellipse
PQ lies on 4acy
c2 .
*10 Obtain the general equation of conies through (3,2), ( 3,2), (2, 3),
Show that this family contains a parabola, and find its equation.
( 2, 3).
[Use Ex. 16(e), no. 26 (i).]
The normal to y 2 = 4oa; at the point P(t) meets the parabola again at
and the normal at Q meets the curve again at R. Prove that the equation
of the parabola which passes through P, B and touches the given one at Q can
*11
Q(d),
be written
X{y 2  lax)
where A = (d t)
of y 2 = 4ax and
/4dt.
is
= t 2/t.]
line joining
46
0,
GPM II
684
20
20.1
straight line
chapter
is
to obtain
We
on the straight
line
and
circle, in
material required.
AB2 = r\ + r\  2r
t r2
cos {6 1 
2 ).
Fig. 200
Fig. 199
200)
sin (ft,^)
i
\r x r sin (0  d x )
If
P lies
show
between
and B, then
i.e.
lies
on
AB produced
similarly
by
sin(0 2 0)
rinffflj

r2
or on
rx
BA
same
result holds.
..
20.13]
685
of
the line
(c), is
obvious from
a.
first principles.
Let the perpendicular from O to the given line have length p and
angle a with Ox. If P(r, 6) is any point on the line, then from
triangle ONP we have
(ii)
p = rcos(0 a).
make
This
(cf.
Fig. 201
ordinates (p,oc).
When converted to
Ax + By = C becomes
Aeosd + Bsvcid =
(iii)
Example
Any line perpendicular to
(iii)
Qf
i.e.
i.e.
AaiD.d + Bcos6
=
r
C
r
462
686
20.2
The
[20.2
circle
Let P(r, 6) be any point on the circle with centre C(p,a) and radius a
(fig. 202). From the distance formula (20.11),
a2
Hence the
circle
= CP 2 = p 2 + r 2 2prcos(doc).
has equation
r2
2pr cos (6 a) a 2
2
.
P(r, 6)
Fig. 203
Fig. 202
If C
lies
on Ox, then a =
r2
(b)
If O
lies
on the
so that
(c)
If
lies
the equation
fig.
then p
a2
2a cos (d a) represents a
then
lies
is
2
.
a; the equation
2arcos(6 a) =
on Ox and
is
the equation
2pr cos 6 =
circle,
r2
0;
on the
0,
through O.
circle
circle,
2a cos 6, which
is
becomes
then a
easily written
and p =
a;
down from
203.
20.22
Chord
P P2
by
(2a cos
lt
2a cos d x cos (6 X a)
.
cos (20,
at
p=
=p=
a) + cos a =
2 ),
so that
cos (20 2 a)
is
+ cos a,
(i)
20.23]
and hencef
(20 2 a),
26 1 ex.
a=
i.e.
From
(i)
+ 6%
chord becomes
By letting
n1 da 2
ID
r cos (0
> 0!
687
cos
2>
n
a
2a
cos X cos u %
(u)
2^) = 2a cos 2 6 X
viz.
(iii)
20.23 Examples
(i) Simeon's line.% From any point O on the circumcircle of triangle
perpendiculars are drawn to the sides. Prove that their feet are collinear.
ABG,
Choose O for pole, and the diameter through O for initial line. The equation
of the circumcircle is then r la cos 6.
Let A, B, G correspond to the values = a,/?,y. Then the chord BC has
equation
2^ cos ^ cqs y = r COjg ^Q_p_y^
lie
is
(ii)
= rcos(0 a ft y).
Extension of
ex.
(i).
If
cos/ff, oc+fi).
(2acosa
triangle
(iv)
ABC.
ABC, then A, B, G, D can be selected three at a time in 4 ways, and hence there
are 4 Simson lines wo O corresponding to the 4 possible triangles.
By the Remark in 20.13, the foot of the perpendicular from O to (iv) is
ft+y + S),
lie
on the
is
8,
y + 8+a),
a +ft).
line
rcos(0 a ft y 8).
t The general solution (1.52(3), ex. (iv)) is 2d1 a = 2nn (2d2 a). The sign + is
P 2 to be distinct. The sign gives
inadmissible since we are assuming
1;
a = 9l + 8 nn. Using this value of a and the fact that cos mr = ( 1 ) B , we find that
p = 2o( l) n cos dx cos a and cos (0 a) = ( 1)" cos (6 X 2 ), so that the result
unchanged.
Robert Simson (16871768).
(ii) is
688
[20.3
Exercise 20(a)
1
(6)
Show
holds.
that the area of the triangle whose vertices are A^r^O^), 2?(r2 ,02 ),
{r2 r3 sin (0,0,)+!",^ sin (0 x 0t ) + rx r2 sin(02 #1)}. Deduce the
C(r3 ,0z ) is
condition for A, B,
G to be collinear.
Prove that the perpendicular from (rlf X ) to the line r cos (0 a) = p has
length {p  rx cos (dx a)}.
4 A variable line through a fixed point
meets three given lines at points
Plt P 2 P3 On this line is taken a point P such that
3
OP ~
Prove that the locus of P
is
OP* OP* OP
'
another straight
line.
Write down the polar equation of (a) the circle whose centre is (p, a)
and which touches the initial line; (6) the circles of radius a which touch the
5
(i)
initial line
at the pole.
why
Explain
(ii)
the circles r
acos(0 a), r
gonally.
is
circle r
+ b sin 0.
2pr cos (8 a) +p 2
a cos
circle r 2
a chord of the
0,
prove that
OPQ, ORS
ON be
circle r
2a cos 6 at 6
= a has equation
for
which
11
Two
2a cos {6 a)
circles
respectively.
= a cos 6 + b sin 6
20.3
(i))
by pure geometry.
Choose a focus
8 to the
S of the
all
nondegenerate conies
initial line.
LU have length 21, and let P(r, 6) be any point on the conic.
20.31]
689
In
figs.
Fig. 206,
= SP = e.PM = e.ND
(e(SDSN)
in 204
(e(SDrcoad)
in 204, 205
e{SD + 8N)
in 205
in 204, 205,
ercoad l
e(rcoadSD)
in 206
e(SNSD)
ercoad
in 206,
in 206
A\N
S
1
i
= SL =
Fig. 204
since
Fig. 205
Fig. 206
e.SD. Hence
in figs. 204, 205
206
in
fig.
= 1 + e cos 6;
= 1 + ecosd.
^
(J>r
~^jP
\
>
Fig. 207
If we replace r
fig.
by r and 6 by 6 n
=
I
we
+ e cos (6 n) = 1 e cos 6,
1
i.e.
(i.e. if
= 1 + ecos# becomes
= 1 + e cos d
r
690
as for
= l + ecos# will
remote from S if r is allowed
figs.
[20.32
Ijr
nondegenerate) conic
is completely represented
by
Ijr
+ e cos 6.
Remarks
(a) If
DS
the line
discussion.
represent
all
proper conies
The equation
(/?)
(in
towards S)
by
equivalent to replacing
Ijr
if r is
Ijr
1 ecosd;
= 1 + e cos 6,
+ ecosd
it will
obtained from
negative values of
Ijr
is
taken as
in the
completely
fig.
206, also
same
represent the
conic,
Ijr
= 1 + ecos# and
is Ijr
6 in
the equation
is Ijr
+ e cos 6;
When
1,
is
= 1 + e cos (6 a).
(1)
r,
above
all
r).
completely represents
71
new equation.
l/r
the conic
For,
= l+e cos 8
is
written
= l + cos0 =
2 cos 8 40.
Fig. 208
Fig. 209
It is represented completely
n <
by
6 < n
and
>
0.
20.33]
(2)
sented
When
(3)
is
When
by
given by
<
the conic
1,
n <
e
>
1,
an
is
ellipse
which
691
completely repre
is
6 ^
the conic
tt
and
>
0.
is
>
0;
n sec 1 e <
< n + sec 1 e
and
<
0.
Fig. 210
20.33 Examples
(i)
(a)
If
(r,
6) is
rcostf
Hence this
so that
SD =
l/e.
= SD =
Z/e.
e cos 0.
we
Jl""""
6'
D'
Fig. 211
If e
<
1 (fig. 211),
Fig. 212
then
rcos(7T0)
= SD' = DD'SD
2a
/2a
692
and since
l/a
= 1 e 2 (see Ex.
20(6), no.
1), this
2
\
II = Zl + e2
e\le
ele
/
>
1 (fig.
11
rcos0A = i
r cos 6
is
ee 2 l
= SD' = SD+DD'
= + s
I
2a
11
;(
e2
e\
since l/a
212),
rcostf
[20.33
becomes
1+
2a\
t)
l/
ee 2 l'
*(ii)
its
length
is
Fig. 213
and so
its
equation
is
(20.13)
Since 6
l/*J(e
rcos
+ sin 1  = r~
2 rr*

e)
2n + sin 1
V(e
1)
1
r cos Id sin  1
/
\
i.e.
r cos id sin 1  1
/
\
tt
= j~
2
V(e
1)
(iii) Two conies have a common focus. Prove that two of their common chords
pass through the intersection of the directrices which correspond to that focus.
The
= 1+e cos
and
=
r
+ e'cos(0 a).
20.34]
we obtain
e cos
=V
which
is
the equation of a
e'
cos (0 a),
Since
line.
it is satisfied
also satisfied
by the points
I
693
(r, 6)
which
e cos 6 =
satisfy
V
0,
e'
by the
points
common chord.
(r,
6)
which
Clearly
it is
cos (6 a)
0,
(i).
This
common
= 1 + e cos 6
r
(see
Remark
(fi)
ecos#H
I
e'cos(0 a)
20.34
common chord is
0,
directrices.
cos 0+ 2? sin
^
=
r
let
Then
^?cosa + gsina
= 1 + ecosa
and
pcosfi + qainft
= 1 + ecos/?;
{p e)cosa + 2sina
i.e.
1,
(p e)cos/#+gsin/?
1.
Solving these,
sina
= 7
pe
r
sin/?
sm(ayff)
and
q
*
The chord
^
cos fi cosa
^ =
sm(ayfl)
1#
cosA(a
tf)secA(a/?)
2V + r/
2V
r'
...
sin M<x
2V
n
m
 r/
B).
+ rp) sec *(a
2X
.
'
PQ is therefore
e cos
+ sin
sin J(a
4 /?)},
694
i.e.
e cos 6
[20.34
7.
e cos
# + cos (d a).
see
Examples
(i)
The tangents
at the points
P,
corresponding
to
%(ot+j3).
satisfy
6 + cos (0 a)
e cos
6 + cos (d fi),
cos {6 a)
= cos (0
from which
6 = (a + /?).
Remark. The general solution is d a = 2nn(dfi). The sign +
admissible since P, Q are assumed to be distinct; the sign gives
i.e.
is in
+ Ha+fi).
Since n < a < n and 7r < /? < 7T, also at < (a +
^n, and so there is
no loss of generality by taking n because any point in the plane can be
6
specified
(ii)
If P,
nir
are not
on
and
Hence ST
of ex.
(i),
ST bisects PSQ.
PST = Ha+fi)a =
Hfi a)
TSQ = fi^a+p) =
(/?a).
(fig.
214),
bisects
PSQ
(internally).
PST' = i(a+yff)a =
and
T'SQ'
= fi\{a+P) =
(/?a)
^
a),
so that
(iii) Find the equation of the circumcircle of the triangle formed by three tangents
a parabola, and verify that this circle passes through the focus.
Taking the parabola as l/r = 1 +cos#, the tangents at the points A, B, G
corresponding to 6 = cc, /?, y are
to
cos
+ cos (0 a),
etc.
20.34]
Tangents at
cos (a
ft)
r
i.e.
ft),
+ cos \{ft a) =
by
2 cos
ex.
\ct
(i),
695
and hence
cos \ft,
at the point (Zsec a sec \ft, \{a +/?)). Similarly the other meets are
(Z sec \ft sec
+ y ))
(Z sec
\y sec a, (y + a))
a + ft 4. y\
2
which by 20.21
(6)
'
S.
Fig. 214
(iv) If a variable chord of a conic subtends a constant angle at the focus, then
tangents at the extremities of the chord meet on a fixed conic, and the chord itself
touches another fixed conic.
Let the angle subtended at the focus be 2ft, and take the angular coordinate
of one extremity to be a ft then that of the other is a + ft. The chord thus has
equation
;
cos ft
e cos
696
This
is
a to the
[20.34
conic
= 1 + ecos/fn cos0,
(l)
21 cos {$.
extremities of the chord are
ecos0+*cos(0
ex.
= a and l/r =
(i))
+ cos/?.
This point
lies
+ cos B,
e cos
IseoB
i.e.
an
+ e sec B cos 0,
(u)
Exercise 20(6)
1
Z/(le
= 1 + e cos 6 (e
and
2
)
2
3
>
4= 1)
Z/V(le
<
1,
2
).
1.
4 If P(r1
a) is
an extremity of a
that
J_ + J__2
SP SQ ~ V
i.e.
that
reciwm
is the
harmonic mean of
the segments of
any focal
chord.
5 Prove that the sum of the reciprocals of the lengths of two perpendicular
focal chords of a conic is constant. Express this constant in terms of e and I.
l/r
= 1 + e cos
8 Find the equation of the chord of l/r = l + cos(0 y) which joins the
points 6 = a, /?. Deduce that the tangent at tx is l/r = e cos (6 y) + cos (0 a).
9 Find the
I
common points
V3
r(
of the conies
^3 + cos 0)
and
^3
If the tangent at
697
Zcosa
a
= 1 + ecosa.cosc/.
,
14 Two conies have a common focus, about which one of them rotates. Prove
that the common chord touches a conic which has the same focus as the given
conies, and whose eccentricity is the ratio of the eccentricity of the fixed conic
to that of the rotating one.
15
An
ellipse of eccentricity e
common
latus rectum,
and 8
and a parabola
is
common
their
focus. If
common
tangent
touches them at P, Q respectively, prove that SP, SQ are each inclined to LL'
at an angle sin 1 {(l + e)}. Explain geometrically how the two cases arise.
16. P, Q are the points of the conic l/r=l + e cos 6 corresponding to 6 = ct, fi.
If the tangents at P, Q meet at T, prove that T has coordinates
1
P,
If P,
Q meet
'
H*+P)).
fi)
^
/
ecos#j = cos^ ^).
ecosfljj
.
given
by 6 =
cos4(a fi)
Use
l/r
= 1 + e cos d
>.
a, ft, the
^ ecos#^
chord
is
=cos^0
no. 16.]
20
Show that the tangents at the points 6 = a,/i of l/r = 1 +ecos# are perpendicular if and only if e 2 + (cos<x + cos/?) e + cos(a ft) = 0. [Change the
equation of the tangent into cartesian coordinates.]
*21 Prove that the meet of perpendicular tangents to l/r = 1 + ecos0 lies on
the locus 21(1jr ecostf) = r(l e 2 ). If e 4= 1, show that this locus is a circle
concentric with the given conic (the director circle). What is the locus when
e=
1?
(iii).]
698
Show
*23
l/r
a to
e cos (6
[See 20.14, ex.] If this line passes through the point a, prove that
= esina/(l + ecosa).
Deduce that
the
normal
at
a has
equation
esina
1
+ e cos a r
*24 Prove that the tangent at A(p, a) to the curve r =f(6) has equation
fdr\
cos(0a) (
1
sin(0a)
[If P(r, 6) is any point on the tangent at A, apply the sine rule to triangle OAP,
and use cot $5 = drjrdd (8.14).] Deduce the equation of the tangent to
l/r
1 + e cos
at
<x.
(p,
a) to r
it
iy
+ r\2 + 2r x r 2 cos
and M(r,6)
and
2 }
tan 6
2
n
If
(r 2 ,6 2 ),
jOPr )
is
it
at
is
the midpoint of
?\
sin
+ To sin
AB.
ricos^ + racos^a
P P2
lf
n respectively.
r=l
circles r
2a cos (6 a), r
a)
and
2b cos(dfl).
circle
whose diameter
is
the join of
(&,/?).
cos Qr
rx
(rlt dj)
cos (6
to the circle r
2a cos d
is
X)
(i)
Prove that the locus of the mid points of chords through the focus
of l/r
locus
is
coordinates,
(i)
699
given by
circle
SPT.
9 Find the value of a for which the tangent at a to Ijr = 1 + e cos 6 is parallel
to the initial line. Hence find the locus of the ends of the minor axis of this
ellipse as e increases
from
to
1.
l/r
1.
conic
l/r
= 1 + e cos 6
1 1 Two parabolas have a common focus and their axes perpendicular. Prove
that the directrix of either passes through the point of contact with the other
of their common tangent.
12 Find the locus of the foot of the perpendicular from the pole to the line
= e cos 6 + cos (6 a) when a varies. Interpret the result geometrically.
l/r
Prove that
*/ the foot of
PQ
variable chord
of the ellipse Z/r = 1 + e cos 6 with focus S is parallel
14
to the major axis. If its extremities are the points 6 = 2a, 2fi, prove that
cos (a + /?) + ecos (a /?) = 0.
If the internal bisector of PSQ meets PQ at K, prove that
QTr
SK
=
lies
2SP.SQ
spTsQ
coaiPSQ
'
MN
47
GPM
II
700
21
Coordinates in space
Fig. 216
On
Fig. 217
sense along
it
turned in the sense from Ox towards Oy, label this direction Oz.
We use
21.12]
701
and that
wo
x, y, z, the points A, B, O
on the axes can be determined, and the box then completed. The
corresponding point
of coordinates x, y, z
numbers
x, y, z
wo
and
to
P in space.
Zj),
and so
book
x is
on.
the plane
xOy by
ON
X
Fig. 218
Fig. 219
472
702
[21.2
xON, measured positively in the sense from Ox towards Oy. P is then determined by the triplet of numbers (p, z), called the cylindrical coordinates of P.
<j),
(2)
p and z can be replaced by polar coordinates (r, 6), where r = OP and 6 is the
angle zOP, measured positively from Oz towards ON. P is then determined by
(r, 6, 0), the spherical coordinates of P.
4*
Fig. 220
If the sphere in fig. 220 represents the Earth and xOy, xOz are the planes of
the equator and Greenwich meridian, then 6 is the colatitude and <p is the
longitude of P.
Fundamental formulae
21.2
PN PN
xOy be
let
The coordinates of
2
2
and of N2 are (x2 ,y2 ,0).
X,
are (#1,^,0),
wo
xi> Vi)>
(*2> y*)>
so
N N\ = {x
X
Through
since
it
P2 B
lies
meets PX
X,
axes Ox,
are
NN
in the plane of P N and P N
draw
PR
parallel to
say at R.
X;
2,
Fig. 221
P P2 B,
P P\ = P R* + P2 R* =
triangle
their
Oy
tnat NX N2 is given by
cartesian coordinates
{z x
z 2 f + N2 N\
= (*i  z 2 2 + (*i  *2 2 + (2/1  y%Y>
)
and hence
P P = <J{(x X
2)
+ (yx 
f + (z x  z2
2
)
}.
O is J(x + y 2 + z 2 )
2
21.22]
703
NN
NN
Fig. 223
Fig. 222
If the division
is
internal
z
(fig.
z,
z 2 z
from which
I*
^ ++ k^
Za
.
Similarly,
we
find that
zOx, that y
{lyx
Alternatively, since
and
has the same x and ^coordinates as P, the last two results follow
If the division
is
external
z
z
from which
(fig.
223)
(15.14).
we have
zx _ k
z% V
5
= \
l
2
.
704
[21.22
P P2 in
X
has coordinates
llxx
+ kx %
l+k
lyx
+ ky %
l+k
'
+ kz^\
l+k )
lz x
*
Examples
(i) The line joining .4(1,2, 3) and J5(3, 0, 2) meets the coordinate planes
xOy, yOz, zOx at P, Q, B. Find the coordinates of these points, and the ratios in
which they divide AB.
The point
dividing
AB in the ratio k
ll
+ Zk
[l + k
It lies in the plane
has coordinates
3Z + 2AA
21
T+k'
'
)'
+k
P has coordinates
0),
and divides AB
and only if its
31.
Hence
(ii)
The four
the
the
corresponding face.
If the vertices are
etc.
The midpoint
P P P3 P
lt
2,
if
and
'
+ xs y1 + y 2 + yz
,
/'
'
[xi+xz
3
\
Zi+g 2\
3/1+3/2
z1
yx
z1
3; its
coordinates are
+z 2 + zz + zi\
)*
'
Hence
)'
1).
+g 2 + g3 \
'
+ Vi + Vz + Vi
(cf.
(x1 + x2 + xa + xi
\
C4
let
M of P, P, has coordinates
3,
and the
results
follow.
Exercise 21 (a)
If the negative teaxis is labelled Ox', and similarly for Oy', Oz', give the
signs of the coordinates of points in each of the octants
1
Oxyz,
Ox'yz,
Ox'y'z,
Oxy'z,
Oxyz',
Ox'yz',
Ox'y'z',
Oxy'z'.
(iv)
=
=
0;
(ii)
(v) z
z;
=
=
0;
x;
(iii)
(vi)
705
= 0;
= =
yt
3 In fig. 217 prove that PA = <J{y 2 +z*), and give expressions for PB, PC.
4 Prove that the points (2,3,7), (5, 1, 5), ( 10,8,7) are the vertices
of an isosceles triangle, and find the length of the base.
Show that ( 1,
5
(13,
 1,
6 For A{Z,
and
2,
7 Prove that the points (1, 3, 4), (2,  1, 3), (7, 1, 8), (6,
vertices of a parallelogram, and find the lengths of the diagonals.
5, 7)
are the
8 Find the ratios in which the coordinate planes divide the join of the points
(3,
5,
and (4,
2)
7, 1).
Show that the point (6, 2, 0) lies on the line joining ( 4, 6, 2) and ( 9, 8, 3),
(7,
5,
2).
10 Prove that the lines joining the midpoints of opposite edges of a tetrahedron are concurrent at the centroid of the tetrahedron. [With the notation
of 21.22, ex. (ii), X 8 and 3 4 are a pair of 'opposite edges'.]
PP
PP
Change of origin. If new axes Px x', Px y', Px z' are taken through Px parallel
to Ox, Oy, Oz respectively, show that the new coordinates of P(x, y, z) are
{x',y',z') where x' = x x y' yy lt z' z z x
11
12
217, prove
Obtain the formula for the distance X
(i)
PP
(ii)
to
OP8 = x*+y2 + za
2
origin
Pv
13
Express
x,
is
y in terms of p and
<j>,
and conversely.
(i)
k;
(ii)
<j>
a;
(iii)
a?
=k
<j>
*15 Express
* 16
If k,
cc,
(i)
(iv)
21.3
=
=
x, y, z in
are constants,
k;
cc
and
= /?;
(v)
=
=
(iii)
cc;
k and
= /?;
(vi)
=#
k and
we
is
Through
draw a line
denned to be
P towards Q.
OA of unit length parallel to and in the same
706
sense as
PQ
(sensed) line
(fig.
224).
We
OA
call
[21.32
PQ.
OA
will
be determined
customary to write
I
m = cos
cosa,
n=
/?,
cosy,
and
call
{I,
parallel to
+ m 2 + rc 2 =l.
(i)
Fig. 224
Fig. 225
Remarks
(a) If the line is described in
a),
has n
i.e.
(/?)
(77
is
zOx has
i.e.
is
the point
y)),
m, n).
(it
I,
0.
is
another
line,
through 0, then B
be
is
B towards S, let
OB is the corresponding unit ray
{1',
m', n'}. If
the point
(V, m',n').
21.33]
707
AB = 22cos0.
2
AB2 = (l
AB
2
,
1.
Comparing
we find
cos
(ii)
if I = V, m = m' n =
= m', n = n'.
and only
= V,
and 'antiparaUel'
n'\
if
and only
if
is
by a
set of
m, n.
of the line. For
I,
\n,
f$
\n,
y=
\2 (p 2 + q2 + r2 ) =
0.
Xr
1.
Hence when the direction ratios p:q:r are given, the direction cosines
can be written down as
2
2
{pN(P +q* + r ),
In any problem
it is
qlyl(p*
+ q* + r 2 r^(p 2 + q2 + r2 )}.
),
b'
'
708
[21.33
meant to be
direction cosines or
direction ratios.
We remark
line.
Examples
(i) If 6 is the acute angle between two lines having direction ratios p:q:r,
p' q' : r', then by equation (ii)
:
a
COS U
V(2>
+? +
r*)
V(P'
+ ?' 2 + r'*)
(ii)
cos a
The
(iii)
P is
x/OP,
or antiparallel
'
The
lines
OP
if
q r
:
= p'
OP, OQ have
direction ratios I:
q' : r'.
projections of
cc
l9
y 2 yi,
z%
z.
z x Thus
.
1.
mm,
V m'
:
n'.
Find
direction
Let
and only
if
'
'*
p:qir
r.
and
1.43 (2)),
Vp + m'q + n'r =
condition
0.
we have
= mn'm'n:nl'n'l:lm' l'm.
Exercise 21(6)
For .4(3,
2, 5),
B(
3),
 1,
B(3,
6, 4),

2,
(
through
b, c.
5(6,2,9),
9
From P{x, y, z) a
direction cosines
where
=5 Ix
1, 3),
0(5, 1,1).
{I,
p = ON, and
+ my + nz.
drawn
to
m, n} at N. Prove that
perpendicular
write
down
is
meet the
21.4]
*1 1
p3
709
N to OP in no.
from
9,
find the
prove that
'
Pi
?i
*i
p%
q%
r2
?s
rs
Pa
0.
true. [Use
is
Theorem II of 11.43.]
*13 If Ox, Oy, Oz and Ox', Oy', Oz' are two systems
5f, Z?" of rectangular axes with the common origin O,
let Ox', Oy', Oz' have direction cosines (^.w^Wj),
(l 2 ,m 2 ,n 2 ), (l ,m ,n ) wo S?.
3
3
3
(i) Prove that
Z*
and
i
a is
+m m +w
a
2 W3
l3 l1
+m m +n
3
m m
3 ),
n1
(n ly
n 2 n3
,
are
Fig. 226
+ + ll = mf +m+m = n*+n+n* =
and
m1 n 1 +m2 ni +m
14
Rotation of axes.
If
has coordinates
n3 =
(i)
x'
[x'
n^+n^+n^ = m
l1
(x, y, z)
= ll x + m1 y+n1 z,
y'
wo 7 and
+ l 2 m 2 + l 3 m3 =
= l2 x+m 2 y+n2 z,
(ii)
z'
wo Sf",
0.
prove that
= l3 x+m3 y+nz z.
9.]
l1 x'
+ la y' + l3 z',
y', z'
by llt
why
l% , l3
in
x 2 +y 2 +z*
(i),
(ii).]
verify this
algebraically.
21.4
The plane
by
710
[21.41
(pl,pm, pn).
P(x, y, z)
z) is
by
m, n},
{I,
N(pl,
pm, pn)
and so
+ (y pm) m + (z pn) n = 0,
Ix + my + nz p(l 2 + m 2 + n 2 = p.
(x pi)
i.e.
The equation
Ix
which
Fig. 227
+ my + nz = p,
is satisfied
by the
z.
Ax+By + Cz + D = Q
is
Let Pj and
equation; then
W6get
first
of these
0,
Ax2 +By2 + Cz z +D =
H^) +B ^)P+cP ^) +D
X
in the ratio k
Ax + By + Cz + D =
satisfy
By
is
arbitrary, this
(i)
by
Since k
0.
means that
'
:
'
also satisfies
0.
all
(ii)
P^
(ii).
definition a plane
line joining
equation
(ii)
is
a locus in space
it lies
(or surface)
entirely
some plane.
0.
711
THE PLANE AND LINE
This shows that the direction A B C is perpendicular to the direction
x1 x2 :y 1 y 2 :z 1 z2 i.e. to PX P^ for any line PjP2 in the plane.
Hence the direction A B C is perpendicular to the plane.
21.43]
independent conditions.
The planes
same plane
and only
are parallel if
A'
B'
if their
C'
A=B = C =k
normals are
parallel,
i.e.
(say) '
0.
These parallel planes will be the same plane if and only if they have
z ); and then
Ax + By + Cz + D =
0,
Jc{D) + D'
so that
k{Ax
0,
if
i.e.
+ By + Cz + D' =
)
0,
D'/D.
and only if
'
_ IT _ (7 _ >
A ~ B ~ C ~ D'
^1/
i.e. if
will not
be equal.
I:
lx+my + nz = p,
,
even when I, m, n are direction ratios and not direction cosines. Since
the plane passes through {xx ,yx> z^,
lxx
By
+ myx + nz 1 = p.
subtraction,
l{x
0.
(iii)
712
(x, y, z)
is
some plane
therefore represents
tion
[21.45
l:m:n
It is linear in x, y, z
(iii).
(21.42(1)),
by
(21.42(2)). It is satisfied
{xx ,y x ,z x ).
and
direcis
the
required equation.
21.45 Plane
P,P2 P8
Method
Suppose
equation
1.
is
has normal
.
n; then
by 21.44 its
0.
l(x2
 xx + m(y 2  yx + n(z 2  zx ) =
l(x3
 x x + m(ys  yx + n(z3  z x =
Ehmination of
I,
m, n from
x ~ xx
this
yVx
x x y%~ y\
xz ~ x i Vz~ Vl
x%
P1 P2 P3
0.
~ zi
z%
~ i\
ZZ
~ Zl
Ax + By + Cz+D =
This
equations gives
2.
and
Method
 xx + m(y  yd + n(z  zx =
,
l(x
The points
this plane
is satisfied
by Px
0.
and Ps so
,
Ax x + By x + Czx + Dz=0,
Elimination of A, B, G,
0,
Ax3 + Byz + Cz z + D =
0.
gives
xx
y1
zx
y%
^3
Vz
%3
21.46J
713
on the
coordinate axes
The required plane passes through the points (a, 0,0), (0,6,0),
(0, 0, c). Any plane through (a, 0, 0) has an equation of the form
l(x
This
is satisfied
by
a) + my + nz =
0.
al + bm =
Hence
al
al + cn =
0,
= bm =
0.
cn,
l\m\n = \\\.
so that
b c
i.
by
the
Ax+By + Cz + D =
0,
A B
the planes
by
is
therefore given
cose/
In particular,
C,
0,
(i))
'
+ B'* + C" 2
AA' + BB' + CC =
0.
i.e.
A:B:C = A':B':C'.
Exercise 21(c)
Find the locus of points equidistant from (2, 1, 3) and (1, 3, 2).
2 Find the equation of the plane which bisects AB and is perpendicular to
CD, where A{Z,  2, 1), J?(5, 4,  2), (7(2, 4,  3), D{  1, 3, 2).
1
3
is
0)
714
[21.5
 1),
(2, 2,
Find the equations of the two planes through (0, 4, 3) and (6, 4, 3),
other than the plane through O, which make intercepts on the axes whose sum
5
zero.
is
6 If ^1(3, 0, 0), B(2, 3, 0), 0(1, 1, 1), find (i) the angle between planes
(ii) the angle between line BC and plane OAB.
OBC,
OAB;
At
XO Y =
three points O,
X,
respectively.
q r has equation
:
z1
Xi
yyx
Xo
which is
0.
parallel to
p'
q'
r'
0.
13 Sides of a plane. Find the value of k I if the point dividing PlP2 in the
= 0. Deduce that x a are on the
ratio k I lies in the plane Ax + By + Cz +
same or opposite sides of this plane according as Ax1 + By1 + Cz1 +D,
Ax 2 + By z + Cz 2 + have the same or opposite signs.
:
P P
,
21.5
The
line
A straight line in
planes
(6)
(c)
by
(a)
of direction ratios.
in direction
l:m:n
Let P(x, y, z) be any point on the line; then a set of direction ratios
x x1 :yy 1 :z z 1 Since l:m:n is also such a set of
x x1 y yx
:
z1 =
n.
21.52]
715
These equations are satisfied by any point of the line, and are therefore
the equations of the
line.
'
xxx = zzx
xxx = yyx
"
represent
two such
We refer to
(i)
'
planes.
Example
Equation of the line
PP
X
PX P% are
xx x % yx 1/ 2 zx z 2 Hence the
y yx
xxx
2/1
line
zzx
2/2
x = xx + M,
(i)
above equal to
= y1 + \m,
zx
A,
we have
+ Xn.
(ii)
is
parallel to
= 0. The equation
(i)'
x = Urn
z t = 0.
then means
and
(i)',
with a similar
This
is
We
continue to write
=  =
as alternative to
48
gpmii
716
[21.52
Examples
Find
(i)
A(3,
7,
2)
to the line
x 10 _ y+8 _ z + 5
~
~~~T'
3
the coordinates of its foot N.
The parametric equations of the
and find
line,
its
= 3A+10,
= 4A8,
= A5.
AP are
3A + 7:4A15:A3.
AP
3:4:
will
1) if
and only
line
if
from which A
is
obtained
AN
22 + 3 2 + 62
49,
AN = 7.
.\
(ii) Find the image of the point P(3,  5, 2) in the plane 2x + ly  3a = 27.
By definition, the image P' of P in the plane is such that PP' is perpendicular
parametrically
or ^
J
P is
xZ
_
~2~~
2A + 3,
3,
the equation of
_ z2
~ 3'
y+5
7
= 7A5,
oi.o
=  3A
+ 2.
This point will be P', the image of P, provided that the midpoint of P'P
on the plane, i.e.
2(A + 3) + 7(A5)3(A + 2) = 27,
lies
from which A
(iii)
Find
the
2.
image of the
(7, 9,
 4).
line
3
x2
__ _ y__ 1 _ z +
5
in the plane x 2y + 3z = 5.
find (a) the point Q where the line and plane meet; (6) the image P' of
the point P(2, 1, 3) of the line; (c) the required line as the join of P', Q.
(a) The parametric equations of the given line are
We
a;
The line
3A + 2,
= 2\+l,
(3A + 2)
i.e.
hence
is (5, 3, 2).
for
= 5A3.
which
 2(2A + 1) + 3(5A  3) =
5,
21.53]
(6)
P'(4,
3,
(c)
(ii)
717
3).
g2
x5
___ _
__ _ yS
_
(iv)
coplanar.
to be
lines
xa
yb
y b'_z c'
m'
n'
x a'
'
for
some
A,
for
some
fi.
(a
and
also
(a'
Hence there
exis