You are on page 1of 618

5B

ma

boh

of

WE
or

UNIVERSITY
UFO

WORKS OF

PROF. MANSFIELD MERRIMAN


PUBLISHED BY

JOHN WILEY & SONS,


53 East Tenth
London:
Street,

New
Limited.

York.

CHAPMAN & HALL,

Treatise on Hydraulics.

and a Manual for Engineers.

Text-book for Engineering Colleges Octavo, cloth. Price $4.00.

A Text-book on
A Text=book on
The Strength

limns, and Shafts.

the Mechanics of Materials and of Beams, Col= Octavo, cloth. Enlarged edition. Price $4.00.
Octavo, cloth.

the Method of Least Squares.


Price $2.00.

En-

larged edition.

of Materials. An Elementary Text-book for Manual Training Schools. Duodecimo, cloth. Price $1.00.

Text=book on Roofs and Bridges.


Jacoby.
In Four Parts.
I.

By Professors Merriman and


Price $2.50.

Octavo, cloth.

Part
Part
Part

Stresses in Simple Trusses.

II.

Graphic Statics.
Bridge Design.

Price $2.50.
Price $5.00.

III.

Part

IV.

Higher Structures.

{In Press.)

Price $2.50.

A Handbook

for Surveyors. By Professors Merriman and Brooks. Pocket-book form, morocco cover. Price $2.00.

Higher Mathematics.
leges. cloth.

Edited by Professors
Price $5.00.

Text-book for Classical and Engineering ColMerriman and Woodward. Octavo,

HIGHER MATHEMATICS.
A TEXT-BOOK FOR

CLASSICAL AND ENGINEERING COLLEGES.

EDITED BY

MANSFIELD MERRIMAN,
Professor of Civil Engineering
in

Lehigh University,

AND

ROBERT

S.

WOODWARD,
in

Professor of Mechanics

Columbia University.

SECOND EDITION, REVISED.


FIRST THOUSAND.

NEW YORK:

JOHN WILEY
London:
1898.

& SONS.
Limited.

CHAPMAN & HALL,

Copyright,

1896,

BY

MANSFIELD MERRIMAN
AND

ROBERT

S.

WOODWARD.

ROBERT DRUMMOND, ELECTROTYPER AND PRINTER, NEW YORK.

Y
PREFACE.

In the early part of this century it was possible for an industrious student to acquire a comprehensive if not minute knowledge of the entire realm of mathematical science. The
of that time, like Lagrange, Laplace, and were about equally familiar with all branches of pure Gauss, and applied mathematics. Since that epoch the tendency has been constantly towards specialization and additions to pure
;

more eminent minds

theory along with extensions of applications have been made with increasing rapidity, until now the mere quantity of information available presents a formidable obstacle to the simul-

taneous attainment of the breadth and depth of knowledge which characterized the mathematician of a generation ago.

would appear, however, that this obstacle is due to the bewildering mass of details rather than to any considerable Hence the increase in the number of fundamental principles.
It

who seeks to gain a comprehensive view of the mathematics of the present day needs most of all that sort of guidance which fixes his attention on essentials and prevents him
student

from wasting valuable time and energy


essentials.

in the pursuit of

non-

During the past twenty years a marked change of opinion has occurred as to the aims and methods of mathematical The old ideas that mathematical studies should instruction.
be pursued to discipline the mind, and that such studies were ended when an elementary course in the calculus had been In our best covered, have for the most part disappeared.
classical

and engineering colleges the elementary course


ill

in

IV

PREFACE.
is

now given in the sophomore year, while lectures and work in pure mathematics are continued during the seminary It is with the hope of meeting the junior and senior years.
calculus

existing
class

demand

for a suitable text to

be used

in

such upper-

editors enlisted the cooperation of the authors in the task of bringing together the chapters of this

work that the

book.

and

Each chapter, so far as it goes, is complete in itself, intended primarily to give a clear idea of the leading principles of the subject treated. While the authors have been
is

guided by general instructions issued by the editors, each has been free to follow his own plan of treatment. It will be found
that certain chapters adopt the formal method usual in textbooks, while others employ what may be called the historical

and

method. glance at the table of contents will that the chapters of the work present a considerable variety of subjects, thus affording teachers and students an
intuitive

show

opportunity to select such topics as

maybe

suited to their time

and

are given for solution, numerical examples of the application of theory to physical science are freely introduced, and the footnotes set forth much sugtastes.

Numerous problems

gestive matter of a historical and critical nature.

The

Editors.

NOTE TO THE SECOND EDITION.


THIS edition differs from the first only in the correction of those typographic errors which have been detected and in the
alteration of a few paragraphs

which seemed obscure.

Readers

of this edition are requested to bring to the notice of the editors

any remaining
January
i,

errors

which may be discovered.

1898.

CONTENTS.

Chapter

I.

THE SOLUTION OF EQUATIONS.


By Mansfield Merriman,

Professor of Civil Engineering in Lehigh University.

Article

I.

2. 3.

Introduction Graphic Solutions

Page

3
5

The Regula

Falsi

4.
5.

Newton's Approximation Rule


Separation of the Roots

6
8

6.
7.
8.

Numerical Algebraic Equations Transcendental Equations


Algebraic Solutions The Cubic Equation

10
13
15

9.

17

10. 11. 12. 13. 14.

The Quartic Equation


Quintic Equations Trigonometric Solutions

19
21

24 27
28
31

Real Roots by Series Computation of All Roots

15.

Conclusion

Chapter

II.

DETERMINANTS.

By Laenas Gifford Weld,


Professor of Mathematics in the State University of Iowa.

Article

1.

Introduction

33

2. 3.

Permutations
Interchange of Two Elements Positive and Negative Permutations The Determinant Array 1 Determinant as Function of n Elements of Determinants Examples
Notations

34
35

4.
5.

36 36

6.
7. 8.

37
38

39

9.

10.
11. 12.

Second and Third Orders Interchange of Rows and Columns Interchange of Two Parallel Lines

40
42
43 43

Two

Identical Parallel Lines

VI

CONTENTS.
by a Factor Line of Polynomial Elements Composition of Parallel Lines Binomial Factors Page
44

Article 13. Multiplying


14.

44
45

15. 16. 17.

Co-factors; Minors

18.
19.

20. 21.

Development in Terms of Co-factors The Zero Formulas Cauchy's Method of Development


Differentiation of Determinants

46 47 49
51

52 54
55

22. 23. 24. 25.


26.

27.
28. 29. 30. 31.

Raising the Order Solution of Linear Equations Consistence of Linear Systems The Matrix Homogeneous Linear Systems Co- factors in a Zero Determinant Sylvester's Method of Elimination

56
58

60 60
62
63

The

Multiplication

Theorem

Product of

Rectangular Arrays Reciprocal Determinants

Two

65 68

69

Chapter

III.

PROJECTIVE GEOMETRY.

By George Bruce Halsted,


Professor of Mathematics in the University of Texas.

Article

1.

The Elements and Primal Forms


Projecting and Cutting

70
72

2. 3.

Elements

at Infinity

72
74
:

4.
5.

Correlation and Duality

Polystims and Polygrams

74
77 80

6.
7.

Harmonic Elements
Projectivity

8.

9.

Curves of the Second Degree Pole and Polar


Involution
Projective Conic Ranges

82

87
88
91

10.
11.

12. 13. 14.

15.

Center and Diameter Plane and Point Duality Ruled Quadric Surfaces Cross Ratio

94 96
98

104

Chapter

IV.

HYPERBOLIC FUNCTIONS.
By James McMahon,

Associate Professor of Mathematics in Cornell University.

Article

1.

2. 3.

4.

Correspondence of Points on Conies Areas of Corresponding Triangles Areas of Corresponding Sectors Characteristic Ratios of Sectorial Measures

107

109

109

no

CONTENTS.
Article
5.

Vll

6.
7.

Ratios expressed as Triangle-measures Functional Relations for Ellipse

Page

8. 9.

Functional Relations for Hyperbola Relations between Hyperbolic Functions


Variations of the Hyperbolic Functions Anti-hyperbolic Functions

no in in
112

10.
11. 12.
13.

114 116
116
118

Functions of Sums and Differences Conversion Formulas


Limiting Ratios Derivatives of Hyperbolic Functions
Derivatives of Anti-hyperbolic Functions Expansion of Hyperbolic Functions

14.
15.

119 120
122
123

16.
17. 18. 19.

Exponential expressions.

...

124
125

Expansion of Anti-functions Logarithmic Expression of Anti-functions

20.
21.
22.

The Gudermanian Function


Circular Functions of Gudermanian Gudermanian Angle Derivatives of Gudermanian and Inverse Series for Gudermanian and its Inverse

127 128
128

129 130
131

23. 24. 25.

26.
27. 28.

29.

30.
31.
32.

Graphs of Hyperbolic Functions Elementary Integrals Functions of Complex Numbers Addition Theorems for Complexes Functions of Pure Imaginaries Functions of x -f *> in the Form X-\- iY

132
135

138

140
141

143
*45

33.
34. 35. 36.
37.

The Catenary The Catenary of Uniform Strength The Elastic Catenary The Tractory The Loxodrome Combined Flexure and Tension
Alternating Currents Miscellaneous Applications Explanation of Tables

14?
*

148

*49
x

50

151

*53
158 l6

38. 39.

Chapter

V.

HARMONIC FUNCTIONS.
E. Byerly,

By William

Professor of Mathematics in Harvard University.

Article

1.

2.
3.

History and Description Homogeneous Linear Differential Equations

l6 9

172

4.
5.

Problem Problem Problem

in

Trigonometric Series in Zonal Harmonics in Bessel's Functions

174
x

77

l8 3

6.
7.

The Sine Series The Cosine Series


Fourier's Series

l88
x92 T 94

8.

Vlll

CONTENTS.
9.

Article

Extension of Fourier's Series


Dirichlet's Conditions

Page 196
198

10.

11. 12.

Applications of Trigonometric Series Properties of Zonal Harmonics

20a
202 205

13.
14.
15.

Problems

in

Zonal Harmonics

Additional Forms

207
208

16.
17. 18. 19.

Development in terms of Zonal Harmonics Formulas for Development Formulas in Zonal Harmonics Spherical Harmonics.
Bessel's Functions.

209 212
213 213
215

Properties 20. Applications of Bessel's Functions


21.

22. 23.
24.
25.

Development in Terms of Bessel's Functions Problems in Bessel's Functions. Bessel's Functions of Higher Order Lame's Functions
Tables

217

220
221

221

222

Chapter VI.

FUNCTIONS OF A COMPLEX VARIABLE.


By Thomas
S.

Fiske,

Professor of Mathematics in Columbia University.

Article

1.

Definition of Function

226 227 228

2.

3.

4.
5.

6.
7.

Representation of Complex Variable Absolute Convergence Elementary Functions Continuity of Functions Graphical Representation of Functions Derivatives

229 230 232


233

8.

9.

10.
11.

12.

Conformal Representation Examples of Conformal Representation Conformal Representation of a Sphere Conjugate Functions Application to Fluid Motion
Point at Infinity Integral of a Function

236
238

244
245
.

246

13. Critical Points 14.


15.

250
256 257
261

16. 17. 18.

19.

Reduction of Complex Integrals to Real Cauchy's Theorem Application of Cauchy's Theorem Theorems on Curvilinear Integrals
Taylor's Series Laurent's Series Fourier's Series

262
264
267

20. 21.
22. 23.

269
271

273

24. 25.

Uniform Convergence One-valued Functions with


Residues

274
Critical Points

278 282

26.

Integral of a One-valued Function

284

CONTENTS.
Article 27. Weierstrass's
28.

IX

Theorem

Page 287
292 298

Mittag-Leffler's Theorem 29. Critical Lines and Regions

30.

Functions having n Values

300

Chapter

VII.

DIFFERENTIAL EQUATIONS.

By W. Woolsey Johnson,
Professor of Mathematics in United States Naval Academy.

Article

I.

2.
3.

Equations of First Order and Degree Geometrical Representation


Primitive of a Differential Equation

303 305

307
308
311

4.
5.

6.
7. 8.

9.

Exact Differential Equations Homogeneous Equation The Linear Equation First Order and Second Degree Singular Solutions Singular Solution from the Complete Integral
Solution by Differentiation Geometric Applications; Trajectories

312

314
317

320
322
325

10.

11.
12. 13. 14. 15.

Simultaneous Differential Equations Equations of the Second Order

327

330
333

The Two

First

Integrals

16.
17.

Linear Equations Linear Equations with Constant Coefficients Homogeneous Linear Equations
Solutions in Infinite Series

336
338

342

18. 19.

344

20.

21. 22. 23. 24. 25.

Systems of Differential Equations 349 First Order and Degree with Three Variables 352 Partial Differential Equations of First Order and Degree... 355 359 Complete and General Integrals 362 Complete Integral for Special Forms Partial Equations of Second Order 365 Linear Partial Differential Equations 368

Chapter

VIII.

GRASSMANN'S SPACE ANALYSIS.


By Edward W. Hyde,
in

Professor of Mathematics

the University of Cincinnati.

Article

1.

2. 3.

Explanations and Definitions Sum and Difference of Two Points

374
375

4.
5-

6.
7. 8.

any Number of Points Reference Systems Nature of Geometric Multiplication


of

Sum Sum

of

Two Weighted

Points

378
381

386

390
392

Planimetric Products

The Complement

399

X
Article
9.

CONTENTS.
Equations of Condition and Formulas Stereometric Products

Page 405
410

10.

11. 12.

The Complement

Space Addition of Sects in Solid Space

in Solid

416

419

Chapter IX.

VECTOR ANALYSIS

and QUATERNIONS.
Lehigh University.

By Alexander Macfarlane,
Lecf.urer in Mathematical Physics in

Article

1.

Introduction

425

2.

3.

4.
5.

6.
7.

8. 9.

Addition of Coplanar Vectors Products of Coplanar Vectors Coaxial Quaternions Addition of Vectors in Space Product of Two Vectors Product of Three Vectors Composition of Located Quantities
Spherical Trigonometry Composition of Rotations

426
432

439
443

444
449
453

457 463

10.

Chapter X.

PROBABILITY and THEORY OF ERRORS.


By Robert
S.

Woodward,

Professor of Mechanics in Columbia University.

Article

1.

Introduction

467
471

2.

3.

Permutations Combinations
Direct Probabilities

473

4.
5.

476 479 482 484


487

Probability of Concurrent Events


Bernoulli's

6.
7. 8.

Theorem

Inverse Probabilities
Probabilities of Future Events

9.

Theory

of Errors

490
491

10.

Laws

of Error

11.
12.

13.
14.

Typical Errors of a System Laws of Resultant Error Errors of Interpolated Values


Statistical Test of

493

Theory

494 497 504

Chapter XI.

HISTORY OF MODERN MATHEMATICS.


By David Eugene Smith,
Normal
College.

Professor of Mathematics in Michigan State

Article

1.

Introduction

508
511

2.

3.

4.
5.

Theory of Numbers Irrational and Transcendent Numbers Complex Numbers Quaternions and Ausdehnungslehre

513
t

515

517

CONTENTS.
Article
6.
7.

XI

Theory

of Equations

Page 519
524

Substitutions and Groups

8. 9.

Determinants
Quantics Calculus

526
52S
531

10. 11.

Differential Equations 12. Infinite Series


J3. 14. 15.

535

539 543

Theory

of Functions

16.
17.
18.

and Least Squares Analytic Geometry Modern Geometry Trigonometry and Elementary Geometry Non-Euclidean Geometry
Probabilities

550
552
558

563
565
o

19.

Bibliography

568
57 1

Index

HIGHER MATHEMATICS
Chapter
I.

THE SOLUTION OF EQUATIONS.


By Mansfield Merriman,
Professor of Civil Engineering
in

Lehigh University.

Art.

1.

Introduction.

In this Chapter will be presented a brief outline of methods, not commonly found in text-books, for the solution of an

equation containing one unknown quantity.

Graphic, numeric,

and algebraic solutions will be given by which the real roots of both algebraic and transcendental equations may be obtained, together with historical information and theoretic
discussions.

algebraic equation is one that involves only the operaIt is to be first freed from radicals so as tions of arithmetic.

An

make the exponents of the unknown quantity all integers; the degree of the equation is then indicated by the highest exponent of the unknown quantity. The algebraic solution of an
to
algebraic equation is the expression of its roots in terms of the literal coefficients this is possible, in general, only for linear, quadratic, cubic, and quartic equations, that is, for equations
;

of the

first,
is

second, third, and fourth degrees.


all its

numerical

equation

an algebraic equation having

coefficients real

numbers, either positive or negative.

For the four degrees

THE SOLUTION OF EQUATIONS.

[CHAP.

I.

above mentioned the roots of numerical equations may be computed from the formulas for the algebraic solutions, unless
they
fall

under the so-called irreducible case wherein

real

quantities are expressed in imaginary forms.


th algebraic equation of the n degree may be written with all its terms transposed to the first member, thus
:

An

+ax
x

n-x
-j-

a 1x

i n ~'

-f-

-f-

an -

+a
is,

o,

and, for brevity, the first member will be called f(x) and the The roots of this equao. equation be referred to as f(x)

tion are the values of

x which

satisfy

it,

that

those values of

that reduce f(x) to o. When all the coefficients alt ati . . .aM are real, as will always be supposed to be the case, Sturm's
real roots,

theorem gives the number of

provided they are un-

equal, as also the number of real roots lying between two assumed values of x, while Horner's method furnishes a con-

venient process for obtaining the values of the roots to any


required degree of precision. transcendental equation

is

one involving the operations


as, for

of trigonometry or of logarithms,

example, ^-(~

COS;tr

>

or a 2x
of

general method for the literal solution -\these equations exists but when all known quantities are

xb*

= o. No

expressed as real numbers, the real roots may be located and computed by tentative methods. Here also the equation may

be designated asf(x)

= o,

and the discussions

in Arts. 2-5 will

apply equally well to both algebraic and transcendental forms. The methods to be given are thus, in a sense, more valuable
than

Sturm's theorem and Horner's process, although for It should algebraic equations they may be somewhat longer.

be remembered, however, that algebraic equations higher than the fourth degree do not often occur in physical problems, and that the value of a method of solution is to be measured not

merely by the rapidity of computation, but also by the ease with which it can be kept in mind and applied.
Prob.
1.

Reduce the equation

(a

+ x)* +
unknown

x)l

=
all

2b to an
integers.

equation having the exponents of the

quantity

Art.

2.]

GRAPHIC SOLUTIONS.

Art.

2.

Graphic Solutions.
of the real roots of

algebraic equations

two simultaneous be found by the methods of may plane analytic geometry when the coefficients are numerically expressed. For example, let the given equations be
**

Approximate values

+y = a\

x*

- bx = y - cy,
The
c\

the

first

Drawing two rectangular axes


scribed from

representing a circle and the second a hyperbola. and OY, the circle is de-

OX
a.

with the radius

coordinates of the

center of the hyperbola are found to be

OA = \b

and

AC = \c,
the

while

its

diameter

BD = *J &

from which

two

branches

may

be described.

The

intersections of the circle

with the hyperbola give the real values of x and y. If

^=
and

I,

= 4,

and

3,

there

are but two real values for

x
y,

two

real

values

for

since the circle intersects but

one branch
here

of the hyperbola

Op

the positive and the negative value of x, while rnn


is

Om

is

the positive and

pq

the negative value of y. When the radius a is so large that the circle intersects both branches of the hyperbola there are four real values of both x and y.

By a similar method approximate values of the real roots of an algebraic equation containing but one unknown quantity may be graphically found. For instance, let the cubic equation 3 This may be x -\- ax b = o be required to be solved.*
written as the two simultaneous equations

y
pp. 47-49

x\

ax

-\-b,

*See Proceedings of the Engineers' Club of Philadelphia, 1884,

Vl

IV,

THE SOLUTION OF EQUATIONS.

[CHAP.

I.

and the graph

of each being plotted, the abscissas of

their

points of intersection give the real roots of the cubic.

The

curve
I

should be plotted upon cross section paper by the help of a

= x*

/
-J>4?

table of cubes

then

OB

is

laid off

equal to

b,

and

OC equal

to a/b, tak-

^___nl

ing care to observe the signs of a and


b.

The

line

joining

B
b

and

C
is

cuts

the curve at /, and hence qp


real root of x*
-f-

the
the
will

ax

= o.

If

cubic equation have three

real roots the straight line

BC

intersect the curve in three points.

algebraic equations of higher degrees may be graphicFor the quartic equation ally solved in a similar manner. Bz C o, it is best to put z A*x, and thus z* Az*

Some

reduce

it

to the form x"

+ +
#*

bx

o;

then the two

equations to be plotted are

y
the
first

= x* + x\
may be drawn

y=bx +
once for
all

c,

of

which

upon cross-section

paper, while the straight line represented by the second be drawn for each particular case, as described above.*

may

This method
tions
;

is

also applicable to

many

thus for the equation Ax sin^r sin^r write ax o; then y

transcendental equaBs\i\x o it is best to

of a table of sines, while

through the origin. curve represented by y

is readily plotted by help ax is a straight line passing In the same way ax x* = o gives the

= y=

y = x%

the parabola represented by the intersections of which determine the real roots of

= a* and

the given equation.


Prob. 2. Devise a graphic solution for finding approximate 3 o. \zQ. values of the real roots of the equation x*-{- ax -\- bx*-\- ex -f- d Prob. 3. Determine graphically the number and the approximate

values of

the

real

roots

of

the
,

(Ans.

Six real roots,

equation

arc

x
456

8 sin
.)

o.
[

159

430

and

a^

* For an extension of
see Phillips and Beebe's

this

method

to the determination of

imaginary roots,

iphic Algebra,

New

York, 1882.

Art

3-]

THE REGULA

FALSI.

Art.

3.

The Regula
for

Falsi.

One

of the oldest
is

methods

an equation
ple that
if

the rule

known

computing the real root of as " regula falsi," often called

It depends upon the princitwo numbers *, and x% be substituted in the expression/^), and if one of these renders/^) positive and the other renders it negative, then at least one real root of the equation Let the figure represent a f(x) = o lies between x and #s
x
.

the method of double position *

The point- X, part of the real graph of the equation y /{x). where the curve crosses the axis of abscissas, gives a real root of the equation f(x) o. Let OA and OB be inferior and

OX

trial

which are determined either by superior limits of the root or by the method of Art. 5.

OX

Let

Aa

and Bb be the values

of

J{x) corresponding to these limits. Join ad, then the intersection C of


the straight line ab with the axis OB gives an approximate value

OC
is

for the root.

Now

compute

Cc and

join ac, then the intersection


still
x

D gives a value OD which

closer

to the root

OX.

Let x and x^ be the assumed values OA and OB, and let x i) and/(;tr2 ) be the corresponding values oi f(x) represented

by Aa and Bb, these values being with contrary signs. Then from the similar triangle AaC and BbC the abscissa OC is
x%

~
A*t)-A*.)

By a second application x is computed, and by continuing


K

^A*>)-A**)~~ of the rule to x and x


x

*^
%
,

A**)-A*y
another value

the process the value of


2

can be obtained to any required degree of precision.

As

an example
trial

be found by
ham ben

o. x* 5* -f- 7 /(>) that a real root lies between

let

Here it may 2 and 1.8.


was by Abra-

*This originated

in India,

and

its first

publication in Europe

Esra, in 1130.

See Matthiesen, Grundzuge der antiken und moder-

en Algebra der litteralen Gleichungen, Leipzig, 1878.

h,

i~

7 iT

THE SOLUTION OF EQUATIONS.


jr,

[CHAP.

For

=
and

2,

f{x)

=
falsi

5,

and

for *,
is

1.8,

then by the regula

there

found x 3

/) = + 4-304;
1.90 nearly.

Again, for
with
jTj

1.90,

f(x
;r 4

z)

f(x^) give

= + 0.290,
1.906,

and these combined


is

which

correct to the

third decimal.
sin x 0.5 = o. example let f[x) = arc;r shows that there is but one real root,, Here a graphic solution and that the value of it lies between 85 and 86. For x, 85V and for x^ = 86, /(.r 2 ) = -f- 0.00342 then 0.01266, fix ) o 00090. Again, by the rule x = 85 44', which gives f(x ) = combining the values for x^ and ,r3 there is found x\ = 85 47',

As

a second

which gives /(^ ) = Lastly, combining the values 0.00009. for x^ and x t there is found x b = 8547 -4, which is as close an
4
/

approximation as can be made with five-place

tables.

In the application of this method it is to be observed that the signs of the values of x and f(x) are to be carefully regarded, and also that the values of f(x) to be combined in one

have opposite signs. For the quickest the values of f{x) to be selected should be those approximation having the smallest numerical values.
operation

should

Prob.

4.

Compute by the

Also those of x*

sin

2X

= o.

regula falsi the real roots of

xb

0.25

= 0.

Art.

4.

Newton's Approximation Rule.


for
is

Another useful method


the real root of an equation
If

approximating to the value of

that devised

by Newton

in

1666.*
of

=f(x) be the equation

real

in the figure represents a o. root of the equation f(x) Let OA be an approximate value of

curve,

OX

/b
to the curve; then

OX, and Aa the corresponding value oif(x). At a let aB be drawn tangent

OB

is

another approximate value of


infinitas, p.

OX^
I

* See Analysis per equationes numero terminorum


of Horsely's edition of

269, Vol.

Newton's works (London,

1779),

where the method

is

given

in

somewhat

different form.

Art.

4.]

newton's approximation rule.


of f(x)
;

Let Bb be the value


let

the tangent

bC be drawn

corresponding to OB, and at b then OC is a closer approxima-

tion to

the process may be continued. df(x)/dx. Let/'(-*0 be the first derivative of/(jtr); ox,f\x) x OA in the figure, the value of f(x ) is the ordiFor x

OX, and thus


x

nate Aa, and the value of f\x^)

is

the tangent of the angle

aBA

this tangent

is

also

Aa/AB.
(2(7

Hence
by

AB = /*,)//'(*i)

and accordingly

(9j5

and

are found

which
to

is

Newton's approximation

the closer value

is

third application found, and the process may be conrule.

By a

tinued to any degree of precision required.

For example,
ative is/'(;r)

let

f(x)
10^.

+ 5^

-}-

= o.

The

first

deriv-

$x*

-f-

Here

it

may

be found by

trial
x

that
2
1.92.

For x 2 is an approximate value of the real root. = 60, whence by the rule x = 5, and /'(#,) /(*,) =
a

Now

for

x^

1.92

are

found

/(#)

0.6599
1.906,

and
is

f(xj) 29052, whence by the rule correct to the third decimal.

which

As
the

second example
derivative
is

let

f(x)

first

f\x) =
by

= x* + 4 sin = o.
;tr

Here

2x

4 cos x.

An approximate

by a graphic solution is .*= u^cx/. For .*, = 1.94, 1.94, corresponding to about as 0.03304 and /'(*,) = 5-3 2 3> whence by the rule /(.*:,) a second application x = 1.9328, which x>= By 1.934.
value of
either
trial

x found

or

corresponds to an angle of
In

uo

54f

the application of Newton's rule it is best that the assumed value of x, should be such as to render /(*,) as small
as possible,
will fail
if

and

also /'(*,) as large as possible.

the curve has a


is

maximum

or

The method minimum between a


falsi,

and

b.

It

seen that Newton's rule, like the regula

and algebraic equaapplies equally well to both transcendental that the rule itself is readily kept in mind tions, and moreover

by help

of the diagram.

8
Prob.
5.

THE SOLUTION OF EQUATIONS.


Compute by Newton's
jx

[CHAP.

I.

braic equation x*

+6=
-f-

rule the real roots of the alge-

o.

Also the real roots of the trans2

cendental equation sin x

arc

0.

Art.

5.

Separation of the Roots.

roots of an equation are of two kinds, real roots and imaginary roots. Equal real roots may be regarded as a special class, which lie at the limit between the real and the imaginary. If an equation hasp equal roots of one value and q equal roots of another value, then its first derivative equation has 1 1 roots of the first value and q roots of the second p
value,

The

and thus

all

common

to both primitive

the equal roots are contained in a factor and derivative. Equal roots may

hence always be readily detected and removed from the given For instance, let x* gx* -\- ^x -\- 12 = o, of which equation.
the derivative equation
factor of these
is

^x*

\%x

-|~

=o

as

is

two equations, two


-

of the roots of the primitive

equation are

+2

The problem

of determining the

number

of the real

and

imaginary roots of an algebraic equation is completely solved by Sturm's theorem. If, then, two values be assigned to x the

number

of real roots

between those

limits

is

found by the same

theorem, and thus by a sufficient number of assumptions limits may be found for each real root. As Sturm's theorem is known
to
all

who

read these pages, no applications of

it

will

be here

given, but instead an older method due to Hudde will be presented which has the merit of giving a comprehensive view
of the subject,

and which moreover applies to transcendental

as well as to algebraic equations.*


If

abscissas

any equation y = fix) be plotted with values of x as and values of y as ordinates, a real graph is obtained
intersections with the axis

whose

OX give the

real roots of the


See CEuvres

* Devised by Hudde in 1659 and published by Rolle de Lagrange, Vol. VIII, p. 190.

in 1690.

Art.

5.]

SEPARATION OF THE ROOTS,

o. equaf ion fix) Xg've three values

Thus

in

OX for

three real roots.

the figure the three points marked The curve which

represents

= fix)

has points of

maxima and minima marked

A, and

inflection points

marked B.

Now

let

the

first

deriva-

tive equation

dy/dx=f'(x) be formed and be plotted in the same manner on the axis O'X'. The condition f'(x)= o gives
the abscissas of the points A, and thus the real roots O'X' give To ascertain if a o. limits separating the real roots of fix)
real root

OX lies between two values


=

of

O'X' these two values

are to be substituted infix): if the signs oi fix) are unlike in o lies between the two the two cases, a real root of fix)
limits
;

if

the signs are the same, a real root does not

lie

between

those limits.
In like

manner
f

d*y/dx* = f"{x)
limits

the second derivative equation, that is, be plotted on O" X" the intersections give
if
>

which separate the

real roots of

f\x)

o.

It

is

also

seen that the roots of the second derivative equation are the abscissas of the points of inflection of the curve y = fix). To illustrate this method let the given equation be the
quintic fix)

- $x* +6x + 2 = o. The first derivative 6 = o, the roots of which are equation is /'(*) = 5x* -i$x* + - 1.59, -0.69, +0.69, 4- 1.59. Now let each approximately
=
x*

of these values be substituted for

in

the values

- 00

o,

and

00

and
:

let

the given quintic, as also the corresponding values

of fix) be determined as follows

10

THE SOLUTION OF EQUATIONS.

[CHAP,

I.

bs

oo

I.59,

- O.69,

O,

+ O.69, +
+4.7,

I.59,

00

/(*)=-

00,

+2.4,

-0.6, +2,

+1.6,
and

+00.
1.59,

Since f{x) changes sign between x one real root lies between these limits

=
;

00

^=
0.69

since f(x) changes sign


real root lies
%

between x

=
;

1.59

and x^

0.69,

one

between

these limits

since f(x) changes sign

between x

and

= o,

one

real root lies

not change sign between x a nary roots is indicated, the

between these limits; since f(x) does = o and x A = 00 a pair of imagisum of which lies between -f- 0.69
,

and

00

e^ e* a second example let f(x) 2elx is fix) e* derivative equation

As

4 = 0. The first o, which has two

roots e*

For x
;r

= i and e* = = 00 /(jr)
,

o,

the latter corresponding to

^
***

=
;

00

is

negative; for e*

= i, /(^)
is

is

negative
-

for

-f-

00

f[x)

is

negative.
real roots.

The equation

**

=o

has, therefore,

no
first

When

the

derivative equation

not easily solved, the

second, third, and following derivatives may be taken until an equation is found whose roots may be obtained. Then, by

working backward,

limits

may be found
In

in succession for

the
the

roots of the derivative equations until finally those of

primative are ascertained.


ess

many
is

cases,

it is

true, this procit

may prove lengthy and

difficult,

entirely; nevertheless the

method
Zx

and one

in

some

may

fail

of great theoretical

and practical value.


Prob.
positive
6.

Show
Show

x that e

+ e~

4
1

has two real roots, one

and one negative.


7. 1

Prob.

that x*

-{-

= o has

x+

o has no real roots; also that

two

real roots,

one positive and one negative.

Art.

6.

Numerical Algebraic Equations.


th

algebraic equation of the with all its terms transposed to the

An

degree

may

be written
:

first

member, thus

xn

+ a.x"- + a x +
1

M ~2

+a

n _x

+a

Art.
arid

6.]

numerical algebraic equations.


the coefficients and the absolute term are real
is

11

if all

numfirst

commonly member may for brevity be denoted by f{x) and


itself

bers, this

called a numerical equation.

The

by /(V)

= o.

the equation

following principles of the theory of algebraic equations with real coefficients, deduced in text-books on algebra, are here recapitulated for convenience of reference
:

The

(i)

If

is

a root of the equation, /(*)


if

is

divisible
x

by x

xt ;

and conversely,
equation.
(2)
(3)

f(x)

is

divisible

by x

x lt then x

is

a root of the

An equation of the n th degree has n roots and no more. xn are the roots of the equation, then the prodIf x x%%
x ,

is equal tof(x). a the sum of the prodequal to ucts of the roots, taken two in a set, is equal to -f- # 2 the sum of
x

uct (x
(4)

x )(x x The sum of

t)

(x

xn )

the roots

is

x \

the products of the roots, taken three in a set, is equal to so on. The product of all the roots is equal to an and to -f- a n when n is even. odd,
(5)

a3

and
is

when n

The equation /(a;)

=o

ing

second term by substituting y (6) If an equation has imaginary


its

may be reduced to an ajn for x,*


roots, they
1.

equation lack-

occur in pairs of

the

form/
(7)

qi where / represents

sign

is

equation of odd degree has at least one real root whose opposite to that of a n
.

An

equation of even degree, having a n negative, has at least two real roots, one being positive and the other negative.
(8) (9)

An

variations in the signs of

complete equation cannot have more positive roots than its terms, nor more negative roots than
in signs.
If all roots

permanences
(10)

be

real, there are as

many

posi-

tive roots as variations,

and

as

many negative

roots as permanences. f

In an incomplete equation, if an even number of terms, 2w, are lacking between two other terms, then it has at least 2m say
*

By substituting y
remove

-\-py

+ q for x, the quantities / and q may be determined


means
of a quadratic equation,,

so as to

the second and third terms by

the second
fifth

and fourth terms by means of a cubic equation, or the second and


in 1639.

terms by means of a quartic equation. The law deduced by Harriot in 1631 and by Descartes

12

THE SOLUTION OF EQUATIONS.

[ChaP.

I.

imaginary roots; if an odd number of terms, say 2tn -f- i, are lacking between two other terms, then it has at least either 2tn -f- 2 or 2m imaginary roots, according as the two terms have like or unlike
signs.*

(11) Sturm's

theorem gives the number of

real roots,

provided

that they are unequal, as also the tween two assumed values of x.
(12) If a r
is

number

of real roots lying be-

the greatest negative coefficient, and


coefficient after
limits

if

as

is

the

greatest negative
real roots lie

between the
is

(13) If an
efficient,

the

first

all changed and (as \} !j0ui*m&y 5*7 negative and a r the greatest negative co-

is
1

into

x, then

ar

then a r
first

is

a superior limit of the positive roots.

If

a k be

the

negative and a s the greatest negative


x,

coefficient after

is

changed into

of the negative roots.

&

then a* /v/w

1 is

a numerically superior limit

(14) Inferior limits of the positive


l

found by placing x = z~ and thus obtaining an equation f(z) whose roots are the reciprocals of f(x) = o.

and negative roots may be

=o

is

z r where r (15) Horner's method, using the substitution x an approximate value of x enables the real root x to be comx ,
x

puted to any required degree of precision.


application of these principles and methods will be familiar to all who read these pages. Horner's method may

The

be also modified so as to apply to the computation of imaginary roots after their approximate values have been found. t

The

older method of Hudde and Rolle, set forth in Art. 5, is however one of frequent convenient application, for such algebraic

equations as actually arise in practice. By its use, with principles (13) and (14) above, and the regula together falsi of Art. 3, the real roots may be computed without any

assumptions whatever regarding their values. For example, let a sphere of diameter and

specific gravity

* Established by

DuGua;

see

fSheffler, Die Auflosung der algebraischen


n,

Memoirs Paris Academy, 1741, pp. 435-494. und transzeridenten Gleichung-

Braunschweig, 1859; and Jelink, Die Auflosung der hoheren numerischen

Gleichungen, Leipzig, 1865.

Art.
-

7.]

transcendental equations.

13

float in water,

mersion.

The

and let it be required to find the depth of iimsolution of the problem gives for the depth v

the cubic equation

As

a particular case let


x*

D=
3^
-f-

2 feet

and

-=0.65; then the

equation
2.6
is

=o
is

to be solved.

The

first

derivative equation

3**

6x
is

=o

whose roots are o and


tive root greater than 2.

2.

Substituting these, there


less

found

one negative root, one positive root

The

and one posiof the question physical aspect


2,

than

excludes the

first

and

last root,

By

(13)

and

(14) an inferior limit of this root


0.5
x

that

it lies

between

and for x^ 2, f(xt ) = For x 3 =i.35,/(^,) =


regula
falsi
is

and the second isto be computed. is about 0.5, so and 2. For x = 0.5, /(jr,) = -j- 1.975,
1.4;

then by the regula

falsi

#,= 1.35.

decimal,
Prob.
of

gives xK the correct depth of immersion of the sphere.


>

0.408, and combining this with x, the 1.204 f eet which, except in the last

8.

The diameter
is is

and which
10 feet
38.x:

to

whose length is 200 feet discharge 100 cubic feet per second under a head
of a water-pipe
real

ft>

M>

given

101
3!

= o.
7.

by Find the value of

the

root

of
x.

the
'

quintic
-

equation
JL

"

$2.

:i

21

^v

Art.
Rules
(1)

Transcendental Equations.

to (15) of the last article have no application to trigonometrical or exponential equations, but the general prin-

2-5 may be always used in attempting their solution. Transcendental equations may have one, many, or no real roots, but those arising from probciples

and methods

of Arts.

lems

in

physical science must have at least one real root.


of such equations will

Two

examples

be presented.

cylinder of specific gravity g floats in water, and it is If required to find the immersed arc of the circumference.
this

be expressed

in circular

measure
sin

it is

given by the transo.

cedental equation

f{x)

2ng =

14

THE SOLUTION OF EQUATIONS;


first

[CHAP.

I.

cos x derivative equation is I o, whose root is any even multiple of 2n. Substituting such multiples in f(x) it is found that the equation has but one real root, and that

The

this lies
is

substituting \n, \it% and n for x, further found that this root lies between f n and n.

between o and

2;r;

it

a particular case let g 0.424, and for convenience in using the tables let x be expressed in degrees; then

As

fix)

=x = 166

57

.2958 sin

152

.64.
=ss
x

Now proceeding by the regula falsi (Art. 3) let x 180 and and /(.r =-58. 16, whence ^=135, giving /(*,) =+ 27 36
.

2)

166

For x

f(x 3 )

.469,

and hence 166

is

an

approximate value of the root. Continuing the process, x is found to be i66.237, or in circular measure ^=2.9014 radians.
tal tension of a

a second example let it be required to find the horizoncatenary cable whose length is 22 feet, span 20 feet, and weight 10 pounds per linear foot, the ends being sus-

As

pended from two points on the same level. If / be the span, s the length of the cable, and z a length of the cable whose weight equals the horizontal tension, the solution of the problem leads
to the transcendental equation the numerical values,
7 (10 e
is

s=

L
2z

1
2z )

\e

z,

or inserting

IOV

- e~ T )z
first
-is\

O
is

the equation to be solved.


/

The
L

derivative equation

/<(*)

10/ = -V -e*)+-[e>+e

-i_\

')
is

= Q,
less
x

and

this substituted in f(z) 20.


x

shows that one

real root

than

about

Assume n =15, then /(.s^) =0.486 and f\z ) =0.206, whence by Newton's rule (Art. 4) z = 13 nearly. Next for
2

z^

= =

13, /(#,)

=
13.

0.0298 and f'(z 9 )


1

0.322,
/'(<&,)

whence z

13.1.

Lastly for z 3

f(z 3 )

=0.0012 and

= 0.3142,

whence

13.096,
6 e

which
e

is

a sufficiently close approximation.


is

The

horizontal tension in the given catenary


*Since
e~
,

2 sinh 6,

this

equation

hence 130.96 pounds.* may be written n0 iosinhQ,

where

I02 -1

and the solution may be expedited by the help of tables of


See Chapter IV.

hyperbolic functions.

Art.

8.]

algebraic solutions.
9.

15

Prob.

Show

that the equation 3 sin

2x

o has but

one

real root,

Prob.
2x-\- log

10.

and compute its value. Find the number of real roots of the equation 10 000 = o, and show that the value of one of them is

4995-74-

Art.

8.

Algebraic Solutions.

Algebraic solutions of complete algebraic equations are only possible when the degree n is less than 5. It frequently

happens, moreover, that the algebraic solution cannot be used to determine numerical values of the roots as the formulas
expressing them are in irreducible imaginary form. Nevertheless the algebraic solutions of quadratic, cubic, and quartic

equations are of great practical value, and the theory of the subject is of the highest importance, having given rise in fact
to a large part of

modern

algebra.

solution of the quadratic has been known from very early times, and solutions of the cubic and quartic equations were effected in the sixteenth century. complete investigation of the fundamental principles of these solutions was, how-

The

ever,

given by Lagrange in 1770.* This discussion showed, th if the general equation of the n degree, f{x) =0, be deprived o, that the of its second term, thus giving the equation /(j)
first

expression for the root

is

given by

y=
in

cos,-]-

oo\

oo

~^n-i

>

which n

is

cession, each

the degree of the given equation, th roots of unity, 1, e, e\ of the n

go is, in
. .
.

suc-

e n ~\

and

Slt sit

s M _, are

the so-called elements which in soluble cases


of the n
th
I

are determined
instance,
if

by an equation
3

degree.

For

the equation
00

is

of the third degree or a cubic,

the three values of


a?t

are
e,

l,

gd=

-i + i-/ 11! =

go

= -i-iV- 3 = e

*>

Lagrange

* Memoirs of Berlin Academy, 1769 and 1770; reprinted in CEuvres de See also Traite de la resolution Vol. II, pp. 539-562.
(Paris, 1868),

des 6quations numeriques, Paris, 1798 and 1808.

16

THE SOLUTION OF EQUATIONS.

[CHAP.

and the three roots are expressed by


y%
in

= h+s
3

>

y%
s2
5

^i

+ e\

iyi

es,

which

5,

and
9).

are found to be the roots of a quadratic

equation (Art.

go are the n roots of the binomial equation n be odd, one of these is real and the others if n be even, two are real and n are imaginary 2 are imagi1 o are -f- 1 and 1 those nary.* Thus the roots of go* 1 o are given above those of go* of go 1 o are

The n
1

values of
If
;

oo

= o.

-|*b

I, -fI

!>

an d
2
,

where
is
|

i
1,

is-v/

i-

For the equation

Go

= o the
e, e
1,

real root
e
3
,

denoted by

4
;

to find

and the imaginary roots are these let go" 1 be divided

=0

by

giving
4
1

GO

-\-GO-\-GO-\-GO-\-l=

O,

which being a reciprocal equation can be reduced to a quadratic, and the solution of this furnishes the four values,

=-i(l~ V~$ + V_IO-2i 5), e* = -l(i-V S - V-io-2Vj)>


6
/

=- j(l+ V5 + fC 10+2^
= -i(i+
1

3
<?

^5
3

^-10 +
1,

2V5),

where

it

will

be seen that
5

e.e

and

e .e

as should

be

the case, since e


In order to

1.

method

let it

solve a quadratic equation be of the form

by

this

general

x
and
let

+ 2ax + b =
a,
(

o,
it

x be

replaced

by y

thus reducing
b)

to

y_
Now
the product of
(

a*

_
y
x

o.

the two roots of this are

-f- s t

and y %

s%%

whence

y
a

s^)

and (y

-f%

st ) is

f - s = o.
Thus the value
*

of j

is

given by an equation of the

first

degree,

The values
1

of a> are, in

" vectors " drawn from the short, those of the n


first

center which divide a circle of radius unity into n equal parts, the
GOi

vector

being measured on the axis of real quantities.

See Chapter X.

Art.
s
7

9.]

the cubic equation.


b;

17

a*

and since x
1

-\-y,

the roots of the given

equation are
*x

= - a + y/d

-b,
th

*9

=-a

JlF^lt,

which

is

the algebraic solution of the quadratic.


of the
I

degree upon which the solution of the equation of the n th degree depends is called a resolvent. If such a resolvent exists, the given equation is algebraically solvable but, as before remarked, this is only the case for
;

The equation

quadratic, cubic, and quartic equations.


Prob.

n. Show

that

the

six

6 th

roots

of

unity are

-f- i,

+i(x+ v^3), -*(i- t^). ~h -i(i+


Art.
9.

^J,

-i(i- V=^).

The Cubic Equation.

the result
cubic be

All methods for the solution of the cubic equation lead to commonly known as Cardan's formula.* Let the

x
and
let

-f $ax* -f

Tfix

+ 2c =

o,

(i)

the second term be removed by substituting y x, giving the form,

a for

y + 3fy+2c=o,
B and C are
b,
7

(i')

in

which the values of

B=-a +
Now by
1

C=a*
es,

%ab

c.

(2)

the Lagrangian
1

method

of Art. 8 the values of

y
,

are

y =zs
in

+S
and
e
a

y, ==

eV9

7,

e^ 2

which

Forming the products


e"

are the imaginary cube roots of unity. of the roots, and remembering that
1

and

-|-

-f-

= o,

there are found

at* +y,y,+y>y* For the determination


tions from
s
6 3

= - 3v. = + $b> yj>y* = s? + s?= -2C.


s,

of

and

s2 there are

hence two equa-

which
9

results the quadratic resolvent


o,

+ 2Cs B = and thus s ={-C+V& +')*>


%

s9

= (-C-V#+Cy.
published by Cardan in 1545.

(3)

* Deduced by Ferreo in 1515, and

first

18

THE SOLUTION OF EQUATIONS.


of the roots of the cubic in

[CHAP.

I.

One

therefore

is

/i

= (r c+/FF?')
is

+H-* ^ +
/
,

"*)*.

and

this

the well-known formula of Cardan.

algebraic solution of the cubic equation (i) hence conand C by (2) in terms of the given coefficients, sists in finding

The

and then by

(3)

the elements s and s a are determined.


x

Finally,

*, =

(*,

*.),

*,= *,=
When B
is
9

ft*.

*- i(s,

+ + i V'^is, - sj, + - i^/~^r - sj,


s >)
s,)

(4)

3 ( Sl

which are the algebraic expressions of the three

roots.

C* is negative the numerical solution of the not possible by these formulas, as then both s and j a are in irreducible imaginary form. This, as is well known, is
cubic
k

the case of three real roots, s a pure imaginary.* When B*


x

+
-f-

s^

C*

s7 being a real, while s is o the elements s and


1 x

is

s^
**,

are equal, and there are two equal roots, a while the other root is xx 2O.

xt

=x =
3

-f-

When B

-f-

is

positive the equation has one real

and

two imaginary
cubic

roots,

and formulas

(2), (3),
(1).

and

(4) furnish the

numerical values of the roots of

For example, take the


5

= o, = +4 whence by comparison with (1) are found a = 1.5, b Then from (2) are computed B = 1.75, c^+3.125. c 2.5. These values inserted in (3) give s = +0.9142, ^ = 1.9142 ^ = + 2.8284. 1.0 and ^ thus s 4- s = Finally, from (4) - 1.0 = +0.5, x = 1.5 = 2 + 2.4495/, x, = 1.5 +0.5 + 1.4142 V^~3 /:= ^ = 1.5 + 0.5 1.4142 4 ~3 = 2 2.4495/,
x
%

4.5^

+ \2x

which are the three roots of the given cubic.


* The numerical solution of
cosine
is

this case

is

possible

whenever the angle whose

C/

4/

can be geometrically trisected.

ART.

10.]

THE QUARTIC EQUATION.


12.
2

19
o.

Prob.

Compute
5.76.x

the roots of x*

2x

of x

-f-

o.6^
13.

+ 4.32 =

5=

Also the roots

o.

its altitude 6 inches and the diameter of placed with vertex downwards and one fifth of its volume is filled with water. If a sphere 4 inches in diameter be then put into the cone, what part of its radius is immersed in the water ? (Ans. 0.5459 inches).

Prob.

cone has
It is

its

base

5 inches.

Art.

10.

The Quartic Equation.


in

The quartic equation was first solved who separated it into the difference of two
in 1637 resolved
it

1545 by Ferrari,

squares.

Descartes

two quadratic factors. Tschirnhausen in 1683 removed the second and fourth terms. Euler in 1732 and Lagrange in 1767 effected solutions by
into the product of
roots.

assuming the form of the

All these methods lead to


first

cubic resolvents, the roots of which are

to be found in

order to determine those of the quartic.

The methods
similar, first

and Lagrange, which are closely reduce the quartic to one lacking the second term,
of Euler

y + 6B/ + 4Cj> + D = o;
and the general form
of the roots being taken as
3
,

= + Vs + V7 + V7 - VI, - V7 y, = + V7
y
x

3,

y y
2

= - Vs. + VJ, - Vi = -V7 - v7 + Vs.,


x

3 ,

the values slt s9 st are shown to be the roots of the resolvent,


,

+ $Bs* + i(9^ 9.

)*

~ iC* = o.
is

Thus the
solvable

roots of the quartic are algebraically expressed in f

terms of the coefficients of the quartic, since the resolvent

by

the process of Art.

Whatever method
final

of solution be followed, the following

formulas, deduced by the author in 1892, will result.* Let the complete quartic equation be written in the form
x"

+ 4ax> + 6bx* +4cx + d=o.

(1)

* See American Vol. XIV, pp. 237-245. Journal Mathematics, 1892,

20
First, let g, k,

THE SOLUTION OF EQUATIONS. and k be determined from


s

[CHAP.

g=a*-b, h=b
Secondly,
let /

c'

-2abc

+ dg,

k=

%ac-V-\d.

(2)

be obtained by
h

= \{li + VW+T>) + \{h v,

VF+J*)*
\2gL

(3)

Thirdly, let u,

and
v

be found from
(4)

u=g+/,
Then

= 2g-J, w = 4u* + $k

the four roots of the quartic equation are

x
#

x
x

= = = =

+ Vu + ^v + Vw, a + Vu *v + Vw,
a a

(5)

Vu
Vu

-\-

Vy

Vw,
Vw,

Vv

in which the signs are to be used as written provided that 3 c is a negative number; but if this is positive all 2a lab

radicals except

Vw

are to be reversed in sign.

These formulas not only serve


discussion of the quartic
tions to be

made
is

for the complete theoretic but they enable numerical solu(1), whenever (3) can be computed, that is, when-

ever A* -\-k*

positive.
roots.

For
If
%

this case the quartic has

two

real

and two imaginary

there be either four real roots or

four imaginary roots J? -f- k is negative, and the irreducible case arises where convenient numerical values cannot be obtained, although they are correctly represented

by the formulas.

As an example
and
let
it

let

a given rectangle have the sides

/ and

q,

be required to find the length of an inscribed recwhose width is m. If x be this length, this is a root of tangle
the quartic equation

_ (#

_|_ tf _|_

m )x
i

+ ^pqmx

(/ + tf

n?).*

= o,

and thus the problem is numerically solvable by the above formulas if two roots are real and two imaginary. As a special case let p = 4 feet, q = 3 feet, and m == 1 foot then
;

x*

2jx'

-f-

48^

24 = o.

*"

4-1.

On** '*

"M

ART.

11.]

QUINTIC EQUATIONS.
(i)

21

By
and

comparison with

are found a
(2),
3

^=24.

Then from
is

^=
is

+ -V3.6067.

= o, b =+ g

1
-,

41.,

^
(3)

= + 12,
4

an(j

Thus ^* -j-

positive,

and from

the value of /
12.6067,

From
161.20.

(4) are

and

w= +
x x x
x

now found, u = +0.8933, v=

Then, since

c is positive, the values of the

four roots are, by

(5),

x9
A

+ V'12.607+ 12.697 = + 4.085 = - 0.945 + V12.607 - 12.697 = + 0.945 + 0.30/, = 0.945 V 2.607 12.697 = + 0.945 0.30/,
0.945
feet,
1

= =

0.945

1/12.607

+ 12.697 =

5.975 feet,

Each of evidently the required length. these roots closely satisfies the given equation, the slight discrepancy in each case being due to the rounding off at the third
is

the second of which

decimal.*
Prob.
14.

Compute

the roots of the equation

xA +

'jx

+ 6 = o.

{Ans.

1.388,

1. 000, 1.

194

1.

701/.)

Art.

11.

Quintic Equations.
fifth

The complete
ally solvable, nor

equation of the
is it

degree

is

reducible to a solvable form.

not algebraicLet the

equation be
x*

+ $ax* + $bx* + $cx* + $dx + 2e

o,

and by

substituting

/+ S#/+SCf+SDj> + 2E =
according to Art.
$*
*

for

let

it

be reduced to
o.

The

five roots of this are,

8,

y,

y,
y*

y*
y>
in which
if
e, e',

= + + s + so = &i + \ + e\ + \f = + e% + es + e\, = eV, + es + eV + e's = e% + e % + s + S


*,

6**i

t,

*>

are the imaginary fifth roots of unity. Now the several products of these roots be taken there will be
e
3
.

e*

* This example is known by length of a strut in a panel of the

civil

engineers as the problem of finding the


truss.

Howe

22

THE SOLUTION OF EQUATIONS.


(4) of

[CHAP. I*

found, by

Art.

6,

four equations connecting the four ele-

ments

s lf s
4

s3

and

s<,

namely,

B = s,s
C

+ Va> + sfr + s \ + s<\, D = V'd + V*i + V*i + *A - *iV s*s 3


t

*i**

+ ^W.

degree for
s 1 si
s^s 3
s,

50,

V< + *,V. + ^

V + ^Vt)
4

>

but the solution of these leads to an equation of the 120th

However, by taking quantity, a resolvent of the 6th degree is obtained, and all efforts to find a resolvent of the fourth degree have proved unavailing.
or s*

or of the 24th degree for s\ b * -C as the -\- s 3 -f- s 3

unknown

Another line of attack upon the quintic is in attempting to remove all the terms intermediate between the first and the last. By substituting y* + py -\-q for x the values of p and q be determined so as to remove the second and third terms. may by a quadratic equation, or the second and third by a cubic equation, or the second and fourth by a quartic equation, as was first shown by Tschirnhausen in 1683. By substituting y* -{-py* -f- qy -f- r for x, three terms may be removed, as was shown by Bring in 1786. By substituting y*-\~Py -\- qy -\-ry-\-f for x it was thought by Jerrard in 1833 tnat f ur terms might be removed, but Hamilton showed later that this leads to
y

equations of a degree higher than the fourth.


In 1826 Abel gave a demonstration that the algebraic soluof the general quintic is impossible, and later Galois

tion

published a more extended investigation leading to the same conclusion.* The reason for the algebraic solvability of the
quartic equation may be briefly stated as the fact that there exist rational three-valued functions of four quantities. There
are,

however, no rational four-valued functions of five quantities, and accordingly a quartic resolvent cannot be found for
*
Jordan's Traite des substitutions et des equations algebriques; Paris, 1870.

the general quintic equation.


Abhandlungen iiber die algebraische Auflosung der Gleichungen von N. H* Abel und Galois; Berlin, 1889.

ART.

11.]

QUINTIC EQUATIONS.

23
of the quintic of these is the

There are, however, numerous special forms whose algebraic solution is possible. The oldest
quintic of

De

Moivre,

f + SB/ + $&y + 2E = o,
which
sx
is

solved at once by making


;

equations

then

=: s s x 4

and
t

s,

2
,

and

s A are

found, and

=s
are
e's 1

= s = o in the element = + s,\ from which


9

s,"

-\- s t

or

y
while

=(-+ V& + &)* + (- E - VW+&)\


other
roots

the

=
4

es%
.

e\

eV 4

and j 6

+ e^
if

If

+ e\ y = eV, + e ^ B + ^ be negative,
3
,

4 ,

this quintic

has

five real roots;

positive, there are

one

real

and four imaginary

roots.

When any

relation, other than those expressed


t , , ,
,

by the four

element equations, exists between s st st s4 the quintic is solvable algebraically. As an infinite number of such relations

may be

stated,

it

follows that there are an infinite

number

of

In each case of this kind, however, the cosolvable quintics. efficients of the quintic are also related to each other by a
certain equation of condition.

solution of the quintic in terms of one of the roots of its resolvent sextic was made by McClintock in 1884.* 6 6 By this method ^ x s2 \ s3 \ and ^ 4 are expressed as the roots of
,

The complete

a quartic in terms of a quantity / which is the root of a sextic whose coefficients are rational functions of those of the given
quintic.

Although

this has great theoretic interest,

it

is,

of

course, of little practical value for the determination of


ical

numer-

values of the roots.

By means

of elliptic functions the complete quintic can,


first

-shown by Hermite in 1858. For this purpose the quintic is reduced by Jerrard's transformation to the form x $dx-\-2e = o, and to this form can
however, be solved, as was
b

also be

reduced the

degree.

modular equation of the sixth Other solutions by elliptic functions were made by
elliptic

* American Journal of Mathematics, 1886, Vol. VIII, pp. 49-83-

24

THE SOLUTION OF EQUATIONS.


in 1861

[CHAP.

I.

and by Klein in 1884.* These methods, the help of tables, have not yet been sysby though tematized so as to be of practical advantage in the numerical
Kronecker
feasible

computation of

roots.

Prob. 15. If the relation ^4 b b * s3 s4 s% show that sf

+
b

= Vt - 2 E. =
of

exists

between the elements


3

Prob. 16.

Compute
3

the roots
-f-

also those of y

ioy

2oy + 6 = o.

+ ioy + 2qy + 6 = o, and

Art.

12.

Trigonometric Solutions.

When a cubic equation has three real roots the most convenient practical method of solution is by the use of a table of If the cubic be stated in the form (1) of sines and cosines.
Art. 9, let the second term be removed, giving

f + $By +2C= o.
Now
suppose y=2r sin
3

0,

then this equation becomes


/?

C
0+2-i

8 sin 0-)-6-8 sin

= o,

and by comparison with the known trigonometric formula 6 sin 6 -\- 2 sin 30 = o, 8 sin' 6
there are found for r and sin 30 the values
r
in

=V
always

By

sin

30

C/

B\

which

B is

negative for the case of three real roots


is

(Art. 9). of sines,

3$ being computed, 3# and then 6 is known. Thus,


sin 6,

Now sin
y
2

found from a table

= 2r

= 2r

sin (120

6),

= 2r

sin (240

0),

are the real roots of the cubic in y.\


* For an outline of these transcendental methods, see Hagen's Synopsis der
I, pp. 339-344. 2 z is negative and numerically less than C' as also when In positive, this solution fails, as then one root is real and two are imaginary. this case, however, a similar method of solution by means of hyperbolic sines

hoheren Mathematik, Vol.


f

When

is

is

possible.

See Grunert's Archiv fur Mathematik und Physik, Vol. xxxviii,

pp. 48-76.

fV

Art.

12.]

trigonometric solutions.

25

For example, the depth of flotation of a sphere whose diamis 2 feet and specific gravity 0.65, is given by the cubic - 3*2 2.6 = o (Art. 6). Placing x = equation x* 1 this y reduces to/ 0.6 = 0, for which B = i and C 37 =+0.3. Thus r = 1 and sin 30 = Next from a table of sines, 0.3. 30 = 17 27', and accordingly 0=5 49'. Then
eter

+ +

y y
Adding
1

=2 =2

sin

49/

sin 125

J,

= 2 sin
*
a

245

= -f 0.2027, 49' = + 1.6218, 49' = 1.8245.


x
are
feet,

to each of these, the values of


feet,

x
and

= + 1.203

= + 2.622

= -0.825

feet

first

evidently, from the physical aspect of the question, the of these is the required depth. It may be noted that the

number 0.3 is also the sine of 162 u', but by using this the three roots have the same values in a different order.
the quartic equation has four real roots its cubic reIn this case the formulas of solvent has also three real roots.
Art. 10 will furnish the solution
if

When

the three values of / be ob-

tained from (3) by the help of a table of sines.

The

quartic

being given, g, h, and k will always be negative for four


r

k are found as before,


real roots.

and the value of

Then

=V

k,
;

sin

30

=
is

h/r\

and 30

is

taken from a table


/ are
0,

thus

known, and the three

values of
/
x

r sin

/f

= r sin(i2o +
of u, of v,

0),

/s

= r sin

(240

0).

Next the three values

and of

those selected which give u, w, and v ties. Then (5) gives the required roots of the quartic.

w are computed, and Vw all positive quanti-

As an example,
Art. 10, and let/

take the case of the inscribed rectangle in


3 feet,

= 4 feet, q =
>

m=
156

Vi$

feet;

then the

quartic equation
jr

is

i_5i jr

+ 48 VT$x

= 0.

26

THE SOLUTION OF EQUATIONS.

[CHAP.

I.

Here a
"

o, b

8J-,

=+ 8, ^=
3$ to be

*f*,

= + 12 ^13, and d = 156. and = The trigonometric


/

-jL-

Next work
/

now
sin

begins; the value of r

is

found to be

i607
/,

0.7476; hence from the table 3$ = 4823 and // The three values of / are then computed 40

+ 4f,

and that of
,

by logarithmic

tables,

and found to
/,

be,
/4

= + 1.250,
of
?/,

= + 3.1187,
w are

=- 4.3687.
it is

Next the values

v,

and

obtained, and

seen that

only those corresponding to lx will render all quantities under these quantities are u the radicals positive 1 9.75, v 5-75>
;

=
x

and
*, =

w=

192.0.

Then

the four roots of the quartic are


*,

8.564,

x,

=+ 2.319,

= + 1.746,

= + 4.499

feet,

which only the second and third belong to inscribed rectangles, while the first and fourth belong to rectangles whose
of

corners are on the sides of the given rectangle produced.

Trigonometric solutions of the quintic equation are not b a, and the quintic of possible except for the binomial x
general trigonometric expression for the root its second term isy=2r 1 cos^,+2r2 cos at vof a quintic lacking and to render a solution possible, r and r2 as well as cos d
x
,

De

Moivre.

The

and cos#2 must be found; but these


,

in
:

general are roots of


3

equations of the sixth or twelfth degree in fact r, is the same as the function s s t of Art. II, and r* is the same as s^s 3
.
x

Here cos#, and cos# may be either circular or hyperbolic cosines, depending upon the signs and values of the coefficients
2

of the quintic.

Trigonometric solutions are possible for any binomial equation, and also for any equation which expresses the division of
an angle into equal
parts.
,

Thus the
which

roots of

+ =0 are
1

cos^ 30 The roots


where

i sin

m 30

in

m
5

has the values


6

1, 2,

and

3.

of

$x*-\-$x

2 cos

= o are
4.

2 cos

(m

72+$)

has the values


17.

o, 1, 2, 3,

and

Prob.

Compute by

a trigonometric solution the four roots of

the quartic

xA

+ 4#

24^

76^

29

= 0.

(Ans.

6.734,

1.550,

+ 0.262, + 4.022).

Art.

13.]

real roots by
1 8.
z

series.

2?

Prob.

$bx

+ 5^ x
when
b

Give a trigonometric solution of the quintic equation


2e
\

to* the case of five real roots.

the roots
0.2864,

and

e=

Compute
1.3952*

0.752798.

(Ans.

1.7940,

o-WI,

i-97 10 -)

Art.

13.
in

Real Roots by

Series.

The

value of

as an infinite series.

any algebraic equation may be expressed Let the equation be of any degree, and
first

power

by dividing by the coefficient of the term containing the of x let it be placed in the form
a

= x+ bx + cx + dx' -f ex*+fx* +
%
z

Now
By

let it

be assumed that x can be expressed by the series

= a + mc? + na

-\-

pa

-\-

qc?

inserting this value of x in the equation and equating the coefficients of like powers of #, the values of m, n, etc., are

found, and then

x=a -

ba'

+ (2b' -c)a -(sb -sbc+d)a


2

-{-(i4b

-2ib'c + 6b^+y' -e)^

-(42^-

84

V+28V+28^ -

>jbe- 7cd

+ f)a

+.

.,

In order is an expression of one of the roots of the equation. that this series may converge rapidly it is necessary that a

should be a small fraction.*

apply this to a cubic equation the coefficients d, e,f etc> % 0.6 o; For example, let x to o, 3* x hence a = 0.2^ form is 0.2 this reduced to the given \x\
f

To

are

made equal
o, c
-J,

= and x = 0.2 + i

then
.

0.2

+i

0.2

etc.

= + 0.20277,

which

is

decimal place.

the value of one of the roots correct to the fourth This equation has three real roots, but the

series gives only

one of them

the others can, however, be

found
is

equation

one root approximate values are known. Thus, l.6 there results an about +1.6, and by placing 0.0218,. whose root by the series is found to be in
if

their

x=y +
x
%

and hence

+ 1.62 18
is

is

another root of
J. B.

3*

*This method

given by

Mott

in

The Analyst,

1882, Vol. IX, p. 104.

28

THE SOLUTION OF EQUATIONS.

[Chap.

I.

Cardan's expression for the root of a cubic equation can be expressed as a series by developing each of the cube roots by
the binomial formula and adding the results. Let the equation be y* -J- ^By -\- 2C o, whose root is, by Art. 9,

= (_ C + VJ? + Cy +(-C- VBr+C>)\


series,

then this development gives the

^
*
in
v
'

2 2

2.5.8,
2.3.4

2. 5. 8. 11.

14, 2.3.4.5.6

/'

which r represents the quantity (* O/Sw* If r the equation has two equal roots and the third root is 2( C)$. If r is numerically greater than unity the series is divergent, and the solution fails. If r is numerically less than unity and
sufficiently small to

make a quick convergence, the series will serve for the computation of one real root. For example, take % 2 and C 6 6x the equation x o, where 3
hence r

+ =

B=

1/8

1,

and one root


0.01235

is

2.8845(1
is

0.00051

0.00032)

2.846,

which

correct to the third decimal.


is

In comparatively few

cases, however,

this series of value for the solution of cubics.

other series for the expression of the roots of equations, particularly for trinomial equations, have been devised. One of the oldest is that given by Lambert in 1758, whereby

Many

the root of

xn

+ ax

=o

ascending powers of b/a. Euler and Lagrange. These series usually give but one root, and this only when the values of the coefficients are such as to
render convergence rapid.

developed in terms of the Other solutions were published by


is

1771), pp. 143-150,

Prob. 19. Consult Euler's Anleitung zur Algebra (St. Petersburg, and apply his method of series to the solution of

a quartic equation.

Art.

14.

Computation of all Roots.


and valuable method for the solution of was developed by McClintock, in 1894, by

A comprehensive
equations by
series

ART.

14.]

COMPUTATION OF ALL ROOTS,

29
this

means

of his Calculus of Enlargement.*

By

method

all

the roots, whether real or imaginary, may be computed from a The following is a statement of the method as single series,
applied to trinomial equations
:

Substitute

= nAx + B be the given trinomial equation. = By and thus reduce the equation to the form ' = nay + where a = A/B Then B positive, the
Let x
n n n~ k

if

is

roots are given

by the
l

series

= co -{-co -* a + oo 1
l

2k

(i

2k-\- n)c?/2

-\-oo

-^{\

4k-\-n)(i

4k-\-2n)(\

4^+3^/4 !+...,

in
If,

which

0,1

however,

represents in succession each of the roots of unity. n is negative, the given equation reduces to

= nayn ~

1,

and the same

series gives the roots


I.

if

co

be

taken in succession as each of the roots of


In order that this series

n convergent the value of a k~ H thus for the quarmust be numerically less than k~\n k) the value of a must and k where n tic y* 3, 4 1, 4ax

may be

be

less

than 27-*.

/=
in

apply this method to the cubic equation x'=z$AxB\ Bx. It then becomes n 2, and put y 3 and k place and the series is I where a A/B\ lay

To

y=

co

+ Go*a

$Goa*

+ iafa*

-(-

which the values to be taken for co are the cube roots of 1 3 2x For example, let x 5 =0. or 1, as the case may be. Here #=0.228, this reduces toy =0.684 y+i.
Placing
co

y= $*x,

and as

this is less

than 4-* the series


is

is

convergent.

Making

i,

the

first

root

y sss
American

I -J-

0.2280

O.OO39

+ O.OOO9 =

I.2250.
p. 3; also

Vol. See Bulletin of American Mathematical Society, 1894,


Vol. Journal of Mathematics, 1895,

I,

XVII, pp. 89-110.

SO

THE SOLUTION OF EQUATIONS.

[CKAP,

I.

Next making go = and the corresponding

+iV
root
is

3, go

is

3,

found to be

Again, be the conjugate imaginary of the second. each value of y by 5*,

= - 0.6125 + O.3836 V^. \ \V 3 the third root making go =


y
x
2.095,

is

found to

Lastly, multiplying
I-I3 6

= - 1.047
x
3

V^,
to

which are very nearly the roots of x*


In a similar manner the cubic
-\3

=
i

= o. 2^ + 5 =0 reduces
2x
5
I,

0.684J/

1,

for

which the

series is convergent.

Here

the three values of

go are,
*

in succession,

y = 0.388

+i

3>

anc

the three

roots are

+4 ~ 3 y= 0.777
J
'

and

1. 1 37*'.

When
fails

all

the roots are

real,

the method as above stated

The given equation can, because the series is divergent. k roots by one however, be transformed so as to obtain n of the general series and k roots by another. As application
an example,
tion this
is

o. 330 to be written in the form

let x*

243^

-f-

For the

first

applica-

x=z S_
243
for which n

330
243'

and k

2.

To make

the last term unity

place

ss

-v,

and the equation becomes

whence a

= 330 /3.243'.
8

These values of

k,

and a are now

inserted in the above general value of y, and go made unity; thus ^=0.9983, whence x x =1.368 is one of the roots. For

the second application the equation


*>

is

to be written

= _33V, + 243> T *" 243

Art.
for

15.]

conclusion.

31

which n

= 2 and k =

3.

Placing

243*7,

tms becomes

whence #
values of

and

go is

110/243', and the series is convergent. These and a are now inserted in the formula for y, n, k, 1 in succession, thus 1 and made giving two

values for y, from which x^ 14.86 and other roots of the given cubic.

16.22 are the

McClintock has

method applicable

The equation
na
.

is

also given a similar and more general to other algebraic equations than trinomials. reduced to the form y n na (J>y 1, where

<py

Then

denotes all the terms except the first and the the values of y are expressed by the series

last.

^=(+

(w I

-"0^.^+^

~M

^^
n

~w

(0^)
n

.-j

+{^in

^^- ^-i\+
when
the conver-

which the values of go are to be taken as before. The method is one of great importance in the theory of equations, as it enables not only the number of real and imaginary roots
to be determined, but also gives their values gence of the series is secured.
Prob. 20.
quartic x*

Compute by the above method


10

all

the roots of the

+*+
this

= 0.

Art.
While

15.

Conclusion.

to the theory of Chapter forms a supplement

equations as

in college text-books, yet the to it has prevented the discussion and debrief space allotted Chief among these velopment of many interesting branches. of is the topic of complex or imaginary roots, particularly their graphical representation and their numerical computation. are required in the solution Although such roots rarely, if ever, is a matter of problems in physical science, their determination

commonly given

of

much

theoretic interest.

It

may be

mentioned, however,

32

THE SOLUTION OF EQUATIONS.

[CHAP.

I.

that both the regula falsi and Newton's approximation rule may, by a slight modification, be adapted to the computation
of these imaginary roots, approximate values of
first

them being

obtained by

trial.

A
which

method

of solution of numerical

algebraic equations,

may

be called a logarithmic process, was published by

in deriving

Graffe in 1837, and exemplified by Encke in 1841.* It consists from the given equation another equation whose

roots are high powers of those of the given one, the coefficients of the latter then easily furnishing the real roots and the moduluses of the imaginary roots. The method, although

without doubt one of high practical value, as logarithmic tables are used throughout; moreover, Encke states that the time required to completely solve an equation of
little

known,

is

the seventh degree with six imaginary roots, as accurately as can be done with seven-place tables, is less than three hours.
algebraic solutions of the quadratic, cubic, and quartic equations are valid not only for real coefficients, but also for

The

imaginary ones. In the latter case the imaginary roots do not The method of McClintock has the necessarily occur in pairs.
great merit that
it

nary coefficients;
Prob. 21.
tion

it

is applicable also to equations with imagiconstitutes indeed the only general method

by which the roots


x
3

in

such cases can be computed.


series the roots of the equa-

Compute by McClintock's
1

ix

= o.

Prob. 22. Solve the equation cos

equation x

e*

= o.

x cosh.* -j-i o, and also the answers see Crelle's Journal fur Mathe(For
pp. 1-62.)

matik, 1841, Vol.

XXII,

*See

Crelle's Journal fur

Mathematik, 1841, pp. 193-2480

Art.

1.]

INTRODUCTION.

33

Chapter

II.

DETERMINANTS.
By Laenas Gifford Weld,
Professor of Mathematics in the State University of Iowa.

Art.

1.

Introduction.

early as 1693 Leibnitz arrived at some vague notions regarding the functions which we now know as determinants.
this subject, the first account of which is contained in his correspondence with De L'Hospital, resulted simply in the statement of some rather clumsy rules for elimiin

As

His researches

nating the unknowns from systems of linear equations, and exerted no influence whatever upon subsequent investigations in the same direction. It was over half a century later, in

formulated an intelligible and general definition of the functions, based upon the recognition of the two classes of permutations, as presently to be set forth.
first

1750, that Gabriel

Cramer

Though Cramer failed to recognize, even to the same extent as Leibnitz, the importance of the functions thus defined, the
development of the subject from this time on has been almost continuous and often rapid. The name " determinant" is due
to Gauss, who, with

Vandermonde, Lagrange, Cauchy, Jacobi, and others, ranks among the great pioneers in this development.
Within recent years the theory of determinants has come

into very general use, and has, in the hands of such mathematicians as Cayley and Sylvester, led to results of the greatest
interest

and importance, both through the study of special forms of the functions themselves and through their applica*

tions.*

list

of writings

on Determinants

is

given by Muir

in

Quarterly Journal

of Mathematics, 1881. Vol.

XVIII, pp. 110-149.

34

DETERMINANTS.

[CHAP.

II.

Art.

2.

Permutations.

The

be arranged

various orders in which the elements of a group in a row are called their permutations.

may

Any two
orders
:

elements, as a and
ba.

ab and

third, as
in

c,

b, may be arranged in two may be introduced into each


:

of these

two permutations
;

three ways

before either element,

or after both

thus giving

3X2 = 6 permutations of the three

manner an additional element may be introduced into each of the permutations of i elements in (i -\- 1) before any one of them, or after all. Hence, in ways general, if Pt denote the number of permutations of i eleelements.
In like
:

ments,
jP4

P = (r+ 1)/^ = 4 X 31=4!; and, n Pn = n{n


i+l
is,

Now,

P = 3x2x1=3!;
%

hence

being any integer,


i)(n
2)
.
.

That

the
all

number

of permutations of n elements

is

!.

For
If

integral values of n greater than unity, n

is

an

even number.
the elements of any group be represented by the different letters, a, b, c, ., the alphabetical order will be considered
. .

as the natural order of the elements.

If
:

represented by the

same

letter with different indices, thus

a tt a

or thus

a',
is

a", a'"

.,

the natural order of the elements

that in which the indices

form a continually increasing

series.

elements, whether adjacent or not, standing in their natural order in a permutation constitute a permanence

Any two
in

standing
inversion.

the reverse of the natural, an in the permutation daecb, the permanences Thus,
is
;

an order which

are de, ae, ab, ac

the inversions, da, dc, db,

ec, eb, cb.

into

of the elements of a group are divided even or positive permutations, in which the number of inversions is even and odd or negative permu-

The permutations
two
classes, viz.:

tations, in

which the number of inversions

is

odd.

ART.

3.]

INTERCHANGE CF TWO ELEMENTS.

35

When
number

the elements are arranged in the natural order the of inversions is zero an even number.

#,,

Thus, the even or positive permutations of the elements a a % are a <z 2 a 3f #a a 3 alf a a, # a
t 3
;

while the odd or negative permutations are a 3 a t a lt a, a 3 a a2 a a%


x

Art.
It will

3.

Interchange of
that
if,

Two Elements.

now be shown

elements of a group, two of


class of the permutation will

any permutation of the the elements be interchanged the be changed.


in in question.

Let q and

be the elements
all

Then, represent-

ing collectively

P those which
%

fall

the elements which precede these two by between them by R, and those which follow
of the

by

any permutation

group

may
in

be written

PqRsT.

Of

the elements R

supposed to be r
of
"

number,

let
q,
q,
s,

represent

h the
i

number
"
"

"
"

an order higher than " " " " lower


" "
" "

j
k

"
"

lower
higher

"
"
s.

"

order of the elements qRs or T. can affect their relations to the elements of either Then, passing from the order PqRsT to the order
It is evident that

no change

in the

PRqsT
changes the number
this to the order

of inversions

by

{h

i)

and passing from

PsRqT
again changes the
j

number

of inversions

by (J
(

k)

I,

the
'

minus

Sign
(

bdng USCd
in the

aS q

Ser
is,

0rder than S

The

total

change

number
in

of inversions

due to the

inter-

change of the two elements

question

therefore,

36

DETERMINANTS.
i

[CHAP. IL

But since

=r

h and k

== r

j\ this

may be

written

A h +J ~
which
is

r)

an odd number for

all

admissible values of

//,/,

and r

Hence, the interchange of any two elements in a permutation ch anges the number of inversions by an odd number, thus
changing the class of the permutation.

Art.

4.

Positive and Negative Permutations.

Of

all

half are even

the permutations of the elements of a group, one and one half odd.
this,

To
tation.

prove

write out

all

the permutations.
in

Now choose
each permu)

any two

of the elements

and interchange them


be the same
set of
(

The

result will

before, only differently arranged.

But each

permutations as even > permutaQ

tion of the old set has been converted into an

the new.
tations as
Prob.
(i)
i.

Hence, in odd that


;

either set,
is,

even ( one in there are as many even permuj

one half are even and one half odd.

Classify the following permutations:


;

bcdea
a"
a* a'

(2)

m vi

II iv;

(3)

fteyZafi;. (8) F.Tu. M. Th. W.; (7) x x xt x;x;x% xt Prob. 2. Derive the formula for the number
(4)

a"

a'"; (5)
\

(6) (9)

knimlj; 52413;
fiKviX.

of permutations of

n elements taken
Prob.
3.

at a time.

(Ans. n\/\n

m)\.)

How many
may
Show Art.

natural order
n\/tn\(n

combinations of m elements arranged in the be selected from a group of n elements? (Ans.

m)\.)
4.

Prob.

that o!

1.
(

y\~y\[

.'!:_.

5.

The Determinant Array.


:

elements arranged in n vertical ranks or columns^ and n horizontal ranks or rows, thus
n*
.

Assume

a( a,"

a^

'

a "

(M)

Art.

6.]

determinant as function of

n*

elements.

37

In this array all the elements in the same column have the same superscript, and those in the same row the same subscript. The columns being arranged in order from left to right, and

the rows likewise in order from the top row downward, the position of any element of the array is shown at once by its
indices.

Thus, a"'

is

in

the third column and the

fifth

row

of the above array.

a n(M) diagonal passing through the elements a/, a%", that passing is called the principal diagonal of the array (M) the secondary diagonal. The posia n \ a n . '\ #, through
.

The

tion occupied
position.

by the element a/

is

designated as the leading

Art.

6.

Determinant as Function of

n*

Elements.

array just considered, inclosed between two vertical thus bars,

The

dl a;'
f

ai aj

a^ a^
a
{n)

a'
is

a"

ments called
as follows
:

used in analysis to represent a certain function of its n* eleThis function may be defined their determinant .*

Write down the product


diagonal, taking

of the elements on the principal


:

them

in

the natural order; thus

This product

is

called the principal term of the determinant.


in this principal

Now

undisturbed. To such of possible way, leaving the superscripts the n\ resulting terms as involve the even permutations of the to those involving the odd subscripts give the positive sign
;

permute the subscripts

term

in

every

This notation was

first

employed by Cauchy

in

1815.

See Dostor's

Theorie des determinants, Paris, 1877.

38

DETERMINANTS.

[Chap.

II.

permutations, the negative sign. The algebraic sum of all the terms thus formed is the determinant represented by the?

given array.

Art.

7.

Examples of Determinants.

Applying the process above explained to the array of four


elements gives
a,

a.

a:a m

a% a

pansion,
~ a
t
l

As an example of a determinant may be written


~ a
ft
x

of nine elements, with

its

ex-

~ a

tti
x

a a2 as
x

It is evident,

-f-

<z a

a^a^al"

a 'a,"a9
x

'"

- a^a^a;".

+a

a^

(2}

from the mode of

its

formation, that each term

of the

expansion of a determinant contains one,

and only

one,,

element from each column and each row of the array.


It

neous function of
with respect to minant. Thus,
respectively.

follows that every complete determinant is a homogeThe degree of this function, its elements.
its

elements,

is

called the order of the deter-

(1)

and

(2)

are of the second and third order

The
article
is

definition of

once more

a determinant given in the preceding^ illustrated by the following example of a


its

determinant of the fourth order with

complete development
a
x

a b
x

cx

a, b, c3

a% b%

c%

a, b A c t

d dt d d
x

+ a bj d + a bf/l
x

a b% cK d
x

b,c,dt -f

a b t c3 dz
x

afi^d,
afi c% d%
x

aJ> x c %

+ +
-f

aj> x c k d%

-f-

a 3 b c^di
x

ab
3

c 4 d^ -\- aj> c z
x

+ +
-f-

ajb % c

a%bfxd%

a% bj d
K

d a bcd aj>j d
aJb A c
A 9
x

aJb % c
%

d c d aJ> aJb c d
x

-\- ajb,1 c x
K

d d aJ) c d a 6,c d
t
A

(3)

Art.
It

8.]

NOTATIONS.

3D
the columns are ranked

will

be noticed that,

in this case,

alphabetically instead of by the numerical values of a series of


indices.

Art.

8.

Notations.
is

Besides the notations already employed, the following very extensively used
:

U
0*1

0,8
2

'

'

01.

0*

0, a m
This
is

a
;

called the double-subscript notation

the

first

subscript

indicating the rank of the row, the second that of the column. Thus the element # is in the second row and the third column.

The

letters are

sometimes omitted, the elements being thus

represented by the double subscripts alone.*

Instead of writing out the array in full, it is customary, when the elements are merely symbolic, to write only the prinThis is called cipal term and enclose it between vertical bars.
the umbral notation.
is

Thus, the determinant of the


a: a
in)
n

nth.

order

written

or,

using double subscripts,


|

#n

22

>

These

last

two forms
|

are sometimes

still

further abridged to
|

a
^oectively.
Prob.
nants:
5.

and
|

a hH

Write out the developments of the following determi-

w
though he

a, b

* Leibnitz indicated the elements of a determinant

in this

same manner,

made no use

of the array.

40

Art. 9]

SECOND AND THIRD ORDERS.


first

4L
in

Beneath the square array let the


order, as

two rows be repeated

shown

in the figure.

Now write down six terms, each


the product of the three elements lying along one of the
six oblique lines parallel to the diagonals of the original square.

Give to those terms whose

ele-

ments
tive

lie

on

lines parallel

to __,

the principal diagonal the posisign; to the others, the

*
-f-

negative sign. the required expansion.

The

result

is

Ap-

plying the method to the determinant just written gives a K c% a Atv a b. c iVs ajb^ aj.c, aj>x c% After a little practice the repetition of the first two rows will
I

be dispensed with. The above methods are especially useful in expanding determinants whose elements are not marked with indices, or
in evaluating

those having numerical elements.

No

such sim-

ple methods can be given for developing determinants of higher orders, but it will be shown later that these can always be

resolved into determinants of the third or second order.


Prob. n. Develop the following determinants: o n a h m

(i)

h b

f
c

g f
(4)

(s)

*,.r,

(7)

cos

(8)

cos

42

DETERMINANTS.

[Chap.

II.

Art.

10.
in

Interchange of Rows and Columns.


the development of the determinant a
|

Any

term

{H)
x
j

may-

be written
**'

a!

a .()
I, 2, 3,.
.

in

which hij

./is

some permutation

of the subscripts
.

.n+

/. Also, let Designate by u the number of inversions in hij v be the number of interchanges of two elements necessary to bring the given term into the form
.
.

a
in

(<?)

W
in

(0

which the subscripts are arranged


.

the natural order, while


(M)
',

pqr

/ is

a certain permutation of the superscripts

", '",

odd.

This permutation is even or odd according as v is even or But u and v are obviously of the same class that is,
;

both are even or both odd.


. . .

Hence the permutations


;

hij

and the term will have the and pqr / are of the same class same sign, whether the sign be determined by the class of the
permutation of the subscripts when the superscripts stand in the natural order, or by the class of the permutation of the
superscripts

when the order


that the

of the subscripts

is

natural.

It follows

same development

of the determinant

array will
Art.
6,

if, instead of proceeding as indicated in the superscripts of the principal term be permuted, the

be obtained

subscripts being left in the natural order, and the sign of each of the resulting terms written in accordance with the class of

the permutations of

its

superscripts.

Passing from one of these methods of development to the other amounts to the same thing as changing each column of
the array into a row of the same rank, and vice versa. Hence, a determinant is not altered by changing the columns into cor-

responding rows and the rows into corresponding columns.

Thus:

a"
aj a n
"
.

am
()

a n{n)

a^>

a,

ART.

11.]

TWO IDENTICAL PARALLEL


therefore,
is is

LINES.

43

Whatever theorem,
columns.

demonstrated with reference

to the rows of a determinant

also true with reference to the

The rows and columns


called lines.

of a determinant array are

alike

Art.
If

11.

Interchange of Two Parallel Lines.


parallel lines of a determinant be interchanged,

any two

the determinant will be changed only in sign.


For, interchanging any two parallel lines of a determinant array amounts to the same thing as interchanging, in every

term of the expansion, the indices which correspond to these lines. Since this changes the class of each permutation of the indices in question from odd to even or from even to odd, it changes the sign of each term of the expansion, and therefore
that of the whole determinant.
It follows

from the above that

if

any

line of a

determinant

be passed over ;// parallel lines to a new position in the array the new determinant will be equal to the original one multiplied

by

m
(

i)

to the leading position over the (k i) preceding rows, and by passing the /th row This the sth column over the (s i) preceding columns.
{s)

The element ak

may be brought

being done the determinant

is

multiplied

by
l)*+*,

(_
which changes
even.
its

,)*-

(_ i)~

(-

sign or not according as {k

s)

is

odd or

The when (k

{s) is called a positive position position occupied by a k a negative position when {k s) is odd. s) is even
;

Art.

12.

Two
in

Identical Parallel Lines.


parallel lines are identical

A determinant
is

which any two

equal to zero.

For the interchange of these two

parallel lines, while it

44

DETERMINANTS.
will in

[Chap.

II.

changes the sign of the determinant,


value.

no way

alter its

The

value then,
13.

if finite,

can only be zero.

Art.

Multiplying by a Factor.

Multiplying each element of a line of a determinant by a given factor multiplies the determinant by that factor.
Since each term of the development contains one and only one element from the line in question (Art. 7), then multiplying each element of this line by the given factor multiplies each term of the development, and therefore the whole deter-

minant, by the same factor.


It follows that,
if

contain a

common factor,
*.,

the elements of any line of a determinant this factor may be canceled and written
;

outside the array as a factor of the whole determinant

thus

# 31

m a. ma

2i

a,

a m ..ma ni

.**

which the elements of any line have a common ratio to the corresponding elements of any parallel For this common ratio may be written line is equal to zero. Its outside the array, which will then have two identical lines.
determinant
in

value

therefore zero (Art. 12). determinant having a line of zeros


is

is

equal to zero.

Art.

14.

Line of Polynomial Elements.

A determinant
the

sum
of

sum

having a line of elements each of which is two or more quantities can be expressed as the two or more determinants.
of

Let

{b

*.

+b bj + b:> {b>>

b:

,,,)
.

Cx

(1)

ct
c3

a3

{Bt

-b + b
'

...)
if

be such a determinant.

Then,
i

B =b
i

-bi^br.

Art.

15.]

composition of parallel lines.


of the determinant

45
is

any term of the expansion


ah BiCj
. . .

A
cj

ak
ah

hi c s

ip
.

ah
.

bl
;

b(' cj

. .

2)

the expansion of A are obtained by permuting the subscripts h, i,j\ ... of a h B{ But permuting at Cj the same time the subscripts of the terms in the second member of (2), and giving to each term thus obtained its
in
.

The terms

proper

sign, there results

J = \a B^...\
1

\aAc ...\-\a
3

b;c % ...\

+ \aA' c
,

...\...

which proves the theorem.

Art.
If

15.

Composition of Parallel Lines.


line of a

each element of a

determinant be multiplied by

ment

a given factor and the product added to the corresponding eleof any parallel line, the value of the determinant will not
;

be changed thus: a u a a lt
.

av
a>

*n *
#

On
2

+ ***)
+ ma
ni )

x
;

#*, '!

#, #, "-H2 "M3

0*,

(a M%

a nn

appear upon resolving the second member into two determinants (Art. 14), one of which will be the given deThis
will

terminant, while the other, upon removal of the given factor,


will vanish

because of having two identical

lines.

In like

manner any number

of parallel lines

may

be com-

in any way the elements to which added multiples of corresponding elements from other For example, a l>n is equivalent to parallel lines.

bined without changing being taken not to modify

the value of

the determinant, care

are

*n

(^11

..

- ma * +'")

*!*

a xn

-m(a

i?

+ Xa +
15 )

.)

46

DETERMINANTS.

[Chap.

II.

Art.

16.

Binomial Factors.
is is

A
and

determinant which
b,

of

such that

if

a rational integral function of a substituted for a the determinant


factor.

vanishes, contains {a

^)asa
a*

For example,
a-\-r
b

Jee

p*
1

9-pp
is divisible

b-q
q

+r
r

by

{a

b).

To prove this, let the expansion of any such determinant be written in the form

A
the coefficients
,

= m,
iy

-j-

m & + wi-P? +
x .

m m

tn

being independent of

a.

Now
Hence,

when

is

substituted for a the determinant vanishes.

o = m + mjb
%

-\-

mj?

Subtracting this from the precedfng gives

= m {a - b) + m,(a* t

F) +...
is

This being divisible by (a


Prob.
13.

b),

the theorem

proven.
:

Prove the following without expansion

(0

x
o

X y
o

(2)

my mnz nz
(3)

+
b
c c

a
a
b

+
c

a -\-b a

(4 )

*_t a
c

+a

Art.

17.]

co-factors

minors.

47

Prob. 14. Prove that

x 1 x ix 3
1
1

a y a y

b
x

b
b

a y^

43

DETERMINANTS.

[Chap.

II.

row of the array (Art. 7). If, therefore, in selecting the elements for any term, any other element than a/ be taken from the first column, the one taken from the first row must be zero. Hence, the only terms which do not vanish are those which
contain the element a/.

Moreover,

in the

not vanish, a/ is multiplied by (n from each column and each row of

terms of the expansion of (1) which do 1) elements chosen one

an

.a
()

(2)

There are
written
class of the

{n

1)!

,V/"

such terms, any one of which may be a ^ n) tne s,*S n being determined by the
1, t f

permutation of the n subscripts

j\

I.

But

since this is of the same class as the permutation of the (n 1} a t (n \ a^al'a-" subscripts i,j\ .'../,- the sign of any term,
.

of the expansion of (1)

is

sponding term,
ax
'

a" a/"
a2

the same as the sign of the corren) of the expansion of (2). ,a} Hence,.
\

(3)
.

a 7 #a

Art.

18.]

development

in

terms of co-factors.
is

49

The

co-factor thus obtained

represented by the symbol

the sign-factor of which,


the symbol
itself,

(is

1)*+*, is

intrinsic,

i.e.,

included in

which

The

co-factors
:

of

accordingly written as positive. the various elements of \a u a n a are as Si


\

follows

An =

50 group.

DETERMINANTS.
In
all

[Chap.

II.

of them, then, there are

nX{n
is

i)! or

;/!

dif-

ferent terms of the determinant, which

the whole number.

Hence,
|

a^ = aiAi + a^'A,"
|

+ aj*Ai*\
.

(i)

Similarly (Art. io),


|

a
|

==

0VA

+ **AP +
Ai
n
\

+ aPAt
(i) or (2),

(2)

Any

determinant may, by means of either

be

re-

solved into determinants of an order one lower.


these formulas

Since, in

A k\

or

A^,

A^

s)

are themselves

determinants, they may be resolved into determinants of an order still one lower in the same manner. By continuing the
process any determinant may ultimately be expressed of determinants of the third or second order, which
easily
in

terms

may be

expanded by methods already given (Art.


let
.
I

9).

A=

For example,
I

a b%
t

c3

4
4

be required to develop the determinant 1, gives Applying formula (1), letting k


it

A =

a,

c*

Upon

a second application of the

same formula

this

becomes

Art.

19.]

the zero formulas.


19.
i

51

Prob.

Develop the following determinants:

to

52
Similarly,

DETERMINANTS.

[CHAP. II.

p and
fl,^

being different superscripts,

+ aiA +
Cauchy's

+ *ia.CP =

o.

Art.
It is

20.

Method of Development.

frequently desirable to expand a determinant with reference to the elements of a given row and column.
n) and the given row Let the determinant be A a} and column the ht\\ and pth respectively. Then is A h(p) the
| |

co-factor of a h(p\ the element at the intersection of

the two
p)

given

lines.

The

co-factor of
{s

any element a

s)

of

will

be

designated by
(n
2).

\ this being

a determinant of the order

The

required expansion

may now be
:

obtained by

means

of the following formula,


|

due to Cauchy

mf*
2,
.
.

= mJ*AP- 2aPaiBP.
.

(i)

in

which k
.

I,

I,

-f- I,
.

n,

and

I, 2, ..

./

i,

p -\- I, .. n, successively. To prove this, consider


terms
a
p)

that
of

{s)

is

the

of
.

the

expansion

which contain
{p)
.

aggregate of all the product


.

These terms are included in ah A Now, every term in the expansion which does not contain ah must contain some other element ah from the /ith. row and also some other from the pth column, and thus contains the prodelement a p) uct ali ai But this product differs from ah ak only in the order of the superscripts and is, therefore, in the expansion of
ak
{s)
{f>)

{s)

p)

s)

{p)

{s)

A, multiplied by an aggregate of terms differing in sign only Bk is the coefficient from that multiplying a h ip) a k Hence, in the required expansion. of ah a k
(s)
i$)

(s)

{p)

In the formula ah {p)


are also {n

[p)

gives (n

i)

terms of A.
{s)
{s)

There

taining (n
(n
i)
I

-\-

ah a k(p) Bk each coni)* such aggregates as terms. The formula therefore gives 2) a n terms, which is the complete (n i) {n 2)
,
\
!

expansion.

When

the expansion

is

required with reference to the ele-

Art.

20.]

cauchy's method of development.

53

ments

of the first

explicitly,
\a W|

column and the becomes


tf,
x

first

row the formula, written

= a/A/- afa/'B," - a^a B m - ... - */*,, - a;a;'B:' - a;<x;"B:" - ... - a 'a^B^


%
3

- an'a "BH" - aJa^'BJ" - ... - a;ay>B*\


x

(2)

in

which

(s)

has intrinsically the sign

i)*

+s .

Cauchy's formula is particularly useful in expanding determinants which have been bordered such as
;

-Q
u a \\ a \i a \%
i

(3)

U % a%x a
u % a 3l a
3,

ait
a S3

Applying formula

(2) to this

determinant gives

-Q =

"

^22 ^23

54
(4)

DETERMINANTS.

Art.

22.]

RAISING THE ORDER.

55
written

Prob. 25.

Show

that Cauchy's formula

may be

J==la. w

6A ah W daW
22.

Sa^a^

d*J

da^da^

Art.
' the elements at

Raising the Order.

Since, in the expansion of the determinant (1) of Art. 17 aj do not appear, these may be replaced
t
.
.

by any quantities whatever, as Q, value of the determinant thus


; :

T, without

changing the

a/ o a % a,

o
a,
,

o
()

a/ o

au'aj'aj"...a
Similarly,

Ta" a"
()

...a

O
o

G*," o
*,'*,"*.'" ...*,<>
%

...

R La '"...a

a' a" a'" un un un


in

a n(M) u
. .

TNa " ...a


f

which Q, R,

and Z,

N are any quantities whatever.


)

Finally,

^aJ'...aH ?r <*i

n) -

'

a,

o
t

o o

Qa
a' n-i <* h-i a" u

"...o

ari-i u (M-I) o u

T N...C a^
that

the elements on one side of the principal diagonal are zeros the determinant is equal to its principal term, and
is, if all

the elements on the other side of this diagonal

may be

replaced

by any quantities whatever.

By what

precedes,

.
x

in,
t

an

a n ()

Ta'..a, ()

56

DETERMINANTS.

[Chap.

II.

Hence, a determinant of the nth order

may be

expressed

as a determinant of the order (n -\- i) by bordering it above by a row (to the left by a column) of zeros, to the left by a

column (above by a row)


writing
I

of elements chosen

arbitrarily,

and
con-

at the intersection of the lines thus added.

By

tinuing this process any determinant

may

be expressed as a

determinant of any higher order.


Prob. 26. If
all

onal are zeros, what

the elements on one side of the secondary diagis the value of the determinant ?

Prob. 27. Develop the determinant

a h
h b

g u

o
o
t

f
c

v o

g f
U V
o
{s)

w
o
t

w
o

s
{k)

a s and au {k) o is Prob. 28. A determinant in which a k Prove that every skew-symmetric detersaid to be skew-symmetric. minant of odd order is equal to zero.
r

Art.

23.

Solution of Linear Equations.

Of the many analytical processes giving rise to determinants the simplest and most common is the solution of systems of
simultaneous linear equations. a ;x>

Thus, solving the equations

+ a;>x" = kJ
:

by the methods

of ordinary algebra gives

a; a,"

- a^r

"
*/*,"

-*

Art.

23.]

SOLUTION OF LINEAR EQUATIONS.

57

numerator of the fraction giving the value of x' is formed from this denominator by replacing each coefficient of x' by the corresponding absolute term. Similarly for x"
.

The

difficulty of solving

systems of linear equations by the

ordinary processes of elimination increases rapidly as the number of equations is increased. The law of formation of the
roots explained above
is.

however, capable of generalization,


all

being equally applicable to

complete linear systems, as

will

now be shown.
Let such a system be written

a> x -f a ,"x"
f

<*?
a n 'x'

" <' x

+ +

4-

*w*w =
n)

*,

+<
.

{n)

">

>

(I)
.

+ aj'x" +
D=

+ a^x =
()
*- ()

Kn

Now
equations

form the determinant of


;

the coefficients of

these

thus

a a
x

()
x

()

an

a,
s)

M
a,}

and

let

A^

be the co-factor of a

in

this

determinant.

The

function
i

+ a w AJ +
(Art. 18)
.

+ a,}* A ^
n

is

equal to

D when/ = s
Af\ Af\
.

to zero

when p and

are

different superscripts (Art. 19).

Then, multiplying the given


]

equations by
of

respectively, the

sum

of the

in which resulting equations is a linear equation x is equal to Z>, while those of all the other

the coefficient

unknowns

vanish.

The sum
jR*

is,

therefore,

K Af*
M

+ *AW +
a*
,
.

+K A
is

S)
'

2)

But the second member

of this equation

what

D becomes

& by

upon

replacing the coefficients

the absolute terms

k9

a?\ .-.:.# of the unknown in order. ** Hence, ,

58

tf

;b
|A<6. a.
-

Art.

24.]

CONSISTENCE OF LINEAR SYSTEMS.

59

obviously depend upon some relation among the coefficients. This relation will now be investigated for the case of [n i) linear equations involving n unknowns. Let the equations be

+ ... + <*

-v()

-_

K.

*v + ... +
If

=/cK

the above equations are consistent the values of the obtained by solving any n of them must satisfy the
equation.

unknowns

Solving the

preceding article, obtained in the last equation, and clearing of fractions, the result reduces to (Art. 18)
y
. .

remaining n equations by the method of the xP* thus substituting the values of x', x"
first
.

E~

a/ an
a n+i
.

()

= o,

(*)

+i

^+i

which

is

the condition to be

fulfilled

by the

coefficients in order

that the given equations

may

be consistent.
linear equa-

Hence the condition of consistency for a set of tions involving a number of unknowns one less than
of equations
is

the

number

that the determinant of the coefficients and

absolute terms, written in the same order as in the given equaThis determinant is called the resultant* tions, shall be zero.
or eliminant of the equations.

Thus the equations


x-{-y-{-z

x-\-y

= o, x

y-{-z

= 2,

= 4, x-\-y-\-z = 6

are consistent, for the reason that


I

o.

60

DETERMINANTS.

[Chap.

II.

Art.

25.

The Matrix.

Assume
greater than

r linear equations involving n unknowns, r being n. as follows


:

al'x -p...
an 'x'
ajx'

+ af>j* =

*,, 1

+ .:. + aM*P> = K +
. . .

nt

+ a^x = K

r.

The

minant

consistency of these equations requires that every deterof the order (n+ i), formed by selecting (n i) rows

from the array whose elements are the coefficients and absolute terms written in order, shall be zero.
the elements of the array expressed thus
If
:

fulfill

this condition the fact is

a/ ...a:

o;

()
x

<)
*
. .
.

a r() u
.

K,

/C.

Kr

the change of rows into columns being purely arbitrary. The above expression is called a rectangular array, or a matrix.

Art.

26.

Homogeneous Linear Systems.


of the given

Let the equations

system be both linear and

homogeneous

thus

al x> 4.

.-,

. ;

.J.

a**&*

= o, ]
o.
J

au'

x'

+ a^x^ =
= o/E.

Representing the determinant of the coefficients by E, the general solution, as given by the formula (3) of Art. 23, is

x^
That
is, all

tions have

the unknowns are equal to zero, and the equano other solution than this unless

E = o.

Art.

26.]

homogeneous linear systems.


in this

61

But
only
in

the

knowns may
becomes

unknown is obtainable indeterminate form o/o. The ratios of the unbe readily obtained, however. For, dividing each
{s

case the value of each

equation through by any one of these, as x

the system (i)

X*

JT^

-1 *

jw-r

-t^i

xw

-r a

M
{s+ J >

2>

r< w >

xis)

-t-

-t-^

- - ^.

Now

the condition
(2)

E = o establishes the
1)

consistency of the
ratios (Art. 24),

equations

involving the {n
x>_

unknown

&>

'" x^
*P

'

x^
x<*>

'

" x

'*'
;

o the given equations (1) are consistent that is, Hence, if ii the values of the above {n 1) ratios obtained by solving any of them will satisfy the remaining equation. Any n (;/ 1)

among themselves the ratios thus determined " x are will satisfy the given equations. Thus, if x x " Xx \ A being any n such quantities, so also are \x A x
quantities having
{n)
',
Q
y .

(u

factor whatever.

The determinant
geneous
the system. When the
of

of the coefficients of the given


is

homo-

linear equations

called the resultant or eliminant of

unknowns

of equations is greater than the number the conditions of consistency are expressible in

number

the form of a rectangular array, as in Art. 25. As an example, consider the five equations

2x

2>y

+2= x +y

>

4*
z

y
first

~~
>

7X

+W+*~
o.

= o,

$x

Sy

+z=

Dividing each of the


for the

two equations by z and solving

x y two unknowns - and gives

_3
5

62

DETERMINANTS.

[Chap.

II.

or

x\y\z\\2

:3

5;

and any three quantities having these ratios will satisfy the two equations, as 10, 15, and 25. That the third equation is consistent with the first two is shown by the vanishing of the
determinant

2-3 1=0. 4- -7 3
1

I
1

the equations are consistent the determinant of the coefficients of any three of them must vanish that is,
If all
;

4-

= o.

Art.

28.]

SYLVESTER

METHOD OF
is

EI

WINATION.

63

Hence, when a determinant


of the elements of

equal to zero, the co-factors

any

line are proportional to the co-factors of


line.

the corresponding elements of any parallel

Art.
Let
it

28.

Sylvester's

Method of Elimination.*

be required to eliminate the unknown from the two

equations

a% x%

+ ay

-f-

ax
x

+ a, =
-\-

O,

by -fThis
will

= o.

be done by what is called the dialytic method, the invention of which is due to Sylvester. Multiplying the first of the given equations by x, and the second by x and x* successively, the result
is

a system of five equations,


%

viz.:

ax
%

a 3 x*

-\-

ay a x a ay + ay + a x by -f b x + b bx b x + by
-\-fx

-f-

-f-

by + b x" + by
x

= o, = o, = o, = o, =a

The

eliminant of
x, x*,

these five equations,


x*
is

involving the

four

unknowns

x\ and

(Art. 24)

64

DETERMINANTS.
first

[Chap. IL

obtained which involve as unknowns the

(m

-\-

i)

powers of

x.

The

of the order {ni

eliminant of these equations is a determinant -\- n), which is of the nth. degree in terms of

the coefficients of the equation of the mt\\ degree, and vice The law of formation of the eliminant is obvious. versa.

The same method may be used


the variables from a pair of

in

eliminating one or both

homogeneous equations.

an example, from the equations


2x*

As

let

it

be required to eliminate the variables

$*y

9J
first

=o
x

and

3^

jxy

6y x
;

= o.
the second

Dividing the

by y\ and multiplying by

by y*, and multiplying by


in all, five

twice in succession, there result,

X X
equations involving
,

X
,

-r,

and

Eliminating

these four ratios gives

E~

25 09
o

o
o o

3-76
o

3-7-6

,3-7-6
consistent.

the vanishing of which shows that the two given equations are

Prob. 31. Test the consistency of each of the following systems


of equations:
(1)
(2)

x+y+2Z=g, x+y
x y
2Z

z=o, 2xy-\-z=$, *
z

3)'+ 22=1;

=0,
o,

(3)

- xf = ix*y
2x
-{-y

= o, 2X 2y + - 5/ = o. Sx*y + Sx/
y
42

$y

o;

Prob. 32. Find the ratios of the unknowns in the equations


2z

= o,

4W

= o,
+1)

2iv

+x
[i

5V

+ * = o..n]

Prob. 33. In the equations


ak'x'
-f-

+<**<*>*<">

+a
:

(n

+W
::

5s o,
:

prove

that

*':...: x

(n)

x< n+1)

M'

M M M<n+
:

where

Art.

29.]

THE MULTIPLICATION THEOREM.


is

66

i)'

*M

the determinant obtained by deleting the *th

column

from the rectangular array

M~

a/ ...
a n '... a

a<*>

a^

M+1)

a***
r

Prob. 34.

From **+*?
p

+ *! = "* + ?+* = Mx+Xy + nz


q

66

DETERMINANTS.
result of

[Chap.

II.

The

such transformation

is
%
>

(.Ai (*. Ai

+ iAtK + fa-Ai + *M*% = + **h)*i + (M + .A.K + iA. Ai + <*.At


Now

o-

(3)

The vanishing

of the determinant

*A,
tf
a

+ ** *n*n + ad>
x*
;

(4)

is
is,

the condition that the equations (3) may be consistent that the condition that they may have solutions other than

the equations (3) may be consistent because of the consistency of the equations (1), in which
,

ux

=O=

(Art. 26).

case the determinant

a%% an
vanishes.

Or, this condition failing, and the


l

thus having no solution other than x =. o (3) will still be consistent if the equations if the determinant

= x%y

equations (1) the equations


so
;

(2) are

that

is,

KK
vanishes.
(6),

<

6>

vanishing of either of the determinants (5) or It follows therefore, causes the determinant (4) -to vanish.
(5)

The
(6)

that

and

are factors of (4)


(4)
, .

and since

their product

and

the determinant

are of
. ,

the coefficients an

blx ,

the same degree with respect to . . r, they are the only factors.
.

Hence,
a x\ a xi

Art.

29.]

THE MULTIPLICATION THEOREM.


process indicated by this formula *
:

67

The
follows

may be

described as

the determinant \p ltH \> which is the product of determinants \a 1>n and \& un \, first connect by plus signs two the elements in the rows of both \a 1>n and b 1<n Then place the first row of \a un upon each row of \b 1>n in turn and let
\

To form

each two elements as they touch become products. This is the first row of \pu \. Perform the same operation upon \b hn
\

with the second row of \a hH to obtain the second row of \p 1>n and again with the third row of |tf 1)M |to obtain the third row
\
\

f|A.|; etc

Any

element of
Pks

this

product

is

*Ai

+ a**K +

+ **A..

(9)

the two determinants to be multiplied together are of different orders the one of lower order should be expressed as a determinant of the same order as the other (Art. 22), after

When

which the above

rule
of

is

applicable.

The product

two determinants may be formed by

columns, instead of by rows as above. In this case the result Thus the product of the deis obtained in a different form. terminants (5) and (6) by columns is
a iA%
Prob. 41.
(1)

+ a*A*

a iA*

+ a*&*
:

Form

the following products

(3)

Art. Prob. 42. Generalize the last example (see Prob. 22,

18).

Prob. 43.

By

forming the product

68

DETERMINANTS.
of

[Chap. Ik

show that the product


squares,
is itself

two numbers, each the sum of four

the

sum

of four squares.

Art.

30.

Product of Two Arrays.

process explained in the preceding article may be apwhat is conventionally termed the product of two rectangular arrays. It will appear, however, that multiplied to form

The

plying two such arrays together by columns leads to a result radically different from that obtained when the product is

formed by rows. Let the two rectangular arrays be

and

M.A.
,AAs
is

The product
=

of these

by columns
a
x

*.Ai

+ aj^

J>x >

+ aj>

%l

aj u
^.3^13

+ aj>

%l

*iAi

+ *t'Ai
A
is

#iAs

iAi

&*J>*t

The determinant
uct of
of

plainly equal to zero, being the prod-

two determinants formed by adding a row of zeros to one the given rectangular arrays and a row of elements chosen
In general, the product by columns of two rectangular

arbitrarily to the other.

arrays having

rows and

ft

determinant of the n th order

columns, m being less than whose value is zero.

is

Multiplying together the above rectangular arrays by


the result
is

rows,,

J'

lAl

+ *lAi + *lA.

*xxK

At

iAi

a^a x%
In the same

+
manner
it

*,A.
&*A*

a,, a.

KA
bj>.

may be shown that the product by rows of two rectangular arrays having m rows and n columns, m being less than n, is a determinant of the m th order, which may be expressed as the sum of the n \/m {n m) determinants
!

Art.

31. J

RECIPROCAL DETERMINANTS.

69

of the arrays by deleting {it m) columns, each multiplied by the determinant formed by deleting the same columns from the other array.

formed from one

Art.

31.

Reciprocal Determinants.
replacing each element of a

The determinant formed by

given determinant by its co-factor is called the reciprocal of the given determinant.* Thus, the reciprocal of
a,, a.
IS

AA
**

.A.
.A,

a,a

av

A 21-"- 22 A
nl
.

dnxdn-i.

(1,

A Am
two determinants
is

The product
6.4=

of these

(tn\Au-\-.

.-\~a nn

Am

a n iA<n-\-.

.-\-a nn A 2n

a m A n \-\--

-\-a nn

A nn
is

Each element on the


equal to S (Art.
19).
18),

principal diagonal of this product


all

while

the other elements vanish (Art.


or

Hence,
6.

= So...
o d
. .
.

o (w)

<T,

6'

oH o

<?

That
equal to

is,

the reciprocal of a determinant of the n ih order


th

is

its (n

i)

power.

* The term reciprocal as here used has reference to the algebraic transformation concerned in the passage from point coordinates to line coordinates, called
reciprocation.

The

reciprocal of a determinant

is

also called the determinant

adjugate.

70

PROJECTIVE GEOMETRY.

[CHAP. Ilk

Chapter

III.

PROJECTIVE GEOMETRY.
By George Bruce Halsted,
Professor of Mathematics in the University of Texas.

Art.
i.
'

1.

The Elements and Primal Forms*


determined

A
'

line

by two points on

it

is

called

straight.'
2.

On any two
join.'

points can be put one, but only one, straight,

their
3.
it is

A surface
'

determined by three non-costraight points on

called a

plane.'

lie all on one and only one plane, their 'junction.' on a plane, so does their join. 5. If two points lie 6. The plane, the straight, and the point are the elements

4.

Any

three points, not costraight,

in projective
7.

straight
It

points.

geometry. is not to be considered as an aggregate of is a monad, an atom, a simple positional concept as


'

primal as the point. It is the 'bearer is the bearer of any planes on it.

of

any points on

it.

It

is

8. Just so the plane is an element coeval with the point. It the bearer of any points on it, or any straights on it.

9.

A point A
'

is

the bearer of any straight on


is

it

or any plane
called

on

it.

10.

point which

on each of two straights

is

their

cross.'

* This Chapter treats Projective Geometry entirely by the synthetic method. Metric relations are not considered, and nothing is borrowed from Analytic

Geometry.

ART.

1.]

THE ELEMENTS AND PRIMAL FORMS.


Planes
all

71
all

11.

on the same point, or straights


'

with the

same
their

cross, are called

copunctal.'
lie

12.
'

Any two
meet.'

planes

both on one and only one straight,


join, planes

13.

Like points with the same


called costraight.

with the same

meet are
14.

plane and a straight not on


their
'

it

have one and only one

point in
15.

common,

pass.'

Any

three planes not costraight are copunctal on one


'

and only one point, their apex.' 16. While these elements, namely, the plane, the straight, and the point, retain their atomic character, they can be united into compound figures, of which the primal class consists of
three forms, the
17.
*
'

range,' the

'

flat-pencil,'

the

'

axial-pencil.'

The aggregate
'

of all points

If point-row,' or range.' intersection.' it is called their


'

on a straight is called a a point be common to two ranges,


is

18.
'

piece of a range bounded by two points

called a

sect.'

19.

The aggregate

of all coplanar, copunctal straights

is

called a 'flat-pencil.'
point.'
20.

The common
plane
'

cross
4

is

called the

'

pencil-

The common

is

called the

pencil-plane.'

piece of a flat-pencil
called an
angle.'

bounded by two

of the straights,

as

sides,' is

21.
'

The aggregate

of all planes

on a straight

is

called an
*

axial-pencil/ or 'axial.'

Their

common

meet, the
of

axis,' is

their bearer.
22.

sides, is

piece of the axial bounded called an axial angle.'


'

by two

its

planes, as

23.
24.
of,
it is

Angles are always pieces of the

figure, not rotations.

No

use

is

made

of motion.

If

moving point
its

is

spoken

to be interpreted as the

mind
of

shifting

attention.

25.

When

there can be no ambiguity of meaning, a figure

in a pencil,

though consisting only

some

single elements of
so,

a pencil. Just the complete pencil, may yet itself be called may be called a range. certain costraight points

separate

72

PROJECTIVE GEOMETRY.

[CHAP.

III.

Art.
26'.

2.
'

Projecting and Cutting.

project from a fixed point (the projectionvertex ') a figure, the original,' composed of points B, C, D, etc., and straights b, c, d, etc., is to construct the projecting
'
' ' '
'

To

straights

MB, MC, MD, and


is

the

'

'

projecting planes

Mb, Mc,
and

Md.

Thus

obtained a

new

figure
'

composed
'

of straights

planes, all
27.

on M, and called an
'

eject
ju

of the original.
'

To

cut
'

'

by a

fixed

plane
of

(the

picture-plane
/?,

')

figure, the

subject,'

made up
Thus
is is

planes

y,
/*/?,

6,

etc.,
jud,

and

straights

b, c, d, etc., is

to construct the meets

py,

and
'

the passes

jub, p.c, jud.

obtained a new figure com-

posed

of straights

and points,
' '

all

on

}x,

and called a

'

cut

of

the subject. If of the subject is


28.
'

an eject of an original, the cut the subject an image of the original.

(the

Axial projection. To project from a fixed straight m projection-axis '), an original composed of points B, C, D,
construct the projecting planes

etc., is to
is

mB, mC, mD.

Thus

obtained a
called an

new
'

figure

composed
'

of planes all

on the axis m,

and

axial-eject

of the original.

29.

To

cut

by a

fixed
/3,

straight
etc.,

m
is

(to 'transfix

')

a subject

composed
mft,

of planes

y, #,

to construct the passes


is

my, md.
'

The
'

cut obtained

by transfixion

a range on
'
'

the

transversal

m.
fixed

30.
(7\)

primal figures are called projective when one can be derived from the other by any number of

Any two

projectings and cuttings.*

Art.
31.
It is

3.

Elements at
for every
is

Infinity.
in

assumed that

element

either of the
in

three primal figures there


others.

always an element

each of the

*Pascal (1625-62) and Desargues (1 593-1662) seem to have been the first to derive properties of conies from the properties of the circle by considering the fact that these curves lie in perspective on the surface of the cone.

Art.

3.]

elements at infinity.

73
'

32.

On

each straight

is

one and only one point


others are
*
'

at infinity,'

or

'

figurative' point.

The

proper
'

points.

Any

point going either way (moving in either sense ') ever forward on a straight is at the same time approaching and receding

from

its

point at infinity.
its

The

straight

is

thus a closed line

compendent through
'

point at infinity.

33.
34.

Parallels

'

are straights on a
in
it

common

point at infinity.

Two
sect

proper points

divide a range into a finite sect

and a

through the
it

proper point in
('rays').

infinite. Its figurative point and a divide a range into two sects to the infinite

35. All the straights parallel to each other on a plane are on the same point at infinity, and so form a flat-pencil whose penSuch a pencil is called a parallel-flatcil-point is figurative.
'

pencil.'
36. All points at infinity

on a plane
Its

lie

on one straight at

infinity

or

figurative

straight.*
is

straight on the plane


straight.

with any proper the point at infinity on the proper


cross

straight in

same plane have all a the straight at infinity on which common, namely, are the figurative pencil-points of all these pencils.
37.

Parallel-flat-pencils

on the

38.

Two

planes whose meet

is

a straight at infinity are

called parallel.
39. All the planes parallel to

each other are on the same


axial pencil

figurative straight,
infinity.

and so form an
axial
is

whose axis

is

at

Such an

called a parallel-axial.

40. All points at infinity

and

all

straights at infinity lie

on
is

a plane at infinity or figurative plane.

This plane at infinity

common
Prob.

to
1.

all parallel-axials,

since

it is

on the axis of each.

From each

of the three primal figures generate the other

two by projecting and

cutting.

infinity.

* This statement should not be interpreted as descriptive of the nature of In the Function Theory it is expedient to consider all points in a

plane at infinity as coincident.

74

PROJECTIVE GEOMETRY.

[CHAP.

III.

Art.
41.

4.

Correlation and Duality.


*

when every element with one and only one element of the other. paired Correlation is a one-to-one correspondence of elements. The paired elements are called 'mates.'
figures are called

Two
is

correlated

'

of each

42.

Two

figures correlated to a third are correlated to each


just

other.

For each element of the third has

one mate

in

each of the others, and these two are thus so paired as to be themselves mates.
a plane, any theorem of configuration and determination, with its proof, gives also a like theorem with its
43.

On

proof,

by simply interchanging point with

straight, join

with

cross, sect with angle.*

This correlation of points with straights on a plane

is

termed a
Prob.

'

principle of duality.'
is

Each
'

of

two
9

figures or theo-

rems so related

called the

'

dual

of the other, f

2. When two coplanar ranges ///, and m are correlated as 9 cuts of a flat-pencil M, show that the figurative point of v or Q 9 or of the the one is mated, in general, to a proper point

>

other.

Prob.
1'.

3.

Give the duals of the following:


'

2'.

3,.

Two coplanar straights determine a flat-pencil on their cross. Two coplanar flat-pencils determine a straight, their concur/ Two points bound two explemental sects.
'
'

Prob. 4. To draw a straight crossing three given straights, join the passes of two with a plane on the third.

Art.
44,.

5.
'

Polystims and Polygrams.


is

'

polystim

a system of
'),

44'.

'

'

polygram

is

a system
(*

n coplanar points
all
('

('

dots

with

of n coplanar straights

sides

'),

ranges connectors ').

the

they determine

Assume

that

no

the flat-pencils they determine ('fans'). Assume that


all

with

three dots are costraight.

no three

sides are copunctal.

* Culmann's Graphic Statics (Zurich, 1864) made extensive use of duality. Reye's Geometrie der Lage (Hannover, 1866) was issued as a consequence of the
Graphic Statics of Culmann.
f In Analytic

of the

Geometry the principle of duality consists in the interpretation same equation in different kinds of coordinates point and line or point

and plane coordinates.

Art.

5.]

POLYSTIMS AND POLYGRAMS.


i)

75

In each dot intersect (n

In each side
fans, going

concur (n

1)

connectors, going through the reSo there i) dots. maining (n


are n(n
45,.

through the remain-

ing (n

1) sides.

So there are

i)/2 connectors.

n(n
3,

i)/2 fans.

For n greater than


will

the
in

connectors

intersect

45'. For n greater than 3, the fans will concur in straights other

points other than the dots. Such intersections are called 'codots.'
46,.

than the
are called
46'.

sides.
'

Such concurs

diagonals.'

There are
i)(n

There are
i){n

n(n

2){n

3)/8 codots.

n(n
nals.

2){n

3)/8 diago-

Proof of
connectors.
[n(n

46,.

In a polystim of n dots there are n(n


intersect in

i)/2

These connectors
i)/2

i)/2][(
;

i]/2

= n(n

i)(;z

- n-

2)/8
i)/2

points
things,

i.e.,

the

number

of different combinations of n(n

two at a time. But some of these

ones are codots.


each dot.

Now

intersections are dots, and the remaining (n 1) of these connectors meet at


is

Therefore each dot

repeated (n

1) in

2)/2

times; or the

number

of times the connectors intersect in


i)(n
is

points not codots, i.e. in dots, is n{n Therefore the number of codots
2

2)/2.

n(ti

i)(;/

n
a

2)/8

n{n

i)(

2)/2

= =
47,.

[n(n
(/*

i)/8][^
\){n

n2-4(n- 2)]
3)/8.
47'.

2){n

A set of

n connectors may-

set of

;/

fans

may be

be selected in several ways so that two and only two contain each one of the n dots. Such a set of connectors is called a
*
'

selected in several ways so that

two and only two contain each one of the 71 sides. Such a set
of fans of fans.
48'.
is

called a

'

complete

set

complete
48,.

set

of connectors. are
(n
1) 1/2

There

There

are

1)

\/2

complete

sets of connectors.

complete

sets of fans.

Proof of
single

48,.
(

dot

In a polystim of n dots there are through any and hence (0 i){n 2)/2 1) connectors,

pairs of connectors.

Consider one such

pair, as

BC and

BE.

76

PROJECTIVE GEOMETRY.

[CHAP.

III.

The number
from
{n

of different sets (each of n


etc. [there

2 connectors)

to

E
is

through A, D, F, G>
(n
3)
!,

being {n

3)

such dots],

i.e.

3) things.

Hence

the

number of permutations of number of complete sets of conthe


is

nectors having the pair

BC and BE

[n

3)

Therefore the

whole number of complete


(n
49,.

sets of connectors
(

is

I)(

_ 2)[( _ 3) !]/ 2 _
49'.

_
is

j) !/2 .

In any complete set of connectors, when n is even, the


first

In

any complete

set

of

fans, when n

and
' '

the

(/2 + i)th
'.
'

are

even, the first and ' the (/2+i)th are called opposite.'

called
x
.

opposite

5o A tetrastim is a system of four dots with their six connectors.

tetragram is a system of four straights with their six


50'.

'

'

Each

pair of opposite

fans.

Each
in

connectors intersect in a codot.

concur
three
'

These three
the
'

codots
'

determine

pair of opposite fans a diagonal. These determine the diagonals


'

codot-tristim

of the tetra-

diagonal-trigram

of the tetra-

stim.

gram.

correlated polystims whose paired dots and co51. dots have their joins copunctal are called 'copolar.'
52.

Two

Two
and

tersect
4

correlated polystims whose paired connectors inhave their intersections costraight are called

coaxal.'
53. If

two non-coplanar

tristims be copolar, they are coaxal.

For

since

AA' crosses BB\


two non-coplanar

therefore

AB and A' B'

intersect

on

the meet of the planes of the tristims.


54. If

tristims be coaxal, they are copolar.


f
r
,

For

The

these four points are coplanar. three planes ABA'B', ACA'C\ BCB'C are copunctal. Hence so are their meets AA\ BB\ CO,
intersects
55.

since

AB

AB

taking the angle between the planes evanescent, is seen that coplanar coaxal tristims are copolar and then by reductio ad absurdum that coplanar copolar tristims are coaxal.

By

56.

If

are said to be

two coplanar polystims are copolar and coaxal they complete plane perspectives.' Their pole and
*

Art.

6.]
'

harmonic elements.
center of perspective' and the
'

77
axis of

axis are called the


perspective.'
57.
If

two coplanar

tristims are copolar or coaxal, they are

complete plane perspectives.


If two coplanar polystims are images of the same polyfrom different projection vertices Vlt Vif they are comstim For the joins of pairs of correlated plete plane perspectives.

58.

points are all copunctal (on the pass of the straight V F2 with the picture plane), and the intersections of paired coni

nectors are

all

costraight (on the meet of the picture plane


original).

and the plane of the


Frob.

5. In a hexastim there are 15 connectors and 45 codots. In a hexagram there are 15 fans and 45 diagonals. Prob. 6. If the vertices of three coplanar angles are costraight,

make three tetragrams whose other diagonals are copuncby threes four times. [Prove and give dual.] Prob. 7. The corresponding sides of any two funiculars of a given system of forces cross on a straight parallel to the join of the poles of the two funiculars.
their sides
tal

Art.
59.
lie

6.

Harmonic Elements.
If

Fundamental Theorem.

two correlated tetrastims

on different planes whose meet is on no one of the eight dots, and if five connectors of the one intersect their mates,,
then the tetrastims are coaxal.
fixed

For the two

pairs of tristims

by the
If

five pairs of intersecting

connectors being coaxal

are copolar.
60.

Hence the

sixth pair of connectors are coplanar.


if

the tetrastims be coplanar, and

five intersections

of

pairs of correlated connectors

are costraight, this the

coplanar

case can be

depend upon the other by substituting for one 6f the tetrastims its image on a second plane meeting
to
first

made

the

on the bearer of the


the axis

five intersections.

61.
If of

If

two

a figurative straight, the theorem reads : correlated tetrastims five pairs of mated connectors
is

are parallel, so are the remaining pair.


62.

Four costraight points

are called

'

harmonic

points,' or

78

PROJECTIVE GEOMETRY.
if

[Chap.

III.

a harmonic range,'
'

the

first

and third are codots of a

tetra-

stim while the other two are on the connectors through the
third codot.
63.

By

three costraight points and their order the fourth


is

harmonic point

uniquely determined. For

if

the three points

in order are

A, B,

C,

draw any two


Join

third through

B to

cross these at

CK, crossing
join

AM at N.

K and M respectively.
from B.

straights through

A, and a
Join

LN crosses the straight

CM, crossing AK at L. Then the ABC, always at the same point D,

the fourth harmonic to

A BC separated

64. In projecting from a point not coplanar with it a tetrastim defining a harmonic range, the four harmonic points are projected by four coplanar straights, called harmonic
'

straights' or a 'harmonic flat-pencil.'


65. The four planes projecting harmonic points from an axis not coplanar with them are called 'harmonic planes,' or a

harmonic

axial-pencil.'

harmonic primal figure gives always again a harmonic primal figure. 67. By three elements of a primal figure, given which is the
66. Projecting or cutting a

second, the fourth harmonic


68.

is

completely determined.

Defining harmonic points by the tetrastim distinguishes

ART.

6.]

HARMONIC ELEMENTS.

79

two points made codots from the other two. Yet it may be shown that the two pairs of points play identically the same
role.

First,

arated two

from the definition of four harmonic points each sepmay be interchanged without the points ceasing to

be harmonic
codots.

[or,

if

A BCD

is

ADCB, CBAD,

and

CDAB\

a harmonic range, so is also For the first and third remain

Second, to prove that in a harmonic range the two pairs of separated points may be interchanged without the four points is a harmonic range ceasing to be harmonic [or, if

ABCD

(and therefore
the joins

ADCB, CBAD, DABC, BCDA, and DCBA] Through the


:

and CDAB), then

also

is

BADC
O

third codot

draw

AO

and CO.

These determine on the connectors


four

NK, KL, LM, and


tively.

MN

new

points, S, T, U, V, respec-

The

tetrastim

KTOS has
B; hence
In like

for
its

two codots

and

C,

and

has a connector though

must pass though D.


the tetrastim

remaining connector manner, the connector


y

TS
of

UV

MVOU must

of each of the tetrastims


fore

B and D are
The

pass through D and LUOT and VNSO through B. Therecodots of a tetrastim STUV with the remain-

a connector

ing connectors, one through A, one through C.


69.

separated points

points,' as also are


'

B and

D.

conjugate Either two are said to be 'harto the other two.

and

are called

'

monic conjugates with respect


Prob.
8.

To

accessible cross

X of two given straights n and

determine the join of a given point


n'.

with the in-

80

PROJECTIVE GEOMETRY.

[Chap.

III.
'

at

D and D'

Through

M draw any two


.

straights crossing ndXB and B\ and draw Join Z>Z? and Z/i?', crossing on A. Through third straight crossing any

and Join ^'Z>" and


n at

B"

'

at Z>".

>'",

Then crossing at Z. is the join required.


Proof.

LM

The

tetrastim

XBMD
XB'LD'
does

makes AB'C'D'

AB"C"D".

and cutting the eject are costraight.* C", C", and Prob. 9. Through a given point to draw with the straight-edge a straight parallel to two given parallels.

a harmonic as range, But projecting AB"C"D" from X, by AB'D' gives a harmonic range. Therefore

Prob. 10.

To

inconstructible join

determine the cross of a given straight m with the x of two given points and N'. Join any two

points on m with JV and N' giving b and b' on JV d and </' on N*i Join the crosses db and </'' On a take by 0.
y
f

any third point


ing with

N
A

joinin b"
in

and with
quired.

T/

d"

Join the crosses

b'

d" and
8,

<W

by

/.

Then Im

is

the cross re-

[From
range.

Prob.

by

duality.]

Prob. 11. Cut four coplanar non-copunctal straights in a har-

monic

12. On a given straight determine a point from which the of three given points form with the given straight a harmonic ejects

Prob.

pencil.

Art.
70.

7.

Projectivity.

Two
If

jective.

primal figures of three elements are always proone be a pencil, take its cut by a transversal. If

the bearers of

ABC

and

AB C
' '

be not coplanar, join

AA\
'.

BB\ CC\

and cut these joins by a transversal, ;/z. Then ABC and A'B'C are two cuts of the axial mAA' mBB' mCC
,

Surveying may be solved by the application of the preceding principles, but such application has not been found advantageous in See Gillespie's Treatise on Land Surveying, New York, 1872. practice.
in

* Numerous problems

Art.

7.]

projectivity.

81

If

the bearers are coplanar, take on the join

projection vertices
at

M and
,

M\

Join

MB
.

and
Join

&'\ join
at

MC and M' C
Then

M B\ crossing
f

A A'

any two
crossing
.

crossing at

C"

B" C"

AA'

A".

ABC and

A' B'

C are images of A" B" C"


of

71. If any four harmonic elements are taken in one of two projective figures, the four elements correlated to these are also

harmonic.

For both

ejects

and cuts

harmonic figures are

themselves harmonic.
72. Two primal figures are projective if they are so correlated that to every four harmonic elements of the one are

For and cuttings which derive A' B' the same projectings from ABCv\\\ give D from D. Therefore A'B'CD, is harmonic. B C D' is harmonic. Therefore D is D But by hypothesis

correlated always four harmonic elements of the other.

C
1

'

'

'

'.

73.

If

two primal

figures are projective, then to every con-

secutive order of elements of the

a consecutive order of
a bearer.
74.

one on a bearer corresponds the correlated elements of the other on


primal figures having three elements For two self-correlated elements

Two

projective

self-correlated are identical.

cannot bound an interval containing no such element, since they must harmonically separate one without it from one
within.
75.

Two

ranges are called

'

'

perspective

if

cuts of the

same

flat pencil. flat pencils are perspective if cuts of the same axial Two axials are perspective or ejects of the same range. pencil, if ejects of the same flat pencil.

Two

A
flat

range and a flat pencil, a range and an axial pencil, or a pencil and an axial are perspective if the first is a cut of

the second.
76,. If two projective ranges not costraight have a self-correlated point A, they are perspective.

76'.
flat

two coplanar projective pencils not copunctal have


If

a self-correlated straight a, they


are perspective.

82

PROJECTIVE GEOMETRY.

[Chap.

III.

Let the join of any pair of f cross the correlated points at V. join of any other pair

BB

CC

Let the cross of any pair of correlated straights bb' join the cross of any other pair cc' by m.
Cutting the two given
cils
flat

Projecting

the

two

given
are

pen-

ranges from
tive

V, their ejects

by

m, their cuts are identical,

identical, since they are projec-

since

and have the three

straights

they are projective and have the three points tna, mbb\
mcc' self-correlated.

VA,
lated.

VBB\ VCC'
Art.

self-corre-

8.

Curves of the Second Degree.


non77'.

77r

If

two

coplanar

If

two

coplanar

non-

copunctal flat pencils are projective but not perspective, the


crosses

costraight ranges are projective but not perspective, the joins of

of
' '

correlated
'

straights

form a

gree,' or

range of the second deconic range.

correlated points form a pencil of the second class,' or 'conic


pencil.'

'

78,.

If

two copunctual

non-

78'.

If

two
flat

copunctal

non-

costraight axial pencils are projective but not perspective, the

coplanar tive but


planes
'

pencils are projecnot perspective, the

meets of correlated planes form a conic surface of the second


'

of

correlated
*

straights

order,' or 'cone.'

form a pencil of planes of the second class,' or cone of planes.'

All results obtained for the conic range or the conic pencil are interpretable for the cone or cone of planes, since the eject of a conic is a cone and the cut of a cone is a conic.
79.
8o,.

On the cross A of

of correlated straights a

any pair and a


x

80'.

On

the join a of any pair

of correlated points

and

of

Art.

8.]

CURVES OF THE SECOND DEGREE.


flat

83

of the projective

pencils

the projective ranges u and take two points and

x
.

The ejects

abc and a b x c x
x

being projective and having a pair of correlated


straights a, a x coincident, are perspective, both be-

ing ejects of the range on the join of the crosses a


,

bb

and

cc x .

Any
joined

point with

D
V
by

of
d,

u,
is

then

correlated

to

the
of

and V draw two and u


x
x

straights

cross of u x with the join and the cross du t


.

d
x

The

cuts

ABC

and

ABC
X

Any

D joined to

its

so de-

being projective and having a pair of correlated points A, A coincident, are perspective, both
x

termined, gives the conic pencil

p K,

a straight of

being cuts of the pencil on the cross of the


joins

2f

BB

and

cc,.

Any straight d of ^crossing u at B>, is then


correlated
to

the join of x with the cross > of u and


x

the join

DVV

Any d crosses its d so deterx

mined, at P, a point of the conic


range
8 1,.
k.

The

pencil-points V, Vx

8i'.

The

bearers u, u
their

of the

of the generating pencils pertain to the conic, since their join


crossed by the element correlated to it in either pencil
X

generating ranges pertain to the


conic, since

cross

uu

is

VV

is

joined to the element correlated to it in either range by its bearer.

at

its

pencil-point.

84
82,.

PROJECTIVE GEOMETRY.

[Chap. Ill

The
x

lated to

VV
its

correstraight on ' is called the tan-

82'.

The
is

to u u
x

point on u correlated called the contact on


'
!

gent' at V.

on

V is

Every other straight join with a second


straight, as #,

Every other point on u is its cross with a second straight of


u.

point of the conic.

the conic.

83^

On

any

on

83'.

On any

any point second element


with the join
84,.
t

of

the
is

conic, its cross

its

any second element


the cross u u^
x

straight a

point, as V, on of the conic, its


is its

join q with

V VV
X

From
}

VV AMZ
P.
join

of

the five given points construct a sixth,


>

84'.

From

the

five

The

cross

of u with the
cross

straights u, ,, a, struct a sixth

DD

q-y

rlt of
,

K conThe

given

VP, and the


.

or p.

of u

are costraight with the join VX Therefore* the three with F2

join

with the cross up, and the join dx of V with the cross u p are copunctal with w 2
of
x
x

opposite pairs in every complete set of connectors of a hexastim

whose dots are


sect in

in a conic inter-

opposite pairs in every complete set of fans of a hexagram whose sides


f

Therefore

the

three

three costraight whose bearer is called a


straight.'

codots
'

are in a conic concur

in

three

Pascal

copunctal diagonals whose bearer is called a Brianchon point.'


'

This hexastim has sixty Pascal straights, since it has sixty complete sets of connectors. 85,. The ejects of the points of a conic from any two are projective.
86,. five

chon

This hexagram has sixty Brianpoints, since it has sixty


sets of fans.

complete

cuts of the straights 85'. of a conic by any two are projective.


86'.

The

By
is

of

its

points a

By
is

five of its straights

conic
87,.

completely determined. Instead of five points

conic
87'.

completely determined. Instead of five straights


given the two bearers pairs of correlated

given the two penciland three pairs of correpoints

may be

may be

and three

If one given the join of the pencilstraight points, then four points and a tangent at one of them are given.

lated

straights.
is

points. If one given point is the cross of the bearers, then four straights

and a contact point on one of them are given.

Thus by four of its points and the tangent at one of them a


*
Pascal, 1640.

Thus by four of its straights and a contact-point on one of


f Brianchon, 1806.

Art.

8.]

CURVES OF THE SECOND DEGREE.


completely determined.
of
at
its

85
completely deits

conic

is

them a conic
termined.
88'.

is

88^ By three

points

By

three of

straights

and the tangents


the conic
is

two of them

completely deter-

and the contact-points on two of them the conic is completely


determined.
Interpreting a pentagram hexagram with two sides coinciding gives: In every com89'.

mined.

89^ Interpreting a pentastim as a hexastim with two dots


coinciding gives: In every complete set of connectors of a pentastim whose dots are in a conic, two pairs of non-consecutive

as a

whose
pairs

plete set of fans of a pentagram sides are in a conic, two


of

non-consecutive

fans

connectors

determine by their

determine by their two concurs


a point on which is the join of the fifth fan-point with the contact-point on the opposite side.

two intersections a straight on which is the cross of the fifth


connector with the

tangent at the opposite dot. Thence follows the solution of the problems

90^ Given

five

points

of

90'.

Given

five straights of

conic, to construct tangents at the points, using the ruler only.

conic, to find contact-points on the straights, using the ruler only.


91'.

9 1 j.* pair of

The

hexastim

with

The hexagram

opposite connectors re-

of opposite fans replaced

with a pair by con-

placed by tangents gives:


intersections of the
pairs in

The

two opposite every complete set of

tact-points gives: The concurs of the two opposite pairs in every complete set of fans of a tetra-

connectors of a tetrastim with dots in a conic are both costraight


with the crosses of the two pairs
of tangents at opposite dots. tetrastim with dots in Or:

gram with

sides in a conic are

both copunctal with the joins of the two pairs of contact-points

on opposite
Or:

sides.

A tetragram with sides in a

a conic has each pair of codots costraight with a pair of fanpoints of the tetragram of tangents at the dots.

conic has each pair of diagonals copunctal with a pair of connectors of the tetrastim of contacts

on the

sides.

The
92,.

figure

for

91,

and that

for 91'

are

identical,

and

called Maclaurin's Configuration.

(See page S6.)


92'. The contact-points of a conic pencil are a conic range.

The

tangents of a conic
* Due

range are a conic pencil.

to Maclaurin, 1748.

86
93.
all

PROJECTIVE GEOMETRY.

[Chap. III.

The

on a curve, a

points of a conic range may now be conceived as-, conic curve,' their bearer. The straights of
'

the

corresponding

conic

pencil,

tangents of this conic range,

may
on

now
this

also be conceived

as all

same conic curve on which


contact-points.
is

are their

Consedual to

quently the conic curve


itself,

and

so the principle of dual-

ity

on a plane receives an impor94. It follows

tant extension.

immediately from
all

their generation that

conies are

closed curves, though they

may

be compendent through one or

two points
'hyperbola with none,
95.
If

at infinity.

With two
is

points at infinity the


;'
'

curve

called

with one, 'parabola;* *

ellipse.'

gents to 'without'
*

point has on it tanthe curve, it is called


a

the

curve;

if

none,

within

'

the curve.
all

The

contact-

point on a tangent

is

'

on

'

the curve

other points on a tan-

* The generation shows that a straight cuts the curves in two points and Hence the that from any point two tangents to the curves may be drawn.
curves are of the second order and of the second class, that
is

they are identical

with the conies of analytic geometry. o,. Analytically the equations P-\-XQ P' -f- XQ' o, where P, Q, P', (7 are linear functions of point coordinates,

represent two projective pencils, the correlated rays corresponding to the same Hence the locus of the intersection of correlated rays is reprevalue of A. sented by PQ' Q o, a second-degree point equation. Projective ranges

'

are represented by

= R XS
-j-

o,

R'

AS'

= o,

where R,

S, R', S' are linear


is

functions of line coordinates.

The envelope

of the joins of correlated points

represented by

The

a second-degree line equation. projective generation of conies is developed synthetically in Steiner's

R S'

R' S

= o,

Theorie der Kegelschnitte, 1866, and in Chasles' Geometrie superieure, 1852. For the analytic treatment see Clebsch, Geometrie, vol. 1, 1876.

Art.

9.]

pole and polar.

87

gent are without the curve. Every straight in its plane contains innumerable points without the curve, since the straight crosses every tangent.
of them,

Prob. 13. Given four points on a conic and the tangent at one draw the tangent at another.

Prob. 14. If the n sides of a polygram rotate respectively about n fixed points not costraight, while {n 1) of a complete set of fanpoints glide respectively on (n 1) fixed straights, then every remaining fan-point describes a conic*

Prob.
of

15.

In any tristim with dots on a conic the three crosses


the

the

connectors with

tangents

at

the

opposite

dots

are

costraight. f

Prob. 16. If two given angles rotate about their fixed vertices so that one cross of their sides is on a straight, either of the other
three crosses describes a conic. \

Prob. 17.
straights

Construct a hyperbola from three given points, and


its

on

figurative points.

Art.
96.

9.

Pole and Polar.

Taking every tangent to a conic as the dual to its own contact-point fixes as dual to any given point in the plane one polar,' of which the point is the particular straight, its
'

'pole.'

polar of

Z
B

reference to any given conic, to construct the any given point in its plane. Put on the given point two secants crossing the curve, one at A and D, the other at
97.

With

and

C.

The

join of the other codots

X and
as

Y of ABCD

is

the polar

of Z,

must always be the cross 5 of the tangents at A and D, and also the point which D and A harmonically separate from Z (given by each of the variable
change
this polar, since
it

Varying on

either

secant,

ZBC,

does not

tetrastims
98.

BXCY).
join of

The
is

any two codots

of a tetrastim with dots

on

a conic
*
f
\

the polar of the third codot with respect to that

Due to Braikenridge, 1735. From Pascal dual from Brianchon. Given by Newton in Principia, Book
;

I,

lemma

xxi,

under the name of

'the organic description " of a

conic.

88
conic,

PROJECTIVE GEOMETRY.

[CHAP.

III.

and either codot

is

the pole of the join of the other


its

two.

Any

point

is

harmonically separated from

polar

by

the conic.

with ruler only the tangents to a conic from a point without, join it to the crosses of its polar with the
99.

To draw

conic.
ioo,.

Two

points

are

called

100'.

Two

straights are called

'conjugate' with reference to a conic if one (and so each) is on the polar of the other.
ioij.

'conjugate' with reference to a conic if one (and so each) is on


the pole of the other.
101'. All

are

All points on a tangent conjugate to its contact-

tact-point

straights on a conare conjugate to its

point.
io2j.

tangent.

The

points of a range

102'.

The
are

are projective to their conjugates on its bearer.

pencil

straights of a flat projective to their


its

conjugates on
103'.

bearer.

103^ With reference to a given


conic,

the
all

'

polars of

kerncurve,' points on a second


'

the

With reference to a given the 'kerncurve,' the poles conic, of all tangents on a second conic

conic

make
is

bearer

a conic pencil, whose of the polarcurve


'

make
is

a conic range, whose bearer the polarcurve of the second


'
'

the second conic.

conic.

Prob. 18. Either diagonal of a circumscribed tetragram is the polar of the cross of the others. Prob. 19. A pair of tangents from any point on a polar harmonically separate
it

from

its

pole.

Prob. 20.

A pair of tangents are harmonic


on
their cross

conjugates with respect

to any pair of straights respect to the conic.

which are conjugate with

Art.

10.

Involution.
throw
')

104. If in a primal figure of four elements (a

first

any two be interchanged, then the other two, the result jective to the original.
[That is, ABCD a BADC a Let ABCD be a throw on m.
eject

is

pro-

CDAB

a DCBA.~]
it

by a

straight (m')

on A.

Project The cut

from

V.

Cut

this

is

AB'CD'.
by

Now

project ABCD from

C.

The

cut of this latter eject

V B is

Art.

10.]

involution.

89

B'BVH. Project B'BVH from D and cut the eject hyui. The cut is B'AD'C, which is perspective to BADC.

105. Two projective primal figures of the same kind of elements and both on the same bearer are called conjective.' When in two conjective primal figures one particular element has the same mate to whichever figure it be regarded as be'

longing, then every element has this property.


If
is

AA'BB'

is

projective to

A'AB'X, then by

104,

AA'BB'

projective to

AA'XB', and having

three elements self-corre-

lated,

they are identical.

106.

Two

conjective
('

figures such that


')

the elements are

mutually paired
ple,

coupled

form an

'

Involution.'

For exam-

the points of a range, and, on the same bearer, their conjugates with respect to a conic, form an involution. Every
eject

and every cut of an involution is an involution. 107. When two ranges are projective, the point at
is

infinity
its

of either one
*

correlated to a point of the other called

vanishing point.'
' '

108. When two conjective ranges form an involution the two vanishing points coincide in a point called the center of

the involution.
109. If

two

lated elements, these are called the


involution.
for

figures forming an involution have self-corredouble elements of the


'
'

An

involution has at most two double elements

were three

self-correlated, all

would be

self-correlated.

a primal figure of four elements is projective with a second made by interchanging two of these elements, they harmonically separate the other two.

HO.

If

For project the range

ABCD

from

Fand

cut the eject

by a

90
straight on A.

PROJECTIVE GEOMETRY.

[Chap.

III.

The

cut

AB'CD'
tive

is

projective

to
is

ABCD,
projec-

which by hypothesis
to

ADCB. Therefore ADCB is perspective to


AB'CD'.
cross

So

VCC

is

on the

of the joins

DB' and

BD'.
of the tetrastim

VD'XB' while A
',

So and C

and

are codots

are on the connectors

through C, in. If an involution has two double elements these sepaLet A and C be rate harmonically any two coupled elements.
the double elements.
therefore by
1 1

the third codot.

Then ABCB'
is is

is

projective to

AB'CB

no ABCB'
involution

harmonic.

completely determined by two couples. 7\ A'AB' ... is For the projective correspondence AA'B completely determined by the three given pairs of correlated elements, and since among them is one couple, so are all corre2.
.

An

lated elements couples. 113. When there are double elements, then the elements

no couple are separated by those of another couple. Inversely, when the elements of one couple separate those of
of

another, then the elements of every couple are separated those of every other, and there are no double elements.
114'.

by

The

three pairs of op-

posite fan-points of a tetragram are projected from any projec-

tion-vertex by three couples of an involution of- straights.

The three pairs of opconnectors of a tetrastim posite


114,.

are cut

by any transversal

in

three couples of a point involution.*

*Due

to Desargues, 1639.

Art.

It.]

PROJECTIVE CONIC ranges.

9t

Let

QRST

connectors

RT and

be a tetrastim of which the pairs of opposite QS, ST and QR, QT and RS are cut by
in

any transversal respectively

and A',

and

',

C and C\
'

From
5.

the projection-vertex

the ranges

ATPR and ACA B

are perspective.

But

A TPR and ABA' C


is

are perspective from

Therefore

ACA'B'
and
6?

projective to

ABA'C, and

therefore
(

to^'C042?(
are i? and
115.
i?',

104).

Since thus A and

^4'

and

are coupled, so

105)

To

construct the sixth point


given,

C of
The

an involution of

which
at R.

five points are

which take any two points

draw through C any straight, on and T. Join AT, B'Q crossing


join

Join BT, A'Q crossing at 5. bearer of the involution in C '.


Prob. 21. Find the center

RS

cuts the

of a point involution of

which two

couples

AA'BB'

are given.

and iV on m are harmonically sepaProb. 22. If two points rated by two pairs of opposite connectors of a tetrastim, then so are they by the third pair.
Prob. 23. To construct a conic which shall be on three given points, and with regard to which the couples of points of an involution on a given straight shall be conjugate points.

Art.

11.

Projective Conic Ranges.

116. Four points on a conic are called harmonic if they are projected from any (and so every) fifth point on the conic by four harmonic straights.

two conies are called projective when so correlated that every four harmonic elements of the one correspond to four harmonic elements of the
117.

A conic

and a primal

figure or

other.
If a conic range and a flat pencil are projective, and element of the one is on the correlated element of the every conic is projected from other, they are called perspective.

118.

every point on it by a flat pencil perspective to it. Inversely the pencil-point of every flat pencil perspective to a conic is,

on the

conic.

S2
119.

PROJECTIVE GEOMETRY.

[CHAP.

III.

pencils respectively Therefore any three perspective .elements in one can be correlated to any three elements in the other, but this completely pairs all the elements.
to
120.

Two

conies are projective if them are projective.

flat

Two
two

different conic ranges

on the same bearer have

at most
121.

self-correlated elements.

in

Two different coplanar conic ranges with a point V common are projective if every two points costraight with
correlated.
V.

pencil on
;

For both are then perspective to the flat Every common point other than V is self-correlated but V only when they have there a common tangent. They can have at most three self-correlated points.
122. If a flat pencil

V are

V and conic range k are coplanar and but not perspective, then at most three straights of projective the pencil are on their correlated points of the conic but at
;

least one.

For any

flat

pencil

M perspective
in general a

and with it determines must have in common with k every point which lies on its correlated straight of V, So if more than three straights of V were on their correlated points of k, the conies would be identical and V perspective to k.
Again, since every conic is compendent, and so divides its plane into two severed pieces, therefore the two different conies
if

projective to F, second conic range which


is

to k

they cross at their cotnmon point

M must
it

cross again, say


correlated,

at P.

and so

In this case the straights VP and VP is on the point correlated to

MP are
on
k.

In case they do not cross at their common point M, the corresponds to the common tangent at M, and so straight

VM

to the point
123.

correlated to

it

on

k.

projective conic ranges on the same curve form an involution if a pair of points are doubly correlated. Besides be any other two correlated and the couple V let

Two

AA
X

points, so that

B B B corresponds AA
x

to
u.

AA

and

BB

call U,

and

its

polar

A AB r The Project AA B
x

cross

of

from

Br

Art.

11.]

projective conic ranges.

93
X X

Project

A AB
t

from B.

The

ejects

are projective, and having the straight self-corre) (or The crosses of their correlated elelated, so are perspective.
X X

B {AA B) and B(A AB } BB BB


X X

ments are therefore


correlated straight

costraight.
X

But the cross of

BA
X

with

its

BA

is

known

to be on u, the polar of U, the

cross of
is

AA

with

BB

Likewise the cross of


x

BA
X

with

BA

on

u.

of

CU

Therefore the point C correlated to C with the curve. So C and C are coupled.
x
.

is

the cross

124. If

two conic ranges form an


all

involution, the joins of


'

coupled points are

copunctal on the

involutioncenter.'

125. Calling projective the conic pencils dual to projective

conic ranges, if these ranges form an involution, so do the all costraight pencils, and the crosses of coupled tangents are

on the

'

involutionaxis.'

pencils forming an involution are cut by each Two of their straights in two ranges forming an involution.

So two conic

conic ranges forming an involution are projected from each of


their points in
126. If

forming an involution. the involutioncenter lies without the conic bearer


flat pencils,
it

two

of an involution,

has two double elements where

it is

cut

by

the involutionaxis.
127.

To

construct the self-correlated points of two pro-

jective conic ranges on the three points of k, and


x ,

same
x ,
x

conic.

Let A, B,

A B C

their correlated points of


X
X X

be any kv

The

projective

flat

pencils

A(A B C ) and A {ABC) have


X
X X

AA

to a range on self-corresponding, hence they are perspective of AB and A B with the cross of AC the join u of the cross
X

94

PROJECTIVE GEOMETRY.

[CHAP.
this join

III.

and

C.

The

crosses of the conic

and

n are the

self-correlated points of

k and k

on a conic, the six points where a straight not on a dot cuts the conic and two pairs of
128. If the dots of a tetrastim are

opposite connectors form an involution. For the two flat pencils in which the two crosses of

with the conic, P, Pi9 and two opposite dots R, T, are projected from the other two dots Q, S, are projective, and consequently so are the cuts of these
flat
X

pencils
X
X

by
X

tn\

that

is,

PBP A PBP A
X

a PA P B 7\P B PA
X X

But

PA P B A P B PA
X
X

Therefore

129,. Conies on which are the dots of a tetrastim are cut by a transversal in points of an involu-

to 129'. Copunctal tangents conies on which are the sides of

a tetragram form an involution,

tion.

At

its
is

transversal

double points the tangent to two of

The double
of
point.

straights touch
at the

two

those conies

pencil-

those conies.

Prob. 24. The pairs of points in which a conic is cut by the straights of a pencil whose pencil-point is not on the conic form an
involution.

Art.
130.

12.

Center and Diameter.


of a point at infinity with
is

The harmonic conjugate

respect to the end points of a finite sect


sect.

the

'

center

'

of that

pole of a straight at infinity with respect to a certain conic is the center of the conic.
131.
'
'

The

polar of any figurative point the conic, and is called a diameter.'


132.
'

The

is

on the centre of

133. If a straight

crosses a conic the sect between

the

crosses

is

called a

'

chord.'
is

The
134.

center of a conic

the center of

all

chords on

it.

The

centers of
all

diameter are

chords on straights conjugate to a on the diameter.

135. Two diameters are conjugate of the figurative point on the other.

when each

is

the polar

ART.

12.]

CENTER AND DIAMETER,

95

136.

The tangents

at the crosses of a straight with a conic


is

cross on the diameter which

a conjugate to that straight.

137. The joins of any point on the conic to the crosses of a diameter with the conic are parallel to two conjugate diameters.

138. Of two conjugate diameters, each is on the centers of the chords parallel to the other and if one crosses the conic, the tangents at its crosses are parallel to the other diameter.
;

center of an ellipse is within it, for its polar does not meet the curve, and so there are no tangents from it to the
139.

The

centre of a parabola is the contact point of the The centre of a hyperbola lies without the figurative straight. The tancurve, since the figurative straight crosses the curve.
curve.

The

gents from the center to the hyperbola are called asymptotes.' Their contact-points are the two points at infinity on the
'

curve.
140. If a

gents at its crosses can

diameter which cuts the curve be given, the tanbe constructed with ruler only, and so
straights conjugate to the diameter.

however many chords on


141.

Every flat pencil is an involution of conjugates with Hence the pairs of conjugate diamto a given conic. respect eters of a conic form an involution.
If

the conic

is

a hyperbola, the asymptotes are the double

straights of the involution.


eters

totes

Hence any two conjugate diamof a hyperbola are harmonically separated by the asympand since the hyperbola lies wholly in one of the two

explemental angles made by the asymptotes, one diameter cuts the curve, the other does not.

one pair of conjugate diameters of an ellipse is always separated by any other pair. Any one pair of conjugate diameters of a hyperbola is never separated by any other pair.
142.

Any

and

tangent to a hyperbola cuts the asymptotes at A then the contact-point B is the center of the sect AC, since the tangent cuts the harmonic pencil made by the diameter through B the conjugate diameter and the asymptotes, in
143. If a
y

the harmonic range

ABCD where D

is

at infinity.

Just so the

96

PROJECTIVE GEOMETRY.
is

[CHAP. IIL

center of any chord by the asymptotes.


144. If a point

the center of the costraight sect bounded

is

the center of two chords


is

it

is

the center

of the conic, for its polar


145.

the figurative straight.

As many

structed

With
ruler

points as desired of a conic may be conthe ruler alone. by the aid of one fixed conic all problems solvable by

and compasses can be solved by ruler alone, that is, by pure projective geometry. For example Of two projective primal figures (say ranges) on the same bearer, given three
:

pairs of correlated elements to find the self-corresponding ele-

ments, if there be any. V of the given conic.


projective conic ranges.

These

Project the two ranges from any point ejects are cut by the conic in

Of these determine the

self-correlated

points by

127.

The ejects cut the bearer of the Project these from V. original ranges in the required self-correlated points.
Prob. 25. Find the crosses of a straight with a conic given only

by
for

five points'.

Prob. 26. Given a conic and

its

center, find a point

two given points A, C, the center of the sect


Prob. 27.

B such that AB shall be C.


two
coinitial sects

is

The join of the other extremities of to the join of their centers. parallel
Prob. 28. In an ellipse let CA, CB with the curve.
f
.

and B be crosses of conjugate diamThrough A' the cross of the diameter conjugate to CA with the curve draw a parallel to the join AB. Let Then CB' is the diameter conjugate it cut the curve again at B to CB.

eters

Art.
146!.

13.

Plane and Point Duality.


5

On

a plane are 00 points,

146'.

On

a point are
oo
1

00

planes,

'

point-field.'
147,.

a
1

'

plane-sheaf.'
147'.

The

oo

planes of a sin-

The

gle axial pencil

have on them

all

gle range

have on them
3

points of a sinall the


so

the

points of point-space; 3 there are just oo points.

so

planes of plane-space;
are just oo

there

Point-space

is

tridimensional.

planes. Plane-space is tridimensional.

Art.

13.]

PLANE AND POINT DUALITY.

97
is

148.

With the

straight as element, space


2

of four dimen-

sions.

On

a plane are

oo

straights,
1

On
On

a point are

oo

straights,

a 'straight-field.' On a straight are

straight-sheaf.'
00
1

planes,

a straight are
00
3

00

points,

and so

00

straights.
2

and so
2

straights.
oo
2

On

each of the
00

a plane are the


oo
4

points on straights of a
00

On

each of the
oo
2

planes on

a point are the


straight-field;
oo
4

straight-sheaf; so there are just


straights.
i49j.

straights of a so there are just

straights.
149'.

planes determine a straight, their meet.


i50j.

Two

Two

points determine a points determine a

straight, their join.


150'.

Two planes determine


on
their meet.

an

Two
Two

axial-pencil

range on
planes
151'.

their join.

15

j.

Two

bounding

bounding
sect.

points

determine an axial angle.


152,.

determine a
152'.

A
it

not on
pass.
i53j.

plane and a straight determine a point, their


axial
its

A
it

point and a straight

not on

determine a plane.

An
flat

plane not on

pencil and a bearer deter-

153'.

on

its

A range and a point not bearer determine a flat

mine a

pencil.

pencil.
154'. Three points determine a plane, their junction. 155'.

154^ Three planes determine


a point, their apex.

155^ Three planes determine a plane-sheaf.


156^

Three points determine

a point-field.
156'.

iVo

coplanar straights

Two

copunctal straights

are copunctal.
157.

are coplanar.

Any figure, or the proof of any theorem of configuand determination, gives a dual figure or proves a dual theorem by simply interchanging point with plane. Thus all the pure projective geometry on a plane may be read as geometry on a point.
ration
Prob. 29. If of straights copunctal in pairs not then all are coplanar.
all

are copunctal,

Prob. 30. On a given point put a straight to cut two given straights. Prob. 31. If two triplets of planes afty, a' ft'y' are such that the meets fty and ft'y', yoc and y'a\ aft and a' ft' lie on three

yy'

planes a", ft", y" which are costraight, then the meets <xa' y ftft\ are coplanar.

98

PROJECTIVE GEOMETRY.

[Chap.

III.

Prob. 32. Describe the figures in space dual to the polystim and
the polygram.

Art.
158.

14.

Ruled Quadric Surfaces.


' '

The joins of the correlated points of two projective whose bearers are not coplanar form a ruled system ranges of straights no two coplanar. For were two coplanar, then two points on the bearer tn and two on the bearer m would all four be on this plane, and so m and m coplanar, contrary
x
x

to hypothesis. 159. Let the straights n,n iy // 2 be any three of the elements of a ruled system, and iV2 any point on ;/ 2 Put a plane on iV2 and the straight n if and let its pass with n be called N. The
.

straight

NN^

cuts n> n iy

n.2 all

three.

ranges of the ruled system (on the bearers


straight iv7V2 (or
pencils,

Projecting the generating and ) from the

Q)

as

axis produces

which having three planes m 2 n, sponding, are identical. Therefore every pair of correlated points of the ranges on m and ;;/, is coplanar with ;;/ 2 that
x

two projective axial mji mjt^ self-corre,

is,

m^ cuts every element of the ruled system.

By varying
all

the

oo

the point N^ 00 straights are obtained, all cutting straights of the original ruled system and making

on every two projective ranges. Of the straights so obtained no two cross, for that would make two of the first ruled system
coplanar.

a 'ruled

Either of these two systems may be considered as generating Each of the surface,' which is the bearer of both.
is

two systems

completely determined by any three straights From of the other, and therefore so is the ruled surface also. the construction follows that the straights of either ruled
the straights of the other in projective ranges. So any two straights of either system may be considered as bearers of projective ranges generating the other system, or

system cut

all

indeed the ruled surface.


160.

two

each point of this ruled surface are two and only straights lying wholly in the surface (one in each ruled

On

Art.

14.]

ruled quadric surfaces.

99
is

system).
also
surface.

So a plane on one

straight of the ruled surface

on another straight

of this

161. If in the

two generating

projective ranges the point at infinity of one is correlated to the

point at infinity of the other, the ruled surface is called a hyper'

bolic-paraboloid.'

The
is

join

of

these figurative

on the figurative plane. points Therefore the plane at infinity

common

cuts the surface in a straight and so has a second straight in with the ruled surface.

That a hyperbolic-paraboloid has two straights in common with the plane at infinity may also be proved as follows
:

n,

and let Call the bearers of the generating ranges m and lt n be any two elements, and /the element at infinity. By 159 the ruled surface may be considered as generated by the
x

But all these straights straights on the three elements n, n /. must be parallel to the same plane, namely, to any plane on/. On /and each one of these straights put a plane these planes
x

a parallel-axial-pencil, and cut any two of the original elements in projective ranges with the figurative points correlated. Therefore the figurative straight joining the figurative
points of n and
162.
#, is

make

wholly on the ruled surface.


all

From
plane.

161 follows that

same ruled system on a hyperbolic-paraboloid are


the

straights pertaining to the parallel to


'

same

Such planes
is

are called

asymptote-planes.'

hyperbolic-paraboloid completely determined by two nonand an asymptote-plane cutting them. To coplanar straights
get an element cut the two given straights by any plane parallel to the asymptote-plane, and join the meets.
163.

Three non-crossing

straights, all parallel to the

same

Let m, plane, completely determine a hyperbolic-paraboloid. be the given straights. The passes of planes on tnv ///, %

100

PROJECTIVE GEOMETRY.

[CHAP.

III.

with

and

are projective ranges

whose

joins are a ruled

system.

But from the hypothesis one of these planes is parallel to both m and m.. Therefore their points at infinity are correlated and the ruled surface is a hyperbolic-paraboloid.
164. If

two non-coplanar projective ranges be each

axially

projected from the bearer of the other, two projective axial pencils are formed, with those planes correlated on which are If A, A be correlated the correlated points of the ranges.
l

the meet of correlated planes. Thus two projective axial pencils with axes not coplanar genIf the whole figure be cut by a plane, erate a ruled system.
points, then the straight
1

AA

is

this will cut these axial pencils in

two projective
will

flat

pencils,

and the conic generated by these


surface.

be the cut of the ruled

So every plane

cuts

it

in a conic or a pair of straights.

straight not wholly on the surface can cut it in more than two points. The surface is therefore of the second degree

Hence no

(quadric).

the plane at infinity cuts the ruled surface in a pair of If not, it is called a straights, it is a hyperbolic-paraboloid.
If

hyperboloid of one nappe,' a ure of which is here shown.


1

fig-

allel to

164J. Copunctal straights parthe generating elements of

a hyperboloid of one nappe are on a cone. Copunctal straights parallel to

the generating elements of a hyperbolic-paraboloid are on a

system of two planes.

For the figurative plane cuts the hyperboloid of one nappe in a conic curve, but cuts the hyperbolic-paraboloid in two straights and each of the copunctal straights
;

goes to a point of the figurative cut. 165. Each straight in one ruled system of a hyperboloid of

ART.

14-]

RULED QUADRIC SURFACES.


is

101

one nappe
loid

parallel to one, but only to one, straight in the

other ruled system.

Of the

straights on a hyperbolic-parabox
,

Let n and n any two elements of parallel. one ruled system, be the bearers of the generating ranges R and R,. If V is the vanishing point of R, then the straight on

no two are

V parallel
1

to n

is

an element of the other ruled system.

But

for the hyperbolic-paraboloid


66.

Fis

itself

a figurative point.
is

Any
'

straight of
'

one ruled system on a ruled surface


of the other ruled system.
cut by
167'.

called a
167!-

guide-straight

A
of

ruled system
its

is

ruled system

is

pro-

any two

guide-straights in

jected from any two of


cils.

its

guide-

projective ranges.

straights in projective axial pen-

For

if

m,

m m
it

be any three guide-straights of the ruled

and in projective ranges the system, the planes on m^ cut joins of whose correlated points are the elements of the ruled be projected axially from t system. Again, if the points on and * the meets of the planes so correlated are the ele1

ments

of the ruled system.

168.

Four
if

straights of a ruled system are called

harmonic

harmonic points by one (and so straights three straights, no two coplanar, by every) guide-straight. By a fourth harmonic is determined lying in a ruled system with
they are cut in four
the given three and on a fourth harmonic point to any three costraight points of the given three.
of one cutting the ruled surface in a straight n of the other ruled system and consequently also in a straight ruled system has in common with the surface no point not on
169.

A plane

For any straight from such a point cutting both these straights would lie wholly on the ruled surface and so therefore would their whole plane, which is impossible. Any third straight coplanar with m and n on their cross has no second point in common with the surface and so
one of these
straights.
;

is

a tangent,

and the plane of

m and n

is

called tangent at their

cross, the point tnn.

102

PROJECTIVE GEOMETRY.

[CHAP. IIL

The number

of planes tangent to the ruled surface

and on

a given straight equals the number of points the straight has in common with the ruled surface, that is two so the ruled
;

surface
170.

is

of the second class.

Project the two generating ranges of a ruled system

from any projection-vertex Fnot on it. The eject consists of two copunctal projective flat pencils. The plane of any two correlated straights is on an element of the ruled system. All
such planes form a cone of planes.
points of contact of these planes with the ruled surface The planes tangent to a ruled surface at are a conic range. the points on its cut with a plane form a cone of planes.
cut of a hyperbolic-paraboloid by a plane not on an element has on it the meets of the plane with the two figu171.

The

The

rative elements,
is

and so
in

on the plane,
is

is a hyperbola except when their cross which case it is a parabola. The figurative

plane

a tangent plane.

172. The planes tangent at the figurative points of a hyperboloid of one nappe are all proper planes, copunctal and forming a cone of planes tangent to the asymptote-cone of the
'
'

hyperboloid.

Each element

to the asymptote-cone

is

parallel

to one element of each ruled system.

Any
it is

nappe cuts

plane not on an element of the hyperboloid of one it in a hyperbola, parabola, or ellipse, according as parallel to two elements, one, or no element of the asympis,

tote-cone, that
rative conic
173.
If

according as

it

has

in

common

with the

figu-

on the hyperboloid two points, one, or no point.

an axial pencil and a ruled system are projective,


in

they generate

general a

'

twisted cubic curve,' which any

plane cuts in one point at least and three at most. For a plane cuts the ruled system in a conic range perspective to it,
of

which

in

general three points at most

lie

on the correspond-

ing planes of the pencil.


174.

The

ruled quadric surface

is

the only surface doubly

Art.

14.]

ruled quadric surfaces.

103
of the

ruled.

two so united ruled systems is one most noteworthy discovered by the modern geometry.*
figure of
175.

The

Let u lt

find the straights crossing four given straights. u % u t be the given straights. Projecting the range on ux from the axes u 2 and u % gives two axial pencils, each
u.t ,
,

To

and consequently projective. The meets of perspective to their correlated planes are all the oo straights on u ,u^, u %r and form a ruled system of which u lt u 2 u z are guide-straights.
t

The two
'

two same kind and on the same bearer

projective axial-pencils cut the fourth straight u K in [Two projective primal figures of the conjective ranges.
'

are called conjective.]


,

If

now
uK
in

a straight

of the ruled system crosses u A

then the two

correlated planes of which this straight

is

the meet must cut

the same point, which consequently is a self-correspondSince there are two ing point of the two conjective ranges. such (the points common to u t and the ruled surface), so there
are

two

given that for the two self-correlated points of two conjective ranges.

straights (real or conjugate imaginary) crossing four Their construction is shown to depend on straights.

This important problem


'

straights,

what

is

common

the four-dimensional space of to four straights ? is the analogue


in
'
' '

the space of points, what is common to three points?' and its dual in the space of planes, what is common to three planes?
of the

problem

in

fundamental diversity, but also, as the inhercompared to points-geometry and planes-geometry,


It

shows not only

their

ently quadratic character of straights-geometry.


Prob. S3- Find the straights cutting two given straights
parallel to a third.

and

Prob. 34. Three diagonals of a skew hexagram whose six sides are on a ruled surface are copunctal. are Prob. 35. If a flat pencil and a range not on parallel planes on the points of the range parallel to the projective, then straights a hypercorrelated straights of the pencil form one ruled system of
bolic-paraboloid.

*See Monge, Journal de

l'Scole polytechnique, Vol. I.

104
Prob. 36.

PROJECTIVE GEOMETRY.

[CllAP. III.

What

is

the locus of a point harmonically separated


?

from a given point by a ruled surface

Art.
176.

15.

Cross-Ratio.

Lindemann has shown how every one number, whether


or irrational,

integer, fraction,

or

may

be correlated to

one point of a straight, without making any use of measurement, without any comparison of sects by application of a unit He gets an analytic definition of the cross-ratio of sect.*
'
'

Then this expression is applied to four copunctal straights. Then is deduced that the number prefour costraight points.
viously attached to a point on a straight is the same as the cross-ratio of that point with three fixed points of the straight. Thus analytic geometry and metric geometry may be founded

without using ratio in its old sense, involving measurement. Thus also the non-Euclidean geometries, that of Bolyai-Lobachevski in which the straight has two points at infinity, and that of Riemann in which the straight has no point at infinity,

be treated together with the limiting case of each between them, tire Euclidean geometry, wherein the straight has one

may

but only one point at

infinity.

Relinquishing for brevity this

pure projective standpoint

and reverting to the old metric usages where an angle is an inclination, a sect is a piece of a straight, and any ratio is a number; distinguishing the sect AC from CA as of opposite
*

sense,' so that

AC =

CA, the

called the cross-ratio of the range

ABCD

where

and B, called conjugate be looked upon as the extremities of a or externally by C and again by D.\
* Von Staudt
in

[AC/BC]/[AD/BD] is and is written [ABCB] of the cross-ratio, may points


sect divided internally

ratio

projective definition of number, and thus

Beitrage zur Geometrie der Lage, 1856-60, determines the makes the metric geometry a consestated in the Mathematical Colis

quence of projective geometry. f The fundamental property of cross-ratio


lections of Pappus, about

is

370 A.D.

The

cross-ratio

the basis of Poncelet's

Traite des propriety projectives, 1822, which distinguishes sharply the projective

and metric properties of curves.

.Art. 15.]

cross-ratio.

105

177. If

un

ABCD

respectively be the straights abed co-

punctal on V, then

A C/BC= A A VC/AB VC or A C/BC = iA V. VC sin (ac)/\B V. VC sin (be). AD/BD = A A VD/ABVD = iA V. VD sin (ad)/bV. VD sin (bd).
[A BCD]

Therefore

[sin (ae)/s'm (&)]/[sin

^/sin (&/)]

Thus

as the cross-ratio of any

on any

pencil u(a/3yd) may transversal.


178.

flat pencil V[abcd] or axial be taken the cross-ratio of the cut ABCD

Two
As
1.

inversely two equicross primal


179.

projective primal figures are 'equicross;' and figures are projective.

approaches the point at


cross-ratio

infinity,

AD/BD
is

ap-

proaches
equals
180.

The

[ABCD] when
ABC,

figurative

A C/BC.
Given three costraight points
to find

D so

that

On any ma)' equal a given number n (-|- or ). on C take A' and B' such that CA'/CB' =n; A' and straight B' lying on the same side of C if n be positive, but on opposite sides if n be negative. Join AA', BB', crossing in V. The For if parallel to A' B' on Fwill cut AB in the required D. D' be the point at infinity on A'B\ and ABCD be projected
[ABCD]
from
V, then

A' B'CD'

is

a cut of the eject

so
,

[ABCD]
181. If

[A' B'CD']

=A

C/B C-n.

then Z\ coincides with Z>. be complete plane perspectives, four figures costraight points (or copunctal straights) in one are equicross with the correlated four in the other. Let O be the center of

[ABCD]
two

= \ABCD^

182. If

any pair of correlated points and N' another pair of correlated points lying on the straight OMM' whose cross with the axis of perspective is X. Then [OXMN] = [OXM'N'].
iV^

Let perspective. of the two figures,

M and M' be

[OM/XM]/[ON/XN] = [OM>'/XM']/[ON> /XN'\ Therefore [OM/XM]/[OM' /XM'] = [ON/XN]/[ON' /XN']. That is, [OXMM'] = [OXNN']; or the cross-ratio [OXMM']
'

That

is,

106
is

PROJECTIVE GEOMETRY.
all

[CHAP. IIL

constant for

pairs of correlated points

M and M' taken


Since

on a straight

OX on
OLL

the center of perspective.

Next

let

and L' be another pair of correlated points and


with the axis of perspective.

the cross of

LM
if

and

L M'
'

OXMM'
is

cross on some point Z of the axis XY, therefore be projected from Z, the cut of the eject by OY
;

is

OYLL'. So \dXMM']-=\_dYLLJ]
constant for
It is called
all

or the cross-ratio

\OXMM'\

pairs of correlated points.

the 'parameter' of the correlation. When the is harmonic, and two i, the range parameter equals correlated elements correspond doubly, are coupled, and the

OXMM'

correlation
183.

is

'involutorial.'

When
is

the correlation

is

involutorial

and the center of

the figurative point on a perpendicular to the perspective axis of perspective, this is called the 'axis of symmetry/ and
the complete plane perspectives are said to be
184.
'

symmetrical.'

When

the correlation

is

involutorial

and the axis of

perspective is figurative, then the center of perspective is called the 'symcenter,' and the complete plane perspectives are said to be symcentral.'
'

With the
to
it.

Prob. 37. In a plane are given a parallelogram and any sect. ruler alone find the center of the sect and draw a parallel

Prob. 38. The locus of a point such that its joins to four given points have a given cross-ratio is a conic on which are the points. Prob. 39. If the sides of a trigram are tangent to a conic, the
joins of two of its fan-points to any point are conjugate with respect to the conic.

on the polar of the third

Prob. 40. If from any point of the sect between the contactpoints of a pair of tangents to a parabola straights be drawn parallel to these tangents, the join of their proper crosses with the tangents
will

be a tangent.

ART.

1.]

CORRESPONDENCE OF POINTS ON CONICS.

10?

Chapter IV.

HYPERBOLIC FUNCTIONS.
By James McMahon,
Assistant Professor of Mathematics in Cornell University.

Art.

1.

Correspondence of Points on Conics.


way
for a general treatment of the hyperis

To

prepare the

given on the relations between hyperbolic sectors. The method adopted is such as to apply at the same time to sectors of the ellipse, including
bolic functions a preliminary discussion

the circle;

the hyperbolic and circular functions will be obvious at every step, since the same set of
of
in

and the analogy

connection with either the hyperbola convenient to begin with the theory of correspondence of points on two central conics of like species, i.e. either both ellipses or both hyperbolas.
or the ellipse.*
It is

equations can be read

To
1

2 B^ conjugate radii of any other central conic of the same species be two points on the curves; and let their coordilet 2
1
, ;

be conjugate

obtain a definition of corresponding points, let radii of a central conic, and O^A 2

O A lt
x

P P
l , x ,

nates referred to the respective pairs of conjugate directions

be (x

jp,),

(x 2

j/2 );

then,

by analytic geometry,
9

~
a,'

'

b?

ai

6,'

W
is

* The hyperbolic functions are not so named on account of any analogy " The with what are termed Elliptic Functions. elliptic integrals, and thence
the elliptic functions, derive their

name from

the early attempts of mathemati-

cians at the rectification of the ellipse.


.

... To

a certain extent this

because we employ the name hyperbolic function to de. disadvantage; note cosh u, sinh u, etc., by analogy with which the elliptic functions would be ." (Greenhill, Elliptic merely the circular functions cos <p, sin cp, etc.
. .
.

Functions,

p. 175.)

108

HYPERBOLIC FUNCTIONS.
if

[Chap. IV.

Now

the points

P P
1
,

be so situated that
(2)

a,'

b*

the equalities referring to sign as well as magnitude, then If are called corresponding points in the two systems. 3

P
Q

P
Q

be another pair of correspondents, then the sector and

tri-

angle

O Q are said to correspond respectively with the These definitions will apply also sector and triangle 7 O^Q^. then referred when the conies coincide, the points x 2 being to any two pairs of conjugate diameters of the same conic.
x
x

P P
,

In discussing the relations between corresponding areas it convenient to adopt the following use of the word " measure": The measure of any area connected with a given central conic is the ratio which it bears to the constant area of the triangle
is

formed by two conjugate diameters

of the

same
x

conic.
x
x

For example, the measure

of the sector

A OP

is

the ratio

sector
triangle

A OP A OB
x

Art.

3.]

areas of corresponding sectors.


to be

109

and
x x

is
x

A OP

and

A OB
x
x

regarded as positive or negative according as are at the same or opposite sides of their
x

common

initial line.

Art.

2.

Areas of Corresponding Triangles.


ii
i

P Q be (x ,y (x ',y '\ and let those of their correspondents P g be (xit y ) (x^y/); let the PfiJ2i De T T and let the measuring tritriangles P O Q A O B A O B be K K and their angles aolt g? angles
For,
let

The <

areas of corresponding triangles have equal measures. wi immjSUbww u hii mm H M W mm m


i
ii

iii

i i

ii

the coordinates of

x ,

),

lt

it
%

2 ;

then,

and

analytic geometry, taking account of both direction of angles, areas, and lines,

by

magnitude

T\

_ \{xy - x;y )sm


f
x
x

gd x

_ x,
a
i

K\
7\

i^A sm

y^ __ */ y^
&i
i

'

&i
l

_ j(xjr 'xM
t

sin

a>,

_ x\
a2

y/
bt

_ x^^
a t b^
(3)

\<*J>* sin g? 2

Therefore

J3.

= ~\

Art.

Areas of Corresponding Sectors.

The
For

areas of corresponding sectors have equal measures. conceive the sectors Sl9 S, divided up into infinitesimal
;

corresponding sectors then the respective infinitesimal corre-. sponding triangles have equal measures (Art. 2) but the of these infinitesimal given sectors are the limits of the sums
;

triangles,

hence

^ ^ k = k;
x x x

(4)

ures

In particular, the sectors A O Pxs A,0,P2 have equal measfor the initial points A l9 A, are corresponding points.
;

It

may be proved
if

conversely by an obvious reductio ad


initial

points of two equal-measured sectors correspond, then their terminal points correspond.

absurdum that

the

Thus

if

any

radii

OA
x

x ,

O^A, be the

initial

lines of
x x,

two
t

equal-measured sectors

whose terminal

radii

are

OP OP

110

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

are corresponding points referred respectively to the pairs of conjugate directions Ot lt O B and vA v O^B^;
xJ
2

tken

P P

x ,

that

is,

*i

y:

y*

Prob.

i.
x

Prove that the sector

to the mid-point of joining spectively, to the median as

P\QV

POQ
x x

is

bisected by the line

(Refer the points


axis of

P Q
lt

re-

common

opposite conjugate directions as axis of y, are then corresponding points.)


Prob.
Prob.
2.

x, and to the two and show that Q

x ,

Prove that the measure of a circular sector


its

is

equal to

the radian measure of


3.

angle.
elliptic

Find the measure of an

quadrant, and of the

sector included by conjugate radii.

Art.

4.

Characteristic Ratios of Sectorial Measures.


x

Let

A OP =
x x

PM
X

vS,
x

be any sector of a central conic; draw


x ,

let
x

O
x

M = x MP
x

ordinate to
xi

OA
X

i.e.

parallel
x
x

to

the

tangent

at

=y,, O A =#,, and

OB

bx

then the ratios

xja yjb
x
,

the conjugate radius are called the charac. x


x

the given sectorial measure S /K These ratios are constant both in magnitude and sign for all sectors
teristic ratios of

of the

same measure and species wherever these may be

situ-

Hence there exists a functional relation beated (Art. 3). tween the sectorial measure and each of its characteristic
ratios.

Art.

5.

Ratios Expressed as Triangle-measures.

and its complementary triangle are measured by the two characteristic ratios. For, let the triangle A O P and its complementary triangle P O B be denoted by
triangle of a sector
x
x

The

x ,

7V; then

K-

^^
l

sino?

_^

(s)

ART.

7.]

FUNCTIONAL RELATIONS FOR


6.

ELLIPSE.

Ill

Art.

Functional Relations for Ellipse.


its

The
for
all

functional relations that exist between the sectorial


characteristic
ratios

measure and each of


elliptic,

are

the same

in-

cluding circular, sectors (Art. 4).

Let/*,,

be corresponding points on an ellipse


2

and a

circle, referred
dix

to the conjugate rections O A O


x

and

A0,B^,

right angles

let

the angle Afi^P^

the latter pair being at 6 in radian measure; then

K,

1/7

'

(6)

= cos -1
at
hence, in the
ellipse,

sin

bn

[*.

= *,

by Art.
-^

cos

= sin
measure of the
4, b x

K,
Prob.
4.

(7)

Given x

\a\\ find the

A1O1P1.
Prob.

Also find
5.
is

its

area

when a

3, a?

elliptic sector

6o.
whose

Find the characteristic


\n. Write

ratios of

an

elliptic sector

measure
Prob.

6.

down

the relation between an elliptic sector and


5.)

its triangle.

(See Art.

Art.

7.

Functional Relations for Hyperbola.


the case of the hyperbola
S,
-XI,

The

functional relations between a sectorial measure and

its characteristic ratios in

may

be

written in the form


,
,

cosh

= sinh
two

;
lines

and these express that the


right.

ratio of the

on the

left is

a certain definite function of the ratio of the two areas on the

These functions are

called

by analogy the hyperbolic

112

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

cosine ana the nyperbolic sine.

Thus, writing u for


V
'

S /Kv
x

the

two equations

X _

cosh u

7^ =

sinh u

(8)

serve to define the hyperbolic cosine and sine of a given sectoand the hyperbolic tangent, cotangent, secant,, rial measure u and cosecant are then defined as follows
;

tanh u

= =

sinh u

cosh&

7,
,

coth u

cosh&
sinh
i

w
(9>
>.

sech u

csch u

cosh u

sinh u

The names
"
h-sine," M

of these functions

may

be read " h-cosine,"'

h-tangent," etc.

Art.

8.

Relations between Hyperbolic Functions.

the six functions there are five independent relations, so that when the numerical value of one of the functions Four of is given, the values of the other five can be found.

Among

these relations consist of the four defining equations fifth is derived from the equation of the hyperbola

(9).

The

< K
giving
cosh*
11

'

sinh &

1.

(10}

By

a combination of
2

some

of these equations other subsidi(10) successively

ary relations

by cosh

u,

may.be obtained; thus, dividing sinh u, and applying (9), give


1
2 tanh u

coth u

= sech = csch

u.

Equations

serve to express (9), (10), (11) will readily

the

value of any when tanh u is given,

function in terms of

any other.

For example,
2

coth u

tanh u

sech u

=J

tanh

u,

Art.

8.]

relations between hyperbolic functions.


i

113

cosh u

=
V
i

I
,

sinh u

=
y
I

tanh u
,

tanh #
tanh 2 # tanh u

tanh #

csch u

a/

The ambiguity in the sign of the square root may usually be removed by the following considerations The functions cosh u, sech u are always positive, because the primary char:

acteristic ratio

x /a
x

the abscissa

M
x

is

positive, since the initial line

OA
x

and

are similarly directed from

on which-

ever branch of the hyperbola may be situated; but the functions sinh u, tanh u, coth u, csch u, involve the other characx

is positive or negative according as or opposite signs, i.e., as the measure y and b x have the same u is positive or negative hence these four functions are either Thus when any one of the funcall positive or all negative.

teristic ratio
x

y /b
x

which

tions sinh u, tanh


sign, there
is

?/,

csch

u,

coth
in

no ambiguity
;

given in magnitude and the value of any of the six


ti,

is

but when either cosh u or sech u is hyperbolic functions is ambiguity as to whether the other four functions given, there
shall

be

all

positive or

all

negative.
as the ratio of

The hyperbolic tangent may be expressed


two
lines.

For draw the tangent


/;

line^C=
tanh u

then

=y
b

114

HYPERBOLIC FUNCTIONS.

[Chap. IV.
u.

Prob. 7. Express all the hyperbolic functions in terms of sinh Given cosh u = 2, find the values of the other functions.

Prob.

8.
r,

Prove from eqs.


sech u

10,

n,

that

cosh&>

sinh

u,

cosh#>i,

tanh u

<

<

1.

Prob.

and area of sector A OP = 3; find the sectorial measure, and the two characteristic ratios, in the elliptic sector, and also in the hyperbolic sector; and find the area of the triangle A OP. (Use tables of
cos, sin, cosh, sinh.)

9.

In the figure of Art.

1, let

OA = 2 OB=i, AOB =
,

6o,

Prob. 10. Show that coth u, sech u, csch u may each be expressed as the ratio of two lines, as follows: Let the tangent at make on the conjugate axes OA, OB, intercepts OS n\ m, in R, let the tangent at B, to the conjugate hyperbola, meet

OT =

OP

making BR =

I;

then

coth u
Prob.

l/a,

sech u

m/a,

csch u
is

n/b.
u.

n. The measure

of segment

AMP

sinh u cosh u

Modify
8, 10.

this for

the ellipse.

Modify

also eqs.

10-14, an d probs.

Art.

9.

Variations of the Hyperbolic Functions.

Since the values of the hyperbolic functions depend only

on the sectorial measure,


half

it

is

ations,

tracing their varito consider only sectors of one

convenient,

in

rectangular hyperbola, whose and to take the conjugate principal axis OA as the common initial
of a
radii are equal,
line

of

all

the

sectors.

The

sectorial
00,

measure u assumes every value from


through
o,

to

-f-

00

as the terminal point

comes in from infinity on the lower branch, and passes to infinity on the upper
branch
;

that

is,

as the terminal line

OP

tion

y
is

x,

swings from the lower asymptotic posiIt is here assumed, to the upper one, y = x.
the sector

but
as

proved

in Art. 17, that

A OP

becomes

infinite

P passes

to infinity.
u,

Since the functions cosh

sinh u tanh u, for


y

any position

ART.

9.]

VARIATIONS OF THE HYPERBOLIC FUNCTIONS.


are equal to the ratios of x, y, evident from the figure that
t,

115

of
a,

0P
it is

to the principal radius

cosh

o=l,
become

sinh

0,

tanh

0=0,
//,

(15)

and that

as u increases towards positive infinity, cosh


infinite,

sinh u

are positive and as a limit thus


;

but tanh u approaches unity

cosh

go

00

sinh

00

co

tanh

00

1.

(16)

Again, as u changes from zero towards the negative side, cosh u is positive and increases from unity to infinity, but
sinh u

negative and increases numerically from zero to a and tanh u is also negative and increases negative from zero to negative unity hence numerically
is

infinite,

cosh

00)

= 00

sinh

(-

00)

00,

tanh

00

1.

(17)

For intermediate values

of u the numerical values of these


16, 17,

functions can be found from the formulas of Arts.

and

are tabulated at the end of this chapter. general idea of their manner of variation can be obtained from the curves in

abscissa,

Art. 25, in which the sectorial measure u is represented by the and the values of the functions cosh u, sinh u, etc.,

are represented by the ordinate.

The
Thus

relations

between the functions

of

u and of u are
8.

evident from the definitions, as indicated above, and in Art.

cosh
sech

() = + ()= +
(

cosn u
sech

>

smn
csch

(~~ u )

~~

sm ^ u
csch

>

?/,

tanh

u) =

()=
(

u,

>
J

(18)

tanh

u,

coth

= u)

coth

u.

from
u,

Prob. 12. Trace the changes in sech u> coth u, csch u, as u passes 00 to -}- 00 Show that sinh #, cosh u are infinites of the
.

same order when u


cosh u are
of u.)

is infinite.

(It will

infinites of

an order

appear in Art. 17 that sinh infinitely higher than the order

Prob. 13. Applying eq. (12) to figure, page 114, prove tanh
tan

u,

A OP.

116

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

Art.

10.

Anti-hyperbolic Functions.

The equations

cosh

u,

y -r

t
//,

sinh

tanh

u,

etc.,

may
u

also be expressed

by the inverse notation u


t

cosh -1

x
-,.

sinh

_1

y
,

tanh -1 -T,

etc.,

which may be read: "u

is

the sectorial measure whose hyperbolic cosine is the ratio a" etc. or " u is the anti-h-cosine of x/a," etc.
;

to

Since there are two values of

u,

correspond to a given value of cosh u,

with opposite signs, that it follows that if u be

determined from the equation cosh u = m, where m is a given number greater than unity, u is a two-valued function of m. The symbol cosh" 1 tn will be used to denote the positive value
of

u that

satisfies
1

symbol sech"
satisfies

the equation cosh u m. Similarly the will stand for the positive value of u that

///. the equation sech u -1 _1 coth -1 functions sinh //z, tanh f,

The
;;/,

signs of the
1

other

csch" ;, are the

same

as the sign of m. Hence all of the anti-hyperbolic functions of real numbers are one-valued.
Prob. 14. Prove the following relations:

cosh" 1

/^

sinh" 1

Vm*

1,

sinh" 1

//*

=
,

cosh" 1 Vm*

-f-

1,

the upper or lower sign being used according as _1 1 cos" Modify these relations for sin negative.
.

is

positive, or

2,OB i,AOB 6o; find 15. In figure, Art. i,let OA the area of the hyperbolic sector AOP, and of the segment y -1 is 3. from the tables for cosh.) if the abscissa of (Find cosh
Prob.

AMP

Art.
(a)

11.

Functions of Sums and Differences.

To
sinh

prove the difference-formulas sinh u cosh v cosh (// v)

cosh (u

= = cosh u v)

//

sinh v,
v.

cosh v

sinh u sinh

I
)

09)

Let

OA

AOQ
sector

be any radius of a hyperbola, and let the sectors AOP, v is the measure of the have the measures u, v\ then u

be the radii conjugate to OA, OQ; Let OB, the coordinates of P, Q, Q' be (x j/,), (x, y), (x' y') with reference to the axes OA, OB; then

QOP.

OQ

and

let

ART.

11.]

FUNCTIONS OF SUMS AND DIFFERENCES.

ikr
117
[Art.
5.

sinh

= sinh _

sector

^
sin
z>

= tria "g le ggg

K-^Ji

^)
x

<

_
~~

j^i
x

yx
x
;

\a x b sin a?

b ax

b ax

= sinh

& cosh

cosh u sinh #

cosh (u

^)

cosh

SeCt r

^P =
y
b:

triangle

/W
bx

[Arfc

$>

KxyyS) sin
Jtf,^ sin g?

&?_

y ^ _#<
^
a
x

a;
(20)

but

since Q, Q' are extremities of conjugate radii

hence
z>.

cosh (u

v)

= cosh u
is

cosh ^

sinh u sinh
is

In the figures u

positive

and v

positive or negative.

Other figures
in

may may

be drawn with u negative, and the language


all.

the text will apply to

In the case of elliptic sectors,


1

similar figures

be drawn, and the same language win apply, that the second equation of (20) will be x'/a y/b except
x

x;

therefore
sin (u

cos {u
(b)

= sin u cos v = cos u cos v v)


v)

cos u sin
-f-

v,
v.

sin

u sin

To

prove the sum-formulas

sinh (u

cosh

+ v) sinh u cosh v + cosh u sinh + v) = cosh a cosh v + sm ^ w smh v


(19)

v,
-

(21)
S

These equations follow from

by changing v into

118

HYPERBOLIC FUNCTIONS.
for sinh
(

[CHAP. IV.

and then

v) t

cosh

v),

writing

sinh

v,

cosh v

(Art. 9, eqs. (18)).


(c)

To prove v
(

that tann

-= (uv) = itanhtanhv
v
'

(22) ; V

Writing tanh

s>)

=s

cosh (u

v,

expanding and dividing


z/,

v)

numerator and denominator by cosh & cosh


tained.

eq. (22)

is

ob-

Prob. 16. Given cosh u

=
.

2,

cosh

3,

find cosh (u

-\- z>).

1.

2.

3.

Prob. 17. Prove the following identities: 2 sinh u cosh sinh 27/ 2 2 2 1 -f- 2 sinh u 2 cosh u cosh 211 sinh ^ cosh'tt 2 2 1 2 cosh -#, cosh u 2 sinh -. 1 cosh u

= =

+
,
,

+
z/

1.

4.

tanh }*
.

=
1

sinh

5.
6.

sinh 2U

sinh 3#

7. '

cosh

-4'

= 2 -_ cosh 2U tanh u = 3 sinh w 4 sinh cosh i-i. u = + anh i* sinh t^\~r, ,

cosh u tanh u

= cosh

=
1

/cosh :

i\*
.

sinh u

\cosh u -j2 i-4- tanh u


.

1/

tanh u
3

-j-

= 4 cosh'tf

3 cosh ^.

t,

tann -#

sinh z;)=cosh (& -f- v) sinh (# -jsmn )(cosh v (cosh u=v= 9. Generalize (8); and show also what it becomes when a 2 2 2 2 sin 7. sinh ^: cosh a: sin v 10. sinh ^: cos'_v
8.

v).
.

11. 12.

cosh -1 /#
sinh
_1

cosh _1
sinh
-1
>z

==

cosh~ 1 \_mn
_1

y(m
1

i)(tz

i)J.

sinh

[w y
?

yi + *?i
(22),.

Prob. 18.

What

modifications of signs are required in (21),

in order to pass to circular functions

Prob. 19. Modify the identities of Prob. 17 for the same purpose.

Art.

12.

Conversion Formulas.

To

prove that
cosh
cosh u

j+ cosh u
1

2 cosh \{u

-\-

u 9 ) cosh \(u
&,)
2)

u 7),
9 ),

cosh u t
-j-

2 sinh \{u
x

sinh #, sinh #,

sinh

it,

sinh u %

= 2 sinh \{u = 2 cosh \{u

+ + +&

sinh |(#,~

cosh J-(,
sinh
(?/,

# a ), # a).
^

a)

Art.

13.]

limiting ratios.
it

119

From

the addition formulas

follows that
v)
z/,

and
v

= 2 cosh & cosh cosh + v "" cosn (&?>) = 2 sinh & sinh v, = 2 sinh & cosh v, sinh (& sinh ( ^) 2 cosn w smn ^> sinh (# + z ~~ smn (^ ~ *0 v = u %t u = \{u + then by writing u-\-v = u iy u
cosh (u
(

-f-

v) -J)

cosh (u

-f-

-f-

z^)

a ),

(,

# 2 )>

these equations take the form required.

Prob. 20. In passing to circular functions, show that the only modification to be made in the conversion formulas is in the algebraic sign of the right-hand member of the second formula.

cosh 2U ... Prob. 2i. Simplify -r. r J

smn
2

+ cosh av r:

cosh 2U
,

-f-

cosh av

-f sinh av
2

cosh

cosh av
y).

Prob. 22. Prove sinh

.*
2

sinh
.*
2

.y
2

sinh (x -\-y) sinh (x sinh


2
2

Prob. 23. Simplify cosh Prob. 24. Simplify cosh

cosh
2

sinh ^.
2
ji>.

cos y

+ sinh * sin

Art.

13.

Limiting Ratios.
approaches zero, of
tanh u
u
in form.

To

find the limit, as u

sinh u

which are then indeterminate

By

eq. (14), sinh

u> u>

tanh u

and

if

sinh u and tanh u


inequalities,
it

be successively divided by each term of these follows that , Jrr ^ L vJCk*


1

^ <

sinh u ^

<
u

cosh u

sech u

<

tanh U

<V

but when u-^zO, cosh u


lim. sinh

= _

I,

sech

=: i,Tience

Hm. tanh ^
u
==.0

1.

u == o

(24)

\V

120

HYPERBOLIC FUNCTIONS.

[Chap. IV.

Art.

14.

Derivatives of Hyperbolic Functions.

To

prove that
^(sinh
(a)
ft)

du
d(cosh
ft)

= =

cosh

ft,

(*)

du
d(tanh
ft)

sinh

ft,

w
(rf)

du
<^(sech
2/)

= sech
=

2/,

(25)

du
d(cotri
ft)

sech u tanh

ft,

w
(/)
(<z)

du
<f(csch
ft)

csch

ft,

du
ft,

csch u coth

ft.

Let y

= sinh + ^ w ~~ smn * jy = sinh = 2 cosh \{2u ^) sinh \Au, .sinh JJ Ay -~ = cosh -f $-;
(ft

-j-

.,

ZJft)

Take the

limit of

both

sides, as

Au

= o,
ft)
'

and put

lim

lim.

dy = Hi = </(sinh ^ cosh + i^) = cosh

Ay

(ft

ft,

sinh
lim.

\Au 7 A iAu

(see Art. 13)

dT(sinh u)

du

= cosh

ft.

()

Similar to

(a).

dftanh u) ~
dfa

d
'

sinh

2/
ft

<& cosh
cosh u
2

sinh
8

8
ft

cosh u

cosh

=
?/

sech*

Art.

15.]

derivatives of hyperbolic functions.


Similar to
(c).

121

j(V)
.

u w ~^^ = ^-^ii^ = -^ih^ = - sechatanh


^(sech u)
i

sinh

"-

(/)

Similar to

(e).

It thus appears that the functions sinh u, cosh u reproduce themselves in two differentiations and, similarly, that the circular functions sin u, cos u produce their opposites in two
;

differentiations.

In this connection

it

may be

noted that the

frequent appearance of the hyperbolic (and circular) functions in the solution of physical problems is chiefly due to the fact
that they answer the question What function has its second derivative equal to a positive (or negative) constant multiple of the function itself ? (See Probs. 28-30.) An answer such as
:

y = cosh mx is not, however, to be understood as asserting that mx is an actual sectorial measure and y its characteristic ratio but only that the relation between the numbers mx and y is the
;

same

as the

known
and
its

relation

bolic sector

characteristic ratio

between the measure of a hyperand that the numerical


;

value of y could be found from a table of hyperbolic cosines.


Prob. 25.

Show

that for circular functions

the only modifica-

tions required are in the algebraic signs of

(b), (d).

Prob. 26.

Show from

their derivatives

which of the hyperbolic

and

circular functions diminish as u increases.

Prob. 27. Find the derivative of tanh u independently of the derivatives of sinh u, cosh u.
Prob. 28. Eliminate

equation^ = A cosh mx + B sinh mx, and prove that

the constants by differentiation from the

dy/dx

= m?y.

Prob. 29. Eliminate the constants from the equation

and prove
tial

=A that d^y/dx' =
y
down

cos

mx

-f-

B sin

mx,

my.

Prob. 30. Write

the most general solutions of the differen-

equations
d*v

d*y

d*y

122

HYPERBOLIC FUNCTIONS.
15.

[CHAP. IV.

Art.

Derivatives of Anti-hyperbolic Functions.


-1

^(sinh
(a)

x)

dx
l

Vx*
x)
__
i

+i

w
(A W

d(cosh.~

dx
1

^(tanh- x)

"1

dx
4coth~ ^) _ dx
-1
1

- *_]*<,'
(26)
i

(d\ 1 ^

-i

?3TJ^
i

w
,

^(sech

*)

<r
-1

* Vi
#)
I

-.

^(csch

x Vx*
(#)

+
&,

Let

ss sinh"" #,

then

= sinh

dx

= cosh

</

=
{b)

Vi

+ sinh
&

udu

Vi

-\-

x* du,

du

= <&:/ VT+1?.
2

Similar to

(a).

(c)

Let

tanh" x,
2

then

.r

(i

tanh

&)afo

(i

= tanh u, dx = seen u du du = dx/i x*. x*)du,

(d) Similar to

(c).

("

4f:, =sK)=7/&-')'%iij(e).
1

(/) Similar to
Prob. 31. Prove
^(sin" x)

_
Vi
_ ~
1 1

d(cos~

x)_
Vi

dx
^(tan
-1

1'

dx
1

x*

x)

djcot- x)

dx

+ x*

dx

+x*'

Art.

16.]

expansion of hyperbolic functions.

123

Prob. 32. Prove


tfsinh

_,x

dx
.

^cosh

x -

dx

</tanh

x
a

-=
a'

adx
l

~|
,

x 5 _\x <a
1

</coth-

# -=
a

dt^tr

"1

xs

a Jx>m

Prob. 33. Find ^(secli" x) independently of cosh -1 *.

x is real, prove that coth* Prob. 34. When tanh and conversely; except when x 1. nary,

-1

is

imagi-

Prob. 35- Evaluate

sinh"

cosh

-1

x
,

-^-,

when x

cc

Art.

16.

Expansion of Hyperbolic Functions.

For

this

purpose take Maclaurin's Theorem,


/'(o)

/()

= Ao) +

+2

',

y"(o)

+ j,

/'"(o)

+
.

.,

and put f(u)


then

= sinh u, f{u) cosh &, f"(u) = = 0, /'(o) = cosh o = /(o) = sinh


1,
.

sinh
.;

,...,

hence

sinh u

=u+
2

u*

+ -j u"
4
:

-f-

(27)

and

similarly, or

by

differentiation,

cosh u

& H 214! a +
r

(28)

By means
cosh

of these series the numerical values of sinh u y

u, can be computed and tabulated for successive values of the independent variable u. They are convergent for all values

of #, because the ratio of the nth term to the preceding

is

in

the

first

case
2){2n

u /{2n
>

\){2n

2),

and

in

the second
less

case

u*/(2n

3),

both of which ratios can be made

than

unity by taking n large enough, no matter what value u


have.

may

124

HYPERBOLIC FUNCTIONS.

[CHAP. IV.
:

From

these series the following can be obtained by division

tanh u
sech u

u
i

\u

-j-

T jU

-f-

j^V*'
6

= - u* + ^u* TVo^ coth & = + i^ y* + -gls^ 8U CSCh W = - -V + J ^ - T AVtf'+


i
3
4

+ +

.,

(29)
i

These four developments are seldom used, as there is no observable law in the coefficients, and as the functions tanh u, sech u, coth u, csch &, can be found directly from the previously

computed values

of cosh u, sinh u.

Prob. 36. Show that these six developments can be adapted to the circular functions by changing the alternate signs.

7^
cosh u

Art.

17.

Exponential Expressions.
(27), (28)

Adding and subtracting


1

give the identities


u*

+ sinh u = + u 4- 2!
sinh u
1

ri?

3!

+ 4! u* +
%

u
,

cosh u

uA

-1? 2!

-u*

3!.

+ r^* ^4!
u
(>

e~

u
,

hence cosh u
tanh u

J(^

+ e~%
e
u,

sinh #

e~

u
),

= eu

sech u

2
e -\-e'
u

r
u,

(3)

etc.

e -\-e~

The

analogous exponential expressions for sin

u,

cos u are

cos u

\e +e~
e
ui

ui

ui
),

sin

(***

e~

ui

(i

=V

1)

where the symbol


jc in the

stands for the result of substituting ui for

exponential development

This

will

be more
28, 29.

fully

explained in treating of complex

numbers, Arts.

Art.

18.]

expansion of anti-functions.

125

Prob. 37. Show that the properties of the hyperbolic functions could be placed on a purely algebraic basis by starting with equations (30) as their definitions
;

for example, verify the identities

sinh

(
2

u)
1,

sinh u, sinh (u
,

cosh

u)

cosh
-j-

u,

cosh u
2

sinh

u=

+ v) = sinh
^
2

u cosh v

cosh u sinh v%
.

^ (cosh

Tj

mii)

= tn

cosh mu,
sinh

(sinh

mu)
'-

=m

sinh mu.

Prob. 38. Prove (cosh u


2

-f-

)*.

= cosh

nu -f sinh nu.
u,

Prob. 39. Assuming from Art. 14 that cosh


differential

written

how

to

= Aeu + Be~ u where A, B are arbitrary constants show determined, B in order to derive the expressions for cosh u y
y
,
;

equation d y/du*

= y,

sinh u satisfy the

whose general solution may be

sinh u, respectively.

[Use

eq. (15).]

Prob. 40. Show how to construct a table of exponential functions from a table of hyperbolic sines and cosines, and vice versa.

Prob. 41. Prove u

log,,

(cosh u

-f-

sinh u).
is

Prob. 42. Show that the area of any hyperbolic sector when its terminal line is one of the asymptotes.

infinite

Prob. 43.
2

From

the relation 2 cosh u

= eu + e~ u prove
(n

n_1

(cosh #)

=cosh nu+n cosh {n2)u-\-\n{n\) cosh

4)^4-..

.,.

and examine the last term when n is odd or even. Find also the corresponding expression for 2* _1

(sinh u)

Art.
.

18.
-1

Expansion of Anti-Functions.
x) j

Since

tdTsinh

7tt^ = _Iy + Ii 2
I

fw = (i+ "
,

9N

24

,._I3I je 246

> .

hence, by integration,
sinh

23' 245

z T"T" 2467
1

>

V3

/'

the integration-constant being zero, since sinh" x vanisheswith x. This series is convergent, and can be used in compu-

126
tation, only

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

x >!,
^sin

is

when x < I. Another series, convergent when obtained by writing the above derivative in the form

dx
i i

2 X*

4_I3I_I3iI, T 2 4^ 2 46/ r,
2 t

"I

"J'
.

.-.

sinh"

C+log x+- 2x 2

~l
2

4 4**

,+i 2

1 4 6

3.

i--.
6

(32) y

6;tr

where C is the integration-constant, which will be shown in Art. 19 to be equal to loge 2.

development

of similar

form

is

obtained for cosh

-1

x; for

rf(cosh-x)

_t=1 /

_iW

-ir14.II4.II-L4.II 5 16T 1 4. *L T 2^ T 24j4T 246j "J'


hence

eosh-^C+log.-I^-Ii^-Ilii-...,

(33)

In in which (7 is again equal to log, 2 [Art. 19, Prob. 46]. -1 order that the function cosh ^ maybe real, x must not be
less

than unity; but when x exceeds unity, this series vergent, hence it is always available for computation.

is

con-

Again,
*

^^
tanh"
1

^-1+^ + ^4 *+-.,
9

and hence

= x+-*
3

-f \x*
5

I* +7 +
7

(34)

From
sech
-1

are (32), (33), (34)

derived:

= cosh -1

x
x
x*

=C

log 5

I.3.* -

2.2

2.4.4

\ 2.4.6.6

1.3. 5. X D

9
. .
.

, \ (35) K0DJ

ART. 19

LOGARITHMIC EXPRESSION OF ANTI-FUNCTIONS.

127

csch

-J

= sinh-

IJ_-lI.1J_I!I_L_i
2 $x* 2

4 $x

246

jx
;

= C _,og * + _-_l_ + _ l_l__... :


coth- *

(36)

= tanh- I = I +
Show Show

-L

+ JL. +

-L

..

( 37)

Prob. 44. Prob. 45.

that the series for tanh -1 x, coth -1 x, seen -1 x, that one or other of the two developments of the
is

are always available for computation.

inverse hyperbolic cosecant

available.

Art.

19.

Logarithmic Expression of Anti-Functions.

Let
therefore

cosh
1,

u,
I

then

vx

x-\- \ x
u,

cosh u

-f-

= sinh u; sinh u = eu
I
,

and
Similarly,

= cosh" x, log {x T sinh -1 ^ = log (* + Vx ~+ 1).


l

-j-

Vx

i).

(38)

(39)

Also

1 sech -1 ;r= cosh" - == log

-V"
!

(40)

csdi"

sinh" 1 !

log

+ Vl + x
X

l.

(41)

Again,

let

128

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

Prob. 47. Derive from (42) the series for tanh" 1 .*; given in (34). Prob. 48. Prove the identities:

log*=2tanh"
log sec
log tan

=tanh _1
-1
2

=sinh" ^(^

^ _1)=cosh" l(^+^ _1 );
1
'

x
x

== 2

tanh

tan \x\ log esc

2 tanh'
2jc

tan*(ar

-j-

ix);

tanh -1 cos 2^
20.

sinh -1 cot

cosh -1 esc 2X.

Art.

The Gudermanian Function.


of sectors of the
It is

The correspondence
cussed in Arts. 1-4.

same

now convenient

species was disto treat of the


different

correspondence that
species.

may
,

exist

between sectors of

v a on any hyperbola and ellipse, are said to with reference to two pairs of conjugates O A correspond

Two
1 ,

points

P P
,

,.

and O^A %

respectively,

when
(44)

xja
and when
jj/ x
,

= ajxv
sign.

jj/ 2

have the same

The

sectors

A O Piy
x
x

are then also said to correspond.

sectors of central conies of different

Thus corresponding species are of the same

sign and have their primary characteristic ratios reciprocal. Hence there is a fixed functional relation between their re-

The elliptic sectorial measure is called spective measures. the gudermanian of the corresponding hyperbolic sectorial
measure, and the latter the anti-gudermanian of the former. This relation is expressed by

or

= gd S,/K, v = gd u, and u = gd
SJK,

-1

#.

(45)

Art.

21.

Circular Functions of Gudermanian.


terms
its

The

six hyperbolic functions of u are expressible in

of the six circular functions of

gudermanian
cos
v,

for since

=
a,

cosh

u,

=
a,

(see Arts. 6, 7)

in

which

u,

v are the measures of corresponding hyperbolic

and

elliptic sectors,

Art.

22.]

gudermanian angle.
cosh u
sinh u
sec
*

129

hence

v,
1

[eq. (44)]

= VsecV = tan v tanh u = tan v/szc v = sin v, coth u = esc v, sech u = cosz/, csch u = cot z\
The gudermanian
if

(46)

is

sometimes useful

in

computation

for

sinh u be given, v can be found from a table of instance, natural tangents, and the other circular functions of v will give the remaining hyperbolic functions of u. Other uses of this

function are given in Arts. 22-26, 32-36.


Prob. 49. Prove that gd u

Prob. 50. Prove

Prob. 51. Prove

= sec _1 (cosh u) = tan _1 (sinh u) = cos -1 (sech u) =sin~ (tanh u), -1 -1 gd~ v = cosh (sec v) = sinh (tan v) = sech" (cos v) = tanh _1 (sin v). = ^n, gd o = o, gd 00 = Jzr, gd( 00) gd" 0=0, gd-^iar) =00, gd- (-l ) =
1

7r

Prob

52.

Show

that gd u

and gd~

x z'

are

odd functions of

u, v.

Prob. 53. tion tanh \u

From

the

first

= tan \v.

identity in 4, Prob. 19, derive the rela-

Prob. 54. Prove

tanh" J (tan u)

-J

gd

2fc,

and tan _1 (tanh ^)

= ^ gd _1 2^.

Art.
If

22.

Gudermanian Angle

a circle be used instead of the ellipse of Art. 20, the gudermanian of the hyperbolic sectorial measure will be equal
to the radian measure of the angle of the corresponding circular This angle will be sector (see eq. (6), and Art. 2, Prob. 2).

gudermanian angle but the gudermanian function v, above defined, is merely a number, or ratio and this number is equal to the radian measure of the gudermanian angle 6, which is itself usually tabulated in degree measure thus
called the
;

as

i%ov/7t

(47)

130

HYPERBOLIC FUNCTIONS.

[Chap. IV.

ed

Prob. 55. Show that the as follows:

gudermanian angle of u may be construct-

Take the

principal radius
.

OA

initial

of an equilateral hyperbola, as the as the terminal line, and

OP

line, of

the sector whose measure

is

u;

from M, the foot of the ordinate of P, draw tangent to the circle whose diameter is the transverse axis;

MT
is

then

AOT

Prob. 56.

the angle required.* Show that the angle


.

never exceeds 90

Prob. 57. The bisector of angle M bisects the sector (see Prob. 13, Art. 9, and Prob. 53, Art. 21), and the line AP. (See Prob. 1, Art. 3.) Prob. 58. This bisector is parallel to TP, and the points T,

AOT

A OP

are in line with the point diametrically opposite to A.

Prob. 59. The tangent at ordinate of T, and intersects Prob. 60. The angle

P
is

TM on the tangent A. APM half the gudermanian angle.


at

passes through the

foot of the

Art.
Let
then

23.

Derivatives of Gudermanian and Inverse.

= sec v = sec v tan vdv = sec vdv =


v
_1

gd

u,
u,

gd

_1

v,

cosh

sinh u du,
du, sec vdv.
(48)

therefore

^(gd

v)

Again,
therefore

dv
d(gd u)
sinh u

= =

cos v du

sech u du

sech u du.

(49)

Prob. 61. Differentiate:

y
y

= =

gd

u,
-f-

tanh u sech u

gd

u,

y y

= =

sin v

+ gd"

v,

tan v sec v

-f-

gd

-1

v.

* This angle was called by Gudermann the longitude of u, and denoted by lu. His inverse symbol was It; lnus u H(^)' (Crelle's Journal, vol. 6, 1830.)

Lambert, who introduced the angle 0, named it the transcendent angle. (Hist, Hottel (Nouvelles Annales, vol. 3, 1864) de l'acad, roy de Berlin, 1761). called it the hyperbolic amplitude of u, and wrote it amh u, in analogy with the Cayley (Elliptic amplitude of an elliptic function, as shown in Prob. 62.
Functions, 1876)

made

the usage uniform

by attaching
in

to the angle the

name

of the mathematician

who had used

it

extensively

tabulation and in the

theory of

elliptic

functions of modulus unity.

Art.

24.]

series for

gudermanian and

its inverse.

131

Prob. 62. Writing the "elliptic integral of the

first

kind"

in

the form

J
k

Vi

/c

sin

0'
is,

being called the modulus, and

the amplitude; that


/c),

=
show
u
cos

am

u,

(mod.

that, in the special case

when K

=.

1,
y

am

= =

gd~ 0,
sech
u,

tan

am u am &

gd u

sin

am

tanh

&,

sinh u;

and that thus the

elliptic

functions sin

am

u, etc.,
is

degenerate into

the hyperbolic functions,

when the modulus

unity.*

Art.

24.

Series for Gudermanian and its Inverse.


u,

Substitute for sech

sec v in (49), (48) their expansions,


'

Art. 16, and integrate, then

= ugd-'z, = v +
g du

+ &u - j%^u + K + +y^ +


%
1

(50) (5i)

constants of integration appear, since gd u vanishes with and gd" ^ with v. These series are seldom used in compuu,
1

No

tation, as

gd u

is

best found and tabulated

by means

of tables

of natural tangents and hyperbolic sines, from the equation

gd u

= tan -1 (sinh

u),

and a table of the

direct function can be used to furnish the


;

numerical values of the inverse function

or the latter can be


-1

obtained from the equation,


-I

gd

z>

sinh

-1

(tan v)

cosh

( sec

v )>

To

obtain a logarithmic expression for gd"V, let

gd"V
* The relation gd u

= u,
1),

= gd u,
name
the function

= am

u,

(mod.

led Hoiiel to

gd

u,

In this the hyperbolic amplitude of u, and to write itamh u (see note, Art. 22). connection Cayley expressed the functions tanh u, sech u, sinh u in the form
sin

gd

u, cos

gd

u, tan

gd

u,

and wrote them sg


1); etc.

u,

eg
u,

u, tg u, to

correspond
tan

with the abbreviations sn u, en u, dn u for sin

am

cos

am

u,

am

u.

Thus tanh u
It is

sg u

sn u, (mod.

well to

note that

neither the

elliptic

received their
special case.

names on account

of the relation existing

nor the hyperbol'c functions between them in Sl

(See foot-note, p. 107.)

132
therefore
sec v

HYPERBOLIC FUNCTIONS.

[Chap. IV.

= cosh u, sec v + tan v = cosh u =


u,
-\-

tan v

=
#)

sinh u,
,

-f-

sinh u
-|-

= eu

sin

*"

cos ($7t
sin (J
-f-

cos v

tan (iw

+ iv),
(52)
t

v)

gd
-r.
,

'?,

== log, tan (i?r -f j?).

^ Prob.
,

gd 2 63. Evaluate

# U

u~\
,

5
f

gd

V
.

z>~|
(

l=o
sin

_}v=o
fifth

order,

Prob. 64. Prove that gd u when u o.

is

an infinitesimal of the

Prob. 65. Prove the relations

\n

+ \v = tan"VM
25.

\n

\v

tan"V" M

Art.

Graphs of Hyperbolic Functions.

Drawing two rectangular axes, and laying down a series of points whose abscissas represent, on any convenient scale, successive values of the sectorial measure, and whose ordinates represent, preferably on
the same scale, the corresponding values of the
locus traced

function to be plotted, the out by this

series of points will

be a

graphical representation of the variation of the function as the sectorial meas-

Art.

25.]

GRAPHS OF THE hyperbolic functions.

133

ure varies.

The equations
:

of the curves in the ordinary carte-

sian notation are


Fig.

Full Lines.

Dotted Lines.

A
B C

D
Here ^
is

= y= y= y=

cosh x,
sinh x,

tanh x,

= y = y=
y

seen
csch

x x x

coth

gd

x.
u,

written for the sectorial measure


u, etc.

and y

for the

numerical value of cosh


variables x,

It is

thus to be noted that the

y are numbers, or ratios, and that the equation x merely expresses that the relation between the y numbers x and y is taken to be the same as the relation between a sectorial measure and its characteristic ratio. The

cosh

numerical values of cosh//, sinh u, tanh?/ are given in the tables at the end of this chapter for values of u between o and
4.

For greater values they may be computed from the developments of Art. 16.

The

curves exhibit graphically the relations

sech u

=
<
1,

csch u

coth u

=
tanh u

cosh u

sinh u

cosh u
sinh
(

sech u

>

I,

tanh u
cosh
(

>
u)

I,

gd u

>

\n,

etc.;

u)

tanh
cosh

= u) =

sinh u,

tanh
sinh sinh
(

u,

gd

u)

= cosh u, = gd u,
csch
(

etc.;

0=1,

= 0,
00)

tanh

= 0,
,

(o)

=00

etc.;

cosh

00)

= 00,

00

tanh

00)

1,

etc.

dy/dx = cosh x,
Its direction

sinh x is given by the equation slope of the curve y that it is always positive, and that showing the curve becomes more nearly vertical as x becomes infinite.

The

of curvature
is

proving that the curve

obtained from dy/dx* concave downward when x


is

sinh x,

is

nega-

The point of inflexion is tive, and upward when x is positive. at the origin, and the inflexional tangent bisects the angle
between the axes.

134

HYPERBOLIC FUNCTIONS.

[Chap. IV.

The

direction of curvature of the locus


%

by (Py/dx

sech

is

sech x{2 tanh 2

i),

and thus the curve

is

given con-

cave downwards or

upwards

\ 4|M' \^
.

8 according as 2 tanh

negative or positive. flexions occur at the

x i is The inpoints

= tanh" 707, = .881, = .707 and the slopes of y


x
1

the

inflexional

tangents

are

1/2.

The curve y

csch

is

asymptotic to both axes, but

approaches the axis of x more rapidly than it approaches the


axis of y, for

when x
not

*$,

is

C
that

is

so small as

.1.

The

only curves

.1,

but

it is

till

cross at the points

.881, y =

y
I.

= csch x, y = sinh x

y=

10

Art

26.]

ELEMENTARY INTEGRALS.
cosh
*

135
2;

Prob. 69. Solve the equations:

gd x

= x+

sinh

= U~
"

'

\n.

Prob. 70. Show which of the graphs represent even functions and which of them represent odd ones.

Art.

26.

Elementary Integrals.

The
14, 15,

following useful indefinite integrals follow from Arts.

23:
Hyperbolic.
Circular.
,

1.

sinh u

du

cosh

Ain

u du

=
=

cos u

2.

y cosh

udu

sinh 0,

/"cos u

du

sin #,

3.

Aanh udu =
/ coth u

log cosh

a,

Aan udu
/ cot
^/

log cos

4.

du

log sinh

log sin

5.

/csch #</#
sa

= logtanh-, J csc u du = = sinh-^csch a), =


gd
u,

log tan

cosh-^csc

),

6.

/ sech u du

I sec

udu

gd

-1
&,

,/^
a

osh,f,
.

=cos1

-,

9. *

.x P;dx P dx ~1 I / -^ u a x J x<a =-tanh~ -, * a a / -^-j 5 a -\-x

=-tan- ,*
1

* Forms 7-12 are preferable to the respective logarithmic expressions (Art. 19), on account of the close analogy with the circular forms, and also because they involve functions that are directly tabulated. This advantage
appears more clearly
in 13-20.

136

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

r -dx l
10.

/ *

x*aJ x>a
a

=-COth,

1 ,

a
i

-, v / a **
.

aa h- xr dx
.

=a

cot

-1

-, a

/
l2
-

dx
,

=
-2

seen

x
,

P
/

=
J

sec

.x r dx r dx i / / J xVa'+x =~csch- a' d x Vx* _ a a " Va' +x


,

x
CSC

' ,

From
derived:

these fundamental integrals the following

may be

/dx +
Vax*

v==

=,= sinh~ Va 2dx+c


.

-= +
ax
dtfT -f-

b
%
,

Vac-b
.

a positive. ac> V
'
.
.

; '

-=cosh~

#
,

Va

Vb'ac
-1 -

^positive,

ac<b;

cos

ax -\,

b
,

a
1

Vtfac
"

a negative.
b

/dx
v <r

=
#

**"+2t*+<
i

^^
<2^4-^

ax

tan

-\-

yz=y

ac>
\

Vbac
coth1

=
Thus
>

,/Ti-^ Vbac

~===, ^ < Vbac


,

a
,

^+

>
,

|/^

_^

f~^ZT~Zn\
1

=coth- (*
,

2)1

=coth- 2

coth-^

= tanh- (.5)-tanh- (.3333)=:. 5494-. 3466^.2028.*


e

P^+
dx

=-tanh-(^-2)J r tanh-o-tanh-X. 5 )

=-

-5494-

(By interpreting these two integrals as areas, show graphically that the first is positive, and the second negative.)
IK. D

(a-x) Vxb
.

2
.

Va-b

tanh

lxb
\ /
\J
7, a-b
y

*For tanh -1 (.5)

interpolate between tanh (.54)

.4930, tanh (.56)

.5080

(see tables, pp. 162, 163);

and similarly

for

tanh- 1

(.3333)-

Art.

26.]

elementary integrals.
2

137

or

tan

xb
-7
,

\l

or
b

coth- \ /
2
,

, 1

lxb

the real form to be taken.


/c l6

(Put

=z

and apply

9, 10.)

dx

(a-^Vb^-^T^^
2

lbx \/b^'
2

ox

Vba

coth-

. .,

lb/ <^ /

*'
,

or

tan

'l
.

\/

/.*:
1
;

the real form to be taken.


17.

yV By means
a cosh

a *? dx

= -*(* - a

)*

- -a

cosh" -.
is

of a reduction-formula this integral

easily

made

to depend on 8. It may also be obtained by transforming the expression into hyperbolic functions by the assumption

#,

when

the integral takes the form


I (cosh 2u

a* I sinh'

udu=

i)du

= -a*(sinh 28 211) = J#'(sinh u cosh u u),


#,

which gives 17 on replacing a cosh u by


2

and a sinh

by

{.r

a 9 )*.

The
it

-evident, as

geometrical interpretation of the result is expresses that the area of a rectangular-hyper-

bolic

segment
sector

AMP

is

the difference between a triangle

OMP

and a
,

OAP.
2

g.

/V - *)*</* = -*(rt

- x'f + -tf

sin" -.

i 9.

20.

yv + ^>^ = =A Aec
3

^
I

**)*

+ J* sinh_i J
2
1

0^/0

+ tan2 ^)*^ tan ^ = i tan 0(i + tan 0)* + J sinh" = sec tan + 4 gd" 0.
t
1

(tan 0)

21.

sech

</=
What
Show

-J

sech

tanh u

-j-

4 gd

?/.

Prob. 71.
Prob. 72.

is

the geometrical interpretation of 18, 19?


9

that / (tf*

2 bx

+ <r)fe reduces

to 17, 18, 19,

138
respectively:

HYPERBOLIC FUNCTIONS.

[Chap. IV.

when a
is

and when a

is- positive, with ac P. positive, with ac

< b*

when a

is

negative

>

Prob. 73. Prove

/ sinh u tanh u du

sinh

gd u

/
Prob. 74. Integrate
(J?

cosh u coth u du

=
1

cosh

+ log tanh
+
2X

+ 2^ + 5)"^,

Prob. 75. In the if be the length of arc measured from the vertex, and the angle which the tangent line makes with the vertical tangent, prove that the intrinsic equation of 3 -1 the curve is ds/dcp 2p sec <p, s p sec (p tan -f-/gd 0.

+2X + sY dx, s parabola y = 4px,


(*
2

(*"

+ $fdx.

Prob. 76. The polar equation of a parabola being r referred to its focus as pole, express s in terms of 0. Prob. 77. Find the intrinsic equation of the curve y/a

a sec*\Q y
cosh x/a%

and of the curve y/a

= log sec x/a.


cosh""
1

Prob. 78. Investigate a formula of reduction


also integrate

f or /
1

cosh n xdx;

by parts

.*

dx, tanh"

1
.*:

dx,

(smh~ xydx; and


/ cos ^sin

show that the ordinary methods

of reduction for

^^:

can be applied to / cosh w x sinh M x dx.

Art.

27.

Functions of Complex Numbers.

As vector quantities are of frequent occurence in Mathematical Physics and as the numerical measure of a vector in terms of a standard vector is a complex number of the
;

form x-\-iy,

in

which

x,

are real,

and

stands for

1;

it

becomes necessary

in treating of

any

class of functional oper-

ations to consider the

meaning of these operations when performed on such generalized numbers.* The geometrical definitions of cosh//, sinh u, given in Art. 7, being then no longer
it

applicable,

is

necessary to assign to each of the symbols

* The use of vectors in electrical theory is shown in Bedell and Crehore's The advantage Alternating Currents, Chaps, xiv-xx (first published in 1892).
of introducing the
into the differential equaProc. Elec. Congress, 1893; while the additional convenience of expressing the solution in hyperbolic functions of these complex

complex measures of such vectors

tions

is

shown by Steinmetz,
is

numbers

exemplified by

Engineers, April 1895.

Kennelly, Proc. (See below, Art. 37.)

American

Institute

Electrical

Art.

27.]

functions of complex numbers.

139

cosh (x -f iy), sinh (x iy\ a suitable algebraic meaning, which should be consistent with the known algebraic values of cosh x, sinh x, and include these values as a particular case

when y

assigned should also, if possible, be such as to permit the addition-formulas of Art. u to be made general, with all the consequences that flow from them.

= o.

The meanings

Such definitions are furnished by the algebraic developments in Art. 16, which are convergent for all values of u, real or complex. Thus the definitions of cosh (x -J- iy), sinh (x -f- iy)
are to be

cosh (x

iy)

= + L(x + iy)' + 2
i
I

-fx

'

iy)'

+
(52)

sinh (x

iy)

= (x+ iy) +

+ -^(x

+ ...
+
come

From
sinh (x
tion,

these series the numerical values of cosh (x iy\ iy) could be computed to any degree of approximaare given.

when x and y

In general the results will

out in the complex form*


cosh (x
sinh (x
-f- iy)

-\-

=a =c iy)

-f- ib,

-f- id.

The

other functions are defined as in Art.

7,

eq. (9).

Prob. 79. Prove from these definitions that, whatever u

may

be,

cosh

u)

= =

cosh u

sinh

u) =

sinh
u.

du

cosh u

sinh #,

-r-

du

sinh #

= cosh

-7-=

cosh

mu

m* cosh mu,

-r-r

sinh

w =

m* sinh

/.+

*It

is

to be

borne

braic operators

in mind that the symbols cosh, sinh, here stand for algewhich convert one number into another; or which, in the lan-

guage

of vector-analysis,

change one vector into another, by stretching and

turning.

themselves in Mathef The generalized hyperbolic functions usually present 2 2 matical Physics as the solution of the differential equation d'^/du <, where <p, m, u are complex numbers, the measures of vector quantities. (See

=w

Art. 37.)

140

hyperbolic functions.

[chap. iv.

Art.

28.

Addition-Theorems for Complexes.

The addition-theorems for cosh ( -f- v), etc., where u, v are complex numbers, may be derived as follows. First take u, v
as real numbers, then, by Art.
II,
v\

cosh (u -\-v)

hence

i-,(

+ sinh u sinh + vf +. ..=(i+u' + .)(i + f+.


u cosh v
..

= cosh

.)

+ (* +
,

^+
3

)("

+3 V+--)

^ are any real numbers. It This equation is true when be an algebraic identity. For, compare the terms must, then, of the rth degree in the letters u, v on each side. Those on the
left

are

\(u-\- v)

and those on the

right,

when

collected,

form an rth-degree function which is numerically equal to the former for more than r values of u when v is constant, and for more than r values of v when u is constant. Hence the terms
of the rth degree on each side are algebraically identical functions of u and v.* Similarly for the terms of any other degree. Thus the equation above written is an algebraic identity, and
is

true for

all

values of
its

u, v,

whether
it

real or

complex.

Then

writing for each side

symbol,

follows that
-{-

cosh (u

-f- 7;)

= cosh u cosh v
v,

sinh u sinh v\

(53)

and by changing v
cosh (u

into
v)

cosh u cosh v
is

sinh u sinh

v.

(54)

In a similar manner
sinh (u
v)

found
cosh u sinh
v.

sinh u cosh v

(55)

In particular, for a

complex argument,
(56)

cosh (x
sinh (x

iy) = cosh x cosh iy + sinh x sinh iy,) = sinh # cosh 27 cosh sinh iy. /y)
;tr

* '* If two rth-degree functions of a single variable be equal for more than r values of the variable, then they are equal for all values of the variable, and are
algebraically identical."

Art.

29.]

functions of pure imaginaries.

141

sin u, cos u,
u,

Prob. 79. Show, by a similar process of generalization,* that if exp u \ be denned by their developments in powers of then, whatever u may be,
sin (u
-f-

v)

cos

-f v)

exp

(z*

+ v)
3

= sin u cos v + cos = cos cos sin = exp exp v.


z/
z> z*

u sin

z>,

sin v,

Prob. 80. Prove that the following are identities:

cosh u

sinh u
-[-

cosh &
cosh &

sinh

sinh

= = exp ^, & = exp


1,
( (

u) f
u)],
)].

cosh u = J[exp a -f exp sinh u = i[exp & exp(

Art.

29.

Functions of Pure Imaginaries.

In the defining identities

cosh u

=
=

-\

T u*

-I

7 u*

4-

. ,

2!

4!
-

sinh #

&

-|

tU*
3!

-|

-f-

. ,

5*

put for

the pure imaginary

ty,

then
.

cosh iy
sinh z>

=
= =

/ + JLy _

=
.

cosjj/,

(57)

z>

+
~

(z

+ -|^ +
4

*\y
and,

-jj/
z)/

+ ~j\f tan
j/.

=*
.J

sin j,

(58)
(59)

by

division,

tanh

* This method of generalization is sometimes called the principle of the " permanence of equivalence of forms." It is not, however, strictly speaking, a
"principle," but a method; for, the validity of the generalization has to be demonstrated, for any particular form, by means of the principle of the algebraic identity of polynomials enunciated in the preceding foot-note. (See Annals of Mathematics, Vol. 6, p. 81.)
f
cal

with t*

The symbol exp u stands when u is real.

for "exponential function of

," which

is

identi-

142

HYPERBOLIC FUNCTIONS.

[Chap, IV.

These formulas serve


functions.

to interchange hyperbolic

and

circular

cosine of a pure imaginary is real, and the hyperbolic sine and tangent are pure imaginaries. The following table exhibits the variation of sinh u, cosh u,

The hyperbolic

tanh
u

u,

exp

u,

as u takes a succession of pure imaginary values.

Art.

30.]

functions of x
cosh (x
sin (x
/

-f-

iy in

the form
/.*),

X+
*#),

Y.

143

iy)
*

cos (#

= cos (y =F = / sinh = cosh (y qp

(j> =F

*#).
a;

Prob. 84.

From

the product-series

for sin

derive that

for

sinh

sin^

=
=

4-J)(i-^)(i-^ )...
5

sinh

*(,

J)(,

+ -,) ( + pp)
x

..

Art.

30.

Functions of *

+ *>

in

the Form X+z'F.


#
*"

By

the addition-formulas,

cosh (x
sinh (#

iy)

cosh

;tr

cosh y/
cosn iy

-f-

sinh

sinh sinh

iy,
iy->

-f~ iy)

smn *

+ cosn
*

= sin y, sinh cosh zj = cos y, but = cosh x cos y -\-i sinh x sin 7, hence cosh (x iy) sinh (x yO = sinh x cos J + cosh # sin y. c Thus cosh (x + iy) = a + ^, sinh (x iy) # = cosh cos y, # = sinh ^ sin y, = cosh x sin j. = sinh x cos 7,
zj/

-f-

(60)

-f-

if

-\-

-\- id,

then

^r

(61)

<^

From

these expressions the complex tables at the end of

this chapter

have been computed.

Writing cosh z

on Argand diagrams, in complex numbers the usual way, by the points whose coordinates are (x, y), {X, Y); and let the point z move parallel to the j-axis, on a
given line
tions

= Z, where z x z, Z be represented Z

-f- iy,

Z = X-\- iY;

let

the

m, then the point

will describe

an ellipse

whose equation, obtained by eliminating y between the equa-

X = cosh m cos y, Y= sinh m sin y,


(cosh

is

my

(sinh

my
foci is

and which,

as the

parameter

varies, represents a series of

confocal ellipses, the distance between whose

unity.

144
Similarly,
if

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

the point z move parallel to the ^r-axis, on a given will describe an hyperbola whose equaline y n, the point obtained by eliminating the variable x from the equations tion,

X= cosh x cos

n,

Y=

sinh

sin n, is

X*
(cos

F
(sin

nf

nf

and which, as the parameter n varies, represents a series of hyperbolas confocal with the former series of ellipses.

close intervals
;

These two systems of curves, when accurately drawn at on the Z plane, constitute a chart of the hyperand the numerical value of cosh {in -f- in) can be bolic cosine read off at the intersection of the ellipse whose parameter is with the hyperbola whose parameter is n*
85.

P^r^Prob.of curves systems


'

that, in the case of sinh (x -J- iy) % the above two are each turned through a right angle. Compare the chart of sin (x iy). iy) 9 and also of cos (x i sin ay sinh 2X , , x Prob. 86. Prove the identity tanh (* iy) cos 2 cosh 2jC

Prove

+
,

Prob. 87. If cosh (x

-\- iy),

= a + ib

be written

in the

+ "

modulus

and amplitude" form


r
2

as r(cos
2

i sin 6),

r exp W, then
2

= c? -f P = cosh x sirfy = b/a tanh x tan y. tan


sin (x

cos

j^

sinh x,

Prob. 88. Find the modulus and amplitude of sinh (x

iy)>

iy),

exp (x

+
+

/.
-\-

Prob. 89.
also sinh

The
is,

functions sinh
sinh (u

period 2in; that

2tn)

(u+^irt) =i
(

cosh
/?r)

#,

sinh u, cosh

= sinh u, cosh cosh (^ + ^/7r) =


cosh
u.

u,

cosh u have the pure imaginary cosh 2in) (

;.

sinh ^, sinh (u-\-in)>

Prob. 90. The functions cosh"V/z, sinh~V/ have multiple values at intervals of 2in, but each has a unique value (called the principal n for the value) in which the coefficient of i lies between o and

former, and between

\n and

+ \n for the latter.


t

* Such a chart is given by Kennelly, Proc. A. I. E. E., April 1895, and is used by him to obtain the numerical values of cosh {x-\-iy) sinh {x-\-iy), which present themselves as the measures of certain vector quantities in the theory of
alternating currents.

(See Art. 37.)


1.2;

The

chart

is

constructed for values of

and of y between o and

but

it is

available for

all

values oiy t on account of

the periodicity of the functions.

Art.

31.]

the catenary.

145

Art.

31.

The Catenary.
is

A flexible
curve in which

inextensible string

points, and takes up a position

of equilibrium

suspended from two fixed under the

action of gravity.
it

It is

required to find the equation of the

hangs.

AP
T

be the weight of unit length, and s the length of arc measured from the lowest point A then ws is the weight of the portion AP. This is balanced by the terminal tensions,
;

Let

acting in the tangent line at P, and

in the horizontal

tangent.

Resolving horizontally and

vertically gives

T cos
in

H,

T sin
ws
s

ws,

which

is

the inclination of the tangent at P; hence

tan0=i=ir=-,
where
c
is

written for

H/w,
;

constant horizontal tension

the length whose weight therefore


s ~7_ %t
'

is

the

dy

_s
c'
.

ds

dx _
c

ds
Vs'

dx

dx
,
.

Y
.

+
=

c*'

x c

= sinh -1

s
c

x
c

-, sinh

s
c

dy y j-, dx v

x
cosh -,
c

which
axis of

is

x drawn

the required equation of the catenary, referred to an at a distance c below A.

The

following trigonometric
:

method

illustrates the use of


s

= c sec d(p hence dx, = = dss'm 0, = <;sec tan (pd<p; thus x=c gd" <f>,y = c sec 0; dy, = sec gd x/c = cosh x/c and whence y/c = sec = tan gd x/c = sinh x/c. s/c
d(t>
;
1

the gudermanian 2 c sec gives ds

The

" intrinsic

equation," ds cos 0,

c tan 0,
;

Numerical Exercise. A chain whose length is 30 feet is suspended from two points 20 feet apart in the same horizontal find the parameter c, and the depth of the lowest
;

point.

146

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

The equation
by putting 10/c

s/c

= z,

sinh x/c gives \^/c may be written i,$z

ining the intersection of it appears that the root

= sinh lo/c, which, = sinh^. By examthe graphs of/ = siting, y = of this equation z = 1.6, nearly.
1,5,3,

is

To

find a closer

approximation to the

in the

form f(z)

sinh z

i.$z

root, write the equation

o,

then,

by the
.0244,
.0024,

tables,

/(1.60)

=
=

2.3756

2.4000 2.4300 2.4600

/(1.62) == 2.4276
/[1.64)

2.4806
it is

= = =+

.0206;
o,

whence, by interpolation,
z

1.622

1,

10/z
is

found that /(1.6221) ==

and

6.1649.

The

ordinate of either of

the fixed points

given by the equation

y/c

cosh x/c

cosh 10/c

cosh 1.622 1

2.6306,

from tables; hence y


vertex

16.2174,

=y

and required depth of the

10.052$ feet.*

Prob. 91. In the above numerical problem, find the inclination


of the terminal tangent to the horizon.

be drawn from the foot of the Prob. 92. If a perpendicular is equal to the conordinate to the tangent at prove that is equal to the arc AP. stant c and that Hence show that

MN

is the involute of the catenary, and has the propthat the length of the tangent, from the point of contact to the erty axis of x, is constant. (This is the characteristic property of the

the locus of

NP

MN

tractory).

Prob. 93. The tension at any point is equal to the weight of a of the string whose length is equal to the ordinate/ of that portion
point.

Prob. 94.
feet

An

wide and 10

feet high;

arch in the form of an inverted catenary f is 30 show that the length of the arch can be
z

obtained from the equations cosh z


* See a similar problem
Britannica.
in

=1,

25

sinh

z.

Chap.

I,

Art. 7.

f For the theory of this form of arch, see

" Arch "

in the

Encyclopaedia

Art.

32.]

catenary of uniform strength.


32.

147

Art.
If

Catenary of Uniform Strength.

the area of the normal section at any point be made proportional to the tension at that point, there will then be a

constant tension per unit of area, and the tendency to break will be the same at all points. To find the equation of the

curve of equilibrium under gravity, consider the equilibrium of an element PP whose length is ds, and whose weight is gpaods,
f

where

go is
is

the section at P, and p the uniform density.

This

balanced by the difference of the vertical components weight of the tensions at Pand P\ hence

d(T sin

(p)

d(T cos 0)
therefore

= gpoods =o
;

T cos

H, the tension
if

at the lowest point,

and

Again, then by hypothesis go/oo

T= //sec 0.

go

=
sin

be the section at the lowest point, T/H sec 0, and the first equation
0)

becomes
Hd(sec
or

= gpoo
ds,

sec <pds,

cd
c

tan

sec

stands for the constant H/gpoo^ the length of string is equal to the tension at the {of section oo ) whose weight lowest point hence,

where

ds

c sec <pd(p,

s/c

gd-'0>

the intrinsic equation of the catenary of uniform strength.

Also

dx

hence

= ds cos = c d(p, dy = ds sin x y = c log sec 0,


c<f>,

= c tan

d(p

and thus the Cartesian equation


y/c
log sec x/c,

is

in which the axis of

is

the tangent at the lowest point.

Prob. 95. Using the same data as in Art. 3 r find the parameter gd s/c (The equation x/c < and the depth of the lowest point. becomes s, 10/V gd 15A, which, by putting 15/V gives

148

HYPERBOLIC FUNCTIONS.

[Chap. IV.

gd

f(z)

=f = gd z
z.

From
*,

the graph it is seen that z is nearly 1.7. If then, from the tables of the gudermanian at the

end of

this chapter,

= + .0447, 1667 = + .0129, /( -75) i-i79 6 1.2000 = 1804 /(1.80) = .0196, z = 1.7698 and c = 8.4755. whence, by interpolation, Again, and 1.1799 radians y/c = log* sec x/c but x/c = 10/c = 1.1799 = 67 36' 29"; hence y = 8.4755 X .41914 X 2.3026 = 8.1798, the
/(1.70)
I

1.

1780

1.

1333

1.

1.

required depth.) Prob. 96. Find the inclination of the terminal tangent.

Prob. 97.

Show

that the curve has two vertical asymptotes.

go,

Prob. 98. Prove that the law of the tension T, and of the section at a distance s, measured from the lowest point along the
is

curve,

H=
is

GD
<z>

= cosh
,

-; %
c

and show that in the above numerical example the terminal section
2.85 times the

minimum

section.

Prob.

W= H
An

=c

99. Prove that the radius of curvature is given by cosh s/c. Also that the weight of the arc s is given by sinh s/c, in which s is measured from the vertex.

Art.

33.

The Elastic Catenary.

elastic string of

ral state is

uniform section and density in its natususpended from two points. Find its equation of

equilibrium. Let the element dcr stretch into ds; then, by Hooke's law, ds dcr(i -f- A.T), where X is the elastic constant of the string

hence the weight of the stretched element


gpoods/{i -{-XT),

ds,

= gp coder,

=.

Accordingly, as before,

d(T sin
and
hence
in

= gpoods/il + A T), T cos = H = gpooc, sec 0), ^(tan 0) = ds/(i +


0)
jj.
;

which

pi

stands for XH, the extension at the lowest point

Art.

34.]

THE TRACTORY.
ds
s/c
a

149

therefore

^(sec

-f-

sec3 0)^0,

= tan

-f-

/*(sec

tan

+ gd"

0),

which
of the

is

the intrinsic equation of the curve, and reduces to that

o. The coordinates x, y catenary when }x may be expressed in terms of the single parameter by put2 ds cos ting dx -f- jx sec (p)d(p, <r(sec 2 ds sin dy /* sec 0) sin <:(sec ^0. Whence

common

= =

_1

x/c

gd

//

tan 0,

j/c

= sec

yu

tan 2 0.
result of

These equations are more convenient than the eliminating 0, which is somewhat complicated.

Art.

34.

The Tractory.*

the equation of the curve which possesses the that the length of the tangent from the point of conproperty tact to the axis of x is confind

To

stant.

Let FT, P'T' be two consecutive

tangents
c,

such
let

that

PT=P'T' = = t\ draw TS
to FT'-,
is

and

OT

perpendicular

then

if

evident

that

PP' = ds, it ST' differs

from ds by an infinitesimal of a higher order. Let FT make with OA, the axis of y; then (to the first order of an angle
infinitesimals)

=
is

= TS = cd(p = cos c sin 0, = (gd


PTd<p
used

TT'
t

cos 0; that

is,

cf)dt y

= c gd
i

-1

0,

-1

sin 0),

= c cos 0.
all

This

a convenient single-parameter form, which gives


is

* This curve
p. 242)
;

in Schiele's anti-friction

pivot (Minchin's Statics, Vol.

i,

theory of the skew circular arch, the horizontal projection of the joints being a tractory. (See "Arch," Encyclopaedia Britannica.) The gd t/c furnisher a convenient method of plotting the curve. equation <p

and

in the

150
values of x,

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

as

increases from o to \n.


is

The

value of

s,

ex-

pressed
ds

in

the same form,


r

found from the relation


c

= ST = dt sin
=

tan

cpd(fi,

=c

log,,

sec 0.

the point A, o, /=o, /=<:. o, ^ o, ^ Cartesian equation, obtained by eliminating 0, is

At

=
-1

The

=
If

-1

gd

(cos

-1

-J

sin (cos

-\

cosh

-1

/i

-.

= gd u,
jy

& be put for //*:, and be taken as independent variable,. u tanh u, y/c =. sech #, s/c log cosh #. x/c

Prob. 100. Given

/=

2C t

.265SY,

.*

1.0360*:.

show that At what point

0= 75
is t

= cl

35', s

1.3249^,

Show that the evolute of the tractory is the catenary. Prob. 92.) (See Prob. 102. Find the radius of curvature of the tractory in terms and derive the intrinsic equation of the involute. of ;
Prob. 101.

Art.

35.

The Loxodrome.

the surface of a sphere a curve starts from the equator in a given direction and cuts all the meridians at the same
angle.

On

To
:

find its equation

in latitude-and-longitude co-

ordinates

Let the loxodrome cross two consecutive meridians

AM, AN in
let
J/jV

the points

Q;.

PR
;

be a parallel of

lati-

tude let OM x, MP = y = dx, RQ = dy, all in radian measure and let the angle MN MOP=RPQ = a; then tan a = RQ/PR, but PR = MN cos MP* hence dx tan a = dy sec y, and x tan a = gd y, there being^
;

_I

no integration-constant since y vanishes with x; thus the required equation


*
is

gd (x tan

a).

Jones, Trigonometry (Ithaca, 1890), p. 185.

Art.

36.]

combined flexure and tension.

151

To

find the length of the arc

OP:

Integrate the equation


s

ds

= dy esc a,

whence

= y esc a.

To illustrate numerically, suppose a ship sails northeast, from a point on the equator, until her difference of longitude is 45, find her latitude and distance
:

Here tan a
radians: s
40.980.
If

I,

andj/

= gd x = gd \n =
radii.

gd (.7854}= .7152
is

=y

V2

1.0114

The

latitude in degrees

modified as follows
tion

the ship set out from latitude y l9 the formula must be Integrating the above differential equa:

between the limits


(*.

(^,, j/,)

and (x yt ) gives
g d ">2

- O tan
-1

gd">iJ

hence

gd~y =
7

gd

^,

latitude can be found

ence of longitude are


(fi

-|- <X when the initial latitude and the differThe distance sailed is equal to given.

*i) tan a, from which the final

fi) csc

radii,

a radius being 60

i8o/7r nautical miles.

In this projection the meridians are parallel straight lines, and the loxodrome becomes the straight line y' x tan a, hence the relations between the coordinates of f x corresponding points on the plane and sphere are x

Mercator's Chart.

=
~

y'
is

= gd
"
;

y.

Thus the

latitude

tabulated under the


the values of

name

magnified into gd y, which of " meridional part for latitude

"

is

y and of y' being given in minutes. A chart y constructed accurately from the tables can be used to furnish graphical solutions of problems like the one proposed above.
Prob. 103. Find the distance on a
(30

rhumb

line

between the points

N, 20

E) and (30

S,

40

E).

Art.

36.

Combined Flexure and Tension.


is

A beam
other, and

that
is

built-in at

one end carries a load

P at
Q

the

also subjected to a horizontal tensile force apfind the equation of the curve plied at the same point; to Let x, y be any point of the its neutral surface assumed

by

152

HYPERBOLIC FUNCTIONS.

[Chap. IV.

elastic curve, referred to the free


is

end as

origin, then the

bend-

Px. Qy ing moment for this point notation of the theory of flexure,*

Hence, with the usual

dx*

dx

=
l

?i*{

mx).

m=P

Q
2
,

Q EI

which, on putting/

mx =

u,

zxAd %y/dx %

=d

n/dx becomes

whence
that
is,

= A cosh nx B sinh nx, = mx A cosh nx B sinh nx.


-f-

-f-

-j-

The
must be

arbitrary constants A,

are to be determined

by the

terminal conditions.
zero,

At

the free end

y=o;

hence

and

dx
but at the fixed end,

= mx + B sinh nx, _f- = m + nB cosh nx


y

I,

and dy/dx

= o,

hence

B=
and accordingly
y

m/n

cosh

nl,

= w^r
x

;;z

sinh
;

nx
~.

cosh nl

To

obtain the deflection of the loaded end, find the ordinate

of the fixed

end by putting
deflection

=
v

/,

giving

= mil

tanh/z/). '
of a cast-iron beam,

Prob. 104.

Compute

the deflection

2X2

inches section, and 6 feet span, built-in at one end and carrying a load of 100 pounds at the other end, the beam being subjected
to a horizontal tension of 8000 pounds.

[In

this

case

/=

4/3,

E=
*

15

io

6
,

Q = 8000, P =

100

hence n
^5(72

deflection

= ^(72

50 tanh 1.44)

1/50,

m=

1/80,

44-69)

-34* inches.]

Merriman, Mechanics of Materials (New York,

1895), pp. 70-77, 267-269,

Art.

o7.]

alternating currents.

153

Prob. 105. If the load be uniformly distributed over the beam, say w per linear unit, prove that the differential equation is

= @y ~ ^ wx^ or = n *(y - mx fa and that the solution isy = A cosh nx B sm\\nx -\-mx
EI
~ci?
l

*)>

-\-

-\

^-.

ti

Show

also

how

to

determine the arbitrary constants.

Art.

37.

Alternating Currents.*

In the general problem treated the cable or wire is regarded as having resistance, distributed capacity, self-induction, and although some of these may be zero in special leakage
;

cases.
circuit

The
of

line will also

be considered to feed into a receiver


;

any description
insulated.

clude the particular cases

in

and the general solution will inwhich the receiving end is either

grounded or

electromotive force may, without loss of generality, be taken as a simple harmonic function of the time, because any periodic function can be expressed in a

The

Fourier series of simple harmonics. f


current,

The E.M.F. and the


will

which

may

differ

in

phase by any angle,

be

supposed to have given values at the terminals of the receiver and the problem then is to determine the E.M.F. circuit
;

and current that must be kept up at the generator terminals and also to express the values of these quantities at any intermediate point, distant x from the receiving end the four
;

line-constants being supposed known,

viz.:

R = resistance, in ohms per mile, L = coefficient of self-induction, in


C=
capacity,
in

henrys per mile,

farads per mile,

G
It is

coefficient of leakage, in

mhos per mile4


that
if

shown

in

standard works
foot-note Art. 27.

any simple harmonic


Chapter V, Art. 8. Steinmetz writes
j for

* See references
\

in

Kennelly denotes these constants by r, /, c, for G, and he uses C for current. coL, K for ooC, Thomson and Tait. Natural Philosophy, Vol
of Sound, Vol.
I.

g.

I.

p. 40;

Rayleigh, Theory
p. 214.

p. 20;

Bedell and Crehore, Alternating Currents,

154
function a sin

HYPERBOLIC FUNCTIONS.
(pot -\- 6)

[CHAP. IV.

be represented by a vector of length

a and angle Q, then two simple harmonics of the same period 27t/gd, but having different values of the phase-angle 0, can be

combined by adding their representative vectors. Now the E.M.F. and the current at any point of the circuit, distant x from the receiving end, are of the form
e

=e

sin (oot

-j- #),

=i

sin (oot

-j-

6'\

(64)

values t v and the phase-angles 0, 0\ These simple harmonics will be represented by the vectors ejd, ijd' whose numerical measures
in

which the
all

maximum

<?

are

functions of x.

complexes j sin 0'), which will be denoted by e, i. The relations between e and l may be obtained from the ordinary equations f
di
,

are the

e x (cos 6 -\-js\n 0)*, ix (cos 0' -\-

^de

de

_.

di

,^

for, since

de/dt

ooe

cos

(oot -f- 0)

ooe x sin (Got -{Goe 1

-\-

^n) >

then

dfe/dfr will

be represented by the vector


x

/0

-fl

\n

and

di/dx by the sum of the two vectors Ge /0, Cooe /0 \n ; whose numerical measures are the complexes 6V, jooCe-\ and thus the relations similarly for de/dx in the second equation
;

between the complexes

e, 1

are

5 = (<?+>cy,
*In
f

= (*+>)*

(66)t

electrical theory the symbol j is used, instead of i, for |/ 1. Bedell and Crehore, Alternating Currents, p. 181. The sign of dx
is

is

changed, because x
leakage, G,
is

measured from the receiving end.


is

The

coefficient of

usually taken zero, but

here retained for generality and sym-

metry.
%

These

relations

derives

them from

first

have the advantage of not involving the time. Steinmetz For instance, principles without using the variable /.

he regards
to
i,

R -f joaL

as a generalized resistance-coefficient, which,

when
in

applied

gives an E.M.F., part of which


i.

with

Kennelly

calls

R+

phase with i, and part the conductor impedance; and JgoL


is

in

quadrature
-f-

JaoC the

dielectric admittance; the reciprocal of

which

is

the dielectric impedance.

Art.

37.]

alternating currents.

155

Differentiating and substituting give

(67}

and thus

^~,

are similar functions of x, to be distinguished only

by

their terminal values.


It is

now convenient

to define

two constants

m ^m
i

by the

equations*

w = R +jgdL)(G +jooC)
>

m?

= (R +JgoL)/(G +J&Q
then be written

(68)

and the

differential equations

may

the solutions of which are f

=A

cosh

?#;tr -|-

B sinh

;/z.r,

= A'

cosh

;;z;tr

-f~ i?'

sinh

mx

wherein only two of the four constants are arbitrary


stituting in either of
cients, give

for subcoeffi-

the equations

(66),

and equating

(G+jgdC)A = mB\
whence
B'
let

(G+ja>C)B
A'

= mA\

= A/m v

= B/m =

. x

the assigned terminal values of e, o gives be denoted by E, J; then puttings

Next

i,

at the receiver,.

whence

B = mj
e

B'

= E/m
mx

A, I A\. and thus the general solution is

E=

= E cosh mx + mj sinh =
i

mx,
|

I cosh

-f-

~E sinh mx.

(7o>

variable length
f

* The complex constants m, ;i are written z, y by Kennelly; and Steinmetz writes v for m. x is written Z 2
, .

the-

See Art.

14,

Probs. 28-30; and Art. 27, foot-note.

156
If desired,

HYPERBOLIC FUNCTIONS.

[Chap. IV.

these expressions could be thrown into the ordinary complex form X-\-jY, X' -\-jY', by putting for the letters their complex values, and applying the addition-theorems
quantities X, Y, X', Y' would then be expressed as functions of x and the repre 2 2 sentative vectors of e, i, would be e /0, i /#', where e' -|;

for the hyperbolic sine

and

cosine.

The

=X

#,

= X" +

Y'\ tan 6

7 Y/X, tanl =Y'/X.

For purposes
mulas

of numerical computation, however, the for-

(70) are the most convenient, when either a chart,* or a table,f of cosh u> sinh u, is available, for complex values of u.
2 ohms per mile, Prob. 106. \ Given the four line-constants: 20 millihenrys per mile, 1/2 microfarad per mile, 0; and given go, the angular velocity of E.M.F. to be 2000 radians

R=

L=

C=

per second; then

= 40 ohms, conductor reactance per mile; R -\-jooL = 2 + 40/ ohms, conductor impedance per mile; 00C = .001 mho, dielectric susceptance per mile; G + jooC = .001/' mho, dielectric admittance per mile; 000;' ohms, dielectric impedance per mile; {G-\-jooC)~ = = [R jooL){G -\-jooC) .04 .0027, which m
ooL
l

-\-

-f-

is

the

measure of .04005

/177." 8';

therefore

m=
l

measure of .2001 /88


200

34'

= .0050 + .2000/,
ohms
per mile.

an ab-1
,

stract coefficient per mile, of dimensions [length]

mm = m/(G + jooC) =
Next

5/'

let the assigned terminals conditions at the receiver be: 1000 volts, whose phase may be taken o (line insulated) and as the standard (or zero) phase; then at any distance x, by (70),

/=

E=

= E cosh mx

1
y

sinn
x

mx

in

which

mx

is

an abstract complex.

Suppose it is required to find the E.M.F. and current that must be kept up at a generator 100 miles away; then
* Art. \ See Table II. 30, foot-note. The data for this example are taken from Kennelly's article. %

Art.

37.]

alternating currents.

157
(.5
20/'),

e=
but,

iooo cosh
89,

by Prob.

cosh

+ 20/), = 200(40 /)-' sinh + + 20/) = cosh (.5 + 20/ 67tj) = cosh + 1.15/) = .4600 + .475/
(.5
1

(.5

(.5

obtained from Table II, by interpolation between cosh and cosh (.5 1.2/); hence

(.5

1.17)

460

+ 475/' = ^(cos 6 +j sin


=
sinh

6),

where log tan 6 = log 475 log 460 = .0139, = 460 sec # = 625.9 volts, the required E.M.F. *,
Similarly sinh (.5
-+-

= 45

55',

and

207)

(.5 -f- 1.157)

= .2126 +

1.02807V

and hence
l

= --(lOO +7)(.2I26 4- I.O287) = 9.70684, log tan 0'


let it

~2

(4046

+ 20607)
6/'

26 #'/i6oi amperes, the phase and magnitude of required current.

where

#'=

= 1,(008 = 4046 sec 59', ^


x

4-7

sin 6'),

= 2.77

Next
e

be required to find
-f-

e at

8;

then

1000 cosh (.04

1.67)

=
>

10007

smn

C4 + '3j')f
Interpolation be-

by subtracting %7tj\ and applying Prob. 89. l tween sinh (o 07) and sinh (o j) gives

sinh (o
Similarly

-|-

.037)
-

= 00000
I

-J-

.029957.
.030047'.

sinh

(.1 -f

ZJ)

ooo4

-j-

Interpolation between the


sinh (.04

last

two gives

+ 'OS J) = .04002 + .029997.


29.994- 40.027

Hence/ =7(40.02 +29.997')=


where
log tan 6
volts.

=^(cos

0-\-jsin #),

.12530, 6

126

51',

<?,

=
x

29.99 sec I2 ^

'

==

01

Again,
e ==

let it

be required

to find e at

=
4~

16; here

1000 cosh (.08 4" 3*27)

=
cosh

1000 cosh (.08 4- .067),


(.1

but cosh (o 4- -067)

= .9970 4"

j\

6/)

1.0020 4- .0067;

hence

cosh (.08 4" 06J)


e

= 1.00 10 -f- 00487,


#),

and
where 6

10014-4.87

= <?,(cos 0+j sin


Thus
at

180

17', ^,
is

1001 volts.

a distance of about

16 miles the E.M.F.

the

same

as

at the receiver,

but

in

opposite

158
phase.

HYPERBOLIC FUNCTIONS.
Since
e is

[CHAP. IV.

proportional to cosh (.005 4- .2j)x, the value of


.

which the phase is exactly 180 is n/,2 15.7. Similarly the phase of the E.M.F. at x = 7.85 is 90 There is agreement in phase at any two points whose distance apart is 31.4 miles.
for

In conclusion take the more general terminal conditions in

which the line feeds into a receiver circuit, and suppose the current is to be kept at 50 amperes, in a phase 40 in advance of the electromotive force; then/
5o(cos 40
)

+/ sin 40 = 38,30 + 32.14/;


-j-

and
e

substituting the constants in (70) gives

1000 cosh (.005

-j-

.y)x

+ (7821

62367) sinh (.005

-f- .2j')x

460+475/ -4748+9366/= -4288+984iy=^(cos 0+/sin

0),

where This is

10730 volts, the E.M.F. at sending end. 113 times what was required when the other end was insulated. 17 Prob. 107. If the receiving end be grounded, that is if =0; and if a current of 10 amperes be caused to flow to ground; find
33', e
l

6=

the E.M.F. and current to be kept up at the generator. Also compute these quantities, and their phases, at the distances 7.85,
15.7, 31.42, 94.25 miles

from the receiver.

Prob. 108. If self-induction and capacity be zero, and the receiving end be insulated, show that the graph of the electromotive force is a catenary. Prob. 109. Neglecting leakage and capacity, prove that the
solution of equations (66)
is
1

/, e

= E+

(R

+ jooL)Ix.
how

the sending end, show equations (65), (66) are to be modified; and prove that 1 _ _ sinh m IQ sinh mx, 1 7 cosh mx cosh mx a

Prob.

no.

If

x be measured from

=E

~E

mx

where

refer to the sending end.

Art.
1.

38.

Miscellaneous Applications.

The
The
tf(sinh

y = J/(cosh+logtanhJ), in which M=
2.

length of the arc of the logarithmic curve length of arc of the spiral of

ax

is

i/loga, sinh w

=y/M.

Archimedes r

= aO

is

=
3.

2u +- 2u), where sinh u


1

0.

In the hyperbola x^/a* y /& 1 the radius of curva2 2 ture is p=(a* sinh u-{-d>* cosh iCf/ab', in which u is the,

measure
4.

of the sector

A OP,

i.e.

cosh u

= x/a,

sinh u

y/b.

In an oblate spheroid, the superficial area of the zone

ART.

38.]

MISCELLANEOUS APPLICATIONS.
parallel plane at a distance
is

159

between the equator and a


(sinh 2u eccentricity, n
7r^
5.

5=

is

2ii)/2e,

wherein b

the axial radius,

the

and/ parameter of generating ellipse. length of the arc of the parabola 2px, measured from the vertex of the curve, is / i/(sinh 2u-\-2u), in which sinh u tan 0, where is the inclination of the termiy/p
ey/p,

The

nal tangent to the initial one. 6. The centre of gravity of this arc

is

given by
411)
;

llx

=/ (cosh
a

i),

64/y

p* (sinh 411
is

and the surface


7.

of a paraboloid of revolution of inertia of the

5= 2n yl.
its ter*
pi

The moment
is

same
3

arc about

minal ordinate

I=

pi\_x/(x

2x)

-{-

^T p N], where

is

the mass of unit length, and

JV
8.

\ sinh 2u

\ sinh 4

-\-

-^ sinh 6u.

The

centre of gravity of the arc of a catenary measured


is

from the lowest point

given by

4fy=
in
its

2
<r

sin h

2u
;

+ 2u), fx = c\u sinh u


is

cosh u

1),

which u

= x/c
3

and the moment of

inertia of this arc

about

terminal abscissa

/ = /^ (tV
9.

sinh

3?/

sinh u

11

cosh

11).

of bars are given in RayApplications to the vibrations Vol. I, art. 170: to the torsion of leigh, Theory of Sound, to the flow of heat prisms in Love, Elasticity, pp. 166-74; Fourier Series, pp. 75-81 to wave and electricity in Byerly,
;

motion
Bassett,

in fluids in

Rayleigh, Vol.
arts.

I,

Appendix,
384; to

p. 477,

and

in

Hydrodynamics,

120,
in

the

theory of

Maxwell, Electricity, arts. potential in Byerry p. 135, to Non-Euclidian geometry and many other subjects

and

and VI. Several Gunther, Hyperbelfunktionen, Chaps. V are worked out in Laisant, Essai sur les numerical examples
in

172-4;

fonctions hyperboliques.

160

HYPERBOLIC FUNCTIONS.

[CHAP. IV.

Art.
In Table
y

39.

Explanation of Tables.

I
y

sinh u cosh u

the numerical values of the hyperbolic functions tanh u are tabulated for values of u increasing
.02.

from o to 4 at intervals of may be used.

When

u exceeds

4,

Table

IV

Table II gives hyperbolic functions of complex arguments,


in

which
cosh (x
iy)

=a

ib>

sinh (x

iy)

= c zh

id,

and the values of a, b, c, d are tabulated for values of x and of y ranging separately from o to 1.5 at intervals of .1. When interpolation is necessary it may be performed in three For example, to find cosh (.82 -f- 1.342') First find stages. cosh (.82-]- i-30 by keeping at 1.3 and interpolating between
:

jj/

the entries under x

= .8 and* =

.9

next find cosh

(.82

i-4*)>

by keeping y = .8 and x
cosh (.82
in

at 1.4

and interpolating between the entries under .9, as before; then by interpolation between

1.31)

and cosh
at .82.

(.82

1.42) find
is

cosh( .82

-f- 1.342),

which x

is

kept

The

table

available for

all

values

of y,

however
-\-

great,

by means

of the formulas

sinh (x
It

2in

= sinh^r,
x
is

cosh (x-\-2i7t)
greater than
1.5,

= cosh

x, etc.
sel-

dom

does not apply when occurs in practice.

but this case

This table can also be used as a com-

plex table of circular functions, for


cos (y
ix)

= a =F
x

tb,

sin

(y

ix)
is

= d -.ic
by

and, moreover, the exponential function

given
d),

exp
in

iy)

= ac

i(b

which the signs of c and afare to be taken the same as the sign of x, and the sign of i on the right is to be the product of

the signs of

x and

of

on the

left.

Table III gives the values of v

gd

manian angle

0=

180

v/ir, as u changes

u, and of the guderfrom o to I at inter-

Art. 39] vals of .02,


intervals of

explanation of tables.
from
.1.
I

161

to 2 at intervals of .05,

and from

2 to

4 at

In Table

cosh

u,

IV are given the values of gd u, log sinh u, log as u increases from 4 to 6 at intervals of .1, from 6 to
.2,

7 at intervals of

and from 7 to 9
be

at intervals of

.5.

In the rare cases in which


sary, reference

more extensive
In the

tables are neces-

may

made

to the tables* of
first

Gudermann,

Glaisher, and Geipel and Kilgour.

the Guderman-

ian angle (written k) is taken as the independent variable, and increases from o to 100 grades at intervals of .01, the corre-

sponding value of u (written Lk) being tabulated. In the usual case, in which the table is entered with the value of u, it gives
circular functions

by interpolation the value of the gudermanian angle, whose would then give the hyperbolic functions
of u.

When
is

is

large, this angle

is

so nearly right that interthis

polation

not reliable.

To remedy

inconvenience Gu-

dermann's second table gives directly log sinh u, log cosh u> log tanh u, to nine figures, for values of u varying by .001 from
2 to
5,

and by

.01

from

to 12.
e*

Glaisher has tabulated the values of


nificant figures, as

and

e~ x to nine sig,

varies
10,

by

.001
I

from o to

.1,

by

.01

from o
these

to

2,

by

.1

from o to

and by

from o to 500.

From

the values of cosh x, sinh

are easily obtained.

Geipel and Kilgour's handbook gives the values of cosh;tr, sinh x, to seven figures, as x varies by .01 from o to 4. There are also extensive tables by Forti, Gronau, Vassal, and there are four-place tables in Byerly's Callet, and Hoiiel
;

Fourier Series, and

in

Wheeler's Trigonometry.

In the following tables a dash over a final digit indicates that the number has been increased.
in Crelle's Journal, vols. 6-9, 1831-2 (published separately Theorie der hyperbolischen Functionen, Berlin, 1833). Glaisher in Cambridge Phil. Trans., vol. 13, 1881. Geipel and Kilgour's Electrical Handbook.

Gudermann
title

under the

162

HYPERBOLIC FUNCTIONS.
Table
I.

[Chap. IV.

Hyperbolic Functions.

u.

Art.

39.]

TABLES.

163

Table
u.

I.

Hyperbolic Functions.

164
Table
II.

HYPERBOLIC FUNCTIONS.
Values of cosh
(x

[CHAP. IV.
(x
-4-

iy)

and sinh

iy).

Art.

39.]

tables.
Table
II.

165

Values of cosh

(x -f iy)

and sinh

(x -f

iy).

16G

HYPERBOLIC FUNCTIONS.
Table
II.

[CHAP.
(x -f
iy).

IV*.

Values of cosh

(x

+ iy)

and sinh

Art.

39.]

tables.
Table
II.

167

Values of cosh

(x -f iy)

and sinh

(x

-\--iy.)

168

HYPERBOLIC FUNCTIONS.
Table
III.

[Chap. IV

ART.

l."|

HISTORY AND DESCRIPTION.

169

Chapter V.

HARMONIC FUNCTIONS.
By William
E.

Byerly,

Professor of Mathematics in Harvard University,

Art.

1.

History and Description.

What

is

known

and development

as the Harmonic Analysis owed its origin to the study of concrete problems in various

branches of Mathematical Physics, which however all involved the treatment of partial differential equations of the same
general form. The use of

Trigonometric Series was


in

first

suggested by

Daniel Bernouilli

1753

in

his

researches on the musical

vibrations of stretched elastic strings, although Bessel's Functions had been already (1732)

employed by him and by Euler

in dealing with the vibrations of a heavy string suspended from

one end; and Zonal and Spherical Harmonics were introduced by Legendre and Laplace in 1782 in dealing with the attraction of solids of revolution.

was greatly advanced by Fourier in 1812-1824 in his remarkable work on the Conduction of Heat, and important additions have been made by Lame (1839) anc by a

The

analysis

host of modern investigators.

The

differential equations treated in the

problems which

have just been enumerated are

170

HARMONIC FUNCTIONS.
;

[CHAP. V.

for the transverse vibrations of a musical string

dt

ox

oxl

for small transverse vibrations of a

uniform heavy string sus-

pended from one end

ox*

"*"

ay
;

&*

u>

which

is

Laplace's equation

and

for the conduction of heat in a

homogeneous

solid.

Of these Laplace's equation (3), and (4) of which (3) is a special case, are by far the most important, and we shall concern ourselves mainly with them in this chapter. As to their interest to engineers and physicists we quote from an article
in

The
"

Electrician of Jan. 26, 1894,


is

by Professor John Perry:


is

There

a well-known partial differential equation, which

the same in problems on heat-conduction, motion of fluids, the establishment of electrostatic or electromagnetic potential, certain

motions of viscous

fluid,

certain kinds of strain and stress, currents


elastic solids, vibrations of flexible

in a conductor, vibrations of

strings or elastic membranes,

and innumerable other phenomena.

The

equation has always to be solved subject to certain boundary

or limiting conditions, sometimes as to space and time, sometimes


as to space alone,

and we know that


is

if

we

obtain any solution of a


Further-'

particular problem, then that

the true and only solution.


is

more,

if

a solution, say, of a heat-conduction problem

obtained

by any person, that answer is at once applicable to analogous problems in all the other departments of physics. Thus, if Lord Kelvin draws for us the lines of flow in a simple vortex, he has drawn
for

Lord Rayleigh
which
is

us the lines of magnetic force about a circular current: if calculates for us the resistance of the mouth of an

organ-pipe, he has also determined the end effect of a bar of iron

magnetized; when Mr. Oliver Heaviside shows his match-

ART.

1.]

HISTORY AND DESCRIPTION.

171

less skill

and

familiarity with Bessel's functions in solving electrois

solving problems in heat-conductivity 01 the strains in prismatic shafts. How difficult it is to express exactly
th-e

magnetic problems, he

distribution of strain in a twisted square shaft, for example,

and

yet

how

easy

it

is

to

understand thoroughly when one knows the

perfect-fluid analogy!

How

easy, again,

it is

to imagine the electric

current density everywhere in a conductor

when

transmitting alter-

nating currents

viscous-fluid analogy, or even the heat-conduction analogy! " Much has been written about the correlation of the physical sciences; but when we observe how a young man who has worked

when we know Mr. Heaviside's

almost altogether at heat problems suddenly shows himself acquainted with the most difficult investigations in other departments
of physics,

we may say

that the true correlation of the physical

sciences

lies in

the equation of continuity

~ a ,/diu

W^ay + aW*

av

av\,

In the Theory of the Potential Function


of

in

the Attraction

Gravitation, and

in Electrostatics
(3) is

Vxn Laplace's equation


tion, at

and Electrodynamics,* the value of the Potential Func-

any external point {x, y, z\ due to any distribution of matter or of electricity; in the theory of the Conduction of

Heat

in a

homogeneous

solid f

is

the temperature at any

point in the solid after the stationary temperatures have been established, and in the theory of the irrotational flow of an

incompressible fluid % V is the Velocity Potential Function and (3) is known as the equation of continuity.
If

we

use spherical coordinates,

(3)

takes the form

si

xVZ\rV), 1 r?L 3? '"iinT


f

"49 ^ d
1

sin'

0B<p' J

n_ c W
. '

rrt

Boston. * See Peirce's Newtonian Potential Function. See Fourier's Analytic Theory of Heat. London and
Brunswick.

New
1895.

York, 1878

or Riemann's Partielle Differentialgleichungen.


%

See Lamb's Hydrodynamics.

London and New York,

172

HARMONIC FUNCTIONS.
if

[CHAP. V.

and

we

use cylindrical coordinates, the form

*V

r\^\^Z^Z=o.
1

(6)

In the theory of the Conduction of Heat in a homogeneous solid,* u in equation (4) is the temperature of any point {x, y, 2) of the solid at any time t, and a is a constant deter-

mined by experiment and depending on the conductivity and


the thermal capacity of the solid.

Art.

2.

Homogeneous Linear Differential Equations.

general solution of a differential equation is the equation expressing the most general relation between the primitive variables which

The

consistent with the given differential equation and which does not involve differentials or derivatives. general solution will always contain arbitrary (i.e., undeteris

mined) constants or arbitrary functions.

particular solution of a differential equation

is

a relation

between the primitive variables which is consistent with the given differential equation, but which is less general than the
general solution, although included
Theoretically,
in
it.

every particular solution can


in

be obtained

from the general solution by substituting

the general solution particular values for the arbitrary constants or particular functions for the arbitrary functions but in practice it is often
;

easy to obtain particular solutions directly from the differential equation when it would be difficult or impossible to obtain the
general solution.
(a) If a

problem requiring
is

for its solution the solving of a

determinate, there must always be given in addition to the differential equation enough outside condidifferential equation
all the arbitrary constants or functions that enter into the general solution of the arbitrary

tions for the determination of

equation
tial

in dealing with such a problem, if the differencan be readily solved the natural method of proequation
;

and

ART.

2.]

HOMOGENEOUS LINEAR DIFFERENTIAL EQUATIONS.

175

cedure

is to obtain its general solution, and then to determine the constants or functions by the aid of the given conditions.

happens, however, that the general solution of the differential equation in question cannot be obtained, and then,,
if determinate, will be solved, if by any solution of the equation can be found which will also satisfy the given outside conditions, it is worth while to try to get particular solutions and so to combine them as to form a

It often

since the problem,

means a

result

which

shall satisfy the given conditicns

without ceasing

to satisfy the differential equation.


(J?)

A differential
if

degree were regarded as algebraic unknown quantities. If it is linear and contains no term which does not involve the dependent variable or one of its derivatives, it is said to be linear and

first

equation is linear when the dependent variable and

it

would be
its

of the

all

derivatives

homogeneous.
All the differential equations given in Art.
I

are linear and

homogeneous.
(c)

If

a value of the dependent variable has been found

given homogeneous, linear, differential equation, the product formed by multiplying this value by any constant will also be a value of the dependent variable which
will satisfy

which

satisfies a

the equation. the terms of the given equation are transposed member, the substitution of the first-named value
to

For
to the

if

all

first

must reduce that member


value
the
fore
is

zero

equivalent to multiplying each


substitution

substituting the second term of the result of


factor,

first

by the same constant

which there-

may

be taken out as a factor of the whole


factor being zero, the product
is

first

The remaining
equation
is

member. zero and the

satisfied.

of the (d) If several values

dependent variable have been

the given differential equation, for if the sum of the values their sum will satisfy the equation in the equation, each term of the sum in question is substituted

found each of which

satisfies

174

HARMONIC FUNCTIONS.

[CHAP. V.

will give rise to a set of

terms which must be equal to zero, and therefore the sum of these sets must be zero.
(e)

generally possible to get by some simple device solutions of such differential equations as those we particular have collected in Art. I. The object of this chapter is to find
It
is

methods
any

of so

combining these particular solutions as to

satisfy

given conditions, which are consistent with the nature of

the problem in question.

This often requires us to be able to develop any given func-

which enter into the expression of these conditions in terms of normal forms suited to the problem with which we happen to be dealing, and suggested by the form of
tion of the variables

particular solution that

we

are able to obtain for the differential

equation.

These normal forms are frequently sines and cosines, but they are often much more complicated functions known as
Legendre's Coefficients, or Zonal Harmonics; Laplace's .Coefficients, or

drical

Spherical Harmonics; Bessel's Functions, or CylinHarmonics; Lame's Functions, or Ellipsoidal Haretc.

monics;

Art.

3.

Problem

in

Trigonometric Series.
problem
:

As an

illustration let us consider the following

large iron plate

n centimeters

thick

is

heated throughout

to a uniform temperature of ioo degrees centigrade; its faces are then suddenly cooled to the temperature zero and are kept
at that temperature for
5

seconds.

What

will

be the tempera-

ture of a point in the middle of the plate at the end of that

time?

Given

a*

=0.185

in

C.G.S. units.

Take the
and the axis

origin of coordinates in
of

one face of the plate perpendicular to that face, and let u be the
in

temperature of any point


ing begins.

the plate

seconds after the cool-

suppose the flow of heat to be directly across the plate so that at any given time all points in any plane parallel
shall

We

ART.

3.]

PROBLEM

IN

TRIGONOMETRIC

SERIES.

175

to the

faces of the plate will have

the same temperature.


;

Then

it

depends upon a single space-coordinate x


(4),

o and

=0, and

Art.

I,

reduces to

^=^.
Obviously,

(1)

and and we need


ditions

= u = u =

ioo

when

o,

(2)
(3)
;

o when ^ o when
.r

= o, =
?r

(4)

to find a solution of (1) which satisfies the con-

(2), (3),

and

(4).

We
we
tion
is

shall begin

shall use a device

linear

by getting a particular solution of (1), and which always succeeds when the equaand homogeneous and has constant coefficients.
u

Assume*
tute in
(1)

v', where fi and y are constants; substiand divide through by e^ x+yi and we get y = # 2 /f

eP

and

if

this condition is satisfied,


efi*+**fi*t
/?.
i

x eP +y* is

a solution of

(1).

then a solution of

(1)

no matter what the

value of

11

We can modify the form of Let fi = fxi,\ then u = e^ ^ n e^ = e~ ^^'e~


a
"'**'

this solution
is

with advantage.
is

xi

a solution of (i\ as

also

By

(d),

Art.

2,

u
is

= e-+**
is

e~ a

W cos px
2<

(5)

a solution, as

also

*- "m*

f.

= *-V

Sin

^
It

6)

and

/* is

entirely arbitrary.
must be tested

* This assumption must be regarded as purely tentative.

by substituting
f

in the equation,
i

and

is

justified

if it

leads to a solution,

The

letter

will be

used to represent 4/

1.

176

HARMONIC FUNCTIONS.

[CHAP. V.

By giving different values to p. we get different particular solutions of (i) let us try to so combine them as to satisfy our conditions while continuing to satisfy equation (i).
;

it px is zero when x = o for all values of n if yu is a whole number. If, then, we write is zero u equal to a sum of terms of the form Ae~ aimn sin mx, where m is an integer, we shall have a solution of (i) (see {d), Art. 2)

e~ a ^ n sin

jjl

when x =

which

satisfies (3)

and

(4).

Let
u

this solution
ait

be
4ait

=A e~
1

sin

x
.

+ A,e~
(7)

sin

2x

+ A e~
3

9an

sin

3*

+ ...,

(7)

A lt A %i A tf
When
u
If
t

being undetermined constants.


reduces to
;r

= o, =A

sin

+ ^s sm

2;r

+ ^3 sin
A^

$x

+
.

(8)

form
series

now it is possible (8) we have only

to develop unity into a series of the

to substitute the coefficients of that

each multiplied by 100 for

l%

in (7) to

have
(2)>

a solution satisfying (1)


(3),

and

all

the equations of condition

and

(4).

We

shall

prove

later (see Art. 6) that

sin

" sin

^x

~
-f-

sin

%x

for all values of

x between o and x

n.

Hence our

solution

is

^[f-^ sin

+ -e~
of u

9aH

sin

3*

+V

25 " 2 '

sin

5*

. . .
|

(9)

To
to

get the answer of the numerical problem

we have only
seconds.

compute the value


is

when x

and

As

there

no object

in

place tables will

more than

going beyond tenths of a degree, foursuffice, and no term of (9) beyond


Since sin 2
71

the

first

will affect the result.

I,

we have

to

compute the numerical value

of

Art.

4.]

problem

in

zonal harmonics.
a3
10

177

400
7t
7 log a

e~ an

where

= 0.185

and

5.

log

log eft log log e & &


log log log
If
9'

= 9.2672 = 0.6990 = 9.9662


Q.6378 y j/

log 400

2.6021

colog
10 10

colog e**
.

=
=

9.5059
9.5982
,

10 10

^ =
f"

9.6040

10

g ^

=1.7062
50- 8.

= 0.4018

of

n centimeters

the breadth of the plate had been c centimeters instead it is easy to see that we should have needed

the development of unity in a series of the form

.
x

nx

sin

+ A, sin
,

2nx

+ A, sin
,

%nx

+ ....

Prob. 1. An iron slab 50 centimeters thick is heated to the temperature 100 degrees Centigrade throughout. The faces are then suddenly cooled to zero degrees, and are kept at that temperature for 10 minutes. Find the temperature of a point in the middle of the
slab,

and of a point 10 centimeters from a face

at the

end of that

time.

Assume
.

that
.

4/ - sin

ttx

r-

\7tx

sin

~ sin

?rx
,

from x

o to

c.

Ans. 84.o; 49.4.

Art.

4.

Problem in Zonal Harmonics.


let

As

a second

example

us consider the following problem

Two

equal thin hemispherical shells of radius unity placed together to form a spherical surface are separated by a thin
air. charge of statical electricity is placed upon one hemisphere and the other hemisphere is connected with the ground, the first hemisphere is then found to be at poten-

layer of

tial 1,

At what

the other hemisphere being of course at potential zero. " field of force" due to potential is any point in the
shall use spherical coordinates

the charge?

We

and

shall let

Fbe

the

potential required.

Then Fmust

satisfy equation (5), Art.

178

HARMONIC FUNCTIONS.

[CHAP. V.

But since from the symmetry


independent
of 0,
if

we take

of the problem is obviously the diameter perpendicular to the


-

plane separating the two conductors as our polar axis,


zero,

is

and our equation reduces to

rftrV)
dr"

9
.

Si (

"^ sin 6

"4) _
dV
r

(l)

Fis given on the surface

of our sphere, hence

V=/(0)
where /(d)

when
and f(6)

I,

(2)

if

< 6 < -,

=O

if

<6<

ir.

Equation
an

(2)

infinite distance

and the implied conditions that V is zero at and is nowhere infinite are our conditions.

To
which
is

find particular solutions of (1)


is

we
is

shall use a

method
where

generally effective.

Assume*
6,

that

V = RG
we
get

a function of r but not of


r.

and

a function of 6 but

not of

Substitute in (1) and reduce, and

rd\rR)
dr*

d \ sm

6
-de)
.

(3)

R
Since the
first

sin 6

dd
and the

member

of (3) does not contain 6

second does not contain r and the two members are identically Let us call this equal, each must be equal to a constant.
constant, which
is

wholly undetermined, m{in

-\- 1)

then

R^
r

d%rR)

-w^ro-^r- = m{m+l);
1
,

s '"

de)

whence

,'

m(in

-\-

i)R

= o,

(4)

dJ smedQ\
\

and

lel

^Te'-^e
* See the
first

+ *0*+i)e. = o.

(5)

foot-note on page 175.

Art.

4.]

PROBLEM
(4)

IN

ZONAL HARMONICS.

179

Equation

can be expanded into

d'R
r--j-i

dR + 2r^- - m(m +1)^ =


,
,'

0,

and can be solved by elementary methods.


solution
is

Its

complete

R
Equation
(5)

=.

Arm + Br~ m -\

(6)

can be simplified by changing the independ-

ent variable to

x where x
'

cos

0.

It

becomes
1

( dx\J r[

^S] + m{m +

)&

= 0l
is

(7)

an equation which has been much studied and which


as Legendre's Equation.

known

We
tive
(7)

;//, which is wholly undetermined, to posiand we can then get particular solutions of whole values,

shall restrict

by the following device

Assume*
coefficients.
.

that

can be expressed as a

sum

or a series of

terms involving whole powers of x multiplied by constant

Let

2anxn
i)a n x
n -'J

and substitute
n(n
-f-

in (7).
M

We

get
n

2[n(n

i)<V

+ m(m + i)a x =
n
]

o,

(8)

where the symbol 2 indicates that we are to form all the terms we can by taking successive whole numbers for n.
Since
(8)

must be true no matter what the value


any given power
of x, as for instance

of x, the

coefficient of

xk must
,

vanish.
(k

Hence

+ 2){k + i)a - k(k -\-j)a + m(m + i)a =


k+t
k k

o,

and
If

a k+ ,

m{m+i) k{k-\-i) * (k+l){k+2 )

fe)

now any

taken,
If

= k = m,

set of coefficients satisfying the relation (9) k 1Ea kx will be a solution of (7).

be

then ak+%

= o,
first

a k+i

= o,

etc.

* See the

foot-note on page 175.

180
Since
it

HARMONIC FUNCTIONS.
will

[Chap.

V..

answer our purpose


.

if

we

set of coefficients that will

obey the condition

pick out the simplest: (9), we can take

a set including a m Let us rewrite

(9) in

the form
.

**

~ ~ (*+2)(*+ !)*, + - k)(m + k (m =m


1)

( IO>

I)'

We
*""
Um - A

get from (10), beginning with k

2,

-~

m(in

2\2m-i)
_ ~ m(m
\){m

a""
2)(m
3)

~2.4.(2m- i)(2^-3)

*m%
$)(m

""- ~~
If
is

~ m(m
is

i)(m

2)(m

2.4.6. (2m
even we
a. t .

i)(2m

- z){2m -

4)(m
5)

5)

""

'

etC

'

m
am

see that

the set will end with a

if

odd, with

xm L

mitn
2. {2m
I

1)

\x
1)

9 m-.
,

mjm '

i)(m

- 2)(m - 3)
i)(2m
3)

2. 4.

^_ _
t

n
"'J'

(2m

where a m is entirely arbitrary, is, then, a solution found convenient to take a m equal to

of (7).

It is

(2m
and
it

i)(2w

3) ...

will

be shown later that with this value of a m


1.

when

.r

=
.

@
than

is

a function of

x and

contains no higher powers of

xm

It is usual to write it as

Pm (x).
Pm (x)
from the

We
formula

proceed to write out a few values of

P (** ^-W- <

2m -

)(

2m

- 3)
\

r 1m
.

Mm2.

1)

[J

(2m -if (2m 1)

*f(

i)(m-2)(m-s)
i)(2m
3)

^ m _<

,j jx

2.4.

(2;;z

ART.

4.]

PROBLEM IN ZONAL HARMONICS.


have
I
:

181

We
Pjx)
x

= or P (cos 6) = P {x) = x or P^cos 6) = cos - I) or P (cos 6) = J( 3 cos Plx) = (3* P (x) = i($x> - 3*) or P (cos 60 = i(5 cos 0- 3 cos 6\ (12) ^W = K35^ -30^ +3) or P (cos 0) = 4(35 cos # - 30 cos + 3), ~ 70x* + 15*) or ^.(*) = K6 3* = K63 cos 70 cos #+ 15 cos /'.(cos 0) We have obtained Q == /^(.ar) as a particular solution of (7), and Q = Pm (cos 6) as a particular solution of (5). Pm {x) or Pm (cos 0) a new function, known as a Legendre's Coefficient,
I,

0,

i),

I-

6>

61

6). j

is

or as a Surface Zonal Harmonic, and occurs as a normal form


in

many important

problems.
is

V'=
is

rm

Pm (cos 6)
6)

a particular solution of

(1),

and r m Pm (cos

6)

sometimes called a Solid Zonal Harmonic.


Q

V = A P (cos
satisfies (1), is

+ A rP {cos + A^P^cos + AyP (cos6)+...


l

6)

6)

(13)

not infinite at any point within the sphere, and

reduces to

V=A P (cos + ^(cos + A,P,(cos + A P (cos6)+...


0)

6)

6)
3

when

(14)

1.

v= Af,(cos6) r
satisfies (1), is

AfXcosd)
r*

PJjcos0) z r

~*

r-.-

(15)
is

equal to zero
If

when

not infinite at any point without the sphere, 00 and reduces to (14) when r r 1.

then
(14),

form

we can develop f{6) we have only to put

[see eq. (2)] into a series of the

the coefficients of this series in

A,, get the value of Ffor a t within the sphere, and in (15) to get the value of Fat a point point without the sphere.
place of the
,

... in (13) to

182

HARMONIC FUNCTIONS.
shall see later (Art. 16, Prob. 22) that
if

[CHAP. V.

We
o

f(0)

for

<

<

and f{6)

o f or

<

<

n,

Hence our required


l

solution

is

V= - + \rP,(cos ti)-\\ -r'P (cos


s

(f)

+ I?- 2
at

-^

C0S

^---

<">

an internal point

and

V=

^+4 r \-\P 2r
=

7 {cosB)-- 8 2 r ---\pU:ose) J 3V

+I7-^?^C0S *)
?t an external point.
If

< l8 >

and

= o,

(17) reduces to

To two
is

decimal places

F=

0.68,

and the point r

=
4

o-

at potential 0.68.
If r

and

(18)

and Table

I,

at the

end of

this,

chapter, give

and the point r


If

5,

7t

is

at potential 0.12.
is

the radius of the conductor


in (17)

a instead of unity,
(18).

we

have only to replace r by

and

Art.

5.]

PROBLEM
2.

IN BESSEL's FUNCTIONS.

18o

Prob.

ameter

is

One half the surface of a solid sphere 12 inches in dikept at the temperature zero and the other half at 100 de-

grees centigrade until there


at

is no longer any change of temperature within the sphere. Required the temperature of the any point center of any point in the diametral plane separating the hot and
;

cold hemispheres
axis of

symmetry

of points 2 inches from the center and in the and of points 3 inches from the center in a di-

ameter inclined at an angle of 45


Ans. 50
;

to the axis of symmetry.


;

50

73.9; 26.i

77.i

22.q.

Art.

5.

Problem

in Bessel's Functions.
:

As a last example we shall The base and convex surface of

take the following problem a cylinder 2 feet in diameter and 2 feet high are kept at the temperature zero, and the upper base at 100 degrees centigrade. Find the temperature of a

point in the axis one foot from the base, and of a point 6 inches from the axis and one foot from the base, after the permanent
state of temperatures has been set up.
If

or, representing the temperature by u and observing that from the symmetry of the problem u is independent of 0,
;

base

we use cylindrical coordinates and we shall have to solve equation (6),

take the origin


Art.
1

in

the

(1)

184

HARMONIC FUNCTIONS.
;

'

[CHAP. V.

contain r

therefore the
is

first

member
we can

cannot.
write
(5)

Hence each

member

of (5)

a constant, and

R
when
7

~dr*

+ rR
d*Z -dz

dr~~

Z dz

=~
1

'

jj

is

entirely undetermined.

Hence
and

^Z
x

o,

(7) x 7

d*R

\dR _+__ +
x

,*

= 0L
general solution
is

(8)

Equation

(7) is easily solved,


~

and
2
,

its

Z = ^^* + ite ^ Z = C cosh (/*#)


We can
reduce
(8) slightly

or the equivalent form


Z> sinh
(/*#).

-(-

(9)
it

by

letting //r
.

= x, and

becomes
,
<

d*R

u + * =sa 25?+52F
.

\dR

I0N >

Assume, as in Art. 4, that 7? can be expressed in terms of whole powers of x. Let R = 2a n x n and substitute in (10).

We

get

2[n(n

\)a n x

n "

+ na x
n

n -

-J-

<V M] = o,
x.

an equation which must be

true,

no matter what the value of


of x, as

The

coefficient of

any given power


i)a k
a

~*,

must, then,

vanish, and

k(k
or
-

+ ka + a ^ + ^_ =
h
2

= o,
(11)

o,

whence we obtain

^_ =
3

k*ak

as the only relation that need be satisfied by the coefficients in order that 2akxk shall be a solution of (10).

R=

If

= o,

a k _^

0,

ak _

0,

etc.

We

can, then, begin with

= o as

the lowest subscript.

Art.

5.]

186
If
i

HARMONIC FUNCTIONS.

[CHAP. V.

now we can develop

unity into a series of the form

S,/.(Mxr)
ITO

+ B,/,(M,r) + BJj^r) + ...,


/^" r
T
,

"=

\~B. sinh (/u.z)

Linnr(^

B sinh ujz) T / (lv)+ )+iiHhT2^N


,

"1
-

'

(I5)

satisfies (i)

and the conditions

(2), (3),

and

(4).

We

shall see later (Art. 21) that

it

/Ax)

= -

*& ax
J
(l6)

~
for values of

\w/wyiu&i + iUK) +
r

''

<

I.

Hence

=
is

200

r /.(^r)

sinh
sinh

(/i,*)

L^/,0^)
At

^
1

7.(w)

sinh

(/y )
"*"
* '

(2/i.)

^/X^)

sinh

(2//,)

J ^ 7>
"|

our required solution.


the point r

= o, z =
sinh

(17) reduces to
_i_

= 200 r

___iiiih
/i,/
1

^
(

L^/X/O
L/ij/,0*,)

sinh (2/1J

()"t)

sinh

2 /0

"J
'J*

cosh

"
/*,

/i./X/iJ'cosh

/7,

since
If

7 () =
we
III, at

and sinh

(2.ar)

2 sinh

^ cosh

x.

and

use a table of Hyperbolic functions* and Tables II the end of this chapter, the computation of the
is

value of u

easy.

We

have

/i(/0 = 0.5190
colog
"

^=2.405

= =/X^)
yU 2

5.52O

0.3402
9.2581
0.4683/*

= 9.6189 7<X> = 0.2848


/*,

10

colog
"

/*,=

10

/.(*.) =

M cosh /*,= 9.2530

10

coshyw a

7.9037

10
10

9.1567
* See Chapter IV, pp.
tions.

10

7.6301W

162, 163, for a four-place table

on hyperbolic func-

Art.

5.]

problem in bessel's functions.


(/*i/(/0 cosh Mi)' 0*i/i(/O cosh

= ^y =
1 1

O.1434
0.0058

0.1376;

a=i3.8

At

the point r

= J, z =
7^>

I,

(17),

reduces to

,,-icef

~
I

^^
/i,/, (^9 )

-I/i f

L MxJiiMd cosh

/*,

cosh

1 J

/ (/0 = 0.6698 log/(/0 = 9-8 2 59 colog MiA(Mi) cosh//,


9.1567
8.9826
/.(*/<.)

10
10
10

= - 0.1678

log /.(i/O colog A/AM*) cosh /I,

188

harmonic functions.

[chap. v.

Art.

6.

The Sine
3, it is

Series.

As we have

seen in Art.

sometimes important to be

able to express a given function of a variable, x, in terms of sines of multiples of x. The problem in its general form was first

solved by Fourier in his " Theorie Analytique de la Chaleur"

and its solution plays an important part in most branches of Mathematical Physics.
{1822),

Let us endeavor to so develop a given function of x,/{x), sin x, sin 2x sin $x, etc., that the function and the series shall be equal for all values of x between o and n.
in

terms of

We
fix)

can of course determine the coefficients a lf # a a %


,

an

so that the equation

=a
;

sin

+#

sin

2x

-f-

sin

$x

-f~

-|-

an

sin

nx

(1)

shall hold

O and n

any n arbitrarily chosen values of x between good for we have only to substitute those values in turn
for

in (1) to get

n equations of the

first

degree, in which the n co-

efficients are the

For instance,
3 Ax,
<i).
.

only unknown quantities. we can take the n equidistant

values Ax, 2Ax,


lot

nAx, where

Ax

=
n-\1
a

and substitute them

in

We get /{Ax) = a

sin

Ax + #
Ax -f- ^

sin

2Ax -f- a
-f-

sin
sin

$Ax
nAx,
6 Ax

an
s

J[2Ax)

=a

sin 2

f{$Ax)

=a =a

sin 3

Ax -f-

+ 2nAx, + a^ sin 6Ax + a sin gAx + a sin $nAx,


a

sin

^Ax -f- a

sin

an
3

sin

1. .

(2)

-\-

/(nAx)

sin

n Ax

-f-

a^ sin

2nAx

-f- <^ 3 -f-

sin 3z/jtr
sin

-4-

an

M^r,

^ equations of the first degree, to determine the n coefficients a a t a% an Not only can equations (2) be solved in theory, but they can be actually solved in any given case by a very simple and
x ,
,

Art. 6

the sine

series.

189
a,n

ingenious method due to Lagrange,* and any coefficient be expressed in the form

can

am

--T-2E A kAx)

K ~n

sin (*

A*)-

(3)

If

now n

is

indefinitely increased the values of

for

which

(i) holds good will come nearer and nearer to forming a conand the limiting value approached by a m will tinuous set probably be the corresponding coefficient in the series required
;

to represent fix) for

all

values of

x between

zero and n,

Remembering
question
is

that (n

+ \)Ax = n,

the limiting value in

easily seen to be

am

= - ff{x) sin

tnxdx,

(4)

This value can be obtained from equations (2) by the lowing device without first solving the equations
:

fol<

Let us multiply each equation in (2) by the product of Ax and the coefficient of a m in the equation in question, add the form of the resulting equation equations, and find the limiting
as n increases indefinitely.

The
sin

coefficient of

any
.

a,

aK

in

the resulting equation


sin sin

is

kAx

sin

mAx Ax + sin
-\- sin

2k Ax uk Ax

2mAx Ax -\nmAx Ax,


. .

Its limiting value, since (n

+ i)Ax =
sin
.

ar,

is

/
but
it

sin

kx

mx dx

f sin kx sin mx.dx = i/*[cos (m


if

k)x

cos(^

+ *)x]dx = o

and k are not equal.


Riemann's Partielle Differcntialgleichungen, or Byerly's Fourier's and Spherical Harmonics.

* See
Series

190

HARMONIC FUNCTIONS.

[CHAP. V.

The
A*(sin

coefficient of a m
3
2

is
2

mAx -j- sin 2mAx -\- sin imAx -\is

+ sin

nmAx).

Its limiting value

ysin
o

w^ .dx =
2

The

first

member
sin
;;/

is

/{Ax)

J^r

Jjr -\-fi2Ax) sin


+/(/zz/;tr) sin

2m Ax Ax +
.

mnAx

z/^r,

and

its

limiting value

is

fix) sin

mx.dx.
by the
final

Hence the
as n
is

limiting form approached


is

equation

increased

/ fix) sin

mx dx =
.

# wt

Whence
as before.

tf

= -J

fix) sin

mx.dx

(5)

This method
equation
f{x)

is

practically the

same
a

as

multiplying the

=a

sin

-f-

a^ sin

2x

-f-

sin

$x

(6)
tt.

by

sin w;r. akr


It is

and integrating both members from zero to

this article are in

important to realize that the considerations given in no sense a demonstration, but merely estab-

lish a probability.

elaborate investigation * into the validity of the development, for which we have not space, entirely confirms the results n the o and x formulated above, provided that between x

An

* See Art. 10 for a discussion of this question.

Art. 6

the sine
is

series.
infinite

191

and single-valued, and has not an ber of discontinuities or of maxima or minima.


function
finite

num-

It is to

be noted that the curve represented by

y=fix)

need not follow the same mathematical law throughout its length, but may be made up of portions of entirely different
curves.

For example, a broken

line or a locus consisting of

finite parts of several different

and disconnected straight

lines

can be represented perfectly well by

a sine series.

As an example
Here

of the application of formula (5) let us take

the development of unity.


fix)

1.

am

=
.

/sin

mx dx
.

/[
/
sin
ir

mx ax =
T

cos

mx
.

sin

mx .dx

m (1
if

cos nnt)

w
(

m [1

l) ] J

m
if

is

even
is

=
Hence
I

m
.

odd.
sin

4 /sin x
(

$x + -^-+ -sin

xx

+^
sin
,

Jx

+...).

(?)

It is to

be noticed that
c.

(7)

gives at once a sine development

for

any constant

It

is,

_4f/ sin;r
Prob.

sin

3*

s in

$*

..

4-

Show
Tsin
2\

that for values of

x between zero and


xx
sin

7t

w
,
.

(a)

x
.

=
.

sin
2
jc

2x

Li

+
.

sin

4X

4
sin

J
.

(*)/W =
.,.

4 Tsin

sin

![_

r+1
,

%x

sin 5 a:

nx

+---J

192

HARMONIC FUNCTIONS.

[CHAP. V.

if

f(x)

= x f or o < x <

-, and f(x)

=n

f or

- < x

<

tt.

w /(*) =
2

Tsin

jc

2 sin

2.2:

sin

xx

sin s.x

2 sin 6jc

sin nx

if

fix)
(</)

for o

<x<

71
,

and /(#)

7t

o for
2

<x <

/r.

sinh
sinh

x
tt

=
sin
2

n
(e)

x
5

sin

2x

-\

sin

xx
17

sin

ax

10

x*

=
?7 sin
x

Z[\t

7 sin
7.

~
\J
?J

sin 3

*~

sin

*+
-J*

Art.

The Cosine

Series.
in a series

Let us now try to develop a given function of x


of cosines, using the

method suggested by the

last article.

Assume
j\x)
b^-\-b cos
x

x -\- &,

cos 2x -\-b % cos 3^


(1)

-f-

(1)

determine any coefficient b m multiply and integrate each term from o to n.


IT

To

by cos

mx .dx

I b cos
w

mx

dx

= o.
and k are not equal.

I b k cos

kx cos

w^ dx=o,
.

if /

/7t
b m cos
2

#z;tr

dx

#w

if

#z is

not zero.

it

Hence
if /

w =

/ / f(x)

cos

mx ,dx>

(2)

is

not zero.

Art.

7.]

the cosine
get b multiply
(i)
a

series.

193
to n.

To

by dx and integrate from zero

Jb.dx
IT

Z>

7T,

1 b k cos

kx dx
.

= O.

IT

Hence
which
if

~Jf{x)dx

(3)

is just half the value that would be given zero were substituted for m.

by formula

(2)

To

save a separate formula

(1) is

usually written

f(x)

= ib + b

cos

cos 2x

+b

cos 3*

.,

(4)

and then the formula


ficients.

(2) will

give b as well as the other coef-

Prob.

5.

Show
7C

that for values of

x between o and
,

7t

w
,
.

* w

=
x

--!(-7-+-7
8 /cos 2X
l

/,n

(4) /[*).=

+^L + * _ ___^_ _+_- + ? _-+...


/COS
. ,

COS

7.X

COS KX

);

cos6x

cos io#

\
J,

if

/(*)
, .

=
..

a:

for o

<x<
2

and f{x)
COS 3^

=
,

7T

x
<3:

for
2

<

n;

1,2 W /M - j+jH
%

/COS

.#

COS
s

-)'
for

if

/(*)

for o

<x<

and fix)

=o
1

- < x <
2
7T -f- 1)

n^

(^) sinh

= 21
n

(cosh

n
2.2:

1)

(cosh

cos

-\

(cosh

1)

cos

(cosh
COS

-j-

1)

cos 3*

\;.

n*

/cos

COS 2X

3JC

cos 4X

194

HARMONIC FUNCTIONS.

[CHAP. V.

Art.
Since a sine series
of an

8.

Fourier's Series.
an odd function of
the development in such a series must hold good from

is

odd function
ft

of

x
it

=
is

to

ft,

except perhaps for the value


is

o,

where

easily seen that the series what the value of the function.
if

necessarily zero,

no matter
see that

In like manner
its
ft

we

f{x)

is

an even function of
valid from

development

in a cosine

series

must be

to

ft.

be developed into a Trigonometric n and n. series to which it is equal for all values of x between be the given function of x. It can be expressed Let/(^) as the sum of an even function of x and an odd function of x

Any

function of

x can

by the following device

jy)=
=Q identically; but

Mz* + M^k=Jl
'
'

(I)

-A

no t changed by reversing

the sign of

x and

is

therefore an even function of x\ and

when

we

reverse the sign of x,


its

^
2

is

affected only to the

extent of having function of x.

sign reversed,

and

is

consequently an odd

Therefore for
J\ x ) -h J\
j

all

values of

x between x
-\-b % cos

n and n
2x

_ _^ =
ft t/
o

_|_

ft

cos

+
.

cos ix

where

bm

cos

mx dx

and

A*) -i

A
-

) '-

=a

sin

x -f- #

,..-, +
3

sin

2x

^ 3 sin $x

-f-

where

am

2 ft

}f(x)j\^^
2

X) sin
.
'-

mx

.
.

dx.

Art.

8.]

Fourier's series.
little.

195

b m and a m can be simplified a


bm
71

Jy

'

'

U/
It

X 2

^-cos

mx. dx
It

j
o

Ax )

cos

mx dxJ
;r,

rJA
o

x)

cos tnx d-r


.

;
|

-J

but
ir

if

we

replace

^ by

we
f{

get
cos

jA

x) cos

w* dx=
w

x)

fnx.dx^

JAX
*

Q0S inx.dx,

fl-

an d we have
In the same

I
It

Ax

cos

w;lr

dx

way we can reduce


it

the value of a m to

I
it

Ax
+
-\ir

sm mx >dx.

Hence

Ax = - +
)

&o

^i

cos *
sin

^a

cos

2;p

-|-

tf,

a^ sin

+ ^ cos 3^ -f 2# ^s sm 3 X +
-f~

>

2)

where

#w

J Ax
7T

cos

w;tr

d*>

(3)

and
and and
this
tc.

tf

sin ** w = l-fA*)

*
x between

(4)

development holds

for all values of

The second member


The developments
development
Prob.
e*
6.

of (2)

is

known

as a Fourier's Series.
are special cases of

of Arts.

6 and 7 x from
1

in Fourier's Series.

Show

that for
1

all

values of

to

n
.

2 sinh 7rf~i
7t

cos #-4
2

.1
5

cos

2.*

\_2

10

cos $x-\

.1
17

cos4#+..

~]

196
2
,

HARMONIC FUNCTIONS.
sinh
it

[CHAP. V.

Ti

sin

sin 2JC 4-

sin

3^

sin

4*

-\-

Prob.

7.

Show
*,

that formula (2), Art.

8,

can be written
COS (2* COS (3*
1

f(x)

=-

COS#,

+C
2

COS (X

fix)

+C +
<T

- A)
A) +

where.
Prob.
8.

cm

(,

)*

and

fim

= tan"
8,

-~"

Show

that formula (2), Art.

can be written
sin
(

f(x)

= ~c
2

sin /5 9

+c

sin (*

+ A) + +c
and
fim

'.

2X

+ A)
+ /? +
3)
.

sin (3a;
1

where

cm

{a

J + b,*)*

= tan'

Art.

9.

Extension of Fourier's Series.

In developing a function of
is

into a Trigonometric Series it

often inconvenient to be held within the narrow boundaries

=
Let

n and x
it

it.

Let us see

be required to Trigonometric Series which shall be equal to f(x) for c. c and x of x between x

we cannot widen them. develop a function of x into


if

all

values.

Introduce a new variable


z

7t

X,
c,

which

is

equal to
z
)

n when x

and to n when x

c.

f(x)
(2), (3),

= /(
and
(4).

can be developed in terms of z by Art. S y

We

have
cos *
sin

f[~z)

=
2

+*
-\-

+^
-f-

cos 2 *
sin
2^r

#2

+^ +^
.

cos $ z
sin

+ 3# +

(1)

where

bm

J f\zj cos mz

<&,

(2)

Art.

9.]

extension of Fourier's

series.

197

and
and
(1)

'

am

//(
n

A
n

sm

m%

dz

(3)

holds good from z


its

to z

n.
(1)

Replace z by
f(x)

value in terms of
cos
c

x and
f'

becomes

b u,t
-f-

cos
*
.

\-

0^

o 3 cos J

'

+ a sm+tf,sm
l

+ a, sin

+..,.;

(4)

and

(2)

and

(3)

can be transformed into

= =

y
c

/(*) cos

j-dx>

(5)

and

(4)

holds good from

ff( x x =
c

sin

^r-dx,
x

(6)

c to

c.

In the formulas just obtained c may have as great a value as we please so that we can obtain a Trigonometric Series for

f(x) that

will

an interval as we
It

be equal to the given function through as great may choose to take.


if

can be shown that


00

this interval c
its

is

increased indefi-

nitely the series will

approach as
00

limiting form the double


is

integral

I f(X)d\ I cos a{X


00

x)da, which

known

as a

Fourier's Integral.

So that
+00
so

/(*)

-~f f(X)d\ f cos at{\

- x)da

(7)

for all values of x.

of

of Fourier's Integral and for examples use in Mathematical Physics the student is referred to Riemann's Partielle Differentialgleichungen, to Schlomilch's

For the treatment

its

Hohere Analysis,

and

to

Byerly's

Fourier's

Series

and

Spherical Harmonics.

198
Prob.
9.

HARMONIC FUNCTIONS.

[CHAP.

Show

that formula (4), Art.

9,

can be written

f(x)

-ct cos

+c

cos
(

Ay +A cos
3

^-^/?,j
3

+ , cos(^-/? )+...,
where
Prob. 10.

C*

= (" + bm y

and

/?,

=
9,

tan" 1

-.

Show

that formula (4), Art.


'1

can be written

/(*)

c% sin Po

sin

(^T

+ Aj +

'1

sin

(^ + AJ
-^.

where

^=

)*

and

/? w

tan -1

Art.

10.

Dirichlet's Conditions.

In determining the coefficients of the Fourier's Series representing f(x) we have virtually assumed, first, that a series of the required form and equal to f(x) exists
it is uniformly convergent ; the results obtained as only provisionally established.

and second, that and consequently we must regard


;

It

is,

however, possible to prove rigorously that


=.

finite

and single-valued from x

to

7t

if f(x) is and has not

an infinite number minima between x


(2),

of (finite) discontinuities, or of

maxima

or

n and x

n, the Fourier's Series of

Art.

for all

and that Fourier's Series only, is equal to f{x) n and ?r, excepting the values of values of x between
8,

x corresponding n and n and


;

to the discontinuities of f{x),

and the

values.

that

if

c is a

value of

corresponding to a

discontinuity of f{x), the value of the series

when x

iJQ[A* + )+/(
x

)];

and
is

that

when

= c is = n or

it

the value of the series

$'[/(*)

+/(

*)].

This proof was first given by Dirichlet in 1829, and may be found in readable form in Riemann's Partielle Differentialgleichungen and
in Picard's Traite

d'Analyse, Vol.

I.

Art.

10.]

dirichlet's conditions.
deal of light
is

199'

A good
we

thrown on the peculiarities of

trigo-

nometric by the attempt to construct approximately the curves corresponding to them.


series
If

construct

y=a

sin

x and y

=a

sin

2x and add the

ordinates of the points having the tain points on the curve

same

abscissas,

we

shall ob-

y=
If

a sin
1

x -f
%

a^ sin 2x.

now we

construct
x

y
a2

=a
sin
-f-

those

of y = a

sin

x
t

-f

sin ^x and add the ordinates to 2x we shall get the curve

y
By

=
x

sin

a9

sin

2x

-\-

a%

sin $x.

continuing this process we get successive approximations to

=a

sin

+#

sin

2*

tf

sin

$x

sin

^x

'

200
tion to this curve.

HARMONIC FUNCTIONS.

[CHAP. V.

In each figure the curve y the series, in question are drawn in continuous and the approximations

and the preceding approximation and the curve corresponding to the term to be added are drawn in dotted lines.
lines,

Prob. ii. Construct successive approximations to the series given in the examples at the end of Art. 6. Prob. 1 2. Construct successive approximations to the Maclaurin's
Series for sinh x,

namely x

x-\

x
-j

-|

Art.
(a)

11.

Applications of Trigonometric Series.

at potential zero,
tential
is

Three edges of a rectangular plate of tinfoil are kept and the fourth at potential I. At what po-

any point

in

the plate
solve
is

Here we have

to

which, since the problem

Laplace's Equation (3), Art. two-dimensional, reduces to

I,

a^+
subject to the conditions

W=
"

>

(I)

V=

o when x

V=o V=o v=i


V

"

= o, = a, y = o, y =
x
t-

(2)

(3)
(4)

(5)

Working as in Art. 3, we readily get sinh/?jj> sin fix, sinh fiy cos fix, cosh /3y sin fix, and cosh fiy cos fix as particular values of satisfying (1).

sinh

^
Cc

sin

^
CL

satisfies (1), (2), (3),

and

(4).

V = ^\
is

[-sinh

^Lsinh^ a

-^~+
'

sinh^
3 sinh

*1
it

*L*
a
1

"I
sin

+ T
when y

J
=
b.

(6>

the required solution, for (7), Art. 6.

reduces to

See

ART.

11.]

APPLICATIONS OF TRIGONOMETRIC SERIES.


harp-string
is
;

201

(b)

initially distorted into

a given plane

curve and then released

find its motion.

The differential equation for the small transverse vibrations of a stretched elastic string is

as stated in Art.

i.

Our conditions

if

we take one end

of

the string as origin are

y y
-^

o when

=
=

"

x *
/

o,

= =
=

(2)
(3 )
(4) (5)

/,

y
Using the method
of (1)

=fx
3,

"

o.

of Art.

we

easily get as particular solutions

y j

= =

sin

#r
/for

sin

a fit,
tf/?/",

cos

sin

and

y = y =

sin

/far

cos

tf/^,

cos /?* cos a fit.

sin

j- cos
.

satisfies (1), (2), (3),

and

(4).

am

mnx
1

sin

cos

mnat
;
,

\p)

where

am

= j Cf{x) sin ^^
;

dx

(7)

is

our required solution


9.

for

it

reduces to/(*)

when/ = o. See

Art.

Prob. 13. Three edges of a square sheet of tinfoil are kept at


potential zero, and the fourth at potential unity Ans. 0.25. is the centre of the sheet ?
;

at

what potential

Prob. 14.

kept

what

Two opposite edges of a square sheet of tinfoil are at potential zero, and the other two at potential unity Ans. 0.5. is the centre of the sheet ? potential
at
;

Prob. 15.

Two

adjacent edges of a square sheet of tinfoil are

202

HARMONIC FUNCTIONS.

[CHAR V.

kept at potential zero, and the other two at potential unity. what potential is the centre of the sheet ? Ans. 0.5.
Prob. 16.
harp-string
is

At

Show

that

if

a point whose distance from the end of a


is

-th the length of the string

drawn aside by the

player's finger to a distance b from its position of equilibrium and then released, the form of the vibrating string at any instant is given

by the equation
2bn
(n
1

^
)

nnt

mnx

sin " n* ^-* Km* n

sin

cos

m7iat\ mnat\

>

Show from this that all the harmonics of the fundamental note of the string which correspond to forms of vibration having nodes at the point drawn aside by the finger will be wanting in the complex note actually sounded.
iron slab 10 centimeters thick is placed between and two other iron slabs each 10 centimeters thick. The temperature of the middle slab is at first 100 degrees Centigrade throughout, and of the outside slabs zero throughout. The outer faces of the outside slabs are kept at the temperature zero. Required the temperature of a point in the middle of the middle slab fifteen minutes after the slabs have been placed in contact. Given a % = 0.185 m C.G.S. units. Ans. io.3.

Prob. 17.*

An

in contact with

Two iron slabs each 20 centimeters thick, one of which temperature zero and the other at 100 degrees Centigrade throughout, are placed together face to face, and their outer faces are kept at the temperature zero. Find the temperature of a point in their common face and of points 10 centimeters from the common face fifteen minutes after the slabs have been put together. Ans. 22.8 15.! ; i7.2.
Prob. 18.*
is

at the

Art.
In Art.

12. t

Properties of Zonal Harmonics.

as a particular solution of Art. 4] by the device of assuming Legendre's Equation [(7), that z could be expressed as a sum or a series of terms of
4,

= Pm {x) was obtained


then

the form a n x n and


* See Art.
f
3.

determining the

coefficients.

We

The student should review

Art. 4 before beginning this article.

Art.

12.]

PROPERTIES OF ZONAL HARMONICS.

203

can, however, obtain a particular solution of Legendre's equation

by an

entirely different

method.

potential function for any point (x, y, z) due to a unit of mass concentrated at a given point (xx ,yx% ,) is

The

V=
and
[(3),

(I)

this

Art.

must be a particular solution of Laplace's Equation is easily verified by direct substitution. 1], as
transform
(i)

If

we

to spherical coordinates
1

we

get

V=

Vr

=
sin 6 sin d x cos

2rr,[cos

cos B

-\-

(00,)]

+r

%
x

(2)

as a solution of Laplace's Equation in Spherical Coordinates


[(5),

Art.

i].
x >
x

the given point (x xt y z ) is taken on the axis of X, as it must be in order that (2) may be independent of 0, 0, O, and
If

^=
is

r* V,-,

2rr, cos
4.

.-v

+r

.
x

13)

a solution of equation (1), Art. Equation (3) can be written

^('"V
and
if

08

*+)"*'

(4)

is

less

than

r,

2^1

cos 6

can be developed
- be this series,
1

into a convergent

power

series.

Let 5?/
6.

/,

being of course a function of


(1),

Then

V=2pm r
I

w
is

solution of

Art.

4.

in the equation, and rememSubstituting this value of bering that the result must be identically true, we get after a

slight reduction

204
but, as

HARMONIC FUNCTIONS.

[CHAP. V.

we have

seen, the substitution of


[(7),

= cos

6 reduces this

Art. 4]. Hence we infer that the coefficient of the mth. power of z in the development of i 2xz-\-z*)~ iz a function of x that will satisfy Legendre's >(i

to Legendre's equation

^equation.
(i

2XZ

+ Z )"* =
2

[i

2(2X

*)] -*,
;

and can be developed by the Binomial Theorem the coefficient m of z is easily picked out, and proves to be precisely the function of x which in Art. 4 we have represented by Pm (x), and have called a Surface Zonal Harmonic.

We
if

have, then,

t l -2xz+S)-*=P,{x)+Pl(x).*+P,(x).S+PJLx).e+.. (5)
the absolute value of z
If
(I
is

less

than

I.

x=

I, (5)

reduces to
9
l
.

_ 2g+ j?)-i = P (l)+P (t).i+PJil).S+ />(!) *> + but (1 - 2*+ f)-*=(i--*)- =i+s+ J +*>+...;
1

hence

Pm (l) =
Surface Zonal Harmonic

I.

(6)

Any
two
of

may

be obtained from the

next lower orders by the aid of the formula


(n

+ l)PnU*) ~ (2* + WU*) +


1

Pu-

(*)

= O,

(7)

which

is

easily obtained,
is

and

is

convenient when the numerical

value of x

given.
z,

Differentiate (5) with respect to


(

and we get

il

z L+* = p>w +2PM -' +3PM'*+----

Avhence

^1

2XZ-\-S y

or by
(I

- 2*8 +
+ (s

(5)

+ Z/fc) .* + 3/>,(*) ." x){P,(x) + />,(*) * + ^.(*)


)(/,(*)
*

+)=

0.

(8)

ART.

13.]

PROBLEMS
(8)
is

IN

ZONAL HARMONICS.

205

Now
power

of z

must vanish.

identically true, hence the coefficient of each n Picking out the coefficient of z and

writing it equal to zero, we have formula (7) above. By the aid of (7) a table of Zonal Harmonics is easily com-

puted since we have


for

P (x) =

1,

and

P {x) =
x

x.

Such a

table

cos

is

given at the

end of

this chapter.

Art.

13.

Problems in Zonal Harmonics.

In any problem on Potential if Fis independent of sa that we can use the form of Laplace's Equation employed in
Art. 4, and
if

the value of

Fon
or as

the axis of ^fis known, and

can be expressed as

2am rm

^ 3+I> we

can wr ite out

our required solution as

V=2am r~Pm (cos8) each of for since Pm(i) =


1

or

V^^ ^'^

value on the axis


satisfies

and as we have seen

these forms reduces to the proper in Art. 4 each of them

the reduced form of Laplace's Equation.


let

As an example,
of electricity placed
lar disk,
tial

us suppose a statical charge of units on a conductor in the form of a thin circu-

be required to find the value of the PotenFunction at any point in the " field of force" due to the
let it

and

charge.

The
from
its

surface density at a point of the plate at a distance

centre

is
<T

M
4#7T Va*
.

and

all

points of the conductor are at potential


(

See Pierce's

Newtonian Potential Function

61).

The

value of the potential function at a point in the axis

ot the plate at the distance

x from

the plate can be obtained

without difficulty by a simple integration, and proves to be

V = M cos-*.x
2a

x*

+ a*
,

v ' (1)

206

HARMONIC FUNCTIONS.

[CHAP. V.

The second member


series.

of (i)

is

easily developed into a

power

M cos
2a

x* x*

- a*

+a
.

MYn

x*

x*

;.

Hence

-lJi>.(cos )+-...] 5
is

(4)

our required solution

if

<a
<0

and

<

-, as

is

F=

jj-

- VP, (cos + -

P (cos
t

0)

-lJ/>.(cos)

+ ...]ifr>,

(5)

The series in (4) and (5) are convergent, since they may be obtained from the convergent series (2) and (3) by multiplying the terms by a set of quantities no one of which exceeds one
in absolute value.

Pm (cos
The

6)

always

lies

For it will be shown in the next 1. between 1 and

article that

Prob. 19. Find the value of the Potential Function due to the attraction of a material circular ring of small cross-section.

value on the axis of the ring can be obtained by a simple

integration,
ring.

and

M
=

is

if

is

the mass and c the radius of the


c

At any point

in space, if r

<

Vm
and
if

y [/'.(cos
>
c

1 0)- -

p,(cos 6)

+ 1^ /,(* 60 -

],

ART.

14.]

ADDITIONAL FORMS.
I
2

207

T= - f-^ (cos 6) C
Lr Art.
14.

-3 P (cos 2

6)

+ 2 -^ r (cos 0) 0>
T
.

4V 4

'

.1 J

Additional Forms.

have seen in Art. 12 that m (x) is the coefficient of (#) m in the z 2xz-\- #*)-* in a power series. development of (i
(l

We

2XZ

~
Z*)

= =
01

[I

*(*'

+
K

e~
1

')

+
^*0

(I

**')"

]-* "*

we develop nomial Theorem


If
(i

(i

ze ')-*

and

(i

ze~ 9i)-l by the Bi-

their

product

will give a
is

development

for

2xz

+ z*)~l.

The

m coefficient of z

easily picked out

and reduced, and we get

PJcos

6)
1

=
I

1.3.5... (2m 3.4.6... 2m


i
'

i)

n 2C0sw,g

+2

Wl

i.(2 W -.)

C08

^-

2) g

3-

(*-)

I.2.(2W-l)(2W-3)

V cos(

_ 4) g + 17
I

1 J

(I)

is odd the parenthesis in (1) ends with the term conif is even, with the term containing cos o, but cos taining in the latter case the term in question will not be multiplied by the factor 2, which is common to all the other terms.

If

Since
positive,

all

the coefficients in the second


0)

member
#

of (1) are
its

Pw (cos

has

its

maximum

value

when

= o, and

value then has already been shown in Art. 12 to be unity. 1. Obviously, then, its minimum value cannot be less than

we integrate the value of Pm {x) given in (11), Art. 4, (b) If times in succession with respect to x, the result will be
differ

tound to

from

'3-5

^-^.

^ {2m)\
(

2m

0/
171th.

\x*

W m
i) '

t.

by terms m-

volving lower powers of

than the

Hence

PJfi

^^~

I*

(*)

208'

HARMONIC FUNCTIONS.
Other forms for

[CHAP.

VI.

(c)

Pm {x\
m
i)

which we give without demon-

stration, are

PJ

^~
=

m
>

dx

tv + y + /

i_

(3 >

IT

PJx)

\f\* +

Vx^i

cos <] V0.

(4)

P,(*)

= l/
^j/

#
[*

.
.

V -T

COS 0] m+1

(j)

(4)

and

(5)

can be verified without difficulty by expanding

and integrating.

Art.

15.

Development
as in Art. 4,

in

Terms of Zonal Harmonics.


the value of the Potential

Whenever,

we have

Function given on the surface of a sphere, and this value depends only on the distance from the extremity of a diameter,
it

of

becomes necessary to develop a function the form

of 6 into a series

Aflcos
or,

d)

+ ^(cos + Aflcos +
6)

0)

what amounts to the same

thing, to develop a function of

into a series of the form

AJ>.{?)

+ Aflx) + A fix) + ....


r

is entirely analogous to that of development treated at length in Art. 6 and may be solved by in sine-series the same method.

The problem

Assume
for
1
1

f(x)
1.

= A P (x) + A P {x) + A P (x) +


x

(1)

<x <
1.

Multiply (1) by

Pm (x)dx and
1

integrate from

to

We
1

get
_

^ P ff{x)Pm(x)dx "s\A. f m (x)P {x)dx\


n

(2)

Art.

16.]

formui as for development.


shall

209

We

show
i

in the

next

article that

j
-i
i

Pm (x)Pn (x)dx = o,

unless

m=n

and that

f[PJx)ydx -1

^.
1

Hence
It is

A m = ^-t- ff(x)Pm {x)dx.


in Art. 6, that

(3)

important to notice here, as

the method

we have used
mining

Am

obtaining so that the equation


x

in

Am

amounts

essentially to deter-

Ax)
and

==

A P {x) + A P {x) + A fix) +


x

+ A P {x)
n

shall hold
I,

good for n -\- I equidistant values and taking its limiting value as n

of
is

x between

indefinitely in-

creased.

Art.

16.

Formulas for Development.


in

We

have seen

Art. 4 that z

= Pm (x)
-f-

is

a solution of

Legendre's Equation

-j-\ (1

x*)

m(m

+ 1)2 =

o.

(1)

Hence
and

^[(l
(l

- x*)

^?]
(3)

+ m(m + i)Pm {x) =

o,

(2)

i[ -^-2^]+^ +I ^ W =a
Multiply
(2)

(3 >

by

PW

W and

by

Pm (x), subtract, transpose,


1

and

integrate.

We

have

\m{m + I) - n(n +

i)]fpm (x)Pn (x)dx


-1

210

HARMONIC FUNCTIONS.

[CHAP. V,

=[^K.-)"-f
-1

-^)<.-^'i

];,

-1

by integration by

parts,

o.
i

Hence
unless

fPm (x)P (x)dx =


n
-i

O,

(6)

m=

n.

If in (4)

we

integrate from

to

instead of from

to

\r

we

get an important formula.

Pm (x)P* (x)dX= n
"* v

'

/(;

r-=i)
I.

~,

n(n-\-

(7)

i)

and as a special

case, since P^-*-)

unless

m=
get

o.
i

To

f[Pm {x)Jdx

is

not particularly

difficult.

By

(2),

-1

Art. 14,
*

~i

flP4*)7d*^ttf
v
'

dx~

dx-

dx

9>

-1

By

successive integrations
a

by

parts, noting that

dx__(^

i)

m contains (x*

1)*

as a factor

if

k < m, and

Art.

16.J

formulas for development.


K

211

that

dx* m

= (2m)l

we

get
i

-1

*'

-1

f{x*
-i

i)

m dx

f(x
-i

m
i)

(x

+
i)

i)

m dx

=
m-\i

fix J
v

' i)

\ {x+ -r

m+1dx

{2tn)\

+U

2
(

+l

(m

\y
!'

I}

(2m

1)

Hence

-i

/W)]W* = -^rj.
'

<i I)

Prob. 20.

Show
m-\

that /

Pm (x)dx = o
x
-

if tn is

even and

is

not zero

(- i)"^

m\m-\-

3-5-7---M
\)

if

is

odd

.4.6

..

.(m

i)

Prob. 21.

Show

that^ [AC*)]V*
x.

2m \

'

Note that

[^(tf)]*

is

an even function of

Prob. 22.
,/(.*)

==

Show that if f(x) from x = o to x = x,

o from

to

= o,

and

Prob. 23.

Show

that i?(0)

= ^ Bm Pm (cos

0)

where

i?w

fF(d)Pm (cos

6) sin

dd.

212
Prob. 24.
esc 6

HARMONIC FUNCTIONS.

Chap

V-

Show

that (cos

y [1 +
Show
a
!

(j)V,

(?)

+ 9(^5) A(cos

0)

"+
.].

See

(1), Art. 14.

Prob. 25.

that
(2 "

**

=
1.3.5

.(a

+I )[

l}/>
<

+ (in - 7
1

+ (" - 3) 3 " + '* ~ '>


( *>

2JL

~^-

(x)

/>._,(*)

+ ...].

Note that

t/

fx Pm (x)dx
n

1
2 "V//

/
.-1

fx* ^
1

x
dx m
\'~

^"-dx,

and use the

_i

method

of integration by parts freely.

Prob. 26.
at

Show
and
if

that

if

V is

the value of the Potential Function

any point

in a field of force, not

imbedded

ling matter;

'

/{&) when r

in attracting or repel-

a,

V=2A m P M (cos6)if r<a


and

2jJ^Pm{a

V) if

r>

a,

where

Am =
Show

-^J-ff{0)P
2
if
;

nt

{cos 6) sin Odd.

Prob. 27.

that

V= c

when r

=a

F=

if

<#

and

F=
r

if

>

dr.

Art.

17.

Formulas

in

Zonal Harmonics.

The
tion

following formulas which we give without demonstramay be found useful for reference
:

^^= (2n-i)P..,(x)+(2n-s)PU')+(2n1

9)PU'}+----(i)

fp{x)dx =

_L_[ P,_ (*) - />+,(*)].


J 1

(3)

Art.

19.]

spherical harmonics.

213

Art.

18.

Spherical Harmonics.

In problems in Potential where the value of Fis given on the surface of a sphere, but is not independent of the angle 0, we have to solve Laplace's Equation in the form (5), Art. I, and

by

a treatment analogous to that given in Art. 4

it

can be

proved that

V = rm cos n<b
where
ju

sin

M
,

and

V=

rm sin nd> sin M 6

r^~^>

= cos

6,

are particular solutions of

(5),

Art.

1.

The

factors multiplied

by

r m in these values are


<p

known

as

Tesseral Harmonics.

They

are functions of

and

6,

and they

play nearly the same part in.unsymmetrical problems that the Zonal Harmonics play in those independent of 0.
YJji, 0)
is

= A.Pm(p) +

m=i

l (A

cos

+B

sin

0)sin

^*

(IfA.

P jM

known

as a Surface Spherical

Harmonic

of the #zth degree,

and

V=r~Ym

(fii,<p)

and

V = -^1
1.

(,

<p)

satisfy Laplace's Equation, (5), Art.

The

Tesseral and the Zonal Harmonics are special cases of

the Spherical Harmonic, as is also a form m (cos y) known as a Laplace's Coefficient or a Laplacian y standing for the angle between r and the radius vector r, of some fixed point.
;

refer

For the properties and uses of Spherical Harmonics we the student to more extended treatises, namely, to

Ferrer's Spherical Harmonics, to Heine's Kugelfunctionen, or

to Byerly's Fourier's Series and Spherical Harmonics.

Art.

19.*

Bessel's Functions.
in Art. 5 that
i

Properties.

We

have seen

= J {x) where

/.w = i-J + i^ - ir ? J^
* The student should review Art.
5

+ --.

(1)

before reading this article.

214
is

HARMONIC FUNCTIONS.

[CHAP.

a solution of the equation

^+^ + *=
.

d*z

dz

0:

(2 >

and we have called J&x) a Bessel's Function or Cylindrical

Harmonic

of the zero order.

JA) ~
is

dx

2L

2.

4^2. 4

2
.

2. 4
order,

.6\8^

Kl)

called a Bessel's

Function of the

first

and

is

a solution of the equation

dx*

+ x dx ^+(>-^=>
f

(4>

which

is

the result of differentiating (2) with respect to x.

table giving values of (x) and /,(^) will be found at the end of this chapter. If we write J&x) for z in equation (2), then multiply through by xdx and integrate from zero to x, simplifying the
resulting equation

by integration by

parts,

we
o,

get

xdj%(x)

or, since

J (x)
x

dx AT (J
-

+ fxj,{x)dx =

dx
fxj,(x)dx

= xJSx).
(2),

(5)

If

we

write J^x) for z in equation

then multiply through

by

xi

j--i

an d integrate from zero to x, simplifying by inte-

gration

by

parts,

we

get

or

fx(/l*)ydx

f [(/.(*)) + (/,(*))"].
'

(6>

-Art 20.]
If

applications of bessel's functions.


replace

215

we

x by fix dz
2

in (2)
1

it

becomes

dz

(See

(8),

Art.

5).

Hence z

= J^x)
(j)J
Q
t

is

a solution of
(/* L x)

(7).

If

we

substitute in turn in

^ x) and /
K

for

z,

mult

tiply the first

equation by

x/ (^ x),
first,

the second by

x/ (/uK x)

subtract the second from the


parts,

simplify by integration

by

and reduce, we get

JxJl^xK x)J^x)dx
-^KaJljA,a)f^ K a)~}x aJl^ K a)J^}x,d)\
K

(8)

Hence
/,(/**)

if /i K

and

// t

= o,

are different roots of /(/*#)

= o,

or of

or of

im/Aim)
a

\J (iia)

= o,
o.

J xJ^Kx)JX^x)dx =
:

(9)

We give without demonstration the following formulas, which are sometimes useful
ir

J (x) =
x

- fcos(x
n

cos

cp)d(p.

(:o)

J (x) = -

I sin

cos (x cos

<p)d<p.

(11)

They can be confirmed by developing


grating, and comparing with
(1)

cos (x cos 0), inte-

and

(3).

Art.
(a)

20.

Applications of Bessel's Functions.


of Art.
5 is
:

a special case of the following The convex surface and one base of a cylinder of radius a and length b are kept at the constant temperature zero, the

The problem

temperature at each point of the other base is a given function of the distance of the point from the center of the base re;

216

HARMONIC FUNCTIONS.

[CHAP. V.

quired the temperature of any point of the cylinder after the

permanent temperatures have been established.

Here we have

to solve Laplace's

Equation

in

the form

or
(see Art.
5),

r dr

oz

subject to the conditions

=o u = o

when
"

u=f(r)"
u

= o, r ~ a, z = b.
(1),

Starting with the particular solution of

sinh (/Az)J (Mr),


5,

(2)
.

and proceeding
of

as in Art.

we

get,

if

/*,,/*,,//,,..

are roots
(3)

/.(/**)

= o,
%
z

and

f(r)

= AJ

Ml r)

sinhfos) ^\sinh (/^)

+ AJ.{ r) + AJ^ + sinhOi^) ^ + "%inh (///) / '^ +


r)
,

(4)

r;

r)

'

'

(5

'

instead of keeping the convex surface of the cylinder at temperature zero we surround it by a jacket impervious to when r heat the equation of condition, 21 a, will be re-

()

If

placed by

= O when r = a,
oV
.

or

if

=
r

sinh (^iz)/
a,

(^ir)

by

=o

when

dr
that
is,

by

/xj^a)
/:(/"*)

o
(6)
.

or
If

= o.
are roots of

now
If

in (4)

and

(5) /Jt

/*,,/*,,..

(6), (5) will

be the solution of our new problem.


instead of keeping the convex surface of the cylinder at the temperature zero we allow it to cool in air which is at
(c)

the temperature zero, the condition u


replaced by

= o when

= a will

be

or

\-hu

O when r

= a, h

being the coefficient

of surface conductivity.

Art.

21.]

development

in

terms of bessel's functions.

217

If

sinh (?z)J9 (f*r) this condition

becomes
r

P/Af**)

+ hj^jxr) = o
ahj^d)
,

when
o.
.

=a

or
If

H*/i(j<"*)

(7)
(7), (5) will

now

in (4)

and

(5)

yw,

//,

yw 3

are roots of

be the solution of our present problem. It can be shown that

- o, = /iW
/.(*)
>

(8)

(9)

and
have each an

xJ (x)
x

\Jlx)

=o

(10)

infinite

number

of real positive roots.*

The

earlier roots of these equations can


difficulty

be obtained without serious

from the table for Jix) and /,(#) at the end of this

chapter.

Art.

21.

Development
now

in

Terms of Bessel's Functions.

We
article.

shall

obtain the developments called for in the last

Let

Ar)
fjtt
,

= Ajfar) + A /&S) + AJ (Mt r) + ...


t
9

(1)
o,

^ ^
,

etc.,

being roots of

of

MaSXM<*)

= o, or Vo(^) = o.
(/*tf)

of /,(//#)

or

To determine any coefficient A k multiply (1) by rj (jii k r)dr and integrate from zero to a. The first member will become
a

frf{r)/.{n t r)dr.

Every term
19,

of the second

member

will vanish

by

(9),

Art.

except the term

At

fr\J,(fi,r)Ydr.

JrUAwWr = jjf4./J&J*t=
by
(6),

j([/.(w*)]"+[y.(w*)]')

Art. 19.
* See Riemann's Partielle Differentialgleichungen,
97.

218

HARMONIC FUNCTIONS.
a

[Chai\ V.

Hence

A =
k

The development
Arts. 6 and 15).
If J* t
,

(i)

holds good from r

= o to r = a
(2)

(see

Mt

M%> etc -> are roots


'

o{/ (jua)
o

= o,

reduces to

Ak

7uh&%J'rJW*Jvyir
oi/ (jia)
t

'

(3)

If Mi, M*y M*> e tc, are roots

o, (2)

reduces to

If /i,,
(2)

ju 3

M%y

etc -> are

roots of

^aj {^a)
x

\J%{j*a)

o,

reduces to

^ =
4

^+jJ^UA^)r / rArV 0v)dr


'

(5)

For the important case where /(r)


a
a
%
fi k

=
a

frf{r)Jl^k r)dr=

f rjJjA#)dr=l- fxJ (x)dx = ^-J ^ d) (6)


Q
x

by

(5),

Art. 19; and (3) reduces to


2

(4)

reduces to

Art. 21.]

development

ix

terms of bessel's functions.

219

Prob. 29. An iron cylinder one meter long and 20 centimeters diameter has its convex surface covered with a so-called non-conducting cement one centimeter thick. One end and the convex
in

surface of the cylinder thus coated are kept at the temperature zero, Given that the conthe other end at the temperature of 100 degrees.
ductivity of iron is 0.185 and of cement 0.000162 in C. G. S. units. Find to the nearest tenth of a degree the temperature of the middle point of the axis, and of the points of the axis 20 centimeters from each end after the temperatures have ceased to change.

Find also the temperature of a point on the surface midway between the ends, and of points of the surface 20 centimeters from each end. Find the temperatures of the three points of the axis, supposing the coating a perfect non-conductor, and again, supposNeglect the curvature of the coating. Ans. ing the coating absent.
O

Or>
;

15 .4

40

.85

72 .8

Oo-O 15
;

.3

40

.7

72 .5

o .0

.0

.3.

Prob. 30. If the temperature at any point in an infinitely long cylinder of radius c is initially a function of the distance of the
point from the axis, the temperature at any time must satisfy the

equation

du ~-

3 a (d*u (- 3

1
-\

du\
I

(see Art.

1),

since

* it is

clearly in-

dependent of z and

<p.

Show

that

+ Af-"*j far) +
%

.,

where,

the surface of the cylinder is kept at the temperature o and Ak is the value zero, //, // 9 , M 3 , . are roots of J^pc) if the surface of the given in (3) with c written in place of a cylino and A k is obare roots of JJju) der is adiabatic /* //,, jtf 3
if
, .
.

tained from (4); and if heat escapes at the surface into air at the tem> ^JX^ C ) perature zero /*,, ju a j/t , ...are roots of pcjjj**)
,

and

Ak

is

obtained from

(5).

Prob. 31. If the cylinder described in problem 29 is very and is initially at the temperature ioo throughout, and the

long con-

is kept at the temperature o, find the temperature of a point 5 centimeters from the axis 15 minutes after cooling has begun ; first when the cylinder is coated, and second, when the coating is Ans. 97.2 o.oi. absent. Prob. 32. A circular drumhead of radius a is initially slightly
;

vex surface

distorted into a given form which is a surface of revolution about the axis of the drum, and is then allowed to vibrate, and z is the

ordinate of any point of the

membrane

at

any time. Assuming that

220

HARMONIC FUNCTIONS.
satisfy the equation

[CHAP. V.

must

^3

+
/

~j,

subject to the con-

ditions z
/

when

a,

o when
/^<tf
)

o,

and

= /(r)
/i 9 tf

when

o,

show

that 2

= A^J^y)
. .
.

cos

+ AjXl*/) cos
an(*

where

/*,, fi %% f*% ,

are roots of

jX^a

^* nas

tne value

that if a drumhead be initially distorted as in not in general give a musical note that it may be problem 32 that in this case the initially distorted so as to give a musical note that the periods of the various vibration will be a steady vibration musical notes that can be given are proportional to the roots of o, and that the possible nodal lines for such vibrations (x)

given in (3). Prob. 33.

Show
will

it

are concentric circles whose radii are proportional to the roots of

/.(*)

o.

Art.
If in a

22.

Problems in Bessel's Functions.

o when

problem on the stationary temperatures of a cylinder z = o, u = o when z = b and u = f{z) when r = a,


y

the problem

is

easily solved.

If in (2),

Art. 20, and in the corfit,

responding solution z

can readily obtain

= cosh (Mz)/ (Mr) we replace M by we z = sin (/xz)J (/uri) and z = cos (Mz)f (M r i)
(1),

as particular solutions of

Art. 20; and

and

is real.

^T
fc=i

knz

where

A k =-r
9.

knz
I

f(z) sin

7-

dz
>

(2)

by

Art.

knri\
b
7
I

Hence

= "^

k7TZ^\
sin

r-

(3)

is

the required solution.

Art.

24.]

lame's functions.

221

table giving the values of

J (xi)

will

be found at the end

of this chapter.
Prob. 34. A cylinder two feet long and two feet in diameter has bases kept at the temperature zero and its convex surface at 100 degrees Centigrade until the internal temperatures have ceased

its

Find the temperature of a point on the axis half way between the bases, and of a point six inches from the axis, half way between the bases. Ans. 72. i; 8o.i.
to change.

Art.

23.

Bessel's Functions of

Higher Order.

If we are dealing with Laplace's Equation in Cylindrical Coordinates and the problem is not symmetrical about an axis, functions of the form

**
very much

**
'

2r(*+l)L

2'(+l)

2\2\{?l+

1)0+2)

the same part as that played by (x) in the play are known as Bessel's Functions of preceding articles. They the nth. order. In problems concerning hollow cylinders much

more complicated functions


of the second kind.

enter,

known

as Bessel's Functions

For a very

brief discussion of these functions the reader

is
;

and Spherical Harmonics for a much more complete treatment to Gray and Matthews' admirable treatise on Bessel's Functions.
referred to Byerly's Fourier's Series

Art.

24.

Lame's Functions.

Complicated problems in Potential and in allied subjects are usually handled by the aid of various forms of curvilinear coordinates, and each form has its appropriate Harmonic Functions,

which are usually extremely complicated. For instance, Lame's Functions or Ellipsoidal Harmonics are used when
solutions of Laplace's Equation in Ellipsoidal coordinates are Toroidal Harmonics when solutions of Laplace's required
;

Equation

in

Toroidal coordinates are needed.

For a

brief introduction to the theory of these functions

see Byerly's Fourier's Series

and Spherical Harmonics.

HARMONIC FUNCTIONS.

[Chap. V.

TABLES.

223

224

HARMONIC FUNCTIONS.
Table
II.

[Chap. V.

Bessel's Functions.

TABLES.

225

Table
n

III.

Roots of Bessel's Functions.

226

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP, VI.

Chapter VI.

FUNCTIONS OF A COMPLEX VARIABLE.


By Thomas
S.

Fiske,
in

Adjunct Professor of Mathematics

Columbia University.

Art.
If

1.

Definition of Function.
of them,

two or more quantities are such that no one

when any

values whatsoever are assigned to the others, suffers any restriction in regard to the values which it can assume the quantities are said to be " independent."
If

one quantity

is

so related to another quantity or

to

several independent quantities, that whenever particular values are assigned to the latter, the former is required to take one or

another of a system of completely determined values, the former is said to be a " function " of the latter. The quantity or
quantities upon the values of which the value of the function " " of the independent variables depends, are said to be the
function.

function

is

" one-valued "

when

to every set of values as-

signed to the independent variables there corresponds but one " value of the function. It is said to be " ^-valued when to

every set of values of the independent variables n values of the function correspond.

The "Theory

of Functions

"

has

among

its

objects the

study of the properties of functions, their classification according to their properties, the derivation of formulas which exhibit the relations of functions to one another or to their independ-

ent variables, and the determination whether or not functions


exist satisfying assigned conditions.

Art.

2.]

REPRESENTATION OF COMPLEX VARIABLE.


2.

227

Art.

Representation of Complex Variable.


quantity
is

A variable
real

and imaginary

values.

capable, in general, of assuming both In fact, unless it be otherwise

specified, every quantity " " form

w
V
by
is

is
I,

to be regarded as having the

complex

u-\-v

u and v being

real.

It is cus-

tomary to denote
tity thus
:

V
If

i,

and to write the preceding quan-

iv.

zero,

is

real

if

is

zero,

is

pure imaginary."

A quantity z = x
i

-\-

iy

is

" said to vary " continuously


x x

when
x

between every pair of values which it takes, c = a -f- ib and c * = a *~\~ M% tne value of z varies in such a manner that x and y pass through all real values intermediate to a and #a b and b respectively.
x ,

x give to a variable quantity z iy a graphical representation by drawing in a plane a pair of rectangular axes and constructing a point whose abscissa and ordinate are
It is usual to

= +

To every value of z will correrespectively equal to x and y. a point and, conversely, to every point will correspond spond
;

point and value, then, may be interchanged without confusion. When z varies continuously the

a value of

z.

The terms

"

"

graphical representation of its varia" tion, or its path," will be a continuous

This graphical representation is line. of the highest importance. By means of it some of the most complicated
propositions
sion.

given an exceedingly condensed and concrete expres-

may be

r cos putting x quantity, the point

By

0,

r sin

where r

is

a positive real

z
is

= r(cos 6

-f- i

sin 6)

referred to polar coordinates. " absolute value or " modulus of

The quantity
z.

is

called the
(^J,

It will often
z.

be written

and

" 6 will be called the argument" of

223

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

function is sometimes considered for only such values of each independent variable as are represented graphically by the In the study of functions points of a certain continuous line.
of real variables, for example, the path of each variable

resented by a straight or y o.

line,

is repthe axis of real quantities,, namely,

Art.

3.

Absolute Convergence.
means
of infinite series

The
is

representation of functions by

one of the most important branches of the theory of funcIn


it is

tions.

that

problems, in fact, it is only by means of series possible to determine functions satisfying the condi-

many

tions assigned

and to obtain the required numerical


will
If

results..

Frequent use

be made

of the following theorem.

Theorem.

the moduli of the terms of a series form a

convergent series, the given series is convergent. wQ -\- w Let the given series be r (cos 0, i sin # ), w, r (cos 6 a in which w

W=
r

=
x

-f-

wn -\-

By

hypothesis the series


Its

R=

+r +

+ +r +
n

i sin #,)

is

terms being all positive, the sum of its first m convergent. terms constantly increases with m, but in such a manner as to approach a limit. The same will be true necessarily of any
series

formed by selecting terms from R. The sum of the is composed of two parts, terms of the series

first

W
x

r cos 6o
z{r sin 6

+ r cos 6 + r sin 6 +
X
x

+ rm + rm

_, _x

cos 9m .
sin #,_,),

and each of these in turn may be divided into parts which have all their terms of the same sign. Every one of the four parts
thus obtained approaches a limit as m is increased for the terms of each part have the same sign, and cannot exceed, in absolute value, the corresponding terms of R. Hence, as m is
;

increased, the
limit
;

sum

of the

first

tn

terms of

approaches a

which was to be proved.


series,

the moduli of whose terms form a convergent

series, is said to

be " absolutely convergent."

Art.

4. J

elementary functions.
i.

229
*
. .

Prob.

Show

'

that the series


if
|

I
I.

-+

+V +

+ zn +

is

absolutely convergent,

<

Art.

4.

Elementary Functions.

In elementary mathematics the functions are usually considered for only real values of the independent variables. In the case of the algebraic functions, however, there is no difficulty in assuming that the independent variables are complex.

The theory
fore,
is

of elimination

shows that every algebraic equation

can be freed from

radicals. Every algebraic function, theredefined by an equation which may be put in a form wherein the second member is zero and the first member is

rational

and entire

in

the function and

its

independent variables.

Besides the algebraic functions, the functions most often occurring in elementary mathematics are the trigonometric and

exponential functions and the functions inverse to them. The definitions, by which these functions are generally first intro-

duced, have
familiar series,
e*

pendent variables

no significance in the case where the indeare complex. However, the following

z* = expz= i+z+ z* + z* +-+..., z 4 j


.

C0S2=

z __+_+....,
z*

a
3
I

.9
'

z
J

1
.

7i

which have been established for the case where the variables are real, furnish most convenient general definitions for exp z,
cos
z,

and
finite

sin

every the equation

z these series being absolutely convergent for value of z. Defining the logarithmic function by
%

e\o%
it

ex p (Jog g}

follows that
a*

= ezl s

sa

exp(z log

a).

230

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VL.

The

following equations also are to be regarded as equations,


:

of definition

tan z

sin 2

Art.

5.]

continuity of functions.

231

variables

quantities of moduli less than some determinate positive quantity 3, the value of the function is altered by a

by

quantity of modulus less than any previously chosen arbitrarily small positive quantity e.

function of one independent variable

is

said to be con-

tinuous in a given region of the plane upon which its independent variable is represented, if it is contiuuous at every point
in that region.

From the

principles of limits,

it

follows that

are continuous at a given point, their

uct are continuous at that point. quence, every rational entire function of z

two functions sum, difference, and prodAs an immediate conseif

is

continuous at

finite point every such function can be constructed from z and constant quantities by a finite number of additions, subtractions, and multiplications.

every

for

Let a function of a single independent variable be continuous at c, and let it take at that point the value /, different
from
zero.

Suppose

also that at
/ -\-

any other point c-\-Ac\X\z

function takes the value


I

At.

Then
At

At
At
|

t{t+ At)
/
|

If it

be assumed that

<

|,

the modulus of the preceding

difference cannot exceed

\t\(\t\-\At\y

and

will,

therefore, be less than e

if

\jt\<

up
l+e\t

Hence
at a point

if
c,

a function
its
if

is

continuous and different from zero

at

once that
is

It follows reciprocal is also continuous at c. two functions are both continuous at c, their
c,

ratio

continuous at

unless the denominator reduces to zero

232
at that point.

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

But every

rational function of z
It is

may
its

be expressed

as the ratio of

two

entire functions.

therefore continuous

for all values of z except

those for which

denominator

vanishes.

Consider the function expz,


pZ-\-b.Z

gZ

Hence

if

\Az\<\,

\A Z \'

but the limit of the third


proaches zero.
of
z.

member
is

is

zero

when
all

Hence exp z
that cos z

continuous for

finite

\Az\ apvalues

Prob.

5.
z.

Show Show

and

sin z are

continuous for

all

finite

values of
Prob.

6.

that tan z

is

about the origin as a center with a radius

continuous in any circle described less than \n. o^ 2~-

Art.
It

6.

Graphical Representation of Functions. ? -Functions


in Art. 2 that a

was shown

plane suffices for the complete

graphical representation of the values of an


able.

In the same way

it is

independent variconvenient to use a second plane

to represent graphically the values of

any one-valued

function.

For example,

if

w =jf[z)

be such a function, to each point

x
u

-f-f-

iy of the

iv of the function.

of the point

independent variable will correspond a point " This point u -f- iv is called the u image If w is a continuous function of z, then -f- iy.
in

every continuous curve in the ^-plane will have an image the w-plane, and this image will be also a continuous curve.

Consider the expression u

-f-

iv

x" -f-

-f-

2ixy.

Here

Art.

7.J

DERIVATIVES.
-f-

233

x*

y and v =

2xy.

Since to every value of z correy

spond determinate values of x and y and consequently determinate values of u and v, this expression falls under the general definition of a function oiz.
It is

evidently continuous.

Every
of
it

straight line
is

=
1

parallel to

the axis of

converted by means y into a parabola v \f{u f).

Prob.

7.

Find the family of curves

into which the straight lines parallel to

the axis of y are converted by means of

the function u

x*

~~

^"

+ 2txy.

Show

that

no two curves

of this family intersect.

Art.
Let
tt>

7.

Derivatives.
*.

= f(z)
is,

be a given function of

If

//

is

an "

infiniif

tesimal," that

a variable having zero as

its limit,

and

the

expression

h
has a
finite

determinate

limit,

remaining the same under

all

possible suppositions as to the way in which //approaches zero, " derivative" of the function^) at this limit is said to be the

the point
at
z.

z.

In this case
is

f{z)

is

said to be " dvu


-7-.

"

monogenic
is

The

derivative

written f'{z) or

A function

said

to be monogenic in a region of the plane of the independent variable if it is monogenic at every point of that region.

Consider

now

the circumstances under which a function

iv

may

have a derivative at the point z


is
;

-\- iy.

If z be given a real increment, x Ax y is unaltered, so that Az

changed and

into

-f-

Ax, while

A iv
~Az~

Au

Av
'

Ax

Ax'

234
If,

FUNCTIONS OF A COMPLEX VARIABLE.


is

[CHAP, VI.

on the other hand, z ment, Az iAy, and

given a purely imaginary incre-

Azv

'

_"

Au
iAy

Az
If

Av Ay

'

the second

members

of these equations approach deterzero,

minate limits as
are equal,

Ax and Ay approach

and

if

these limits.

du, l-9^___ z du

dv dy

dx^ dx~

dy

Hence, equating

real

and imaginary

parts,

du

dv

dv _

du

dx~dy'
which are necessary conditions
It

dx~~~dy'
for the existence of a derivative.

can be shown that these conditions are also sufficient.

the increment of the independent variable be entirely arbitrary, no supposition being made as to the relative magni-

For

let

tudes of

its real

and imaginary
is,

parts.

Then

the differential of

the function, that

that part of the increment of the function

which remains after subtracting the terms of order higher than


the
first, is

W
Hence
(

dx'

\dy

dy

'
I

du

+ _ dx + idy
du
idv

\Yx

.dv\
t

^x)

dv\dy + (du + %)^ V \dJ


.

,1\
l

lt dx I+ ;i
is

3f
i"*f *t

which, by virtue of the conditions written above, either member of the equation

equal ta

du

.dv_
'

_ -du

dv dy

dx

dx
is

dy

The

value thus obtained

independent of

-i
dx

or,

what
z.

is

the

same

thing, of the direction of approach to the point

The

Art.

7.]

derivatives.

235

existence cf a derivative of the function zv depends, therefoie,

only on the existence of partial derivatives r J

-*-.

dx

^, dx

-r-,

;r

ay

dy

satisfying the specified equations of condition.

The same equations

of condition

express the fact that

w=

x and
some

supposed to be an analytical expression involving and having partial derivatives with respect to each, y, involves z as a whole, that is, may be constructed from z by
u
-\- iv,

series of operations, not introducing

or

y except

in

the

combination x

y may

In other words, they indicate that x and -\- iy. both be eliminated from <p(x, y) by means of the

w=

x -\- ty. This property, however, is not sufficient equation z to define a function as monogenic, for not every function which possesses it has a derivative with respect to z.*
monogenic function is necessarily continuous that the existence of a derivative involves continuity. For, if
;

is,

limit "^

^-^ =/(*)>

it

follows that

where

rj

approaches zero with

//.

Hence

f(z)

is

the limit of

f(z-\-h) when h approaches


point
z.

zero, or f(z)

is

continuous at the

following pages relate almost exclusively to functions which are monogenic except for special isolated values of z. Functions which are discontinuous for every value of the independent variable, and functions which are continuous but admit

The

no derivatives, have been


real variables.-)*

little

studied except in the case of

* For an interesting illustration of a non-monogenic function constructed from z by a series of arithmetical operations, see the expression given at the bottom of page 298. fin this connection see G. Darboux, Sur les fonctions discontinues, Annates de l'cole Normale, Series 2, Vol. 4 (1875), PP- 51-112. For a systematic
treatment of functions of a real variable, see the
treatise

German

translation of Dini's.

by Luroth and Schepp, Leipzig, 1892.

236

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

Art.
Let z

8.

Conformal Representation.

start from the point z% and trace two different paths a given angle at the point z and let z and z% be arbiforming
> x

trary points on the

first

and second paths respectively.


-fi'sin
t

Then

z
x

= r,(cos 0,

r e i6 \
x

where r denotes the length

of the straight line joining z

and

and

0,

denotes the inclination of this line to the axis of

reals.

In the

same way,
z
a.

for the point #a


2

there

is

an equation

z%
If

r3 (cos

sin

9)

9 rj* :

now

is

region

of the ^-plane considered, to the points z

one-valued monogenic function of z, in the z gt corre,

x >

spond points
the equations

w w w
,

and

for these points

can be formed

w
From
once
that,

w = p/*

w,

w = p/f*.

the supposition that w is monogenic, it follows at when z and z % are assumed to approach zQy
x

limit

v^Z^i = H m it ^zJHs. z
z,

zx

If

the

members
in the

of this equation are not equal to zero,

it

may

be put

form
limit
limit

*!*.

Art.

8.]

CONFORMAL REPRESENTATION.

231

or

limit

&**A.

limit

*V**-H ^
a

Hence

limit

(0,-

2)

limit

(0,-

f)

and the images in the w-plane of the two paths traced by z form at w an angle equal to that at z% in the ,2-plane. Accordingly, if z be supposed to trace any configuration whatever
in a portion of the ,2-plane in

which
dz

is

determinate and not

equal to zero, every angle in the image traced by w will be equal to the corresponding angle in the -s'-plane. If, for exambe divided into infinitesimal ple, such a portion of the w-plane
triangles, the

corresponding portion of the -plane will be divided in the same manner, and the corresponding triangles
will

be mutually equiangular.

Such a copy upon a plane, or


is

upon any surface, of a configuration in another surface a " conformal representation."

called

The modulus
"

of the derivative

dw
~dz

Aw
limit

Az

is

the

Its value, which, in general, changes from magnification." point to point, may be obtained from the relations

dw
dz

-+'-'+'
"

dx dy

dy dx
representation has interesting ap-

The theory of conformal plications to map drawing.*


p. 500,

* For the literature of the subject, see Forsyth, Theory or Functions, and Holzmiiller, Einfuhring in die Theorieder isogonalen Verwandschaften und der conformen Abbildungen, verbunden mit Anwendungen auf matheniatische Physik.

Qrlfl
23S

Ajprf/

nnfyr^oCAU ftwtLAsUn**>
[CHAP. VI.

FUNCTIONS OF A COMPLEX VARIABLE.


9.

Art.
Case

Examples of Conformal Representation.


Let

I.

w=z s+c.

This function

is

formed from the


Putting

independent variable
for w, z,

by the addition of a constant.


-f-

and

c,

respectively, u

ivt

x -\-

iy

and a -f-

tb,

one ob-

tains

=x

-f- a,

=y

-J- b.

Any

configuration in the ^-plane appears, therefore, in the seaplane unaltered in magnitude, and is situated with respect to

the axes as

if

it

had been moved

parallel to the axis of reals

through the distance a and parallel to the axis of imaginaries through the distance b. The following diagrams represent the transformation of a network of squares by means of the relation

w=

-\- c.

Case
c

II.

Let

cz.

Writing

w
:

pe**,

re ie

and

rx ^\ the following equations result

p=
The

rt r,

0=0, +

*.

origin transforms into the origin, all distances measured from the origin are multiplied by a constant quantity, and all straight lines passing through the origin are turned through a constant angle. See the following diagrams.

Art.

9.]

examples of conformal representation.

239

Case III

Let
zv

zv

=
e*e

e*.

Writing z
^(cosjy

=x

-f-

iy,

the function

becomes

iy

=
/,

i sin y).

\ t

Every straight formed into a

line

circle

= p =

parallel to the axis of


J* described

is

trans-

about the origin as a

center, the axis of

y becoming the unit circle. Points to the of the axis of y fall without the unit circle, while points right / to the left of this axis fall within. Every straight line jc

tan /9 parallel to the axis of x becomes a straight line v/u The accompanying diagrams* the origin. passing through
exhibit in a simple

manner the
(z -f

periodicity expressed

by the

equation

exp
where n
is

2nni)

exp

(z),

any positive or negative


in

integer.

the w-plane, excluding the origin, correevery point number of points in the ^r-plane. These spond an infinite points are all situated on a straight line parallel to the axis of

To

*The
treatise.

figures of this

and the following example are taken from Holzmilller's

240
y,

FUNCTIONS OF A COMPLEX VARIABLE.


it

[Chap.
If

VL

and divide

into segments, each of length 27t.

z be one
is

of these points, the general value of the inverse function

log VD == if

-}" 2/ZZ7T,

where n
If

is

any positive or negative


line

integer.

beginning at the origin be drawn in the w-plane, there will correspond in the ,2-plane an infinite number

any straight

27T-

37T 2

of straight lines parallel to the axis of x, dividing that plane To any curve in the w-plane into strips of equal width.

which does not meet the


the .s-plane an infinite

line just

number

of curves, of

drawn, will correspond in which there will be

one

in

each

strip.

Case IV.
sin {iy)

Let

w=

cos 8.

Writing

w=

and employing

as equations

of definition

= x iy> cos {iy) = cosh^,


-j-

tv,

-\-

i sinhjj/,

the given function takes the form

-\-

tv

cos^r coshjj/
coshjj, v

i sin

s'mhy.

Hence

= cos x

sin^r sinh/.

ART.

9.]

EXAMPLES OF CONFORMAL REPRESENTATION.


straight line,

241
is

Any

t v parallel

to the axis of y,

trans-

formed into one branch of a hyperbola,

=
cos
/,

sin t x

I,

having

its foci

at the points

+
+
'

and
is

i.

Any

straight line,
ellipse,

y=

t%

parallel to the axis of x<

transformed into an

=
sinh
/,

I.

cosh"/,

having

its

foci at the

line equal in taken once.


is

same points, any segment of the straight length to 2n corresponding to the entire curve By means of these confocal conies, the w-plane

divided into curvilinear rectangles, the conformal representation breaking down only at the foci, where the condition
that

dz

-^

should be different from zero


of the function, expressed

is

not

fulfilled.

The

periodicity

by the equation

242

FUNCTIONS OF A COMPLEX VARIABLE.

[Chap. VI.

tf-plane serving to

determine

its

image

in

the w-plane as an

intersection of orthogonal curves.

Case V.
the relations

Let
u

w= =x
a

3
.

Writing

w=
$xy
x,

+
y

zv,

+ iy

3*/,

follow at once.

If

one of the variables

be eliminated from

these two equations by means of the equation Ix my n o, a straight line in the s-plane, equations are ob-^ representing tained representing a unicursal cubic in the w-plane.
'

kvulIP"

By

putting

p(cos

the

relations

=
ra

+
=

sin 0), s

r(cos 6

-\-

i sin #),

r\

3#,

are

obtained.

Hence the
1/frwtVr*

circle

2ar cos 6

#a

= =

c*

gives the curve


pi

2#p* cos

a*

c*,

center.

which enwraps three times the point corresponding to the The accompanying figure represents this transformation, the straight line feg giving the curve feg.

To

each point in the w-plane, excluding the origin, at which

dw
dz

o and the conformal representation

is

not maintained,

Art.

9.]

EXAMPLES OF CONFORMAL REPRESENTATION.

243

there correspond three points in the #-plane, having for their

arguments

00
3

-|-

2n

0+4^
:

respectively.

Any straight line


from

drawn from the


images

origin in the w-plane will have, therefore, three

in the ^-plane, viz., three straight lines diverging

the origin, Any continuous curve

and dividing the plane into three equal regions.

in the w-plane not meeting the line just drawn will be represented in the ,2-plane by three curves, of which one will be situated within each of these regions. In the

figure here given are exhibited the three conformal represen-

tations of a square
tv

If

t lt u the w-plane by lines u / 2 parallel to the axes. tlt v the relation between and z be reversed, and z be

formed

in

taken as a function of w, z will be a three-valued function, its values giving rise to three branches which will remain distinct

and continuous except when

becomes equal to

zero.

Prob. a

8.

If

w=
w=

-\

show

that circles in the 2-plane having

common
Prob.

center at the origin transform into confocal ellipses.


z
i

9.

If

+*

-,

show that the

axis of reals in the 2"-plane half of the s-plane

transforms into the circle

\w\

1,

and the upper

into the interior of this circle.

244

FUNCTIONS OF A COMPLEX VARIABLE.


10.

[CllAP.

VL

Art.

Conformal Representation of a Sphere.


its

Let OPO' be a sphere having

diameter 00' equal


length to unity.

in

Conin

^ O and
the

struct tangent planes at


O'.

Draw

plane rectangular axes

tangent

at

Ox
the
as

and
other

Oy

and

in

plane draw axes O'u, parallel to

Ox
the
0'ir

and

measured
direction,

in

same
ured

and

parallel to
in

Oy

but measdi-

a contrary

O'

by a straight line, and let <9Pand produce it to meet the plane uO'v in w. From the similar triangles O Oz and 00' w
'

rection. Join any point z in the plane xOy to Draw 0' z meet the sphere in P.

00' Oz -_ = _-,

OO

Ow
|

or

n nl = 757Pt Oz.Ow 00
I.
|

that

is,

.
|

w = rp =

an observer standing on the sphere at O rotation about 00' from O'u toward O'v is positive, while to an observer standing on the sphere at O such a rotation is negative.

To

Hence

/_xOz=
The

/_uO'w,
:

or

0.

following equation results


'

wz = p?v *+
z(

fl)

I.

^-planes are therefore conformal representations of one another. Any configuration in one plane can be formed from its image in the other by an inversion with respect

The w- and

ART.

11.]

CONJUGATE FUNCTIONS.

245

to the origin as a center, combined with a reflection in the axis of reals. Such a transformation was termed by Cayley a " quasiinversion."

points at a great distance from the origin in one plane are brought near together in the immediate neighborhood of the origin in the other plane.

By

it

Since the line O'Pz makes the same angle with the plane as with the plane xOy, any spherical tangent to the sphere at

angle having
z.

its

vertex
is
it

The sphere

projected into an equal angle at thus seen to be related conformally to the


also so related to the plane uO'v.

P at P is

plane xOy, and


\.

must be

jjl

\jlmA

representation of the sphere upon a tangent plane in uthe manner described above is termed a " stereographic proWhen to this representation is applied a logarithmic jection."

The

JU

transformation, that

is,

one inverse to

the

transformation
.

V ./.

the preceding article, the so-called described in Case " Mercator's *Jyi 'Am****.projection" is obtained, ./.
III of

Art.

11.

Conjugate Functions.

The

real
ivy

w=

and imaginary parts of a monogenic function, have been shown to satisfy the partial differential
9

equations

_ dv ~
dy

dv__

_du
dy
partial

d*

d*~

At any

point, therefore,

where u and v admit second

derivatives, one obtains

dx^~df-'
that
is,

9? + cy-

tion for
this

the functions u and v are solutions of Laplace's equatwo dimensions. Any two real solutions / and q of

equation, such that p-\-iq is a monogenic function of x -f ty, are called " conjugate functions." * Thus the examples of Art. 9 furnish the following pairs of conjugate functions
:

Maxwell, Electricity and Magnetism, 1873,

vol. 1, p. 227.

246

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP.

VL
;

-\- a,

y -\- b

cos

x cosh y,
is

rx r cos (0 l -jsin x sinh y


in

6),
;

r x r sin (dx
3

0)

cosj,

t*

wny

^r

3^y,

3^rjA

y.

The second

pair

expressed

to cartesian coordinates

polar coordinates, but may be transformed by means of the relations

= Vx*+y\

cos 6

= -

sin 6

one of two conjugate functions be given, the other is thereby determined except for an additive constant. Let u y Then for example, be given.
If

dv

=
dx

dx -4dy

dy

=
and therefore the value
of v
is

du, du, dx -\ ay. J% dx dy

c lt v c^> obtained by assigning conequations u stant values to two conjugate functions, represent in the It w-plane straight lines parallel to the coordinate axes.

The

follows that the curves which these equations define in the


,2-plane

intersect at right angles.


x

the quantities c and obtained and c and


;

c
a

two orthogonal systems

Consequently, by varying of curves are

be taken as orthogonal curvilinear coordinates for the determination of position in the ^-plane.
x

may

Prob. io. Show that if p and q are conjugate functions of u and where u and v are conjugate functions of x andy,fi and q will be conjugate functions of x and y. Prob. ii. Show that if u and v are conjugate functions of X and y\ x andjy are conjugate functions of u and v.
v,

Art.

12.

Application to Fluid Motion.

Consider an incompressible fluid, in which it is assumed that every element can move only parallel to the ^-plane, and has a velocity of which the components parallel to the coordi-

Art.

12.]

application to fluid motion.

247

nate axes are functions of


of the fluid
is

x and y

alone.

The whole motion


in

known

as soon as the

motion

the .s-plane

is

ascertained.

any curve in the ,2-plane is given, by the "flux across the curve"* will be meant the volume of fluid
in unit

When

which

time crosses the right cylindrical surface having

the curve as base and included between the ^-plane and a parallel plane at a unit distance.

is

The flux across any two curves joining the points z and z the same, provided the curves enclose a region covered with the moving fluid. For, corresponding to the enclosed region,

Let z be there must be neither a gain nor a loss of matter. fixed, and z be variable. Let ip denote the flux across any curve
z
z,

reckoned from

left to right

for an observer stationed at z

and looking along the curve toward z. If /, m be the direction cosines of the normal (drawn to the right) at any point of the curve, and /, q be the components, parallel to the axes of the velocity of any moving element, the value of tp will be

tp

/ tA

{lp -f-

mq)ds,

where the path of integration

The

is the curve joining z and z. a one-valued function of z in any region tp within which every two curves joining z to z enclose a region covered with the moving fluid.

function

is

If

z moves in such a manner that the value of


it

tp

does not
it, i.e.,

liMi
*

vary, " stream-line."


ip is

will trace a

curve such that no fluid crosses


tp

The curves

const, are

all

stream-lines, and

If p and q are continuous, z be given infinitesimal increments parallel to x and y respectively, one obtains

called the

" stream-function."

and

if

4^y

If

now

the motion of the fluid be characterized, as


* Lamb's Hydrodynamics (1895),
p. 69.

is

the

248

FUNCTIONS OF A COMPLEX VARIABLE.

[Chap. VI.

case in the so-called " irrotational" motion,*


of a velocity-potential 0, so that
j

by the existence .V '^

\
a

30
:

90

pis

the following equations result

_ a^

a*
Hence
is

"ay a*~

90
a^'

monogenic function of ^r -j- iy. The curves -(- *^ which are orthogonal to the stream-lines, are const.,
a

called the "equipotential curves."

Consider, as an example, the motion corresponding to the The equipotential curves are given by the z*. functionf

w=

equations

<b= u = x
the
Y''

^xy=z const.,

stream-lines

by

the equations - v

'v|n

= 3*y y

const.

In the following figure the stream-lines


are the heavy lines,

while
tential

the

equipocurves are

dotted.

The
in

fluid

moves
origin,

toward the

which

is

called

"

cross-point," from three directions, and

At the cross-point flows out again in three other directions. the fluid is at a standstill, since at that point the velocity, for
which the general expression
is

* In irrotational motion each element


strain, but not to rotation.

V+'
is

subject to translation

and pure
by-

f F.

Klein

Riemann's Theory of
p. 3.

Algebraic

Functions

translated

Frances Hardcastle (1893),

Art.
is

12.]

APPLICATION TO FLUID MOTION.

249

equal to zero.

The

stream-lines in the figure represent the

motion of the

each of six different angles, as if the fluid were confined between walls perpendicular to the ^-plane.
fluid in
It is of

importance to note that


i,

if

the function considered

be multiplied by
-f
i<p.

are interchanged, since the function


tp

the equipotential curves and stream-lines <p-\-itp then becomes

An

example

of particular interest

is

w=
Let z
a

u log a

>

14

2.

c*w*<**?r

dc

1^-

= r/*, z =
/i

-j-

r9 ^>; then

log

-i,

tf

n{6

a ).

The
tems
lines,

curves u

= const., v = const,
which

of circles, either of

may

form two orthogonal sysbe regarded as the stream-

the other constituting the equipotential curves.

The
finite

velocities are everywhere, except at the points


If

a,

and determinate.

the circles

rjr%

a is as the stream-lines, each of the points const, be taken as the stream-lines, one If the circles 0, a

const, be taken a " vortex-point."

-0 =

250
of the points

FUNCTIONS OF A COMPLEX VARIABLE.


a
is

[CHAP.

VL

a "source," the other a "sink."

In the

latter case, besides the

simple electrical

hydrodynamical interpretation, a very illustration is afforded by attaching the poles

of a battery to a conducting plate of indefinite extent at fixed points of the plate.

two

As another example may be As

taken the relation

w =.

cos z~

has been shown, the curves x const, form a system of confocal hyperbolas, while the curves y const, form an

as stream-lines.

orthogonal system of ellipses. Either system may be regarded In one case the motion of the fluid would be

such as would occur

a thin wall were constructed along the axis of reals, except between the foci, and the fluid should be impelled through the aperture thus formed. In the other case
if

the fluid would circulate around a barrier placed on the axis of reals and included between the foci.

Besides their application to fluid-motion, conjugate functions have important applications in the theory of electricity

and magnetism * and

in elasticity/)-

Art.

13.

Critical Points.
of z.
It

Let w be any rational function the form

can be written in

where f(z) and


This function

<p (z)

are entire and without

common
number
root.

factors. of suc-

is finite

and admits an

infinite
z,

cessive derivatives for every finite value of


of the equation
(z)

except the roots

= o.

Let a be such a

Then the

reciprocal of the given function is finite and admits an infinite number of successive derivatives at the point a. Such a point
*

J.

J.

Thomson, Recent Researches


Theory of

in Electricity

and Magnetism

(1893),

p. 208.

+ Love,

Elasticity (1892), vol.

1,

p. 331.

Art.
is

13.]

critical points.

251

called a "pole."

Any

rational function having a pole at

can be put by the method of partial fractions in the form

where
and

lt

.,

Ak

are constants,
at the
a.

point the "order" of the pole, and the function is said to have for its value at a infinity of the k\\\ order. In accordance with
the definition of a derivative,
a.

ip(z) is finite

The

being different from zero, integer k is said to be

does not admit a derivative at

From
become

the character of the derivative in the immediate


a,

neighborhood of
to

however, the derivative

is

sometimes said

infinite at a.

has a pole of the first trigonometric function cot order at every point z nnt, being zero or any integer posi-

The

tive or negative.

log (z a) has for every finite value of z an infinite number of values. If z a is writz, except a, i& ten in the form Re ,

The

function

w=

w=
where log
integer.

log

R -\- z(Q -f- 2m7t)j


is

R is
If

real,

and

zero or any positive or negative

remain

line, beginning at a, S will but will vary. The images in the w-plane will fixed, therefore be straight lines parallel to the axis of reals, dividing

z describes a straight

the plane into horizontal strips of width 2n. If now the ^-plane is supposed to be divided along the straight line just drawn,

and z

varies along
it

any continuous path, subject only


cannot cross this

to the

restriction that

line of division, there will


in

be a continuous curve as the image of the path of z


strip of the w-plane.

each

Each

of these images

is

said to corre-

spond to a "branch" of the function, or, expressed otherwise, the function is said to have a branch situated in each strip.

The

line of division in the ^-plane,


is

which serves to separate

the branches from one another

" called a cut."

252

FUNCTIONS OF A COMPLEX VARIABLE.

[Chap. VI.

the point z a no definite value is attached to the As z approaches that point the modulus of the real function. part of the function increases without limit, while the imagi-

At

nary part

is

entirely indeterminate.
a,

Let z be an arbitrary point, distinct from


log

and

let

-j-

t@

-\-

2mni

be any one of the corresponding values of the function. Suppose that z starts from z and describes a closed path around
the point a, the values of the function being taken so as to give a continuous variation. Upon returning to the point z
the value of the function will be
log

or

+ z& + 2(m + log R + t@ + 2{m


is

\)ni,

i)7ti,

described in a positive or negative direction. By repeating the curve a sufficient number of times it is evidently possible to pass from any value of the function
at z

according as the curve

to

any
it

other.

When
in this
it

a point

is

such that a

-s'-path

en-

closing to another value,

may lead

manner from one value

of a function

is

called a " branch-point."

In the case

of the function here considered, the point z a "logarithmic branch-point," or a point of


discontinuity."

is

called

"logarithmic

The

function

w=

log

~-\ where
y

f{z)

and

(p(z)

are entire,

(p{z)

has a point of logarithmic discontinuity

at

every point where

either/^) or

<p(z) is

equal to zero.
t

For, writing
.

f(z)
<j)(z)

= A(z - a Yi* - mfr = B{z btf^z brf*


be written
log {z

the value of

w may
$

log

- + 2p m
m

- am

2qn n

log

(z

b n ).

Art

13

critical points.

253

Take now the function


proach
minate.

w=

e*.

It

for every value of z except z

o.

has a single finite value If z is supposed to apis

zero, the limit of the value of the function

indeter-

For

let

+ iq be perfectly arbitrary, and write


-f-

If

now a

ib is

the reciprocal of p
h

-\- iq,

so that

~~
.

"?

the preceding equation

may be
.

written

_i_

x%

r.

But whatever the value of the integer m, q 2mit may be substituted for q without altering the value of c -f- id, and hence
both a and

may be made

less

than any assignable quantity.

given function e* therefore takes the value c -f- id at points a 4- ib indefinitely near to the origin. point such that, when z approaches it, a function elsewhere one-valued tends toward

The

an indeterminate limiting value


gularity."

is

called

an " essential

sin-

Prob. 12.
singularity.

Show

that for the function e^-* z

is

an essential

i_

variable

considered as a function of a real continuous for every finite value of z, and the same is true of each of its successive derivatives. Show that when it is o is an essential regarded as a function of a complex variable, z
Prob.
13.

The

function e

z*

is

singularity.

In order to illustrate

still

another class of special points

take the function


IV

V(z

- aX*

a^)...{z

an ).

254

FUNCTIONS OF A COMPLEX VARIABLE.


This function has at every
finite point,

[CHAP. VI.
a^,
. . .

except a iy

a tli

two
it

distinct values differing in sign.

al9 at

takes but a single value, zero. let a straight line of indefinite extent be drawn in , . . . , au

At these points, however, From each of the points

such a manner that no one of them intersects any other, and suppose the ^-plane to be divided, or cut, along each of these
lines. Along any continuous path in the -S'-plane thus divided the values of the function form two distinct branches.

For, writing

=re
x

iQ
i,

a2

i6

r^e

*,

an

rn e' en ,

the function takes the form

w=
No

Vr

r2

rn

ei

closed path in the divided plane will enclose any of the 3 n after points a lt a9 , . . . , an and the quantities continuous variation along such a path, must resume at the
,
l
,

point their original values. No such path, therefore, can lead from one value of the function at any point to a new value of the function at the
initial

same
cuts
traces

If, however, the point. are disregarded and z

in

a positive

direction,

a closed curve including an odd number of the points a <? 2


x

a, and

not intersecting

itself,
X ,
, . . . ,

then an odd number of

the quantities 6n are each increased by 27r; and a the value of the function is altered by a factor ^BHU*^ and In the same way any closed path deso changed in sign.
scribed about one of these points, and enwrapping it an odd number of times, leads from one value of the function to

the other.

On the other hand, a simple closed path enclosing an even number of these points, or a closed path which encloses but one of the points, enwrapping it an even number of
times, leads back to the initial value of the function.
It
fol-

Art.

13.]

CRITICAL POINTS.

255
.

lows that each of the points a

a branch-point. Any point in the ^-plane, closed paths about which lead from one to another of a set of different values of a function, the
x
,

#a

aH

is

number

of values in the set being finite,

is

called an

"

algebraic

branch-point."

As

a further illustration, consider the function

w=z
which
is

-f- {z

a)^,

a root of the equation of the sixth degree,


2(z
a)zu*

w*

t>ziv"

+ 3<s^

z(z

a)w

+ (*
=

of

=o.

function has at every point, except z o and z a, Six branches are thereby formed which six distinct values.

The

can be completely separated from one another by making cuts from the points z = o and z = a to infinity. Putting w for the cube root of unity, these six branches can be written
V \V w z -\(z a)' f ^ -f- a? {z -\- go{z Hf, ss s ay, w a) V \V W = ^- V + G? W = Z + V* ^) The branches w, and w w and zt\, w and w are interchanged

1/2

si/ 8 '

-j-

(z

a)

1/2

1/3
3

'

1!

9 2/

3,

&

by

a small closed circuit described about z

=o,
.

while a small

circuit described

w w w
tf

%%

a permutes cyclically the branches w A w6 and also the branches

about z

All of the special points examined above, poles, points of

logarithmic discontinuity, essential singularities, and branchpoints, are called critical points.

In

fact,

a function, or a
"

branch of a function, is said to have a " critical point at each point where it fails to have a continuous derivative,* or about
impossible to describe a circle of determinate radius within which the function, or branch, is oneit is

which as a center

valued.

Any
*

point not a critical point

is

called an " ordinary point."

Continuity and, therefore, finiteness of the function are implied in the existence of a derivative.

256

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP.
is

VL

An

ordinary point at which a function reduces to zero a "zero" of the function.


If in

called

a certain region of the ^-plane there are no critical for a given function, the function is said to be " syneo points " " tic" or holomorphic in that region. If in a certain region

^hfa ( the only


"

critical

points are poles, the function


in that region.
is

is

said to

be

"

meromorphic

Under

similar conditions a

branch of a function

also

described as holomorphic ot

meromorphic.
Prob. 14.
(z

When w and
that
if

z are connected

by the relation

g=

z describes a circle about h as a center, describes a circle about g as a center, an angle in the 2-plane hav-

Kf show

ing its vertex at h is transformed into an angle in the w-plane f times as great and having its vertex at g and that z h is a branchexcept when t is an integer. point of
t

Art.

14.

Point at Infinity.

Ou-i

modulus
it

In determining the limiting value of a function when the of the independent variable z is increased indefinitely,, is usual to introduce a new independent variable z' by the
is

relation z

This

o. i/z\ and consider the function at the point z' equivalent to passing from the .s'-plane to another plane,,

the ^'-plane, related to the former by the geometrical construction described in Art. 10. It is often very convenient, however,, to go further and to substitute for the -plane the surface of the

sphere of unit diameter touching the ^-plane at the origin. Nodifficulty is thus introduced since, as explained in the article
just cited,

any configuration in the ^-plane obtains a conformal representation upon the sphere; and the advantage is gained that the entire surface upon which the variation of the indeis studied is of finite extent. The point of the sphere diametrically opposite to its point of contact with o. the -plane coincides with the point written above as z'

pendent variable

It is

called the point at infinity, z

sphere approaches it at the same ^-plane recedes indefinitely from the origin.

since a point on the time that its image in the


00
,

.Art. 15.]

integral of a function.
point at infinity

257

The
point.
point,

may be
f

either an ordinary or a critical


*'-

For the function e* for example, it is an ordinary For a rational entire function of the th since e* = f'.
it

degree

is

a pole of order n.

Consider
in

it

for the function

the preceding article. \/(z (z )(z Let a circle of great radius be described in the ^-plane inclosing a n Its con formal representaa7 . all the branch-points a
.
. . x

a9)

a H ), discussed

tion

on the sphere

oo point z branch-point or not, according as the function changes value or( not when the curve surrounding it is described, that is accord.

be a small closed curve surrounding the This point must, therefore, be regarded as a
will

ing as

n,

the

number

of finite branch-points,
is

is

odd or even.

When
total

the point at infinity

taken into account, then, the

number

of branch-points of this function

The
point

character of the

point z = oo

always even. for this function can be deis

termined directly, by changing z into


z'

i/z'

and considering the


(p(z) -77-7 >

= o.
Show
that z

Prob. 15.
<p(z)

00 is

an ordinary point for


if

where

and

rp(z)

are rational
ip(z).

and entire

the degree of <p(z) does

not exceed that of

Art.
Let
z,

15.

Integral of a Function.

w =/(z) be a continuous function of a complex variand suppose z to describe a continuous path L from the point z to the point Z. Let a series of points z z ... ,zn be taken on L, and let /,/,,..., / be points arbitrarily chosen on the arcs z z lf z z zn Z respectively. Form the sum
able
t

S * .'fc - *.)A*.) + (*. - *VW +


If

+ (*- z

n )f{tn ).

now

nitely in
*
It is

number of points z lt ., zn be increased indefisuch a manner that the length* of each of the arcs
the
. .

assumed
the

in

may be
or,

associated with the portion of


is

regard to every path of integration that the idea of length it included between any two of its points,
is

what

same
if

thing, that the path

rectifiable.

This condition

is

evi-

dently satisfied

the current coordinates

and^j/

can be expressed in terms of

258

FUNCTIONS OF A COMPLEX VARIABLE.


1
,

[Chap. VI,

z z 1 ,z za ...,zn

approaches zero as a
proaches a

limit,

the

sum 5
is

finite limit

which

apinde-

pendent
z

of the choice of the points z lt


/
,

..., zn and

/,,..., tn

For take any other sum

(*.'-*,W)+...
formed
for
in a similar

manner.

Suppose

the sake of greater definiteness that the points #,,... and */, . follow one another on the line L in the order
.
.

and form a third sum

in

which both

series of points occur.

It

may

be shown that as
. . .

the

number

of points in each of the series *

and

*/, ...

is

increased, the differences


zero,

S"

S and S"

both approach

from which

it

follows that the difference

S
S"

S' has a

limit equal to zero.

For example, the

difference

has

the value

<?L $H*>
If

- *.)[/(r.) -At.)l+W - *,>I/W -Atil + W - <)[/(r,) - Ml +


of the quantities

. .

M denotes the upper extreme


l/W -M)\,
\A*,)

^.,- 2'JDi
fc* -2-*)f

-m\.
be
less
1

\M) -M)\
than

the modulus of
7l/[!^

S"
1

S will

-^| + |<-5r H-|<-Vi+...].


dx
and
dy
are

any parameter
/
2

so that

continuous.

For then the integral

4/akr

-f-

dy

is finite.

See, in this connection, Jordan, Cours d'Analyse, 2d

Edition, Vol.

I., p.

100.

ART.

15.]

INTEGRAL OF A FUNCTION.
z
is
\

equal to the chord of the arc z^z^ and must therefore be less than or equal to this arc, and a similar result holds for each of the quantities *,' z *t Hence *,'

But

\z x

'

\S"-S\<M/
where
the
/

number
/(n)

denotes the length of the path of integration. When of points of division on the line L is increased, the

differences

- AU\
since

/0.)
f(z)

- /(',).
is

/W - A*,),

approach
zero.

zero,

continuous,

accordingly

decreases indefinitely and the difference S"

approaches

The
strated,

limit,
is

the existence of which has just been demoncalled the integral of f{z) along the path L. It is

written

I f(z)dz.

The

definition here given

is

similar to that

given for the integral of a function of a real variable. It is unnecessary to specify the path of integration when the indeit

pendent variable is restricted to real values, since in that case must be the portion of the axis of reals included between

the limits of integration.

The following well-known principles, applicable to the case of a real independent variable, may be readily extended to the
general case
:

1. The modulus of the integral cannot exceed the length of the path of integration multiplied by the upper extreme of the modulus of the function along that path.

2.

The independent

variable

may

be altered by any equait is

tion of transformation, but L', the path of integration in the

transformed integral, must be such that

described by the

new

variable while z describes L.


If

3.

F(z)

is

any one-valued function having everywhere

f(z) for its derivative, the equation

fjWz = F(Z)-F(z^
must be
true.

260

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

To
form

prove the third principle, write F{Z)

F(z

in

the

F[Z)-F{Zl,)+F{zu)-F(Zn _,) +

+ F(s,)-F(z;)+F(z,)-F(z

a ).

Since the derivative of F(z) \sf(z),

F(*m+ l )

F{z m )

[/(*)

tj

m ](zm+1
is

- zm

),

where
zm
.

t]

m has zero for

its limit

when zm+1

made

to approach

Hence
F(z
)

F(Z)
or, since

= limit

2f(z m )(zm+1

zm ) -f limit 2r}m (z

l+1

zm)

the second term of the right-hand

member

is

equal

to zero,

F{Z)-F{z,)
If

= ff{z)dz.
L

no function F{z) fulfilling the preceding conditions is known, the value of the integral requires further investigation.

Consider as an example the integral

taken from the

I to the point z I, the path of integration being point z half of the circumference of a unit circle described the upper

about the origin as a center.

Writing z

exp

(iff),

z will

describe the required path while 6 varies from

n to

o.

The
dz
z

equations -,

e~ 2iB

dz

ieiedd,

= ie-dd = i cos
Hence
+1 J

dd

+ sin

dd

id (sin

6)

- d (cos

6),

follow at once.

for the

path specified

C
_1

= i Cd (sin 6)
IT

Cd (cos
V

d)

2.

application of the direct and more familiar the same result gives
:

The

method

~~
z*

"~"

z J, =I

zj s= _i~

ART.

16.]

REDUCTION OF COMPLEX INTEGRALS TO REAL.


of reals

261
limits of

For a path along the axis


integration this result
is

between the

unintelligible.

The

discontinuity of

dz
the differential, -3, at the point z z
tion of such a path
;

= o,

prevents the considera-

and that the


is

result should

be negative

when the

differential

at every point of the path positive

has no significance.

introduction of the complex variable furnishes a perfectly satisfactory explanation of the result.
Prob. 16.

The

Show

dz
that the integral of

along any semi-circumis

ference described about the origin as a center

equal to ni.

Art.

16.

Reduction of Complex Integrals to Real.

The

integral

may

be written

in

the form

I {u-\- iv){dx
or,

+ ufy),

separating the real and imaginary terms,


/ (udx

vdy)

-f- i

I (vdx

-f-

udy).

calculation of the integral may be reduced to the calculation of two real curvilinear integrals.

Hence the

The equations
du

_ dv

du

__

dv

'dx~dy'

dy~

dx
iv should

which express the condition that u-\express also that

be monogenic,

udx

vdy,

vdx

-\-

udy

are the exact differentials of


x, y.

two

real functions of the variables

Consider the case where these functions are one-valued.

262

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

Denoting them by P(x, y) and Q(x, y) respectively, the


gral

inte-

may be

written

\I\X, Y)
(x ,y
)

- J\xyM + iWX,
initial

Y)

Q{x.,y.%
re-

and (X, Y) being the

and terminal points

spectively of the path of integration.

Art.

17.

Cauchy's Theorem.

Cauchy's Theorem furnishes the necessary and sufficient conditions that a one-valued function f(z), having continuous
partial derivatives with respect to

x and

y, should yield within

a region bounded by a continuous closed curve a one-valued integral, that is, an integral the value of which, when the lower
limit
is

fixed,

depends simply on the upper


It will

limit,

and not on the

be more convenient, before considerpath to demonstrate the following lemma: ing Cauchy's Theorem,
of integration.

Lemma. Let A be a portion of the .s-plane, having a boundary 5 which Consists of a closed curve not intersecting itself, or of several closed curves not intersecting themselves or one
every point of the region A, including its boundary 5, a function Wot the real variables x and y is onevalued and continuous and has continuous partial derivatives
another.
If

at

f,f, the

relations
^

^*J?*)
(2)

fw*--f*jfc*
exist,

the integrals in the first members being taken along the boundary in the positive direction, and those in the second members being taken over the enclosed area.

normal
*

Denote by A the inclination to the axis of x of the exterior at any point of the boundary,* that is, the normal drawn
It is

assumed

that the

boundary has a determinate tangent

at every point.

Art.

17.]

cauchy's theorem.

,03
ft

Xtvwl
to the right as the boundary
line
is

OAL*-

K~hu4

described in a positive direction. If any straight parallel to the axis of x be traced in the direction of increasing values of x, at each point where
it passes into the area A, cos A. is negative, and therefore in the first member of

At each

negative. point where this straight line passes out of the area A, cos A, and there(i)

dy = cos \ds

is

fore

dy

in the first
(i),
is

member
positive.

of equation

Hence
of
entials

in

the

first

member
differ-

equation

(i)

the

Wdy
y

corresponding

to a given value of y, and taken in the order of increasing values of x have signs which, compared with the signs of the

corresponding values of W, first differ, then agree, and so on alternately. In order now to compare the integral in the first member of equation (i) with the integral in the second

member,

it

The sum
form

necessary to take dy as essentially positive. of the differentials in the first member, correspondis

ing to a fixed value of

y must
%

therefore

be written

in the

dy{where

W,+ W.-W,+
.

1V -...), t

are the corresponding values of taken in the order of increasing values of x. But performing now in
lf

W^

the second
to x, the

member

of equation (i) an integration with respect


is

same result equation (i) become

obtained, so that the two

members
verified.

of

identical,

and the equation

is is

obtain equation (2) the same necessary in this case to observe that

To

method
if

used.

It is

a line parallel to the axis of y is traced in the direction of increasing values of y, at each point where it enters A, dx in the integral of the first
If the

J<3

boundary

of a given region

is

not of this sort, the theorem holds for any


is

interior curve of

which

this

assumption

true.

264

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

member must be
negative.

this line passes out of

taken as positive; and at each point where A, dx in that integral must be taken as

By means
easily proved.

of the preceding

lemma, Cauchy's Theorem


be stated as follows
:

is

This theorem
If,

may

Theorem.

on the boundary of and within a given region

A, a one-valued function
ative f\z)
is

=/(#)

is

monogenic, and

its

deriv-

continuous,* the integral

j f(z)dz

taken along

the boundary

is

equal to zero.
integral in the

For writing the


/

form

wdz

= j (iidx
gives

vdy)

f\udy -j- vdx),

the preceding

lemma

but since at every point of

the given integral reduces to zero.

Art.

18.

Application of Cauchy's Theorem.


it

From Cauchy's Theorem

follows that,

if

two

different

paths Z, and Z 2 lead from the point z Q to the point Z, and if along these paths and in the region inclosed between them a given function f(z) has no critical points, the integrals of the
function taken along these two paths are equal. For two such taken together, one described directly, the other repaths
versed, constitute a closed curve, and the integral taken along
* Otherwise
expressed, the one-valued function /(s) has no critical points on
is

the boundary of or within A, or f(z)

holomorphic

in

A.

ART.
it is

18.]

APPLICATION OF CAUCHY'S THEOREM.

265

equal to zero.
of integration

path

is

But, since reversing the direction of the equivalent to changing the sign of the

integral, the equation

is

obtained.

The
theorem

result just established


:

may be

stated in the following

Theorem I. If a function is holomorphic in any simply connected region bounded by a continuous closed curve, the integral of the function, from a fixed lower limit in that region to any point contained therein, is independent of the path of
integration, and
is

a one-valued function of

its

upper

limit.

region whose boundary is composed of disconnected is not necessarily characterized by the property stated in the theorem. Take, for example, the function

curves

w
and suppose

\/{z

a
|

)(z

<*,)...(*
|

a n ),

that o . . . am \. With the ori\at ]at as a center, construct a system of concentric circles Clf gin C..., C, C{ passing through axi C3 through a and so on.

<

<

<

<

Denote by S the region inclosed within the first circle Clt by that inclosed between C and C and so on, the portion
Q
t
t
.

of the plane exterior to the last circle Cn being denoted by Sn At an initial point z interior to one of these regions, assign to

w one

of the

two values

possible,

and consider the branch of

w resulting from a continuous variation.


. .

Then however z may

will be a monovary within any such region, this branch of and its derivative will be continuous. Having genic function,
it is evident that in ., a n regard to the branch-points a lt a t the regions Sof S9 ... it will be one-valued, and in the regions Thus in the regions 5a St it will be two-valued. Slt Ss
, ,
,

the branch
not.

fulfils

does

The

the required conditions, but the boundary It may be theorem is applicable only to 5
.

every other region two paths may be drawn the same two points such that the branch is not onejoining valued in the enclosed portion of the ,s-plane.

observed that

in

266

FUNCTIONS OF A COMPLEX VARIABLE.

[Chap. VI.

Tl/

>

Theorem II. lif(d) is holomorphic in any simply connected region S bounded by a continuous closed curve, the integral
$>
> /

I f(z)dz,

taken from a fixed lower limit z in that region to any


contained therein,
is

point

a holomorphic

function of

its.

upper

limit.

Let

at the point

-\-

dZ

be any path from z to Z. When the upper limit is to> dZ, L followed by a straight line from can be taken as the path of integration. Hence

Z+

nz+dz

nz

nz+dz

fo-P,
\A*)-AZ)Y>
The modulus
is

S*Z+dZ

pZ+dZ
z d,+J

Z =AZ)J

The
term
is

first

term

is

equal to f(Z)dZ.

of second

equal to or less than


of \f(z)
is

M\

dZ\, where

treme

f{Z)\ along the line


zero.

But since f{z)


approaches

continuous, the limit of

M the upper exdZ^ joining Z to Z M when Z dZ


-\-\-

is

Hence

"*A*y* - f*A*V*
where
r/

\AZ)

+ nZ,

approaches zero with dZ. The integral therefore has /(Z)fora derivative, and is holomorphic in 5.
In the case of a region bounded by several disconnected closed curves, of which one is exterior to all the others,.

Cauchy's Theorem

may be

stated in the following form

Theorem
region curves

A C

Let a function j\z) be holomorphic in a bounded by a closed curve C and one or more closed
III.
C%*
*

x%

interior to C.

The

integral of f(z) taken

along

will

integrals
u*

be equal to the sum of its taken in the same direction

^\
the curves
(7,,

along the curves Clt Ct > p or t h e integral of f(z) taken

in

positive direction completely around the

is equal to zero. But boundary of are then described in the direction oppo-

ART.

19.]

THEOREMS ON CURVILINEAR INTEGRALS.


which

26?
the curves

site to that in

is

described.

Hence

if

all

are described in the

same

direction, the result

may

be written

c c fjw* =f/w* +f
If

dz

there

is

included between two curves

but one interior curve, so that the region A is C and Clf the integral taken along
x

every closed curve containing C but interior to C has the same value, viz., the common value corresponding to the paths

C and

C\.

Art.

19.
I.

Theorems on Curvilinear Integrals.


If f{z)

Theorem

be continuous

in

a given region except

at the point a, the integral I f{z)dz, taken


,

around a small

circle

having its center at a, will approach zero as a limit simultaneously with the radius r of the circle c, provided only
lim (z
d)f{z)

=o

when

= a.
a)f{z)
c,

For

let

the upper extreme of the modulus of (z

on

the circle c be denoted by


a

M.

Then

at

every point of

\-

M
2tzM.

and consequently

mod^ f(z)dz
Theorem
closed curve
II.

J ds
v
-,

The

integral / fa

r-, ny

taken around any J

(z

a)

when n
.

I.

C containing the point a, is equal to zero, except When n = I, this integral is equal to 2ni.
of

For the value

the integral will be the same

if

any

circle described

integration.
varies from

about a as a center be taken as the path of Let then z a reiB where r is a constant and

=
e

o to

2rc.
#

The
r-

integral

becomes

/2ir

(n-l)iO

rn-i

/ J

dd

%G8

FUNCTIONS OF A COMPLEX VARIABLE.


I.

[CHAP. VI.

which reduces to zero except when n ==


is 27rz,

If

I, its

value

whence

SA
z

dz
:

27tl.

a function holomorphic in a given closed curve the interior of which is wholly region S, C within S, and a a point situated within C, then
III.
a.

Theorem

If f{z) is

^c

&-dz
a

= 27tif{a).
v J

r,

For describing about a as a center a small circle c of radius the equation

^c z
is

*Jc

za
c,
t]

obtained.

But

at every point of

where, by choosing r sufficiently small, the modulus of be made less than any fixed positive quantity. Hence

may

^cza
zero.

vc z

^cz

but by the preceding theorems the first term of the right-hand member is equal to 27tif(a), and the second term is equal to
the equation of the theorem just established be differentiated with respect to a, the following important formulas, expressing the successive derivatives of a holomorphic function
If

at a given point, are obtained

Art.

20.]

Taylor's series.
integrals in the
first

269

The
finite

members

of these equations are all

and determinate for every position of a within the curve Therefore any function holomorphic in a given region admits an infinite number of successive derivatives at every
C.

interior point. Each of these derivatives being monogenic must be continuous. Hence the following:

Theorem IV.
there exists an
f(z) which are
y

If f(z) is

holomorphic within a given region,


of successive

infinite

number

derivatives of

all holomorphic within the same region. Denote by r the shortest distance from the point a to the

curve Cm

Then

M be the upper
length of C.

at every point of this curve \z

a\

> r.
and
/

Let
the

extreme

of the

modulus f(z) on

Then

mod

/c (z

-a)

n+l

dz

(n)

270

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

From

the preceding

article,

denoting a variable point on

2nt

Jc

-f

= jl

27iiJ c

fm^v 1+ __ Z-a\_ ^Z-a^


,

1 +^l_+ T (C-a)^(C-)'*(C--^J

= A") + */'(<*) + 7^r/"<) +


I
.

7-

+ T-fzf'W + R
1
2i
. .

'

71

where

r - _L f

" +1

+ 2niJc (Z-aY \Z-a-t)

AQ

r _

By taking n sufficiently great the modulus of made less than any given positive quantity. Let

R may

be the upper extreme of the modulus of f(z) on the circle C, p the a or radius of C. Then modulus of t, and r the modulus of Q

be

27t

Jc

r n +\r-p)
its

< r-p\rj
when n

which, since p

<

r,

has zero for


f

limit

= oo

Writing now z for a-\-t Taylor's Series becomes

is convergent and the equality is maintained for z included within a circle described about a as a every point center with a radius less than the distance from a to the nearest

The

series

critical

point of f(z).

When

is

equal to zero, Taylor's Series takes the form

M=

/(o)

+ *f(d) + ^/"(o) +
in

+ _!_/<-)(o) +...-,
This form
is

expressing/^)

terms of powers of

z.

known

as Maclaurin's Series.

Art.

21.]

Laurent's

series.

271

Art.
Theorem.
region be the

21.

Laurent's Series.
bounded by two

Let

S, a portion of the ^-plane


1

concentric circles

and

72

be situated

in

the interior of the

which a given function f(z) is holomorphic. If a common center of the two circles, and a -j- t a point interior to S, f(a-\- 1) can be expressed in a
}

in

convergent double
f{a

series of the
tn

form

+ = =2AJ". m
/)
oo

oo

With a
^

as a center construct a circle

sufficiently small to be contained within the region 5. If then C be the greater of


x

the two given

circles, it follows

from Article 18 that

2ni*'c ^
i

a
Article

2ni^c%

2ni

J'

But from

icj^.'Ufc

^
=f{a

i_r
whence
yv

/(ck

t)>

2ni Jc

^-

-t

27tiJc*Z- a

i
written
:

The two

integrals of the right-hand

member may be

^Cv

where

AQdZ =2 / 2V^ m^(Z-ay+\Z -a-t)' p - _L AC *Y ACW "> ~


tf 1
+l

t n+1

27ziJc>t

n+ l

But

|/|

<|C

*|

at every

(Q-a-t)' point of C and


l%

|/|>|C

.*)

at

every point of

so that R, and i? 3 both have zero for a limit

272

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP.

VH

when n
in the

oo

The

value of f(a

t)

can therefore be expressed

form
f(a

+ t)=A + A + AS + Af + ...
1

m+1 the function f(z)/(z a) phic for both positive and negative values of written

Since in the region

is

holomor-

m A m may
f

be

a * m - -L C 2t7zJ
c

f^ dr - a)"* a^ &
1

where

is

any

circle concentric

with C, and

and included

between them.

The series thus obtained is convergent at every point a 4- 1 contained within the region S. It is important to notice, however, that when the positive and negative powers of t are considered separately, the two resulting series have different
regions of convergence.

powers of

containing the positive converges over the whole interior of the circle C ;
series
x

The

while the series of negative powers of / converges at everyThe region 5 can be regarded, point exterior to the circle 7a therefore, as resulting from an overlapping of two other
.

regions in which different parts of Laurent's Series converge.

Writing z for a
f(z)

-f- /,

Laurent's Series takes the form

= A + Az - *)+

Alz
1

+ A _,

z-

a)

- a y+... + A_ (z - a)- +
2
2

Consider as a special numerical example the fraction


Z

{Z If \z\

I) (Z

2) (Z
all

3)

2(2

I)

+ 2~ 2(Z

3)

<

1,

three

terms of the second member, when

developed

powers of z, give only positive powers. If 1 < < 2, the first term of the second member gives a series\z\ of negative descending powers, but the others give the same
in

series as before.

If 2

<

\z\

<

3,

the

first
3,

and second terms


three terms give

both give negative powers.

If \z\

>

all

Art.

22.]

Fourier's series.

273

negative powers, and the development of the given fraction can contain no positive powers. Thus a system of concentric annular regions is obtained in each of which the given fraction
is

expressed by a convergent power-series.


results for every function

Laurent's
is

Series gives analogous morphic except at isolated points of the ^-plane.

which

holo-

Art.
Let

22.

Fourier's Series.
in a region
go,

w = f(z)
is

be holomorphic

and

let it

be

periodic, having a period equal to

so

that/^

+ #<) =/(^)
; .

any positive or negative integer. Denote by Sn the obtained from 5 by the addition of nco to z and supregion S 5 Sn ... S_ n 5_ pose that the regions

where n

X ,

such a manner as to form a continuous strip in which, of course, the function w will be holomorphic. S, Draw two parallel straight lines, inclined to the axis of reals at

meet or overlap

in

an angle equal to the argument of cw, and contained within the The band T included between these parallels will be strip 5. interior to S. 1 4 o wholly
2iriz

By means of the transformation z' = e w the band T in the ,2-plane becomes in the '-plane a ring T' bounded by two
z
concentric circles described about the origin as a center, z and Since is holomorphic -f- hod falling at the same point z\ in a region including T, and

dw
dz'

dw

dz

gd

2iriz
'

dw
dz'

dz dz'

2ni
will

regarded as a function of z' Hence, by Laurent's Theorem,

be holomorphic in

T\

w= w 2
the quantity a
in this

m=

<x>

= -<

A mz h\

being the preceding equation becomes

in the general formula of the preceding article case equal to zero. Substituting for z' its value,

w = 2 A me

274

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

where

rwdz' A = ^iJc^ = vJ/


i i
2

z+w _ 2m " e

wd*'

In the latter integral the path is rectilinear. Denoting its independent variable by C for the purpose of avoiding confusion, the value of

w becomes

M = -^C
m=-oo
I

e~

/{Z)dZ
2 w=co

/+ w
r^sr*

/+ w

imn

i/

/^c+^cs + -^
.

ic

2//Z7T.S

^
2M71Z

/^ w cos 2mnQ
2ni7lC
sin

-/(eye

co

sin

/+w
/
b

V(CKC

=1

G
Art.
Let the
term
of
series
is

23.

Uniform Convergence.
-\-

W= w
5.

u\

+ w^

-\-

ze/ n

each

which

a function of

of a given region

z, be convergent at every point Denote by n the sum of the first n

terms of W.
tive quantity

If it is possible,
e,

whatever the value of the

posi-

to determine an integer v, such that

whenever

> v

\W- Wn \<e

at every point of S, the series

is

said to be uniformly con-

vergent
in

in the region 5.

Uniformly convergent
exactly the same
of terms.

series can in

many respects be treated


finite

manner

as

sums containing a
series,
itself

number

uniformly convergent which are continuous functions of z, is


I.

Theorem

the terms of
a continuous

function of

z. z,

For

at

any point

W may be written

in

the form

Art.

23.]

uniform convergence.
n

275
r
,

JV=

W + R;

and

at a

neighboring point z

W= W

'

+ R'.

Hence

W-W'=W -W '+Rn n

R',

and
|

IV-

W'\ =|

- W:\ + \R\ + \R!\.


and
\R'\
.

But by choosing n
be made
less

sufficiently great, \R\

may both
Having

than any given positive quantity

chosen n thus, n becomes the sum of a finite number of It is then continuous, and, by making functions. continuous
\z'

z\ less

than a suitable quantity


.

d,

\W

n '\

ma y

De W<n

made

less

than -

But, under these suppositions,

\Wis,

W'\<e.
z.

therefore, continuous at the point


II.

Theorem
series

If all

the terms of a uniformly convergent

W = W + w, +
o

+w

-f

are continuous, the integral of the series, for any path L situated in the region of uniform convergence, is the sum of the
integrals of
its

terms

Sl

Wdz

=
e

Sl

w ^z +X Widz +
n
it is

Wnd*

For, writing

W= W + R,
may
be, \R\

however small
so chosen,

<

possible to choose n so that, 6 at every point of Z. If n be

fWd^fwjz + fRdz..
But,

by

Article 15, denoting

by

the length of the path L,


e/,

mod / Rdz <


which, when n

00

has zero for


lim

its limit.

Hence

f Wdz =

fwjz.

276

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI
.
.

Theorem
is

III.

If

the series
series
'

JV= w

-\-

w -\x

+w

n -\-

..

convergent, and the


,

dw __ dw, "" dv>i ~ '"'r n t '" dz ds"r ~dk"

is

of

uniformly convergent in a region S, and if further the terms are continuous in S, will be the derivative of W.

For, integrating
in S,

W from a
w,{a)

to z along a path

contained

J W'dz
of

w (z)

+w

n {z)

wn {a) +
is

^W{z)-W{a\
But the derivative
of the first

member

W,

which must

also be the derivative of the second

member, and therefore


is

W.

An

immediate consequence of the preceding theorems


:

the following

Theorem IV.

If

the terms of the convergent series

+ w, -f
dw
\

+w +
n

are holomorphic in a given region of convergence, and if the series


ytrr
"

contained in the region

dw
dz

"*"

dz

"t"

T
i

dw
dz

T
i

'

"

"

is

uniformly convergent, ^Fwill be holomorphic for its derivative. S, and will have

in

the region

To
is

illustrate

by an example that uniformity

of

convergence

essential to the preceding theorems, take the series

W=
At
the point z
is

-iI

*"(i-

*)

each term is continuous, and the series the value 1/2. The series is, however, convergent, having i< discontinuous at z For, writing it in the form

w=-

L_\+(<~i

L.\

Art.

23.]

uniform convergence.
first

277

the sum of the

n terms
"

is

seen to be

+z

But

is

the limit of

when n
\z\

= oo
I,

and

is

therefore
at every

unity at
point for
If

every point z for which

<

and zero

which

\z\

>

I.

now

this series

upon a

circle

be considered for the points within and described about the origin as a center with an
z
n

assigned radius less than unity, the remainder after n terms,


or
I

W=
n
I

-f z

can,

by a J

suitable choice of n be
y

made

absolute value than any given quantity. In such a region, then, the series converges uniformly, and, by Theorem similar result holds I, can have no point of discontinuity.
less in

for the region exterior to any circle described about the origin as a center with an assigned radius greater than unity.

By means

of

Theorem
is

II given

above

it

can be shown that

For, assuming the notation used unique. in the determination of the series, the series is uniformly con-

Laurent's Series

vergent in the region included between any two given circles concentric with C and Ca both being interior to C and exx

terior to

Cr
:

Suppose, now, that two such series are possible


m

f(a
,

t)=m

= -

2 A mr
oo

oo

=
,tt

= _oo

2 A m 'r.

Divide by /n + 1 and integrate along any circle described about rtasa center and included in the region of uniform converg.
ence.

The

integral I

- n

dt for such a path

is

zero, except

when

m=

n;

the integral I t~ l dt

2* Jr.

Hence

for such a path,

278

FUNCTIONS OF A COMPLEX VARIABLE.


it

[CHAP.

VL

from which
identical.

follows that

An = An

',

and the two

series are

Art.

24.

One-valued Functions with Critical Points.

function holomorphic in a region 5 and Theorem I. not equal to a constant, can take the same value only at isolated points of
vS.

For

in

the neighborhood of any point a interior to S,

by

Taylor's theorem,
/(*)

-A") =

(*

**'(<*)

+ ^=^-"/"() +
.

constant over the entire circle of convergence of this series, the derivatives /'(a), f'\ci), cannot all be

Unless/^)

is

equal to zero.
zero.
J

Let f {n \a) be the

first

which

is

not equal to

Then
v

'

[_i

'

v
. . .

(n

-\- i)

If

\z

a\

modulus

of the

be given a finite value sufficiently small, the first term of the series within the brackets will
all

exceed the sum of the moduli of

same result For values


finite

will

hold for every


is

still

the other terms, and the smaller value of \za\ +

of

z,

then, distant from a


f(a)
different

amount, /(^)

by less than a certain from zero.

If, on the other hand, the function is constant over the entire circle, described about a as a center, within which Taylor's series converges, it will be possible, by giving in succession

new

positions to the point a, to show that the value of the function is constant over the whole region 5.

Theorem
in a

II.

given region

Two functions which are both holomorphic 5 and are equal to each other for a system of

points which are not isolated from one another, are equal to each other at every point of 5.

For let/(^) and

<p(z)

be two such functions.


<p(z)

By

the pre-

ceding theorem, the difference/^) zero at every point of S.

must be equal to

Art.

24.]

one-valued functions.
III.

279
in

Theorem

function which
is

is

holomorphic

every

part of the -s-plane, even at infinity,

constant.
g,

For, a being any given point, whatever the value of

As)

= /(a) + (* - ay '(a) + ... + ^'^ ,/

*(<*)

But by Article 19, r being the radius of any arbitrary circle being the upper extreme of the having its center at a, and modulus of f{z) on the circumference of this circle, (9'^(p*\

M
N

mod,

f \d)

, lM)/ n

nM
.

<

But

is

always

finite,

Hence f {n \d)

for all values of n,

and r may be made indefinitely and


f{z)=f(a).

great.

function/^), holomorphic in a region 5, is equal to zero at the point a situated within S, the function can be expressed in the form
If a

Theorem IV.

f{z)

= (z- ay<f>(s),
<p(z) is

a positive integer, and different from zero at a.

where

is

holomorphic

in

5 and

For

in the

neighborhood of the point a, by Taylor's Theorem,

A>)=A*) + (
Let/
is
(w)

-}/+
J 7

(a)

be the

first

of the successive derivatives at

a which

not equal to zero.

Then

f(z) v

(z

m
a)
\

.2.

^m
. .

-L
t

'

1.2

(w+
is

-(z
1)

a) J

I,

which

is

the required form.

The

point a

a zero of f(z) and


f

is its

order.
If

the point a is a critical point of a given function /(s), but is interior to a region 5, in which the reciprocal of j\z) is holomorphic, the function can be expressed in

Theorem V.

the form

where

is

a positive integer, and x{z)


of a.

is

holomorphic

in

the

neighborhood

280

FUNCTIONS OF A COMPLEX VARIABLE.

[Chap. VI.

For by the preceding theorem

jA
where
<p{z)
is

= (* -

w
tf)

0<,
at z

holomorphic and not equal to zero


f(z\ J \9

a.

Hence

in

l
'

^_ a yn
-f

0(^

^ _ ay

*&

Further, since

a region

of finite extent including the

point a
X(*)
f(z) JK J

=A

(z
.
.

-) + ...,
. '

=
(z

dl L - q) m ^

n h ^z -a ^ +(g\
A-dauzi _L

a being an ordinary point

for tp(z).

The
ing no

point a

is

a pole of f(z)

and

is its

order.

Theorem VI.
finite

function, not constant in value, and hav-

points except poles, must take values near to every assignable value. arbitrarily
critical

For suppose that f(z) is such a function, but that it takes no value for which the modulus of f(z) A is less than a given
positive quantity
e.

Then

the function
i

/(*)
will

-A
is

be holomorphic
III, is

in

every part of the ^-plane, which, by


a constant.

Theorem

impossible unless f(z)

function f(z), having no critical point except a pole at infinity, is a rational entire function of z.

Theorem VII.

For the only

critical

point of

is

a pole at the origin.

Hence

/- #+...+4 +0W
where
<fiz) is

holomorphic over the entire plane, including the

is consequently equal to a constant (f)(z) point at infinity. The given function therefore can be written in the form

f(z)

= A m + ...+A z + A
z>
1

.Art. 24.]

ONE-VALUED FUNCTIONS.

281

Theorem VIII.
are poles
is

A function f{) whose


z.

only

critical

points

a rational function of

The
other
;

poles must be at determinate distances from one anotherwise the reciprocal of f{z) would be equal to zero

for points not isolated from one another. The number of poles cannot increase indefinitely as \z\ is increased; for then the

reciprocal of

would have an

infinite

number of zeros

indefi-

nitely near to the origin.

The
.

total

number

fore

finite.

Let

a, b,

denote them.
in

of poles is thereIn the neighborhood

of a the function can be expressed


m
(z
4X
'

the form

'

a)

a
In the neighborhood of
b>

being an ordinary point for cf>(z). in the form <p(z) can be expressed
u
*
'

'

'
'

(z

b)

b
*p(z).

a and
this

b being both ordinary points for

Proceeding

in

way the given function will be expressed as the sum of a finite number of rational fractions and a term which can have
no
critical

point except a pole at infinity.

This term

is

rational entire function.

the function f{z) has no zeros and no critical points for finite values of z it can be expressed in the e^ z \ where g(z) is holomorphic in every finite reform f{z)
If
t

Theorem IX.

gion of the

-s'-plane.

For

f(z) -~~

can have no

critical points

except at

infinity, since

in every finite region of the ^r-plane f(z) and f'{z) are holomorHence, choosing an arbiphic and f{z) is different from zero.

trary lower limit z

the integral

/
is holomorphic in every finite region. sequently must take the form

The

function /[z) con-

^82

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

where

g(z)
If

= k(z) + log f(z

).

Theorem X.
if

two functions/^) and

<p(z)

have no

criti-

cal points in the finite portion of the ,2-plane

these poles are identical in position and in functions, and their zeros are also identical in position order, there must exist a relation of the form
f{z)

except poles, and order for the two

and

<p(z)^\
finite

where g(z)

is

holomorphic
ratio of the

in

every

region of the ^-plane^


no>

For the
critical

two functions has no zeros and

points in the finite portion of the .s-plane.

Art.
If a
is

25.

Residues.
critical

one-valued function has an isolated

point

a, it

by Laurent's series in the region comprised between any two concentric circles described about a with radii
expressible
less

than the distance from a to the nearest


in

critical

point*

Hence

the neighborhood of a
/(s)

= A. + A (s-a) + A (*-ay + ...


l

+ B (z-a)- + B&-a)-> + ...


i 1

The
If

coefficient

of (2

a)'

in
a.

this

expansion

is

called

the

"residue" oi f(z) at the point

any closed curve

C including

the point a be drawn in the

region of convergence of this series, and f{z) be integrated along C in a positive direction, the result will be

J (fz)dz = 2niB
The
theorem
following
:

may be

regarded as an extension of Cauchy's

Theorem

I.

If in

one-valued function

a region 5 the only critical points of the f ... the /(.s) are the interior points a, a
,

Art.

25.]

residues.

283

integral

j f(z)dz
is

taken around

its

boundary

in a positive

direction

equal to

fA*)d* = 2*i(B + +
>

...),

are the residues of f(z) at the critical points. where B, B\ For the integral taken along C is equal to the sum of the integrals whose paths are mutually exterior small circles de.

scribed about the points a, a!

>

The

following theorems are immediate consequences of the


:

preceding
only

Theorem
critical

II.

If in

a region having a given boundary

C the

points of the one-valued function f(z) are poles

interior to C, an equation
-dz

exists,

M
in

cA J}. z) denoting the number


order. |p^^l

2iit(MN)
of zeros

and

N the number of
number
of times

poles within C, each such point being taken a

equal to

its

For

the neighborhood of the point a

/( Z )

(*

)-0(*)j|j^yiv
a,
if

where

<p(z) is finite
if

and different from zero at


is

and
a
is

is

positive integer

a zero, a negative integer

a pole.

Hence
/'{*)

= m
z

,,

0Xf)
<P(zY

f(z)

a**

fj^
\

2L-*

integrand, therefore, has a pole at every zero and pole of f(z), and its residue is the order, taken positively for a zero, and negatively for a pole.

The

Theorem
roots.

III.

Every algebraic equation

of degree n has

For
zn

let f{z)
~
l

a zH
x

represent the first member of the equation an o. Since f(z) has no poles in the
.

+ =

284

FUNCTIONS OF A COMPLEX VARIABLE,


the

[CHAP. VI.

finite part of the ,2-plane,

number

of roots contained within

any closed curve

C will

be given by the integral

'

ni

**?* /(*) Jc A
is

But taking

for

a circle described about the origin as a

center with a very great radius, this integral

j_
27Z

k-' + (-y- + * + a i? +* Vc
e has zero for a limit
|sr|

...

dg

= J_
oo
.

Cndz
Hence the

2nl Jc z
limit

vj

where

when
is

\z\

of the preceding integral, as


Prob. 17.

increased,

is u.

is, if

00 is an Show that if z ordinary point of f(z), that /(s) is expressible for very great value of z by a series containing only negative powers of z, the integral o(/(z) around an infinitely
is

great circle cient


is

equal to 2iti into the coefficient of

This

coeffi-

called the residue for z

00

Prob. 18.

Show

that the

sum

of all the residues of f{z), of the


is

preceding problem, including the residue at infinity,


zero,

equal

to

f^l
io.

Prob.
is

If

tt\

is

a rational function of which the numerator

of
,

4,

degree lower by 2 than the denominator, and if the zeros ,...,# of the denominator are of the first order, show that

=
A
>

o.

*-

Art.
It

26.

Integral of a One-valued Function.


in Article 18 that,
if

was shown

a function/^)

is

holo-

integral taken from a fixed morphic in a given region lower limit contained in 5 to a variable upper limit z is a oneS, its

valued function of z within 5.

If F{z)

is

a function which

z and is one-valued takes a determinate value F(z ) at z while z remains within S, having at every point f(z) for its
derivative,

the
).

integral
If

of f(z)

from z

F(z)

F(z

F (z)
x

is

another function

equal to these confulfilling


to
is

Art.

26.]

integral of a one-valued function.


can be written also
in

285

ditions, so that the integral of f(z)

the

form (z) F^z^), the functions F(z) and a constant term for 3,$) - ^< ^o) - *\
x
l

F (z)
x

differ

only

by

Suppose now that f(z)


has isolated
critical points

is

still

one-valued in S, but that


interior to 5.

it

#,,#,,... paths from z to z, which inclose between them a region con., will give integrals identitaining none of the points a lf a 2
,
. .

Any two

Let the two paths L lf L include between them and consider the integrals along a single critical point aK these two paths. The integral along L will be equal to the
cal in value.
;

integral along the composite path


i

L^L^L, where the exponent


is

indicates that the corresponding path

reversed

for the

integral curve, or" loop," including the critical point a K , and, assuming that it is described in a positive direction about a KJ the intex

along L'

is

equal to zero.

But

L~

is

a closed

gral along
at a K
.

it is

equal to 27tiBKi where

K is

the residue of f(z)

Hence
ff{z)dz

= 2niB
L
,

K -\-

ff{z)dz.

If

now
,

the two paths Z,,

from z to z include between


y

them
utive

several critical points a K a ky a^ ., draw intermediate Z3 Lmy so that the region between any two consec., paths
.
.

The integral paths contains only one critical point. L will be equal to the integral along the composite path along L L~ L^ Lm ~*L m L~ L, since the integrals corresponding to
x
X

Z _1 Z,, L L '\
a i t

.,

.,

L m L m L~ L are all equal to zero. But L L~\ L m L~ are all closed paths or loops, each including
X
,

~1

a single critical point, so that, assuming that each is described in a positive direction and that denote the resiK Ky B^,

B B
,

dues of f{z) at the

critical points,

f^Mds =
It

27ti(BK

+ B, + B. +

.)

+ f A*YL

has been assumed in the preceding that neither of the paths L lt L intersects itself. In the case where a path, for

286

FUNCTIONS OF A COMPLEX VARIABLE.

[Chap. VI.
.

example

lt

intersects itself in several points c%% c%t

.,

it

is

possible to consider Z, as

made up

of a path

Z/

not intersect-

ing itself, together with a series of loops attached to Z/ at Each of these loops encloses a single the points clt c v critical point a K and, if described in a positive direction, adds
.

Each such loop described in a to the integral a term 27tiB K 2niBK It is evinegative direction adds a term of the form
.

dent that the form of each loop and the point at which it is attached to Z,' may be altered arbitrarily without altering the
value of the integral, provided no critical point be introduced In fact all the loops may be into or removed from the loop. as attached to L* at z regarded
.

It

eral path that can

can be proved by similar reasoning that the most genbe drawn from z% to z will be equivalent, so

far as the value of the integral

is concerned, to any given path preceded by a series of loops, each of which includes a sin-

gle critical point


tive direction.

and

is

The

value of the integral

described in either a positive or negais therefore of the

form

f
where

L f{z)dz
. .
.

+ zni{m B + m^ + ...),
x
x

m m^
lt

are

any integers positive or negative.


.

As an example
critical

consider the integral /

The onlv
is

point

is

= a. Any path

whatsoever from z to z

equivalent to a determinate path, for example, a rectilinear path, preceded by a loop containing a and described a certain

number

of times in a positive or negative direction. If w denote the integral for a selected path, the general value of the If now a straight line be drawn integral will be w -f- 2nni.

joining zQ to a, and if along its prolongation from a to infinity the .sr-plane be cut or divided, the integral in the #-plane thus

divided
stricted,

is

one-valued.

But, with the variation of z thus re-

any branch of the function log (z a) is one-valued. Select that branch, for example, which reduces to zero when z and its 2=za-\- I. It takes a determinate value for z

Art.

27.]

weierstrass's theorem.

28

derivative for every value of z

is

Hence, denoting

it

ct>

by Log
.

(z

a),

J j^T = Log iz-d)- Log a


For a path not
restricted in

(z

- a) =

Log ^-~^.
of the inte-

any way, the value

/
-ator is of

dz
z

a a

Log z
is

2nni

log

Prob. 20. If -S-v


tp(z)

a rational function of 2 of which the numer-

a lt

<z

.,

degree lower by 2 than the denominator, and if the zeros # of the denominator be of the first order, show that

ft8*=2-P&
fc. where
2<p{a v )/tp\a v )

log

= o.

(See Prob. 18, Art. 25.)

Art.

27.

Weierstrass's Theorem.
z,

Any

rational entire function of


.

points a lf a

.,

a my can be put
*,)"
,

in the

having form

its

zeros at the

A{z- a^(z where

...(#
.
.

a m y>,

., n m are positive integers. function which has no critical point in the generally, any finite portion of the ^-plane and has the points a lt ., a m as

is

a constant and

More

its zeros, is of the

form

where g{z)

is

holomorphic

in

every

finite region.

The
without

extension of this result to the case where a function


finite critical

points has an infinite


It is effected

<Jue to Weierstrass.

number of zeros is by means of the following


of isolated points

theorem

Theorem.

Given an

infinite

number

a lf

288

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

aa , ...,<?,..., a function can be constructed holomorphic except at infinity and equal to zero at each of the given points
only.*

For the given points can be taken so that

a n increasing indefinitely with


|

n.

Consider the
f

infinite

product

4w = $(i
where

fjt

<f,

Pn {z)

denotes the rational entire function

Any

factor

may

be written

in

the form

(i

- L)*
dz an
z

-*s(

-i)

+F

But since
g
(l

z \-

~"aj~

r ~*A
. .

* " " ~
~an

"

z*

r
*/>

z * dz

no}

a*(a n -zf

the path of integration being arbitrary except that it avoids as the points a lt a ., the product may be expressed
ZZ**

W
,

in

which

ij) n (z)

zn dz

a nn (a n
it

- z)'
will

In any given finite region of the ^-plane


to

be possible
indefinitely

assume that
n.

z\

< am

since

\a\ increases

with

Divide the product into two parts,

The second

part

is

equal to
tn
.

e * The following proof


Vol. II.
is

taken from Jordan, Cours d'Analyse, 2d edition,

Art.

27. j

weierstrass's theorem.
00

289

00

Consider the series 2ip n (z) and 2ip(z), each term of the secm tn ond being the derivative of the corresponding term of the first.
In the given region
;

p
CX

v*

*.'(*)

< ij Ki-i-py
,

00

Each term

of ^tp n (z)

is

accordingly less in absolute value than

the corresponding term of a convergent geometrical progres00

sion independent of

g.

The
oo

series 24>'(*)i therefore,

converges

uniformly.

The

series ^tp n (z) also converges, since

\U*)\= mod
where
/

fj'W*< W^Z\-p)>
jo,

denotes the length of the path of integration.

ink

a,

By Theorem

IV, of Article 23, the series 2ip(z) represents


tn

in the given region a

holomorphic function.
tn

The

exponential

also

must be holomorphic.
27 (1

The

other part of the product

- lW>
of factors
is

containing only a

finite

number

morphic, vanishing at all of the points


situated within the given finite region.

a lf But

everywhere holoav ., which are


.

this region
fulfils

may

be extended

arbitrarily.

The product
it

therefore

the re-

quired conditions. In the preceding demonstration

none

of the given points

ati a

was tacitly assumed that was situated at the origin.


necessary merely to mul-

To

introduce a zero at the origin tiply the result by a power of z.

it is

The most

general function without finite critical points

290

FUNCTIONS OF A COMPLEX VARIABLE.


.

[CHAP. VI.
.,

having its only zeros at the given points a lf a%} be expressed in the form

an

. .

.,

can

yw = <*un[i - -\ f[z) aJ
i

fj.*)

where g(z) is holomorphic except at infinity; for the ratio of any two functions satisfying the required conditions is neither
nor zero at any finite point. means of Weierstrass's theorem it is possible to express By any function, F(z), whose only finite critical points are poles as the ratio of two functions holomorphic except at infinity. For,
infinite

construct a function
.

tp(z)

The product F(z) tp(z) The given function can,

having the poles of F{z) as its zeros. no finite critical point. <p(z) will have

therefore, be written

Ft A

#*)

which

is

the required form.

it

In applying Weierstrass's theorem to particular examples, will rarely be found necessary to include in the polynomials
as

PJz) so many terms


given above.

were employed

in

the demonstration

It is quite sufficient, of course, to

choose these

the product converge polynomials in any way which will Factors of the for finite values of z to a holomorphic function.

make

form

aj
such a manner, are called " primary

where

Pn(z)

is

chosen

in

factors."

application of Weierstrass's Theorem take the resoThe zeros of sin z are o, lution of sin z into primary factors.

As an

7t,

27t,

.,

n7r,

....
I
\

Consider factors of the form


z

nit'
,

so that PJz) contains only one term

and

W) =

-I

nninn

z)

ART.

27.J
00

WEIERSTRASS'S THEOREM.

291

The

series

2$M'(g) m

will

converge uniformly in any region at

every point of which

=
\z\

p <

mn

for, since

nn{n7t

s)
\

\nn\l

mnl

each term term

is

less in

absolute value than the corresponding

of the series

war/
series J2tp n '{z).

A
series

similar result holds for the


m

-m

The two

$4,4*),
are also convergent; for
\tp n {z)\

21>4s)

cannot exceed the upper ex/,

treme of

\tpn(z)\

multiplied by

the length of the path of


z.

integration from the origin to the point

These

series

accordingly represent holomorphic functions in the region for

which

=
\z\

p.

Hence the expression required


sin g

is

= se*'>n( 1

-00 \

nnj

)e

the value n

= o being excluded
next article that
gd 1

from the product.

It will

be

shown

in the

e^

I.

Prob. 21. If
the doubly
is

and

gd 3

be two quantities not having a real

ratio,
rr

infinite series of
if

which the general term


2.

is 7

absolutely convergent

p>

Hence show
Z
-Itil

that the product

-=*4-3
where
go

= moo

-\-

nco 7

defines a holomorphic function in any finite

region of the s-plane. This function is Weierstrass's sigma function, and is the basis of his lystem of elliptic functions.

292

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

Art.

28.

Mittag-Leffler's Theorem.

Any one-valued function f{z) with isolated critical points *?,,#,,... can be represented in the neighborhood of one of these points by Laurent's series viz.
; :

+ M =A. +
is

A,(z,{g

O + A,(g -.) + ... )-'

+ Biz
1

a.)"

Hence
where 0(#)

/(*)

= <&*) + G.( ~in a region

),

holomorphic
J

containing the point a n

and

GH \
.

is

holomorphic over the whole plane excluding


is

the point a n

If

an
;

a pole oif{z),
it

Gn

l
J

contains a finite
infinite
series.
If

number

of terms

otherwise

is

an

the

number

of critical points
at

is finite,

and the function

Gn

each such point, by subtracting the sum of these is functions from/^) a remainder will be obtained which has no

formed

critical

point in the finite part of the plane.


series of

This remainder

ascending powers G{z) convergThe function f(z) can thereing for every finite value of z. fore be written in the form 1* >*]

can be expressed as a

analogous to the expression of a rational function by means of


partial fractions.

The

extension of this result to the case where the


is

number
Let a xi

of critical points

infinite is

due to Mittag-Lefrler.

of the one-valued func,,..., ani ... be the critical points tion /(#), and suppose that

kl
\an
\

<l*il

<

A*\ <
is

>

increasing without limit

when n

increased indefinitely.

Let, further,

GM (
.

be the series of negative powers of

.4
?

A
>ju

*^

<*>i

"

TAk

Art.

38.]

MITTAG-LEFFLER'S THEOREM,
in

293

a n contained

the expansion oif(z) according to Laurent's


.

Series in the neighborhood of an

The
,

function

GJ

\ having no

critical

point except at

a n may be developed by Maclaurin's

series in the

form h-^YO

series will converge uniformly within a circle described about the origin as a center with any determinate radius Within the same circle Maclaurin's series, applied fin < \an \.

and the

to

Gn

'\

\2

aj

],

the derivative with respect to z of

GJ
\2

a nj

A,

converges uniformly.

Hence,

for

any point within the

circle

Fn (z)
GJ
\2

representing the

first

terms of the development of

aj

by Maclaurin's theorem,

Fn \z)

its

derivative,

and

less in

R! remainders which by a suitable choice of v absolute value than any given quantity.

may be made
so that

Choose the positive quantities


the series

Elf E
.
.

>

. . .

EU9
.

. . .

+E

-\-

+ En +

is

convergent.

also in connection with each of the points

alf att

. ,

Choose aMI .,
.

an integer v such that

ft
\

z
\

<A <

|*h

anc*>

m
]

general,

1x10(1

t^-ty

" fjw <e*>

mod

[^t^y

" a,w <Em ]

'

294

FUNCTIONS OF A COMPLEX VARIABLE.


Consider now the series

[CHAP.

VI*

finite region of the plane, the points ax% a 3 . ant . . . excluded. Since \a M increases indefinitely with n, it is being possible, in any finite region of the #-plane, to assume that

in

any

|*|

<P<
m
sum.
be

am\*

Separate from each of these two series

its

first

terms.

These terms

will

have

in

each case a

finite

The remaining terms

of either series taken in order

absolute value than mt B m + h respectively, than each of the quantities p my pm + Ac\z\ w*>*Oiw~ ft* k*. a-Vma each of the series "/'"^s^^ cordingly,
will
less in less
.

being

lf

is

a ny

absolutely convergent for every value of z except a lt a 2 ,. r It is evident, further, that in any given finite region,
.

from which the points a lt a

an

are excluded, the

two
of

series

converge uniformly.
is

In such a region any term of

either series

holomorphic
first

Article 23, the

and, therefore, by Theorem IV of these series defines a holomorphic


;

function.

The

point a n

is

an ordinary point for the difference

&> since in

[>by its

'*#>]

=
[a*)

sfcdbeG
may be
.
. .

m
an
is
.

neighborhood

this difference

developed as

a convergent series containing only positive powers of z In the same way each of the points a lt a aHi ... ordinary point for the function
,

an

This function, therefore, can have no

critical

point except at

infinity, and must be expressible as a series G{z) containing only positive powers of z and converging uniformly in any
finite

region of the ,2-plane.

Hence

the function f(z)

may be

put

in the

form

Art.

28.]

mittag-leffler's theorem.

295

M=
in

G{z)

+2

[_G{izrj)

/)],
is

which

the character of each critical point

exhibited.
z.

As an

application of Mittag-Leffler's theorem consider cot

Its critical points are z

= o,
is

zr,

27r,

....
;

In the neighin the neigh-

borhood of

borhood
cot z
z

= o, of z =
is

cot z

holomorphic

and

tztt,

being any positive or negative integer,

rnt

holomorphic.

The
j

series

^
in

+ 00

nn'

which

for finite

an arbitrary positive integer, is not convergent values of z, even when \z\ < m. The series
is

.^r
is,

!+-] = -^

=^=
\

nnl

however, absolutely convergent at every point for which For the modulus of any term is equal to \z\ <m.

_
V

1*1

v(i-W). ^7T>
7Z7Z7

and, therefore, less than the corresponding term in the series


z\ *\
\

win)

A similar result holds

for the series

\_z -f- rnt

nnA

easy to see now that the reasoning employed in the demonstration of Mittag-Leffler's theorem may be applied to
It is

show

that the series

296

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

^
oo

\_z

nn

'

nnS

where the summation does not include n


tion holomorphic
O,
7t>

= o,

defines a func-

in

any

finite

region of the ^-plane, the points

2?r,

being excluded.

The

difference

cot

~ ,_i_^r_!_ + _LV^ J nn nn J z
x

\_z

can have no

point except at infinity. It must, therefore, be expressible as a series G(z) of positive powers of z>
critical

having an

infinite circle of
I

convergence.

Hence

cot *

=
is

(*)

+ - + j>
oo

p
1

h-

"1

The next
that,
if

step
is

to determine
its

(*).

It is to

be observed
zero, since

G(z)

a constant,
z.

value

must be

not a constant, differentiation z) ( of the preceding expression for cot z gives


cot
cot
If G{z) is
i

,,

sin z
It follows,

z
into z

TZ
-\-

{z

nny

by changing z

n, that

G\z +
Hence G\z)

n)

G'(z).

if
;

is periodic, having a period equal to n and as the z traces a line parallel to the axis of reals, G'{z) passes point again and again through the same range of values. But G\z),

being the derivative of G(z\


value of
*.

is

holomorphic for every


infinite,
if

finite

It can, therefore,

become

at

all,

only
)

when

the imaginary part of z is infinite. If z be written in the form x -f- ty, the value of G'{z) may be expressed as

^ '

_ ~

j_4r-

(x

-\- iy)*

2 ^y ( cosx

sm x V
lyi

(x-\-ty
first

nnf
and

\(cos 2x-\-i sin


last

2x)e

When
member
}-->-

jy

the

erms
it

of the

second

vanish.
f

In regard to the series


2
i

can be proved that,


^C*g -\
<f t

-^

iU

Art.
for

28.]

mittag-leffler's theorem.
I4v is

297

which y is finite and different from any given region an integer r can be found such that the sum of the moduli zero, of those terms for which ||> v is less in absolute value than

any previously assigned quantity e. As \y\ is increased the modulus of each of these terms is diminished. The modulus 00. of their sum, therefore, cannot exceed e when y = But the sum of any finite number of terms of the series when^= Hence the limit of the whole series is zero. G'(z), is zero. Hence, by Theorem III, therefore, never becomes infinite. It follows that Article 24, it is constant, and is equal to zero.
&(z)
is

equal to zero.
expression for cot z
is

The

accordingly

J+SfcJ nn
The
strass's

nnj

logarithmic derivative of the product expression for sin st given in the preceding article as an example of Weier-

theorem,
cot *

is

= #'(*) + ?-+'*[ ^- \-2


v

00

nn h nn J

T
Making z

Hence g{z)
its

in that

expression

is

a constant.

= o,

value

is

seen to be unity.

Prob. 22.

From

the expression for cot z deduce the equation

+
cosec z
8

ec

= 3> *Z

where the summation does not


Prob. 23.

- nnf exclude n = o.
7

;*>

(z

Show

that the doubly infinite series

jb,

vj^

where

00

mao -f

points are z

noo^

defines a function
is

whose only

finite critical

go.

This function

Weierstrass's ^-function.

(Com-

pare Problem 21.)


Prob. 24. Prove that

JM =~
J

^
s

log

<r(*).

*.M,

^ %u

[i$l-M} -i^CP

298

FUNCTIONS OF A COMPLEX VARIABLE.


Prob. 25. Prove that

{CHAP.

VL

'(z)

2^7

r- , 3

where the summa-

tion does not exclude go

o.

Art.

29.

Critical Lines and Regions.

The

functions whose properties have been considered in the

preceding articles have been assumed to have only isolated That an infinite number of critical points may critical points.

be grouped together in the neighborhood of a single finite point is evident, however, from the consideration of such ex-

amples as

^H/^y) "fr* w = cot ,

#7/ y -/

f'-Vu^
osec

/'

^r3

=^

jr^.-

In the former an infinite

number

neighborhood of the

origin.

of poles are grouped in the In the latter an infinite num-

ber of essential singularities are situated in the vicinity of the a. point z

easy to illustrate by an example the occurrence of lines and regions of discontinuity. Take the series *
It is

The sum

of

its first

n terms
* z'

is
1
.1

>

which converges to unity

if

\z\<

I,

and to zero

if

\z\>

I.

Hence the
series.

circle

|*|=I

is

a line of discontinuity for this

Consider

now any two

regions

5 and S
t

the former situated

within, the latter without, the unit circle.

Let

<p(z)

and

ip(z)

be two arbitrary functions both completely defined in these


regions.

The

expression
<P(z)V(z)

+ tp(z)[i-0(z)-]
See Weierstrass, Abhandlungen aus der

* This series

is

due to

J.

Tannery.

Functionenlehre (1886),

p. 102.

Art.
will

29.]

critical lines and regions.


<p(z) in

29D
In regions comthe same

be equal to
expression

S and
1

ip(z)

in

pletely separated from one another by a


literal

critical line,

may

thus represent entirely independent

functions.

For a single continuous region, however, in the interior of which exist only isolated critical points, the character of the
function in one part determines its character in every other Let 5 be such a region, and assume that its boundary is a part.
critical line.

a critical
series, viz.

In the neighborhood of any interior point a, not point, the given function is expressible as a power
:

f{z)

= /(a) -f (* - a)/' (a) + ... + JfLZi2L/W(,) +

will converge uniformly over a circle described a as a center with any determinate radius less than the about
It serves for the distance from a to the nearest critical point. all its successive derivatives at any point

This series

calculation of f(z) and

b interior to this circle.

From

the preceding power series, ac-

cordingly, can be obtained another

M=
center.

Ab)

+ t'~ *)/'(*) + ... + }'~ 2


I
. .

6y,

ft

J^m +

representing the f(z) within a circle described about b as a In general, the point b can be so chosen that a portion of this new circle will lie without the circle of convergence of

the former power series. At any new point c within the circle whose center is b, the value of the function and all its successive derivatives can be calculated
series
;

and

so, as before,

power

can be obtained convergent in a circle described about c as a center and, in general, including points not contained in
either of the preceding circles.

continuing in this manner it will be possible, starting from a given point a with the expression of f(z) in ascending powers, to obtain an expression of

By

the same character at any other point k which can be connected with a by a. continuous line everywhere at a finite distance

from the nearest

critical point.

It follows

that the character of

300

FUNCTIONS OF A COMPLEX VARIABLE.

[CHAP. VI.

the function everywhere within

S can be

determined completely

from

its

hood

expression in ascending power series in the neighborof a single interior point.

It will

be impossible by the process just explained to derive


in

any information
S.

regard to the function at points exterior to given above, furthermore, shows that a complete definition of f(z) within 5 may carry with it the definition of an entirely independent function without 5.

The example

As an example

of a function

having a
series
-f-

critical region con.

sider the function defined


I

by the

.,

+ 22 + 2Z* + 2Z*
;

X. ?i

+%^
is

01+-'

which represents a function without critical points in the For points on or without this interior of the circle \z\ = I.
circle the series
is

divergent

and, further,

it

impossible to

obtain from

it

an expression converging when

=
\z\

i.

The

function thus defined, consequently, exists only in the region interior to the unit circle. By changing z into \/z a series

is obtained, representing a function which has no existence in Functions in connection with the interior of the unity circle. which such regions arise are called " lacunary functions." *

Art.

30.

Functions Having n Values.

Let the function


gion

a value
at z
,

beginning remain in the

take at the point z of a given re/ Suppose that along any continuous path, and subject only to the conditions that it shall
0)
.

w =/(z)

interior of

S and
.

shall

isolated points

t ,

a%t

w is

not pass through certain continuous and has a contin-

uous derivative.
ferent from
*
{Q

impossible, when z traces such a path, difto return to the point z so as to obtain there a value of
If
it is

w \ w is

one-valued in the region S.

On

the other

Morley, Theory of Functions (1893),

PoincarS, American Journal of Mathematics, Vol. p. 119

XIV; Harkness and

Art.

30.]

FUNCTIONS HAVING n VALUES.

301

hand, certain paths

may

lead back to z with

new values
x ,
, .

of

w.

w has Suppose that at each point of 5, except a <z a n different values, and that starting from such a point 89 and a i9 f tracing any continuous curve not passing through a
. .

the several values of

each of which
the

is

give rise to n branches % x characterized by a continuous derivative.


,

w w

wnT
In

a k any one of the points a x , a% , . . . neighborhood these branches are said to be distinct or not, according a? small closed curves described about this point lead from each value of
of

back to the same value again, or cause some of the branches In the latter case the point is a branch to interchange values.
point.

About any branch point a k


cle
;

as a center describe a small cir-

and suppose

that, starting

from any point of

it

with the

value

corresponding to a certain branch, the values ... are obtained by successive revolutions about ak , Wp^Wy the original value being reproduced after p revolutions. Introduce

wa

now

new independent
z'

variable
k )'K

z'

such that

= (z - a

It

can be shown that when z makes one revolution about


f

ak

z'

gin

makes only one /th part of a revolution about the oriof the ^-plane, and that to a complete revolution of z
.

about the origin of the ^-plane correspond / revolutions of z about a k Considering then the branch wa as a function of z' r
the origin cannot be a branch point, for whenever z' describes is reproduced. a small circle about it, the value The a branch 7va must accordingly be expressible by Laurent's

series in the

form
+ 00
hn
y

or,

substituting for z'

its

value,
1

2
.
.

Wa = A + A,(z - a + A,(z - atf + + A_ (z - a y*+A_ (z - a j"*+


k)
.
1

This expression makes plain the relation between the different

302

FUNCTIONS OF A COMPLEX VARIABLE.


in the

[CHAP. VI.

branches of a function

neighborhood of a branch point.

When
of
its

the development of a branch in the neighborhood of one branch points gives rise to only a finite number of terms
is

containing negative powers, the branch point

called a

"

polar

branch point." Consider the functions

P =z Wy + W + + Wu P = w w + w,w + + wn _ wn
x

. .

Pm =
Each
of these functions

WSiVa
is
.

w.
in

unchanged
.

value

when several or all


is

of the quantities n are interchanged, and %i xt sequently a one-valued function of z within S. Hence
.
,

w w

con-

w must

satisfy an equation of the

th degree,
%

+ Pjf-* + p

of-'

+P =
n

o,

the coefficients of which are one-valued functions of z having only isolated critical points within 5. When the entire ^-plane can be taken as the region 5, and those branch at which
points polar branch points, the only other critical points being poles for one or more branches, the functions u H tt ..,,P are rational functions of z. In this case is an algebraic function of z.

the branches do not

all

remain

finite are

P P

Art.

1.],

EQUATIONS OF FIRST ORDER AND DEGREE.

303

Chapter VII,

DIFFERENTIAL EQUATIONS.
By W. Woolsey Johnson,
Professor of Mathematics in the U. S. Naval Academy.

Art.

1.

Equations of First Order and Degree.

In the Integral Calculus, supposing y to denote an unknown function of the independent variable x, the derivative of y with
respect to

is

required to find the value of y as a function of x. words, given an equation of the form

given in the form of a function of x, and it is In other

%=A*\
of which the general solution

dy

= Ax)dx,

(i)

is

written in the form


(2)

= /A*)**,
such reduction

it is the object of the Integral Calculus to reduce the expression in the second member of equation (2) to the form of a

known

function of x.

When

is

not possible,

the equation serves to define a new function of x.


In the extension of the processes of integration of which

the following pages give a sketch the given expression for the derivative may involve not only x, but the unknown function

or, to
it

write the equation in a form analogous to equation

(1),

may be

Mdx

+ Ndy =

o,

(3)

in which

M and N
;

are functions of

x and y.

This equation

is

in fact the general

form of the

differential equation of the first

order and degree either variable being taken as the independent variable, it gives the first derivative of the other variable

304
in

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

terms of

x and y.

So

also the solution

is

not necessarily an
is.

expression of either variable as a function of the other, but


generally a relation between x and implicit function of the other.

y which makes either

an.

When we recognize the left member of equation (3) as an " exact differential," that is, the differential of some function o
x and y
tion
s

the solution

is

obvious. For example, given the equa-

xdy -\-ydx
the solution

= o,
is is

(4)
(5)

xy =

C,

where
is

C is

an arbitrary constant,

obtained by

" direct inte-

gration."

attributed to C, the result " x' 1 is a particular integral thus jj/ a " particular integral of equation (4), while the more general relation expressed by" complete integral." equation (5) is known as the
;

When

a particular value

In general, the given expression


act differential,

Mdx -\- Ndy

is

not an exless

and

it

is

necessary to find

some

direct

method
of the

of solution.

of solving a differential equation, order and degree is, when practicable, to " separate the variables," so that the coefficient of dx shall contain x
first

The most obvious method

only,

and that of

dy,
(1

only.

For example, given the equation


(6)
(1 -\-

- y)dx + (1 + x)dy = o,
x)(i

the variables are separated by dividing by

y).

T Thus
,

dx
:

'

h
1

dy =l_

o,

-\-x

y
and hence
y)
log
(1

Each term

is

now

directly integrable,
(1

log

+x)

c.

solution here presents itself in a transcendental form,, but it is readily reduced to an algebraic form. For, taking the exponential of each member, we find
'

The

=e =

C,

whence

1 -L-

= C(i

y),

(7>

where

C is

put for the constant ^.

ART.

2.]

GEOMETRICAL REPRESENTATION.
verify the result in this form

305

To
tion

we

notice that differentia-

gives dx =

Cdy, and substituting in equation (6)

we

find

which
is

C{i

-y)+i+x = o,
(7).

true
1.

by equation

Prob.

Solve the equation dy -\-y tan

x dx

= o.
b

Prob.

2.

Solve

$dx

+ py = a\ J
= ^}-. x +1
~ C- r dr

y=C cos x.) Ans> l+A _ ^ax\


(Ans.
by
( Ans. \

Prob.
Prob.

3.

Solve

fdx

y y

=
1

ex

^)
j

4. Helmholtz's equation for the strength of an electric current C at the time / is

L are given constants. Find the value of C, dethe constant of integration by the condition that its initial termining value shall be zero.
where
,

R, and

Art.

2.

Geometrical Representation.
of a differential equation

The meaning

illustrated by supposing simultaneous the rectangular coordinates of a variable point. ient to put/ for the value of the ratio dy dx.
:

may be graphically values of x and y to be


It is

conven-

Then

P being
its

the moving point (x, y) and <p denoting the inclination of path to the axis of x, we have

p *
The
between p, x and
y

= dy = tan -fdx

0.

given differential equation of the first order is a relation y, and, being of the first degree with respect determines in general a single value of p for any assumed to/, values of x and y. Suppose in the first place that, in addition
to the differential equation, we were given one pair of simultaneous values of x and y that is, one position of the point P.
y

Now

let

P start

from

this fixed initial point

in either direction along the straight line

and begin to move whose inclination

306
is

DIFFERENTIAL EQUATIONS.

[Chap. VII.
initial

determined by the value of p corresponding to the

thus have a moving point satisfying values of x andjj/. moves the the given differential equation. As the point values of x and y vary, and we must suppose the direction of

We

its

motion to vary

in

such a way that the simultaneous values

In of x, y, and p continue to satisfy the differential equation. is said to that case, the path of the moving point satisfy the
differential equation.

The point may return to its initial thus describing a closed curve, or it may pass to infinposition, ity in each direction from the initial point describing an infinite
branch of a curve.*

The ordinary

cartesian equation of the

path
If

of

Pis

a particular integral of the differential equation.

no pair of associated values of x and y be known, Pmay be assumed to start from any initial point, so that there is an
unlimited

number

of curves representing particular integrals

of the equation.

These form a "system


is

of curves,"
"

and the

complete integral form of a relation between


This parameter
is

the equation of the system in the usual


x, y>

and an arbitrary

parameter."
It is

of course the constant of integration.

constant for any one curve of the system, and different values of it determine different members of the system of curves, or
different particular integrals.

As an

illustration, let us

may

take equation (4) of Art. be written

1,

which

dy ~

y
x'

dx
the axis of

Denoting by 6 the inclination to

of the line joining

P
is

with the origin, the equation


equivalent to tan

tan

0,

and

therefore expresses that moves in a direction inclined equally with

OP to either axis, but


*

on the other
is

When
why

the form of the functions

M and AT

is

unrestricted, there

no

reason

either of these cases should exist, but they

commonly occur among

such differential equations as admit of solution.

ART.

3.]

PRIMITIVE OF A DIFFERENTIAL EQUATION.

307

side. Starting from any position in the plane, the point thus moving must describe a branch of an hyperbola having the two axes as its asymptotes accordingly, the complete
;

integral

xy

=C

is

the equation of the system consisting of

these hyperbolas.
Prob.
5.

in a direction always perpendicular to OP, equation of the system of curves described.

to move Write the differential equation which requires and thence derive the

Prob.

6.
?

What
If

is

the system described

when
(Ans.

is

ment

of 6

Prob.

7.

x*/=Q
(x*

the comple-

26,

show geometrically that the system described


find the differential equation.

consists of circles,

and

(Ans. 2xydx

y*)dy.)

Art.

3.

Primitive of a Differential Equation.

Let us now suppose an ordinary relation between x and y, which may be represented by a curve, to be given. By differentiation we may obtain an equation of which the given equation
is

of course a solution or particular integral.


this with the given equation

But by

combining
be found.

ential equations of

any number of differwhich the given equation is a solution may For example, from

we
of

y = m(x
obtain directly

a)
t

(1)

2ydy

= mdx
dx

(2)
(2)

which equation we have

(1)

is

an integral; again, dividing

by

(1)

2dy

=
is

y
and
of this equation also (1)
If in
it is

JZT^

(3)

an integral.

be regarded as an arbitrary parameter, equation (1) the equation of a system of parabolas having a common

axis and vertex.

The

differential equation (3),

which does not

contain m,

is

satisfied

by every member

of this system of curves.

308

DIFFERENTIAL EQUATIONS.
(i)

[CHAP.

VIL

Hence equation

equation (3), the variables are already separated.

the complete integral of as will be found by solving the equation in which

thus regarded

is

Now
it

equation
of

(3) is

obviously the only differential equation


(1)

independent
is

which could be derived from

and

(2),

since

" differIt is therefore the the result of eliminating m. " and in this point of view the ential equation of the system
;

integral equation (1)

is

said to be
(1)

its

"primitive."

Again,
stant,
it is

if

in

equation

a be regarded as the arbitrary con-

a
is

common

the equation of a system of equal parabolas having Now equation (2) which does not contain a axis.
it

is

satisfied by every member of this system of curves; hence the differential equation of the system, and its primitive equation (1) with a regarded as the arbitrary constant.

is

Thus, a primitive is an equation containing as well as x and an arbitrary constant, which we may denote by C, and the
is

corresponding differential equation

a relation

between

x,

yr

and p, which is found by differentiation, and elimination of C if necessary. This is therefore also a method of verifying the comFor example, in plete integral of a given differential equation.
verifying the complete integral (7) in Art. I we obtain by differIf we use this to eliminate C from equaentiation 1 Cp. is equation (6); whereas the process before tion (7) the result

employed was equivalent to eliminating p from equation


thereby reproducing equation
Prob.
Art.
2, 8.

(6),

(7).
7,

and

Write the equation of the system of circles in Prob. derive the differential equation from it as a primitive.

Prob. 9. Write the equation of the system of circles passing b), and derive from it the differthrough the points (o, b) and (o,
ential equation of the system.

Art.
In Art.
I

4.

Exact Differential Equations.


is

the case

mentioned
is,

in

which

Mdx

-f-

Ndy

is

an

" exact differential," that

the differential of a function of

andjy.

Let u denote this function; then

du

Mdx -\-.Ndy,

(1)

ART.

4.]

EXACT DIFFERENTIAL EQUATIONS.


the notation of partial derivatives

309

and

in

d M= d* A

N^P. dy
dydx
dxdy
y
J

Then, since by a theorem of

partial derivatives

'

dy

d*
fulfilled

This condition must therefore be

by

M and N

in

order that equation (i) may be possible. When it is fulfilled Mdx -\- Ndy = o is said to be an " exact differential equation,"

and

its

complete integral

is

= C.
/)dy

(3)

For example, given the equation x(x


-\-

2y)dx

+ (**
y\

= o,
and

x{x

-\-

2y),

N = x*

2#,

= 2x
To

the

condition
function

(2) is fulfilled,

and the equation

is

exact.

find the

u,

we may integrate Mdx,

treating^/ as a constant; thus,

in

which the constant of integration


result of differentiating this
is

Y may

be a function of y.

The

x*dx

+ zxy dx +
written

-|-

x*dy

dY,
;

which should be identical with the given equation therefore, dY y^dy, whence Y = \y % C, and substituting, the complete integral

may be
is

The

result

more

readily obtained

if

we

notice that

all

terms containing x and dx only, or y and dy only, are exact hence it is only necessary to examine the terms differentials both x and y. In the present case, these are containing
;

2xy dx -f- x*dy, which obviously form the differential of x*y whence, integrating and multiplying by 3, we obtain the result
;

above.

The complete

integral of

any equation,

in

whatever way

it

310

DIFFERENTIAL EQUATIONS.
in

[CHAP. VII.

was found, can be put

the form u

C,

by solving

for C.

can be obtained, Hence an exact differential equation du which must be equivalent to the given equation

=o

Mdx + Ndy =
here supposed not to be exact. must therefore be of the form

o,

(4)

The

exact equation du

=o
(5)

ju(Mdx

+ Ndy) =

o,

where and y.

}x is

Such a

a factor containing at least one of the variables x " factor is called an integrating factor" of the

given equation.

For example, the


I,

result of differentiating

equation

(7),

Art.

when put
(1

in the

form u

C, is

(1

-y)dx-\-{\

+ x)dy _
3
'

-J)

so that
is

is an integrating factor of equation (6). It to be noticed that the factor by which we separated the

(1

y)~

variables, namely,
factor.

(1

y)~\i

x)~\

is

also

an integrating

an integrating factor can be discovered, the given differential equation can at once be solved.* Such a factor is sometimes suggested by the form of the equation.
It follows

that

if

Thus, given
the terms ydx
act,

(y

x)dy-\-ydx

= o,

xdy, which contain both x and y, are not exbut become so when divided by either x* or y*; and be-

cause the remaining term contains y only, y~* is an integrating The resulting integral is factor of the whole expression.

^ogy
Prob. 10.
x~*
is

+^ =
x

C.

Show from the integral equation in Prob. 9, Art. an integrating factor of the differential equation.
1

3,

that

Prob. ii. Solve the equation x(x

+ $y*)dx -\-yiy + 3^)dy = (Ans. x* + 6xY +/ =


c.)

o.

* Since

we have
fl

juM and #iVin the exact equation (5) must satisfy the condition (2) r a partial differential equation for ju; but as a general method of finding

this

simply comes back to the solution of the original equation.

Art.

5.]

homogeneous equation.
5.
1

311

Prob. 12. Solve the equation ydy-\-xdx-\

(Am.
Prob. 13. If u

^t^ + tan"
2

^*.) x

is a form of the complete integral and /* the corresponding integrating factor, show that m/( u ) is tne general

=c

expression for the integrating factors.


Prob. 14. Show that the expression x a yP(mydx nxdy) has the km ~ ~ a kn ~ ~P; and by means of such a factor integrating factor x y
1
1

solve the equation

y(y

-f-

2x )dx

+ *(**

2y)</y

o.

(Ans. 2x*y

y*

Prob. 15. Solve (x*

-\-

y'*)dx

2xydy = o.

(Ans. x*

= ^.) = ex.) y*

Art.

5.

Homogeneous Equation.

The

differential equation

and homogeneous when x and j of the same degree

N are
or,

Mdx + iVi/y =

is

said to

be

homogeneous what is the same

functions of
thing,

when

dy t-

is

expressible as a function of -.

If in

such an equation
v,

the variables are changed from

x and y
and

to

x and

where
vdx,

=y

whence

= xv

dy

= xdv

-\-

the variables

x and v
(x

will

be separable.
-|-

For example, the

equation
2y)dx

ydy

=o
indicated

homogeneous dividing by x,
is

making the substitutions


2v)dx
-f-

and

(1

v(xdv
7

-f-

vdx)
o.

= o,

whence

dx h x
s

vdv
-~

{v

\f
1)

Integrating,

log

x+

log (v

^
C.

= C;

and

restoring^/,

& Ky log (y

x) '
Afa^/

The equation Mdx

-f-

= o can

always be solved when

312

DIFFERENTIAL EQUATIONS.
the
first

[CHAP. VII.
is,

M and N are functions of


of the form

degree, that

when

it

is

(ax
For, assuming

+ by + c)dx
x

-\- (a!

x
r

+ b'y + c')dy =
-\- k,
it

o.

=x

-\-

h, y = y

becomes
f

(ax'+ b'y'+ ah

+ bk+c)dx'+{a'x'+ b'y'+a h+b' k+c')dy' =o


k,

which, by properly determining h and


{ax'

becomes

+ by')dx' + (a'x' +
b

b'y')dy',

homogeneous equation. This method fails when a:

a'

b',

that

is,

when the

equation takes the form

{ax

-{-

by

)*&"

+ [w(^^ + ^) +

</]d^/

=o

but in this case if we put z = ax -f- ^, and eliminate be found that the variables x and z can be separated.

jj>,

it

will

Prob. 16. Show that a homogeneous differential equation represents a system of similar and similarly situated curves, the origin being the center of similitude, and hence that the complete integral may be written in a form homogeneous in x, y, and c.
Prob. 17. Solve

xdy

y dx
>jx -f-

^(x* -f y*)dx

o.

(Ans. x*

=c
2

2cy.)

Prob. 18. Solve ($y

i)dx -j- (jy * (Ans. (7

Prob. 19. Solve (x Prob. 20. Solve (1


the

-\-

y*)dx

- + 2xy dy = o.

3^ -f- 3)^ =; o. * - i) 5 i)\y

+ xy)ydx + (1

(Ans. .r o by introducing xy)xdy

= = ex.) y*
f.)

new

variable 2

= xy.
-j-=ax + by+c.
6.

(Ans.

a;

Cye*y.)
bx
.)

Prob. 21. Solve

(Ans. abx+&y-\-a-\-bc=Ce

Art.

The Linear Equation.


is

when (one of the variables, say x, being regarded as independent,) it is of the first degree with respect to y, and its derivatives. The
linear equation of the first order

A differential equation

" said to be " linear

may

therefore be written in

the form

Art.

6.]

THE LINEAR EQUATION.

313

where

and
is

are functions of

only.

Since the second

a function of x, an integrating factor of the first be an integrating factor of the equation provided To find such a factor, we solve the equation it contains x only.
will

member member

+/>which
is

done by separating the variables

dy
;

thus,

Pdx

whence log^

=c

Pdx

or

y
Putting equation exact equation is
Pd

Ce-f**.

(3)

this

in the

form u

c,

the corresponding

e* *(dy

+ Pydx) =
Pdx

o,

whence r

is the integrating factor required. the general solution of equation (i) is factor,

Using

this

ef

pdx

=f/

Qdx

+ C.

(4)

In a given example the integrating factor should of course

be

simplified in form
(i

if

possible.

Thus

+ x*)dy = {m-\- xy)dx


y
;

Is

a linear equation for

reduced to the form

(i), it is

dy

dx
from which

i+x^
n x dx

\y

m
i+x*'
i
,

/
The

integrating factor

is,

therefore,

efPdx

=
is

whence the exact equation


dy
v(i

xy dx
(i

mdx
'

+ **)

+ o'~(i+*')

314
Integrating, there

DIFFERENTIAL EQUATIONS.
is

[Chap. VII.

found

y
or

*
,

= mx + CV(l + x*).
linear,

An
for

equation

is

sometimes obviously
dy
-y- -f-

not forj, but

some function

of y.

For example, the equation


tan

= x sec ^
y
;

when

multiplied by cos
is e*,

takes a form linear for sin

the

integrating factor

and the complete integral

sin

=x
dy
-j- -f-

-\-

ce

~x
.

In particular, the equation

Py

n Qy which
,

is

known

as

" the extension of the linear equation,"


linear for

is

readily put in a

form

I_n
.

Prob. 22. Solve

x a dy +

(i

2x)y

= x\
sin

(Ans.

= x*(i + ***).)

Prob. 23. Solve cos

x~-

-\-y

1 -|-

x
^

= o.

(Ans. jy(sec

+ tan x) = x +
= sin x = x* +
2
-\-

.)

Prob. 24. Solve

cos
</#

-f-

y sin

1.

(Ans.

c cos #.)
1

Prob. 25. Solve


Prob. 26. Solve

= x*y*
dx

xy.

(Ans.

-5

+^*.)

i ^ = xy + y

,-,.

(Ans.

-=
x

- * + ce~h>\\ v '
f

Art.
If

7.

First

Order and Second Degree.


first

the given differential equation of the

order, or re-

a quadratic for p, the first step x, y, in the solution is usually to solve for p. The resulting value of / will generally involve an irrational function of x and y;
lation
is

between

and p,

so that an equation expressing such a value of p, like

some

of

those solved

in

the preceding pages,

is

not properly to be re-

Art.

7.]

FIRST

ORDER and second degree.


first

315

garded as an equation of the


case

when the expression whose

In the exceptional degree. root is to be extracted is a

perfect square, the equation is decomposable into two equations properly of the first degree. For example, the equation

when

solved

= 2A*'+f) x y for p gives 2p = -, or 2p = -;


*y(i+4/)
x

it

may

therefore

be written

in the

form
(

2px

- y)(2py - x) = o,
either

and

is

satisfied

by putting
dy

y
2x
l

dy

x
2y

dx

dx
x*

The

integrals of these equations are

y*

= ex

and

2y

C,

and these form two entirely


equation.

distinct solutions of

the given

As an

illustration of the general case, let us take the

equation
(I)

*/'=/,

or

%=
'

Separating the variables and integrating,

\/xVy=V~c,
and
this equation rationalized

(2)

becomes

{*

- yf -

2c(x

+ y) + =
c*

o.

(3)

thus a single complete integral containing one arbiconstant and representing a single system of curves; trary namely, in this case, a system of parabolas touching each axis
is

There

at the

same distance

from the

origin.

The

separate equa-

tions given in the form (2) are merely branches of the

same

parabola.

Recurring
tion of the

now

ential equation, as given in Art. 2,


first

to the geometrical interpretation of a differit was stated that an equa-

degree determines, in general, for assumed values of y, that is, at a selected point in the plane, a The equation was, of course, then supposed single value of p.

x and

316
rational in

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

x and y* The only exceptions occur at points for which the value of p takes the indeterminate form that is, the equation being Mdx -\- Ndy = o, at points (if any exist) = o and = o. It follows that, except at such for which no two curves of the system representing the complete points,
;

integral intersect, while through such points an unlimited

num-

ber of the curves

may

pass,

forming a

"

pencil of curves." f

the other hand, in the case of an equation of the second degree, there will in general be two values of p for any given
point.
{4,
1), p

On

Thus from equation

J-;

(i) above we find for the point there are therefore two directions in which a

point starting from the position (4, 1) may move while satisThe curves thus described fying the differential equation. If the same values represent two of the particular integrals.
of x and
sult
is

be substituted
c,

a quadratic for

complete integral (3), the reI, and these giving c = 9 and c

in the

determine the two particular integral curves,

Vx -j- Vy

3,

and Vx

Vy =

1.

In like manner the general equation which may be written in the form

of the second degree,

Lp
where L, M, and
point for

+ Mp + N =
two

o,

N are one-valued functions of x and y, repreintersect in


real values.
is

sents a system of curves of which

which

found to have two

any given For these

points, therefore, the complete integral should generally give

two

real values of
its

c.

Accordingly we

may

assume, as the

standard form of

-equation,

P<*
* In
fact
x

+ Qc + R =

o,

p was supposed
not in this

P -=s\vr
degree.

x would

a one- valued function of x and y\ thus, connection be regarded as an equation of the first
to be

f In Prob. 6, Art. 3, the integral equation represents the pencil of circles pass-

ing through the points (o, b) and (o, tion is indeterminate at these points.

b)\

accordingly/

in the differential

equais

In

merely a node of one particular integral. J> is indeterminate at the origin, and this point o, which passes through it. integral, xy

some cases, however, such a point Thus in the illustration given in Art.
is

2,

a node of the only particular

ART.

8.]

SINGULAR SOLUTIONS.

317
y. the differential

where P> Q, and


If

are also one-valued functions of

x and

there are points which

equation, they will also

make p imaginary make c imaginary in

in

the integral.

Prob. 27. Solve the equation p* gral to the standard form.


(Ans. (y
2

+y =

and reduce the inte-

-j-

cos x)c*

2c sin

-f-

cos

o.)

Prob. 28. Solve yp o, and show that the intersect2xp y ing curves at any given point cut at right angles. Prob. 29. Solve (x*

+
8.

i)p*

i.

(Ans. <rV

%cx0

i.)

Art.

Singular Solutions.
sometimes
is

A differential
admits of what
is

equation not of the first degree " " called a singular solution that
;

to say,

solution which is not included in the complete integral. For that the system of curves representing the complete suppose

Every point A of this envelope with a particular curve of the complete inis a point of contact therefore a point moving in the envelope when tegral system
integral has an envelope.
;

has the same values of x, y, and / as if it were moving through A in the particular integral curve. Hence such a point satisfies the differential equation and will continue to satisfy it as long as it moves in the envelope. The equation
passing through
of the envelope
is

therefore a solution of the equation.

As an

illustration, let us take the


is

system of straight lines

whose equation

y
where

= cx +

",

(l>
differential equation-

c is the arbitrary parameter. derived from this primitive is

The

*=/*+-, P
of

(2>

which therefore

(i) is

the complete integral.


(i),

the lines represented by equation values of c% are the tangents to the parabola

Now

for different

/ = 4**.

(S)

318

DIFFERENTIAL EQUATIONS.

[Chap. VII.

A point

moving

in this

parabola has the same value of/ as if it were moving in one of the tangents,
(3) will

and accordingly equation be found to satisfy the


be noticed that for

differential equation (2).


It

will

any point on the convex side of the parabola there are two real
values of p for a point on the other side the values of p are
;

imaginary, and for a point on the curve they are equal. Thus
its

equation

(3)

expresses the

y which must exist in order that (2) as a quadratic for p may have equal roots, as will be regarded seen on solving that equation.
relation

between x and

In general, writing the differential equation in the form

Lf + Mp + N = o,
the condition of equal roots
1

(4)

is

M -\LN =

o.

(5)

which is the " discrimi" of equation (4), frequently admits of separation into nant factors rational in x and y. Hence, if there be a singular solution, its equation will be found by putting the discriminant of

The

first

member

of this equation,

the differential equation, or one of


It

its factors,

equal to zero.

does not follow that every such equation represents a soluIt can only be inferred that tion of the differential equation.
a locus of points for which the two values of / become Now suppose that two distinct particular integral equal. curves touch each other. At the point of contact, the two
it

is

values of/, usually distinct, become equal. The locus of such " Its equation plainly satisfies the points is called a tac-locus." but does not satisfy the differential equation. An discriminant,
illustration
is

afforded by the equation

ART.

8.]

SINGULAR SOLUTIONS.
is

319

of which the complete integral


discriminant, see equation
(5), is

y*
t

-f-

(x

cf
)

= a*,

and the

y*(y
a,

s= o.

= o, the first two of which satisfy the differential equation, while y = o does not.
This
is

satisfied

by y = a, y =

and y

The complete integral represents


a with center on the axis of
x.

in this case all circles of radius

Two of

these circles touch at

every point of the axis of x, which is thus a tac-locus, while a constitute the envelope. y a and y

The

discriminant

is

the quantity which appears under the

radical sign

therefore

values of

when the general equation (4) is solved for/, and it changes sign as we cross the But the envelope. / remain real as we cross the tac-locus, so that the
1

discriminant cannot change sign. Accordingly the factor which indicates a tac-locus appears with an even exponent (as y in

the example above), whereas the factor indicating the singular solution appears as a simple factor, or with an odd exponent.

A simple

factor of the discriminant, or one with an

odd ex-

ponent, gives in fact always the boundary between a region of the plane in which/ is real and one in which/ is imaginary For the two nevertheless it may not give a singular solution.
;

arcs of particular integral curves which intersect in a point on the real side of the boundary may, as the point is brought up

to the boundary, become tangent to each other, but not to the In that case, since they cannot cross the boundary curve.

boundary, they become branches of the same particular


gral forming a cusp. " called a " cusp-locus
is of

inteis

boundary curve of this character the value of p for a point moving in

it

course different from the equal values of/ at the cusp, and therefore its equation does not satisfy the differential equation.*
Prob. 30.

To what
?

curve

is

the line

= mx

-f*

|/(i

w
3

always tangent
Prob. 31.
* Since there

(Ans. jp

x*

=a

.)

Show
is

that the discriminant of a decomposable differ-

no reason why the values of/ referred

we

o has not in general a singular conclude that the equation Lp* -f- Mp -fsolution, its discriminant representing a cusp-locus except when a certain condition
is fulfilled.

N=

to should be identical,

320

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

it

ential equation cannot be negative. Interpret the result of equating to zero in the illustrative example at the beginning of Art. 7.

Prob. 32. Show that the singular solutions of a homogeneous differential equation represent straight lines passing through the origin..

Prob. 33. Solve the equation xp*


(Ans.

x2

2cy-\- ac*

=o

2yp
;

+ ax

o.

singular solution y*

= ax"*,}

Prob. 34.

Show

that the equation/*

+ 2Xp y=o

has no sin-

gular solution, but has a cusp-locus, and that the tangent at every cusp passes through the origin.

Art.

9.

Singular Solution from Complete Integral.

the complete integral of a differential equation of the second degree has been found in the standard form

When

PS+Qc + R = o
R

(!)

(see the end of Art. 7), the substitution of special values of x and y in the functions P, Q, and gives a quadratic for c whose roots determine the two particular curves of the system which

pass through a given point. If there is a singular solution,, that is, if the system of curves has an envelope, the two curves which usually intersect become identical when the given

moved up to the envelope. Every point on the entherefore satisfies the condition of equal roots for equavelope
point
is

tion (1),

which

is

ff- 4 P =

o;

(2>

and, reasoning exactly as in Art. 8, we infer that the equation of the singular solution will be found by equating to zero the
discriminant of the equation in c or one of its factors. Thus " the discriminant of equation (1), Art. 8, or ^-discriminant," is the same as the "/-discriminant," namely, y 4 ax which
1
y

equated to zero
straight lines.

is

the equation of the envelope of the system of

But, as in the case of the /-discriminant,

it

inferred that every factor gives a singular solution.

must not be For ex-

ample, suppose a squared factor appears in the ^-discriminant. The locus on which this factor vanishes is not a curve in crossing which c and

/ become

imaginary.

At any

point of

it

there

Art.
will

9.]

singular solution from complete integral.

32f

be two distinct values of p, corresponding to arcs of par-

ticular integral curves passing

through that point

but, since-

there

is

but one value of

c,

ticular integral,

hence the point

these arcs belong to the same paris a double point or node.

The

locus
it

is

therefore called a " node-locus."

The

factor repre-

does not appear in the /-discriminant, just as that representing a tac-locus does not appear in the ^-discriminant.
senting

Again, at any point of a cusp-locus, as shown at the end of Art. 8, the two branches of particular integrals become arcs of
the same particular integral the values of c become equal, that a cusp-locus also makes the ^-discriminant vanish.
;

so*

conclusions established above obviously apply also ta equations of a degree higher than the second. In the case of

The

the ^-equation the general method of obtaining the condition for equal roots, which is to eliminate c between the original and
is the same as the process of finding the " of a system. or " locus of the ultimate intersections envelope of curves in which c is the arbitrary parameter.

the derived equation,

suppose the system of curves to have for all values of c* a double point, it is obvious that among the intersections
of

Now

two neighboring curves there are two

in

the neighborhood-

of the nodes, and that ultimately they coincide with the node,, which accounts for the node-locus appearing twice in the dis-

criminant or locus of ultimate intersections.


*

In like manner,,

It is noticed in the second foot-note to Art. 7 that for an equation of the degree p takes the indeterminate form, not only at a point through which all curves of the system pass (where the value of c would also be found indeterSo also when the equation is of minate), but at a node of a particular integral. the th degree, if there is a node for a particular value of c, the n values of c at.
first

the point (which

mine

-f- 1

is not on a node-locus where two values of c are equal) deterarcs of particular integrals passing through the point; and there*--

fore there are

-j- 1

distinct values of

at the point,
is

which can only happen


all

when p

takes the indeterminate form, that


is

to say,

when

the coefficients

of

the /-equation Cwhich

of the nth degreed vanish.

See Cayley on Singular Sot-neory of Singular Solu-

lutions in the "Messenger of Mathematics.

New

Series, Vol. II, p; 10. (Collected-

Mathematical Works, Vol. VIII,

p. 529).

The present
its

tions was established by Cayley in this paper and

continuation, Vol. VI, p. 33^

Sec also a paper by Dr. Glaisher, Vol. XII,

p. 1.

322
if

DIFFERENTIAL EQUATIONS.
is

TChap. VII*

a cusp for all values of c, there are three intersections of neighboring curves (all of which may be real) which ultithere

mately coincide with the cusp; therefore a cusp-locus 'will appear as a cubed factor in the discriminant.*
Prob. 35.

Show

equation must be straight


Prob. 36. Solve
is

that the singular solutions of a homogeneous lines passing through the origin.

3/V
-j-

2xjp

+ 4/
y

x"

= o, and show that there


is

a singular solution and a tac-focus.

Prob. 37. Solve yp*

2Xp

o,

and show that there


(Ans.
2
2

an

imaginary singular solution.

2cx

-f-

.)

x )p' 1 Prob. 38. Show that the equation (1 a system of conies touching the four sides of a square.

represents

discriminants.

o examine and interpret both Prob. 39. Solve yp* \xp -\-y 3 2 8* ) $x y* -\-y* (Ans. f -f- 2cx{$y* o.)
;

Art.

10.

Solution by Differentiation.
an equation of the second order, that

The
the
first

result of differentiating a given differential equation of

order

is

is,

it

contains the derivative

d %y
-*-?
;

but,

if

it

does not contain

ex-

plicitly, it may be regarded as an equation of the first order for the variables x and/. If the integral of such an equation can be obtained it will be a relation between x, p, and a constant

of integration

c,

by means

of

which

can be eliminated from


f

the original equation, thus giving the relation between x, y and c which constitutes the complete integral. For example, the equation

j|+ **=*+/;
* The discriminant of Pc*

(1)

+ Qc

-f-

R=o

represents in general an envelope,

no

further condition requiring to be fulfilled as in the case of the discriminant


-f-

of Lp*1
there

Mp

-f-

N=

o.

Compare

the foot-note to Art.

8.

Therefore where

an integral of this form there is generally a singular solution, although Ip* -j- Mp -J- ^V = o has not in general a singular solution. We conclude, thereare one- valued functions of x fore, that this equation (in which Z, M, and
is

and y) has not


.Series, Vol.

an integral of the above form in which P, Q, and R are one-valued functions of x and y. Cayley, Messenger of Mathematics, New
in general

VI,

p. 23.

Art.

10.]

SOLUTION BY DIFFERENTIATION.

323

when

solved forjy, becomes

y
whence by

= x + ltf;
I

(2)

differentiation

+^-Iand
-

(3)
its inte-

The
gral

variables can be separated in this equation,


is

vp- C - e"
Substituting in equation
(2),

we
.

find

y
which
is

x ~\

6+> _ w
q
e
(1).

the complete integral of equation

This method sometimes succeeds with equations of a higher degree when the solution with respect to p is impossible or
differential leads to a form which cannot be integrated. equation between p and one of the two variables will be ob-

tained by direct integration when only one of the variables is explicitly present in the equation, and also when the equation In the latter is of the first degree with respect to x and y.
case after dividing by the coefficient of y, the result of differentiation will be a linear equation for x as a function of p, so

that an expression for x in terms of p can be found, and then by substitution in the given equation an expression for y in

Hence, in this case, any number of simultaneous values of x and y can be found, although the elimination of p
terms of
p.

may be

impracticable.

In particular, a homogeneous equation which cannot be solved for p may be soluble for the ratio y x, so as to assume
:

the form y

= x<p(p).

The

result of differentiation

is

p=<t>{p)+x<p\pf>
in

which the variables x and p can be separated.

Another

special case

is

of the

form
(1)

y = P*+AP),

324

DIFFERENTIAL EQUATIONS.
is

[Chaf- VII.

which

known
is

as Clairaut's equation.

The

result of differ-

entiation

which implies either

* +f'(p) =o,

or

dp

-=0.

The
first

elimination of / from equation (i) by means of the of these equations * gives a solution containing no arbiis,

trary constant, that


differential

a singular solution.
for

The second
is

is

a.

equation

(i)

equation gives the complete integral


;

its

integral

c,

which

in

y = c*+A<)-

(2)

This complete integral represents a system of straight lines, the singular solution representing the curve to which they are An example has already been given in Art. 8. .all tangent.

differential equation

is

sometimes reducible to Clairaut's


less

form by means
variables.
It

of a

more or

obvious transformation of the


that an equation of

may

be noticed

in particular

the form

y
is

=
;

nxp

+ 00, /)

sometimes so reducible by transformation to the independent n and an equation of the form variable z, where x = z

y
double transformation

= nxp+<t>{y>P)y
= yn
.

by transformation to the new dependent variable v


of

the form indicated

when

the last term

is

a function of both
7>V

may x and jj/as well

succeed
as of p.

Prob. 40. Solve the equation


solution and a cusp-locus.

*P

+
-j-

3/
c

find

a singular
-(x-\-c)*,)

(Ans. (x-\-y

= i)*

Prob. 41. Solve 2V


(Ans.

= xp

a
-\
,

and

find a cusp-locus.
3

\2acxy -f 8ry

\2x1y l

i6ax*

= o.)

* The equation is in fact the same that arises in the general method for the See Art. 9. condition of equal roots.

AKT.

11.]

GEOMETRIC APPLICATIONS
a*)/? 2xyp 1 circle x

TRAJECTORIES.
y*
a*

325

Prob. 42. Solve (x*


(Ans.

The

+/ =
#y

-\-

o.
its

a
,

and

tangents.)

Prob. 43. Solve ,y

=
(Ans.
s

xp
<:

+ x*p*
-J-

= o,

and

+ ^x*y =

o.)

Prob. 44.

Solve / 8y* = o. 4#>p = c(x c)*; 2jy = 4X* andjy = o are singulai (Ans. 7 also a particular integral.) solutions; y = o
-f-

is

Prob. 45. Solve

x\y

px)

=yp\

(Ans.

ex 1

+ c\)

Art.

11.

Geometric Applications
of a curve

Trajectories.

Every property
its

which involves the direction of

tangents admits of statement in the form of a differential equation. The solution of such an equation therefore determines the curve having the given property. Thus, let it be required to determine the curve in which the angle between

the radius vector and the tangent is n times* the vectorial Using the expression for the trigonometric tangent of angle.
that angle, the expression of the property in polar coordinates is

= tan nd.
dr
Separating
integral
is

the variables and


rn
in

integrating,

the complete

= cn sin nd.
is

The mode

which the constant of integration enters here


shared by
all

shows that the property in question bers of a system of similar curves.

the

mem-

The
eral

solution of a question of this nature will thus in gen-

be a system of curves, the complete integral of a differential equation, but it may be a singular solution. Thus, if we express the property that the sum of the intercepts on the axes made by the tangent to a curve is equal to the constant a, the
straight lines

making such

intercepts will themselves consti-

tute the complete integral system, and the curve required is the singular solution, which, in accordance with Art. 8, is the

326

DIFFERENTIAL EQUATIONS.
lines.

[CHAP. VII.

envelope of these
to be the parabola

The

result in this case will

be found

Vx + Vy

Va.

application is the determination of the "orthogonal trajectories" of a given system of curves, that is to say, the curves which cut at right angles everyvcurve of the

An

important

given system. The differential equation of the trajectory is for at every readily derived from that of the gfven system of the trajectory the value of pis the negative reciprocal point
;

of

have therevalue in the given differential equation. fore only to substitute p~ for / to obtain the differential
its
l

We

equation of the trajectory. For example, let it be required to determine the orthogonal trajectories of the system of parabolas

having a

common

axis

and vertex.

The
is

differential equation

of the system found

by eliminating a
2

xdy = y dx.

Putting

in place of -~, the differential equation of

dy
the system of trajectories
is

dx

2xdx -\-ydy
whence, integrating,

= o,
= c\
system of similar
ellipses

2x*+y*
The

trajectories are therefore a

with axes coinciding with the coordinate axes.


Prob. 46.
is

Show

that

when the
its
;

system

of the second degree, but will be identical


it

differential equation of a system discriminant and that of its trajectory


if it

represents a singular solution in

one system,

will constitute a

cusp locus of the other.

Prob. 47. Determine the curve whose subtangent is constant and (Ans. ce*=y.) equal to a.

Prob. 48.
r
M

Show

that the orthogonal trajectories of the curves

=c

sinnB are the same system turned through the angle

about

the pole.

Examine

the cases n

1,

2,

and n

-J.

Prob. 49.

Show

that the orthogonal trajectories of a system of

Art.

12.]

simultaneous differential equations.


through two given points
radical axis.
is

3*2?

circles passing

another system of

circles.

having a

common

Prob. 50. Determine the curve such that the area inclosed by any two ordinates, the curve and the axis of x, is equal to the product of the arc and the constant line a. Interpret the
singular solution.
-

-e
a
).)

(Ans. Prob. 51.


onal.

The catenary^

c= \a{e a
is

Show

that a system of confocal conies

self-orthog-

Art.

12.

Simultaneous Differential Equations.


of n equations

A system

between n

-f- 1

variables

and

their

differentials is a "determinate" differential system, because it serves to determine the'/z ratios of the differentials; so that,

taking any one of the variables as independent, the others vary in a determinate manner, and may be regarded as functions of

the single independent variable. Denoting the variables by x, y> z> etc., the system may be written in the symmetrical form

dx
where X, F, If any one
tials

X~Y~~Z~
Z
.

_dy

dz _
'

may

be any functions of the variables.

of the several equations involving two differencontains only the two corresponding variables, it is an and its integral, giving a reordinary differential equation
;

lation

between these two


until n

variables,

may

enable us by elimina-

tion to obtain another equation containing

and so on

integral

two variables only, equations have been obtained.


_ ~~

Given, for example, the system

dx

x
The

"

_dy
'

dz
'

>'*

relation between dy and dz above contains the variaand z only, and its integral is bles^/ z = a. y* (2)
1

dx and dy

Employing this to eliminate z from the it becomes dx _ dy

relation

between

328

DIFFERENTIAL EQUATIONS.
integral
is

[CHAP. VII.

ot which the

y-Y V .(/
The

+ a) = bx.
and

(3)

(3), integral equations (2) of integration, constitute the complete solution. It is in like manner obvious that the complete solution of a system of n equations should contain n arbitrary constants.

involving two constants

Confining ourselves
-extension of

now

to the case of three variables, an

the geometrical interpretation given in Art. 2 presents itself. Let x, y, and z be rectangular coordinates of starts from any given Preferred to three planes. Then, if

position A, the given system of equations, determining the ratios dx: dy\ dz, determines the direction in space in which moves. As moves, the ratios of the differentials (as deter-

mined by the given equations)


to

move

in

such a

way

will vary, and if we suppose as to continue to satisfy the differential

equations, it will describe in general a curve of double curvature which will represent a particular solution. The complete solution is represented by the system of lines which may be

thus obtained by varying the position of the initial point A. This system is a " doubly infinite " one for the two relations
;

between x, y, and z which define it analytically must contain two arbitrary parameters, by properly determining which we can make the line pass through any assumed initial point.*

Each
.-surfaces

of the relations

between

x,

y and

z,

or integral equa-

tions, represents

by

itself

a surface, the intersection of the

two

being a particular line of the doubly infinite system. equation like (2) in the example above, which contains only one of the constants of integration, is called an integral of the

An

^differential

system,

in contradistinction to

an " integral equa-

*lt
S)l x.,y,

is

assumed
z.

in the explanation that


is

X, V, and Zare one-valued functions

and

There

then but one direction in which

can

move when

^passing a given point, and the system is a non-intersecting system of lines. But [if this "is not the case, as for example when one of the equations giving the ratio
of the differentials
is

of higher degree, the lines

may form an

intersecting sys-

tem, and there would be a theory of singular solutions, into which we do not
Jiere enter.

ART.
tion

12.]

SIMULTANEOUS DIFFERENTIAL EQUATIONS.

329
integral

"

like (3),

which contains both constants.

An

represents a surface which contains a singly infinite system of lines representing particular solutions selected from the doubly

Thus equation (2) above gives a surface on those lines for which a has a given value, while b -which may have any value whatever in other words, a surface which
infinite system.
lie all
;

passes through an
lines.

infinite

number

of the particular solution

integral of the system which corresponds to the conb might be found by eliminating a between equations (2) stant and (3). It might also be derived directly from equation (1)
;

The

thus we

may

write

dx

x
tion between
its integral,

_dy ~~ _ _dz
'
'

dy
'

'

+ dz +z y
z
is

_du
u
9

in which a new variable u

=y

-\-

introduced.

The

rela-

dx and du now contains but two

variables,

and
(4)

y+z

= bx,
;

and this, together with is the required integral of the system the integral (2), presents the solution of equations (1) in its The form of the two integrals shows that in standard form.
this case the doubly infinite system of lines consists of hyperbolas, namely, the sections of the system of hyperbolic cylinders represented by (2) made by the system of planes represented

by

(4).

A system
tain

of equations of

symmetry may
Since

which the members possess a cersometimes be solved in the following

manner.

dx _ dy _dz _ \dx -f- jxdy vdz_ 3T~~ Y~~~Z~~ XX+pY+vZ'


if

we

take multipliers

A,

//,

v such that

we

shall

have

XX+ M Y+vZ = o, hdx + pdy + vdz = o.


first

If the expression in the

member

is

an exact

differential,

330

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

direct integration gives an integral of the given system.

For

example,

let

the given equations be

dx

dy

dz
Iz

mz
/,

ny

nx

ly

mx

'

and n form such a and z. Hence we have


Idx

set of multipliers,

and so

also

do

x,

and

also

+ mdy + ndz = o, xdx + y dy z dz =


-f-

o.

Each

of these

is ar\

sxact equation, and their integrals


Ix -f-

my -\-

nz

and

*+/+**=
namely, the line passing through the origin and whose
/,

The doubly infinite system constitute the complete solution. this case of circles which have a common of lines consists in
axis,

direction cosines are proportional to


Prob. 52. Solve the equations -5
~
,

m
,

and
,

n.

dx

dy

dz

2xy

2xz
i

and

interpret the result geometrically.

(Ans.

y=az,
.

x*-\-y* -\-z'

=bz.)

Prob. 53. Solve

dx -

+z
dx

=
z

dy
-\-

f x
tf(x

dz

-\-y

(Ans.

+y + z)=--=
dy
r

-A-.)

Prob. 54. Solve D^

r
,

dz
7
rr
.

c)yz (Ans. x'

(b

\c

+/ + z* = A, ax* -f bf + cz* =

a)zx

(a

b)xy

.)

Art.

13.

Equations of the Second Order.

A relation
tives of

between two variables and the successive deriva-

one of them with respect to the other as independent

variable

is called a differential equation of the order indicated For example, the highest derivative that occurs. by

is

an equation of the second order, in which

is

the independent

ART.

13.]

EQUATIONS OF THE SECOND ORDER.


first

331

variable.

Denoting as heretofore the equation may be written


,)

derivative

by/,

this

(I+ * i; + */+w* =
and
this, in

'

(I>

connection with

which defines /, forms a pair of equations of the first order, connecting the variables x, y, and /. Thus any equation of the
second order
is

equivalent to a pair of simultaneous equations

of the first order.

When,
tain

as in this example, the given equation does not con-

explicitly, the first of the pair of equations involves

only

the two variables

x and/

and

it is

further to be noticed that,


first

when

the derivatives occur only in the

degree,

it is

a linear

equation for/.

Integrating equation (i) as such,

we

find

p = ->*
and then using
this value

v{1
in

\^
(2), its

(3>

of/
x

equation

integral

is

y = c,-mx + c
in

log [x

+ y(i + x')l
is

(4)

which, as in every case of two simultaneous equations of the first order, we have introduced two constants of integration.

An

equation of the

first

order

readily obtained also


in

when the independent variable is not explicitly contained the equation. The general equation of rectilinear motion
dynamics affords an
illustration.

in

This equation

d*s
is

=/(j),

where
force

denotes the distance measured from a fixed center of


line of motion.
It

upon the

may be

written

dv
at

= f[s), in
Eliminat-

connection with
dt

= v,

which defines the

velocity.

ing dt from these equations,


integral
is

we have vdv=f(s)ds, whose


the "equation of energy" for

-JV

/ f(s)ds

-f- c,

the unit mass.

The

substitution of the value found for v in the

332

DIFFERENTIAL EQUATIONS.
t is

[CHAP. VII.

second equation gives an equation from which terms of s by direct integration.

found

in

The
order

is called
;

integration, such as equation (3) above, a "first integral" of the given equation of the second it contains one constant of integration, and its complete
result of the
first

integral,

which contains a second constant,


"

is

also the "

com-

plete integral

of the given equation.


is

all its
is

when, terms being transposed to the first member, that member the derivative with respect to x of an expression of the first
is,

A differential equation of the second order


a function of x,

"

exact

"

order, that

y and p.

It is

obvious that the

terms containing the second derivative, in such an exact differential, arise solely from the differentiation of the terms containing/
in the function of x,

y and/.

For example,

let it

be

required to ascertain whether

<>- 4+'=
The terms
in

is

an exact equation.
arise

question are

(1

x*)

ax
2

which can

only from

the differentiation

of (1

x )p.

Now

subtract from the given expression the complete derivax*)p,

tive of (1

which

is
'

2 (i-x*)^--2x dy (I X)

dx>

*dx>

the remainder
that of xy.
ential,

ax -\- y, Hence the given expression

is

x-f-

which

is

an exact derivative, namely,


is

an exact

differ-

and

{i-*^ + *y = c
is

(6)

the

first

integral of the given equation.

Solving this linear

equation for y, we find the complete integral

y
Prob.55.

Cl

+ C^(l

-X^).

(7)

= Solve(i-^)g-^| 2. -1 x)* (Ans. y = (sin

-f-

c x sin"

-f-

cv )

Art.

14.]

THE TWO FIRST INTEGRALS.

333

Prob. 56. Solve Prob. 57. Solve

g=&
-~

(aus.
a

+^

.)

=a

*
2

j.
at

(Ans.

Py

=A

sin

bx

-{2

B cos Ar.)

Prob. 58. Solve

y? + (g)" =1.

(Ans.

/ = * + ** + <V>

Art.

14.

The Two First

Integrals.

We
tween
tion

have seen

integral of
x,

in the preceding article that the complete an equation of the second order is a relation beand two constants c and c % Conversely, any relax

between

x,

y and two

arbitrary constants

may

as a primitive,

from which a

differential equation free

be regarded from both

The process consists in arbitrary constants can be obtained. first obtaining, as in Art. 3, a differential equation of the first
order independent of one of the constants, say c^ that is, a relation between x, y,p and c and then in like manner eliminating
,
x

from the derivative of

this equation.

The

result
f

is

the equa-

tion of the second order or relation

between x

y,

p and

q {q

denoting the second derivative), of which the original equation is the complete primitive, the equation of the first order being
the
is

first

integral in

which

ct

is

obvious that

we

can, in like
x, y,

the constant of integration. It obtain from the primimanner,


cq
,

tive a relation

between

p and

which

will also

be a

first

integral of the differential equation.

Thus, to a given form of


first

the primitive or complete integral there corresponds two


integrals.

infinite

Geometrically the complete integral represents a doubly system of curves, obtained by varying the values of c
t

and of

independently.

arbitrary,

we

regard cx as fixed and <r3 as select from that system a certain singly infinite
If

we

system

the

first

integral containing c

is

the differential equa-

tion of this system, which, as explained in Art. 2, is a relation between the coordinates of a moving point and the direction

of its

motion

common

to

all

the curves of the system.

But

334

DIFFERENTIAL EQUATIONS.

[Chap. VII.

the equation of the second order expresses a property involving curvature as well as direction of path, and this property

being independent of

cx is

common

to

all
is,

the systems corre-

sponding
infinite

to different values of cv that

to the entire
this

doubly

system.

moving

point, satisfying

equation,

may

in any direction, provided its has the proper curvature as determined by the value of q path derived from the equation, when the selected values of x> y

have any position and move

and/ have been

substituted therein.*
(7)

For example, equation


sents an ellipse having
its

of the preceding article repre-

the lines

1,

center at the origin and touching as in the diagram c x is the ordinate of the
;

1, and c 2 that of the point in point of contact with x If we regard c as fixed the ellipse cuts the axis of y.
y

which
and
<r

as arbitrary, the equation represents

the system

of

ellipses
(6) is

touching the two lines at fixed points, and equation


differential equation
like

the
In

of this system.

manner,

if c^ is

fixed

and

arbitrary,

equation (7) represents a

system of ellipses
fixed
1.

cutting the axis of y in and touching the lines

points

x=

The

corresponding differential equation will be found to be

(y- Xp)V{l-X*)=C,.
Finally, the equation of the second order, independent

of c

and

the equation of the 2 doubly infinite system of conies f with center at the origin, 1. and touching the fixed lines x
[(5) of the preceding article]
is

If

the equation
is

is

of the second or higher degree in q, the condition for

equal roots

a relation between x,

y and/, which may be found

to satisfy the

given equation. If it does, it represents a system of singular solutions; each of the curves of this system, at each of its points, not only touches but osculates It is to be remembered that a singular soluwith a particular integral curve.
tion of a
tion; for
first
it

integral is not generally a solution of the given differential equarepresents a curve which simply touches but does not osculate a set

of curves belonging to the doubly infinite system.


f Including hyperbolas corresponding to imaginary values of
c%.

Art.

14.]

THE TWO FIRST INTEGRALS.

335

we

But, starting from the differential equation of second order, may find other first integrals than those above which correto c 1

spond

and

c^.

For instance,

if

equation

(5)

be multiplied

by/,

it

becomes

which

is

also an exact equation, giving the first integral


(1

-*v +/ =
integrals

',*.

in

which

c s is a

new constant
first

of integration.

Whenever two

have thus been found inde-

pendently, the elimination of p between them gives the complete integral without further integration.* Thus the result of eliminating

/ between
cx

this last
(6),

equation and the


is

first

inte-

gral containing

[equation
2c
x

Art. 13]

/
which
is
is

xy

+ f,V = c?

c*

obvious from the

therefore another form of the complete integral. It first integral above that c 2 is the maximum
it is

value of y, so that of ellipse inscribed

the differential equation of the system in the rectangle drawn in the diagram.

comparison
a

of the

two forms

of the

complete integral shows


is

that the relation between the constants


If
first

ct

= c*

-\- c*.

the form

is, put in the constant will disappear on differentiation, and the result will be the given equation of second order multiplied, in general, by an integrating factor. can

integral be solved for the constant, that


y, p)

<J>(x,

c,

We

thus

find

any number

of integrating factors of an equation

already solved, and these

may

suggest the integrating factors

of more general equations, as

illustrated in Prob. 59 below.


in Art. 10 to the

* The principle of this method has already been applied


solution of certain equations of the
first

order; the process consisted of forming the equation of the second order of which the given equation is a first integral

{but with a particular value of the constant), then finding another and deriving the complete integral by elimination of p.

first

integral

336

DIFFERENTIAL EQUATIONS.
ji

[CHAP. VII..

Prob. 59. Solve the equation

+ <?y = o in the form


B sin ax;
also inte-

=A

cos ax

-f-

and show that the corresponding integrating factors are


grating factors of the equation

where

is

any function of x; and thence derive the

integral of this

equation.

(Ans. ay

sin

ax

I cos

ax

Xdx

cos ax I sin ax

Xdx).

Prob. 60. Find the rectangular and also the polar differential equation of all circles passing through the origin.

(An,(^

+/ )g=
Art.

[
15.

+ (|y](,|-4

an d

-)

Linear Equations.
of

any order is an equation of first degree with respect to the dependent variable y and each of its derivatives, that is, an equation of the form
the

A linear differential equation

where the

coefficients

Pn and

the second

member

X are

functions of the independent variable only.

The
in the

form

solution of a linear equation is always supposed to be y z=.f(x) and if y is a function which satisfies the
;

equation, it is customary to speak of the function yl% rather than of the equation y y as an "integral" of the linear equaK ,

tion.

general solution of the linear equation of the first order has been given in Art. 6. For orders higher than the
first

The

the general expression for the integrals cannot be effected


integral

by means of the ordinary functional symbols and the sign, as was done for the first order in Art. 6.

The

solution of equation (1) depends

upon that of

dnv

dn ~^v

Art.

15.]

linear equations.

337

integral of this equation will contain n arbiconstants, and the mode in which these enter the exprestrary sion for y is readily inferred from the form of the equation.

The complete

For

let

be an integral, and

cx

an arbitrary constant
(2) is c x

the

re-

y = y that is, it is zero therefore c y is an integral. So too, if y is an integral, j/ is an integral and obviously also c y + j/ is an integral. Thus, if n distinct integrals y n can be found, y%i ,y + + cn y n y=cy + (3)
putting
x \
;

sult of putting

= cj

in

equation

times the result of


l
1

x ,

will satisfy the equation, and, containing, as

it

does, the proper

number

of constants, will be the complete integral.

Consider
it,

now equation

(1);

let

Fbe
o.

a particular integral of
of equation
(3),

and denote by u the second member


the complete integral

which

is

when

X=

If

y=Y+u
be substituted
in

(4)

equation

(1),

the result will be the


of putting

sum
the

of the
first

results of putting

y=

Fand

y=u

of

these results will be X, because Fis an integral of equation (1), and the second will be zero because u is an integral of equation
(2).
it

since
is

Hence equation (4) expresses an integral of (1); and contains the n arbitrary constants of equation (3), it

this equation

the complete integral of equation (1). With reference to Y is called " the particular integral," and u is The particular integral called "the complementary function."

contains no arbitrary constant, and any two particular integrals may differ by any multiple of a term belonging to the comple-

mentary function. If one term of the complementary function of a linear equation of the second order be known, the complete solution can be found. For let y be the known term then, if y = y v
x
;

be substituted in the first member, the coefficient of v in the result will be the same as if v were a constant it will therefore be zero, and v being absent, the result will be a linear equa:

tion of the

first

order for

v',

the

first

derivative

c*f v.

Under

338

DIFFERENTIAL EQUATIONS,

[CHAP. VII.

the same circumstances the order of any linear equation can


in like

manner be reduced by

unity.

very simple relation exists between the coefficients of an exact linear equation. Taking, for example, the equation of the second order, and indicating derivatives by accents, if

P.S + PJ+P y = x
t

term of the integral will be P y' Subtracting the derivative of this from the first member, the remainder is P ')y' Pt y. The second term of the integral must (P
is

exact, the

first

therefore be

{P

P ')y
is

subtracting the derivative of this ex-

pression, the remainder,

(Pt

P/

-\-

P ")y, must vanish.


9

Hence

P -f- P" = o
'

the criterion for the exactness of the


result obviously

given equation.

A similar
x
d^y
(3

extends to equa-

tions of higher orders.


x

Prob. 61. Solve

dy + x)-j-

-f-

3V

=
2
7

o,
z

noticing that e*
$x* -f 6x
2y
3

is

an integral.
Prob. 62. Solve (*
5

(Ans.
72 a

cj?

+ c (x
-f-

-f-

-\- 6.)

x)-jr$

+ 2(2*

i)~-

-f-

= o.
-f-

(Ans. (x

i)

= c,(x*

6x* -f 2X
2 sin

$
B-^

4X log x)

cjc*.)

Prob. 63. Solve-^3


( Ans.

+ cos
fe
sin

2 #-^-

y
~

cos
sin9

= Sm

sin 26.

= e~

sin 6

\cfi

c% )d0

c% e

\\

Art.

16.

Linear Equations with Constant


Coefficients.

The linear equation with member zero may be written


x

constant coefficients and second


in the
. .

form
.

AJPy + AJXr y +
in

+A y =
n

o,

(1)

which
n

D stands for
=
,

the operator
is

-3-,

D*

for -r%, etc., so that

indicates that the operator 2 mx e memx since Demx

to be applied n times. Then, wV", etc., it is evident that if

ART.

16.]

LINEAR EQUATIONS, CONSTANT COEFFICIENTS.


in

339

be substituted the factor emx will be


e

mx

equation
1

(i),

the result after rejecting

A
(1)

w"

+ A.m"- +
equation
.
.

+A =
n
is

o.

(2)

Hence, and
;

if

m
if

satisfies

(2),

mx e

an integral of equation
(2),

w
y

mn

are n distinct roots of equation

the complete integral of equation (1) will be


3= c x

W+

cj**

cn e

mn
*.

(3)

For example,

if

the given equation

is

(Py

dy

the equation to determine

m
x

is

m?
of

which the roots are


is

m m =
3JC

2
2,

= o, m =
9 a *-*.

therefore the in-

tegral

y
The
tion.

= c^
f

-J-

general equation (1)


in

form f(D) .y = o,
equation
is

which

may be written in the symbolic denotes a rational integral funcf{m)

Then equation
(m

(2) is

= o,
{m

and, just as this last

equivalent to

m^(m

t)

m n = o,
)

(4)

so the symbolic equation

f{D) .y = o
m,)

may

be written
(5)

(D

m )(D
t

...(>-

m n )y = o.

This form of the equation shows that it is satisfied by each of the quantities which satisfy the separate equations

(D
that
is

m^)y

= 0,

(B

m^)y

=o

(D

m )y = o
n

(6)

to say,

by the separate terms

of the

complete integral.

If

two

of the roots of equation (2) are equal, say to

of the equations (6)

number

of integrals

become we must

mlt two and to obtain the full identical, find two terms corresponding to
x
;

the equation

(D
in other

- m Yy = o
integral.

(7)

words, the complete integral of this equation of which


is

:= em x
i

known, to be one

For

this

purpose

we

340

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

v. t put, as explained in the preceding article, y Dem xv e m x {m x v -f- Dv) therefore tiation, Dy

=y

By

differen-

(D
In like manner

tn x )e**v

= e m *Dv.

(8)

we

find

{D

m m^fe ^v

= e m xD v.
l
^

(9)

Thus equation

(7)
is

is

transformed to
2
;

D'v = o,

of

which the
is

complete integral

= c,x -\-c hence y = e^x {c x-\- c


x

that of equation (7)

%).

(10)

These are therefore the two terms corresponding to the squared


factor

(D

mY
x
1

in

f(D)y

= o.

evident that, in a similar manner, the three terms corresponding to a case of three equal roots can be shown to
It is

be

^(c^x + cjc -j- c ) and so on. The pair of terms corresponding to a pair of imaginaryroots, say **, = a-\- ifi, mt ~ a ifi, take the imaginary form
s
f

Ci gi*+iP)x _j_ c

^-mx

t**(c/t*

+ c e-#*).
t

and imaginary parts of e# x and Separating changing the constants, the expression becomes
the real
e ax (A

*-**,

and

cos fix-\-B sin fix),

(11)

For a multiple pair of imaginary roots the constants A and must be replaced by polynomials as above shown in the case

of real roots.

When
be made
the
first

coefficients

member of the equation with constant a function of X, the particular integral can also to depend upon the solution of linear equations of
the second
is

In accordance with the symbolic notation introduced above, the solution of the equation
order.

dy
-j^
is

ay

X,

or

(D

- a)y = X

(12)

denoted by

(D

a)~*X, so that, solving equation (12),

we have

^X=e'fe-**Xdx
as the value of the inverse symbol whose meaning
is

(13)

"that

ART.

16.]

LINEAR EQUATIONS, CONSTANT COEFFICIENTS.

341

is converted to by the direct operation a" Taking the most convenient the symbol expressed by special value of the indefinite integral in equation (13), it gives

function of

x which

the particular integral of equation


ticular integral of

f{D)y = X is

(12). In like manner, the pardenoted by the inverse symbol

TfpfX.
fraction

Now, with the notation employed above, the symbolic

may

be decomposed into partial fractions with constant


:

numerators thus

in

f(D) which each term


regarded

TTF^X = D n

'

n m X+ D
'

N X + m^
a
'

N + -ft^X* (14) ^' D m


'

is

to be evaluated

by equation

(13),

and

may be
ary

(by virtue of the constant involved in the

indefinite integral) as containing

function.

one term of the complementFor example, the complete solution of the

equation

3---*
y

is

thus found to be

= \#*fe-**Xdx - \e~*fXdx.
a

When

is

power

of

the particular integral

may be

found as follows, more expeditiously than by the evaluation of

For example, if x* the integrals in the general solution. the particular integral in this example may be evaluated by development of the inverse symbol, thus
:

_ _ ~ y ~ D*-D-2 X a__I 2
I

i+KZJ-JT)

* The validity of this equation depends upon the fact that the operations in the second member of expressed

f{D)

= (D -

mi )(D

-m,)

+ ...+(D-m M
is

)
if

are commutative, hence the process of verification

the

same as

the equation

were an algebraic identity. This general solution was published by Boole in It had, however, the Cambridge Math. Journal, First Series, vol. II, p. 114. been previously published by Lobatto, Theorie des Characteristiques, Amster-

dam,

1837.

342

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

that, in this case, it is only necessary to carry the development as far as the term contain-

The form

of the

operand shows

ing

D\

For other symbolic methods applicable to special forms of we must refer to the standard treatises on this subject.
Prob.64. Solve

dy

dy

4^3

-3^+J=o.
(Ans.

= e*x (Ax + JB) + ce~x

.)

Prob. 65. D

Show

that

7n^e AD)

ax

-^-\** /(a)

and that
Prob. 66. Solve
2

(Z>

+ )y = e* + sin
1 -f-

y^

sin (ax

+ ft) = jA^. sin (ax +


2x
-f-

ft).

sin

x.

(Compare

Prob. 59, Art. 14.) sin (Ans. y

=A
17.

+ B cos x

i<^

sin

2.3;

\x cos

#.)

Art.

Homogeneous Linear Equations.

The

linear differential equation

AX
-

"S- +
etc.,

^"& +---+^y =
is

o,

(1)

in

which

A A x%
,

are constants,

called the

"homogene.

m ous linear equation." It bears the same relation to x that mx the equation with constant coefficients does to e Thus, if m will divide m be substituted in this equation, the factor x y=zx

out from the result, giving an equation for determining ;;/, and the n roots of this equation will in general determine the For example, if in the n terms of the complete integral.

equation
u d*y x*--4

ax

dy + 2X-fdx
.

2y

=o
1) -f-

we put y
(m

= xm

the

result

is

m(m

2m
and

= =

0,

or

The Hence
is

i)(m -f- 2) roots of

= o.
this

equation

y = cx+
x

are

m =
x a

/# 3

2.

the complete integral.

Equation

(1)

might

in fact

have been reduced to the form

with constant coefficients by changing the independent vari-

Art.

17.]

homogeneous linear equations.


0,

343

able to

where x

=e

e
,

or 6

= log

x.
in

We may

therefore at

the preceding article that the terms corresponding to a pair of equal roots are of the

once infer from the results established

form
(cx

c%

log x)x

m
y

(2)

and

also that the

terms corresponding to a pair of imaginary


cos

roots,

i/3,

are
(J3

xa [A

log x)

-\-

B sin

[ft

log x)\

(3)

the two classes of linear equations considered in this and the preceding article is more clearly

The analogy between


when a
.

xD is used for the operation of single symbol the derivative and then multiplying by x, so that taking xm mx m It is to be noticed that the operation x*D 2 is not a the same as # or xDxD, because the operations of taking the
seen

derivative

and multiplying by a variable are not


is,

"

commu-

tative," that
7

their order

is
;

not indifferent.

We

have, on the

contrary,

x D*

= $(#
2

1)

then the equation given above,


2)7

which

is

(x

D*-\-2x>2]/
o

= 0,
($

becomes
[${&
1)

+ 2#

= o,

or

i)(S

+ 2)y =
m.

o,

the function of

$ produced being

the same as the function of

which

is

equated to

in finding the values of

linear equation of which the first member is homogeneous and the second member a function of x may be reduced

to the form

A*)-y
The

=&

(4)

particular integral may, as in the preceding article (see eq. (14)), be separated into parts each of which depends upon the solution of a linear equation of the first order. Thus,

solving the equation

xdy dx

ay

= X,

or

($

- a)y = X,
iXdx
'

(5)

we

find

W^~'

X = xafx

~a~

6>

The more

expeditious method which

may be employed

344

DIFFERENTIAL EQUATIONS.

[CHAP. VII.
:

when A* is a power of x is illustrated in the following example d v dv Given x* -4 The first member becomes homo2-7- = ^ dx dx geneous when multiplied by x, and the reduced equation is
2
.

($

3^ = *.
roots of f{&)
r

The

=0 are
r
,

and the double root


%
x

zero,

hence
8

is c x -f- c^ -f- c Since in log *:, we infer that in operating upon ^ we general J\$)x =f(r)x may put # = 3. This gives for the particular integral

the complementary function

S
but
fails
fall

-3^

.3

9 #

'

3*

now

with respect to the factor # 3.* back upon equation (6), which gives

We

therefore

=z

x jx-'dx
3

=x

log x.

The complete

integral therefore
3

is

y = ti* + + c
c,

lo g

+ ix
y

lo g

Prob. 67. Solve 2^ ^~4

d"*y

+ Zx

dy
~j

3J
(Ans.

?-

cx x

+ ^" + +*'.)
8
2

Prob. 68. Solve (* Z>

+ 3 *Z> + JD)y = -. (Ans. ^ = c + * log x + (log


2
3

<r

.x)

+ J(log *)

.)

Art.

18.
in

Solutions in Infinite Series.


this article to illustrate the

We

proceed

method by
series

which the integrals

of a linear equation

whose

coefficients are
in

algebraic functions of x terms are powers of x.

may
For

be developed
this

whose

purpose

let

us take the

equation

is a term of the complementary function having an indeterminate coefficient; accordingly the new term is of the same form as the second term necessary when 3 is a double root, but of course with a determinate coefficient.

* The

A failure occurs because x'

^RT.

18.]

SOLUTIONS IN INFINITE SERIES.

345

which

is

known
the
first

as " Bessel's Equation,"

and serves to define

the "Besselian Functions."


of this equation m the result is the single term Ax m ri Ax m+\ A{n? )x
If in

member

we

substitute for_^

(2)

the first term coming from the homogeneous terms of the equation and the second from the term x*y which is of higher If this last term did not exist the equation would be degree.
satisfied

by the assumed value of/,


first

if

w were

determined so as

make the Bx~ n Now


to
.

term vanish, that is, in this case, by Ax n or these are the first terms of two series each of
if

the equation. For, a term containing x m+2 thus y A


satisfies
,

which

xm

we add to mJr2 -\- A x


x

the value of
,

the

new term

will give rise, in the result of substitution, to terms containing ^m+2 ancj x m+4 respectively, and it will be possible so to take that the entire coefficient of xm+2 shall vanish. In like

manner the proper determination coefficient of x M+4 in the result of


on.

of a third

term makes the

substitution vanish,

and so

We therefore at once assume m m y = 2 A,x m+ar = A x + A xm+2 + A,x +


x

(3)

in

which r has

all

integral values from

o to

00

Substituting

in equation (1)

2[{ (in

+ 2r)

}A rx

m+2r

+ A xm+2
r

I)

= o.
,

(4)

The

coefficient of each
;

power

of

in this

arately vanish

hence, taking the coefficient of

equation must sepxm+2r we have


o.

[(m + 2ry-n*]A + A _ =
r
r
t

(5)

When

= o,
Ar

this reduces to

m*

n2

= o,
it

which determines
gives
'x
'

the values of m, and for other values of r

=~
(m

+ 2r + n){m + 2r - n) Ar
l

the relation between any two successive coefficients.

For the

first

value of m, namely

n, this relation

becomes

346

DIFFERENTIAL EQUATIONS.
in

[CHAP. VII,

whence, determining the successive coefficients


the
first

equation

(3),.

integral of the equation

is

Ay,

= =

AX[i

j^pj

~,

+
(

+l)(w
is

+ 3)

^5771

-...}

(7)

In like manner, the other integral

found to be

Bj,

Bs-\i

+ ^-^ + {n _
is

,)(_
l

2)

^, + -],

(8)

and the complete integral

=Ay
t

-f-

Bjv *
For example,
n

This example illustrates a special case which may arise in If n is a positive integer, the second this form of solution.
series will contain infinite coefficients.
if

2,,

the third coefficient, or

is infinite,

unless

we take

B = o,

in

which case
solution

is

indeterminate and
will

This

we have a repetition of the always occur when the same powers of


the case in which

x occur

in the

two

series, including, of course,

new integral has equal roots. For the in such cases the complete treatises must be referred to.f
of obtaining a
It will

mode

be noticed that the simplicity of the relation between

consecutive coefficients in this example is due to the fact that equation (1) contained but two groups of terms producing of when Axm is substituted for as in exdifferent

powers

x,

pression

(2).

The group containing

the second derivative

necessarily gives rise to a coefficient of the second degree in m, and from it we obtained two values of m. Moreover, be-

cause the other group was of a degree higher by two units, the

assumed of x\
*

series

was an ascending one, proceeding by powers


function of the nth order usually denoted by n is the value n if n is a positive integer, or generally by i T{n-\1).
of these functions see

The Besselian

of y\ above, divided by i n n\

For a complete discussion


Bessel'schen
fA

Functionen, Leipzig, 1868; Todhunter's


1875, etc.

Lommel's Studien iiber die Treatise on Laplace's,

Lame's and Bessel's Functions, London,


regular solution and log x, and

solution of the kind referred to contains as one term the product of the

is sometimes called a "logarithmic solution." See also American Journal of Mathematics, Vol. XI, p. 37. In the case of Bessel's equation, the logarithmic solution is the "Besselian Function of the

oecond kind."

Art.

18.]

solutions in infinite series.

347

In the following example,

d*y

dy

there are also two such groups of terms, and their difference of degree shows that the series must ascend by simple powers.

We
The

assume therefore

at

once

y=
result of substitution
is

2A
m+r- 2

xm ^.

(io)

%[{(m+r)(m+r-i)-2}A rx
Equating

+ a(m+r)A xm+
r

r-I

~\= o.

(u>

to zero the coefficient of

m+r~ 2
,

{m-\-r-\- i)(m
which,

+
(m

2)A r

+ a{m + r

i)A r _ 1

= o,

(12)

when

= o,
o,

gives

+ \){m - 2)A = o,
m-\-r
1

(13)

and when r

>

Ar

The

roots of

\m+r+i){m+r- 2) Ar'T equation (13) are m = 2 and m =

-"

(I4)
1;

taking

m = 2,

the relation (14) becomes

whence the

first

integral

is

A,yt

= As[i -

\ax

Taking the second value

^V - ~ m=
i
f

6 '*'+--

(V|)

equation (14) gives

whence
gral
is

B =
x

-2?

and

i?3

o *; therefore the second

inte-

the finite expression

^,=^-[i-^>4i-f].
*B
3

(16)

would take the indeterminate form, and


is

if

value, the rest of the series

equivalent to

yi,

we suppose it to have a finite reproducing the first integral.

348

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

When

the coefficient of the term of highest degree in the


it

result of substitution, such as equation (u), contains m,

is

possible to obtain a solution in descending


this case,

powers

of x.

In

occurring only in the


;

solution can be found

it

degree, but one such would be identical with the finite


first

In the general case there will be two such soluintegral (16). tions, and they will be convergent for values of x greater than unity, while the ascending series will converge for values less

than unity.*

When

the second

member

of the equation
in

is

power

of x,

the particular integral can be determined


in a similar

the form of a series

For example, suppose the second member of equation (9) to have been x*. Then, making the substitution as before, we have the same relation between consecumanner.
tive coefficients; but

when

o,

instead of equation (13)


-2

we

have

{m

+ i){m

m 2)A x

= x*
This gives
(14),

to determine the initial term of the series.

m=

2\

and

% hence, putting the particular integral f


;

A =

m=f

in

equation

we

find for

= J\ _
7

lllax 9.3

2''* 7 + 9-H-3.5 <*V - 1


.
.

-I

A linear equation
of transformations
;

remains linear for two important classes first, when the independent variable is

x, and second, when for y we put ax e~ v be substituted an example of the latter, let y vf{x). After rejecting the factor e~ ax the in equation (9) above.

changed

to

any function of

As

>

result

is

(Pv

dv

2V

= 0.

Since this differs from the given equation only in the sign

*When there are two groups of terms, the integrals are expressible of Gauss's " Hypergeometric Series."
f If the

in

terms

second

member
if it

is

a term of the complementary function (for ex-

any integral power of x), the particular integral will ample, take the logarithmic form referred to in the foot-note on p. 346.
in this case,
is

ART.
of a,

19.]

SYSTEMS OF DIFFERENTIAL EQUATIONS.


infer
.

349

we
-|

from equation (16) that

it

has the

finite integral

Hence the complete


in

integral of equation (9)

can

be written

the form

xy

*= c x {2

ax)

+ c e~
7
. .

ax

(2 -f- ax).

Prob. 69. Integrate in series the equation

d*y -~

-j-

xy

= o.
.

(Ans.

J= ^(x- 3

y+^_

>

)+Il{x-ff +'~j*x>t

.).)

Prob. 70. Integrate in series

x*.

+ x -~
l

-f-

(x

2)y

o.

Prob.

= e" x {x~ + 1 + ix),


l

71.

Derive for the equation of Prob. 70 the integral and find its relation to those found above.

Art.
It is

19.

Systems of Differential Equations.


in Art. 12 that a

shown
its

ential equations of the first

determinate system of n differorder connecting n 1 variables

complete solution as many integral equations conthe variables and also involving n constants of intenecting 1 variables would be a The result of eliminating n gration.
single relation
in general the

has for

between the remaining two variables containing* n constants. But the elimination may also be

effected in the differential system, the result being in general an equation of the /zth order of which the equation just men-

tioned

is

two equations

For example, if there were the complete integral. of the first order connecting the variables x and

with the independent variable /, by differentiating each we should nave four equations from which to eliminate one variable, sayjj/, and its two derivatives* with respect to /, leaving
a single equation of the second order between

x and

t.

It is easy to see that the same conclusions hold if some of the given equations are of higher order, except that the order of the result will be correspondingly higher, its index being in
2 * In r general, there would be equations from which to eliminate n * n derivatives of each, that is, ( 1 quantities -f- 1) i)(

variables and

leaving a single equation of the

th order.

350
general the
tions.

DIFFERENTIAL EQUATIONS.

[Chap. VII.

tions

sum of the indices of the orders of The method is particularly applicable with constant coefficients, since we have a

the given equato linear equa-

of solution for the final result.

general method the symbolic notation, Using

the differentiations are performed simply by multiplying by the symbol D, and therefore the whole elimination is of exactly

the same form as

if

the equations were algebraic.

For

ex-

ample, the system

dy
when

dx
2t,
is

dx

dy
dt

~df"~~dt
written symbolically,

*dt^
2J9y

3*

= o,

(2D'

4)y
2

Dx =
x
y

2/,

+ (4Z>-3)* = o;
2t

whence, eliminating

2D -4 2D
which reduces to

-D
4^-3

(D-i)\2D-\-3)y =
Integrating,

V-

the particular integral being found as explained at the end of Art. 16.

by symbolic development,

The value

of

x found in like manner is x = (A' + B'ty + C'e~* - f


function, depending solely upon the determembers,* is necessarily of the same form

The complementary
minant of the
first

as that for y, but involves a

new
is

set of constants.

The

re-

lations betv/een the constants

found by substituting the

values of

x and y

in

one of the given equations, and equating

to zero in the resulting identity the coefficients of the several

terms of the complementary function. In the present example we should thus find the value of x, in terms of A, B> and C to be
y

x
the final equation
;

= (6B - 2A in

2Bty
this

\Ce-*
is

-J.

* The index of the degree


it is

D of

determinant

that of the order of

not necessarily the

sum

of the indices of the orders cf

the given equations, but cannot exceed this sum.

ART.

19.]

SYSTEMS OF DIFFERENTIAL EQUATIONS.

351

In general, the solution of a system of differential equations depends upon our ability to combine them in such a way as to form exact equations.

For example, from the dynamical

system

W~ X
where X,
Y,

'

dt*-

'

dt'-^'
and
z,

w
/,
,

Z
.

are functions

of x, y,

but not of

we form

the equation

dx ,dx
,

dy dr + -4-d dt + dtd,dz dt dt dt dt
.

Az-

dz

'

'

-., VJ Xdx + Ydy 4- Zdz. '


. '

The
that

first

member
is

is

an exact

differential,

a conservative
is,

field of force

the second

and we know that for member is also exact,

it

the differential of a function

of x, y,

and

z.

The

integral

;[()"
is

+"+>*+*
which
is

M
as the

that

first

integral of the system (1)

known

equation of energy for the unit mass.


Just as in Art. 13 an equation of the second order was regarded as equivalent to two equations of the first order, so the

system (i)in connection with the equation defining the resolved velocities forms a system of six equations of the first order, of " which system equation (2) is an " integral in the sense explained in Art.
12.

Prob. 72. Solve the equations


ear in
/.

Prob.

= dy = at as a system iinmx (Ans. x = A cos mt+B sin mt, y = A s'mmtB cosmt.) n y = e * JL z = 73. Solve the system
r,
,
.

dx

my

1-

_j_

(Ans.

nx y = Ae + Be'"*

+n -, z = - nAenx + nBe

" nx

f
_|

.) '

Prob. 74. Find for the system

-^

= x<p(x,y),
/a dy

-^~

= y<P(x,y)

first

integral independent of the function 0.

dx

352
Prob. 75.

DIFFERENTIAL EQUATIONS.

[CHAP; VIE.

The approximate
when
dy
,

of a pendulum, d?x

equations for the horizontal motion, the earth's rotation is taken into account, are

gx
are of the
x

d*y

dx

gy

show

that both

x andy
-\-

form
%

A
Art.

cos n x t

B sin n t + C cos n t + D sin nj.

20.

First Order and Degree with Three Variables.

variables x,

The equation of the first order and degree between three y and z may be written Pdx + Qdy + Rdz = o, (1)

where P,
equation

Q
is

and

are functions of x,

y and

z.

When
in Art. 4,

this

exact, P,
u, of x,

and

R
z
;

are the partial derivatives of

some function

y and

and we derive, as
d

dP = dQ
dy
d*'

sG^atf ~
dz
dy'

= &P
dz

dx

M ^

for the conditions of exactness.

In the case of two variables,

the equation is not exact integrating factors always exist;, but in this case, there is not always a factor u such that piPy

when

juQ and /jR (put in place of P, Q, and R) will satisfy all three It is easily shown that for this of the conditions (2). purpose

the relation

p (dQ
must

dR\, n (dR

8A,

(dP

dQ\

(3)'

exist between the given values of P, Q, and R. This " of equation (1).* therefore the " condition of integrability

is>

When
grated by

this condition

is

fulfilled

equation

(1)

may be

inte-

first supposing one variable, say 2, to be constant. o, and supposing the constant Thus, integrating Pdx -j- Qdy of integration C to be a function of z, we obtain the integral, so

*When

there are

more than

three variables such a condition of integra-

bility exists for

independent.
ditions.

each group of three variables, but these conditions are not all Thus with four variables there are but three independent con-

Art.

20.]

first

order and degree, three variables.


Finally,

353

far as

it

depends upon x and y.

by comparing the

total differential of this result with the given equation

we

de-

termine

dC in

terms of z and

ds,

and thence by integration the


exact

value of C.
It

may

be noticed that when certain terms of an

equation forms an exact differential, the remaining terms must It follows that if one of the variables, say z also be exact.

can be completely separated from the other two (so that in equation (i) R becomes a function of z only and Pand Q functions of x and y, but not of z) the terms Pdx -f- Qdy must be
thus rendered exact
if

the equation

is

integrable.*

For example,

zydx
is

zxdy

y*dz

= o.

we

an integrable equation. Accordingly, dividing by y*z, which notice separates the variable z from x and y, puts it in the exact form dz ydx xdy

y
of which the integral
is

o,

=y log cz.

Regarding x, y and z as coordinates of a moving point, an integrable equation restricts the point to motion upon one
of the surfaces belonging to the system of surfaces represented

by the

other words, the point (x, y, z) moves in an Let us now consider arbitrary curve drawn on such a surface. in what way equation (i) restricts the motion of a point when
integral
;

in

it is

not integrable.

The

direction cosines of a
y

moving point

are proportional to dx, dy and dz\ hence, denoting them by and n, the direction of motion of the point satisfying /,

equation

(i)

must

satisfy the condition

P/+ Qm + Rn = o.
It is

(4)

connection an auxiliary of lines represented, as explained in Art. 12, by the system simultaneous equations

convenient to consider

in this

dx

_dy _dz ~p~~Q~~R'


* In fact for this case the condition presses the exactness of Pdx -}- Qdy.
(3)

(5)
its

reduces to

last

term, which ex-

354

DIFFERENTIAL EQUATIONS.

[Chap. VII.

The
-of

this

direction cosines of a point moving in one of the lines Hence, desystem are proportional to P, Q and R.
A,
ju,

noting them by

v,

equation
-)- p-in -f-

(4)

gives

XI
for the relation

vn

=o

(6)

between the directions of two moving points, whose paths intersect, subject respectively to equation (1) and
question therefore intersect at right angles; therefore equation (1) simply restricts a point to move in a path which cuts orthogonally the lines of the auxilito equations
(5).

The paths

in

ary system.

Now,

if

there be a system of surfaces which cut the auxiliary

lines orthogonally, the restriction just

mentioned

is

completely

expressed by the requirement that the line shall lie on one of these surfaces, the line being otherwise entirely arbitrary.

This

is

the case in which equation (1)

is

integrable.*

restriction can only

when the equation is not integrable, the be expressed by two equations involving an arbitrary function. Thus if we assume in advance one such relation, we know from Art. 12 that the given equation (1)
the other hand,

On

together with the first derivative of the assumed relation forms a system admitting of solution in the form of two integrals*

Both

of these integrals will involve the

assumed function.

For

any particular value of that function we have a system of lines satisfying equation (1), and the arbitrary character of the function

makes the

solution sufficiently general to include

all lines

which

satisfy the equation/)-

Prob. 76.

Show

that the equation

is

(nx (mz ny)dx lz)dy (ly mx)dz = o and infer from the integral the character of the auxilintegrable,

*It follows
(5),

that,

equation

(3) is the

with respect to the system of lines represented by equations condition that the system shall admit of surfaces cutting

them orthogonally.

The

lines of force in

any

field

of conservative forces form

such a system, the orthogonal surfaces being the equipotential surfaces. f So too there is an arbitrary element about the path of a point when the
single equation to which it is subject is integrable, but this enters only into one of the two equations necessary to define the path.

Art

21.]

PARTIAL EQUATIONS, FIRST ORDER*


(Compare the
illustrative

355

iary lines.

example (Ans. nx

at the
Iz

end of Art.
C(ny

2.)

mz).)
x

x e (i-\-cz).) e dz o. Prob. 77. Solve yz*dx z*dy (Ans. yz Find the equation which in connection with^ - f(x) Prob. 78. forms the solution of dz aydx -J- bdy.

Prob. 79.

Show

that a general solution of

ydx
is

(x

z)(dy

dz)

given by the equations

y
(This
is

<?K*),

jy

(*

*)0'(*).

an example of

"

Monge's Solution.")

Art.

21.

Partial Differential Equations of First Order and Degree.

variables

Let x denote an unknown function of the two independent x and y, and let dz dz
partial derivatives

denote

its

a relation between one or both


is

of these derivatives "


ferential equation

and the variables


of the
first

" called a partial dif-

order!

A Value

of z in terms of

x and y which with its derivatives satisfies the equation, or a relation between x, y and z which makes z implicitly such a " function, is a particular integral." The most general equation
of this kind
If
is

called the " general integral."

only one of the derivatives, say/, occurs, the equation


differential equation. For \i y is becomes the ordinary derivative of a constant, p
;

may

be solved as an ordinary

considered as

z with respect to x therefore, if in the complete integral of the equation thus regarded we replace the constant of integration

by an

arbitrary function
all

of y,

we

shall

have a relation

which includes

particular integrals
It will

sible generality.

be found that, in

and has the greatest poslike manner, when


arbi-

both p and q are present, the general integral involves an


trary function.

We

proceed to give Lagrange's solution of the equation of

356

DIFFERENTIAL EQUATIONS.
first

[Chap.

VIL

order and degree, or " linear equation," which written in the form
the

may be
(I)

Pp+Qq = R,
y and
z,

P,

and

R
is

denoting functions of x,

y and
and
a,

z.

Let u

= a,
to

in

which u

a function of x,

a constant, be an

integral of equation (i).

and

respectively,

Taking derivatives with respect we have


.

+ du = ^

du

'

&+&t =
,

3*

du

'

and substitution

of the values of

/ and

in

equation

(i)

gives

the symmetrical relation

Consider

now

the system of simultaneous ordinary differ-

ential equations

dx __dy
Let u
tem.

~P~Q~~R
=a
be one of the integrals (see Art. 12) of
its

_dz

(3)

this sys-

Taking

total differential,
-du

dx 4'

a
dy
-dy J

~ du dz = o
:

-\-

-dx

dz

and since by equations (3) dx, dy and dz are proportional to P, Q and R, we obtain by substitution

di P+ dy
which
is

Q + d^ R
(2).

-'
It

identical with equation

follows that every

system (3) satisfies equation (1), and conversely, so that the general expression for the integrals of (3) will be
integral of the

the general integral of equation

(1). (3),

Now
v
is

let

= b be

another integral of equations


satisfies
(2).

so that

also a function

in Art. 12,

equation each of the equations u a, v

which

As
is

explained the equation

of a surface passing through a singly infinite system of lines

belonging to the doubly


tions
(3).

infinite
is

What we

require

system represented by equathe general expression for any

Art.

21.]

PARTIAL EQUATIONS, FIRST ORDER.

357

none

surface passing through lines of the system (and intersecting C is such of them). It is evident that f(u, v) =f(a, b)

an equation,* and accordingly/^, v), where / function, will be found to satisfy equation (2).
solve equation (1), we v b of the auxiliary system

is

an arbitrary

Therefore, to find two independent integrals u a,

(3),

(sometimes called Lagrange's


(4)

equations,) and then put

<p(v),

an equation which
Conversely,
(4),
it

is

evidently equally general with/(z^, v)

= o.

may

be shown that any equation of the form


rise

regarded as a primitive, gives


of

to a definite partial

differential equation

Lagrange's linear form.

For, taking

partial derivatives with

respect to the independent variables

x and y, we have

-+r

>

(. )

[-- +

-/

J,

and eliminating cf) (v) from these equations, the term containing/^ vanishes, giving the result

du 9
(5)

which

is

of the

form Pp
is

-f-

Qq

= R.jwhose

* Each line of the system

characterized by special values of a and b which

we may

call its

coordinates, and the surface passes through those lines

coordinates are connected by the perfectly arbitrary relation J\a, b) C. These values of P, Q and R are known as the " Jacobians " of the pair f of functions u, v with respect to the pairs of variables y, z\ z, x and x, y re;

spectively.

Owing

to their

analogy to the derivatives of a single function they


:

are sometimes denoted thus

358

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

As an
tion be
for

illustration, let the

given partial differential equa-

(mz lz)q ny)p -\- (fix which Lagrange's Equations are dx dy


,

= ly
dz

mx,

(6)

nx Iz mx' ly ny These equations were solved at the end of Art. integrals there found being
Ix
-f-

mz

(j\ w
J

12,

the two

my

+ nz = a

and
system

x*

-\-

y*

-\- z*

fr.

(8)

Hence
sists of

in this case the

of "

Lagrangean lines" con-

the entire system of circles having the straight line

-==for axis.

(q)

The

Ix

general integral of equation (6) is then 2 0(V -\- y* -f- ), -\- my -f- nz

(io)

which represents any surface passing through the circles just mentioned, that is, any surface of revolution of which (9) is the
axis.*

Lagrange's solution extends to the linear equation containing n independent variables. Thus the equation being

P^Lmp^L+
the auxiliary equations are

4-p*L-r
_ dxn
dz

dx

dx^

= o is

the condition that <p (a function of x, v

and

z)

is

expressible

d(x, y,

Z)

identically as a function of

u and

v, that is to say, that <p

=o

shall

be an

in-

tegral of

Pp -f- Qq R. When the equation Pdx -f- Qdy

-f-

Rdz

is

integrable (as

it

is

in

the

above example; see Prob.

may be put in the form C, represents a singly infinite system of surfaces which the Lagrangean lines cut orthogonally therefore, in this case, the general integral may be de76, Art. 20), its integral,

which

V=

fined as the general equation of the surfaces which cut orthogonally the

system

Conversely, starting with a given system V'= C, u =f(v) is the general equation of the orthogonal surfaces, if u a and v b are integrals of
C.

V=

dx I /

dx

= dy

'I hv

dz

I dz

Art.

22.]

complete and general integrals.


t

359

and

if

c 1}

?/

=c
/(

u n s= cn are independent integrals,

the most general solution

is

uu )

= o,
(Ans. xy
-\-

where /is an arbitrary function.


Prob. 80. Solve xzProb. 81. Solve (y Prob. 82. Solve (x
-f|-

yZ
-f-

=
(z

Xy

z*

= /(-].]

+ y)(p - q) =
(Ans.
z)p -\-y(z

z)p

+ x)q = x
5.

y.

(^+j)
(Ans.

log^
z(x
-j-

Prob. 83. Solve ^(^

x)q =

x=f(x+y).)
y).

y -\-

= f(xyz).)

Art.

22.

Complete and General Integrals.


in

the preceding article that an equation between three variables containing an arbitrary function gives rise to a partial differential equation of the linear form. It
follows that,

We

have seen

when

the equation

is

not linear in/ and

q,

the

general integral cannot be expressed by a single equation of the form 0(?/, 7/) o; it will, however, still be found to depend

upon a single arbitrary function. It therefore becomes necessary to consider an integral having as much generality as can be given by the presence of arbiSuch an equation is called a " complete intrary constants. " it contains two arbitrary constants {n arbitrary contegral stants in the general case of n independent variables), because
;

this

is

the

number which can be eliminated from such an equaand


its

tion, considered as a primitive,

two derived equations.


z*

For example,
a and

if

(x- ay

+ (y-dy + = k\
O,

b being regarded as arbitrary, be taken as the primitive, the derived equations are

+ zp =
of

y
1)

-j-

zq

= O,

and the elimination

a and b gives the

differential equation
,

*V + f +

of which therefore the given equation

is

a complete integral.

360

DIFFERENTIAL EQUATIONS.

[Chap. VII.

Geometrically, the complete integral represents a doubly inin this case they are spherical surfinite system of surfaces
;

the plane of xy. In general, a partial differential equation of the first order with two independent variables is of the form
faces having a given radius
in

and centers

F(x,y, z,p

q)

= o,
=
o.

(i)

and a complete integral

is

of the

form
(2)

f(x, y,

z, a, b)

In equation (1) suppose x, y and z to have special values, the equation namely, the coordinates of a special point becomes a relation between p and q. Now consider any sur-

face passing through


(1), or, as

of

which the equation

is

an integral of

we may call it, a given " integral surface " passing through A. The tangent plane to this surface at A determines values of p and q which must satisfy the relation just menConsider also those of the complete integral surfaces [equation (2)] which pass through A. They form a singly in-

tioned.

system whose tangent planes at A have values of p and q which also satisfy the relation. There is obviously among them one which has the same value of p, and therefore also
finite

the same value of

q,

as the given integral.

Thus there
at

is

one

of the complete integral surfaces

which touches

the given

follows that every integral surface (not included in the complete integral) must at every one of its points
integral surface.
It

touch a surface included


It
is

in

the complete integral.*

velope of a singly infinite


integral system.

hence evident that every integral surface is the ensystem selected from the complete
Thus,
in

the example at the beginning of this article, a right cylinder whose radius is k and whose axis lies in the plane of xy is an integral, because it is the envelope
* Values of x, y, and
z,

determining a point, together with values of p and

q,

determining the direction of a surface at that point, are said to constitute an

"element
"
plete

of surface."

The theorem shows


all

in the

sense of including

complete integral is the surface elements which satisfy the

that the

'

com-

differ-

ential equation.

The method
is

of

grouping the "consecutive" elements to form

an integral surface

to a certain extent arbitrary.

Art.

22.]

complete and general integrals.

361

of those
tegral

among the spheres represented by the complete inwhose centers are on the axis of the cylinder. If we
center of the sphere describe an arbitrary curve in we shall have the general integral in this ex-

make the

the plane of xy

ample. In general,

if

in

a and

b,

such as b

equation (2) an arbitrary relation between <p(a), be established, the envelope of the

singly infinite system of surfaces thus defined will represent the general integral. By the usual process, the equation of

the envelope

is

the result of eliminating a between the two

equations
f(x, y,
*, a, <p(a) )

= o,

a ^r

Ax

>

y>

*>

a <Ka)
>

= o.

(3)

*'

These two equations together determine a line, namely, the ultimate intersection of two consecutive surfaces." Such
"

lines are called the


tion.

characteristics

"

of the differential equaof

They

are independent of

complete integral, being integral In the surfaces which pass through them touch one another. illustrative example above they are equal circles with centers
in the plane of
"

in fact lines

any particular form along which all

the

xy and planes perpendicular

to it.*

The example
lution

also furnishes an instance of a " singular so-

analogous to those of ordinary differential equations.


"

*The characteristics are to be regarded not merely as lines, but as linear elements of surface," since they determine at each of their points the direction of the surfaces passing through them, Thus, in the illustration, they are circles regarded as great-circle elements of a sphere, or as

elements of a right

be likened to narrow hoops. They constitute in all cases a The surfaces of a complete integral system contain them triply infinite system. all, but they are differently grouped in different integral surfaces.
cylinder, and

may

If
its
is,

we

arbitrarily select a curve in space there will

in

general be at each of

points but one characteristic through which the selected curve passes; that whose tangent plane contains the tangent to the selected curve. These char-

acteristics (for all points of the curve)

the selected curve


itself

and

it is

the only one

form an integral surface passing through which passes through it unless it be

a characteristic.
is

Integral surfaces of a special kind result

when

the se-

reduced to a point. In the illustration these are the results of rotating the circle about a line parallel to the axis of z.
lected curve

362

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

For the planes z = k envelop the whole system of spheres the complete integral, and indeed all the surrepresented by
tion exists

When a singular solufaces included in the general integral. it is included in the result of eliminating a and b from equation (2) and its derivatives with respect to a and b,
that
is,

from

/='

df

df

d^=>

Zb=

(4>

but, as in the case of ordinary equations, this result clude relations which are not solutions.

may

in-

Ix

-f-

Prob. 84. Derive a differential equation from the primitive a, where /, m, n are connected by the relation my -f- nz

m*

= =

1.

Prob. 85. Show that the singular solution of the equation found in Prob. 84 represents a sphere, that the characteristics consist

of

all

general integral therefore represents touch the sphere.

the straight lines which touch this sphere, and that the all developable surfaces which

Prob. 86. Find

cos general integral above I -\-m relation between the parameters.

the integral which results from taking in the 2 2 2 6 (a constant) for the arbitrary

Art,

23.

Complete Integral for Special Forms.


integral of the partial differential equation

A complete

F(x,y, z,p,
contains two constants, a and
b as an arbitrary parameter, finite system of surfaces, of
it b.

q)
If

=o

(1)

a be regarded as fixed and

is

the equation of a singly in-

which one can be found passing

through any given point. The ordinary differential equation of this system, which will be independent of b, may be put in the form dz pdx qdy, (2)

in

which the coefficients/ and q are functions of the variables and the constant a. Now the form of equation (2) shows that these quantities are the partial derivatives of z, in an integral
of equation (1)
;

therefore they are values of

p and

q which

ART.

23.]

COMPLETE INTEGRAL FOR SPECIAL FORMS.


Conversely,
if

363
in

satisfy equation (i).

values

of/ and q

terms

of the variables and a constant a which

satisfy equation (i) are

such as to

make equation

tem

of surfaces,
if

words,

we

of a sys(2) the differential equation In other these surfaces will be integrals. can find values of p and q containing a constant a
(1)

which

satisfy equation

and

make dz

= pdx

-f-

qdy

inte-

grate, we can obtain by


gral,

direct integration a

complete

inte-

the integration introducing a second constant.

There are certain forms of equations for which such values p and q are easily found. In particular there are forms in which p and q admit of constant values, and these obviously
of

make equation

(2) integrable.

Thus,

if

the equation contains

/ and q only, being of the form


F(P,
q)

= O,

(3)

we may put/
Equation

= a and

b,

provided
b)

F(a,
(2)

= o.
-f-

(4)

thus becomes

dz

= adx

bdy,

whence we have the complete z = ax


in

integral

+ by +

c,

(5)
(4) so that
a,

which a and b are connected by the relation and c are equivalent 'to two arbitrary constants.
In the next place,
if

the equation

is

of the form
(6)
10,

*=P* + qy+f(p,q),
which
of
is

analogous to Clairaut's form, Art.


q are again admissible
z
if

constant values

p and
this

they satisfy
9

= az + by+f(a
c,

b\

(7)
this equation is

and

is itself

the complete integral.

For

of the

form z

= ax + by +

tions between the three constants.

and expresses in itself the relaProblem 84 of the preced-

ing article

is

an example of

this form.

In the third place, suppose the equation to be of the form


F(z,p,q)

= o,

(8)

364
in which neither
~qz= ap,

DIFFERENTIAL EQUATIONS.
appears explicitly. If be a function of z determined from
F(z, p, ap)

[CHAP. VII,

x nor y

we assume

will

= o,

say

<p(z).
-f-

(9)

Then dz

pdx -|-

qdy = o becomes dz = <p(z)(dx

#^), which

is

integrable, giving the complete integral

that in which, while z does not explicitly occur, it is possible to separate x and p from y and ^, thus putting the equation in the form

'+*=/&+*
fourth case
is

(I0)

/,(*,/) =/.(?,
If

9)-

(")
to a con-

we assume each member of this equation equal stant a, we may determine/ and q in the forms
and dz

/ = (pix, a), q = {y, a). = />^r + ^y takes an integrable form giving 8 = f^** a dx + f^y* a dy + b
(p t
^
^
-

(12)

(*3)

frequently possible to reduce a given equation by transformation of the variables to one of the four forms considered
It is

in this article.*

For example, the equation

x^p*

-\-

y* q

= z*

may

be written
\* -dx)

\z

dy'

general method, due to Charpit, of finding a proper value oip consists of establishing, by means of the condition of integrability, a linear partial differential equation for/, of

*The

which we need only a particular

integral.

This

may

be any value of / taken from the auxiliary equations employed in Lagrange's process. See Boole, Differential Equations (London 1865), p. 336 also Forsyth, Differential Equations (London 1885), p. 316, in which the auxiliary equa;

tions are deduced in a

more general and symmetrical form, involving both p

and
as

q.

These equations are


z,

in the

in fact the equations of the characteristics regarded concluding note to the preceding article. Denoting the partial deriva-

tives of F(x, y,

p, q)

by X, Y,
dy

Z,

P, Q, they are

dx

~P

dz

~Q

Pp+Qq

dp

dq

~XTZp
Any

-~~YTz~q
;

See Jordan's Cours ^'Analyse


ential Equations

(Paris, 1887), vol. in, p. 318

(New York,
q,

1889), p. 300.

relation involving

Johnson's Differone or both


j>

he quantities p and

combined with

F= o,

will furnish

proper values of

ART.

24.]

PARTIAL EQUATIONS, SECOND ORDER.

365
}

whence, putting x' i, which p' -\-q'*


%

=
is

log x, y' =s log y, z' log z it becomes. of the form F(p\ q') o, equation (3).

Hence the
it

may

integral is given therefore be written


z'

by equation
or

(5)

when
c,

a* -f- b*

= I?

;r'

cos

+y sin a-\a
i/Sin a^

and restoring

;r,

j,

and

z,

that of the given equation


^-j^cos

is

2 Prob. 87. Find a complete integral for/

(Ans. s

Prob. 88. Find the singular

0* = = # sec a + solution of z = px

1.
jy

tan

or -{-

.)

-\-

qy + pq.

(Ans. z
Prob. 89. Solve by transformation ^

.#y.)

2^'. (Ans. ^

## 4-

y+
)i

b.y

Prob. 90. Solve z(p' Prob. 91.

(Ans. *l

= (* +

y.

)l

that the solution given for the form F(z,p, q) a. o is a singular represents cylindrical surfaces, and that F(z, o, o) solution.

Show

+ (^ + a + ^.) =
=

Prob. 92.

complete integrals of pq
(Ans. 2*

Deduce by the method quoted = px + qy.

in the foot-note

two

=(

+ ^) +A and 5 = a? +j> ^(* +


5

<*)

+ &)

Art.

24.

Partial Equations of Second Order.


in

We

have seen

solution of a partial differential equation of the


;

the preceding articles that the general first order de-

pends upon an arbitrary function although it is only when the equation is linear in / and q that it is expressible by a But in the case of higher orders no general single equation. account can be given of the nature of a solution. Moreover,
consider the equations derivable from a primitive conarbitrary functions, there is no correspondence between taining their number and the order of the equation. For example, if
and
q.

when we

Sometimes several such

relations are readily found

for example,

for

the equation

zpq we thus obtain the two complete integrals z = (y + a)(x + and 4* =fe+ ay + /?Y.
d)

36*6

DIFFERENTIAL EQUATIONS.

[CHAP. VII.

the primitive with two independent variables contains two arbitrary functions, it is not generally possible to eliminate them

and

their derivatives
first

equations of the
let us

from the primitive and its two derived and three of the second order.

Instead of a primitive containing


take an equation of the
first

two arbitrary

functions,

arbitrary function.

This

may

order containing a single be put in the form u (p{v)

u and v now denoting known functions of x, y, z, p, and q. eliminated from the two derived equations <p'(v) may now be
as in Art. 21.

Denoting the second derivatives of z by


r

~s*

a^

'

_ _av_

a*f

~d*d/

ay
V,
z,

the result

is

found to be of the form

Rr + Ss+ Tt+U(rt-s*)=
in

(i)

which R,

S, T> U,

and

are functions

of x, y,

p,

and

q.

reference to the differential equation of the second order " the equation ti intermediate equation of (p(v) is called an

With

=
:

the

first

order

"
it is

analogous to the
It

first

integral of an ordi-

nary equation of the second order.

follows that an inter-

mediate equation cannot exist unless the equation is of the form (i); moreover, there are two other conditions which must exist between the functions R, 5, T, and U.
In tained

some simple cases an intermediate equation can be obby direct integration. Thus, if the equation contains

derivatives with respect to one only of the variables, it may be treated as an ordinary differential equation of the second order,

the constants being replaced by arbitrary functions of the other variable. Given, for example, the equation xr p xy,

which

may

be written

xdp

pdx

= xy dx.
x when
divided by x*,

This becomes exact with reference to

and gives the intermediate equation

p=yx\ogx + x<p(y).

A second

integration (and change in the form of the arbitrary

function) gives the general integral

= \yx* log x + x*<p(y) + ip(y).

Art.

24.]

partial equations, second order,

367

Again, the equation p -\- r -\- s gives the intermediate equation


'

is

already exact, and

*+/+? = * + 0C?)
The
auxiliary equations* are
,

which

is

of Lagrange's form.

dx
of which the
first

= dy = =

,,

second,

its

a, and eliminating y gives x the form integral is of

x from

the

a-\- <p(y) -\-i~*b.


integral

Hence, putting b

= tpia), we have for the final z = x + 0O) + e ''i>(x - y),


is

in

which a further change

made

in the

form of the arbitrary

function 0.
Prob, 93. Solve
t

= e* + ey
= y\

(Ans. z

=y(e y

- e x + (p(x) + e y tp(x).)
)

Prob. 94. Solve r

+/

(Ans. z

= \og[e xy ip(y)

e'

xy

]-\- tp(y).)

Prob. 95. Solve y\s t) (Ans. z


Prob. 96. Solve ps

= x. = (x +y) logy + <p(x) + ij>(x +y).) (Ans. ^ = (p(y) + ^(2).) qr = o. =

Prob. 97. Show that Monge's equations (see foot-note) give for Prob. 96 the intermediate integral p <p{z) and hence derive the
solution.
* In Monge's method (for which the reader must be referred to the complete treatises) of finding an intermediate integral of

Rr+ Ss+
when one
exists, the auxiliary

Tt

= V =
Vdx dy

equations

Rdy*
are established.

Sdy dx

Tdx*

= o, = pdx

Rdp dy + Tdq dx
-f-

These, in connection with


dz
qdy,

form an incomplete system of ordinary differential equations, between the five But if it is possible to obtain two integrals of the variables x, y, z, p, and q-. in the form u = a, v = />, u = <p(v) will be the intermediate integral. system

The

first

of the auxiliary equations


If

is

a quadratic giving two values for the ratio


will

dy.dx.

these are distinct, and an intermediate integral can be found, for

ach, the values of

jntegrable,

p and q determined from them and give the general integral at once.

make dz =p>dx-{-qdy

368

DIFFERENTIAL EQUATIONS.

[CHAP.

VIL

the intermediate integral/

Prob. 98. Derive by Monge's method for fr o 2pqs -\- p*t q <p(z), and thence the general integral.. (Ans. y x(p(z) ip(z).)

Art.

25.

Linear Partial Differential Equations.

variable and

Equations which are linear with respect to the dependent its partial derivatives may be treated by a method
in the case of

analogous to that employed


equations. ent variables

ordinary differential

We

shall consider only the case of

two independ-

x and

y,

and put

d*

dy

DD

% so that the higher derivatives are denoted by the symbols , s % D' etc. ', Supposing further that the coefficients are
,

constants, the equation

may

be written

in

the form
(1)

f{D,D')z
in

= F(x,y),

denotes an algebraic function, or polynomial, of which the degree corresponds to the order of the differential

which

Understanding by an "integral" of this equation an explicit value of z in terms of x and y, it is obvious, as inArt. 15, that the sum of a particular integral and the general
equation.
integral of

/(>, D')z
will constitute
is,

=o
(1).

(2)

an equally general solution of equation

It

however, only when f{D, D') is a homogeneous function of and D' that we can obtain a solution of equation (2) containing " 11 complearbitrary functions,* which solution is also the

mentary function

"

for equation (1).

Suppose then the equation to be of the form

A
and
*
let
It is

B +A

us assume
assumed

+ --- +A "B=' z = <p(y + mx),

(3)

(4}

that such a solution constitutes the general integral of an

equation of the th order; for a primitive containing more than n independent of the arbitrary functions cannot give rise by their elimination to an equation
nib. order.

Art.

25.]

linear partial equations.


a.

369

where mis

constant to be determined.
-\-

From

equation

(4),
'

= mcp\y mx) and D'z = <p\ y mx), so that Dz = mD z, IPz = m*D'*z, DD'z = mD'^y, etc. Substituting in equation (3)
Dz
-\-

and rejecting the factor D' nz or

{n

\y
.

-\-

mx),

we have
o
(5)

Ajn -f A m
n
A

n~ l

-f A n =
i}
.

for the determination of m.

If

m m
xi

mn are distinct roots

of this equation,

^
is

0i( y

+
=

,*)

+0 (7 + ^r^) +
a

+ 0O + ***)
-

(6)

the general integral of equation

(3).
_a

For example, the general integral of


found to be z
cp(y
-\-

=o

is

thus

x)

~\-

ip(y

x).

Any expression
;

of the

form
it

Axy + Bx -f- Cy -f- Z> is


expressible as the

a particular integral

accordingly

is

sum

of certain functions of

x + y and

y respectively. The homogeneous equation


form
x .

(3)

may now be
{D

written sym-

bolically in the

(D
in

m D'\D - mJT)

m nD')z = o,

(7)

which the several factors correspond to the several terms of the general integral. If two of the roots of equation (5) are equal, say, to m lt the corresponding terms in equation (6) are
equivalent to a single arbitrary function.
integral

To form

the general

we need an

integral of

{D
in addition to 4>{y

- m DJz = o
x

(8)

+ m^x).
x

This

will in fact

be the solution of
(9)

(D-m D')z=
for,
if

we operate with

rty + ms); D' upon both members


Writing equation

of this

equation,

we

obtain equation

(8).

(9) in

the

form

m q=
x

<p(y

+ mx\
dz
-\-

Lagrange's equations are

dy _

m
giving the integrals^ z

0( y

m^x)'
-j- b.

mx =
x

a, z = x(p(a)

Hence the
(10)

integral of equation (9)

is
x
x

= x<p(y + m x) + tp(y + m x),

370

DIFFERENTIAL EQUATIONS.
<fi

[CHAP. VII.

and regarding
If

also as arbitrary, these are the

two independ-

ent terms corresponding to the pair of equal roots.


has a pair of imaginary roots the corresponding terms of the integral take the form

equation

(5)

m = /z

iv*

<p(y -\-pix -\-ivx)

-j-

ip{y

-\-

^.x

ivx),

which when
terms.
If

and

tp

we

restrict

functions contain imaginary ourselves to real integrals we cannot


are real

now

say that there are two radically distinct classes of intebut if any real function of y -j- jax -f- ivx be put in the grals
;

form X-\-iY, either of the


integral of the equation.
2

real functions

X or

Fwill be an

Given, for

example, the equation

a^

3 a *

of

which the general integral


*

is

0( y

+ ix + *P(y - ix
)

to obtain a real integral take either the real or the coefficient


of the imaginary part of any real form of cp{y -\- ix). Thus, if # we find ey cos^r and ey sin^r, each of which is an (j){t)

integral (see Chap. VI, p. 245).

corresponding case of ordinary equations, the particular integral of equation (1) may be made to depend upon the solution of linear equations of the first order. The
in the

As

inverse symbol j:

jyF(*> y)

m the equation
p

corresponding

to equation (14), Art. 16, denotes the value of z in

(D
For

mD')z

F(x, y)

or

mq

= F[x, y).
b
x

(11)

this equation

Lagrange's auxiliary equations give


z

+ wix =

a,

j F{x, a
is

mx)dx -\-

= F (x, a) -f- &,


(12)

and the general


z

integral
x

= F {x,y + mx) + (p(y -f mx).


is

The

first

term, which

the particular integral,

may

thereinte-

fore be found

by subtracting

mx

from

in

F(x,

y),

Art.

25.]

linear partial equations.

371

grating with respect to


result.*

x and
f

then adding

mx

to

in

the

For

certain forms of

F(x

y) there exist

more expeditious

methods, of which we shall here only notice that which applies to the form F{ax -\- by). Since DF{ax -f- by) aF\ax -f- by)

and D'F(ax

-f-

by) ==
is

wheny~(Z>, D')
t =

bF\ax -f- by), it is readily inferred that, homogeneous function of the nth degree,

f{D, D')F{ax
That
is,

+ by) =f(a, b)F*\ax +


the operation of f(D,

by).
r

(13)

if

ax

-f- by,

equivalent to multiplication by rivative, the final result being

f(a

b)

D ) on F(t) is and taking the nth. det.

still

a function of

It follows

that, conversely, the operation of the inverse symbol upon a function of / is equivalent to dividing by /(a, b) and integrating

n times.

Thus,

/ah) F{ax + by) =xh)ffWhen ax -f- by


equation
(5),
is

/**
is

(I4)

a multiple of y
fails

-f-

mx
t

where w,

a root of

this

method
is

ing symbolic of which the solution

factor, giving rise to

with respect to the correspondan equation of the form (9),


equation
(10).

given
d*z

in

Given, for ex-

ample, the equation d*z d*z


.

or

(D

a^+a^Jy- ay^ - D') (D + 2D')z = sin


function
is

Sm{x

-^ +Sm ^+^ - y) + (x + (x
.

sin

y).

<p(y -\- x) -\- tp(y part of the particular integral arising from sin (x

The complementary

2x).
y), in

The
which
aris-

i~b

1, is

/
this

/sin tdf
theorem
y)
is

=
-

- sin (>

y).

That

* The symbolic form of

D -m>' F(x
-

'

= emxD

'f

mxD

'^(x, y)dx
ev*xD'

corresponding to equation (13), Art. 16.


addition of

The symbol

here indicates the

mx

to

in

the operand.

Accordingly, using the expanded form

of the symbol,

***D>FKy)

(i+mx^

^~

-f

. .

.)

F(y)

= F(y + mx),

the symbolic expression of Taylor's

Theorem.

372

DIFFERENTIAL EQUATIONS.
is

[Chap. VII.

ing from sin \x -\-y) which

of the form of a

term
^

in the

com-

plementary function
tion (10)
is

is

n_ n

COS

+^' w licl1 b y equaintegral of

\x

cos ix-\-y).
is
)

Hence the general


{*

the given equation

<P(y

+ *) + $(y - 2x +i sin
=
is still

y)

i * cos
t

(*

+ fi-

o the symbol f(D, D') though the equation f{D, D')z and >', can be separated with respect to not homogeneous
ll in

into factors, the integral

the

to the several symbolic factors.

sum The

of those corresponding

integral of a factor of
;

the

first

degree

is

found by Lagrange's process mD a)z o {D


,

thus that of
(15)

is

z = e ax (p(y

-j-

mx).

(16)

But in the general case it is not possible to express the Let us, howsolution in a form involving arbitrary functions.
ever,

assume

z=cehx+ky
y

(17)

where c, h, and k are constants. and D'ehx ^ ky =kehx lrky substitution xJrky = o. Hence we have a solution of the form cf(h, k)^

= h^x + Since Z)e in f(D, D')z = o gives


hx

+ ky

ky

(17)

whenever

//

and k

satisfy the relation

AK
c

k)

= o,
obvious that

(18)

being altogether arbitrary. write the more general solution

It is

we may

also

z=
where k
F{Ji) is

2ce,lx + F^y

(19)
(18),

derived from equation

and

and h admit

of an infinite variety of arbitrary values.

fhx

+ FWy

Again, since the difference of any two terms of the form w t h different values of h is included in expression
j

(19),

we
is

infer that the derivative of this expression with respect

to

//

also an

integral,

and

in like

manner the second and

higher derivatives are integrals.

For example,

if

the equation is d*z dz

l?~d~y^'

Art.

25.]

LINEAR PARTIAL equations.


is

373
classes of in-

for which equation (18)


tegrals of the forms

I?

o,

we have

** + h \(x

+ 2hyf + 2y)\
//

f* +

*"'[(*

+ 2/^) + 6>(^r + 2hy)\


3

In particular, putting
cxx c,(x
t

= owe obtain the algebraic integrals

2y), c 3 (x

+ 6xy), etc.

The

variable coefficients

solution of a linear partial differential equation with may sometimes be effected by a change of

the independent variables as illustrated in some of the examples below.


Prob. 99. Show that if m terms of the integral are x*<p(y
1

is

a triple root the


x

corresponding

+ m x) ~V *^(jH- ^)+ X(y+^i x)'


tfz
tfz
2
2

Prob. 100. Solve

ci Prob. 101. Solve

Prob.

+2 + =? = 0(*) + ip(x + j) + #(* +jf) log*.) (Ans. z n 102. Solve (Z> + 5_>Z>' + 6D )z = {y - 2x)"\ = <p(y 2x) + tp(y 3*) + x log 2*).) (Ans. z
1

2^^ dx 9^

d*z

^
2 2

3^-^ dxdy
d'*

= o.
I
.

dy

JM

(jy

d*z

Prob. 103. Solve

d'z

dz
-f-

dx

5-^dxdy

2=0.

dy

Prob. 104. Show that for an equation of the form (15) the solution given by equation (19) is equivalent to equation (16). Prob. 105. e Solve
,

-.

d*z

1
3

dz

d*z
3
.

r~i * 3^ a

^ r 9#

-5^3

/ dy

dz :r- by transposiy* dy
1
.

tion to the independent variables x* andjv

Prob. 10 6. Solve

^g +

^+/g- =

o.

374

GRASSxMANN's space analysis.

[Chap. VIII*

Chapter VIII.

GRASSMANN'S SPACE ANALYSIS.


By Edward W. Hyde,
Professor of Mathematics in the University of Cincinnati.

Art.

1.

Explanations and Definitions.

algebra with which the student is already familiar deals directly with only one quality of the various geometric and mechanical entities, such as lines, forces, etc., namely, with
their magnitude.

The

Such questions as How much? How far? long by an algebraic operation or series of operations. Questions of direction and position are dealt with indirectly by means of systems of coordinates of various kinds.

How

are answered

In this chapter an algebra* will be developed which deals directly with the three qualities of geometric and mechanical
quantities, viz., magnitude, position,
ric

and direction.

A geomet-

quantity may possess one, two, or all three of these properties simultaneously; thus a straight line of given length has
three, while a point has only one.

all

The geometric quantities with which we are to be concerned are the point, the straight line, the plane, the vector, and the
plane-vector.

When
will

" the word " line

is

used by

itself,

" a " straight line

definite length will


* The algebra of

be always intended. portion of a given straight line of " be called a " sect though when the length
;

this

chapter

is

a particular case of the very general and

comprehensive theory developed by Hermann Grassmann, and published by him in 1844 under the title " Die lineale Ausdehnungslehre, ein neuer Zweig der Mathematik." He published also a second treatise on the subject in 1862.

Art.

2.]

explanations and definitions.


is

375
line will fre-

of the sect

a matter of indifference, the

word

quently be used instead. Similarly, a definite area of a given " plane will be called a plane-sect."
a point recede to infinity, it has no longer any significance as regards position, but still indicates a direction, since all lines
If

finite points, and also through this point at are parallel. Similarly, a line wholly at infinity fixes infinity, a plane direction, that is, all planes passing through finite

passing through

points,

and

also

through this line at

infinity, are parallel.

Thus

a point and line at infinity are respectively equivalent to a line direction and a plane direction.

quantity possessing magnitude only will be termed a " scalar " quantity. Such are the ordinary subjects of algebraic
analysis, a, x, sin
0,

logz,

etc.,

and they

may

evidently be

in-

trinsically either positive or negative.

The

letter
will

T prefixed

to a letter denoting
its

some geometric

be used to designate quantity magnitude, always positive. Thus, if sect, then TL is the length of Z, and
portion of a geometric quantity be called its " unit," and will
letter

absolute or numerical

be a

TP is

and Pa. planethe area of P. That


sect,

whose magnitude is unity will be indicated by prefixing the

thus

UL =
.

unit of

= sect

one unit long on

line Z-.*

Hence we have TL
Art.
2.

UL

=s L.

Sum and Difference of Two


we

Points.

In geometric addition and subtraction


,

shall use the or-

=, but with modified significance, as will dinary symbols -{-, in the development of the subject. appear
Every mathematical, or other, theory rests on certain fundamental assumptions, the justification for these assumptions
* The word "scalar" and the use of the
introduced by Hamilton in his Quaternions.
letters T and U, as above, were T stands for tensor, i.e., stretcher,

and

TL

is

the factor that stretches


is

UL

into L.
|

The notation

for absolute

magnitude

not used, because the sign

has been appropriated by Grassmann

to another use.

376
lying in

grassmann's space analysis.


the
its

[Chap. VIII.

harmony and reasonableness

of

the resulting

theory, and

accordance with the ascertained facts of nature.

Our

first

commutative laws hold


tion, that
is,

assumption, then, will be that the associative and for geometric addition and subtrac-

whatever A, B,
)

C may

represent,

we have

+ B + C= (A + + C^zA +(B + C) = A + C+B = (A + C) + B.


We
shall also

assume that we always have A A = o, and that the same quantity may be added to or subtracted from both sides of an equation without affecting the equality.

Now

let/,

,/> 2

be two points, and consider the equation

A +A -A =A + (A -A) = AIn this form

(0
in

we have an

identity.

Write

it,

however,
>

the

form

A -A +A = (A -A) +A = A
appears that/,
it.

(2)

and

it

/,

is

an operator that changes/, into


this

^> 2 by being added to

Conceive

change oi p into/ 2 to
Y x
;

take place along the straight line through p and /> 2 then the operation is that of moving a point through a definite length
or distance in a definite direction, namely, from/, to/> 2 This has been called by Hamilton " a vector," * that is, a operator
.

carrier,

because

it

carries/, rectilinearly to p^.

to

it

the

name

Strecke, and

some

writers

now

Grassmann gives use the word

" stroke " in the

same
x

sense.

Again, p^

is

difference that can exist

the difference of two points, and the only between them is that of position, i.e.

a certain distance in a certain direction.

as

Hence we may regard / -/, as a the operator which moves p over


2
x

directed length, and also


this length in this direc-

tion.

Writing^

/,

e,

equation

(2)

becomes
(3)

A + e=A* See the


first

of Hamilton's

discussion of equation (2) will be found.

Lectures on Quaternions, where a very Also Grassmann (1862), Art. 227.

full

ART.

2.]

SUM AND DIFFERENCE OF TWO POINTS.


of a point

377

Thus the sum


the
first

by the length

of

and a vector is a point distant from the vector and in its direction.
~~'

Since

A A
is
>

~"

(A

A)>

**

of a vector
direction.
If

a vector of the

same length

appears that the negative in the opposite

A A
is

or

A = A> A must

coincide with

because

there

now no

difference

between the two points.

The

A -A

question arises as to what, if any, effect the operator should have on any other point A> that is, what is the

value of the expression

A A+A?
it

We will

assume that

is

some point A> so that we have


(4)

A-A+A=A,
^

A-A=A-Aamount and
This implies that the transference from direction as that from

A to A

1S

the

same

in

to

A>

that

is,

that

A> A> A> A

are the

order.

four corners of a parallelogram taken in Thus equal vectors have the same

length and direction, and, conversely, vectors having the same length and direction are equal.

Note that

parallel vectors of equal length are not neces-

sarily equal, for their directions

may

be opposite.

Equation

(4)

may

also be written

A+A=A+A,
so that, whatever meaning
points,
if

(5)

may be assigned to the sum of two we are to be consistent with assumptions already

of either pair of opposite cornerof a parallelogram equal to the sum of the other pair. points The sum cannot therefore depend on the actual distances

made, we must have the sum

apart of the points forming the pairs, for the ratio of these two distances may be made as large or as small as we please.
If
is

bea scalar quantity, ne will denote that the operation to be performed n times on a point to which ne is added,
is,

that

the point will be

moved n times

the length of e

hence

378

grassmann's space analysis.


is

[Chap.

VIIL

ne

a vector n times as long as

e,

and having the same or the


n.

opposite direction according to the sign of In the figure above, let

A-A =
.Then
e,
2

e i>

A A=e
/,

A-A =
-p =
%

e 3>

A A=e

+ e = A -A + A -A =A -A +A ~A = A~ A =
by Eq.
if

*.

(5)

since,

4,

A = A T Ae,

Also,

ea

=/,

e4

(6)

two vectors are drawn outwards from a point, and Hence, the parallelogram of which these are two adjacent sides is completed, then the two diagonals of this parallelogram will represent respectively the sum and difference of the two vectors, the sum being that diagonal which passes through the origin of the two vectors, and the difference that which passes through
their extremities.*

Again, hence the

A -p
sum
if

+/,

-p +A
t

-/,
zero.
;/

=o=

e,

+ + (- e
e,

t) ;

of three vectors represented


is

by the

sides of a

triangle taken around in order

Similarly,

p p9f
it

,p n be any

points whatever taken as

corners of a closed polygon,


that

we

shall

have
;

(A- A)+(A- A)+(A- A)+


is,

+(A - A-,)+(A-A)= o
By
is is

the

sum

of vectors represented
is

order about the polygon

zero.

by the sides taken in "taken in order" is not


must be observed

meant that any


in

particular order of the points

forming the polygon, which simply that, when the polygon

evidently unnecessary, but formed, the vectors will be

the operators that will move a point from the starting position along the successive sides back to this position again, so that

the final distance from the starting-point will be nothing.

Art.

3.

Sum of Two Weighted


**,A+ wtA>
15.

Points.!
scalars,

Consider the sum


that
is,

in

which m and m^ are


x

numbers, positive or negative, and


* Grassmann (1844),
f

p p
it

are points.

Grassmann

(1844),

95,

and

(1862), Art. 227.

Art.

3.]

sum of two weighted points.

379

The

scalars

and

will
i

be regarded as values or weights


2
.

assigned to the points

p and/

When any

weight

is

of unit

value the figure I will be omitted, so that p means I/, and is called a unit point. Occasionally, however, a letter may be

used to denote a point whose weight

is

not unity.
consider the weights

To

assist his thinking,

the reader

may

initially as like or unlike parallel forces acting at the points.

we

shall

In order to arrive at a meaning for the above expression make two reasonable assumptions, which will prove to
viz., first,

be consistent with those already made,


a point, and second, that
of the
its

that the

sum

is

weight

is

the
this

sum

of the weights

two given

points.

Denoting

sum-point by/,

we
(7>

write

*i

A+
m (p
l l

.A

Transposing, we have

= m + **t)A / p) = m (p
(
i

a ),

or
,

a -/ _/-a
Both members of
(8) are vectors,

8V

and, being equal, they must r

by Art.

4,

be

parallel.
2
.

This requires that/ shall be collinear

with /, and

Also, since /,

/ and /

are vectors

whose

to

lengths are respectively the distances from/, to p and from/ . / 2 it follows that these distances are in the ratio of m^ to
,

Hence, / is a point on the line /,/ 2 whose distances from/, and / 2 are inversely proportional to the weights of these points.

We shall call If m and m^


x

/ the mean point


but
if

of the

are both positive, (8)


;

two weighted shows that/ must


negative, let

points.
lie

be-

tween /, and/ 3

one, say m^,

is

=m^
(9>

Thus
*i(A
7)

**/(A

/).

and / is on the same side of each point, that is, its direction from each point is the same. Also, since its distances from the

two points are inversely as their weights, / must be the point whose weight is greatest.

nearest,

380
Case when

grassmann's space analysis.

[Chap. VIII.

-|-

dition equations (7)

m^ = o, or m % = and (8) become


f

m
A)

*
x

With

this

con-

0*iA+*,A = * i(/i
and

/A =//

= o-A

(10)

(")

Thus/

the same direction from each point, that is, not between them, and yet is equidistant from them. This reis

in

which evidently

p iy quires either that the two points shall coincide, that is, p % satisfies (10) and (11); or else, p and/a being
i

Thus different points, that/> shall be at an infinite distance. the sum is in this case a point of zero weight at infinity/)Eq.
(10)

shows that a zero point


Art. 2 that
let

at infinity

is

equivalent to a
1.

vector, or directed quantity, as stated in Art.

It

has been

shown

in
;

coincide
general,

/, us consider the equality of weighted points in Hence, by (7), there is found i p^. p say
*
x

A =A

tne condition that

p and

=m
l

m iPi m
l

~~
;

since

nence since / cannot be zero, m *P% = m ~ m ^)P = o, or m = m and therefore m (p p = o, or, Therefore, any /, =p r m^o, p p = 0, that
(
i
>

u)

is,

if

two points are equal,

their weights
is,

positions identical, that

they are the

must be the same and same point.


difference of the

their

Exercise

1.

To
3/,

find

the
:

sum and

two

weighted points

and pt

3A
2
p"
1

+ A = 4A
3 _
~~*j

3A

- A = 2/,
as

and the mean points are


*f
B Pi

shown

in

T
ip

the figure.

The

reciprocals of the

2p>
viz.,
-J,
,

from p 9 distances of p, p v and are in arithmetical progression, hence the points |,


,

p,

form a harmonic range.


Exercise
*Grassmann
I

2.

Given a circular disk with a circular disk of

(1862), Art. 222.

Compare

the case of the resultant

of

unlike parallel forces of equal

magnitude.

'

Art.

4.]

sum of two weighted points.


;

381

half

its

radius removed, as in the figure

to find the centroid

of the remaining portion.

Take/,
then

at

of small circle,

A = |(A
I
:

at center center of large circle, at the point of contact and The areas of the two cir-f- A)-

t$

cles are as
if

call

them

and

4.

Then

it is

as
1

P = [4A - i(A + A)] + 3 = (7A - A) + 6.


Prob.
1.

there were a weight 4 at

A> and a weight

at

hence

Show

that

A> A>

m iP\ + w aA>

anc

^,A

~~

w aA

are

four points forming a harmonic range.

Prob.
disk
;

2.

An

inscribed right-angled triangle

is

cut from a circular


is

show that the centroid of the remainder of the disk


(37T
2 sin

at the

point

2a)

A sin 2a
of the triangle,

$(7t
if

sin 20L)

is

the center of the circle,


its

p % the opposite vertex

and a one of

angles.

Art.

4.

Sum of any Number of

Points.

As
points

in

the last article


is

whose weight
;

equal to the
n

we assume the sum to be a point sum of the weights of the given


n
(
1

thus,

2mp = p2m.
n

2)

Let

be some fixed point, and subtract


;

e2m
1

from both

sides of (12)

thus

we have
e)

2m(p
1

= (/
n

e)2m, 1

(13)

an equation which gives a simple construction for


If

2m = o,

then

m =
1

2m, and

2mp =

m,p,

+ 2mp = m\

A-~

2mp

2m

J>

(i4>

382

grassmann's space analysis.

[Chap. VIII.

so that the sum becomes the difference of two unit points, or a vector whose direction is parallel to the line joining p with
x

the mean of
length
is

the other points of the system, and whose times the distance between these points. Since
all

designated as p if it follows that the line joining any point of the system to the mean of

any point

of the system

may be

all

the others

is

parallel to

any other such

line.

If

2mp =
i

o,

equation (14) shows that/j is the mean of all the other points of the system, and, since any one of the points may be

taken as
others.

plt any

point of the system

is

the

mean

of all the

Let n

3 in (12)

and

(13);

then
2

+ W *A + m%P% = Oi + ^ + (15) *i(A - ')+ mlP~ + *(P -')=(**i+* +in )(p--e), (16) and/ on the line joining the point m p + m p with/,, and therefore inside the triangle AAA ^ tne ms are a ^ positive.
**iA
f*]}p,

e)

is

If

be negative and numerically

less

than

-\-

then/

have passed across the line p^p 2 to the outside of the triIf m and m^ are negative and their sum numerically angle.
will
x

less

than
%

m
,

s,

then

will

have passed outside the triangle


z

have crossed p^p s and p p v through p e must evidently always be in the plane p p^p v
i.e., it

will

The

point

As

a numerical example

let

m = 3,
x

;/z

= 4, m =
3

5,

so

that (16) becomes

p-e = f (p -e)+ 2(A - e) - |(A s

')>

Now, p p

since e
3

= |(A

in the figure.

may be any point whatever, put e=p then p ) -f- 2(p p ), and the construction is shown p -p = |(p /,), and J p = 2{p% /,).
a
;

As another example take/ = 4/,


(13),

-f"

5A ~ 2A

^A>

or,

by

making

e=p P -A = 4(A -A) + 5(A -A) - 2(A - A) = A - A+A A +/ - A-

Art.

4.]

sum of any number of points.

383

When any number of geometric quantities can be connected with each other by an equation of the form 2m/> = o, in which the m's are finite and different from zero, then they arc said to
be mutually dependent, that is, any one can be expressed in terms of the others. If no such relation can exist between the

/ /

*
quantities, they are independent.

We

obtain from what has

preceded the following conditions:

That two points

shall concide,

**iA

+
be

.A

= 0= o.
o.

(17)

That three points

shall

collinear,
*./.

A+ A+
That four points
f*xPx
It

(18)

shall

be coplanar,

+ **%Px +

+ tn p =
K h

(19)

follows that three non-collinear points cannot be connected by an equation like (18) unless each coefficient is

separately zero. Similarly four non-coplanar points cannot be connected by an equation like (19) unless each coefficient is
separately zero.

The
trated.

significance of these statements will be presently illus-

The
vectors
:

following are corresponding equations of condition for


shall

That two vectors

be

parallel,
9 a

*ii

+*e =

o.

(20)

384

GRASSMANN

SPACE ANALYSIS.

[Chap.

Vim

That three vectors

shall

be parallel to one plane,


n,e,

*ii

+ne =
3 z

o. 2,

(2i>
or

These conditions follow from the results of Art. equations (17) and (18) by regarding the e's as points If in addition to (21) we have
n
t

from

at infinity..

+ +n =
n,
3
.

o,
if

(22) radiating from a point,.

the extremities of the three vectors,


will ex

be collinear
e9 =3
e,
,

for, let e

e3

e%
*,(',

e%

=e
'.)

e%
*.('.

e%

=
'.)

be four points so taken that then (21) becomes ,


;

+
x

+ "Ae* 3

'.)

= O,

or

by

(22)

ex

+ n^ + n = o,
es

which by
It

(18) requires el9 e^, ez to

be

collinear.

may be shown

similarly that
i

=o

(23)

are the conditions that four vectors radiating from a point shall have their extremities coplanar.

Exercise

3.

Given a triangle

e^e x e^

and a point p plane; pe cuts e e%


x
x

in

its

pe cuts

e% e9 in

in qot q >pe% cuts


x

Vi

^ft&cuts^in/.,.
,

cuts ^ a <? in /, and q q cuts * *, in p a to show that


x
:

p ,p and/,
x ,

are collinear.

Let/^v.+V.+V,;
^3 coincide respectively with n e -f- ^/2 ,
,

then ^
lies

^,

/,

n
,

and

+ n,e
if

with q

etc.

Hence,

p x oJ x ,yof y
x

because
x

on the

line joining t%

are scalars,
x

A = *<A + *', = yi,n e +


x
x

hence

(*,

-y
^'s

)e

+ [x

+y {n e + - >,*,>, - n,{y +y )e = o.
,*,)
% %
x

,*.)

Now

the

are not collinear,

and yet are connected by a

ART.

4.]

SUM OF ANY NUMBER OF POINTS.

385

of the form of equation (18); hence, as was there shown, each coefficient must be zero accordingly
relation
;

whence we
hence

fa

= *xyj*x = y> + ?x = find x x =n nv = n e n e and similarly (n - Vu (*. .)a = vrVf *)/. = /.


* -ysh
>
:

,)/,

Adding, we have
(*,

,)A

+ (, = 2ne
,

.}?,

+ (. 2

,)A

=o

therefore,

by

(18),

p of p lt /
2

are collinear.

Exercise

4.

Let /
e^e x e%
:

-i-

^w

be any point

in

the plane

of the triangle

points of the sides


to e p, e p, and ej>
y

*-/

show that eu e^ and


,

lines

through the middle

e e x of the triangle parallel

meet

in a point

/=
By the
to p'
is

[(,

*,)*.

+ (*. +

.y,

+ (. +
;

,yj

+ 2^.
x

conditions the vector from the middle point of e

e^

a multiple of the vector e

p hence

/ = #', +
or,
%)

'.)

+ *('. ~P) = i(e + O + j(^


Q

-/),

substituting value of/,


a

/ = 4(', + e + x{e - 2ne+2n) = i(e + e )+jie -2ne+2n). hence i)2n 4- n (y [{x x)]e + n (y x)e - Sfi* + n& - *)> = O + [(*
x

therefore, as in the previous exercise, each coefficient


zero,
It

must be

of

and substituting we find p' as above. follows also that the distances of p' from the middle points the sides are the halves of the distances of p from the oppo-J,

whence x

=y =

site vertices.

Prob.

3.

Show

that e

= $2$ is

collinear with

p and p' of Exer-

386
cise 4.

GRASSMANN

SPACE ANALYSIS.

[Chap.

V I1I.

linear with the

Also that, by properly choosing/, it follows that e is colcommon point of the perpendiculars from the vertices
sides,

on the opposite

and the common point

of the perpendiculars

to the sides at their middle points.

Prob.

4.

Given two

and an ellipse, as in the figure, with p% and px Radii of circles 4 and 1, axes of ellipse 2 and 4, small circle and ellipse touching large circle at e 9 and e respectively, e e e 2 an equilateral triangle: show that the cencircles

centers at *

troid of the remainder of the large circle, after the small areas are removed, will be at

= tV(iK ~ A Prob.
5.

2/1 )=-BV(S9^

4*i

3'.).-

If I of a sheet of tin in the

shape

of an isosceles triangle be folded over as in the figure, show that its centroid is given by

3^ = *[35('t
Prob.
6.

+ *,)+
-J

"'Je ne e i e 3
1

If

a tetrahedron
of
its
,

have a
off

tetrahedron of

plane parallel to e a e t e 2

volume cut and one of

by a
of
x

its
,

show

by a plane parallel to e e % e % that the centroid of the remaining solid is at


off

volume cut

/= dn>(227*o +
Art.
5.

175*3

+ 239(^1 +

',) )

Reference Systems.
e
e xt e^

Let p be any unit point,

three fixed unit points,

and w, x,y

scalars

then, writing

p
we must have
also,

= we, + xe + ye
x

(24)

because/

is

a unit point,
(25)

w + x+y-i,
and
for

p
it

is

the

mean

of the weighted points

we
e %i

xext yex

The
x
;

point/ may occupy any


is

position whatever in the plane e Q e e 9

on the

line joining we,

xe with
x

and by varying

y and

w + x,

w
x remaining
constant,

p may

be moved along

Art.

5.]

REFERENCE SYSTEMS.
from
x

387

this line

oo

to

-f- oo

while by varying the ratio

the

oo to along e e lf point we -j- xe may be moved from and thus the first line will be rotated through 1 80 degrees, and / may thus be given any position whatever in the plane.

points to which the positions of other points may be referred is called a reference system, and the For reasons that will aptriangle e e^ 2 is a reference triangle.
of unit

A system

pear

the double area of this triangle will be taken as the unit of measurement of area for a point system in two-dimenlater,

sional space.
Similarly, in solid space, taking a fourth point e a
,

we

write
(26)

p
which
implies also

we,

xe

+ye

ze

w-\-x-\-y-\-z~\\
as

(27)

and p may be shown

above to be capable of occupying any

position whatever in space by properly assigning the values of w, x,y,z\ so that <?,... et form a reference system for points
in three-dimensional space.

The tetrahedron e^ete

is

called

the reference tetrahedron, and six times its volume will be taken as the unit of volume for a point system in three-dimensional space.

Eliminating

between

(24)

and

(25),

we have
(28)

P=?
from which
it

+ *(', - O +jK<. ~ O,
e

may

also be easily seen that

in the plane e% ex ev

Writing/
p = xe
t

= p,
2
,

-e =

p may

be any point
e,

e l}

e2

<28) becomes

+ ye
we
t

(29)

and

e,, e 2

form a plane reference system for vectors.

Similarly, from (26)

and

(27)

find
,

9
and 6
three.
If, in

= **x +?e + *,
system

(30)

ea

e 3 are a reference

for vectors in solid space,


in

any vector whatever being


equations (29) and

expressible

terms of

these

(30),

the reference vectors are of

388
unit length

grassmann's space analysis.

[Chap. VIII.

and mutually perpendicular, we have unit, normal reference systems, and in this case z., z 2 z will generally be used
,

instead of

e,,

e9
5.

e,.

Exercise

To change from one


e
,

reference system to an-

other, say from

e lt e^ to e

',

et

'
%

et \

The new

reference points must be connected with the old

ones by equations such as


e

/.'.'

+ J/' + lf{%

ei

.'.'

+ *// + *V'.
t

Then any point/ = x e -{- x e -f- x e will be expressed terms of the new reference points by substituting the values
x x

in

of

etc., as

given.
x ,

If

points, e9 , e
ex
'

e^

may

are given in terms of the old be found by elimination. Thus, if e 2le y


e \ e
x

et

'

'

2me,

e^

=
K

2ne, we have at once

/o

390

grassmann's space analysis.

[Chap. VIII..

Art.

6.

Nature of Geometric Multiplication*

idea of geometric multiplication is, that a product of two or more factors is that which is determined by those factors.

The fundamental

Thus, two points determine a line passing through them, and also a length, viz., the shortest distance between them

hence /,/,

=L

is

the sectf drawn from


rectilinearly

p to/
x

or generated

by a point moving

from

to p^.

The student should note


p^p^ and/> 2 the sect p^p t

carefully the difference

between

they have the same length and direction, but confined to the line through these two points, while the vector/, p is not. The sect has position in addi-

is

tion to the direction

and length possessed by the vector.

Again, in plane space, two sects determine a point, the intersection of the lines in which they lie, and also an area, as
general In solid space, however, two lines do not, in but two sects, in general, meet, and hence cannot fix a point this case, determine a tetrahedron of which they are opposite
a unit point.
;

will

appear

later, so

that Z,Z, 3 == p, in which

is

not

in

edges.
It

appears, therefore, that

product
in

may have

different

interpretations in spaces
will consider separately

of different dimensions.

Hence we

products, and those in

plane space, or planimetric solid space, or stereometric products.

products

Products of the kind here considered are termed " combinatory," because two or

combine to form a one of them. This is the any fundamental difference between this algebra and the linear
factors

more

new quantity

different from

associative
special case.

algebras of

Peirce,

of

which quaternions are a

the various products that may arise, we will give a table which will serve as a sort of bird's-eye view of the subject.
in detail
* Grassmann (1844), Chap. 2 f See Art. I.
;

Before discussing

(1862),

Chap.

2.

Art.

6.]

NATURE OF GEOMETRIC MULTIPLICATION.

391

In this table and generally throughout the chapter

we

shall

use/,

A>A>

etc., for

points;
?/, Tj lf

e,

e3

etc.,

for vectors; L,
;

x ,

etc., for sects, or lines; etc., for

etc,, for

plane-sects, or planes.

Also/,/,,

lf plane-vectors etc., as used in this

and P,

table will not generally be unit points.

The products are arranged in two columns, so as to bring out the geometric principle of duality.
Planimetric Products.

AA = L.

392

grassmann's space analysis.

[Chap. VIII.

Laws

of

products are

Combinatory Multiplication. assumed to be subject to the

All

combinatory

distributive law ex-

pressed by the equation

A(B+C)=AB + AC.
The planimetric product of three points or of three lines, and the stereometric product of three points or planes, or of four points or planes, are subject to the associative law. That is,
In Plane Space
:

AAA =AA -A = A AA
AAA = A A A = A A A

L L%L = L,L .L = L L,L


x

In Solid Space
;

ftV. = P.
;

tt = Pf,

ft

AAAA = A AAA =AA -A/*


The commutative law
hold.

PPPPP PPPPP
may

pp

of scalar algebra does not, in general,

Instead of

this, in the products just given as being asso-

ciative, a

law prevails which

be expressed by the equation

AB=
from which
it

- BA,
two
single

follows that the interchange of any

factors of those products changes the sign of the product.*

Since vectors are equivalent to points at law holds for e^e^ and rjjjjjf

oo

the associative

Art.
Product of
Art.
6,

7.

Planimetric Products.
Points.f

Two

This has been


its

fully defined in

and

it is

evident from

nature as there given that


(32)

A A = -A AIf

x ,

this

becomes p p
x

o,

which must evidently be

true, since the sect

is

now

of

no length.
(33)

Also,
* Grassmann

A(A -A) =AA -AA = AA(1862),

Chap.

3.

Grassmann

(1862), Arts. 245, 246, 247.

ART.

7.]

PLANIMETRIC PRODUCTS.
2

393

But/

is

a vector, say, e
e

hence
\

A = AA
vector by a point fixes the point.
its

(34)

or the product of a point and a vector rection and magnitude of the vector
;

a sect having the dior, again, multiplying a


is
it

position

by making

pass through

To

find

under what conditions pp'


3

Take any other


and write /

will be equal to /,/ 2 the plane space under consideration, point / xj> % +xj> p' =-y,p y,p, x,p yj> with
.

in

the conditions for unit points

2x = 2y = o.
x x
n

Then
If this
jf 3 jj/ 3

pp'
is

y
l

y,

AA +
xy
x

y* y

PJ*

AAy* y>

to reduce to
x

s a y* x*y unless the coefficient of /,/,

/,/, we must have the third condition = o, which requires that x = y = o,


a %

is

to vanish also.
If,
.

Thus //' must


in

be
x

in

the same straight line with/,/,.


x

moreover,

addition

x y^
to p'
is

AA

is equal 1, x^y /,/2 when, and only when, the four points are collinear, and distant from/ by the same amount and in the same direc-

we

shall

have pp'

Hence pp'

tion that/,

is

from/,.

Art. 6 the product is what In solid space they would is determined by the three points. fix a plane, but, as we are now confined to plane space, this is

Product of Three Points.

By

not the case.

points evidently fix either a triangle or a will be parallelogram of twice its area, and the product taken as the area of this, or an equivalent, parallelogram.

The

AAA

This area
it
is

is

taken rather than that of the triangle, because


it

what

is

generated by/,/, as
till it

is
3
.

moved

parallel to its

initial

position

passes through
x

We have pjtj> ./ / -/, .p p, = -AAA. so that we go around the triangle in the opposite sense the sign is changed. As this product possesses only the properties of magt a 3 3

=p

if

nitude and sign

it is

scalar.

Write /

2xp, p'

^yp p"
y

*2zp

then

394

grassmann's space analysis.

[Chap. VIIL

x x x
\ x

PP'P"

y, y*P,P.P*\

(35)

Art.

7.]

PLANIMETRIC PRODUCTS.

395

determined by

and Z 2
//,/,.

is

then the same that

we have

given as the value of

We
is

write therefore

4A = AA -AA = AAA
The
third

A-

(38)

member

of (38)

not to be regarded as derived

from the second by ordinary transposition and reassociation of the points, for the associative law does not hold for the four
points taken together, since //,/ -A simply results from the definition of L
as a

=
X

Tne
It

tnird

member
be taken

L^

may

model form which


to

will

be found to apply to several other

cases, for instance

(38)

when

points and lines are intert

changed throughout. Thus, if/, = L L and /, = Z Z we have (39) AA = L L Z Z = L L L Z For take //and //so that/,// = Z, and/ //= L p p is
2
. l

evidently

some multiple

of

say

nL

hence

AA = *. = ^( A A -A A') (AA
=-

AA')
by
(38),

= ^AA'-A)-(AAA'-A)>
2

-AAA'- AAA' -AA> because /,/,// and

AAA' are

scalar,

~-

(AA AA' -AAO


2
.

4,,

by

(38),

= Z Z,Z Z
cated, but

which was to be proved.

Product of Three Sects.

The method has


:

just

been indi-

we may
,

Z,,

Z
,

and
l

let

Let the lines be proceed thus be their common points. Take /,/,,/-,
also

scalars n Q

n% so that
,*>*, ./,/,

=
}

n9 p

etc.,

then

Z Z,Z =
2

AA AA =
.p ip

n * n Jh

A A A A -AA
2
-

= - n n^ .p,p p
uct of this kind, the

= ^^(AAA)

(40)

" " " * Grassmann to a prodregressiv applies the terms "eingevvandt and first term being used in the Ausdehnungslehre of 1844,
in that of 1862.

and the second

See Chapter 3 of the

first,

and Chapter

3^

Art. 94, of the second.

396

grassmann's space analysis.

[Chap. VIII.

Product of a Point and


let Z,

Two
then

Sects.

Let/ be any point and


(40

and

be as

in (38)

pL,L
It

= p.p p, -Pop, = A AAA A = AAA ./ApL


.

has been here assumed that

uct

is

not associative, for


x ,

pL L Z is
x

p L
.

L^.

The prod2

the line

times the

a different meaning from that assigned in (41). As a rule, to avoid ambiguity, the grouping of such products will
scalar pL

be indicated by dots. Product of

Two

Parallel Sects.

Let them

be^e and
.

np 3 e;

then, as in (38),

A*
that
is,

np % e

= n .p e .p e =
x

n.ep

ep t

n
00

ep

e,

(42)

a scalar times the

common
point, and

point at

Addition and Subtraction of Sects.


sects,

Let Z, and

be two
that

A = AA A = PoP, then A + A =AA +AA = A(A +A) = zp p,


;

their

common

and

so taken

(43)

being the mean of/,

and/

2;

hence the sum

is

that diagonal

of the parallelogram which passes through

Also
(44)

A-A=A(A-A),
by
Z, and Z 2 If the two sects are parallel let
.

so that the difference of the two passes also through p and is parallel to the other diagonal of the parallelogram determined

them be n p
x

and # 2/ 2 e;
(45)

then

*iAe

+ n,P^ =
is

(Mi

+ M.) = (, +

,)/

so that the sum

a sect parallel to each of them, having a length equal to the sum of their lengths, and at distances from
inversely proportional to their lengths.

them
If

n the two sects are oppositely directed and of equal length, and the sum is w2
x

(A ->.e)
Avhich, being the product of

,(A

-A).
is

(46)

two

vectors,

a scalar area.

Art/7.]

planimetric products.
j

397
,

Consider next n sects p e ,p 3 e 2 arbitrarily chosen point; then


1

pn en and

let e

be some

= e 2e + !(/ - *,)e. (47) |/e The second term of the third member of this equation, being a sum of double vector products, that is, a sum of areas, is itself

i/ =

e.Se

- e 2e +

equal to the product of any two non-parallel vectors of suitable lengths. Therefore, a and /3 being such vecis

an area, and

tors, write

2e

= a/3. 2pe = e a + aft = (e a and 2(p


e )e

Hence
fl)

(47)

become
(48)

Let q be some point on the


%

line

2pe

then

q2pe =0 = qe a + qa/3 = qe a + a/3, by (37) hence qe a = a/3 = /3a.


Q

The

figure presents the geometrical

mean-

ing of the equation, and hence it appears that qa{ 2pe) is at a perpendicular distance from
e

of

afl

Ta

_ ~

2{p

e )e
'

T2e

(49)

It is easily seen that a sect possesses the

exact geometrical

properties of a force, namely, magnitude, direction,

and position, and the discussion of the summation of sects which has just
been given corresponds completely to the discussion of the
sultant of a system of forces in a plane.
re-

In this algebra, then,

the resultant of any system of forces is simply their sum, and this will be found hereafter to be equally true in three-dimensional space.

The

expression
e )e of

in (46)

corresponds to a couple,
this equation proves
is
,

as does also the

2(p

(47);

and

the proposition that any system of forces in a plane

equiva-

lent to a single force acting at an arbitrary point, e

and a

couple.

Equation (49) gives the distance of the resultant from

this arbitrary point.

Exercise

7.

To

find x, y, z

from the scalar equations

398

grassmann's space analysis.

[Chap. VIII.

Multiply the equations by add hence


;

pvi p%%
3

and

respectively,

and

x^ap
i

-f-

y~2bp
i

+ z~2cp = 2dp.
:

ii3
. .

Now

2ap, ^bp

etc., are points

multiply the equation just

written by

2ap.2bp; thus

z2ap ^bp^cp
.

2ap 2bp 2dp,

because
z

2ap 2ap
.

= o,

etc.;

therefore

= 2ap.2bp.2dp+2ap. 2bpcp = [>,,

<K|>

<rj,

x and y
points.

a very simple proof of the determinant solution. Of course will be found by multiplying by the other pairs of
e2
f

Exercise

8.

Forces are represented


of the sides of a par-

by given multiples
;

allelogram Let the parallelogram be double the triangle e e ev and the forces
tt

determine their resultant.

k,e n e,

+ke
x

(e%

+
=

k,e,(e

e)

(k Q

ex x )e

+ k = 2pe + (k + k + (k + k )e e
a

e2 e

)e e t
x

t 9.

Multiply by e then

e x to find

where the resultant cuts

this line

or */, cuts the resultant at the point


[X^

+
x

A,),,

_(*,

+ *,>,]
t

-5-

(6,-i,).

triangle at [(,
at
9 x

+ kfc - {k + k )e (, + - k k and - {k + *,>,] - (k - k [(k + k )e then each of the three points Suppose k = k = ^ =
x]

Similarly the resultant cuts the other sides of the reference


-43
x

a ).

,;
;

just found recedes to

infinity

but

in this
**)>

case

2pe reduces
s y stem

to
is

2^(y,'+^+ V.)equivalent to a couple.

2 &fa'*X'*

and the

= 2, k.~ 3, & = 4; when ^ = 3,^ = ^=2,^=


k
1

Prob. ii. Construct the resultant of Exercise 8 2, k 2 1, k 3, k a a

when k

i,

and when k

= k^ =

1,

ka

4;

when
2.

ART.

8.]

THE COMPLEMENT.
.
.

399

Prob. 12. There are given n points /, .p n to find a point e such that forces represented by the sects ep ep^ etc., shall be in
\ x ,

equilibrium.

(The

equation of equilibrium

is

2ep = e2zp =

ep

o.

e coincides with the mean point of the/'s.) Prob. 13. If a harmonic range *,,/, e^ p' be given, together with some point e not collinear with these points, show that

Hence

<V,

(Let p
Art. 3.)

=me
x

-f-

m^e^

<V/ = - *o P*% '.Am e as and p' = w/,


% %
,

in

Exercise

2 of

Prob. 14.

Show

that the relation of Prob. 13 holds for any four


e e tf

points whatever taken respectively on the four lines e e lt e p, If the four points are all at the same distance from e Q Q p'.
,

show

that the areas e e p, etc.,


x

become proportional

to the sines of the

angles between

e ex

and

p, etc.

Art.

8.

The Complement*

Taking point reference systems, or unit normal vector reference systems, as in Art. 5, the product of the reference units taken in order being in any case unity, the complement of any
reference unit
is

the product of

all

the others so taken that

the unit times

its

complement

is

unity.

To

find the

complements

of quantities other than reference


:

units the following properties are assumed (a) The complement of a product is equal to the product of the complements of its factors.
(b)

The complement

of a

sum

is

equal to the
is

sum

of the

complements of the terms added together. (c) The complement of a scalar quantity

the scalar

itself.

Considering now the point system in plane space ew e lt e% with the constant condition e eA e^= 1, the sides of the refer-

ence triangle taken in order are the complements of the opposite vertices, and vice versa.

The complement
line, as
\p,

of a quantity

is

indicated

by a

vertical

read,

complement

of p.

* See Ausdehnungslehre of 1862, Art. 89.

400

grassmann's space analysis.


lA

[Chap. VIII*

Thus

= '/t k = <Vo>
k.= v,i
li

kA=l(W=s ^i k.^=kk) =
'i*

ko^

= l(k =
a)

^;

For
I

^ |^

= W, =
I

which agrees with the definition

'A = k

e%

?%'%

V, =

V
=
o

Vi = - VA
(

eo

*o>

by

(a)

and
I

W=

(38)

ko

ki
(*:)
;

k = *A
eQ [e t

'A Vi = V/J' =

= VA

which

agrees with

e e2 e

e \ea

= e^e^
have, by
(b),

Next take any point/,

^7^, and

we

\A=2t\?=W>+W.+^=W,(l
Thus

j")g

J)

= A-

(50)

the complement of a point is a line,* which may be easily constructed by the fourth member of (50), which expresses this line as the product of the points in which it cuts
the sides
this
e e1

and

e e^

of the reference triangle.


4, it

Comparing
\/> 1

equation with Ex. 3 in Art.

appears that

above

is

related to the point

e
-.

as the line/

of Ex. 3

is

to the point

2ne.

Hence

\p
1

may
e
>

be found by constructing this line corin

responding to

as

shown

the figure of Ex.

3,

Art. 4.

Again, the line \p may be shown to be the anti-polar of p with respect to an ellipse of such dimensions, and so placed upon e e e 9 that, with reference to it, each side of the reference
x
x

triangle
it

the anti-polar of the opposite vertex.* From this appears that complementary relations are polar reciprocal
is

relations.

Take any

point

p = 2me, and we have


2

= 2/m =
*See Hyde's Directional

2me.2l\e=zpt \pit
Calculus, Arts. 41-43 and 121-123.

(51)

Art.

8.]

THE COMPLEMENT.
is

401

so that this product


sign,

commutative about the complement

true of all such products when the on each side of the complement sign are of the same quantities

and

scalar.

This

is

order in the reference units.

Take

for instance
is

the product

AAlAAhave/,/),

This

is scalar
is

because \p ap A

a point, so that the


;

whole quantity

equivalent to a triple-point product


I

and we

|aa

= IAA AA = AA lAA)
(
|

==

AA AA> by (a) and


i

such a quantity be taken as/^, \p 9 it is neither (c). If, however, scalar nor commutative about the sign for, \p % being a line, the product is that of two lines, that is, a point, and
.
;

AA A = - A AA = .
I

(A

AA).

(52)

Such products as we have just been considering are called Grassmann " inner products," * and he regards the sign by as a multiplication sign for this sort of product. Inasmuch, as these products do not differ in nature from those however, heretofore considered, it appears to the author to conduce to
|

simplicity not to introduce a nomenclature which implies a new For instance, p\q will be treated as. species of multiplication.

the combinatory product of / into the complement of not as a different kind of product of p into q.

q,

and

The term co-product may be

applied to such expressions,,

regarded as an abbreviation merely, after the analogy of cosine for complement of the sine.

Consider next a unit normal vector system, tion we have

By the

defini-

k==

..

1*.=
i
J

1(10= -*i*

because

\i 1

z,t 3

I,

Also,

2,

z
1

Mi==
i

z3

z
l

i-

Next

iet

t
i l

+ m^i,

and

e2

i
1

-f n t i 2

* Grassmann (1862), Chapter

4.

402
then,

grassmann's space analysis.

[Chap. VIII.

by

(b)

and
I

(c),

ex

=m
it is

ix

+ m, = m - m
z
|

t,

ttx.

(53)

By

the figure
x

evident that
|

e, is

a vector of the

same

length as e

and perpendicular to
.

it,

or, in

other words, taking


it

the complement of a vector in plane space rotates

positively

through 90

The
of
e
1

co-product
sides are
,

e,

e2

is
I

whose
is

e,

and

the area of the parallelogram, two e 2 drawn outwards from a point if


;

parallel to |ea

this area vanishes, or e |e 9


x ,

= o;

but, since

|e2 is

to e 2

perpendicular to e 3 hence the equation

e,

must

in this case

be perpendicular
(54)

e1
is

\e,

=
e,

the condition that two vectors

and

e 2 shall

be perpendicu-

lar to

each other.
e,
|

The co-product
called

e,

which
,

will usually

be written

e,-,

and

the co-square of e, is the area of a square each of whose sides has the length Te hence
x ;

r6l
x

=^K=7^
x

(55)
e
x

z,

Let a and a a be the angles between i and and e 2 respectively, as in the figure. Then
e
x

and between

e,

= mn x

m.i n

Te 7e2
x

sin

(or,

-a

x ),

(56)

the third

member

being the ordinary expression for the area of


.

the parallelogram e,e 2


6,
|.6,
x

Also
i
%

= {m + m i^(n - n = m n + mjt = Te
ix %
x

x)

7e2 cos

(a,

x ),

(57)

the last
sin (90

member being found as a) cos ( a ). -f or,


x a

before,

remembering that

^ ^

(57)

we

let e2

e,,

whence n

tnx

and nt

=m

we

have

Te
If
2

7^ =
ar 2
,

7e

1,

= sin

= V5 = *; + <. then w, = cos a tn = sin a


x ,

(58)
x

,n

= cos ^

2,

and equations

(56)

and
x

nometrical formulas sinfo

- a) =

(57) give the ordinary trigosin a2 cos ax cos a, sin ^,

Art.

8.]

THE COMPLEMENT.
{a,
<x t

403
sin
.

and cos

== cos

cos

or,

+ sin a
2 )'

or,

Squaring and
2
.

adding (56) and (57), there results

Te
Attention
is

.T*e

e;ie*

= (e,e + fe

e2 )

(59)

which the student may have = AC, in which already noticed, that such an equation as AB AB and AC are combinatory products, does not, in general,
called to the fact,

imply that B = C, for the reason that the equation can usually be satisfied without either factor being itself zero. Thus pL = pL 2 means simply that the two quantities which
x

A(BC)0

are equated have the

to have an infinity

same magnitude and sign, which permits of lengths and positions, when p and L
x x

are given.

The equation p

p,

=p

% ,

or

A (A

~~

A)

/, being unit points, implies, however, at 00 that is, a vector.


,

that/ 2

=A

un ^ ess

A A

anc*
ls

Exercise

9.

triangle
its

whose

sides are of constant length

moves
Let

so that two of

vertices remain

on two fixed

lines

find the locus of the other vertex.


e e
x

and

e Q e,

be the two fixed

lines,

Let pe be pertriangle. e = xe and pendicular to p'p", p' - = ye, - xe - e9 je2 then /' /"

and pp'p" the

7( je 2
ditions.

constant, by the con.are, ) constant = Also, Tp'e

=c=

say,
e

and Tep

constant
U{e
e

= nc,
mc
9

say.

Hence
~~~

-p'

Tp'e

- p') =

VG
.

Xe

^^
xe
x

__

m{ye,

#,),

and

similarly

= n\(ye
l

).

Therefore
x

p-e = p = xe + m(ye, - xe + n
)

(ye,

an equation which, with the condition T( ye,

- xe xe =
x

),

c,

or

e2 -

2xye

e2

+x

e x

= c\

determines the locus to be a second-degree curve, which must in fact be an ellipse, since it can have no points at infinity.

Let us rearrange the equation

in

p thus
i

p = #[(1

m)e

n\

ej

+y[me

+ n\ ej = xe + ye\ say,

404
so that e

GRASSMANN

SPACE ANALYSIS.
e'

[Chap. VIII.

(i

m)e

n e and
\

me

+n

ea

then multiye'e and the equa-

ply successively into


pe'

and

e'\

therefore

pe

=
in

xee\

Substituting these values of

x and y

tion of condition,
2

we have
|

el (pe)
.

+ 2e,

,.

pe pe'
.

+ e^pej = c\ee')\

a scalar equation of the second degree in p.

Exercise
sides
:

10.

There
if

show that

given an irregular polygon of n forces act at the middle points of these


is

sides, proportional to

them

in

ward or
Let

else all inward, these forces will


e

magnitude, and directed all outbe in equilibrium.


, . . .

2e n be a vertex of the polygon, and let 2 v 2e 2 Then the midrepresent its sides in magnitude and direction. dle points will be e -{-e lf en -\-2ex ~\- e 2 etc., and, using the
,

complement

in

a vector system,
I
I

we have
e2 +(^o+2e
1

2/e

= OvK) *+('+*<+*)

+2e +e )
2
3
.
.

e9

+
2e+2 + 2
n
I
l

(*o

+ 2e +
x

+ 2e_

+ + OK
. . . ..

2e +

2e,

2e

+ ... + 2e _
n

\en

\ 2

2;e-f-

12 ej=

o,

which was to be proved.

Exercise

II.

A line

two

passes through a fixed point and cuts fixed lines at the points of inter;

section perpendiculars to the fixed lines

are erected

find the locus of the inter-

section of these perpendiculars.

Let the fixed

lines

be

e e
;

and

e Q e2

and the fixed point


at

e -f- e 3

the moving

line cuts the fixed lines

in p'

and p n

'.

points perpendiculars erected meeting in/.

which

are

Let/
then p

= p,p'
:=
J

= xe ,p"e = ye
1

Te,

= 7e

=i;:

xe

-{-x'\e

^e 2 -f-j/

e2

whence p\e,= x and p\e

y..

Art.

9.]

equations of condition, and formulas.


e Q -\~ e % , p' ,

405

Also, since

(. -

e *)(y e *

p"
)

are collinear points,

= *y*fy+y*A
jj/,
.

+ *w"

or, substituting values of


.

^ and

p| e 2 e 3 e 3 P|< ; p|et 6,6, an equation of the second degree in

+ p|
p,

e,

e 3 e, ss O,

and hence representing

a conic.
Prob. 15. If a,
the formula a*
2
1

=b +c
%
1

b, c

are the lengths of the sides of a triangle, prove 1 2bc cos A, by taking vectors e,, e 9 and
,

equal to the respective sides. Prob. 16. If e e and e e* are two unit lines, show that the vec# e i)(^ tor perpendicular from e on the line (e ^*) is

abe.e*
:
.

...

(be*

(be^

ae

ae.), of

which the length

.. ...

abee
7=
r-.

is

)-

T(be*

ae

From

origin

this derive the Cartesian expression for the perpendicular upon a straight line in oblique coordinates,
00
2

from the

ab sin

-f-

(a

b'

2<zb.

cos 00)^, go being angle between the axes.

me *,, me -+- ne*, me* -f- ne , be 17. taken on the sides of the reference triangle, then the sides of the ne^), etc., will be respectively paralcomplementary triangle, (me lel to the corresponding sides of the triangle formed by the assumed
Prob.
If three points,
x
|

points (me,

* 9 )>

me *

+ ne

o)>

etc

Art.

9.

Equations of Condition, and Formulas.

Several equations of condition are placed here together for convenient reference some have been already given others follow from the results of Arts. 7 and 8. When we have
:
;

AA = o iA + n *P* =

44 =
or
;

o,

>

+
3

n 2 L,

= o,
o,
o,
J

(60)

the two points coincide

the two lines coincide;

AAA = o1
or

444 =
or
;

2np

= o,

2nL =
1

(61)

the three points are collinear


e.e,

the three lines are confluent.

= O,

or

n e
l

n,e,

= O,

(62)

the two vectors are parallel (points


*i

at infinity coincide);

[.

(63)

406

grassmann's space analysis.


;

[Chap. VIII,

the two vectors are perpendicular

Z,|Z,
either point lies on the com-

= o,

(64)

either line passes through the

plementary

line of the other.

complementary point of the


other.

If

we

write the equation

*\ e i
ei

+ x ev
*

e1 1

is

the projection of p on

parallel to e 2
.

and

;r2 e a

is

the

projection of

p on
;

e 2 parallel to e a

Multiply both sides of the

equation into e 2

therefore pe 2

ex e

or

x
x

pe a -~

e,e 2 .

Similarly, multiplying into e

we have

pe,

e e lt or 2 2

= pe,

-5-

e 2 e,,

whence
e,

.pe 2

e 2 .pe,

The two terms


the projections of
respectively.*

of the second

member
,

of (65) are therefore


e 2 parallel to e,,

pon

e,

parallel to e a

and on

Let

e,

and

e2

be unit normal vectors, say,


i.p\i
\i.

and

\i\

then (65)
(66)

becomes
p
or,
if
i
x

pi= i.p\i+ip.\i;
1

and

be used instead of

and
|

z,

P= VPK +
Again,
therefore
e,e 2
.

-Plv

(67)

in (65) let

p=

e3

clear of fractions,

and transpose

e3

e2 e 3

e,

+e

e,

e2

= o,

(68)

a symmetrical relation space.

Let 7e,

between any three directions in plane 7e 2 = 7e = 1, and multiply (68) into |e


3

3,

thus

e^ +
is

e2 e 3

e,

e,

+e
cos

e,

e2 e 3
1

= o,
a
sin
/?,

(69)

which

equivalent to
sin (a
/3)

sin

/3

cos

the upper or lower sign corresponding to the case


* Grassmann (1844), Chapter 46 and 47.
5

when

e\,

is

(1862), Art. 129.

Hyde's Directional Calcu-

lus, Arts.

Art.

9.]

EQUATIONS OF CONDITION, AND FORMULAS.


e1

407
in (69)
i

between

and
,

e2

or outside, respectively.

Writing

ea

instead of e 2

we have
eje 2
e 2 |e 3 .e 3 |e 1

+ee
3

.e 2 e 3

= 0,

(70)

These formulas being for any which gives the cos (a /?). three directions in plane space, are independent of the magnitude of the angles involved.
There
is

given below a set of formulas for points and lines,

complementary pairs, and all placed together for arranged convenient reference, the derivation of them following after.
in

/=(AAA)" [A -/A A + A -/A A + A -/AAL Z=( AAA)-'[4 LL L + L LL,L + A ZZ.ZJ /^(AAArClAA-AA + IAA-/IA + AA-A /IAU
1

(70

z^ZAATUAA-^iA+IAA.^IA+IAz; AA- A A = - A AAA + A -AAA = A A A A A AAA. 44 44 = - A 444 + 4 444


. . .

(72)

(73)
J

4-

AAA- L

.L Lt Lt
t
1

AA-

|*3=

A 4 if, A 4W Jf; AW,

^ AW
_
?i

'

AW Al AW Al
I

A ft
a I?.
Al?.
not given, but
^'s.

r,

',

The complementary formula


Derivation of Equations (71)

to (77)

is

may

be obtained by putting Z's and M's for /'s and


(77).

Equation
this

(71).

Write

p=xp
then

-\-

xp
1

-\-x% pv and multiply


or
2

equation by

pp
l

AA/ = *oAAA>
pp
-^

= /A A ^ AAAfind
-5-

Multiplying similarly by
*,

= /AA

-*-/.AA and

= /AA

and by p p v we

A A A-

The

substitu-

408

GRASSMANN

SPACE ANALYSIS.
first

[Chap. VIII.
is

tion of these values gives the


similarly obtained or p's in the first.

of (71),

and the second

may

be found by simply putting Z's or

Equation
values of

(72).
;

Write/
,

multiply into \p o
x

thus/|/ x and x% and substitute.


(73).
x

= x \p p + x p,p + x = %*P*P\Pv Find in the


x

p p and
o iy

same way

Equation

Write p p .p p
2 s

-~
x

xp

by PA* therefore pp% .p,p^.p*p, = Xpp% p or, by Eq. (38), = *PP*Pi = *P%PP or * PiPtPv MultiplyA/A we find _y =AAA> anc * on substituting obtain the ing by pp For the second put p 'p % p p A = *p s -\-ypv and first of (73).

+ ypv and multiply

AAA
x

>

=~
9

proceed

in a similar

way.
first

Equation

(74).
(75).

In the

of (73) put

p pK
%

= \q

Equation
Equation

In the fourth of (73) put


kit

AA =A A =
(76).
;

A=

|ft

Multiply (75) by/,. Equation (77). In the first of (72) put byAAA-ftft then

for/, and multiply

AAA

ftftft

= ftft AA
I

ft

IA+ ftft lAA


Alft
A'ftl

ft

|A+ ftft |AA


Alft

ft

lA

=A
by

Alft
ft-

Alft

Alft

Alft
is

+Alft-

Alft

A'ftl

+ Alft-

Alft Afft
A'ft

(76),

which

equivalent to the third order determinant of

equation (77)* Exercise 12.


2

To show

the product of two determinants as

a determinant of the same order.

Let p
then

= 2le, p = 2me, p = 2ne,q = 2\e, q=2fxe, q,=2re;


x

pp

p,

[/

x1

J, ?

()

*** etc. Sub'1*1 ft *!*, these values in (77), we have the required result. stituting solution may also be obtained directly without the use of (77).
[ft

/o A o

+ l*K> A

^==
2

[A

//

vj

also

w o*

+
2

Let the

q's

be as above, but write p

2lq,p

= 2mq,p =
2

2nq.

Then

p p p,=2fq.2mq.27tq=[/
o
x

,]?.?,?,=[/

J[A

ju

fj.

* Grassmann (1862), Art. 173.

ART.

9.]

EQUATIONS OF CONDITION, AND FORMULAS.

409

Also /,

= l^\e + l 2fie + l^ve


t

with similar values (or p

AAA

g lvQ tne resu lt.


in

product form.*

which on being substituted in Equation (77), however, exhibits the a very compact, symmetrical, and easily remembered
t

and/

Exercise
x

13.

Show

that the sides

AA

AA> P%P%
,

tne

tr *-

angle p p^p % cut the corresponding sides \p 3 \p plementary triangle in three collinear points.

x ,

\p 2 of the

com-

The

three points of intersection are, using (74),

AA-lA=-A-AlA+A-AlA.AA-lA = -A-AlA+A.AiA. AA -!A = A-AlA+A-AlA> of which the sum is zero,


showing that the points are
the same
confluent.
collinear.
It

may

be shown

in

way

that the lines joining corresponding vertices are

Exercise 14. If the sides of a triangle pass through three fixed points, and two of the vertices slide on fixed lines, find the locus of /\
the other vertex.

iy

Let the fixed points and lines be p p L iy L and /, /, /" the


s ,

%
2

"

,/ " p/

\ * \
\

7>^r >'p\

_.-""/

M
\

vertices of

the triangle, as

in

the

o; p' figure. p'p zp" cides with pp^.L^ and p" with pp^. Z 2 hence substituting o, the equation of the locus, which, being {pp L^)p z (L^. p 2 p)
;
.

Then

coin-

of the second degree in/,


Prob. 18.

is

that of a conic.

Show
(73).

that

if

are collinear, then the locus of

the three fixed points of the last exercise p breaks up into two straight lines.

Use equation

Prob. 19. If the vertices of a triangle slide on three fixed lines, and two of the sides pass through fixed points, find the envelope of
the other side.

(This statement
is,

is

reciprocally related to that of

Exercise 14, that


* These methods

lines

and points are replaced by points and

may

be applied to determinants of any order by using a

space of corresponding order.

410
lines respectively,

grassmann's space analysis.

[Chap.

VIIK

and the resulting equation will be an equation of the second order in Z, a variable line.) Prob. 20. Show that if the three fixed lines of Problem 19 are confluent, then the envelope of L reduces to two points and the line joining them.

Art.

10.

Stereometric Products.
in solid

The product
plane space.
plane, and

of

two points
7.

space

is

the same as in

See Art.

Product of Three Points.


also, as in

Any
;

Art.

7,

an area

hence

three points determine a is a plane-sect

AAA

or a portion of the plane fixed by the three points whose area is double that of the triangle AAAIt may be shown, in

the manner used


in this plane,

in

Art. 7 for the sect, that no plane-sect, not

can be equal to AAA* and that any plane-sect in this plane having the same area and sign will be equal to AAA**' is not now scalar. Of course

AAA
of

Product

Four

Points.
5 x

Any

2^
*

y ^p\ x _\/ \
\

four non-coplanar points determine a tetrahedron, say J


'

7s^
4

\ ume S* \/
x

AAAA- anc
of
f

s*

x times the
value of

volis

this
r

tetrahedron

ta ^ en

the

the

product, because this is the volume of the parallelepiped


i.e.

generated by the product

AAA

the parallelogram A>A

when
Pi

it

moves
e,

A=

If/,

A A= A A= AAAA = AAA*" = AA*V = A<*V'. c2^, A = <27* A = 2me,p = 2ne, then m AAAA = 2&e2le2me2ne = [k v/A

333 000
it

parallel to its initial position " e then e',

from/, to p v

Let

(78)

ot

/,,

,]

(79)

any two quadruple products of points differ from each other only by a scalar factor, that is, they differ only in magnitude, or sign, or both; hence such products = o, the volume of the If are themselves scalar.f
from which
appears that

AAAA
f

tetrahedron vanishes, so that the four points are coplanar.


* Grassmann (1862), Art. 255.

Grassmann

(1862), Art. 263.

ART.

10.]

STEREOMETRIC PRODUCTS.

411

Product of
area as in Art.

Two
7,

Vectors.

The two

vectors determine an

tion, so that the

but they also determine now a plane direcproduct e^e., is a plane-vector, and is not scalar
Also,

as in plane space.
differs

from pe

direction, that

is,

namely, toward a certain

e^ differs from p e^ has a definite


x

<G

now

just as e

area and plane

line at infinity, while

e^ is

fixed in position

by passing through/,.

Equation

(37) there-

fore does not hold in solid space.

Product of Three Vectors.

Three vectors determine a

parallelepiped as in the figure above, and ee' e" is therefore the volume of this parallelepiped. Any other triple vector
let e l

product can differ from this only in magnitude and sign. e i be such a product, and write l
.

=x e +^ e +xe =
1 1

2xe,

33
e
f

For

= 2 ye,
*,

e"

= 2ze

then

412

grassmann's space analysis.

[Chap. VIII.

Product of a Sect and a Plane-Sect.


P, and
let

Let them be

and

be their

common
and

point; take A,

A* A

so tnat

Z=AA
point

an<^

PPaPtPr L
is

evidently determine the

A' an d

also the parallelepiped of

which one edge

is

L
of

P, so that the product should be Hence we write these two factors.

and one face

made up

LP = AA AAA = PP J J*
PL =P.PJ,
If

A*>

-Pop!

= AAAA
is

-A

LP.

]
it

< 8 3>

is

parallel to P,

at infinity, and, replacing

by

e,

{83)

becomes

PL =
Two

LP=

eA

eAA

= eA AA

'

'

Product of
let
x

Plane-Sects.

Let them be

and

^
and
faces,

P =
and and

be their intersection, while and are such points that Lp and P, Lp2 then P and P9 determine the line L
x

also a parallelepiped of

which they are two adjacent

Pf^Lp^.Lp^Lp^.L^-Pf,
If P,

(85)

and

are parallel,
t]
;

to a plane-vector, say to

at infinity, and is equivalent hence, substituting in (84),


is
'/

PA = >ZA
;
,

?A = ?AA

= By

^and
(83) this

(86)

Product of Three Plane-Sects.

(85)

must

be the square of a volume times the common point of the three planes or, if p A> be taken in such manner that

A = AAA, ^ =AAA. ^3 = AAA P,/^, = 023 031 012 = 023 0123


2
.

A A

then
.

01

= (AAAA) -A
3

(87)

the suffixes being used instead of the corresponding points. If be at infinity, the three planes are parallel to a single line,

and may be written


above.

ep.1

etc.,

and then treated

as

Product of Four Plane-Sects.*

Let the planes be

and
etc,

let

-A

be the four
n
. .

three by three.
;

n%

common may be so
*

points of the planes taken taken that #

P = AAA>

then

PPPP =
l

nji x n % n %
9

123. 230. 301 .012


(88)

= *.,* ,(AAAA)'
* Grassmann (1862), Art. 300.

ART.

10.]

STEREOMETRIC PRODUCTS.

413
?; 2

Product of

Two

Plane-Vectors.
;

Let

tf l

and

be two plane-

vectors or lines at infinity and e, and e2 so taken that

let
17,

= ViV*
because
7,

ee

6e 2

= =

be parallel to each of them,, ee lf ?; 3 = ee 2 then


,

,,. 6 as

^ffc,
e,

(89)

and
of

7/ a

determine a

common

direction

and a

paral-

lelepiped
e,

which three conterminous edges are equal to

e 2 , respectively.

Product of Three Plane-Vectors.


rf^VtVt

Take

e lt e 2
2 3)
1

e 3 so that

=*
. .
.

e 2 e3

636,

,,

= w(e e e
2

(90)

The
77,
.

e 3 are common to the plane-vectors e, taken two by two. ?/, Several conditions are given here together which follow

directions

from the results of

this article.

AA = o,
Two
points coincide.
o,

P,P,

= o,

AAA =

Three points

collinear.

AAAA = AA-AA = L,L = o,


2

Four points coplanar


lines intersect.
e,e 2

two

o,

Vectors

parallel.

6^6, = o, Three vectors


one plane.

parallel to

Sum

of

Two

Planes.
line,

sect in their

common

P, = Lp,\

and take p and/, so that


l

P = Lp
l

lf

then

/>+/>= Z(A
p
being the

+ A) = *M
Also

(96)

mean

of

and p 9

P,-P, =

L(

A -J>,);

(97)

whence the sum and difference are the diagonal plane through Z, and a plane through L parallel to the diagonal plane which
is

itself parallel

to L, of the parallelepiped determined

by

414

grassmann's space analysis.

[Chap. VIII.

and

If

TP = TP P
l

9i

P, will evidently be the two

bisecting planes of the angle between them. planes may also be written

The

bisecting

-^YpIf

or

P,TP,P,TPV
let
tf

(98)

the two

planes are parallel,


is,

be a plane-vector
line at infinity,

parallel to each of them, that

their

common

and

let/,

may

write

If

P = n^ptf, P ~ n pjj, whence P* + P* = KA + A)* = (a + n S~Pv= o, this becomes -f- n^


1

and / 2 be points
2

in
2

the respective planes; then

we

(99)

^,

+ ^, =

(/>,

-/,)?>

(100)

the product of a vector into a plane-vector and therefore a


scalar,

by

(80).

Two plane-vectors may be added similarly, have a common direction, namely, that of the
to both of them.

since they will

vector parallel

If two tetrahedra e e e^e s and e^e^e^e^ are so 15. that the right lines through the pairs of corresponding situated vertices all meet in one point, then will the corresponding faces

Exercise

cut each other in four coplanar

lines.

The

given

conditions

are
**'*'
-

equivalent to e e

'
.

e e/
t

=
by

****'

V/ = V'
012

Vt'

**i

Vi'-V,'
e^e x e^

=
.

e /*

'/,'

Two
*t *i e%

of the intersecting lines of faces are

eje^'cj

and

*{*%*%* an(^'
.

equivalent to the fourth condition above, since quadruple-point products in solid space are associative. Similarly all the other pairs
.

= 0123. o'V^i'

(93),

^ these intersect, we must accordingly have, 0V2' 123 1V3' = o = 012 123 o'i'2' 1V3'
.

i2i 2

/
,

the last factor of which

is

of intersections

Exercise

may be treated. 16. The twelve bisecting

angles of a tetrahedron fix eight points, the


six.

planes of the diedral centers of the

inscribed and escribed spheres, through which they pass six

by

The sum and

difference of

two

unit planes are their

two

ART.

10.]

STEREOMETRIC PRODUCTS.

415

Let the tetrahedron be e9 ex e% e%i and bisecting planes, by (97). Te e 2 e 3 etc.; then a let the double areas of its faces be

A =

pair of bisecting planes will be

"

-^-

or e%e x {A %e%

e % ).

The

e e A^^). pair through the opposite edge will be e 3 9 (A Q which the six internal bisecting If there be a point through

planes pass, it must be on the intersection of these two planes taken with the upper signs, and we infer by symmetry that it
3

must be the point 2Ae.


e % en (A e
l l

Another

internal bisecting plane

is

-f-

e 9 ),

which gives zero when multiplied into 2Ae,

as do also the other three.

To

obtain
that

all

signs, so

the points we have only to use the double Ae A e^ A^e^ A 3 er This they are
x

gives eight cases, namely,

+++ + + -+
The

++++ -

++ +

-+++
+

eight apparent cases that would arise by changing all the signs are included in these because the points must be essentially positive.

negative signs, hedron must be greater than the fourth face. It will be found on trial that six of the bisecting planes will pass through

Moreover, no positive point could have three because the sum of any three faces of the tetra-

2(

Ae) with any one of the above arrangements of sign.

Prob. 21. The twelve points in which the edges of a tetrahedron are cut by the bisecting planes of the opposite diedral angles fix eight planes, each of which passes through six of them.

Prob. 22. The centroid of the faces of a tetrahedron coincides with the center of the sphere inscribed within the tetrahedron whose vertices are the centroids of the respective faces of the first tetrahedron.
Prob. 23. If any plane be passed through the middle points of two opposite edges of a tetrahedron, it will divide the volume of the tetrahedron into two equal parts.

416

grassmann's space analysis.

[Chap. VIII*

Art.

11.

The Complement

in Solid Space.

relations in a unit

According to the definitions of Art. 8 the complementary normal vector system are as follows
:

= k= k=
l*i

*.*.

Mi
f

i.

1(10= lv.= 1(10 = Im.= 1(10=


v.

*.
z
3

Let

J/z

then

6
1

= AVs + /.., + hh\ =


e is a plane-vector.

j&h

- W(4 - 4A
which
is

( I02 >

so that
|

The

*^&

hh

drawn in isometric projection, shows that the two vectors /jZ 2 lj / z,, whose prodand l i s
figure,
x
x

^U

uct

is /,

.
|

e,

are both perpen;

dicular to e

for the

first is
,

/ /,*, -|- 2 z a perpendicular which is the orthogonal pro-

to

jection of
z
3
,

eupon
e,

z z
x
ft

,and to

and therefore

is

also per-

S^A'l
second
is
1 1

pendicular to

while the
*

h l * anc* to h> anc therefore perpendicular to / i to e. Hence e is a plane-vector perpendicular to e and, since |(|e)= e, the converse is also true, i.e. the complement of a
;
|

plane-vector

is

a line-vector normal to

it.

The figure shows that e is equal to the vector diagonal of the rectangular parallelepiped whose edges have the lengths
/,,

/3

/s ,

hence

Te

= V'lJTVTV.
l,t,
/,, 3

(103)

Multiply equation (102) by e; therefore


e\e

= (/,, + + )(/,M, + Wi + '.'.0 = /,+/, + /,= re =6,

(104)

so that the co-square of a vector is equal to the square of its The product e|e is that of a vector e into a planetensor. vector perpendicular to it, as has just been shown it is there;

ART.

11.]

THE COMPLEMENT

IN SOLID SPACE.

417

fore a

volume which is equivalent to Te T\e; hence, by (104), e]e= Te T\ e = T*e> or Te = T\e. Hence, the complement
.

of a vector in solid space

and having the same


tude.*

a plane-vector perpendicular to it tensor, or numerical measure of magniis


3

Let a second vector be


e e'
I

e'

=2mi

then
3 3

= l m + /,, + l m = e'\e.
x

(105)
|e',

is

e into the plane-vector e|e', being the product of the volume of the parallelepiped in the figw|
is,

Now

ure, that

TeTe'
f.

sin (angle

between

and

|e')

TeTe' cos
i

Hence
1

e\e'

= e'\e=f m +l m +I m =TeTe' cos f.


i
2
z 3

(106)

If

7e

Te'

I, /,

...

/t , *,.".

?/z 3

are di-

rection cosines, and (105) gives a proof of the formula for the cosine of the angle between

two
ee

lines in

terms of the direction cosines of the


case
x

lines.

We
and,

have also

in this

(/,,

ljn

z
|

-f-

(Jt

m - /m
t
a

2)

i
|

+ (/,, - l m
x

i
\

%t

taking the co-square,

{eeJ-=
If

(sin

f^= (/,,-

W+fe-W+W^-W.
e|e'=o,
e',

(107) (108)

e
is

is

parallel to the plane-vector perpendicular to perpendicular to e', as is also shown by (106).

that

is,

Let

rj

=
I

e, rf

=
|

e'

then

I7|V=
and
is

|e.e'

= e'| e =

e\e'

= TeTe

cos f

71?

7Y

cos

J, (109)

^l^^o
=
O,

the condition of perpendicularity of two plane-vectors.


e\rf

(no) Also
(

either

or

1/

e
|

= O,

II]t )

the condition that a vector shall be perpendicular to a planevector, for the first means that e is parallel to a vector which is
is

* Grassmann (1862), Art. 335.

418

grassmann's space analysis.

[Chap. VIII.

that perpendicular to 7', and the second vector which is perpendicular to e.

rf is parallel to

a plane-

stereometric vector Equations (71) (77) of Art. 9 become formulae if e lt et etc., be substituted iov p p^ etc., and i^, //,, For instance, (76) gives the vector formulas etc., for L Z.,, etc.
,

x ,

x ,

v.kk' =
For lack

ik'
*,k'

.k'
ViV*\Vx'v t

'

vAVx
V*\Vi

Vx\V%

.k'

vM

(112)
in

of space

no treatment of the complement

point system in solid space is given. To prove the formulas of spherical trigoExercise 17. cos a == cos b cos c -f- sin b sin c cos ^2, and nometry
sin

_ sin
sin
e,,

sin ^4

B
ea
,

_ sin
sin (7

'

Take three

unit vectors

e s parallel to the radii to the

vertices of the spherical triangle, then # (angle bet. e 2 and e ), A (angle bet. 6,6, and e^), etc. In eq. (112) put e,6, for e/e,';
3

hence

6,6,

efy

sin

sin c cos

^ =
<:.

e ? e2 e 8
.
1

e3

e,

e3

= cos #
.

cos b cos

Again,
T{e
or
x

e e,)
x

sin

= 2^6,6,6, e = Te e,e = r(e e e,6,)= T(e e sin B = sin # sin b sin b sin c sin ^4 = sin a sin
x

.
i

6 8 e a );
7,

whence we have the second


Exercise
in
18.

result

by dividing by

sin

a sin b

sin

c.

Show
cos

that in a spherical triangle taken as

Exercise

17,

2.

= '-r^ f-^, whence derive Ue e 3 ) 7\(/ e 9


1

the ordinary value

/sin 5 sin (5 ..

#)
<:

sin

sin

Expanding, the numerator becomes


the denominator 4/2(1
E/e.eJ
E/e.e,,

-(- C/6,6,1

#6,6,,
is

and

+ Ue^] Ue e
x

2 ).

Also there
is

obtained

-f

^'

^-.

The remainder

left

to the stu-

76,6,76,6,
dent.
Prob. 24. If 6,, e 2 , e 3 drawn outward from a point, are taken as three edges of a tetrahedron, show that the six planes perpen,

Art.

12.]

ADDITION OF SECTS IN SOLID SPACE.


all

419

dicular to the edges at their middle points

pass through the end


1

of the vector p
gestion.

~
2e
fc
i

fc

r(
s

e2 e 3

6*+
x)
1

e3 e

. 1

ej

ex

e 3 ? ).

(Sug-

We

must have (p

ie

6,

= o,
|

with two other similar

expressions.)

Prob. 25.

Show

that

pendicular vectors,

ee' and ee'. e are three mutually pere, no matter what the directions of e and e'
|

may

be.
,

e 2 e 3 be taken as in Prob. 24 let A Q be the xt area of the face of the tetrahedron formed by joining the ends of these vectors, and 2A = Te a e x etc.; also Angle between e e q and e^g, etc.: then show that we have the relation, analogous to

Prob. 26. Let

that of Prob. 15, Art.

8,

2 A 2 A, cos 2 A 3 A cos 6 t %A X % cos 6 V 6 3 are right angles, this becomes the space-analog of the proposition regarding the hypotenuse and sides of a right-angled

'=

A?+A?+A*.

6-

If 6 X

e )(e a Z(e 2 (Suggestion. 2A, e,).) Prob. 27. There are given three non-coplanar lines e i e e e^ , *r e 8 ; planes cut these lines at the sum of the squares of right angles, their distances from e being constant. Show that the locus of the
triangle.
x x ,

common
if

Te

point of these three planes

is

Te%

re

(p\ e

a
1

+(/o| e 2 ) -f (p| e 3 ) =<r

3
,

1.

Art.

12.

Addition of Sects in Solid Space.


space will not
in

general intersect, so that their sum will not be, as in eq. (43), a definite line. For let p e and/ 2 e 2 be any two sects: then
1 l

Two lines

in solid

A e + P& = p,e +/>,, + e (e + 6.) - e (e + e.) - '.(e, + + (p - e )e + (/, - * )e


i
x x x

e,)

that

and

a sect passing through an arbitrary point e a plane-vector, the sum of the two in the equation. The
is,

the

sum

is

sum cannot be

A = A + xe
hence

a single sect unless the two are coplanar for let * et* e s being a vector not parallel to e^, -\-f *
;

ex x

p % e%

= p e + (/, + xe + ye + ze )e = A (*, + e + *6, (e, -f e + *e e = (A + *0 (, + e.) + ^e e


x
x

2)

and
less

this

cannot reduce to a single sect unless z

= o, that
is

is,

un-

ex

and/

e 2 are coplanar.

Since a plane-vector

a line at

420
oo
,

grassmann's space analysis.


the

[Chap.

VIIL

sum

of

two
and a
of

lines

may
.

always be presented as the sum;

of a finite line
If

line at oo

is equal to the sum of any other two, their products will also be equal, that is, the twoFor let pairs will determine tetrahedra of equal volumes.

the

sum

any two sects

-ft

Lt =s L

+ L\

then squaring

we have L

L^

=LL
3

since

L,L
the
let

o, etc.

An

infinite

number
is

of pairs of sects can be found such that

sum

of each pair

a given pair be

+A

equal to the sum of any given pair; for and take a new pair t
-ti

+ *A)(*A + *,<0 0\A +J'*A)(*\e + v e = te, +jw}?. + (*.. +y^p,e, + (x u + y v )p e + (xji +y,z\)p,e
(*iA
%

%)

..

This

will

be equal to the given pair


x

^x+^Vi = *.* +.JW

we have =i, and x u% +y vt = xt u


if
x

+y v = o).
%
t

Since there are eight arbitrary quantities with only four equations of condition, the desired result can evidently be ac~

complished in an infinite number of ways. Let p e ,/ a e 2 p n en be n sects, and and e any point, then
.

let

be their sum,

5=
the

2j>e
1

= e,2e

- e,2e + 2pe = e,2e + 2{p - e.)e,

(u 3 >

and a plane-vector as before. e )e is parallel to 2e it may be written as the prodIf 2(p uct of some vector e into ^e, that is, e'^e, when the sum becomes S = e^e -f- e'^Ee = (e -\- e')^e, a sect, because e -f- e is. In no other case does 5 reduce to a single sect. If a point.
of a sect

sum

^e
e

o.

ing S,

becomes a plane-vector. Of the two parts composthe sect will be unchanged in magnitude and direction if
5"

be moved to a new position, while the plane-vector will in It is proposed to show that a point q may general be altered.

be substituted for
dicular to ^e.

such that the plane-vector

will

be perpen-

Writing

and, for
so that

S = q2e-(q- e )2e + 2<J> e = 2e = brevity, putting q


/>,

-e
a,

)e,

2(p

e )e

\/3 r

S = ga

pa

\fi,

(114)

Art.

12.]

addition of sects in solid space.


for perpendicularity,
( ft
|

421

-we

must have
or

XI 5) pa .\a=a.p\a p. a- = \fta. ( The second member is obtained from the first by substituting in eq. (74) p for p and a for /> 3 and q in accordance with the statement at the end of Art. II. If in (1 1 5) we make p a = o,
x

pa)

a=O=
| |

by (in), ft a pa

.
|

a,

iy

will

be the vector from

to q taken perpendicularly to a,
2

say

p
Since
found.

= \aft-^-a -=q -e
i

(116)

a and

ft

are known, the

required
(75),

point has been

Multiply (115) by a; then, using


2 2 ap a = pa a
.

=a
|

ft a

\ft.a-

a a
.
|
|

ft,

whence,

substituting in (114),

S=,
This

2 + 4g.\a = ,* + <*-** .\2<.

(lift

may be called the normal form of S* The sects of this article represent completely
hence
all

the geometric

properties of forces,

that has been

shown applies

immediately to a system of forces in solid space. We have only to substitute the words force and couple for sect and planevector.
called

The

resultant action

by

Ball in his

Theory
is

of any system of forces is 5, of Screws " a wrench." The con-

dition for equilibrium

5 = o, which
and

gives at once
;

^e = o
since otherwise

we must e )e, which is an impossibility. The line q^e is the central axis of the system of forces 5. Lack of space forbids a further development of the subject,
but what has been given
adaptability of this
in this article will indicate

2(p - e )e = o have e^e = ^{p

(118)

method

the perfect to the requirements of mechanics.


e
x

Exercise

19.
a)

Reduce p
(p

e,(e

+e +
x

e )e,

+ A e = 5 to normal form. For convenience + {p 2

its

e )e,.

suppose p and

to be taken at the ends of the

common

per-

*Grassmann

(1862), Art. 346.

422

grassmann's space analysis.

[Chap.

VIIU

pendicular on/,6,
/,
<?

= =
t

and/

e2
;

and moreover
then
t

let ^

(p,

e,)

z
|

e,

r|e,

= o.

i(A

+/)

Accordingly
.

5=

e (e,

e.)

+<e,

- e,) = g(e +e,)+

fei+&gSLpS>

,^+^y

By (i5) *

-, = =

1^ = - <*
-6
(e,

J -lfc
,Dy

>

,.( e

,)|( 6i

+ ej*
5
is

^'

e,*-e, (6,+ ,)"

Hence the normal form

of

Exercise 20.

tetrahedron

e e lf e9e%

Forces are represented by the six edges of a find the S, reduce to e Q e 3 **?,, e t e lt ex e%
,
,

normal form, and consider the special case when three diedral
angles are right angles.
l

5=
t

e (e
l

+ +
ea
1

et )

= ^( +e + )+(e eXe = e M. + e + O + e e +
t t
t * 2 3

)= ^(e +e +e
3

3)

+ e e + e e + e e+ + (e -e -e
t
t

)(e 3
x

e3 e 1

e.e,

in

which

e,

=e

,.

etc.

Hence

For the rectangular tetrahedron


e3

let

e,

ai
x

e,

bi^

,,

z,

being unit normal vectors.

Then we

find

+ ^qiy
Exercise 21.
is

-l( gg

+ ^ + g 0-

A pole

50 feet high stands on the ground and

held erect by three guy-ropes symmetrically arranged about attached to its top and to pegs in the ground 50 feet from the pole. The wind blows against the pole with a pressure of
it,

50 pounds in the direction

p when
r

is

at the

bottom of

Art.

12.]

addition of sects in solid space.

423'

the pole, and


in the ratio

p
:

divides the distance between two of the pegs


find the tension

on the guys and the pressure

on the ground.
Evidently only two of the guys will be in tension let their pegs be at e and e 3 and let e% be at the top of the pole, and
; x ,

ftl

the weight of the pole.


of equilibrium
is

Then p

tl
*.

and the equation

^ 5'

(*.+')('

/)

2l\e.-p)

^
|

2 St(P ~~ e

(x+w)e IX'.-p)*' Tej


)
|

e%

ye % e,

ze,e a
%

+Tes+Te ert.)

Tes%

50,

rVl =
;

Te % e t

=
J

5 o l/i,

T(p

t{^^> -e)
if

7\

J m{e,-e,)+n{e,-e
=
U(e%
e
);

)\

50
j

Time

-\-neJ,

e =[/(e
1

it

and ea

then T(/

eQ )

Vm*

-\-

n*

mn,

because

e,-

1
c,

I,

equation of equilibrium

and e,|e = cos 120 becomes


3

\.

Hence the

v/#

+
x -\-w _

w
y
z

r2

V2

Multiply successively by et e lf e etJ and ^,, and


25

we

obtain

m
y and

-j-

w 4^2

n V2

Vm*
-f-

-\-

n*

mn

z being the tensions, and

the upward pressure.

Prob. 28. Three equal poles are set up so as to form a tripod, and are mutually perpendicular; a weight w hangs upon a rope which passes over a pulley at the top of the tripod, and thence

down under
e
. x
.

a pulley at the ground at a point

^le, in which
j

e s are at the feet of the poles,

and 2/

if

the rope

is

pulled

424
so as to raise w,

grassmann's space analysis.

[Chai

III.

show that the pressures on the

poles, supposir-

the

pulleys frictionless, are

Prob. 29. Six equal forces act along six successive edges of a

cube which do not meet a given diagonal; show that if the edges of be the magnitude of each the cube be parallel to z,, z 2 z 3 and z z if the diagonal taken be 3 ), force, then a parallel 2F\ (i
,

to

z,

S= + +
z
2

+ +

Prob. 30. Three forces whose magnitudes are 1, 2, and 3 act along three successive non-coplanar edges of a cube; show that the

normal form of

6* is

('.+tf, ,+ 3',)+*l(',+", 3.)Prob. 31. Forces act at the centroids of the faces of a tetrahedron, perpendicular and proportional to the faces on which they act, and
all

S=

+ i,-A.)(, +
all

directed inwards, or else

outwards; show that they are

in

equilibrium.

Art.

introduction.

425

Chapter IX.

VECTOR ANALYSIS AND QUATERNIONS.


By Alexander Macfarlane,
Lecturer in Electrical Engineering in Lehigh University.

Art.

1.

Introduction.

Vector Analysis is meant a space analysis in which " the vector is the fundamental idea; by Quaternions" is meant a space-analysis in which the quaternion is the fundamental

By

"

"

one whole; and in this chapter they will be treated as such, and developed so as to harmonize with one another and with the Cartesian
idea.

They

are in truth

complementary parts

of

subject to be treated is the analysis of quantities in space, whether they are vector in nature, or quaternion in nature, or of a still different nature, or are of such a kind that

Analysis.*

The

they can be adequately represented by space quantities.

Every proposition about quantities in space ought to remain true when restricted to a plane just as propositions about quantities in a plane remain true when restricted to a
;

Hence in the following articles the ascent to the straight line. of space is made through the intermediate algebra of algebra
the plane. Arts. 2-4 treat of the more restricted analysis, while Arts. 5-10 treat of the general analysis. This space analysis is a universal Cartesian analysis, in the

same manner

as algebra

is

a universal arithmetic.
it

ing an explicit notation for directed quantities,

By providenables their

general properties
particular
drical, or polar.

to be investigated independently of

system of coordinates,
It also

has this

any whether rectangular, cylinadvantage that it can express


se

*For

a discussion of the relation of Vector Analysis to Quaternions,

Nature, 189T-1893.

4^0

VECTOR ANALYSIS AND QUATERNIONS.

[CHAP. IX.

the directed quantity by a linear function of the coordinates, instead of in a roundabout way by means of a quadratic function.

The

different views of this extension of analysis

which have

been held by independent writers are titles of their works


:

briefly indicated by the

Argand, Essai sur une maniere de representer les quantites imaginaires dans les constructions geometriques, 1806.
Warren, Treatise on the geometrical representation of the square
roots of negative quantities, 1828.

Bellavitis,

Moebius, Der barycentrische Calcul, 1827. Calcolo delle Equipollenze, 1835.


lineale

Grassmann, Die

Ausdehnungslehre, 1844.

Algebra, 1849. O'Brien, Symbolic Forms derived from the conception of the translation of a directed magnitude. Philosophical Transactions,
1851.

De Morgan, Trigonometry and Double

Hamilton, Lectures on
Quaternions, 1866.
Tait,

Quaternions,

1853,

and Elements of

Elementary Treatise on Quaternions, 1867. Hankel, Vorlesungen iiber die complexen Zahlen und ihre

Functionen, 1867.
Schlegel, System der Raumlehre, 1872.

Houel, Theorie des quantites complexes, 1874. Gibbs, Elements of Vector Analysis, 1881-4. Peano, Calcolo geometrico, 1888.

Hyde, The Directional Calculus,


Heaviside, Vector Analysis, in
"

1890.

Reprint of Electrical Papers,'"

1885-92.

Macfarlane, Principles of the Algebra of Physics, 1891.

Papers

on Space Analysis, 1891-3.

An
of his

excellent synopsis "

is given by Hagen in the second volume der hoheren Mathematik." Synopsis

Art.

2.

Addition of Coplanar Vectors.


is

By
and

a "vector"

direction.

It is

a quantity which has magnitude graphically represented by a line whose

meant

ART.

2.]

ADDITION OF COPLANAR VECTORS.

42?

length represents the magnitude on some convenient scale, and whose direction coincides with or represents the direction of

Though a vector is represented by a line, its Exphysical dimensions may be different from that of a line. are a linear velocity which is of one dimension in amples
length, a directed area which is of two dimensions in length,, an axis which is of no dimensions in length.

the vector.

A vector will
magnitude by Greek letter, as

be denoted by a capital
italic letter,

italic lejter, as
its

B*

its

a small
/3.

as

b,

and
bfi,

direction
rp.

For example,

B=

R=

by Sometimes

a small

it is necessary to introduce a dot or a mark / to separate the specification of the direction from the expression for the magnitude ;f but in such simple expressions as the above, the difference is sufficiently indicated by the difference of type.

system of three mutually rectangular axes will be indicated,, as usual, by the letters i,j, k.

The

tude and direction.

analysis of a vector here supposed is that into magniAccording to Hamilton and Tait and

other writers on Quaternions, the vector is analyzed into tensor jUldjnyfcyectfir, which means that the tensor is a mere ratio
destitute of dimensions, while the unit-vector
is

the physical

magnitude. tude and direction

But

it

will

is

be found that the analysis into magnimuch more in accord with physical ideas,,
difficult to

and explains readily many things which are

explain

by the other analysis.


its components have a and are applied simultaneously; common point of application or it may be such that its components are applied in succession, each component starting from the end of its predecessor.

vector quantity

may

be such that

An example

of the former

is

found

in

two forces applied simulin

taneously at the same point, and an example of the latter


*This notation
is

found convenient by electrical writers

in

order to harmo-

nize with the Hospitalier system of symbols and abbreviations. for this purpose in the author's Note on Plane Algebra, \ The dot was used
1883; Kennelly has since used

for the

same purpose

in his electrical

papers

428

VECTOR ANALYSIS AND QUATERNIONS.


rectilinear displacements

[CHAP. IX.

two

made

in succession to

one an-

other.

Composition of Components having a common Point of Let OA and OB represent two vectors of the Application.

same kind simultaneously applied


c parallel to
join

at the point

O.

Draw BC

OC.

OA, and AC parallel to OB, and The diagonal OC represents in mag-

nitude and direction and point of application and OB. This principle the resultant of

OA

was discovered with reference


Take the
rection of
direction of

to force, but

it

applies to

any

vector quantity coming under the above conditions.

OA

for the initial direction

the di-

any

other vector will


initial
it.

angle round which the order to coincide with


/,/0, OB by 2 //V ure it follows that

be sufficiently denoted by the direction has to be turned in

Thus

OA

may be denoted by
fig-

OC

byf/J.

From

the geometry of the


,/, COS d.

/*=/,' +/' +
and
hence
tan 6
/,

+/,

cos

;
,

OC =

//

+ / + 2/,/, cos
resultant
is

/tan

^ ^"^
12

Example.
3/60
.

Let the forces applied

at a point

be 2/0 and
1

Then the
3Q'-

^4 + 9+

J-

/tan

-,

= 4- 3 6/36
If

the

first

component

is

given as/j/0 lf then

we have

the

more symmetrical formula

OC = ^+/,< + 2/,/, cos (,-*,)


When

^XXt/Xi

the components are equal, the direction of the resultant bisects the angle formed by the vectors; and the mag-

nitude of the resultant

is

nent on the bisecting

line.

twice the projection of either compoThe above formula reduces to

OC =

2/

cos

& /?. 2/2

Art.

2.]

addition of coplanar vectors.

429=

Example. The resultant of two equal alternating electromotive forces which differ 120 in phase is equal in magnitude to either and has a phase of 6o.
Given a vector and one component, to find the other component. Let OC represent the resultant, and OA the component.

Join

AC
AC.

and draw

parallel to

The

line

OB OB
it

equal and
represents is the only
gives

the component required, for line which combined with


as resultant.

,/
A
'

OA

OC

~^

The

line

hence the rule parallelogram formed " Reverse the direction of the then compound is, component, it with the given resultant to find the required component."
;

OB is identical with the by OC and OA reversed

diagonal of the

Let

f/f) be the

vector and
is

fjo one component

then the

other component

fj, = */+/,--

2//,

e9*l -./+/ 9 CM

Given the resultant and the directions of the two compoThe resultant nents, to find the magnitude of the components. is represented by OC, and the directions by OX and OY. From C draw CA parallel to OY, and CB
parallel to
off
is

OX

the lines

OA
OB

and

OB

cut

It represent the required components. evident that and when com-

OA

pounded produce the given resultant OC,


and there
is

only one set of two components which produces


;

a given resultant hence they are the only pair of having the given directions.
l

components
directions.

Let//0 be the vector and /6 and /0 the given


9

Then

/ +/. cos (0, - e =/cos {o - e \ / cos (0, - 6 +/ =/cos (0, - 0),


x

from which
ft

it

follows that
{cos "
(0

~f

)
I

cos

- cos

(0,
a

- 6) cos -6 (0,
X

(0,

430

VECTOR ANALYSIS AND QUATERNIONS.

("CHAP. IX.

For example,

let

100/60
.

/30
1

and /90 be given

then

/.=

IOO

cos 30 4- cos 6o'

Composition of any Number of Vectors applied at a common Point. The resultant may be found by the following graphic construction Take the vectors in any order, as A, B, C. From the end of A draw B' equal and parallel to B, and from the end of B' draw
:

and }B equal

parallel to C\ the vector from the beginning of A to the end of is the

resultant of the given vectors.


~2

This follows
parallelo-

by continued application

of the
is

gram

construction.

The

resultant obtained
is

ever the order; and as the order

the same, whatarbitrary, the area enclosed

has no physical meaning. The result may be obtained analytically as follows

Given

/,/#,

+// +/Jl +
%

+f

/ln.

Now
Similarly

/,/*,

=/,cos

6,/q

+/

sin

^l\
/^,

fJJK
n

=/

cos

/o

+/

sin 6,

and

f /J> = fn cos
2\f/J\

/o
6\

+fn sin 6 /n

Hence

j^/cos

/o

{^/sin

6\

/.

= V(2f cos
In the case of a

Off

(^/sin BY tan-

-^^"

mon

of simultaneous vectors applied at a compoint, the ordinary rule about the transposition of a term in

sum

For example, if A -\-B + C 0,then + C = B, and B + C = - A< etc. This is permissible because there is no real order of succession among the given components.*

A +B=

an equation holds good. .C, and A

* This does not hold true of a


sion.
It is

sum

of vectors

having a real order of succes-

a mistake to attempt to found space-analysis upon arbitrary formal

ART.

2.]

ADDITION OF COPLANAR VECTORS.

431
of

Composition of Successive Vectors.


successive vectors partakes Let tion than of addition.

The composition
A

more

of the nature of multiplicaa.

Abe

vector start-

ing from the point O, and B a vector starting from the end of A. Draw the third side OP, and from O draw a vector equal to B, and from
its

/
{^-'''

extremity a vector equal to A.

The
were

line
so, it

OP

is

not. the
also be
-\-A

complete equivalent of

-\-

the complete equivalent of

B B

if
-f-

it

would

A.

But

-f-

B and B

determine different paths; and as they go oppositely around, the areas they determine with OP have different signs. The
diagonal OP represents A ered independent of path.
vectors, the
is
-f-

only so far as
as
is

it

is

consid-

For any number


so far
it

of

successive

sum

independent of
point
of the
is

path
first

the vector from the

initial

to the final point of the last.

This

also

true
as to

when the

successive vectors

become

so small

form a continuous curve.

the curve

OPQ
The The
is

and the vector

OQ
is

area between on the path, and depends

The

has a physical meaning.


Prob.
is
i.

resultant vector

123/45

and one component


200 /75

100/0
Prob.

find the other component.


2.

velocity of a

body

in

agiven plane

is

and

one component
Prob.
virtual

100 /25

find the other

component.

3. Three alternating magnetomotive forces are of equal find the revalue, but each pair differs in phase by 120

sultant.

u ^ w( , (Ans. Zero.) ,7 Prob. 4. Find the components of the vector 100/70 in the direc-

rV^

tions 20

and ioo.
5.

Prob.

Calculate the resultant vector of 1/10

2/20

3/30

4 /40

Prob.

6.

Compound
i2o)/i2o

the following magnetic fluxes: h sin nt

-f-

sin (nt

h sin (nt

24o)/24o.

(Ans. %h/nt.)

laws; the fundamental rules

must be made

to express universal properties of the


is made to apply formal laws to an analysis of these quantities.

thing denoted.

In this chapter no attempt

directed quantities.

What

is

attempted

is

432
Prob.
7.

VECTOR ANALYSIS AND QUATERNIONS.

[CHAP. IX.

Compound two
sin nt
'
.

alternating magnetic fluxes at a point,

a cos

?it

/o and a
8.

(Ans. a /nt.)

Prob

tive forces

Find the resultant of two simple alternating electromo100/20 and 50 /75


'

Prob. 9. Prove that a uniform circular motion is obtained by compounding two equal simple harmonic motions which have the

space-phase of their angular positions equal to the supplement of the 4*m' f**"ftime-phase of their motions.
'

Art.

3.

Products of Coplanar Vectors.

When

all

the vectors considered are confined to a

common

plane, each

may

be expressed asthe
i

sum

of

two rectangular

components.
right angles to

Let

and/ denote two


;

directions in the plane at


x

one another

then

A = a i -\- a J, B = b i + bj
x

R=xi-\-j/j.

Here

and

are not unit-vectors, but rather

signs of direction.

Product of two Vectors.

Let

A = aj>-\-aJ and B = b[-\-b%f

be any two vectors, not necessarily of the same kind physically. We assume that their product is obtained by applying the
distributive law, but

we do

not assume that the order of the

factors

is

indifferent.
l

Hence

AB = (a i + aJ)(b i-\- bj) = afiji +


x

aj>jj+ aJ)jj-\-aJ?JL
algebra, that the

If

we assume,

as suggested

by ordinary

square of a sign of direction is -j-, and further that the product of two directions at right angles to one another is the direction

normal to both, then the above reduces to

AB = a b + A + OA
tf
l
l

aj>^k.

Thus the complete product breaks up into two partial a 3 b 9 which is independent of direcproducts, namely, a b which has the axis of the plane for and (a b i tion, aj)^)k
i 1

direction.*
* A common k is that i is an operator,/an explanation which is given of ij operand, and k the result. The kind of operator which i is supposed to denote is a quadrant of turning round the axis i ; it is supposed not to be an axis, but a quadrant of rotation round an axis. This explains the result ij'= k, but
unfortunately
it

does not explain

ii

for

it

would give

ii

i.

Art.

3.]

PRODUCTS OF COPLANAR vectors.

433

Scalar Product of two Vectors. By a scalar quantity is meant a quantity which has magnitude and may be positive or

negative

but

is

destitute

of

direction.

The former

partial

product is so called because it is of such a nature. It is denoted by SAB where the symbol S, being in Roman type, denotes, not a vector, but a function of the

A and B. The geometrical meanSAB is the product of A and the orthogonal projection of B upon A. Let OP and OQ represent the vectors A and B
vectors

ing of

draw

QM

and

NL

perpendicular to OP.

o^_^._^n

Then
(OP)(OM)

= (OP)(OL) + (OP)(LM),

-H#**.3L
let A denote a Corollary I. of application the velocity of its point force and then denotes the rate of working of the force. The result is the

= a b +a b SBA = SAB. For instance,


l
l

SAB

same whether the


velocity on the

force

is

projected on the velocity or the

force.

Example I. A force of 2 pounds East -f- 3 pounds North is moved with a velocity of 4 feet East per second 5 feet North second find the rate at which work is done. per

2X4+3X5= 23 foot-pounds per second. The square of any vector A = a? + a* Corollary


2.
9

a*.

independent of direction it is an essentially positive or signless quantity for whatever the direction of A, the direction
is
;
;

of the

other^ must be the same; hence the


2.

scalar

product

cannot be negative.

Example
velocity 64

feet

down per second


energy then
s

stone of 10 pounds mass is moving with a 100 feet horizontal per

second.

Its kinetic

is

(64

+ ioo

foot-poundals,

434

VECTOR ANALYSIS AND QUATERNIONS.

[CHAP. IX.

a quantity which has no direction.


s

The

kinetic energy

due to
k

the

downward

velocity '
is

is

10
2
;

64
2

and that due to thehoriis

zontal velocity
tained, not

lOO

the whole kinetic energy


addition,

ob-

by

vector, but

by simple

when

the com-

ponents are rectangular. Vector Product of two Vectors.

The

other partial product

from

its

nature

is

called the vector product,

and

is

denoted by

VAB.
product
is

Its

of

geometrical meaning is the and the projection of B which

perpendicular to A, that is, the area of the parallelogram formed upon A and B.
1

Let OP and OQ represent the vectors A and By and draw the lines indicated by the It is then evident that the area figure.
a b%
x

of the triangle s= -H^A ,,)

OPQ =

\a

a^

\b x b%

(#,

b )(b 2
x

#a ),

Thus

(a x b9

gram formed by which it lies.


It

ajb^k denotes the magnitude of the paralleloA and B and also the axis of the plane in

follows that

VBA =

VAB.

It

is

to be observed

that the coordinates of

whereas

and

Example.
inches, then

A and B are mere component vectors, B themselves are taken in a real order. Let A = (ioz -f- iff) inches and B = (5*+ l2 J)
(120

55)^ square inches; that is, 65 square inches in the plane which has the direction k for axis.

VAB =

ab cos .a/3, expressed as aa and B as bft, then SAB where aft denotes the angle between the directions a and /3.
If

is

Example.

The

effective electromotive force of


is

100 volts
8

per inch /90 along a conductor 8 inch /45 cos /45 /90 volts, that is, 800 cos 45 volts.
the direction

SAB =

100

Here /45

indicates

a and /90

the direction

ft,

and /45 /90 means


of

the angle between the direction of 45

and the direction

90

Also
-which
is

VAB= ab sin
normal to

aft

aft,

where
that

aft denotes the direction


is,

both a and

ft,

their pole.

ART.

3.]

PRODUCTS OF COPLANAR VECTORS.

435
is

Example.

At
.

a distance of 10 feet /30

there

a force of

IOO pounds /6o

The moment

is

NAB
.

we 10 X 100 sin =2 1000 sin

/30 /6o pound-feet 90 / /90

30 pound-feet 90 / /90

Here 90 / specifies the plane of the angle and /90 the The two together written as above specify the normal k.
Reciprocal of a Vector.

angle.

By
The

meant the vector which combined with the


duces the
product
-f1.

the reciprocal of a vector is original vector proreciprocal of

is

denoted

by A' a~ and
. l

Since
/3
I.

AB =

ab (cos #/?-[- sin

ap

a/3), b

must equal

must be

identical with

in

order that the product

may

be

It follows

that
I
<2

^
;

__

aj,-\-aJ ai #
1

The
let
cal,

reciprocal
2/3

and opposite vector


then
is its

is

A~\
and

In the figure
/? is its

OP =
and

be the given vector


(

OR =

OQ =
R
<

recipro%.
I
1

/3)

reciprocal

>

opposite.*
If

P
l

>

Example.
feet

A =
'

10 feet East
feet

feet

North,
x

A~

s=s

East

125

North

and

A~

= :^

r feet

125

125

East
125

feet North.

Product of the reciprocal of a vector and another vector.

A~

B = \AB, a

= ~A a A + *A + {*&

aA)<*J0}>

= - (cos aj3
* Writers

-f-

sin aft

aj3).

who

identify a vector with a quadrantal versor are logically led to

define the reciprocal of a vector as being opposite in direction as well as reciprocal in magnitude.

436

VECTOR ANALYSIS AND QUATERNIONS.

[CHAP. IX..

Hence SA~*B

-cos a '6 and

NA~ B =
X

-sin afi.ati. ' r a

B=

Product of three Coplanar Vectors. Let A = aj -f a*Jy bj -f- bj, C cj -f- c*j denote any three vectors in a

common

plane.

Then

(AB)C = =

{(*A
{<*

a A)

(*A

aJ>M(c i
x

+ cj)

A + <*A)ki* +
latter

cJ)

(*A - "A)(- v

c x j).

The former

partial product

means the vector

C
;

by the scalar product of


partial

A and B

multiplied while the

product means the comple-

mentary vector

of C multiplied by the magnitude of the vector product of A and B^ If these partial products (represented by OP
total product, the total product will be the resultant of OP and OQ.
.

and OQ) unite to form a


represented

by OR,

is also expressed by SAB C, where the the vectors to which the S refers; and more point separates, analytically by {abc cos aryfj/^

The former product

The

latter

product

is

also expressed

by (VAB)C, which
is

is

equivalent

to

V(VAB)C, because

NAB

at

right

angles

to C. It is also expressed by abc sin a/3, afly, where a/3y denotes the direction which is perpendicular to the perpendicular

to

a and
If

/?,and y.
is

the product

formed after the other mode

of association

we have

A(BC)

= (*,* + </>(Vi + V.) + = (Vi + + a*J) +

WW
aj is

W +(fo

aj)(b

WW
3/0
^4,

c%

bj )k
x

~~*iJ)

=zSBC.A +VA(VBC).
The
vector
x

aj
a 9 j\

the opposite of the complementary


latter partial

vector of a i

Hence the

product

differs

with the

mode

of association.

ExampleLet A =

i/o

+ 2/90

B=

+
B,

4/90

C=

5/0

+ 6/90

The

fourth proportional to

C is

\^>

ART.

3.]

PRODUCTS OF COPLANAR VECTORS.

437

(A-*E)C

= lX

3
l

X4
+;,
1
{

5/o!

+ 6/90
j

4-2X3 _ 6 ^ + "T 2
If

5 /go_o

=
Square

13.4/0^+11.2/90.
of a Binomial of Vectors.
it

-\-

denotes a

sum

entirely equivalent to the the square of any vector is a positive But resultant vector C. scalar, hence the square of A -\- B must be a positive scalar. of non-successive vectors,
is

Since

and

are in reality

components

of

one vector, the

square must be formed

after the rules for the products of rect-

angular components

(p. 432).

Hence

(A+By = (A+B)(A+B), = A* + AB + BA + B\ {T/th - - V& A= A* + B* + SAB + SBA + NAB + VBA,

= A' + B + 2SAB.
1

This

may

also be written in the

form L

^
;

+&

2a b cos <*#

is

But when A -\- B denotes a sum of successive vectors, there no third vector C which is the complete equivalent and consequently we need not expect the square to be a scalar quantity.

We

observe that there

is

a real order, not of the factors,


;

but of the terms in _the-binomial

this causes

both product

terms to be AB, giving

(A

+ B) =
2

A'
2

=A
The
side,

+ 2AB + B +B* + 2SAB + 2VAB.


7

scalar part gives the square of the length of the third

while the vector part gives four times the area included
side.

between the path and the third

Square of a Trinomial of Coplanar Vectors.

Let

-\-

B -{-

C denote a sum
;

of successive vectors.

The product terms must


of the vectors in the triC,

be formed so as to preserve the order nomial that is, A is prior to B and

and

is

prior to C.

438

VECTOR ANALYSIS AND QUATERNIONS.

[CHAP. IX.

Hence
{A

+ B + Cf = A' + B* + C + 2AB + 2AC + 2BC,


= A* + B* + C' + 2(SAB + SAC + SBC), + 2(VAB + VA C + VBC). S(A+B+Cy = (i)
<*/*
2

(1)
(2)

Hence

= c? + &* + f + 2 ^ cos + 2 ^ cos ^r + 2 b cos /^y and V(^+Z?+0 = (2) = \2ab sin a/? + 2ac sin a;/ + 2^ sin /3y\. a
ft

The

scalar part gives the square of the vector from the bec ginning of A to the end of C and is all that exists

when
'

the vectors are non-successive.

The

vector

B part
A

is

four times the area included between the

successive sides

and the resultant side of the

Note that

polygon. it is here assumed that

V(A

+ B)C = VAC +

VBC, which

is

the theorem of moments.

Also that the prod-

uct terms are not formed in cyclical order, but in accordance with the order of the vectors in the trinomial.

Example.

Let

3/0*

B=

S /30

C=

7 /4$

find the

area of the polygon.

iV(AB+ AC + BC), = i{ 15 sin [o /30 + 21 sin /o /45 + 35 = 375 + 742 + 4.53 = 157.
Prob.
10.

sin

/30 /45

},

of 1000 dynes

At a distance of 25 centimeters /20 there /8o; find the moment.

is

a force

Prob. 11. A conductor in an armature has a velocity of 240 inches per second /300 and the magnetic flux is 50,000 lines per square inch /o; find the vector product.
(Ans. 1.04

io T lines per inch per second.)

Prob. 12. Find


directions 0.8141 E.

the sine

and cosine of the angle between the 0.5807 N., and 0.5060 E. 0.8625 N.

Prob. 13.
10 feet /30
,

When
what

a force of
is

200 pounds /270

is
?

displaced

by
the

the work done (scalar product) of the negative sign in the scalar product ? meaning

What

is

Art.

4.]

coaxial quaternions.

439

Prob. 14.
feet E. per

A mass of 100 pounds is moving with a velocity of 30 second feet SE. per second; find its kinetic 5 energy,

Prob.
feet

15.
;

force of

10 pounds /45

is

acting at the end of

8-

/200

find the torque, or vector product.

Prob. 16.

The

radius of curvature of a curve

is

2/0
-J-

+ 5/90
.17/90
.)

find the curvature.

(Ans. .03/0
the

Prob.

17.

Find

fourth

8/o 3 /9o_, and 6/o_ 5 /9o_. Prob. 18. Find the area of the polygon whose successive sides are 10 /30 9/1 00 8/180 7/2 2 5
,
,

proportional

to

10/0

+ 2/90

Art.

4.

Coaxial Quaternions.
"

is meant the operator which a " quaternion changes one vector into another. It is composed of a magnitude and

By

a turning factor. The magnitude may or may not be a mere for ratio, that is, a quantity destitute of physical dimensions the two vectors may or may not be of the same physical kind.
;

The

turning

is

in a plane, that is to say,

it is

not conical.
a

For
;

the present all hence all the quaternions considered have a

the vectors considered

lie in

common plane common axis.*

Let

A
A

and

be two coinitial vectors

the direction normal

to the plane

may

be denoted by

fi.

The operator which

changes

into

consists of a scalar multiplier

and a turning round the axis /3. Let the former be denoted by r and the latter by fi e where denotes Thus R = r/3 eA and recipthe angle in radians.
,

#/

^^

par

rocally 7

A=

e -/3~ R. r

Also
9

Lr = r/3 A

and

A =

-/3~

may be expressed as the sum of two one of which has a zero angle and the component operators, other an angle of a quadrant. Thus
The
turning factor
ft
e

j3

cos B

/5

+ sin 6

p*/\

* The idea of the "quaternion" is due to Hamilton. Its importance maybe judged from the fact that it has made solid trigonometrical analysis possible.
It is the

cally

most important key to the extension of analysis to space. Etymologi"quaternion" means defined by four elements; which is true in space in plane analysis it is defined by two.

440

VECTOR ANALYSIS AND QUATERNIONS.

[CHAP. IX.

the angle is naught, the turning-factor may be but the above form shows that the equation is homogeneous, and expresses nothing but the equivalence of a

When
;

omitted

given quaternion to two component quaternions.*

Hence
and

= r cos 6 + r sin ^p + q.F* r/3A = pA -\-q/3"/A = pa a qa


rfi
.

2
.

p"'

-{-

(3

n/2 a.

The

relations

between r and

6,

and

/ and
1

q,

are given

by

r=
Let

Vp*

+ q\

tan-

^.

denote a sine alternating electromotive Example. force in magnitude and phase, and / the alternating current in

magitude and phase, then

E= (r + 2nnl. fi*/%
(3

where r

is

the resistance, / the self-induction, n the alternations

per unit of time,


sentation.

and

denotes the axis of the plane of repre'

It follows that

E rl -\- 2nnl J-'E = r-{- 2nnl


.

fi"/*I;

also that

/?*/*,

is, the operator which changes the current into the electromotive force is a quaternion. The resistance is the scalar part of the quaternion, and the inductance is the vector part.

that

Components

of

the

Reciprocal of a Quaternion.
l

Given

R=(p + q.^/*)A,
then

A=

p + q./F/*
(P

+ q.p"*)(p-q.p*) R _p q. p R
,%

p-g.p/*

* In
for |/

the
i.

method of complex numbers

(3*/z is

expressed by

i,

which stands
it

The advantages

of using the

above notation are that

is

capable

of being applied to space, and that it also serves to specify the general turning factor /3 e as well as the quadrantal turning factor ft*/*.

Art. 4

coaxial quaternions.

441

Example.
that from

portant to obtain

E=

Take the same application as above. It is im/ in terms of E. By the above we deduce
(r

n 2tt;//. /3 /*)I

w+
their resultant to the

(27tn/y

+ (27tniy-If

Addition of Coaxial Quaternions.


several vectors to a constant vector

the ratio of each of


given, the ratio of
is

is

same constant vector


Thus,
if

obtained by tak-

ing the

sum

of the ratios.

then

= (A + Qn /3"*)A, 2 R = 2p + (2?) /**/} -4,


&n
{
.

and

reciprocally

A ~
(2py
Example.
In the case of

+ (2gy *"
a

compound
"
r%

circuit

composed

of a number of simple circuits

in parallel
3

- r -innl ^_ r
x

.p*l*
i
I

- 2nnl
a

p"*
t
'

'

_|_

(2^)V

r2

+ (2)V,

'

therefore,

2;/= 2
\
[

^-=-

E
r wji'P''* r -\-(27tnyi*
i
j

=
and

2[ a \r

-\-(2 7tnyi )

~\
a

2nn2rT1

>

reciprocally

^ -

s(
i

WJ +

>r^(
/

/W*

2 r^ \

t y y + (***? \2 r> + (2xnyr) 7 + {2 rnyH

21*

Product of Coaxial Quaternions. If the quaternions which change A to R, and R to R' are given, the quaternion which
y

changes

to R'

is

obtained by taking the product of the given

quaternions.
*This theorem was discovered by Lord Rayleigh; Philosophical Magazine,

May,

886.

See also Bedell

&

Crehore's Alternating Currents,

p. 238.

442

VECTOR ANALYSIS AND QUATERNIONS.


Given

[CHAP. IX.

and
then R'

rr>'/3 e

+ q. P )A R' = r''R = (/ + p*/*)R, + A = {(// + (pq' +/q)


r/PA
(p
n/2

R =
e'

q' .

qq')

ft*) A?

Note that the product

is

formed by taking the product of

the magnitudes, and likewise the product of the turning facThe angles are summed because they are indices of the tors.

common

base

/?.*

Quotient of two Coaxial Quaternions. ternions are those which change A to R, and

If

the given quato R' then that


y

which changes
Given

to R'

is

obtained by taking the quotient of

the latter by the former.

and
then

R = r/FA = (p + g. fi"/*)A = r'fi*A = (/ + q' F'*)A R' = - p '-*R


R!
.

= (P' + t.
_ (//
Prob.
19.

+ qq') + {pq'-P'q) P* P f+
q>
;

^-%p

axei
R,

alternating electromotive force is 200 volts, the resistance of the circuit is 10 ohms, the self-induction is required the 3-J-0 henry, and there are 60 alternations per second
current.

The impressed

(Ans. 18.7 amperes /


If in the

20

42'.)

r=-

fo<

Prob. 20.

above

circuit the current

is

10 amperes, find

the impressed voltage. Prob. 21. If the electromotive force

is

no

volts

/B and the cur-

rent is 10 amperes /0 \n, find the resistance and the self-inducthere being 120 alternations per second. tion, number of coils having resistances r v rti etc., and Prob. 22.

/2 find the impressed etc., are placed in series electromotive force in terms of the current, and reciprocally.

self-inductions lx

*Many writers, such as Hayward in "Vector Algebra and Trigonometry," and Stringham in " Uniplanar Algebra," treat this product of coaxial quaternions as if it were the product of vectors. This is the fundamental error in the
Argand method.

Art.

5.]

addition of vectors in space.

44$

Art.

5.

Addition of Vectors in Space.

space can be expressed in terms of three independent components, and when these form a rectangular set the directions of resolution are expressed by i,j\ k. Any variable vector

A vector in

R may

any constant vector

be expressed as R ~ rp = xi-\-yj -\-zk, and B may be expressed as


l>/3

B=
In space the

= bj + bj-\-b,L
for the direction involves

symbol p

two

ele-

ments.

It

may

be specified as

P= s + y + ir
where the three squares are subject
to the condition that their

xi

+ yj + zk
specified

sum

is

unity.

Or

it

may be

by

this notation, <p//0,

a generalization of the notation for a plane.

The

additional

angle

0/
we

is

introduced to specify the plane

in

which the angle

from the
If

initial line lies.

are given
:

in the

form r<p//0, then we deduce the

other form thus

R = r cos 6 i -f- r sin If R is given in the form


.

6 cos <p.j-\-r sin 6 sin

k.

+ yj'+ zk, we deduce R= Vx* +/ + tan- j tan ^f +


xi
l

z*

z>

For example, aa io cos 45.

B
i

io

30//45_
cos 30 ./

-f io sin 45

+ 10 sin 45
// //

sin

30

Again, from
.*.

32

+ 4/+

$k we deduce
1

##
V9

+ 16 + 25 tan"
.

5 ~

//

bn" tan

1 ,

V41 -L-

=
To
applied at a
vectors or,

6 7.07 ,5i.4/ / 4-9

find the resultant of

any number of component vectors

common'

point, let

R R
lt

i?M

represent the

444

VECTOR ANALYSIS AND QUATERNIONS.

[CHAP. IX.

then

and

<