Anda di halaman 1dari 455

CAMBRIDGE

UNIVERSITY

PRESS

CI
ffl

a
F.

GERRISH

PURE MATHEMATICS
AN

c
m
>

VOLUME

II

ALGEBRA

TRIGONOMETRY
>
H
n

COORDINATE GEOMETRY

This two-volume text-book on Pure Mathematics has been designed to cover the whole of

the mathematics required for Part

London

B.Sc. General

of the

Degree and for any

first-

year degree courses containing subsidiary pure

mathematics.

the only book of

is

It

addressed

primarily

student.

also covers

It

matics required

Diploma

the

to

its

kind

General Degree

most of the Pure Mathe-

by the recently introduced

Mathematics of the Mathematical

in

Association.

This book does not assume a complete mastery


of earlier work, and

much

of

it

school

pupils

preparing

for

suitable for

is

the pure mathematics required

by grammar

scholarships

in

Natural Sciences, as well as for the mathematics

However, though

specialists.

aimed to supply a
students at these

Mr

Gerrish has

need of examination

real
levels,

he has treated his

subject in a serious fashion and this book

cram book

sense a

goes a

little

for examinations.

is in

no

Thus he

beyond the requirements of the

examination syllabus.

By a

natural division the subject-matter

conveniently into

two volumes which,

falls

despite

occasional cross-references, can be used inde-

pendently as separate text-books:


I.

Calculus and diiVerential equations, with appli-

cations

to

topics such as areas,

centres of gravity,

moments of

arc-lengths,
inertia,

and

geometry of plane curves.


II.

Algebra and convergence, trigonometry and

coordinate geometry of
sions, in

when

two and fhree dimen-

which calculus methods are

illustrated

instructive.

Finally

the

book contains

many

sets

of

exercises for the reader', often with hints for


solution.

Each chapter includes a Miscellaneous

Exercise for revision purposes.

Answers are

iho provided.

15 s

One or 1 600+ items f fee


from Ihe Internet Archive
uploaded by artmisa

-8

sm

-a my -mi

-3. SEP.

NAME

This book

is

due for return on or before the

last

date

shown above.

-0 4^

8EDGLEY PARK COLLEGE,


PRE8TWIGH.

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.

CAMBRIDGE UNIVERSITY PRESS


1 <><><>

Printed in Great Britain at the University Press, Cambridge


{Brooke Crutchley, University Printer)

CONTENTS
General Preface

p a g e xy ji

Preface to Volume

II

xi x

References and Abbreviations

xxi

Chapter

363

10.1

10. Algebra of Polynomials


The remainder theorem and some consequences

10.11

363

Long

division; identities, p. 363. 10.12 The remainder


theorem, p. 364. 10.13 Factorisation of a polynomial;

'equating coefficients

10.2

',

p.

366

Exercise 10(a)

ggg

Polynomials in more than one variable

370

10.21 Extension of the preceding results, p. 370,


10.22 Symmetric, skew and cyclic functions, p. 371

Exercise 10(6)
10.3

373

Polynomial equations relations between roots and


:

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

10,41 Further examples,

equations, p. 380.

10.5

two

Exercise 10 (d)

382

The

382

10.51

h.c.f, of

The

theorom,

two polynomials

h.c.f. process, p. 382.

p. 385.

10.53

10.52

An important algebraic

Theory of partial

fractions, p. 386

Exercise 10(e)

388

Miscellaneous Exercise 10(f)

388

Chapter 11. Determinants and Systems of


Linear Equations
11.1

Linear simultaneous equations

Two

equations in two unknowns, p. 391. 11.12 Three


equations in three unknowns, p. 392, 11.13 Structure of the
solutions, p. 392
11.11

391
391

CONTENTS

VI

11.2

page 393

Determinants

11.21 Determinants of order 2, p. 393. 11.22 Determinants of


order 3, p. 394. 11.23 Other expansions of a third-order
determinant, p. 395. 11.24 Properties of A, p. 396.
11.25 Examples, p. 397

11.3

11.4

Exercise 11(a)

399

Minors and cofactors

401

11.31 Definitions

and

alien cofactors, p.

403

notation, p. 401.

11.32 Expansion by-

Determinants and linear simultaneous equations


The

404

when A = 0:
11.43 Homogeneous

case

11.41 Cramer's rule, p. 404. 11.42


inconsistency and indeterminacy, p. 406,
linear equations; solution in ratios, p. 407

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

The binomial theorem


12.12 Properties of the binomial expansion, p. 420.
12.13 Examples, p. 420

12.2

Exercise 12(a)

422

Finite series

422

12.21 Notation
for

summing

binomial

and

12.22 General methods


12.23 Series involving
12.24 Powers of integers, p. 426.

definitions, p. 422.

finite series, p. 423.

coefficients, p. 424.

12.26 'Fraction' series, p. 429.


trigonometric series, p. 431. 12.28 Mathematical

12,25 'Factor' series, p. 428.

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

infinite series, p. 436,

12.32 General

CONTENTS
12.4

vii

Series of positive terms

page 43^

Comparison tests, p. 439. 12.42 d'AIcmbert's ratio


1243 Speed of convergence of a aeries, p. 446.
12.44 Infinite aeries and infinite integrate, p, 447
12,41

test,

p. 443.

12.5

Exercise 12(e)

442

Exercise 12(f)

446

Exercise 12(g)

450

Series of positive

and negative terms

451

12.51 Alternating signs (theorem of Leibniz), p. 451,


12.52 Absolute convergence, p. 452, 12.53 The modified ratio
test, p. 456.
12.54 Rogroirping and rearrangement of terms of

an

12.6

infinite series, p.

457

Exercise 12(A)

459

Maclaurin's series

460

Expansion of a function as a power series, p. 460.


12.62 Expansion of c x sinx, cosx, log(l +x), (1 +x) m , tan -1 a?,
12.63 Note on formal expansions, p. 464
p. 461.
12.61

Exercise 12
12.7

465

(i)

Applications of the series in 12,62

466

12.71 Binomial series, p. 466. 12,72 Exponential series, p. 470.


12.73 Logarithmic series, p. 472. 12.74 Gregory's series and

the calculation of n, p, 476

12.8

Exercise 12(j)

468

Exercise 12(A)

472

Exercise 12(1)

475

Scries

and approximations

476

12.81 Estimation of the error in s = s nt p. 476. 12,82 Formal


approximations, p. 479. 12.83 Calculation of certain limits,
p. 480

Exercise 12 (m)
Miscellaneous Exercise 12

481
(w)

482

Chapter

13. Complex Algebra and General


Theokt of Equations

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

13.14 Third stage: logical development, p. 489.


13.15 Importance of complex algebra, p. 492. 13.16 Further
possible generalisations of 'number', p. 493

page 493

Exercise 13 (a)

13.2

The modulus-argument form of a complex number


13.21 Modulus and argument, p. 494. 13.22 Further
definitions, notation, arid properties, p. 490. 13.23 The

494

cube

roots of unity, p. 497

493

Exercise 13(6)

13.3

Applications of the Argand representation


13.31 Geometrical interpretation of
13.32 Constructions for the
p. 499.

490

modulus and argument,


sum and differ once of z v z v

13.33 The triangle inequalities p. 501.


13.34 Constructions for the product and quotient of & XJ z, p. 502.
13.35 Harder examples on the Argand representation, p. 504

p. 500.

500

Exercise 13(c)

13.4

Factorisation in complex algebra

507
3

13.41 'The fundamental theorem of algebra, p. 507.


13.42 Boots of the general polynomial equation, p. 507.
13.43 Principle of equating coefficients, p. 509. 13.44 Repeated
roots and the derived polynomial, p. 509. 13.45 Equations

with

13.5

'real' coefficients;

conjugate complex roots, p. 510

Relations between roots and coefficients


13.51

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

Factorisation in real algebra

Roots of the general polynomial equation, p. 515.


13.62 Location of roots in real algebra, p. 510
13.61

520

Exercise 13(e)

13.7

Approximate solution of equations

(further methods)

521

The method of proportional parts, p. 521. 13.72 Horner's


methods p. 522. 13,73 von Graeffe's method of root -squaring,
13.71

p.

523

Exercise 13 (/)

524

Miscellaneous Exercise 13 (g)

52o

Chapter

CONTENTS
de Moivre's Theorem

14.

ix

astd

some

Applications
14.1

p.

528

The values of (eos# + i sin

14.12

14.2

page 528

de Moivre's theorem
14.13 Examples,

p. 529,

530

Exercise 14(a)

532

Use of the binomial theorem

534

14.21 oos m

sin # in terms of multiple angles, p. 534.


14.22 oasnd, sinnfl, tanri.0 as powers of circular functions,
14.23 tan (O L + 2 + ...+& n ), p. 537
p. 535.
fl

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.

14.34 sin n(9, p. 542

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

between roots and

coefficients, p.

546

Exercise 14(d)
14.5

547

Finite trigonometric series:


14.51 Cosines

and

summation by C + iS

sines of angles in A. P., p. 548.

548

14.52 Other

examples, p. 549

Exercise 14(e)
14.6

Infinite series of

550

complex terms. Some

single -valued

functions of a complex variable

550

and absolute convergence, p. 550. 14.62 The


infinite c,.i\, p. 551.
14.63 The exponential series, p. 552.
14.64 The modulus -argument form of expz. p. 554.
14.61 Convergence

14.65 Euler's exponential forms for sine, cosine, p. 554.


14.66 Examples, p. 555. 14.67 Generalised circular and
hyperbolic functions, p. 557, 14.68 Relation between circular

and hyperbolic functions: Gsborn's

rule, p.

558

Exercise 14(/)

556

Exercise 14 (g)

560

Miscellaneous Exercise li(h)

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

of the equation of a straight line

570

through [$v jfj, p. 570.


15.21 Line with gradient
15.22 Gradient form, p. 570. 15.23 General linear equation
Ax + By + C = 0, p. 571. 15.24 Intercept form, p. 571.
15.25 Line joining P lt P 2 p. 572. 15.26 Parametric form for
the line through [v lr y ) in direction 6 p. 572. 15.27 General
parametric form, p. 573. 15,28 Perpendicular form, p. 573
,

15.3

573

Further results
15.31 Sides of

a point from a

line, p.

line, p. 573.

15.32 Perpendicular distance of

575

575

Exercise 15(a)
15.4

Concurrence of straight

576

lines

15.41 Lines through the meet of two given lines, p. 576,


15.42 Condition for concurrence of three given lines, p. 577

15.5

Exercise 15(6)

578

Line-pairs

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 line-pair
to the meets of the line
$ = 0, p. 581.
15,54 Line-pair joining
Ix + my 1 and the locus s = 0, p. 584

15.6

Exercise 15 (c)

585

The

586

circle

15.61 General equation of a circle; centre and radius, p. 586.


15.62 Circle on diameter I\P 2r p- 586. 15.63 Tangent at Pv
15.64 Chord of contact from P 1? p. 587.
p. 587.
15.65 Examples polar, p. 589. 15.66 Orthogonal circles, p. 591
;

15.7

Exercise 15(d)

592

Conies

594

15,71 Definitions, p. 594. 15,72 The equation of o very conic is


of the second degree, p. 595. 15.73 Change of coordinate axes,
p, 595,

15.74

K eduction

of 3

to standard forms, p. 598

Exercise 15(e)

600

Miscellaneous Exercise 15 (/}

601

Chaptee
16.1

The

CONTENTS
The Parabola

16,

locus y 1

page 603

4ax

003

Focus -directrix property,


representation, p. 603
36.1

XI

p. 603.

16.12 Parametric

Exercise 16(a)
16.2

go 5

Chord and tangent

605

Chord Px P 2 p. 605. 16.22 Chord t^, p. 606.


16.23 Tangent at F 1} p, 607.
16.24 Tangent at the point
16.25 Tangency and repeated roots, p. 008.
p. 607.
16.26 Examples, p. 609
16.21

>

t,

Exercise 16(6)
16.3

610

Normal
p.

612

Normal at the point

16.31

t,

p. 612.

16.32 Conormal points,

612

Exercise 16(c)
16.4

614

Diameters

615

16.41 Corporal definition, p. 615.

16.42 Diameters of

parabola, p. 615

Exercise lG(d)

616

Miscellaneous Exercise 16(e)

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.

17.12 Focus- directrix property of

17,13 Focus-directrix property of

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.

Other ways of obtaining an

17.17 Circle

and

ellipse, p.

622

ellipse

623

17.21 Auxiliary circle, p. 623, 17.22 Focal distances, p. 624.


17.23 Orthogonal projection of a circle, p. 625, 17,24 General
properties of orthogonal projection, p. 626

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

Chord and tangent

page 630

Chord PjPj, p. 630. 17.42 Chord 0$, p. 631.


Chord
17.44 Tangent at P p. 632.
p. 63 L
17.45 Tangent at 0, p. 632. 17.46 Tangent at t, p. 632.
17.47 Examples, p. 633
17.41

17.43

Exercise 17 (b)
17.5

Normal
p.

636

Equation of the normal,

17.51

17.6

634

p. 636.

17.52

Conormal

points,

636

Exercise 17(c)

637

The

637

distance quadratic

17.62 Chord having midpoint I\, p. 638. 17,63 Diameters,


17.64 Conjugate diameters, p. 639
p. 639.

Exercise 17

C41

(d)

Miscellaneous Exercise 17 (e)

Chapter
18.1

18.

The Hyperbola

The hyperbola -g fj
18.1

Form

18.2

The

1;

645

asymptotes

645

of the curve, p. 645.

definition, p. 645.

18.14

642

18.12 Asymptote: general


18.13 Asymptotes of the hyperbola, p. 646.

bifocal property:

SP-S'P =

2, p. 647

Properties analogous to those of the ellipse

+
tir

18.3

18,32 The point


18,34 Another algebraic

18.33 The point 0, p. 650.


representation, p. 650

18.5

Parametric representation
18.31 Hyperbolic functions, p. 649.

18.4

D2

648

649
t,

p. 650.

Chord, tangent, and normal

651

Exercise 18(a)

651

Asymptotes further properties

653

an asymptote meet the hyperbola on!y


once, p. 653. 18.52 The equation of the tangent at Pi tends to
the equation of an asymptote when P -* co along the curve,
p. 653. 18.53 The family %ya*-y*fb 2 = ko hyperbolas has the
same asymptotes for all k, p. 654
18.51 Lines parallel to

18.6

The conjugate hyperbola


18.61 Definitions, p. 654.

Exercise 18(5)

654

18.62 Conjugate diameters, p. 655

656

CONTENTS
18.7

Xlii

Asymptotes as (oblique) coordinate axes

18.71

xy

18.73

The rectangular hyperbola,

2
,

p. 657.

page 6o7

18.72 Parametric representation, p. G5S.


p.

6S9

Exercise 18(c)

659

Miscellaneous Exercise 18

Chapter
19.1

The General Conic;

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.

19.12 Notation, p, 664,

666

Joachimsthal's ratio equation.

666

19.21 The ratio quadratic for $


19.22 Sides of
0, p. 666.
eonic, p. 666. 19.23 Tangent at
667.
19.24 Pair of
p.
v
tangents from t ,p. 667. 19.25 Chord of contact from v

Examples; polar of P l wo s = 0, p. 668.


19.27 Chord whose mid-point is PJf p. 669. 19,28 Diameters,

p. 668.

19.26

p. 669.

19.29 Conjugate diameters, p. 670

19.3

The distance quadratic

670

19.4

Tangent and normal as coordinate axes

671

Exercise 19(a)

672

19.5

Number

673

19.6

The equation s = ks'


19.61 Number of possible intersections of two conies, p.
19.62 s = ks' p. 674. 19.63 Degenerate cases, p. 676.

of conditions which a conic can satisfy

674
674,

19.64 Examples, p. 677.

19.7

two

Equations of a type more general than

conies, p. 679

= ks'

680

Exercise 19(&)

681

Miscellaneous Exercise 19(c)

682

Chapter
20.1

19.65 Contact of

The

20.

Polar Equation of a Conio

684

684

straight line

20.11 Distance formula, p. 684. 20.12 Line joining two points,


p. 684. 20.13 Line in 'perpendicular form', p. 685.

20.14 General equation of a line, p. 685

20.2

The

686

circle

20.21 Polar equation, p. 686. 20.22 Chord


tangent at v p. 686. 20.23 Examples, p. 687

of r

2acoa0;

Exercise 20 (a)

688

CONTENTS

XiV
20.3

Conies: pole at a focus

page 68S

20.31 Polar equation of all non-degenerate conies, p. 688.


20.32 Tracing of the curve Ijr = 1 + ecos0, p. 690.
20.33 Examples, p. 691. 20.34 Chord and tangent, p. 693

Exercise 20(6)

696

Miscellaneous Exercise 20 (c)

698

Chapter 2L Coordinate Geometry in Space:


the Plane and Line
21.1

Coordinates in space

700

21.11 Rectangular cartesian coordinates, p. 700


coordinate systems, p, 701

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)

Direction cosines and direction ratios of a line


21,31 Direction cosines, p. 706. 21.32 Angle between two
21.33 Direction ratios, p. 707
p. 706,

Exercise 21
21.4

700

705
lines,

708

(&)

The plane

709

21.41 Equation of a plane in 'perpendicular form p. 710.


21.42 General linear equation Ax + By +
0, p. 710.
21.43 Conditions for two linear equations to represent the same
plane, p. 711. 21.44 Plane having normal I twin and passing

Cz+D-

through t , p. 711. 21.45 Plane P^Pj, p. 712.


21.46 Intercept form, p. 713. 21.47 Angle between two planes,
p.

713

Exercise 21
21.5

The

713

(c)

714

line

21.51 Line through Pj^ in direction l:m:n, p. 714.


21.52 Parametric equations of a line, p. 715. 21.53 Line of
intersection of two planes, p. 717. 21.54 Distance of a point
feom a plane, p. 718. 21,55 Areas and volumes, p. 719

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

21,71 Geometrical introduction, p. 727. 21.72 Length of the


common perpendicular, p. 728. 21.73 Equations of the common
perpendicular, p. 729. 21.74 Alternative method, p. 729.

21,75 Standard form for the equations of two skew lines, p. 731

Exercise 21(f)

732

Miscellaneous Exercise 21

733

(g)

Chapter 22. The Sphere; Spherical


Trigonometry
22.1

733

Coordinate geometry of the sphere

736

22.11 Equation of a sphere, p. 736.

22.12

Some

from pure geometry, p. 736.


737. 22.14 Examples, p. 738

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

general principle, p. 742.

circle, p.

22.22 Spheres through

742

Surfaces in general

744

Exercise 22(b)

745

The

747

spherical triangle

22.41 Some definitions and simple properties, p. 747.


22.42 Sides and angles, p. 748. 22.43 Polar triangle;

supplemental relations, p. 749.

22.5

22.44 Area, p. 750

Triangle formulae

751

22.51 Cosine rule, p. 751. 22,52 Sine rule, p. 752.


22.53 Supplemental formulae, p. 754. 22.54 Triangb on the
general sphere, p. 755

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 two-volume text-book 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

a serious treatment of the subject, written to

been evident at this

fill

a gap which has long

The author

believes that there is

way

to further progress.

no other
book addressed primarily to the General Degree student which covers
the ground with the same self-contained completeness and thoroughlevel.

ness, while also indicating the

On

the

any examination is from above',


the book has been constructed so that those students who do not
intend to take the subject Mathematics in Part II of their degree
course will find included some useful matter a little beyond the
prescribed syllabus (which throughout has been interpreted as an
examination schedule rather than a teaching programme); while
those who continue with Mathematics will have had sound preparaprinciple that 'the correct approach to

tion.

As

it is

the author's experience that

a degree course have received hasty

many

students

and inadequate

who

training,

begin

a com-

knowledge of previous work has not been assumed.


Although written for the purpose just mentioned, this book will
meet the needs of those taking any course of first-year degree work in
which Pure Mathematics is studied, whether at University or Technical College. For example, most of the Pure Mathematics required
for a one-year ancillary subject to the London Special Degrees in
plete

Physics, Chemistry, etc.

is

included,

two years' work ancillary to Special

and

also that for the first of the

The relevant matter


some of Part II) of the B.Se. Engineering Degree is
covered. The book provides an introduction to the first year of an
Honours Degree in Mathematics at most British universities, and
would serve as a basis for the work of the mathematical specialist in
Statistics.

for Part I (and

Grammar School. Much of the material is suitable for pupils


preparing for scholarships in Natural Sciences.
By a natural division the subject-matter falls conveniently into two
the

volumes which, despite occasional cross-references can be used


independently as separate text-books on Calculus (Vol. I) and on
Algebra, Trigonometry and Coordinate Geometry (Vol. II). According
,

to the plan of study chosen, the contents

may

be dealt with in turn,


b-z

GENERAL PREFACE

XViii

or else split up into

two or even three courses of reading in Calculus,


Algebra-Trigonometry and Geometry taken concurrently. Throughout
it has been borne in mind that many students necessarily work without

much

and

hoped that those of even moderate


be able to use this book alone.
A representative selection of worked examples, with explanatory
remarks, has been included as an essential part of the text, together
direct supervision,

it is

ability will

many sets of 'exercises for the reader' spread throughout each


chapter and carefully graded from easy applications of the bookwork
to 'starred' problems {often with hints for solution) slightly above the
ultimate standard required. In a normal use of the book there will
with

not be time or need to work through every 'ordinary* problem in


each set; but some teachers welcome a wide selection. To each chapter

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 cram-book giving

drill

in examination tricks. However,

many part-time and evening


students in the Te clinical Colleges unfortunately are) may have to
postpone the sections in small print and all 'starred' matter for a
those

are pressed for time (as so

later reading.

Most of the problems of 'examination type* have been taken from


Pinal Degree papers set by the University of London, and I am grateful to the Senate for permission to use these questions. Others have
been collected over a number of years from a variety of unrecorded
{and hence unacknowledged) sources, while a few are home-made.
It is too optimistic to expect that

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

cellence of their printing work.

P.

DEPARTMESX OT PHYSICS AND MATHEMATICS


TECHNICAL COLLEGE
KINGSTON -UPON -THAMES
THJE

GEBBISH

xix

PREFACE TO VOLUME

II

Although for convenience of reference the pagination and section


numbering are continued from Volume I, this does not imply dependence of the present volume on the first. Apart from occasional backward references, Volume II is a self-contained text-book on Algebra,
Trigonometry and Coordinate Geometry of two and three dimensions,
in which calculus methods are illustrated when instructive.
Beginning with a chapter which leads from revision of previous
algebraic work to an introduction to formal algebra, an early start
on determinants can be made in the next, thereby assembling all the
equipment necessary for the subsequent geometry (which throughout
is real and euclidean) Certain widely-used general theorems on systems
of linear equations have received more explicit statement and
emphasis than is customary at this level.
In Chapter 12 the passage from finite to infinite series lays the
foundations of convergence \ a subtle subject so often misunderstood
.

and mishandled by

beginners. It

is

hoped that the many somewhat

negative cautionary remarks will stimulate rather than shake the


reader's confidence.

Chapter 13 sets complex numbers on a logical footing by first briefly


mystery

retracing the historical steps in their development from

through 'diagram' to the concept of 'ordered number-pair*. The


is often approached from only one of these standpoints, but

subject

the present inclusive treatment combines the advantages of all.

The

opportunity is also taken of setting up a general theory of factorisation


and polynomial equations. Despite some repetition, it is felt that this

can be appreciated only

after the provisional

clusions in 10.13, 10.3 has been realised.

The

nature of the con-

following chapter, con-

cerned with trigonometrical applications of the preceding algebraic


theories,

also

introduces some genuine functions of a complex

variable.

An anticipated criticism

of the book

is

that complex numbers and

the complex exponential have been introduced too late, with a con-

sequent loss of freedom of method in the Calculus section in topics


such as the solution of linear differential equations. This delay was
intentional,

and could almost be claimed

as a special feature; for the

author believes that (with the exception of confessedly symbolic

PREFACE TO VOLUME

XX

II

methods in which "anything is fair') only confusion of principles can


arise by incautiously mixing the real and the complex, especially in
calculus techniques.

Since

many

students

come

to the course regarding Coordinate

Geometry as merely a mixture of 'graphs', Pure Geometry, and easy


Calculus, an attempt in Chapter 15 has been made to review the
'known' parts of the subject more as an illustrated account of linear
and (real) quadratic algebra. Use is made of oblique axes when
appropriate.

In the next three chapters, which contain a fairly detailed treat-

and hyperbola by means of their


'standard' equations, the emphasis is on parametric methods and

ment of the parabola,

ellipse,

the consequent elegant applications of the

(real)

theory of equations

given in Chapter 10. Joachimsthal's ratio equation and the distance


quadratic are also used incidentally.

Owing

to the algebraic analogy

between ellipse and hyperbola, the discussion of the latter is centred


mainly on properties of the asymptotes.
The unifying influence of a systematic treatment of 'the general
conic' by Joachimsthal's method is too valuable to omit. This and
the powerful s = ks' principle' are the themes of Chapter 19.
A short chapter on polar equations, first revising the straight line
and circle in this form, develops in more detail their application to the
conic, and thus gives the geometrical complement of the calculus
methods illustrated for various 'polar' curves in Volume I.
In principle, Chapter 21 returns to the fundamentals of cartesian
coordinate geometry, this time for three dimensions. The treatment,
designed to emphasise whenever possible the analogies between the
two- and three- dimensional cases, while also pointing out important
contrasts, revises the methods of linear algebraic geometry, and may
well be left until fairly late in the course. Finally the sphere is briefly
treated^ first geometrically by methods resembling those already used
for the circle in Chapter 15, and then trigonometrically.
e

xxi

REFERENCES AND ABBREVIATIONS


In a decimal reference such as 12.73(2),
12 denotes
12.7 denotes

12.73 denotes

12.73(2) denotes

chapter (Ch.

12),

section,

sub-section,
part.
in the

same

refers to

equation

refers to

worked example

(ii)

in the same section.

refers to

worked example

(ii)

in sub-section 4.64.

Ex. 12 (6), no. 6

refers to

problem number

wo

means

with respect

(ii)

ex.

(ii)

4.64, ex.

(ii)

In the

text,

matter in small type

(ii)

{other

'

omitted at a

first

reading

6 in Exercise 12

(b).

to.

examples) and in starred worked examples is


'

section,

than 'ordinary' worked


subsidiary,

and may be

if time is short.

In an exercise
no. 6

refers to

a 'starred' problem

problem number 6 in the same Exercise.


either

depends on matter in small type


in the text, or on ideas in a

or

is

for the solution of a problem,

later chapter;

above the general standard of


difficulty.

matter in
matter in

. .

. .

is

a hint

is

explanatory comment.

363

10

ALGEBRA OF POLYNOMIALS
10.1

The remainder theorem and some consequences

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

and remainder obtained by dividing ax 2 + bx + cby x k.

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

Observe that the remainder

the original expression with x replaced

is

by

k.

If only the remainder is required in the above example, we can


proceed as follows. Let the quotient be Q{x) (a function oix) and the

remainder be

R (a constant). Then we have


ax2 + bx + c

If

we now put x =

k in

this,

{x-k)Q{x) + R.

we obtain

ak 2 + bk + c
since the

(i)

term involving Q(k)

is

=R

(ii)

zero.

Remarks

may

be objected that the division process has established


x except k, so that the substitution x = k
is not justified. The validity of (i) for all x can be shown by the following
method, which also identifies R. We have
(a) It

relation

(i)

for all values of

{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 right-hand side has the factor x k. If the other factor

is

denoted by Q(x), then


(ax2 + bx + c)- (ak 2 + bk + c)

which

= (x-k) Q(x),

is equivalent to
with R given by
We may regard the 'long division as a process for constructing
(i)

(ii).

'

(ft)

from the given polynomials x k and ax 2 + bx + c another polynomial


Q(x) and a constant R which satisfy (i) for all values of x. (A similar
principle, justified in 10.13, ex. (v), holds for

although when the divisor


of degree

The

n 1

or

is

of degree

n >

any pair of polynomials,

1,

R will be

a polynomial

less.)

which holds for all values of x, is called an identity


an equation in x, which holds only for certain special
values of
The sign = between two polynomials in x is used to denote
'equal for all values of
and is read 'identically equal to'. Thus
(i) would be written
relation

(i),

in x (in contrast to

ax 2 + bx + c

= (x- k) Q(x) + R.

(i)'

The result of substituting a; = k would still be written with the ordinary

sign because

we may

ever,

a relation between special numbers only.

(ii) is

continue to use the sign

are in fact identically equal, unless

= even when

we

How-

the expressions

specially wish to emphasise

their identity.

10.12

The remainder theorem

We now obtain the remainder when the general polynomial


=p

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
)

By direct multiplication we can verify that,

+ ... +p n - 1 (x - k).
for

(iii)

any positive integer

m and any x and k,


xm -km

(When x

4= k,

= (x- k) {x-1 + xm ~ 2 k + xm ~s k 2 +...+ km ~x

this is equivalent to the

the second bracket.) Thus


(iii).

a;

&

is

).

sum-formula for the

g.p. in

a factor of the right-hand side of

Denoting the other factor f by Q{x), we have


p(x) -p(k)
t

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

Hence the remainder when the polynomial p(x) is divided by x k is


number p(k) obtained by substituting k for x in this polynomial.
Also from (iv) we have
The factor case. If p(k) = 0, the polynomial p(x) has x k for a
factor. Conversely, if x k is a factor, then p(k) = 0.
the

Examples
(i)

+nx- 6.

Factorise 2xs -llx 2

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)

= 2-11 + 17-6

p{2)

= 16-44 + 34-6 =

To find the other


2x*-lx + 3, so

Find

a:-

1 is

not a factor.

/.

a factor.

polynomial by x 2; the quotient

by

a and b

if 6x*

remainder) by 2x 2 + x

+ ax 2 + bx 2

(iii)

7,

Find

the

remainder when p(x)

Since the divisor

say

Ax + B.

= 1.

p(x) has degree greater than

lfQ(x)

is

is divided

by (x a)(x b), a

4= 6,

where

2.

quadratic in x, the remainder will in general be linear in x,


the quotient, then (cf. 10.11, Remark (yff))

= (x-a)(x- b) Q(x) +Ax + B.

a:

p( a )

= Aa + B.

6:

p(b)

= Ab + B.

Solving these equations for

A
so the remainder

are factors of the

which give

a + &-,

Put x =
a;

2x-l
0,

is

p{x)

Put

is divisible (i.e. exactly

1.

= a-b-8,
=

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.

Since 2x3 +x~l = (x + l)(2x-l), both x+1 and


given polynomial p(x). Hence p( - 1) = andp(J) =

and so a

is

(a:-2)(a;-3)(2a?-l)

factorisation

the values of

divisible, without

/.

= (x-2)(2x*-lx + 3),
=

(ii)

0,

factor, divide the

p(x)

on completing the

#: 0,

A general result of this kind is proved in

13.62(1).

26-2

ALGEBRA OF POLYNOMIALS

366

[10.13

10.13 Factorisation of a polynomial; 'equating coefficients'

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)

when x has any one of the n distinct values

an then
,

= Po( x ~ a i) (x-a 2 )...(x- an

P(x)

(v)

).

by hypothesis, therefore x a x is a factor


of p(x). Let the quotient when p(x) is divided by x a x be Q n -x{x)\
*1
it will be a polynomial in x of degree n 1 whose first term is p^x
Then
p(x) = (x- a t ) Qn ^{x)
Proof. Since p(a x )

11

Since p{a 2 )

by

hypothesis,

we have

= ((h-c^Qn-Mand therefore
a x (also by hypothesis), hence Qn -i(a 2 ) =
x-a 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=

Qn-ifr)

and so

p(x)

= (x-a 2 )Qn - 2 (x),

(x -

ax )

(x

- a 2 Q n - 2 {x).
)

can continue step by step to remove the factors x a3


until we reach x an after which the other factor will be Q (x) whose
only term is^ a; = p The result (v) follows.
Corollary I (a). A polynomial of degree n in x cannot be zero for

We

more than n distinct values of x.


Proof. Expression (v) cannot be zero when x takes any value
different from a v a 2 ...,a n for no factor would then be zero, and
,

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

this case the expression reduces to

p k Xn

+P k+1 Xn

-k - 1

+..-+P n

(Pk*

0),

a polynomial of degree n k in x. Hence by Corollary I (a),


cannot be zero for more than n k values of x. However, we are

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

thus leads to a contradiction, and so only the

first

alternative

is

possible.

when p = p x =

Finally,
for all x,

i.e.

p(x)

...

= pn =

0,

the given polynomial

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

values of x, or for all values of x, then

Pi

qo,

Pn

$v

qn

>

For we have
1
n
(Po-q )x + {p 1 -q1 )x^- +... + (pn -qn ) =

0,

and the results follow from Theorem II or Corollary II (a).


Corollary 11(6)

is

the basis of the

principle*)"

an identity between polynomials.

cients' in

polynomials of degree n are equal for

all

of 'equating

It asserts that if

values of x

(i.e. if

nomials have identical values for corresponding values of

x),

two

the poly-

then they

they are identical in form). Without this


would be conceivable that two quite distinct polynomials

agree term
result it

by term

coeffi-

(i.e.

might take the same values


numbers.

same

for the

x,

as x ranges over the real

Corollary II (c). //
~
p xn +p 1 xn 1 + ... +pn =
(where

then

1
m
q x + q^-

+ ...+qm

m > n)for more than m distinct values of x, or for all values of x,

= qm _ n _ 1

p = qm _ n

In particular, the polynomial on the right has


The proof is like that of Corollary II (b).

degree n.

Examples
(i)

Find

constants a,

n =
8

b, c,

dfor which

an(n +

1) (n

+ 2) + bn(n + 1) + cn + d.

f Already illustrated in 4.62.

p n = qm

ALGEBRA OF POLYNOMIALS

368
The right-hand

[10.13

side

= a(n 8 + 3n 2 + 2n) + b(n z +n) + cn+d


= an3 + (3a + b)n* + (2a + b + c)n + d.
This

is

identically equal to

and only

if

3a + 6

1,

i.e.

(ii)

nz

Show that x% cannot

2a + 6 + c

0,

1,

if

= 3,

1,

d=

0,
a*

0,

0.

be expressed in the form

a(x+l)(x + 2) +b(x + 3)(x + 4:).

= 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.

two equations give a = 0, 6 = 0, and these values fail to satisfy the


The three conditions cannot be satisfied, and so x 2 cannot be written in

last

the form stated.


(iii)

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)

relations between a, b, c by equating coefficients of the various

powers

of x.

When x =

a,

the left-hand side reduces to

Hence the polynomial

1.

(x c)(x a)
(b c)(b a)

(x b)(x c)
(a b){a c)

(x a)(x b)
(c

a)(c-b)

zero when x = a. Similarly, it is zero when x = 6 and when x c. Thus this


quadratic in x vanishes for three distinct values of x. Hence it vanishes for all
values of x.
By equating coefficients of x2 , x and the constant terms, we obtain

is

(a-6)(a c)
(a

6+c
b)(a c)

(a

6) (a c)

(6-c)(6-a)

(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)

as a factor, obtain relations between

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;-l-afc,

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

The process of dividing/(a;) by g(x) gives a quotient Q(x)


a remainder B(x) of degree less than n such that (a) holds for
except possibly those for which g(x) = 0. Thus the polynomial

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).

had we proceeded in a different way,


obtained another pair of polynomials q(x), r(x) such that
It is conceivable that,

f(x)

where r(x) has degree

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

2 6a8 -a; 2 -19:c-6.

4:X + x + 6 =
2

5 Find the values of a

x* + 2x 3 + x 2 - 4.

0.

and

6 if

6a; 8

+ 7a; 2 + ax + b is divisible by 2x 1 and by

x+1.
6 The remainder when (xl)(x 2)
a and b.

divides x* + ax3 + b

is

a;+l.

7 A polynomial gives remainder 2x + 5 when divided by (x 1) {x +


the remainders when it is divided by x 1, x + 2 separately.

2).

Find

Find

8 A polynomial gives remainder 2 when divided by x + 1 and remainder 1


when divided by x 4. Find the remainder when it is divided by (x + 1 (x 4).
*9 A cubic polynomial gives remainders 5x 7, 12a; 1 when divided by
)

x2 x + 2, x 2 + x

respectively.

Find the polynomial.

When divided by x 2 + 1 a polynomial gives remainder 2x + 3, and by x 2 + 2


gives remainder x + 2. Find the remainder when it is divided by (x 2 + 1) (x 2 + 2).
* 10

Find values
11

of a,

b, c,

d for which

n 8 + 6n = an(n-l)(n-2) + bn{n-l) + cn + d.

12 x3

a(x+2)* + b(x+l) 2 + cx + d.

13 If x* 6x3 + ax 2 + 30* + 6 is a perfect square, find a and 6 and write


the square roots of the polynomial.

down

ALGEBRA OF POLYNOMIALS

370

2x + l

14 Express

If a,b,c are

^
16

15 Express

m the form
.

(x-l)(x + 2)

b
H

-D m the form
.

2x+l

-.

x*+l

prove

a(x b) (x c) b(x c)(x a) c(x a)(x b)


_i
L

-+ =
+(a b)(a c)
{c a)(c b)
(b c)(b a)

+ b + x) (b + c + x)
(b c)(b a)

(a

x-1 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)

18 (i) Prove that there cannot be two different quadratic expressions in x


which take the values A, B, C when x has the distinct values a, 6, c respectively,

Verify that

(ii)

(x-c){x-a

(x-b)(x-c)

{x-a)(x-b)

B
c
c)(b a)
(b
(c a)(c b)
)

(a

b)(a c)

has the required properties, and deduce from

(i)

that this

is

the only such

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

If a;4 + px + q contains a factor (x a) 2 prove 27p 4

(i)

256q3 and express

x, y,

we may arrange

q in terms of a.
(ii)

10.2

Prove the converse of

(i).

Polynomials in more than one variable

10.21 Extension of the preceding results


If p(x, y)
it

is

a polynomial in the two variables

according to powers of x, say as

%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

ax2 + 2hxy + by* + 2gx + 2fy + c


a

a',

b',

c',

+ 2h'xy + b'y* +
+ Wv +
f=f, g = g', h = h'.

a'x*

Arranging both expressions as quadratics in


ax* + 2(hy + g) x + {by* + 2fy + c)

a'x*

x,

+ 2{h'y + g')x + (b'y* + 2f'y + c')

t The proof in 10.12 is valid when p j> x ...,pn are polynomials


number of variables); therefore the deductions in 10.13 still hold.
,

<>'>

in y (or in

any

ALGEBRA OF POLYNOMIALS

10.22]

Equating
a

we have for all y:

coefficients,

2(hy + g)

a',

371

2(h'y + g'),

by 2 + 2fy + c

b'y 2

+ 2f'y + c'.

By equating coefficients of like powers of y in the last two identities, we obtain


all

the results stated.

Show that 2x 2 + 5xy Zy 2 -x + lly-6 has

(ii)

Since

2* 2 + 5xy - 3y*

(x

linear factors,

and find them.

+ Zy) (2x - y),

try putting

2x 2 + 5xy-Zy 2 -x+lly-6

(x

+ Zy + a){2x-y + b),

where a and 6 are constants to be determined.

The right-hand side


2x

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

,
= -\, 36-a=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+Sy-2)(2x-y + Z).
Factorise

(iii)

Regarding

xy(x + y)

+ yz(y + z) + zx(z + x) + 2xyz.

a quadratic polynomial in x whose


put x y: the expression becomes

this as

nomials in y and

z,

+ yz(y + z) - yz(z -y)- 2y2z Hence x + y is a factor. Similarly, x + z is a factor.

coefficients are poly-

0.

Next, regarding the expression as a polynomial in y whose coefficients are


polynomials in x, z, we may put y = z and show that y + z is a factor.
Hence (y + z) (z + x) (x + y) is a factor. Since this product and the given
polynomial each have total degree 3, any other factor h must be numerical,
LHCL

SO

xy(x + y)

+ yz(y + z) + zx{z + x) + 2xyz =

k(y + z)(z + x) (x + y).

0,

We obtain k =

1.

To obtain k, we may either substitute numerical values for x,


y

1,

1) in

both sides, or compare coefficients

10.22 Symmetric,

skew and

(e.g.

of x 2y).

y, z (e.g.

cyclic functions

In 1.52(4) we defined a homogeneous function of two or more


variables,
(1)

variables

and we now give further useful

definitions.

function (not necessarily a polynomial) of two or more


is

symmetrical in these variables

if it is

unaltered

by the

interchange of any two.

For example, x + y and x2 + y2 are symmetrical functions of (x,y);


a + b + c, bc + ca + ab are symmetrical in (a, b, c); and the expression
in 10.21, ex. (iii) is symmetrical in (x,y,z). The polynomials x y + z,
a2 + b % + 2c 2 are not symmetrical.

ALGEBRA OF POLYNOMIALS

372

[10.22

Bemarks
(a)

The only symmetrical function of (x, y, z) of the first degree is a


ofx + y + z. For if Ix + my + nz is unaltered by inter-

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,

linear function reduces to

The most general symmetric polynomial in (x, y, z)


second-degree terms can likewise be shown to have the form
(/?)

k(x 2 + y 2 + z 2 )
(2)

consisting of

+ l(yz + zx + xy).

A function of two or more variables is skew or alternating if an

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)

+ yz(y -z) + zx(z - x)

skew functions of their respective variables.


The reader should satisfy himself that the product

are

or of two skew functions is

of two symmetric

a symmetric function, while

symmetric and a skew function

is

the product of

skew.

Considerations of symmetry, skewness and homogeneity often save


algebraic manipulation,

and will be helpful in Ch.

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

by long division, or as follows.


The given expression and the factor x + y + z are both symmetric

directly

in (x,y,z),

therefore so is the remaining factor.


Since the given polynomial is homogeneous of degree 3 and x + y + z is homogeneous of degree 1, the other factor must be homogeneous of degree 2 in (x, y, z).
By Remark (/?) above, the most general symmetric homogeneous polynomial of degree 2 in (x, y, z) has the form

k{x2 + y % + z 2 )

+ l(yz + zx + xy),

and so
x* + y* + z z -Zxyz

(x

+ y + z) {k{x 2 + y* + z 2 + l(yz + zx + xy)}.


)

ALGEBRA OF POLYNOMIALS

10.22]

We can find the constants k,

Hence

by substituting numerical

0,

0,

give

k;

0,

y=l,

give

2{2k + l},

xP + y^+z 3 3xyz

(3) Cyclic expressions.

:.

= - 1.

Consider again

+ yz(y -z) + zx(z - x),

letters x, y, z are

of a circle as shown.

values:

= (x+y+z) (x* + y 2 +z 2 yz zx xy).

xy(x - y)

and suppose the

373

When

the

arranged around the circumference

first

term xy(x y)

is

given, the next

can be obtained from it by replacing each letter by


the one which follows it on the circle. Repetition of
the process gives the third term, and further repeti-

back to the first term. The given sum


thus consists of the term xy(x y) together with the
two similar terms obtainable from it by cyclic interchange ofx, y, z\ it can be written
tion leads

Lxy{x-y).

The number of letters involved has either to be stated explicitly, or


must be clear from the context; otherwise the S-notation is ambiguous.
Observe that the complete sum is unaltered if we replace x by y,
y by z, and z by x; for this cyclic interchange merely alters the order
in which the three terms occur. The expression (y z) (z x) (x y) has
the same property.
Definition. An expression in x, y, z which is unaltered by cyclic
interchange of these letters

in (x, y, z).
similar definition can be given when there are
is cyclic

more than three


For convenience of reference and comparison, it is desirable
to write expressions with their terms in cyclic order whenever possible.
Thus we prefer to write bc + ca + ab rather than ab + ac + be.
Remark (y). To test an expression for symmetry or skewness we
interchange letters two at a time. A cyclic interchange involves change
of all the letters. Thus a function may be cyclic but not symmetric,
e.g. I,xy{x-y), bc 2 + ca2 + ab 2 See also Ex. 10(6), no. 16.
letters.

Exercise 10(6)
Prove that the following expressions have linear factors, and find them.
1

2x2 -Zxy-2y i + 7x + 6y-'t.

x*-y i + 2xz-14yz-48z*.

ALGEBRA OF POLYNOMIALS

374
3

[10.3

Find the values of a for which 2a; 2 5xy 3y 2 x + ay 3 has


[The factors must be of the form (x 3y + 3k) (2x + yllk).]

linear

factors.

4 Show that x 2 + Axy + 3y 2 + 2x 2y + 6 does not possess linear


Using

the

remainder theorem, together with considerations of degree, symmetry

and skewness when


5

(x + y + z) 3

helpful, factorise

-(x3 + y3 + z 3 ).

+ zx(z 2 x 2 +xy(x 2 y 2
x(y-z) + y(z-x) 3 + z(x-y) 3
(x + y + z) 5 -{x 6 + y 5 + z 5

6 yz(y 2
7

factors.

).

).

9 Write the expressions in nos.


10 Write in
(i)

full

Zx 2

6, 7

in the S-notation.

the following expressions, assumed


lLx 2 y 2 ;

(ii)

S6c(6-c);

(hi)

(iv)

to involve three letters:

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(6-c)

a(6

(iii)

12 Prove S(6 - c) 3

=
2

0;

-c

Xbc(b-c)

(ii)
2
)

= Sa 2 (6-c);

(b-c)(c-a)(a-b).

3(6 - c) (c - a) (a -

6).

Sa 8 (6 2 -c 2 = -(b-c)(c-a)(a-b)(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

Prove that a symmetric or skew function of three variables

[If/(a, 6, c) is

any

a b in the example in 10.22(2). deSa 3 (6 c) 3 (iii) Use the result of

Similarly, deduce the factors of

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)

Polynomial equations: relations between roots and

coeffi-

cients

10.31 Quadratics: a
If the equation ax

reader will

know

summary

+ bx + c =

that

has roots

a+/3 =

a,

fi

a/?

-.

He will have used these symmetrical relations to


of other symmetric functions of

(possibly equal), the

a and

ft

calculate the values

(such as

a 2 +/? 2

a//?

+ /?/a),

and to construct quadratics having prescribed functions of a and ft


as roots (e.g. 'form the equation whose roots are 3a /?, 3/? a').
We now extend this work to cubic and quartic equations.

ALGEBRA OF POLYNOMIALS

10.32]

375

10.32 Theory of cubic equations


If the distinct

numbers

a,

/?,

satisfy

ax3 + bx 2 + cx + d

(i)

0,

then by Theorem I of 10.13,

ax3 + bx 2 + cx + d

numbers a,
Theorem x-cc and x-fi are
If the unequal

division the other factor

= a(x-a){x-fi)(x-y).
ft

satisfy

then by the Remainder

(i),

factors of the left-hand side,

'

Similarly,

that the factor x-fi appears twice in (ii).


the roots of (i) we mean that the right-hand side of
'

(ii)

if a,

holds in

/?,

all

and by

seen to be linear and of the form a(x y).


are the roots of (i) we mean that y = fi, so

is

By the statement

Thus,

(ii)

by

3
(ii) is a(x - a)

(not necessarily distinct) are the roots of

cases.

a are

'a, a,

Expanding the right-hand

side

(i),

by

then

direct

multiplication,

ax3 + bx 2 + cx + d

Equating

axa -a(a+/3 + y)x 2 + a(fiy + ya + afi)x-aafiy.

coefficients,

we find

a + p + Y = -^,

PY + Y + P =

Conversely, the cubic having roots x

a,

(x-a)(x-fl){x-y) =
i.e.

in

^
ft,

a PY =

P)

y is

0,

x3 -(a+fi + y)x 2 + (py + ya, + afi)x-otfiy

0,

which

and

the coefficient of x3

is

the coefficient of x2

is

the coefficient of x

is

the constant term

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;

they are equivalent to the information that


of(i)'.

'a, /?

and y are the roots

ALGEBRA OF POLYNOMIALS

376

[10.32

Examples
(i)

If a,

/?,

are the roots of pa? + qx +r

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)

= 0-2?- =

a/>y

a is a root of the given equation,


pot 3

there are similar relations for

/?,

+ q<x + r =

0;

y. Adding,

pSa 3 + gSa + 3r =
Sa 8 =

.-.

0,

By the example in 10.22 (2),


+ y 8 -3a/y = (a+/? + y) (a 2 +/#2 + y 2 -/?y-ya-ayff)

Second method.

+/ff

0,

Sa =

since

0.

3r

Sa 3 = 3a#y =

.-.

P
(d)

From the given

pa.

Adding

this to the

0,

+ qu + ra =

0.

equation, >a 8 + qoc + r

two

similar relations for

fi,

y,

2>Sa 8 + gSa 3 + rSa 2

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 = Sa-a = -1-a. Similarly y + a = -1-/?, a + /? = 1-y. 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:

In the given equation put y

(-l-2/)
i.e.

377

+ (-l-2/)2-24(-l-2/)-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

this is the required equation;

but see the

when the roots

\y
so

by

64cc a

squaring,

^-3j =

*(?-.)
y\y

The argument used

in (a)

Remark at the end

of 13.53.

shows that

16 2

^1-^

=M

a
,

y)

y* 38t/ a +

which reduces to

4% 16 = 0.

10.33 Quartic equations

By reasoning

as in 10.32

we

find that if a,

ax* H-foe3 + 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

8 are the roots of

Sa = --,

i.e.

fiy8+y8a + 8<xfi + afiy

/?,

Qi

Again notice the alternation of the

signs.

Applications of the theory given in 10.31-10.33 to coordinate

geometry are illustrated in 16.22 (2);

and elsewhere.

16.26, ex.

(i);

16.32; 16.12, ex.

(ii);

ALGEBRA OF POLYNOMIALS

378

[10.4

Exercise 10(c)
If a,

ft

are the roots of x 2 5x + 3

a 2/? + a/? 2

(i)

(v) (a-/?) 2 ;

2 If a,
roots are

(ii)

are the roots of

/?

(i)

(iii)

^ + ^;

(vi)

i/ a ,

calculate

(a-l)(/?~l);

a3 +

(vii)

+ 2* 4 =

3a; 2

a2

(ii)

0,

/?

(viii)

a 2 +/? 2

a4 +

/ff*.

a+

/?+ 1/a.

1//?,

Write down the conditions for both roots of ax + bx + c =


2
z
14a; 24 = 0, calculate
/?, y are the roots of x +

to be positive.

construct the equation whose

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)

5 If a, /?, 7 are the roots of x + 4a; + x 2


whose roots are
3

Sa 2

(iii)

a" 2

0,

a(/?

/?~ 2 ,

construct the equation

+ 7), /% + <*), 7(+A);

7~ 2

6 If a, /?, 7 are the roots of 6a; x 6x 2 = 0, form the equation


roots are a + \, /? + \, y + \. Hence solve the given equation.
3

has two roots equal, prove


7 If x 3 + BHx + G =
roots be a, a, /?; then 2a +/? = 0.]

G 2 + 4H3 =

3Hx + G = are in a.p., prove G = 0.


+ px 2 + qx + r = 0, find the condition for the roots

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)

10 If a, /?, 7 are the roots of a; 3 + px 2 + qx + r = 0, construct the equation whose


roots are a 2 fiy, P 2 ya, y 2 ocfi, and comment on the special cases (i) p = 0,
2
2
2
(ii) q 0. Also write down the value of (a - Py) (fi - yet) (y - a/?).

a 2 -fiy = a 2 + a
J_
I

3
= a +r =

pct 2 q<x

= -(pa + q).
,

~|

_|

has roots a, /?, 7, 5, construct the equation whose roots


If xi + px 3 + q =
a + ji + y, ft + y + 8, y + d + ct, 8 + a + fi.
has three equal roots, prove that p 2 + 12r = and
12 If a; 4 + px 2 + qx + r =
2 =
value of the repeated root is 3g/4p.
that
the
and
show
32pr;
9g
11

are

10.4

Elimination

10.41 Further examples

When a
is

system of equations is given and the number of equations


number of unknowns, then in general the equations
be satisfied unless the coefficients are related in some way.

greater than the

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

some further examples.

give

can be laid down; but see 11.43.

identity.

Eliminate x, y, z from

x + y+z

a,

From the identity


x

b2

x3 + y 3 + z3

c3 ,

xyz

= d3

(10.22, ex.)

+ y + z -3xyz =
3

+ y + z)(x 2 + y 2 +z 2 -yz-zx-xy)
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 =

).

).

Use of theory equations.

nfrom (y-mx) 2 = a 2m 2 + b 2 (y nx) 2 = a 2n 2 + b 2 mn = 1.


The first two equations show that m and n are the roots of the quadratic in t

Eliminate m,

= a2t 2 + b 2
+ 2xyt + (6 2 - y 2 =

(y-tx) 2
(a

i.e.

-x

2
)

0.

The third equation shows that the product of these roots


b

-y =

1, so

a2 x2
x 2 +y 2

i.e.

is

a2 + b 2

For the geometrical interpretation see 17.47,


The method illustrated in ex. (ii) can sometimes be used to solve a system of
equations as in ex. (iii) following. See also Ex. 10(d), nos. 12-14.
(ii).

(iii)

Solve for x, y,z:

x + ay + a 2z

a*,

These equations show that

x + by + b 2z
a, b, c

Since the coefficient of

Hence

is

c*.

zero, the

sum

0.

of the roots

is

zero,

and hence the

(a + b + c). Then

x = product of roots = abc(a + b + c)


y = sum of the products of the roots taken in threes
= abc (a + b + c)(bc + ca + ab);
z = sum of the products of the roots taken in pairs
= bc + ca + ab (a + b + c) (a + b + c)
= -(a 2 + b 2 + c 2 + bc + ca + ab).
x = abcZa, y = abc-(Za) (Hbc), z = E(a 2 + 6c).
f It

27

3 is

x + cy + c 2z

are three roots of the quartic in

t*-t2z-ty-x

remaining root

bi ,

has already been used in this book.

GPMII

ALGEBRA OF POLYNOMIALS

380

Common root

10.42

of two equations

The necessary and

(1)

sufficient condition that the quadratics

ax 2 + bx + c
have a

[10.42

0,

a'x 2 + b'x + c'

common root is
{ab'- a'b)

-b'c)

{be'

-c'a) 2

{ca'

is a value of x which satisfies both equations,


then by treating them as simultaneous equations in x 2 x and elimin-

Necessary. If there

ating

first x,

and then x 2 we obtain


,

(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

so that the condition follows.


Sufficient. If ab'

a'b

=(=

the condition

0,

(ca'-c'a\ 2

be'

Put A = {ca'-c'a)l{ab'-a'b)\ then A 2 =


{ab'
{ab'

- a'b) A 2 =

be'

- b'c,

- a'b) {aX 2 + bX) =

{ab'

a{bc'

(i)

-b'c

{be'

-b'c)/ {ab' -a'b),

- a'b) X =

{ab'

a'b) (aA 2 + bA + c) =

0;

and

ca'- c'a.

- b'c) + b{ca' - c'a)

c{ab' a'b)
Hence

can be written

and

on simplifying.

similarly

{ab'-a'b) (a'A 2 + 6'A + c')

0.

a'b #= 0, these show that x = A satisfies both quadratics.


a'b = 0, then the given condition (i) shows ca' c'a = 0.
Hence a:b:c = a' :b' :c', and so the quadratics are not independent.
Since ab'
If ab'

In this case the result holds trivially.


Remark. Direct calculation will verify that
4{{ab'

- a'b) {be' - b'c) - {ca' - c'a) 2}


=

The condition

(i) is

{b 2

(6

- 4ac) (6' 2 - 4aV) - {2ac' + 2a' c - bb'f.

therefore equivalent to

- 4ac) {b' 2 - 4aV) =

{2ac'

+ 2a' c - bb'f.

(ii)

ALGEBRA OF POLYNOMIALS

10.43]

381

When the given equations are of degree higher than the second,

(2)

we ehminate the highest powers step by step until two


with a common root are obtained. For example, if

quadratics

/ = ax3 + bx 2 + cx + d =
and

have a common

root,
(

root

if

like

(i)

then this root also

(iii)

satisfies

pb aq)x2 + (pc ar)x+pd

p / ax
.

0, i.e.

(iv)

0.

and p.f ax.g = is also a


root of g =
and / = 0. Hence / = 0, g = have a common
and (iv) have a common root, and a condition

common

Conversely, a

common root

= px 2 + qx + r =

of g

and only

if (iii)

will express this fact.

10.43 Repeated roots

With the notation of


p(x)

where g{a)

4= 0.

root

a,
r.

// p(x)

0,

(x-ct) r g{x)

(1

<

<

n),

not a factor

We say that the polynomial p(x) has a repeated

and that the equation p{x) = has an r-fold


r equal roots a, a root x = a, of multiplicity r, or a root x = a
If r = 1, we call # a a simple factor of p(x), and x = a a

simple root ofp(x)

p'(x)

g(x).

x a of order

of order

suppose that

Then by the Remainder Theorem x ocia

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:)

r(x a) r_1 g(x)

+ (x a) r g'{x)

= (x- a)'-1 {rg(x) + (x - a) gr'(z)}.


Since ^(a)

=# 0,

the contents of the last bracket are non-zero

when

= a is a root of p'{x) = 0, of order exactly rl.


Conversely, if p'{x) =
has a root x = a of order rl, then provided
x = a satisfies p(x) = 0, it is a root of p(x) = of order r.
of order s. Then the
Proof. Suppose x = a is a root of p(x) =
preceding theorem shows that a is a root of p'(x) =
of order 51.
Hence rl =31, i.e. r s.
x

a.

Hence

a;

These two results show that the necessary and


to have a repeated root is that p(x) =
for p(x)

common

sufficient condition
0, p'{x)

have a

root.
27-2

ALGEBRA OF POLYNOMIALS

382

[10.5

Exercise 10(d)
Eliminate

from

2 x = a(l-t 2 )/(l + t 2 ),y = 2bt/(l + t 2


+ l/t?.
2
3 Eliminate x y from x + y = a, x + y 2 = b 2 x* + y* = c 4
4 If a 2 + x 2 = b 2 + y 2 = ay bx = 1, prove a 2 + 6 2 = 1.
[(a 2 + a: 2 + (& 2 + 2/ 2 )-2(GM/-&a:) = 0, i.e. (x + b) 2 + (y-a) 2 = 0, ;. x = -b and
V = ]
5 If a3cc + 6 3 i/ = xy, b 3x a 3y = x 2 y 2 and x 2 + y 2 = 1, prove a 2 + b 2 = 1.

t-ljt, 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

m, n from the equations in no.


fi from

Z,

and Im =

b.

10 Eliminate A,

a2 + A
11

62

+A

a 2 +^

Solve the following equations for

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

sufficient condition for

x2 +px + q

and

Spqr + p r + q =
3

0.

sufficient condition for the following to

+px + q =

0. (Cf.

Ex. 10(a), no.

20.)

0. (Cf.

Ex. 10(a), no.

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,

have a common root

Find

x. y, z:

x + by + b 2z

15 Prove that the necessary


3

6 2 +/t

18 Find the common root condition

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

last result in 10.43

any) of p(x)

0,

we must

shows
find

that, to find the repeated roots (if

all

factors

common

top(x) andp'(x).

ALGEBRA OF POLYNOMIALS

10.51]

We now show how to find the

common

highest

383

factor (h.c.f.) of

any

pair of polynomials f(x), g(x). It will be convenient to write deg/

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

for 'the degree of f(x)

Suppose deg/ =
f(x)

by

g(x),

=g(x)q x (x) + rx (x).

f(x)

The degree of rx (x) must be less than that of


degrx ^ n 1. We can therefore divide g(x) by r x (x),
q 2 (x) and remainder rz (x), where

3W

getting quotient

(ii)

degr2 ^ n 2.

i.e.

Similarly, dividing rx (x)

rJx),

and hence

g(x),

=rx {x)q 2 (x) + r 2 (x)

g(x)

and degr2 < degr1?

(i)

where

by
.

r x (x)

r 2 (x) gives

quotient q3 (x) and remainder

n~\
(m)

=r 2 (x)q3 (x)+r3 (x)


,

and deg rz ^ n 3.
Proceeding thus,

and
where

rs+1 (x)

applications:]:

First,

0.

we eventually
V-aO*)

=rs- 1 (x)qs (x)+rs (x)

rs_ x (x)

The

rs (x)qs+1 (x)

stage

is

suppose f(x) and g(x) have a

identity

(s)

+ rs+l (X),

(8+1)

reached after at most

n+l

of the successive division process

f(x)

The

obtain

(i)

rx (x)

=fx (x)h(x),

common factor h(x),

g{x)

so that

gx (x)h(x).

shows that

= f{x) - g(x) qx (x) =

h(x) {fx (x)

- g x (x) qx (x)},

is

is also a factor of r x (x). Identity (ii) then shows that h(x)


a factor of r2 (x), and so on. Finally, identity (s) shows that h(x)

is

a factor of rs (x). Hence

so that h(x)

any common factor

off(x), g(x) is also

a factor of r8 (x).

lc(x) is a factor of rs (x). Then identity (s+


a factor of rs _ x (x); hence by identity (s), k{x)

Secondly, suppose that

shows that

k(x) is also

(A)

t Cf. 10.13, ex. (v).


% This is seen from the degrees of the successive remainders.

1)
is

ALGEBRA OF POLYNOMIALS

384

a factor of r8_ 2 {x)


g(x),

and hence by

etc.

Thus from

(i) it

is

any factor of rs (x)

From

(B)

it

and g(x);
must also be a

see that k(x)

is

a factor of

a factor off(x). Therefore


is

a common factor of f(x) and

follows that rs (x) itself

and (A) shows that every

f(x)

we

(ii)

[10.51

g(x).

(B)

must be a common factor of


common factor off(x) and g(x)

factor of rs (x). Consequently,

(a) if rs (x) == constant,

then f(x) and g(x) can have no algebraic


g(x) are coprime polynomials;

common factor, and we say that f(x),

a polynomial, then rs (x) is the (algebraic) highest


common factor f off(x) and g(x). It has been found as the last non-zero
(6)

if rs (x) is

remainder in the process of successive division just described (the h.c.f


process or Euclid's algorithm).

Examples
(i)

Find

the h.c.f. of
2a;

- 4a;8 + 5a; 2 + 2a; -3 and Xs - x* + x* + 4a; 2 - 2x + 3.

We begin by dividing
2(x6 -x i + x 3 +

4x*-2x + S)

by

2a;

- 4a; 8 + 5a; 2 + 2a; - 3,

in order to avoid introducing fractional coefficients. The introduction or


removal of numerical common factors at any stage is permissible since this

does not alter the algebraic factors, f


g

2xl -4x3 + 5x 2 +
2x* + 2x 3 - 6a; 2 +

2x-

18a;

2a;

-6a; 3 +lLr a -16a;- 3

-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

- 2a;4 + 2a;8 + 8a; 2 -4a; + 6 = /


- 4a;4 + 5a;3 + 2a;2 - 3a;
8
2
2a; 4 - 3a; + 6a; - x + 6
2x* - 4a; 8 + 5a; 2 + 2a; -3
xa + x 2 -3x + 9 = r 1
x3 -2x 2 + 3x
3a; 2 -6a; + 9
3a; a -6a; + 9
= r,

the last non-zero remainder ra viz. x 2 2x +


,

Test the equation


4a; 5

- 20a;4 + 25a;8 + 10a; 2 - 20a; - 8 =

for repeated roots, and hence solve it.


repeated factor off = 4a; 5 20a;4 + 25a; 8 + 10a;2 20a; 8

3.

is

Numerical factors are regarded as irrelevant; Xrt (x), where


f
constant, would also be called the h.c.f.

common factor
is

any non-zero

ALGEBRA OF POLYNOMIALS

10.52]

H.C.F.

- 16a;3 + 15a; 8 + 4x - 4). We therefore


process to/and
4
3
2
g s 4a; - 16a; + 15* + 4a;-4.

of / and /'

5(4a; 4

4a; 4

-16a; 3 +15a; 2 + 4a;-4

4a; 4

- 14a;3 +
- 2ar*+
- 2a;3 +

The
it

h.c.f. is 2a; 3

4a; B

-4a;-4

that

begin by applying the

- 20a;* + 25a; 3 + 10a; 2 - 20a; - 8 = /


-16a;4 +15a;3 + 4a; 2 - 4a;
- 4a;4 + 10a;3 + 6a; 2 - 16a;- 8
- 4a; 4 +16a; 3 -15a; 2 - 4a; + 4
- 6a;3 + 21a; 2 -12a;-12
Remove 3:
3
2a; - 7a; 2 + 4a; + 4 = r x
4a;

+ 8a;
lx 2 - 4a; 8a; 2

7a: 2

385

r9

7a; 2 + 4a; + 4. By

using the Remainder Theorem,

we

find

has a factor x 2; hence


H.CB-.

From

10.43

it

(x -2) (2a; 2 -3a; -2)

follows that

(a;-2) 2 (2a;+

x 2, 2x+l are repeated

and that x = 2, x = \ are roots of/


there can be no other roots.

in/,

of orders

1).

factors of orders 3, 2

3, 2.

Since/ has degree

5,

An important algebraic theorem


The identities (i), (ii), ... in 10.51 show that

10.52

ri

ra

and so

on.

=
=

f+bo g,

where

(hf+brf,

where

ax

We thus see that


rs

where a, b are polynomials in x.


Writing/ = rs <J>, g = rt f, then by
1

where 0,

\Jr

//

Theobem.

af+bg,

(C)

(C)
a<j>

+ bxjr,

factor. Since a<fi + bijr has no algebraic


are coprime.
constant, then we may divide (C) by rs and get the important
If f, g are coprime polynomials, then other coprime polynomials
1),

hence

B exist for which Af+ Bg =

A,

= -qx
b t = l + q1 qa

have no algebraic common

factor (being identically


(a)

= 1,
= -q 2

a, b

1.

If rs is a polynomial in x, then we have:


The h.c.B". of two polynomials f, g can be written in the form af+bg, where a, b
are coprime polynomials.
polynomial is said to be irreducible when it has no algebraic
Definition.
factor of lower degree than itself.
Cobollaby. If f is irreducible and g, h are polynomials such that f is not a
factor of g but f is a factor of gh, then f must be a factor of h.
Proof. ^ Since / is not a factor of g, the degree of any common factor of/, g
is less than the degree of / and is therefore zero since by hypothesis / has no
such factor. Hence /, g are coprime and, by the theorem, polynomials A,
exist such that
, _
,

(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

result is 'intuitively obvious'.

ALGEBRA OF POLYNOMIALS

386

[10.53

Remarks
(a)

The polynomials

a, 6 in

a,

(where c

is

identity (C) are not unique, for clearly

and

a-\

any polynomial) would

0,

also satisfy (C).

There are arithmetical results (important in the Theory of Numbers)


analogous to the preceding; they can be formulated by replacing 'polynomial',
'irreducible', 'degree' by 'integer', 'prime', 'magnitude' respectively.
(/?)

10.53

Theory of partial fractions

An application of the theorem in 10.52 is to prove the possibility of resolving


a rational function into partial fractions. In 4.62 we stated and illustrated the
practical methods for doing this in any particular example, but we did not
attempt a general justification of our statements.
Definitions. The rational function f/g, where / and g are polynomials in a;, is
called (a) irreducible if /, g are coprime; (6) proper if deg/ < deggr.
Iff, g have H.c.s\ h (found as in 10.51), then / = hfx and g = hgx so that
(except for those values of x which make g = 0) f/g = fx /gx emdf1 ,g1 are coprime. We shall therefore always assume that the rational function is irreducible.
Iff/g is not proper, we can express it as the sum of a polynomial and a proper
fraction in the form q + r/g, by finding the quotient q and remainder r when
,

/ is divided by g.
Theorem. If p x p 2 are coprime polynomials,
f/(PiPz) can oe expressed uniquely in the form

the (irreducible) rational function

+ r-,
rx

3+

Pi

Pi

where qis a polynomial and rx /p x r 2 /p 2 are irreducible proper fractions.


Proof. Since p x , p 2 are coprime, polynomials A, B exist such that
,

APi + BP2 =

_/_

J(APl + Bp = Bf + Af^
2)

PiPz

By

division,

where deg rx

<

Bf =

q x p x + rx

deg p x and deg r 2

Pi'

PlPi

Pi

and Af =

q2p z + r2 ,

< degp 2 Hence


.

?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
,

that f/(p x Pi) is irreducible.


To prove uniqueness of the decomposition
?+

+ =
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

a factor of the left-hand

p2

side, therefore

387

a factor of

is

(r2

r2 )p x

As p 2 is prime to p x by hypothesis, it follows from the corollary in 10.52 that


p 2 must be a factor of r2 r2 This is impossible unless r2 r 2 = 0, for p 2 has
higher degree than r'2 and r 2 and so certainly degp 2 > deg(r2 2 ). Thus
r2 = r2 and similarly we can show r'x = rx It then follows that q' = q.
.

p n are polynomials every two of which are coprime,


rational function f/(p x p 2 ...p) can be expressed uniquely

Corollary. If p lt p 2
then the (irreducible)

in the form
q-\

1-

Pi

...

Pn

Pi

r n /p n are proper and irreducible.


where qisa polynomial and rx \p x
This is proved by repeated applications of the preceding theorem:
,

Pi(Pi---Pn) to give
gi

then to p 2 (p 3

. . .

pn

first

to

+ -+

J1

Pi

Pz---Pn

to give

f
~
= ?2+
+;
p 2 P 3 ..-Pn
^*2

/i

p 2 ...p n

on. The uniqueness is proved by the same argument as in the theorem.


Remark. For their application to a given rational function f/g, the preceding
results depend on the factorisation of the denominator g into irreducible factors
PiP 2 ...p n (of. 4.63). In 13.61 we shall prove that a polynomial g(x) can be
factorised into the product of linear and irreducible quadratic factors like
(x a) r {(x b) 2 + c 2}*. Assuming this, the above corollary shows that

and so

f- =

q+

^ +S-
X
{(x
(x-a)
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.

(x-a)<f> 2 (x)+A x ,

these results,

we havef

= A +A x (x-a)+A 2 {x-a) 2 + ...+A m (x-a) m

Hence the proper fraction rx /(x a) r can be expressed as


Aft

J^L-t

a) r
any polynomial

Similarly,

<fi(x)

where A Q
4>(x)

(a;

$(x) can be written

{(x-b)* + c*}4> x (x)+A x + B0>

are constants. Proceeding likewise with

(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

{{x-b) 2 + c 2 }'

By

(iii),

A ^x + B

x + B1

+ c2}

s.x

(x-b) 2 + c 2

K
'

'

theorem shows that each of the

in proving the

unique.

(iv) is

combining

{{x-b^ + c 2 }*- 1

The argument already used


reductions

fraction rj{(x b) 2

Hence the proper

or less.

[10.53

and

(ii), (iii)

(iv),

we completely prove

in 4.62 (assuming, of course, the facts in 13.61;

the statements

made

but see the Remark in that

section).

Exercise 10(e)

Find
1

the h.c.f. of

x 3 + 3x z -8x-24, x3 + 3x 2 -3x-9.

2 2x 3 + 7a; 2 + lOx + 35, 2xi + 7x 3 - 2x 2 -3x +


3

- a: 2 + 4a: + 15,

2a: 3

a;

14.

+12a;-5.

Test for repeated factors, and hence factorise completely

a:

- 9a: 2 + 4a; +12.

x*

- x 3 + 4a; 2 - 3x + 2.

6 x - 3x 5 + 6a;3 -3x 2 -3x + 2.


Test for repeated roots,

and hence

+ 54

*9 Find polynomials A,

12a:

+ 4a:

-45a:

solve completely

8 x6 - 5x 3 +

0.

of least degree

^(2a;3 -3a; 2 + 4a;-l)

5a:

-2 =

0.

such that

+ B(a: 2 + 2a:-3) =

1.

Miscellaneous Exercise 10(/)


1 Prove a(x y ) 2hxy always has linear factors.
2 If 3x 2 + 2Axy + 2y 2 + 2ax 4y+ 1 has linear factors, prove that A must
satisfy A 2 + 4aA + 2(a 2 + 3) = 0. Is this sufficient?
2

Establish the identity


(i)

Expand

{x

+ y) to
3

Hx z 3xyz = (Sa;) (Lx 2 1>yz) as follows:


show x 3 + y 3 = (x + y) 3 3xy(x + y).

Deduce that x 3 + y 3 + z 3 3xyz = (x + y) 3 + z 3 3xy(x + y + z), and facsum of two cubes.


(i)
Verify
that x 2 + y 2 + z 2 yz zx xy = %{(y z) 2 + (z x) 2 + (x y) 2 }.
4
3
3
3
(ii) If x + y + z > 0, prove x + y + z
> 3xyz unless x = y z.
3
3
3
(iii) If x + y + z = 0, prove x + y + z = 3xyz.
5 If a: = b + c a, y = c + a b, z = a + b c, prove (using no. 4 (i))
(ii)

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 ) (b-c);

7 Use the identity

(2a:) 2

).

Sa: 2

10
[Solve the

a;

first

two equations

for p, q, using the third.]

2.

ALGEBRA OF POLYNOMIALS

389

11 'The equation (x l) 2 = A(a; 2/i) (x 4) has equal roots.' (i) If /t has a


given value, prove this statement holds for A = and one other value, (ii) If
A has a given non-zero value, prove the statement is true for two values of
ft only if A < 1. If A = 15, find the two sets of equal roots.

to divide the distance


12 Find the condition for the roots of ax 3 + bx + c =
between the roots of a'x % + b'x + c' = internally and externally in the same
ratio.

*13 If {a1 x 2 + b 1 x + c1 )/(a i x* + b 2 x + c 2 ) takes the same value


by aa x 2 + b a x + c3 = (6 2 > 4a3 c3 ), prove that

when x has

the

values given

a1

a2

&l

&jj

63

C\

C2

Cg

"

0.

[If the value taken is k, then (a 2'k a 1 )x 3 + (b 2 k b 1 )x + (c 2 k c 1 ) =


must have
the same roots as aa x 3 + b a x + c3 = 0. Hence corresponding coefficients are

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 ,

integral multiple of 7r?

If a line cuts the curve x = at 3 , y


the values tv t2 tz , prove Ti2 ta = 0.
15

= at?

ia.

three points for which

has

16 If one root of x* + ax + b
prove that the roots are

is

twice the difference of the other two,

= 0.

-136/12a, 136/3a, -136/4a, and that 144a3 + 21976 s


17 If a,
of p, q.
18

ft,

are the roots of x

+px + q =

0,

express a +fi
i

+ yt

in terms

Obtain the equation whose roots exceed by 3 the roots of


x* + 12x*

+ 49a;2 + 78x + 42 = 0,

and hence solve the given equation.


19 If a, ft are the roots of ax 2 + 2hx + b
roots are 1/tx, 1//3.

20 EUminate

x2

A,

y*

fi,

0, find

the quartic equation whose

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

= only when x = 1 if n is even, and


- 1) 0.]
Find k so that 2a;4 3a; 2 2x + k = has (i) a double root; (ii) a triple root;

cannot have a repeated root. [p'(x)


p'(x)

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

a double tangent, to the curve y = x 2 (x 2 + 4x 18).


has two double roots.]
[In (ii), the equation x 2 (x 2 + 4x 18) mx c =
and if ax 2 + bx + c, a'x 2 + b'x + c' have a common factor,
26 If 2ac' + 2a''c =
prove that at least one of these quadratics is a perfect square. [See 10.42 (1),
Remark.]
27 If a, b, c, d are constants such that ad 4= be, and /, g, p, q are polynomials
in x such that p = af+bg, q = cf+dg, prove that /, g and p, q have the same
h.c.f. What happens if acZ = bet
inflexional tangent,

(ii)

*28 If p{x,y) is a symmetrical polynomial in {x,y) having factor x y, prove


that actually (x y) 2 is a factor. [p(x,y) = Hars (xry s + x syr ). The conditions
for p to be zero when x = y are the same as those for dp/dx to be zero when

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* +

The identity in no.

shows that

BHx + G =
*/ y, z

(a)

0.

can be chosen so that

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 -Gt-H
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.

C 2 + 4H3 = 0, show that the right-hand side of the identity in


no. 3 becomes (x + 2y) (x y) 2 so that (a) has three roots, two of which are
equal, viz. - 2y, y, y.
3
2
(iii) If Q + 4H < 0, numbers satisfying (6) do not exist. The following
(ii)

If

trigonometrical method can then

Put x

= kcosd

in (a),

be used.

and choose h so that k 3 :3HJc

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 So-called, although discovered

by

Tartaglia.

391

11

DETERMINANTS AND SYSTEMS


OF LINEAR EQUATIONS
11.1

Linear simultaneous equations

11.11

Two

equations in two unknowns

The usual method of solving


ax x + bx y

cx ,

a2 x + b 2 y

c2

by elimination shows that


(ax b 2

-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

4= 0, these give unique values for x and y.


a x b 2 a 2 b x = 0, we may assume throughout that a x and b x
= b x the first
(and likewise a 2 and b 2 ) are not both zero; for if a x =
= c x which is either false or trivial. Two cases
equation becomes

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.

The given equations

is

not zero,

therefore have

no

(i)

gives a

solution,

and

are said to be inconsistent.


(b)

If a x b 2 a 2 b x

cx b 2

c2 b x = ax c2 a2 cx =

0,

then

(i)

gives

no

information. Since a x b x are assumed to be not both zero, suppose


,

ax

The

#= 0.

first

of the given equations can then be solved for x in

terms of y. When y
x be xn then

is

given the value y

X~

ax

Cl
+ b 2 y - c2 = ^ a
CT

a2 x

let

the value obtained for

Since

CT

2 Cl

+ b 2 y - c2
x

~ ai C2 +

a l & 2 ~ CT2 frl) t/0

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)

DETERMINANTS, LINEAR EQUATIONS

392

[11.12

coefficients are 'proportional', so that the

shows that corresponding

equations are not distinct.)


If

Geometrically, the given equations represent straight lines.

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,

parallel, while in (b) the equations represent the

11.12 Three equations in three

to

same line.

unknowns

Here we obtain only the general form of the results corresponding


(i), and postpone discussion of details until 11.4.
Elimination of z from the first and second of

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

and ehmination of z from the second and third gives


(a2 c 3

-a 3 c 2 )a; + (& 2 C3-&3C a )^ =

d 2 c 3 -d6 c 2

Now ehminate y from these last two equations:


{(b 2 c 3 - 63 2 (a x c 2 - 2 c x - (b x c 2 - b 2 c x (a 2 c 3 )

(b 2 c3

- 63

2)

(d x c 2

On simphfying we find that the


c 2 (ax b 2 c 3

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.

a i h c 3 - a i &3 c 2 + a 2 &3 c i -a2 b x c 3 + a3 b x c 2 - a3 b 2 c x


should find similarly that this

expressions for y

3 c 2 ).

x is

and that the right-hand side reduces to c 2 times a


The solution for x therefore has denominator

We

and

is

also the

(ii)

denominator in the

z.

11.13 Structure of the solutions

Instead of proceeding to higher eliminations,


properties of the expression

we notice the following

(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)

DETERMINANTS, LINEAR EQUATIONS

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,

find that the

number of interchanges

the term a3 b x c2 the number

terms for which the number


those for which it
that the

is

is

is

odd. Similarly, for

even. It is easily verified that all the

is

odd are preceded by the

evenf have

(For a given term

number of interchanges

is

either always

it

sign

while

can be proved

odd or

else

always

even, no matter how the rearrangement to natural order is carried out.)

Hence the sign of each term in

(ii) is decidable by the parity of the number


interchanges
of
of its suffixes from the actual to the natural order.

Determinants

1 1 .2

11.21 Determinants of order 2

The symbol

A=

ax

bx

a9

b9

called a second-order determinant, is defined to

mean a x b 2 az b x The
.

following five properties^ of A are easily verified.

The value of A
That is, if
(1)

is unaltered

A'

then A'

A.

by interchange of rows and columns.


a-,

b2

We call A' the transpose of A.

By use of this result, any property proved for rows extends at once
to columns,

and conversely. The following

will therefore

be stated for

rows only.
(2)

Interchange of two rows alters only the sign of A.

That

is,

a2

b2

ax

bx

= -A.

A=

(3)

// two rows are identical, then

(4)

Multiplication of any one row by k multiplies

For example,

t
X

is

kb x

a9

ft,

A by

k.

= M.

reckoned as even.

With the wording

minants

kax

0.

(11.22).

used, they hold without modification for third-order deter-

DETERMINANTS, LINEAR EQUATIONS

394
(5)

Addition

to

[11.22

any row of a multiple of another row does not

alter

the value of A.

For example, by adding k times the second row to the

we obtain

a x + ka 2

+ kb a

bx

a9

first

row,

A.

b9

11.22 Determinants of order 3

The symbol

ax

A=

a 2u

b9

called a third-order determinant,


c2

defined to

is

-&1

c3

a2

Co

as

c3

mean

+ CX

a%

b2

a3

b3

(i)

which (by using the definition of the second-order determinants)

is

equal to

~ b i( a 2 c 3 ~ % c 2) + Ci(a 2 6 3 - a3 b 2
= a 1 & 2 C3-a 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

referred to as the expansion of A.


c3 are called the elements of A; with
The numbers a x b x c v a 2
our notation, the suffix denotes the row and the letter denotes the
column in which a particular element lies.
The remarks in 11.13 show that the expansion of a third-order
,

determinant consists of 6 terms, each of which involves an element from


each row and each column but no two elements from the same row or the

same column. The sign before each term


according to the rule in 11.13 (/?).

shortly as

The

^a

definition expressed

by

define determinants of fourth

how

(i)

bj ck

ax

is

determined by the suffixes


write the expansion of

above can be readily generalised to

and higher orders

the second-order determinants in

coefficient of

is

We may

(i)

(see 11.7). Observe

are constructed

the determinant obtained from

from A: the

A after deleting the

row and column containing a x the determinant in the middle term


is got by omitting the row and column containing b x from A; and
;

Since the elements a x b x c x appear as


multipliers of the second-order determinants, the expression (i) is
consequently referred to as the expansion of A from the first row.
similarly for the last term.

DETERMINANTS, LINEAR EQUATIONS

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(8-42)-5(4-21) + 3(12-12)
= -51.
(ii)

-h

+9

= a(bc -/) - h(ch -fg) + g(hf- bg)


= abc + 2fgh - a/ 2 - bg 2 - ch\

The reader should now do Ex.

11 (a), nos. 1-10.

11.23 Other expansions of a third-order determinant

We may arrange the expression

(ii)

of 11.22 according to elements

from any row of A and the corresponding second-order determinants


formed by deletion as described above. Thus, grouping by elements of
the second row,

A = -a a (6 1 c3 -63C 1 + 6 a (a 1 C3-a3 c 1 )-C2(a1 63-a 3 6 1


)

= -a 2
which

is

b3

+h

c3

ax

cx

ax

bx

as

c3

az

bz

(iii)

the expansion of A from the second row. Similarly,

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)

the expansion from the third row.


Likewise,

we may arrange

A=
=
is

(ii)

by elements of any one column

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*

the expansion from the

first

+ 0-

(v)

column.

The reader should not attempt

to memorise these results because


they are all easily obtainable from the definition of A in expanded form.
28

GPMII

DETERMINANTS, LINEAR EQUATIONS

396

11.24 Properties of

We now

[11.24

show that the

properties (l)-(5) in 11.21 hold for third-

order determinants. In view of the

first, all

subsequent properties

hold for columns as well as for rows.

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,

~ & s c i) + %( & i c 2 - &2 c i)

= A by the line preceding

(v) in 11.23.

Proof of (2). Interchange of two rows is equivalent to interchanging


two of the suffixes 1, 2, 3. Results (i), (iii), (iv) show| that the sign of

changed in every case.


start from the definition;
and third rows gives

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)
)

on arranging by elements of the first row of A; and this is clearly A.


A similar direct proof would hold for other row interchanges.
Proof of (3). By interchange of the two identical rows we obtain
A, by (2). However, interchange of identical rows clearly leaves the
same determinant A as before. Therefore A = A, so A = 0.
Proof of (4). Since each term in the expansion of A contains exactly
one element from each row, multiplication of each element in a given
row by k causes every term in the expansion to be multiplied by k.

Hence the value of the new determinant is &A.


Proof of (5). We expand the new determinant by the row which
f Using the corresponding property already
sary.

proved for second order when neces-

DETERMINANTS, LINEAR EQUATIONS

11.25]

has not been mentioned. Thus,


first

row,

we

get

a x + ka 2

397

if

we add k times the second row to the

bx

+ kb 2

cx

+ kc 2
(vi)

The row

'not mentioned'

expand from
b t + kb 2

it,

cx

the third; using an equation like

(iv)

to

we see that the new determinant is equal to


a x + ka 2

+ kc2

60

is

+ kc

cx

ax + ka 2

a2

bx

+ kb 2
b2

+ c.

a9
b9

by using Property (5)


Again by equation (iv),

CL

C2

for each of the second-order determinants.


this expression is A.

Remarks

many extensions of Property (5). Three are indicated


in Ex. 1 1 (a), no. 22; but see the Remark about random manipulations
(a)

There are

after no. 29.


(/?) The operation by which determinant (vi) above is obtained
from A can be denoted by r1 -> r x + kr 2 With this notation, properties
(4) and (5) can be combined in a single statement:
.

//

rt

-^k 1 r1 + k2 r2 + k3 r3

A -> kt A

then

(i

or 2 or

3).

11.25 Examples

ways of simplifying a determinant when


direct application of the definition would be clumsy owing to the large
numbers or heavy algebra involved,
Properties (2)-(5) give

(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

ra -> ra rx , this becomes

(These two steps

may be

condensed into one by saying r8


'

r.

r2

followed

28-2

DETERMINANTS, LINEAR EQUATIONS

398
The

last

determinant

is

now

expanded, according to the definition, as

easily

-29

35

[11.25

+ 86

= 35.3-29.0 + 86(-3)
= - 153.
(ii)

Prove

+c

By rx

->

l*x

+ r2 the determinant

is

a+b

+a

0.

equal to

a+b+c a+b+c a+b+c

(a + b

=
(iii)

+ c) a

on removing the factor (a + b + c)


from the first row,

since the last determinant has

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
,

y-x

z
z

x
xz

= (y-x)(z-x)
y 2 + xy + x 2

Xs

by removing the

factor y x

pandingt from the

first

from column

2,

and

row, the only non-zero term

%
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

(y-x) (z-x)

+ xz + x 2

z2

z*

+ xz + x 2

(y

f Before expanding a determinant


as here. Also see ex. (v) below.

it is

on

factorising the last bracket,

on arranging

cyclically.

helpful to get zeros in a

row

or column,

DETERMINANTS, LINEAR EQUATIONS

11.25]
(iv)

399

Without expanding either determinant, prove

a2

ca

ab

c2

be

Introducing a factor a in row

we have

be

6 in

1,

a2

a3

63

row 2, and c in row

abc

ca

bz

abc

62

63

a6c

aa

a"

62

6s

abc

abc

abc

ab

3 of the left-hand side,

on removing the factor abc from the

first

column of the middle determinant,

(v) 'Triangular' determinants.

ax
62

o2

62
63

C8

Ca

&3

the product of the elements in the leading diagonal. Similarly, a determinant


whose elements below the leading diagonal are all zero is thus readily evaluated.

Remarks
(a) In expanding a determinant from a given row (or column), the amount
of calculation is reduced if row- and column-operations 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?.

Verify the following equations by expanding each side.

a2

62

c2

ca

ab

ax

by

cz

zx

xy

be

ax

by

cz

yz

abc

111

y
be

ca

ab

DETERMINANTS, LINEAR EQUATIONS

400

a+A

10 Verify that

+A

+A
= \* + (a + b + c)X 2 + (bc + ca + ab-p-g*-h*)\ + {abc + 2fgh-af i -bgi -ch*).
Using

the properties of determinants, evaluate the following.

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*-bc-ca-ab).

bx

19

a2
22 Prove

X*

Deduce an identity by comparing this


Prove

16

111

18

cab
21

13

13

a b a+b

17

result

(6

with that of no.

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

are the values of

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-*

2-a;

4-jk

0.

c x + c 2 + c8 and remove 9 x; then r 2 -> rt rlt r3 -> r8 rx .]


,

DETERMINANTS, LINEAR EQUATIONS

11.3]

401

24 Without expanding, prove that


be

1
1

ca

ah

and hence evaluate the

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

(b-c) (c-a){a-b)(x-y).

ca

[Use the result of no. 21.]

JfA,B,G are the angles of a triangle, prove that


1 cos C cos B
1 cos A =0;
cos G
1
cos B cos A

26

and by expanding, obtain a


a

relation

between cos A, cosB, cosC. [Use

b cos C + c cos B, etc.]

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

^tote <Ais reswft cw

28

(i)

equal to the

is

sum of 8

deteraainants. [See no. 27.]


(ii) If each element of A consists of the sum of three terms, how many
determinants are there in the expression for A as a sum of determinants?

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

tempted to say that D is formed from A by the operations c t -> c x + Ac 2 c 2


c 2 + /tc3 and c 3
c3 + w^. A choice of A, [i, v such that
Xfiv = 1 makes D = 0, and appears to prove that A = 0; but the compound
manipulations of columns just described are a misuse of Property (5) and have
Remark.

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

Minors and cofactors

11.31 Definitions

and notation

The expansions of A from


in 11.23)

all

its

various rows or columns (written out

follow a pattern similar to the original definition: the

three terms each consist of an element from the particular

column from which the expansion

is

row or

being made, multiplied by the

[11.31
DETERMINANTS, LINEAR EQUATIONS
second-order 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-

We now introduce a notation which will make all this systematic.


Definitions

The minor of an element of A is the second-order determinant


obtained from A by deleting the row and column in which that element
(a)

For example, the minors of b x a2 are respectively

lies.

(b)

The

a2

c2

az

c3

>

element in the

cofactor of the

the minor of that element multiplied


cofactor of b x which
,

(_

is

c3
ith.

by (-

the element in the first row, second column,

is

is

a2

1)1+2

az

^3

that of c z

row and jth column is


<+
the
1) *. For example,

ax

=+

(-1) 3+3

c3

a2

&2

Cofactors are 'signed minors'.


c 3 in A by the
We denote the cofactors of ax b v c a
expanvarious
The
A
C
A
B
C
letters
capital
corresponding
2 ,...,
z
v
x

Notation.

sions of A can
first,

2,

lf

now be written concisely and uniformly: those from the

second and third rows are

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)'

The reader should verify the last two as in

11.23.

These six results are

summarised in the following rule.


The expansion of a determinant from any row {or column) is the sum
of the elements of that row {column) each multiplied by the corresponding
cofactor.

DETERMINANTS, LINEAR EQUATIONS

11.32]

11.32 Expansion

by

403

alien cofactors

Consider the determinant

a2

b2

a9

b9

a3

^3

two rows identical, and consequently is zero. By expanding


from the first row according to the definition, we have

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=

u-2

u2

a*

bo

In this way a total of 12 expressions each equal to zero can be formed by


sum of the elements in any row (or column) each multiplied by
the cofactor of the corresponding element from a different row (or
column). These results, which may be called 'expansions' of A by
'alien cofactors, are useful when employed in conjunction with those
taking the

'

of

1 1 .3 1

for expansion of

A by

'

true cofactors.
'

Examples
(i)

A=

If we interchange rows and columns, we obtain the same determinant; A is said


to be a symmetric determinant, meaning that it is identical with its transpose.
The cofactors are

B = ac-g
C = ab-h
A = bc-f
F = gh-af, G = hf-bg, H =fg ch.
2

Owing to the symmetry, we shall get only three


and six expansions by alien cofactors:

different expansions

cofactors,

A = aA+hH+gG = hH + bB+fF = gG+fF + cC;


= aH + hB + gF = aG + hF + gC
= hA + bH+fG = hG + bF+fC
= gA +fH + cG = gH+fB + cF.

by true

DETERMINANTS, LINEAR EQUATIONS

404

[11.4

This determinant arises frequently in coordinate geometry, and the use of


the above relations often saves much algebraic manipulation. The reader is
invited to verify some of them by direct substitution (see 11.22, ex. (ii)).
(ii) The proof of Property (5) in 11.24 can be expressed more concisely in the
notation for cofactors.

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

can be done in a similar way.

(a)

Determinants and- linear simultaneous equations

We now return to the subject of systems of linear equations

(11.1),

and apply our knowledge of determinants to the problem of solving


them.
11.41 Cramer's rule

Consider again the system

ax x + bx y + cx z

a2 x + b 2 y + c 2 z

= d2

a3 % + b3 y + c3 z

dx

d3

Multiply these equations by the cofactors of a x a 2 a 3 in


,

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

alien cofactors the coefficients of

Hence
dx
X

bx

d2

d3

b3

cx

Co

sum d x A x + d 2 A 2 + d 3 A 3 is the expansion from the


column of the determinant obtained from A by replacing its

because the

first
first

DETERMINANTS, LINEAR EQUATIONS

11.41]

405

column by the elements dv d2 dz appearing on the right-hand


,

sides

We may therefore write

of the given equations.

Az = A

(1)

where

(1)

A with its first column modified as just indicated.


the given equations by the cofactors of b x b 2 b s in A

denotes

Multiplying

^=

and adding gives


and

=|=

0,

A
= -
A
(1)

The

case

(3)
.

the solution of the given equations

A=

will

A< 2>

'

A(2)

=A

Az

similarly

If A

is

=A

unique and isf

(3)

Z
'

be considered later

'

(11.42).

Example

5,

3x-8y + 6z =

4,

Sx-2y-9z =

12.

2x + 4y-Sz

Solve

Here

A=

= 6x

2x 3x

-1

-4

-1

-3

by cx -^Cj-Cg,
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.

f Strictly, what we have just proved is that, if there are numbers x, y, z


satisfy the three given equations, then these numbers also satisfy Ax
A (1> , Ay

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.
.

DETERMINANTS, LINEAR EQUATIONS

406
Similarly

(2)

and

5-3

12

The

11.42

case

3 x 49,

so

-h

2 x 49,

so

z-

3-8

-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 >.

follows that, if the system possesses a solution,

A<

A<

2>

A<

3>

we must have

0.

(i)

If one or more of the determinants A(1) A(2) , A (3) is non-zero, 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

one cofactor in A is non-zero.


4= 0,f and consider the two equations in two

a^ + ^y =
a2 x + b 2 y

unknowns J x, y:

dx c x z,'
(")

d 2 c 2 z.

C3 4= 0, these can be solved for x and y (as in 11.11) in terms of z. When


given the particular value z let the values obtained for x, y be x , y
We may actually solve (ii) for x and y, and show by direct substitution that
the values x Q y z do satisfy the third equation of the system. More elegantly,
however, we can prove this as follows.
Since (x ,y ,z ) satisfies (ii),
Since

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.

these values satisfy the third equation of the system

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,

f This can always be made the case if necessary,


rearranging the equations.
% Associated with the elements in G 3
The reader may try this for himself.
.

+ Az -A< 3 = C3 u.
>

by renaming the unknowns and

DETERMINANTS, LINEAR EQUATIONS

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

but at least one element of

Then the

When

y,

Lis non-zero.

of the given equations can be solved


let the value
z are given the values y z
first

Xq

and

so

a2 x

+ b2 y + c2 z d2 =

^i-fri2/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

The given equations

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 left-hand 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

the numbers on the right-hand sides are

obtain the homogeneous system

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

DETERMINANTS, LINEAR EQUATIONS

408

Theorem
x, y, z

I. If the homogeneous system


which are not all zero, then A = 0.

Proof. Proceeding as in 11.41,

which

satisfies

is satisfied

[11.43

by values of

we find that any set of values

(x, y, z)

the given equations must also satisfy

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=

In the particular case z

= 1 we obtain the non-homogeneous equations

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.

have a common solution, say (x y ), then the corresponding


homogeneous equations have the solution (x y 1) which is certainly
not all zero, whatever x y may be. Hence
Cobollary I (b). If the non-homogeneous equations possess a common
If these

solution, then

A=

0.

direct proof of this result

Corollary

1(c).

If

is

4= 0,

indicated in Ex. 11
the

(b),

no. 14.

non-homogeneous equations are

inconsistent.

For by Corollary I (a), the only solution of the homogeneous system


so (x, y, 1) can never be a solution for any choice of x, y.

is (0, 0, 0),

sometimes called the eliminant of the non-homogeneous


A = is the necessary and sufficient relation between the
coefficients in order that the three equations in two unknowns x, y can
is

system; for

be

satisfied simultaneously (cf. 10.41).

Geometrically, the results of Corollaries 1(6),

with concurrence of lines in a plane


to

Theorem

7/ A =

by values of x,

are associated

Theorem I we have

As a converse
II.

(c)

(15.42).

0,

then the homogeneous equations are satisfied

y, z not all zero.

First proof. The discussion in 11.42 shows that, when A = 0, we have a


solution containing at least one arbitrary unknown which can be chosen to be
non-zero. (The homogeneous equations are automatically consistent because
they have the solution (0, 0, 0).)

Second proof. Independently of 1 1 .42, we can prove the theorem by


actually writing down a solution of the required type.

DETERMINANTS, LINEAR EQUATIONS

11.43]
(1)

Suppose that at least one cofactor in A is non-zero, say Bz

x = XA Z

= AB3

a solution not all zero, where A is non-zero but otherwise


For
(alien cofactors),
ax A z +b x B z + c x Cz =

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 non-zero, say c x

x
Choose

Ac 1?

Let

=# 0.

in

=j=

0.

Then

= AC3

409

by

arbitrary.

hypothesis).

are zero, but that at least one element

y have arbitrary non-zero values

x,

= /ic x

(so

and

4=

/i =# 0).

z to satisfy the first equation:


z

-(Aaj +/&!).

With these values, the left-hand side

of the second equation

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

hypothesis. Similarly the third equation

is satisfied

by the above

values.

Two homogeneous equations in three unknowns


If A z Bz Cz are not all zero, then the equations
(2)

solution in ratios.

ax x + b x y + c x z
have a non-zero 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

where if a particular denominator is zero, the corresponding numerator


must be interpreted as also.
This simple result will be useful

later, especially in

Ch. 21.

DETERMINANTS, LINEAR EQUATIONS

410

Exercise 11(6)
Solve the following systems of equations by
1

2x-y + 5z =
x + ly-IOz =
a? + f/ + z =

*4 If a,

6, c

2.

Zx + y- 4z

2,

x-2y-3z =
Zx y + 5z =

1,

of determinants.

vise

x-y-z =

2,

5x-y + Zz =
oj + y z =

1,

4.

13,
5,
3.

are all different, solve

x+y+z

aa;

1,

+ 6t/ + cz =

= d2

a 2a; + b 2y + c 2z

(2,

[Use Ex. 11(a), no. 21.]


5

Prove that the following system of equations


Bx + y-z

[Solve the
the third.]

Show

first

two

for x,

x-4y + 2z =

0,

y in terms of

z,

inconsistent:

is

4x-Zy + z =

9,

and

l.

substitute in left-hand side of

that the solution of the system

xyz

3x + Zy-z

0,

2x + y-z

Q,

indeterminate, and can be written

is

where A
7

is

1 + 2X, y= 1-A,

arbitrary. [Solve the first

Show that the

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.

8 Find the values of A for which the equations

have a solution other

=
=
V2 + V22/ + (A-2)z =
than x = y = z = 0. Find
\x + y + j2z

0,

x + Xy + j2z

0,

also the ratios x-.y.z

which

correspond to each of these values of A.

9 Find the values of A for which the equations


3x + Xy

5,

Ax 3y

are consistent. Solve the equations

10 Solve

(if

4,

when A has

3x y

=l

these values.

possible) the equations

= 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=

geneous equations for

a, b, c.

DETERMINANTS, LINEAR EQUATIONS

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.

then the non-homogeneous system of equations


has the solution (<x 2 ,a); use
Q, lx + my + n =

root,

px + qy + r =

0,

11.43, Corollary 1(6).]

14 Find from

the condition that the equations

first principles

a 1 x + b 1 y + c1

a2 x + b 2 y + c 2

0,

are satisfied by the same values of x


the last two for x, y in the form

b2

c2

b3

c3

a3

substitute these solutions in the


If the

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,

has a solution other than x = y = z = 0, show by means of no. 14 that


[If z 4= 0, the equations can be written a^x/z) + b x (ylz) + c x = 0, etc.]
Verify in detail that

16

solve

satisfy the equations.

when A

[The solution

#= 0,

is

Az

the solutions given in 11.41 actually

given

Ax = d^i + dzAz + dsAs,

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

we have examples in which a

expressed as a product of factors without expanding

Here we consider various methods of factorising a deterwhen other methods

minant. Direct expansion should be used only

cannot be further applied.


(i)

Use of the Remainder Theorem.

Factorise

A=
29

a2

62

c2

GPM II

DETERMINANTS, LINEAR EQUATIONS

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.

where k must be numerical because


nomials in

a, b, c

We

A = (b-c)(c-a)(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.

k(a z + 6 2 + c 2 ) + l(bc + ca + ab),


numerical.
where k, I are
Clearly the expansion of A contains no term in a4 while the factorised form
does unless k = 0. Hence
,

A=

By comparing
a, 6, c,

l(bc

coefficients (e.g. of 6 2 c 3 ), or substituting particular values for

we find I = + 1. Hence

A=
(iii)

+ ca + ab) (6 c) (c a) (a 6).

(6

c)(c a)(a b) (bc + ca + ab).

Use of row- and column-operations

to

make known factors appear

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,

DETERMINANTS, LINEAR EQUATIONS

11.6]

=
=

(b

(6

+ c+a

+a

a)(c a) (b c)

+a
+ ca + a2

by c x

by

r 2- r i

= (b-a)(c-a)(b-c)

b+a

c 2 and

removal of 6 c,

r2

a2

cx

413

c) (c a) (a 6) (6c + ca + a6)

on expanding and rearranging

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*

<SoZe the following

systems of equations by use of determinants.

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

If the elements of A are functions of x, then dAjdx is equal to the sum


of the three determinants obtained from A by deriving one row at a time

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
29-2

DETERMINANTS, LINEAR EQUATIONS

414
Proof. It

is

[11.7

convenient to use the notation (mentioned in 11.22)

A = I a
l

for the expansion of A.


cZA

^ = S

bj ck

Then

da,

.,

^ + Sa^

dbi

6
rfx

dc k

^-

+ Sa' 6

The first sum is the expansion of the determinant obtained from A


by replacing the top row a x \, c x by dajdx, dbjdx, dcjdx; the other
two sums are interpreted similarly. This proves the theorem.
A similar result holds with 'rows' replaced by 'columns'.
,

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=

when x = a, (x a) 2 is a factor of A by 10.43.


Second solution. Subtract any one of the rows from the remaining two rows.
When x = a, each element in each of these new rows vanishes; hence by the
Remainder Theorem x a is a factor of each row, so (x a) 2 is a factor of A.
Determinants of order 4

11.7

dx

The symbol

d2
d3
d.

called

a,

fourth-order 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

a direct generalisation of the definition of 'third-order determinant' in

11.22.f

We mention fourth-order determinants because they arise occasionally in


the coordinate geometry of three dimensions. It can be shown that
(i) the properties (l)-(5) in 11.21 continue to hold;
(ii) the determinant can be expanded from any one row or any one column;
(iii) the concept of cofactors is still valid and useful;
(iv) there are expansions by true and alien cofactors.
They can be used to solve systems of four linear equations in four unknowns
by the obvious extension of Cramer's rule (11.41), and in elimination problems.
'

t Determinants of order

'

n can be defined

similarly in terms of those of order

n 1.

DETERMINANTS, LINEAR EQUATIONS

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,

w are polynomials in x whose degrees do not exceed


U V w
A=
w

3,

prove that

w"
is

also a polynomial in x of degree not greater

than

[Derive

3.

A wo x four times,

rejecting zero determinants at each stage.]

6 Prove
a2

dxdy

X
7 If

A=

X
X2

by rows, regarding

[Derive

2(y-x).

prove

3A

SA

8A

ox

oy

oz

z2

8/8x + 8/8y + d/8z as operating

on the rows.]

8 If

8A

A=

find

8x

xs

y*

8A
8y

8A
8z'

zA

Evaluate the following fourth-order determinants.

*9

*10

-3

*11

-1

*12 Prove

= pA- (Al 2 + Bm 2 + Cn 2 + 2Fmn + 20nl + 2Hlm),

where the notation is that of 11.32, ex. (i). [Expand by the last column, then
by the last rows of the third-order determinants so obtained.]

DETERMINANTS, LINEAR EQUATIONS

416
*13 Solve

by determinants:
3x +

2/

-5z

+ 2z + Zt

y + Bzx + 2y
Prove

(i)

(ii)

z--2f

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;

Obtain identities by considering the cofactors of


cients of Xs x 2 x in the product.
,

a?,

x 2 x in
,

Miscellaneous Exercise 11(e)


1

Without expanding the deteraiinant, explain why

is

the equation of the line joining the points (x u yj, {x it y a ).

Without expanding, prove

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-

DETERMINANTS, LINEAR EQUATIONS

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"

Show that a and a + 6 + c are

(6

factors of

+ c) 2

62
(c

ca

+ a)

c2

62

and hence

+ 6)

By factorising the determinants, prove that


2
a+6
a2 + 6 2
1
1
a + b a 2 + 6 2 a8 + 6 8
a 6

c2

a2

62

c2

cos 6 are factors of

10 Prove that sin 6, cos 6, sin

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

find the remaining factor.

11

Express

ap + br

aq + bs

cp + dr

cq + ds

as the sum of four second-order determinants,

X
(This gives a rule for multiplying

and hence show that it is equal to

two second-order determinants.)

Solve

12

x+1

2x

Sx-2

2x

13

0.

X2

a3

62

as -6

(assuming that

a, b, c are

non-zero and distinct).

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

DETERMINANTS, LINEAR EQUATIONS

418

Show that g,f,

15

cannot be found so that the equation

x 2 + y 2 + 2gx + 2fy + c
is satisfied

by the values

Show that

16

(1, 1), (0,

- 2),

of

(2, 4)

=
Interpret geometrically.

(x, y).

the equations

2x + 3y

can be consistent whenever

= 1, Ax-2y
4c 156c 439 > 0.

3x + Xy

4,

c satisfies

17 Discuss the equations


a;

when

(i)

+ + 3z = 4,
2; (ii) a = 1,

+ 2?/ + 4z = 5. a-y + 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.

19 Prove that, for two values of A, the equations

= 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

thus chosen, find the

general solution for each of the values of A.

20 If/,

g,

h are functions of x which


d

L ==f+g + 2h,

^=

satisfy the differential equations

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,

and show/<r)(a;) = Oforr >

4.

Using Maclaurin's theorem, prove


f(x)

and deduce that


(x + a)(x + b)(x

Vr

(r

+ c)V3

=/(0)

1, 2, 3)

(Cf.

+ VyX + V2 x2 + Vs x3

is

the coefficient of xr in the expansion of

Ex. 11(d), no.

15.)

419

12
SERIES
The binomial theorem

12.1

The reader should already be familiar with the work in this section;

we include it for revision and completeness.


12. 11 If n is
(x

a positive

integer,

+ a) n = xn + n Cx xn ~ xa + n C2 xn ~ 2a 2 +

where nCr denotes the

number

+ n Cr xn-^ar +...+an

. .

of selections ofr different objects from n, viz.

n(n 1)

...

(n r+ 1)
'

r\

Proof. Consider the product (x + a t ) (x + a 2 )


it

we must multiply each term

...

in each bracket

(x + an ). To expand
by the terms in the

other brackets. This can be done systematically as follows.

we obtain xn

If we multiply the x from each bracket,


If

we

multiply the a-term from one bracket and the a;-terms from

the other brackets, and do this in


(a 1

all

possible ways,

+ a 2 +...+an )x

we obtain

n-1
.

Next, multiplying the a-terms 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

taken two at a time. The number of terms in this

therefore nG2

from three brackets and


ways gives

Similarly, multiplying a-terms

from the

en-

coefficient is

rest in all possible

{a x a 2 az

+ a1 a 2 ai +

..)xn

ar-terms

~3
,

~
where the number of terms in the coefficient of xn z is n C3
In general, if we multiply a-terms from r brackets and aj-terms from
.

the remaining n r brackets in

all

possible ways,

(a1 a 2 ...ar +...)w

where the coefficient consists of n Cr terms.

-r
,

we get

series

420

by multiplying a-terms from

Finally, the product obtained

brackets

ax a2

is

...

an

[12.12
all

We now put a1 = a2 = ... = an = a. The original product then


becomes (x + a) n In the expansion, the coefficients of cn_1 xn ~ 2
xn ~z
xn -r ... become
.

nC a 2
2

na,

nC

and the constant term becomes

az

n C ar
r

an This proves the result stated.


.

12.12 Properties of the binomial expansion

The expansion of (x + a) n consists of n + 1 terms.


The (r + l)th term is n Cr arxn -r and is called the general term.
The coefficients of terms at equal distances from each end of

(1)
(2)
(3)

the expansion are equal.

For the (r+ l)th term from the beginning is


(r+ l)th term from the end is n Cn _r an -^xr and

nC

arxn ~r while the


,

nC

r\{n-r)\

= n Cw ~r

This fact saves work in numerical calculations.

by r =

(4) If

If n

is

odd, there are two (equal) middle terms given by r

(5)

We can obtain the expansion of (x + a) n in ascending powers of

is

even, there

is

a middle term, given

x by interchanging x and a in the


{x + a) n

This

is

a for a in

(x - a) n

= xn - n Cx ax

71

-1

. .

+ xn

(i)

Find

when x is small compared with a.


we obtain

+ n C2 a 2x n ~ 2
+ - l) r nCr arxn ~r +... + (- l) n an

the coefficient of x in the expansion of (2x

This will involve x 9

if

^_^

+ x- 3a; 2

5
)

(2c)

18 -'

18a'(

-3

r213 "ral8 ~ ar
)

13 2r
13

Expand

Bjx) ls

is

i3Cr

(2

Examples

The general term

(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.

= 9, i.e. if r = 2. Hence the required coefficient is


C 2 (-3) 2 2 n = 13.3 3 .2 12
.

+ x 3x 2 6 in ascending powers of x as far as the term in x*.


= {2 + *( 1 - 3a;)} 6
= 2 5 + 8 C X 2 lx{ 1 - 3a;) + 6 C2 2 82 1 - 3a;) 2 + 6 C3 2 2a;3 1 - 3a;) 3 +

(2

. .

series

12.13]

where the unwritten terms


such terms, we have

all

421

contain no power lower than

32 + 5. 16(x- 3a; 2 ) +

- Qx) + ^^4z*+ ...

8x 2 (l

32 + 80* - 160a 8 - 4408 +

On ignoring all

x*.

. .

after collecting like terms.


(iii) Find the numerically greatest term in the expansion of(l 3*)' if x =
J.
The terms in this expansion are alternately positive and negative, but the
numerically greatest term will be the same as the corresponding term in the

expansion of (1

+ 3a;) 7

If the latter expansion

is

u1 + ui x + u3 x + ...+uB x7
2

7!

then

ur

= _ __

7!

r!(7-r)!

(toy
/
v
'

(r-

1)1

(8-r)!

1
(3*)*v
'

= 8-r 3x
r

.-

if

l.

Now uf+1 ^ ur according as ur+1 jur ^


3(8 -r)

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
,

than the other terms. Thus the 4th term


7

(iv)

Cs(-i) 3
n

Prove

is

the greatest. Its value

is

-35x(f)3.

Gr _Y + n Gr =

n+1 G

r.

This can be verified directly, or obtained from the identity

+ 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.

When the coefficients n Gr in the expansion of (1 +x) n have been calculated,


the above relation shows how to obtain those in (l+x) n+1 The following
scheme, in which each number is formed by adding the two immediately above
.

it, is

called Pascal's triangle:

(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

(l-x){l+x) i

in ascending powers as far as the

6
)

term in xl

Write down and simplify the

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

Evaluate correct to 4 places of decimals

12 Find r

1 \x

+ x) 20

(1-04) 5 ;

(i)

is

(ii)

+ x - 3a; 2

(0-98) 5

7
)

~
twice the coefficient of xf x

Find the greatest term in the expansion of (5 + 4a;) 11 when x


14 Find the greatest coefficient in the expansion of (5 + 4a;) 11
r
15 Obtain an identity by equating coefficients of x in
13

(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

16 Consider the coefficient of xr in

17

x) m ( 1

where

cr

denotes

Cr

+ c + c +...+c =
2

^L,

18 Prove
(1

+x) n+1 -x n+1

(l+x) n + x(l + x) n - 1 +...+xr {l +x) n ~ r +

Deduce an identity by equating

Show that

19

(2

...

+x n

coefficients of x

+ ^3) 5 + (2 JS) 5

is

rational,

and

find its value.

20 Prove that
v

x + n(x + y)+

n(n
v

1)'

(x

n
+ 2y)+
x

n(nl)(n 2)

-(x + 3y)

+ ... to (w+1) terms

=
12.2

2 n ~\2x + ny).

Finite series

The reader
progressions,

will already have met arithmetical and geometrical


and perhaps other simple series. We now extend this

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

of the finite series

%, u 2

.
,

un

in the language of 2.71 ; but when a sequence


to its sum-sequence sn , it is usually called a series.

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.

lim sn provided this limit exists, and

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
= l-x

t^;

summing

Given a

series,

our problem

terms.

The methods

first

of

(x 4= 1)

...

when n-^ao

12.22 General methods for

the

sum

of the geometrical progression

is

(2.72), so sn -> 1/(1

a).

finite series

to find a formula for sn the


,

(l)-(3) listed

below

will

sum

of

be illustrated

in this section.

known

power series and its


sum-function (followed perhaps by substitution of some particular
(1)

Derivation or integration of a

value for
(2)

finite

x).

The

difference method. If

a function /(r) can be found so that

ur ^f{r+\)-f{r),
then by taking r

n,

1,

n 2,

. .

.,

2, 1

un-i =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)

Use of complex numbers and de Moivre's theorem. This method,


many trigonometric series, will be considered in 14.5.

applicable to

(for

result).

series

424

[12.23

12.23 Series involving binomial coefficients

As our standard form of the binomial theorem we take

= l+ n C1 x + n C2 x2 +...+ nCr xr +...+xn

(l+x) n

Used

thus, the

numbers

nf]

nfj

nf)

are called binomial coefficients,

and when the index n

is

nfl

and are sometimes written

0'

C).

all,

(:>

G)

evident from the context and

the same for

is

they are abbreviated to


c l>

c 0>

Thus

(1

The

+ x) n =

c 2

Cn-

'

+ qa; + c 2 a; 2 +

. . .

+ c n xn

(i)

following methods are useful in dealing with series involving

binomial coefficients:

Express the given


binomial expansions.

series as

(a)

a combination of two or more

Obtain the given series by derivation or integration of an


on the standard binomial expansion (i).
(c) Construct a function in which the given series is the coefficient
of a certain power of x, and evaluate this coefficient in another way.
(6)

identity based

Examples
(i)

Sum

Method

+ 2c1 x + 3c i x2 + ... + (n+l)c n x n

(a)

+ c1 x + c 2 x 2 + ...+c n x n + (c1 x + 2cz x i + ...+nc n x").


The first bracket = (1 + x) n The second is
n(n\){n 2)
n(n 1)
nx + 2--x* + 3
-x* + ...+nx n
The sum

(c

= nx\\ + {n-l)x +
{

(n-l)(n-2)
^-

.
,)
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)

Multiply both sides of the standard identity


c

a;

+c

a;

+ c2 a; + ... +
8

(i)

c n a; B+1

by x:

= x(l+x) n

1:

Derive both sides

wo x:

which gives the same

(iii)

not a power

series.,

2c 1

Sum

Method

+x) n + nx{l+x) n ~ 1

+ 2c1 + 3c 2 +... + (n+l)c B

(1

result as before.

Sum
is

425

+ 2c1 x + 3c2 x 2 +... + (n+l)c n x n =

This

series

12.23]

(ii)

but

it

can be obtained by putting x

...

+ (n+l)c B = 2"- 1 (n + 2).

3c a

in ex.

+ ic x + ic 2 + ... +^ c.
n+

(b)

Integrate both sides of identity

(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

= -j- (2+ - 1).


n+1

n(n 1)...2.1

(n+l)n

(n+l)n(n-l)

^
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).

+ c1 a; + c 2 a;2 + ... + c B a B )(c + c1 :c+...+c B

"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)

~ x on the left-hand side


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)!

(n-l)!(n+l)!'

which is the value of the required sum.


Ex. 12 (a), no. 17 is another illustration of Method

(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

+ x) n_1 and deduce the sum of

n( 1

+ 2c 2 + 3c3 +...+nc n

c1

4 Find CJ-2C2 + 3C3 -... + ( -l) n_1 nc n


n
[Method (a); or write
5 Prove c + 3cx + 5c 2 + ... + (2w + l)c = (n+ 1) 2
backwards and add, as for summing on a.p.; or derive x(l + x 2 ) n .]
6 Prove 2 + 3c x o; + 4c 2 2 +
+ (n + 2) c n x n = (1 +)"- 1 {2 + (n + 2)a;}.
.

. . .

7 Prove

c cr

+ CiCr+1 + c 2 c +2 +

. . .

8 Using the identity (l+x) n (l x) n

What is the

value

if

9 Prove

cl

Sum

10

c c1

Find

11

when w

is

even,

(2n-l)\
,
,
_a _
=
+ Scl+...+ncl
+ 2cl
{(n-1)!} 2
,

+ 2c 1 c 2 + 3c 2 c3 + ...+nc n _ 1 c n
a;

(l-a; 2 )", prove that

oddl

is

x*

n
++ +
r2 Co+

F4 Ca

Ci

(w
'

and deduce that


Co

Ci

K_

. ,
1U
..to(n+l)terms
,

3.4

2.3

1.2

*12 Prove

c2

<h

2+ 2 -n-3

(n+l)(n + 2)

+ _ i)n-i! Cn = A + J+ + ... + I.
i.

...

+ c n x n ~ x = {(l+x) n - l}/x; integrate from - 1 to 0.]


6
13 (i) Find the sum of the coefficients in the expansion of (1 + 3a;)
6
(l + 3#) 6 = a + a1 x + ...+a 6 x and put a? = 1.]
(ii) Calculate a + a 2 + a 4 + a 6 and a i + a z + a s- L See no 2
[c x

+ c2 x +

. . .

[Write

*14 Let
(i)

ar

(l

+ a; +

2
t;

+ a1 x + a 2 x 2 + ...+a 2n x 2n

Deduce other expansions by writing

ljx, a; in

place of x,

and prove

a 2n -r-

(ii)
(iii)

Deduce the expansion of ( 1 + x 2 + # 4 n


Use the identity 1+ x 2 + x*) = ( 1 + x + x 2
)

al-a + 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

12.24 Powers of integers


n

We consider sums of the form ^p where p =

or 2 or 3.

r=l

n
(1) -2 r.

r=l

This

is

an

a.p.,

and the usual method gives


n
r=l

427

series

12.24]

n
(2)

2>2

First method.

From the identity

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

/.

Snz + 12n 2 + 4n,

r=l

2>2 =

and

+ 1) (2n + 1).

r=l

Second method. Using

(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

n3 + 3n2 + 3n = 3r2 +fw(tt + l) + n,


r=l

which leads to the

(3)

Sr 8 We
.

result just found.

have

r=1

Summing

+ 1)} 2 - {(r - 1) r} 2 = 4r3


this for r

1, 2,

. .

.,

n,

we get

{n(n+l)} 2 -0 2

= 42^,
r=l

a
S* = ia (+l)

so

r=l

Observe that this result

is

the square of

r=l

The sum could


(r

also

be obtained by using the expansion of

+ 1) 4 r4 together with the results of (1),


The preceding

series

have been summed by using the

method, described in general in 12.22


series
30

(2).

which can be treated in

this

(2).

difference

We next consider some other

way.

GPMH

SERIES

428

[12.25

12.25 'Factor' series

Sum

(i)

1.2 + 2. 3 + 3. 4 + ...+w(n+l).

The general term is ur = r(r+

1).

Consider

r(r+l)(r + 2)-(r-l)r(r+l)

This

is

r(r

3r(r+l)

+ l){(r + 2)-(r- 1)}


= Zur
.

a suitable difference relation, and by summing

for r

1, 2, ...,n,

32r(r+l) = w(ra+l)(n + 2)-0,


r=l

r(r+l) = n(n+

i.e.

1)

(w +

2).

r=l
(ii)

The

Sumton

1.2.3 + 2.3.4 + 3.4.5 +

terms

general term isf


r(r

+1

wr =
)

(r

r(r

+ 1) (r + 2).

....

Consider

+ 2) (r + 3 - (r - 1
)

r(r

+1

(r

+ 2)

r(r+l)(r + 2){(r + 3)-(r-l)}

4r(r+l)(r + 2)

4wr

Hence

r(r+
r=l

1) (r

+ 2) = n(w+l)(w + 2)(ra + 3)-0,

|; r ( r+ i)( r + 2)

i.e.

= in(n+l)(w + 2)(w + 3).

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

.2.3 + 2.3. 5 + 3.4.7 +

...

to

n terms.

ur = r(r+l)(2r+l)

= r(r+l){2(r + 2)-3}
=
/.

**r

r=l

22
r=l

r(r+l)(r + 2)-32 r(r+l)


r1

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)

by exs. (i), (ii),

(n + 2)

after factorisation.

2r8 + 3r 2 + r,

f Here and elsewhere we write down the simplest formula which


the terms given; cf. Remark (a) in 2.71.

is

consistent with

series

12.26]

n
r=l

r=l

r=l

2 x n2 (n +

429

n
r=l
2

1)

+ 3 x \n{n + 1) (2n + 1) + |n(n + 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.

12.26 'Fraction' series


(1)
(i)

ill
++

Examples.
o
tfwm

by

method

terms.

r(r+l)

r+1

this decomposition is of the form f(r) f(r + 1). The


therefore applicable, so that on putting
1,2, ...,n

"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)

Partial fractions are less suitable here because their use


fractions.

would lead to

three

We may decompose ur into two fractions by omitting the last factor:


l

11

(r+l)(r + 2)

r(r+l)

Hence

1.2

and the required


sum
H

{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

may be reducible to the above.

(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.
30-2

'

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

2(n + 3)(n + 4)'

riP'r-l' r=l^"24.
(2)

= 2wr

find

series

~~4~n + l'
2

+ 3)(r + 4)

(r

"

Vr '

+ 2)(r + 3)

[12.26

method will sum all

Direct use of partial fractions. This

summable by the

difference

method, although often

'

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)

and by the usual method

(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+l--5S
rir+3

n+1

r=l*'

r+1
n

22- + 32
r-ir
n 1

r=i

2
-

n+1 n+1 n+2 n+3

37

n+1 n+2 n+3

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

the factors in the denominator successive terms of an a.p.,


in the form

and then express the numerator


2r 2

+ 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)'

fraction can be dealt with separately.

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

cancels out. This will always be the case

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.

12.25 clearly have no


above do possess such a sum

series in 12.24,

limiting sum, but those in exs. (i)-(iv)

to infinity (12.21), viz.


(i)

Some

12.27

(ii)

(iv) ft.

a,

(iii)

trigonometric series
of sines or cosines of angles in a.p.

(1) Series

C=

(i)

i,

cosa + cos(a+/tf) + cos(a + 2yff) + ...+cos{a + (n-l)/ff}.

Multiply both sides

by

then since

2 sin

2 sin /?ttr = 2 sin %/3 cos {a + (r - 1) ft}


= sin {a + (r - J) fl} - sin {a + (r - f

by using one

of the formulae for products into sums,

the form/(r +

1) f(r),

we have by summing for r

S=

The

. .

0}

.,

n:

a product.

difference into

C = coa{* + $(n-l)/3}
(ii)

1, 2,

this expression is of

= sin {a + (n - ) fi} - sin {a = 2cos{a + (w-l)/?}sinn/?

2 sin \p G

by converting the

and

sin-Jnyff
"

sin/ff

sina+sin(a+yff)+sin(a + 2 ff) + ...+sin{a + (w-l)^}.

result

S=

sin{a + (n-

Sm
l)yff}

^f

sinf/J

can be obtained in various ways:

(b)

Put a 7r for a in ex. (i). (This shows that the multiplier


G must also suit S.)
Multiply both sides by 2sin/? (the same factor as for G).

(c)

Derive the result of ex.

(a)

2sin/ff for

finding

(iii)

(i)

wo a.

G'

= cosa-cos(a+^) + cos(a + 2yff)-...+(-l) n_1 cos{a + (n-l)yff},

S'

= sma-sin(a+

ff)

+ sin(a + 2^)-... + (-l)"- 1 sin{a-|-(n-l)/?}.

These can be deduced from exs.


or found directly

respectively by writing 7T+/? for


2 sin
+ /?) = 2 cos

(i), (ii)

by using the multiplier

/?;

series

432

[12.27

Other useful differences. Results like

(2)

= cot0-2cot20,

tan0
cosec 26

cot 6 cot 26,

tan 6 sec 26

tan 26 tan 6

are easily verified,

and can be used

sum

to

suitable trigonometric

series.

Example
(xv)

Sum

+ tan - sec - + tan - sec - + ...ton terms.

tan - sec

wr = tan-sec

If we write 0/2r for 6 in the last difference just mentioned,


u.r

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)

Find a formula for each


1

of the following sums.

23 +

3
... + (2n + l)
+
2
2
2
2 + 3 4 + ... to n

terms

if

l2

+ 3 2 + 5 2 + ... + (2w-l) 2
even;

is (i)

(ii)

odd. Give a formula

which includes both cases.


4 Prove that the sum of the products in pairs (without repetitions) of the
first n integers is -^n(w 2 1) (Bn + 2). [If s n is the required sum, then

(n
r=l
5

Expand

(r

\2

=2*+2> 2

Z')

.]

r=l

+ l) 4 r4 Using the formulae for


.

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

l)+r = (r l)r(r+ 1) +r,

12.25.

n
obtain
f3
r=l

2/i

*"(*+
7 Calculate
r=+l

1).

8 Express r(r + 2) (2r

1)

in the

form

ar(r + 1) (r + 2)

hence calculate

2 r r + 2) (2r
(

1).

r=l

9 Calculate

2 (n-r) (r+1).

r=l

[ur

= n(r+ 1) -r(r+ 1).]

+ 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

3.5.7 + 5.7.9 + 7.9.11

+ ....

[Method of

12.25.]

l +2 +3
l +2
I
+ ^ +

+ 2.5.8 + 3.6.9 + ....

14

+ + ^^ +
1.4

4.7

7.10

16

2r+l

Use partial fractions


18
r

19

r ti r(r

>Sm the following

to

7
h....

7.9.11

sum to infinity,

r +\
+ 2) (r + 3)

n terms.

sin 20 + cos 20 sin 30 + cos 30 sin 40 +

20 cos

+....

1-

5.7.9

sums, and also the

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

...

= {sin (2r + 1) + sin 0}.]


[cos rd sin (r + 1)
= 4 + cos 20.]
[cos 2
+ cos 2 20 + cos 2 30 +
22 cosec 20 + cosec 40 + cosec 80 +
23 tan0 + 2tan20 + 4tan40 + 8tan80+....
24 sec 2 0+4sec 2 20+16sec 2 40 + 64sec a 80+.... [Use no. 23.]
25 sec sec 20 + sec 20 sec 30 + sec 30 sec 40 + .... [tan(r+ l)0-tanr0 = ....]
26 Prove that tan -1 (r + 1 tan-1 r = cot -1 1 + r + 2 and hence sum
cot"1 3 + cot- 1 7 + cot"1 13 +
+ cot-1 1 + n + ra a
is
the
sum
What
to infinity?

21 cos 2

. .

..

. . .

),

. . .

).

12.28 Mathematical Induction

This

is

a general principle

for proving a given statement

which

involves a positive integer n.

Let

TC

always
is

be such a statement. For example, n might be 'n(n+ 1) is


or 'the sum to n terms ofl.l! + 2.2! + 3.3! + ...
<f>

divisible by 2

(n+l)\-V.
by assuming the truth

If

(i.e. if

of

<j)

we can prove

the truth of

corresponding statement for the next integer), and

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
\

continue thus until any positive integer

The

<f>

the statement for any particular integer always implies the

n is

reached.

principle holds for negative integers also, since a statement

n is equivalent to a statement about the positive integer n.

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

the induction hypothesis.


fc+ 1

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),

the induction hypothesis,

is similar to the induction hypothesis, but with k + 1 instead of k. Hence


the result holds for n = k, then it also holds for n = k + 1.
When n = 1, the result is true because the left-hand side is l 2 = 1 and the
right-hand side is $ 1 2 3 = 1.
Hence by Mathematical Induction the statement is true for all positive

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

similar to the induction hypothesis, but with

result holds (trivially)

when n =

1.

or

k+ 1. Hence

+...+ a k +\

k+1

for k.

The binomial theorem

follows

Induction.
(iv)

of

Cr ar

12.13, ex. (iv). This expression is also correct

coefficient

1,2,...,^;)
k

(x

8.3 2 *.

by Induction.

some

in the expansion of (x + a) k+1

a-fc-r+iig

by

(3

relation shows that f(k+ 1) is also divisible by 8.


which is in fact divisible by 8. Hence the result follows

the binomial theorem

a) k

32fc+2_ 32*

8, this

Suppose the theorem holds


{x

Leibniz's theorem on d n (uv)/dx"

was proved by Induction in

6.62.

by

series

12.28]

435

Remark. Mathematical Induction applied to the summation of a


series is equivalent to the difference method. For to prove that, for
all positive integers n,

we assume that,

some

for

+^ _

^+^+

k,

...+U k =f(k),

and then show that

u x + Uz+ ...+uk+1 =f(k+l).

u1 + u2 + ...+uk+1 =

Since

(u1 + u2 + ...+uk )

+ uk+1

=f(k) + uk+1

showing that

this is equivalent to

* =/(*+!)-/(*).
so that/(r)

a suitable difference function.

is

Exercise 12(d)
Prove
1

by Mathematical Induction.

the following results

+3+5+

..

+ (2n-l) = n 2

2 1.1! + 2. 2! + 3. 3! + ..ton terms


.

+ 4 + 7 + ... + (3n-2) =
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.

= b + u1 and 2u n+2 = u n +u n+1 (w ^


3w = a{l - + 6{2 + - )},

2u 2

by

by x + y.

1),

and

1).

n_1
n terms = ^,3 + ( mx
1)
2ra

?/2n-i jg divisible

= a + 6,

= (n+l)!-l.

find limw.

prove

n->oo

9 Prove
cos

sec

+ cos 20 sec 2 +

. . .

+ cos n0 sec" =

[Use the answer to construct a difference function: see

s
|

"*"^

_j

sin 6/ cos"

Remark in

12.28.]

10 Prove
cosec a cosec (a + /?)

to

n terms

is

+ cosec (a + /?) cosec (a +


+ cosec (a + 2/?) cosec (a + Bfi) + ...
cosec fi{cot a cot (a + w/ff)}.
2/ff)

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

and has the value s,

n-><

sm to infinity

of the infinite series T,ur (12.21).

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

number of terms', but is the


such a sum. If we attempt to treat infinite series like finite
ones (i.e. without consideration of convergence), then paradoxical
ordinary sense of 'the result of adding a

limit of

results

may

arise.

For example, suppose x

>

and consider the

series
1

l+x + - + x* + 1z +xao + - + ....


1

One

is

tempted to regard

this as the

sum of two

l+x + x2 + x*+...

whose sums to

infinity are (see

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

and hence the original


Another example is

series

has no

sum to infinity.

i++i+i+i+*+-= (i+i++ -) + (i
from which

+
i+i+

i.e.

= 1+4 + *+...,

a result clearly wrong since each term on the


one on the right.

When

sn does

left is less

than the corresponding

not tend to a limit when n -> oo, there are various


by the following example.

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;

when n->oo, and so s -> 1/(1


and so sw -> + oo; we say $n is properly

then xn ->

x > 1,
divergent to +oo;
(ii)

437

+ x + x + x + ....
n )/(l -x). Hence by 2.72
-x
(1
2

TC

< 1, then #n -> + oo when n

sn

even and xn -+

is

- co when

oscillates infinitely;

x = 1, the series is 1 1 + 1 1 + ..., so that sn = if n


and 1).
is even and s n = 1 if n is odd: s TC oscillates finitely (between
If x = 1, the formula for sn is meaningless, but the series becomes
1 + 1 + 1 +
., so that sn = n and s n -> + oo when n->oo:s n is properly
(iv) if

. .

divergent to

+ oo.

In this book we shall not give a detailed study of the


behaviour of infinite series, and it will usually be sufficient if we
employ the term 'divergent' to mean 'not convergent'. Thus our
(3) Divergence.

'divergence' will cover proper divergence to +oo, proper divergence


to

oo,

We

finite oscillation

abbreviate 'Sr

and infinite

is

means Hur is divergent

oscillation.

convergent' to

(i.e.

ty

Lur

c'; similarly,

'wr d'

not convergent)'.

12.32 General properties


(1)

IfZu r converges, thenMmun =

0.

n><x>

For by hypothesis sn -+s when n ->


therefore un -> s s =
when n->oo.
(2)

The converse of this

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

the series 1>ur may

series

438

[12.32

Given a positive integer n>\, a positive integer p can be found so that


< n < 2 P and therefore (since all terms of the series are positive)

2 J'- 1

S^p-i

When n

-> 00, also 2 p

shows that the

00, i.e.

< *n ^ s

1/n ->

Remark.

cx>

because Sg*- 1

when n

is

(properly) divergent (to

+ 00), although

-> 00.

We have shown that %n ->

of T,ur but */ un -+> wAew


2wr
would be contradicted).
Thus Sr/(r + 1) d because ww = n\{n +

does not imply convergence

-> 00,

6e divergent (otherwise (1)

->

1)

when - -> 00.

(3)

Convergence or divergence of a series

(a)

removing a finite number of terms from the series;


multiplying every term by a non-zero constant.

(b)

This

uiiuu.

series

sometimes called the harmonic series,

un

a*-

p ^ cc; hence s n ->

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.

If1iUr Svr both converge,

to

converges to as + bt, a

For

m\ where m is a fixed number; i.e. from some

series

and

sums

s, t

respectively, then T,(aur

+ bvr

b being constants.

if

r=l

r=l
n

then

2 (a>ur + bv = asn + btn


r)

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
,

speaking, the divergent parts

may cancel out.

If sn steadily increases when n increases, then T,ur either converges


or properly diverges to + 00. If also sn is bounded, then Ymt is convergent.
(5)

This follows from 2.77. In particular,


terms, then sn

is

if

Swr

consists of positive

certainly increasing since s n+1

sn

+ un+x >

sn

Hence
a series of positive terms is either convergent or properly divergent to

+ 00.

439

series

12.4]

The behaviour of the

series considered in 12.2

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

may show that the

divergence. Although these

do not help us to evaluate

We recall that

its

statement 'limsn

(2.71) the

series converges,

they

sum to infinity.

s'

means that

if

any

->co

positive

number e however small is given, then a number m


e) can be found such that, for all n ^ m,

(in general

depending on

e <

sn

<

s + e.

Series of positive terms

12.4

We suppose all terms of our series are positive in this section.


whose terms are

negative can be included

all

factor

-1

12.41

Comparison

by

first

Series

removing the

(see 12.32 (3) (6)).

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 ^

cvn for all

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

the terms vr are positive,

for all n.

r=l

v1 + v2 +

By

...

+vn <

hypothesis,

Wm + Um +1+---+ Un <
<
<
Hence for
sn

all

C{vm

+ Vm+1 +...+Vn

c(v 1 +...+vm +...+v n )


ct.

n ^ m,

= u i + ...+OT + ...+ < ux + ...+um _ x + ct = K,

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.

The right-hand inequality shows that

un <
Taking

n ^ m.

for all

the result follows.

(a),

Test for (proper) divergence.

(2)

If

= l + e in

+ e) vn

(1

1>vr

and

i>,

either

(a)

or

(b)

un ^

n ^ m, where

c is

>

cvn for all

lim (ujvn )

(but

not

a positive constant,

0),f

n-*-<x>

then 1>ur d.

Proof of

Since T>vr d, then (see 12.32

(a).

vm + vm+1 +

By

+ vn

. . .

when

c(m

Swr

sn

.*.

^+

+ um +

. . .

>

(3)
(i)

m+1 +...+f;ft )

-> oo.

+ un -> oo when

n->co,

From the left-hand inequality in the proof of (1) (b),


un >

- 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.

for all positive e sufficiently small;

we may

and the result follows.

Standard comparison

series are

the g.p. Zaf, which c

if

<x <

and d

if

x ^

(12.31 (2));

the harmonic series S(l/r) which d (12.32 (2)).


Another is E(l/rp ); for various values of p this gives a whole family
(ii)

of

series,
(hi)

W
/

which we now consider,


> 1 and d if p ^

X(l/rp ) c if p
5

=
<

111
+ + +

i^

2^

3^

(^ ^)

t The result may be false


yet E(l/r2 c (see (3) below).

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

the inequality is clear since the first, second, third,

. . .

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)

= 1 +2^1+ (2^i) + (g^l)

3+

<

terms >

71

{l

- (^i)

/{*
}

0n summing the G p "

~ 2^}

'

for all

'- 1

>

if

1,

l-(i) 2

common ratio () p_1 is less than 1.


Hence when p > 1, sn is bounded; so by 12.32 (5) the series c.
(b) The case p = 1 gives the harmonic series already investigated

since then the

in 12.32

(2).

p < 1, then mP < n when n > 1, so that


by comparison with S(l/r), S(l/rp ) d whenp < 1.
Alternatively,
series)

s
n

If p

<

=
0,

if0<p<

11
h-

IP

2p

1
h

...

nv

then clearly sn -

>

IjnP

If

(c)

1/n.

Hence

then (without referring to the harmonic

=
np
1

> nx

-> 00

when

w -> co.

Hence S(l/rp ) d when p <

00.

1.

Examples
(i)

Examine

the series

1.3.4

2.4.5

3.5.6

2n+3

Here

n(n + 2)(n + 3)

[When n is large, n 2n/(n.n.n) = 2/n2 ; so for large n, behaves like 2/n 2 .


It is therefore likely that Ew n c, and we try the comparison test for convergence.]
Using the 'inequality' form of the

2n + 3

Sn

n(w + 2)(w + 3)

n.n.n

u n < 3/na if n > 3. Taking v


(being the series S(l/r p ) whenp =

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

Alternatively, using the 'limit'

and taking v n

test,

l/w a as

before,

n^oo v n

\n(n + 2) (n + 3)/

n-yoo

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 -

+ 2 + ...+w _ n(n+l) _ n+1

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)

Show that the following series do


i

un

not converge because

f+f+ H....

lf!-l^ + l^?.
2

a;

a;

-+> 0.

a;

V(r+1)-Vr*
Using

the result of 12.41 (3)

(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.

ascertain the behaviour of the following series.

i+i+^+^+^+.-.

10

comparison

which of the following

+ ...C if >
a;

17 Use the argument in no. 15 to prove that

1,

r>

11(x r jr\)

o for

all

0.

<

If Swr c, prove that

18

Ewa and 2{wr/( 1 + ur )} also c.

11,

prove Z(wr/r) also


19 IfZw*c,
r

20

443

series

12.42]

-<

Verify that

(i)

Deduce that

|_\

therefore

>

2.

=2

r=2 (r-l)r
1

1111,

and has sum to

(This shows that 2(l/r 2 ) converges

<

Prove

*(ii)
v

(r l)r

-<
r=lf

0,

(r+1) 2

'

<

r+1

<

2.

infinity a

when

"J

<

2.)

1,

r*

and deduce that


1

?*/

when

r2
sn

r(w--\ ^

c.

[Use the comparison test.]

+ 2i+--- + (^I)i + <

S^

<1 + 25 +

---

11
^-

(This shows that l/(n+ 1) < s-s 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
:

. . .

fixed positive integer.

12.42 d'Alembert's ratio test

The

tests in 12.41 required

an auxiliary

be stated involves only the given

series

Evr that now


;

series itself.

1iUr is convergent if

either

(a)

or

(b)

un+1 /un < k <

for all

= <

lim (un+1 jun )

n ^ m, where k is

constant,

1.

n-*-ao

HtUr is (properly) divergent if

either

(a')

or

(&')

un+1 jun ^

for all

lim (un+1 /un )

n ^ m,

= 1>1.

n->ao

Proofs
(a)

By

hypothesis,

um+l ^ kum>
um+z < kum+1
and so

on.

s$

k 2um

from the previous

from the previous line;


um+3 ^ kum+2 ^
Hence each term of the series

um+l + um+2 + Um+Z +


is less

line;

than or equal to the corresponding term of the

g.p.

kum + k*um + k3um +


31

GPM II

to

series

444

< k <

which converges since

verges, since the first

by

[12.42

hypothesis. Therefore

terms u 1 + u2 +

+um

...

do not

Swr

con-

affect its

behaviour.
(b)

Given

>

0,

there

is

< l+e

l-e <
Since

<

= I+e

from
(a')

1,

we have

m such that

a number

+e <

for all

n ^ m.

for all e sufficiently small.

Taking

in the right-hand part of the inequality, the result follows

(a).

By

hypothesis,

um+l ^ um>

um+2 ^ um+l ^ um>

e^ c

Hence for n ^
we have un ^ um so that i*n -I* when n -> oo,
proving non-convergence of
The series.must actually be properly
divergent since it consists of positive terms (12.32 (5)).
(&')

Since

>

1,

we have Z e >

for all e sufficiently small.

left-hand part of the inequality in the proof of

^1>1

and the

result follows

by

for all

(6)

The

then shows

n > m,

(a').

Remarks
(a)
if

The complete condition (a)

is essential

the result

may be false

we merely have

^1<1

for all

n > m.

Thus although E( l/r) d,


*

/i

n+l/ n

un

"

n+1

<1

foralln.

No fixed k independent of n can be found for which


^i-1 < k
because n/(n +

1)

and

k<l

can be made as close to

as

we

please

by taking n

sufficiently large.
(/#)

The

tests give

no information i/lim {un+1 /un )

1.

For example,

W->co

d and 2

but in both cases un+1 /un -> 1 when n -> oo.


If use of the 'limit form of the ratio test leads to the limit 1, then

S(l/r)

(l/r 2 ) c,
'

we must revert

to the 'inequality' form since this can give a decision

series

12.42]

when the

form

445

For a power series it is usually


best to treat this exceptional case from first principles.
{y) No conclusion can be drawn from non-existence of the limit of
un+il un- Thus for the seriesf
'limit'

is

inconclusive.

1+1 + i+ i+1 + 1+
2 42 2 3 4 4 2B 4

*"''

successive values of un+1 /un are


9

Jl
23'

so that

un+1 jun

_L

25

JL

93

27

'

29

oscillates infinitely.

vergent (because, roughly,

it is

the

I
1 1 1
2"^42"^2^^"4*"^"""^

95
'

The

97

*"'

'

series is nevertheless con-

sum of two convergent


1

2 2n_1

i 2

q.p.'s):

a-royi-an
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.

= n+1 x -> x when n

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.
,

31-2

series

446

[12.43

When x = 1 the test gives no information; but then the series is 1 + 2 + 3 +


which clearly d.
Hence the series c when < x < 1, d when x ^ 1. (See Ex. 12 (/), no. 12 for
a method of finding the sum to infinity.)
,

. . .

Exercise 12(/)
Test the following series.
1

22

1+

Tl
3

If a

32

2i

and

42
3i

23

22

+ '-

*+*+*+

6 are positive, prove that

q+1 (a+l)(2a+l) (a+ 1) (2a+ 1) (3a+ 1)


+
+
+ "'
6+1 (6+1) (26+1) (6+1) (26+1) (36 + 1)
converges

if

<

and diverges

if

6.

Test the following, assuming x to be positive.

1+
r.
a;

+ ii + Si + ""

a;

i-2

x*

9
10

+ --

3!

+ (l + 3)a: + (l + 3 2 )a: 2 + (l + 3 3 )a:3 + ....

a+l

a;

+
a:

+2

x+ %

-x

*(ii)

+ ...

(aj+l).
'

By showing that
1.3.5... (2n-l)

2.4.6... (2n)

prove that the


*12

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!

8 a;+2 3a; 2 + 3 3a;3 + 4 3a: 4 +

series in

(i)

diverges

>

1.2.4... (2n-2)

2.4.6... (2n)

when x =

_
~

2n'

1.

By deriving the identity


l+a; + a; 2 +...+a: n

1x

1x

obtain a formula for 1 + 2a? + 3# 2 +


+ nx n ~ x , and hence verify from
principles the conclusions of the worked example in 12.42.
. . .

first

12.43 Speed of convergence of a series

A series is said to be 'rapidly convergent


mation to the sum to
sense the g.p.

infinity s for fairly

l+x + x 2 + x3 + ...

good approxismall values of n. In this

'

if sn

(\x\

<

gives a

1)

SERIES

12.44]
is

rapidly convergent

when

\x\ is

small, since

\xn

and x
|

447

xn

small for small n. It converges less rapidly

is

when x

is

just

less

than

1,(1 /r

2
)

1.

On

the other hand, Ex. 12(e), no. 20

(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 \
'

'

'

Similar language would describe relative rates of divergence.

d'Alembert's test for convergence consists fundamentally of com-

paring the given series

1ur with a g.p.

rapidly than any g.p., the test

If Hur

happens to converge less


this is the case with

is ineffective;

Z(l/r 2 ) (and indeed with any of the series T,(l/rp ) for


usually so

when lim (un+1 /un

p> 1),

and

is

1.

Tests of greater delicacy than d'Alembert's can be formulated, but

no

finality

can be attained since

it

can be shown that, however


which that test is

delicate the test, a series can be constructed for

In 12.44 we give a test which is of no fixed standard of


delicacy, and which deals particularly easily with series like 2 (l/rp ),
S{l/r(logr) p } for which the ratio test is indecisive.

ineffective.

12.44 Infinite series and infinite integrals

The

'sum to infinity of a series' (as the limit of a


and 'value of an infinite integral' (as the
limit of the corresponding finite integral, 4.92) are somewhat similar.
For a certain class of functions f(x) we now give a theorem relating
definitions of

certain 'finite' sum, 12.21)


'

in

'

which convergence of one implies that of the other.

series

448

[12.44

The Maclaurin-Cauchy integral test.


is continuous and steadily decreases

(1)

If f{x)
oo

to zero

for x

1,

then

/*oo

2 /(**) and

r=l

converge or both diverge.

f(t) dt either both

When they con-

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.

< x < r + 1, then

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

The argument is illustrated geometrically by considering the sums of


the inner and outer rectangles in fig. 124.
/*n

/oo

of n,

/() eft is an increasing function


J
cannot attain its limit; therefore for each n,

Suppose!

(a)

it

f(t) coexists.

Since

jj(t)dt<jf(t)dt.

From

the

left

of

(ii),

f(r
r-2

< f7(0
Jl

series

12.44]

449

so that
r-2

is

bounded. Hence

(2.77) S/(r) converges,

and

S/(r) ^ f"/W*+/(l).
Jl

r=l

now follows from the right

It

of

(ii)

by letting n

-> co that

r=l
(6)

If S/(r) converges, then

V/(r)<
r=l

r-l

P /(f) iB < /(r),

By the right of (ii),

r=l

Jl

so that

bounded and therefore tends to a limit, say

f(t) dt is

Z,

when

,-> oo.

Sincef

/()

<ft

is

an increasing function of x, then if n

When X->co,

also

and so

oo,

/()

Since the series converges

exists, it

X < w,

<

rn

f(t)dt<jJ(t)dt<jj(t)dL

(c)

rx

rn-i

<ft

->

Z,

/(f)

i.e.

<Zf

exists.

when and only when the

integral

follows that divergence of one implies divergence of the

other.
(2)

The function

$(n)= if(r)r=l
steadily decreases as

Proof.

By

f(t)dt

when

and f(l) for

all n.

f(t)dt<f(l);
Jl

necessary because the fact that

a positive integer) does not by

rx

between

rn

S/(r)r=l

is

Jl

n increases, and lies

(ii),

0<
f This step

{ f(t)dt

itself

j 1 f{t)dt->l when

imply the existence of

X -> oo, X being a continuous variable)

cf.

f
I

n-+ oo

{n being

f(t)dt (the limit of

Remark (fi)

in 2.71.

Thus

J
j cos nt dt

for all integers n,

does not exist since

f
/

J 1

and hence its limit when w-> oo


1

cos

ittdt

is 0;

but

= sin nX, which oscillates when X-> co.


f

cos ntdt

series

450

[12.44
rn+1

also

0(n+l)-0(n) = /(n+l)-

f(t)dt

<

Jn

Hence <p(n) decreases as n increases, but < $(ri) < /(l).


Corollary. It follows that lim <fi(n) exists and is non-negative.

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)

ljx in the corollary,

= l + + +...+--logn.

${n)

Hence <fi(n) tends


and S y < 1.
(ii)

to

some

Discuss the series

With/(a;)

then

limit

1 /r p )

if

by means of the integral

if

p=

1,

test.

1,

=j=

f*/(f)*
while

(Eider's constant: cf. 4.43 (8))

= _I-(l_X-M*)

/(*)

<ft

logX.

If p > 1, the integral tends to l/(p l), and if


< p S
to +oo, whenX -> oo. Hence S (l/r) c if p > 1, D if p ^ 1
The general theorem also shows that when> > 1,

p1
00

In fact y\

J >

1,

which

is

the integral tends

(cf.

12.41(3)).

p1

rP
r =l

a better lower estimate when p

>

2.

Exercise 12(g)
Prove

2^ diverges.

2 Discuss 2

rlog(2r)

If a

>

0,

prove that the


1

a2
lies

between

4 If a

>

sum to

lie.

(a+l) 2

+ (a

+ 2) 2+

""

11
+

a2

must

'

infinity of

and l/a+ 1/a 2


find numbers between which the sum

1/a

0,

r(logr) J

a2 +

l2

a 2 + 2 2+

'"'

to infinity of

series

12.5]

Sketch the graph of y

= logx, and show that

logxdx

<

Jl

Deduce the value of lim

451

n
^,logr
r=2

log x dx + log n.

<
Jl

{(n!) 1/n /n}.

n-yoo

and negative terms

Series of positive

12.5

The special difficulty with such series is that sn is not necessarily


a steadily increasing function of n, so that the principle in 12.32 (5)
does not apply directly.
12.51 Alternating signs (theorem of Leibniz)

The simplest kind of series with terms of mixed sign is that in which

the signs are alternately


//

(i)

+ say
ui ~ u2 + % u* +

the terms are alternately

where each u

is positive,

un+1 < ur

(ii)

un steadily decreases to

i.e.

Km un =

(iii)

zero,

n->oo

then the series is convergent.

Prf

By

(ii),

s 2n

K-u

each bracket

+ {u3 - 1*4 +
)

2)

is positive.

. . .

+ {u2n _ x - u2n ).

Hence

s 2n is

a steadily increasing

positive function of n. Also


S 2n

= Ul ~ U 2 ~ Ua) ~ (^4 M s)
(

each bracket

is

positive

by

(ii),

(u2n-2 ~ U2n-l) ~ u 2n>

and u2n >

Therefore by 2.77 s 2n tends to a limit, say


Further,!
_
s 2n+l

and

since lim u2n+1

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.

even or odd, sn -*l when n -

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

converges because the terms satisfy all three conditions of

The argument used in the proof shows that the sum to


between \ and 1; for s 2n < u x = 1, and

the theorem.
lies

%i = (l- + - + ...+
(

Also see Ex. 12

(h),

no.

><--*

-S;)

8.

If any one of the conditions


verge. Thus since the series
i

infinity

(i), (ii), (iii) is

omitted, the series

may not

con-

+ *+i+i+-,

d, c respectively, the 'difference series'

diverges (12.32(4)); conditions

(i), (iii)

are satisfied

by

but not

it,

(ii).

12.52 Absolute convergence

Consider the pairs of series


i

iii ^" ,iii


AT

The

+ i+i+i + -;

22

32

42

i-i

2a

'

i-

32

4a

"

of each pair converges, by 12.51; but in the corresponding


where all the terms are made positive, the first diverges (being
the harmonic series) while the second still converges (12.41 (3)).
It is convenient to distinguish between series which remain convergent when all their terms are made positive, and those which do not.
Definition. If the series TiUr is such thatf S \ur converges, then 1iUr
is said to be absolutely convergent (a.c).
Observe that nothing has been asserted in this definition about the
convergence of T>ur itself; it is a different series S \ur (i.e. Ewr with
all its terms made positive) which converges.
first

series

Examples
(i)

(ii)

1U )

^ + f 4+
1 i+
1

,111
- +
+
1

-2

3-2-

4l

is

not

A.c.

because

is a.c.

-" lsA

because

C! -

because

t See 1.14 for the

+ + % + % + ... diverges.
1 ++
... converges.
1

,111
+_ + _ + _ +

meaning of

\x\.

...

converges.

SERIES

12.52]

453

The importance of absolutely convergent


they behave like

series

Theorems II, III in 12.54.


Although the definition

we can

series is (roughly) that

of positive terms. For example, compare


states nothing

about the convergence of

an absolutely convergent
convergent in the ordinary sense. More explicitly
Ei*r ,

7/2

in fact prove that

series is also

\ur converges, then 2i*r also converges.


\

Before giving the proof of

Suppose that the

this,

we

illustrate

the underlying idea.

series 2i*r in question is

Ul + U2 U3 + Ui -U5 -1*6 + 0-W8 + U9 -U10 + U11 + U1Z + ....


Consider the following two

%+

i*

+ i*4 +

+0 +W3 + O

series:

+u9 +

+0 +0 +

+1*5 + 1*6 +

+ i*u + i*ia +
+i*10 +
+0 + ....

+ 1*8 +

The first is formed by replacing all negative terms in the given


by 0's; the second is obtained by replacing all positive terms by

series

and

changing the sign of the negative ones.

Compare each of these

%+
which
with c

1*

series of positive

terms with the

series

+ 1*3 + 1*4 + 1*5 + 1*6 + + 1*8 + 1*9 + 1*io + %i + %2 +-">

is

convergent by hypothesis. The comparison test (12.41

shows that each

series converges.

Hence by

(1))

12.32 (4) their

difference (which is the given series) also converges.

General proof. Consider the series


v 1 + v 2 + vs +

+ w2 + w3 +...,

where we define

and
(Thus relative to the given series Si*r Ei>r is the non-negative part, and
Swr is the non-positive part with its sign changed.)
\ur \. Comparing each of Svr 2t*?r (which
Clearly vr ^ |i*r and wr
,

are series of positive terms) with 2 \ur which is given to be convergent,


\

the comparison test with c


li{vr wr

converges;

i.e.,

shows that they converge.

since ur =

vr wr for all r,

Hence

Si*r converges.

The property just proved gives a possible test for convergence of a


terms with mixed signs, viz. prove it is a.c. For example,

series of

series

454

[12.52

However, the test is 'one-sided' because


a series which is not a.c. may yet be c e.g. 1 1 +
J+...is not a.c,
but is in fact c. Such a series may be called conditionally convergent
converges because

it is a.c.

(o.c.)

or semi-convergent.

Examples
*(iv)

The

'

triangle inequality 'ffor infinite series.

If YaUt is A.c, its

sum to infinity is numerically less than or

equal to the

sum of

Slid:

With the above notation we have ur


Sur Hwr converge,

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

sincef both these

K|.

r=l

The device used in

example can

this

\( f(x)dx\
a

\J

>

a and f(x)
if(x)

is

Then/1 (a;),/2 (a;)

be employed to prove that

also

where b

sums are positive,

<

continuous;

if

/(a)>0,

if

f(x)

=sj

( \f(x)\dx,
Ja

Ex.

cf.

7 (a), no. 5.

Define

-f(x)

0,

are non-negative functions,

\m\ =f (x)+f (x),


t

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

Discuss the series

1+ * +
/or aZZ values of x.
If a; = 0, the series
t See 1.14,

both integrals being non-negative,

fi{x)dx,
Ja
I

is 1

(i).

+ + +

. .

x2
2!

x3,
3!

+-

which converges to the sum


%

By

4.15(9).

series

12.52]
If x

>

the terms are positive and

0, all

455

we can use d'Alembert's ratio test:

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,

the series converges for all positive x.


If x < 0, write x = y so that y > 0. The series becomesf

3!

2!

By the case just considered, this is A.c. for all


The given series

positive y,
therefore converges for all negative x.

Consequently, the series converges for


12.32(1) it follows that

also c.

2.74)

(cf.

xn

it is

all x.

N.B.From

lim

and hence

for all

x independent of n.

n-oo n\

(vi)

x +

Discuss the series

for

4- +

...

all values of x.

w2
First consider the series

2/

+ ~2

where y ^ 0. If y = 0, the series


If y 4= 0, the ratio test gives

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

y-series therefore converges for

A.c. for

which converges to sum

. . .

when

0.

n->oo;

the series converges.

1 <x<

1,

and thus

<

<

hence the given

1;

series in

also converges in this range.

= 1, the given series is 1 | +


which c (12.51).
i+
= 1, the series is 1 | i i
which is properly divergent to
oo (it is 1 times the harmonic series).
If x < 1, each term of the series is negative, but its numerical value is
Hence the series
greater than that of the corresponding term of 1 + + J +
properly diverges to oo.
If x > 1, then x = 1 + c where c > 0, and
x n = (l+c) n = l+nc + in(n-l)c z + ... > n(w-l)c 2
If x
If x

. . .

since all the terms of this binomial expansion are positive. Therefore

> i(n l)c


n

2V

'

and sot+

The series cannot converge in this

x^

when

we

- oo.

case.

\x\ +

\x\ 2

f In fact

are considering

X This also follows at once from 2.73.

\x\ 3

. .

..

SERIES

456

[12.53

Summarising, the aeries c for 1 < x < 1, and d for x ^ 1 or x


range 1 < x < 1 is called the interval of convergence of this series.

The modified

12.53

>

\.

The

ratio test

We introduce here a modification of d'Alembert's ratio test which


makes

directly applicable to series

it

// either

or
then

Hur

(a)

(b)

u,n+l

^ k <

for all

=I<

lim

whose terms are not

n ^ m, where k is

all positive.

constant,

1,

converges.

For either condition ensures that S

\ur c,
\

applying the original ratio test to this

being what is obtained by-

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).

we can say is that 2 \ur d, i.e. that Y,ur is not a.c;


not prevent 1>ur from converging.)
Condition (a') implies that \un ^ \um for all n ^ m, so that Mn -f0
(At

first

sight all

this does

when n-^co. Given e > 0, condition


number m such that \un+1 > (l e) \un
\

follows from

(a')

implies that there

(&')
\

for all

is

n ^ m, and the result

as in the proofs in 12.42.

Examples
(i)

Discuss the series of 12.52,

ex. (vi)

in this way.

x the series converges to sum


when x 4= that
If

0.

|a;|

n+l

Since

u n = ( l)" -1 a; n /n, we have

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)

Discuss the series

l+mx-\

m(m

1)

m(m-l)(ra-2)

#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

cannot be decided by means of the tests at our disposal.


can be proved that

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.

and rearrangement of terms of an

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

(1-1) + (1-1) + (1-1) + ...,

+ + + ...,

i.e.

converges to zero. If we remove the brackets

we

obtain the series

1-1 + 1-1 + 1-1 + ...,


which diverges since

we

and

between

s B oscillates

1.

If we insert brackets as

+ (-l + l) + (-l + l) + ...,

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>

Then we wish to prove that 2^ converges to s.


Given e > 0, there corresponds a number N such that,

2U

^8 <

e.

for all

n ^ N,
(i)

r=l
f

They

but this

is

will hold if

we bracket

relatively trivial.

or rearrange only a finite

number of the terms;

SERIES

458
Let n v be the

N. Then

to

first

(i)

[12.54

integer of the above sequence which

holds for

all

2s

is

greater than or equal

n v and so
,

Tim

2u

s <

(ii)

r=l

m > p (since the integers n +1

for all

integers greater than

n +2

2u
r=l
(ii)

are a selection from the set of all

Tim

result

sum can be grouped in any way,

Since terms of a 'finite'

Hence

n P ).

2
i=l

becomes

<

for all

m ^ p;

i=l
i.e.

Y&i converges to

s.

The converse

of this theorem is false : if a series with the terms grouped is


convergent, the series obtained by removing the brackets may be divergent.

This

illustrated

is

by ex. (i), which also shows that there is no general theorem


Theorem I; but see Ex. 12 (h), no. 14.

for divergent series like

Example
If s n denotes the

(ii)

sum

of n terms of the series

W+W+i--,
and

tn

denotes the

sum of n terms of
1 _l_l_l_i_l_l-Ll
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

<.-(i-i)-i+(i-*)-i+(i-A)-A+.-- + (2^ri--4 n _2)

= *-* + *-*+...+

4n 2

4n

The first series is known to converge, say to


and therefore t3n -> %s. Since also

Wl =

hn+ *

therefore t n -> %s

Example
series

(ii)

when n

*3n

s;

hence when n

4n

-> 00, s 2n -> s

+ 2^T[-*2 S

= hn +
2nTl ~ 4^T2

-> 00; this is the result stated.

illustrates that, if

a convergent

series is rearranged, the

may converge to a sum different from the original.

It

new

can be proved that

can be rearranged so that the new series


(Riemann's
rearrangement theorem). For series of positive terms the situation is simpler;
and we shall prove that the same is true of a.c. series.

a given non-absolutely convergent

series

will converge to any pre-assigned limit ; and it can be arranged to diverge

series

12.54J

Theorem

II.

manner without

459
any

convergent series of positive terms can be rearranged in

altering the convergence or the

sum to infinity.

Swr be a convergent series of positive terms, with sum to infinity s.

Proof. Let

Let Zw^ be a rearrangement of Swr and write


,

Each term of Zt^ appears somewhere


a number
(all

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

m such that s' ^ sm

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

and its sum is s'.


term of Zwr appears somewhere in Hu'r the preceding

~Lu' certainly converges,


r

Similarly, since each

argument will prove that s < s'. Consequently s' = 8.


N.B. Observe that we could not prove s ^ s' first because we do not know
whether Hu'r converges or not until we have proved s'n < *.
Theorem III. An a.c. series can be rearranged in any manner without altering

the absolute convergence or the

\u'
r

is

sum

to infinity.

a rearrangement of 2wr then certainly


a rearrangement of \ur \, and Theorem II applies.

Proof. If Hu'r

is

2>u' is A.c. since


r

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

seriesf of positive terms,

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

1-V2 + W3-V4+--

logi+log|+logf+logf+log^+logi|+.-

+
1

2.3

1.2

3.4

4.5

+h

-f+i-f+-&-H+....

6 State which of the series in nos. 1-3 are a.c.


7

Show that 2 {(cos rx) jr2} converges

Under the conditions of Leibniz's rule

for all values of x.


(12.51),

Discuss the behaviour of the following series for

xz

x6

*-3 + 5-7 +
a;

2*

x2

10
'

<s <ux

all values of x.

x3

x7

prove ux u t

x6

x1

*-3! + 5r7l + '-

x*

+ 32 + 4^ + ""
f See the General Proof in 12.52.

32

GPM

II

series

460

[12.6

S^l + i + i+.-. + ^af.

*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

series of positive terms.

conditionally convergent, prove that (with the notation of 12.52)

|~

Hvr and ?jwr diverge to +oo.


n
also

r=l

Theorem I holds for a

I)

m+1

i=l

and deduce that

11.

r=i rj

prove that (with the notation of 12.54, Theorem

for all r,

if

>wherea

Ifs

2 V =

Sw

r=l

Tiur ,t n
r=l

\ur \,

then s n -> sand

+oo;

r=l
~1

i(n-*n)-

r=l

*16 If the series

\ + % \ + \ ...

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

12.61 Expansion of a function as a

We

proved in 6.52 that

tinuity

and

f(x)

where

power

series

if f(t) satisfies certain

derivability for

<

= /(0) + xf>(0) + ^/"(O) +

conditions of con-

then

x,

. . .

+ ^-_^/(-i)(0) + Bn (x),

(see 6.54(1))
r

fx _
(*)

(Lagrange's remainder),

f(n)(0x)

x^l-d)
(n-1)

71

-1
n
f( \dx)

and 6 is some function of x and n

(in

(Cauchy's remainder)
general not the same function in

the two forms of remainder) which satisfies

< 6<

1.

In 6.53 we showed that (under suitable conditions)


a polynomial approximation tof(x).

by an

this result gives

We now use it to represent f(x)

infinite series.

Suppose that for each positive integer

n,

however

large,

(n

\t) is

series

12.62]

O^t^x;

continuous for

then the

Maclaurin's theorem hold for

/(0)

We have

sn (x)

461

all

conditions

for

of

validity

orders. Consider the infinite series

+ ^/'(0) + ...+^^(0) + ....

=/(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;

converges and has sum-function f(x)

(\x\

and call this series the expansion off(x) as a power series


interval

\x\

<

for

in

when

<c),

(ii)

x over the

c.

now be applied to obtain series for some of the


elementary functions; the results are important, and will be discussed
This idea will

in detail in 12.7.

only
of

We use the results on wth derivatives obtained in 6.6;

when such convenient formulae

are

known

is

direct discussion

Bn (x) possible.
some elementary

12.62 Expansion of
(1)

The exponential

ex

series.

then fin) {x)


Lagrange's remainder is
If f{x)

functions

ex for

every positive integer n, and

Rn (x) = ^e* (O<0<1).


Since xn jn\ ->

when n->co for any fixed value of a; (2.74),


when n -> oo, for all x. As/ (0) = 1 for each r,

Rn (x) ->

e*=
(2)

+ *+^ + ^+.. .++...

The sine and cosine

If f(x)

mainder

therefore

(r)

sin a;,

(allx).

series.

then fin) (x)

sin.

(x

+ Inn), and

Lagrange's re-

xn

is

Bn (x) = ^sm{dx + \n7r),


\Rn (x)

-L-L-

->

when

n -> oo,

for all x.
32-2

series

462
Also fr\0)

[12.62

and /(^(O) = ( - 1 y- 1

sin* =

^-^T+^t-

+ (-

1 )r

for each

r,

~l
(

so

2 r-l)!

(a//

'

Similarly

cos*=

T^e logarithmic

(3)

If
so

^Ji + ^T-

/(a;)

in)

{0)

+(- 1 ) r "

(2r _2)!

+ "'

(aW ^*

series.

= log(l+z), then fn\x) = (- l)"- 1 ^- 1)!/(1 + x) n] and


= l) n_1 (n 1)!. We are supposing a; > 1, for otherwise
(

+ would not be defined in part of the range x ^ t ^


When ^ x ^ 1, Lagrange's remainder

log (1

vn

When x <
remainder

0,

{l

+ 0x) n

when

l-S^) < - ->

shows that

n->oo.

xj(l + 6x) may be greater than


.

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.

Cauchy's remainder can be written

^
so that if

because

1 < x <

(1

\x\

0,

\Rn ( x )

Hence

Rn (x)

case

<x<

<
^

j^j

+ 0x\ > \-6\x\ >

(see 1.14)

l-\x\.

when n-+oo. (Cauchy's remainder


1

also covers the

already treated.) Thus

+ *) = *-^+^-... + (-l)'-+...

(-1<*<1).

1 < x < 1 (see 12.52, ex. (vi)),


most that can be proved concerning its sum to infinity.

Since the series converges only for


this is the

f)

6)j{\ + 6x) is a positive proper fraction and


\l

log(l

r w / \ -i
= <- 1 )"- 1 rTfe(rr^)

series

12.62]

463

The binomial series.


lff(x) = (l+x) m then
(4)

f
If

is

n\x)

= m(m - 1)

. . .

(ra

- w + 1) (1 + x) m ~ n

a positive integer, Maclaurin's formula terminates when


gives the usual binomial expansion containing m + 1

n = m+l, and

m is not a positive integer,


> m only if x > 1.

terms. If

when n

then f {n\x) will be continuous

Cauchy's remainder can be written

When

|g|

<
an

by

1,

1) n
= (m- 1)-... (m-n+
^
x ->

(n-1)!

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.

m(m 1)J*2_ *...

+ mx+
+

If

when

a positive proper fraction, and

[(1-IjcI)-

(l

m(m-l).(m-r + l)^

m is a fraction with even denominator,

(1

(M<J)

+x) m has two values

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
/<")()

(-l)"- 1 (w-l)!(l+* 2 )-* n sin(ncot- 1 a;),

(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)

!.

2 )~ in sin (w cot" 1 (&)),

when

- oo

if

|as|

<

1.

SERIES

464
8

Hence

tan

x?

ac

[12,63
8r-1

= *__ + _-... + (

<xr

(|*|

l).

The value of tan -1 x in the range \n to + ^n is intended (i.e. the principal


the sum of the series is clearly
value when 1 < x ^ + 1), since when x =
zero. No more than the above can be proved about the sum because the series
-

converges only for

12.63

<

\x\

see Ex. 12 (h), no. 9.

Note on formal expansions

In Ex. 6 (6), no. 15 we considered the function y


proved that it satisfies the differential equation
(1

by

%=

and

0,

that the Maclaurin coefficients ar

while

tan -1 a;. Having

+ a?) yn+2 + 2(n + l)xyn+1 + n(n+l)yn =

we put x = and showed


are given by
Since a

1, all

= - n {n + l)an
coefficients

(yr ) x= o

with even

suffix are zero;

successive application of the above recurrence formula to

we find that

those with odd suffix,

= ~ I)"" 1 (2n - 2)

GW-i

Assuming

that

(n

1, 2,

. .

.).

y possesses a Maclaurin expansion, we thus obtain

for it the series

Since this converges for


to expect this series

\x\

1 (see

to have y

Ex. 12 (h), no. 9), it is reasonable


its sum-function when

tan -1 a; for

However, as we have nowhere considered Rn (x), we have


not proved by this argument that the series has sum-function tan -1 a;,
nor even that tan -1 x can be expanded at all as a power series.
\x\

is

1.

It can be proved that, within its interval of convergence, every power series
the Maclaurin series of its sum-function. 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

These equations do not prove that


example, if
=
C(x)

*W(0)

'(0)=/'(0)

= /<'>(0),

(x

0),

C(0)

(ii)

....

s(x) is identical with/(a;) for \x\

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 sum-function. 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

Assuming the exponential series, state

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 *

cos (x sin a)}

e x 008 * cos

6 cosa;cha;.

(isma+ na),

dec

and deduce that

for all x,

e XC0BCt cos (x sin a)

By showing that R n (x, h)

*8

coa(x + h)

x3

x2

= l + a;cosa + cos2a + cos3a + ....

cosa;

prove that

in Taylor's theorem,

->

nsmjc cosa:+ suiir ...


2

x and h. [See Ex. 6 (6), no. 24.]


*9 Obtain a Taylor series for sin (x + h).

for all

10 If y

ch (sin -1 x), prove

(l-x*)yi-xy x -y

and

ch(sin _1 a;)

If

(l-x*)y n+i -(2n+ l)xy n+1

a;

(k+l)(k + 2) ak+2 + (n 2 k 2 )ak


If n
n;

and

is

if

12 If 2/

an

even

(k

(1

aj 2 )ya xyi + n 2y =

0.

0,1,2,... ).

show that the assumed expansion is a polynomial of degree


1)*" w 2 (n 2 4).
and greater than 2, prove a4 =

integer,

n is

0.

cos0 and y = cosn# where n > 1, prove


Assuming that y = o + at x + a t x2 + ..., prove
If

(n a + l)y n

+ a l x + a i x% -\-

obtain an expression for a 2n , and prove a an+1


11

+ 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

{n+l)(n + 2)a n+i + (2n+l)(n+l)a n+1 + (n2 +l)a n =

and hence determine the expansion up to the term in x6

0,

series

466
12.7

[12.7

Applications of the series in 12.62

12.71 Binomial series


/-v

(1)

I ,

m(m
1)'^(m r+l)
...

The expression

called a binomial coefficient.

is

The expansion of

reduced to the standard one in 12.62

(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

+ 2a;) -3 in ascending powers of x, stating when it is valid.


There

will

and then the

be an expansion in ascending powers of a; if


(r + l)th term is

<

\2x\

1, i.e. if \x\

<

J,

t-)(-)-(--r+D (2xy . <-ir:-tr + 2)


if

if\

- _ i)r (r +
(

+ 2)

2rxr

1.2

= - l) r 2 y - 1 (r + 1) (r + 2) xr
(

We can obtain the first four terms by writing r


thisformula:
-s = 1 _ 6x
+ 24x *-80x* + ...
{1 + 2x)
(ii)

.Find

<7&e 9ft

term in the expansion of(l

and the 9th term

(iii)

(putting r
/

1\

if

3 successively in

<
\.

we expand in powers ofl/x,

= i(i-i)"'.
i

8 in the general term)

\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\

4x)~ 2 if\x\ >

Since [4a?j > 1, the expansion can be made only


i.e. in descending powers of x.

i-^)- ! = (i-i)"

(1

V{100(1

+ 0-02)} =

10(1

+ 0-02)*= l + i(0-02) + ^^(0-02) 2 + ^'

f Cf. 12.23; the

+ 0-02)*,
~*' ~

symbol m Cr could not be used unless

(0-02) 3

+ ....

m is a positive integer.

series

12.71]

467

Since we require the result to be correct to 4 places, we must work with


5 places; and owing to the factor 10, (1-02)* must be found to 6 places. Calculating the terms shown, we find
(1-02)*

= 1 + 0-01 -0-000,05 + 0-000,000,5+...


= 1-009,950,5...,

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
(l-a;) 5 (l-a:4 )~ 6 = (l-5a;+10a; 2 -10a;3 +...)(l + 5a;4 +...)
= l-5*+10a; 2 -10a;3 + ...

Expand

Since

v)

jc )

on neglecting
(

Hence

10-0995.

all

terms in x* and higher powers.

Find the coefficient of xr in the expansion ofl/{(2 x)(l + 3x)}in ascending

powers of x, stating

the condition of validity.

By resolving into partial fractions we find

11

(2-)(l + 3a0

~ 72-C + 7 1 + 3*
= ^(i-4^)- 1 +f(i + 3)-1

bracket can be expanded in ascending powers of x if \%x\ < 1,


the second can be so expanded if |3a;| < 1, i.e. \x\ < . Hence both
can be expanded if \x\ < \, and then the coefficient of of is

The

i.e.

|a;|

first

<

2;

(-l)(-2)...(-r)

14

r!

=
(2)

IV

(-l)(-2)...(-r)

2)

r\

k - 3)- = \

*<*>'

Summation of series

- ( - sr *}

reducible to binomial expansions.

In the result

(l- a; )^=l + (-3)(-^) +


,

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

above shows that the sum to

result

(l-i)-3

[12.71

infinity of this series is

8.

many

which the factors in the


numerator and denominator of each term are in a.p., while the
number of such factors is the same but increases as we proceed along
the series) can be reduced to binomial expansions, and hence their
sum to infinity written down.
Conversely,

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+

-"-

The factors in the numerators now ascend by the common difference 1,


whereas for a binomial series they descend by 1. This can be adjusted here by
introducing signs as follows:

and this series is clearly the expansion of


Hence the required sum to infinity is 1 +

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

...

3" f (-|) 3 + -.-}-l

= (!-!)-*-! = V3-L
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
(1-x)*

2 (l-x)i.

X
2+x

+ 3*)-*.
1+x2

(1

(1-a:) 8

(4

jc )*.

SERIES
Find the named terms in
8

469

the following expansions.

8th term in (1 - 2a;)- 8 if \x\ <\.

9 10th term in

(1

- 3a;)- 2 if \x\ >

10 What is the first negative coefficient in the expansion of (3 +


[Consider the sign of ar+1 /ar .]
If ur

*11

the

is

(r

+ l)th term in the expansion of


ur

+ 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 (1-x)- 10 whena;

Use

the

or terms in the expansion of

when x = f [Use the argument

21 ' 3

binomial series

16

Expand

+ x + 2a;

V 98

(iii).]

when x =

to calculate correct to four places of

15 (4-08)*.

17

in 12.13, ex.

14 (1 +x)~*

|.

\.

decimals

as far as the term in x*.

2 )-*

State the condition that the following can be expanded in ascending powers of x,

and find the

coefficient of

xr in

the expansion.

2 + 3a;

18

(l-a;)(l
a;

19

+ 2a;)

1
1

*22 (l+a;+a; 2 + ...+a;m - 1 ) B

if r

23 Find the coefficient of x

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 + 4T8-4XT2 + --

3.4.5
3.4
JJ_
+ ""*
+
+
2T4 2.4.6 2.4.6.8

< 1,

00

+ 3r-3)a;r-1 when

r=l
[r2

'

\_~

to infinity of

1.3

Verify that

Hence

+x + x

<m.

and of x2r+1

+ 8 + 8Tl6 + 8Tl04 + "--

26

2r

/j=.
the

T _ l-x l

{l-x)*(2-z)'

Find

6-a;-a; 2

+ 3r-3 = -3 + 2r + r(r + 1).]

||'<1.

SERIES

470
*29

[12.72

By

writing Pascal's triangle (12.13, ex. (iv)) as


we observe that the nth vertical column
gives the coefficients in the binomial series for (1 x)~ n
Justify this rule by proving

shown

here,

12
13

1
1

(rwr--/)- (r)

4 6 4
5 10 10

12.72 Exponential series


(1) Irrationality

of

e.

Taking x

in the series of 12.62 (1),

111

e=1 +
From

H + 2! + 3! + - + H + ""

this it is possible to calculate e to

accuracy (see the example in

any

specified degree of

6.53).

Since this series consists entirely of positive terms,


1

1!

2!

n\

The

error in this estimate


1

(n+l)\

+ (n
,

+ i)\{

(n+l)\{

{n

+ (n

+ 2)\

=
<

is

1+
1

+ 3)\
,

n+2
n~+l

(n + 2) (n + 3)

(n+1)! l-l/(+l)

n+1 _

(n+l)\

+ ")

+ (n+l) 2+ ")

+ ...

by summing the

g.p.,

n\n'

These results can be used to prove that e is an irrational number,


that e cannot be expressed in the form p/q where p, q are integers.
For suppose if possible that e = p/q; then from the above inequalities
with n = q,
,
,
,
,
,
1
1
1
,
p
i.e.

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)!.

that the integer

is

between / and I+l/q, which is impossible since 1/q is


p.(q
a proper fraction. Hence e cannot be written as p\q for any pair of
1)! lies

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

= 1 + a; log a + ^ (log af + (log a) 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

(3) Series reducible to exponential series.

write

down the general term

of the given

first

step

is

always to

and then put

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)!

(r-l)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

(r-l)l

rl

+J}2

r=l

+ 2) 2

+ 4r + 4 = r(r- l) + 5r + 4,

(r-2)l
oo

(r

[12.73

1!

if

GO

r=2

= 9 + e + 5(e-l) + 4^e-l-^
= 10e-4.
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

of each of the following series.

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!

12.73 Logarithmic series


(1)

There are several important forms of the result

log(l+*)
(a)

Replacing x by

*-- + ---+...

(6)

1 < x <

tJC^

(-1

<x^

1).

x,

log(l-z)
provided

{](^

1, i.e.

x2

provided

x3

1 ^x<

1.

Subtracting the above two results, which are simultaneously

valid if

1 <x<

1,

log(l+aO-log(l-#) = 2^x + ~ + ^+...y

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
=
1-x

1+x p+l
l-#
p

p + 1 - log p =

and the left-hand

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.

more rapidly convergent than

are

either the

used for calculating natural


logarithms, and (d) for calculating logarithms of consecutive numbers.
standard series or

(a)

separately;

(c) is

Example
(i)

Find

the coefficient of x r in the

expansion of log ( 1

+ a;2
a;

if |*|

<

1.

l+x
= log= log(l+*3 )-log(l+ar).
1+x
z

log(l-a; + a: a )

We

have

log(l+ar)

a?

- + -... + (- l) r_1
2

and

log(l+a:)

If r

is

If r

is

-...

+ (- I)'"1

not a multiple of 3, the coefficient of xr


a multiple of 3, the coefficient of xr is

is

...

-!-

....

l) r_1 /r.

(2) Series reducible to logarithmic expansions.

Examples
(11)
K

1.2

'

2.2*

3.2s

(- .

. .

12

= -log(l-i)
If the series
general term.

is

...

=log2.

not immediately recognisable, we begin by writing down the

series

474
(iii)

Sum to

[12.73

111

infinity

275

+ ""

377

r(2r+l)
1

2r+l

on using

2r

2r+l'

partial fractions,

where the denominators are now consecutive numbers. Then

^{-+-*>+- + (s-5Tl)}
-> 2{1

The sum of the series


Find

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)

and the sum

-1

when

s$

<

r+1

sn
"

w+1

-> 1

when n

the problem has been solved completely.

Sum

-> oo,

is 1.

It is easily verified that the series does not converge unless

(v)

1.

1,

u.r

/l

|--l)log(l-a;)+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]

and we find by the usual method that A =

2r

2r+l

2r + 2

= l/_
Sn=

Hence

\.

L_V

2r + 2/

/ 1

\2?~ 2r~+l) ~ 2 r?i \27+I~ 2r + 2j

B = \, G

\,

2\2r+l

2r+lJ

2\2r

475

when

-i(log2-l)- |(log2-l + )

-> oo,

= f-log2.
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(l-4a;).

log(l+a;) 2

4 log(l + 5a; +

6a; ).

(1

6 log

-a;)log(l +x).

+x+
.

7 log(l-a; + a; 2 -a;8 ).
8

Expand log

9 If

>

|a;|

+ 2x + 3a; 2

as far as the term in xl .

expand log(l+a;)-loga;

1,

in powers of 1/x, giving the general

term.

10 Taking x
of decimals.
11

Taking x

12 Use series
13

\ in 12.73(6), prove log 2


\ in

(d)

series (6)

and no

Prove 2 log 7 + log 2


<r

and using no.

10. to

prove log 3

0-6931472 correct to 7 places

10,

prove log 10

2-3025851.

1-098612.

2 log 10 a, where

=
=

0-02 + i(0-02) 2 +i(0-02) 3 +

...,

and deduce that log 7 = 1-945910.

Ill

Find the sum


14

to infinity of the following series.

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=r-3

3.4.5

x*

33

x*

4r-2

4r-l'

4a;

5.6.7

^r1 - y4r/ - \4\2r


hr^T1 ~ 7T
2rJ

written ] (-r^z - -r^z +


2 \4r 3 4r 2 4r

6a;

'

consider lim *
|

J
8a;

series

476

24 If y

2x*

1,

prove (under conditions to be stated) that


1

i^*2*
25 If

J7r < a <

sin 2 a

[12.8

3a

3?/

52/

+ "*'

prove

+ sin4 a + ... + -sin2n a + ...


n

=
x

26 Iff{x)

2 2 sinH<x

2 3 sinHa
7T -+-2

= x + -+- + ... and

<

\x\

1,

prove/

2 n + 1 sin 2n

ia

+....

to

2x \
^^1 =

2/<<r).

12.74 Gregory's series and the calculation of


If we put x =

in the result of 12.62 (5),

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 we now prove. Let a


tan 4a

which

is

write 4a

tan

and a; =
Thus on putting x =

\ and

2 tan a 2 ~|
_ 4 tana IV
~ l-tan 2a/ |_ ~ \l-tan 2 aj
J
(

2a

a number just greater than


= \ti + tan -1 x, then
|

result

Jar,

tan -1 -J, so that tana

2 tan 2a

on the

1.

Hence 4a

tandzr + tan- 1 *)

is

just greater than

\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

12.81 Estimation of the error in s

Given a convergent
to infinity

=sn

infinite series,

by taking the

first

n terms.

we can approximate

to

its

sum

If the series has been obtained

series

12.81]

by an

477

application of Maclaurin's or Taylor's theorem as in 12.61, the

remainder term R gives the error in the approximation (see 6.53);


but owing to the indeterminate nature of the number 6, an estimate
of R has to be

made in practice. When R cannot be obtained explicitly


when the sum-function
of the given series is not known, the method in 12.72 (1) can be used.
because of difficulty in calculating f {n) (x), or

Examples
(i)

If
Ifwetake
we

= x - + -... + (- l) n

log(l+z)

then the error

is (see

12.62 (3))

(n+l)(l + 0a) n+1

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

<

+ 6x\ > l-d\x\ > 1- \x\ - 1+x,

and

(ii)

|-R+i(a;)|

<

ex

\B n (x)\

X
(n-1)!

(x n /nl)e 9x

and

x
<e
n!

When x <

we can assert

all

error is

n\

(n+l)!

e 9x

<

and

that

which is of no help since we are trying to approximate to

The

is

0,

e x.

(n + 2)\

+ "'

x*
_ xn / + x +
+ \
~n\ \ w+l (n+l)(n + 2) '"/'
,

and when x >

this is less

xn /

wl\
if

than

n+l

< n + 1, on summing the

approximation

is less

than

x2

g.p.

Thus
xn
n\

The argument

(n+l) 2+

_
=

'"J

if

>

xn

nll-xl(;n+1)

and n > x 1, the

error in the

n+1
n+ 1x'

just used is essentially that in Proof (a) of d'Alem-

bert's ratio test (12.42).

If ur+il uA < ^ < 1 for all r > m, the error in taking s


numerically less than \un+1 \j(l k).
\

is

===

sn (n

> m)

33-2

series

478
For when

n> m

K+a |<* K+1

K+s| < * K+2I <


The

error after

[12.81

n terms

|,

& K+l|>

is

numerically less than or equal to

+ un+2 + mtc+3| +

Kn+ll

'

is

un+l + un+2 + Un+Z +


which

etC

<

>

12.52, ex. (iv))

(cf.

|t

Wl |(l+* + *+...)
1

U,n+1

1-&

on summing the g.p., since < k < 1.


Another useful result is based on Leibniz's rule (12.51).
If for all n > m the terms of a series alternate in sign and
decrease to zero in numerical value, the error in taking s

==

steadily

sn (n

> m)

cannot exceed the modulus of the first term neglected.


Let the series be

u 1 -u2 + u 3 -... + (-l) m- 1 um + {-l)^um+1 +...,


where for
in

> m)

all

n > m, un >

is

The argument

and un+1 < un The error

x
- 1 )
{un+1 - un+2 + un+z r

after

}.

. .
.

in the proof of Leibniz's rule shows that, for

un+l ~ un+2 + un+3 ~

En

converges to a limit

where

fixed,

< En < un+1 Hence


un+1

the numerical

value of the error does not exceed

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

>-

from the second term onwards) the signs alternate;

2r 1
,

r+1

(0-01)

/
(

0-01

<

0-02,

479

series

12.82]

the terms steadily decrease numerically. Hence the error is numerically not
greater than
4 A
*-*-.-#
3.5
1
(0-02)4
(0-02) 4 = 6-25 x 10- 9 .
2*4!
4!
-S

Formal approximations

12.82

The

following examples illustrate methods of obtaining approxi-

mations, but no error estimate


Ifxis small, find the

(i)

best

W
i

(a)

is

made.

approximation for log ( 1

+ x) which has the form

x(l+ax)

ax

'

+ bx'

'

+ bx+cx2

By expanding both sides of


log(l

ax

+ a;) =
1

when x is

small,

+ bx

we have

x-$x 2 + %x3 - ... = ax{l-bx + b 2x 2 -...)


= ax abx 2 + ab 2xz
approximately for all small x if a = 1 and ab =

This will hold


The expansions differ in their x3 -terms since ab 2
log ( 1

+ x)

(b)

i-

A similar method could be

correct to order

+X

\, i.e. b

|.

Hence

2x
==

Ji

rrom

4=

x2

used, but the following

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

We can obtain three equations for a, b, c by equating coefficients of


a
from which a

and

= 6-|,
6

1,

= c-\b + \,
= \. Hence

= -ic + i6-i

since it can be verified that, for these values, the coefficient of


is non-zero, this approximation is correct to order x*.

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

e* for ease of printing,

*"{ -L + t +0 $))
1

=e i+
[

{-i + 3s-

+o

^{-i + i +o (^)F +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

approximation we expand the left-hand side as far as terms in x:


1

+ 2x = 1 +p,

For the second approximation, expand as

.'.

far as

\p.

x2

= 1+p, i.e. 2x{l + \x) = p.


Using the first approximation! x = \p, this becomes
l + 2x + $x*

+ ip) =p,
x = \p{ 1 + 1^)- 1 = ip - &p 2
2a*l

so

12.83 Calculation of certain limits

Example
.

ex

Find

lim

-l-x

0 --log(l+aO

We have

ex

x*

= 1 + x + + a8

where

11

^ + ^+| +
4!

3!
/

....

5!

Id 2

\x\

<3
if

3!l-iH

1*1

<

4.

log(l

Similarly

where

\b\

+ a?) = x +bx z

< o+

T + iir+

<i(i +

=
f

a;
|

V-r
31-|a;|

a;
|

+...)

if

To keep the working

N<111

linear in *.

series

12.83]
ex

Hence

-l-x

ix*

481
l(x2

i + ax ->
,

when

-+ 0.

The above argument is often given incompletely as follows:

Mm
x_

ex 1 x
-log(l + a;)

lim
x->o

x
i

2i

+-

= lim

x-+0

1.

The result is easily obtained by two applications of l'Hospital's rules

(6.9).

Exercise 12(m)
Prove

the statements in nos. 1-4.

The error in taking log{(w+ l)/n}

==

2/(2n+

1) is less

than

6n(n+l)(2ra+l)*

2 The remainder after n terms of


1

1+

Ii

lies

|aj|

<

1,

8i

+-

and

nf n\

the error in taking

1+x
1-x

i log^
is less

2i

^
(H n+l/w!
\

between

3 If

=?
'

xs
Z + + ... +
3

than the numerical value of

prove that log


2 log 3 + 0-20067
*4 The error in taking
1 1 ==

a:

2n+1 /{(2w

a;

"" 1

2n-l
;

+ 1)(1 x2 )}.

Putting n

2,

correct to 5 places of decimals.

**l-*+*-... + <-D-^
is

numerically less than l/(2n+ 1). Deduce that the


\it correct to only two places of decimals.

first

50 terms of the series

give

Use the expansion of (1 -^r)~* to find <J2 correct to 7 places of decimals.


By considering (l + ifs)*, show $2 = 1-2599211 correct to 7 places of

decimals.

Verify the following formal approximations for small x.

4 lllx)
7 V(1+a!)

2n + (n +

l )a

^n*-l
r

2n + (n-l)*" .12*

af

'

SERIES

482
8 e x /*J(l

+ 2x) = 1+x 2 f# 3

Write down the corresponding approximation

for l/{e x j(l-2x)}.

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
\
/

12 Find an approximation of the form a + bx for (8 + 3)*/{l + ( 1


x being small.
13

Find an approximation of the form a + bx + cx 2

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

1-03(1

-xf.

15 Neglecting x*, prove that e^logfl+a;) = log (1 x). Hence find an


approximation to the root of e x log ( 1 + x) = which lies between and 1

= 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

small, prove that

Calculate the following

_
18

lirftits.

a;-log(l+a;)

^
x

lim
x-+0

'

trx

19

lim

a x l

>

(a

0).

a:->0

Miscellaneous Exercise 12(h)


n

Prove
r= or!(n r)!

2n

.!

+ c 1 + ... + c) 2 = 1 + 01 + 2 "C2 + + 2 "<72B


Calculate eg + +
+ c 2 and prove
2
2
c 2 + 2c 2 +...+rac 2 = n(c + c +...+c

2 Prove
3

(c

. . .

. .

).

(i)

Verify the identity

(l+x}

and use
*(ii)

= l+x{(l+x) n - x + (l+x) n - 2 + ... + (l+x) + l},

~
~
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)

+ (3 + 5) + (7 + 9+ll) + (13 + 15+17 + 19) + ...,find

the number of terms in the first r 1 brackets;


(ii) the first and last terms of the rth bracket;
(iii) the sum of the terms in the rth bracket.
terms of 1 + 3 + 5 + , and show that the sum
(iv) Write down the sum to
n
of the terms in the first n brackets is t 2 , where t n = ^r.
r=l
\2
/ n
n
*" 3 =
r
Deduce from (iii) and (iv) that
(
I
V=l /
r=l
(i)

. .

SERIES
Find

sum to n terms and to

the

infinity of

2.3.4

tan- 1 (2r+

+
+
T+2 l + 2 + 3 '"*
If r > 0, prove tan- 1 (l/2r8 =

483

1)

+ 3.4.5 + 4.5.6 + ""

- tan" 1 (2r- 1). Hence find

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).

+ 2cosa: + 2cos2a; + ... + 2cosna;)sina; =


w
f sin(n + i)a5
= it

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

odd or even. Deduce that


/71
%7T

sin nx

dx

=n

or

!q 8W.X
Jo

according as

n is an odd or even positive

integer.

r n /sin na;\ 8

Prove

(ii)

Jo \sina;/

for all positive integers n. [Square the relation in

sin0

(i),

and use Ex. 4 (Z),no. 30 (i).]


6

= 2cos-cos- ... cos-.

12 Prove that

Deduce the values of


n
(i)

Use Mathematical Induction


13

-2 4 + 3 4 -...

dn
14

dx n

(x*- 1 e1 "")

(ii)

r=l

l4

Slog cos-

to

to

prove the following (nos. 13-16).

w terms = ( - l) n " 1 in(n 3 + 2n 2 -

= - l) n eW/af+ 1
(

1).

U =
+ +
+
nTi nT2 - *n>

15 If

Wn=1 _ i +

then

Deduce that limw n

6
6
m S ^tan-.
tan ^5
S
^
z
z
z
z
r=l
r=l

+ ... + _i__ii

log 2.

t In nos. 10, 11 the expression (sin ma;)/(sin a;) is defined to be


the definition is 'completed by continuity' (see 2.64).

i.e.

when x =

0;

SERIES

484
16 If

u n+% (a + /?) u n+1 + <xfiu n


(a -/?)

17 If s n denotes the

sum

for

w>

+a

then

1,

(a"- 1 - yff"- 1 ) w 2 - ayff(a- 2 - /?- 2 ) u v


of n terms of
16a; 16

2x 2

x
1

8a
4a*
+ 1+a;
^ ^
t
1+* r

+ 1+a; 2 + 1+cc* +

1r.

= 2" and x =t= 1. Deduce that


prove that s n = xj{\ x) mxm j{\ xm ) where
the series converges when |a| < 1. [Either use Induction, or take logarithms
and derive the identity obtained from
1
_ 1+x _
1-x ~ 1-x 2 ~

18

Assuming that

+ a; 2 _
1-x1

(l+a:)(l

19

+ 3i + 52 + - =

Find the

(i)

+ a + ^Z + ... - ~,

provethat

coefficient of a:4

(U)

,111
+- =
-2 +
1

and of x*

3i-ii

12-

in the expansion of

(l+x+x +x 3 + x i

st

How many solutions in non-negative integers has the equation

(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

A circular arc subtends an angle 6 at the centre of the circle, and a,

Find numbers x, y, z independent of d, a, b, c such that the length of the


approximately ax + by + cz, correct to order 6 e [Use the series for sin x.}

tively.
is

Find the sum to

infinity of

3.9
3.9.15
23 l + r + r-rr + r-rr-^7+-
8 8.16 8.16.24
_

b, c

and of \ of the arc, respec-

are the lengths of the chords of the arc, off of the arc,
arc

stated),

)x* + \(a*+P )x +....


3

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

r-2
r(r+l)(r + 3)

^Tl irzi^'"
16

29

1.2.3

,234
+

28

r+l(r-l)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

Taking p = 26, 31,


that a log f +6 log
35 Denoting

485

by

+l

49, calculate log 5 to 3 places of decimals. [Choose a, 6, c so

|K clog f = !og5.J
Pr the number of arrangements of n unlike things taken r

n
at a time, prove
no. 2.]

2 "^r = the largest integer less than e (n


r=l

!).

[Use Ex. 12 (m),

486

13

COMPLEX ALGEBRA AND GENERAL


THEORY OF EQUATIONS
13.1

Complex numbers

13.11 Extension of the real

number system;

<J(

1)

Consider the six equations

=
2x =
2x + 3 =
2x

=
x =
2
x +l =
x2

(i)

6,

5,

(ii)

0,

(iii)

9,

(iv)

2,

(v)

0.

(vi)

knew nothing about fractions or negative numbers, we should


(i) and (iv) are both satisfied by x = 3, and

If we

only be able to assert that

that the others have no solution

numbers which
If

we
(i)

that

we could find no natural

are subsequently acquainted with fractions

of numbers
that

(i.e.

satisfy these).

and

is

(i.e. if our list


extended by admitting rational numbers), we can say

(iv)

would remain

are satisfied by f while


,

(ii) is

satisfied

by f The others
.

insoluble.

Not until signed numbers are introduced can we solve (iii) by


x = f and then (iv) is found to have two solutions f We still have
no solutions for (v), (vi).
.

To enable us

to say that (v) has a solution (the necessity for doing

we introduce the number *J2 which can


be non-rational. Even then, (vi) has no solution.
It will now be clear that what can be said about solutions of equations (i)-(vi) depends entirely on what we are prepared to call a
'number'. In other words, the possibility of solving a given equation
depends on how far the concept of number has been extended. Fractional
and negative numbers enable all linear equations with integral
coefficients to be solved, but not all quadratics.
If we attempt to solve (vi), written in the form x % = 1, in the same
manner as for (v), we should obtain the formal solution x = ^J(-l);
this

was indicated in

be proved

1.11),

(see 1.11) to

'

'

that

is, ^/(

a solution

1)

is

existed.

what we should write for the solution of (vi) if such


Whether or not (vi) can have a solution depends on

'

COMPLEX ALGEBRA

13.12]

487

the meaning of 'number', so that a question like 'Does


is futile

until

1) exist?

*J(

we specify what kind of 'numbers we are going to work


'

with.

Hence we have two


meaningless symbol; or

leave ^( l) as a
further the concept of

alternatives: either (a)


(b)

we

generalise

still

we

'number' to include such expressions as this. It is clear that, unless


we choose (6), we can never have any complete theory even of quadratic equations. To arrive at our generalisation we follow the historical
sequence of development; we use the symbol i as an abbreviation
(introduced

by

Euler) for ^(

- 1)

in 13.12, 13.13.

13.12 First stage: formal development

At first i was regarded as something mathematically disreputable!


whose use, nevertheless, gave results quickly and simply to problems
which could be solved more laboriously by other (respectable) means.
The procedure was formal: i was treated like an ordinary algebraic
symbol (with the peculiarity of satisfying i 2 1), and was used in
the spirit of 'press on and see where we get'. Cardan (1501-76) was
probably the first to do this.
Accordingly we should have (a, b, c, d being 'ordinary numbers'):
(a

+ bi) + (c + di) =

+ c) + (b + d)i

(a

(i)

by regrouping and taking out the factor i;


(a

by

first

+ bi) (c + di) =
=

ac + adi + bci + bdi 2


(ac

bd) + (be + ad) i

(ii)

multiplying out the brackets, and then using

= 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.

Squaring both sides

assert that a positive

number

number and is a contradiction unless both are


and d b = 0. Hence

equal to a negative

zero,

i.e.

ac
if

a + bi

= c + di,

These examples show

how

then

and

expressions involving

=
*

d.

(iii)

would behave

assumed to apply to them.


Owing to the mystery associated with it, i was named an imaginary
number, and expressions involving it imaginary expressions; while
if the rules of ordinary algebra are

f Earlier, negative numbers carried a similar stigma.

COMPLEX ALGEBRA

488

[13.13

when

ordinary numbers and expressions weref called real

was

contrast

desirable.

This intuitive approach, however unsystematic

many advances

in fact lead to

it

may appear,

did

produced results which


could be verified subsequently by orthodox means.
because

it

13.13 Second stage: geometrical representation

In the

by

we

i,

get iia

spirit of 13.12

we

see that if any real

obtain an imaginary

number

= la = a, which

successive multiplications by

real

number

-a

number a is

multiplied

multiplying again

by

i,

we

again a real number. Thus two

is

convert

ia;

into its negative.

J.

^x

Fig. 125

Consider now the geometrical representation of this result. Suppose

a is positive, and let OP be the corresponding


segment of the a;-axis. Then two multiplications by i have the effect
of rotating OP, originally along the positive asaxis, through two rightfor definiteness that

angles (say in the counterclockwise sense, to agree with trigonometrical

conventions) to position OP'. It

one multiplication
right-angle.

by

is

as turning

therefore reasonable to interpret

OP

counterclockwise through one

Hence the imaginary number ia is represented by a poin|;

P" at distance
apply when a

a,

is

measured along Oy

(fig.

126). Similar considerations

negative.

This geometrical representation was published in 1806 by Argand,


although others had the idea a little earlier. In it we are regarding i

but as an operator which rotates lines OP


counterclockwise about the origin O through a right-angle. (Observe
that ordinary numbers can also be interpreted as operators in the
process of multiplication: the 2 in 2a as the operator which doubles

not as a number at

the length of

OP

all,

and preserves the

and so

We

OP

2a

as the

and reverses the

sense;

sense; the

operator which doubles the length of

in

on.)

can now represent real numbers by points on the (positive or


and imaginary numbers of the type ia by points on

negative) #-axis

/-axis. Can expressions of the form a + bi


be represented? If we interpret this as corresponding to a step of
t And still are (see 1.11).

the (positive or negative)

COMPLEX ALGEBRA

13.14]

489

by a turn counterclockwise through one


and a step of length b parallel to Oy, we end up at what
we should refer to in graphical work as the point (a, b), say Q (fig. 127).
Although the interpretation of * as an operator and Argand's

length a along Ox, followed


right-angle

representation of a + bi both have important applications, the significant thing at this stage is that

we

are led

away from the vague

'imaginary expression' a + bi towards the familiar and precise idea

P"
ia

\
\

Fig. 126

Fig. 127

of the coordinates of a point


a,

in a plane,

(a, b)

b taken in a definite order. It

is

from

i.e.

a pair of numbers
we can re-

this notion that

construct the theory of 'imaginary' numbers on a logical basis.

13.14 Third stage: logical development

Our aim now is to make a fresh start by setting up and developing


an algebra of ordered number-pairs^. That is, we begin with the definite
idea of a pair of numbers, and as far as possible try to retrace our
steps to get at * indirectly. This approach is contained in the work of
Gauss (1777-1855) and others.
(1 )

We

write our number-pairs, usually called complex numbers, as

[a, b] where a, b are ordinary (real) numbers. As we are about to invent


a new kind of algebra, we can make the rules as we please provided
they do not contradict each other or lead to contradiction. However,
our purpose is to make the new algebra as much like ordinary algebra
as possible; and so, with a view to preserving the structure of the formal
results (i)-(hi) of 13.12, we define the fundamental operations with

number-pairs as follows:
(a) Equal number-pairs.
[a, b]

[c,

d]

if

and only

if

a=

and

d.

f The reader is in fact already skilled in dealing with ordered number-pairs of


another kind, viz. ordinary fractions: see Ex. 13(a), no. 26.

COMPLEX ALGEBRA

490

Addition of number-pairs.

(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,

multiplication to be the number-pair


[x,

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.

solve these simultaneous equations

_bc ad7
V = ~c? d?'
+

ac + bd

+ '

Vac + bd bc-ad~]

fcfl-L ?Ta? ,^ T? J.
we can show

that complex

the laws of ordinary algebra. For example,

[0j

For the hypothesis and the


ac bd

and so

can be defined in terms of


y] such that

d]

using the above definitions

numbers obey
>

define

[0, 0],
r

By

we can

are not both zero,

for x, y:
*

(2)

y]

[xc yd, yc

this gives

Hence

we can

such that

[a,b].

Similarly, the quotient [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],

the opposite of addition,

is

x+c

If

[ac

[c, d] to mean the number-pair

which by

By

[a

mean

to

is

[x,y]

By

mean

to

Multiplication of number-pairs.

(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 =

&e [0, 0].

show that

0,

a 2c 2 + b 2d 2 + b 2c2 + a 2d 2

COMPLEX ALGEBRA

13.14]

Hence at

491

one of the real numbers a + b c + d must be 0; and


a = = b, or c = = d, or both. The result follows.
Ex. 13 (a), nos. 19-25.
2

least

this implies

See also

Application of definitions

(3)

[a,0],[c,0]gives

[M

[a, 0]

+ [c,0] =
x

to number-pairs of the form

(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

numbers a are structurally

the

same.f

Owing to this fact, we agree to abbreviate [a, 0] to a, and in particular

we write

2 for +2,

for [1, 0]. (There is a precedent for this:

In definition

agree to write

because the (unsigned) natural numbers and the

etc.,

positive signed

we

numbers behave exactly the same


(c) let

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];

and according to our agreement about abbreviation of the right-hand


side, this

Thus
is

can be written

n9

2
[0, l]

1,

number-pairs

and

an element whose square


[0, 1] to i; then we have

(6),

[a, 0]

+ [0,6]

[a, 0]

+ [b, 0] [0,

1]

= a + bi
etc.,

is

Let us abbreviate
for which i 2 = 1.

viz. [0, 1].

definition
[a, 6]

If

= - 1.

in the algebra of number-pairs there

equal to

By

r_

by

definition

by our

we now rewrite definitions


we recover the results (iii),

(a), (b), (c),


(i), (ii)

abbreviations.

replacing

and

all

numbers

in

[a, b]

by a + bi,

respectively in 13.12; but

these results are consequences of our definitions


abbreviations,

(c),

them

now

and agreement about

are actually number-pairs

although they appear to be ordinary numbers.


f

We say that there is

[a, 0]

and the

34

real

an isomorphism between the complex numbers of the form


numbers a.

GPM II

COMPLEX ALGEBRA

492
Since
[a, b]

[b,

0][0, 1]

[0, 1][6, 0],

as a + ib instead of a + bi

amount to writing a for a + Oi,


(4) Conclusions.

V( 1)

as

easily verified,

is

when
for

convenient.

+ li, and bi

= 1,

(if it

we can

write

Our conventions

+ bi.

for

In the algebra of real numbers ('ordinary algebra '),

does not exist. In complex algebra (the algebra of ordered

1 which do satisfy
so that this pair i has the property which *J( 1) would have

number-pairs) there are pairs abbreviated to


2

[13.15

i,

existed) in real algebra.

Correct results in complex algebra are obtained

by applying the

laws of ordinary algebra to the number-pairs in their abbreviated

form a + bi, with the replacement of i 2 by 1 whenever it occurs.


That is, the intuitive work of 13.12 is correct, provided we realise
that

we are handling abbreviations for pairs.

Remark. The relations


complex algebra; e.g. it

'less
is

than', 'greater than' are not defined in

meaningless to write 3 + 2i

particular, the terms 'positive', 'negative' cannot

< 4 + 5i. In

be applied to

complex numbers.
13.15 Importance of complex algebra

Complex algebra provides a method for proving results of real algebra.


is so for two reasons.
(i) A single equation in complex algebra is equivalent to two

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,

definition (a) in 13.14.

to be zero. Because of the exact correspondence between complex

numbers

[x,0]

and

real

numbers x

(13.14(3)),

we deduce a

relation

in real algebra.

These remarks will be illustrated in this chapter and the next.


Complex algebra offers generality and completeness. For example, we
can assert that in complex algebra every quadratic equation has two
roots (possibly equal).

Since complex numbers arose from the attempt to solve all quadratic

equations (and in particular the equation x 2 +

0),

we may wonder

whether consideration of equations having degree greater than 2


would lead to some higher form of complex number, say 'super-

complex numbers'. The answer is no, because in 13.42 it will be


proved that in complex algebra every polynomial equation of degree n

COMPLEX ALGEBRA

13.16]

493

n roots (i.e. the roots are themselves complex numbers). Thus


complex algebra has a finality which the real number system lacks,
and no further generalisations can be obtained in this direction.
has

13.16 Further possible generalisations of 'number'


(1) Ordinary algebra is 1 -dimensional, i.e. any real number can be represented by a point on a line, say the axis of x.
Complex algebra is 2-dimensional, since a complex number is represented
by a point in a plane, with reference to two axes.
Is there a 3-dimensional algebra an algebra of ordered triplets [a, b, c]?
Hamilton (1805-65) considered this problem and showed that no genuine
'algebra' could be constructed, although a system of elements (which may be
called vectors) of the form ai + bj + ck where a, b, c are real numbers and i, j, k
denote [1,0,0], [0,1,0], [0,0,1] would obey some of the usual laws, and is
useful in 3-dimensional mechanics., for example.
However, Hamilton did construct an algebra of ordered quadruplets [a, 6, c, d]
or quaternions, in which all the usual laws are obeyed except that xy =# yx. We
could similarly think of an 'algebra' of n-vectors [ax ,a 2 ...,a] or even of
infinite sequences [a v a 2 a 3
.].

Further,

(2)

. .

we may consider arrays with more than one row,

e.g.

real numbers. With suitable laws of combination, the


algebra of rectangular arrays or matrices would be obtained. Although such an
algebra was originally investigated for its own sake, it is now widely used,
e.g. in theoretical physics and much geometry. Nor is there need to restrict
the elements to be real.
The reader must seek elsewhere for development of these suggestions.

where the elements are

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) + (5-3t).
+ 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

which agrees with


Solve nos.

7, 8,

13. 14

12

+ bi)(c di)
+ di)(c di)

(ac + bd)
c2

+ (be ad) i
+ 2
<2

'

1)

by this method.
34-2

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

Find the values of o

for

which

z%

Zi

z is of the

form x + 0*.

16 If J(a + bi) = x + yi and x > 0, prove x % y %


express ^/(3 + 4i) in the form x + yi.

X+ Yi.

17 Express (x + yi) 3 in the form


expression for (x yi) 3

= a 2 and

2xy

6.

Hence

Write down the corresponding

18 Prove {(

- 1 i V3)} 3 =

1.

B?/ direct appeal to the definitions

numbers

+ [c, d] [c, d] + [a, 6] (commutative law for addition).


x [c, c?] = [c, d] x [a, 6] (commutative law for multiplication).
([a, 6] + [c, cZ]) + [e,/] = [a, 6] + ([c, cZ] + [e, /]) (associative law for addition).
([a,b]x[c,d]) x[e,f] = [a,b] + ([c,d]x[e,f]) (associative law for multi-

19

[a, b]

20

[a, 6]

21

22

in 13.14(1), verify that complex

(a)-(c)

satisfy the following algebraic laws.

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]

. . .

26 Consider ordered number -pairs {a, 6} of integers


but with 6 4= 0) which satisfy the following laws:

(positive, negative or

zero,

(a) {a, b)

(b)

{a, 6}

(c)

{a, 6}

{c,

d}

if

and only

if

{c,

d}

{ad + be, bd};

{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

Obtain expressions for


is

isomorphic to

(fractions).

(We

(i.e.

way.)

The modulus-argument form of a complex number

13.2

Modulus and argument

13.21

We

saw

in 13.13 that the

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

When x + yi is so expressed, and r is restricted to be positive (cf. 1.62),


then r is called the modulus and 6 the argument of x + yi, respectively
written \x + yi arg (x + yi) f We may think of x + yi as the number r
|

in direction

'

6\ a

generalisation of the idea of 'directed

number

'

as

presented in elementary algebra.

customary to write z forx + yi, and to write r = |z|, = argz;


z may be called the number of the point P in Argand's diagram, and
sometimes P is called the affix of z.
Frequently it is necessary to express a complex number, given in
the form x + yi, in terms of its modulus and argument. From equaIt

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 many-valued function, and we define its
principal value to be that in the range ttkO^tt. (The single
equation tan#

yjx does not determine argz because

it

gives two

values for 6 in this range.)

Examples
(i)

Express (5 + i)/(2 + Zi) in modulus -argument form.

+ i)(2-U) _ 13-13t _
*'
~ 4+9 ~
(2 + 3i)(2-3i)
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 modulus-argument 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

called the amplitude of

x+yi and

is

written eon (x + yi) or

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

remark here that the complex number z, represented by the


point P(x, y), can also be associated with the displacement OP in the
sense from
towards P. Given the axes, there is an exact correspondence between the complex number x + yi, the point P(x,y), and the
vector

OP. Also

see 13.32.

and properties

13.22 Further definitions, notation,

We agreed in

(1)

form x + Oi to x.

13.14 (3) to abbreviate a complex

number z

of the

customary to go further and call z real although


this double use of 'real' is most unfortunate. Similarly a number of
the form + yi is shortened to iy and called purely imaginary.
If z

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

x-axis;

and

positive or negative.

corresponds to some point on the y-axis ;

and arg z \n according as y < 0.

= 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)

Real and imaginary parts. In the complex number z

If z lf z 2 are expressed in modulus-argument form as

r(cosd + ismd),

then from
it

follows

s(cos0-Msin^),

r(cos 6 + i sin 6)

s(cos

+ i sin

<p

<j>)

by equating real and imaginary parts that


r cos

s cos

Squaring and adding gives

and

<j>

2
,

r sin 6

so that r

s sin

<j>.

s since

r,

s are positive.

The two equations now become


cos d

cos

<f),

sin 6

sin

<j>,

by some integral multiple of


denote the principal values of argz l5 argz 2 then 6 =
so that in general 6,

<f>

differ

2ir.

If 6,

(f>

The process of deducing from

zt

moduli. That of deducing argz x

z2 that

argz 2

is

\z x

|z 2

is

called taking

called taking arguments.

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 left-hand side is also a value of the right, and conversely.

COMPLEX ALGEBRA

13.23]

Conjugate complex numbers. Given a complex number z

(3)

the

497

number x yi

obtained by changing the sign of

and

conjugate of z

i is

called the

when the previous

written z (or sometimes z*

is

= x + yi,

notation

is required to denote 'average value'). Conjugate complex


numbers have the following properties.
(i) The product zz = (x + yi) (x yi) = x 2 + y* = |z| 2 is 'real'.
(ii) The sumz + z = (x + yi) + (x-yi) = 2x = 20t(z) is 'real'.
(iii) The difference z z (x + yi) (x yi) = 2yi = 1iJ(z) is purely

imaginary.
(iv)

\z\

when

but

|z|;

and negative',

z is 'real

= argz, except
form x + Oi where x < 0.

principal values) argz

(for

of the

i.e.

In the a;y-plane the points which represent z,


If z x
zx

z2

z are

images in the #-axis.

= z, then z is purely imaginary.


= z a then xx = x 2 and yx = y% hence x1 iy1 = x 2 iy2 i.e.
The process of deducing from z x = z 2 that z x = z 2 is called

(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 ,

From the relation

(4) Properties of\z\.

zx z2 ,

shows that

Z X Z 2^ Z l Zi)

z i z %\

Hence
\

since both sides are positive


2

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|

The cube roots of unityf


is

a number whose cube

is

+ 1,

then z3

(z-l)(z 2 + z+l)

hence either z -

or z 2 + z+

by

0, i.e.

0;

13.14

(2).

From the latter,

= -f, z+| = i|V 3 and 2 = K-i*V 3 )Since cos tt = \ and sin 7r = \ <J3, the first of these roots
modulus-argument form cos \n + i sin tt. We denote this by cj.
(z+)

A simpler approach will be given in

14.13, ex.

(ii).

has

COMPLEX ALGEBRA

498
Also,

cosfTr

= | and

cos f71 + i sin fn. It

^ow
(jfi

= \ ^3,

sin \-n

[13.23

so the second root

is

also w, as is clear from the original formula for z.

is

(cos f 77 + i sin fit) 2

cos 2 f7T sin 2

cos jn + i sin f/r.

7r

+ 2i sin 7r

cos

7r

+ z + 1 = can be written z = to, w 2 There are


roots of + 1 in complex algebra, viz. 1, (o, a> 2 The

_
2
A
2

Hence the roots of


thus three cube

z2

relations

often simplify calculations.

Example
Direct expansion shows that

= a3 + 6 s + c8 3abc.

c x + c 2 + c 3 reveals the factor a + b + c.


Since the expression on the right can be written

The operation c x

a3 + (bw) 3 + (ceo 2 ) 3 - 3a(b(o) (co 2 ),


a + bo) + co) 2 Likewise, since the expression
a3 + (6w 2 ) 3 + (cw) 3 3a(6w 2 ) (ceo), a + 6w 2 + ceo is a factor.
We now have three complex linear factors, so that any further factor must
be numerical. Comparison of the coefficients of a3 shows that it is + 1. Hence
similar reasoning reveals the factor

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 -bc-ca-ab)
(see 10.22 (2),

(a

example; or Ex. 11
2

+b +c
2

+ b + c) (a + bo) + cco 2

(a),

no. 20),

-bc-ca-ab =

(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 1-i.

a, 6,

ar,

6+iJ3
V
4_2

2+i
(l+i)^.-

9 (cosa + *sina) + (cos/fl + isin/?).


y are real, express a 2 + b 2 in terms of

^.

If the coefficients in the equation z l +pz 3 + qz 2 + rz + s


there is a purely imaginary root, prove r 2 +p 2s = pqr.
11

^3

are 'real',

and

COMPLEX ALGEBRA

13.3]

12 If

(z
|

- 1 )/(z + 1

2,

prove that the point P which represents z lies on the

^+^+^+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

find the locus of

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*/ +

can be written in the form

/ )

= (a- <ob) (a - (o 2b), etc.]

*19 Prove that (a3 + b 3 + c 3 -3abc)(x 3 + y 3 + z 3 -3xyz) can be written in the

+ Y3 + Z 3 3X YZ, where either


Y = cx + ay + bz, Z = bx + cy + az,
(i) X = ax + by + cz,
Y = cx + by + az, Z = bx + ay + cz.
(ii) X - ax + cy + bz,
or
[In (i), (a + ba) + <xo*)(x + y<o 2 + z(o) = X+ Ya) 2 + Z(o, etc.]
*20 If ar (r = 1, 2, 3) denote the cube roots of - 1, prove
(l+xx^x^Dil+x^ + x^Dil+x^+x^al) = (l+x 3 )K
[A typical factor is 1 - 8a 3 )/( 1 - xa) = 1 + 3 )/( 1 - aa).]
21 If the equations x3 = 1, gkcs + 6a; + c = have a common root, prove
a3 + b 3 + c3 -3abc = 0.
[One of the three roots 1, to, to 2 of 3 = 1 satisfies the other equation, so
2
10
(a + b + c) (aw 5 + bo + c) (aw + ba> + c) = 0.]

form

tc

a:

13.3

Applications of the Argand representation

13.31 Geometrical interpretation of


If z

is

represented

Also

\z\

J(x

+ y2

is

represented

(a?
|

by the

by P, then

modulus and argument

a +
)

- y2

j{(x 1 -x2 ) 2 + (y 1 -y 2 ) 2 }

'distance formula' of coordinate

geometry.

Fig. 128

by OP.

COMPLEX ALGEBRA

500
If

the principal value of arg z, then

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

and represents the angle of turn from Ox towards


L(Ox,P% Px ).
Examples (some

X,

written

loci)

is the fixed point z x

and P

is the variable

point

z,

then

z x cis the circle with centre Px and radius c ;


(ii) the locus arg(z z
= is the half-line PX P for which /_(Ox,Px P)
x
(The other half is given by arg (z z x = n + a.)
the locus

(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

(x-xJt + iV-Vi)* = (z-*a) 2 + (2/-2/ 8


by

If P,

(iv)

radius

are the points

w-

z,

2z + 3

4i and P moves on a circle of centre

2Z +

B-U. Then

tt>

- 3 + 4 =

|w-(3-4*)|
since

\z\

r.

Therefore

(1)

sum and

Similarly,

a3 +

zx
j"

Since

2r.

difference of z lf z2
z2 ,

and complete the parallelo-

projection of

OP3 on Ox

projection of

OPx + projection of PX P3

projection of OP\ + projection of OP2 f

m=
,

2r

by projecting on the

+ z2

\2z\

Sum. Let Px P2 correspond to z x


Then
x

P2 OP P3

x3

z3

2z, so

Q lies on the circle of centre (3, 4) and radius

13.32 Constructions for the

gram

and

locus of Q.

r, find the

Write

i.e.

a
)

simplifying.

xi + xi) + %i + 2/2)

so that

OP2

y-axis

is

we

find y3

= x x + Wi) + x z + Wz),
(

P3 represents the sum z + z

equal and parallel to

= yx + y2 Hence

P P8
1

2.

and in the same sense.

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

sum-construc-

corresponding to z x z %

y,

/
L

/
/

.Jp*
/

p*
Fig. 129

As OP

is

equal

13.33

The

and in the same sense as

to, parallel to,

can represent z x z2 by the

displacement

l5

triangle inequalities

By applying the theorem that

'the

sum of two sides of a triangle is


OPx P3 we have
,

OP3 ^OPx + Px P3
OP3 ^OPx + OP2

i.e.f

and hence
this

|z,

we deduce

+ z2 ^
\

\z t \

+ \z2

\.

the following, exactly as for real numbers

\z x

-z 2 >

\\z x

|z 1

+ z2 ^

||^l|

\-\z 2 \\,
I^H'

These inequalities should also be verified geometrically from


The
do

so,

we

P2 P v

greater than the third side' to triangle

From

P2 P

results

can

all

we first observe

be proved algebraically, without appeal to a

fig.

figure.

that
0t(z)

= **S

f Equality is included in case 0,

V(*

P P
lf

+ 2/ 2 = M
)

(and hence also

P3

129.

are collinear.

To

COMPLEX ALGEBRA

502

with equality when and only when z


where x 2s 0. (Similarly J{z) < |z|.)

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 |*

complex numbers. As M{z x z 2 )

Z\Z % are conjugate


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 |.

|,

modulus of a complex number is


Equality occurs if and only if z x z 2 is 'real and positive',

positive signs being chosen because the


positive

by

definition.

a2z 2 where a is 'real'.


The other results can also be established

i.e.

zx

similarly.

13.34 Constructions for the product

modulus-argument form.

It is convenient to use the


(1)

P P
x,

and quotient of z z2

Product. Let z x r(cos 6 + i sin 6), z 2 = s(cos


have polar coordinates (r, 0), (s, <f>). Then

zx z2

=
=
=

+ i sin

(j>),

+ i sin 6) (cos (p + i sin


rs{(cos 6 cos sin sin 0) + *( sm # 008 ^ + 008 # sin
rs(cos 6

rs{cos (0 +
is

so that

<j>)

<j>

Therefore z x z 2
{rs,d +

<f>

<p) + i sin ((9 +

represented

(j>)}

<j>)}.

by the point which has polar coordinates

(f>).

A be the point whose polar coordinates


are (1,0). Draw triangle OAPx and construct triangle OP2 P3 directly
similar to OAP1 (fig. 130). Then
To

construct this point, let

=
i 6J'

ops = oPl .OP^ rs

80

x6p3 = #ap2 + p2 6p3 =

Also

a*5p2 + ^0Pi

Hence P3 is the required point.


The above calculation verifies the
shows that (cf. the end of 13.22 (2))
arg

Ordinary equality

may

(z x z 2 )

result

\z 1 z 2

arg z x + arg z 2

not hold

when

all

=
=

<p

+ e.

Iz-J

|z 2 |,

and

three terms have their

principal values; e.g. if


zx

z2

cos fn + i sin

7r,

then

z x z2

also

cos %n + i sin tt,

COMPLEX ALGEBRA

13.34]

503

and the principal values are


arg Zi

We

arg z 2

arg {z x z 2 )

fn,

= - n.

observe that multiplication of a complex number z 2

r(cosd + i&md) turns the corresponding vector

and multiplies

its

length

product-construction in

We

by r.

z4 z 2 and so we can use the


making Px take the role of P3 there.
therefore construct triangle OPx P2 and then make triangle

If z 4

(2) Quotient.

OP^A

by

OP2 through angle

z 1 /z 2 ,

then

zx

(1),

directly similar to this

Alternatively, the construction


Zj

z2

131).

(fig.

can be obtained from

_ 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 <j> + sin

cos

= - {cos (6 -

<j>)

+ i(sin
2
<p + sin

sin 0)

+ i sin (0 -

cos

cos

sin <j>)

(j>)}.

This calculation

verifies

that

K|/|z 2 [, and shows that

with the usual understanding about many-valuedness.

COMPLEX ALGEBRA

504

[13.35

Remark. It also follows that


y

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

P3P2 Pi in the sense from P P3


2

towards P^P^

Fig. 132

on the Argand

13.35 Harder examples


*(i)

Interpret geometrically the equations


-

a;

(6)

(k 4=1).

we have P2 PP1 = a. Hence the locus of P is an


and containing angle a.
prove that the locus of P is a circle (a circle of Apollonius with

By the Remark in

arc of a circle through


(6)

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

We are given that APjPB =

where A, B, represent

#= 1,

AB internally at H and externally at K in the ratio k:l.

drawn assuming k >

1;

the reader should illustrate the case k

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 right-angle. Since H,
are fixed
points, P lies on the circle having HK for diameter.

Algebraical proof.

From the hypothesis,


Z - Zl 2 = F|z-2 a
|

(x-Xi)* + {y-Vi)

i.e.

and

since k

4= 1

this represents

2
|

= k*{x-x 2 )* + k*{y-y 2 )*;

circle (see 15.61).

t The implied change of notation from


convenience in writing.

P Ps
lf

to

A,

respectively

is

made

for

COMPLEX ALGEBRA

13.35]
It is easy to

505

prove that, if the centre is O, the radius OH of the circle is given by

OH* = OA. OB.


For since

AH/HB = AK/KB,

follows that

it

OA-OH OH -OB
OA + OH~ OH + OB'

AHjAK = HB/KB,
20 _ 2 OH

and so

20H~ 20B

by properties of equal ratios, which gives the result.

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.

the circles are orthogonal.

Prove that the triangles P^P^P^,


Zi

P P P6 are directly similar if and only if


4

Z4

22

Zs

(a) If the condition holds, then by taking moduli, P 1 PJP3 P 2 =


and by taking arguments and using the Remark in 13.34,

in the same- sense.

f 8-P 2-^*1

and are

P4PJP 6 P6

= P 6-P&P

Hence by the

sides proportional', triangles P^^P^P^,

'common

test

angle and containing

P P P6 are directly similar.


4

the triangles are directly similar, then the above two


z i )((z3 z 2 ), (z 4 z 5 )l(z 6 z 5 ) have the same modulus
and the same argument. They are therefore equal.
The condition can be written
(6)

Conversely,

if

relations hold, so that (z 1

z1

0,

z2
1

By

rx

ri + z 2 rs followed

by

r 2 -> ra + z 5 r3
Zl

Z2

2-2

Z4-Z5

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,

P3 are such that Px P2

= 2P 2 P3

+ 2z 2

zx

2z 2

zx , z 2 , z3 .

Write down the complex number represented by the point dividing

in the ratio k

P if 2 <

What is the locus

Find the greatest and

6 Prove that

of

|z

+ 1 - 2i\ <

least values of

+ l|

if

3?

(i)

\z

4|

if

\z\

1; (ii)

|z

+ z 2 2 + |zi-z 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.

|z-3|

find the relation

(i)

+ z2 =

|zj.

prove that argz x and argz 2

z i~ z a|

(principal values intended),

(ii)

If arg{(z 1 + z 2 )/(z 1

2 )}

differ

by

\n, prove

or

\it
|Z]J

\ir

|z 2 |.

[Treat geometrically.]

8 Verify the following construction for the roots of z 2 2az + b 2 = 0, where


are real and < a < b. With the origin O for centre, draw a circle of radius 6.
From the point on Ox at distance a from O draw a perpendicular to Ox cutting
the circle at P, Q. Then P, Q represent the roots.
a, 6

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

10 Given the points representing z x z 2 construct those representing the two


values of V( 2 i z 2),

11

is

any

chosen so that
12 If
bisects

point,

and on the

P OP =

PxOP 2 =

on OPx as diameter points


Prove z\ cos 20 = z x z z cos 2 0.

circle
<}>.

G is the centroid of triangle PX P 2 P3 and

[Use Ex. 15

x (?.

no.

(a),

4z x

+ z 2 + z3 =

0,

P P
2,

are

prove that

1.]

13 If A is real and z = 4A + 3*(1 A), prove that P lies on a straight line.


As A varies, prove that the least value of |z| is 2-4, and interpret geometrically.
14 If z x 2 = z 4 z 3 prove that Px P t Pi Pl is a parallelogram and that the
,

point \{z x + z 2 + z3 + z 4 )

is its

centre.

that the point P 2 where z 2 = (l+AiJz! and A is real, lies on the


Find A if argz 2 = argZj^ + i^r, principal values
perpendicular at
x to OP x
being intended.

Show

15

16 If zx z 2 are the roots of z 2 az + a? = 0, where a is a complex number, prove


that the points zx , z 2 are the vertices of the equilateral triangles drawn on
opposite sides of the line 0.4.
,

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

l/z 2 + l/z 3 prove that OP 2 P x P3 is a cyclic quadrilateral.


z 2 )/{0 z 2 = (z x z s )/(zs 0); take arguments.]

is (z x

[The

COMPLEX ALGEBRA

13.4]

507

20 Interpret geometrically the equation

On

*21

the sides of triangle

directly similar to each other.

[By

13.35, ex.

(ii)

bc

drawn

are

and properties of equal


xc

so Hoc

ABC

triangles

BOX, CAY,

Prove that the centroids of ABC,


ya
ca

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-

zz z is directly similar to the triangle

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 = Iza-Zilandargftzs-ZiV^i-Sa)} = + \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 left-hand 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

Factorisation in complex algebra

13.41 'The fundamental theorem of algebra'

We remarked in 13.15 that complex numbers cannot be generalised


by attempting

to solve equations of degree higher than

2.

This

is

consequence of the following theorem.

In complex

algebra the polynomial equation

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,

assume it here; the reader may


the complex integral calculus.

later

study a proof which depends on

13.42 Roots of the general polynomial equation

Defining 'repeated root of order r* as in real algebra (10.43), and


reckoning such a root as r roots, it is now easy to prove the following, f
Theorem I in 10.13. Both the
f This result should be compared with the weaker
Remainder Theorem and Theorem I are valid in complex algebra, their proofs being
unchanged.
oe

GPMII

COMPLEX ALGEBRA

508

In complex

algebra the equation

[13.42

has exactly

(i)

roots.

By the fundamental theorem there is a root z

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;

since z ol1 divides each bracket, to

i.e.

(z-<x 1 )f1 (z)

where fx (z)

is

0,

w-1

a polynomial of degree

in

whose

z,

first

term

is

71-1

(as is clear by starting the division).


PqZ
Again by the fundamental theorem, fx (z)

necessarily distinct

from

ol ),
x

and by

/i(z)

where

a polynomial in
Continuing thus, we find
2 (z) is

p(z)

where n-1 (z)

is

degree

(z- ai )(z-a 2 )

= a2

(not

+ k,

zn 2 .

n 2, beginning with p

in

or

= Po( z ~ ai) (2 - a 2

. . .

and begins with p

z,

(z

cc n

has n roots a l5 a 2

z;

...,an

z; it

say. Therefore

- aJ.

(z

This shows that p(z) has n factors linear in


the equation p{z)

(z-a^J/^z),

...

a polynomial of degree

P(z)

has a root

(z-oc 2 )fz {z),

z of

must therefore be of the form p

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(A-ai)(A-a 2 )-(/?-aJ

and

since

cc2
f}

by

4=

If the roots are not all distinct,

where nx + n 2 +
is

. . .

4-

0,

hypothesis, at least one of the factors ficx,^

fi a n must be zero (13.14(2)),


Hence p(z) = has exactly n roots.

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

unique (apart perhaps from the arrangement of the factors). For


p(z)

an

takes the form

= p^z - ctj** (z - a 2 )

n k = n and ot lt

coincides with

i.e. /?

(z

- /?2 r

. . .

(z

if also

(iv)

., /?. are distinct, then each /? must be an a, since otherwise from


while from (iv) p(fir ) = 0; similarly each a must be a (i. Suppose

. .

COMPLEX ALGEBRA

13.43]

509

then (iv) can be


= oc /?2 = a 2
fi x
(v)
= p (z - a x )"* (a - a )* (z - a*)**,
It now follows that mr = nr (r = 1,2, ...,&) E.g. if 7% < n x then from (iii) and
mi
(v) we should obtain, after dividing out p (z oc x
m*,
TO
n
n
n
* = (z - a 2
(z - a
(z - a
(z - oc x
(z - a 2
which is impossible since the left-hand side is zero when z = cc but the rightthe notation has been chosen so that

witten

x,

p(z)

. . .

. . .

ft )

>

. . .

fc )

hand

side

is

not.

13.43 Principle of equating coefficients

We can now deduce results for complex polynomials corresponding


to

Theorem II and its corollaries in

1 0. 1 3.

The enunciations and proofs

are the same. In particular, the 'principle of equating coefficients'


valid for polynomials in

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 )-

Consider the expression

When

expanded,

it is

a polynomial in A of degree

The given equations imply that


Hence it must be identical with

3,

A3
when A = Alt A 2 A 8

with leading term

this expression is zero

-(A-A 1 )(A-A a )(A-A3

=
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

).

13.44 Repeated roots

The

),

z3 .

and the derived polynomial

definitions of 'repeated factor', 'repeated root' in 10.43 are

retained for complex polynomials. Continuing to write


p(z)

we

ss

zn

~
+p 1 z n 1 + ... +pn

define the derived polynomial of p(z) to be


p'(z)

= p^nz *-1 +p 1 (n - 1) z n ~ 2 +...


1

The reader may wonder why we do not give a


<p'(z)

denotes

lim

p(z + h)

definition like the following:

_ p{z)
,

35-2

COMPLEX ALGEBRA

510

[13.45

means that and n\ tend to zero


is complex (say h = g+irj) and h
separately and in any manner.' There are difficulties underlying this procedure;
they can only be fully examined in dealing with general functions of a complex
variable, a subject which may be studied at a later stage.
where h

From

the definition given

it is

straightforward to verify the usual

and in particular the product rule,


of complex polynomials'. The theorems
10.43 remain valid in complex algebra

rules of the differential calculus,


in this 'differential calculus

on repeated roots given in

since their proofs are exactly as before.

13.45 Equations with 'real' coefficients; conjugate complex roots

// the

(1)

and

if a

is also

coefficients

+ fii

(where

fi

p p ly

=f=

pn

0) is

in the polynomial p(z) are 'real',

root ofp(z)

0,

then

its

conjugate

a fii

root.

The condition for a

Proof.

fii

Po(a+Pi) n +p 1 {oc + /3i) n

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

occurs, this equation can be written

P + Qi =
where P,

0,

are real polynomials in (a, fi).

This condition is equivalent to the separate equations P

Hence also P Qi =

0* =

0,

Q=

0.

which (because p(z) has real coefficients)


is equivalent to the statement p(a fii) = 0, i.e. a fii is a root of
p(z) = 0. See Ex. 13 (d), no. 7 for an alternative proof.
If the coefficients in p(z) are not all real, the argument breaks down
at the last step. For example, if p(z) is the linear polynomial z (1 + i),
then clearly z = 1 + i is a root of p(z) = 0; but 1 i is not a root since

p(l-i)

0,

= (l-i)-(l + i) = -2i #

The

correct statement in the general case

p(z)

0,

where

p(z) is the

0.

is that a
fii is a root of
polynomial obtained from p(z) by changing

For if p(oc + fii) = P + Qi, then we


by changing the sign of i everywhere on both
sides f (because the equality was obtained by using only the property
2
i = 1 of i, and this continues to hold when * is replaced by i);
this would give p(afii) = PQi. In the case of entirely real coeffithe sign of

must

still

in all the coefficients.

get a true result

cients, p(z)

= p(z).

It is also easy to prove the following.


f In short,

by

'taking conjugates'.

COMPLEX ALGEBRA

13.45]

Conjugate complex roots occur

For if p{z)
a root

oi-fii.

to the

Remove

same

order.

=a

has an r-fold root z

511

(yff

4= 0),

then

it

also has

the factor (z-a,-fii)(z-oc+fii) from p(z),

has an (r- l)-fold root a+fii, and


therefore also a root <xfii; and by removing the corresponding factors
from/1 (z) we obtain 2 (z), say. Proceeding thus, we arrive at a poly-

obtaining fx (z), say. Then/^z)

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

r-fold root of p(z)

0.

Example
Prove that the equation

has

ft*

62

6a

x ax

x a2

x an

n roots in real algebra if al9 a2

. .

.,

a n are all different,

the

Vs are non-zero, and

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

(a-ar a +^ 2
)

sum is non-zero, we have /? =

0.

Hence all the roots are of the form

In real algebra there are thus n roots.

a theorem in real algebra on surd roots of an equation


with rational coefficients which can be proved by an argument like
(2)

There

that in

is

(1).

If the polynomial equation p(x)


coefficients

a b^jc

is

Proof.

and if x
a root.

= a + b*Jc

yjc

(in real algebra) has rational

root (where

^c

a surd), then also

is

for

or zero. This implies that

surd

a + b Jc to be a, root of p(x) = is, when


of the form P + Q ^jc = 0, where P, Q are rational numbers

The condition

simplified,

is

P=

would be equal to the

and Q =
rational

0, for

otherwise

contradiction. Since

p(a-bjc) =

P-QJc = 0-0 Vc =

a b <Jc

is also a root. Also see Ex. 13 (d), no.


a + b *Jc, a bjc are called conjugate surds.

f That

is,

if

=f=

number P/Q, which

8.

root of the corresponding polynomial equation

(z-a1 )(z-ai )...(z-an )f(z)

= 0.

0,

the
is

COMPLEX ALGEBRA

512

[13.5

Relations between roots and coefficients

13.5

13.51 Symmetrical relations

In 13.42 we proved that

p
so that

zn

+p 1 zn

a lt a 2

~1

+pn s p

+...

(z

- ct^ (z - a 2

. . .

(z

- aJ,

a n are the n roots of the equation p(z) = 0. On


it becomes (cf. the proof in 12.1 1)

multiplying out the right-hand side,

- (Sa) z"- 1 + (S ai a 2

{z n

zn

~ 2 -...

Hence on equating coefficients of zTC_1

2a =

n~2

. . .

aj.

and the constant terms,

Sa, a 2

S ai a 2 a 3 = -^,
and

+ (- l) n

ai a 2 ...an

= (-l)^;

in general,

^YPrlPo sum f t ne products of the roots taken r at a time.

This work generalises the results of 10.3, which are theorems of

and are valid only if the equation under consideration


possible number of roots. The above results in
complex algebra hold without any such restriction because every
polynomial equation of degree n has exactly n roots.
real algebra,

has the

maximum

13.52 Unsymmetrical relations

We remarked in

10.32 that the symmetrical relations do not help

us to solve the given equation because they are equivalent to the

information that
if

'

cl^cl^, ...,an

some non-symmetrical

given, then

it

are the roots of p(z)

relation

between some or

all

0'.

However,

of the roots

is

may be possible to solve the equation.

Example
Solve (in complex algebra) the equation
3x*

+ 17a; 3 - 5a; 2 + 8a; + 12 =

given that the product of two roots is 4.


Let the roots be a, fi, y, 8, and suppose

a+fi + y + d
(There

is

4.

We have

fiyS+y8a + doifi + cc/3y

-f,

afiyd

= f but we shall find that this


4 and the third relation, yd = 1.

also the relation Ea/?

From a/? =

a,/H

0,

is

4.

not needed.)

COMPLEX ALGEBRA

13.53]

513

The second relation can be written

a/% + <S)+ytf(a+/?) =
+ (a+/?) =
4( r +

so that

-f,
-f.

<>)

This together with the

relation can

first

be solved to give

7 + ^=1.
4
+
+^# = 0, viz. |, 6; and
w2

Hence

a,

ft

are the roots of

roots of x % x + 1

a;

w,

0, viz.

y, 8 are the

13.53 Transformation of equations


Properties of the roots of an equation are often conveniently dis-

cussed

by

considering another equation whose roots are related to

those of the given equation in a

(iii)

less by k than, (iv)

p(x)
(iii)

0.

The

We now show how

known manner.

to form the equations whose roots are

(i)

reciprocals of,

(ii)

k times,

squares of the roots of the polynomial equationf

process

(ii) is

known

as 'diminishing the roots

by

&',

as 'multiplying the roots

and

by k\

as 'squaring the roots'.

(iv)

Applications are given in 13.72, 13.73.

Write y

(i)

nomial equation

If

ljx.

the equation p(l/y)


is

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

the required equation reduces to

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.

= 1 gives the equation whose roots are the negatives


= ('changing the sign of the roots').
=
x k. When a satisfies p(x) = 0, a k will satisfy
y

of those of p(x)
(iii)

Write

p(y + k) = 0.
(iv) Write y

= x 2 When a satisfies p(x) =


.

This can be rationalised

(i.e.

0,

a 2 will satisfy p{*Jy) =

freed of square roots) as in ex.

(ii)

0.

below.

Examples
(i)

Transform

4a; 3 + 6x + 5 = so that the coefficient of the


introducing
without
fractional coefficients.

the equation 3a; 4

leading term becomes

Multiply the roots

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

equally applicable to equations other than polynomial ones.

COMPLEX ALGEBRA

514

[13.53

By inspection, the coefficients will be integers when h =


such value). The new equation
4
2/

where y

3 (this

is

the smallest

then

-42/s + 54y+135

0,

3x.

Find the equation whose

(ii)

is

roots are the squares of the roots of

3x3 -x* + 2x-B

(a)

0.

Write the given equation as


x(3x 2 +

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.

*Jy satisfies (6), i.e. the given equation (a). In either


the square of a root of (a).
Further examples of the transformation process are given in 10.32, ex. (ii)

Thus

either +*Jy or

event, y

and Ex.

is

10(c).

Exercise 13(d)

The algebra
1

is

complex.

Prove that the equation


1+i! +

z2
2i

z3

z"

+ 3T + - + ^=

has no repeated roots.

2 Solve
3

z4

4z 2 + Sz 4 =

Given that 2 + ^/3

is

0,

given that

+ i is a root.
= 0,

a root of x9 2a; 2 7x + 2

solve the equation

completely.

4 Prove that w is a repeated root of


hence solve the equation.
5

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.

6 The linear equation az + b = with real coefficients a, b has a root a


and therefore also a root <x~fii: two roots for an equation of degree 1.' Explain
'

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

not always be the case; e.g. the square of each root of x3 x =


yet the root y = 1 is not the square of a root of the adequation.
Indeed, there is not a unique result in this instance: the method of ex. (ii) gives
2
y(y I) = 0, and clearly y\y 1) = also satisfies the requirements. Such ambiguity
can arise whenever the squares of the roots of the given equation are not distinct.
f This

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

*8 Prove the theorem about conjugate surd roots in 13.45(2) by a method


7. Also prove that such roots occur to the same order.

similar to that of no.

9 Ifa,# y, S are the roots of x* +px3 + qx 2 +rx +s

0,

find Zafiyia+fi+y).

+px3 + ga; 2 + rx + s = is equal to the product


Prove
three.
ps + r) 2 = s(q + s + 1 2
* 1 1 Find the sum of the cubes of the roots of x 5 + x 2 + x 1 = 0.
One

10

root of

a;

of the other

Solve the follouring equations.

12 x 3
13

5a;

16* + 80 =

if

xi + 3x3 -5x i -6x-S

the
if

14 x* - 2xz - 2 1*2 + 22a; + 40

sum of two roots is zero.


sum of two roots is -2.

the

if

the roots are in a.p.

by k, and find for what value


Increase the roots of xs + Sax 2 + Bbx + c =
of k the new equation contains no quadratic term. (This process is called
'removing the second term from the given equation'.)
15

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;

+ 8a;8 -18a; 2 -16a; -3 =

into one with integral coefficients of which that of the leading

term

is

+ 1.

18 If a, f$, y are the roots of x3 +px2 + qx + r = 0, find the equation whose


(y- l)/y.
roots are (i) a(fi+y), /3(y + a), y(x+P); (ii) (a- l)/a, (/?19 If the roots of x n

are

1,

ax a 2
,

prove that

n _ 1?

. .

{l-a 1 )(l-cc 2 )...(l-cc n =


)

[Construct the equation whose roots are

*20 If a, /?, y are the roots of x3 +px + q

1,7
[y

+2 =

(fi*

1, 1 a
0,

n.

lf ....]

form the equation whose roots are

7* *fi
ay pa

7 p
+ y 2 + 2/?7)//?7 = a 2 /^ = - a 3 /q = pa\q + 1.]

13.6

Factorisation in real algebra

13.61

Roots of the general polynomial equation

(1)

We have proved

that complex roots of an equation with real

coefficients occur in conjugate pairs (13.45 (1)). Since

(z-a-j3i)(z-a + fii) = (z-a) 2 + /?2


the result (ii) of 13.42 shows that any polynomial p(z) of degree n with
real coefficients can be expressed in the form
2
{(z - b,f + cf}, (vi)
P(z) s P Q {z - a x ) (z-az )...(z- ak ) {(z - 6 X ) + cfj
where
k + 2l = n,
,

. . .

t Cf. 10.11,

Remark

(yff).

COMPLEX ALGEBRA

516

and

all

tinct,

constants

are real.

The

all linear (the

case

a, b, c

and may be

factors
I

[13.62

shown need not be

dis-

0) or all quadratic (the case

0).

(2)

there
p{x)

Writing
is

p(x)

= p xn +p 1 xn ~1 + ... +pn

an identity

=p

{x

- a x (x-a 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

must be even, and so either all factors


number of linear ones will be present.

Remembering that

in real algebra

will

no quadratic factor can be

be

zero,

we have the following result.


In real algebra an equation of even degree has either no roots or an
even number of roots; and an equation of odd degree has an odd number
of roots {and therefore at least one).
Remark. Just as equation (ii) in 1 3.42 shows that, in complex algebra,
every polynomial can be resolved into linear factors, so equation (vii)
above shows that factorisation in real algebra requires no more than
linear and irreducible quadratic factors. Both results are existence
theorems, i.e. they tell us that the factorisation can be done in a certain
manner, but they provide no process for actually doing it. The results

are of great theoretical importance.

13.62 Location of roots in real algebra

Rational roots of an equation with rational coefficients. By the


method of 13.53, ex. (i), the given polynomial equation can be trans(1)

formed into one in which the

coefficient of the leading

other coefficients are integers.


If the polynomial p(x) = xn

all

+ p^

coefficients,

11

-1

is

well

Proof. Suppose

known and is

h\k

is

+ 1 and

+ p 2 xn ~ 2 + ... +p n has integral

then every rational root must be

ofPnThis result

term is

We suppose that this has been done.


an

integer

which

is

a factor

often used (e.g. in 10.12, ex. (i)).


a rational root of p(x) = 0. Without loss

COMPLEX ALGEBRA

13.62]

of generality
factor

we can assume
0. Then

that

h,

517

common

k are integers with no

and that k >

^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.

Thus the rational root is in fact


~

p n = - hih"-1 + Pl hn -* + ... +pn - 1 ),

a factor of

Change of sign of a polynomial.


If the polynomial p(x) is non-zero when x = xx and when x
between xx and x2 is
the number of roots ofp(x) =
(2)

x2 then
,

odd ifp(x^), p(xz ) have opposite signs,

and

even or zero

if p{x^),

p(x 2 ) have

the

same

signs.

These results are 'intuitively obvious' properties of all continuous


(cf. 2.65). For a polynomial p{x) they can be proved as

functions
follows.

By result

(vii)

in 13.61,

2>(a)

=p

(x

- a x {x-a 2 )...(x- ak
)

g(x),

a product of factors of the form (#-&) 2 + c 2 or else (in


the case when p(x) consists entirely of linear factors) is 1. In either

where

g(x) is

event, g(x)

is

positive for all values of x.

If p(Xj), p(x2 ) have opposite signs, the linear factors cannot be


absent,

and (a?x ax

all

(xx a 2 )... {x 1 ak ), {x 2 a {x 2 a2 )... (x2 ak


signs. Now x t a,x2 a have like signs unless a
)

must have opposite


ak lie between
lies between xx and x 2 Hence an odd number of a1} a 2
xx and x2
Conversely, if an odd number of roots he between x1 and x2 then the
above products must have opposite signs, so that also p{xx ), p(x 2 )
have opposite signs.
I p(x1 ), p(x2 ) have like signs, it follows that the number of roots
between x x and x2 cannot be odd, i.e. it must be even or zero.
.

. .

COMPLEX ALGEBRA

518

[13.62

Examples
Let

(i)

be the numerically greatest

ps

two or more are numerically equal)


p(x)

x n +p 1 x n

Ip^-i + .-.+p^x+Pnl ^

Then

|a;|

>

~1

+p n

+...

one such coefficient

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

Given any positive proper fraction, e.g. \, we shall have


such that \x\ > 2 p s + 1 = K', say. Hence if
> K\
\

p{x)

|i? g |/{|a;|

is

x n (l

+ d)

\x\^K,

for all

where

<

\0\

1} ^

if

J.

This shows that p(x) has the same


x ~$t
and all x < i:
that p(x) is dominated by
for large (positive or negative) x. Thus
p(if), p( K) have opposite signs if n is odd, and the same sign if n is even;
since
can be as large as we please, we have proved the theorem in 13.61 (2).
The symbols p{co), p{ <x>) are customarily used to denote '#>(i), p(K)
sign as x n for all

we say

K
where K

is

arbitrarily large'.

Solve the example in 13.45 (1) by a 'change of sign'' argument.

(ii)

We may suppose the notation to be chosen so that a < a % <

. . .

<

a n Con.

sider the polynomial

p(x)

(x - a x ) (x - a 2 )

bl{x-a2 ) ...{x-a n ) + b\{x-a1 )(x-as )

+ b%(x - ax

. . .

. . .

(x - a n ) f(x)

(x - a n_!)

+ k(x - a

x)

(x

(x-a n + ...

...

- a2

. . .

(x

- a).

When x takes the values


00
p(x) has the

In

this

=|=

0, it

aa

...

0"n-\

an

(-1)"" 1

(-1)"- 2

...

-1

+1

k.

same sign as
k(-l) n

negative.

ax

sequence there are exactly n changes of sign, whether k is positive or


Hence p(x) = has at least n roots. Since p(x) is of degree n when
cannot have more than n roots. The number of roots is therefore

precisely n.
(3) Rollers

theorem for polynomials. This

general theorem in 6.21; see 6.23(1),


there.

and

is

a special case of the

also the deductions

An algebraic proof is indicated in Ex. 3 (e), no.

(2), (3)

14; also see no. 15.

COMPLEX ALGEBRA

13.62]

519

Examples

Find the number of roots of x* - 2a;3 + 6a:2 - 10a; - 8

(iii)

locate

them between consecutive

Writing

=x - 2x* + 6a; - lOx - 8,


l

p(x)

then

p\x)

4a; 3

in real algebra, and

integers.
2

- 6a; 2 + 12x - 10 =

2(x -

2
1) (2a;

- x + 5),

the factor x

1 being discovered by trial.


Hence p'(x) = has one root in real algebra. By 6.23(3), p(x) = cannot
have more than two roots; and if p(x) = actually has two roots, the root x = 1
of p'(x) =
must lie between them. We therefore consider the sign of p(x)

CO
+

X
Sign of p(x)

+00

and the other greater than 1.


By trial we find that p(0) < 0, p( 1) > 0, so there is a root between and 1.
Similarly we find p(2) < 0, p{3) > 0, so the other root lies between 2 and 3.

Thus p(x)

(iv)

has two roots, one

Find the range

than

of values of k for which the equation

p(x)

has 4 distinct

less

= x*-2Qx 2 + 4:Sx-k =

roots.

p'{x)

Possible roots of p{x)

8
4(a;

-13a;+12)

4(a;-l)(a;-3)(a; + 4).

= must be separated by roots of p'(x) = 0, viz. 4,


+oo
4
3
-oo
x
1

1, 3.

Value of p(x)

+oo

-352-A;

23-A;

-9-k

+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.

= has 4 distinct roots which all


1 and + 1.
has roots +1, 1 each four-fold.
Putting f(x) = (a; 2 -l) 4 then f(x) =
has roots +1,-1 each three-fold (by 10.43) and a root x = a
Hence/'(a;) =
between +1,-1 (by Rolle's theorem). Thenf(x) = has double roots +1,-1,
a root ft between 1 and a, and a root y between a and + 1. Similarly, f"'{x) =
has simple roots +1, 1, and roots between 1 and fi, ff and y, and y and + 1
has a root between each of
5 distinct roots altogether. Therefore f w (x) =
these, i.e. it has at least 4 distinct roots all lying between + 1 and 1. Since
(v)

lie

Prove that the equation d^x^l^/dx1

between

f w(x) is of degree 4, it can have no other roots.


The expression {d n (x a l) n /dx n }/2 nn\ is denoted by

n (x) and called the


Legendre polynomial of degree n. An argument similar to the above shows that
has exactly n distinct roots, all lying between 1 and + 1. These
n (x) =
polynomials arise in Mathematical Physics.

COMPLEX ALGEBRA

520

Exercise 13(c)

The

Find the rational roots


1

algebra is real.

(if any) of the following equations.

= 0.
3a; 4 - 4a; 3 + 6a: + 5 = 0.
3a^-lla; 2 + 9a;-2

Determine the number of

(real) roots of the

2a: 4

+ 5a; 8 + 14a; 2 -3a;-54 =

0.

following equations, and locate each

between consecutive integers.

= 0.
-2a: + a;-10 = 0.
If a t < a 2 < a3 < a4 <

4 x* + 2x 2 + Sx-l
6
7

x-x 3 + x 2 -2 =

0.

< a6 prove that the equation


(x a 1 )(x a 3 )(x a 5 + k 2 (xa 2 )(xa i (xa e =
a6

has 3 distinct roots for any real


8 If a
roots a,

and p < q, prove that al(xp) + b/(x q) = x has


0, 6 >
y such that a. <p < ft <q<y.

>

ft,

9 If ax a z ,

n
i
prove that the equation 2
r=l x-a 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

by applying Rolle's theorem to f(x) = (x a t )... (x a n ).


[We may assume ax < a 2 <
< a n without loss of generality.]

(ii)

. . .

10

(i)

If p n

<

and n

is

p(x)

even, prove that the equation

=p

x n +p 1 x n - 1 +...+p n

has at least one positive and at least one negative root.


(ii) Prove that the number of positive roots of p(x)
if

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

Prove that x + 3a; 3 =

12

0.

8a;

-5a4 -40a; 8 - 50 =

0.

has only one root a, and hence that

x* +

6a; 2

-12a; -9

has just two roots.

>

14 If a

1,

prove x5

5a*ar

+4 =

has 3 roots.

What happens when a <

Find the range of kfor which the following equations have 4 distinct roots.

Illustrate

the results by sketches.

15 xl -14x z +

17 If p

>

24x-k =

q and a,

ft

0.

are the roots of x


f(x)

- 16a; 8 + 6a; 2 + 12x-k =


2px + q = 0, prove that

16
2

3a; 4

= xz - 3px 2 + Zqx-r =

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

2(q -p*)* + {pq - r),

/(a)

Prove that x* + ax + b

etc.]

has 3 distinct roots if and only if 276 a + 4a8

<

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 right-hand side
< b|-M n /{H-l}> so 1 < |p,|/{|a;| - 1}, i.e. \x\ < l + \pt \. Thus any root
numerically greater than 1 must be numerically less than 1 + \p,\.]
\

*20 Replacing 'numerical value' by 'modulus', verify that the results of


13.62, ex. (i) and no. 19 hold in complex algebra, even when the coefficients
pr are not real.

13.7

Approximate solution of equations (further methods)

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

The method of proportional

parts

For a straight line, the increment of the ordinate is proportional to the


increment of the abscissa. For a curve, a small arc will usually not depart far
from the chord joining its ends, so that over this arc the ratio
(increment of ordinate)/(increment of abscissa)

not differ much from the gradient of the corresponding chord.


These geometrical considerations lead to the principle of proportional parts

will

'

'

the increment of any continuous function is approximately proportional to the


increment of the variable, the range of this

An analytical formulation and an estimate of the error are given


in Ex. 6(e), no. 15.
Suppose now that numbers a, b have been
variable being small.

=A>

found such that /(a)

0, f(b)

-B<

0.

Geometrically, the points (a, A), (6, B) are on


opposite sides of Ox, and preferably close to
Ox (a wise choice of a, b will secure this). The
line joining these points has equation

y-A =
o

(b,B)
Fig. 135

(x-a),

cuts Ox at x = c where c = a + (b a) A /(A B). This expression lies in


value between a, b and is certainly a closer approximation to the root otf(x) =

and

than either a or 6.
Newton's method

is equivalent to replacing the arc by the tangent at one


extremity (6.73 (2)). The present method replaces the arc by its chord.

COMPLEX ALGEBRA

522

[13.72

Example
Find the

root of

By

we

trial

this root is 1

xz 2x 2

= correct to 3 places of decimals.


= 3,/(2) = 2, so there is a root between

find/(l)

and

2.

If

+h, then by 'proportional parts'

J(l + h)-f(l)

/(2)-/(l)

'

2-1

+ h)-l
and since /(l +h) = 0, this gives 5 = S/h, i.e. h = 0-6.
A better approximation is therefore x = 1-6. Since
/(l-6) = -l-l, /(l-7) = -0-487, /(l-8) =
the root actually

lies

between

1-7

(l

'

and

/(l-8)-/(l-7)

1-8-1-7

1-8.

/(l-7

+ ft)-/(l-7
+ fc)-l-7

= 0-0677, and a better approximation


Applying the process once again,
1-8-1-768

032

is

1-768.

^
"

0-009

Hence h = 0-001195, and x = 1-76919 =

0-009.

places of decimals. (Taking a = 1-768, 6


shows that the error is less than 0-0002.)

13.72 Horner's

'

"

1,e

since /(1-768)

^ /(l-768 + ft)-/(l-768)

0-241
"

+ h, then

0-487

so that h

/(l-8)-/(l-768)

is 1-7

(l-7

'

0-719

If the root

+0-232,

1-8, c

1-769 to three
1-769 in Ex. 6(e), no. 15,

method

Suppose that the equation p{x) = has a root 2-76 .. and has no other root
whose integral part is 2. Thenp(2), p(3) have opposite signs; this fact, discovered
.

by trial, gives the first figure of the root.


The first figure 2 is now removed from subsequent calculations by diminishing
the roots of p(x) = by 2. The new equation then has a root 0-76 .. .. To avoid
decimals, we multiply the roots by 1 0, so that the new equation has a root 7 6
The figure 7 is discovered by showing that the last equation has a root
between 7 and 8. The roots are then (iiminished by 7, and next multiplied by 10.
The figure 6 is then found by trial. The process is repeated until the required
number of decimal places has been found.

. . .

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 + y-3

Multiply the roots

by

0.

10:

+ 30t/ 2 + I00y - 3000 =

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.

Multiply the roots

by

10:

+ 510z 3 + 66,7002 - 487,000 =

2s

523

By trial this has a root between


= 2 6, i.e. 2 = t + 6; we obtain
t

Multiply the roots

by
t

and

7.

0.

Diminish the roots by

+ 528* 2 + 72,928* - 68,224 =

6,

putting

0.

10:

+ 5280* 2 + 7,292,800* - 68,224,000 =

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.

Hence the required root is x =

1-769.

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

this is too large; writing


0(z)

23

+ 5102 2 + 66,7002 - 487,000,

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

von Graeffe's method of root-squaring

This method approximates to

the roots in one process, although the signs


can also be used to find complex roots, but
applicable to polynomial equations only.

have to be sorted out at the end.


is

36

all

It

GPMII

COMPLEX ALGEBRA

524

[13.73

Given a cubic x + ax + bx + c = 0, whose roots are a, /?, y (say), we construct


the cubic whose roots are a 2 /? 2 , y 2 by putting y = x 2 the equation becomes
2

*Jy{y+*>)

-(cm/+c),

+ (6 - 2ac) y - c 2 = 0,
^-f^ + By + C = 0,
=
A 26-a 2 B = b 2 -2ac, O = -c2

and by squaring,

i.e.

where

+ (26 - a 2

n times, we obtain an equation whose


|/?| > |y|; then for n sufficiently large,
compared with /? 2* and y 2", so that a 2" is approximately

If the root-squaring process is applied


roots are a 2*, ytf 2", y 2". Suppose that |a| >

a 2"

will be very large


'minus the coefficient of the quadratic term in this equation.
Since /? 2" is large compared with y 2", the coefficient of the linear term will be
approximately (a/?) 2". Similarly the constant term is approximately (a/?y) 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 = 26-a
-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 1-7; and this is the
only real root (e.g. see Ex. 13(e), no. 18). Since ctfiy = 2, it follows that
yffy = 2-r 1-7 = l2,i.e.r 2 = 1-2 where/? = r(eos# + *sin0), y = r(cos0 isin#);
== 1-1. 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=

1-7692;

1-0632.

Referring now to the given equation and considering the sum of its roots,
a + 2rcos0 = O, from which cos 6 = -0-83215 and 6 = 146 19'. Hence
r(cos 6 i sin 6) can be found. The three roots are approximately
1-7692,

- 0-8847 0-5896*.

Exercise 13(/)

The

algebra is real.

Solve the following equations approximately, correct to at least 2 places of decimals.

2 3a: 3 -9a: + 2 =
x*-2x-5 = 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].

= 20x [2 roots, each to three places].


= 1*5. [This has a root between 1 and
log a; = log 1-5.]
*11 Find the smallest positive root of x = tan-1
9 10*

*10 x x

2; write the

equation as

a;

a:,

correct to four places of

decimals.

*12 Find the smallest positive root of sin a;

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.]
.

Miscellaneous Exercise 13(g)


1

If c

and the number (1 + i)/(2 + ci) + (2 + 3*)/(3 +i) is represented


by a point on the line y = x, prove c = 5*J21.
=l, prove (1 + c + *s)/(l +c is) = c + is, and write down a

is real,

in the an/-plane

2 If

c2

+ s2

similar result for

3 If

+ s + ic)/( 1 + s ic)
2
+ ra 2 +n a =l and m + m =
1+iz
l + im

(l+l)z,

4 If z = r(cos0+isin0) and a = p(cosa+tsina), calculate


ofr, p, 6, a. Deduce that |1 -az\*- |z-a| 2 = (1 -r 2 ) (1 -/> 2 ).
5

If |zi-,|

<

|z 2 |, prove

6 The number z
and radius is 1.
(i)

(ii)

>

i|z 2 |;

l^ + a,!

(ii)

represented by a point on the

>

circle

|z 2 |.

whose centre

is 1

+ 0*

(i)

If

|z

z| =

P P

right -angle
Zl

|af

in terms

Represent the number z 2, and prove (z 2)/z = i tan (argz).


Construct the point representing z 2 and prove that
(a)

is

(i)

a| 2

\z

x,

|z|

and

(6)

arg(z

1)

arg(z 2 )

f arg(z

z).

PP

represent z x z 2 and the line x a is turned through a positive


into the position P-iP^ prove that
3 represents
t

about

+ ^Zjj-Zx).
(ii) Two opposite vertices of a square are 1 + 2i, 3 5i. Find the numbers

of the other vertices.


8

(i)

If Jc

is

a real constant, interpret

zx

+ k(z 2 z x

geometrically.

P1 OP

E,
2 meet PiP 2 at
and
is the mid-point 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

isosceles triangle, the angles at


(8

P P

- z 2) a =

2,

by

the vertices

P P P3
x,

2,

of an

each being %(n a). Prove that

4 (*3 ~ 2i) ( zi ~ z z) sin2 ia

36-2

COMPLEX ALGEBRA

526

10 Interpret the following as loci:


|z

(i)

+ 3*| 2 -|z-3i| 2 =

arg

(iii)

- =

12;

(ii)

*(iv)

Jtt;

|z

+ K| 2 +|z-H| 2 =

|z-z x

|z-z 2

10fc 2

(Jfe

>

0);

1.

The numbers z lf z 2 are connected by the equation z x = z 2 + l/z 2


IfP 2 describes a circle of radius a 4= land centre O, show that P1 describes

11

(i)

the ellipse

%1

(l+a 2
If z 2

(ii)

y2
2

(l-a 2

=
2

^"2

'

r(cos0 + isin0), determine the locus of

when 6

and

\tt

varies.

liu + iv =

12

show that the curves in the ay-plane which


constant are (in general) orthogonal systems

is real,

constant, v

= (w + 2c)j{w 2c) and c is real, prove that when


then w = 2c cos 0. Hence show that if z describes the circle
describes the segment of the a?-axis between 2c, once in each

c(eos 6 + i sin 0),

|z|

If {(z + c)/(z c)} 2

13
z

where a

ajz

correspond to u
of circles.

c,

then

direction.

14 If z, w are represented by P, Q and zw + w z+1 = 0, prove that when


= cos 6 + i sin 6, then \w\ = tan0 according as =S 6 < 71,
If
describes the circle \z\ = 1, prove that Q describes the y-axis, and indicate
corresponding directions of motion.

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

semi-diameters. Prove z 2 +z|

(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.]

real axis. [Clear fractions, equate real

In

the following,

17 If x

a>

= a + b,

and w 2 denote the cube roots of + 1, w 4=


y = a + bco, z = a + b(o 2 prove that
,

= 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 ).

does f(x) +f((ox) +f(o) 2x) represent?

r=l
Calculate similarly f(x)

+ (of((ox) + (o zf(u) 2x) and f(x) + w 2/(waj) + o>/(w 2a;).


(z a), prove that
19 If 2" + p x
+ ... +p n = (z Oil) (z a 2
2
2
2
2
-...)
(l-p
(l + a )(l+a )...(l + a
=
+ (^i-p 8 +P6--) 2
2 +p 4
n
2
[l+p <+i? 2 + ...+p = (1-<M) ... (l-a n ). Put<= i, then multiply.]
20 Find conditions for z 2 + (a + &*) z + (c + a*) = to have a real root, where
*(ii)

zn x

. . .

a, 6, c,

a"

21 If z
Sx* +

are real.

a,

+ fti is a root of z 3 +pz + q =

2px-q =

(p, q real),

22 Find the equation of lowest degree with rational

1,2- ^3, V2 - 1

prove a

real root of

coefficients

which has

is

0.

for roots.

23 Prove 3a;5 5a; 3 + k =


*
if k > 2 or k < 2.

has three real roots

if

2 <

<

2,

and one

real

root

24 Find the condition for

3a;

+ 4pa;3 + s =

to have

no

real roots (p, q real).

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,

using the substitution y

527

6,

has one root; and that for


where b is a certain number

solve the following equations^ in

complex

algebra.

26 12** - 4a;3 - 41a 2 -4a: +12

= 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;

has an infinity of roots. If n

is

a large

show that

and that a

pairs of roots he in the neighbourhood of


closer approximation to these roots is

(2n + 1) n 2 e-+*)#.

[Express th x in terms of e.]


f Since each
equations.

is

unaltered

when x

is

replaced

by

1/x,

they are called reciprocal

528

14

DE MOIVRE'S THEOREM AND


SOME APPLICATIONS
de Moivre's theorem

14.1

14.11 If n is

an

integer (positive or negative), then


(cos

d + i sin d) n

if n is rational, then cos nd

cos nd + i sin nd ;

+ i sin nd is one of the values~\ of


(coad + ismd) 11

Proof,

Let

(i)

nbea positive integer. By direct multiplication,

d + i sin 0) (cos

(cos

<j>

+ i sin

<f>)

6 cos $ sin d sin <f>) + i(sin d cos <p + cos

(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)

applications of the above result. Proceeding step

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

nbea negative integer,


(cos

cos (20)

becomes

cos nd + i sin nd.

say

+ i sin 0) w =

= m. Then

(cos 6 + i sin 0)~m

~~

(cos# + isin0) OT
1

cos md + i&mmd
f-

See the definition in Case

(iii)

below.

+ i sin (20).

DE MOIVRE'S THEOREM

14.12]

by Case
(cos

so

(i),

since

m is a positive integer. Now

md + i sin md) (cos md i sin md) =


,

*
.

529

cos to# + i sin top

cos m# i ammd

=
=

cos

cos 2 md + sin 2 md

ly

m#) + i sin m#) by trigonometry,


cos nd + i sin nd
since w = m.
(cos d + i sin 0) n = cos nd + i sin nd.

Hence again

(iii) 2>e n 6e rational, say w = ^/g where (without loss of generality)


we can suppose that p and q are integers and q > 0. So far no meaning
has been assigned to the expression zplq when 2 is complex. We now

define the values of zplq to

be the roots of the equation a

zp .

Consider

pd

/
I

cos

+*sm
.

2/

cos jo^ + i ainpd

by Case

(i),

(cos d + i sin 0) p

by Case

(i)

is

Hence by the

(ii)

a positive integer,

according as the integer p

positive or negative.

pd
i

sin

14.12

or

is

definition just given,

pd
cos

since q

is

a value of

(cos d + i sin d) plq .

The values of (cos 8 + 1 sin O)^*

Still supposing that p, q are integers and q > 0, let s(cos


+ i sin 0)
be any one of the values of (cos d + i sin d) plQ This means that, on
raising each expression to the qth power,
<fi

+ 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

as in 13.22 (2) that

cos pd + i smpd.

= 1 and q<j> = pd + 2k7r, where k


modulus of a complex number, it

sa

zero. Since s is the

by definition real and positive, so that sa = 1 implies s = + 1.


Taking k = 0, 1, 2,
q 1, we obtain the q expressions
. .

. ,

pd
cos

2hn\

pd

+*sinH- +
.

as possible values of (cos d + i sin d) Plq

2kn\
(1)

These q values are distinct, because angles given by k


differ by 2 \k x k z njq, which is less than 2n since A; x
|

=
|

kx k
,

<

q.

&2

DE MOIVRE'S THEOREM

530

No further

[14.13

values are given by taking other values of k, because

other value of k differs from one of the values

0, 1 , 2,

. .

any

q - 1 by some

integral multiple of q.

Consequently, (cos 6 + i sin 6) vla has exactly q different values, viz.


q- 1. Any other set of q integral
(i) given by k = 0, 1, 2,

those in

values of k could be chosen, provided they give q distinct values of (i)

we often take k = 0, + 1, + 2, ....


The q values (i) are represented in
the Argand diagram by points
e.g.

on the unit

circle

\z\

1:

(Ox,OP1 )=pe/q,
and

arcs

PX P

2,

Pq-iPq

P^Pz,

sub-

tend angles 'Znjq at the centre 0. Thus


PX P2 PZ ... Pa is a regular g-sided poly-

gon inscribed in the unit circle.


Since every complex number

r(cos#-Msin0),

Fi8- 136

can be written in the form

where

tt<6^7t,

the above shows that the values of zplq are

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

spaced around the

circle

\z\

Definition. It will often

rplQ

be convenient to abbreviate cos0 + isin#

to cis0.

14.13 Examples
(i)

Solve the equation (z+

l) n

z".

If the left-hand side were expanded, this would reduce to an equation of


degree n 1 in z. Hence in complex algebra there will be n 1 roots.
Since z = is not a solution, the equation can be written

= 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\

cos|0 +

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,...,n-l).

1 solutions.

Alternatively, instead of quoting the general theory,

stage

(a)

1).

kn\

1 /

These are the n

\
1)

2^\

1,

h*sui

*sin

Jen

1,

kn\
kn

cos =
n
n

kn
cos

Hence by the theorem (with index


kn
n
2*zsui

(
I

h2*sm

2kn\

l.e.y

1+* sin

A , ft
0,l,2,...,n-l).

(k

2kn\

|+*sin

in 14.12,

'

we may proceed from

above by writing
cos 2kn + * sin 2kn',

hence by de Moivre's theorem,

z+1

2kn

cos

which takes distinct values


(ii)

for

Show that the roots ofz n =


zn

From

=
z

which takes

gives

for the case k

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,

Remark. Since the sum of the roots of z n

h*sm

2n\

a",

a2

a n_1

=
1 =
l + a + o^+.-.+a"-

0.

f This step of passing to half-angles

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

in 13.23. Observe also that

a n ~r =

[2{n-r)7i\
| |

cis

{2rn\

>

n
and hence a n_1 =

a,

a n-a = a 2

....

2rn
cos

2rn\

2rn

r
oc ,

Also see Ex. 14(a), no. 21.

If p, q are coprime integers (i.e. have no


q values of (cos 6 + i sin 6) PlQ can be written

*(iii)

the

ism =
n

cis|27T

common factor

other than 1),

cos^(0 + 2for) + *sin^(0 + 2for),


where k

0, 1,2,

...,q

(ii)

1.

These q expressions each satisfy z" = cospd+iainpd, and are therefore


values of (cos0+isin0) 1
They are all distinct; for if
'/ <1

'.

cis

-(d + 2k 1 ir)

cis-(0 + 2fc a 7r),


3

then as in 13.22(2),

P (d

+ 2k1 n)--(d + 2k z n) = 2mn


3

?
for

some integer m,

has a factor in

i.e.

common

k2 < q, this shows that q


p(k x k 2 ) = mq. Since
with p, which is impossible because p, q are coprime.
\

Remarks

cispd has exactly q roots, the set of values of (ii)


(i), in some order.
1 /9
is understood to
(/?) If p, q are not coprime, the function (cos0 + *sin0)
mean {(cos0 + isin0) } 1/<r , and therefore has q distinct values, viz. the values
1
(i) of (coapd + i smpd) !". However, in this case the expression cis {(p/q)(6 + 2kn)}
does not take q distinct values, and consequently, does not represent all the
values of (cos0 + isin0) ,'A'. See Ex. 14(a), no. 8.
(a) Since

must be

the

the equation z 9

same as

the set of values of

'

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,

2 cos (md nd)).


r/

+ cos0 + *sin0) 8
.. ,,
(cosf *sin/)*4
=

cis

prove

(inn no), n

bemg an mteger.

= cis0, prove (x 2n l)/(x* n + 1) = *tann0, where w is an integer.


Find the modulus and the principal value of the argument of ( 1 + i V3) 8
6 Find the three cube roots of 2i 2.
7 Obtain all the values of {(J2 + 1+ i)/( *J2 + 1 *)}* in the form a + bi.

4 If a;
5

DE moivre's theorem
down

8 Write

+ i sin 0)*.

(cos

+ i sin 0)i and the two values of


+ 2ifor)} has only two distinct values, and

the six values of (cos

Observe that

cis {f(0

therefore cannot completely represent the

cos 3 0-3cos0sin 2

function.

first

By

9 Simplify (cos0 + isin0) 8 in two ways.


parts, deduce that

cos30

533

and

sin 30

equating real and imaginary

3 cos 8

sin 6 - sin 8 0,

and express tan 30 in terms of tan 0.

10 If z

cis 0,

express in terms of 0:
1

(i)

\;

z+ -;

(ii)

(iii)

z-\;

z" + ^;

(iv)

z--~.

(v)

Solve completely the following equations.


z"

11

= -l.

13

(z-l)"

15

z 2n

12 (z+l) n + (z-l) n

= z.

14 (1+z)"

0.

(l-z) B .

2z n cos ncc + 1 = 0. [First solve as a quadratic in z".]


By expanding 1 + z) n where n is a positive integer, and putting z =

16

prove

cis 0,

\nd

(2 cos 0)" cis

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*" cos Jw7T

. . .

2* B sin Jtmt.

= a + a 2 +a 4 7 = a 3 + a 5 + a 8

/?

Write down the quadratic having roots

2n

fi,

7,

19
cos

If

cos0 + cos0 + cos^

4=n

87r

prove /ff+y

'

sin0 + sin0 + sin^

and

30+ cos 3^ + cos 3^ 3008(0 + ^ + ^) =

and a

= 1 and

and deduce that

+sm + sm = +JVv 7
sm
7
7
7
.

2* n+1 cos Jn7r. Writing

+c1 x+c 2 x 2 +...+c n x n

prove

/?y

2 n ~* I cos

2w

r=
17 If w

S n Cr cos =

Deduce that

0,

prove

that

similar result for sines.

= cis 0, y = cis 0, z = cis ^ and use Ex. 10 (/), no. 4 (iii).]


What conditions have to be satisfied by z in order that the

[Put x

20

senting

all integral

powers of z should

and (ii) be finite in number? Mark in


number z such that there are only three
z,z a , z 8 ,

points repre-

on a circle with centre the origin,


a diagram the points which represent a
(i) lie

distinct points given

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

Use of the binomial theorem

14.21 cosm 8 sinB 6 in terms of multiple angles (m, n being positive


integers or zero)

This transformation

sometimes needed for the integration of


= cos 6 + i sin 6, then

is

circular functions: see 4.82. Writing z

cos0 isin#,

2cos0 = z+-,

so

Also zn

cosw# + isinn0 and


zn

2*sin0

of

l/zn

(i)

cosw0 iainnd, so that

zn

1
-
sin nd.
n =2i

+1 = 2 cos n6,

(ii)

By means

and the binomial theorem, the given function can


powers of z and 1/z. By (ii) the resulting expansion

(i)

be expanded in
can be expressed in multiple angles.

Examples
(i)

Express cos8

From

in terms of circular functions of multiple angles.

(i),

(2cos0) 6

= (z+^y
=
=

=
.-.

(ii)

*4) +6 (*<+ ^ +15 (* 4) +2


s

+ 15(2 cos 20) + 20 by


= -^(cos 6(9 + 6 cos 40 + 15 cos 20 + 10).

^z +

=
cos

sin4

(ii).

-^ (z-^j*

= (*+!)

sin4 6 in terms of multiple angles.

(2cos0) 5 (2isin0) 4

.*.

2 cos 60 + 6(2 cos 40)

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 70-4 cos 50-4 cos 30 + 6 cos 0).

de moivre's theorem

14.22J

535

We

observe that coan d can always be expanded in terms of cosines


because it depends on the expansion of (z+l/z) n
Since sinm
.

depends on {z\jz) m which

involves terms like zr +ljzr

even,

and terms

like z

\\z

if

odd, therefore

is

expressed in terms of cosines or sines according as

14.22 cos

an

/18,

sin

#18,

is

can be

m is even or odd.

tan /18 as powers of circular functions (n being

cos n& + i sin nO =


cn

if

integer)

We now reverse the process in


=

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~^ -

c^s - (^j c"" 3*3 +

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 right-hand side

by

cn ,

t*nn0 =
where

we obtain

U/.
1-t

W
WW

tan#.

";

(v)

[%'-.

The formulae (iii), (iv) can be transformed by means of the relation


1 to show that cosw-0 can be expressed entirely in powers of
cos/9, and so on; see for example Ex. 14 (6), no. 20, and ex. (iii) below.

s2

+ c2 =

Examples
(i)

Express cos 56 and sin 60/sin 6 in terms of cos 6.


cos 56

sin 66

c8

=
=

c8 - 10c8 ( 1

=
=

cs*

- c2 + 5c( 1 - c2
16c8 - 20c8 + 5c.
c 8*

c8s 8

2
)

cs 8

s{6c8 - 20c8s 2 + 6cs 4}


{6c8 - 20c8 (l

sin 60
/.

c 8* 2

sine'

32c

-c 2 + 6c(l -c 2
)

- 32c + 6c.
8

2
)

},

DE MOIVRE'S THEOREM

536

[14.22

Write down the formula for tan 50. What equation


tan 50 = 0; (6) 56 = Jct? Solve each of these equations.

(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, tan-7r, 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

a 5 -5a:*-10:r8 +10a; 2 + 5a;-l

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

and this takes distinct values for r =

The required roots are therefore


1'

'

2 ' 3 ' 4 )-

in (6) could also be solved algebraically

by the method

in

Ex. 13 (g), nos. 26-28.


*(iii)

Ifnis

odd, prove that sinn0 can be expressed in the form

sinn0

and find blt

61 s + 6 8 s 8 +

...

+ b n sn
= (r 2 n 2
,

b n . Also prove that (r + 1 ) (r + 2) 6r+a


(iv) and the relation c 2 = 1 s 2 ,

br ,

and hence find

68 .

From formula
sinw0

=
(^j

(l-s 2 )K-i) s - f^j (l_ s 2)i(-3) g 3 + f^j (l- a *)H-*) 8 *- ....

Since each of n 1, n 3, ... is even, this can be expanded as a polynomial in s


involving only odd powers of s and having degree n, for
b

coefficient of s n

= ~ I)*"""
(

" " I)*""*


(

- _ i)fcn-i )2 n-i by
(

(j)

+ - D*"-
<

Ex. 12 (6), no.

2.

DE MOIVRE'S THEOREM

14.23]

Also
&!

537

/n\

= II =

coefficient of a

n.

By deriving the identity ahind =

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

r-2 c 2

- r&r *r-1*]

= S [r(r - 1) 6r sr " a - {r(r- 1) br + rbr} a']


= X [r(r - 1) 6r sr - 2 - r a6r sr];
n

r=l

r=l

- n 2 S 6r *r = S[r(r-l) &r sr_a - r 26r ar].

hence

Equating

coefficients of r,

- n26r =

(r

from which the required relation


2.3&8
14.23 tan (e t

cos(0 1 +

(cos 6X

+ 62 +

+...

+ 2) (r + 1

follows.

= (l-n2 )n,

. . .

&r+a - r 26r ,

Taking r
68

i.e.

1,

we have

= Jn(l-w2 ).

+ 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 +...),

2r denotes the sum

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

cos<9 1 ...cos^n (l-S 2

+ S 4 -...),

cos tfj... cos tfJSi-

(vi)

and

By

The formulae

n)

(vii)

+...).

division,

tan(01 +

. . .

+ ...+0J =

(vi)-(viii)

dn

6. All

include

^~^^ S~;;
5

(iii)-(v)

as the special cases

these can be written

down

when

any
would cease

easily for

particular value of n, the terms continuing until they

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.

sin4 0cos6 0d0.


o

9 Express cos 60 in terms of (i) cos 0;

(ii)

sin 0.

10 Express sin 70 in terms of sin 0.

Express cos 70/cos

11

in terms of sin 0.

12 Solve completely the equation cos 50 + 5 cos

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.

+ cos 70)/( 1 + cos 0) =

(x

Hence prove

0, 1, 2.

-x ~2x- l) 2
2

[Express

in half-angles.]

*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

down the expression for tan 30 in terms of t = tan 0. By first solving

are tan ^77-, tan -^n and


roots of x3 3x 2 Sx+ 1 =
What are the roots of x 2 4x + 1 = 0?
18 (i) If a + 2 + 3 + 4 = mr, where n is an integer, state the relation between
h, t2 t3 4 where tr = tan r
tan04 are the roots of ti + at3 + bt 2 + ct + d = 0, express
(ii) If tan0 1}
tan (0X + 2 + 3 + 4 in terms of the coefficients. What can be said about the
angles if c = at
19 If tana, tan/?, tany are the roots of ax 3 + x 2 + bx + 1 = 0, prove that

tan0

1,

show that the

tan |7T.

a + /?

is

an

integral multiple of tt.

(i) Prove that cosw0 can be expressed as a polynomial of degree n in


~
~
the last term being
cos0 of the form ac n + a n_ i c n 2 + ...+a n_ Sr c n 2r +
(ii)
Prove
that a n = 2 n_1 (iii) If n is even,
a or ax c according as n is even or odd.
put = \n to show that a = ( l)* n if w is odd, consider lim (cosn0/cos 0) to

*20

prove a 1 = n( l)^"- 1). (iv) By deriving twice wo and equating coefficients


of cr , prove (r+ 1) {r + 2)ar+2 = (r 2 n 2 )ar (v) Hence expand cosn0 (a) in
descending powers of c; (6) in ascending powers of c, first for n even and then
.

for

odd.

14.3

We

Factorisation

now

Theorem

give further applications of the results in 10.12 and of

I in 10.13, which remain valid in complex algebra. Just as

DE MOIVRE'S THEOREM

14.31]

539

we have used 2/(r)

to denote /(l)+/(2)

+ ... +f(n),

here

it is

con-

venient to write

n/(r)-/(l)/(2).../(n).
r=l

As with the E-notation, the range of values of r can be omitted when


from the context.

clear

x-l

14.31

We

have xn

if

xn

cis2?rr, i.e. if

cis (2r7r/w).

For

n it follows that x cis {2m jn) are distinct factors of xn 1


1 2,
Since r = k and r = n k give the same value to cis(2nr/%), we
may use r = 0, 1, 2, instead of r = n, n 1 n 2,
This gives
Case (i) : n even. Take r = 0, 1, 2,
1),

. .

. . .

\n
to

pairs of factors, together with the single factors corresponding

and \n; in all, there are 2{\n 1) + 2 = n


The factors corresponding to r = k are
/

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
\

The factors corresponding to


Hence
for

2kn\

is

x cos

h*sm

factors.

xn

(x

0,

hi.

|w are respectively xl,x +

1) (# + 1)

* n_1 /

# 2 2a; cos

2kn

+ 1\

l.

(i)

(w 1); this gives


n odd. Take r = 0, +1, 2,

=
1 + 2{%{n
n factors altogether. Arguing as before, we find that
1)}
Case

():

*(n-l)/

fornodd,

14.32

xn +

xn -l

(x-l)

[x*-2xcos

2kn

+ 1).

(ii)

if xn = 1 = cis (2r 1) n, i.e. if a; = cis {(2r 1) n/n},


xn + 1 =
where r is an integer or zero. Distinct values of x are given by any n

consecutive values of r.
37

GPMII

de moivre's theorem

540
Case

(i) :

As

even.

to pair off conjugate complex

we wish

in 14.31,

[14.33

Now the factors

linear factors so as to get a real quadratic product.

X CIS
.

are conjugate if (2r-

with the values

we

2r-l
71,
n

r'

+ (2r'l) =

1)

r=l,

pair

X CIS

2k-l

X ICOS

and their product


for

+ *Sin

- 1, -2,

is

2k-l
n

- {%n- 1).

X |COS

2k-l

xn +

in l

%
II \x

njn}

(ii)

r=l,
with

r'

factor x +

1.

Then

for

\(n +

odd,

factors

2k-l
n

\
7T\.

+ 1 Hence
.

2xco&

2k

tt+

\
1

(iii)

1 .

We pair off the values

odd.

The value

l&m
.

TT

&=i\

Case

The

altogether.

\
7T\,

x* 2x cos {(2k

even,

therefore

+ (\n 1)} = n factors


= k, r' = (k 1) are
.

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(n-l)

2,

-1, -2,

gives 2r

1)

as before

xn +l

-%(n-3).

n, to

which corresponds the

real

we have

= (x+l)

x^-lx" cos na + 1
The equation x 2n 2xn cos

Hn-l)/
2k -1
\x 2 2x cos

H
k=i

n+l).

(iv)

14.33

(x n

+1 =

is

quadratic in x n

cos na) 2 = 1 + cos 2 not = sin 2 na,


xn

.'.

Hence the 2n values of x which

cos na

satisfy it are

2rn\

i sin na.

2rn\

cos

where

0, 1,2,

...,n

integral multiple of tt.

1;

these are distinct unless

na

is

zero or an

de moivre's theorem

14.33]

541

The product of the conjugate factors corresponding


[X

cos

2kn\

x*

rt

2k7T\)(

-\

n-i

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\W-1xcos

(v).

+l},
)

equal to
in -II

(x-l) 2 (x+l) z

W~2xcos

fc-i

2kn

\*

+ 1}
)

because the values k = 0, k = \n give the factors {x l) 2 (x+ 1) 2 respectively,


while factors corresponding to k and n k are equal. Taking square roots (in
which the positive sign is chosen since in real algebra all the quadratic factors
are positive and x n 1, x % 1 always have the same sign when n is even), we
obtain formula (i) of 14.31. If n is odd, we likewise deduce formula (ii).
Similarly, by putting a = n/n in (v), we obtain formulae (iii), (iv).
,

(ii)

Divide both sides of (v) by x n

xB H
Putting x

2cosraa

*n

-M
TT

n
+
2cos|a

{a?H

*-ol

2hn\\
1}.

J)

cis
is0,
8,

2cosn0 2cosna

(2cos0 2cos (oH

fc=ol

n-i
cos nd cos net

\
/

H {cos0 cos laH


ft=ol

2" -1

2kn\\

(iii)

In

(v)

put x

and a =

sin 2 w/?

n 1

J] 2|l-cos(2y6f +

n1

2 2 "- 2 II sin 2
o

)}.

J)

= 2 n - 1 *n
o

sin

2kn\\
-jl,

klt\

ft

\
1

sinn/?

2/?:

2(l-cos2n/0

.'.

)},

})

(V+ \
nf
\

37-2

DE MOIVRE'S THEOREM

542

[14.34

To decide the sign of the right-hand 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
\

taking logarithms of the modulus of each side of this last result,


/? is not zero or an integral multiple of 2n,

assuming that

nl
loglsinwytfl

fc=0

nl

wo

Deriving

(v) de

ncotnfi

S cot IAH
fe=0
\

= (n-1) log 2+ Slog

jfc7j-\

sin(^ +

/\

lvn\

nj

Moivre's and Cotes's properties^

of the circle.

A A X A 2 ... A n_ 1 is a regular n-sided


polygon inscribed in the circle of centre
O and radius a; P is a point such that
OP = x and /.(OP, OA ) = 0.
Since sides of the polygon subtend angle
2n/n at O, we have
L(OP,OA r =
)

Also,

by the

6+

2m /n.

cosine rule,
Fig. 137

PA* = x 2 + a*-2axcos[d
(" +

v)n-ll

PAl.PA\...PAl_ 1 =

=
by a

slight extension of

formula

(v).

2
II |a -2aa:cos|0 +

j+a
2rn\

x 2n - 2a nx n cos nd + a 2n

Hence de Moivre's property:

PA PA ...PA n_ = V( " - 2a nx n cos nd + a 2n


In particular, if P lies on OA Q then 6 = and so
PA .PA ...PA n_ = \x n -a n
A
If OP bisects A n_ OA^ then 6 = njn, and so
2

).

\.

PA^.PA x ...PA n_x = x n + a n


These

last

two

results are

14.34 sin nG, n

is

as Cotes's properties.

odd

of 14.22, smnd can be expressed as a polynomial in sin 6 when n


odd; the leading term is b n sin" 6, where b n = ( l)**"-" 2"- 1 The polynomial
zero when sinnfl = 0, i.e. when nd = rn, 6 = rnjn and so sin0 = sin (rnjn).

By ex.

is

known

(iii)

f These illustrate geometrically the results of 14.31-14.33.

DE MOIVRE'S THEOREM

14.34]

Hence

for r

0, 1, 2, ...,n

1,

543

sin 6 sin (rn/n) are distinct factors of sinntf;

consequently

n _!
( sin 6 sin
&
,

sin n#

^
r=l\

w/

(vi)

I.

This can be expressed alternatively as follows. The


sin (m/n) are given

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 #

Divide each factor following sin# in

where
sin 6

(vii)

(vii)

I.

sin 2 (9

~ am? (kn/n))'

foA

by sin 2 (kn/n); then

^D/

A sin/

sin nd

sin 2

(sin 2

JJ
fc=l

independent of 6 and can be determined by dividing both sides by


letting 6 -> 0. We obtain

J. is

and then

A=
.

sinn#

= n,

lim
d^O sin d
;

i(n-l)

and so

sinw#

= nsin0

TT
1

fc-L !

am 2$

r-tt^ 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

identifying the product

i(n-l)

ns

n
ft=l

- s 2 cosec 2 =
n/
]c7t\

ns + %n(l

-w2 )s3 +

...

+& B s B

Equating coefficients of the various powers of s gives identities involving


cosec 2 (Jen /n); e.g. from those of s 3 ,

cosec 2

2j
ft=l

K-D

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

= l + cot 2 < l+^-a

= m/n where n is odd, and sum for r =

1, 2,

\(n

1):

de moivre's theorem

544
Writing

and using formula

(ix),

we have
7T

<
Since the series S(l//" 2 )

some

to

limit s

1)

sequence

for

n-l
< ^ "J" + **-
7r

12.41 (3), therefore sm tends


any manner, and in particular through the
Letting n -> oo, the above inequality gives

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

cos \n + cos |n + cos f7T

and
s

real quadratic factors of x

2 Find the

is (x 7

for x and

2/3

n-l

5 sinn/?

= -

6 cosn/?

= (-l)*2 n ~ 1

2 n_1

l)* n

in formula (v) of 14.33, prove

2~ 1 IIsinj^ + r=0
I

2w

n1

].
)

rjjX

JJ cos

-i

>

if

even.

is

(2r+l)n\

]J cos{,ff+-

if

is

7 Obtain a result by deriving no. 4 logarithmically

n-l

2 cosec

/ =

/
2

r7T \

/? H

r=0

9 Prove ch nx cos na

by

n
o;

and then put x =

ev

n_1 f

[Divide formula ( v)

/J

.]

sum of n partial fractions.

[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/?.

10 Express # n_1/(# 2n 2# n cos noc + 1) as a


(v) logarithmically wo a.]
11

^.

(2r+l)n\

4 cosw/?

or 2/3+n/n for

the following.

8 Prove

4a; 4

l)/{x 1).]

By substituting

cos \it cos %n cos ftf

+16.
quadratic factors of x6 + x5 + xi + x 3 + x i + x+l. [The

Write down real

expression

2cos

(iii), (i)

By putting x =

2n

/2r+l

\)

n\),

xn

J)

xn

with 2n instead of w.]


in the results of no.

cis

n-l/
(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

2 n_1 sin 6 JJ ( cos 6 cos


r=l \

\
)

77A

nj

n
,

de moivre's theorem

14.4]
13

Obtain a result from no. 12 (i) by putting = 0.


dividing both sides of no. 12 (ii) by sin0 and then letting

(i)

By

(ii)

Prove

n-i

Vn = +2"- 1

II sin

r=l
14

545

By deriving no.

(i)

12

logarithmically, prove

(i)

nt&nnd _ n ~ 1

2j

r=: QCOsd coa{(2r+\)7il2n}'

sin0

What result is obtainable similarly from no.

(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"

Show that the solutions

16

= i tan

|sin w0|

of

+ x) %n +

2r-l

Deduce that

(l

+ a;) 2n + (l-a;) 2n =

and hence prove that

" 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\

*17 Prove that

cosn0 cosraa

2r7r\\
/
2 n_1 JJ <cos0 cos |a + }>

r=ol

by using Ex. 14 (6),

14.4

no. 20

and arguing as

/;

in 14.34.

Roots of equations

14.41 Construction of equations with roots given trigonometrically


(i)

Form the equation whose roots are cos fn,

= %n,
equation
equation

(ii)

%Tt,
3
4a;

is

tt

all satisfy

3a; = is

satisfied

therefore

Form the

by

a;

cos fn, cos n.


Since cos 30 = 4 cos 3
cos \tj, cos fn, cos f

+1 =

= cos 30.
40

m is any integer or zero,

It

3 cos 0,
-

the

The required

equation whose roots are cosfTT, cos

Consider the equation cos 40

where

cos 30

is satisfied

cos f7T.

by

= 2mn 30,

i.e.

= \mn

by

(this includes

both solutions

obtained from the double sign ).


Writing x = cos 0, the equation becomes
8a;

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

this are given

by

^ ^ ^

CQS Q

the required equa-

de moivre's theorem

546
(iii)

Form

the equation

The equation tan


zero.

Writing

whose roots are tan \n, tan tt, tan

W=

tan 6,

Removing the root t =

becomes

it

we

0,

it

formula

is

tan

any

%tt.

integer or

(v) of 14.22)

see that the equation

tan {\mr), where n

(see

\tt,

- 36*6 + 126*4 - 84< 2 + 9 =

+ 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

*4 - 36# 3 + 126a 2 - 84a; + 9

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.

obtained by using relations between roots and

14.42 Results

coefficients

If the results of 13.51 are applied to equations like those found in


14.41, various trigonometrical relations are obtained.

Examples
(iv)

Prove sec ftt + sec %tt + sec fzr

6,

and find sec \tt + sec ftt + sec f7r.

The equation whose


will

have roots sec ftt,

sum of its roots is

roots are the reciprocals of the roots of that in ex. (i)


sec %tt, sec tt. This equation is x 3 6x 2 + 8 = 0, and the

6.

Since sec \tt sec %n,


sec \tt + sec tt + sec %n =
(v) Calculate

From ex.

(ii),

sec %tt

2/

=x

the equation having roots sec ftt, sec

2
,

+ 4x -4x-8 =

i.e.

2/

f7r,

= sec ftt,

= y 1,

i.e. 2/

- 24t/

= z + 1,
23

which has roots sec |7r

-4)

16(2/- 2)

becomes

-21z 2 + 35z-7 =

tan 2 f7r,

etc.

is

is

2
,

+ 80^/ - 64 =

this

sec far

0.

the equation having roots sec 2 f7r, etc.


2/(2/

Writing

sec ftt

tan 2 ftt + tan2 ftt + tan 2 f7T.

x
Putting

= sec %tt, and

6.

0.

(after reduction)
0,

The sum of the roots

is

21.

hence

DE MOIVRE'S THEOREM

14.42]
(vi)

Prove that cos 3a

547

4 cos a cos (a + tt) cos (a + f7r).

Consider the equation cos 30 = cos 3a, which


where k is any integer or zero. It can be written

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 the equation whose roots are

cos \tt, cos \tt, cos ^n.

Form

the equation whose roots are cos \n, cos ftt.


[Consider cos 56 = 1.]

3 By considering cos 30 = cos 26, construct the equation


gos%tt and cos7r. Hence find cos 36, cos 72 in surd form.

4 Show that x

2m
2 cos

-yy

(r

1, 2,

. .

.,

5)

are the roots of

x6 + xi -4x 3 -3x 2 + 3x+l


[Consider cos 66

cos 56; remove the root x

whose roots are

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

the equations obtained in exs. (i)-(iii) of 14.41 to prove the following.

= 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

Deduce that YJ cot a H


I

r=0

a;B
{

0, 1, 2,

n-1
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

Finite trigonometric series:

The method of summing


illustrated in the following

certain trigonometric series which is


examples gives results in pairs, and is also

applicable to infinite series (see 14.66, ex.

14.51 Cosines

and

(i)).

sines of angles in a.p.

Write

C=

+ cos (a + 2fi) +

cos a + cos (a +fi)

and
8 = sina + sin(a+/?) + sin(a + 2/?) +
Then

G + iS =

cisa + cis (a + /?)

cis

a + cis a

cis

a{l

on summing the
multiple of 2n.

cis /?

. . .

...

+ cos {a + (n 1) /?}

+ sin{a + (n 1)/?}.

+ cis (a + 2/?) + ... + cis{a + (w- l)fi}


+ cis a(cis /?) 2 +

. . .

+ cis a(cis

- (cis^) w}/{l - cisytf}


g.p.,

provided

cis/? 4= 1, i.e. J3 is

not an integral

Now|

l-(cis^) n

1-cisn/?

= 1 cos nfi i sin nfi


=

2 sin 2 \nfi 2i sin \nfi cos \nfi

= 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

\% cos {a + $(11 -1)0}, # =


sinf/?
1

^^ cisla + ln-l

sin \nB
.

+ i sin

sin InB

f
+ |^
1
\ C sin{a
sinf/?
.

, x

zero or a multiple of 2n, then from the original series clearly

G = wcosa and S

wsina. Cf. 12.27, exs.

(i), (ii).

t The following reduction is shorter than multiplying numerator and denominator


by the conjugate of 1 cis /S to make the denominator real.

DE MOIVRE'S THEOREM

14.52]

549

14.52 Other examples

Sum 1 +

(i)

a;

cos 0+ a; 2 cos 20 +

Write

G-

terms, where

x is real.

l+a;cos0 + a: 2 eos20+...+a; n - 1 cos(n-l)0

andf

S=

Then

C + iS = 1 +
=

ton

...

a;

sin

a;

cis

+
+

a;

a:

sin

cis

20+ ... +a: n - 1 sin(ra- 1) 0.

20 + ...+ a:"- 1 cis (w- 1)0

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

05" cis n0) {l-a cis (-0)}

(l-a:cis0){l-a;cis(-0)}
to
1

since cis 0, cis

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

Prove that sin a +

(ii)

real part of this result,

a;" cos n0 + a3 n+1 cos (n 1)0


1 2a; cos + x 2
when w - oo, and the series has a sum to infinity
^
(7

103 008

w _kjo
If

the denominator real,

n+1 cis(n 1)0


0) +c

_ 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

cisa|l

^ cis^+^

We do not write S = 1 + x sin 6+


1

cos (a +

2/?)

sin(a + 2ytf)

cis (a

cis

+ sin (a + nf})

. . .

+ ^/?).

+ cos (a + n/?)

+ ... + sin(a + W/fl).

+ 2/?) +

a (cis /?) 2 +

(cis/?) 2

(2cos/ff) n sin(a

. . .

...

+ cis (a + w/?)
+ cisa(cis/?) n

+ ... + (cis/?)"J

... because at the next step


+*, which does not appear as a factor in succeeding terms.

C+iS would start

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 /?)"

cos \fl cis /?)"

n
(2 cos \fi) cis (a
(2 cos J/?)" sin (a

+ in/?).
+

on equating imaginary

parts.

Exercise 14(e)

Sumsina sin(a+/?) + sin(a + 2/?) sin(a+ 3/?) + ...

2 Prove

+ cos0 sec0 + cos20 sec 2 + ... +cosn0 sec"0 =

Calculate Y.

to 2w terms.

S1X1 { Th

sin

X)

-J~

cos"

sin 2r0.

r-lW

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) af-1

If a

+ 3 cos a + 5 cos 2a + + (2n 1) cos (n 1) a = n.


and
3 sin a + 5 sin 2a + ... + (2n 1) sin (n 1) a = n cot a.
6 Find the sum to n terms of sinJ. + (6/c)sin2J. + (6 2 /c 2 )sin3^. + ....
b < c in the triangle .ABC, deduce that the sum to infinity is (c/a) sin G.
1

7 If

is

not zero or an integral multiple of n, find the

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

Further examples appear in Exs. 14

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).

of complex terms. Some single-valued funccomplex variable

Infinite series

tions of a

14.61 Convergence

and absolute convergence

Limit of a complex function of n. If sn = o~n + irn and if o*n -> o~


and rn ->T when n -> oo, we say that sn -xr + *t when n->co and write
(1)

lim 5w
n> co

o"

+ *r.

This definition of limit is consistent with that in 2 7 1 for real functions of n


such that,
given a positive number e however small, there is a number
'

'

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"n-o-|

we have

<r B ->

(and

cr

For the

(2)

<

+ i(T n -T)\ <

is satisfied

|(o--o")

|<r n

+ |t-t| <

-<r]

for all

5s

N, then

2zr where
,

= 2

zr

n
zr

+ *(t-t)| <e,

a;,.

+ iyr and a;r yr are real,


,

= 2 r + * 2 2/r =

sav

+ * T>

"

r=l

r=l

r=l

By (1), 5n tends to a limit if and only if <rn

rn both tend to limits

n -> oo. Accordingly we give the following definitions.


2zr is convergent if 2#r and Syr both converge.
If YiXr Y,yr converge to sums X, Y, then Z = X + iY
,

infinity of

2zr

is

the

when

sum

to

(3) If the series (of real positive

2zr is

e.

since

similarly r n -> r).

series

Sn

-<r)

551

terms)

|zr

converges,
j

we say that

absolutely convergent (a.c).

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

and consequently (by 12.52) also converges. Similarly


converges, and hence by definition 2zr converges.

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

1-z

l-z

r(cos0 + *sin#), where r


zn

Ifr

(just as for

0,

we have

rn (cosnd + iainnd).

< l,thenlimrn = 0;andsince

therefore both rn cos nd

>

|r

n cosw#|

rTC

and

|r

n sin?&0|

and rn sin w# tend to zero when n ->

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

when n^-oo, and again

= 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

(1-z)- 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,

and its sum

1).

when the series of moduli


when and only when

the case

Hence the ranges of convergence and of absolute

convergence are the same.

14.63

The exponential

series

Z2

Write

(1)

r(cos 6

+ i sin 6). Then since the series


1

converges for

all r,

23

+ l! + 2! + 3! + -

+ Ti + 2! + 3! + -

the series
z

z2

zz

+ T! + 2! + 3! + -

for all complex numbers z, and hence converges for all z.


Denoting the sum-function of series (i) by expz, or by exp (z) when

is a.c.

necessary,

we have

(z
+n+
1

1!

Remark. The notation

is

real series

whose sum-function
in printing.
series

(i)

having

+ --- + n\)
2!
iTi

We

ex is

(ii)

occasionally written expo; for convenience

usually avoid writing ez for the

many

z.

+ n + 2! + 3! + -

because later work (not in this book)

infinitely

for all

chosen because of the analogy with the

zn \

z2

sum

to infinity of

shows that ez is a function

values for each non-real value of

z,

and

it is

sum of a convergent series, which


However, when the index notation eB is used,

clearly undesirable to use this for the


is

necessarily unique.

it is

Case

method of
more suggestive than exp (iy).

to be interpreted as expz; thus in the symbolic


(iii),

the notation eiv

is

5.43,

de moivre's theorem

14.63J
(2)

>

By

using the definition

now be shown

can

it

(ii),

553
that expz

obeys the functional law

= 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

rt-w^fi

n\n\

the terms of total degree *

' :

(s-2)!2!

+ J1

. . .

2!n!

We now add up these terms 'by diagonals


(s-l)!l!

,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

for these are precisely the

2
p,qP '"'

for all p and q for which p + q > n, p ^ n, q ^ n;


terms not included in the preceding summation by

is

diagonals.

In exactly the same way we have


pPpQ
s n (r2 )

We know

that,

when n

->

oo,

* n (r 1

sjfi) -> e s s n (r2 ) -> e

Hence the left-hand side of (v) tends to

From

+ r2 =

and

zero, so also y\

s n (r1 + r2 )

- ->

v)

-+ eri+r*.

0.

(iv),

K( 2 l)*n( z a)-*n(2l + 2 2)| =

12

^ N'W' ^gl!^
But when w

- oo, n (2 x )

hence the result

(iii)

when

n->co.

->exp2 1 s B(2 2 ) ->exp2 2 and s n (z 1 + z i ) -> exp(z1 + z 2 );


any two complex numbers z x 2 2

follows for

DE MOIVRE'S THEOREM

554
14.64

By

[14.64

The modulus-argument form of exp z


expz

(iii),

exp{x + iy)

= exp x exp (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/

the brackets are finite

3
,

Since

series.

by

12.61 (2)

cost/

"3! + ^"'

y
+ ^-...
=1-1
2/

y
y
amy = y __+|__...

and
for all (real) values of y,

we have by letting n-+co that

exp {iy)

lim sn {iy)

n> oo

=
expz =

Therefore

ex (cosy + i sin 3;).

This result shows that exp z

exp

(z

cosy + isiny.

is periodic,

(vi)

with period^ 2m. For

+ 2m) = exp {x + (y + 2n) i}

ex {cos (y

ex (cos

+ 2tt) + i sin (y + 27r)}

y + i sin y)

exp z.

14.65 Euler's exponential forms for sine, cosine

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

{exp {iy) - exp - iy)},


^
2i
(

(vii)

where we observe the factor l/2, not


These formulae, often written
cos y

\{eiy + e-iv ),

1/2, in

sin y

the last result.

^ (eiv - eriy

),

known as Euler's exponential forms. However, they are equivalent

are

to the infinite series expansions of cosy, siny stated above.

14.66 Examples
(i)

Find the sum to

infinity of

l+cos0 tan + ^ cos 20 tan 2 6 + ^ cos 30 tan3 0+....


Write

tan 0,
t

C = l+cos0+ cos 20 + cos30 + ...,


2

S=

and

(ii)

if

+ +
2 z

3 3

where

cis 0,

= exp {tz)

for all z

and

exp (sin

+ i sin

tan 0)

e sln0 cis(sin0 tan0).

G=

....

e sln0 cos (sin

all real

tan 0).

Trial exponentials: the case of failure. In 5.33(2) we remarked that the


solving y" + ay' + by =
by the substitution y emx will fail

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

has no (real) roots, i.e. if a 2


complex algebra the roots are m

application of 5.32

(2), II,

< 46. Putting


= \a ip, so

the general solution would be

e -Ux e ipx + ft g-ipas)


_ e -iax [A(cospx + * sin pa;) + B(cospx i sinpx)}
e -tax {A' cospx + B' siapx},
where A' = A + B, B' = i(A B). This solution is identical in form

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

byC + iS. From formula (vi) we have (after a slight change of


e<*

= cos* +i sins.
GPMII

de moivre's theorem

556

and we

If u, v are derivable functions of x,

[14.66

define the derivative of

u + iv to

be u' + iv', then


d

{coax + iaiax) = sinx + icosx = i(coax + ismx),


dx
d

= 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

_f. I gin fa)

(for real

e ax cog

a cos fa _ & gin fa)

constants

a, 6)

fa _j_ 1 g ax 8m faf

+ i eax (a sin bx + b cos 6a;)

= eax {(a + ib) cos bx + *(a + ib) sin &#}


= (a + ib)4a+ii)x
law d{emx )\dx = roe"M! already known from 4.41 (5) for all
also holds for complex ones. Equivalently, we may write
.

Thus the

stants m,

e(a+ib)xdx

where

c is

a+ib

e(o+t6)a!

real con-

+ C)

a complex arbitrary constant.

This result permits rapid calculation of certain integrals involving combinations of exponential and circular functions. For example, put

C = j eax cosbxdx, S = jeax sinbxdx;


C + iS = jeax (cos bx + i sin bx) dx

then

e(.a+iU*dx

a + ib

a ib
a* + b

e ax (cos

e(a+ib)x

+C

bx + i sin bx) + c.

Equating real and imaginary parts,


a cos bx + b sin bx ax

Z
e
+c
G-

Ex.

Cf.

L,

a sin bx b cos bx ax_

S=
e + c2

4(e), nos. 28, 29.

A similar method, combined with integration by parts, will apply to


x n eax
cf.

Ex. 4 (m), no.

cos,

sm

fadx;

35.

Exercise 14(f)
Simplify the following.
1

4 exp(*) + exp( t).

exp {(a

6 exp {(a +

7 Expand

3 exp(logr+i0).
+ i7ra).
exp {cis 6} + exp {cis 6)}.

2 exp(l

exp(?ri).

008

a sin ( sin a) in ascending

powers of x. [Use no.

6.]

DE MOIVRE'S THEOREM

14.67]

557

Find the sum to infinity of the following series.


8 ccsm0 + (x 2 /2!)sm20 + (a:3 /3!)sin30-|-....
9 cos a - cos fi cos (a {!) + ( 1/2 !) cos 2 /? cos (a + 2/?) - ...

S see'/? sin r/?.

10

eos0 + cos30 + cos50+....

11

5!

3!

r=lT"

i^.

12

r=l

13 If w = exp(z/a) where w = u + iv, z = x + iy, and a is real, find the locus


of w corresponding to each of the lines y = 0, y = \Tia in the z-plane.

u + iv = exp (x + iy), find the locus of the point (u, v) in the w-plane
x = constant; (ii) y = constant. Verify that these two loci cut ortho-

14 If

when

(i)

gonally.
15

Criticise the following

argument.

dy

= sin x +
dx

1.

If y

cos x + i sin x, then

*(cos# + isina;)

A e ix

by 4.41, Remark (a), y =


Hence cosaj+isina; = e ix

Since y

%y\

when x =

0, therefore

14.67 Generalised circular


(1)

cos x

therefore

A=

'

and hyperbolic functions

Consistently with formulae


cos z

sin z

(vii),

we define

+ exP ~ * z )}>

|{exp

(viii)

and then tanz to be

{exp

(iz)

exp iz)},
(

sinz/cosz, etc.

These definitions are equivalent to


cos z + i sin z

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!-

Directly from these definitions

it

can

now be shown

that the

formulae relating trigonometric functions of a real variable continue


to hold for a complex variable. (The definitions were chosen with a
view to making this so.) This can be done as in real trigonometry if
we first verify the addition theorems together with results concerning
special angles, e.g.

cos0=l,
and

sin0

0,

cos

z) =

cos z,

sin(

z) = sinz,

also periodicity properties.


38-2

de moivre's theorem

558

[14.68

For example, to prove that


sin (z r

+ z2 =

sin z x cos z 2 + cos z x sin z 2

we should use the definitions


into exponential functions

to the translation of the


(2)

to translate each

(viii)

term on the right

and verify that the right-hand side reduces

left.

We define
ch z

{exp

(z)

+ exp - z)},

sh z

|{exp

(z)

- exp - z)},

(x)

and then thz

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! +

can be verified that

formulae relating hyperbolic

all

functions of a real variable continue to hold for a complex one.


ever, in contrast to the real case, definition (x)

of exp z show that ch z and sh z have period


j.1

and exp ( ni) =

also th z has period ni,

+ 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.

and hyperbolic functions Osborn's

14.68 Relation between circular

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)}

= - i{exp - z) - exp (z)}


= ishz

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

In any formula relating circular functions of zx z 2 ... we may replace


zl5 z 2 ... by iz 1} iz 2 ... and then use results (xii) to translate it into
hyperbolic functions of z x z 2 .... In the course of this translation each
product of two sines becomes i 2 multiplied by a product of two hyper,

bolic sines. Since the formulae all hold in particular when the variables

we have the justification of Osbom's rule stated in 4.44 (4).


An example of the detailed verification is the following:

are real,

ch (A

+ B) =
=

cos {iA

+ iB)

cos (iA) cos (iB)

sin (iA)

sin (iB)

= chA chB-i 2 ahA ahB


= ch.A chB + ahA ahB.
Examples
cos (x + iy)

(i)

sin (x + iy)

and

\mi(x+%y)

=
=
=
=

cos x cos (iy)

cos a;

sin x sin (iy)

chy iainx shy;

sin x cos (iy)

+ cos x sin (iy)

sin a; ch.y + icosx shy;

sin (x + iy)

cos (x + ry)

sin (x + iy) cos (x iy)

r-

cos (x + iy) cos (x %y)

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
= (l-sin 2 a;)ch 2 + sin 2 a;(ch 2 2/-l)
= ch 2 y sin2 x
/

Similarly

Thus

< ch2 y for all x.


cos z| 2 = cos 2 x + sh 2 y ^ sh 2 y for all
|sh?/| < |cos2| < chy for all z.

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) =
%

T sin(*a;) cos y - cos (ix) sint/


%

= shx cosy + ichx siay,


shz z = (ahx cosy x) + i(ohx amy y),
which is real if ch x sin y = y.

'

DE MOIVRE'S THEOREM

560

Clearly y

satisfies this

equation which,

of 7T, can be written

chx =

smy

if

[14.68

not an integral multiple

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

Prove that ch (iz)

By using results

cos z, sh (iz)

isinz.

in 14.68, prove that

(xii)

+ B) = shA chB + ch^ shB


th(A+B) = (thA+thB)/(l+thAthB)
sh(^L

and

follow from the corresponding properties of circular functions.


3

Use the

definitions (viii) in 14.67 to prove that cos 2 z + sin 2 z

sin(2 1

Excess

+ 22 =
)

the following in the form

>

0,

prove that thy

and

a + ib.

5 sec (x + iy).

4 ch (x + iy).
7 If y

sinz x cosz 2 + cosz x sinz 2

|tan(a;

+ *t/)| <

th.

(x

+ iy).

cotht/.

8 Determine the general forms of z for which

(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.

9 Find the general solution of sin z


10
11

[Equate imaginary parts.]

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 =

Interpret these as loci corresponding to the lines x


respectively.

= constant,

= constant

DE moivre's theorem
13

Expand

561

(sh z sin z) in ascending powers of z.

U8eC+iS to find the sum to infinity of


M

sin 26

14

sin 40

sin 60

41

cos 96

cos 50

cos0H

15

h..-

21

1-....

\-

61

9!

5!

Miscellaneous Exercise 14(A)


1

If x

cis a

a;

-2/

x+y

and y =
..

* tan

prove that

ciafi,

and

p)

+ y)(xy-l)

(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)

(cos ft + i sin /?)}" + {(cos a i sin a) (cos fi i sin /? )}"


= A n 2 n + 1 sin n |(a- 5) cos^(a+/?),
y

= l)* n if n is even, and A n = l)*< n + if n is odd.


Solve x 6 + xi +x* + x* + x+ 1 = 0.
0.
Find the solutions of x6 = 1 for which 1 + x + a; 2

where A
4

(i)

(ii)

1)

=}=

5 If n is a positive integer, show that every root of (z+


purely imaginary. If the roots are represented by x
a
origin, prove OP\ + OP| +
+ 0P2 = 2n(2n - 1).

P P

is

. .

l) an
.,

+ (z l) 2n =

P2n and

is

the

. . .

6 Show that the roots of (z


circle of radius f,

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

and that the roots are

(r

33

0, 1, 2, 3, 4).

2m\

4cosI.
]J[ (5
5 /
31 r=1 \

are integers, prove that 1 + a* + a 2* +


+ a(n-1)fc
is zero otherwise. [Sum the g.p. when

and n,

. .

a multiple of n, and

4= 1.]

8 Express (1 t 2 )/{(l ot)(l bt)} in partial fractions when the non-zero


constants a, b are (i) unequal; (ii) equal. Taking a = ft -1 = cis0 and a suitable
value of t, deduce that if < <j> < \n,

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

cot0. [Method of 6.61,

'

(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 (0 + Jw) + cot (0 + \tt).

cot

cos 60 + 6 cos'40- 9 cos

as a polynomial equation in cos 6,

3/9

+15 cos 20 -27 cos + 14 =


(9

and hence

find all solutions

and

between

inclusive.

12 If a x + a 2 +

. . .

sec

If

13

lt

2,

+ a 2n =
sec

2it,

. . .

prove that (with the notation in 14.23)

sec

2n

= 1 E 2 + 2 4 ... + 1)"
(

6 3 are roots of tan(0 + a)

and if A

tegral multiple of n,

4= 1 ,

prove

and 2n which

'

acos

+ 2/&cos0sm0 + &sm

prove tan ^(6 1 +

+ 6S +

4)

2h/(a

2n .

= A tan 26, no two differing by an in a for some integer n.


X + 2 + 6 3 = nn

*14 If 6 X 6 2 63 6 4 are values of 6 between

in

(0

Express

11

2tt

Xl

satisfy

+ 2grcos0 + 2/sin0+c =

b).

[Put

0,

tan 0 to obtain a quartic

*.]

u n = # 2n 2a; n cos na + 1 by methods of real algebra as follows.


Prove u n+1 = 2xu n cos a C 2m_x + (x Zn + 1) uv
Use Mathematical Induction to prove that u x = x 2 2x cos a + 1 is a

15 Factorise
(i)

(ii)

factor of
(iii)

(iv)

a;

2cos(a + 2/77-/n) + 1

is

also a factor of

un

for

any

unaltered by replacing a by ct + 2mjn.'\


Verify that r = 0, 1,2, ...,n 1 give distinct factors.

integer

16

un

Deduce that
r.

Solve

[u n

(z

is

+ 1) 8 -z 8 =

0,

and prove that


3

+ 1) 8 -

(z

= ^(2z + 1)

+ 42 + cosec 2

II
r=l

Hence show that


3

16(cos 16 6 - sin 16 6)

cos 26 YL (cos 2 20 + cot 2 \rn).

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

AP .BP .CP .DP .EP .FP < 2a

6
.

18 Prove
sin 50

16 sin

sin (0 +

\tt)

sin (0 +

|7r)

sin (0 +

and by considering lim (sin 50/sin 0) deduce that

sin

ftf)

\n sin

sin (0 + $n),
\ir

= + \ *J5.

0-*O

19 If

w = u + iv,

prove that the point

= a + M/, w = (expz- l)/(expz+ 1), and \tx <y


the circle \w\ = 1.

(u, v) lies inside

<\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 = (z-l)exp(-*a) + (z-l)- 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

logr + i(d + 2nn), where

is

any

cos (u + iv), prove

(l+x) 2 +y 2

and

(ch + cosw) 2

(1

-x) 2 +y 2

= (chv-cosw) 2

< tt) in the preceding, find cosw and cht> in


If a; = cos 0, y = sin0 (0 <
terms of cos \0, sin \d, justifying the choice of sign when square roots are taken.
= coth (x + iy), prove v = - sin 2y/(ch 2x cos 2y). Show that
24 If u +
=

iv

{1

- exp (2a; - 241/)}- 1 - {1 - exp (2x + 2m/)}" 1


00

and deduce that

if

<

0,

then

= -22e

2ra!

sin 2ry.

r=l

25 If to = z 2 sketch in the Argand diagram the path of w when z describes


the sides of the rectangle whose vertices are a, a + ia (a real), starting at
O and moving counterclockwise.
26 (i) Sketch the graph of y = thx. Show that the equation thas = Aas has
,

two

real non-zero roots if

<A<

1.

Prove that the equation tanz


A < 1; but that if A has any other
(ii)

<

Az has two purely imaginary roots

if

real value, all roots of the equation are

real.

Find the sum


27 cos 2 a +

to infinity of

^ cos

2a + ^ cos3 3a +

28 a?sin0+ sin30 + |-sin50 +


3!

. . .

...

(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)

By considering exp (a;) + exp {<ox)+ exp (a) 2x) where w 3 =


1

*(ii)

= exp (ax + ibx)

r cis 0.]

+ |.+?L + !_ + ... =

^{e* + 2 e-** cos (%x V3)}

What results are obtained by

[Use Ex. 13 (g), no.

l,w

4= 1.

prove

for all real x.

considering

(a)

exp (x) + (0 exp ( (ox) + w 2 exp (o) 2x)

(b)

exp (a;) + w 2 exp (wa;) + (o exp (w 2a;) ?

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 ^.

of 8.12.] Deduce that

~y\ds 2 )

\ds 2 )

~d^ds 2 ~dsds 2

'

8(d), no. 8.)

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

Advantage of oblique axes

Although

it is

usual to refer cartesian coordinates to two perpen-

which use of oblique axes


more convenient. For example, in proving properties of the triangle
by coordinate methods, two of the sides can be chosen for coordinate
axes. The coordinates (x, y) of a point
P are then its signed distances from Oy,
Ox measured parallel to Ox, Oy respectively. The angle xOy (measured
counterclockwise from Ox) is denoted
by a).

dicular axes Ox, Oy, there are problems in


is

In sections 15.1-15.5 we revise the

main results about 'the straight line',


and extend them to the case of oblique
axes when this does not introduce any

Fig. 139

not the axes are oblique', they are

Notation. In Chapters 15-19

The
same in form whether or

essential complication.

reader should observe which results remain the

marked by the

we

sign

{&>}.

use the convention that

has

coordinates (x v yj, and so on.

15.12 Cartesian

We

and polar coordinates

obtain the relations analogous to those in 1.62 for rect-

first

angular axes, taking

as pole

and Ox

as initial line

(fig.

PN perpendicular to Ox; then


r cos 6 = ON = OM + MN = x + ycosco,
r sin 6 = NP = y sin
to.

tt

Hence

tan a

also
r2

tysinw

x + y cos a)

x% + y 2 + 2xy cos o>.

140).

Draw

ELEMENTARY COORDINATE GEOMETRY

566

[15.13

Example
If the line

PiP 2 is of length r and makes angle


r cos

(a; a

rsin#

6 with

Ox

(fig.

141),

then

- a^) + (y 2 - yj cos w,
=

(y 2

yi) sin o>.

Fig. 141

15.13 Distance formula

From

the above example, or directly from the cosine rule applied

to triangle

PPQ
X

PX P\ = r =
2

(fig.

(x 2

141),

- Xj) 2 + (y 2 - y x 2 + 2(x2 - xx

15.14 Section formulae

(y2

- yx

cos o).

{to}

These give the coordinates of the point dividing PX P2 in the ratio k


Let P(x, y) be the required point, so that PX P\PP2 = kjl.
(1) Internal division.

triangles

With the construction shown

PX PQ, PP2 R are similar; hence

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,

ELEMENTARY COORDINATE GEOMETRY

15.15]

In each expression observe that the number


to the segment

multiplied
(2)

k >

P P,

is

by that

of

as in

fig.

which corresponds

by the coordinate of P2 while

multiplied

is

Pv

External division. Lettering

I)

k,

567

142,

we

so that

fig.

143 (which illustrates the case

have similar triangles

still

PX PQ,

PP^R,

Pil^M
~

PR'

k
i.e.

=
~

from which

Similarly

'

l-k

'

l-k

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

division in the negative

we agree that PX P PP2 shall be reckoned posiPX P2 internally and negative when the division is

if

both cases are covered by the same formulae,


x

=
l

15.15 Gradient of a line

The

xz
+
+ (-k)

formally the same as that of

and I were understood to be

result of (2)

+k

viz.

= iyi+ty2
l + k

{to}

gradient of a straight line

is

defined to be the tangent of the

angle which the line makes with the positive #-axis.

Since this

'

ELEMENTARY COORDINATE GEOMETRY

568

[15.16

makes no reference to Oy, we have (as for rectangular axes)


two lines with gradients m x m 2 are parallel, then m 1 = m 2 if
they are perpendicular, then m x m 2 = 1; and in the general case an
angle between them is given by
definition

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

15.12, the gradient of

tan0

15.16
(1)
(r 3 ,

3)

Area of a
One
;

fig.

145.

similarly for

+ (y2 - y x

cos ft>

P3

Area of triangle

Let the polar coordinates of

Then by

r 2 cos d 2

and

(x 2 - x )

triangle

vertex at 0.

see

PxP2 is

(2/2-2/1) sin w

P2 P3
,

be

(r 2 ,# 2 ),

15.12,

= x2 + y 2 cos (o,

= y2 sin &>,

r2 sin d 2

OP2 P3 = OP2 0P3 sin P2OP3


.

r2 r3 sin(0 3 -<9 2 )

\r2 r3 sin #3 cos # 2

i(^2

+ 2/2 cos w

(2/3

2 r3

cos

sin 6 2

sin w )

- %{y 2 sin w) (# 3 + y3 cos w)


=
_
-

\ sin &>(a; 2 /3 x3 y2 )
1

2/2

after reduction

smw.

2/3

Bemarh. According as the vertices O,

P2 P3 are named in counter,

clockwise or clockwise order, the angle 6 3 6 2

and hence the expression


tively. If we consider

is

positive or negative

for the area is positive or negative respec-

a signed area which is positive when the vertices

are named counterclockwise and is negative otherwise, then the above


result gives the area of

OP2 P3 in magnitude and sign.

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

2/2-2/1

2/2-2/1

""l
Ps
xi
X2
XZ

Vi
2/2

Vz

sinw

Vz-Vi
sm<y

Vz -2/i
1

sin to

by the operation c -> c x + c3 followed by c a ->


step is verified by expanding from the first row.

c 2 + /x c 3 the middle
;

ELEMENTARY COORDINATE GEOMETRY

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

Forms of the equation of a

15.21 Line with gradient

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

).

is satisfied by the coordinates of any point P on


what we mean by 'the equation of the line'.

This equation, which


the required

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

for oblique axes.

form

The above shows that any line with gradient m


has an equation of the form y = mx + c. Conversely, any equation
(

1)

Rectangular axes.

which can be put into the form y

= mx + c

represents a straight line of gradient

For when x
locus; also

(where m, c are constants)

making intercept

= 0, the equation shows that y = c, so


{y c)j(x 0) = m, so that the gradient

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

(y-c)sinw

x+(y c) cos to'

c) sin (to 6)

x sin

(i)

ELEMENTARY COORDINATE GEOMETRY

15.23]

and

571

sin 6

= -r. a\ x + e
sin (o)
6)

The equation of the


no longer represents
If the equation y

line is therefore still of the form

= mx + c is given, then m =
and

Hence

this

independent of

is

= mx + c,

but

now m

the gradient of the line.

for the gradient of the join of P{x,y)

which

'

(x,y).

(y c)jx,

and the expression (i)


becomes msinw/(l +wcosw),
equation represents a straight line

(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,

the axes being

a>.

The gradients of the lines are


m,

sin o)

sin a)

l+m

1+miCosw

cosa)

an angle between the lines is given by tan _1 {(/ 1 /* 2 )/(l +/* 1 /t 2 )}


unless fixfi 2 = 1 (in which case they are perpendicular). After reduction this

By
is

15.15,

found to be

+ (m(m+ m m
,

tan -1

the lines are perpendicular if 1


if

m =m
x

+ (* 1

If

Ax + By + C =

and

are parallel

is

B=

the equation of a straight

but

#= 0,

{to}

y = (AjB)x CIB, which

the equation can be written

=|=

15.22

If
is

cos a) +

2.

15.23 General linear equation

by

sin to

mm
+ m )cosw + m m = 0,
2)

>

2)

line.

the equation can be written x

= C\A which
y

the equation of a straight line parallel to Oy.


Since these are the only possibilities,

it

follows that every linear

equation in x, y represents a straight line.

15.24 Intercept form: line

making

on Ox, Oy

intercepts a, b

{to}

Let the gradient form of the equation of the required line be


Since the line passes through the points (a, 0), (0, b), we

y = mx + c.

nave

Hence

= ma + c and
c

c.

and m = bja, and the required equation


b

is

tv

^ y
- + |=1.
a b

i.e.

39

GPMII

572

ELEMENTARY COORDINATE GEOMETRY

[15.25

CB

+ l/CB

Example

ABC is

is constant.

triangle with sides

Prove that

CA,

along given lines, and 1/CA

AB passes through a fixed point.

Choose the lines along which CA, CB lie


to be Ox, Oy. Let the coordinates of A, B be
(a, 0), (0, 6); then by hypothesis 1/a+ljb = k,
where k is constant.
The equation of
is xja + yjb = 1. The
above condition shows that this line passes
through the point (\jk,\jk), which is independent of a, b.

AB

Fig. 147

15.25 Line joining

Pv P2

For rectangular

(1)

{to}

axes, if P(x, y) is

any point on the

line,

then

V-Vi = Vi-Vz
ry*
*As

since each expression


This result

(2)

nix + c, then

is

is

/y*

/y

rv*

equal to the gradient of the

also true for oblique axes.

For

if

line.

the required line

is

= mx x + c and y 2 = mx z + c, so that by subtraction


x
y-y 1 = m(x-x 1 and y 1 -y i = m(x 1 -x 2 );
t/

and the equation


(3)

is

obtained by division.

In either case the result can be obtained as follows. Consider

the equation

y
yi

0.

y%

by expanding from the first row), and


therefore by 15.23 represents a line. It is satisfied by (x^y^) and by
(x2 y 2 ) (since then two rows become identical). Therefore it represents
It

is

linear in

x and y

(as is clear

(This form of the equation

the line

could also be seen from the example in


15.16.)

15.26 Parametric form for the line through


(jtj,

yt)

in direction 8

be any point on the line, and let


=
PX P r; here r is a signed length, so it will
be positive for points of the line on one side

Let

of Px and negative for points on the other.

Fig. 148

The axes being rectangular,

x x 1 = rco&6 and y y x =

r sin 6.

ELEMENTARY COORDINATE GEOMETRY

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

form in the treatment of

shall use this

no simple analogue

conies.

There

is

for oblique axes.

15.27 General parametric

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),

iax,y and therefore represents some

15.28 Perpendicular

line

(by 15.23).

form

Let the perpendicular from O to the


and make angle a with Ox. Then the

line

have length p

foot

(essentially positive)

N of the perpendicular has polar


I

N
P/

\P(*. y)
A

>\
\ x

Fig. 149

coordinates (p,cc). Let any other point


(x, y) and polar coordinates (r, 6); then

p=

rcos(0 a)

P of the line have cartesian coordinates

= rcosd cosa + rsin# sina


= x cos a + y sin a,

so that the line has equation

= 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

on the axes are

15.24.

Further results

15.31 Sides of a line

{co}

The point P((lx 1 + 1cx 2 )l(l + k),


whose equation is ax + by + c =
a(lx x

(fo/i

2 )/(^

+ ^)) ^ es on

+ kx 2 + b{lyx + ky 2 + c(l + k) =
)

tne une

if

0,
39-2

ELEMENTARY COORDINATE GEOMETRY

574

h
t

i.e. if

ax,+by1x + c

^r
ax 2 + by
2 +c

P P

If

lf

...

(*)

on opposite sides of A (fig. 150), then P divides P\P2


is positive; hence ax 1 + by 1 + c, ax 2 + by 2 + c have

2 lie

internally

[15.31

and

so kjl

opposite signs.

Fig. 151

Fig. 150

Similarly
so ax1

if P\,

P2 lie on the same side

151), kfl is negative

(fig.

and

+ byx + c, ax2 + by2 + c have the same sign.

It follows that the converses of these statements are also true.

Examples
(i)

P8 P

Menelatis's theorem. Let the line ax + by + c


jP]P 2 of the triangle 1 2 3 in the ratios k

PPP

X,

u n = ax n + by n + c
we have by

(i)

{n

=
:

divide the sides

l lf

k2

1,2,3)

above:
&2
'

u9
&2 ^3

Multiplying,

1.
ij

(ii)

Ceva''s theorem. If lines

l3

1%

P Q P%Qi, P3Q3
P P P are conP P P3 Pi, P\Pi
1

through the vertices of triangle

X,

current at (h, k) and divide


2
3
in the ratios kt lx k2 l2 k3 l3 then since the
equation of X Q X is (see 15.25(3))
,

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,

ELEMENTARY COORDINATE GEOMETRY

15.32]

On

multiplying,

575

Ki

/Cg

K3

Zj

Z2

Z3

15.32 Perpendicular distance of a point from a line

Let d denote the length (essentially positive) of the perpendicular


from Px to ax + by + c = 0, and let the foot be Pa Then
.

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

can be written (analogously to


a{x z - xx )

Square and add

(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

'

where the sign


ax x + by x + c > 0,
If c

4=

is

line,

make the expression positive: +


if ax 1 + by1 + c < 0. Also see Ex. 15(a), no. 15.

chosen to

and the equation of the

then a x x + b x y + c
the

will

if

line is written so that c is positive,

be positive when

and negative otherwise, by

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

f The main purpose of this exercise


solving locus problems.

= h + kcoso).
is

to illustrate the use of oblique axes in

ELEMENTARY COORDINATE GEOMETRY

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).

Two fixed lines Ox, Oy are cut by a variable line at A, B respectively;


Q are the feet of perpendiculars from A, B onto Oy, Ox. JiAB passes through

4
P,

PQ will also pass through a fixed point.

a fixed point, prove that


5 If the equal sides

that

BE CF = AB

AB,

AC of an isosceles triangle are produced to E, F so


EF always passes through a fixed point.

prove that

PM, PN are drawn from P onto Ox,


PM + PN = 2c.
7 OM + ON - 2c.

From a

Perpendiculars

Oy. Find the locus of


8

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.

any two straight lines are drawn to cut one fixed


10 Through a fixed point
Ox at A, and another fixed line Oy at G, D. Prove that the locus of the
meet of BG,
is a straight line through O.

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

15.41 Lines through the


If L,

lines

meet of two given

U are linear functions of

x, y,

lines

{w}

then the equation

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

by any point which

the lines L

straight line parallel to both (as

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.

One condition will determine the constant

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 mid-points 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

and AC, BD,

OE are its diagonals.

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

intercept theorem of elementary geometry these will cut Ox,


points A'(a, 0), B'(b, 0), C"(0, c), D'(0, d), and their equations will be

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,

(d c)x + (a b)y = ad be.

i.e.

The mid-points of BD,

AC are

(6, d), (a, c),

and these

clearly

lie

on

this last

line.

15.42 Condition for concurrence of three given lines

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.

ELEMENTARY COORDINATE GEOMETRY

578
If

two of these

point (obtained

say the second and third, the

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

2)+M C 2 a 3- C 3 a 2) + C l( a 2&3- a 3&2) =

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

that (with the notation of 11.31)

A=
Hence

CiCi + CgOa

+ CgOg =

0.

*/ the lines are concurrent or parallel, then

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

This converse argument fails if a 2 b z a z b 2

A=

0.

first line

because then

a 2 b z -a z b 2

0;

0.

but a similar argument

applies to the lines taken in a different order unless also

az b 1 a1 b z
in

which case the three

and

a 2 b z az b 2

lines are parallel.

0,

Hence when A =

the lines

are either concurrent or else all parallel.

Thus the condition A =

is

necessary and sufficient for concurrence

or parallelism; see also 11.43, Corollaries I

(6), (c).

Exercise 15(6)

Find the equation of the line concurrent with 2x 3y 3 =


and
x + Sy 15 = and passing through ( 2, 1); verify that it passes through
1

the origin.

2 Find the equation of the

5x-2y =
3

4,

8x-ly + 5 =

line of gradient

f which

is

concurrent with

0.

Find the equation of the perpendicular from the meet of the

3x- ly =

2,

4x + 5y

9x+ 10y+ 15 = 0.
lines which make numerically

lines

to the line

4 Find the equations of the


on the axes and which are concurrent with Zx + 2y

0,

equal intercepts
2x y + 3 = 0.

ELEMENTARY COORDINATE GEOMETRY

15.5]
5

Show that for

(i)

579

numbers k the equation

all

2x + 3y-l

+ k(3x-y +

4:)

represents a line passing through a fixed point,


(ii) Find the point of intersection of the lines

2x + 3y-l + a(3x-y + 4:)


b(2x +

3y-l)-c(3x-y + 4)

=
=

0,
0.

6 Find the equation of the line joining the meet of


3x + 4y-l

3x- 5y +

to the meet of

37

= 0, 2x-3y+l =
= 0, 5x-2y-8 =

0.

Test the following sets of lines for concurrence.

=
=

2x-y-l - 0, 1x-5y+13 = 0.
x-By
+ 2 = 0, 3x + 8y-6 = 0.
0,
(p + l)x + (p-l)y+p = 0, (q-l)x + (q+l)y + q = 0, x = y (p * g).
rc-^ + aJ* = 0,x-t 2 y + at* = 0, (^yi + fl-M^ = o(i + ) (*i * < 2 )are concurrent, find
If the lines x- 2y + 3 = 0, 2x- 3y = 0, ax + y+ 1 =
3x-2y +

<t

8 5x + 2y-4:

9
10
11

the value of

0,

a.

12 If the lines ax + 2y+ 1 = 0, bx + 3y+ 1 = 0, ca; + 4?/+


prove that a, b, c are in arithmetical progression.

b 3 , 3cx + 2y

are concurrent,

for concurrence of the distinct fines 3ax + 2y


[Use theory of equations.]

Find the condition

13

3bx + 2y

c3

= a3

*14 Find the condition for concurrence of the distinct fines

x sin 3a + y sin a
[If (h, k) is the

a,

a:

sin

3^ + y sin/?

common point, then 6 =

(& + 3ft) sin 6 + a =

<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,

lines are concurrent

If

*(ii)

+ b-l y + c 1 + m{a 2 x + b 2 y + c 2 + n{a3 x + b z y + c z =


)

a,

absent, the roots

(i)

lines,

a.

must satisfy h sin 3d + k sin 6 =

Since the term in sin2 6


sin a, sin/?, siny of this cubic in sin0 satisfy 2 sin a = 0.]

4h sin3

i.e.

distinct

(A)

0,

or else all parallel.

use 11.43, Theorem II to show that numbers

I,

m, n not

all

zero exist such that

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).

Line-pairs

15.51 Equations which factorise linearly

A single equation in x, y may represent two or more lines.


x z = can be written y = x and therefore represents
+ x, y = x.
2
2
can be factorised as (x + y)(x + 2y) = 0, and
(ii) x + 3xy + 2y =
sents the line-pair x + y = 0, x + 2y = 0.
(i)

the pair

of lines y

so repre-

ELEMENTARY COORDINATE GEOMETRY

580

[15.52

x + Sxy + 2y + Sx + 5y + 2 = can be written (x + y + 2) (x + 2y + 1) = 0,


therefore represents the pair x + y + 2 = 0, x + 2y+l = 0.
(iv) x 3 + 3x 2 y + 2xy % = 0, i.e. x(x + y)(x + 2y) = 0, represents the three lines
2

(iii)

and
x

0,

x+y

0,

aj

+ 2?/ =

0.

In general, the equation

(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

the given equation, then either

+ b2 y + c2 =

or both.

+ 2hxy + by 2 =

the equation

0,

of these equations, then it satisfies the original

x(ax + 2hy)

is

0,

representing a line-

pair.

If b

4= 0,

the equation can be written

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
,

equation represents two distinct lines through O.

When h % =

ab, the

equation

sents the single line hx + by

is

(y/x + h/b) 2

and therefore repre-

We

say conventionally that the


equation represents two coincident lines, or a repeated line.
When h 2 < ab, the quadratic for yjx has no roots. The original
equation is satisfied solely by the values x = 0, y 0. Hence if

h2 <

0.

ab, the equation represents the origin only.

The three general results

just given include the case 6

disposed

of at the start.

The work of

15.5 so far

is

valid for oblique axes; in the following

we

shall suppose the axes to be rectangular.

(2)

Angle between the lines ax 2 + 2hxy + by 2

have equations y
has roots

and

so

m m
x,

a.

x x,

2 x,

0.

If the separate lines

then the above quadratic for yjx

Hence

(m 1 m 2 ) 2

= (m 1 + m 2 2 4m 1 m 2 = (h 2 ab).
)

ELEMENTARY COORDINATE GEOMETRY

15.53]

581

The required angle is given by

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

Ae Zir&es are perpendicular;

and

conversely.
(3) Bisectors of the angles

of the perpendiculars from

between these lines.


(x,

By expressing equality
m1 x = 0,y mi x = 0,

y) to the lines y

the equations of the angle-bisectors are

y-m

V(!+i)
Hence the two

bisectors

y-m x
-V( 1 + W D*
2

have equation

{y-m^f {y-mz x) 2 =
l+m|

1+raf
i.e.

(1

+m|) (/-m 1 a;) 2 -(l + w&f) {y-m % xf =

(raf- m|) a; 2 2(m! -

i.e.

If

4=

2,

by

satisfies

as of course

(2),

(1

-*/ 2 )

it

in x,y.

0.

= 2(l-m 1 m a )a;2/,
=

(a-b)xy.

should from elementary geometrical considera-

x2

xy

6=0.
1

line-pair 5 =

(1) Parallel line-pair

is

)xy {m\ ml)y'2

'

can be written

The general

This

the perpendicularity condition at the end

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.

through the origin. Write

= ax 2 + 2hxy + by 2 + 2gx + 2fy + c =

0.

(i)

the standard form of the general equation of the second degree

BA

ELEMENTARY COORDINATE GEOMETRY

582

[15.53

If s can be factorisedf in the form

= (A^ + Brf + CJiAzX + Bzy + Cz),

then

(i)

we

so

By equating coefficients of the second-

represents a line-pair.

degree terms in

also

(ii)

(ii),

= A1A2

2h

= A X B2 +

= B X B2

X,

have
ax 2 + 2hxy + by 2

Therefore

*/

(A 1 x + B 1 y)(A 2 x+B 2 y).

represents a line-pair, then the equation

(i)

ax 2 + 2hxy + by 2

(iii)

represents the parallel line-pair through 0.

This fact can be used to calculate the angle between the lines

which

is

the same as the angle between

When h 2 =

ab, the lines

this condition implies

Cx C2

be equal to
(2)

If

Ax

(i)

=j=

(i),

(2).

to represent

a line-pair.
(ii); and if
means that (i),

represents a line-pair, s can be written in the form

and

A% ^

(i.e. if

regarded as a quadratic in

Now

found in 15.52

may be coincident or parallel; for although


2 these ratios may or may not

A X :A 2 Bx

Necessary condition for s

(i)

(iii),

(i),

x,

Bf

~A

#= 0,

since a

= AXA2

),

can be solved in the form


C-t

y ~T'

x=z

written as a quadratic in

Ba

Co

~T y ~T'

(lv)

x, is

ax 2 + 2(hy + g)x + {by 2 + 2fy + c)


so

this

0,

(v)

- (hy + g) V{(% + g) 2 - a(by 2 + 2fy + c)}].

Hence (i) can be solved in the form (iv) if and only if the quadratic
y under the square-root sign is a perfect square, i.e. if

in

(h 2
is

- ab) y 2 + 2(gh -af)y+ (g 2 - ac)

a perfect square. This


(gh

(which holds
gh af=

is

so if and only if

- af) 2 - (h 2 - ab) (g 2 - ac) =

even when h ab =
2

0, for

0,

the condition then implies

also), i.e. if

a(abc + 2fgh - af 2 - bg 2 - ch 2 )
j-

0.

This will not be the case in general: see 15.74.

ELEMENTARY COORDINATE GEOMETRY

15.53]

Since

we

are assuming a

the condition

4= 0,

is

abc + 2fgh - a/ 2 - bg* - ch?


If a

but b

quadratic in
If a

0, b

y,

4= 0,

then equation

(i)

and the same condition


0,

but h

equation

4= 0,

v+

which,

j.

(vi)

0.

can similarly be solved as a

(vi) is

obtained.

(i) is

2hxy + 2gx + 2fy + c


i.e.

x+ l y + lL =

0,

if factorisable, is

H) H) f=

so that

2fgh-ch*

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)

in 11.22, (vi) can

We have therefore shown that

to represent a pair of straight lines is

(vi).

be written

A=

(vii)

0.

The determinant A is called the discriminant of s,


Remark (a). The necessary condition A = is
to represent a line-pair. For A is zero when
a

if

would not be of second degree.

the necessary condition for

By ex.

583

= b=l,

c=f = g =

or of equation

(i).

not sufficient for

(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).)

Point of intersection of the line-pair. In a numerical case we


should resolve the second-degree equation into linear factors (e.g. as
(3)

ELEMENTARY COORDINATE GEOMETRY

584

in 10.21, ex.

For the general case

(ii)).

(i),

[15.54

equations giving the point

of intersection (sometimes called the centre or vertex of the line-pair)

can be found as follows.


Writing (i) in the form

(v), we see that to a given value of y there


correspond in general two values of x; but if y is the ordinate of the

meet P, then

(v) will give

only one

value of x, viz. the abscissa of P.

Now when
root

is (see

has equal roots, that

(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.

by 2 + 2(hx +/) y + (ax 2 + 2gx + c)

P we have

at

Equations
see ex.

hx+f

hx + by+f

i.e.

(i)

Remark

(viii), (ix) suffice

0,

(ix)

0.

to determine

P as the point (GjC, FjC);

in 11.32 for the notation.


(/?).

Since

(i)

can also be written

(ax + hy + g)x + (hx + by+f)y + (gx+fy + c)


it

follows that the coordinates of

Equations

0.

(x)

(viii)-(x) are therefore consistent,


(vii)

as a necessary condition for s

intersecting line-pair.

Compare

0,

P must also satisfy

g%+fy+c =
of 11.43 gives

(2)

above, where

and Corollary

(vii)

1(6)

to represent an

was shown to be

necessary for any type of line-pair, not necessarily intersecting. Also


see 15.73, ex.

(ii).

15.54 Line-pair joining

to the meets of the line Ix + my

= 1 and

the locus j =

In general the equation s = represents a curve, but the following


argument holds in particular when the equation represents a line-pair..

ELEMENTARY COORDINATE GEOMETRY

15.54]

Consider the homogeneous equation in

ax 2 + 2hxy + by 2 + 2(gx +fy)


It

is satisfied

(Ix

585

x, y:

+ my) + c(lx + my) 2 =

0.

by the points which satisfy both the equation of the line

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-

points. Since the equation is

through O. Hence

lines

it

pair.

Exercise 15(c)
[Rectangular axes.]

Find the area enclosed by the lines x 2 + 4xy + 2y 2

2 Show that the area enclosed by the


ax* + 2hxy + by 2

and the line x + y =

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

{ax2 + 2hx 1 y1 + by\)IJ{{a-b) 2 + Ah 2}.

which are perpendicular to the


4 Find the equation of the lines through
ax 2 + 2hxy + by 2 = 0. [If (x, y) lies on either of the required lines, then
( y, x) lies on one of the given lines.]

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 =
angle-bisectors, prove that h(a' b')
6 If the

lines

a'x 2 + 2h'xy + b 'y 2

0,

have the same

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

Show that the

find the angle


(i)
(ii)

following equations each represent pairs of straight


between them, and their point of intersection.

2x 2 + lxy + 3y 2 -4x-ly + 2 =
15x 2 + xy - 2y 2 + 3x - y = 0;

lines;

0;

xy-3x + 2y-6 = 0;
x 2 + 4xy-2y 2 + 6x-12y-l5 = 0.
9 Write down the equations of those line -pairs through the origin which are
(iii)

*(iv)

perpendicular to the line-pairs in no.

8.

[Use no.

10 Find the equation of the lines joining


lines 4a; 2 + 16xy - 12y 2 - Sx + 12y - 3 = 0.

4.]

to the

meet of x + 2y

3 with the

* 1 1 Show that the line 2(g -g')x + 2(/-/') y + (c - c') =


is a diagonal of the
parallelogram formed by the line -pairs s = ax 2 + 2hxy + by 2 + 2gx + 2fy + c = 0,
' = ax 2 + 2hxy + by 2 + 2g'x +
2f'y + c' = 0. [Consider s s' = 0, which is linear
and is satisfied by the points common to s = 0, s' = 0.]

*12 If the line -pairs s


equation.

0, s'

in no. 11 possess

line in

common,

find its

ELEMENTARY COORDINATE GEOMETRY

586

The

15.6

circle

15.61 General equation of a circle; centre

and radius

Referred to rectangular axes, the equation of the

C(M) and radius r is


When simplified,

{x

this equation is of the

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 xy-term.

by completing the square for the terms involving x and

for those involving y, equation


(*

<7)

(ii)

can be written

+ (2/+/) 2 =

+/ 2 -c.

9'

If g 2 +f 2 c > 0, this represents the circle with centre ( g, /) and


radius J(g 2 +f 2 c). If g 2 +f 2 c = 0, it represents the single point
(

ff>

/); and

if

+f 2 c <

0,

the equation does not represent any

locus.

Example*
Referred to oblique axes at angle o), the circle with centre C(h, k) and radius r
has equation
(x-h)* + (y-k) 2 + 2(x-h) (y-k) cosw = r a ,

which

is

of the form

x 2 + y 2 + 2xy cos (0 + 2gx + 2fy + c

15.62 Circle

on diameter

0.

P P2
l

Let P(x,y) be any point on the required


by 'angle in a semicircle'. Since
gradient of

PX P = UzMl

and

circle;

gradient of

- xx

This equation, which


the

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)

by the coordinates of any point P on

therefore the equation required.

(See also Ex. 15(/),

ELEMENTARY COORDINATE GEOMETRY

15.63]

15.63 Tangent at

Let

Px (xx

yx )

the tangent at

be a given point on the

circle

(ii)

perpendicular to the radius

is

+/) - y\ -fyx + x(x x + g)-x\- gxx

xxx + yy x + gx +fy

i.e.

Since

Then since
and CPX has

in 15.61.

CPX

gradient (y1 +f)l{x 1 + g), the equation of the tangent

i.e.

587

is

x\ + y\ + gxx +fy x

0,

Px lies on the circle,


x{ + yl + 2gx 1 + 2fy 1 + c

and so the equation of the tangent at

0,

Px can be written

= ~ 9xx ~fVi - c

xxi + VVi + Qx +fV

xx x + yyx + g{x + x x ) +f(y + y1 ) + c

i.e.

>

(iv)

0.

Remark. This result can be remembered by the following rule of


alternate suffixes. 'Write xx for x z ,x + x for 2x, etc., in the equation of
the

circle:

xx + yy + g(x+x)+f(y + y) + c

and attach the

0;

suffix 1 to alternate variables.'

It should be emphasised that this rule

and

is

only an aid to memory,

certainly does not constitute a proof that (iv)

the tangent at

Other applications

will

is

the equation of

appear in later chapters.

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

Chord of contact from P,

x is

outside the circle, two tangents can be

their points of contact are

P2 P3
,

chord of contact of tangents from

Px if
P2 P3 is called the

drawn from

then the whole line

Px We now find the equation of this


.

chord.
40

GPM II

ELEMENTARY COORDINATE GEOMETRY

588

For simplicity we consider the


the origin. The tangent

Since

P2 P

at

centre

is

nas equation

xx z + yy 2

Px lies on this line,

= 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

which shows that the coordinates x

equation

i.e.

that

P2 lies on the line xxx + yyx =

2
.

Pi

Fig. 156

Similarly,

chord

P2 P3

P3 lies on this same line, which must therefore be the


Hence the chord of contact from Px to x 2 + y 2 = a 2 is
^1 +

2/2/i

= a2

The same argument applied to the general


x

circle

+ y + 2gx + 2fy + c =
2

shows that the chord of contact from

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.

Remark. The equation xx x + yyx = a 2 has two quite distinct


meanings according to whether Px lies on the circle x 2 + y 2 = a 2 or
outside it. When Px lies on the circle, it represents the tangent at Px
;

when Px lies

outside,

it

represents the chord of contact from

remarks apply to the general

Px

Similar

case. See also (1), (2) in 15.65.

Examples
(i)

Find

the equation of the tangents from

The chord of contact from O

(fig.

to

x 2 + y 2 + 2gx + 2fy + c

157) has equation

gx+fy + c =

0.

0.

15.65]

As

ELEMENTARY COORDINATE GEOMETRY

589

in 15.54, consider the equation-

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 line-pair. Hence it represents the pair of tangents from 0.

Fig. 158

Fig. 157

Pair of tangents from Px to x 2 + y 2 = a 2


When Px is given, let P a be any other point in the plane. Any point on P X P 2

(ii)

will divide it in

some ratio k

I,

and hence

/ lx x

This point will

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

+ yl-a 2 )(xl + y 2 -a 2 = (x^ + y^-a 2


)

This equation shows that the point


(x

+ y -a
2

2
)

(x

2 (flJ 2

+y - a
2

2
)

Vz)

2
.

nes on *ne locus

(xx x

+ yyx - a 2

2
)

which must therefore be the combined equation of the tangents from


i.e.

of the pair of tangents. See also 15.73, ex.

x,

(i).

15.65 Examples; polar


(1) Tangents at the intersections of the
through P x meet on the line xxx + yyx = a2

circle

x2 + y 2

a 2 with variable chords

40-2

[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

and this shows that P 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

that part of the

is

on

lies

xx x + yyx

the line

AP

PP

k'

the complete

2 is

P meets the circle x* + y = a at A, B; P is


AB in the same ratio (one internally, the other extern-

Since by hypothesis
1 :P 1 B
so that A, B divide the line X

say

outside, the locus of

outside the circle.

variable chord through

chosen so that
ally), f

the locus of

circle,

entirely outside the circle. If


line

AP

2
.

:P 2 B, hence also P X A :^4P2 = P1 P: PP 2


internally and externally in the same ratio,
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

has coordinates of the form

+ ky 2\
+ k )'
divide AB harmonically.
lyi

ELEMENTARY COORDINATE GEOMETRY

15.66]

and

lies

are

k'

on x

+y =
2

2
;

hence equation

(v) of 15.64, ex.

xx x2 + y x y 2 -a 2

Hence

P2 lies on the line xx x + yx y a 2

Px

If

lies inside

the

which

lies

within the

line

0.

the whole
the locus of

2 is

which

line,

lies

com-

that part of the

2 is

circle.

xx x + yyx

a2

is

called the pole of the line

x is

lies outside,

We now unify the complementary results in (1),

(3) Definitions.

The whole

and its roots

the locus of

circle,

pletely outside the circle. If


line

holds,

(ii)

sum is zero,

V. Since their

591

called the polar of

= 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.

Pt lies on the circle, the polar coincides with the tangent at P x


The polar exists for every point except the centre (0, 0).
If

Reciprocal property. If the polar of x passes through 2 then the polar


passes through
(For if xxx + yyx = a 2 passes through (x if y 2 ), then
x
x 2 xx + y 2 y x = a2 , which shows that x lies on xx z + yy2 = a 2 , the polar of 2 .)
(8)

of

15.66 Orthogonal circles


(1)

The angle between two intersecting circlesf is the angle between

common point.

their tangents at a

(It is

easy to prove geometrically

that the angles at the two intersections are equal.)


If these tangents are perpendicular, the circles are orthogonal. In
this case the tangent to
(2)

one

a radius of the other.

circle is

Condition for

x 2 +y 2 + 2gx + 2fy + c

0,

x 2 + y 2 + 2g'x + 2f'y + c'


to be orthogonal.

The centres are C{ - g, -/) and C'{ - g', -/'),


and the radii are <J{g 2 +f 2 -c), V(^' 2 +/' 2 -c').

Fig. 163

If the circles cut orthogonally at P, the triangle

angled at P, and so
i.e.

(9

Conversely,

(9

+P -c) + (g'

2gg' + 2ff'

when this

condition
g

is

cp2 +

- g'Y + (f-f) 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-

ELEMENTARY COORDINATE GEOMETRY

592

to both sides the previous equation

is

The condition

is

by the converse

obtained; hence

of the theorem of Pythagoras, triangle

C'PC is

therefore both necessary

[15.66

right-angled at P.

and

sufficient for ortho-

gonality.

Example
Prove that in general

there is only one circle orthogonal to three given circles.

If the given circles are

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//i-c =

=
=
-c
+
2^3 2//3

299i + 2fA- c

and

cx ,
c 2
c3 .

If the determinant

A=

91

92

fi

9z

fz

non-zero, 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

through O whose centre


[Use tangent _L radius '.]

circle

0.

is (p, q).

'

2 Show that the chord of x 2 + y 2 = a 2 whose mid-point is (x x yx ) has equation


xxx + yy x = x\ + y\. [If P is any point of the required chord, OP x _L PP\.~\
,

3 Prove that the mid-points of those chords of x 2 + y 2 + 2gx + 2fy + c =


which pass through Px lie on the circle (x x 1 )(x + g) + (y y 1 )(y+f) = 0.

[CMCP
4

(i)

1 .]

PxP 2 of x 2 + y 2 = a 2 Deduce the


Also find the tangent at Px by using Calculus.

Find the equation of the chord

equation of the tangent at P x

(ii)

= n touches x + y 2 = a 2 find its point of contact. \xxx + yyx = a 2


+ my = n are the same line if x x jl = y x /m = a 2 /w.]
6 Find the condition for y = mx + c to touch x 2 + y 2 = a 2 and deduce that
the lines y = mx a*J(l+m 2 touch the circle for all values of m.
5

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 right-angle 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

ELEMENTARY COORDINATE GEOMETRY


(A, 0), its equation is y = m(x A) where m is chosen so

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

Prove that the direct

r\

have equation

r2 )x (a^ a 2 rx )} 2 = {(%

{("].

(rx

2
2)

2
2)

}y

2
.

Find the equation of the transverse common tangents.


10 If the line of centres of
x* + y*

+ 2gx + 2fy + c =

cuts the axes at A, B,

x*

0,

show that the

+ y* + 2g'x -r 2fy + c' =

circle

on diameter

AB is

(g-g')U-S')^+y 2 = (gf'-rfHig-rtx-V-Dy}*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
)

through another fixed point. [Use oblique axes.]


12

A circle passes through (h, k) and cuts orthogonally the circle


=

x2 + y* + 2gx + 2fy + c
Prove that

its

centre

lies

on the

0.

line

2(h + g)x + 2(k+f)y

fc

+ F-c.

13 Find the equation of the circle orthogonal to x 2 + y 2 + 2x 2y + 1 =


# a + y 2 + 4x 4y + 3 = and whose centre lies on the line 3x y 2 = 0.

and

14 Find the equation of the circle orthogonal to

x2 + y2
15

5,

x 2 + y 2 + 6x+l

Prove that an angle between the

x* + y* + 2gx + 2fy + c

(supposed to intersect)

is

COB 6

given

x i + y 2 -<ix-4ty + l

and

0.

circles
s'

0,

x 2 + y 2 + 2g'x + 2f y + c'

by
2gg' +

2V(fl'

+/

2ff'-c-c'

-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.)
.

orthogonally, prove that it also cuts


cuts s = 0, s' =
17 If a circle or =
orthogonally each member of the system s + As' = 0. Interpret the case A = 1
by first showing that the centre of a lies on the line ss' = 0.

18 Prove that the length of a tangent from


is

to a 2 + y % + 2gx + 2fy + c

V(a* + y\ + 2fla?i + 2/yx + c).

such that the tangents from it to the circles s,


19 (i) Prove that a point
are equal (notation of no. 15) lies on the line 2(g g')x + 2(ff')y + (c c') =
i.e.

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

P is only that part of this line which is outside both circles.

ELEMENTARY COORDINATE GEOMETRY

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 + ly-12 = 0.

Conies

15.7

15.71 Definitions

The

P in a plane such that the ratio of its distance

locus of a point

from a given point 8 to

its

distance from a given line d

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

the foot of the perpendicular from

to the line d, the

definition is equivalent to the relation

SP = e.PM.
The name

'conic'

is

an abbreviation

for 'conic section'. Originally

the ellipse, parabola and hyperbola were obtained as the sections made

by a plane drawn perpendicular to any one generator of right circular


cones with acute, right and obtuse vertical angles respectively. Later
they were defined by different sections of the same right circular cone,
the ellipse, parabola and hyperbola being respectively the curves of
section by a plane making an angle with the base less than, equal to,
or greater than that made by the generators. If the cone is a double
one, the hyperbola will consist of

A plane parallel to

two separate parts or branches.

the base will give a circular section; one through

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

line-pair,

It can
bolic, or

and

a single point. Hence with this approach, a

a single

line,

and a

circle,

single point are all conic sections.

be provedf that, for each plane which gives an

elliptic,

para-

hyperbolic section, there exists in that plane a fixed point

8P = e.PM;

i.e.

any point P on the curve satisfies the law


these 'conic sections' have the focus-directrix

a fixed line d such that

properly and are therefore conies in the sense of our definition.


'

We shall not

'

do so in

this book.

ELEMENTARY COORDINATE GEOMETRY

15.72]

The equation of every conic

15.72

If the fixed point

the definition

SP =
2

(x

S
e

PM

of the second degree

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

this is clearly of second degree in


In our definition we made no reference
to the relative positions of 8 and d. In

we suppose 8

is Ix

becomes

When simplified,

particular, if

595

lies

on

d,

x, y.

we

choose d for y-axis and the perpen-

dicular at

for aj-axis. Then, if P(x, y)

any point on the

8P 2 =

e2

.PM 2

x2 + y 2

According as

is

expressed by

e2x 2 , i.e.
e

is

locus, the definition

$ 1, this

(e

l)x 2

Fig. 164
.

equation represents a line-pair through 8,

the (repeated) fine y = 0, or the single point #(0, 0).


Hence, consistently with our definitions, we must regard an intersecting line-pair as a degenerate hyperbola (since c
single fine as a degenerate parabola

as a degenerate ellipse

(e

<

1 for

(e

for both),

>

for both),

and a single point

both). These conclusions agree with

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.

We now turn to the more difficult converse problem of interpreting


the general second-degree equation

ax2 + 2hxy + by 2 + 2gx + 2fy + c

by showing that

We

forms.

can be reduced to certain standard


prove that every equation of the second

this equation

shall eventually

degree represents a conic (including the degenerate cases just mentioned), a parallel line-pair,

15.73
(1)

we

circle,

or nothing.

Change of coordinate axes

When a locus is specified by some property (such as SP = e.PM),

express this as an equation between the coordinates of any point

(x, y)

on the locus, referred to a pair of coordinate axes. Some choices


(e.g. those having regard to symmetry of the locus) will lead

of axes

ELEMENTARY COORDINATE GEOMETRY

596

when an equation wo one

to a simpler equation than others. Hence,


set of axes is given, it is often desirable to

simplify the equation. This can be done

changing the

(a)

(a translation

[15.73

change the axes in order to

by

origin, leaving the directions of the

axes unaltered

of axes)

changing the direction of the axes, leaving the origin unaltered

(6)

(a rotation

of axes);

both together.
Remark. When a problem involves equations of two or more
(c)

in general

way;

it is

loci,

possible to simplify only one of the equations in this

e.g. see 16.12, ex.

(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_

change the origin

h}

to the

y>

= y-k

point

(h, k)

we

+ h for x
Oy isf(x, y) =

substitute x'

and y' + kfor y: the locus whose equation wo axes Ox,


becomes f(x' + h, y' + k) = wo axes O'x', O'y'
We usually omit the dashes in the new equation, and say that
becomes f(x + h,y + k) = 0.
f(x,y) =
The above work remains valid for oblique axes an example occurred
;

in 15.16(2).

Examples
(i)

Pair of tangents from Px to x z + y 2 = a 2 (Cf. 15.64, ex. (ii).)


Changing the origin to P^x^y^, the equation of the circle becomes
.

= a\
x* + y* + 2x1 x + 2y1 y + (xl + yl-a 2 =
(x+x1 )* + (y + y 1

i.e.

0.

The equation of the pair of tangents from the new origin is


(xl

+ yl-a*)(x* +

y*)

= fax + ytf)*.

(see 15.64, ex.

(i))

ELEMENTARY COORDINATE GEOMETRY

15.73]

597

Referred to the original axes, this equation becomes

+ 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

Point of intersection of the line-pair ax 2 + 2hxy + by2 + 2gx + 2fy + c

(ii)

Suppose the
becomes

lines

meet at Px (x x yx ). Changing the


,

a(x + x x ) 2 + 2h(x + xx )(y + y x )

ax

i.e.

origin to

x,

+ b(y + yx 2 + 2g(x + xx + 2f(y + yx + c =


)

0.

the equation

+ 2hxy + by + 2(ax x + hyx + g) x + 2(hxx + byx +/) y


+ (ax 2 + 2hxx yx + by2 + 2gxx + 2fyx + c) =

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

The first two equations deterrnine Px As in Remark (/?) in


.

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,

then from triangle

r cos (6

x'OP =

and

<f>.

OPN,

+ 0) = r cos cos r sin


= x' cos 6 y' sin 6
(f>

sin $

ELEMENTARY COORDINATE GEOMETRY

598
since

from triangle OPN',

Thus,
for

r sin (0

to rotate the

<j>)

= rcos^ and y' =


=

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

axes through angle

+ y' cos

x and x' sin

x'

[15.74

for y: the locus f(x, y)

y sin 0, x sin + y cos 0) =

0.

The reverse transformation can be obtained


either by solving the above two equations for x', y', in terms of
y;
or by writing 6 for 6 and interchanging (x y'), (x, y) in the
r

above equations;
or

by considering

x'

r cob {{d

+ <!>)- 6} =

y'

Both transformations can be read

= rsm{(d + <f>)-6} =

off from the

....

scheme

y'

x'

cos 6

sin 6

sin 6

cos 6

Change of both origin and direction of axes. By combining the


substitutions found in (2), (3) we see that, under the most general
change of rectangular axes, the locus whose equation is f(x, y) =
becomes
,

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

substituted the linear function x cos 6 y sin 6 + h,


to powers

and products of

linear

and similarly for y;


functions correspond powers and

products of the same degree.


15.74 Reduction of 5 =

to standard forms

By rotating the axes through angle 6,


s

==

the equation

ax 2 + 2hxy + by 2 + 2gx + 2fy + c

(i)

becomes
a(x cos

y sin 6) 2 + 2h(x cos y sin 6) {x svnO + y cos 6)


+ b(x sin 6 + y cos 6) 2 + 2g(x cos 6 y sin 0)
+ 2f(x8in.0 + yco80)+c =

0,

(ii)

ELEMENTARY COORDINATE GEOMETRY

15.74]

in

which the

coefficient of

xy

is

- 2a cos sin + 2A(cos 2 - sin 2 0) + 2b cos


= 2(6-a)sin0 cos + 2ft cos 20
= (6 - a) sin 2d + 2h cos 26.
If 6 a

4= 0,

this will

be zero when

tan 20
if b

-a =

\ir

takes the form

sin

satisfies
*

2ft

be zero when cos 20

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)

can be written in one of the forms

a2
according as

can

pi

Q%

(iv)

+B y + B) =A + B-[
(

4= 0,

is so known.

then by completing the square,

By changing the origin to the point - G\A, -F/B),


Ax2 + By2 = X,
where A = G 2 \A + F 2 jB - C.
If A

(ii)

now arise.

and

4=

the

Giving

satisfying the appropriate condition, equation

where A,B,C,F,G are known in terms oia,b,c,f,g,h since

The following

599

A /A,

/?

a2

'

I?/A are

/?

'

a2

both positive, of opposite sign,| or both

negative.

// A

0, (vi)

can be put into one of the forms

a 2x2 - fity2 =

0,

a 2x 2 + [S*y* =

B have opposite or the same signs.


and # 4= 0, equation (iv) can be written
If A = 0, B
F2 \
F\
C
/
I

according as A,
(b)

=}=

Changing the origin to the point

(_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

ELEMENTARY COORDINATE GEOMETRY

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,

parallel lines, a repeated line, or nothing,

forming equation

(iv)

the cases considered in

We

becomes

(iv)

By 2 + 2Fy + G =
which represents two

....

(vm)

cannot have

F+

0,

or */ j?

=^=

0,

A*

0,

by rotating the axes through

then by transtt,

we obtain

(6), (c).

A =B=

would not be of second


and hence (see 15.73 (4)) neither would (i). Thus all admissible
possibilities have been considered.
It has now been shown that, by a suitable change of axes, the
0,

for then (iv)

degree,

general equation

(i)

of second degree can be reduced to one of the

following standard forms:

^+^ = -1,
2/

ccx*+]3y*

0,

= aV(a 2 -/?).

Clearly the fourth equation represents nothing; the fifth represents

an intersecting

line-pair,

a repeated

or a single point;

line,

and the

sixth represents a pair of parallel lines, a repeated line, or nothing.

The
first

following three chapters are respectively concerned with the

three

loci,

which

will

be proved to be conies in the sense of the


'

'

definition given in 15.71: see 16.11, 17.12

and

17.13.

Exercise 15(e)
1

By

a change of

origin, the points

1,3),

(4,

2) become

(a, 5), (3,/?);

find a, ^.

2 Find the new equation of the locus x 2 xy 6y 2 Zx +14y 4


is changed to (2, 1).

when

the origin

3 Find the new equation of the locus x 2 y 2


through angle \ti.

a 2 when the axes are rotated

4 By rotating the axes through angle a, show that p in the equation


x cos oc + y sin a = p is the length of the perpendicular from O to this line.
5 Find through what angle the axes must be rotated so that the equation
2
2
2
2
7a; + 4xy y =l becomes of the form ax + by = 1.

ELEMENTARY COORDINATE GEOMETRY

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

- 48xy + 51y 2 + 86a; - 48y + 1 8 =

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

the choice of coordinate axes.


x'

and

x'dA

Thus

similarly for y'.

= xcoad + yaind + h,

(xcoad + yaiad + h)dA


(x' ,y') is

the point

(x, y)

referred to the

new axes.]

Miscellaneous Exercise 15(/)


1

If

the points

M3 + M1 + M2 =

{at\,at\),

{at\,at\),

Conversely,

0*

are

(at\,at\)

show that when

collinear,

prove that

this condition is satisfied the

three points are collinear. [Use theory of equations.]

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

= 0, xcosSfi + y acos/tf = 0, a;cos dy + y acosy =


= 0, and prove that this condition is also sufficient.
is
4 If the line-pairs ax 2 + 2hxy + by 2 = 0, a'x 2 + 2h'xy + b'y 2 = have a line in
common, prove (ab' -a'b) 2 + 4(ah' -a'h) (bh' -b'h) = 0. [Use 10.42(1).]
is perpendicular to one of the lines
5 If one of the lines ax 2 + 2hxy + by 2 =
a'x 2 + 2h'xy + b'y 2 = 0, prove (aa' -bb') 2 + {ah' + b'h) (a'h + bh') = 0.
and x 2 + 2qxy y 2 = are such that one
6 If the lines x 2 2pxy y 2 =
pair bisects the angles between the other pair, prove pq = 1.
form an isosceles triangle,
and ax 2 + 2hxy + by 2 =
7 If lx + my + n =
prove that h(l 2 m 2 = (a b) Im. Find also the further condition if this triangle
arcos3a + 2/ acosa

cos a + cos /? + cos y

is equilateral.

[The line is parallel to an angle-bisector of the pair, so its gradient

Z/m must satisfy {1 (y/x) 2}/(a b) = (y/x)/h.]


represents
8 If ax 2 + 2hxy + by 2 + 2gx + 2fy + c =

a pair of lines meeting at


A, and parallel lines are drawn through O to meet these at B, G, find the equations of the diagonals OA ,
is a square, prove a + b =

BC of the parallelogram so formed.

If the parallelogram

and h(g2 -f 2 - fg(a - b). [OB J_


9 Prove that the equation m(x 3 3xy 2 + y 3 Sx 2y =
)

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
)

10 If A, B, G are fixed and PA 2 + PB 2 + PC2 is constant, prove that the locus


of
is a circle whose centre is the centroid of triangle ABC. [Choose O at this

centroid; then
11

HxA =

4x-3y + 2a =

12

2y^.]

Find the centre and radius of the

incircle of the triangle

whose

Bx- 4y+ 12a = 0, 3 + 4y- 12a =


Prove that for all constants A and p, the circle
{x-a)(x-a + \) + {y-fi)(y-/3+p) = r2
0,

0.

bisects the circumference of (x cc) 2 + (y fi) 2 = r2


Find the equation of the circle which bisects the circumference of
.

x2 + y2 + 2y-S =

and touches the

line

xy

at the origin.

sides are

ELEMENTARY COORDINATE GEOMETRY

602
Show

13

that the line (x a) cos 6 + y sin 6

= r touches the circle


2
2 =

{x
a) + y
r2
,

and

state the coordinates of the point of contact.

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;

the equation of the line-pair through


Ox. Explain why the locus

and

which

is

(ii)

(x-x 1 )(x~x t + {y-y1 )(y-y 1 =


)

passes through the vertices of the rectangle formed


deduce that this locus is the circle on diameter PiP2

by these

line -pairs,

and

16 If the orthogonal circles

x 2 + y 2 + 2g1 x + 2f1 y + c1

have centres A,

x 2 + y 2 + 2g2 x + 2f2 y + c 2

0,

B and cut at C, D, prove that the circle through A, B,

2(x

+ y + 2( 9 1 + g 2
2

If the equation of the circle

x + 2(/x +/2 ) y + (c x + c 2 )

on diameter

= r\\AB 2 where r

is

D is

0.

CD is written in the form

x 2 + y 2 + 2gx x + 2f1 y + c1 + X{2( gi -g z )x + 2(/x -/2 ) y + (c x - c 2 )}


prove A

G,

the radius of the

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

Show that the

line -pair joining

to the meets A,

= 1 has equation
(a I 2 x 2 2lmxy + (6 m 2

B of the line Ix + my

with the conic ax 2 + by 2

If

0.

A OB is a right-angle, deduce that AB touches the circle (a + b) (x + y =


2

19 Find the equation of the line -pair joining

circle x 2 +y 2 + 2gx + 2fy + c = 0. Hence


circumcentre of the triangle formed by Ix + my =

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.

vary, but remain equally inclined to the


on a fixed line through O.

show that the circumcentre

varies

20 Explain why the equation


(x 2

+ 2hxy + y 2 + (?(x+fty)(x + y+l) =


)

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

the triangle formed

by the

lines

ax 2 + 2kxy + ay 2

Obtain the equation of the lines joining


x y b with the curve x z + y 3 = 3axy.
21

0,

x y+1

0.

to the points of intersection of

603

16

THE PARABOLA
The locus y2 = 4ax

16. 1

16.11 Focus-directrix property

Since the equation y 2

4ax can be written in the form

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

symmetrical about Ox because

the equation

is

y by y; Ox

unaltered

is

by replacing

called the axis of the

O is the vertex. Without


generality we may always

parabola, and
loss

of

suppose a >
all

0;

a;

then positive for

points on the curve.

The parabola

= and hence y =

meets Oy where x
it

is

therefore touches

Oy at 0, and Oy is
Fig. 167

called the tangent at the vertex.

The chord LL' through S


angles to the axis
it

is

at right-

called the latus rectum; its equation is

cuts the curve at the points

(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.

16.12 Parametric representation

For any point

(x, y)

ofy 2

2x

so that y
41

2at

and x =

4ax other than

(0, 0),

at 2 ; (0, 0) is also

given thus

when

GFMII

0.

604

Hence each point on y


for just

one value of

THE PARABOLA
= kax has coordinates
Also each value of

t.

[16.12

of the form

(at

2
,

2at),

determines exactly one

The equations

point of the curve.

at 2

2at

(i)

therefore give a proper parametric representation (see 1.61); they

can also be written

The point P(at 2


t

:2t

....

1.

(n)

2at) is briefly referred to as the point

the parameter of P.

is

9
2

x:y:a =

of the curve;

By using these coordinates, the condition that

P lies on the curve is automatically satisfied without reference to the


cartesian equation;

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

This cuts the curve at points

for

l(at 2

a) + my =

0.

which

-a)+m.2at =

0.

be the chord t^ if the roots of this quadratic in t are tx and


product of the roots is 1, the required condition is tx t^ = 1.
This necessary condition for a focal chord is also sufficient. For
then tx and t 2 are the roots of a quadratic equation of the form
It will

pt 2 + 2qt-p

which shows that the points


clearly goes through the focus
(ii)

tx ,

lie

t2

on the

line

px + qy ap =

0,

and

this

A circlef

which

satisfy J

0,

aHl + 2a(2a + g) t* + 4aft + c =

0.

aH* + 4aH 2 + 2agt 2 + 4aft + c


i.e.

= 1,

0,

x* + y* + 2gx + 2fy + c
t

if

Since the

(a, 0).

Concyclic points on the parabola.

cuts the parabola at points

<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

Suppose there are four intersections, given by

numbers are the roots of the above

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 +

give concyclic points.

Exercise 16(a)
'

Find the gradient of the chord


t

is

from the directrix

is

If

is

4ax.

t2 .

one extremity of a focal chord PQ, find the length of PQ.


the mid-point of a focal chord PQ, prove that the distance of

2 If the point
3

'The parabola'' in this exercise means the curve y 2

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 mid-point of QR lies on the parabola y 2 = 2a(x + a).
5 The chord PQ subtends a right-angle 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

two common points


t1

(iii)

If there

is

only one

+ t2 =

2m/l,

tx

and

2,

t1 t 2

prove that

n/al.

common point, prove that am 2 =

In,

and conversely.

the parabola at A, B, C, D. Show that the chords


are equally inclined to the axis. [Use no. 1 and ex. (ii) in 16.12.]

circle cuts

9 Prove that the locus of the centre of a


x 2 + y 2 2ax =
is the parabola y 2 = ax.
for a suitable S and a*.]
SP =

PM

10

circle
[If

AB,

CD

touching Oy and the circle


is the centre, show that

A variable circle passes through a given point A and touches a given line

Prove that the locus of

its

centre

is

directrix.

16.2

Chord and tangent

16.21

Chord

I.

a parabola, and identify the focus and

P P2
t

The gradient of the chord is


Vi-y* = 4a(yi-y2 ) = a
y\-y\
yi+yt'
41-2

THE PARABOLA

606

since y\

4axx and y\

4ax 2 The equation of the chord is therefore


.

yVi =
This equation
(x2

2 ).

is

To make it

x-,).

Ix

(1)

not symmetrical in the pair of coordinates (x x y x ),


,

of fractions:

so, first clear

2/(/i

By using the

[16.22

+ 2/2) ~ 2/i - 2/i2/2 = 4 - 4^1-

condition y\

ax x again,

this simplifies to

lax - (yx + y 2 )y + yx y2

0.

(ii)

Remark. To obtain (ii) we have appealed once to the condition


= 4ax2 that P2 lies on the curve, and twice to y\ = 4axx Contrast
the directness of the work in 16.22, where the use of parameters
automatically ensures that Px P2 lie on the curve.
For another way of proving (ii), see Ex. 16 (6), no. 1.
y\

Chord

16.22

tx t2

Since the gradient of the chord

(1)

is

2at x -2at 2
at\

at\

'

the chord has equation


2

y-2atx = 77- (x-atl),


t,

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)

This line cuts the curve at points t for which


lt

Since the fine

0.

the theory of quadratics) let the required

(2) Alternatively (using

chord be| Ix + my + a

and t 2

l~r t 2

and

\\l.

Tfb

must have roots

Hence

^-

t t2
x

tx t2

and the required chord is


x
t x t2

which

is

equivalent to

1 tx

tx t

y+a

0,

(iii).

f There is no loss of generality in taking the last term to be


of a line contains only two independent constants.

a, since

the equation

THE PARABOLA

16.23]

the numbers

(3) Alternatively, since


t

for

This represents the required chord


t

given by

16.23 Tangent at

(iv), i.e. t

the tangent at

x is

to substitute

because

tx t2

and

2t

iv )

it

cuts the curve at

For curves other than the

the points

0,

use the parametric equations x y a = t 2


and thereby construct the linear equation

x - Wi + h)y + at ih

to

are the roots of

-{t x + t 2 )t + t x t 2

we may
2

x,

607

circle,

the tangent

the limit of the chord

is

defined as a limit:

PX P2 when the point P2 tends

along the curve; but see 16.25.

The tangent at Px to y2 = Aax


equation (ii) by letting y2 -> y v and is
(1)

therefore obtainable from

is

4ax-2y1 y + yl =
Since y\

0.

4axv this can be written

yyx

2a(x + x1 ).

(v)

Observe that this can be written down from the equation y 2 = ax


by applying the 'rule of alternate suffixes' (see 15.63, Remark).
the gradient of the tangent at

(2) Alternatively,

calculating dy/dx

can be found by

when x = x x y = y x Thus from y =


2

2y

and the tangent

at

dy
dx

4a

dy
dx

and

4ax,

2a

=
y

Px has equation
2a.

y-yi = (x-xi)>
which
(v)

is

equivalent to the limit of

(i)

when y2 -> yx and reduces


,

to

above.

16.24 Tangent at the point


(1)

Letting

Omitting the

->tx in equation

suffix,

(iii),

we

the tangent at the point

x-ty + at 2 =

0.

obtain x tx y + at\=0.
t

is
(vi)

THE PARABOLA

608
(2) Alternatively,

[16.25

the gradient can be obtained by calculus:

dy
dx

_ dy jdx =
=

2a

dtj dt

2ai

and then the equation written down


y 2at

l'

as

^(x at 2 ).

Geometrically, the parameter

the cotangent of the angle

is

made by the tangent-line at the point t with the ic-axis


16.25 Tangency

for tan i/r

ijr

1/t.

and repeated roots

For a curve whose equation

is

a polynomial in

the simul-

(x,y),

taneous solution of this equation and that of the chord


,

P P2 leads to
1

equations for x and y containing the factors


(x

respectively, or to

- xx

(x

- x 2 ),

(y

- yx

(y

- y2

an equation in a parameter

containing

(*-*i)(*-f.).t

Since the equation of the tangent at

the chord

P1 PZ

when

tends to

equation of the tangent

is

will give equations for x,

or an equation for

P is the limit of the equation of


x

along the curve, hence

if

the

solved with the equation of the curve,

y containing the factors

containing

(t

tx

(x

xx

2
)

(y

yt

it
2
)

2
)

Thus tangency corresponds to repeated roots, and is often dealt with


thus rather than by direct appeal to the limit definition. For example,
the line Ix + my + n = meets the parabola where alt 2 + 2amt + n = 0,
and will be a tangent if and only if this quadratic has equal roots,
i.e. if

am 2 =

More

In; cf.

Ex. 16(a), no.

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)

of 16.12, the limiting circle will touch the parabola

4 other hmiting cases can


be considered similarly (see Ex. 16 (6),nos. 16-18). See also Remark (a)

at the point t x

and cut it again at points f3

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

which has no polynomial

factors.

THE PARABOLA

16.26]

609

16.26 Examples
(i)

The tangents

at

t lt t 2

meet at the point

{at 1 t 2 , a(t

2 )}.

This result can be verified by direct solution of the equations

x hy + atl

x t 2 y + at\

0,

Alternatively, these two equations express that


quadratic in t
2
at

y
a

so that

-ty+x =

d.

and

are the roots of the

0,

+ tz and

tt

tx t 2 ,

giving the required intersection (x, y).


The coordinates can be remembered by an extension of the rule of alternate
suffixes applied at at*, 2at.
'

(ii)

on

The

'

orthocentre\ of the triangle formed by three tangents to

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^,

and this line meets the directrix x = a where y = a^ + ^ + t^ + t^t^. The


symmetry of this result shows that the other two perpendiculars meet the
directrix at this same point, which must therefore be the orthocentre of the
triangle of tangents.
(iii)

(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.
(mc-2a) 2

mc

2 2
,

i.e.

Hence for all m


Alternatively,

4= 0,

the line

mx y + c =

tangent at some point

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.

f The orthocentre of a triangle is the point of


drawn from the vertices to the opposite sides.

intersection of the perpendiculars

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)'

from which the equation for m follows.


Taking r 2 = \a 2 the equation becomes
,

m + m -2 - 0,
i

m=

so that

1.

The common tangents

x + a,

y
Tangents from

(iv)

(m 2 -

i.e.

x ;

1)

(m 2 + 2)

0,

are thus

x a.

the chord of contact.

given point P x at most two tangents can be drawn to the parabola.


For the tangent at the point t will pass through (x lf y x ) if

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)

>

Pair of tangents from Px

The condition

for the point


l lx x

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 =

(JoachimsthaVs ratio equation for the parabola).


The argument in 15.64, ex. (ii) shows that the pair of tangents from
equation
2
2
2
{y 1 y-2a(x1 + x)} = (y -4ax 1 )(y -4ax).

has

Exercise 16(6)
1

Verify

that,

when

simplified, the equation

{y-yi)(y-y*)
is linear

Deduce

in x and y ; and that

it is satisfied

the equation of the chord

2 Deduce from equation


if

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.

3 Obtain the equation of the tangent at the point


quadratic equations.

by using the theory of

4 Prove that the chord whose extremities are given by the roots of
= has equation ux + vy + aw = 0.

ut2 + 2vt + w

t The set of points (xv yi) for which y\


of the parabola.

> baxx may

therefore be called the outside

THE PARABOLA
5 Find the equation of each of the tangents from

611
(

2a, a)

to y

4ax.

on the parabola is PN, and the tangent at


6 The ordinate of the point
meets the axis at T. Prove that the vertex bisects TN.

7 If the tangent at P meets the axis at T, prove SP = ST.


is the perpendicular from a point P of the parabola to the
8 (i) If
directrix, prove that the tangent at P bisects angle SPM. [Use no. 7 and pure

PM

geometry.]
(ii)

light at

Deduce the property of the parabolic

from a source of
[Angle of incidence

reflector: rays

S leave the surface in a beam parallel to the axis.

angle of reflection.]

9 If the tangent at

P meets the directrix at Z, prove SZ _L SP.


and QU is parallel
UP

is the tangent at P,
10 PQ is a variable focal chord;
to the axis. Find the locus of the mid-point oiPU.
11

Show that any one


(i)

(ii)

(iii)

tangents at
tangents at

12 If P,
triangle

13

of the following statements implies the other two.

PQ is a focal chord;

(i)

P and Q are perpendicular;


P and Q meet on the directrix.

are the points

x, t

v and tangents at P, Q meet at T, prove that


If this area is always 4a 2 and the locus of T.

TPQ has area \a2 (t 2 t^) 3


If the chord

tx t %

subtends a right-angle at the point

-M a

+4 =

t,

prove that

0.

(ii) If the circle on chord PQ as diameter cuts the parabola again at H


and K, prove that the chords PQ, HK make equal angles with the axis.
14 A circle cuts the parabola at A, B, C, D; tangents at A, B meet at T, and
tangents at G,D meet at T'. Prove that the axis bisects TT'. [Use 16.12, ex. (ii).]
15 A circle passes through the vertex of the parabola and cuts the curve
again at A, B, G. The tangents at B, G meet at T. Find the ratio in which AT
is divided by the axis.
16 (i) If a circle touches the parabola at A and cuts it again at G, D, prove that
the tangent at A and the chord GD are equally inclined to the axis.
(ii) If circles touch a parabola at a fixed point, prove that the
chords not through this point are parallel.

common

*17 (i) If G ->


in no. 16 (i), then by 8.42 (2) the limiting circle is the circle
is the point t lf prove that the remaining
of curvature of the parabola at A. If
has parameter 3 x What can be said about the inclination of
intersection
and the chord AD1
the tangent at
circle cuts the parabola at A, B, C, D. If the circles of curvature at
(ii)
these points cut the curve again at A', B' , G', D' respectively, prove the latter

points are also concyclic.

*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 mid-point 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

*23 Prove that tangents at the extremities of a variable chord through Px


meet on the line yyx = 2a(x + x1 ). [Method of 15.65(1).]
*24 A variable chord through Px meets the parabola at A, B; P2 is chosen on
it so that Px and P 2 divide AB (one internally and the other externally) in the
same ratio. Prove that P2 ^ es on *ne lme VVi 2a(x + x1 ). [Method of 15.65(2);
use the ratio quadratic in 16.26, ex.

(v).]

*25 Defining the polar of x wo y % = 4a* to be the line yy x = 2a(x + x t ), prove


that if the polar of x passes through 2 then the polar of a passes through x
What is the polar of the focus?

*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

Since the tangent at the point

has gradient

has gradient

l/t,

the normal there

t. Hence the equation of the normal at t is


y 2at
tx

i.e.

= t(x at 2

+y =

2at + ats

),

(i)

Example
Prove that

the

normal

at

so that

ts,

meets the curve again at the point t2/t.

If the other intersection

is

the point

gradient of normal at

then the normal at

s,

=
t

from which s = t 2/t. Also


16.32

also the chord

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

Since this equation

tx
is

enquire

+y =

cubic in

t,

how many normals can be

The normal
2at

+ t3

at

will pass

through

there are at most three points on the

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.

16(c), no. 15.

Suppose that the normals at the points


(x

).

Then these numbers

is

are concurrent at

are the roots of the cubic

)t-y =

at3 +(2a-x

Since the term in

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 ,

The point of con-

to be conormal.

given by

2a-xQ =

dLtx t2

atx t2 tz

(iv)

The necessary condition T,t x = Ofor conormal points is also sufficient.


In a general way this is clear because only one condition is needed for
concurrence of three lines. In detail, given t x t2 t3 we can define
numbers x and y by equations (iv) above; then, provided
,

h + 2 + t3 =
t

we see that t lt

2 , t3

>

are the roots of the cubic

(ii),

which expresses that

the normal at each of the corresponding points passes through (x y


,

If
(x

two roots of the cubic


)

coincide. If t 2

ts ,

By

8.53, ex.

normals

(i.e.

(i),

the

(ii)

then

).

are equal, then two of the three normals from


t x = 2t 2 and (x
y ) is given by (iv) as

-2a + Zat\,

-2at\.

these equations show that (x ,y ) lies on the envelope of the


of the parabola. From any point on this locus only two

evolitte)

normals can be drawn.


Three normals from a point can coincide only when < x
zero (since 2^ = 0); and then (iv) shows that {x ,y )

distinct

is

<2

is

the point (2a,

t3 , i.e.

when each
0).

Examples
(i)

The

circle

through the feet of three concurrent normals also passes through

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 equation of the circle through the feet of


current at (h,k). (See also 19.7, ex. (i).)
(ii)

The

feet of the three

normals are the roots of the cubic


at s

+ (2a-h)t-k =

0.

(v)

If the required circle is x 2 + y 2 + 2gx + 2fy + c


satisfy (cf. 16.12, ex. (ii))

0, its

aH* + (4a 2 + 2ag) t 2 + laft + c


Since by ex.
given by

The

(i) t

cubics

The required

is

0.

one such meet, therefore


at*

(4a + 2a) * + 4/

meets with the curve

0.
0,

and the other three are


(vi)

now have the same roots, and so


4a + 2g = 2a-h, 4/ = k.

(v), (vi)

circle is therefore

x 2 + y 2 - (h + 2a) x - \ky
(iii)

normals which are con-

0.

Prove that the normal at P x has equation 2a(y y x ) + yx (x xt ) = 0.


Hence prove that the feet of the normals from (h, k) to the parabola lie
curve xy + (2a h)y 2ak = 0.
(a)
(6)

on

the

By

(a)

gradient

16.23 the tangent at

has gradient 2a/ylt so the normal at

yJ2a, and its equation is

has

y
y-yi = - a {x - x J'

Let P x be the foot of a normal drawn from


on the normal at Px

(b)

(h, k) lies

(h, k)

to the curve.

Then since

2a(k-y1 + y 1 (h-x1 =
)

which shows that

lies

0,

on the locus
2a{k-y) + y{h-x)

0.

Exercise 16(c)

PN

is the ordinate of a point


on the parabola, and the normal at
1
= 2a.
the axis at G. Prove that

P cuts

NG

2 The tangent and normal at


SG.

meet the axis at T,

G respectively.

Prove

T8 =

3 The line through the


axis meets the normal at

mid -point

P in

of a focal chord
V. Find the locus of V.

PQ and parallel to the

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

satisfy the quadratic


(ii)

+ st + 2 =

If the normals at

t lf t z

0.

[Use 16.31, ex.]

meet on the curve, prove that

fx 2

2.

Conversely, if t1 tz = 2, prove that the normals at tx t2 meet on the


curve. [Let the normals meet the curve again at s lt s 2 respectively; use 16.31,
ex. and the condition to show s t s 2 .]
(iii)

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

are the extremities, then


2/(i

+ h) = m,

i.e.

t1

+ t 2 + {- 2/m) =

0;

meet on the normal at the (fixed) point 2/m.]


11 If normals at P, Q, B are concurrent, and the chords PP', QQ', BR' are
parallel to QB, BP, PQ respectively, prove that the normals at P', Q', B' are

hence the normals at

tx , t
2

also concurrent.

12 Prove that the normals at

a{t\

t lt t 2

intersect at the point

+ tx t2 + t\ + 2),
t3 from S x =

[Solve directly; or eliminate


13

= - atitjfa +

(a;,

where

2 ).

and equations

y),

(iv) in 16.32.]

variable chord PQ subtends a right-angle at the vertex. Tangents at


is the mid-point of PQ.
meet at T, and the normals at P, Q meet at N;

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

16.41 General definition

The locus of the mid-points of a system of parallel chords of a


is

conic

called a diameter of the conic.

The parallel chords are sometimes

called ordinates to the diameter.

In 17.63 the above definition will be reconciled with the usual meaning
of 'diameter'.
16.42 Diameters of a parabola

The chord

has gradient m, where

m = 2/(

of this chord has ^-coordinate a{t t + t 2 ),


points of

all

chords of given gradient

i.e.

+ 2 )- The

mid-point

2a/m. Hence the mid-

satisfy

2a/m, which

is

THE PARABOLA

616

[16.42

therefore the equation of the diameter bisecting chords of gradient m.

Thus

the diameters of

Also see Ex. 16

(d),

a parabola are

no.

straight lines parallel to the axis.

6.

Example
Prove that

the tangent at the extremity

is parallel to the chords

P of a diameter

which the diameter

bisects.

The diameter y = 2a jm meets y 2 = 4ax where


x aim 2 i.e. at the point P(a/m 2 2a/m). The tangent
,

at

(which

is

the point 1/m) has gradient


1/(1 /m)

and

is

= m,

therefore parallel to the chords of gradient m.

Fig. 168

Exercise 16 (d)
1

Write down the equation of the diameter of y 2

2x 3y+ 1 = 0.
2 Write down the gradient

x which bisects chords

parallel to

the diameter 2y + 3

of the chords of y 2

2x which are bisected by

0.

3 Prove that tangents at the extremities of any chord of a parabola meet on


the diameter which bisects this chord.

4 V is the mid-point 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

focal chord as diameter touches the

directrix.

6 By considering the ^-coordinate of the mid-point of the chord y = mx + e


of y 2 = 4ax, show that the locus of the mid-points of chords of gradient
is
the line y = 2a /m.

Miscellaneous Exercise 16(e)


1

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.

2 Obtain the equation of the

circle

through the points (p,

0), (q, 0), (0, r).

A circle passes through a fixed point, and the chord cut off by it from a given
line is of constant length.

Two

senses.

Prove that the locus of its centre

is

a parabola.

common axis and their concavities are in opposite


If any line parallel to the common axis meets the parabolas at P, Q, prove
parabolas have a

that (provided the latera recta are unequal) the locus of the
another parabola.

mid -point of PQ

is

4 Prove

that

ax 2 + bx + c represents a parabola, and find

its

vertex

and

the

length of the latus rectum.

5 Prove that the common chord of circles


as diameters passes through the vertex.

drawn on any two

and cuts it again at G


circle touches the parabola at
6
that the axis bisects the line joining
to the mid-point of CD.

focal chords

and D. Prove

THE PARABOLA

617

7 G is the centroid of a triangle inscribed in a parabola; G' is the centroid of


the corresponding triangle of tangents. Prove that GQ' is parallel to the axis,
and that the parabola divides GO' in the ratio 2:1.

P meets the parabola again at Q, and the axis divides PQ


Prove that the ^-coordinate of P is 2ka/(l k). For what value
of k is P an extremity of the latus rectum?
9 Normals at the ends of a focal chord PQ cut the curve again at P', Q'.
Prove that P'Q' is parallel to QP and that P'Q' = 3QP.
8 The normal at

in the ratio k:l.

prove that the meet of tangents at


on the locus y\x + 2a) + 4a 8 = 0.
11 A,B,C,D are concyclic points on the parabola; AB is a focal chord; AG is
the normal at A. Show that the axis divides BD in the ratio 1 3.
12 Find the locus of a point P(h, k) such that two of the normals drawn from
10 If normals at P,

P,

lies

either

Q meet on the curve,

on a line

parallel to Oy, or

to the parabola y 2

it

4ax are perpendicular.

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.]

14 Prove that chords of y 2 = 4ax which subtend a right-angle at the point tQ


of the curve all pass through the point (a(t 2 + 4), - 2at ). (Cf. Ex. 19 (c), no. 2.)

A variable triangle is inscribed in y =

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

+ 2(yx sin 6 - 2a cos 6) r + (y2 - 4axx =


)

y = 4ax because it gives the distances r


of the points of the parabola on the line through x in direction 6.)

called the distance quadratic for


x

0.

18 Prove that the chord of y 2 = 4ax whose mid-point 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.
.

The equation x xx cos 6 = y y1


:

sin 6 of the chord

{x-x1 )jy1 =
19

Deduce from no. 18 that

Px is y 2 2ax = yy 2axx

m is the line y =

(ii)

becomes

(y-yjfia.-]

the locus of the mid-points of chords through


the locus of the mid -points of chords of gradient

(i)

2a/m.

20 Find the locus of the mid-points of chords of y 2

y =
2

4ax which touch

46a;.

21 If parallel chords are divided so that the product of their segments is


constant, prove that the point of division lies on a parabola congruent to the
given one.

THE PARABOLA

618
22 Using no.

obtain the equation of the tangent at

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)

(6k + 54) x + (8k - 53) y + (k 2

+ 29).
By choosing k so that the lines
3x + 4y + k = 0, (6k + 54:)x + (Sk53)y + (k 2 + 29) =

are perpendicular, express the equation in the form


(a)

PN

If PM,
are the perpendiculars from P(x, y) to the (perpendicular)
2 = 2PN.
3x + 4=y + 1 = 0, 4x-3y + 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

then the terms of second degree in

ax 2 + 2hxy + by 2 + 2gx + 2fy + c

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
line-pair 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)

and use the result of no. 26 (i).]


given parametrically by x =f(t)/h(t), y = g(t)jh(t). Prove that

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)

Deduce that the tangent at the point t has equation

[r 3 -> (ra

r 2 )/e, and

f(t)

g(t)

h(t)

f'(t)

g'(t)

h'(t)

let e -> 0.]

=0.

619

17

THE ELLIPSE
17.1

The

17.11

SP = e.PM

We now

loci

consider the locus denned

8P e.PM

by

(e

>

0, e 4= 1),

is the foot of the perpendicular from


where 8 is a given point and
P to a given line d (15.71). Choose 8 for origin, and let the given line
then have equation x = k (k > 0). The equation of the locus referred

to this set of axes

is

therefore

x 2 + y2

i.e.

2 (l

- e + 2e
2

e 2 (x
2

k) 2

kx + y

e2k2 .

Fig. 169

On

'

completing the square with the terms involving x, this becomes


'

e2 k \

1-e2 /

e4

e2k 2

(l-e 2 )

(l-e 2 )
Change the

origin to the point

by the formulae
x

x'

e 2k

l-e,2'
!

42

+ l-e
2

y'-

GPM

II

11

THE ELLIPSE

620

[17.12

Omitting dashes, the equation of the locus becomes

V-e*)x*+y* =

x2

(i)

^+p=

17.12 Focus-directrix property of

In the equation

~.

y2

we can assume that a >

and b > 0, and also that a #= b (otherwise it


would be the equation of the circle with centre
and radius a).
Without loss of generality we can therefore always assume a > b > 0.
Equation (ii) will be the same as equation (i) if
e

a9

On

dividing these

we

<

hence

<

r.9

ft

'

T^-

get

- e2 =

j
and

~52 ,

e2

i.e.

a2

<

1;

1.

Wo the new (dashed) axes, S has coordinates

and the equation

a;

becomes

A;

_
Hence
e

times

<
is e,

<

(ii)

its

is

whose focus

_a

&

the locus of a point whose distance from 8(ae, 0)

(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.

distance from the line x

Therefore

1.

e 2&

aje,

17.13 Focus-directrix property of -=1 v=


2=

The algebra of

17.12,

with 6 replaced by

x2

y2

b2

a
is

& 2 shows that


,

(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 b > 0; but we must


In equation (iii) we may assume a >
allow a $ b, because (iii) is not symmetrical in x and y like (ii).
17.14 Second focus

and

directrix

For both curves the condition

SP = e.PM is expressed by

(x-ae) 2 + y 2

x 2 2aex + a 2e 2 + y 2

i.e.

By adding 4:aex to both sides,

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
c-axis

A, A' are called the vertices of the ellipse.


The hyperbola cuts Ox at the points A(a, 0) and A'{ a, 0); but it
does not cut Oy. A A' is called the transverse axis; it has length 2a.
The points A, A' are the vertices of the hyperbola.
For either curve, the chord through S or S' at right-angles to A A'
is

called

a latus rectum.

We now

consider the ellipse in detail,

until Ch. 18.

17.16

Form

of the

=1

ellipse
a-

From the equation we have

and leave the hyperbola

THE ELLIPSE

622
so that

[17.17

a < x < a for all points on the curve.

Similarly

b^y^b

on the curve. Hence the ellipse lies in the rectangle whose


sides are x = a, y = b.
When x = a, the equation shows that y% \b 2 = 0, which has the
repeated root y = 0. Hence the line x = a touches the ellipse at A {a, 0)
it is the tangent at the vertex A. Similarly the line x = a touches
the ellipse at A'; and y b touch it at B, B' respectively.
Since e < 1 therefore ae < a and so S lies between O and A similarly
S' is between
and A'. Also aje > a, so the directrices cut Ox at points
D, D' beyond A, A' respectively. We have
for all points

OS =

OS'

ae,

OD =

OD' = -

(iv) f

B'

Fig. 170

17.17 Circle

When b =

and
a,

ellipse

the equation

(ii)

becomes that of the

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 focus-directrix
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

when b -> a. The


that e->0; hence by (iv), S and S'
directrices recede along the a;-axis. The ellipse

consider the ellipse represented

by

(ii)

= a (l e 2 shows
2

relation b

approach

O and

the

tends to become circular in form, since

OP2 =

x 2 + y*

t Thus the eccentricity e


are 'off-centre'.

a 2 (l-^)

= OS/OA,

+ y* = a 2 +

and indeed

(l

~) y

indicates to

->a\

what extent the

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.

Other ways of obtaining an

ellipse

17.21 Auxiliary circle

The

circle

ellipse. Its

on diameter

equation

is

A A'

called the auxiliary circle of the

is

x 2 -\-y2

_
a2

Fig. 171

Let the ordinate


cut this circle at Q.
ellipse

and

circle.

NP of any point P on the ellipse be produced to


are called corresponding points on the

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

PN QN = b:a, then the locus ofPasQ varies on the circle is an


:

For if Q is the point (x, Y), then x 2 + Y 2 = a 2 and


point (x, y) where y = bY/a, so the coordinates of P satisfy
ellipse.

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

From the similar triangles RPN, OQN,

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 semi-axes 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.

SP + S'P = e{PM + PM') = e.DD' = e(^j =


Hence

the

sum

2a.

of the focal distances of a point on the ellipse is 2a (the

bifocal property).

Conversely, the locus of a point


an ellipse with foci S, S'.

SP + S'P

such that

constant is

Proof. Choose the mid-point of SS' for origin,

S'S; then let

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 + y 2} = 2a- J{{x - c) 2 + y 2}.

+ 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.

and take Ox along


Supposing that

4a 2 - 4ca.

Again squaring,

a\x 2 - 2cx + c 2 + y 2} = a 4 - 2a 2cx + c 2x 2


hence
^1

ie

x2

1.

< PS + PS' = 2a, i.e. c < a and so a2 -c 2


The above equation therefore represents an ellipse with

triangle PSS', SS'

positive.

a2 - c2

y2

^+~/-- =
a* a z c2

From
is

^jx2 + y2 =

THE ELLIPSE

17.23]

625

a2 -c 2 If e is its eccentricity, then since a -b


we have c = ae. Hence S, S' are the points ( ae, 0),
h2

=a
i.e.

2 2

the

by

17.12,

foci.

Example
Mechanical description of the ellipse: pin-construction'
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.
'.

'

17.23 Orthogonal projection of a circle

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

obtain formulae relating the coordinates of

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 right-angles to Ox.

Choose the

on

it

Wo

these axes, suppose P{x,y) in

the angle between the planes a,

is

a;'

The

circle

+y =
2

and

projects into P'(x',y') in a'. If

a', i.e.

y'

between Oy and Oy', then

y cos 0.

a in a therefore projects into the curve


v'

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

of a circle of radius a onto a plane


angle cos -1 {bja) with the plane of the circle.
is the orthogonal projection

If

making

we rotate the plane xOy about Ox until it coincides with plane


we obtain fig. 171 of the ellipse and its auxiliary circle.

x'Oy',

17.24 General properties of orthogonal projection


(

lx'

Effect

1)

on angles.

+ m sec 6y' + n

viz. {nil, 0),

0.

A line projects into a line, since Ix + my + n =


This line and

and make angles


tan 6

<j>,

<j>'

projection cut

its

Ox

at the

becomes
same point,

with Ox, where

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

y-Vi
and

projects into

(y'

=f

(^i)'(x-x1 ),

y'J sec = f'(x'j)

(x'

x'j),

i.e.

y' - y[ = /'(a) cos 6

This is the tangent at

to the curve

(x[, y[)

y'

- x[).

(x'

= f(x') cos 6, which is the projection

of y =f(x).

By

(1),

a normal does not in general project

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

cos 0, r sin (j> cos 0)


r 2 (cos 2

+ sin 2

cos 2 0)

r 2 (l-sin 2 0sin 2 0);

to P'Q', therefore

0P'x

PQ and P'Q', so that P'Q' = PQ cos PxOP'v i.e.


P = PQ V( 1 - sin 6 sin 2 0)

is

the angle between

(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, mid-points project into mid-points. On
intersecting lines the ratio of lengths is altered.
(4) Effect

on areas. Since

orthogonal projection reduces areas in the ratio cos 6:1.


(5) By means of these results and that in 17.23, properties of the ellipse can
be deduced from certain properties of the circle by orthogonal projection.
In general only central properties can be obtained in this way; we cannot expect
to prove focal properties thus, since foci are foreign to the circle (see 17.17).

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

the circle it is true that ON.OT = OP 2 , but this cannot be


generalised into a property of an ellipse because the lengths concerned do
not he along the same or parallel lines.

N.B.

17.3

Parametric representation

17.31 Eccentric angle

Let
point

<f>

be the point on the auxiliary

on the ellipse (17.21), and

let

circle

corresponding to the

^ = NOQ

(called the eccentric

628
angle of P).

Then x

THE ELLIPSE
= ON = acos0; and since NQ =

[17.32

asin^, we have

y=

{b\a)NQ = bsm<j). Hence the coordinates of each point on the


x2 la 2 + y 2 /b 2 = 1 can be expressed in the form (acos^, ftsin^).
As increases from to 2n, Q describes the auxiliary circle and P
describes the ellipse. Two values of
which differ by an integral
multiple of 2n give the same point of the ellipse.
Conversely, for each
the point (a cos <}>, b sin <fi) lies on the ellipse,
as is clear by substituting in the equation. The point whose coordinates
ellipse

<f>

are (a cos 0, 6 sin (f>)

The point

17.32

By putting t
z

Hence

referred to as the point

is

tan \<$ in the coordinates of the point 0,

1 t2
= acos0 = aj-^,

for each

t,

a
\

on the

ellipse

2/

2t

= 6sin0 =

6j-j-^.

the point
/

lies

<f>.

1-t 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
;

A a, 0) is not given by any value of


'(

as the limit

when

although

t,

it

can be obtained

t->oo.

This algebraic representation can also be obtained directly; for


since

b2

t
therefore

l-xla
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
/l-t*\
|-

+ 62

l
|

2t

\
-)

cuts the ellipse at the points

+2ga

1-t*
-

for

which

2t

+ 2/6- + c =

0,

- 2ga + c) tl + 4/6<8 + 2(26 2 - a 2 + c) t z + 4fbt + (a 2 + 2ga + c) =

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

(positive, negative, or 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

of the ellipse are

3,

concyclic, then

is

also sufficient: see

Ex. 17 (a),

no. 11.

Exercise 17(a)
'

The

ellipse''

means

the locus

x2 /a*+y 2 /b*

1.

Find the length of a latus rectum of the ellipse.


2 With the notation of 17.15, P is any point on the ellipse and the perpendicular from P to A A' has foot N; prove that PNZ :A'N.NA = OB*: OAK
Find a similar relation involving N', the foot of the perpendicular from P
1

to BB'.
3

If

P is the point (x, y) on the ellipse, prove SP = a ex and S'P = a + ex.

4 Use the bifocal property to prove that BS = BS' = a.


Given an ellipse and its major and minor axes A A' BB', construct the
',

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,

(b/a) cot \(d +


b(l *1 *2 )/a(1 + * 2
of SP is an ellipse whose

6 Prove that the gradient of the chord

6<j>

7 Prove that the gradient of the chord

<!< a is

is

<f>).

).

8 Show that the locus of the mid-point


the mid-point of OS. State the lengths of the semi-axes.

centre

is

9 Prove that the mid-point of the chord 6$ has coordinates


(a cos (0 + 0) cos $(0-0), b sin (0 + 0) cos $(0-0)).

10 Find the locus of the mid-point of the chord 00


= 2a, where a is constant.
(ii)

when

(i)

2a;

1 1 Prove that the condition E0 X = 2nn is sufficient for the points


V 2 8 4
of the ellipse to be concyclic. [The circle through v 2 3 cuts the ellipse again
at 04 prove 0 4 is a multiple of 2n, so that 0^ and 4 give the same point.
,

<f>

Cf. 16.12, ex.

(ii).]

circle cuts an ellipse at A, B, C, D. Prove that the chords AB, CD are


12
equally inclined to the major axis (and therefore also to the minor axis). [Use
17.32, ex. and no. 6.]

and cuts
13 If a circle touches an ellipse at
are equally inclined to each axis.

AC,

it

again at

G and D, prove that

AD

14 Circles touch an ellipse at a given point. Prove that their


not through that point are parallel.

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 mid-point of a chord
of an ellipse whose centre is O;
meets
the ellipse at P;
is the mid-point 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

*19 The chord PQ has mid-point M; the tangents at P, Q meet at T. Prove


that O, M, T are collinear. If this line meets the ellipse at R between
and T,

prove

OM.OT = OB

*20 Deduce the area of the

with semi-axes

ellipse

a, b

from that of a

circle

of

radius a.

Further examples using orthogonal projection appear in Ex. 17

17.4

Chord and tangent

17.41

Chord P,P2

The line joining

Since

and P2

lie

and

on the

(d).

P2 has equation

ellipse,

^+

- + p-L

and

By subtracting and then rearranging,


y\-y\

b2

so that

x\-x\
a2

~ =

x 2 -xx

a 2 y2

The equation of the chord PjP2

'

1.

yt

therefore becomes

b *
x2 T
+ x*1/
"
t
a2 yz + yi
I.e.

i-e-

{x-xx )(x2 +xx )


S

^K + ^) +

t
I

a2
p(2/i

(y-y 1 )(y2 +y1 _


Av,
To
)

b2

+ 2/2 =
)

^+^+l-

(i)

THE ELLIPSE

17.42]

631

Alternatively, consider the equation

{x-xx ){x-x% )

(y-y1 ){y-y2 )
b2

a2

a/

'

When simplified it is linear in x and y, and therefore represents a line.


by the points
be the chord PxP2

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<

The chord joining


,

(a cos

y-bam.6
y
Y =
,

<j>,

b sin

and

<p)

&(sin0-sin0)
-j

- a cos g>)
Y
,

(a:

a(cos0-cos0) v

_ 26cos(0 + 0)sin(0-0)
- - 2a
2arinlf/9-nA)
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 + 0)},

-cos(0 + 0)+|sin(0 + 0)

i.e.

17.43

Chord

t2

This line will be the chord


2,

<j>)

(ii)

x t2

if

the above quadratic has roots

_ 1+ua

^+
t\* 2

which

0.

and then

Hence

for

2t

(l-ua)t 2 + 2vbt + (l+ua)

i.e.

sin (0 +

cos(0-0).

cuts the ellipse at points


1

and

+ sin

ft fa

The line ux + vy + 1 =

and

2vb

ua,

= _ i((l + y_J_ = _il^.

THE ELLIPSE

632

[17.44

The required chord is therefore

U-2)^ + (*i + *2 )f= 1+M2


Alternatively, the

= tv t =

chord

x t2

cuts the ellipse at the points given

which are the roots of the quadratic


*

-(*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.

and so represents a line. It cuts the ellipse at


i.e. at t = t x t 2 and therefore is the chord t t
x 2
,

This can be obtained by making


giving

2?h.2m =
a2

since

(t
)

(iv)

(i-g-(* 1+ y| + M2 (i + 3 =
It is linear in

tx

0-

Consider the equation constructed from


*

(iii)

62

and y2

"^

"

a2

->

yx

in equation

9'

Px lies on the ellipse. The tangent at Px is therefore

which can be written down by the usual 'rule of alternate suffixes'.


Alternatively, equation (v) can be found directly by calculus.
17.45 Tangent at 4>

Letting 6 ->

(J)

in equation

(ii),

we obtain

x
v
-cos ^+^sin^
r
a
b

This could also be obtained directly by calculus. Cf. Ex. 17


17.46 Tangent at

Letting

t2

->

(vi)'

1.

(a),

no. 17.

in equation

(iii),

we

obtain (after omitting the

suffix 1)

(l-* 2 )- + 2*|
a
o

as the equation of the tangent at the point

(vii )

t.

THE ELLIPSE

17.47]

633

can be found by using the theory of quadratic


equations, or by calculus, or as a particular case of the result in
Alternatively, this

Ex. 16 (e), no.

28.

17.47 Examples
Condition for

(i)

= mz+c to

x2
touch

y
+
=

1.

cuts the ellipse at points for which

The line

x%
a2

(mx + c)* _

~h

62

(am 2 + 6 2 ) x* + 2a2mcx + a 2 (c2 - 6 2 )

i.e.

The
roots,

line will

touch the

ellipse if

i.e.

and only

if this

0.

quadratic in x has equal

= o2(a2m2 + ft2) (c2 _ ft2))


= a%i 2 + 6 2

ca

i.e.

Hence for all values of m the


y
touch the

lines

= mx V(a2m2 + 6 2

ellipse.

Locus of the intersection of perpendicular tangents : director circle.


(i) passes through the point (# >2/o)> then

(ii)

If a tangent found in ex.

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

and having radius A/(a2 + 6 2 ) = AB


(fig. 170). It is the locus of points from which perpendicular tangents can be
drawn to the ellipse, and is called the director circle of the ellipse (by analogy
with the parabola, for which the corresponding locus is the directrix: see

This

is

circle concentric

Ex. 16 (6), no. 11

with the

ellipse,

(ii), (iii)).

The argument used in


(iii) Chord of contact of tangents from P x
show that the chord of contact from Px to the ellipse has equation
.

15.64 will

THE ELLIPSE

634

Exercise 17(b)
1 Use equation (i) in 17.41 to prove that the mid-points 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

is a focal chord if and only


equivalent to tan yd tan 0

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.

2a, prove that the chord d^>

is

parallel to the tangent at a,

6 For a system of parallel chords prove that the


of the extremities

is

and

sum of the eccentric angles

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

[Method of 17.47, ex.

(i);

x 2 /a 2 + y 2 /b 2

and only

if

if

or compare with 17.45, equation (vi).]

9 If m is the gradient of a common tangent to the circle x 2 + y 2 = c 2 and the


x 2 /a 2 +y 2 /b 2 = 1, prove m 2 (c 2 2 )/(a 2 2 ). Deduce that four common
tangents exist only if b < c < a.
ellipse

10 Prove that tangents from

2, 3)

to 4x 2 + 9y 2

36 are perpendicular.

PN is the ordinate of a point P on the ellipse, and the tangent at P meets


the maj or axis at T. Prove ON .OT = a and obtain the corresponding property
11

for the

minor

axis. (Cf. 17.24 (5), ex.)

The tangent at P meets the

PSZ is a right-angle.
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.

(Cf. 8.14, ex. (iv);

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

+ V-sin o =

square and add these equations.]

is

on the auxiliary

V
cos <b
a
i

ae

sin
b

<p

circle.

[The

THE ELLIPSE

635

19 The chord PQ subtends a right-angle 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

Prove that the tangents at the points


1

[Use equation

(vii)

tx , t2

x/a _ 1+xja _

in 17.46,

meet at

(x, y),

where

yjb

and theory of quadratics.]

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

[The chord of contact from T(xlt yx ) to the

ellipse

touches the

circle;

use no.

8.]

24 Prove that the tangents at the extremities of a variable chord through Px


meet on the line xxja? + yyjb 2 = 1. What follows by taking Px at a focus?
*25

Given the points Pv P2 prove that the point dividing


be on x2/a2 + y2/b 2 = 1 if
,

(^iHH^iHHM(Joachim8thaVs ratio quadratic for the


*26

PP
x

in the ratio

will

By

taking

on the

ellipse,

1 )''-

ellipse).

deduce the equation of the tangent at

Pv

[See Ex. 16(6), no. 26.]

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.

A variable chord through P cuts the ellipse at A and B, and P is chosen


P and P divide AB (one internally and the other externally) in the
same ratio. Prove that P lies on the line xxjat + yyjb = 1. [Method of
*28

so that

15.65(2).]

*29

(i)

Denning the polar of Px wo

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

17.51 Equation of the

normal

The reader should verify that the normals


and t have respective equations

x-x 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

passes through the point

2ath -b{l-t 2 )h

i.e.

bkt*

+ 2(ah + a 2 - b 2

Since this equation

from a given point

is

) t

2(a 2 - b 2 )

(h, k) if

* (1

+ 2(ah -a 2 + b 2 )t-bk =

quartic in

t,

at

(h, k) to the ellipse;

0.

(iv)

most four normals can be drawn

and (apart from coincidences)

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

there will be either

the constant term

is

minus the

Y>t x t 2

coefficient of 4

and

we have

t^t^ = 1.

(v)

These necessary conditions for the points tv t2 tz 4 of the ellipse to


be conormal (i.e. such that the normals at them are concurrent) can
also be shown to be sufficient. This is reasonable, because two indepen,

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.)

4 Find the locus of the mid -point of PG.


2
2
is a normal to 2x + 3y
5 Prove that 3x yl =
coordinates of

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>

again at the point diametrically opposite to the fourth. [If


diametrically opposite, then ft = 4 + 7r by Ex. 17 (6), no. 2. Since

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,

intersect at points for


2

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 }\

and that to each root of this quartic in y corresponds a single x given by


{(a 2 b 2 )y + b 2 y x } x = a 2 x x y. Deduce that in general four normals can be drawn
from Px to the ellipse.
17.6

The

17.61

The line through P1 ( 1

distance quadratic

equations
^

y^) in direction

6 (15.26) has parametric

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

d + b 2 cos 2 6) r 2 + 2{a2yx sin

a?b 2

+ b 2x x cos d) r
+ (b 2xl + a 2y\-a 2b 2 =
)

0.
43-2

(i)

THE ELLIPSE

638

[17.62

This quadratic in r gives the (signed) distances from P of the points


x
ellipse which he on the line through P\ in direction 6, and is
called the distance quadratic for the ellipse.

on the

Example*
Newton's theorem.

The product of the


(x\

y\

roots of
// cos

it

(i) is

fl

// \~a?~

For chords Pj^QR,


tions 6, d'

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>'

sin 0'\ //cos

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

OK, OK' lie

in directions 6, 6'.

P Q.P R _ OK*
P Q' ,P R' ~ 0K'
X

Hence Newton's theorem:

'

It is a generalisation of the 'product property of chords of a circle, and reduces


to this property when b = a. Also see Ex. 17 (d), no. 16.
By letting -> Q and R' -+ Q' we see that the ratio of the tangents from
'

is

Px

equal to the ratio of the parallel radii.

17.62

Chord having mid-point

If the chord through

P1

in direction 6

is

roots of the distance quadratic will be equal

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.

of the chord which is bisected

= y-Vx
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

above has gradient m, then

the mid-points of all chords of gradient

Hence

62

a 2m

According to the definition in 16.41, this locus is


all chords of given gradient m. It is a straight
of the ellipse, and is thus a
line which passes through the centre
'diameter' in the usual sense.
Cf.

Ex. 17

no.

(b),

1.

the diameterf bisecting

Examples
Tangents at the extremities of a diameter are parallel

(i)

chords which

to the

the diameter bisects (the ordinates to the diameter).

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^, i-e- - The result follows.

Tangents at the extremities of any chord meet on the diameter which

(ii)

bisects

that chord.

PQ have gradient m, and let the tangents at P and Q meet at


PQ is the chord of contact from (x^y^), its equation is

Let the chord


Since

0i>2/i)-

xxJaP+yyxlb 2 = 1, and hence its gradient is m = b^xja^. Therefore


lies on y = b^xjahn, which is the diameter bisecting PQ.
(iii)

The sum

{xlf y x )

of the squared reciprocals of two perpendicular semi-diameters is

constant.

Choosing O for pole and Ox for initial line, the


polar equation
2
sin2 0\
/cos

(^- + -H =

ellipse

x 2 /a 2 +ys /b 2

has

1-

If P on the ellipse has polar coordinates (rlt X ), then the coordinates of an


extremity Q of the perpendicular semi-diameter are (r 2 X + \ir), where from
the equation,
cos 2^
1
,

7\~~
1
-r

and

r|

a*

62

2
+ iTT) = sin 2 ^
= cos^ + ^tt) sin^
2
2
1

1111

cos 2
1

62

_ + _ = _+_

Addmg,

17.64 Conjugate diameters

By 17.63, the diameter bisecting all chords parallel to the diameter


=
y mx (i.e. having gradient m) is y = m'x, where m' = Wjahn,
f

Strictly, the

diameter

is

that part of the locus which

is

inside the ellipse.

THE ELLIPSE

640

mm' = b

i.e.

ja

The symmetry of

diameter which bisects

all

[17.64

this relation

shows that the


= m'x

chords parallel to the diameter y

y = mx.

is

Two diameters which are such that each bisects

all

chords parallel

to the other are said to be conjugate.

Hence y = mx, y

m'x are conjugate diameters

mm =
It

is

from the

clear

if and only if

~.

a2

definition that the axes of the ellipse are a pair of

conjugate diameters; they are the only perpendicular pair.

Examples
//

(i)

<f>

6,

<[>

are the extremities

P,D of a pair of conjugate semi-diameters, then

\tt.

For the gradients of OP,


conjugacy condition

OD are respectively (b/a) tan 6, (b/a) tan


^2

is

tan 6 tan 6Y =

a2

<J>,

and the

^a
-,

a2

from which tan 6 = cot = tan (0 $n) and hence 6 = \n.


The result shows that the corresponding points P', D' on the auxiliary circle
subtend a right-angle at O; i.e. to conjugate diameters of the ellipse correspond
<j>

<f>

perpendicular diameters of the auxiliary


*(ii)

an

Perpendicular diameters of a

conjugate diameters of

mx, y m'x, where mm' = 1, are the diameters of the


= a 2 they project into the lines y = m x x, y m'x x where (see
= m cos 6, m'x = m' cos 6, and so mx m'x = cos2 d = 6 2 /a 2

If

circle project into

ellipse.

x* + y2

circle.

circle

17.23)

are perpendicular diameters of the circle, then all chords


parallel to fi are bisected by A. Hence in the projected figure, all chords of the
ellipse which are parallel to ji' are bisected by A' (see 17.24); i.e. A' and /i' are
conjugate diameters of the ellipse.
It follows that properties of diameters and conjugate diameters can often
be conveniently proved by orthogonal projection from a circle.
Alternatively, if A,

*(iii)

through

First

Then

/t

The tangent at T to an ellipse meets the diameter OP at H, and the


T parallel to the conjugate diameter OD meets OP at K. Prove that
OH.OK = OP*.

draw the

(fig.

175)

figure for

circle,

replacing 'conjugate'

by

line

'perpendicular'.

OK. OH = OT 2 = OP 2 and since this can be written in terms of


,

on the same line, the result follows by projection on any plane


throughO. (The example in 17.24 (5) is a special case of the property just proved.)

ratios of lengths

*(iv) Ellipse referred to

With the notation of


theorem

(17.61, ex.) that

a pair of its conjugate diameters as (oblique) axes.


fig. 176, we have by taking P x at V in Newton's

qy

q-qs

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

PV.VP' = {OP-OV)(OP + OV) = a\-x\


where ax = OP, the above equation becomes (on writing b x = OD)
a\ x 2

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

2 Write down the gradient of the chords of 3a; 2 +


by the diameter y = Zx.

In

the following,
are P', D'.

2 which are bisected

OP and OD are conjugate semi-diameters whose other extremities

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 equi-conjugate 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.

7 Deduce from nos. 5, 6 (iii) that the (p,r) equation of the


centre as pole is a 26 2 /p a = a2 + 6 2 2 [OP = r; eliminate Oi>.]
.

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,

O V lie along conjugate

diameters.

Nos. 12-15 are easily solved by orthogonal projection.

QR is a diameter, and P is any point on the ellipse. The tangent at Q meets

* 12

PR at T.

Prove that the tangent at

P bisects TQ.

*13 OP, OD are conjugate semi-diameters; the tangent at


of the ellipse at Q and R. Prove PQ.PR = OD 2

P meets

the axes

PQ

*14

is

HK

a diameter and

at Y. Prove that

XY

is

is

QK cut at X; PK, QH cut

any chord; PH,

parallel to the diameter conjugate to

PQ.

By projecting a circle and a circumscribing square, prove that the tangents


the extremities of conjugate diameters PP', DD' form a parallelogram of

*15

at
constant area 4ab. (Cf. no. 6 (ii).)

*16 Prove Newton's theorem (17.61, ex.) by projection. [When written in


the form

Rt
^ p^, p^
OK' OK ~ OK' OK'
'

involves ratios of lengths on parallel lines.]

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

*18 Prove that the mid -points of variable chords through

xxl

yy 1
|

a2

b2

_
~

x2

y2

a2

b2

he on

'

Show that this is an ellipse whose axes are proportional to

those of the original

and whose centre is (%x lf ^y ).


*19 Use the distance quadratic to obtain the equation of (i) the tangent at
Pi [two roots r = 0]; (ii) the pair of tangents from Px [equal roots r],
*20 Obtain the equation of the chord having mid-point Pt from the ratio
ellipse

quadratic in Ex.

PP PP
X

1 7 (6),

and PX A

no. 25. [If the chord is

AP

Z,

and hence are

AP P

the roots k

2,

0, kjl

are the ratios

= .]

Miscellaneous Exercise 17(f)


1

of

The

lines

P when

bx + aty ab

0, btx

ay + abt =

meet at P. Find the locus

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.

Q are the points 6, $ on the ellipse x 2 /a 2 +y 2/b 2 =


Q meet at T. Prove that triangle TPQ has area
ab sin8 \(<f> 6) sec J(0 6).

*4 P,
at P,

[The corresponding area for the

circle is

and the tangents

1,

2A0PT A0PQ; use projection.]

Y is the foot of the perpendicular from S to the tangent at P.


and SY meet on the directrix corresponding to S.
5

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

a given point inside a given circle of centre O and radius a; Y is a


on the circumference, and p is the line through Y perpendicular

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)

Deduce that p touches an

ellipse

having S for focus and the given

circle

for auxiliary circle.

*8 Give the result corresponding to no. 7 (iii)


What happens when S lies on the 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.

centre of an ellipse coincides with the vertex of a parabola, and they


have a focus in common. Prove that the common tangents meet the common
axis at its intersection with the other directrix.

10

The

11

(i)

Prove that two tangents can be drawn from

and only if x
(ii)

+y

-1 >

to x 2 /a 2 + y 2 /b 2

[Use 17.47, ex. (i).]


the ratio quadratic (Ex. 17 (6), no. 25) prove that

fa

From

/b

1 if

0.

if

then Px and Pa lie on opposite sides of the ellipse. (Accordingly, the set of points

for

which x\ja2 + y\jb 2

>

is

the outside of the

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 line-pairs
O parallel to TP, TP' and to ST, S'T. By showing that they have the
same angle-bisectors, deduce that ST, S'T are equally inclined to TP, TP'
13

through

respectively.

14 A point P varies so that the chord of contact from P to x2 ja2 + y 2 /b 2


touches the ellipse x 2 /a 2 +y 2 /b 2 = 1/k 2 Find the locus of P.

15 The chord PQ of x2/a2 +y 2Jb 2 = 1 touches the parabola y2 = 4&X. Prove


that the tangents at P, Q to the ellipse meet on the parabola a 2cy 2 + b*x = 0.

THE ELLIPSE

644

16 Prove that points whose chords of contact are normals to the


x 2 /a 2 + y 2 /b 2 = 1 lie on the curve a/x 2 + b 6 /y 2 = (a 2 -b 2 ) 2

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 mid-point (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 line-pair 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 semi-diameters 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

22 Lines are drawn through


x 2 /a 2 + y 2 /b 2

a 2x 2 + b 2 y 2

perpendicular to the tangents from

to

If these lines are conjugate diameters of the ellipse, prove


the curve a 2x 2 + b 2y 2 = a* + b*. [Use the equation of the pair of

1.

that x lies on
tangents in no. 12.]

Show that each of the following equations represents an


semi-axes, eccentricity, foci,

23
25

ellipse,

and find

and directrices.

&x + 2) 2 + \(y-l) 2 = 1.
3(x-y+l) 2 + 4:(x + y-l) 2 =

24 8x2 + 9y 2 - lQx-Qy
12.

63.

the centre,

645

18

THE HYPERBOLA
The hyperbola

18. 1

^-p=

asymptotes

Form of the curve


We now resume the discussion, begun

18.11

hyperbola

^2

in 17.13-17.15, of the

oi2

5- =

i-

7i

From

x2

^=

the equation,

so that

1,

a for all points on the curve;


lines x a. Further, since

y
6

is

y2
4-p >

\x\ $s

between the

we see

that

when x ~> + cc

i.e.

no part of the curve

lies

or x ->

a2

oo,

'

then also y -> oo. The curve

therefore unbounded.

When x a, the equation gives y 2 /b 2 = which has the repeated


root y = 0. Hence the line x = a touches the hyperbola at A {a, 0): it
is the tangent at the vertex A Similarly, the line x = a is the tangent
at the vertex A'{ a, 0).
Since e > 1, we have ae > a and aje < a. Hence S lies on OA
produced beyond A and if the directix x = aje cuts Ox at D, then D
lies between O and A. Similarly S' lies on OA' produced beyond A',
.

and D'

lies

between

and -4'.

18.12 Asymptote: general definition


(

1)

We say that Px -> oo aZowgr a curve if one or both of its coordinates

xx yx tend to oo while satisfying the equation of the curve.


For example, Px -> oo along the parabola y2 = 4ax when xx -> oo and
-> oo, or when x 1 -> co and 2/i -> oo Px
oo along the curve
-> 0, or
y = l/x 2 when xx -> oo and yx -> 0, or when x x -> oo and
when xx - and yx - oo; but Px cannot tend to infinity along the
,

THE HYPERBOLA

646
2

x /a

ellipse

point

+y

/b*

^ a and

have
(2)

An

\yx

because for every point

1
\

[18.13

P1

on

this curve

we

b.

asymptote of a curve

a line such that the distance of a

is

of the curve from this line tends to zero

when

-> oo along

the curve.

In

may mean

this definition, 'distance'

the usual 'perpendicular

some fixed
The informal use of the term 'asymptote'

distance' or the 'oblique distance' measured. parallel to


direction, e.g. the y-axis.

in 1.41,

Remark

(a), is

consistent with our definition.

18.13 Asymptotes of the hyperbola


If

P1 lies on the curve, then x{\a% y\\b % =

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 line-pair, is therefore
62

a2
Alternatively, the equation

y
If y

can be written

(i)

a a \*

/
1

-^)

'

= mx + c is an asymptote to the part


b

y
then

Using the binomial

2/carve-2/asnnPtote

series,

= +-x[l

a 2 \*
-

when

we have for large x that

"

+ (|-")" +

'

-+CO.

THE HYPERBOLA

18.14]

and

tend to zero when x - oo only if c


is an asymptote. Similarly, y

this will

the line y = bx/a


the other part

a2 \

= a X \ l -x*)

647
and bja m =

=
= bxja

0. Hence
an asymptote to

is

i
'

Since
for

any point on the hyperbola, therefore y 2 <

y\\

hence

within that double angle between the lines y

lies entirely

the,

hyperbola

= bx/a which

contains the x-axis.

The angle between the asymptotes is 2tan _1 (6/a). Figures 177,


178, 179 illustrate the cases a < b, a = b, a > b respectively.
When b = a, the asymptotes are perpendicular and the curve is
called a rectangular hyperbola. The equation (i) becomes x2 y2 = a2
and the eccentricity, given by b 2 = a2 (e 2 1), is e = <J2.
;

18.14
If

The

Fig. 179

Fig. 178

Fig. 177

bifocal property:

lies

SP-S'P= 2a

on the branch which

contains the focus 8, then

S'P-SP = e.PM'-e.PM

M'

= e.MM'

=()
=
If
S',

2a.

P lies on the branch enclosing

Fig. 180

then

SP-S'P = e.PM-e.PM' = e.MM

2a.

'

THE HYPERBOLA

648

[18.2

Conversely, the locus of a point P such that S'P SP = 2a is one


branch of a hyperbola having foci 8, S' and transverse axis of length 2a,
viz. that branch which contains S.
Proof.

The algebra of 1 7 .22 shows that, with the same choice of axes,
a2

a2 c 2

From triangle P8S' we have PS' < PS + 88',


2a

<

so that a

c.

= PS'-PS <SS' =

Hence we can write

b2

c2

i.e.

2c,

a 2 and
,

the equation be-

comes that of the standard hyperbola (i). Since b 2 = a 2 (e 2 1), we


have c 2 = a 2 + b 2 = a 2 e 2 and so c = ae. Therefore 8, 8' are the points
( + ae, 0), i.e. the foci. Finally, since SP < S'P, P lies on the branch
which contains S.
,

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 = AB-BP = AB-(l-PC)
l.

= (AB-l)+PC,

AP PC is constant, viz. AB
AB
the locus of P is that branch

therefore
If
=#

I.

I,

of

a hyperbola with foci A. G which encloses G.


If AB = I, then AP - PG, and P lies on the

Fig. 181

perpendicular bisector of AG.

18.2

Properties analogous to those of the ellipse

^+p= 1

Since the equations of the ellipse and hyperbola differ only in the
sign of 6 2

it

follows that

many

properties of the hyperbola can be

written down from the corresponding results for the ellipse by replacing
b2

by b 2

All such properties can be proved independently

methods used
(i)

Chord

for the ellipse; the following

PX P2

is

a summary,

by the

THE HYPERBOLA

18.3J
(ii)

Tangent at

Px

(ui)

Normals

Pl

649

= L

^+^

0.

The contact condition for y = mx + c is c 2 = a 2 2 6 2 provided


\m\ < bja. The condition remains significant when m = bja, for then
c = 0; but the lines y = foc/a do not cut the hyperbola anywhere
(iv)

(see 18.13).
(v) If

bja <

m<

b\a,

the lines

= mx *J(a 2m 2 b 2

touch the hyperbola.


(vi)

Director circle.

The locus of the intersections of perpendicular

tangents to the hyperbola

the

is

circle

x2 + y2

= a2 b 2

provided a >
(vii)

b. If a < 6, there are no perpendicular tangents.


Chord of contact from Px
:

b2

a2
(viii)

Pair of tangents from

Px

L_t_A

x l_yl_A 3
2
a2 b 2
J [a
(ix)

b2

Chord having mid-point

xx i
a2

Px

= (xx2i_yyi_iY
2
[a

'

Wi _ x l
b2

a2

yl
'
b2

lie on
(x) Diameters. The mid-points 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

18.31 Hyperbolic functions

The equation

(i)

in 18.11

0; but since ch0 >


of the hyperbola.

all

0,

is satisfied

by x = ach0, y =

bsh<fr for

these equations represent only one branch

THE HYPERBOLA

650

[18.32

Analogously to the treatment in 17.32, put t

x
If t

now

is

1
t2
=a +
1
T

fi

oo.

(If

and +

18.32

b-

2t

1 T

then

unrestricted, these equations represent the whole of the

curve except the point A'(


t ->

= thf^;

still

but this can be obtained by letting


it can take only values between

0);

denotes th \$,

1.)

The point

The general representation above can be found

directly.

For since

the equation can be written

IHH(H
= ~-- =
x/a + 1

xja-l

f
therefore

y/b

._

y/b

(H

Hence

-:|:1 =

i.e.

! fi

a b

(l

+1

t,

say.
J

<

:1 '

+ t*):2t:(l-t*).

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 \
i-t*'l-t 2 )

(j>,

by x = a sec $,y =

tan 0) as the point

tan \<f>,

we obtain the

point

18.34 Another algebraic representation


Since equation

(i)

b tan <j> for all ^.

can be written

in 18.32.

THE HYPERBOLA

18.4]

we may put

x_

651

xyl
a

From these we find


x
These equations represent the whole of
t = +l gives A(a, 0), and
= 1 gives A'( a, 0). When t increases

the curve;
t

from oo to +oo, the parts traced


P'A'
t

p
Fig. 182

are:

-1 <

AQ

PA

A'Q'

< -1

<

<

<

Chord, tangent, and normal

18.4

To each of the preceding representations corresponds a standard


form of equation for chord, tangent, and normal. By methods already
illustrated in Ch. 17, the following results can be obtained.
Chord 6$: -cos(0-0)-|sin(0 + 0)

-sec0 ^tan^ =

Tangent at 0:

$
t:

ax cos
2atx + b{l

The reader should

<j)

+ by cot

+ t*)y =

<f>

= a2 + b 2

2(a 2

</>).

1.

(l+t 2 )--2t%= l-t 2


a
o

t:

at

cos %($ +

(i+M2 )|-(*i+yf = !-m2

kh-

Normal

+ 62

1-t 2

'

verify these; also see nos. 13-15 of the following

Exercise, which contains examples similar to those already given for

the ellipse together with some properties of the asymptotes of the


hyperbola.
Exercise 18(a)
1

is

If

P is the point (x, y) on the hyperbola, prove SP = ex a axidS'P = ex + a.

2 A shot is fired from A so as bo strike an object B. The sound of the firing


heard at P, n seconds before the sound of the destruction of B. Find the locus
44

GPMII

THE HYPERBOLA

652
of P. [Let

AB;

to

u=

velocity of sound, v

velocity

component of the shot

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)

State the order in which the hyperbola


/

+t

is

traced

by the point

2bt \

i-t 2 )
as

(ii)

(6)

oo to + oo.
What is the condition for the points tlt t2 to

increases from

(a) lie

on the same branch;

be diametrically opposite?

on the ellipse x 2 /a2 + y 2 /b 2 =


5 P, P' are the points <p,
are A, A'. If
meets A'P' at Q, prove that the locus of Q

AP

x 2 /a 2 -y 2 /b 2

1
is

whose

vertices

the hyperbola

1.

6 If the points
prove that

6,

are the extremities of a focal chord of x2 /

<j)

tan \d tan \<p

_e

or

/b 2

1,

e
.

touches x 2 /a 2 y 2 jb 2 = 1 if and only if


7 Prove that Ix + my + n
aH 2 -b 2m 2 = n 2
8 (i) The tangent at any point P of the hyperbola meets a directrix at Z.
.

Prove that
(ii)

PZ subtends a right-angle at the corresponding focus.

Deduce that tangents at the extremities of a

focal

chord meet on the

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).

of the perpendicular from a focus to an asymptote


10 Prove that the foot
on the auxiliary circle and on the corresponding directrix.

lies

[OF = OS cos FOS =

...

a; also

OS.OD =

a2

= OF 2

so

ODF

is

right-

angle.]

The line joining the focus S to the point P on a hyperbola is parallel to an


1 1
asymptote. Prove that this asymptote, the directrix, and the tangent at P
are concurrent.
12 The point P on the hyperbola is such that the tangent at P, the latus
rectum through S, and an asymptote are concurrent. Prove that SP is parallel
to the other asymptote.

With

the representation in 18.34, obtain the equation of

13 the chord

then

S^tfjj

14 the tangent at

tx t 2 .

16 Prove that

if

and

the normals at the points


t

x t % t z t

15 the normal at
t lt t 2 , ts , 4

t.

of 18.32 are concurrent,

= 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.]
\

18 If the normals at <p lf


4 are concurrent, prove
a
is some integer (positive, negative, or zero).
,

where n

0 X = (2n+l)n

THE HYPERBOLA
Show that the feet of the normals from P to x

18.5]
19

curve a 2y(xx x) + b 2x(y 1 y) = 0.

20 If the points

18.5

t lt t , t ,
2
3
4

653
%

la

are concyclic, prove S(f1 + *a <8 *4 )

\b %

1 lie

on the

0.

Asymptotes: further properties

18.51 Lines parallel to an asymptote meet the hyperbola only once.

Any line parallel to

the asymptote x/a + y/b

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

parallel to x\a y\b

=
for

has equation

k.

which

2t

+ l-t 2 ~

'

+ t) 2 = k{l-t 2 ),

= -l

1+* =

or

jfc(l-J).

no point of the curve, there is a unique inter(k l)/(k + l). A similar result holds for lines
0.

18.52 The equation of the tangent at


asymptote when x -> oo along the curve.

tends to the equation of

an

The equation xxx \a 2 yy x \b 2 =


b2

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

asymptote is sometimes denned as 'the limit of the


Px -+co along the curve'. For curves whose
equations are algebraic, it can be shown that the two definitions are
equivalent; but for general curves a line may be an asymptote
Remark.

tangent at

when

44-2

THE HYPERBOLA

654

[18.53

according to the definition in 18.12, but not according to that above

because the equation of the tangent

may have no limiting form when

Px -> oo.
18.53 The family x 2 \a 2 y 2 \b 2

k of hyperbolas has the same asymptotes

for all k.

For the equation can be written in standard form as


y
ka

kb 2

'

and the asymptotes have equation

^_ll_
ka 2

which

is

kb 2

ie
'

t-t-o
~
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 as-coordinate 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 mid-point 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

the asymptotes is bisected at the point of contact.

18.6

The conjugate hyperbola

18.61 Definitions

In 17.15 we defined the transverse axis to be the intercept A A' made


by the hyperbola on Ox. Although the curve does not cut Oy, it is
now convenient to consider the points -6(0,6) and B'(0, b) on this
axis of symmetry and (analogously to the ellipse) to call BB' the
conjugate axis of the hyperbola.

Two hyperbolas are said to be conjugate when the transverse and


conjugate axes of one are respectively the conjugate and transverse
axes of the other.
It follows that the hyperbola conjugate to

THE HYPERBOLA

18.62]

655

IS
1.

b2

For, with

(ii)

Oy

for #-axis and Ox' for y-axis, the equation of the conjugate hyperbola is x 2 jb 2 - y 2 ja 2 = 1 On rotating these axes clockwise
.

through a right-angle to positions Ox, Oy, the equation becomes


2
2
2
y \b -x \a 2 = l,i.e. (ii).

The two hyperbolas have the same asymptotes,


x2
a2

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

fax + brf + cj (a2 x + b z y + c2 ) =

A,

a1 x + b 1 y + c 1

0,

where A is constant; for this equation differs only by the constant A


from the equation of the asymptotes, viz.
{a1 x

By

writing

+ b 1 y + c x ){a2 x-\-b % y + c z ) =

for A,

we

0.

obtain the equation of the conjugate

hyperbola.
(y)

In particular, any hyperbola whose asymptotes are the co= 0, y = has equation xy = A.

ordinate axes x

18.62 Conjugate diameters


(i)

wo

// a pair of diameters are conjugate

the conjugate hyperbola.

wo a hyperbola,

they are also conjugate

THE HYPERBOLA

656

The argument which leads

[18.62

to property (xi) in 18.2 can be applied to

and gives the same conclusion independently of k: y = mx and y m'x are


conjugate diameters of each of the hyperbolas (iv) if and only if mm' = b 2 /a a
The result stated concerns the cases k = + 1, 1.
.

(ii)

Intersections of the hyperbolas with conjugate diameters.

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)

Extremities of conjugate diameters.

The points of intersection are

called the extremities of the diameters, despite

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>

i.e. where x = atan0 and hence y = 6 sec


Thus P(a sec
D(a tan 0, b sec (j>) are extremities 6f conjugate semi-diameters.
<fi.

<j>,

6tan0) and

Exercise 18(6)
1

If

e,

e'

are the eccentricities of a hyperbola

1.

l/e a +l/e' 2

OP,

OD

and

its

conjugate, prove

are conjugate semi-diameters of the hyperbola x 2 /a 2 y 2 /b 2

1.

Prove

the following.

2
3

OP 2 ~OD =
2

a 2 -b\

The mid-point of PD lies on an asymptote.


PD is parallel to an asymptote. 5 The area of triangle OPD is %db.

6 Prove that the tangents at the extremities of a diameter are parallel to


the conjugate diameter.

7 Prove that the tangents at the extremities of two conjugate diameters


on the asymptotes.

intersect

and their diametrically


8 Deduce from nos. 6-7 that tangents at P,
opposite points P', D' form a parallelogram of constant area 4a6, whose vertices
lie on the asymptotes.
9 Deduce from no. 6 that the part intercepted on the tangent at
is bisected at P.

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

12 Find the equation of the hyperbola whose asymptotes are x 2y + 1 =


and dx-y + 2 =
and which passes through the point (0,1). [18.61,

Remark

(yff).]

*13 Given that the equation 2x % 5xy By 2 + x + lly 8 =


represents a
hyperbola, find the equations of the asymptotes and the equation of the conjugate hyperbola. [Choose A so that 2a; 2 - 5xy - 3y* + x + 1 ly A = represents

line -pair.]

18.7

Asymptotes as (oblique) coordinate axes

18.71

xy = c2

The product of the perpendiculars from a point Px on

Fig. 184

to the asymptotes x/a + yfb


<H

+ 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

'

and the equation becomes


x'y'

J(a

+ 6 2 ).

Omitting dashes, this can be written

xy
where

c2

|(a 2 + 6 2 ),

and

c2,

the standard equation of the hyperbola

is

referred to its asymptotes as (oblique) coordinate axes.

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

totes are parallel to

18.72 Parametric representation

The equation xy =

c2

can be written

=
so that x
(ct, c/t),

ct

and y =

c/t.

say,

*,

For each value of t except t = the point


2
t, lies on xy c
and to each point of

referred to as the point

the curve corresponds just one value of

t.

can also be written

x:y:c

l)

The parametric equations

: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',

prove that the chord

t x t^

has equation

x + t1 t2 y

and hence that the

tangent at

c(^

+ #a

is

x + thj

2ct.

),

<

THE HYPERBOLA

18.73]

659

Example

A tangent to a hyperbola and


totes form

the

asymp-

triangle of constant area.

The above tangent at P(ct,c/t) meets


Oy at Q(2ct, 0), B(0, 2c/t). Hence

Ox,

OQ =

2ct,

OR = 2c/t,
OQR is

and the area of triangle

^OQ.OR sin

(i)

2c 2 sin a>.

This work also shows that


is the midpoint ofQR; cf. the example in 18.53.

18.73

Fig. 185

The rectangular hyperbola

When

o)

= \n the asymptotes Ox, Oy are perpendicular, and the


= c 2 is rectangular (18.13). In addition to the results in

hyperbola xy

18.72, the usual

method now shows that the normal at t has equation

Px-ty = C(^-1).
Example
The

orthocentre of

triangle inscribed in

a rectangular hyperbola

lies

on

the

curve.

The chord tt% has gradient

similarly,

and only

chord
if

ta tt

1/<3 4

has gradient

These chords are perpendicular

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 hh-Lhh- Hence this point tt = l/Ma^s ^s the orthocentre of
,

triangle Ma^s'
It is also clear

from the relation M2M4 = 1 that the point


centre of triangle Ms*4> and so on. The set of four points tx t2 tz
,

the ortho-

*x

is

tt

on the curve

may be called orthocentric.


Exercise 18(c)
In the following examples the sign {<y} indicates results which are true for the general
hyperbola xy = c 2 The reader who wishes to avoid oblique axes may prove them
.

for the rectangular case only.

+ my + n =

{w}

The line

{to}

Tangents at

Ix

t lt t 2

is

meet at

a tangent

if

and only if AcHm = n 2

THE HYPERBOLA

660

The chord Px P s is c2x + xx x 2 y = c 2 (xx + xt


The tangent at Px is xyx + x x y = 2c 2 Could this be written down from
xy c2 by the 'rule of alternate suffixes'?
5 {o)} The chord of contact of tangents from P x is xyx +xx y = 2c 2
3

{(t)}

{(o)

*6

{to}

*7

+ (xyx + xx y-2c2 )kl+(x L yx -c 2 )l 2 = 0.


{(o) The pair of tangents from P x is 4(^2/! c 2 )(xy c2 = (xy x +x y 2c 2
x
The normal at P is aja^ yyx = x\ y\.

Joachimsthal's ratio quadratic

is

(x t y s -c 2 )k 2

2
)

9 The distance quadratic

is

+ (xx sin + y x cos 0) r + (a?!^ c2 = 0.


The chord having mid-point Px is xy x + x x y = 2x yx [Use no. 9 or no.
The diameter bisecting all chords of gradient m is y = mx.
y = mx, = m'a; are conjugate diameters if and only if m + m' 0.
2
[Begin as in 18.71.]
{&>} The conjugate hyperbola is xy = c
2

r sin 6 cos 6

10
11

12
13

3.]

2/

14 Conjugate semi-diameters are equal in length and make complementary


angles with the transverse axis; and the asymptotes bisect the angles between

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.

quadrilateral circumscribes a hyperbola. Prove that the line


joining the mid-points of its diagonals passes through the centre of the

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
right-angle at R, prove that the tangent at
is perpendicular to PQ.

20 Concyclic points.
(i)

Show

hyperbola xy

that the circle x 2 + y 2 + 2gx + 2fy+d


= c a at points t for which
c 2 *4

+ 2gctz + dt 2 + 2fct + c 2 =

cuts the rectangular

0.

If the points tx , t2 , ts , tt are concyclic, prove <i< a *8 $4 = 1.


(iii) By considering the fourth intersection of the hyperbola
through the points t x , t2 , ta , prove the converse of (ii).
(ii)

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.

22 If circles touch a rectangular hyperbola at a fixed point, prove that their

common chords not through this point lie in a fixed direction.


*23 (i) Prove that the circle of curvature at* meets the rectangular hyperbola
again at the point 1/t*.
(ii) Points Px , P , P a , P 4 on the rectangular hyperbola are concyclic. If
2
Q r is where the circle of curvature at Pr meets the curve again, prove that

Qi are also concyclic.


24 The points A, B, G, D on a rectangular hyperbola are not concyclic. If
the circles BCD, CDA, DAB, ABC meet the curve again at A', B\ C, D'
Qi> Qs> Q,

THE HYPERBOLA

661

respectively, prove that the mid-points of the chords

A A',

BB',

CC, DD'

lie

on another hyperbola having the same asymptotes as the given one.


25 Prove that the circumcircle of triangle tx 1 2 13 cuts the rectangular hyperbola
again at the point diametrically opposite to the orthocentre of this triangle.
[Use 18.73, ex.]

26 If a circle and rectangular hyperbola meet at A, B, C, D, prove that the


orthocentres of triangles BCD, CDA, DAB,
are concyclic. Prove also
that the tangents to the hyperbola at G,
meet at a point on the diameter
which is perpendicular to AB.

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.
.

P meets the rectangular hyperbola again at Q.

28 The normal at
the mid-point 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

=}=

The axes being rectangular, prove that the equation y =

(ox + b)j(cx + d),


represents a rectangular hyperbola whose asymptotes are x = djc,
a/c. What are the coordinates of its centre? [Use 18.71, ex.]

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

a hyperbola whose asymptotes are

parallel to the given lines. [Use oblique

axes.]

33 A variable chord of xy = ca passes through the fixed point (p,q). Prove


that the locus of its mid-point is another hyperbola, and give the equations of
its asymptotes.

Miscellaneous Exercise 18 (J)


the angle
1 If
eccentricity is sec %o>.
o) is

between the asymptotes of a hyperbola, prove that

its

PN is the ordinate of a point P on x /a - y jb = 1 NT is a tangent from


+ y = a If P is the point ^, prove NOT = and
2

N to the auxiliary circle x

<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

= ca projects into another rectangular hyperbola.

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.
.

[Use the converse bifocal property.]

THE HYPERBOLA

662

is a fixed point in the plane of a given circle of centre A, and


6
is a
variable point on the circumference;
meets the circle again at Q, and the
parallel to
through meets
at R. Prove that the locus of is an ellipse
or a branch of a hyperbola according as
is inside or outside the circle.

BP

AP

AQ

P to x /a y /b 2 = 1 meets the asymptotes at Q, R;


VR are parallel to the axes of the hyperbola. Prove that V lies on one of

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 rectangular hyperbolas xy


8
that

The tangents

PQ

at P,

to

9 The tangents at two points P, P' on a hyperbola meet one asymptote at


Q, Q', and the other at R, R'. Prove QR', Q'R are parallel.
10 For the hyperbola x = %a(t + -1 ), y = \b(t _1 ), find the equation of the
chord joining the points t given by ut2 + vt + w = 0.
1 1 Prove that tangents to y 2 = kx at (kt\, kt^ and (kt\, kt2 ) meet at the
point (kt^,
+ tj).
The tangent to y 2 = kx at any point
cuts the hyperbola x 2 ty 2 = k 2 at
U and V. IfQ,R are the points of contact of the other tangents from U, V to
the parabola, prove that the chord QR touches the circle x 2 + y 2 = k 2

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 mid-point of this chord lies on

touches the circle x2 + y 2

c2 .

Prove

14 The normal at a variable point P on the rectangular hyperbola xy c a


meets the asymptotes at Q, R. Prove that the mid-point of QR lies on the curve
4x3y 3 + c\x 2 y 2 ) 2 = 0. Explain why this is the same locus as in Ex. 18 (c), no. 28.
15 Prove that the tangent and normal at P on a hyperbola bisect the interior
and exterior angles (respectively) between SP, S'P. [Either prove SO = e SP
and use pure geometry (Ex. 17(c), no. 3); or use r' r= 2a and calculus
.

(8.14, ex. (iv)).]

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
,

other a hyperbola. [Express the equation as a quadratic in A, say /(A) = 0; its


discriminant reduces to (a 2 6 2 x 2 + y 2 ) 2 + 4x2y 2 which is positive (unless
,

THE HYPERBOLA
x

663

= a b and y = 0),
2
> 0,
0, /(-a
2

so there are distinct roots X 1 > A a Since /(oo) > 0,


the roots satisfy a 2 < A < -& 2 , -ft 2 < A respectively.] (From 5.72, ex. (ii) or from no. 17 it follows that this ellipse and hyperbola cut orthogonally at P.)

/(-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)

17.64, ex. (iv).J

Give the centre, semi-axes, eccentricity, foci, directrices, and asymptotes of the
following hyperbolas.

23

V-4z

25

144# 2 - 25y 2 + 50y=l 69.

36.

24 %{x+l) 2 -\(y + 2) 2

I.

664

19

THE GENERAL CONIC;

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

= ax 2 + 2hxy + by 2 + 2gx + 2fy + c =

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.

the last paragraph in 15.74. In this

be convenient to refer to the locus

methods

a parallel line-pair.
begin the chapter by applying to the locus
it is

as a 'conic',

circle or

the general

for finding chords, tangents, etc. already illustrated in-

dividually for the parabola, ellipse,

and hyperbola. The work will thus


and will contain as particular

be a summary and unification of conies


'

'

many results already obtained for the special standard forms of


equation of these curves. Had the general case been treated first,
cases

considerable repetition would have been avoided; but dealing with

the special forms separately increases the appreciation of the methods


themselves.

Some

of our work will apply unmodified

represents a line-pair.

remain

when

The reader should consider whether the results

significant in this degenerate case.

The second part of the chapter

is

independent of the

first,

and

is

concerned with a principle which permeates the whole of coordinate


geometry. The student who is pressed for time may turn at once to
19.5 etseq.

19.12 Notation

To

simplify the writing

(a)

We continue to put

(6)

We

8i}

we introduce the

= ax 2 + 2hxy + by 2 + 2gx + 2fy + c.

write

= axi xj ^h{xi yj + x^) + byt y} +

Clearly siS

following notation.

sj{

+ xs +f(y4 + y} ) + c.
)

THE general conic; s=ks'

19.13]

The

(c)

by

fy;

result of omitting the suffix j throughout in (b) is

denoted

thus
8i

This

665

is

axi x + h(xt y + xyi ) + byi y + g(x + xt ) +f(y + yi) + c.

what would be obtained from

by applying the

'rule of

alternate suffixes' (15.63, Remark).

For example,
sn

s 12

= axx x2 + h(xx y 2 + x2 y x + by x y 2 + g(xx + x 2 +f(yx + y 2 + c,


= ax\ + 2hx2 y 2 + by\ + 2gx2 + 2fy2 + c,

s22

ax\ + 2hx1 y 1 + by\ + 2gx x + 2fyx + c,


)

sx

ax x x + h(xx y + xyx ) + by1 y + g(x + x 1 ) +f(y + y 1 ) + c,

82

ax 2 x + h(x2 y + xy 2 ) + by 2 y + g(x + x 2 ) +f(y + y 2 ) + c.

Notice that

and

st

(axt + hyt + g)x +

Si

(ax + hy+g)xi + (hx + by +/) y<

(fee,

+ byt +f) y + {gxi +fyt + c),


+ (gx +fy + c).

There are similar forms for sif


is linear in x and y, the equation
Since
.

The geometrical interpretation of s,


5

is

n=
5 12 =

is

Not only does

represents a line.

the equation of the conic;

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

discussion of the general conic as

concise as for the various special forms, but

it

provides a suggestive

means of obtaining and memorising some of the results themselves;


e.g. see 19.13. The reader should check each of the following general
results with the corresponding ones obtained for the special equations

used in Chs. 16-18; but

19.13

first

do Ex. 19

(a),

nos. 1-3.

of s =

Chord

Consider the equation


^1

It

is

linear in

a;

and

y,

^'2

*12*

and therefore represents a

line. It is satisfied

THE GENERAL CONIC; S = ks'


= because Px lies on s = 0; and it is

666

by

Px

for sxx

since s 22

Hence

0.

P2

Letting

Px is s =

at

we

along the conic,

the above equation,


s

the equation of the chord

it is

2s

i.e.

0.

JoachimsthaPs ratio equation

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\ +

the equation of the tangent to

+ kx 2 2 + 2h{lx x + kx 2 (Z^ + ky 2 + b{ly x + ky 2 2


+ 2g(lx 1 + kx2 (l + k) + 2f{ly x + ky 2 (l + k) + c(l + kf =

a(lx x

by

ratio quadratic for s =

The point

i.e.

satisfied
.

0.

19.2

It will lie

obtain 2s x

Hence

PX P2

[19.2

0,

2hx2 y2 + by\ + 2gx2 + 2fy2 + c) k 2

+ 2{ax x x 2 + h{x x y 2 + x 2 yx + byx y 2 + g{x x + x2 +f(y x + y 2 + c} kl


+ (axl + 2hxx yx + by\ + 2gxx + 2fyx + c) Z 2 = 0,
)

s22 fe2 + 2s 12 fe/ + s 11 /2

i.e.

This quadratic in k:l

P]P2 and

0.

0.

Its

Px A ^4P2 for
intersection A of the

roots are the values of

the points of

Joachim-

is

sthaVs ratio equation for 5

line

the conic.

Since the

quadratic will have either two, one,


or no roots, f a line meets a conic in
two, one, or

Kg- 186

no points.

The above work applies even if s = represents a line-pair.


We now make a sequence of deductions from the ratio equation.
19.22 Sides of a conic

When

and s22 have opposite signs, the product of the roots is


Hence just one root is positive, so that the conic meets
internally only once. We say that Px and P2 lie on opposite sides
sxl

negative.

PP
X

of the conic. Also see the


f Unless sn
all

k and

I,

i.e.

= 0,

s lz

every point

and
if

Remark in
28

PP
X

= 0,
2 lies

19.24.

which event the quadratic is satisfied


on s = 0, which is therefore degenerate.

in

for

THE GENERAL CONIC;

19.23]

19.23 Tangent at

= ks'

667

Px is a fixed point on s =

and so one root of the


0, then s n =
The second root, which corresponds to the
remaining intersection of PX P2 and the conic, will also be zero if and
only if s 12 = 0. In this case P2 will lie on the tangent at Px and the
condition s 12 =
shows that P2 lies on the line s x = 0. Hence the
If

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.

+ hy1 + g)x + (hxx + by x +f)y + (gx x +fy x + c) = 0,


and will represent the same line as Ix + my + n = if and only if the coefficients
(ax x

are proportional,

and

for

i.e. if

some A we have

=
=
gx 1 +fy 1 + c-Xn =
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,

A from these, we obtain


a

0,
0.

(using

an obvious extension of

0.

When expanded this becomes (see Ex. 11 (d), no. 12)


Al z + Bm* + Cn 2 + 2Fmn + 2Gnl + 2Hlm =
19.24 Pair of tangents

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

He on any tangent from Px but this equation shows that P2 lies on


s lx s = s{. The equation of the tangents from Px is therefore s n s *f.
\

Remark. By writing s n s = s\ in full, it can be shown by using 15.53 (1) and


15.52(1) that this second-degree equation will represent two distinct intersecting lines, a repeated line, or the single point x according as As n
0, where

denned by equation

in 15.53(2). Hence for a non-degenerate conic


(i.e. A + 0), s lx s = s\ represents a pair of lines ifP lies on that side of the conic
1
for which As n < 0. The outside of the conic s =
can therefore be denned
algebraically as the set of points x for which As n < 0.
is

(vii)

45

GPM II

THE GENERAL CONIC;

668

= ks'

[19.25

Chord of contact from P,

19.25

P (supposed outside the conic)


P P The tangent at P nas equation s2 = 0; and since
it passes through P we have s
= 0. This shows that P2 lies on the
v
12
line 5 = 0. Similarly s 13 = 0, so that P3 also lies on s = 0. Hence the
equation of P P3 is s = 0.
Method 2. The points of contact of tangents from P satisfy both
Method

Let the tangents from

1.

touch 8

at

2,

= and s xl s = s\. Hence they also satisfy s x = 0. Since s x is linear,


s =
is the equation of the chord of contact of tangents from P
v
s

19.26 Examples; polar of P,

wo s =

(1) Tangents at the extremities of a variable chord through P x meet on s x = 0.


Let the tangents at the extremities A, B of such a chord meet at P2 Then
the chord of contact from P2 is AB, whose equation is therefore s 2 = 0. Since
this line passes through Px s 12 = 0; and this condition shows that P2 lies on
.

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

Let a chord through x cut the conic at


and B. The point a such that Pj
divide
in the same ratio {one internally and the other externally) lies

0.

2 is

called the harmonic conjugate of

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

+ hy1 + g)x + (hxx + by 1 +f)y + (gx x +fy x + c) = 0.


= and hxx + by x +/ = 0; in general

ax k + hy + g

x.

is

sx

is

P wo

there

satisfying these conditions.

The argument shows (without

line-pair s

not exist

unique point
(3)

outside s, the polar coincides with the chord of contact from


on s, the polar coincides with the tangent at x
the polar of x is

sx

(ax + hy + g)x 1

0.

modification) that the polar of

Since

+ {hx + by+f)y l + (gx +fy + c),

the polar of x always passes through the vertex of the line-pair (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 line-pair 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

Chord whose mid-point

19.27

is

P,

Let the extremities of the chord be A and Pa

and the ratio quadratic has one root

2k8 li + U 11

0.

0,

(fig.

188)

The other

Then s 22 = 0,
by

root, given

A and since A divides P1 P2 externally in the


must be k\ls= \. Hence s 12 = n which shows
that P2 lies on *ne locus s x = s n Since this is linear, and is also satisfied
by Pl5 it is the required chord. Thus the chord ofs = wJiose mid-point
corresponds to the point

ratio

2, this

root

is

is s t

= su

Example
Show

that the mid-points of all chorda of s

the curve s

sx

which pass through

x lie

on

Let 2 be the mid-point of such a chord, which will have equation * a = aa


ia = s 2t This shows that
Since the chord passes through
x
a lies on the
locus s x = s which, being of second degree, is also a conic; clearly it passes
.

through

Pr

19.28 Diameters

Let

Px be the mid-point of a chord of gradient m. Then this chord

has equation

45-2

THE general conic; s = ks'


Its equation is also s x = s xx which can be written
670

[19.29

ax x (x - x x ) + h{xx (y - yx ) + yx (x - x x )} + byx (y - y x )

+ g(x-x 1 )+f(y-yx =

0.

Hence

which shows that

lies

on the

(ax + hy + g)

By

ax x + h(mx x + yx ) + bmyx + g +fm

0,

line

+ m(hx + by+f) =

0.

definition (16.41) this is the equation of the diameter bisecting

chords of gradient m.

all

19.29 Conjugate diameters

The diameter

bisecting all chords of gradient

m has gradient

a + hm
h + om
a + h(m + m') + bmm'

so that

The symmetry
gradient

in

0.

m, m' of this relation shows that the diameter of

bisects all chords of gradient m'.

Two

such diameters

are conjugate.

19.3

The

The

distance quadratic

line

through

x
It

meets

Px in direction 6 has parametric equations


x x + r cos 6,

y = y x + r sin 6.

at points for which r

is

(i)

given by

a(xx + r cos d) 2 + 2h(x x + r cos 6) (yx + r sin 6)

+ b(y x + r sin d) 2
+ 2g(xx + r cos 6) + 2f(yx + r sin d) + c =

i.e.

+ 2h cos 6 sin 6 + b sin 2 6)


+ 2r{(axx + hyx + g) cos 6 + (hx x + byx +/) sin 6} + s xx =

0,

r 2 (a cos 2 6

0.

(ii)

Px of the points
Px in direction 0.

This quadratic in r gives the signed distances from


of the conic s

which he on the

line

(i)

through

Examples
Chord whose mid-point 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 *

(axx + hyx + g) cos 6 + {hx x + by x + /) sin 9

0,

THE GENERAL CONIC;

19.4]

= ks'

which gives the direction cos# :sin# of such a chord. If


this chord, then from equation (i),
cos 6

sm.6

671

(x,y)

is

any point on

'

+ g)(x-xx + (hx 1 + byx +/) (y - yx = 0.


which has mid -point Px
This is therefore the equation of the chord of s =
it can be arranged in the form s x = s u (cf. 19.27).
and

(axx + hyx

so

(ii)

Segments of a chord.

If the line (i) cuts the conic at Q, R, then


roots of (ii) we have

P Q.Px R = rx r2 =
t

(a)

by considering the product of the

a cos2 d + 2h

eg

n e + b gin2

For chords QB, Q'B' through P x in given directions

6, 6', it follows

that

P Q.P R _ a cos2 6' + 2h cos B' sin d' + b sina 6


P Q'.P R' ~ acos + 2/&cos0sin0 + &sin
which is independent of the coordinates of P Hence the ratio of the product of
'

'

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

which is independent of 6. Hence the ratio of the products of the segments of


two parallel chords drawn through given points Px P2 is constant for all
,

directions.

19.4

Tangent and normal as coordinate axes

Sometimes

it is

convenient (as for example in 8.42(1)) to choose

the tangent and normal at a particular

yj

point for axes of coordinates; this point

thus becomes the origin.


If s

passes through 0, then c

0.

= touches y = at (0, 0), the equation


ax% + 2gx + c = must have both roots zero,
so that also g = 0. The equation of the conic

If

is

therefore

ax

o
2

+ 2hxy + by + 2fy
2

0.

Fi8- 189

Example
Fregier's point.
is a fixed point on a non-degenerate conic s = 0, and PQ is a chord which
subtends a right-angle 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.

THE GENERAL CONIC;

672

= ks'

[19.4

Choosing axes as above, let PQ have equation Ix + my = 1. (This form of


equation for PQ is legitimate since the hypothesis that PQ subtends a right angle at O implies that PQ does not pass through 0.)
The line-pair OP, OQ has equation
l

'

ax 2 + 2hxy + by 2 + 2fy(lx + my)

by

15.54. Since

OP JL OQ, then from

(i)

0,

15.52 (2)

a + (b + 2fm)

0.

4= 0, then m = (a + 6)/2/+ 0. Hence Ix + my = 1 passes through


(0, 2//(a + 6)), which is a fixed point on Oy, the normal at O.
If a + b = 0, then the perpendicularity condition becomes fm = 0. If / = 0,
the conic s =
is an orthogonal line -pair
excluded by hypothesis; in such a
= 0, the equation of PQ
case every 'chord' subtends a right-angle at O. If
is Ix = 1, which is a line parallel to Oy.

If a + b

Exercise 19(a)
Write out in full the expressions for & xx s Xi when s denotes
,

^ + ^-1-

2 y -4ax.

xy-cK

+ my + n)(l'x+m'y + n'); and if u Ix + my + n, u' = I'x + m'y+n',


s ia = ^{u-^u^ + u^u ^.
5 Interpret the equation slx + 2s12 + s = 0.
6 Obtain the equation of the tangent to s = at Px (i) by Calculus; (ii) from
s n = si by taking P x on s = 0.
7 Prove that the normal to s = at Px has equation
(2/ - Vx) (o^i + %i + g)~(x- x x (hx x + by x +/)
8 Find the condition for the lines px 2 + 2rxy + qy z =
to be parallel to
conjugate diameters of * = 0.
9 Prove that the line-pair joining O to the meets of x 2 + y 2 = r 2 and
ax 2 + 2hxy + by 2 = 1 is
*4

(lx

show that

2iS

a-^jx 2 + 2hxy+(b-^jy =
2

Show

0.

that these will be conjugate diameters of the conic


1

2
_ 2{ab-h

if

a+b

r2

Hence obtain the equation of the equi-conjugate diameters of

ax 2 + 2hxy + by 2
10 Writing

and
Use
11

w = gx +fy + c, verify that


= ux + vy + w, s x = u1 x + v y + w = ux + vyx + w,
s xx = u x +v y x + w
1

the distance quadratic for s

no. 10.

x.

in nos. 11-13.

Obtain the equation of the tangent at

by using

1.

u = qx + hy + g, v ~ hx + by+f,

x,

and reduce it to the form s x =

THE GENERAL CONIC;

19.5]
12

{ux (x - xj)

+ vx {y - yx )} =
2

Using no.

(ii)

and hence
1

673

to s\

10,

show that this is equivalent to (s l

= 8Sn

is

{a(x -xtf + 2h(x - Xl )(y- yi )

+ b(y- y^} s u
s 11 2 = (s 2s1 +811 $n
.

A line through O cuts 8 =

Prove that the equation of the pair of tangents from Px

(i)

OR is (i) the arithmetic

that

= ks

at

P and Q. A point R is chosen on this line so

mean;

(ii) the geometric meanj (hi) the harmonic


mean, of OP, OQ. Find the locus of R in each case.
is the point ^ on the ellipse x 2 /a z + z /b 2 1. Write down the equation
14
y
of the chord PQ for which KP, KQ are parallel to the axes, and find where it
meets the normal at K. Deduce from the example in 19.4 that chords which
subtend a right-angle at the fixed point are concurrent at

<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

through a fixed point on the tangent at

parallel to this tangent.

Number of conditions which a

19.5
If k

is

conic can satisfy

any non-zero constant, the equations

and

ax 2 + 2hxy + by2 + 2gx + 2fy + c

Ics

Jcax 2

represent the

same

the other. Since

+ 2khxy + Jcby 2 + 2hgx + 2kfy + he =

locus; for if

a, b, c, f, g,

so that one coefficient in ks

h are not
is 1.

then

satisfies one,
all zero,

it

also satisfies

we can always choose h

The remaining

five coefficients

(i.e.

the five ratios a:b:c:f:g :h) can be chosen so as to satisfy five conditions,

but in general no more. The equations which express these


may have more than one set of solutions but, since they

conditions

are polynomial equations, the

number of

be

sets will

finite if

the

conditions are independent.


It follows that a conic can be chosen to satisfy five independent

conditions,

and that the number of such

conies

is finite.

Examples
(i)

(r=

In

general a unique conic can be

drawn through

five given points

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

the equation of the required conic.

(r

1,

. .

.,

5),

THE GENERAL CONIC;

674

= ks'

[19.6

c, f, g, h can be eliminated from the above equations


giving the required equation in determinant form.
of finding the equation of a conic through five given points

Alternatively, a, b,

together with s

An
will

easier

way

0,

be illustrated in 19.64, ex.

(i).

If three of the five points in ex. (i) lie on a line


which therefore consists of two lines I, V.

(ii)

conic,

I,

then

must be part of the

For no line can meet a non-degenerate conic in three points (see the footnote
on p. 666); and the only line-pair 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'

of possible intersections of two conies

The equations of the two given


s

s'

conies can be arranged in the

+ 2fy + c) = 0,
+ 2(h'y + g')x + (by + 2f'y + c') =

ax + 2(hy + g)x + (by


2

a'x

form

0.

The coordinates of any common point of these must satisfy the


equation in y obtained by elimination of x. Since this equation will
in general be quartic in y, there are 4, 2 or

(apart from possible coincidences).

By

no possible values of y

eliminating x 2 only,

we

see

that to each value of y corresponds at most one value of x, given by


2{(a'h

-ah')y + (a'g - ag')} x

+ {(a'b - ab') y 2 + 2{a'f- af) y + {a'c - ac')} =


Hence, ignoring coincidences,
two conies

the

number of

0.

possible intersections of

is 4, 2 or 0.

19.62 s = ks'

This equation has already been discussed


sent lines (15.41) or circles (Ex. 15
(1)

If s

0, s'

the equation of a conic,

are conies

is

and k

when

and s' =

constant, this last equation is

0, s'

repre-

no. 16).
is constant, then s

which passes through the meets

For if P satisfies both 5


Since k

(d),

ks' is also

{if any) ofs and s'

then it also satisfies s = ks'.


of second degree and therefore
0,

represents a conic (or nothing).


(2)

// s,

s'

meet in four points A, B, C, D, then any conic

these points has

If

Px

is

on

an equation of the form

cr,

o~

through

ks'.

k can be chosen so that s

ks'

passes through

lt

THE GENERAL CONIC;

19.63]
viz.

= sn ls'u

(assuming that

below). For this k, the conies

and hence coincide

is

not on

<r and s =

(19.5, ex.

= ks'

s'

675
the

0: see

ks' have five

points in

Remark

common,

(i)).

may

Exceptions to the last statement

arise

when

three of the five

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(fi-ky)

s-Jcs',

and the theorem is still valid.


Remark. The only conic through the meets A, B, C, D of s and s'
which does not have an equation of the form s = ks' is s' itself. This
exception can be avoided by using As = AV, where A and A' are
independent constants; but usually the equation s = ks' is adequate.
For different k or A A', the equation s = ks' or As = A represents
a family or system or pencil of conies through the intersections (if

any) of s and

s'.

19.63 Degenerate cases

The results in

19.62 apply

when one or both of s,

s'

degenerate, and

these cases are particularly useful.


(i)

kecfi.

This represents a conic through the meets

of the distinct lines

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

its intersections (if

to have double contact with s along

a =

the equation of a conic

any) with a

(fig.

191).

0.

It

is

said

THE GENERAL CONIC;

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

When ft - t in (i), we have


= 0, and touching

any) with a

192).

If one meet of a and s coincides


with one meet of ft and s in (i), we obtain a conic meeting s at a point
on a and at a point on /?, and touching 8 at the meet of a and /? (fig. 1 93).
(iv) s

kaft,

where a,

Fig. 196

ft meet on

s.

Fig. 197

(v) s = foxT, where a passes through the contact of r. If one meet of


a and s in (iii) coincides with the contact of r, we obtain a conic meeting
s at a point on a and touching s at the meet of a and r (fig. 194).
2
(vi) s = Jct
If a -> t in (v), we get a conic touching s at its
.

contact with r
(vii) aft
(fig.

196).

(fig.

195).

= kyS is

a conic through the meets of the lines a,

ft,

y, 8

THE general conic; s=ks'

19.64]

= ky 2

677

y in (vii), we have a conic touching the lines


<x, /? at their meets with
y (fig. 197).
2
2
represents the two lines a = ^/fcy, which pass
(ix) a = ky
through the meet of a, y if this exists.
In the next two cases s' represents a single line.
(x) s = kcc is a conic through the meets (if any) of s and a.
(viii) a/?

(xi) s

&t

is

If S ->

a conic touching

at its contact with

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.

no. 15; Ex. 16

(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):

Hence any conic through the first four points

This passes through (15,

(ii)

Find

is

1.8 = k. 15.6, i.e. k =


_ 3y _ 6){x + 5y _ 5) = 8xy

and

Ix

+ my + n =

0;

0.
(vii))

The required

conic

formed by

the lines

0.

Let the tangent at O to the required circle be px + qy 0. Then by


equation will be
ax2 + 2hxy + by 2 (Ix + my + n)(px + qy)

1 9- 63, (iii)

its

provided

we choose p and q so

that

apl b qm and

2h

= mp + Iq.

If we arrange the above three equations in terms of p


(Zoj

By

and

q,

we have

+ my + n) xp + (Ix + my + n)yq (ax + 2hxy + by =


Ip mq (a b) = 0,
)

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

as the equation of the required circle.

is

the equation of the circumcircle of the triangle

ax2 + 2hxy + by2

(by 19.63,

6) if

therefore

(0, 1), (15, 6).

0,

THE GENERAL CONIC;

678
(iii)

= ks'

[19.64

Goncyclic points on the ellipse

l + yl _

The chords 0i0 2 and


a

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 ).

be concyclic if and only if one conic of the system

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)

+ 0J + cos J(0 3 + 4 sin


+ 2 + 3 + 4 = 0,
+ 02 + 03 + 04 = 2rwr

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

are then equal

?/

2)

cos |(0 3 +

4)

if

- ^ sin (0 X +

2)

sin |(0 3

4)

& j-^ cos

( 0!

2)

+ ^ sin

J( 0!

2 J

cos nn ,

and here the coefficient of k is non-zero.


Hence if S0 X = 2nn, k can be chosen uniquely to satisfy the second condition.
Thus the relation S0 X = 2nn is necessary and sufficient for the points x 2 3 4
of the ellipse to be concyclic. Compare the example in 17.32.
<j>

*(iv)

Normals

at the extremities of the chords Ix

ellipse

The

a2

b2

conic

a2

b2

+ my =

1, 1'x

+ m'y =

of the

- + vf--l = k{lx + my-l){l'x + m'y-l)

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

must belong to the above system.


and the constant terms, we have

no. 10. Consequently this conic

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),

normals at the extremities of the chords

are concurrent, then

aHV =

b %mm'

0.

THE GENERAL CONIC;

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.

Since the equations of the chords of contact AB, CD are respectively s x = 0,


= 0, the conic is 8 = Ax * a by 19.63, (i). This passes through lt so s u = ks n sl2
Clearly this is satisfied
hence k = l/s12 , and the conic has equation s12 s =

byP2

*(vi) Pascal's theorem.

on a

If

1, 2, 3, 4, 5,

6 denote six points

conic, then the meets of the lines


12, 45;

23, 56;

34, 61

are collinear.

For suitable k and I, the line-pair (12) (56) has equation


k{25) (16), and the line-pair (23) (45) has equation
Z(25)(34). The points X(12,45) and F(23,56) lie on
both pairs, but do not lie on s = or on (25). Hence they
lie
on the line &(16) = i(34), which passes through

=
s =

Z(34, 61).

The argument holds even when * = is a line-pair 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

Although in general two distinct conies intersect in at most four


A,B,C, D, there may be coincidences of the following types:

points

AAGD
AACC
A A AD
AAAA

2-point contact at

double contact along

(figs.

AC

A
4-point contact at A
3-point contact at

192, 193),

(fig.

191),

(fig.

194),

(fig.

195).

Since a circle can be determined to pass through only three general


points,

we may expect that in general a

3-point contact with a conic at


limit

when

and cutting

it

By

circle

cannot have more than

regarding such a circle as the

along the conic of a circle touching the conic at

at P,

we

see

from 8.42 (2) that

the circle having 3-point

A is the circle of curvature of the conic at A.


= at A is s = Tear, where
A
=
=
is any chord through A. If
is the tangent at A to s and a
r

contact at

conic having 3-point contact with s

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

a and t must be equally

inclined to the axes.

t For examples of exceptions, see Ex. 19(6), nos. 9, 10.

THE GENERAL CONIC;

680

S = ks'

[19.7

Example

= 4ax at the point


= x ty + at 2 = 0, and
2at) with gradient ljt is

Find the circle of curvature of y 2


The tangent at the point t is t
chord through

{at 2 ,

y 2at

t.

The required

- 4ax =

The

x 2 and y 2 are equal (the


k = 1/(1 + t 2 ), so

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.

already absent). This gives k

circle is

curvature

k(x + ty- Zat 2 ) (x-ty + at 2 ),

where k must be chosen so that the


is

The

circle is

y
xy-term
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

found to be 2a(l + 2 )*. These results can of course be obtained by

the methods of Ch.

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 mth-order 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.

Equations of a type more general than

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

The problem in 15.54 is of this type.


Sometimes the equations are combined in such a way that the deductions

constant.

do not hold

for all the

common points.

Examples
Circle through the feet of conormal points

(i)

The

where

(h, k) is

Hence they

4ax

and

xy + (2a h)y 2ak

the point of concurrence of the normals

also satisfy

xy2 + {2a _ h)yi=z 2aky>

+ (2a h)y 2 =
4a( a + y 2 )-(2a + h) 4ax x 2 +y2 -(2a+h)x =
x(4kax)

and
This

on a parabola.

points in question satisfy the equations

is

the required

circle; cf. 16.32, ex.

(ii).

2aky,
2aky,
iky.

0,

(see 16.32, ex.

(iii) (6)).

THE GENERAL CONIC;

19.7]
(ii)

Find

= ks'

681

the equation of conies through the meets, other than the origin, of

and

s'

ax 2 + 2hxy + by 2 + 2gx + 2fy


a'x

+ 2h'xy + b'y + 2g'x + 2f'y =


2

0,

where fg' *f'g.


The given equations can be written

= -y(by + 2f),
= -y(b'y + 2f).

x(ax + 2hy + 2g)


,

x(a'x + 2h y + 2g')

Coordinates other than


<r

(0, 0)

which

satisfy

(ax + 2hy + 2g) {b'y + 2f)

both these equations also satisfy

- (a'x + 2h'y + 2g') (by + 2f) =

0.

This is therefore one conic through the meets of*, s', and does not pass through O.
Clearly
<r

+ ks + k's' =

(i)

also passes through the same points for all k, k'.


If 8, s' meet in three points other than O, then k, k' can be chosen to make the
conic (i) pass through two other general points. Hence any conic through the
three points has an equation of this form. The system (i) is called a net of conies.

Exercise 19(6)

Find

the equation of the conic through

(0,0), (0,1), (3,0), (2,1),

(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

4 Find the common chords of x 2 + y 2 = 25 and x 2 + xy + y 2 36.


5 Two chords are drawn through a focus of an ellipse, and a conic is drawn
through their extremities and the centre of the ellipse. Prove that this conic
cuts the major axis in another fixed point.
6 (i) Prove that, for any c,
(l+t 2 )(y 2 -4ax) + (x-ty + at 2 )(x + ty + c)

the equation of a circle which touches y = Aax at the point t.


(ii) PSQ is a focal chord of a parabola.
Circles are drawn through the
focus S to touch the parabola at P, Q respectively. Prove that these circles
cut orthogonally.
2

is

A circle has double contact with ax + by + c =


2

contact

is

parallel to either

or y

0.

2
0. Prove that the chord of
[Let the chord be Ix + my + n = 0.]

8 Find the equation of the circle touching the ellipse x 2 /a 2 + y 2 /b 2 =


end of a latus rectum.
9 Find the equation of the circle which has 4-point contact with

x2

circle

has equation s

at each

- + Vv- =
[The

kr 2 where t
,

at(,0).

= x a.]

10 Show that a circle can have 4-point contact with a parabola only at the
vertex.

THE GENERAL CONIC;

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

cos0 -sin0 cos 20 J

0,

and find its centre and radius.


12 Find the circle having 3 -point contact with xy =
deduce the centre and radius of curvature at this point.

ca

at the point

13 Prove that the circle of curvature of ax 2 + 2hxy + by 2


a{x 2 + y 2 ) = 2y.

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

By subtraction, deduce that E0 t =


condition for the points

*15

(i)

1?

2,

+ COS J(0j + cos (0 X -

2)

OS (0 3 -

4)

2)

cos (0 3 -

4)

0.

(2n +

03> 04 to

1) ir is a necessary (but not sufficient)


be conormal. (Cf. 17.52, ex.)

Write down the equation of a conic circumscribing the quadrilateral

whose opposite

sides are the lines a,/?; y, 8.


the length of the perpendicular from
to the line a = 0, prove
that p = constant x a
(iii) If a conic circumscribes a quadrilateral, prove that the ratio of the
product of perpendiculars from any point
of the conic onto two opposite
sides, to the product of perpendiculars from
onto the other two sides, is
constant (a result due to Pappus).
(ii)

If p

is

*16

Show

that the feet of normals from

x 2 + (2a-h)x

iky. [See 19.7, ex.

(h,Jc)

to y 2

4ax

lie

on the parabola

(i).]

*17 Explain why the conic H\(x t 2 y + at\) (x tz y + at\) =


passes through
the vertices of the triangle formed by the tangents to y 2 = 4ax at t x t2 3
Find the ratios fc x k 2 k 3 for which this conic is a circle, and verify that the
circle passes through the focus. State this result as a geometrical theorem.
,

Miscellaneous Exercise 19(c)


2
2
1 A pair of tangents to ax + by = 1 intercepts a constant length 2c on Ox.
Prove that the point from which the tangents are drawn lies on the curve

by 2 (ax 2 + by 2 ~

1)

ac 2 (by 2

l) 2 .

2 Prove that chords of y 2 = 4aa? which subtend a right-angle at


are concurrent at (a(jj + 4), 2a ). [Use 19.4, ex.]
3

Through a fixed point O a

line is

drawn

to

meet a conic at

A point Q on OPP' is chosen so that


1

OQ 2
Prove that Q

lies

on a

OP

1
2

OP' 2

'

conic. [Use the distance quadratic]

(at\,

2at

and

P'.

THE GENERAL CONIC;

683

ks'

4 If ax 2 + 2hxy + by 2 + 2gx + 2fy + c =


represents a line -pair, prove that
the pair (other than the coordinate axes) passing through the meets of these
with the axes has equation ax 2 + 2(2fg/c h)xy + by 2 + 2gx + 2fy + c = 0.
in four points
5 The lines 6a; 2 + ay 2 =
(a + 6 4= 0) meet the conic s =
A, B, G, D. If the diagonals not through O of the quadrilateral ABGD are
perpendicular, prove that
h
ab
g

b-a 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

7 Prove that the points of intersection of x 2 2xy + By 2 + 9x 16y + 24


and 5a; 2 + 6xy y 2 43a; + 60y 62 = are concyclic. [Eliminate xy.]

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 line-pair y 2 = 2x 2 Show that when
this rectangular hyperbola cuts the ellipse orthogonally.

ellipse

9 A circle through the origin touches the rectangular hyperbola xy = c 2 and


meets it again at P, Q. Prove that the foot of the perpendicular from the origin
to

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

+ (tx + y- at 3 - 2at) (6x + y- ad 3 - 2a6) =

Prove that its axis is parallel to SN, where S is the focus


the point where the normal at
toy 2 = 4ax cuts Oy. [In

/4dt.

the last part use 6

is

= t 2/t.]

P to the conic obtained in ex. (v) of 19.64 is the


AB and CD. [Use Ex 19 (a), no. 4.]

*12 Prove that the tangent at


Px to the meet of

line joining

46

0,

GPM II

684

20

POLAR EQUATION OF A CONIC


The

20.1

straight line

Polar coordinates and equations have been used in various parts

The purpose of this


and use the polar equation of a conic to prove

of the book, particularly in calculus applications.

chapter

is

to obtain

We

geometrical properties, especially focal ones.


sections

on the straight

line

and

circle, in

begin with short

which we assemble the

material required.

20.11 Distance formula

Given two points A(r v X ) and B(r 2 ,d 2 ), the length of AB can be


found by applying the cosine rule to triangle AOB (fig. 199):

AB2 = r\ + r\ - 2r

t r2

cos {6 1 -

2 ).

Fig. 200

Fig. 199

20.12 Line joining two points

Let P(r, 6) be any point on AB. If P


(fig.

200)

sin (ft,-^)
i

\r x r sin (0 - d x )

If

P lies

show

between

and B, then

^ AQB = ^ AQp + ApaBj


lrx r2 sin {6 2 -d x )

i.e.

lies

on

AB produced

similarly

by

+ \rr 2 sin {0 2 -6),

sin(0 2 -0)

rinff-flj
|

r2

or on

rx

BA

produced, the reader should

considering areas that the

same

result holds.

..

POLAR EQUATION OF A CONIC

20.13]

685

is satisfied by any point (r, 6)


AB, it is the equation of the line.
Remark. The line joining O to the point (p, a) has equation =

Accordingly, since this equation

of

the line

This fact, already noticed in 1.63

(c), is

obvious from

a.

first principles.

20.13 Line in 'perpendicular form'

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.

15.28) is the equation of the line in 'perpendicular form'.

Fig. 201

Remark. The foot

of the perpendicular from

has polar co-

ordinates (p,oc).

20.14 General equation of a line

When converted to
Ax + By = C becomes

polar coordinates, the general linear equation

Aeosd + Bsvcid =

(iii)

Example
Any line perpendicular to

(iii)

has an equation of the form

Acoa(d + $7r)+Bsin(d + in) =

Ax+By = G has equation


-Bx + Ay + C = 0,

For any line perpendicular to

Qf

i.e.

i.e.

AaiD.d + Bcos6

=
r

Acos(0 + 7r) + jBsin(0 + 77)

C
r
46-2

POLAR EQUATION OF A CONIC

686
20.2

The

[20.2

circle

20.21 Polar equation

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(d-oc).

has equation
r2

2pr cos (6 a) a 2

2
.

P(r, 6)

Fig. 203

Fig. 202

We notice the following particular cases.


(a)

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

oir = 2a cos 6 tangent


;

Let the required chord have equation


therefore satisfied

by

(2a cos

lt

2a cos d x cos (6 X a)
.

cos (20,

at

rcos (d a). This

p=

6 X ) and (2a cos 6 2

=p=

a) + cos a =

2 ),

so that

2a cos d 2 cos (0 % a),


.

cos (20 2 a)

is

+ cos a,

(i)

POLAR EQUATION OF A CONIC

20.23]

and hencef

(20 2 a),

26 1 ex.

a=

i.e.

From

(i)

+ 6%

we now have p 2a cos 61

chord becomes

By letting

n1 da 2

ID
r cos (0

-> 0!

687

cos

2>

and the equation of the

n
a
2a
cos X cos u %

(u)

we obtain the equation of the tangent at Pv


r cos (0 -

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^

and hence by the Remark

in 20.13, the foot of the perpendicular from O to


Similarly, the other feet are

BG has polar coordinates (2a cos ft cosy, ft + y).


(2acosy cosa, y+a),
These three points clearly

lie

is

(ii)

= rcos(0 a ft y).

known as the Simeon line of O wo

Extension of

ex.

(i).

If

cos/ff, oc+fi).

on the line whose polar equation is

2a cos a cos ft cosy


This

(2acosa

triangle

(iv)

ABC.

D is another point on the circumcircle of triangle

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

corresponds to the value 6


(2a cos a cos ft cosy, oc+ft+y). Similarly, if
the feet of the perpendiculars from O to the other Simson lines are
(2acosft cosy cos#,

ft+y + S),

(2a cosy cos S cosa,

(2a cos 8 cos a cos ft, 8


Clearly these four points

lie

on the

is

8,

y + 8+a),

a +ft).

line

2acosa cos ft cosy cos#


This result

rcos(0 a ft y 8).

likewise capable of extension.

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 (1687-1768).

(ii) is

POLAR EQUATION OF A CONIC

688

[20.3

Exercise 20(a)
1

(6)

Sketch the figures for the eases in 20.12 when

BA produced. Verify for each that the equation

Show

P lies on (a) AB produced;


(i) still

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

6sin(0 a) cut ortho-

gonally.

6 Find the centre of the


7 If OPQ

is

circle r

OP.OQ = p 2 -a 2 Deduce the


through O, OP.OQ = OR. OS.
.

+ b sin 0.
2pr cos (8 a) +p 2

a cos

circle r 2

a chord of the

0,

prove that

'product property' that, for chords

OPQ, ORS

ON be

the perpendicular from


to the tangent at A^dj) on the
2a cos 6, and P(r, 6) be any point on this tangent. By expressing ON
in two ways as r cos (6 26x ) and rx cos lf obtain equation (hi) of 20.22.
8 Let

circle r

*9 Prove that the normal to r

2a cos 6 at 6

= a has equation

rsin (2a 6)= asm 2a.


10 Find the equation of the chord joining the points of the circle
r

for

which

11

Two

2a cos {6 a)

Deduce the equation of the tangent at V


2
meet at O, and a line through
meets them again at P, Q
Find the locus of the mid-point of PQ.
6lt

circles

respectively.

12 Find the condition for the line IJr


= 2c cos 0.

= a cos 6 + b sin 6

*13 Prove the Simson line property (20.23, ex.

20.3

(i))

to touch the circle

by pure geometry.

Conies: pole at a focus

20.31 Polar equation of

Choose a focus

8 to the

S of the

all

non-degenerate conies

conic for pole,

corresponding directrix for

and the perpendicular from

initial line.

Let the latus rectum

LU have length 21, and let P(r, 6) be any point on the conic.

POLAR EQUATION OF A CONIC

20.31]

689

204 and 205, P lies on the same side of the directrix as S.


where P and S are on opposite sides, will arise when P is on
that branch of a hyperbola which is remote from S. We have

In

figs.

Fig. 206,

= SP = e.PM = e.ND

(e(SD-SN)
in 204

(e(SD-rcoad)

in 204, 205

e{SD + 8N)
in 205

in 204, 205,

ercoad l

e(rcoad-SD)
in 206

e(SN-SD)

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.

The last equation represents the branch of a hyperbola remote from S.

^
(J>r

~^jP
\
>

Fig. 207

If we replace r
fig.

by r- and 6 by 6 n

207), then the equation Ijr

=
I

we

use the notation of

+ e cos (6 n) = 1 e cos 6,
1

i.e.

(i.e. if

= 1 + ecos# becomes

-= 1 + e cos d
r

POLAR EQUATION OF A CONIC

690
as for

= l + ecos# will
remote from S if r is allowed

204 and 205. Hence the equation

figs.

represent the branch of a hyperbola

[20.32

Ijr

to take negative values.


It follows that, if negative values of r are allowed, then every proper
(i.e.

non-degenerate) conic

is completely represented

by

Ijr

+ e cos 6.

Remarks
(a) If

DS

the line

initial line, this is

discussion.

represent

the sense from

The equation becomes

all

proper conies

The equation

(/?)

(in

towards S)

by

equivalent to replacing

Ijr

if r is

Ijr

1 ecosd;

= 1 + e cos 6,

Indeed, the two equations

+ 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

but any chosen point of the

conic will of course correspond to different pairs of values of

the two representations. This fact


(y)

is Ijr

6 in

the equation

is Ijr

+ e cos 6;

20.32 Tracing of the curve

When

1,

is

= 1 + e cos (6 a).

wise through angle a gives the

(1)

r,

used in 20.33, ex. (hi).


The equation of a conic whose major axis is inclined at angle a

to the initial fine


line,

above

proper conies (with the same proviso about

all

r).

allowed to take negative values.

completely represents

71

wo its axis JSD as initial

new equation.

l/r

the conic

For,

rotation of the initial line clock-

= l+e cos 8

is

a parabola whose equation can be

written

-= l + cos0 =

2 cos 8 40.

Fig. 208

Fig. 209

It is represented completely

-n <

by

6 < n

and

>

0.

POLAR EQUATION OF A CONIC

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.

a hyperbola. The branch enclosing

is

7r + sec -1 e < 6n < n sec -1 e and

>

0;

and the branch remote from $ by

n sec -1 e <

< n + sec -1 e

and

<

0.

The reader should check the above statements.

Fig. 210

20.33 Examples
(i)

(a)

If

Eqitations of the directrices,

From the focus-directrix definition, SL = e.SD,

(r,

6) is

any point on the

rcostf

Hence this

so that

SD =

l/e.

directrix corresponding to 8, then

= SD =

Z/e.

directrix has equation


I

e cos 0.

*(b) For central conies


as follows.

we

obtain the equation of the directrix remote from

Jl-"-"""

6'

D'

Fig. 211

If e

<

1 (fig. 211),

Fig. 212

then

rcos(7T-0)

= SD' = DD'-SD
2a

/2a

POLAR EQUATION OF A CONIC

692
and since

l/a

= 1 e 2 (see Ex.

20(6), no.

1), this

2
\
II = Zl + e2
e\l-e
el-e
/

Hence the equation


H
If e

>

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'

1. The equation is thus the same for each case.

Equations of the asymptotes of a hyperbola.

*(ii)

The asymptote whose cartesian equation is y = bx/a is inclined to Ox at


angle A, where tan A = b/a and hence cos A = aQ(a 2 + b 2 = 1/e. Therefore the
perpendicular from S to this asymptote makes angle Zn sin -1 (1/e) with Ox;
)

its

length

is

Fig. 213

and so

its

equation

is

(20.13)

Since 6

l/*J(e

rcos

+ sin -1 - = r~
2 rr*
|

e)

Similarly j the asymptote y

1), this can be written


rcos

2n + sin -1

V(e

-1)

= 6sc/a has polar equation

-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

conies can be taken as


I

= 1+e cos

and

=
r

+ e'cos(0 a).

POLAR EQUATION OF A CONIC

20.34]

If we subtract these equations,

we obtain

e cos

=V

which

is

the equation of a

e'

cos (0 a),

Since

line.

it is satisfied

satisfy the equations of both conies, it represents a

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

both the equations

cos (6 a)

0,

and these represent the directrices corresponding to S, by ex.


chord therefore passes through the meet of these directrices.
The first conic is also given by
I

(i).

This

common

= 1 + e cos 6

r
(see

Remark

(fi)

in 20.31). Proceeding similarly, another

ecos#H
I

e'cos(0 a)

which also passes through the meet of the same

20.34

common chord is

0,

directrices.

Chord and tangent

Let P, Q be the points of the conic Ijr = 1+e cos 6 corresponding


to 6 = a, 6 = /?. Since by 20.14 the polar equation of any line can be
written
.4

cos 0+ 2? sin

^
=
r

let

the required chord have equation

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
p-e
r

sin/?

sm(a-yff)

and

q
*

The chord
^

cos fi cosa

^- =

sm(a-yfl)

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

+ sec (a /?) {cos

cos \{a +fi)

+ sin

sin J(a

4- /?)},

POLAR EQUATION OF A CONIC

694
i.e.

e cos 6

+ sec i(a - /?) cos [d -

For an alternative method, see Ex. 20 (b), no.

[20.34

7.

By letting J3 -> a we find that the tangent at the point 6 = a is


-

For a calculus method,

e cos

# + cos (d a).

Ex. 20 (b), no. 24.

see

Examples
(i)

The tangents

T(r, 6), where 6

at the points

P,

corresponding

to

a,fi meet at the point

%(ot+j3).

For the coordinates of T


e cos

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)

by an angular coordinate which lies between + n.

With the notation

If P,

nir

are not

on

and

Hence ST

of ex.

(i),

ST bisects PSQ.

different branches of a hyperbola

PST = Ha+fi)-a =

Hfi -a)

TSQ = fi-^a+p) =

(/?-a).

(fig.

214),

bisects

PSQ

(internally).

If P, Q are on different branches of a hyperbola, let Q be on the branch remote


from S. Then /? is the angle xSQ', and -|(a + /?) is angle xST', as shown in fig. 215.

PST' = i(a+yff)-a =
and

T'SQ'

= fi-\{a+P) =

(/?-a)

^-

a),

so that

ST bisects PSQ externally.

(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.

POLAR EQUATION OF A CONIC

20.34]

Tangents at

A and B meet where 6 = \{<x +


-

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))

These three points satisfy the equation


^

a + ft 4. y\
2

which by 20.21

(6)

(R, \(ct + ft + y )),

represents a circle of radius

'

R = \l sec \a sec /? sec y, centre

and passing through the pole

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

+ sec/? cos(0 a),

ecos/?.cos0 + cos(# a).

POLAR EQUATION OF A CONIC

696
This

is

the equation of the tangent at


I cos B

a to the

[20.34

conic

= 1 + ecos/fn cos0,

(l)

whose eccentricity is e cos /? and latus rectum

The tangents at the


-

21 cos {$.
extremities of the chord are

ecos0+*cos(0

and these meet where (by


on the conic

ex.

= a and l/r =

(i))

= ecos0 + cos(0 a B),


e cos a

+ cos/?.

This point

lies

+ cos B,

e cos

IseoB

i.e.

an

+ e sec B cos 0,

(u)

which has eccentricity esec B and latus rectum 2Zsec /?.


The conies (i), (ii) have the same focus and directrix as the given one.

Exercise 20(6)
1

If a, 6 are the semi-axes of l/r

Z/(l-e

= 1 + e cos 6 (e
and

2
)

Give the corresponding results when

2
3

>

4= 1)

Z/V(l-e

prove that when

<

1,

2
).

1.

What locus is represented by a = r sin 2 0? Sketch this curve.


What is represented by kjr = a + 6 cos + c sin 0?

4 If P(r1

a) is

focal chord of l/r = 1 + e cos 0 show


Q has coordinates of the form (r2 n+ a). Prove that

an extremity of a

the other extremity

that

J_ + J__2
SP SQ ~ V
i.e.

that

/&aZ/ i7ie Zaiws

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.

6 Prove that mutually perpendicular focal chords of a rectangular hyperbola


are equal in length.
7 Obtain the equation of the chord joining the points of
which correspond to 6 = a,fi from the result in 20.12.

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

2r{V3 + cos (0 + tt)},

and show that they touch there.


10 Prove that the tangents at the extremities of a focal chord meet on the
corresponding directrix.
11

If the tangent at

P meets the directrix at Z, prove that PSZ is a right-angle.

POLAR EQUATION OF A CONIC

697

12 A chord oiljr = 1 + e cos subtends an angle 2a at the focus. If the bisector


lies on the conic
of this angle meets the chord at M, prove that

Zcosa

a
= 1 + ecosa.cosc/.
,

13 P is any point on a conic with focus S. A line through S making a given


in T. Prove that T lies on a conic which
angle with SP meets the tangent at
has the same focus 8 and corresponding directrix as the given conic.

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

intersect at the ends L, L' of

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

\cos (a - fi) + e cos (a


17.

P,

If P,

Q meet

'

H*+P)).

fi)

Q are points on a parabola with focus S,


at T, prove that SP.SQ = ST*.

and the tangents at

= a, fi of the parabola ljr=l+ cos 0. Prove that


18 (i) P, Q are the points
the tangents at P, Q contain an angle a /?|
(ii) Tangents to a parabola with latus rectum 21 contain the constant
angle <f>. Prove that they meet at a point on a hyperbola having the same
focus as the parabola, and give the latus rectum and eccentricity. [Use no. 16.]
|

* 19 Prove that the chord of contact of tangents from (rx , 6 t ) to


has equation
v

^
/
ecos#j = cos^ ^).
ecosfljj
.

[If its extremities are

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?

*22 Tangents are drawn to the parabola l/r = 1 + cos d at 6 = a, fi, y, 8.


Prove that the circumcentres of the four triangles so formed lie on the circle
6

which passes through the focus. [Use 20.34, ex.

(iii).]

POLAR EQUATION OF A CONIC

698
Show

*23
l/r

that the equation of any line perpendicular to the tangent at

a to

= 1+e cos 6 has an equation of the form


kl

e cos (6

+ in) + cos (6 + \n a).

[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

esin0 + sm(0 a).

*24 Prove that the tangent at A(p, a) to the curve r =f(6) has equation
fdr\

-cos(0-a)- (
1

sin(0-a)

[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.

*25 Obtain similarly the equation of the normal at


to obtain the result of no. 23.

(p,

a) to r

= /(#), and use

it

Miscellaneous Exercise 20(c)


1 A, B are the points (r^dj),
Prove that

iy

+ r\2 + 2r x r 2 cos

and M(r,6)

and

2 }

tan 6

A variable line meets n given lines through

2
n

If

(r 2 ,6 2 ),

jOPr )

is

constant, prove that

it

at

is

the mid-point of

?\

sin

+ To sin

AB.

ricos^ + racos^a

P P2
lf

n respectively.

passes through a fixed point.

r=l

Find the equation of the common chord and of the

circles r

2a cos (6 a), r

4 Obtain the polar equation of the


(a,

a)

and

line of centres of the

2b cos(d-fl).
circle

whose diameter

is

the join of

(&,/?).

5 Prove that the chord of contact from


cos 6

cos Qr

rx

(rlt dj)

cos (6

to the circle r

2a cos d

is

X)

[Use the cartesian equation of the circle and chord.]


6 If PSP' and QSQ' are mutually perpendicular focal chords of a conic,
prove that 1/PS .SP' + 1/QS .SQ' is constant, and express this constant in
terms of I and e.

(i)

Prove that the locus of the mid -points of chords through the focus

= 1 + ecos0 is r(e 2 cos 2 0 1) = eZcos#.


(ii) By changing this equation to cartesian

of l/r

locus

is

coordinates,

a conic with the same eccentricity as the original one.

show that the

POLAR EQUATION OF A CONIC


8

(i)

699

If the tangent to the parabola a = r cos \d at the point


T, prove ST = a sec 2 \a. SP.
2

given by

= a meets the initial line at


(ii)

Find the equation of the

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

10 Prove that the line


if and only if (a - e) 2 + 6 2

to

1.

l/r

= a cos 6 + b sin 6 touches the

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

focus S to the directrix meets the conic at A.


the perpendicular from S to any tangent lies on the
tangent at A, then the conic is a parabola. [Do not prove the converse result.]
13

The perpendicular from the

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
=

and deduce that

lies

2SP.SQ

spTsQ

coaiPSQ

'

on a parabola whose vertex is S.

*15 PQ is a focal chord of a conic, and its mid-point is M. The normals at


is parallel to the (major) axis. [Use no. 7 (i)
P, Q meet at N. Prove that
and N.]
and Ex. 20 (6), no. 23 to prove that r sin 6 is the same for

MN

47

GPM

II

700

21

COORDINATE GEOMETRY IN SPACE:


THE PLANE AND LINE
21.1

Coordinates in space

21.11 Rectangular cartesian coordinates

The method of locating points in a plane by coordinates x, y referred


two intersecting lines Ox, Oy (15.11) can be extended to points in
space. Clearly three axes of reference will be required, and for simplicity we shall choose them to be mutually perpendicular.
Take any two intersecting perpendicular lines; let them meet at 0.
to

Fig. 216

On

Fig. 217

each select a positive sense, indicated by an arrow, and label

draw a line perpendicular to both


Ox and Oy (and therefore to the plane xOy), and choose the positive

these directions Ox, Oy. Through

sense along

it

to be the direction of advance of a right-handed screw

turned in the sense from Ox towards Oy, label this direction Oz.

Then we have constructed a right-handed system of rectangular axes


Ox, Oy, Oz.f The planes xOy, yOz, zOx determined by these axes are
also mutually perpendicular, and are called the coordinate planes.
Given a point P in space, we can draw the perpendiculars PL, PM,
PN to the planes yOz, zOx, xOy. The rectangular box having OP for
diagonal and PL, PM, PN for edges can now be completed; let its
t If the direction of any one axis is reversed,
only right-handed axes in this book.

We use

we should obtain a left-handed system.

THE PLANE AND LINE

21.12]

701

be OA, OB, 00 along Ox, Oy, Oz respectively. The


OA, OB, OC can be specified in magnitude and sense by
signed numbers x, y, z, and the ordered triplet of numbers (x, y, z) are
edges through
distances

the rectangular cartesian coordinates of P


particular, observe that

and that

wo

the axes Ox, Oy, Oz. In

L has coordinates (0, y, z), M(x, 0, z), N(x, y, 0)

A is (x, 0, 0), J3(0, y, 0), C(0, 0, z).

x, y, z, the points A, B, O
on the axes can be determined, and the box then completed. The

Conversely, given three signed numbers

corresponding point

P is the extremity of the diagonal through O.


P in space there corresponds a unique set

It follows that to each point

of coordinates x, y, z

numbers

x, y, z

wo

the axes Ox, Oy, Oz;

corresponds a unique point

The three coordinate planes

and

to

each ordered set of

P in space.

divide space into eight regions called

each corresponds a distribution of + and signs in the


coordinates (see Ex. 21 (a), no. 1), the number of such possible signdistributions being 2 s = 8. In fig. 217 all coordinates of P are positive.
octants; to

A in fig. 217 is the point in which Ox is met


P drawn parallel to the plane yOz. Hence PA
is perpendicular to Ox, so that A is the orthogonal projection of P
on Ox. Similarly, B and C are the projections of P on Oy, Oz respectively. The coordinates of a point P are therefore the (signed) lengths
Finally, observe that

by the plane through

of the orthogonal projections of OP on the axes.

Notation. In the remainder of this

coordinates are (xv yv

Zj),

and so

book

x is

the point whose

on.

21.12 Other coordinate systems


(1) Cylindrical

the plane

xOy by

polar coordinates. We may replace the coordinates (x,y) in


and
polar coordinates (/?, <j>), where p =
is the angle

ON

X
Fig. 218

Fig. 219
47-2

THE PLANE AND LINE

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),

Spherical polar coordinates. Further, in the plane

(2)

zONP the coordinates

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

21.21 Distance formula

Given two points Px and P2


plane

PN PN

xOy be

let

the perpendiculars from them to the

The coordinates of
2
2
and of N2 are (x2 ,y2 ,0).

X,

are (#1,^,0),

Regarded as points of the plane xOy,

wo

xi> Vi)>

(*2> y*)>

so

N N\ = {x
X

Through
since
it

P2 B

lies

meets PX

X,

axes Ox,

are

-x2 Y + {yx -y2 )\


N,

NN
in the plane of P N and P N
draw

PR

parallel to

say at R.

X;

2,

Fig. 221

From the right-angled

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

In particular, the distance of P(x, y, z) from

2
)

}.

O is J(x + y 2 + z 2 )
2

THE PLANE AND LINE

21.22]

703

21.22 Section formulae

Given two points Pv P2 let P{x, y, z) be the point dividing PX P2 in


the ratio k:l; then PX P\PP2 = k\l.
Draw PX X P2 2 PN perpendicular to plane xOy. Through P
draw PR parallel to X 2 to meet P2 2 at R; and through i\ draw
PX Q parallel to 2 to meet PN at Q.
,

NN

NN

Fig. 223

Fig. 222

From thesimilar trianglesPj PQ,PP2 R we have PQ\P2 R = P^/PPj.


(i)

If the division

is

internal
z

(fig.

z,

z 2 -z

from which

222), it follows that

I*

^ ++ k^

Za
.

Similarly,

we

by dropping perpendiculars from Plf P2 P to plane yOz,


x = (Z#1 + &a;2 )/(Z + &); and by perpendiculars to plane
,

find that

zOx, that y

{lyx

+ ky2 )f{l + k).

Alternatively, since

N divides ^iVg internally in the ratio k

and

has the same x- and ^-coordinates as P, the last two results follow

from the section formulae in the plane xOy


(ii)

If the division

is

external
z
z

from which

(fig.

223)

(15.14).

we have

zx _ k
z% V
5

= \
l

2
.

There are similar expressions for the x- and ^-coordinates of P.

THE PLANE AND LINE

704

Hence, with the convention in 15.14 (3),


the {signed) ratio

[21.22

the point dividing

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

xOy if and only if its z-coordinate is zero, i.e. 2k =

P has coordinates

0),

and divides AB
and only if its

31.

Hence

internally in the ratio 3 : 2.


cc-coordinate is zero, i.e. I

The point lies in yOz if


= 3k.
Hence Q is (0, 3, 5J), and divides AB externally in the ratio 1 3.
The point lies in plane zOx if its ^-coordinate is zero, i.e. 1 = 0. Thus R is
(3, 0, 1) and coincides with B. This is evident since B already lies in zOx.
:

(ii)

The four

a tetrahedron to the centroids of


a point one-quarter the way up each line from

lines joining the vertices of

opposite faces are concurrent at

the

the

corresponding face.
If the vertices are
etc.

The mid-point

P P P3 P
lt

2,

if

be the centroid of triangle Pj^P^P^

and

(? 4 is the point dividing


0. has coordinates ,

'

+ xs y1 + y 2 + yz
,

/'

'

MP, in the ratio

[xi+xz
3
\

Consider the point

Zi+g 2\

3/1+3/2

z1

yx

Ex. 15 (a), no.

z1

3; its

coordinates are

+z 2 + zz + zi\
)*

'

Hence

)'

which are symmetrical in those of P x P 2 P8 P4 Therefore


dividing each of the lines G^Pi, G 2 P 2
3 P Z in the ratio 1
,

1).

+g 2 + g3 \

'

+ Vi + Vz + Vi

(cf.

O dividing #4 P4 in the ratio

(x1 + x2 + xa + xi
\

C4

let

M of P, P, has coordinates

G is also the point

3,

and the

results

follow.

The point G of concurrence is called the centroid of the tetrahedron PjPaPjPj,


and the lines P-^Gx,
P4 #4 are called its medians.

Exercise 21 (a)
If the negative te-axis 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'.

THE PLANE AND LINE


(i)

(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.

1) is the centre of the sphere through the points (4, 6, 1 1),


- 1), - 3, - 10, 2).
- 1, 2), B(5, 3, - 6), C( - 1, 1, 7), D(9, 1, - 11), prove that AB

Show that ( 1,

5
(13,

- 1,

6 For A{Z,

and

2,

3), (6, 14,

CD bisect each other.

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),

and also on the line joining

(7,

5,

1), (4, 16,

2).

10 Prove that the lines joining the mid-points 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

By using triangle ONP in fig.

217, prove
Obtain the formula for the distance X

(i)

PP

(ii)

to

OP8 = x*+y2 + za
2

by first changing the

origin

Pv

The notation of 21.12

13

Express

x,

is

used in the following examples.

y in terms of p and

<j>,

and conversely.

*14 If k and a are constants, what points have


and = a?

(i)

k;

(ii)

<j>

a;

(iii)

a?

=k

<j>

*15 Express
* 16

If k,

cc,

(i)

(iv)

21.3

=
=

x, y, z in

terms of r, 6 and 0, and conversely.

are constants,
k;
cc

and

= /?;

what points have


(ii)

(v)

=
=

(iii)

cc;

k and

= /?;

(vi)

=#
k and

Direction cosines and direction ratios of a line

21.31 Direction cosines

We now consider how to specify the direction of a line, i.e. we seek


the space analogue of 'gradient'. Whereas the latter

the tangent of an angle,

we

is

shall find that in space the cosine is the

important trigonometrical function.


Let PQ be a line in space, described in the sense from

Through

draw a line

denned to be

P towards Q.
OA of unit length parallel to and in the same

THE PLANE AND LINE

706
sense as

PQ

(sensed) line

(fig.

224).

We

OA

call

[21.32

the unit ray corresponding to the

PQ.

The direction of OA, and therefore that of PQ,


by the coordinates of A. Since these coordinates
projections of

OA

will

be determined

are the orthogonal

on the axes Ox, Oy, Oz, hence A is the point


where a = xOA, /? = yOA and y = zOA. It is

(cos a, cos/?, cosy),

customary to write
I

m = cos

cosa,

n=

/?,

cosy,

m, n} the direction cosines of PQ (and therefore of any line


and in the same sense as PQ).
Since OA has unit length, the distance formula shows that

and

call

{I,

parallel to

+ m 2 + rc 2 =l.

(i)

Fig. 224

Fig. 225

Remarks
(a) If the line is described in

the sense from

corresponding unit ray through


(cos

a),

line parallel to the plane

Similarly, a line parallel to

has n

cos(7T /?), cos

( cosa, cos /?, cosy),

i.e.
(/?)

(77

towards P, then the

OA', where A'

is

zOx has

i.e.

is

the point

y)),
m, n).

(it

I,

yOz has a = \tt, so that 1 = 0.


m = 0, and one parallel to xOy

0.

(y) In particular, the coordinate axes described in the positive


senses Ox, Oy, Oz have direction cosines {1. 0, 0}, {0, 1, 0}, {0, 0, 1}.

21.32 Angle between two lines


If BS

is

another

line,

its direction cosines

through 0, then B

be

is

B towards S, let
OB is the corresponding unit ray

described in the sense from

{1',

m', n'}. If

the point

(V, m',n').

THE PLANE AND LINE

21.33]

707

The angle between the {sensed) lines PQ, BS (intersecting or not) is


denned to be the angle AOB for which < AOB ^ it (fig. 225).
Let AOB = 6; then from triangle AOB, in which OA = 1 = OB,
we have by the cosine rule that

AB = 2-2cos0.
2

Using the distance formula for AB, we also have

AB2 = (l-

+ (m- m'f + {n- n') 2


= Z 2 + m2 + n2 + V 2 + m' 2 + n' 2 - 21V - 2mm' - 2nn'
I')

= 2 - 2{IV + mm' + nn'),


since by (i), Z2 +

+ n2 = 1 and similarly I' 2 + m' 2 + n' 2 =

AB

the two expressions for

2
,

1.

Comparing

we find

= U'+ mm' + nn'.

cos

(ii)

In particular, the directions {l,m,n}, {l',m f ,n'} are perpendicular


if and only if IV + mm' + nn' = 0. They are parallel (in the same sense)
if
I

if I = V, m = m' n =
= m', n = n'.

and only

= V,

and 'antiparaUel'

n'\

if

and only

if

21.33 Direction ratios


It

is

often convenient to specify the direction of a line

by a

three numbers which are proportional to the direction cosines

set of

m, n.
of the line. For
I,

Any such three numbers are called direction ratios


example, the line in the plane xOy which bisects the angle xOy has
a =

\n,

f$

\-n,

y=

\n, so that its direction cosines are {1/^2, 1/^2, 0};

a convenient set of direction ratios for this bisector would be 1


In general, suppose I m n = p q r;f then I = Xp, m = Xq,n
where A is determined from equation (i) by
:

\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 )}.
),

important to observe whether the numbers

is understood to mean that o = 0; if


f An equation a:b = a':b' in which a' =
= 0, the equation means that 6 = 0. An equation a:b:c = a':b':c' in which (for
example) c' = means that a:b = a':b' and c = 0, and so on. Cf. 11.43(2).

b'

'

708

THE PLANE AND LINE

[21.33

meant to be

direction cosines or

p, q, r specifying a direction are

direction ratios.

We remark

the two senses of a

that direction ratios do not distinguish

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)
:

{pp + qq' + rr ')


'

a
COS U

V(2>

+? +

r*)

V(P'

+ ?' 2 + r'*)

where the sign is chosen so that the expression is positive.


In particular, the lines will be perpendicular if and only if
pp' + qq' + rr'

(the 'perpendicularity condition').

They will be parallel


If

(ii)

cos a

The

(iii)

P is

x/OP,

or antiparallel
'

The

lines

OP

if

q r
:

= p'

P P on Ox, Oy, Oz are x


P P i is x x :y y :z z
x

OP, OQ have

direction ratios I:

q' : r'.

are {x/OP,y/OP,z/OP}; for

A convenient set of direction ratios for OP is x

projections of

ratios for the line

cc

l9

y 2 yi,

z%

z.

z x Thus
.

1.

mm,

V m'
:

n'.

Find

direction

OB which is perpendicular to both OP and OQ.

OR have direction ratios p


lp+mq+nr =

Let

and only

if

(x,y,z), the direction cosines of


etc.

set of direction ratios for


(iv)

'

'*

Solving these for p q r (see


:

p:qir

r.

Then by the perpendicularity

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)

C(5, - 3, 7), D( - 3, 5, 5), prove AB CD.


- 1), C( -4, - 3, 0),D(6, -2, 4), prove AB JL CD.
Prove that the points - 5, 2, 7), - 9, 3, 6), (3, 0, 9) are collinear.
Prove that (2, 3, 0), (1, 5, 2), (3, 7, 1), (4, 5, 1) are the vertices of a square.

Find the direction

For .4(3,

2, 5),

2 For A{2, -4,


3

B(

3),

- 1,

B(3,

6, 4),

||

2,
(

ratios of the lines

through

O which make equal angles

with the coordinate axes.

6 Calculate the angle between any two diagonals of a cube.


7 Find the angles between the diagonals of a rectangular box whose edges
are a,

b, c.

8 Calculate the angles of the triangle whose vertices are A(2,

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.

line through O with


has coordinates (pl,pm,pn)
direction ratios for PN. Hence prove that

drawn

to

m, n} at N. Prove that

perpendicular

write

down

is

meet the

THE PLANE AND LINE

21.4]

is the foot of the perpendicular


10 If
coordinates of M.

*1 1

If the directions p q 1 : rlt p a q2 r2


:

p3

709

N to OP in no.

from

9,

find the

q3 ra are parallel to the same plane,


:

prove that
'

Pi

?i

*i

p%

q%

r2

?s

rs

Pa

0.

[They are perpendicular to the normal to the plane.]


*12 Prove that the converse of no. 11

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*

+ m*+n* = l\+m\ + n\ = Zf+m + n =

and
i

a is

+m m +w
a

2 W3

l3 l1

+m m +n
3

m m

(ii) Verify that


l 2 l3 ), (m^
2
the direction cosines of Ox, Oy, Oz
that
,

3 ),

n1

(n ly

n 2 n3
,

are
Fig. 226

wo f", and hence

+ + 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 =

(x' t y', z')

= l2 x+m 2 y+n2 z,

projection of OP on Ox'; use no.

(ii)

z'

wo Sf",

0.

prove that

= l3 x+m3 y+nz z.

9.]

Also prove that

l1 x'

+ la y' + l3 z',

[Symmetry; or multiply x',

= ml x'+m 2 y' + m3 z',

y', z'

*15 Explain geometrically

by llt

why

l% , l3

in

x 2 +y 2 +z*

(i),

= nx x' + n2 y' + n 3 z'.

add, and use no. 13

= x' 2 +y' 2 +z' 2 and


,

(ii).]

verify this

algebraically.

21.4

The plane

In this section we obtain the equations of planes determined

by-

show that the general linear equation in


x, y, z is the equation of some plane in space. The reader should note
resemblances to some of the equations obtained in 15.2.
various conditions, and

THE PLANE AND LINE

710

[21.41

21.41 Equation of a plane in 'perpendicular form'

Let the perpendicular from


to the given plane have foot N. If
ON has length p and direction cosines {I, m, n}, then is the point

(pl,pm, pn).
P(x, y, z)

any point of the plane,


then PN0N. Hence the direction
given by xpl:ypm:zpn is perIf P(x, y,

z) is

by

pendicular to that given

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

coordinates of any point of the plane,

by the

consequently the equation of the plane.


in x, y,

z.

Ax+By + Cz + D = Q

21.42 General linear equation


(1)

is

We observe that it is linear

Let Pj and

be two distinct points which satisfy the above

equation; then

Axx + Byx + Cz1 + D =


Multiply the

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)

+ h), the second by kj(l + h), and add;

by

which shows that the point dividing

Since k

0.

means that

'
:

'

also satisfies

0.

all

(ii)

points of the line

P^

(ii).

definition a plane

line joining

equation

(ii)

is

a locus in space

any two points of


represents

it lies

(or surface)

such that the

on the locus; hence

entirely

some plane.

The normal to this plane has direction ratios A:B:C. For if


P1 and P2 are any two points in the plane, then by subtraction of
equations (i) we have
(2)

A (xx - x2 + B{yt - y2 + C(z1 - z 2 =


)

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]

At least one of the coefficients in (ii) is non-zero. On dividing by


it we obtain an equation containing three arbitrary coefficients. This
confirms the geometrical fact that a plane is determined by three
(3)

independent conditions.

21.43 Conditions for Ax + By + Cz +


to represent the

The planes

D = 0, A'x +B'y + C'z + D' = Q

same plane
and only

are parallel if

A'

B'

if their

C'

A=B = C =k

normals are

parallel,

i.e.

(say) '

The equation of the second plane can therefore be written


k{Ax + By + Cz) + D'

0.

These parallel planes will be the same plane if and only if they have
z ); and then

a point in common, say (x y

Ax + By + Cz + D =

0,

Jc{-D) + D'

so that

k{Ax
0,

Hence the planes are identical

if

i.e.

+ By + Cz + D' =
)

0,

D'/D.

and only if
'

_ IT _ (7 _ >
A ~ B ~ C ~ D'

^1/

i.e. if

corresponding coefficients are proportional; in general the latter

will not

be equal.

21.44 Plane having normal

I:

mm and passing through P,

By 21.41 the plane with normal in direction l:m:n has an equation


of the Jform

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

- Xj) + m{y -yj + niz-

0.

(iii)

THE PLANE AND LINE

712

This relation holds for any point


therefore

(x, y, z)

of the required plane, and

the equation of the plane.

is

Alternatively, consider the equation

some plane

therefore represents

tion

[21.45

l:m:n

It is linear in x, y, z

(iii).

whose normal has


Hence it

(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

y-Vx

x x y%~ y\
xz ~ x i Vz~ Vl
x%

as the equation of the plane

P1 P2 P3

0.

homogeneous system of three


z

~ zi

z%

~ i\

ZZ

~ Zl

Let the required equation be

Ax + By + Cz+D =
This

equations gives

2.

P2 P3 must satisfy this, so

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,

Axz + Byz + Czz + D =

0,

Ax3 + Byz + Cz z + D =

0.

D from this homogeneous system of four equations

gives

xx

y1

zx

y%

^3

Vz

%3

as the required equation. It is easy to reduce this fourth-order determinant to


the third-order one just obtained, and vice versa.

THE PLANE AND LINE

21.46J

713

21.46 Intercept form: plane making intercepts a, b, c

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.

(0, b, 0), (0, 0, c) if

al + bm =
Hence

al

al + cn =

0,

= bm =

0.

cn,

l\m\n = -\\\-.

so that

b c

and the required equation is


x "y z
-+!+a o c

i.

21.47 Angle between two planes

The angle between two

intersecting planes (a dihedral angle) is

by

between the perpendiculars drawn in each to


clearly equal to that between their normals.

definition the acute angle

the

common line, and is

Given two planes

Ax+By + Cz + D =

A'x + B'y + C'z+D'

0,

their normals have directions

A B

the planes

by

is

therefore given

cose/

In particular,

C,

0,

A' B' C. The angle 6 between


:

(see 21.33, ex.

(i))

(AA' + BB' + CC)


2
2
V (A* + B + C ) y/(A

'

+ B'* + C" 2

the planes are perpendicular if and only if

AA' + BB' + CC =

0.

They are parallel if and only if their normals are parallel,

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

Find the equation of the plane through ( 2,4,


1) and (6, 3, 2).

parallel to the line joining (3, 5,

0)

and 3,2, 9) which


(

THE PLANE AND LINE

714

4 Obtain the equation of the plane through

[21.5

- 1),

(2, 2,

(7, 0, 6), (3, 4, 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,

Y on a horizontal plane, where OX = c = OY and

sunk to meet a seam of coal at depths p, q, r


Assuming that the coal face is plane, find the angle between this
plane and the horizontal, and the distances from O of the points where this
plane meets OX and OY.
8 If A(a, 0,0), B(0,6,0), (7(0, 0,c), P(a,b,c), prove that the angle between
OP and plane ABC is sin" 1 {Bq/p), where p 2 = a 2 + 6 2 + c 2 andg~ 2 = a~ 2 + &~ 2 + c~ 2
Also prove that the distance between the feet of the perpendiculars from and
P to ABO is <J{p* - 9g 2 ), and find the angle between planes PBO, PGA.
9 Find the equation of the plane bisecting PX P Z a* right-angles.
*10 Show that the two detenriinantal equations in 21.45 are equivalent.
1 1 Prove that the plane through P 1 and P 2 which is parallel to the direction
\tt,

vertical shafts are

respectively.

q r has equation
:

z1

X-i

12 Prove that the plane through


q r, p' q' r' has equation
:

y-yx
Xo

which is

0.

parallel to

each of the directions

x-xx y-y x z-z x


p

p'

q'

r'

0.

13 Sides of a plane. Find the value of k I if the point dividing P-lP2 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)

space can be determined

the join of two points

(c)

by

(a)

the meet of two

a single point together with a set

of direction ratios.

We begin most conveniently with case (c).


21.51 Line through

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

for the line is

direction ratios, hence

x x1 y yx
:

z1 =

n.

THE PLANE AND LINE

21.52]

715

This can be expressed in the formf

x-x1 ^ y-y1 = z-z1

These equations are satisfied by any point of the line, and are therefore
the equations of the

line.

Notice that two equations are necessary to determine a line in space:


a single linear equation represents a plane. The two equations (i) are
in fact the equations of two planes, each of which contains the line;
G

'

x-xx = z-zx

x-xx = y-yx

"

represent

two such

We refer to

(i)

'

planes.

as the symmetrical equations of a line.

Example
Equation of the line

PP
X

(case (6) above).

Direction ratios of the line


has equations

PX P% are

xx x % yx 1/ 2 zx z 2 Hence the

y yx

x-xx

2/1

line

z-zx

2/2

21.52 Parametric equations of a line

Putting each of the ratios in

x = xx + M,

(i)

above equal to

= y1 + \m,

zx

A,

we have

+ Xn.

(ii)

These are the parametric equations of the line through Px in direction


l:m:n, A being the parameter (cf. 15.26$). They are particularly
useful when dealing with intersections of the line with a line, plane,
or other surface.
t If the line

is

parallel to

a coordinate plane, say to yOz, then w

= 0. The equation

xx1 :yy1 :zz 1 = l:m:0

(i)'
x = Urn

z t = 0.

(see the footnote on p. 707) that x x^.y


consistent with the geometrical situation, since every point of the line has its
z-coordinate constant, say z = z x : the line lies in the plane whose equation is z = z x .

then means

and

(i)',

with a similar

This

is

We

continue to write

= - =

as alternative to

understanding when two of I, m, n are zero.


% Unless I, m, n are direction cosines, A is not the distance of the point
(x x +\l, y x +Xm, z 1 +Aro) from P x but it is proportional to this distance.
;

48

gpmii

THE PLANE AND LINE

716

[21.52

Examples
Find

(i)

the length of the perpendicular from

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,

obtained by putting the ratios in

its

equations equal to A, are

= 3A+10,

= -4A-8,

= A-5.

Calling this general point P, the direction ratios of

AP are

3A + 7:-4A-15:A-3.

AP
3:-4:

will

be perpendicular to the given

1) if

and only

(whose direction ratios are

line

if

3(3A + 7)-4(-4A-15) + l(A-3) = 0,


= 3. Hence the foot N of the perpendicular from A to the line
by putting A = 3 in the coordinates of the general point P; this

from which A
is

obtained

gives N(l, 4,-8).


It now follows that

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

to the plane and is bisected by it.


Since the normal to the plane has direction ratios 2

the normal through


6

parametrically
or ^
J

P is

x-Z

_
~2~~

2A + 3,

3,

the equation of

_ z-2
~ -3'

y+5
7

= 7A-5,

oi.o
= - 3A
+ 2.

This point will be P', the image of P, provided that the mid-point of P'P
on the plane, i.e.
2(A + 3) + 7(|A-5)-3(-|A + 2) = 27,

lies

from which A
(iii)

Find

the

2.

Hence P' is the point

image of the

(7, 9,

- 4).

line

3
x-2
__ _ 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,

cuts the given plane at the point

(3A + 2)
i.e.

hence

is (5, 3, 2).

for

= 5A-3.

which

- 2(2A + 1) + 3(5A - 3) =

5,

THE PLANE AND LINE

21.53]

The method of ex.

(6)

P'(4,

-3,

(c)

shows that the image of

(ii)

717

P in the given plane is

3).

The line P'Q has direction ratios

- 1, so its equations can be written

g-2
x-5
___ _
__ _ y-S
_
(iv)

Condition for two linen

Suppose that the

coplanar.

to be

lines

xa

yb

y b'_z c'
m'
n'

x a'

'

are coplanar. Then either they are parallel, or they intersect.


n = V m' n'.
If they are parallel, then I
If the lines intersect, then the common point has coordinates
:

+ Al, b + Am, c + An)

for

some

A,

+ pi', b' + pm', c' + fin')

for

some

fi.

(a

and

also

(a'

Hence there

exis