You are on page 1of 110

A . I.

MARKUSHEVICH

THE REMARKABLE
SINE FUNCTIONS
THE REM ARKABLE
SIN E FU N C T IO N S

BY

A. I. MARKUSHEVICH
Moscow State University

Translated by Scripta Technica, Inc.

Translation Editor: Leon Ehrenpreis


Courant Institute of Mathematical Sciences
New York University

NEW YORK

AMERICAN ELSEVIER PUBLISHING COMPANY INC.


1966
Originally published as
z a m e c h a t e l 'n y y e s in u s y
Nauka Press, Moscow, 1965

SOLE DISTRIBUTORS FO R G REAT BRITA IN


ELSEVIER PUBLISHING COMPANY, LTD.
Barking, Essex, England

SOLE DISTRIBUTORS FO R T H E C O N T IN E N T O F EU R O PE
ELSEVIER PUBLISHING COMPANY
Amsterdam, The Netherlands

Library of Congress Catalog Card Number: 66-29781

CO PY RIG H T © 1966 BY A M E R IC A N ELSEV IER PU B L ISH IN G C O M P A N Y , IN C .

A L L RIGHTS RESERVED. TH IS BOOK OR ANY P A R T T H E R E O F


M U S T N O T B E REPRO DU CED IN ANY F O R M W IT H O U T T H E W R IT T E N
P E R M ISS IO N O F T H E P U B L ISH E R , A M E R IC A N ELSEVIER P U B L ISH IN G
C O M P A N Y , IN C ., 52 V AN D ERBILT A V EN U E, N E W YORK, N . Y. 10017

M A N U FA C TU R ED IN T H E U N IT ED STATES O F A M ER IC A
ANNOTATION

The p resen t book rep resen ts an attem pt at uniform co n sid er­


ation of the various sin e s (the circu la r and hyperbolic s in e s ,
the le m n isc a te sin e, and the sin e amplitude of Jacobi) as sp ecia l
c a s e s of the s o -c a lle d gen eralized s in e —a ll of them functions
defined as the in v er se of so m e integral. It req u ires a certain
amount of m athem atical training and is intended for su fficien tly
prepared read ers who have m astered m athem atical a n a ly sis as
taught in u n iversity c o u r se s.

v
CONTENTS___________________________________________________

Preface ix

1. GEOMETRIC DEFINITION OF CIRCULAR, HYPERBOLIC AND


IJiMNISCATE FUNCTIONS 1

2. GENERALIZED SINES 13

3. INTEGRATION IN THE COMPLEX PLANE 25

4. ELLER’S METHOD OF DERIVING THE ADDITION THEOREMS 41

5. FURTHER STUDY OF COMPLEX VALLES OF THE ARGUMENT 49

6. ZEROS AND POLES. SIMPLE AND DOUBLE PERIODICITY.


THE CONCEPT OF AN ELLIPTIC FUNCTION 73

Index 99

VII
PREFACE

Of the functions studied in elem en tary m ath em atics, the


trigon om etric functions stand out by virtue of th eir g eo m etrica l
d efinition. Without stopping for inconsequential variations of the
sa m e id ea, we m ay sa y that the sin e and co sin e are introduced
as coordinates of a point on the unit c ir c le . The independent
variab le is treated as an angle or arc of a c ir c le .
In the p resen t book, we sh a ll show how it is p o ssib le , by
beginning with other cu rves (such as the eq u ilateral hyperbola
or B ern ou lli’s lem n isca te (a curve having the form of a fig u re-
eight), to define in tere stin g and im portant functions analogous
to the trigon om etric functions, sim ila r to them in so m e resp ects
but p o s s e s s in g certa in new c h a r a c te r istic s. T h ese functions are
ca lled r e sp ectiv ely h yp er bo li c and le m n is c a te functions. In
analogy with them , we sh a ll refer to the trigon om etric functions
as c i r c u l a r functions.
We sh a ll con sid er a ll th ese functions as sp e c ia l c a s e s of the
g e n e r a l i z e d sine, that is , the function in v erse to the function
defined by an in tegral of the form

H ere, the circu la r sin e correspon d s to the c a se m = — I, n — 0;


the hyperbolic sin e correspon d s to the c a se m = l,rt = 0; the
lem n isca te sin e correspon d s to the c a se m = 0, « — — I. If
m — 1 — k2 and n = k7 (for 0 < k < 1), we obtain for the function
in v e r se to that defined by the in tegral the s o -c a lle d sine a m p l i ­
tude of Jacob i. The problem of the pendulum leads to a study of
this la st function.

IX
X PREFACE

To study the p rop erties of a ll th ese functions in a uniform


way, we need fir s t of all to define them as functions of a com ­
plex variable and then esta b lish an addition theorem for them .
This path was already taken at the end of the eighteenth century
by the young Gauss in his Tagebuch.
The reader is assum ed to have a fa m ilia rity with the e le ­
m ents of analytic geom etry and d ifferen tial and in tegral calcu lu s.
The n ecessa ry m a teria l on integration in the com plex plane w ill
be given in the p resen t book though proofs w ill be om itted.
The ultim ate purpose of the book is to acquaint the reader
not p o s s e s s in g an ex ten siv e knowledge of the theory of functions
of a com plex variable with the sim p le st rep resen ta tiv es of the
c la s s of ellip tic functions, nam ely, lem n isca te functions and the
som ew hat m ore gen eral Jacobian ellip tic functions.
In con clusion , we warn the reader that this book is not in ­
tended for light reading. He m ust read it with his p encil in his
hand.

The Author
CHAPTER 1

G eom etric D efinition o f Circular, H yperbolic,


and L e m n i s c a l e F u n c t i o n s

1. The circu la r functions are defined and their p rop erties


are d iscu sse d in h igh -sch ool m athem atics c la s s e s . H ere, we
sh a ll b riefly expound fa m ilia r facts in a form that w ill be con­
venient for us to bring out the s im ila r itie s and d is s im ila r itie s
sharply.
Let us begin with the unit c ir c le

*2+ y2= b (i)

The arc length from the point C to a point A (see Fig. 1) is


defined up to an in tegral m ultiple of the total circu m feren ce of
the c ir c le , nam ely, 2nk. H ere, k is the number of com p lete
revolutions about the coordinate origin d escrib ed by a point on

Fic. l.
2 THE R E M A R K A B L E S I N E F U N C T I O N S

the c ir c le as it m oves from C to A. We take the counter­


clo ck w ise d irection as p o sitiv e.
The coordinates x and y of the point A are independent of the
number of com plete revolutions m ade. T h erefo re, they are
p eriod ic functions of t with period 2n. By definition,

x — cos t, y = s in i. (2)

From this it follow s that, for arbitrary t,

cos21 -)- sin21= 1. (3)

It follow s from the sym m etry of the c ir c le about the jc-axis


that

cos(— t') = cos t t sin(— t) — — sini. (4)

If we a lso u se the sym m etry about the b isecto r of the fir s t and
third quadrants (by virtue of which the a b sc issa and ordinate of
any point on the unit c ir c le are interchanged), we s e e that

cos ( £ — / j s i n /, sin[— — t j — cos t. (5)

Each of the form ulas (5) enables us to obtain a definition of


eith er of th ese two functions in term s of the other; for exam p le,
we can define cos t in term s of sin t (assum ed already defined).
As consequences of form ulas (4) and (5), we can d eriv e a ll the
sim p le “ reduction” form u las. Thus, for exam p le, we obtain the
follow ing p airs of form ulas; the derivation of each pair after
the fir s t u ses the p reced in g form ulas:

cos (— -j- tj = cos — (— oj = Sin (— t) — — sin t,

sin(— f /j = sin [y —( —/)j — cos(— t) = cos t\

cos (n —t) — cos [y f ^y —Ojj = — sin (— — — — cos t,

sin (n - t)=r sin ^ -)-■( y — i'jj — cos ^y — sin t\


cos (rt -f- /) cos [.t — (— t)\ — — cos (— t) = — cos t,
sin (rt -? t) —■sin [rt — (— 01 = sin (— t ) = — sin i;
cos (2n -j~ /) = cos [rt 4- (n f- /)] — — cos (n -)- t) = cos t,
sin ( 2rt -j- t) = sin [n 4- (rt -f~ i)[ — — siri (ji -j- t) = sin t.
CIRCULAR, HYPERBOLIC, AND L E M N I S C A T E F U N C T I O N S 3

We note that the secon d pair of form ulas can be d erived im m e­


d iately from the sym m etry of the c ir c le about the y -a x is and
that the third can be derived from the sym m etry about the co ­
ordinate origin. The fourth pair e x p r e sse s the p erio d icity ,
m entioned above, of the circu la r functions.
Defintion (2) and equation (1) enable us to answ er the question
regarding the number and nature of the solution of equations of
the form

sin t — a or cos t — a.

w here a is a rea l number.


For exam p le, con sid er the equation

sin t = a. (6)

We may a s se r t that, for |a | > 1, this equation has no solution


at all. For |u | < I, th ere e x ists at le a st one solution in the
in terval 0 < t < 2n. S p ecifica lly ,

if a = 1, the only solution is t = —;

if 0 < a < 1, th ere are two solutions q and t2;


fu rth erm ore, 0 < q < — and — < t.2 = n — q < jx;

if u — 0, there a re two so lu tio n s, nam ely, q — 0 and t2 — n;

if —1 < u < 0, th ere are two solutions q and t2\


fu rth erm ore, n < q < — and — < q = 3n — q < 2rr;

if a = — 1, there is only the solution t =

Of co u rse, in the c a s e s of a equal to 1 or —1, we can speak


of two solu tion s that coincide. If q + t2, the corresponding
valu es of the co sin e are negatives of each other:

COS q — cos t2.

G eom etrically, solution of equation (6) red u ces to finding points


on the c ir c le (1) of given ordinate y —
- a and the evaluation of the
lengths of the a rcs term inating at the points found (see Fig. 2).
4 THE R E M A R K A B L E S I N E F U N C T I O N S

A ll that has been said in the p resen t sectio n is w ell illu strated
on the graphs of sin t and cos t. The construction of th ese graphs
depends som ewhat on the definitions accepted (see Fig. 3).

FIG. 2. FIG. 3.

We note that we might interpret the param eter t in the d efin i­


tion (2) g eom etrica lly not only as the arc length CA but a lso as
tw ice the area of the secto r OCA:

1= 2 area OCA. (7)

H ere, we need to regard the secto r as produced by a m oving


radius displaced from the position OC to the p osition OA. If the
radius undergoes k com plete rotations in the p ro cess (where k
can be eith er negative or p ositive (according to the d irection of
the rotation), the area swept by it w ill be equal to kx. Thus, we
have the term 2kn in the definition of t given by form ula (7).
2. Now, instead of the unit c ir c le , let us con sid er the unit
eq u ilateral hyperbola

x2 — y2 — 1. (8)

Equations (1) and (8) resem b le each other much m ore than do
the cu rves rep resen tin g them (cf. F ig s. 1 and 4). In fact, this
is putting it m ildly. Equations (1) and (8) differ from each other
only by the sign in front of y2.. On the other hand, the cbrves
shown in F igures 1 and 4 are not the le a st bit like each other.
Y et, it is the equations that we m ust tru st rather than our e y e s .
The sim ila rity in the equations en tails a fa r-rea ch in g sim ila rity
in the p rop erties of the cu rv es, which we d isco v er when we
com pare the circu la r with the hyperbolic functions.
CIRCULAR, HYPERBOLIC, AND L E M N I S C A T E F U N C T I O N S 5

FIG. 4.

Let us denote by t tw ice the area of the se c to r OCA:

t = 2 area OCA\

We sh all con sid er the area of OCA p ositive if the rotation from
OC to OA is ca rried out in the counterclock w ise d irection and
we sh all con sid er it negative in the opposite c a s e . If the point A
d e sc r ib e s the right branch of the hyperbola shown in F ig. 4
as y in c r e a se s from — co to -f-co, then t in c r e a se s a lso from
— oo to -j- co. The coordinates x and y of the point A can be
regarded as sin g le-v a lu ed functions of t. We define

x = coshf, y=-sinh/. (9)

H ere, the abbreviations “ c o sh ” and “ sin h ” are for the e x p r e s ­


sion s consinus hyper bol ic us (hyperbolic cosin e) and sinus h y p e r -
bolicus (hyperbolic sin e).
It follow s from the definition that, for arbitrary t.

cosh2<—sinh2<= I. ( 10 )

U sing the sym m etry of a hyperbola about the x -a x is , we


conclude that

cosh .(—[) =cosh l, ■- sinh (—/) = —sinh t ( 11 )

(cf. the graphs of cosh t and sinh * in Fig. 5).


The equation

sinh (t) - a ( 12 )
6 THE R E M A R K A B L E S I N E F U N C T I O N S

has one and only one root for arbitrary rea l a. The equation
cosh t a (13)
has no root at a ll if a < 1, one root (nam ely, if a — 1, and
two roots t x and /2 (---— ?,) if a > 1.
G eom etrically, the solution of equations (12) and (13) can be
reduced to finding points of given ordinate y or given a b sc issa
v > l on the right branch of the hyperbola (10) and to the evalua­
tion of the area s of the s e c to r s corresponding to the points
found (Fig. 6).

3. B efore defining the lem n isca te functions, we note that it


is p o ssib le to give to the circu la r functions a g eo m etric in terp re­
tation different from the interpretation given in sectio n 1.
S p ecifica lly , let us look at the c ir c le of unit d iam eter tangent
to the x -a x is at the origin and lying in the upper half-plane:
x2-|-y 2 — y = 0 (see Fig, 7). O bviously, the chord 0/1 connecting

the end points of the arc 0.4 of length t is equal to sin t. T h ere­
fo re, it would be p o ssib le to construct the theory of circu la r
functions by defining
CIRCULAR, HYPERBOLIC, AND L E M N I S C A T E F U N C T I O N S 7

sin t — CM, t = length 0 4 ] 'I - °^(1.4)

It follow s from this definition that sin t in c r e a se s from 0 to 1


as t in c r e a se s from 0 to rt/2. As t in c r e a se s beyond n/2 , the
sin e d e c r e a se s and van ish es as t = n. (Under the conditions
stated , n is the circu m feren ce of the c ir c le .)
Let us com plete our definition by ag reein g that sin t changes
sign every tim e that it p a sse s through the value 0 as t in c r e a se s.
Then, we need to take the length of OA in form ula (14) with the
m inus sign for n < t < 2a and with the plus sign for 2a < / < 3it,
etc. Only under such a supplem entary convention w ill our new
definition coincide with the preceding one, and only then, in
p articu lar, w ill the period of the sin e be 2n (rather than n as
would be the c a se if we con sid ered the length of the chord OA
p o sitiv e for all p o ssib le values of the arc OA, which would then
differ by in tegra l m u ltiples of a).
The second circu la r function, the co sin e, is defined in term s
of the sin e by the equation

cos/ = s i n ^ — t'j. (15)

4. Let us look at B ern ou lli’s lem n isca te shown in Fig. 8.


This is the curve co n sistin g of all points in the plane such that
the product of th eir d ista n ces from two given points, called the
foci of the le m n isc a te , is constant. In the p resen t c a s e , the
points (1/ 2 , 1/ 2) and (—1/ 2 , —1/ 2) are chosen as the foci and
the constant d istan ce product is 1 /2 . T h erefore, the equation
of the lem n isca te is of the form

x
V ) 2‘

o r, after squaring both sid e s and sim p lifyin g,

(x2 -f- y2/ — 2xy 0. (16)

Thus, B ern ou lli’s lem n isca te is a fou rth -ord er a lgeb raic curve.
Let us introduce polar coordinates r and q , taking the x -a x is
as the polar a x is. Then,

x — r cos <(■. y —r sin fp.

Thus, the equation of the lem n isca te in polar coordinates


becom es
8 THE R E M A R K A B L E S I N E F U N C T I O N S

r2 = sin2<p. (17)

The variab le r attains its m axim um v a lu e,n a m ely , 1 at rp— ji/4 .


It follow s from equation (17) (or (16)) that the lem n isca te
is sy m m etric about the coordinate origin. F u rtherm ore, it is
sym m etric about each of the diagonal lin es b isectin g the angles
between the coordinate axes: if x = a and y — b sa tisfy equation
(16), s o do the p airs (— a, — b), (b, a), and(— b, — a).

5. Suppose that a m ovable point /VI sta r ts at the origin and


m oves in such a way as to d escrib e that portion of the lem n isca te
lying in the fir s t quadrant, m oving in the p o sitiv e (i.e ., counter­
clock w ise) d irection . Since it is being d isp laced along an arc
tangent to the y -a x is when it fir s t returns to the origin , it is
natural to require that it be m oving along an arc tangent to that
axis as it en ters the next (third) quadrant in the cou rse of its
m otion. A ccordingly, we m ust take the clo ck w ise d irection for
the p ositive d irection of m otion of the point as it m oves around
that portion of the lem n isca te in the third quadrant. When the
point is finishing its fir st com plete trip over the en tire le m n is­
cate and is beginning a new one, the point lea v es the third and
en ters the fir s t quadrant along an arc tangent to the .v-axis. The
m otion thus d escrib ed is then repeated.
Let us denote the length of the en tire lem n isca te by 2<o. Thus,
the length of the arc OMC is equal to « /2 . Just as in the ca se
of a c ir c le , we sh a ll a ssu m e that the variable point M m oves
from an in itial point 0 of the arc to its term in al point A after
som e number of com p lete co u rses over the en tire lem n isca te
in eith er the p o sitiv e or the negative d irection . Thus, the length
t of an arb itrary arc OA is determ ined only up to an in tegral
m ultiple of 2 m . T h erefo re, r, the length of the chord OA, is a
p eriod ic function of t with period 2 m .
CIRCULAR, HYPERBOLIC, A ND L E M N I S C A T E F U N C T I O N S 9

We com plete the definition of this function by a g reein g that it


changes sign ev ery tim e it p a sse s through the origin with con­
tinuous in c r e a se in t.
In analogy with the circu la r sin e, we sh a ll r e fe r to the func­
tion that we have just defined for all rea l values of t as the
le m n isc a te sine and we sh a ll denote it as follow s:

O A — r — s\t. (18)
H ere, the le tte r s " s i ” are the in itia ls of the ex p ressio n sinus
le m n is ca ti cu s. We note in particular that

S l|= r l . (18')

From the sym m etry of the lem n isca te about the origin and
the convention adopted regarding sig n , it follow s that s i t is an
odd function (see F ig. 8):

sl(— () = - s l t (19)
From the sym m etry about the b isecto r of the fir s t and third
quadrants, we have (see Fig. 9)

Sl (to — t) = Si/. (20)

FIG. 9. ■

A nalogously, from the sym m etry about the b isecto r of the


second and fourth quadrants, we have (see F ig. 9)
-Sr
'' si (co-H) ■= — Sl/. (21)

A ctually, (21) is a sim p le consequence of equations (19) and (20):

sl (ft) -)-£) = Sl [ft) — (— 0] ~ sl (— t) = — sl t.


10 THE R E M A R K A B L E S I N E F U N C T I O N S

The lem n iscate co sin e is defined in term s of the sin e by

Clt = si f-J - /). (22)

From th is, it follow s in particular that it, too, is a p eriodic


function with period 2<o. However, in contrast with s i t, the
function cl t is an even function. To s e e th is, note that, by using
s u c c e s s iv e ly (22), (21), (19), and again (22), we have

cl (— t) ----- si (— + /) = — si -I- /) = si (23)

The read er can e a sily verify the follow ing id en tities:

cl (to — t) — — cl/, cl ((0 -)■ t) — cl t. (24)

The definition of the lem n isca te functions and equation (16)


enable us to answ er the question as to the number and nature
of solutions of equations of the form s i t = a or cl t = a, where
a is a real number.
H ere, we have a p erfect analogy with the ca se of circu la r
functions (cf. sectio n 1).
F or exam p le, let us con sid er the equation

si t a. (25)

We may a s se r t that, for |u | > 1, it has no solution at a ll. For


| a | < 1, there e x ists at le a st one solution in the in terval 0 <; t < 2co.
S p ecifically,

if a — 1, th ere is the sin g le solution t = y ;

if 0 < a < 1, there are two so lu tio n s, and /2;

th ese are such that 0 < t x < -- and y < t2 = a>— /, < to;

if a = 0, there are two solutions t x— 0 and t2 — to;

if — 1 < a < 0 , there are two solu tion s q and t2;

th ese are such that < t x< - - an d^ - < t2 = 3to — q < 20 ;

if a = — 1, th ere is the sin g le solution i = - y .


CIRCULAR, HYPERBOLIC, AND L E M N IS C A T E F UN CT IO NS II

O bviously, in each of the c a se s a = 1 and a = — 1, we can speak


of two solu tion s that coin cid e.
We note a lso that, if q j=t7, the corresponding values of c l t
are negatives of each other:

cl q = — cl tT

G eom etrically, solution of equation (25) red u ces to finding the


points of in tersectio n of the lem n isca te (16) and the c ir c le
x 1 + y2 = a2, to the ch o ice of those points on the in tersec tio n for
which r = OA has the sa m e sign as a, and, fin ally, to the c a l­
culation of the corresponding arc lengths.
A ll that has been said in the p resen t sectio n is illu stra ted in
the graphs of s i t and c l t, which, in appearance, resem b le the
sin e and co sin e graphs (se e Fig. 10).

FIG. 10.
CHAPTER 2

G eneralized Sines

6 . Let us now show that a uniform approach is p o ssib le to


the study of the sin e s (and with them the co sin es) defined
g eo m etrica lly in Chapter 1. With this end, we shall c o n s id e r
each type o f sine a s a function i n v e r s e to s o m e integral.
Let us begin with the circu la r sin e. At the end of sectio n 1,
it was noted that the variable t can be interpreted as tw ice the
area of the circu la r se c to r OAC (see Fig. 1). If x and y — sin i
are the coordinates of a point A, then x — \ 1 — y- and, con­
sequently,

area OAK — —y \ ! 1 — y!.

The area of the cu rv ilin ea r trapezoid OCAK is calculated from


the form ula
y
area OCAK ~ f I (!r\ j V 1 _ h2^rl-
0

Integrating by parts, we obtain

y y

y y

13
14 THE R E M A R K A B L E S I N E F U N C T I O N S

from which we get

Consequently,

y
S-= area OCA ^ area OCAK —area

or

(26)

As the variable y in c r e a se s from — 1 to 1, the variable t in­


c r e a se s from — n / 2 to n / 2 . We can calcu late n / 2 from the
form ula

(27)

C on versely, as t in c r e a se s from —n / 2 to n / 2 , the variable


y — sin / in c r e a se s from —1 to 1. Thus, we can regard the func­
tion y = si n/ in the in terval (—n / 2 , n / 2) as the function in v erse
to the in tegral (2 6 ). As before, the function cos t is defined by
the form ula

7. Let us npw look at the hyperbolic sin e sinh t. In defining


it g eo m etrica lly (cf. sectio n 2), we interpreted the variable t
as tw ice the area of the se c to r OCA (see F ig. 4). We have

S = a re a OCA —area OCAK —area OAK.

By using the equation for a hyperbola x2 — y2 = 1, we find

area OAK — -^xy = - ^ Y l - { - y ' 2y


GENERALIZED SINES 15

and

y y
area O C A K = j ’ = J Y 1 —f- Tl2 ^t|-
o o

Integration by parts y ield s

y y
j yr+^dr]=YnrY y - j ^ = - ^ =

dr\
= V 1-+■ y3y — I V1+ h2dn + / V\

so that

y y
area OCAK = f 4x1 = ^ ] / T T f y + ± f
h n V ~t *]

C onsequently,

5= 1f dT>
2 J0 yr+y

or

t = 2S = di\ (28)
I /I T ?

As the variable y in c r e a se s from - o c to -+- o o , the variable t


a lso in c r e a se s from — o o to - j - o o sin ce

f <ni
J/ Vi
r + if -- °°-

C on versely, as t in c r e a se s from — cc to ~(-co, the variab le y


a lso in c r e a se s from — o o to + o o . Thus, we can regard the
function y = sin h t on the en tire real axis (— o o , + o o ) as the
function in v erse to the in tegral (28).
16 THE R E M A R K A B L E S I N E F U N C T I O N S

8. In what fo llo w s, we sh all ch oose a procedure of exp osition


in which the in teg ra ls (26) and (28) w ill be the original functions,
and the functions sin t and sinh t w ill be secon d ary, that is ,
obtained from the preceding. We sh a ll define the la tter functions
as the in v e r se s of the in teg ra ls. Of co u rse, when such d efin i­
tions are m ade, the in teg ra ls th em selv es can be regarded as the
functions in v erse to sin t and sinh t and we can introduce the
follow ing notations for them:

y v
t~ j - dr* —arcsiny and t = / - . =arsinhy. ( )

In the fir s t c a s e , the sy lla b le “ a r c ” is appropriate because


tw ice the area of the se c to r OAC in the unit c ir c le is m easured
by the sam e number as is the arc AC. In the second c a s e , the
sy lla b le “ ar” stands for “ a r e a ” . In the c a se of the hyperbola,
how ever, th ere is no such sim p le relation ship betw een the area
of the se c to r OAC and the arc AC as in the ca se of a c ir c le .
F inally, form ulas (29) do not as yet provide us with new
m eans for calcu latin g the corresponding in teg ra ls. For exam ple,
the fir st of them sim p ly a s se r ts that the in tegral is the function
in v er se to sin /, but we have decided to define sin t its e lf as the
function in v er se to the integral!
How ever, for the second of the in teg ra ls (2), th ere is an
analytic rep resentation that does not depend on our d efinitions.
S p ecifica lly , it is given by the follow ing form ula, the validity
of which can be shown by differentiating:

< - / 7 = a T = i.(> + y r + ? ) . wo)


• V 1+

This enables us to obtain a form ula for the in v e r se function


y = sinh t without needing to r e so r t to the in tegral (38) ev ery
tim e we need to speak of the hyperbolic sin e. S p ecifica lly , it
follow s from (30) that

V i + y2+ y = e1.

so that

1
Vl + / - y - e
VT+f + y
GENERALIZED S IN E S 17

If we subtract the second of th ese equations from the f ir s t, we


obtain

y = sinh / = (31)

Thus, it would have been p o ssib le at the very beginning to


define the hyperbolic sin e from form ula (31) as half the d iffer­
en ce of two- exponential functions.
Then, form ula (10) for cosh t y ield s

Since + e~')/2 > 0 and cosh t is a lso p o sitiv e (we r e c a ll that


x = cosh t is the a b s c is s a of a point on the right branch of the
hyperbola (Fig. 4)), we have

cosh t = e -~r>g— (32)

T his form ula d efin es the hyperbolic co sin e as the average of


the exponential functions e‘ and e '. In Chapter 5, we sh a ll s e e
that the circu la r functions can a lso be ex p ressed in term s of
exponential functions though only with im aginary argum ents of
the exponentials.
9. Let us turn to the lem n isca te sin e s i /.w h e re the variable
/ is treated as the arc length of the lem n isca te. Let us u se polar
coordinates for the in tegral ex p ressio n for t. In polar coordin­
a te s, the differential of the arc length is given (see F ig. 11) by:

In the p resen t c a s e , equation (17) yield s

ip = 2* arcsin ( r 2).

From (29) (with y = r7), we obtain

dq _ r
dr y I — r4
18 THE REMARKABLE SINE FUNCTIONS

Consequently,

d t=
1—r‘

s o that

(33)

As r in c r e a se s from —1 to - f !, the in tegral (33) in c r e a se s from


— o)/2 to o)/2 , where o>/2 has the value

(34)

C onversely, as t in c r e a se s from — w/2 to « /2 , the variable r


in c r e a se s from —1 to 1. Thus, we can define the function
y = sl/ in the in terval (— w /2 , a>/2) as the function in v er se to the
in tegral (33). As b efore, we sh a ll define the function cl t by the
form ula

FIG. 11.

10 . Let us com pare the definitions of the th ree sin e s given


in this chapter: (1) v = sin/ (the circu la r sin e ), the function
in v er se to the function
GENERALIZED S IN E S 19

and defined in the in terval (—n / 2 , n /2), w here

1
ji __ I* dr\

0
v \ —v1 ’

(2) y = sinh t (the hyperbolic sin e), the function in v er se to

y
f rfTl

and defined on the en tire rea l axis; (3) y = s i t (the lem n isca te
sin e) the function in v er se to

y
f rfri
J v

and defined in the in terval (—u / 2 , w /2), where

I
Ci) /' rfll
"2 s

O bviously, a ll th ese c a s e s w ill be included if we su cceed in


studying the function y = s{t) in v er se to an integral of the form

t= f , dy= , (35)
J V\ + mr? + my*

w here m and n are any rea l num bers. Let us a g ree to ca ll such
a function a “ g en eralized s in e .”
For m — — 1 and n — 6, we have the c ircu la r sin e; for m - - 1
and n = 0, we have the hyperbolic sin e; for m = 0 and n - — 1, we
have the lem n isca te sin e.
If m = — ( l- f - 62) and rt — k2, w here 0 < & < 1, the polynom ial
1-|- mvi2- f «r|4 takes the form (1 — if) (1 — kh\l). In this c a s e , the
function s(t) is .th e Jacobian ellip tic function with modulus k. It
is called the sine am plitude and is donated by sn (b k) or, m ore
b rie fly , sn t.
The function sn t is encountered, for exam p le, in the prob­
lem of the pendulum. This problem c o n sists in study of the
20 THE R E M A R K A B L E S I N E F U N C T I O N S

o sc illa tio n s, in a v ertica l plane, of a heavy sp here of m a ss m


at the end of a fine thread of length I (see F ig . 12). Suppose
that this thread with the attached sphere is held at an angle
0O from its equilibrium position and that the sphere is then
re le a se d with zero in itia l velo city .

When the strin g is p a ssin g through the p osition that m akes


an angle 0 with the v ertica l, the v elo city of the sp h ere is
v — i /id/ lit and its k inetic energy is

m v2 _____ m l2 / d % ' 2

— "ITVST/ '

From the theorem dn kin etic e n erg ies, this value m ust be equal
in the p resen t c a se to the work done by gravity (we n eglect air
resista n ce):

m gB 0B = m gl (cos 0 — cos 0O).

Thus,

ml2 j d&\2 . n a.
— = m gl (cos 0 — cos 0O).

so that

dt = — — • —r dQ =
V 2g V cos 9 — cos 90

(The minus sign is due to the fact that the angle 0 fir s t d e c r e a se s
as t in c r e a se s). Integrating, we obtain
GENERALIZED SINES 21

o ft,
dQ f J_ f ^9
t= —
Y cos 9 — cos 0o 2e B
■’ Y COS 0 — COS 9(,

_ J_ ______ M_______
2
| / sin-’ — — sin2 —

Let us m ake the change of variable

o %
2 'b.
sin — sin

Obviously,, ri has a sim p le geom etric interpretation, nam ely,

AC
n — A„C '

We obtain

If we denote the constant

by (rep resen tin g the instant the pendulum p a s se s through


its equilibrium p osition ), we can rew rite the la st form ula in the
form

'hi
' V
» 7
g JJ
« Y
.. /
(1 — ')*) (I “ sin ifj

from which we get

‘hi
o
f ..............
1. (I If ) ( I — sin- 11- j
= V I i! n
22 THE R E M A R K A B L E S I N E F U N C T I O N S

and, consequently,

(i
m i --

n =■
' Sl' 1 f si"

Thus, we have ex p ressed the quantity q ch a ra cterizin g the d is ­


p lacem ent of the pendulum at the instant / from its equilibrium
p osition in term s of the Jacobian ellip tic function with modulus
k — sin (r.,,/2).
11. It is natural to require that the polynom ial

P 1111 --- 1 >nr\2 -)- m\

in form ula (35) not be a perfect square. O therw ise, VP(Tp w ill
be of the form and the integral (35) w ill be ex p ressed
eith er in term s of the arc tangent (if q > 0) or in term s of the
logarithm of a linear fractional function (if q < 0). In each of
th ese c a s e s , exam ination of the in v erse function would yield
nothing new. T h erefo re, in what fo llow s, we sh a ll assu m e that

rri- —- An 4=0. (36)

We a lso need to d istin gu ish betw een the follow ing two c a se s:
I. The polynom ial POP has no real z e r o s. This m eans that
eith er the seco n d -d e g ree trin om ial 1 -j~ mz ~f-nz'1 has no rea l
zer o s (that is , m2 -An •. u) or its zero s are negative rea l num­
bers (that is , m5 - 4n > 0, n > 0. m > 0). In this c a s e . Pop > 0 for
arb itrary (real) ip T h erefore, the integral (35) is defined and
rep re se n ts a str ic tly in crea sin g function on the entire rea l
a x is. We denote the ex p ressio n

by A. For /; ----- 0, this in tegral d iv erg es and -4 - oo. For n 4-- 0, it


con verges and 4 is a finite p o sitiv e num ber. Since the, in tegral
(35) in c r e a se s continuously from — 4 to 4 - 4 as y in c r e a se s
continuously from — jo to 4 -co , the in v erse function y —- syt) is
defined in this c a se in the in terval (— A, — .4) and it in c r e a se s
continuously in that in terval from - to -j- o. The hyperbolic
sin e (for which .4 —oo) provides a very sim p le illu stra tio n of
this c a s e .
GENERALIZED SINES 23

II. The polynom ial P(r\) has real zero s (either two or four
such). This m eans that m2 — An > 0 and that eith er n < 0 or
m < 0 < /i. Let a denote the sm a lle st p o sitiv e zero. Then, — a
denotes the negative zero of sm a lle st absolute value. T h erefore,
the polynom ial P(rp has no zero s in the in te r v a l)— a, a). Hence,
P(r\) does not change sig n in that interval; it rem ains p ositive
sin ce I > 0. [We note that P(r\) b ecom es negative when
il (in creasing) p a sse s through the value a.] The in tegral (35) is
a function that is defined and str ic tly in crea sin g in the in terval
(— a, a). We denote the ex p ressio n
(1

by a (which is > 0 ) . Since the integral (35) in c r e a se s continuously


from —a to a as y in c r e a se s continuously from —a to a, the
in v er se function y — s(t) is defined in this c a se in the in terval
(— a, a) and in c r e a se s continuously in that in terval from —a to
a. The circu la r sin e (for which u 1 an d a — n/2) are sim p le
illu stra tio n s of this c a se .
In the ca se of the Jacobian function, in which

1 -(- m f = (1 — if) (1 — ft'V) (0 < k < 1),


we have

This la st quantity is called the com plete ellip ti c integral of


the f i r s t kind in L e g e n d r e ’s n o rm a l f o r m . It is denoted by K(k)
or, m ore b riefly, by K. Special tables have been com piled for
it. We note that th ere is a sim p le ex p ressio n for w in term s of
K ( l / y r 2):

If we now set 1 — if = y2, we obtain

CD dy
2 VI
- y d fi - 4 y2j
24 the: r e m a r k a b l e sine; f u n c t io n s

Thus, if we define the functions sin t, sinh t, s i t, and sn (t , k)


as the functions in v er se to the in tegrals referred to above, only
the function sinh t is defined on the en tire rea l a x is. The func­
tions sin t, s i t, and sn i7, k) are defined only in the finite
in tervals

n n u>
■K

r esp ectiv ely .


In the follow ing chapter, we sh a ll show how we can adhere to
our b a sic plan of exposition in t e r m s of the i n v e r s e s o f integrals
and s t ill extend th ese definitions not only to arbitrary rea l
in tervals but to the field of com plex values of t. How ever, when
we do th is, the product form ulas and the p eriodicity of the
functions sin t. and s i t, of which we spoke in Chapter 1, w ill not
be im m ed iately obvious. N on eth eless, an addition th eorem , which
we sh a ll obtain in Chapter 4, w ill enable us to esta b lish e a s ily
not only th ese but many other facts that go far beyond what we
might d isco v er if we confined o u rselv es to a g eo m etric d efin i­
tion of th ese sin e s.
CHAPTER 3

I n te g r a tio n in th e C o m p lex P l a n e

12. As we have stated , it is p o ssib le to study com p letely the


p rop erties of the various sin es and their s im ila r itie s and
d ifferen ces if we treat them as functions of a com plex variable
t : - o -/t, where « and r are real numbers and i is the im aginary
unit. Of co u rse, when we do this we cannot then interpret t as
area or arc length, nor can we interpret the corresponding
value of the sin e in question as the ordinate of a point or the
length of a chord. At the b asis of the definition is the integral

the upper lim it of which is som e com plex number . a - iv,


w here u and v are real num bers.
Let us a ssu m e, just as in Chapter 2, that m and n are real
num bers such that m: — An —0. The in tegral (37) is a particular
c a se of an integral of the form

(38)

where the integrand is a rational function of r and the square


root of a polynom ial of no higher than fourth d eg ree. Such an
in tegral is called , in gen eral, an elliptic integral (provided the
integrand does not reduce to a rational function of z and the

25
26 THE R E M A R K A B L E S I N E F U N C T I O N S

square root of a polynom ial of second or low er d egree). In


p articu lar, for the in tegral (37) to be e llip tic , it is n e c e ssa r y
that nil—4n^=0 and that n=k0. The name “ e llip tic ” is due to the
fact that the length of an arc of an e llip s e is ex p ressed in term s
of an in tegral of the form (38). In p articu lar, if P /V + y2/b- = 1
(where (b±a) is the equation of a particular e llip s e , the length
of the arc CA (see F ig. 13) is rep resen ted by the in tegral

cm „ / y i + ( £ ) • * * = ./' / 1+ y c - -
X X

a
_ ! f a * - ( a ' - h ) x ______ (/ r
a J y w — - y r r 7j '

13. Let us pause to s e e how we should in terp ret the integral


(37) in the c a se in which w is a com plex number and what prop­
e r tie s this in tegral p o s s e s s e s . H ere, we sh all make certain
a sse r tio n s without proof. P roofs can be found in m ost textbooks
on the theory of functions of a com plex variab le.
Let us se t up a C artesian coordinate sy ste m in the plane.
We sh all regard i.v, y\ as the im age of the com plex number
z —- x i iy. In what fo llow s, we sh all apply the term “ point”
both to a point in the com plex plane and to the com plex number
corresponding to it.
Let us connect the coordinate origin z = 0 with any point
w — u^- iv by draw ing a continuous curve L of finite length from
the origin to the point w (se e Fig. 14). (The fact that the curve
sta rts at the coordinate origin is of no sig n ifica n ce for the
d erivation s that we are going to m ake.) Let us partition /. into
a rcs by choosing points zQ U, z v z 7........ z n~\' z n = iW numbered
in the d irection of m otion along L from 0 to a>, and let us
N T E G R A T I O N IN THE C O M P L E X P L A N E 27

form the corresponding approxim ating sum for the function


1 / V T + mz'2 ■+- nz*\

V (39)
u (Zk — **-!)•
* -1 + mzk + nzu

Since )/" 1 + mz2 -f- nz'1 is a double-valued function, let us agree


that, for z —- z 0~ 0, we take the p o sitiv e square root of 1. Sup­
p o se a lso that, if the value of the square root has been chosen
at any point z k__„ that value at the point z k is chosen that w ill be
the c lo se r to the value a t z 4_,. Then, for an arbitrary partition
of the curve £, the sum (39) w ill have a definite value. Let us
now take over fin er partitions of L so that the length of the
lon gest arc approaches zero. It can be shown that the sum (39)
w ill approach a d efin ite lim it. It is called the in tegral of
l / \ r \ + mz-~+liP along L and is denoted by

f \rr___
■'
dZ_____ =- ■
\ + mz2 + m z4
(40)

If the point w lie s on the rea l x -a x is (that is , if w---u and


u — 0), if the polynom ial I -f-mz-^-nz1 does not vanish anywhere
on the in terval [0, «], and if L coin cid es with this in terval, then
all the values of z, that is , z0 — 0, z t, , z n, a re rea l num bers,
the values of y \ mz2 -|~ nzA are real and p o sitiv e, and we have
sim p ly the usual definite integral

0■! | 1 L mx- -)-■ /;.v4


28 THE REMARKABLE S IN E F U N C T IO N S

In analogy with the gen eral c a s e , it is custom ary to w rite the


in tegral (40) in the form

(41)

without indicating the curve L over which the integration is


taken (that is , for the points of which the approxim ating sum s
(39) are form ed). H ow ever, such notation ca u ses som e in ­
d efin iten ess. S p ecifica lly , for two d istin ct curves Lx and L2 that
unite the points 0 and w, the correspon d in g in teg ra ls (40) may
have d istin ct valu es. H ow ever, in the study of the theory of
functions of a com plex v a riab le, it is shown that this can happen
only when the curves Ly and L2 en c ir c le one or m ore zero s of the
polynom ial 1 4- mz2-)- nz* (at which the integrand b ecom es infinite)
one or m ore tim e s. [See Fig. 15, where p, and p2 rep resen t two
of the four zones of that polynom ial and the curves Z., and L2
together e n c ir c le both th ese zero s tw ice.] H ere, the d ifference
betw een the two values of the in tegral depends on just how many
zero s are en circled and how many tim es they are en circle d .
T h erefo re, the in tegral (41) has in finitely many d istin ct values
for every value of w; that is , it rep resen ts a m u ltip le-valu ed
function F(w), any one of the values of which we sh a ll denote by
the letter t (where t is a com plex number):

(42)

FIG. 15.

Let us suppose that w0 does not coincid e with any zero of the
polynom ial 1-\- mw1 rvw*t that is ,
I N T E G R A T I O N IN THE C O M P L E X P L A N E 29

1 -)- riw* 0.

Then, we can draw a c ir c le K around w0 that w ill a lso e n circle


no zero of that polynom ial (see F ig. 16). Let us ch oose any
curve L0 connecting z 0 with wlt. For points w in sid e the c ir c le
K, let us take cu rves form ed by the union of L, and the line
segm en t drawn from w0. to w. Under th ese supplem entary condi­
tio n s, the in tegral (42) d efines a sin g le-v a lu ed function of -w
in sid e the c ir c le , which coincid es with som e sin g le-v a lu ed
branch of the function, F (to) inside the c ir c le K. Its value at the
point w is equal to the in tegral along L, plus the in tegral along
the segm en t to0to. To sw itch to any other branch, we need only
rep lace the curve L0 with another curve L0 to which another
value of the in tegral at the point w0 corresp on d s. It is ea sy to
s e e that the corresponding branches d iffer by a constant term
in sid e K. One can show that each of th ese branches p o s s e s s e s
a d erivative at an arbitrary point w in sid e the c ir c le K; s p e ­
c ific a lly .

di ______ I______
= lira (43)
ctw ( til W\ —w 1^1 mw1 f- nw{

In other w ords, the theorem that the derivative with resp ect
to the upper lim it of integration of an in teg ra l is equal to the
value of the integrand at the corresponding point rem ains valid
for the com plex integral (42).
F u rth erm ore, one can show that none of the values of our
in tegral that are assu m ed at any point w l in the plane can
coincide with any of the values of the in tegral assu m ed at
another point w2 (where •w2=ji=wI). In other w ords, to a given
30 THE R E M A R K A B L E S I N E F U N C T I O N S

value t of the in tegral th ere correspon d s only one point w as


the upper lim it of integration.*
All this ju stifie s our regarding w in form u la (42) as a s in g le ­
valued function of the com plex variable t. That is , we may
r e g a r d the u p p e r li m it of integration as a function o f the value
o f the integral. Let us denote this function by

w s (t). (44)

As b efore, we sh a ll ca ll it a gen eralized sin e.


It follow s from (43) that

dw =~ y 1-j-- mw~ -|- nw‘1dt. (45)

With regard to the m u ltip le-valu ed n ess of the in tegral (42),


which resu lts from the fact that to a sin g le value of w there
correspon d s in finitely many d istin ct values of t, this property
leads to the follow ing a sse rtio n for the in v er se function (44).
A sin g le value of w is obtained for in finitely many d istin ct
values of t. In connection with th is, we sh a ll se e in Chapter 6
that the function s(t) is a p eriodic function of t.
14. Let us com pare the values of the in tegral (42) c o r r e ­
sponding to two points stand — w sy m m etric about the coordinate
origin . Let us connect th ese two points with the origin by curves
L and L' that are sy m m etric about the origin (see Fig. 17), and
let us integrate over th ese cu rves from the origin to the points
•m and — w , re sp e c tiv e ly . If we partition the curve L into arcs
by m eans of the points z()---0, z t, z.2....... z n ------w and evaluate the
corresponding approxim ating sum
n
(46)

and if we partition the curve /.' into sy m m etric a rcs with the
aid of the sym m etric points - r,,-•=-n. z v - v......... zn^ — w
and again evaluate the approxim ating sum ,
n
V _ (4 6 0
1
k- ) i* 1--- —/?

*A detailed proof of thi.s fact can be found in E . Goursnt, A C o u r s e i n M a t h e m a t i c a l


, vol. II, New York, Dover, 1059. However, this proof assumes familiarity
A n a ly sis
with the general theory of analytic functions and the fundamentals of the theory of el­
liptic functions.
I N T E G R A T I O N IN THE C O M P L E X P L A N E 31

the values of the second sum w ill differ in sign from the p reced ­
ing one. But the fir st of these has the lim it

f ^ = £ L =
■■ V \ + mz- + nzA

which is one of the values of

■' } 1 m z-' -j- n z *

and the second has the lim it

./ I 1 -L nz- 4~ nz'
L'

which is one of the values of


•{if
f - J'
o■' V 1 -f- m z- -L n z '
r

It is clear from what has been said that th ese values d iffer from
each other only in sign . In other w ords, if

li ’I 1 f1 mz- ■ n z 1

then

• 1' i f- mz j- n z A

(We are still assu m ing that the integration is over two cu rves
sy m m etric with resp ect to each other about the point v 0.)
From this it follow s that, if the value of the upper lim it -u>
corresponds to the value of the in tegral /, then the upper lim it
:■:> correspon d s to the value — i. In other w ords, if

i' {!) Ul
32 THE R E M A R K A B L E S I N E F U N C T I O N S

then
s (— t) — — w.

that is ,
s (— t) ■— — s (t). (47)

This m eans that the function s(i), rep resen tin g the in v er se of
the in tegral (42), is an odd function.

15. In the p recedin g sectio n , we com pared the values of the


in tegral along the cu rves L and L', w here I! was obtained from
L by a rotation through the angle jt about the coordinate origin.
Now, let us com pare the values of the in tegrals along the curves
L and L", where L" is obtained from L by rotating it through
an angle n/2 about the coordinate origin (see Fig. 18). We note
that this rotation correspon d s to m ultiplication of each point of
the curve L (that is , the com plex number rep resen ted by each
point) by the number i.

FIG. 18.
I N T E G R A T I O N IN THE C O M P L E X P L A N E 33

O bviously, if the curve L connects the origin with the point


w, then L" connects the origin with the point lw.
Let us partition the curve L into a rcs sep arated at the
points z0_ = 0, z v z t......... z n-i' zn = w • For the function
l / V l + mz2+ Tz~\ the corresponding approxim ating sum is

2 ■(«* z k-0- (48)


V* i 4* mz\ 4*;

The points obtained from the a b o v e-listed points by m eans


of a rotation through n/2 about the coordinate origin lie on
the curve L". T h ese are the points iz0 = 0, l z v iz2........
lz n = lw. The corresponding approxim ating sum for the sam e
function \ / \ r\ + m z 2 + mz* is of the form
n
(49)

If the partition of the curve^ L is varied in such a way that the


length of the lon gest a rc z k^izk approaches 0, the fir st approx­
im ating sum w ill approach the lim it

/’ d.2r

J y \ T mz2T

which rep resen ts one of the values of the m u ltip le-valu ed


function

J
f ____ dz
V~ 1 4 - m z 2 nzA
o

C om parison of (49) and (48) shows that (49) can be regarded


as the approxim ating sum constructed for the sam e curve but
for a d ifferent function, nam ely, the function
i

V1—mz\ -f nzl
T h erefore, the lim it (49) is equal to the integral
34 THE R E M A R K A B L E S I N E F U N C T I O N S

which rep resen ts one of the values of the m u ltip le-valu ed func­
tion

dz
I
f Y 1—mz2 ■-)- mz*

On the other hand, this sam e lim it m ust be equal to the


in tegral
I*_____ dz_____
■; Y 1 + m z2 -j- n z *

which is one of the values of

r dz
■' Y
0 r 1 1 + mz2 + mz*

sin c e the approxim ating sum (49) was constructed for the curve
L" and the function

______ l______
Y 1+ rnz‘

From what has been said , it follow s that for every value of the
in tegral
iw
f dz- - - -
J Y 1 + mz2+ nz*

there correspon d s the equal value of the in tegral

dz
1
/ Y 1—mz2+ nz*

In other w ords, if

dz
= t and f — - dZ = r.
/ V1 mz2-f- mz< ^ Y 1— nz2-|- nz*

then t — it* under the hypothesis that the integration in the fir st
in tegral is over the curve L" obtained from the curve L, over
INTEGRATION IN THE C O M P L E X P L A N E 35

which the second in tegral is taken, by rotating L through an


angle n /2 about the coordinate origin.
Let us denote s*(t) the function in v er se to the integral

«
f ______ dz______
J y
o
1 — m z 2-f- n z
'
*
O bviously, s"(f) is a lso a generalized sin e. It follow s from the
equation

dz
/ V 1 — mz2 + nz*

that w* — s*(t"). A lso , /w = s(/)and t — l f . T h erefore,

a (0 = s (It") = tw = ts* (/*).

F in ally, if we drop the a ste r isk in the notation /*, we have

a (//) = /s’ (f). (50)

This relation ship connects the values of two gen eralized


s in e s , the fir st rep resen tin g the in v erse of the integral

W
f
•'
dz-~
y i + mz2 t nz4

and the second rep resen tin g the integral

to
f dz-~. -
o V \ —mz2-F' nzA
./

16. U sing th ese fa cts, in what follow s we shall con sid er


each of the sp e c ia l c a s e s of the g en eralized sin e as a function
of a com plex variab le. S p ecifica lly , the circu la r sin e w — sin t
is the function in v erse to the integral

W
36 THE R E M AR KA BL E S I N E F U N C T I O N S

the hyperbolic sin e w = sinh t is the function in v er se to the


in tegral
W

the lem n isca te sin e w = s\t is the function in v erse to the in te­
gral
W

the Jacobian ellip tic function (the sin e am plitude) w = s n ( t , k) is


the function in v er se to the integral

If t a ssu m es rea l v a lu es, we can a s s e r t that the functions


sin t, sinh t, and s i t defined here coincide with the functions of
Chapter 1 only in th ose in tervals in which th ese la st functions
are in v er ses of the corresponding in teg ra ls. For sin t, this
m eans that —n/2 << t < ji/2 and, for s i t, that — w/2 < t < u /2 .
Only for sinh t are we without any such restrictio n ; in this
c a s e , the in tegral has an in v erse for the en tire rea l axis (cf.
Chapter 2).
From sectio n 14, it follow s that each of th ese functions is
odd:

sin(— t) — — sin t, sinh (—1) = —sinh/,


(51)
si (— t) = — si t, sn (— t) — — sn t.

To use the d erivation s of sectio n 15, we need to con sid er the


function s(t) togeth er with the function s ' (t). The change'from
s(t) to s*(t) corresponds to change of sign of the co efficien t of
z- in the polynom ial 1 4- mzl -j~ in the integrand. To s e e this
c le a r ly , we p resen t the follow ing table of type s(t) and s*(t)
functions correspon d in g to the four particular c a s e s that we are
con sid erin g.
N T E G R A T I ON IN T H E C O M P L E X P L A N E 37

Integral Inverse Integral Inverse

W w

J
f dz
Y 1 — rnz■-j- n z 1
s(t)
J
f dz
V 1 — rnz1 -j- n z *
s*(t)
0 0
w w

J
f dz
Y 1 — -r2
sin / f
J
dz
sinht
0 Y i+ ^
0
w w

J
f dz
sh t
J
f dz
sin t
0 Vl+ ^ 0 V\ — z2
w w

J
f dz
\ r \ — z*
sl { f
J
dz
si t

i
0 0
w w

J
f dz
sn (1. k)
J
f dz
+ z 1) ( i -j- k2z 2)
sn* (t. k)
0 V (1 — Z-) (1 — Vz' 1) 0

If we apply to each of th ese particular c a s e s the gen eral form ula


(50), we find

sin(it) = i sinh I, (52)

sinh (it) = i sin t. (53)

si [it) = i si t. (54)

sn (it, k) = / sn* (t, k). (55)

It is ea sy to s e e that form ulas (51) follow from th ese la st


form u las. For exam p le, from form ulas (52) and (53) applied
s u c c e s s iv e ly , we obtain

sin (—t) = sin[i(/t)J = isinh(d) = i ■(sinf = —sin*.

17. The r e su lts of sectio n s 14 and 15 can be obtained in a


very sim p le manner if we a ssu m e that the fa m ilia r rule of
change of variable under the in tegral sign is a lso applicable to
in tegrals of functions of a com plex variab le. In co u rses on the
theory of functions of a com plex v a riab le, it is shown that this
rule is indeed applicable under quite broad con d ition s—conditions
that are s a tisfie d by the exam ples we have been con sid erin g.
38 THE R E M A R K A B L E S I N E F U N C T I O N S

In the in tegral

0■I y l + m z2 +' nz'

let us make the change z = — £. We obtain

■/ V l +- mz* + nz* ' ■> V \ + ml-- f- ’-V '

Turning to the in v erse functions, we conclude that

w — s(t), —w — s ( — /), i .e ., s s (/).

If we make a different substitution, nam ely, z — i%, that is ,


£ = iz, we obtain

d. ____ [_d_£____
J VT+m: nz' J V [ —m ’Q1 -\- nl}

When we turn to the in v er se functions, we conclude that

nu— s{t), — i.e ., s(it) — is(t).

Let us use once m ore the rule for change of variab le, this tim e
to ex p ress the function sn‘ (/, k) rep resen tin g the in v er se of the
integral

dz
/ (0 < k < 1),
V(\ + ^ )('l,+ ™4b

in term s of the Jacobian function sn t with appropriate m odulus.


To do th is, we make the follow ing change of variable z = l / V l — .
A fter som e sim p lifica tio n s, we obtain

W I 11-
f ______ i f - ______ _________ dl_________
■! Y ( \ + z *)(\+k*z*) / V ( T - '
I N T E G R A T I O N IN THE C O M P L E X P L A N E 39

Let us s e t k' ~ Y l — k2 > 0 (where k' is called the modulus


com plem entary to k). Then, the equation

f _______ d£______
t=

which in d icates that -121= sn*(/, k), can be rew ritten

V 1 f W1

from which we get

sn (t, k')

or

sn2 (t, k')


w sn* (t. k) =
f \ — sn2 (t.JP)'

T h erefore, equation (55) can be replaced with the equation

,
= I ~ r ={!, rk )
. . s n
sn lit, /!) (55')
y 1 — sn - (I, k')

w here kr — ^~T~~k2.
CHAPTER 4

E u le r ’s M ethod o f D e r iv in g th e
A d d itio n T h e o r e m s

18. At the b a sis of a ll trigonom etry lie the two addition


th eorem s
sin(a-)-fl) = sin a cos fl d-cos a sin 0. )
(56)
cos (a-(-P) = cos a cos (1— sin a sin p. J

With the aid of the relation ship betw een the sin e and c o sin e,
we can rew rite th ese addition th eorem s in such a way that the
value of sin(ad-p) is ex p ressed only in term s of sin a and sinp
and the value of cos(a-)-p) is ex p ressed only in term s of cos a and
cos p:

sin (a -f- p) = sin a — sin2p -f- |/T — sin2a sin p.


(56')
cos (a -j- p) = cos a cos p — Y * — cos2 u V 1 — cos2P-

We do not w rite the ± sig n in front of the rad ical, but it should
be understood that the sign is taken according to the value of the
circu la r function ex p ressed by that root. For exam ple,

• A, . n II a Y3
V 1 -s in 2 6 = cos 6 = — .
, r . 7 T 3 R 3jx V 2
1/ 1 — s u r -~= cos — — — — - etc.

Each of form ulas (56') e x p r e s s e s a lg eb ra ica lly the value of the


function of the sum of two num bers in term s in the values of

41
42 THE R E M A R K A B L E S I N E F U N C T I O N S

that sam e function of the two individual num bers. Form ulas of
this type are called alg e b ra ic addition th eo r em s and we say that
a function for which such a theorem is applicable p o s s e s s e s or
obeys an algeb raic addition th eorem . Thus, the sin e and cosin e
obey algeb raic addition th eo rem s. Other exam p les are the linear
function y -----at and the exponential function y e1. T h ese are
exam p les sin ce

a (a P) — a c t - f - a p , e a+fl = ea ■c®.

19. Our next problem w ill be to d erive an addition theorem


for the function w = s ( t ) in v er se to the in tegral

t= f ■■
•J Y 1+ mz2+ nz4

In p articu lar, we sh a ll obtain addition th eorem s for sin t,


sinh / , s i t, and sn(f, k ) . In this sectio n , we sh all ob serve the
gen eral behavior of the solution of this problem .
Suppose that a and p are arbitrary com plex num bers and let
us define y==a-t-p. We set

u = s ( a), v = s(P), w = s ( a - j - p).

This m eans that

dy
a --
Y 1 + my' + ny* ■J V \ -t- m y " + ny*
o
(57)
__ / ' _____ dy______
■J T T + m y5 T ny'

If a and p vary in such a way that th eir sum y rem ains constant:

a -j- p = y = const, (58)


then
d(a + P) = 0,
that is ,

d
______ dy______ _____ dy_____ = 0.
V \ + my- + ny>
o Y 1 + my- + n y 1
E U L E R ' S METHOD OF DER IVI NG THE A D DI T IO N T H E O R E M S 43

or

(59)

T his is a d ifferen tial equation connecting the valu es u ---s(a)and


n=:s(p) subject to the condition (58). We sh a ll show below that
equation (59) has an algebraic solution and we sh a ll find it. M ore
p r e c ise ly , we sh a ll find an algeb raic function F(u, v) that is
equal to u at v = 0 (and equal to v at u — 0), the d ifferen tial of
which van ish es by virtu e of condition (59) (or (58)). In other
w ords, we sh a ll show that (58) im p lies that

dF(u, v) = 0 and F («, v) = C = const. ( 60 )

To find the constant C, we s e t p = 0. Then v = s (P) van ish es and


F(u, v ) coin cid es with u = s(a). But, if p = 0, it follow s on the
b a sis of (58) th a ta = y and hence u = s(a) = s(y), so that

C = F(u, «)|p=0 = «la=v = s(Y)-

Thus,

F(a, v ) = F[s (a), s (P)] = s (y) = s (a -f- P).

T his relation is independent of the value of the sum a-f-P = Y-


C onsequently, we sh a ll obtain an algeb raic addition th eorem for
the function s{t):

s(a-|-P) = n s (a ), s(P)l-

Thus, the problem of d eriv in g an algeb raic addition th eorem for


the function s(t) red u ces to finding an in tegral of equation (59)
p o s se s s in g the p ro p erties listed .
2 0 . To find an a lgeb raic solution of equation (59), we sh a ll
u se a method proposed by Euler in his Institutiones C alculi
In teg ra lis, Vol. I (1768).* We sh a ll begin with the follow ing
fou rth -d egree a lg eb ra ic equation relatin g the v a ria b les u and v
and containing three arb itrary param eters A, B, and C:

u2-|~ v2 Au2v2-f- 2Buv — C2= 0. ( 61 )

*Here, we shall use a simplified form of Euler’s method that will be sufficient
for our purposes.
44 THE R E M A R K A B L E S I N E F U N C T I O N S

If we differentiate this equation, we obtain

(u 4 Bv -f- Auv2) du (u 4 Bu -f- Au2v) dv — 0. (62)

But, if we rew rite (61) in the form

(Av24 1) u1-)- 2Bvu -|- (v2 — C2) — 0,

m ultiply both sid e s by Av2-\-\, and com plete the sq uare, we


obtain

[(/tv2 4 \) u-\-B vf — [(C1 — v2)(Av2-\- 1 ) + S V | = 0,

from which we get

u-\- Bv Auv2= } / C1-\- (B2+ AC2— 1) v2 ~ Av*. (61')

Again, we do not w rite the ± sign but we understand which­


ev er of the two values of the square root coin cid es with the
ex p ressio n on the left. (An analogous rem ark w ill apply a lso to
the functions that we sh a ll con sid er below when we encounter
square roots.)
Equation (61) is sy m m etric in u and v. T h erefore, if we
r e v e r s e the ro les of u and v, we obtain from (61')

v H- Bu -f Au2v = \ rC2+ (B24- AC2 — 1) u2 — A.A. (61")

If we su bstitute (61;) and (61") into (62), we obtain

Y'C2+ (B2-j- AC2— i j v2^ ' d u


4- Y C- -f i ff + AC'2 — \)u2 — Au* dv = 0

or

du
V"C2 (62+ AC- — 1) u2 Au*
dv (59')
0.
Y C ‘ 4 (B2 4 AC1 — \)v< — Av *

Since the d ifferen tial equation (59') is s a tisfie d for a ll values of


u and v that a re related by the algebraic relation (61), it follow s
that (61) is an algeb raic solution of (59').
E U L E R ' S METHOD OF DERIVING THE ADDI TI ON T H E O R E M S 45

Let us now choose A, B, and C so that equation (597) w ill


coincid e with equation (59) of the p receding sectio n . To do th is,
we only need to se t

B74 AC7 - 1 = mC7. A = — nC7.

Let us ex p ress A and B in term s of m, n, and C. Then,


equation (59') (after we m ultiply a ll term s by C) takes the form

(59)

and its solution (61) is

u74- v2 — nC7u7v2 -(- 2 V T + mC2 nC] uv — C2= U, (63)

w here C is an arb itrary constant.


Since we w ish to rep resen t (63) in the form (60) (that is , in
the form F(u, iO = C), let us so lv e equation (63) for C. We
obtain s u c c e s s iv e ly

[u2+ v7 — (nu2v2 1) C2]2 = 4(1 ~|- mC2-\- nC4) u2v2,


(1 — nu7v2)2C4 — 2 [(1 -)- nu7v2) (u2+ v7) -j-
4- 2mu7v7] C2 4- («2 — v7)2 = 0.
_ (1 + nu2v2) (u 2 -f- v2) + 2 mu2v2
(1 — nu2v2)2 '
V [(1 + nu2v2) (u 2 -(- v2) 2 mu2v2)2— ( u 2 — v 2) 2 (1 — nu2v2)2
(1 — nu2v2) 2
__ u2(1 + mv2-)- nvA ) 4 - v 2 (1 4 - mu2+ nu*) .
(1 — nu2v2)2 '
2uv V(l 4- mu24- (1 4“ 4~nv*)
+ (1 — nu2v2)2

s o that, fin ally,

u 1 4- mv 2 4 - nv4 4 - v V 14 “ m“ 2 4- nu< = c.
1 — nu2v2 (64)

The function

F(u, v) = u \ r 1 4 - m v2 4 - nv1' 4 ~ v V 14- mu2- f - nu*


1 — nu2v2

is algeb raic. At n = 0, it becom es u and, at « = 0, it b ecom es v.


We leave the read er to v erify that its differential is of the form
46 THE R E M A R K A B L E S I N E F U N C T I O N S

du dv
dF (u , v) = 0 ( « , w)
V '\ + mu2+~ru? y 1+ W 1+ nv*

w here Q(a, v) is so m e nonzero algebraic function. T h erefore,


the d ifferen tial equation (59) im p lies

dF( u, v) = 0, or F(u, v ) ~ const.


Thus, the ex isten ce of an algebraic solution of equation (59)
p o s s e s s in g the p rop erties indicated in sectio n 19 is esta b lish ed .
From th is, as was shown in sectio n 19, we get an algebraic
addition theorem for s(t) in the form

s(a + P) = F [s(a), s(P )l,

that is ,

: (a) Y 1-(-ms2(ft)-)-ns4(p) -fs (P)K 1-\-rns1(a)-fns4(a)


s(a -4P) = 1 — ns2(a) s1 (P)
s2 (a) — s2 (P)
(65)
5(0)11 4 ms2 (P) -t- ns1(P) - s ( P ) l l + ms2 (a) + ns* (a)

21. From the gen eral addition theorem (65), which is valid
for the function s(t), we obtain as sp e c ia l c a s e s the addition
th eorem s for sin t, sinh t, s i t, and sn (t , k). C orresponding to
the circu la r sin e are the values m ~ — landn = 0; corresponding
to the hyperbolic sin e are the values m = 1 and n ■= 0; c o r r e ­
sponding to the lem n isca te sin e are the values m — 0 and
n = — l; corresponding to the Jacobian function a re the values
m — ( I -f- k2) and n = k2 (for 0 < k < 1).
A ccordingly, from form ulas (65), we obtain the follow ing
addition theorem s:
For the circu la r sin e,

sin (tt ~f- p) — sin a | i — sin5 p -)- sin p \ 1 — sin2a: (65')

for the hyperbolic sin e ,

sinh (a ~U /3) —sinh a VI 4- sinh2 ft -A- sinh (3 VI 4~ sinh2 a: (65")

for the lem n isca te sin e,


s! (I V l si4 p + si P 1-4 — si* a
shu •
I -f- si- a si-’ p
si2 a — si2 p
(65")
si a | / l - s l < p - s i p V \ — si4 a ’
E U L E R ' S M E T H O D OF D E R I V I N G T H E A D D I T I O N T H E O R E M S 47

for the function sn(f, k),

sn (a + P) =
__ $n ay i + m sn2 p + n sn1 p -f- sn p 1 + m sn2 a n sn4 a __
1 — n sn2 a sn2 p
sn2 a — sn2 P
(65IV)
sn a 1 + m sn2 p -(- n sn4 p — sn p Y 1-1- m sn2 a -)- n sn4 a

For brevity, we w rite sn(a-f-p) instead of sn(a-f-p. k) and sn a


in stead of sn(a, k), etc.
It is im portant to note that a and p in th ese form ulas are
a r b i t r a r y c om p le x n u m b er s except that the a and p in form ulas
(65'") and (65lv) are such that the denom inators of the fractions
are nonzero.
CHAPTER 5

F u r t h e r Stu d y o f C o m p lex V a lu e s
o f th e A r g u m e n t

22 . For rea l values of /, we have the form ulas

c o s / ^ j A — sin2/, cosh / = V l -\- sinh2 /, (66)

w here corresponding to the value / = 0 is the value of the


rad ical equal to 1. Thus, form ulas (65 ) and (65 ) take the
form s

sin (a -f- p) = sin a eos p -(- sin p cos a, (67)


sinh (a -f- j8) = sinh a cosh /3 -(- sinh p cosh a. (68)

From th ese and from form ula (66), we d erive the addition
th eorem s for the circu la r and hyperbolic c o sin e s. S p ecifica lly ,

cos (a -f- ft) = \ f 1 — sin2(a -f- P) =


= |/(sin 2re-fcos2a) (sin'2p-j-cos2P) —(sin a cos p f- c o s a sin (b2 =
— |/ s i n 2 a sin2 p — 2 sin a sin p cos a cos p + cos2 a cos2 p =
= ± (sin a sin p — cos a cos p).

To choose the proper sign , we need in this ca se only to set


p = 0. We s e e that we need to take the minus sign on the
right:

cos (a -f- p) = cos a cos p — sin asin p. (69)

49
50 THE R E M A R K A B L E S I N E F U N C T I O N S

A nalogously, for cosh (a-bp ), we obtain, to begin with, the


form ula
cosh (n - f (3) = Vi -\~ sinh2 (a -|-/8) =
= V(cosh2 a —sinh2 a) (cosh2 /3 —sinh2 j8) (sinh a cosh/S ~V cosh a sinh/3)2=
= Vcosh2 a cosh2/3 -j- 2cosh a coshjS sinh a sinh 8 sinh2 a sinh2 /S =
= - (cosh a cosh /S -(- sinh a sinh /3).
For p = 0, we obtain cosh a on the left and ± cosh on the right.
T h erefore, on the right, we need to choose the plus sign:

cosh (o -(- /3) = cosh a cosh/S -)- sinh a sinh /3. (70)

Returning to form ulas (66), let us rep la ce t with it. We


obtain

cos (it) = vT —sin2(it) = V1 sinh2 1 = coshi, i


> (71)
cosh (it) V14- sinh2 (it) = Vl —sin21 —coshf. )
(In the right-hand m em b ers, we choose the plus sign s in c e , for
t = 0, the values of both circu la r and hyperbolic co sin es are 1.)
If follow s from form ula (66) that both co sin es are even
functions:

cos(— t) = Y 1—sin21 — cos t, cosh (—t) = Vl ~ j- sinh2 1 = cosh t. (72)

From the addition theorem (67), we a lso conclude that the


form ula cos / = s i n ( n / 2 —t) is valid for all com p lex values of
t. To s e e th is, let us set a = n/2 and p = — t in (67). We obtain

sin (-j —■tj = 1 • cos (— 0 + s|n (— t) • 0 = cos t. (73)

H ere, if we rep la ce t with n/2 — t, we obtain

cos (^-— ^j = sint. ' (74)

Thus, form ulas (5) of sectio n 1 are valid (for a ll valu es of t


both rea l and im aginary) if we define sin t as the function
in v er se to the in tegral
F U R T H E R S T U D Y OF C O M P L E X V A L U E S OF T H E A R G U M E N T 51

and define cos t as the square root of (1 — sin2/) (with the con­
vention that co s 0 = 1). Since, under such a definition sin t is an
odd function [ s e e form ulas (51)| and cos t is an even function
[see form ulas (72)], form ulas (4) of sectio n 1 a lso rem ain valid.
But, as was shown in sectio n 1, from form ulas (4) and (5), we
get all the “ reduction” form ulas of sectio n 1, in p articu lar, the
p eriod icity of the circu la r functions. T h erefo re, in what fo llo w s,
we sh a ll use the form ulas of sectio n 1 for arb itrary com plex
values of t:

cos (a — /) = — cos t, sin (a — /) = sin/;


cos (a -(-/) = — cos t, sln(a-|-/) = — sin<;
cos (2a -)-/) = cos t, sin (2a-f-1) — sin t.

To su m m a rize, we may say that the new d efinitions that we


have made (beginning in Chapter 2) for sin t and co s t have led,
for rea l values of t, to the functions that w ere defined on the
en tire rea l axis in sectio n 1 on the b a sis of geom etric con­
sid era tio n s. We note that, for the hyperbolic functions sinh t
and cosh t, we do not need to go through the analogous proof.
This is true becau se the d ifferen ce betw een the c a s e s of the
circu la r and hyperbolic functions c o n s is ts , with regard to the
p resen t m atter, in the fact that sin t, treated as a function of
the real variable in v er se to the integral

was defined d irectly only on the finite in terval —a /2 ■< t a/2


and we had no inform ation, to begin with, regardin g its behavior
outside that in terv a l. The hyperbolic sin e , on the other hand,
treated as the in v erse of the integral
52 THE R E M A R K A B L E S I N E F U N C T I O N S

was defined from the very beginning on the en tire rea l axis
(see Chapter 2).
If we s e t a = o and p = in form ulas (67)-(70), w here o and
t are rea l num bers, and u se (52), (53), and (71), we obtain

sin(cr-)- h) = sincr coshr -| -i sinh r • coscr,


cosier-4- ir) —cos cr cosh r —i sin a ■sinh r,
, (75)
sinh(<r-l- ti) = sinher cos i sin r • cosh a,
cosh (a -)- u) = cosh a cos r -(- i sinh a ■sin r.

T h ese form ulas ex p ress the circu la r and hyperbolic func­


tions for an arb itrary com plex value of the argum ent t = o + ix
in term s of th ese functions of the rea l variab les o and t .
23. In sectio n 8, we derived ex p ressio n s for cosh ^and sinh t
for r e a l t in term s of the exponential function:

e , - L e ' ( . e‘ - e ~ ‘
c o s h i= ----------- , sinh<=---------- .
2 2

From this it follow s that

e l = cosh (-)- sinh I . (76)

Since we defined cosh t and sinh t above the arb itrary com ­
plex t = a-\-iT., form ula (76) en ab les us to define the exponential
function for an arb itrary com plex value of t. U sing this de­
finition, we obtain, with the aid of form ulas (75),
ea+!T -=cosh(CT-l~*r)4~/,sh(<7-t-u) =
= cos r(coshcr-|- sinh a) -j- i sin r(cosha --(- sinh a).

Now c o s h + s in h a = eCTin accordance with form ula (76). T h erefore,

e‘ = e0+lx = (cos t I sin t ) . (77)

If we rep lace t with It' in (76), we obtain

e lt = cosh(ii")-(-sinh(fi ) = cosi i s int ' .

o r, om itting the p rim es,

e11 = cos t 1 sin /. (78)


F U R T H E R S T U D Y OF C O M P L E X V A L U E S O F T H E A R G U M E N T 53

This form ula is due to E uler. H ere, if we rep lace t with — t,


we obtain

e~lt = cos t — (sin t. (79)

F rom (78) and (79), we d erive two other form ulas of Euler:

el t -\-e 11
cos t — sin t = (80)
2

Let us verify that the addition theorem

ea+fl = ea . efl' (81)

w here a and p a re arbitrary com plex num bers, holds for the
exponential function e‘ defined by form ula (76). We have

ea+^ = cosh (a /9) 4- sinh(a -(- /3) =


= (cosh a • cosh (8 -|- sinha • sinh/3) -|- sinha • cosh/3 sinh/3 • cosh a) =
= (cosha -|- sinha) (cosh/3 sinh/3) = ea ■ .

[H ere, we u sed form ulas (70), (68), and once again (76).] In
p articu lar, l = e° = ea~a = eae a. From this it follow s that the
value of eu is alw ays nonzero no m atter what the value of the
com plex number a.
24. Let us turn to the lem n iscate co sin e cl t. In sectio n 5,
this function was defined for rea l values by the form ula

c l/ = si (82)

Let us take (82) as the definition of cl t for arbitrary com plex t.


If we s e t a = oi/2 and p = —- 1 in (65") and note that s i w/2 = l
[form ula (18')] and si (— t) = — si t [form ula (51)], we obtain

1/ I/“ ,\ V l — si' t / 1 — si2 t/ o o \


c U ““ S1(,2 l ) ~~ 1 —V 1+ si2/ •^

The value of the ra d ica l is chosen in such a way that cl 0 = 1.


Equation (83) can be w ritten in the sy m m etric form

sl^ t —
p cl“ ( -|—si2 1 • cl21 — 1. (84)
54 THE R E M A R K A B L E S I N E F U N C T I O N S

From th is, we get

• <8 3 '>

We note that, if we rep lace t with it, we obtain from (83) with
the aid of (54)

r i - si-’ (U) 1-Psl2/ _1_


cl (it) V 1 + si2 (it) V 1 -slh cl t
(85)

F u rth erm ore, if we rep lace an arb itrary com plex number
t in (83) with — t, we get, by virtue of (51),

= j q r § f r l H / T R T = d '- (86)

Thus, c l t is an even function. With the aid of (83), the e x p r e s­


sion s (65") take the form

si a cl P (1 + s l 2 P) + s ! P c l a ( l - f s l 2 a)
Sl(U+ f,)n=---------------- l ~ s l 2asl2fi----------------- =
si2 a — si2 p (87)
si a cl p (I + si2P) — si p ■cl a (] -p si2a)

To esta b lish the addition theorem for the lem n isca te co sin e
as sim p ly as p o ssib le , let us fir st rep la ce t in (82) with « /2 —
We obtain

c, (t ~ /) = sl / - (82')

From (82), it follow s that

cl (a -f- p) =. si — a] — pj = s! [ ( - — p) - a ] .

Let us rep lace a by w/2 — a and p by —p in the fir s t e x p r e s ­


sio n s (87) and let us rep lace a by m/2 — p and p by — u in the
second ex p ressio n . If we then u se form ulas (82) and (82'). we
obtain

c ! a c i p ( l - - t - s l 2 P) — si a si P (1 T d 2 a)
cl (a -j- P)
“ 1 cl2a • si2 p "
cl2 p —si2 a
( 88 )
cl a cl p (1 f s i 2 a) + si a si p (1 -f- cl2 p) '
F U R T H E R S T U D Y OF C O M P L E X V A L U E S O F T H E A R G U M E N T 55

In con trast with the ex p ressio n s (87), th ese la st fractions


a re not rep resen ted sy m m etrica lly with resp ect to a and p.
H ow ever, the form ulas are of cou rse valid when we r e v e r s e the
ro le s of a and p.
25. We have already noted that all the " red u ctio n fo rm u la s”
of the circu la r functions are con sequences of form ulas (4)
e x p ressin g the ev en n ess of the function cos t and the oddness
of the function sin t and of form ulas (5) ex p ressin g the equality
of the values of cos q and sin q if — n/2. The lem n isca te
form ulas p o s s e s s analogous prop erties: cl t is an even and
s i t is an odd function [cf. form ulas (86) and (51)]. F u rth er­
m o re, the values of cl q and s i t2 are equal if q 4 q = .- w/2 [cf.
form ulas (82) and (82')]. T h erefore, the corresponding " red u c­
tion form u las” obtained by the natural substitution of s i for
sin , cl for c o s , and to for it a re valid:

cl (to — cl t, si (CO — t) =-. si t;


cl ho 4 t) = — cl t, s! (co -j- t) —. —si t;
cl (2to-f- 0 = cl t, si (2to 4 - 1) — si it).

It follow s from th is, in p articu lar, that the lem n isca te func­
tions are p eriod ic functions with period 2«. Now we m ay a s se r t
that, for real values of i, the lem n isca te functions defined by
con sid erin g the in v er se of the integral

on the entire rea l a xis coincide with the lem n isca te functions
introduced in sectio n 5 by way of a geom etrica l approach.
The identity (84) show s that, for rea l values of i, the values
of the lem n isca te functions

x = cl / and y = si t

can be regarded as the coordinates of a point A lying on the


follow ing fou rth -ord er curve (see Fig. 19):

x2+ y2 + x2y2= b (84y)

Together with the point A (x . y), let us con sid er the point
A' (I, y), where I = 1/ x. In accordance with (85) and (54), we have
56 THE R E M A R K A B L E S I N E F U N C T I O N S

— = cl (It). y = Si t = -J Si (It).
^ X cl t

Obviously, I and y sa tisfy the equation obtained from (84') by


rep lacin g x with l / \ but leaving y as it was:

^ + y2+ ^ = 1' or ^2- y 2- ^ v = i- (84")

F I G . 19.

We have thus obtained a new fou rth -ord er curve (one con­
s is tin g of two unbounded branches as shown in F ig. 19). The
coordinates of an arbitrary point A on it a re ex p ressed in term s
of the values of the lem n isca te functions of purely im aginary
argum ent it:

i = cl [it), y T si (it).

The variable t has an e a sy geom etric interpretation. Spe­


c ific a lly , from equations (84') and (84"), we have

(For d efin iten ess, a ll the points a re taken with p o sitiv e a b sc issa .)
T h erefore, for the a rea s of the cu rvilin ear trap ezoids OCAK
and OCA'K, we obtain
y y

u o
F U R T H E R S T U D Y OF C O M P L E X V A L U E S O F T H E A R G U M E N T 57

from which we get

t I = —[area OCAK -(- area OCA Kl-

Such a geom etric interpretation of t in this method of r e p r e ­


sen tin g the lem n isca te functions is valid both for the c a se of a
real param eter t and a purely im aginary p aram eter it.
The addition theorem s enable us to ex p ress the valu es of the
lem n isca te functions for an arbitrary com plex variable / — a -4 - ix
in term s of the sam e functions of the rea l variable o and x.
S p ecifically, when we se t a — a and p ~ i x in the fir s t of e x p r e s ­
sion s (87) and u se the form ulas

si (ix) = I six, (54)


(85)
cl v(ix)' = —
clj—
T,
we obtain

,, , . . si a (1 — s i 2 t) ~|" «$1 t cl a (1 + s i 2 a)
si fa 4 - ix) = ----- — ----- ---- jTi---- — 2— -------- .
1 — si2a si2 t

In this equation, let us rep la ce 1 -sl'-'x with cl2 x (1 -f-si2x) in a c ­


cordance with form ula (84). After dividing through by c l x, we
obtain

si (o-f- ix) = si o cl t (1 4- sl2 T ) T ( sl


I — sl2 <J sl2 T
t cl 0 (1 -)- sl2 a )
(89)

A nalogously, the second of ex p ressio n s (87) y ie ld s, for « — o and


P= A ,

sl2a -f- sl2 1


sl (0 + ix) sl CTcl T (i 4- sl2 t) — i sl T cl 0 (1 A sl20 ) (89')

In a s im ila r m anner, we d erive from form ulas (88) the relation

cl o cl x (1 4- sl2 t) — i sl ct si t (L cl2a)
cl (o + ix) -- 1 — cl2a sl2 t

__________ 1 — sl2 CTcl2 X___________


(90)
cl n cl T ( 1 4- sl20) -4 i sl Osl T (1 4 d 2 T)
58 THE R E M A R K A B L E S I N E F U N C T I O N S

26. Let us now turn to the form ulas for the Jacobian ellip tic
functions. The b asic function sn(t, k) is used to define two other
Jacobian ellip tic functions:

cn (/, A) = 1 ^ 1 —sn2(t, k), dn (/, k) - - \ 1—k2sn2(t, h). (91)

H ere, we take values of the rad icals that assu m e the value
1 at t — 0, that is , at which sn(0, k) -- 0, and we extend the d efin i­
tion of th ese functions to the com plex plane in such a way that
continuity and d ifferentiability are ensured. Thus,

cn (0, k) = dn (0, k) — 1.

The relation s (91) can a lso be rep resen ted in the form s

sn2(/\ 4) J c n 2(i, k) = 1. k2sn2(t, ft)-f dn2(t, k ) = 1 . (91')

It follow s from form ulas (91) that cn(f, k) and dn(/, k) are
even functions

cn(— t, k) = cn (t, k), dn(— t, k) = d n ( t , k). (92)

The notations for the Jacobian functions sn t and cn t under­


sc o r e their sim ila rity with the functions sin t and co s t, r e s p e c ­
tiv ely . If the modulus k = 0, th ese functions reduce to sin t and
cos t. The function dn t then red u ces to the constant 1.
To s e e th is, note that for £ = 0, the integral

becom es

T h erefore, the in v er se function w = sn (t. k) becom es w=rslnL


Form ulas (91) yield

cn(t, 0) = cos t, dn (t, 0) = 1.

With the aid of form ula (91), we can rep resen t form ula (55')
in the form
F U RT H E R S T U D Y OF C O M P L E X V A L U E S OF THE A R G U M E N T 59

'S o t - <93>

w here ft' — Y 1 — k? is the com plem entary modulus (with resp ect
to ft). Form ulas (91) now enable us to e x p r e ss cn(lt, ft) and
dn(*b ft) in term s of the corresponding functions of the com p le­
m entary modulus:

cn(«. 1+ i S ^ = _ > _ ; (94)

dn(«. * ) = y 1 - * ’ sn*(tf. « ) = ) / 1 =

V 1 —sn9 (t, ft') -F (I — ft'2) sn2 (t. ft') _ _ dn (t, ft')


^ cn (t, ft') cn (t, k') ‘ (95)

If in form ulas (93)-(95), we rep lace t with it (and hence it


with — t), we obtain

,, »•. , sn («<. A )
sn (/. k) — — I — - ■
/
cn (it. k )
1 dn (it. ft')
cn (t , ft) =
cn (it. k') dn(t , ft)-
cn (//, ft')

Noting that the com plem entary modulus k' van ish es for ft = 1,
we obtain, with the aid of form ulas (52) and (71),

sn (it, 0) sin (;7) slnh t


sn (*, 1) =
cn (it, 0) cos (it) cosh* ’
1 _ 1 _ 1
cn (t. 1)
cn (it. 0 ) " cos (it) cos h (/) ’
dn (it. 0) _ 1 _ _ J ___
dn (t. 1) = cn (it. 0 ) cos (it) cosh*

C onsequently, for ft = 1, the Jacobian functions reduce to


ratios of hyperbolic functions. The corresponding graphs (for
rea l values of t) are shown in Figure 20. Thus, the Jacobian
functions snit. ft), cn(/, ft), and dn(f, ft) for continuous variation
in the modulus ft from 0 to 1 form , as it w ere, a bridge from
the circu lar functions sin t and co s t and a lso the function
id en tically equal to the constant 1 to the hyperbolic functions:
(sinh *)/cosh t, 1 /c o s h t, and l/ c o s h *, r esp ectiv ely .
60 THE R E M A R K A B L E S I N E F U N C T I O N S

F IG . 2 0 .

27. With the aid of the functions cn(t, ft) and An{t, ft), the
addition theorem (65,v) for sn(/. ft) takes the form

su (a -f' p. ft) =
__ sn (a. k) cn (p. £) dn (p, k) 4- sn (P. k) cn (a. ft) dn (a. ft) __
1 — ft2 sn2 (a, ft) sn2 (ft ft)
„ ______________sn2 (a. ft) — sn2 (p. ft)______________ (9 6 )
sn (a. ft) cn (p. ft) dn (P, ft) — sn (P, ft) cn (a, ft) dn (a, ft)

To d erive addition theorem s for cn (t, £)anddn(Y, ft) from (96),


let us u se the follow ing id en tities, the validity of which can be
checked by u se of form ulas (91'):

(sn n cn p dn p sn p cn a dn a)2 4~
4 - (cn a cn p — sn a sn p dn a dn p)2 =
= (dn a dn p — ft2 sn a sn p cn a cn p)2 4~
-f- ft2 (sn a cn p dn p 4 - sn p cn a dn a)2 = (1 — ft2 sn2a sn2 p)2.

We obtain
cn (n 4- p , ft) = Y 1 — sn2 (a 4 p, ft) =
| 4 ( 1 — ft2 s n- ’ a s n 2 p ) 2 — ( s n a c n P d n P 4 - s n p c n a d n a ) 2
‘ 1 - ft2 s n 2 a s n 2 p

___ c n a cn p — s n a s n p d n a d n p (97)
1— ft2 s n 2 a s n 2 p ’

dn (a f p. ft) ■
— — ft2sn2 (a 4~ P. ft) —
( (1 --- ft ■s n 2 a s n 2 P ) 2 — ft2 ( s n a c n P d n p 4 - s n p c n a d n a ) 21*
1 — ft2 s n 2 a s n 2 p

_ d n a d n P — ft2 s n a s n p c n a c n p
(98)
1 — ft2 s n 2 a s n 2 p
F U R T H E R S T U D Y OF C O M P L E X V A L U E S OF T H E A R G U M E N T 61

T h ese addition th eorem s enable us to ex p ress the values of


the Jacobian e llip tic functions for an arbitrary com plex number
l = o + /i in term s of the sam e functions of the rea l variab les
a and t . H ere, we a lso need form ulas (93), (94), and (95). The
fir st of equations (96) y ield s
sn (a k) —
___ s n ( a , k) cn (it, k) d n (it, k) - | - s n ( i t , h) c n ( a , k) d n ( a , k)
' 1— k 2 s n 2 ( a , k) s n 2 ( i t . k)
sn (a, k) dn ( t , k') - ) - I s n ( t , A ' ) c n ( t . k') c n ( a , k) d n ( a , k) (99)
c n 2 ( t , k') - f k2 s n 2 ( a . k) s n 2 ( t , k')

w here k' = [ / l — k2. The second of ex p ressio n s (96) enables us


to rep resen t the sam e resu lt in a different form :
sn (o -j- it, k)
k) c n 2 ( t, k')
s n 2 (a, s n 2 ( t , k') ( 9 9 7)
sn (o, k) d n (t. k') — i s n ( t, k ’) c n (t, k') c n ( a, k) d n ( a, k)
In the sam e way, by u sin g form ula (97), we se e that
cri ( a -j- i t , k ) =
c n (a, k) cn (t, k') — i s n (o, k') s n ( t . k') d n ( a , k) dn (t, k') (1 0 0 )
c n 2 (t, k‘) - ) - k2 s n 2 ( a . k ) s n 2 ( t , k')

The read er can use form ula (91) to verify that the sam e
resu lt can be rep resen ted in the form
c n ( o —|— i t , k )
c n 2 ( o. k) -(- k'2 s n 2 ( a , k) s n 2 ( t , k') ( 1 0 0 7)
c n (a. k) cn (t. k') - f i s n (o, k) s n ( t , k') d n ( a . k) dn (t, k')

A nalogously, for the function dn t, we obtain


dn ( o - h i t . k ) —
d n (0 , k) cn (t, k') dn ( t , k') — / k2 s n ( a , k) c n (a, k) sn ( t, k') __ _
c n 2 (t. k') -)- k2 s n 2 ( o, k) s n 2 ( t . k')
k 2 c n 2 ( a , k) + k'7 c n 2 ( t , k') (101)1
dn(a, A)cn(T, /i')dn(T , k') -j- s n ( a , A) c n ( o , A ) s n ( T , A ' )

2 8 . Let us se t
1
--------------------- = K(k^r=K. (102)
J V(\ — y7) (1 ~ A2y2l

w here the integration is along the segm en t of the rea l axis


betw een the points 0 and 1, so that K(k) is a p ositive rea l num­
b er. It follow s from (102) that

sn [K (k ), k\ = 1. (103)
62 THE R E M A R K A B L E S I N E F U N C T I O N S

T h erefore,

cn [K (k), k\ = V 1 - sn2 [AT(ft). k\ = 0


and
dn [ K ( k) , k ] = Y l — k* sn2 [AT (ft). k\ = V 1- *2= k' - (1 0 4 )

For sim p lic ity of w riting, let us drop the sym bol for the
modulus k. If we then se t a — K and p = — t in form ulas ( 9 6 ) - ( 9 8 ) ,
we obtain

sn K cn t dn / — sn / cn K dn K cn / dn t cn t
sn (K 0= 1 — k2 sn2 K sn2 t 1 — k2 sn2 t dn t ’
(105)

cn K cn t 4 sn K sn / dn K dn t _
cn (K — t)
1 — k2 sn2 K sn21
_ k' sn t dn / , sn t
(106)
~ 1 — A2 sn2 / k

dn K dn / -(- k2 sn K sn / cn K cn t _ k' dn / k’
dn (K — t)
1 — k2 sn2 K sn21 1 — k2sn21 dn t ’ (107)

If we now rep la ce / by — t, we obtain


fn t
sn (K + /) = ~dnY = sn ^ ~

cn (K + /) = — k' = — cn iK — /); (108)

dn (K r 0 = dn (K — t).

In the theory of Jacobian e llip tic functions, form ulas (105)-


(108) play a role analogous to the reduction form ulas in tr ig ­
onom etry.
If we rep lace t by t K in form ulas (108), we find further
“ reduction fo rm u la s” :

sn ( / -f-2/C) = — s n /, |
cn (t 2 /0 -= — cn t, / (109)
dn (t -f- 2/C) = dn /. J
From the la st form ula, it follow s that the function dn t is
p eriod ic with period 2/0 F inally, in the fir st two of form ulas
(109), let us rep lace t by / -f 2/0 This gives us

sn ( / -|- 4 /0 == sn /, cn (/ 4K) — cn /. (H O )
F U R T H E R S T U D Y OF C O M P L E X V A L U E S OF T H E A R G U M E N T 63

Consequently, sn t and cn t are a lso p eriod ic functions—in


th eir c a s e s , with period 4/C.
29. The form ulas of the p recedin g sectio n enable us to
ch a ra cterize com p letely the behavior of the Jacobian functions
of a rea l variable without any other calcu lation . Let us begin
with the function y = sn(f. J k ) (for 0 < k < 1). From the definition
of this function, sp e c ific a lly , as the function in v er se to the
in tegral

* - /’ - n- ^_____

it follow s that it in c r e a se s m onotonically from 0 to 1 as t


in c r e a se s from 0 to

i
K{k) = f _______ _________
0 /(l-V )

(cf. sectio n 11). The fir st of form ulas (108) show s that the
graph y = sn (/, k) is sy m m etric about the straigh t lin e / = /C
and hence has the shape of a half-w ave in the in terval 0<j/<;2A :
(se e Fig. 21a). F u rtherm ore, if we rep lace t by — / i n the fir st
of form ulas (109), we get

sn (2K — /) = — sn (— /) — sn t.

By com paring sn (2K + / ) with sn (2/C — 0. we obtain

sn (2K -(- /) = — sn (2K — t).

This m eans that the graph of y = sn(/, k) is sy m m etric about


the point (2/C, 0). Consequently, in the in terval 0 it has
the sam e shape as does a sin e wave in the in terval 0 ^ / < 2 n
(see Fig. 21a). Since we know that the function sn(/, k) is
p eriod ic with period 4K, we can now proceed to con stru ct its
graph.
To determ ine the nature of the graph of the function y — cn i/, k),
we can begin with the relation cn (/, k) - - y 1 — sn-<7, k), from
which it follow s that this function d e c r e a se s m onotonically
from 1 to 0 in the in terval 0 / K. F u rth erm ore, the graph
is constructed just as above, by u se of form ulas (108)-(110).
We s e e that this graph looks much like that of the co sin e (see
F ig. 21b).
64 THE R E M A R K A B L E S I N E F U N C T I O N S

FIG. 21.

In an analogous m anner, we obtain the graph of the function


y = dn (t , k). The form ula

dn (t, k) = Y 1 — k 2 sn2 (t , k)

show s fir st that this function d e c r e a se s from 1 to 1 — k2 = k'


in the in terval 0 < t < K. The graph is then extended to the
en tire real axis by using the appropriate relation s (108) and
(109). It is shown in Figure 21c.
It follow s from the above an alysis that a ll rea l zero s of
sn (t, k) [that i s , the real roots of the equation sn (/, £) = 0] are
exhausted by points of the form t — 2mK{k), w here m is an
arbitrary in teger.
A nalogously, the rea l zero s of cn(7. k) are of the form

/ - -- (2in — 1)K (k).

The function drift, k ) has no rea l zero s at a ll (if 0 <r k < 1).
30. Let us show that the gen eralized sin e s(t) can in all
c a s e s be rep resen ted in term s of the functions sin t and sinh t
or in term s of the Jacobian functions. S p ecifica lly , let us show
that, for n -- o (where n is the co efficien t of z4 in the ex p ressio n
F U R T H E R S T U D Y OF C O M P L E X V A L U E S OF THE A R G U M E N T 65

1 - f mz 2 ~f n z 4), it can be ex p ressed in term s of sin t or sinh t


and that, for n =£ 0, it can be ex p ressed in term s of the Jacobian
functions.
If n ~~ 0 and m =£ 0, then I -f- mz 2 can be w ritten in the form
1 ± k 2z 2 , w here a. > 0. Then, if we rep resen t the relation
w = s(t ) in the form

W
t - f ■ '/_
J V \T Iv ’
0

and make the change of variable t, — kz, we obtain

Kw

t
dt
V\ ± U

From this we have eith er k w =sin(?,Z) or k w = sinh {kt). Thus, in


the c a se in question,

s (/) = -i- s i n (kt), or s(t) = — sinhU, (). (I ll)

Suppose now that n =t= 0 and, as alw ays, that m2 — An 0, We


have two c a se s:
CASE I. m2 — An > 0. H ere, the roots x x and x., of the equa­
tion

x 2 4 - mx -j- n ~ 0

a re real and d istin ct. We have

x 2 -j- mx -f- n —- (x — A',) (x — x

or, if we rep la ce x with 1 / z - ,

I t mz'2 n z 4 --- (1 — t , 2 2) (1 x.,zl ).

At this point we need to con sid er three su b ca ses: (a) a-, and v.
both p o sitiv e, (b) x x and x 2 of opposite sign , (c) a, and x, both
negative.
Subcase (a) : We can assu m e that a, and aa have the values
k2 and |i2, w here k > p > 0. Then,

I - j - m z 2 -(- n z 4 = ( 1 — k 2z 2) ( l — \ i 2z 2 ).
66 THE R E M A R K A B L E S I N E F U N C T I O N S

C onsequently, the equation w = s(t) takes the form

7(1 - A>z-)(1

If we make the change of variable i ^ = l z , we obtain

?----------- .where 0 < k


V ( l- C 2) ( l - n 2)

Thus, in this su b case

w ----- 5 (t ) — s n 0 , t , k). ( 112 )


A

Subcase (b)- We can a ssu m e that x, = k1 andx, = —p-, where


h and ii are both p o sitiv e. We obtain the follow ing decom position
into factors:

1 ■+■mz- f m ' = (l - 1 - x,r-) ----- (\ -)- p-z2).

If, in the in tegral

we make the change of variable C— ]/ 1 —l 2z2, we obtain

l : /•*..*
1 ■' if;

I'd - C ) ( l a- ;q

w here

0 .. k 1.
V 4- h

If we se t
F U R T H E R S T U D Y OF C O M P L E X V A L U E S O F T H E A R G U M E N T 67

we obtain
V I - }.’w 7

K- f
0
_______ dj

or

|/ l — X2w 2 = sn (k - y X2+ u2 1, k) =
_ cn (YX2 -f- |x; t, k)
_ dn ( / X2 + \x2 t, k) [cf. (105)]

in su b case (b), so that

cn- i \ rK7 1, k) k' sn (]/ X3T |c’ i, it)


W — 5 (/) —
w dn y h- 4- (i! I- h i i. dn (y a2- ^ 2 k)

Noting that, in the p resen t c a se .

k' = ]/ 1 - if2 = - r -
V X2 4-1*3

we can rew rite our la st resu lt in the form

1 sn IV X3 4 |i ; t. h )
(113)
1 V +- n : dn 11 X- + (i31, k)

For X = p = 1, the in tegral


XL'

J
o f(l d'-t-n2*2)

coin cid es with

. yr
T h erefore, s(t) — si (/), and we obtain

si t ± " I1 w ( H 3 y)
L? _L_j
68 THE R E M A R K A B L E S I N E F U N C T I O N S

For cl t, we obtain

= c n ( / 2 /.-p L ). (113")

Thus, the le m n is c a te functions can be expressed, in t e r m s


o f the Jacobian functions with modulus 1/V~2.
Subcase (c): When x, and x 2 are both n egative, they can be
rep resen ted in the form s — X 2 and — \ i 2, and we can assu m e that
X > > 0. T h erefore,

1 -(- mz2-f- nz4 = (1 — x 1z 2) (1 —x 2z 2) = (1 X2z 2)(1 (i2z 2).

In the integral

t= I - r

we make the change of variable I = X z / \ r \ (cf. calcu lation


in sectio n 17, where X= 1 andp = £). We then obtain

ii
t rw here 0 < k < I-
u V (1 - £ 2) (1 - * 2£2)

From th is, we get

Xw
sn (Xt, k),
/r p iv
or
sn (Xt, k) 1 sn (Xt, k)
w = s it) — — —— (114)
X K 1— sn2 (Xt, k) X sn (W, k)

It rem ains for us to con sid er


CASE II. m- — 4n < 0 (obviously, here n > 0). Since the roots
x, and x , of the equation

x 2 -j- m x -j- n = 0
F U R T H E R S T U D Y OF C O M P L E X V A L U E S OF THE A R G U M E N T 69

are im aginary, the roots of the equation

44+ ^ 3+ " = 0.

defined by

are a lso im aginary. F urtherm ore, th eir rea l parts are nonzero
(because oth erw ise the sq uares of the roots would be rea l num­
b ers). Noting that si- A £3 and 14 m ust a lso be p a irw ise conju­
gate and changing the num bering if n e c e ssa r y , we have

= — £2 = « + £* = — £4 = “ — ^ (a > 0, (1 > 0).

In the decom position

V -f rng + n = (I - y (i - 12) (I - U) {I - 10
let us rep lace I with l A . We then obtain the follow ing d ecom ­
p osition into factors:

I + mz2 -f nz4 —
= | l + ( a f t ' P ) ^ l l l - R a — < P ) 2 ] ( 1 — ( a | - < P ) 2 | | l — (a — «P) z\ =

= 11 -(- 2az -j- (a3 +- P2) 22| 11 - 2az -f- (a2 + p2) z2| .

O bviously, a2 + p2=lA *" (from which we get 0 < a < V~n). Con­
sequently, our in tegral is of the form

/= / .
(i V (1 4- 2 a z + Y~rTz}) (1 - 2nz 4 V'TTr'y)

Let us show that it reduces to one of the c a s e s con sid ered


above when we make the substitution

When we make this substitution and perform som e m anipulations


in the in tegral, we obtain
70 THE R E M A R K A B L E S I N E F U N C T I O N S

1+V n

I •V n •

/ ■
A ( v n — a ) + ( y n f < i ) t 2] [ * V « + a ) + ( V n — a ) t 2]

I+ V n w

l i ) tt IV

dt.
F ./ V o + * 2t,2) ( l + | i ! ? )

, -a I >0.
Vn —a V V~n + <i

Let us se t

d'Q ■---= A .
VO 4 -W )(l+ l» V )

Then,

1+V n w

i - !■ n t#(
P/ + A = / .. ■
./ | ^ ( l + A2£2)(1 + i ' v >

Thus, this brings us to Case I (c). T h erefore, in accordance


with form ula (114),
■]
1+ V nw 1 sn (pt -f A. h)
4 acn (P* -+- A, k) ’
l —- y n w
w here
k = yrv=p ^

From this we g et, finally.

^ lt) = _ L . sn (P< -+ >1, fe) — A. cn (P< + A fe)


4_ sn (p/ -f- A. k) 4 - cn (P< -f- A. k)
Vn
FURTHER S T U D Y OF C O M P L E X V A L U E S O F T H E A R G U M E N T 71

Thus, we have shown in this sectio n that the g en era lized sin e
with n=h 0 can be ex p ressed in term s of the Jacobian functions
with appropriate modulus k and, for n — 0, in term s of the cir c u ­
lar or hyperbolic sin e. All this w ill give us the right in what
follow s to confine o u rselv es to a study of only th ree sin e s
(together with the corresponding co sin es): the circu la r, tr ig ­
on om etric, and Jacobian (the sin e am plitude). H ow ever, together
with th ese, we sh a ll a lso con sid er the lem n isca te functions sin ce
it is for this sp ecia l c a se that the p rop erties of the Jacobian
functions m anifest th em selv es in the sim p le st m anner.
CHAPTER 6

Zeros and P oles.


S im p le and D ouble P eriod icity.
The Concept o f an Elliptic Function

31. If we com pare form ulas (75), obtained above for circu ­
lar and hyperbolic functions with form ulas (89), (89'), and (90)
for the lem n isca te functions or with form ulas (99)-(101) for the
Jacobian functions, we s e e a sharp d ifferen ce betw een form ulas
(75) and the other six . S p ecifica lly , the circu la r and hyperbolic
functions are rep resen ted as polynom ial functions of sin a,
cos a, sinh a, and co sh ;o, w hereas the lem n isca te functions w ere
rep resen ted as fraction al ex p ressio n s in term s of s i a, c l o,
s i t, and cl t and the Jacobian functions w ere ex p ressed as
fractional ex p ressio n s in term s of the Jacobian functions of the
rea l variab les o and t .
T h erefore, the functions sin /, cos t, sinh /, and co sh t are
defined and have fin ite values for ev ery com plex /. On the other
hand, the functions sib cl/. sn(/, k), e n d . k) and dn(/, k) a re not
defined for th o se values of / at which the denom inator of the
fraction in question van ish es.
Let us look into this phenomenon m ore c lo se ly , beginning
with s i /. The denom inator of the fraction in form ula (89) is
equal to 1 — si2 a • si2 t. It van ish es if and only if

si a • si x ± 1. (115)

But si a and si t are rea l num bers that do not exceed 1 in ab so­
lute value. (We r e c a ll from sectio n 5 that si a is the signed

73
74 THE REMARKABLE S IN E F U N C T IO N S

length of the chord of the lem n isca te.) T h erefore, equation (115)
is equivalent to the two equations

s! o ± 1 and s it- - ± 1. (1 1 6 )

It follow s from the sa m e geom etric representation of si a that


the condition sla = 1 is equivalent to the condition o = w/2 -+- 2/na>,
w here m is an arb itrary integer (2o> being the period of slo).
Since slo is an odd function, the condition o = —1 is equivalent
to the condition o = — w/2 -f- 2mu>. Thus, the denom inator of the
fraction (89) van ish es at all points with coordinates

o — ± — + 2m o), t = ± — - f - 2/i o )

(m = 0, ±1, ±2. n = 0, ±1, ±2, . . .)

and only at such points.


The corresponding com plex num bers t = a-[-li are exhausted
by the form ula

t — (4m± 1)y + (4 « ± l) - y -

But if m ranges over all in tegers 0, ±1, ±2, ± 3 ........the quan­


tity 4m ± 1 a ssu m es a ll odd values ±1. ±3, ± 5 ..........
T h erefore, the values of t can finally be rep resen ted in the
form

f = (2 p -l)f-K 2 y -l)4 * (117)

w here p- = 0. ±1. ± 2, . . . and q - - 0, ±1, ± 2 .........


Let us look at one of th ese points. l f o = - ( 2 ^ - l)io /2 ,th e n
slo — (— l)p l, as can e a sily be derived from observations made
above. T h erefore, sl2o = l , and form ula (83) y ield s clo = 0.
A nalogously, for t = (2 <7 — 1) w /2, we have s l t = ( - l ) ‘, ‘ l and
c l i - 0 . If we substitu te th ese values into the right-hand sid e
of form ula (89), not only the denom inator but a lso the num erator
in the form ula w ill obviously vanish and we cannot draw any con­
clusion as to the behavior of si / as — l) u /2 - ) - ( 2 q — l)wi'/2
im m ed iately. Form ula (89') sa v es the situation. H ere, the
denom inator of the fraction is equal to 0, just as b efo re, but the
num erator s12o-|-sI2t = 2 and thus is nonzero. From this it
follow s that si t approaches co as t —>i2p — 1) <o/2 — |—(2<y— 1) ©f/2.
(H ere, we u se the fact that the lem n isca te functions of a real
variable a re continuous.) We can now extend the definition of
si t, to the en tire com plex plane by settin g
ZEROS AND POLES 75

sl [ ( 2 p - l ) 5 - K 2 9 - l ) - £ ] = oo.
p = 0, ±1. ±2, <7= 0. ±1. ± 2 .......... (118)

If we refer to a point in the com plex plane at which a func­


tion b ecom es infinite (that is , at which the function approaches
oo as the argum ent approaches that point) as a pole of the
function, we may now say that form ula (118) provid es a ll the
in finitely many p oles of the lem n isca te sin e.
A nalogously, it would be p o ssib le to d eterm ine all the poles
of the lem n isca te co sin e from form ulas (90). H ow ever, it is
sim p ler to start with form ulas (82):

Cl t sl ( - — t 'j .

It follow s from this form ula that cl t approaches oo as t ap­


p roaches a point a if and only if s l( u /2 — /) approaches oo as
<s>/2 — t approaches o>/2—a. From th is, we get, just as above,

J — a ■= ( I p — 1) ^ 4- (2q — 1 ) ;

that is ,

a — 2 (— p 1) -fj- -f- 12 ( q -)-• 1) 11— ■

If we rep lace 1 in this equation w ith // and if we rep la ce


— q -\- 1 with q', we s e e that a ll the p oles of cW are rep resen ted
in the form

t = p'<* + {2qr - 1) — (119)

w here p' and q' a re any in teg ers.


32. By an analogous procedure, we find the p oles of the
Jacobian functions. Let us begin withsn(^, k) (where 0 < k < 1).
The denom inator of form ula (99) is of the form

cn2(T, k 2 sn2 (a, A)sn2(T, k'). (120)

Since the values of the Jacobian functions are real when o


and t are r e a l, vanishing of this sum is equivalent to the two
conditions

cu (c. k') = 0, sn (o, fc)sn(i, k ’) — 0.


76 THE REMARKABLE S IN E F U N C T IO N S

But if cn ( t , k') — o, then sn ( t , k ' ) = ± I [cf. (91)]. T h erefore, we


s e e that (120) van ish es if and only if the two equations

cn ( t , k') — 0, sn (o , k) = 0 (1 2 1 )

hold sim u ltan eou sly. From this it follow s in accordance with
sectio n 29 that

a = 2mK (k) and t = (2n — \)K(k')-,

that is ,

t = 2mK (A) -f- i (2n — 1) K (A').

Denoting K (A) by K and K (A') by K', we rew rite th ese valu es in


the form

t = 2mK ±L(2n - \ ) K ' . (122)

w here m and n are any in teg er s.


How ever, if we substitute the values that we have found for
o and t into form ula (99), we s e e that not only the denom inator
but a lso the num erator van ish es [sin c e sn (o, A) = cn (t, A')—0],
T h erefore, this form ula does not enable us to draw any con­
clu sio n s im m ed iately regarding the behavior of sn (t. A) as
t —►2mK -T (2n — 1) iK'■ Form ula (99') com es to our aid. Since
sn (r, k ' ) = ± 1 when cn(x, k') — 0 [by the fir s t of form ulas (91)],
we have sn2(t, A ') = l, and form ula (99') show s that sn (/, £)->oo
as t approaches any of the points (122). C onsequently, the
valu es found for t rep resen t the p oles (all of them) of the
Jacobian function sn {t, k).
By an analogous p rocedure, we can s e e that the valu es (122)
constitute a ll the p oles of the functions cn (t, k) and dn (/, A).
Furtherm ore, the fact that the p oles of th ese functions m ust
coincide with the p oles of sn (t , k) follow s from id en tities (91).
33. In what fo llo w s, we sh a ll a lso need to know the zero s
of the functions that we a re studying, that is , all the points in
the com plex plane at which the functions vanish. Let us begin
with the circu la r sin e . From the fir st of form ulas (75), we have

sini = sin (a -(- ir) = sinacoshr i sinhr c o s c t . (757)

From this it follow s that the equation

sin t = 0 (123)
ZEROS AND POLES 77

is equivalent to the sy ste m of equations

sin a • coshr,= 0,
(124)
sinh r ■cos a = 0.

Let us look at the fir s t of equations (124). Since co sh r =£ 0


(rec a ll the g eo m etric rep resentation of co sh x in sectio n 2 or
form ula (32): co sh t = e‘ + e~t ) / 2 ) , we conclude that s in o = 0.
T h erefore, o = mji, w here m = 0, ±1, ± 2 ......... C onsequently,
coso — (—\)m 4= 0. The secon d of equations (124) y ield s sin h x = 0 ,
s o that r = 0. (R ecall the geom etric rep resen tation of sinh r in
sectio n 2). Form ula (31) leads to the sam e con clusion . Thus,

t = a -f- lx = mu, m — 0, ±1, ±2, . . . . (425)

This ex p ressio n takes ca re of a ll the zer o s of sin t. We s e e


that none of them is im aginary. Extension of the definition of
the circu la r sin e to the com plex plane d oes not yield any new
z e r o s.
We leave it as an e x e r c is e for the read er to u se the second
of form ulas (75) to show that a ll the zer o s of co s t are r e p r e ­
sented by the e x p ressio n

/= m = 0, ±1, ± 2 .......... (126)

F in ally, by the sa m e p rocedure, we can find the zer o s of the


hyperbolic functions. H ow ever, it is sim p ler to achieve this end
by u sin g the relation s

sin(h) = j'sinht, cos (it) = c o s h t.

It follow s from the fir st of th ese that the equation sinh t = 0 is


equivalent to the equation sin(//) = 0. T h erefore, lt — mn and
hence

t — (— m)nl = n n l.

w here n is a lso an arbitrary in teger. In just the sam e way, we


can show that the zer o s of co sh t are given by the form ula

t = (2 n - I)— ,

w here n is an arb itrary in teger.


78 THE REMARKABLE S IN E F U N C T IO N S

34. Let us now seek the zero s of the lem n isca te sin e . U sing
form ula (89'), we conclude that the equation si t = si (a i t ) = 0
im p lies that sl2o si2t = 0. Thus, we obtain the two equations

From this we get o - = m o v T = nm (where m and n a re arb i­


trary in teg er s). Consequently,

I = m<s> na>l, m — 0, ±1, ±2, n — 0, ±1, ±2. . . . . (128)

H ow ever, for si a — 0 and s I t = 0, the denom inator of the


fraction (89') van ish es. T h erefore, form ula (89') does not te ll
us the values of si/1 at the points that we have found. Let us turn
to form ula (89) for s1(cj-(-/t). H ere, the num erator of the fraction
is equal to 0 when equations (127) are sa tisfied but the denom i­
nator is equal to 1. T h erefo re, si (mm b nmi) 0 and, hence,
form ula (128) does indeed give the zero s of si/1 [a ll of them
sin c e form ula (89') im p lies that the conditions found are n ec­
e s s a r y for vanishing of si/].
To find the zer o s of cl/, the sim p le st procedure is to use
form ula (82):

From this form ula, it follow s that equation cl / = 0 is equivalent


to the equation s l ( « / 2 — /) = 0. T h erefo re, « / 2 — / = mm-f-nml.
From this we get

t — [2 (— m 1) — 1] —-\- (— n) — (2m' — 1) y -[- n’asl, (129)

w here m' and n' are arb itrary in teg er s. Form ula (129) provides
a ll the zeros of cl t.
35. Form ula (99') is convenient for finding the zero s of
sn (/, k). From this form ula, we s e e that a n e c e ssa r y condition
for sn (o -|-/T. k) to vanish is

sn2 (o, £ ) c n 2 (x , A ' ) + s n 2 (T , k' ) = Q.

This condition is equivalent to the two equations


sn (a, k) cn ( t , k ') = 0,

sn ( t , k') — 0.
ZERO S AND POLES 79

But the second of th ese, together with (91), im p lies that


cn ( t, k ' ) ± 1 A 0. T h erefo re, the n e c e ssa r y conditions for
sn (a -+- a) to vanish are

sn (o, k ) — 0,
(130)
sn (t, k ' ) — 0.

T h erefore (cf. sectio n 29),

o — 2mK {k), x — 2nK (k') ;


that is ,
t — 2mK -J~ 2nlK' (131)

w here m and n are arb itrary in teg ers.


How ever, form ula (99') is not suitable for evaluating sn(h k)
at th ese points s in c e , when conditions (130) are sa tisfie d , not
only the num erator but a lso the denom inator of the fraction
rep resen tin g s n (t, k) v an ish es. T h erefore, let us turn to form ula
(99). H ere, the num erator is equal to zero but the denom inator
is nonzero [we re c a ll that end, k')— ± 1], Thus, (131) rep resen ts
the zeros of sn(/, k) (all of them).
With the aid of form ulas (100') and (100), the read er can
show that

t=-.(‘2m - ])K + 2niK' (132)

provides all the zer o s of enp, k), and he can use (101) to show
that

t---=(2m — \) K {2n — 1)i.K' (133)

y ield s all the zer o s of dnp, k) (where m and n a ssu m e a ll in te­


gral valu es).
36. With regard to the p eriodicity of the functions that we
are studying, up to now we have con sid ered only th eir rea l
p eriod s. H ow ever, in extending the definition of the functions to
the com plex plane, im aginary periods can a lso a r is e . An exam ­
ple of this kind is the exponential function.
This function e‘ as a function of a rea l variable is not
p eriod ic. H ow ever, it does have the purely im aginary period
2rU. To s e e th is, let t ~ o + t i , where o and i are real num bers.
Then, from form ula (77),

gt __ ea+iT_ ea (cos x d- / sin t).


80 THE REMARKABLE S IN E F U N C T IO N S

Let us rep lace / with / + 2n/. Since 14- 2ni = o (x 4- 2n) I, this is
equivalent to rep lacin g t with t-)-2 n . We obtain

e‘ +2ni = e<r+(T+2n)i _ ea [cos ^ _|_ 2n) -(-/ sin (x -j- 2n)\ =


= e° (cos x -)- i sin x) = e1

for arbitrary t. From this it follow s that 2m is a p e r i o d of the


function e1.
Let us re c a ll that a number A is called a period of a func­
tion / ( / ) if the equation
f ( t + A) = f(t)

is sa tisfie d for every t for which the function is defined.


If f (t) has at le a st one nonzero period, it is said to be a
p e r i o d i c function. It follow s from the definition that, if A is a
period of f( t), then any in tegral m ultiple of A is a lso a period
of /(/).
It was shown in sectio n 22 that 2n is a period of the function
sin t, treated as a function of a com plex variab le. C onsequently,
any in tegral m ultiple of 2jt is a lso a period of that function. Let
us show that the co n v erse is true: Any period of sin t is an
in tegral m ultiple of 2n. Suppose that A is a period of sin t.
From the identity

sin(/ + A) = sin t

we obtain, for 0,
sin A = sin 0 = 0.

Consequently, A is zero of the circu la r sin e. A ccording to


form ula (125), A m ust be equal to so m e m ultiple of n; that is ,
A —mn for som e in teger m. Let us denote by k the rem ainder
obtained when we divide m by 2. Then, m=-2 p k, where p is
an in teger and k is eith er zero or 1.
T h erefore,

sin (t A) — sin (( -)- kn -}- 2pn) = sin (/ feix).

The fact that A is a period of sin t im p lies that

sin(i f-k:i) = sin t

for arbitrary t. If k ~ 0, this condition is sa tisfie d . If k = 1, it


cannot be sa tisfie d id en tically sin ce sin((-(-^) = — sini. F in ally,
ZE R O S AND POLES 81

we conclude that fe = 0, that is , that A = 2pn, which was what we


w ished to show.
In an analogous m anner, we can show that any period of co s t
is an in tegral m ultiple of 2ji. In sum m ation we may a s s e r t that
the tran sition from a rea l to a com plex variable does not bring
about any new periods in the study of circu la r functions.
The situation is d ifferent when we study hyperbolic functions.
F orm ulas (52) and (71) provide the quickest path to what we
w ish to show:
sin(d) = ( sinh g cos (it) = cosh<,
or
sinh(d = —;sin(u), cosh(t) = cos (it).

From th ese form u las, we have, for an arbitrary com plex num­
ber A ,
sinh (( A) = —i sin (it -)- iA), cosh (t -\- A) = cos (it -(- iA).

T h erefore, A is a period of the hyperbolic functions if and only


if /.I is a period of the circu la r functions, that i s , if iA --- 2pn,
w here p is an in teger o r, in other w ords, if A = (— p) 2ni = 2nip't
w here p' -= 0, ±1, ±2, ±3, . . . . C onsequently, the hyperbolic
functions sinh t and co sh t are p eriodic functions, and all th eir
p eriods are given by the form ula A-^2p'ni, w here p' is an
arb itrary in teger.
We s e e that the tran sition to a com plex variable d isc lo se d
the p eriod icity of the hyperbolic functions (these are not periodic
if we con sid er them as functions of a rea l variable).
37. Although the periods 2p.i of the circu la r functions are
all rea l num bers and the periods 2p'ni of the hyperbolic func­
tions a re purely im aginary (except for the period 0), both th ese
c la s s e s of functions have in com m on the fact that any period of
the function in question is an in tegral m ultiple of so m e sin g le
num ber. In the c a se of the circu la r functions, this number is
2.i (or — 2jt); in the c a se of the hyperbolic functions, it is 2ni
(or — 2ii). If ev ery period of a periodic function is an in tegral
m ultiple of som e sin g le period, known as the fundamental period,
the function is said to be a singly p eriodic function. Thus, the
circu la r and hyperbolic functions are sin gly p eriodic functions.
Let us turn to the lem n isca te functions, beginning with the
lem n isca te sin e. Noting that si w = 0 and
82 THE R E M A R K A B L E S IN E F U N C T IO N S

we obtain the fir s t of ex p ressio n s (87)

s i ( t - f - to) l— — si t. (134)

We s e e that to is not a period of si t . From form ula (134),


it follow s that

s i ( t - f - 2(D) = si \ ( t + w) + 0 )] = — si ( / + w) =- si t. (135)

T h erefore, 2co is a period of the lem n isca te sin e (as we noted


in sectio n 25), and hence ev ery in tegral m ultiple of 2w is a lso
a period. H ow ever, periods of the form 2/i<i> do not exhaust all
the periods of the lem n isca te sin e ! To s e e th is, let us note that
si (o/) — i si a) = 0 and cl («/) = l / c l «■ — — 1. If we se t a = t and
p = /w, in the fir s t of form ulas (87), we obtain

si (t - j - io/) — — s i / . (136)

From this it follow s that is not a period of s i/. How ever,


we conclude from this sam e form ula (136) that

si if 4 2<nt) - - si |(t -j- ou) -fid/] — — si if t a)/) s i /; (137)

that is , 2wi, and hence any in tegral m ultiple of 2w/, is a period


of si t . Thus, the le m n isc a te sin e p o s s e s s e s in finitely many
rea l and in fin itely many purely im aginary p erio d s. From this
it follow s that si t , being a periodic function, is not a sin gly
p eriod ic function. To s e e th is, let us suppose the opposite,
nam ely, that each of its periods is an integral m ultiple of som e
nonzero fundamental period a. Then, th ere m ust e x ist nonzero
in tegers m and n such that

2(o — ma, 2(o/ na .

But this would then im ply that l — n / m , that is , that i is a real


num ber, which is not the c a s e . Thus, si / belongs to the sp ecia l
c la s s of p eriodic functions that are not sin g ly p eriodic.
By u sing form ulas (134) and (136), we conclude that

si (/ ~f- (o -)- /(o) = si |(/ T i(o) -+- (<>] = — si (/ -(- to) = si t.

T h erefore co-j-lw is a p eriodic of s i/, as in any in tegral m ultiple


Of (0 T /(D.
38. From the p eriods that we have found, we can construct
new ones by com position. For exam p le, from the fact that 2 moo
ZER O S AND POLES 83

and n ( go 4- ho) a re both periods of si t (where m and n are both


in te g e r s), it follow s that

2mo> -)- n (to 4- go() (138)

is a lso a period of si t.
Let us show now that every period A of the function si t can
be rep resen ted in the form (138) for so m e in tegral values of
m and n. Of c o u rse, for those periods that we have already
found, this is obvious:

2cd = 2 • 1 • to -)- 0 • (u -f- o>/), (m = 1, n = 0);


2di/ = 2 ■ — 1 ■(i) -)- 2 • (o) 4~ (0/), ( m — — 1, d - 2 ) ;
0) cot — 2 • 0 • (D-f- 1 • (to -j- io(). (m = 0, n = 1).

H ow ever, our a sse rtio n is of a gen eral ch aracter and it


sta te s that si l has no other periods than th ose that can be
rep resen ted in the form (138).
Suppose that we have the identity

si (t 4 /l) = sl/. (139)

Setting / = 0 in this identity, we find

sM =r 0.

From this it follow s that A is a zero of the function si t ; that is


[cf. (128)],

A = mu + rual,

w here m and n are in teg er s. Let us rew rite A in the form

A = (m — /i)u4-n(w + w0 = L n (w 4" Gtf).

Let k denote the rem ainder obtained on dividing p by 2; that is ,


p = 2r~\-k (where r is an in teger and k is eith er 0 or 1). Then,

A — 2r«) n (m 4 w/) -f- kw.

When we su bstitute this value into (139) and note that 2/-co 4-
-f- n ( go -4- (ai) is a period of the form (138), we conclude that

si (t 4 - £«) = si t
84 THE REMARKABLE S IN E F U N C T IO N S

id en tically. This condition is sa tisfie d by only one of the two


p o ssib le valu es of ft, nam ely, ft = 0 [that ft = 1 is im p o ssib le is
shown by form ula (134)].
Thus, any period A of the function si t is of the form

A = 2r<o n (oj -(- <o<),

as we w ished to show.
The above a n a ly sis has brought out the fact that an arbitrary
period of si/ can be rep resented as a lin ear com bination, with
in tegral c o e ffic ie n ts, of the two fundamental periods 2w and
o)-(-(i)/. P erio d ic functions p o s se s s in g such a property are
called doubly p er io d i c. Thus, sl(/) is a doubly p eriodic function
with fundamental p eriods 2u> and o> +■ w/. The reader may v erify
that we could a lso have chosen w —o« and w - f «/, for exam p le,
as our fundamental p eriod s.
By using the form ula

cl t = si ^ - /J,

the read er can e a sily d erive from what has been shown that cl t
is a lso a doubly p eriod ic function with the sam e fundamental
p eriod s.
39. An analogous a n a ly sis can be made for the Jacobian
functions. Let us con sid er, for exam p le, sn(/, ft). In sectio n 28,
it was shown that

sn (f -T 2 K) = — sn t,

from which it follow s that

sn (t f 4 K) = sn t ;

that is , AK is a rea l period of sn /. Let us show that sn r a lso has


im aginary p erio d s. From form ula (131), which gives a ll the
zero s of ft), we conclude that t — 2 i K ' \ s a lso a zero:
s n i i ,

sn (2/A'/ , ft) = 0. (140)

F u rth erm ore, from form ulas (94) and (95), we obtain, for t — 2K',

____1____ dn (2K\ ft')


(141)
c n (2iK\ ft) = dn (2iK'. ft) cn (2K', ft')
cn (2K' . ft')
ZEROS AND POLES 85

But from form ulas (109), it follow s that, for r — 0,

cn (2K. k ) = — cn 0 = — 1, dn (2K, k) = dn 0 = 1,

or, if we rep lace k by k' and hence

_______dt______ ______ dt .
K=
I V (i — «s)( i ~ k W )
by
Y ( \ — / 2) ( 1 — k ,2tr)

we have

cn (2K'. k') = — 1, dn (2K'. k ' ) = 1.

T h erefore, form ulas (141) yield

cn(2(K', k) = — 1, dn(2lK'. k ' ) = 1. (142)

If we se t a = t and p=2/Af' in the fir st of e x p ressio n s (96)


and u se form ulas (140) and (142), we obtain

s n ( ( + 21K', k) = sn(t, k ) c n ( 2 IK\ A ) d n ( 2 IK', k ) = s n ( t , k). (143)

From this it follow s that 2iK' is a period of sn (t, k). From what
has been said in the p resen t sectio n , it follow s that the com plex
number

A = 4Km -\-2iK'n, (144)

w here m and n are arbitrary in te g e r s, is a period of sn(t, k).


Let us show that num bers of the form (144) exhaust a ll the
p eriods of sn (t, k). H ere, we can reason as in the ca se of the
le m n isc a te sin e (cf. sectio n 37). Let A denote any period of
sn (t, k). Then,

sn (/ -f- A) = sn t (145)

id en tically. If we s e t <= 0 in (145), we obtain

sn A = 0;

that is , 4 is a zero of the function sn i. T h erefore [cf. (131)],

A = 2mK 2iK'n,
86 THE REMARKABLE S IN E F U N C T IO N S

w here m and n are in teg er s. Let r denote the rem ainder ob­
tained when we divide m by 2:

m = 2P + r t

w here p is an in teger and r is eith er 0 or 1. Then,

A ^ A p K + I n l K ' + 2 rK .

If we su bstitute this ex p ressio n into (145) and note that ApK -\-2nlK'
is a period, we s e e that

s n ( / - f - 2 r K ) — s n /.

This relation cannot be sa tisfie d id en tically if r = 1 sin ce


sn(t ~\- 2K) — s n / . T h erefo re, r = 0 , so that

A — \ pK 2niK',

which was what we w ished to show. m


From what we have proven, it follow s that sn(/, k) is a doubly
p eriod ic function and we can take 4 K and 2iK' as the fundam ental
p eriod s. A nalogously, one can show that cn ( / , k) is a doubly
p eriodic function with fundamental periods 4 K and 2K 2IK'
[another convenient ch oice of the fundamental periods of cn (/, k)
are the numbers 2K — 2iK’ and 2K -(- 2iK'\ and that the function
dn (/, k) is a lso a doubly p eriodic function with fundamental
p eriods 2K and AiK'.
40. Let us rep resen t on the com plex plane a ll the periods
of s i n / and let us draw p a ra llel lin e s through th ese points. T h ese
lin es partition the plane into strip s (se e F ig. 22), known as
p e r i o d s t r i p s . Obviously, th ese strip s fill the en tire plane
without any gaps and without any overlap. The inclination of
th ese lin es can be chosen a rb itrarily. Let us make the sim p lest
ch oice and draw them p a ra llel to the im aginary a x is. If a point
/ is allow ed to vary in such a way as to a ssu m e a ll values in
one of the period s tr ip s , the point t -\-2pn (where p is an integer)
w ill assu m e a ll of the values in another period strip . Since
s i n / a ssu m es the sam e values as / and t -\-2pn, a ll the values
assu m ed by the circu la r sin e in one of th ese strip s are repeated
in any other strip . T h erefo re, for exam p le, when we seek the
roots of the equation

sin/ A. (146)
ZEROS AND POLES 87

w here (a and p real) is a given com plex num ber, we


need only find its roots belonging to any one of th ese s tr ip s . The
other roots are obtained from those found by d isp lacem en ts of
the form t' — t-\~2pn (where p is an in teger). They occupy in
each of the str ip s the sam e p osition s as the origin al roots do
in the origin al strip .

t i U4n

-Zn 0 Zn ‘m pn

F IG . 22.

Let us show that, for any 4 - - - a ~f-/p, equation (146) has


exactly two roots in each period strip . (T hese two points may
m erge into a sin g le point, ca lled a multiple root.) H ere, we
need to adopt a convention a ssig n in g the boundary line between
two strip s to one or the other of th ese str ip s . Thus, let us
a ssig n each such line to the strip lying to the right of it, so that
the points of any strip sa tisfy the in eq u alities

2np ^ o < 2,i(p f I) (— oo < t < -)- o o ) .

For exam p le, the points / — 0 and t — l belong to the strip


0 < o < 2a, and the points 2a and 2 n - f / belong to the strip just
to the right, nam ely, the strip 2a < o < 4a. R ep resen tin g sin/
accordin g to the fir st of form ulas (75), we rew rite equation
(146) in the form

sin <7 • cosh r-f- i cos • sinh t - a -j- i/3. (146')

We w ish to show that th ere e x ist only two roots t a + ix


sa tisfy in g the condition 0 < o < 2a. O bviously, (146') is equiva­
lent to the sy ste m

sincr ■coshr —a, cos a ■ sinhr~ /j. (146^)

Let us s e t sin a = x, s i n h t — y. Then, th ese equations take the


form s x Y 1 -j- y 2 = a and |/1 — x 2y = p or
88 THE REMARKABLE S IN E F U N C T IO N S

jc2 —
(- x ‘2y 2 = a2, y2— je2y2 = p2. (147)

We are seek in g the real solutions of (146"). T h erefore,


0 < a:2< 1 and 0 < y 2. T h ese conditions are sa tisfied only by the
valu es

a J +■ P2 - 1 + V^A _ a 2- f PJ + 1 —
(148)
2 ’ 2

w here

A= (a2 -|- p 2 — l)2 + 4p2 = (a2 + p 2 + i)2— 4 a 2 (149)

and ]/A denotes the p o sitiv e value of the square root.


It is obvious from (149) that the values that we have found
a re nonnegative. To show that x2< 1, con sid er the d ifferen ce
1 — *2:

, J.2-_ / A - ( a » + p»- l ) _

It follow s from (149) that this d ifferen ce is a lso nonnegative.


The equation

y 2 = sinh2 +

y ie ld s two values of t that differ only in sign . C orresponding to


them are two values of opposite sign , sinh t and —sinh t , of
the hyperbolic sin e but only a sin g le value of the hyperbolic
co sin e.
The equation

x2 - sin2 a =
a’ + P’ + l - ^
2

is sa tisfied by two values of sin o. H ow ever, as the fir s t of


equations (146") sh ow s, we should take only the one w hose sign
is the sam e as the sig n of a (sin ce cosh t > 1 > 0). We denote
this value of sin o (which is the only one p o ssib le) by jc0:

sin a = x0, |* 01 < 1.

F urtherm ore, corresponding to this value of sin o are two


values of a in the in terval 0 < a < 2jt, which we denote a, and o2.
ZEROS AND POLES 89

The correspon d in g values of coso, and cosa2 d iffer in sign:


cos a, = — cos a2.
Since the product cos o sinh x m ust be equal to p, r e p la c e ­
m ent of by o2 in the second of equations (146") m ust be a c ­
com panied with the replacem ent of one value of x by its negative
— r. Thus, we have seen that there e x ist exactly two p airs of
values of o and x that sa tisfy (146") or (146) under the additional
condition that 0 < a < 2:t:

0[, x and o2, —x.

The corresponding points t are

q == o, + it, t2 = o2 — lx. (150)

H ere, o, and o2 are determ ined by the conditions

sinOy — ±, .y f —
a2 +
- P2 + 1— V~K

(for j ~ \ , 2), the sig n of sin Oy coin cid es with the sign of a,
0 (jy < 2n, and ± x is determ ined by the condition

,inh„±

We sh a ll not w rite out exp licit form ulas for o and t although
this would not be difficult.
We note that the points f, and t.2 are sy m m etric about .x/2 if
u > 0 and that they are sy m m etric about 3ji/2 if a < 0 .
An analogous a n a ly sis could be made for the circu la r co sin e.
[The sim p le st procedure for doing s o would be to u se the fact
that cos t = sin (n/2 — t ).
Let us ob serv e b riefly how the r e su lts are changed if,
instead of the circu la r sin e, we con sid er the hyperbolic sin e.
H ere, all the p eriods are purely im aginary, except for 0:

0, ± 2it/, ± 4jx/, . . . •

T h erefore, for our period str ip s, we take str ip s lying p a ra llel


to the real axis (see F ig. 23). The identity

sin (it) == i sinh t


90 THE ffEMARKABLE SINE FUNCTIONS

im p lies that the equation


sinh/ = A, t = a-\-ir, 0 < r ^ 2n,

is equivalent to the equation

si n ( / / ) = sin (— x -|- io) =■ si n [ (2a —x) -(- /a] = IA,

w here 0 < o' = 2n — x < 2a, We have already exam ined this equa­
tion, By using the resu lt obtained for s i n t; we s e e that the equa­
tion sinh t = A, for an arbitrary com plex number A, has two and
only two roots in each of the period strip s of the hyperbolic
sin e. (These roots may m erge into a sin g le double root.)

41. Let us indicate on the com plex plane a ll the periods of


the lem n isca te sin e:

2mco ~j- n (to -f- /o>), m and n integers.

In our stu d ies of the periods of the circu la r and hyperbolic


sin e s in the p reced in g sectio n , we partitioned the plane into
period str ip s . H ere, we are d ealing with a doubly p eriodic
function and, instead of period s tr ip s , we naturally have period
p a ra llelo g ra m s. One of the p o ssib le ch o ices of a sy ste m of
period p arallelo g ra m s for si / is shown in Figure 24. As a point
t a ssu m es all the values in one of th ese p a ra llelo g ra m s, the
point t-\-2m uv±n (<■>-f- no), w here m and n are any fixed in te g e r s,
a ssu m e s all the values in another p a ra llelo g ra m . Since

si [/ + 2 ffIM -f-rt(o ) /m) ] = si /.

the values of si t assu m ed in one of the period p a rallelogram s


are repeated in ev ery other p ara llelo g ra m . T h erefo re, for
exam p le, in seek in g the roots of the equation
ZERO S AND POLES 91

sl/ = A. (151)

w here /4 = a -)-/p is a given com plex number, we need only find


a ll the roots belonging to any one of th ese p a ra llelo g ra m s, for
exam p le, the p arallelogram with v e r tic e s 0, 2to, 2co-(- (a>-|- /co),
and All the other roots of equation (151) can be found
from th ese by m eans of d isp lacem en ts of the form

In each p arallelo g ra m , th ese other roots occupy the sam e p o si­


tion as do the roots found in the original p arallelogram .

// // / / /
- p ------- p - - y —
/ / ' x / /
- J ------------- p- -------------- ------r — * — y-
/ / r i- 4u) + ( w + iu/) /
/ v // /
~ P/----- p -/ - - r P/ '-
- p ------------- - * ---------------- - • --------------- p -
/ / */■
/ / / // //
/ /
/-*fw /-2co / 0 / 2 cj ’"/
- p --------------* - --------------- p —
/ / / /j? e / c s / f
/ /
------p ------ -I r
'/ // // a ; b /

FIC. 24.

Let us show that, for any /4 = a-f-/p, there are two and only
two roots in a sin g le p arallelogram (these two may m erge into
a sin g le point, i .e ., a double root). H ere, we again need to adopt
so m e convention regardin g the boundary points that are com m on
to two or m ore p a ra llelo g ra m s. Let us a ssig n to each period
p arallelogram th ose points lyin g on the low er base (other than
the right end-point) and the points on the left sid e (other than the
upper end-point).
In the p ara llelo g ra m s I, II, III, and IV shown in Figure 24,
for exam p le, the point A belongs to I, the point B to II, the
points C and G to III, and the points D and £ to IV.
Returning to the th eorem m entioned above, let us fir s t prove
the follow ing proposition:
The rectan gle 0 < o < 2<d, 0 < t < u (see F ig. 25) a lw ay s con­
tains two and only two ro o ts o f equation (151).
92 THE REMARKABLE SINE FUNCTIONS

The origin al form ulation of our a sse rtio n w ill follow from
this sin ce the function si t a ssu m es, by virtue of its p erio d icity ,
the sam e values in the trian gle OAB outside the p arallelogram
OBEC, as in the trian gle CDE.

B H ___ E
kJl
t+ZCdf

2 uj

F IG . 2 5 .

Let us ex p ress si f — si (a-f-ft) in accordance with form ula


(89) and rep lace (151) with the equivalent sy ste m

si <j cl t (1 + sl2 r)
1 — s l2 a sl2 t
(152)
sl x cl a (I -f sl2 a)
1 — sl2 a s l2 t

Let us ex p ress the values of d o and cl x, r e sp e c tiv e ly , in


term s of sl a and sl t in accordance with form ula (83) and let us
se t s l o —- x and sl t —- y . Then, the sy ste m takes the form

x V 1 —y‘
l — x 2y J
(152')
y
1 — x ' Jy 2

We are seek in g only rea l solutions of this sy ste m that sa tisfy


the conditions j*| < 1 and 0 ^ y < l (sin ce s l xd>0 f o r O < x < w ) .
Let us square each of equations (152'), add the r esu lt, divide
both num erator and denom inator of the sum obtained by 1 — x 2y 2,
and then divide the whole equation through by 1 — x 2y2. This y ield s

x2-\-y': — ( I — x 2y 2), a2y2 (1 — x A) -- p2jc2(1 — y,)= 0. (153)

We note that this equation rem ains unchanged if we sim u ltan ­


eou sly r e v e r se the ro les of x and y on the one hand, and a and p
on the other. E lim in a tin g * 2 from (153), we obtain
ZEROS AND POLES 93

P2y8 - [ (a* + p2)2- l | / - 2 ( 2 a H P 2) / +


(154)
+ l(a2+ p2)2- i]y2y p 2 = o.

If we divide through by y4 and make the substitution

we obtain the follow ing quadratic equation for u:

p V + I (a2 -j- p2) — 1] u — 4a2 = 0. (155)

Since the conditions of the problem adm it only values of y


that are rea l and do not exceed 1 in absolute value, the quantity
u — l / y ’1 —y2 m ust be a nonnegative number. T h erefo re, we ob­
tain from (155) the sin g le a d m issib le value of «:

^ [ ( a - ’ + P2)2- ! ] 2 f 1 6 a ^ - [ ( a 2 + p T - l ) . n (156)
2p2 ^ U

Now, to find y, we have the equation

y4q-uy2 — 1 = 0 . (157)

Here again, only one root, a p o sitiv e one, is p o ssib le for y2.
O bviously, this root is le s s than 2 in absolute value (sin ce the
sum of the two p o ssib le values of y2, which is equal t o — u, is
n onpositive). C onsequently, it w ill not exceed unity:

yV (158)
0 y2

Since y = si t > 0 (where 0 < x < c o ) , this gives us the only


p e r m issib le value of y.
A nalogously, if we se t l/x~ — x2~ v and u se the sy m m etry
(m entioned above) of the sy ste m (153), we obtain

_ yr [ (a2 + P2)2 - i p 4- Hkt-P2 - [ (a2 + P2)2 - 1) ^ (159)


2a2 ^

and hence the only p e r m issib le value of x2:

n .- >-> 1 (160)
2
94 T H E R E M A R K A B L E S IN E FU N C TIO N S

From the value that we have found for y = si t:

„ ^ i f V u 2— 4 — u ^ ,
0^y= y -------2------

we find two values t, and t2 of t in the in terval 0 < t < <o. T h ese
two values are sy m m etric about w/2:

y
„ . r dt , <u . to
0 T, = / . , T, = (0 — T, > -s- .
.! V i — t< 2 2
0

We note that the correspon d in g values of cl t, and clx2are nega­


tiv e s of each other (cf. sectio n 5). For d efin ite n e ss, let us
suppose that y — si t, = si t 2 > 0. (We su g g est that the reader
h im se lf exam ine the c a se y — 0.) Then, from the second equa­
tion in the sy ste m (152), it follow s that the sign of cl o coin cid es
with the sig n of p. T h erefore, for any two p o ssib le valu es a, and
o| such that

i i ' 1 / ~ V v 2+ 4 — v
si o, = s! o , = y ---------- j -----------’

w here o1> 0 , oj > 0 . o, + 01 = 0), we should ch oose that one (which


we assu m e to be o,) such that the sig n s of d o , and p are the
sa m e. A nalogously, of the two values o2 and o2 such that
/- —■■

s i o 2 = s i o2 = —
,
y
1 / y v 24- 4 —v
— 2— ’

w here co -s;:o2, <0 < o 2 and a24- a,' =- 3co, we choose the one (which we
assu m e to be 0-,) such that the sig n s of cl 0., and p are the sa m e.
From what was said in sectio n 5, we can s e e that 0! -f-o.> = 2o>.
(We begin with the fact that 0 < Oj <; w, co < o2 < 2co, sino, = — sino2,
and cl o ,--•=cl o2.)
To su m m a rize, we have shown that the sy ste m (152) can be
sa tisfie d under the conditions 0 < 0 < 2w, 0 < x < w by only two
valu es tj and r2 of t such that

si T, = Si T, =r y > 0, 0 < cl T, = -- cl T., (T, -j- T., =; (0),

and by two values a, and o2 of 0 such that

si o, -- — si a, > 0, cl o, :== cl 0 , (o, -f- 0i — 2d)),


ZEROS AND POLES 95

H ere, the sig n of cl o, (and that of clo2) coin cid es with the sign
P-
Since the fir st equation of the sy stem (152) im p lies that the
sig n of slo • cl x co in cid es with the sign of a, it follow s that the
sig n s of si o and cl x m ust be the sam e if a > 0 and opposite if
ti< 0). This m eans that the value o, is com patible with only one
of the two values x, and x2, nam ely, with -q in the c a se a > 0 and
with t2 in the c a se a > 0. S im ilarly, the value a, is com patible
with t2 or with x,.
What a ll this amounts to is that th ere are two and only two
points q and q in the rectangle 0 < o < 2d), 0 < x < (i) that sa tisfy
equation (151). In each of the p o ssib le c a s e s , nam ely,

+ ( j - o 2+ ix2 — (2co—op-f-ftw— x,) (a > 0),


q -= a, tx2, q —- a., + It, = (2u — o,) + /(o) x2) (a < 0)

the points q and q a re sy m m etric about the cen ter m-)- i'w/2 of
the rectan gle.
If both th ese points do not fall sim u ltan eou sly in the p a r a lle lo ­
gram OBEC, then one of them , le t us sa y , q, belongs to the t r i­
angle OAB and the other, q , belongs to the p arallelogram OBEC
(Fig. 26). But, in this c a s e , the point q = qq-2(i)1 fa lls in OBEC,
and we again have two roots of equation (151) in that p a ra llelo ­
gram . In this c a s e , they a re sy m m etric about the point 2w + no/2.

/ ______ v v
0 C
F IG . 26.

42. An analogous an alysis could be made for the Jacobian


functions with arb itrary m odulus. H ere, we would need to show
that, for arb itrary A, any equation of one of the follow ing form s

sn (q k ) - - - A, cn (/, k) —- A, dn (/, k) = A

has two and only two roots (which may m erge into a sin g le
m ultiple root) in the correspon d in g period p ara llelo g ra m . How­
e v e r , the proof would require som e tedious ca lcu la tio n s, and we
sh a ll not go through it.
96 THE REMARKABLE S IN E F U N C T IO N S

We p refer to conclude our exposition with a b rief su rvey of


the definitions and general p rop erties of ellip tic functions.
T h ese functions constitute a further gen eralization of the func­
tions that we have been studying. At the b a sis of this study is
the concept of an en tire function of a com plex v a riab le.*
A function / ( / ) of a com plex variable is ca lled an en tire
junction if it is defined and differentiable at every point of the
com plex plane. E xam ples of en tire functions are a ll poly­
n om ials, the exponential function (e‘), the circu la r functions sin/
and cos/, etc. It can be shown that the c la s s of en tire functions
coin cid es with the c la s s of functions that can be rep resen ted by
everyw h ere-con v erg en t power s e r ie s :

f ( t ) — a o~\~ a j~ \~ ~ir a n f l +

A function q>(/) of a com plex variable is called a m ero m o r-


phic function (from the Greek words uepou “ fraction ” and
uop<pf| “ form ” ) if it can be rep resen ted a s the quotient of two
en tire functions:

Exam ples of m erom orphic functions a re all rational func­


tions (i.e ., the quotient of two p olyn om ials), the functions
s e c / = 1 / c o s t , and tan / = ( s i n /) /c o s /. It turns out that si/,
cl/, sn (/. k), cn (/, k), and dn (/, k) are a ls o m erom orphic functions.
H ow ever, this a sse r tio n req u ires a sp ecia l proof and does not
follow im m ed iately from the fact that th ese functions can be
ex p ressed by fraction s of the form s (89), (90), (99), (100), and
(101). The d ifficulty lie s in the fact that the num erators and
denom inators of th ese fractions are not en tire functions of the
com plex variable t — a + lx. The proof that the lem n isca te and
Jacobian functions can be rep resen ted as ratios of en tire func­
tions and ways of finding such en tire functions are sp ecia l
questions that would take us beyond the sco p e of the p resen t
book. Of co u rse, an en tire function / ( / ) can be regarded as a
sp e c ia l c a se of a m erom orphic function sin ce / ( / ) = / ( / ) / 1 .
H ow ever, such a function has a finite value at ev ery value of
/ without exception, w hereas a m erom orphic function of the
form <p(/)= g(t)/hjt) has, in g en era l, p oles at which it becom es
in fin ite. O bviously, ev er y point / = /0 at which ft(/0) = o but
g ( t n) 0 is a pole of the m erom orphic fu n ction al/).

*See, for example, the author's E n tire F u n c tio n s, New York, American Elsevier
Publishing Co., 1966.
ZEROS AND POLES 97

A m erom orphic doubly p eriodic function is ca lled an elliptic


function The above im p lies that the lem n isca te and Jacobian
functions are ellip tic .
One of the b a sic th eorem s in the theory of ellip tic functions
a s s e r ts that ev ery nonconstant ellip tic function m ust have at
le a st two p o les (which may m erge into a sin g le m ultiple pole)
in each of its period p a ra llelo g ra m s. From this it fo llow s, in
p articu lar, that no nonconstant en tire function can be ellip tic
(that is , doubly p eriod ic).
We sh a ll refe r to the number of p oles of an e llip tic function
in its period p a rallelogram as the o r d e r of the e llip tic function.
The Jacobian e llip tic functions sn(i, k), cn (t, k), and d n ( i , k) (in
p articu lar, the le m n isc a te functions si t and cl t) that we have
been studying are seco n d -o rd er functions. The r e su lts obtained
by perform ing any of the four rational op eration s, nam ely,
addition, su btraction, m ultiplication, and d iv isio n , on any two
e llip tic functions with the sam e fundamental periods 2w, and
2m.2 y ield s another e llip tic function with the sam e p erio d s. In
this way, we can obtain functions of a rb itrarily high o rd ers. Let
p (p > 2) denote the order of a function / (t). It can be shown that
the equation

f (t)~ A

has exactly p roots in any of the period p a ra llelo g ra m s of the


function / , no m atter what the value of the com plex number A.
(H ere, m ultiple roots are counted in accordance with their
m u ltip licity.)
We s e e that the th eorem s on the number of roots of the equa­
tions

s 1/ —
~ A, cl t = A, sin (t, k ) — A. cn (t, k) — A, dn ( t , k)— A

in a period p a ra llelo g ra m are sp ecia l c a s e s of the o v era ll


prin ciple.
Without stopping for other p rop erties of e llip tic functions,
let us point out in con clusion that ev ery e llip tic function obeys
an algeb raic addition theorem . W eierstra ss proved a theorem
that in a certa in s e n se is the co n v erse of the precedin g one: If
a m e r m o m o r p h i c function <p(t) p o s s e s s e s an a lg e b ra i c addition
th eo r em and is not a rational function e i t h e r o f t o r o f an e x ­
ponential function o f the f o r m eal (where a is any c o m p le x con­
stant), the function cp is n e c e s s a r i l y an ellipt ic function. (See,
for exam p le, the book refe rred to at the beginning of this s e c ­
tion.)
98 THE REMARKABLE S IN E F U N C T IO N S

A ll the gen eral sin e s that we have been studying obey addi­
tion th eorem s. They are eith er ellip tic functions (as is the ca se
with s i t and s n ( / , k)) or rational functions of an exponential
function, as is the c a se with

sin t —e -it
2/

and

sinh t —----------- .
2

T h ese facts can be regarded as illu stra tio n s of W e ie r str a ss’


theorem .
INDEX

A d d itio n t h e o r e m 24 F u n c tio n s o f a c o m p le x v a r ia b le , th e o r y
d e r iv a t io n o f 41 o f 26
a lg e b r a ic 42 F u n d a m en ta l p e r io d , d e f in e d 81
A lg e b r a ic e q u a tio n , f o u r t h - d e g r e e 43
A n a ly tic f u n c t io n s , th e o r y o f 3 0
A r c le n g t h s , c a lc u la t io n o f 11 G o u r s a t , E. 30
A r g u m e n t, c o m p le x v a l u e s o f 49

H y p e r b o lic fu n c t io n s , c o m p a r is o n o f 4
g e o m e t r ic d e f in it io n o f 1
B e r n o u l li ’s l e m n is c a t e 7 r a t io s o f 59

Im a g in a r y a r g u m e n ts 17
C a r t e s ia n c o o r d in a t e s y s t e m 26 I n t e g r a l s , i n v e r s e s of 24
C h o rd , le n g th o f a 25 I n s t it u t io n e s C a lc u li I n t e g r a lis 43
C ir c u la r f u n c t io n s , g e o m e t r ic d e f in it io n I n v e r s e fu n c tio n s 38
of 1
stu d y o f 81
th e o r y o f 6 J a c o b ia n e l l i p t i c fu n c tio n 19, 2 2 , 36
C o m p le x n u m b e r s , a r b it r a r y 47 t h e o r y o f 62
C o m p le x p la n e , in t e g r a t io n in 25 J a c o b ia n fu n c tio n 38
C o s in e , h y p e r b o lic 5
C o s in e g r a p h 11
K in e tic e n e r g i e s , t h e o r e m o f 20

D o u b ly p e r io d ic fu n c tio n s 84 L e g e n d r e ’s n o r m a l fo r m 23
L e m n is c a te , fo c i o f 7
e q u a tio n o f 7
sy m m e tr y of 9
E l lip t ic fu n c tio n , c o n c e p t o f 73 L e m n is c a t e c o s i n e , d e fin e d 10
o r d e r o f 97 L e m n is c a t e f u n c t io n s , g e o m e t r ic d e f in i­
th e o r y o f 30 tio n o f 1
E l lip t ic in t e g r a l 25 L e m n is c a t e s in e 9
f i r s t kind 23 L en g th o f a c h o r d 25
E n tir e fu n c tio n , d e fin e d 96
c o n c e p t o f 96
E u le r ’ s fo r m u la 5 2 , 53 M e r o m o r p h ic fu n c tio n ," d e fin e d 96
m e th o d 4 1 , 43 M u ltip le r o o t , d e fin e d 87

99
too THE REMARKABLE S IN E F U N C T IO N S

P e n d u lu m , d i s p la c e m e n t o f 22 S i n e s , stu d y o f 13
p r o b le m o f 19 c ir c u la r 13
P e r io d o f a fu n c tio n 80 g e n e r a liz e d 19
P e r io d p a r a l le l o g r a m s 90 h y p e r b o lic 5 , 19
P e r io d s t r i p s , d e fin e d 86 l e m n is c a t e 19
P e r io d ic fu n c tio n s 55 S p h e r e , v e l o c i t y o f 20
d e fin e d 80 k in e t ic e n e r g y o f 20
P e r io d i c i t y , d o u b le 73
s im p le 73
U n it e q u ila t e r a l h y p e r b o la 4

R e d u c tio n fo r m u la s 2 , 55
W e i e r s t r a s s ’ th e o r e m 9 7 , 98

S in e a m p litu d e 19
S in e g r a p h 11

You might also like