Você está na página 1de 818

A

TREATISE

THE

ON

THEORY

OF

FUNCTIONS

BESSEL

BY

Gf"N. 'WATSON, Sc.D.,


PROFESSOR

OB'

MATHEMATICS
FELLOW

LATELY

IN
OF

THE

TRINITY

F.R.S.
OF

UNIVERSITY

COLLEGE,

BIRMINGHAM

CAMBRIDGE

'

1"^
CAMBRIDGE
AT

THE

UNIVERSITY
1922

PRESS

Cp

PREFACE

book

THIS

has

functions

of

adapted;

while

they

be

value

of

encounter

such
of

collection

required

in recent

While

which

functions

consequently

endeavoured

also

special,which, although

practicalapplications;and

in
form

appropriate

with

it necessary
Since

and
of

these

the

be

to

myself, it

has

for

reference

part,

of

convenient

been

with

been

have

of

theory

Bessel

far

general

within

or

required

possible,in

as

these

have

aims, combined
has

bounds,

made

intelligibility.

that

than

rather

general theorems,

properties

of

Physics.

so

of the

Analysis by

regard

of

existence

whether

development

Modern

to

Physicistswho

likelyto be

of

book

compatible

most

Course

in the

expounded

is

as

is, for the

book

the

concise

as

the

would

fairlycomplete, I

as

given,

breadth

of

cation
appli-

provided by

which

The

the

interest, are
are

The

size

of

regard

results

purposes.

necessity for keeping

the

as

for

and

all formulae,

theoretical
such

is

of results

researches.

account

an

include

to

without

than

of Mathematical

would

Mathematician

Pure

give

to

for the

scope

of

admirably

are

large order) which

of

problems

theory

series.

collection

functions

been

has

real variable

wider

first is the
the

the greater abstruseness

by

in various

endeavour

my

of their

course

(especiallyof
years

rather

Mathematicians

of

be demanded

to

seems

functions

of

compilation

The
of

processes

of Fourier

theory

in the

functions

functions

Bessel

of

view.

Bessel

purpose

time

increasing number

the

to

Bessel

the

this

functions

in the

object is

fundamental

same

theory of

trigonometrical functions
second

the

offer at

of parts of the

The

For

variables.

complex

objects in

two

applicationsof the

of

development

with

designed

been

Professor

treatise

as

refer the

to

of functions

theory

Whittaker

standard

reader

work

original

to

sources.

It is desirable
as

the

defined

by

by
to

have

Weber
For

felt

of

mathematicians

authority of
kind.
between

The

the

second
the

of

the

two

second
who

kinds

second

function

the

to

kind, namely

sentimental

and

Hankel's

this.

which

kind;
use

and,

in

Weber's
who

my

has

function
use

any

the

functions

and

(iraf

other
use

the

for

two

which

been

kind;

kinds

was

and

pleasing
but

three

standardizing the

weight

function

of Wi'ber's

regarded

(Jubk'r

authority of

greater
one

and

have

second

necessity

opinion,the

parallelismwhich
of Bessel

of the

function

I have

function

the

it would

reasons

first is the

The

mathematicians
is the

here

subsequently by Schlafii, by

used

historical

prevented
the

of

and

justifiedin using

considerations
function

attention

draw

function

canonical

Nielsen.

to

tlie group
than

of the

function

the

second

exhibits

(cosine and

sine)

PREFACE

functions.

igonometrical
is

jrpolation
ich

CO

has

been

previously
book
at

end

of

in

of
been

eclectic

In

the

munber

/
*

value.

given

the

an

vigilance
labours
To

every

complexity,

clerical

mistakes

such

of

to

these

has

my

remove

offer

my

treatise

been

with

Mr
and

to

the

unfailing

grateful

this,

as

have

21,

1922.

the

to

theory
I

equation,

have

seemed

C.

T,

patience,

is

of

probably

useless

undetected;

but

diminished
Preece

obscurities
staff

it

have
the
in

by

and

Mr

to

been

University
a

work

of

great

to

the

criticisms

the

T.

A.

of

the

Lumsden,
greatest
who

Press,

have

typographical

thanks.

G.

August

every
omitted

which

remained

considerably

colleagues

and

such

other,

or

errors

gentlemen

assistance,

those

inserted

slight,

memoirs

this

doubtless

though

Riccati's

of

in

given

consciously

not

been

bibliography

have

however

topic

references
have

The

to

have

work.

scheme.

analytical

mistakes,

profess

XX,

numerical

proofs

possible,

which

Chapter

section

cited.

only

of

by

section,

the

that

as

not

related

inserting

each

sources

device

in

of

inferred

contribution,

the

to

with

do

end

inevitable

mentioned,

are

of

extent

the

at

complete

as

While

the

of

results

the

of

any

original

regard

general

of

no

of

/whose

the

to

case

the

an

III

be

to

made

functions

with

to

that

find

found

and

the

not

in

it.

existence

connexion

which
is

been
in

containing

relevant

it

given

has

Bessel

functions;

the

hope

be

which

memoir

be

book

the

will

memoir

any

those

necessarily

omissions

in

but

enunciated;

are

the

or

the

function

in

give,

to

treatises

is

Tables
Weber's

of

use

policy

my

memoirs

any

the

third

in

possible

make

to

seems

It

made

The

N.

W.

CONTENTS
CHAP.

PAGE

I.

II.

III.

IV.

V.

VI.

VII.

VIII.

IX.

BESSEL

THE

FUNCTIONS

BESSEL

BESSEL

BEFORE

1820

COEFFICIENTS

FUNCTIONS

38

DIFFERENTIAL

EQUATIONS

MISCELLANEOUS

85

PROPERTIES

INTEGRAL

OF

REPRESENTATIONS

ASYMPTOTIC

OF

EXPANSIONS

BESSEL

OF

FUNCTIONS

OF

POLYNOMIALS

FUNCTIONS

BESSEL

132

FUNCTIONS

BESSEL

LARGE

ASSOCIATED

BESSEL

160

FUNCTIONS

194

ORDER

WITH

225

BESSEL

FUNCTIONS

271
.

X.

FUNCTIONS

ASSOCIATED

WITH

BESSEL

FUNCTIONS

308
.

XI.

ADDITION

THEOREMS

358

XII.

DEFINITE

INTEGRALS

373

XIII.

INFINITE

INTEGRALS

383

XIV.

MULTIPLE

XV.

XVI.

THE

INTEGRALS

ZEROS

OF

NEUMANN

450

BESSEL

FUNCTIONS

SERIES

4/

LOMMEL'S

AND

FUNCTIONS

VARIABLES

XVIL

XVIIL

OF

TWO
522

KAPTEYN

SERIES

SERIES

551

FOURIER-BESSEL

OF

AND

DIN

576
.

XIX.

XX.

SCHLOMILCH

THE

OF

TABULATION

OF

TABLES

618

SERIES

BESSEL

BESSEL

FUNCTIONS

LIST

GENERAL

OF

OF

654

665

753

BIBLIOGRAPHY

INDEX

FUNCTIONS

789

SYMBOLS

AUTHORS

INDEX

QUOTED

791

706

CORRIGENDA

^Page

62,

line

for

11,

-^^{hzf-K'

"

"

(i^)^'""

yectc^

"

'^

m=0

"^
5/1=0

"

"

Page

91,

line

the

from

foot

the

of

for

Brassiue"

"

read

Brassinne."

"

page,

Page

Page

228,

327,

in

line

formula

(2),

from

the

/or

"

foot

sec^^"

98720

of

the

read

/or
page,

"

"

Bruhns

.sec^

78720

"

read

^."

"

Bruns."

CHAPTER

BESSEL

Riccati's

I'l.
The
a

In

of

which

is derived

The
in

fact

Bessel

paper*

that

he

has

In

curves

Bernoulli

refers to

Analysis of

states,

1697):

Vellem

"

in mille

Ego

petuo

equation

of the

occasione

recordor

quam

John

gives in

he

num

banc

years

order

second

wrote"

and

aequationes alias
sed

discoveryhad

construi

See

Leihnizens

" Ibid.

p. 65.

occurs

In

1694.

type and

states

reducing

to

Leibniz

dy

x-

"

dy

(Jan. 27,

yydx

xxdx.

Problema

the

perlinear

to a

yydx

sicut

reducendo

was

435

"

xHx

tentavi"

x'-dx in qua

aequatio

aequationem

simple step
of the

nun-

maneret

ad

banc

to

solve

equation

of

the

last

the

first

437.
+

y'^dx (i-dy
=

quae

an

separationem

per

(p.436).

Folge (Matliematik), iii.


to
effectively,

was,

writes

dx"."

made, it

Dritte

he

equation
(Nov. 15, 1702): "Qua

invicem,

se

possit noudum

procedure

1704, James

form

improbum

mefvm

separavi illas

been

gesamellte Werke,
Bernoulli's

type
in

and

tentaveris

memoratae

indeterminatas
:

the

in

later he succeeded

puhlicata Lipsiae, 1694, pp.


proposita aequatio differentialis haec

indeterminatarum

of this

it. Thus

solve

et

Eruditorum

Acta

t "Esto

Riccati's

1697

between

equation in series,and so to obtain the solution


order as the quotient of two
power-series.
*

equation)

transformation.

Bernoulli

equation

an

inabilityto

differentio-differentialem||
ddy ".y
this

Bessel's

as

yydx-'rxxdx,

"

scire

Te

ex

simpliciterdifferentialis

When

(known

Riccati's

particularsolution

equation of

transmutavi, sed operam

potui

separare

his

once,

porro

lusit." Five

as

elementary

an

an

example,

an

equation,which

the

formas

as

theory of

itf.

than

more

first order, known

published by

dy
and

the

order

to
Leibniz, written
letters:J:

various

with

usuallydefined

the second

was

gives,as

solved

not

the

of

equation by

in

which

Bernoulli

is

function

Riccati's

from

on

1826

intimatelyconnected

equation

earliest appearance

this paper

BEFOUE

is

equation of

linear differential

functions

Bessel

of differential

type

equation.

FUNCTIONS

differential
equation.

theory of

certain

take

new

(Halle, 1855), pp.


variable

50"87.

defined

by

the

formula
1

du

dx

~-^
_

iu the

equation

IIThe
in

d!/ldx

connexion

x^ +

between

y", and

then

this

equation

to

replace u by
and

y.

special form

of

Bessel's

equation

will

be

"4-3.
W.

^
B.

F.

seen

THEORY

OF

And, in fact,this form


Bernoulli

James

"Reduce
terminus

within

BESSEL

of the

solution

communicated

was

(Oct.3, 1703)

year

in the

aequationemdy=yydx-\-xxdx ad
seriem exprimitur,
ita

per

^~

to

Leibniz

by

terms*:
following
fractionem

autem

cujusuterque

3.4.7

[CHAP. I

FUNCTIONS

3. 4

3.4.7.8.11

7
.

8
.

1]

12

pff.

15

8. 11

12

15
.

]6. 19

X^-

X^

4^

3.4

3.4.7.8.11.12

divisione

series

quidem actuali
tam
non
progressionis

quae
ratio

^
Of

"^

"^

37377

in

pf,o

3.4.7.8.11.12.15.16

conflari possunt, sed

unam

in qua

facile patescat,scil.
"^

37373737 57777Tn

3.3.3.7.11

that time, mathematicians

at

course,

.*"!"
4^

"

3.4.7.8

their energy, so far


terms,
obtainingsolutions in finite
concentrated

differential

equationswere concerned, on
Bernoulli seems
and consequently
James
to have received hardlythe full credit
entitled him.
his discovery
to which
Thus, twenty-two years later,the paper f,
in which Count Riccati first referred to an equationof the type which now
followed by a note;]:
bears his name,
was
by Daniel Bernoulli in which it was
stated that the solution of the equation^
as

aa-"
was

unsolved

hitherto

dx + uudx
The

problem.

of the solution :
anagram
occultis involuta
characteribus

"

an

hdu

ended

note

with

in

announcement

an

Solutio

problematisab 111.Riccato proposito


24a, 66, 6c, 8rf,33e, bf,2g, 4"h,SSi, 61, 21in,

2Qn, 16o, Sp, hq, l7r, 16s, 2U, 32w, ^x, 3y, +,

-,

",

=,

4, 2, 1."

to have been solved ; but Bernoulli


published
anagram
appears never
of the anagram.
his solution 11of the problem about a year after the publication
of a set of values of n, namely
The solution consists of the determination

The

"

4!m/(2m"1),

where

soluble in finite terms;


The
with
*

of which the
is any integei',
for any one
will
be given in
the details of this solution

the fact that Riccati's

Suppl.

viii.

X Ibid.
other

66

x'"''dqdu

xiudx:q,

"

73.

The

form

in

which

Kiccati

took

tlie

.r".

pp. 73

"

75.

of his

members

" The

(1724),pp.
=

of

equationwas

gesamellte Werke, Dritte Folge (Mathematik),iii. (Halle,1855), p. 75.

See Lcihnizens

f Acta Eniditorum,
equationwas
where

by Daniel Bernoulli,combined
more
generaltype than
slightly

of Riccati

the work

prominencegivento

equationis
""4-1,4*11.

reader

Daniel

family

should

Bernoulli

mentioned

John, Nicholas

"

observe

and

that

solutions

the younger

had

been

obtained

by

three

Nicholas.

that the substitution


b dz
U=--rz

givesrise

to

an

equation which

IIExercitationes
pp. 465"473.

quaedam

is

dx

easilysoluble in series.
(Venice,1724),pp.

mathematicae

77

"

80;

Acta

Eruditonim,

1725,

1*2]
John

BESSEL

Bernoulli's

equation*has

stillmore
It is

generaltype of

equationof

any

where

P, Q, M

It is

the

The

of Riccati

without

beingassociated
but also with
solving,

Riccati's

generalised
equationto

name

he discussed

give the namef

the form

given functions

are

supposed that

if P=0,

equation.

to

customary

now

1826

BEFORE

resulted in the

only with the equationwhich

not
a

FUNCTIONS

neither

nor

equation is reducible

of
is

x.

If B=0, the equation is linear;


zero.
identically
form by taking 1/?/as a new
varialjle.

to the hnear

last

it is reducible to the general


equationwas studied by Euler;]:;
of
the
second
order,and this equationis sometimes reducible
equation
Bessel's equationby an elementarytransformation (cf.
""3"1, 4"3,4'31).

linear
to

Mention

should

be made

here

of two

memoirs

Euler.

by

In the

it
first"
is proved that, when a particular
integral
t/iof Riccati's generalised
equation
is known, the equationis reducible to a linear equationof the firstorder by
replacingy hy i/i+ l/u,and so the generalsolution can be effected by two
solutions are
quadratures.It is also shewn (ibid.p. 59) that,if two particular
known, the equationcan be integrated
by a single
completely
quadrature;and
this result is also to be found
of these theorems

1'2. Daniel

Bernoulli

of theorems

number

\villbe

givenin

Bernoulli's rnechanical

In 1738 Daniel
a

in the second

on

||of the

Chapter

two

papers.

brief discussion

iv.

lirohleni.
memoir

publisheda

IT containing
enunciations
**

heavy chains. The eighth


oscillantis. Sit catena
uniforniiter

the oscillations of

of
of

AG
Jiguracatenae
uniformiter gravis
flexilissuspensa de puncto J.,eaque oscillationes
et perfecte
in situm AMF\
facere uniformes
catena
fueritque
intelligatur:
perv'enerit
FM
sumatur
^:
n
ejus
x,
longitudocujuscunquepartis
longitudocatenae
fit
ut
valoris+f
these is

follows:

as

De

"

*-

See

James

-I

4/i/i

Bernoulli, Opera Omnia,

ii.

t The

the solution
term

'

by

Kiccati's

Daniel

equation

(1763),[published 1770], p.

XIX.

t Institutiones
the

reduction, see
Comvi.

IIIbid.

IX.

"Theoremata

suspeusae," Comm.
**

Luc.

t+ The

James

'

was

is

of

used

solution

1054"1057

by D'Alembert, Hist, de

V Acad.

Sci.

solution

the

which

Ii. des Sci. de Berlin,

242.

ii. (Petersburg,
1769), " 831, pp. 88"89.
Iiiteiiraliii,
Bernoulli's letter to Leibniz already quoted.

oscillationibus

Acad.

it is stated that
and

given.

Acad.

de

in tinite terms,

Petrop. viii. (1760"1761),[published1703], p.


(1762"1763),[published1764], pp. 163"164.

" Novi

Calculi

Bernoulli

0.

4.9.1G.25/r'

(Geneva, 1744),pp.

point of Eiccati's problem is the determination


resembles

V etc.

"

4.9.16n^

4.9n3

coriwrum

file flexili

Imp. Petrop. vi. (1732"3),

connexorura

lu connexion

82.

et

catenae

verticaliter

[published1738], pp. 108"122.

cit. p. 116.

length of the simple equivalent pendulum is

with

ii.

1"2

distantiam

OC

calculo
brevissimo
say: "Invenitur
alios."
littera n infinitos valores

autem

last series is

The

described

now

as

infinite number

of

?i

of order

and

zero

this function

that

I....

proxime 0'691

function

Bessel

and the last quotationstates


'\/(x}n);

argument

"

4.9.16.25/1^

4.9.16n'

4.9?i3

4)in

x^

x'^

to

on

goes

x^

XX

~w

Habet

puncti F ah linea vertical! 1, dico fore


punctiubicunque assumptiM ab eadem linea verticali aequalem
extremi

Ponatiir porro distantia

He

[CHAP.

FUNCTIONS

BESSEL

OF

THEORY

has

an

zeros.

f proofsof his theorems soon afterwards; in theorem


published
by consideringthe forces actingon
VIII, he obtained the equationof motion
also obtained by
of length x. The equation of motion
the portionFM
was
Eulerjmany years later from a consideration of the forces actingon an element
Bernoulli

of the chain.
the substance

followingis

The

densityof

Let p be the line

height

above

of

chain

horizontal

The

and

(supposed uniform)
in its undisturbed

let T

be the tension

position. The

the

obtain

for
equation 8T=gp8x by resolvingvertically
integralof the equationis T=gp,v.

The

length8x.

the chain

:
investigation

point of the chain

the lowest

transversal,we

of Euler's

of the tension

component

; and

displacementof the element

is,e"ectiYQly,'
T{dyldx) where
the

so

motion

equation of motion

being

element

an

at

y is the

of

zontal)
(hori-

is

p"^S="^^)
If

substitute

we

for T and

proceed to

the

df^
If

lengthof

is the

the

find that

limit,we

dx\

dx)

'

simple equivalentpendulum

for any

normal

one

we
vibration,

write

y=-.n(?)si"(f"^?),
where

If

and

xlf=u,

constants

are

obtain

we

and

the solution

^=

,
1

tions

On

the

Continent, the functions

t Covim.
+

Acta

took

the

distance
notation
p and

{x\f)is a

dv\

dx

dx)

in the form

"

^i

".

"

1.4

."r.

solution

of the

equation

f
of Bernoulli's

series,namely

^"

1.4.9

""""

1.4.9.16

usuallycalled cijlinder
functions,or, occasionally,
/"ncfur Math. lxix.
(1868),p. 128; see also Math. Ann.

are

after Heine, Journal


of Fourier -Besisel,

(1871),pp.

111.

then

609"610.

Acad.
Acad.

weight

Petrop. vii.
Petrop.
of

(nottwice
lias been

(for d).

v.

(1734"5), [published 1740], pp. 162"179.


1 (Mathematiea),(1781),[published
pars
1784],pp. 157"177.

length e
the

of the

chain

distance)fallen by

to
a

be E, and

particlefrom

followed in the text apart from

the

he
rest

defined
under

to

be

the

gravity in

significanceof

and

the

measure

second.

Euler
of

the

Euler's

introduction

of

1*3]

BESSEL

FUNCTIONS

1826

BEFORE

"t

where
,

constants.

are

Since

If a is the whole

finite

is

lengthof

when

j;=0, C must

the chain,y

0 when

a?

zero.

a, and

the equation to determine

so

/'is

"fi
+

\7f^T:\p 1.4.

be

"

C and

=
"

"

0-

"

9r'

By an extremely ingeniousanalysis,which will be given fullyin Chapter xv, Euler


proceeded to shew that the three smallest roots of the equation in a// are 1-445795,7-6658
and 18-63.
values are
[More accurate
1-4457965, 7-6178156 and 18-7217517.]
In

the

solution
law

memoir*

(inthe

Euler
immediately followingthis investigation
of series)of the

form

of formation

of successive
stated

however, been

equation ^

coefhcients

\\i

obtained

but his
\-\-v=-0^

"

the

general
of the

statement

is rather

in his Institutiones

incomplete. The law of formation had,


ir. (Petersburg,
Integralis\,
1769), i^977,

Calculi

pp. 233-235.

I'S. Euler

mechanical

vibrations

The
1764.

of

arrived at the

He

jjvohlem.

stretched

membrane

investigated
by

were

Euler

in
;J:

equation
1 d'z
"ldP^_drz Idz
?~df-~d?'^r(h''^
r-d(l"'

where

is the

coordinates

and
(r, "/");

are

obtain

To

where

a,

A, ^, B

of

dr-

solution of this

11

u-r

This
of order

differential

/3; and ^
for

,7:nye2-(

equationis now

may

have ||any

Acad.

Petroj).v. pars
t See also "" 935, 93G (p. 187
+

in

Novi

" The
IIIf
in view

function

(d-

/3-\

\e-

r-

of r;

and

the result of

equation

is finite at the

originis given on

p. 256

'

2
.

constant

argument

known

(".+

1 )(n +

3) e^

'

'

1.
as

Besscl's

equationfor

of the values 0, 1, 2,
factor

"

the series is

ar\e. The

functions

....

now

called

Bessel

of a
periodsof vibration,2-7r/a,

(Mathematica),(1781),[published1784], pp. 178"190.


which
associated eiiuation
of an
et seq.)for the solution

will be

" 3-52.

Coinm.
reason

/3 were

dr

coefficient of order ^ and


Acta

Idu

equationwhich

omitted

an

is

is the differential

written^in placeof 2/3+

has been

Save

and

+ B),
(/5(/)

it is

of Euler's memoir;

sin

(at+ A)

constants

are

d-a

The

sin

of this value

substitution

discussed

depending on

constant

point whose polar


the densityand

solution he wrote

normal

is

the

t at

of the membrane.

tension

where

time

displacementat

transverse

Petrojj.x. (17G4),[published17t)6J,pp. -243" -200.


why Euler made this change of notation is not obvious.

not

Acad.

an

integer,the displacemeut would

of the factor sin

(/30+ B).

not

be

one-valued

function

of

position,

THEORY

circular membrane
from

This
a

of radius

the consideration

BESSEL

OF

that

with

fixed

vanishes

[CHAP. I

FUNCTIONS

boundary* are

when

to be determined

a.

investigation
by Euler contains the earliest appearance
coefficient of generalintegral
order.

Bessel

1'4. The researches

ofLagrange,Carlini

and

in

Analysisof

Laplace.

Only a

few years after Euler had arrived at the generalBessel coefficient


in his researches on
in an
vibratingmembranes, the functions reappeared,
astronomical
motion

of

of the

problem. It was
planetabout the

inverse

anomaly

at

sun

the relations between

square,
the eccentric

and

M^E

giverise to

by Lagrangef in 1770 that,in the elliptic


the focus attracting
accordingto the law

shewn

the

-"iimE,

(6

'^'*
"

vector

r, the

mean

the forms

assume

a{l-"cosE),

y.

An"inuM,

and

^m=o
2''+''^m\{n+

jB"cos"ilf,
orbit, and

of the
eccentricity

the

and

"'""'"'

my.'

S
w=l

semi-majoraxis

the

are

i6^+

l +

n=\

radius

expansions

E^M^
in which

the

anomaly E, which

^^^^^

2"+^"" m\

(n + m)l

for w
1, 2, 3. The objectof the expansions
Lagrange gave these expressions
in
is to obtain expressions
for the eccentric anomaly and the radius vector
=

of the time.

terms

In modern

^"
It

noted

was

these formulae

notation

2/" (ne)/n,5,

by Poisson,Connaissance

memoir

Poisson

would
account

by Lefort,Journal
should
corrected,

Cf.

Journal,

xi.

(e/n)J^ (ne).
[published
1833],p.

6 that

ae

(1846),pp.

142

152, in which

"

an

made

error

by

also be consulted.

of the approximatevalue of A.^ when n is large


investigation
1 is due to Carlini :|:;
is not rigorous
"
"
(and it
though the analysis
be difficult to make
it rigorous)
it is of sufficient interest for a brief
of it to be given here.
e

Bourget, Ann.
xxi.

Sci.

de

VEcoIe

norm.

siij}.in.

(1866), pp. 55

"

95, and

Chree, Quarterly/

(1886), p. 298.

t Hist, de VAcad.
III.

de Math.

remarkable

and

is

des Terns,1836

written

are

R. des Sci. de

Berlin, xxv.

(1769),[published1771], pp. 204"233.

[Oeuvres,.

(1869),pp. 113"138.]
X Riceixhe

siiUa

(Milan, 1817).
col. 197"254

reprinted in

This

courergenza

work

was

delta

serie

translated

che

into

serva

German

alia

soliizione

by Jacobi,

{Werke, vii. (1891),pp. 189"245]. See also two


Math.
Ann, xvii. (1880),pp. 531"544, 545"560.

papers

del

problema

Astr.

by

Nach.

Scheibner

di
xxx.

Keplero

(1850),

dated 1856,

14]

BESSEL

It is easy to shew

Define

Hence

when

on
we

is

''^u*i
then
J"=2?i"-ie-^"''7"!

largeeither

should

(?""')
we

m^

or

w^ and

expect

du/de must

or

dujde

consideringthe highestpowers of n in the


find that
1. It is consequentlyassumed
a

of

powers

where

this
substituting

u^^ duJdf

that

and
0(?j2o)

and
("") respectively;

of the last differential

admits

v.yjn
+

Ui +

of

an

expansionin

equation,

descending

powers

of

-'^

??o= "

".

series in the differential

that

; so

du^, +

independent of

are

...

the coefficients of the various

zero

be

to

large.

in the form

Wq, "i, u^.,

On

be

various terms

where

A^ is a solution of the differentialequation

the formula

1826

BEFORE

??,

?"i

equationof

we

t^
1

the first order and

equatingto

find that

"

If t"=0

by

that

FUNCTIONS

"

:,,
f-

and

therefore

["c7.
M|log-p^-^^,^"^/(l-.^')+l}-i
+

and, since
be taken
From

the value
and

of

An shews
of

constant

no

formula
Stirling's

that

^iide n log^ewhen

is to
integration

it

now

follows

at

6"exp
An

is

...,

small,the

upper

sign must

be added.
once

that

{nj{\-e^)}

v/(U)."t(l-e2)i{l+V(l-e2)}"'
this is the result obtained

and

further

much

formulae
which

for

by

A^

Carlini's

Aleissel

in the

case

by

This method

Carlini.

of

has
approximation

been

carried

(see " 8-11), while Cauchy* has also discussed approximate


of comets
moving in nearlyparabolicorbits (see " 8'42),for

approximation is obviouslyinadequate.

has justbeen givenis much


of which
account
an
investigation
than the arguments employed by Laplacefto establish the
plausible
approximationfor i?".
The

more
sponding
corre-

which
and the method
investigation
givenby Laplaceis quiterigorous
the value of B^ is modified by
is of considerable importancewhen
he uses
by
taking all the coefficients in the series to be positive or, alternatively,
that
an
supposingthat e is a pure imaginary.But Laplace goes on to argue
be
variables
may
approximationestablished in the case of purelyimaginary
crainte in the case of real variables. To anyone who is acquainted
used
sans
The

"

'

'

with

reasoningwill
*

486"489.

fallacious character of such

be evident.

Comptes Rendus,

t Mecanique
pp.

theoryof asymptoticseries,the

the modem

xxxvni.

(1854),pp. 990"

Celeste, supplement,

t.

v.

9;i3.

[firstpublished 1827].

Oeuvres,

v.

(raris.1882),

OF

THEORY

[CHAP.

FUNCTIONS

BESSEL

is based on the principle


portionof Laplace'sinvestigation
increase
in which the terms
terms
steadily
that, in the ease of a series of positive
decrease,the order of magnitude
up to a certain pointand then steadily
be obtained from a consideration of
of the sum
of the series may frequently
the order of magnitude of the greatest term
of the series.
The

earlier

of this principle,
Stokes, Proc. Camb. Phil.
see
applications
and
VI. (1889),pp. 362"366
[Math,
Phys.Papers,v. (1905),
pp. 221"225], and Hardy,
Proc. London
Math. Soc. (2)ii. (1905),pp. 332"339;
(1905),pp. 97"101.
Messenger,xxxiv.
A statement
of the principle
was
given by Borel,Acta Mathematica, xx. (1897),pp. 393
For

other and

recent

more

Soc.

"

394.

The

the

of the principle
appliedto
following
exposition
b}^Laplacemay not be without interest :
The

.seriesconsidered

is
+

J,

is

/x, where

largeand

fi is the

has

+ 2m)?i"-^-'"-^f"
I (?;,

~^,"=o2"

"

in which

fixed

fj.is

so

Now,
that, to

-'"

'

+ "i)!
=i"""i!("

positivevalue.

greatestintegersuch

The

greatest term

is that for which

that

4^ (n + ,x){n + 2^
and

example considered

2) ^ (n + 2n)nh%

approximatelyequal to

if ?"", denotes

the

general term

in

it is easy
II,S^\

theorem
verifyby Stirling's

to

first approximation,-^^'r^qi-,
where

'

log2=-2V(l4-.2)/(ne2).
Hence

+ 2q + 2q^+ 2q^+ ...}


B"^'^)r^Uf,{l

~2"^^/W(l-g)},
.

since*

nearly equal

is

to

1.

Now, by Stirling's
theorem,

{ns/(H-62)}

e"-i exp

and

e"exp{nv-(l+.^)}
|MI"li)l^

5"(i)"^

so

The

inference

which

Laplacedrew

V
This

Trn'

approximateformula happens to

for this restriction is not

from

this result is that

{l + V(l-e-)}"

be valid when

apparent, apart from

but it is difficult to
necessary),
prove it without
"

of.

The

formula

l +

^^^f~s'N{^-q)}

Bromwich,

formula

in

may

Theory of Infinite Series," .51.


the theory of elliptic
functions,

be

inferred

It is also

the

"

Modern

Analysis,%21-5l.

(though the

fact that

it is

using the methods


from

consequence

of Jacobi's

reason

obviously
of contour

general theorems

^3(0|r)=(-;r)-H3(0|-T-i);
see

"

on

series;

transformation

1-5]

FUNCTIONS

BESSEL

1826

BEFORE

to have been dubious as to the validity


(cf." 8*31 ). Laplaceseems
integration
about
real and
of his inference because, immediately after his statement
of
he
that
he had
mentioned, by way
confirmation,
imaginary variables,
another proof;but the latter proofdoes not appear to be extant.

of Fourier.

researches

1'5. The

by Fourier*, La

classical treatise

appearedthe

In 1822

Bessel

de la Chaleur; in this work

functions

of order

Theorie

zero

analytique
in the discussion

occur

motion of heat in a solid circular cylinder.It is


symmetrical
120) that the temperature v, at time t,at distance
by Fo-urier (""118
of the

shewn
X

"

from

satisfies the equation


cylinder,

the axis of the

fd'-v

1 dv

CD

Kda--

xdx

dv
dt

K, C, D

where

Density of

and

Heat
the Thermal
Conductivity,
Specific
respectively
solution
and he obtained the
material of the cylinder;

denote
the

^2^4-

2-

""

2^4^6,

where

and

mCDjK

has to be

at

the

boundary of

the

chosen

so

hv + K

that
0

{dvldx)
=

Conductivity.

h is the External

where
cylinder,

that
309) by Rolle's theorem
proceededto give a proof (""307
and
of
real
roots
the values of m hasf an infinity
the equationto determine
that
no
incompletebecause he assumes
complex roots. His proof is slightly
true of integral
which have been proved for polynomials
are
certain theorems
by Hurwitz^
functions; the defect is not difficult to remedy, and a memoir
has the objectof making Fourier's demonstration
quiterigorous.
Fourier

"

formula

("313)

for the

"

The

discussed

will be

formula

of order

and

zero

in

7i

"

"

-.

"

"

"

COS

ttJo

...

2-. 4^6-

2". 4-

(a

sm

x) dx,

it is a special
case
years earlier by Parseval";
Bessel's and Poisson's integrals
(""2'2,2"3).

of what

as

depositedin

the archives

greater part of Fourier's

of the

French

found

in the

Institute

on

researches

Sept. 28, 1811,


des

VAcad.

de

Mem.

its

"" 5-6, 9'6o.

proved some
known

now

are

-\-7^

"

TT

2-

been

function

Bessel

fraction

givenby Fourier, namely

formula
1

had

quotientof

of this
generalisations

derivate ;
Another

"thatFourier discovered the continued

also be mentioned

It should

Sci.,

contained

was

crowned

and

iv.

on

in

memoir

Jan.

This

6, 1812.

(1819),[pubHshed 1824],

pp.

185"

is to be

memoir

555;

v.

(1820),

pp. 153"246.

[published1826],
of
t This is a generalisation
+

Math.

" Mem.
statement

another
a

Ann.
des

of the
paper

by

etranijers,

theorem

on

of

i.

Fourier

this little known

general solution

quoted in "

statement

1-2.

(1889), pp. 246"266.

xxxiii.
savans

Bernoulli's

(1805),pp.

639"648.

This

which

is sometimes

constants

des

writer, Mem.

Laplace's equation

in

savans

form

also

paper

etrangers,

called
i.

contains
Parseval's

the

(1805),pp. 379"398,

involving arbitrary functions.

formal

theorem
tains
con-

10

[CHAP.

FUNCTIONS

BESSEL

OF

THEORY

function into a series of Bessel functions of


arbitrary
also examined
order zero was
by Fourier (""314"320); he gave the formula
for the generalcoefficient in the expansionas a definite integral.

expansionof

The

much

examined
validityof Fourier's expansion was
Math.
Schlafli,
471"494;
viii.
(1875),pp.

The
Math.

an

Ann.

Ann.

246"269

Serie di Fourier, i. (Pisa, 1880), pp.

1'6. The

in

be dealt with

expansionwill

This

researches

Math.

Hobson, Proc. London

Math.

Young, Proc. London

and

(1909),pp. 359"388;

recently by Hankel,

more

Diui,

(1876),pp. 137"142;

x.

Soc. (2) xviii.

Soc. (2) vii.

(1920),pp.

163"200.

Chapter xviir.

of Poisson.

unsymmetricalmotions of heat in a solid sphere and also in a


were
investigated
by Poisson* in a lengthymemoir publishedin
cylinder
In the problem of the spheref, he obtained the equation
The

where

the distance

denotes

and
integer(zeroincluded),

mode, which
a

is

function

solution of the

the centre, /? is a constant, n is a


is that factor of the temperature, in a

from
R

of the radius

QQg

Jo
he discussed

(" 3'3) that


order

by

positive
normal

Poisson

that

the

the

cases

definite

QQg
jy^sin^"^+^ft) dw
(^y,p

0, 1, 2 in detail.

integralis

(save for

It will appear

factor)a

subsequently

Bessel

function

of

^.

shewn

was

1823.

equationis
j.n+i I

and

It

vector.

solid

In the

problemof

{ibid.
cylinder
p. 340

the

X"

et

is
the analogousintegral
seq.)

{h\ cos (o)sin-^codto,

cos

.'o

where

n=0,

1, 2,

is now
integral
In the

and

...

known

case

large;the
Let
Then

Jq (k)

J^ {k)is a

Journal

de

cylinder.The

integral
("2-3).

0, an

obtained

cos

I'Ecole R.

di Torino,

xxv.

whom

mentioned

{kcos

"

w) da,

J^ Ik)

cos

sin (k

last

integral

the variable

w) da.

cos

solution of the equation

Pohjtechnique,xii. (cahier19), (1823),pp.

t Ibid. p. 300 etseq. The

are

Poisson's

as

the axis of the

from

importantapproximateformula for the


was
352) when
by Poisson (ibid.,
pp. 350
i
s
the
substance
his
of
following
investigation:
n

and its derivate


is

is the distance

(1821),pp.

1835),pp. 366, 369.


+
See also Eohrs,
not used by Poisson.

equation

532

in " 4-3.

"

was

534, and
See

Proc. London

also

Math.

also studied
has

since

Poisson,
Soc.

v.

La

by Plana, Mem.

been

studied

Theorie

(1874),pp.

by

249"403.
della R.
numerous

Mathematiqite
136"137.

Tie

Accad.

delle Sci.

writers,some
de la Chaleur

notation

of

(Paris,

Jg{k) was

12
and

so,

by equating to

the various

zero

[CHAP.

FUNCTIONS

BESSEL

OF

THEORY

find that
we
coefficients,

A'^-Ib,
A'-=-,-^,A.i^B,...
A":

and

hence

expansion of

the

/ttX^
r/

/"tt

integralis

Poisson's

cos
(/"
0,)cfo,

cos

9.25

+
2.^:W3+-r^'^'
^^j LVl-8|-278^^
9.25

9^

\_

right are not convergent, the researches


of
subsequent writers are a necessary preliminaryto the investigation
the latter portionof Poisson's investigation.
But, since the series

It should

be mentioned
W.

given by
expressedthus
first

was
was

on

r
-

E.

the

that

Hamilton, Trans.

for the

Irish Acad.

R.

generalterm
xrx.

of
the

and
Lijischitz
of
significance

in the

(1843),p. 313;

-^

1 [0]" ([ \ff (4^)-" cos (2/3 \nn


-

expansion
his result

\n\

the expansion as semi-convergent; the -expressions


[0]~"and
as
Ijn ! and {-\) { %)"".{-n + \).
interpreted

he described

and

to be

(2/3sin a) da=

cos

formula
explicit

an

jsin

[-i]"

are

"

A result of

some

memoir*, is that

which
importance,
the generalsolution

and

It follows at

by Poisson
equation

of the

Jo

where

obtained

was

2/

that the

once

generalsolution

d-y

dy

dx-

dx

e-''^cos"^(y ^. ^

.'o
This

result

subsequent

constants.

are

Jo

7o

of the

equation

.^
_

"

is

in

g-7wcos"o iQg(^ sij^2^) ^^

quoted by Stokesf as a known theorem in 1850, and it is


that he derived his knowledge of it from the integral
likely
givenin Poisson's
but
the
fact that the integral
memoir;
is substantially
due to Poisson has
been sometimes
overlooked|.
*

Journal

de

was

I'EcoIe

general integral of

R.

Polytechnique, xii. (cahier 19), (1823),p. 476.


partialdifferential equation was
given in

associated

an

The
an

correspondiug
earlier

ibid. p. 227.

t Camb.
J See

Phil.

Trans,

Encyclopedic

ix.

(1856),p. [38],[Math,

des Sci. Math.

u.

28

and

("53),p.

Phys. Papers, iii. (1901),p. 42].

213.

memoir,

r7]

BESSEL

The

1'7.

bear

his

that

the

which

in

Bessel

written

was

name

in

Ba

expression
In

memoir

the

defined

be

be

to

in

function

is most

function

which

is of

function
After

that

it

and

Berliner

Jan.

29,
t

formula

the

"

1'4.

function

Ij/^

Atti

1824

this

in

1894,
R.

[published

Accad.

1826],

Storungen,

the

from

and

266;

"

dei

pp.

Lincei,

1
"

aus

date
der

of

defining
" lO'l)

(see
h

will

is

the
the

Ij/^ but

not

Chapter

be

III.

so

numerous

chronological

account

order.

Fourier
briefer

for

became

the

to

(3)

of

this

iv.

the

(1880),

memoir,
der

Bewegung

been

ha"

1858

account

{Transunti),

The

52.

welcher

of

in

logical
of

time

204

integer
values

functions

the

which

adapted

an

abandon

to

results

studied

be

on

the

is not

is not

will

stage

pp.

of

non-integral

systematic

researches

planetarischen

which

at

della

for

u) du.

"

com|)iled

early history
pp.

259

263.
"

Untersuchung

Sonne

entsteht,"'

is

1824.

Berliner

J This

in

given

systematically

integral

investigations

Mittheilungen,

Ahh.

der

series

the

many

when

study

Bessel

of

by Maggi,

Theils

by
of

Bessel's

ii.

Lommel

account

given

most

theory

Beni

des

now

shewn

is

u) du

sin

k sin

"

obtained

and

interest

the

Wagner,

was

with

(hu

cos

worth

convenient

seems

historical

An

by

time

develop

to

Chapter

defined

the

motion

had

r-'"

integer

an

which

the

memoiri*

which

he

integral :|:

detail

in

given

ne

"

investigated

functions

llM,

cos

(nu

compared

^r"

took

Bn

sin

the

planetary

detail
earlier

an

in

e-

Bessel

1824

//

He

13

Jo

should

of

the

by

sin

B^

for

in

vector

nTT

this

in

but

1824,

where

examined

radius

the

of

expansion

1826

BEFORE

Bessel.

of

researches

memoir*

The

FUNCTIONS

Ahh.

integral

given

by

[published

1816"17
occurs

in

Poisson,

the

expansion

Connaissancc

1819],
of

des

49"55.

pp.

the

eccentric

Teins,

1825

anomaly

[published

with

1822],

the

p.

notation

383.

of

" 1-4,

CHAPTER

BESSEL

THE

of

them

who

proved

and

fourteen

the

had

of argument
coefficient

the

by

of the

Bessel

and

order

that

the

Bessel's

grals
inte-

Hansen

by
Some

:J:

similar

be

is

developed

into

is called

expansion

it is denoted

Laurent

the

by

series,

the

Bessel

symbol /" (z),

that

so

eH'-d=

(1)

i t-J,,{z).
-cc

n=

development,

this

establish

To

series

absolutelyconvergent
with
the exception of
together,their

product

arranged according to

ZeitschriftfiirMath,

namely

pointed
but

the

are

to

that

of

out

that

found

t Berliner

given

Ahh.

X Ermittelung

1824

Abh.

II.

sur

(1855),pp.

la

e.g.

Sternwarte

determination

When

ii.

text

is

into

these

Soc.

Seeburg

Ital.

165.

"

For

(Modena),

denotfed

Math.

Ann.

of t,

gent
absolutelyconver-

an

an

multiplied

are

it may

so

expansion

somewhat

by Jj,,^what

similar

(1820),p.
we

Traces

(1871), p.

in.

an

be

of the

excepted.

xviii.

universally adopted.

now

Storungen

137

t, t

for all values

series

have

into

expanded

series,and

we

say,

and

^:

(1857),pp.

Schlatli,

des

is to

of

following Hansen,
the

expanded

t.

of t ; that

[published 1826],

250"251.

of

be

of t ; and

powers

absolutelyconvergent

Phys.

Absoluten

der

theil, [Schrifteu der


Memoire

in

elsewhere,

can

Prullani, 3Iem.

Schlomilch,

definition
be

be

e^'^ can

that

ascending

for all values

und

e^cosO^ggg

observe

powers
an

powers

is valid

(1), which

form

zero,

is

of

e~^^'^

descending

of

series

I.

nition,
defi-

("2-22).

P^ in the

n, and

his

f. It

by

memoir.

coefficients

can

of

is due

Bessel

discovered

Schlomilch's

is to

equal to

are

defined

expansion,was
Jacobi

from

defined

theorem

converse

in the

procedure

functions

previouslybeen

that

coefficient

of t; the

function

thus

this function

that

shewn

It will be

the

functions

of

ways

in this work

This

expansion.

properties

several

are

adopted

the

in 1836

published

been

fundamental

propertiesof

publication of

generatingfunction

The

qua

they

will be

certain

coefficients

before

years
had

results

which

the

equal to

are

which

many

mentioned

be

should, however,

in

of the

There
coefficients.

method

; the

incidentallythat

integralsby

definite

discussion

Bessel

as

derived

coefficients.

is the

coefficients

the

as

Schlomilch*,

to

known

functions

definingthese
define

chapter

of functions

set

this

object of

The

COEFFICIENTS

the Bessel

of
definition

The

2"1.

II

503.

now

write

of

Hansen's

expansion,
It must
as

be

J,^{2z);
notation

148.

p. 22.
in
:

Ellipsen
Gotha,

"perturbations

von

beliebiger Excentricitdt

1843], p.
absolues

106.

See

also

the

und
French

(Paris, 1845), p. 100,

and

Neigung,
tion,
transla-

Leipziger

2*1, 2-11]
If in

BESSEL

THE

write

(1) we

1/tfor

"

COEFFICIENTS

t, we

15

get

1t=

-Xi

00

n=

on
a

n
replacing
by n. Since
comparisonof this formula

the Laurent

"

with

is any

J,j(0)as

integer
integral.

an

formula

"

e^zit-llt) J^ (^)+

(3)

function is unique*,

that

derived

l^n+ (

Bessel

by

be written

(2) it is evident that (1) may

From

(- r Jn {Z),

where

expansionof

(1) shews

/-" {2)

(2)

i-trj.niz),
:c

"

from

his definition of

in the form

J^^(^)_
Y ^-n^^

problems of

elementary results concerningJ,^(z)has been given by Hall,The Analyst,


of elementary applications
account
of these functions to
84, and an
been
has
Mathematical
Journal
compiled by Harris, American
Physics
of

Math.

(1912),pp.

A
I.

suinmary

(1874),pp.
XXXIV.

The
IX.

of

81

"

(1909),pp. 433
The

unity has

of order

function

2"11.

391"420.

441,

"

been

in connexion

encountered

with

the

by TnwieYe, Notiv.

steepest

curves

de Math.

Ann.

the surftice z=y

on

(4)

{bx--y^).

ascendingseries for Jniz).


I

for /" {z)in the form of an ascending


series of powers
expression
explicit
the
series
for
is obtainable by considering
thus
exp {hzt)and exp ( ^2/^),

An
of

"

exp
When

is a

integeror
positive

rightwhich, when
rise to

associated

with

^-2-^ L_I^Z_^.
V

11^(^-1/0}=

onlyterm of the firstseries on the


generalterm of the second series gives

the

involvingt^ is the term


always one term for which

for which

term

there

is

terms

for all the values of m,

we

the

zero,

has

Jn{z)= S^

For, if not,

zero

(-m-i ajj^
This

result

could

was

be

expanded

that
particular,

integratedit round
noticed

this value.

On

and, since

0,

these
associating

productis

ni\

the result

(1)

coefficients (say,in

/;

{-\zr

m={i{n-\-m)\
therefore have

that the coefficient of V^ in the

see

I {\zf^
We

of

?"')

circle with

/r

into

were

not

Laurent
zero.

centre at the

by Cauchy, Coniptes liendus, xiii.

'

If

in t, in

series
we

then

origin,we

which

some

of the

multipUed the expansion by

should

(1841), p. 911.

obtain

contradiction.

16

where

is

givenby

integeror
positive

of the series

terms

II

are

the formula
(

z''

Jn {z)

(2)

first few

The

zero.

[CHAP.

FUNCTIONS

BESSEL

OF

THEORY

z^

2,7-;-,
1

2M

(n + 1)

2M

z^

"^

"""}'

(n + l){n+ 2)

particular

In

^0 (2) -I

(3)

2^

22

4''

2^

42

6.

volving
negativeorder,we select the terms int'^ in the productof the series representing
exp (^zt)and exp ( hz/t),
where n is stilla positive
integer.The term of the second series which, when
associated with the generalterm of the first series givesrise to a term in t~"'
obtain the Bessel coefficients of

To

"

is the term

for which

whence

obtain
evidently

we

; and

have

we

the formula

anew

J.n{z)

| 2"! (2),namely

(-rJ,,(z).

that,in the series (1),the ratio of the (m

It is to be observed
is

so

(n + m)],and this tends to zero as


jz^l{7n
ratio test for convergence,
values of z and n. By D'Alembert's
the series representing
J" (z)is convergent for all values of z
oi
is an integral
function z when w= 0, + 1, + 2, " 3,
to the mth

term

711

"

It will appear later ("4"73)that Jn{z) is not


and so it is a transcendental
function ; moreover,

transcendent,that is

an

l)th term
for all

cc

it follows that
and

n, and

so

it

algebraicfunction of z
it is not an
elementary

finite combination
of
as
a
say it is not expressible
and
logarithmic
exponential,
algebraicfunctions operated on by signs of
indefinite
From

to

integration.
(1) we

can

obtain two

(cf Chapter xvi)

multipleof

which
inequalities,

in the discussion

of series whose

of

ance
importgeneralterm is a

are

some

Bessel coefficient.

Whether

useful

be real

or

complex,we

have
00

\Jn{z)\^\lz\^ X
IjZ I

_2jf_l_"^
nl

and

so, when

(4)
This
854

; a

have

0, we

1/" (.)I,
result

was

,"=o

w!(w-|-l)'"'

exp

1ii^"
(i^')
,

exp

(J I. n.

given in substance by Cauchy,ComptesRendus, xiii.


inequality,
namely

(1841),pp. 687,

similar but weaker

Jn{z)\^^~'--exp{\z\-^),
n
was

given by Neumann,

Theorie der BesseVscken Functionen

(Leipzig,
1867), p.

27.

2-12]
all
By considering
found

BESSEL

THE

that

J"W

be observed

It should
bounded

domain

in the

terms

terms

follows from

that the series


variables

of the

if S, A and

For
f-})lane.

are

The

the

on

and

and

or

both

(1) converges
contain

not

exceed
the

in absolute

and

uniformlyin
originin the

value the

sponding
corre-

uniformityof the convergence


apply to the series obtained

expansion2^"/" {z),whether
z

the

if

considerations

Similar

2'1

does

do not
expansionof exp {\zi)
exp {^zjt)
and
of the product exp (^/2a)exp {hR/8),

of the

right in "

positiveconstants

the test of Weierstrass.

The

"'^P^^'"

which

by term-by-termdifterentiations
performed with respect to z or t
2" 12.

"fe^'(l+").

(Ji^VU

|ff|^exp

where

the

of the series for J^ {z) except the first,


it is

the terms

(5)

any

17

COEFFICIENTS

the differentiations be

t.

formulae.

recurrence

equations*
2n

Jn-i (Z)+ Jn+l(Z)

(1 )

Jn {z),

--

Jn-Az)-Jn+dz)

(2)

'^Jn(z\

functions are useful in constructing


Tables
contiguous
coefficients;
formulae.
they are known as recurrence

which

connect

Bessel
To

prove

three

former, differentiate the fundamental

the

of

expansionof " 2-1,

namely

with respect to t ;

we

get

h^z{l+

i nt'^-'Jniz),
llt^)e'=''~"'^
=

SO

that

1^(1 + 1/^^)i
)l=

V'J.Xz)^ i
"

"Xl

the left is arrangedin


on
expression
in
equated the two Laurent series,which

If the
are

"

nV'-^Jn{z).
X

powers
are

of t and

coefficients of P~i

it is evident
equal,
identically

that

^Z {Jn-,
(z)+ Jn^ (z)} nJn (z),
=

which
*

is the first of the

Tlnoughout the work

formulaef.

primes

are

used

to

denote

the

derivate

of

function

with

respect

to

its argument.

f
The

Differentiations are permissiblebecause (g 2 -11)the resultingseries are uniformly convergent.


because Laurent
expansionsare unique.
equating of coefticients is permissible
W.

B. F.

18

Again,differentiate

expansionwith respect to

the fundamental

M=

\{t-\\t)

that

so

n=

By equatingcoefficients of
(2) immediately.
results of

The

Jn (z)+

(4)

zJn (z)
-

Jn (Z)

Jn {Z)

-Z

Jo {z)

formulae

The

(1) and

Bessel,Berliner Abh.

(4) from

1824,

Jn+i(^).

this manner,

but he obtained

formula

(8)

which

obvious

the

und

The

may
method

direct differentiation

(2)by

be

others

derived

derived

of

Phys. II. (1857), p.

^'~ip= ^

"

By

which

Schlomilch

,.C",is a binomial

where

reduce

to

./,(z).

[1826],
pp. 31, 35.

Schlomilch, Zeitschrift
fur Math,

138.

discovered

were

proof given

here

Schlomilch

of the series for

is due

by
to

proved (1) in
J" {z).

{ibid,p. 143) from (2) is

(-)'""c../"-.^."(.X

coefficient.

inductions

from

(5) and (6),we

have

{z-'''Jn{z)]={-r^Z"^-Jn^,,,{z),
[jj-^

(10)
n

Bessel
an

is any

integerand

is any

positive
integer.The

formula

{ibid.
p. 34).

example of

the results of this section observe

that

zJ^ {z) ^Jo (z) zJ^ {z)


=

U,{z)-8J,(z)

zJ,(z)

{-Y-'nJ^,(z)+ {-)^'zJ^^^,{z)

n=l

4
n

smce

formula

Jn-, (z),

n=0, (1) is trivial while the other formulae

case

(7)

As

obtain

In the

to

we
identity

^^{z-Jn(z)}=Z-J,,_,{z),
-z-Jn+d^).
^Jz-J^{z)}

(6)

where

t^Jniz).

to
equivalent

(5)

and then

(1) and (2) are


adding and subtracting
Z

are

-X

either side of this

(3)

These

i
"i=

CO

on

II

-00

PJn{z)^

[CHAP.

FUNCTIONS

BESSEL

OF

THEORY

zJ^K+i(z)-*0

as

-*

(-r-'nJ.^,{z),
l
cc
,

hy ^ 211 (4).

(10) is due

20

THEORY

frequentlyconvenient
and writing2'7r B for

It is

BESSEL

OF

"

to

Since

tegratio
modify (1) by bisectingthe range of inproceduregives

/"'"

cos(nd-zsine)de.

integrandhas period2'rr,the

the

II

0 in the latter part. This

J""(0)=-

(2)

[CHAP.

FUNCTIONS

be transformed

first equationmay

into

Jn{z)

(3)

^ [^'"'^
cos{ne-2sind)de,

^"n- J

where

is any

To

a.

angle.

(1),multiplythe fundamental

prove

integrate*round

which

contour

encircles

expansionof " 2'1 (1) by ^~"~^ and


the originonce
counterclockwise.

thus get

We

dt.

m"n"i

J
The

on
integrals

the contour

Take

be taken

be

to

27r -t-a

to

"Zv +

result

givenby

Hansen

former

f in the

equationtake
part,replace0 by

In this

write t

It is thus found

a.

n;

and

e~'^,so that 0

that

"

case

0.

bisect the range of


0. This proceduregives
tt,

"

r
J",^(^)=i_^g[e'^ne-zsmO)
which

equation(2),from
Various

modifications

Jn{z)
If 0 be

COS

"

TT

I "^%^"(""-^sin^)"(^,
^n(z)=^"7*

and

for which

one

circle of unit radius and

from

decrease

to

(5)
a

the

t"^e^'^'-''''dt

^(^)=^(
27n,\

(4)

may

rightall vanish except

the

obtain the formula

we

so

27n

,"=-x

be

(1) may

of Bessel's
nO

J0

replacedby tt

sin
TT

^ in these two

is odd, the latter when

is

even,

sine)

obvious.

by writing

"0sin(^sin ^) (Z^.

j |"

the former
integrals,

the other

in the

j^0^

obtainable

"

'

"

deduced, is now

are
integral

(zsin 0)d0

COS

line-

and,
integration

being

changessignwhen

unaffected

in each

case

therefore

and

l"'^
sin nQ
.

Jn ("s^)
~

"

sm

{z sm

^) d0
(n odd).

sin n0 sin {zsin

Terni-byterm integration is permitted because the expansion is uniformlj'convergent on the


the symbol J'""*"'
to denote
It is convenient
to use
integrationround a contour encircling

contour.

the

d0
0')

point

once

counterclockwise.

t Ermittelung der abioluten

Storungen (Gotha, 1843),p.

105.

"2-21]

THE

BESSEL

Jn (z)=

COEFFICIENTS

I'""

nO

cos

(z sin 6) dd

cos

^-'^

(7)
=

the latter parts of

J"n{z)

it

'/"(e)
=

(-)*"

[Note.

It

was

due

are

nT?

sin

(2cos 77)dri

{n odd),

11

cos

(zCOS 7;)c^t;

*( 7 cos

(n even).

"j

results

COS
1

"^

last two

(6) and (7),it is found that

r?"

(-)^*" ^^
"

(9)

sin ^)c?(9
C^'
^

Jo

replaced
hyh-rr- i] in

(8)

??^ cos

cos

IT

If ^ be

(71even).

/'^'^

The

21

substantially

shewn

by Parseval,Mem.

+
^~lP"
^-

^y

Jacobi*.

to

des

etrangers,i. (1805),pp. 639"648,

savans

that

and

so, in the

will be
and

special case

in

seen

Poisson's

"

Z-

"

2-3

z-

4-

in which

that

r^r-^,+ ...=

"

r5~T'

4-

5i

COS

D-

TV

(a sui A')a.r,

0, (2) will be described

Parseval's

as

integralrepresentations of J^ (2),namely
identical when
"=0, so a specialname

two

integralbecome

integral.It

Bessel's

integral

this

for

is

case

ju.stified.]
The

"

2-12

reader

will find it

(4) from

2 "21.
Two

Bessel's

Modificationsof ParsevaVs
formulae

integralformula

oljtain

to
interesting
integral.

" 2-12 (1)and

integral.

which
are
involvingdefinite integrals
worth notice.
The first,
namely

Jo {s'{2^-f)\
=

to

formulae

closelyconnected

ParsevaPs

with

are

(1)
is due

(afterBessel)the

Bessel +.

The

simplest method

'^-

''"'

{zsin 6)dd,

cos

of

proving

it is to write

the

the

expressionon

rightin the form


/"""

Stt J

expand

in powers

f:(y

cos

cos

e + iz sin 6)''"+ ^de


then

Catalan

to

Journal

of

(1) obtained

fur Math.

xv.

See

also

Anger, Neiwste

"

^)" ;

i, namely

c(i+-)cos0

had

Schriftcn

12"13.

limits

-z

and

[Ges. Math.
0 and

Naturf. Ges.

der

{(1 z)sin 6) dd,

cos

with
in

1 +z

for

IVerke,
factors

Danzig,

913.
(1854),pp. 910
t Berliner
Ahh., 1824 [published1826], p. 37. See al'io An^er,
Ges. in Danzig, v. (1855),p. 10, and Lominel, Zeitschrififilr ^Math.
t Bulletin deV Acad. 11. de Belgique, (2)xli.
(1876),p. 938.

Comptes Jte7idns,xxxviii.

.,

if

r
T{it+\)
(??+ 1)

by substitutingI

(1836),pp.

integralsactuallygiven by Jacobi

Z'"^

difficulty.

J""(2^"
v'2)

specialcase

2r(n+|)r(^),

^)2"dB

iz sin

'

-n-

"n-

is

Q/ cos

due
integral,

(2)

the formulae

use

0,

follows without

other definite

"tt

6 + iz sin 6 and

the formula
The

of ^

and

?/

respectively.

(1891),pp. 100"102]
I/ttreplacingthe factors
vi.

v.

(1855),p. 1,

aud

the
2/V.

Caucby,

"

Keiieste Schriften
iind

xv.
P/ji/.v.

der Xatiirf.

(1870),p.

151.

22

Catalairs

independentlyby using the

be established

integralmay

formula

z'W-^)

[CHAP. II

FUNCTIONS

BESSEL

OF

THEORY

27ri

that

so

00

^.m
ytti

-1"=".
2
./o(2i\/~-)=
m-o(w!)2

r'
.' (0 +

^Jn

t-"'~'e'dt

-,

271-?,"=o"2! ./

/"("+)

?^^
'^
+

exp

by taking

the contour

to

be

unit circle

-L /

"

+ ^6exp {e^-9

the result then follows

'^lde,

by bisectingthe

of

range

nitegration.
2"22. Jacohi's

series,which

Two

by
^

in
expansions
are

Jacobi*.

The

" e'^ in the fundamental

J,(z)+

connected
with Bessel's integral,
were
closely
them
method
of
is
to
obtaining
simplest
expansion" 21 (3). We thus get

of Bessel coefficients.

series

covered
diswrite

J,n (^)cos2nd " 2i 2 Jo"+i{z)sin {2n + \) 6.


"=0

n=\

On

the
adding and subtracting

we

find

two

results which

are

combined

in this formula,

00

(^sin ^)

/o (^)+

(1 )

cos

(2)

sin(^sin^)=

J^n (z)cos 2n6,

+ 1) ^.
Jo"+i(^)sin(2?i

71=0

Write

^TT

"

77 for

6, and

get

we

00

(3)

{zcos 77) Jq {z)+

cos

22

J^n {z)cos 2nr],


(")""

sin

(4)

Anger

Jacobi's

t.

cosines

of

procedure was
multiplesof ?;, and use

which
integrals

In

view

are

seen

of the

to

Journal

X Neumann,

Fourier's

xv.

with

as

is defined
n

der

to

is

(1836),p.
BesseVschen

to

12.

in

were

sin (s cos

obtained

r;)into

later

by

series of

the coefficients in the form

(1) and

(3)

are

it is convenient

not

of

formed

to introduce

be

equal to 2 when n is not zero,


employment of this factor,which

[Ges. Math.

hi

others

integrals.

terms,

The

zero.

obtain

Bessel's

the other

Schrifteu der Naturf. Ges.


Theorie

rule

the first terms

fact that

fur Math.

t Neueste

given by Jacobi, while the


to expand cosucosj;) and

be associated

law
accordingto the same
Neumanns
\ "", which
factor
and to be equal to 1 when

i (-fJ"2^+i(^)cos(2?i
+ l)77.

(2cos 77)=

results (3) arid (4) were

The

Danzig,

Funcilonen

Werke, vi. (1891),p. 101.]


(1855),p. 2.
(Leipzig,1867),p. 7.

v.

2-22]

THE

will be of

BESSEL

in the

frequentoccurrence

23

COEFFICIENTS

enables
sequel,

write

to

us

(1) and (2) in

the compact forms:


00

(5)

cos

{zsin 6)

(6)

sin

sin ^)= 1, eon+iJm+i{z)^v^{'^n


+ l)d.
(2^

e^n

J-m {z)cos 2nd,

00

If

put ^

we

0 in

find

(0),we

1=5

(7)

71

If

differentiate

we

e,,J,n(2).
=

of times

(5) and (6) any number

before

putting^

0, we

series of Bessel coefficients. We


for various polynomials
as
expressions
different method
shall,however, use a slightly
subsequently(" 2'7)to prove
is any positive
into a series of Bessel coefficients when
m
that z'^ is expansible
that
is
This is a
integer.It is then obvious
any polynomial thus expansible.
gated
which will be investiof an expansiontheorem, due to Neumann,
case
special
in Chapter xvi.
obtain

For

will

the present, we

before d is put

z=
n

6,"+i

(9)

sin 2^2

(10)

cos

results

in

The

be strict

are

(2n + 1) Jsn+i(^).

while, if 6 be put equalto ^ttafter

Note.

differentiations of

two

4^

11- J, (z)

3-^Js

due

The

(5) and (6),then

J, (z)+ 6' J, (z)

[22J, (z)

(z)+

5-

.],

J, i^)-"""}"

Lommel*.

to

in

expressionexp {^^(i- 1/0} introduced

sense.

once

equalto 0, there results

(8)

These

differentiated

merely notice that,if (6) be

t associated

generatingfunction

^-l is not

generatingfunction

(,iJn(z)is

with

e,i^"*A
{^)-

(1=0

If this

If

we

expressionbe

solve this differential

result

-S',
by using the

equation we

recurrence

formula

"

2-12

(2),we

have

get

jyh^it-mj^^^d,.
s=ei-(t-iin+l(^
^e^^(t-vn

(11)
A

called

equivalentto

this

was

given by Brenke,

Bull. American

Math.

Soc.

xvi.

(1910),

pp. 225"230.

Studien

ilher die Bessel'schen

+ It will be

seen

in

Chapter

xvi.

Fiiitctionen
that

(Leipzig,1868),p.

this is

form

11.

of "Lommel's

function of

two

variables."

24

THEORY

OF

[CHAP. TI

FUNCTIONS

BESSEL

integralfor the Bessel coefficients.

2"3. Poissoiis

before the appearance of Bessel's memoiron


perturbations,
planetary
Shortl}'
of
Conduction
the
work
Heat*, in the
had publishedan important
on
of the types f
of which he investigated
course
integrals
Poisson

(z cos 6) sin-'^+iddS,

cos

Jo

cos

(zcos 6) sin^'*dcie,

-JO

He proved that these integrals


are
positiveintegeror zero.
which
and gave the investigation,
solutions of certain differential equations!
has alreadybeen reproducedin " 1"6,to determine
an
approximationto the
0.
latter integral
in the special
when z is largeand positive,
"?
case
where

is

We

shall

and, in view
describe
case

now

that

prove

to
appropriate
importanceof Poisson's researches,it seems
the expressions
the right
for J^i^). In the
on
" as Poisson s integrals
reduces to Parseval's integral
0, Poisson's integral
("2"2).

of the

It is easy to prove that the expressions


under consideration are
Jn{z)',for,if we expand the integrandin powers of z and then
we
term-by-term||,

cos

T^.'o

equal to
integrate

have

(zCOS 6)sin2" 66.6

\^

d sin^*^OcW

,,

(_)m^2m 1.3.5...

cos^'"

(2m)! Jo

7r,,,=o

(2n

1). 1

3
.

(2m

...

1)

"

",=o

(2w)!

1.3.

5.

2.4.6...(2??+ 2m)

..(2/1-1) 2
m=o

and

\,
^,^,,1
2"^+^'"'
+
m

{n

.,,

ni)I

the result is obvious.

Journal

t Ibid.
examined

p.

de I'Ecole E.

293,

by Euler, Inst.

elegant, and

more
+

E.g.

on

p.

xii.
Polytechnique,
(cahier19), (1823),pp.

et seq. ; p.

his

Calc.

340,

et

seq.

Int.

ii.

(Petersburg,
1769),Ch.

study of them

is

more

Integrals equivalent
x.

to

249"

i03.

them

had

" 1036, but

previouslybeen

Poisson's

forms

are

systematic. See also " 3-3.

300, he proved that, if

Jo
then

satisfies the differentialequation


"PE

" Nielsen,Handhuch
BesseVs

IIThe
to

second

der

Theorie

integral,but the above

series to be

Poisson,ibid.

der

n{n

l)
^

Cylinderfunktionen
(Leipzig,1904),

nomenclature

seems

p.

51, calls them

preferable.

integratedis obviouslyuniformly convergent; the procedure adopted is

pp. 314, 340.

due

2-3, 2-31]

BESSEL

THE

also observed*

Poisson

25

COEFFICIENTS

that

gizcos e^^^^n

0dd=r

(zCOS 6) siri""ddO;

COS

.' 0

when

this is evident
left and

derived
integral

the

We

consider the arithmetic

we

it

^ by
by replacing

tt

the

0.

j.^;ii|"j^/%-.'sin"^

/"(.)

slightmodification

of this

formula, namely

suggestedimportant developments(cf." 6'1) in

has

"

on
integral

thus get

(2)
A

from

of the

mean

the

theory of

Bessel

functions.
It should

also be noticed

(4)

that

{zcos 9)sin2" dd0

cos

(zcos $) sin-'*Odd

cos

Jo

Jo

['"cos
(zsin 6) cos-'* 0d6,

Jo

givesrise to
expressions

and each of these

modified

form of Poisson 'sintegral.

of Bessel's and Poisson's integrals


was
interesting
application
t
he
formula
who
multiplied
by Lommelf
4n-[47i2-22|...(4w2-(2m-2)-}
^
^

obtained

An

"

cos2n^=

by

:i

^^sm-"*^
.

-Tii^,

(-)"*

,,

It thus follows that


{zcos 6) and integi-ated.

cos

4??-[4n--2-}...
[4n--(2m-2)-|j;"(^)

"
,

m=0

J,is somewhat

It is

seen

lU:

integral.

which
integral,
follows :
as
elaborate; it is substantially

that ./"{z) is equalto


proof,

The
Bessel

investigation
of Poisson

Bessel's

2"31.

Poisson's

was

given by

differentiation that

on

2'

cos

6 sin-'*~i6

cos

dd

sin-"-- d
1)
^

his

the
actuallymade
but, as he points out on

Poisson

sin'-"-!-!
0

sin^'*+^6 sin

"

(2n

{zcos 6)

~\

^
X

2w sin^**6 +

{z cos 6)

sin-^"+-0
;^-^
2n
1

cos

(z cos 6),

statement
p.

ft

340, odd

(p. 293) concerning the integralwhich contains


throughout
be replacedby even
powers
may
powers

analysis.
f

Studien

Berliner

[Ges. Math.

objectedto

Functionen
(Leipzig,1868),p. 30.
p. 13.
Jacobi, Journal fiir3Iath. xv. (1836),
[published1820], pp. 36"37.
formula,
givinghis proof ("2-32)of Poisson's integral
(1891),
p. 102], when

ilher die BesseVschen


Ahh.

Werke,

1824
vi.

the artificialcharacter

of Bessel's

demonstration.

and

when
integration,

hence, on

^ 1,

"

(2n

1)

(zcos 6) sin^'^--ed0

cos

If

-^^^--

I"cos(^

T^-rrv-w-r.

(z cos 0) sin^" OcW

cos

["cos
(z

0) sin-"+- OdO

cos

0.

(n
z(j)

1)

(n 4- 1)
(n)+ Z(f)
2n(f"

the
J^ (z)satisfy

that (f)
(n) and

6) sin- ^fZ^

cos

c^{n),

that

shews

the last formula

so

-.2n

write

we

now

[CHAP. II

FUNCTIONS

BESSEL

OF

THEORY

26

formula.

recurrence

same

0,

it is evident that
But, by usingBessel's integral,

(f"{0) Jo(z),
=

^ (1)

and

when

11

{z

cos

from

0, 1, 2, 3,

the

f^

7^

"

.)

^^
1^^"

"

^H

^^^ ^^^^

have

formula, we

recurrence

c*^^

J^ {z),

J^ {")

..

investigation
of Poissons

Jacohis

232.

6) sin- 6d0

cos

cos
6) cos Odd
(^^

sin

induction

by

so,

r-TT

integral.

into
problemof the direct transformation of Poisson's integral
necessitates
attacked by Jacobi*; this method
was
successfully
integral
The

of Jacobi's transformation

where

cos

fjb
=

"

6. We

the

use

formula

sin-"-i ^
c?'"-i
"

Bessel's

"

shall

1.3.

..(2/1-1)
^
^

5.

=(-)"-i

this formula

assume

"

smnd,

for the moment,

and, since

been previously
to have
we
simpledirect proof of it seems
published,
2'323.
givean account of various proofsin ""2'321

no

shall

"

If

observe

we

jx, vanish when

z'^

COS

/i

that the lirst n"


+

B) sin2" ddd

{zcos

1 derivates

1, it is evident

that,by

2"

cos

of
n

(-r

cos

Journal

also Jom-nal

fiirMath.
de Math.

{Zi, Invr)

(1836),pp. 12"13.
196.
(1836),pp. 195

xv.
1.

"

_^

we
integrations,
partial

{zyC)(1
^

with respect to
(1 "/a-)'*"*,

[Ges. Math.

have

yu,-)'*-*
fZ/x

^\^^^dfx.
Werke,

vi.

(1891),pp. 101"102.]

See

28
2*322.
The

Liouville's

Let y

Liouville of Jacobi's formula

{\ -^-}^~- and let I) be written

Difi'erentiate this

times

equation n

follows

then

nAD"-^7/
f-^,

that

A and

where

determine

To

B
A

ascendingpowers

so

compare
of 6.

that

and

since

"4

thence

we

have

the

(-)"^i

1.3.

Lagrange'sexpansion,which
A

ce

Now

the usual

conditions

take

of

/(2)=

of z qaa
singularities
of J{\
z^)in powers
"

fJn

/x +

to

sin 6 when

v^(l-m^),i.e. to
of h

at A

are

convergent

(1

^/(1

\)t-i

it follows

that

\ /

e^'^; and

both

when

|h

,2

in powers
z-) {czjOjj.)
.

so, when

\and

| \are
z

6 is

cZ"~^ sin

"^""^

of h.

But

it is evident

of the coefficient of /i"


~

in the last

coefficient of A"-"

in the

ex-

that

e^"^-e-"'^
^

"

Analysis," 7-32.

"_,
'

expressionestablishes

Jacobi's formula.
Cf. Modern

unity.

o?/i""i

pansion
ex-

^
is the

cz _{\-he'')---{\-he-'\'- 1.3.5...(2".-1)
I
/""
^^'3m
2.4.6...(2?t}

consideration

real,the

less than

-"

A?

"

A^-0.

2^/i+ k^)]ih^

-"

2"~i.(?i-l)

pansion of J{1

then
A/'(s),

"

^^^-'^
^-^(1-2^/^+/.^-)'

^"

and

that, if s

of h is
3

^"""^

is

(1 -^2)^ "^'(2) V(1 -A

function

Now

Hence

n6.

convergence*.

being supposed that ^' {z)reduces


The

sui

SchUiJii's
proofof Jacobi's transformation.

firstrecall

subjectto

(_)""-!

..(2"-l)
^
^

1.3.5.
,

We

..(2/1-1),

5.

result,namely

c?"-isin2"-i(9

2 "323.

^cos??^,

I)"-'^yis obviouslyan odd function of B, B is zero.


the coefficients of 6 in the expansions of D"-'^y and A sin??(9 in
to be
The term
involving 6 in D"-^y is easilyseen

constants

are

0.

/1 sinw^

Z""~iy

Hence

it

as

d/dii;then obviously

for

and

is

a-'^^)^-^-M^="-^^X^S+'''^^="^

but

so

transformation.

proof of JacohVs

proofgiven by

[CHAP. II

FUNCTIONS

BESSEL

OF

THEORY

the truth

of

2'322-2-33]

nxdx

/ (coscc)cos

}0

S
m

(" )"*a.w/"^'+-'"'
(cosx) dx,
0

"=

coefficient of p^^-'"^in the

in asexpansionof Jn{t)/Jo(t)
cending
of t, has been studied by Jacobi*.
To establish it,integrate
powers
the left n times by parts ; it transforms ("2'32)into
on
expression

in which

the

29

expansion

formal

The

COEFFICIENTS

of Jacobi's transformation.
application

An

2'33.

BESSEL

THE

is the

a,"

1
"

1.3.

sin-"^ is replaced
by

and, when

f""
I /"^'
(cosx) sin-^xdx,

rx

j^

5. ,..(2?i-l)Jo
..(2/1-1)

series of cosines of

ofx, this
multiples

becomes

2/1

2n(n-\)
2.4.(/...(2.)/o"'^^^^""
(?i-fl)(w 2)
.

r-

COS

ZX

^rr-.

n-^\

We

COS

dx.

\X

cos
4a;,
integrate
/""(cosa?)COS 2.r,/'"'(cosa.')
by parts, and by
a
rrive
this
of
at
a formal
repetitions
evidently
expansion
process, we
is
in
of the type stated. When
the
a
cos
polynomial
/(cos a;)
a?,
process
valid.
obviouslyterminates and the transformation is certainly
now

...

continual

determine

To

the values of the coefficients a," in the

If

(cos x) cos

nxdx

expansion

(-)'"am/"'"^'""
(cosx) dx

thus obtained, write

/(cos x)
accordingas

is

even

or

sin (tcos x),


cos
(")-'*
(t cos x), (" )i""-i'

odd, and

Jn {t)

that am

SO

It has
in
to

been

stated

cos

be

an

providedthat

that

no

*
.

it can,

lim

Jilr Math.

also Jacobi, Astr.

expansionis

a/j6" jis

Nach.

xv.

.5

less than

of

cos

x,

when

say

6"co.s"a;

the smallest

this will not

[Ges. Math.
(1836),pp. 25"26
94
col.
[Ges. Math.
(1849),

xxviii.

/(cos x) is a polynomial
is
stricted
merely re/(cosx)

valid when

however, be established

function
integral

0; the

Journal

" 2-2 (8) and (9) that

(-)'"a,,^"+-^'"
{(-r /o (t)],

the

of
investigation
practical
importance.
=

from

has the value stated.

J^ (0

deduce

we

be

root
positive

given since

Werke,

vi.

Werke,

vii.

of the
it

seems

equation
to

be of

(1891),pp. 117"118].
(1891),p. 174].

See

30

THEORY

Bessel

formulafor the

24.

The addition

The

coefficients possess an
expressedin terms of Bessel

be
may
which was

in

simplestway

of

provingthis

which
z.

Jn {y + z)
This formula,

J",(y)Jn-,n{z).

J,,{y+ z)=

The

formula by

coefficients of y and
is
and Lommelf,

first givenby Neumann*

(1)

coefficients.

addition

Bessel

[CHAP. II

FUNCTIONS

BESSEL

OF

CO

result is from the formula

" 2*2 (4),which

gives

/" {y + z)=

e^'2'+^'"-""'"'dt

"-1

^-~.

rto+)

1
1

CO

V^-''-'J,n{y)e^'^^-'!*^dt

'

V.Trl
'Ztti

...

00

/"(0+)

oo

27n

m=

00

00

J, (y)Jn-m (z),

-ao

1)1=

changingthe order of summation


analysis
; and this is the result to be

Hansen's

series

of squares

and

the producton
By expressing

j I )?(

00

the

be

given in Chapter xi.

productsof Bessel coefficients.

addition formula
Specialcases of Neumann's
earlyas 184-3. The first system of formulae is
fundamental
expansion" 2'1 (1),so that

(,r

the third line of the

established.

of this expansionwill
generalisations

Numerous

2'5.

in
integration

and

on

rightas

givenby Hansen J as
obtainable
by squaringthe
were

Laurent

series in t,and

the coefficient of ("" in the result to the coefficient of P


of the

the left,
we

on
expression

equating

in the Laurent

pansion
ex-

find that

00

Jn{2z)=
In

particular,
takingn

(1)

J, (2^)

Theorie

der BesseVschen

t Stiidien Uber
Ann.
+

and

in.

Jo^(^)+

Functionen

die BesseVschen

(1871),pp.

Jr{z)Jn-r{z).

have"

0, we

J;^(z)
(-)'"
=

(-)'"
e, J,'(2).

(Leipzig,1867),p. 40.
(Leipzig,
1868),pp. 26"27

Functionen

gave

see

also

Math.
Schlafli,

der absoluten
Storungen (Gotha,1843),p. 107 et seq. Hansen
did
of (2)in which n
only the specialcase
\. The more
general formulae

Ermittelung
he

135"137.

Loramel, Stiidien ilber die BesseVschen

Functionen

" For brevity,


J,f{z)is written in placeof

(Leipzig,1868),p.
(z)}-.
{J^^

33.

not
are

give (4),
due

to

2-4-2-6]

generalformula

the

From

Jni^z) =1

(2)

find that

we

J'

coefficients of

the Bessel

Jr {Z)Jn-r (^)+

)"=()

when

31

COEFFICIENTS

BESSEL

THE

./,(z)J,^, (z),
(-)'"
l

negativeorder

are

removed

by using" 21 (2).

since
Similarly,

exp

1,

1/01 exp [iz(-

{^z (t

t +

1/01

it follows that
2 2

J,H^) +

(3)

/.n^)

in

l.

cc

+ 2 2 J, (^)/,"+,.
(^)
(-)'-j:,(^)J,"_,.(^)

(4)

!"

0.

the coefficient of P* in the La


Equation (4) is derived by considering
expansion; the result of consideringthe coefficient of f-'^+iis nugatory.
A

very

?"

2'6.

1, 2, 3,

was

by

from

Hansen.

J- r

e^ine-zsm9"^0^

so

To

reduce

by

the

this double

integralto

singleintegraltake

new

variables defined

equations
6-cf"

so

is real,

"2-2(5) that
J,,(2)

and

for Jn"(z).
integral

Neumanns

It is evident

noticed

...,

(3),namely that, when

\Jr{x)\^lls/%

|Jo(^)kl,

(5)
where

of

importantconsequence

2x, e

"t" 2f,
=

that

It follows that
f,-2usin^cosx(^;^c?x/r,
e-'"'^
=^^,ii

J,;-;
(2)
where

the field of

is
integration

the square

for which

integrandis unaffected if both x ^^^^ "^ ^^'" increased by tt, or


decreased by tt, the field of
is increased by tt while \/r
is simultaneously
evidentlybe taken to be the rectanglefor which
may
Since

the

if x
gration
inte-

BESSEL

FUNCTIONS

J^n (22 COS

C^X.
;j;)

replace% by ^tt+ ^,accordingas

;^ is acute

32

THEORY

OF

[CHAP.

II

Hence

1
.'0

"n"

If

we

obtain the

obtuse,we

or

result
"in-

J,-^{z)

(1)

I'"
J,n{2z sine)dO.

IT

'his formula

may

is the result

I fi

written

obviouslybe

actually
given by

3laborate transformations

Neumann*.

It

derived

was

the addition-theorem

from

which

by

him

by
given
proof of

will be

justbeen given is suggestedby the


which was
jdition-theorem
publishedby Graf and Gublerf.
different form of the integral
if we perform the integration
a
3 obtain
aspectto X instead of with respect to yfr.This proceduregives
The

2.

proofwhich

in the form

has

"

Jn' (^)

^ f

Jn (^)

TT-

^0 (22 sin f) e^"'l'dyjr,

1
=

f"

^0 (22 sin i/r)


COS

2n\lr

dyfr

1
=

"^

"

f"

"

result which

attributed
Schlafli;|:

2"61. Neumanns

By taking the
rightin powers of

and

dylr,
2)i\lr

to

Neumann.

for J^ (2).

series

formula

Jo {'22sin ylr)COS

Jo

"2'6(1),expanding the Bessel coefficient on the


then integrating
term-by-term,Neumann
" shewed

that
1

Ttt

S
Jn'(z)=-\
Jo
TT

\)n "2H+2)n

;"=o

^ (-Y{2n

\,^
m\{2n

ciri
''"

2n+2*;i

"

dd
-f

"

m).

2my.{^2Y"+-'^

'

m=o

'

der

Theorie

Einleitung

J The
" 2Iath.

formula
Ann.

in the

BesseVsehen
in die Theorie
is
iir.

an

Fiaictionen

1867), p.
(Leipzijj,

der BesseVsehen

immediate

(1871),p. 603.

Leipziger Berichte, xsi.

m)\ {{n+ m)\\-

m\{tn+

consequence
The

70.

Funktioiien,11. (Bera, 1900),pp. 81

memoir,

of

equation 16
in which

(1869),pp. 221"256.

on

"

p. 69 of Neumann's

this result

vv-as

given,was

85.
treatise.
first

lished
pub-

2-61, 2-7]

BESSEL

THE

This result

written

was

COEFFICIENTS

Neumann

by

33

in the form

Toz^

(1) Jn'i^)

T,^
+

{2n

1)

(2w + 1)(2/1+ 2)

1.2.

where
2/1 +

2/rr2'
(2/1+1) (2/1+ 3)

4)'

(271+ 2) (2/1+

(2)

(2//+ 1) (2/^+ 3) (2/^+ 5)


'

(2/i+ 2)(2/"+4)(2;z+ 6)'

This

generalexpansion(due to
expansionis a specialcase of a more
functions
Bessel
as a series of powers
with
productof any two
coefficients
comparatively
simple
("5"41).
for the
Schlafli)

2'7. ScJddinilch's
We

shall

obtain

now

series

0 has

of sin- 6.

in

cerning
"2*22con-

These

is any

alreadybeen givenin "2"22(7).


for

cos

2//^ and

sin

(2/i+ 1)^

expansionsare*

(2 ,,",).
i^(_).M^+--;^f

2".=

cos

foreshadowed

was

"2-22(1)and (2) substitute

expansionsin powers

of Bessel coefficients.

series of Bessel coefficients,


where

of z"^ in a
expansibility
positive
integer.The result for m=

their

result which

the

the

In the results

in

expansionof z^

('2??+

"

(n

1)

sV

Sin

The

results of substitution

(zsin 6)

cos

J, (z)+

=.

are

J," (z)

\I

(-f

y*/"
t^^'(2 ^)4
sin

1
If

we

that it is
'

the series

rearrange

to
permissible

the

on

power

series in sin 0

(assuming

have

do so),we
S

rightas

x)

+ .9-l)!
(-)H2sin^y^^i^ 2//..(/i

2JJ..(.)jJ^^-^A^^i"^^^_^^, ^J.M\,
cos(.sm^)^|/.(.)
"

.X

^ (-y(2sindr+'
S --^
sin(2'sni6')=
rvi
^
^
+ 1)!
(2.5
,=0
"

"

ns

"

Cf.
W.

B.

F.

Hobson,

Plane

(2//+ l).(/i+ 5)!


{ ^,
-^
i^J^
"271+1
^n^
7
{n-sy.
{n=s

Trigonometry (1918),""80, 82.

"

If

of sin 6 and equate coefficients,

left-hand sides in powers

expand the

we

[CHAP. II

FUNCTIONS

BESSEL

OF

THEORY

34

find that

we

n=l

(^
(i^""=i'^4S^-^'""^)'
1-2.3.-)

(,,)."

itn+l).(n

The

first of these is the result

into the

The
He

how

can

make'use

we

by

Neumann

of the

of the

double

series

of the double

convergence

^2nJrl)\

^^'^

the series for sin {zsin

the series of moduli

The

be

bined
com-

is

series.

^^'^ ''^' 72^" ^ ^' ^" ^ ^^'

6) we

that

see

similar

may

the rearrangement
justification;

needs

now

I2 sin
6
"'"
"^ |2"+i
I
I

"i

.)

givenby Schlomilch*.
which was
givenexplicitly

generalformula
f and Lommel:|:.

""

so

1, 2,3, were

"

and

inequalities

with

the

establish the absolute

I'^2,,
!"
1 {z)
in connexion

; the others

alreadyobtained

(1)for which

to obtain

rearrangemeut

permissibleif we
If

of

cases
particular

years later
The

0, 1, 2,

(.

singleformula

also shewed

some

j^^^^^^^^

^7-

v''

n=s

sy.

is convergent.

The

(2.+ 1)!

sinh

(Iz

sin 6

series for

cos

I, 128 +
I
I^'l

('A I 2 |2'l
exp(^|2|)
pxn

1z |2),
\)exp (|(2sin 6) may

be treated

in

manner.

somewhat

elaborate

analysiswhich has justbeen given is avoided in


proofby induction,but this proof suffers from the fact that it is
tion.
supposedthat the form of the expansionis known and merely needs verificathat
Lommel, we assume
If,following
Lommel's

az\m-

Zeitschrift
filrMath,

Theorie
"j-

X
IS

Studien

der

und

BesseV schen

Phys. ii. (1857),pp.

n-l)l

140"141.

Functionen

iiber die Bessel'schen

given later in this section.

(m+2n).(m

(Leipzig,1867),p. 38.
Functionen
(Leipzig,1868),pp.

35"36.

Lommel's

investigation

36

were

obtained

gave,

as

where

of

or

2p times

with

The

(1857), p. 141, and

he

of the

series for which

in $^, and

term

so

since terms

equidistantfrom
The

equal

truth

from

manner

reader

Schlomilch

"

^1^ !

-^2

will

expanded

2(-V'

"'

(2m;

in

asc-endingpowers

of 0,

Ci'2//i-2X-)^

t=,"

z^I^S!,

the beginning and


equation (1) is now

of
2-22

when

jn"p,

it is sufficient t"3 evaluate

The

is that for which

of the summation

1). To prove the first formula, take the equation " 2-22 (1),differentiate
It is thus found that
respect to 6, and then make 6 equal to zero.

terms
no

the last term

coefficient and

binomial

Di=o

similar

n.

i*^

are

Phyi.

i^(m"

is ^rfi 1

und

^C'tis

contain

by Schlomilch, ZeiiscArift fiir Math,

the value

[CHAP. U

FU^'CTIOXS

BESSEL

OF

THEORY

the end

of the summation

e"dent,

and

equation

with

,2} is

to

respect

proved

in

(2).

easilyestablish

the

following specialcases,

which

were

stated

by

|13J, (^)+33 Ja (z)+


(4)

-,

bKJ.,(.)+

22 ./,(2)+ 42 .7,(2}-5-6^ Jg (")+

...=!(5+^3)^
=
.

l^,

l2.3.4J3(2;^4.5.6J5(2:-i-6.7.8^-(2)
+ ...=M

of

2-72.

Xeumanns

From

Schlomilch's

even

Bessel

"

take

the

Q\'^

integratewith respect

that

(1)

series of squares

of Bessel coejjicierits.

expansion
(2m +2n).(27n

(when

rn

"

+/(-!):
..

.^

to

find that

6, we
^

so

as

expansion("2-7) of 2^ as a series of Bessel coefficients


order,it is easy to derive an expansionof 2*^ as a series of squares of
coefficients,
by using Neumann's
integralgiven in " 2"6.

Thus, if we

and

expansionof z^

(2m-2n).{2m-\-n-l^

0)

(i^)-=i^

("2m
V

9n).(2rn^n-l):

"2-72]

result

This

and

BESSEL

THE

this

was

is

true

special

As

cases,

given

by

when

Neumann*.

An

it

for

0,

37

COEFFICIENTS

then

alternative

reduces

to

form

is

Hansen's

formula

of

have

we

2-=

^n-J^-'Z),

f".

(3)

o
.

If

we

differentiate

(1),

use

it=3

"

2-12

m\{rfL-V)\
,

"

an

'

(2"i

and

then

it

rearrange,

Leipziger

is

existence

Berichte,

was

xxi.

indicated

(1869;,

by

p.

readily

found

that

(2rft+"-2)!

2n-l).

whose

expansion

--"

(2;

226.

,.

,,

Neumann.

[Math.

Ann.

iii.

(1871),

p.

585.]

"

2"5.

CHAPTER

FUNCTIONS

BESSEL

3'1.

TJie

The

Bessel

of two

which
coefficients,

be

required to
have

functions
This

effected

analysishe

linear differential
definition

by

shall

equation which

of the

Bessel

function

r(^
the

for

which

values

functions

the

v,

The

had

shall

Bessel

finction of order

course

desirable

to

integralvalue
We

d' Analyse,

n,
not

called

iiher die

ii.

i^

Math.

Following Lommel,
being

glance

reserved

come

whether

this

(Paris,1840),
i.

into
a

(1869),

for

to

as

Bessel

course

of the

Lommel's

wasf

''''"' ^'^^'

by

effected

Lommel

by

of

real

by HankeliJ:
;
of

means

an

tension
ex-

2 "12.

with

order

whose

"

solution

solution

to

odd

an

Bessel's
v,

of this

reduces

as

of

Plana

and

integer

treatise.
in

occurs

that

1'6

is half

Lommel's

which

certain

solutions

and

differential

then

define

equation ; it is of

Jn (z) when

p.

assumes

not

0,

necessarily

an

v.

p. 1.

integer) were

studied

byDuhamel,

Coum

469.
the

symbols

integers.

This

in

this

is true

(.^-.-^)y

equation

118"121.

use

use

differential

of the

(Leipzig, 1868),

Functionen

(with v

pp.

we

general

result

of

solution

order

equation for functionsof order

Bessel's

BesseVschen

"

yet

be

to

discuss

resembling

Ann.

and

(realor complex) number"

Studien

so

in the

integral;

being

publicationof

.^g ^g
will be

been

n.

shall therefore

t Integrals

"

of

functions

select such

(1)
which

definition

in this section.

comparing " 3'8

the

before

values

defined

were

on

unrestricted

an

argument

replace the integer n

now

equation by

the

"

observe,

century

of

complex

investigatedBessel

nearly half

so

discussed

formulae

will

functions

defined, is

is to be

Jv{z)

to

less than

recurrence

reader

We

extension

of order
of the

Poisson

exceeds

R{v)

of

hitherto

right is convergent for general complex values


apparently contemplated only
"\. Lommel

the

integralon

function,

''''"

and

Poisson's

^^
^'
C'^''"
1)^(1)

J. {z)

functions

ii, are

has

whose

by Lommel*,

the

that

but

generalisethese

now

of
generalisation

shewed

Chapter

(complex) variables.

effected
a

in

is unrestricted

unrestricted

of two

was
generalisation

function was
of his

integer. We

an

discussed

were

of which

n,

differential
equation.

BesseVs

generalisation
of

variables, z and

III

v,

distinction

country.

for unrestricted

denote

fj. to

is

unrestricted

customary

It lias
functions

the
or

on

obvious

the

numbers,
Continent,

advantage

for functions

of

the

symbols
it has

though
of

shewing

integral order

at

only.

3-1]

BESSEL

Let

us

construct

now

assumed

form

for such

solution of

solution is

39

FUNCTIONS

(1) which
series of

is valid

the

near

ascendingpowers

origin;the

of z, say

00

where

the index
that Co is not

For
so

and

"

^111 -^

"

the coefficients c,n

"

be determined, with the proviso

to

are

zero.

differential operator which

brevitythe

(1) will

be called V^,

*^,^^U,=

^.,,=

(2)
It is easy to

The

111

the

expressionon
"

if

111=0

rightreduces

of

in the two

of

series

{{a+ ly V-]
+ 2)--v~\+Co
c,\{a

c,n

If,then, these equationsare

(4)

values of
Now

i^; for Cq

=0

C,n_-"=

have

CmZ''+"'=Co{or-v')z'^.
series can be a solution
postulated
and ^" vanishes onlyfor exceptional

that the

is not

zero,

z.

consider the ntth

written

v"}+

satisfied,
we

this result,it is evident

"'l)
only if a

equationin

the system

(3) when

It

1.

ui"

be

can

in the form
c,a (a"v

it determines

so

"

?/

or

integer(when
We

a-Y
a=

When

then
the

is

by

c"i_o

0,

of c"i_o for all values of


that is,unless
negativeinteger,

2v is

v) or unless

{a + mf

constants

formal

+
III)

c,",in terms
v

"

III)
(a +

v-

"

and

processes

procedure actually produces a

is

does

c,,^have

easily

solution

seen

not

been
to be

greater than

m
"

2v

negativeinteger(when

values
disregardthese exceptional

3-5),and
*

[(a+ mf

V,

From

obtained

=0

(3)

unless

rightcancel.

the

on

that the coefficients of

so

equations

c,

and

of the first series,

the first term

to

the coefficients Cm

choose

we

givesthe system

This choice

_,,

that

see

2",
namely Co (a- i/-)
corresponding
powers

of

in

occurs

that

of

determined
convergent

of the differential

for the moment

when

vanish

is

??i

l, 2, 3,

negative
i").

(see ""8'11,
It
....

now

by the following analysis, the series


and

that the formal


so
differentiable,

equation.

equations(3) that 01
of Co by the equation

the

follows from

03-65=

in terms

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

40

=0, and

...

(-rco

that

c^m

is

pressible
ex-

(5)
a

"^

Co2-

formal

see

v, we

solution

",r1 ml{v

l){v + 2)...{v+ m)

equation(1). If

of

take

we

obtain

we

~v,

second

solution

formal

Co'2- 1+

(6)

'

,riml{-v

(5),so that the

Cq and

constants

in

Cq

m)_

Cq, because

the

2)...{-v

placeof
without

carried out

be
evidently

obtaining(6) can

l){-v

written

In the latter,cj has been

of

(4) that

from

is

take

satisfied ; and, if we

system of equations(3) is now

The

to the existence

reference

independent.

are

independentof z may be assignedto


of obtainingsolutions
but, in view of the desirability
define them by the formulae*
V
n, we
Any

procedureof

Cq and

the constants

values

reducible

to

Cq';

Jn{z) w^hen

-^

series

The

(5) and (6)may

be w-ritten

now

(-)'"(i^)""^'"'
,"romir{-v + m + l)'

(-y'^(l^)''+2m
,"tom ir{v + m + l)'
"

In the circumstances
of powers converge
differentiations are
which
two

solutions of

the

{z

these series
integer,
so
term-by-term
in the analysis
f by

an

0 excepted)and

operationsinvolved
and
legitimate,
consequently

are

so

we

have obtained

equation(1).
series defines

two

and

z, of the

argument
/^
{z). Since
symbol
v

of z,

2v is not

The
permissible.

first of the

The
order

considered,namely when
for all values

obtained

theywere

the present, 2i^ is not

an

function

kindX;
first

is unrestricted

the
integer),

and

called
the

Bessel

function

functionof
by

is denoted

(apartfrom the condition

second

series is

that,for

evidentlyJL.^{z).

is defined
the functionJt.{z)
Accordingly,
hy the equation

It is evident

from

" 2-11

that this definition continues

integer(zero included),
positive
a

identical with
*

For

X Functions

up

function

of

to hold

when

integralorder

Bessel coefficient.

propertiesof

+ Which,

Bessel

to the

the

Gamma-function,

present, has been

of the second

and

see

Modern

Analysis, ch.

xii.

purely formah

third kinds

are

defined

in ""

3-5, 3 54, 3-57, 3*6.

i^

is

being

3-11]

BESSEL

symbolic solution
interesting

An

FUNCTIONS

of Bessel's

41

equationhas

been

givenby

Cotter*

in the

form

[l+z" D-'^ z-'^"-^B-'^ 2"+^]-^


{Az'+ Bz-"),
where

djdz while

and

This

constants.

are

be derived

may

by writingsuccessively

[D{zB-2v)+z]z''^ 0,
=

L"-^z]z''i/=-2vB,
+ z--''D-h''^\j=
-2vBz-~\
zD{z-''y)
[zD-2v

result.

givesCotter's

which

" o\,

two

order is

whose

3'11. Functions
In

of Bessel's

cases

It will

now

is half of

that

is half of

values of
odd

an

If

is

we

take

general

zero.

find that

=0,
1

ni

and

in the

(r + i)^

+ 2)
[Cx.l(2r
+ 2r
\c,n.('"n

'

be included

of " 3'1, we
-^in the analysis

,,.

an

let
integer,

positive
integeror

omitted
temporarily
is
(ii)when
integer,

odd

v.

ir

where

integer.

(i)may

case

When

odd

generalised
equationwere

be shewn

integerf
theoryfor unrestricted

an

halfof an

(i)when

consideration,namely

from

)+

c",_,

{m"i)

0,

so

"-*""""

(2)
which
If

c,..--

is the

2.4...

value

(2m)

+ 5)
(2r + 3)(2?-

2ni +

(2r +

and

'

replacedby

are

1)

^,

take

we

"

obtain

we

when

givenby " 3'1

of Com

2''+*r(r+ |)'

the solution

,,Zomir (r
which

is

denoted
naturally

the

by

^)'

that the definition of

so
symbol Jr+ki^),

" 3-1 (S) is stillvalid.


If,however,

take

we

"

(o..l(-2;)
[Cm
As

before,Ci,

(m

m
"

c-^,...,c.2,_i

become

=0.

"

2r)+

"

c,"_2

all zero, but the

are

0
.

"",.+!+ Cor-i

0.

equationto

when
c.i,-+i',

this

f
-'"+'

\in"r

"

Froc.

t The

f^

(2r + 3) (2?-+ 5)
E. Irish. Acad.
cases

combine

xxvn.

to form

+
(27/i

) 2
.

(A),(1909),pp.
the

case

c.",-~iis

determine

0,

value of
equationis satisfied
by an arbitrary
is defined by the equation
and

Cm

(",^j^

(3)
/

determine

|,the equationswhich

"

in which

(2m

4
.

2/-)

157"161.
'2v is

au

integer.

"

r, c.."n+i

42

the

v=-r-l,

Jy{z) be defined by " 3-1 (8) when

If

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

solution

now

structed
con-

is*
Co

2-'--^r

It follows that

(1

r) J_r-i{z)+

2'-+^r (r +

|)/,+"(^).

in the definition of J^

modification

no

C2.+1

when

{z)is necessary

is that the
of the solution in this case
peculiarity
negativeroot of the indicial equationgivesrise to a series containingtwo
constants, Cq and Car+i, i.e.to the generalsolution of the differential
arbitrary
equation,

"{r

v=

\);

the

real

fundamentalsystem of solutions of Bessel's equation.

3"12. A

that,if yi and 3/2are two solutions of a linear differential


denote their derivates with
order, and if ?// and 2/2'

It is well known

equationof the
respect

to

the

if the

does

vanish

not

then, either
the two

one

second

the

independentvariable,then
Wronskian

if the

Wronskian

solutions vanishes

of the two
a

independent
linearly

are

determinant^

and
identically;

solutions is

solutions

does

vanish

else
or
identically,

identically,
the ratio of

constant.

If the Wronskian

does

vanish

not

then
identically,

any

solution

of the

differential equationis

Cg

constants

; the

in the form c, 2/1+ c^ y-z where Cy and


expressible
solution under
consideration
depending on the particular

solutions 2/1and
For

3/2are

brevitythe

then

said to form
of t/i and

Wronskian

We

now

being used
proceedto

M|yi,i/.},

it is necessary to

when

system.

y^ will be wi'itten in the forms

m^z[y"yo],
the former

fundamental

are

the independentvariable.
specify

evaluate

im[j.{z),j_.(^)}.
If

we

multiplythe equations
V, /__,{z)

and
by J^ (z),J"_^(z)respectively
which

may

be written

0,
subtract

V, J, {z)

the results,we

obtain

an

equation

in the form

o,
j^[zm{JAz),J-A^)]]
=

lu connexion

Sci. di

Torino, xxvi.

with

series

(1821),pp.

representing this solution,see

Plana, 2Iem.

della E. Accad.

delle

519"538.

t For references to theorems


concerning Wronskians, see
(" 23), p. 109. Proofs of the theorems
quoted in the text
74.
DifferentialEquations (1914),"" 72
"

Encyclopedle des Sci. Math. u.


given by Forsyth, Treatise

are

16
on

44

concerning Bessel
but

the

discussion
are

65"72

; XXII.

(1904),pp.

Trans.

Royal

Soc.

Math.

Soc.

The

be

of these

results

to

found

in

analogues is

of

80"85

of this work.

105"118,

(1905), pp. 1"28,

(2)ll. (1905),pp. 192"220;

of
generalisation

399-408;

Proc.

main

London

(2) III. (1905),pp.

function, obtained

Bessel

the

functions,

Jackson's

Edinhurgh Math. Soc. xxi. (1903),pp.


Edinburgh, xxv. (1904),pp. 273"276;

Proc.

1"23.

by increasingthe

of the series,will be dealt with

of the terms

of factors in the deuominators

basic Bessel

theory of

the scope

Royal Soc.

Proc.

Edinhurgh, xli.

obvious
sets

outside

series of papers,

the

analogues in

their

(2)I. (1904),pp. 361"366;

more

number

have

functions

[CHAP.Ill

FUNCTIONS

BESSEL

OF

THEORY

see
Cailler,Mem. de la Soc. de Phys. de
generalisation
in the shape of Bessel functions of two
Oeneve, xxxiv,
(1905),p. 354; another generalisation,
variables,has been dealt with by Whittaker, Math. Ann. LVii. (1903),p. 351, and Perfes,
170.
ComptesRendus, CLXi. (1915),pp. 168

in

"

In connexion

4"4.

with

this

"

General

3'13.
The

series which

closed domain
R

of J^,{z).
properties
defines

of values of

(v)" 0],and

\v\^

[theoriginnot

domain

bounded

in any

For, when

Jy{z)converges

and uniformly*in any


absolutely
being a pointof the domain when

of values of

v.

\z\"$ A, the test ratio for this series is

and

4Z

m{v
whenever

is taken

to be

m)

{m

in

root of the equation


greater than the positive

m--mN-l^-'
This choice of

N)

"

beingindependentof

Q.

and

z, the

result stated follows from

the test of Weierstrass.

Hencej-J^ (z)is an analytic


functionof z for all values of z {z 0 possibly
and it is an analytic
beingexcepted)
functionof v for all values of v.
=

An

is that term-by-termdifferentiations
of this theorem
importantconsequence
and integrations
(with respect to z or v) of the series for J^ {z)are
permissible.
An

due
inequality

to Nielsen

where
and

I^ i"
11/0+

This

1is the

result may

with

the

Bromwich,

t Modern

1 1,
"l,-^'^!,
-

Iko

of the numbers

+ l \}

+ 2|,|i/
+ 3j,
|i/
+ l|,|i'

be

function

z", which

is

as

factor of J^

"

2*11

"

(5);

it should

be

pared
com-

specifica{z),needs precise

Theory of InfiniteSeries,"82.

Analysis, "5"3.

3Iath. Ann.

(1902),p.

exp

here,namely

proved in exactlythe same


way
which will be given in " 3*3.
inequalities

the
Finally,
*

smallest

be noticed

% should

494.

Lii.

is.
(1899), p. 230; Nyt Tidsskrift,

(1898),p. 73;

see

also Math.

Ann.

lv.

3-13,3-2]
tion. We

BESSEL

define

it to be

TT

"

it is necessary

values of arg z,

3'2. Tlie

"

phase (or argument)

is

this range of
of the process to be carried out.

the function

will be made

J^{z)outside

formulae for J^{z).

of the
generalisations*

("2'12)are

as

for the Bessel

formulae

recurrence

efficients
co-

follows:

(1)

J.^,{z)-vJ.^Az)

(2)

J".-:{z)

^JAz)

2J; (z),

J.+,{z)

(3)

zJ;(z) + vJ..(^) ^J.-A^\

(4)

zJJ (z)

These

of

^w.

arg

to "continue"

mention
explicit

recurrence

Lommel's

the

exp (v logz) where


that
so

value
givenits principal

When

45

FUNCTIONS

vJ, {z)

of

are

the
precisely

being the

substitution

To

them,

prove

we

dz'

form

same

as

" 2"12,the onlydifference

the results of
number

of the unrestricted

for the

integern.

first that

observe

"

zJ,.+,(z).

^^'

dz

,Zo ^"^"^
/

00

\V{v+ni-rl)

"

\iti

^iv"l-\-2m

,"=o

When

we

the

differentiate out

2-^+^'".m!r(i.+ m)

the left,
we

producton

at

once

obtain

(3).

In like manner,
J

dz^

"^^^

CO

"

Vn

,2W

dz,r=o^^^'"'.m\V{v^m

l)

C_yft^2
V

-^ 2"+"'"-' {in -\)\V{v+

111

1)

00

C_y"+i 2'-'"+i
r

"r=^2''+'^'"+im!
.

whence

when

p. 533.

ilher die BesseVschen

is half of

an

odd

2)

-z-^J,^,{z),

(4) is obvious; and (2) and (1) may


(8) and (4).

Studien

(z/+

Functionen

integerby Plana, Mem.

be

obtained

by adding and

(Leipzig,1868),pp. 2, 6, 7.
della

R. Accad.

ddle

Sci.

Formula
di

Torino,

(3)was
xxvi.

tracting
sub-

given
(1821),

46

THEORY

We

obtain

now

can

the

[CHAP. Ill

FUNCTIONS

BESSEL

OF

formulae
generalised

[Z^J.{Z)\ Z^-^-J.-m(Z),
(^)"
[z-^J.(z)] (-)- ^-'-/.+".(^)
(^)"

(5)

(6)

when
by repeateddifferentiations,
obtained

Lommel

described

(1) has been

formula

The

his

results from

all these

has been

which
integral

positive
integer.

is any

in

of
generalisation

Poisson's

" 3'1.
in the construction

used*
extensively

of Tables

of Bessel functions.

/^_i (z) and Ji_^ (z)in


By expressing
and

(4),we

formula

of

(1) and (2)is that,if Q,,(z) J^ [z\ then


=

this formula

was

discovered

3'21. Bessel

imaginaryparts

real,have been discussed in

extended
subsequently

See

seq.

also

it,Studien
Math.

Neumann,

obtained

detail

where
J"^ifj,(x),
by Lommel J,and his

v, fx

and

results

were

by B6cher".

the
after defining
particular,
equation
11

real functions

by
K^^^{x)and Si,^^(x)

the

the results

" iS",A^)]+ {^^^M


(^)" iS,,^(x)}
^, {K,,^(x)

(1)

2(v"ifi) +
X

d
,j.

A",+i,^
(x)

(3)

'S^^+i,K
(^) 'S^.-.M
(^)+ -S^'Vm
(^)'

-o

dx

(2)

K,^^(x)+ K'\^^{x),

See,

Math.

Ann.

iv.

Math.

Ann.

ra.

"

Annals

IIThe

of

consequences

of the function

some

In

Lommel

et

functionsof complexorder.

real and

The

various
48

600.

(1871),p.

III.

by Lommel, who derived


(Leipzig,1868), pp.

Functionen

iiher die BesseVschen

are

^^.-i(^)-"?.+i(2)
y";(^);

(8)

Ann.

^~^^f^

(2) of "3-12.

interesting
consequence

Au

(z) and J'"^,


{z) by (3)

(z)
/, (z)/i_, iz)+ /_, (z)./,_,

(7)
from

formula

derive Lommel's

can

of J""

terms

e.g.

Lommel,

of Math.
for

reason

established

in

"

3'3.

Ahh.
Milnchener
xv.
(1884"1886), pp. 644"647.
(1871),p. 105. Some associated formulae are given
(1871),pp. 481"486.
vi.
(1892),pp. 137"160.
insertingthe factor on the right is apparent from

in " 3'G3.

formulae

which

will be

3-21, 3-3]
with
no

BESSEL

other formulae

numerous

great importance,and
in which

memoirs

In the

of this

solutions

in which

equationin

47

of like character.

consequentlywe
published.

they were

case
special

FUNCTIONS

the

results

merely refer

0, Bessel's

form

These

the

be of

to

seem

I'eader to

the

equationbecomes

of series

Boole*

given by

were

many

years ago,

3'3. Lormnel's
shall

We

shew

now

that, when

R{v)"

integral
of Poissons
-

cos

type.

^, then

(^
6)
f"TmvT)//os

J. {z)

(1)

sin'^'
6 dO.

cluded),
proved by Poissonf that, when 2v is a positiveinteger(zerointhe expression
the rightis a solution of Bessel's equation
on
; and
the delinition of J^(2)for positive
was
as
expression
adoptedby Lommel;]:

It

this

expression
of J^{z) hy an

was

values of

^.

Lommel
subsequentlyproved that the function, so defined,is a solution of Bessel's
formulae
of 5^3"2 ; and he then
generalisedequation and that it satisfies the recurrence
defined /" (i)for values of
in the intervals { h, -f), (" f,
#),(-#, -f),
cessive
hy sucof " 3'2 (1).
applications
v

"

"

...

deduce

To

the

the definition of

(1) from

of the series for

generalterm

Jt,{z)
adoptedin this work,
manner:
Jyiz) in the following

(-yjhzy

(-y^(^zy+'"'

r(i;+ -|)r(m

f^

we

form
trans-

^)

m\V{v

providedthat R(v)"
Now

when

\)

V{v+h)V{hy{'2.my:

R(v)^^,

\)

-l^p'-Hi-ty-i
(2r/i)!

^eoiiverges
uniformlywith respect to
be integrated
term-by-term; on
may
Tripos,

the series

Pliil. Trans,

"^.

n"=\

T{v

throughoutthe interval (0,1 ),and so it


adding to the result the term for which

of the Royal Soc. 1844, jj.239.

See

also

question set

in the

Mathematical

1894.

f Journal de VEcole B. Pohjtechnique,xii. (cahier 19), (1823),pp. 300 et seq., 340 ct seq.
the right
21/ is an
odd integer,the expressionon
Strictlyspeaking, Poisson shewed that, when
the
Bessel's
is
solution
derived
of
the
from
appropriate
equationby
multipliedby ijz a
equation
change of dependent variable.
J Studien

iiher die Boi^eVscheu

Fuiictioneii

(Leipzig,18G8), jip. 1

et seq.

48

0,

namely

f-^il

ty^dt, which

"

is convergent,

the

usingthe

more

terms

result stated

(2m)!

the

unaffected

the

analysisnecessary
elaborate.
The
simplestprocedure seems
omitted and integrate by parts,thus

m=2

jo

to

be to take

m=2v

establish

to

(2m)!

Vi^dt

by parts a
integrating
in
integration the second
on

of the

(which are

second

time.

line of

analysisis

On

series.

It follows that,when

-^^^'^

Obvious

adding

obtain

convergent),we

(2)

J, (z)

^^^

'^^^^^

equation is

the series with

0.

little

the first two

J^

^
2

/^ ;, {i-ty"-Kdt
j

i,"=2 (2m)!

result.

^, then

^^
C^"'*
r(z.l'i)r(|)
~

transformations

last

placeof

and
interchangeof the order of summation
of the uniformityof
permissibleon account
\
integrals
correspondingto the terms m
0, m

the

"

the

^ in

"

and

The

the desired

R{i")"

f=sin-^

substitution

by writingtt

ov-^^

convergence

that, when

r^'

(2m)!

by making

follows

integrandis

fact that the

"^"R(v)"h,

When

find

Ave

Jo

'^(""-r(.
i"^o
+ i)r(i)Jo'
whence

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

of this

^'

''''

'' ^^

result,in addition

to

"^^^
^^'^

(1),are

the

following:

r( n-^
p.,^^.^!fpp.(zt)dt,
1

cos

^*'
^'^""^ ^'^^
Wi:'^
Ijr(i)
"

r(.

^"^^

(5)

IJ''cos(^cos^)sin-^d^
"-^-(^)=rA^t)r(i)

(6)

J.(^)

The

formula

(7)

^*^^''
=

r(/^+i)r(^)

obtained

J, (2)
=

by

of (5),namely
partial
integration

^:;^^;;^YT I sin(2cos(9)sin^''-^5'cos^rf^,
1
1
(i;+ ^)

is sometimes

g,"cosegin2''^(;?6'.

useful ; it is valid

(^)

only when

R(v)" ^.

3-31]
An

with

BESSEL

FUNCTIONS

expansioninvolvingBernouUian
help of the expansion

49

has
polyn(juiials

been

obtained from (4)by Nielsen*

the

in which

(^"(^)denotes

the ?ith BernouUian

of
[Note. Integrals

polynomialand

izt.

the

studied before Poisson by Plana,Mem. della R.


type (3) were
Accad.
delle Sci. di Torino, xxvi.
(1821),pp. 519
538, and subsequently
by Kummer,
Journal fur Math. xii. (1834),pp. 144
147; Lobatto, Journal filrMath. xvii. (1837),pp.
"

"

and

363"371;

Duhamel,

Cours

d' Analyse,li.

(Paris,1840),pp.

substantially
equivalentto J^,(z),defined by
function,

x) =1
(jx,

(1

cos
v^)'*

V.V

118

the

"

121.

equation

do,

investigated
by Lommel, Archiv der Math, imd Phys. xxxvii.
(1861),pp.
problem of obtainingthe differential equation satisfied by

WAS

The

349

"

3G0.

convei'se

z^

[^"f''i,v-af-^
iv-^J-^ dv

126.
In
by Lommel, Archiv der Math, und Phys.xl. (1863),pp. 101
Inst.
also
Calc.
Int.
li. (Petersburg,
Euler,
integralsee
1769), " 1036,
der linearen Dijferentialgleichungen
Petzval,Integration
(Vienna,1851),p. 48.]

also discussed

was

connexion
and

with

"

this

3'31.

derived from
Inequalities

Poisson's

From

" 33 ((3)it follows that,if v

be real and

greater than

i, then

"

\I(z)\
^pj!ff|p(X)j^^exp
sin^^edd

I/. (2)I

(1)

By usingthe

formulae

recurrence

" 3*2 (1)and (4),we

deduce

in

similar

that

manner

l-A'(^)K^||l
|^i)|}expr/Wl(."-i).

(3)

"

^
cos
By using the expressionf{2/(7r^)}2

(1) is valid when

"

character

Math.

can

Ann.

Analysis," 7'2;
t The

reader

general theorem
w.

B. F.

be obtained

lix.

in

(1904),p.

Nielseia
should
will be

uses

have

for

(z)it may
./_j

be shewn

that

^.

-^hese inequalities
should
obtained in "3-13. When
v

integral.

inequalities
compared with the less stringent
of a more
is complex,inequalities
complicated
of
portance
but they are
no
the same
great immanner,
be

108.

The

notation

used

iu

the

text

is that

yiven iu

formal

proof of

Modern

different notation.
no

in verifying this
difficulty

result.

more

given in " 3'4.


"*

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

50

of Poisson's integral.
Gegenhauersgeneralisation

3-32.

formula
integral

The

C^" (t)is the coefficient of

in which

ascendingpowers
R(v)"-^ and n
reduces

to

In the

of

is due

a,

0, 1,2,
integers

When
...

2at +

a^)-"in

is valid when

formula

the

Gegenbauer*;

to

is any of the

Poisson's

expansionof (1

in the

a"

0, it

obviously

integral.
in which

case
special

the form

assumes
h, the integral

TV

(2^
j j ^'''
'

Jn+i(z) {-iT

(2)

""'

'

^" (^"^^)^^^ ^ '^^

'

equationhas

this

To prove

and

J^n {z)

times

(_

-y.V{v

and

reader

^^^'

''"^-^^'
ft""*"

the

will find it instructive

to

case
special

establish

(3) by

\.
induction

with

the

aid

of the

formula

ci^^(o=(2.+.o

-^2)^?^.

"c/ (0- (1

Wiener

Milnchener
pp. 365

^"

Gegenbauer'sequationis

of

(n+1)

memoirs

'

hjri-lv+n)

is evident.

givenby Rayleigh"in

recurrence

"*'"

dt-

f.f('
v"^t)

^^^

symbolicform

The

the form

have

'-*"(-)

was

in
integral

^^~^'^

Gegenbauer'sresult

this

take Poisson's

n-\-\)V a) jJ

Viv

we

studyby Whittakerf.

that:|:

(")

detailed

the result is

by parts;

it is known

whence

subjectof

Gegenbauer's formula, we

n
integrate

Now

the

been

"

lxyii.
Sitzuvgsherichte,
(2),(1873),p. 203; lxx. (2),(1875),p. 15. See also Bauer,
di Mat. (2)xii. (1905),
0. A. Smith, Giornale
v. (1875),p. 262, and
Sitzungsberichte,
The
studied
373.
function
has
been
extensively
by Gegenbauer in a series of
C^" (t)
in the Wieiier Sitzungsberichte; some
of the more
important results obtained by him are

given in Modern
t Proc.

London

t Cf. Modern
" Proc. London

Analysis, " 15-8.


Math.

Soc.

Analysis,"
Math.

xxxv.

(1903),pp.

198-206.

15 '8.

Soc.

iv.

(1873),pp. 100,

263.

See

"" 6-17, 10-5.

52

TPIEORY

[CHAP. Ill

FUNCTIONS

BESSEL

OF

and so the
periodicfunction of yjr,
integrandis an integral
be taken to be a and a + 27r,where
with respect to yjr
may
integration
This follows from Cauchy'stheorem.
(complex)number.
arbitrary
NoAv

the

limits of
a

is

an

thus get

We

I'^r

-'^'

JJru)^

6 sin
e^'^^'^'^'^^^'^cos-''-^

dd^lrdd

Ja
7rr(z^).'o

ii^Z / I
""

"

i^+a)
e^^sinecos

cos^--!

e sin

Odyfrdd.

7rr(i/)Jo.'o
We

define

now

OT

so

the

by

cos

pairof equations
Z

"

cf), IS- sin

cos

sin cf),

that

Jv{^)=

t,/
ttI

[i{Z"

exp

(j/)Jo Jo

iz sin (^sin ^ sin

sin dcos-dr(i")

cos

^ sin

cos-"-!

this formula

between

onlydifference

The

J^ (tij-)^\^

is in the form
with
analysis

the

"

"

i{Z~

^ sin

cos^"-!
i/r]

cos

Od-^dd

are

cos

(j))liz sin (^.m,

cos

n
"f))

cos

"^)cos

"

iz sin

"

"/"I,
.

iz sin

"

sin 6.
"^cos -y^

is
expression
iZ

\,
{v) j oJ 0

cos

that the result of

ttI

the formula

factor ; and we now


the steps of the
retrace
exponential
the steps are
factor. When
modified form of the exponential

i{Z

so

(9c^'v|rcZ^.

of the

i{Z

last

[m sin 6

exp

retraced the successive exponents

The

and

^]

exp

iz (cos^

"

6 + sin

cos

"^sin 6

cos

"/r),

the steps is
retracing

\iZcos 6

"

iz

(cos(^cos 0

sin rf)sin 6

cos

"\|^)]

sin2''-ii/rsin-^''^rf"/rr
and

consequently
Gegenbauer'sformula
[Note.

this

The

device

section,to evaluate

Sci.,1789, p. 372, and

The

3"4.

We
when

hy

shall

ineans

using transformations
seems
integrals

definite

of
to

Poisson, lle'vi.de I'Acad. des

polarcoordinates,after
be due to Legendre,M^m.
Sci. iii. (1818),p. 126.]

the

of

manner

de I'Acad,

des

expressionof J"("j^i)
{z)in finiteterms.

the importanttheorem
that,
integral
is halfof an odd integer,
in finite
the function
terms
J^iz)is expressible
and trigonometrical
of algebraic
functionsof z.
now

It will appear
is not

of

is established.

so

recondite

deduce

later

from

("4-74)that, when

expressible
; but
character

Poisson's

than

of

course

the theorem

this

has not

theorem

converse

which

is

such

now

about

value, then J^,(z)

is of
to

be

much

more

proved.

3-4]

BESSEL

Solutions

[Note.
by

earlywriters

various

1765),p. 76 that

equation satisfied by

the

[1820],
pp.

solution of the

of differential

in finite terms

equationsassociated

with

tained
ob-

"/" (2)were
"

it was
observed
by Euler, Misc. Taurinensia,iir. (1762
in finite terms; while
equationfor e""/^^j(2)is expressible
"

solved
was
t/j,^.i.(2)

2-

53

FUNCTIONS

in finite terms

by Laplace,Conn,

des

Terns,

and

v. (Paris,
MecMnique Celeste,
1825),pp. 82"84
; by Plana, J/em.
Torino,x.x.vi. (1821),pp. 533
534; by Paoli,Mem. di Mat. e
di Fis. {Modena),xx. (1828),
and also by Stokes in 1850, Trans. Camh. Phil.
pp. 183"188;
Soc. IX. (1856), p. 187 [Math, and Phys. Papers, II. (1883), p. 356]. The
investigation
be given is based on
the work
of Lommel, Sttulien ilber die Bessel'schen
which
will now
Functione7i (Leipzig,
1868), pp. 51
56.]

1823

245"257

delle Sci. di

della R. Accad.

"

"

It is convenient

then, by

when

" 3-3

to

be

to

and
integer(zero included),
positive

(4),

integrateby parts

we

degree 2n, tiie process


To

restrict

2" +

times; since (1

times
r

"

polynomialof

terminates.

then

culated
observe that if d''{l
we
"t-y^/dV be calexpression
the
from Leibniz' theorem
(1 + ^)^
only
by writing(1 t^y^ (1 ^)"'
which
does not vanish at the upper limit arises from differentiating
n
the other factor
the factor (1"^)", and therefore from differentiating
times; so that we need consider only the terms for which r'^n.

simplifythe

last

"

term

is
(-)""

"

'

2-"~'"

Hence
dV

and

I 92*1"

d'-(l-tT

similarly

{-y-

df

i=-i

Cn

"

(2H-r)

It follows that

(i^"n+l
sjir

'hn (^)

i"r+l 22"i-""^. !

2w

S*

(-)"+!e"

z''+\{r-n)\ {2n-r)l

,.=",

and

,="^'--i.(r-")K2n-r): j'

hence

(1) /n+j(^)

,':,rl{7i-r)l(2zy'^^
,7, rl {n r^

his result may

(2) /,n+i

V(27r^)
be written

in the form*
"i"

"=(~)^ (z
sin

^ /?7r)S

"

,.0

(-)'".
(n+2r)!
(2r)!(n-2r)!(2^y^'-

"k{n-i)

cos{z

"

luTr)
}"=()

compact

Soc. Sci. de

method

of

Bruxelles, xxix.

(2^)'_

obtaining this formula

(1905),pp.

110"143.

(-Y,(n+2r+l)l
(2r + !)!("is

2?--

given by de la Vallee

1)1{"Izy'+\
Poussin,

Ann.

de

la

54
In

have

we
particular

the former

the

hence, from
n

.^

sm

formula

recurrence

(^~
j (^^

J|(^)

^,

of these results is also obvious

Again,from

and

(^"
j

/j {z)

(3)

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

cos

series for

the power

from

5J;
Jj iz).

have

we

(1),

jr'^A?i+ ry^_

,^0 r!(n-r)!(2^)'-

(- ^")'-'^
(7?+ r) !

_.^

,ro rl{n-r)l(22y
"

But, obviously,
by induction

we

can

express

\zdzj
as

polynomialin 1/zmultiplied
by e*'^,and
",..

("iY-''(n

12

r)\

,"

f d

"^,

have

must

we

so

\" e"''2
z

would
for,if not, the precedingidentity
e'^ "j),
(z)

e-''

Avhere ^j {z)and

lead to

result of the form

4"o{z) 0,
=

in Ifz; and such


are
polynomials
^o ("2)

an

is obviously
identity

impossible*.
Hence

it follows

I
r=o

thatf

t'-^.Q^+ r)!
! ("
r): (22)'-

(-tr-"-(n

.^

,"0

! (n

r)!

r)! (25r)'-

{-T{2'jryz-^^(^^^\
{-y(2'jTz)KI_"_,(z).

Consequently
^

(4) J^n-l{z)=
*

Cf.

V(27r^)L

^'"+'^(" +

olZ

/");

"
.

,=0

! (n

r): (2^)'-

Hobson, Squaring the Circle (Cambridge, 1913),p. 51.

t From

the series

-i^'
it is obvious

that

r(i) ,"tom!i.f...(m-i)'
J

(-iy+'\{n

r)l'

,lz)

-i

"

\TrzJ

cos

z.

,.ror

! (?i
-

r)! {2z)

3-41]
and

BESSEL

FUNCTIONS

55

hence

(5)

J_,,_i
{z)

("2r)!(n-2r)!(2^

lo

kTTZ

(^ + ^ /"7r) i

sin

"

(2r +

r .=0

In

we
particular,

1 ) ! (rt
-

2r

1) ! {2zy^^\

have

J_i(.)=(-Vcos^,
./_3,(^)=(iy(-^^^_sin^
\7rzJ \

(6)

"

\7r^/

We

have

half of

an

The

expressedin
integer,
by means

now

odd

explicitexpressionof

numerical

results contained

(1873),pp.

3*41.

any Bessel function,whose order is


of algebraic
and trigonometrical
functions.

number

of these

letter from

functions

Hermite

be

can

written

Gordan, Journal

to

down

fur

Math.

from
Lxvi.

303"311.

Notations

Functions
branches

in

finite terms

forfunctionswhose

of the

order is

types J"(n+h(z) occur

halfof an

with

such

odd

integer.

frequencyin

various

of Mathematical

Physicsthat various writers have found it desirable


them
to denote
by a specialfunctional symbol. Unfortunatelyno common
notation has been agreed upon and none
of the many
existingnotations can
be said to predominateover
the others. Consequently,
apart from the summary
which will now
be given,
the notations in question
will not be used in this work.
In his researches
Sac.

Royal

CLViii.

surrounded
vibratingsphei-es
by a gas, Stokes,Fhil. Trans, of
and Phys. Papers,iv. (1904),
made
use
(1868),p. 451 [^Math.
p. 306],
on

the
of

the series
"*"

2.imr
which

is annihilated

by

the

opei'ator
d'^

?i(n + l)

^.

dr^

This

denoted

series Stokes

+"""'

2.4.(mr)2

by

dr

the

;-^

and
symbol f^ (?")

he wrote

e
(r)+ "S';
v/." "S'"
i'"'7"
( r),
e"'"'/;,
-

where

S,^ and

;SV/are

zonal

surftxce harmonics

so

that

\/^"is annihilated

by

the

total

operator
dr'

dr-'

dr

"

and

by

the

n(n-"r\)
;-

partialoperator
c2

{
.

.7)}

Math. Soc. iv. (1873),


by Rayleigh,Proc. London
Soc.
LXXii.
[Sckmtifc
(19Q3),
again
Royal
pp. 40"41
pp. 93"103, 253"283,
that
trivial
Ruyleigh
change
Papers, v. (1912),pp. 112"114], apart from the comparatively
would have written /" (ijnr)
where Stokes wrote f^ (r).
In this notation

Stokes

and

was

followed

Proc.

56
lu

order

obtain

to

Sn =(-)""*"!
Sn in the

and
analysis,

of his

course

origin,Rayleigh found

finite at the

solution

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

to take

it necessary

then

"/'" (-0"^^m,SJ"

j J,,+ ^(mr).

twrj

IX p

"'^/'''^
(i^^"

and

that

[e
-j^^

J^^^(r)

{271+ 3)

in his earlier papers,

(1884),pp. 139"149;
pp.

519

pp. 486
Math.

xxx.

^^"(^)
=

it is evident

T{n

"

"

/" ( ir)].
-

of the

found

types e+''"/"
(" t")which

in his treatise

Soc.

(1882), pp.

xiii.

51

f\

\n,
f"
o
+
(-)"i-3.5...(2,.
K

"

-^T^i-'^"+j(^)
found

it convenient

to

also Proc.

Hydrodjaiamics and
the notation f

on

^" iz)

"

modify
London

this

ingly
notation, and accord-

Math.

Eayleigh,Phil.

Phil. Trans,

while

149

"

(1909),pp.

Trans,

161

665

has been

736, and

"

V."(2)
=

Cn {^)
=

it is

the
certainly

best

to

Papers,
A, (1904),pp. 87"110
[Scientific
t
he
symbol
f,i{z)
by Xn{^)- Love,
rejjlace

A, (1915),p.

ccxv.

'

CCIII.

it convenient

Soc.

notation

of the Royal Soc.

found

""J'

5)

,7+r^

^" (2)
=

Chemie,(4) xxxvii.
(1912), p. 29
191
this
notation is
(1913),p.
;

and

(1901),

xxxii.

-^^^

of the Royal

yet another

XXVIII.

Soc.

*-

-"

(1912) pp.

^^^^

f '^Y

l).(^_^J
^.
-,\

he used

zdzj

V.

xv.

"

(d
while

212;

"

(1883),
; Phil. Trans, of the Royal Soc. CLXXiv.
A, (1906),
by Eayleigh, Proc. Royal Soc. lxxvii.
(1912), pp. 300
312],and by Love*, Proc. London

120"150

that

189

66;

"

^"^^^^1.3.5...
+ 3)'''2.4(2?i
+ 3)(2w +
(2n + l)L^~2(2?i

SO

to write

it convenient

that

%)

however.Lamb
Subsequently,
pp. 11"20,

^"

(1899),pp. 308"321.

this notation

With

e"-

27"43

followed

was

Papers, v.
[Scientific

499

"

Soc.

he

2.4.(2?i + 3)(2?t+ 5)

Math.

(1885),pp.

xvi.

and

549;

"

London

Froc.

+
/" (ir)

Lamb
origin.

finite at the

solutions

requiredfor

i" +

ir

"

for the combinations

notation
simjile

^^

'

In order to have
are

3-4 that

"

It follows from

112

omitted

the factor

"

)" and

wrote

by Sommerfeld, Ann. der Physikiind Chemie, (4)


pupils,namely March, Aim. der Physik mid
Ann. der Physik und
Chemie, (4) XLI.
Rybczyiiski,

used

of his

two

and

(Wj,+.(.)=.-i(^-^-^J
"

{hTrZ)^
+
iJ_
[J,,
+, {^) ( )"
-

adapted

for the

on
investigation

(Z)],
electric

which

waves

was

the

subjectof their researches.


\

"

(d )
"J-

stated,

he

t This
p. 82 ;
not

modified
is

nearer

the definition
the notation

except that Heine

in his

memoir, Journal

defined

e~^^
"

but, as

in bis later work.


used

by Heine, Handbuch

(z)to be twice
i^,j

fiirMath.

lxix.

the

(1869),pp.

der

i. (Berlin,
Kugelfimctionen,
1878),

expression on
128

"

141.

the

right in his treatise,but

3-5]

BESSEL

notation

Sommerfeld's
used

and

Vn

v" +

is
in

)"nv"

modified form
slightly
of
and f,,
place \/r"
; see

light,K. Danske Videnshahernes


(1898),pp. 405"502.]

second solution

57
of the notation

Selskahs

of

3*5.

FUNCTIONS

his memoir

used

by

L.

Lorenz,who

reflexion and

on

refraction

i.
Skrifter,
(6)vi. (1890),[Oeuvresscientijiques,

order.
of BesseVs equation
forfunctionsofintegral

("3'] 2) that,whenever v is not an integer,


a fundamental
system of solutions of Bessel's equationfor functiong^joforder v is formed by
the pairof functions ./^{z) and J_^ {z). When
is an integer(= n),this is no
v
of
the
relation J_n {z) (")'*
J^ {z).
longerthe case, on account
It has been

seen

It is therefore necessary to obtain a solution of Bessel's equationwhich is


linearly
independentof J",{z);and the combination of this solution with Jn{z)
will

fundamental

give a

The

system of solutions.

which

solution

the full details of the

will

be

now

constructed

involved
analysis

obtained

was

in the construction

by Hankel*;
firstpublished

were

by Bocherf.
An
his

constructingHankel's solution was discovered by Forsj'th


;
on
generalmethod of Frobenius,Journal fiirMath, lxxvi.
(1874),
235, for dealing with any linear difterential equation. Forsyth'ssolution was
in his lectures on
differential equations delivered in Cambridge in 1894, and it
been published in his Theory of Differential
Equations, iv. (Cambridge, 1902),
Treatise
and
in
his
on
Differential
102,
Equations (London, 1903 and 1914),

alternative

method

is based
jjrocedure
214

pp.

"

contained
has

since

pp.

101

"

Chajjtervi.

note

of

the

1.

It is evident

that, if

negativeor zero),the

be unrestricted,and

if

be

integer(positive,

any

function

J,.{z)-{-YJ_^{z)
is

solution of Bessel's

vanishes

when

functions

equationfor

of order

; and

this function

n.

so
Consequently,
long as

i^

n,

the function

J,{z)-{-TJ_Az)
V

is also

solution of Bessel's

assumes

undetermined

an

We

shall

equationfor functions
n.
form| when v

and this function

li"/,W-(-)"/-.(^)_
V-n

r-*n

and

we

shall shew

Math.

Ann.

t Annals

it is

that

(1869),pp

r.

of Math.

vi.

solution

of Bessel's

equationfor

functions

of

469"472.

(1892),pp.

See

85"90.

also

und

filrMath,
Niemoller,Zeitschrift

Phys.

(1880),pp. (;5-71.
X The

essence

the equation when


the

of order

evaluate

now

XXV.

"

integer n and

of Hankel's
p

is not

which

has

is
investigation
an
a

integer,which
limit when

v-*-n.

the
assumes

construction
an

of

an

undetermined

expressionwhich
form

when

is

satisfies

equal to

58
order

and

independentof Jn (2);
linearly
solution required*.

that it is

be the second

to

[chap. Ill

FUNCTIONS

BESSEL

OF

THEORY

It is evident

so

that it may

that

dv
as

"

dv

of the differential coefficients exist f.

since both

v-^n,

be taken

Hence

Mz)-{-YJ..{z)

lim

exists; it is called
To

Bessel function

it
distinguish

from

"

second kind of order

of the

which

functions

other

are

n.

also called functions

of

kind it may be described as HankeVs


function.FollowingHankel,
shall denote it by the symbolJ "^ni^)so that

the second
we

(1)

lim

{z)

Y"

and

also

(2)

dJ,{z)

Y,(^)

It has

be shewn

to

now

Since the two

_,_y,dJ-,{zy
^

dv

that Y,; (z)is

functions J""

dv

(/=?l

solution of Bessel's

functions
analytic

(z)are

equation.

of both

and

order of

v,

the

and

differentiations on J^.^(z) with respect to z


v
performingpartial
of indifference ". Hence
the
of
the result of differentiating pair
equations

is

matter

with respect to

When

may

be written

1-^

J-;
dz^

-o

dv

dz

combine

we

^
dv

{z-

V-)
^

"

"

the results contained

dJA^)

IvJ

;r-

dv

in this formula,

(z)
/

"^
^^v

U.

find that

we

_^_y^^^J-A^y
2v{jA^)-{-rJ-A^)],
=

dv
*

The

reader

limitingform

will realise

+ See "3'1.

X The
function

is

It is conventional

in many

symbol
equal

that,given a solution

of this solution

coefficients while
Of course,

dv

to

solution
to write

parts of the

to

are

differential

\Yith

written

variations in v are not


theorj',
actuallyused by Hankel,

was

I/tttimes Hankel's function

" See, e.g. Hobson, Functions

of a Real

equation, it is

corresponding limitingform

differentiations

differentiations with respect


which
r,j(2),

of

of the

respect to

as

obvious

that

equation.

total differential

partialdifferentialcoefficients.
contemplated.
as

is used

("3-54).

Variable

not

of the

(1921),"" 312, 313.

in this work

to

denote

60

(^)

is

term-by-termdifferentiation

the result of

and

y%^)

[chap, iti

FUNCTIONS

BESSEL

OF

THEORY

log(1^)_ flog r(l.+

1)

7/1

.-

{m !)2

m=0

derivate
logarithmic

is customary, the

t/tdenotes, as

where

+ l)},
{logi^-l/rO"^

Gamma-

of the

function *.
Since

('"+ !)"

0"V'

when

"

bj using
may
The convergence
replacedby m.
concerninganalyticfunctions.

(1)

Yo(^)=2

(2)

Yo(^)

S^

See Modern

\^]J

Yo (z)

The

solution of Bessel's

\'\J ^{m

1)

")

1)

I-

that
1

is

be noticed

J^^-y^T^
|i

reader will observe

5"3.

to

V'''

log(1.)}/o (^)

of

{iog(i^)-^(,,,+l)},

m=0

(8)

Yo(2)
is
\//-(m-|-l)
theorem
general

series for

series in which

consequence

Analysis,"

expansionare

of the

convergence

immediate

an

S/
log(i.)../o-(^)-

the

...

ratio-test for the

is also

of the

forms
following

The

2, 3,

m=\,

D'Alembert's

be estabhshed

(log2

Yo (z)+

7) /o (^)

functions

equationfor

of order

The

zero.

expansionof

this function is

VW!)

m=0

This

function

Neumann,
But

adopted as

was

the canonical

obtained

was

Euler's result in his

as

notation

own

2'^

kind of order

44; see "


equation,long before,
by

of Bessel's

is that the

rii\

of the second

function

1867), pp. 42
(Leipzig,

solution

-"

(m \y

m=l

Theorie cler Bessel'sche?i Functionen

the series

(-)-(|^n

"

generalsolution

of the

zero

by

3'57.

"

Euler

f.

equation

xxddy + xdxdy + gx'^ycx^=0


6Ag'-

2Aq
IS

22Ag^

"

100.4/

,.3"

.t'4"
+ etc.

''

n?

1
.

8?i5

27w^

1.8.

9'

9'

,,.2)1
_

nn

'^
a

"

.3?"+

Analysis,Ch.

-^^^.r-"
1.4h^"

7l?l

Modern

xii.

^(1):
where

7 denotes

Euler's

t Inst. Calc. Int.

11.

4?r

1.4.

"P''
"

%i^

1.4.

"7,

^(m

l)=

27

a;3"+
9?i"
1.4.9.

64?i9

9^

.r*"-etc.

.i;3"
+
1.4.9.

16?i8

that, when
+

-+...

Ua'

16?i8

^"_etc.,

is

positiveinteger,then

+--Y,

....

(Petersburg, 1769), "977,


i.

It is to be remembered

constant, 0-57721.57

(1781) [published1781], pars

Mathematica, pp.

pp.

2.33"235.

186"190.

See

also

Acta

Acad.

Petrop.v.

3-52]

BESSEL

and

wliere A

in the first line

numerators

548

22

100

16

gave the

...

We

shall

Y,i(^),where

successive

11

548

7.22-9.6,

25

1 00

etc.

in

rii-CTm

"

ascendingseries

an

and

the

definition
of

Hankel's*

expansionof the more


generalfunction
positive
integer.[Cf.equation(4)of " 3'5.]

obtain

now

determine

!)
'^-^,
ml
7nJ

1 ) a-m

(2'"+

expansionof Yn(^)

The

3*52.

to

the formula

evidentlyexpressedby
O-m

100

3528

2(1
^
\1

If
is

followinglaw

5.6-4.2,

22,

this law

61

3.2-1.0,

He

arbitraryconstants.

are

FUNCTIONS

is any

It is clear that

"Zo ^v [m ! r (i'
+

dv

1)

(_)".(1^).+.

X
^om.^.r {v +

when

where

v-^n,

is

I)

III

^^^

'"

positive
integer.That

'dJA2)

^)

(1)

i-f'{hzY+-'''{\

is to say
1

cv

evaluation

The
of

(i/r

I'

at

the series for J_^

break

in the

{z)into

and

in the former

part

0 ^

We

;"=H7/i!r(-:/

l)

l)'

r (v

in) sin (v

"

"

m)

TT

^^

m+l) 1)

"

0,l,% ...,n-\.

pole

replace

we

Now, when

iii

of the

parts,thus

T(-v

for which

terms

two

low! r{-v

tedious because

little more

\dJ_Az)ldv\=nis a

of

TT

n,
'

~d
dv
=

r (:^ 7/0sin (v
((|^)-''+-'"
-

JJf='

TT

[{^z)-''+'"'r(v-m)
yfr(v
{tt-i

m) ir]

"

sin (i/-m)7r+
III)

r (w
(l^r)-"+27n

m)

cos

cos

(v"m)

tt

-tt-^

TT.
(/I III)
-

3/ai/t. Anil.

i.

(1869),p.

471.

log(^z)sin (i/-?")7r}]"="

BESSEL

OF

THEORY

62

[chap. Ill

FUNCTIONS

Hence
"n+2m

n-i(-)nr^n-m)(^z)
2

dv

\.//l: ,_.,{-loga^)
!
w^!
+

("

=n

"/r(-7?

m)

1)1,

that is to say

(2),

az;

"(

x[log(-|-^)-A|r("i+l)}
when

replace by
m

we

??

?7i

have

combining (1) and (2) we

On

series.

in the second

Hankel's

formula, namely
(

00

'"'
m

,"=o

{7+ log(1^)1/,,(^) (1z)- f


-

(-)'"(l^)"+^^^il1

In the first term

{m

0) of

the last summation,

so

...

111

1)}

^^^~^^;r^^'
(i^)""
1

the

in {| is
expression

-.

Lommel*)
(following

%{^)-

(4)

7n)\

-+1

convenient
frequently

\n+im,

,111

11

It is

{n

{2log{^z)-^lr{7n+l)-ylr(n +

ml

'

\m

to write

-J^{z)\ogz,

dv

that
n

'^"^^^

(5)
when

negativeinteger,
^^ (z)is defined by
right.
is

the

on

We

Y" (z)

2 J,

complete solution

contained
936

expression

of

(z)]ogz+ %, {z)+ {-r 3-n {z).

+ ay
-j-^^

was

given in

logarithmicfactor) by Euler, Inst.


of this equation are

Cede.

the form

Int.

li.

of

series

(partof

which

(Petersburg,1769),"" 935,

solutions

x^ Ji (2ai^2)j ^4 Yi
Euler
this

the limit of the

thus have

(6)
The

+ l)}
{log2+ 'v/r(j;4-??l

nZ^miriWm^l)

also gave

(2aia;i).

{ibid.
"" 937, 938) the complete solution

oi xi

Q);solutions
-j^-\-ay
=

equationare

x^ Ji (4a2 ^i), .^i


Y2
*

Studien

ilber die BesseVsehen

Functionen

x^).
(4a*
(Leipzig,1868),p.

77.

of

3-53, 3-54]

BESSEL

The

3-53.

is

the

function

integer.The

an

values of

odd

of the second

of the

when

this statement, observe

prove

VTT

J^ {z)cos

p-n
sni

I'lr

have

Y,, [z) + lim

Y" (z),

when

evident

is
equation,
has any

(i)v

integer
The

function

second kind
v

is

vw

J^z)

Y,(4

Y^(^),defined

value for which

should

reader

function

either

form
by (1) or by the limiting
equationfor functions of order v both
is not an integer,
and when (ii) is an
combined
with " 3*5 (3).
equation(2)

2v

type) of order v; and

avoid

2,

be careful to observe
of

v,

is continuous

The

i'

except when z is zero; and, in


Y" (2),as v^^n, uniformly with

n,

Weber-Schldfi
functionof the

("3"5o)

Sept.

Journal
function

Ann.

Journal

dated

function

of the

is any

Ann.

fur
1872

filr Math.

i.

slightly
by Weber
the
failure
producedby
is half of an odd integer.

which

i' and

given by

was

in order
Schlafii^'

of the

definition when

to

the

(1869), p. 472.

Math,
and

(1873),p. 9; Matli.
respectively.In a
10-5, dated
(1873),pp. 75

lxxvi.

Oct.

lxxv.

of the second
(U Mat.

kind

second

modified

was

or
negative.
integer,
positive

second kind.

the inconveniences

Math.

of the

that,in spiteof the change of form, the


at

order of the function


*

Bessel function

the definition fails only when

Y^(2) approach their limits /" (2)and


except in the neighbourhood of 2
0, where n

definition of the

Hankel

"

fact,Jt,{z)and

The

"

J^y {")'

"

Y,, (^),defined in this way, is called

Y^ (2),qua

3*54.

J_^ (z)'

"

solution of Bessel's

(of Hankel's

The

respectto

completein
by the limit

integer.

an

Note.
function

that

the latter result follows from

is defined

an

"

(" )"cos

limY.(^)

of that

restricted
un-

proved that

(2)
It is now

for

is half of

failure is

vtt
v

VJT
v

we

its

that

J^,{z)cos

{-)" lim

so

Hankel*

when

The

function

lim

(3)

cos

and

onlywhen

it is easy to reconcile this definition with

Tre""^
lim Y^

integerand

an

case, the

the rightand
on
expression
definition of " 3*5.

To

adopted by
excepted)is
was

vanishingof sin 2p7r.

of the
the

is

but, in the former

case

defined

sin 2v7r

definition fails both

because
integer,

has been

"

'

the latter

kind

definition which

values of 2v
(integral

(1)
This

63

of ^,(z).
definition

Hitherto
order

FUNCTIONS

1872

"

kind

Ann.

paper

May

vi.

(1873),p.

written

1872, Weber

(see "" 3'57, 3-58).

(2)vii. (1875),p.

17 ; this paper

is dated

Oct.

148.

few

4, 1872.

These

months
had

used

papers

before

are

these,

Neumann's

64

THEORY

The

which

function

adoptedby Weber

was

J^ {z)cos

VTT

sin
is

symbol K,

so

formula*

by the

J_y (z)

"

viT

integer).

an

^tt; and
that,with his definition,

Schlafli,
however, inserted
the

the canonical function of the

as

of functions of the firstkind

in terms
second kind is expressible

(or the limit of this,when

[CHAP. Ill

FUNCTIONS

BESSEL

or

his function

denoted

he

factor

by

^^^J.^(^)cosz^^-J^)_

^^

sm

VTT

Subsequent writers,however, have usuallyomitted this factor ^tt,e.g. Graf


and Gubler in their treatise f,and also Nielsen, so that these writers work with
Weber's

function.

The

by
symbol K is,however, used largelyin this country, especially
denote
function
to
different
Bessel
of
3"7),
a
Physicists,
("
completely
type
and so it is advisable to use a different notation. The procedurewhich seems
to produceleast confusion is to use
the symbol Y^,{z)to denote Weber s function,
of NielsenJ,
after the manner
and to adopt this as the canonical function of
the second

in

kind, save

order
integial

instances

rare

the

when

the insertion of the number

saves

ir

function

of

in certain formulae.

thus have

We

JA^)^o^v'--J-A^)

n ^ {z)
^

(1)
^
^

cos^

(2)

lim

F,.{z)

[Note.

Schlafli's function

pp. 85"90,

and

pp. 23

Sehafheitlin

30.

"

the

Tre""^'

Sehafheitlin

which

Mathews

Y" (.).

been

used

Math.

(1892),
(1895),
and Heaviside
Weber's
function
with the sign changed,so
use
we
(with Nielsen) denote by Y^ {z) is written as
Yy (z)by
Gn (z)by Heaviside||.
7i)as
Annals

of

by Bocher,

Math.

viii.

of

Annals

(1894), pp. 57"61;

vi.

ix.

-.

" and (when

Gray and
symbol Y".

J^i^)^^^^'^-J-^i^)

has

by McMabon,

function

smyTT

that

of Hankel's

use

sometimesH

Weber's

use

function,and

they

denote

by

it

the

Weber's
definition was
by an integral(see " 6-1) which is equal to this expression; the
expression (with the factor hw inserted)
was
actuallygiven by Schlafli.
+ Einleituiujin die Theorie der BeHsel'schen
Funktionen, i. (Bern, 1898), p. M et seq.

% Nielsen, as in the case


Y" [z),Handbuch

of

index, thus
are

obvious

objections to such

" See, e.g. Journal


Bessel'schen
IIProc.

change
^] A

functions, writes

Theorie

der

the

number

indicatingthe

order

Cylinderfunktionen (Leipzig,1904),p. 11.

notation, and

we

it for the

reserve

obsolete

function

as

an

There
used

by

("3-58).

Neumann

other

der

Funktionen

fur

Math.
cxiv.
(1895),pp.
(Leipzig,1908).

31"

44, and

other

Royal Hoc.
in

papers;

also Die

Theorie

der

liv.
(1893),p. 138, and Electroma"inetic. Theory, ii. (London, 1899), p. 255;
sign has been made from his Electrical Papers, ii. (London, 1892),p. 445.

Treati'^e

on

Bessel Functions

(Loudon, 1895),pp.

65

"

66.

3-55]

BESSEL

FUNCTIONS

Lomrael, in his later work, used Neumann's

function

iiher die BesseVsclien Functi")7ien

in his Studien

65
of the second

kind

1868), pp.
(Leipzig,

85

"

(see" 3'o7),but
86, he used the

function

Yn {z)+ {^l^{n+^)
+ log2} Jn (z),
l-rr
F" (z) is the function

where

(" + i) makes
\|/-

of the term

presence

of Weber.

Julius, Archives

complicated;see

One

the

disadvantageof
formulae

recurrence

x.xviii.
iVeerlandaises,

this function

is that

for the function

(1895), pp.

221

much

the
more

225, in this

"

connexion.]

Heine's definition
of the functionof the second kind.

3"55.

definition

The

givenby Heine* of the function of the second kind possesses


some
advantagesfrom the aspect of the theoryof Legendre functions ; it
of Mehler's formula ("o'71),namely
enables certain generalisations
Km

P" (cos din) /" {O),


=

expressedin a compact form. The function,which Heine denoted by


in terms
the symbol Kn {z),is expressible
of the canonical functions,
and it is
differs only
equalto "\irYn{z) and to
|Y"(2); the function consequently
in signfrom the function originally
used by Schlafli.
to

be

"

The

of Heine's

use

the function

function

{z). In this form

it Gn

This

function

revival of the

(which will
cosine

been

has

of Heine's

use

function

become

and

sine;
the

to

tabulated first by AldisJ


extensively
Beports,1913, 1914 and 1916.

unfortunate,both
distinctly

seems

kind

and

pointsof resemblance
adoption|jof Jn{z)and 6-'"
(2)as

Vu (z)present many

so

use

the Continent
many
years ago
in their treatise t, and they term

the

that
of

cosz

and

-^tt sinz

as

other than that

senses
symbol G^
writers,namely Heaviside, Proc. Royal Soc. Liv. (1893),p.
Dougall,Proc. Edinhuryk Math. Soc. xviii. (1900),p. 36.

two

and

canonical
It

is

also be

(aswas

the
the
parable
com-

pointed

justexplainedby

138

and

of the fact

functions

must

account

on

account

on

to the relations between

canoniciil functions.

{z)has been used in

that the

out

also

apparent in Chapters vi and vii) that the relations between

more

./"(z)and

on

Mathews

of the second

of functions
existingmultiplicity

functions

and

subsequentlyin British Association

Airey", and

of the

the

died out

to have

seems

used by Gray
occasionally

was

stated in

at least

" 3-54),

treatise has been

pointedout by Morton,
signof y in formula
had apparentlybeen copied
" 3"51 (3)supra. It was also stated by Morton that this error
noticed by GrayH) J. J. Thomson,
by various other writers,including(ashad been previously
and
Researches
in
Rec"iit
Electricity
Magnetism (Oxford,1893), p. 263. A further error
Note.

Nature,

An

error

in

signon

(1901),p.

lxiii.

Haiidbuch

.i Treatise

Proc.

der
on

p. 245
the error

29 ;

of Heine's

is

equivalentto

Kugelfanctionen, i. (Berlin,1878),
Bessel

Functions

pp.

change in

185

"

the

248.

(London, 1895),pp. 91, 147, "242.

(1900),pp. 32
RoijulSoc. lxvi.
(6)xxii. (1911),pp. 658"663.
the historical point of view there
"

43.

" Phil. Mag.


IIFrom
and
has

also
not

for
even

11 Nature,
W.

B. F.

function

usiug Neumann's

is

but

something
Heine's

to be said for

function, being

using Hankel's
more

modern

function,

than

either,

this in its favour.


xlix.

(1894),p.

359.
5

THEORY

66

noticed

Heine

hj

called

for

The

which

formulae

those which

satisfied

are

(1)

p. 248

on

to

confusing notation employed

most

same

satisfied

are

by J^ (z); they are

the

form

same

follows

^Y,{2),

Y,_,(2)+Y^^,(2)
=

"

by Y^ (2)are of
as
consequently

F,_,(^)-F,+,(^) 2IV(^),

(3)

2Y:{2)^vY,{2)

(4)

zY:{2)-vY,{z)

formulae

in these

function
To

function

the

2\\_,{2),

-zY,.^,{z\

be

may

replacedthroughoutby

the

Y.

them

prove

" 3'2 (3) and (4) in

take

we

[z^J",{z)]
=

^
if we

(J^o ^'^'o)-

(2)

and

the function

symbol Ko to denote
W
symbol Kq denotes

the

nses

the

formulaefor V^ (2)and Y^ {2).

Recurrence

recurrence

to be due

seems

of his treatise Heine

p. 245

on

^TTJ'oin this work, while

3-56.

as

work

in Thomson's

Morton

by

[CHAP. Ill

FUNCTIONS

BESSEL

OF

z^

multiplythese by

/,_! (^)

cot

vir

[z^/_, {z)\

and

cosec

and

vrr,

then

/_,+!{z);

-z-

the forms

have

subtract,we

[z^F, {z)]
=

dz

(3) follows

whence

at

z"

F,_, (2),

Equation(4) is derived

once.

in

similar

from

manner

the formulae

{z-"J, (z)]
=

By

addition
The

and

formulae

but since F^

replacev by

z-"

subtraction
are,

{z)and

proceedingto

the

so

./,+!
(z),
of

[z-^/_, {z)]
=

j^

(3) and (4) we


the

far,proved on

its derivatives
limit when

obtain

are

tends

to

(2) and (1).

hypothesisthat

continuous
an

./_,_,{z).

?-"

functions

is not
of

integralvalue

v,

integer;

an

the result of

n, is

simply to

n.

Again,the effect of multiplyingthe four equationsby 7re'"'''sec vir, which


is equalto 7re'''='=^'"^
the functions Fby the functions
sec
{v " l)7r,is to replace
Y
throughout.
In
Studien

tlie

case

of functions

iiher die BesseVschen

instructive
Neumann's

to establish

them

of

integralorder, these formulae


(Leipzig,1868), p. 87.
such functions directly
from the

Fimctionen
for

connected
investigation

with

the formula

were

The

given by Lommel,
reader

series of

"

(4)will be discussed

will find it

3"52.

in

" 3-58.

68

It

shewn

was

typefor 7'"'(z).

of Poissons
integral

The

3-571.

by

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

that

Poisson*

gixcosa,
^Qg (a;sin- co)do)
Jo

is

equationfor

solution of Bessel's

an

associated

The

identified

with

if

so,

Ave

the

"(-)"/," (-2)
^
=

of summation

order

TT

at

once

F""

(1)
from

which

The

{z cos 6) log(4 sin^ 6)dO

we

2?i~i

observe

of the

cos27i^

that, since

Parseval's

the definition of

integral
("2 2) and

['"cos
{z

6) log{iz sin- 6) dd,

cos

of summation

and

non-uniformly convergent
(

integrationhas
^

near

To

0.

)"Jin {^)lnis convergent, it follows from

(-)"^2"(2)/'i= lim

a-*-l-0

Journal

associated
Cours

a-".l-0

TT n=l

y 0

Abel's theorem

Theorie

ii.

(Paris,1840), pp. 122"124,

and

difficulty
% that

,a"cos2n^

cos

{zcos

6)'

de. VEcole

d' Analyse,

examined,

this

B. Polytechnique, xii.
(cahier19), (1823), p. 476. The
partialdifferential equation had been given earlier {ihid.p. 227). See

(1857),pp.

(-)"a" J2"(2)/"= hm

2
n

be

to

now

overcome

'"

2
n=l
1=1

result is obvious.
order

is

./0

the formula

{z)

Neumann's

change

because

with

obtain

rh
cos

this result combined

de

{z cos 0) z

cos

TT

F'"'{z),
we

be

can
integration

"cos2/?6'

/"*-

from

and

that

2 2 ^~^n

analysisby
given here, with

Jo

nir

that

assume

deduce

changed,we

the

of detail.

in matters

and

F""'(z); and

function

the

have

" 2'2 (9) we

From

of

this result is of sufficient interest to be

slightmodifications

some

(z sin 6) log(4^^cos^ 6) dO

cos

by Neumannf

Avhich he obtained

argument

integral
-

was

and

zero

in the form
of the integral
expression

Stokes obtained an
subsequent!}^
ascendingseries (see" 3'572).

and

of order

functions

Spitzer,Zeitschriftfur

solution
also

Math,

,.

ctd.

und

of

an

Duhamel,

Fhys. ii.

16.5"170.
der

fiir Math,

BesseVsr.hen
imd

Cf. Bromwich,
:J:

Functionen
(Leipzig,1867),pp.
Fhys. xxv. (1880),pp. 65 71.
Theory of InfiniteSeries," -51.
"

45

"

"9.

See

also

schrift
Niemoller,Zeit-

3-571, 3-572]
Now, since

is less than

1,

FUNCTIONS

69

(a"cos 2n6)/n does

with

integration(by comparison

of

range

BESSEL

2a"),and

uniformly throughout the


converge
the interchangeis permissible
; that

so

is to say
2

""
/'*'"
/ cos (2cos 5)

n=i

fk"

2
2;i(9
d6"-

cos

/iN

I
^

'^

Hence

22LW

^_^^

n=i

If

--

cos(scos^)log('l
-2acos2^-|-a2)c/(9.

lim

ijjj^

proceed to

now

,"

do

have

we

We

(scos ^)

1
=

a"cos2?i(9

"
,

cos

^.^,g

^^log (1

(_ ^jQg

2a

2(9+ a'-i)
C^5.

cos

that*

shew

"

cos(scos^){log(l-2aco.s2^

0.
a2)-log(4asin-^)}(;(9

a-*'l-0.'
0
It is evident
and

that

26l+ a2-4asin2

1 -2acos

log(1

so

bound

Hence, if A be the upper


I

cos

^) {log(1

(sCOS

2a

2a

2^ -j-a-)
^

cos

t of jcos

(1 -af^O,

log(4a sin^ 6).


6) \when

{zcos

0^6

^ ^tt,

have

we

log (4a sin'^6)]d6

-2(9
+ a2)

cos

^ .4 I

{log(1

2a

cos

26 + a'-)
-

log(4a sin^ 6)]d6

"

a"

1.2
/AnI

cos
"-'^"

-^"

"1

log(1/a)

log(2 sin d)\dd

^^

"=l

2/(^

*7rJlog(l/a),
term-by-term integrationbeingpermissiblesince

1.

a"

Hence, when

\ cos {zCOS 6) {log(1 2a cos 26 + a^) log(4o sin'-6)\d6


Jo
a-"-l -0; and this is the result to be proved.
'

as

"i7r/llog(l/a)^0,

1,

a"

Consequently

1
i (-)"'^2''("^)=
lini
_

^
TT

and

'

[-"cos(2COS^).log(4asin2^)c

"''cos(2cos(9).log(4sin2^)rf^.

,,^

justified.
interchangeis finally

the

reader

The
combined

with

"

3"5

integral.
d^y

differential

The
m

is

Stokes

stated

solution

equationconsidered

This

constant.

is Bessel's

(presumably with

f
The

t If
+

p.

value
is

Trans.

42.]

by

Stokes

equation for

reference

1850

functions

was

of order

Poisson) that it

to

^
was

dy

.,

^-m-i/

zero

known

and

0, where

argument -im:.

that the

general

was

J
*

integralfor J^ (z)

(5).

Stoles' seriesfor the Poisson-Neumann

3-572.

Poisson's

this result from

to deduce
will find it interesting

of this limit

real,^
Camh.

""

i^c+D log {zsin2 6)}cosh

was

1 ; if not, A
Phil.

Soc.

(mz

cos

6) dO.

ix.

assumed

by Neumann.

{z)\\
exp
Papers,
(18-56),
p. [38]. [Mathematicaland PhijsicaJ

{]I

iii.

(1901),

70
It is easy to

^n {CThe

notation,the value of the expressionon

that,with Neumann's

see

series

logz)J" (imz)+ 2Z)

WiC+D

it is

(,^'i; J
"

dd,
it may

be

that

order
kind, of integral

Bessel function of the second

The
Neumann*

of F""'{z)by induction

in terms

(1)

is
this

^if^

nY^^'^ (z)

of the

formula

recurrence

...

^)}.

from

by

the formula

^F"'^+i)(z\

type

same

defined

n, was

as

"2'12(4).It

is evident

equationthat
F(") (^)

(2)

(- ^r

F'"'(z)satisfies the

Now

of F""' {z).
definition

Neumanns

3-58.

from

forroula fi-om which

recurrence

logsin

i^(l^
2#^|i7rlog2
-j^%os2"^logsin^c^^

which

right

equal to

cos^'^0

-j^r-^

obtained

by parts, Stokes
and, by integrating

Stokes

by

,1=0

deduced

the

i^W iimz).
log(4m)} Jo {imz)-^\tTD

expandedinto

expressionwas

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

(4
(^)""")

equation

and, if we

apply the operator-f

-"r-

theorem,

and

we

equationn times, and

use

Leibniz'

get

so

This

equationis at

once

reducible

(4)
and

this

to

so

to

V, F"") {z)

F'"' (z)is

Again,(3) may

solution of Bessel's
be written

0,

equationfor

functions

of order

?i.

in the form

Theorie

its order

t The

der BesseVschen

is not

an

Functionen

(Leipzig,1867), p.

integer.

analysis is simplified
by taking iz-

i,

zdz~

so

that

_d
d^'

51.

The

function is undefined

when

3-58, 3-581]

BESSEL

FUNCTIONS

71

that

SO

whence

obtain another

we

(5)

\^

When

combine

we

formula

recurrence

/iFe^){z)

^F"-^'

(1) with (5) we

at

(2).
the other

deduce

once

recurrence

ormulae
F'"-i' {z)+

(6)

F'"+') {z)

F'") (z),

(7)

F""-i' (^)

F'"+"

satisfies the
ConsequentlyF*"' (^")
and

Y" {z). It follows from

" 3-57

F"") (^)

(8)

(^)

same

r/T''""iz\
^ -*
.

formulae

recurrence

i TT F" (^)+ (log2

iY" (^)+ (log2

7) ./,(4

und

The

which reduce.s"to
integral,
Zeitschrift
fur Math,
by Spitzer,

definite

expansionof F'"' {z).

Neumanns

3*581.

Jn{z\ F"(2)

7) /, {z)

solution of the equation Vh (y) 0 in the form of a


integralof J^3'571 when n
0, has been constructed
iii.
(1858),pp. 244-246; cf. " 3-583.
Phys.

the

as

(2) that

of the
generalisation

been

" 8"57 (1) has

formula

givenby Neumann*;

it is

s"

To
the

2m)

111
where

(-)-^ {n

+
9

+"""""""'

establish this result,we

^o

^-

first define the

functions L^

(z) and

Un (2) by

equations

(2)

Z.,(.)

/. (.)log.

(3)

U,,(z)

.,

/,.{z)+

^-~^, ^
j^ ;^^

/"

(^),

SO

that
We

F")

Zo (^)

(z)

shall prove

(z)
(4) X"^.:

and

then
*

Ln{z) and

the
t^^C^)satisfy

L,: (z)+ injz)L" (z), Un^,{z)

(1) will be evident by induction

Theorie

die BesseVschen
and

that

f/o(^).

der

BcsseVschen

Functionen

Functioncn

(Leipzig,1867),

1868), pp.
(Leipzig,

Haentzschel, Zeitschriftfiir Math.

from

82"84

xxxi.
uiidPIii/.^.

recurrence

formulae

U,: {z)+ {niz)U,,(z\

" 3-58 (2).


See

p. o'i.

Otti,Bern

also

Lommel,

Stmiien

Mitthellungen,1898, pp. 31"

(1886),pp.

25"33.

iiher
3") ;

THEORY

72
that

It is evident

dz\ z'' \~^^^dz\ z'" ]


Jn^.(z)log .

z"

-^

dzl

z''

J
=

Sn

-"-^

("Y\m

*"

Z"^,(Z)+ Un^,(Z)

dz

z'^

since the

induction

and

that it vanishes

The

The power

the

equation(1) is therefore
sefies for

Ln jz)+ Un {z)

{"""-"(Z)

rightvanishes
values
for all integral

expressionon

the truth of

3'582.

z^

0, it is evident

when

of

Hence

??.

by

established.

Un(z).

Un(z),which was defined in " 3'581 (3)as


has been expressedby Schlafli* as a power
coefficients,
coefficients,
namely

function

establish this result,observe

series of Bessel

series with

simple

in:{n + m)i

1)1=0

To

(" + m) /",.",,(.)}

J^ "^(;^:^ {^^Jn,.m-.(z)
"

-"

z''

have

part of (4) is proved. It follows from " 3'58 (2) that

the second
y"^+'"(Z)

and

dz\

m{n^m)

,^=1

"i^^^

"

^n

and

J^___|l+___^^_^^^

(4) is proved. To prove the second part, we

''dz\z"'

(z)

^^"^^^^^

"

the first part of

,,t^{n-m).m\dz\z'''^-'^

^"+i

f ^^^
^,in-m).ml

-Jn^.{z)\ogz +

and

[CHAP, III

FUNCTIONS

BESSEL

OF

that it is true

when

7i

by " 3"51 (3) and


expressionon the

the
differentiation,
|3"57(1);and that,by straightforward
formula as that of " 3"581 (4) for Un (z);
recurrence
rightsatisfies the same
equation(1) is then evident by induction.
Note.
of

It will be found
2w

(|2)"

in the

to establish this result by evaluatingthe


interesting
expansionon the rightof " 3-581 (3).
*

Math.

Ann.

in.

(1871),pp. 146"147.

coefficient

3-582-3-6]

BESSEL

reader will

The

(2)

easily
prove

now

%. (z) {7
=

(3)

F"" {z)

(4)

47rYn{z)

3 '583.

The

different from
observe

L"(z) +

to

Un {z),

of

obtain

y(")

"3-571

{z).
for

F(")

(2) was

generalised by Lommel,

(Leipzig,1868), p. 86,

Lommel's

^ly^^
j

J, (z)log.=

so, from

sin d) cos^"
(.-

cos

^-^"

with

result in Neumann's

that, by difierentiation of Poisson's integralfor J^ (2),we

^
and

%, (z)+ {log2-y]J, (z),


%,(z).

Functionen

Neumann's;

formulae
following
-

Ln {z)+

formula

73

log2}J, {z)

fiber die BesseVschen

Studien

the

of Poisson^s type for


intagral

Poisson- Neumann

The

FUNCTIONS

notation

notation, we

ratlier
first

have

{log(^cos^ d)-^{,

J)}dO,

" 3-582 (3),

r^fi^WMf

'"

^'"'
and

^")
=

(1)
hence, since \^(^) \//=

r(")

(1)

{z)

^y

^.

log2

"

^'''''' ^) ^'^""'^ '1"S^^'^^^

^'^^

it is to be remembered

coefficients and

powers

of

(^sin ^) cos2" 0 log(4 cos^ 6) d6

COS

Jn (2)+ /^"(2),
{^ {n + h_)-y\r
(i)}

that Z"

as
(2)is expressible

finite combination

of Bessel

z.

of the

3*6. Functions

the formula

'"

,1,

in which

have

log2, we

(n + 1) y} c/^ + Ln (z),
"/'

third kind.

theory of Bessel functions,especially


those which are based on Hankel's researches (ChaptersVI and Vll)on integral
and asymptoticexpansionsof J^ {z)and Y^ {z),two coiiibinarepresentations
of frequentoccurrence.
tions of Bessel functions,namely J^.{z)
are
"iY^,{z),
in the theoryof "Bessel functions
also present themselves
The combinations
of purelyimaginaryargument" ("3"7).
In

developmentsof

numerous

consequentlyseemed

It has
functions

Jv{z) "iY^{z)

as

the

desirable

standard

to

regardthe pairof
and he
Bessel's equation,

Nielsen*

solutions

of

to

functionsof the third kirid; and,in honour of Hankel,


The two functions of the third kind
Nielsen denotes them b}^the symbol H.
defined by the equations
f
are
them

describes

"^"^"

is

Since
in which
*

\^j

the functions enter


det K.

Ofversigt oier

t Nielsen

uses

der

the

Vaiiskc

same

and
linearly,
Vide.nskahernes

have

the
Y^{z) satisfy

" 3-54 (1),we

with

tsm^TT

J^{z) and

J,,(z) iY^ (z).

sides
the right-hand
integer,

an

hucli der Theorie

Hf^ (z)

J.,(z)+ iY, {z\

combined
these definitions,

From

When

l'^
(z)

(1)

as

"

are

to

recurrence

be

ismi/TT

replacedby

their limits.

formulae

3-56),
(""3-2,

since the functions of the third kind


Selskabs

(Leipzig,1904), p.
Cylindcrfunktionen
11/ [z).
symbols Hi''{z),

Forhandlinger, 1902, p.
IG.

125.

Huiui-

[gHAP.Ill

'

74
linear functions

are

these

same

Hence

we

at

can

it follows
of J" (z)and Y^ (z),
coefficients)

(withconstant
formulae

recurrence

dH^^\z)

".

rf^

"

satisfied

are

"'-1^

"^'

,,^

.,.

dz

.;r,(2)

^a)."^

(6)

that

of the third kind.

functions

dHf\z)

,,.
^

by

the following formulae

write down

once

FUNCTIONS

BESSEL

OF

THEORY

.^^^-.jy^i)
^(^"(.),z^^^^-rH''\z)
-zH%(z).
(.)
-.

dH^!;\z)

dnf{z)

a)

,,.

Hm

Note.

266 ; (6)xiv.
(1897),i").
Eayleigh on several occasions,e.g. Phil. Mag. (5)xlhi.
iv.
[Scientific
Papers, (1904),p. 290; v. (1912),pp. 410"418],has used the

(1907),pp. 350"359

denote

symbol D^ (z)to
3'61.

the function

the
connecting

Relations

It is easy to obtain the


function in terms
of functions
that

of the formulae

some

which

Nielsen

calls

three kinds

followingset

(2)

^iriH

[z).

of Bessel functions.

of formulae, which

of the other two

kinds.

express each
reader will observe

The

simply the definitions of the

are

functions

the

on

left.

H^H^)+ ^f(^)
r,

,-,,

^^^

Y.^z) -YA^)

'^''^'^-

+ e-^''^Hfiz)
e^^^H^\z)
F-,

^^^

"^-A^)-

Y
.,x
^^^^^^

.ox

^"^^

...

F, (^)
'

sirTW

_/..(^)-/-.(^)cos^7r

Yi

'

_e^''^Hl'^{z)-e-^-'Hf\z
",

^^

g-j^^^

^r-.(^)-e--"r,(^)
HW^s_J-A^)-e-'"''JA^)
sin

sin

T/TT

jim,.^e^"'JAz)-J-A^)
"

(7)

From

^tt

'

H^^\z)-Hf\z)

/.(^)cos.7r-J_.(^)

"

(6)

VTT

sin;;7r

i^-.l^)-

(5)

(^)cos

cos

^STT^tT

^Y.,{z)-e^-iY,{z)

isinyTT

(5) and (6) it is

H^]l(z)
=

obvious

e"-

'

i/tt

sinvTT
that

^i^^(z),

?;(z)
=

'

e"

7/f^ (z).

76

" 3-54(1) that

It is clear from

m,

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

[J,(z),n i^)]
=

cosec

_2_

Wi

vir

[J.(z),J_. (^)l

"

irz'
result is established

This

of

hypothesisthati^

be

integeror
system of solutions.

siderations
con-

Hence

not.

an

integer;but

an

r, (^)l 2/(7r0),

mi{JA^\
v

is not

that

shew
continuitj^

(1)
whether

the

on

F^ {z)alwaysform

Jv{z) and

mental
funda-

that

It is easy to deduce

ra{/.W.Y.Wl

(2)

jjjj^,

and, in particular*,

im[Jn{z),^n{z)]=2lz.

(3)
When

we

Y^ (z),it is found
Wi

(4)
so

that

of the

third kind

of /"

in terms

{z)and

that

[H?(z),hT

functions

the

the functions

express

of the

solutions for all values of

{z),F, {z)]

C^)}=-^i(m{J.
third kind

also form

ii/iTTz),

fundamental

system of

v.

Math.

Soc.

formulae

connected

(1) and

(3) have been given by Basset, Proc. London


ential
they are readilyobtainable by expressingsuccessive differcoefficients of J^iz) and
peated
JV (s) by reYt,{z)in terms of J^"(2),JJ {z),and r^(2)',
the earlier ones
differentiations of Bessel's equation. Basset's results (of which
are
frequentlyrequiredin physicalproblems) are expressed in the notation used in this
work by the followingformulae :
Various

with

(1889),p. 55;

XXI.

(5)

J, (z)IV

(z)- 1\ {z)JJ' (2)=

(6)

j; {z)} v

(--) yj {z)J," (z)

(7)

J. (z)IV" (2)

(8)

j; (z)IV" iz)

(9)

J."{z)JV" (z) Vu'iz)J.'" (2)

(10)

J, {z)}v-)(-') y. (z)

(11).

(^)
j; {z)r,(iv)
(z) JV {z)^.("")

J-2,

^)
(^1
n (2)J.'" (2) ^.
('^-1)
r; (z)J.'" (z) ^2(I"
1)
=

'

'

A (1
-

J^^Kz)=^A,
(1

Throughout

these

formulae

1\ may

be

multipliedby -sini/7r; and J^, Yy may


expressionson the rightare multipliedby

be

Cf.

Lommel,

Math.

Ann.

iv.

2i/2+3 ,\
+
V
^-

/v*+ Uv'^

(^^4

replacedby J_"

are

if the

replacedby

expressionson

the

right

throughout if the
S^^\
ZT^^^^

2i.

(1871),p. 106,

and

Hankel, Math.

Ann.

vni.

(1875), p. 4-57.

3-7]

BESSEL

An

associated

Math.

Ann.

formula, due

to

FUNCTIONS

Lomtnel*

77

Math.

Ann.

(1871), [".106, and

iv.

Hankel,

(1875),p. 458, is

viii.

J,{z)Y,,^{z)-J,,^{^z)l%{z)^-\.

(12)

irz

This

is

proved

Bessel

3'7.

iu the

differs from

which

in

occurrence

way

as

"

(7).

32

functionsof pwely imaginaryargument.

differential

The

same

equation

Bessel's

equationonlyin the coefficient of y, is of frequent


it is usually
Physics;in such problems,

problemsof Mathematical

desirable to present the solution in

J^ (iz)and ./_^(iz)or J^,(iz)and

real form, and


unsuited

Y^ (is)are

the function e~-'"'^''


J^

However

is regardedas

its phase,
not with reference
of the function
of arg z,

J^ (iz)would

it

by

the

systems

for this purpose.

(iz)is a real function of

of the equation.It is customary to denote

When

the fundamental

which

is a solution

symbol /^ (z)so

that

convenient to define
complex variable,it is usually
the
of
value
to
principal
arg iz,as the consideration
but
with
value
reference
to the principal
suggest,

that

so

(- vr " arg z^^ir),


(/" (z) e-i"-'J" {zif\
\I, {z) e^'"^*'
J\,(ze--^'^'), (ITT " arg ^ " TT ).
=

introduction

The

"

argument
that

of the

symbol /;,(z) to denote "the function of imaginary


It should be menuse.
tioned
f and it is now in common
of
before
the
Basset's
work, NicolasJhad
publication
years
of the symbol F^,{z),
but this notation has not been used by

is due to Basset
four

suggestedthe

use

other writers.
the Continent
on
as
compared
positionsof Pare and Applied Mathematics
are
remarkably illustrated by the fact that, in Nielsen's standard
the second
solution K^iz), which will be defined
treatise ", neither the function Iv{z)i
nor
in
of
their
in physicalapplications.
is
mentioned,
even
spite
importance
immediately,
The

with

relative

this

The

country

function

I-"{z) is also

solution

of

(1),and it is easy

to

prove

(c"

"3-12)that

^(3)
*

Lommel

m
gave

the

[L,{z\ /_. (.)}

corresponding formula

for Neumann's

^-^^
.

function

of the second

kind.

first publishediu 1880.] Basset,


(1889),p. 11. [This paper was
f
but he subsequentlychanged
order to be t+"J"((^),
of integral
in this paper, defined the function
it, in his Hydrodynamics, ii. (Cambridge, 1888), p. 17, to that given iu the text. The more
definition is now
recent
universallyused,
+
Sci. de VEcole
Ann.
norm.
supplement, p. 17.
S2ip. (2)xi. (1882),
der Cylinderfunktionen
der Theorie
(Leipzig,1904).
" Handbuch
Fvoc.

Camb.

Pliil. Soc.

vi.

78
It follows that, when
a

fundamental
In the

of functions

case

by
The

system of

the methods

function

an

order,a
integral

of

""3*5

of

lim

K^ (z)

(4)

solution has to be

second

structed
con-

3"54.

"

will be

Kn (z),which

solution,is defined by the

second

[CHAP. Ill

FUNCTIONS

the functions 7^ (z)and /_^ (z)form


integer,
solutions of equation(1).

is not

BESSEL

OF

THEORY

adopted throughoutthis
equation

work

as

the

^-//_,(2)-/,"
V

"

equivalentdefinition (cf " 3'o)is

An

It may be verified,
by the methods
the order v is equal to n.

Macdonald*,

by

the

solution of

defined, for unrestricted

K^,{z) has been

function

The

" 3'5,that /r" {z)is a

of

values

(1)when

of

v,

by

equation
K,{z)

(6)

^7r^-''^''^~^''^.

Sm

it may
and, with this definition,

I^TT

be verified that

Kn{z)=\miK,{z).

(7)
It is easy to deduce

from

h! (iz) iTTie-*'"''
H[,(iz).
^TTie-^-""'

K, (z)

(8)

(6) that

physicalimportanceof the function Ky(z) lies in the fact


solution of equation(1) which tends exponentially
to zero
z-^^
as
values. This fundamental
positive
property of the function will be
The

that
x"

it is

through

established

in " 7-23.

^,,(2)is due to Basset,Proc. Camh. Phil. Soc. vi. (1889),p. 11, and
his definition is equivalentto that given by equations(4) and (5); the infinite integrals
by
in "i^6'14, 6:15. Basset subsequently
which
he actuallydefined the function will be discussed
The

definition

modified
18

pp.

"

order

In

Gray
omit

19, and

and

to

his definition of the function


his final definition is

obtain

Mathews

the factor

of

function

in their

1/2",so

equivalent to

which

work, A

Hydrodynamics,il. (Cambridge,1888),

in his

-"

r-,

satisfies the

Treatise

that their definition is

on

-^^

same

^-^
"

recurrence

Bessel Functions

formulae

as

/^ (z),

(London, 1895), p. 67,

equivalentto.

ira/_^)_a7j^)-|
The

of this definition to functions

only simple extension

of unrestricted

formula

K^
*

Proc.

(2) in- cot


=

London

Math.

I/TT

{/_^(2)-/^(3)},

Soc.

xxx.

(1899), p. 167.

order

is

by

the

3-71]

Analysis," 17'71)but

(of.Modern
it vanishes

whenever

formulae

is

1v

will be used

function

been omitted

although it

function

factor is not

undesirable

so

those

disadvantageof

that it would

it is

the serious

as

A"^{z);

have been

not

the
satisfying

connected with

"

advantageousifa

K^,(z)are

not

of

common

occurrence.

Iv{z)and K^(z) analogousto


Bessel functions. The proofs
ordinary

formulae
for the

for

left to the reader.

are

^L (Z),/C-x (Z)

(2)

/.-,{z)+ L^, {z)

(3)

zlj (z)+ vl, {z) zl,_,{z), zK: {z)+ vK, (z)

(4)

zi: (z)

/C+i (^)
=

vh (z) zl,+,(z), zKJ {z)


=

/C-i {z)+ /C+,{z)

1: {z\

vK, (z)

^^ (')'

2K: (z),

zK,_, (z),

zK,^, {z),

[z^K.{z}] (-)-^''-^._,"{z\
{z^L(z)] ^"-'"/.-m{z),(^^)"
(^)"
=

\zdz) \

s"

z^-^'^^

(8) I^n{z)

\zdz) \

'

K: (z)

(7) I,'{z)=L(z),

id)

factor ^n had

/"(^)and Ky{z).

/.-:(Z) L^. {Z)

The

rence
recur-

the

(1)

^^^

same

change,and the presence of the


factor in Schliifli's
function
corresponding

to make

the presence of the


of /^(i)and

now

of the formulae

5)

that
disadvantage

but in view of the existence of extensive tables

inadvisable

now

give various
constructed in ""3*2 3*6
shall

the

linear combinations

3'71. Formulae
We

suffers from

integer. Consequentlyin this work, ^lacdonald's

has

the definition of

from

(" 3'54)because

79

/^ (s).

as

of Macdonald's

"

this function
odd

an

of formula (8)shews
inspection

An

FUNCTIONS

BESSEL

formulae
integral
following

are

'

Z^^'"

K, (z),

(^zY

liijJL

onlywhen
cos

R(v

6) sm'''dde

e^^^o^^sin-"Odd

^ifll"
=

valid

I'^osh
(z

r7"XTVr7rx

K_,{Z)=K"(Z).

Iniz),

lAz)

Z^

/'''cosh
(z cos d) sin^'edd

[ (1

cosh (^0 dt.


t-y-'-

/,

h)"

80

These

results

due

are

InH (^)

(10)

(-)"-(n+ r)!

,%r\{n-r)\{2zr

sj{'lirz)

I
(^i 4- 7'")

^^

{Z)
I-(n+h)

(11)

(-Y(n + r)l
:^
z
^(2-rrz)
I r=-orl{n-r)l(2zy

,^orl(7i-ry.{2zyy

(n + r)\

(^Ve-S
2z)

^",,(^)

,.ZQr\(n-r)\(2zyy

e^

^^

(12)

Ill

also have

We

to Basset.

[chap.

FUNCTIONS

BESSEL

OF

THEORY

'

r=Qr\{n-ry.{2zy

(18)
K. (A

(14)

^i^"
f (m

{m
{logi^z) \y\r

log(iz) 7. {z)+

(15)

2,"=o

1),

m!(^^)"-""
/'l9\n+2jn

00

m!(M

,=0

Hi)!

1)

^^/r(n +

)},

+ 1

7n

"

Z'o(^)

(16)

e^''"^0

{log(2^ sin^ ^) + 7]dO,

(17)
K, (^e"-')

(18)

7C {z)

e-'"-'

iri ^^'^"^ /^ (^),


sin

(19)

5ie{7.,(2),
A^(^)}=-1/^,

(20)

/. (z)K,.^,(z)+ /,+,(z)K^ {z)


integralinvolved

The

integralsinvolved

memoir

"Zur

The
pp.

Theorie

"

in

been

(9)and

the

der Nobili'schen

recurrence

formulae

have

used

by

he added

Soc.

been

Hertz

yet another

in which

notation

^=0;

Math.

Soc.

discussed
der

he

an

odd

those

described

by

"

in his

Riemann

Physikund Chemie, (2) xcv.

also discussed

Camb.

the

Phil. Soc.

(1899), pp.

xxix.

(3) II. (1920),pp. 8

Dissertation, 1880
to

(cf." 3-572).

Stokes

series in (14) were

order is half

in his Berlin

by

given by Basset, Proe.

of Japan,

of this type whose

discussed

Farbenringe,"Ann.

19; by Macdonald, Proc. London

Functions

and

(16)has

"

Aichi, Proc. Phys. Math.

were

in

139, in the specialcase


series for /q(2).
130

(1855),pp.
power

1/z.

The

I/TT

110

"

ascending

vi.

(1889),

115; and

integer,as in equations(10)and
{Ges.Werl-e,1. (1895),pp. 77
"

in

"

by.

19.

341.

(12),
91

J;

3-8]

BESSEL

class of functions which

functions of order

be defined

may

(1 )
x

ments

we

ber

{x)+ i bei {x) /o {^iVO

ber

Hence

we

first examined

were

by

Thomson*; they

ber and

bei denote

definitions

I^,
(^vVO,

real functions.

expressedby

For

complex aro-u-

the formulae

(z)" i bei (z) ./"(zi\I"i) h{z^J" i).


=

have

g|-....
bor(.)=l-y;^

(3)

(2!)-^(6!f

third kinds have been


The

'

""""

(10!)-^

second and
functions of any order of the first,
and Whitehead".
effected by Russell:J:

Extensions of these definitions

W.

by

equationj=

adopt the

(2)

zero

the

is real,and

where

in certain electrical problemsconsists of

occurs

arguments have their phasesequalto {iror ftt.

Bessel functions whose


The

81

functionsber {z) and bei {z)and their generalisations.

TJiomsons

3-8.

FUNCTIONS

to

kind of order

functions of the second

pair of equationsresembling(2), the

by Russell by
I^ being replacedby the
defined

were

zero

function

function K^, thus


ker

(5)
Functions
to

{z)" i kei {z) K^ (zVi i).


=

of unrestricted order

defined

were

i"

by

third kinds, thus

Bessel functions of the firstand

(6)

ber,(z)" i bei,(z) J, (^e**'^'),

(7)

her,(z)" i hei,(z)

It will be observed

(8)

of

(z)
=

"

hir hei

(z), kei (z) ^tther (z),


=

" 3"7 (8).

due
series,
following

The

if,""(ze^^'"').

that||

ker

in consequence

to

Russell,are

without

obtainable

:
difficulty

+ lTrhei{z)
ker{z)= -\og(^z).her(z)

(9)

with reference

Whitehead

Presidential

Address

to

(1890), p. 492.]
Papers,
f In the case of fLinetious

the

Institute

of

Electrical

Engineers,1889.

[Math,

and

PIiijs.

in.

of

order, it is customary

zero

to omit

the suffix which

indicates

the order.

Phil.

Mag. (6)xvii. (1909), pp. 524" 552.


" QuarterlyJournal, xlii.
(1911),pp. 316"342.
IIIntegralsequal to ker

(3)XXII.

{z)and

kei

(1884), p. 450 [Ges. Werke,

\V. B. V.

(z)occur

i.

in

memoir

by Hertz,

Ann.

der

Physik

nnd

Cheinie.

(1895),p. 289].
6

82
kei

(10)

(z)

ber (z)
{z) l-rr

bei

log(1^)

E.Wtw^"'"'^'^-

also been

It has
ber^

Eussell

by

ber2(2)+bei2(.)

(11)

"

had

this result

but

that

first few

the

of

terms

expansion

the

of

thus
(2)have simple coefficients,

bei^

(j)+

observed

[CHAP. Ill

FUNCTIONS

BESSEL

OF

THEORY

l+MV"+ii^+Mf,
obtained, with

previouslybeen

in the
5-41); the coefficient of (l^)^"

expansion on

...,

(cf.

notation, by Nielsen

different

the

l/[(mIf (2m) !].

rightis

expansions involvingsquares and products of the general


functions have been obtained
by Russell ; for such formulae the reader is
and also to a paper by Savidge*.
referred to Russell's memoir
Numerous

Formulae

the results of

analogousto

; it is sufficient to quote the

Whitehead

(12)

ber_" (z)

cos

vir

(13)

bei_" (z)

cos

vtt

(14)

her_^(2^) cos

vtt

(15)

hei_^ {z)

vir

reader

The
been

worked

The

by

sin

will be

functions

B. A. Smith

3*9.

"

here
following

sin

vtt

her^(z) sin

vir

her,,(z)+

vtt

"

cos

discussed

by

[hei^(z) bei^(z)],
"

ber^ (z)],

[her^(z)

"

he\^,
{z),

hei,,
(z).

the

able to construct

of order

ptt

bei,,
(z)+ sin

which

formulae

recurrence

have

Whitehead.

lengthby

at

out

ber^(z)

been

have

""8-61, 3"62

in

examined

unityhave recentlybeen

detail

some

f.

The definition
of cylinder
functions.

Various

theoryof

the

have studied
and Nielsen",
writers,especially
SonineiJ:

analyticfunctions

of two

variables

the
satisfy

"^^{z)which

general
pairof

formulae

recurrence

(1)

'^._,(5)+ -^#.+1
(^)

"

^. (^),

(2)

"^.-1(^)

in which
are

and

satisfied

in

hy

Functions
Sonine

'^Vi(^)

unrestricted

are

each

which

complex variables.

of the three kinds

satisfy
onlyone

in his elaborate memoir

2*^; (^x

of Bessel

of the two

brief account

These

formulae

recurrence

functions.

formulae

are

also discussed

of his researches

will be

Chapter X.
*-^

Phil.

Mag. (6)xix.

"fProc. American
X Math.

Ann.

" Handbttch

xvi.

(1910),pp.

Soc. of Civil

49"58.

Engineers,

xlvi.

(1920),pp.

375

"

425.

(1880),pp. 1"80.

der Tiieorie der

Cylinderfunktionen(Leipzig,1904),pp. 1, 42

et seq.

by
given

84

THEORY

Some

writers*,

Fourier

(cf. ""

Bessel

Bessel
the

Jn{z)

Bessel

Fourier-

function,

call

the

of

Jn{z)

seems

work

of

desirable

of

order

any
to

do

to

customary

Bessel's
but

so,

in

predecessors

his

used

were

associate

name

invention

the

with

because

also

before

long

of

of

them,

the

great
and

simple

time

Bessel
because

only

not

advance

made

compact

notation

by

functions.

Bessel's

(1836),

name

p.

associated

was

[Oes.

13

with

Werke,

Math.

(1891),

vi.

by

functions

the

p.

Journal

Jacobi,

"Transcendentium

101].

definitis

integralibus

determinandis

in

usus

que

coefficients

generally
the

on

it

1'4),

1-3,
become

has

for

called

who

Ill

function.

Although

it

Heine

following

[CHAP.

FUNCTIONS

BESSEL

OF

ill.

exposuit

fur

Math.

/j.* naturam
Bessel

in

xv.

varios-

commentatione

celeberrima."

recent

more

series

of

letters

Nicolas,

E.g.
Jonrnal

cylinder

function.

fiir

controversy
in

Ann.

Math.

Sci.

lxix.

the

on

Nature,

LX.

de

VEcole

(1868),

pp.

norm.

p.

to

name

(1899),

128.

101,

sup.

Heine

applied

be

149,

(2)

xr.

also

to

174;

(1882),
seems

the

functions

(1909),

Lxxxi.

is

p.

to

be

found

in

68.

supplement,
to

be

responsible

for

the

term

CHAPTER

IV

DIFFERENTIAL

4'1. Daniel
The

solution

EQUATIONS

Bernoulli's

solution

givenby

Bernoulli*

of Riccati's equation.
of the

equation

-^^az''
+ bif

(1)

-^

dz

in

consisted

shewing that

0'

while

of

"l-

h have

and

means

"

when
_S

any

the index
_12

_8-

and
algebraic,
exjxjnential

justgiven are comprisedin

is

zero

or

Ifi

f, then

the

Ifi.

equationis
The

logarithmicfunctions.

soluble
values

by

of

the formula

" 1

Im

where

of the values

_12.

values

constant

has any

integer.
positive

Bernoulli's method

of solution is as

follows

If

be called the index of the

it is firstprovedthat the general


equationJof index
equation,
of
index
into the generalequation
N, where

is transformable

'^

N=-

(3)

1'

proved that the generalequationof


generalequationof index v, where

it is also

and

into the

(4)

n-

index

is transformable

4.

the
because
equationof index zero is obviouslyintegrable,
4 is integrable.
variables are separable.
Hence, by (4),the equationof index
this
If
tinued
Hence, by (3),the equationof index
| is integrable.
process be conthe
arrive
at
we
by usingthe transformations (3) and (4) alternately,
set of soluble cases
given above, and it is easy to see that these cases are
comprisedin the generalformula (2).
Riccati

The

"

"

pp.
n

Exercitationes
473

is not

475.

"

The

restricted to be

t It is assumed
be

quaedain iiuithematicae
notation
an

used

by

(Venice,1724),pp.

Bernoulli

has

been

77

80;

Acta
;

Eruditorum,
and

in

this

1725,

analysis

integer.

that neither

Ji is

nor

zero.

If either

were

zero

separable.
t That

"

slightlymodified

is,the equation in which

r;

and

h have

arbitraiy values.

the variables

would

obviously

86

OF

THEORY

Daniel

4*11.

Bernoulli

[CHAP. IV

FUNCTIONS

BESSEL

of Riccatis equation.
transformations

procedurehave been indicated,we


transformations are effected.
requisite

that the outlines of Bernoulli's

Now

proceedto givethe analysis


by which
Take

index

equationof

the standard

" 41 (1)as

the

and make

the substitutions

^"+'
"

[Note.
The

n.

The

substitutions

factor

+ \ in the

The

Y'

possiblebecause

are

denominator

is that the transformed

presence

-^

'

included

the values of
among
Bernoulli ; the effect of its

inserted

by
simple than

not

was

equationis

1 is not

"

more

if it

omitted.]

were

equationbecomes
1

clY

V-2

,7*7

^^

'

that is
dY

where

in

Again

The

where

and

+ 1) ;
n/(?i

h{n + \YZ^

" 4"1 (1) make

this is the

"

4 ; and

"

transformations

equationis soluble
solution in

generalequationof

the substitutions

index

N.

"

this is the

generalequationof

described

in

in the

The

When

the

stated.

cases

"

The

2.

finite number

in

this

v.

effected,and

proceduredoes

equationwith

equationwith

vjz,and

the

with
equation,

""4'1, 411

which

of transformations

To solve the

therefore

are

But

limiting
form of Riccatis

processes described
Riccati's equation,the value to

" 4'] (2),is

" 4"1

index

the

so

give the

not

compact form.

4" 12.

write y

aY\

equationbecomes

The

index
of the

index

"

2 is

"

2.

appliedto
continually

are

index

the

index

tends, when

not
consequently

types hitherto under

in

soluble

by

-*"

consideration.

2, namely

"

equationbecomes
dv
z

tt

bv-

dz

and

this is

an

equationwith

Hence, in this
of

the variables

limiting
case,

elementaryfunctions.

Riccati's

separable.
equationis

still soluble

by

the

use

13]

4-1 1-4-

This

solution
If

[). 185.

which

is

implicitly
given by Euler,Inst.

was

write

we

EQUATIONS

DIFFERENTIAL

(cf." 4-14)y

^^^
'T^-'

~"r

87

Calc. Int.

(Petersburg,
1769)," 933,

ii.

equation which

determines

r)

is

homogeneous, and consequentlyit is immediately soluble.

Euler

does

although he
which

not

gave

both

connects

mention

to

seem

the

the solution of the

equation of

any

limitingcase
homogeneous

Riccati's

of Riccati's
linear

type with

linear

equation explicitly,

equationand

the transformation

equation.

It will appear subsequently


(""4*7 4-75)that the only cases
the cases
Riccati's equationis soluble in finite terms
which have
are
"

examined

This
than

theorem, due

converse

soluble

cases

will
notation),

First transform
constants

follows

as

(1)

Riccati's

Define

new

ah

-c\

cases

^'

'"'"'"

those in which

are

variable

by

the

so

zu

equationin

'

odd

integer.

equation

-^

that the

2q-2;

1/q is an
1

variables and

equationis

the soluble

recondite

cases.
specified

" 4"1 (1),by takingnew


equation,

-v/h,

(2)
and

in the

more

the transformed

equationis soluble

much

solution of Riccati's equationin the


a
constructing
changes
by Euler*, and this method (withsome slight
explained.

be

now

C(jurse,

of

devised

was

of the values

one

of Riccati's equation.

Elder's solution

method
practical

been

(orboth) is zero.

Liouville,is,of

to

now

_9

or

that the

Bernoulli's theorem

4"13.

in

in which

also the trivial cases

has

_S_K.

_4_4-

0-

and

the index

; that is to say, those in which

in which

dw
ciz

is

2c.^-

A solution in series of the last

(?

) ^^^'-''^'^
=

0-

equationis

providedthat
Ar+,

(2qr +

q +

l){2qr +

8^c(r +

~A7
-Nov.

[1764],pp.

Comm.
154"169.

Acad.

Petrop.

1)

'

viii.

(1760" ] 761) [176:^],pp.

:5" 63

and

ix.

(1762-1763)

88

THEORY

and
+

so

1/(27H+ 1);

this

and

^m^"*'" if q has either of the values

with the term

the series terminates

[CHAP. IV

FUNCTIONS

BESSEL

OF

solution*

proceduregivesthe

examined

Bernoulli.

by

generalsolution of Riccati's equation,which is not obvious by this method, was


by Hargreave, QuarterlyJournal, vn. (1866),pp. 256"258, but Hargreave's
given explicitly
two years later Cayley,Phil. Mag. (4)
form of the solution was
complicated;
unnecessarily
tion
vii.
348"351
(1894),
xsxvi.
Papers,
[Collected
pp. 9"12], gave the generalsolu(1868),pp.
chief
diiference
between
the
Euler's
r
esembles
s
olution,
in a form which closely
particular
The

the two

solutions

being the

reversal of the order of the terms

of the series involved.

simpler form of the equation than (2),because he took


Cayley used a slightly
multiplesof both variables in Riccati's equation in such a way as to reduce it to

(5)

0.

Cayley's
generalsolittion of Riccati's equation.

4'14.

We

,-2_^2,-2

constant

have

equationis reducible

that Riccati's

justseen

the form

to

dz

given in " 4'13 (2);


this equation,which
equation.
When

make

we

and

shall

we

is to

the

explainCayley'sfmethod
regarded as a canonical form

be

and, if f/jand

U^

are

respect to

the
system of solutions of this equation,

form

of Riccati's

constants
arbitrary

and

equationis

primesdenote differentiationswith

z.

To express

U^ in

Ui and

finite form,
V

so

that the

in

lu

equationsatisfied by zy
proceedingin ascendingpowers
^-1
q{q-\)

1-

write

we

{cz^jq),

exp

is " 4-13

(4). A solution

of this

we

When

Phil.

take

the memoirs

(iriKSi^l)^,2,2.

.,^+

q{q-l)2q{2q-l)

J This

cV9

Uj^to be exp (czi/q)


multiplied
by this series.

the index

of the Eiccati

Mag. (4) xxxvi.

by

equation

of z'i is

(q-l){Sq-l)(oq-l)
q{q-l)2q(2q-l)Sq{Sq-l)
and

Riccati's

0 ;

fundamental

of

solving

the equationbecomes
(\ogv)/dz,

c^z^-^-^v

solution of the canonical


g'eneral

Avhere Cj and C^ are

substitution!
r]

(1)

of

now

Euler

(1868),pp.

which

were

equation is
348"351

2, equation (4) is homogeneous,

[CollectedPapers, vir. (1894),pp. 9"12].

cited in " 4-13.

is,of course, the substitution used

in

1702

by

James

Bernoulli; cf. "

1-1.

Cf. also

4-14]

DIFFERENTIAL

Now

U"

EQUATIONS

equation(1) is unaffected by changing the signof

U,

^
exip("cz'i/(j)

integer.
multipliedby
Ui, U^ form
If q

ratio

the

Since

qiq

(q-l){3q~l){5q-l)
q(q-l)2q{2q-l)Hq{Sq-l)

U^

fundamental

the

were

equation(1) in

when

q is the

U^ is the

algebraicfunction

an

so

take

we

(q-l){3q-l)

series terminate

both of these

and

c,

^q{q-l)2q{2q^'''''

'"

1)

and

89

of
reciprocal

of an
reciprocal

odd

exponential function

exp

of z-J, it cannot

of solutions of

system
an

odd

(fz"i +

"

be

constant

...

positive
{2cz'^lq)
;

and

so

(1).

we
negativeinteger,

should

write

the form
"P

(v/z)
0,
-c'{l/z)-"i-'{vlz)
d{l/zy
=

it follows that

whence

where

Fi

j^

V2

and

detail in
The

""4-4

reader

equation,when

It is to be
to terminate

"

4-42

and

constants, and

(+ cz^/q)

exp

series which

The

see

72 are

"

have

been

now

l)

q{q

obtained

l)2q{2q

will be examined

...

l)

in much

greater

4-42.

should

have

it is soluble

noticed

with

c-z-"i+

-q{q

that

the term

uo

the following
in constructing
solutions
difficulty

of Riccati's

in finite terms.

the

series

the
as
"'1,U2 (or l"i,V.^,

before the first term

Glaisher,Phil.

Trans,

which

of the Royal

has

Soc. cLXxn.

zero

be) are supposed


may
factor in tlio numerator
;

case

(1881),p.

773.

90

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

Among the writers who have studied equation (1)are Kummer, Journal fur Math. xii.
(1834),pp. 144"147, Lobatto, Journal fiirMath. xvii. (1837),pp- 363"371, Glaisher (in
the memoir
reference has just been made), and Suchar, Bull, de la Sac. Math, de
to which
" 4-3.
France,xxxii. (1904),pp. 103"116
; for other references see
The

reader
soluble

obtainable

by taking j

2 in

second

the

(1),and

included

it is not

so

of Riccati's

form

equationwhich

This is

reduced
easily
independentvariable.
To solve the

examined

was

Schlafli
equation,

arrived

the

at

generalsolution

of the
2/

solution of

(1) is

m!

r(a

equationin

y is

between

function

1, 0

Later

Ann.

c.,F{-a-l,t)

cd-^'F

(-

a,

t)

somewhat

s, defined
1
'*'
2

as

the series

a^
"

a^

1
"""

2^2+1)

2T 3

"

0^2+17(2
+

"^

""*

2)

'

used
of Schlafli's function,was
evidentlyexpressiblein terms
by Legendre,
of his proofthat tt is irrational.
(Paris,1802), note 4, in the course

the function
his Math.

di Mat.

was

studied

notation

(witha

different

Papers(London, 1882), pp.

(2)i. (1868),p.

series (" 1"1)is to be associated

t This

t).

de Geometrie

fragment in
*

a,

equation and Bessel's equation is thus


is necessary ("4*43)
tedious investigation
Cayley'ssolution and Schlafli's solution.

between

Elements

1)

Riccati's

is

'

evident ; but

The

exhibit the connexion


Note.

new

then

c^F{a,t)+

rendered

c,F(a, t)+ c.d-"F (-

cV^+^Fja
"

connexion

pn

F(a,t)=

1=0

which

was

'

00

to

equation

Iff

The

Schlafli

wrote

dt

The

by

" 4-13(2) by taking "t-^'la as

of

the form

to

''"^

the

cases.

^^t^-t-^-Ht\

(1)

and

the soluble

among

coMonical form of Riccati's equation.


Schldflis

4-15.
The

that

and

diately
0, the equation(1) is homogeneous and immeq
Bernoulli (" M) is
order equation solved by James

that when

will observe

should

be

232.

with

The

reader

will

Schlafli's solution

compared

with

notation) by Clifford;see
346

see

rather

the notation

posthumous

349.

"

that
than

of " 4-4.

James
with

Bernoulli's

solution

Cayley'ssolutiou.

in

92
reader

The

the

majority of

Univ. Nach.

And

the results

5, 7, and

1898, nos.

of the equationdiscussed
generalisation

obvious

(1)

Euler*.

by

either P

is

" 41

Qy^Ry^

P +

ox

are

any

Set. Math.

pointed out by Enestrom, EncyclopMiedes


of this type namely
equation
special
It

in

given functions of z. This equationwas investigated


R is identically
It is supposed that neither P nor
zero; for,if
R is zero, the equationis easily
integrable
by quadratures.

P, Q, R

where

4.

1899, no.

Riccati equation.
generalised

4*2. The
An

short paper
a
by Siacci,Napoli Rendiconti, (3) vir.
monograph on Riccati's equation,which apparentlycontains
of this chapter,has been producedby Feldblum, Warschau

also consult

should

139"143.

(1901),pp.

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

was

nxx

dx

nyy dx +

"

xx

dy

ii.

16, " 10, p. 75, that

xy dx

studied
pn'mi gradtts
by IManfredius, De constructione aeqiiationmndifferentialum
(Bologna,1707),p. 167. "Sed tamen haec eadem aequationon apparet quomodo construiab
bilis sit,neque enim videmus
quomod5 illam integremus, nee quomodo indeterminatas
was

invicfem

separemus."

The

equation(1) is easilyreduced to
order,by takinga new
dependent variable
1 d

(2)
The

equationthen

C^^^*

if in
Conversely,

the

logu

dz

(/^

dR] du

Tin

rv

generallinear equationof
dhi

/,.

(4)

the second

order,

du

P^d?-^P^dz'-P''''=^'

(where

jJo,Pi, p^

are

given functions

(5)

it

equationto

determine

of

write

z),we

e/.'/d2^

y is

^J _ft_"!j,_y.

(6)

dz

which

defined

equationof the second


by the equationf

becomes

/r,\

the

the linear

is of the

Riccati

type

same

equationwith

as

po

/)o

of the generalised
completeequivalence
equationof the second order is consequently

(1). The

the linear

established.
The

equationsof

1896, pp.
*

Nov.

Johnson,
t This
Calc. Int.

Comm.
Ann.

Acad.

of Math.

is the
II.

this section have

been

by Anisimov,

examined

[Jahrhuchilber die Fortschritte der'Math.

1"33.

Petrop.viii. (1760
ni.
(1887),pp. 112

generalisation of James

1761) [1763],p.

"

"

1896, p. 256.]

32 ;

see

also

short

paper

by

W.

W.

115.

Bernoulli's

(Petersburg, 1769),"" 831, 852,

Univ. Nach.

Warschau

pp.

substitution

88, 104.

("1-1). See

also

Euler, Inst.

4-2,4-21]

DIFFERENTIAL

Elder

4'21.

93

concerningthe generalisedRiccati equation.

theorems

EQUATIONS

solution of the
that, if a particuhir
by Euler*
Riccati equationis known, the generalsolution can
be obtained
generalised
solutions are
known
the general
by two quadratures; if two particular
solution is obtainable by a singlequadrature
f. And it follows from theorems
if
and
Picard
three
discovered by Weyr
solutions arc known,
that,
particular
be effected without a quadrature.
the generalsolution can
It

has

To

prove

and

shewn

been

he
the first result,let i/,,

write y

equationin

^/v- The

yo +

solution
particular

is

'^ {Q+2R"j,)v
+

of which

0,

the solution is
V

and, since
To

of

{f{Q+ 2%o) dz] + JR

exp

l/(y

"

the truth
t/o)"

"

y^ be two
y

dz

of the first theorem

the second, let y^ and

prove

{J{Q+ 2%,0 dz]

exp

0,

is manifest.

solutions, and
particular

write

ih

"

"

y-Vi

The

result of

"

dw

3/0-yi

and, when
and

dz

we

"

dz

substitute

for

last

Ih'^v

equationis

Vq

(dyjdz) and {dy^jdz)the


equationis reduced to

(y^'^^ y^
-

1 dz

"

y in the
-

p _^ q

Qya + Ry^^,the

P +

dy,

_^

{w"\y-

1) for

{yiW "y")l{w
substituting

values

iv

"

F +

Qy^+ Ry^-

1 diu
-r,

tu

so

that

where
we

tu

that y

see

exp

|/(i?y" Ryi)dz],
-

w,
integration.Hence, from the equationdefining
is expressedas a function involving
a
singlequadrature.

is the constant

r,

dz

of

solutions alreadyspecified,
prove the third result,let y^ and y^ be the
le^2 be a third solution,and let c' be the value to be assignedto c to make
To

y reduce

Then

to y.,.

yo-

c_

y-ya

yo

^
'

y-yi
this is the

and
*

t
Math.

Nov.
Ihid.
XL.

Gomm.
p.

59,

in
integral

Acad.

and

(1850),p.

Fetvop.
ix.

361.

viii.

form

u-i-i/^'

free from

quadratures.

(1760" 17G1) [1763],p. 32.

(1762"1763) [1764],pp.

163"164.

See

also

Minding,

Journal

fUr

94

generalsolution

It follows that the

in
expressible

the form

four solutions,obtained

by

then the cross-ratio


C\, Co, C^, 0^ respectively,

the values

givingC

is

y-i,yi be any

that, if ^/i,y-u

it is evident

Hence

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

{yi-y2)(y3-yA)
(2/l-2/4)(2/3-2/2)
is

; for it is

of z
independent

equalto
(6\-a)((73-C,)
(C.-QCCa-O,)'

In

the obvious

spiteof

noticed

been

until

it does

of this theorem,

character

not

have

to

seem

fortyyears ago*.

some

Riccati equationmay be derived from


of the generalised
properties
has given
linear equation("4*2). Thus Rafifyf
of the corresponding
properties
methods
of reducingthe Riccati equationto the canonical form
two
Other

|
these
form

the methods

correspondto
by changesof

of
propertie.s

Various
functions

the

been

have

"'

^(f);

reducinga linear equationto its normal


dependent and independentvariables respectively.

the

obtained

of

solution

by

of Riecati's

C. J. D.

equation in which P, ", R are


far Math. xxv. (1843),pp.

1356; cxxviii.
(1899),pp. 410
Comptes Reiidus,xcvi.(1883),pp. 1354
de V Assoc. Francaise
Jamet, Comptes Rendus
(1901),pp. 207
(Ajaccio),
228;
Autonne,

"

"

Fac. des Sci. de Marseille.,


xii.
The

behaviour

hagen, Nieuw
The

(1902),pp. 1

"

equation of

of P, "^,R
near
singularities
Wiskunde,(2)VI. (1905),pj). 209
"

the second

order

whose

23

37

"

412; and

"

Atm.

de la

21.

of the solution

Archiefroor

rational

Hill,Journal

primitiveis of

has

been

studied

by

Falken-

248.

the type

Cir)i+C2r]2+ C3rj3
^

where

the Riccati

no.
Toidouse,IX. (1895),

and

Comptes Rendus,
*

de

Weyr,

VEcole

Ahh.
norm.

is devoted

Noiiv.

Ann.

cxxxvix.

hdhm.

Ges.

s^ip.

(2) vi.

to

equation has various


t

of integration(which is an
obvious
generalisation of the
equation),has been studied by Vessiot,A7m. de la Fac. des Sci. de
6 and by Wallenburg,Journal fitrMath. CXXi.
(1900),pp. 210
217;

constants

Ci, c^, c-^ are

primitiveof

C1C1+ C2C2+ CSC3'

the

(1903),pp.

1033

"

Wiss. (6)viii. (1875"1876), Math.


Mem.
i.
Picard's
(1877),pp. 342"343.
thesis, in

theory of surfaces

and

applications.

de Math.

1035.

(4)ii. (1902),pp.

529"545.

twisted

curves"

p.

30; Picard,

which

theory

the
in

Ann.

result

which

is

Sci.
tained,
con-

Kiccati's

4-3]

DIFFERENTIAL

Various

4'3.
The
Bessel's

EQUATIONS

95

of Bessel's equation.
transformations

about to investigate
derived
are
equationswhich we are now
transformations
of
the dependentand
equationby elementary

from
pendent
inde-

variables.
first type which

The

where

is

unrestricted

an

shall consider

we

The

constant.

equationis of fi-equent
occurrence
is
problems,
usuallyan integer.
p

and, in such
physicalinvestigations,

in

is*

equation has been encountered


Theory of Sound by Poisson,Journal
Stokes, Phil. Trans, of
pp. 249"403;

in the

of Heat
and the
Tlieory of Conduction
xn.
Polytechnique,
(cahier19), (1823),
the Royal Soc. 1868, pp. 447"464
[Phil.Mag. (4)
XXXVI.
(1868),pp. 401"421, Math, and Phys. Papers, iv. (1904),pp. 299"324]; Eayleigh,
Soc. iv. (1873), pp. 93"103, 253- -283
Math.
Proc. London
[Scientific
Papers, i. (1899),
in
in
which
2
The
the
occurs
specialequation
138,
139].
Theory of the Figure of
p
pp.
The

de

VEcole

the Earth;

Ellis,Camb.

see

Since

Journal,

Math.

equation(1) may

li.

be written

d(uz-i)

^d-{uz-^)

(1841),pp. 169"177,

193"201.

in the form

.",,.

_,

,1..,,

its general.solution is

(2)

ii

zi%+.{ciz).

the equationis equivalentto


Consequently

Bessel's

equationwhen p is
unrestricted,and no advantageis to be gained by studyingequationsof the
form (1) rather than Bessel's equation.But, when
tions
the solujj is an integer,
of (1) are
"in finite termsf" (cf." 3 4), and it is then
expressible
desirable to regard(1) as a canonical form. The relations between
frequently
various types of solutions of (1) will be examined
in detail in ""4'41
4'4:).
"

Tiie second

type

of

the

dependent variable

The

roots

vz~P

equationis
which

of the

indicial

we

thus

are

generalsolution

(4)
"'"

Mem.

See

t This
Las

Plana, Mevi.
was

just been

di Fis.

known
made.

della B. Accad.
della
to

transformation

equationhave
and

"

p, and

zero

so

we

of
root.

write

dv
2

zP+^'",,+,(ciz).

delle Sci.

Soc. Italiana

Plana, who

p+l

(1) by

is
v

di Mat.

2p

from

indicial

equationof (1) are


led to the equation
c?-y

the

the

makes

/Q\

of which

derived

di Torino,

studied

xxvi.

(1821), pp. 519"538,

(1828),pp. 183"188.
equations (1)and (5)in the paper

delle Sci.

andPaoh,

xx.

to which

reference

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

96

Ann. Sci. de VEcole nm-m.


sup.
Equation (3),which has been studied in detail by Bach,
L.
Ann.
Lorenz,
see
certain
investigations;
in
physical
(2) III. (1874),pp. 47"68, occurs
i. (1898),pp. 371"
[OeuvresScientifiques,
(188.3),
der Physikund Chemie, (2)xx.
pp. 1"21
Solutions
of equation(3)
287"291.
396];and Lamb, Hydrodynamics (Cambridge,1906),""
examined
been
by Catalan, Bulletin
fractions (cf.
"" 5-6,9-65) have
in the form of continued
See also Le Paige,ibid. (2) XLi.
(1871),pp. 68"73.
R. de Belgique,(2)xxxi.
de VAcad.
935"939.

(1876),pp. 1011"1016,

Next,

we

in its normal

derive from
We

form.

(3),by a change of independentvariable,an equation


where q
write z
l/{2p+1), the equationthen
^^Iq,
=

becomes

'^^,-cV^-'v

(5)

0,

its solution is

and

(6)

When
taken

to

into the

is absorbed

factor

constant

a'i/qy'^'^^'^m,){ci^'^lq).

symbol '^,the

solution may

be

be

r*"^l/(2,)(c^"W(Z)"
has

Equation (5),which
Mem.

uher die BesseVschen

Studien

encountered

Tonno,

in

1868), pp.
(Leipzig,

Functionen

of the researches

interesting

account

been

studied by Plana,
Cayley, Phil. Mag.
9"12]; and Lommel,

112"118.

been

system of equationswhich has now


by Glaisher*, whose
systematically

The

" 4-14, has

(1821),pp. 519"538;

xxvi.

Papers,vii. (1894),pp.
[Collected

348"351

(1868),pp.

(4) XXXVI.

alreadybeen

delle Sci. di

della R. Accad.

constructed

important memoir

equationshave been studied from a different aspect by


regardedthem as degenerateforms of Lame's equationsin

the invariants g.^and

433"438;

contains

an

Haentzschel
which

both

of

zero.

by Glaisher should
Messenger,viii. (1879),pp.

followingpapers

The
pp.

g-^ are

cussed
dis-

of earlier writers.

The
who

has been

also be consulted
20

Proc.

23;

"

Phil.

London

Mag. (4) xliii.


Math.

(1872),
(1878),

Soc. ix.

pp. 197"202.
It may

be noted

that the forms

of

equation(1) used by

various

writers

are

as

^J+y=^^^"
d^R

-j^"

(Plana),

n{n + \)

a-u="

",
"

follows:

/p

"

(Glaisher).

u.

of a
by Greenhill|in his researches on the stability
Equation (5)has been encountered
variable
u
nder
When
of
the
the
cross-section
action
of
cross-section,
pole
gravity.
constant, the specialequation in which j f is obtained, and the solution of it leads to

vertical
is

Bessel functions
*

Phil. Trans,
and

Dublin

of order

of

the

Math.

Royal

1
.

Soc.

Journal,

clxxii.

(1881),pp. 759

"

828 ;

(1854),pp. 272"290.
Math,
und
t Zeitschrtftfiir
Phij". xxxi.
(1886),pp. 25"33.
Phil. Soc. iv. (1883),pp. 65"73.
X Proc. Camb.
ix.

see

also

paper

bridge
by Curtis,Cam-

4-31]

DIFFERENTIAL

Lommel's

4'31.

transformations
of Bessel's equation.

examined
types of transformations of Bessel's equationwere
by
two
of a somewhat
occasions;his earlier researches* were
special

Various
Lommel

97

EQUATIONS

on

type, the later f

much

were

In the earlier

general.

more

that
after observing
investigation,

the

generalsolution

of

is

(2)

3/

Lommel

proceededby

generalsolution
which

is

^'"g;(^),

direct transformations

where
z^''~''9$\{jz^),

a,

to construct

/3,7

it will be sufficient to quote, is that the

the

equationwhose
His

constants.

are

result,

generalsolution of

is

(4)

it

/3 0, the generalsohition of (3) degeneratesinto

When

and

when

unless jiv \H
The

into
0, it degenei'ates

zero.

solution of

(3)was

be sufficient to quote the

givenexpHcitlyby Lommel
followingfor refei'ence :

in

nvmierous

specialcases.

It will

^^-'^A^.
cfe^+.-rf^+n.^'-?)^*=^'

(5)

"=.i''%^.(v/.).'

0;
,g* (l_,)J+l"

(7)

ctz-

az

(9)

^2 /32/.-23

(11)

P^"zu0;

An

2^''-"'2?,(7^^).

--'

0;

u=

u=zi^i(izi),2H#j("ul).

of Stokes' researches

account

zi-^i/c,^)
(yz?).

on

the

solutions

of

equation (11)will

be

given in

"" 6-4,10-2.
*

Studien

(1871),pp.
t Math.
w.

B.

liber die

Bessel'scken

Functionen

(Leipzig,1868),pp. 98"120;

Math.

Ann.

475"487.
Ann.

F.

xiv.

(1879),pp.

510"536.

iii.

98

THEORY

Lommel's

OF

later researches

appeared at

about

the

same

to

the

two

by Pearson*, and several results are common


procedurewas to simplifythe equationf

d^jylxi^)] 2v-ld[ylx(z)]
d[y^{z)Y
yjriz) d^Jriz)
of which

the solution is

reduction

the

y}r{z)

adequateto

'

%(^)]%(^)

^ ^

%(^)

-'^

J -^

equation

take

"^(^) ^'(^){%(^)Plt(^)}"-^-

(14)
If

^^

x(^)Ja^

i^iz)

define the function (f)


(z)by the

It Avillbe

Lommel's

equationbecomes

l[ylr(z)
Now

memoir

papers.

(^)

as

x(^){^i^)V'^Mf(^)}-

(z)
[yjr

dz^

time

("4'3)

(12)
On

[chap. IV

FUNCTIONS

BESSEL

we

eliminate

dz"-

(^)"it is apparent

^{z)

that the

generalsolution

of

dz^\_4^\"l,{z)\
2cf"{z)
l

yfr'
(z)]

+
4"\^}r'(z)yfr{z) + |^2(^)_^.+ ^'ir{z)\
j

1J

'Z

IS

(^(z)yfr(z)]

As

"^^)

special
case,

if we

take (p(z)
=

1,it is seen

that the

generalsolution

2-

of

IS

(18)

y=snir(z)/f'{z)}."^^{.jr(z)}.

Next, returningto (13),we

take x

(z) [y\r
{z)Y-\ and
=

we

find that the

general

solution of

is

(20)
*

Messenger,

y=W{z)Y'i^^{y^{z)].
ix.

(1880),pp. 127"131.

The

functions

(z)and i/(z)are

arbitrary.

100
4-3 this is

By "

where

is

in
integrable

integer;

an

if

finite terms

that

so

J)4s

(l-r)^}

(2)

,.

The

in the trivial
equation is also obviouslyintegrable

2="

compact notation,invented

is convenient
write

the
expressing
subsequently

for

and

now

{0L)n
The

notation

which

by

series which

{a +

...

will be used

and

Pochhammer*

2)

Pq,

...,

"

(a)"

1.

("i)n
{Cli)n ("p)n
"

"

"

rr~Y~Y~\

''

"

\Pi)n \Pi/n

"

"

"

7T\
\Pq)n

,.

particular,
^"{Ph

i -i"^

,F,(p:z)=

functions defined

by

2;(p)2

^.{ph

z-

0^^-\p)n

n=0

The

modified

In

2.

b}^Barnesf,
We shall
investigated.

1),

n-

ap, pi, p.,

m=

is,in general,

...,

be

to

are

00
ITT

pJ^gK^i,0(2,

and

.4=0

cases

for series ofhypergeometrictype.

l)(a + 2)

ci{a+

^^^^

of Pochhamvier

The notation

4-4.

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

ni{p)n

the first three

series

are

called

generalised
hyper-

geometricfunctions.
It may be noted here that the
differential equation

and, when

p is not

an

Various

iF^(a; p; z) is

an
integer,
independentsolution

z^-^

It is evident

function

.,F,{a-p

l; -I-p;

of this

solution of the

equationis

z).

that

have been studied


integralrepresentationsof functions of the types ji^j,
0F2,0-^3
3Iath. Ann.
xli.
(1893),pp. 174"178, 197"218.

by Pochhammer,
*

Math.

Ann.

xxxvi.

t Proc. London
of the

suffixes

_p and

Math.

(1890),p. 84 ; xxxviii.
Soc. (2)v. (1907),p.

q before and

after the F

(1891),pp. 227, 586,


60.

The

to render

inodilieation
evident

587.

due

Cf.

" 4-15.

to Barnes

the number

is the insenion

of sets of factors.

4-4, 4-41]
4 '41.

We

DIFFERENTIAL

Various

shall

EQUATIONS

solutions in series.
examine

now

various solutions of the

^-c^u
and

obtain

101

relations between

in Pochhammer's

equation

1-^-2 u,

them, which

will for the most

part be expressed

notation.

It is

supposed for the present that p is not a positiveintegeror zero,


since the equationis unaltered
and, equally,
by replacing
p by
^3" 1, it is
supposedthat p is not a negativeinteger.
"

It is

alreadyknown ("4-3)that the generalsolution*


solutions
givesrise to the special

this

zP+'

The

which

.oF,{p + r, i c'z');
be written

equationmay

is

and
z^^S'p+'^iciz),

.oF,{l-p; i c'z').

3-P

in the forms

suggestedby the fact that the functions e*"^^ are


original
equationwith the right-handside suppressed.

solutions of the

are

When

^ is written

(^
When

for

1) (^

jt)

the
{djdz),

solve these in series

we

+ l; 2p
zP+'e^'.^FAp

^p+i Q-cz

last

,^F,{p+ l; 2p

2c^^

jj) {ue^'')"

(weT")

we

are

z'P e''

2; 2cz) ;

o.

four expressions
for u:
following

led to the

2; -2cz);

pairof equationsbecome

,F,(- p

z-^' e''-

-2cz);

-2p:

.iF^{-p\ -2p;

2cz).

the left are


of series,
the two expressions
on
Now, by direct multiplication
And the expressions
in ascendingseries involving
zV'^'^,
z^"^-,
zP'^\
expansible
....

on

the

involve
rightsimilarly
the

are

powers

when

same

2p

is not

we
integer,

an

2cz)

e'"

^oFAp
e''.,F,(-p; 2p

J^)

2cz)

These

formulae

due

are

to

Kummerf.

values of p, the truth of (2) is obvious


We
*

to consider

have

now

It follows

from

"

3-1

that

the

cases

fUr Math.

xv.

(1836),pp.

138"141.

of

2p^2;

2cz)

ic'^')^

,F,(- p

2p

2cz)

oF^{i-p; \"fn
(1) has been provedfor general

replacing
p by

when

specialinvestigationis

integer.
t Journal

e-"

When
on

I ;

sets

have

must

,F, (p

of the two

none

....

e'^'.,F,(p+l;2p-v2-

(1)

Since

z'^, z^-p,z--'',

2p

is

an

-p

l in

"

(1).

integer.

also necessary

when

is half

of

an

odd

102

the solutions which contain z'P


p has any of the values i, f |, ...,
3-52),
factor have to be replacedby series involvinglogarithms(""3-51,

When
a

as

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

of

only one solution which involves only powers


equation(1) still holds.
previousreasoning,
of the
values 0, 1, 2,
a comparison
When
p has any of the
and

is

there

By

z.

lowest

powers
obvious

...

of

in the solutions shews

involved

that there

relations of the form

no

are

(1) stillholds ; but it is not

that

z-P,F,{\-p- ic^z"-)z-Pe'\FA-p- -'2p: -2cz)

We
and

shall

(2) which

^; \"z^

are

not

zero.

of (1)
have to give an independentinvestigation
consequently
of series.
depends on direct multiplication

In addition

Note.

which

constants

are

z-Pe-'\F,{-p; -2p; 2cz)+ hzP+\F, {p + ^ ; \c'z%

kj,ko

kzP+\F,ip

where

the

to

should

researches,the reader

Kummer's

consult

the

tions
investiga-

Papers,
(1868),pp. 348"3.51
[Collected
by Cayley,Phil. Mag. (4)xxxvi.
Phil.
(1872),pp. 433"438;
9"12] and Glaisher,Phil. Mag. (4) XLiii.

of the series

(1894), pp.

VII.

Trans, of the Royal Soc.

The

the solutions in series.

between

Relations

4'42.

(1881),pp. 759"828.

CLXXii.

equation
e'\F^{p+ \\ 2p + 2; -2cz)

which

part of equation(1) of " 4-41, is

forms

due

generalformula

to

p;

of the

case
particular

more

which
presently)

e^.FAp-a;

p;

0,

and p subjectto certain conventions


(whichwill
have to be made
when a and p are negative
integers.

holds for all values of

be stated

2; 2cz),

Kummer*

.FAa;

(1)
which

e-'\F,{p+ \; 2p

that p is not a negative


integerand then the coefficient of
^" in the expansionof the productof the series for e^ and iF^{p a; p ; -t) is
We

first suppose

"

nC"i {p

"=o(w-m)! w!(p)^

w!(p)nm=o
1=0

"),"(1

n)r,

,(1 -a-n)n

nl(p)n
(a)"
'

ifwe
in

firstuse

the numerator;
result

The

values
*

III.

Vandermonde's

!(p)n

theorem

f and

then

reverse

the order of the factors

and the last expression


is the coefficient of ^^ in iFi(a,p', 0-

requiredis therefore

established when

and p have

generalcomplex

J.

Journal

(1874),p.

fur Math.

xv.

(1836),
pp.

138

"

141;

see

Ann.
also|,Bach,

Sci. de VEcole

norm.

sup.

(2)

55.

t See, e.g. Chrystal,Algebra, ii. (1900),p. 9.


X Another
Trans.

Camb.

the theory of
on
(1908),pp. 254"257.

proof depending
Phil.

Soc.

xx.

contour

integrationhas

been

given by Barnes,

4-42]

DIFFERENTIAL

103

EQUATIONS

unless
equation(1)is obviously
negativeinteger,
meaningless
of (1) in these
also a is a negativeintegerand |a ]" |p |. The interpretation
will be derived by an appropriate
circumstances
limiting
process.
p is

When

so
negativeinteger(= N) and let p not be an integer,
is valid. The
series iF-^{"N; p: ^) is now
a
that the precedinganalysis
sists
^) is an infinite series which conterminatingseries,while ^F^ip + N ; p;
in
\vhich
the
earlier
followed by terms
factors p + N,
of iV + 1 terms

First let

be

"

"

p +

N
p +

li^'i
(p + iV
is any

; p ;

then

^11'*^

"

-\-l,N

the last term

in which

then

we

the

does

numerator

to

see

01 =.F,(-.Y;

replaceN hy
-M-

ordinarycase

and

"

^)^

and

from

have

M+2;

^)

This could

,F,(-N;
have

been

value, and
integral

an

with

this result combined

(.5)

Cf.

far

^^^^"*"\
^-^^+-,
...

the term

as

the

vanishing

though

as

^^\ i.e.

their ratio

that

-M;

^ by
,F,{N

Ol

have

^,we
M;

-J/;

(2),(3) and (4)

we

-^1
1

e^.FiN +1;

^v

H-

2; -^),

deduce

fi'om (1)by givingp-a


directly
making p tend to its limit.

derived
then

2 ;

that

0^=e^^F,{Iy-M",-^r:-n-^.

-M;

v. (1871),
pp.77"
Gayley,Messenger (oldseries),
462],and Glaisher, Messenger, viii. (1879),pp. 20" "23.
*

in

.^^M+l^ (,Y+

of (1) we

,F,(M-N-\-l-

y^

in

factor,w^hile

zero

proceednormallyas

being cancelled

N\{M-N)\

an

-0^,

contain

not

wnth terms

denominator

-M:

,F,iN-M;

(4)

form
M, and the limiting

"

that the series is to stop at the term

this convention, it is easy

When

-*

^)-]=e^,F,(N-M;-M;

means

and

,F,(-N;

(3)

equality.

of

With

....

the limit when

it is to continue

factors in numerator
one

2,

that the series is to

symbol ']means
and
(^^'~^\

be written*

symbol 1

As

...

"

N,N
integers

in w^hich the

were

both

,i^,(-iV; -.";

(2)

in

in the denominators.

I, p +2,

p, p +

be cancelled

can

cancelled,the series for i^i(- iV; p; ^) and


functions of p near
continuous
M, where
p

'may proceedto

we

(1) may

"

of the

Hence

the

in the numerators

in the sequences

...

these factors have been

When

of

2,

the later factors of the sequences

with

82

Papers, vm.
[Collected

(insteadof

a)

(lB95),pp.4oS"

104

THEORY

We

examine

next

which

forms

[CHAP. IV

FUNCTIONS

BESSEL

equation

the

^\F, (^ +

(6)

OF

1 ;

2p

2 ;

2cz)

o^^i
(^

f ; \c'z%

which

equation(1) in " 4'41,and

of

the remainder

is also due

to

Kummer*.
the coefficient of
suppose that 2p is not a negativeinteger,
the productof the series on the left in (6) is
If

we

(- 2r

"

{p

{-Y

1)",

Now

"

2'"

expansionof (1

l{n

(1
2t)-P-'

and
t)'^+^'-^\

it is

so

m)

2t)-v-'(1

dt

r"-i
^)"+2i"+i

no+)

-".\

"" in the

equalto

1
\

(1

(1

!/2)-p-h"-'"-irfM,

t/(l t)and the contours enclose the originbut no other singularities


integrands.B}^expandingthe integrandin ascendingpowers of u, we

of the

"

that the

see

2;)-!)"_",

l)n-m is the coefficient of

"C,rt
(^J+ l)m ("n-2p-

/"(0+)
/"(0+)

27r*

where

lX"(-n-

"""

(p

(2p + 2)," (2p + 2)" ,"^o

{n-m)\m\

,"=o

2,"

{czy'in

integralis

if n is odd, but it is equal to ~^i~~^

zero

when

is

even.

Hence

it follows that

jczT'

^
_

and

this is the result to be


When

the

make

we

p tend

limiting
process

same

lim

,F^{p +

\:

proved.
to the

as

value of

negativeinteger, N,

we

find

by

before that

2p+2;

2cz)

,F,(1-N;

2iY;

2cz) 1

^'^

(-V-'(N-1)]N^
^

i2N-2)l(2N)Y

(- ^"^"""' ^^^ (^' 2^5


"

2cz).

'

It follows that

oF,{^-N; ic'z"-)e'''.,F,{l-N;2-2N;

-2cz)n

(-Y(]V-l)\Nl

^^''^''''" ^^^^^"
(2^)!

"^

""

(2i\r-2)!
If

we

change the signsof

"

and

throughoutand

2iV^'-2c^"-

add

the results

so

obtained,

find that

we

(7)

2.,F,{^-N;

lc-^z')=e".,F,(l-N; 2-2N+

e-''
.

-2cz)1

^F, (1

iY;

2N; 2cz) ~^

Journal
filr Math. xv. (1836), pp. 138"141.
In connexion
Barnes, Trans. Camh. Phil. Soc. xx. (1908),p. 272.

with the proof given here,

see

4-43]

DIFFERENTIAL

the other terms


the

course,

For

on

"

see

The

rightcancelling
by

proofof

of

use

Rummer's

equation(1).
with

This

is,of

different notation.

formulae, by the methods

gration,
inte-

of contour

6"5.

equation.
Sharpes differential

equation

.g |

(1)

which

J-^r+},
(icz)in finite terms

for
expression

Barnes'

4 "43.

the

105

EQUATIONS

is

(. + 4),

0,

of Bessel's equation for functions of order zero,


generalisation
of
the
reflexion of sound by a paraboloid.It has been
occurs
theory
which
reduces to
investigated
by Sharpe*,who has shewn that the integral
unityat the originis
a

in the

y=G\

(2)

(io^{zGo^6+A\o^(tot\d)d6,
Jo

where
"iff

(3)

['"cos
(A logcot

hd) dd.

Jo
This

is the

its convergence

write

(4)
^

'^

It is easy

to

values of A

from

see

for which

cos

cos

(^0+.

j/(^)

of the

'

tanh

and
(f),
tanh

To

vestigat
in-

it becomes

"^)
^

cosh(/)

./(,

this form

integral(" 2'3).

of Parseval's

appropriatemodification

that
integral

for

it converges

(complex)

1, andf

"

"

cosh ^ttA.

IT

in great
integTalhas been investigated
successive
given elaborate rules for calculating

The

y in powers
A

of

simpleform

reader should
*

by Sharpe

coefficients in the

and

of the solution

have

no

(which

was

e"'2,Fi(i+ iU;

expansionof

1;

2{z).

xii.

(1884),pp.

66"79

result.

; Proc.

Comb.

Phil. Soc.

pp. 101"136.
e.g.

has

givenby Sharpe)is

not

this
in verifying
difficulty

Messenger, x. (1881),pp. 174"185

t See,

he

z.

2/

The

detail

Watson, Complex Integrationnnd

Canchtfs Theorem

(1914),pp.

64-

-65.

x.

(1900),

106

OF

THEORY

the second.

Equationsof order higherthan

4"5.

[CHAP. IV

FUNCTIONS

BESSEL

equationof any order,which is soluble


its possibility
of Bessel functions,has been effected by Lommel*;
by means
depends on the fact that cylinderfunctions exist for which the quotient
Wy (z)l^_y(2)is independentof 2.
The

of

construction

differential

functions

of the

Each

possess
Now

and

is the

this

form

fundamental

hence

2v.

and

2n,

(ii)if

or

have

we

result that the functions

that 7^"=

so
(" )"c^",

ic exp

values
possible

2^^

Jn(y V^)?

{r-rriln). (r
obtain

we

that
=

0, 1, 2,

solutions

2n

of

1)

n"
,..,

these

(2),and

system.
we
H^'\,+i,

Next, if '^"+.denotes

and

off

value such

all

if

integer,n,

an

Yn,

either J,ior

By giving7

is

"

1.

solutions

has any

(i)if

order

obtain Lommel's

equationwe

^/2)are
2i^Yn{'y

where

2?i +

if 9^n denotes

Hence

From

either

case

) K-.n
(Iy)-^*'""
(7 V^),

rightis of

the

on

not.

in the form

2^"^^ (7 ^2)

cylinderfunction
This

integeror

an

" 3'9 (5) is written

when

(1)
the

is

iT,'''
{2), H^'^^(z)

of the third kind

the functions

[""2-31, 3-5]; and


it ("3"61),whether

property

this

order, possesses
Yn{z), of integral

and

Jn{z)

2'^'"''^^
H^^\+),
(7 ^^z)is

have

that
so
'^_(n+A)e'"+^"^'^n+i,
=

solution of

d^-^_(^cr^
where

has any

value

the solutions

and
*

Studien

ilber

so

that

such

obtained

die BesseVschen

g^
7^''+! c^"+i Q-^n+\)n%^

-ic
form

exp
a

Functionen

+ ^)},
{r7rij{n

fundamental

^j^g^^

(r

0, 1, 2,

2?i)

system.

(Leipzig,18G8), p. 120;

Math.

Ann.

11.

(1870),

pp. 624"635.

has

The

been

more

generalequation

discussed

by Molins,

Mem.

de VAcad.

des Sci. de

Toulouse, (7)viii.

(1876),pp.

167

"

189.

108
These

four

give the following equationsand

cases

(6)

(8)

g|,"g|.l";

^^

w;

These

to be the

seem

^*{"^i(^)
+ #i(^2)},

onlyequationsof Nicholson's type which are


case
2, the equation(5)is homogeneous.
n

in the

their solutions

u=z-^{"^i{izi)+WU"^h],

of Bessel functions

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

soluble with
Nicholson's

the aid

general

equation is associated with the function


2 +

^3-2fx

2fi

"^^V4-2/x'
4-2/i'
4*6.

z^-'^

l+2fi

4-2/x' (4-2^)V

equations.
Symbolicsolutions of differential
mathematicians

Numerous

given solutions

have

of the

equation" 4'3 (1)

namely

"--=^-^^".

(1)
in

symbolicforms, when
connected
intimately

are

It has been

j:)is
with

positive
integer(zeroincluded).These

the

(|4'3) that

seen

formulae

recurrence

the

generalsolution

for Bessel
of the

forms

functions.

equationis

2i%+^{ciz);
and

the

from

recurrence

formula

" 39 (6) we

have

{ciz)].
[z-i'S^
lj"j'

zi%^,(ciz) (- ciyzP^^
=

Since

where

any

function
cylinder

and

/3 are

of the form

as
'W^{ciz)is expressible

constants, it follows that

the

generalsolution

of

(1) may

be written

(2)

"

.P"(4Y?"!!"^"r.
\zdzj

where
*

modification

of this,due

/8'= /3/c.This
ajc,
"

to

may

Glaisher*, is

be

seen

aV^+/3'e~'^^
once.
by differentiating

Phil. Trans,
remarked
of the Royal Soc. clxxii.
(1881), p. 813. It was
by Glaisher that
equation (3)is substantially
given by Earnshaw, Partial
DifferentialEquations (London, 1871),
p. 92. See also Glaisher, QuarterlyJournal, xi. (1871),p. 269, formula
(9),and p. 270.

4-6]

DIFFERENTIAL

109

EQUATIONS

equivalentto (2) was set by Gaskiti as a problem* in the Senate House


published by Leslie Ellis,Camb. Math. Journal, ii.
Examination, 1839 ; and a proof was
195, and also by Donkin, Phil. Trans, of the Royal Soc. CXLVii.
(1857),
(1841), pp. 193
the
of
c^
that
the
In
the
set
tion
soluas
was
43
57.
s
o
by
Gaskin,
sign
question
clianged,
pp.
involved circular functions instead of exponentialfunctions.
Note.

result

"

"

Next

""/)".'
z-f

operatingon

function

function

the

operators z^

with

to

to

is

Equations (1914),"
It is easy to

from

see

reversal of the
zP+'

at

the

reverse

now

The

Nouvelle

CLXXii.

X It

function

of

z.

no

2) (^

(4-T
=

^*'"^" ^

was

2^)+ 4) (^

2jj+ 6)

:^,
the

on

(^

operators in the last result,and

of the

left)

2) (^- 4)
-

(^

...

(^

2p

(^

2) ^

e-"z'+^)e
e-'^^"-'"]
(e-o^)...
(e-"^).
[^e^e^^

part

2)

dzj

by

get

e"^-^'" [(^ +2^-2)

second

of (5).
time, by repeatedapplications

the

with those
e~"^ (beginning

functions

previousprocedurewe

problem

Cambridge

2^

order

^p+i
*

a) Z,

(5) that

successive

past the operators one


We

presents

e^i-P)B(^

bring the

we

Z is any

e"' ./(^

33.

when

specialdifficulties when/ (5) is a polynomial in


present investigation.See, e.g. Forsyth, Treatise on Differential

this formula
in the

case

and

constant

proof of

is the

as

it is supposed
right,
applicationof the

.|^a.

.",

The

the

write

/(^) (e"'Z)
a

1/^-^"-before

symbolicformula

the

use

(5)
in which

the

on

(d/dz).

then

the operator

multipliedby

is

It is convenient

S,

that
Glaisherf,

to

z'-

In

and

due

symbolictheorem,

^""(:^J
ii^,dz)

(*)

that

the

shall prove

we

2p

4)

e-t^i'-^t"]

z-i'^-

of

question 8, Tuesday afternoon, Jan. 8, 1839

; see

the

p. 319.

UniversityCalendar, 1839,
Math.
Corr.
ii. (187(3),
pp. 240"243,

349"350

and

Phil.

Trans,

of

the

Eoijal Soc.

(1881),pp. 803"80.5.
was

remarked

by Cayley, QuarterlyJournal,

Glaisher, that differential operators of the form

s"+'

xn.

(1872),p. 132, in

^-2-", i.e.

",

obey

footnote

to

paper

the commutative

by
law.

no

THEORY

proved. If

this is the result to be

and

dz)

',dz

When
the

replace
p hy p

we

2P+1

(6)
we

of

1, we

find that

z-P

I
"

"cz

dz)

'

1p"\

,d y+^

/3'e
"cz

(xe'^ +

\fdz)

^i'+3

solutions of the

that

zP+^V

(8)

see

+ ^edyae'^'

/
"

(7)

The

(2) and (3) with the aid of (4) and (6),we


forms :
in the following
(1) is expressible

transform

generalsolution

[chap. IV

FUNCTIONS

BESSEL

OF

Z2P

equation
d^v

2p

dv
c^v

"

dz-

0,

dz

correspondto (2),(3),(7) and (8) are

which
[(3)of "4-3],

(9)

^e-"'

yae^^+

!^"-P+i

zdzj

d, \^+^

(10)
(11)

(12)

-\

dz)

Z'

z^'

ae"^
d_\P+'
'Tzj

'

4-

^'e'

^2p

obtaining(7)is due to Boole, Phil. Trans, of the


Royal Soc. 1844, pp. 251, 252 ; Treatise on Differential
Equations (London, 1872), ch. xvii.
also
423
Math.
and
Dublin
see
Journal, IX. (1854),p. 281.
425;
Curtis,
Cambridge
pp.
solution (9)was
first given by Leslie Ellis,Camb.
The
Math.
Joxirnal,11. (1841),-p\).169,
de Math.
xi.
193, and Lebesgue, Journal
(1846), p. 338; developments in series were
obtained from it by Bach, A^m.
Sci. de VEcole norm.
sup. (2) in. (1874),p. 61.
A

diflferent and

direct method

more

of

"

dP'V

symbolic solutions

Similar
John

equation -1-5

vi. (1886
Hopkins UniversityCirculars,

A
pp.

for the

transformation

of the

solution

-c-z^'i~^v=0

discussed

by Fields,

7),p. 29.

"

(9), due

to

Williamson, Phil. Mag. (4)

xi.

(1856),

364"371, is
(13)

{ae''-\-Sie-").
c'""(^^.-\

-1

is derived

This
functions
We

of

thus

from

the

equivalenceof

obtain

the
1

equivalenceof

when

operators -",-",

they operate on

being supposed that


manifest.

the

followingoperators

:ci)"j]=""^"="*{C:^*)"a
=

is then

the

cz.

^2^ +

it

were

the

(p.^y-p +
-{cz)

_\cz^ccj czj

operators operate

on

function

\_\cdcj cj'
of

cz

and

Williamson's

formula

4'7]

DIFFERENTIAL

Liouvilles

4*7.

EQUATIONS

HI

classification
of elementarytranscendental functions.

givea proofof Liouville's generaltheorem (which was mentioned


of solving
in "4-12)concerningthe impossibility
Riccati's equation"in finite
in
classical
the
discovered
cases
terms
by Daniel Bernoulli (and the
except
form
index
shall
of
of Liouville's* theory
we
"2),
give an account
limiting
of a class of functions known
transcendental functions;
as elementary
and we
notation for handlingsuch functions.
shall introduce a convenient
Before

we

"

For

write

brevitywe

f
I (I(z)),

k (z) l{z)^ logz,

I,{z)

e,{z) e(z)

e.(z) e(e(z)),

I (I,
(Z)),

I,(z)

...,

e-Az)

(e,(z)),

,,f(z) ,f{z)=Jf{z)dz, ,,f(z) ,[,f{z)],,,/(,)

l,,/(^)},

e',

fimction

of

is then

said to be

....

elementarytranscendental function\

an

function
algebraic

of z and of functions of the


as
an
expressible
types
where
the
functions
Cr^iz), '?rX(^).
auxiliary
^{z), -^(z),xi^) ^^^
lr(j)(z),
of
of
second
and
of
in
terms
set
a
^^
functions,and so on;
auxiliary
expressible
providedthat there exists a finite number n, such that the nth set of auxiliary
functions of z.
functions are all algebraic

if it is

The

order

of

elementarytranscendental

an

function

is then

defined

follows:

as
inductively

function of
(I) Any algebraic

is of order

zero|.

(II) If /,.
[z)denotes any function of order
of functions

of

r,

then

function
algebraic

any

of the types

lfr{z),efr{z\ ,f,(z),f.(z), fr_,(z),...f(z)


(intowhich

at

least

of the first three enters)is said to be of order

one

(III) Any function is supposedto be expressedas


order.
possible
order

Thus

of order

not

elfr(z) is to be
?"

replacedby

?'

1.

function of the lowest

fr{z),and

it is

function of

2.

with this and the followingsections,the reader should


study Hardy,
of Infinity
(Camb. Math. Tracts, no. 12, 1910). The functions discussed by Hardy
inider
of a slightly
restricted character than those now
more
consideration,since,for
were
to
his Btirposes, the symbol s is not required,
and also,for his purposes, it is convenient
of
the
postulate
reality the functions which he investigates.
In

connexion

Orders

It may

be noted

merely remarked
*

Journal

de

that

Liouville

that it had

Math.

ii.

many

(1837),pp.

study propertiesof the sj'mboly


propertiesakin to those of the symbol I.

did not

56"10.5

t It is supposed that the integralsare


X "Une
" For
not

be

fonction
the

as

in.

(18.38),
pp.

523"547

(1839),pp.

iv.

423"45(5.

all iudeiinite.

fiuie

explicite."
irratioiKtl
investigation,
functions.
algebraic

purposes

regarded

in detail,but

of this

powers

of :, sucli

as

z^

of

course

must

112

of the solution of the

character

of the
investigation

The

linear differential
equations.
concerning

theorem*
first

Liouvilles

4-71.

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

equation

d}u
,

transcendant

x{z) is a

in which

orderf n,

has been

made

by Liouville,who

of order ni + \, where
which is of orderf n,
^ n, then either there exists a solution of the equation
in the form
expressible
else there exists a solution,u^, of the equation
If equation(1) has

of

theorem
following

has established the

or

^^=^^%(^"'

(1)

solution luhich is

(2)
where

u,

(f)^{2).ef^{z),

of 4"^{2)does

the order

order fu,,and

is of
ft,.{z)

transcendant

exceed

not

fj.,and

/m

is

such that n^fi":m.

equation(1) has a solution


function of one
/m+i(^)is an algebraic
of
?/m ("^X^fm(^)as well as (possibly)
If the

Let

m.

concentrate

us

shall first shew

(I) We
m

1, then

solution

the types 6 and

can

how

1, let it be

functions

functions

whose

of

order

fm+i{^); then
the types If^(z)^
does

not

function of
particular
accordingto its type.

on
or

more

or

attention

our

let it be called ^, ^

three types,and

of order

"

that,if (1) has

to

prove
be constructed

which

does not

one

solution

involve

exceed
of the

of order

functions

of

^.

function of 6;
F(z, 6),where F is an algebraic
of order ni + 1 which
in F
(otherthan 6 itself)
occurs

let /,"+i
(z)
For, if possible,

and
is

any

function

of

independentof
algebraically
it is easy to shew

Then

^^'

'^^^

dz'

that

dz^^f^Xz)dz

df,,{z)Yd-^F\d \

\
^

6.

{z)
(/,"

"^

f 8^

dz

beingsupposedthat z and
differentiations.
the partial

it

are

dOdz
\

df,,{z)\-\dF
"^

{z)
^dz (/,"

the

dz

\\de

in

independentvariables

^ ^^ ''

"

performing

expressionon the right in (3) is an algebraicfunction of 6 which


when
6 is replacedby lf"^{z).Hence
vanishes identically
it must
vanish
for all values of 6 ; for if it did not, the result of equating it to
identically
would
whose
zero
algebraicfunction of transcendants
express lfm{z)as an
orders do not exceed ni together
with transcendants
of order wi + 1 which are,
ex
algebraically
independentof 6.
hyputhesi,
The

Journal

t This
t

Null

de Math.

iv.

phrase is

used

solutions

are

(1839), pp. 435"442.


as

an

abbreviation

disregarded ;

if

of
were

"

elementary transcendental

of order

less than

function

n, then

would
,

less than

n, which

is

contrary

to

hypothesis.

dz''

of order

".-'

be of order

4-71]

DIFFERENTIAL

113

EQUATIONS

the rightof (3) vanishes when 6 is replaced


the expression
on
particular,
this change is made the
constant
; and when
hy 6 +c, where c is an arbitrary
the left of (3) changesinto
on
expression
In

which

is therefore

is to say

That

zero.

^^-^

(4)
When

differentiate

we

-F{z,0

d'Fjz, 0

c)

'

solutions of

(1) for

all values of

^0' Ke,

(1). For

of

consequentlysolutions
either F

they do

If

put

we

0 after

be called

brevitythey will

Fg form

and

ol' (1)

system of solutions

fundamental

or

not.

they do

not,

have*

must

we

Fe
where

"""

'

""""

Now

If

c)

d-F{z,e)

dF{z, d)

are

find that

c, we

independentof z.
become
expressions

these
performingthe ditferentiations,

which

dc'

ac
are

with regardto
(4) partially

dF(z, e

0.

c).x(2)

is

independentboth

of

AF,

and

On

0.

we
integration

find that

F:=(^e^',
where

involves

(of order

transcendants

independentof 6. But this is


algebraically
F and
function of 9; and therefore
algebraic
of solutions of (1).
Hence

in
Fg^ is expressible

where
in 6

and B

(with constant

and "I"2
are
"l"i

the roots

of the

-^ ^,e^^
^.e'^^

only value
/3

0 ; and

of F

which

then F is

be

B.

".

the form

linear

equation

e'^^["i",
+ "i".d\,
nature

same

as

while
cj)^

and

jB are

"

0.

function
algebraic
linear function
"f 0.
is

an

of 6 is obtained

function of the type ^,

wc

can

prove

when

that

linear function of '^.


*

w.

Ax

"

if f^,^^^
{z)involves
Similarly,
it must

equationof

equation
X-

The

an

BF,

or

of the

functions

by

this may be regarded


as
is
its solution

and
coefficients)
F

where

AFe

Now

constants.

are

of F and Fg

terms

Fee

exceedingm + 1) which are


because e^^ is not an
impossible
Fq form a fundamentalsystem

not

Since

must

involve

0, b'ecannot

be

zero,
identically

114

It follows that,in

write

may

of order

^p,q(z)are

involved

we

the functions
m

functions of the types 6 and ^,

fm^^{z)involves

(z),
te, {z)d,{z)...dp{z) ^, {z)% {z)...\ (z) ir,^
,

/"^,{z)
of order

as

that
so
linearly,

it involves them

where

far

so

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

in them

onlyfunctions

the

of the type 0.

are

in

of the terms
any one
function of O^,6.,, ^i, ^2,
Take

1 at most, and

(z)which
f,n+i

is of the

highestdegree,qua

let it be

and

""",

...

e,(z)6,{z)...dp {z).^,{z)...^Q {z) tp, Q {z).


.

used, it follows that


Then, by arguments resemblingthose previously
d

is

solution of

function

-f

(z) is a solution of (1).


(1); i.e.i/rp
^

is either
'\jrpQ(z)

But

of order

the types 6 and

of order not

function

which

involves

functions

exceedingm,
of the

else it is

or

type 0

and

of

not

^.

In the former

repeat the process of reduction

we

case,

order,and in the latter

case

we

that

see

solution

some

to

functions of lower

of the

equationis

an

function of functions of the type @.


algebraic

of order greater than


is

solution which
where
"^^{z),

of order

the transcendant

G is

of order
Then

yu.+

fi

and 0^

prove

our

solution which

of functions

n,

then

attention

on

(1) has
it has

of it

power

of the

order which

an

that, whenever

greater than

is a transcendant

solution of order

{z)is of

ef^{z)only in having a

concentrate

and
ef^{z),

then either it has

algebraicfunction

an

shall next

transcendant

We

n,

f^ {z)is of order

(II) We

where

provedthat,if (1)has

have therefore

We

else it has

type ef^{z)and

does not exceed /x.

solution

solution
as

or

which

which

is

involves

factor.

transcendant
particular

of the

form

then the
an

solution may be written in the form G{z, 0),


postulated
function of 0 ; and any function (otherthan 0 itself)
algebraic
which occurs
in G is algebraically
independentof 0.

it is easy to shew

that

Ovr

^[/m"(^)+{/m'W]^^3-G^.x(^)

the rightis an algebraic


on
function of 0 which vanishes
expression
when
0 is replaced
and so it vanishes identically,
by ef^{z),
by the arguments
it vanishes when
used in (I). In particular
0 is replacedby c 0, where
c is
of
But
its
value
is
then
z.
independent
The

^^

-'

(? (5,c

0)

{z).

116

THEORY

Liouvilles

4*72.
We

have

second theorem

'J
(^) i^ of order 7i]has

of order greater than

scendant

differential
equations.

"%W

linear

concerning

IV

that, if the equation

justseen

(1)
[in which

[CHAP.

FUNCTIONS

BESSEL

OF

then

n,

is

which

solution

have

it must

elementary tran-

an

solution of the form

0M (^)e/;{z),
where

/u.^

equationhas

If the

/;.

solution for which

Liouville's theorem, which


order

(logu^jdz is

of d

than

more

shall

we

equalto

solution

one

value be chosen, and

the smallest

has
/u.

of this type, let


let it be

called

u^.

prove, is that,for this solution,the

now

n.

Let
d

log u,

^
~

dz
and

then

is of order yu.at most

If N=n,
satisfied

by

If

N"n,

then

^1
+

solution whose

the

order N

^'

not

exceed

is

is gi'eater than

n.

iV

an

1.

"

If t contains
our

attention

on

We

than

more
a

call the ijrsttliree transcendants


transcendant

one

function
particular
t

solution of (2). The

used

and

this is

solution

differentiation with

for

in

" 4"7l,we

jF (z,6

all values

respect to

of

solution of (1) ; and

c=o

we

have
U2

so

itiJFg

dz,

that
du.y

of

(1) is

independent

l{exil"jF{z,e
c)dz]

write

c) dz,

_dc
is also

concentrate

find that,if iV^" n, then

find that the function

c, we

does

respectively.

type d, we
we

order

c)

solution
corresponding
exp

6, ^,

of this type, and

F(z,d
a

of the

the types

of

F(z,e).

By arguments resemblingthose

is also

equation

(^)'

function of at least one


transcendant
algebraic
of transcendants
whose
";fN-i{z),
ef^-iiz)and (possibly)
IfN-ii^),
Now

^/x.

t, namely

(2)
has

let the order of t be iV, where

requiredis proved.

theorem

the

dui
,

"

u.

of

z.

Hence,

defined

as

by

4-72,4-73]
But

DIFFERENTIAL

the Wronskian

of any

EQUATIONS
solutions of

two

ih'Fe
where

G is

If 0

u,

Hence

Mj is

an

consider any

function

so

(1); and

we

write

Cr{z,"),
used

in

" 4-71

it., defined

and

those used

earlier in

as

0=1

have

that

da,

This

is

Wronskian

and
constant, C'l,

so

function, not only of


Consequentlyu-^ is an algebraic
?r,but also of those of type 0 which

the types 6 and


^/l is

The

N, if N

^ ^

"

4"73.
We

shall

this is the theorem

to

that the

by writing
*

See

e.g.

t Journal
concerned

"various

Forsyth,
de Math.

with

Treatise
iv.

the

to
modifications)

"1841),pp. 1"13,

36.

1
,

on

Bessel's

hence the order

no

integral.
algebraic

uz~

i^

==

"v

"

is an

function
algebraic

of

z.

\-

DifferentialEquations (1914),"65.

(1839),pp.
general

equation

no

We

/a +

be established.

(otherthan a null-function)which
integral
firstreduce the equationto its normal form

has

therefore

is of order

u,

be greater than

cannot

has
that BesseVs equation
theorem-]-

shew

now

that N

shews
And

Liouville's

in t ; and

occur

of

n.

contradiction

d([ogu,)ldzis n.

all the transcendants

that
hypothesis

This is contrary to the

of order N.

where

was

we

expJG^(^,cQ-y)dz

du;

of

in t ;

occurs

dc
solution of

C :^0, and

in t.

occur

type " which

find that the function


7)

hypothesis
; so

it is an algebraic
function
similarly

which

of the

by arguments resemblingthose

this section, we

is

is contrary to

sJiCIF,).

of the types 6 and

and,

so

C,

algebraic
functionof 6 ; and

of all the functions

Next

(1) is a constant*; and

constant.

0, J" is independentof ^,which

117

case

429"435

; vi.

in which

equation was

Liouville's

(1841),pp.

4"7.

i^) is

polynomial; the

given

any

in his later paper.

first investigation

application(with

Journal

de

Math.

vi.

118

THEORY

OF

[CHAP. IV

FUNCTIONS

BESSEL

This is of the form


dht

where

xW="^^-i^^-i.

(2)

let Bessel's equationhave an algebraic


integral
; then (1) also
possible,
which
this
integral,
integral.Let the equation
algebraic
expresses
w,
f
unction
of
be
z
algebraic

If
has

an
an

as

64{u,z)

(3)
where

is

in

polynomialboth

and

ii

in

it is

; and

supposed that S"

is

irreducible*.
Since

is

solution of

have

(1) we

MuuMi-'l"4uz^uUz-\-r9lzz-S"r^^-6"^^ux{z) ^.

(4)

The

equations(3) and (4) have


(3) satisfy
(4).

of

For, if not, the left-hand


have

Let all the roots

is

rational function

of

for which

Let

any

this

Hence

z.

(3) be

of

; and

is not

sum

such value of

than

s" would

Mj, lu,

S4

itself,and

all the

roots

of ii)would

this would

be reducible,which

be

is contrary to

Then, if s is any positive


integer,

w,^.

...

there

hence

(3) and (4) {(luafunctions

sides of

factor other

highestccrmmon
polynomialin u and in
hypothesis.
a

root, and

common

is at

least

value of

one

exceeding

not

zerof.
be taken, and

let

M
"

tf

"

\
M

Also

let

Tf ,

(5

1)

(s

1)

'du.

^i z/^*-'"r'^X
\

m=\

where
prove

1, 2,

r=

...

Since

s.

u^, lu,...

u^j

are

dz J

all solutions

:Jof (1),it is

easy to

that
dW

'^"=,r.,

(5)

dW

~^=^Vr^^+

(6)

.^4 has

That

If not, all the

roots

Because

satisfied

is to say,

(4)is

of

no

factors which

(3) would

by

r{s-r+l)x{z)W,-u

be

are

polynomials in

(r

or

zero.

all the roots

of

(3),qua

equation in

".

in

or

1, 2,

in both

...

"

and

1)

z.

4-73]

DIFFERENTIAL

W^ is a rational function

Since
that

so

EQUATIONS

,..

of z, it is

"

A^ and

where

Bn,q are
only in

values
integral

Let the

constants,

of

a,y)in Wr is 1/(2

1/(2

there is

Hence

contradiction

higherpower

that there

shews

are

?"

have

the left of (7) than

no

terms

(5) and (6) it is easy

From
which

Tfo.is

in

of

term

of

(2r).(s

geometricseries
of

degreesX

2 and

"

is odd.

When

is even,

is then necessary

by

(,"
-

highestdegree

to

that the

shew

equatingcoefficients of
s\

-\A,.

,F, (1,

oFi{h,-^s; ^

the left vanishes


that

the term

analysisgiven by Liouville,

t The

point,because
in view

he
of

IV.

is; i

only of

the

hyper-

z^~'^ in

(7) is

hs] l)y.

hj; 1)

0,

(.9 1)
"

apparently

the fact that

only when
of

Journal

overlooked

the

here

is

X,

highest degree
de

Math.

based
Mem.

on

^^^

is

in

zero^f*.

xi^)

(1839), p. 435; see also Genocchi,


lxxxv.
(1877),pp. 391"394.
; Coniptes Eendus,

1-

7,

algebraicfunction
delle

made
Sci.

to

seems

fail at

failure

vanishing. The

suggestion

Accad.

^^

(1841),p.

vi.

possibilityof

('"^*^
(z)+ J'tn-h
jf^^j

subsequent part of the proof given


299"362

hj] l),-+a,

theorem,

Vandermonde's

It is to be remembered

!"; i

highestdegree on the left and rightof (7) are


Hence
TFo vanishes
which is impossible.
A. respectively,

XA^.sl

on
expression

\)..^,(i,

2r +

...

Math.

of

that the term

shew

1.3.0

pp.

\A",{'^s-^2)z^-\....

of

^^

The

are*

that the first ?" + 1 terms

the result of

is to say
so,

highestdegree in

of

be taken.

\Ak.s\

and

must

"""

3)

1 indicates
to

are

whenever

That

aq)~^ in W^

"

expressionfor W^

this

that the terms

see

possibleto

1) (s

is odd, the terms

Bn,q(z

identically.

complicatednature

more

the suffix

If

of the type

right.This

the

on

highestdegree in W^r+\ is

\A,z"^-' 2.4...
where

of

^;,^M.3...(2r-l).s(s-2)...(s-2r+2).

induction

An

it is

"

-K

\A",z^-\ A^sz\

simpleinduction

to

in Ifo, l^i" ^2, 1^3.

occur

A^z\

By

vanish

\j{z a^y.

.9.

...

J.a=^0, because

if it does not

last term

that

assume

now

may

1, 2,

on

n=

We

in TFo be

(5) and (6) that the highestpower

where
UqY'^'',

"

"

positive

assumes

ciq^ 0.

occurs

"

from

n
integers,

are

and

which
l/(^ a.q)

easy induction

by an

A.

the last summation

highestpower

It follows

and

in partial
fractions,
expressible

__?".?

119

of

z,

by "

3-4.

by Liouville,Journal
di

Torino,

xxiii.

this
evitable
in-

seems

The
de

(18(51)),

120

THEORY

We
s

proved that,when
in the form
Wn is expressible

have

is even,

therefore

where

does
J.o,s

[CHAP,

FUNCTIONS

BESSEL

OF

is odd, Wo vanishes,and that,when

~"

()

vanish.

not

expresses the coefficients in an


in terms
of powers of the roots, it appears that M must
of the sums
in the form
0 is expressible
and that the equationrS4 {u,z)
From

IV

Newton's

which

theorem

equation
be even,

'1 ii''-'^'-}%.(l/z)
0,

u^'+

(8)

r=l

where

^^

the functions

When

solve

we

of

series of

where

is

vanish

not

the constant

the series which

And

l/z,we

find that

in the

positive
integerand,

because

of

ascendingpowers
in the form
is expressible

(8) in

each of the branches

polynomialsin l/z.

are

of

case

branch

one

in the functions

terms

at

^,.are

least,Cq does
not

all

zero.

of the form

are

00

?n

convergent*for all sufficiently


largevalues of

are

When

substitute the series into the left-hand

we

the coefficient of the constant


be

Co must

zero,

thus obtained

On

4"74.
We

shews

the
now

are

contrary

is

in

what

in

so,

just been
equationhas no algebraic
integral.

order

of

prove Liouville's theorem f that Bessel's


has no
solution (excepta null-function)

in finite terras
expressible

" 4"73,we

an

odd

by
integer.

Bessel's

reduce

of

means

equationto

elementarytranscendental

its normal

form

"="^"^)-^{z)

Now

-I

write d

p(p

t Journal

t,and

"

p=

"

^.

have

we

^^^.,i_"0-l).o.

Goursat, Cours

convergence

and
l)/2''

(logu)/dz

(2)
*

find that

for every branch,


proved. The contradiction

has

positionto

(1)
where

(1),we

in finite
Bessel's equation
terms.
impossibility
of integrating

functions,if 2v is not
As

to

side of

in the result is Co, and

term

that Bessel's

equationfor functions
which

z.

of
de

d' Analyse,ii.

series derived
Math.

vi.

(Paris,1911),pp. 273"281.

in this

manner

(1841),pp. 1"13,

from
3G.

an

Many

treatises

algebraicequation,

assume
tacitly

the

4-74]
Since
has

order zero, it follows from

(z)is of

If

has

(2)

t, as
integral,

an
an

have

i.e.it must

is of order zero,

which

Since

irreducible

an

is

az

(4)
in the

As

polynomialin

solution

expresses

this

0,

t and

z.

have

solution of (2),we

have

must

algebraic
integral.

an

S4 (t,z)
S^- is

(2)

let the equation which


algebraicintegral,
of
function
algebraic
z, be

(3)
where

" 4'72 that,if Bessel's equation

finite terms, then

in
integralexpressible

an

121

EQUATIONS

DIFFERENTIAL

[x (^)

^*
1']

0.

of " 4*73, all the


analysis
corresponding

branches

of t

satisfy
(4).

of t, and let three


than two branches
First suppose that there are
more
values of u (defined
as
of them be called ti,to,U, the corresponding
exp ^tdz)
.solutions
of
all
Wronskians
functions
and
the
These
are
so
(1)
being "i, Uo, u^.

du-i
""'

'"''di
~

constants, which

are

Now

du2
~dJ

will be called

is
1.2

"

not

Mo

Cj ^ 0, and

would

U0U3

(and similarly
UsUi
Ui

therefore

u^

is

Next

suppose
LetTthe branches

and
that

be

so

=
.

has not

has

two

Mj

[F'+
be fixctorised so

Mo)is an

function
algebraic

oi

of

more

z.

than

branches, so
U and

This, as
two

we

have

seen

are

rational

4t7F=0.

and
integers,

"4-73,

quadraticin t.
functions of z.
By

that -SW{t,2) is
are

in

branches.

that

is constant, k,^ and

z.

V=Az^U{z-a,;)'''t,
where

-"

find that

^^^
V

have

and

U " \/V, where

in (2) we
substituting

Let

equation(3) would

Gi/{ts to).

function
algebraic

an

be the case,

cannot

the

zero,

were

so

But

and

,,

{is" 1.2}'.

U-iUs

"

therefore be reducible.

ii^Us

Therefore

dui
~d^

''"'

"

Jz

dvo
K'S -J-

"J

because, if it

zero,

pairof equal roots, and


Hence

'''

C\, C^, 0^.

dtt-i

ri

(J

^3

duo

dii-^

verifythat

it is easy to
'

and

dui

'''^~'''~cU'

'

k,, and

a,, are

not

zero.

122

From

rr
_

4^

and

consider the

Now

that

is evident

is greater than

of them

of the

principal
part
of the

none

numbers

4, which

that Kg is 0 or
the numbers
Kg are

"

Again,if we

consider

principal
part

the

the

1 in ^

"

Hence

have

t cannot

as

for
expression

the

where

An and

values

onlyin

Bn^g are

constants,

that X

"^

highest

Kg.

consider

and

fractions
partial

\l{z ag)cannot

of t in

IS

Ug ^ 0.
find that

(2) we
Anz-

(h=-k

(^q)

5,,,

^^
\Z

that

B\

^1^5=
consider

"

we

may

-PSp"1^
=

Ojg)J

0.

side

near

a^ we
the first and that

see

that

0,

principal
parts near

take ^_i

J'

1.

{A_y-A_,==p(p+1);

K=l,

"v-^,

the

positive

ii assumes
integers,

X, are

principal
part of the left-hand
in
^ to a higherpower
than
occur

if we
Similarly,

be

'

the

"

be rational.

it must

so

Zg {Z~ ag)-

and

and

l^^+

n,q\^

It then

that the

see

we

n.

'"^'''

Since p=

all

Hence

-K

this value

substitute

n=-K

SO

oo

(z),so

in

Anz''+

the last summation

nAnZ^^-^

we

one

1
n=

If

consideration.

from

near

2, and, if any

"

equation

branches

two

as

many

t=

we

It

a^.

hypothesis.

let
Accordingly,

If

less than

near

z) is reducible,which
/-4{t,
consequently

It follows that \JV is rational,and


is contrary to

the left

2.

"

cancel with

in V must

power

equalto

have

we

U,

excluded

both

are

be

the
satisfy

Kg + ^fCq
SO

of (5)

on
expression

can

k^

2 it must

"

'

4"{z- Clq)

the first member

into
by substituting

then

^9

""?

IV

that

(5) it follows

of

member

the second

[CHAP.

FUNCTIONS

BESSEL

OF

THEORY

without

-p

0 and

oo
,

0, ^o'
loss of

find that

we

-l.

generality.

follows that
u

z-Pe^''U(z-ag).
(J

if we
Accordingly,

must

this

have

replaceu by

solution

which

polynomialdoes not

z'P

is

vanish.

e^'^

in

(1),we

polynomialin

z,

see

and

that the

equation

the constant

term

in

124

THEORY

Solutions

4"8.
The

equation.
ofLaplace's

(" I'S) to
which
equation,
We

shall

Whittaker's*

the vibrations of

Bessel coefficients arise in

how

shew

an

problem of

in the

occurs

now

with

in connexion

be

generalBessel coefficient has been


equation,equivalentto Laplace's

the

of
analysis

in

first appearance

seen

[CHAP. IV

FUNCTIONS

BESSEL

OF

circular membrane.

natural

from

manner

equation
Laplace's

solution of

d-'V

d'V

d'V

The

solution in

questionis
^^

(2)

in which
In

/denotes an

/(^

^*

^^^

*^

*V^i^

^*' ^) ^"'

"TT

of the two

function
arbitrary

solution
a
particular,

variables involved.

is

ixcosu

ek(z+

iysinu) QQ^.^^^^^,^^

'

TT

in which
If

we

k is any

and

constant

is any

integer.

coordinates,defined by
cylindrical-polar

take

cos

(f),y

p sin

the

equations

^,

this solution becomes

gfcz
.

gikpcoi{u-4,)
(jQg y^n^du

e^^ I

"It

manner

and

this is

analytic

near

(v + 0) dv,

j^^y QQg

j,j0^y^

27ri'" e*^ cos

In like

2e^^ I eifcpCOS"
QQg

by " 2'2.

ei*p'="s"
cos

"IT

m(j" /," (A;/3),


.

solution is

"IT

equal to 27rt"* e^"^sin m^


the origin.

Both
J",^
(A:/?).

of these

solutions

are

nates,
coordiAgain,if Laplace's
equationbe transformed!^^ cylindrical-polar
it is found

to become

'^

dp"*

Monthly

Notices

of the R.

A.

'^
dp'^
p- d(f"^- dz^~

S.

lxii.

'

(1902), pp. 617"620;

3Iath.

Ann.

lvii.

(1902),

pp. 333"341.

t The
W.

simplest method
Thomson, Camb. Math.

of

effecting the

Journal,

iv.

transformation

(1845),pp.

33"42.

is

by using Green's

theorem.

See

4-8, 4-81]
and

DIFFERENTIAL

solution of this

normal

EQUATIONS
which

equationof

125

e*^ is a factor must

be such

and, if the solution is to be one-valued, it must


independentof (f),
is an
where
to
ni
minteger.Consequentlythe function of p
factor of V must
be annihilated by

is

"

that

be

equal

which

is a

'

dp^
and

therefore it must

be

dp

multipleof

p'

./,"
(kp) if it is

to be

analytic
alongthe

line p =0.
We

thus obtain

the solutions

anew

e^^

md)

J,n(kp).

sm

These

solutions

Maxwell's

of

derived

been

have

method

Hobson*

by

harmonics
differentiating

the sohition

from

with

respect

to

^^J"(kp) by

Clerk

axes.

Laplace'sequationinvolvingBessel functions has been obtained by


coordinates
the equation in cylindrical-polar
Hobson
from
by regardingcjdz
{ibid.
p. 447)
obtained
The
solution
is
as
a symbolic operator.
so
Another

solution of

^-

sin

arbitraryfunction

where/(2)is an
a

of the

function

function
London

of

kind

involves
of this solution when '^'"j
interpretation
Other
solutions
a
involving Bcs.sel
question.
been
given by Hobson, P)-oc.
arbitraryfunction have

shall

examine

now

the

"^

"^

the

from

I'TT

wtere
In

'

of the
of propagation
velocity

waves,

/(^

denotes

sin

cos

"TT.

F denotes
Loudon

Math.

Ann.

an

t
(1904),pp. 122"137,

sin

cos

+ ct, u,

v)dudv,

of the three variables involved.

is
+

1/sin

function
arbitrary

(1902),
and

silli} + Z":osu

+ ct)

^ /f^f^i\dudi*

(I

Matli. Soc.
lvii.

+ y sin

arbitraryfunction

an

solution
a
particular,

Froc.

equationis

i'TV

"ikixsmucasi"

where

dt^

cr

aspect.

same

Whit taker's f solution of this

(2)

the

motions

wave

dz'

dy'

represents the time and

492"494.

motions.

equationof

dx'

in which

(1895),pp.

xxvi.

of ivave
equations

^^

to

Soc.

Solutions of the

4*81.

is open

operator actingon an
xxiv.
(1893),pp. 55"67;

an

Math.

We

second

"'"V'^rf^

but the

xxii.

pp.

of

and

v.

(1892),pp. 431"449.
See
also Havelock, Proc.
H4o.

342"

xxxvi,
Watson, SIr.ssrnijcr,

(1907),pp.

London

98"106.

Math.

f^oc.

(2) ii.

The
is the

lies in the fact that it

solution
particular

this

physicalimportanceof
generalsolution in which

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

126

the

the

all have

waves

frequencykc.

same

and let ("",


be the
y\r)
{x,y, z) be (r,6, "^),
for
which
the
referred
to
new
axes
direction
(m,v)
angularcoordinates of the
polaraxis is the direction {6,4")and the plane -^ 0 passes through the
Now

of

polarcoordinates

let the

cos

ft)

sin

sin

(f)) sin

(v

(n,v)

{v

cos

shew

that

(f)),

sin ^fr.

to be

Sn (u,v)sin

we

then

may

u, where

notes
S,,de-

write

Sn {d,(/)
; Oi, ir),

in (w, ^) of

Sn is a surface harmonic*

co

"

Sn {U, V)
where

sin 6 sin

{u,v) of degreen

in

surface harmonic

cos

function
arbitrary

take the

Now

cos

trigonometrythen
spherical

of

formulae

well-known

^-axis. The

degreen.

thus get the solution

We

Vn

-TT.'
0

Sn is a surface harmonic

Since

sin codwdyjr.
Sn(d, (});ft),i/r)

gikrcosm

e^^"*

of

in

degreen

write

we
("","\jr),
may

Sn {6,"f";
(0,ir) An (0,(f")Pn (cOSft))
=

11

-\- S

sin mylr\Fn"'
cos
[A J"'\6,(f))
(cos(o),
m^fr + BJ""^6,(i))

and """'""
An^"'^{6,
(^,"^)are independentof
"/")
An{6, 4"),

where

to

and

-v/r.

with respect to yjr,


we
Performingthe integration
get

Vn

lire'^'^^An
(0,(t") e^fc'-cos.op^^
^qq^^) sin (odo)

=(2'7r)U-e''^"'^'I^^^An(

"

by " 8-32.
Now
t

by

the

the

equationof

same

factor,leaving6 and

"^unaltered

of the constant

Hence
to say,

lim (^"~''
Vn) is

is
r''An{d,"f")

motions, and
This

follows

motions

wave

is unaffected

i.e.if we
factor,

constant

so

k which
a

if we

and

t
multiplyr
{0,(f))
may
t.
r and
multiplies

multiplya;, y, z and
constant
by the same

that An

solution of the

equation of

be

to

be independent!

motions, that is

wave

solution

(independentof t) of the equationof wave


is consequently
a solution of Laplace's
equation.Hence An (^,"/")
from

the fact that

Laplace's operator is

an

invariant

for

changes of rectangular

axes.

t This

taken

is otherwise

obvious, because

S,^may

be taken

independent of A-.

4-82]
is

DIFFERENTIAL

surface harmonic

of

EQUATIONS
If

degreen.

to be any such harmonic,


An id,"/))

eikcty-hj

we

127

it to be

assume

to
permissible

take

obtain the result that

we

^kr)F,r (cos6)

^^^

vid"
sm

is

solution

of the equationof wave


frequencykc.

motions*; and the motion

representedby

this solution has


To

that

justifythe assumption

/
di- V

A,^{6,cf))
may

of the

solution

the normal

construct

,bV\
,cV\

drj

be any

equation of
/

0
c

sin 6 d0

^dV\
^dV\

dd

surface

a-'I'

1
^

harmonic

of

degreen,

we

motions

wave

sin^ 6

i^d^V

dcfi^ c"- dfi

'

cos

which

has

factors of the

form

t;'*"'

The

mcb.

factor which

involves

6 must

then

he of

sm

the form

P""*(cos6) ; and the factor which

involves

is annihilated

by

the operator

l)U-^r^
-(r'^"yn(n
dr
so

that if this factor is to be

Theorems

4'82.

analyticat

derived

dr

from

the

originit

solutions

be

must

of

the

multipleof e/^+^{kr)j^h:

equationsof

Mathematical

Physics.
It is

to
possible

Bessel

of the

equationor

Thus, if we

(or,at any rate,

prove
functions

render

to

by a comparisonof
motions.
equationof wave

cerning
probable)theorems convarious solutions of Laplace's

take the function

e^^J^i
+ a"
[k^J{p'

lap

"

^)],

cos

by making a change of originto the point(a,0, 0),we


coordinates.
of Laplace's
equationin cylindrical-polar
factor and it is analytic
at all pointsof space. It
a
in the form
expect it to be expansible

see

that it is

solution

This solution has e*-'^


as
is therefore

natural

to

00

e^^ J.ot/o
{kp)-1-2 2
?"

{A,ncos mj)

B"i sin m^) J,"(kp)

Assuming

the

consideration
in p and
/We

a,

observe that the function under

of this expansion,
we
possibility

is

an

and
function of cf),

even

A,n is of the

form

so

B,n

0 ;

c," is

Cm.Jm{ka),where

and, from the symmetry

independentof

p and

a.

thus get
+ aJo \k's/{p"

-lap cos

S e,"c,"Jm0"'p)
'fn{ka)cos

^)}

mcf).

111=0

expand both sides in


coefficients of (k-pacos "^)",
we
If

we

powers

Cf.

and

cos

get
C,)i=

of p,

Bryan, Nature,

i,

lxxx.

(1909),p.

309.

and
(f),

compare

the

128

and

expansion*

led to the

are

we

so

[CHAP. IV

FUNCTIONS

BESSEL

OF

THEORY

00

"/))} S f,",/m(kp) Jm (ka)cos 7n(f,,

2a p cos

+ a-Jo {k\/(p-

of which

which
take e'*"'^*+^*,

Again,if we

to

be

given in " 11-2.

is a solution of the

motions,

equationof wave

moving in the direction of the axis of z from


representsa wave
we
with frequencykc and wave-length27r/Ar,
expect this expression
to
00
00
be expansible
f in the form
which

and
+

proofwill

formal

more

I Cni^^J^^^^(kr)Pn{cOS
(^^"^kct
6),
where

is

Cn

constant

that

; so

r^y

gikrcose
=

\KrJ

If

we

the coefficients of

compare

nl

and

so

c"

^;

4*83.

Solutions

of the

be

analysis
justexplainedhas
equation
d^-V

"

normal

function

solution
of

and

of this

been

dxp^

dx.r

d'^V

c-

dt-

equationof frequencykc
vW-

by

the

case

of

which

is

as
expressible

"2-+

"""+

VX

the operator

j^

such

by Hobson" to

only,where

be annihilated

so

dimensions.

d^ p-1
and

extended

...

ofp

space

d^V
1-

r=

must

given in " H'o.

d-V
1

dx{'

find that

expansion:}:

equationin

wave

The
the

side,we

,1=0

proofwill

formal

more

each

(kr cos 6)^ on

''"2'"+*r(n
+ |)'2".(w!/'

\icr/

of which

n=0

thus led to the

are

we

c,,i'^/"^.
(kr)Pn (cOS6).

(jr

solution,containing
time-factor e^*^^
a
must

be of the form

e""'''^^^p.,,(kr)/{kr)i^p-^K
*

This

t The

is due
tesseral

to

Neumann,

harmonics

X This

expansion is due

" Proc.

London

Matli.

Theorie
do

to

Soc.

not

occur

der

Bessel'sclten Functionen
because

Bauer, Journal
xxv.

(1894),pp.

the function

fur Math.
49

"

75.

lyi.

(Leipzig,1867), pp.

is symmetrical

(1859),pp. 104,

about
106.

59"65.

the axis of

z.

4-83]

DIFFERENTIAL

describes the

Hobson
of rank p

such

succeeded

Hobson

in

be written in the form

combined

notation

By using this

function
a
as
ikr)/{kr)^^P~''^
quotient'(^jip-a)
cylinder

function may

with

an

example of

provinga
equalto

Xp

no

cos

when

shall consider

we

a"

odd

an

for

'

%ir

expansionfor

an

p],
"/")

cos

regard(^as beingconnected with Xp by the equation


is a solution of the wave
by e'^'^*
equation,
0. This function multiplied
it is expressible
write p
r sin "p,
as a function of
we
p, "^,t and of

it is convenient

and

\k s/'if+

^j-dimensional
space,
of
functions
cylinder
integersimultaneously.

of theorems

half

such theorems
J

where

the concept of

number

order and of order


integral
As

129

EQUATIONS

to

other coordinates.
Hence
e''*"*./
{kV(r-+

is annihilated

by

by

dp

dr

dec./

'

are

cf) d(f)

sm

r-

functions which

normal

"/")
jp]

cos

the operator

dr^
Now

2ar

the operator

dp'^
that is to say,

a-

annihilated

r-

d(f)'^

by this operatorare

of the form

oa

Fn (cos^ !^)is the coefficient*

where

(l -2a

By
J

the

used
reasoning

+
{k\l{r-

a^

2ar

"

in

cos

of aV' in the

cos

" 4*82,

we

expansionof

+ a'y-^p.
(f)

infer that

"^)\p]

1
=

/N^ow

{ka)Pn (cos"t"p).
i]^r)
Jn+,,,.,
^n-/"+i^-i
{kap=Hh^^^^
all the

expand

on
cos
{k'^ar
"^)'*

each

side ;

Bessel

functions

and

equate the

coefficients oi'

find that

we

2^

2" r

A^

(?t+ i/J
-

)
^

"

^M+hp-,
so

that An
*

So

,^ 1

1^ ^^^
+

2flP-^(n +

ip^

2"+iP-i r {n +

h^p-l)V

{hp

n\ r

(1^

1)

'

1).

that, in Gegenbauer's notation,


I'u (cos"p\p)^

W.

lp)f

15. K.

C^/'-^cos./.).
!)

130

OF

THEORY

expansion

thus obtain the

We

{kV(r2
/jp_,
(r^+
2iP-

a^

2ar

"f))]
(f))^^'

2ar

+a^-

"

[CHAP. IV

FUNCTIONS

BESSEL

cos

cos

^^ i
(*^
'/\;f_,

*p-l)e/,^i^-i(^^r)/^,p_,(A:a)Cf-^(

which holds for Bessel functions of


proofof this expansion,
analytical
will
all orders (thoughthe proofgiven here is valid onlywhen }) is an integer),
be givenin " 11'4.
An

4"84.
Two

Batemari's solutions

of the generalised
equationof wave

solutions of the

systems of normal

rr

a^

8^

motions.

equation

d-V

1 B^F
_

have been
the two
If

also established

connexion

between

systems.
take

we

variables p,

new

equationtransforms

the

who

investigated
by Bateman*,

normal

Xi

^2

a,

^y

cos

"^ defined by

the

-yfr,

^,

Wx

cos

p sin Xr

^i

(^

sin -\/r,

equations

into

solution of this

equationwith frequencykc

is

J^ (kpcos "t")
J^ {ka sin "f")
e' ('*x+''"^+^fO^
where

"I"is any

constant.

Further, if we

write
p

so

that(r,
^) form
%, i/r,

?" cos

of

system

sin

(f",(T=r

"^,

polarcoordinates,equation(2) transforms

into

^,g.
'

^ir.??7,l?!r
ar^

dr

r-

cot"/"-tan"/)aF
"^

d(l"^

df

r^

a'^F

"*"

Now

normal

annihilated

are

K^" +
or-

solutions of this

7, 5^ +

Messenger, xxxm.

cos-

"^a^'

equationwhich

a^Fia^F
~

r-

have

sin- "/"
a-"/r-c^ 1)"

'

e""('^x-l-'"/'-t-^cO
as
a factor

the operator

by

rdr

1
^

r^

'^ +

^
v

loT^

6
(cot
v
r

[a"^2

(1904),pp. 182"188;

"

tan

"i)
r/
^^

Proc. Lo7idon

-^r~

Math.

-.

^^^^^
Soc.

sm'

)"
,
'

"j"\

(2)in. (1905),pp.

111"123.

CHAPTEK

PROPERTIES

MISCELLANEOUS

this

have

shall discuss

chapter we

place in the
namely
common,

found

not

feature

one

in

preceding chapters,and

that

they

all obtainable

are

indefinite

some

(^)
1%"+^'^.
dz

(1)

l~z--^^

(2)

"^, {z)dz

these formulae,
generalise

To

which

by

which

have

but
of

processes

integrals.

" 3'.9 (5) and (6) at

formulae

recurrence

functions

Bessel

character.

shall first evaluate

The

propertiesof

some

two

definitely
elementary
We

FUNCTIONS

BESSEL

integralscontaininga singleBessel function.


Indefinite

5*1.
In

OF

z^+^

the

to

results

^C+i {z),

'e,^ (z).

z-"^'

lead

once

consider

j%''^^f(zy^^{z)dz;
integralbe equal to

let this

+ B(z)K^,(z)l
z''+^{A(z)"^^(z)

where
The

(z) and
result

{z) are

to

be

determined.

of differentiation

^"^Vl^) K

(z)^

z"^'

is that

\a' (z)"^^{z)+
+

In
take

order

(^)

that

{z) and

B' (z),and

{z) may

{z)

z"-^^

^''^"^^{z)-A{z)^^
(^)|

[B'{z)-^^.^j
{z) +

depend

not

on

the

{z)"W. {z)\.

cylinderfunction, we

then

f{z)^A'{z)^^^^A{z)^B{z).
Hence

it follows
'

(3)

z^^^

that

^^
\^"{z)
+

B'

(z) -f B

z^^^

This

result

formula
die
due
pp.

(with

BesseVschen
to

was

obtained

diflPerent
Functionen

Nielsen, Nyt

by Sonine, Math.

notation) had

been

(Leipzig,1868),

Tidsskrift,ix.

(z)].
^^ {z)dz
[E {z) ^^ {z)+

Ami.

p. 70.

(1898), pp.

{z)%\^, {z%

(1880), p. 30, though

xvi.

obtained

Some

73"83

developments
and

Ann.

di

ber

and

equivalent

an

previously by Lommel,

Studien

of formula
Mat.

(3)

vi.

iiher

(3) are
(1901),

43"46.

For

some

associated

Quarterly Journal,

XLii.

integralswhich
(1911),pp.

338

involve
"

340.

the

functions

bei,see

Whitehead,

5-1, 5-11]

MISCELLANEOUS

THEOREMS

reduction formula, which


following

The

is

133

obvious

an

of (3),

consequence

be noted:

should

z'^^''",{z)dz

(4)

ifji'V-

r z"--'K

'C+i {z)+ {^J^-v)zi^^,{z)'\.


[^'^^'

5" 11.

two cylinder
containing
integrals
functions.

Lommel's

which
simplestintegrals

The
derived

the Wronskian

from

{z)dz

contain

formula

"

of

312

are

those

which
integrals

contain

^^^^

functions

(2),namely

/_. {z)j; (z)

/. iz)J'_. (.)

Bessel

two

TTZ

which

gives
dz

"

(I)

J^^{z)

2 sin

VTT

./.,
(Z)J_, (Z)

2 sin

VTT

^^

^^

/, (z)

'

from " 3"6o (1),


similarly,

and

dz

^^

V^{z)
2J^{z)'

; zJ^H^)

(3)

'

77

zJ,{z)YAz)-2'''^
JA^)'
dz

(5)

two

given

TT

zY:-{z)
reader

The
any

J-.{z)
J^ {z)

IT
^

are

should

have

cylinderfunctions
due

to

Lommel,

(3) with the result due


Some

of the
Math.

to Euler

order

same

Ann.

verified

which

in the denominator.

was

quoted

"

Math.

Ann.

The

to

xiv.

The

fornmlae

reader

should

actually
compare

1 '2.

Lommel*,

by differentiation that,if y

in

due

f (P-Qyy'}'^^

then

similar

(1871),pp. 103"116.

rv.

more

once

F,(2r)'

in evaluatingthe
difficulty

no

results,also
interesting
of Bessel's equation.
genei'alisations
It is at

J^ (z)

are

obtained

the
and "?;satisfy

"'''""

(1879),pp. 520"536.

from

equations

134

THEORY

Now

apply this

^v denote

I,

result to any

equationsof

two

functions
cylinder

any two

[chap. V

FUNCTIONS

BESSEL

OF

" 4"31 (17). If


have
we
respectively,

the type of

of orders \x and

".,^^^^.,.^^Ap^
dz

(6)

dz

cj^'iz)
4"(z)

'^ ^'^"' +^^l^(-)n


4|^P(7)I
-2;p(7)^
"

where

functions
arbitrary

are
(z)and y{r(z)
(f)

This formula

As

use.
generalto be of practical
multiplesof 2^,say kz and Iz. It

be

{k^-l^)z-

^^.

is too

and ^{^(z)
to
(f){z)

(7)

of

IX'

"

take

special
case,

is then

found

that

%(kz)'"^Mlz)dz

\k%^, {kz)f {Iz) m^ (kz)^?,+i


-ifji-v)9^^{kz)%\ {Iz).
(Iz)]
-

The
k
(ii)

If

take ix=

we

(8)

{kz)

This formula
It becomes
numerator

v,\i is found

6
^

may

{tz)dz

(i)/u,i/,

cases
special

is

that

k''-V

be verified

nugatory when

the expression
on
by differentiating

k=l, for the denominator

is then

the

right.

zero, while the

constant.

If this constant
when

stillfurther in two
the left simplifies

on
expression

l.

is omitted, an

of THospital's
rule
application

shews

that,

l-^k,

(9)

z"^^{kz)%\ {kz)dz

result of

[kz^^^,(kz)'6; {kz)

kz90^(kz)f^V+i{kz) '^^(kz)#^+1(kz)}.

The

-^

using recurrence

formulae

to

the derivates

remove

on

the

rightof (9) is
(10)

z^^ (kz)9^^(kz)dz

\z-'[W^ (kz)%% (kz) 9^^_,(kz)%%+, (kz)


-

-^,^,{kz)"f,-,{kz)]

5-11]

MISCELLANEOUS

of these formulae

Specialcases

K' {kz)
f~^

dz

(11)

{kz)f
^~z'W^
_^

are

135

^z' {'e;~
{kz) '-^^_.
{kz)'^,^,{kz)]

|(l^\ 9B,'{kz)

Iz^-

a 2)

THEOREMS

{kz)dz

{2'^^^
{kz)'f
\z"-

(/.^)J

'^;^

_^

{kz)+ W^_, {kz)Wf_^_,


{kz)

9?,^,{kz)^_^^,{kz

the

latter

function
To

equationbeing obtained

of order

obtain

is found

"

by regardinge~'^'''-Yi}
^{kz) as

cylinder

fi.

different class of

take
elementaryintegrals

k=l

in

(7) and

it

that
^" ^^'*'
^'"^^' ^^"^
iic.)^=r fe-^(fe)-g,

"^"^'^'-"^'^^'
^'^'

"

(13)

^S^{kz)^,{kz)
The

result of

(-14)

?^

making

z/

in this formula

^-/a

^
(^"^)6. ^ (A;^
T

""

last

'

~a

'

0 fJi

^g'M(A:^)^^.(A^-^)
^M+i(^)l
_^

^^ (yr,^)

(15)

we

mtegrating,

and

then

re-

have

j ./m-H^^)^
=|^{/..,(A-,^)!i|M(^-^)-/.
2^^.M^-4
+

method

of

equivalent to (11) are


(Paris,1822), "5^318"319,

formulae

Lommel's
Various

result will be

obtainingthis

Results
Chaleur

V,^^^^=2/.t^(^)

kz.

Arfalternative

other

by multiplyingthe equations

~^"^\^) respectively,subtractmg and

case
special

placingz by
As

'

'""^'

equationis also readilyobtainable


^.'^,(^) 0,

"

2~ 1

The

is

of this section

as

seem

old

in the
to

as

Fourier's

treatise,La

of functions

case

have

been

givenimmediately.

of order

discovered

Analytiquede

Theorie
zero

before

but

the

none

la

of the

publicationof

memoir.

specialcases

of the

formulae

Napoli Rendiconti,(2)ni. (1889),pp.


XII.
(1902),pp. 99"108.

91"99

have
and

been

worked

out

in detail

by Chessin, Trans. Aaid.

by Marcolongo,
of St Louis,

Sci.

136

THEORY

cylinderfunctions;

Indefimte integralscontainingtwo

5'12.

[CHAP. V

FUNCTIONS

BESSEL

OF

LommeVs

second method.
An

alternative

integrals
justdiscussed.

of the
form

been

has

method

The

consists in

method

^ {z"^^ (z)#. (z)]


=

this method

By

suitable for numerical

more

Lommel*

givenby

their values

some
evaluating

are

obtained

in

computation.

adding the

results

two

(z)W,^, (z)+ %^, (z)W'.(z)]


{'^,

-z"'

(p +

v) z'"-'1^^{z)W,{z),

^+

z""-^ ^^+, {z)%\^, {z),

^{p-^-v-2)

that

so

(p +

v){~z^-'
%\ (z)^,{z)dz+

/x +

(p-fM-vz^

and

then
Thus

(1)

2) r

(2)
As

^^+, (z)W

,+1

(z)dz

{z)+ ^,+, (z)^,+1(z)l

givingspecialvalues to p.
we

have

fz-"^-''-^%^Az)'^^^,{z)
dz

'^e(^)%%
J ^'^""-''+i
cases
special

(^)f/^

g^^^^_^^^

c^^

['

'

z-^-+^'^C^
{z)dz

(4)
Again,if p

be made

zero
,

{^, (z)^^ (z)+ -e+^ (z)'^^^,


(zy^.

4^ l^,'(z) "^W, (z)},


^^^("^,^(z) "^^,+,(^)}.
+

it is found

that

that,by summing formulae


if.+

^'^+ '^'^-^^
^'^'^"^^^^"^'

of these

"^V+i(^)
f^-'"-'

(3)

(5)

ZP-'

re (z)f

^"^^
^'^^^
2(/7^+l)

so

for

K (z)%\ {z)+ ^^^, {z)%\^, (z),

of this type, we

get

%\ (z)W,(z)^-(p,
2n) j'
W,^,,(z)^?.+,
v)f
(z)^
+

Math.

Ann.

xiv.

(1879), pp. 530"536.

5'12-5-14]
if /u.
particular,

In

0,

z^

137

THEOREMS

MISCELLANEOUS

dz

j %\{z)'"n{z)~

(6)

r,

"-ll

r-

'^0(^)^0 (^)+ 2
m

where

1, 2, 3,

there

But

seems

+ '^niz)'^n{z)
'",n{z)'^\n{z)
I

to be

no

....

j%{z)%\ {z)^
For
I.

xi. (1881),p.
Rayleigh,Phil. Mag. (.5)

(1)see

for

Papers,
{Scientific

217.

(1899),p. 516.]

So7mie's

5 "IS.

in
expressible

be

extended

by Sonine,Math.

Ann.

(1880),pp. 30"33,

xvi.

that

the form

-f {"/^(^)}
{0 (^)}
{^ (^)H^ (s)-2?^^,
(^)-^M i'^(^")J;5.

but the results

been

of conditions

the discussion

may

functions.
containingtioo cylinder
integrals

analysisof " .5-1 has

The
to

of

specialcase

simpleformula

+ C

+i" (.-)
{0 (z)}
^^^,{^ {z%
-gf^^j
(z)"^^{0 (.-)}
%-^^^{V.(3)}

are

too

complicatedand

not

important to justifytheir
sufficiently

insertion

here.

reduction formula.
Scliafheitlins

5 '14.

reduction

formula

for

I'z"^9^%''{z)dz,
is

which

natural
and

Schafheitlin*

^v{z)
To

as

varies

appliedby
(" 15"6).

the formula

obtain

of the formula

extension

we

him

to discuss

observe

" 5"1 (4),has been discovered by


of
the rate of change of the zeros

that

'

2"^{z'-v')'i^J'{z)dz

--fz^W.i^)\z^^^z^m^)dz
'

-^ [_ ^^+2^"^_
(^)r^; (^)]+
Now, by
+
(/j.

^;2
I {^^+-2

(^)+ (^

1 , ^^+1

-i^^
(^)'^-;(^)|dz.

integration,
partial

[z"^+''^\''{z)]
S)\'z'^+"(^J-'(z)dz
=

Berliner

1%'^+'
{z) +
K' (z)[z"^:

v. (1906),
p. 88,
Sitzunf/nberichte,

dz,
{z"- v')K (^)l
-

138

THEORY

and

[CHAP.

FUNCTIONS

BESSEL

OF

so

+ 1)
{fjL

'^/^{z)dz
I%'^+^

Hence,

r^'^+^K"

{z)]+

'^,(^)-g^;(^)dz.
(^'^ v'^)
['2'^+^
-

substitution,

on

{lji+\)\%"^{z'--v')9^-'{z)dz
-^/^(2) (/x+ 1) z^^' '^v(z)'^J (z)]
[^'^+"

1%^^+-'
9^.(z)%%' (z)dz

{fi+3)

[V+^ W^

and

{z9S:
[^'^+1
(^)
-

this is the reduction

5'2.

Expansionsin
shall

We

in " 2'7.
(1^)'"^
Chapter XVI.
result of

The

" 27

'i'/{z)
J^*^

dz.

v'\

+
1) '^.(^)}^

in

at

^f^

z^9Sj'{z)dz
^'

^'^+^

[z'
-

7.^+

H/^

(^)].
^)1'^^.^

question.

of Bessel functions.

series

discuss

now

i (m

formula

for

for

dz

+ lf-v^z"^+'^ii,'{z)]
{hAjJi

{z) (/x+ 1) {1{^l+ 1)^

2)['
{ti+l)[v'-\{fi+
z"^+^^i^fj'{z)dz
+

(z)W;{z)
1%'^+'^,

-22.^1

find that

we
By reari'anging

{fji
+

z"'+'"^.{z)
"^J {z)+ z"^+-''g?,2
{z)

"W:^(z)-(fi+1)
[z"'+'

+ If
{{fi

of the

tained
simplestexpansionsof the type obgeneraltheoryof such expansionsis reserved

some

The

of the expansion
suggests the possibility

once

00

+
(i.)^=s(^i"MIV
.")^^^^(,)_

(1)
is due

which

establish the

To

is

Gegenbauer*

to

series of

and

is valid when

expansion,observe

fi is not

negativeinteger.

that

analyticfunctions which converges uniformlythroughout any


of the ^;-plane
(cf " 3'13);and since

domain

bounded

J^+on(Z)]
{(i^)"^

it is evident

(^z)~'^

(z)-(fl+
{nJ^+^n-l
-^_^
^^

that the derivate

of the series

now

under

n) J^+2n+l(z)],
consideration

i"n

is

1=

n-

Wiener

Sitzungsberichte,lxxiv.

^"

(2),(1877),pp.

124"130.

0,

140

By

we
analysis,

of the

slightmodification

[CHAP. V

FUNCTIONS

BESSEL

OF

THEORY

that, if k is

prove

may

any

constant,

oi^i
{fi+

This

formula

in"

n, -n\

+ 2n)J^,+^n
1\ A--)
(yu,
{z).

some
requiredin establishing

will be

generalexpansions

more

11-6.

+ h)].
\\/{z
of {z + A)**''./^
expansions

LommeVs

5'22.

It is evident

+ h)],
A)~^"J"^[s/{z
qua

that {z +

" .3"21 (6),w^e

Am

00

J. {V(^+ h)]=1
(.+/o^''

These

formulae

(2) we

deduce

are

from

h, is analytic
combined

[z-i^j.
(v^)l

-^

-.

some

yirzj

(4)

--^

sin

z-\-h

+ 2zt)
^/{z^=

0; and

so,

in

J.n-i(z),

m.

,n=o

"

when

^ in (1)and
^
slightchanges in notation,
take

we

2zt)=^

^(z^

cos

(V^).

{.i^/.
(V^)]

If

" 3*4, after making

f+Wl
m

dtie to Lommel*.

l^\

(3)

(Jm

+ ^)} is analyticexcept
{\/('2'
Again, (^+ A )"'/,.
providedthat |A |" j^^ |,we have

(2)

have

{z + hr^^j.yiz +h)]

(1)

of

theorem
by Taylor's
consequently,

for all values of the variable,and


with

function

./i_^(z),

"

equation(4) being true only when |i|"^|2^j. These formulae are due to
Glaisherf,who regarded the left-hand sides as the generatingfunctions
odd integer,
associated with the functions whose
order is half of an
just as
efficients.
(^ 1/0} is the generatingfunction associated with the Bessel coexp f^^'
"

Proofs

been

givenby

is somewhat
*

of (3) and

Studien

Glaisher ; the

algebrainvolved

in

of this
investigations

nature

formidable.
ilber die BesseVschen

by Bessel, Berliner

Abh.

t Quarterly Journal,
Trans,

(4) by direct expansionof the right-handsides have

of the Royal

Soc.

1824
xii.

Functionen

[1826],p. 35,

(1873),p.

clxxii.

136

(Leipzig,1868),pp.
for the Bessel

; British

Association

(1881),pp. 774"781,

813.

11

"

16.

Formula

(1) was

given

coefficients.

Report, 1878, pp. 469"470.

Phil.

5-22]
shall

We
In

(1) replacez

J, {W(l

(5)

// by

and

modifications

various

enumerate

now

h)\ (1 + hr

(1) and (2).

of

^"2^and then

and

z-

141

THEOREMS

MISCELLANEOUS

;;,

-^

/.^",(^),

and, in particular,

If

/.(^V2)^2^^

(6)

from

In like manner,

providedthat |.^^|"
A'

make

we

A-)H (1 + AO-i" i

when

is

an

1 + 0,

the series

W'i

integer.If

rightis convergent. The convergence


is not an integer,
then, for largevalues

the

on
v

is obvious
of ?",

"^

in]

TT

the condition

is any

^-^^^/.-,"(^),

m=0

is R {v) " 0, and

for convergence

the convergence
satisfied,
when

find that

1, we

"

-^

find,by Abel's theorem,

we

"??l!

Hence

1.

A--*-l+0

providedthat

make

J^^-Az).

(2),

J. {W(l

(8)

If

then

and
(5) by (1 + A:)^"

divide

we

"7f

when
Consequently,

is absolute.

if the

condition

is

R (v) " 0, and also

integer,
S

(9)
if

In like manner,

^-^

R(v)"

^f^J.-,"{z)

0.

also when

1, and

"

is aiii/ integer,
we

have

/. (W2)

(10)

2-^"

^^

J,_",{z).

III--

It should

be observed

for functions
and

j^"I"

1 ;

of the
so

that functions
first kind

in

of the

may

be

substituted

(8) providedthat |A j" |^ |

(1),(2),(5) and

that

111 =

(12)

r, y{2
(z + h)i''

h)]

(13)

F. [zV(l

AOJ
=

^i^^]:"'^!-"-)
Y,_,,wz),

5
(1

"*-!

Y^ {zv(i

S
Z^)^"
/"

(14)

kind

second

^7"^^iW-" (--),

t ^
^01 (1 + k)-^-"
=

}-;_,"
(4

142

THEORY

[CHAP.

FUNCTIONS

BESSEL

OF

the functions of the second kind as a


may be proved by expressing
of functions of the first kind ; by proceedingto the limit
linear combination
These

when

tends

to

integralvalue, we

an

they hold

that

see

for functions

of

order.
integral
the

correspondingresults
substitute the symbol W

By combining (11) (14) with


"

kind, we

first

the

that

see

we

may

for functions
for the

of

symbol

throughout.
last formulae

These

234

(1918),pp.

by Lommel,

noted

were

Chapter xi.
that
242,
they are

"

Stadien, p. 87.

of

some

Numerous
generalisations
by Airey, Phil. Mag. (6) xxxvi.

observed

It has been

given in

will be

of them

in calculations

use

connected

with

of

zeros

Besael functions.

When

that, when

(5) and (13),and

combine

we

|X^

"

(_\m

x^

when
and, in particular,

("\m

results

two

X-' S

/. (X^)

"g',^",
{z),

is unrestricted,
00

These

find

we

il^\m

(\-2
-\\m
U__^^_Ji-Ai^

00

(16)

replace\/(l-\-k)by X,

1" 1,

^. (M

(15)

then

(\2
^^

-\\m.(lp'\m

^''^

described*
fi*equently

are

J.^,n{z).

theorems
multiplication

as

for

Bessel functions.

(when

observed

be

It may

is taken

the result

that

equal to

(2/zr

-{a-l)\

(17)

of

treating(14) in

i-i^

?n=0

An

alternative

proofof

xiii.
Sitzungsberichte,
see

formulae) Wagner,
A

[Note.
Lommel

Bern

Ann.

special
years

before

27r

as

(8) is that

the

of formula
the

F"_,"(4

has been

aid of the methods

pp. 115

given by Bohmer, Berliner


of complex integration
;
extensions

(fornumerous

of the

119; 1896, pp. 53"60.

"

(1),namely that in which

publicationof his treatise ;

see

1, was

discovered

der Math,

Archiv

area

consisted, in

taking the integral

I I c,os,{^r
cos

of the circle ^~+

with resi:)ect
to
integrating

sin (^r cos

difif'erentmethods.

rj is

^i^-^")

cos{^cose)iimU{l-P).riim6}~^^

^^
7 _i
(-)'"(lrsing)2"^
J^^i(?-cosg)

(2?n+ l)!

"7r"i=o
-

ml

TO=o

Schafheitlin,Die

two

di
6
_-v'(i-{=)'''sin

6 + r^r sin 6)

e.g.

6 + r]r sin Q) d^d-q

and evaluatingit by
rf'"l.,

_i

See,

by

xxxvii.
"

result of

The

way

"

(1904), p. 108, and

lix.

^
the

same

356.

His method

over

''''

formula
multiplication

Mittheilungen,
1895,

case

seven

(1861),p.

the

(1913),p. 35, with

also Nielsen, Math.

the

integern)

an

Theorie

der

Besselschen

"=

b/

'

(/-cos^)'"*!
Funktionen

(Leipzig,1908), p.

83.

5-23, 5-3]

MISCELLANEOUS

changing to polarcoordinates

the result of

and

'2n

"TT

{ri}cos

cos

{(f) d)}p dp d(f)


=

"

these equationswe
If we
compare
and r^ sin- ^.]
r'^cos''^^

obtain

expansionof a

Bessel

5*23.

The

From

formula

Bessel functions
If

fjb and

The

/""^ r^

"

which

"n

the

(f))
pdpdcf)

I)

case

functionas
has

(rpcoh

cos

./

i/

l with

series

deduced

and

replacedby

of Bessel functions.
series
interesting

an

of

represents any given Bessel function.

unequal,and

are

(1) in

(7),Lommel

5'22

^"

"^"n'

143

{p,(p)is
1

/"jr ri
I
I

1
-"

THEOREMS

jx is not

we
negativeinteger,

have

repeatedseries is absolutelyconvergent; consequentlywe


have
it by replacing
n, and then we
p by tn

may

arrange
re-

"

and

hence, by Vandermonde's

This

formula

1868), pp.

putting

22
=

"

case
23, in the sjjecial

0, it is found
=

/i

/i=0,

ilher die Bessel'schen

with
0; by differentiating

have

we

Lommel's

MJ

"

An

compared

addition

with

Neumann's

(1)

of

i
7/1

the

"

|2; |" j^ ], being unrestricted.


similar but more
generalformula
%\{z

to
*

ilv

"

3'571.

"

of any

order is

J.-"^{t)J,n{z),
00

t)=
m

is due

"

to Be.ssel functions

This

(2)

lit

expansiongiven in "

" 2'4

J,{z + t)=

where

^^'ff
"f
;/'^

formula for Bessel functions.

of the formula

extension

An

formula

is due

to Schlafli* ; and

%\-,n{t)J,.{z)
X

Soninef.

Math.

Ann.

then

"J/l^\

53.

and

formula

be

(Leipzig,
v

l"l

m=0

1^ Y, (z) ./o(^) {y+ log(1-0}


+
should

respect to

i ^-^~^{"r-''J,..,n{z)
J,{z)\0g{hz)-''^f^-}^
"}

(3)
This

Functionen

that

and, when

Studien

given by Lommel,

was

knY,{z)

(2)

theorem,

in.

(1871),pp.

13-')" 137.

^'^'''-x^'i-

(1^80), pp. 7"8.

144

THEORY

It will first be shewn

the

that

series

where

r,

R, A

When

in

sin

when

ascendingorder

Ju-m(i)Jm(^)
positive,

is largeand

uniformly

7? ^ I^ I^ A,

r,

numbers
unequalpositive

are

and

is

rightof (1)

the

on

functions of both
convergent series of analytic

I2^ I^

[CHAP. V

FUNCTIONS

BESSEL

OF

vir.i^Ry.ir/Ry

is

(m

of

magnitude.

comparablewith
v)

and

of the
the convergence
series for {l"rjR)". When

series is

comparable with that


largeand negative{=

is

of the
"

binomial

n), the

general

is comparablewith

term

V{v
the

and

uniformityof

n+\).n\

follows for both

the convergence

of values

sets

of

iii

by

the test of Weierstrass.


is

Term-by-termdifferentiation

il-P) i
\"jt

so
consequently
permissible*,

J.-.n{t)Jrr,{z)=l

CZj^=-tx^

\J\-n,{t)
J^{z)

m=

"x"

"

""

J^-m (0 [Jm-l{2) Jm+i (^)}


-

rearrangement, that all the

on

J' m{z)]
,_,n{t)

it is seen,

-ao

1,1=

and

that

terms

on

the

rightcancel,so

that

Hence, when

12^ ]" ji| the

series

2
m

of

and

with

F{z

which

respect
+

is

to

1),then

we

put

0,

we

that

see

"

oc

and
{t)^J^{t),

the

truth

evident.

Again,if the signsof

and

in

in=

and

when

(3)

function
analytic

as
a
expressible
functionof z-\-t only,since its derivates
and t are
identically
equal. If this function be called

m=

If

J^,-m(t)Jm {z)is an

this result is combined

have

"

with

(1),we

see

that

I Y^_rr,(t)J,n{z).

Y,{z + t)=
m=-

(1) be changed,we

Cf. Modern

CO

Analysis, "5-8.

of

(1) becomes

5'4r]

MISCELLANEOUS

with

this is combined

When

reader will

The

(1),equation(2) becomes

J,^,n{t)J,,{z\

'i^,{t-z)=

%%^,,(t)J,n {Z),

Ann.

is

iiber die BesseVschen

Studien

Functionen

by Graf,
given* explicitly
of these formulae will be
generahsations

integer; while (4),(5) and (6) were

an

(1893),pp.

XLiii.

|,

";

"

given by Lommel,

results,(3) was

1868),when
(Leijjzig,
Math.

i"

|^ |"

F,+",(0 "/".
(4

2
)"

evident.

that,when

Y,{t-z)=
these

method

same

J,{t-Z)=

(6)
Of

the

readily
prove by

(4)
(5)

145

THEOREMS

141

142.

"

Various

given in Chapter xi.

of Bessel functions.

Products

5'4.

ascendingseries for the productJ^ (z)J^ (z) has been givenby various
stated to be due to Schonholzerf,who
writers;the expansionis sometimes
published(in1870)
publishedit in 1877, but it had, in fact,been previously
examined
been
has
More
by Orr",while
by Schlafli^.
recentlythe product
has givenexpansions
(c" " 5-42)for productsof the forms
Nicholsonjl
The

l^^(z)Yn (z).

J^ (z)Yn (z) and


In the

presentsection

the differential equationsatisfied

shall construct

we

by the productof two Bessel functions,and solve it in series. We shall


of series.
by direct multiplication
("541) obtain the expansionanew
Given

differential

two

equationsin

the

productvw,

have

we

y"

v"iu

-{I

where
*

See

also

[Jahrbuch
yXJeber

Epstein,Die

vier

iiber die Fortschritte

Gubler, Einleitung

Nielsen, Ann.

in

die

Theorie

Sci. de VEcole

396, Meissel obtained

2v'tu',

respect to

some

z.

Reclinungsoperationen mit BesseVschen


1893"1894,
pp. 845"846].

norm.

who
der
sup.

1904), p. 20.
derfiinktionen(Leipzig,
p.

J)y

Ftniciionen

attribute

the

According

expansion to

Funktionen,

BessrVschen

(3)xviii.

l^mndertmgen

(1901),p. 50
to

Nielsen,

Integratioimceges
include

Math.

Ann.

Graf

Ann.

de

Frogramm,

Math.

(4)ii. (1902),

1862.
that he

uses

iii.

IIQuarterly Journal,

xliii.

and

(Bern, 1900), pp. 85"87, and


der Theorie der CylinHandbuch
ii.

None.

series for productsin the Iserhhn

des

Schonholzer

A trivial defect in SchlaHi's proof is


(1871),pp. 141-142.
only when R(fi + i" + l)"0.
contour
integralwhich (ashe pointsout) converges
93"100.
pp.
" Proc. Camb. Phil. Soc. x. (1900),

(Bern,1894),

der Math.

authorities

The

2v'w' + vtu"

bestimmterlntegralemit Hiilfevon

die Auswertlmng

(Bern,1877),p. 13.

differentiations with

primesindicate

forms

their normal

dz-

az-

if y denotes

then

(1912),pp. 78"100.
10

146

THEORY

OF

^Jy"(^

It follows that

^) 2/1

2^"^' + 2v'rv"
-2Ivw'-2Jv'w

and

y'"+ 2(1

hence

J)

+ 4//
2/'"

(1)

to shew

but, if I ^ J, it-iseasy

This

Appell,Comptes Rendus,
To
a

Orr

form ; both

2/'2/=0;

in connexion

with

(1),see

211"214.

results to Bessel's

applythese

normal

viv').

the equation
satisfies

equation used by Git;

(1880),pp.

xci.

{I- J) (v'w

differentiation that

by

differential

of the

is the form

I =J, y

when

case
Hence, in the special

to

(/' +J')y

[CHAP. V

FUNCTIOjNS

BESSEL

the equationhas
equation,

Nicholson

and

effect the

to be reduced

reduction

by taking

it
z^^^{z) as a new dependent variable,but, for purposes of solution in series,
is simplerto take a new
independentvariable by writing

d^_d^

^-^^ ^^'' ^'^^^ ^^^

that

so

^"

equationsatisfied by J^ {z)Jy (z),when

the

Hence

5?

(^=- .=)"=
.^^ 4e-,}

2 (2^=- .'"
-

/x- ^ v', is

0,

that is to say

[^^-2(fi'+ v')^2

(3)
and

the

(fM' vj] y

+ 4e-"'
(^ +

1)(^

2)3/

0,

equationsatisfied by J^ {z)J"" (z)is


0.

^(^^-4z;2)2/+ 4e2"(^+ l)?/

(4)

Solutions

in series of

(8) are

m=0

where

" /j^"

a=

and

4(a
Cm

If

2m-l)(a

"

{a +

fx +

take

we

+ 2ni
/m +

){cc+

/J,

"

2m) (a

2m)c,"_i
+ 2m) {a
yu,+
z/

"

'
"

fi

"

2m)

and*
1

^'~2''+''r{fi
+ i)r(v
we

obtain

^
,"^oml
and

(-)'"(1^)'^+'^+""
T(fi +
+ m + l)r {/jl
+ 7n
{fi+
V

the other series which

signsof

i)'

the series

either fx

or

or

are

both

solutions of
/x and

v.

v+2m
+

l)

l)r (v +

(3) are

obtained

1)'
by changing the

148

earlier

procedure does not


given by Nielsen, Math.

mode

obvious

This

of

writers;it was

[CHAP. V

FUNCTIONS

BESSEL

OF

THEORY

been

have

to

seem

Ann.

noticed

by

of the

any

(1899),p. 228.

lii.

obtained
by Bessel,Berliner Abh. 1824,
Jq{z)cosz and J(i{z)ii\mwere
and
for
results
J^ (2)sin 2 were
and
the
J^{z)cos,z
corresponding
39,
[1826],pp.
schen
Functionen
iiher
die
BesseV
Studien
Lommel,
deduced
from
Eoisson's integralby
ber
the
functions
and
bei have
deductions
Some
16
18.
concerning
1868),pp.
(Leipzig,
been made
by Whitehead, QuarterlyJournal, XLii. (1911),p. 342.
series for

The

38

"

"

if we multiplythe series for J^(az) and J^ibz),


we
generally,
a'^h"
is
coefficient
of
(-)"'
(|2r)^+''+-""
expansionin which the
More

an

fow!T{v

l).{m -n)\ r{fi +

m-n

tt'-"'
oF^(- m,

'

obtain

l)
m;

//-

1/

1 ;

6-/a^)

+
r(yLt

m!

and

l)r(i/ + l)

so

{hazY (Ibz)"
r(^ + l)

(2) J^(az)J,{bz)
=

2^1 (- m,
^ (-)'"(|a^)^'"
Jlo
mWiji
this result

and

in
expressible
One
another

whenever
simplified

is the

case

case

we

is the

then

Ave

I-

/ox

the

1; "/a')

l)

hypergeometricseries

a,

which

/i- =

has

alreadybeen

v".

the formula*

use

r("-^-fl)r(i)

-1)=

that

see

case

ia,providedthat

Fia ^- a-^4-1and

hiiTT
cos
i-y" (^azy''+^

m-

(4) J_.(az)L(az)=

If

(6)
an

we

take

e^'^"
in

2
=oml

(3) we

Cf,

the

Kummer,

(-)'"(^a^)2''+^'"
m-^
l)r{i' + 2m+

(-)- (i"^r

which

T{v

ber,2(^)+ bei.H^)

expansionof

is

compact form.

of reduction

case

In this

be

can

-fi-m;

cos

1)'

(I.- lm)7r

find that
S

^^

i2i/+4W

~-^,

leadingterms

Journal fur Math.

Avere

xv.

given in " SS.

(1830), p. 78, formula

(53).

discussed ;

5-42]
The

MISCELLANEOUS

(1906),pp.

XV.

from

490

of the
been

series have

Some

types J^^ {z)and J^^ (z)J

By giving/*
e--"

(7)

by Nielsen,Attidella R. Aecad. dei Lincei,(5)


Monatsheftefar Math, und Phys. xix. (1908),pp. 1G4" 170,
differential equation satisfied by J^,{a.z)J^v{bz).

and

497

"

consideration

discovered

(3),(4),(5) were

formulae

"

/._, {zsin ^)

in

by

importance.

is easy to prove

that

^'"''
^^"'^"^y\) (2 sin By-^ S rT^^TZT^
\^v

1-

M=o

in which

of this formula

case
special

to be of any

(2),it

The

(1910),p. 55, for products of the

given,QuarterlyJournal, XLi.
but they are too cumbrous
(/:),

values
special

the

149

THEOREMS

2i/ is

n)

integerhas

an

given

been

Hobson*.

been

outline of

kind.

productsJ^ {2)F" (z),J,n{z)Yn (z),and Y,n(z)F" (z)


is an
the subjectof detailed studyby Nicholsonf ; the following
modifications.
with some
his analysis

have

We

{J^(Z)/. (Z)]
I;

TtJ,(z)Yn {Z)
=

where

second

series for the

The
have

Bessel functions
of the
involving

Products

5'42.

is to be made

equalto

I [J,(^)/-. (^)l,

(-r

performed.

after the differentiations have been

Now

|;{/.(^)J.(^)}=l0g(|^)./.(^)^.(^)
2

r=0

rlT{fi +

x{f(fjL+

2r +

l)r{fM

l)-ylr{/u.+v

l)r{v+ri-l)
1)

+
(z.
i/r

?"

1)}

and

[J,{z)J_. (z)}
1^
=

logiiz).J, {z}/_, (z)


+
(i^)^--+^'T(/^-i^
(_)'"

50

irlr

(fM-

1)

+
+ 2r -\-l)-^lr{fl-v
{^|r{f^-v

parts, S and

divide the last series into two

We

1) r (fi+

.=0

r=o

l)

2r +
r

(-

l)-^jr{-v +
In

the

1)
r

former

l)]

part

we

r=n

have

^^"

Proc.

Lotidon

Math.

1 (-i' +
Soc.

xxv.

l)

(1894),p.

0(3 ;

see

also

Cailler,Mem.

de

l"i Soc.

de

Phijg.de

xxxiv.
(1902"1905), p. 316.
had been
The
expansion of Jo(z)Yo(z)
t Quarterly Journal, xliii. (1912), pp. 78"100.
p. 21.
1904),
(Leipzig,
der Theorie der Cylinderfunktionen
previouslyby Nielsen, Handbuch

Geneve,

giveu

150

OF

THEORY

undetermined

while in the latter part there is no


r

is

replacedin

this part

by

+ r, it is

[CHAP. V

FUNCTIONS

BESSEL

to be

form

evaluated.

When

that

seen

"

--^

(;. n + r
(l^r-"+^'r!r(;a+

2^

\.to
X

r\{n+r)\T{fi

+ ?i
(yu,
-v/r

1)

1)!

l)

2r +

+ r
(yLt
"/r

r+\)

+ "
[21og(i^)+ 2"/r(yLt

1),(n

7r/,(^)J.(^)

(1)

l)
+

1)

-^/rO^+r+l)-^/r(r+l)}.
The
m

0, 1, 2,

by

and

continuous

function

rightin (1).

series for

The

of /a at yu- m where
series for 7rJ'm{z)
Yn{z) is obtained by replacing
fx,

the

so

...,

the

on

rightis

the

expressionon

series
by constructing

be calculated

Yj,i
{z)F" {z)can

for

'd-J"y.{z)J"^{z)'
dfxdv
in

similar

have

been

5*43.

integral
for J^ {z)J^ {z).

The

by

; the

" 5*41

has
integral

this value when

0,l,2,

...

provided

R(fi + v)"-l.

It is then

evident

J^{z)J^{z)
that,when

that

change of
difficulty.
This

p. 263.

formula

the

order

is due

to

T^

(fi-v) 6 dO,

l,

"

J^ (z)J, (z)

the

COS

--"

R{/jb+ v)"

(1)

integral
("2*6) for Jn^(z)is obtainable

formula*

to the result of

so

which is extremelylaborious,
analysis,
will not be repeatedhere.

givenby Nicholson, and

applyingthe

details of the

of Neumann's
generalisation

that

The

manner.

m,

"=

fj.

('^J^+,
{2z
cos

of summation

Cauchy

6) cos (fi v) dde


-

Jo

; for

and

proof by

integration
presents

contour

see
integration,

no

Modern

serious

Analysis,

5-43-5-51]
If

be

MISCELLANEOUS

integerand R(/ji n) "


positive

"

J^ (z)J,,(z)

(2)

151

1, then

"

^-t)!['V^_,
{2zcos d) cos
if /a and

is also true

this formula

and

THEOREMS

are

both

(ix+ n)\ede,

but
integers,

are

otherwise

unrestricted.
Formula

given by Schlafli,Math. Ann. iii. (1871), p. 142, when yi"v are both
cxi.
general formula is due to Gegenbauer, Wiener Sitzungsherichte,
(2a),

(1) was

integers;the
567.

(1902),p.

The

5*5.

A natural

expansiono/QzY'^''as

series

formulae

of the
generalisation

ofproducts.

of Neumann

and Gegenbauer
(""2'7)

("5-2)is that
+ i)r(.
(i,w..-r(,.
(2^r

a^
(i)

i)

r(/x

i)

The

formula

is true

if p, and

proofappliesonly if i2 (^ +

" 5"2

From

multiplyby

we

6
cos'*+''

^^

(p

cos

"

following

^/^+,+,",(22cos^).

it is clear
integrate,

and

(/u-p)Odd=

cos

^"T^

"

v)6

"

Jo

77"

the

1.

"

have

we

(zcosey+''= s
If

1) "

i'

but
negativeintegers,

not

are

"

from

"5"43that

/^

"

"i'

the result follows

and

of fx and

the result

is at

formula

The

Lxxv.

5'51.

the left ; for other


on
by evaluatingthe integral
continuation.
may be established by analytic
deducible

once

from

formulae

given by Gegenbauer, Wiener

values

herichte,
Sitzungs-

(2),(1877),p. 220.

series

LommeVs

of squares of Bessel functions.


Lommel*

expansionderived by

An

t/,,+m(2),
'-/^+"i,(2^)

from

the formula

2vdJ^^{z)

-j"irz~'='^^-'^'^-'^^^'^^^

2(v + 2n)dJU,,,(z)

r,

/,x
J-^_i {z)

IS

~r-

so

a^

71=0

that

2JzJ\_i(z)dz

i (v + 2n)J\^,,,(z)

p. 532

The
;

results of this section


Milnchener

Ahh.

xv.

will be fouud

in Math.

(1886),pp. 548"549.

Ann.

11.

(1870),pp.

632"633

xiv.

(1878),

.152

Hence, by

have

" 5'11 (11),we

..

Iz^ {./^_i
(z) J,-,(z)J, (z)}

(1)

takingzero as the lower limit


beginningof the series,it may be

when

on

If

take in turn

we

have

we

+ 2n) J%+,n{z),
(i/
0

by adding on

R{v)"0\

that the restriction

seen

terms

so

at the

is superfluous.

R{v)"0

add and subtract the results

h;,v %,and

[CHAP. V

FUNCTIONS

BESSEL

OF

THEORY

obtained,

("3*4)
z

(2)

-=

sm2.^

(3)
while, by taking7-

1.

"

(_)n("+i)/.^^^^(^)^
that

see

we

(w + i)/Wi(4

Iz^ [J:(z)+ J,'(z)]

(4)

Another

formula

of the

(2n + 1)J%"^, (z).

type is derived

same

"

'v+n

the
by differentiating

series

("3')
;

M=0

for it is evident

that

"^

00

2^

J-

end

'v+n \2)

e,i Jv+n

Z"

(^)^

i^)

v+n

00

and

so.

Avhen

R(v) "0,

+ J,^,(z)]
J,(z)lJ,_,(z)

2vJj^{z)/z,

obtain

we

J"+n (2)[Ju+n-i
(^) '/^-f-n+i
i^)}

"n

modification

of Hansen's

formula

("2-5),

namely
(5)

enJ-\^,,{z)2v\ J.'{t)~.

2
"

.'0

An

of this formula, namely


importantconsequence
bound
for |J^ {x)\,will be given in " 13 42.

By takingy

i, it is found

en

an

upper

sin^ t
2 r^
+

sni

TT

^.cZ^
2t

so

(6)

I J\+,{z)

where, as usual,the symbol Si denotes


Lommel

in the third of the memoirs

-Si{2z),.
TT

n=^0

by

of

TTJo

2 sin^^

and

value

that

/-n+i {z) -\
"

the

the "sine
to

This
integi'al."

which

reference

result is given

has been

made.

5-6]

MISCELLANEOUS

5"6.

Continued

Expressionsfor quotientsof
immediatelyfrom the

it is at

Bessel

functions
formula

recurrence

are

givenby ^3'2(1);thus,

i^V{^(1^+ 1)1 i^V{(^+ i)(^-f2)1

J^_,iz)

\z-/{(v+rn~l)(v+m)}
This formula

is

easilytransformed

/,_, (z)
results

+ ni)
^z J^+rn^i(z)l{v

i^

"

These

fractions

be written

AM=W^

^^^

continued

as

apparent that

once

(u
^ ^

153

fraction
formulae.

deducible

if the formula

THEOREMS

2vlz-2(v

J"+,"(2)

"

into

l)/z-...-2(v + m)!z

./,+",
{z)

for

generalvalues of i^; (1)


Bessel* for integral
values of v. An equivalent
result,due
then
that, if Q" {z) /,+!{z)l[^z
J" {z)],
true

are

was

to

'

discovered

by
is
Schlomilchf,

Other

(4)

formulae, givenby Lommel:|:,


are

J^+^{z)

Z"'

J,{z)

^^

2{v + \)--l{v

Z"-

2)-2{v\-^)-...-2{v

^2{v+

'

JJz)
Bessel functions

The

Z"

m)-

Zj,+,n+l{2)
J.^m{z)
'

'

l)-2(v

in all these

2)-...-2(,^

formulae

may

J.+.ni^)

m)-

obviouslybe replacedby

any

functions.
cylinder
It

assumed

was

so
neglected,

b}^Bessel that, when

Berliner

writers ;

Graf,

see

Ami.

Abh.

the

00

last

quotientmay

be

Mat.

des

(2) xxiii.

t Zeitschrift
fiirMath,

only.
I Stiidien iiher
und
Math,
Phys.

+ l)] \z^'l[{v^l){v+2)]
\z"-l\v{v

(1824),[1820],p.

Kncydopedie
di

-^

that

JAz) _hzlv
1
J,_,{z)

/m

Sci.

31.

Math.

Formula
11.

(2) seems

28, " 08,

(1895),p. 47.
uiid Plnjs.11. (1857),
p.

142

p. 217.

not
A

to have

_....

been

given by

ditt'erent form
slightly

considered

Schlomilch

the earlier
is used

by

integral values of

pp. 292"297.

die
xxx.

BessePsclien

Fiinctiunen

(1858),p. 332,

and

(Leipzig,1868), p. 5;
Giinther, Archiv

der

see

Math,

also
uiid

Spitzer, Archir
Phys-.

lvi.

dcr

(1874),

154

THEORY

[CHAP. V

FUNCTIONS

BESSEL

OF

though it happens to
assumptionis justifiable,
be so, and a rigorous
proofof the expansionof a quotientof Bessel functions
fraction will be given in " 9'65 with the help of the
into an infinite continued
theoryof Lommel's polynomials."
obvious

It is not

that this

"

[Note.
the fraction
tends

The

tends

pjom

to that

to

limit
be

limit ; this may

as

The

reader

continued
pp. 483
been

will find

fraction

sinw,

memoir*

considered

q,a

discussion

elaborate

an

by Perron,

of the

fractions

types

considered

and

231"247

(1860), pp.

that

case

^'"^"'

though

^"'

"

oi J^{z)/J^_i(s)as
representation
xxxvii.
Sitzttngsberichie,

equation,depending on such
by Wilton, QuarterlyJournal,XLVI. (1915),pji. 320

continued

LVii.

the

on

MUnchenei'

contiguous hypergeometric functions


Math.

a,"=-l.]

m,

of Riccati's

504; solutions

"

between

in

that,even

by taking

seen

jt?"i=m +

the
necessarily

it is not

"i-*-co

is

obviouslycorrect

is not

assumption

the

why

reason

been

The

323.

"

connexion

the relations

noticed

necting
con-

Journal

by Heine,
Lvur.
fiir Math.

Christoffel,Journal

(1907),

representation,have

in this section and

has

filr
(1861),

pp. 90"92.

Hansen's

5*7.
It

was

stated

expression
for J^(z)as
Hansen

by

limit

ofa hypergeometric
function.

j-that

i-4fJ
"^'"=j",i"Tl)-'^-(^'''^^

"

We

shall prove this result for general(complex)values of


throughcomplex values.
fjLtend to infinity
If X=

This is

numbers

For
term

and

the (m+ l)th term


I/tj,
_i

m!

V{v

continuous

(lessthan

\2^)
+

function
2

n
+

1)

of h and

of the

expansionon

when

the

and

rightis

[(l+ rS)(l+ r7,)].

1'

77 ;

and, if 80

the series of which


|~i),

arbitrary
positive

770 are

it is the

{m

l)th term

verges
con-

uniformlywith respect to h and ?; whenever both |S |" So and 1 77 1^ 770.


the term
in question
is numerically
less than the modulus
of the {m + 1 )th
of the (absolutely
convergent)expansionof

the
Since
*

1/8,fi

and

This

uniformityof

the convergence

the convergence
memoir

is the

follows from

is uniform, the

subject of

paper

of the

sum

by Nielsen,

the test of Weierstrass.

Milnchener

terms

is

continuous

Sitzungsberichte,xxxviii.

(1908),pp. 85"88.
t Leipziger Abh. 11. (1855), p. 252 ; see also a Halberstadt
[Jahrbuch ilber die Fortschritte dcr Math. 1909, p. 575.]

dissertation

by

F.

Neumann,

1909.

156
We

take

Murphy's formula
P" (coszjn) o/^i
(=

and

modification

that

then

\z\ ^ 2\n\, and

when

,F, (where

is

we

can

the two

compare

1 ; 1 ; sin^'^z/n),

applicablewith

the

slight

series

,F, (1/So,
1|^ +

^B^

1 ; 1 ;

|^

^),

number
less than f j^ |~^
and the comparison
positive
arbitrary
\n\" l/8o-The details of the proofmay now be left to the reader.

Bqis an

is made

n,

1 ; 1 ; sin- ^z/n);

precedingsection
the inequality

use

we

n,

the

reasoningof

the

[CHAP. V

FUNCTIONS

BESSEL

OF

THEORY

when

the series for P" (cosz/n)


positive
integer,
to appealto Tannery'stheorem
to complete
terminates,and
firstnoticed by Giuliani ; the earlier writers took for
the proof This fact was
of the passage to the limit.
grantedthe permissibility
When

is restricted to be

it is convenient

In the

generalised
Legendre functions (of unrestricted order '7"i),
depends on whether the argument of the functions is between
of

case

the definition
+

1 and

"

P"so

not

or

for real values of

(between

.F, ((cos
^) ^^^^^^
=

l;m

.,

0 and

have

ir)we
l;

|.r/.),

sm^

that

lim"-P,r"'(cos-)=j;"(^),

(2)
but otherwise,we

have

Pn'"^ (cosh ~]

p
so

j^^oPi(^^^\^

n,

1;

-\-1 ;

smh^

|-^/"),

that

(3)

lim n^-P,r-^

(cosh
-)

/,,{z).

The
from

formula for functions of the


corresponding
the equationwhich expressesf
Q""*in terms
?i~'"

sin

lim

(4)
This formula

n IT

second

Qn"'[cosh

kind may be deduced


of P""* and P"~"' ; it is
=

Km (2).

has been

it is most
given(with a different notation)
by Heine ;]:;
the integral
of Laplace's
easily
provedby substituting
type for the Legendre
function,proceedingto the limit and using formula (5) of "6-22.
*

Cf.

Bromwich,

+ Cf.

Theory of InfiniteSeries," 49.


Barnes, QuarterlyJournal, xxxxx.
(1908),p. 109; the equation is
sin
_"

2r

??^

sin

mir

n)

P~"^

nir

g/"

in Barnes'

notation, which

t Journal

fiirMath.

lxix.

is

in this work.

adopted
(1868),p.

131.

"

'"

^"

T-

r(l-m

n)

T{l

n)

5-72]

MISCELLANEOUS

different from those justdiscussed,is


formula, slightly

Another

(5)

lim P,,

"

this is due

Laurent*, and it may

to

Pn (cose)
The

[Note.
to be

g^)

/o (2.);

provedby usingthe

be

second of

Murphy's

namely

formulae,

used

157.

THEOREMS

generallybelieved
Thus

Bessel functions.

Functions^ Lamfs

it

that there
stated

was

Functions

Bessel functions]
are
[i.e.

and

with

in

the main

luith Mehlers

^^).

be

emphasized because it
Legendre functions and
his Eleuientan/Treatise on Laplace's
(London, 1875), p. vi, that "these
subjectof this book."]

Functions

BesseVs

associated
Integrals

5"72.

no

tan-

must

connexion

Todhunter

connected

not

of this section

was

by

1 ;

n, -n;

formulae

of the

existence

^e 2F1 (-

cos"

between

formula.

the formulae
of the last
completelydifferent method of establishing
and
this
Mehler
method
section was
also,later,by Rayleigh;
givenby
depends
of
thus
a
on
use
integral,
Laplace's
A

":

P,j (cos6)

\ (cos 6

i sin 6

cZd)
"/))"^

cos

1 T"
g" log (cos0+i

__

sine

"/))
^^Jj

cos

Since

uniformlyas

when

go

-^

i sin

log{cos{zjn)+

0 ^

{zjn)cos 0}
have

tt, we

at

iz

-^

cos

once

1 f"

lim
Heine

of

f and de
Laplace'stype

Bessel functions
results in

" 6'22
has also

Mehler

P" (cos2/n)

Ball.I have
for

made

Legendre

when

/'^

2
"^

If

be shewn

it may
y\r^

n"p

by using the

the

limit

integral

cos

(9)}

])resentssome

^|rd^

/"^cos
"

-.u,

--r:y,
'^

to

Mehler- Dirichlet

that
2

passage

his

{n + ^) (f)d(f)

cos

J 0 V{2 (^os(p

P"(coss/")^bvit the

to

of Y^,{z).
representations
integral

his formula

given a proof of

third kinds ; reference will be made

deal with

we

d(f) J^ (z).

similar passages to the limit with integrals


In this way
Heine
functions.
has defined

and

of the second

e'' '"''"^

\'V

because
little difficulty

the

integralis an

proper
im-

integral.
Various
*

Juuriml

enoneous

de
on

Journal

have

formulae

which
givenrecently
384"385;

(3) i. (1875),pp.

Math.

account

been

of

Math.
J'iir

arithmetical

an

lxix.

Nacli.

cxxviii.

(1891),col.

formula

actuallygiven by

Laurent

is

error.

(1808),p. 131.

pp. 383"386.

t Ai"tr.

the

in which

exhibit the way

1"4.

See

also

Journal,
Sharpe, (juarterli/

xxiv.

(1890),

158
the

Legendre

Thus,

formal

approachesits limit
expansiondue to Macdonald*

function

its

as

degree tends

to

infinity.

is

(n + 1)-" (cosi^)-""[Jm (^')

where

[CHAP. V

FUNCTIONS

BESSEL

P^-'-lcos^)

(1)

a;

(2?i+ 1) sin ^6.

upper limit for the error due to


Legendre function of largedegreeby a Bessel function,aref
exhibit

formulae,which

Other

OF

THEORY

(2) Pn (cos7])"

an

replacing

Qn (cOS7])

ITT-l

|(^+ 1)tan 77}" iY^ {(n+ 1) tan 77}]


V(sec7;)e" ("+*)""("-tan')) [/,,

4"9i
\/(sec11)

Pn (cosh^j

(3)

(4) Qn (coshf)

(^)

^-^-^

/o (nf) +

j)
e-("+4)(f-tanh
^(gech I) 7^0{(w+

i) tanh ^|

"6"3V(sech|).e-"^^+^)^
"^

i? (n) +

where, in (2),0
less than

^ir,and, in (3) and (4),f ^

^ ?; "

unityin absolute

its real part is

magnitude,and

0 ; the numbers

be

complex
may
the proof of these results is too

positive.But

i,
6^,62,6s are

providedthat
lengthyto be

givenhere.
Th

The

expressible
by Hansen's formula as a
to investigate
methods
hypergeometricfunction has led Olbricht;):
Bessel's equationis expressible
confluent form of equations
a
as

"

fact that

limit of

which

by

formulae of OlbricJd.

5 73.

associated with
If

Riemann's

take the

we

Bessel

function

is

P-functions.

equation

S-^|-(-'^"
of which

and

fundamental

the

compare

system of solutions is the pairof functions

equationwith

the
'a,

pU
Proc.

obtained

London

Math.

previouslyby

+ Watson,

Tram.

the

Soc.
same

Camh.

equationdefined by
b,

1,

/3', 7',

(2) xiii. (1U14),pp. 220"221

writer,Proc.
Phil.

Soc.

xxii.

London

Math.

(1918),pp.

pp. 151"160.

X Nova

Acta

the scheme

c,

^,

A,
*

o.

Caes.-Leop.-Acad.
(Halle),1888,

pp. 1"48.

associated

; some

Soc.
277"

xxxi.

308 ;

results

(1899),p. 269.
Messenger, xlvii.

had

been

(1918),

5-73]

MISCELLANEOUS

THEOREMS

159

namely
fl-"-"-

d^y
^

d^;^

-^-^'

1
^

f"a^(g -h){a_^

"

] dz

"

{h -c){by3/3^

c)

1-7-y]

"

7y(c-a)(c-6)

a)

"

_^

_^
2'

"

"

=0,

(2: a) (s' b) {z
-

"

that

see

we

latter reduces

the

while

0,

h, c, /3,/3',7, 7' tend

remain

finite

We

thus

(their sum
obtain

a=

the

7'

"

J/

0,

2^/3,

fi,

^,

/i,

Pl

scheme

for

lim

those

which

Note.

have

It has

(1886),p. 31, that

whose

when
zero.

solution

the

the

^2

{6- and

h +

7'

0.

2i/3,
z

'y,
7

"p,

7,

Hansen's

formula

-4a/3,

but

constructed

is

0,

a-hv,

schemes

observed

7,

00,

^^".

v+l-a-^.

they

are

will be

by Haentzschel,

of

no

great importance and

sufficient

Zeitschrift
fur

examples.
Math,

unci

Phys.

xxxi.

equation

(/m), may
{% 4-3) is iih^"^v

invariants

directlyfrom

ii/,

been

now

been

^,

[_ij,^
l^-\v,

other

given

/3/3' 77'

7 4-

before.

0,

pJ

(;^::)
has

7' as

J^ {z)derived
r

Olbricht

/3 + /3'and

that

way

/^,

IX,

V-

of 7 and

values

same

fx,

"

is

scheme

\-v-ii,
the

such

while

"

scheme

lim

with

+ ^Q-}.
^[{ix+ i-)-

similar

Another

o!

IX,

"

infinityin

to

I7,

if

former

being 2/a+ 1)

lim

where

the

to

c)

"

and

g-^ of

the

be

derived

Weierstrassiau

by

confluence

Lame's

from

eUiptic function

are

made

equation

to

tend

to

VI

CHAPTER

REPRESENTATIONS

INTEGRAL

Generalisations

6'1.

this

In

Poisson's

of

Bessel's

functions

in

applied

Chapters

and

vii

It

that

happens

the

shall

we

integralwhich

formulae

singleexponentialunder
examine

shall therefore

integralswill
formulae

and

integral are

Bessel's

of

more

and accordingly
integral,

type, deferringthe study of integrals

is due

of

choices

large.

Poisson's

in substance

the circumstances

Hankel

to

" 3*3 (4),since

is

" 3"3

expressiblein

are

can

contour

approximate

integralsign,while

the

functions, Avhich

circular

is

with

be obtained

investigationof generalisationsof

give

now

the

or

Poisson's

The

6"2.

shall

we

simplestof

The

"

to

type

applicationsof

study integralsof

now

Bessel's

of

the

obtain

to

suitable

elegantformulae

integrals.The

applications of

than

elementary character

of

viii

asymptotic expansions for /^ (z) when

integralsassociated

integral("2'2). By

definite

as

FUNCTIONS

contour

integration,
largenumbers

Bessel

BESSEL

integral.

various

study

integral(""2'3,3"3)and

express

also be

Foisson's

shall

chapter we

of the contour
which

of

OF

the

*.

this formula

other

of

terms

in which

Poisson's

contains

formulae

contain

exponentials.

two

We

integralsof the type

contour

2^1e'^'Tdt
J

are

solutions

of Bessel's

not

of z, and

that

of

the

equation;

it is
and

end-points,a

h, are

complex

is

function

numbers

of t but

independent

z.

operating on

" 3"1, is

in

V", defined

TdA

'

of

result

The

e'-'

integralwith

gu'fj(i

^2~)^n + (^2v+

fb
J

Math.

and

Ann.

AV(2)

expressed

detail

i.

is due
in

(1869), pp.

473

Schladi,

Ann.

to

the

notation

by Gubler, Zurich

the

theory

pp.

235

"

Bruxelles,

of the

262
xxis.

linear

lvi.

The

485.

"

di

explained

operator

Mat.
in

discussion

4-15.

(1903), pp.
140

432
"

equations
"

143.

444.

See

Te'-'^
Ttdt

i2vVl)Tt-^^{T(t^-l
dt,
of

(2) i. (1868), pp.


"

1) ("+'

xxxni.
Vierteljahrsschrijt,

differential

(1905), pp.

Bessel's differential

follows:

as

^"+2

the

e^~^'T{t'-l)

t"''+i

are

that

supposed

which
also

The

the

for Iv(z}
corresponding integi-als

though

232"242,

integrals have

(1888),pp.

147

also

"172,

they satisfy,by Graf,


de

la

Vallee

Poussin,

been

and,
Math.
Ann.

Schlafli's

results

examined
from

the

Ann.

de

in great

la

aspect of

xlv.

(1894),

Soc.

Sci. de

6-1]

INTEGRAL

we
by a partial
integration.
Accordingly

if T, a, h

are

so

chosen

161

REPRESENTATIONS

obtain

solution of Bessel's

l)Tt,
eiztT(t^-l)
^^{T{t^-l)]^{2v
+

The

former

of these

that

so

it is

that

we

0.

multipleof
integration,

constant

the

path of
^

1)"+
e''^*(^2

that

closed circuit such

is

choose

may

initial value after t has described the circuit,


or
at

that

equationsshews

(t^" I)""*,and the latter shews


either

equation

that

returns

"

that e'^ {t- 1)"+

so

to

its

vanishes

each limit.
A

of the first type is

contour

passinground
figure-of-eight
clockwise.

the

point

if

and round
t
And,
we
suppose
of
is
the
second
t
hat
contour
real
of
the
a
z
positive,
temporarily
type is
part
both the points
which starts from -|-ooi and returns
there after encircling
one
t

counter-clockwise

-1,-1-1

counter-clockwise

(Fig.1

and

"l

If

Fig.2).

we

take

a,

"

1, it is

Fig.1.

Fig. 2.

necessary

to

suppose

that

R{v

h)"0, and

we

merely obtain

Poisson's

integral.
To make
of ^

1 and

"

axis

on

We

the

the many- valued function (i"


^ + 1 to vanish at the point A
"

rightof

therefore

^=

we
1)""*definite*,

where

the contours

take the
cross

phases

the real

1,

proceedto

examine

the contour

integrals

"(i+, -1-)
+

of the values i ^| ^,
It is supposed that v lias not one
"1, and both integralsvanish, by Cauchy's theorem.
,

at

w.

B. F.

tt

CCI

...

; for then

the

integrandsare analytic

11

162

THEORY

It is

differentiations

analyticfunctions

are
integrals

In

order

expand

the

under

to

in
integi-and

the

convergent with

respect

/"(!+,-1-)
J

Now

odd;

or

we

2
m

and so,

1)""^is an
takingthe

or

even

left with

the

To

from
u

"

"

1) "
0

0 ;
1

to

on

the

evaluate

the

the

is

even

origin,
we

are

/"(!+)

rightoi

contour

taken

Jo

(2m):

/
;
{'^lll)

u^"^-Hn-iy-idu,

Jq

right,we

then
the

on

the

1 vanish

"

when

be

temporarilythat
the straightline

assume

deformed

into

first part, going from

second

part, returningfrom

in each case, the

"

and

\.

the

may

twice;

l)"-*dt

f^'" (f^,

/"(1+)
(_\7nf.v+2)n
^

integralson

{\"u)e~''',and on
{\ u) 6+"',where,

We

accordingas

rightvanish, and

the phasesof
\/u; in the last integral

the real axis

i? (i/
+

i^

s
7/1

on

a"

is

the

on

(_\m^v+2m

00

of t

1)""^dt

symmetricalwith respect to the

t'^ (P

JA

equation
-

writingt

r-

of the series

efe*(^f- 1)"-*dt

JA

r{,\+,-l-)

odd function

an

-1-)
/"(1+,
2"

on

ml

to be

contour

that the alternate terms

see

then

dt=

f^it^"

It follows that

,;in2:v^m

oo

^^t (f-ly-i

z-

v.

integralin terms of Bessel functions,we


series being uniformly
of z, the resulting

the contour.

on

of

for all values

powers

to t

convergent,
permissible.Also, both

integralsign are
of

first

the

express

are
{")" 0, both integrals

that, when

phase oil

is

"u

to

we

have

0, we

have

1,

1 to
zero.

thus get
f(l+)
ti""-h(xi

r\

1)^-*du

tcn-h (1
{e-(""-*)'r^' e(''-*)T'}

2t

Now

both

sides of the

cos

VTT

du
u)"-^-

^^--

and

be observed

to

[CHAP. VI

FUNCTIONS

BESSEL

OF

"

"

"

equation

f"^'""-"
("- 1)-'

rf. =

2i

cos

analyticfunctions of v for all values


this result,which
continuation*,
analytic
for all values of
persists
are

I"^^"ili""i)

of v; and

by the generaltheoryof
proved when R(v + ^)"0,

so,

has been

v.

-R

Modern
Analysis," 5-5. The reader will
("+ i)" 0, by repeatedly using the recurrence

P+^.-4

(u

1)"+-*

du

is obtained

find

it

possible

+J'"-"_1
"

nA-h

f"^\,"-i("

J0

u-1

the

result, when

i)v+"H du,

the formula
by integrating

(m
|^{u'"+^(u-l)''+"+ill)u"'-^(u-l)''+" (,
the

obtain

to

formula

Jo
which

also

integral is then expressed in


has a positivereal part.

terms

of

au

integralof the

same

"

!) M"'-i(u_!)"' "-*;

type in which

the

exponent of

164

THEORY

Thus, when

and

R(z)"0
r"'i

-^is not

positive
integer,

/"(-!+.1+)

7/V'"rtVi "y

"

(2)

BESSEL

2/

[chap. VI

FUNCTIONS

OF

27rtr(|)

This

also obtained

equationwas

Next

by

Hankel.

consider
f(-i+,i+)
,

where
of

co

00

esp(-!to)

defines
or negative.This integral
angle,positive
is analytic
when

is an

which

acute

iTT +

"

and, if z is

subject

into the second

analyticcontinuation

of

ZttI

has any

jR

(2)"

0 and

(v+ ^)"0

we

can

the
an

have

we

value between

0);

e^^Hf- ly-Ut,
"^tt

By giving a suitable value*, we


any assignedvalue of arg z between
When

function

J cciexp(-iuy)

{^)

for

|arg z j" ^tt,the contour


justconsidered. Hence
over
by the new integral

that

defined

that

so

contours

be

J-t,(z)can

J-.(^)='^%
yrnV
arg

K^TT

arg

of the two

of values of arg z;

range

(3)
where

ft) "

the further condition

to

be deformed

extended

^ir+

obtain

can

tt

"

take

may

and

o)

and

the

(o.

of J_^(z)
representation

a
tt.

contour

to

be

that

shewn

in

Fig.3,

/\

Fig.3.

supposed that
By takingeach straightline
in which

it is

the radii of the circles


in the contour

made
ultimately

are

we
separately,

small.
indefinitely

get

^"^^^s^T/:" '-'"'-")-*

J-.{z)-

+
*

If

iw Ibe
of J

increased
-I,

(z)valid

in

series of

for any

stages to an
preassigned value

e^"e'^'(''-i)(l-"2)i/-i^^1
appropriate value
of arg

may

(greaterthan

be obtained.

a
Jtt),

sentation
repre-

6-11]
On
"

bisectingthe

third

(4) J-A^)

which

Weber

t in the

the formula

to

integrals
by it,
Gubler*, which corresponds

is

l^\-^^{l
+ t^^r-hdt+rcoB{zt
^J;jf[smvTT
vn).{l-t^yL
Jo
J

{" + 2)

and, if this be combined

formula

the various

^,
i

n(^)

and replacingt in
integration
for J^v{z),due
a formula

integralfor ./^(s)
;

(5)

of

obtain

we
t, "t, t, it respectively,

to Poisson's

path

165

REPRESENTATIONS

INTEGRAL

""

K^J

with

Poisson's

2{^zr
=

"r(v+*)r(i)
was

it is found
integral,

that

I e-'i{\-irt'^Y-hdt\
j\m{zt).{\-t^-)''-hdt-

also discovered
values
of integral

case

by Gubler,though
of

it had

been

stated by
previously

v.

has gone before the reader


After
closelyconnected with (1),namely
what

should

have

in obtaining
a
difficulty

no

formula

(6)
in which

rightof

it is
i

supposed

the

that

phase

of t'^"\

vanishes

when

is

Modifications
of HankeVs

6 '11.

of

contours

" 6"1

By making
Zurich

und

the

into the contours

which

the contours

(1888), p. 159.

See

also

are

to
parallel

the real

Graf, Zeitschriftfiir Math,

and

(1893),p. 115.

xxxvm.

Phys.

portionsof

those

xxxiii.
Vierteljahrsschrift,

t Journal
pp. 86"00.

on

Fik'. 5.

Fiff.4.

Phys.

the real axis

integrals.

contour

Taking R {z)" 0, let us modifythe two


shewn in Figs.4 and 5 respectivefy.

on

l.

filrMath,
The

iii.

lxxvi.

formula

(1873),p. 9.

was

(1892),pp. 142,

examined
234.

Cf.

Hayashi, Nyt

in the

case

i-^O

Tidssk

riftfor

Math,

xxiii.

is, (1012),

by Escherich, MonatsheftcfiirMath,

166

THEORY

axis

off to

move

obtain

BESSEL

OF

(so that
infinity

[chap. VI

FUNCTIONS

integrals
along them

the

tend

to

zero),we

followingformulae:

the two

(-1-)

(1+)
X

(2) J_.U)

-1+coi

r(^-v) e"- (i0)


27rtT(|)

r(i+)
(1+)
,

/"(-!+)

X
I

./"1

l+ooi

In the firstresult the

many-valued functions

phase of
phase of f-

be 0 at J. and

t^"1

the

1 is 0 at JL cmd

"

To
in the
^-

avoid confusion
same

1 is +

"

to

(3) J.A^)

to have

formulae; and when


(1) is of course

i?,the formula

at

TT

the

phase of ^- 1 interpreted
it is supposedthat the phase of
unaltered,while (2)is replaced
by

27rt r

(-1-)

e^^{t-- \y-^dt-\-e-vm\
l+

alteration

e^'^{t--iy-^dt
-1

OOl

in

the

00

direction of the contour

the
integrals,

In the last of these

convention

has been

determiningthe phase

reversed

of i^

"

has

the insertion of the factor g-slv-jj^ri

necessitated
On

"

(1)

(1+)

the

the

and

interpreted
by taking

B, while in the second

at B.

it

"

it is desirable

in both

way

is

at

tt

00}

to be

are

be +

to

comparingequations(1) and (3)with "

3"61

equations(1)and (2),we

that

see

H.^H^)

(4)

^^\-"r}a'fl''^'
e^^^{t^-iy-idt,

(5)
*""* ^

unless
We

is

in
integer,

an

can,

however, obtain

value (w),from
continuous

which

consideration

functions

of

ir""i'
{z)
=

near

and

-1

oot

equations(1)and (3)are

case

(4)and

(5) in

the

case

when

not
v

Thus

n.

-JTl

(t)

r(|-n).(i^)^
iriVil)

for ^"*^'{z).
similarly

has

of the fact that all the functions

lim fr;i'(z)

y^n

\2.'

1+ooi

fi^^)

e'^^if-iy-idt,
1+ooi

independent.
an

integral

involved

are

6-12]
As

INTEGRAL

REPRESENTATIONS

167

of " 6'1, the


of (4) and
corresponding
analysis
ranges of validity
extended
round
the
and
contours
the
by swinging
using
theoryof

in the
be

(5)may

continuation.
analytic
if

Thus,

(6)

^77"

"

"y

f TT

"

6)

'^1' i-

both

providedthat, in
^TT4- ").

"

27r and
If

and

^"' (^'

1)^

cotexp(

"

obtained

thus

27r,and

^^^'

"-

!w)

"

lies between

of ^^'^'(5)when

of H^^-^(z) when

arg

has

arg

^tt+

"

any

and

"

has any value


value between

tt.

increased

be

o)

round

tt

"

\2)

(6) and (7),the phase of

are
Representations

between

1)"--dt;

^^-vwrf^I

^;=) (z)

(7)

^"* (i'

have

Itt,we

"

have

we

'^\.;ViV

^.'^'(z)

while, if

f TT,

"

the

beyond

singular points of

the
the

limits stated, it is necessary


integrand,and numerical

make

to
errors

the

are

coil

contours

liable to

occur

of the integralsunless great care


is taken.
Weber, however,
interpretation
the
to determine
(1890), pp. 411"412,
adopted this procedure.Math. Ann. xxxvii.
with H,m (s),HyC^) (z).
of " 3-62 connectingH^C^)( z\ Hj:^)( .-)
in the

Note.

The

formula

2ii\{z)

makes
Hy('^){z)-HJ~'){z)

possibleto

it

Hankel
obtained
and
in this manner
loop integrals,
be
of
view
w
ill
in
this
not
reproduced
investigation
Yn(3);
Hankel's other method
which has been described in " 3-52.

offunctionsof the
Integralrepresentations

6"12.

In the formula
between

tt

"

and

27r,and

that

that the

phase of

contour; and
(1)

(o

has any

given value

^,

shall write

we

e-^^i z-^ (

t-l=
so

3^^(2)in

" 3-52 for


greater simplicityof

Itt" /3" Itt.

"

Then

phase of
equation

define /3 by the

the

express

series of

third kind.

that the

" 6-11 (6) suppose

arg
so

the

of

terms

has
mulae
for-

it follows

H^^ (z)

m).

increases

from

tt

/3 to

tt

/3 as

t describes

the

immediatelythat

-^^

7,f,

(- uy-^

e-

-.

1 +

du,

value. Again,if ^ be a given


phase of 1 + ^iulzhas its principal
of
affords a representation
this formula
acute
or
negative),
angle (positive
in which
the sector of the 2r-plane
i/'^"i"
{z)valid over
where

the

_-

+ /3"
-Itt

arg

zk^it

13.

168

[CHAP. VI

FUNCTIONS

BESSEL

OF

THEORY

from " 611 (7),


Similarly*,

/3 is any

where

or
negative)and
angle(positive
|7r+ /3 " arg ^ " Itt+ ^.

acute

t,by " 3-61 (7),H^^^'^{2) e""' iT^D (^),it follows that we lose nothing
to deform
that R(v + ^)"0; and it is then permissible
so
v
by restricting
taken twice; for the
into the line joiningthe originto 00 expt/3,
the contours
tend to zero
at the origin)
taken round a small circle (with centre
integrals
Since

the radius of the circle|.

with

deformingthe

On

where

/3may

the

specified
manner,

{^^l.^,;^
I

HJ'" (.)

(3)

(1) in

of

contour

be any

l)"0,

from

In like manner,

find that

(l 2i)^-.

.-" "-*

and
or
negative)
angle(positive

acute

R{v

we

Itt+ /3 "

arg

"

/3.

tt

(2),

rxexp'jS
2 \^ g"Hz"ivTr"iTT)

iu\''~^

*''
i."w=Q VTTTirl.-"''-K^-s)

w
where

^ may

be any

R(v
results (3) and

and
or negative)
angle(positive

acute

l)"0,

|7r+ yS.

/3"arg^"

yet been

proved when

odd positive
integer.
(where ?j 0, 1, 2, ...)
similar work of " 6"11,that (3) and (4)are true when
it follows,as in the somewhat
^,
also
be obtained for such values of
results
The
by expanding the integrands
", #,
may
in terminatingseries of descendingpowers of z, and integrating
term-by-term;the formulae
are
so obtained
easilyreconciled with the equationsof " 3' 4.
The

in view

But

of the

(4) have

-f7r

not

continuitynear

of the functions

2u

is

an

involved

....

importancein the
generalformulae (3) and (4) are of fundamental
of J"" (z)for largev^alues of \z\. These
discussion of asymptoticexpansions
of the formulae will be dealt with in Chapter Vll.
applications
The

arg

u=2z

l3 (so that arg z is restricted


cot 6, it follows that

To

There

obtain

formula,write
e~^" 2-^ ( h),
f-h l
=

is

an

even

to be

function

Cf. Modem

an

If we
Schafheitlin".

acute

and
angle),

"

this

seems

be

to

to

of

no

then

r^-cos''-*(9.e-"'"^-''^+^'^

2"+'^"

rT(.^,^

,^.

is due

of the formulae

useful modification

,^

,^

2e"'"
(1 -^iufz).

simple direct proof that

v.

Analysis,"

12-22.

"

Journal

fur Matlu

cxn-.

(1894),pp. 31"44.

take
write

6-13]
and

INTEGRAL

REPRESENTATIONS

that

hence

008"-'^$. sin
li''

2^1 ^^

(7)

169

(z-vO

hO)

These

formulae, which

of

valid

_.^^^^^g
^"

sin^-'+i^
^''(^)-r(i;+i)r(i)Jo

,.

rf^,

onlywhen

R(v + ^)"0, were


of the zeros
of Bossd
functions
appliedby Schafheitlin to obtain properties
obtained by him from the consideration that the
(""15'32
15"35).They were
the rightare solutions of Bessel's equation
which behave in the
on
expressions
the
near
manner
appropriate
origin.
are

course

"

integral I

The

u^'^^{l + ii)'^~^
du, which

e~"^

is reducible

integralsof the types

to

Jo

occurringin (3)and (4)when


(1904),pp.

Lix.

The

ix

v,

has been

studied

this section

of differential

are

also discussed

6*13.

detail

some

The

from

by Nielsen,Math.

the aspect of the

equationsby Brajtzew, Wm'sckau

1, 2 [Jahrhuchiiher die Fortschritte der Math.

nos.

in

Ann.

89"102.

of
integrals

solutions

Mehler-Sonine
generalised

1903, pp. 575

theory of

Polyt.Inst.
577].

"

Nach.

totic
asymp-

1902,

integrals.

Some

definite integrals
elegant
maybe obtained to representBessel functions
of a positive
variable of a suitably
restricted order. To construct
them, observe
is less than |, it is perthat, when z is positive
(= x) and the real part of
missible
to take ")
Itt in " G'll (6) and to take "
^tt in " 6'11 (7),so
that the contours
those shewn
in Fig.6. When, in addition,the real part
are
of V is greater than
the contours
to deform
(afterthe
|, it is permissible
of " 6"12) so that the first contour
consists of the real axis from + 1
manner
v

"

"

to
"

1 to

twice while

taken

00

"

taken

00

second

the

contour

consists of the real axis from

twice.

-^.^

r,^^

N.
/

Fig. 6.

We

thus obtain the formulae


^
Ri^)

(,.;)
=

I/f^(-3)
(^,)
=

(i^^lM

[V*{t'

e-.(-*t--)

^"^Skl^^-^
(1

the second

(1

iriv

being derived

from

(o)

dt,
1)^-^

it? 1)"-^dt,
e=(-5)-)^e-'-'
-

.'1

t by
" Gil (7)by replacing

1.

170

THEORY

replacez; by

In these formulae

1^

and

the transformation

use

formulae
then

['* e^^*dt

iT.W (a;)

(1)

-l"R{v)"^,

and

x"0

follows that, when

givenby " 3-61 (7). It

[CHAP. VI

FUNCTIONS

BESSEL

OF

=iY{l-v)T(i).{lxy] {t'-iy-^i
'

so

e-^'^^dt

[*

that
dt
f'"sin (a:;^)

(3)

/. (^)

(4)

Y^ {x)

r(i-^)r(i).(i^)v'i
(f^-i^i'
2

Of

these

results,(3) was

Other

specialcase.

Symbolicformulae

"When

R{z)"0

where

the

denotes

If D

and

of 1

phase

Ann.

"

due

(i/
+^)

is any

Ann.

v.

(1872),p. 142, in the special

0 and

(4) in the same


given in " 6-21,

Maodonald.

and

0, it is evident

i^ lies between

and /
{djdz)

Mat/t.

^^.^ ^^^^^

(1880),p. 39, gave both (3) aud


the Mehler-Sonine
integralswill be

Hargreave

to

"

^^^^^

xvi.

of
generalisations

6*14.

given by Mehler,

v-0, while Sonine, Math.

case

^^ J, ^^^

(xt) di

r"" cos

from

formula

" 6-11 (6)that

\tt.

"

polynomial,then

f(it).e^^^=f{D)e^\
and

so, token

is

v+\

have
we
positive
integer.,

-r("+i)r
When

is not

of
representation

("''+""'

positiveinteger,the last expressionmay be regarded as a symbolic


^^O {z\ on the understanding that/(Z))(e"'^/2)
is to be interpretedas
a

',

0 dt.
e^^'/(^"

Consequently
(1)
and

SO

similarly

that

B^K^-s(z)=

ly

(i+i)y-j"i!iizi^2^,

THEORT?"

172
in which

phasesof

the

both

^^ are

" 8-71 (9),

Now, from

zero.

have

we

and

so

^^

7_, (z) /. (z)

(3)

that is to

"

'

obtain the formula

we

^^i^i^^
fe"
Jo
(i/

K, (z)

(5)
^

result set

formulae

Hobson

by

as

all valid when

are

Basset's

6*16.
When

sinh^''^d^,
The
will

for K^, {xz).


integral

and
positive

is

'^"

t)

Tripos,1898.
problem in the Mathematical
R{v + l)"0 and [arg^ |" Itt. The reader
from " G'll (6).
(4) directly

find it instructive to obtain

^
1

(*"-!)'-**,
/C(^)='^^|4^/V'

whence

Iarg

^^iMiy

say*

(4)

of 1

and

t--l

[CHAP. VI

FUNCTIONS

BESSEL

OF

is

complex

number

derived
for H^^^
{xze^-"^)
the integral
I" l-TT,

subjectto

from

condition

the

"611 (6)may

be written

in the form

of a circle from p to
taken round arcs
integral,
Hence, by Cauchy's
oo
tends to zero
as p
by Jordan's lemma.
p gijTTi-iargz^
the line
until it becomes
theorem, the path of integration
may be openedout
iu, the phase of
which R (zt) 0. If then we write zt
{u-jz^)1 is tt
on
in the ?t-plane.
at the origin

Now, when

the

R(v)^-l,

It then

follows from
=

(.r^e^-O
iT^l^

J..V.(^^-i)"^-^

2r(i)

e-^^'^du

V{v+\). {Izy p
and

so

we

have

Basset's formula
,^

valid when
Basset
*

The

,_^_V{v

\).{"zy r""

co^xu.du

\)^0, x "0,\ arg z\"\7r. The formula was


integralvalues of
only,by regardingKq{x) as

R{v

f, for

"

"

" 3-7 (8) that


iTTie-i"-*

K, {xz)

"

integralon the rightwas examined


(1855),pp. 130"139.

in the

case

by Kiemann,

Ann.

obtained
the
der

Physik

Chemie, (3)xcv.

t Proc. Camh.

Phil.

Soc.

vi.

(1889),p.

Hydrodynamics,

ii.

by

limit of

(Cambridge, 1888),p.

19.

und

6-16, 6-17]

REPRESENTATIONS

INTEGRAL

173

it by the corresponding
Legendre function of the second kind and expressing
of
limit of the integral Laplace's
type (Modern Analysis,
" 15"33).The formula

for

is obtainable
Kn ("^^)
Basset

of
by repeatedapplications

similar
a
investigated

also

formula

for

the operator

"

j-

I^,(xz),but there is an

error

in his result.
mathematicians
before Basset.
integralon the rightin (1)was studied by numerous
Poisson (see " 6"32),Journal
these investigators
de I'Fcole Polytechnique,
were
ix.

The

Among

Catalan, Journal

(1813), pp. 239"241;

de Math.

(1840),pp.

v.

la Soc. R. des Sci. de

de

corrections,Mem.

some

de Math.

110"114

(reprintedwith

Liege,(2) xii. (1885),pp.

(1843),pp. 20, 21; ix. (1844),pp. 193"210;


97.
These writers evaluated
u.
1848),pp. 96
Studien,
(Leipzig,
Analytischen
is
when
ianite
a
terms
in
positiveinteger.
j/+|
Serret, Journal

viii.

"

Other

who

writers

(1873), pp.
; J. J.

815

360

365

"

Phil. Soc,

Camh.

Ellis,Trans.

Leslie

be mentioned

must

QuarterlyJournal,

cos

.Ku

(1882),pp.
Oltramare, Comptes

and

provedby

writer

last named

213"215

Glaisher, Phil. Trans, of the Royal Soc.

Thomson,

Journal,xx. (1885),pp. 250"260


XXIV.
(1895),part il. pp. 167"171.
The

(1849), pp.

viii.

SL-hlomilch,
the integral

Enneper, Math.

CLXXii.

377"381

xviii.

Rendus

de

Ann.

lxii.

232

vi.

(1881),pp. 792"
Coates,Quarterly
V Assoc. Frangaise,

that
integration

contour

du

31); and

"

Malmsten, K. Svenska V. Akad. Handl.


Acta Reg. Soc. Sci. Upsula,x. (1832),
p.

are

(see" 7-23); Svanberg,Nova

(1841),pp. 65 "74

20

(-)"-'

/e--"-^Vi"\-j
["(i"-'

TT

r c^"-!

(-)"-i7r

e-

+pY_
The

former

of these results may

be obtained

/.

cos

and

obtainable

the latter is then

Whittakers*'

6" 17.

Formulae

xu

the equation
by differentiating

du

^^

ire

'^P

by using Lagrange'sexpansion.

of Hankel's
generalisations
in

of the type contained

equationshould

p=l

be constructed

integrals.

" 3'32 suggest that

solutions of Bessel's

in the form

zi\'e^''Tdt.
J

be shewn

It may

by

the methods

" 6'1

of

that

v,.|."/%.""|^
r*

and

so

the

functions
each

end

integralis a
of order v"\

d"T

("

solution if T is
and

dT

solution of

the values of the

Legendre'sequationfor
at
integrated
part are the same

of the contour.
*

Proc.

London

Math.

Soc.

xxxv.

(1903),pii.

19S"

206.

174

the contour
Legendre function Q^_j(^),
may start
is
where
acute
or
an
negative)
angle(positive
iexp(" ?'")),
satisfies the inequalities
to be

If T be taken
end

and

at

+00

providedthat

the

Itt+

"

If T

taken

be

arg

^TT+

z"

Py-^{t),the

be

to

"

of P^_i (^ at
singularity
logarithmic

to
joining
impossible
in
considered
the special
"3'32; for a detailed
case
in the generalcase, see " 10"5.

We

proceedto

now

"

take various

is

(when

"

0).

possible;but the
an
|
integer)

contour

same

take the line

it

makes

[CHAP. VI

FUNCTIONS

BESSEL

OF

THEORY

to

t-

is not

"

as

except in

contour

discussion

of the

integral

in detail.

contours

First consider
/"(-!+,1+)
zi

e'''Q,..(t)dt,
J

phase of

the

where
contour

crosses

1 and

1^1
=

the similar

t is

ootexp("

the

at

zero

/(o)

the real axis. Take

the

point on
the contour

to

expand Qv-^{t)in descendingpowers


of " 6'1,that
analysis

rightof ^ 1 at
lie whollyoutside
=

of t.

It is thus

which

the

the circle

found,as

in

r(-i+.i+)
(lz\^-e-^^"^^'!'^^

J^ (^)

(1)

^Hn^n
^

I'

and

"

\2)

""*

xi

exp

Q-i (0 dt,

2u))

therefore

J-A^)

(2)

-^l,^.

^''
J

""^va/

If
kinds

combine

we

and

these formulae

r
i (
cos
(I)
2 )

TT

use

Q-^-k (t)dt

iu")

the

relation*

connectingthe

two

find that

Legendre functions,Ave

of

ixiiexp

VTT

J
J ooiexp(ia.)
ooiexp(iai)

Again,consider
"(1+)

zH

e'''Q,_^{t)dt;
J

at:
"K

i AT-n
exp ((
7

"

uj)

},.\)
i

this is a solution of Bessel's


the line R

rightof
z

-^-\- coi.
^

"

00%,

Hence
we

be taken to lie on the


and, if the contour
equation,
is 0 [z^exp {"a\z \)]
{t) a, it is clear that the integral
as
the integral
is a multipleof ^^"' (^).Similarly
by making
=

find that
/"(-1+)
?iU

"'~'Q,_,{t)dt

J ooiexp(
Tlie relation, discovered

"

i"o)

is
by Schlafii,

p"(.) =

*^{Q"

(.)-"_"_,(.)}:

TT

cf. Hobson,

Phil.

Trans,

of the Royal Soc.

clxxxvii.

(1896),p. 461.2

6-2]
is

INTEGRAL

REPRESENTATIONS

multipleof iT/-'(z). From

H^^H^)

(4)

*-

V2/

C^z^h
p-h'v+h)ni r(-l+)
^

H^^^ {z)

with

(TLz\h pk(v +

TT

this is also obvious


The
of

1 is

"

from

which
integral

h)Tvi

r{-\

COS

VTT

e^'^P._i (0 dt;

j-exp{-ioj)

rjj

(3).
differs from

the point+
(6)onlyb}^encircling
inside such
analytic

(5) and (6),arg {t+ 1) vanishes where the


the rightof
1, and, in (5),arg (" 1) is
tt
In

on

"

"

6*2. Genei^alisations
We

system

Schlaflif.The
to

We

the

Hankel's

take

that

at

"

contour.

the real axis

crosses

point.

of Bessel
representations
contour
integralsdue to
which

formula

fundamental

in
integral

Bessel's

various

integralsand

definite

of

contour

1 instead

of BesseVs integral.

examine

shall next

Q,_,(t)dt,

Schlafli's relation,

since the integrandis

zero

e-'

l^.\^.^^^^^
(2/

H.^^ (z)

(6)

e"^Hl_,{t)dt,
exp(-iaj)

00!

"

(1) it is then clear that

of

^^p.i. I

hence, by " 3"61 combined

and

consideration

7*

(5)

175

will be obtained

of functions

case

well-known

whose

order is

by

Sonine*

and

is

reduced
easily
integer.

an

of the
generalisation;]:

functions

Eulerian

second

integral
1

r(o+)

l)^^i].^
^

V{v
in which

the

phase of

increases from

tt

"

to

tt

'""^*^^'

"

as

describes the contour, and

then

\^

2
Consider

'

^
TT

e^ dt.

obtained

the function

t-"-^-^

the signsof
by interchanging

summation

the right;it is
on
integi'ation

and

/"{0+)

Z-)

function of z for all values of z, and, when


expanded in
analytic
ascendingpowers of z by Maclaurin's theorem, the coefficients may be obtained
with regardto z under the integral
signand making z zero
by differentiating
after the differentiations ". Hence
This is

an

S LJ_llfZ_
t-''-'exp\t-~[dt=
'/h!

-00

Matlmnatical

t Ann.

di Mat.

also be consulted.

Hopkins
+

Cf.

Collection,

'it)

v.

(Moscow, 1870) ; Math.


His

Ann.

memoir, Math.

(2)v. (1873),p.

204.

In

Graf, Math.

addition,see

t-''-^--U'dt,
J

"j=o

UniversityCirculars,xiv. (1895),pp.
3Iodern Analysis," 12-22.

Ann.
20

"

lvi.

-00

xvi.

An7i.

(1880),pp. 9"29.
(1871),pp. 134"149,

iii.

(1903),pp. 423"432,

and

21.

" Cf. Modern

sliould

Chessin,

Analysis,"" 5-32,

4-44.

John

176

THEORY

and

so

have

we

at

"^'W=y/_j-'--pK4*
result,which

This

When
it passes
real axis.
On

the first to

was

writingt

Sonine's

given in

was

Again,writingu

J, (Z)

(3)

jarg^ [" ^tt. This

valid when

In this formula

Fig.8, with

about

angle arg

an

find that, when

|arg

the

originuntil
the negative

with

\" |7r,

earlier paper

e2Sinh",-K!.
^^y^

^r".

is the first of the results obtained

take the contour

vertices at

oc

(p.335).

have

e'",we

the contour

by Sonine;

2^-L""~''"'exp|i^(i.--j|^z^.

^"'^^)

form

then

Izu, we

by Schlafli,was rediscovered
pointout its importance.

may swing round


direction making

Iarg z\"^7r, we
in a
to infinity

(2)

in

discovered

was

the latter writer

This

[CHAP. VI

FUNCTIONS

BESSEL

once

(1)

and

OF

"

to

iri,
"

consist of three

iri,rri and

sides of

by

Schlafli.

as
rectangle,

m.

ni

-TTl

Fig. 8.
If
on

write t +

we

for

tti

the lines joining0 to

(4)

/.(2)=-

valid when
If
if R

we

jarg
make

(v)" 0, so

is still true
The
second

tri,
we

f''cos(i/6'-^sin^)rf^-^H^^
["e-^'-^^'^ht
^^^

\" ^tt.

the rightis continuous


on
and,
" ^tt,the first integral
arg^
also is the second, and /^ (z)is known
So (4)
to be continuous.

when

-^

^r

is

pure

imaginary if R (v) is positive.

integrals
justdiscussed
memoir;

to the real axis and + id for w


parallel
s generalisation
of BesseVs integral
get Schlafli

the sides

iv on

in that

memoir

modifications
appropriate
or
(positive
negative)and

were

examined

he obtained

of the contour.
if

For

methodicallyby Sonine in his


definite integrals
numerous
by
acute
if
be
angle
example, yfr an

6-21]

replacedby
(tt+^|^)^. By takingthe contour

the contour
+

00

corners
as

REPRESENTATIONS

INTEGRAL

in

be

(3) may

177

which

goes from
be three sides of

one

to

i,(tt+ y}r)
i,and
y}/)i, (tt -v/r)
modification of (4),
at

cc

(v

"

"

viip

"

with
rectangle

obtain,

e'^''*"''"-'^^*'cos(i^^-^cos-v|rsin^)rf0

e-"'* sin

f"

ITT

Again,if we take -^ to be
be replaced
by one which

anglebetween

an

passes from

oo

0 and

tt,

,,^.,,

the contour

to go
(|7r4 "y\r)i

"

Jo

TT

so

"

fn

Ju{z)

(5)

i to
(tt yfr)

"

(tt+ -v/r
)i,we

co

"

in

(3)may

and
(|7r4-\|r){,

find that

we

riT + ^

(6)

1
1["

./^(^)
=

1
^_

cos(i^^-^sin6')r/6?

f"

g-zsinh"sin^-.,?
g-j-^
COsh t COS
(2r

ttJ 0

providedthat |arg
When

"
{i")

\is less than


and

is

both

and
"\\r

tt

"^

I^TT

dt,
J^a/t)

^.

"

positive
(= x),we

take

may

"|r

cosh
(a;

in the last

formula,and get*
fi'^

(7)

./^(a;)
=

Another
the contour

by

Jordan's

cos

1 r^

{yO

"

sin

e""' sin

0)dd +

c/^.
I^'tt)

"

out
importantformula,derived from (1),is obtained by spreading
axis on the rightof the origin;
until it is parallel
to the imaginary
this is permissible
if R{v)"
lemma
1, and we then obtain the
"

formula

J.(^)=^-^^-J t-''-^exi^]t--\dt,
4^1
J

(8)

^TTl

in which

may

which
Integrals

(1899),pp. 121"161

Whipple, Proc.

resemble

W.

London

Math.

Soc.

(2) xvi.

94

(1917),pp.

substitute Schlafli'sintegral
" 6-2

we

rightof

the

of

importancein

Soc.

the

tion
investiga-

Math.

Proc. London

of

Soc.

xxx.

(2)i. (1904),pp. 393"414;

HI.

"

second and third kinds.

(4) for both of the Bessel functions

equation
=

v-TT

/_,.{z)cosec

vtt,

find that
C^

fir
"n

basis of many

is the
integral

Math.

Jackson, Proc. London

Y^ {z) /" {z)cot


we

this

given in this section are


a
prism ; see Carslaw,

those

lightby
H.

which represent
uf the
functions
Integrals

6-21.

the

have any

of the diffraction of

on

"

value
positive
investigations.
c

Sonine's

If

Y, {z)
=

cot

COS

VTT

J (,

{v6
-

sin d) dd

cosec

cos

pit

(r6^+

cos

VTT

W.

13. V.

Cf.

Gubler, Math.

sin

dt
e-"'-''"'^''

Jo
*

^)dB

.'0

Aiiit.

xi.ix.

f"e"'-""'!''
dt.
J^

(18',)7),
pp. 583"584.
^-^

178

OF

THEORY

Replace^ by

tt

^ in the second

"

[CHAP. VI

FUNCTIONS

BESSEL

the

on
integral

and
right,

it is found

on

duction
re-

that
1

(1) Y, (2)
=

/'""

sin {2 sin

C"

-\ (e"*+

e-v6)d6-

e""* cos

v-n)g-^inhf ^^^

discovered by Schlafli (who actually


formula,practically
ing
gave the correspondformula

By

means

which
function),

for Neumann's

of this result

can

we

TTlj

Iarg zIk^tt;

when

is valid when

\" ^tt.

evaluate

_x

take the contour

for we

[arg

be rectilinear,
in
as

to

li

"

Fig.9,and

Fig. 9.

write
the

iri for

t, id, t +

"

the three parts of the contour

on

then

we

see

that

is equal to
expression
1

"] ("00

and

p"vni

/""

this is

equalto /" (z)+ i Y^{z)from


Hence, when |arg z\"^7r, we have

formula

roo

(1) combined

with

" 6"2 (4).

".( "'e^sinhw-w^'
^^^

(2)

jy^d'(z)

(3)

JY,(^)
(z)=

.!
TTlJ

"'

e^ ^'"h

fi^

-uw

-co

Formulae

discovered by Somnierfeld, Math. Ann. XLVii.


equivalentto these were
(1896),
The only difierence between
these formulae
and Sommerfeld's
is a rotation
of the contours
through a rightangle,with a correspondingchange in the parametric
variable; see also Hopf and Sommerfeld, Archiv der Math, und Phys. (3) xviii.
(1911),
"

pp. 327"357.

pp. 1"16.

By

an

obvious

change of

(4)

ir,(')
{z)

(5)

H.^^ (z)

"

variable

I
1

we

M-"-^

roc

A.,
^

may

write

(2) and (3) in the forms

expU^[u
-

exp(-7ri)

,-.-.

exp

du,
^^j
/

|.(u-i)|
.

du

"

180

OF

THEORY

Modifications
thus found

of

w
by
by replacing

obtained

(2) and (3) are

[CHAP. VI

FUNCTIONS

BESSEL

"

^iri; it is

that*

HJ^'H2)

(8)

ei2coshw-vw^^

"

2g" i'"ri fao+iTT

gizcosh

ir^(-'
(Z)

(9)

qqqY^ J^W

dw,

Jo

TTl

d^
g-izcoshw-,'W

r-

TTl

-oc+hni

2ei
g-izcosh

qqq]^ j/|y

dw,

TTl

providedthat |arg zIk^tt.


(=") in (6) and (7)
by taking^^ positive
and "l"R{v)"l.
of Jordan's lemma
A double application
(to circles of
shews that,in such circumstances,we may
largeand small radius respectively)
take
^7rin (6) and
^tt in (7). It is thus clear,if u be replacedby
Formulae

(o

"

of

interest
special

arise

"

that
ie'',

(10)

HJ'^Ux)

e^^coshi-.'^^^^

r-

ijkvni

(11)

ff/"^(x)

hence, when

"

J^,(x)=
-

F^(a;)
=

"

1^

g-ixcoshtcogh i,t dt,


.

"TTl

00

"

"

iit(i^)
"
1, we

J 0

have

C'^

sin

TTJ

(13)

cosh z^^f/",

Opkvrri I"ao

roc

g-ixcosht-ut fli

0 and
2

(12)

J0

7n

TTl

and

cosh
e'-'^

__

j_oo

TTl

(x cosh

"

^ vtt) cosh
.

vt

dt,

-I

(" cosh

cos

"

j/tt)cosh
.

i/^ c?^;
.

and, in particular
(cf."6'13),

F.w=-?r5^*,

(15)
.

when
The
Math.

replacecosh

we

last
Ann.

Proc. Camb.

two
xvi.

formulae

by
are

^.
due

to

Meliler,Math.

aud
(1880),p. 39, respectively;

Phil. Soc.

viii.

(1895),pp.

122

slightlydifferent form of (14) has


if in (14) we
(1901),pp. 369"384;
write

also been

discussed

SoDinc,

by Basset,

Journal,
given by Hardy, Quarterly
find that
J{ah),xt^au
+ bjw,we

The

reader

will find it instructive

A(cos^)

^
=

77

combined

(1872),p. 142, and

xxxii.

^^)^^=.yo{2v/("6)}.
j'%in("u

(16)
Note.

they have

v.

128.

"

been
x

Ann.

with

the formula
*

Cf.

to obtain

f'- sin(n +

(14) from

the formula

^|)(^

je ^/{2(cos^-cos0)}"^

5^5-71 (1). This

was

Mehler's

Coates, QuarterlyJournal,xxi.

originalmethod.

(1886),pp.

183"192.

6-22]

INTEGRAL

which
previousanalysis

is

easilymodified

(2) and (3) are

formulae
If in

(3) the
"

SO

that

prove

(!"""'""-"""
dw.

valid when

arg^:

^tt if E

be three sides of

to

(v)"

0.

with
rectangle

corners

that
I"^

'

ir

Sin

VTT

e'^'^'^cosvede

e-'^^^^'f-"'
dt,

that
2 sin

/_. (z)

hence, when

and

jarg^

77 77

e-^c"**h"
cosh vt

L {z)
=

"

K,

(5)
a

to

as

00"17}

iri,it is found

IJ2)=-

(4)

is taken

contour

x
Tri, iri,-rri,

"

so

"

=J"-V
-iTT%

00

givenfully;theyare
mainly in terms of the

|arg z\"^'ir,

/, (z)

(3)

at

of sufficient interest to be

K^ (z)are

of " 62
analysis

hence, when

The

in the discussion

F{a, t)of " 415.

function
The

involved

arc

Schlafli*,though he expressedhis results

to

and

of the

modifications
of I^ (z)and

due

181

/"(^) and Ky(z).


representing
Integrals

6 "22.
The

REPRESENTATIONS

dt,

^tt,

{z)=re-^coshtcosh

jjt
.

dt,

Jo
obtained

formula

From

by Schlaflif by

the results

evaluate

can

e^ cosh

z\"^7r.

\ arg

For

it is

cc

"

d^

that
r^+TTi]

r=o-7r/

"I

_!-. J
27^^ [j_^_^,-

27nJ
"

w-uw

"ni

easilyseen
=

tions.
elaborate transforma-

of somewhat

means

justobtained, we
27ri

when

"

^.j

d^fj
gZCOS\MV-VW

J oo-ni)

TTi

e-"^'''^''-'''dt
+ I,(z)

"l-TTlJ
_:

1i
(2)v. (1873), pp. 199"205.
(2) v. (1873),pp. 199"201
(1868),p. 131, as the definition to

Ann.

di Mat.

Ann.

di

Math.

Lxix.

Mat.

I/TTl

+ L{z\
[I_^{z)-L{z)]

sm

vir

this

which

formula

reference

was
was

used
made

by Heine, Journal Jur


in " 5-72.

[CHAP. VI

FUNCTIONS

BESSEL

OF

THEORY

182

and

hence

e^'''I.Az)-e-'"''I,{z)

oo+wi

g0COShW-vW(^,;y

(6)

"

27rlj

Again,we

KAz)

(7)

^r

hence, by the processes used in

and

e-'^^'^"-''Ult,

" 6-21,
r

n
fee exp ( ill)
/""=exp(-i".)
1
-

^"(^)

(8)
when

"

TT

"

fi)

"

u"^exp

and

TT

vir

the form

(5) in

write

may

2i sin

-oc-TTl

'

-) du,

Oexpiuj

^tt+

"

1\)

iWw

"

"

arg

^tt+

"

(o.

Similarly
r

sinz^TT /'""exp(,r-a"),"^

(9)

e""'

.s

/_, (^)

e-''"

-r/x

I, (z)

it-"-^ exp

7^

this is valid when


The

0 "

aX

27r and

for the formulae

contours

J0exp{-7r+"o)j

^tt+

"""

(8) and

arg

(9) are

"

^tt+
shewn

ix)

nz(u+-h
V

du ;

"^^/j

co.

in

Figs.13

and

14

respectively.

Fig. 14.

Fig. 13.

Further, when

(= x) and
positive

is

"1"

round
be swung
it
is
thus
axis
found
that
imaginary
;
in

(8) may

the

path of integration
half
of the
positive

R(v)"l,

until it becomes

the

Iv

-^v(^) le"^""'
=

so

y"""^ exp

^-lixiv-

that

(10)
and,

on

(11)

K^ (x)
=

changing the signof

^^^
le-i""j e-f^sinh(-vt

v,

K^ (x)

I ei""^*r
J

e-'*sinht+vt ^f^
00

6-23]

INTEGRAL

From

these results

(12)

we

REPRESENTATIONS

that

see

IvTT.K^ (x)

cos

183

g-'^sinht ^q^]^ ^^

that

SO

K^,(x)

(13)

f"^

(a;sinh 0 cosh

cos

COSjl^TTJo
these formulae

and

In

all valid when

are

"

0 and

;4

(^^,

"l"R(v)"

i.

particular
K, [x)

(14)

{x snih 0

cos

dt

result obtained
It may

by

be observed

Jo

Mehler*

"
{z^)

0.

whom

exp

among

(cahier16), 1813,

p.

integralon

The

the

be mentioned

may

-I)

^ze'
=

j^ { "}

provided that

the substitution

make

A'.(^) I {^y

(15)

(t"+

in 1870.

if,in (7),we

that

),/

: 0

^^^^

T,

we

find that

been

righthas

studied

Poisson, Journal

237; Glaisher, British Association


and

bj

de V Ecole

Report, 1872,

maticians,
mathe-

numerous

ix.
Pol^technique,

pp.

15

"

17; Proc.

Bull, des Sci. Math.

(2) xvi. (1892),


Kapteyn,
(1880),pp. 5"12;
whicli
has
the
discussed by
in
values
and
were
integrals
special
\
f
pp.
odd
half
of an
Int.
415
when
Inst.
Calc.
is
iv.
and,
1794),
v
(Petersburg,
;
Euler,
p.
been
evaluated
has
Exercices
de
Cidcid
t
he
i. (Paris,
by Legendre,
Integral,
integer, integral
1811), p. 366; Cauchy, Exercices des Math. (Paris,1826), pp. 54
56; and Schlomilch,
The
280.
gration
Journal fur Math, xxxiii.
integralin which the limits of inte(1846),pp. 268
958
has
been
examined
xii.
Binet,
Rendus,
are
(1841),pp.
by
arbitrary
Gomptes
Phil. Sac.

Camb.

41

in.

The

44.

"

"

"

"

962.

6*23.
The

Hardy

of Du
formulaefor integrals

Reyinond's
type.

integrals
r=c

r-a

/j,2
_

I sin
Jo
in which
of Du

Bois

./" "

Bois

0, "1"B

sin

"

i""^

dt,

gO.
cos

t
.

cos

{v)" 1, have

been

^""^

"

dt,

Jo

examined

by Hardy f

as

examples

Reymond's integrals

r txtf^'^t.f-'dt,

Jo

cos

a differential equation
f{t)oscillates rapidlyas ^ 0. By constructing
of Bessel
them in terms
in expressing
of the fourth order, Hardy succeeded
them is to make use of the results
functions ; but a simplerway of evaluating

in which

of

6-22.
""6-21,
*

Math.

Ann.

t Messenger,

(1881),p. 182.
(1911),pp. 44"51.

xviii.
xl.

184
If

it is clear that

replacet by xe\

we

sin ^ sin

t"-''dt

"

sin (^e"*)e"*c?^
{xe^)

sin

x"

Jo

[CHAP. VI

FUNCTIONS

BESSEL

OF

THEORY

-"

roo

j^ "v

J
=

hence

and

_,

(2a;)],

have

we

-rT-^
sinisin^.^''-irf^

(1)

"00

7i
2e-^'"^'
iT^(2a;) 2ei'"^^

and

{2x)
[7rtW'""'f/'^_^*
(2.2;)TTte^"-*H^^^j,

i^"

gf*^^
g-2ia;cosli(gSiasinhf g-2ia;siiilitj
fg2ij;cosh"

+ /_,(2^)-/,(2./0][/.(2.t")-J"_,(2.r)

similarly

(2)
When

^ cos

cos

these formulae

zero,

become

+ A'"(2a.O,
sinisin^.^
|7rFo(2a-)
t
t

(3)

.'0
"

(4)

cos

^'"^^ IttFo (2a;) K, {2x).

cos

6 "24.

it

Jo

[/_.(2^) /.,{2x)+ /_,.{Ix) 1" {2x)].

value
special

has the

cZ^

r-i
.

extension of BesseVs integral.


Theisinger^s

curious

extension

of Jacobi's formulae

of

obtained

has been

" 2*2

in the

of

case

and

Ji{x) by Theisinger,Monatskefte
fiirMath, und Phj/s.xxiv. (1913),pp. 337
shall now
formula which is valid for functions
of Theisinger's
give a generalisation
V

"

where
If

341

Jq {x)
; we

of order

h"v"\.

is any

it is obvious

positivenumber*,
^

J^(x)=
'"' '

^^fj [""
e-"'^"'"cos(A'cos^)sin2''^c?^

j^

A)r(i) jo

r(r

integralthat

Poisson's

from

'-"f/

"

^r(r+|)r(i)jo
^^^

(i-e-"^8ine)cos(^cos^)sin2''^"Z^.
)
\
I
^

/"in"

Now

sin-" "9o?^
(9)

COS
(1 -e-"^sinfl)
(.rCOS

}0
Ttt

contour

find that the last

1-exp

above

origin,and

sinh ix sin ^

cos

u;

6) sin-" ^cZ^

5)
"?2
/._-l/A^''

(.-!/.)}
(lag;.origin. Take

the

write

\^

+tan

z"

the

parts

^^,
sui

T"

(x cot

r,-

sui

ix tan

(^)

^0)

e""^'cot-" d)
.

f ^-^--^W^^-^^^t'^)
sin Gr
l"^).e-

"-'

tan

/o

"

4A I

Jo

sin

/I
/I
JLN
cot d)
cot (/")
(ia.r
cos
(iax
\_
4.

4.

In

cot^"

-^

sui(.rcot0)

[-'"

2'

of the contom*;

-r

"

sin (.rtan

vn)

Theisinger's
analysis,a

^^

sin

-"^

+ ^"

to be the real axis with

contour

the two

on

2/

expressionis equal to

/"!

passes

the

at

;rs

sin

the

"

[x

where

"

snili

^
^^'V
j_t sinhl:|.rz(2-l/2)}

__

indentation

g"

^^TT

aarsin^

sin

is

au

"p

-^

c^
^sm0

r-",
(.rcot 0)

cot^" d,

even

c?6

i"^)

'-.

integer.

-^-^
sin

we

an

thus

6-24-6-31]
and

INTEGRAL

REPRESENTATIONS

185

therefore

(1)

.wis/

"

"

-^.(""")=

e-"^-^"'e

(.rcos^) sin'^"^c^^

cos

""

j
The

case

to

(h-un)
sni

(I

hecause

I'^A,

fails when

l differs from

cot

"

the radius

with

zero

cot 0) cos (Ur


(ia.v
"

transformation

tend

not

"*"'

sin

the

he works

^i/(*") f
^-fej^I^^-

with

"

The

form

sui

the indentation

does

given by Theisingerin the

(7) which

3"3

'^

(.rcot (Ji)

integralround

of the indentation.

(1) because

_^^
(f_^50)
^ot^^ "i

gives

'"

(2)

"""'=''"^

f-""

"

4,

,1

/I
(iax cot (b

i.j\

(^a.rcot 0)

sm

xj

cos

T-

jo

d cos
cos
("""
e) sin2''-'.^

"ii

X-

sin'''
(i^ tan

vtt)

"

"^

T-

sni

^ cW

i(i)

deb
,,,

fJ-icot-"--

.,

(a-cot (/))

(h ^-J"

^sni^"^

providedthat ^"v "f.

6*3.

The

Three
obtained

of K^{z).
equivalence
of the integral
representations

different types of
in

""6*15 (4),6"22

which represent K^ {z) have


integrals
and
G"16
(5)
(1),namely

now

been

e--^'""'"*coshi/".c?^,

Jo

demonstrated in 1871 by
directly
of the second and third as early
Schlalii*;but Poisson proved the equivalence
of the
1813, while Malmsten
as
gave a less direct proof of the equivalence
second and third in 1841.
We
proceedto describe the three transformations
in question.

The

equalityof

6'31.
We

the first and

arises from

(1)
r("+^)
.

be established

may

We

have,of

at first to taket
*

more

t The
tion

or

of the

Ann.
by Schliifli,

analysisused

di Mat.

(2) v. (1873),

{\z)^
j ^_,^

^y_.^^^ ^r

"

j,

e-^coshs cosh vOde

j"

'

and
of K^ (z),
comparison of two of the integralrepresentations
that
of
2"323.
by analysisresembling
^

course,

to suppose

-~h"R

{v)"

tliat R

{z)"

0 to

secure

convergence,

and

which

it is convenient

1.

filrMath. xvii. (1837),pp. i-iS" 242, but it is


elaborate than Scblatii's invesligation.
for larger values of 11 (c)eitlier by the theory of analyticcoutinuaresult is established

earlier

Au

much

abstract

201, to prove the relation

"

which

was

Scldiifli's
transformation.

first give an

pp. 199

second

by the

proof is

use

of

due

recurrence

to

Kuinmer,

fornuilae.

Journal

186

THEORY

Now

where

on

define S

by

and

x^l;

the

cosh

0, we

J,
that,by

see

have

we

u,

integrandin

last factor of the

Replacingx by

so

equation

theu^ii t=x-{x-l)

expanding the

[CHAP. VI

FUNCTIONS

BESSEL

OF

of

powers

and

term-by-term.
integrating

that

uFii)

(coshd-ty

partialintegration,

^'^^^""'"^^
1 e-^^^i^esinh ^sinh r^cZ^
/""e-^eoshecoshr^c/^^^"^^^","'"^^

jo

r(i)

jo

"FC*)

r(i-.)jo

F(I^)/,i/"'""'"-'""'

rci-.-) ji
2"

of the

the transformation

6*32.
The

to Poisson*

is true

when
of the

course

of

and
If

we

by

presents
integrations

no

and
great theoretical difficulty,

transformation.

proof that

is due

of the

zt"

is established.

Poisson's
direct

order

re-'^f"-iy-idt;
J

the inversion

jo

j-.

^""'r(i)

Journal

de VEcole

x"0
|arg2|"-|7r,
proofthat R{v)"^
an

appeal to

replace^ by

suflficientto prove

new

ix. (1813),pp.
Polytechnique,

and

R{v)"

and

|arg

"

-g,

|"

formulae

recurrence

variable

jo {w^+z-^y^h

defined

jtt,

and

by

but

it is convenient

and

to derive

the

the

The

239"241.
to

equation
in the

assume

the result for other

values

theory of analyticcontinuation.

equation v^x^e'^,

we

see

that

that

f%x".r_i,/.v..
pcos(^,70rfu_lr(l)
"2,-i/.v^^
*

See

also

Paoli,Mem.

di Mat.

di Fis.

della Soc.

Italiana

delle Sci.

xx.

(1828),p.

172.

it is

188
and

the first is

of the third

unboundedness

multipleof

the second

C is

independentof

C, make

and

x-^0

then

(.)r (^) _Cri2v)

sotnat

R{p)"0,

follows,when

requiredtransformation

'

z'"

2z^^T{v+ i)

if we

the

use

duplicationformula

function.

for the Gamma

in finite terms
expressible

for it is

"

transformation

of Malmsten's

consequence

,.

"

,,

",,,

immediate

An

constant

(u^+ z^)

the

he

JX

-^

determine

To

.v.

/
and

of the

In view

that

so

J 0 {U +2')"
where

third.

the first must

that

it is obvious

x-^+oo,

as

and

of the second

linear combination

consequentlya

[CHAP. VI

FUNCTIONS

BESSEL

OF

THEORY

is

r,

that

"^

[
,

cos

xu

du

^^

^"

2m

equal to
roc

^' ^^
2^-i{(n-l)!}^j,
'

"-i(2a's)'"(2/i-"i-1):

^g-w

given by Catalan,
evaluatingthe integralis simpler than a method
is not rigorous in all its
Journal de Math. v. (1840),pp. 110
114; and his investigation
is discussed
by Serret,Journal de Math. viii. (1843),pp. 20,
stages. The transformation
also
193
see
216;
Cayley,Journal de ,Math. xii. (1847),p. 23G
21; IX. (1844), pp.
This

method

of

"

"

Papers,i. (1889),p. 313.]


{Collected

Airi/sintegral.

6*4.
The

integral
r

{t^" xt) dt

cos

Jo

of Light in the
appearedin the researches of Aiiy* "On the Intensity
which are
of a class of integrals
neighbourhoodof a Caustic" is a member
tabulated by Airy
in terms
of Bessel functions. The integral
was
expressible
laborious. Later, De Morgan "(but the process was excessively
by quadratures,
obtained a series in ascending
powers ot'x by a process which needs justification
either by Stokes' transformation (which will be explained
immediately)or by
the use of Hardy'stheoryof generalised
integrals^.

which

Trans.

Camb.

Phil.

Soc.

vi.

(1838),pp. 379"402.
cos

In

Airy used the

form

(iv'^mro)die,
-

thif?is easilyreduced

but

f The result was


(1849),pp. 595"599.
+

to the

communicated

Quarterly Journal,

xxxv.

integralgiven above.
to

Airj'on

(1904),pp.

22"66

March

11, 1818

; 2'rans.

Camb.

see

Trans.

Phil. Soc.

Camh.

xxi.

Phil.

Soc.

(1912),pp.

viii.

1"48.

6-4]

INTEGRAL

It

noticed

was

Stokes

and

The

Stokes*

also obtained

values of x, both

with

by

reader

will observe

189

is annihilated by the
integral
operator

asymptoticexpansionsof
negative.

and
positive

the

integralfor large

that Stokes' differential


forms

{loc.cit.p. 187),but

Stokes

that the

the

of the transformed

one

REPRESENTATIONS

of Bessel's

integral is identical
equation for Aii-y's
equation (^ 4-3). This fact was noticed by

made
of it until Nicholson,Phil. Man.
specialuse was
(6)
of Bessel functions of
XVIII.
(1909),pp. 6"17, expressedAiry'sintegraldirectlyin terms
orders
considerable importancef; see
"7'.. These Bessel functions have latelyassumed
(1910),p. 267, and the approximate formulae described in " 8'43.
Weyl, Math. Ann. Lxviii.
no

Subsequently Hardy, QuarterlyJournal, XLi. (1910),pp. 226"240, pointed out


between
Airy'sintegraland the integralsdiscussed in 5^.^
6"21,6-22,and
of Airy'sintegral("" 10"2
various generalisations
examined
10'22).

the

nexion
con-

he then

"

To

observe
we
Airy'sintegral;]:,

evaluate

that it may

be written in the form

1 r*
^

consider this

Now

exp {if

ixt)dt.

integrandtaken alongtwo

of

arcs

circle of radius

with

the arcs terminating


and pe^'^',
at p, pe"'^''
origin,
pe^^respectively.
tend
these
to
The integrals
arcs
oc
zero
along
by Jordan's lemma, and
asp
obtain
Stokes'
transformation
we
hence, by Cauchy'stheorem,
the

at

centre

-*

rcoexp^TTt

/"CO

{f "xt)dt

cos

-^
2

Jo

lr
"^

.'0

.'CO exp f TTJ

[e^""'
exp (- T^
second

the contour

of the

originand

the third

"

exp {iV""

e^'^'
^t)

ixt)dt

^'^'

integralconsists

exp (- T^ "

of tvro

dr ;
i'^'xr)]

emerging from the


re^'^*
for t on these
by writingre^'^^

integralis obtained

rays

rays.
series
Now, since the resulting

Jo
*

pp.

Trans.

(-

exp

73

" e-s"''
xt) dr

convergent, it may

are

^-^

Phil.

Camb.

329"349.]

See

also

Soc.

ix.

Stokes'

(185ii),
ppletter of

t'"
'^""'-

111=0

12, 1848,

ScientificCorrespondence, 11. (Cambridge,1907),pp. 159


t The
account

pp. G09"

X The
Soc.

Sci.

functions

occur

in

problem (with
619; xxx.
(1915),pp.

of the

Bruxelles, xvi.

" Bromwich,

dr,
(- 7-*)

[Math,

and

Airy,

Sir

to

Pli,/s.Paper.-",11. (1883),
G.

G.

Stoke/^,Memoir

and

160.

problem concerning the stabilityof motion of a viscous fluid; an


(1914),
bibliography)is given by Kayleigh,Phil. Mag. (6)xxvin.
329"338.
Papers, vi. (1920),pp. 266"275; 341"349.]
[Scientific

Cf. Plardy, loc.


(1892),pp. 150"180.

integral is convergent.
de

"

exp

that"

.0

166"187.

May

be shewn

Theory of InfiniteScries, " 176.

cit. p.

228,

or

de la ValltJe

Poussin, Ann.

de la

190

OF

THEORY

and

so

cos(f

(Im
(+ a;)*"cos

^
S

xt)dt=

^^="^

^
"J

Avhich

a;

(1)

to be

{t^ Xt)dt

cos

(" ^)" sin f (m

De

by

"

".

1) TT

r
.

When

Morgan.

of Bessel functions,we

terms

is to be taken

!
{^vt+ ^j)/m

the series

obtain

the

on

Nicholson's

right

formulae, in

:
positive

/2x\/a-

^TT'\/(^x)

,x

T"'exp(-Tnc?T
Jo

(" i^)
,^=0^?^!
r(?n +

This is the result obtained


in
expressed

r=^

i7r

f)
TT
^^"

ml

,,^^0

'

are

[chap. VI

FUNCTIONS

BESSEL

J-"[373/

J,

i2x\Jx

V'3V3,
2x \Jx
/2x.\/x\

'2x\/x\
f2x\/x\

(2)

^-^l-3V3J-^n"3V3J
\Jx

Barnes'

6*5.

/2ijc^/x\

"

integral
representations
of Bessel f unctions.

of a type introduced by Pincherle*


and Mellinf,
Barnes^
By usingintegrals
of Bessel functions which render possible
has obtained representations
an
easy
4"42.
formula of "
proofof Rummer's

Let

consider

us

the residue

of
r

at

2m

r,

where

of the residues is

?'

0, 1, 2,

(2m

s) {izf

This residue is (")'"


(iz)'^"''^^'lrl,
so the sum

("^'n^amg-iz^

Hence, by Cauchy'stheorem,
J, (z)e"''
if the

(v +

that the
that

suppose

s)

"

the

2^'" .mWiv

^.

"

R{v)

When

"

|, and

"

choose

this last condition


r

(2m

~o 2^"'.mir(v
rn=t
is convergent and

Rend,

Lincei,

ser.

t Mellin
J

Gamb.

of this

del R.

has

Istituto

type, see

given a
Trans,

Barnes,

ds,

the

l)

be

contour

verified,by using

so

that, on

it,

is satisfied the series

s)

l)

equalto

4, Rendicoiiti,

Phil.

It may
points0, 1, 2,
are
integrals
convergent.

00

"

formula
Stirling's

r(2m-s).{izy
/"""+'
i

2'Tri^=0

encloses

contour

Now

(hzy 5
=

Lombardo,

(2) six. (1886),pp. 559"562;


(1888),pp. 694"700, 792"799.

iv.

summary
xx.

of his

(1908),pp.

Proc.

London

Atti

researched,Math.
270

Math.

"

279.
Soc.

For

(2)v.

della

Ann.
lxviii.
(1910),pp.
bibliographyof researches
65.
(1907),pp. 59
"

R.

Accad.

305
on

"

dci

337.

inte;,Mals

6-5]

onlypolesof

The

calculate the

we

Gauss.

to

integrandinside

the

191

If therefore

the

contour

of the residues at these

sum

change the

we

order

have

we
integration*

and

of summation

due

formula

well-known

the

by

REPRESENTATIONS

INTEGRAL

0, 1, 2,

at

are

When
....

find that

we
poles,

that

so

J^ {z)e--

(1)

1^1

relation.

is Kummer's

which

('^ I
f^^r+1)

,h {z)e^'

(2)

'^'^" +

R{v)

"

which

; 2z. +

are

v,

2iz),

1 ;

2iz).

relations

\, are

"

1 ;

find that

we

Y%'ll) i

analyticfor all values of


true.
continuation,they are universally
of

; 2r +

In like manner,

formulae, proved when

These

and

connectingfunctions
by the theoryof analytic

so,

to represent Bessel functions by integrals


in
possible
To
do
involved.
consider
is
the
function
f
actor
this,we
exponential

It is also

which

no

T{-v-s)T{-s){lizy^-^^,
function

qua

of

5.

It has
5

residue at

The

0, 1, 2,

the

polesat

2,

....

is

the residue at

"

vir

in

i^

(-)"'(2 ^)'"^'"'
m\ T {v -{-m + 1)

'

sin

'

is

-iri-"

(-)"(l2)-''+='"
'

sin

ml

vTT

(v +

1)

'

that

so

ds,
(lizY^''
^-^.ir{-v-s)r(-s)

7re-i"''+i'"'
i/,'-'
(z)

(3)

and, in like

where
the
in
X

manner,

Treic+it'^''
^,w

(4)

"

l,-v

...; -v,-v

TTJ^

while

points

the contours

start

{z)

from

~[r(-v-

and

return

to +

integrandcounter-clockwise. When |arg


be opened out, so
(4) the contours
may
i. If

(5)

we

s) r (- s)(- yzy+'' ds,

iz

to

the directions of the contours

reverse

7re-i"'+^'-'
H/'^

(z)

^. f'^"'
r (ItTI J

Cf.

Bromwich,

hir

\"

as

the polesof
encircling
in (3)or jarg (" iz)\" |7r

after

oo

start
we

from

oo

i and

find that

ds,
s) V (- s)(^izy^''

-c-jai

'Theoryof InfiniteSeries, " 176.

end

at

192

THEORY

OF

BESSEL

[CHAP. VI

FUNCTIONS

providedthat \arg izlK^ir; and

number
c is any
positive
providedthat \arg (" iz)|" ^tt;and, in each integral,
is parallel
to the imaginaryaxis.
exceedingIl{v)and the path of integration
There is an integral
resemblingthese which represents the function of the
and the argument
R(v)"0
firstkind of order v, but it converges only when
in questionis
The integral
of the function is positive.

in the

it is obtained

and

Barnes'

the

offunctionsof the
representations

^vK^)e

Consider

for the

2i.^/7r}^
r(s
the

now

{2zy

third kind.

Gamma

function

we

write

may

l)r(2i.+ s+l)sins7r"

integral
"'

2iVt.
in which

precedingintegrals
; the reader
integrandis 0 {\sl"""^).

the contour, the

formula
By using the duplication
results justobtained in the form

\^}

the

as

way

js |is largeon

will notice that,when

6'51.

same

T{-s)r(-2v-s)r(v

i).{2izyds,

_c

the

differs from the integrandin (1) by a factor which is


integi-and
in s. It is to be supposedtemporarily
that 2v is not an integerand
periodic
is so drawn that the sequences of poles0, 1, 2,
that the path of integration
;
while the sequence
of
2j^,1
2v, 2
2i^, lie on the rightof the contour
lies on the left of the contour.
In the first
^, "v~^,
poles
f,
shall
shew
if
iz
the
taken
round
circle
that,
we
a semiplace,
jarg
|" |7r,
integral
of radius p on the rightof the imaginaryaxis tends to zero
y:) ;
as
p
...

"

"

"

...

"

''

"

"

"

"""

for,if s

have

we
pe'^,

i).(2,"r
,r(-.)r(-2..-")r(.+.H).(2,-.).^p^^"/r
(s) (2i/
1)
(2i^ s)
+ . +

5 +

sm

STT

sm

TT

and, by Stirling's
formula,
r{v

i).(2izy

''^r(s
+ i)r{2i^

i)

pe'^log(2iz)-(v

and the real part of this tends to


term

with
of
p

s
cos

and

is

When

logp.
p
jsin 6 j}and the
\ exp \piT
times the integrandis
6

cos

log122 [
"

this tends to

p sin ^
"

as

"

pe'^)(logp

when

"

^
tt "
-|-

id)+ pe^^ h log(27r);


-

in

because the dominant


^tt,
^tt,|sin s-k \is comparable
the real part of the logarithm

pOamO

"

6 is nearlyequalto "
dominant

arg 2iz
/a

"

term

pcosd\ogp
if | arg iz i"

|7r.

pcosd

-{-

"

2p\^\nd\,

6-51]

INTEGRAL

Hence
the

times the

round
integral

to pass between

REPRESENTATIONS

integrandtends

all alongthe

to zero

the semicircle tends to

193
and
semicircle,

if the semicircle is drawn

zero

so

so
as

(and not through)the polesof the integrand.

It follows from

Cauchy'stheorem

that,when

|argiz\" |7rand

2v is not

an

then
integer,

r'
J

be calculated

may

2v

(- s)r (-

s)r (v +

i).{2izyds

"x"i

"-

residues at the

the
by evaluating

poleson

the

rightof

the

contour.

The

residues of
r

at 5

and

TT

r ((- .9)

2^ +

"

(v +

7)1

1) (2/^)-^
.

respectively

are

1) {2iz)'"'

2/. -s)r{v+s

r (-

TT

;/

1) (2i")-2''+"'

sin 2viT
and

rn

!F

+
(2i/

'

1)

sin ^vir

nil

r(" 2i^+

1)

hence

-Pr r
=

r{-s)T{-2v-s)T(v

s+l).{2tzYds

^^+i' -^^"
snr2;^r(2TTT)-^^^^+^-

H2^)-vr^r(i-z.)

g"

TT^e'^

sin 2t'7r
It follows that,when

i,,..(,,

(2)

|argiz |" f tt,

^-'"-'""=-(^-)-(2-)"

77 '^
-j-.i

X
"

and

when
similarly,

2/v

r (i.
+
.9)

s +

1) i2izyds,

s) r (r+

.9 +

1) (- 2{zyds.

jarg (" iz)\" f tt,

F.'^'(2)

(3)

(- s)r (-

XI

giiz-.'Tr)
cos (i^tt)
(22:)"
.

r ((- .s)

2i'

-XI

The

restriction that

is not

to be an

integermay

be removed

by a limiting
process, but the restriction that
integercannot be removed, since then poleswhich must
would have to coincide with poleswhich must
contour
manner

AV. B. F.

2i" must

be
be

on

on

in the usual
not

the

be

an

rightof

the left.

13

odd
the

VII

CHAPTER

EXPANSIONS

ASYMPTOTIC

Approximate formulae for Jv{z)-

7*1.
In

Chapter

ascending powers of the argument


well adapted for numerical
are

series

since the
large compared with 4 (i^+ 1), 4 (i^+ 2), 4(^ + 3),
fairlyrapidlyfor such values of z. But, when |-^ I is large,the
slowly,and an inspectionof their initial terms affords no clue

z^ is not

converge
converge

; when

i^

is

possiblethe calculation
Bessel's equation when
There

the

of

the

be

calculated

without

of formulae

fundamental

difiiculty.

which

recondite

aspects of the problem

latter,and

investigationof

problem
the

render

of solutions

system

of

large.

it is

be considered

to

that

made

was

behaviour

until

postponed

however, be mentioned

It must,

the

to

large,the expressionsfor Jj^^{z)

to

series

this statement

exception to

one

determination

values

; the

is

than

more

is

is not

functions

..

when
largeis very different from the investigation
former
investigationis, in every respect, of a more
V

The

character

of the

reallytwo

are

when

large.

the

chapter is

this

objectof

is

Y^,{z). There

integerwhich

an

("3'4) enable

in finite terms
The

{z) and

of J^

values

approximate

the

obtained

were

multipliedin some
computation when

z,

series

These

by log z.

functions

Bessel

representationsof

various

III

of series of

in the form
cases

FUNCTIONS

BESSEL

OF

by

the

Chapter

of Jq {x), for

some

is not

recondite

viii.

first step towards

Carlini*

gation
investi-

the

solution

before

years

of

Poisson'sf

large positivevalues

of x,

was

published.
The

formal

expansion obtained

by

Poisson

was

1^

3^

1-

5^

-Hsin(^-i,r).|^^-3J^^
+

when

is

large and

and

since

positive. But,
Poisson

to

be
*

Ricerche

1817).

his method

regarded as suggestiveand

An

siilla convergenza

t Journal

de

of Jv

Mathews,

of these

account

VEcole

on

che

serie

investigations has

to

Bessel

Poisson's
Functions

series

of

the

on

right are

serva

than

alia soluzione

already been

given

(London, 1895), pp.

34

"

vergent,
con-

in

dominant

the

term) is

convincing.

del

prohlema

di

Keplero (Milan,

in " 1*4.

(cahier19), (1823), pp.

investigation of Jq (^) has

not

remainders

obtaining the

ingenious rather

Pohjteclinique,xii.

{x) similar

Treatise

della

the

investigationof the

no

gave

series,his analysis(apart from

since

..

350

been
38.

"

352

see

constructed

" 1-6.

An

by Gray

vestigation
in-

and

196

of the formula

generalcharacter

The

of Weber, 3Iath. Ann.

the researches

and

memoir;

vi.

by Lommel,
publicationof

Studien

indicated

(Leipzig,1868), just before

Functionen

iiber die BesseVschen

F" (z)had been

for

[CHAP. VII

FUNCTIONS

BESSEL

OF

THEORY

the

(1873),pp.

146

"

149

Haukel's
also be

must

mentioned.

asymptoticexpansionof K^{z)
at an
earlydate by Kummer*
asymptotic)
formula
of
addition
the
the
corresponding
also contains
known
paper by Malmsten:]:
expansionof K^ {z).
The

study of

close

the remainders

(and proved to be
investigated
with
was
reproduced,
for /^ {z),
by Kirchhoff f ; and a littleof the asymptotic
an
investigation
was

; this result

asymptotic expansions of J^ {x\ Fq {x\ Iq {x)

in the

Sci. de I'Ecole norm.


Ann.
by Stieltjes,
Kq{x)
sup. (3) in. (1886),
been
extended
Didl. des Sci.
of
his
have
and
233
by
Callandreau,
analysis
252,
parts
pp.
of
include
functions
110
to
Math. (2) XIV.
(1890), pp.
114,
any integralorder; while
when
the variables are
results concerningthe remainders
complex have been obtained by
Weber, Alath. Ann. xxxvii.
(1890),pp. 404"416.
has

and

been

made

"

"

expansionshave

The

1906, pp.

239

"

265, and

also been

Ann.
investigatedby Adaraoft",Petersburg
1905.
by Valewink|| in a Haarlem
dissertation,

Inst,

polyt.

Investigations
concerningasymptoticexpaosionsof t/^[z)and F" {z),when
to be most
l^iis largewhile v is fixed,seem
simplycarried out with the aid
of Poisson's type. But Schlafli*[has shewn
that a largenumber
of
of integrals
of integrals
results are obtainable by a peculiartreatment
of Bessel's type,
Barnes**
has
discussed
the
while, more
recently,
asymptoticexpansionsby
of the Pincherle-Mellin
means
integrals,
involvinggamma-functions,which
examined

were

||6"5,6"51.

Asymptoticexpansionsof HJ^^ (z)and HJ-^ (z)afterHankel.

7'2.
We

in

shall

obtain

the

asymptoticexpansionsof the
third kind, valid for largevalues of \z\;the analysis,
apart
modifications,will follow that givenby Hankel ff.
Take

valid

R{v

now

The

-|7r"/3"|7r
l)"0.
expansionof

and

the factor

^'^

Journal

fur Math.

Svenska

xvii.

V. Akad.

" See the Jahrbuch

(1837),pp.
Handl.

lxii.

Tra)is.

tt Math.

Camb.
Ann.

Phil.
i.

of the

slight

some

(1869),pp.

xx.

228"242.

f Ibid,

(1841),pp.

^, provided that

xlviii.

(1854),pp.

1907, p. 492.

di Mat.

270"279.

of

is

"""'

65"74.

der Math.

(1908),pp.

491"495.

3,^ +

"

2:4:1^

H Ann.
Soc.

arg^^

in descendingpowers
(1 + liu/z)"'^

iiber die Fortschritte

IIIbid. 1905, p. 328.


**

^tt-f-/3"

-^

2z

t K.

from

" 6'12 (3),namely

the formula

when
+

functions

(2) vi. (1875),pp. 1"20.

348"376.

7-2]

ASYMPTOTIC

EXPANSIONS

197

since this

but

expansionis not convergent all alongthe path of integration,


replaceit by a finite number of terms plusa remainder.

shall

we

For

all

values
positive
integral

"2zJ

V^i2/

"^'

",=o

It is convenient

to

take p

any

effect of this choice is that, when

the choice has been

8 is

sin S,

"

2iz

S.

"

so

that

2S.

"

""'^^-h
"

"I

independentof

TT,

so

^e-!'"''):(sinS)""''-^-i'
^p,
=

z.

its expansionfor (1 + ^iu/zy-^and integrating


term-by
substituting

On
term,

is

Ap

^2-77

"7r

is restricted

given,z

consideration,and

under

2{z/
say, where

choose

ut\

ut

1-

then

we

made, then

1-

ii

'

(Itt+ /3)

"

2S ^ arg

TT

"

for the values of t and

and

"

2tz/

satisfies the
"

When

(p~\)l\2izJ Jo

j/3|^|7r S, Iarg^The

have*

we

largethat R(v"p
l)^0;
inequalities

so

positive
angle S which

of p,

find that

we

|)
_^
l).{2izy

+
(|-i/),".r(i^
"p"i

Z,

mlT{v

^^(j)'

where

^^

(1) I "

i?
^'

(I

^).

(1

i(3

exp

ty-' dt

.-U

^-(p-iy:r(v+i)(2iz)p\]o

I^^v+p- ^^f^
;

^Bp.\zrp,
where

5^

is

function

Hence, when

when

is such

and
angle";
function
The
*

due

of the

formula

Cf. Blodern
to

Graf

pp. 86"87.
Math.

the

Soc.

and

t Cf. Modern

v,

8 which

p and

and

R(v-p-l)"0

that

tt

independentof

R(v+l)"0,

Analysis, "5-41.

valid when
The

Modern

(1920),pp. 6"9

as

die

acute
being any positive
symbol which denotes a

|-^ |

oo

^^

this may

R(p-p-l)"0;

of this form

use

in

Gubler, Einlcitung

of z~J^

Theorie

der

of the

z.

have

we

2S ^ arg 2^ ^ 27r
28, 8
is the Bachmann-Landau

(1)is also

Analysis, " 2-1.

is

symbol 0
order of magnitudef

Cf. Whittaker,
xxxviii.

of

MacRobert,

BcsseVschen

(7;/(/.pp.

seen

by

to be
expansion seems
Fiinktioncn, i. (Bern, 1S9S).

binomial

Analysis (Cambridge, 1902), " 161; Gibson,


and

be

10"19.

Proc.

Edinburgh

198

THEORY

[chap. VII

FUNCTIONS

BESSEL

OF

^)"0 and then taking an integerq so largethat


which is contained in [ ] in (1) is then
B,(^i,q ^)".0; if the expression
rewritten with q in placeof p throughout,it may be expressedas p terms
each of which is 0 (^"^0or o {z~^')
followed by q"p
+ 1 terms
; and the sum

supposingthat R(v

"

"

"

"

of these q
In

similar

work)

can

we

is therefore 0

p + 1 terms

"

{z~v).

(by changing the sign of

manner

from

deduce

throughoutthe previous

" 6'12 (4) that

+^^^
ir,W(^)=('A'j%-^.-4--i:r)
"J'
_,r=o m!(-2.-.)-

(2)

277 +

Hankel,
If,following
{v,m)

28 ^ arg z^ir

of -values of

is

given

now

-1h.

write

we

(i

(-)

that the domain

and

providedthat R{v + \)"0


by the inequalities

v\n il + v)y,

ml

{v +

!r

(i/
-

{4i^^ 1^}{4i/- 3-^} {4i^^ (2m


-

"

"

1)
+ I)

m
m

1)^}

2-'^.m\

become
expansions

these

2 \^

(3)

Sl=W^'^-i

TV 2f /

''"^

For

^Ay e-M3-A.^-i^,

ir;^)(^)

(4)

brevitywe

write these

(5)

(v,m)

0(z-p)

equationsthus

WzJ

(2t2)"'

,^.0

(6)
Since

(v,m) is

" 3'61 (7),which

of

V,

"

when

|
z

function

even

of

of order

it follows

v,

"

"

with

the

the real

of

sponding
corre-

part of

^^" (^)

(;^)'

(8)

m^

(A)*

(z)

anglejustless than

Stt.

when

i-J

..F,U
+ .,l-v;

^'"'^--'"^"''^'

e-^^-i"'-^-^ .^0

"

tt

"

arg

"

"

(i
+

^,

27r,and (8) when

valid for all values

(6) are
of

i/

formulae

functions,the expansions
generalised
hypergeometric
are

(7)

(7)is valid

the

from

that the restriction that

is unnecessary.
So the formulae (1)
is confined to one or other of two sectors

In the notation of

of which

t,

functions of the third kind of order

connect

fimctions
exceeds

an

"

2^-^
'

27r "

arg

"

tt.

7-21]

ASYMPTOTIC

7 '21.

199

AsymptoticexpansionsofJ^{z),J-^(z)and Y^{z).
combine

we

the formulae
Bessel

(which express

functions of the third

of

functions

" 7-2,we

deduce

of the

from

first and

second

(-r

"
"

cos

"

(z

vir

'-^to

irz

-^A^)'^

7H

COS(^

" 3-61
of

{Izf
^

\
=

''

'

{^zf"i+i

()

(-)'" (^.2w)
"

'

^-^
"

2
cos

irt+l
(^2r)-

J-. {z)

{v,2m)

(-)"^(^',
2//i+l)'

f 7r). N

I i^TT"

"

"

^^-^

sm(2'-iz/7r-i7r). S

"

\lTZj

(3)

of

in terms

(2#-

-sm{z-h"rr-\Tr).
(2)

kinds

kind) that

'/.iz)

(1)

the formulae

"

If

EXPANSIONS

m-=0

"

1
/
1
%\Vl{z-^\vtt
-\it)

F_.(^)^

case

of functions

Y, ^z)

oi

order
integral

2??i + i)"
(-)"^(/',

co^{z

"

\mr

"

(2^)-"'+i

(-)"* {n,2m)
(2^)-"^
.

s\n{z-\mT-\'Tr). %
+

(-)'".(n,2m +1)'

{'TT).S
m

formulae

\-"^}

only),

1)1

These

C0s(2'
+ |z/7r:^7r).S

2lT

(5)

1)'

(in the

{-f.{y, 2m)

sin(^+ |z^7r ;f7r).2


+

and

.^-^-

1
.

(4)

(-)"'"
(^.2m
^

^^^-^

all valid for

largevalues of |^ |providedthat |arg z\"Tr;


and the error
due to stoppingat any term
is obviously
of the order of magnitude
of that terra multiplied
however, this factor Ijz vcaxy
by Ijz.Actually,
be replaced
of HJ^^^(z)and
by Ijz-; this may be seen by takingthe expansions
further than the last term
H^'--'
(z) to two terms
requiredin the particular
combination

with

are

which

we

have

to

deal.

dealt with when


which
are
" 7"2, the integrals
R{v)" -I represent H^,^^^
(z) and H^^-^(z),but, when R(v)"-1, the integrals
from which the asymptoticexpansionsare
derived are
those iv/iichrepresent
of J^,{z)
of
and H^"^^^(z).
treatment
This difference in the mode
H^^'"_^{z)
writei'S to think* that
and Yt,(z)
led some
for such values of i' seems
to have
formula (1) is not valid unless R{v) "
^.
As

has

been

seen

in

"

Cf.

Sheppard, Quarterh/ Journal,

xxiii.

(188:)),p.

223

Searle, QnarterhjJournal,

xxxix.

to have
originatedfrom Todhunter, An Elementary Treatise
(1908),p. ()0. The error
appears
(London, 1875), pp. 312"313.
Laplace'sFunctions, Lame's Functions and BesseVs Functions

on

200
The

[CHAP. VII

FUNCTIONS

BESSEL

OF

THEORY

asymptoticexpansionof Jo {2) was obtained


round a rectangle
e*'^'
(indentedat
(1 t^)~^
theorem
1 + X i. Cauchy's
givesat once
-

and

by Lipschitz*by
" 1) with

grating
inte+ 1

at

corners

.'0

-1

"

ei""^ e-"'+""^zr ^ (2

ui)"-dn

0,

then proceedson the lines alreadygiven; but in order to


analysis
obtain asymptoticexpansionsof a pair of solutions of Bessel's equationit
which
involves at some
method
to use
a
stage the loop
seems
necessary
discussed in Chapter vi.
integrals
the

and

the initial terms


explicitly
equations(1)" (4); they are as follows :

be convenient

It may

involved in

(-y^ {v, 2m)


.r^o
i^^T"'

to note

expansions

3^)

(4.V"-1-)(iv-

in the

2!

(8^)^

"^

4!(8^)^
(4t/^ 1-)(4i/- 30 (4i/- 5'^)
3! (82^)3

^ (-yn.{v,2m + l.)_4i"--l1182
(2zym+i
""=o
The

reader

should

notice that

J,^{z)+ J,^^Hz)o^2/(m),
a

formula

given by Lommel, Stttdien,


p.
The

Note.
of

method

which

by

F,i(2)in his Studien, pp.

respect to

but

of

93

asymptotic expansion with


It should

be noticed

that

endeavoured
to obtain the asymptoticexpansion
the expansions of J^^ (z)with
by
differentiating
97,
that the term-by-term differentiation of an
known
now
Lommel
was

"

it is

course

67.

respect to
Lommel's

later

parameter raises various theoretical difficulties.


work, Math. Ann. iv. (1871),p. 103, is free from

which occur
These errors
have been enumerated
in his earlier work.
errors
algebraical
Archives
The
221
225.
xxviir.
by Julius,
A'eerlandaises,
asymptoticexpansions
(1895),pp.
of '-Ai
Ann.
of Math. viii. (1894),
(^) and T"{z) have also been studied by McMahon,
282.
fiirMath, und Phys. xiv. (1903),pp. 275
61, and Kapteyn, Monatshefte
pp. 57
the

"

"

"

novel

asymptoticexpansionshas been discovered


some
importancein the analytictheoryof the
of the exsuch investigations
the dominant
terms
pansions

of
application
3'ears : they are

in recent

divisors of numbers.

In

these

of

adequatefor the purpose in view. This fact combined with the


consideration that the theoryof Bessel functions forms onlya trivial part of
it seem
desirable merely to mention
in questionhas made
the investigations
and
the more
recent
the work of Voronoifand WigertJ
papers by Hardy".
are

Journal

\ Ann.
Math.

fiirMath.

Sci.

de

Kongresses
Acta

in

lvi.

VEcolc

norm.

sup.

Heidelberg, 1904,

Mathematica,

%Quarterhj Journal,
pp. 1"25.

(1859),pp. 189"196.

xxxvii.
xlvi.

(1904),pp. 207"268,

(3) xxi.
pp. 241

459-534;

Verh.

dcs

Int.

245.

"

140.
(1914),pp. 113
(1915),pp. 263"283;
"

Proc.

London

Math.

Soc.

(2) xv.

(1916),

7-22]

ASYMPTOTIC

Stokes'

7"22.

!arg2^|"7r.If
between

and

tt

"

/.,(z)
=

we

201

phenomenon.

" 7-21 (1) for J^{z) was

formula

The

EXPANSIONS

established

for values

took axg z to lie between


0 and
have
tt)we should consequently

e-"^'-A

27r

of

such

that

lies
(so that arg^'g"''''

(2'e~"')
2

CO

e"

"'

(ire

cos
TT

"

i/TT

"

tt)

"/-HI

that,when

so

"

arg
/

"

2m
^ (-)"'"(J^,

Wi

(2lp

III

The

and
the

27"
(-)'".(!/,

sin(^+ |i^7r
+ Itt) S

"

shall

make

now

derived

expansionsof "7'21 are

throughoutthe sector in
asymptoticexpansion

close examination

which

tt

"

from

"

arg

iy

=11

expansion is superficially
quite different

" 7"21 (1). We

1)'

cos(^ + li/7r+ i7r) S

this

'lir,

"

TTZ

and

^-

--^^

ze'

expansionof
change.

from

the

of this

the formula

27r,the function HJ^^ (z)has

"

\m

The

expansionfor //^''-^
(z),namely
corresponding
^;^) (^)oo

(1)

"

-Hz- 4"

J-)

('\'"'\

TTZ

is,however, valid for the


valid for the sector

"

arg

E,^'^
(z)
=

and

this

(2)

sector
2^ "

cos

'Iir we

VTT

27r " arg z " tt. To obtain an expansion


the formula of " 3"62 (6),namely
use

H,^-^{ze-^')
+

e"^'

gives
-('B-ii/jT-iTr)
H,('Hz)c^(~Ye

{v,m)
..o(2izy"

^(-r.(^,-)
2cos...(iY.^-i-^\lT4yJ

The

F,'"{ze-^%

"i

\-'"^

expansions(1) and (2) are both valid when 0"arg^"7r;


them has the asymptoticexpansion

.'

now

the

difference between

2cos^7r.(" ) e''^+i-+i-" S
and, on

account

oi lower' order

U'-W

"

is
this exi)ression
the series,
multiplies
duo to stopping
than the error
of magnitude{when \z\\blarge)
of the factor e'^ which

202
at

OF

THEORY

term
definite
stop at the pth

expansion(1); for

term.

Hence

when

occurs

used in

0"arg2^"7r,

where

this

is 0

error

between
discrepancy

onlyapparent, since the

when
(e~^^z~P~^)

(1)and (2),which
(1) has to be

series in

its remainder.

with
conjunction

Generallywe

is

the

[CHAP. VII

FUNCTIONS

of the

any

we

BESSEL

have

the constants

have

values which

depend on the domain of values


increased (or decreased)
while
assignedto arg z. And, if arg z is continually
I^ Iis unaltered,the values of Cj and Co have to be changed abruptlyat various
stages,the change in either constant
being made when the function which
it is negligible
multiplies
compared with the function multiplyingthe other
I {z) is positive,
That
is to say, changes in Cj occur
constant.
when
while
in
when
is
I
changes Cg occur
{z) negative.
It is not
Cj

and

Ca are

c,

Ci

where

difficult to prove
follows :
as

le2p{"'+i)"

Co

^"^i^^',
|e-^+i)

c,

p is any

This
Stokes
which

confined

functions which

Soc.

to

possess

discontinuous

in which

le2?"(''+i)-^

him

Bessel

in

arg

"

{2p + 1) tt],

[2p7r"

arg

"

(2p + 2) tt],

asymptoticexpansionsof
involved

which

constants

discovered

was

by

in the

by

simpletype*.

asymptoticexpansionof the analyticfunction


in (March?) 1857, and the discoverywas

Stokes

made

at three o'clock in the morning. See Sir George


The
i. (Cambridge, 1907), p. 62.
Scientific
Correspondence,
h
is
the
Trans.
Camb.
Phil.
t
:
are
xniblished
following
discovery
are

and

Stokes

third of these

to

seems

have

been

the last paper

Acta

Math.

; 283"298

written

by

(1902),pp. :39.3-"
(1905),
pp. 283-287.]

xxvi.

; V.

Stokes.

and K^(z).
of Iy,{z)
Asymptoticexpansions

7 '23.
The

"

series of papers.
It is a phenomenon
functions,and it is characteristic of integral

Xl. (1871),pp. 412"425;


(1864),pp. 106"128;
[Math, and Phys. Papers,IV. (1904),pp. 77"109

The

the constants

X.

397.

assignedto

1 ) TT

discontinuity
of the

discovered

was

apparentlyone of those
Gabriel Stokes,Memoir
papers

e2^(''+^)-^,
[(2^
-

by

fact that the constants

Jy (z)are

of the

discussed

was

is not

that the values to be

or
integer,
positive
negative.

phenomenon
and

The

Ci, c^

" 7'2 (5) combined

formula

shews
K^ (z)and H^^^'"
(iz),

at

once

with

equation" 3"7 (8),which

connects

that

^^'W~y ''"!"(2ir

(1)

(")'

1 H^

"

'

2!(8^)-

1!82

CO

Cf.

Bromwich,

t Stokes
the latter

Theory of InfiniteSeries,"

illustrated the

being

those

change

associated

with

with

133.

the aid of Bessel

functions

Airy'sintegral("6'4).

whose

orders

are

0 and

"

204

THEORY

be

quoted here.

functions of

The

of the third kind which

are

order had

zero

[CHAP. VII

EUNCTIONS

to deal with

it convenient

Russell*; he found

BESSEL

OF

been examined

the

previously
by

logarithms!of the functions

involved,and his formulae

be written

may

as

follows:

iber(^)^expg(^)cos
^^

\hei(z)V(27r^)sin

'^ ^'

|ker(z)^expa(-^)cos
^"'^

^'

^{-Iz/Tr)sin^^

[kei(^)

where
13

25

_z

"^^^^~V2"^8^V2
384^^/2
^^^

The

Iarg

\" |7r in

These

the

of

case

results have

(1910),p.

been

the formulae

Hadamard's

result

the

of

case

(1) and

expressedin

modified

form

by Savidge,Phil. Mag. (6) xix.

expansions.
modification
of the asrjmptotic

which

so
expansions,

that

is of

the reader should

no

importance is due to
modifythe various asymptotic
luith a negligible
convergent series together
mainder
re-

that it is possible
to

formulae
have

theoretical

considerable

theybecome

The

term.

will-be stated for real values of the

in making
difficulty

the modifications

but
variables,

appropriate

complex variables.
We

u,

" ^ir in
|arg^-j

are

(2).

Hadamard:[:;he has shewn

to

3842V2

162'^

8^v'2

51.

7*25.
A

of
validity

of

ranges

V2

""'

128^*

take firstthe

of

case

when
/"(";)

"

\.

"

When

U^)

ilxy
ede
ii^V^n
f'^^^''""sini) (i)Jo

rf
1 (i^
+

(- '"'^" ""^'
nSm) ir'^

"

2(2xye''

"

i^-v)
^

the last result


We

may

being valid

write this

equationin

e^
_

because

[^w2.+2m
exp (_ 2u-x)du,

where

y denotes

Phil.

is convergent.
integrals

the form

r{r

l, 2x)
l).m':{2xym

"incompleteGamma-function"

of

'

Legendre".

Mag. (6) xvii.

t Cf. the similar


Bull,

the

the series of

I (^-v\,^.y(v+

V(27r")^=o

'"
"^'~'-

Jo

"""

replacesin |^ by

have

we

we

(1909),pp. 531, 537.


procedure due to Meissel, which

de la Soc. Math,

de

France,

xxxvi.

will be

(1908),
pp.

77"85.

explained

in

" 8-11.

" Cf, Modern

Analysis, " lG-2.

7-25, 7-3]

e~-^-)which
0{x''^"-'^h
In the

is

of the

case

EXPANSIONS

205

of x, the difference between

largevalues

For

is

ASYMPTOTIC

o(l) for each integralvalue of

n.

we
ordinaryBessel functions,

take

the

expressionfor

the function of the third kind


/

so

\-S plix-iv:r-j;ntfee

'i),\v-h

that
fl

(2)
and

!."(,)
=

(A)',

-^-^';'"'_7"''+'"+.%
0(..-.-n

,-,.-i..-i"

similarly
^ir-xj

(y + I)

! ("

2ta;)^

these results it is easy to derive convergent series for the functions


of the firstand second kinds.
From

for functions
gave the formulae
of any order exceeding | is obvious.

Hadamard
functions

of order

zero

only ;

but

the extension

to

"

7'3.

Formulae

for

in the

the remainders

asymptoticexpansions.

in
shewed
which
that the remainders
investigation
order of
the asymptoticexpansionsof H^,^^^
{z) are of the same
{z) and i/^"-'
magnitude as the first terms neglected.In the case of functions of the first
In

and

" 7'2

second

theorem

to

we

an

gave

kinds, it is easy

to

the effect that when

obtain
v

more

is real*

and

exact
x

is

and

rather

remarkable

positivethe remainders

asymptoticexpansionsof J"^ {x) and Y"t,(x) are


condite
remore
and, by a slightly
numericallyless than the first terms neglected,
("7"32),it can be proved that the remainders are of the
investigation
neglected.
sapie signas the first terms

after

Let

certain stage in the

us

write

We

"-'!(i-v"r+(i-i^.n'^-^'"-)

may

take

v^O

without

losing generality.

206

[CHAP. VIT

FUNCTIONS

BESSEL

OF

THEORY

that

SO

(1) /i,

{x)={-^) [cos(oc+ IvTT Itt)P

(x,v)

Sin

{x^-lvir-Itt)Q {x,v)\

cos

{x + lvir- {it)Q (x,v)].

yrrx/

/ 2 \*

(2) F"^ {sc)

"

l{v)

the variables

real,and

are

that, if p

It follows

[-2x)
and,

V^2x)
1 ;

It follows

'

be

taken

on

\.2ixJ

|, there

\2ixJ

{2p)l

in this formula, we

real,
^"^' since 6o is obviously
"

exists

1 ^

'

have*

\2x)

{2p)\
^o^

'

1.

that
integration

^^Zo

(2p)! (2xyp r(p

(2"0 ! i2xy^^

doe-" zf+2p-* du

'

^)Jo

'

the

if+'^P-^
du

T{v

2p^ I),

in the expansionof P (x, v) does not


after p terms
in absolute value, providedthat 2p" v
^.

(p

l)th term

"

the formula

From

(l +"'Y""=

2x)

V
find in

I e-"

Jo

that the remainder

see

exceed

of

"

value, such that

,Zo (2m)! \2ixJ

\h

we

^tt since

since

and

we

large that 2p^v

so

ml

2xJ

take h to be

may

1-

results combined

|^o1^

where

^2^^

"r=o

adding the

on

so

exceedingunityin absolute

6, not

number

of " 7"2,we
and, in the analysis

yrrx

Now

+
[sin{x^-lvir-^tt)P (^,i^)

similar

{x,v) does not

(i

"r=o

V)m

(" MV"

"^

\2ixJ

ml

e.Jl- v),^+,/+ U\'P+^


{2p + l)l \2ixJ

that the remainder

manner

exceed

the

{p +

after p terms
in absolute

l)th term

'

in the

expansion
value,providedthat

2p-^v-%.
These

results

were

reproduced by Gray
p. 70, but small
Camb.

This

(1859),pp.

xvii.

was

189"196.

Math.

Ann.

in their Treatise

have

been

pointed out

i.
on

(1869), pp. 491


Bessel Functions

in both

"

494, and

were

(London, 1895),
investigations
by Orr, Travis.

(1899),pp. 172"180.

of if^,
{x) we

case

result

Mathews

inaccuracies

Phil. Sac.

In the

given by Hankel,
and

obtained

in

have

the formula

rather

different

mariner

by Lipschitz,Journal

filrMath.

lvi.

7-31]

ASYMPTOTIC

EXPANSIONS

207

and

and, when

j)'^v

where

^i "$ 1, and

"

\, the

"

last term

be written

may

integration,

so, on

m)
v^^jv,
?H

where

0 "$ ^o "^ 1 when

This is

by

the

methods

; the

the fact that

after the

7'31.
The

(1 +

of

manner

(1 +

" 7'3

proved not only that


with Jo (x),Yf,
(x)and

"

(2^)pJ
'

|.
those

obtained

for P

{x,v) and

the greater exactness

why
and does not
^ut/xy~P~iis positive
+ (1
liut/xy-^'i
liut/xy~P-i.
reason

Q (x,v)

is secured
oscillate in

is,of

sign

The 7'esearckes
results of

;^

result than

exact

more

same

course,

jj ^

(2^-)'"

{v,p)
'

on
of Stieltjes
Jo{x),Y^Xx) and Kq(x).

put into

who
preciseform by Stieltjes*,
in the asymptoticexpansions
associated
the remainders
less than the first terms neglected,
Kq (x)are numerically
were

but also that the remainders

have the

more

sign as those

same

terms.

also examined
and we
shall not reproduce
Iq (x),but his result is comj'licated
Stieltjes
It is only to be expectedthat /q(."")
is intractable because in the dominant
expansion
the terms
all have the same
sign whereas in the other three asymptotic expansions the
alternate in sign.
terms

itt.

It is evident

from

0)=^r

{x,

Q (x,0)
In these

formulae

" 7 "3

the definitions of

u-i

e--

f%"

w-i

{(1+ |m)-i + (1
{(1+ iuO-i

dv,
|V")-*]

(1

im)-^}du.

replace(1 + ^iu)~^by
2 r^TT

that

d(f)
1 "

^^'

Si^^ ^

'

norm.
(188G),pp. 233"252.
sjyj. (3)in.
by Schafheitliii, Jithre.fhericlit der Deutschen
/^(.r)has also been examined
for tlie
to use
Lagrange's form
129, but he appears
Math.-Vereinigung,xix. (1910), pp. 120
remainder
in Taylor'stheorem
when
it is inapplicable.

Ann.

t The

Sci. de. VEcole

function

"

208
evident

It is then

that

1
i iu)--^
+ (
^
^

1 +
(
^

Iiu)~^
=

'^

-^

TT.

"

['"{1
xu^
_

2
^ttsin-*^

1 +

where

[CHAP. VII

FUNCTIONS

BESSEL

OF

THEORY

(-)^-'(Art^sin" "^)?'-i

sin^ (/"
+ ...+

+ 1 '*'sin" 0)}d(j",
i-y iiu'sin" (/))^V(1

integer(zero included).
positive

jj is any

Now, obviously,

where

1 ; and

0 and

0 lies between

l{(l+ l|u)-4+ (l-lm)-4}

If

Jo

hence

^^^^f^(in)"-...
l-^^(iu)^
+

function
positive

the

multiplyby

we

l+^u'sm-"f)

e~"*

and

m"^

it is evident
integrate,

that
(^i;

^(,a;,u;

'^ ^

2!(8;r)'^

(2p

2)! (8^)^-^
l^3^5^..(4"-l)2

where

"

^i " 1, and

result which

had

we

and
integer(zeroincluded);
positive
proved for P {x,0).

p is any

to be

from
Similarly,

this is the

the formula

find that
l-.3^52

r-

(2) Q(^,0)-

+,^

118^+3,(8^)3
"""

()'

l-.3^5^..(4^-3)(2p-l)!(8^)^P-i
l^3^5^."(4p
,

^'^

(2^ +

l)l)!(8a;)^^+i
+

'

"6.2"1, and p is any positive


integer(zeroincluded); and this is the
result which had to be proved for Q {x,0).
where

In the

case

oi

took
Kq{x),Stieltjes

^'JJ

K, {x)

e-"^

and

replaced
by
(1 + lu)~^
^
^

"i'^

By

and
justexplained,
an

obtain the
a

method

"-^ (1 +

du,
lu)-'-

dd)

; the

Jtfsm-'"|)
givesagain the result

TT.o

method

the formula

1+

procedure then follows the

of

" 7*3.

svicceeded in applying the result of Stieltjes


to
ingeniousdevice,Callandreau*
results
for functions of any integral
oi'der ; but we shall now
corresponding
explain
which is effective in obtainingthe preciseresults for functions of a,ny real order.
*

Bull,

des Sci. Math.

(2)xiv.

(1890), pp. 110"114.

7-32]

ASYMPTOTIC

signsof tJieremainders

The

7*32.

with J" ix)and


It has

in the

209

asijmptotic
expansionsassociated

Y^ {x).

alreadybeen

functions

two

EXPANSIONS

and

P{x,v)

simplertype.

shall

We

that J^ {x)and

seen

Q{x,i')which
extend

now

Y^ {x)are

in terms
of
expressible
asymptoticexpansionsof a
of Stieltjes
("7'31)so as to

have

the result

shew

that for any real value* of the order v, the remainder


the expansion
of P (a;,
v) is ofthe same
signas (inaddition to
less thanf)the

|; a
{p + l)thterm providedthat '2p" v
restrictions
The
which
2])"v f
"

p enable the theorem

on

be stated in the

to

terms

of

beingnumerically
result
corresponding
these conditions lay

"

holds for Q (x,v) Avhen

after p

manner:
following

In the

oscillatory
parts of the series for P {x, v) and Q {x, v),the remainders
less than, the first
terms
sign as, and numerically
of the same
neglected.

are

By

modification
slight

"""^ Kl
^^^l^^
f"^~"'

P G'P"
^)

and, exactly
as in

11(1+ iuo-i

" 7 '3,we

(1

of the formulae

shew

may

reader will

when

2p

"v

see

that

|,the last term

"

du,
ii")"-^!

liuty-^p+i (1
-

the

on

rightis of fixed

dt
^iuty-'p^i]

if

establish the theorem

can

we

+ (1
'"'O"-^'

{^m)l

(-)^(l-^Wi^r-Y^-^
^^,^_, ,^^

The

have

that

m=o

" 7"3,we

(_r-(h ^'Uihuy^^

iiuy-^ Ji

of

we
can
prove that,
and
its
sign
signis that of

{-y.{l-i'),,auypi(2p)i.
It is

sufficient to
clearly

-^1(1

(1
ty^-'I i {(1+ iiutY-'i"+'^
-

2p-v-i^o
is

this

Now
positive.

that

shew

is equaltoij:
expression

I
-

In-

{2p-v-^)T{2p-p-l)]o^

li
typ-'~
^
^

in

" 7-3

we

may

take

{2p

"

ru
+ i)JoJo
"

,^y^
1

As

Jo

r/9
1

I^x-v-"-^
\e-^"+='"'"'
-

f f"(1 ^y'""^""^

dt
1iuty-'P+i]

{2p-v

:r.

i-^O without

I V^-"-^

e-^

loss of

^i'^^2^w^)^-'d\dt
"

ty-"-sin (hXut)dtdX.

.'o

^)Jo

proved in g 7"3.
X Since isin (hXut)\^i\ut, the condition
iofiniteintegral.

[(1

rf\dt
e-^ '^-i'""}

generality,

t This has already been

W.

B. P.

2p"v-h

secures

the

absolute

convergence

of the

^^

210

THEORY

Now

(1

second

OF

mean-value

theorem,

f (1

number

Jo

Since

(2p

I
"

"

2/5

I'

"

into

"

we
^) is positive,

f (1

consideration

'under

succeeded

in

transforming

dt
iiuty-'p+^

the

is

positive.

(1

and
integrandis positive,

so

the

expression

is to say,

That
1

0.

ii {(1+ ^iuty-'p^i (1
typ-"-

that

.'0

in which
infinite integral

an

have

hence, by the

1, exists such

r sin (IXut)dt ^
"-

"

0 and

f,between

0'^~'sin (iXut)dt=

of t; and

decreasingfunction

typ~'is a monotonic

"

[CHAP. YJL

FUNCTIONS

BESSEL

{(1+ liuy-

liuy-i]

p^i(-r

"

(I

v)Uiur-

nZo
0:^0 when

(-y.(i-v\p{h^y^
^

i^my.

{2p)l

alreadybeen seen ("7"3)that in


these circumstances
0 ^ ^ ^ 1 ; and then, on multiplying
|^ {^ 1. Consequently
obtain the property
the last equationby e~^^ it""^and integrating,
at once
Ave

where

stated for P
The

2i

K'

2jw"i/ |. And

(x,v).

corresponding
property
i"'^'-^ (1
-

the details of the


Note.

{l"Uuy-^

it has

"

The
Hes

for

Q (x,v) follows from the equation

i"')'-'i -,!o
=

'"J:t

will easily
be suppliedby
analysis

analysisfails
between

and

the reader.

when

if we
take p
0, but then the phase of
"h"v"^
and
so
"h(v-^)7r,
l{{l+liuy-h + {l-Uuy-h} has the
the same
signas
manner,
^{{l+^iu)''~i{\ -iiu)''~h}/iha,ii
=

signas unity,and, in like


^^^ hence P{x, v) and Q {x, v) have the same
sign as
2 (""""")'''
expansions,so the conclusions are still true ; and the conclusion
i"./"f if^ 0.
same

"

the

first terms

is true

for

in their

Q {.v,
v) when

7*33.

Weber's

formulae for the

remainders

in the

expansionsoffunctions

of the third kind.


Some

which are satisfied by the remainders


in the asymptotic
inequalities
o
f
and
hg^ve
been givenby Weber*.
These inequaliJT^'^i
expansions HJ^''(z)
(^^-^
ties
their importanceto the fact that theyare true whether 2 and v are real
owe
which we
shall give it will be supposedfor
or complex. In the investigations
that
is real,though it will be obvious that modifications of detail
simplicity
to complex values
only are adequateto make the mode of analysis
applicable
;/

Math.

Ann.

xxxvii.

(1890),pp.

404"416.

212
where

functions

of

(z)
(z)^,"''(z)+ B' (z)ZT,'^'

0,

(A'

(z)and B (z)are

^ H/^^ (z)

Y'(z)^ H^^

(z)+ B' (z)

=^

that

chosen

so

[CHAP. VII

FUNCTIONS

BESSEL

OF

THEORY

e'-"-i"
(iTT.)-^

/(-2i^U

i'^"

'

It follows that

and

J'^)

{z) ^7ri^TTzrie^;-J"

A'

|l^

H/^^ (z),

so

(.
{iTT
tt/J

A(z)==A-i,
J. is

where

S;" {z; p)

for
expression

similar

B (z),and

the
By considering
difficult to

Hence

we

behaviour

J'^)

"^^^

and

write Hankel's

may

l-2i(. + 0P

of both

that A=l

see

it follows that

hence

g-'X^-J"
{^'rrz)\
{z)]
[AH,^'^
{z)+ 5^,'-^'

(^J
-h^'rrp.{v,p)l
is not

^"^^'

constant.

obtain

We

^"^ ^'
l\^
{-2i(z

0}-^e^-^^^^-^-i^^

sides of this

equationas

^^

-^

it

oo
,

0.

in the forms

formulae

"^

^"'^ ^^^

"

the remainder

where

^"'""^'""''"1-"'"
(^
(")"

72^*"may

be defined

the

by

(^
i7rp.(.,^.)j^

i?/'

Since

(z)^ 0,

see
we
inequalities,

we

have

that the

'

i')+ ^i'"'!'

Equation
^^-

1_2,"(,_^,)}.

+ i^),and
\z+ t]"^\J{r^
so, by using the crude
modulus
of the last integranddoes not exceed

Hence
+ i2)-i(p+i)
dt,
I$ 2'-p G'p I{v,p) jr (7-'
Ijs^"i'

Jo

when

and

so,

and

similarly

These
the
and

are

p^l,

we

have

the results obtained

analysisno hypothesishas
p;

in this respect Weber's

other writers.

by

been

Weber

made

and

it will be observed

that in

concerningthe relative values of

results differ from

the results obtained

by

7-34]

ASYMPTOTIC

7*34.

Approximationsto

When

the argument

(with the

in it

the

small ; at

remainders

in the

of

time

to

after the smallest

the

argument

ingeniousmethod for meeting these


shall explainthe method
we
by Stieltjesf;
method.

and
Jo ("^^)

to

apply the

We
so

the results which

state

be

term)

is not

particularly
sufficiently
largefor the

slowly.

very

converge

may

numerical

An

and

asymptoticexpansions.

remainder

same

ascendingseries

Kq{x)

213

Bessel function is not very large*,


the asymptotic
well adaptedfor numerical
computationbecause the smallest

expansionis not
term

EXPANSIONS

in detail

devised

appliedto the function


by Stieltjes
by applyingthe

obtained

were

difficultieswas
as

Yf^{x).

transformation

indicated

in

" 7"31

the formula

to

" 6'15 (4),

that
e-^" du

e-^

e-^V2
TT

i'*" e-'^'^dedu
.',)
.'o "*(! + I'usin^^)
r

/"" rjTT p-xu

p-1

hi sin' Oy^dddu

-^(-

IT

IT

X^ r

Now
when

the value of
is

6-="''{\U BIXV d)V


{"^
dd
1

/Now,
e~^

is

as

of the series for K^

increases from

0 to

oo

2) and then decreases


1

where

^j)+

(when

increases from

"

oo

to

to

ue-^"

oo
,

"

The
about

t Ann.

range

of values

4 to about

of

.r

under

norm.

sup.

nearlyequal to

remainder

that

then

and

zero

and,
=

2x

after the

a,

from 0 up to
write

so

we

maximum
"

"

e-'-^'\
for similar reasons,

wc

write

e"'''.

contemplation for
(8)in.

the

so

the

functions

10.

Sci. de VEcole

is

increases
^ue~^"'

sin- 6

from

choose p

(x),we
=

consider

to

numericallyless than unity;

du.

is least
(0,ni)l{2xy"'

where

^-T^

,-,

in order
large; accordingly,

smallest term

dd du

.'oi""(l+i"/sin-^)

for which

^"rnr

v^'{0,ni)
+ {-y R,
L".
(2*')'"
=0

."X

K'^xj

iri

"i

x/T
Jo Jo w^l
+|"sm-^)

(1886),pp. 241"252.

is
^'o(-Oi"'i "'"-o(-''}
.7o("'")"

214

domain

The

(|,77)plane; and it is

of the

the whole

becomes
integration

of

[chap, vn

FUNCTIONS

BESSEL

OF

THEORY

that

found

i.^

TT*

-00

("%s

-00

where
a

by
the

'

It follows* that

arithmetic.

largevalues

for

asymptoticexpansionof R^

^0,2

24'

^,"

tedious

rather

some

^"

"

0,0

x\^e-"-^

of p

"

'

the dominant

terms

of

are

c^o,2+ cr2,o

R^'^2

'p

ITJ

that

so

(2)

"^' +

|_2

TT/

It is easy to

J9

that the

to

stopping at

V)

"

'

of the smallest

one

...

is

terms

roughly half

of the first

omitted.-

term

In like
in

due

error

that

theorem
verifyby Stirling's

'{-)
so

?V

1,
A~2(--V^-'7.+

proved that,if P {x,0) and Q {x,0) are


Stieltjes

manner

defined

as

" 7*3,then

^X"=^9|=;^^(-)'i^;",

i'(^,0)

(3)

-s"-"'"ff:r^'^(-)^ie;

.3(..o)

(4)

""=0

where
'''

R^(P)r^(^\^^

(5)

TT/

^""

"

p is chosen

of this character

similar

formulae

have

und

Phys.(3)xx. (1913),pp.
Reports,1913, 1914.
It would
Bessel

be of

functions

to
functions),

J,
^

L2
_2

"""

vw
Vtt/
so

as

be

to

nearlyequal to

and

x,

is defined

to

p.

Eesults
some

1_2

i^^'"-.
--

providedthat
\iQ.X

240"244

zero

order

Bessel functions

tabulatingBessel

actuallyused

interest

some

of

useful for

are

been

to

(as well
of

(1914),
pp.

(3)xxil.

extend
as

the

functions

for that purpose

in the critical range;


der Math,

by Airey,Arcliiv

30"43

and British Association

has established for


Stieltjes
other
logarithmicintegraland some

results,which

for the

order.
arbiti'ary

Cf. Bromwich, Theory of InfiniteSeries,"" 133, 137,


proved in " 8'3.

and

174,

or

the

lemma

which

will be

7-35, 7-4]
Deductions

7"35.
If

ASYMPTOTIC

^.

which

6 in the formulae

2 tan

(2^)"+* f*-sin"-*

Schafheitlin's

resemble

It is obvious

215

s integrals.
from Schafheitliri

replacea by

we

EXPANSIONS

" 7 "32,we

of

^ sin

deduce

(i/ 1) 6/
-

,^

of "6*12.
integrals

these results that

from

P (.'T,
V) " 0,

Q (*",
i^)"

interesting
consequence

(-

Q ('^,
v) " 0,

An

that

0.

"

,.

"

f)

(h "

i'

"

f)

"

1)

(_

of these results is that

we

can

"

that

prove

Q (x,v)/F (x,v)
is

function
increasing

an

function
For

of

when

"

of

when

h;"v

"

"h:

and

it is

that

decreasing

^.

"

have

we

Q'{x,v)P{x,v)-P'{x,v)Q{x,v)

where
n//i

.N

^^^' ^^
so

(sin^ sin d))"-*


^^^^ ^

(cos^cos.Jj^^^
,

.,

-.^

,,

*^"

'"'

^"

"

ix

*^^ ''''^" ^^'^'

"

"

that
+ P(c^,^)
F(0, cf")

If
add

we

the

has the

'^)
^Pi-(^^"
^^'^"^^^7^
^

^^^

t-^^

in the
interchangethe parametricvariables 0, cj)
results so obtained we see that,when
l"v" %,
the
result.
and
this
same
signas I" v ;
pi'oves

7'4.
In

"

^) (^

double

"^)-

and
integral
integral

the double

of Bessel functions.
expansions
investigation
ofthe asymptotic
Schlaflis
memoir

which

hardlyto

seems

its importance deserves, Schlafli*


elaborate

(2

of
investigation

the

have

received

has

which
recognition
elegantbut somewhat
the

given a very
asymptoticexpansionsof

the fvmctions

of the

third kind.

integralformulae from which he derived these expansionsare


is not so well
of Bessel's integral
generalisations
; althoughBessel's integral
the asymptoticexpansionot
adapted as Poisson's integralfor constructing
The

the

Ann.

di Mat.

(2) vi. (1875),


pp. 1"20.

importance of this memoir

is

The

only standard

recognisedis the treatise

by

work
Graf

on

and

Bessel
Gubler.

functions

in which

216

Jt,{z)when
in

OF

THEORY

largeand v is fixed,yet Schlafli's method not onlysucceeds


in a neat and
but also it expresses the remainders
expansion,

is

[CHAP. VH

FUNCTIONS

BESSEL

the
obtaining

compact form.
Schlafli's procedureconsisted in

^.j

H-"-

the
selecting

and

of

contour

of
takingintegrals
i7-e-

in
integration
-

He

and

being 0

(I) Let

tt

took

two

contours, the

and

it is

supposedthat

firsttake the

us

p is

where

way

is

tt

it,the phcifie*
of

that,on

l/w)
for the

constants

and
positive

contours
respective

is real.

write

pe*",

l +

6 is real,and

and
positive

du,

such

2 +

be

phase to
u

type

|" (u ^)
J

exp

ire'"(?/
is constant.

the

then

+ pe'^)
re^'p^e-'^/il

is

is

and
negative,
Hence

its

consequently
equalto

conjugatecomplex.

have

we

sin (a + 2^)

..,,

^^

Next

choose

sin
a

sin ^
.".
'

(a + ^)

sin

parametricvariable "^such

new

(a + ^)
that

"/) 2^
=

and

a-7r,

then

!("-(/))
i (a + 0)

cos

(2)

cos

Now,
from

^ varies from

as

the

originat

at
infinity

angle(tt

an

"

angle

an

"

"

(tt

"

"

ot)to (tt

a) with

(tt
"

laid

|(a

cos

a),u

"

a) with the

If this restriction is not


than

-rsin-"^

(m-1)2
'

traces

the

"

out

0) cos |(a + 0)
contour

real axis and


positive

real axis,providedthat
positive

on

the

contour

passes

to

emerging
passingto
0 "

"

27r.

more
infinity

once.

We

shall

for
specified

value

a,

as

laythis

now

formula

restriction

on

and

" 6*22 (9),providedthat

then
we

the contour

give to

and

is of the type
arg

the

same

is permissible.

It follows that
e""'

/_^ (re'*)
-

e'^^i

I^(re^")
=

'2isin

where
*

The

"method

Chapter

vir

is defined in terms
reader
of

viii.

will find it

of

2^IZ'j'""
(" ")1
1^''''"
% ^'^''P

which

"

"^by equation(2).

to
interesting

steepest descents"

compare
is

the

applied to

general methods
obtain

various

of this section

with

the

asymptotic expansions in

7-4]

the

ASYMPTOTIC

and so, replacing


to replacing
u by 1/u,
Changing the signof "^is equivalent
of
the thii'dkind,we have
the left by its value as a function
on
expression

11 /-e'""
L
r 'or"
+ u") exp

) X
(.S)ei""'i/,w (j-e'("-^^-)
=

(2) it follows

From
as

0 to

ir

"

a:

and, if we

?-e'"'(" lf/u

when
positive

is

the contour,

on

we

write

have
dt

dt

du

0 to +

t,

^'

"

from
varies monotonically
"/)

that t is

^^

d(f".
^]\^^^

from
"re"^{i( 1)-/hincreases steadily*

that

so

217

EXPANSIONS

"

'

H
the range

(a

re'"

of values of arg

+ u.-i)
^/{rt)
Tfi() e-H'^-")^("i

being less

than

tt.

Next, by Cauchy'stheorem,

it is convenient
outside

to

take the

the contour

because

point^=1 inside the contour, but ^


the originis a branch-point.

must

be

It follows that
iii+ u-i i

'" +" !/"+'

^--Ht-1)(^^

_
_

Hence
i/"+, i+)

r
r ("+.
e-J'"^'exp(re''^)
"

rr

/rx

^^

m
"

Now

",

it is evident

2" is any
to

On

making
1

that
fP
(-)y l;p

^mpn
(^_y,i

that p exceeds

both

R(v

in the last

this substitution

"

(wjththe

will be convenient

h) and R("v

r (i;-I-m
"^^"-^^

notation

of

(9

'^^~r(v-m+

i).

4-

A)

I (v,m)

i,).(2vil) (2m)!

that

deduce

" 7'2),we

"

quently
subse-

integrandand observingthat

1/M.+,1+)
/"("+,

"2^j

^""H?- l)"^^c?^

+ ^^/(7-e'-)
(^- 1)-^

positive
integer(zero included). It

suppose

g-i^r*

27r'-'?;r^e*- JoJ

p-i

where

"x

^^

r j'-i

\i

("V"

(i'w)

where

-^^*'
"

J
2"W(27r),'o
t?

SincG

sin0(l+2cosacos0

sin-"*
^^

"

d(j)cos

cos

(T^n:p^(7^e^'M(^^^"^+T^"^}

(p

"

(cos a

+
~"

"

cos

cos-"/))

(/")-

"'

218

OF

THEORY

[chap. VII

FUNCTIONS

BESSEL

First consider

p is

When

to

contour

so

largethat

be

as

circles tend
If

the two

on

both

R(v

"

^) and R{"v

take the
we
-|),
and
small
large

"

Fig.15; and when the radii of the


the integrals
respectively
along them

in

shewn
oo

it exceeds

and

tend

to

zero.

write

we

now

to

^"^p-id^

/"("+,1/U+, 1+)

rays

all that survives of the

(whichare

(-)Pcos

contour),we

TT

.-kP-"-*
(1

vir

find that

Jo

tx

xy-^'''^d.v

{I

x)/{re^'')

Fig.15.
Now

the numerator

modulus

of the

of the denominator

values of

it is

never

integrandis positive(when
is

less than

never

less than

|sin

1 when

is

^tt"

"

and the
real),
|7r; for other

!.

Therefore
^ft

d'

do \cos

vir

^
l^7rrV(2
1"

e-ittP-i xP-''-i(1

dxdt
a:)P+''-i

^TJ")j 0 J 0

where

j^o 1is 1 or |cosec


is complex,it is easy to

a
see

\accordingas

V^

is negativeor

that
cosi^TT

(7)

cos

i^

6o\{R(v),p)j

(vtt)1
~

cos

'

{27')P

6^I{v,
p) I^ {2r)P,
When
positive.

220

THEORY

and

[CHAP. VII

FUNCTIONS

BESSEL

OF

hence, if now
t

re^''(u"lf/u,

find that

we

expresseda second solution of Bessel's equationin a


consequently
be deduced ; and the analysis
form from which its asymptoticexpansioncan
(z),the final result being that,when
proceedsas in the case of 77^'^*
We

have

-f7r"arg^"-|7r,

^.'M.)=Q^--"--f:!:[i^;^^"i

(13)

as
l{z)^0 or l{z)^0,
|02.\does not exceed 1 or |secarg^ ]accoi'ding
providedthat v is real and l) + l"\v\; and R (d^)" 0 when I(z)^0. If is
complex the form of the remainder has to be modified,justas in the case of (8).

where

functions

of

"

as

was

that

in this

it is unnecessary

v,

exceed

must

in (3) and (11) are


that, since the integrands

be observed

It should

to suppose
investigation
based on
of
investigations
integrals

in

necessary

even

R {v)

Poisson's type.

investigation*
of asymptotic
expansionsof Bessel functions.

7"5. Barnes

of functions.of
asymptoticexpansions

The

which

formulae

Barnes'

from

obtained

were

r{~s)V{-2v-s)V{v

the third kind follow


in

6'51.
""6*5,

Let

immediately
consider

us

i){-^izyds

"cci"v"p
'

T{~s

(-2iz)-''-i'
J

If

"

r{-s
"

cc

Di-2izyds.

have

v+2))r(-s-v+p)

r(s-p

i)(-2izycis

T'

v+p)r(-s-v+p)r{s-p

iarg (" iz)\^%Tr 8, we

\r(-s

v+p)r(-s-v+p)r(s-p

i)e^''-^y''^ds

is convergent
integral

the last

and

"

and

so

the first integral


of all is

{(-2iz)-'-P\
.

But, by the arguments of


of the residues at the poleson
TT^H^^^^{z)l{f^^~'"'^
cos
VTT

"

"

^,

"

"

%,

"

...

is
" 6"51,the first integral
the

(22^)"]
plus
V"

p +

(-)"'r (1/+

Tram.

Camb.

\.

Phil.

"

"

^Tvi timcs

The

residue at

^) r (-

Soc.

xx.

the

1/

i^

it is

so

the

sum

equalto

of the residues at

sum

"

times

'Im

rightof the contour, and


"

h)

(1908), pp. 273"279.

??i"

is

7-5, 7-51]
and

ASYMPTOTIC

jarg (" t" |^ f-n- S

when

so,

221

EXPANSIONS

"

(iy,...,-.p,;i.g",o(..-.)this is

and

//^'-'
(2')
may
The

reader should

obtain

to
possible

been

^ by
by replacing

of
" 7'2. The investigation
throughout.

notice that,althoughthe determination

of the remainders

magnitude
of the

be constructed

in

result obtained

the

equivalentto

is

of the 07'der

of

transparently
simple,it is not

formulae, concerningthe magnitude and the sign

concrete

remainders, which

by

this method

ultimately
suppliedby

are

the methods

which

have

considered.
previously

7*51

ofproductsof Bessel functions.


Asymptoticexpansions

asymptoticexpansionsof the four


J"y(z) in which the coefficients have simple forms, even
productsJ"fj,{z)
for this is that the productsi/^"'
when
//^"^(2^)and
The reason
(2^)
fx
for
which
no
H^^i (z)H^'"'(z) have asymptoticexpansions
simple expression
in the tv/o expansions
exists for the generalterm ; the leading terms
are
It does

not

possibleto

seem

obtain

v.

2e"2*'~THA*+.'+i)-'
^

2fx-+ 2p'

TTZ

4^12

however, do possess simple


(2)^^""(^),
{z)and//^"-'
ir^"^'
products
{z)H^^-^
deduce asymptoticexpansions
can
asymptoticexpansions
; and from them we
The

for

J^{z)J^{z)+Y^{z)Y,{z)
and

for

J^ (z) F, (z)
-

V^ (z)./,{z).

the expansionsis by Barnes' method,


simplestway of constructing
in
A
consideration of series of the type obtained
just explainedin " 7'5.
The

" 5"41 suggests that


27n

examine

should

the

integral

/";r(2.+i)r(^"i;-..)r(t^-.)r(''7

the contour
contour

the

we

and

contour

is to be chosen
the
;

polesof

and

so

that the

polesof

the other four Gamma

(2s + 1) lie on

the left of the

functions lie on

the

rightof

not
are
temporarilysupposed that /u,, and fi "
is convergent
integrandhas no double poles.Tlie integral

it is

so that the
integers,
providedthat Iarg (?2)I" |7r.

i^

222

THEORY

First

evaluate

is
m

equal
+

to

minus

^{/x+ v)

lie to the

the

of

sum

round

the

right of

the

residues

at

the

VII

enclose

the

to

contour

; the

originalcontour
poles,and

these

[CHAP.

the

expression
residue

at

is

v)'7r'm\ r{/x

sin /ATT. sin v7r.sm(fi +

It follows

FUNCTIONS

BESSEL

integralby swinging

poles which

of

sequences

the

OF

')n

l)r{v

l)r i/x+

m+

l)

that

27rt

sin fXTT sin

vir

sin (i;

sin

yti)7r

"

sin

(,a+ v)7r

(/a i/)tt
"

sm(/ji+ v)'Tr

{J,(z)J,{z)+Y,(z)YAz)\
sin

+ v)7r
{/J,

^3g-j(M+w"|2cosit7rcosi/7r+tsm//i+ i')7r}
^S

^^^-

^"7

"

\r
t^
/
\
^
1 Az)-ru.
(z)
^

\)

(z)J^ {z)\

TT^

[[/,(^)j,(^)+F^(^)r,(^)}
2sin|(/x + v)Tr
cot

i (/. I/)TT {/^ {_z)F, (^)


-

F^ {z)J, (^)}]

-2cos|C+^)J{^-^-"-^-(-"^^-(-"^'^+

By writing
arg iz I and

"

tan

(^) F, {z)
i (/. ^) TT {./,
-

for i

|arg (" iz)\are

cot

the

throughout
less than

H/*

") '^

Y, (z)/, {z)]].
deduce

analysis we

fvr,i.e. if jarg^: j"

"

!-^(Z) I^.(^)

tt,

1
,.

that, if both

then

(2)^. (^)l]

7-51]

ASYMPTOTIC

EXPANSIONS

223

and

results hold

^) TT

[J,(z) r. (z)

F" (2)/. (z)]]

.) (-^^ .)
("-^^

These

J (^

tan

.in

..

(i.r

ds.

for all values

of fi and v (whether integers


or
not)
providedthat,in the case of the former u+ v and fj, v are not even integers,
and. in the case
of the latter /jl+ v and fu, v are not odd integers.
"

"

We

obtain

now

(1) and (2) after


We

firsttake

the

the

be

p to

J -xi-p

-xi

"

either

(when

is equal to 27ri
integrandis inserted)
the polesbetween
the contours.
Since

residues at

the left of

on

tegrands
integerso largethat the onlypolesof the inof
and
then
F
R{s) "2)-j are poles
(2s+ 1) ;

an

the left of the line

on

the functions

asymptoticexpansionsof
of " 7'5,
manner

times

the

of the

sum

r~'~'f(s)(izrds=^0{z-^P-i),
J

deduce

we

that the

txi-

"

when
asymptoticexpansions,

(3) [J^(z)J, (z)+ Y^ (z)n (2)]

cot

11

Sm

^) 'r

z\

"

ir, are

[J,(z)Y. iz) /, (z)Y, (z)]


-

V
TT

"

"^

H/^

[arg

TT^

r(-'y
r(^V",i)r5g+,"+i)
,^.(-rT(^%."+i)

+,.+i

iTZ-

I,
i-'^;
-\)
..F.(^'+i.
^"+1,
'-/+i.
.".^;7"
^
^
2
2
^^-y
V 2
I (yu, i^)
sm

"

TT

and
.

(4) [./^
{z)J, iz)+ Y^ {z)F, (z)]+
2
TT^COS

//i+
//A

^{iJb v)7r
"

^'^

IZ
1/

l
1

tan

\{y,-v)'rr.
{z)F, {z) J, {z)Y,{z)]
\_J^

^l"V
/A

?-'

"

l
1

V
^'

"

\
1

ix+l1
/i+

"

^
/i.
"

1
"\

v
i^
.

.
_

'

'

'

'2'

W*

224

THEORY

the

In

special

in

Formula

(5)
(6)

formulae

is,
pp.

and

der

(4)

VII

to

{1.3.5...(2m-l)l^\

(-r{1^3^5"""(2.^-l)}-

by

stated

Walter

to

by

(2.)-

of

Camb.

by

A.

p.
Phil.

repeated

(Leipzig,
Gregory

Lorenz,

(1898),

I.

P)-oc.

Orr,

Cylinderfunktionen

attributed

however,

discovered

scientijiqties,

transformations

on

der

Theorie

Lodge,

while

Soc.

integrals
1904),

x.

the

(1900),

247
"

British

Selskahs

general

more

p.

by

given

was

245

pp.

Vidensk.

Danske

K.

435],

99.

Nielsen,

the

Association

proof
Hand-

expansion

Report.,

(5)
1906,

494"498.

is

not

method

Jy'

{z)
an

to

easy

after

such

[Oeuvres

v^ere

depends

which

It

of

reduces

formula

(2m)!

been

have

to

seems

(1890).

vi.

(3)

(5)

last

JHz)+YHz)^-''~
Jo(^)+io(^)-^^^-^

Sh-ifter,

huch

the

v,

[CHAP.

FUNCTIONS

particular,

m
(b)

of

jx

JH^)+Y^^{z)^^

(5)

and,

when

case

BESSEL

or

is

used.

Yy-

estimate.

number

any
An

{z)

estimate

will

exactly
of

alternative
be

given

in

terms

the
these

method
in

magnitude

" 13"75,

of

or

asymptotic
obtaining
and

it

will

the
then

the

sign

of

the

mainder
re-

expansions

when

asymptotic

expansion

be

possible

this

to

form

CHAPTER

A'lII

FUNCTIONS

BESSEL

OF

LARGE

ORDER

functionsof large order.

8"1. Bessel

of descriptive
subjectof this chapteris the investigation
properties,
equalitie
includingapproximate formulae, complete asymptoticexpansions,and inThe

of various

functionsconcerned

are

results

Carlini's

formula

with

with

Bessel

functions ; and

primaryimportancewhen
many

of the

results

the

the orders

happen

to

perties
pro-

ofthe

be

true

orders.
positive

shall first obtain

associated

of

large,though

are

of all

for functions
We

types connected

will be examined

which

which

are

of

("1'4). Next,

purely formal
we

shall

character,

obtain

certain

approximate formulae with the aid of Kelvin's* "principleof stationary


the contour
shall examine
discovered by
we
integrals
phase." And finally,
to obtain asymptoticexpansions
when
Debyet ; these will be employed firstly
the variables concerned
of
to obtain numerous
are
real,secondly,
inequalities
to obtain asymptoticexpansions
varyingdegreesof importance,and thirdly,
both the order and the argument are complex.
of Bessel functions in which
in which
and sc are positive,
it is
v
dealingwith the function /^ (sc),
consideration
have to be divided into three
found that the problems under
classes,accordingas x/v is less than, nearlyequal to, or greater than unity.
in the corresponding
Similar sub-divisions also have to be made
theorems
with complex variables.
concerned
In

The
and

trivial

problem

is fixed,may

of

determining the asymptotic expansion of Jy (z),when

he noticed

It is evident, by

to the

this
l/v^(27r);

tV|
=

large

logi^z)-{v

that
expansion of " .3-1,

^)logv}.
V

where

is

here.

theorem
applyingStirling's

Jy (z) exp [v+

result

has

been

pointed out

by Horn,

Math.

Ann.

Lll.

(1899),

p. 359.
for functions of largeorder,
of approximate formulae
physicalapplications
writers may be consulted:
Macdonald, Proc. Royal Soc. lxxi.
(1903),pp. 251
theJi^Howing
A
258 ; Lxxii.
; Phil. Trans, of the Royal Soc. cxx.
(1904),pp. 59"68
; xc. A (1914),pp. .50"61
und
der
Ann.
xxx.
Chcmie,
(1909),
57"136;
Physik
(4)
Debye,
11.3"144;
pp.
(1910),pp.
Ann. der
Rybczyiiski,
(1912),pp. 29"50;
March, Ann. der Physikund Chemie^ (4) xxxvii.

[Note.

For

"

Phil.

Mcifj.(5)xxiii.

In connexion

with

the

(1887),pp.
principle,see

Math.

W.

B.

Ann.
F.

lxvii.

Stokes,

(1883),p. 341.]
(1909),pp. 535"558;

[Math, and Phys. Papers,


t

252"255.

[Math, and Phijs.Papers, iv. (1910),


pp. 303"306.]
Trans.

Camb.

Phil. Soc.

ix.

(185G), p. 175 footnote.

ii.

Milnchener

xl.
[5], (1910).
Sitzungsberichte,

15

226
und

Physik

pp. 516"537

Love, Phil. Trans,

of

the

Royal

Soc.

waves

Mag. (6) xix. (1910),


131 ; Watson,
(1915),pp. 105

ccxv.

"

The

546"563.

Royal Soc. xcv. A (1918), pp. 83"99,


the problem of the propagation of electric
largelyconcerned with attempts to reconcile
Proc.

8'11.

Nicholson, FMl.

(1913),pp. 191"208;

Chemie, (4) XLi.

[CHAP. VIII

FUNCTIONS

BESSEL

OF

THEORY

the

over

theoretical

works

sui'face of the

with

all deal

quoted

with

earth, and

are

experimentalresults.]

firstextension of Carlini's formula.

MeisseVs

approximation ("1-4) obtained by Carlini is the first term of the


asymptoticexpansionof a Bessel function of largeorder ; subsequentterms
in the expansionwere
formallycalculated by Meissel, Astr. Nach. cxxix.
manner
:
284, in the following
(1892),col. 281
The

"

(1)

function

{z)by

if we

define

then

equation(1) transforms

of

now

we

equation

+ ZU
{z)-v-'il- z")
[u{z)Y'\
=

that, for largevalues of

assume

iio, u^, u.2, ih,

(Z)

VUq +

on

functions

denote

of

which

...

independent of

are

128(1-2^)-

16^ + 368^3 + 924^5 + 374^' +


"5.=

it is
integration,

found

'

(cf." 1'4)that

jlog^'^^
^,^

^ ^

V(l

(16

1 5122-

"'"5T60l"{
322-^ +

l}i log(1

^^)

42^ + z'

((1-^2)^ j

2 + 32^
24i;

-U-

ISz^

32(1-2-)'

8(1-^^)^'

64^ + 560^3 + 456^5+2,52'

+ ^^

8(1-^-^)^

(z)dz

series

...,

W2=^

"

2(1-^^)'

4^+10^^

j\

2882"

v,

by

the coefficients of the various


powers

zero

z-)]/z,Ui=ir7z.

Uo=Wil-

on

(z)is expansiblein

find that

the left,we

Hence,

+
Un/v+ u-s/v^

Ml +

in (2) and equatingto


substituting
of

v,

0.

of v, thus

descendingpowers

where

the

i);

into

z''{u'{z)+

(2)
If

v^{l-z^)JAvz)

,-^^-_+._^_
a

be written

equationmay

It is clear that Bessel's

2322"

-z'-f
lQv''{l
-

36542-*

375^"
_

(1-^^)^
+

132"

z-^)

228

THEOKY

OF

BESSEL

where

(1) P.

sec"'/3)
^^(4sec^y8
+

(32 sec^ yS+

-^"^

g"'

288

(768 sec^yg+

/3 + 232

sec^

sec*/3+

41280

(2)

Q,

/3-)-13

sec"

/3)

sec"y9+ 17493 sec"/3

14884

^8+

103

375 sec"

yS)

sec^o

sec^^ /3)

'

"

"

"

3sec=/3)
z;(tan^-/3)-^||^(2
+

57601/^

(16

o^^-Va

322od0j^*

1512

sec-

3654

/3

(256 +X8720 sec-y3+

/3

sec*

1891200

1914210

sec"/S

sec^/3+ 4744640
sec^yS+ 67599

sec^"

/3)

....

J^ {y sec /8)in

To determine

^TT,and compare
type givenin " 7*21
to

the results
;

of these

terms
so

obtained

that, as /3

see

we

P.
we

sec"

4242

+
'

and

[CHAP. VIII

FUNCTIONS

with

take yS to tend
we
expansions,
the expansionsof Hankel's

\'k,

"

", Q.~z^(sec/S-^7r),

"

infer that

(3)

ir,w iysec /3)

(4)

^,*="(i/
sec
/3)

e-^^+'^Q.'-^'"',
a/(^-^)
a/(^-^^)
"

e-^--^"Q-+^'''^

"

It follows that

(5)

J, (v sec y8)

(6)

F. (^'^^^^
P^ and

where

("8'41)that
and

Y^ (vsec
Formulae

which

easilybe
values

rigorousmethods

the

reader

(Q,

'"'''''^^

^^"' ^'^)'

"

ivr),

valid for small values of yS,i.e.asymptoticexpansions

are

equal,cannot

The

e-P"'cos

by (1) and (2). It will appear subsequently


these formulae are actually
asymptoticexpansionsof t/^(y sec /3)
/3)when v is largeand ^ is any assignedacute angle.
Y^ (z)which

that, for such

defined

Q^,are

of Jt,(z) and

y/'(?^^)
v/(^^)

are

obtained

of the

from

will observe

valid when

by

this method

that the
8*11.

are

both

largeand

but it will be

variables,approximationscan

Schlafli's extension

correspondinganalysisof "

and

of Bessel's

approximation has

been

be

nearly

are

in

" 8'2

obtained

by

seen

integral.

carried

two

stages further

than

in

8-2]

FUNCTIONS

The

Note.

dominant

OF

in the

terms

LARGE

ORDER

229

expansions(5)and (6),which

be written

may

in the

form

J^{x)

(7)

M^cos{Q^-^7r),

}\ix)

M,sm{Q,-^7r),

1'

3f"

where

Qv

"

the

publicationof

Physical Optics, K. Danske


i. (1898),
[OeuvresScientifiques,
pp. 421

Videnskahernes

had

obtained

sin

arc
^'^) ^vir-\-v
'J{-'^'^

been

memoir

two

before

years

on

The

Lorenz

procedure of
has been

which

proved

in

[_

^.^L
then
^

to

the exact

use

whence

the

from

the Wronskian

Subsequent researches on
of the Royal Soc. ccx.
(1907),pp. 697"707
; (6)xix.
(1911),pp. 67"80;

closelyconnected

1906, pp.

8*2,
and

of the result

consequence

.r*

formula

follows without

(1910),pp.

131

been

that

prove

difficulty.

Lorenz

by

144, and

"

(1910),pp. 228"249;

(8),has

" 3'63 (1),to

{^-l}^

the lines laid down


A

of

are

publishedby

A.

due

to

Macdonald, Phil.

Nicholson,Phil.

Mag. (6) xiv.


Math. Soc. (2)
result concerningQ^ i-Qv, which
Lodge, British Association Report,

516"537;

(2) xi. (1913),pp. 104"126.

with

in

trxAI^^'

x-y.-\

approximation .stated for Qv

Trans,

is

L. Lorenz

'

easilydeduced

by

(6) vi. (1890).


Skrifter,

equation

Q.

IX.

"^'-

dx
is

granted that,as

2.4

.'-

'

which

paper

SelsJcahs

" 7 '51,

iro:

and

Meissel's

436.]

"

take for

to

was

{vjx),

Proc. London

494"498.

principleof stationary
i^hase.Bessel functionsof equal order

The

argument.

of stationary
principle
phase was formallyenunciated by Kelvin* in
of the principle
connexion with a problemof Hydrodynamics,
though the essence
is to be found in some
much
earlier work by Stokes f on Airy's
("6'4)
integi-al
and Parse val's integral
and
also
in
a
posthumous paper by Riemann^.
("2-2),
The

The problem
integral

which

propounded

Kelvin

?/

-"/
"In

which
time

to find

[m [x

an

approximate expressionfor

corresponding
Phil.

The
wave-length 27r/??i.

Mag. (5)xsiii.

t Cavib. Phil.

Trans,

Werke

Ges. Math.

the

dm,
tf{m)}]

disturbance
at
expresses the effect at placeand time {x,t) of an impulsive
waves
(0,0),when f{m) is the velocityof propagation of two-dimensional
to

cos

was

principleof interference

set

forth

placeand
in

by

water

Stokes

Papers, iv. (1910),


pp. 303"306.]
(1887),
pp. "252" 255. [Math, and Phijs.
pp. 341, 351.]
Papers,ii. (1883),
(1856),pp. 175, 183. [Math, and Phijs.
(Leipzig,1876),pp. 400"406.
ix.

230

BESSEL

OF

THEORY

[CHAP. VIII

FUNCTIONS

and wave-velocity
of group-velocity
suggested to Kelvin
Rayleigh in their treatment
outside
the
that,for largevalues of .v tf{m),the parts of the integral
-a, n + a)
range (/x
of interference if /a is a value (or the value)of m
account
of values of m are
on
negligible
and

"

which

makes

"^[m{x-tf(,m)}]
=

(/^-a, /x + a), the expression7n{x"tf{m)] is then

In the range
three terms

and

0.

of its

replacedby

the

first

expansionby Taylor'stheorem, namely

that, if*

it is found

v/[-Mm/"(m) + 2/'(m)]'

/"
then

c/o+ 0-2}
{",i2y'(/x)

cos

7rv^f-2r;Mr(M)
+ 2/'(M)}]

M'

C0S,V/'(M) + i7r}
+ 2/'(;.)}]"
v/[-27^^[M/"(M)

replacedby

the

case

to

and

oo

o-,

which

are

largeeven

though

be

small,have

been

oo
.

that in
foregoinganalysisthat Kelvin's principleis,effectively,
the
the
is
function,
importantpart of
integral
of the integralof a rapidlyoscillating
xohich the phase of the trigonometrical
that part of the range of integration
near

It will be

due

limits for

integralthe

In the last

from

seen

functioninvolved

is

the

stationary^.

subsequentlybeen noticed J that it is possibleto give a formal


for a largeclass of oscillating
mathematical
functions,
proofof Kelvin's principle,
formula due to Dirichlet.
" of an integral
by usingBromwich's generalisation
It has

The

of the
Let
let ybe

j,M

of Bromwich's

form

to
principle

Bessel functions

be

follows

the

tions
applica-

f t''-'sin t .dt=

r a;''-^F(x)smvx.dx^F(+0)

The

as

adequatefor

"

fji"

method

1, the sines may


which

has

sin ^/jltt;
(+ 0)T (fi)

Jo

Jo

and, if0

is

will be

when x^O;
functionof x which has limited total fluctuation
Then,if"l"fj,"l,
functionof v which is such that vy^- cc as v-^ oc

F{x)
a

which

theorem

replacedby cosines throughout.

he

justbeen explainedwill

now

be used

to obtain

an

for a
at m=fjL;
m{.x-tf{m)) has a minimum
the sign of the expression under the radical is changed.
maximum
in the writingsof Cauchy. See
search reveals traces of the use
of the principle
t A persistent
16 of his Theorie
de la propagation des Ondes, crowned
Sept. 1815,
e.g. equation (119) in note
21/"?m.presentes par divers savants, i. (1827).[Oeiivres,
(1)i. (1882),p. 230.]
*

This

t Proc.

is the

Camb.

" Bromwich,

appropriatesubstitution

Phil.

Soc.

xix.

when

(1918),pp. 49"55.

Theory of InfiniteSeries,"

174.

8-2]

FUNCTIONS

which

This formula,
largeand positive.

by Cauchy*, is

discovered

was

is

231

ORDER

LARGE

Jv{v) when

for

approximateformula

OF

r(i)

JA^)

(1)

^2

^1

2^
.

3^ TTV^

of stationary
of the principle
paratively
phase,comby means
investigated
by Nicholson, Phil. Mag. (6) xvi. (1909),pp. 276"277, and Rayleigh,
recently,
Papers,v. (1912),pp. 617"620]; see
Phil. Mag. (6)xx. (1910),
[Scientific
pp. 1001"1004
42"48.
Soc.
Camb.
Phil.
xix.
Proc.
also Watson,
(1918),pp.
This

From

and

has been

formula

that

" 6"2 (4) it is evident

obviously
sm

vir

77

^^
g_^,j+sinhi!)
-rr

Jo

Hence

J^ U)
Now

let

(f) 6
=

I"cos[v(6

,.(0

(l/v).

Imi

But

then

sin 0, and

"

6)}cie +

sin

"

6l-*.0

COS

"

6"

in the interval (0,tt),


cos
6) has limited total fluctuation
hence, if (f)'^/{l

and

"

itfollows
from Bromzvich's theorem

cosi/d)

"

^"
0

that

6^

cos

"

then

(1) follows

It stillhas
this result

we

to

be

at

,_a

,,

""

v(pd(p

cos

once.

proved

observe

'

I 9
-^'
0

r^x r(i)cos^7r,

and

r*

,,

-a9'^"

"^^/(lcos 6) has

that

total fluctuation

limited

; to

establish

that
d

(p-ismeg{d)

^s

tl-cos^J 3(l-cos^)'_

d~6

";(^)=illl^^^3(^-sin^),

whelT
so

5f(0) 0, ^(7r-0)=+oo,

that

^'(^) (1
=

therefore, by integration,
g{d)^0

and

monotonic

'

and

it is

Comptes Eendus,

Cauchy's methods

obviouslybounded.
xxxviii.

will be

(1854),p.

given in " 8"21.

cos

+ cos ^) ^ 0,
^)2/(l

when
The

993.

O^e^n.
result

Consequently 0*/(l-cos^) is

rpquiredis therefore proved.

[Oeuvrcs, (1) xii.

(1900),p. 663.]

proof by

232

THEORY

By

and

it may

tedious

some

also be

obtain

proved that

JJ(v)=^^^o(v-i);

(3)

associated formula

an

[CHAP. VHI

FUNCTIONS

to
integrations
by parts*,it is possible
approximation,
namely

of

means

second

BESSEL

OF

is

r.w~-M).

(4)

terms,

give the dominant


machinery
powerfulanalytical

these

which

asymptoticexpansions,of
with the
will be investigated

The

aid of

more

results

in ^ 8-42.

Meissel's third

8'21.

expansion.

to
integral
just discussed has been used by Cauchyf and Meissel:]:
is a largeinteger.
obtain the formal asymptoticexpansionof Jnin) when
w
obtained by Cauchy and (in
be explainedhow this expansionwas
It will now
of the promore
a
complete form) by Meissel; the theoretical justification
cesses
in " 8 '42.
employed will be investigated

The

Taking the

formula
1

Jn {n)

let

write 6

us

sin 6

"

f'^

{?i
(6

cos

"

sin

"

small values of t,
^t^; it then follows that,for sufficiently

/9_/_l_l/3i
u
i -f "^"qI
"

-r

_V-x

/5i
-r

Y4^^t'

"

Xo

1,

^\, Xa

1213.
7 20 7 200000'

"

*-5

Xi

~"

tioo

^^

'

/2m+l

asio^' ^^

A"jt,

""

and

^)|cW,

tt^tswuT)'

"""*

It follows that

Jn(n)

-r\

'^
.

When

is

^nt^is largeat
large,
J,,(7i)
~

(2m

l)X",tAcos(Ut')^dd.

[ m=0

limit,and Meissel inferred that

the upper

(2"i+

1 ) X",

SeeProc.

Camb.

Phil.

t Comptes Rendus,

161"164,

pp.
+

Astr.

formula

Soc.

xxxviii.

(?
.

'7r,H=o

Cit"

r"

cos

(^n f) dt,

.0

(1918),pp. 42" i8.


(1854),pp. 990"993,
xix.

1104"1107.

[Oeuvres, (1) xii.

(1900),

167"170.]

Nach.

(1),Meissel

cxxvii.

(1891),col.

stated "Sclion

vor

359"

362;

cxxviii.

dreissigJahren

war

(1891),col. 145"154.
zu
folgendenFormel

ich

Concerning

gelangt."

8-21, 8-22]

OF

FUNCTIONS

LARGE

233

ORDER

a
("6'4); and hence, by
integral"
"generalised
signindicating
term-by-termand usingEuler's formula, Meissel deduced that
integrating

G is the

where

:i: X.,r("m

/.(n)--

(1)

also gave

Meissel

exhibits

approximationfor A,",valid

an

of the

divergentcharacter

the

-)

f)

cos(im-hi)7r.

when

is

large;and

this

approximation is obtainable by the theory developed in the memoir


des fonctions de tr5s grands nombres," Joiirnal de Math.
I'approximation

of

The
"Sur

tion
approxima-

expansion (1).
Darboux,

(3)iv. (1878),

5"56, 377"416.

pp.

AVe consider
be

the

monogenic) are

and

near*

of 6
singularities

the

which

pointsat

the
"(127r)^

t=

fim

function

of t ; the

2/-7r and

where
(12r7r)5,

/=

"1,

"2,

"3,

...

are

}^
t
(367r)3il
+

(127r)-^i

approximation to X,h

theory of Darboux, an
jrithe expansionsof the two

fails to

singularities
(where 6

expansion of

in the

terms

the

6^

dominant

"27r

By

qua

is the

the last formula

comprisedin

functions

of the

sum

coefi"cients of
;

that is to say

that

(367r)*

'^

-^

(2/n-i)
^

2
1
*.*."...

\,"

"'

(2m

1)! (12,7)*'"

2r(2?n,-l-")
sfm'
+ 2).(127r
3^r(t)r(2?rt
and

formula,
by Stirling's

so,

("^)

"

in

2j"

(18)^r(f)(m+l)Ml^'r)^"'
is Meissel's

This

loc.

by Cauchy,

an

approximation of

the

same

character

was

obtained

cit.,
p. 1106.

of Kelvins
cqyplication

The

8"22.

approximation;

to J^,(vsec^).
'principle

of stationary
phase has been appliedby Rayleighfto obtain
principle
acute
angle,
an
approximateformula for J^ {v sec /3)where /3is a fixed positive
and V is large
I.
The

in

As

" 8*2

have

we

J,(vsecl3)

fcos[v(6

/^

TJ"

and

(a minimum)
/3sin 6 is stationary

sec

"

Write

"

from
*

These

are

t Phil. Mag.

sec

0 to
the

13sin 6

Royal

lxxi.

"

tan

which
singularities

(G)

xx.

(1910), p. 1004.
Phil.

Trans,

(1903),pp. 251"258;

/3sin 6)}cW+

Oil Iv),

when

j3.

that ^ decreases
so
/3-t-"/",

/3 and then increases

J See also Macdouald,


Soc.

sec

as

to zero

6 increases from /3 to

as

6 increases

tt.

nearest
to the origin,
Papers, v. (1912),p. 620.]
[Scientific
of the RoijalSoc. ccx. A (1910),pp. 131"144;
lxsii.
(1904),pp. 59"68.
are

and

Proc.

234

THEORY

[CHAP. VIII

FUNCTIONS

BESSEL

OF

Now

{v(6

cos

"

/3sin 6)]dd

sec

/"T-^ + tan^"

dd

^^

tan

"

/3)}
j-. dcfi.

JO

_^tan^-/3

and

(10

{i^
("^+ ^

COS

as

"^V(2tany8)
rf"/)

/S.

"

in the range 0^0


has limited total fluctuation
Hence, t/ (^*(dd/dcf))
BroniwicK

from
follows

[v(0
pcos

that

theorem

/3sin 6)]dd

sec

+ /3
[v("/"

cos

vi/ tan

and

f^7r,it

/3)}//gl

tan

[v(tan^

cos

/3

"

-^^^

/3)

tt],

so

{v(tan /3

^)
V(i.^tan^)

cos
^
/.(.sec^)^

(1)
The

^tt}
'

formula

F.(.secffl~^'"'-'";,f^r^^:'"
v(2-i''''tan;S)

(2)
"^

"'\

is derived
The

in

similar

reader

r-/

from

manner

will observe

" 6"21 (1).

that these

the dominant

are

in Meissel's

terms

expansions" 8'12 (5),(6).


complete the rigorousproof of

To
limited

these formulae

have

we

to shew

that

has
(f)^
{dd/d(j")

total fluctuation.
the square

Now

of this

is equalto
function,namely 0 {d6ld(f))^,
^

/3sin ^

sec

3 +

tan

^
_

(1
say.

sec

But
cos

^ (1

cosec

sec

cos

The

and

so

k{6)

decreases
from

B increases

when

/3cos

/3cosee

6)"^-2 ((9-sec/Ssin6
^ (1 sec (3cos 6)^

/3+ tan ^

numerator, k (d),of this fraction has the differential coefficient


"

as

(9)2

/3cos

0 ^ ^ ^

and

cos

/3cos

steadilyas

^tt to

6 cosec^ ^ (1

B increases

since

k{B) 0
h' {B) changes sign once
tt

from
when

/3cos

sec

to

6)^,

^tt,and

^"^7r,

then

increases

it follows

that

(from negative to positive)in

steadily
h'{B)^0

the

range

0t$"9^7r.
Hence
|Jh (B)|is monotonic
(and decreasing)when 0 ^ ^ ^ ^, and it has one stationary
and continuous
{6)\ is bounded
point (a minimum) in the range ^ ". B "
; since !s,fh
0 ^ B ^ ir, as had to be
when 0 "$ 0 ^TT it consequentlyhas limited total fluctuation when
proved.
tt

236

[CHAP. VIII

FUNCTIONS

BESSEL

OF

THEORY

integranddoes not
oscillate rapidlyon
it ; and so we
expect that an approximate value of
may
of the integrandin the
from
the integral
will be determined
a consideration
neighbourhoodof the pass : from the physicalpointof view, we have evaded
the interference effects (cf.
" 8'2)which occur with any other type of contour.
The

contour

The

mode

has

the

8'6,8"61

that

so

the

will be
asymptoticexpansionsfrom the integral
specialfunctions which will be studied in ""8'4 8 '43,
of

of derivation

from
clearly

seen

selected*

been

now

"

but it is convenient

enunciate

to

this stage

at

in provingthat the expansionswhich


subsequently
asymptoticin the sense of Poiucare.

be useful
are

Lemma.

when
F{t) be analytic

Let

\t\r^a

8, luhere

"

will

f which

lemma

will be obtained

0, B "0

let

and

00

IT I^ a, r beingpositive;also, let \F (t)\" Ke^'',luhere K


^ a.
positivenumbers independentof t, ivhen r is positiveand
asymjjtotic
expansion
when

e-"-" F

(t)dr

is valid in the

1arg
where

is

!^ ^TT A,
-

number.
arbitrary
positive

an

It is evident
such

z/

the

largeand
\v\issufficiently

when

of Poincare

sense

Then

are

(m/r)v-'^"'

:i:a,nT

'^

Jo

and

that, if M

be

fixed

any

a
integer,

K^

constant

can

be found

that
M-l

whenever

^ 0 whether

^ a

or

+ i2j/,
e-''^a"iT""*/^'-^dT

e-^'F{T)dT= 2
^0

therefore

and

J 0

ni=l

e^^dr
\Rm\^\ \e-"'' \ K^r^^^'^^-^

where

JO

K,T{Mfr)/{R(v)-b]^/^

providedthat
valid

remains

compared

when

even

with

is the

{v)" b, which
have

We

v.

6 is

function

therefore

Jo

For

the lemma
an

Burkhardt's

account

j)i

London

such

cosec

that

A.

R{v)

"

The

analysis

b is not

small

proved that

is established.
of researches

article in the

t Cf. Proc.

of

\v\"b

% a^T (m/r)v-"^''+ 0 {y-^^l"-),

e-''F{r)dT=
so

when

31-1

/""

and

case

in which

the

contour

der

Math.

Wiss.

Encyclopddie

Math.

Sac.

(2)xvii.

(1918),p.

is the
ii.

133.

real axis

(1916).

see

pp. 1343

"

1350

of

8-31]

FUNCTIONS

construction

The

8'31.

It has been
with

shall

we

method

whether

steepest descents
with

In accordance

On

contours.

examine

now^

of

explainedin " 8*3, we

function

the

once

sinh

have

seen

these three

(I)

When

that

cosh

find

can

completesolution

It will be sufficient to confine

iv

it

discussed

that

0,

are

descents,as

viv,

"

equation

v={)\

"

three

distinct

to

cases

consider,

We
equal to 1, respectively.

or

sider
con-

iv,where

u,

that

2?i7rt.

attention

our

such

(1) is
+

sinh

lu

to the
"

stationary
pointsf" a; at
the equation
so

is zero, and

vw

is
w

viv)

"

0.

real,and this equationbecomes

are

cosh
so

and

z/sechct,

/ {x sinh
W^rite

of steepest

number
positive

imaginarypart of

to be

of the contour

of

these pointsthe

functions associated

in turn.

cases

xjv" 1, w^e

the

real.

are

stationary
pointsof

shall have

we

a;

then

the method

to solve the

(2)
and

less than, greater than,

xjv is

the variables

positive,
of the contours
appropriatefor the
in "|6'2,6"21.
types investigated

any

in which

237

hypothesisthat

the

first to find the

have

of ^v, i.e.we

it is at

when

of
principles

(1)
and

contours*

of the

are

qua

ORDER

""6-2,6'21 that the various types of


of the form
represented
by integrals

be

along suitable

taken

of Debyes

LARGE

in

seen

J^ {x) can

OF

sin

cosh

"

0,

or

vcosha
,

coshw

"

-.

(S)

siU'y

givesa divergentintegral.We therefore consider the


contour
spond
given by equation(3). To values of v between 0 and tt, correpairsof values of it which are equal but oppositein sign; and as v
value of u steadily
increases from 0 to tt, the positive
increases from ot to + x
The

/-'

paper,

The

contour

contours

Math.

+ The

parallelto

Ann.

investigatedin this section


i.xvii.

(1909),pp. 535

"

are

those

which

were

discussed

558, except that their orientation

of taking stationarypoints other


imaginary axis.

effect

the

than

"

would

be to

is

in

Debye's earlier

different;cf. "6-21.

translate

the

contour

238

THEORY

BESSEL

OF

[chap. VIII

FUNCTIONS

The

equationis unaltered by changing the sign of


symmetricalwith regard to the axes ; the shape of
between
in Fig.16.
and v
v
tt is shewn
=

tt

"

and

the

so

contour

the part of the

is

contour

Fig. 16.

If

sinh

cosh

"

(sinhw

"

that t (which is real on


it is easy to verify
increases in the directions indicated by the
As
+

to 0 and

00

along the

travels

from

contour

increases to +

then

shewn

curves

in the

figure)

arrows.

iri to

"

oc

since,by

; and

the

cosh a),

"

tti,

decreases

from

" 6"2 (3),

27rtJoo-,ri
have

we

obtained

J^(cc)when

cc

and

from

contours

solution

phenomenon, on

to

"

a:jv " 1, we

find

can

we

take the

acute
positive

(sinhtu

that,replacingiv by
(5)

cosh

10

"

cos

cos

"

yS

of the

second
Stokes'

that

of the

e"|uation

0,

i/3.
which

contour

/3) sin /3

sin /3+

angle/3 such

roots

now

are

"

(y

"

sin?;

we

/3cos /3,

iv,the equationof the


li

concerning

discussion

/9,

sec

pointi/3,the
stationary
/

detailed

give information concerninga


problem is complicatedby
stationary
pointson the contour.

tv""

are

so

The

information

this

which
points,
stationary
cosh

When

1.

derive

""8'4,8*5.

a;=v

the relevant

can

iri

of the two

account

(4)
and

we

xlv"

equation;but

of Bessel's

(II) When

largeand

are

which

from

curve

givensubsequentlyin

integralwill be
The

contour

/3)cos ^

is

obtain

is

8-31]

OF

FUNCTIONS

of

for values

Now,

0 and

between

sin ^ + (v
has

minimum

one

values of

(v

0 and

between

of

for values

of

infinite when

are

The
half of

is 0

t;

or

"

0 and

00

is

; for

zero

other

these

v.

equation(5) givestwo
coincide only when
v

y8cos /3) (sinhw


"

"

iri,T decreases

tt,

sin

givenby equation(5) is as

(sin/3

it is easy to verifythat
directions indicated by the
to

real values
=

/3. They

tt.

shape of the curves


Fig.17 ; and if
T

sin

the value of the function

and
(equalbut oppositein sign),

"

(i' /3)cos l3 "

between

/3

tt,

sin /3+

Hence,

the function

/3)cos

"

/3)at which

tt,

239

ORDER

LARGE

(which is

arrows.

from

oo

As
to

real
w

on

"

cos

in the

then

upper

13),

the curves) increases

travels

and

shewn

alongthe

contour

increases to +

oo

in the

from
and

"

so

we

Fig. 17.

have

obtained

curve

when
ZT;,'^'
(.*")

will be
of the" integral

from
x

and

which
v

are

["6"21 (4)]we
largeand x/v "

can

1.

derive information
The

cerning
con-

detailed discussion

given in "" 8'41, 15'8.

stationary
point"i/3, we should have obtained the
to
in the lower half of Fig.17, and the curve
oc
shewn
curves
going from
discussed
00
TTi givesan
integralassociated with HJ-^ (w); this also will be
obtained
form a fundamental
in " 8"41.
The
two
now
system of
integrals
that there is a marked
distinction between
solutions of Bessel's equation,
so
the case x/v " 1 and the case
xjv " 1.
If

we

had

taken

the

"

"

240

THEORY

(III)
from

The

(I)or
are

in which

case

from

OF

(II)by takinga

as

limitingcase

The

/3 equalto 0.

cosh
shewn

they are

be derived

may

or

[chap. VIII

FUNCTIONS

now

curves

either

to be

sidered
con-

0 and

"

(6)
and

BESSEL

in

v/sin
v,

Fig.18.

Fig.18.

from

curves

the

from

to

"

which

curve

"

given in ""8-42,8-53,8-54.

will be

Geometrical

8*32.

the
concerningB.y^{y)and B.}'^'^{v)
by considering
information
while
is
obtained
oc
+ mi,
concerningJ^,(v)
from
vrz'.The detailed investigato
x
Trt
x
+
tion
passes

information

obtain

We

of Dehye'scontours.
p-operties

Bessel

interestingi-esult which will be found


functions (5^
15'8),and which is also used

which

will be stated in " 8'43,is associated

An

"

to

of
important in dealing with zeros
proving certain approximate formulae
with the second of the three contours
cussed
just disis
that
the
the
branch
question
slope*of
from

in

exceed J3.

in

question,

It is evident

the

that,for

curve

du
sinh u-^r
.

sin (v

^)

"

^^

"

(v

sm^

has

sin(i; ^)secv-(i;-j3)cos/3
"

8) coti

"

V cos

'.

"

do

But

in

" 8-31). The theorem


and does not
iri is positive
of

(Fig.17

be

to

..

positivederivative

the

cos^tan^v, and

hence

it

follows that
sin (w

"

has the

same

for the

curve

Proc.

"

{v ^) cos
"

V cos

signt as v ^. Therefore since v ^ and


consideration,
dvjdu is positive.
"

"

/3
are

both

positiveor

both

negative

under

Camh.

Phil.

(3=0, the slope is 0

on

Soc.

xix.

(1918),p. 105.
originand

the left of the

better results of this

type

is obvious

from

+ This

^)

exist.
a

figui-e.

Since, in the limitingcase


is

^/3 immediately on

the

(Fig. 18) in which


right of the origin,no

8-32,8-4]
Again, to

OF

FUNCTIONS

dvjdu does

that

prove

exceed

not

241

ORDER

LARGE

write

y/S,we

'

sin

that

and then it is sufficient to prove

+ l ^0.
3^/.'2(^")-^^2("i;)

the

Now

the left (whiclivanishes

expressionon

J^

''^

[(t' ,3){sin2(. +

a\
(v-B)cos,B-\
^
^'

when
positive

has the

has

This

0"v

~z

and

it

since

so,

Ijetween

value
positive

of

of r, there

the values of Wj and

0 and

^ 8'31

in

signan

same

tt

and

Next

0, it is
v

i^

(3,it follows that

it is

not

neo-ative.

obtain

now

the

asymptotic

the argument is less than

" 8'31 (I);


are

to
and it is clear that,corresponding

will be called u\ and

values of w, which

two

w.,

imaginarypart,and

supposedthat

sech

/0"o)"0.

"

-7-7
dr j

^t,

Ami.

lxvu.

(1909),pp.

a,

in

ascendingpowers of r.
in
it follows that the expansionof
this expansion,
a.)-;by reverting
{w
and

Wi

tu.^ in

"

",=o

asymptotic expansions contained

B. F.

^^

(dr

J0

expansionsof

drjdtvvanish when lu
a term
a beginswith
lu
of the form
expansions

by Debye, Math.

g-*'

^^
'Itti

and

"i=oW
The

a.

discuss the

we

powers of
obtain
we

w.

0-

therefore

differ onlyin the signof their

./,{v sech a)

when

then have

We

SintJe

ixjsitive

is

shall

we

l{w,)"0,

where

beinglargeand positive.

retain the notation

it will be

{v /3)].

/3,and consequently

the results obtained

the order,both

Wo;

sin

"

expansionof the function of the firstkind in which

any

cos

expansion*of J^ {v sech a).


asymptotic

The

From

We

the result stated.

proves

8"4.

5-^.,,

(v)has the
Therefore,since y\r'

l3 for its only minimum

sin ^

the derivate

+ 3cos2i;

(sin-i'-(-3cos^y)-

"tt.

signas

same

"

^) has

r--

sm^v

-^~,

i7sin/3-3cos vsin(i;-/a)
-'
i

4sin*";cos^

"

-x-

positive denvate

the

has

,,

,,

v]cos /3+ sin^

cos-

sin2
ly
But

when

535"

in this section and

m+l
in Si

8'41, 8-42

were

established

55B.

16

242

BESSEL

OF

THEORY

and, by Lagrange'stheorem, these


values

of

IT

valid for

expansionsare

small
sufficiently

Moreover

I.

dr

mwA
r(0+'0+V^wA

1
a,n

i{m+l)
['d^Jr

27ri
("+)

dw

=JLf
27nJ
The

[chap. VIII

FUNCTIONS

circuit in the

r-planeis necessary in order to disposeof the


circuit round oc in the w-planecorresponds
fractional powers of
; and a single
circuit
double
the last contour
to a
round
the originin the r-plane. From
integralit follows that a^n is the coefficient of l/{w a) in the expansionof
^-iiin+i)jjj ascendingpowers of w
thus enabled
to calculate the
are
a; we
double

"

"

coefficients a,,,,.
Write

IF and

0.=

"

Co

"

| sinh

a, Ci

and

so

(cosh W

"

| cosh

1)

"

cosh

"

coefficients in this

(sinhTT

Tf)

"

c,3F^+...),
a,

c^

^^ sinh

"

expansionof {co+

coefficient of Tf'" in the


The

have

TFHco + CiTf

where

sinh

"

we

Ci Tf +

expansionwill

Therefore

a,
c,

W^

be called

a", is the

("+i'.
...)-*

ao(m), ai(?)i),
azivi),

...,

have

we

ao(m)

Co-^("'+^),
+ 1
Gi]
"271! -^o;
m

771

1) (m

'co

2.1!
m

(m

C.2

S)

Cf

2^2!
+ l)(m
(?)i

Cs

3)

2ciC2
"

2^2!

Co'

l)(m

(m+

(1)

+ 5) d^
3)(/?i
3{

2^3!

a,{m)-c,

2.1!

-Co

V Co'

2^2!

2^3!

(m

CoV
Co

Co-'

1) (?H + 3) (m

5) (m

2*. 4!

On

substitution

we

f aQ=a"^

(0)

I Oi

I tto

"!

(1)

do

(2)

(2)
\a,

":;(3)

a,

I tta

(4)

"

"

"

find that
("

sinh

a)~*,

(" I sinh ot)~*{Icoth a},


(" 1 sinh a)~^{^ -j^coth^ a},
-

("

coth
I sinh a)~^ {y^^

"

^j coth* a},

(- i sinh a)-t{y^g /Jj.coth^a


-

'

^%%\ coth^a},

7) Cj^
'c*

244

THEORY

8'41.

[CHAP. VIII

FUNCTIONS

BESSEL

OF

asymptotic
expansionsof J" {vsec /8)and Y^ {v sec ^).

The

asymptoticexpansionof a Bessel function in


less than the order,both being large; we shall now
which the argument was
obtain the asymptoticexpansionsof a fundamental
system of solutions of
Bessel's equationwhen the argument is greater than the order,both beinglarge.
In

" 8"4

obtained

we

the

retain the notation

We

positivevalue
passes from

"

of

there

t,

to

oo

of

it is clear that,corresponding
to any
of

values

two

are

these

+7^^;

go

" 8"31 (II);

the

lying on

will be called w^ and

values

which

contour

it will

and

w^,

supposedthat

be

R
We

e-*"

-.

except that

r^\dr,
dr

[ar

The

analysisnow
proceedsexactlyon
is replacedthroughoutby i/3,
and the Bessel

v8ec^.

0.

\~

J i)

TTi

(wo)"

have

then

HJ^^^(v sec yQ)


Avhere

(wj)" 0,

the

"8*4

lines of

function

is of the

third kind.
It is thus

found

that

{div.
^

Q-Xr

determine

To
arg

the

[{iL\i^)lr]

--"-

"

\dr

phase

5 TT

as

dr]
of a^, that

0, and

^^

a"

af"^^

"t=o

is of ("

we
|tsin/3)~*,

observe

that

so

ei'VV(ism/3).

Consequently
(taii/3-^)-i"
g..i
tan
^'i^vir

In like manner,
in the real axis of

(2)*

If

(4)

is

we

(^)(. sec

B)

have

(1) and (2),we

Aj
|)
{\) {\vitan /8)+

^m

"

'

precedingcontour

that

valid when

are

r(m

the reflexion of the

^ri*!-^^^!!^
^

largeand positive,
we

combine

yS)^.^o

by takingas contour
the w-plane,
find
we

In these formulae, which


and

"

ir,""(i/sec/3)^

(1)

to make

r(m

is

|)
fixed

A^
angle

acute
positive

the substitutions

find that

J^ {v sec yS)

-^

2
/
^x
1
Z
cos(i/tany9-^/3-|7r)
.

vir

tan

/3

m=o

(-y"r(2m
r(i)

|)

A
2m

'(iJ/ tan ye)-"'

sin(.tan^-.^-l.)J/^^iM).

8-41, 8-42]

FUNCTIONS

ORDER

245

-^^r
"

7,

dominant

/D
(z/tan /3

sin

"

in these

terms

stationary
phase in " 8*21

(-)'"r(2m
-^

o
V
1
z^/3
1
tt) S

expansionsare

which

formulae

been

have

established

give adequate approximationswhen

to

argument

and

order of the Bessel

however, possible

to

the

use

variables to be

and

complex

are

the

nearlyequal. It is,
pansions
determiningasymptoticext
hat
arise
no
ha]3pens
complications
by
are

complex.

numbers

of

large modulus,

such
that

that
z

\z

"

"

v\ is
in
(z^),

be small.

shall write
i'

and

argument

for

will appear that it is necessary to assume


of low rank in the expansions
order that the terms
may
We

order and

""8'4,8"41 obviouslyfail

concerned

method

same

in

large.It

not

principle

shall discuss the functions

Accordinglywe

where

the

(or /S) is small, that is when

function

in these circumstances,and it

supposingthe

by

Asymptoticexpansionsof Bessel functionswhose


nearlyequal.

The

A^^---.

those obtained

8*42.
are

i)
^

=^--

"

of

LARGE

F"(i'sec/9)~

(5)

The

OF

it is convenient

to

z(l -e),

temporarilythat

suppose

|arg^!"|7r.
We

then

iy"^'*
(z)

(1)

where
and

have

".

is that shewn

the contour

in

Fig.18;

w)

dtu,
zeiv]

this contour

on

sinh

"

is real

negative.
write

We

and

the values of

and

W2,

[z(sinhlu

exp

real
We

(2)

of which

w^

"

sinh w,

to any
corresponding
is a complex number

value of r will be called lUi


positive
with a positive
real part, and w., is

negativenumber.
then

have

H^^''
(z)

e-'"-

"

jexp

(zeiu^)
-~

exp

{zew.^-j^)"dr.

246

THEORY

The
obtain

OF

expansionof t in powers
expansionsof the form

of

{zeiu^)
^

exp

BESSEL

beginswith

T-s

ttr

don

valid when

To determine

7H

we

h,nt^"*,

small.
[t jis sufficiently

the coefficients b^awe


+ )
/"(0+,0+,0

observe

that
^^

^^yjS

^^P^"^^^^"-(l^)r
^

Gtti

^""P ^'''''^

67ri]

hence

'^

(IT

are

in lu^,
and

term

2 ei('"+i'-'6,"TH

these

,"=o

T-3
exp(^6m,)5^'

and

[chap. VIII

FUNCTIONS

(w

(m+i)

"'"+^)
iv)i

sinh

"

of ^ 8'4, a singlecircuit in the r-plane


analogousinvestigation
circuit is necessary to disposeof the fractional
and the triple
is inadequate,
the originin the r-planecorresponds
of t; a triplecircuit round
powers
circuit in the ?y-plane.
to a single
As

in the

It follows that bniis


in the

equalto

igJc^+D^i

multiplied
by

the coefficient of w'"

expansionof
exp

The coefficientsin this

"

expansionwill be

w)/^'}"*^"*^^).
called

6o(//i),
6i(m),b.^im),
...

that

It is easy to shew

fbo{m)

lu
(zeiu){(sinh

6^^'^+'\

~2~

J'

60

(3)

o.K^u)

For

brevitywe

e^z^

(vi+ l)ez

60

124

120

write

b,,{m) m^+'^Bm{"z\
=

.50400

J"

so

that

8-42]
that*

so

By {ez)

ez,

B,{ez)

h^'z'-j'-Eez,

B, {ez) j^^z^
=

[Bs{0)

We

+ "|toe^"
ij^e'z^

Bs{0)
-^^Sjjj,

"

jjiisooo' -'"io(0)"

"

6 5

5-ioOOOO-J

have

then

diUi

ex^(zewyp
dr

00

e^^+'^-^6''"'+'^
B,,,{ez)Ti''\

lT-i

"i

dw"

00

-p
exp (zew.)

e"'"+i)-''
6i('"+^"
5"

T-' 2

(zeiv)(dw/dT)]satisfies the
[exTp

and

+ ^^,
^B,{ez) i-,^'z^-i-^e-^z'

(4)

conditions

(e^)r^'",

of the lemma

of

i"

^,w (^)

(5)
and

00

o^

"8'3.

of 5 8'3 that

the lemma

It follows from

,^1'^

e3"-+"'^^B,, (ez)
smh{m

e-3(-+i)-5,,(6^)sini(^"i-Hl)7r.-

l)7r.

similarly

H^^H^)^-^

(6)

deduce

We

at

that

once

ram

'^

(V)

J.{z)

(8)

i^(^)~-o;l^ (-r^5,"Msin4(m

From
1)1

247

ORDER

LARGE

OF

FUNCTIONS

is

the

1)

l)7r.-4^^^

Cauchy-Meisselformula " 8-21 (2),it is to

be inferred that,when

large,
^^"'^^^

^^^

(_)m (-|)s

""

but there

to

seems

dominant

The

be

terms

no

very

in

(7) were

'

r(").(7/i+ i)Hl27r)3'"

simpleapproximateformula
by Meissel, in

obtained

for

Kiel

5"i {ez).

Programmi,

1892;

obtained

which seem
by
similar results,
" 8-43,
some
terms
in (8) as well as in (7) were
Koppe in a Berlin Programml, 1899. The dominant
also investigated
by Nicholson, Phil. Mag. (6) xvi. (1908),pp. 271"279, shortlybefore the
ap^aranceof Debye's memoir.
to resemble

and

The

values

the

same

the

as

integration is
The

values

of

(0), L'lo(0) were


Bo(0), /"'2
...

apart from

col. 3.59"362;

of the contours

use

analysisgiven

to evade

of

the

in this

section.

those

stated

in

were

(1891),
given by Meissel, Astr. Nach. cxxvii.
Meissel's analysis (of.
" 8-21)is substautially
The

object

of

using the methods

.Bo{ez),
B-,[ez]and

(1916),p. 524.
t See the

See

Jahrhuch

the Jahrhuch

of

contour

produced by using generalisedintegrals.


B^ (ez)will be found in a paper by Airey,Phil. M(h;. (0)xxxi.

the difficulties

ilber die Fortschritte

der Math.

iiber die Fortxcliritte der

Math.

1892, pp. 476"478.


1899, pp. 420,

421.

248

OF

THEORY

We
has

consider the

next

far been

so

BESSEL

extent

imposed on

which

to

formulae

[CHAP. VHI

FUNCTIONS

which
jarg2^|"^7r,

condition

the

(5) (8),is removable.


"

The

of the integrandin (2),qua function of t, are the values


singularities
of T for which Wi (or w^) fails to be a monogenic function of t, so that the
the values of
to those values of w for which
are
singularities
corresponding
t

dr/dw
They

therefore the

are

points
T

where

all

assumes

0.

2u7ri,

values.
integral

It is

to swing the contour


permissible
consequently
through any angle t)
less than
and
then
a
or
we
right angle (either positively
negatively),
obtain the analyticcontinuation
of ^^'^'(z) or
the range
H^^'-^
(z) over
^TT 77 " arg^: " ^TT 7]. By giving77 suitable values, we thus find that
the expansions
(5) (8) are valid over the extended region
"

"

"

"

TT

"

If

confine

attention

"

"

arg

TT.

to real

that the solution of the problem is


we
see
variables,
quite complete ; we
asymptotic expansionsof J^ (x)valid when x and
are
largeand (i)xjv " 1, (ii)
x/v " 1, (iii)
|^| not largecompared with x^. But there
transitional regions between
and also between
and in these
are
and (iii),
(i)and (iii)
(ii)
transitional regionsxjv is nearly equal to 1 while \x
is
these
transitional
In
v\
large.
regionssimple expansions(involvingelementary functions only in each term) do not exist.
But important approximate formulae
have been discovered
by Nicholson,which involve
Bessel functions of orders " J. Formulae
of this type will now
be investigated.
we

our

have

not

determined

i*

"

formidae valid
Approccimate

8'43.
The

failure of the formulae

Nicholson*

to

second
investigate

manner
following

In the

""8*4

in the transitional

8'42

"

regions.

approximationsto

Bessel's

regionsled
integralin the

of functions

case

of

Jn (^)
and, when

of

in the transitional

COS

{nd

"

n,

sin

6) dO,

it follows from Kelvin's


nearlyequal(bothbeinglarge),
of stationary
principle
phase ("8*2) that the importantpart of the path of
is the part on which 6 is small ; now,
this part of the path,
on
integration
sin 6 is approximately
equalto ^" \6'^.It is inferred that, for the values of
X

and

and

order
integral

under

are

consideration,

Jn{x)'^IT

cos

{nd -xd

IxO')dd

}Q

rQos{ne-xd-^lxe^)dd,

TJ" Jo
TTJO

XXIV.

Phil.

Mag. (6) xix.

(1916),pp.

239"250.

(1910),pp. 247"249;

see

also

Emde,

Archiv

der

Math,

und

Phys. (3)

8-43]
and

FUNCTIONS

the

when

is
expression

last
"

n,

and, when

"

OF

of

one

LARGE

ORDER

249

Airy'sintegrals
("6*4). It

follows

that,

n,

Jn{x)^^\^ ^j{/_i

(2)

J:.i,

where

the arguments

The
Nicholson

of the Bessel functions

the

on

rightare ^ {2{x

correspondingformula for F" {x) when x"


; with the notation employed in this work

)i)|=/a,'

"

also found

was

by

it is

F,.(.)~-j^i^l'{J_,-J,}.
3^

(3)
chief

disadvantageof these formulae is that it seems


impossibleto
of validity
and the order of
determine, by rigorousmethods, their domains
introduced
in using them.
magnitude of the errors
The

With

view

and
integrals,

remedying

to

discovered

this

method

defect, Watson*

which

actuallyit is very laborious),


by means
Nicholson's are obtained togetherwith an
The

method

of

limit for the

upper

iunction whose

Bessel

be written

i^sech a) may

in the

J^ {v sech a)
where

"

sinh

beingchosen

the contour

Debye.'s
theoretically
simple (though
which
formulae
analogousto
errors

involved.

:
employed is the following

for
Debye'sintegral

x{=

examined

is

"

"

its argument

e"^'' div,

exceeds

formf
"

(coshlu
that

so

order

"

is

-iTTt

1)

cosh

"

IX, -TTI

on
positive

(sinhw

"

iv),

it.

expanded in ascendingpowers of w, Carlini's formula is obtained


the lowest,
when
we
approximate by neglectingall powers of lu save
obtained
and when
0, Cauchy's fornnila of "82(1) is similarly
a
|ry"^sinha;
of w save
the lowest, \iv^.
all powers
by neglecting
If

is

"

"

suggest that

considerations

These
first two

terms,

givean

not

The

^w'-sinh

Proc.

Camb.

Phil.

is deducible

shall

we

where

t This

^vf

"

cosh

the

a,

approximationvalid throughoutthe

which
integral

whether

namely
"

nlay

to examine

it is desirable

"

Soe.
from

region.

is therefore
investigate

| IF- sinh
xix.

first transitional

(1918),pp.

"8-31 by making

^ W'^ cosh

a,

96"110.
a

change of originin the "--plane.

250

THEORY

and

in the

the contour

If TF

it.
on
positive
hyperbola

complex variable W is so chosen that


is the right-handbranch of
-{-iV,this contour
C^tanha

and this

that

It therefore has to be shewn


CO

is

the

it^^=iF^

of the third order with

has contact

curve

[chap, vin

FUNCTIONS

the

planeof

BESSEL

OF

Debye'scontour

at

the

origin.

approximationto

an

/"ocexpCaTrO

+7ri

e-^'Uhv

e-''^dW.

is

differ by
integrals

These

ro

Jo)

for
of an upper bound
problem is reduced to the determination
\[d{tu"W)ldT]\. And it has been proved,by exceedinglyheavy analysis
which will not be reproducedhere,that
and

the

so

{lu

W)

Stt sech a,

"

di
and

so

(div

dW]
dT

{dr

Gtt

i
,

Hence
xexp(i7r2)

dW^-^,
V

where

|^j|"

To

1.

evaluate

the

" 6"4),modify the


W

tanh

"

If we

making angles"\'ttwith

and

tanh

{\vtanh^ a) f
Jo
"

into two

contour

write W""

e^"^exp

right(which is of the type discussed in


lines startingfrom the point at which

the

on
integral

becomes
the respective
on
^q-^I-^^
rays, the integral

tanh- o}d^
{-\v^'-\ v^e^'""'

exp

e~^''exp

the real axis.

(^V

tanh^

a) / exp {"^v^^

I v^e

"

^"^

tanh'^a\d^.

Jo

Expand the integrandsin powers of tanh- a and integrateterm-by-term


procedurewhich is easily
justifiedand we get on reduction
"

"

"77*tanh
and

hence

we

exp

tanh* a) [/_i{^v tanh^ a)


(^i^

obtain the formula


tanh

(4)

sech a)
Jt,{v

7r\/3

exp

[v(tanh a

^ tanh=^a
+

where

[^i |"

I^(^vtanh^*a)],

"

1. This is the

more

form
precise

"

tanh^ a)
"x)}K^(lv

oc
a)},
{i/(tanh
8^ii^~'exp

of Nicholson's

"

approximation(1).

252

expand the integrandsin


found

[CHAP. VIII

EUNCTIONS

BESSEL

OF

THEORY

and
of tan- /3,integrate
term-by-term,

powers

it is

that
eXp jTTt

OO

g-^"" dW
00

"

"

tan /8exp ("


^-rri
X

V3

equatingreal

(5) J^ {v sec yS)


+

and
itan

:3-Han

F^ (i^
sec
13)

(6)

/Q)

vi tan*

/i (^i^
tan* yS)]
[e-i-'
J_j (ii/tan* /S)-h e*'^'

Tre^-^ tan

On

/ tail ;3

"

-* tan

tan

^^^ ^_

^^^^ ^, ^^,^,
^^^ ^,^^3^^_
^^^.

imaginaryparts,it is at

found

once

that

/3cos [v(tan/S

^tan*/3 /3)}[/-j-f /j]

/3sin {i/
(tan/3

yQsin [v(tanyS

tan*
1

/3cos {1/
(tan/3

tan*

tan*

yS)|[/-j

/3

J{\+ 24^^/^',

/3)}[./_-H J^.]
/S /9)}[/_
J^]+ 24^3/1^,
/8

rightis
tan* ^; and |0.^ and j^gl are both less than 1. These are the more
precise
1^1/
forms of Nicholson's formulae (2) and (3); and theygive effective approximations
the right.
terms
on
except near the zeros of the dominant
where

the

It is

argument

of each

of the

functions

J"i

upper limits obtained


of the actual values of the errors.

highlyprobablethat

in
largely

Bessel

excess

the

8*5.

of J^ {vx)luhen
Descriptive
properties*

The

contour

which
integral,

"

obtained

was

a;

in

on

the

for the

are

errors

1.

"8"31(I)to represent

yieldan asymptoticexpansionof the function.


But the contour
is really
of much greater importancethan has hitherto
integral
of a function,whereas an
is an exact representation
appeared;for an integral
asymptoticexpansioncan onlygive,at best,an approximaterepresentation.
And the contour
1)
integral(togetherwith the limitingform of it when x
is peculiarly
information concerning
well adaptedfor givinginteresting
/" {vx)
when
V is positive.

Jv {v sech a) was

shewn

in

" 8*4

to

In the contour

take
integral

%u
so

that

With

logr,

the contour

equalto

its

and
positive

write

log{?'e'^|,

0.

selected,
X

is

be

to

sinh

conjugatecomplex,and

"

the

path

of

is its own
integration

flexion
re-

in the real axis. Hence


foc+-rri

qv

The

(2)XVI.

results of this section

(1917),pp.

150"174.

are

(X sinh w"w)

^y

in rather greaterdetail
investigated

in Proc. London

Math.

Soc.

8-5]

FUNCTIONS

Changing

find that the

the notation,we

is

"

sin 6

that

x^md{

log

\
X

last

This

6-

for /", the value of (lu

is made

this substitution

and, when

so

the contour

equationof

so

253

2^

1
+

ORDER

LARGE

OF

'-

cot ^

sin ^0

be
expressionwill invariably

V(^'

denoted

"

*' sm^

the

by

x'sinh w) is

6).
symbol* F(6,x),

that

-re-^'P^''-^Ue,

J,(vx)

(1)

under
by differentiating
it is found that
justified)

.//(")=

(2)
is also

Before

from

J^'{vx)

proceedingto

is convenient

easilydeduced
=

to set

^-.
'liri

the

X-

"

easily

sm-

equation

''-"''
e" ^"'^''"'^

sinh

lo

dw.

further results

obtain

concerningBessel functions,it
oi F(6,x). The reader will
properties!

record various

on

is

'\J{U'
^p^^ldff.
ifV-"..
u)
TT

This

integralsign (a procedurewhich

the

and

easilyverifythat

4^,o,.)
-^-^.o.

(4)
SO

that

F{e,x)^F(0,x)^F(0,l)

(5)

0;

also

and

^
.

n/n

"

Next

we

^'

sin 6

shall establish the

more

abstruse

cos

property

F(e,x)^F(0,x)-\-^(e^--x''sm'e)/^(l^x').
'

(7)
To

"

prove

it,we

shall firstshew
^
,a

This

function

will not

"

that
a;^sin ^

COS

be confused

with

Schlafli's function

defined

t It is supposed throughout the followinganalysisthat 0".r^l,

in

0^0

" 4'15.
^tt.

254

BESSEL

OF

THEORY

[CHAP. VIII

FUNCTIONS

It is clear that

"^/(l+x%

g(7r,x)=l
that,if g{d,x),qua

so

that

value

greatest value when

6 had

that, no

not

at

or

"

(^o'
-

tt,

its

then

tt,

/^

sin^ d^f

therefore
g

so

0 and

^'sin ^oCos ^o)^

(^0

sin- ^o)^

X-

6q between

2^0

A'- cos

(6,^

value

its greatest value

\J{1+X'). If, however, g{6,x) attained

less than

1"

and

of 6, attained

function

be

would

{6,x) ^g {do,x) \/{l


=

where

matter

exceed

g(d,x)

af-cos

2^o)^ \/(l+ ^),

attains its greatest value, that

value

does

V(l +-'^'')-

Hence

dF(0,^)

d-x^sinlcos.e
_

and

so

whence

(7) follows

at

once.

of the
Another, but simpler,
inequality

prove

this,observe that
^{0'
^-^^^^
"

and

type is

F(d,x)^F{0,x)+^d'^{l-x').

(8)
To

same

integrate
; then
From

the

these results

and // (px)qua
J^{i"x)

we

x^

sin^ e)^e

V(l

^),

is obvious.
inequality
are

now

in

functions of

to
position

obtain

theorems

concerning

v.

Thus, since

dJ^(vx)

i1

dv

rF(d,
dd
x) e-"^^''""^

0,

TTJO

decreasing
integrandbeing positiveby (5),it follows that Jt,(vx)is a positive
is a positive
// (I'a;)
decreasing
function
functionof v\ in like manner,
ofv.
the

Also, since

OV

TT.'0

integrandbeingpositive
b}^(5),it follows that e"^^^-^'J^{vx) is a decreasing
is e"^'*^-*)
//(i/a;).
functionofv; and so also,similarly,
the

8-51]

OF

FUNCTIONS

LARGE

ORDER

255

have

Again,from (8) we

Q"vF[0,X)j-ir
g-vF{0,X)j-ir
exp

J, (vx)^

"^

1
^')1cie
[- i z;(9-V(
-

{}

r
g-vF(0,x) /-co
CO

exp[-^ve\/(l-x-)]dd,

"
A

SO

that
g"vF(0.x}

(1
The

\/{2'7rv)
a;^)*

expressionis easilyreduced

last

so

values
positive

o^

for all*

The

v.

for JJ

result
corresponding

approximate expression
expressionis always in error by

Carlini's

8-11) for J^{vx); and


(""1*4,
excess,

Carlini's

to

{vx) is derived

from

(7). Write

e--x-^in-e=G{d,x),
G for

replaceG {6,x) by

and

brevity.

Then
^

2xJ: {vx)

fV^^(^- 'Ei^

[G (6,x)]-icW

-'

cL"

Tr.'o
fv^
g-vFtO,:r)
T
.

exp

G-^dG
I- ^vG/'^il+ x'^)]

exp

[- ij.{?/^(l
+^2)1 Q-hdG,

g-vF{0,x} rao
^

and

so

xJJ {vx)^ e-^^'f.^) (1

(10)
The

of the factor

absence
It is

in

to
possible

very similar
This

Wfi shall

large values
t Cf.

^^^^'"^

will be

It is evident
of

Froc.

is remarkable.

shall establish

we

is less than

concerninga single

its order.

that, when

the lemma

prove

the rate
*

argument

-16-^

lemma

the denominator

F {6,x).
concerning

now

^^^

results which

the

whose

Lemma

8"51.

the formula

,^"2)V^/(27^I^).

manner.

concludes

Bessel function

The

prove

\/(l"a;-)from

used

^^'^^^

immediatelyto

0^0

^tt,

then

^ ^-

^^-^"^sin-^

prove

an

importanttheorem

cerning
con-

of increase of J^ {vx).

from

Debye's expansion that

the

expression is in

v.

London

0 ":a;^l and

Math.

Soc.

(2)xvi.

(1917),p.

157.

error

b^-excess

for

siificientUj

256

OF

THEORY

If

sin- 6)

\J{"^ X"

BESSEL

shall first prove

H{6, x),we

dO
is

function
non-decreasing

function
non-decreasing

{6

sin 6

X-

dd

6?--A'-sin-^

sin 6

of

"

cos

of 6.

differential coefficient of this last function

The

sin^ d)(1

cosec^ O-l-^
d)-^[{d^-

cos

that

of 6 ; that is to say that

(l-^cot^)is

[CHAP. VIII

FUNCTIONS

2a;2(1

^ cot

a^)

^='cot ^ cosec'^ 9

6*

(^2cosec-

of 6 is

6){d cosec

i
-

Of

cos

sin^ 6){\- x-)

sin- (9(1- a,-'^)-],

in this expression
is positive
and every group of terms
(or zero)in consequence
of elementarytrigonometrical
inequalities.
first observe

we
trigonometrical
inequalities,

establish the

To

(9+ sin ^

(i)

(ii) 6 +

sin (9cos

(9

26-

0 $ ^ ^ tt,

sin'-(9^0,

(9-2(9-1

cos

that,when

cot

"9^

0,

(iii)sin^-^cos^-^sin^'^^O,
because

the

the

expressionson

left vanish

when

and

have

the

positivedifferential

coefficients

(i) 2(cos^-"9-isin^)2,(ii) 2 (cos^


and

(9cosec

^)2, (iii)sin

(9((9-sin ^

cos

^),

then
(92cosec2 6-6'^ cot 6

6-1

cosec-

{6-cosec2 ^

sin^ 6

1)(1

^ cot

^)+

cosec

(sin6

^)+

cosec

(sin^

-6

con

6-

1 sin-"*
6) ^ 0,

cos

^ sin^ 5) " 0,

^2cosec2^-l-Jsin2^
6 cosec^ 6

so

{6+

sin 6

cos

sin^

6-26'^

are
proved.
inequalities

that the

consequentlybeen

It has

shsAvn that

re
de
where

the variables

entiations with

understood

are

regardto
d

^ "'

\m

bo 6 and

to

^. It is

now

[FH

obvious
d

x,

and

primes denote

differ

that
{F')

rj

dd
this inequality
from
integrate

and, if we

0 to

6, we

get

F'H

0
_

Since

and
of H

the

F' and

truth

{6,x)

HJH'

vanish

of the lemma

in the last

when

becomes

inequality.

is equivalent
to
0, this inequality

obvious

when

we

substitute

the value

8-52J

FUNCTIONS

The monotonic

8"52.

OF

LARGE

property of J^(vic)fJ^(v).

prove a theorem of some


and 0 ^if ^ 1, then J^ {ikv)/J^
is fixed,
(v)is
We

shall

to the effect that,if oc


importance,
nun
-increasing
functionof v, when

now

is positive.

will
[The actual proof of the theorem
since
some
small
n
umber),
positive
arbitrarily
X-

257

ORDER

only wlien S $./""!, (wliere8 is


introduced in the jjroof
contain
expressions

is obvious

denominators; but the theorem

in their

lie valid

true

in

seen

for values

(1)

'J

--,

dJ^,{vx)

observe that,with the usual notation,

^Te-'Fi^'^^dylr,

f'^

Stta-Jo

[G^l^"^')}"^^^"
e-"^"'^'-)
dO,
dd
'

differentiate under

the

integral
sign,

F{ylr,x)e-''^^*'''^dAlr,

dv

"n"

cKjivx)

dvdx

as

it is

Ov

ox

J,{vx)=

we

; moreover,

fact that

" 0,

-^

-,

dpooc

To establish this result,we

dx

an

and
^ 8 since e"^*"-^) J^ (i/.r)

that

{yx)

dJ^iix)

.r

It will first be shewn

and, when

0 ^

functions of
when x is sufficiently
small
non-increasing
its real importance to the
Chapter xvii, the theorem owes
of x in the neighbourhood
of uniiT/.]

e-''F(o,ce)iJ^
(^)are
will be

when

an

27rxjo

27rxj

^^^
j_.d_G{ex)^_^,^,^^^

^"'

dO

^"''^^''"'
"fj^^

"l^

F{0, x) {G (6,x)]-i
0

_,dG{d,x)
{G(e,x)]i^^l^-^,F{e,x){G{e

V
TTX

'

dd

J 0

.xe-'"f"'."(W

by parts the
integrate

if we

of the two

integrals.

it follows that

Hence

'

former

dx

dvdx

ztt^xJoJo

dv

where

'dF(e,x)

n{e,ir) 2{Gid,x)]i

dd

[dF{d,x)

:^2[G{d,x)\^

dd

F(ylr,x)-F(0,x)dG{9, ^
2G(e,x)

dd

F(0, x) F(e, x) dG (0,x)


dd
'"G{d,x)
-

^0,

by usingthe inequality
x) ^F{0, x) combined
F{-\\r,
W.

B. F.

with the theorem

of

^ 8"51.
l?

258

THEORY

Since

(0,yjr)is

that is to say,

BESSEL

[chap. VIII

FUNCTIONS

the repeatedintegral
cannot
negative,
proved that

be

not

have

we

OF

.d'~Ju(vx) dJv{vx)dJ^(vx)

T/

dx

negative;

dv

that

so

J^ivxyr^0.
this inequality
between
Integrating

dJy(vx)

the limits

1^

and

1, we

get

/^

dv
that

so

dJy(vx)

are
J^{i'x)and J^"(t')

Since

this inequality
positive,
may

both

be written

in the form
3

(2)

dv

[J,(vx)/J,{v)}^0,

and

this exhibits the result which

is

function
non-increasing

write

If,for brevity,
we

F(e)^ log

(1)
the formulae*

The

proved,namely that J^(vx)/J^(v)

be

v.

of J^(v) and JJ{v).


Properties

8 "53.

(2)

of

to

was

for

F{6)

in

placeoi F{d, I),so

-^^r"

"'

^(^'
-

that
^'"' """

are
Jv{v) and Jv'{v)

.;(.) lj7^^^^ia|^^-.."
,/""4/"-....,
=

in the

first term

expansionof

{6) in ascending powers

shall prove a series of inequalities


leadingup
function of 6.
result that F {6)16^is a non-decreasing

4^7(^ \^'^) '^^^

^^'^

of 9 is
to

the

'

We

that

shall firstshew

dd\
To prove

this

we

observe

cot

value

It is to
V.

be understood

e)id^Y
,

P^-

that

\{\-e

F' id)

e-^

that Jl,'(v)means

ua,

.-,

the value

of

,i". Q

when
dJ" {.r)ld.r

.r

lias the

particuhir

260

THEORY

[CHAP. VIII

FUNCTIONS

ofJ^{v) and J^'{v).


properties

Monotonic

8"54.

BESSEL

OF

It has

alreadybeen
functions of
decreasing

("8"5) that

seen

It will

v.

the

functions

be shewn

now

J^ (p) and

JJ (v) are

that both v^J^,


(v)and

v^JJ{v)

steadilyincreasing functionsof v.

are

To

the first result

prove

dv

StT J

observe

we

that

TT

Oe-^P^^)
"Stt

^ I

(^)

dd
(^)l^""^'^'^

3^

.' 0

"J7'

"0,
since the

integrated
part
positive.
v^ Jv (v)is

Hence

vanishes

each

at

function
increasing

an

of

("8"53)the integrandis

; and

therefore

0-44731.
v^J,{v)"lim {i.-^
/,(!/)] r(i)/(2^3^7r)

(1)

In connexion

this result it may

with

be noted

Ji (1) 0-44005,
the second

To prove

e--ne)

Stt

by " 8*53 (5),and

so

the
result,by following

\v^j; (v)]^2i'-^
dv

that

2^8 (8) 0-44691.

limit and

method

same

we

find that

sin2 0)

^(^2
_

v^ J^ iv)is

function
increasing

an

of

v.

Hence
v-^ J:

(2)

It is to be noted

J J (v)]
{v^

lim

{v)"

is that the

is

0-41085.

which

is

0-32515,

4^8'(8)

0-.38854.

property of v^JJ {v)jJ^{v).

The monotonic

A theorem

(|)/(2*
tt)

that

Ji'(1)

8"55.

3^ r

more
slightly

recondite than the theorems

justproved

quotient

functionof v.
steadily
increasing
It is not

asv-^-Qo,

possible to deduce

and
f (v)~(p{v),

propertiesof/(j')in

if

these monotonic

is monotonic,
(j){v)

the absence

of further

propertiesfrom
nothing

information

can

be

the

asymptotic expansions. If,

inferred

coucerning /(j*).

coucerning monotonia

8-54, 8-55]

FUNCTIONS

prove this result we


for the four functions

To
8*54

OF

the

use

LARGE

261

ORDER

integrals
alreadymentioned

dv

Taking the parametricvariable


placeof 6, we find that

in

"" 8"53,

dv

in the first and

third

be -^ in

to
integrals

where

n, {6,^)

sin^ 0)

IF' (0)V(^-'

l^~^^^ (6)
F

sin ^

cos

^
,

"

M^_-_^_^)

The

by " 8'51.
(cf " 8"52);
0 and

(",r (6) F(e)]

to

function
overcome

XT/

'j-^^ \ifr(f)

CT

F(n.

does not seem


to be essentially
n^{d,-yfr)
positive
this difficulty,
interchangethe parametricvariables

it will be found

when
yjr,

that

Now, from the inequality


justproved,
n, (d,f)

Hi ("/.,
d)

"

2 sin- -^

sin i|rcos x/r


+ i/r
^|^^
-

^2

^^p'

,.

-p,

ti,,\)

^*^^ ^^^

P/Z3M

/zdn

^sin^cos6^-2sin-^
,

^^^^^""

^V(^--sm-.)

Since 0-^
of

^J{0- sin- 0) and


-

0i?"(^) F{0) are

0, the ftictorsof the first term

in the

sum

functions
("8-53)increasing
or
the rightare both positive

both
on

and third terms


negative;and, by
and therefore
Hence
+ i\ {^, 0) is positive,
Hi {0,~^)
positive.

both

which

""8-51,8-53, the

establishes the result stated.

second

arc

both

262

THEORY

OF

BESSEL

[CHAP. VIII

FUNCTIONS

of Bessel functionsof largecomplex order.


Asymptoticexpansions

8 '6.

(""8-31 8'42) b}-Debye in connexion with ./^(.r)


and x are largeand positive
and Y^{x) where
extended*
were
subsequent!}^
of complex variables. In the following
which is,in
to the case
investigation,
detailed than Debye'smemoir, we shall obtain asymptotic
some
respects,more
expansionsassociated with J^ {z)when v and z are largeand complex.
results obtained

The

"

supposedthat jarg

It will firstbe

where

and

between
and

TT,

i^ and vjz if

real and

/3 are

while

"

to
integrals

The

is

be

that

cosh 7

to

sinh

"

(v
The

replacew by

we

/3)cosh

shape of

a cos

the

the

integrandis
equationis

slopesof

(u

curve

sinh
a.)

"

the
as

moves

away

and

Rf(w)

has

the two

from

As

rcc

"

e^/(""dw,

at

sin y8

sin /3+

branches

shall therefore

we

vestigat
in-

be written

cosh

sin

in the form
cosh

sin /3

0.

(u -a){v- y8)sinh

Itt + ^arc tan (tanh a

cot

(tanh o

cot

acute

an

cos

/3

0,

through that pointare

jTT -!-^arc tan

denotes

and

7,

Ifij).

/3),
y8),

creases
angle,positiveor negative; Rf{iu) in-

away from 7 on the first branch, while it decreases as w


The increase (or decrease)
is steady,
7 on the second branch.

tends to +

(or

"

oc) as

tu

moves

off to

unless
infinity

the

curve

double-point
J.

Miinchener

That
"j+

tan
moves

second

stated

were

arc

(a,/3)is

near

"

where

iv,this equationmay

the

to

ic.

cosh
{(u ay -{v- /8)-}
so

correspondence

-one

whose

curve

e-zf{w)^^^

If(w)
If

one

/3 is restricted

rcc-H

stationary
point of
the

_oo

-t
=

is

".r^^\-'-f^'^'^
dw,

i7;2)(^x
=

complex. There

we

TTlJ

f(iv)

1/3),

are
investigated

H,^'^{z)=

where

cosh (a +

shall write

we

lie between f 0
suppose
may have any real value. This restriction prevents zjv from
has already("8"4) been investigated.
1 and 1, but this case

lyingbetween

cosh 'y

"^Tr, and

and AV(.r)
xl.
Sitzungsherichte,
[5],(1910);the asymptotic expansions of /^(.r)
explicitly
by Nicholson, Phil. Mag. (6)xx. (1910),pp. 938"943.
is to say 0"/3"7r.

will be

seen

later,this is the exceptionalcase.

8-6, 8-61]

(i)and (ii)denote

If
marked

in Fig'.19,

givenby

the

that

of which

contours

that of

portionare

shall write

we

"""* J

identical with
analysis

by

263

ORDER

(i)

by 7),it is found

LARGE

of the

the whole

those numbers

with

TTi

and

OF

FUNCTIONS

(ii)

(exceptthat 2/3is to be

" 8'41

replaced
(z) and S^,'-'
asymptoticexpansionsof S^^^^
(z) are

the formulae

(1)

"SV'^'
(z)

(2)

S,-"{z)

"where

'

v/(- h v-wi tanh 7) ",10

'

{hv tanh 7)'"

(^)

(|)

y
-_

'

s/{- h,v-rritanh 7) ," "

arg

and the value of arg

(" | viri tanh 7)


i sinh

("

7) which

sinh

(" i

+ arg

arg

'

(- 1 1' tanh 7)"^

lies between

^tt

"

7),

and

^ttis to

be taken.

(i)

Fig. 19.

The

values of ^4o,A-^,A.^,

are

...

^i

^0=1,
-4.

(3)

It remains
and

tIs

to do this

to express

coth=
5"ff

7 +

iff^ coth^

if^,'"
(^)in
{z) and ff^'^'

study of

intensive

an

i-Acoth^7"

the

If{^)

"SV'"
(^);

of "S^'i'
(2)and

terms

which

on

curve

7,

^/(7)

is necessary.
8-61.
The

The form

equationof

(1)-

{v

"

the

/3)cosh

curve

cos

"

a) sinh
"

where

{u,v) are

when

sin |8

cosh

coordinates

sin

and

0"

cosh

/3"

sin /3

TT-v

and

acute

tt

y8 are

written

though many

angle are

stilltrue

for

and

/S,we

of the results which


when

^ is

an

obtuse

shall also at
will be

angle.

0,

tt.

the

that 0"/8"i7r,
an

Cartesian

equationis unaltered by a change of sign in both u


in which a ^ 0 ; and since the equationis
first study the case

Since
shall

current

complex.

are

in the last section is

introduced

y8+ {u

the variables

when

of Debyes contours

and

a,

we

unaltered

first suppose

proved when

(3 is

264

THEORY

the left in

the expression
on
brevity,

For

d(f){u,v)
=

--~

sinh

(1) will
.

B^

sm

[chap. VIII

FUNCTIONS

BESSEL

OF

v).
(f)(u,

Since

smh

"

be called

sm

v,

ou

it follows that, when


the

is

vanishes
given,dcp/du

has at most

two

real roots ; and

0 "

?' "

0,

of these is infinite whenever

one

is a

multiple

have*

we

TT,

(f"{-X ,v)
(a,v)
(f)
and

root

one

so

of the

cosh

(f){+x ,v)

OC

"

is less than

and

"

"

equationin

finite root

of the

(^(w,0) 0,
seen

that, in each

to +

OC

as

tends to + 0

When

is

minimum

"

the

a,

becoming equalwhen

both

0.

is less than

to

tt

"

0, and

a,

so

for values of

when
configuration

vanishes
^,d(f"{u,v)ldu

the

largerroot

tends

at

li

0 and

lies between

o, and

"

hence

"

tt.

^(v/, /3) has


"

value
2 cosh

It

0,

consider

or

tt

curve

Next

/3|" 0^

justless than 0 or just


the equation(f)
(u,v) 0 has a largenegativeroot. The shape
is therefore roughlyas shewn
by the continuous lines in Fig.20.

greater than
of the

+ sin

"^(;^,7r)0,

this root

case,

CC

"

equations

it is

u,

the other is greater than

the
By considering

sin

{{v /3)cos fS

4"(% v)

a.

"

There

are

now

sin /8(1

two

is

so

TT.

When

at

value of u, and

only one

equationin u,
(p(w,v)

of

for

cases

"

/3cot /3

"

to consider

/3cot y8

"

"

tanh

tanh

a)

accordingas
a

or
(I)positive
(II)negative.
*

minimum

Since

d(p (a,v)ldv=cosh a. (cos^

value

zero

at

/J.

cos

v), and

this has

the

same

sign

as

i"

j3,(p (a, r) has

8-61]

FUNCTIONS

The

domains

OF

of values of the
1

is

/3cot /3

"

numbered

in

"

for the

tanh

cosh

in

Fig.21; in
expressionis negative;the

the

complex v/z

265

t/3 for which

numbered

are

2, 3, 6a, 6b, la, 7b

domains

ORDER

complex y

positive(in the strip0"$/3^7r)

domains

LARGE

1, 4, 5

(a + ^yS)have

the

the
responding
cor-

numbers

same

Fig.22.
N^

6/"

6c?

Fig.21.
When

(I)

y8cot /3

"

that the

curve

never

is that the

curve

after

so

dotted

upper

has

real root

no

between

0 and

(f)("a,v)has
that

so
negative,

Also
so

-h (a,

that the

/3
"

the real axis, the


on

27r) has

^V'"'
{z)when
Next
=

we

"

have

("
negative,the equation(f)

a,

v)

(cos /3

at

at

does not

cross

pass off to

a,
v

its value

/3,and

"

0 and

lies between

v).

cos

"

there

/3

"

is

27r.

its value there is negative,

and

'lir; hence, after

jB
"

irl,as shewn

by

crossing

the dotted

curve

discussion

the

the line

0, 0
to

"

associated

curve

with

/3$ Itt.

consider

"/"
(a,v)

on
expression
u

of the j^artof the

what

happens to

the

curve

crossingthe

after

TT.

Since
and

cosh

maximum
0

"

27r,for

"

must

Tjiiscompletesthe

therefore
onlypossibility
shewn
off to
x
as
by the

rightof Fig.20.

the

line

/3. The

"

singlemaximum
v) is negativewhen

"^(" a,

curve

is

v)jdv

a,

(p(u,v)

purve

real axis goes

tanh

"

/S)is essentially
positive,

"

the line

/3cot /3

"

is positive,
"^(u,

crossingthe
Fig.20.

{"
d(f)

Therefore

tanh

"

crosses

in

curve

When

(II)

Fig.22

the

cosh

[{v /3)cos /3
when
rightis positive
a

"

"

sin
/'

sin

(3,the

/3},
curve

never

crosses

a.; also

0 {u,iitt) {u
=

"

a.)sinh

sin /3+ (mr

jS)cosh

a cos

y8+ cosh

sin /5,

266

and

this is

when
positive
the rightmust
on
infinity
When

i.e.when

(a,/3)lies in

found

that the

passes off to

7r

curve

BESSEL

OF

THEORY

"

numbered
any of the domains
27r
does not cross
v
^, and

Fig.23.

Fig.23

Fig.21, it is
after crossing

1, 2 and 3 in
the

so

"

in

shewn

as

tti

corner

go off to

"

of

(tt /3)cot y8" 0,

which

curve

in the north-east

shewn

as

tanh

that the parts of the

a, so

lie

[chap. VIII

FUNCTIONS

by

curve

broken

curve.

Fig. 23.
We

have

now

numbered

6a in

to

consider

Fig.21.
1

and

("
(f)

a,

v) has

negative.The
u

"

Now

In such
tanh

"

maximum

after

curve,

a,

happens

(a,/8) lies in

when

the domain

circumstances

"

2tt

v
crossing

v=27r

0 ;

(tt /S)cot /3"

"

at v

until it has got above

("
"f)

what

tt,

"

^, the value of "})


("

"

a, 'Itt

remains
consequently

/3)being

the

on

rightof

y3.

in the intervals
v) is increasing

(;g,27r-/3),(27r+ /3,47r-)S), (^tt+ y8,Gtt


let the firstof these intervals in which

/3),
.

be
positive

it becomes

+ /8,2i/7r+27r-yS).
(2il/7r

Then

^ (u,^Mir

and
positive,

so

the

therefore go off to

2ir"
curve

has

^)

cannot

meet

minimum

cross

the

the left,and

on
infinity

-00

it cannot

lower
go to infinity
horizontal line in more

at

line

"

v=

a, at

which

2M'Tr + 2Tr

consequently
goes

"

its value is

it must

to

+(2M+l)7ri;
than

this,for then

than

two

points.

the

complete curve

would

268
The

reader

pass from
the

region 1

regionsla
Soc.

to
interesting

iri and

cc

from

The

differences between

and

76 appear

/3 |7r,the contours
expansions appropriateto

that the

so

for the

the formulae
been

have

to

regions6a and 66 and


by Debye, and by Watson,

overlooked

also for the


Proc.

Royal

91.

8'5) to Bessel coefficients in


(""8"11,
complex has been effected by Kapteyn* who has
1 is not
zany value, real or complex,for which

argument is
that, when z has
the

real

go

of Carlini's formuhi

extension

shewn

7^^'to

for J^ (nz).
Kapteyns inequality

An
which

that,in the critical case

prove

qc

"

[CHAP. VIII

FUNCTIONS

valid.

are

A, (1918),p.

xcv.

8*7.

to

oo

"

Note.

will find it

'BESSEL

OF

THEORY

"

then

numberf
positive

z''eKp{n\/(l
-z^)}
{1+ V(l 2')]^

\Jn(nz)\^

(1)

This
not

formula

precisethan

in the denominator

appear

powerful for
To

is less

the purposes

obtain

Carlini's formula

right,but nevertheless
it is required;]:.

for which

in which

the contour

be chosen
If

formula
integral

the

r'*-i

^-^.

{Inz{t

exp

circle of radius e", where

e"^'^,we

J^ (nz)

1/0}dt,

is

positivenumber

to

get

[n [^z(e^e'" e-"e-'")

exp

Now, if M be the maximum

iO]]dO.

value of

id]\

Iexp [\z(e"e'^ e-"e-'^)-u-

on

{irrn)"^
(1"z-)* does

inequalityis sufficiently

subsequently.
write t

we

is

the

/'""+'

J,^(nz)

the factor

the

on

consider
inequality,

the

because

the contour, it is clear that


M-.

IJn {nz)I"
But

if ^

where
/je'",

and
positive

p is

is real,then

\z {e^e'^ e-"e-'^)
-

is

|p fe"cos {a+ 6)
this attains its maximum

and

"

"

e~"

cos

i9

(a

the real part of

6)]
"

u,

value when
tan

coth

"

tan

a,

its value is then

and

p
*

Sci. de I'Kcole

Aim.

t Since

both

sides

inequalityis still true


radicals

J See

norm

(1)are

when

z-

accordingto tlie way


Chapter xvii.

sup.

of

V(sinh^u

sin- a)

(3)x. (1893),pp.

continuous
1 is

when

: for such
positive

in which

"

u.

91^120.

approaches the
values

approaches the cuts.

real

of z, either

that

the

given to

the

axis it follows

sign may

be

8-7J

OF

FUNCTIONS

Hence,

choose

now

u
[np v/(sinh-

exp

minimum,

the

equation*

sin- a)
// +
/v/(sinhof m, when

function

qua

sinh

as

small

as

this method.

this choice of
2

and, by takingz

it may

be

"

is chosen

cosh

u
\/(sinh-

"

"

root
positive

of

'

+ sin- a)

cosh

(cosh2ii

"

be real,it is clear that the

to

be the

to

proved that

\/(l "s^Osinh

e-''^),

"

positive
sign must

be taken

in

Hence

ambiguity.

and

be

expression

has

the

nu].

"

the right may


on
expression
attainable by
strongest inequality

With

sin- a)

that the

so

get the

in order to
possible

The

269

ORDER

values of a,
for all positive

j^
IJji(npe"^)
We

LARGE

sinh
[1+ x/(l z"")]

cosh

e-"'

e"-'^

so
z

log

exp V(l

i^^"
log

1 +

V(l

V(sinh-u

2^)^

2')

sin^ a) \exp v'(l

g2M

sinh^

'
"2")

g2la

sin- a

sinh
=

and

it is

now

V(sinh-u

as

both

iz exp

of this

interesting
consequence
1-J I$ 1 and

I1
To construct
=

the domain

pe'",and define

by

z'^)

the

at

;Jn

(nz) i ^

the last

cosh

write
is satisfied,
inequality

p'

that,when

exp v(l "2')

l+^/(l-z')
then

p
This

sin- a)
u +
\/(sinh-

equation is

long

sin- a)

once

so

"

equation
u

"

'

u
+
\/(sinh-

shews
previousanalysis

"

V(i-^')

in which
u

\/(l "2")
I

"

sinh

The

u,

"

is that
inequality

"

sin- a)

( 1 +v(i-^T'

\z exp \/(l
^^

before

li

clear that

/h (n^) I^
An

cosh

quadratic in sinh-

"

with

0.

one

positiveroot.

as

270

[chap.

FUNCTIONS

BESSEL

OF

THEORY

VIII

It

that

follows

2u

As

sin^

p-

sinh

increases

from

sinh

{u cosh

sinh

"

u).

'

2u

sin-

1-1997

to

increases

from

0
z

from

creases

and

to

the

on

Fig.

in

is shewn

theory
When
the

take
TT

is then

It

0-6627434

arc

of

The

24.

Fig.

24

tan

of

(coth

in

; it

will

of the

order

contour

domaiu

series

Kapteyn
the

of

boundary

Bessel

integraiion
w

a)

tan

clear

exp

that

\/(l

de-

^^)

"

^1

be
real

/""

\J

exceed

considerable

but

positive
radius

of
If

axis.

take

we

cosh
'

1 +

is
circle

of

not

This

origin.

curve

unity.

in

importance

(u

not

6"

to

cos

6"

on

be

these

2a cosh

(v

rays

get

"vv

do

u)

ex-^

"

{v

u)]

dv\
)

vahie

is

given

by Plummer,

Dynamical

Astronomy

(Cambridge,

1918),

p.

we

inclined

we

rays,

integer

an

b\- two

terminated

|^ |
v)-

i-estricted

so

This

the

xvil).

function
to

be

to

prove

the

containing

curve

| J" (nz) j certainly does

which

the

to

sini/TT

=$3/"

oval

an

(Chapter

\J^{vz)\4:M''-{-

and

....

inside

and

to

...,

47.

CHAPTER

POLYXOxMIALS

The

objectof

in various

occur

ASSOCIATED

WITH

definition
of Neumanns

The

9'1.

IX

EESSEL

FUNCTIONS

On(t).
polynoynial

this

chapteris the discussion of certain polynomialswhich


connected
with Bessel functions.
types of investigations

first of these

The

in Neumann's*
polynomialsto appear in analysisoccurs
function /(^)
problem of expanding an arbitrary
;inalytic
into a series of the form Xa^.Jni^)-The function On{t),which is now
usually
is
defined
called Neumanns
the
coefficient
of
as
enJn {z) iu the
jJolynomial,
expansionof l/(^ z) as a series of Bessel coefficients f, so that

of the
investigation

"

.-^

(1)

Jo (z)0, (t)+ 2.1,(z)0, (t)+ 2J, (z) 0, (t)+

this definition

From

it will

and

...

^n-Jn{z)Onit).

expressionfor the function,


explicit
then appear
that the expansion (1) is valid whenever
\z\" ^t
obtain this expression,
that |2;|
"
assume
|^|and, after expanding
ascending j)owers of z, substitute Schlomilch's series of Bessel
("27) for each power of 2'.
shall derive

we

an

order

In

l/{t z)
"

to

in

coefficients
This

proceduregives
1

^ ^_

1
=

7I

Assuming
*

Tluorif

dcr

2M

"

e2m'J2m{z)+

ni=0

l f'

Maih.

Lxvii.

deriv^tlie

(1867), pp. 310


differential

"

'6\\.

Neumann's

equation which

will

m-l)l

Js^2m{2)l.

'"i

repeatedseries

(Leipzig,1867),

Functioiicn

"

(./n0

that the

for the moment

BeascVschen

(s + 2m).(s

-f^j :i

pp.

is absolutely
convergent

"

lo, 83

see

also

Journal

:|',

fiir

procedure,after assuming; the expansion (1),is


solve it
to
be given subsequently (" !)-12)and

to

in

series.
f In anticipation of " 16 11, we observe
by substituting for l/(( :) in the formula

that

the

expansion

of

an

arbitraryfunction

is obtained

{z-^)f[t)(lt

X Cf. Pincherle's
pp. 224"225.

rather

more

general investigation,Reiuliconti

Jt. Lst.

Loiiibardo,[2)xv.

(1882),

272

effect

we

is

BESSEL

OF

THEORY

s by
rearrangement by replacing

series of Bessel coefficients ;

1
=

Accordinglythe

functions

e"

0.(0

(3)
It is easy to

(4)

We

("

the

equations

1):

"^

(2"

2)

before there is any

4
.

(2"

2) (2n

"

"

"

4)

denominator

of a
possibility

r
factor

negative.

or

have

"

by

l/".

i +
^^;^^

and the series terminates

being zero

defined

that

see

On (t)

e.

^,,

I ^n {z)

'

,n-2m+i

It**"

series
2m, and the rearranged

"

n.(n-m-l)l)
|"i(.-i)2"-^r"-i
^ ".
",

On{t) are
1

^x

thus get

we

..

f2m^2m (^)+

[CHAP. IX

FUNCTIONS

to

now

consider

"

2"

should

be convergent.

(4),we

have

,"!o

r,i

{s+ 27n).{s+ m-l)\

r7i"+"i1 \

To

for

is that the series

\l{t z). A sufficient condition


"

of rearranging the repeated series


permissibility

the

prove

'"'''+2'''
f.^/
If

^\

that this is

I''**

2-

^'"
"

actuallythe

^,"!o
-n^Hs

case,

observe

we

^ir^l) !

that,by "

2*11

^^^iil^l)

"2(^l2i)' + 2"'{exp(i|0P)}/(2m)!
"(i|.i)"exp(il^|2).
Hence
2'

"

^,I|7^I

s=l

''

f"

^
\,m=0

+ 2?w).(s+ OT-l)!,
(.s

^r~
"'

,,1

\z\'

"

^l"^..2m(^)|U2
^iexp(i|0p)
,

"

I'

s=l

,,,

",

Ig Iexp (^ Iz P)
^
"

I^Ki^l-MI)
The
under

absolute

of the repeatedseries is therefore established


convergence
the hypothesis
that i^|"!i{j.
And
the expansion(1) is valid when
so

\z\" \t\,and the coefficients


by (2) and (3).

of the

Bessel

functions

in the

expansionare

defined

It is also easy to establish the uniformity


of the convergence
the
(1) throughout
regions\t\^R, |^^ |^ r, where R "r"0.

of the

pansion
ex-

9-1]

ASSOCIATED

When

these

POLYNOMIALS

the
satistied,

are
inequalities

273
moduli

of the

sum

of the terms

does

not

exceed

Since

the

(g+ 2m).(g + m-l)!

i -^

the

follows from

convergence

right is independent of

the

expressionon

(^r)"-^2'"exp
(|-/-2)1
exp (ir^)
and

", the

uniformityof

the

of Weierstrass.

test

a Bessel function
by Neumann
of the secoyid
used (cf.
kind*; but this tei'm is now
""3"53, 3*54)to describe a certain solution
of Bessel's equation,
and so it has become obsolete as a description
of Neumann's
of degree n + 1 in l/t,
function. The function 0"(0 is a polynomial
and it is
polynomialof order n.
usuallycalled Neumanns

The

function

called

0"(Owas

in Neumann's

If the order of the terms

\n

"

\{yi

or

"

found

once

for

\)"m

S^(l,_,.)!(iO--

(6)

(n~

t^

The

^^

-I

results

in

^"

the methods

" 21

of

(8)

1 ""^)
(n-

t'

3^

in the formula

be combined

mav

Ctf-

(n odd)

equations(5),(6) and (7) were


By

^^ '^'"^

/,)_l^^V^^,Jii(iZi+"^-i)'

These

in

that

0" it)
5

(5)

polynomialis reversed by writing


(2),accordingas n is even
or
odd, it is at

givenby

Neumann.

provedthat
easily

1, it is

\enOn{t)\^i.{n\).{^\t\)"^exip il\t\%

(9)

enOn(t) i^.(nl).{H)"^{l+ d),

(n"l)

"

where

|^ j^ [exp(^ \t f)

From

the series

l]/(2/;2).
-

it follows that the series San Jn

these formulae

whenever

(^)On (t)is convergent

is absolutely
convergent ; and, when
-an{zjt)"'

is outside

the circle of convergence


of the latter series. anJn{^)0,i(t)does not tend to
and so the former series does not converge.
,as^7i
oc
zero
Again,it is easy
-^

to

prove that,as

*;

en

By analogy with

oo

-^

the

W.

B. F.

"

Legendre function

'

Cf. Modern

^"Sl
|l^^

Jn {Z)On {t)

"

71

of the second

(/^-^)|
,

kind, y" {t),which

is such

that

Analysis, " 15"4.


18

274

OF

THEORY

and

[CHAP. IX

FUNCTIONS

BESSEL

which

hence

it may be shewn*
that the pointson
the circle of convergence
at
either series convergesfare
identical with the pointson
the circle at

which

the other

series is convergent. It may


also be proved that, if either
series is uniformlyconvergent in any domains
of values of z and t, so also is
the other series.
Since

series

the

functions
analytic

the

rightof (1) is a uniformlyconvergent


\z\" \t\,it follows by differentiation J that

on

when

(-)i.{p + qV._

,.r..

^ ^

where

p, q

It may

are

dPJn(2)d"iOn(t)

"

^^/n

{t-zf+9+i

be convenient

to

l/f;

Os (t)

+ 24"/t*,
S/t'

0, {t)

5/t'+ 120/^ + 1920/^".

We

0, (0

V^

0, (0

l/t+ IQ/t'+ 192/i^

4/^^

polynomial 0" (0, for


1898, pp. 4, 5.
Mittheilunyen,

The

shall

obtain

1)0.^,(0 + {n + 1) 0,Ut)

0,^, (0

(2)

been

calculated

by

^^^"^B!W,
^''Onit)
^"\~
(,,,^1)
=

0"+,(0

20,/ ( t),

(n ^ 1)

-0,(t)=Oo'(t).

(3)
The

15, have

...

the formulae
^

(1) ("

0, 1, 2,

formulae satisfied
hy On(t).

recurrence

now

following
expressions:

0, (t)

coefficieuts in the

9*11

'

^^q

record the

placeon

Otti,Bern

^^p

positive
integers(zero included).

any

0" (t) 1/t,

The

series of

first of these

stated

Math. Ann.
iii. (1871),p. 137, and
proved b}by Schliifli,
LXV.
(2),(1872),pp. 33
35, but the other two were
Sitzungsberichte,
was

Gegenbauer, ^Viener
proved some
years earlier by Neumann,

"

Theorie der BesseVschen

Functionen

(Leipzig,
1867),

p. 21.

Since

them, but
are

shall not repeat


we
earlyproofsconsisted merelyof a verification,
mulae
forgive in their placean investigation
by which the recurrence
derived in a natural manner
from
the corresponding
formulae
for

Bessel coefficients.

Taking |^ |" |^ |,observe that,by " 91 (1) and " 2-22 (7),
{t-z)^ enJn {z)On (0
"=0

It is sufficient to

use

=
-

e" COS*

the theorems

that, if 2?)"is convergent,

llbjn^ is absolutely convergent.


t This was
pointed out by Pincherle, Bologna Memorie,
X Cf. Modern

Analysis, "

|n7r Jn {z),
.

n=0

o*33.

so

also is

(4)iii. (1881

"

"Zbjn, and

2), p. 160.

that

then

276

BESSEL

OF

THEORY

By combining (1) and (2) we

at

(4)

ntOn-x (t)-{n--l)Onit)

(5)

ntOn+^(t) (n^
-

If ^ be written

1) (^ +

(w

(7)

(/I+ 1) (^

The
the

by

equivalentformulae

(n-l) tOn (t)+

-(n

(0

formulae

1) On (0

1 ) On

(t)

sin' i?i7r},

[tOn+^
(t)

sin^ Ititt.

become

{iOn-,(0

^i

sin- Inir,

l) tOn (t)+

sin^ l^iir]

order
integral
polynomialof negative

Neumann

by Schlafli*

defined

was

equation
0.n(t)
definition

this

With
values of

From

?i

formulae

(1) (7)

valid

are

"

all

for

integral

equation^satisfied
hy On(t).
differential

The
the

the

i-rOn(t).

n.

9*12.

1) (^

" 9'11 (6) and (7),it is clear

formulae

recurrence

1) On (0

1) {- ntOn+,{t)+

(^

{tOn(t)

"

sin''|n7r}

j~ X

that

r(^

ft

and

the

(8)

(^

obtain

once

these
{djdt),

for t

(6)

1 ) 0"

[CHAP. IX

FUNCTIONS

2) On+i(t)+

cos^

"

+
2^7r}

sin^ ^mr

sin^ ^nir,

the differential
On (t)satisfies
equation
consequently

(^

On(t)
1)^On (t)+ (t^ 71')
-

d^y

dy

/,

n-

"

is
and

so

the

only solution

sin- |??7r.

7?

sin- hiir

t-'Wn(t),

of (1) which

equation

"^^-^^

y=On{t)

+
^/iTT

cos^ hiir

1\

-d-'-tIV--^)y

"i"

cos-

of the differential

generalsolution

It follows that the

is

as
expressible

terminatingseries

is On (t).
convenient

It is sometimes

^-^^

to write

dt'^tdt^V

(1) in

f-

the form

^"^^'

where
/ox

^"^

(**"^^")

Moth.

Ami.

t Neumann,
Lxvii.

/A_i^/^'

^""^^^

(1867),p.

Tii.

(1871),p.

Theorie

314.

der

[nit'.

(71odd)

138.

Bessel'schen

Functionen

(Leipzig,1867),p. 13;

Journal

fiirMath.

9-12, 9-13]
method

Another

and

ASSOCIATED

of

POLYNOMIALS

constructing the differential equation is

f (J^
'

^^^^" ^^^
i^"""^*^"

so

277

+^

a;

2z^

f2"'^2"
(2), 2=

"-2

22

e2n

that

r^2
'"'}

("-2)3 ("-2)2

Now

observe

to

"-Z

+ l)"^2n l(4
l(2^i

00

and

hence

t+

fi

"/"(2).

e",9'"(0

'2

"i=0

Therefore
^'

e",./"
(2)

5 ^4
"""

^ "*"^'

^" ^^^ ^'"^"


^^^]
""'j
~

"

7(=0

On
as

equating to

" 9"11,we

in

the coefficient oi

zero

obtain

at

Neumann's

9'13.

It has been

the difterential

once

shewn

by

with

On{z).

that, if G be any closed contour,

Neumann*

\ 0^{z) On {z)dz

(1)

just
identity,

equationsatisfied by On (t).

associated
integrals

contour

side of this

left-hand

the

J^iz) on

^"

{m

0,

and

i^n)

Jc

! J^{z)On{z)dz

(2)

(m?^n')

0,

J c

[Jn (z)On (z)dz

(3)
k is the

where
the

The

firstresult is obvious

of

Om{z) On{z)

is

Cauchy'stheorem,
and the residue
origin,

is at the

simplepoleat

prove

the second

similar

and
origin,

the

manner

of the contour

because

the

there is

zero.

the

the residue at

larity
onlysingu-

grand
onlypole of the intethis pointis l/e".

V, {zOn(z)} z'gn (z)

0,

and
and /,"(z)respectively,
by zOn (z)

subtract.

If U {z) be written in

.dJjAz)
,.d{zOn(z)}_^
(
''^^

^""^^^
the result of

dz'

(Z)+ ZU{Z)
*

Theorie

{W?

dz

'

the form

assumes
subtracting

Z'U'

round

result,multiplythe equations

V,,J^ {z)

/-

of

from

third result follows in

The

To

of

the number

originover

27rikl"n,

circuits
positive
negativecircuits.

of the number

excess

der BesseVschen

Zj,n{Z)On {Z) Z'gn {z)Jm (z),


?i^)
=

Functionen

(Leipzig,18G7), p.

!!".

placeof

278

THEORY

and

BESSEL

or

[CHAP. EX

FUNCTIONS

hence

[zU {zyic+ (w'

The

hence

(2) when

deduce

we

Two

due
corollaries,

(4)

^-.

(5)

^ n^.
Ann.

Jn {X + y) 0^ (y)dy

"

The

m^

Schlafli,Math.

to

+ )
/"(0
/"(0

first is obtained

the second

9*14.
It

was

follows

by making

Neumann's
stated

by

Neumann*

'"
=/;

We

shall

now

prove

by

where

is any
of (1) will then
a

"

2-4

{x).

"

(1),namely

of variable.

change

that

^""' +

^^'^^^""'
-

induction

^'"'"

e-'du.

formula
equivalent

the

ia

/"ooexp

l\

On {z)

(2)

the formula

QC

obvious

an

)"J^

Jn+p{x)J_j,{y),

2
"

(^)+ (

,,

("^),

integral
for On{z).

0. (.)

(1)

J^,

that

are

Jr,-m (*")
+ (-)'"^n +

by applying(2)and (3) to

P=

(1871),p. 138,

iii.

0^ (^ +y) Jn Lv)dy

"4("+.y)=
and

(z)dz.

integrated
part vanishes because U(z) is one- valued, and the integral
for all values of z ; and
rightvanishes because the integrandis analytic

the

on

dz=\
{z)J^
z^^g^
J

rr) \ J^ {z)On {z)

[[t+ V(l

anglesuch

that

|a. +

+[t- V(l
"^)}"

arg

2^

|" ^tt;

on

e-' d^,
r-'))"]

writingt

u/z,the

truth

be manifest.

modification

equation(2) is

of

""

On (z) I

(3)
To

prove

(2) we

cosh
e-^sinhfl
f '"{e""
+ (-)""-"*}

observe

that

foaexpia

"

te-'^dt;

Jo

Jo

by usingthe

ia

fee exp

e-''dt, Oi(^)=

Ooiz)=
and so,

Odd.

Jo

formula

recurrence

Too

" 9'11 (2),it follows

that

we

may

write

expia

"\in{t)e-''dt,

On(z)=
J 0

where

(4)

(0
it) 2t"Pn
(t) (t"n-i
"\"n+,
-

0,

and

(5)
*

Theorie

p. 312.

"^o(0=l. "PAt)
=

der BesseVschen

Functionen

t.

(Leipzig,1867),p. 16;

Journal

fiir Math,

lxvii.

(1867),

9-14]

solution of the difference

The

4",(t) A[t
=

where

and

equation(4) is

+ 1)}"
-\-B\t-V(l + ^0}".
^{t^-

independentof

though they might be


B
(5) shew, however, that A
\\ and the

are

conditions

The

279

POLYNOMIALS

ASSOCIATED

n,

functions
formula

of t.

(2) is

established.
This

given in

proofwas
/"oc exp

written

have

symbolic form

by

Sonine*

who

wrote

where
(^"(Z)) {\lz)

/
'/:

^" {t)e~"' dt, D standing for {djdz).

of this result is due


completelydifferent investigation
is based on
the expansionof " 9"1 (1),which we
analysis

whose

we

ia-

to
now

Kapteynf,
write in

the form

When

j^ i" !2^ i we

"

"=o

have

-r,=

if p be

so

chosen

exp

J0

ii\ du

\^

("=-oo

that
1

=K^-^)-

w
z

It follows that

^-r
shall

We

now

will be sufficient

shew
to

=r[

be made

du.
i^i"4^:"i2i!.^"(f)
e~"

interchangeof summation
that, for any given values of ^

shew

and

is justifiable;
it
integration
(such that | C 1" I^ |),

Z"'

small by taking
arbitrarily

large| ;
sufficiently

\u"s/{u^+ z^)\^^{u
and

and

that the

n=N^\

can

now

\z\),

so

|2!" Jul

Math.

t ^rm.

Ann.

Sci. de

xvi.

(1880),p. 7. For

VEcole

J Cf. Bromwich,

norm.

sup.

similar

investigationsee "
syinbalic

(3)x. (1893),p.

Series, " 176.


Theonj of Infinite

108.

G-l-l supra.

280

OF

THEORY

Therefore, since !f |" i2 |,we

BESSEL

have

{u"J{u^ + z')Y

the

and
when

the left can

expressionon
and

[CHAP. IX

FUNCTIONS

be made

IC^+^ Iexp {Iz \+ i IC?}

small by taking N
arbitrarily

sufficiently
large

fixed.

are

|^ j" |^^ | we

Hence, when

have

"3

.'0

b"=-""

ienJniOOniz),

"

is defined
0" (i^)
by the equation

where
.

and

it is easy to

?i+

1.

the

that 0"

see

{z),so defined,is

polynomialin \jz of degree

in powers of z and integrated


term
term, it is easy to reconcile this definition of On{z) with the formula " 9'1
When

9' 15.

integrandis expanded*

Sonine's

investigation
of Neumanns

by
(4).

integral.

An

and suggestive
of a generaltype
extremelyinteresting
investigation
of expansionof l/(a z) is due to Soninef; from
this generalexpansion,
Neumann's
formula ("9*1) with the integral
be derived without
of " 9'14 can
Sonine's generaltheorem
is as follows:
difficulty.
"

Let
so

that

(w) be an arbitrary
functionof w
a/t
is the functioninverse to yjr.
-//v

Let Zn and

and, if yjr
{w)

cc, let lu

beingassumed

that the series

on

the

^nAn,

rightis convergent.

Suppose that for any given positivevalue o{ x,\w\"\-^{x)\ on


C surroundingthe origin
and the pointz, and \'w\"\^{x)\on
curve
*

Cf. Hobson,

t Mathematical

X This
Wiss.

II.

Plane

Trigonometry (1918)," 264.


(Moscow), v (1870),pp.

Collection

but
slightly,

modified

^ (x),

"

A^ be definedby the equationsl

Then

it

is connected

323

"

382.

Sonine's

notation

o.

closed

closed

has

been

the

and //i
his.
symbols \f/
are
with Laplace's transformation.

(Analysis)
{m"),

pp. 781"784.

See

Burkhardt, Encyclopadie der

Math.

9*15, 9-16]
curve

surroundingthe originbut

'

n-=OJ

'"^,f^' f

Then

W'^

J r]

[JC

"

the pointz.
enclosing

not

J C

27ri Jo

281

POLYNOMIALS

ASSOCIATED

tu

-/fi{x)

"

QZ"il(W\"aiX

2'rriJ J

"

jf^{x)

Jo

provided that

that the various transformations


In order to obtain

(w

take
expansion,
1/w), ^ (x)^x"

if it is assumed

permissible.

are

Neumann's

yfr{iv)
and

the result is established

(a); and

R{z)"R

^(x" +

),

then

""

11=

-'X
QO

1,

Since

^"

(-)" ^-n

dx,
+ 1)}-"]
^(ar + 1)]"+ (-)" [x" V(*'-'

e--"" [[x"

we

at

Sonine

so

notes

the

that

he

memoir

328) that

(p.

expansion of l/(a-2) converges when |2 |" ja j; and


of his generalexpansion.
gives further applications

The

9*16.

series

The

intesfral
o'

obtain Neumann's

once

(2),which
(- )"e"i!"0"

is

generating function

of t except zero.
Kaptejn*
converge for any value
of Borel, in the following
:
manner
series after the method
1

'""
,

part of his

generatingfunctionof On {z).

does not

in the later

/^

'""

"^

n.(n +

"

'^

"i=0

n=()

inj

however, has

with

"summed"

m-l)\t~''
.

(^2*/

{n + ^).{n+m)l t^"-^'^

"

\i''

"(=o

associated

'''^

_1

_1

Nieuw

1 + r^

Archief

^ n.{n

""

m-l)\

f^"'(l+t^)_\

{2m)

voor

Wiskunde

t-'"'

'^

+
'(l+i'^)
(2TO+ 1)! i!^'"

pp.
(2),vi. (190.5),

49

"

55.

0"(2),
the

282

OF

THEORY

BESSEL

[CHAP. IX

FUNCTIONS

"

-57

-Tj

is

convergent

long as (1

so

fi)zjtis

"

"

and

"-

this

integral

negative.

not

00

There

is

in verifyingthat the series


great difficulty

no

"

)"f"^"0"(3)is

an

asym-

n=o

ptoticexpansion of the integralfor small positivevalues of t when |arg 2 1" tt, and
integralmay be regardedas the generatingfunction of On (2). Kapteyn has built up
of the

theoryof

The

9 "17.

function

Neumann's

the

much

ihis result.

from

of Kapteyn s typefor On{nz).


inequality

possibleto deduce from


which closely
resembles
by On (n^;)
8-7.
in "
It is

We

so

Neumann's

satisfied
integralan inequality
satisfied by J^inz) obtained
inequality

the

have
"

^" (^^^)
=

the

2^ f 1^^^^^'^^^'
^

being a
path of integration

where

that

+ {w
^'^^"

in the

contour

value of the radical is taken

e-""- dw,
V(w^ + ^-)}"]

w-plane,and

which

so

givesthe integrandwith

the

greater modulus.
Now

is

the

pointof
stationary

and
\/(l "2^),

so

"

where

is one
path of integration
stationary
point.
If
the
ta

the

surface of the type indicated

stationary
point is
+

for which

cc

are

at

lower

the

only pass

level than
z

(2)
Hence, since

" 8*3

in

on

the

integrandis greatestat

is constructed

the surface ; and

over

both

the
w

the

ty-plane,
=

and

the pass if

exp V(l

Z-)

be drawn when
can
(2) is
joiningthe originto infinity
and since the integral
involved in (1)is convergent with this contour,
satisfied,
it follows that, throughoutthe domain
in which (2) is satisfied,
the inequality
a

contour

^')
V(l
On(nz)"r^^
\/(l 2^)
1 +

(3)

is satisfied for
character

as

some

the

constant

value of A

of " 8"7.
inequality

exp
;

""'

"

and this is an

of
inequality

the

same

284

THEORY

The

differential

OF

equationof

BESSEL

which

[CHAP. IX

FUNCTIONS

J." "(")is

solution is

where

The

generalsolution

of

(8) is An, (t)+

'^^+"(0-

1"-'

"

Of these results,
(3),(4),(8) and

due

(9) are

Gegenbauer ;

to

he

and

also

proved that
(10)
where

"

An, (t)e-* dt

formula
corresponding

The

a"

in the

I
"tti J
"

formulae
following

for Neumann's

On (t)e^^^dt
may

polynomialis

i^ cos

\narc

also be mentioned

I Ajt,
"{z) An
Jc

(12)

"

;*"""""*

(11)
The

expansionof (1 2az -\-a^)'"\ this


the residue of (t^^^)"*
is easily
An,v{t)at the origin.
provedby calculating

Cn {z)is the coefficient of

formula

(v) (v + n) C." (z),

2" i- r

y{z)dz

cos

z],

(m

0,

and

^ n)

'

[2-''J,^,n(z)An,Az)dz0,

(13)

(m^ :^w=)

J c

( z-''J,+niz)An,,{z)dz27rik,

(14)

.'c

and k is the excess


of the number
C is any closed contour, vi
0, 1,2,
circuits over
the number
of negativecircuits of C round the origin.
of positive
where

The

firstand

the second

third of theSe last results

is derived

from

the

^i'+mdv+m{z)= 0,
whence

(m

we

are

the method

provedby

of

" 9'13

equations
^v+n{z^ An,v\Z)\= Z
"

gn,v\Z),

find that

n) {2v\- m

n)

z-"

J^+ni(z)A

J c

9*3.

n,

{z)dz

\ z''-" gn,

{z)dz
{z)J"^+,"

0.

J c

polynomialSn (t).
Schldfli's

polynomial0" (t) was


polynomialcloselyconnected with Neumann's
of some
of its
investigated
by Schlafli. In view of the greater simplicity
Neumann's
convenient
it
rather
than
it is frequently
to use
nomial.
polyproperties,
A

9-3]

t" ^:^^l^l^)}(lt)--^-m^
(u^l)

S,"(t)

(1)

polynomialis

of the

Schlafli's definition*

So(t)

comparing (1) with " 9-1 (2),we


=

substitute

we

the former

l)Sn-^ (t)-^(n

2)H-'

the

by using
Neumann's

(0
l) .9"+,

cos-

nt-'

of course,

Sn{t),without

of

definition

n-TT,

S,,(t)
-

2t-' cos^

Other

get

be

proved by elementaryalgebra
of
appealingto the properties

is

polynomialof negativeorder

true

for all

from

values
integral

of

w.

forms

(3) and (4).

of the

formulae

recurrence

which

be derived

may

(5) are
(t)
nS" (t) tS,,'

2 cos-^ Imtt,

(8)

tS"_,(t)

(9)

(t) 7lS,,
(t)+ tSn (t)
tS,,+,

wfc

these formulae
for t (d/dt),

write ^

(10)

{^+77)Sn(t) tSn_,(0-2

(11)

(^

n) Sn (t)
=

UTT.

become
i
COS'-^
7l7r,

2 COS^ i

tSn+,(t)+

2 COS-1

WTT.

Jt-followsthat

(^"- n-)Sn {t) t('^+


-

and

/^tt.

formula, pointedout by Schlafli,


interesting

easilyderived

If

polynomial.

definition of Schlafli's

The

and

nS^'(t)

and, with this definition,


(4) and (5) are

is

U-'

S,,(t)

the result to (4),we

add

2 and

(4) and (5) may,

formulae

The

The

" 9-11 (1)

formulae

recurrence

the latter,

multiplythis by

we

hmr.

cos2

that
-

and from

that

once

0,j(t)in the

Sn+^it)+ .S;_i{t)

(4)

If

for the functions

find from

and (2),we

i (n

at

see

|/iSn (t) tOn (t)

('3)
If

0.

On

nil

"=o

(2)

285

POLYNOMIALS

ASSOCIATED

so

(12)

Sn{t) is a

"

l-

n) Sn-i (t)+

-In

2t sin'^
^ UTT

4-

{t)+
^'-*S'"

solution of the differential

Math.

Ann.

in.

2u

|mr

cos-

i nir,

cos-

1 ?i7r.

equation

'',+{PH-)u^-lt sin^ \
t'^i^^+t
*

cos^

mr

(1871),p.

138.

2n

from

(4)

286
It may

be convenient

BESSEL

OF

THEORY

to

the

record

placeon

S, (t) 2/t+ 16 It',

S, (0

+ 768/t',
S, (t) 2/t+ 48/t'

S, (t)

:
following
expressions

8/t'+ 96/f,

+
12/t'

[CHAP. IX

FUNCTIONS

384/^^

7680/^.

by Otti,are
generaldescendingseries,given explicitly

The

(|n + m-l):
^
2 ,il_.^.i..,,n
(ln-m)l{ity

Sn{t)=

(13)

(neven)

,=1

_2

coefi"cients in the

{ii" 1^)
-

^^

"

errors)by Graf and

"

1,2,

formulae

12, have been calculated by Otti,

...

obvious
reproduced (with some
Funktionen,ii. (Bern,

are

der BesseVschen

Theorie

24.

1900), p.

9"31.

Formulae

the polynomialsof Neumann


connecting

alreadyencountered
and Schlatli,
namely

have

We

for
polynomial "S'"
{t),

14; Otti's
1898, pp. 13
Mittheilungen,
Gubler, Einleitungin die

Bern

(/?.^1-){n? 3-)
+""""
^i
.

-^+

'

t-

The

2n(n- -2-)(n- -4^-)

2n{n"-2'-)

2n

Neumann

formulae

two

\n Sn (t) tOn (t)


=

"

and

Schldfli.

connectingthe polynomialsof

cos'^ |?l7r,

Sn-^(t)+ S,^^(t) Wn{t),


=

of which

is

the former

an

immediate
from

functions,and the latter follows


other

connectingthe

formulae

we

eliminate

cos-|w7rfrom

definitions

formulae.
due

are

" 9"3 (3) and

we

either

of the

number

shall

now

discuss

" 9"3 (8) or (9),we

find that

(1)

Sn-At)-Sn'{t) 20nit),

(2)

S"^,{t)+ S,;{t) 20n{t).

Next,

on

summing equationsof

(3)
and

.Sf"
{t)
=

-2

^^T'^*Sf'"_""_i
(t)+

sin^ i

UTT

find that

: S, it),

hence

(4)

Sn {t)+ S,,_,(0

's'^

2
m

" 9'3 (5),we

the type

Comptes Rendus,

cxxv.

(1897),pp. 421"423,

n-m-i

(t)+ S, it).

860"863;

of

Crelier*; they are

to

alreadyobtained, and

derivable from the formulae


easily
the more
importantof them.
When

functions

two

of the

consequence
the recurrence

BernMittheilungen, 1897, pp. 61"96.

9-31, 9-32]

ASSOCIATED

POLYNOMIALS

287

Again from " 9"3 (7)and (5) we have


4 [On-,it)+ On^. {t)] Sn-, (t)+ 28,,(t)+ S,,^,
(t)
S"
\Sn
{S,^2(t)
(t) Sn^,(t)}
(t)]4"Sn(t)
=

that

SO

Sn" (t)+ S, (t) 0"_, {t)+ 0,,+,


(t).

(5)

is the most

This

of
interesting
formulae

summing

Again,on

(6) On {t)

-2

of the

0'"_.","_!

(0

0"(t)+ 0n-At)

-2

9'32.
The

type of "9'11(2),we
sin-^1

Sn(z)

'rr

III

stated

Gubler's

by

Graf*

nir

0, {t)+

when

obviously
the

7i

the

+ 0,{t)+0,{t).
0',,_,,,_,(t)

sum.

proof being suppliedlater

is most

IT

temporarilyby

Y,n {Z)

"n

is

the

,"

{z)

"ft=-"-l

n-l

irJ.n-,{z) S
m,=

it

Yn+i{z)

n"

TT

now

and

it is clear that
(f}n{2),

?i-l

"

If

l.

n+1

Jn+i{Z)

in Graf

readilyproved by induction;

also,by "3'63(12),when

0, and

m-

Y,,{z)-'7TYn-,{z) 1
+ l

-{2n7rlz)Jn(z) S
in

we

Oo (0,

nfl

at

| n-Tr

cos^

[Jn(^)ym(2)-J"n(z)Yn(^)}

formula

rightbe denoted

Now

find that

-n

in 1893,

treatisef.This
true

on

Graf'sexpressionof Sn(z) as
peculiar
summatory formula

(1)

sum

Crelier.

by

hence

(7)

was

obtained

-.11=

and

the formulae

"

J ,,{z)

)i+\

Y,,(z)+ ('In-rr/z)
Yn{z)
"

m=

I
-

/,"(4
n

the rightby suppressing


or
on
terms
modify the summations
inserting
all
end
that
the
summations
the beginningand
fi'ora
to n ; and
run
so
n
coefficients
of
the
the
then see
that
and 2 F," (2)
sums
!"t/"i(2)
complete
"

vanish.

It follows that

(z)
{z) {2n\z)
(f)n
(2)+ "/)"_!
"^"+,
7r/"+,{z)\F"+, {z)+ F_"_j {z)] IT F"+i(2){./"+j
{z)+ /_"_, {z)\
^
IV, ^z) [J,,
J"_j {z)[F" {z)+ Y_, {z\\+'TT
+
(z) J_n {z)]
|i^ (_ i)"}!/"_^(^)r,^(^) /" ( .) F,_^ (2)}
-

TT

4^~^cos--^?i7r,

formula
by "3'63(12);and so (^n{z)satisfies the recurrence
by Sn{z),and the induction that "i"n(z)Sn{z) is evident.

which

is satisfied

Math.

Ann.

xliii.

(1893),p.

t Einleitiuig in die Theorie

dcr

138.
Besserscli'm

Funkliou""n,11. (Bern, 1900), pp.

34

"

41.

288

THEORY

9'33.
If

Creliers

r X

On (^)

and

ia

exp

[[t+ V(l

o-

1z\
exp

"_ f"

Hence

/a

e-^^

[{^+ v(l

dt

;^+

P)Y'\e-^' dt,

t^r +{t- V(l

t^r] dt

V(l+^^)]"-|^_^(l+f2)j"_^^
^

V(l+"0

rfi.

'^-^^^-Jo
equation,which

Sn{z)=

write

^=')

Ann.

iii.

(1871),p. 146, in

the form

"{enB-^-y^e-ne^e-ZBinhdcld^

in Crelier's researches*

is fundamental

We

v(i

given by Schlafli,Math.

was

(2)

of which

we

shall

give an

now

outline.

temporarily
Tn

and

it follows that

(1)

This

"^2^Jo

t^)Y+{t- V(l

find that

^'

''I

[CHAP.IX

FUNCTIONS

"9'14(2),namely

by parts,we
integrate

On {z)

BESSEL

integral
for Sn{z).

the formula

take

we

OF

{t+ x/(l+ t^)}^


-{t- V(l

t')Y\

then
T

-*"

71+1

"^fT"T
-t
-*-

0
'-')

"

-* n"

"

that

so

"in+i
=

1
-..

2t +
-

-*

and

^^

2^ +

continued

quotientof

fraction

having

so

the suffixes n,
It follows

"

It

follows that

...,

number

2vXl

the

2t}n

of elements

that|TJK(2t)n-i is independentof
r,

Tn^^jT^is

K{-lt,2t,...,2t)n-,'

denotingthe

..+2r

that

Tn+i_ K{2t,2t,
Tn"

+ 2^+.

elements.

simplecontinuants^

two

^'^2^

^;
the

n/ J-n"i

therefore

in the continuants.
; and

since

K{2t\=\,

have

we

T,,
=

Comptes Rendus,

cxxv.

t Chrystal,Algebra,
X Since

ii.

all the elements

this abbreviated

notation.

-l^{\+t^).K{2t)n-,,
860"863

(18!)7),
pp. 421"423,

; Bern

Mittheilungen,1897, pp. 61"96.

(1900),pp. 494"502.
of the continuant

are

the same,

the continuant

may

be

expressedby

9-33,9-34]

ASSOCIATED

289

POLYNOMIALS

and hence
exp ia
/"a3

K{2t)n-^e-''dt.

Sn{z)=2

(3)

Jo
to obtain
possible
of continuants.
Sn (z)by usingproperties

From

this result it is

9*34.

expansionof Sn{t + z) as
Schldfiis

We

shall

to

series

recurrence

formulae

for

of Bessel coefficients.

Schlafli*

that, when

\z\"' t\,

expanded in the form

be

Sn (t+ z) can

result due

the

obtain

now

all the

I S,_",{t)J,,(z).

Sn(t + z)^

(1)

00

"

values of n
this formula for positive
simplestmethod of establishing
It is evidently
true when
n
is by inductionf.
0, for then both sides vanish ;
the rightis equalto
when 7i
on
1, the expression
The

I {^",,_i
(0 +

20o (t)./"{-z)h

J,,(- z)
'SV, (t)}

m=l

e",0,n{t)Jra{-z)
0

2l(t+ z)

8, (t+ z),

by " 9-1(1)and "9-3(7).


Now, if
0, 1, 2,

n,

...

the

assume

we
we-

S^_, {t + z}- 2S,:(t+ z)

have

we

induction

"

? {Sn-,a-^
it)
-

the

"

is established ;

used the obvious

Math.

Ann.

left to the reader

t The
W.

extension

B. F.

to

"

Qo

J,",
2.S""_",
(Z)
{t)\

to

obtain the second

line in the

analysis,

result that

(1871),pj:).139
(of." 9-1).
iii.

z).
z)=^^s,,(t

s,:{t+
*

of orders
(1) for Schlafli's polynomials

have

S,,+,
{t+ z)

and

of

truth

"

141;

negative values of

tlie examiuation

follows

on

the

of the convergence

of the series is

value?,by " 9-3 (6).


proof for positive
10

290.

THEORY

BESSEL

OF

[CHAP. IX

FUNCTIONS

The

expansion was obtained by Sclilafli by expanding every term on the rightof (1) in
ascendingpowers of z and descending powers of t. The investigation
given here is due to
concerned
with a more
was
Sonine, Math. Ann. xvi. (1880), p. 7 ; Soniue's investigation
as
generalclass of functions than Sehlafli's polynomial,known
hemi-cylindrical
functions
(5510-8).
When

make

we

of

use

equation" 9'3 (7),it is clear that, when

i On-rn{t)J,niz).

On{t + z)=

(2)

m=

This

[^^ |" |^ |,

00

"

Lxvi.
proved directlyby Gegenbauer, Wiener Sitzungsberichte,
(2),(1872),
in
of
220
z
ascending powers
by Taylor'stheorem, used
223, who expanded 0" {t+ z)
obvious formula
[cf.j^9'll(2)]

pp.
the

was

"

(3)
and

2"

"^'
dtp

( )- ,C", 0"_p+2,H{t),
^^2^

,"=(

rearrangedthe resultingdouble
Graf's

It is easy to deduce

series.

results

\z\ "

(validwhen

\t\),

00

(4)

Sn{t-z)=^

S"^,"(t)J.Az),

(5)

On{t-z)=

On^,n(t)J,n(z).

m=

definitionof Neumann's

9'4.

The

The

problem
of

polynomialiln{t)-

analyticfunction into a
arbitraryeven
coefficients was
f by the
suggestedto Neumann

expanding an

of Bessel

series of squares
formulae

of

^00

" 2'72,which

express

even

any

of

power

as

series of this type.

preliminaryexpansion,correspondingto the expansionof l/(t z)


and the function n" (i) will
given in " 9"1, is the expansionof \l{t-"z'^)]
that
be defined as the coefficient of enJn{z) in the expansionof l/(f- z^),
so
The

"

"

,T^

(1)

Jo'(^)^0 (0

"i

To

obtain

^J^ (^)-^1(0

2/2^(z)a

(0

and, after
" |i |,
explicit
expressionfor n"(^),take ^'j
in ascendingpowers of z, substitute for each power of z
l/(^^ z"^^
an

"

mil

pp.

'Math.

Ann.

BesseVschen

xliii.

(1893), pp.

Functionen

Ill

"

142;

see

also

Epstein, Die

(Bern, 1894). [JuJirbuchilber

die

vier

Fortschritte

liechnung"operationen
1893"1891,

der Math.

845"846.]
t LeipzigerBerichte, xxi.

(1869),pp.

221"256.

[Math.

Ann.

in.

panding
ex-

the

(1871).pp. 581"610.]

292

By reasoningsimilar
that the domains
the

are

BESSEL

OF

THEORY

that

to

it is easy to shew
of the series Sa^ J^ {z)n,i{t)and 2a,i{zjty^

given at

of convergence

the end

" 91,

of

same.

The

have

reader should

in verifying
the
difficulty

no

curious formula, due

Kapteyn*,

to

n"(o

(10)

The

9*41.
The

-2^/;'o..(J|-,)rf^.

formulaef 07' n" {t).

recurrence

to " 9"11 (2) and (3) are


corresponding

formulae

|n"'(o
^H^-5^-??4",

(1)

"

(2)

(2/0n/ (0

-^

(3)

(2/0Ho'(0

There

to

seems

Neumann

Take

and

be

2z/{t' zj
-

t-^

2n, (0

is that described

-zi

(z)

The
in

method

by

which

| 9'11.

observe

that

{J\-,(z) /Vi (z)]/"'


-

regard to t,and with regardto

z,

that

Jo (z)/"'(z)Ho {t)+ zl

zl

[J\_,{z) J\^, {z)\n" (0/n


-

[J^^.,{z)-J\+,{z)]{n^{t)-n,{t)]|n.

comparing these res'dts,it is clear

(0-

m,

expansion" 9*4 (1),and

differentiations with

"

Tn}

expansionof " 2*5,

2 J"o{z)J:

by

iHo {t) in, (0,

these formulae

the fundamental

find

71

simpleanalogue of " 9*11 (1).

no

obtained

that,by Hansen's

On

("^2)

We

[CHAP. IX

FUNCTIONS

e"Jn'{z)^n{t)+

n=0

that

\J\^,{Z) J\^AZ)\ .[Unit)-^o{t)]ln


=

0.

M-1

On

selectingthe coefficient of Jn{z) on the left and


" 9'1),we at once obtain the three stated formulae.
(cf.
*

Ann.

LtipzigerBerichte, xxi.

Sci. de VEcole

norm.

sup. (3)x. (1893),


p. 111.
(1869),p. 251. [Matli.Ann.

iii.

equating it

(1871),p. 606.]

to

zero

9-41, 9-5]

ASSOCIATED

POLYNOMIALS

293

Oegenbaner's
generalisation
of A^eumanns
polynomialI^nCOIf we expand z'^'^^l{t
z) in ascendingpowers of ^ and replaceeach power
series
of productsof Bessel functions given in " 5"5,
a
z by its expansionas
find on rearrangement (by replacing
s hy n
2m) that
9*5.

"

of
we

"

2'"+" +

"

i |, r(ju+ |^+l)r("/ + ^g+ l) {fM+


(m=o~ r{fi+v^s+l)

t'-"'

s^o

00

()

JfjL
+ ^s +

" A" 9/x +

(ft+

2?n

"i"

(w=0

\^)

+ As +

??i

supposedthat |^ |" |i | and

theoretical difficulties than

then

are
consequently
equation

(1) Bn,^,,{t)
=

the

m)

+ -^n-m+l)T(fj,+
) r (i/
n

-'Im

v-{-n-

m)

l)

of

greater

" 9"1.

defined by
polynomial5".^_^(^),

the

-^3

formulae, none
be noted

l)V{v + \n-m

l)V{fi+ v+n-m)

by Gegenbauer*; it
investigated
of which are of a simplecharacter.

polynomialwas

It may

ml

corresponding
rearrangement

^''V{lx+ \n-m

recurrence

2m)V in-Vv

the rearrangement presents no

led to consider

We

This

2m

the

+
m\V{iJi,^-v

it is

\^) f

//)r (/A+ i?i

1/

"

"

i'

satisfies various

that

B,n.,.At)=^nt^n{t).
of Gegenbauer'sformulae are worth placing
generalisations
following
record. They are obtained by expanding the Bessel functions in ascending
on
series and calculating
the residues.
(2)

The

^. \^'''^\-^
J,{^t%m6)B,"

(3)
(4)

^.^

J. {2tsin

t-"

2'^+"

(/i+

t;

,,^,At)dt=i).

At) dt
(l")B,n.,,,
1 )r

2/0 r (/x+

(/A+

1/

w) sin" (f"

_
~

n\ r(/i +
X

In the

(5)

3^2(-

n,

1)

sin- "/)).
+ \,\^JL^\v^r\\
^l^\,ii.^"V'^n\\^l"\"\v

in which

case
special

i^+

/x

i', this

reduces

to

^ r"V''/,(2"sin(^)i^,";...(0rf^=2-=''(
LlTl J

This

formula

may
*

be stillfurther

Wiener

equalto ^ttor ^tt.


by taking"/"
specialised

(1877),
pp.
lxxv, (2),
Sitzungsberichte,

218"222.

294:
9'6.
The

The

genesisof Lommel's*

the

which

are

obtain

for
explicit
expressions

The

'polynomial
Rm,v{z).

known

result of

J^ (z) /,_i(z)
{2vlz)
-

used to express Jv+m (-2^)


in terms
of J^ (z) and
linearly
coefficients in this linear relation are
polynomialsin 1/z

obviouslybe

J"-i(z); and

[CHAP. IX

FUNCTIONS

formula

recurrence

.7^+1
{z)
may

BESSEL

OF

THEORY

as

LortimeVs

polynomials. We

shew

proceed to

how

to

them.
the system of

.Z^+i
{z),J^+2i^), Jv\-m-\
i^)from
eliminating
"

"

"

equations
7^+^+1{z)

{2(v + p)lz]J,+p{Z)+ ./,+p_i


{Z) 0,

is

seen
easily

1,

1,

-22-i(i/+ ffi-2),

0,

0,

1,

0,

0,

0,

0,

0,

0,

0,

-2z-'^{v+m-\\

By expanding in
Jv+m (2)is unity;

cofactors of the
and

1,

0,

1,

0,

0,

0,

0,

cofactor of
row

1)

cofactor of

we

see

that

the

cofactor of

0,
1,

-23-i(i/+ m-3),

-2z-i{v

0,
0,

1,

J^-i (2)is this determinant


(" )"^~'
and

-22-i(v+l)

J" (z)is
(" )"'~^

-2z-^{p+m-2),

1,

of the last

first column,

the cofactor of

-2?-i(j/+ m-l),

modified

l),

-2z-^v

by the suppression

column.

("y"~^J^(z) is denoted

thus defined,is called Lommel's


I^m,v{z),
and it is also of degreeni in v.
The

771

...

0,

Jy{z\
J,_i{z)-{2v/z)J,(z),

The

0, 1,

be

to

Jv^m{z\

The

(p

and
by the symbol ("y'^Em,y{z)'"
polynomial.It is of degreem in \jz

effect of

the last row


of the determinant
and column
suppressing
by
which Rm,v ("2')
is defined is equivalentto increasing and diminishing
m
by
and
the
cofactor
of
is
so
J^-i
/
2,"_]
(")'"~^
unity;
{z) (")*"-!
"+i {z).
v

Hence

it follows that

Jy+m {Z) Jv {Z)R,n, {Z) + /^_i{z)i?,"_i,


^+i {z)

"

Math.

Ann.

iv.

(1871),pp.

108"116.

0,

9-6, 9-61]

ASSOCIATED

295

POLYNOMIALS

that is to say

It is easy to see that*


fraction
the continued

2z-'(p+ ni-1)
The
He

function

by

somewhat

coefficients
which

way

It had
recurrence

'

2z-'

elaborate

(v +

2)

2z-'

(i'+

induction

8)

by

of

means

equation(1).

polynomial

to determine

simpler

the

Bessel functions in the

two

by Bessel,Berliner Abk., 1824, p. 32, that, in


formulae,polynomials ^b_i (2),At-i (z)exist such that
observed

9-62

2^-'

coefficients in the

; it is,however,

by usingthe series for the productof


will be explainedin " 9'61.

been

[cf."

where

of the last convergent of

numerator

defined by Lommel
was
R,n,^^,(z)
for the
an
explicit
expression

derived

then

is the
R,n^^{z)

consequence

of the

(8)]
^n-l (z)B,,{z) A,,(z)B"_i{z)=
-

"

^^

^,

,^^^

It

should

Lommel's

be

that

noticed

notation

they

(^0

the

same

shews

P"

"

(x)J

The series for LommeVs

It is easy to

see

Crelier|use

ix)

notation

which

differs from

P"
,

ix)J

{x\

liolynoviial.

J_^_,"(^),qua
)'"
formulae as /"+,"{z);

that ("

recurrence

write equation (i) in the form

,/

9'61.

and

Graft

^y,g^^

integerm, satisfies
of " 9'6 also
analysis

function of the
and

hence

the

that

{Z)
(-)'"./_,_,^

(1)

(z).
./_,{Z)R,n.,.(z)
+ ./_,+,
(Z)Rra-l,.+l

and
(z),
(z)and " 9-6 (1) by ./_^+i
Multiplythis equationby ./,_i

add

the results.

It follows that

(z)+ (-)- /_,_,"(z)/,_!(z)


(2) J,+" (z)./_,+!
=

[Ju(z)/_,+!(z)+ ./_,(z)/,_,(z)]
R,n,u{z)
2 sin

VTT

Rm,v{z),
TTZ

Cf.

t Ann.

Chrystal, Algebra,
di Mat.

(2) xxiii.

11.

(1900),p.

(1895),pp.

502.

45"

fi5; Einh'ilang

tionen,11. (Bern, 1900),pp. 98"109.


t Anu.

di

Mat.

(2) xxiv.

(1896),jip.

131"163.

in die

Theorie

der

BesscVschen

Funk-

296

have

by " 3-2 (7). But, by " 5-41, we

when

Ave

^Qn\V{-

in

1) r(i/+ n)

?i +

last summation

in the

replacen

[CHAP. IX

FUNCTIONS

BESSEL

OF

THEORY

b}-m

Now

p + 1.

it is clear

that
+ 2j)+ 1) !
{V+})m+p+^ (??t

(m+jj

and

when

so,

the series for the

combine

we

2)p

the Bessel functions,


we

productsof

find that
2 sin

vir

"

izT^H-jn

")" (i^)-'"+^^-i

sin

(-)"(m -n)\r{v+
"|"'

vir

n) (|^)-^+^*-^
,

_
~

for which

the terms

("

on

V~

"to

vanish

n"^m.

! (m

n)

of the presence

account

on

'

2n)ir{v

of the factor

+ n)n in the numerator.

When

is not
7?

n^

^s"(-r On

f.\
ii,,,,{z)-

^d;

infer that

we
integer,

an

^^^^

^^"(^Y

n)l r (v + m
n\{m-2n)\r(v
-

^
c

n=o

But

the

(^ +

definition
original

of

"^

n) (|^)-'"+^"

n)

''""1
(

,)-^n^-^n

(t*+ ";

R,n"iz),by

of

means

determinant, shews

that

or
not;
R^^{z) is a continuous function of v for all values of v, integral
.and so, by an obvious limitingprocess, we infer that (3) is a valid expression
for Rjft"(z)
when
is an integer.When
even
z/ is a
negativeintegerit may
be necessary to replacethe quotient
v

^(^

^n

n)

r{v+n)

n-v-n
r{-v-m

^^
^_^",
^
^
^

+
+

l)
+ l)

in part of the series.


The

series

(3) was

result,in
Archiv

der Math.

given by Lommel, Math. Ann. iv. (1871),pp. 108


notation,had,however,been publishedby him

different
7md

Phys.xxxvii.

(1861),pp.

354

"

of
result,dependingon the equivalence
interesting
firstnoticed by Graf*, namely that
was
(4)
R,n,{z) (-)'"Rm, -.-",+! {z).

mentioned,

Ann.

di Mat.

(2) xxiii.

111;

an

ten years

valent
equiearlier,

355.

An

"

(1895),p.

56.

the

quotients
just

9-62]

(Iz)-'" oF, (I
(5) 72,"_
(2) (v)"i
=

orders

[^^ -2{{v
where

and
+

i"

mf

(v

and

(d/dz);

is
i?", (2^)

so

+
+ 4"z^(^'^
If}-']

solution of the differential

2v +

2)(^

2i;

m) (^

,u

p. 251
pp. 332

2) ;

equation

4^-^^(^+ 1)^

0.

Ann.
stated by Hurwitz, Math.
xxxill.
(1889),
equation equivalent to this was
a
given by Nielsen, Ann. di Mat. (3) vi. (1901),
lengthy proof of it was
; and
from
be obtained
from the proof just given,may
334 ; a simple proof,differing
"

fornuila

(5).
Various

9*62.

of LommeVs
'properties

We

proceedto enumerate
in his
publishedby Lommel

polynomial.
which
concerning Rm,i'{z),

theorems

some

functions

the
satisfying
=

replacedby any
and, in particular,
are

r,_i (z)E",_",
,+, (z),

F, {z)R", {z)

};,+",
{z)

(1)

formulae

recurrence

same

were

of 1871.

memoir

In the firstplace,
" 9'6 (1) holds if the Bessel functions
other

3^

2)]y

+
An

z"-).

"

[(^ + m) (^ +

(6)

+ my -{vIf]^'^ + {{i'

in, \ -v-m;

of

l,it follows from 1 5"4

"

v,

functions of
productsof cylinder
that it is annihilated by the operator

linear combination

is a
R,"^v{z)/2

Since

I m, -\m;

have

(cf.|"4*4,4*42 ), we

of Pochhammer

In the notation

297

POLYNOMIALS

ASSOCIATED

it follows that

whence

(2) 7,+,,(z)o\_, {z) J,^,,(z)F"_i (z)


R,n,A^)[l\{z)/,_, {z)
-

Next, in

vhy

" 9'61 (2),take


The

m.

be

to

J, {z)F,_,(2)|= 2R,nA^)l{irz).
-

integer; replacem

even

an

becomes

equationthen

2 (-)'"sin
+ J-.-,. {Z)j_,n.l+.{Z)
(3) J,+,n(z)Jm+i-.(z)

VIT.

and, in'the

speciafcase
v

}2,we

R.^,n,.-m{z)l(-7r

get

{z)
J%n^,(z)+ J^_,"_i

(4)

by 2//",and

(-)- i?,,,,i_,"
iz)l(7rzl

that is to say

(2^r--""(2m-n)!(2m-2")!
^,
^
^"
.,,,

J^"+. {z)+ ./-_,"_"


(z)

(5)

"

"

"

This

is the

order is half of
In

of the asymptoticexpansionof " 7-51


case
sj)ecial
odd integer.
an

we
particular,

have

J%(z)

JU(z)

--,

irz

/,

/,

(6)

45

225,

when

the

298
Formula
direct

published in 1870 by Lommel*


expansions (" 3"4)

(5) was

[CHAP. IX

FUNCTIONS

BESSEL

OF

THEORY

derived

who

it at that

time

by

of the
multiplication

^...iWT(-)".-^-,-i
by

followed

of

cases
special

As

/ 2

\*

(7)

\5

2 \^

={~)

(-)"^J_rn-i(2)

j sin
(^^

R"t,
J (^)+

from

deduce

By squaringand adding we

2 \^

/
^

cos

product.

have

" 9*6 (1) and " 9"61 (1),we


/ 2

the coeflQcients in the

to determine

lengthy induction

somewhat

w-(i)*(T.T^-."-;i/'"-

R.^-i,
I (z).

(4) thatf

(z) (-)'"R^m,i-,H (^).


R%n,^ (2)+ R\n-:,^,

(8)

replacem by

if,in " 9*6 1 (2),we


Finally,
replacev hy v m, we get

integer2m

the odd

1 and

then

"

J^^m+i \Z)J

(9)

\Z)

-v+m+i

J^^jn^i(z)
i^z)

J_i,_^_i

"

(3)v. (1901),p. 23, is that

di Mat.

result,pointed out by Nielsen,Ann.


interesting

An

sin v7rR2,n+i,"-,n
{z)l{'rrz)
(-)"""

if

we

have

of
identity

any

the

type

/m (2)"/i, ("2) 0"


+

the

where

functions

"

/,"(2)are

m=o

algebraicin

z,

we

at

can

infer the

once

identities

two

0,
f,n{z)-Rm.v{z)

l(2)sO,
/",(2)/i!".-l,"

m=0

7rt=0

the form

in
identity
by writingthe postulated

(2)} 0,
/,"(Z){J^{z)R,n,v{l)-J.-\{z)Rm-l,v^i

"

7)1=0

" 4-74 combined

and

with

observingthat,by
algebraicfunction. Nielsen pointsout
pp. 331

"

340, that this result leads

in this

to

many

(3),the

" 3*2

memoir, and

formulae for Lommel's

9 '63. Recurrence

Jf+m (2) Jv {z)Rm, Az)=

by

in his

to

seem

(3) vi. (1901),

series of Lommel's
der

Handhuch
be

Theorie

of sufficient

polynomial.

1 and

i^

"

1 ;

on

Jv-\{z)Rm-\

v+i

comparing the

(z),
two

for
expressions

that
=

Math.

t This

Ann.

result

11.
was

der

practical

(^)lR,n,(z).
+ i?^+i_
Jy-X{z) [R,n_,^
(z)} [Jy(z)+ J"^_2
^_i
y+, (z)
*

an

formula

In the fundamental

replacem and
Jv+m {z),we see

its sequel,ibid.

expansions in
interesting

will be found
of these formulae
polynomials; some
Cylinderfunktionen(Leipzig,1904), but they do not
importance to justifytheir insertion here.

"

(2)is not
quotientJ^-i {z)lJv

(1870),pp.
obtained

627"632.

by Lommel,

Math.

Ann.

iv.

(1871),pp.

115"116.

300

THEORY

9"64.

Three-term

It is

possibleto

which

the

from

theoryof

the

from

deduce

continued

polynomials.
formulae

recurrence

by Crelier*.

[CHAP. IX

FUNCTIONS

Lommel
connecting

relations

discussed

has been

BESSEL

OF

relations

The

class of relations

by Crelier

obtained

were

fractions.

J^^m{z)and i^," {z),qua functions


three contiguous
formula connecting
the same
of m, satisfy
recurrence
precisely
of the arguments of " 9*6 (modifiedhy replacing
functions;and so a repetition
shews that
Lommel
the Bessel functions by the appropriate
polynomials)
First observe that

"

(^)(^)Rn-i.f+m+i
Rm+u,i'{z)Rm,^{z)Rn,v+m(^) J^m-i.f

(1)

Next

that

" 9-63 (2) shews

"

in

" 9-63 (2) replacem by m


; it is
{nt+ v)/zfrom the two equations

and

then

eliminate

and

by v+l,

that

seen

\^) Rm,v \^) -f^'m"2,i'+i K^/y


Rm"\,v\Z)"tl'm-\,v+\
~

"

and

so

1.

"

is

the value
It is

the left is unaffected

on

independentof
consequently

(^) 1,
Rm,Az) Rm.f+l(^) Rm+hA^) Rm-l,^+l
due to Bessel (cf " 96) in the special
case
v
essentially

\^)
Rm, \Z)R^n-n+i,
v+n

Rni+i. \Z) Rm-n,


v

so

replacedm by

had

if in " 9"63 (2) we


generally,
found that
should have similarly
More

and

v\^)Rm"n, v+n \^)


Rjn-A,

the value of the function

we

i/

by

+ n,

Rm, \^) tlm"n"\,


v-\-n V'Jy
v

and

into

by changing m
m

"

givenin

Replacem

If

and

since its value when

result

(4)

"

0.

R,n,("2')
Rm-n+\,v+n {^) J^m+i,{^)Rm-n,v+n \^) Rn~i, \Z),

(3)
a

the left is unaffected

on

It is

1.

"

\Z)

+n

consequentlyindependentof
find from " 9-63(10) that
is Rn-,,,{z),
we

result

we

and, since its value when

?";

into

by changingm

Crelier's formula

have

unity,we
(2)

of the function

and

different form
n

by

Lommel

by

1 and

f.

1 in this

?n-

it is found

and
equation,

that

Rn,v\2')Rm,v{z)Rm-n-'i,v+n+\{z)
l^m-n-\,v+n^i{z)
Rin-'i,v{z)
tween
rewrite this equationwith p in placeof n and eliminate Rm-\,v{z)be=

"

the two

we
equations,

that

see

\^)
{^)~ Rp, {z)Rm-^n-i,
Rn, (^)Rm-2}--i,v+p+\
v+n+i
{z)Rm-p-i,v+p+i
(^)J
(^) Rm-n-i,v+n+i
(^)Rm-n-\,v+n+i
R,n^^{z)
\^Rm-p-2,v+p+\
V

""

J^m,v\Z) iin"p"i,v+p+i \Z),

by (3).
we

If

obtain

we

transform

the second

factor of each term

Crelier's result {loc.cit.p.

(5) Rn, (z)Rp-m-i,


v+m+i
V

{z)

"

Ann.

di Mat.

(2)xxiv.

(1890),p.

Rp, iz)Rn-^m-i,v+m+i
v

136

of

" 9'63 (10),

143),

by means

et seq.

{z)
\Z)Rm, v\Z) -tl^;"-ji-i,
f+TH-l

^^atJi. Ann.

iv.

(1871),p.

115.

9-64J

ASSOCIATED

POLYNOMIALS

301

generallinear relation of the types considered by Crelier;it


three polynomials
Rn,,(z),Rp^^{z)which have the same
R,"^^{z),

This is the most


connects

any

parameter

the

and

argument

same

The

z.

formula

be written

may

more

symmetrically

that is to say
S

(7)

^"
Rn,A2)Rp-m-^,pJrm+\{z)
=

1)1, n, p

similar

functions

which
be obtained
may
orders differ by integers.
If we

result

whose

connects

three

any

eliminate

Bessel

'/"+m-i(z) between

equations*

the

\^)
v+n
]''
'J

\Z)

v+p

^^

'J

'J v+in

i^n"m,v+m \^)

'J v+in"i

\^) "ti'n"m"i,v+tn+\
\^)"

\^)
\^) ^^p"}n,v+7ii

'J v+m"\

\^) -'^p"m"i,i'+m-\-i
\^)"

v-\-tii\^)

find that

we

\Z) iip"),i"\,v+m+\
(-2') "' v+p \^) ^n"m~l,v-\-m+i\^)
"

"v

'J

^^

'J

the last
m,

",

n,

v+m

\^) V^n"m^u+in,\^) -t^p"m"-i,v+m-\-\


\^)
J^p"m^f+ni\^) ^n"m"i,i'+vt+i
\^)\

v+m

\^) -ti'p"n"l.i'+n+i
\^) i

is obtained
expression
V by 0, n
m,
m, p
"

"

of (5) derived by replacing


case
special
respectively.

from
i"

It follows that

n, p

m,

and

obviouslywe

can

the

prove
S

(9)

generalequation

more

^,
'^^+n{2)Rp-jn-i,i'+m+i(^)
=

'

m,n,p

where

^denotes

The
Graf

last two

and

Gubler

BesseVschen

[Note.

formulae
hint

use

have

to

never

seem

If

(2) to

we

eliminate

and/SO, replacingv hj

Jv-m\Z)=

resultingequation,
"j

positivevalues

result is

the

of the

{z),

(z)I^m~\,v

\Z)i

have

we

(z)Rm-2,

deduce

v-m

equation

parameter

(-)+"i'-l (-)"?n

1,

+ \

v-m

{Zj-

that

"

that

(^)+ "^i/
+ m-1

(z)Rin~-",
v+1

find that

we

"^v-m {z) Jv (z)R


that is to say

"^'v

By using " 9'G3 (10),we

equations

m,

109.

{^) Jv {Z)R,n-\, (Z)-Jv-\ {z)Rm--2. +

J^ 4.

"

the

J^-i {z)from

m-l

simplifythe
J" (2)

It is

though
previouslystated explicitly,
equations,Einleitungin die Theorie der

Funktionen,ii. (Bern,1900), p}). 108,

been

of such

the existence

at

W^
and

cylinderfunction.

any

-m,

i^9-6

(z)-J^-\ (z)R-m-l,

(1),which

??;,is still true

supposed temporarily tliat


symmetrical,this restriction

may

for

is the snmllest
be removed.

has

hitherto

v+1

(^)"

been

considered only for

negative values.]
of the

intepiers?;(, ",

See also the note

p;

but since the final

at tlieeiul of the section.

302

shall

We

limit

Hurwitz

9'65.

Lommel

of a

[CHAP. IX

FUNCTIONS

polynomial.

that

prove

now

BESSEL

OF

THEORY

(iy-^^".-i(^)
j^ (^).

lim

(1)

252, to discuss
(1889),pp. 250
appliedby Hurwitz, Math. Ann. xxxiii.
has an
of the zeros
assigned real value (" 15"27). It has also
the reality
oiJy{z) when
been examined
52, and by Crelier,Bern
by Graf, Ann. di Mat. (2) xxiii. (1895),pp. 49
result

This

was

"

"

pp. 92

1897,
Mittheilwngen.,

have

" 9-61 (3) we

From

96.

"

'

"=o?'!r(z' +

r(i/ + 7?i+l)
Now

?i

l)' (m-2w)!r(y

l)

write

{m

n)\ r (//+

"

"

1)

{m-2n)ir{v
so

^"''''"''

l)

that
6

If

of 6

numerator

n){m
m){v +

"

{v +

"

"

"

(m, n)

is

numericallyless

"

...

denominator, providedthat
Hence, when

7i "

N, and

1)'

the

in the
in the

N.

"

"

\v\,then each factor


factor
corresponding

in

than

1)

2?/ +

(m
1) ...{v+ m
1)

be the greatest integercontained

now

(m

(m, n)

2N,

"

I6 {m, n) \" 1,
while,w^hen

has any

fixedvalue,
lim
in-*-

is

the theorem

and

Again, since
of values

of

the

(by the

from

'
"=o

!T

(j/+

/I

1)

is established.

of 2

convergence
test due

that

Tannery'stheorem*

"!r(l/+7l+l)

of Hurwitz

1,

"-);"^^""'",

absolutely
convergent, it follows

m-*x"n

{m, n)

Since

to

'

^^

in any

is uniform

"

bounded

that the convergence

it follows
Weiei'strass),

domain

of

+ i)
(i0)''+'"i?,",,+i(2)/r(v+'/"
to its limit is also uniform

Cf.

t An
excluded

in any

bounded

domain

of values

Theory of InfiniteScries," 49.


small
the origin
region of which
arbitrarily

of

z.

Bromwich,
from

this domain

when

[v]^

0.

is

an

internal

point

must

obviouslybe

9-65, 9-7]

ASSOCIATED

the theorem

From

POLYNOMIALS

of Hurwitz

it is easy

fraction for t/^_i


{z). For, when
{z)IJ"

VI-*-

by " 9-63 (1).

On

carryingout
JXo

,/_1-'"h

"

derive

to

./"{z) ^ 0, we

2pz-'

inn

303

have

-^'m,
v+i \Z)

y.

the process of reduction

\Z

infinite continued

an

and

that
noticing

"^2{v+m-l)z

"

9^-1

-'"'"i.v+m
v^y

(I'+ m)

find that

we

1
=

2vz-'

and

+ 1) ^-^
(i.

(2/+ 2) ^-1

(v + m)

...

'

^-1

hence
1

(2)
This

procedure avoids
of the continued

J.-^{^
be

neglected;

9"7.

The

In order
convenient

the

"^V+

is due

to

Graf, Ann.

to discuss

of the
properties

di Mat.

the

wi-s-oo,

last

modified

m-iv^n

-^

(2)

(1895),p.

52.

of Lommel

zeros

it
polynomials,

is

in the notation, for the

define the

"

polynomials.

by making a change
polynomialscontain onlyalternate

gm,v\Z)

7(i

(2)xxirr.

follow Hurwitz

{-yV

(1)

necessityof jn-ovingdirectlythat, when

modified notation for Lommel

to

(i.+ 2)^-1

fraction

the method

Accordinglywe
equation*

so

(17+ 1)^-1 -2

--2vz-^--

that Lommel

reason

element

may

2i.^-i

Lommel

{v +

r{v

powers

of the variable.

polynomialgm,v{z)by
+

m-

the

l)z'''

n+l)

that

(2)

By making the requisite


changes in notation
will easily
obtain the following
formulae :
(3)

g.n+,,Az) {v +

(4)^

gm+^,.-l {Z)

""9"63,9"64,the

reader

\)g"r,.{z) zg",_,^^{z), ["9-63 (2)]


-

vg,n,Az)

in

["^'OS (1)]

(-)"
^5''m-i,.'+i

^^.-..(^)+i/".-i-.,.-:(4["9-(^'K7)]
^^{^''i/,.,.(^)|

(5)

ym+-2

(6)

{z~"'-'
g,u, (z)}
=

dz

g,n+,,.-1

{z)

{z)
(7) gm,Az)g,"+,"+,
{z)-g,"^.^^{z)g,n-i,y+i
=

grn-,:, (z),
.

z'''go,v (z)g,^^+",+i
(z).

case
[A special

This

notation

differs in

unimportant detailf* from

["9-63 (4)]

the notation

used

of

" 9'64 (5).]

by Hurwitz.

304

OF

THEORY

These

results will be

down

the

BESSEL

requiredin

the

analoguesof all the other

sequel;it will
formulae

eliminatingalternate
importance.The eliminant is

The
some

result of

{if+ m)
where

c,n

(z)

{v +

C,n {z)ffm,

{2)

gm+2,u

thus obtain the set of

We

{z) {v +
+

from

2)

the

2)Z'

iU

of

(z),

1z].

"

equations
:

(^)
{v + 4)^G,.

Ciiz)g,^,{z) (i/+ 6) z''g.2,^{z),

write

(3) is

system

g"t_o^

c, (z)g,^,{z)-(u +

to

9'64.

"

(i.+ 2)g,^ {z)


,

(8)

functions

be necessary

not

|"9*6

of

I) [(v+ m){v

[CHAP. IX

FUNCTIONS

4")z'g^,{z),
.

"
=

2m

(v+

2) g^^

c.^rn-^(z)g^_^^

{z)

^(z)-(v+

c.^ {z)g^^

(v + 2s) (/2S+2,(z)

2s +

2)z^g^^o,{z),
"

{z).
^{z)-{v + 2m) z'^g.^^n-i,
v

The

9*71.

shall

We

1/ "

"

cdl

2, in which

u'hen
ofg.2m,v{z)

zeros

giveHurwitz'

now

cl^'^
of gzm,v(z)

zeros
"

reality
of the

case

all

theyare
negative.

is

2.

"

that luhen

also that

of them

one

his theorem*

proofof

real; and

exceeds

"

positive,
except when

is a polynomialin z of degree m, we
observingthat g-2m,v{2)
that the set of functions g."m,v{z),
form
"7m-2,.'(^),
g2,v{z),
ga,v{z)

After
shew

"""

of Sturm's functions.

the real
To

Sufficient conditions

of the set of relations

existence

oi

zeros

prove

that the

We

with

case

are

set

(i)the

(ii)the theorem

that

(/2ot,v(^)-

function
the

of the real variable z,

except

at

the

quotientis discontinuous.

l/'"''"
tA
^^2m,2m-2,
g'^m~2,u(z)
dz [g.an-2,v{z))

have

-r-

W^

where
from

for this to be the

(8),combined

shall

alternate,it is sufficient to prove that the quotient

zeros

of the denominator, where

zeros

" 9"7

alternate with those of


g^m-^^viz)

is a monotonic
g2m,v{z)lgim-2,v{z)

and

2, the

"

r, ..

{z)
g,-, (z)g's,
y

g^,u

| 9'7 (3) it follows

(z)g'r,{z);
"

that

+ (v + 2m) 5l21,"i-l.2"n-2,
^^""m,2,n-2
g-2m~2,A^)
-

Z" S(?B2m-3,m-4
+ (r +
^512R.,,"_i.2m-2
=

SO

2""

2) g%m-^,y{z),

that

^^"m,^n-2 g'^-2A^)
=

{v-^ 2m) 1. {v + 2r) z'-^'^^-' g%r-^,.{z),


r

and
when

therefore,if
V

"

"

is expressible
as
^ 1, ^^2m,-im-2

2.
*

Math.

Ann.

xxxiii.

(1889),pp.

254"256.

sum

of

positiveterms

9-71, 9-72]
The

ASSOCIATED

graph that the real

of

zeros

.-

Sturm's

theorem

The

reason

quotientg^m,
I.

(1918),"

that the number

The

of

on
g^m^vi^)
of the number
of alternations of sign
end of
at the right-hand
i^),
(2')
^'o,."

"

"

zeros

of zeros
is the excess
and not the deficiency
is that the
why the mimber
is
and
function
a
not
of z,
an
function,
(^)
decreasing
increasing
{^)/g2m-2,
version of Sturm's theoi-em. See Burnside and Panton, Theoryof Equations^
v

96.

0, oc

of

of alternations at the left-hand end.

arrangements of signsfor the

The
"

in the usual

as

the number

over

g.2m-2,i"(z)
separate those of ^a^, (^)-

any interval of the real axis is the excess


in the set of expressions
go.m,v{z),
g2m-2,
the interval

305

property is therefore established,and it is obvious from

monotonic

It follows from

POLYNOMIALS

are

or

upper

negative
; and

as

follows

lower

of functions when

has the values

signsare

to

be

taken

of Hurwitz'

the truth

set

accordingas

from

is obvious

theorem

-I-1

is

or
positive
inspection

an

of this Table.

9"72.

Negativezeros

Let

be less than

^"

when
ofg.2m,^,{z)
-

"

2.

"

integers
2, and let the positive

defined

be

by

the

inequalities
"

It will
has

no

has

Provided
zero.
negative
2m is positive.

It will firstbe shewn


with those of

that*, ivhen

negative
zero; hut that,when

one

that

be shewn

now

This

2s

2s"v"-

that the

"

2.

liesbetween

liesbetween

that,in each

negativezeros

2.s and

1 and

2s-

case,

25
2s

1, g^mA^)
2, g.im,v{z)

is taken to be

so

large

(ifany) of g-mA^) alternate

g2"i-2,v{^)-

proof differs

from

the

proof given by Hurwitz;

see

Proc.

London

Math.

Soc.

(2)xix.

(1921),pp. 266"272.
20
W.

B. F.

306

THEORY

By

i(^)f
"'('2^)
+5'2w+i,i'"
9'^^-^,''\^)
1^5'aOT-i,

that

9im,v\Z) \g2m-\,v-A^) 9'2m-\,v\^)\


~

"

[CHAP. IX

FUNCTIONS

quoted in " 9'7,it is clear

of the formulae

means

BESSEL

OF

^- go,n^2,v{^)92^n^^,v-i{z)-g2m~i,v-i{z)92m,
(v + 2m) g\n_-^^^{z)

{v+ 2m) g%n-i,u(z)z-'''~'goA^)9i,"'+27n-i{z)


-

^(v

27n){g\,"^,,^{z)-z"-'^'']

"0,
and
providedthat v + 2m is positive
the
postulated, quotient

is

The

existence

negative.Therefore,in

the alternation

decreasingfunction,and

is

of the system of

of the

the circumstances

is evident.

zeros

equations" 9'7 (8) now

shews

that the set

of functions

+92s-i,M,
-92s-2,A^)"
form

The

functions.

of Sturm's

set

these functions

signsof

+, +,
and

there

are

functions

when

alternations

of

is

^^

+, +,

....

-,

sign.

oo

+,

are

(-)*,

...,

When

is zero,

the

signs of

the

are

", ",

the

(-y9o,,(z)

.-.,

signsbeing

upper

",+,

...,

when

taken

...,(-)*,

+,

"2s

"

"

1, and

2s"

"

the lower

signs

there are
2s"
1 "v""2s"2;
s and
s + 1 alternations
being taken when
2s
has
of signin the respective
Hence, when "2s"v"
l, 92m,v(z)
cases.
2s
2s
has one
1 "
"
2, g2m,vi^)
no
negative
negativezero ; but when
therefore
stated
is
The theorem
zero.
proved.
"

"

"

"

9*73.

Positive and

complexzeros

As

" 9"72,define

the

in

It will

be shewn*

now

"

"

when
of g.^^t,{z)

positive
integers by
-

2s

"

that when

7/ "

2s

"

the

paper

This

lies behueen

"

proof is

of

cited in " 9"72,

more

elementary character

than

"

2.

inequalities

2.

2s

"

"

"

2s and

but that,when v lies between


zeros;
positive
Provided
1 positive
2s
zeros.
that,in
g"im v{z)has m
that m+
V is positive.
has

"

the

"

2s

2s"

"

"

each case,

1, gim,v{z)

and
m

2s

"

is so

proof given by Hurwitz

"

2,

large

see

the

CHAPTER

The

and

functions Ji,{z)

FUNCTIONS

BESSEL

WITH

ASSOCIATED

FUNCTIONS

10"1.

H.

and

E^ (2^)investigated by Anger

F.

Weber.
the propertiesof various functions
chapter we shall examine
certain
definitions are
representationsof Bessel functions.
suggested by
defined by integralsresembling Bessel's
shall first investigatefunctions
this

In

and

Poisson's

integral,and,

functions

connected

with

of which

by Lommel,
The

It is defined

the

by

J, (^)

This

function

It follows

from

distinct.

took

" 6"2 (4) that,

function

the

of the
limit

upper

investigatedthe

(1880), pp.

cussed
^TT

noted

by Anger

B in the

"

is

that

the

i
=

I"sin(v0

should

reference

function

"

terms

of

co"{v6-ziim6)dd=z-

der

la Conn,

give Anger's
t Zurich
of

name

des
to

cos2

Naturf.
V"

Additions

vn

range

of

cos

E"

J^ (2)and

integration,

by Lommel,

to researches

cob

6)dd

(vd -zsin

Jt,(z)+ sin

(j/^-2 siu ^)
p. 15

vtt

(1855),pp.

v.

"Z^

2iT
cos

vn

(2

(cf." 10-12),but

E^

"

f) siu
as

actually dis-

was

(2); for, if

/""^
I cos

--

'

which

replace 6 by

we

get

we

he

(21/77-1/^ +

sin

^) 0?^

(z).
"29. It

was

shewn

by Poisson

that

I'TT,

did

no

more

it

seems

reasonable

function.

xxiv.
I'iirteljahrsschriJ't,
(1879), pp.

Hv(z).

also

0) dO.

(i*^- 2 .sin 6) d6

cos

Ges,. in Danzig,

Temps, 1836,
the

sin

also be made

27r_/o

Schriften

he

equation

/""

Neueste

but

v.

and

Weberf,

are

/'2'r
1

half of the

to

the

by

order

n.

J"(^) is

function

and

F.

H.

by

by Anger*,

the

27r; and

functions

two

studied

was

of argument

later

defined

integer,the

be

integralvalue

the

has

J" (^)

as

function

easilyexpressiblein

right-hand

27r_/o

integral.

183-208.

/"Sn-

1
"

be

may

Bessel's

^) f/^.

when

an

integralto

E^(e)

this function

with

Ann.

type

discussed

was

connexion

is not

It

suggested by

sin

Jn(z)

to

when

the

E, (z)

xvi.

J^ (z),is

Anger's

function

(2)

Math.

several

propertiesof

particularcase.

1'cos(i/^

same

of

as

function

similar

obviouslyreduces

convenientlydescribed

In

examined,

gral
inte-

class of functions, first defined

study a

are

We

equation

(1)

he

functions

be

to

discussing the

shall

!'"(z) we

Bessel

first function

after

whose

33

"

76.

Weber

omits

the

factor

l/iriu

his

nition
defi-

10-1]
To
in the

ASSOCIATED

FUNCTIONS

309

expand J"(z) and E^(^) in ascendingpowers


and proceedthus :
integrals
I

sin" 0 sin v6d0"

cos"*

^VTT

sin

cos
(f)
v(f)d(j)

cos"'
0

TT

!sin

^j/tt

2'"r(im-|i/ + l)r(im
a

formula

due

0 sin {^vir+ vcf))


dcf)

by

^tt+ (}ifur

hw

"

of z, write

ii.+ l)'

Cauchy*.

to

In like manner,
'^

"

n
sm'"rn d

aja
/'^cf^

cos

Jptt

! COS

'"

"

^i-

"

=--

z,

-"

But, evidently,
I

J,(^)=so

00

\i

\m,

^lin fn

( _\tn

sin^"*^cosi/^(^^+-

^2m+l

'^V

:i

,,

fn

sin^"+' ^ sin ?/^fZ6',

Tvr

that

(3)

J^ {z)

{-y-(^zr

^vrr X

cos

=0

(m

|j.+ 1) i' (m

sinii'7r S
r

=0

and

^2/+ 1)

(m

ij/+

1)r (/w+

1;/+

1)

'

similarly

(4) E, (^) sin ii^TT S


=

(-)-(l^)^
+
or(m-iz/+l)r(m
cos

"

1
|z/7r

r (m

,"=o

These

results may

be written

J^{z)

sin

22

"
_

i;2

22
("

Ll'-^'
1

"

COS

(6) E.(^)

(l'-I^-)(S--Z^2)(12_j,2)('32_j,2)(52_y2)

vir

Z*

2-

1--

9i_

VTT

T.+,-^
J/-

(2--i;^)(42-i;-)
^""'

I'TT

The

memoir)

equivalentto

formula
in

17, 1854;

see

Mem.

(l^-I/^XS^-V^)

_r--V-

IT

Results

these

were

correspondingto (5) was


to Cauchy which
was

letter

ComptesRendus,
sur

...

^2)(4-2^2) {2?-v^)(^^-^
(G^-j,^)
j.2)

vir

TT

COS

"

s^

VTT

1 +

(-r (W^^'
+ f)
1) r (m + ii/

1^^+

in the alternative forms

smvir

(5)

ij.+ l)

les

xxxix.

""

(12_j,2)(32_j,2)(52_j,2)

givenby Anger

and

Weber.

given by Anger (beforethe publicationof his


communicated

(1854),pp.

to

the French

Academy

128"135.

integralesdefiniea (Paris,
1825),p. 40.

Cf. Modern

Analysis,
p.

263.

on

July

310
For

it is convenient
("10'7),
subsequently

will be apparent

which

reason

[CHAP. X

FUNCTIONS

BESSEL

OF

THEORY

to write
z^

z^

"*"

(7)

Sq^v\Z)

(8)

s_,^,{z)

,.2\/Q2_,.2\
/'12_
(p_j;2-)(32_j,2^(1^-7^0(3"
i/2^(52-I/-^)

-.^
I2_y2

"""'

1
-

-,

z^

^-

^,,^22

^2)

(z)

s_i

E,

(z)

-^

(z)

So,.

7r

(11)

v6

cos

from

formulae
following

the

(^).

5_i,,

TT

It is easy to deduce

(^'X

i;(l-coSi'Tr)^

l-4-cosy7r
-ci

'

TT

TT

(10)

"

5o,

/iA\

"

j^sini"7r

sini/TT
/
\
T
J,
{z)

(9)

"

^2-)

^2-)
(42

have

and, with this notation,we


/n\

(22

^2

cos

(z sin 6) dd

cos

sin d)dd
(2^

sin

(^ sin 6) dd

sin

sin 6) dd
(2;

"

sin

(1

i^tt

these results
s_i

(2),

Jo
sin vd

(12)

"

vrr)

cos

"

s_i

(z),

j,

Jo

(13)

sin vd

sin

i'tt

""

(z),

Jo

(14)

vd

cos

(I

cos

1/

sin

i/tt)*" (2^),
.

"/

riff

(15)

(16)

cos

cos "^)
rf(^
i'(^cos (^^

cos

v(f"sin (zcos 0) cZ(^

sin

(1906),pp.

Also, Hardy, Messenger,xxxv.

158

6)

(v^-2sm

sm

proved that, when

10*11.

"

is

positive,
zero,

or

equal to hn

is

-r

-."

COS

fi"^

wI/TT

Jt,{z)+ J^t,{z)=

"

COS

,\

-r

J^{z)

"

"

)j"

is 1, 0

or

Anger'sfunctions.

COS

{z COS d")d(p,

Jo
,

J_^{z)=

vd)

4 sin Ai/7rr*''
/

where
(2),

" 10*1 (9),(10),(15) and (16) that


TT

{2)

?;" Jn

negative.

the formulae

from

4
/I

(1)

-3-,

his functions
with
Webe7-'s formulae connecting

It is evident

620.

"

real,it

is

611

the integral
166, has investigated

"

Jo

namely
section,

LVii.
(2),(1868),pp.
by Unferdinger,Wiener Sitzutigsberichte,

examined

accordingas

in this

(n6-cosd)dd,

has

COS

lesin*
J

and

discussed
resembling the integrals

Integralssomewhat

been

-|-i/7r
(^'X
s_j

^vir ^o (z).

cos

Jo

have

"

'

"

,.7,

cosv(f"sin{zcos(b)d(p,

"

TT
.

10-11, 10-12]

ASSOCIATED

(3)

E^(z) + E_^(2)

(4)

E^ (z)

It follows

on

J, (z)

1/0sin (^cos A) d^,

cos

"

^^

I
'

addition
cot

E_^ (z)

311

FUNCTIONS

(zcos 0) dxf).

v"b cos

cos

that

IvTT[E^(z)

(s)} I tan |i/7r


{E^(^)+ E_^ (z)\

E_^

that

so

sin

(5)
and

I'TT

J" (z)

cos

i'tt

E^ (z)

E_^ {z),

similarly
sin

(6)
The

formulae

The
Weber

recurrence

have

been

It is evident

E^ (2:) J_^ (z)


=

(5) and (6) are

cos

"

due

i^tt

J^ (z).

Weber.

to

fornmlaefor J^(^) and E"(^).

Recurrence

10'12.

vir

which

formulae
determined

by

functions of

Anger

J^ (2)
=

(cos

TT

j0

{v6

cos

sin

dd
6^)

Zj

f":^{sin(i;^-^sin^)}r

-"

2 sin

vir

TTZ

and

--"R^{z)=~V
f

E,_, (z)+ E,+, {z)

")sin {vO

6'

cos

^\^^.{cos{vO

TTZ

sin

sin 6) dd

d)\de

f'C'

2(1

coSfTr)

"

TTZ

It is also very easy

that

to prove

j._, (^)

(s)
j.,+,
-

j; {z)

1e,_,(^)-e.+,(^)-2e;(^)
From

these results

we

deduce

the

2i/
,

(2)

j,_,(^)-j,+,(^)-2j;(^),

I/)J, {z)

(^

(4)

("^ v)J, (z)


-

0,
o.

sm

TTT

J._,(^)+ J.+:(^) -J.(^)-

(3)

eightformulae

(1)

and

that
integrals

(2)
J"_i (z)+ J".+i

by the

Weber.

the definite

from

satisfied

are

J,_i (2)+

-^^

(sini;7r)/7r,

zJ^+, {z) (sinj;7r)/7r,


-

312

OF

THEORY

E,_,(^)-E,+i(^)

(6)

(^ + z.)E, (2)

(8)

(^
^,

Next

E,_i (^)+ (1

v)E, {z)
=

[CHAP. X

FUNCTIONS

2E;(^),

(7)

where

0E,+, {z) (1

i/7r)/7r,

cos

i^ttVtf,

cos

usual,stands for z(djdz).

as

the differential

construct

we

y2)j^ (^) (^

(^2

so

BESSEL

""

equations;it is evident

that

J,)[^j^_j (z)+ (sin7/7r)/7r}

(^ +

{vsin v'Tr)l'Tr

J"_i (z)
z--)

"

"

z'^Jy,
(z)+ (zsin v7r)/Tr (vsin z/7r)/7r,
-

"

that
T7

/ \
V^J^(^)

(9)

(^

sin
I')

7"7r
.

TT

We

also have

C^' v')E, (z) {'^-v) {zE,_,(z)+ (1


=

so

"

(^ +
^^

E^_i (^)
i^)

2^

E^ {z)
"

z(l +

cos

i/7r)/7r}

cos

(1

i^7r)/7r

cos

y (1
i/7r)/7r
"

cos

"

v'ir)l'ir,

that

V,E,(^)

(10)

vr

TT

Formulae

obtained
equivalentto (9) and (10) were
by Anger, Neueste Schriftender
and
xxiv.
Danzig, v. (1855),p. 17
by Weber, Zurich Vierteljahrsschrift,
formula
(9) had been discovered earlier by Poisson (cf." 101).
47, respectively;

Naturf.Ges. in
(1879),p.

10" 13.

in
Integralsexpressible

terms

of

the

functionsof Anger

and

H. F. Weber.
It is evident

from

the definitions that

rex^{"i(vd-zsme)}de.

J^(z)"i-E,{z)=-

(1)

with formulae obtained


of this result,combined
By means
definite integrals
in terms
is possible
to express numerous
Thus, from " 6*2 (4) we have
Bessel,Anger and Weber.

(2)
when

larg^

e-''"-"inh"dt
the

"|7r;

result

^-

is valid

in

(3)

when

|arg^|

|7r,provided that

have

r
0

e-'^-^-i''
dt

-^
sin

I'TT

622, it

{J,(z) J, (z)],

"

of the functions of

R{v)"0.

Again,we

""6"2

I/_, (z) J., (z)},


-

10-13, 10-14]
SO

that, when

ASSOCIATED

combine

we

FUNCTIONS

313

(2) and (8),

"

f e-2"i"h"cosh

(4)

dt

vt

h7r tan

{J,(z) J, (z)] \iv|E, {z)+ F, (z)},

vtt

e-"'"''' sinh vt dt^hir

(5)

cot

| i/tt {J,(z) J, (^)j


-

[E,(^)+ F, (z)].

tt

Jo

I e~^^"'^^^
cosh
integi'al

The

pt

dt has

ah-eadybeen

but
("6'3);

evaluated

Jo
^

.'0

does

not

powers

of

in a simpleform; its expansionin ascending


expressible
from the formula of " 6'2'2 (4),

be

to

appear

be obtained

can

2 sin

2 /"'^

/_^ (z)+ /. (z)

f"^

VTT

g-^^^o^'^*
sinh vtdt,

;'^f^^+

e^"="*^
cos

but, since
T"

i"

/I

.-.

"'""'"'"

The
9"33

formulae

i^TT

"

I'

"

7/i

,\

is an

2-

'

be evaluated

cannot

nugatory when

(2) (5)are

'""

"

consideration

under
integral

the

(" )*"sin

W(ir+^=^'('

7,1

"" ''"

-^"-^

in any

simpleform

but
integer,

from

"

*.

""6'21,

have

we

f'e"f"

(6)

i"h'

dt

l {.%(z) ttE, (z)

TT

Y, (z)},

Jo

g-nl-z.uAU ^^

(7)

(_y.+

(^)+ ^E, iz)+


{,S;^

TT

F" (^)}.

io
The

integrals

associated

'^

./0
been

have

noticed

sinh 0 rf",
.' U'
'

Whipple,

Proc. London

10-14.

"in

COJS

these types

Jo

sui

Math.

Soc.

(1885),p.

in researches

occur

(2) xvi.

cosh ")c/i!
(.f

e""^

by Coates,QuarterlyJournal.,xx.

of
integrals

Various

COS

e-"'

(1917),p.

on

diftVaction

by

prism ;

see, e.g.

106.

oflargeargument.
of Anger- Weber functions
expansions
Asj/mptotic

It follows from

" 101

(2)that, in order

to

obtain the

is largeand |arg^' |"^-7r,it


l^-j
asymptoticexpansionof the integrals
of J"A^)

260.

when

gTvt

"

zsmh

asymptoticexpansion

is sufficient to

obtain

the

^Jf

Jo
To

carry

sinh
*

out

t in
i^f/cosh
See

shall
we
investigation
of
series
ascendingpowers

this
a

expand

cosh

t
vt/cosh

and

of sinh t.

Anding, Sechsstelline Taft'lnder Be-fttelschen Ftinktionen imanhiriren Arfiumeuts (Leipzig,


Moth.
die Furtschritk' der Matli. 1911, pp. "l'J3" 494],and Takeuehi, Tohoku

1911) [Jahrbuchiiber
Journal,

first

xviii.

(1920),pp. 295"296.

314
If e^^=

have, after the

u, we

SO

II

-*"

^"

" 7"4,

of

manner

^h

^'"-^\y

"

[CHAP. X

FUNCTIONS

f(u + .llu+)

W*" +

BESSEL

OF

THEORY

l/uh ]

v\t

\d^,

that

coshz.^

^i"'-^{^-l)d^

r("+.i/"+.i+)

c?r.

"

(2m)!

TT

and, if
to

take

we

so

in

be that shewn
(wH-,l/M+.l

largethat R(p + ^"^v) " 0, and


Fig.15 of " 7 "4,we find that

i''~i

^P+

then

take

the contour

d^

(^- 1)^-1 {(t-l^'-4fsinh*"]

2iri]

(-)Pcos^y7r (^xP-^^-^jl -xY+^^-^dx


_

Jo

TT

If

and

{-)Pqo^Ivttr(p+i

^i ^1, since

0 ^

1 + 4ic (1

It follows that, when


cosh vt
_

"

TT

in the form

^i/)r(p 4-^-^1/)

a;)sinh'-^ ^

R{p "\-\"\v)^Q,

COS^I'TT 'i'
(-rr(m

cosh t

"

(2jO)!

TT

where

4a;(l a;)sinh2i

be written

real,the last expressionmay

t are

1 +

1.
we

have

KWr(m
(2m)

m=o

^-i.)

^2 ,i"h ^""

(2;?)!
For

complex values

the condition
reader

0 ^

will be

sections of

^j :";1

familiar

Chapter

Similarlywe

equationhas to be modified by replacing


by a less stringentcondition, in a way with which the
in view of the similar analysisoccurringin various

of

and

this

vii.

have
1

nu+,iiu+)

316

OF

THEORY

BESSEL

[CHAP. X

FUNCTIONS

ofAnger-Weber functions
Asymptoticexpansions
oflargeorder

10*15.

and

argument.
We

obtain

of a type similar
asymptoticexpansions,
in Chapter viii, which represent J^ {z) and
expansionsinvestigated
when
iV \and |z \are both large.

shall

view

In

now

of the results obtained

asymptoticexpansionsof

in

" 1013,
integrals

the two

it will be

the

to

adequate to

E" {z)

obtain

-f
TJo

in

As

Chapter viii,

write

we

where

/3:$ tt and

0 ^

(I)

We

is not

first consider
/

cosh

1/""

zsinh"-7/

j g"
'""Jo

increases
steadily

shall take this function


It is easy to shew

of ^

that

"
T

as

z("coshy+sinh

from

0 to

00

variable

new

t) ^/

When
vizis positive.

it is supposedtemporarily
that

t cosh 7 + sinh t

iri.

integral

ttJo
in which

cosh 7,

nearlyequal* to

the

g"

(a + t/3)

as

cosh 7 is positive,

^ increases from

0 to

00

we

t.

is

of t, except possibly
when
a monogenic function
(2%+ 1 ) Trt cosh -y+ sinh-y+ y cosh y,

where
is an
of these vakies of t is a real
n
none
integer;and, when coshy is positive,
positivenumber ; for,when y is real,(2?i+ 1) Tr/cosh y does not vanish,and, when y is a pure
the imaginary axis and the originis not one
the singularities
on
of
are
imaginary ( 2/3),
them
since y is not equal to iri.
=

The

expansionof dtfdrin ascendingpowers

a "-m

;(o+)

where

and

a",
'"

so

powers

a,"

"

liri J
.

rrrr,

we
particular

l/t in

expansionof

cosh

2(1

of

in

cosh

7)^'

'

24(1+

cosh 7
cosh 7)7'

" 83,

we

are

now

in

positionto

write

expansion

(1)

fV''*-^"*"h"(^"
i lML?i"
~

Expansions valid

near

ascending

54 cosh 7 + cosh- 7
720(1 + cosh 7)10

225
'3

general theorem

'

7'

'

t"^*"-!

the

the

1+

the

2iri

dt
/"(+")
j
t^'"+"

have

1
_

From

T--""+i dr

is the coefficient of

of t. In

'

^
dT

T-

.
"

is

T-

dt

"

r^

of

Tri

are

obtained

at the end

of this section.

down

10-15]

317

FUNCTIONS

ASSOCIATED

far,been established
; it has, so
expansionis valid when vjzis positive
that jarg^-j
"
the hypothesis
^tt,but, by a process of swinginground the
on
the range of validity
the
in the r-plane,
to cover
contour
may bo extended
domain in which |arg z \" it.
This

Next,

If

that cosh 7 is real.


for its equation

sinh

modifications

the

consider

we

write

we

sin y8+

It

sinh

studythe

in which

curve

cosh

by

"1"(ti,
v) is unaffected

Since

cosh

cos

cosh

-f

which

on

curve

sin

by methods

is real has

0.

resemblingthose

write

sin /3 +

u+iv,

by abandoningthe hypothesis

the

has to be examined

shape of this curve


" 8'61. For brevitywe
The

of

caused

cosh

of

change

It is evident

" 0.

cos

sin

cf"(u,v).

sign of

both

that the

curve

and

has the

first

we

o,

originas

its centre.
Since

it follows that, when


value of u, and

has,

at

most,

multipleof
0

When

the

so

has

and, when
"

"

"

/B

ir,

00

t')
=

value of

the maximum

cosh

(J)(+

"

sin /5 {1

tanh

"a

contour

does
"

not

the

meet

00
,

y)

CO

"

^{u, v),qua function of

line

"

tt)
=

"

/3

(tt

is at

u,

"

ir

/3)cot /S}.

0 has
or

no

real root, and

so

the

(by symmetry)

the
line

fB.

providedthat the point (a, /3) lies in one of


"1"(a, v)=0
1, 2, 3 in Fig.21 of " 8-61, the contour
the shape of the contour
continuous
curve
indicating

Hence

the

is

being

the equation^ {u, /3


negative,

TT

only one

is infinite whenever

of these

one

If this is

for

have

TT, we

"

"

sin v,

tt.

the value of "I"(u, v) then

a,

real roots ; and

two

sin /3 + sinh

assignedvalue, d(i"/bu vanishes

any

equationin

"I"("

sinh

9"I"{u,v)/du

the domains
lies as
when

bered
num-

in
a

Fig. 25,
is positive

Fig. 25.

and

the broken

curve

increases is marked

the

by

an

shape when
arrow.

a.

is

negative;the

direction in which

318

OF

THEORY

It follows that
domains

the

BESSEL

[CHAP.

FUNCTIONS

is valid when

expansion(1)

(a,^) lies in any of the

1, 2, 3.

Next,

have

we

consider

to

the

asymptoticexpansionwhen

(a,/3)does

not

To effect our purpose we have to determine


lie in any of these domains.
the
"t"(w,v)
0 which
destinations of the branch of the curve
passes through the
=

origin.
Consider
'^

first the

(a,v) has maxima

at

"y =
V

the

cannot

curve

to

the

is the smallest

strip
"

a "

cannot

curve

line

/3 and minima

"

and

when
positive
line

the

cross
cross

goes downwards
until it gets below

tanh

cross

goes off to

function

the

the

the

below

the

tt.

"

originis therefore
v=

^Ktt

"

which

point at
pointat
+

"

which

confined

/3,where

which

{(2K + 1) tt

line

(2K

"

in
infinity

/?}cot /9"

1) ir

0.

^, and

it

so

the direction of the line

"

crosses

the

^Kir.

get

we

(ML?^

r'^^',-.t-.sinh.^^^ L 5

get

we

(2-)!a.
l[^^^^%-.-"i..h.^,^l
V

(3)

TTJo

7r""

By combining these results with those


expansionsfor the domains 6a
asymptotic
If,however, y3is obtuse and
w

the
L

is

the

"

-atanha-{(2Z+

e-"^-

-"'^

/"-x+(2i+i)7r;

in

" 8'61,

below

(a,tt

"

line

/Sjcot y8"

goes below the


13)and it does not
v

"

(2L

0.

get
S

^dt^-

l-^'''
,

get
1

the

which

branch

along the

obtain

we

7a.

and

1) tt

while,if a is negativeand ^ is obtuse, we

l\

z"^"-'

obtained

and /3 is obtuse, we
Hence, if a is positive

the
positive,

origincannot cross the line


i"-axis again,so it must
go to
is the smallest integerfor which

at

(5)

"

the line

while, if a is negativeand /3 is acute,

(4)

"

where

at v

above

at

(2)

cross

the line

and /3 is acute,
Hence, if a is positive

axis of

The

is greater than

which

integerfor
1

The

that "^(a,v) is

branch

The

TT.

it cannot
similarly

and

TT,

"

1) tt

and /S is acute.
positive

"

it follows
positive,
Hence

{2n

is

(2??+ 1) tt + /3,
fi tt)is now
being greater than the preceding;and since (f)(a,

minimum

each

in which

case

/9mV
.-."-"i..^'*~i
S "--|i^".
00

1) tt,

10-15]

ASSOCIATED

FUNCTIONS

319

By combining these results with those obtained


asymptoticexpansionsfor the domains 4, 5, 6b and
Since

give the

not

(1) is the onlyone

formula

for a," when

consider the

(II) Next

Qvt-zsm\it
^^

only difference between this


the sign of cosh 7; and so, when
have
5 in Fig.21 of "861, we

a,/ is derived from

1106.

r*
/

g-zi-"coshy+sinh")
(^^

and
7

the

previousintegralis the change in


1, 4,
any of the regionsnumbered

lies in

cosh 7

'

be

cosh 7,
9

(1) failed when

this

deal with

after the method

of

(1- cosh

7)^'

^'

"24(1

" 8-42.

"

so

cosh

that

-cosh

7)

the

'

pansion
previousex-

write

we

case

vrjo

when 7 is small,justas
significant
was
nearlyequalto iri.

z{\

"

e),

It is thus found

"

sinh^,
that

oC^

TTJo

'Sir ' 0

and

namely

^'

expansionfails to

To

(1854),p.

".

This

y is zero,

by changing the signof

a",

,_
_

shall

TtJo

The

the

integral

TTJo

^"

we
practical
importance,

largeand

xxxviii.

i"^-'

where

is

by Cauchy, Comptes Rendus,

obtained

obtain

^-^-^'"^(^)

was

is of

" 8"61, we

76.

expansionsin greater detail.

other

approximate formula

An

which

in

m=0

hence

(7)
A
p. 313.

result

e"'

equivalentto

S
"""'^'rfi~^

this has been

.i^a,m+i,

given by Airey,Proc. Roi/alSoc.

"

xciv.

A, (1918),

320

THEORY

10*2.

Hardy

OF

BESSEL

generalisations
of Airys integral.

considered
by Airy and
integral
manner:
by Hardy* in the following
The

If

sinh

Stokes

has

("6'3)

been

generalised

"^,then
{2
1

cosh 20

46'-+ 2

2 sinh

3"/"

85=*+ 65

2 cosh

4(^

16s* +165^

I,2
and

[chap. X

FUNCTIONS

sinh

h"\)
=

32s" + 405=* +

10s,

generally
2

^?'jj
nc^ {2s)\F,(- ir.,i
=

In ; 1

l/s%

SI

the cosh

sinh

or

Now

being taken accordingas

is

even

odd.

or

write

Tn (t,
a)

si^,(- in, 1

P.

1
-

;1

7j

4a/"0.

that

so

f^ +
r,ao)
T.,{t,a) t^ +
T, (t,a) P+

Then

the

2a

f]at

iaf

three integrals
are
following

"

Gin

(1)

'2o?

generalisations
f of Airy'sintegral
:

{a)=i^ Tn it,a) dt,


cos

Jo

Sin(a)=

(2)

sin

Tnit,a) dt,

Jo

Bin

(3)
be shewn

It may

;[that

(")=["
a)}dt.
exp {- Tn (t,
the

first two

or
negative)if n
(whetherpositive
is complex ; and it is
a
converges when
function of a.
an
integral

real

When

is

of Bessel

functions

other two

involvingthe

Before
which

; but

when
function

to

the

obvious
fairly

third

is

integral

that Ein

(a) is

in terms
expressible
first only is so expressible,
the
are

of H. F. Weber.

we
evaluatingthe integrals,

reduce

indeed

is odd, the

But

2, 3, 4

three functions

the
integer,

even

an

convergent when

are
integrals

Cin (a) and Sin (") when


Gin (a)+ iSin (a)

observe
a

functions
integral

exist

is real ; for take the combination

exp

that

{iTn {t,a)}dt.

Jo
*

Quarterly Journal,

t The

xli.

in
sine-iutegral

pp. 168"182.

[3Iath.and

the

(1910),pp. 226"240.
case

"

was

examined

by Stokes,

Phys. Papers, ii. (1883),pp. 332"349.]

X Hardy, loc. cit.,p. 228.

Camb.

Phil.

Trans,

ix.

(1856),

10-2, 10-21]

ASSOCIATED

when
taken round an arc of
integral,
the origin(the arc being terminated
by

Jordan's lemma, the

By

with

radius R

at

centre

tends
complex coordinates R, Re^'"'"'),

with

321

FUNCTIONS

to

zero

circle of

the

points

R^^oc

as

therefore

And

rooexp(j7r?7")
+ iSin (a)
Cin (cc)

ei-^'/n exp

where

te"^"''"; and

"

of

integralfunction

an

be treated

Cin (cc) i Sin (a) may

combination

dr,
{- T" (t,oLe-"'"'^)]

integralis

last

the

{i1\ {t,a)}dt

exp

in

similar manner,

The

a.

and

the

result is then evident.

ofAiry-Hardy integrals
of even

'The evaluation

10'21.

To evaluate

Sin {a), Ein{a) when n is even,


integrals
Cin{oL),
and then, making the substitution
a is positive,

the three

that
temporarily

suppose

in the

we
integrals,

order.

2a* sinh

find that,by

we

{ujn)

" 6'21 (10),

2a* r""

Cin (a)+ iSin (a)

sav

Cin (a) + iSi, (a)


If

we

In

^^^

{e*"^'/"
J-vn (2a*")
-

equate real and imaginaryparts,we

e-*-V" J,^^
(2a*")}
.

have

similar manner,

""4 (a)

2a* T"

that,by

exp

"

"
so

du
i cosh u) cosh {ujn)
(2a*"

iriain-' e*'''"
iTi/^^^'
(2a*"),

that is to

exp

"

(" 2a*" cosh u) cosh {u/n)du,

" 6-22 (5),

(3)

^V"(a)

(2a*/n)iiri;."(2a*").

Tljese results have been obtained

on

the

that
hypothesis

is positive;
and

functions of a which reduce to


expressionson the rightare the integral
Hence,
or negative.
Cin (a),Sin (a)and Ein (a)when a is real,whether positive
when a is negative
the equations
(1),(2),(3)are stillvalid,so that,for example,

the

we

have

^*" ^"^
"

whether
W.

vilT(,n + l-lln)
2wsin(j7r/r0,,?o
a be positive
or
negative

B. F.

"

"

J,,
pil r{"u

l +

Vn)\
2T

'

322

OF

THEORY

Hence, replacinga by

"

BESSEL

/3,we

that, when

see

[CHAP. X

FUNCTIONS

and
/3 is positive

is even,

then

" 4-31 (9) that,when


annihilated by the operator

It follows from

8i,i(a) are

and

that

Einia) is annihilated

the

of the

case

because
the definitions,

from
directly

the

twice under

evaluate

it is difficult to

the

not

obtain

obtained
integrals

Gin (a) when

is odd,

and
Gin{oL)

this result*

by differentiating

convergent.

of Airy-Hardy integrals
of odd

The evaluation

10"22.
To

signare
integral

functions

^,2Qn-2

functions

first two

the

the operator

by
-Z

In

is even,

we

suppose

order.

temporarilythat

is

and then, by " 6"22 (13),


positive,
2a-

Gin (a)

r^
cos

"

n
.

du
(2a^"sinh n) cosh {iijn)

That

is to say,

ft"(")

(1)

??^^2ilW!0^.,"(.2".")
j/_""(2ci") /"" (2al")i.
-

2n sin

device

Using the

explainedin " 10'21, we

It follows that the

and, whether

be

(^tt/t?)

even

see

equation" 10-21 (4) is true


or

yS is positive.

that, when

whether

odd, Gin (a)is annihilated

by

be

even

or

odd ;

the operator

J^^(-)"n--=a"-^
+

for all real values of


*

It has

been

integrals."

o.

proved by Hardy,

loc.

cit.,p. 229, with

the

aid of the

theory of "generalised

324

THEORY

and

OF

hence, for all real values

17"2

(7)
This

It should

be noticed

'^^'^
(")
"*'~i

^'

Stokes

in the

^'

case

w\

{sin(Uhi)
'

X
=

The

a, and

"

formulae

his methods
in the

of

v-

Cauchys

In connexion

(- l)-^"'^+''
}

(- 1)*'"+''}

(2/3i")-J,/,,
{/_,/"
(2y3i")l

precedingthree sections
obtainingthem were different and
n

real.

/3 are

of the

case
special

10'3.

and

3.

X
{Tr/n)

nsm

"

{sin{^M
^^^^^^

where

"* '""'*

that

Si, (a)+ (-)4-+^)


Ein (a)

(8)

[CHAP. X

FUNCTIONS

of a,

givenby

equationwas

BESSEL

due

are

he gave

to
some

Hardy, though
of them only

3.

numbers.

with

of Bessel's integi'al
which was
defined
generalisation
studied by Giuliani (see " 10"31),
it is convenient
by Bourget,and subsequently
class of functions known
to investigate
a
as
Cauchy'snumbers.

The
the term

in

typicalnumber, N_nic,m,is defined by Cauchy*


independentof i in the expansionof

of t.

ascendingpowers

the last two

are

Cauchy'stheorem
1

(1)

supposedthat

n,

k, and

are

the coefficient of

of
integers

which

negative.

not

It follows from

It is

as

^-..M.

,"(0+)

that
1 \^" /

\""

t-'-'lt+^){t~-)dt
2;".J
27r
cos*
+ {-y e"^*}
I [e-'"^

Orn+k

r-rr

cos

TT

It is evident
if it is
*

pp.

from

d sin'" BdS

(^niTT

"

nd) cos*

6 sin'"Odd.

Jo

the definition that

is zero
iV^_n,fc,m

if "n-\-k

is odd

or

negativeinteger.

Comptes Rendu"",
682"687,

850"854.

xi.

(1840),pp.

473"

475, 510"511;

xii,

(1841),pp. 92"93;

xiii.

(1841),

10-3]

ASSOCIATED

From

N_nX,n. i-)^ NnX^a


=

results,togetherwith

These
numbers
The

formulae

recurrence

(3)

('*)

-'"'"",t,m

_n,

immediate

they are

formulae

k. in

^^-n-rl,
k-1,
-^' n+i,i-,m"

^^

"

t'-''(t+

ijtf-'(t

r"

(t+ ijtf{t

1/0"'

"'"''(t +

1/0^'
{t

of the

of numbers

terms

different class of

existence

to

)ii "

"(-i"
i,fc,

n"

"

+ r"-i
i/t)""

1/0"~'

1'""'^

Cauchy'snumber
^-n,o,mtypes H^n.k.O'

of these formulae

successive

of the identities

1/0'"

means

-^'

By

^-n-i,k-l,

{t+ i/tf(t

which

are

r"

from

givenby Bourget*.

consequences
=

(-)""*N^Xm-

recurrence

be calculated,were

may

325

that

(1) it is seen

(2)

and

FUNCTIONS

is

any

1/0'^

(^+ 1/0*(^

I/O'""'-

in
ultimately
expressible

formulae, also due

recurrence

{t+ i/tf-'(t

Bourget,owes

to

equation

the

d f

l\

It follows that

+7) }*
h^)\ {'-7)s {'-"('

27ri (1

by

{m

(5)
and

differentiation

On performingthe
integration.
partial
1 ) N_n^k,m

nN-n,k-l, m+i

nN_n.k+i,m~i

(A'
-

see

we

1 ) i\^_",
k-2.

that

m+2,

similarly
1 ) -^^-n,
m
/t,

+
(^"

(6 )

Developments

due

of Math.

Chessiu, Annals

to

(m

x.

1 ) iV_",
^.+0

6), pp.

(1895

"

"

",_2.

2, are

r=0
.s

-^ -n,

(8)
These

may

Another

be deduced
formula

by

due

k, ?rt=

induction

to Chessin

)'gC^. iV_"

"

3_2,-,

fr, ))i-8'

(3) and (4).

from
is

A^-n,k,.n=

(9)

i-YkCp-r-mOr,

("=0

where

hU-+m

"

the coefficient of
n). This is proved by selecting

{t+iiifxit-i/ty".
*

Journal

de Math.

(2)vi. (1861),pp.

33"54.

t"

in the

product

its

326

OF

THEORY

The

10"31.
The

JnA^)=
is

Giuliani.

the

of Bessel's integral
generalisation

2^.j'"^'r"-(^
^)'expji^(^-^)|
+

and
integer,

an

[CHAP. X

FUNCTIONS

get and

our

is defined by
Jn,k{z)

function

(1)
where

functionsofB

BESSEL

k is

positive
integer.

It follows that
1

Jn

(2')
=

jfc

f'"

^r-

Ztt

and

J-,,^
(,-)
=

r(2

d)\ (2 cos Qf dd,


.

The
pp. 42

^(2) has
"/",

Jo

(1888), pp.

151

satisfied

the function.

by Bourget, Journal de Math. (2) vi. (1861),


astronomical
Giornale di Mat.
applications
; while Giuliani,
constructed
linear differential equation of the fourth
a

[Note.
contains

by

An

171, has

"

earlier paper

propertiesof

another

studied

been

55,for the sake of various

"

XXVI.

order

function

by Giuliani, Giornale
of
generalisation
1
-

we

from

(1887),pp.
integral,
namely
xxv.

198

"

202,

to be

seem

incorrect.]

powers

of 2^,we

deduce

from

" lO'B

that

(1) that

(4)

= J"(4
J",o(^)

Again

from

" lO'S (2) and (3) it is

evident

that

J_,,,(2) (-)-^-J",(^),

(5)

(6)

(z)
Jn,k{z) Jn-xk-i{2)+ Jn+i,k-i

',

and, if we

take ^'

1 in this

(7)
These

Bessel's

Mat.

expand the integrandof (1)in

it is evident

di

/ cos("(9-2PsinP^)(:/^,

parts of the analysisin this paper


If

6'/cos
(?i6'-^sin 6*)fZa

cos

TT

and

sin

"

-"

therefore

(2)

but

{" i {nO

exp

formula,

J",,iz)
=

results

obtained

were

^J,,(z).
and

by Bourget;

the

reader

should

have

no

in provingthat
difficulty

(8)
Other

(9)
(10)

2/Va^)
formulae

recurrence

J,,,+,
{z)
=

Jn-,,ic{z)Jn+,A').

(due

(z)
^ j:,,,^,

to

Bourget and

^J^"}1

Giuliani

(^)},
{^,^_,
/,^^,_,
, (^)
_

4/'Vt_,(z) /".,(z) 4/",,_3


{z).
=

are
respectively)

10-31]
The

ASSOCIATED

differential

FUNCTIONS

equationis most

327

simply constructed

the method

by

used

Giuliani ; thus

by

^n Jn,k(^)

f'"d

{- (n +

IT

;Tn
dd

2k

in

and

cos

(z)-"
IkzJ'n^j,

z^\u

^ [ \jQ {nd

{z)+
2kzJ'n^k

6) sin {nd

cos

sin

6)](2 cos df

dd

Of-' sin OdO

6) (2 cos

6) sin {jid

cos

IT

{nd

cos

sin

sin

6)1
(2

df-^ sinj^f/^

cos

6')
;^ {(2cos

sin d]dd,
^)^-^

so

V, Jn,k{Z)

2kzJ'n" {Z) kK/,,_k


(Z)+

4.k (k

1) /,,,_,
(z).

d^
this

Operatingon

equationby

(^,l)IV, ./.,,(z)
+

and

hence

(11)

we

2^

was

1, and

using(10),it
k

Giuliani's

follows that

(k

1) /,,,(z),

equation

+ (/^+ 2)^ n^}


(z)+ {'2z'
J'\, (z)+ {2k + 5) ^J"'",,
J"n^,{z)
-

(2^"+ 5) zJ'n^k
{z)+ {z'+ k

It

(z)+ k^Jn,k
{z)\
2kzJ'"^k

have

;7^,

also observed

by

Giuliani

'2-

{z)
n')J^,^.

0.

that

S e,",Ln,k{z)
2nd
cos
(12) e'-"i""(2cos6?)^-=
rt

this is verified

by applying Fourier's

i S

rule

e.n+iJ2n+i,k
{z)sin {2n + l)d;

(cf." 2'2) to the function

on

the

right.
A

somewhat

(1883),col.

The

most

"

similar
8.

n^N

^^^^

function

it follows

J{z;

V,

by Bruhns,

Jc)has been studied


is defined by the series

important property

(14)
whence

This

function

{z;

v,

of this function

k)-J{z;

V,

J^+l)

is that

j^^^^^f^:^^ff^f:^y

that
//
"^^''

z^_

Astr. Nach.

"2vJ^^.2m{z)

''''^^-,"?,(v
+ 2".-2)(.

+ 2m

2)'

civ.

328

THEORY

10*4.

OF

BESSEL

[CHAP. X

FUNCTIONS

definition
of Struve's function'H.^{z).

The

completelyexamined the functions defined by integrals


it is natural to investigate
defined by
a function
resemblingBessel's integral,
an
integralresemblingPoisson's integral.This function is called Struve's
functions of this
function, although Struve investigated*
only the sjDecial
of the generalfunction have
type of orders zero and unity. The properties
been examined
at some
lengthby Siemonf and by J. Walker;):.
Now

that

have

we

Struve's function

H^ (z),of order

v,

is defined

2(izy
r(i;+ i)r(|).io

providedthat R(v)

"

by

the

equations

sm(zcosO)sin'''Odd,

-h.

similar
By analysis

that of

to

(1^)"

" 3'3,we

have

(-)"*227ft+".m!
r(i),"=o(2m + l)!r(^ + m + f)'

that

so

(2)

H,(^)=

The

R(v)

function

exceeds

"

H^ (z)is defined

this

It is evident

not.

or

by

It is easy to

all values of

v,

whether

function
integral
resultingexpressionis also an

that

the
and, if the factor (^z)"be suppressed,
function of

equationfor
H^(z) is an

of

tegral
in-

z.

[cf ""2-11 (5),3-121 (1)]that

see

^'^

"'W=IWPT?Ti)"'^'*'

where

(4)
and
*

1 1is the smallest of the numbers

I^0

Mem.

de VAcad.
Imp. den Sci. de St Petersbourg, (7) xxx.
(1882),no. 8; Ann. der Physik,
und
1016.
See also Lommel,
Archiv
der Math,
(1882), pp. 1008
(1861),
Fhijs.xxxvi.

(3)XVII.

+ fl,li'
+ li,
+ f|,li'
|i/

"

"

p. 399.

t Prograinm, Luisenschule
pp.

X The
this

Berlin, 1890.

[Jahrbuch ilber die Fortschritte

der

Math.

1890,

340"342.]

AnalyticalTheory of Light (Cambridge, 1904), pp.


with the exception of (3),
section,
(4),(10)and (11),are

892

"

there

895.

The

given.

results contained

in

10-4]
We

and

can

obtain

thus

formulae

recurrence

329

FUNCTIONS

ASSOCIATED

similarly
^
c?^

b-H
^

"

(-)-(2m + l)^^
^^to2''+-^-+ir (m + f) r (v +

(5V=
^
^^

^)

/_y"+i22m+2

"

1^ (m
^,""_i2'^+^+-^

I) r (i;+

1)

l)r(i)"^"H.+iC^

~2''r(z.+

H,_. (.). H,"

(6)

H._.(.)-H."(.)

(.)+

2H,^,(2),

have

we
particular

^ {m, (^)]

(9)

J^llV-^^^
2H/,.)-p^(i|]l^^.

=.

(a + WH,(")

(7)

^ H.

(.)

(5)

In

find that

comparing these results,we

On

zU, (z),

{Ho(^)}
=

C^-^ V-')H, (2)


-

v) |^H,_i(z)}

(^

z('^-v

l)II,_,{z)

to

V,H,(^)

p. 218.

been

This

bears

L^ (z)which

function

J^(s)has

studied

function

(in the

is defined

L. ^^ (2)=

(11)
'

should

reader

case

by

the

^(^'+ i)^(|)"

have

to

Struve's

function

Journal^
0) by* Nicholson, Qxiarterly

XLn.

See

also

(1911),

equation

#)r(:.+m

+ A) r (1) / 0
(,.

in
difficulty

f)

^
"'^"''''
"'^^"^^'''*^'* ^-^

^^'

"
{%")
"\.

obtaining the

fundamental

propertiesof

function.
*

/^ (z)bears

as

*^^^

no

relation

same

being valid only when

integralformula
The

the

",=or(w

the

^"""'*^^^'

(z)satisfies the differential equation

(10)
The

(^).

H,

r(i.+ |)r(i)
that H^

have

Again,from (7) and (8),we

so

Gubler, Zurich

xlvii.
(1902),p.
I'ierteljahrsschrijt,

421.

this

330
10*41.

The

It

noticed

[CHAP. X

FUNCTIONS

BESSEL

OF

THEORY

for H^(^).
loop-integral

" 10*4 that the integraldefinition of H^ {z) fails when


limit. We
does not converge
at the upper
R{v)": ^, because the integral
avoid this disability
in placeof the definite
a loop-integral
can
by considering
integral.
was

in

"

Let

take

us

(f

ly-^-sin

zt

dt,

Jo
where

the

phase of

the contour
t

f^

l vanishes

"

the

at

the real axis,and

crosses

the

pointon

rightof

does

the contour

1 at which

enclose

not

the

point

-l.

If

"

r(i+)
-

dt

zt

2i

cos

the

phase of

(1)

^^ is

"

R(v)"

H.

zt
.

dt,

^^

{z)

have

^, we

"

"

;Vil
(2 )
1

"f' ^y-"-"i"

'i

dt-

'

"' 0

equationare analyticfunctions of v for all* values of v ;


generaltheoryof analyticcontinuation,equation(1) holds for

sides of this

by

the

all values of
From

v.

this result,combined

(2)
To

t-y-^sin

zero.

in

so,

Jo

Hence, when

Both

(1

VTT

Jo
where

ment
seg-

ly-i sin

{t-

and

into the

\, we may deform the contour


suppose that Ti{v)"
(0,1) of the real axis,taken twice, and we find that
we

" 6"1 (6),we

deduce

that

{z)=^^^ ;;Vi\-'^e-' {t^

j^{z)+ iH,
transform

with

this result,let

be

"

any

1 )"'-dt.

or
angle(positive
negative),

acute

We
then deform
+ w.
^tt-f "" and |-7r
the four parallel
lines
the contour
into that shewn
in Fig. 26, in which
make
with the imaginaryaxis. It is evident that, as the lines
an
angle
the
real
axis
the integrals
to
off to infinity,
move
parallel
along them tend to
The integral
zero.
alongthe path which starts from and returns to 1 + so ie~''"
is equal to i/^w {z); and on the lines through the originwe
write t
iu, so

and

let the

phase
"

of

lie between

"

"

that

on

them

(f- 1)"-*
=

eT {"'-h)W

(1 + ^c^y-h,

It follows that

/, (z)+ in, {z)

The

isolated

undetermined

values

form.

i/.a)(^)+

|,

|,

...

are

jT^^y^p
^^^j^
excepted,because

the

e-'

(1 + u^-^ du,

expressionon

the

rightis

then

an

332

OF

THEORY

If

write

we

ix in

(3),where

BESSEL

we
positive,

is

deduce

and, by considering imaginary parts,we

which

R{v)"\,

that

sin

du
{xu) (1 +i{2)''-2
.

jo

r(i/+*)r(i)

given by Nicholson, QuarterlyJournal,

result

that,when

see

{W

lty{x)=-I_^{x)-

(6)

[chap. X

FUNCTIONS

(1911),p. 219, in

XLii.

the

specialcase

in

0.

Tlie

10*42.

asyni])totic
expansionof

jz \is large.

lating
asymptoticexpansionwhich may be used for tabuthe argument z is large,
the order v being fixed.
Struve's function when
Since the corresponding
asymptoticexpansionof V^,(z)has been completely
in Chapter Vil, it follows from | 10"41 (4) that it is sufficient to
investigated
determine
the asymptoticexpansionof
We

As

shall

in

now

obtain

{z)when

H^

an

^ 7 '2,we

have

^2sv-J

P^^i-y^'.ik-vXr.U

(-1)

am

mlz''

uHy-p-^
r
f^5^//-')-^?)

(-)^.a-^)pw^
+

(pWe

take p
for which

so

largethat

R (v

"

h)^0, and take S

"

S, Iargi:
|/3|-$^7r"
so

that

is confined

then

yS|$ Jtt S,

"

"

^7r

"

28.

have
1 +

so

be any

plane for which

2S ^ arg

TT

"

We

of the

the sector

to

to

ill\/t\,
1! ^

sin

"

5V

6,

arg

{I "

ill\Jt
"'rr.

that
1 +

say, where

~Y

I
^

A^ is independentof

It follows

on

e^-i/WI

(sin8)2i2W-2p-i A^,
=

z.

that
integration

/"
J

-^

7)1

mlz^

ao

where

^p

-^

"1^--

0 (^-2^).

exp ip

e-"

c?w
it'^P

dt.

positive
angle

10-42, 10-43]
We

{arg z\"Tr

that, when

deduce

providedthat R{p"

333

FUNCTIONS

ASSOCIATED

and

in

-[-\)^0;but, as

is

|^

large,

" 7'2, this

last restriction may

be

removed.
This

asymptoticexpansionmay

'

p,

^1

7r"i=ol (^ + 1

"

in the form

"-

^\

4-

'

Tm

F,(^)+

V (m

/'-I

H,(^)

(2)

also be written

"+1
v^n.-.+i

w)(^2-y"*

Q(^'^~^M-

the
that, if is real and z is positive,
difficulty
in the asymptoticexpansionis of the same
terms
after
sign
numericallyless than the first term
neglected,provided that
as, and
established
by the method used in " 7-32.
'R{'p-\-\"v)'^^.This may be
It may
remainder

be

proved without

'p

asymptoticexpausion* was given by Kayleigh,Proc. London Math. Soc. xix. (1888),


case
0, by Struve, Mem. de VAcad.
Imp. des Sci. de St Petershourg,
(7)
and
Ann.
und
der
Chemiey
-p.
no.
101,
XXX.
(.3)xvii. (1882),p. 1012 in the
Phys.
(1882),
8,
was
case
\; the result for general values of
given by J. Walker, The Analytical
Theory of Light (Cambridge, 1904), pp. 394"395.
The

in the

p. 504

i'

has any of the values ^, %,


terminatingseries and Y^ {z)is also
If

that, when

is half of

...,

then

is expressible
(1 + u-/^-)""as

in a finite form.
It follows
expressible
odd positive
in terms
H^ {z) is exjDressible
integer,
In particular

an

of elementaryfunctions.

[Hj(^)=(-)"(!^),
cos

^^^

l-r,r
.X

10"43.

^^^

2\

/2\V

"

COS^

order.
asymptoticexpansionof Sti'uve'sfunctions
of lai-ge

The

obtain

asymptoticexpansions,of a type similar to the


in Chapter viil, which
expansionsinvestigated
represent Struve's function
H^ {z) when j \and |z \are both large.
shall

We

now

As

usual, we

shall write
V

cosh

(a + t/3)
=

cosh 7

shall confine the investigation


to the special
case
we
and,x^r simplicity,
cosh 7 is real and positive.The more
by the methods
complex may be investigated

which

of

no

see

generalcase
" 8'6

used in

importanceand it involves
great practical
For

an

asymptotic expansion of

Kayleigh,Phil. Mag. (6)viii.

the associated

(1904),pp.

481"487.

some

in

in which
and

cosh 7 is
" 10' 15, but it is

rather intricate

analysis.

integral

Papers, v. (1912),
[Scientific
pp.

20G"

211.]

334

THEORY

The

method

t3"pe,and

steepestdescents has

of

not,

be

to

" 10"41 (3),we

of

of the formula

In view

consider

of
integral

an

of Bessel's

the

Poisson's

type.

integral

dw

L-^z(l
+u'-y
which

[CHAP. X

FUNCTIONS

appliedto
to one
previousinvestigations,

in the

as

BESSEL

OF

write in the form

we

div

f'
"ZT

where

cosh 7

that

It is evident
so

log(1 + tv^).

r,

function

qua

of

has

w,

stationary
pointswhere

two
7 is equaleither to a or to lyS,
points
give rise to the stationary

that, since

which

have

cases

e-y,

be considered,

to

(I) e"", (II)e""^.


consider

Accordinglywe

zjvis greater than

(II)in which
(I) When
increases

is

from

real

to

oc
,

decreases

then

the
separately

to e"^

In order to obtain

a,

first increases

cosh

"

which

zjvis less

is real when

from

0 to e~"

"

1, and

is real,and, as

cosh

increases
log(1 + e^")and finally

than

1.

number
positive

(I)in

cases

alongwhich

iv

o.log(i -\-e~^),
to

ao

increases,
we
continually
suppose
the
that w
moves
originto the pointe~'^,and
leaves the real axis at right
then starts moving along a certain curve, which
and increasing.
which t is positive
on
angles,
To

variables

in the

it has
We

now

double

sinh ^,

on
^-plane,

cosh
and

of this curve,

^ sin

1^+

?;

which

2 cosh

make

6""

^'7"

sinh

^0,

is real,has for its

(tanh ^ tan 1;),

tan

arc

equation

write
2

arc

tan

J,

(tanh ^ tan 77)

cosh ^ sin
examine

the values of

0, A, B, C have

^goes from

increases from

i^(f,r})as ^

traces

-t]

out

the

whose
rectangle

corners

complex coordinates
0,

As

to

point*at ^o-

"

and

it is convenient

real.

^, rj and ^0 are
curve

by writing
w

The

destination

find the ultimate

change of

where

contour

along the real axis from

first

arc

0 to A,

sinh 1,

arc

sinh 1

F{^, t])\s,
equalto

| tti,

2 sinh

^iri.

f,and
|/cosh'-

0 to 1.
*

Except

when

0, in which

case

it has

triplepoint.

this

steadily

10-43]
When

^ is

this

increases
steadily
establish

To

Note.

this

U(l+2fi)
r~

1
"

dt
t+

because

2t^
5-

^j

When
from
is

tt

as

cosec

tan

r]

rj

observe

^ *"'

2fi (2 + fi)

derivate
t,has the positive

C;

to

''"

"'"

with

^tt.

\t+7^

vanishes

0 to

that

(t + 2fi

t^Tiu^)

77,

increases from

J2 and

^ goes from

7r/\/2as

1 to

\
^
'^''

^''

"

7^^
V \/2

BG, F(^, 77)is equal to

^is on

to
irls^'I

tan

arc

result,write

t, which

tan

arc

from

335

AB, F (^,??)is equalto

on

\/2
and

FUNCTIONS

ASSOCIATED

sech

ir

^,and this increases steadily


when
^ is on CO, F {^,-q)
finally

and

zero.

the curve,

Hence

on

part of the

inside the

curve

{^,77)is equalto sech

except at the double

rectangleOABC,

the

which

rectangle

must

a, cannot

pointon

the side OA

pass from

this double

emerge
; and

from
so

point

to

the
the

singularpointC.
The

contours

and

broken

in

the

continuous

for
2t'-plane

curves

which

has

the

0, i

values

shewn

are

in

Fig.27 by

respectively.

Fig. 27.

Cdnsequentlya

in the

contour

part

of the real axis

the

singularpoint i; and,

0 to +

joiningthe
as

on
tf-plane,
to
e""
origin

traces

out

which
and

is real,consists of the
from

curve

this contour,

00
.

It follows that, if the

expansionof d^/ch in

^i=
ar

"i=o

r'"

powers

of

is

this

pointto

increases

ft'om

336

THEORY

BESSEL

OF

[CHAP. X

FUNCTIONS

then

o"j

and

10-4

hence, by

have

(1),we
/

XT

(1)

"

(II) When
0 to

CO

62

2 cosh 7,

6i

-^--^^"

V
S

6 cosh- 7

"

63

20 cosh^ 7

4 cosh 7,

"

a
pure imaginary(= i^), t is real and increases
travels alongthe real axis from 0 to 00 ; and so

";
.-

Hence, from

''^"^

2(*^)''

is

7
as

that

It is easy to prove

bo= I,

r
\
H,(^)~-i/,(^)+p"

/-.x

from

(1

dn,
".)'-"

steadily

/;.-"
[^^^ *"|
4r.

(3) it follows that

" 1041

i ^'^H.(.)~F.(^)+^//^/";,,,

(2)

providedthat jarg

\" ^tt. This

result

of values of arg z, after the

domain
From

results
corresponding

the

expectedthat

these results

be extended

can

of

manner

in the

somewhat

wider

" 8"42.

theoryof

valid for suitable

are

to

Bessel functions,it is to be

domains

of

complex values

of the arguments.
In

we
can
pcirticular,

that,in the

prove

Ii"(vx)

(3)
when

1 1is large,|arg
1-

expressionon

[Note.
should

we

that

we

get

10"44.
When

"

lO'l

is

If in

have
a

(I)we

obtained

case

"

right.

had

taken

The relation between


the

order

?;

is

two

as

from

magnitude

of

..,^
^^
''""^"^*

w"0

crosses

H,i (z)and

of

to

?"; =

e~" and

placeof

the line

thence

iJy(2). This

to

w=

"

i,

indicates

/3=0.]

E" (z).

positiveinteger(or zero),Ave can deduce from


H" (z)by a polynomialin z; and when n
functions differ by a polynomialiu 1/z.

(4) that E,i (z)differs from

the
negativeinteger,

is of the order

containingiJy (2)in

phenomenon

purelyimaginary argument,

.,,,7

.^

the contour

the formula

error

^''i''

"

of

I^ (vx)

fixed,and the

"

the

of Stokes'

of functions

ri+j(i+x^)

the

is

x
|" ^tt,

^
times

case

"

10-44, 10-45]
For, when

is

positive
integeror

337

FUNCTIONS

ASSOCIATED

have

we

zero,

r{hu

\)T{\m

\)

'

and
CO

"

"=o

and therefore,since

Jn(z)

,"

{\m

(1

1)r

pn) r (?i+

(i//i
+

1)

'

have

Jn(z),we

\"mri (l"\n+m

"

|m)

that is to say

(1)

E,.

In like manner,
=

??

"

L-)"ft

T/ie si^n of Str live

10'45.
We

when

E_. (.)

(2)

{s)=l%
ii^""^Ml"L^H,. (.).

shall

is

negativeinteger,

H-" (4
^"'-"-^Vtf""'"'
-

function.

prove the

result that lrl^(x)


is positive
when a; is
interesting
and
has
v
alue
than
This
to
or
positive v
result,
greater
equal ^.
any positive
which was
1, is derivable from a
pointedout by Struve* in the case
definite integral
(which will be established in " 13'47) which is of considerable
importancein the Theory of Diffraction.
now

i^

To obtain the result

parts and then

we

see

by

elementajymethod, we integrate
" 10'4(1)by
that,for values of exceeding|,
an

iw"
cos
cos
(.a;
6) sin-""^6
V{v + h)V{\) \i

(2z; 1)

'

/=

^^"^'fx
1
1

(2^' 1)

n
/
/ 1 X
(^ + 2)^ (i) I

4^

"

tvWtt

COS

['"cos
{x
cos

{x cos d) sin-"--^ cos

d) sin-"--^ cos

Odd]

h
sin^''--'6'
^ 11
cos

ddd"^

J
-

cos

{x cos 6)}dO

^0,
since the
*

Mem.

integrandis positive.
de

VAcad.

Imp. des Sci.

proof given here is the natural


w.

B. F.

de

St

extension

Petersbourg, (7)xxx.
of Struve's

no.
(188'2),
8, pp.

100"101.

The

proof.
22

338

OF

THEORY

When

be

cannot
h, the partialintegration

is less than

[CHAP. X

FUNCTIONS

BESSEL

performed ; and, when

|,we

have

(^.)*(l-cos.^')^0,

Hi(.^")
=

is

the theorem

and

completelyestablished.

proved in " 10"42 with that of


expansion which was
comparison of the asj'mptotic
H^ {x) is positive
largeand positive,
Y^ (,v)
given in " 7"21 shews that, tc/ien x is sufficientiji
that Hk (.r)is not one-signedwhen
term
of the
if
" ^ ; for the dominant
" 3 and
of
is
H^
{x)
asymptoticexpansion
A

j'

1/

^
(
I sin (x
'
\lTXj

or

accordingas

i/

"

or

i^

The

"

theorem

function
is positivefor
sufficiently
largevalues.

functions

of x,

values

for the

an

essential

positivevalues

difference

between

i and

"

i/

Struve's

asymptotic expansions of Chapter vii


Y^ (x)are not of constant sign.

extended

more

function

result

not

merely

and

Bessel

that, for sufficiently


large

shew

J^ {x)and

take the

we

proves the
of x when

integral.
Theisinger's

10 "46.
If

indicates

theorem

The

of this section

all

i-n)

that Struve's
for

Ivn

"

"

r(.'+*)r(i)

equation

7r2

/"in-

J -hn
f

/"*

r.

JUi'piQ

te'e

A-

1\"1

l + iz dz
'

IZ

and

the

choose
+itau

z=.

be

the

imaginary axis,indented

the

at

then

write

di"

pT

"

{h (^) I"oix))
=

+ *0)
{x cot ^) logtan (iir

cos

-^^

so
"

/o (^) Lo i-v)="i

(1 )
a

origin* and

find that

1^, we
TT

and

to

contour

formula
If

order

cos

given by Theisinger,MonatsheftefUr

^^

{x tan "/")
logcot {h"i")
Math,

und

Phys. xxiv.

(1913),p.

341.

we
find,on changing
replacex by .vsiu 6, multiply by sin 6, and integrate,
i
n
the
the
on
integrations
right,
absolutely
convergent integral

we

the

of the

-^ | f

'"

I
so

El {x tan "^)logcot (|0)

(^ (^ "^" ^) ^o (-^sin 6)}sin


'

6de

that

'"

(2)

El {x tan 0) logcot (*0)

cos

expanding the integrandon


Theisinger.

on

The

presence

indentation.

of the

-^

the

rightin

logarithmic factor

powers

ensures

of

the

(f)
(p
x.

f
2'

\-e-'
'

This

curious

convergence

result is also due to

of the

integralround

tlie

340

is thus

function

The

-uUr L

/cos
_

equalto

/,

-vw)

'"'

'""
"""^'-'^'^

t=l,

I r(m-i/
i^Try

(m
(^^) 2-"/r
jlog

and

be

"- '"*
^"""".^-'

that*, near

it is known

Now

to

seen

[CHAP. X

FUNCTIONS

BESSEL

OF

THEORY

1) + A/r
(m

|)r(m

i/

i;

i) /I

ty

1) + Vr(m

1)1
,

since
+ i)
"'-"\,-.n-.
/i-^r^^^^''^-^r(/.

2/ii^M+l

"^",-"..-"fL^M'",,J'-'r(M+i)
i:
V
2

obtain

we

the

"2'^^'^+i

expansion
asymptotic
iiT,"'(^)

(7) W, (^)

GC

+
TT

V(27r2^)
+

functions

Some

which

(m
yfr

^ + v)

"

^iri]

yfr{m+J)-\og2z"

satisfyequations of the same


general type as (1) have
of the Coll. of Sci. Imp. Univ. Japan, iv. (1891),p. 310.

been

noticed by Nagaoka, Journal

The

10'6.

functions
composingYn{z).

reader will remember

The

order,may
integral
IX

be written

that the Bessel function


in the form

'""-^(n-m-\)\

of the second

("3"52)

'

,,

!)-"/.(.+
I^'"g(i^)-V'(''"+
+.!"Jid'+l)!
The
which

series
has

the

rightmay be expressedas the sum


thus
fairly
simplerecurrence
properties,
on

Cf.

kind, of

Barnes, QuarterlyJournal,

xxxix.

",.

1)1.

of four functions,each

(1908),p.

111.

of

10-6]

ASSOCIATED

341

FUNCTIONS

where

(2)
m

and

(of." 3-.582)

(-rd^f^"""

Un{z)=

(3)

The
pp. 142

functions
"

The

due

are
investigations

to

studied

by Schlafli,Math.

Tn {z)

Otti*

to Gi*af and

and

["(i
TT-d)

indicated

establish this result,observe

(z sin

sin

Ann.

by

the

ill.

(1871),

equations

Gublert.

by
simply represented

Tn{z) is most

function

(4)
To

been

T'n(2) ^i-nd Un{z) have

147, though he used the sHghtly different notation

recent

more

+ l)--"/r(l)}
{i/r(/|+w

?u!(n + "i)!

=0

the definite

integral

nO) d9.

that

(-y"(|2)"+-"*

Tn{z)

de

"i"_j"_j

{m

+
e) r (?i

1 +

e)Je=,
-r^y^

1
27rt

./_i

(/i+ 2m)!
27n\^".';"_i
2

TT

'7^^

where

t has been

e"'9(-i" sin

2,

V^+'^^

e=0

/"'"+"
(-)""(i^y^+--"^
(1 + tY+'''^
log^

j_i

(n + 2m)!
Lae,""_j,,_j

Jo

,.

CW,

"

(n + 27H)!

m"-irt-i

replacedby

^f +"-'".(^-Itt)

;;

"^^

f'^^^

eC-^'-'^ii.

It follows that

Now

(-

iz sin

^)"+2'"

(- iz sin ^)
jcosh

(n even)
(n odd)

_
~

(h

,^"_j"_j

and

2?/i)!

[sinh(-

iz sin

6)

so

TTl

Bern Mittheilungen,
1898, pp. 1"56.
Einleitimgin die Theorie der BesseVscken Funktionen,
treatise,
87, should also be consulted.
pp. 77
t

"

11.

(Bern,1900),pp.

42

"

09.

Loramel's

342
If 6 is
second

of

obtained
The

"

at

once.

for Un (z)is obtained by observingthat


integral
corresponding

r (1 +
~d_I (-)'^(1^)"+^"^

_9e,"=o

(o)

from

p-"w

-rr^

7/i

e)

"=0

e)j-,+,(^)}

that

sin

(n(9
-

sin

formulaefor Tn{z) and

Recurr-ence

" 10"6 (4)we

[^e-""-^smhf^^,

^) rf^ + (-)"

Jo

TT

From

(n +

6)

fiog(i^)-t(i)K"W
+

10*61.

1 +

"=0

deduce

" 6*2 (4),we

!r

Ki^)-^r(i

ae
so,

[chap. X

FUNCTIONS

obtained by considering
^ in the integral
only the
replacedby tt
is
is due to Schlafli,
the formula (4),which
the two
exponentials,

Un {Z)

and

BESSEL

OF

THEORY

(Jn(z).

that

see

Tn-^{z)+Tn+^{z)-(27l/z)Tn(z)
=

sin (zsin ^
(Itt (9)
-

n^) {2cos

2n/z]dO

r (hir-6)^{cos(zsin d-ne)}de

-"

TTZ

civ

"

4
=

COS^

Jn (z),
^

??7r

integrating
by parts and usingBessel's integral.

on

Thus

(1)

T^ {z)+
Tn-. {z)+ 7^,,+,
{z) (2n/2)
=

Again
and

T"' (^)
=

/"'^

I (Itt

"

"

^)sin

Invr J" {z)]\z.


{cos^
-

{z sin ^

cos

nd)(^5*,

"

so

(2)
From

it follows that

these formulae

(3)

(SV+ n) Tn {Z)

(4)

(^

and

hence

(5)

"

(cf.
" 1012)

) Tn {z)

zTn-, (Z)

2 COS^

-z

Tn+,{z)+

IITT

2 cos^ * nTT

2 J"

(z),

2/" (^),

find that

we

V,jTn {z)
=

[zsin2 1

?i7r

cos^

Iwtt}
-

4h

Jn (z).

10-61, 10-62]
With
for Jn

ASSOCIATED

the aid of these formulae

{z),Y" {z)and

combined

deduce

Sn [z),we

343

FUNCTIONS

with

the

" 10"6

from

formulae
corresponding

(1)that

(6)

Un-, {z)+ Un^,{z)

{2n/z)Un (z) (2/z)J,,(z),

(7)

f^"_:(Z)

Un+,{Z)

(8)

(^ + n) Un {z)

(9)^

(^-^0

U,(z)

V,,Un(z)

reader

may

to

T_n (z)

-zUn^,(z),

-2zJ,^,(z).
the definition,
" 10'6

(3).

T_n(z),of negativeorder,by the


find that
^ in the integral
we
replace^ by

we

tt

f"(1
7^

"

^) sin (z sin

^) sin (^ sin

(ivr

3-58(2)]
[of.""3-58(1),

the function

define

of " 10"6 (4). If


equivalent

{21Z)J, {z),

from
verifythese directly

It is convenient

and

U,: (Z)

zUn-. [z)+ 2J, {z),

(10)
The

"

nO) cW

nO }"

"

inr)dd,

so

T_,,{z) (-r+^TA^).

(11)
We

define

now

it is then

found

i-Y {Un (Z)

t7_,{Z)

Series

shall

We

now

for Tn{z) and


shew

how

Tu (Z)+ Sn (z)].

Un{z).

to derive

the

expansion

Tn{z)= t -{Ju+2nAz)-Jn-,,niz)]

(1)

from

of

that

(12)

10-62.

hold for all values

U^niz) by supposhig" 10"6 (1) to

" 10-6 (4). The

method

"''

which

shall

we

is to substitute

use

sin 2m0

^---0=^
TT

integralfor Tn (z),and then integrateterm-by-term.This procedure


since the Fourier series does not converge uniformlynear
justification,
^d 6
justquoted is untrue for these two
iT, and, in fact,the equation

in the
needs
^

values of 0.
To
series

justifythe process*,let
converges uniformly when

S and

S ^ ^ ^

be
tt

small positivenumbex's.
arbitrarily

S, we

"

^^

{\7T-6)The

pp. 257"

analysisimmediately following is
-262. The

value

of the constant

due

an

integer mo

Since

the

that

2m(9

sm

m=i

find

can

such

"f,
m

to D.

Jackson, Palermo

is 1-8519...

Rendiconti,

xxxii.

(1911),

344

OF

THEORY

throughout the
of 6 between

8 ^ 6 ^

range

0 and

-M^

2m0

sin

,
=

'

{l+2cos2^

Jrr

"^

sni

the

2cos2i"i!}o?^

f(M+h)^

07

Jtt I

ao;

./(23f+l).f.

'

the second

^tt,by

sin

dt=

theorem, since tlsint is

mean-value

it is easy

of .r~isinjp

graph

to

that

see

the

last

expressioncannot

'-

"

jo

...

function.
(increasing)

By drawing
exceed

6 and

0 between

"

] ^

value of

SUl

[h^s\n('2,M+\)t
^

some

g
(

iTT ;

monotonic

+ 2cos4^

t
|''i'rsin(2i/+l)"
dt

all values

"l

m=l

for

Again, for

8, for all values of M exceedingWq-

"

have

we

tt,

ir

[CHAP. X

FUNCTIONS

BESSEL

dx in absolute

value

if this be called hnA.

we

have

Tn{z)

I
J0

2
^

m=i

sm(2Sin^-"^)rf^
"*

sin 2?u^l

-^^
/TJ(i:r-^)[
2

'

[sin
(zsine-

nd)dd

"i

m=l

sin
U

J TT-Sj 11'

(2

sin 6

"

n6)\d6

"

m=l

"-{7r.48 + (7r-2S)f}^,
where

is the upper

Since

(2J.8

4-

bound

of

|sin {zsin 6

n6) |

small, it follows from the definition of


e) 5 is arbitrarily

an

infinite series that*

I r 'i'^-?^
sin (zsin

Tn {z)=-

and

nd) cW

Wn+2m (-3")Jn-2m \^)h


"

I3r

the result is established.


It will be remembered

series of Bessel

and
as

that, in
power

10"63.

Uniz) has alreadybeen


coefiicients by the equation

" 3-582, this

series

that

definition

given in

" 10 Q

t)=
m=

This

expansion was

(z)

(3).

series

of Bessel coefficients.

expansion

Tn(z

(1)

("3'581) as

identified with the definition of Un

was

Graf'sexpansionof T^ {z+ t) as

It is easy to obtain the

defined

discovered

Tn-,,{t)J^\z),

"00

by Schlafli,Math.

Aitn.

iii.

(1871),p.

146.

10-63, 10-7]

that

" 10-6(4),it is evident

for,from

T;^(^+ ^)

["(
1

TT

TTl

sin

6) sin {tsin e-nd-\-z

345

FUNCTIONS

ASSOCIATED

j/i

0) dd

cc

I ./,"(^)sin (^sin ^^["(1,^-^)


j

TT

by using " 2"1

(71-m)l9)r/^,

CO

series under

since the

the order

and

of summation

vergent,
integralsign is uniformlyconbe changed,and the
integration
may

the

result is evident.
The

proof of the formula


the
; it depends on

There

pp. 425

444, as

"

includes

as

(1)
where
this

V,y
k and

kz"-+\

(^ + 1).

+ \r-- v^-]
[{f,
{(^+ 3)^ .^j

,r-

(-)-(iy"+--^

=kz^-^

^,

+ l)r(l^
I (-)"n(l^)em+2r(l;.-l^

,_i

,"to V{l^Ji-\v+ m+^,)^{\^l-^\v


the

rightare

is evidentlyundefined
s^^^{z)
odd negativeinteger*.Apart from
(1) is evidently

an

solution of

(3)

2/

In like

manner

the

|. + |)
+

^)

in the form

written

function

is

on
brevitythe expressions

The

integralof
particular
with
is
z'^^'^,
beginning

that

of

equation,
proceedingin ascendingpowers

/x +

It is easy to shew

constants.

/* are

(2)

For

S^^y(z)and s^^t,{z).
functions

specialcases the polynomialszOn {z) and


and Schlafli,
derived by Loramel, Math. Ann. ix. (1876),
was
of the equation
integral
particular

function,which

8n {z)of Neumann

plicated
com-

more

{z+ t).
equallysimple expressionfor "/""

no

genesisof LommeV

The

10"7.
A

to be

seems

given by Graf, Math. Ann. xliii. (1893),p. 141, is


with " 2-4.
of the series of " 10-62 combined

use

either of the

when

this restriction the

numbers

general

+ A-S,,,(0).
'^,(2r)

generalsolution

of

IS

y-z-'-^''-'Hr^,{z)+ks,,A^))-

(5)
*

The

solution

of the

equation for such

values

of

/x

and

is discussed

in

" 10-71.

346

THEORY

Next

let

us

consider

the

assume

as

parameters."We
where

(z)and

BESSEL

OF

solution of (1)

of

determined

J,{z)A'(z)+ J_A^)B'(z)

that

see

;7r

z'^ J_^

sm

Hence

(6)
where

i/

of

B{z)

z"^ Jy

2 sin

sm

{")dz.

vTT

(1) is

J^ {z)

z^

J_^ {z)dz

J_^ {z)

J^ {z)dz

z^

vK

the lower limits of the

a
Similarly

{z)dz,

v-TT

solution

equations

kz^-\

lTT

A(^)

of

0,

using" 3" 12 (2),we

the

by

J',(z)A' (z)+ J'_, {z)B' {z)


On

of "variation

method

the

by

solution*

functions

{z)are

[chap. X

FUNCTIONS

are
arbitrary.
integrals

of (1) which

solution

is valid for all values

of

v,

whether

integersor not, is
(7)

i^'T^ F, {z)

z^

J^ {z)dz

./,{z)

z^

Y^ (s)dz

real
It is easy to see that, if both of the numbers
positive
/x + i^ + 1 have
If we
expand
parts,the lower limits in (6) and (7) may be taken to be zero.
the

integrandsin ascendingpowers
as a
rightin (6) is expressible
power
than

"

/m "

is

odd

an

Ave

the

that

see

containingno

we
integer,
negative

the

expressionon

of z
powers
neither
(3),it follows that, since

series

Hence, from

z^'^^,
z^^^,z^'^'^,

numbers

of z,

other

of the

have

must

TT

(8) V..(^)

J^ {z)

In

(9)
and

in this formula
It should

(10)

(z)dz

zi^ Jv

J.

was

supposed that

of the second

l7r

we

{z)dz

Jo

result it

functions

s^^Az)

z'' /.^

VTT

obtainingthis

introduce

we

2 sin

Y^ {z)
may

be observed

z^

kind,

we

J^ {z)dz

proceedto

is not

integer;but

an

if

find that

J^ {z)

the limit in

that,in Pochhammer's

z^

Y^ (^)dz

making

notation

an

integer.

("4"4),

V^(^)-(^_^l)(^^^_^l)
+

x,2^2(l; \ii-\v^\,\ii^\v
*

Cf.

is not

Forsyth,Treatise
an

on

\\ -\z').

DifferentialEquations (1914),"66; it is supposed temporaiilythat

integer.

The

generaUsationof this result,obtained by replacingzt^+'^ in (1)by an arbitrary function


of 2, was
given by Chessin, Comptes Rendus, cxxxv.
(1902),pp. 678"679; and it was appliedby
him, Comptes Eendus, cxxxvi.
1126, to solve a sequence
of equations resembling
(1903),pp. 1124
f

"

Bessel's

equation.

348

THEORY

OF

BESSEL

[CHAP.X

FFNCTIONS

It will appear in " 10"75 that the series (1),by means


of which S^^"{z)
is
defined when
either of the numbers
is still
odd positive
integer,
/x + y is an
when
the numbers
odd positive
not
are
significance
integers.It
fx "v
in fact, an asymptoticexpansionof S^^y{z)valid for large values of
yields,
of

the variable

z.

Recurrence

10*72.

It is evident

formulae satisfied
hy LommeVs

functions.

" 10-7 (2) that

from

that is to say

Again,it is easy

verifythat

to

[z"Sm, {z)\={^l + v-\)z'' S^_i,

,_i

{z),

that

so

(2)
and

s'^ {z)+ {vjz)


s^^y(z) (/x+
=

1) s^_,,^_i (z),

v-

similarly

(3)
On

^(z) (fl-V-l)
sV. (z) (viz).9^,

and adding these


subtracting

results

{2v/z)
s^^^(z) (^ + v-l)

(4)

+ V-l)
2sV,.(^) {fJ'

(o)
The

S^i, ^+1

"

we

.9^_,,
,_,

obtain

(z).

the formulae

(z)-{fx-v-l)

S^_,,^.i{z)+ {fM-v-

s^,, ,+1

1) S^_i.^+i (z).

will find it easy to deduce from " 10"71 (2) that


the type s^"(z) may be replacedthroughout these formulae
reader

the type

"S^_(z);
^

the functions

of

functions

of

by

that

so

(6)

(z),

S,^,,(z)
=

z'^^^-{(^ + iy-

"^|,SV.(z),
.

0)

+
(Z)+ {viz)
S,,,(Z) {fji
S',,.

(8)

(z),
S'^^^z) (vjz)
S^^,(z) (fM-v-l) 5f^,.,+,

1) "^^_,._,
(z),

V-

{2vlz)S,^,(z)^(fi+v-l)S^,,^_,{z)-(fi-v-l)S^.,^,^,{z),

(9)

2S',,,(Z)(,M+ V-1) ^,_,,,.,


(Z).
(z)+ifjL-V- 1)"^^_,.+x

(10)
These
due

to

not

in all

formulae

Lommel,

"

(1) and (6). They are


of proving them were

completelysatisfactory.

cases

10' 73.

be transformed
in various ways by using
may
Math. Ann. ix. (1876),pp. 429
432, but his methods

LommeV

when
functionsS^^y(z)

fi

"v

is

an

odd

negative
integer.

undetermined
form when /j, v or fi + v
an
" 10"71 (2) assumes
define
is an
odd negativeinteger*.We
in terms
of
can
easily
'S^"_2p_i,v(^)
10'72
which
of
(6)
"
gives
'S"_i,^(^)
by a repeateduse
The

m
(1)

formula

"

.,x

9
^.-.p-^,Az)

(.yn,.-.p^.n
(-)PS^_,,^{Z)
"^^
+
_-^ 2.^^ (_^^^^^(^ _^)^^^ 2..^,(1 ^)^
"

Since S^, (z)is an


negativeinteger.
v

odd

even

function

of

v,

it is sufficient to consider

the

case

in which

fi-v

is

an

10-72,10-73]
We

lini

^"_i,{2)

(2)

of

limitingform

the

S^^i"(z)by

define

next

349

FUNCTIONS

ASSOCIATED

72

" 10

(6),namely
'

+ 1/ + 1)j
y +
1)(/U,
1_(/L6

"

"

^^^_l

The

(which

numerator

when

fx

1, and

v"

so,

function
analytic
theorem*
by L'Hospital's
is

of /x

an

near

/x

l)

"

vanishes

S^_,^^(z)^
=

Jfl

Z,V

'
.

P"1

verifythat

it is easy to

Now

IJ.=

\,Zo{m+

.-1

(2logz

iy.r(v+m

2)

+ y}r(v+l)-^|r
{m
^|r{\)

2)

(v +
yjr

2)}.

Also
d

+ f)r(l^
^{2'^+ir(l/.-l^

ii^+ |)cosl(^ + z0 7r}

y^i

M=i'-i
_

2"

r(i.+ 1) sin

ii. +

I) rCl/i+

vir

+l)
l^|r(v

|-f(1)+

{log2 +

^7rcotvir},

and

'd_

r (l/x
{2'^+!
-

d/jb
and

(/^ ^)77}

+ l),
2''-^7rr{v

i'"

it follows that

(-r(i^)""
i=o^"i!r{v + 'in+

i^T(z.) S
"Sf,_,,(^)

(3)

{2\oghz

this formula, which

y}r{v +

integer)we
positive

to

apjjears

be

1)

1)

f (m

2"--^tt T

1)}

nugatory whenever

only when

is,in effect,
integer,
nugatory
a

1)cos

jn

hence

and

ii; +

i'

0 ; for when

by

define the function

v
v

"

is
7i

(v)Y, (z),

negative

(where

the formula

(2),
(z) 'S-w-i,
"S^_"_i,_n
=

in which
To

the function

discuss the

case

rightis defined by equation" 1073 (1).

the

on

in which

i'

V,

which

(z)-

'S-i,o(2),

gives

take the fm-uiula

0, we

^,{s''^'-^%^2,0(^)}

M=-l

Since

{^
S, ,,,,{z)=^z--^{V
+ 2"

it

f)}-^2

'^J

+ i)}2{cosAyxTTJ'o(z)
{r (l/L.
-

Sin

i/xTTJo (--)),

on
reduction,that
follows,

Cf.

Bromwich,

Theory of InfiniteSeries," 152.

is

350

OF

THEORY

10*74.
From

Functions

BESSEL

[CHAP. X

FUNCTIONS

in terms
expi-essible

of LomtneVs

functions.

descendingseries givenin " 10'7l (1)it is evident that Neumann's


in terms
of Lomrael's functions by the equations
poljmomialOn{z)is expressible
the

0.,,n
{z)
{Z) (If2) S,^
0^,^+1
^n. (Z),

(1)
and

Schlafli's polynomialSn{z) is

S^ (z)

(2 )

to
possible

of Lommel's

{0"J, (z)

^1-"

the

equations

2,"

the

express

in terms

similarly
expressible
by

(z), S^,^^(z) 26'o,o,"+,


"Sf_i,
(z).

4m

It is also

+ 1 )/^}
{(2//i
{z),
/S",
,,n+x

zi^ J^

importantintegrals

{z)dz,

functions ; thus

Y^ {z)dz

z"^

have

we

^^_i, (z)] zJ,.,{z)8^_,^ (z)+(^-v-l)


=

,_i

,_;

zJ, {z)^^_2.{z),
,

/,_i {z) Z-- S^^ (z)]


[z'-'
=

On

" 10'72
(8)
and

from the rightof


eliminating/S^_j^_i(2:)
that
(6),we find by integrating

[' J,{z)dz
z^

the

proofsof

(4)
and,

(z)
{fi+ v-l) zJ^ (z)"SV_3.
._,
shew

nature

same

and using
equations,

zJ,^,(z)/Sf^.
(z),

that

generally,
{^ v-l) z^, (z)S^_,
j^z'^K{z)dz
+

of

Specialcases
the functions

(6)

these

{,^ v-l)zY,(z)S,_,,,_,(z)-zY,.,(z)S^,^(z),
j%:^Y^iz)dz

more

(5)

+ {fi
zJ, (z)S^_,{z)
+ v-l) zJ,_,(z),Sf^_,,
(z).
,_i

the

on

f z^,,, (z)

0)
Of these

(1876), pp.

these

formulae

rightreduce

dz

to Neumann's

I-J^%"

z^

r ''^2,0+,
(Z)dz

"

444;

be

(6) and

noticed

by choosing/u,and i" so that


Schlafli's polynomials,
thus
or

(z)0,,_,(z) "^^,_,(z)0,",(^)l
-

(7)

were

Cylinderfunktionen
(Leipzig,
1904), p.
It should

^Z {'(^,,ri+,
(Z)S^" (Z) %rn (z)^2m+i(z)].

(1),(3),(4) and (6) are


results,
425

obtained

are

(z) z%\^, (z)^V (z).


-

,_,

that

contained

in Lommel's

given by Nielsen,

paper, Math. Ann.


Handbuch
der Theorie

100, but his formuljie contain

Lommel's

some

function,in those

ix.

der

misprints.
eases

when

it is

10-74, 10-75]

ASSOCIATED

351

FUNCTIONS

in
expressible

finite terms, is

equivalentto Gegenbauer'spolynomialof " 9"2.

The

connectingthe

functions

formulae

2''r{v + m)

lit i

are*
2m

+
z

(^)
A

2''+^r(^^+m + l)

X-

2w

i/

l^

.,

It follows that the most


in

generalcase in which the integral


(5)is expressible
functions is givenby the formula
elementaryfunctions and cylinder

of

terms

("')

(^ .+..

(^) dz

^.p(^^^,^^

,.

2/M

The

function

defined

2m

the series

by

+ 2'"
(-)'"2''

lor(i/
+ 2??i+

)((.

has been studied

in great detail

analogousto those of
generalfunction seems
Tiie

integral
v

Messenger,xxxiii.
incorrect

H.

i)

insufficient

.-s

dt has

been

58 ; and

pointed out

the

justifyan

to

i)*''+7,i^^'*

r(..+

Voung

Bessel functions,but

(1904),p.

(as was

by W.

4:Z'

this function

increase

of

account

of them

studied

(when

by Kapteyn)

and

is

he stated that,when

possesses

many

the value

more

here.

integer)by

an

n, its value

for

properties

Lommel's

simplicityover

H. A.

Wel)b,

is 0" (z). This

generalvalues

of

is

is J

{"'^'uA-2)-^'%Ku{-Z)}Iz,
when

Ji (v)"

0 and

The

10'75.
We

shall

|arg {-z')\"7r.

asymptoticexpansionof S^"(z).
shew

now

then
positive
integers,

by Barnes' method" that, when /x "v are


admits of the asymptoticexpansion
S^^^{2)

"^

(;. 1)^
-

^.^

[(f, If
-

v^-\
{(^ 3)-^

z"-

when

not

odd

V-

z*

i^' jis largeand |arg z\ "7r.

Let

take the

us

_^
/Stti
The

integral

ni-lfi
p.-Phi
j
_co/_^,+j

contour

r(^

+
"

is to be drawn

^v-sjTi^-lfM-^v-s)

+ ^v) r (I
^/j,

"

by taking^j to

h/J' ^f)

TTJ-Uy-'ds
'

"

be

an

onlypolesof the integrandon the left of the contour


functions being on the rightof
polesof the Gamma

sinsTT

integerso largethat the


are
polesof cosec sir, the
the contour.

Oegenbauer, Wiener
Silzungsberichte,lxxiv.
(2),(1877),p. 126.
+ QuarterlyJournal, xliii.
177.
(1911),pp. 161
I Cf. Gubler, Zurich Vierteljahrsschrift,
xlvii.
(1902),pp. 422"428.
"

" Proc.

London

Math.

Soc.

(2)v. (1907), pp. 59"118;

cf.

"" 6-o, 7-5, 7-51.

352

THEORY

OF

BESSEL

integralis convergent when


that it is 0 (zf^'-P).
difficulty
The

[CHAP. X

FUNCTIONS

[arg

|"

tt,

and

it may

be

without

seen

from the
be shewn
It may
when
that the same
integrand,
at

centre

p
tends

that R

"

through any

function
asymptoticexpansionof the Gamma
round a semicircle,
of radius R with
integrated
f on the rightof the contour, tends to zero fis R-^oo
provided
in such a manner
that the semicircle never
to infinity
passes
the
the
of
integrand.
polesof
"

the

It follows that

expressiongiven above

is

equal to

the

sum

of the

residues of
^'^

'
"

r(^-ifi
at

the

iv)r{i-ifi-iv)

sinsTT

s
points

0,-1,-2,. ..,-(^-1),
1,2,3,...,
1
2

2A*
2/^

When

we

^^1

2^''

_1,,
3_1,,
2/^2*''
2

2^"

2A''+2'"

1,, _li,
"

'

calculate these residues

(-)'"r{l-ifi

5_1,,
_!,.
2/*
2'^---)
2^^2*^'

find that

we

^v + 711)r (I

l/i

1
1;

m)

_j

,_,

(-rr(i

ifjL-iv)T{i + ^f.+ iv)qz)^

T{l"v-\-m)
r(|-|yu.+ |i^)sini'7r^=oWi!

2"-^7rra+ia-i.)

-^

(-rd^r-^

so

that

sini/TT
X

/_, (z)
[cosl(/jL-v)7r.

by " 1071 (2),we

and

so,

and

this is

to
equivalent

have

the

cos

J, (z)]
l(fi+ i')'7r.
=

the formula

asymptoticexpansionstated

in

(1).

0 (z'^-'P),

10-8]
10'8.

353

Henii-cylindrical
functions.

Functions

the single
recurrence
Sn{z) which satisfy
=

combined

formula

2S,/ (^)

S"_, (^) S,+, (^)

(1)
with

S,{z) -S:{z)

(2)
have

FUNCTIONS

ASSOCIATED

studied

been

great detail by Sonine*.

in

They

will be called hemi-

functions.
cylindrical
It is evident

that S^i{z)is

in
expressible

the form

S, {z)=/" {D) So (z),


.

where

d/dz and fn (D) is a polynomialin D of degreen

fn (^)satisfiesthe
combined

and the

polynomial

formula

recurrence

with

/o(^) l, /(^)
=

It follows

by

-e

(cf.
" 9"14) that

induction

/n(^)=H{-^
and

v(r+i)i"+{-i-v(r+i)}'a

therefore

S"{z)=^[{-D

(3)

^{D-^+l)]-+{-D-,/{D-^+ l)\n].s,{z).

values
(1)holds for negative

If it is supposedthat

S_n{z)

(4)
To obtain

an

alternative

J''^^'

of n, it is
easy to

sinh t, and

{-Qmhnt

"1

that

{-rSn{z).

to (3),put ^
expression

thenf
(wodd)

2l ^

'

see

4l

"

"

{n even)

n{n'- 1-)

n
^

'r----^^

[-1,^-

3;

(^odd)

Hence
a-

n-

+
So {z) + ^ So" (z)
2|--" \-y

"

(5)

S"(^)

(n^

"^

"

So {z)

So

{z}-

It is to be noticed that On {z),


T,^(z)and En(z)

Sn(z),Un {z)and
It should

H" {z) are

be remarked

that

not

the

functions of
*

Math.

no

....

xvi.

Hobson,

(1880),pp. 1"0,
Fhme

{n odd)

formula

^S.(^)

intrinsic interest than Lommel's


gi-eater

Ann.

t See e.g.
B. P.

(n even)

are

singlerecurrence
=

w.

.,.

hemi-cylindrical
functions,
functions.
hemi-cylindrical

+ X.+,(^)
S,_,(^)

givesrise to

So'^{z)+

but

2^)
^

41

polynomials.

71"80.

Tiigonometry (1918),"264.
23

354

We

theorem

TJie addition

10*81.

shall

BESSEL

OF

THEORY

functions.
for hemi-cylindrical

establish Sonine's

now

importantexpansion*

S",(^+ 0=

(1)

expansionis valid when z


the pointz, which does not

function

inside

as

or

contour

contain

whose centre
is
circle,
largest
of the hemi-cylindrical
singularity

any

circle C with

centre

such that So

(^) has

no

singularity

the circle. Then

on

Ziri

The

Jn{t)S,^-n{z);
00

"

consideration.

under

Take

-\-t lies inside the

the
at

[CHAP. X

FUNCTIONS

^"z

the contour, and

uniformlyon

series converges

iTTl

"

,i=o

we

have

^-^d^
^. l.nJn{t)fni-^)\
V dzj
^-z
2in

,j=o

"^

so

d\

"nJn{t)fn[-J-^j^m{z)dz

=J^.,./.("/"(-|)/.,.(j'
it is easy to

But

verifythat

(f).
2/n(- ^)/" (^) =/r"-"(f)+ (-)"/""+"
so

that

S^(0

i Jn{t)[S,n-n{t)
+ {-T^,n^n{t)],

^o(OSm(^)+
M

Sonine's

whence

It should

type than

formula

be noticed

is obvious.

that,if S" {z) denotes

function,namely
hemi-cylindrical

function

one

which

of

general

more

merely satisfies the

equation
the equationS-^{z)
satisfying

without

fn
and

so

the

formula

the formulae
*

XVII.

Math.

Ann.

of

(-^)S", {Z)

xvi.

934

and

(1880),pp. 4"8.

See

stillhave

S,n-n (z)+ (-)" S",+" (z),

(1) is still valid. We

""5-3,91,

(1880),pp. 85"86.

we
So'(^),

"

thus

have

an

alternative

proofof

10-63.
also

Konig, Math.

Ann.

v.

(1872), pp. 310"340

; ibid.

356

OF

THEORY

BESSEL

given,and, after definingct^ (z)by (6) and (7),we

F,(z)

(9)

and

are

independentof
It remains

z,

Y, (z)\d,+ iTT

J, (0

that

c^

and

c?"can

be chosen

F,{z) given by (9) satisfies (1) and (2),or (what


then
that it satisfies (3) and (4). If (3) is satisfied,

zJy-i(^)

c.

""

^TT

I Y, (t)

m^

the

method

(0 dt\
,

be

to

be

that the value of

so

the

to

comes

taken

thing)

same

(t)dt ]" ^irzJy (z)Y^ {z)^^ {z)


-

dt [ + ^TTZF" (Z)J^ (Z)Tff^ {z)


F^_i {Z)\d^
-k-^TT { J^ (t)
-ST^ (t)

zJ^_^{z)j(
+

^,

constants; and Ct, and dt,may


arbitrary
though they will,in general,
depend on v.
be shewn

to

(5) by

J,(z)\c,-l7rY,{t)^,(t)dt

where

solve

solution is

The

of variation of parameters.

[chap. X

FUNCTIONS

_1,

Y,_,{t)7n^_,{t)dt\

'
2 TT

"v"\

^F^-i {z)\d"-i+ l-TT J,_i(0

(t)dt\-a^

^^-1

{z),

that is to say,

^J"^_i
{z)

Y" (0
-l'^\\

--1

^F,_i(2) d^

But

(/,_!
+ Itt

it is easy to

^.

(0

Y^_, (t)t^,_, (0}dt

{J,(t)^At)-

J,-, (t)CT,_i (Ojdt

ofp (z)=

0.

verifythat

(z)/3,_i{z) ^, {z)", {z)]

since

zJ^_^{z)\c"
+

and

c^_i

{z)^, (z)

F, {z)a, (z)

4- Itt /,_! {z)/3,_i


fZ^_,
(^) J^ {z)a, (z)

" 3*63 (12),reduces

Itt

i(ll)

d,

c^_i

d,_,

^TT{F^_i (a)yS^_i(a)

Itt

[J,_,(b)/3._,(6) -/.(b)a, (b)]}

far as (3) is concerned, it is sufiicient


so
Consequently,
the diiference equations
satisfy
c^

a,(z) 0,

F^ (a)a^ (a)]}

to

Itt[F^_i (a)/3^_i(a)

c"_i +

^F,_, (^){o5, d,_,

|(10)

(z)'^,_,{z\

t^,_i

(3) is satisfied if
-

this condition,by
-

so

Itt F,_i {z)/S._,{z)

^F^_i {z)\d^

zJ^_i(z){c^

to

and
satisfied;

(6) and (7)are

isr,

._,

dz

to

F^ (a)"^ (a)),

(6)/3,_,(6) /, (6)", (6)}


{J,_,
;
-

choose

c"

0.

and

d^

10-82]
and

ASSOCIATED

the

difference
the

difference

condition

(1) and
as

^^

of

by

of

are

in

is

(8)
solution

of

two

same

satisfied, then

are

solutions

sufficient,

be

may
well

as

given

the

of

these

(4).

whose

type

throughout)

i/

if

that,

verifying

solution

condition

the

so

is

by (9)

existence

the

for

replaced

equations

and

known*;

difficulty

no

(with

(z) given

of F^

These

If,

have

equations

value

as

will

reader

357

FUNCTIONS

as

of

pair

regarded
necessary,

functional

tions
equa-

(2).
00

-^

f^(z)
S

where

0, then

"

we

0(z^-%

make

may

so

that

solution

general
F, (z)

(12)

7ri(i/) and

Some

Note.
be

of

of

formulae

deduce

to

(0

["J,

ItT

periodic

of

Mat.

(t)

which

(3)

of

dtl
,

with

period

17

pp.

"

unity.

(2) 07ili/are

satisfy equation

(1901),

v.

dtl

(t) 7Z, (t)

functions

functions
di

^.

Thus,

31.

to

from

{z)+F,

(2vlz) F(z)

^,{z)-

2g, {z)lz,

that

F,,,,{z)

(13)

ItT,
(l/)

A)in.

paper,

["F,

type
F"

it is easy

Itt

properties

earlier

the

(v)

rt"_i

arbitrary

are

interesting

Nielsen's

in

found

set

773(1/)

L,

have

we

written

be

F, (Z)

where

and

zc
,

cti,

c""i,

may

/, {z)

h^~

oo

-*-

0{zi-^),

Cy

the

g^{z)

F,{z)R",,{z)-F,_,{z)R,,.,,,^,{z)

n-l

the

first

and

the

Math.

An

terms

two

series

is the

of

account

Soc.

(2)

11.

the

on

right

particular

various

(1904),

are

complementary

gv^-miz)Rn-m-hv^m

function

of

the

given

by

\{z);

difference

equation,

integral.

memoirs
pp.

the

(2/2)

438"469.

dealing

with

such

equations

is

Barnes,

Proc.

London

CHAPTER

XI

ADDITION

The

11*1.

THEOEEMS

of addition

general nature

theorems.

proved ("4"73) that Bessel functions are not algebraicfunctions,


obvious
in Chapter vii
it is fairly
from the asymptoticexpansions obtained

It has been
and

that they are


they are not simply periodicfunctions, and, a fortiori,
in
with
functions.
accordance
theorem
a
Consequently,
doubly periodic
that

to Weierstrass*, it is not

Jv (Z) and

of

addition

as

Jy (z). That
in the

theorems

There

however,

are,

addition

the

is to

strict

the

In

that

say,

as

due

algebraicfunction

an

functions

do

not

possess

term.

which

of formulae
of functions

case

"\-z)

Bessel

of the

sense

for functions

formula

Jv{Z

express

classes

two

theorems.

; and

coincide

possible

to

not

are

of order
first kind

of the

commonly
the

zero

described
classes

two

is

00

Jo {'\/{Z'^
+ Z-

2Zz

"

^)}

cos

7)1

which

has

The

Heine

on

|,who

addition

Neumanns
shall

We

as

integral;

confluent

is due

another
of the

form

to

Neumann

proof is
addition

f,

due

to

theorem

functions.

Legendre

11*2.

of this formula, which

the formula

obtained

m(f",

of Parseval's

transformation

" 4"82.

in

simplest rigorous proof

depends
for

indicated

already been

e^nJm {Z) Jm i^) COS

establish

now

theoi'eni^.
result

the
00

^0 M

(1)

CT

all the

The

1893) ;
Ch.

xni

variables

theorem

see

Phragmen,

for

proofs

XHandbuch

Analysis, "

""
vii.

"

to

of

2^

2Zz

have

cos

0),

general complex

Schwarz'

edition

(1885), pp. 33"42,

and

of

values.

Weierstrass'

Forsyth, Theory

(Berlin,

lectures

of Functions

(1918),

Functionen

Kugelfunctionen,

i.

70.

(Leipzig,1867), pp.

-lO

(Berlin, 1878),

340"343;

pp.

"

cf.

" 5-71

and

Modern

15'7.

" In addition
XVI.

der

sJiZ^+

theorem.

BesseVschen

der

Theorie

of the

in

Math.

Acta

supposed

are

stated

was

7/10,

write

where, for brevity,we

and

^mJ'm {Z) J,n {z)COS

to Neumann's

(1880"1881),

pp.

201"202.

treatise

cited in " 11-1, see

Beltrami, Atti della R. Accad.

di Torino,

-3]

11-1-11

We

ADDITION

THEOREMS

(Parse val's integral)

take the formula

If''
^TT

which

the

of 6 with

/"'"

^TTJ

-n

is valid for all (complex)values of

function
analytic

359

sr

and a, the

We

period27r.

_^

integrandbeinga periodic
suppose that a is defined by

next

equations
sin

ts

and

it is then

cos

"

(f),

-ur

cos

a.

sin "^,

apparent that
\

,/"(ct)
=

["
.

X"

exp

2^ S
}H

^) sin ^

^^ cos

"

i^

sin

cos
^]c?^
cf)

((^-6)f^0
gj/it9+(zsin

/,"(2)
X

^TT
)".= =
HI

[i(Z

X
:

"
-

"l

the

interchangeof

./",(^)J..(^)e-'^
X

"

the order

the

uniformityof convergence

the

of
periodicity
If

the

of summation

for which the values


group the terms
formula.
immediatelyobtain Neumann's
The

formula;

J511

see

LXi.

of order

227, four years

"

of

differ

"

before the

were

only in sign,we

obtained

by Clebsch,

of Neumann's
publication

'4.

addition

Neumann's
v

(1863),pp. 224

Graf'sgeneralisation
of Neumanns

11*3.

order

for Bessel functions

correspondingformulae
filrMath.

the next

integrand.

we

Journal

followingfrom
integration
from
step following

and

of the series,and

theorem

has been

form'tda.
extended

to

functions

of

arbitrary

different ways.
The
extension
which
to be of more
seems
is due
to
Graf*, whose
importance in physicalapplications

in two

immediate
form uJ^ is

iZ
(^

this formula

and

ze~^)^^

J^{^).\^^\

(1)

is valid

"

ze^

""

J,^,"(^)/,"(^)e-*,

III =-00

providedthat

both

of the numbers

are
\ze^''^\

less

thanj^l.
*

pp. 59

Math.
"

p. 241.

61.

xliii.
(1893),pp. 142
specialcase of the result

Aim.
A

"

144

has

and

Verhandlungen

also been

obtained

der

Schiceiz. Natitrf. Gen.

by Nielsen, Math.

Ann.

lii.

1896,

(1899),

360

THEORY

BESSEL

OF

[CHAP. XI

FUNCTIONS

but, two years after it was


integration,
proofis based on the theory of contour
G.
T. Walker, Messenger,
xxv.
pubhshed, an independentproofwas givenby
(1896),
pp. 76
80; this proof is applicableto functions of integralorder only, and it may be obtained
from Graf's proofby replacingthe contour
by definite integrals.
integrals
Graf's

"

To

the

prove

is convergent in

m=

"

generalformula, observe that the series on the rightin (1)


the circumstances
and so, if arg Z"a, we have
postulated,

'x

^T!"*

is

r
exp\lZ(t--)\

s-7

(,

-a)exp(-"a)

in interchanging
the
specialdifficulty
*.
integration

there

where,

-07

For

-sr

all admissible

w-contour

We

to start

then

+ z^
\/{Z'^

is taken which

makes

"

2Zz

cos

(f"),

of the square

root

^^0.

of z, the

values

or negative.This
positive

and

{Z- ze''l')/t
^ju,

-uTU,

that that value

supposednow

Z when

ze-''^)t

usual,

as

it is

-^

of summation

write

{Z

and

/J

order

no

Now

J,,(z)e-^'Ut

"---'

w=-oo

from

phase

determination

and

end

at

oo

"

of

is
"ay/Z

of

-nr

exp("

renders

now

it

where
1/9),

acute

an

to
possible

y3

arg

angle,

take the

zsr.

have

J." (Z) J, (.)."-

i-.l^^^)'r

m=-oo

27rl V

CT

exp

h.

J-ooexp{-ip) [

^
')]
(u
V
uj }
-

v''+'

(Z-ze-'^\^^
this is Graf's result.

by " 6-2 (2); and


If

we

define the

anglei/rby

Z
where

yjr
^^

as

relation indicated

"

-^

by

cos

(3)

"

sin

yjr,

may

be written

00

we
yjr,

t
Wl=

Cf. Bromwicb,

Graf's formula

and

e-^'*M^)=
*

nr

i /.+,.(^)/,.(^)e'"'*

(tsr)

changingthe signsof ^

"

real values of the variables,we

Fig.28), then

and, on

-v/r,z sin (f"

cos

OT

equations

{so that, for

e"^^/.

(2)

the

"

have

J.^m{Z)Jm{z)e--^'\
00

Series,"
Theory of Infinite

176.

obtain

the

11-3]

361

THEOREMS

ADDITION

it follows that

whence

(4)
olll

""
jn=

oill

00

"

Fig. 28.

change the signsof

If,in this formula,we

and

in,

deduce
readily

we

from

that

8'54

(5)

sin

and

'

'

"^

sm

,"-_"

so

(6)
sm

The

formula

of
applications

If

given by

(5) was

also Sommerfeld, Math.

Ann.

and

sill

Qo

"

in his treatise in the

Neumann

due

are

(1894),p. 276; ibid.

XLV.

the formulae

replaceZ, z

we

)"

XLVii.

Schwarzschild, Math.

to

Ann.

lv.

equationsby iZ, iz

in,these

specialcase

(1896),p.

356.

Some

0;

(1902),pj). 177

and

i-^

see

physical
"

247.

respectively,

it is apparent that

0)
sin

1M=
)"

"rr,
"

Olll

00

00

(8)
bill

Of these

iij=" =
)H

stated

(7)was
results,

!5J11

cr
X

"

by Beltrami, Atti

della R. Accad.

di

Torino,xvi.

(1880
"

1881), pp. 201"202.


The

should
results,obtained by taking(^=|7r,
special
following

9^,{m)(,o%v^=

(10)

S
{-rK+,m+i{Z) J,m+,{z),
'^,(B7)sini;r/r=
?Ji

cos

""^,

"

{-y^9^.+,,n{Z)J,"^{^\

(9)

where

be noticed

=-st

sin

yfrand

\z\" \Z\.

of these formulae the reader is referred to


physicalinterpretation
the papers by G. T. Walker
and Schwarzschild
; it should be observed that,in
in which v is an integerand the only functions involved are
the special
case
of the first kind, the inequalities
|2^e***\"\Z\ need not be in force.

For

the

362

THEORY

BESSEL

OF

[CHAP. XI

FUNCTIONS

theorem.

11 '4.

addition
Gegenhauer^'s

The

of Neumann's
addition theorem
was
generalisation
the
of
before
publication
by Gegenbaiier*nearly twenty years

second

obtained

of

type

Graf's paper.
If Neumann's

of

" 111

is differentiated

times with

respect to

find that

"^,we

cos

formula

Jm+n {Z) J,n+n(^)d""COS (m + ?i)0


d (cos"f"y'
z''
,",'Zo'^^''Z""

Jn ("r)

-^^
tsr'^

(1)
This formula

"

e.

order
by Gegenbauer to functions of non-integral
differential equations(see" ir42); but
of the theory of partial
by means
Soninef gave a proofby a direct transformation of series,and this proof we
shall now
reproduce;it is to be noted that, in (1),^ is not restricted (as in

" 11 '3)with
We

extended

was

reference

Z.

to

take Lommel's

expansionof " 5'22,namely


j.wa+h)]

^ (- hh)pj.^A^)
^

and
in

if we
replacef and h by Z'^+ z^ and "2Zzcos(})respectively;
it is found that
for brevity,
placeof J^ (otj/ct"

J.+pW(Z^+2')}
(Zzcos"t")P

I (-yzP+^COSP"f" J^p^giZ)

"
^

by

and

further

write O

^ro,to 2^.p\ql

of
application

Lommel's

'

Z'^^

expansionwith ^

and

r(p

k)

replacedby

Z^

2l

But, by

" 5-21,

v+p+2k
ql
J,+p+g{Z)_ 9
T^
29
{:,k\{q-k)\

p +

iy"^^^'^^^^'

and

Tiv

p-^q+k

so

4 (-)g {v + v^ 2k) r (v +

P +

k) zP^'i

cos^

Z"

+ q+k+l)
^ro^ro^ro 2'^plkl{q-k)ir{,"-\-p

the

J,+j,+^Zl

tripleseries

on

the

right being absolutelyconvergent, by comparison

with
00

00

/)=0g-0

qi

S
A:-0

Z'^^^
r (r + p + A;)
zP^-^'i
/.\^^
r (j,
+ p ^ 2k) r{v+p
2P+-2q+2kj)i
^.)!
_

Wiener

t Math.

Sitzungsberichte,lxx.
Ann.

xvi.

(2),(1875),pp. 6

(1880),pp. 22"23.

"

16.

q + k +

i)

364

THEORY

OF

BESSEL

[CHAP. XI

FUNCTIONS

in obtaining
though it is used implicitly
some

of the results

givensubsequently

in this section.
Unlike

the

is so
" 11"4, the formula is true only when l^'l
small that both the inequalities
|ze^^^ \"\^\ are satisfied;but, in provingthe
formula, it is convenient first to suppose that the further inequalities
formulae

of

+ z''\,\z\"\Z\
\2Zzcos(f"\"\Z^

satisfied.

are

We

then

Lommel's

use

expansionof " 5'22 (2) in

P!

i)=0

which

is valid when

It is then

^ ^

found

^
^
-

"

|^|.

alterations
by making slight

(-)y+i{v ^- 2i + 2k)V {-

"3oA/t=o

|A

P+

.;i,C""X
2"^

{v+p

~,.=oA=o

SO

the

q-

of 1 11"4
analysis

2k)T{v
p\k\
r

p +

Z"

n) ^P-H*+^" cos^ (t"


jZ)
J-,.p-,k
Z^

k) cosy ^ J_,_p_^(Z)J,+p+,k
(z)
Z^

+
(i/

k) cos

"'"^^

requiredresult

Z"

%^
^-7/^^

is established

z^

J-^-m (Z) J^+m (z)


"/"

{m-2k)lk\

(.) i (-r (^ + m)

under

that

k) zP+-'9 cos^ "f"


J-,-p-^{Z)

2^^-^-p\k\n\Y{\-v-p-k)

piok%

2-' r

'2k)r(-v-p-2k-

2''+"'-2^{v + m)
I "4"^(-)m-fc

in the

2-^'iplk\(q-k)ir{l-v-p-k)

+
(-)P+'(i^

I I i-y^^

the form

z"

G^" (COS,/,),
the conditions

\2Zzcos"f"\"\Z'+ z^\, \z\"\Z\.


Now

the last

is an
expression

circle of convergence

analyticfunction

when

lies inside the

of the series*

I (y + 7n)Z-'"-"^
,"=o

and

of

z'^

Cm''(cos(b)

r(l -v-m)T{\-\-v^-m)

this circle is the circle of convergence

'

of the series

0^" (cos"^).
(^Ij

given series converges and represents an analyticfunction of z


is satisfied,
providedonlythat j2re*"*j" |^ |; and, when this pairof inequalities
is also an analytic
function of z.
J-^ (ra-)y''5r''
Hence

the

Cf.

" 5-22.

11-41]

365

THEOREMS

ADDITION

Hence, by the theory of analyticcontinuation,(1) is valid through the


of the domain

whole

of

of values

for which

I^e*** I" i^ I
"

replacey by

If in (1) we

Again,if
values of

^^-P 2" r

(3)
and

(2),we

that, for the domain

see

of

(.) 2

%^" %M

(. + m)

C," (coscf"),

generally.

so,

^)

(4)
If in

ir4

consideration,

under

now

find that

(1) with "

combine

we

i^ we

"

make

(8) we

and

^-0

Co"(cos 0)

2^r(.)

lim

use

(.+^)%i^)^%^c',/(cos"^).
the formulae

(cos 0)}
[r {v)(v + m) C,,,"

cos

mcj), (m i^0)

find that

we

00

Fo(t3-)=S

(5)

cos
",nY,n{Z)J,n.(z)
m(f).

m=0

formulae

The

(1) and (2)have

not

been

given previously;but (3) is due

to

Gegenbauer,

in his treatise (save that the functions


"," were
rej^laced
given by Neumann
to
with
has
been
formula
also
T
he
examined
an
functions
the
equal
(3)
integer
Ft"*)).
by
464.
der Kugelfunctionen,
Some
i. (Berlin,1878), pp. 463
ments
developby Heine, Handhuch
due to Ignatowsky, Archiv der Math, und Phys. (3) xviir.
of (4)are
(1911),pp. 322

and

(5) was

"

"

327.

If

replaceZ, z

we

this section

we

^-#

(6)

and

by iZ, iz

ot

and

in the formulae

of

" 11'4

and

find that

2^ r

tJT-

(.) S
,"

(-)'"(. + m)
o

%^^
%^^
Z"

C^" (cos"/"),

Z-

(7)

J^_-Ap
2''r{v) i (-)-(.+,n)^^^^"%^6;/(cos

(8)

^"

Of
pp. 156

2^ r

(.) i (. + m)

^^^ ^^^

C.,r(cos"^).

Math. Soc.
formulae, (8) is due to Macdonald, Proc. London
Neumann
in
the
specialease
given by
157; while (6) and (7) were

these
"

xxxii.
v

(1900),

366

THEORY

The
in the
2

formulae
case

and

(b

If

^.

V(/-'+

are

of

special
physical
importance
kr
0 for Z.
and
by writingkxt,

become

cos
6)
fl"'-2arcos^)
a-

Z;v'(?^
+

cos

^ ^^

V(r^+

a-

2ar

"

""

2ar

-2ar

cos

6?)

(9)

cos

{" AVC?*^+
v(^

a^

"

0-"

2ar

2or

cos

cos

^)

^)}

l)^^iii"iA^^'Ii"i"^)p,"(

(2,"+

\ "

171=0

These

point on
centre

formulae

of

are

importance

the axis of harmonics

is at the

at

If "f)

TT, we

If 0

|7r,we

Z=z.

the

6=0,

reader

from

Ann.

the
lxxv.

of (4) were

which

from a
pulsationsemanate
in presence of a sphere whose
origin,

(1914),p.

pointedout

141

et

seq.

by Gegenbauer,and

have

have

and

is taken
"if^

alreadyobtained

formula

\r

problems in

origin. Cf. Carslaw, JJatk.

followingspecialcases
worth recording
:

are

in

distance

The

V'*

V"

fi.=0

exp

[CHAP. XI

FUNCTIONS

the notation

change

we

that the formulae

sin k vX/^ +

.^^

BESSEL

and of this section

" 11"4

of

see

we

OF

should

consult

to

("o'o) by

be /",

different method;

Gegenbauer,Wienei'

in this connexion

lxxv.
Sitzungsherichte,

(2),

(1877), p. 221.
More

taking Z=z,
generally,

^=^0, 'fe'Y/",,we

cases
Gegenbauer,loc.cit.givesalso special
(b
TT.
(^ ^ TT.
=

have

of this formula, obtained

by taking

11-42]
Again,it

that*, if B{v)"

be shewn

can

0/
I sin-" (^Cm" (cos(/))

providedthat R{v)

so,

and,

fV(-^'+
T" '^".'

Kx

simple proof of

functions

"

\,

"

V2/

2/

0)*"

^'

2^^

4")]
^

cos

first kind,

of the

proof for

direct

"

.,.

^^

^z,

.^

'^r(2"/+ 7H) '^Um(^)"/.+.,


(^)
Z"
2"
2"-^m!r(i/)

in which
the cylinder
0 and
m
given by Sonine, Math. Ann. xvx. (1880),
the first kind is due to Kluyver, Proc. Section

f, in the specialcase

this formula

functions

are

Another

p. 37,

y'^^'V)

[~22''-J(^
+ m).m!{r(i/)p

-\-m)
{'Iv

generally,

more

/ -,

it
_

2:2-2^^008

.'o (^^+

^,

"

(cos"^)c^0"

"0

and

367

THEOREMS

ADDITION

was

of

functions

indirect proof,
^755. An
xi.
te Amsterdam,
(1909), pp. 749
of ScL, /{. Acad, van
Lxxxv.
is due to Gegenbauer, Wiener Sitzungsherichte,
(2),(1882),
depending on " 12'13(l't,
Wet.

"

pp. 491"502.
An

[Note.

Lxxiv.
(2),(1877),p.
Sitzuiiffsberickte,

of

of (4),which

interestingconsequence

then
integration,

was

noticed

127. is that, if jse***


\"\^\

by Gegenbauer,
throughout

the

Wiener

contour

fcf. " 9'2)


^

(18)

"

-"-^

,"

(z)"

2" r

"

results of

of this formula, resembling the


Specialcases
equal to 0 or tt.]

The
to

method

obtain

the

used

by Gegenbauer,

It consists

transformation.

of

theorem

addition

in

Wiener

"

11"4

proving

" 9*2,are

obtainable by taking 0

theorem.

addition

of the
investigation
O'egenbauer's

1142.

g,/ (coscj,).

(.) (.+ m)

LXX.
Sitzungsberichte,
so
quite
easy

(2),(1875),pp.

is not

that

is

solution

to

of the

justify

as

"

16,

Sonine's

differential
j)artial

equation
d^a.

2i/+l

3q

d2j2

z^

d"p^

2;/cot(^9Q
~

dz^

assuming

and

that

i2 can

dz

be

expanded

z^

'

d(f)

in the form

1 ^,".C,/(cos"/)),

0=

m=0

is independent of 0,
whpre-iJj,!

and

is
C'"/(cos(^)

polynomial of degree m

in cos"/);it

follows that

2.cotc^l}c,/(cos0)
{|,+
*

Gegenbauer,

London

Math.

Soc.

Wiener

Sitzungsberichte,lxx.

(2)iv. (1906),p. 472;

cf. also

443, and Bateman, Proc.


(2),(1875),pp. 433
Barnes, QuarterlyJournal, xxxix.
(1908),
p. 189;
"

246.
Soc. (2) xvii.
Math.
(1919), pp. 241
Anabjsi.%" 15-51 and Proc-. London.
(16)has been given in the specialcase i/^O by Heaviside, Electromagnetic Theory,in.
disguised form.
(London, 1912),p. 267, in a somewhat

Modern

t Formula

"

368
is

OF

THEORY

constant

of a'" in the

BESSEL

[CHAP. XI

FUNCTIONS

multiple of C^ (cos^), and so C^" (cos0) may


expansionof (1 2a cos 0 + a^) ". And then 5^,

be taken

"

qua

be the coefficient

to

of 2, satisfies the

function

diflferentialequation

multipleof z""./^^"(2),the
beinganalyticnear the origin.
that

so

i?m is

of

considerations

From
a

symmetry

i2=

Om

"

z^

Z"^ cos'^
A

in
(f)

similar
less

seems

11 '5.
The

process

used

was

e^"""*

by Gegenbauer

in the

case

to

of functions

establish

"

11-41

(3),but

coefficients of

the

analysis

of the first kind.

addition

(2/1 1)
(")*^i^

t"

theorem.

Jn^ {Z)Pn (COSC^)


by Gegenbauerf,
generalised

by Bauer* as earlyas -1859; it was


the expansion
2''r(i;)i {v + m)

g"cos.*

(2)

i^

^"-^"^y^
Cj' (cos"^)
;

-2^"

m=0

of" this expansionin which


case
obviouslythe special
j.
fundamental
the
becomes
when
expansion
v-^0, the expansion

result is

In the limit
of

Z, is

formula

obtained

Bauer's

by comparing

form of the
degenerate

The

discovered

who

in the

of

function

5"i, qua

-^

only ; and h^ is determined


expressionon the right.

and

convincingthan

(1)
was

and

of

that

VrrCK^O"(p),

"

rr;

"^

m=0

6,"is a function

inferred

Gegenbauer

equation not

that

multipleol Z~'"Jy^yf^{Z\so

where

of this differential

solution

other

t"

"2-1.

Gegenbauer's expansionis deducible from the expansionof " 11 '41 (4) by


00 ; it is then
apparent from " 11"41 (9)
multiplyingby Z""^* and making Z
of the expansionis that it gives
and (10) that the physicalinterpretation
the axis of
on
the effect due to a train of plane waves
coming from infinity
in a form suitable for the discussion of the disturbance
harmonics
produced
the
with
of
at
introduction
centre
the
origin.
a sphere
by
-*

proofof the expansionconsists in expanding z" e'^''^"^^


simpleanalytical
the series of Bessel functions
for each power
in powers of z and substituting
suppliedby the formula of " 5'2 ; we thus find that
A

nV

cos 1^
gJZCOSC
"

i"-cos" (b
z.

"

__

w=o
*

t" cos"

_
"
"

Journal

t Wiener
Lxxv.

2"'^

w!

(^

2"+'* i^v+

2^') r (z/+
.

filrMath. lvi. (1859),pp. 104, 106.


lxviii.
Sitzuvgsberichte,
(2),(1874),pp. 355"367;

(2),(1877),pp. 904"905.

/I

^')

r-j

Lxxiv.

v-it-n-\-'ik\Z)-

(2),(1877),p. 128;

and

11-5]
If

we

repeatedseries by writingn

the

rearrange

fci{m

w=oA-=o

2" r

369

THEOREMS

ADDITION

(v)

Ik, we

"

deduce

that

Zh;)l

"

(^)C',/(cos"/)),
(i^+ "i) t'"-/,+",

m=0

and

this is

Gegenbauer'sresult.

Modified

forms of this

(3)

gzcos*

2" r

(v)
m

(4)

3=

Gegenbauer,are

^
-

"

+ "0
('^

cos(^cos"^) 2''r(,0 5 (-y^.{v

(5)

to

^-^^"1^
Cm" (cos"/"),

+ m)
(i.
0

{v) % (-r

e-~^cos"^ 2" r

due

also
expansion,

m=0

^-^^^^
G^n"(cos "/"),

2m)'^^^"^!^C''^(co

"^

"

(6) sin(^cos"/))2''r(i;)S (-)'" (i.+

2/h +

1)

'J"^^
i'+2m+i
"^

m=0

(7)

2"

:"

+ 2m)
(i^

^?i!

m=o

?:i"^"^|p^i"%i!i
/;."c"*,.."(e"s,)."=.,c/,=

(8)
The

\^)
'
'

(cos0),
,^' C'',",+,

last is

different method
These

which was
of Poisson's integral,
generalisation
when
in " 3'32. It is valid only
R{i')" "^.

formulae

to which

to

are

reference has

Equation (1) was


consideration

be found

pp. 363

on

"

365

of the first of

obtained

by

Gegenbauer'smemoirs

just been made.

obtained

of sohitions of

by Hobson, Proc. London


Laplace'sequationin space

Math.

Soc.

of 2i/+ 2

xxv.

(1894),p. 59, by

dimensions, 2i/ +

being

integer.

an

by
(f)

cos

generalset

more
cos

cos
(f)

of formulae

sin
-f sin cf)
"/"'

with respect to

may

be

derived

from

cos
(f)'
multiplyingby
-v/r,

(2) by replacing
grating
inte-

sin^""^^jr,
and

yjr.The integral*

sin'-""^
C'^i"
+ sin "^sin cf)'
cos -yjr)
(cos(^cos "/)'
-"/rfZ-v/r

which

exp

is valid when

Cf.
W.

that

sin^""^-i/r
cos
[iz(cos(^cos ^' + sin ^ sin "/"'
rf\^
-v/r)]

R (i')
" 0, shews

^-''(^^"'^)^"''(c"^^')"
^"rir'i^l'^^'

2--

{r (.)}^2

Gegenbauer, Wiener

B.

F.

^Itr^
'%^
'71^9

Sitzungsbertchte,lxx.

^'"''
^'""'
^"^^^
^"^' '^"
'^'^^

(2),(1874),p. 433;

cii.

(2a),(1893),p.
24

942.

370

THEORY

and

[CHAP. XI

FUNCTIONS

so

,"

12^
{z sin ^
t/"_j
Hill

t/ "_i

{z sin

The

This
v

sin

(D

sin
"f)

Sill

,,

proofconverges onlywhen K {v)" 0, but


of v, by analytic
continuation.

was
given by Bauer,
generalformula is due

the

\;

in the

for all values

result

"/"')
ro

0')"

used
integral

result is true

case

BESSEL

OF

Munchener

also Bateman,

192; see
(1899),pp. 189
to
Kapteyn,
Gegenbauer
"

v.
(1875), p.
Sitzungsheriehte,

Gegenbauer,Monatsheftefur

to

xxxiii.
Messenger,

Proc. Section of

Sci.,K. Acad,

(1904),p.
Wet.

van

te

Math,

182

and

the final

in the

263

tmd

Phys. x.

letter from

Amstcrdam,lY. (1902)

pp. 584"588.
of the formula

specialcases
Interesting

(10)

"TTi
z

2"

the expressionon
given by Bauer in the

11 '6. Bateman
shall

to

find that

case

(1) ^0 J"^(2^008
0

"*""'^ ^^ "^ ^"^^'^''


^"^
^^^'^'^
^^^
"" ^"^
^ ^"^
"

V"

symmetric

function

of

and

Z; this formula

also

\.

expansion.

establish the

now

/f27r)

the left is

that

We

we
integrate,

{''
Jv-\(^ sin 0) e'^cos.^sin'^"(^rfc/)
J

so

by taking"/"'
equal to "^or

multiplyby e^^^"^^sin^"^ and


put ^' equal to ^tt,

and,if we

was

obtained

are

cos

generalexpansion

"J")"/^
(2^sin ^

sin

"I")

a)

"^cos*^

CDS'"

^ sin" "^sin" ^

(-)"(/i+

.!r(;.4-. + l){r(.+ l)r^


X

which

o^i{-n,ix-^v +

l\

2'

is valid for all values of fx and

1) /^+^+27i+i
(^)

2;? +

".F.(-.,^+

n4-l;.

l;sm-c/")

1 ; sin- "l"),

with

the

exceptionof negativeintegral

values.
Some
discovered
solutions
We

of the results of

by

Bateman*

from

consideration

of the

proceedto
Messenger, xxxiii.

generalised
equationof wave
givea proofof the expansionby
(1904),pp.

of this

cases
"11*5 are special

182"188

; Proc.

London

of the
motions
a

Math.

which
expansion^
two

types of

examined

in

normal

" 4'84.

direct transformation.

Soc.

(2)in. (1905),pp.

was

111"123.

^!

~^ro

^-r

*=o

{v

*=*^-.r*"
0

"

r) \(r

(y

l)r {t

r) \{r

1),. (^

(i/ +

r!

00

(-

"), (/^ +

..

(.

.,

v-\-n+\:

cos'^

at

we

^)"_"

(-

1)

cos

4"

cos'^

(/"sin

"I") cTj, {z sin

COS

sm^

"i"

sm"

2^1

which

(-

??,

/A

Bateman's

2^

form

J^

of

sin=

c/,)

2^1

(-

fi-n,v-vn-\-\;

2^1

(-

/I,

(/x

/i

1/

v+l\

t-

sin-

^)

sin-

4)).

$)

'

n!

result

the

obtain

once

\-

1)

x(-r^^^~j|"..i^i(-n,A*4-i/

from

1),,_"

^._^,"

{-Y^~^^^.,F,{-n,fi

J^{z

r)\ (^

^i^,.

Hence

^z

r)\ {u

1),

"

^22

l)r+s-u

S-t)\{u-r)\{^l+

r=o

M=o

\)r+B-^,

^=0

r) ! (/a

s-ty.{u-

XI

."=,"

"

l)r (t

[CHAP.

FUNCTIONS

BESSEL

OF

THEORY

372

j;

the

^^

sin^

l;

1;

(.+

sin^

$),

expansion

is

evident.

1)

"^)

'^^

^^"
^

XII

CHAPTER

INTEGRALS

DEFINITE

integrals.
types of definite

Various

12'1.
In this

shall

chapterwe

various
investigate

which
integrals

definite

contain

the

integral
are
integrals
by
sign,and
evaluated ai-e, for the most part,of an obvious character; the onlynovel feature
is regarded
is the fairly
systematicuse of a method by which a double integral
surface integral
a
over
portionof a sphere referred to one or other of
as
a
those
The
most
are
two
integrals
interesting
systems of polarcoordinates.
These
discussed in ""12-2
12-21, which are due to Kapteyn and Bateman.
recondite
to be of a much
for no very obvious reason,
more
seem
integrals,
discussed in this chapter; their real sigcharacter than the other integrals
nifican
in
described
work
from
the
recent
has become
Hardy
by
apparent
in which the upper
and importanttypes of integrals,
" 12-22. The numerous
is infinite,
deferred to Chapter Xlil.
limit of integration
are
either Bessel functions
which

or

functions

of

which

methods

finite limits. The

have

similar character

under
the

"

reader

The

and

Sonine

here

may

due to
very importantintegral,
has alreadybeen established in " 11*41,namely

be reminded

Gegenbauer,which

r^^.{V(^^ + .--2Z.coscA)}

of the

..,,,.x,i^..

.^.

+ m)
^7r^(2^/

'^.+,"
(^) /.+,"(^)
Z"

"l^-KmlViy)
12' 11.

is valid when
of

in terms

The

formula

was

i^(/i)and R(v)
integral
involvinga

both
an

stated in

and it may
Schafheitlin:[:,
Math.

Ann.

J Die

Theoria

unaware

of

(1880),p. 36;

xvi.

(1884),p. 979.
t Nieinv Archief
been

'^"^

function

^^
^'
2TC+l)/r^^

/,+.+,(z)

(1)

and

integral.
first
finite

formula

The

which

Sonine's

z"

voor

der

'^^"^' ^ "^'""'' ^^^'

exceed

any Bessel
of lower order.

-1, expresses

Bessel function

different form by Sonine*, Rutgers-f


slightly
be proved quitesimplyby expanding the intea

see

also

Gegenbauer,

Wiener

Sitzungsberichte,lxxsviii.

("2),

IViskunde, (2)vi. (1905),p. 370.

BesseV sclicn Funktioncn

previousresearches

on

what

(Leipzig,1908),p.
he describes

as

new

31.

Schafheitliu

integral.

seems

to have

374

OF

THEORY

grand in

of

powers

[CHAP. XII

FUNCTIONS

thus
term -by-term,
integrating

and

BESSEL

f'/^{z sin 6)sin'^+^^ 0082"+!6'c;^

'Jo
2''r(i^+i)

U 2'^+''+-'"mir(/Lt
+ m-M)r
/

00

the truth of the formula


It will be

sin'"*^6

R{v)

"

the factors

the factor,we

as

1 and

"

fi is

+
T(/u,

l) in

ni

should

the denominators.

removed

have

we
unrestricted,

(-)f V^ (2

sin

which

J^ {z sin ^)/, {zcos ^) tan

(4)

'^+1

R(v)"

"
(/jl)
"

Hence, when

had taken

z
---^^r(/.)-

H^_j (2).

(1) is

from
ddd

Vfi

This may be proved by


making use
integrating
term-by-term,and finally
when

we

^^^^^^

d)sin"-^ ^^61

obtained
easily

is

If
I

have

\, we

"

integrand

have

i"^
.^

(3)
formula*

factors

the

particular,
by takingv

factor sin'^+i6 in the

f*V^(zsin 6)sin^-'^^ cos2''+^Ode

"

(2)
In

is obvious.

that the effect of the

observed

is to eliminate

/i^ V+''+2ni+i

m!r(yu.-F2/-fwH-2)'

^=0

and

\m

(2/+ 1)

i!-^^(^X

expanding/^ (z cos ^) and


of Lommel's
expansion

givenin " 5"21.


The

functional

F^ and F^^^
LIX.

i,

(1904),pp.
Some

in

529"552.

specialcases

Istituto Lomhardo
p. 92.

functions to be determined,
equation,obtained from (1)by substituting
by Sonine,Math. A nn.
placeof the Bessel functions,has been examined

of the

formulae

Rendiconti,(2)xiii.

of this

section

have

been

given by Beltrami,

(1880),p. 331, and Rayleigh,Phil. Mag. (5)xil. (1881),

Papers,I. (1899),p. 528.]


{Scientific

It will be obvious

to the reader that Poisson's

is the specialcase
integral

of

(1)obtained

by taking/x=-^.
For
papers

develoimientsof the formulae of this section,the reader


by Rutgers,Nieuiv Archiefvoor Wishmde, (2)vi. (1905),pp. 368
some

"

should
373

consult

(2)

vii.

two

(1907),

pp. 88"90.

The

12" 12.

proofof Sonine' s first


integral.
geometrical

proofof the formula of the precedingsection depends on


in " 3 '33)of integrating
of the surface of
the device (explained
a portion
over
various axes
of polarcoordinates.
a unit spherewith
An

instructive

If (I,m,
*

n)

are

Due

to

the direction cosines of the line


Eutgers,Nieino Archief voor

joiningthe

Wiskunde, (2)vii.

centre

(1907),p. 175.

of the

12'12]
sphereto

an

375

INTEGRALS

DEFINITE

of surface dco whose

element

0, it is evident from

longitudeand co-latitude
of Poisson's integral
that
application

an

^ and

are

"

r(fi + l)r (i)(1zy+'

f J"^(^ sin O) sln'^+id cos^-'+iOdd


J
0

I"IV^s'""'cos0gin2f^+i^cos2''+
(|2X+''+'

Jo .'o

(1^)"+"+'
11

e^^hn"!^71-"+'d(o

11
{lzY+''+'

e''^"
dco
l'^7)1'"+'

0 (,og2^ ^
gin-'^+2."+2
ji V^cose
(^zy*''+'
Jo

2r(/. +

the truth of Sonine's

and

An

Jo

f)

Bessel functions which

be evaluated

can

by the

is

device*

same

i.

is obvious.

formula

integral
involvingtwo

c^^
sin^-'+i
"/)(/(/"

(9cos^-'+i6 dO,
J,(z sin26) /, (z cos2 6) sin^-'+i

J 0

in which, to
If

convergence,

Riy)

"

I-

"

write

we

w^

and

secure

sin^6^+ cos^^

1 11*41 (16),we

use

see

I 'lA^
^_1M!_^ [*''

2 sin-^ cos^^

that the

cos

"^
=

sin^ 26

cos^(j",

is equalto
integral

"

(Izy

2^"+^ r

so

f'^ri'^
J^{^v/(l-sin2^cos"(f))|

-(l-sm-^^cos-^)^-'^""
2^-^^r(:+i)r(l)JoJo
2--r(.

^,/,

"

o,

7^

7^

i)r(i)jj,^o,.^o
(1-^^)*"

(z/+ I) r (1)jj ,^0,

2-^^ro.

sin^^+i6 cos'"-^'e sin^"6 deb dO

."o

^)r(i)j-iJo

(1

n')^"

J, (2 sin 6) sin-'+i^

cos-"

cos-" 0 dd dcj)
(/"

thatfinally,
by |12-11 (1),
(1)
*

I*V,{z sm'0) J, {z

cos'

^cZ^
6 cos-=''+^
6) sin'^''+^

integralhas been evaluated by a different


kunde, (2)vii. (1907),
p. 400; cf. also " 12-22.
This

\^^''+^r\v
l)z^
'

method

by Rutgers, Nietiw Archie/ voor

Wis-

376

THEORY

Some
have

integralswhich
the subjectof

simple example

As
reader

may

researches
of

an

that, when

prove

[CHAP. XII

FUNCTIONS

this,but which

resemble

beeu

BESSEL

OF

much

are

by Bateman, Kapteyu

which
iutegi-al
R{v)"
i,

Rutgers;

be evaluated

may

cos
{x^ t^)h''
-

t
.

/, W(x^

The

second

{z
\y,

sin

6)JaZ

is valid when

A
and

'

t)sm

0) by B6cher, Annals

of Math.

(pau(((ji.
viii.

(1894),p.

136.

finite
integral.

equationby

cos

both

and, in fact,he obtained


of the

de\dce, the

same

formula

(1)
which

given (with

Sonines

12'13.

evaluate,

12 "2.

the left in the form

2''+ir(v+|)r(i)jo jo
was

the

"

^^=2^?r(l
a)

1^)}

by writingthe integralon

formula

by

see

"

This

difficult to

more

and

d) sin'^+^^

then

of

" 12'11

simple method of proving the formula


verifythat the terms of degree fi+ v

'^^^^^^^^^^

"

1, is also due

from

it

by

to

Sonine*

dividingboth sides

Z ^0.

making

to

(fi)and R{v) exceed

the formula

Z" and

cos-'+^Odd

is to
+ 2m

expand

the

integralin

the left combine

on

to

powers

of

and

form

+ 22)'"
(-)'-g^Z'-(Z2
'

ml
The

proofby

Wo
the

this method

exceed

"

Ti
^

of

|,

we

unit

7n

sphere. Under

2)

formula
the

that,with the notation

see

is left to the reader.

establish Sonine's

proceedto

surface

+
T(fjL

over
by integrating
portionsof
that
R(fi) and R{v)
hypothesis
of " 12'12, we have

/M(^sin^)/,(^cos^)sin'^+'^cos''+^^(^^

w/iyw

\\ JI e''^""^^'=""*+'^"=""^"=""'''sin2'^+i^cos2''+i^sin=

Jo

^izi+izncos^i
11-"+^sin^" -fdcody^
^y^iH.' OJ .' m"0,

n"0

If

J oJ J n^O.l^O

dwdyjr

e^"'n"(^cos*+zsm"/.cos.A)cos2M^sin2''+i0sin'''+-^sin
J

JO.'O

Qizm+iZlcos^'
"2M [iu+iginS.^

-hw

iT|

II

J.

JJ

n'^0

gisuie(zn+ZD^^2ucos-''6sm-''+-edQ)dd

?)i"0

I/

gisin

(zl+Zm)

0 sm'"'+^ ddw dd
j^2u(josSf^

e"sinesiii.#,ucos^+zsmWcos2''"f)sin"/)Cos=^^sin-''+*

Math.

Ann.

xvi.

(1880), pp. 35"36.

12-13]

377

INTEGRALS

DEFINITE

function of
function involved here is a periodic
analytic
the'exponential
the
limits
of
with
and
integration
yjrwith period27r,
so, by Cauchy'stheorem,
respect to yjrmay be taken to be a and 27r + a, where a is defined by the
equations
Now

"ST cos

and

for

+ a
"\/r

z,

-CT

adopt these
becomes
the
integral
-v/r, triple

\/(^'+ ^')- If

OT

we

q\^ Sin

sni

cos

sin

limits of

Z,
and
integration,

Sm
(^COS''' Q
"/)

A" coS""

+"-

Sm^"

then

write

f/^,
f/"/)
c/-v/r

integral
may also be obtained from its precedingform by replacing
On retracing
the steps of the analysis
with these
^
by tsr and Z by zero.
reduce
the
t
o
substitutions we
triple
integral
and

this

iff /'TT f

TT

"f"
sin""l/r
g/-Grsin0cos
sin"^'^+'
^ C0S-''+i^sin-'^"/)
f/0fZx/rc/(9,
J

+ 1)
r(i-'

r(^

i)r(i) p
Jo.'o
r(i;+l)

and

obtain

we

J j "i^i),n^O

Sonine's

formula

by

comparison of

the

initial and

final

expressions.
Sonine's

proof of this formula was based on the use of


would
and the process of making it rigorous
integrals,
own

infinite discontinuous
be

longand

tedious.
The
and

"

formula

i^i^(i/)"
Sonine's

In

be extended

may
"

1, by

the domains

in which

"h'^R(fi)"-l,

multiplyby
by ^/{Z'+ ^-" 2Zl^cos (f"),
from
11-41
follows
It
"
(16) that
integrate.

formula, replace Z

+ ^--2Z^cos0)i^
sin'''(/)/(^-

(2"^

to

continuation.
analytic

and

[V^ {zsin d) J, (Zcos 0)J, (^cos6) sin'^+i^

cos

6(16

"

^^^-r
r /M..^.{v(^^+^--+r-2irrcos(^)}
^
^
+ Z^' + ^"
2Z?cos"^)i"^+''+^'
2''r(v+ ^)r{^j}o (^"^

^'

providedthat
R{^)"-1,
This
section

result is also due


the

reader

(1903),pp. 442, 443.

should

to

Sonine, ibid. p.

consult

Macdonald's

R{v)"-h
45.

In

connexion

memoir, Proc.

with
London

the

formulae
Math.

Soc.

of this
xxxv

378

[CHAP. XII

FUNCTIONS

integral.
finite
Gegenhauer's

12*14.

first of Sonine's

the

resembles

somewhat

integralwhich

An

BESSEL

OF

THEORY

integrals,

namely
'^

cos

{zcos
has been

evaluated
method

usingthe

cos

6 sin

J^_j{zsin
y\r)

0/ (cos6)sm''+i6dd,
-v/r)

sm

It is thus

by Gegenbaiier*;we shall adopt our


unit sphere.
of integration
a
over

procedureof

normal

that

seen

r gzzcosflcos^
j^_^(2sin d sin

C/ (cos0) siw+iddd
i/r)

Jo
(-^zsin '\lrY~^C'^ f'^
^
e'' ^*'"' '""
r. /
N -n; 1

**'^"

*+'*"

"'"'*^ C/
"''

6 sin-"-'"i dd. cZ^


sin^-"
(cos6*)
^

{v)i {^) JoJo

iU sin
A

i|r)_"-*

(J^)l(i)

J.

m"0

(i^ sin aI^)"-*

Qiz(?COS

^)

*+"nsin

(^^v
(^)^j2,.-i^^^

(^z sin "\!ry-*ri''r-''

xrTAx
Hy
1
A

(2)

{v)

(^) """(^) J 0

0) cos-"-"

6 sin

d^
^c^"^

+ -f)}cos2"-"^sin(9rf(/)fZ^,
e'^"""^c"*"^C/{sin^cos(0
J 0

penultimateintegrandis

since the

cos

Jo

NrT/ix

T^/

e^^sinecos(*-^) c^..
(gin^

Jo

periodicanalyticfunction

of

"^ with

period27r.
If

find

we
predecessor,

immediate

instead
usingthe last integral
analysis,
is equalto
that the original
integral

the steps of the

retrace

we

(^0smijr)"-"
I ^.^^^^^ (I
(^) i (2)

n'"-'
y\r)

day

""o

11
^^^f^^^}\

sin

yjr-m

COS

of its

e'^" Or" (n cosf-l

sin

wi-"-i dw
yjr)

p;%ucosec^.(cos"/rcos^-sin"|rsin^cos"^)si
^(^^sm^r)"
I
I
"'O-'o
(2)

(^)

Now, by

the addition

6
cos
0/ (cosyjr

,V.

Mo

"

sin

sin
yjr

theorem
^

Vi/^

cos

for

Gegenhauer'sfunction,

(jj)
^

(2v + 2/j
-

"

Sitzungsberichte,lxxv.

Wiener

t This

(1893),p.

was

942.

(2),(1877),p.

proved by Gegenbauer, Wiener

221

1) sm^^

sm'^-v^

(cos^|r)
C^*(coSc^).
:;(cos^) C;;:;
and

lxxxv.

Sitzungsberichte,lxx.

(2),(1882),pp. 491"502.
(2),(1874),p. 433; cri. (2 a),
"

380

OF

THEORY

BESSEL

[CHAP. XII

FUNCTIONS

that is to say

r~J^{t)J.{z-t)dt
2 i (-r

(2)

n=(i

importantdeduction

An

from this result is that,when


M

[z

(/x)"

0 and

R{v)

"

"

1,

[z

Jo
so

J-^+,+,"+,(4

.'0

J 0

that

'^^^"^.
fMt)JA^-t)^

(3)

Jo

This formula

is due

Bateman*;

to

independently
by Kapteynf, who
It will be observed

that

specialcases had
values of
integral

some

considered
deduce

can

we

/"

(2),combined

from

obtained

been
/x

and

only.

" 2-22 (2),

with

that
sin 2,
rj^(t)J.^(2-t)dt
j J^(t)J,_^(z-t)dt

(4)

Jo

(/x)" 1, and when

"1"R

when

"1"R

and
v
By interchanging
/x with
then
R(v) are both positive,

and

Jot
It

which

(3)

with

"

from

{/x)"
2-

J, (z)-

cos

z,

respectively.
^

"

in

(3),we

vj

\fl

give the

to

unnecessary
Kapteyn deduced

describe the

complicatedinductions

somewhat

the

that,if R (fi)

see

in which

special
on
disquisition
by Rutgers;): the subjectof
case

ix

the formulae

v=\,

ot

by
to

generally.

integrals^.
Kapteyn s trigonometrical

simplerformula

seems

12'21.

Jo

than

those

justconsidered

is

1 cos{z t)Jo{t)dt=zJo{z).

(1)

"

Jo
To

prove this,we
verified that

put the left-hand side equal to

u,

and

then

it is

easily

d^u
J-

and

therefore
w

where

and

are

constants

Jo C^)+

2"

of

-4

cos

z-\-

B sin z,

integration.

(1905),p. 120. Some similar integrals


occurringin the theory
the
ibid.
iv.
writer,
(190G),p. 484.
integralequations
by
(2)
sapje
K.
Section
Akad.
Proc.
of
Wet.
Sci.,
te Amsterdam, vii. (1905),p. 499; Niemv
van
t
Archief
Mim.
20"25
de la Soc. E. des Sci. de Liege, (3)vi. (1906),
5.
no.
voor
Wishunde, (2)vii. (1907),pp.
;
405.
385
Nieuiv
Archief voor
Wislainde, (2)vii. (1907),pp.
X
" 3Iem. de la Soc. E. des Sci. de Liege, (3)vi. (1906),no. 5.
*

of

Froc. London

Math'. Soc.
are

(2)ra.

examined

"

12-21]
Now, Avhen

and

J-

so

is small,

the result is established.

0, and

381

INTEGRALS

DEFINITE

(1) by differentiation that

It follows from

\\\xx{z-t).J,{t)ilt
zJ^{z\

(2)

Jo

and, by

integration,
partial
sin

I sin (z-t) Ji (t)dt

(3)

zJ^ {z).

Jo
formula

The

sin

(4)

{z

1)^^

-^

is valid when

of the

precedingsection

is

character,and the result

elaborate

more

requiredto

write

(-r J,^,n-,^
{z),

Mw

(/i)" 0, is of

which

We

dt

Jo

it.

prove

Jo(z-t)J^(t)dt,

v=

Jo

and

then

we

have

J
dz-

{J" (z-t)

v=

t)\J^ (t)dt

j; {z)

j;{z)
\^^'^^^JMdt
+

Jo

J, (z

"

j;(z)
fMz-t)'^dt
+

fiJ^(z)/z,
.

by " 12-2.
of variation

the method

By

of parameters

(cf "7'38),we
J

[^
v

Acosz-\-Bsinz

fx\ sm(z -t)

-^

when

^^^

is small,it follows that, when

we

obtain

the

(/x)" 0,
=

0.

requiredresult.

(4) with respect to


By differentiating
(5)

{z''+%

_^

A=B
Hence

(t)
dt,

Jo

and, since

"

deduce

f'cos
(Z

"^

t)

dt=-

we

find that

(-)" en J^+,n{z).

that

382

THEORY

12"22.
As

f*" J^

(1)

is valid when

which

method

The

it involves

is

the

"
(fi)

i and

elaborate

more

of infinite

use

XI

of

we
evaluatingfinite integrals*,

shall

(cf." 12'12)

6) J, {zcos2 6) sinV

[CHAP.

FUNCTIONS

powerful method

very

the formula

iz sin2

BESSEL

of evaluatingfiniteintegrals.

of

typicalexample

give a proof of

now

method

Hardy's

OF

^^~^l '^''^r,l\^
V^^^^i^

6 cos^" ^^6d6=

R{v)" -h.

than

other

any

integralscombined

method

described

with

applicationof

an

iu this

chapter,because

Lerch's

theorem

null-functions.

on

fin

Let

sin2 6) J,
J^ {zr"'

from

By changing
v^13'2 (5)we

exp

r2 0

12 '3.
A

Chessin's

the truth

substitute

the formula

in

x"^"
t)J^ {zx^)

^dx

of Lerch's

obtained

been

by Chessin,

I must

express

curious

this

dy

to

American

Journal.,xvi.

of the

ascending

series for Y,,

(5),we

obtain

question,namely
2(y + logi2)J"(2)-

thanks

my

publicationof his

many

(-y^t)Jy.{zy^)
yV

+ m

rii-^n

-/
*

exp
0

theorem, /i(r) is identically


equal

\'m-''-Kn-m-r)\
in

m=o

the

using

then

this result in the coefiicients

Y"(2)

(1)

and

the formula

from

we

(x,y)

for Y" (2).


integral

(1894),pp. 186"187,

if

f^ (r),

of the formula.

integralfor Yji (2)has

curious

coordinates

Cartesian

to

modification

obvious

an

this establishes

q ^os^" ^^6d6

/ exp(-j-2i;)./2(r)o?j-,

f^(r); and

x^

^{^2^.+

.'LI

'

hence, by

exp

2.' + 3

(r,6)

t"\I{z)\,

c??A (?")

"

Jo

and

coordinates

that, whenever

see

polar

{zr^cos2 6) r^'*+

integrals
;

dt.
\-t

to Professor

developments

owu

and

the reader

Hardy

of it. The
use

may

for

communicating

method

was

it to evaluate

used
the

the

method

by Ramanujan

integrals examined

to

me

before

to evaluate

earlier in

chapter.

t Acta

that,

Mathematica,

if /('")
is

xxvii.

continuous

(1903),pp.
function

of

j
for all

sufficiently
largepositivevalues

839
when

"

The

352.
?" "

0,

form

such

of the

that

bX'p(-rH).f(r)dr^O
of t,

then/(r)

is

zero.
identically

theorem

required here

is

CHAPTER

XIII

INFINITE

which
integrals
the

under

chapter is

this

subjectof

The

integralsign.
they

The

functions

methods

of

consist, for the

(I) Expanding the Bessel

of

similar character

evaluating such
part, of the

most

function

functions

or

classes of infinite

various

of
investigation

the

either Bessel

contain

numerous

very

integrals.
types of infinite

Various

13*1.

INTEGEALS

integralsare

followingdevices

of its argument

in powers

not
:

grating
inte-

and

term-by-term.
(II) Replacing the Bessel function
of the

and
integrations,

then

the

function

order

weighty

the

Bessel

function

integralsign, they

the

integralof
order

(VI)
contour

order

of

function

under

spiteof

interest

in many

to

Math.

Wiener

the

as

integral of
the

order

product
single.

of the

orders

but

of the

tegrations
in-

by

Neumann's

argument

same

be

may

replacedby
("5"43), and

formula

the

has

care

the

Ann.

xvi.

the

values

taken

length

(1880),pp.
has

Sitzungsberichtc.

been

Mathematician,

possibleto
to

incompleteness

with

procedure

been

the
not

the

give

chapter

out

the various
selection

by

the most

examples by

chapter,its
on

And

than

more

several

integralsign are
they are of extreme

but

Physics.

will be worked

evaluate

of this

of the

under

functions

of Mathematical

it is not

the

systematic manner.

Pure

the

branches

that

unfavourably
t This

Sonine*

changed.

integralsinvolving Bessel

whose
important integrals,

occur

(cf " 12*1), and

function

investigatedin

sooiumerous

In

the

carrying out

integralsign may be replacedby


functions, and
integralof Barnes' type ("6*5) involving Gamma
has
the integrationsis then changed ; this very powerful method

importance

methods

order

same

replaced by

formula

integrationsis then

only of great

most

then

systematicallyby

out

Bes.sel's

of
generalisations

integralsign,the product

the

under

Bessel

The

Infinite

are

of the

of different

single Bessel

of the

previouslybeen

not

functions

two

the

of the

changed f.

product

one

carried

be

may

by Gegenbauer's

(V) When
as

been

functions

Bessel

two

is then

occur

integrations.

the order

memoir.

(IV) When
under

by

integral,
changing

the

and
integrations,

of the

integrations;this procedure has


in his

Poisson's

carrying out

(III) Replacing the Bessel

integral,changing

by

length

finite

must

each

types
of

the

suitable
method.

be contrasted

integrals.

33"60.

carried

out

by Gegenbauev

in

number

of

papers

published

in the

384

OF

THEORY

The

It

shewn

root

HankeVs

generalisations.

by Lipschitz*that

(!)"
where

[CHAP. XIII

FUNCTIONS

with
integral
ofLipschitz,

13*2.
was

BESSEL

ly.KU(U)dt
-^^.
=

limit of inat the upper


tegratio
(a) " 0, and, in order to secure
convergence
both the numbers
That value of the square
R{a " ih)are positive.
is taken which makes
|a + \/{a-+ lf)\"\h\.
R

this result is to replacethe Bessel


simplestmethod of establishing
Parseval's
and
then change the order of the integratio
("2-2)
integral
by
without
which
be
It
is
thus
a
justified
procedure
difficulty.
may

The

coefficient
"

found

that
roc

TT

./0

cie
IT

the formula

is

consider

Now

ih

"

cos

1/V(a"+ 6"),

and

jQ

proved.

the

generalintegral

more

je-"W,{bt)t''-Ult.
in all its generality
by Hankelf, in a
integralwas first investigated
time as the appearance
memoir
of
publishedposthumouslyat about the same
These
writers proved that,if R(/j,+ v)" 0, to
two
papers by Gegenbauer:[:.
and the previousconditions concerninga
at the origin,
secure
convergence
then .the integral
at infinity,
and b are
to secure
is
satisfied,
convergence
equalto

This

-A
-^

To

establish this result,first suppose

that

I6 I" ja |. If
term,

we

expand

the

'

/^+^+^,
'

a'^r(i.4-l)
(I.4-1)

'

l;_^^

"

'

b is further

integrandin powers

of b and

aV
restricted

so

that

integrate
term-by-

find that

we

I e-"'UAbt)t"^-^dt= X A-iir^^-^ rJ
I
m=ow'

Jo
/o

r"i=o

'

Journal

{v +

ncTVlv+

l)Jo

1)

t'^-"'^"-"'-'
e-^' dt

a'^+''+^'"

(1859),pp. 191"192.
(1875),pp. 467"468.
ibid, lxxii.
1 Wiener
(2),(1876),pp. 343"344.
Sitzuvgsherichte,lxx.
(2),(1875),pp. 433"443;
lu thejormer, the special case
ix"v-^\ was
investigatedby tlie integralgiven in "3*32; in the
the generalresult by substituting Poisson's
integral for Jv(bt).
latter, Gegenbauer obtained
t Math.

fur Math.

Ajin.

viii.

lvi.

13-2]

since |6 [" |a |,and


absolutely,

final series converges

The

385

INTEGRALS

INFINITE

the process

so

of

Hence
is justified*.
integration

term
term-by-

re-^^J.{bt)t^-'dt

(2)

Jo
{fi+ v)

(^b/ayr

(iM+

fi +

6^

The

R (a) " 0 and

proved onlywhen

yet, been

result has, as

I6 |" Ia

|;

but, so longas merely


R

(a + ih)"

continuation,(2) is true
analytic

for this

Again,by usingtransformations
forms :
written in the following

I"e-"'
/. (ht)V'-'

(3)

R (a

functions
analytic

then both sides of (2) are

be

and

ih)" 0,

"

of h ; and

extensive

more

the

by

so,

of
principle

of values of h.

range

functions,(2) may
hypergeometric

of the

dt

Jo
r (^ + ,.)
i^

(Way

ai^Tip + l)

M"i
^^_ /._ir^-^
a-J
'\

(^byVifi + v)

ff^+
"f^^^+"^r{v+l)- '\ 2

~(a'+

l-f^

^,

'

q^
2

_
'

,
'

.
_

'

^^
a-

b^-

i^.
'

-,

,
"

'

'

bV

a' +

by Gegenbauerf in expressingtoroidal
of (2) are
functions as infinite integrals
cases
; special
requiredin various
researches,of which those by Lamb J may be regardedas typical.
physical
Bessel

By combining two

(4)

e-"t

"

cot

vir

cosec

is

obvious

general

XProc.
also

under

W.

2^1

^^^

2'^i

the

V.

B. F.

i ;

v+

'^

"
'

"

"

"

..

,.,

"

a^

"

rs
6V

of this formida are due to Hobson,


cases
{a " ib)" 0; special
and
Heaviside, ElectromagneticTheory,iii.
(1892),p. 75,

xxv.

of the

specialcases
interesting

formulae

far discussed

so

may

be

Theory of InfiniteSeries, " 176.


involve

terms

London

that

",
;
2
'\ ~^r~2
("2^.^,-2^i(M-.')r(i-v)'

Soc.

toroidal

functions

and

"

766; Gegeubauer

Bessel

functions,

as

also

expressedseries,

integralswith

Bessel

integralsign.
Math.

Macdonald, Proc.

Phil. Soc.

c.
Sitziaigsberichte,
(2),(1891),pp. 745

t Wiener

functions

to deduce

85.

that

Cf. Bromwich,

whose

it is easy
functions,

J-

VTT

Math.

(London, 1912),p.
It

used

l\(bt)t''-^dt

providedR{ix)"\ Il{v)\and
Proc. London

been

(2) has

formula

"

The

(1886),pp.

Soc.

London

xxxiv.

Math.

(2) vii. (1909),pp. 122"141.


(1902),pp. 276"284;
443 and Basset, Proc.
Soc. xxxv.
(1903),pp. 428
"

See
Camh.

425"4.33.
25

386

THEORY

obtained

by choosingfi and

functions.

OF

that

so

BESSEL

the

Thus, by taking /x equal to

hypergeometricfunctions

+ 1

or

obtain

2, we

reduce

to

elementary

the results

j\-a^JAbt)t^dt
^^^^l^i^,
b'-y+i
^tt'

(5)

Jo

(a'+

jo

provided that R{v)"


These
pp. 433

formulae

Camb.

obtained

also

Other

respectively.

Lxx.
by Gegenbauer, Wierier Sitzungsberichte,
(2),(1875),
and
noticed by Sonine, Math. Ann. xvi.
(1880),p. 45;
Hardy,

(1912),p. 12; while Beltrami, Atti della R. Accad.

xxi.

(1880"1881),

various

"

'

b-^y+^^TT

obtained

were

Phil. Soc.

Torino,xvi.

{a^+

", ^(i')"

"

443; they were

"

Trans.
di

[CHAP. XIII

FUNCTIONS

^^.

specialformulae

specialformulae

delle Sci.

203, and

(4) ii. (1880),pp. 461"505,


Bologna Alemorie,
by taking^ 1 and v to be any integer.

has

are

W("^+'''-'"'
f%-.....(60^'
vb"

(7)

W{a'

e-''iJ^{bt)dt

(8)

b^)-a}''

143,
by Pincherle,Bologna Memoric, (4)viii. (1887),pp. 125
from the generalisedform of Bessel's integrals
(" 6'2)by
Laplace'stransformation
(cf." 9"15). This aspect of the subject has been studied by
Proc.
London
Math.
Soc. xxxv.
Macdonald,
443, and Cailler,3le'ni.de la
(1903),pp. 428
iSoc. de Physiquede Geneve.,
368. The dift'erentialequations
xxxiv.
(1902 1905), pp. 295
satisfied by (5) and (6), qua functions of a, have been examined
by Kapteyn, Archives
Merlandaises,(2)vr. (1901),pp. 103"116.]

[Note.

that these

It

observed

was

"

derivable

are
integrals

"

"

"

(1863),p. 46, as
not

to

seem

series for it

be

tdt
Jo (bt)

obtained

was

limit of

series of

capable of

Legendre

being
by using the expansion

evaluated

cosech

7rt

Lxxii.

(cf." 14"64). The integraldoes


terms, though it is easy to obtain a

functions

in finite

fiirHath.

Journal

by Neumann,

Trt

Sinn

(2"+ ^)^t.

n=0

series which

Some

more

converges

of the
integrals

(1873),pp. 92"102;

and

same
more

b is

rapidly(when

large)will

generaltyi3eare given by Weber,


recentlythe formula

+ ^)
rJAbt)t''dt^{2b)''r{u
I

Jo
which

13*21.
It

e^i-l

Jn

is valid when

Soc. R. des Sci. de

The

be obtained

Pt{v)"0 and |7(6)|""-,


Liege,(3)vi. (1906),no. 9.

been

in

"

13-51.

fiirMath.

Lxxv.

+
n=i(7i2^2

has

Journal

^,'y+i'

obtained

by Kapteyn, Mem.

de la

as
Legendrefunctions.
Lipschitz-Hankel
expressed
integrals

by Hankel that the hypergeometricfunctions which occur


integrals
justdiscussed are of the special
type associated with Legendre
in terms of toroidal
the integrals
functions;subsequently
Gegenbauerexpressed
functions (which are known
and a
to be expressible
as
Legendre functions),
was

noticed

in the

littlelater Hobson*

gave
*

Proc.

the formulae
London

Math.

Soc.

in
xxv.

some

detail.

(1893),pp.

49"75.

388

THEORY

BESSEL

OF

[CHAP. XIII

FUNCTIONS

expresses a function of cosh a in terms


generalformula of the same
type is

of

function

of coth

The

a.

more

QlZl(cosha)
f"e-^'^o"i^"^/^m^'^-irf^
cos

(6)

sin

'o
In these formulae, R{fji1^
On

replacing?/ by

i/

"

and

v) "^
in

VTT

a*
TT V(2'^)-sinh'^-*
(/i+ i^)

(cosha) "\.

find that

(6),we

P*"^(cosha)

r-

e-t^'''^'^K,{t)t^-'dt=^|{h^'^").T{^JL-v)T{^JL^
,

(7)

* a.

Sinn'*

J
^

and this formula


If

we

is valid when

take cosh

result

deduce

0, we

(/x)"\

given by

Heaviside*

case

hence, if v

1.

T'

TT

sin

"

arc

"^^

so, when

\I

sinh

v'Cot^1)

^'"'^ ^"^

-^^^-^"^

va

snih

V(l

"*^)

*rc

cos

_________

=._^^__.

{b)\"l,

r""

"

j^sm{bt).Koit)dt

/n\

(11)

of these is due

former

sinh

find that

replacea by "zb, we

we

VTT

0,

/:

The

"

0.

Jo

and

(cosha.)"

that

in the

[^

,".

If

fi=l, (7)becomes

When

and

and
(i/)\

/JKAt)t--^dt=2'^--rl^^)r('^y

(8)
a

to

/7

r^

sinh 6

arc

-j^^-^.

Basset,Hydrodynamics^ii. (Cambridge, 1889), p.

32.

from the aspect of


[Note. Various writers have studied the Lipschitz-Hankelintegrals
have
we
coordinates,
potentialtheory; to take the simplestcase, if (p,0, z) are cylindrical
1

/
It is

suggestedthat, since

function
potential
it is

e~P' J,,{^t)is

on
function,the integral
potential

the left is a

real space except the originand that ont the plane


of a unit charge at
be the potential
it is inferred that it must

finite at all

pointsof

equal to 1/p,and so
of the integral
to precludethe possibility
origin.But such an argument does not seem
at
the
being a potentialfunction with a complicatedessential singularity
origin,and so
this reasoningmust
be regardedas suggestiverather than convincing.

2=0

the

ElectromagneticTheory,

t On

the axis of z, the

in.

(London, 1912),p.

integralis equal to

269.

constant

divided

by \z\.

13-22}
various

For

researches

the reader

Hafen, Math.

consult

may
based on

Ann.

(1910),pp. 517

lxix.

"

of
integrals
For

537.

this section,

ments
develop-

some

potentialfunction

the

(1912),p. 94.]

Bateman, Messenger,sli.

see

the aid of the

potentialtheory with

on

389

INTEGRALS

INFINITE

f
m

Applications
ofthe addition formula to

13"22.

the Lipschitz-Hankel
integrals.

It is easy to deduce from the results of the


with "|11"41
(16) that,if all four of the numbers
and

21/)" 0, while

R{}i +

(1)

sr

is written

in

placeof

precedingsections
R(a"ib"ic)
26c
+ c^/(b'
"

^^

,F,r-~^-,'^
^
2
2

"v

"

'

reduces

function
hypergeometric
so

we

to

'

l;-~-]sm'''cf"d"t
a-^
'

elementaryfunction

an

if /x

or

2 ;

("^i|^)

J, m J, (ct)
/;.-.
^
dt

case

'

have

(2)
The

then
cfi),

^^

'\
"7ra"^+"'r{2v+l).^o

and

positive

are
cos

e-"' ./,(bt)
./,(ct)f^-' dt

-V{v-\-h)r{h)]Jo

The

combined

^l

be derived

2 may

Q,.,

from

this

differentiation with

by

respect to

a.

of theiu,have been examined


These formulae,or specialcases
by the followingwriters:
505; AtH della R. Accad. delle Sci. di
Beltrami, Bologna Memorie, (4)ii. (1880),pp. 461
Torino, xvi.
(1880 1881), pp. 201
205; Sommerfeld, Konigsherg Dissertation,1891;
"

"

"

Gegenbauer, Monatsheftefiir Math,


Moth. Sac. xxvr.
London
(1895),pp.
By taking /x=

"

1, v

l in

so

that the

Phys.

v.

(1894),p.

55 ;

and

Macdonald, Proc.

find that

(1),we

/.^-at'-'il^;cu^

"Itt

t-

und

257"260.

;^/("2
+ 2-2cos"/))-a}(l+cos0)(^0,
J

,,

encountered
by Rayleigh,Phil. Mag. (5) xlii.
integralon the left,which was
is
an
as
iv.
expressible
(1904),
elliptic
integral.
[Scientific
Papers,
p. 260],

(1896),p. 195
An

(S^

which
integral

may

be associated

at

I, {bt)K,(ct)dt

(1) is
de

"icos

.'0

.0

+ (b + c)--4"bcsin'd]'
\/{a-

by Kirchhoff* as earlyas 18.53; the reader should have


with " 11"41 (16); it
in deducing it from " 13"21 (10) combined
difficulty

This
no

with

was

discovered

is valid if all the numbers

R(c"b"
are

ia)

positive.
*

Journal

fiirMath,

xi.vin.

(1854),p. 364.

390

THEORY

somewhat

OF

BESSEL

[CHAP. XIH

FUNCTIONS

result,namely

similar

I e-"'U^-''J^ibt)J^{ct)dt

(4)

J0

(v+ ^)r (^)

("2+ 2iac

cos

"^
-

c2 cos2

'

+62)^+^
"/,

and
is valid when
It (fi)""h, is due to Gegenbauer, Wie?ier
R{a"ib"ic)"0
Lxxxviii.
(2),(1884), p. 995. It is most
Sitzungsherichte,
easilyproved by substituting
of Poisson's type for the Bessel functions.
In the memoir
cited Gegenbauer has
integrals
also given a list of cases
in which
the integralon the right is expressible
by elementary
which

functions

(cf." 13-23).

13'23.

Gegenbauers

deductions

from

the

and
integrals
ofLipschitz

Hankel.

Gegenbauer,Monatshefte
filrMath, und Phys.iv. (1893),
401, is obtained by combining the results of " 13"2 with the integral
pp. 397
formula of " 5*43 for the productof two Bessel functions;it is thus possible
to
which
involve two Bessel functions by
integrals
express certain exponential
functions *. The generalresult obtained
of trigonometrical
of integrals
means
by Gegenbauer is deduced by takingthe formula
A

formula

due

to

"

2 r*''

J^(bt)
(bt)
=

ft.

cos
(^btcos (f")
{/jL v)(bdeb,
Jy,+v
"

"

ttJ 0

from
it by e'^"-^
f^^" and integrating
multiplying
if R (a)"\I (b)\and R (ti
+ v) "
^,then

0 to

; it is thus found

that,

re-''''^J"(bt)J^(bt)t"'-^''dt=[ j^e-'''^J^^^{2btcos4))t''+
Jo

"^Jo Jo
=

f'"[6--"*/,.+.
{2bt

cos

cos
(/ji-v)(f).
(j")i''+''
dtd(f"

TTJO Jo

The

inversion

of the order

2 r^- (46cos

0^+" Tifi+v

;;-Jo
(4"-

46- cos-

of the

l)
'''
^^
V^
c{"y-^-i
+

integrations
presents

no

'^ '^"^^^-

great theoretical

difficulties; hence

(1)

e-"-^'J^{bt)J,{bt)t"^+Ult

Jo
COS
T{fl-\-V+ \)b^^" fi' COS"^+''(f)
(fl v) "f)
*^"
TT^
Jo (a'+ b'-cos'(f"y+''^i
-

This

result,in the specialcase

Beltrami,Atti delta R. Accad.


As

particularcases

of

(1) take

/i

0, had

equal to

been

obtained

0 and

to

It is found

1.

'2nbJ{a^-\-b^)'

'

an

earlier note

previouslyby

(18S0" 1881), p. 204.

See also

Torino,xvi.

1 and

matjo c-^'^^JHbt)dt

e-^'''Ji{bt)Jo{bt)tdt
i

Jo

(3)
^ ^

fi

^~^

(2)

in which

delle Sci. di

(2"^+^'^)^-2(a^+6^)i:
^62v'(a2+ 6-^)
by Gegenbauer,

ibid. pp. 379

"

380.

that

13-23, 13-24]

correspondingformula

and

f^-^"^A (bt)h
in

similar

of

deducible

results of

the

from

{Jahrbuchiiber die Fortschritte

"

R{^-) " R{u)

result

The

not

"

formula

is of

we
satisfied,

are

to

as

obtained

type than

-at

'"
2T(v
is at

exponentialintegralformulae given
these formulae,for,since the conditions "

J^ (0 dt
f-i^

the formula

for

the

limitingcase of
by " 13"2 (3)

values of
integral
pletely
by Soninef ; and the com-

Weber*

have

''-^^ lim f

whence

by

l)'

generalvalues of v
provedby Schafheitlin :{:.

recondite

more

r(iji)

2''-'^+ir(l/-iyLt+
was

^,

also

was

be established

13'2 ; it may

of convergence

extended

was

generalresult

(3)

by
integralsin (2),(3) and (5) are expressible
formulae
in
detail
the
results
some
give
;
this section were
given by Meissel, Kiel Programing 1890.
der Math. 1890, -^-g.521
522.]

^,_^+i

in which

in

fonnulae

the

did

he

j,(t)dt_

(1)

The

:
'

formula

The

V.

^i5'
"i''!Yy
2Trb{a'-b'^)

integral,
afterSchafheitlin.
infinite

Wehers

13"24.

manner.

that

by Gegeabauer
but
eUipticintegrals,
stated

was

{ht)tdt

is bja. The
1:^of the elliptic
integrals

the modulus

and

be modified

(4)may

means

(2)that

from

deduce

R{a)"\R{b)\,

It

is.bls'{a^
+ b^)- Beltrami's

/"-"-'"^('")*=W{5t-6?-

Replacing b by ib,we

where

is

(4)

(5)

and

K
integrals
complete elliptic

of the

modulus

the

where

391

INTEGRALS

INFINITE

.F.fs,!^:^^';.^!;!

l)"'-^V2'

obtained.

once

evaluatingthe integralis to substitute Poisson's integralfor the


used by
Bessel function,and then change the order of the integrations
; this is the method
established
first
is
for a
the
result
but
the
because
is
intricate
Schafheitlin,
analysis
of
formulae
extended
the
values
and
recurrence
of
of
and
then
limited range
use
by
/x
and partial
integrations.
A

direct method

of

Analyticaldifficulties are, to a large extent, avoided


of Schafheitlin.
instead of the definite integrals
integrals
*

Journal

Smith's

fur

Prize

Math.

lxix.

(1868),p.

Math.

Ann.

xvi.

(1880),p. 39.

Math.

Ann.

xxx.

(1887),pp.

" Cf.

230.

question, .Jan. 29, 1867.

Bromwich,

The

specialcase

[Math, and

157"161.

Theory of InfiniteSeries, " 172.

in which

Phys. Papers,

v.

by using contour
If

we

was

(1905),p.

suppose
set

that

by Stokes

347.]

as

"
(fi)

0 and

"^,

R(v)"

then

we

have

(the integrals
convergent)
being absolutely

jy^^'

"- '^'~'

i)r(i)

2t sin /MTT. r

"

'''

Jo ^^^'^^^^^

2-r(.

[CHAP. XIII

FUNCTIONS

(. + i) r (i)J ,.

2-^

(- 0--^

..

BESSEL

or

THEORY

axis taken
The

twice, and

at

we

into the

the contour

deform

" 0, we
(yu)

^^^

(-|yLt)

By the theoryof analyticcontinuation,this result is


R(v + %).
subjectedto the singlerestriction i?(/u,)"
When

^" ''^''''^^^'

/a

half
positive

the Weber-Schafheitlin

obtain

once

valid when

and

t- are

of the real

formula.

integral*
(- ty-i^^'

be

may

treated

analysisis

that

exactlythe

same

cos

(tcos 6) has

to

sin

"

be

2i sin
_
"

providedthat R{fi)" R{v


{fx)"

the

"

{tcos 6),and

factor

so,

by

has to be
|/i7r

cos

fx-rr.T

'

%) and R (/j,)
^ 0.

contour

may

the real axis taken

deformed

be

into the

half
positive

of

twice, so that

r(|/.)tan(l/x7r)

tlAt)dt

r
.,,.

providedthat "\"R{ix)^0
take /a

we

0, v=l,

and
see

we

R{ii)" R (v)+ 1.
that

Hi (t)dt_^

(3)

xTT.
f

This

combined
result,

with

ir
"^

used

was

sin

the

tan Ifx-rr
(|/z)
2''-'^-^ir(i/-iyLi+l)

(- 0 dt

r t"+)H.

If

"

in

that

j+x {-ty-"'^^
R

replacedby
the
integrals),

only difference

I/att.

It is thus found

When

the

manner

(adapted for contour

Euler's formula

replacedby

in

the

asymptotic formula

H.i{2t)dt

(\

cos(2a;-|-i"r)

by Struve,Ann.

der

Physik und

(1882),p. 1014, to

Chemie, (3)xvii.

tabulate

i_ r'A^mdt
for both

small

and

largevalues

of

x.

The

last

integralis

of

importance in

the

Theory

of

Diffraction.
*

the

Generalisations

integrals of

Videnskahernes

this

obtained
section

Selskabs

by replacingBessel
and

in

many

other

Skrifter,
(7)v. (1910), pp.

functions

by Lommel's

integrals are
1

"

37.

discussed

functions

by

("10-7) in

Nielsen, K.

Danske

13-3]

(1) under
[Note. Bj difi'erentiating

(4)

formula

for functions

Weber's

result

J^{t)\ogtdt=-y-\og%;

of the second

und

Phys.

i.

kind, correspondingto (1),is

r (hn
jhfi)

P l\{t)dt^ V

(5)

(hfi-v)iT

been

has
i'

magnetic
given by Heaviside, Electro-

0.]

and
integral
firstexponential

Weber's

13'3,

p)cos

provided that I R {v)\" R {}i-v) "'i. This result


Theory,in. (London, 1912), p. 273, when

its generalisations.

formula
integral

The

Jo {at)
l^i"

(1)

exp

deduced

was

obtain

105"112.

(1890),pp.
The

integralsign we

the

investigatedbj Lerch, Monatshefte


fiirMath,

also been

has

this formula

393

INTEGRALS

INFINITE

Weber*

by

tdt
i-if-t')
his

from

It is

exp

double

" 14-2. This


chapterby containingthe square of the
It is supposedthat |argp |" jtt to secure
complex number.
discussed

^^ (-^3)

in

equallyeasy

those

function.
exponential

but

convergence,

will be

earlier in the

considered

variable in the

Hankel'sfmore

to prove

which

integralformula

integraldiffers from

is an

unrestricted

generalformula,

r J,(at)ex^{-p-t^).t'^-'dt

(2)

To

2p^r(.+

by a direct method.
supposedthat:|:

^^-^
^-^^(i^
l'-4p^j

"

secure

l)

the

at

convergence

it
origin,

must

now

be

R{lJi+ v)"0.
To

obtain

[ /],,d

I0

"

Jo
is convergent,it is

Jt,{atyi^powers
*

an

Journal

fiirMath.

lxis.

Iexp {-pH-)1 1i'^-iIdt


.

evaluate

the

givenintegral
by expanding
integrating
term-by-term,

to
permissible^

of t and

that, since (by " 7 '23)

observe

the result,we

(1868),p.

227.

Weber

also evaluated

(2)in

the case;U

2, v being

integer.
f Math.

Ann.

viii.

(1875),p.

469.

See 'also

Gegenbauer,

Wiener

Sitzungsherichte,lxxii.

(2),

(1876),p. 346.
% This

restriction

may

be

disregardedif

we

replacethe

definite

integral /
.'

/"|('+)

integral
" Cf. Bromwich,

Theory of InfiniteSeries, " 176.

by the

contour

394

THEORY

It is thus found

BESSEL

OF

that

J, (at)exp i-pH') f^-'dt

this is
If

rightin (2),we

{-pH') dt

("4"42) to

the function

the

on

find that

the

so

In

we
particular,

exp

providedthat R{v)"
by Sonine, Math. Ann.
discussed in " 13"47.
In

order

that

finite terms

whenever

yu,

i/

"

is

an

even

have

JV. {at)

(4)

i-pH^)
This

1.

"

the

t^^^dt

^)

("^^^^P ("

integralis the

(1880),pp. 35

xvi.

38;

"

to Kummer's
susceptible
replacevhy 2v, we

second
then

"

investigations
are
applications

basis of several

hypergeometricfunction

if we

transformation

of these

some

rightin (2) may

the

on

("4"42),we

take fi

be

\] and,

"

find that

|V,.(aOexp(-^^i^).rf"=
^exp
{-^).L
{^),

(5)
result
If

f+i^+^-i exp

first transformation

in
integralis expressible
positive
integer.

"

r J^{at)ex^{-p-f).t^-^dt

(3)

and

'

the result stated.

equivalentto

apply Kummer's

we

^'^

)^\

and

[CHAP. XIII

FUNCTIONS

we

givenby

Weber

replacei/ by

in the
"

j;, it

case

^.

is easy to

that

see

\^Y.,,{at)ex^{-pH'')dt

(6)

.' 0

when

\R(v)\"^; and, if we
r*

make

-*

tan
-r-r

(8)

VTT

=
,

Jo
when

we
positive),

Y,Aat)dt

(7)

0, (a being now

f^

\R{v)\" ^,hy using" 7*23

and, in particular,

rY,{t)dt0.
=

find that

396

THEORY

13*32.

GeneraUsations

When

the Bessel

of the

not

The

BESSEL

of Weber's

only method

[CHAP. XHI

FUNCTIONS

second

exponentialintegral.

in

functions

order,it is

same

simpleform.

OF

of the type just considered


are
integrals
usuallyimpossibleto express the result in any
of dealingwith the most
generalintegral

"^

I ./"(at)J, (ht)exp
is to substitute

the series of

integrateterm-by-term,but
In the special
in which X
case
is equalto
(hay-"

by

An

for the

" 11*6
it
v

.^,/,

case
exceptional

when

occurs

{lx+ v-\-l
^

fi-\-v +
'

\ +
'

It will

This
XXX.

to

result

-n

then have

we

"

a-\

'

1. ^

1./^ +

"

!-"

-p2J'

of this formula

specialcases

R{^

-^

have

v) "0, then

T(,M

2'^+" r

(fL+

obtained

was

(1882),p. 91, in the

by Struve, Jle'm. de
l ;
specialcase /x
=

z/

v)r{^)

^)r (fi+ ^)r (v + 1)


(7)
Imp. des Sci. de St Pe'tersbourg,
then
the
is
on
equal
right
expression

V Acad.
the

4/(37r).
In

that, when

^M+v

i')"0.

/x +

iJb+v
2

/'Mt)JAt)^^^

(1)

i^ R(\+

integral
involving
productsof Bessel functions.

be shewn

now

,,

13"7.

"" 12'11,

b-+aHin-0^

-^-r-^-^^

by using the expansionof " 5"41. Some


been investigated
by Gegenbauer+.
13 '33. Struve's

and

"

J^ (at)J, (at)exp (- pH-) t^-'dt

"^^

fab sin 6\

7-

the results of

on

functions

to give the result here.


unnecessary
has shewn that the integral
fx, Macdonald*

"

,",/,"

based

Bessel

product of

seems

r^'^

transformation

t^-' (It
(- jft-)

it is first convenient
evaluatingthe integral

both exceed h. It then follows from


{i')

to

suppose
3-3
that
(7)
"

that R

and
(/u.)

(2^.-1)
2'^-''-Vr(/x-hi)r(i;-F|)

J^(t)JAt)^.^^(2/x-l)
^+,
c=^

rh-

ch-Q[jx(ts,i'a.d)s,\n(tsv[\d")
^^
^^
^

I
.

.,.,/,

^^

cos-''"-

"

J0

,"

cos-""-

"

"

"^sin

t'

Proc.

t Wiener

London

Math.

Soc.

xxxv.

Sitzungsberichte,lxxxviii.

(1903),p. 440.
(1884),pp. 999"1000.

-a
^

"

sin

"

7/1
dd

jj
;*
dt.
d(p

13-32, 13-33]

{tsin 6) sin {tsin ^) does

In view of the fact that t~" sin


the

smaller

\jt-and

of

exceed

not

cally
numeri-

sin 6 sin 0, the

repeatedintegralconverges
integrations
may be changed.

the order of the

and
absolutely,

397

INTEGRALS

INFINITE

Since*

6) sin {tsin (/")

sin {tsin

we

i-TTr

^TTI
J

6 cos-''-=="/)
sin^ 6 sin

cos-'*--

-I-

^ cos-"--

COS-''--

(2v-l)

sin 6 s\n^(f)d(f)d6.
(/)

have

['"cos^''-^
sin (^j
(j)
j cos-'*-^ ^

cos-''-^

cos-'^--

sin^

(9sin^ Odd

(9g?^1
dcf)
cos^-'-i

(^

cos^'*-^

Jo

"'0

(f)ddd(f"

we
integration,
partial

{6 ^ (^)

"|i7rsin(/), (^^(^)

^^

is equal to
integral
triple

find that the

But, by

i^irsin 6,

Jo

(^sin- "^(/(/"

_V{^i+ v-l)T{^)
"2r(^ +
other

The

""

*)

i/

integralis evaluated

in the

same

and

manner,

v-l)V{%){{2^,-l)
/"-JAt)"L{t)^^_V{^l

the result stated is evident.

whence

merely R{fi + v)"0

of //.and V for which


continuation.

It may

be sliewn

in

similar

extension

The

that, when

manner

j
result

This

By using "
'"

also obtained

was

10'45

we

find

by

then
R{fi+ v) is positive,

also

2'*+T(iu+ i/+i)r(/i+i)r(i/+ i)'

t^"^"

i"

{^v-l)]

r(;x+.)r(A)

r-H^(OH.(0^^_

have

we

the range of values


the theoryof analytic

over

is obtained

so

by

Struve

that, when

{ihid.
p. 104) in
and
(/x)

the

case

\j.

v=

1.

iy)exceed |,

H^ (0 H. (0 dt
(2^-l)(2v-l)

r*

p'- [h^{l-

cos

{tsin 6)\ {1

"'"'
"'"
~2'"+''-"^7rr(/L(,+|)r(i/
+ ^-)

Now,

if a and

/3are

it appears
positive,

from

cos

cos''^'*-''
6 cos-"--

consideration

0 sin 6 sin (j)dddcj}dt.

of

(1-e"'^)(l-e3'^)

!"
This

result is

{tsin 0)}

^^

-' "^

easilyproved by

dz

contour

integration.

398

OF

THEORY

round

BESSEL

of the real axis and


contoyir consisting
""

/" (1

(1
(at)}

cos

(301

COS

largesemicircle

f""sin

,
_

Jo

The

R{fi+ v)"
The

be evaluated

way

of values

range

R (v) both

of

fi

in

but

similar manner,

the results

integral
of Weber

of

are

and

for which

R{v)"l

and

great interest*.

no

Schafheitlin.

supposed to be positiveto
limit, was
investigated
by Weber, Journal
80, in several special
cases, namely,
and

(i) x

Soninet, Math.

Ann.
did
Some

o,

(ii) \

\,

xvi.

he

(1880),pp.

lay any

51

stress

the

on

later the

/ti=

convergence
Math. lxxv.

not

at

the

(1873),

0, v="h.

for which

discontinuities

examine
which

it is convergent, by
the

integralin
when

occur

very

and

heitlin
Schaf-

was

his

fur

-\,

52; but he did

"

integral
years
preliminaryanalysisrests to a
differential equations.
equal.

I, but

secure

are

evaluated, for all values of X, ^ and

integralwas

great detail,nor

linear

I , then

integral

in which

become

exceed

cit

^^,

r JAai)JAJ"t)
^^^

The

and

/^
/J"^^^^^^^

Jo

"

in

(1).

as

_j^

the

over

13 "4. The discontinuous

upper
pp. 75

tripleintegralevaluated

1.

integrals

The

to the

same

extended

be

this may

equal

and
that, if R (ju)

manner

it,that

t^

in the

rH.wj,w^,^
J

may

in like

will prove

reader

(3)

and

is

the

above

(at)smj^t)

Jo

^^

tripleintegralunder consideration
proving (1),and consequently(2)is established
Hence

[CHAP. XIII

FUNCTIONS

investigated
very thoroughlyby
undue
the theory of
extent
on

somewhat

by GublerJ^who used a very


of contour
however, it seems
impossible
integrals
eleganttransformation
; unfortunately,
in which X =^ 0. The analysisin the
to adapt Gubler's analysisto the more
general case
specialcase will be given subsequently(" 13'44).
The

1890

specialcase

related

Some

[Jahrbuch

in which

integrals have

ilber die Fortschritte

t See also " 13-43 in connexion

discussed

was

been

evaluated

der Math.
the

with

in 1895

by Siemou,

1890, p. 341]

researches

of

Progrnmm,

Ltiisenschule.

Berlin,

Gegenbauer,

Wiener

Sitzungsberichte,

(2),(1884),pp. 990"991.

Lxxxviii.

X Math.
Math.

Ann.

" Math.

An)t.
xxx.

Ann.

der BesseVschen

(1887),pp. 161
(1887),pp. 582"583,
xxx.

xlviii.

(1897),pp.

37

"

"

178.
and
48.

The

question of priorityis

by Schafheitlin,Math.
See

also Graf

Funktioyien, ii. (Bern, 1900),pp. 136"148.

and

Ann.

xxxi.

discussed

by Sonine,

(1888),p.

Gubler, Einleitung

156.

in die Theorie

13-4]
The

INFINITE

which
first investigation

conditions

The

INTEGRALS

shall

we

399

giveis based

the results of

on

are*

for convergence

[R{,jL+v+l)"R(\)"0,
it

beingsupposed,as alreadystated,that

(a

and

shall first suppose


that the former conditions are
The
shall also take b " a.
analysisis greatlyshortened
constants

I2a

by

the

v-X

1,

=fjL

I'

supposedthat

It will be

b)

positive.

are

We

a, /3,7 defined

" 13*2.

and
satisfied,

we

by choosing new

equations
l,

i\

/3,

y-a-

'
j^

1.

"

these relations hold down

the end

to

of

" 13'41.

that

It is known

c-^+oJ

Jo

integralon the left is convergent ; now, when c has any assigned


the rightis convergent for complex values of 6;
on
value,the integral
positive
and
the
is an analytic
function of 2^ when
we
resulting
integral
replaceb hy z
and
\I{z)\" c.
R(z)"0
since the

Now]^^6-*

dt

^piz-,

mir(y

[n"=o

Jo

m!r(7

",=o

m)

m) Jo

providedthatf
m!r(7

^=0

is

absolutely
convergent;
\z\"c.
Hence, when

12^ |"

m)

and

Jo

it is easy

shew

to

that

this is the

when

case

c,

(ia)"-^r(2a+2m)

J,/

"

,"rom\r(ry+m)

ty-'^-^

r (a
+ c-)'^+'"
(ft-

1)

a"
X

2-Fi(a

m.

"

/3
"

m;

"

/S+1;
ft-

It follows

from

the

asymptoticexpansions of
R{lii+

are

sufficient to

secure

t Cf. Bromwich,

convergence

Theory

when

b,

the Bessel

l)"R

that

the conditions

(\)--0

provided that

Series, " 176.


of Infinite

functions

^-^z

is

an

odd

integer.

c-

400

THEORY

and the

OF

function
hypergeometric

r(a-/3

i)r(-i)

BESSEL

on

the

R^

the moduli

Now

do

(1

"

in absolute

exceed

not

where
\/x)~'^-'^"\

moduli

of the terms

in the

of the terms

value

rightmay

/^

I,1

the terms

Hence

^M

"

in the

/3
"

f ; a?)

m;

in the

of
expansion

V*-)"^"^"'"VVa;-

in the infinite series which

exceed in absolute value the terms


*

expansionof
the
]and 12/3-11;and, similarly,

do not exceed in absolute value the alternate terms

(1

/Q

_Li

_L

replaced
by*

the alternate terms

is the greaterof |2a


of
in the expansion
7?i

be

of
expansion

oi^i
(a+

[CHAP.XIII

FUNCTIONS

has been obtained do not

of the series

{laf-^r(2a + 2m)
i^)Y+2"i-i
(-)""(

Lir(l-/5-m)r(a

l)|

where

+ c-). But
c'^l{a^
c, and

I2: 1" V(a^+ C-)


"

this last series is

absolutely
convergent when
it represents
function of z in this domain.
an
analytic

theoryof analytic
Hence, by the general
continuation,

that
provided

(-)""
(W+^'"-'' (lay-^r (2" + 2m)

satisfiesthe three conditions

R{z)"0,
Now

take C to be

\I{z)\"c, \z\" ^{ar+ c=)


-

number
positive
6

and take 0

"

small that

so

V(a-+ C') G,
-

C, so that also
h

"

c.

Cf.

Forsyth,Treatise

"

+ c")
s/(a-

on

Differential
Equations, (1914),
" 127.

c.

yi

13-4]

formula we may take z


h, and when this has been
integral
nse
fonctionsmajorantesjustas before,we find that the resulting
of an absolutely
its terms
less than the terms
convergent series

in the last

Then

done, if we
series has

r(|)(i-.v/X)-^-^-

(_)m (i5)Y+2m-]
Qa)--^ r (2a H- 2m)

401

INTEGRALS

INFINITE

r=o wITT (7 + m)

(a' + C'Y^''-'^

(1

/S

r (a +
III)

\)I

V(h-^-m)T{a
where

.Y=(7V(rt=+

with respect to
c

0 is the

-*

of

when

as

same

series converges uniformly


Weierstrass,the original
therefore the limit of the series when
": c ^ C, and

the value of the series when

therefore

have

We

m)

02).
test

Hence, by the

0.

proved that

00

"Zo m\r{y
and

V {a
a^''+^'^

m)

j3+ 1)

"2

'

therefore

J,.p(at)Jy_i{bt)^^^
(_)m 5v+2m-i

(2a + 2m)

(i)

V
_

r (7 + III)2--^+v+2m-i
",ro'//i!
It has therefore been

^^^

shewn

"^^

Jo

^r-a-^

(^a+^+2m i^ (1

,M) r (a +

/5

A)

that

2v-"-^a"+^r(7)r(l-/3)-- ^V '^'^'aV'

that is to say

bTilfi+h^-h^

f^JAat)JAbt)
^'^-

Jo

providedthat 0 " 6 " a,


obtained by Sonine and

and

a
interchange

and

If

we

that,when

"a

"b

and

that the

iv-"-^

b,and also

the

/u, and

F.

i)

This is the result

v,

find

throughoutthe work, we

is convergent, then
integral

a"^-Pria)

"^~2v-"-^6-^'^-"+U'(7-a)r(a-/3+l)
X

\y. B.

*/.-iz.

Schafheitlin.

"

J.

l)

is convergent.
integral

[-J.-,{at)Jy-r(bt)

...

^'^^

l^(z.+ ])r(i\

2^a-^+^

^^

oi^jfa,

7 +

1 ;

/^+

1 ;
26

^,j
-

402

THEORY

OF

BESSEL

[CHAP.

FUNCTIONS

XHI

j"

Now

it

happens that

so

the

the rightin (1) and (3)are not


on
expressions
of the same
function. There is consequently a

the

analyticcontinuations
in the formula when
discontinuity
this phenomenon in some
detail.
TJte critical case

13*41.
In the

of the

case

h; and it will be necessary

to

examine

of the Weher-Schafheitlin
integral.
under

now
integral

when
consideration,

b, we

have,

before,

as

.'o

c^

assuming that R{im-\-v^ I) " R{\)"0,


Now

to

secure

convergence.

consider

Ja-(^
(at)Jy-^(at)

._,,

is

When

complex variable

2a

(z)"

we

with

this
integrateterm-by-term,

the

series is convergent.
resulting
thus get, by

(z)positive.

expand the integrandin ascendingpowers


procedurebeing justified
by the fact

may

and

We

'^^'

ty-^^

where

t^

Q J a

of

that

using" 5'41,

/,^1J Jg-p (at)/.y-i(at)


"^

e-^t(_)m

l*3C

CO

"=0

m\

^2a-Hom-lT (g
/g+ ry + 2m)
(^^),-p+y+2;"-l
^^
T
T
m
-H
+ 1)
(a
yS + 7 -h m)
(a
(7 + m)

V (2a 4- 2in)T (a
yg + 7 -h 2m)
(-)"^(^ay-^^y^-^-'
,"t0 z''^^^^m ! r (a y8-I-m -h 1 ) r(y -h 7/i)T(a /3-F 7 + m)
X

Now
and

the

its value, when

so

the left is

on
integral

has

of the series

the

on

analyticfunction of z when R (z)" 0,


the small positive
value c, is the analytic
continuation
right.
an

But, by Barnes' theory*,the series


may

be

represented
by

the

the

on

rightand

^,a+,s r

(a

/y-F 6- + 1 ) r

its analytic
continuations

integral

r (2a -f 2s) r (a
(^a)"-^+y^-i

"

2lTi]

"

(7 -h s) r

/3-f-7

(a

/3-h

2s)

V(-s)ds;
7 +

s)

and this
It is
ensure

when
integral
representsa function of z which is anal3'tic
arg^^ |" tt.
that the contour
consists of the imaginaryaxis with loopsto
si:pposed
that the polesof V (" s) lie on the rightof the contour, while the poles

(2a + 2s) and of T (a

of r

When
as

to

\z\"2a

enclose the
*

Proc.

London

we

yS+

evaluate

may

poleson
Math.

7 +

2s) lie on
the

the left of the


Soc.

the left of the contour.

integral
by modifyingthe contour so
the residues
and evaluating
contour

(2)v. (1907),pp.

59"118.

See

also

"" 6'5, 6-51 svpra.

404

OF

THEORY

and

BESSEL

[CHAP.

FUNCTIONS

XIII

hence

t^

unless

(4)

This

0.

obtained

2r(p

should

be

" 13'4,namely

from

compared

that, when

2"^a''-P-^r{a-p)
T{p
b

"

l\+l)Y\

X, ^

0, the functions

a,

2^1 U-l\,-p-l\\ a-p:

--,

F,(a-l\p+l-^X;
a+p^l;
rightin (4)and (5)do

the

on

not

tend

-^
to

limits

the contour

A, is zero,

becomes
integral

[r (2a + 2s)}-T (- s)
(^ay''+''-'
2^iJ-ooi^'"+-* V{a+p + s+l)r(a-p
+ s)r {2a + s)
/"=-*"

and

formulae

b.

the other hand, when

On

general

+ l)T{lX-p)
"2^b'^+p-^+^r(a+p

when

more

JO

Since

"

the

^\)r(^x-2J)'

a,

Jg+p{at)Jg-p-i{bt)

f
,

with

dt

j^

and, when

^\+i)r{a

the residue

at

s=

is (" )^/(2a).

"

*'

It follows that

"-^^' a^)'
a^-PT(.-p).pl^^^['''-i'

'

j^Ja+p(at)J"_^_i(bt)dt

(6)

accordingas

"

a,

a,

"

a.

,F,(a,-2r,a-p;
it is evident
tuhen b"
The

that tJievalue
"

0 and

result of

(7)

taking X

0.

l in

(2) is

^_ypu2a),

Since

l)

(-)PplT(a-p)/V(a),

dt

sin

2
^ {v- ^)
J^{at)J,{ca)'~j

is the

mean

of its

limits

IT

Jo
which

lulienb
of the integral

a+

'

IT

is also

easilyobtained by insertinglimits in v$5"11 (13); this formula has been


discussed
in great detail by Ka^jteyn, Proc. Section
Wet. te
of Sci., K. Akad. van
103; Archives Neerlandaises,
(2) vi. (1901),pp. 103
Amsterdam, iv. (1902), pp. 102
"

116.

"

13-42]

INFINITE

of the

Specialcases

13*42.

405

INTEGRALS

discontinuous

of interest

integral.
obtained

values
b}^givingspecial
when
To save
to the constants
\, ju.,v in the precedinganalysis.
repetition,
the firstis its value for h " a, the second
three values are given for an integral,
for
third
b"
when
the
values onlyare given,the first is
for h
two
a;
a, and
cases
special

Numerous

are

"

following
"--

J^(at)J^{bt)

(2)

"

^\
r
/
Ju.(at)
sin
^^

for all

correct

are

the

[R if.)" 0]

r/x~'sin{yu,arcsin(6/a)!,

J,
bt

,,

"I

^^^

""

[fji
{b

'

JAat)cosbt,

--^-^7^

values

(3)

the

[i{h/aYlfM,

^^

(1)

and

integrals
convergent.
c
importantspecialases* :

the most

are

for h^a;

make

which

values of the constants


The

second

the

for h ^ a,

the value

at

/J'

i/^"^
V(6-^ a^),p

si^

"

|"-^cos{;.arcsin(6/a)},
h/JLTT
[R (/x) 0]
[f^{b+ ^(b-' a')]''
a"

cos

"

'

'

sin

arc
Ifi

sin

(b/a)]

v'(a-^ b')
-

I J^ (at)sin bt

(4)

dt

or

[R (f.)"

2]

}2f^7r

aJ"-cos

+ ^(b-'-a')\'^'
y(Jf-a').{b
arc
{/J,

'cos

J^{at)cos

(5)

bt dt

-Ice

sin

(b/a)]

o/^

sin h/xTT

[R(p-)"-l]

0,

b-7

"

,.

^/(b'-a"').[b+^/(b'-a')\'^
of

Specialcases

precedingresults

are

0;

(6)

Jo (at)sin bt.dt=

-j

(7)

Jo (aO cos bt.dt=

-Uo

0.
These
p. 77,

are

two

known

formulae, which
as

WeheYs

determiningthe potentialof
*

The

Numerous

in
integrals

t The
Smith's

former

given l"y Weber t, Journal filrMath.


factors; they are associated with
electritied circular disc;.
were

discontinuous
an

specialeases are given by Nielsen, Ann. di


(4) and (5)diverge for certain values of ix when

other

was

known

prize examination

to

Stokes

many

question in Feb.

X Cf. Gallop,QuarterlyJournal,

xxi.

years
1853.

earlier,and

Mat.
a

(3)xiv.

230"231.

(1873),

problem

(1908),pp.

82

of

90.

"

h.

was,

in

fact, set by him

[Math, and Phijs.Papers,

(188(J),
pp.

lxxv.

the

v.

as

(1905),p. 319.]

406

THEORY

OF

f J^ (at)J,.,(bt)dt

(8)

[CHAP. XHI

FUNCTIONS

is

formula
special

Another

BESSEL

\1/(26),

"
[R ifjC)
0]

(O;
if we

and

put fi=l,

obtain

we

Weber's

Jo

result of

The

other

U-

putting ;li

specialformulae

have

Wins.

been

lxiii.
Young, LeipzigerBerichte,

method

Another

of

ii.

useful
369

in the

theory of

Fourier

series

by

387.

"

and

given by Hopf

be noted.

(1) must

When

f^

""

Sommerfeld, Archiv

der

have, by " 5*51 (5),

0, we

"

dt

e"^-. + "(.i-)=2^/ JvHt)-

2
71=0

.' 0

and

the

factor; see

(1),(1916),p. 109.

found

been

discontimious

(1911),pp. 1"16.

of formula

consequence

BirichleVs

as

(1911),pp.

evaluating(5) has

v/iid Pkys. (3)xviii.

Math,

1/7

,1/6.

(8) is known

J^in

der Math.
Encyclopadie

by Voss,

Some
AV. H.

{ibid.,
p. 80),

rMat)J,{bt)dt
\ll(2b)

(9)

article

result

"

1,

so

1^.(^)1=$!, IJ. i(aO1^1/^2,

(10)

of Hansen's
positive;this is an interesting
generalisation
discovered
Miinchener
Abh.
xv.
by Lommel,
(1886),pp. 548

providedonly that
(" 2 "5)which
The

was

he

reader

may

find it

to
interesting

the Weber-Schafheitlin

Bateman's

*,
integral

theorem.

Gegenbauer's
investigation
of the Weber-Schafheitlin
integral.

13*43.

specialcase in which
Gegenbauerf found that by his
in a simplemanner.

the Bessel

In the

If

deduce

^^'
f^J,{at)[l-J,{bt)]'^=
[log(6/a),

(11)
from

inequality
549.

"

R{'iv+1)"R

method

(X) "R{v

h)

we

functions

Weber's

are

of the

integralcould

order,

same

be evaluated

have

dt

J, {at)J. {bt)
=^

/--J, [tV(a^ + 6" 2ab cos 0)} sin-"


cpdxpdt
+ b'-2abcos"p)i
^)T{h).K Jo t^-''{a'

{-lab)"
r{v

(aby
2^ r
*

Messenger,

Messenger,
t Wiener

xlii.

xli.

(J/

{v

1)

(1912),p.

101

I
'

IX + I)

(1)r
;

for

"

'

(iX +
a

i) j 0 (a'+

proof of the formula

(1913), pp. 92"93.

Sitzimgsberichte,lxxxviii.

(2),(1884),p. 991.

sin-" (f"d(f)
^-^
-

by

2ab

cos

another

(/))''-*^+^
method,

see

Hardy,

13-43]

INFINITE

by " 13'22.

Now

When

the

from

"

rightis

the

on
expression

the

a,

407

INTEGRALS

formulae

recurrence

{n-v2,M- 1)(1 -^"^)-i"-iC^_,(^),


^ 1(1-z^)-h^C,,^(^z)]
1) (1 .-)^-+'^-i
r",+,(^),
G,,^{z)]
^ {(1 z'-f-+"^
=

we

{n +

that

see

l)|'

(n +

(\

z^y-i C^n^,{z)dz

-C

(1

~\{l s")i"+^CV
z"-)"-^^i''+^
-

{n + 2,x-2v

f'

1)

(2)\
dz

dz
(1 -^-""^)''-*a"'^(5)

-1

-^=""""' i{(i
="2':;!t-/!/'
-.')-"'c,

that

so

sin-"(f)d({)
C^l+l(cos(/))

2fi-2v-l)(n

^(n +

{2v +
Hence

7i

sin-"(f)d(f).
G'^n-i
(cos(f))

it follows that

and

2fi-l)
l)

l){n

this agrees
The

with

the result of

,fJv
-

l\ + H

method

extended

more

range

By expandingthe

(1)

{2v + 1) " -R (X,)"

finite integral
in powers

./.(at)J. (bt)
^

-,

and

with
The

by

is valid whether
this form

"b

of

"

Gegenbauer's; but
avoid the necessity
of

as

to

establish the result

over

1.
we
cf),

cos

obtain

the formula

^~

2z/ + 3

(2v+\-X

of the result the

reader will find it

putting6

or

"

^^,^ ^^^_^,,,^^^^
-

4a-6-

^+l;(a^T6^0'
-,

^"^^^1"4
which

^^Y

" 13-4.

the same
given here is substantially
he used ilightly
in order
more
complicatedanalysis
appealingto the theoryof analyticcontinuation to
the

IX; v+l;

'

b. This

'

result

was

givenby Gegenbauer,

is masked.
discontinuity

to
interesting

in the finite integral.

examine

the

critical

case

obtained

408

OF

THEORY

Ouhlers

13"44.
The

BESSEL

[CHAP. XIII

FUNCTIONS

of the Weher-Schafheitlin
investigation
integral.

integral
.-00

dt
J^ (at)/, (ht)
Jo
will

now

first to

investigated
by the method due
consider the more
generalintegral
be

(-

t^

be evaluated
cannot
though this integral
It is first supposedthat R{v)"0,
methods.
b are positive,
and a "b.
as usual, a and
even

that the

in
R

simplemanner
by Gubler's
I; and,
(k) "^, R(fM X)"
a

"

"

of Bessel's integral,
generalisation
givenby " 6*2 (2),it is evident
is equal to
integral

the

Mv

27riJo
We

It is convenient

JAo^t)J,{ht)^^

-'0

From

Gubler*.

to

take the contour

/I

2-"-^

^^

shewn

as

in

expHhtlz-

Fig.29

to

the

meet

-]}-dzdt.
circle

j^ |

1 and

the

FiK. 29.

line

R(z)

only at

" i; and

then, for all the values of

and

t under

consideration,

R{hbt(z-l/z)]^0;
and

the

since
absolutely,

repeatedintegral
converges

0^
,-v-l

idt

dz\

Jo

is convergent. The
we

order of the

integrations
may

therefore

be

changed,and

have

Jj,(at)J,{bt)
Lrr'r^M,,^i.
dt
0

If

we

t^

27ri j

CO

j0

t''

H'-i)

write

b(z-l/z)
Math.

Ann.

xlviii.

-a(^-l/0,
(1897),pp.

37"48.

y"V"l

dtdz.

13-44]
and suppose

b\^ Kummer's

oi^iU

of z, qua

sin

"

X: /i +

1;

c^^,
j-^-^oj

Fig.29; and the


of

function

the unit circle

curve

on

different

the different jmrts of it. The


is the acute
w
anglefor which

on

e-'", where

only at

path is the

is irreducible because

curve

taken

^, are

Since

describes its contour.

is chosen, the

the greater modulus

the rightof the circle in

meets

1, /i +

X +

path described by ^,when

the

the value of ^ with

curve

have, by " 13'2,

transformation*.

consider

branches

j^j ^ 1, we

"~.

Next

for which

that that vahie of ^ is taken

dt

409

INTEGRALS

INFINITE

co.

Now

both

functions of X

and the final contour


integral
original
when
1, so long as a^^b. Hence
R{X)"
"

providedthat R{/x) "

Next

are
integral
analytic
\ =0,
we
takef
may

the

write ^=

zr

1; and then

"

have

we

then

and

br

T(bT + a)

"2

=
~

(ar + 6)

'

tiT

'

and

the

in

is that shewn

contour

originclockwise,and

to

returns

have

half of the real axis,we

Fig.30;

from

the contour

where

"b/a;

arg

it starts

b/a,encircles the
the positive
crosses
-

-0.

dz

adr

c,.

bince

2T(6T

",0(l+ n
find

we

(on reversingthe direction of the contour) that

I"'/^
(at)J, (bt)dt
J

^r~.

f'"^'
ri'-'-'^-i'
(br+

ZttI J

b"

-1,1

Journal

fur Mat

Ii. xv.

/'"o+'

(183G),p. 78, formula

(1908),pp. 115"119.
t If X:^0, the hypergeometric function
and

the

analysis becomes

dr
(ar + 6)*c+'--i"
a)-i"'+'^+i"

27r?'o''+\'
_i
*

a)'

intractable.

does

b-

(57). See
not

in

also

x-iiv+ft+ii

Barnes, QuarterlyJournal,

general reduce

to

an

xxsix.

elementary- function,

410

THEORY

If

OF

[chap, xni

FUNCTIONS

BESSEL

the values of the

and substitute
expand in ascendingpowers of h-jaEuler-Pochhammer
then Gubler's result
integrals,
we

//x+i^

i^

ii +

V-

A-^^

is manifest.

Fig. 30.

The

modification
of the

13"45.

Weber-

K,(at)J^(ht)^^^

integral

.'0

which

converges

terms

of

unlike
To

ii R{a)"\I

(b)\ and R{v

that

it has
integral

evaluate

when
discontinuity

no

it,expand J^(bt)in powers

order

that

the

result

convergent series. By using"


^

in
(fi)\ is expressible

"\)"\R

hypergeometricfunctions,like the Weber-Schafheitlin

16 j" ja j in

Schafheitlin
integral.

Jo

13"21

of

6" r (A y

b.

of 6, assuming

that
temporarily
term-by-termintegrationmay be a

(8) it is found that

n=on\l {v +

t"

but
integral,

l)Jo

^x 4- ^At +

^) r (^1/
-

'-"-X,4-;u,-|-l

A-

"

"

'

1\

i/i -I-i)

b-

/i+1
v

\\

a-

and, in particular,

f"("o^(.o..^-^-.e=^^""'"^

(2)

providedthat
Formula

J? (i/+

(1) was
*

1) "

i2
,

(/a)
jand R (a) "

givenby

Heaviside*

when

I(b)\.
fj,=

0 and

A,

Electromagnetic Theory, iii. (London, 1912), pp. 249, 268, 275.

is 0 and

"

1.

412

observed

It has been

a,

b, c

the left in (1) is


on
integral
Legendre functions. The expression
may be
the
manner:
following
rightin

of
the

the sides of

are

sin

[CHAP. XIII

FUNCTIONS

that the

Macdonald*

by

in terms
always expressible
derived from the integral
on
When

BESSEL

OF

THEORY

by
triangle,

sin

I "^
=

the substitution

sin 6

^A

have

we
rA

h-

(a-

2 he

c-

sin-" "/"
d(j"
(f"y-''-^

cos

0
A

(2bcy-'-^ I {cos(f"-COS Ay-"-^

sin-" (j)d(fi

Jo
rhTT

sin-''-'hA
2-''-'^''-^bcy-''-'

^^6'
(l-sin-|A sin-^)''-Hin-"6'cos-'^--'-i

JO

i'^-'

^A^^'''^f^^^~''\,F,(^
^;sinnA),
vA-i']f^

{bcy-^-'sin''^-'

and

and
therefore,if R (yu,)

we
triangle,

^, and

"

a,

b, c

the

are

sides of

have

j^J, (at)./.(60 -L (ct)

t^-^ dt

(4)

If,however,

a^

"

(6 + cf, and

P:,-"^^^ )"

-^^:^^,,,,

write

we

"b'-

a-

we

(v) exceed

(i^+ i)

c-

"

26c cosh S^-,

have

I (a--6--c=+

"/))'^-"-isin-"^(Z(/)

26ccos

Jo

S4 +
(26c)'^-''-i
I"(cosh

^V-'-'^;f^-fv-

(260cosh

so

that,when
/-\

(o)

where

^\

j^K
2bc X

/j^N

from

-\-b-+

a-

c-;

and

in

result is also due


Proc.

London

Math.

^.^

Q^_^(A ),
--^-j^^-i^-j-^-

this formula

a,

b,

may

the four numbers

R{a

positive
; this

^)

1 ; sech-^

^ 13*45 (2)that

J. (ct)t^" dt
(at)J. (bt)
=

sinh'*-*c^^.i-j^
(6ey~^cos^'7^.
Q ,(cosh^).

7^

^x^i

deduce

we

providedonly that

are

"^

J^{at)J,ibt)JAct)f-''dt=^-^

In like manner,

(6)

have

we
(6 + c)-,

"

a-

(I'M-1)

{^^^ 1, '" f^;

.,F,

sin-" "/)c7"^
"/))'^-'-i

cos

" lb "
to

Soc.

ic)

Macdonald.
(2)vii.

(1909),pp.

l-i2"

149.

be

complex,

13-46]

apparent discrepancybetween

The

[Note.
donald's

INTEGRALS

INFINITE

is

paper

these

fornnilae

difi'erent definitions

of the

consequence

413
and

the formulae

adopted for

of Mac-

the function

Q^"^;

" 5-71.]

see

Other
formuhx

three Bessel functions may


involving
z by x, multiplying
(1),replacing
by

formulae

" 11"G

2 /p {x cos

and

be obtained

by taking

6)1x'^

integrating-.
It is thus

found

that
dx

/"^
.

(7)

./^(" cos ^

/^ (" sin (^sin 4))Jp (" cos ^)


(I")

cos

"

"^

Jo

^j
.

cos'^

sin" 6 sin" 4"

cos'^ O

cos''

^W-'-Tip

l)[V{v

l)Y

1,

00

X
n

r(^a+|/'4-ip-lX

111+

X2F,(^
v

"

2-^1("

n, fx +

i"

2F^{-n, /j,+

R
0 is not

cos

Some

equal to

(/i+
+

of this
specialcases
of Sci.,K. Acad,

Some

extensions
...

it is easy

(^

cos

of formula

to shew

l;

l;

p +

"

p +

n;

l;

cos-

1: sin- (/")

I; sin- ^)

v+

R (\) "

2) "

/x

"

^~

"

0).

"

been

given by Gegenbauer

Amsterdam.,

(3) have

iv.

in

letter to

(1902),jip. 584

"

Kapte}-n,

588.

been

given recently
by Nicholson*.
arrangedin descendingorder of magnitude

numbers
positive

a,",are

"

Wet. te

can

result have

Proc. Section

If tti,rto,

"

"

when
and

"

1)

that, if

then
dt

(8)

JJaJ)

n
.

71
1

^-

0;

the

of establishing
method
this result is by induction, by substituting
sjjnplest
Gegenbauer'sformula of " 11 "41 [on the assumption that R{i')" h] for
and then changing the order of the integrations.
./^((/"i_ir)
J,.{a,"t),
"

When
a
or

"!, a..,

polygon,the
III

...

""i

are

such

that

be

they can

is intractable unless
integral

the

lengthsof the sides of


(thecase alreadyconsidered),

4.

Quarterhj Joiinud,

in i^13-48;

xlviii.

(1920),pp.

321

"

329.

Some

associated

integralswill be discussed

414

THEORY

When

BESSEL

OF

foi'm the sides of

tti, a.2, as, a^ can

1 6 A-

IT

("!+

[CHAP. XHI

FUNCTIONS

write
we
quadrilateral,

tts+

Oo

2a,i),

ttj

"

71=1

is the

that A

so

of the cyclic quadrilateral


with

area

be

evaluated

in

sides a^, a.2,a3, a4.

0: but
simple form onl}-*when
its value, it is simplestfirst to obtain an expression
for the integral
to deduce
and
deduce
the
value
for
0
when
"
continuation:
the
v
R(v)
|,
by analytic
different forms accordingasf
value of the integral
assumes

integralcan

The

a4 $ a2+

th +

i.e.accordingas
We

write

A- $ aj
ct-

o/

a.;-+

2a2 ^s

"

cio

a^ a^.

^,

cos

tts,

and

replaceJ^,(a.,t)
J^(aJ) by

Gegenbauer'sformula, so that

Jo
Jo^^^^^^^"^ V\^^\)V^,
=

_(a,asr(a^a,r;^

or

aa +

"3

We

when

y._ ^.i.-j r^._

=ai

a^ and

,^

"

the upper

yi.-*

limit

sin^"A#

by

aj +

a^

this expressionwill
A\\i{a^a.2asa^\

be

"

ot

is the smaller.

whichever

(o,

limit for

that the upper

We

"57

write

called

+ a4)(fli

",)-

is 1

a-

or

(ao

'

^3 )-

\\k.
continuation (unlessaj + 04=00-^
carry out the process of analytic
the integrals
divergeat the upper limit if t- 0), and we get
now

a;,,

"=i

"

Ts-

so

the lower limit is givenby

where

sm-c^^^^

^^

Jf)(ant)tdt
+ "4)'
I[[("i
-

-.

^']{^'
-

(Oi

a,y][zT-^(a, 03)-}
+ fls)'^
{(rt2
t^i]-^t:r(^^
-

TT".'
rlorl/i-

";^.

Hence
1
rx

(9)

Jo ("nO ^C?i

n
n

j^Njaia-iOzCii)

Itt^A

4
=

K
'

.TT^/^^a-.a^asai)\y/(aia20sa^)/
denotes

where

whose

modulus

For

t We

other

the

of
integral
completeelliptic
is less than unityis to be taken.

values

still suppose

of

that

it is

expressibleas

"j^"._,^ 03^04.

the first kind, and

hypergeometric function

of three

that

variables.

one

13-47]

INFINITE

415

INTEGRALS

has also evaluated

Nicholson

{JAat)Y^,
when

and

R{v)"0

case
special

The

0.

"

of the last,so
1

and

as
simplestprocedureis to regardthe integral
equal to

that it is
1"

in

"

,M

,-,

sin'-''
(16
(j)

hence*

(10)

^13'47.

I^/^ediscontinuous

integralsof Sonine and

Gegenhauer.

Several discontinuous
Schafheitlin

type, have

modifications

first example^ which

The

(1)
^

of these

of a more
integrals,
generalcharacter than the Webersome
investigated
by Soninef and Gegenbauer;!:;
of
in
are
integrals
importance physicalproblems.
been

shall take

we

is due

to

Sonine, namely

\'j."fit)^^^'^'t^"dt
(^2 ^-)-"
'^^

Jo

("." 6)

0,

To

and R(v)
convergence, a and b are taken to be positive
b, then we take R (v)" i? (/x+ 1) " 0. The number
2 is an

"R(fx,)"

secure

if a

reduces
complex namber, and the integral
when
is
z
zero.
integral

to

case

that, if 0

1 ;

unrestricted

of the Weber-Schafheitlin

involved being absolutely


we
integrals
convergent Ij,

The

"

see

from

" 6'2 (8)

0, then

j^Xbt^^^t^'^t^-^dt
t^ + 2'

J^ {bt)f'^'

u-"-'

exp

la { u

"

dudt

(o"

27rm'^+i'
*

An

between

arithmetical
0 and

t Math.

Ami.

X Wiener
" This

error

is obtained
xvi.

in Nicholson's

du.
liu

has

work

been

corrected.

The

result

for values

oi R

(v)

by analyticcontinuation.

(1880),p.
is also

"la

_";

38

et seq.

Sitzungsberichtc,lxsxviii.
formula

u
b'-)

exp

(1884), pp. 990"1003.

investigated by Cailler,il/e'm.

de

la

Soc.

de

2^hys. de Geneve,

xxxiv.

(1902"1905), pp. 348"349.


IIThe
this

convergence

is absolute

condition, the formula

only when

is to be established

{v)" R(ijl+ 1}"0; for values


by analyticcontinuation.

of

not

covered

b}'

416

THEORY

When

OF

indefinitely
great semicircle on the right of
integralalong this is zero; but, when a ^6, we
then

obtain the formula

we

"-

be evaluated

may

imaginary axis, and the


to apply" 6"2 (8),and

have

in

similar

cu

% |^')f-'-v^.w("=+6=)!

manner.

We

;^\in4ivaluat

1
f,and that R (/x)"
suppose that a and h are positive
is
convenient
it
to
that
integral
!arg2'!"i7r,
though
suppose

the

an

stated*.

/;J. (^o^i^fy

(2)

the

into

integral

related

[CHAP. XIII

FUNCTIONS

involved in the last integral


maybe deformed

h the contour

a "

BESSEL

"

the

subsequentlyextend

of values

range

of

\a.rgz\"^tt

to

we

may

by analytic

continuation.
From

that the

" 6*22 (8) it follows

on
integral

(2) is equalto

the left of
~

t- +z-

ha[u

"

^Jo

dudt

+
u

.0

{a-+")u

'''^"""^P
S^^lo

make

arg

"

z^^

Jo

\^

i7r.-i-''--i' '-

\ir. If

put

we

iy,where

y"0,

find that

we

y J'

bi^ {\/(a-

du
'2u

this is the result stated.

by " 6'22 (8); and


Now

az'

2a

62\) V-/X-1

[/.-.-.
{yV(a^
l^^^^^^l

providedthat R(v) " 1 ; and it is supposedthat


the singularity
indentation
above
t
y by an
is given to v'(i- y-)which makes
interpretation
=

"

+ ")}l
6^)i tF._,_,{y^/(a^
-

the

avoids
path of integration
the singularpoint,and that
the expression
when
positive

t"y.
If

we

axis and
In

had
the

above
*

pp.

signof

should

iy,we

i would

have

have

been

had

the indentation

below the real

changed throughout(3).

particular

(A\
^^^

where

put

]0

exp{-"\/(^" .V^)];.

T ^^^^
(ht\
^"

For

t With

of
physicalapplications

_
"
'

w^^'

the upper or lower sign is taken


the axis of y.
or below

122"

exv\+il|^l{a"^+b^)]

"

"

this

integral,see

accordingas

Lamb,

lil.

certain

limitations,a and

b may

be

7u,^:rb^r~

complex.

Proc.

the

London

indentation

Math.

Sac.

passes

(2) vii. (1909),

13-47]

INFINITE

last formula

The

Bateman,
If in

(with the lower sign*)has been used in physicalinvestigations


by
Phjsik unci Chemie, (4) xxviii.
(1909), pp. 682
683; see also
and OpticalWave-Motion
(Cambridge,1915),p. 72.

der

A7m.

Sommerfeld,

Electrical

"

divide

(1) we

by

6'^and

make

and R{\v
providedthat fOO
established independently
by the

0 and

"

In

R{ix)"

\)"R{iju)"

"

obtain Sonine's formula

h-^i),we

1; this might have

"

been

method.

same

have

from (2) we
Similarly,

if a

417

INTEGRALS

I.

(5) replacev by 2v, a by

2 sin 6 and

from
integrate

0 to 6

^tt.It

follows that

Jo

this is valid when


The

R{v
the

on
integral

TT^^-'-'^-i

V) " R ijx)"

"

sin^+^d*

1.

"

2/-

+ 3, and
2//-

Jo^if-\-z^

we

^.

...-.

of ^:; but the

powers
then have

^ ^^

f^^^!M^l + fl)}
*

(S^
^^^

.'o

in
rightis easily
expansible

of interest is when

case

^AtL"J:}
|-^^^..-M-:(2^sm^)^^.

r J-''Ml"^^ rwdt=
'{t''+z''-y

n\
'^^

only

i)

H
(^z^
'^^-2z^^7^^'^-''^'
-

that

so

"

I"^ ^Y('^^^^"y-^da

(9)
and

these

valid

are

if R (v) " h

(when y=l) by

different

manner

theorem

luhen
thatij:,

v"^

of considerable value in the


Some

variations

multiplying
by

the upper
*

the

arc

due

last formula
it we

established

was

deduce

the

in

important
is
integral

Diffraction.

discontinuous

integralare

with respect to
integrating

obtainable

b from

Cf. "
w.

der

beingb
integral

to Professor

Love

for

or

pointing out

a,
to

whichever
me

the

is the smaller.

of emj^hasiziiig
desirability

Phijsikniul Clicmic,(3)xvii.

(1882),pp.

1010"1011.

10-45."

v.. F.

by

0 to b.

ambiguity of sign.
t Ann.

that

limit in the last

My thanks

Struve's
Il^{x) is positive.

x"0,

Theory of

then

l-'^!rPp
H. (2z)

Struvef; and from

of Sonine's

h'^'^^
and

It is thus found

and

The

27

418

THEORY

If i

put

a, the

"

on
integral

sin 6 and

Jo

to

1) "

replacea by

we

from

deduce

to

[CHAP. XHI

FUNCTIONS

rightseems

intractable,
but, when

b"

a,

we

that

'

z"

providedthat R{v
If

the

BESSEL

j,,,(tt)'^^^^M^t^dt=^^^^^M,
b^+^
{t'-hz"')^"

(10)
^

OF

00

i*

after

"
(/u,)
"

in (1) and

dividingby

1 ; this is

then

of Sonine's

one

take a^b

and

integrals.

with respect
integrate

find that, tvhen


we
""-""',

is restricted to he

positive,

u^"-^

(v^+ b-)"

2"-"-' Jo

by " 13'3 (4),and

)^^m(^";

(11)

"h

providedthat a
positive
may now

and

(12)

see

we

that

(v+ 2) "

may

(lo)

i^(v+ 2)" J?(/a)"0

first method

"
(fj,)

1; the

'^^ ^'

restriction

that

is

in the form

and

at-

^^

in two
by
generalised
ways
Neumann's
and
integral
substituting

has

6"a,
of

^^

been

the second

Bessel function.

gives
^

J^(6")

]/^^^^) it'^z^)^^

2^-'r{v)

be written

by the usual methods


Gegenbauer*,
(cf." 13-1)for
integi'al
Gegenbauer's
The

"^^

(^2_,.^2)j.+i

be removed.

(10),which

Formula

where

thence

^^

+ W
(^2^^2^i(A

^^^ ^cos(X-.)"^#rf^
^ioio^'^(^^^-^^M-T^iM^-

2"^-a^(/A)/^(a^)JA(a^)
6

providedthat
*

"

2a and

Wiener

22 (i/
+ X +

f) "

lxxxviii.
Sitzungsberichte,

i2 (/i)
" 0.
(2),(1884),pp. 1002"1003.

420

anglebetween s^ and
the angle 6,nbetween
Now

let

Pni't^';
ch,
the

from

the distance

then, in the two-dimensional

cim+i,

and

tt

"

"2"

"m) denote

""")

the

that
probability

starting
pointshall be
and

dPn(r;

problem,all values

of

equallyprobable.

are

tt

lies between

that the distance

[CHAP. XIII

FUNCTIONS

BESSEL

OF

THEORY

that the

so

stretches

)i

probability

Sr is

less than r,

after

...,an)^

a^,ao,

dr
evident

It is then

that

Pn (r;a,,a.2, ...,an) .-^-^;^i


=

^"_2 assume

^i,^o,

where

...,

only such

assume

| |

all values

values

"

"

"

Id^n-i dOn-o

between

tt

"

and

tt,

dO, dO^,

while

6n-i is

to

make*

as

Sn ^ r,

for each

set

Now

of values of ^i,^o,

...

dn-2-

("13-42)

^';";;^
rrj,(rt)j,{s,,t)dt=\l'
[0,
r)
Jo

and

so,

if this discontinuous factoris inserted

of values of

range
We

change the
remembering that

be taken

6n-imay
order

of the

S"ii

we

{Sn

ji_2

(n
be (" tt, tt).

the
\)-twple
integral,

with respect to ^,j_i


and
integrations

^S)i_i ft,jcos

ft'H

Jo {Snt)dt ddn-i
Ji (rt)

'27rr

-TT.'O

t, and,

t/ji_i,

J, (I't)
Jo (s"_i
t)Jq (a"t)dt

.0

by " 11"41 (16).

We

make

next

S'n"i

^^

the substitution

S",i_2 T

ft'"i_i

cos
"^5)j_2
ftji,"
1

C/ji_2,

with respect to dn-2- By repetitions


of
performthe integration
deduce
t
hat
we
ultimately
r--c

Pn (r ; fli

"o

...,an)

this is

this process

\ J, (rt)U
.'0

and

"

get

and

to

in the

"

J^ {a,n
t)dt,
1

Kluyver'sresult.

W^e shall

consider the

corresponding
problem for space of p dimensions.
In this problem it is no longerthe case
that all values of 6"iare equally
likely.
If generalised
is regardedas a co-latitude)
are
polarcoordinates (inwhich d,,,
solid angle contains 6^ onlyby the factor
used, the element of generalised
now

sii\P~^6^1d6m" and

6,nvaries from

polaraxis

us

enables

to

It is to be remembered

0 to

disregardthe
that 5"i is

tt.

The

factor

function

symmetry

dependingon

with
the

of the variables di,

6^,

respect

to

the

longitudes.
...

^^-i

"

13-49]
If

P"(9';ai,ao,

less than

deduce,

we

r,

the final distance is

that
probability

the

a^Jj) denotes

...,

421

INTEGRALS

INFINITE

before,that

as

P"(r;a"a.,...,an\jy)
[i {-kp-i)

make

s" "

.'0.'W=]

extends

the values

over

of ^"_i

r.

tactoi- which

discontinuous

The

.'c.'o

with respect to O^-i


integration

v/here the
which

1 W)

we

now

is

introduce

then, since, by " 11'41 (16),

and

infer that

we

F,,(r

"

a,

a,,

a,

...,

|p)

[Ti^Wr^

|J(irtP'^ J,,{rt)U^^i^^^]

'^'-

all equalto a, and n is large,


we
"" are
a^, cu,
displacements
may
The
of
the
value
the
integral
by Laplace's*
important
approximateto
process.
which
is
of t,
for
for
such
values
t
the
is
the
small,and,
of
part
integrand
part

When

so

the

that

...

("13-3)

P, (7-;
a,

a,...,

(^rt)i'-'
Jip(rt)
(-'-|^')
jT^/J

\j))

approximationhas

This process of

been

Pearson
jj 2, jj 3, while
with the problem.
connected
the

cases

13'49.
The

has

dt

exp

The discontinuous

carried much

further

publishedvarious

by Rayleighin

arithmetical

tables

integrals
of Gallopand Hardy.

integral
^

is convergent if
condition

must

be

(x+ty

{z + tr

and R{/x + i')"


positive
replacedhy B. {fi+ v) " 0.
and

are

I; when

b the last

by Gallop,
^ has been investigated
specialcase of the integralin which /x 0,
has been
which
b
and
the
in
232
case
a
234;
Quarterly Journal, XXI. (1886), pp.
The integralis
Math. Soc. (2) vii. (1909), pp. 469.
by Hardy, Froc. London
investigated
and
Schafheitlin.
Weber
o
f
discontinuous
with
the
be
associated
integrals
to
obviously
The

i'

"

La

theorie

recognised

as

analytique drs Probabilites


somewhat

disguisedform

(Paris,1812), chapter

of the method

of

in.

steepestdescents.

The

process

may

be

422

OF

THEORY

the

evaluate

To

is

Hardy

that

so

function

all the

t in

Write
evaluated

Poisson's

and

integralmay
which
integrals
let

and

first let

us

substituted

will be used

R{v)"

for

the

by
^,

"

second

vergent.
conabsolutely
the integral
to be

are

^= Z, so that

z"

discovered

take

t"

(16)"

[-J,{a{Z t)]
1 .C
~izv^'^^'

f"

r (. + 1) r (i) 1

2.(16)"

^^^

[^ ^

r(i;+ i)r(i)Jo

'''^''^
{bt
cos

i2aY
case
special

^''
/o
(,W^TV^)

" 13

of

"4

"^^''''''^^^^^

'''

cos
(bZ cos (f")sm''^(f"d(l,dt
(ji)

cos

t"^

.0

'' ^"^

at

be

a^b,

method

the

generalcase,
and

placeoi t+^,

[CHAP. XIII

FUNCTIONS

becomes

{Z + ty

by

the

that

effective;
suppose

R{/jl)"^,
Bessel

integralin

BESSEL

^^'-'''' ^'' '^' ^) ^^' ^'^'

(2).

in a simple manner
is expressible
only when /i=:|,a
integral
when
considered by Gallop,
6, the case considered by Hardy.
or
a
This

case

results

two
easilyobtain Gallop's

We

f ^^^"^^^^
Mbt)dt

(1)

"00

sina(2^+ 0

f"

'7TUbz),

{b%a)

-T

/7.^

.du

f^cosuz

7.

,,

Hardy'sformula

and

r (;^+ .) r

J. {a(^+ 1)]J. [a(^+t)]


J-. {z + tr
i^+ty
r

...

^^^

r(/.+ i)r(^

(i)
+ i)
^aj

The

reader

will find it

obtain

to
interesting
"

(z

^y+^'-h

(1)by integrating

gai{z+ t)

/'

Jo i^i)dt

z-k-t

round

the contour

has to be

The

formed

supposed that

by

the real axis and

there is

indentation

an

indefinitely
great

an

at

"2

when

semicircle

above

is real.

integi'al
J_3o

has also been

considered

+ t

by Gallop.To
*

evaluate
'

1z

z+t'

it,we

observe

that

it ; it

13-5]
and

INFINITE

the

so

be written

integral
may

in the form

ro

/""

{- sin a {z + 1)]Jo {bt)dt


j

423

INTEGRALS

sin

(^ + 1)Jo (^0 f^^

Jo

00

sin"("_M)
J

J"

.'-00

^^^

Jo

r CO

cos

fZ^+
(6^)

ra

.'

cos

if

dtdu
{z+ t)jQ (bt)

sin

"2;

00

/"CO

2z\

"

Jo

^^
^^^^

raj

r a

sin ai Jq

az

sin"(^+i)j

"

u(z + t)Jo(bt)dtdu

cos

.'o .0
ra

Too

cos

Jo

Hence, when
^

t)

^^

"

-^

sin ut Jq (6^)
cZ^cZm.

a"b,

\t\sma(z

r*

("i)

C'^

j I
J J

sin a^ Jo (6^)
dt + 2z

a^

2
,

J.
Jo{bt)dt=
^
^

cos

V{a2

/""

az

Y-

b-)

cos

22

"

""^

--"

it^

j^dit

"

V(w'

rare

az

sin
-

cosh "/6
_

but, when

^,

,^

sm(^6cosh^)rf^,
V(^^6-^)-^-"Jo

b,

"

["^ \t\sma(z
j~^^

"

/-\

(o)

t)

,;,.,,
0.
'-Jo(bt)dt

j^

13-5.

b)

evaluated by
integrals
Definite

largenumber
of the forms
integrals

of definite

contour

can
integrals

integration.
be

evaluated

by considering

^.j0 (z)H^^'^iaz)
dz, ^.U(z)
'^^(bz)JZ,")(az)dz,
taken

round

suitable

function,and that

is

it is

contours;

supposed that 4^{z)is

an

algebraic

positive.

of two
contours
are
appropriate
types. We take the first type when
the contour
is
0 {z)has no singularities
except polesin the upper half-plane;
taken to be a largesemicircle above the real axis with its centre at the origin,
which joins
togetherwith that part of the real axis (indentedat the origin)

The

the eiids of the semicircle.

pointsin the upper halfis derived from the first type by inserting
loopsstarting
plane;the contour
from and ending at the indentation, one looppassinground each branch point,
inside the contour.
has no singularity
so that the integrand
We

take the second

more

with
integrals

type when

powerful method

(cf " 13"1)

one

of the

branch

which

is effective in

evaluating

signis to substitute for the


integral
discussed in " 6'5,and change the order of
integrals

Bessel functions under

Bessel function

^ {z) has

the

424

THEORY

BESSEL

OF

[CHAP. XIII

FUNCTIONS

integrations
; since the integrandin "6'o(7) is 0 (x"'^),
qua function of ^,
small positive
the
double
where 8 is an arbitrarily
number,
integralusually
when
the
does
and
the
original
integral
interchange
so,
converges absolutely
theoretical
difficulties.
no
produces
the

HankeVs

13*51.

Bessel

one
integrals
involving

function.

the more
abstruse integrals
which
will be
investigated
he
evaluated
class
discussed in ChajDter
of
definite
a
large
xiv,
integrals*
by
considering
Before

Hankel

the first type of contour

described

in

" 13"5. In this integral,


a
is
with
a positive
a complex number
m
integer(zeroincluded),r
positive,
and
imaginarypart,
positive
taken

round

is

is

\R(v)\"R{p)"

2m

the convergence
tends to zero; and (as a

first inequality
secures

The

of the indentation

tends to

that

inequality

second

the

ensures

as

zero

the radius tends

the

of the

when
integral

the radius

of Jordan's

consequence

integralround

the

lemma)

largesemicircle

infinity.

to

of the integrandinside
onlysingularity

The

the contour

is the

pointr.

It

follows that
1

(ax)
{ir,"^'

scP-'

J^iJQ

(2)

^~2'KiJ

0.

can

Hankel's

be

(x^
(^'

expressedin

It is thus found

BesseVschen

"fThe

was

form

neater

sin

2h + 2,
integralwith v
n, p
Functionen
(Leipzig,1867),p. 58.
evaluation of integralsof this character

is effected in " 13 -G.

by writingr

ik, so

that

l(p-v)77. F, (ax)]
t^^^^.^^

publishedposthumously, Math.

the

?-2)w+i

thatf

[cosl(p-v)-7T../,(ax)+

work

r'.yn+r~

xc~^dx

F, (ax)]
i{l- e(p-'')-0

of
investigation

functions

" 3-62 (5) that

(^f.

result

(k)"

/'""+'
^"-^.g,"'((i^)c^^

["[(1
+ gCp-")-')
J, (ax)+
J

This

if,"''
(axe^')}

(^2_r2)'"+i

It follows from

(1)

ep"'

"

Ann.

m=2n

which

A partial
401.
(1875),pp. 458
Theoric
der
given by Neumann,

viri.
was

contain

"

only

one

of the

two

Bessel

13-51]

(3)

{cos

VTT

J, (ax)

VTT.K {a:c)}

sui

f"^ x^

The
For

an

of

The

+ F)'"
(."2
2)n

(4) to the

,"

^.^^

'

,"

{ah)

" 1, and the latter when


(i/)

is net

an

integer,see

"

13-6

1"

"
(i^)

2m

-^
.

(2).

specialformula

been

xJ^{ax)dx_

II.

specialformula

another

Ko{ah)

/'"Yo{ax)dx

jo

transformation
complicated

of

_
~

^'

:^ + i^

repeatedintegrals.

integralsresemblingthose just given

though it is most
Thus.

xviii.

19; while

(Cambridge,1889),p.

^^)

Some

dynamics.,
(1881), p. 194, and Basset, Hydro(1911),
Nicholson, QuarterlyJournal, slii.

3/ath. Ann.

bj Mehler,

pointed out

p. 220, has obtained

K^

"'

when

r-

by

"

2'"

/i

f"

case

TN

has

k"

is valid when

former

a'"

",

J0

extension

^Jy (ax)dx

^^."_^j^,yn^i^^

^'^"

of this formula

followingspecialcases

notice the

should

reader

The

425

INTEGRALS

INFINITE

convenient

them

prove

observed

has

Nicholson

to

using Cauchy's

without

cos
(ax cos 6)
f"^f^-""

2
_

+ IC-

X-

established

here,

theorem.

that

Jn{ax)dx
0

be

may

TT

^' +

Jo

T^

,^^
U.l/1//^

1^'

a^

.' 0

'^-j^[I,{ak)-lMo{p.k)\,

that
by " 10'4 (11),provided

(^^^^^

(7)
More

if R (v) "
generally,

/. (ak)

and

and

since,by

of

and

providedthat R{y)"'2
/.(a^)-L.(aA-)
=

.t"

cc

so
positive;

that

^{Io(ak)-I.,(ak)].

have

(2),

is

both

sin- ddd,
f'^-"^--^
j.,^^2itra^

(k) are

|,we

"

L. {ak)

form
special

A;

/i'""

it follows
positive,

that

j_."^1^^
-j.^^^^^^,^j^

^^'^^

_ki~''r' Jy{ax)-t in,, (ax)


TT

j
_

r^

I'"v

TT

cc

^'"

kdx

J. (ax)+ i (1 -e^-')llAct A')]


[{I+ e^-^)
^^^,

426
and

so

have

we
ft

formula

the

CC

f4

B,

a"0,

0 and

(k)"

Jo
ii R(v)"
k

choose

we

0.

(a)"

" 616 (1),

have, by

(k) " 0, we

xu.dudx

_r(v ^) p p (2a)''
f x''K^{ax)dx
'^^+k' ~~r(i7 JoJo (x"- k^{u'
+

cos

Jo0

Jo

-~2krJl)
TT

4
use

we

(9)
^
^
and

j0

been

has

R(v)

(when

Jo

^,

"

^^

-j,_ ^^^
'

I'TT

|,

0)

due

are

Nicholson, and

to

the

also

last has

Heaviside.

f"=^"!^
k^

Jo

^ and that

"

"

cos

x^ +

x"

investigated
by Gegenbauer*.

been

and

integral
^

R(v)"

I/TT

when

+ ^^-

i^ (v)"

formulae

given by

The

[II.^,(ak)-Y_,{ak)],

cos

therefore,when

These

(w^+a-^r+i

^r^^-^
^-^.(a^Oc/a;
^'

a'y+'-

ZjgV-1

" lO'-il (3). Hence,


"

(jM- i) r""(2aye-''^du

TT
_

when

of the

order

'

k'

x' +

and

"

that R

so

in the

(ax)dx

r"^x^Ky

If

change

is
integral

similar

Avhich converges

^^

theoretical difficulties.

integrations
presentsno great
somewhat

[I^(ak) L^ {ak)\,

"

^.,

The

R(v)"2.

-^"

^__

where

/yt

[cos\v'K.J^{ax)+ sin \v'k H^ (ax)]^"

(8)

[CHAP. XHI

FUNCTIONS

BESSEL

OF

THEORY

x' +

k^

and

x"

T(v

r(v

(k) are

both

To

evaluate

we
positive;

^)r(l)Jo '("

x-

it,we
then

suppose
have

that

k'

i ) n'' ./ 0
^)r(^)kj,

so
"

f '^f^

(11)
*

he

Wiener

omitted

formula
p.

11.

are

The

pp. 422"424.

the term

also incorrect.
correct

result

{/,(ak) L, (ay^OI-

L,, (ak); and

-^

(2),(1876),p. 349. Gegenbauer's result is incorrect


consequentlythe results which he deduced

Sitzungshericlite,
lxxii.
to insert

similar
was

error

was

made

given by Gubler,

by Basset, Proc. Camb.


Zilrich

Phil.

Soc.

xlvii.
Vierteljahrsschrift,

because
from

vi.

his

(1889),
(1902),

428

OF

THEORY

Hence, when

we

BESSEL

and

vir

of

expand ^^'^'(az)in ascendingpowers

7j-gi(p-"'-2fi)ir?
sin

[chap. XIII

FUNCTIONS

V(h_p+iv

(^a)"kP+''-^'^-'-S

+ 1)
(/J,

w!r(y

=0

1) r

??i

find that

z, we

m).(laky
(^p+ ^1/ /A + 7?i)
-

therefore
xP~'^dx

r^'

(1) I [cos(^p

sin {Ip -Iv


IT. J, {ax)+
|i/--f.i)

2siny7r.

fx)it. Y^ (ax)]

,j,2\^+i

r(//.+1)

+ ^v-fMy
lr(v+i)r(ip

'

a^k-

"

2
T{i-v)r(^p-^v-fiy'^'{

Fig. 31.

It is natural

to

enquirewhether

singleBessel function cannot


method
of evaluating
it is the
HankeVs

method

by

Hankel*

instead of

of

only effective

is

they contain a product


been investigated
integral

that

singleBessel function,have

by applyingCauchy'stheorem

a'^h "0,

the

explainedin " 13 '(J.

will be

" 13'51,except
to

the

ir;i"{az)dz

if
in which

which

that

seems

contains

a productof Bessel functions.


integrals
involving

Integrals
resemblingthose
of Bessel functions

this type which

integralof

be evaluated; it

13'53.

the

with
7' is a
positiveinteger,
complex number
positive
imaginarypart,^^ denotes any cylinderfunction of order fi, and
m

\R{v)\+ \R{fi)\"R{p)"2ni
*

Math.

Ann.

viir.

(1875),pp.

461"467.

4".

13-53]
i" a
[When (^^/n
inequahty.]

429

INTEGRALS

INFINITE

Bessel function of the firstkind, \R{fj.)\


may

be

in
replticed
by "It{fj.)

this

When

h,the presence

of the

in order
inequality

above

term* in the asymptotic


expansion
non-oscillatory
that we must
2m -f 4 by 2m + 3 in the
replace
the integral,
when taken round a largesemicircle
zero
as the radius tends to infinity.

to make

the real axis,tend to

The

contour

to

of that

manner

of

shews
integrand

be taken

is that of

" 13"52

; and

if

proceedin

we

section,
we

^^^

"

-^.I

rf~'
^^'" (a^-e'^Ol
{hx)jy,'"(cix) e^-i9^^(""e-0
['2^
27riJo
^^i^^2_^.2yn+i

(1)

'

'^^

'

limit

non-zero

as

{az)
-g?^
=

then obtain the

(2) 2^.

The

^^,-

that, when

1)6 pointedout

It must

semicircle tends to

'

of this result

cases
special

Numerous

we

of

(1)in which

2,n

take fM

as
according

the

large

0 and

is a
'"'^

Bessel function

; it is

mention
special

ccclor

and

and

R{/j,)"\H (v)j

/7

-1

we

see

that

{i7riJ,(br)HJ'"(ar),

^^f^^'
^

in the
existence of the disconti niity

for this integral


was
expression

Hankel.

modifyformula (3) we
[^

Y^ (ax)]

"

2.

R(p)"

^ (/j, v)tt

sm

^ 6.

pointedout by

(^)

["

we

"

"

providedthat a^b-"0

If

ronnd
6, tlie integral
ii
and,
infinity;

{hx)[cosl(fx-v)'7r.J^
(ax\+
Jy.

The

(az),
c, ff^i'-)

{az)^

r "^

we

"

the radius tends to


c, H^m

Hankel.

givenby

are

2?H + 3 and

^J'^ (axe-i)]
[%\ {ax')H,i^){ax)+ "W {axe-^)
(^^23^2)^+1

case
particular

(3)

,-.

formula

new

of the firstkind deserves

If

the

find that

that,if a

see

^ 6 " 0 and

{k)" 0, then

^M (bx)[cos]j{P' v)7r.J^ (ax)+


"

"

^o

7i

sin |(^

"

y) tt
=

Since

H^'-')
{a~2(az)H^"'^i
~

c-

"

^'^ ''""''when

dx
F^ {ax)]

r^{bk)K, {uk).

j" |i large.

430

OF

THEORY

More

BESSEL

generally,
takingequation(1) with

(^)

[CHAP. XIII

FUNCTIONS

0 and

-I^{hh)K,{ak)kf"-\

multiplyby
0 to tt;

(7)

replacephy

2ac cos 6),where


-\-cby \J{a'^
a
nd
with respect
cos
6)^",
integrate
11"41
Gegenbauer'sformula, "
(16),

sin-"^/(a^
+ c^

find from

we

AC'^)

m( V
^

X-

p -\- v,a

"

2ac

"

Jo

be

process may
a"b-"r 2c, then

c"

"

b,

0 from

to

j-^^g
i(p + fj,)TrJ^
(ax) + sin ^ (p + /i)tt Y,.(ax)]dx

fC
=

This

+ s\u\ip-{\GO"\{p-^
fJL-v)Tr.J^{ax)
ij,-v)Tr.Y^{ax)]dx

"

^TTT'^

In this result

have

Wfj, J^, we

repeatedas

often

as

jj j^ (^^^^^-^
f "^"^"'^'mC^^)
h^\p
|-^Qg
J

-I^ {bk)I, (ck)K, (ak)kp-\


and

please;

we

we

find that, if

"

(8)

pi-"{N-l)v]7r.J, {ax)

+ k-

"

sm^{p

+ p, +

{X

"

1 ) 2^i
TT

Y^ (ax)]dx

4 (bk)fl/, (c"k) K, {ak)


.

"

k''--.

Again,by considering

2^-f,^"[n^'^(^^)]^''''("^)
round

factors of the

(^"

used, where
previously

the contour

^{p

providedthat a"^[R{b)
lip + 1/ii
"

i^

of the firstkind
of the first kind
for
necessity

is

{p-h

differ in the different

generalresult

Y^ (ax)]dx

-[U

K^ {ak)k"'I^ (bk)]

"
(a*))
R(v)r

on
th^'integral
integer,
th^eftinvolves functions
of productsof functions
the integrals
only;a result involving"
of this type was
who overlooked the
given by Gegenbauer,
even

an

this restriction

extension

of Hankel's

(cf "13"51).
results is obtained

by considering

H,'^{az)
f.o-i-^.l^x/(^^+r-)]

round

IT
'Sfi- }.")

and

/x

^/j.-p)7r.J,{ax)

sin

b and

more
slightly

-l^osi^ip
J,-r^,U^^"^(^^''-)]
4-

An

the

obtain

product,we

both

the contour, where

^ ft"

0, m

is

and
integer,
positive

+ R{M').
\R(v)\"R(p)"2m-i-^/i

It follows that

1
2m+i."it

\^.(Ij

13-54]

431

INTEGRALS

INFINITE

and, in particular,

result obtained

(1880),pp.

XVI.

in

much

of Nicholsons

interesting
consequence
Nicholson*, namely that, when

method

by

Math.

Ann.

and

are

this result is obtained

which

Jo

integral.

of Mehler's

An

The

by Sonine,

manner

56"60.

Generalisations

13*54.

elaborate

more

p~

is

follows

as

due

to

-"r h'

-r [^-r^
k~
J

TT

of " 13"ol (5) is


integral
positive,

Jo [a\/(r+ ^'

cos
-P^''
"/")}
^dp.

J 0 P' +

repeatedinteg)-al
convergent double
may be regardedas an absolutely
make
since the integrandis 0 (p~-)when
a change
integral,
p is large.Now
of originof the polarcoordinates by writing

This

and

we

(fi k

cos

1
zz/

and

cos

0,

p sin

sin

6,

have
7^

r/

/N

[^ J,{ar)rdr

JAar)rdddr

this is the result to be established.


To

the
generalise

result consider

-z^-'H,^'^(az)dz
(^ + 4kY^'
taken

round

the contour

shewn

in

Fig.32.

Fig. 32.

It is

supposedthat

is

and, to ensure
positive,
convergence,
\R{v)\"R(p)"iR{fi) + \K

Quarterly Journal,

xlii.

(lUll),p.

224.

432

THEORY

OF

BESSEL

[CHAP. XIII

FUNCTIONS

It is also
the

supposedthat |arg k\" ^ir,and

the

loopsin

the contour

surround

points

By analysis
resemblingthat
(2)

J,^
ft

of

" 13"52, the

reader

will find that

+ Bm{lp-^v-^,Ji)7^.Y,{ax)^]
[cos{\p-h^v-2^l)7^.J,{ax)
,

,,

_,,,

^,

,"",

."",,^

._,.

_^,,.

..,,_,_,

^^_^^^^^^^^
I

(^ ^2)"-^-

TT

'-

".

^i

_r:^

cQs

+ l)r(^/j+^z;-/i+i??i)
l) [",=o"i!r(y+7?i
2sinz^7r.r(yu,+

00

'^

;"=ow!r(-y+m
If the series

on

that the former


or

if /3

or

if p

"

"

The

""

cos

TT

l)r(^|0-^i/-;i + i??i)

the

rightare compared with those given in " 5"41,it is seen


productof Bessel functions if p 2 ?/= /i,+ |
Up"
ijl-\-\
yu.+ |, while the latter is so expressible
+
t|.
/A
is expressible
as a

"

'2

1/

"

"

which
corresponding
integral
in " 13'6.

contains

singleBessel

function will be

considered

Sonine's

13* 55.

of definite

number

Sonine, Math.

Ann.

xvi.

of
integrals,

(1880),pp.

most

specialforms

66, can

"

"liriJ
a

to
integral

generalcontour

be evaluated

were

by

given by
the method

contour

be taken

is

pi u+k)

round

63

which

integration.

of contour
The

integrals.

(z + ky

of
consisting

the parts of the circles

\z\ 8, \z\ E,
=

lines arg
of these cu'cular arcs*.
terminated

by the

(" ^)

" tt, and

the lines which

and

extremities

supposed that ;n is an integerand k is not a negativereal number.


=8 tends to zero as S^-0, providedthat R(p)"\R(v)\,
round
z
integral
round \z\
R tends to zero
ii
the integral
x
as
providedthat

It is
The

jointhe

^-

R(p)"
By Cauchy'stheorem

we

+f.

have

27rij
*

Cf. Modern

Analysis, " 6-2;

or

" 7*4 supra.

13-55]
and

thus

433

INTEGRALS

INFINITE

have

we

"

7^ 1
0

2i cos

+
^^^^ip + v)'n^^'^'^^

In

particular,
takingm

(2) ki'-^H/^(k)

0, we

consider the

we

2i

z^tt cos

cos

sin (p
t'F^(,r)

ott]

z/)tt]c/^'.

"

integral

find that

we

(k)
(3) kfl-'H,^-^

[/,{w){sin(/J+ ;')TT

"

-^

If

take p

we

1, t-

0, we

r"sin("+A;)

The

F" (A;)
=

last two
More

results

to

w.

B.

F.

See

"^

and

-l"R{v)"h,

we

get

k^[JAl-)"iY,.{k)],

r^

1
_

--^

v-tt

j^
.

""

J (,,
+

,x

,ft,
i.)
from

(6) and (7) of " 13-42,Sonine deduced

-^ r
Jo

'^k dx+X

1^^
sin (k
TT

f^

cos

(5) that

6) d0,

.'0

^0 (^)f^^ +
On (a;+ /t)

alsoLerch, MonatsheftefUr Math,

und

r^"'

sin(A'cos^ |?i7r)cos7i^rf^
"

"

''^ .' 0

""3-56(2),9-11 (2),
2

(8) Yn(k)

-_-^^

from

tt]dx.
z/)

Sonine*.

to

F" {h)

hence

ott!

f cos(^^-)^^_

cos

IT

and

"

cos

Sonine.

u""ig the formulae

(7)

^
-

cos(a;+ A-)

-J.-

By wr.t,"g

ptt

IC

-T

a;

.le

result also due


-D

.'0

"^^J^^a^)dx=:^^
2i
+

TT

and

due

are

takingp
generally,

(6)

n,\
TT

(5)

cos

get
2

{A\

2i

sin (p
il"^(a;)

"

"

[J^ (ic){sin(p + v)7r+

."

iYy (x)sin (jO v) tt]o^a;.

get

If

p7r}

cos

vtt

("^
+ A;)"'+^

j0

Phys.

i.

(1890),pp. 105"112.
28

434
A

13*6.
We

function,under
order of

the
reversing
As

for the
integi-als
by substituting
the integral
sign,the definite integralof " 6'5, and
the integrations.

evaluate

sliall now

Bessel

integrals.
of evaluatingdefinite

method*

7iew

definite

various

of
integral

first example consider the

af~'^ J^

R{v)"(),
is

real

R(2fi + 2)"R(p

number,
(positive)

v)"0:

in order that the

x^^-^^axy-^^

r(-g)

[" p''

integral
may

this is evaluated

When

the

poleson

22^+3-pr(^+

hypergeometricfunctions

"~-'^'[2

the

on

in certain circumstances ; the former


2 +

By

the

or

'^^"

^'

'

4/-

Bessel functions

to
+

+ 2, the latter
2/ji

(1) is valid when

of analytic
continuation
priuciple

particular,
taking p

/"

obtained

formula

enclose the

I..

""

to

^"''

reducible

right are
if p

'

-R{p)"R(p)"2R(p,)
In

as

(ax) dx

^^^^
2+iv-^p)-^^'V^^-^-

so

find that

we

+ l)r(v+l)
2''^'r(fx

(ah- 1)

the contour

(by swinging round

the contour)

rightof

xp''- J^

if p

converge.

Tj-g)

r'

-~

The

type

is equal tof
integral

The
1

(ax)

Hankel's

it is at first supposedthat

in which

and

[CHAP. XIII

FUNCTIONS

BESSEL

OF

THEORY

by

+ ^

A-^

2, we

find that
a'^k"-!^

j; {ax) dx

method

another

by Sonine,

Math.

Ann.

xvi.

(1880), p. 50;

it is

valid when

-l"R{v)"2R{ix)
*

This

Math.

method

is due

(1896),p. 233];

to

Lerch, Rozpravy, v. (1896),no.

he shewed

no

other

t The

use

change

has

been

of the

made

order

%.

23

[Jahrhiich iiher

die

Fortschritte

that

^..

but

sr(s + i)

of the method.

of the

integrations may

be

justifiedwithout

difficulty.

der

436

OF

THEORY

as
Finally,

BESSEL

[CHAP. XITI

FUNCTIONS

example suggestedby "13'55,we

an

shall consider

ai'~^ J^,(ax)doc

(x + ky+'

in which

0 and

"

|arg k\"

It is first to be

tt.

R(v)"0,

supposedthat

R(fi + l)"E(p

v)"0.

is equal to
integral

The
1

xp-'(iaxy+^

r(-s)

pp

dsdx

j27riJo
(- .) r (p +

-l^r

2.) r (^

^^,p-M-l
sin

(p +

I'

p.)TT

"

r (/u,
+ ?n
(^a^O'^+^-P+'"
! r (1 /x + i
ip + i m

wi

I'

firstseries reduces

The

Lommel's

by
expressible

['^"'
x"
,_.

2.)

^^"..^^_._. ^^

(-)"^daXO"^"" r (p + i; + 2m)
m^o wz! r(j^+ m + 1) r(p + 1/- p, + 2m)

(/i+ 1)

"

4- 1

Jt,(ak) when

to

/x

1 )sin

^ (p +

i
+ f) r ( /A

i^

J.

At

1
-

J^{aa)dx

irk"
r_,

p +

^m

f)

0, and the second series is then

we
(cf." 10"7).In particular

functions

:?n)
tt

"

,.

have

txt

providedthat "^"R{v)"^.
reader will find that

The

be evaluated

chaptermay

largenumber

of the

discussed
integrals

in this

this method.

by

Integrals
involving
productsof Bessel functions.

13*61.
If

integralinvolves the productof two Bessel functions of the same


it is likely
.that the integral
of the same
order),
argument (but not necessarily
is capableof being evaluated either by replacing
the productby Neumann's
integral
("5'43)and usingthe method justdescribed,or else by replacingthe
productJ^ {x)Jy ix)by
an

r(-g)r(/x +
27rij_oe,r(/x
+ 5+i)r(i.

\_ p"

in which

r(/x+

i/

the
+

2s+

by usingthe
The

polesof r(" s) are

reader

1) are
method

may

of

+
s

2g +
+

i)(^^-)^+-+^-

i)r(/x +

i/

.s'+

i)

^'

the

the left ; this

on

find it

on

t/

rightof the contour while those of


derived from " 5'41
is easily
expression

" 6'5.
obtain-ing

to
interesting

evaluate

3(^~^Jft.
{ax) Jy {ax) dx

/:
by

these

methods.

final element

in each

The

result is

function

combination

is a^k^.

of two

functions

of the type

^F^

and
,

the^

13-61]

INFINITE

integralformula,obtainable

Aaother
r"

437

INTEGRALS

by replacingJ^ {bjx)by

is
integral,

an

p-i

(1) /

dx
J^ {ax) Jt,{blx)

Jo
+

-H"^^'~2"
"2-^''-''+^r(v+i)r(i^+iv-ip
+ i)-"

^'^ "2"

^'

-16-

16
This

is valid when

and

generalformula
had
137; specialcases
The

positiveand

are

Rao, Messenger,
given by Hanumanta
XLVii.
given previouslyby Cailler,Mem. de la Soc.

was

been

(1902"1905), p. 352; Bateman, Trans. Camh.

XXXIV.

Hardy,

and

An

the

of functions

case

example of
interesting
I

lim

discussed

who

contains

^)-iri2-vy^
4^!^l,)^^^^^'si8-i^(^^
+

be Avaitten in the form

may

""xfT-/Nr

1 dx

^icsinI'TT

/x

It (2)
Jv(l^)

f (" + 1)

f (2

sin
-g^i^;

exp {

logipl.

yTrl dx

r'

/"

,.

T^

,
_

/-i+cc*

r(-s)r(25

s{s+l)V{v
^-rrij^-^i
1
=

"ri w. (n + 1) ! r(y

this series is
To

obtain

small and
at 0 and

an

]argp
"

1, and

"

r, /"

"'^"^
,

i,\

rfs

2)

s+\)V{-l-v

s} (2^;)--'

(27i+ l)!(2p)-2"

^
"S

")

easilyproved that

r="

and

productis

"^

It is

xxr.
(1912),pp. 185, 186;
(see
|,
" 6'23).

"

"

r""

"

{-p'x')J^ {x)/i_, {x)

exp

which

134

Phys.de Geneve,

Phil. Soc.

of orders

integral*which

an

(1918),pp.
de

an

function
integral

1) r(2
of

1/

n)

'

Ij}).

valid when
of the integral,
jja iis
asymptoticrepresentation
j" jtt, we observe that the last integrandhas double poles
2, |, 3,
simplepolesat
f,

This integral was


brought
problem of Diffusion of Salts in

"

"

"

"

to
a

my

notice

circular

by Mr

C.

G.

cylinder of liquid.

Dar.vin, who

encountered

it in

438

THEORY

Hence

OF

[chap. XIII

FUNCTIONS

find

we

f^+--'-' T(-s)T(2s

J_

BESSEL

S-jriJi^^i s(s +
sin

l)r{v

2)
l)r(2-v^s)

ds_

(:2py'

VTT

+ l)-ir(2-j,)-2\0g2p][l
^^^^^^^^--^^{ir(2)-f(p

(1

2v)sin

VTT
,

^' ^

27ri'(l-z;)
and

2v(l-v)p^]

(-Y(273Yr(hi)

^
"r3 2n (n

2)r (v +

-^n)

{2-v-

n) (n

2) !

so

(2) lim

exp

J, (x)./i_^
(-2f-x-)
(a;)
"

'^"

sm

VTT

4"7n"{l

v)

"

{l+ 2\ogi8-yJr(v+ l)-f{2-v)]

sini^TT,
1

--^;^

In the

0, we

VTT

I.)

"

+
f^^-^

find that

Jo (*")
Ji (*")
f-af)
-;;^

exp (-

X-

6^0

s -^
Zo 7r= {2m

lf+

13'7.

27/)sin

"

(_)n(2^)"r(l"l)

case
special

(3) lim

"

27rWl

(1

{ir{2)-^{r(v+l)-ylr{2-p)-2\og2p]f--^

[r(m
+

f)l3(2/9)=""+3
1 )2(27W+ 3) (2m + 1 )!
+

"

Integralreprenentations
ofproductsof Bessel functions.

result is obtainable
Gegenbauer'sformula of " 11-41 (16) an interesting
the
cylinderfunction to be of the first kind and substituting
by taking
result of " 6'2 (8) for the function under the integral
sign.

From

the

This

proceduregives

2^r(. +

and

if we

(1)

i)r(i)':^^'^^

change the

J,{Z)J,{z)^^^.\

R {v) "

"

of the symmetry

in

by making \z'\-^\Z\.

and

and

be replacedhy R{v)"
obviously

account

limit

exp-ji^
V

27rlJnC"Xl

This result is provedwhen


may

we
integrations,

order of the

Z.

"

find that

i^r~\-^''\J)J2t
"

but the former

1, and the latter may


It is also

restriction

to
permissible

be

removed

proceedto

on

the

13-7, 13-71]

By usingthe

results of

providedthat R(p)
formula

The
{)p. 152

"

"

" 6"21 (4) and (5),we

l and

"

(1) was

obtained

the

theoryof

155, from

439

INTEGRALS

INFINITE

find in the

same

further

by Macdonald,

Froc. London

Soc.

Math.

change

For

(2) XIV.

the

remove

these

physicalapplicationsof

rightin (2) and (3) was


exponentialfunctions.
the

Macdonald,

see
integrals,

I'he

of K^ {Z)K^,{z)as
expression
obtain

shall next

productK^{Z) Ky{z) as
K^, namely

formula, due

to

(2) and
form

in

given by Sonine,
London

Math.

Soc.

to

integral.

Macdonald*, which

represents the

it when

prove

Z,

of the

type

when

Iarg Z \"7r, jarg z\"7r and


it is convenient

an

integralinvolvinga singlefunction

an

is valid for all values of

This formula

but

Proc.

formulae
modified

(1915),pp. 410"427.

13'71.
We

on
integrals

being to

(1900),
Gegenbauer's

xxxii.

equations,and he deduced

linear differential

of the

modification

objectof the

that

|^ |" |Z |.

integralby reversingthe steps of the analysiswhich we have given. The


given by Macdonald, though they are also to be found in a
(3) were
Sonine's memoir, Math. Ann. xvi.
(1880),p. 61.

the

way

|arg (Z + z)\" ^tt;


z

have

positivevalues X,

x,

and

theory of analyticcontinuation;the formula, which is


to be associated with " 11"41
importancein dealing
(16),is of some
obviously
to prove the formula
with the zeros
of functions of the type K^(z). It is possible
used in proving "11"41(16);
without
tie rather elaborate transformations
the following
proof,which differs from Macdonald's, is on the lines of " 2'6.

to

extend

it

the

by

By "6-22(7)we

have

y-^{,(X)K,

(x)

\j^ r
'i J

00

e-('+")-

A-cosh

^-..co^hu

^nci^^

GO

"

U)

Q-2i'T-Xcosh(T+U)-xcosh{T-

dJJdT.

2j-c.
_.

If

be taken
(A'e^+ .re~^)e^'

I
*

Proc.

as

variable v, in the

new

Q~Xco9MT+U)-xcosh(T-U)

London

Math.

Sor.

xxx.

integral

^JJ

(1899),pp. 1G9-*171.

440

OF

THEORY

Z2 + a;^ + 2Za; cosh 2^1

1 f
it becomes

and

so

exp

Macdonald's

13"72.
We

obtain

once

positive.

are

at

we

been

developedby Nicholson*,

functions

and

the

easilyproved are

K^ {z)K, (^)

(1)

T,

to

of Bessel
representations
integral
of
5"43.
Neumann's
integral "

of this type have

most

are

series of

associated with

formulae

which

two

discuss

to be

are

||^^'l ^dT,

ofproducts.
representations
integral

Nicholsons
now

and

the variables X

when

theorem

shall

The

'

with respect
performingthe integration

on

dv

e-2.r-(x./oco.h22'
^I"J"
exp

which

^j^H

"

have

we

K, (Z) /C {x)
and,

[CHAP. XIII

FUNCTIONS

BESSEL

K^+^ {2zcosh t)cosh (/J.-v)tdt

Jo

+ v)t dt,
ir^_"(22 cosh t)cosh {/j,

Jo

Iarg z\" ^tt,while

when
To

obtain

/x and

these formulae

K^ (z)K, (^)
=

"

00

unrestricted.

are

" 6'22 (5) which

use

we

^""^^ t+coshu)

e-~

shews

that

cosh fitcosh

vu

dtdu.

CO

is absolutely
repeatedintegral
convergent, and it may be regardedas
double integral.
the transformation
In the double integral
make
The

t+u
and

2T,

t-u

2U,

it is apparent that

K^(z)K,{z)
But

^j

2 cosh

fi(T+U)

cosh

U)coshv(Te-2^'^"^h^"""i"^cosh/i(r+
cosh

ifi+ v)T

sinh

v{T-U)

cosh

{/x+ v)T

U)dTdU.

(/Li v) U+
"

sinh

(//, v) U
-

cosh

+ v) U
(yu, i/)jT cosh {fu,
-

+ sinh

(/u, v) T sinh (/x+ v) U.


-

The

of these four terms


to the last two
integrals
corresponding
obviously
vanish; and, if we interchange
the parametric
variables 2"and Uin the integral
to the second of the four terms, we
obtain the formula
corresponding

K^ {z)7C {z)
=

If

\\ \

e-^' """i^ T

cosh

with respect to U
integrate
w
ith
integrate
respect to T we obtain
we

*"QuarterUjJournal,

we

cosh {fi+
obtain

cosh

(,m-v)U

the firstform of

the second

xlii.

T
/')

(1911),pp.

form
220"

of (1).
223,

dT d U.

(1),and

if we

13-72, 13-73]
The

441

INTEGRALS

INFINITE

formula
'f

I^ {z)/, {z)

(2)

/^+,{2z cos d)cos {fj^-v)edd,

TT

which

R(fx,+ v) exceeds

is valid when

is at

"1,

once

deducible

mann's
Neu-

from

formula.
If

Ioiz)K,{z)

(3)
More

m)

"

'

"

rCiz+ i) r^- r

2
,

"

providedthat

72 (z^)" 1 ; and

|"

"

in

I,{z)K,{z)

(5)

then

replace/n

and

and

by m

"

i',

6) COS (m

v)0 cW.

(8) with " 6'16 (1) we

combine

0) cos vdcW.

K^^ra (2^ COS

we

cos

1, then

If

and

"

find that

we

v,

i'^K,(2z

^^^-^

/",{Z)K, {Z)

(4)

take fi

if we
generally,

find that,ii \R{v

we

change the signof

0 and

take fjb

we

find that

(4zY

cos

(h cos ^) cos i^^

,.

particular
Ja(u)du

result of which

generalform

more

Nicholsons

13'73.

.'o \/(u-^
+ 42^)'
has been

given in " 13'6,formula (3).

integral*
for J^-{z)+ IV {z).

The

integral,
correspondingto

Take

the formula

just discussed, which represents


It is first necessary to
J^-{")+ IV (yz)is difficult to establish rigorously.
that the argument is positive
assume
(= x), and it is also necessary to appeal
to Hardy'stheoryof generalised
such principle,
in the course
or
some
integrals,
of the proof
("6-21)
I

ZT^d)
ix)

the

e-^sinh

resulting
integralconverges

continuous

function

of X.

i/^(i)
(x)

^yj^

'
-co

integrandtends to zero as \iv\-*x on


is inserted,
factor exp {" \w'-}
an
exponential
uniformlywith regard to A.,and so it is a

the

the contour, it is clear that when


the

w-utv

in which

manner

+77?"

,' OC

"

-TTl

From

those

Hencef
lim

"

exp

|- \w-]e-^sinhw-w"^l^^,^

Phil. Mag. (6)xix.


(1910),p. 234; QuarWrlij Journal, XLii. (1911),p. 2'21.
Soc.
Phil.
Trans. Camb.
Hcardy,QnarterhjJournal, xxxv.
(HlOi),pp. 22"66;
1
48. In this integral
(as distinguished from tho.se which follow)the sign lim
with the hitegralsign.

t
pp.

"

xxi.

(1912),

is commutative

442

OF

THEORY

the

By Cauchy's theorem,
so
(w) ^ir,
long as \ has

we

[CHAP. XIII

FUNCTIONS

contour

be

may

deformed

value ;
assignedpositive

an

into

writingt +

the

line

^Trifor w,

get

^4^

lim

if^(i)
(x)

In like

Hardy'snotation.

f
J

TTi

K-^+O

in

BESSEL

{-\(t

exp

manner
Avni

/-co

TTt

with

factor
impliedexponential

an

Since

{" X, (u

exp

conditions
requisite
convergence
productof the two integrals

.'
"

(in

-x^

the

regardthe

which

CO

ei{t)and

(^
e'-i-cosh^-.^
^Trif]

-OD

are

fulfilledwhen

"0,

we

may

"CO

for the

e.^iu)stand

^iri)-].

"

as
exponentialfactors)

double

integral
J

We

-co

"

thus find that

with

the

the substitution

t+

[- X (t + Itti)- \ (u
-

2T,

"

.'

with

an

"X

impliedexponentialfactor
exp

Make

"X

"

2U

.'

lirif].

and

then

"

factor
impliedexponential
exp

In view

{-

2XT-'

(U

2\

liriy].

of the absolute

it may be replaced
of the integral,
by
convergence
the repeatedintegral
in which the integration
with respect to U is performed
that
first,
so

\_J0

"

r r
.' 0 J

with

an

impliedexponentialfactor
exp

in each

{- 2\T- -2X{U

e-2'-xsmhrsinhC^+2.-r^^(^2T
X

case

equalto
i-rriy}.

444

OF

THEORY

BESSEL

[CHAP. XIII

FUNCTIONS
.

Now, qua function of T,

(1^,
K^ {2x sinh T)\
T is small,and

when

since the

we

so

proceedat

may

is convergent
integral
^

T)\

{2x sinh

the limit

by making

B -^0,

e"

^-^

clT

Jm=o

convergent, the result of making

are

to

once

[Oh--

sinh Ty},
[(log

and, since the integrals

\~ K^

tt' {J,+ F,^{x)]


{cc)
1-

A,

0 is

-*

/iTo
(2^ sinh T) (e"2"^'

dT.
e^"'^)

J0

It is therefore

proved that, when

A 1^Ko

J,'(x)+ 17 (x)

If

we

of 2,

replacex by ^, both
providedthat R (z)"

have

we

cosh 2vTdT.

sides of this

equationbecome analyticfunctions
Hence, by the theoryof analytic
continuation,

0.

JJ"(2)+ 17 (2)
=

providedthat
Another

{z)"

"

Ko (2zsinh t)cosh 2vtdt,

0.

formula
integral

^i^

J, {z)

To prove
we

(2^'sinh T)

the result

(1)

(2)

"0,

which

=-^r
F, {z)^-^h^

this formula,we

replaceby x),and

be established

can

first suppose

by the

K,{2z sinh 0

that

is

method*

same

-'"'

is

dt.

variable (which
positive

then

/.(-)^^-F.(..)^4^
OV

OV

2ir''

^^^

j I ("

~.G
27r-i

-^l^r r
"

For
202.

"T^

the full details of the

00

"

dv

,'

f(

"

^^

dv

e'^ fcosh"-cosh")
tri)

^''"''
(2^^ + "^Oe^'"'"''"''

g-K'+")
.

(o/^^u)

e-^" ^-

g2.T ^^f/'^J

(f^^^t^)

00

e-2'*sinh
(2 f^ + TTl)

analysis, see

Watson,

Proc.

Tsinh

Royal

Soc.

xciv.

a,

(1918), pp. 197-

13*73]

INI'INITE

INTEGRALS

445

have
we
Now, T beingpositive,
Tniuh
e2'"8i"i"
irif]

(2 C/ + iri)exp [-2\{U+l
"? '

^ijj

e--i^sinh
rcosb^-^y
{v+ iri)
exp [- 2\ (v+ -rri)-]

and, since ^e-2.^sinhrcoshvjg


last

f^j^ function

.^^

of v, it may

be

proved that

the

of T such

that

is
integral
27ri

where

the constant

its integral
with

impliedin

e-2^sinh Tcosh

V(ly_^Q

symbol 0 (A.)is

the

respect to T from

0 to

(^x),
function

is convergent.

xi

In like manner,

{2U

e-2"
irif]

2 V exp

iri)
exp {- 2X ( f7 +

./

Hence

^i"'^ rsinh

^m

(- 2\v-)e---^ """'! ^"="^*i dv


"

0.

CO

it follows that

/"

ifo(2"sinh^)e-2''^c^r.
TT.'o

The

extension

complex with

the

to

in which

case

real part
positive

It should be mentioned
of

reader

may

find it

is made

that formula

of
descriptive
properties

The

the argument
as

to
interesting

functions is

(1).

(2) is of importancein

of Bessel

zeros

in

of the Bessel

the discussion

functions.

that

prove

CV

"(22-^2) J"(^)^^-n(.)^
(^) dv
Cv

and

hence

that

^LlJ""

(3) j; (z)

_.

^-^

r,' (z)

Cv

Other

(4)

formulae

JL
=

Ov

which

nz-

be established

may

"

/"(^2cosh

2T-

v^)Ku (2.rsinh T) e-'^"'^dT.

J (,

by

the methods

of this section

are

j^{z)j,{z)+r^{z)rAz).

\ / A'^_^(2ssinh0.{e^''+"^'cos(^-r)7r
+ t'-('^'')'}c/",
+

{^)

J^{z)i\{z)-J,{z)Y^{z):

sin

{ii
"

i^K
J

these

are

valid

when

published.
previously

R(z)"0

and

{2ziimht)e^''-^"^^dt]

\R (ijl-v)\"1; they

do

not

appear

to

have

been

446

OF

THEORY

Deductions

13"74.

-^o(f)is

Since

Nicholsons

from

XIII

integrals.

decreasing*function

[CHAP.

FUNCTIONS

BESSEL

of

" 13"73 (1) that

^, it is clear from

Jj'{x)^-Y,^{x)
is

decreasingfunction

this function

Since
shall

of

for

is

of v, when

when
approximatelyequal to 1l{Trx),

is

positive.

is

large,we

investigate
[JJ'{x) + 17 {x)]

and

real fixed value

any

it is

that

prove

increasingfunction

oi

decreasingfunction
when
v"\.

of

v"\,

when

and

that it is

an

It is clear that

^\J^{x)YHm
+

f"[K,
{2x sinh

-"

T)

2a;sinh TK^'

{^^ sinh T)}cosh2vTdT

K, (2x sinh T) tanh Tcosh

2vT

TT^

integratingthe

on

-J

dT,
cosh2z^r-^{tanhrcosh2

K,{2xsmhT)

second

in the

term

Hence

integralby parts.

^Jx{J.^{x)+Y.^(x)]]
r

Now

--,

integrandis
establishes
Next

2vT

increasingfunction of X
negative or positiveaccordingas

XT

\ tanh

Ko i2x sinh T) tanh Tcosh


is

an

{tanhT-2v
when
2y "

X
1

or

tanh

"

2vT] dT.

0, and
"

2i^ "

so

the

1 ; and

last
this

the result.

we

prove

that, when

x'^v'^0,

{x^--p^)^[J,^{w)+Yj^{x)]
is

increasingfunction

an

If

omit

we

under

the

of

x.

positivefactor

consideration

we

(x^
"

from
v^)~^Itt^

the derivate

of the

pression
ex-

get

sinh
[xK, {2x sinh t)+2 {x"- v"")
-

t
.

Ko {2x sinh 0} cosh 2ut dt,

Jo
to establish

and
is

the theorem

positive.
*

This

is obvious

from

the

formula

stated

it is sufficient

to

prove

that this

integral

13-74]

last

integrateby parts the

twice

We

447

INTEGRALS

INFINITE

the

portionof

second

in the

term

thus
integral

("^-xsinh t)cosh

sinh t /C

1v'

Ivt dt

sinh t sinh '2vtK^'{2x sinh t)

-VI

d
=

Jo
z=

"

"

-Iitdt

f^^

{sinh^AV (2a;sinh ^)|sinh

-y-

sinh 0} "inh
[sinhtlu (2a.'

"

Jo

2vtdt

tK,^{2x sinh i)sinh Ivtdt

2x sinh t cosh

cosh 2vt
sinh ^ cosh tK^^{2x sinh i()
0

.^"

[sinh^ cosh tK^

;^

sinh
:[^ cosh 2f 7^0(2a;

{2x sinh

^)]cosh

2vt dt

"^

2*= sinh " cosh-

tK^' (2x sinh 0] cosh 2vt dt:

Jo

step is producedby using the differential

the second

after
simplification
equation

the

^/lo"(z)+ K; (z) zK, {z)

The

2x sinh- 1 K^

[-

consequentlyreduces

discussion

under
integral

0.

{2x sinh t)

2./.sinlrtKo

"

to

2i^tdt

cosh
{2x sinhi)]

J 0

a-sinlr^

"

,x

Zvt

sinh t)
A'o(2a;

a;

1
rr
Ao(2a' snih t) cosh
,r.

"

cos)sh

"

2 sinh" t cosh 2vt +

Jo
X

==

and

this is

sinh t)[tanh^t cosh


A'o(2a;
the

positivebecause

v:

dt

2f^ +

'

"

^,

cosh 2vt

"

dt

(cosh i

2v sinh^ i sech ^ sinh

; hence
integrandis positive

2vt]dt,

the differential

coefficient of

(x)+ Y,'(x)]
(x^- v^)i{J^^
-

the result is established.

is positive,
and

the functions

Since the limits of both

(x)+
{x)\, {x' v^)i{J,'
(x)+ F,-^
{.7,^

a;

are

it follows
2/7r,

from

the last two

results that when

i'

|,

"

deduced

a;

(.^O}

n^.-^"^^H-)+lV(.^)"^.
{X- 1^2)1

(1)
An

}V

was
elementary proof of the last inequality(with various related inequalities)
formula
the
from
v.
(1906), \). 8(5,
by Schafheitlin,Berliner Sitzungsberichte,

(cf." 5-14)

'

where

%-^+

+
{x) aJ^ (,v)
'g'^
=

62)
-

.r

)\ {x).

[|^^ ''#;(-^O} (lI)'g';^(,^) r^y (,,)


+

+ 2

448

The

next

which

consequence

is that, when

is

shall deduce

we

d \\ (v)

written

:j

*-("')
"

dv

0-

"

"

dv

observe

this result,we

in the

of " 13*73
integrals

dJ^jv)

"

"

obtain

the

from

XIII

positive,
^"'(^)

To

[chap.

FUNCTIONS

BESSEL

OF

THEORY

that

the

expressionon

the

be

left may

form

dv

dv

-Y^{x)

J.{^)

dJ" (x)
dx

dx

TT

T)e-"'^dT

k\(2vsmh
TTV

e-'KJ'Ivsinh^^dt

1TTV

But,

for each

and so, since

value of t, 2vsmh{ltlv) is a decreasingfunction


positive
K^ix) is a positive
decreasingfunction of its argument, we see

1-1
is

decreasingfunction

of

e-' Ko
and

v,

v\J^{v)
r
^

v,

that

^\dt

sinh

therefore

l^{v)

,-

dv

dv

.dJ.{v)['

dY^jv)

lim

(2v

of

"

4
=

lim

0,

lOoTry-

by using the asymptoticexpansionsof " 8*42

; and

this establishes the result

stated.

The

13*75.

of J^ {")+ F/ {z).

asymptoticexpansion

the asymptoticexpansion of
It is easy to deduce
Nicholson's formula obtained in " 13"73, namely

/^^{z)-vYy^iz)

for

w^e

"

K^ (2z sinh 0 cosh

Y^'^z) from
t/^-(^)4

2vt dt

have, by " 7*4 (4),

.,-"
cosh2^
,nApm)
^^ ^.^^,",^
coshf

COS

where

and, when

\R^\"

i --i^
i
^

I'TT
o

cos

is real and p is

It

so

(2m)!

,"=o
m-

{vrr)

^^i^^^^^^^^2^sinh=n^
^'"^
^'
(2^)!

largethat

p +

^"

v,

Up

lies between

^^^^^2^i'sinh^^^

0 and

13-75, 13-8]
We

at

asymptoticexpansion

the

deduce

once

JJ^{z)+ 17 (z)
~

du,

-1)1*1^^;

3. ..(2m
be

(z)" 0, but it may

provedwhen

and, if v is real and

Iarg z\ Kir]

U--

Jo

{,-^ni)i

m=0

2 11.
r,'(.)~^

+
J.^(z)

is of the

after p terms

[ K, (2zu)

2-

by " 13-21 (8),

that is to say,

this is

"

i m^^
;^,
^

(1)

449

INTEGRALS

INFINITE

the wider

over

and j) exceeds
is positive,

signas,

same

extended

"

\,the

range
remainder

less than, the {p + l)th


numerically

and

term.

13 '8.

Ramanujans integrals.

have been obtained by Raman


ujan*from an
integrals
extraordinary
theorem f to Cauchy'swell-known
formula
of Fourier's integral
application
Some

which

is valid if i^ (^ +

where

"
i^)

The

1.

shews
application

formula,it is seen

In

of

then

applyingthis

fJ-^'-^e^'^i^
\t{ofe-i"fei")\l
"^-+'
[^!2
U.-t^+ylia)"'""
'''"""'"

t"

TT

cos

for other real values of t, the

is zero.
integral

particular

(2)

In vi6w

of the

"

J^+^{x)J,_^{x)d^ J,^,{2x).
=

'"

researches

of

March, Aim.

(1912),
Chemie, (4) xxxvil.
C/iemie,(4) XLi. (1913),pp. 191"208, it
Phi/sikund

der

50 and
Ann. der Physik ^md
Rybczynski,
character
quitelikely
that,in spiteof the erroneous
these integrals
evahiated by Ramanujan may prove to be
theoryof the transmission of Electric Waves.

pp. 29"
seems

QuarterlyJournal, xlviii.

Cf. Modern

B. F.

(1920),pp.

of the
of the

analysisof these writers |,


highestimportance in the

294"310.

Analysis, "" 9-7, 11-1.

t Cf. Love, Phil. Trans,

W.

y, and

and

that

ii "7r"

{\t\"'ir),

0,

is any real number.

By expanding in ascendingpowers

(1)

that

of the Royal Soc.

ccxv.

a,

(1915),pp. 123"124.
29

CHAPTER

XIV

MULTIPLE

connected

Problems

14" 1.

INTEGRALS

with

multipleintegrals.

subjectsof this chapter and the last is more


of mere
than one
degreeproduced by the insertion of an additional integralsign.
with the discussion of integrals
concerned
of perfectly
In Chapter Xlii we
were
of auxiliary
of the variable and of a number
definite functions
parameters; in
the functions
under
the integral
the integralswhich
to be discussed
are
now
arbitrary. Thus, in the first problem which
sign are to a greater or less extent
the
involves a function which has merely to satisfy
wall be discussed, the integral
of
conditions
of being a solution of a partialdifferential equation,and
having
differential coefficients at all pointsof real three-dimensional
continuous
space.
difference

The

the

between

of Fourier's
integral
subsequent problems,which are generalisations
restrictions
has to satisfy
even
more
formula, the arbitraryelement
general
such as having an
absolutelyconvergent integral,and having limited total
In

fluctuation.

infiniteintegrals.

Weber's

14" 2.

integralswhich

The

incidentally;but
because

be considered

now

formulae

of

in fact,discovered

were,

involve

Bessel

functions

only
somewhat
investigatethem
fully
Chapter Xiii may easilybe derived from
of the results of
as
by Weber
specialcases

desirable

seems

of the

many

them, and

it

will

to

this section.
researches*

Weber's
effect that

of the

solution

based

are

upon

result discovered

equation of Conduction
d^u

d-u

2X^/t,y+2Y^%z

d-u

du

by Fourierf to

the

of Heat

is
u

\r

^{a:

where

"l" is

Weber

an

arbitraryfunction

first

proved that, if

of its three

exp

2Z^/t)

{- (X-^+

Z')]clXdYdZ,

variables.

{x,y, z) is restricted

Y' +

to

be

solution

of the

equation
5'^

a-"i"

d-"^
,,.

Journal

t La
one
vm.

term

filr Math.

Theorie
on

(1809^

the
pp.

lxix.

Analytique
right had
235"244.

(1868),pp.
de

222

la Chaleur

previously been

"

237.

(Paris,1822), "
solved

372.

The

by Laplace, Journal

simpler equation
de

I'Ecole

with

only

iwlytechnique,

452
we
(f),

to

OF

THEORY

[CHAP. XIV

FUNCTIONS

vanishes because d^/d(j"


is supposedto
integral
position.The first term on the rightgives

that its

see

function

valued

BESSEL

of

be

one-

T-S

1
0

sni

clef),

^^77
ou

this

and

small
small by takingS sufficiently
arbitrarily

be made

can

since

o"l"/(sin^a^)
is continuous

therefore bounded.

and

f "rft3"(r)^
^l(-'-"^V^-"
d

small by taking8 sufficiently


small, and
arbitrarily

be made

can

is

therefore

it

zero.

Consequently

*i^

(3)

that

so

A sin kr + B

(r)

CT

i=,-"W

cos

0.
kr

and

where

all values of r, A
of

; since

constants

are

and

r^-O,

we

have

must

bt

(r)and its derivate

the

constant

same

are

continuous

for

values for all values

r.

If

make

we

that

see

0,

4"'Tr^{x,y,z)/k.

Hence*

2k
and

(1) of the

rdr

't"{x,y, z),
(" k'-t)

exp

result.

of /', then
00

(4)

jsin A;r

"

change to polarcoordinates shews that,if 4" {x,y, z) is a solution


continuous
type alreadyconsidered,and if /(?')is an arbitrary

similar

function

"

j0
^/{'Trt')

this establishes Weber's


A

of

exp (

)
00

Too

Too

.'

"

a"

"

ci"(Z,

F, Z)fy[{X

xf

(F- yy

(Z

dXdYdZ
zy]-\

CO

\ f (r)sm

kr
.

dr.

.0

The

reader

will have

concerningabsoluteness
*

This

integralis

most

respect

to

k.

of

easilyevaluated
I

with

in enunciatingsufficient conditions
difficulty
the various changes in the
to make
convergence

no

exp

17

the well-known
by differentiating

cos

kr

dr

J{irt)
exp ( K^t)
-

formula

14'3]

MULTIPLE

permissible.One
integrations
bounded

f{r)
where
A

such

the variables tend to

as

conditions

of

set

and
infinity,

0{r-v), (r^O);

453

INTEGRALS

f{r)

0{r-%

should

be

that

(r

-*

),

jtx 3, q"\.

simpler formula established at about the same


of the polarcoordinates (r,6) which
has
differential coefficients at all pointssuch that 0-^r-^a, whose
and which is a solution of the equation
if ?"

is that ^

somewhat

6) is a function
(/",

l^u

time

Weber*

by

continuous
value

at

is

first and
the

that,

second

origin is u^,

d^u
,"

then
y

when

0 ^

(1)

fF

J0

givenby
(R, 4") is

integral.

{R,^) J,[usl{R'+

Neumann

an

in his treatise f

with

respect to

(2)

cob{^

publishedin

the formula

reduces

of

F{r),

to

integral
J than (1).

Fourier's

(2) to functions of any order,namely


=

F(r),

Jo

by Hankel". In

modified
when

.'o

Jo

In this formula,

to

rudu rF(R)J, (uR) J, (ur)RdR

(3)

though a
value^^^f;

1862.

function is independentof "I",


we
arbitrary
tegratio
and then perform the ini-epeated
integral,

presents a closer resemblance

extension

effected

(^y^^Rid^dR)

the

rF(R)Jo (uR) Jo (ur)RdR

udu

Jo

The

-^Rr

tegratio
arbitraryfunction of the two variables {R, 4"),and the inthe planeof the polarcoordinates (R, ^) is a double integration.

in which
case
special
by
replacethe double integral

result which

r'

-77

In the

was

Stt i^uJq (^t),

of Neumanns

discussion

[^udu r

over

formula

Jo

was

(r,6)dd

proof of this is left to the reader.

General

14'3.
The

The

a.

u
77

"

form

of the

theorem

(""14'5

a
"\, (3) is actually

v=

that v^

this result it is apparently


necessary

case

"

\,

"

14'.52)is valid for all real

of Fourier's formula.

much
more
(2) and (3) are, naturally,
easy to prove than (1);
the proof of (3) is of precisely
the same
character as that of (2),the

The
and
"

formulae

Math.

Ann.

i.

(1869),pp.

t Allgemeine Lostvig

Korpers,welcher
151.

Cf.

von

zwei

Cf. Modern

iibcr

den

stationdren

einea

Tcmperatnrzustand

Kugelfldchenbegrenzt wird
Sitzungsberichte,xcv.
(2),(1887),pp. 409

"

410.

Analysis, "9"7.
vm.

(1875),pp.

476"

homogencn

(Halle, 1862),pp. 147

nichtconcottrischen

Gegenbauer, Wiener

" Matft. ^HH.

8"11.

de.i Problentes

483.
.

"

454

OF

THEORY

[CHAP. XIV

FUNCTIONS

BESSEL

arbitrariness of the order of the Bessel functions not

introducing
any

additional

complications.
(2) and (3) as
FollowingHankel, many writers* describe the integrals
"Fourier-Bessel
"Fourier integrals"
or
integi-als."
On

proving(1);

and

this

at

of the various

researches
As

give a proofof (3) before


of the
stage it is convenient to give a brief account
the formulae.
writers who have investigated

of its greater

account

stated, Hankel

alreadybeen

has

transformed

generalformula (3). He

CRF{R)dR

lim

Jo

k ^.co

the

was

first

writerfto give the

into
integral

the

J,{uR)J,{ur)udu

RF(R)[RJ,+,{\R)J,{\r)-rJ,+,{Xr)J,{\R)
K-

lim
A

then

shall

.0

and

we
simplicity,

.'0

oo

-*

"

appliedthe

second

theorem

value

,2

the

integrandjust as
the same
in the evaluation of Dirichlet's integrals.
Substantially
proofwas
laid
the
who
fact
that
the value
stress
on
important
given by Sheppardj
of the integral
which
depends only on that part of the i^-rangeof integration
is in the immediate
neighbourhoodof r, so that the value of the integralis
when R is not nearlyequalto r.
independentof the values which F(R) assumes
A

after

to

proof,based on the theoryof discontinuous integrals,


the formula ("13'42)
givenby Sonine",who integrated

different mode

has been

mean-

of

multiplication
by

from

(R) RclR,

0 to

so

"

r-'+i

f"I J,+i(ur)J, {uR)F(R)

.'0

RdRclu

Jo

to

as

get

fR''+^F(R)dR;
Jo

both sides with respectto r, formula (3)is at


then, by differentiating
obtained; but the whole of this procedureis difficult to justify.

and

once

proofof a more
directly
physicalcharacter has been given by Basset ||,
but, accordingto Gray and Mathews, it is open to various objections.
A

the

proofdependingon
by WeyllT.
The
*

extension

t A statement
Math.

dated
+

Ann.

vi.

Hankel's

QuarterlyJournal,

IIProc. Camb.
Functions

H Math.

of

memoir

xxiii.

Phil. Soc.

deducing (3)from

but

v.

this

lxvi.

is effected

is dated

(1886),pp.

(1909),p.

324.

(1) when

is

probably later than

was

(1889), pp.

(London, 1895),pp. 80"82.


Ann.

been

constructed

by replacingthe

cited later in this section.

mode

(1873),p. 149,

1872, while

's formula, which

of Hankel

See e.g. Orr's paper


of

equationshas
theoryof integral

an

integer was

Hankel's

made
by Weber,
researches, since it is

1869.

223"

42-3

"

'244.
133.

" JIath.
See

Gray

and

Ann.

(1880),p.

x\i.

Mathews,

Treatise

47.
on

Bessel

14-3]

455

INTEGRALS

MULTIPLE

obtained

Bessel functions

cylinderfunctions,was
by arbitrary
in
discussed
""14' 5 14"52.

it will be

and

by Weber*,

"

by Orrf to replacethe Bessel functions by any


functions, the w-path of integration
being a contour which avoids the
cylinder
used by him appear to be divergent,
of the integrals
so it is
*, but some
origin
attempt has been

An

made

criticism
The
his results are
correct.
same
say to what extent
appliesto the discussion of Weber's problem in Nielsen's treatise. It will be

difficult to

("14'5) that if,as Nielsen assumes,


of the
the integral
signare not necessarily
not, of necessity,
convergent.
It should

the

on
expressions

that, if

stated

be

rightin (2) and

l[F{r-0)

same

the

F{r

0)\,

in Fourier's theorem.

justas

the

For
Ueber

two

of F{R),
point of discontiimity
(3) must be replacedby:!:

be

cylinderfunctions under
is
type, the repeatedintegral

the

shewn

die

researches

recent

more

the reader

by Neumann,

should

consult

his treatise

Entwickelmigen
Kreis-, Kugel- unci Cylinder
-functionen
fortschreitenden

nach

1881).
(Leipzig,
Neumann's
formula

formula

involvingLegendre

objectin

that he

view

obtained

(1) was

functions ; in fact,it

obtained

the formula

lim
it does

but

by Mehler" as
of

was

of

limitingcase
apparentlywith
a

this

" 5"7l,

P,, [cos{zin)] /" {2),


=

a
rigorousproof on these lines
easy to construct
direct method
of proof is given in a difficult memoir

not

seem

(cf.

by
" 14-64).A more
Fourier's
the generaltheoryof integrals
Du Bois Reymond|| on
resembling
which
shall
The
we
(""14'6 et seq.)is based
givesubsequently
proof
integral.
these researches.

on

pointedout that the


Bois Reymond which

SubsequentlyErmakoffH
from a result obtained by Du

variables of Fourier's theorem

two

J_
=

Ermakoff

"

[" p j
deduced

for

formula

is also derivable

is the direct extension

variable,namely

one

yir(X,Y) cos [a{X

x) -\-i3{r- i/)]{dXd F) dad/S.

by changing to polarcoordinates by

the formula

to

of

means

the substitution
a

and

cos

(3

0),

with respect to
the integration
effecting

Math.

Proc.

J The

Ann.

vi.

Cailler,Archives

of

xxvii.

the iutegral at
des

" Math.

Ann.

v.

IIMath.

Ann.

iv.

co,

co.

(1873),pp. 146"161.

Boyal Irish Acad,


value

sin

Sci.

a,

(1909j,pp.

point

of

205"2-18.

{Soc.Helvetiqne), (4)xiv.

(1872),pp.
(1871),pp.

been

discontinuity has
(1902), pp.

examined

with

some

care

350.

B47"

135"137.
3G2"

390.

H Math.

Ann.

v.

(1872),pp. 639"

G40.

by

456

OF

THEORY

BESSEL

[CHAP. XIV

FUNCTIONS

be the polarcoordinates corresponding


to the Cartesian
(?',
"^)and (R, "i")
coordinates (x,y) and {X, Y) respectively,
the formal result is fairly
obvious
^ (X, Y) by F{R, ^) ; but the investigation
when we I'eplace
by this method
If

the
be

since it
difficulties,

without

is not

to be

seems

infinite
taken over
an
repeatedintegral
taken over
replacedby a repeatedintegral

by no means
in
rectangle
the

the (a,/3)plane may

of

area

that

easy to prove

indefinitely
great

an

circle.

which

function F {R, ^) is not continuous, the factor F {r,(f")


arbitrary
the
occurs
on
rightin (1) must be replacedby the limit of the mean

value

of

If the

This

F(R, 4") on

what

givenin ""146

he knows

formula

of the

which

is

"

Fourier

theoryof

more

reader

The

14"63.

recondite

centre

than

this
might anticipate

has been

examined

HankeVs

14'4.
The
Hankel
stated

by Bateman,

8^0.

the

proof

result from

series.

(3),namely

F{R)dudR
J.iu-r)-^^^^^
I \IJ^^

(4)

when
(r,cf))

at

in his treatise of 1881, and

in effect,
provedby Neumann

was,

will be

8 with

circle of radius

Proc. London

Math.

F{r\

Soc. (2) iv. (1906),p. 484

cf.

"

12'2.

repeatedintegral.

of Neumann's
generalisation

effected

(c" " 14"3)in

be

follows

as

Let F

(R)

the

case

of

integralformula which was


functions of a single
variable,may

by
formally

be

an

functionof
arbitrary

the real variable

subjectto

the

condition that

rF{R)s/R.dR
Jo
is

absolutely
convergent; and
less than
|. Then

eocistsand
be not*

let the order

of the

0) + F(r-0)},

Bessel

functions

"

TOO

rx

\ F(R)J,(uR)J,{u7-)RdR

udu

(1)
J0

l{F(r

.'0

providedthat

the

number
positive

lies inside

an

interval in which

F{R)

has

limited total fluctuation.


Hankel's proof,and it
proofwhich we shall now giveis substantially
is of the same
generalcharacter as the proof of Fourier's theorem ; it will be
the proofof Fourier's theorem givenin Modern
set out in the same
manner
as
of lemmas.
Analysis,Chapter ix. It is first convenient to prove a number
The

more

It

seems

extended

not

unlikelythat

range

of values

it is sufficient for
of

i"

would

be

more

to be

greater than

difficult.

1 ; but

the

proof for

the

14-4, 14-41]

MULTIPLE

The

14*41.

analogueof the Riemann-Lehesguelemma.

A result,
which
of Fourier
theorem

resembles

follows

as

the lemma

which

series,and

is

of

Riemann-Lebesgue* in the theory


requiredin the proof of Hankel's integral

is

Let-fI F(R)\/R.dR
absolutely
convergent ;

exist,and

and

let v^

(ifit

is

\. Then,

"

F{R) J, (XR) RdR


It is convenient
assumed

that

divide the

to

F(R) \/R is bounded, and


is bounded

F(R)\/R

let
improper integral)

an

A. -^

"s

it he

cc
,

(1/^/X).

proofinto three parts ; in the first part it is


that b is finite ; in the second

the restriction that b is finite is removed


that

457

INTEGRALS

part

and in the third part the restriction

is also removed.

tegratio
(I) Let the upper bound of \F{R)\/R\ be K. Divide the range of in(o, b) into n equal intervals by the pointscc^, Wo,
Xn-i {^o= ^,
...

Xn

b),and choose

7i

largethat

so

111

where
and

is

small
arbitrarily

an

lower bounds
Write

of

F{R) sJR

number
positive

FiRJ^/'R in the mih


=

F(R,"_^)\/R"i-i+

and

17,^and L^

are

the upper

interval.

(R),so that,when

Wm

interval,|"," (R)\-^U,,^ L,". Now, when

v^-h,

"

both

R liesin the mth

of the functions

of a;,

rx

x^J^{x),

tiJ,(t)dt,
.'o

bounded

are

Let A

when

B be the upper

and

is not convergent
though the integral

0, even

bounds

of the moduli

of these functions.

as

-*

oo
.

It is then

clear that

F{R)J,{\R)RdR\
2
m

F(R,,_,)s/R,,_,
l

/, (XR) ^R

,"

2BnK

Cf. Modern

t The
R

preserves

upper
the

'"

oy,,,{R)J,{XR)sJR.dR

.'Xm-l

Ae

'*

A-

2B

dR
.

Xm-l

\'A.

Ae

Analysis," 9-41.
limit of the

analogy

with

integral may
" 1-1-3 (3).

be infinite

and

"^0.

The

apparentlyirrelevant

factor

458

THEORY

OF

BESSEL

[chap. XIV

FUNCTIONS

large(n remainingfixed after e


By takingX sufficiently
made
and so
be
less than 2Ae/'\/\,
last expression
can
is 0 (1/a/X).
limit is infinite,
choose

If the upper

(II)

has been
the

chosen) the

original
integral

that

so

^\F(R)\^/R.clR"6,
J

and

inequality

the

use

F(R)J,{\R)RdR
then, proceedingas in

F(R)J,(\R)RdR'

"

(I),we

case

choice of

The
mode

A_

F{R)\^R.dR;

get

2BhJl

F{R)J,(\R)RdR
I J

2Ae

depends on

now

of subdivision

of the range
stillindependentof \, and so we

of
can

through the choice of c as well as by the


(a,c); but the choice of n is
integration
infer that the integral
(with upper limit

is stillo{l/\/\).
infinite)

(Ill)

If

F{R) \/R

in

unbounded

is unbounded*,

may
8
intervals
of
such
p

number

enclose the

we

pointsat

which

it is

that

t\ \F{R)\s/R.dR"e.
s

Js

By applyingthe arguments

of

intervals,we

the

(I)and (II)to

parts of (a,b) outside

these

get
RdR
1'f(R)J.(XR)
3^,^
I 25-^(4"iM
I
+

"

A,*

Ja
K

where

is

now

choices of both
that

we

can

14'42.
We

the upper bound


of [F
K and w now
depend on

stillinfer that the

The

inversion

shall next

(R) \\/R
e,

but

outside

the intervals S. The

still independentof X,

are

that, when

repeatedintegral.
v^

\, and

"

F{R)\'R.dR

exists and

Jo

is

absolutely
convergent,then

ndu\ F{R)J,{uR)J,(ur)RdR
0

Jo

k-^aa

providedthat

the limit

on

F(R)]

lim

the
*

so

is o (l/VX).
integral

of Hankel's

prove

VA,

J 0

")uduy RdR,
J,{iiR)J,{ur)

[Jo

rightexists.

Cf. Modern

Aiudysis, "9-41.

460

OF

THEORY

BESSEL

effect this, it is sufficient to

To

( F{R)

f udu
./0

the limits

the

on

IF(R)

udu

is not

the

point of

J, (uR) J, (ur)RdR

0.

invert the order of the

JO

that, if

prove

(a,h),then

interval*

We

[chap. XIV

FUNCTIONS

in " 14'42, and


as
integrations,

we

find that, if

rightexist,

J, {uR) /. (ur)RdR

\l\j,.
(uR)
I^F{R)

lini

K^:r.

udiilRdR
,/,(la-)

(,.'o

"

.1

"b

X.RdR

lim

F{R) [RJ.+,(7^R)J. (Xr)

Km

-ti-

Jn

the

r'

r-

"

C
XrJ^^,(X7')

lim

both

R^

f'~^^-^J,+, (XR) dR

A./,(Xr)

x^x

Since

rJ,+,(Xr)J, (XR)]

^J;f^^^JAXR)

integrals
r^

Ri

F(R)
F(U)
R'-

Ja

f^ F(R)R

dR,

r^

dR

R'-r'

a-

Riemannabsolutely
exhypothesi
convergent, it follows from the generalised
Lebesgue lemma (" 14'41) that the last two limits are zero; and so

are

du

(R) /, (uR) ,/,(ur)RdR^O

.' 0

providedthat

The

14"44.

is not such

that

be shewn

now

uf I

boundedness

It will

that, as
h

^b.

J^(uR) J^(a7^)uR^dudR.
J 0

\^~

the

cc

repeatedintegral

TA

Jt,(u R) Jt,(ur)uR
remains
is
to

bounded, providedthat

permissible
for a
r

as

Let
of the

\.

and

b have any

values.
(bounded)positive

of which

functionsofX

one

firstconsider the

dominant

terms

rA

(uR

cos

"

hi'TT iir)cos

(ur

"

^vir

"

du dR
4-7r)

"

TT'^rJaJo
1
TT

f''[s\nX(R"r)

\/rj
1

cos

[X(R

R-r

\/'r

AC'-)-) sin
X{a-r)

r)

"

R +
A(Hr)

cos

dx
TT

It

tend

(or both)may

integralobtained by takingthe
asymptoticexpansions,
namel}'
us

rb

00

b to be

and

dudR

J 0

Kia
It is

r)

permissiblefor

(X

vtt]
"

cos

TTT

dR

vtt)

dx +

^'

li to be infinite.

cos

vir

b+

log
a

14-44]

461

INTEGRALS

MULTIPLE

sin

is bounded
first integral

The

a?

because

dx is convergent ; and

the

\^

is bounded
integral

second

Cos

because

\'(jC

dx

under

now
integral

convergent;

\%

and

so

the

VIT)

"

is bounded, and

consideration

its limit,as

-*

is

x
,

the limit of

_1

dx

providedthat
But
h

we

+ cos

log

VK

A(f(-)-)

this limit exists.


write

may

,-A

J^ (uR) Jv (ur) uR- dudR


a

J 0

TT

('00

[WuJAnR)-h{in-)^!{Rr)

'\/rJ
a

{uR

cos

"

\viT jtt)cos {ur


"

"

\v7r jTr)]dudR

"

"

""

(uR) J, (wr)^J(Rr)
ll-rruJ,

Now,
R

of

an

T-

"

(i^i^ |i'7r Itt)COS

COS

of R and r, and

of

when

is

it is

so

and
positive

has
integral

third

("r

"

"

^vrr jtt)]dudR
"

"

i/TT" jtt)c?uc?ii.

"

the right,
the
on
integrals
(with respect to it)which
integral

function
The

\v7r jTt)cos {ur

"

first is the

of the

of values

domain

\/r J " J

TT

(uR

cos

"

been

converges
continuous

with respect to
integi-al
uniformlyin any positive

(and therefore bounded)

bounded.

shewn

be bounded, and

to

it converges

to

limit whenever
"Hb-r)

sin

da
X

\{a-r)

does

so.

The

second
4"v-

integralmay
-

be written

'' 00

sin

y^TT \/r

in the form

uR

{uR

"

|i/7r Jtt)cos {ur


"

co?,{uR \vTr jtt)sin {ur

-\

"

4z;2-l

[''

"

where
to zero

-^

^vir \it)"{"0{llu-)
"

dudR

functions

00

of X and
"

"

(X),(po(X) and ^3 (X) are


(j)^
as

\v7r ^ir)

lcfy,{\)
cP,{X) dR,
^cf,^{X)
+

47r \/r J

"

"

7^ which

tend

uniformly

462

OF

THEORY

rb

[CHAP. XIV

FUNCTIONS

values
positive

for all bounded

Hence,

BESSEL

of a,

r\

h, r, the integral

J^ (uR) Jv (ur)uR^ diidR


Jn
is bounded

A.

as

it converges

; and

oo

^-

J0

r\{b-r)

to

sin

limit whenever

a;

ax
^

J k(a-r)

does

so.

Proof of HanleV

14"45.

Since F

has limited fluctuation in

(R)

point,so

also has

F{R) \/R

where

(-^)^^^

%2

Xi

small that

so

; and

been

proved,the

monotonic

(-^)^^^

interval of which

an

therefore

number
choosinga positive

After

have

actual

proof

is

theorem

of Hankel's

theorem.
integral

lemmas
preliminary^
quitesimple.

that all the

Now

we

is

internal

an

write

may

functions.
(positive)
increasing
choose
we
arbitrarily,

has limited total fluctuation

F(R)

number
positive

in the interval (r

B,r

"

8)

also

and

(r+ 8)

%i

%i

(r + 0) "

X2{r+8)-X2{r
If

second

apply the

we

e-|
0)"e]'

xi

;^, (r

mean-value

/i

0)

0)

theorem,

| intermediate in value between

number

0'

0 and

8) " e]
^.C/'- 8) " ef

%i

(r

'

find that there exists

we

S such

that

dudR

(R) J^ (iiR) J^ (ur)u"JR.

\ J^ ("^) J^ ("0

(^'+ ^")\

Xi

[Xii^'+ S)

%i (^*+

0)}r
J r+f

t"

Since

Jo
as

X,

-^

00

sin

of the

of the modulus

as

-*

exceed

go

I J. (uR) J, (ur)

\/R dudR.

dx

-^

^-tt,

while

from

" 14'44

8 remains

that the first term

fixed.

Ce in absolute value,where

And

the second

C is the upper

(cf." 14-44),
repeatedintegral

Hence, if

J^(uR)J^(ur)uR^dudR C\/^r,

Km

it follows that
lim
A. -".

exists and

is

00

r
.'r

remainingfixed,it follows

righttends to a limit
the rightdoes not
on

dudR

^R

cc

"

1 x^(R)JAi(R)J.(ur)uR'^dudR
Jo

equalto C^Xi(^ + 0)l\/r.

on

the
term

bound

14-45]
We

MULTIPLE

x^i-^)^'^^

treat

lim

then
exists and

is

I J^{uR)

apply similar reasoningto

./,(ur)uR^ duclR

C.,/\/7;

F{R) J,{aR)J^{ur)uRdudR

equalto
C,F{r

We

also

infer that, if

we

"

lim

and

similar manner,

the interval. (;" h, r) ; and

463

INTEGRALS

have

now

Ci and

evaluate

to

0) + C,F{r-0).
C.,.

the

By

theory of generalised

have

we
integrals*,
ri

rr+S

rcc

JAyR)Ju{iir)nR^dRdu

Jo Jr

\/r

fr+S
=

lim

lim

dRdu
Jy(uR)J^,(ur)uR'-^

I
exp (" |)-it^)
rr+S

/'x

I "exp (" jj-U-)J^ (uR) J^ (ur)uR^

p^O

dudR

.' 0

by "13-31(1).
of

Now, throughoutthe range


/.
"

^^'\

as

j9

^!l
i?r)
v/(7r

2psJ{Rr)

6'i lim
=

"

"

r*^"'
exp

^^'
0/)]exp(|
\2if)
/.V2M
^^

.-...,.

^^

-t-

'

"r-

exp

-^ "

X-) dx

"

,"

;H^}"dR
\,

similarly
1

Co

lim

exists and

lim

therefore shewn

We^iave

is

/""

Co,

see

dx
^"^)

^.

that

fV

./,(ur)uRdudR
(i?)./,(ui^)

equalto

Hardy, Quarterly Journal,

and

exp ("

"f"

^\F(r+0)

C,

4"p' J

Km

+,

0.

Hence

and

n
'

V2/V

j,

integration,

"14-52.

xxxv.

(1901),pp.

F(r-0)}.
2-2

"

66.

For

difiereut method

of

calculating

464

THEORY

OF

BESSEL

But, if this limit exists,then, by

udu

Jo

also exists and


stated in

is

[CHAP. XIV

FUNCTIONS

" 14"42,

F{R) J, (uR) J, {ur)RdR

JO

it ; and

equal to

so

have

we

proved Hankel's

theorem,

as

" 14"4.

to be due
seems
generalisedintegralsin the proof of the theorem
to
of such methods
Sommerfeld, in his Kouigsberg Dissertation,1891. For some
applications
with the generalresults of this chapter to the proUente des moments
combined
of Stieltjes,
81
recent
88.
XLVii.
(1918),
Hardy,
a
by
see
Messenger,
paper
pp.

The

of

use

"

Note

14'46.

proofof his

HankeVs

on

proof given by Hankel


inadequately. The first is in the

of his formula

The

lim

which

he

theorem.
to

seems

discuss

two

points somewhat

of

discussion

/ F{R)J^{uR)J^{ur)uI{dudR,

replacesby

\^x,

li--r-

he substitutes the first terms of the asymi^totic


integral,
consideringwhether the integrals
expansionsof the Bessel
arisingfrom
a fatal objection
to the proof),
the second and followingterms
(which seems
are
negligible
oi\R
the consequences
and without considering
vanishingat the lower limit of the path of
In order

this

approximate to

to

without

functions

integration.
is of

point,which

second

The

by
pi'oving

and

is

if ^

this does

not

the method

0, he takes

shall

discuss

now

the order
The

number.

lower

since it is convenient
definiteness

For

The

and

of

are

theorem

of
integrals

du

this is
must

be

zero

if

^ tends

bounded

if

to

^-"-0 as

positivelimit
X-^x

to

any

and

cylinder
functions.

the type

uR) "W,(Mr)R dR,

F{R)^,(

of the unrestricted

limits of the

givethem

to

we-

is any real*
cylinderfunction '^^,(2')
will be specified
integrals
subsequently,
values which depend on the value of v.

shall suppose
=

cr

justexplained that
for granted that it

^,.(z)
where

I J^ (uR) Jy (ur)uRdudR

J r+f

of HankeVs

in which

character,is in the discussion

prima facieobvious.

seem

14*5. Extensions
We

it

similar

lim
A^*

after

that

+ s\n(i.Y^ (z)],
{coscc.J^,(z)

constants.

subsequent discussion

is

simplifiedand

no

generalityis

lost

by assuming

that

v^O.

14-46-14-51]
The

analogueof

the

465

INTEGRALS

MULTIPLE

("14-41),namely that

lemma
Riemann-Lebesgue

[ F(R)^R.dR

providedthat

Ja

the
be provedby precisely
absolutely
conver^gent,
may obviously
of " 14-41,providedthat a ": 6 " oo
and

exists and
methods

is

^a ^

\a"0
The

theorem

theorem

is that the

1,

if v"h
in
slightly

has to be modified

" 14-44

of

if 0 ^

form.

modified

The

repeatedintegral
"b

r^

(uR)'^,{ur)u^/R,dudR
a

is bounded

while

X-^cc

as

functions of \ which

finite limits

have

for
though it is permissible
Also the

it to be

and
repeatedintegral

X,

as

[Note.

both

or

If the two

is

positive,

O^t-^^,

oscillate as A,

qo

in the

cylinderfunctions
the integral

repeated integralwere

not

of the

same

considered

type, i.e.if we

'

f -^^(uR) ^^ (wr)

?t

it would

dx

converge

number

integral

J \(a-r)

both

The

oc

-^

when

zero

the

be

fixed;as in "14-44,a and h may

remains

be found

s/R

du
.

dR,

that the convergence

of this

integralnecessitates

the convergence

of the

integral
fK(h-r)

-cos.r

dx;

/
and

so, if X

(a

14-51.

"

r)-^-0as

X-^

The extension

the
Retaining

the
,

is divergent*.]
repeatedintegral

theorem

of Hankel's

notation of

""14-4

"

when

14-5,we

0 "^v

shall

%^.

now

prove the

following

theorem.

Let^j F(R)
letO^v^l.

l^

(1)

This

du

point

B. F.

be

and
convergent integral,
absolutely

an

'J^hen

was

F{R) 9^,(uR) ^^,(ur)RdR

overlooked

(Leipzig,
1904),p. 365,
w.

exist and

\/R .dR

in his

by Nielsen, Handbuch

expositionof Hankel's

der

Tlworie

der

Cylinder
Junhtionen

theorem.
30

466

THEORY

OF

BESSEL

number
providedthat the positive
limited total jiuctuation.
As

in

I
Jo0

" 14"42, we

shew

may

lies inside

interval in which

an

F(R)

has

that

F{R)9^, (uR) "W,{ur)R

udu
du
u

[chap. XIV

FUNCTIONS

dR

Jo

the limit

providedthat
But

now

we

observe

-""

the

on

uR
\^F{R)\^'i^,
(uR) "W,iiir)
i

lim

JO

du dR,

rightexists.

that

'^,{uR)''^,{ur)udu
1

uR
R'

'^^,+1
{uR) "W,(m-)

ur

r'

%\+, (ur)"W,(uR)

2a" sin

TT

Hence

sin (a +
sin

r^"

R''r''(R^ -r')'

vir

infer that, if r is not

we

R^"

vtt)

pointof

the interval

{a,h),then

(R)

f '^^(uR) "",{iir)uRdudR
sin

TT

as

00

-^

Now
that F

; and

choose

(R) has

the last

so

has
repeatedintegral

arbitrary
positivenumber

an

F{R)

take

divide the

{R)

the second

e,

limit when

and

(r

if r"R^

0) I"

if r-8^R"r.

X.

oo
.

choose

B, r

"

-^

small

so

8)and

so

that

8,

^ 'i^,{uR)'^,{ur)ududR,

F{R)

Jo

8),(r

then

0)\"e

into
i?-pathof integration

(0, r

Apply

F{r

Jo
and

limited total fluctuation in the interval

F(R) -F{r

Now

J" R''r''(R^-r^)

y-TT

four parts,namely

8, r), (/",
r +
8),(r + S, x ).

mean-value

theorem

in

as

" 14'45,and

find that

we

IK (iiR)"^^(wr)uRdudR
2cr2 gjj^Qj gjjj^Qj
_^ j^^^ fJ'*'"'*
TTSm

i'o

1^73-

r+S

i^(r+ 0)Vr.

F(r

0) Vr

v,

R""
r'
I
F(R)dR
]r+slR''-'r''(R^-7-)
"v

r\

'"',{uR)%{ur)uR^dudR

J r-sJ
+

"^,(uR) ^, (ur)uRhladR

468

OF

THEORY

these

and

BESSEL

[chap. XIV

FUNCTIONS

all

Weber's

evaluated

be

absolutely
convergent. They may
of " 13"4, and hence
discontinuous
integral
are

as

cases

of

find that

we

'

'

^*

K+, (ur)]^, {ur)

V^^

'^.+z(uh)
{6"+^

(a + pir) sin (a + vtt).V {}"-\-\"


p)
r (i/
22p sin pTT sin fTT
+ l)r(|0+ 1)

o--?-" sin

.^,{v^\-

62P

sin (a + pir + vtt) F (1


p)
2-f sin (pTT+ vk) sin i/tt r (1
+ p +
I/)r (i/
"

a-r

sin

iFiil-p,

]f.v+2p
limit of this

when
expression,

(r-r"

sin

vtt)

TT

algebra;
simply\a-r^.
some

In like

and

so

have

we

du

^TTcot

and

the

be shewn

vr

Jttcot (a + i/tt) -^{\) -\-y^{" v) \,

"

"

last

when
expression,

-*-/"

0, is

proved that

F(R)^,(uR)^,(ur)RdR
a

TT

providedthat

14*52.

sin

generaltheorem

VTT

iv

F{R)dR,

R''-'r''{R^-r'-)

sin ccY^(z)]
;

stated at the

beginningof

the section.

]Vehe7-'sintegral
theorem.

It is evident

from

" 14-.51 that,if

1
J

convergent, where
lim

R'"

vtt)

[cosaJ^ (z)+

a-

this is the

sin (a +

and
":^,F {B.)\%subjectto the conditions stated in " 14-4',

0 ^v

*^"{z)

(1)

to

that

la^-{F(r+ 0) + F{r-0)]

and

limit of the

2a- sin
=

it may

manner

1)

log(l-2)
h-J 21og^-:^,i^(l,-.:l
V

t'TT

after

^J

0, is reducible

-^

sin

l-v;

p\

-V-

The

i)r(2p)-

"

"

sin (a

r(.

i.^V,.

p.,

"

fluctuation and

r^,^,(Xr)"^,(Xi^)]

+ 0) + i^(r-0)},
|(7={i^(r

lies inside

F{R)

absolutely

^^,

F{R) [RW.^, (\R) ^^ (Xr)

is

0, then

provided that

exists and

F(R) 'JRdR

interval

an

is defined

functions lies between


cylinder

to
"

be

m.

zero

and

\.

which

when

$i^

{R) has limited


"

a,

total

if the order of the

14-52]
We

shall

469

INTEGRALS

MULTIPLE

establish the truth of this formula

now

cylinderfunctions

for

of

unrestricted order.

It is

r%^, (Xr) "",(XR)]

[RW,^, (\R) W, (Xr)

Let

^.{R,r; \)-"P,_,(R,

by

and so,

["[$, (R,

lim

(2)

\)

r;

analogueof

the

the

from

easy deduction

an

formulae

recurrence

"j"^
(^^ ,.

^y

that

''S',.,(X7-)%M\R)],
j^S'^f.-dXR)%M\7^)
+

the

^A^

(" 14-41),we

Riemann-Lebesgue lemma
X)

"P,_,(R, r

dR

X)]RF(R)

have

0.

of this result,
Hence, by adding up repetitions

where

hm

(3)
7i

lim
A-".oo

that

so

of the

one

X)

^,^n{R,r;

lim

K^^x

This result is
indicated in
To

dR

X)]RF(R)

0,

v "
integers

liesbetween

RF{R)dR

|,and

la' {F{r + 0) +

then from (1)

(r

0)],

and so, for all real values of

(4)

(R,
"t","n

integer.
positive

is any

Choose

{R,r; X)
["I",

deduce

v, we

from

X)RF{R)dR

^,{R,r;

(3) that

la^{F{r + 0) + F(r-0)\.

due
practically

to

Weber*, and it

obtained

was

by

the method

" 14'46.

obtain

form, let

the result in Weber's

P,(,) F, (r)/, (z) ./.(r)F. (z),


l"^.
(^) Y. (R) J. (s) J. {R) F, (z).
-

^"'^

Then
_

'W,(uR)

XI

"7

-^^(ur)-^^

and
u

the

(R'

W, (uR) u du,
?"-) W, (U7-)

du

au

the left is also

on
expression

equalto

[iiC+ii'lR)% (wr) r%+, {ur)W,{uR)]


-

uR[Y, (R) /.+!("E)


ur[Y^ (r)J^^,(ur)

J. (R) F,+i(iiR)]
[F, (r)J,(ur)
J, (r)F,^,(ur)][F, (R) ./,(uR)

uY, (R) Y, (r)[RJ.+, (uR) J, (ur)

J, (r)F, (ur)]
J^ (R) F..(uR)]

rJ,+,(ur)J, (uR)\

(R) F.,(r) ./,(r)F, (R)][RJ,+, (uR) F, (ur) R Y,+,(uR) J, (ur)


+.^i {./,
(ur)Y^ {uR)]
(ur)J^ (uR) + r./^+i
?'Fi,+i
-\u[J, (R) F, (r)+ J, (r)F, (R)][RB,+^ (uR) D, {ur) rD,+,{ur)D, (uR)
-

"

RD,+, (uR) D, {ur) + rD,^,(ur) D, (uR)]

rY,+,(ur)F, (uR)],
nJ, (R) J, (r)[RY,+,(uR) YAkv)
p. (.) J. (.)+ F. (.),
-

where

\DAz)
*

Math.

Ann.

vi.

J.(^)-YA^)(187H), pp. 14(3"161.

470

OF

THEORY

Now

suppose

BESSEL

[CHAP. XIV

FUNCTIONS

that
'^

f(R)BdR
is

exists and

consider

; and

absolutely
convergent

\iml'^f{R)\
'^,(ur)%\(uR)uRcludR.
with respect to
integration

the

Carry out
by the sum

replacethe integrated
part
above, divided by R^
r^.

written

of the four terms

u, and

"

J,{R)Y,(r)-J^(r)l\(R)

^.,

si-

is hounded

r, and

near

containing
r,
tend
terms
of

limited

to

use

we-

\-^y:

as

zero

total

the third group

(1),we

of terms

get

we

which
pairof integrals

left with the result that


ther^efore

are

r f {R) [^
'^^(ur)^, (uR) uR

lim

r-

in any hounded interval


fiuctuation
the integralscorresponding
to the second
group
,hy the generalised
Riemann-Lehesguelemma.

it followsthat

Correspondingto
happen to cancel.
When

has

"

du dR
.

i l^.HO + IV(r)} {/(r + 0) +/(r

0)},

that is to say

(6)

udu \"
l'^
f{R)%\{ur)%\{uR)RdR
=

i {/.M'O+ F.Hr)} {/(r+ 0) +/(r


.

functions are defined by (5),and


cylinder
/(-R)has limited total fluctuation.
the

in which
in which

of functions

14"6.

lies inside

details of notation,this is the result obtained

Apart from
case

Formal

of

an

by Weber

0)1,

interval

in the

order.
integral

of Neumanns

statement

theorem.
integral

the theorem
which will be the subjectof
precisely
It is convenient to enunciate
discussion in the sections immediatelyfollowing.
obtained by replacing
with Du Bois Reymond's* generalisation,
the theorem
which
satisfies
function
certain
function
the Bessel
by any
generalconditions.
is as follows :
theorem
The generalised
We

shall

(I)

Let

now

state

'^{X, Y)

vaj'iables {X,

Y), which

he

hounded

arhitrary
functionof
integral

iX, Y) {X' + rO"'

exists and
*

2Iath.

is
Ann.

the

pair of real

is such that the douhle

"

(dXdY)

ahsolutely
convergent.
iv.

g{t)=Jo{t),and the
published in 1881.

(1871),pp.
conditions

383

"

390.

Neumann's

(I)"(III)are

formula

(cf." 14-3)is obtained

those
substantially

given in Neumann's

by writing
treatise

14-6, 14-81]

471

INTEGRALS

MULTIPLE

"^'(X,F) is expressedin terms of polar coordinates,let it he


have the property tJiat(for all values
and let F {R, "I")
denoted hy F (R, "I"),
in
of ^ bettveen " tt),F(B, "I"),
qua functionof R, has limited total fluctuation
(II) When

the interval (0, oo

and
let tJiisfluctuation

); and

also F (+ 0, "I")
be

integrable

functionsof^.
in the interval
of F (R, "I")
(III) If Q{R, ^) denote the total fluctuation
with resjjectto "l" as R^^O,
unifoj'm.hj
{" 0, R), let fl (R, 'i')tend to zero
with the exception*
the whole of the interval (" tt, tt),
of values of "P
throughout
in a number
of tuhose angles may be assumed arbitrarily
of sectors the sum
small.
this condition necessitates that
^n(R, "t"),
{+ 0, (i")\
sectors.
F (R; ^")^- F (+ 0, "I")
uniformlyexcept in the exceptioncd
Since

\F(R, ^)

variable R, such
(IV) Let g{R) be a continuous functionof the positive
and ivhen R
cc
that g {R) \/R is bounded both ivhen R-^0
-^

f-K
=

.'o

dt

f"

q{t)tdt G{R),

Let

and

G{t)" he convergent.

let

-'o

'

I" ["["

^^

udu

Then

is convergent,and

is

(Z, F)

\u^(X'^+ F'^)}(dXdY)
.

equalto
.'0

'Where

means^
{+ 0, "!")

^F

^
Ztt

Before

justas

proving the

in the

14*61.

i-

case

main

Fi+o,^)d^.

-"

theorem,

of Hankel's

we

shall prove

number

of Lemmas,

integral.

analogueof the Riemann-Lebesguelemma.

The

Correspondingto the result of " 14-41, we have


unbounded
an
of
domain\ surroundingthe origin,
\
cc
interior jwintor a boundarypoint,
then, as

the theorem
which

the

that

if T

originis

not

is
an

(dRd^)
o(l).
'j^FiR,ct")GiXR)^-^^
=

The

that the reasoning is applicableto the case


(which
objectof the exception is to ensure
bounded
is
one
outride
in
which
"if
by
a
region
zero
(A',Y)
physicalimportance)
the
the
the
on
inside
and
constant
region,
originbeing
is, say, a positive
analytic curves

is of considerable
or

more

boundary
t The
of

F(R,

die nach

of the

region.
value
expressioninvolvingthe mean
Ueber
made
Neumann,
was
by
a point
F(B, 4")
Entivickelungen (Leipzig,1881),
Cylinder-functionenforUchreitenden

discovery that the repeated integralis equal to


when
"f")

the

origin is

Kreis-, Kugpl- und

of

an

of
discontinuity

pp. 130"131.

X For

origin.

instance

might

be the whole

of the

plane outsile

circle of radius

5 with

centre

at the

472
observed

It will be

is

this

that

BESSEL

OF

THEORY

theorem

[CHAP. XIV

FUNCTIONS

of

much

weaker

^ 14'41, in view of hypothesis(II)of " 14-6. The


be* 0 (v^X)for certain values of R, and this seems
O (XR) may
used
in
" 14"41 inapplicable.
type

theorem

of

To

prove

of "t",
F{R,

than

the

of this is the fact that

reason

to make

arguments of the

the lemma, suppose first that T is bounded.


Then, for any value
mono"t")
may be expressedas the difference f of two (increasing)

functions

tonic

character

the

in
F(R, "l")
If Rn and

whose
"l"),
%2 {R, *^).
interval (0,R).

(R,
'x^i

R^

are

the second

follows from

values of R

the extreme

value theorem

mean-

is the

sum

total fluctuation

of

for any particular


value of "i",
it
value of R2 between
that,for some

Ro and R^,

j^X^ (R,^)

(XR)

("^R)^
"^)j^G

(i^o,

X,

{R.

(XR) ^^
^)j^G

Git)"^.
t

"^E,

dt

[^

choose

Xx

+ xAR.,"^)
G(t)^
X^{R"c^)\
J
J
KE,

Since

X, so

is

(t)
"

convergent,if

is

arbitrar}^
positivenumber,

an

we

can

largethat

rG{t)^

"e,

for all values of

less than

not

the smallest value of Rq. Also

I^ (Xi{R, ^)
Ixi {R" "E")

+ IIF(^ 0, ^) I
IF(+ 0, "!")}

^Xi{^,^)-hF{+0,^)
and

l\F(+0,^)\,

similarly

whence

it follows that

F(R,^)G(XR)^^^^
R

2e

r [x,{co,^) + x-^('^^^)+F
J

"TT

"71

and, since F {"{"


0, ^) is bounded, this
small,and
sufficiently
we
may proceedto the
e

some

This

is the

limit when
=

be made

independentof
the outer

{R) Jf,(R); then


imposed on F (R,4")are
Analysis,"3-64.

case

of the conditions

t Cf. Modern

when

it is

can

(+ 0, ^) I}d^^

small by taking
arbitrarily
the outer boundary of T. Hence
boundary tends to infinity.

(R)=EJi (R). It
superfluous.

is

bj uo

means

impossiblethat

'

14-62]
infer that, if T has

We

made

be

can

MULTIPLE

theorem

to

shall next

We

are

Cudur
J

r
J

-r.

providedthat
a

number

where

repeatedintegral.
of the integral

^V{X,Y).g[usJ{X''-+Y')].{dXdY)

f" r ^(X,Y)rg{uW(X"~+Y-'}]udu{dXd

the limit

on

the

rightexists.

of X and

of e, there exists

value
arbitrary
positive

any

(3 such that

is the upper

We

this is the

that

given value

any

of Neumanns

and

-X

lim

For

of

prove that the existence and absolute convergence

conditions
sufficient

Jo

modulus

small by taking \ sufficiently


arbitrarily
large;
be proved.

TJie inversion

14'62.

boundary,the

outer

no

473

INTEGRALS

then

bound

of jg {u) jf^u.

have

\^F{R,
^)g{uR)

udu

-f
J
r-jT

r
J

-n

roo

(dRd^)

(J

r
J

-TT

F{R,^)g{uR)R{dRd^)udu

rx

f"f F(R,

udu

^)g{uR)

(dRd^)

J 13 Jo
"^^

F(R,(p)g{uR)R(dRd(^)udu
J

-n

rr:F(R,^)\u'^duR^dRd^)

^aT

J -ttJ

a[ j^ r
Jo

^^

"

r-rr

-TT

for

small
arbitrarily
R

lim

r
J

on
integral

exist.

values of

e, we

infer that

(dRd^)

udu

J 0

the

\F{R,"P)\Ri(dRd"l^)uidu

fi

C'-i-

F{R, ^)g(uR)
0

-IT

".

Since this is true


""c

the left

-TT

"

f Jor F{R,
J

^)g{uR)

udu

R
.

(dRd^^),

existingbecause

the limit

on

the

rightis

assumed

to

474

THEORY

it follows that,if the limit

Hence

-ir

lim

A-".30

14'63.

The

We

now

are

proofof Neumamn's
in

choose

uniform, in such
choose

stated in

then

the sectors
a

bounds

be B and

(dRd^)

r F{R,^)G(XR)

J0

-rr

theorem.
integral

the theorem
due to
difficulty
prove without
firsttake an arbitrarily
small positive
number

in which

w^ay that the

We

to
position
" 14'6. We

small that II

so

take the upper

to

rightexists,then

and

the

on

F(R,^)g{uR)R(dRd^)

udu
.'0

Neumann

[CHAP. XIV

FUNCTIONS

BESSEL

OF

the convergence
of H (R, "I")
is
to zero
of their anglesexceeds 27r
We then
e.

sum

{R, "!")"

"

in these sectors

Avhenever

8;

and

we

of

(R, ^)

\F{R,^)\

f"(?(w)

and

"

C.

then

apply the

r^

second

mean-value

theorem.

have

We

dR

jj,{R,^)G(XR)'^
G (XR) ^
X^ (+ 0, ^) /^
r^

dR

where

0 ^

^^

C^

(x.{8,"!")

X.

rlTl

j^G(xRf-

(+ 0, "I")1

8.
"^

Now

AS

dR

G(\R)

^j,

"2G.

G{u)

Hence
"*

dR

F(R,^)G(\R)~

and

less
\r]\is

is less than
Hence

F{+0,^)

-tt

where

f"^
=

than

2BG

2eC

inside the sectors

in the

dii

G{ii)-

v,

in which

is uniform,

convergence

sectors.
exceptional

it follows that

"itJ 0

"27r.2eC+".2BG
=

Hence,

for

G (XR)
F(R, "!")
"n

J 0

2eC{27r+ "}.

largevalues of \

^^^^1^^
0, "^) f'(?
27ri"li?'(+
(u)
-

Ji

"

Jq

"2eC(27r

B)

o(l),

that is to say
lim
A

-*"

du

I
j^j^F(R,^)G(\R)^^^^^-27rmF(+0,^

00

2eC

{27r+ B).

476

limiting

The

"

OF

THEORY

5'71

R,

defined

are

radius

the

used

process
of

the

R,

that

indefinitelylarge, and

and
"t"),

by F(R,

XIV

result

of

variables

new

equations
kS,

kO,

function

(k, ", "l"),


{k, 6, cfi);the

coordinates

the

suggested by

substantially cylindricalcoordinates

are

is that

is made

so

[CHAP.

FUNCTIONS

Mehler

by

sphere

the

by

BESSEL

becomes

P,i(cos7)

the

of

points

position /(", ^)

of

approximately

equal

with

polar

is then
to

noted
de-

Jo (117:7
/k),

where
OT-

We

led

thus

are

to

now

1/k

write

we

differential

the

as

F(r,6)

But

this

Thus,

are

so

potential

surfece

necessitates

sphere
/c

00

the

with

the

by

method

he

stated

abgeleitet habe,

the

method

"Die

ist nicht

cause
rigorous proof, be-

(cos 7) sin 0cZ""de

k), which
that

we

surface

transition

of
from

value

put

may

of the

convergence
the

the

assumes

series

the

on

sphere

to

series

to

an

the

f(6, "/")on
in

the

series
of the

surface

plane, by making
integral,iff one

difficulty.

discovered
that

"

theorem

from

theoretical

Neumann

which

corresponding

possible that

of

require j ustification.

/(e, ^)P"

of the

transition

It is

the

discussion

basis

the

that

(when

sphere,

and

integration (taking

an

F{R,^)Jo(u^)RdRd^,

made

be

steps which
know

we

function

of considerable

(pY^"irr

the

of

by

get

we

hardly

can

many

although
X

is

summation

result.

procedure

there

"^).

K'

the

[" udur

(^

cos

J-TT

replace

element),

is Neumann's

which

and

u,

2Rr

.'0

"*7'"

M=0

'h/k

"{- r-

equation

the

K-^'X

If

R-

which
his

has

integral

Methode,

durch

just

been

described

formula

in

welche

ich

voUstandig strenge."

1862.

diese

is the

method

Concerning
Formel

so

his

eben

XV

CHAPTER
ZEROS

THE

will be

functions,which

problems,connected with the zeros of Bessel


in this chapter.We shall beginby proving
investigated
mainly concerned with the fact that Bessel functions

quitegeneraltheorems
of zeros,andwith
have an infinity
functions.

Next,

the relative situations of the

shall examine

we

of Bessel functions.

zeros

classes of

various

are

FUNCTIONS

BESSEL

connected tvitlithe

15 '1. Problems
There

OF

the

of
reality

the

zeros

of different

of Bessel functions

zeros

order is real, and discuss the intervals in


whose
(and cylinderfunctions)
either by elementarymethods
or
which the real zeros lie,
by the use of PoissonSchafheitlin integrals.
Next, we shall consider the zeros of J^ {z)when v is not
real,and proceed to represent this function as a Weierstrassian
necessarily
product. We then proceedto the numerical calculation of zeros of functions
consider the rates of growth of the zeros
with
of assignedorder,and finally
of cylinder
functions of
the increase of the order,and the situation of the zeros
of the results
unrestrictedly
largeorder. A full discussion of the applications
is
contained in this chapterto problemsof Mathematical
Physics beyond the
such
will be made in the
of
this
references
to
book, though
applications
scope
of the chapter.
course

Except

in

""15'4

15"o4, it is

"

supposed that

the order

v,

of the f mictions

is real.
under consideration,

The

of functions whose

zeros

themselves

to

discussion

the
particular

of

zeros

more

^
"

order

is half

readilythan

-^^
dx

"

the

have

an

odd

of other

zeros

been

integerobviouslylend
functions.

investigated
by

In

Schwerd

and

of
the zeros
by Rayleigh*;and more
recentlyHermitei* has examined
Jn+\{x).The zeros of this function have also been the subjectof papers by
used the methods
who
of Sturm; but it has been pointedout by
Rudskiij:
Porter and by Schafheitlin"
of Rudski's results are not correct, and,
that some
in particular,
his theorem
of Jn^^(^) lies
smallest
that the
positivezero
between
Such a theorem is incompatible
\{n + 1) and \ (n + 2) is untrue.
with the inequality
given in "15"3 (5) and the formrdae of ""15'81,15"83.
tt

Schwerd,

Die

tt

Beugungserscheinungen

(Mannheim,

1835); cf. Verdet,

Math.
rjujsique, (Paris,18G9),p. 260; Kayleigh,
i.
t Archiv der Math,
und
20"21.
PInjs.(3) (1901),pp.
X Mem. de la Soc. R. des Sci. de Lie/ie,(2)xviir. (1895),no. 3.
Fizyczne,iii. (1892),pp. 69"81.
[Jahrbnch ilber die Fortschritte
Froc.

I.

" Porter, American

(1900),p. 304.

Joiirwil

of Math.

xx.

London

Sac.

See

iv.

Lemons cVOptiquc
(1873),pp. 95"103.

also Prace

der Math.

Matonatijczno-

1892, pp. 107"108.]

(1898),p. 198; Schafheitlin,Journal

fiirMath,

cxxii.

478

THEORY

The Bessel-Lommel

15"2.

BESSEL

OF

theorem

the

on

Bernoulli*

of J^,(z).

zeros

and

Fourierfthat Jo (z)has an infinity


of real zeros
analysisof Parseval's
; and a formal proof of this result by an
It
observed
is due to Bessel^.
was
subsequently
by Lomniel" that
integral
Poisson's
t
o
for J"^(z),
Bessel's arguments are
immediatelyapplicable
integral
of Rolle's theorem
to
|"
application
^ |. A straightforward
providedthat
theorem
is then adequate to prove Lommel's
that J^ (2) has an
x^''J^{a:)
of real zeros^for
infinity
any givenreal value of v.
It

by Daniel

[CHAP. XV

FUNCTIONS

stated

Wcas

i'

"

and

;/;

lies between

J^(x) is

mir

number

odd

an

function

continuous

i;

"

(0,2, 4, ...),and

of m,

has

consists in provingthat when


investigation
| " ^ |,
and (m +|)7r,then J^(x) is positive
for even
values
is negativefor odd values of m, (1, 3, 5, ...).Since

Bessel-Lommel

The

of

of

each

in

zeros

when

it is obvious

x^O,

intervals

of the

that

J"(x)

27r),
(|7r,
tt),(|7r,

(Itt/Stt),
....

Some

preciseresults

more

To prove

of

similar character

theorem, let

Lommel's

transformations of Poisson's

obvious

integral,
r2m+9

^)r{i).(2m

r2m+9

and

sgn J. (x)

sgn

so

the last

Now

j^

{(2m +

15"36.

0 ^ ^ ^ 1 ;

then, by

^ttu
''"'
ey
1^-]^-"
-

l,jj.y^

cQg

^"-

|^2m + 6)"- ..^j^-'


-

be written

integral
may

cos

dy]o

"

have

we

r{v

given in "" 15*32

(ni+ 1$)tt where

2{l7ry
^''^""^

will be

in the form

7)1

If

cos

(-)-.,. \

where

write

we

now

{(2m + ey

2r

(2/-

1 +

1 "

"

TTlf

du,
^^^_^^^-___
^^_^

U, and then put

uy}"-^

[{2m + ey

(2r

uy]''-^
=f,.( U),

it is clear that

ff-iU)
sin ^TrU.dU,

v,=

.'o

"[ function of
and, since IIv ^^,fr{U) is a positive
increasing
*

Acad.

Comm.

t La

Theovie

Berliner

" Studicn
IIThis
convergence
% The

then

Sci.

Imp. Petwp.

Anah/tique

Abk., 1824,

de

point at

of the
reader

(1732"3) [1738], p. 116.

la Chaleur

(Paris,1822),"803.

p. 39.

liber die BesseVschen

is the

vi.

r.

which

Functioiwti
the copdition

(Leipzig,1868),pp.
v

65

"

67.

$ ^ is required ; the condition

"

-^

ensures

the

integral.

will prove

this without

(6^)with respect to
differentiating/^

any
r.

by regarding;" as
difficulty

continuous

variable and

15-2-15-22]

OF

ZEROS

BESSEL

479

FUNCTIONS

It follows that
and

so

+ (v,"_" v,"_3)
[(-)^ [vj + (i',"v,,,^,)
+ ...]]

+ I^tt) sgn
sgn J^ (i/iTT
=

sgn(-l)'^,

since v"/ is

obviouslynot negative.

is to say, when

That

|"

"

z^

|,
+,

and

from
The

this result Lommel's


of e/j(x),as

zeros

well

method.

Bessel-Lommel

lies in the intervals


the intervals

more
slightly

It
.

zero

result for t/j{.v)


is that
and

The

the

that

C. N. Moore, Annals

of Math.

Trans.

by Whewell,

between

2 and

15'21.

The

Camh.

Phil.

^72,and that the function

non-repetition
of zeros

function

function

Soc.

is

positivewhen x
negative when x lies in
gated
Ji{x) has also been investi-

it is

(2) ix. (1908),pp.

just stated are of a less exact nature than


refined analysisby Schafheitlin ("!^
15-33

noted

was

Baehr's

(0, tt),(f7r,
Stt),(fir,
5tt),...,

by

alreadystated.

manner

have been investigated


of J^ (.r),
bj Baehr, Archives
which
resembles
the
358, with the help of a method

"

....

results

The
of

351

(i^Tr,
-Irr),
(-V-TT,
67r),
-irv),
{l-rr,

in this way

0, 2, 4, ...)

those

as

JVeerlandaises,
vii. (1872),pp.

follows in the

theorem

(m

156

162.

"

the results obtained


"

ix.

H(, (2)has

with the aid

15*35).
lias a
(1856),p. 156, that /q {'*"')
real

some

zeros.

of cylinder
functions.

It is easy to prove that %^y{z)has no repeatedzeros, with the possible


exception
it would
of the origin*.
For,if^t^^,(z)a.nd
'^Jiz)vanished simultaneously,

follow,by repeateddifferentiations of the differential equationV^ ^^ (z)


that all the differential coefficients of W" (z)would vanish at the common
zero

"

^^ (z)and WJ

and then, by
(z),

The

15'22.
It will

now

be

shewn

that

This result is sometimes


interlaced with
To

prove

of

if ".

first of

between

the result

use

we

This

these

the

zeros
positive

of Jy

(x)

formulae

recurrence

a"-^/,+l{x\
that

shews

each consecutive

is

"

[x-^J,{x)]

oi x"^'^Jyix); and
*

the

J^ (if),

those of J^^i(x).

there is at least
./^(,7)

OP""

of

"""

expressedby sayingthat

the

of

zero.
identically

zeros
jv,-!, are the positive
1,
magnitude,then, if "
1,

be

of Bessel functions.

interlacing
of zeros

arrangedin ascendingorder

are

theorem,^;, (z)would
Taylor's

0,

one

pairof

the result is

specialcase

of

between
zero

theorem

now

of
zeros

/,+! {x)]
\x^^^
=

a;""

each

A-+"

consecutive

and
1/^+1
(a;),

./,(.c)
;

pair of

the second

shews

of a'""^^
t/^+j
ix) there is at least

of

zeros

one

that
zero

obvious.

proved by Sturm,

Journal

de Math.

1.

(1S3G),p.

109.

480

OF

THEORY

If i^^"

1, the

the smallest

nearer

The

result

This

but

^^ (x)and

fundamental

the

and

smallest

the

zero

of Jv i{x) is
+

of Jt,{x).

zero

concerning interlacingof positivezeros

cylinderfunction*

[CHAP. XV

FUNCTIONS

obviouslystill interlaced

are

zeros

the origin than

BESSEL

contiguousfunction

simpleproperty

is

obviously true

for any

real

"^^ ^ (^'),

of Bessel

functions

appears
century agof, when

never

four
proved until about a quarter of a
which
mathematicians
the proof
has
publishedproofsalmost simultaneously;
which
is
due
t
he
other
been
to
Gegenbauer;):
just
given
proofs,
and,Porter|;
elaborate
of
a
character,
are
were
given by Hobson|| and
slightlymore
been

have

to

Vleckir.

van

It has been

that, since

Porter

pointedout by

2
r

7-/\

(^+1)7-

J^ {x) + J^+o {x)

/\

J^+1 {x),

(a^)and J^+o(x) have the


positivezero of J^{x) the functions .7^+1
same
(x) alternates in
sign; but at successive zeros of J^ (x) the function i/^+i
odd
and
there
number
between
each conof zeros
of t/^^.,
an
secutive
so
are
sign,
(x)
functions
of
t
he
zeros
ofJy(x); interchanging
/^+2(^)and
pair positive
J^ (x) throughoutthis argument, we obtain Porter's theorem that the positive
at

any

of

zeros

J^+o{x) are

15 "23. Dixon
A

result of

interlaced

theorem

with

the

on

those of J^,(x).

interlacing
of zeros.

more
slightly
generalcharacter than the theorem of " 15"22
\, and A, B, G, D are
namely that, when v"
constants
such that AD=f^ BC, then the positive
of AJ^.{x)+ BxJJ{x) are
zeros
interlaced with those of CJ^ {x)+ DxJJ (x),
and that no function of this type
have a repeatedzero
other than x
0.
can

is due

C. Dixon**,

A.

to

"

The

latter part of the theorem

deducible

from

is an

immediate

of the formula,

consequence

" 5'11 (11),


J^ (x),

j J,-(t)tdt -^x

d{J,ix)]

.1

xJJ (x)
d

\xJJ {x)]

dx

'

dx
i

for the
would
*

is
integral
vanish

real

at

is

when
positive

repeated

of

zero

cylinder function

is

an

and the
positive

AJt,{x)+ BxJJ

expressionof

on
expression

the

right

(x).

the form

aJ,(.T)+^Y^(x}
in which

a,

j3and

t Cf. Gray and

are

real,and

Mathews,

is

positive.

Treatise

on

Bessel

Functions

(London, 1895), p.

Phys. viii. (1897),pp. 383"384.


iv. (1898),
pp. 274"275.
Math.
Soc. xxviii.
IIProc. London
(1897),pp. 372"373.
IT American
Journal
of Math. xix.
(1897),pp. 75 85.
also Bryan, Proc. Camb.
Messenger,xxxii. (1903),
p. 7; see
+

Monatsheftefilr Math,

" Bulletin

American

Math.

-50.

uiid

Soc.

"

**

PhiLSoc.vi.

(1889),pp.

248

"

264.

15-23,15-24]

ZEROS

BESSEL

OF

part of the theorem,

To prove the former

'^^''^~

0'{x)
^
'
^

(x)

GJJ^)+^JJ

AJAx)-\-BxJ:{x)'

0 {x) is monotonia.

so

with the

"

^)o

rw

[AJ, {x) + BxJJ {x)Y C,

and

'

\ii'^,
f J^'(0 tdt,
\
D

=
"

The

therefore interlaced
are
(a;)

of ^

zeros
positive

from

and
positive
poles,

that,if

observe

we

then

481

FUNCTIONS

this result the former

part of the theorem

is obvious.
If the function

is replaced
J^,{x)
by a

real

function a./^,
(x)-\-j3Y^,{x)
cylinder

have

we

x'W^ ix)

j'gf,{x),

%V (0 tdt

!*" d;^(x)
I dx

'2vj3
(a sin

[x^i^J
{x)] +

'

/3cos vtt)

vtv

IT Sm

VTT

dx

1 "
and
" 1 ; and so the theorems
non-repetition
concerning
providedthat
for A9^^{x)+ Bx'WJ (x)and C'(ff^{x)
true
of zeros
+ Dx^ifJ {x)
are
interlacing
providedthat /3(a sin vtt + /3cos vir)is positive.
y

"

Again,since
%^.{x\

d^'S.ix)dlx'^Jix)]

ix

'

.dx

or

l,B

0,G

15-24.

for

The

" 15"22
v,I)

zeros

'gf;{x)+ 0^9^J-"
{x)],
[{x- v'')
-

exceeding+ \/v-,whether

lies between

"

'^.{x)

of Dixon's

specialcase

of order
of cylinder
functions

in which

v.

functions of the
{x)be any distinct cylinder
zeros
positive

are

a/, {x)+ ^Y, {x\

same

order ;

interlaced*.

%\ {x)

%".(x)%%' (x) '^%(x)-gf;(x)

then

theorem

-l.

shall prove that their


If

is the

of zeros
interlacing

Let 'Wv(x) and ^^


we

not.

result of

The

1
=

dx

is true

the theorem
and 1

x^J(x)

J",(x)+ h \\ {x),

2(aS-/37)

TTX

of ^^ (x),
^J (x)has opposite
zeros
that,at consecutive positive
and therefore,
from the last equation,
signs;that is
'"y{x) has opposite
signs,
it is known

Now

to

say

'^(")has

zeros
positive

an

odd

inimber

of

zeros

each

between

"Wyix) has an odd number


similarly
of '"' {x); and so
zeros
pairof positive

of '^^{x);

each consecutive

consecutive
of
the

zeros
zeros

pairof
between
must

be

"

interlaced.
If

take

we

deduce

we
*

one

that ail real

Nova
Olbricht,

Acta

of this result with


w.

B.

F.

functions to
cylinder
functions have
cylinder

of the

be
an

function of the firstkind,

of positive
zeros.
infinity

1888, pp.43
{Halle),
Caes.-Leop.-Acad.
"

some

instructive

iS, has

giveu an

elaborate

diagrams.
31

cussion
dis-

482

LommeVs

15'25.

extension

An

which

zeros

theorem

of

is that,if the order

of Jv{z).

zeros

Fourier*, that the function J^iz)has

to

exceeds

1, then the

"

functionJv{z)

has

luhich

zeros

no

no

theorem

real.

not

are

reality
of the

real,has been effected by Lommel-f-, The extended

not

are

due

theorem

the

on

[CHAP. XV

FUNCTIONS

BESSEL

OF

THEORY

To

that a is a zero
of J^{z)
theorem, suppose, if possible,
from
series
for
is
the
not
real. It follows
a
J^,{z) that a
pure
Lommel's

prove
is not

which

then

imaginary,because

(-)'"(ia)^

",^Qm\T {v-\-m + 1)
would

be

series of

Op be

Let

i/

complex number

the

J^,{z)is

J^ {z),because
Since

terms.
positive

"

conjugateto

real function

tJ^(at)JAaot)dt=
so, since

and

a-

that

so

olq is also

zero

of

z.

dJ^(a^w)

J-

"

dJ^{ax)'

a,

" 5"11 (8) that

1, it follows from

"

of

"o

"

a,,-,
''1

tJ,{at)J,{a^t)dt0.
=

Jo0

The

the left is

integrandon

Hence

the number

Similar
function

be

AJv{z)-\-BzJJ{z)has

These

{AjB) +

exist,and

cannot

arguments | may

when

zeros

a.

and
positive,

used

to

all its

so

obtained

have

we

is

the theorem

shew

that,if

and

contradiction.

proved.

are

real,except that it has

zeros

real and
two

"

1, the

purely imaginary

v"0.

results follow from

the series for

-^

j^ (2)}combined
[z-^IB

with

the formula

that

^^^^q\

1:tJ^{^t)J^{Mdt
which

is satisfied if /3and

15'26.

^0

are

any

0,

of AJ^ {z)+ BzJ^'

zeros

analogue of Lommel's

The

theorem

(2)such

forfunctionsof the

second kind.

It is not

to prove by the methods


of " 15"25 that"V"{z) has no
possible
complex zeros in the region||in which |arg^r j" it. But it has been proved
real part,other than
by Schafheitlinl that Fq (~)has no zeros with a positive

the real
*

La

zeros.

Theorie

Analytique de

la Chaleur

(Paris,1822)," 308;

see

also

Stearn,QuarterlyJournal,

(1880),p. 93.

XVII.

t Studieii iiber die BesseVschen


X See A. C. Dixon, Messenger,
" Or,

arg

vanish

Proc.

London

Archiv
the

(Leipzig,1868),p. 69.

(1903),p. 7.

generally,"^^(2).

more

iiWhen
cannot

Functionen
xxxii.

"

Math.
der

complex

is half

Soc.

Math,
zeros

1^ (2) e"^""^
l\ (

tt,

unless

nnd
of

xxx.

of

an

odd

(1899),pp.

2)" 2/ cos virJ^( 2),and hence, by " 3-(53 (1),1^ [z)


integer. This type of reasoning is due to Macdonald,
-

165"

179.

Phijs.(3)i. (1901),pp. 133"137.


Yi (z)to

similar

treatment.

In this paper

Schafheithn

jects
also sub-

484

BESSEL

OF

THEORY

Hence, for sufficiently


largevalues of
the

^raihv)

curve

one

^ ^i^s in the finite part of the

"

(|,v) is
"^,"+i

negativewhen

so

zero

other of the closed branches

or

Hence

as

m-^00

and
the ^-plane,

for which

those
(i.e.

[CHAP. XV

FUNCTIONS

the

plane,and

^",+i(^,'"?) 0 lies

curve

which

2?n is positive),

the

compose

is
"f)m{^,v)
inside
wholly

^",(^,77)

curve

complex zeros of gzm,^ (0 li" i^ibounded


have limit-points.
consequently

0.

regionsof

the

Now, since,by ""9"65, 9'7,

domain
of the ^-plane,
small in any bounded
by taking
arbitrarily
of
number
that
the
it
follows
from
m
Lagrange'sexpansion*
sufficiently
large,
zeros
of/^(^) in any small area is at least equalto the number of zeros of g^m^^i^)
when m is sufficiently
in that area
large; and so /I (^) has 25 complex zeros.
be made

can

None

of these

pointof
double

two

zero

is real,for if

zeros

of them

one

conjugatecomplex zeros
of/^(^);and/^(^) has no

real it would

were

of ^am, ^(0" '^^^


double zeros.

be

it would

so

count

as

(2s+ 1)
for/^(^)it is seen that,when v liesbetween
have
than
and
it
for
has
cannot
and
more
one
negativezero,
(2s+ 2), then/^(^)
than once
could be made
to change signmore
one
negativezero, then gom,y{^)
be made
to differ from /"(t)by
00
can
as
^ varied from 0 to
[since
g."m,v{K)
small number],and this is impossible.
an
arbitrarily

Again,from

the series

limit

"

"

For similar

have

cannot
jfy(^)

reasons

than

more

2s

complex zeros.

replace^ by ^ z-, so that negativevalues of ^ correspondto purely


imaginaryvalues of z, we obtain the results stated in the case of Jt,{z).
If

we

For
based
Ges.

discussion

a
on

(1911),pp.

XI.

of

ideas,the
in

brief

15*28.

Bourget's
hypothesis.

It has

been

consult
that

paper,

with

zeros

of

polynomials

Lindner, Sitz. der Berliner Math.

Hurwitz

extended

has

Hamburger

conjecturedby Bourgetf that, when


functions

than

all

the

It

in association

should

his results to

Mittheilungen,11. (1890)^

31.

"

(zeroincluded),the

the

reader

be mentioned

It may

functions

generalisedBessel
pp. 25

5.

"

of Bessel functions

zeros

rather different

seems

for
origin,

The

values
positive
integral

that this theorem

simplecases

Jt,{z),
-/"+,",
(2^)have

1, m

has

been

never

of

is

no

common

Cf. Modern

A7in.

Jv {z)Rm, (z) Ju-i {z)Rm-i,f+i (z)


-

A7ialysis," 7-32.
norm.

sup.

iii.

other

proved except (cf " 15'22) in

2.

Sci. de VEcole

zeros,

m.

formula

Jv+m (^)

integer
positive

(1866),pp.55

"

95.

15-28, 15-3]

OF

BESSEL

485

FUNCTIONS

and J^_^{z) have


that,since J" (-2^)

shews
oi

ZEROS

no

common

zeros,

the

common

zeros

the equation
must
satisfy
Jv{z)and J"^4.,"(2^)

The

numbers.
algebraic

be

i.e.they must

truth of

therefore

Bourget'shypothesiscan

proved that ./"{z) has no zeros which are


integer;but attempts to prove this theorem*

numbers
algebraic

be

When
have

common

no

that

no

is half of

an

zeros

so

when

can

is

an

far failed.

it is easy to shew that J^ {z)and Jv+m {z)


integer,
for such zeros
numbers
and it is known
are
algebraic
the equation
J can satisfy

odd
;

number
algebraic

'

cot(^--|^7r-i7r)
since the

have

if it

be established

is

right-handside

in
algebraic

-p

when

is half of

odd

an

integer.

and Legendre that ir- is irrational may be


proofJ given by Lambert
appliedto " 5"6 (6) to prove that ./"{z) has no zero whose square is rational
is rational; and so, from a consideration
of R,n-i,v+i{z),
when
V
Bourget's
4.
3, vi
hypothesisis true in the cases m
The

Elementaryproperties
of the

15"3.

It is possible
to

smallest

the

formulae

recurrence

of the

zeros^

of J, {x).

information
interesting
of J^ {x)and related functions,when v is positive,
zeros
differential equationsatisfied by J^ {x) togetherwith

acquirea

concerningthe
by a discussion
theorems

considerable

shall

; we

such
"concerning

of

amount

establish the

now

truth

of

selection of

zeros.

The

reader will find a more


in various
ijof these theorems
systematicinvestigation
321 ; Archiv
by Schafheitliu,
(1900), pp. 299
notably Journal fiirMatli. cxxii.
Math,
und
III.
137; Berliner Sitzungsherichte,
Phys. (3) i. (1901), pp. 133
(1904),

papers
der

"

"

pp. 83"85.

will be
the smallest positive
of /^ (x),J J {x),J J'(x),
zeros
brevity,
called j^,jj,jj',
The smallest positive
of Y^{x), YJ {x),YJ' (oj),
zeros
will similarly
be called y^, yj,yj',
For

...

....

We

(1)

first prove
/^

jV "

It is obvious
*

I consider

succeeded

that

in

that

from the power


the

theorem

is

V,

jj

"

V.

series for /"


probably

true

{x)and

it is

an

JJ

{x)that

abstruse

these functions

theorem,

and

I have

not

proving it.

Journal
noticed by Porter, American
of 2["itli. xx. (1S9S),p. 203.
was
Hobson, Squaring the Circle (Cambridge, 1913),pp. 44, 51"53.
Sac. (2)xvi.
Math.
related results are due to Watson, Froc. London
(1U17),pp.
" Some

t This

X
171.

Cf.

165

"

486

THEORY

OF

BESSEL

small positive
for sufficiently
values
positive
equation

are

doc

[CHAP. XV

FUNCTIONS

of

and, from the differential

dJ^,{x)

\={v^-a?)J,{x\

dx

it is evident

is positive
and
that,so long as ^ " y and J^,{x) is positive,
xJ^ (a-)
and so J^,(x) increases with x.
increasing,

Therefore,so longas
functions

that

so

Again,from

"

and jj
j^,

a; "

v,

both J^ (x)and xJ^' (x)are


*

cannot

the differential

be less than

equation

vj;'(v)

-j;(v)"0,

and

//'(x)has

so

become

Hence, when

zero.

positive
increasing

v.

negativebefore

has increased

to

the value

from

I v"l,

jj'"r.

(2)
Next, since

the rightis positive


less
Now, if // were
on
so longas a; " z^ + 2.
expression
than ^/[v
the rightwould be negative
when x is equal
the expression
on
(v-f 2)},
less than j")"
and this is not the case.
from a graph,is obviously
to jj (which,
the

Therefore

j;"^[v{v

(8)
Now,

from

2)}.

it follows that

" 15*22

Jv

"

Jv+\

"

Jv+2"

and, as has justbeen stated,

so

that

both

JvijJ) and J"+o{jJ)are

If
positive.

now

put cc=jj in the

we

formula

/,+,{x)

it is obvious

"

Similarly,
by puttingx =j^
(v + 3) /, (x)+
deduce

and

"

J, {x)
^1

J, {x\

-^

that

(4)

we

^^

|1

+ 2)
(i.

"

V{2. {v + 1)}.

in the formula

j^,^^
(^) (^ 1) J^^^(^.)
|lil^i"^0:"?)|
-

that

therefore

^{v(v+ 2)}"j,."^{2(v + l)(v + S)].

(5)
*

Cf. Eieinann,

t When

0";v"l,

(Brunswick,1876), p. 269.
Differentialgleichungen
small values of .r.
J/' {x)is negativefor sufficiently

Partielle

0,

15-3]

ZEROS

In like manner,

and

we

OF

BESSEL

deduce

can

from

-vW-

J\^,{x)

FUNCTIONS

487

the formulae

"^1J, {cc)-{v+l)

J J'{x)

that

(6)

"^\v{v-\)]"jj'"^{ir'-l).

Some

rather better

formulae
complicated
"

Schafheitlin*

than these are


inequalities
thus, from the equation

("*')
=

p+ti

deduced

"

(*'}-^5,
v+\ (j^)
"

v+l

obtainable

by takingmore

\^) ^i,v+i('^7;

that

-R5..+i(j.)"0,
i.e.

3"^ -lQ{v

Since

less
jj^is certainly
be less than

must

3^'

than

the smaller

-\Q{v+

16

+ 4) (z/
+ 5) "
(i/+ 1) (z;+ 2) (i;

0.

4- 4), by results alreadyproved,


+ 2) (i^
f (i/
J"

of the

root
positive

2){v + 4) X'

16

equation

+ 5)
(z/+ 1) (j^+ 2) (1/+ 4) (i/

0,

hence, a fortiori,

and
"

2) {v + 4)j,-+

""V[|(^

(7)
from
Similarly,

the

l)(^ + 5)}.

equation

J^ {x)[R,^{x)+ i^3,
^+, {x)

{x)
4J"'^+4

Rs
-

Schafheitlin deduced

jj " ^{v (v + 2)},

and, when

"

4,

i;"V[z.(.

(9)

inequalities
being derived from
roots of the equation
positive

these

x'-S(u
The

that

(8)

the

i?,^
,+o (x)]
2// (.c)
i?^"^i (")},
[i?o
^^.1 (ic)
(x)
-

discussion

2y

x""+

2v{v

3)i,

the consideration

that

+ 4)
1) {v + 3)(i^

of y^, requiresslightly
more

abstruse

j^'lies between

0.

reasoning.We

use

the result that

JJ^(x) + IV (a-)
is a

function of x ; this is obvious from " 13 73. Hence


it follows
decreasing
that Yv"(x)decreases through the interval (0,jV),aiid so y^ exceeds jV; again,
is
in this interval Y^{x) is negative,
and it follows from " 3'63(1) that Y^(j^,)
since JJ(ju) is obviously
negative.
positive,
Hence

j;"i/,"j,.

(10)
This
with

(withyVreplaced
by i^ -f h) was
inequality
the aid of rather elaborate analysis.
*

Berliner

t Journal

established

in.
(1904),p. 83.
Sitzungsberlchte,
cxxii.
(1900),pp. 317"321.

fur Math,

by Schafheitlinf

488

[CHAP. XV

FUNCTIONS

BESSEL

OF

THEORY

values of cylinder
functions*
Stationary

15*31.

function J"^{x)cos a
Y^ (x)sin a,
cylinder
infinite
of positive
an
^t,{x),has an infinite number
or
zeros, and so there are
Such values of x
number
values of x for which it is stationary.
of positive
where
will be called /u,i,/Uo, /u.3,
which exceed the order v (supposedpositive)
It has

alreadybeen

that the

seen

"

yU.2"

yLti"

shall

We

"g^.W, -^.W,
The

which

firsttheorem

"

"

"^.W,

the sequence

....

shall establish is that

we

this,observe that the function

To prove

"

of
simplerproperties

of the

study some

now

"

Ms

..,

{x) defined

as

x'^^J-'ix)
7)^
v"

TX-"

has the

negativederivate
-2x'^;'(x)/(x^--v'y,

and

" A(/u,3)"
" A (/Xo)
A(/u.j)

so

Since

....

the truth of the theorem


(fXn) ^V" (/*"),

is

is

result
interesting

more

suggestedby

evident.

now

Hankel's

asymptoticformula

("7-21)
%% (x)

This indicates the

when

fin

It

can

is

(^y

{x + a-^v7r-l7r)+0

cos

[^^
.

consistent
of provinginequalities
possibility

with

large.

in fact be shewn

that

(I) The values assumed by (x-" v^)^


|^^(ic)|when
fi-i,-'.forman increasing
sequence ivhose inernhers are
(II) The values assumed
/jLr+2" -..form

by xi\^t,(x)\when
whose

decreasing
sequence

members

takes the values fi^, /m,


less than "J(2/'7r).

takes the values /Xr, f^r+i,


are
greater than \/{2j'ir)

providedthat
(i) v"^ V3,

(ii)yu,""

V-

[^v-+

+ 13)}/(4i/='
V(48i.-^
3).

4 +

Consider the function


A
where

A
"'

{x)W,^ {x)+ 25 {x)% {x)'W: {x)+ G {x)-^/^{x)

(x),B

(x)

{x),C

{A'(x)
-

chosen.
suitably

We

(x),

have

'^:~{x)
(x'- V-)B {x)lx^]
-

^, (x)-^Z (x)
[B'{x)+ A(x)-B (x)/x (x' v')G (x)/x'}
[C (x)+ 2B (x) 2C(x)/x}^:^ {x)
2

+
+
=

(x) are

be

to

i){xYW:-'{x\

where

D{x)
*

Cf. Proc.

London

G'

{x)+ 25 {x)-2G

Math.

Soc.

(2)xvi.

{x)lx,

(1917),pp.

170"171.

15-31, 15-32]

defined

by

(cc)is

providedthat
the

{x)

1, 2,

x"- [B'(x)+

(x)
-

{x)lx]j{x^v").
-

v'-f

{x){x"

+
(3a,^

Of

function
increasing

of

\4"x-v'+
which

0(ic)

^v')" 0,

is therefore less than

2/7r.

that when
we
^V (/"")
see
{ixn) {i^n' v'-)^
then the numbers
v^)^
|'^^(fin)
| form
(/u.,;"

less than

then

(x) are

7i

"

(x)l(x- v'),

lim

Since

(x) and

{x) is an

so

that B

{x- v-)Kthen

2D
and

and
arbitrarily

x-A'

(x)

0{x)
If A

chosen

489

FUNCTIONS

equations
2B

(I)

BESSEL

OF

ZEROS

the values

assumes

increasing
sequence

an

v'(2/7r).
If A (x)

(II)

2D

then

X,

{x)("- v-y

{(4z.-3) X'

X'

8j/- (v-+

1) x'

v' {^v~
-

1)}

"0,

providedthat

4y-

3 and

"

Sv-

{^v- 3) ic"

In this
to

case

those used

the greatest root

exceeds

{v-+l) X'- +

(4z/- 1)
-

equation

0.

function and we can


applyarguments,
{x) is a decreasing
in theorem
(I),to deduce the truth of theorem (11).

of the
investigation
Schafheitlins

15-32.

v'

of the

zeros

similar

ofJo(x).

have been given in "6"12,


it has been
which
integrals
of Jo (x)lie in the intervals
shewn by Schafheitlin* that the onlypositive
zeros
of Yq{x) lie in the intervals
zeros
(mTr+ ^ir,nnr +^7r) and the onlypositive

By

of the

means

{niTT+ ^TT,rnir

where
f-rr),

shall first

We

and
modifications,
of the

give

then

0, 1, 2,

shall prove

type
J,,(x)cos

(where

lies between

Schafheitlin's
From

an

"

h and

of the
insjDection

that, when

rnv

sgn

"

F" (x)sin

"

sm

"

vnr

6 vcos

Journal

the values 0

sgn ("

1)'",

sgn Jo (x)

sgn ("

1)"'.

no

zeros

JiirMath.

in the intervals
cxiv.

by Schafheitlin.
and ^ir of a.

f tt,

to

(sin(x + ^0)}

so

Jq (x) has
Consequently

used

" 6"12 (7),

of

methods

the

confined

formula

it is obvious

|), by

were
investigations

77

and

for Jo{x), with slight


investigation
similar results for cylinderfunctions

Schafheitlin's

w^e

"

(nnr,imr

(1894), pp. 31"44.

f tt).

490

that J^ {x) has

To prove

no

{m

T.

"

"

"^

last

^ vcos

sm

write
tt),

"

"

,,

d6.

is negativeor positive
accordingas
integi-and
0

"

j0

0,

"

TT

Since

1) TT

4-

{mir+ ^vr,imr

2(-)'"+M-^-sin(A^-(f))
^jS
^ ^
J,{x)^ -^"^

then

The

[CHAP. XV

FUNCTIONS

in the intervals

zeros
X

and

BESSEL

OF

THEORY

2(^

"

2"f""e" l-rr.

or

longer;and

of these intervals is the

Itt, the second

"

"

g"

2a; cot

the function

sin 6 Vcos 0
is

Hence

positive
sign,and

has the

of values of 6 than

We

shall

a "

tt

and

0, 1, 2,

"

^ i^TT

Of, 1MT

"

zeros
onlypositive

TT

from the formulae

once

mir

a"

"

"

"

irnr

f tt

,", ,

" 6"12,

vd -\-\6)'\ sgn

(x) is

the

obtain the

Consequentlythe onlyzeros

zero.

exceptionof
possible
result
precise

more

(" !)'"",

"

not

(mTT+ f TT
logarithmicderivate

^vTT" a,

niTT

are

odd

an

the firstif a

that the

1 TT

cosee-^

+ itan

"

number

of

f tt

^vir.

zeros
onlypositive

5 i/TT

is

{2x- sin ^cos^)

"

lie in the intervals

Its

of

|i/7r a,

intervals,and there
{x) lie in the specified

in each interval,with

^" {x)

-'^

[sin{x +

sgn

we

of

a)

"

f cos-^^sinCr+g-r^+i^)

for such values of x, ^^

Next

of the type

have

of ^y

result that the

This result follows at

for,when

so,

^^.

Jy {x)cos CL"Y^, {x)sin a,

the crude

^"^^ ^^

and

^ "v

"

that

c^ (^^

we

in any of the intervals


of Jo(^) 'T-re in the

Schaf heitlin's results to functions

{mir + f TT
shew

"

| " v^\.

"

shall first prove


lie in the intervals

which

(i^ 0) is greater for the second


integralunder consideration is

ofSchafheitlins
type, when

We

where

value, but

|7r).

??i7r

0 ^

The

numerical

same

so

^^ (x)
where

"^)has the

"

for the first set.

extend

now

(|^

Jo(x)cannot have a zero


the only positive
Therefore
zeros

Theorems

15'33.

sin

angle.

acute

2c^ there correspondsa value

0 and

the cofactor of sin

and
positive,
consequently
+ ^TT,mir+Tr).
(niTr
intervals (^mr + f tt,

"f7r and 6 is an

of d between

each value

to

and ^ttfor which


2"/"

between

set

of 6 when

function*
increasing

an

6.

"

a)

zeros

of ^^ {x)

492
have

we

Next

obtain the

we

l 6)]

(-1)*",

sgn

stated is obvious.

the theorem

whence

a-v6

[sin{x +

sgn

[CHAP. XV

FUNCTIONS

BESSEL

OF

THEORY

result
precise

more

that the

of9^^(x)

zeros
onlypositive

lie in the intervals

{miT "^TT
where

(it

a,

"

l7r+ \v7r

We

the

use

ayinir

"

except that, if

...,

in the interval

zero

0, 1, 2,

{vir

-V

(0,IvTr

^ir

"

"

is

^TT^-^vir oi),

"

"

near
sufficiently

a),and there may

there may be
in the interval

to it,

be

one

a).

"

and

notation

same

in

reasoningas

" 1533

only now,

if

6 =/(^),
g-2xcotecog.-i 6'/sin-''+i
of 6 ; but it is

function
an
increasing
f(6) is not necessarily

0"ylr"2ct),then

that,when

observe
result,

this

obtain

To

that

2.^'{cosec2
^ ^og-^Jr~^,
+ cosec^
+ x/,)
{2(j"

(2c^ yjr)}

"

to prove
sufficiejit

n\\r
Tl'zfh
Air)
"/(2(^-x/r)
dy}r
"

"

sin

increasingfunction

because
when

40

is

acute

an

angle;

ij/)

cos
"^)'\
(20 + ^)

exceeds

and

so

log'-.

(20

cos

the former
is

"

y-

by

-y/r)

increasingfunction

an

ahvayspositiveif

it is

positive

i|/'0, i.e.if
=

Ax "

and

(20-+

and therefore,a fortiori,


\|^,

of

increasingfunction,since this function

an

"1

^/r)sin(20-V')J'

+ cosec2 (20
(20 + \//)
[cosec2

is

sin

"

+ cosec^ {2(p ^)] sin (20 + ^^)sin (20


(2"^+ 1|^)
[cosec2

But
an

(2(^ \^)
(20 + ^//-)cos

_cos

is

2^ +

40

this is the

case

Hence, when

when

{{v i) tan2 20
+
!"'

.v"

| " v"^,

the

+ 1)}sin 40,
(2i/

"".

of ^^

only zeros

(x),which

exceed

f i; + 1, lie in

the intervals

(mTT
The

method

which

"

a,

nnr

{tt+ ^vtt a).


"

"

for largervalues
inapplicable

seems

character
oscillatory
method

^TT + \v7r

"

of

sin{a;+

"

is effective for these

v6 +

hO)

largervalues will

s investigations
Schafheitlin
of the
largeorder.
unrestrictedly

We

shall

now

that, if i^

prove

"

J^ {x)cos
which

exceed

|, those
a

"

assumes

be

values.
integer

"

^tt;
explained.
0

to

of cylinderfunctionsof

lie in the intervals


(2i/+ 1) {2v + 3)/7r
"

from

of the cylinder
function

zeros

Y^ {x)sin

+ \'ir,
{niTT a-\-\v'Tr
nnr

ivhere

zeros

now

of the

account

on

6 increases

as

15"35.

of

+ f tt)
|i'7r

15-35]

ZEROS

considered
his

used

method

The

the

is

this result is due

case

lengthyand

obscure

mean-value

by

the

theorem

use

order
integral

he

only,and

of arguments

when

the

sin
Y^ (./")

a,

equivalent
of that
use
explicit

write
precedinganalysis,
J

9ov(^)

(^)COS

"

that

so

Now

9
cot'-''+^

6 increases from

with

that 62 is

tan

arc

$2 and then decreases

0 to

j-.

nearlyequal to ^irwhen

is

largecompared

V.

Now

that

suppose
^iTr

and

^o

\'^^'

cos-'-^i
e

6 increases from

as

62 to ^tt,where

It will be observed

dd

e-2-"cot8ill
creases

as

^^

(. + f) r (i)J0

"

then

choose

"

"c

lies between

It is easy to

h)

hTT {v

"

and

7H7r

"

"

-f-a

(f + I) ^1

"

?/i7r.

verifythat
"2v

-I

2v

TT

TT

2vTS'2'^ '^2^7^that 61 is

positive
angleless

arc

suppose

2'

0..,
providedthat

than

tan

"

"

4i/ + 6

We

h) + f tt,

^tt{v

d^ so that
a;

so

Schafheitlin*;but

to

obviouslydesirable.

in the

As

493

FUNCTIONS

of functions of the firstkind and of

of the second

use

theorem

obtain

to

reasoningis made

to the

BESSEL

OF

that

now

+ 3)/7r,
a;"(2i'+ l)(2j^
so

that ^1 is

less
certainly

Then, by the second


0 and

^such

than

value

mean-

theorem,there

a-z^^

"^'
=

fcot-+"^

e-=xcot.,]

i^)

number

^0,between

^^^^^,,

A |cosj^rJ^a-^^^-^[

^0
{cos(m7r
cos"-^^
+

exists

that

cos-^^sin(^' +
i".

6.2.

e,
{cot-+^
^
*

6-cotM

Journal

cosjx

a-ve,-

filr Math,

cos''+^(^o

6'i

cxxir.

(1900),pp.

299"321.

494

OF

THEORY

Now

{a:+ a-v6

cos

stationarywhen
fraction is equalto

s,m{x

is

-\-a

6
|^)/cos''+^
\6)=^0, and for
-

v6 +

"

such

values

0 the

of

6.
l/cos""^

cos("

Hence

a-i^6?o- 2^o)
-tt-?,

the greatestvalue
numerically

exceed

cannot

[CHAP. XV

FUNCTIONS

of 6,

function

qua

BESSEL

of

6 in
l/cos""*

the interval

(0,^i),

therefore

and

cos{x a-ve-W\
(cos(m^+^0
008"+*
+

"

COS-'+^^o

^1

v9 -\-\6)
is the signof (" 1)'"when
6
Therefore, since the signof sin {x + a"
lies between
6^ and ^tt,we see that, for the values of x under consideration,
sgn W^,(a;) sgn (- 1)'".
=

Hence, when

exceeds

has
{2v + 1) {2v + S)/7r,'^"(a;)

no

in intervals of

zeros

the type

(lUTT
"

and

the

so

^TT,mir

"

(x)which

of ^^

onlyzeros

^rTT

^v7r+ ^tt),

"a

exceed

lie in intervals
+ 3)/7r
(2v + 1) (2i/

of the type

(niTT
"

and

this reduces

hi'ir+

Schafheitlin's

to

reader will observe

The

the smaller

of

zeros

are

^"(^) when

Backer's

result of

discovered

by

from

obtained

zeros

consideration

and

the distance between


zero
positive

consecutive

zeros

0 and

givesno

is

integer.

an

information

it will be apparent in

ofW^ix).

character

questionis that ^^ {x)has

theorem

To

be

f vr),

concerning
" 15"8 that
than {2v + 1) {2v + d)/7r,
and that
concerningthem by using Debye's

large;
less

zeros

different
slightly

Bocher
in

is

hvir

result* when

tJteoremfon the

The

smallest

of

15"36.

"

that this theorem

largenumber
information
can
intei-esting
integrals.
there

|7r,mir

from

those

just established was


of the integral
formula " 11-41 (16).
infinite number
of positive
an
zeros,
does not exceed 2;owhere j^is the

of Jq {x)-

establish this result,write

j^

f'^#0(tir)
d"^
=

0, z=2o

in

" 11-41 (17),and

TT-^^o
{Z) J, (jo)
=

then

0.

.'0

Hence

be

one-signedas ^ increases from 0 to tt, i.e.as


increases from Z
j,^to Z + j^; and so Woi'^) must vanish for at least one
value J of
in the interval {Z "Jq,Z+j^). Since Z is an arbitrary
positive
number
Bocher's theorem
evident.
is now
(greaterthan Jo),
cannot
'"0'o
("sr)

ot

"

or

Schafheitlin

lie in the

t Bulletin

X Cf.

gives(2c + 3) (2v + 5)/7ras

the lower

limit of

specifiedintervals.
Modern

American

Math.

Soe.

Analysis, " 3-G3.

v.

(1899),pp.

385"388.

the values

of

for which

the

zeros

15-36, 15-4]
[Note. By

ZEROS

of Green's

form

OF

BESSEL

theorem,
dv

""

w,

solutions

two

are

By taking

the

at

angle

subtending an
Hence

circle.

the

exceed

j^

distance

second

differential

from

/u^ Weber*

n6 satisfies the

(r)cos

is drawn

with

centre

origin,/In (?')must
such
the

normal.

deduced

that

y^.

Wn

j^

the

along the

that

the

the

on

vanish

consecutive

originof

conditions
requisite
axis of

somewhere

zeros

smallest

at

are

on
a

of them

and
the

distance
does

not

^|

-fcosec

results

the

be .t-+^-

to

curve

(?"(?")
are

of

positive

|l

These

at
tt/?*

zeros

2/n, and

apart less than

continuous

differentiation

indicates
l^l'hv

circle of radius

less than

with

")

circle of radius

any

as,

;r~

cv

dy'-

this result that since

origin,if

s, and

curve

on

"^-\-^-\-u

of

and the
J^{s]{x^-\-y'^)\

inferred from

Bochor

except

at least twice

vanish

must

as

dv

ox''

coefficients inside the closed

du

i u^^

J
where

495

FUNCTIONS

are

of interest

on

account

of the extreme

of
simplicity

the

methods

used

them.]

to prove

15*4.

On

the nuniber

of zeros

of J^{z) in

assignedstripof the z-plane.

an

give the expressionfor Jv{z) as a Weierstrassian product,


and then developexpressions
involvingquotientsof Bessel functions in the
fractions ; but as a preliminary
it is convenient
form of partial
to prove
the
situation
indication
the
which
of
those
to
as
theorem,
zeros
givessome
following
In this investigation
it is not supposed
of 7^(2) which are of largemodulus.
to suppose
that V is restricted to be a real number, though it is convenient
is real the results of " 15"2 to some
that V is not a negativeinteger.When
which will now
be proved.
extent
take the placeof the theorem
shall next

We

Let

C be the

whose

contour
rectangular

"iB
" iB + 1 7ri7 (z'),
where

i? is

We

Sinde

unr

^ i^TT

are

^ tt,

number.
positive
(large)

shall shew
of z""

zeros

vertices

inside
J^,(z)

7/^

is

number

function
integral

J^ (z) is an

z~^

sufficiently
largeintegerthe
is precisely
equalf to ni.

that when

of z, the number

of its

zeros

of

inside

C'is
1

J_^^^^
I dr\og{w-''JAw)}^^^_lf
dw

ZTrlJc
*

Math.

t When
of

zeros

pairof

on
zeros

Ann.
t'

i.

is

the

(18(59),
p.
real

ZTTlJcJ^iW)

10.

negative number

ima^^inary axis

is to be reckoned

;
as

(and

for certain

in sueVi circumstances
a

single zero.

complex

the contour

values

of

v) there

may

be

C lias to be indented, and

pairs
each

496
We

consider

now

BESSEL

It is first to be observed

(")%'"^-i--J-)

(1+

yirwj

iT," (w)
?7i_
^

{w) "and

since the

Now,

that

C,
If^i)(w)

where

[CHAP. XV

FUNCTIONS

four sides of C in turn.

the

all the sides of

on

OF

THEORY

772,
v

^1

.
.

(w)},
N

/J

{1+^,_^(^y)},
(''-^Ye-'-('"'-i--^-)

(w) are

(1/w)when

integrandis an

odd

jw |is large.

function*, we

have,

as

-^

x
,

^iB+lTriI(v)
^

Next

take the

integral
along the

27r j iB

"

";".

27ri j iB+mn+iun+^n

upper

idiu

horizontal

^il{v).

side of G\

this is

equalto

dlU

J^v{w)

{1

imI{v)

V2,u{w)

[1 +

div
(e^"0]

1
m.r

27r

Ki^(.)

+0(l/i?)
^"^"^.^-^^^/

ivr+-^log

as

B^'oo

so

Similarlythe integralalong the


the limit of the integral
alongthe

Lastlywe
this

we

have

consider

to

lower

side tends

three sides

the

to

the

same

considered

now

integralalongthe

value, and
is m-^

^v + l.

fourth side,and

to do

the difference
first investigate

^"-tan("-i..-i.),
which, when

is equalto
\iu\is large,

2ot
rtB+mn+^vn+iTr
riB+mn+^vn+iiT

Now

tan

(w" ^VTT
"

I it)diu

0,

J "iB+mn+\viT-{-\iT

and

so

_1_ n^_

27riJ-11
]^

riB-\-mit-\-iv7,-\-\n
(2v + 1
/'I
r^'^"-^"
dn,
\^"l+0(1)1

27rij_i

+ l)+0(l/m).
i-(2i.

Hence
*

the limit of the

Allowance

is made

for the

round
integral

the whole

in
indentations,
justspecified,

is m-\- 0(l/m).
rectangle
the first step of the

followinganalysis.

15-41]

ZEROS

OF

497

FUNCTIONS

which
that the expression
we
can
ensure
sufHciently
large,
round
the
less than 1 ; and since the integral
rectangle
(l/m)is numerically
it is equalto m.
be an integer,

If
is 0
must

take

we

of

is to say, the number


and the line on which
That

is

BESSEL

of

zeros

2""/^ (z)between

the

imaginaryaxis

exactlym.
Note.

that the
inside

The

approximate

bounded

quoted for the functions of the third kind shew


largeimaginary port; and so all the zeros of J^ (z)lie
parallelto the real axis and at distances from it which are

formulae

have

largezeros
stripwhose sides are
when
j |is bounded.
cannot

i'

15 "41.
is

It

The

of Jy,{z)as
expression

infinite
product.

cm

a
product of
express J^ {z) as
type, each factor vanishingat one of the

possibleto

Weierstrassian

'

simple factors
of

zeros

'

of

Jv{z). In

first to express the


express Jv{z) in this form, it is convenient
series
derivate
of
rational
fractions by Mittagof
z~^Jv{z)as a
logarithmic

order

to

Leffler'stheorem*.
taken
be "jy,i,
to
are
z~''J^(z)
"ji._2"
"jf,3, wheref
the
values
of j.^,yV^o,
R (j"',n)
" 0 and
\^\R
{jv,2)
I
\^\R
(As)
^
R(j",i)
1
j;,3, being all unequal ("15"21).We draw a (large)rectangleD, whose
and we suppose that "ju,"i
vertices are " A " iB, where A and B are positive,
inside
which
the
rank
the zeros
of highest
are
are
rectangle.
The

of

zeros

"""

"""

...

We

consider

now

27riJ J)
Avhere
not

pointinside
negativeinteger.
z

The
"Jv,

is any

the

{w

"

Jv{w)

z)

other
rectangle,

integrandinside

the

than

zero

onlypolesof

the

residue at

and the residues at


is J^+i{z)jJt,{z)

are
rectangle

of

z,

and
Ji,(w),

"jv,i,
tjy,2,

""",

III-

The

[Z + Jv,n

since -//{z)
=

J^^^(z)when

"

z=

"

"j^,
"

are

Jv,n)

j^,nby " 3*2.


,

It follows that

Jv{z)

n=l[z"Jv,n

Jv,n] n=l[z+Jy,n Jv,n

.^^A^^hw,
27ri}J)
*

Acta

Soc. Scient.

t If iv ( "j^ ,J
W.

B.

F.

Fennicae,

0 for any

xi.

value

(1880),pp.
of u,

we

273"293.

choose

j^ ^^

{lu z)' Jv {w)


"

Cf. Modern
to have

its

is

Analysis, " 7-4.

imaginary part positive.


32

498

OF

THEORY

We

take A

We

also to be

largethat

so

be

" 15"4, to

Mir -\-R{\v +

to be at least

take M

function

odd

an

largethat

so

take

can

we

less than, say,

| or greater than

these limits,I

(lu)is

in absolute

is

value.

Ji'ir"
Jn-)I

and

the

when
is not

does not
expression
a
near
zero
arbitrarily

lie within
of

J^ (w) ;

is bounded
asymptoticexpansionoi ^1'2l,J^+i{iu)IJy(w)
within
this
strip.
rectangle

the part of the

the

whenever

2 ; and

bounded

the

that,from

so

it is sufficient to consider

w,

D.

the

jg2l(W"
is* less than

on

integer; and then


positive
is possible
by " 15'4, and
m, which
defined in
functions -ij^i^iv),
Vv,-2{w),

where
\)'ir,

Z^+i{iv)/J^
(w) is bounded

Then

of

of values which

bounded

of D.

right-handhalf

we

is

this function

Since

suitable sequences
be taken
to be

limit, Jy+i(w)/J^(w)can

without

increase

and

that,by givingA

shew

next

[CHAP. XV

FUNCTIONS

BESSEL

is to say
D as
the rectangle
That

J^,+i
(w) is bounded
(tu)/J^
B and

tend

the whole

on

of the

on

perimeterof

infinity.

to

Hence

J"+i{w)
2iri]
]dw(w-z)
z) JJy{w)
Aw)
i)w{w
1

"

therefore t

and

"fV-^/

When

('Ju+.it),]

Jv,n)

{.^"^ Jv,n

Jv,n)

find that

we
integrate,

we

("^ Jv,n

[/

"

\_^f

\] ^

if,

7-

and

hence
z

This

is the

The

formula

This
when

j/

The

of J^ {z)in
expression

formula

may

was

the

also be written

assumed

form.
specified
in the modified

by Euler,Acta

Acad.

Petrop.v.

form

pars 1,

0, aud subsequently
by various writers for other values of

analysisof

this section

is due

(1781)[1784],
p. 170,

r;

cf.

"" 15-5,15-51.

in substance

to Graf and Gubler,Einleitungin die


it was
Funktionen, i. (Bern, 1898), pp. 123
130, and
given
explicitly
by Kapteyn,Monatshefte
fur Math, und Phys. xiv. (1903),pp. 281"282.

Theorie

der

Bessel'schen

"

'

Because

t If
on

the

we

take the

rightconverge

^
"

1 +

rectangleto

have

separately.

its vertices

2at

A"iB,

-A'^iB,

we

see

that

the two

series

500

THEORY

while the residue


2t

OF

[CHAP.XV

FUNCTIONS

BESSEL

is
at_y^,,j

J^ (j,,n
X) 7, (j\n)Jv (jv,n
x)/[J,'
(",")(^'-j\n)]
^"^''v (",n^)'Jv

"

'o

ijv,x) \y
n

\T^

"

^v~\Jv,n)\2

"

{"'^
v\Jv,n)

'

T
^

\_

\Jv,n)

"

V( i
\V'(i
v\Jv,7i} -t u \Jv,nJ)

J'v,n)

'

J.'^(j.,
'Jrj,,
n)(2^ j\ n)
-

and

summing

on

the residues

84

"

functions

as

well

expansion.

this

93 ; Carslaw

Legendre functions

obtain the stated

at once

(w) by
expansion,obtained by replacingJy {xiv)jJy
Math. Soc. (2)xvi.
see
Carslaw, Proc. London
integral,

For a generalisation
of
"^i,
(w) in the contour
(.rw)/^^

(1917),pp.

we

also constructed

has

to certain
appropriate

similar series which

some

Bessel functions,and

as

these series represent the

problems.See also Beltrami,Lomhardo


physical

II. (1899),
(2)XIII. (1880),p. 336; and Lorenz, Oetivres Scientifiqties,
p.

Elders

15* 5.

investigation
ofthe zeros

Green's

Rendiconti,

506.

ofJ^{2\/z).

the smallest
method
of calculating
ingenious
devised by Euler*,and applied
by him to determine
An

of

contain

zeros

of

function

the three smallest

was

zeros

Jo (2^z).
If the

then

zeros

arrangedin ascendingorderfof magnitudebe

Qj, Wg, Cs,

"""

"

by "15-41,
j"(2V^)=n
11=1

(i-f).
"n/

this formula
been stated ("15"41),
already
then
is differentiated logarithmically,
As has

n=l

CC

that |^^ |"


provided

Oj ; and

by Euler; if it

="

S
-;i-logJ'o(2V^)=
dz
"

assumed

was

an

yVd

00

the last series is then

absolutely
convergent.

00

Put

l/a,i"*+^
a-,"+i and changethe
=

order of the summations

; then

d
az

on
ReplaceJq{2\Jz)

,"

each side

z'^

Math.

Acta

Acad.

XXXIII.

t From

Petrop.v.

by

X2+]^2
22
*

12.2^3-'^'"''

pars 1, (1781)[1784],pp. 170 et seq.


365 should also be consulted.

(1846),pp. 363

paper

"

" 15*25 it follows that the

zeros

are

and unequal.
positive

by Stern,Journal fiiv

15-5]

OF

ZEROS

of

501

FUNCTIONS

the

multiplyout
powers

BESSEL

and equate coefficients of the various


product on the right,
the identity;
thus obtain the system* of equations
we

in

o"3

j2

"1

CTi ,

o"5

2580
~

HS'Toci

^4 +

^s

"

i 0-2

0-3 +

0-1,

"

i^n,

o"2 +

"

-rii:^4

0-0

"

jo-a

0-5+

"

aV^i

+ sfoo-i.
aVo"2

1_

0"4

tO-."

5T0

3 0

1 4 4 0 U

"

'

whence

1, o-o

0-1 =

Since

and

"

"

Oi

"

Wo

|,0-3

ii

the

from
By extrapolating

Euler inferred that a^

"!

473

i%*#,

that

Table
following

1 "445795, whence

2Vai

value

2-404824.

writing

for Aj,

S
and

cr",/o-,"+i

"

!/"! 0-691661,
By adoptingthis

yVij, 0-6

o-"r^"""

so

_UL

^
=
cTg

^i,

it is evident

"

Oo

4, cr,

l/a,i"'o-";,
=

usingthe inequalities

then

" o-'in/oi;,
" o-',",cr',n+i
l/ao'"

Euler

deduced

be

results should

These

a:;=18'63,by carryingthe

that

hence

7-6658, and

ci!o=

stage further.

process

that

ai

7-6178, 03=18-72,

02

American

Peirce,Btdletm

and

of Willsou

Tables

derived froEuihe

1-445796,

the values

with

compared

Math.

Soo.

in.

(1898),

pp. 153"155.
The

of

value

accordingto

quartic obtained
magnitude
*

This

given by

is

of the sixth term

system

is

t Mem.

de VAcad.
Camb.

obvious

an

R.

Phil.

Analytiquede

Poissou+

this result

I
by equating to

X Proc.
Theorie

is

oj

Freeman

des

Soc.
la

Sci.
iii.

zero

the

first five terms

quite sutficient

extension
xii.

(misprintedas 1-46796491);
by Largeteau for Poisson by solvingthe

1-446796491

as

calculated

was

to account

of Newton's

system

of the
for the
for

an

series for

J^ (2 Jz)

the

error.

algebi'aicequation.

(1883j,p. 330.

(1880),pp.

Chaleur, p. 310,

375

"

377.

footuote.

Of, Freeman's

translation

of Fourier's

La

502

THEORY

justdescribed
zero
positive

method

the smallest

Taking the

used

was

of

[CHAP. XV

FUNCTIONS

BESSEL

RayleigKsextension of Euler

15*51.
The

OF

fomnula.

independently
by Ray leigh*

late
to calcu-

t/"{z).

("15'41)

formula

and

writing
M

"

J'

o-^**"',

-r^^^
v,n

find,after Rayleigh,that

we

"'^
=

.(1)
=

fl-..(2)
=

-.

5r +

t"'

o-

29

smallest

and

3-831706.

a
as

"

4- 5)
(i;+ 3) (y + 4)(i/
(i;+ ly {v + 2)-'

positivezeros

Immediately

ll

'""'=

o-

The

afterwards

of J^ (z)and

Cayley f

by a process which he attributed


power
method
of solvingan equation.
Graefte's"
method

consists

in

the

of the

squares

equalitywith

of its

roots

it

J,but which

Encke

to

be

to

is more

2-404826

rapidlywhen r is
usuallyknown

_^

when
calculatingo-^*'')

given equationand forming from

deduced

Jj (z) are

be calculated
that o-^(''"
can

noticed

of 2

The

"=" =

O-

-.

is

power

of 2

by startingwith

equationseach of which has for


then rapidly tends to a
predecessor;and a-J'^^
a

of

sequence

the

its roots
ratio of

llj'^''y^i.

Cayleythus

found

to be
o-..'^)

4291^5 ^ 7640i/"+ 53752./3 + 185430i/2 + 31 1387i/ + 202738


+ l)8(,.
+ 2)i(v+ 3)2(r+ 4)2(i'
+ 5)(^+ 6)(v + 7)(v + 8)*
2i6(,.

It

observed

was

by Graf

and

Gubler||that

the value

of o-^t'')
can

easilybe

checked

by the

formula

22'-ii?,/(2;-)!,
,.j('-)
=

where

B^ is the ?-th BernouUian

number

this formula

is

an

evident

consequence

of the

equation

Extensions
have

been

The

of

made

by

Froc. London

by Lamb,

smallest

tabulated

of these results to the

some

of

zero

for various
"/;,(s),

Airey,Phil.

zeros

Math.

of zJi,'
{z)+ hJ^ (z),where

8oc.

values

of

(1884),p. 273.

xv.

h is a constant,

between

Mag. (6)XLi. (1921),pp. 200"205,

0 and

with

1, has recentlybeen

the aid of the

Kayleigh-

Cayley formulae.
*

London

Math.

Soc.

t Proc. London

Proc.

Math.

Soc.

Journal
:;:

" Die

fur

Math.

Auflosung

IIEinleitung

in

xxii.

der

hoheren

die

Theorie

(187i),pp. 119"1-24.
(1874),pp. 123"124.
(1841),pp. 193"248.
v.

v.

numerischen
der

Papers, i. (1899),pp. 190"195.]


[Scientific
Papers, ix. (1896),pp. 19"20.]
[Collected

Gleichungen (Zurich,1837).

BesseVschen

Funktionen,

i.

(Bern,1898),pp.

130

"

131.

15-52]

15-51,

ZEROS

OF

BESSEL

503

FUNCTIONS

of the rth powers


of the roots of an
[Note. The proeedui-eof calculatingthe sum
obtain
the
numerical
value
of its largestroot seems
to be due to
equation in order to
Waring, Meditationes Analyticae (Cambridge, 1776), p. 311 ; other writers who were
before GraefFe are
Euler (cf." 15-5); Dandelin*, Mem.
de
acquainted with such a method
VAcad. R. des Sci. de Bruxelles,ill. (1826), p. 48; Lobatschevsky*, Algebra,or Calculus
of Finites (Kazan, 1834), " 257.]

The

15*52.

of Jq{x).

largezeros

the large zeros


of cylinder
calculating
functions (when the order v is not too large)
is,in substance, due to Stokes-f*,
writers
have, to some
extent, improved on his analysis.
though subsequent
The

effective method

most

Stokes' method
whose

with

the

illustrated by
sufficiently
equation

will be

notation

"

of

7"3.

expansionsof P(a;,0) and

It will be

examplej Jo(rf'),

own

Q{x, 0)

^,

remembered

that

the

asymptotic

are

1.9
",

his

of the

the roots

are

zeros

of

1.9.25.49

25
""

3 !{%xf

Wbx

Q (x,0) is negative
sufficiently
largevalues of x, P (x,0) is positive,
function of x.
the quotientQ(x, 0)/P(x,0) is a negativeincreasing!
and
function
The
cot(^ ^tt) is a decreasingfunction which vanishes when
and so it is obvious from a graph of cot(*' |^7r)
that there
x=n7r
^'7r,
exists a positive
one
zero
integeriV such that when n " N, Jo(x) has precisely
For

"

"

"

in each

of the intervals

the left-hand

from

end

(mr

^vr,mr

"

of the interval tends

Again,if ",-, v,- denote

the

that the distance

^ir),and
to

as

zero

of P

(r + l)th terms

7^

oo

-^

{x,0)

and

of the

zero

Q(x, 0) we

may

write
m-l

6 and

where
*

references

two

Phil.

m~l

du,"

^^. +

certain functions

0^ are

these

owe

t Camb.

(x,0)

Trans,

ix.

to Professor

(1856),pp.

of

Q (x,0)
and

which

V, +

d,v"^,

lie between

0 and

1.

Whittaker.

[Math, and Phys. Pajyers,u.

182"184.

(1883),pp. 350"

353.]

f Stokes
positionof
" The

also considered
the dark

reader

may

bands

for
Airy'sintegral("6-4)and Jj (.f),
seen

the purpose

of

the
investigating

in artificial rainbows.

verify,
by " 3-63, that

its derivate

is

|l_p2_Q2}/p2,
where

P, Q stand

positive.

for

P{x,0), Q(x,0);

and,

by

the

asymptotic expansions,this

is

ultimately

504

THEORY

the

consider

Now

OF

BESSEL

[CHAP. XV

FUNCTIONS

equation
m-l

2
cot (os

Itt)

"

1
r

values,are

lie between

which

numbers

any

0Um

Uy +

it is temporarily
supposedthat 6 and

in which

d^V.n

V,. +

^^^

^j,instead of havingtheir actual

0 and

1.

under consideration involves no functions more


equationnow
cated
complifunctions.
If x were
than trigonometrical
supposed complex,there
in the a;-plane
each of which enclosed one
would be a number
of contours
of
which
exceeded
and
the
the pointsn-rr
cot
modulus
of
the
on
| (a; \'jt)\
^tt
the
on
right.
quotient
The

"

"

Biirmann's

By

the part of the


in

in which

descendingpowers

readilyperceivedthat

independentof 6 and
Now

the

0 and

as

$1

given their

are

the
That
to

WTT

"

independentof

the

the

first
r

"

equationin

of the

iii, we

but

dejDendon

after the mth

proved that

may

the

cc
.

the

bound

Hence, when
zero

one

zero

where

yjr^0

as

then
,

1
,

tan

33

3417

vr

i-^

"
"

Cf. Modern

sideration,
con-

(nearlyequal

is

Q(x,0)

under

0^

of the

is still 0(n~'^'"~^).

there exists

zero

actually

function of 6 and

the first few of the coefficients f,..If


.

panded
ex-

6 and 6^;

are

be written

expansionof
asymjitotic

to calculate

be

write

It remains

inside

the form

coefficients

n-*
as
questionis 0(?i~-"*~^)
actual values which
they have at

is to say, it has been


and its value may
^ir),

Hence

of the

the root

in

of the terms

sum

root
can

1; and it is clear that the upper

0 and

function

of the

one

this root

"

after the 7nth is a bounded

of the terms

sum

have

^tt.

that, when

6^,so

6^ vary between

modulus
and

nir

"

expansionfor

an

equationwould
mr
^tt,and

surrounds
of

the coefficients /^ (^,^i)are

it is

and

the modified

which

contour

thus obtain

We

theorem

I-

Analysis, " 7-31.

15-53]

ZEROS

1073

25

therefore the

and

be solved

equationto

the form

assumes

1073

25

1
,

505

FUNCTIONS

BESSEL

OF

^"-^"^-^^^^8^--384^+5120.^^--The

result of

the
reverting

series is

1
,

The

15"53.

from

and

not

are

consideration

seems

F" (z)sin

"

series for P (z,v) and

a)

(with large
though the

prove the existence of such zeros


fact that they may be calculated as

the

convergent, because

Q (z,v) were

method

r7j7
r {z, V)

of the precedingsection only in


investigation

the

by Stokes'

to

unnecessary

positivereal parts)or

iTT +

"

",

be calculated

real,may
necessarily
of the equation
cot(^ -^VTT

It

1.

n=

cylinderfunction,

of any

largezeros

Jy (z)cos
v

of J^,(x),to at least five

zeros

of cylinder
functions.

that the

see

the

the smallest zero, for which

largezeros

It is easy to

where

all
adequatefor calculating

placesof decimals, except

|7r)s

(wtt ^ Tr)^ 15360 (nvr

384

S{mr-lTr)
This series is

3779

31

proof differs from

the

tedious details.

cylinderfunction of any given order


but the subsequent
calculated after the manner
of Stokes by McMahon*;
was
have
made
memoirs
of Kalahnef and Marshall :|:
the investigation
more
simple
and have carried the approximationa stage further with no
ture
greater expendiThe

for
expression

of work

the

i/
the

Annals

" two

i/^ P (z,v), M sin y^


=

il/

-*

and

i/r
^-

of Math.

ix.

F^ {z)sin

(1895),pp.

23"2.5

[ j Jf
"

; see

also

cos

" Cf.

as

-^

by the
i/r,

of Math.

nnd

xi.

"

^vit

"

-\-a
\'jr

"

Airey,Proc. Plujs.Soc. 1911,

Phijs.liv. (1907),pp. 55"86.


(1910),
(2)
pp. 153"160.
Phil.
xix.
(1910),pp. 228"249.
Nicholson,
Mat;. (6)

Math,
f Zeitschriftfiir
Annals

and

-Q{z, v),

{z

225"232.

of ^, called M

functions

clear that

J^,{z)cos
*

cos

define

we

understandingthat

It is then

in the calculation.

FollowingMarshall
equations

on

of

largezeros

y^).

pp.

219"22-1,

506

THEORY

OF

BESSEL

[CHAP. XV

FUNCTIONS

Again

and, when

j^^ ^Y

tan

arc

differentiate this

we

^tt

^vtt

"

and
equation,

dy^

i/r,

"

"

| 3*63 (3),we

use

find that

2/(7rg)
_

that,by " 7"51,

so

dyjr

'-dz
TO

When

the

rightis expanded

far

as

the term

as

involving

find that

we
l/z^,

the

expressionon

2m^2?n

+ 1073)
(;a-l)(^''-114yL6

(M-l)(At-25)

^-1

dz

2^z^

2V

2""z''

(fi-l)(5fji'1dS5/m'+ 54703/^ 375733)


-

215^

in this

equationfi

in

written

has been

placeof

Aiv^for

on
brevity.It follows,

that
integration,

375733)

+ 54703/^
{^l 1){bij? 1535yLt^
-

"^"""'

7.2""2^
and

so

the

equationto

be solved is*

z-mr-lv'Tr
+ ly
(/i,

If /3

a,

"

....

is

the result of reversion

(/i,-l)(7/^-31) (ya-1) (83/^2 982^

^l-l

^~/3

1)TT

(y^-l)(/^-25)
g-^f^g

At-1

l'7r-\-a^--^^

3779)

"

2^/3

15

2'y8^

(/i 1)(694V

2i"/S'

6277237)

153855/^2+ 1585743/z
-

105.215/3^
Therefore

the

large zeros
asymptoticexpansion

of /^

4i."

(?i+ 1^

This

1797, but
Lvii.

"

^)77
"

(2)cos

"

Fj,(^)sin

+ li/
{(/I

(1902),p. 19.]

the

given by

are

(4i/--l)(28i^--31)

"

1) TT

a}

384

equation (in the case v l) was given by Gauss in


clue is given concerning the method
by which

no

[{n+ ^v

his notebook
he

obtained

\'rr) a]'
-

with
it.

the date

Oct.

[Cf. Math.

16,

Ann.

508

OF

THEORY

Since

J^(j) 0, it

(15),when

"5'11

hence, from

dj

(2)

This formula
as

shews

2i/

^^no?.

from

B. des Sci. de

We

"

'"'

it

stated without
established

was

extend

Math. Ann. x. (1876),


proofby Schlafli,
p. 137 ; aud
different manner
a
by Gegenbauer*, Mem. de la
3, in the case of the smallest zero of Jy{x).

alreadyobtained

the results

{z)sin

^^ (z) J^ {z)cos a"Y^


=

is an

unrestricted real variable,


and

The

extended

Any

'positivezero, c, of ^^

is

theorem

of the real variable


To

crease
of J^ ix)in-

the positive
zeros
positive,

is

in

Liege,(3)ii. (1900),no.

proceedto

where

is increased.

Equation (2) was


deduction

jJ\+,{j)Jo

that tuhen

is not

so
J^+i(j)t^O,
long as j

"

(v)" 0,

dv

[chap. XV

FUNCTIONS

follows that "//(j)

zero, and

BESSEL

follows

of

zeros
positive

a,

(i.e.
independentof v).

is constant

as
(z)is definable

continuous

function
increasing

v.

this theorem

prove

as

to the

the
Soc.

observe

w-e

that

is

function

of

such

that

tan

arc

J Ac)
is constant,

so

that

dc
dv

^
dz

-^j- '\
\J^{z))_

tan

arc

and

arc

s-

tan

0,

cv

z=c

therefoj-e

TTC

Hence, by

dv

a/,.(^)'
J,.{z)^^-Y^{z)

0.

dv

we
"13-73(2),

have

dc

(3)

-r

2c

dv

Since

the

function
A

j0

^0 (2c sinh t)e-^^Ht.

this
integrandis positive,

of

formula

shews

that

is

an

increasing

v.

less

generaltheorem, namely that,if c is a zero which is greater than


the order v (supposed
function of v, has been
then c is an increasing
jjositive),
with the aid of very elaborate analysis.
provedby Schafheitlinf
It will be observed
when

from

the definition of Y^

tends to ant/ negativevalue which


sin (a

The

reader

vitiated by his
t Berliner
(jimg,

XVI.

use

should

note

of Eudski's

that

the

erroneous

272"

279.

tends

to

zero

only

equation

0.

analysis in the latter part of Gegenbauer's memoir


results

Sitzungsberichte,v. (1906),pp. 82

(1907),pp.

satisfies the

vtt)

"

(z) that

"

is

("15-1).
93 ; Jahresbericht

der Deutschen

Math.

Vereini-

15-6]

ZEROS

(3) shews that,when

It should be noticed that

509

FUNCTIONS

BESSEL

OF

is taken to be

complex

real part,then c is an
number
and c is a (complex)
number, with a jiosifive
of '^^(z)vary continuously,
the zems
function of v ; and so, as v varies,
analytic
into
theycan onlycome
i.e.
when
function
of
v,
analytic

existence

and

It follows that the

when
disappear

or

fails to be

an

0.

of 9^^(2)are

zeros
positive

derived from those of

'^^(z)

zero
by process of continuous variation as v varies,except that one positive
whenever v passes throughone of the specified
values.
disappears
negative
a

If
to

we

choose

now

variation of

v,

and

^VTT

when

has varied

the formula

firstkind of order

"

r-

-J

the

zeros
positive

so

are

...

as arranged
regarded

that it

"

to

Macdonald, Proc. Londoii Math. Soc.

him to the discussion of the

appliedby
exceeding 1

(1898),

xxix.

of Bessel functions of the

zeros

draw

we

branches

formula

"

584 ; it was

"

varies from

k is a

tjijeof argument is due

jjp..575

zeros

see

lies between
finally
(o/tt)A; and
k
have
and so
zeros
positive
integer,
disappeared,
zero.
justquotedgivesthe (n k)th.
positive
v

(o/tt)k "1, where


"

positive
zero,
magnitude.

If,however,

If

iir

"

nth

in order of

of

case

is available,
the
("15"58)

7177

This

in the

so

that, as

|
disappear
duringthe process of
which are so largethat the formula
we

zeros

"

of Stokes' type

givesthe

that* O^ccKtt,

so

1, no
(ck/tt)
exceeding

value

any

the

upwards towards

consists of
^t^x
iy) 0, it evidently

curve

from
starting
the

points
both
right,
on

the
x

number

negativehalf of the "-axis and


and y

without
increasing

limit

of

moving
each

on

branch.
If

we

coordinates
take any pointwith positive
line
downwards
terminated
and
a
right

line to the
that the

"^-j
(y)

curve

meets

and draw from it a


(v^t,yo)
the
it is evident
^-axis,
by

each of the lines in the

same

number

of

ofVJ'^(y,^),
qua function of v, which
of
the
number
of ?J?^^
exceed i\t is equalto
zeros
positive
(y)qua function of y
which are less than y^,.This is a generalisation
of a theorem due to Macdonald f,
function to be a function of the firstkind.
who took Vn
0 and the cylinder

points.It follows that the number

of

zeros

Fig.33

illustratesthe

generalshapeof

the

Jx (y) 0, the length

curves

of the sides of the squares being5 units. A much larger


and more
elaborate
who has also
diagramof the same character has been constructed by Gasser|,
constructed

0 is of

'Wxiy)
=

the

This does not

t See

diagz-amfor Yx{y) 0. The diagram for


corresponding
the same
character as that for J^ (y) 0, except that
general
=

lead to any

letter from

Macdonald

real loss of
to

X Bern Mittheilungen,1904, p. 13o.

tl

generality.

Carslaw, Froc. London

Math.

Soc.

(2)xiii.

p.
(1914),

239.

510

THEORY

OF

below the
portionsof the curves
isolated pointson the lines on which

axis of

the

1x

[chap. XV

FUNCTIONS

BESSEL

is

an

consist

odd

merely of
integer.

number

of

L\i_\L\L_V_
Fig.33.

of

[Note. The
then
"gV (z),

will find it

reader

("

-.

hence, if the variables

and

The

by

The
examined
transverse

membrane
Phil.

(6)XXXII.

The

mode

is

2"

Jo
real and

are

der Deutschen

Schafheitlin

15*61.

-,

signofdc'/dvhas also been

Jahresbericht
used

cosh
(c'2

"

c^-v

to
interesting

Math.

c'"

deduce

^
..2)

from

5^13-73 (3)that,if c'is a

/2c'sinh t)e

| |" 0, then
v

'

2'"

c' increases

zero

dt,
with

v.

discussed

(by more
elementarymethods) by Schafheitlin,
xvi.
279; but the analysis
(1907),pp. 272
Ve?'einigung,
"

extremely complicated.]

membrane.
problem of the vibrating
of increase

of the

by Rayleigh*with
vibrations of
is bounded
Mag. (6) xxi.

(1916),pp.

by

the aid of arguments

membrane

54fi

in the form

the lines ^

(1911), pp. 53"58

544"

of J^,(x) when

zeros

0 and

is increased

circular sector.

ir/v(where v

Papers, vi. (1920),pp. 1"5].


[Scientific

[Scientific
Papers,

vi.

has been

of
dependingon properties
of

(1920),pp. 444"446].

"

If the

^), and by
Cf. Phil.

Mag.

15-61, 15-7]
the circle

and

a,

pUicementin

OF

ZEROS

straiglit
edges of the membrane
vibration is proportional
to

if the

normal

sin vd
J^ (rp/c)
where
of the
the

of propagation
of
velocity

membrane

is fixed,the values

boundary is free to

If the circular

vibrations.

of

ap/c are
transversely
they are

move

fixed,the dis-

are

(pt+ e),

cos

is the

511

FUNCTIONS

BESSEL

the

zeros

the

zeros

of

boundary

J^ (x),while if

of JJ

(x).

constraints in the form of clampswhich gradually


introducing
diminish the effective angle of the sector is to increase v and to shorten the
function of r, and therefore
periodsof vibration,so that p is an increasing
is
That is to
and
c
are
(sincea
unaltered)ap/c an increasingfunction of
The

eff'ectof

v.

say, the

of -/,,
(x) and

zeros

By using arguments
number

JJ

v.

given proofsof a
chapterby analj'tical

of this character, Rayleigh has

which

of theorems

(w) increase with

proved elsewhere

are

in this

methods.

The

15-7.
The

of the function

zeros

and z
included),
qualitatively
by
From

of K,.(z).

zeros

K^ (z),where
domain

lies in the
Macdonald

i^

is

number
givenpositive
)arg^ |" |7r,have been

in which

(zero

studied

*.

the

of Bessel's integral,
generalisation
given in " 6"22,it is obvious
that K^ (z)has no positive
and it has been shewn further by Macdonald
zeros;
that Ky(z) has no zeros
for which
|arg2^ |^^tt. This may be proved at once
from a consideration of the integral
such
were
given in " 13'7l ; for,if z 7'e^'^
re~^"-would
be another zero ; but the
a
zero
(r " 0, ^tt" a " ^tt),then z
shews that
integral
=

"

I
jr

,"

rr

,"

f^

2a]

?'-COS

dv

/ ?'- \

j^exp |- -^\k. (-)

K. (r.'^)
/C (r.-'^)

"

"0,
which
If

and

so

is contrary to
is

equalto

K^(z)

Next

has

It may
even

the number

no

^tt,we

^tt and

be shewn

zeros
tt

or

for which

between

the phase of
(z)is negative,

R
"

|7rand

that the total number

integerfnearest
is

have

purelyimaginaryzeros.

study the

we

eitheHbetween

is the

hypothesis.

to

"

^,

zeros
"

lying

tt.

"

of

unless

in this

\\s,an

pairof quadrants
in which case
integer,

^.

"

In the first place,


there
*

Proc.

+ This

London

is not

are

no

zeros

Math.

Hoc.

the number

on

xxx.

the lines SiTgz=


(1899),pp.

165"179.

given by Macdonald.

"

tt,

unless

"

512

is

OF

THEORY

BESSEL

[chap. XV

FUNCTIONS

integer; for K^,(re*'"') e^""^ K^ {r)+ iriI" (r),and, if both the


have
to vanish, Ave must
are
imaginaryparts of this expression

an

real and

the

cos

VTT

the Wronskian

Since

vtt

K^ (r) +

/^ (?") 0.
=

vr

pair of functions on the


vanish simultaneously
unless

of the

(7r/7-)cos
Z'TT, they cannot

left of the
cosi'7r

equationsis

0.

change in phase of z" K^ {z) as z describes a contour


+ tt,
largeand small circles terminated by the lines arg ^
of
these
lines
terminated
circular
the
arcs.
by
parts

consider

Now

sin

K^ (r) 0,

the

of arcs of
consisting
with
the
together
(Cf Fig.15 of "7 -4.)

If the circles be called V and


that the number

it is evident
consideration

of

zeros

of

K^,{z) inside the contour,

z"

change in phase of

the

l/(27r)times

of

equationsbeing \z\ R and \z\ h,


K^,{z)in the pairof quadrantsunder

zeros

equalto the number

is

this is equalto

and

of

their

7,

z^

K^ {z)as

traverses

the contour.
Now

the

change in phase is

arg

[z"A% {z)]

[z^K^ {z)]

arg

S exp

S-^0, the

and

R^cc

Rexf)(-Tri)
+

[z''I{,(z)]

arg

Eexpiri

As

-rri

s.vg[z-K,{z)]

first two

6exp(-7ri)

tend

terms*

27r{u

to

l)

"

and

|^ |is largeor small on the contour,


K^ {z) ^-'-ig-V^TT), 2" 7C {z) 2"-! r {v)
when

because
respectively,
z"

-"

respectively
f.
last two

The

become

terms

cos

TT

lim

vir

K^ (r)+

TT

sin

VTT

K^ (?")is a positive
decreasingfunction of
increasingfunction, and so the last denominator
Now

and
nf^gative,

no

If therefore
when

is

"-X)

1"

if sin

zero

we

take

arc

tan

less
being numerically
signof cos VTT.

the
(coti/tt),

(z)is negativeand |arg z\"


v

has

/^ (r) is

one

vanish

to

-rr

positive

if sin

zero

when

value

ofK^{z)in

zeros

vtt

is

the

its limit
?--"-0,
inverse

function

signas

same

the

in which
pairof quadrants
:]:

is

-k +

"

while

assignedto the
rightanglesand having the

two

of

7^ (?")_

positive.

the inverse function

the total number

Hence

is

vtt

than

I^,(r)

tan

arc

tan

arc

"

(cotvtt),

TT

Iarg

This
z

I^

is evident

from

the consideration

that the

asymptoticexpansion of " 7 "23

TT.

t The

second

X The

two

zeros

of these
of A'o

approximate

{z)are

not

formulae

very

requires modificatiou

far from

the

points

when

l-29=fc0-44;(.

0.

is valid when

15-8]

OF

ZEROS

that
verify

the reader will find it easy to

and

which

is nearest

When

to

^ is

"

function with

(z)"0

consider the

If

write

we

portionof

K^ (z)

and

is

"

the

planefor

which

"

vr

arg^ ^ 27r.

have

we
^e^"*,

"

integer

even

polynomialin z multiplied
by a
and so the number
of
in the finite part of the plane,
is exactly
v
^.

Ky(z)
integer,

an

Next

is the

this number

^.

"

zeros

no

for which

zeros

513

FUNCTIONS

BESSEL

[F, (0
^7re-i-'

+ ^"(1 +

2^2-) J, (^)],

of zeros
the negativepart of the
lyingnear
sequence
which belong to this sequence
of largemodulus
The zeros

K^ (z) has

so

imaginaryaxis.
are
by
givenapproximately
tan

(f

the roots
hvTT

of the

equation

Itt) -{(1+
=

262""')
;

the rightor leftof the imaginary


it may be verified that theyare ultimately
on
axis in the ^-planeaccordingas cos- z/tt is less than or greater than I; i.e.

accordingas

differs from

exist when

does not

sequence
There

is a

15*8.

Zeros

the nearest

integerby

e'^'"'^= 1, i.e.when
"

of

corresponding
sequence

zeros

more

or

less than

is half of

odd

an

the line arg

near

J. The

"

integer.
f tt.

of unrestrictedly
largeor^der.
of Bessel functions

based mainly on integrals


of Poisson's type^
previousinvestigations,
of zeros
of properties
of Bessel functions,,
have resulted in the determination
is not unduly large.This is,of course, consistent with the
when the order
discussed in Chapter vii, are
fact that Hankel's
asymptoticexpansions,
small in comparisonwith the argument of
significant
onlywhen v^ is fairly
The

the Bessel function.


of " 8"31 afford representations
of functions
Debye'sintegrals
of largeorder suggests that these integrals
of
may form an effective means
of Bessel functions of large
the zeros
order; and this,in fact,proves
discussing
of the results which will be obtained
to be the case*.
Moreover, the majority
valid for functions of any positive
order,though theygain in importance
are
fact that

The

the increase of the order.

with

We

adopt the

shall

notation of

" 8'31, so thatf

gvi(tan/S"/3) /" 00
""

H,
where
the

"

sinh

"

complex variable
*

F.

lu
w

+ i tan

shall

use

the

ifi
e

/3(cosh'W

is chosen

+ ni"

Watson, Froc. Royal

t We
W.B.

{v sec /3)

so

Sue.

xciv.

symbols x

and

1),and the

"

that
a,
v

lyU"

dw,
in the

contour

is positive
on

(1918),pp.
sec

"^

planeof

it.

206.

when
/3indifferently

"

v.

33

514

OF

THEORY

If

iv,where

describes the contour,

so

and

BESSEL

du,

e-'^

both

both

increase

as
steadily

that

/;
are

real,u and

are

[CHAP. XV

FUNCTIONS

e'"^

dv

-e

positive.
regard/3 as variable,and define

Hence, if we

"'"''

arg
J

dw

?"
j3

with /3,it
positiveacute angle when /8 0, and to vary continuously
acute
will remain a positive
anglefor all values of y8 between 0 and ^tt; and,
by " 8"32,it cannot exceed J tt, since dwjdu ^ \/3.
moreover,
to be

anglewill

This
the

acute
positive
equation

-^

It is then

evident

be

called ;t^,and

i;(tan/S-yS) +

then

will be defined

by

^TT.

that

^,w(i,sec/3) ilWe^*,
=

is

J"l

where

(not zero);and
positive

J-^(a;)iHilcos^,

r,(") ittsin^.

it is clear that

the

which

values of ^

we

Jy (x)cos

^y {x)

If

V^

tan

arc

ot

equalto

of ^

as

an

a,

derived

v, are

number

x, when

of

the

rightangles.

remains

For

constant.

^^

"

t,/

from

t/,/
%
JJ"(x)
+ Yj"(x)

dx

increases

increases,^

-j-

J^ (x)

Hence,

odd

increases with

that ^

It is easy to shew
have
^

Yy (x)sin

a-

of ^^ (x),greater than

onlyzeros
make

and
steadily,

so, to

each

of the values

for which
^

cme
corresponds

and

(m

I)

zero
onlyone positive

77-

a,

of ^^

(x).

theorem

This is a
shall prove that % is also an increasing
function of x.
of a much
deeper character,since the result of " 18'74 is required

to prove

it ;

Next

we

we

have

thence
d

dx_(^_

(tan/3
-

2/(7rA-)
J^ (x)+ Fy-(x)

/3)
_

^/jx" v')
-

dx
From
lim X

and A

Hankel's
=

lini

asymptoticexpansionit is clear
[x

"

hvTr

arc

so

in which

dx

dx

the

on
expression

^/{x- v~)+ v

lTT

"

"

tan

1
( "^"'(^)
"

arc

"

"

"

the left is

"

"

'

X
.

that
cos

+ | tt
(vjx)

xtt,

acute
positive

angle.

(1/^)]

516

THEORY

smallest

The

15"81.

OF

BESSEL

[CHAP. XV

FUNCTIONS

of Jy {x) and Y" (oc).

zeros

("15'3)that /^ (x)and "" (x)have no zeros in the interval


and it is fairly
obvious from the asymptoticformulae
(0,v),when p is positive,
of the form v + o(v^)when
842
in
that
have
obtained
no
zeros
they
v is large.
|
It has been

notation

the smallest

at

As

yS increases

1 +

f TT,

occurs

solve the

the value of
is 0

(tan/3-^)

{v~^). The

In like manner,

value
of J^

of Y^

zero

formula

of

"

for

of

is

vix

is

root

values

the

zero

with

2"383447, and
approximately

1-855757

an

error

hence

the

find that

the

0(1).

-Q(f,i)/P(^, a

smallest

7.4X 0-931577

zero

Airey's formula

193.

/" {x) when

such

^y tan^'ySat

approximatelyf

has

value

x-v

For

-Q(^,J)/P(^,i),

0'847719,

we

(x)is

for the

(1917),p.
8-42

(v-i).

by solvingthe equation

smallest

the

^v tan^ ^+0

(x) is

z;

XXXIV.

PGi^tan^^,!)-

is the value of

tan(|-i7r)

The

equation

smallest

^tt, the

to

obtained

so

zero
positive

of which

{l/\/v)\

that

so

if we

smallest

tan(^-f7r)

which

angle.

"

Hence,

that,with the

expressionon the left increases from


the expression
the rightdecreases from 0'2679 to 0; the
on
for a value of /3 for which v (tan ^
^) lies between f tt

from

while
(1/^v),

smallest root

shew

J^,(x),

\v(tan/3-/8)-|7r

tan

negativeacute

of

zero

quoted in " 8'43

were

tangent denotes

the inverse

Hence,

and

which

asymptoticformulae
of "15'8,

The

where

seen

of the

of

Jy(x)

such
=

variables,it has

has

derived

was

0(1).

been

given by Airey, Phil. Mag. (6)


by using Debye's asymptotic expansion

that

0{vi)^o{v^).
not

been

proved that Debye'sexpansionis valid,

of t7"(x)
of the smallest zero
terms
gives the two dominant
of
is
smallest
for
the
not the
zero
result which
//
(x)
Airey gives
results
is
that
method
correct
The reason
J^ {v+ ^)
why Airey's
gives
of ( so long as f is o{v\ and in this expansion it is permissibleto

although Airey'smethod

and

the
correctly,

numerical

that of " 15-83.


as
expansiblein powers
substitute Debye's formulae

same

is

formula
between

for

for the smallest


0 and

Jv{v),// (v),JJ' (v),

zero

the smallest

of
zero

given by Airey. This zero


J^ (x),accordingto the value of v.

J^^ix)
of

was

may

he anywhere

15-81, 15-82]
It does not
section.

We

ZEROS

possibleto make further progress by the methods used in this


of Sturm
a
(which have
digressionto explainthe methods
then
givean
equation by various mathematicians),and we shall
result that the two expressions
to the fascinating
which, in this

be

to

seem

shall

make

now

been

applied to Bessel's
which leads
investigation
proved to

section,were
for the

smallest

become

when
negligible

zeros

be

0(1) are

of

J^ (x) and

which

Fy(.f)in

the

errors

obtained

0{v~^),i.e. the

are

errors

large.

is

that approximationsare
reality0 {v~^),so

in

concerningthe smallest zero of J^ (x)has been obtained


the formula
of " 5'43 by Gegenbauer, Wie7ier Sitzungsberichte,
cxi. (2a),
(1902),
p. 571;
of J^v
where 0 "
+
" 1, then the smallest zero
{^)is less than twice the smallest
be one-signed.]
of J^,(x),because,for the latter value of .v the integrandcannot

from

I'

zero

elementary result

An

[Note.
if ^

517

FUNCTIONS

BESSEL

OF

15'82.

Applications
of Sturm's

Various

writers have

'methods.

discussed

Bessel functions

of
properties

of
by means
of any linear
investigation

the

generalmethods invented by Sturm* for the


differential equationof the second order. The results hitherto obtained
in
this manner
of some
are
interest,though they are not of a particularly
deep
of them
have alreadybeen proved in this chapterby
character,and most
other methods.
The

givena

theorem

which

differential

is at the base of the

equationof

the second

in question
is that,
investigations

order in its normal

form

d"u
^

in which

the invariant

/ is

rapidlydo the solutions of


As

an

example

{ibid.pp. 174
v^

of

175) and

"

and
1-be positive

which

is nest

This

an

be

follows

is annihilated
"^'"^('("*')

does

at

once

by

the

that, when

"^

exceed

not

from

the

"

x^-c,

more
slioiitly

pp. 196"198,
in

abstruse

to the

efiect

has

"

~''

^^ "

"

Other

theorems

facts

that

the

function

result is due

that, if

i/-

"

American
Poi-ter,

to

?
"

and

if the

of like nature

due

are

to

(1901),pp. 333"340;

of

zeros

then

...

vii.

the

6-2

of magnitude are Ci, Co, C3,


alreadybeen proved in i^15-8 by another

(1897),pp. 205"213;

of

x'^

ascending oi-der

This

j-,

"

operator (" 4-3)

X'

"

,,

consideration

dxand

more

applicationof this result,we may take a theorem due to Sturm


Bourget, Ann. sci. de Vi^colenorm.
sup. in. (1866),p. 72, that, if
of
which
exceeds
zero
(,^")
"^^
s/{v^
^),then the zero of 'g'^{x)
any

greater than

result

then the greater the value of /, the


positive,
the equationoscillate as x increases.

e^

Journal

^^
i-Cn

of Math.

(1898),
s^{v--^),

xx.

{x),greater than
decreases

as

American

Math.

increases.

method.

Bocher, Bulletin
and

to

Gasser, Bern

Soc. in.

1904,
Mittheilungen,

pp. 92"135.
*

Journal

de

Math.

equations by Sturm's

i.

(1836),pp.

methods

(Cambridge,1912),pp. 163"195.

is

106

"

given in

ISO;
a

an

account

lecture

of recent

by Bocher, Proc.

researches
Int.

on

differential

Congress of Math.

i.

518

OF

THEORY

of Sturm's
Applications

15*83.

proceedto establish a
largeorder which are based
We

of

Let Wj

(x) and

such that,when

U2

{x) he

U2

let Ii and

(x) be

solutions

functions
of largeorder.

to

of results

the

on

(a)

U2

concerningcylinderfunctions
theorem
of Sturm's type :
following

of the equations

in the

(a), u( (a)

u^ (a),

in the interval a^x^h,

In be continuous

continuous

methods

number

[CHAP. XV

FUNCTIONS

a,
Ui

and

BESSEL

and

also let u/

(x) and

interval.

same

Then, if 1^ ^ /o throughoutthe interval*,\u2{x)


\exceeds \u^ (x)1 so longas
in
the
the
that the first
and
zero
interval,
so
zero
a
of tt^(x)
first

lies between

the interval is

ofui(x)in

Further,if w/ (a) has

IUt^{x)
Iin

the interval is

zero
leftof the first

the

on

the

signas u^ (a),the firstmaximum


pointof
the
m
aximum
left
of
first
pointof \ u^ix)|,and,

same

the

on

ofuoix).

moreover

iWi {x)i"

max

To

prove

the theorem

1^2(^)1-

max

f, observe that, so long as u^ix) and u^{x) are

both

positive,
d'Uo

and

so, when

U-i

du2

the

rv

integrate,

we

Since

jr

.J

dui
""^

dx

dx

consideration

under

now
expression

vanishes

at

the lower limit,

have

we

du2

dui

Hence

we

have

ax

therefore

and

U2

that is to say,

(2)

'^

Ui
*

To

(x)

u^

(a)

simplifythe presentationof the proof of the theorem, it is convenient to change the signs
immediately on the rightof j: a; the
if necessary, so that Wi (x)is positive
(.r),
(/._"

of u-y (x) and

signsindicatingmoduli
t The

theorem

is

may

then

be omitted

due
practically

to

throughout the enunciation.

Sturm, Journal

de Math.

i.

(1836),pp.

12-5"

127, 145"147.

15-83]

ZEROS

It follows

519

FUNCTIONS

BESSEL

just before Ui(x) vanishes for


it has remained
positivewhile

that

and
positive,
value

OF

first time

the

has

increased

first part of the theorem

is therefore

is positive
ihat ii2 (fJ-i)
and
the firstmaximum

u.^{x) must

Therefore

max

the theorem
When

{x)

Ml

We

is

that

shall

Uo

construct

ve^ ;

which

values
positive

\ '^W

"

is
of Q

the
u

and

The

results

(1917),pp.

are

in

ing
concern-

is 0

(v^).Our procedurewill
less and
respectively
slightly
question.

i;

equationto

its normal

form

by writing

9r%{ye')
=

^,

'^^(ye^)for

0,

of

(4) is

cylinderfunction have to
coefficient are
equalto "^^(ve^)and
this

be

adjustedso

its differential

0.

for the

Zj and I2 are

t These
XVI.

Bessel's

"

results f

6'^Q.

its differential

reason

ic

obtain

less oscillatory
than
obviouslyslightly
is obtainable by solving
the equation

implied in

constants

coefficient at ^

which

when

generalsolution

The

largeand

postulated
than
oscillatory*

manner

7nore

(x).

justproved to

m^-^'''

since e~^

that

is

Ui(x) is

have

(*)

and

to

"l"^^-^-^
function

small

reduce

(3)
A

(0,/Hj).

rightof /^i.

i^),

k,,

related in the

(x),are

pairsof functions which


than the functions
oscillatory

then

we

u^

theorem

apply the

In the first placewe


=

maximum

interval

the

on

completelyproved.

Y^ (x) when

more
slightly

must
u,2,{x)

say that
than Uj
(x) is less oscillatory

now

J^ {x) and
to

be

have

the

in
positive

(/ii)
^ Uo (/ij)
^ u.2{fx-,)max

Ui

functions,Ui (x) and

two

(x) and

be

point/lo of

in this theorem, it is convenient

be

the

have

Finallywe

Wj

from

proved.

is positive,
well as Ui (a),then Ui(x)must
as
Again,if Ui'{a)
have from (1)
before it vanishes,and at this point,
/Xj, we

and

still

a.

The

so

Uii^) is

use

of these

terms

is obvious

from

consideration

positiveconstants.
obtained by Watson,
supersede the inequalities

166"169.

Proc.

of the

London

specialcase
Math.

Soc.

in

(2)

520
It follows that
when

which

function

BESSEL

OF

THEORY

is

[CHAP. XV

FUNCTIONS

than ^v{x),
less oscillatory
(slightly)

x'^v,'\'"
'

(2e)i
We

(f)a-)* ^. (v)J-,

endeavour

now

to construct

functions which

are

from

(6) is
o/r

more
(slightly)
^^ (x) trapped between

have

may

should

we

of 6 to

function

we

construct

with

function

be determined.

that the suitable form for "yjr


(6)would

" 8'43

is

function, combined
oscillatory

less

stated in " 8*43,suggests that

where

function which

latory
osciltwo

than ^^ (x).
easilyinvestigated

more

for the

formula

"^; (^)/j
a) (i.)3

^y (x),in order that

than

The

(^")

It
be

might

tan^

the

of the

be

result

type*

anticipated

/3,where

sec

/3

e^ ;

but it appears that this function leads to a differential equationwhose solution


is such that its degreeof oscillation depends on the relative values of v and 6,
and

we

The

invariant o/ the

is known

to

information

able to obtain any

not

are

be

equationdetermined

thereby.
by

("4'31)

WW)\

"

tT^T

\^Td-)\i^
+

36

'

and it is requisite
that this should
natural

to test

the value of

"

When

we

replacee* by
(6)
""|r'

tan

have

test

and

hence

we

to

^'''

J-0

is

negativewhen

of

tan^/S.
The

4
sec

(d)]
{yjr'
^,

multipleof

equation in

its normal

the

(6),
t/t

"^

the truth of the

/3" \/24

"

cf. "4'31

36

''"

(^/r{0)\

(17).

/3,

inequality
sin^yS

cos2/3V(H-itan'/3)
sin^/3
cos2/SV(l+ itan-yS)_

3, and

cylinderfunction

form;

0,

^Ir
{6)=,tan/3

13,

3(tan/3-yS)-

tan-

find that

we

3(tan;8-/3)^

(5)

for this value of

(6)
i/r'

'

exceed p^ (e-^ !)" It is consequently


slightly
is
(6) which
givenby the equations
yjr

,|r'(^)=V(e^"-l),ir{0)
by determiningwhether,

^^^J

is taken

it is

so

that

positivefor greater

the

product satisfies

values

differential

15-83]

ZEROS

Hence, since (5) is


is

certain

angle

when

/3

between

521

FUNCTIONS

0, it is

when

true

and

\/{\/24 3}

tan

arc

in the former

Proceedingas
V{3 (1

"

ySo,where

0^/3^
The

Itt.

^p

sexagesimal

We

can

obtain

which
^^^,(w)

of

which

than
oscillatory

now

find that the function

we

case,

/3);[r (f) (i^)i^^ {v)J_j [v(tany3 ^)]


+ r (i)(^)5 '^Z (^)Ji [v(tan/3

cot

more
slightly

zero

BESSEL

of j3,is 59" 39' 24"-27.

measure

is

true

OF

/3)]]

^^^/Sq.

extremelyimportantresult concerningthe

an

is

/^(i/sec/3),so

long as*

greaterthan

v;

for let

smallest

^X.^*be the smallest value

of

makes

(f) (1v)i'^.(v)./_. (I)+ r (^)(1v)^-#; (^)/^ (|)


6o^A-of the equations
26
\,yv^ V (tsin13 /3) I \/

vanish.

Then

give

Since, by Sturm's
this form, we
is

that the value

^\/v-^+

theorem, the

see

(v~-^).

of '^^(x) lies between

zero

of the zero

of
expressions

two

is next greater than

of 'Wv(^)which

in the form
expressible

When

^y (x) is equal to J^ (x) it is easy


functions of orders " ^ that
K
and

the smallest

so

zero

of
v

In like manner,

The

first maximum

f
The

J^,(x),when
+

of

the two

of Y^

zero

0(z^-i).
0

(v-i).
in

similar

i.e. small
t The

X This

entiating
differ-

approximations.

as

value of

which

makes

^fi^'fi+Oiv-i),

maximum
is

on

v^x

of the function
the
-

right of

qo

to 0

as

0-808618

V^,(x) cannot

its first zero

increases

from

.r

(i^-i).

be treated

this follows

at

in this

from

once

of the maximum
value of Jy{.v)qua function
investigation
a paper
by Meissel,A sir. Nacli. cxxviii.
(1891),cols. 435
"

restriction is trivial because


values

of

we

have

because

manner

" 15"3,because

to consider

values

of

^ for which

of

the

reader

438.

v^"^is

bounded

/3.

permissibiHtyof

this follows from

the second

vaRie of f is approximately 0-685548.

its

0 to the first zero.

consult

This

by

manner,

r(f)(ir.)ij.(^)j-j(f)-r(i)(i^)s/;(i.)j-_3(l)
the first maximum
of J^,(x)is at the point

Yy (x) increases from


an

of Bessel

point

first maximum

For

Table

(x) is

constructed
expressions

The^rst

is
large,

be obtained

J^(x) may

verifyfrom

(i/-S),

v'^X 0-9S1577

i.e.at the

is

result is that if ^fi^,^


is the smallest

then
vanish;]:,

should

v'^x 1-855757

the smallest
V

1-926529

to

part of Sturm's

theorem

just fj;iven.

CHAPTEH
NEUMANN

SERIES

The

16*1.
The

this

general terms

These

expansions are

analogous

to

is called

the

to

series

been

We

shall

finallywe
a

reader

-00

2Zz

cos

be discussed

in

expansions
of

are

Chapters

much

and

xviii

xix.

integer;

the

types discussed

in

an

considered*

only

investigationof the

the

more

mann
" 16'14, the description 'Neu-

in

expansions of

Chapter

Ann.

functions

It will be

v.

LV.

(1902), p. 493.
Neumann

as

sufficient

to

quote

"/".
of expanding
possibility

the

defined

by
the

general discussion
consult

may

real variable

J,.,n{t)Jm{z),

S
m=

shall discuss

very

of

The

series ; then

analyticfunction

function

expansions

known

Laurent.

n\

first discuss

Neumann

functions.

s"'-^^""T'"^"^j""..(^).

J,(z + t)=

the

in series

related

or

the well

Taylor and

of

that various

followingformulae

Z"" -{-z-

functions

suggested by Nielsen, Math.

discussed

alreadybeen

-sr^=

is

the

from
has

'

n=o

For

investigationof

Gegenbauerf.

first kind

(^.).,

into

VARIABLES

is the

xvii

analogous to

they will

will remember

reader

Bessel

more

theorems

of series for which

of the

series have

where

Chapter

series, although in fact Neumann

distinguish these

The

here

TWO

type

general series is due

series

or

extent

the

character, and

Neumann

specialtype

To

of

expansion of

series of the

Any

one

some

by

Fourier's

to

recondite

more

OF

series.

and

chapter

contain

analyticfunction

an

FUNCTIONS

types of expansions of analyticfunctions of complex variables

whose

of

LOMMEL'S

AND

definition
of Neumann

object of

various

XYI

memoirs

146; Leipziger Berichte,lxi.


*

Theorie

Wiener

der

we
a

investigatethe

shall

series

Neumann

expansions of
of

various

generalisationsof

by Nielsen, Journal

(1909), pp.

33"61.

BesseVschen

Functionen

Sitzungsberichte,

lxxiv.

with

of
singularities

given

the

coefficients ; and

particularfunctions.
of series of Bessel

all kinds

fur

arbitraryfunction

an

Math,

(Leipzig,1867), pp.

(2),(1877), pp.

125"127.

functions,

(1907), pp.

cxxxii.

33

"

35.
'

138

"

524

THEORY

[CHAP. XVI

FUNCTIONS

analogueof Laurent's

Neumanns*

16* 12.

BESSEL

OF

theor 6771.

Let

and
f{z) be a function of z which is analytic
defined
the
shaped region
inequalities
by

Let

C and

be the contours

formed

by

\z\ R,
both contours
between

^^

beingtaken

the

counter-clockwise

'

then, if ^ be

pointof

the

region

27riJe

t-z

z-t

f(t)On{t)dt+X ~On(z)tf{t)Jn{t)dt.
^Jn{z)l

ring-

have

we
circles,

27riJc

in the

the circles

\z\=r,

one-valued

Consequently
f{z) is expansiblein

the form
00

00

(1)

f(2)=

anJn(z)+

an'0n{2),

where

(2)

an

f f(t)On (t)dt,

9"

a,/

expansionoff{z) in

If the Laurent

=^ ffit)Jn (t)dt.

the annulus

00

f(z)^

bnZn+

is

'

00

I-

^,

have, as in "16'11,

we

C "'f'2"-'.""''-'",-'"'
(""1),

(")

(" ),m

"

m=

Gegenhauer's
generalisation
of Neumann

16*13.

By using the polynomialAn,v{t) defined


the formula given in " 16" 11.
generalised
f{z) is analyticinside and
formed by this circle,
contour
we
If

,,

If

and

the circle

on

expansion.

"9"2,Gegenbauerf

in

\2\ R,
=

and

if G denotes

has

the

have

z''f(t)dt

^'^^""^ (0}/(0
Jo^"^"
2^Jc{
dt,

so
00

Z^f{z)= 2 anJ^+n{z),

(1)

Theorie

Wiener

der BesseVschen

"

316

for

some

(Leipzig,1867),pp. 36 39.
See Wiener
130.
(1877),
(2),
pp. 124
the
of
cases
expansion.
special

Functionen

lxxiv.
Sitzungsberichte,

(1884),pp. 293

"

"

Denkschriften, xlviii.

16-12-16-14]

525

SERIES

NEUMANN

where

(2)

an

providedonly that

is not

negativeinteger.

Maclaurin

the
"16"11,

If,as in

-^.\j{t)AnAt)dt,

expansionoif{z)is

f{z)== t hnZ^\
then

(3)

,,.

,o1"2"

(" +

"nid:4^)t

^n"zm

expansionof "16"12 may

Neumann's

16'14.
or

squares
From

Neumann-Gegenbauer expansionof
products.

is valid when
then
1^1^?',

when

similar

manner.

/mictionas

series

of

expansionof " 9"o,namely


t

which

in
generalised

be

TJie

the

"

m\

';"=0

"

n=0

\z\"\t\, we
the expansion

at

can

infer that,if f(z) is

once

analytic

00

% anJ^+in(2)J,+Uz)
Z"-+''f(z)=

(1)
is valid when

the coefficients

\z\"r, and

are

givenby

the formula

an=^^.j^f(t)Bn.,,,.{t)dt

(2)

formed
by the circle \z\ r. This expansionis due
being the contour
with this,
connected
namely that
Gegenbauer*; an expansionclosely
C

to

f(z)= i a,:J,Hz),

(3)
where

a,:=-^^.^f{t)nn{t)dt,

(4)
Q.n{t)is Neumann's

and
is

even

an

is y-Alidprovidedthat f(z)
polynomial("9"4),
f
function; this expansionwas obtained by Neumann
analytic
.

formula
Gegenbauei''s

has

Matkr(A)II. (1902),pp.
A

derived

second

been

investigatedmore

different from
type of series slightly
from

of

"5'22(7)in

2^ r

(1. + 1) s

the formula

^-' =

recentlyby Nielsen,Nouv.

Aim.

de

407"410.

Wiener

t Math.

III.

(1871),p.

previouslyconsidered

the form

^^^

Sitzungsberichte,lxxv.
A7171.

those

599.

/i.+,(4

(2),(1"77),pp.

218"222.

is

526

THEORY

which

shews

OF

BESSEL

[CHAP. XVI

FUNCTIONS

that

(5)

ftn^"-^'^i

(''+'^"
(iz)i
Jh.+u)(^X

a"

n=0

w=0

where

Expansionsof

this type have

been

the

topicof

detailed

investigation
by

Nielsen *.

Pincherles

16'2.
Let

anJv^n{z) be

this series and


Let

The

theorem

its

its generalisations.

any Neumann

and let the function defined by


series,

continuations
analytic

be called

/(^).

also

function

defined

the associated power


The

and

by f{z)y and
series off(z).

lim

analyticcontinuations

throughoutthe

series converges

Neumann

its

domain

will be called

in which

^i{anJ'"+"(^)1|"1,

n-"-"x"

and

this domain

is identical with

Km
Jl-*-00

by

Horn's

the domain

aniljr^

"/
V

T{p

Neumann

series has

circle of convergence, justlike a


series and of the
of a Neumann

and the circles of convergence


power series,
associated power series are identical.
theorem

that the convergence

of

prove

to

this theorem,

we

Neumann

Pincherlef;but
has
provedthat/ (2^)
off(z)x^.
singularities
power series is due
and, in fact,it can be
a

To

"1,

n+l)

formula ("8-1).
asymptotic

It follows that

The

in Avhich

series resembles

it is

possibleto go much
which are
singularities

no

that of

further,
not

also

write

ctna^Y"^'

Nijt Tidsskrift,ix. (b),(1898),pp. 77"79.

t Bologna Memorie,
pp. 493"496.

(4) iii. (1881),pp. 151

"

180;

see

also

Nielsen, Math.

Ann.

lv.

(1902),

16-2, 16-3]
and

then, inside the circle

f{z)
the

From

for any value


chosen ; and

cos

theorem

be written

due

Hadamard

to

F,(Z)^

T(v

""^

Vtt "ro 2''+" r{v


a

-^^^--^^

(z)may
definingcf)

^^
and

(zt^)
t^] (/"

continuation
it follows that,if (f)
(z)is analytic
theoryof analytic
is suitably
of 2',so also is f(z),
providedthat the path of integration
of (f"
so all the singularities
(z).
oif{z)must be singularities

the series

Now

*,

of convergence

j^^{^(1

527

SERIES

NEUMANN

l)r{v

b"Z"\ F,(Z)=

n+l)

|-)

'

that, if

states

"=0

in the form

F,(Z)= i bnCnZ\

CnZ'\

"=0

M=0

then all the

of F^ (z) are expressible


in the form ^j,
singularities
of F^ (z) and 7 is some
of F.2(z).
singularity
singularity

some

Since

the

of
only finite singularity

the

where

/3 is

hypergeometricfunction

of 0 (z)are singularities
1, it follows that all the singularities
point^
of f{z) are singularities
off(z)y; and therefore all the singularities
off(z)x;
and this is the theorem
which was
to be proved.

is at the

reader

The

should

connected
theorems:!:
in this

have
with

and provingsimilar
in enunciating
difficulty
which
dealt with
are
expansions

no

the other types of

chapter.

The

specialNeumann

Various

16*3.

of Neumann

number

forms, whose

series.

series,in which

coefficients

the

of

simple
is
importantanalytical
properties
of some
such series which are of
giveinvestigations
are

represent functions with

sums

large; we shall
interest.
special

not

now

By usingthe expansion
(^-2f-

cos

2^ +

1)-*=
n

to

It is assumed

be truncated

argument
t

Acta

X For
p. 230

et

that

by

the

R{v

omission

r-^"--P"(cos 2^),

h) is positive; if not, the


of the

terms

for which

several

(v +

series under

A)

is

discussion

negative, but

the

have

general

is unaffected.

Miith.
such
seq.

XXII.

theorems

(1899),pp. 55"64;

Hadamard, La Serie
concerning the expansion of " 16-14, see

de

Taylor (Paris,1901),p. 69.

Nielsen, Math.

Ann.

lit.

(1899),

528

OF

THEORY

Pincherle*

has observed

If

closed
(large)

curve

^TTl J

obtain

^-^
^^.j

w-plane is

e^'^dw

V{K+l)0^Hsin2^)}'

the formula
/"(0+)
e-^'"'''du

1
=

-:-

^"^

the modulus

where

in the

nv

/"(
(0+)

J,,+,(^)P,(cos2^)

{Z)Pn (cos26)
(1) S J.,n+i
M

1.

"

^^^

we

COS

1/t) w, so that the contour


find that
we
surroundingthe origin,
^{t

.n+.v/

so

2t-

"

the circle j"

00

and

'IfV
l^it,
f""';,!?
\/{P
2^+1)

lies whollyoutside

write

we

now

(cos 2^)

the contour

where

[CHAP. XVI

FUNCTIONS

that

i 7.,. (.)P"
n

BESSEL

of the

r{iK-+)

1
=

e-f^sinesnw^j^^

-r-

^TT j

function
elliptic

is sin 6.

expansion
interesting

The

who proved that the series on the rightsatisfied


givenby Jolliffef
derived as
differential equationas Jy- {^z). This expansionis easily
the same
a
specialcase of " 11"6 (1), but the followingdirect proof is not without
has been

interest

Neumann's

By

and, if we

formula

expand Jzv {z \jt)into

i+\-^^'s

T
J,.
{z v")

find

we

on

("5 "43)we

cosec

l)r(2i; + m+l)

;^r(2.
X

2^1 (-

2{2v

i)

TO, 2i; +

7?i

1 ; 2v +

\; t)J2,.+mt+i
i^),

that
integration

the series

2m

(2i^+

-^^^^

have|

Bologna Memorie, (4)vin.

2m+l)

(1887),pp.

J.

125

"

143.

Pincherle

used

functions
elliptic

of modulus

0 in his result.

zh Jv-^ (\z)Jv {^z)was


t Messenger, xlv. (1916),
p. 16. The corresponding expansionof
ix.
by Nielsen, Nyt Tidsskrift,
b, (1898), p. 80.
instead of definite integrals.
X li R{v + ^)"0, we use loop integrals

obtained

16-31]

NEUMANN

529

SERIES

where

T(2u
TT

by

-L

t''-Hl-trK,F,(-m,2v

"n

l; 2j. + 1 ; t)(U

{-Zi' 1) Jo

n
/i

n
l)

4+

7"7
m

-^7o^A
I
+

integrations.
partial

It follows that a", is the coefficient of h'^ in the

expansionof

1
^2. (^
TT

in

^+

/,^(1

t)]-^dt

1.

It
we

^)}-^-i
{1

It; and this expansionis absolutely


convergent when

of

ascendingpowers

IA I"

]^t.
(1

write

we

"

"

-^

-,

hu

"

find that

aji'^

ff-i
{1-h(l-

I ^i-'-i
(1

evident

0 and

that ftoji+i
=

1.3

1
"-

m)-*(1

dt
t)-i(1 + /?e)-i

du.
h-u)-^

Jo

TT

It is now

OJ-"-^(1

that

...(2/i-l)fi

Jo

^"

2.4...(2n)

l^ 1.3 ...(271-1) r{v

,,

^'

''

"

^"
''^~'

i)r(^)

~7r'
and

this formula

16'31.
The

at

(2n)

2.4...

r(z/+

givesJolliffe'sform

once

The Neumann

series summed

of the

l)

'

expansion.

hy Lommel.

series by means
of recurrence
transformingNeumann
been studied systematically
ceeded
by Lommel*; and he has sucthe sums
in obtaining
of various series of the^tyjje
means

effects of

formulae

have
this

by
in which

a^

is

polynomialin

?i.

Tafee the functions


00

M,,,n{z)

"

+
(z.

-In

{V +
1)/:,,

2n +

1) ./,+",,^j
{z),

where

/(

is a function to be determined
/^,i,(t')
*

\v. B. F.

Studien

ilher die Bessel'scheii Functioiten

presently.
(Leipzig,1868),pp. 46

"

49.
,34

530

THEORY

the

By

formula

recurrence

(1)

BESSEL

OF

[CHAP. XVI

FUNCTIONS

have

we

""

{Z)
64y,,n

2m +

{v +
f2y,

{Z)+ J^+2n+2(2)}
1) {J'.+s^

{z\
=f-2m{V + 1)J",(^)+ IM.^yn
"provided that

/;"(v +

the equation
of mixed
/^(v) satisfies
2n +

solution of this

adopt this

We

3) +f,,n(v +2n

l)=2(v

differences*
(v +
2)/,,^^^,

2n +

2n +

2).

equationis

value

of/w(i')and

then

it is found

by

the

method

same

that

{v)J^+i(z)+ 2S4u,m-i
'v,m
{Z) -f-2m-i
(z).

(2)

it follows that

Hence

(z) W2m {V + 1) '/.(Z) \Z%"_, {v)J,+,(z)+ Z'S^^,^ (z\


"S^.,m
=

,n-l

and

a,,rn(z)

z^^

so

S
71

(v + l) ./.{z)
{i^^-^/;,
0

+
^.+1(Z)]
l^^-^'7;"-i
('')

Z^^+^- a., -^

(Z).

Therefore, since
1 f^

00

(z)
S^u.-l
=

(^)
J"+2n+i

^Jo

M=0

have

we

from
similarly,

and

(4)
_

due

(0 dt,
/ "^"'

Recent
to

1^(. +

the

2" +

r,.Tav+n
2)

of
applications

Bateman,

Proc.

Int.

for My^m(z),
expression
+

^)

^."" (.)
r(^^^".,"^^^

Neumann

Congress

series to the solution

of Math.

i.

of

equationsof

(Cambridge,1912),pp. 291

mixed
"

294.

differences

are

532

Therefore,on

integration,
I

C is

where

Hence

"e-"^
""-^'
Ce-'"" +
^
2.0

independentof

we

IJo {v)
-

By takinga

a.

0, we

dv,
(v)/t}

that

see

C=lft.

have

\z t^^+H'^l

%yA^)Jn{a)=^^.j
+

of
majorit}^

I exp U

(^

^ r [J,(2
having a

the terms

v)

(^
-

residue

zero

U [tJ,
dv
{v) J"i{v)\

dt

v) + J,(2

\l{z-a)
\t--^)\

exp

",

the

[CHAP. XVI

FUNCTIONS

BESSEL

OF

THEORY

at

v)]Jo(v)dv,

0.

Consequently
S

nJn (z)Jn (aj

.s

"=i

"

"

-"

-'o(v)dv,

"

"

i- V

that is to say

^ f JA'^A
"

(1)
If
this

7iJ"(^)/" (a)

select the odd

we

we
equation,

and

each side of

on

find that

is

one

of

Kapteyn'sformulae

Z-V

and

r" (Ji

4Jo I
z
=

"\-V

jAz-v)\

'^^^

\-^

J,{z + v)]

{z -v)

z-v

f" (Jo{z+v)

4.J0 (

z-\-v

'^

'\J,{a-v)dv,
J.

z-v

integrate
by parts.

we

Hence

it follows that

i 2n/,"(^)/,,(a)

(3)

which

^) dv.

parts of the functions of

even

^}q

when

i (2n+l)J,n+,('^)J2n+Aa)

(2)

which

j^ ("

is the other of

ni^^L"J^ +'Ili'JZl)\j^^t-v
^ \\lt
4JoJo(-S'

Kapteyn'sresults.

'V

"V

16-4]
reader should

The
when

have

533

SERIES

NEUMANN

in provingby
difficulty

no

similar

methods

that,

(v)" 0,

l(v+

(4)
"

^"^-^I"'^^"^ J^ ("

(")
n)J^^n {z)-/.+"

"^

z"

.'0

^0dv

dv.
2 Jo

The

16*4.

"

Webh-Kapteyn theoryof Neumann


series have

standpointof the theoryof


functions of real variables by H. A. Webb*.
His theoryhas been developed
The
portant
by Kapteynf and subsequentlyby Bateman
theory is not so imit appears to be at first sight,
as
because, as the reader will presently
deal
with
it
has
functions
which
to
must
not
realise,
only behave in a
the
variable
tends
but
also satisfy
to
must
as
an
" x
prescribedmanner
intricate integral
In
functions
the
which
amenable
the
to
fact,
are
equation.
in the functions to which the complex theoryis
to be included
theoryseem
and simple functions have been
constructed
the real
to which
applicable,
variable theoryis inapplicable.
Neumann

been

studied

series.
the

from

The

result

on

which

the

theoryis based
dt

'Jzm+i

yt)^

{t)

211+1

that,if an

so

odd

is that

function

f{x) admits

of

("13'42)

(0

{m^n),

]-i i/A^,

o\

(l/(4w+ 2)

an

expansionof

(m

n),

the type

00

J yX)

",

and

'J271+1 \^)}
"^2tH-l

if term-by-termintegration
is

we
permissible,

df
CLt

r^"

have

C^2)i4-l

\j\t)J.n+i{t)^
=

"

4/1 -h 2

We

therefore led to consider

are

f{x)

(1)

the

(4n -f-2) J,,^,


{X)

we

shall

conditions

establish

I"-^V(0

the

truth

of

this

expansion under

(I) The integral

rf(t)
dt
Jo

exists and

is

absolutely
convergent.
*

dt ;

.'0

H=0

and

that
possibility

Messenger,

xxxiii.

t Messenger,

xxxv.

X Messenger,

xxxvi.

(1904),p.
(1906),pp.
(1907),pp.

55.
122

"

125.

31"37.

the

following

534

THEORY

BESSEL

OF

(II) The functionf{t) has a


'positivevalues of the variable which

[CHAP. XVI

FUNCTIONS

for
differential
coefficient

continuous

do not exceed

all

x.

the equation
(III) The functionf (t)satisfies

'^ {f{v

2/'it)=r

(2)

luhen t does not exceed


We

now

proceedto

1)]dv

the series

sum

7"

(""

oc

JO

(f\

^^^^^fit)dt,

(4/1+ 2) J,n+^(^)I

71

we

x.

8=1.

and

t)+ f{v

Jo

firstinterchange
the order of summation

and

It is evident
integration.

that
00

JfM+l\^) ['Jin
\i)+

"

2rH-2

(01

n=0

uniformlywith respect to t for positive(unbounded) vahies of t,


converges
since |J^n (01^1 ^^^ ^ !^2(1+1
(^)!is convergent. Hence, since f{t)possesses
an

we
convergent integral,
absolutely

effect the

may

and then,
interchange,

by " 16-32,

{2

S=r

fit)

Jo

=2.

(4n + 2) J,,^,ix)J^,

fVoix

IJ'^{fitV)+fit

t^)

v)

Jo
now

rfiv 1)'^
-

V)}dtdv

dtdv.

Jo

the last

transform

^"^"-'^"^
^^'\Li"^^"7+7^f

IfVo (^
+

We

it)]
^

U=o

integral
by using" 12"2,and

then

n''j,ix-v)fiv-t)'^dtdv
Jo.

r IVo iu
J oJ

t)fix u)
-

'^^dtdu
t

\ t/iiu)fix

"

u) du

Jo

jQiu)f'(x u)du.

=fix)"

"

Jo
*

The

first transformation

is effected

by writing
v=x

+ t-u.

we

have*

16-4]

535

SERIES

NEUMANN

Hence
''

(3)

(4/.-F 2) X,+, {x)

v)+f{t-v)\dt
f{x)-\y,{!n-v)W'{v)-\f^^'^
+

Now

so

\ "^^/lOdt

write

that

^(w) is

Ji{t)/thas

of v, since

function

continuous

an

absolutely-

convergent integral.
If then
as

(0,X),

we

to

are

when

S=f{x)

value in such

has any

an

interval

have

must

we

have

rj,{x-v)F{v)dv
0,
=

Jo

throughoutthis

interval ; and,

with respect to
differentiating

x,

F{x)=\J^{x-v)F{v)dv.
Jo
Since

jJj {x

induction

v)\^ l/\/2,it follows by

"

this

from

since
equation,

\F(x)\^^j')F(v)\dv,
A.

that

x""

\F{x)\'i^j^^,
A

where

bound

is the upper

of

the

jF{x) \ in

interval and

is any

positive

integer.
If

we

make

n^-

it is clear that

-^

F{x)

0, and

so

of equation
necessity

the

(2) is established.
of equation(2) for
sufficiency

The
from

expansion*is evident

the truth of the

(3).

cosec
a) is
pointedout by Kapteyn that the function sin (i*?
has consequently
and Bateman
for which
one
equation(2) is not satisfied;
which
determine
criteria
functions
for
endeavoured
to
equation
satisfy
general
discovered.
criteria
been
have, as yet,
(2); but IjO simple

It has been

[Note.

If /(a-)is

not

an

odd

function,we

expand

odd

the two

functions

i{/(-^-)-/'(--0}, 4-*-{/(.^)+/(-.^)}
separately
; and

then

it is easy

by rearrangingthe

to prove,

f{x)=2
f^

where

"o

?i

/W

"^i(I"*' I)^-^j
dv

I
J

providedthat
*

The

the

"

'A (-^Orri
("'")
./"
I*^

"

appropriateintegralequationsare

(but not
sufficiency

the

expansion,that

ClnJni-O-),

f ""

""

second

of
necessity)

the

(" " 0)

"
I

satisfied.]
equation was

proved by Kapteyn.

536

THEORY

Cailler's

16' 41.

OF

BESSEL

[CHAP. XVI

FUNCTIONS

theoryof reduced functions.

series which
has just been expounded
Webb-Kapteyn theoryof Neumann
with a theorydue to Cailler*. This
has several pointsof contact
connectinga pairof functions. Thus, if
theoryis based on Borel's integral
The

CO

"

the function

then

/(2^)^defined

the series

by

00

f{z)R=
"

supposedconvergent
by the integral

If the Neumann

series which

\z\,may

be

represented

termed

y (2")^maybe

of

values

r e-^f(tz)
dt.

function

Cn.n\z'\
0

small
sufficiently

for

f{z)j,
The

the reduced

function{lareduite)
o{f(z).

representsf{z) is
30

f(z)=

anJn{z),

n=0

then

we

have, formally.
r

00

f{z)R

n=0

an\
J

e-f .In{tz)dt

v(n-^^)"ro
Now

"

put
K

and

we

see

that
1 +

Hence, if the Neumann

^- ./ 2^

series for f{z) is 2


M

function of 2

anJn (z),then

the

generating

a,i^"is

w=0

providedthat
More

this function

is

generally,
iff{z)has

near
analytic
a

the

origin.

near
branch-point

the

originof

such

nature

that
00

/(^)= 2 anJ^+n(2),

(1)
then
="

Mem.

de la Soc. de

1 +

c-

Phys. de Gemve,

xxxiv.

zr

(1902"1905), pp. 295"368.

"^

LOMMEL's

16-41, 16-5]

537

functions

if

In like manner,

f{z)=

(3)

)j

a"^''+"J,+"(^),

then
^^

.to2''+r (i.
+

[Note.

1)

rn

If

"""^

v(i

e"^sin"3=

^ ^

v(i

n^

"

a"y" (0),

2
n=l

result,wbich

This

Diffractionand which
been

,"

of considerable

are

defined

are

Circular
The

exhaustively
by Lommel*
Aperture and Diffraction

functions

defined

Vn (w,z),are

UA^U,Z)=

(1)

Vn (^^,
^)

(2)
It is easy to

see

at

from

(-)-

The
n

by

equationmay
is no

(")'"

\Z

00

"

unrestricted
j_for

fW

be derived

from

the

nir

mr\

Z'

+
(^2^

"^n

Jn+U^)

27.-T

"

F_"+, (w, z)

values of

UAw.z)=

Abh.

der

664.

The

in it differs from

math.

in

pp. 607"608.

phys. Classe

first memoir
that

by J. Walker,
functions

and

z-

^^H2

"2-J

precedingequationby replacing

deals

der

with

we

(-r

different

k. b. Akad.
functions

non-integral

"

J".+...(4

the

328,
xv.
(1886),pp. 229
(Miinchen),
integralorder ; and the definition of V,^(w,z)
work is reprofactor ( 1)". Much
of Lommel's
duced
der Wiss.

"

of

Analytical Theory of Light (Cambridge, 1904). The

physicalproblem

of

\Z/

adopted subsequently by

The

define functions

write

m=0

529

symbols Un{w,z)

(^

2
=

in extending(1) to
difficulty

(5)
*

Diffraction

on

1.

There
order

Tr

Un+^ (w, z)

last

type, have

/_"-,,"{Z).

'

theory of

Jn^2m(z),

Iw

rr

problem

" 2*22 (3) that

/AS

the

Neumann's

of

the

by

"i

(4)

\ZJ

Un{w,z)-V,n^,(w,z)=

(3)

as

StraightEdge.

(-)'MQ

set

in his great memoirs

integralorder n, denoted
by the equations

of

theory,was

importance in

by simple series

discussed

Cailler's

from

variables.

functionsof two

functions,which

Two

,T"^,,,

"

immediately deducible
Tripos,1896.]

LommeV

16'5.

""i"=-"

is

in the Mathematical

at

26^(1+^2)
^

tbeu

has

been

noticed

occurrence

by Pockliugton, Nature,

of Lommel's
lxxi.

(1905),

538
The
factor

the

expressionon
is

removed)

an

BESSEL

OF

THEORY

[CHAP.

FUNCTIONS

right is an integralfunction
integralfunction of w.

of (2) gives a
correspondinggeneralisation
is an
v
integer.And
consequentlyit
only when
of the
V" {w, z) for unrestricted values of v by means

of

(when the

z, and

series which

The

XVI

converges

is convenient

to

define

of
generalisation

natural

(3),namely
F, (w, z)

(6)
It is evident

(^^^
+

cos

"^

f/_,+,{lu,z).

that

(7)

J,{z),
U^{w,z)+U,^,{w,z)={^)^

(8)

F, {w,z) + n+,3{w, z)

As

specialformulae,

deduce

we

from

U, (z,z)

(10)

U,{z,z) -V,{z,z)

and

V, {z,z)

l [J^{z)+

hence, by (7) and

U,n+,{z,z)

providedthat

?i

^ 1 in

(11),and

are

Proc.

closelyassociated

Section

of

Sci.,K.

16*51.

The

It is evident

z-^'t (-)'"
jsin

e^",^,

of
generalisation

J"j (s),

sin (m +

formulae,

^) 6 J^j^I ("^))
.

with

Akad.

van

functions, have

Lommel's
Wet.

te

Amsterdam,

(1918), pp.

191

"

vii.

LommeV

studied

(1905), pp.

375

"

functionsof tiuo

(1) that

hence
z-

been

by Kapteyn,
376, and by

199, respectively.

^UA'w,z) --U,+,{w,z),

consequently

these

{-YUn{z'jiu,z).

" 16
by differentiating

9"

;
/^m+i(^)l

(12).

for
equations
differential

(1)

and

TO=0

Hargreaves, Phil. Mag. (6) xxxvi.

and

^ 0 in

cos

j"

l {-T

that, as

md

2
OT=o

which

/"

Vn{w,z)

(13)
functions

S'{-T"^",J^rn{z)\,
l(-)4cOS^-

F^+i {z,z)

It is also to be observed

The

z]

l^xnz;

(1 2)

cos

(8),

U^^{Z,Z)= F"(^,^)

(11)

that

" 2*22

(9)

(^y/_. (z).

vanables.

540

of
integrals

the

case
particular

"^

in which

^r

''''
^"-^^^^"^

0 is of

by

the

have

we

=or(y

l)'

-2m +

of Lommel's

terms

functions

equations

U.{w,0)J^''^'^^^^^^^

(9)

'

r(v-i)
F_^+2(iv,
0)

(10)

Of these results,(1)
The

when

formulae, valid
following
be noticed

cos

U^+,(w,0)

(12)

memoir.

Lommel's

is

integer(zeroincluded).
positive

V',n{w,0) (-r

(11)

given in

(8) were

"

should

'^''"'^''
^^^"
interest ;

some

in
U^ (tu,
0) and F_^+2('^.0) are expressible
variable

particular

U,{w,0)=

so

one

^
(^ c)"

(8)

of

V_v+2{cz,
similarly
z),are

equation

a)

and

[CHAP. XVI

FUNCTIONS

it follows that Z7"{cz,


z),and

Hence

The

BESSEL

OF

THEORY

(-r

|w

'"-i(-)"'(Jw)^'
S
^=o"~(2""")!^.

"

(2m+l)!

^=0

?7_"(w,0)

(13)

cos(|w + ^W7r).

it follows that

Hence

Fo(^^,0) l, Vn+^(w,0)

(14)

F_^_, (i^,
0)

16*53.

i-y^ilwYsm+1

{-r t

^0

(2m

of Lommel's
Integralrepresentations

1)!

functions.

formulae
'^

w"

U^ (w,z)

(1)

J^,_i
(^)

-^^

C^.+i
(w,^)

(2)

^;zi
"3

the

0,

F_.(.,o)=(-)nj;-M^,

(16)

which

(15)

The

are

valid when

integrandsin powers

cos

{|w; (1

^0}
"

^''dt,

^.-1 {zt)sin {iw (1


.

f")]^''(;^,
.

/n

{v)" 0, may
of

and

be
then

verified

immediatelyby expanding

using the

result of

" 12"11 (1) in

lommel's

16-53]

For
performingterm-by-termintegrations.
replacedby the equations

U. {w, z)

(3)

^.4.,
{w,z)

(4)

v,

be

they may

'^"'^
^~ '^^ ''''^ ^^"^^^

^.^^_^ ^.^ ^^^

J,_,(-^0-sin{li^(l-^01-(-0''^""

"

"

increases from

tt

"

to

tt

as

describes the contour.

R {v) " 0,

[^

w"

i U,+,(w, 2)

^.-1 (2t)exp ["^{w{l- 1^)}t" dt.

-^1

' ft
.' 0

By modifyingthis formula
values of
valid for positive
W

we

and

obtain

can

Let

z.

of F^(w, z)
integral
representations
us

consider

/,_i {zt)exp {" liw (1


.

^'^^^~ ^^^^^'

'

U, (w,z) "

of

It is clear that, when

(5)

values

other

2,-^.-i si^ 2.7r

phase of

the

in which

541

functions

r-)]t" dt.

integralconverges at the lower limit when R{v)"0 and


limit when
R{v)"'^, \i w and z are restricted to be positive.
The

the

at

upper

round until it coincides


swing the contour
integral,
with the ray arg t= ^- ^ir,this ambiguity in sign being determined
by the
modification in the contour
is
a
; such
ambiguity in sign in the integral
by Jordan's lemma.
permissible
When
we

the last

evaluate

To

expand the

we

new

in ascendingpowers
integral

of z,

in

as

" 13"3,

find that
["^

t-)}t" dt

./,_i(zt)exp {"^iw{l-

Jo

2"

"

2"

,"=ow!l {v+ m)Jo

,^0

^l

w)

'

that is to say

^ [V._,{zt)exp

(6)
When

we

combine

.t''dt
t-)]

( f ^^ "f-')

in this formula, we

see

Z-

J /._: (^0

"

"

exp

that,if

then

the results contained

z"0, and 0"R(v)"^,

(7)

{" liiv(1

cos

.t^dt
[lw(l- f-)]

cos

I'TT

(^-2^y
+

(8)

^^ [^./,_,(.0 sin {!..(1

.t^dt=
t?)]

sin

(^^-"^^
+

"

iv

"

0,

542

THEORY

It follows at

V,_, (w,z)

(9)

-'^\

F,_, (iv,
z)

at the
convergence
continuation enables us to

n (w,z)

(12)

n_, {w,z)

remove

t^} t" dt.

now

I J,_. {zt)

[Iw (1

COS

providedthat

"m;

and

-l^^_/,_. {zt)sin

"

t"--'

{1..(1

-^^
t^)]

and R(v)" |.
positive

2^ are

formulae
special
following

The

analytic

^^)}

z"-^
=

dt,

that

see

t"

the theory of
unnecessary,
the restriction It (v)" 0.

(11)

.'1

the notation,we

Changing

f)}

/._, (zt)sin {1w (1

^.^

originis

Since

" 16"5 (6) that

/,_! (zt)cos {|w (1

with

.'1

(10)

[CHAP. XVI

FUNCTIONS

(1) and (2) combined

from

once

BESSEL

OF

are

worth

mention

'

Ujn^^jJ^^_^(^^)

(13)

{|0(1

COS

t^)}f^- dt
.

I "/^i-2(^0 sin {i^(1

^^)|t'""-'dt,

.'6

Um+iiz^

(14)

IV^,(^0 cos {i0(1

J^
^

/^iwf-\

(+^j

"2)}t^" dt.

that

see

r=^

./.-,
(.0 exp

(15)

\ J,n-i(zt)sin {^z(1

Again,from (6),we

cZ^
"2)}r-^'^^^

^"-1
,

exp

"

iz-

,,

."-*

_v'iri\

("j^ ^)
+

and, in particular,
r=^
.

'

Soc.

./0
last results should

The

The

be

cos

/w"2x

sin

V 2 /

1 sin / ^^ \

,,

compared

with

" 13'3;

cos

see

\2w/
also

Hardy,

Trans.

Camb.

Phil.

(1912),pp. 10, 11.

XXI.

found

formulae
in

one

16' 54.

or

of this section
of Lommel's

other

Lommel's

(with the exceptionof


two

the contour

are
integrals)

all to be

memoirs.

formulae.
reciprocation

It is evident

from

"16'5(13) that

connected
closely
integer.

with

functions

functions

of the

z)
type U^(z^/w,

of the type U^ {w, z) providedthat

is

are
an

16-54, 16-55]

LOMMEL

relations observe

of such
the significance
appreciate

To
d

(Iwt-) U^ [

cos

2t]+

"

dt

sin

543

FUNCTIONS

(Iwt-) U^+^ (^
.

that

zt
,

sin

(|w^")

;z'

U^ {", zt\

wt

"

U^^2 [-

wt

"

zt
,

iU

zJ, {zt)sin i^wt')(wt/zy-'.

On

find that

we
integration

./,(zt)sin (Iwr-)t'-" dt

(1)
to

(I

cos

"

"^

-kw
2

U^

i".f/."(|,.)
f,',g.O

sm

and, similarly,

(2)

J, {zt)cos {lwt')t'-''dt

-^

w"

sin

Hence

("
|w. f/'^+i

zj "cos

f
zj+ U^^^

"

it follows that
"iv

(3)

|w. UJ"

,-1

J,_, {zt)cos {fw (1

-^^

z"

t')]t" dt

Uo^^ f

^j

sin

"

f
|w C7'i_^

j+

|w ?72_^("

cos

and
n

(4)

/!_, {zt)sin {|w (1

f/i_^(

-2

"

\'W

and

COS

The
paper

by

will find

some

Schafheitlin,Berliner

16*55.

Pseudo-addition

the

"

) +sin

Itv
.

""

\w

"

differ from
integrals
only in the signof the

reader

f/'i_p(

iw.
'"

these

section

.fdt
t')\

Uo^^ I

"

'

\iu

of
correspondingintegrals

the

0 |

preceding

order of the Bessel function.

additional

formulae

concerning

viii.
Sitzungsberichte,

Lommel's

(1909),pp. 62

functions

in

67.

"

formtdaeforfunctionsof orders ^

and

".

Some

very curious formulae have been obtained by Lomrael, which connect


functions of the type ?7^{iv,
z) with functions of the same
type in which the
second
When

variable is
we

zero,

write

providedthat

is

^ in " 16"5o (5),we

f/'i
{w,z) i ill?,
(w, z)

equalto ^

or

|.

get

I exp
(^^Y
V-^TT/

{+ i {Iw

(^j I

{+ i {Iw

zt

dt
^wt-)]

Jo

exp

-\-zt-

Iwf^)]dt.

544
Now

BESSEL

OF

THEORY

[CHAP. XVI

FUNCTIONS

write

(w

zy

4-

("^

and

find that

we

U^ (w,z) "

i t^3 {w,z)

[
j^^y

gT'z

f-Vf

e"^'^

w
V

Hence

we

/""

to he

V(2?/;)'

{w,z)
{w,z)" iU^_

p)}d^

p)}rff

by the conventions*
interpreted
z

"

V(2iy)'

0)}
(u^ (2a,0) + iU^ (2(7,

igT^^

^e^'^{U,{28,0)"iU^{26A))}

-e'^^'l-] I

exp

{+ 0-1(1-^OWf

+ e"^'M

exp

I"

\7r/

and S^-as
o-|'

take

we

(1

have

When

,^

V^

\'

o-i

{" hi (1

exp

'//o

{"

exp

and \Ja,\/S are


respective
integrals,

in the

these

zf

are
integrals

seen

Jo

8i (1

to cancel ; and

so

we

have

d^
f-)}

respectively,
integrals

variables in the last two

new

the

two

results combined

in the formula

U, (w, z) " iU^ {lu,


z)

(1)
and,

as

U, {z,z) " iJJ^_


(z,z)
formulae

^e^^-{Ui{28,0)"iU,{2h,0)],

0) " iU, (4^,0)}.


le^fe{U, (4"z,

16"56.

Fresnel's

It is easy to

see

(1886),pp. 601 "605


AnalyticalTheory of Light (Cambridge, 1904), pp.

(iue to Lommel,

are

reproduced by Walker,

The

Miinchener

Ahh.

xv.

401

they are
"

402.

integrals.
R

" 16"53 (1) and (2) that, when

from

L\^,(w, 0)
so

corollary,

(2)
These

0)}
le^i'[U, (2(7,0) + iU, (2(7,

j^
2^;=^^

t^"-' sin

{^w (1

{v)" 0,

dt,
t^)\

that
f^

^1 )

i'"''
o.-i-p/

"^

(Iwt-)dt

U^{w, 0) coslw

^in

(lwi^)dt

U, {w, Q) sin |w

not

the

/""'"

w"

(2)

COS

X
{^) Jo

*""'

^i^;z:^r7-^

These

are

same

as

the conventions

used

U,+i(w, 0) sin | w,

f7,+i
(w, 0) cos ^w.
by Lommel,

lommel's

16-56]
If

take

we

545

functions

^ and modify the notation by writing^w

we
^iru'-,

that

see

JJ

(3)

/_i (t)
^j~^
(;^J^ ^j'

[U^ (2z,0) cos

dt
(^TTf-)

cos

tdt

cos

U^ (2z,0) sin 2]/V2

[^^i
(2^,0) sin ^

dt

0) cos 2]/V2,
F| {'2z,

and

JJ

sin

(4)

We

thus

dt
(ITTt')

obtain

^I"
(^J'

0) sin z-U^ (2z,0) cos z]/^/2


[f^j
(2^r,

sin tdt

J^ (t)
^j''

dt

[T^j
(2^,0) cos z-V.^ (2z,0) sin z]{^/2.

ascendingseries

asymptoticexpansionsfor

and

Fresnel's

integrals*
ru

ru

I
I cos(^7rt')dt,

Jo
The

dt.
{^7rt-)

ascendingseries,originally
given by Knockenhauer, Ann. der Physikund
are
readilyderived from the f-series,
namely

(2) XLi.

Chemie,

(1837),p. 104,

\7rJ {

while the
are

sin

{)

1
.

"

5
.

5
.

7
.

asymptoticexpansions,due to Cauchy, Comptes Rendus,


with equal ease from the T-series,
namely

xv.

(1842),pp. 5.')4,
573,

derived

Tables

of Fresnel's

constructed

integralswere

R. des Sci. de

52, and

Lommel
which

of Fresnel's integrals
by
givenvarious representations
of
formulae
the
c
ases
special
f

are

\~JAt)dt
J

,L+,n+,{z)

00

Jv+iiyZ)

(v + l){i'+ -^}...{v+

n=o

f JAtXU^

(10)

J
*

Mem.

t It

I
"

is

de

VAcud.

des

series

B. F.

memoir.

has

(9)^

w.

de I'Acad.
by Gilbert,Mem. coihronnees
Ann. der Physik und Chemie,
Lindstedt,

xxxi.
(1863),pp. 1
Bruxelles,
(18S2),p. 720; and by Lomiuel in his second

(3)XVII.

"

2n

1)

'

("l+l"iii^gI"-(-'^^"-l)^,,
Z

Sci.

v,

supposed in (9) that

(1818),p.
7?

(c)"

3.S9.

[Oeuvres, i. (186G),
p. 176.]

1.

3,-,

546

THEORY

These

readilyverified by

are

Lommel

[chap. XVI

FUNCTIONS

BESSEL

OF

Other

differentiation.

also due

formulae

to

are

(11)

r J^{t)dt=2icosl.

Jo

(-rJn+idz)
0

25sin iz

2
M

(12)

2^ sin

Ji{t)dt
=

ti-yjn^iilz)

^z

Jo

2^

"

t i-y-Jn+iiiz)

^z

cos

LM

These

also be verified

may

16" 57.

differentiation.

Hardy's integrals
for Lommel'

fact that the

The

by

(-)"^.+i(i^)
0

functions.

integrals

[^

h\

dt

h\

"

tdt

Jo'"n"'-^'r"^'io''""r-viit^
^

in terms
expressible
suggestedto Hardy f the

elementaryfunctions* when a
consideration of the integrals

of

are

r="

h\

dt

and

are

positive

^\ id*

Jo'''r^ijiTt- Jo^^"r-'^"Jr"i^

"

functions of two
in terms
of Lommel's
expressible
is important
variables of orders zero
This discovery
and unityrespectively.
such functions contain Bessel
because the majority
of the integrals
representing
functions under the integral
sign.

and

he found

If

l/tbe

(1 ,

,",

(2)
and

them

be

to

written

in

placeof

^,
IJ [at^)
+

COS

i''

t, it is

"

h\ tdt

that

seen

I' [U ^)jf^,
+

cos

['

b\dt

1^sm(a"+^)j^="j^("(-)--(^
.

"

sm

/,

a\

tdt

sm

since,by " 6*1 3 (8),

i^sin

TT
.

it is sufficient to confine
We

now

that
+

our

(at
V

j)^ Jo {2^{ab)},
=

I /

attention

^/ib/a),x=2

to

the

case

^/(ab), (9

Hardy, QuarterlyJournal, xxxii. (1901),p.


integralshave their principalvalues.
Messenger, ixxvm.
(1909),pp. 129"132.

the

in which

"a.

write
c

374.

When

|(1

the lower

c')/c,
sign is taken

it is

supposed

548
The

U,{w,x)

providedthat

"w

obvious

substitute

When

the

V2^

2"Jrr72'

functions.

of

I^ I ^ Ity I
=

We

thus get

2^"")'""'/!!'
"!."-)""
('S)A.
-p

is

the contour
,

and

^'"^

is
L\(w, z) and F"(?^',
z) by integrals
representing
of " 6*2 for each Bessel function in
Bessel-Schlafli integral

method

^- ("-^)

which

-~Pj^

"x.

series.
appropriate

the

V,{w,x)

in the form

Integrals
of Gilbert's type for LommeVs

16*58.
An

[CHAP. XVI

FUNCTIONS

BESSEL

be written

results may

last two

(8)

to

OF

THEORY

we

that it lies whollyoutside

chosen

so

the

change

may

order

of summation

"

the circle

and

on

integration

get

U.{w,z)

(1)
Now

".^_^

the residues of the

integrandat
iw

and

^-^^^_^exp(^--)-.
+

\iw

iz-

are

vTri]

so

Making

slightchange in

the notation,we

deduce

^^(^'^)=2^-j_
r+-F/"7^^"P(2-2^)T'

(2)

the points+ iw lie outside the


and, in this integral,
In

to modify the
generalit is impossible

half of the real axis taken


the

integrandat the

the essential

contour.

in

contour

twice,in consequence

exceptionoccurs
and
singularity
disappears,

when

f^^+Ht/wye^*
2^-/.
27rij_"
+t-lw-

of
singularity

0, because

then

dt

'^'(".0)

(2) into the negative

of the essential

origin.The

(3)
and

that

'

hence

(4)

V,(w,0)

=
,

TT

providedthat R{v)"0

and

a.

is

an

acute

.
I -i-U

such
or
negative)
angle(positive

that
)" +

,-du,

argw| " Jtt.

LOMMEL's

16-58, 16-59]
If

is

equal to ^

Formula

or

the

by

(4)was obtained
integrals.

the

integralon

549

FUNCTIONS

Lonimelt

rightin (4) is called

from

the formula

of

Gilbert's

integral*.

" 16"53 (11)by

tion
transforma-

of infinite

(4) it is clear that,when

From
is

numericallyless

and

positive,
V^{w,0) has

are

"

Sni

VTT

The
w

reader
is

and

signas,

same

1
,,",

r{i-u).{hwy

inequalitywas

similar but less exact

when

"_l

jo

TT

the

than

will also observe

that

obtained

by

Vv{i(',
0)/sinvn

is

Lommel.
a

positivedecreasingfunction

of

lo

positive.

16"59.
From

Asymptoticexpansionsof LommeVs

variahles.

functionsof two

it is easy to deduce
asymptoticexpansionsof
integrals
U^(io,0) for large values of \w\; thus, from " 16'5 (8), we

Gilbert's

F^(w;,0) and
have

where
and

p is any

then, by

choose p to be

positive
integer.We

-"

y-"
I

rx

exp

'\

ta

0(w-''-'P),

of " 7 '2,|arg
analysis

in the similar

I?^ Iis largeand, as

2p) "

have

" 16'58 (4),we

{-y F,+,^{w, 0)

when

largethat R{v

so

|"

tt.

Hence

for the values of


When
as,

and

u+2p
is

under

and

the

are

(Iwy^'-^

-2m).

(" )^F^+2^(w,
0) has
positive,

both

be

^""

the

same

sign

(-)P

Jo

r(l-v-2p).{lwy+-'P'

remainder

numericallyless than
It may

consideration.

f^

VTT

TT

that

tv

ZoT{l-

numericallyless than
sin

so

(-r

VAw,0)

(1)

after p
the

proved in

(p

like

(1) has the

same

sign as,

and

is

l)th term.

manner

F, [w, z)^

(2)

in

terms

from

" 16"53 (11) that

J"_,_.,^
(3)
(-)- {z/ivy+"--

when

\io\is largewhile

which
expression
*

3Iem.

and

z are

Ahh.

de
xv.

VAcad.

easy to obtain
signof the remainder.

fixed ; but it is not

givesthe magnitude and

couronnees

t MUnchener

R. des Set. de Bruxellea,

(1886),pp.

582"585.

xxxi.

(18(53),
pp.

"

52.

simple

550

It is evident

BESSEL

OF

THEORY

" 16'5 (6) that

from

[OHAP.

FUNCTIONS

the

corresponding formulae

for

XVI

U^{w, z)

are

(_)m

t^,(w;,0)-'cos(iw-ii'7r)+

(3)

=0

Uy (w, z)

(4)
These

results

The
is

but

2m) (iw)^-''+'"

J,.o.o,"(z).
(-)"^ {z/iuy-"'-''+'

he

did

+
I i/tt)

given by Lommel*,

were

asymptotic expansion

large and

The

positive,has

dominant

readilyderived
Vy {ex,

in

investigate them

not

t increases

from

(12) which

functions
1

to

1 and

or

is fixed, while

shews

of

greater

+ ^vir
^ca;(l ^^)+ (a?i
"

it may

hence

than

"

is

that

(xt + Iv-jr Itt)sin {| ex (1

cos

and

oc

is 0

investigatedby Mayallf.

x)^--^^j("^
the

i/

general (real) values

| 16'53

from

(ex,x), when

of Vy

been

for

term

if c"l,

Now,
as

^z^/w

'

detail.

any

{\w

cos

(i'

"

^'Tr)vary monotonically

"

verified

be

f^}

tions
by partialintegra-

that
c^"""

"

"

virxj

the

next

in the

term

If, however,
has

maximum

it follows

then

^cx (1

{x~^).

t"^)
+ (xt "\- ^ptt

^ it),

"

asymptotic expansion being

e"l,

"

function

lir), qua

hence, by the principleof stationaryphase

1/c; and

at

(a;+ ^ r;7r

cos

\^

Vy {ex, a;)

(5)

of

t,

(" S'2),

that

Vy (ex,x)

(6)

Finally,when
integration,and

c=l,
the

so

-^

\-^x(c
+ -]

cos

-1^^

maximum-point

the

the

expression on

is at

right

in

^vir[
end

one

(6)

must

of

be

the

of

range

halved.

We

consequently have
(7)

Vy(x, x)'^ ^cos(x

This
that

v"

equation,like (5)
"

by using

^; the
the

three

and

(6),has

equations
formula

recurrence

MUnchener

t Proc.

Camb.

"

Abh.
Phil.

be

may
16'5

(8).

^v'7r).
established

been
now

xv.

proved

ix.

the

hypothesis

for all real values

(1886), pp. 540, 572"573.


Soc.

on

(1898), pp. 259"269.

of

CHAPTER

XVII

KAPTEYN

17*1.

SEKIES

Definition
of Kcipteyn series.
series of the

Any

type
CO

11

in which

and

Such

series

the

investigated,
qua
important
such
into

of
possibility

recondite
of

the

character

of

half

square
in the

hands

of

of

much

it

manner,

Schott||they

occupied by

series

17*2.
The

in order

to

advisable

seems

with

their

first made
discovered

in

The

pp.

series

series

related

in the

appearance

by Lagrange^

proved

be of

to

general

of

are

covered
redis-

and

in the

frequent occurrence

with

begin

descriptionof

which

and

will
a

allied

be used

problems

discussed

by

a,

will, for the


set.

X Hist,

de VAcad.

de

h, and

most

VEcole

t Ann.

R.

The

e.

action

the

sup.
ScL

de

of

that

is

xxv.

" Berliner

Abh.

1816"7

il Electromagnetic

simple
which

[1819],

Radiation

pp.

49"55.

(Cambridge, 1912).

discussion
of force

centre

square

law, is as

eccentricity of the
taken

as

of

the

at

the

follows

ellipse

coordinate

zero,

91

"

120.

(1769) [1770],

113"138.]

ellipseare

of the

(3) x. (1893),pp.
Berlin,

the

axis, and

axes

jjart,be assumed
norm.

des

series

Bessel.

by

in the

section

in this

particleunder

semi-major axis, semi-minor

denoted

the

motion.
elliptic

ellipseof

an

series in

develop the theory of Kapteyn

focus,attractingthe particleaccording to the inverse

It

more

series,yet Kapteyn

Neumann

Bessel";and

have

to

Kepler'sproblem
notation

motion

series
series.

Neumann

general,of

in

are,

such

real,
are
problems, in which all the variables concerned
character than the problems investigated
simpleanalytical

more

and

in connexion

are

into

under
motion
the inverse
problems concerning elliptic
Kepler'sproblem may be taken as typical.More recently,

astronomical

by Kapteyn;

occur

theory of

an

the

theory of Electromagnetic Radiation.

The
are

examined

class of

law, of which

modern

propertiesof

century later by

in

occurrence

then

practicalimportance; they
Kepler's problem which was

more

solution

than

in

by Kapteyn f

Kapteyn

put the

to

propertiesof Kapteyn

systematically

arbitraryanalyticfunction

an

endeavoured
was

first

z,

In this memoir

1898.

which

that

to

variable

complex

series.

Kapteyn

they were

fact that

expanding

generallyhe

positionsimilar

Although

are

published in

series,and

the

of the

is called

constants,

cin are
to

name

functions

memoir

of the

question

coefficients
their

owe

pp.

204"233.

[Oeuvres,

iii.

(18(;9),

552

OF

THEORY

BESSEL

[CHAP. XVII

FUNCTIONS

to the
ellipse
of force is taken as originof polarcoordinates,
of force. The centre
centre
the radius vector
to the particle
being r, and the true anomaly,namely the
anglebetween the radius vector and the axis of x, being w. The eccentric
is denoted
the ellipse,
on
anomaly, namely the eccentric angleof the particle
at
was
by E. The time which has elapsedfrom an instant when the particle
the positive
end of the major axis is called t.

the direction of the axis of

axes,

The

anomaly M

mean

would

vector

is defined

performcompleterevolutions
completerevolutions.
as

way

The

(1)
'

the

(2)

iw

(1

"

"

A),
'

e cos

ecosw

/JU^J
\/%l
^
-

smw

*
ellipse
supplythe equations

equations

tan

(3)
^

"

1 +

which

from

the

of
properties
geometrical

angle through which the radius


rotated uniformlyin such a
vector
the time it actually
takes to perform

in

to

of the

the centre

the

as

t if the radius

in time

turn

being from

"

6^)

tan^E,
E

sm

e cos

Vd-e^sinw

"

,sinA=

"

i=i

"

1 +

e cos

zu

form of the equationsof motion


integrated
the equation
of Kepler's
Second
Law) supplies
expression

deducible

are

; and

lytical
(theana-

an

(4)

E-"smE.

Kepler'sproblem is that of expressingthe various coordinates r, w, E,


which determine
the position
of the particle
+, in terms of the time t, that is,
in terms
of M.
It is of course
supposedthat the variables are real
effectively,
is
0 " e^ 1.
a
and, since the motion
as
elliptic
(or parabolic,
case),
limiting
The

solution

of the

problem which

approximatecharacter,because
expansionsof E and r.

he

effected

was

calculated

by Lagrange was

only the

first few terms

completesolution givenby Bessel depends on the


defines E as a, continuous
function of M such that
increasing
M by 27r is to increase E by 27r.
increasing
The

series,qua

in Fourier

The

Dynamics

construction
of

Particle.

t Kepler himself

was

of

functions

these

See,

e.g.

concerned

(4)

effect of

limited total fluctuation is a function


so

such

functions of E

found

in

are

expansible

of M.

equations will be
Plummer,
with

the

an

in the

fact that

more

It follows that any function of -E*with


and
of M with limited total fluctuation,

of

the

Dynamical

any

text-book

on

Astronomy

Astronomy (Cambridge,1918),Ch.

expression of

in terms

of M.

iii.

or

17-2]
In
values

KAPTEYN

553

SERIES

function of M,
is an odd periodic
e sin E
particular
of E, it is expansible
into the Fourier sine-series
sin ^

where

An

An sin nM,

esin ^sin
J

iiMdM

2esin"'cos"i"'

2
+

1177

nM

cos

'-

"

,,,

dM

,,..

dm

dM

"

TTT

dM

J"

mr

nM

cos

'

nir

..dUsmE)
-^

f'^

"

,^dE-dM

"

nM.dE

cos

.'o

n-TT

so, for all real

TT

and

./,,
(ne).

Hence

it follows that
E

(5)

i. ^Jr,{ne)8mnM,

M+

solution of Kepler's
cerning
givesthe completeanalytical
problem conthe eccentric anomaly.The series on the rightis aKapteyn series which
when e" 1,and it is stillconvergent when e
1 ; cf ""8*4,8'42.
converges rapidly

and

this result

The

radius vector

is

similarly
expansibleas
-

^0

(^

l^{l-"cosE)dM

Bo=~

where

^"icos nM,

71

(1
TT

while, when

cosine series,thus

e cos

Ef

dE

^ 0,

Bn

{l-"CosE)

-\

dM

nM

COS

C^(6C0S^),"

2(l-ecos^)sin?iil/
Sin
mr

nir

sin

E)

niU

2e

sin E sin

(nE

ne

"

dE

nir

2e

so

that
2e

""

(6)

=1

4^2-

T
S
"-i

Jn {ne)cos nM.

"

TTT

dM

dM

554

The
tu

OF

THEORY

BESSEL

[CHAP. XVII

FUNCTIONS

expansionof the true anomaly is derived


function of if, and so
odd periodic
an

from

the consideration

that

is

"

% Gn sin nM,

where

C.n,=-

{w

M)s,mnMdM

"

J0

TT

2 (lu

M)

nAri''

COS

2
+

"

,^
nM.

cos

^\

hxr

\dl\i

J Q

,,,

d^^

,"

-,-r,dE

dE

J0

nir

nir

( dw

^j
nil/

cos

dw

f""

"

f"

"

Jo

niT

V(l

e-) f""cos {nE

sin

ne

E)

JQ

mr

This
and

expressionis

Bn.

used

the

The

such

not

simpletranscendent
of evaluatingit

effective method

most

cos

"

the coefficients A^

is due

to

Bessel*, who

expansion
V(l-e-)

^ ^

"

.^^^ ^ ^ 2/2
^

-^

cos

where

2^

cos

/'=

On

as

making

17*21.

-^

the substitution,we

Gn

cos3E+...,
'2f-^

2 r

"^

-\ Jn {ne)+

find at

that

once

/'"[Jn-m(ne)+ Jn-r-m{ne)]

Expansionsassociated

with the

expansions.
Kepler-Bessel

associated with the radius vector, true anomaly


largeclass of expressions
and eccentric anomaly,are expansible
the same
in series of much
type as those
in a systematic
manner
by
justdiscussed. Such series have been investigated
of
and
shall
of
the
few
them;
are
state
a
more
we
now
they
Herzf,
important
write
all obtainable
out in
Fourier's
it
to
and
seems
rule,
by
unnecessary
detail the analysis,
which the reader will easily
for himself
construct
A

First,we

have

a(l-e-)r

cos

iv

ae

"

so

that

'-^-^^^

(1)

and

"

J,' (ne)cos nM,


i?i

next

rsiniu

/-.x

(2)

a
*

Berliner

Abh.

t Astr. Nach.
delta R.Accad.
see

delle

^^

sm

\/{l-e-) ^
=

"

1824

cvii.

n=i

^
Jn
(we)sm

,-.

nM,

[1826], p. 42.

(1884),col. 17

"

28.

Various

expansionshad

pp.249"
Sci.diTorbio,(2) x. (1849),

Cauchy, CoviptesRendus,

xviii.

(1844),pp.

625"643.

332.

also been

In connexion

\_Oeuvres,(1)vni.

given by Plana, Mem.

with their convergence,

(1893),pp. 168"188.]

556

THEORY

More

OF

find

we
generally

\-ecosE

d'""

(5)

M=0

nine),

n=l

dM"^

n^"^ J

dM-'^

[CHAP. XVH

FUNCTIONS

by differentiating
" 17'21 (6) that

d'

(4)

BESSEL

cos

M=ir

(-)"-iw-'"j:"(?ie).

n=l

d
Since

-TYf

be calculated for any

Again,if

and

the left in

on
expressions

value
positive
integral

regard e

we

the

te^.

E^

the

as

of m,

with

(4) and (5) can

sufficient labour.

independent variables,it

is

easily

that

seen

_^f

{sin
^ (1

de

"*

cos

BE

_
~

cos

E)]

sin^ ^

(1

a^

cos

f
i

"

rTri

"

Etv"

E{1

sin

AE. +

cos

"

Sin

r^9^

A^)2[

cos

""

^r-

Be

esinA"

(1 -ecosjE')^
1

A
dlM 1
so

that,by

de

[sm E (1-

therefore,if we

and

-.

with
integrate

we

cos

the lower limit,

as

"

^,

smM

^)

differentiate with

"

cos

(1

last two

expansionsdo

Expressionsresemblingthose
of

on

Radiation
Schott, Electromagnetic

Thus,
(8)
The

as

cases

of

not

appear

to

Ef

cos

The

/" (nx)(Zic.

Jo

"*"

su^M

Ef

r*

find that

we

ilf

cos
~

(1

sin ?iil/

n=i

respect to M,

A"

cos

sin^ a;

(1 -6

_.

^ ^

^
"=i

1
-. "

=-

irj^

""

E)]

e cos

"

sin

If

A"

cos

(6)
^

" 17-21 (6)

n^

nM

cos

have

j Jn{nx)dx.
.'0

publishedpreviously.

been

rightof (6)and (7)have


(Cambridge, 1912) /"as"m.

the

occurred

in the researches

(4)and (5),Schott proved {jhid.


p. 110) that

J^

.'.^," (2ne)

last of these may

In^
^. (2..-)
^^^^,
^iL"^,
J^

be obtained

d.=

by taking M equal to

0 and

iz

in

(7).

17-23]
MeisseVs

17'23.

expansionsof Kapteyns type.

series,
namely
extremelyinteresting

Two

n^+r

n=i

557

SERIES

KAPTEYN

1^+r

(r^+r)(:i^+r)

-^

{M

"=i

If

1^

"*"

(1-^
+ f) (3^+ a

^!l^

have

who deduced
by Meissel
supposedat present f that 0 " e
*

stated

been

it is to be

simplestmethod

The

is to take the Fourier


"

0 to

0 ^ if ^

It is thus found

TT.

them

these
establishing

pansions
ex-

2f^F7r^

n' +

from

series]:
~

(which is valid when

in

cosh (7r-2ilf)f
2'"if_7r

cos

Z\

various consequences
^ 1, and ^ is real.

procedureto adopt

of

M
tt),replace

hy E

2p
"

"

'

sin E, and

from
integrate

that

+ 26sin"')^
^ X^(2ne)^1 r (7rcosh(7r-2"'
-^ui
^.rl '^"^+r
^sinh7^^

f i-

[ttcosh (2^ +

2e

^i)^

COS

^7^

j^.

_
_

TT

-ST

(ttcosh 2^0
fi'^

is an
expression
in the form"
expansible
the last

"

-;

m-

formula

due

f^
of

e,

and

it is

hence

ad

"

r(l+rBr(l-t|)
Z,r{7n + i + i^)r(m +

Cauchy||;and

to

smhTTf

.'o

^1
by

0)

cosh

function
integral

even

^-j

C-^w)!

m=i

^"-

TT^

(2e^ cos
I sinh 7r|

TT.'o [
Now

sinh

i-i^y^

the truth of Meissel's firstformula

is

evident.
The

second

formula
"

follows in like
cos

(2n

from

manner

1) i"_

TT

sinh

the Fourier

series

(Itt M)^
-

"

4|coshl7rr^

"

"=;
*

A)"tr. Nach.

t The

See

"

It is easy

IIMem.

cxxx.

extension

to

(2/i-ir+p
(1892),col. .863"368.
complex variables is made

siir

in

" 17'31.

(Paris,1817), p. 166.
to see
that the term
independent of e vanishes.
les integralesdejinies(Paris,
1825), p. 40. Cf. Modem

Legendre, Exercices

de C'ulc. Int.

ii.

Aiialy.iis,
p. 263.

now

558

THEORY

BESSEL

OF

Now, since the series obtained

^^

to

respect

^,

we

the latter is
The

for

the values

values

\, 2, 3

bounded
and

domain
then

make

polynomialof degree Im,

even

and

1.

"

given by

'

'

polynomial for

of the latter

also gave

any

of times

Meissel

in the

cases

l,2,3, 4, 5;

are

the values

that

Conversely,it is evident
in the form
expressible

l, 2, 3

2~

of the latter

'*'72*

32

??i.=

2~18'

2'
Meissel

an

polynomial were

The

(2) by differentiations with

number

polynomialof degree 2m

of the former

values

is

the former

e;

odd

an

diiferentiate any

may

polynomials*in

are

(1) and

that

thus deduce

We

from

uniformlyconvergent throughout

are

of real values of

[CHAP. XVII

FUNCTIONS

are

"^450"

81

polynomial for

4.,5.
is

polynomial of degree 2m

even

every

oc

71

and

that every

in the form
1, is expressible

of degree 2m"
polynomial,

odd

00

2
n

where

and

""

6"

hnJ2n-i{{2n-\)e],
l

polynomialsin l/n and

even

are

of
1/(2/? 1) respectively,
"

degree2m.

It

series with

Simple Kapteyn

17*3.

stated

was

recondite

in

character

" 17'1 that,

in

Neumann

than

complex variables.
general,Kapteyn
series,and

the characteristic differences between


In the

case

rearrange

the

in the form

double
resulting

is that of the
It is to be noted

zero

thev

of

as

we

that the coefficients of e-'" and

e-""-i in the

r-'
'

1.

more

explainone

now

series whose

are

( y-' ^.
(in!)"-

of

of

types of series.

power
series.

Neumann
original

are

series in the variable,and

power

series

shall

to expand each
possible
general,

series it is,in

of Neumann

the Bessel functions

the two

series

3^
.

(2m

I):'

domain

then
of

respectivepolynomialsare

of
to

vergence
con-

not

17-3]

559

SERIES

KAPTEYN

is

corresponding
property of Kapteyn series
Kapteyn series
+ n)z]
'^anJu+n{(v
The

is convergent and

for the

analyticfunction (cf." 87) throughoutthe

representsan

\/(l "2-)

exp

lim

"

"^'i/ar,

V(l-^')

series obtained

by expanding each Bessel function


is absolutely
convergent onlythroughoutthe domain in which
1
l^l.expV(l-I^P)

while the double


z

in which

domain

of

quitedifferent

in powers

"

and

the

the

first domain

is the

points" i

the circle

than

the former

interior of the

the longestdiameter

which
the

is smaller

latter domain

0'6627434

thus, when

shewn

curve

joins the points + 1,


; while the second

in

" 8'7, in
shortest joins
of

Fig. 24

while

domain*

the limit is 1,

is

the

only the

interior of

0-G627434.

\z\
=

dealingwith Kapteyn series,if we use the method of


expansioninto double series we succeed, at best,in provingtheorems onlyfor
and the prooffor the remainder
of
a
portionof the domain of their validity;
the domain
either has to take the form of an appealto the theoryof analytic
continuation or else it has to be effected by a completelydifferent method.
Hence, when

we

are

example of the methods


Kapteyn'sfproofof the theorem
As

providedthat

lies in the open

This

domain

occurs

Fig.24
Formula

of

as

give

\/(l z")
"1.
^(l-z"-)

the domain

the

following
analysisthat

it is the interior of the

K;

it is

curve

venient
con-

shewn

" 8-7.
(1) is,of

suggestedby

course,

establish the truth of the

To

shall

employed,we

in which

in
frequently

so

to describe it

exp
l +

to be

i J,,(nz),

domain

'

"

have

that

-^^1

(1)

in

which

an

formula

write

we
expansion,

1 + 2

(2) of " 17"22.

J"{7iz) S(z),
=

and

then it has to be

proved that S{z)

J,,(nz)

,^"

1/(1 z).
-

/"(0+)'exip{^z{t-l/t)\
/

Since

Ztti J
*

For

an

of the magnitude
investigation

(1849),pp. 33"39,
+ Nieiiw
X.

(1893),pp.

t
of

this

domain,

see

Puiseux, Journal

de Math.

xiv.

242"246.

Archief
96"102.

voor

Wiskunde,

xx.

(1893),pp.

123

"

126;

Ann.

sci. de VKcole

norm.

sup.

(3)

560

THEORY

that, if we

see

we

that

circle T with centre

[CHAP. XVII

FUNCTIONS

the

at

originof

such

radius

inequality

it the

on

find

can

BESSEL

OF

exp"'^

(2)

{1^(^-1/0}
^'

"1

is true, then
^(z\
^ '

(^\

^^

To

p, u, a,

'

analysisof " 8'7. If 2


pe'",t e""*"**,
then (2) is satisfied for
being positive),

sin- a)

of ^ if

all values

when

and

(p and

1/^} t

[^z(t

exp

recall the

all real

are

1
r'
./,r+)

27ri

(2),we
investigate

where

l+^~^exp 1^^(^-1/01 dt

J^ i

is chosen

value, this value is

v^sinh-u
that

so

the last

"" "

"

0 ;

the

expressionon

left has

its least

(|8'7)
z

.v/(l z^)

exp

"

14-V-^')
'"8|

is

which
domain

negativewhen

K,

satisfied when

|i |

write

\jtin placeof

^^

where

When

of

\t\

combine

in

1-^
27rzJ(^+)

is the circle
we

value
positive

K.

Hence, when

such

that

the

lies in the

inequality
(2) is

e".

Again,if we
^^

find

can

we

lies in the domain

find that

(3) we

exp

{-1^(^-1/0}

'

e~".

(3) and (4) we

9"^(,\-l^{

find that
*+

exp

{1^(^-1/0} dt

t
^''^'^-27ri\r+,y-)t-eK^{lz{t-l/t)}
'

and

2S{z) is the

so

of

hounded

inside the annidus


We

sum

the residues

by

V and

that there is

next

prove
proved this,we notice that this
For

the number

[
2'iri

of

of

the

integrandat

its poleswliich lie

7.

the annulus*, and,

only one poleinside


poleis obviously^

having

1.

polesis equal to

log[1

t-^

J (^r+.y-)

I/O]]

c^log[1

J_
^

{\z{t

exp
dt

^^

r^

exp

27ri J (r+)

{i^(^- 1/01]

^^

dt

rflog[l-^ exp 1-1^(^-1/0}]

^_l_i

dt

dt

27rl J ir-i-\
{r+)

^log[l-r^expli^(f-l/0}]^^
-m

_L

(r+)

27ri J
The
as

the

dt

"

c?log[-" exp

(r+)

correspondingpart of Kapteyn's investigation


does
investigationgiven in the text.

{-1^(^-1/0}]^^^
dt

not

seem

to be

quiteso convincing

17-31]
Now

vanishes
integrals

the firstof these

t"'

then

I?7|"

V, and

on

for,if

write

we

{^ {t l/t)] U,

exp

consideration

under
expression

the

so

561

SERIES

KAPTEYN

be written

may

in the form

71"^

and

the

of
integral

each

i"
{n=0o

J (V+)

of the

term

dt

series involved

uniformlyconvergent

is

zero.

Hence

the

of

number

zeros

of 1

{^z(t

exp

l/t)}in

the annulus

is

equalto
1

Iiri J
It follows that

r+

1 ; and

this residue is

It has therefore been

[\z{t ijt)]

domain

calculated
easily

shewn

of

{1^(^-1/0}

exp

^-exp
=

clt=l.

^+^

i-2

2S{z) is equalto the residue


^ +

at i

that

to

be

2/(1 z).
"

8{z) is equalto 1/(1 z) throughoutthe


"

the ivhole of the


K, i.e. throughout

open

in which

domain

the series

S(z) is convergent.
defining
[Note. It
of

is

K, except at

possibleto

s=

prove

1,but the

that S

to the

{z)converges

proofrequires

an

appeal

to

1/(1 2) on

sum

"

be made

to theorems

the

boundary

of

Abelian

an

type; cf. " 17 '8.]

to the case
17'31. Tlie extension of MeisseVs expansions

We

shall

^^

Zi

now

shew

how

to

obtain the

expansions

p+r''^(p+n(2^+n

n''+^'

ofcomplexvariables.

(i'+r'^)(^'+r)(3--^+r"
+

^^

"Z, (2/1-1)2+r^

which

are

valid when

12+^2

(p+^2^3.+

lies in the domain

and

...,

^.)

^ is

complex variable

which

is unrestricted apart from the obvious condition that ^i must


not be an
integerin (1) nor an odd integerin (2). These results are the obvious
extensions

of Meissel's formulae

of

"17-23.

[Note. The expansionswhen ^ is a pure imaginaryhave to be estabhshed by a limiting


the imaginaryaxis ; since the functions involved in (1)and
process by making ( a|)[)roach
all
are
functions
of
even
is lost by assuming that li{Q is positive.]
no
(2)
^,
generality
w.

B. F.

36

562

In

order

BESSEL

OF

THEORY

formulae, it is first convenient

establish these

to

[chap. XVII

FUNCTIONS

to

effect the

of the
to complex variables of the expansionof the reciprocal
generalisation
radius vector givenby " 17'21 (6). That is to say, we take the expansion
00

1 + 2 S
n

which

that

by the symbol S{z, ^), and proceedto sum it by Kapteyn's


is a real variable and
(explainedin "17'3),on the hypothesesthat "/"
K. We define a complexvariable i|r
lies in the domain
by the equation
(^ y^
=

arc

sec

is real* if
of these values of (/"

None

which

can

any

Q\n

'^

"^,are

givenby cosi/r \Jz,that


=

isj{z^1 )"
"

"

lies in the domain

is

and,

as

creases
^ in-

through real values,t/tdescribes an undulatingcurve


without
be reconciled with the real axis in the -\/r-plane
passing
points.
singular
0

from

over

"

function of

of i/r,
singularities
qua

The

il^,

denote

we

method

Jn {nz)cos
\

to

GO

It follows that if,for

write

we
brevity,

t^

nexp{i^("-l/0},

then
^"^^'^P^
the notation

with

of

so

shall

the methods

By

of

sum

the residues

hounded

lie inside the annulus

hy

then

havingproved this,we

notice

dt

of the integrandat

T and

of

polesis equalto
dt

dt

'(r+,v-)

c^log(l-2"rcos"^+ JJ')

dt

dt

J (r+)

liri

{r+)

(^log[^^exp{-^(^-1/011
dt
dt

dU

00

f7" COS

+ !)"/)
(??.

n.=0

so

It is easy

that

to shew

1e"^ I" 1.

"

dt + 2
dt

9
that such

vahies

those

of

y.

dlog(l-2t/cos(^+ ?70

"^TTt

have

only two polesinside the anmdus, and,


that these polesare obviously
e""^.
t

the number

By Cauchy'stheorem,

in

we

that tJiereare

shew

now

of that section
l-^''

^^-_L/"

is equalto the
2S(z, (/")

which
its jjoles
We

'

" 17"3.

9cr

and

U' t
27rtj(r+)l-2(/cos(^+

the
of (p satisfy

equation

564
and

BESSEL

OF

THEORY

[CHAP. XVII

FUNCTIONS

that
fin
cosh td

roc

e-^* sin2'"-i'Jrdylr=

Jo

"

p-j

"

cos^'"-!

Odd

cosh|7r^Jo

r{r+ii{?'+3^}...{^^+(2m-i)^r
By

substitution

(5) and (6) and writing2^ for ^ in (5) we

in

at

once

infer the

R{0" 0: and the mode of extendingthe results to


all other values of
has already
been explained.The required
generalisations
therefore completelyestablished.
of Meissel's expansions
are
truth of

(1)and (2) when


'(

The expansion

17'32.
With

of z^

aid of Meissel's

the

into

power
any integral
consider
to
even
powers

In the

of

of

case

even

an

series.
it is easy to obtain the
of a Kapteyn series. It is

generalisedformula

expansionof
convenient

Kapteyn

in the form

and

power, z^^,we

odd

separately.

powers

take the

equationgivenby " 17'31 (1)

in the form
+ l-i^)
f2r{n + l+iOr(n
2'7ri
J r^"-ir{l+iOT{l-iO

...

^ ^

27riJ ",t
r (m + 1 +
1

the contour

of

is
integration

when
uniformlyon the circle,
are
integrations
permissible.
the value

now

^^

^^

l-^"o
1

|^\

io

^.

lies in the domain

^,"_,,_,^.
^'
^
Since

both

series

K, term-by-term

(p+n(2^+n...0i^4-n

polesoutside the contour, and so the contour may


into an
is seen
be deformed
and the expression
to be
infinitely
great circle,
the
circle
and
the
equal to unity;but when m " n,
poles" im are outside the
is equalto unityminus the sum
of the residues of the integrand
expression
i.e.to
at these two
poles,
in

"

n,

there

t'or (m

of

j^f
When

the circle

converge

Consider

m'

,"=i

J- f i ^(.-n+^"D^(.

where

J.mi^mz)

no

are

(m

n) I

"~?/i'^"+'.(m-"-l)!"
The

the left of

on
expression

/9,"^\

(]) is therefore equalto


9

m=i

Next

we

V
"i=ji+i

(m

n)\.L,J27n2)

nv"^^^
.

{m

"

n"l)\

evaluate
1

27rt

^(/^ + l+^"O^0^
il^nH

r{m

l+

i^)r

d^
l-^"O
iO ^^'-'"'+'

(m -h 1

17-32]
When

and, if we take the


originis the onlypole of the integrand,
is seen
the expression
to be equalto 1.
infinitely
great circle,

^ n, the

to be

contour

an

But, when

565

SERIES

KAPTEYN

n, there

"

polesinside

no

are

the circle

{^\

\, and

"

the

is zero.
expression
Hence

If

have

we

replacen by

we

1 and

subtract

the

result

obtained

so

from

(2),we

find that

"

and

(m

W'"--.

vi=n

n)!

m=n+i'^n-''+\
(m

1)!

'

so

If

1, equation(3) is at

deducible

once

from

equation(2),without

the

intervening
analysis.
When

givenby
(4)

have

we

" 17"31

to

deal

with

odd

an

power,

the

equation

27ri

rrZi r

^2m"

00

""

and

take

we
z^^~'^,

(2) in the form

deduce

we

(o)

in

similar

1 yam"

(2m-l)^

2m"

Ji {p+?^H3^+n--K2^-iy+n

J,",_,\{2m -l)z\

'^^'

that

manner

^-.^

"^^

"

,,^-^^

(m-ir.(m-^-l)!
=

2^ +

-t-^-'^-^

Hence

The

formulae

which

is

be combined

obviouslyvalid throughoutthe

values 1, 2, 3,
This

(3) and (6)may

into the

domain

singleformula

when

has

any

of the

"

formula

was

discovered

justbeen given,though somewhat


proof.
*

Ann.

sci. de

by Kapteyn*;
artificial,
seems

I'Ecole

norm.

suj).

the

proof

of it which

rather less so than

(3)x. (1893),p.

103.

has

Kapteyn's

seHes

of the Kapteyn
investigation

The

17-33.

[CHAP.XVII

FUNCTIONS

BESSEL

OF

THEORY

566

for

z'

the method

ii

of

induction.

of z^

shall
as

abstruse

1 + 2

which

now

-^

(1)

"

,"

" 17"3 ; it is,of course,


complex,then z lies in
induction

The

the

1 + 2

-t

",

" 17*22

supposedthat, if

(-)"

and

for

is real,then

(^^)'

"

variables
c"jiplex
1

^^ "

1, and, if

in
is

K.

the domain

which

of

t4-.

J-^{mz),

real variables in

provedfor

were

the expansion
investigaiig
advantage of usig no result more

method*

givean alternative
a
Kapteyn series,which has
than the equations

We

will be used

the fact

depends on

"

it

when

the sum,

00

the

f(z),of

Kapteyn series

series 1

"

"

^can be

"

is known,
UmJmi'niz)

S
m

then

ii

F(z)

sum

of the

obtained

by

two

quadratures,the

former

series

w^

m=i

uniformly.To

converges

establish this result,observe

tha

by

term-

by-term

differentiations.
z^
az

az'

,"=i

F{z)
(z^y

that

so

it follows at

that

once

Now, from (1),we

F{z)

SO,

if

J^" (2,nz)
=

Fiz)= X
111=^1

of

z)

by quadratures.

-1^^($)
.

'^'"^/^mz)^
4"i^

(^^)V(^)
^^,
dzl

then

Therefore,in

where

in terms

have

i
and

(l- z')f{z)
;

be determined

can

a,"J,n{mz),

(1-;-)

and

the domain

K,

of

constants

are

If
integration.
A=B

we

mat

0.

Consequently
(2)

2-

i '^""*(^^"-^)
m=i

Watson, Messenger,xlvi.

m"

(1',)17),
pp.

1.50"157

--*0,

we

see

that

17-33]

KAPTEYN

In like manner,

deace

we

SERIES

from

567

(1) that

-i^^,

:.4"^.,K2m+ i)^}

that

and hence
^ ^

mto

"^

expansionsof

The
Now

z*^

nen

is 1

that, foisome

assume

"

(2m
or

1)^

therefore constructed.

are

value
particular

of ?i, z'^ is

in
expansible

the

form
z^

n''

and consider the

K.^n'J'mi'mz),
1

function^
(z)defined by
(zh (n -f 2)" S
"l"
m

By

the process of ditirentiation

^^

the

equation
J,n(mz).
6,","

^"

alreadyused,we

have

z'
az

az-

,re=i

-.')."
j:-("+2)Mi
=("+2)={.'*
+4^}
(n +

On

we
integration

z''+\
2)-^

dedu-; that
'

{z) z''+-+ ^
(ji
=

It is obvious
the

near

that A'

'
=

0 from

log^

B'.

consideration

of the behaviour

of

"f){z)

origin.

Hence

the

expansioof

is
z'^'^^
30

m=l

6m,n+2

where
follows at

once

b induction

"

That

,_

^^^

^"

lt
ti-

that

"

1)

*"""

'2^'V{m + n)J^{2mz)
*

"Zi {2m)^-'

.nf- r

(m

-n

1)

is to say

equation(c of "17-32.The expansionof


in
from the esansion of z ; the analysis

and this is
same

.,

''"

b^,

"

"":,

,7^i;^=rr
(|m

'n,m

^^^

way
reader.

2-"-'

this

is obtained
case

in the

is left to the

568
We

therefore obtain

/.^

/I

which

the

sn_

r (w + m)

~"''"m=o

Jn+^m \{n+ 2m) z\

(n + 2m)"+i.m!

is the

expansionobtained by other
is valid throughoutthe domain
K.

Since

[CHAP. XVII

FUNCTIONS

expansion

..

^2^^

BESSEL

OF

THEORY

methods

in

"17'32; and

the

pansion
ex-

the series

(?i+ 2mf+i.m!

^=

the expansion(4)
absolutely
convergent (beingcomparablewith S 1/wi^),
The
fore
expansionis therethroughoutK and its boundary.
converges uniformly
in
valid (from considerations of continuity)
the
of
and
K,
on
boundary
K.
at the points2^
+ 1, as well as throughoutthe domain
particular
is

The

17"34.

expansionof l/{t z) in

From

the

deduce,

after

lies outside

and

Kapteyn series.

"

obtained in the two precedingsections,we can


expansionof z''^,
Kapteyn*, the expansionof l/(^ z) when z lies in the domain
"

certain domain

whose

extent

will be defined

later in

this section.

Assuming

|^ |" j^^ |

have

we

_1

Now,

that

^ ^

2"w2

"

gexpV(l

if

z'')\
^y

1+^(1-22)
the

repeatedseries

the double

{n -\-m) Jn+2m {{n+ 2m) z}

is

as
expressible

'''

absolutely
convergent

an

double

series if

series
2'*?i-r

"

{n

m) F"+2^"i

"ti"r-o(" + 2m)"+^m!|"|"+i
is convergent. But
double

the

in this series

terms

are

less than

the

terms

of the

series
X

2FexpF2'"F''+=''^_
m!|"r^
\t\{\t\-2V)
"

'

"

providedthat

Ii

["

i m-o

2 F.

Hence, when
^"exp

t\"2

1 +

\/(l z^)
V(1-^-^) ;'
"

and, when
rearrangement of the repeatedseries for l/(^ z) is permissible,
obtain the formula
we
arrange it as a Kapteyn series,
"

r^

(1)

Ann.

"

0" (0

sci. de I'Ecole

i "Bn (t)Jn (nz),


"=i

norm.

sup.

(3)x. (1893),pp.

113

"

120.

we

17-34, 17-35]
where

KAPTEYN

(2)

"Bo{t)=llt,
v" ('^ 2m)^(n-m-l):

1
(o^

era

From

the last formula

by Kapteyn ;

/A

we

deduce

may

theorem

very remarkable

"4"(M-2m)-.(??.- w-1)!

discovered

have

we

dV-

"4"(^n-vi-l)l

and

569

SERIES

therefore,by " Ol (2),

that,by " 9-12(1),

so

(4)
when

^" {t)
=

?i

1, 2, 3,

(1

+ " cos'^\mr
^2)0" (/^O+ sin^ \n'rr

....

in terms
Kapteyn s polynomial^". (^)i" tJierefofe
expressible
polynomial0^ (nt).
It is

to
possible

now

for,by " 8-7 combined

extend

with

the domain

lie in domains

such

(5)

of
validity

the

" 9-17, it follows that the series

(1) is a uniformlyconvergent series


and

of

analyticfunctions

of

of Neumann's

of

expansion(1);
the rightof
on

and

that

n(z)"n(t),

n(z)"n{i),

n(z)^\ l^^^jMlzA^)|

where
The
of the

expansion(1)is therefore valid throughoutthe


inequalities
(5) are satisfied.

[Note.

This

in which

domains

extensive domain
of values
more
gives a somewhat
he
the
tlicorem proved in " 9-17,and
contemplated by Kapteyn ;
ignored
(sincethe coefficients in the series for (""(0 are positive)
wiien |^ |^ 1,

result

I""(0|"""(1^|)=$""(1)
by (4);

so

that

Kapteyn proved

that

17 "35. Alternative
Now

that

"

explicit
expressionshave

been

l~Z

the reader

will find it

an

Cf.

Kapteyn, Nicuw

observed

was

that

1,

2{\-f^)^
Archie/

voor

z) into

Kapteyn

series.

for the coefficients in the

expansion

Thus, if "" {t)be definedas

ways.

1-)On (n")+sin2 \mT +

f^

than

various

interesting
analysisto
7 4-

obtained

"

possibleto verifythis expansion in


n{\-

|i:|:$:l.]

proofsof the expansioii


of l/(t

of

both

(1) is valid when

0(2)"fl(l),

it is

when

cos^

^rm-,

take the scries

0^ {nt)J" {m\

Wiskunde,

xx.

(1893),p.

122.

570

THEORY

substitute
reduce

suitable

the result to

Or

OF

integralsfor the
1/(^ 2) after the
"

differentiate the

again,if we

222

BESSEL

Bessel

[CHAP. XVH

FUNCTIONS

coeflBcients and

polynomials,and

Neumann

" 9'14.

of

manner

expansion twice

with

respect to

we

find that

=(1
2')J^
^nn^"n it)Jn {m),
(^^^2}

\{t-zf\ {t-zf
and

then,dividingby

whence

and

"z^^ and making

of

use

"

17 "3

find that

(1),we

{t^-\){t-zf

if -\f {t-zf

the differential

equationfor "" {t)is easilyconstructed

hence

{f'-\f{t-zf

it follows that

"n{t)
where

A^ and B^

17'4.

"

nil -f)On{nt)

The

sm^hin-irtco^'^Hit +1-"^ {A^Jnint)+ Br^Yn{nt)],


t; but it does

independent of

are

in the form

not

easy to prove

seem

that

functioninto
arbitraryanalytic

expansionof an

An

Bn=().

Kapteyn

series.

"1

We

shall

now

Let

f{z)

he

{z)^

a,

prove

luhere

Then, at

the

:
expansion-theorem
following

functionivhich

^ 1.

all pointsz inside the

analyticthroughoutthe regionin

is

ivhich

region,
-A

(1)

f{z)

i:

Oo +

aJn {nz\

xvhere

^=^.\(Bn{t)f{f)dt,

(2)
and

the

is tliecurve
path ofintegration

This

result

is obvious

when

on

ivhich il (t)

substitute

we

a.

the

uniformlyconvergent

expansion
n

for

l/(t z) in
"

the

equation

J^'^-^iri]
t-z
since

D,{t)
=

on

the contour, while

is inside the contour.


This

theorem

is due

to

Kapteyn.

both

'

D,{z)"l

and

Q,(z)"Cl (t)when

572

THEORY

OF

BESSEL

It is not

difficult* to express S4n,v{t)


in terms
An,v{nt-\hvt),defined in " 9"2.
And
in

the reader will

that

easily
prove

[CHAP. XVII

FUNCTIONS

of

Gegenbauer'spolynomial

if/(^)satisfiesthe

conditions

specified

" 17-4,then
S

z^f{z)

(4)

a",

/,+,,
{(v+ 71)z]
,

where

in which

the contour

of

surrounds
integration

(v +
_1 "^''

(6)

2m)- r(v

n-

the
+

origin
;

n-

m)

an-

and

hence

-2m

".

where

cio,a^,

are

...

[Note. Jacobi

in

the coefficients in the Maclaurin


of his later papers,

one

Math.

Astr. Nach.

series

for/(2^).

(1849),col. 257

xxviii.

270["'es.

"

Werke, vii. (1891), pp. 175


188] has criticised Carlini for statingthat certain
valid
when
But Carlini had some
for his statement
"0'663
are
excuse
expansions
only
\z\
because the expansionsare obtained by rearrangements of repeatedseries which are
valid throughout the
permissibleonly in this domain, although the expansionsare actually
domain
K.']
"

Kapteyn series of the

17'6.

Series of the

second kind.

type

+")
").}
J-.^IC^^
2/3../^.{(^"'
.}
+

have

been

studied in

detail

f. But the onlyseries of this


|, are some
importance
type which have, as yet, proved to be of practical
s
eries
with ix
results required
in
special
v, and with simplecoefficients. The
the applications
obtainable
Meissel's
are
justspecified
expansion
by integrating
It is thus found that, throughout
of I 17"31 (1) after replacing
z by 2 sin ^.
some

by

Nielsen

the domain

K,

\l

^^"'"''
="? (p.p)(2'4...(..'+r)-^

that

so

^^

and
while
manner

Zi
hence

deduce

we

the

2 P-f

^that

of series of

sum

from

n' +

the

Cf.

Ann.

Cf.

^""^2.4'(l+ ^2)(22
+

is a polynomialin z'S,Jn-{nz)ln-'"'
the type 'Zn^^Jn^(nz)may be found

of

in

expansion'2n^^"'J.2n{2nz).
corresponding

Nielsen, Ann.

sci. de VEcole

sci. de VEcole

norm.

sup.

norm.

(3)xviii.

Schott, Electromagnetic Radiation

sup.

(3)xviii.

(1901),p. 60.

(1901), pp. 39

"

75.

(Cambridge, 1912), Chapter

viii.

degreem;
a

similar

17-6, 17-7]
Thus

SERIES

KAPTEYN

Schott*

has shewn

that

(2)

J^'{nz)
^
'

generaltheory resemblingthat

which

is

seems

in

easilyderived

from

"

17 "5

(1) and

in the unlikely
event
constructing,

Kcipteynseries

17'7.

Im

have

that

seen

function

throughout the

the series may

it resembles
a

As

an

Fourier

series

example,let

should

have

but

pansion
ex-

it

difficulty

no

K.

domain

%anJn{^^z)represents

series

K.

But

since,when
real

not

are

analytic

an

is real,\Jn{nx)\"1,

axis,althoughwhen

\z\" 1,

the real axis.

on

be summed

up^f*
by saying

domain

from

to

to
(1/^')

tt.

and

that

it

sembles
re-

the real axis outside K.

on

consider

us

throughout K\

outside the domain

Kapteyn series may


throughout the
power-series

the

|v/"n| l,

Kapteyn

of such

from

requiringthem.

alongthe whole of the


converge at pointswhich

behaviour

The

is valid

converge

the series does not

that

the

is deducible

the reader

of his

wJiich converge

If

we

" 1 6'14

of

to go into details which

unnecessary

z^)

7i'JJ{nz)=^

2'

2^/(1-22)
2v/(l-22)
zH4

(3)

[Tfy

"

n=i

573

the series

^ Jn jnx)

then
that,if (^ -v/^^ sin -v/r,

It is evident

"

-^Jo

series is

since the Fourier

Now, when
then

increases to

such

that the minimum

ir

as

n-

uniformlyconvergent.

decreases
"1, cf)

"

yjrincreases
value

of

yjrincreases

as

from

arc

cos

lies between
(f"

"

and
arccos(l/a-)
If vi be the integer

and

2m7r

"

{m

to the values
the values of -\^corresponding

0,
of

be
"/)

7o, 7i,

27r, -47r,

"liHTr,-2??i7r,

...,

27r, 0

S/",^,/(-i, "1,So,and then

jm,

...

...,

...

TT

(,. 0 J7r
=

"'7m

()

J S,.+l

-Soin-l

(Cambridge, 1912), p. 120.


made
by Professor
suggestion of these analogies was

M"

Electromagnetic Radiation

t The

Hardy.

1) tt,

let

574

OF

THEORY

Now

when

BESSEL

[CHAP. XVII

FUNCTIONS

8r+i)the
yjrlies in the intervals (jr,7r+i)and (S,.,

of the

sum

the integral
signis

series under

+
Ittc^
i"f"'--

Itt^ +

(r + 1) tt^ + (r + 1) ircf),

and, since

l(-^
"

sin

{(yfrX
"

it may

sin t/t)
+ a;
fZ-v|r
^-\/r-

yjr,

cos

sin ""^)+^x^ {-^jr


sin yfr
+ 2a; (i^r
cos
COS "\//'
-^/r),
i/r)^
c^-v^^^i/r*
=

be shewn

"

without

much

"

that
difficulty
m

+ \x
:^x'

{I(8,- 7,^)+

(cos8^

7,)}+

cos

'2'ir^

The

reader

will

that

see

{h, 7,).
-

j-=i

,.=0

largeclass

of

series may

Kapteyn

be summed

by

this method*.

17*8.

The convergence

series

of Kapteyn

on

the

boundaryof the

domain

K.

exceptionof the points + 1, the boundary of K presents no


of Debye'sasymptotic
features of special
interest;because, by means
expansion
of the Kapteyn series Son Jv+n [{v+ n) z]
the consideration of the convergence
With

the

is reducible

to

(^exp V(l -g-)l"

On

^
"

Vn
that of two

and

The

similar

points +

by

'

interest,because

more

of

for ijn.

the

ordinaryasymptotic

for real values of


therebyproducedis filled,

the lacuna

theorem
following

The convergence

\/(l-z'))

1 +

written
seriesfwith \/'^^V'*''

present

fail. But
expansions
the

series

that of the power

Abelian

an

v,

type:

of

both the convergence of % a^Jv+n (^ + ^^)(^f^dthe continuity


to ensure
sufficient
S
0 ^x^\.
of anJv+n {{v+ n) x] throughoutthe interval^
is

is monotonic, with a limit as n-*oc


+ 7i)|^
{n/{v
it follows^
that 'S,anl(v
+ n)^converges; and since,by " 8'54, (v + n)^Ju+n(v+ n)
it follows that %anJv+n {v + n) converges.
is monotonic, with a limit as n
x
Since

lajn^

converges

and

In

this connexion

1901, pp. 127"146,


t If

2"aJ^n^
of the

the

should

researches

by Nielsen, Oversigt K.

aj^n does not tend to


is absolutely
convergent,

asymptotic expansion with

the

zero

and
a

series cannot

so, if

we

remainder

J Due

allowance

has

to be made

for the

Theory of InfiniteSeries, "

converge

replaceeach

Videnskabernes

Bessel

; and

if it does

function

term, the series of remainder

convergent.

" Bromwich,

Danske

Selskabs,

be consulted.

originif
19.

v"0.

by

tend

the first two

terms

is

to

zero

terms

absolutely

17-8]

n)x]

J.^^\iv
+

since

Again,

of

function

is

in

the

interval

does
it

O^ic^l,
*

of

which

increase

not

follows

from

Sa"/^+"

uniformly
^

a-

this

and

sufficient

and

of

theorem

Bromwich,

This

(1917),

pp.

was

reasoning,

2a,i/n^,

does

so

for
its

Abel's

of

type

all
for

of

values

uniformity

the

that

so

[{v

n)x\

interval

the

throughout

continuous)

theorem

it

may

the
to

be

shewn

of

convergence
be

true;

that

the

if

^a^Jv+niv

^dnju-^

theorem

is

is

both

therefore

n)

verges,
con-

necessary
the

best

kindf.

Theory
pointed

171"174.

test

for

increases,

theorem.

the

proves

the

reversing

By

the

therefore

(and

convergent

as

that

convergence

is

575

SERIES

KAPTEYN

of
out

Infinite

by

Series,

Professor

" 44.

Hardy.

Cf.

Watson,

Proc.

London

Math.

Sac.

(2)

xvi.

XVIII

CHAPTER
OF

SERIES

Fourier

18* 1.

consider

to

in the

the

AND

formal expansion of an

his researches

In

FOURIER-BESSEL

the

on

of

expansion

arbitraryfunction.

Conduction

of

Theory

of

arbitraryfunction

an

DINI

Heat, Fourier*

/(ic)of

led

was

real variable

of

,r

form
00

the

denote

ji,j^yjs,

where

of

of Jo

in

(2')arranged

ascending

magnitude.
necessityof expanding

The

Bernoulli's

also in Daniel
Euler's

a,nJo(jm'^),

positivezeros

...

order

/(")=

(1)

problem

of

problem

of

vibrations

of the

arbitrary function

an

chain

this

membrane

arises

manner

under
oscillating

circular

in

and
gi-avity

with

in

initial arbitrary

an

symmetrical displacement (""1'3, 1"5).


order

In

to

determine

sides of

both

"

from

It follows

coefficients

the

a,"

the

limits

plied
multi0 and

1.

that

5"11

l^^i'

inferred

Fourier

hence

expansion,Fourier

(1) by ocJq {jmx) and integratedbetween

and

in the

\Jm)"

"

ti,

that
'

(2)

a,,

j-^^ \

If

magnitude,

of

the

symbols J",,so thatf j'l,


jg, Ja,

...

function

of the

positivezeros

the

denote

f{t)Jo ijmt)dt.

of
significance

the

change

now

we

' 0

fflKJm)

J^ {z),arranged in ascending order

then
00

/(";)= 2

(8)

a,nJ,(jmOo),

m=l

where

("i)

^, .l\f(t)JAjmt)dt.

am=

This

been

indicated

the

coefficients

exists and
*

La

f The

is

to

are

stated

specialcase

validityof

determined

be

by Lommel;]:; but, of course, neither


has
0 does the procedure which
z-=

the

expansion; it merely
the

on

hypothesis that

indicates

the

how

expansion

uniformly convergent.

omission

X Studien

the

Jm)

was

in the

establish

Thiorie

considerabh'

nor

v+i

general result

more

general case

in the

"

J,

tf

Analijtique de
of the

improves
iiber

die

the

suffix

la

appearance

BesseVsclien

(Paris,1822), ""316"319.

Chaleiir

v,

associated

with

ji J2, js,
,

of

the

Functionen

should

cause

,..,

formulae.

(Leipzig, 1868), pp.

69

"

73.

no

confusion, and

it

18-1]

577

SERIES

FOURIER-BESSEL

for the
of the procedureis somewhat
deceptive;
simplicity
reader might anticipate
that,if the function /(a;)is subjectedto appropriate
the expansionwould be valid for all values of v for which the
restrictions,
integral
In

fact the

I tJAjmt)JAint)dt
Jo

is convergent, i.e.when

v^

1.

"

the
of his researches on
pointedout by Dini in the course
the
that
establish
it
to
seems
hypothesis
impracticable
except on
while
and
various
i'^
^;
provingtheorems on
subsequentwriters,
although
the hypothesisv^
^ have stated that the extension of the analysisto
values of v between
1 is merelya matter
of detail,their statements
| and
appear to be open to question.

But,

as

was

it
expansion,
"

"

"

"

fir.stattempt at

The

rigorousproofof the expansions(1) and (3) is


contained in some
notes
humously.
compiledby Hankel* in ]869 and publishedpostA
a
was
Schlaflif
more
given
by
completeinvestigation
year
after the publication
of Hankel's work; and an importantpaper by Harnack:^
differed
which
of the expansion(3) by methods
contains an investigation
from those of earlier writers.
appreciably
a

the

and Schlafli,

of the researches of Hankel

A few years after the appearance

generalexpansion

more

CO

f{^)=

(5)

where

Xj, "2" ^a,

"""

denote

the

KJA^ni^X
l

(in ascendingorder

zeros
positive

of

magnitude)

of the function

when

v^

h,and H is any givenconstant,

"

coefficients in the

The

expansionare givenby

JJ^(X" ) +
{(X"r v-")

(6)

by Dini",
investigated

was

X"r J;^

(X^)}6^

the formula

2\^'

\ tf(t)J. (X,"i)dt.
Jo

The
known

'

as

of the

of determination

mode

\n

boundary condition,'namely

mixed

vanish at
formally

numbers

a?

subjects/(*")to what is
/'{x)+ Hf {x) should

that

1.

The^xpansion (5) was examined by Fourier (when v=i}) in the problem


when heat is radiated from
of the propagation
of heat in a circular cylinder
of H is the ratio of the
in this problem the physical
the cylinder;
significance
of the cylinder
to the internal conductivity.
external conductivity
*

491.
In the
(1875),pp. 471
of
case
as
a limiting
integralformula
"
(3).
t Math. Ann. x. (1876),
pp. 137"142.
% Leipziger Bcrichte,xxxix.
(1887),pp. 191"214;
" Serie di Fourier (Pisa,1880),pp. 190"277.
Math.

Ann.

viii.

of

w.

B. F.

"

coarse

of this paper,

Hankel

obtained

14-4

3Iath.

Ann.

xxxv.

(1889),pp.

41"62.

37

the

578
It

pointedout

was

the insejxion of
contams

which

which

have

expansion(5) must be modified* by


H + v
0: and, althoughDini's analysis
it advisable to
to make
discoveryseems
Fourier's with the expansion.

Hobson^:,who

have

integral
equationsas
On

aesthetic

because

it

theorems
On

are

give rise

The

use

of

set

plex
com-

possible,
by Kneserf
the
expansionby using
theory of

the

is

far

so

great deal

unnatural

as

to

use

be said for this

to

complex variables

procedure,
proving

in

functions of real variables.


concerning
based on
the theoryof integral
equationsare
mmd
of the ultrain the
of suspicion
feelings

theorems
essentially

the other hand, researches

liable to

on
depend ultimately

basis.

grounds there

seems

described

abandoned,

constructed

somewhat

which

been

Cauchy'stheoryof residues.

variables has, however, been


and

now

provedby

are

[CHAP. XVni

FUNCTIONS

that the

this

error,

name

BESSEL

when

rather than

researches

of lemmas

Dini

initial term

an

associate Dini's
The

by

numerical

OF

THEORY

to

uneasy

mathematician.

orthodox

more
complete by the
theory has recentlybeen made distinctly
of W. H. Young|, who
has thrown
new
lighton many
importantmemoir
of
the
modern
the
of
subjectby using
knowledge
theoryof functions of
parts
An earlier paper
real variables in conjunction
with the calculus of residues.
less synthetic
which makes
some
by Filonjl
appreciably
parts of the analysis

The

also be mentioned

must

here.

The

of term-by-termdifferentiation of the
questionof the permissibility
expansionwhich represents a function as a series of Bessel functions has been
discussed
by Fordii, who has obtained importantresults with the help of
(cf " 18"4).
quitesimpleanalysis
More

recondite

due to C. N. Moore**, who, after studying


are
investigations
the summabilityof the expansionby Cesaro's means,
the
has investigated
uniformityof the convergence of the expansionin the neighbourhoodof the
and also the uniformityof the summabilityof the expansion(when
origin,
not necessarily
convergent)in this neighbourhood.

uniformityof the convergence (or summability)of


the expansionin the neighbourhoodof the originneeds rather special
sideration
conwhich
is that it is necessary to use asymptoticformulae for J^,{Xm^)
valid
when
are
\n^ is large;and, as x approacheszero, the smallest value of
is continually
are
increasing.
significant,
m, for which the asymptoticformulae
The

Details

of necessary

also noticed

was

Archill

Proc.

London

Ibid.

modifications

Kirchboff, Berliner

when

will be

^0

American
X.

Math.

Soc.

(1909),pp. 391"435;

iv.

"

The

modification

524.

Cf.

"" 19-21"

(1903), pp.
xii.

given in " 18"3.

Sitzungsberichte 1883, pp. 519

undPhys. (3)vn.(1908),
pp. 123"133;
Math.
Soc. (2)vii. (1909),pp. 359"388.
(1920),pp. 163"200.

(2)IV. (1906), pp. 396"430.

IT Trans.
**

by

the

derMath.

" Ibid. (2)xvin.


ilIbid.

why

reason

Math.Ann.i.xiu.

(1907),pp.

19-24.

178"184.

(1911),pp. 181"206;

xxi.

(1920),pp. 107"156.

477"524-

580

THEORY

and

0^

I f (t)cos mtdt, b^

"

TT

[CHAP. XVIII

FTr"''CTIO"'S

the coefficients

series in which

Fourier

BESSEL

OF

"

.'

IT

-s-

expressedas integrals,

are

r\t)sin

mtdt.

J ^^-^

It is necessaiy to make
between
the types of series
a similar distinction*
which will be dealt with in this chapter:any series of the t\-pe
X

in which

the coefficients a^

be called

series

If,however,

of Bessel

merely form
functions.

coefficients in this series

the

Oi-m

(0,1),the

expressible
by

are

series associated with

if,farther,the series converges

the interval

of constants,

will

the formula"

j\f{t)J,{j^t)dt,

"

the series will be called the Fourier-Bessel


And

given sequence

the

to

series will be described

for any
the Fourier-Bessel

f(x)

sum

as

f{x).
pointx of
expansion

offix).
In like manner,

the series

"=1

where

Xj, X", Xj,

the

are

...

will be called Dints

series

zeros
positive

of

of Bessel functions.

If the coefficients 6,"are

determined

by

the

formula^

j/^ (xj] b", 2x"^


^) J,2(x^) + x.,^2
[(x,";=

tf{t)J, (X" 0 dt

JO

the series will be called the Dini

series associated

with

f (x).

And

for any pointx of the


if,further,the series converges to the sum/' (jr)
interval (0,1),the series will be described as the JXni expansionof f{x).
Some

writers

have

merely a specialcase
there
view

certain

are

somewhat

18 12.
There
a

assume

One

been
of Dini

distinctions

regard Fouiier-Bessel expansionsas


obtainable by making H
x
: but
expansions,
between
the two expansions
which make
this
to

-*

misleading(cf ^ 1826, 1834,

Specialcases
are

inclined

very few

of Fourier-Bessel

Dini

expansions.

expansionsof simplefunctions

in which

the coefficients

simpleform.
function

whose

expansionhas simple coefficients

and

18-35 ).

The
greater part of the terminology is dae
1920), pp. 167"168.

It is

supposed that the integralis convergent

It is

supposed that

the series is modified,

as

to

Young,

for all

Proe.

has

London,

alreadybeen

Math.

positiveintegralvalues of

in " 18'34, when

fl

t-

^ 0.

Soc.

m.

(2) svm.

18-12, 18-2]
in " 15"42.
investigated

"r

"_

the formal

expansions

2J,{jmCC)

oc

(I)

givesrise to

is x", which

Another

581

SERIES

FOURIER-BESSEL

\JmJ

\Jm)

v+\

^"=2

(2)

m-

It will be
0 ^

.r

1, if iT

s:

The

that (1)is valid


subsequently
Cf. "" 18-22, 18-35.
" 0.

seen

reduction

z^

when

0 ^^

1, and

"

(2)when

formula

.'""

"'

easilyestablished,so that the Dini expansion of .r"*-" may be determined when


positiveinteger.The Dini expansion of .r^ ^'i i xnay similarlybe determined;
the generalcoefficient is expressible
of known
functions and
in terms
case
is

is any
in this

/:
0

In order

calculate this when

to

is

an

McMahon*
integer,

has

proposed to

tabulate

the

fuiiction
'

/"*

which

is

specialform

of

+ J-,
,/i(.r)
(.^")
+ Jr,{.v)
4-

of Lommel's

one

of Hankel

The methods

18-2.

{t)dt

and

functions

variables

of two

(""16-5,16-56).

Schldfli.

which were
described in " 18-1 are based on the
investigations
in
his
researches
series of
on
by Dirichletf
trigonometrical
of proceedingis obviouslysuggestedby the
Fourier's type; this method
series are
fact that the trigonometrical
specialcases of the Fourier-Bessel
obtained by givingv the values " \.
expansion,
earlier

The

analysisused

In the

of Fourier's theorem, to prove

case

that

00

cos
/(*") 1*^0+ ^ (a,rt
=

where

","

6,"sin "i.r),

f(t)sin mtdt,

6,,,,
=

that

"

{^+

cos

{x

t)-h cos

"

2{oc t)+
"

...

cos

{cc t)\J(t)dt,
"

2^./,.
sW^'Iq f(t)dt
/W=, .'!!",

.....that

Pioc.

1 f""

f (if) lim

7*?,^

f (t)cos mtdt,

it is sufficient to prove
^-^

American

in
" 10-74(3)

t Journal

Assoc.

conjuuetion
far Math.

1900, pp.

with

iv.

Table

I.

(1829),pp.

42

"

43.

The

(pp. 666"697)
157"169.

tabulation
;

see

is most

Table VIII.

simply effected by using

582

OF

THEORY

the

In

i^^^%^!\f(t)JAj,J)dt,

lim
it is necessary

so

the corresponding
expansion,

is

limit to be evaluated

and

[CHAP. XVLQ

FUNCTIONS

generalFourier-Bessel

of the

case

BESSEL

behaviour

the
investigate

to

of the

sum

I 2J^{j,"a;)J^(jmt)
m

when

is

In

the

that
investigation

the

use

of the calculus of

desirable.

of Dini's

case

examination

An

than

more

\.lm)

'v+i

"

it is in this

large;and

residues is

expansion,the correspondingsum

which

needs

is

of
application
shews

the calculus of residues which

in

""18'3

"

cussion,
amenable
to disreadily
and so we
are
spared the necessityof repeatingthe whole of the
analysisof the Fourier-Bessel expansionwith the modifications appropriate
to the more
generalcase of the Dini expansion.
18"33

The

18*21.
We

that

will be described

the

difference of the two

is

sums

integral.

contour
Hankel-Schldfli

beginthe attack on the problem of Fourier-Bessel expansions


of the function Tn (t,
x), defined by the equation
properties
by discussing
/I

shall now

/J.

^\

0"a;^l,

and

O^^^l,

j:^
\

method

which

will be

of the details of the

many

function

The
Bessel

"

the order
v

The

"

\Jmt)

J-nV'yX)"

where

v\.1m^)

-""

rn

\^)

used

Tl~\

v+\

is real and is subject


to the condition

l^O.

is due

Hankel*

and

Schlaflif,
though
memoir.
recent
suggestedby Young's;):

are
analysis

to

Tn (t,a;)is obviously
as fundamental

expansionsas

'

\Jm)

in the

theoryof

Fourier-

is the function
sin

(n 4- ^) {a: t)
sin ^ (iJO"t)
"

'

in the
In

special
theoryof

Fourier

order

the

to

obtain

series.

formulae

connected

with

requiredit
Tn (t,
x)
which

as

has

the

is necessary to express the mth


residue at jm of a function,of the

at ji,jo,jo,...jn.When
ipoles
*

Math.

t Ibid.

Tn (t,x) which

Proc.

Ann.
X.

win.

(1875),pp.

this has

terra

complex

been

471"494.

(1876),pp. 137"142.

London

Math.

Sac.

(2)xviii.

(1920),pp.

of the

163"200.

done,

are

sequently
sub-

sum

variable
we

for
w,

express

18-21]

FOURIER-BESSEL

583

SERIES

of this function round a rectangle


of which one
of the
integral
sides lies along the imaginary axis while the oppositeside passes between
then moved
off to infinity
to the real axis are
jn and jn+\-The sides parallel
"in opposite
the convergence
of the integral,
directions,
so that,in order to secure
it is necessary to prescribe
the behaviour
of the integrandas |/ {w)| oo

Tn (t,x)

the

as

There

are

(2)

which
three integrands

shall

we

study,namely

[tJ^{xw)J,+,(tw) xJ^ (tw)J,+,{xw)}f{{t''


x')J^^(w)},
-

J^ (xw) Y^ (w)}J^ (tw)jJ^


(w),

(3)

irw

[J^(w) Y" (xw)

(4)

TTiu

Y^ (w)}J^ (xw)/J^(w).
{J^(w) Y^ (tw) Jy (ttv)

The

first of these

and

of Kneser

suitable for discussions in which


and

(4) when

which

x=^t and

0 "x

other

described

was

two

in

"

are

15'42.

integrandsindicates that (2) is


+

2; (3)when*

t "

0 -^ ^"^"

1 ;

^x"t"l.

proceedto verifythat

We

Carslaw

asymptoticvalues of these

study of the

the
integrandstudied by Schlafli;

the

was

the work

suggestedby

"

the

the

all have
integrands

same

residue,namely

^Ju (jmX)JAjmt)/J\+i(jm),
In the

w=jm-

at

case

of

define the

(2),we

functionfg(w) by

the formula

2w

(5)
and

then, if lu

(w)

[tJ^(xtu)J^+i(tiv) xJ^ (tw)J^+i(xw)},

+ 6, where
;"",

6 is small, we

have

+
J. (tv) dJ,( jm) + iO'J," (j,")
=

WJ,' (W)

+ d'JJ (j,n)
e"-j^JJ^
[jmJu" (jm)+
(j",)

using Bessel's differential

It is easy to verify,
by
the rightvanishes; and
of 6-' on
at

...,

that

so

hence

JJ

(jm)]+

"

"

""

that the coefficient


equation,

the residue oi g

(w)l[wJj^(iu)]

j,nis

f)'{Jni)l\jmJJ'-(jm)\,
and

this is

easilyreduced

to

2/^ (jmX)Jv (jmt)IJ\+i


(jm)
formulae.

by usingrecurrence
In the

case

of

residue at

(3),the

j^

is

[J"'
T^i-n
(jm) Y^ (j,nX) J, (jmX) Y" (j^)]J^ (jmt)IJJ(jm)
(jm)
-jrjmJv.
(jmX)Jv (jmt)Yv (j"n)IJJ
2 J^ (jniX)
Jv (j,nt)IJJ'
(jm),
-

by 1 3"63, and
in the
*

same

This

this is the

required;
expression

integrand(4) is dealt

with

way.

is most

easilyseen

by writingthe integrandin

Hv^) [xw]
{HyC")
(ic)
l-rriw

"\The

the

results obtainable

Hut})

if^(2)(it;)
J.,T^(tw)IJv(iv).
(xi(7)

by using the integrand (2) are

Gubler, Einleitung in die Theorie

der BesseV schen

the form

discussed

Funktionen,

i.

in great detail

(Bern,1898), pp.

by Graf

131

"

139.

and

584

THEORY

We

take

next

the

" Bi, An " Bi, where

at

of

contour

[CHAP. XVIII

FUNCTIONS

BESSEL

OF

to
integration

will be made

tend

to

be

with
rectangle

to

and

qo

An is chosen

it is desired

jn

When

"

to

shall take it to be
n

Now
and

so

so

definite value to ^",

assign
An"jn+iwhich
lies between
+
to(n
^v+
^)7r,
equal
is sufficiently
large("15'53).

that

vertices

j" and

we

jn+iwhen

odd functions of w,
that the three integrands
it is easy to verify
are
vanish *.
the three integrals
alongthe left sides of the rectangles

and either
Again, if io= u -\-iv,it may be verified that when v is large,
while u ^ 0, then the three integrands
are
or negative,
respectively
positive
0
and

0 (e-""-^)l"i),
{e~i^-^-t'\% 0 (e-*^-*'!"!),

alongthe upper and lower


assignedvalue of ^,i, the integrals
and t have the relative
when
oo
x
rectangletend to zero as B
have alreadybeen specified.

so, for any

sides of the
values which

We

thus obtain

(6)

r"(",.)-

the three formulae


1

i"-"-'?(''i*"

rAn+'^i

ri

0wJ^ (xOw) J^ (tdw) dddw

Jo
~27ri.'^"_aoi

J"-=(w)
{"d"x-\-t"

(7)

Tn (t,x)

^.

[J.{w) n {xw)

2; xi^t)

''^)-7
^"^

J, {xiu)F, {w)]

(0"t"x"l)
Tn (t,X)

(8)

^.f^"^"u{J.M

'^-'^^

Y.itw) J. (tw) F. (w)]


-

{0"X"t"\)
equation(6) it is easy to obtain an upper bound for jTn{t,x) !; for
is posiit is evident from the asymptoticexpansionof " 7"21 that, when
;^ + -|tive
exist
c
onstants
(or zero) and bounded, there
positive
d and Cg such that
From

'

J (tw) I"
^JAtw)\^

(^"i
(y)

when

exceeds

is
a

on

the line

value

which

g^expli7(^^)1}
^^l^^^i
joiningAn
depends on

"

oc

v.

c,_exp[J/(w)||

u.wi^

to

An

^,^j
cci and

f^O, providedthat

Hence

\it'-^')Tn{t,x)\^^J'j^^^^j\x
so

that
4r

This
*

tends

inequality
givesthe
It is necessary
to

zero

with

to make

the radius

an

upper

bound

indentation

at

of the indentation.

in
the

question.

origin,but

the

integralround

the indentation

18-22]
It is also easy

f r+i

to

585

SERIES

FOURIER-BESSEL

that

see

dt
T, (t,x) (t- .^"-)
-

Jo

{tJ, (xw) J,+,(tw)


and

.r

J^+i{xw)]
J,+i{tiu)

hence

by a consideration of integralsinvolvingBessel functions


in view of the fact that the
the origin,
of the first kind only can
usuallybe made to cover
(11) may be
constant
" 1. Thus
Cj in equation (9) is independent of t in the interval 0^"
[Note.

obtained

Theorems

written

\
valid

when

of the

TTCo

This

O^t^l.

O^.r^l,

involvingfunctions

is not

extension

kind

second

have

Jin [Z

been

so

X"

"

I)

integrals
easilyeffected when
the simplestinequality

because

used

correspondingto (9) is
! Y^ (tw)I" c'l'
{1tw !"" logIiw I+ 1tic \-i}exp {[7 (tw)\},
and

it is

(8)from

tedious

somewhat

matter

to

obtain

simple

bound

upper

inequality.]

this

Equation ("j)was

used

by

Schlafli to prove

sin-4"(^

that, when

sin

but, since the order of magnitude


evident, we shall next evaluate some
difficultiescaused

by

{t

"

sin i tt

"

of the

then
large,

is

sin -rl,j(^
+ ^)

^)

^"^''^"^'^2;^)
x)
^TT
which

integrandin

to the

(i+ x)j

in this

error

'

is not
approximation

integrals
involvingTn {t,x) by

the unknown

error

of

means

be evaded.

may

'

18"22.
The

Tn{t,x).
Integrals
involving

two

fundamental

formulae

'
'

which

shall

we

obtain

now

(1)

lim

[t^+' T,,{t,x) dt

(2)

lim

I't''+'Ta{t,x)dt ^x''.

From

these it is obvious

In the

of

course

'

(0 "

a-",

-^

z)

when

[f^'

Tn (t,x) dt

^x".

is any

1)

(0"x"l)

be apparent that

! t''+'T"(t,x)dt-^x''+i
lies in the interval
0 ^^'^

w^here A

"

(0".r"l)

"

"'"'

"

Jo

uniformlyas

'

proving(1) it wall
xi

follows

that

lim

(3)

as

are

number.
positive

A,

"
,

.,

586

OF

THEORY

We

shall also

BESSEL

boundedness

the
investigate

[CHAP. XVIII

FUNCTIONS

of

I
Jo 0

in the interval in which


Of
pp. 173

^174,and the proofof it,which

though

new,

^1.

given by Young,

results,(1) was

these
"

0 "t

it is contained

It is evident

will

Proc.
be

now

Hobson's

in
implicitly

of

manner

the

former

indentation
Since
reduces

on

the

to

Formula

given,is his.

(2) xviir.

(1920),

(2) seems

to

be

memoir.

r=^^

the

two

rightof

rightinto
equalto

odd function

the

C^n-^^^i2J

2711 J
An-

be

to

avoided

by

an

of w, the value of the first integral

integrandat

/,

the

1'-^'^+'^'
dw
"2J^{a;w)

the residue of the

after
integral

contour

the originhas
integrals,
the imaginaryaxis.

integrandis an

iri times

2J^(a;w)dw

of these

^lA^cc)

the

on

sum

find that it is

1 18'21, we
1

In

the

transform

we

Soc.

Math.

that

f^"'+1
T^ (t,
x)dt=
When

London

so
origin,

that

(xw)diu

WJ^{W)

^i

Now
^n+^i

2J" (xw) dw I

r"*

2c
2ci

"^^-r

exp

{" {\

x)\v\} dv

"

4ci
CiAn (1

x) \Jx'

"

and, from this result,(1) is evident; it is also evident that

x^lt--^^Tn{t,x)dt-x^-^
Jo
tends

uniformlyto

zero

It will be observed

as

-^

that the

x\

was

formallyobtained

Formula
details of the

be

(2) can

proofare

functions of the second

in

kind.

A.

more

^y
\
^\'
\Jm)

Jm**

(0^^"1)

v+i

" 18'12,is an

proved in

rather

"

m=l

which

0 ^^."r^ 1

importantexpansion

x^-^i

(4

longas

so

somewhat

immediate
similar

elaborate)
by usingan

It is easy to

see

consequence

of

(1).

(though the
integrandinvolving

manner

that

S ^''''^''^''Uf'"f':+^(J^-00)
rt''+'Tn(t,x)dt=
Jo

Jm"

w=l

B^x

^^

A"-Bi

f+iiJm)

Jv{W)

588

THEORY

BESSEL

OF

close the range of values of


that the integrals
infer fi-om the Lemma

Moreover, if we

we

[ V^'

["v"^^
Tn {t,w) dt,
bounded

as

when

oo

??. -^

the

on

that 0

rightso

XVIII

"

a;

1,

Tn {t,
a:)dt

Jo

Jo
are

[CHAP.

FUNCTIONS

"

$ 1.

;"

shall consider

Lastlywe

t''+'Tn(t,x)dt,
0

shall prove that,when


functionof n, x and t,as n
and

we

co

-^

"x

It is easy to shew
+ t'^l, i.e.when

by

^ ^ 1 and

"

0 "

is a. bounded
1, this integral

^'

the methods

which

have

used

just been

that, when

then

t^x,

t''+'Tn{t,x)dt
i

Jin" v+iKjm)

lim

{J" {w) Y^ (xtu) J" (xw) J\ (w)]


"

-^

^*

J5-*oo

A"-Bi

lim

b^^

V""^

"/l\

hm

\AJ

B^^^TTIXfv+h

hm

^r"r-,

lirix'!

s^ao

These

are
integrals

section ; and
and

1"

"

To

so

1+^

prove

"

'.

"

"-

J An-Bi
(jQs

lyj ^tW
^

"

Bi

that the

"

r-^
"

il/TT
^
"

,
dw

{tt)

Itt)

.""-'dw.
,

cos{w

"

hi"'rr
"

\7r)

in the Lemma

originalintegralis

ic^l, i.e.when

XW

"

^tt)

"

-,

i^lV \vTr

COS

-.

of the type examined

the

w"lvrr
^"

"

rAn+Bi

J A"-

tw

,..,^^^

{iu-lv'Tr-\ir)

COS

Bi

f-^n+BiCOS (xw
^

,.
-

-r-

TTi*'^
J

Iit)

"------

'-^. I

,.

\AJ

Jv (W)

sin (tw-^virf^'^ f''^"'+^'


sinw(l"x).
^,^.^,.w
V-

/1\

14-]-

bounded

given earlier

when

"

1"j"

in this

^"

1^1

0"t^x^\.

is bounded
integral

when

0 "

r"

^ ^^

1, we

firstshew

that

rt

I t''+'Tn{t,x)dt
*"'"
=

lim
B-*"x"

and

i"'"^^''
' ^

^-^

-^

^*

J.{xw)dw
^
^
^-^^
Y^
J^+i(tw)
{w)] "y
'

^-^
F^+i(tw)
{J^{w) ^^

-^p-

Jv{W)

An-Bi

then

apply the arguments just used in order to approximate to


the right; the details of the analysis
left to the reader.
on
are
integral
It has

been provedthat,ifA
therefore

be

an

arbitrary
number, then
positive

ft

^"+1

Tn {t,x) dt

"u,

J 0

where

U is

independent
of n,

and

t luhen

the

A^o;^!,

A^t^l.

18-23]

results constitute

These

Tn it,x),and

the

in

now

theorems
concerning
necessary preliminary
discuss
to
position
integrals,
involvingTn (t,x),

in the

occur
an

Fourier-Bessel

the

of
investigation
function /'(a?).
arbitrary

which
with

are

we

589

SERIES

FOURIER-BESSEL

expansionassociated

analogue of the Riemann-LehesgueLemma*.

18*23.

The

We

prove that,if {a,b) is any part of the closed interval (0,1),


is not an internal pointor an end point of (a,h),then the existence

shall

such that
and

now

of

the absolute convergence

"b

tif(t)dt
that,as

to ensure
sufficient

are

^cc
,

\'tf{t)Tn{t,x)dt=0{l),
J

0".r
'where-f

^\.
that this theorem

reader will observe

The

that the only part of the

asserts

in
path of integration

f tf{t)Tn (t,x) dt
Jo

is of any

which
the

as
n
significance,

oo

-^

is the part i7i the immediate

vicinity
of

point x.
It is convenient

in three stages. It is firstsupposed


prove the theorem
and that the originis not an
end point of (a, b). In

to

t^f{t)is bounded

that

the second

stage

remove

we

(I)

stage

the restriction of boundedness, and

remove

we

the restriction

in the

third

concerningthe origin.

t~''f(t) F(t){f'-x%

Let

Divide (o,b) into p equal


of \F{t)\in (a, b) be K.
let the upper bound
parts by the pointst^,t2, ^^-i,(^o c*,tp b); and, after choosingan arbitrary
and

...

number
positive

e,

take p to be

so

largethat

S
m

( U""

L",){t,n t,n^^)
"
-

lower bounds

and

Um and L,n J^re the upper

where

e,

of

F(t) in (Y",,_i,
1^)-

"

F{t)

Let
so

th?lt 10)," (t)\^U,n

+ co,,{t),
F{t,,^,)

Lm, in (^,"_i,
tni)-

"

It,is'thenevident that
("!'
J

[tin

tf(t)Tn {t,X) dt^


tn

f+'

(^,,_,
)
i:
m

Cf. Modern

t If

x=

I'f-m

Tn {t,X)(f' of)dt

J t-m-i

t''+'Tn{t,x)(f'-x^)ay,n{t)dt,
J tm-i

Anuli/sin," 9-41.

1, it is,of

course,

supposed

that b"l.

590

OF

THEORY

BESSEL

[CHAP. XVIII

FUNCTIONS

hence, by the inequalities


(10)and (11) of " 18'21,

and

Sci'Kp
\\f{t)Tn{t,x)dt
7rC2^^"(2

"

h)six

"

dt.

+
TTCo

that is to say

2Kp
7rc.^(2-x-b)^xlA,
4ci-

tf(t)Tn(t,x)dt\^
the choice of

Now
at

are

fixes p

An

choose

to
liberty

suitable choice of ^,1,we

when

;
so

may

been

has

p)

chosen, we

"

b)\lx

is 0 (1) as A^
the integral
small. Consequently
arbitrarily
to be proved.
is the theorem

(II)
choose

F(t)

When

intervals /x, such

throughout(a, b),let

bounded

is not

F(t) is

that

bounded

'

is

which

such

therefore

"

largethat An" ^Kpje. That is to say, by


the integral
the left less than
make
on

TTci (2

this

(and

outside

it be

these

-^

and

00
,

possibleto

intervals and

that

f F(t)\dt"".

"^
fl

When

t lies in

one

IX

of the intervals yu,we

the

use

inequality

^''

X) I"
If^^ {t^ X^)Tn (t,

--

is the upper

hence, if K

and

intervals /x, by

bound

applying(I)to

each

TTCa^{2

"

b) ^/x

of ii^(^)
]in the parts of {a,b) outside the
of these

parts,we

have

'(r+ \)Kp

Sci"

tf{t)Tn (t,x) dt

"

irci (2

"

b) \/x

"

small (thusfixingK) and then take A^ to be


sufficiently
the right(and therefore also
the expression
make
on
can
we
large,
sufficiently
the left)
small,and this is the result which had
on
the expression
arbitrarily
to be proved.
If

we

take

"b

dt
t^-f{t)

(III) If
so

exists and

is

absolutely
convergent, we

can

Jo

small that
V

dt

"

e,

and

then, since

we

have

4ci^

rtf{t)Tn{t,x)dt
"

Jo

TTC^

{2

"

"

b)i^x

dt
t--x^

choose

77

18-24]

591

SERIES

FOURIER-BESSEL

(II)that

it follows from

(r + l)Kp

3e
+

"

2'

A,,
K

where

is the

bound

upper

6) when
\F(t)\ in (77,

of

the

intervals

/j, are

omitted.
it follows that the

the left

expressionon
small by takingn sufficiently
large,and
is completelyproved.
Lebesgue Lemma
Hence

TIte Fourier-Bessel

18"24.

shall

We

f{t) he

made

analogue of

arbitrarily

the

Riemann-

expansion.

of
followingtheorem*, by means
expansionassociated with a givenfunction

of the Fourier-Bessel

Let

so

be

can

the

prove

now

the

in. the
functiondefinedarbitrarily

which

the

sum

is determined

interval (0, 1); and

let

letit he absolutely
t~f{t)dt exist and (ifit is an improperintegral)
convergent.
2

Let

am

Yo

[^

tf(t)J^( j,n
t)dt,

7^-^

'J'v+\ \Jm) J

where

^^0.

he any internal pointof an interval {a,b) such that 0"


in (a,6).
such thatf{t)has limited total fluctuation
Let

the series

Then

is convergentand
We

is

its sum

a"

h "1

and

a"iJw{jm^^
l

h,[/(^+ 0) +f{x

0)}.

"

that, by "" 18-21, 18-22,

firstobserve

2
m

a,nJAjm^)

tf(t)Tn(t,x)dt,

Jo

i{f(x- 0) +f(x

lim

0)}

x-\f(x

0)

[%''+^
T,,(t,
x) dt
0)

( f+'Tn (t,x) dt.

ft"^'
[t-^fit) X-^fix

O)}Tn{t,x)dt,

lim

x-'fix

Hence, if

1%"+^
[t-^f{t) X~^f{x

8n {X)

0))r" {t,
X) dt

Jo
+

it is sufficient to prove
of
convergence

that

Sn {x)

-*

as

-*

00

in order to

"x"

(^m "

"^
m

to

the

sum

| {f(x+ 0) +f(x
*

Hobson,

Proc.

London

\ Jm, X)

0)}.
Math.

Sac.

(2)vii. (1909),pp.

387"388.

establish the

592

OF

THEORY

We

BESSEL

discuss

now

I\^^'U-'fit)

X--f{x

in detail,and

reader

the

the
Sn (x) in precisely
function

The
and

so

we

same

f{t)

t'"

such

(0

0)}Tn (t,
X) dt

the
investigate

other

involved
integral

in

manner.

x~''f(x+ 0) has

"

limited

total fluctuation in

(x,h),

write*

may

X\

then

can

"/ (t) x-^f{x + 0) xi (0 X. {t),


functions
{t)are bounded positive
increasing
')(^
r

where

[CHAP. XVIII

FUNCTIONS

^^^1

of t in

{x,b),

that

%i(^+0)
Hence, when

whenever

S not

^t

then

We

arbitrarypositivenumber
exceedingb x, such that

an

number
positive

chosen, there

exists

"

have

!\^+'
[t-\f(t) X-\f{x +

0)]Tn (t,X) dt

is

8.

^x

x.(^ + 0)=0.

-X-^
[t-'fit)

r+"

f{x

Tn {t,
())}
X) dt

J x+S
+

We

now

Next,

between

'

i"-^'%i

be made

the second

0 and

8 such

the three

t"^'x^

(t)T" (t,
X) dt.
the

on
integrals

right.

Riemann-Lebesgue lemma that the


than e by takingn sufficiently
large.

less

mean-value

theorem

it follows that there is a number

that
''"'

"^V^'
["'
X) dt
X^ (0 Tn {t,
J

r^

the

analogue of

the

of the first can


from

(0 Tn (t,
x) dt

satisfied by
inequalities

obtain

It follows from
modulus

X.

0^ + 8)

""+' Tn

(t,
X) dt,

J x+$

and, by " 18'22, the modulus


modulus

of the third

between

the limits 0 and

sufficient!}^
large,

we

of this does

does not
integral
in

2U6.

similar manner,

make

can

exceed

not

exceed

we

2Ue;

and

the
similarly
By treatingthe integral
deduce that,by takingn

the difference between

araJAjmSc)and

numericallyless than (8t/+ 2) e;


Hence, by the definition of

|{/(a;+0)+/(a;-0)}

and

this is

small.
arbitrarily

infinite series,we

an

have

proved that, in

00

circumstances

2 CbmJvijm^)is convergent
postulated,

M/(^
and

this is the theorem

to

be
*

0)+/(^-0)};

proved.

Cf, Modern

Analysis," 3'64.

and

its

sum

is

the

18-25]

FOURIER-BESSEL

The

18*25.

of the
uniformity

convergence

conditions

enunciated

the
Let/(^)satisfy

having limited

continuous {m addition to

593

SERIES

of the

Fourier-Bessel

expansion.

be
"18'24, and also let/(^)
in the interval (a,h).
total fluctuation)
in

to
associated withf(t)
expansion
converges uniformly
number.
the interval {a + A, i
the sumf(x) throughout
A) where A is any positive

Then the Fourier-Bessel

"

This theorem

is

of Fourier

convergence

the usual

analogousto

series *; the discussion of

of the Fourier-Bessel

expansionnear ^
in the firstplacebecause
careful consideration,
more
when
x=
1, and in the second placebecause

rather
untrue

examine

to

of
concerninguniformity
of the conthe uniformity
vergence
0 requires,
1 and near
x=

theorem

the bounds

1 18'22 (1)is

formula

practicable

it is not

of
"f

t''+'Tn(t,x)dt,
'

when

and

the second

for
usingapproximations

Bessel functions

of

kind.

(cf.
"

18*26); but
of

originare

"18-55.

to

small,without

difficultiesin the

The

the

are

of the

case

the

neighbourhoodof

difficulties in the
character ; and

graver

of the

case

are

to

easy

come
over-

neighbourhoodof
is deferred

the discussion of them

"'

shall prove the theorem concerning


of convergence
uniformity
out
through6
of
a
the
of
the
section.
(a + A,
A) by recapitulation
arguments
preceding

We

"

In the first place,


since
of 8 which

choice

(a

A, 6-

Next

involves uniformity
of continuityfthe
continuity
made
in " 18'25 is independent
of x when x lies in

was

A).

we

discuss such

an

as
integral

t"^^ [t-\f{t) c^-\f\x)]


Tn
-

"'

{t,x) dt

J x+S

Since

B is

Lebesguelemma

.,

"

"

independentof
("18-23)that

x,

it follows from

this

tends
integral

the
to

proofof the Riemannas


n-*oc
zero
uniformly

providedthat

r t''^'^{t-''f{t)-x-''f{x)]dt
is

bounded

function of

x.
^

Now

f
J

%C\t'^f{t)\dt+\x--f(x)\
t^-^^t'^f{t)-x--f{x)]dt

:xr^

this is bounded

and

bounded
*

the

Cf. Modern

(a -f-A, 6

"

A) since /(*")is continuous

and

therefore

Analysis, " 9-44.


Analysis,"3'61. It is

namely that it should


interval {a,b).

5,

w.

in

Jo

in this interval.

t Cf. Modern
on

'Jo'

B. F.

be less than

convenient
additional (trivial)
restriction
to place an
A, in order that the interval {x- 5, x + d) may lie inside

now

38

594

other

the
Similarly
and

BESSEL

OF

THEORY

[CHAP. XVIII

FUNCTIONS

in

introduced
integrals

tend to

" 18'24

zero

uniformly,

so
n

tends to

The

18'26.
near

a;

uniformlyas

zero

the

and

the theorem

this proves

uniformityof

the convergence

of

stated.

the Fourier-Bessel

expansion

1.

It is evident

is

?i

pointa;

the
1, so that,at that point,

expansionvanish at
of the expansion
terms

of the Fourier-Bessel

that all the terms

of the

sum

zero.

Since

of a
of convergence
uniformity
of the sum, it is evident
continuity

the

/(l-0)
is necessary

in order that

associated with
We

the

Fourier-Bessel

of the
x

expansion

l.

now

for the
suffi,cient

convergence

is almost
analysis

The

that the condition

convergence
be uniform near

f{t) may

functions involves

in
prove that the conditions that f(a;)is to be continuous
is
combined
with
the
conditions
that /(I)
stated* in " 18"24,
zero,

shall

(a,1) and
are

series of continuous

to

be uniform

throughout(a -I-A, 1).

identical with that of the

precedingsection

take

; we

\t-''f(t)-w-''f{x)]
Tn{t,cc)dt,
I\''+'
J

just as before,and we then divide the interval (0,1) either into three parts
S, or into two parts (0,x
B),(x B, x+ B),{x+ B, 1), if a; ^ 1
(0,X
B),
B. And we then prove that the three integrals
(x" B, 1),if x"^!
(orthe two
as the case
integrals,
may be) tend uniformlyto zero.
"

"

"

"

"

Again,when /(I)

when

Then

0, we

can

choose

Bi so that

Bi^x ^1.

"

the

expression

f{x)-x-''f{x)(\''+^Tn{t,x)
dt
Jo

uniformlyto zerof as w-*-oo when


expressiondoes not exceed (?7-|-l)e for

tends

lies in
any

(a + A,

value

of

1"

8i),and

when

lies in

(1-Si,l).
*

The

interval

t Because

the

(a,b) is,of course, to


integralinvolved tends

be
to

replacedby the interval (a,1).


x"
uniformly throughout (A, 1

the

5i),by " 18-21.

596

THEORY

This theorem
the

same

manner

as

H.

W.

to

the

BESSEL

theorem

[CHAP. XVIII

FUNCTIONS

Young*, and
of " 18'23.

it may be proved in precisely


We
shall write out the proof

jt^f(t) is bounded, with

when

and

is due

OF

K, and leave the reader to


upper bound
the function is unbounded, on the lines of | 1823.

the

when
proof,

Divide

(a, b) into

let the parts be

so

the

equalparts by

points^i t.^,
,

tp^^{to
=

...,

a,

tp

struct
con-

b),

that

numerous

m=l

where

Um and Z," are

Next
and

let

the upper

and

lower bounds

t-f{t) F{t), F{t)


=

of

t^f{t)in (^,"_i,
t^n)-

,_,)""co,n{t),
{t,

then

i ['"'
"irl|r"t^-J,{\t)dtW
\'tf{t)J,{\t)dt
\tU,{\t)"Om{t)\dt

where

c' is the upper

bound

in
\t^J^(t)\

of

reasoningresemblingthat used
and

this is the theorem


The

theorems

(O^a?^ 1) in

to

(0,oo ). Hence, by
the left is o (X~-^),
on
integral

be

can

made

to

cover

the

closed interval

the forms

This is evident

tends to

"

the interval

18*23, the

proved.

of this section

"^ v+i

Hence

be

in

the

when

;f/;./(07,("o.^"={"%

\Jm)

it is remembered

that

in the Fourier-Bessel

general term

series associated

with'f{x)

{aftermidtiplication
hy \/x)throughoutthe inter'val (0 $ a; $ 1)
f^o,^o.n integral
which is absolutely
if^'/i'"^)
convergent; and, if this function
has limited total fluctuation,
the generalterm tends to zero
as
rapidlyas l/jm-

can

zero

18*3.

The

We

of which we
integrals
by means
expansion,
analogousto those which
either in a direct manner
or
proved for Fourier-Bessel expansions,
for Fourier-Bessel
of the corresponding
theorems
expansions.

shall

obtain

have been

by

means

The

methods
application
of the Hankel-Schlafii
consider

now

theorems

Dini

to

Dini's

expansion.

class of contour

concerningDini's

with

expansionassociated

/(a;)is

00

O^JpyKmOC),

Z(
7n

where

Xi,X2, ^3,

magnitude) of

...are

the

positivezeros

(arrangedin ascendingorder

the function

zJ,'{z)+HJ,{z),
*

Proc.

London

Math.

Soc.

(2)xviii.

(1920),pp.

169"171.

of

18-3]
where

and

real constants,

are

coefficients h^

The

and

to be determined

are

tJ;-(X,J) (it

(it

J^ (\J)
t/{t)

Jo

so

the formula

by

"1
'in

597

SERIES

DINI

that

! tf{t)J,(\nt)dt

2\J

Jo

proceedingfurther,we
which
has
Dini expansions,
expansions.

shall

Before

The
which

to certain
explaina phenomenon,peculiar
analogue in the theory of Fourier-Bessel

no

of Dini expansions
is based
investigation
the
of
has polesat
zeros

and, when

It is

be

only to

-^v

at the

zero

the functionhas
negative,

expectedthat

such

0, an

is

+v

of the series,and
li H

0, this last functionhas

Further, \i H

these

contribution

two

should

zeros

has to be inserted
is

negativeand

the

'

be inserted

symbol ^^{x),

'

wh^re

J^Q^x)is

(1) or

(2) in the

of the

account

on

origin.
'

in the

when

zero

an

initial

was

C. N. Moore, Trans.

American

"

must

is defined

For

the

as

expression

0, H + p"0.
be

inserted

Dini

gave

when

corrections

H+v

its vakie

misquoted by Nielsen,

Math.

common

is

(1908),pp. 947"948;

by the

be considered
actually

and
positive

(Leipzig,1904), p. 354.
Cylinderfwnktionen
Mag. (6) XVI.

will

(Pisa,1880), p. 268, but

Dini's formula

will be denoted

is

term

an

i\ 0-

cases
respective

cases
respective

fact that

being omitted.

4-

zeros

"i\n, then

are

"

(^o)J 0

"

that the series which

Dini, Serie di Fourier

and

"

mi

The

terms

.."..";,

purelyimaginaryzeros

ir)1^- (A,o) V^.-

initial terms
so

[Note.

the

at

zero

'

'

{\r

These

to the

"

of the

account

on

initial term

must

contribute

in fact is made.

Jo

If H+v

purely imaginary zeros.

2{v^l)x^\\''+'f{t)dt

"

Phil.

function

origin.

initial term

(1)

x"

of
properties

on

'

was

noticed

the
incorrectly,

Handhuch

of these errors,

der

by

factor

Theorie

der

Bridgenian,
Chree, PJul. Mag. (6) xvir. (1909),pp. 329"331;
^oc. x. (1909),pp. 419"420.]
see

598

OF

THEORY

We

[CHAP. XVIII

FUNCTIONS

the function

consider

now

BESSEL

'2wJ" (xw) J^ (tw)


J^iw) [wJJ (w) + HJ^ {w)]
'

function

This

has

polesSitji,J2,J3,Xj, X^, X3,


...,

The

residue of the function

The

residue

(0 or

...,

"

iXq).

J,^is

at

X," is

at

2 X"j J

(X^ a;)(/^(X^ t)

Jv (^m) V^mJv (X"i)


+ Jy'(X,")
4- HJy

(X^)}
*

+ X"i- Jy- (X,rt)


V^)J"^(X"i)

(X^'^
-

The

residue at the

The

residues at + iX^ when

originwhen

IT

if +

0 is

2/

is

negativeare

both

equal to

2Xo^7,(Xoar)/,(XoO
'

/^' (Xo)
(Xo"-Vv-)/"-(Xo) X(,-

let Dn be

Now
it is not

equalto

numbers

j"i which

Let Sn

number, which

any of the numbers


j^ ; and
does not exceed D^^.

let

be
^*jv

X"+i,

so

chosen

that

the greatest of the

x-H)=l ^MI^f}Z^mil
^^ (^^^)

{t,

"^

is defined
S4o{oc,t)

where

X^ and

lies between

to

\Jm)

v+i

be 0, 2 (1/
+ 1)^"^" or

2\^I,{\x)h{\,t)
//^ (Xo)
4- V-)IJ"(Xo) Xo'*
(Xo-*

'

accordingas

if +

is

zero
positive,

or

negative.

Then, evidently,
iV

We

tf(t)Sn{t,x;H)dt.
"'0

of theorems
leadingup to the result that,when
prove a number
of
the
existence
and
absolute
1,
convergence

shall

0 " ^"

at",Jy(jm^)-^oi^)- 2 bmJA'^mX)=

now

'tV(t)dt
are

sufficient to

ensure

that, as

n^^cc

Itf{t)Sn(t,x;H)dt

o(l).

18-31,18-32]
enables
equation
from
convergence*

This
of

us

599

SERIES

DINT

of Dini's series in respect


properties
of the Fourier-Bessel
properties
corresponding

to deduce

the

the

series.

The contour

18-31.

It is evident from
1

integral
for Sn{t,x\H).
Cauchy'stheoryof residues
CDn+^i

2w

J" {xw) J" {tw)dw


1

'2'iriJ
where
an

the

odd function of w, the


^

n\

(t

immediate

An

2iu J" (xiu)


/" (tw)diu

*"

'

J" (iv)
[wJJ {w) + HJ^ {w)\

-Qoi

value.' The integrand


being
Cauchy'sprincipal
second integral
vanishes,and so we have
cBn+^i
1
2wJ,{xw)J,{tw)dw
'

denotes

symbolP

that

(cf.
" 18-21)is that

of this formula

consequence

|S.((,,;J)|^____i___,

(2)
Cj is

where

of
independent

n,

and t.

Also
r^^+-'-

l\^^^8
(t x-mdt-^^

2Mxw)J^^,itio)dw

and hence

H)dt
t''+^Sn{t,x;
IJ 0

of n,
c^ is independent

where

18*32.

The

{2-x-t)Bn^/x'

and t.

lemma.
x; H) of the Riemann-Lebesgue
for Sn,{t,
analogue

prove the theorem that,if (a,b) is any part {or the


ofthe interval (0,1),then the existence and absolute convergence of
shall now

We

'\if(t)dt
J

are

to ensu7'e
sufficient

that,as

The
theorem
when

"

proofhas

x"

1.

n-^"x
,

o{l),

And, ifb " 1, the theorem is valid when

"

^'

$ 1.

justas in the correspondinggivethe proofof the firststage,


of the reis bounded
and a"0.
The proofs
maining
t-f(t)
constructed by the reader without difficulty.

("18-23)for Tn {t,x). We
stagesshould be

ll

to be divided into three stages

it is supposedthat

.:.:-.;

H)dt
tf{t)Sn{t,x;
providedthat 0

ivhole)

Except

shall

at the

now

point a;=l.

600

THEORY

OF

Let
and

BESSEL

[CHAP. XVIII

FUNCTIONS

t-''f(t)F{t),
=

let the upper


Divide

bound

of

jF(t)\in (a,h) he

K.

(a,b) into p

and, after

equalparts by the points^i ^2,


number
arbitrar}^
positive
e, take

choosingan

{Urn

(^0 a,tp
^j"-i
=

"

"

"

p to be

so

h),

largethat

Lm){t^n"tm-i)"e,

"

m=l

where

17^ and Lm

the upper

are

Let
so

and

lower bounds

of

F{t) in (^,"_i,
^^).",

F{t) F{t,^,)+ "o",{t),


=

that Ia),n

(t)\^ U^i Z", in (tm-i,tm).


-

Then

j\f{t)Sn(t,x:H)dt
J

t"-*-'
Sn

F(t,n^,)
J

(t,x;H)dt+:i

tm-i

r+1 CO,,
m

l J

dt

(t)Sn (t,x; H)

t.m~\

Hence, by "18-31,
Sn (t,
tf{t)
x; H)dt
and

if we

now

take

ec.

(2

largethat Dn

so

[\f{t)
S,,{t,
X

E)

ec^

dt

Ja

and

the

When
and

the

zero

the

magnitude of
proved.
As

the
as

reader

"

l)r

h)\/x

"

2Kp

c^,

we

have

^'')
!
\ {2"x
b) -Jx
"

"

"

expressionon

the left .tends to

"

71

small.
rightis arbitrarily

on
integral

by

the

00

-*-

removed

has

Hence

the

the method

it should
corollary,

restrictions
of

be observed

concerningboundedness
" 18*23, the theorem is completely

that

rh

x^

tf{t)Sn(t,x;H)dt
J

tends

to
uniformly

if 6 ^ 1, where

18*33.
An

Dini's

immediate

the existence

zero

A is

and

an

as

-*-

00

when

O^aj'^l

if6"l, and when

0 ^

a;

^ 1

"

number.
arbitrary
positive

function.
expansionof an arbitrary
of the result of the precedingsection is that
consequence
of the integral
absolute convergence

\^f{t)dt
are

expansionassociated with/(.r)behaves
the Fourieras
same
as regards
(or summability),
manner,
convergence
1).
expansionthroughoutthe interval {0 "x"

sufficient to

in the
Bessel

ensure

that the Dini

18-33]
For it is evident

that

tends

to

zero

as

uniformlyto

when

zero

since

Now,

0 "

after

1 ; and

"

0 ^^ a' ^ 1

be

may

this

by \/x)tends
(multiplied

sum

A.

"

j,^ which

X",, and

that Dn

("15"23),it follows
n

numbers

the

for all values of

when

^x

601

SERIES

DINI

chosen

that

so

|v \

exceed

has the

n-

interlaced

are

value

same

certain stage.

Therefore, since
n

m-N+l

proved that

have

uniformlythroughout(0,1),we
n

x^

Mo(x)+

a^nJAjm^)}
{b,nJi"(\na;)
-

'

'

tends to zero,

uniformlythroughout(0,1

n-^oc

as

A).

"

'
'

is to say, the series

That

CO

m=l

is

uniformlyconvergent throughout(0,1
It follows from

that,when
is

theorems'
'consistency
is 'summed'
by Cesaro's

the series

its

and

'

'

sum

is

x-a-Dio

.iw

associated with

I,

series

1
(0,'

A), the series

"

"

"

-.

summable

"
,

"'

series*

similar method, it

any

in the interval

\jviX),

f{x),is convergent (or is


-

or

means,

zero.

zero.

value of x
if,for any particular

is

concerningconvergent

the

summable
(uniformly)

Hence,

A) and its sum

"

method),then the

by some

"

"

..
.

'
'

'"

"

:"'

".
-

'

-.

is convergent (or is summable

xibmJy(\mX)

xiMo{x)+

by the

same

method) and the

two

series have the

'sum!

same

series (mtdtiplied
the Fourier-Bessel
by \/x) is uniformly
if,further,
interval (a,b),ivhere
an
convergent(or uniformlysummable) throughout
And

0^a"b"l,

'

.,

formly
by \Jx)is uniformlyconvergent (or uni(mtdtiplied
summable) throughout(a,b).

then also the Dini

In

series

y^f(x)has
particular,

limited

'

'

..

Cf. liiomwich,

O^a

total fluctuation

"6"

in (a, b) where

1, /

Theory of InfiniteSeries," 100.

602

THEORY

BESSEL

OF

[CHAP. XVIU

FUNCTIONS

then the series

converges to the

if/(^
such that

all 'points x

at

is

the convergence

o)+/(^-o)}

A^x^b

of Dini's

A, where

"

uniform iff{x) is

The value

18*34.

sum

continuous

series at

in

is

small
arbitrarily
{a,b).

and

1.

of the value of the sum


of Dini's
completethe investigation
the pointx=l;
series by considering
and we shall prove the theorem, due to
has limited total fluctuation in the interval (a,1),the sum
Hobson*, that,if/(a;)
1 is /(I
of the Dini expansionat ir
0).

We

shall

now

We

"

firstwrite

Tr,{t,x;H)=T,,{t,x)-Sn(t,x; H)
-f
"iyt'Q[^tC,tJ

"

"

and

then

y'^m

^2/^

ox

V-) Jt, \A^)

"""

-v

f^m'

T '2 i-\

'J V

\'

K'^n)

have

we

Ai+"
j-

I 27ri
":

'"

J Bn-ooi

T"{t,x-H).

J^ (tw)dw
(^^sc)

wJJ (w) + HJ^ (w)

rJJn+

'Dn+^ilU^(W, t)J,.ixw)

,i

wJJ

(w)

HJy

'

dw

{w)

'

where

(lu,
(}"
a;)s
The

[Note.
of

with

[{wJJ (w) + HJ^ (w)}Y^ (xw)

former

latter when

Tn (t,X

TT

0"
These

of 2'n(t,x;H)
representation

the

[w F/ (w) + HY, (w)}J, (xw)].


is valid when

of T" {t,x\
representations
18-21

(7) and

H)

^51821

are

(8);

analogous to
strictly

the

the fact that there is

R) analogousto " 18*21 (6) is the reason


why Dini
help of the theory of Fourier-Bessel series.]

Now

series

were

no

discussed

consider the value of


't

t-^'Tn{t,l\H) dt
Jo0

when

0 "

"

1.

We

have
^.+1 cDn+Bi

t"^^ Tn

the

Q"t"x^\,

x"t^\.

Tn {t,x) given by "


;

{t,1 ; H) dt

lira

f"
.

^^'+l rD"+Bi

^(^^fj^
l)J,+,{tyj)dw
(iv)+ HJ^ {w)

^-^^

J^^^(tw)dw

im

jB^x

"t

Proc. London

7ri J

Math.

D^-BiwJJ {w) + HJ^{w)

Soc.

(2)vii, (1909),p.

388.

tions
representaformula
in

for

" 18'33

604

to
extending

and

existence

of Dini's expansionin

convergence

vanish

series do not

Dini

that

condition

the

[CHAP. XVIH

FUNCTIONS

interval

an

x=l.

Because
that

uniformityof the

The

18' 35.

BESSEL

OF

THEORY

absolute

f{x)

at a;
identically

in

is continuous^

1, it seems

unlikely

not

with

(a, 1), combined

the

of

convergence

\^f(t)dt,
0

and

the

that

condition

sufficient to

the

ensure

f(x) has limited


uniformityof the

total fluctuation

of the Dini

convergence

be

in (a, 1), may

expansion

in(a + A, 1).
We

that this is,in fact,the

shall prove

The

for the failure in the

reason

Fourier-Bessel

expansion(" 18"26)

case.

uniformityof

near

of the

the

convergence
the fact that

was

f t''+-'Tn(t,x)dt
JO

converge uniformlyto
that, on the contrary,

does not
prove

in

x"

(A, 1),as

in

seen

was

" 18'22.

We

shall

f t''+'Tn(t,x:H)dt
Jo
to
uniforml}'

does

converge
removed.

consideration

that
difficulty
It is easy

of

the Dini
to see,

x"

" 18'26

in

should

then

enable

of the

cause

the reader

to

failure is

see

without

uniformlyin (a + A, 1).

expansionconverges

from

the

(A, 1), and

" 18'34, that

[ t''+'T"(t,x:H)dt
Jo
is the

of the residues of

sum

77^"+!

[{wj; (w) + HJ, {w)]F,+,(tw)

J,+,(tw)]
[w Y: (w) + HY, (tu)}
X

at

Xj, Xo,

.,

X"

plus half

the residues at 0

[f^'

Hence

J^ {xio)l{wJJ
(w) + HJ^ (w)]
,

or

iX^ if H

v^O.

T^ (t,
x; H)dt

Jo
is the

sum

of the residues of
-

and

hence, when

{21w) (H

(w) + HJ, {tu)],


v)J, {xio)/{ivj;

0 "j-^\,

H)dt
t''+'T^{t,x;
*

Without

+
V
x''-"^^

C^^n+^i
,

restriction ou

the value

J^ {xw)
^;

div

{l0)+ HJ,{lv)y

of/ (1-0).

18-35, 18-4]
the

and

DINI

integrandon

right*is of

the

the order of

10-

the

so

when

the

on
integral

iz;

?; 1.

That

magnitude of

{" {l"x)\ I{w) \]

exp

and

605

SERIES

sjx

uniformlyto

rightconverges

like

zero

l/(Dn\/a;)

is to say

\''+'Tn{t,a;;H)dt
0

ficient
uniformlyto x" in (A, 1); and we have justseen that this is a sufconverges
of the Dini series
condition for the uniformityof the convergence
the

associated with

f{t)to
concerning/(O.
The

18'4.
In

sum/(^)

1) under

the conditions postulated

of Foitrier-Bessel expansions.
differentiabilitij

the earlier part of this

written

in {a + A,

obtained

chapterwe

expansionwhich,

an

when

the form

in full,
assumes

/(^)=

(1)

arrJ"{jm,uX)-

"

'

.
.

We
is

shall

study the

now

to
permissible

f (x)= t a.mjm,Jv (im, ")"


u

This

Ford

also
is

not

examined

problem was

Stokes' researches

method

it
given this expansion,

that

deduce

(2 )

to

in which,

circumstances

is analogous
by Fordf, and his investigation
of Fourier series :|:.
differentiability

the

on

the differentiability
of Dini's expansion when
investigated
applicableto other values of H.

It is evident

that

we

can

the

prove

truth

of

(2) if

we

R"

can

but

his

succeed

in

-v,

provingthat
/(^)--/(^')

(3)

and

the numbers

Now

are
J,",^

know

we

the

a,J",,.J.+i(jm,.^y,
l

zeros
positive

of

"(v/x)f{x)admits
that/'(.r)

of the Dini

expansion

00

i
m

inside any interval in which

O^n'Jv+1

\Jm,v ^)

the function

has limited

fluctuation,
providedthat

jy[f'{t)-jf(t)\dt
exists and
*

is

The

is

term

this
infinite;

from

absolutely
convergent.

in wJ"'
shews

(lo)is

that of the Fourier-Bessel

t Traiis. American

important than

more

clearlythe
Math.

reason

the

term

for the difference

" 18-2()).
expansion (of.
iv. (1903),
pp. 178"184.

Soc.

in

J"(iv)except

in the behaviour

in

the limit when

of the Dini

Cf. Modern

expansion

Analysis," 9-31.

606

THEORY

coefficients bm

The

BESSEL

OF

the formula

givenby

are

[chap, xvni

FUNCTIONS

(t) vfit)}J-.+i
(",,, 0
j^", I\tf
-

dt

h,".=

'J ''v+i
{jm,v)
J 0

dt

[t-^f{t)\t^^'J,^,{j,,,,t)dt

if{t)Ju+l(jm.ut)

"J^v+l(jm,v) [

Jm,

tf(t)JAjm,.t)dt

Jm,

^^

(^m

"

providedthat
that this may

Sufficient conditions

(i) ^"^-yXO^O
(ii)/(l-0)

conditions

be the

case

are

t^O,

as

0,
in the open

(iii)
f{t)is continuous
These

0.

combined

with

interval in which

1.

^"

"

the existence and absolute

of

convergence

'y'^j^[t-^f{t)]dt
sufficient to

are

the truth of

ensure

(2) in

/' {x)

interval in which

any

(vlx)f(x)

has limited total fluctuation.

18*5.
A

The

summahilityof Fourier-Bessel

consideration

of

series associated with

the

values

of the

f{x),combined

integral,
suggests that

it is

series.
coefficients in the

with the
matter

no

easy
C
esaro's
of
the
question
summability,by
means,

Fourier-Bessel

of T^^(t,
x) as
expression
to discuss

by

contour

direct methods

of the Fourier-Bessel

the

expansion.

It is,however, very easy to investigate


the
of Riesz* is used to 'sura' the series,and then
inferred with
The
method

the

help of

summabilitywhen the method


the summability(CI) can be
quiteelementaryanalysis.

expressionwhich
of Riesz

will be

lim
and

when

viable

this limit exists,


the

the

as

'sum'

of the

series

by

the

(1

T^)O'mJvijmOo);

Fourier-Bessel

series will be

said to be

sum-

(R).

It is evident

(1)

taken

is

that

(l-^)ar",JAj,n^)=f\f(t)Tn(t,x\
'

m=l

Cf.

-^n'

Hardy,

Proc. London

JO

Math.

Soc.

(2)vui.

(1910),
p. 309.

18-5, 18-51]

607

SERIES

FOURIER-BESSEL

where

T^.{t,x\R)= i

(2)
and

it will be convenient

so

" 18"22

of

manner

18'51.

before

Tn{t,x\R)

function of t and

discuss the

to

make

we

and

x,

by equation(2) of " 18'5, it is a symmetric


of the
shall proceedto establish the properties

is defined
so

we

hypothesisthat
when
properties
corresponding
results alreadyobtained.
function

We

Ji)J2}Jsf
For

'

"'

"

on

J^ {tw)
J-,{xw) F, (w)]

y^

that, when

it is obvious

t "x,

f^'

.Ju{tw)dw

^/

is an odd function of
4"(w,a;)J'^(^w;)/J'p(t{;)
shall

is

of the residues of

sum

"

We

in the

[J^(w) Yy (xiv) J^,{xw) Fy {lu)] ^ (w,x),

since

write

brevitywe

then

and

the

Jn-

and

\R) is the

write down

then

can

we

by interchangingt

^ ^ ^ 1

.r

( "){/.{w) F, {xw)

TTW

^"

0 ^

that Tn (t,x

firstobserve

and

0 ^t^x^l,

the

on

further

after the
of T" (t,
a; \R)
properties
progress with the main problem.

concerningTn(t,x \R).

Theorems

When

(i^fy-"^^p^^}^

obtain

now^

some

upper

w,

for

bounds

(w) \
j4" {w,x) Jy (tw)/Jy
both when

is

axis
imaginary
and

the line

on

joiningAn"
which

the formulae

0 ^ ^ ^ 1, the

signof

a:

"

oo

i to

Iw| is not
I"; 1is not

when
*

as

the

When
-*"

00

t 5; x, the
.

origin.

There

is

need

series of
from

i,and when
are

valid when

0 ^

the
ic

$ 1

the

ascendingpowers of lu
formulae of Chaptervii

'
"

to make

':

-.

".

""

taken along
integrals
no

c"

being immaterial.

shall use
we
inequalities,
derived
and inequalities
large,
small.

will be discussed

To obtain these
when

An

an

the lines

indentation

joining"
at the
...

ii? to

J"

"

origin,because

iB

do

not tend

to

zero

{w,x) is analytic at

608

THEORY

We

firstdeal with

the factor

Jy (tw)

(1)
when

is

either

on

BESSEL

or

We
Jt,{tw)/J^{w).

consider

next

that*

this follows from

contour;

(w, x),which

observe

xvm

l"^^exp{-(l-01/(t^)|l

" 18"21 (9) when |w[ is not small, and


less than j^.
not large(i.e.
We

[chap,

FUNCTIONS

from

is

the

of the type
inequalities
ascendingseries when \w\ is

equalto

{w) F,'--^'
{ayw) ^,"' (xw) H,^^ {w)];
liw {H,^'^
-

it is convenient

when

being valid

former

first

(I) The

make

to

investigations
concerningthis function,the
"^^v^^, the second when v^^.
two

is quite simple. It
investigation

S 7-33 that

follows from

"3*6

and

L Ie-^''"'

^,"i)(xw)I"

(2)
for all the values

of

and

(3)

\xw 1^
under

(w, x) I"

\Hy'"^{xw)\"

consideration

when

h ^v

"

Iw Iis not small, we

use

the

(deduced
inequalities

h Ie^'

h\e-

w\

togetherwith

the

\xw\ is

largeor

and

" 3*6

that

" 7*33

we

deduce

combine

that, whether

the

supposed

their values

t Provided

may,
of

that

"

h -^v ^ J

the

numbers

however, depend
course

inequalities
(6) are

true

whether

that

0 "x

Iw I*-"}
exp {(1-x)\I(w)\ ],
be the

results contained
or

4" (w, x)\"k,(x-^+

(8)
It is

a;-"

whatever
| and \w\ is large,
now

the

Hence,

not.

I^(w, x) I" k^ki{x-i+

(7)

" 7"33)

IW

gtXW
I
ir^")(xw) \"ki{\xivr* + Ixw j-"}
!e"
o"ixw
e^^'2'(xw) \"k^{\xw\-i+\xw j-"}
I

It fullows from

from

inequalities

(6)

If

the

from

^."2"(w) I"

"
ii7."'(to)\

(5)

when

deduce

x'".
\"t"{w,x)\"k3\w\

(4)

Hence

and |w | is not
V ^ i
large,it is easy to
(II) When
J^,{xiv)and Y^ (xw) that
ascendingseries for Jy (w), Y^ (lu),

If

^^.

{(1 x) \I(w)\].

exp

-~~

xiu\^

on

of

in formulae

(3),(4) and (7)we

jxw

|.

^,

x-")exp {(1-x)^I (w)i},

Aj, A-2,A3,
the value

^ 1.

magnitudef

of

...

v,

are

positiveand

independent of

lo,

and

t\

18-51]
when

FOURIER-BESSEL

is any

pointof

609

SERIES

either contour

and

0 "

a;

Hence, by (1),it follows

^ 1.

that

^{%v,x/-j~^j'"^\"ht-Hx-^
x-'')ex^[-{x-t)\I{iu)\\,

(9)
when

and

0^x^\

We

now

return

O^^^l.
the

to

formula
integral

An " iv and + iv in the first and


deduce that,when 0 ^t "x^l,

two

w
replace
by
we
integrals
respectively,

contour

we

^l^^^^^ilf

cfo

.e"

provedthe
consequently

have

second

x \
{t,
R). If

-5^(..-. .-0 /;

Ir. ft . IiJ)i"
We

for T"

inequalities

(o,*".,i),
|r"(,,|^),"2Ai^^i"iC^
^TAt..m"J"^, (o"."""i).
"

(10)

"

(11)
It is to be remembered
make

|a;

"

that

k^ is independentof
desire to do

^ |tend to zero, if we

and

t, so

that

we

may

so.

is required
in order to discuss the behaviour
pairof inequalities
ojfTn {t,X IR) when x and t are nearlyequal.To obtain them, we write
One

other

2.j(l-^J*(",.)-^-,

2'..(",.,iJ)
=

when

vertices

or

in this

^t^xi^l;
+

iAn, A^

the
integral

is taken

contour

be

to

with
rectangle

iA^-

"

It is easy to see that (9) is satisfied whether tv be on the horizontal sides


the vertical sides of this rectangle
on
; and the factor 1
{w/An)does not
"

\/2 in absolute value

exceed

Consequentlythe

at

any

pointof

of the

modulus

the contour.

integranddoes

not

exceed

+ x-")^-!;
kst-^ix-^

and since the

lengthof

and

"

when
similarly,

the

\T,,(t,x\R)\"

infer that,when

-^^^^-^

last four

sufficient to enable
are
inequalities
summability(R) of Fourier-Bessel series. The

the consideration of small values of


to

we

an

W.

but
appreciable
B. F.

0^^^^^

1,

0 "^x -^t ^1,

(13)
The

is QA",

iT.,,Mm"'-^^^^^i^.

"

(12)

the contour

not

to

an

undue

us

to

discuss

adequately

reader will observe

has increased the

lengthof

the

that

analysis

extent.
39

610

THEORY

We

can

[CHAP. XVni

FUNCTIONS

theorem.
analogueofFejer's

The

18'52.

BESSEL

OF

now

that the existence and

prove

the absolute convergence

of

Ciif{t)dt
Jo

to
sufficient

are

summable

(R) at

all

f{x"0)

exist. And

pointsx of the open interval (0,1)


the sum
(R) of the series is
i{/(^

which

at

with

f{t) is

the two

limits
'

0)+/(^-0)}.

concerningFourier

theorem*
analogueof Fejer's

the
is obviously

This theorem

series associated

that the Fourier-Bessel

ensm'e

series.
Since

that,when

series which

"

"

a;

lim [t"-*-'
rt''+'Tn(t,x\R)dt=
Tn{t,x\R)dt

lim

"

the limits

it follows that, when

Hence

"

/'(ir 0) exist,then
+

rt"-^'
Tn(t,x\R)x-''f{x-0)dt+

lim

(R),it follows from " 18'35

is convergentis summable

lim

ft-+'Tn(t,x\R)x-^f(x
+

We

in

now

are

consider

to
position

(l-Jf)J. Umx)

I
m

am

-^n/

the

|{/(^

0)+/(^-0)}.

"S^"
{x 1R), defined

sum

0)dt

I^^"-^^
T, (t, IR) x-''f{x
X

as

0) dt

Jo

-\ t''+'Tn(t,x\R)x-''f{x0)dt,
+

and

shall prove that it can

we

be made

small by taking?t sufficiently


arbitrarily

large.
The

Sn {x \R) is equalto

sum

[%'+'
[t--f{t) x--f{x
-

0)}Tn (t,x\R)dt

Jo
+

I *"+' [t-'fit) x-^f{x +


-

Now,

on

the

that
hypothesis

number
positive

e,

the limits

there exists

/(ic+ 0) exist,if we
number^ 8 such
positive

i\t-'fit) x-^f{x + 0) I"


li
"~''/(0 ^""/(^ 0) I"
-

We

choose

now

positivefunction

sufficiently
largevalues
the

pointsx"8,x
*

Cf. Modern

"(t

0)}Tn (t,X IR)

dt

of n, and

choose

an

e,

{x^t^x^-l),

e,

{x-^t^x-h).

is less than S for


of n, say "r (n),which
the interval (0,1) into six parts by

divide

(n),x.
t Cf. Modern

Anulysis, " 9-4.


i It is convenient

arbitrary

that

to take

6 less than

and

Analysis,"
1

a;.

8-43.

612

THEORY

Since

each

limits

of the

BESSEl.

OF

the

on

[CHAP. XVIII

FUNCTIONS

right is equal to ^x",it

has

now

been

the

limits

proved that
2t

Cim^

\Jm'^)

m=l

is summable

f{x

(B) with

0) exist

"

and

+ 0) +f{x
^ {/{a:

sum

this is the theorem

to

0)}providedthat

"

be established.

the reader should


be able to prove without
corollary,
that,if / {t)is
difficulty
continuous
in {a, b), the summability (R) is uniform throughout the interval in which
Cf. " 18'25.
where A is any positive
number.
a + A^.v'^b-A,
As

18*53.

Uniformityof summabilityof

the Fourier- Bessel series

the

near

origin.
uniformityof the summability {R) of the
Fourier-Bessel
expansionthroughoutan interval of which the originis an
tiplied
end-point.It will be supposedthat the expansionis modified by being mulis
and
it
Avillthen
be
tinuous
throughoutby "^x,
proved that,if t~''f(t) conin the interval (0,h),then the modifiedexpansion
is uniformlysummable
number.
(0,b" A), where A is any positive
throughout
We

shall

Given

we

e,

examine

now

choose

now

can

the

(lessthan A)

that

so

\{tr''f{t)-x-''f{x)}\"e
whenever

Since
taken
We

to

"

S^t^^x

B and

t^O, providedthat

lies in (0,6

of continuity,
this
involves uniformity
continuity
be independentof x.

"

A).

choice of 8 may

be

write

now

Sn{a;\R)=\
f^' [t-^fit) x--f{x)]
Tn (t,
x\R)dt
-

Jo

and

We

jx^Sn {x\R)\ after the

examine

then

express

be omitted

when

x^Sn{x \R)
x

"

8),and

the

as
we

" 18'52.

of six

sum

that

see

of

manner

(some of which
integrals
does
not exceed
jx^Sn(x |R) \

are

to

AnP
2k,e {x^^

(")
x)\ {''-''

^/Cc

3-4,1" X-\-"T(n)

V2
+

In this formula
to be

P-S

any of the limits of

replacedby

zero.

I
Jx

2
3^ "'

(t-x)

X-"

dt

:,+

"x+S

rx

-7f|

dt

dt

dt +
"

'

}x^^{n){t-Xy_

\[t-^f{t)-x-^f{x)]\dt.
which
integration

are

negativeare supposed

18*53, 18-54]
Now

FOURIER-BESSEL

this upper

bound

for

\xiSn(x\R)\ does

and, since x~''f{x)is bounded


small by
arbitrarily

Now

it has

alreadybeen

exceed

not

(because it is continuous),this

choice of

x^Sn{x \R)
Consequently

613

SERIES

which

tends

to

shewn

is
zero

independentof
uniformlyas

can

be

made

ic.

oo

w-*

("18"22) that

Jo
is

uniformlyconvergent in (0,1 A), and


involves uniformity
of summability,
"

so,

since

uniformityof

convergence

x^-^f{x)\
t''+'Tn(t,x\R)dt
Jo

tends

uniformlyto x^f{x)in (0,6

Hence, since x^Sn{x\R) tends

A).

"

to

zero

uniformly,

[ tf{t)Tn {t,x\R)dt

xi

Jo'
tends

uniformlyto

f(x) i f+'^Tn(t,x \R)dt, i.e.to x^f{x) in (0,h

x^-"

"Jo

It has therefore been

A).

proved that
00

S
is

a",x^J^{j,nOc)

{R) in (0,6

uniformlysummable

A) with

"

x\f(x),providedthat

sum

"1

t^f{t)dt

"

'

'

exists and

is

18"54.
We
series

absolutely
convergent, and

Methods

shall

of summing
'

'

is continuous
t~''f{t)

Fourier-Bessel

various
investigate

now

that

methods

of

in

(0,6).

series.

summing

the Fourier-Bessel

*
00

m=0

on

hypotheses(i)that

the

the limits

that
/(a-+ 0) exist,(ii)

Ctif(t)dt

Jo

exists and

is

that
absolutely
convergent, and (iii)

"*

(R).

so
brevityto write /^ (if)in placeof dm^^Jvi'^),
that/"i(a;)
when x lies in (0,1),
uniformlyto zero ("1827) as m-^cc

It conduces
tends

the series is summable

The

factor

uniformityof

x-

to

is inserted

summability

near

merelyin
the

order

origin.

that

the

discussion

may

cover

the

investigationof

614

THEORY

Consider

givesthe

natural method

most

lim
exists and

is

(lJf)f^{a")

(jJAn)-^!, it is evident

Since

[CHAP. XVni

FUNCTIONS

firstthe limit
lim

which

BESSEL

OF

(ofRiesz' type)for summing

the series.

that

(^^^)Ma^

(l-i^)/^(^).

equalto
lim

Again,since/n(a;)o(l),it is easy

to

that

see

2 fm{x)
m

SO

o{n),

that

I A(^)
lim(-^^-^)
\

w-".oo

and

0,

m=l

Jn

therefore
lim

I (l-^)/"(^);
lim
fl-^fVm(^)=
Jr

the limit

the

on

rightexists

in consequence

of the

hypothesesmade

at

the

the section.

beginningof

Again,since

whether

be

o{n) or 0{n), it

and

follows that

0,
Spf-^V".(^)

lim
n-"ao

\nj

Jn

"/

\Jn

so

lim

(l ^)/^ (^O

Consequentlythe hypothesesthat
that the integral

exists and

is

absolutely
convergent

lim

U
(l ^f)
-

the

limits

are

to ensure
sufficient

/(^r+ 0)

exist

(^).
(0"a'"1)

and

that

00

is sutnniahle

(G 1) with

sum

^x^ [f{^+ 0)+/ (^' 0)|~

iif{x) is continuous in (a, h),the summability(C 1)


reasoning,
By the same
is uniformin (a + A, 6
has a limit as t
O.'the
0 and t~"f{t)
A) ; and, if a
summability(Cl) is uniform in (0,6 A).
=

"

"

--*"

18-55, 18-56]

FOURIER-BESSEL

Uniformityof convergence of the

18'55.

615

SERIES

Fourier- Bessel

the

expansionnear

origin.
We

can

has limited
of

by usingHardy'sconvergence theorem*, that,ift^f(t)


total fluctuation
in (0,6),while /(^)is also subject
to the conditions
now

prove,

"18-53,then
00

2^

a^yiX''Ji,\jq^i,x)

111=1

is

uniformlyconvergent in (0,6
Let h

equalto

(t)be

function
auxiliary

an

in

zero

A) with

"

x^ f{x).

sum

defined

to

be

let the Fourier-Bessel

(b,1); and

equalto f(t) in (0,6) and

series associated with

h{t) be

00

Hl

CO

Then, by 1 18*54, S am^c^Jv(jm^)is uniformlysummahle


(0,6

"

with

A)

x^f(x),and, by Sheppard'stheorem
(jm^)^Jvijm^) is a bounded function of

theorem,
2
m

is

is
("18*27),
am/\//i",
and ni.
Hence, by

sum

while
0(l/ni),
Hardy'sconvergence

(C*1) throughout

a^x^J^ijrnx)
l

uniformlyconvergentthroughout(0,h

A), with

"

x^f{x).

sum

Again
2

{am-0im)0C^Jv{jmX)
=

uniformlyto zero in (0,6


Riemann-Lebesgue lemma ("18'23).

and

this tends

and

-JO :

"

A)

as

n^-oo

tends uniformlyto the


% (ijnX^Jv{jm^)

Hence

X^\tf{t)Tn(t,x)dt,
J b

m=\

18 56.

by

an

analogueof

x^f(x)in (0,6

sum

"

the

A)

as

this is the theorem

of Dini
Sunimahility

to

be established.

series.

Except when x=\, the summability(C 1) of the Dini


f{t) may be inferred by combining the results of " 18'33

series associated with


and

"" 18*51

"

18*53.

f for
summability(Cl) may, however, be established independently
all pointsx such that 0 " ^ ^ 1 by replacing
An and the functions J^ (iv)and
which occur
in " 18*5, by Dn and the functions wJJ (lo)+ HJ^ (in)and
F^ {'w),
the details of the analysis
wYJ (w) + HY^ (w) respectively;
may be left to the
l the expression
^ {f{x + 0) +f{x 0)]
reader,and he will find that when x
be replacedby /(I
must
0).
The

"

Cf. Modern

t Of

course

on

Analysis," 8-5.
the hypotheses concerningf(t) which

were

assumed

in " 18-53,

616

THEORY

The

in

of convergence
The

be dealt with

(a,1) may

was

in

dealt with

is of

the

interval (a + A,

in the

same

as

way

1) when

the

/(so)
uniformity

""18"33, 18"35.

series

Dini

summabilityof

[CHAP, XVIIT

FUNCTIONS

summabilityin

the

of
uniformity

is continuous

BESSEL

OF

of Fourier-Bessel

(and

series)
by

cation
modifi-

physicalimportance.Thus, in Fourier's*
of radius unity,
problem of the Conduction of Heat in an infinite solid cylinder
the temperature v at distance r from the axis satisfies the equation
of Abel's method

some

(d^v

dv
J

the

W'-^rd^V

dt~
with

1 dv

boundaiy condition
dv

0,

if the initial distribution


Normal

solutions

condition

of heat

of

is

symmetrical.

differential

the

the
equation satisfying

boundary

are

^0 (^w?')exp (- A;\,,rO'
and

so

the

is

temperature v

givenby

the series f

00

S
where

b,nJo(Knr)exp (-k\Jt),

the coefficients b,nare

to

be determined

from

the consideration

that

00

bmJo(Xmr)
1

m=

is the Dini

series associated

with

that the initial temperature is

f(r). It

the initial temperature

is evident

as
expressible

00

lim
and

this limit exists when

The

18*6.

Z"";
Jo(\nr)exp{

the Dini

shewn

kX^' t);

series is summable

uniqueness
of Fourier-Bessel

It has been

"

by Young J that

series and

the

(R).
Dini

existence

series.

and

the

absolute

vergence
con-

of

\\\f(t)dt
J

sufficient to

are

Fourier-Bessel
be

XII.

that

ensure

if all

the

coefficients
a^ of

associated ivithf(t)are
series)

the Dini

series {or the

zero, then the function


f

{t)must

null-function.
La

Theorie

(1906),pp.

Analytique de la Chaleur (Paris,1822),"" 306"320.


[Scientific
Papers, v. (1912),pp. 338"339]

106"107

1883, pp.519
Sitzungsbericlite,
+ In this
t Proc.

Math.

Soc.

Rayleigh,Phil. Mag. (6)


and

Kirchhoff, Berliner

524.

"

physicalproblem,

London

Cf.

"

(2)xvm.

0, and

so

there

is

no

initial term

(1920),pp. 174"175.

to be inserted.

18-6]
To

this theorem

prove

where

that, when

observe

we

determined

the coefficients a,,,,are

617

SERIES

FOURIER-BESSEL

by

the

0,1,2,

"

we

...,

write

may

formula

'^''v+iijm).
u
0
and

series

the

in (1

boundedly

by t^f(t)and

the

on

A, 1). It is therefore

"

"

oscillates

A) and

permissibleto multiply the expansion

integrateterm-by-term.
that

It follows

all the

0.

Since

uniformly in (0, 1

right converges

integrals

rtu+.p+.f(t)dt

(p =1,2,2,...)

""

Jo

it follows

are

zero,

the

theorem

series

stated
be

can

because

in
It is

is

Dini

(1

"

A,

preciselythe

series

same

associated

possibleto

construct

with

Bessel

of

seems

convenient

and

bm.are

functions, which
to
""

defer

by

corollaiT

pp.

Lerch,
37"43.

to

711

methods
the

Acta

for Dini

theorem

theoreticallysimpler
formly
uni-

fail to converge

not

functions

of the

any

the

"

constants)which

Rieraann's

directlyassociated with
in Chapter
be discussed

Schlomilch's

"

"
.

more

will

examination

of the

Xix

Wl

" 19"7,when

discussion

Mathematica,

(1903),

series,

pp.

and

simple

the discussion

of Schlomilch

xxvii.

"

it

series

00

to

types

resembles

"

Riemann's

it is

series of Bessel

00

'

"

The

and

w=l

theory is, however,

series

and
does

f^'^P

theory of

theory of trigonometricalseries f.
a

way,

"

the coefficients a^

Such

series.

theorem*,

1).

(where

null-function, by Lerch's

for Fourier-Bessel

proved

in

proved

the

t^f{t) is

that

345

"

of the series forms


'

347;

Young,

"

Messenger,

xh,

(1910),

Cf. " 12-22.


'

t Cf. Modern

Analysis, ""

9-6"

9-632.

'

19 '1. SchlomilcJis
In

Chapter

real variable

CHAPTER

XIX

SCHLOMILCH

SERIES

expansionof a function
of a

XViii

dealt with

we

the

real variable.

expansionof

function

f(x)

of the

in the form

00

f(x)= 2 a^J^(jmx),
m

where

/," is the mth

(m

of /"

zero
positive

jm

{z),so that, for largevalues of

lv-i)7r+0

(l/"0.

That

is to say, the argument of the Bessel function in a term


of
the series is approximatelyproportional
to the rank of the term.
In this

shall discuss the series in which

chapterwe

Bessel function

By choosinga

in each

term

is

to
exactlyproportional

suitable variable,such
S

series may

a,ft Jy

the
the

rank

be taken

to

of such

were

in

of the

of the term.

be

{nix).

subsequentlythat it is convenient to add an


is simplified
("19"11;cf "18'33);and the analysis
by making
in the form of the coefficients in the series ("19'2).
of this type

high rank

argument

It will appear

Series

m,

first investigated
by Schlomilch
Fourier-Bessel

great importanceto the Physicistas

*.

initial term
a

fication
slightmodi-

They

are

not

series,though

0) they present themselves


Rayleighf has pointed out that (when v
in
the
vibration of a twoof
naturally
investigation a periodictransverse
dimensional
membrane, if the vibration is composed of an unlimited number
of equalone-dimensional
vibrations uniformly
distributed in direction
transverse
through the two dimensions of the membrane.
=

the series present various features of purely


applications
mathematical
it is remarkable
that a null-function
interest;and, in particular,
be representedby such a series in which
the coefficients are not all zero
can
("19-41).

Apart

from

than Fourieramenable
to analysis
respects the series are more
in common;
Bessel series,
but the two
types of series have many properties
In

and

some

the

reader

will be

right when

he

infers from

comparison

of

the

that the relevant range of values of x is (0,ir) for


arguments j^^x and mx
Schlomilch
series,correspondingto the range (0,1) for Fourier-Bessel series.
*

Zeitschriftfilr Math,

specialcases

j'

0 and

t Phil. May. (6)xxi.

imd

Phys.

ii.

(1857), pp. 155

"

158;

Schlomilch

considered

l.

(1911),pp. 567"571

[ScientificPapers,vi. (1920),pp. 22"25].

only the

620

We
integrations.

order of the

change the

and

[chap. XIX

FUNCTIONS

BESSEL

OF

THEORY

deduce

that

'

g{wsme)dd-f(0)

f'{xsine

Jo

"

TT
.

TT

sin ^

ri' r"

2w

sin "b)sin Odcbde

Jo

cos

dd

dx

~7rj,Jo-^
^^''"^\/(sin^^-sin^;^)
_

sin ^

2a;

X dd

cos

sinx)
.'o
)

TT

/ (^ sin x)

^"^
6'"'^^
c'X

cos

sm

arc

"

Vcosx/Jx

vav

-^

'^

f'{xsinx)Qosxdx

a;

\/(cos-X

/cos
cos

dx

cos'^ ^)

"

Jo
=

and

so, when

/(^)-/(0),
g

(x) is defined by (4),g (x)is a

solution of

/' (x) is a

from (4) that, when


it is easy to verify
limited total fluctuation in the interval

Now
with

(3).
continuous

(0, tt),so

also

is

function

g(x);

and

in the form
therefore,by Fourier's theorem, g (x) is expansible

(x)

^tto+

cos

"m

mx,

""=i

where
2
^m

f^

/"'^

'

/^O) +

ttJo
and

this series for g


Hence

mudu

COS
('^)

"^19

i^

(x) converges

/"'
(u sin "^)(i(^COS mudu,

and
are
permissible,
term-by-termintegrations

f(cc)
=

"-

so

we

have

(x sin 6) dO

S
+
li",o

interval (0, tt).

uniformlythroughoutthe

","

(wa;sin 6)\dd

cos

m=\

"

2'-*'o
m=l

and

this is the

values

obtained

expansionto

be

established.

coefficients am

for the

are

It is easy to verifythat the


those given by
the same
as

equation(2).
When

the

restriction

is removed, the Fourier

concerningthe

series associated

limited

g (x) is no
that
ensures

with

oi f (x)
convergent, though the continuity

f (x)
longernecessarily

total fluctuation

the

of

Fourier

series

19-2]

SCHLOMILCH

'

is

uniformlysummable
the series
integration,

621

SERIES

(C 1) throughout(0,tt); and

hence, by term-byterm

00

is

uniformlysummable
of Hardy'sconvergence

with sumy(^);
(Cl) throughout(0,tt),
theorem*

then

ttrn

is sufficient to

f{x) when
For

application

an

that the additional condition

shews

0{\l\/m)

the convergence
of the Schlcimilch series to
lies in the half
-open interval in which 0 " x ^tr.

the

ensure

further

concerningthe summability of
memoir
a
by Chapman f.

theorems

the reader should

consult

Schlomilch

sum

series,

[Note. Tlie integralequation connecting /(.r)and g (x) is one which was solved in 1823
by Abel, Jotcrnal fiirMath. l. (1826), p. 153. It has subsequently been investigated
J by
Beltrami, 1st. Lombardo
(2) XXV.

(1897),p.

The

equation

Reiidiconti,
(2) xiii.

104

C. fi.Smith, Trans.

I'

"

"

""

./0

(1880),pp. 327, 402


American

Math.

sin 6 "^0dd
f'{x sin dsiiKJ))

./ti

in

"

simply established

3'33

in view

by

Volterra, Ann.

"Soc. viii.

(1907),pp.

di Mat.
92"106.

f (x)-/(O)

"

is most

the method

changing axes of polarcoordinates,


explained
was
by Gwyther, Messenger,xxxiii.
(1904),pj). 97
107, but
of the arbitrarycharacter of f{x) the analytical
proofgiven in the text seems
ferable.
preIn connexion
with the changes in the order of the integrations,
cf Modern
Analysis,

"4'51.

; this method

:"

We

"

"
"

The

have

definition
of Sclduniilch

now

into

"

19*2.

function

of

used

series.

Schlomilch's problem of expanding an


investigated
series of Bessel

function in the {m
valid for the range

4- l)th term

functions

of order zero,

the argument
and

beingproportional

to m;

arbitrary

the

of the

expansionis

of values (0,tt)of the variable.

the functions of order zero by


generalised
by replacing
functions of arbitrary
order v; and a further generalisation
may be effected by
takingthe generalterm to contain not only the function J^ {nix)but also a
Such

function

series may

which

be

bears

to

sine does to the cosine.

function

The

is,of
generalisation

latter

theoryof Fourier series,and


expansionsvalid for the range of

the

The
*

functions which

Gi. Modern

been

made

are

values

thus

("

tt,

under

come
naturally

same

of relation

course,

as

the

suggestedby

led to expect the existence

of

tt)of the variable.


consideration

for insertion

are

Analysis, %%-b.

t Quarterly Journal,
J Some

we

the

kind

the Bessel

xliii.

(1011),p.

34.

interesting apphcations of Fourier's


by Steam, QuarterlyJournal,

xvii.

integral theorem

(1880),pp.

90

"

104.

to

the

integral equation have

622

OF

THEORY

Bessel functions

kind

second

of the

series to be considered

BESSEL

and

be written

may

[CHAP. XIX

FUNCTIONS

Struve's

functions;and

the types of

in the forms*:
"*

T(v

but his
such

of the

proofof
series

defined to be
Series
direct

(ima;)"

r"t=i

^n

Series

1)

a^n

Jv (tnx)+ hm "H-v(mx)

type (with v=0)

former

by Coatesf;
function f(x) into
arbitrary

of expanding an
possibility

the

invalid

be

to

seems

latter type

considered

except in the trivial case

(with period27r)and
periodic

of the

been

have

to

as

zero

a;

oo

-*

greater interest,and

of much

are

to tend

f(^) is

in which

series. They
of trigonometrical
generalisation

will be

they form a
called generalised

ScJilomilch series.

types of investigation
suggest themselves

Two

in connexion

with

generalised
expanding an arbitrary
problem
function into such a series;and the second is the problem of determiningthe
of such a series with given coefficients and, in particular,
the
properties
of analysis
construction
(resemblingRiemann's
analysisof trigonometrical
a
series)with the objectof determining whether
generalisedSchlomilch
Schlomilch

series. The

the coefficients

series,in which

Generalised Schlomilch
Math.
K.

Ann.

Lii.

not

are

series have

(1899),pp. 582"587

of

all zero,

l)een discussed

Ann.

in

Selskahs,1899, pp. 661"665


(3)vi. (1901),pp. 301"329.

di Mat.

the

Nielsen^ has given


Schlomilch
expansionof
great detail

the

but
null-functions,
those which

an

will be

forms

for

his researches

given in

the

this

series of memoirs

are

and

Schlomilch

of

of

1900, pp.

coefficients in

function
arbitrary

construction

actual

represent a null-function.

can

x. B
Nyt Tidsskrift,
(1899), pp.

Videnskahernes

Danske

127"146

first is the

he has

73"81

55"60

by Nielsen,
; Oversigt
;

1901, pp.

the

generalised
with
investigated

series which

represent

different character
distinctly

from

chapter.

which we shall now


of expanding
investigation
give of the possibility
Schlomilch
series is based on
function into a generalised
the
an
arbitrary
for
0 in his memoir
the case
on
y
givenby Filon"
applications
investigation
residues
functions
in series of
of
the
of
calculus
of the
to
expansions arbitrarj'^
functions of given form. It seems
to be of some
importanceto givesuch an
of modifyingthe set of
there is no obvious method
because
investigation!!
The

The

reason

series one-valued

for

insertingthe factor .r" in the denominators

t Quarterly Journal,
J See e.g. his Handbuch
" Proc.

London

is to make

the terms

of the second

(cf." 19*21).

Math.

xxi.

(1886),pp.
der

Soc.

Theorie

189"190.

Cylinderfunktionen(Leipzig,1904),p.
(2)iv. (1906),pp. 396"430.
der

348.

that
i " f " J. The results which
will be proved in "" 19'41
IIIt has to be assumed
suggest that it is only to be expected that difficulties should arise for other values of i'.
-

"

19'62

19-21]

SCHLOMILCH

623

SERIES

Jy{mx), ll^(mx) so as to obtain a set which is a normal


there is no method
for the interval (" tt, tt);and consequently

functions
set

the coefficients in

Schlomilch

the coefficients in

orthogonal
obtaining

of

that
as
expansionin so simplea manner
Fourier-Bessel
expansionare obtained ("18"1).

in which

which
forms the latter part of the chapter,
investigation,
concerning
of null-functions
Schlomilch
is of
the representation
series,
by generalised
the exposition
of Riemann's
character
researches on
as
exactlythe same
series given in Modern
""9'0 9'6"32.
trigonometrical
Analysis,
The

"

The

19*21.
SchU'miilch

application
of

the calculus

of

residues

the

to

generalised

expansion
.

We

shall

explainthe

to discover the
by which it is possible
values of the coefficients in the generalised
Schlomilch
expansionwhich
function /(a;),
the order
of the Bessel functions
when
represents an arbitrary
this has been done, we
lies between
shall not consider the
| and |. When
the discovery
has been made, but we
of the processes by which
shall
validity
series in which
the coefficients have the
that the Schlomilch
prove directly
does converge to the sum
values actually
f{x).
specified
now

method*

"

This

is

theorem

anaJogous to the procedure which


expansion
/(.r) 5ao+
m

the values of the coefficients


and

mx,

adopted

in Dirichlet's

proof of

Fourier's

so
integrating,

(a"iC0S7)i.r-|-/3"(SinmA')

sin

is

in the

"

discovered

are

that

the

by multiplyingthe expansion by

values

of

and

a"j

/3";are

taken

to

be

cos

ms

and

given by

l^v
the

equations
1

/"""

'r

We

and

prove

that it

It conduces

the coefficients have


1

"^

f{t)dt+-

actuallyconverges
to

tdi, (3m

-TV

/"tt

-"

7)1

J -TT*

take the series in which

then

f-^
f (t)sin 7ntdt.
I
'^J
I

f (t)COH

"""=-!

brevityto

to the

f it)con m{x-t)dt,
sum

deal with

/(.r).
the

i H^

pairof

pairof

'^

We

functions

(mx)

(I mx)"
the

values,namely

T'^

J^,(mx) "

instead of with

these

'

functions

Jy(mx)l(^mxy, 'H.^(mx)j(\mxy.

shall write

J^(z)+ i-a,(z)_

(A\

Apart from

in the memoir

presents

no

details of notation, the followinganalysis is due to Filon; it was


given by him,
to values of v between
"h
just cited,for the specialcase
0, but the extension
^
=

difficulty.

624

and
is analytic
"j"v{^)
2; and evidently

that*

so

OF

THEORY

variable

BESSEL

uniform

J^ (7nx)" m^

We

for all finite values

(mx)

(" )"*(pv(nix)is the

that

observe

now

[CHAP. XIX

FUNCTIONS

of the

complex

residue

at

^^

of the function

{xz)
7r"j)v
suxirz

where

0, +1, "2,

; and

so

the

and
origin,
The
and

to

part of

contour

is

is

shall consider

J,

;.^,(^

f
J

ziTi

in which

we

sin

integral

TTZ

its centre

circle,of radius M-\-\,with

integerwhich

an

the

will be made

to tend

at

the

infinity.

to

F{z) is assumed to be one- valued throughout the 2^-plane,


in the finite
at infinity
be analytic
(cf "19"24);its only singularity
at the origin.
the plane is an essential singularity
function

integraltends

the

By Jordan's lemma,
|.
providedthat v "

to

zero

as

tends

to

infinity,

"

residues,that
calculating

It is evident, by
S

(mx) +F{{-y [F (m) "}),

(- 7nx)]
m) (f),

OT=1

is

equalto

the residue at the

originof

-F(z)'^i^^\
sm

irz

that is to say

i (-)"^{F (m) (f",


(mx) +

(2)

F (- m) "j",
(-

mx)]

"i=i

27ri

/"""+'r-...-^*.('t^)
'
I

'

cm
sin

function f{x) into


problem of expanding an arbitrary
reduced to the determination
Schloinilch series is consequently
F{z) in such a way as to make
The

TT'z

TTZ

generalised

of the form of

(0+)

''F{z)'^^^dz
^iz)-

2771

difi"erby
*

The

theorem

constant

insertion
when

fi'om

of the

the contour

TTZ

f(x).

factor
of

Sm

{hzf

in the

denominator

integration completelysurrounds

makes
the

(z) amenable
"p^,
origin.

to

Cauehy's

19-22]

SCHLOMILCH

We

of the functionF {z).

The construction

19*22.

take the contour

now

integral
i^'^^j.^.^'^^^^i^^)

27rzJ
calculate it in

to

in
expansible

"

where

{z) denotes
-v|r"

whose

degreedoes not

the

{z)is

in the

terras

TT^

sin

ttz

the coefficients p^ will be defined later.

sm
("r"+i
TT^)

"=i

7r'^+^cos

nvr

{z)]
j

[2"+i

sm

,
^

^"^

{n+l)\

smirz

,
^

TT^

Jt"7i7i-^COs|?i7r
I

.^x

we
consequently

PrJMxf

proceedto identify
Pn{\ixy

00

,^^V{^n+l)T{^n

For this purpose we


have to
has differential coefficients of all orders at the
coefficients 'p^

by

the

v+\)

/(a;)
~/(0).

that /(a;)
temporarily
and then we define the
origin,
assume

equation

.
.

PnihiT

/'-'(Q"_

We

tt"'

immediatelythat

Ziri

(3)

expansionof

it is evident that,for small values of |^;,


i^(2^),

"=i

It follows

n, and

2.

with

that F

that

this definition of

sm

of those

sum

exceed

reader will observe

With

and

shall suppose

we

^w=i?'"-*#,

(1)

The

sinTT^'

simplemanner,
type*

series of Filon's

dz.

^(^)

"]

and, in order

625

SERIES

this

equationdefiningpn in such a way that the sum


in a compact symbolic
F{z) is defined,is expressible
transformation of the series for F{z) can be effected by expressing

next

transform

r._i2

of the series,by which


form ; the
*

This

type of series is fundamental

expansions;
denotes
w.

the
B.

F.

thus, in his work


sum

of the terms

on

whose

in Filon's

Fourier-Bessel

degree does
.

theory,and

series,sin
not

exceed

ttz
n

is

is not

peculiar to Schlomilch

and '/'"(;)
replaced by z~''J^,{^z)
expansion of that function.

in the

40

626

THEORY

the coefficients pn in
make
use
we
purpose
that
"

We

^.

"

form

[chap.

FUNCTIONS

involves

Eulerian

of

then

BESSEL

which

be convergent, we

they may

\"

OF

onlyas

of the
integrals

exponent. For this

an

first kind, and, in order

shall find that it is necessary

to

that

suppose

have

^_r(i)r(i.

i)

-^r*-"^'^-"(0)"(i
w{r^V"^'

^-

XIX

dt

dt.

i'iv

dt
and

so

obtain

we

P"

where

stands

Now,

in

if we

the

-'')-'-'
I.
1.(1
-,M^(t^)

^2.

for

symbolicformula
dt,

D''f{tu)
M

djdu.

arrange

the series

of z, it is easy to

descendingpowers

"

verifythat

sinhTrD
ylrn(z)D''_

and

du^

i"2"+"

TT

'

Z))

(t"
-

therefore
'

sinhTri)

^"^)=r5^/i(i-"r'-',^[*"
,,,

(5)

7r(D

consideration

Again,a

of

(2) shews

^
and
we

we

able to

are

effect

our

that

Pn tt"

purpose

need

we

cos

"

izy

dt.

M=0

the series

to sum

\ nir

by making

use

of formula

(4),whence

find that

(6)

COS
^ntr
Pn'Jr^

(sinh irD

T{\-v)]^^

We

have

now

to

for the value

the
zero

dtY \

77

l\f{tu)
M

dt
0

for all the coefficients in the


symbolicexpressions
expansionof f{x),but it is necessary to transform
to be
useful forms, by findingthe significance
more

obtained

Schlomilch
generalised
these expressions
into
attached

^^

symbolicoperator
of

z.

"

-"j,

:-t

both for
,

generalvalues

of

and

628

0, "1, "2,

at

OF

THEORY

is concerned,we

...

calculate the residues

and

BESSEL

omit the second terrti

may

TTZ

by the formula
{z)is defined

e-i^^f{tv)dvdt.

dt

Again,from ^ 19'22 (6) and equation(1) of

this section

pnir^^cos^nir r(l)/(0) p
Z.
(r^+ i)! -Tjf:r^]^^-'^'
"

("^^

/;^^

2ra-.,ra)
first term

The

0; when

We

rightin (1),

{xz)
TTff)^
sin

the

on

of
F{z)

where

[chap. XIX

FUNCTIONS

It
H
^ /:/(^^"
'^-

rightin (3) is equalto F (v + l)/(0),except when


vakie of the term
in questionis zero.

thus obtain the

(4) f{x)

dt""'

the

on

0, the

dt^^

^'-

have

we

"

expansion

(0)F(0)-\- i (-)'"[F (m) 4",{mx) + F (- m) cf",


(- mx)].
"f",

m=\

In the

case
special
term/(0)

additional
When

0, the modified

be inserted

must

change

we

functions and

the

notation

/(^)

2^0
=

r(i;+

form

(3) shews

of

that

an

rightin (4),

the

on

for Bessel

normallyused

the notation

to

Struve's functions,the

(5)
where

in which

expansionbecomes
o^mJv {mx)

l)'^Jl

6,"H^ {mx)

{\nfixy

f
"i ^^""-'^
i W jj^''^
/o
\ r(i-l)r(i)
\
'-

(6)

""^^

'''

dt,
dt.

This is the

19 "24.
We
the

prove that,when
when
F {z)is bounded

now

function

will remember

that

the convergence
We

the

expansion.

cc.
of F{z),as \z\-"-

The boundedness

shall

of Schlomilch's

form
generalised

the

function

12 |

-*-

f{x)

is restricted

assumption that F{z) is


integral.

suitable

be the value of arg

whatever

oc

in

bounded

was

made

in

"

z.

manner,

The

19'21

leader

to

secure

of the contour

take the series of

"

19-22

(1),by which

F{z)

defined,namely
originally

was

Pn^niz)

"

-n

'

n=l

When

t When
consequence

is

negativeit is
0,

of the

the

necessary

expression

to

use

for oq has

in value
discontinuity

modified

expressionfor

to be modified

of the

expression

by
on

the

the insertion
the

rightof

integrals; cf. " 19'3.


of the

(3).

term

2/(0),in

19-24, 19-3]
and

divide

where

SCHLOMILCH

parts,namely the first iV

it into two

is the

629

SERIES

and

terms

the

remainder

of the

terms,

integersuch that

iV^7r|2l"i.V+l.
When

JV,the

of

terms

(z)do
"(//""

exceed

not

tt**"

n7r"-l|3h/("!)
! '-LS

12 |"/n!,and therefore,when

^ JV,

'

":

,71+1

When

have

^ JV, we

'

\z\.{n-i)\

|-^n(s)|"

^ ^'

'

tt

sinh

and

12 1

therefore

i0i"=i (n-l)!

Trl^jA+i"=o

|2|"

sinh Tr\z\

tt

bince
7liV-+l

tends

to

12 1

as

zero

12 ]^-

as

it is evident

o)

that

sufficient condition

for

F{z) to

be bounded

is that the series

Qc

-*-

00

\Pn\

2
n=l

should

be

convergent

and

this is the

7.''+

2
IS

that

i{f{x) is such

ease

i|/(")(0)|

convergent.
19*3.

The

expansionof an

ai^hitr^ary
functioninto

Schlomilch
generalised

series.
that

Now

the

forms

coefficients in

of the

expansionhave been ascertained


specifysufficient conditions for

by
the

generalisedSchlomilch

the

Filon's method, it is

easy matter
expansionand then

of the
validity

an

to
to

establish it.
theorem

The
Let

be

which

number

in the interval ("

tt,

we

shall

prove* is as

follows:

such that

and letf{x) be defined


"^"v"\;
arbitrarily
to the following
conditions:
tt),subject^

(I) The functionh{oc),


definedby the equation

h{x)
exists and

(II)

is continuous

in the closed interval ("

functionh(x) has

The

(III) If V

is

^vf{x) + xf{x),

tt,

tt).

limited total fluctuation


in the interval

(" tt, tt).

I the integral
negative

/:
absolutely
convergent tuhen

is

The

expansion is

1904),p.

348;

but

quite inconsistent

certain

t If

with

stated

those

is

Fourier

is

by Nielsen,

formulae

effect of conditions

t The
of

the

which

he

(small)number

Haiidbuch

gives

der
the

tor

either

Theorie

der

or
negative.
j)ositive

Ci/linderfunktionen(Leipzig,

coefficients in the

expansion

seem

to be

given by equation (2).

(I)and

series connected

(II)is merely
with

to

ensure

the

uniformityof

}i (x).

this Lipschitzcondition
positive,

is satisfied by

reason

of

(II).

the

convergence

630

THEORY

Thenf{x)
,

admits
.,

(!)

OF

BESSEL

[CHAP. XIX

FUNCTIONS

of the expansion
^ argJ^ (mx)

lap

^_

f{x)-^-^^^^y^^

6m H^ (mx)

^^"y^

where

w/zew

"

wi

Mir

rn

sec^""*"^
d)

additional
0; the value of ao is obtained by inserting
an

2r(^
on

the
We

rightin

(3)

/(^)

it will be

[Note.

r7-ZW7r^

continuous

sec-+i

of
(absolute)
convergence
(III). It should be observed
be omitted

when

the formula

is

^ [sin-4"{/(^sin"^)-/(O)]]dc^.

(^

the

in this formula

contained
integral

that the

positivein

aggregate

view

'P"d^'^'P-

'^'

only when

is

is valid

givenby

\)f{())

/,
which

equation
integral

sin ^) ^6^ ;
% (a.

cos-

solution is

The

equation (4) may

(2).

discussion of the

If
rTfl^)

condition

on
investigation

proved that

(4) g{x)=^T{v

by

i)/(0)

the first
equationof the system

shall base the

term

of

is securetl

containing /(O) in

terms

of the formula

rW)

positive.]

We

is a
proceedto verifythat the function g (x)defined by (4) actually
solution of (3),by takingg (x)to be defined by (4),substituting
in the expression
the rightof (3),and reducingthe result tof(x).
on
The
2

cos

result of substitution
vir

is

f^ C^'^

d,

cos-

e sec-+^

d"

[sin-6 [f{x

sin 6 sin

"h) /(O)}]d^dd
-

+/(0).
Hence
2

cos

vTT

we

have

fi"'r^

to

prove

that
/7

cos-(9sec-+i(^^[sin-(/"{/(a;sin6's
=/(^)-/(0)-

Replace"^on

the left by

variable

new

sin ;^

defined

-y^

sin ^ sin

by

the

equation

"/",

in the resulting
change the order of the integrations
absolutely
convergent
and then replace^ by a new
variable i/rdefined by the equation
integral,
cos

cos

;^ sin

"/r.

19-3]

SCHLOMILCH

631

that

thus deduce

We

SERIES

d sec-"-^!4"

cos-"

Jo J n

,^

[sin^"
(^{f{x sin 6 sin "/))
-/(O)}]cZc^ci^
,

sin^cos-"^

rf-r

Jx

"i-

II

/-/^MT

,/i

d
sin x) -/(O)}](^x
[sin-;,":
[/(^^

tan-^"
tan-'" ^/.
rf^/.
\|/-rtY.

^^

"

/./

["""''
X l/(^"m X) -/(0)1]Mdx
,

""T. 3^
(75?^iSrs"

"

and

hence

-/(O)},
(^ + i)r(i- 1.){/(..)

ir

the formula

be

to

is evident ; and

established

so,

when

[/ (,c)is

by (4),then equation(3) is satisfied.

defined

Now, by Fourier's theorem,


2

(a;) l^o+

cos
(o.,n

b^ sin mx),

mx

'

'

where
1

f"

TT

'

(u)cos mudtii

g
_

(5)

f"

^m =1
and

it is easy to
total fluctuation

verifythat

mudu;

5^(") Sin

when

f(x) is

in the interval

("

tt,

continuous

tt)so also is

expansionfor g (x) is uniformly convergent when


is an arbitrarily
small positive
number.
has
at

g
+

ir

"

function

with

limited

(x),and therefore the


S ^x^tt

"

S, where

Replacex by x sin 0 in the expansionof g (x),multiplyby cos-*' 0, which


and integrate
an
absolutely
term-by-term: we deduce
convergent integral,
that

once

/"/

and

this
The

those

that

values

of a,^ and

givenby

botn
when

unable

to

formula

be

we

noticed

come

any

ctm^."
(mx)

b^ given by

formula

bmilu {mx)

"tt+S^x^tt
(5) are

"

S.

easilyreconciled

with

(2).

0 (1 /?n)when

make

uniformlywhen

expansionconverges

It should
are

t^o

to

that,by the Riemann-Lebesgue lemma,


is

large.This

deal

progress

with
without

any

seems

to

be connected

Schlomilch

assuming that

","

with

and

6,",

the fact

(" 19"6"2)we are


is convergent (or
S6,"/?yi
appear in " 19'62 to be
series

equivalenthypothesis);this assumption will


Struve's function satisfies
the diff"erentialequationwhich
necessary because
in
is not homogeneous,so that Struve's function is not of a type which occurs
solutions of Laplace's
equation; there would conseequation or the wave
some

632

OF

THEORY

to be

quentlyseem

of

reasons

placedon f{x)
expansion.
have

been

[Note.

Just

as

ivf{x)+ xf'{x)
about

in

is not

to

(II)concerningthe

limitations

which

of the Schlomilch

limited total fluctuation

of

made

all statements

to be

for the

the existence

ensure

condition

then
satisfied,

of series have

convergence

physicalcharacter

in order

" 19-11, if

[CHAP. XIX

FUNCTIONS

BESSEL

in this section up to this point


about summability (Cl).]
statements

replacedby

importantconsequence which follows from the fact that a"i


and 6^ are both 0 (1/m)when
2vf(x)+xf' (x) has limited total fluctuation
and tt, the general
in (" TT, tt),namely,that in the neighbourhoodsof
tt
and so the expansionrepreof the Schlomilch
term
expansionis 0 (1/m''"*'^),
sents
the
continuous
function ; hence
to
a
expansionconverges (uniformly)
the sum/(^) throughoutthe interval (" tt, tt).
is

There

one

"

19*4.

i^epresented
by Schlomilch
Special
functions

There

are

mentioned
and

few

givean

now

simpleseries

very

Schlomilch

Physics(otherthan
series

of various

account

in

occur

researches

the

problem

natural

in which

manner,

Schlomilch

found.

to be

are

Mathematical

problemsof

" 19"1) in which

in

shall

we

series

series.

is

1+
this series is convergent when

e-'"' Jo

and

coordinates,it is a
cylindrical-polar
0.
of space above the planez

(mp) ;
and, if p and
positive,
of Laplace's
equationat

are

solution

denote

all

points

transformations

Various

thus, by

coordinates

changingto Cartesian
1 -f

(1)

series have

of the

^
27r j _"1:,

^ e-^^J,{mp)
=

'^

When
Q){z-V ix

X-

cos

11

-h

iy sin

If this is done,

(2)

1, the

-\-y^ "\-z"-"

1 -f S

"

be

-.

t?/ smw)}

expanded in ascendingpowers

getf

we

e-'-/" {mp)

s"

6) are
(/",

ia;cos
}" ,(2:-f-

u.

where

r"

integrandmay

^
llT

,"=!

coordinates

given by Whittaker*;
using" 2'21,we have

(x,y, z) and

exp

"

",

been

the

'-

J ^T,Z

"

^-^
+

~--^

B^, Bo, B^,


Math.

Anil.

t Cf. " 4-8 and

are

...

lvii.

(z +

iys\T\u

ix

cos

iy sin u)^"~^du

{2m)\

,"=i

polarcoordinates

{p,z),and

+
'^'^-.-^2
+

COS, u

IX

correspondingto

(1903), pp.

Modern

the

Bernoulli's numbers.
341"342.

"" 7-2, 18-31.


Aiiahjsis,

cylindrical
-polar

19-4]

expansionfor 1/(1"e~^) in

the

from

deduce

we

(3)

1 +

=1

e-^-Vc,(mp)

:^

,tiLVi(2w7ri+ 2)-+

It follows that

set

of

of

the

reader may

I 13*2

as

Some

of

is

due
potential
negative)at the originand

examined

"

at

type.

by Nagaoka*

series is derived

such

to

Lijjschitz-Hankel
integral

series of Whittaker's

equal to l/\/(l ^')in

series in

Fourier

been

One

problem of Diffraction.

The

some

y-^_

x' +

electrostatic

find it interesting
to discuss the

limitingform

which

series represents the

other series have

the function

if] ^/[{2m'I^i
zf

X-

charges(some positiveand
imaginarypoints.

The

^rmri)

t +

^.^

00

of unit

2m7ri

that

set

^. 1^"

9+

112,

l-e

whence

givenby Whittaker, is obtained


fractions;this expansionis
partial

also
of the series,

transformation

Another

633

SERIES

SCHLOMILCH

in connexion

the Fourier

from

the interval

with

series for

(" 1, 1).

questionis

(4)

^;r

,^

iTT +

TT

\/(l X-)

COS
t/,,
(?H7r)

WtTT.r,

and

it converges

is any

to

,"

uniformlythroughoutthe

interval

a:

(5)

and
integrate,

e"** and

dx

",axi

le

series

("1, 1) and

so

on

.axi

A), where

we

sm

1 +

q2

let? t

generalresult,valid
sin

are

1=1

when

min

sin

mirx

^y^2^2

interval

/^

a,

(-)'"Jo(m7r)
a"

"

m-TT-

(y + 1) " 0, is

Journal

"

"

J" (a)

TUTTX

(alsodue

rightin (.5)
converges uniformlythroughoutthe
take
1 and 1 as limits of integration.
may

Jo (")

0)

cos

the formula

the

(6)

"

Jo (m-rr)

Hence, for all values (realand complex)of

more

obtain

then

we

1
+

TTJ V(l-^")

The

formulae

A, 1

1 +

number.
positive

Multiplyby
Nagaoka)

("

(i.+

1)

W)

",=i

m"

(a-

of the Coll. of Set.,Imp. Univ. of Japan, iv. (1891),pp. 301


Chnento, (4) i. (1895),p. 1;""2.
quoted by Cinelli,Nuovo

m'TT')

"

322.

Some

of

Nagaoka's

634

THEORY

This

BESSEL

OF

[CHAP. XIX

FUNCTIONS

expansion is also obtainable by expressing

"

as

"^

partial

of

sum

fractions*.
Various
the formula

of the integralon
the left of (5) were
representations
be
the most
of them.
to
quoted seems
interesting

Finallywe

give the

shall

is deducible

from

the Fourier

^ cos(2m

w
by replacing
by

sin d and

example of

an

denotes
then, \if{(t)

the

the

tt

of the

interval

("

^9

f /"arc sin(7r/x)
+

(Jo

"""' arc
"

"

^^^

ix~(2^;r:riy
-

19'41.

Null-functions

shall

now

providedthat
+

when

00

This

0 "

^c

"

/(.rsin^)rf^
(Wj-)J

sin

Sir)dO,

tt^

/7r\

(-j

.,

.,^

-)

as
expressed

; the

tt

that

\ (2x sin e

V(^

^+ i

(1)

lir,and

7r"x"

sin (n/x) ^

Ix

theorem

arc

Schlomilch

the remarkable

prove

see

^tt.

series when

Schlomilch

series,we

rin-

0 to

have

'lir,we

^ Jo [(2m- 1) a;}

/Q\

We

tt

6 from

^(7r-2^sin^)rf^
""

that,when

shall take

tt),we

.'arc

SO

tt,

of

sum

of the Fourier

sum

with respect to
integrating

the calculation

the variable lies outside

series

l)a;

"

"

("

This

by Nagaoka

formula

J,^^|^^ "-^-

"")

As

obtained

cos

series.

that

(-r Jo(wA-) 0,
=

series oscillates when

and

divergesto

x=7r.

theorem

has

analoguein

no

the

theoryof

it is

definitely
known]:that a Fourier cosine-series
function throughoutthe interval (0,tt).
*

Cf. Modern

t This

was

J Cf. Modern

Analysis," 7-4.
as
a problem in

set

the

Mathematical

Analysis, "" 9-6" 9-632.

Fourier
cannot

series,
and, in fact,
represent

Tripos,1895.

null-

636

OF

THEORY

be

may

equation(2). The effect of applyingthe operator


mJ^ {mx)l{\mxY is to multiplythe function by
; and

times

appliedn

to the function

once

therefore,when

[CHAP.XIX

FUNCTIONS

BESSEL

to

"

"

a; "

tt,

0,

that is to say,

providedthat

either
and

(i)"\"v"1ni^^

0 ^

a; "

(ii)v

tt, or

2n "

Nielsen*

and

Math.

0 ^

."

tt.

(1899),
published about the
subjectwere
Videnskabernes
P" (1899), pp. 73"81
x.
same
time, A^t/tTidssh-ift,
; OoersigtK. Danske
values of
three
two
these
661
665.
In
the
of
first
integral
1899, pp.
Selskahs,
papers
third
the
made
in
of
values
t
he
extension
to general
being
paper.
only were considered,
The

pp. 582

formulae

"

587

in this

given
other

; two

section

by

papers

due

are

Nielsen

to

Ann.

lii.

at

this

on

"

for the

Shortlyafterwards t Nielsen gave a formula


of
this formula is easilyobtained from the integral
(-)

by consideringthe
It is thus

behaviour

found

(2)when

x"n

they make

it

Dirichlet's type

/-"^cos(J/+i)..

,
_

of the

that, when

of the series in

sum

integrandat

is

positiveand

z(

7r,

Stt, Stt,

integersuch

q is the

that

(25'-l)7r".r"(2^+l)7r,
then

The

r(v

%T{\)

I {-r,h{r,ix)_

_J__
^^"

l)"^",=,{hmxY

of
impoi'tance

that, when

xT{v

Nielsen's

function

{^n-lYn^-y-h

'1

i)n=x\

lies in the fact that

formulae

is defined

"

X'

for the interval

if the
tt),

f{x)
(" tt,
series
Schlomilch
a
can
throughout the interval
by
represented
the representation
is not unique
of points)
at a finite number
(except possibly
series which are equalto
unlimited number
of Schlomilch
and there are an
the function y (a;)
throughoutthe interval,except at a finite number of points,
togetherwith the originand (when
namely the pointsalreadyspecified
^ " y ^ I) the end-points+ tt.
evident

be

function

"

The

theorem, that the only Schldniilch series with non-vanishing

converse

which
represent null-functionsat all poiiitsof the interval
coefficients
{whe7il ^ " V ^ ^) except the originare constant multiples
'7r"x"7r,
of
"

"

Formula

(1)was

t Ovcrsigt K.
Nielsen, Ann.
J The
any

di Mat.

theorem

Schlomilch

rediscovered

Danske

by Gwyther, Messenger,

Videnskabernes

is untrue

when
than

lof
the

one

for

301"329

; cf. formula

given

can

more

(3). It

(1904),p.

xxxin.

101.

also

later paper
a
by
compHcated results. Cf. " 19-4(9).
would
be interesting
whether
to know

Selskahs, 1900, pp.

(3)vi. (1901),pp.

series other

{-)'^^h{mx)

represent

55

"

60;

see

null-function

when

^--:"'"f.

19-5]
is,of

SCHLOMILCH

of

course,

much

deepercharacter,and

yet been published.We


this theorem

to

up

features,the
series,

discuss

now

analysiswhich

due
analysis*,

it

series of

will be

Riemann, which

to

that

seems

proofof it has
which lead
propositions

used

no

resembles, in its main

is applicable
to

trigonometrical

Theorems

19"5.
We
and

the

shall

637

SERIES

shall

in which
case

concerningthe
discuss the

now

Struve's

of Schlomilch

convergence

special
type

of Schlomilch

series in which

0,

of takingthis parappear; the object


ticular
is to avoid the loss of clearness due to the greater complication
functions do not

in the

of the formulae
in the more
appearance
the
in
the
exceptions, complications
generalcase

Only; those which


"" 19-6" 19-62.
The

series.

series

are

to

now

not

of detail

matters

be considered

With
few
a
generalcase.
in detail
are
complications
will be dealt with fullyin

is
00

i"o+

(1)

in which

the coefficients a^

shall first prove the


condition that a^ Jo ('''*')0
^

analogueof
as

ni

a,n
If the

t, namely that the

of values

originis a point of

{\fm).

the interval in
"m

is

lemma

ni.

,at all pointsof any interval

oo

-*

Cantor's

of

that

to ensure
of uc,issufficient

[Note.

cimJoimx),
l

arbitrarily
given functions

are

We

question,then

the theorem

that

0(l)

obviouslytrue.]
Take

let this

J of the
portion
portionbe called /j.
any

Throughout /j we
a"i,Jo

and,

(mx)

as

ni-^

have

interval which
Let

the

the

and
origin,

"

"

\ir) Q{nix,0) sin {mx


"

"

\'tt)\
;

oo

Hence, for

all

Q(mx,0)^0.

sufficiently
largevalues

of m,

(sayall values exceedingmo)

all pointsof /j.


Now

suppose that a^
leads to a contradiction.

is not

(\/w);

we

have

Cf. Modern

X Since
the

contain

(c" " 7'3)

("^^J [-^i^mx,
0) cos {mx

a,"

not

lengthof /j be L^.

P(mx,0)^l,

at

does

Analysis,"" 9-6" 9-632.


of
function
Jg ("nx)is an even

origin without

loss of

generality.

to

shew

that this

hypothesis

t Ibid. " 9-61.


.r, the

portion may

be

supposed

to be

on

the

right of

638

THEORY

is not

If am

OF

BESSEL

number
(\/m),a positive

Km I"
whenever
mi,

Wg,

mo

and

is

exist such

must

that

\/m

belongingto

Let the smallest member

tris,

certain

of this sequence

sequence
^ceeds both

which

cos(mi'x
^tt)goes through all its phasesin I^,and
portion
f of /j,say /g. such that

unending

called ?h/.

27r/Zibe

Then
be

value

given any

[CHAP. XTX

FUNCTIONS

"

there must

so

\^ | ^S, \sin (niiX ^tt)|" ^


Icos (m^'x l-rr)
-

all pointsof /g. If Xg is the

at

lengthof I^, then L2

I^Text let the smallest member


be
27r/Zy2

and

Then
be

cos

(mz

which

exceeds

^tt)goes through all its phases in lo,and

"

/o
,

both

mj'

there must

so

say I^, such that

X
Icos {7110

Iv) I^ ^ \/S, Isin (m^'x

"

{tt)\^ ^

"

all pointsof I3. If L3 is the

at

vir

^Tr/m^'.

called wio'.

portionof

of the sequence

lengthof I3,then Z3

jTr/mg'.

By continuingthis

I^,...
process, we obtain a sequence of intervals I^,I.-,,
that each is contained in its predecessor
is
therefore
there
a
pointX
;

such
which

lies inside all these intervals,


and

Icos {mX
when

has any

For such

this

pointwe

itt)I^ |-\/3, Isin {inX

"

of the values m^, m^,

values of

at

"

have

\ir)\ ^ \,

mj,

have
consequently

we

ittm /"i^^) I^ I"m Ia/


X

rmrXj

[P (mX, 0) Icos (mX

sin

^tt)] |Q (mX, 0)

(mX

^tt)|]

V34

and
all

this is inconsistent with

pointsof
The

the

Uxj'

hypothesisthat amJoii^^)tends

to

zero

at

/j.

contradiction

which

has

now

been

obtained

shews

that a", must

be

(s/m).
The

converges

It is

t There
we

next

theorem

throughoutany
supposed that
are,

take

shall prove is that, if the Sclildmilch series


condition
interval,then the necessary and suffi.cient

which

m\ ".m^"m^"

in fact, at least two

I^ to

be

we

that

such

portion which

portionsof Ij ;
lies

on

in order

that Zv may

the left of the others.

be

mined,
uniquelydeter-

19-51]

value of
for any positive

that the series should converge


the interval

not)is

or

{mx

\nnrxj

theorem

is evident

series
trigonometrical
0

(amni~^)

"

(m~-);

The associated

19"51.
Let the

of x.

any

Itt)+

7-

sin

"

11

fact that

(m~-) is a convergent

(mx

"

^tt)

tt/y

a^nJo{mx) by

the
a

of

general term

function

of

the

which

is

series.

function.

of the series

sum

at

(whethera point of

"

the

from

differs from

and

"

should be convergentfor that value


This

that the series

cos

TO=i

639

SERIES

SCHLOMILCH

pointat

which

the series is convergent, be called

/(a-).

Let

(1)
Then

F(x)

sum

It is easy

to

associated

with

the

Schlomilch

isf{x).
see

of any intet^al,then
of X.
am

"^^-^).

(x) will be called the function

series whose

For

iaoX^-

Jo (nix)
-^

that, if the series defiling


f{x) converges
the series

as

definingF (x) converges for

all

at

oo

m^-

pointsof

at

all points

all real values

the interval,and

therefore

the

must

("19-5)
am

Again,by " 2*5 (5),for

o('sJm).

all real values of

IJo (^^) k
and

1 5

consequently
am

Jo {in^)
\m^

m?

Since
00

i oi"\

2
=

is convergent, it is obvious

that

Wv

the

series

on

right in (1)

be

convergent.
It is evident,moreover,
that it is uniform

not

onlythat

throughoutany

the convergence
is absolute,but also
domain
of values of the real variable x.

640

OF

THEORY

19 "52. Lemma
We

I.

(x) is the
prove that, if
is f(x),and if
series whose sum

Schlomilch

now

^,

,,.

(1)

functionassociated

"F

shall

[CHAP. XIX

FUNCTIONS

BESSEL

G(x,a)

"

ivith the

2a) -2xF(x)

+ (x-a)F(x+ 2cc)
(cc+ a)F (or.
^,

then

limG(^,a)

(2)
at

any

pointx

ivhich the series

at

defining
f{x) is convergent,providedthat*

dm

It is easy to deduce

G(x,a)

from

^/(a;)

(\/m).

(1) that

^a,x-1^^^

[{x+ a) Jo (mx

2ma)

{x

a)J^ (mx

"

2ma)

"

2xJo {mx)];

"

theorem, it follows that


and, from I'Hospital's

\(x+ a) Jq {mx

lim

lim

"

Jo {mx

"

2ma)

2xJf){mx)]

"

2ma)

"

2?"a)

"

2m

{x

a)Jo {mx

"

"

2/"a)]

xJo {mx).
limits of the individual

the individual

of the series

terms

It is therefore

sufficient to

uniformlywith respect to
has any

a) Jq {mx

"

Jo"{mx) + Jo'{mx)/m

Consequentlythe
are

{x

2ma)

{x + a) Jo'(mx

2m

+
X

[Ju {mx

"

2ma)

om-a.

"^o

"

in

an

definingG(ie,a)

definingxf{x).

prove that the series for G {x,a) converges


interval includingthe pointa
0 when
x
=

that the series

value such

of the series

terms

is convergent.
for/(ir)

that x is positive
It may be assumed, without loss of generality,
f, and we
shall now
shall then take |ct |so small that it does not exceed Ix ; we
prove
that the series for G

uniformlywhen

{x,a) converges

"

|a;^

|a?.

By observingthat
a^
+

a"

+ 2a.)]
\/\x
{x
^
^
^^
=

^t-

Jq (mx " 2ma)

am

im^[x "
Since

not

are

we

assuming

permissibleto infer from


t The
case

that the series

^.

itt'/x,

"

and

x"a+\/{x{x"2a)}

functions

in which

under

0 needs

than

more

" 19'5 that a," must


consideration
no

further

are

\/{x{x" 2a)]]

the
be
even

of

convergence
o

singlepoint, it is

not

{^/m).

functions

consideration.

f (x)at

of

; and

since G

(0,a)

0, the special

19-52]
is

signsthroughoutbeing taken),we

lower

uniformlyconvergent (upper or
that G {x,a) differs from

see

W{x

by

the

2a) Jo (7nw+ Ima)

"

take the

Now

of this

generalterm

^im^'r^(^.
^ 2a) J" (mx

"

\Jx J,,(ma;)]

uniformlyconvergent.
the convergence

of
uniformity

of

series,
namely

2mo) + V(^

2ma)

written down.

has been

the last series which

"

is

of which

each
series,

of two

sum

2a) Jo {mx

\lix

It is therefore sufficient to establish the

641

SERIES

SCHLOMILCH

2a) Jo (mx

2ma)

V^

"

-h ("ix)],

write it in the form

and

ttr
"

|_

\miT)

2x\

y(

cos

"

JV

^mx

imx

l-7r)sin 2ma

"

"

{x?

'6m

4o-)

"

XtmrJ

'

i*"mx

24"

where

m-a.

sin(ma; jtt) cos2ma

(2x\

{a;- 4a-)

?h-

"

{mx)

"

"""VimTT.

"I"{mx +

"

2mo)

4m^a^

1]cos {1/ i,r)


_

|1 Q (^,0)|{y
sin

generalterm is thus expressedas the


proceedto prove that each of the four series,of
generalterms, is uniformlyconvergent.
The

The

series

firsttwo

theory of
uniformlyconvergent
the

deal with

To

2/?ia) ^ {mx

(y) is defined by the formula

"!"^;i/)[P (y,0)

with

ma

theorem, numbers

Itt).

of four terms, and

which

these

terms

we

the

are

provedto be uniformlyconvergent, in connexion


series*; and the third is obviously
trigonometrical
are

from

the test of Weierstrass.

the fourth series,


we

sum

6 and

observe

that,by the first mean-value

6i exist such that

-1"^"1,

Cf. Modern

obtain

various

Analysis,""9'62, 9'621. It has been the general (butnot invariable)custom


to
propertiesof the series without establishingthe uniformityof their convergence.

The

convergence

can

be dealt with

t The
W.

number

B. F.

-l"6'i"l,

in consequence

f (x)is required to deal with the first series ; the second


of the less stringenthypothesisthat
o(,^'w).
""j

6 is

of

of

the series for

function

series

variable

t which

will be introduced

immediately.
41

642

BESSEL

OF

THEORY

[CHAP. XIX

FUNCTIONS

for which
2"I"(mx)

"I"(mx + 2ma)

"

(mx

2?Ha)

"

--J

{mx
I {4"'

2ma

2mat)

"!"'
(mx +

dt
2moit)}

Jo

Since

"1""(_?/)0
=

I 4"mat "!""(mx
Jo

2ma

4tm.^a^"t""(7nx

"

when
(l/^/")

y is
2

is

with

uniformlyconvergent
Hence

2madi).

"

it is evident
large,

with respect to
uniformly

respect to
when

of six series each of which

(x,a)
of the series for Cr{x,a),i.e.it

the terms

of the limits of
equalto the sum
IS equal to xf(x),providedthat the
the lemma
to be proved.

series for /(a;)is convergent ; and

prove that,with the notation o/


that
to ensure
o{\/m) is sufficient

shall next

{x + 2a) -\-(x

(x + a)F

this is

II.

19 "53. Lemma

converges

therefore

and

^x;

lx"a"

"

is

that arn

a.

sum

lim G

We

that

"i""(mx-2mae,)

(x,a) is expressedas the

2ma9t) dt

a)F (x

""19"51,19"52,the

2a)

2a;P

condition

(x)
_

a-*o

for all

values

of x.

only; and we express


to
seen
of six series each of which is easily
the series for a G {x,a) as the sum
cerning
be uniformlyconvergent when
\x, by applyingthe theorems con"\x"a"
in
19'52.
used
which
series
were
"
trigonometrical
As

in

" 19"52,we

values
positive

consider

need

of

Hence
lim

[a G (x,a)]
"

lim

"

(looo)
"

a^o

1 lim
"I
m"ia^o4/"ra
+

this is the lemma


19-54.
We

can

(x

"

a)Jo (mx

"

2ma)
2ma)

"

2xJo (mx)]

0,

and

[{x+ a) Jq (mx

The
now

consideration

to

be

proved.

analogueof Riemann's
prove that,if two
with v
0, and
(i.e.
=

Cf. Modern

theorem*

Schlomilch
with

on

series.
trigonometrical

series of the

Struve's

Analysis,"9-63.

function

type

now

under

absent) converge

644

OF

THEORY

But

BESSEL

"

function

of

Hence
of

logx

the

and

this is the theorem

y=

logx

Now, by " 19"51, P {x) is continuous


are
an

connected

at the

abruptchange of
^

when

f is an

they

so

and

have

"

the

these

so
curve

curves
logarithmic

have

{x) cannot
Lemma

II,

0,

2a

that is to say

be discontinuous

B cannot

a; "

%af)X^

values

same

at

the

all values

for

exceptional

of

in

the

Make

the series

x-^0;

it is
and

so

77,

"

x-*0,

of segments

set

0)=fF'(f-0).

the

linear

(0,tt).

when
Consequently,

because

T(il-T(i-M
vb

exceptional
point;

the constants

points,and
interval

of

because, by
exceptional
point,

an

fF'(f
Hence

be

must

proved.

(0,tt),and

2a

a-*.0

^ (x) must

B.

exceptional
points; and
direction at

of

even

in

F(f+2")-F(|)

lim

(x)

with

curves
logarithmic

be

to

F {x) consists
equationis y
equationsof the type

whose

curve

that

say

be

the left must

on

that the first term

shews
is to

limit

the

so
(C]+ 2a) ^ (f)
(Ci),
{Ci),
(^(Ci 2a) ^ (f"
(f)

This contradiction
negativeor zero.
be zero
throughout (|i,fa),that

[CHAP. XIX

FUNCTIONS

on

;;^

the left has

uniformlyconvergent.
A

is

limit,namely

log x

Therefore

B.

\ogx +

has

limit when

zero.

when
Consequently,

0 ^x^tt,

the left converges

laox- B

"

and the series on

so integrations
uniformlythroughout(0,tt),

term-by-termare permissible.
Replacex by x
by "1211,

sin 0, multiplyby sin 6, and

^aox'-B=

0 ^x

ttm

Sin

mx

rri^x

^tt,
"

cim

sm

mx
^
_

m=l

m^

0 to

^tt.Then,

""'"f'Vo(m^sin6')sin^(^^

m=l

Hence, when

from
integrate

.3/3

19-6]
sin

Multiplyby

and

mx

from
integi'ate

Since

is

","

"

f(w)

j\ TT'tto

a,n

series

the

on

rightat

We

shall

in

0 and

stated at the

at

We

shall first prove that,ivhen


the series tends to zero
cc
m-^
as
is sufficient to

If the

is

behaviour

that

of the
be

f{a:)cannot

pointunless Uq is zero;

this proves

and

this section.

Schlomilch
of generalised

convergence

series.

series

the condition that the

v"\,

interval

all pointsof any

at

(m

\)thterm

of

values

of
of x

that

ensure

a,n

[Note.

it is evident

tt,

beginningof

the
studybriefly

now

(- )'""t).

" 19'41 concerningthe


a;=

the
concerning

Theorems

19*6.

that

series at either

convergent Schlomilch
the theorem

that

S (-rJo(mw)

1+
2

a.

results contained

the

evident

7r*ar,

(\/m),this equationshews

From

it is then

have

must

we

Hence

tt

To

2m"

be

given to

0 to

TT-B

TTfto

2ni^

645

SERIES

SCHLOMILCH

{m^+^X by,

originis a point of

the interval

{m^+i).

in

question,then

that

the theorem

obviouslytrue.]
Since

the series under

consideration

is unaffected

signsof all the coefficients 6," are


interval on the rightof
lost by considering
an

of

if the

call this interval /j ; and, at all

We

a-,nJv
{mx)

h,"H" {mx)

c^
^m

c",

We-^ow
that this
If

a,"

cos

77",

suppose

6,"

h,"are

c,"

that a^

leads
hypothesis
and

v) cos (mx
[P (n),r,

Q {mx, v) sin {mx

"

a,"

have, by | 10'41 (4),

pointsof /j, we
"

(Imxy+'^x/ir

{^m.x)"

where

by a change in the sign


is
also changed,no generality
the origin.

not

to

both

sin

and
a

r}"^

"

|;/7r jtt
"

"

|/'7r jtt
"

"

tjyt)

+ h,n 0 {ni~^),
7/"()]
"

6",are

not

both

o{m"'^^);we

have

to

shew

contradiction.
a
0(7/^"+*),
positivenumber

must

exist such

that

c,"""m"-^^
whenever
m,,

is

m.,, m.j,....

given any

value

belongingto

certain

unending

sequence

646

THEORY

We

now

prove,

OF

BESSEL

in

exactlyas

[CHAP. XIX

FUNCTIONS

" 19'5, that, at

point X

some

of /j, the

inequalities
Icos (inX

J7,n)
i^

"

satisfied whenever

are

is

IvTT Itt

"

has

value

any

sub-sequenceof the sequence

values of

For

which

\/3, Isin (mX

j^

belong to

both

be

reader

The

this

(w/)

sequence

sub-sequencewe

which

ig (x/3

have

1)

the expression
on
negative,

the

rightcannot

be

arbitraril}-

which

is sufficient to prove that a",, and 6.",


of the Schlomilch
series tends to
-t-l)th term

(m''+*)if the (m

may

now

condition that it may


the interval

for any

converge

^^^^^ ^''^
1'"'
\/("D
-

sm

hvTT

(ma;
-

he convergentfor that value

should

19'61.
liBt

us

associated

The

\,ifthe generalised

necessary and

of

a;

sufficient
{luhether
a pointof

take

pointat

"^"v"-^,

which

"

^'^ ^-^
-

^^

^TT 7}"M +

'

of x.

function.
and

i ""

any

value
positive

"

not)is that the series

or

throughoutany interval,the

"^

S
in

(as in " 19"5) that,ivhen

prove

SchldmilcJi sei^iesconverges

at

belongingto

"

all pointsof Jj.

at

zero

is

is the contradiction

This

small.
must

"

"

(m^).

;a,"J,(mZ)+ 6""H,(mZ)l
and, since

^vir ^tt 7;,")


|^ \

"

let the

a,"

sum

Jy (mx)

of the series
+

b^ H^ (7nw)

the series is convergent be called f^{x).

Let

OqX^
11)
Then

'"'^*'"-

Y"(x)

whose

sum

__

81^1^+2)

^
Zi

will be called the function

a,n

J" (mx)

hm Hp {mx)

i^^.(^mxy
associated

with

the Schlomilch

series

'\^fv(x).

It is easy to prove that,ifthe series defining


f^(x) converges at all points
then the series defining
F^ (x) converges for all real values of x.
of any interval,
The

" 19'51

only respect

in which

is that the additional

the real variable

has

to

the

theorem

be used.

proof differs from the analysisused


that IIy(x)/x''
is a bounded
function

in
of

19-61, 19-62]

SCHLOMILCH

647

SERIES

Again,let
a)
(2) G, (cc,

i[(x-^2va + a) F, (a;+ 2a)

Then, justas in " 19'52, we

may

2va.

(x-

cl)F^ {x

2a)

{x,a)

xf^{x)

any

pointx

a",

and

h^

which

at

both

(ii'"

the series

r(i/+ |)r(^)",=i m

definingf^{x) is convergent,iprovided

Further

we

lim
'"'

hmlm

that

is convergent.

that

prove

may

"

.xt^.ix

that the series S

{nV^^)and

'^

"

-^^

"^

a^o

at

(a;)]/a-.

2,x F^

that*

prove

lim G^

a)] 0,
[aG;,(.'r,
=

'

'^

'

a-*0

"

only that
j)rdvided

a.,n

and

both

6",are

whether
(m'"^^),

the series

is

not.

convergent or
19-62.

bmlm

"

analogue of Riemann's

The

theorem.

SchlOmilcJi series of the same


generalised
and have the same
order v (ivhere
"
" i) converge
sum-functionat all
closed
luith
the
interval (" tt, tt)
possible
exception of a finite
pointsof the
the
that
and.
the
is
number
not all
are
origin
of points(it supposed
points"
Struve's funcand ifthe coefiicients
of the terms containing
tions
exceptional
points),
each
all
then
series are
in the two
to
each,
f
equal
sufficiently
nearly
in the two series are equal.
coefficients
corresponding
We

now

can

that,if two

prove

''

"

ir

and

series be

the difference of the two

Let

^tto

let the

+ bm
'^"t-^fi'"i-^)

of this series be

sum

f"{x),so

at all pointsof the interval

to zero

(-

of the series fov


convergence
tt)necessitates the equations

The

(-

TT,

".",

The

series

ikscib,a. 0
^

mean

We

The

can

presence

as

be

X It

statement

seems

finite number

f"(x) converges
of exceptions.

fy(x) nearlyeverywherein

the interval

(m-'+i).

tions
containingStruve's funcbe
is
to
to
interpreted
sufficiently
nearlyequal

to be

are

oo

-*

in such

of the

provedjthat
series

equation satisfied by Struve's


This
"}"

tt)with

(m''+^),"",

discuss the function

now

iov fv{x).It
*

that the series for

that the coefficients of the terms

statement

in the two

tt,

H^ (mx)

on

way

that

is convergent.

Fp(a;)associated with the Schlomilch series


if the interval (|^i,
^2)is such that the origin

rightis

due

to the

lack

of

homogeneity

in tlie differential

function.

will be made

unnecessary

the

definite

''

immediately.

to repeat the arguments

alreadyused

in

" 19"54.

648

THEORY

and

the

BESSEL

OF

exceptional
points(ifany) are

[CHAP. XIX

FUNCTIONS

internal

not

pointsof

the interval,
then

of a;~^'' in the interval. It may then be shewn


in the form of F" (x),
that the exceptional
pointsdo not cause any discontinuity

F^(x) is

and

linear function*

hence

that

deduce

we

(F^{w)

Ax"'''-{-B,

If,(a;) A'\x [-2-+

B',

where

A, B, A', B'
take the

Now

(0"x^tt)

(0"x^-7r)

constants.

are

equation
aoX^

F. {X)

a"i

J" (mx)

6,"H, (mx)

gp^~^

'

mKi^vixy

^^^^^

0, (which has an absolutely


replacea; by " sin 6, multiplyby sin-''^^^/cos-"
from 0 to hrr. The series for F"(a;sin6^)
and integrate
convergent integi^al)
tions
converges uniformlyin this interval of values of 0, so term-by-termintegraare
permissible.
that

It is thus found

sin-''+^^

fi''sin^-''+^0
dd
cos-" 6

anx^

,^

F^(^^^"^)-^^^^^^
8r(.

2).L

"i

Jv (tnxsin 6) + 6,"H, (mx sin 6) sin''+i6

'^"Urn

J0

/Tx
"

when

substitute

we

i,

"m

"

6^(1

and

on

"

"

"

twice
differentiate,

fails to do

of

the left tends to

We

both

so

limit

unless A

infer that ^

is

now

/o\

(2)
when

"

"""

^^^

or

once

x-^0,

as

the

as

but

case

the

when

0^,i'^

"

vr.

series
be, the resulting

may

on
resulting
expression

A' must

manner

that B and

B'

be

zero

be

must

the

equal.

(1) that
+

^"-^

br"(l-

C0.S

mx)

a^af

"

j^

TT^x^'ir.
V

is

zero

=^-Bxr{.+

-,

When

..

l).

the

right

zero.

S^

",

it is permissible
to differentiate (1)twice
(7?i"+^),
if
^ ; but it may only be differentiated once

0, and in like

follows from

r(|-.)"-'ft''r(.+

be obtained

equationmay

then shews
at the origin
F"(ii';)
It

that

a^x"" Ax^-""" V (%)

12-

similar

term-by-termwhen

we

w")

m'

Since "," and 6",are

If

-cos

deduce

O^x^ir

dd

6^

"

{x sin 6) we

for F^

sin r"^-+

cos-"

T{l
v) ^-"""jSin7H.x--F
6,"(1 coswia;)
m'x
,Zx
r(i)

v)

"

12r(i)
When

(^ mxy

wi^

_aQX-V{\

x''^" has

to be

replacedby logx.

,
l)

continuit}'^

19-7]
Multiply(2) by

from
integrate

and

cosmx

b,,

(3)
sinma;

)'"a,"

with

equationis inconsistent

This

{2Br(v

vf-

a^

the series for

Hence

the series
is also

least

at

one

cannot
for/i,(a;)

^ (-r-/.(m^-y

points0, tt,
at

converge

have

which

started must

we

is the theorem

to be

have

We
theorems

that

an

exceptional
point;

pointunless

"" is zero,

so

and

that a,^

shall
in

discussed

the

usual

for Schlomilch
theorems

series in which

concerningthe

"\"v"\
of
representation

series.
by trigonometrical

and
of Riemanns
type concerningseries of Bessel functions
of Bessel functions.

of

"

"

^) may

be

justgiven in ""19 6

whiclv^o
X,,,are

"

arise in the

not
not

the series

the

manner

of Riemann's

gation
investi-

of Schlomilch
investigation
various
19"62, though there are
pointsof detail*,
of Schlomilch
series,due to the fact that J,"
case

is identical with

method

which

after
investigated
series.
trigonometrical
v

by

m=l

(inwhich

of the method

give a very brief sketch


Chapter xviii, namely
now^

00

The

is convergent all the

Theorems

series
We

%b"Jm

series
vanish ; that is to say, the two Schlomilch
coefficients equal. And this
have corresponding

established

analogousto

19"7.
Dims

series

proved.

therefore

null-functions

and

is not

tt

"

proved that, if the

therefore

must
!),"
coefficients
"," and

series

to

zero.

We

with

la"7r=,

reduce

of the

(^"+2)unless

"

Now

( )'"ay.

f^ (x) must
r

then

l)- ^ao7r%

the fact that a", is

25r(;- + l)
and then

and

to tt;

then

integrate;and

and
"" +

tt

"

0.

Again,multiplyby

649

SERIES

SCHLOMILCH

linear functions

of

the method

of

m.

specialdifficultyin discussingany
to
to set them
out
in full,and they do not seem
of them ; but it is a tedious and lengthy process
The
reader
which
the use
of the space
they would require.
be of sufficient importance to justify
consult the papers
in such investigations
by
who desires to appreciatethe details necessary
may
Hoc. x. (1909), pp. 391"435;
Math.
Trans.
American
xii.
C. N. Moore,
(1911),pp. 181"206;
*

XXI.

These

points of

(1920).pp.

detail

107"156.

are

very

numerous

and

there

is

no

650
In

[CHAP. XIX

FUNCTIONS

BESSEL

OF

THEORY

by

it is easy to prove
the jfirstplace,

the method

used

" 19'5

in

that

if the series
00

00

am

Next

and

is

arrangedas

a) F{x

2a)

series in which

2x

F{x)

the mth

{x

pointo

0, and

that its limit when

has

term

is

a^-0

%"a

latter series is uniformly


convergent with respect to
the

equation

the

"^'

Jm"

defining
/(^) is convergent, and
+

write

we

value
has any positive
if the expression

proved that, when

]-"
[(a- 2m

o{s/in).

2
m

be

hm

function ;

F{x)=
It may

is defined by
with/(.t')

associated

the function

then

o{s/m),

the associated

consider

we

interval of values of x, then

throughoutany

converge

a) F{x

for

a",

in

for which

an

the series

l-x)]

factor,then the

interval

containing

"x^~''f{x).

It may also be proved that,whether


the series for f{x) converges
the condition that ","
sufficient
that
to ensure
o{\Jm) is

or

not,

[{x+

2m

a)F{x

2x

2oi)

F{x)

{x

2m

~tx)F {x

2a)]

4a

tends to
The

w'ith

zero

proofsof

a.

these theorems

dejjendon

number

siiiVm+i"

of lemmas

such

the lemma*

as

tliat

sinVm"

is

bounded

proofsin
It

function

the

proofsof the
case
special(trigonometrical)
a

leiumas

follows in the usual

now

function

of

throughoutthe

be constructed

can

in which

the lines of the

on

^.

is a null(cf." 19'54)that, when /(a;)


(0, 1),then F {x) satisfies the differential

manner

interval

equation

and

so

where

t
A

F{x)
and

are

Cf. Modern

t When

Bx--",

equationis valid

This

constants.
*

Analysis, "9"62.
0,

(x)^A

+ B'

logx.

when

"

^1.

652

throughout(0,1),then

null-function

that

B exist such

and

constants

f {x) is

that when

it is inferred

and

[CHAP. XIX

FUNCTIONS

BESSEL

OF

THEORY

F{x)

+5a;-^

\f

(Ax"

Bx-") xJ^ (\nx)dx

J 0

A,,,

"

and

0 when

v'^0.

Now, Avhen

is

large

-/.+i
(^n)

so

(\"-*), J.-1 (X,)

(X"-3),

that,if 5 ^ 0,
iz+l
7ri?X,i

?"n~

'

2''r(i;)
and

this is inconsistent

with

the

equation
bn

since

"

Hence

B is zero, and

therefore

JJHK)
(i-^,^JJ'{K)

hn
This

equationis inconsistent

with

K
unless A
We
cases

(\/n),

^.

"

is zero, since
consider

next

Dini's series

the

J^+i(\n) is not
what

assume

AXn^v+i \\i)'

equation

(\/'')

"

zero

happens when

and

then

is

1'

bn is

zero

or

zero.

negative;

in these

the forms
00

l
00

bolu0^o^)+ 2 bmJ,.(\n^)"
m

respectively.
of the two

In the second
the insertion of
consideration

initial term

an

because

the

cases

previousarguments

the firstof the two

initial term

the

to

be

cases

unaffected

are

needs

inserted

in

more

the

function is

bpX^
_

4(i.+ l)'
and

hence, when

n^l,
/"I

bn-Jj"
(X,/) X,;- I \Ax"
=

'-^^

5 " ^v+-i
+

IFiTTT)

^^

Bx-")-xJ^ {Xnx)dx

^^"| 207Tr)
+

"

by

careful
associated

19-7]

Since

{"i^n}",

hn
and

We
Dini's

infer
series

converge
of

are

finite

equal.

also
of

(0,
and

are

number

1^

that,

first
and

in

as

1), and

equal

at

all

points,

6"

limiting

by

0,
all

for

of

the

series

Dini's

interval

corresponding

series

of

case

considering

values

represent
of

two

points
then

that

cannot

if

that

so

the

functions

Bessel

of

infer

we

that

then

interval

the

{\/n)

653

SERIES

SCHLOMILCH

of

the

n.

of

Bessel

null-function

(with

functions,

throughout

the

same

(0, 1), with


coefficients

in

term

in

the
the

and

H)

exception
two

series

CHAPTER
THE

It is evident

OF

TABULATION

of Bessel

Tables

20*1.

from

XX

Functions

and

consideration

FUNCTIONS

BESSEL

associated

functions.

contained in Chaptersvii,
analysis
that a largepart of the theoryof Bessel Functions
and XV
has been conVIII
structed
numerical
expresslyfor the purpose of facilitating
computations
To the Mathematician
with the functions.
such computationsare
connected
of less interest and importance* than the construction of the theories which
but to the Physicist
numerical
results have a significancef
make them possible;
formulae

which
As

an

it has

fail to convey.

may

of
application

various

considered

desirable

been

of Tables

historical account

togetherwith
published,
will not

reader

requiredto

master

Functions

which

requirementsof the

be concerned

with

have

he realises that it

been

previously
be of the

to

seem

Physicist.

the monotony

previouschaptersand

the

Theory of Bessel Functions,


Chapter,which contains an

insert this

to

of Bessel

Chapterwhen

of this

the

portionsof

collection of those tables which

greatest value for the present


The

of the

and

technical

be read without

can

the efforts

amplif}^
arguments

to

vance
irrele-

lessly
ruth-

so

condensed.
first Tables

The

his memoir
52.
pp. 46
decimals for
"

Jq (x) and J^ (x) were

publishedby Bessel himself in


on
PlanetaryPerturbations, i?e?'/i?ierAbhandlungen,1824 [1826],
tables give the values oi Jn{x) and J^ix) to ten placesof
These
a

of

of values

range

A short Table

of

of

from

and J^ix)
Ji,{x)

four

to

0 to

3"20

placesof

by Airy,Phil. Mag. (3) xviii. (1841),p. 7; its range


with interval 0"2.
constructed
Airy| had previously
the

same

with

interval 0*01.

decimals
is from
a

was
a;

Table

of

constructed
0 to

a;

10*0

2/, {x)lx,
of

scope.

subsequentlytabulated to six placesof decimals by Lommel,


167, with a range from x--0 to .r=20'0
fur
Phys. xv. (1870),pp. 164
Zeitschrift
of J^ (^)/^^ was
with interval 0"1 ; this Table, with a Table
republished by Loumiel,
315.
Miinchener
Abhandlungen, xv. (1886), pp. 312
The

function

Ji{.v)lxwas
und

Math,

"

"

Of. Love. Proc.

Cf. Lord

magnitude
Baltimore
+

"

Camb.

Sac.

"I have

when

formulas

(2)xiv.
no
are

(1915),p.

satisfaction
intended

184.

in formulas

for definite

unless I feel their arithmetical

dynamical

or

physicalproblems."

(Cambridge, 1904),p. 76.


Phil. Soc.

places of decimals, in which


was

Math.

statement

at all events

Lectures

Trans.

London

Kelvin's

given by Schwerd,

Die

v.

(1835), p 291.

the range

is from

Table

of

and
2Ji (.r)/x

0 to the circular

measure

its square,

of 1125"

Beugungserscheimmgen (Mannheim, 1835), p.

146.

to four

or

five

(with interval 15"),

20-1]

TABULATION

OF

655

FUNCTIONS

BESSEL

of the need of Tables of t/"{x) with


In consequence
and .?" for Astronomical
constructed
n
purposes, Hansen
Ji(x) to six placesof decimals with a range from x

interval 01;

Table

to

a;

of

Jo(*')and

of

with

100

der ahsoluten Storungen


publishedin his Erinittelung
in Ellipsen
Excentricitdt und Neigung (Gotha,1843). Hansen's
von
beliehiger
Table was
it
who extended
and also by Lommelf
reprintedby Schlomilch*
to

this

fairly
largevalues

was

20.

These

tables,however,

a,ndJi(x)

twelve

to

Meissel's Table

is

given in
A

to

Table

Nagaoka,
The

value

has

666

by Gray

247

interval O'Ol.

and

Mathews,

266, and

"

'Treatise

abridgement of

an

it

697.

"

50 has been computed by


placesof decimals fovn=l, 2, 3,
Coll. of Sci. Imp. Univ. Japan, iv. (1891),p. 313.

of the

Juurnal

Table

full

of ./((mtt)
to six

...,

computed by

of ./o(40) was

Mag. (4) xiv.


0'2

infra,pp.

with

15-50

40.

(1900),p.

reprintedin

Abhand-

0
Jj (x) to twenty-one placesof decimals, from x
interval 0*1, has been constructed
by Aldis, Proc. Royal Soc.

6"0 with

Lxvi.

0 to

Jo(x)

of Jq (x) and

Table

.j; =

a;

(London, 1^95), pp.

Table

is from

was

Bessel Functions

an

placesof

its range

lungen,1888;

of

supersededby Meissel's great Table


decimals;]:,
publishedin the Berliner

are

W.

from

R. Hamilton

Phi.
ascendingseries,

the

(1857),p. 375.

Ji(x) to six places of decimals


published by Steiner,Math, und

of

been

from

.r

20'l

to

.^

41

jyaiurwiss. Berichte

with

interval

Ungarn,

aus

xi.

O'l

or

(1894),

pp. 372"373.

earliest table of functions

The

xxvi.
Smith, Messenger,

B. A.

decimals

of Neumann's

from

l"00

to

a;

with

functions

interval

second
98

(1896),pp.

placesof
^

of the

and

{x) and
from

constructed

was

101; this is

"

F'"*

0 01

kind

F"'*

^=1"0

Table

to

by
four

{x). Its
to

range is
10"2 with

interval O'l.
A

whose

extensive

table

Report,1914,

pp. 76

more

range

is from

0 to

of
"

these functions

82; this is
a;=

15"50

Table

with

is

ciation
given in the British Assoto six placesof decimals

interval 0"02;

year later a table


of F'"' {x) and F'^' {x)

ibid. 1915, p. 33, in which


the values
published,
of
decimals
for
a
given to ten places
range from

was

were

interval 0*2 and

from

6"0 to

x=

160

with

0 to

x"

(rO with

interval 05.

structed
Shortlyafter the appearance of Smith's Table, an elaborate table was conof
Heine's functions "
by Aldis,Proc. Royal Soc. Lxvi. (1900), p. 41,
should be
of
the
reader
decimals;
Ga{xf~am\ Gi(x) to twenty-one places

Zeitschriftfur

Math,

und

Fhijs.ii. (1857), pp. 1.58"165.

135.
(Leipzig,1868), pp. 127
a
misprint, the correct value of Jg(0-62)being +0-90G18..., not
X Meissel's Table contains
made
in the reprint of the Table hv Gray and Mathews.
A.n ailditioual misprint was
+ 0'90518....
also tabulated
functions were
"s These
by B. A. Smith, Fhil. Mag. (5),xlv. (1898), pp. 1*22
that of his Table of I'W [x)and i'(^)(.r).
128 ; the scope of this table is the same
as

t Studien

iiber die Bessel'schen

Functioneii

"

"

656

THEORY

The

to

placesof

seven

The

interval

001.

decimals

from

Association
matter;

in

By

means

which

to

^=

0 bo

interval

smaller
116

respectively.
"^tt Fj (a;)

6 0 with interval O'l.

a:=

publishedin
table gives the

was

130; this

"

range from
for 191o, p. 33, contains a
15"o with interval 0*5.
for

decimals

Yq (x) and

of the

have

functions

with

and

Fj (x) are

described,it is

0 to

table

a;

to

16'00 with

placesof

ten

seven

with

easy matter
integer.Such tables of

order is any

whose

to

combined

formulae

recurrence

been

now

tabulated

a;

placesof decimals
the British
appreciableadvantage over
t'iifra
; this table has an
trivial
tables make
a
Tables*, in that the auxiliary
interpolation
is
the British Association Tables interpolation impracticable.

in Table

Report
65

x=

functions

The

is from

Report, 1913, pp.

Association

British

functions

of Aldis' Table

range

[CHAP. XX

FUNCTIONS

equal to "^7rYo(x)

are

table of these functions

Another
the

functions

these

that

reminded

BESSEL

OF

the
to

an

of the tables

use

tables of

construct

J^ix)

constructed

were

after their Tables of


reprintedby Schlomilch and Lommel
Jq (x) and J^ (x).SubsequentlyLommel, Milnchener Abhandlungen, xv. (1886),
316, publisheda Table of Jn{x) to six placesof decimals, in which
pp. 315
in Table IV
0,\, 2,...,12; this Table is reprinted
2,...,20, and x
n^O,l,
731. A Table of Jn {x)of practically
the same
also
infra,pp. 730
scope was
155.
(1891),col. 154
publishedby Meissel,Astr. Nach. cxxviii.
and

Hansen

by

"

"

"

much

that he

seems

extensive

more

Table

of

publishedit.

never

computed by Meissel,but it
communicated
it to Gray and Mathews
This table gives Jn(^) to
267
279.

Jn{x) was

He

publicationin their Treatise,pp.


eighteenplacesof decimals when n=0,

for

"

1, 2,

...

60, and x^O,

1, 2, ...,24.

graphsof /" (x)were constructed,with the helpof the


214.
table,by Hague, Froc. Phys. Soc. xxix.
(1917),pp. 211
Some

last-mentioned

"

to

The

Tables
corresponding

The

British

five

Association

placesof decimals

interval O'l and

with

of functions

of the second

Report,1914, pp.83
forf

6"0 to 16"0

86

contains

,13 for

0, 1, 2
with

"

kind

not

are

Tables of

the range

a;

so

tensive.
ex-

Gn{oc)

0 to

6*0

interval 05.

Similar

Tablesf of F*"' (a;)to six placesof decimals (with the intervals


earlier part equal to 0*2) appeared in the Report for 1914,
36.
pp. 34

in the

"

Some

values

Hankel's

of

Nicholson, Proc. London


Table

A
Table

and

IV

Yn{x)

infra.This

In the

course

of

Math.
to

has been

and

Y"(a') had

Soc. (2)

xi.

been

(1913),pp.

given previously
by

113"114.

(or more) significant


figuresis contained

seven

computing

checked,

Gj (.r)was

entries

of

function

computed from

Aldis' Table

of Go

(x)and Gi (x).

Table

the last

Table
of (?"(.r\
I, a small part of the British Association
found
5 %
in about
to be unreliable
digitsin it were
of the

checked.

+ For

the

larger values of

in

the functions

are

not

tabulated

for small

values

of

x.

20-1]

TABULATION

Tables

of

...,

(interval
10) and
the

100). For this range of values of


(interval

1000

to

approximationthat

is of less value than

table in which

the values of

the intervals

and

x,

the

smaller.
considerably

are

Functions

imaginaryargument have been tabulated


Association Reports. The Report for 1896, pp. 99
tained
149, conof /" (x) to nine placesof decimals,its range being from x=0

of the firstkind

in the British

with interval 0001.

5"100

in
publishedpreviously
of this

with

"

Table

to

100

asymptoticexpansion("7 '51)givesso rapidan

Table

657

FUNCTIONS

"

are

BESSEL

to eightsignificant
\ogiQ['\/(^'jrx)
figuresai^e given
.\H^^^i
(x)\']
97. The values assigned
British Association Report.,
1907, pp. 94
0, |, 1, 1|,
6|, and the range of values of x is from ;"= 10 to 100

in the
to

OF

the

of /j(x)of the

Table

Report for 1893, pp.

(with interval O'Ol) was

givenby Gray

229

and

279;

"

scope had been

same

abridgement

an

in their Treatise,

Mathews

pp. 282"284.
Tables

/" (x) and /j (x)to twenty-one placesof decimals have been constructed
by Aldis,Proc. Royal Soc. lxiv. (1899),p. 218. The range of these
is ic= 0 to x
Q'0 with interval 0*1; Aldis also gave (ibid.p. 221) the

Tables

of

values of /"(x) and


Extensive

/j (x) for x=7,

8, 9, 10, 11.

Io{x)and 1^ (x) have been publishedby


And'mg,Sechsstellige
Tafelnder Bessel'schen Funktionen imagindren Arguments
and logio
from x=0
/o (^-c)
{^i(ic)/^}
1911). These tables givelog,o
(Leipzig,
to

with

10"00

x=

with

tables connected

interval O'Ol.

The}-also givethe

values of the functions

^x
[V*"U (")}and logio
x/(2?r^)e-* /o (."),V(27ra;)e-^/j {x), logio
"

for values

of

1),^
(interval

from

200

a;

to

II

Table

lO'O

1000

to

^;

(intervalO'l),.t

with interval O'l, and


5"0 to

The
the

of the

helpof
means

x=GO

B. F.

and

of

of

with

give K^ia;)and

from

thirteen

recurrence
n

interval 0'2.

to

6'0

significant
figures

11
...,

These

Report,1889,
"

computed with

e~-^/i
(x).

has been tabulated


formulae,/"(.")

0, 1, 2,

by

interval O'l.

in their Treatise,
pp. 285
of these Tables has been
figures)
w.

seven

These

and e^K^ (x) in Table II infra were


e-^Ko{x)

British Association
Mathews

=200

constructed

were
type K^ (*')

Aldis' Table, like the values of e~^Io(x) and

figuresfor
to

also to between

12*0 with

x=

values of

By

a-

"

earliest tables of functions

x=

to

values oix.
(interval
10),and for various larger

Aldis,Proc. Royal Soc. lxiv.


(1899), pp. 219"221.
/Tj(x) to twenty-one placesof decimals for values
from

50

and e^^Ii{x)\
713, givesthe values of e~^/o(;c)
infra,
pp. 698
in Aldis' Table.
been computed,for the most
part,by interpolation

these have
The

50"0

I^ {x)]

to

twelve

ficant
signi-

the range of values of x from x


0
Tables of In(x) were
publishedin the

over

pp. 29
288.

"

An

32, and

reprintedby Gray and


abridgement(to five significant

givenby Isherwood,

who

added

to

them
42

658
Tables

Kn{x)

of

of values of

five

to

from

the Mem.

publishedin

for
figures
significant
6*0 with

0, 1,2,

and

Proc, Manchester

10
.

the range

over

Isherwood's

interval 0"2.

Q to

[CHAP. XX

FUNCTIONS

BESSEL

OF

THEORY

Tables

Lit. and Phil. Soc, 1903

were

1904,

"

19.

no.

Tables
Table

e~^In(x) and Kn{x)

of

IV

Meissel, who

of Bessel functions

of

decimals

large order

given in

are

constructed

were

figuresfor
significant

J2n(") to twelve

calculated

has

placesof

seven

736"739.

infra,
pp.

earliest Tables

The

to

?i

by
10,

(1892),col. 284; Meissel also calculated /n(lOOO)


981, ibid, cxxviii.
(1891),
significant
figuresfor u=1000, 999,

cxxix.
11, ...,2\,Astr.Nach.

to

seven

...,

The

col. 154"155.
six

to
n
n

bO

placesof

400

decimals

Gn{n),Gn-,{n)
/"(n),Jn-i{n),FW(w), F'"-i)(?i),

for values

(interval5), ?i

w=100

to

values of

(interval20), n

400

to

of
100

l *to

7?

1),
(interval

50

(interval10),??
1000
to
50), n
(interval

to

1000

from

200

of

100) and for various largervalues


(interval
96.
Association Report,1916, pp. 93

given in

are

n,

200

the

to

2000

British

"

Tables
in Table

of

VI

The

J,i(w),
Jn'{n),Yni''^),
J^/(^0to

infra,pp.

functions

tabulated

ber

brief Table

V2

(x),bei (x),ker (x) and

to

J^{x

to

a;

6*0

nine

sji)to

interval

with

Tables

of ber

1, 2,3,

...

These

0"1.

placesof decimals

publishedin the
reprintedby Gray and Mathews

(1900), pp.

LXVI.

interval 0'2

a;

the

(x) have

been

extensively
currents.
theoryof alternating

twenty-one placesof decimals

by Aldis, Proc. Royal Soc.


x

importancein

kei

(x) and bei (x),computed by Maclean, was published


Phys. Papers,iir. (1890),p. 493. Tables of J^{x \/i)

and

Ji (^ VO

"

given

are

of ber

Math,

by Kelvin,
and

decimals

746"747.

of their

account

on

placesof

seven

42

have

been

extensions

are

for the

range
British Association

constructed

43; their range

"

from

the

of

0 to

a;

Report,1898,

is from
Table

of

with

60

p. 228, and

in their Treatise,p. 281.

ker {x) and kei {x) to four significant


for
{x),bei (a;),
figures
Phil.
publishedby Savidge,
Mag. (6) xix. (1910),

30, have been

p. 53.
The

functions

placesof

hex

ber' (a;)and
{x),bei (a;),

decimals, from

a;

0 to

Association

x^

10*0

bei' (a;)
are

with

68 ; and a Table
Report, 1912, pp. 57
and kei' {x)of the same
scope (except that onlysix
were

tabulated

interval

"

of ker

in the

01

nine

to

British

{x),kei (x),ker' {x)

or
seven
significant
figures
given)appeared in the Report for 1915, pp. 36 88. Tables of squares
0 to a;=
fi:om a;
10"0
productsof the functions to six significant
figures
interval 02 were
121.
givenin the Report for 1916, pp. 118
"

and

with

"

The

functions

Lommel, Munchener
x=l,

2,

have
t7"(M+j)(a;)
Abh.

xv.

50, and (in the

case

been

tabulated

(1886), pp. 644

"

of functions of

to

six

647, for

placesof
n

decimals

by

0, 1, 2, ...,6 with

order)w
positive

7, 8,

14
.

660

THEORY

Bourget'sTables
overlooked
by the
Math.

Wiss. and

articles

of the

writers

Bessel

on

their

Functions

in

authorship has been


der
the Encycloplidie

Sci. Math.

Encydopedie des

the

[CHAP. XX

FUNCTIONS

frequentlythat

reprintedso

been

have

BESSEL

OF

Ji{x) and Jo{x) were


given to six placesof decimals
fiirMath. und. Phys. xv. (1870),p. 167 and Munchener
by Lommel, Zeitschrift
Ahhandlungen, xv. (18S6),p. 315.
The

first five

The

first ten

oi

zeros

of

zeros

computed

Jo{x) were

to

placesof

ten

decimals

by

MeisseljBerliner Ahhandlungen,1888.

of decimals
pp. 153

by

Willson

Peirce,Bulletin American

and

the

155; they also gave

"

eightand

to

zeros

of Jo{x)were
(and their logarithms)

firstfifty
zeros

The

seven

of J, {x) and

the

sixteen

to

this Table
Tables

of the

of roots

for k

"

few of the

1'2,1'5 and

are

Phys.Liv. (1907),
0, while n is given the

Rayleigh,Proc. London Math. Soc. x. (1878),pp.


(1899),pp. 363"364], has calculated that

has

maximum

when

of SxJ^

zeros

the values

"

(1921),

J^,{x)for small

[Scientific
Papers,

x^)xl, {x)lh{oo)

0'4858.

der Math,

Airey,Archiv
the first ten

x'^

of

zero

"

(1

mentions

Airey,Phil. Mag. (6) XLI.

205, has computed the value of the smallest


pp. 200
fractional values of v by Euler's method.

I.

und

of fractional order

function, while

of each

zeros

these

in their Treatise,p. 280.

of functions

in his Tables

places

(1897),

values of Jq {x) were


corresponding
by Meissel*, Kiel Programm, 1890;

fur Math,
by Kalahne, Zeitschrift

86 ; the values taken


pp. 55
values 0, h, 1, f, 2, |.

Dinnik

III.

at
log jJi(j;)|
respectively.

J,,(kx) Yn (x)

constructed

Soc.

equation

Jn {X) Yn (kx)
have been

Math.

ten

Ji{x) and

placesof decimals
is reprinted
by Gray and Mathews

computed

of

of

placesof decimals

first fifty
zeros

The

values

given to

und

Phys. (3) xx.


2J^ (x) and

(x)

"

(1913),p. 291, has computed


2xJo (x) J^ (") to four places

of

"

of decimals.
his

In
Lommel

has

given here.
from

Diffraction,MiXnchener

on

publishedtables

these tables

but

of

memoirs

to

so

are

His

Table

that

numerous

of Fresnel's

with

50*0

x=

connected

with
a

Ahhandlangen, xv.

his functions

detailed account

Jahrbuch

of Meissel's
Table

VII,

ilber die

table,the
p. 748.

and

minima

Fortschritte
zeros

of

of these

der Math.

were
Ji (.r)

will not

integrals
(p.648) to six placesof
0'5 (with auxiliary
tables for

interval

must, however, be mentioned, and with


interpolation)

firstsixteen maxima

of them

1890,

(1886),
variables,

of two

be

decimals

purposes
it his Table of the

integrals.
p. 521.

recomputed

In consequence

(to ten places of

of the

inaccessibility

decimals)for insertion

in

20-2]

TABULATION

Lommel's

for Fresnel's

form

BESSEL

OF

661

FUNCTIONS

was
integrals

different form

Chemie, (3) xvii.

Definingthe

tabulated

was

earlier

by Lindstedt, Ann.

Ph"/sik unci

der

(1882),p. 725.
functions M

{x) and

/oo

f^dt

(x)cos

x^

sin t-dt

{x) sin

x^ +

COS

the

{x) by
"

equations

(x)sin x-,

{x)cos

X
00

and

writingA'

of decimals
The

Lindstedt
{(y+ 2)'""P.

from

function

9 0 with

3/

0 to t/

{x) defined

tabulated

Chemie, (3) XVli.


X

4-0 to 7-0

to

four

and

the limits

are

by Struve, Ann.

decimals

from

from
-

0 to 4*0

7-0 to ll'O

consecutive

zeros

(up to

"

Table

20*2.

of Struve's functions

(x) and H^ {x) which

of H"

of the
Description

be included

(I)

in this book

(II)

to

seem

been

constructed

Tables contained

in this book.

on

the

led to the

magnitude and character


decisions :
following

space did not usuallyadmit


decimals in the tables.

That

have

of J^ (x),has
forty-ninth)
Berichte cuts Ungarn XI.
problem of Diffraction by

pp. 666"697.

is

That

places^

01), from
(interval

0-4).
(interval

given on

Preliminary considerations
to

und

the

publishedby Steiner, Math, und Naturwiss.


367 ; this integral
in the
(1894),pp. 366
occurs
Circular Aperture.
a
Tables

Physik

been

No

der

integi'al
j

in which

places

t,"

(1882),pp. 1008"1016,

table of values of the

N {x) to six

and

interval 0"1.

placesof

(interval
02)

M{x)

as

IT

has been

tabulated

x-,

the tables should be

In
interpolations.

so

of the inclusion of

constructed

it was
particular,

as

to

more

minimise

decided

that

before the

of the tables

than

the

seven

difficulty

table with

a
a
making
interval
with
t
able
to
an
as
002), together
(such
auxiliary
moderatelylarge
would
be more
useful than a table with a smaller
facilitate interpolation,
interval (such as O'Ol),
occupyingthe same
space as the first table and its
in which interpolation
was
impracticable.
auxiliary,

of

662

(III) That
in the last

tenths of

of

second

of

ensure

number

be

of

carried
of

cases

be trivial*.
entered

are

it is not

to

ten

inaccuracy

This

in Tables

does

not

III.

I and

sufficient to tabulate

to

the differences per minute


of arc in a sevenbe as largeas 00002909
the other hand,
; on

because

arc,

sines may
of a second
hundredth

should

publishedresults should

of natural

figuretable

that the

tables of angleswhich
auxiliary
to obtain seven-figure
accuracy,

In order

error

in order to

figureof the

the

apply to

an

computing tables,calculations

in

decimals

placesof

[CHAP. XX

FUNCTIONS

BESSEL

OF

THEORY

'

does not

affect the value

sine

of the

by

considered
seven-figure
accuracy, it was
adequate to compute to nine placesof decimals the sines (or cosines)of the
anglestabulated and then to compute the anglesfrom Gifford's Natural Sines
tables with an intervalf
of 1".
(Manchester,1914) ; these are eight-figure
for

Hence,

0'00000005.

than

more

be in
frequently
anglestabulated may consequently
exceeds
the error
never
digit,but, in all probability,

The
last

hundredth
We

of

now

I consists

taken

the tables in detail.

are

Tables

of Jq (^).J^o(^^),
J\ {^) and
The

interval of 0"02.

values

while
Table;):,

Meissel's

from

(i.e.a

arc).

primarilyof

with

16"00

to

15"50

of

proceedto describe

Table
=

second

unit

the

to

as

error

of

the

Jo{x) and

values of

in Aldis' Table
interpolation

F, {x) from

Ji{x)

of (ro{x)and

to

up

Y^ix)and

Yi{x)

G^ {x) and

computed partlyby
partlyfrom the asymptoticexpansionsof J^ {x)+ Yq-(x) and Ji^(x)+ Y^-(x)
given in " 7*51.
were

tables " give the


auxiliary

The
n

and

In

n=l.

these tables

(exceptfor quite small


with

little trouble

but

the second
The

small

values.

The

tables

differenced

were
were

number

in the last

errors

tables with

No

interval

and

constructing
advantage
+

in the
of any
of

greater number

the

before

interval

be computed
consequently
arg ^"'^'(x) can
less than 16, with the exceptionof quite

than

more

does

have

of decimal

my

not

been

use

thanks

given order

he remarked

resemble

about

to the

one

published; the
have

use

place ;

accuracy
of

purposes

so

was

it is hoped that the


thousand

tables with

obtained

remarks

the
would

that the

entries.
a

larger

labour
be of

of
no

interpolation.

Preussische

Akademie

der Wissenschaften

zu

of this Table.

that it was

circular functions

grew

out

of

frequentlynot
multipliedby

conversation

realised
a

how

damping

with

immediately followingof

course

presuppose

that

is 0

Professor

closelyBessel
factor

decay is slow.
IIThe

found

be

it

in every
of any

then

so

greatlyincreased

very

the increased

constructing the auxiliarytables

of which

exceed

placeswould

angles, and

cordial

to make

unit in the tenth

using the auxiliarytables for

express

10

0'0000009.

removing the last three figures,and

by

error

auxiliarytables of

here

idea of

course

figureretained

smaller

permittingme

for

" The

rarelyin

to anyone

I must

Berlin

of

is about

The

results
ten-figure
of

value

exceed

not

correspondingvalues of Jn{^) and Yn{x) can


immediatelyby the use of seven-figure
logarithmtables.

calculated
*

for any

and arg^""^'(a;)
for
l//,i*"(^)|
the first differences are sufficiently
steady
to be effected
x) to enable interpolations

(^)Ido
|iTo'^'

||of |^w"' (x)\and

values

of

part of the reader; thus, when

the

on

differences of

the reader

by

of

values

values

or

1.

in which

Love,

functions
the

rate

20-2]

TABULATION

The

the various functions

relation between

663

FUNCTIONS

tabulated

{x) as
briefly
by regarding|Hn^^^{x)\and arg Hn^^''
of
Cartesian
coordinates
then
the
pointin a plane;

most

of

BESSEL

OF

and

Yn (")" Thus,

Table
These

the entry for

from

be expressed
may
the polarcoordinates
this

412"

28' 40"-60,

01716508

0-2818259

cos

0-2235215

0-2818259

sin 412' 28' 40"-60.

I also contains

functions

functions

the values of Struve's

included

are

pointare ./"{x)

800,

I, instead

in Table

Hq{x)

being

of

and

Hi(.x').

contained

in

by " 10-41 (4), the difference


separate Table, to facilitate interpolation;
function, and it varies sufficiently
F,i(.r)is a positivemonotonic
H" (^)
"

easy when

to be
for interpolation
steadily
Tables

The

of Struve's

functions

were

is not

small.

computed by calculatingthe

values

l, 2, 3,...,and

then

Hq (^)
calculating
of

Hq (x) directlyfrom the ascending series when x


for these values of ,r from the difl'erential equation "510-4 (10) and
Ho" (^),Ho'" (""),...

and

equationsobtained

by

the

it.
differentiating

and Hq' (a-)


l)yTaylor's
adequate to calculate Ho(.^^)
unit
is then
fiftieths
of
the
oi
to
x.
0*5,0'6,0"7,
Interpolation
at first
the
it
effected by usingTaylor'stheorem
in
This process, though
seems
manner.
same
an
extremelyrapid one when a machine*
sightto be complicatedand lengthy,is,in reality,
is used.
It is very much
effective than
the use
of asymptotic expansions or the
more
British
Association
in
the
Beport, 1913, p. 116. As an example of the
process suggested
the
be
stated
that
the
and e~^/i (x) in Table II
values of e~^/o(.r)
rapidityof
process, it may
took less than a fortnight
to compute ; of course
the time taken
this tabulation
was
over
for interpolation.
appreciablyshortened by the use of Aldis' Table as a framework
A

differential coefficients

few

theorem

are

for the values

...

Table

II consists of Tables
of e^ is

inserted,in

and

the

Ii{x), Kq{x)
functions'!"
I(,{x),

from

Table

of e~* /" (x),e~^ /j (x),e* Kq

with

0 to 16-00

the reader

case

and

K^{x);

Trews.

Table
Camb.

of e^

the

requirethe

functions

are

e^

K^ (.r),

values of
tabulated

interval 002.

is easy in
Interpolation
by differencing
throughoutthe greater part of the range.
The

should

(x),and

constructed

was

Phil. Soc.

xiii.

with

(1883),pp.

the

the
145

"

of the first four functions

case

Newman's

help of

Unlike

241.

the

Table

of e~^,

other

Tables

and care
figures^,
given to eightsignificant
in every entry. Interpohas been taken that the last digitgiven is accurate
this
effected
Table is,of course,
lationjn
by multiplyingor dividingentries
be
not
can
exceeding 0-01 ; such exponentials
by exponentialsof numbers
calculated without difficulty.
in this book, the

10

The
9

machine

recording

t These
i Nine

on

to 18

functions

figuresare

of e^ is

Table

which

the calculations

were

carried

out

is

Marchant

Calculating Machine,

figures.
were

not

given

in

tabulated

parts

because

of the Table

tables of them
to avoid

are

unsuited

for

spoilingits appearance.

interpolation,

664

THEORY

Newman's
number

Table
gives e~^
significant
figuresin

of

the Table

of

the

value

constructed

e^ was

from

Newman's

Table

III

in

of Tables

consists

short

the tenth
third

of

are

is small

and

calculatingreciprocals
;

the values

by using

the first and

places of decimals,

of

latter part of the Table


the process

by

Table

[CHAP.

FUNCTIONS

large number

the

given by Newman
employed because

significant
figuresin

of the

to

of e"!^"^

wei'e
l^onentials

BESSEL

OF

of e^^ and

table

of

e^^

the

rest

actual
half of

was

structed
con-

and

given by Glaisher*,
These

O'l.

all three and

significant
figuresin

the

less than

interval

with

e~*^

but

XX

ex-

the eleventh

zero.

of

J'i(^),Yx{x),\H^^^{x)\and

|argifj'"(a;)|

effected in the manner


are
interpolations
Tables
These
are
alreadyexplained. A Table of e* K^ {oc)is also included.
of importance in dealing with approximations to Bessel functions
of large
order ("8"43),and also in the theory of Airy'sintegral.
same

The

reader

I, and

Table

as

scope

values

easilycompute

can

of J_i

this table

{x) from

by

means

of the formula

/_j {x)

{x)Icos {60"+
I^j"i)

Table

IV

values of

givesthe

and

The

n.

values

of Jn

e~^In (x) and Kn (x) for various


(^),Yn {oc),

values

of Jn

{x) are

corrections,but the remainder

some

values

of

Kn{x) taken

in part
Table

by

Table

from

is Lommel's

Isherwood's

Table

taken

of Table
Table

of Aldis' Tables

means

modifications

some

Hf^ {x)].

arg

from

Lommel's

IV, with

f, is

exceptionof
they have been

of orders

of functions

f, with

the

new;

-f-of J"(n+h)i^)^^^

Table

and

zero

Fresnel's

some
structed
con-

unity.

with
integrals

corrections.

and

VI

gives the values of Jn(n), Yn(n), Jn (n),Yn (n) and


50.
in
n^Yn(n), n^Jn(n), n^Yn (n) for n
l, 2, 3,
Interpolation
of the last four of the eight functions is easy.
=

n^Jn{n),
the tables

...

Table

VII

givesthe

first

of n;

part of this Table

Forty

zeros

of various

Table

VIII

fortyzeros

is taken

of Jn

(^) and

F" (.r)for various values

the

Tables

of Willson

from

cylinderfunctions

givesthe

values

^r

of order

a;

0 to 50 with

minima
used

of the
to

i'

Trans.

Camb.

I must

here

Miinchen,

integrals.The
the

compute

for which

to

the

interval

also

are

Peirce"f".

given.

of

hrYo(t)dt,

Jo(t)dt,

^0

-'O

from

one-third

and

1, togetherwith
former

coefficients

(c"

table

the first sixteen

of maxima

" 18'12) in certain

and

maxima

minima

and
can

Fourier-Bessel

be

series

0,
Phil

Soc.

express

my

Manchester

xiii.

digits in the entries

of

angles in

in the

to

245.

thanks

Literary and

Society for permitting me

trigonometricaltables

(1883),p.

cordial

make

radian

to

the

Bayerische Akademie

Philosophical Society,and
use

of

these
has

measure

greater part of Table

V.

Tables.
made

The
it

der
to

the

Wissenschaften
American

non-existence

of

impracticable to

check

zu

matical
Mathe-

adequate
the

last

TABLES

OF

BESSEL

FUNCTIONS

666

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

668

TABLES

Table

OF

I.

BESSEL

Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

669

670

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

671

672

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

W.

B. F.

OF

BESSEL

I. Functions

FUNCTIONS

of order

673

unity

43

674

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

676

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

677

678

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

679

680

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

681

682

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

684

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

685

686

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

687

688

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

689

690

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

692

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

OF

TABLES

Table

13 02
13 04
13 06
13 08
13 10
13'

Y,W

/iW

13' 14
13 10
13' 18
13' 20
13
13 ^4
13

26

13

2^

2112796

2213665

o-o6i784i

2123920

0-0574892

2134183
2143582

2211959
2210257

13 30
13 3^
13 34
13 3t)

?"^

0-0531781
0-0488525

2152115

of order

i":mi

0-0660609

12

I. Functions

0-0270667

2198455

0-0226902

2185034

2196784

0-0183113

2187466

21

0-0139317

2189025
2189712
2189527

0-0051775

0-0008063
0-0035587

0-0079157

2188473

2186550
2183761
2180108

2200129

951 17
2193454
2191795
2190139

4217547

4221096

13-36
13-38

634"2i'46'^85

4225507

13-40

635" 30' 23'-6o


636^39' o'^38
637" 47' 37-19

^'

13-42
13-44

638"56'i4'-o2
640" 4' 50 -87

4230776
4236901
4243876
4251699
4260365

2177029
2175407

2172174

0-046491
0-0506758
0-0548341
0-0589646

2120188

2170562
2168954
2167350
2165750

"62

0-0791428

13 72
13 74
13
13
13 So

0-0830728

1989768

0-0869640
o
0908150
0-0946243
0-0983905

13 "82
13 "84
13 "86
13 "88/
13 "90

0-1021121

-%

13 92
13 94
13 96
13 98
14 00

-f
+
+

13-52

13-54

429I37I

13-56

645" 47' 55

-52

4303358
43I6I6I

13-58
13-60

2164153
2162559

646" 56' 32 -52


648" 5' 9-55

2160969

649 13 46'-6o
650" 22' 23^67
651" 31' o'.77

4329775
4344I9I
4359404

4375406

13-62
13-64
13-66
13-68

4392190

13-70

652" 39' 37 -89

4409748

4428071

13-72
13-74

4447152

4466981
4487550

2159382
2157799

2156220
2154644

2148374
2146815
2145260
2143708

658" 22' 43' "86


659" 31' 21' 12
660" 39' 58'"41
661" 48' 35'"72

2142159

662"

2140614

664" 5' 50'


665" 14-27'
666"
23' 5' 18
667" 31' 42' 60
668" 40' 2o'-'o4

129955

165249

1797510

0-1200029

0-1234282
0-1267995
0-I30II56

0-1333752

4269869

4280206

"

o-

1772613
1747048

1720824
1693954

1666448

13-50

2149936

13-48

1911585

-+

0-1057877
0-1094160

13-46

2153071
2151502

o-

13-3^
13-34

1932249

1971349
1952173

13-30

641" 13' 27 -75


642" 22' 4 "66
643" 30' 41 '-59
644"39'i8'^54

653"48'i5'.04
654"56'52'-2i
656" 5' 29-41
657" 14' 6'-62

1890191
1868077
1845252
1S21726

13-26
13-28

2130625

13-22
13-24

0-0422817

70

42I70IO
4214472

13 5^
13 54
13' 56
13' 58
13' 60

68

632" 4' 33 -41


633 13 I0''I2

2140229

2055711
2040400
2024302
2007421

4- O

4214062

0-0380493

0-0711725
0-0751755

624" 4'i7'-2i
625" 12' 53 -73

4213044

4-

13-20

13 50

13-18

4220422

2173788

66

4224709

0-0295218

64

0-0252301

0-0630655
0-0671353

629"47'2o'^o8
630" 55' 56-73

13
13
13
13
13

13-16

2188487
2186839
2185195
2183555
2181918

4229868

2178655

2096838
2083936

2180285

13-14

42 1 20 1 4
4212094

2170223

2108924

13-12

4235900

4212807

2163997
2156920
2148996

4242805

0-0209219

13-10

0-0337954

13-08

13 4^
13 44
13
13

13-06

4259222
4250579

13-02
13-04

626"2i'3o'^28

13 40

48

627" 30' 6-85


628" 38' 43-^45

0-0122630
0-0165990

13

25 -99

4' 2'-30

620"38'27'-8o
621" 47' 4-24
622" 55
4o'^7i

2181729

0-0095532

614" 55'

4279106
4268732

2172499
2177551

4290341

0-0314391

6i8"2i'i4'-99
6i9"29'5i'^38

2201807

0-0401645
0-0358056

6i2"38'i3'^5
613" 46' 49 -71

2205176

HiW

{x)

617" i2'38'-63

2203490

2206866

2166575

arg

616"

2159780

unity

2208559

0-0445140

693

FUNCTIONS

BESSEL

2139072
2137533

2135998
2134466

57' 13' 05

13-76
13-78
13-80

4508850
4530871
4553603
4577036
4601 160

13-82
13-84
13-86
13-88

4625965
4651439

13-92

4677571

4704350

13-96
13-98

4731766

14-00

13-90

13-94

694

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

695

696

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

zero

TABLES

Table

OF

BESSEL

I. Functions

FUNCTIONS

of order

unity

697

698

TABLES

Table

II.

OF

Functions

BESSEL

of

FUNCTIONS

imaginary argnment,

and

e^

700

TABLES

Table

II.

OF

Functions

BESSEL

of

FUNCTIONS

imaginary argument,

and

e^

TABLES

Table

II.

OF

Functions

BESSEL

of

701

FUNCTIONS

imaginary argument,

and

e*

702

TABLES

Table

II.

OF

Functions

BESSEL

of

FUNCTIONS

imaginary argument,

and

e^

TABLES

Table

II.

OF

Functions

BESSEL

of

703

FUNCTIONS

imaginary argument,

and

e^

704

TABLES

Table

II.

OF

Functions

BESSEL

of

FUNCTIONS

imaginary argument,

and

e^

TABLES

Table

W.

B. F.

OF

II. Functions

BESSEL

of

705

FUNCTIONS

imaginary argument,

and

e^

45

706

TABLES

Table

II.

OF

Functions

BESSEL

of

FUNCTIONS

imaginary argument,

and

e'"

708

TABLES

Table

II.

OF

Functions

BESSEL

of

FUNCTIONS

imaginary argument,

and

e^

OF

TABLES

Table

Functions

II.

I I -02

11-04

1 1

-06

1 1

-08

I 2 i 0 1

I2I2789

I 21

II-I2

IT-I4

II-I6

ii-i8

o
o

11-22

11-24

11-26

11-28

11-30
11-32
11-34

11-36
11-38

1
o

I2I5039
I2I39I2

II-IO

II-20

159603
II5863I
I 157662
1 1 56694

69
o
o

I20I724

II47I52

1200634

1 1

199547

198463
197382
196303
195228

148096
462

1 1

II45272
144335
II4340I

11-40

11-42
11-44

II94I56
1 193086

11-46

192020

11-48

190956
189895

11-50

II406IO
139685
1 138762
II3784I
II36922

11-52
11-54

1 1

3(^005

1 1

35090

11-56
11-58

II88837
II87782
II86729
II85680

11-60

II84633

II34I78
II33268
II32360

11-62

II83589

11-64

II82548

1 1

11-66

1 1

29649

11-68

II8I509
1 180473

11-70

1 1

II28750
II27852

11-72
11-74

11-76
11-78

II784IO
II77382
1 176357

11^5173

II75335
II743I5

o
o

1 1 24284
11-23398

-80

o
o

12

II'

14

70262-956

II

16

0-3705122

0-3701886
0-3698659
0-3695440
0-3692229
0-3689027

I I

1 1

II

I2028I7

79440

11-82

11-84
11-86^

II73298
II72284

11-88

11-90

11-92

11-94

11-96

11-98

12-00

II42468
II4I538

o
o

II3I454
3055

II225I3

II2I630

II7I272

II70263

II

I987I

169256

1 1

18995

168252
II6725I
1 166252
1 165256
II64262

120750

III8I20

III7248
I I 16378

III

5509

1 1

14643

661

71682-362

II

18

73130-442

II

20

0-3863468
0-3859818
0-3856178
0-3852548
0-3848929

74607-775
76114-952
77652-576
79221-262
80821-638

II

0-3685833
0-3682648
0-3679470
0-3676301
0-3673140

0-3S45320
0-384I72I
0-3838132
0-3834553
0-3830984

0-3669987
0-3666843
0-3663706
0-3660578

0-3827425
0-3823875
0-3820336
0-3816806
0-3813286

91126-142
92967-012
94845-070
96761-068

0-3654344
0-3651240
0-3648143
0-3645054
0-3641973

0-3809775
0-3806275
0-3802783
0-3799302

100709-962
102744-438
104820-013
106937-518

0-3795830

109097-799

0-3638900

0-3626684

0-3792367
0-3788914
0-3785470
0-3782035
0-3778610

0-3623650
0-3620623
0-3617603
0-36I459I
0-3611587

0-3771787
0-3768389
0-3765001
0-3761621

o-3'^57457

II26957
1 1 26064

04
06

67507-906
68871-656

149990
149042

1 1

0-3881873
0-3878171
0-3874480
0-3870799
0-3867128

02

10

1208326
I2072I8
I206II3

n
II

08

0-3718151
0-37I488I
0-37II6I9
0-3708367

61083-680
62317-652
63576-552
64860-883

II

II54767
II53807
II52849
1 15 1894

1
20501
1203912

II

e*

71 -160

I2I055I
1209437

0-3900543

0-3896788
0-3893043
0-3889309
0-3885586

II55730

5094

and

K^{x)

0-3734632

0-3724717
0-3721430

e""

0-3731319

0-3728014

709

FUNCTIONS

imaginary argument,

1669

of

K,

AW

-"/oW

BESSEL

0-3635834
0-3632777
0-3629727

0-3775194

II
II

24
26

II

28

II

30

82454-343
84120-031
85819-368
87553-035

II

32
34

II

36
38

89321-723

II

40

II

42

1 1

44

98715-771

II
II

II
II

48

II

50

II

52

II

54

II
II

56
58

II

60

III30I-72I

II

"62

II3550-I65
115844-030
118184-235

II

"64

II

"66

II

"68

120571-715

II

70

123007-425
125492-340

II

II

74

128027-453
130613-780

II

7^

II

78

133252-353

II

80

135944-229

II

82

138690-485
I4I492-2I8

II

0-3608590
0-3605600
0-3602618
0-3599643
0-3596676

0-3758251
0-3754890
0-3748194
0-3744859

144350-551

II

88

147266-625

II

90

0-3593716
0-3590763
0-3587818
0-3584880
0-3581949

0-3741533

150241-608
153276-690
156373-085

II

92

1 1

94

II

96

159532-031

1 1

98

162754-791

12

00

0-3751537

0-3738216
0-3734908
0-3731608

0-3728318

II

86

710

TABLES

Table

II.

OF

Functions

BESSEL

of

FUNCTIONS

imaginary argument,

and

e^

TABLES

Table

II.

OF

Functions

BESSEL

of

711

FUNCTIONS

imaginary argument,

and

e^

712

TABLES

Table

II.

OF

Functions

BESSEL

of

FUNCTIONS

imaginary argument,

and

e^

TABLES

Table

II.

OF

Functions

BESSEL

of

713

FUNCTIONS

imaginary argument,

and

e*

714

TABLES

Table

To

OF

BESSEL

III. Functions

compute functions of order

FUNCTIONS

of order

1/3,increase

one-third

the

phase by

60".

716

TABLES

Table

To

compute

III.

OF

BESSEL

Functions

functions of order

FUNCTIONS

of order

1/3,increase

one-third

the

phase by

60^.

TABLES

Table

III.

OF

BESSEL

Functions

(4-00)=0-3981444
'/_]/3
y_i/3(4-00)=0-3981444

COS

sin

717

FUNCTIONS

of order

one-third

213" 12' 56'-65


213" 12' 56'-65
=

"0-3330932.
-o-2i8ioo8.

718

TABLES

Table

To

compute

III.

functions

OF

BESSEL

of order

Functions

of order

FUNCTIONS

1/3,increase

one-third

the

phase by

60".

TABLES

Table

III.

OF

BESSEL

Functions

(6-00)=0-3254331
J'_i/3

xcos

y_";(6-00)=0-3254331

xsin

FUNCTIONS

of order

one-third

328" 7' I2''96= +0-2763443.


328" 7' 12-^96= -0-1718736.

719

720

TABLES

Table

To

compute

III.

functions

OF

BESSEL

Functions

of order

FUNCTIONS

of order

1/3,increase

one-third

the

phase by

60".

TABLES

Table

-Hi

OF

BESSEL

III. Functions

(8-00)=0-2819433

7_i;3(8-00)=0-2819453
W.

B. F.

COS

xsin

721

FUNCTIONS

of order

one-third

442" 52'2i''64 +0-0349823.


442" 52'2i'-64=+0-2797667.
=

46

722

TABLES

Table

To

compute

III.

OF

BESSEL

FUNCTIONS

of order

Functions

functions of order

one-third

1/3,increase

the

phase by

60^.

724

OF

TABLES

Table

To

III.

FUNCTIONS

BESSEL

of order

Functions

corajjutefunctions

of order

one-

1/3,increase

the

third

phase by

60".

TABLES

Table

OF

III.

BESSEL

Functions

"^_i/3
(12-00)=0-2302745
r_i,3(12-00)=0-2302745

COS

xsin

FUNCTIONS

of order

one-third

672" I3'8'-I7=+0-1547365.
672" I3'8'-I7= -0-1705373.

725

TABLES

726

Table

To

compute

III.

OF

FUNCTIONS

BESSEL

of order

Functions

functions of order

one-third

1/3,increase

the

phase by 60".

TABLES

Table

III.

OF

BESSEL

Functions

J_l,3(14-00)=0-2132061
r_j.3(14-00)=0-2132061

COS

xsin

727

FUNCTIONS

of order

one-third

786" 51'26-/78 +0-0837943.


786"51'26-/78 +0-1960494.
=

TABLES

728

Table

To

compute

III.

OF

BESSEL

Functions

functions of order

FUNCTIONS

of order

1/3,increase

one-third

the

phase by

60"

TABLES

Table

III.

OF

BESSEL

Functions

(16-00) =0-1994442
"J-113
y_j,3(i6-oo)=0-1994442

xcos

xsin

729

FUNCTIONS

of order

one-third

901" 29' 3'-42 -0-1993773.


901" 29' 3''42 -0-0051662,
=

730

TABLES

OF

Table

BESSEL

IV.

Values

FUNCTIONS

of

J" (.r)

732

OF

TABLES

Table

IV.

BESSEL

Values

FUNCTIONS

of

J^{x)

and

7"(a:)

TABLES

OF

Table

BESSEL

IV.

Values

FUNCTIONS

of

Yjx)

733

734

OF

TABLES

Table

BESSEL

IV.

Values

FUNCTIONS

of

Y^{x)

TABLES

Table

The

numbers

points. For
first 14

are

in

[ ]

are

the

BESSEL

OF

IV.

numbers

example, the integralpart


given.

of
of

FUNCTIONS

Values

of

digitsbetween

is
Y-^^{o-\)

735

Y^{x)

tlio last digitsgiven and the decimal


number
containing19 digitsof which the

736

TABLES

Table

OF

TV.

BESSEL

Values

FUNCTIONS

of

e-^/"(a;)

TABLES

Table

W.

B. F.

OF

IV.

BESSEL

Values

737

FUNCTIONS

of

K^ {x)

47

738

OF

TABLES

Table

IV.

BESSEL

Values

FUNCTIONS

of

K,, [x)

740

TABLES

Table

OF

BESSEL

V.

Values

FUNCTIONS

of

(x)
Jj^i^^i)

TABLES

Table

OF

V.

BESSEL

Values

741

FUNCTIONS

of J

c^+

) {x)

742

TABLES

Table

OF

BESSEL

V.

Values

FUNCTIONS

of J

[x)
^ /,j^. ^\

TABLES

OF

Table

V.

BESSEL

Values

743

FUNCTIONS

of

J^^^

{x)

744

TABLES

Table

OF

V.

BESSEL

Fresnel's

FUNCTIONS

integrals

TABLES

Table

OF

V.

BESSEL

Fresnel's

745

FUNCTIONS

integrals

Maxima

integrals

J^x{t)dt

{n

minima

and

of Fresnel's

0-779893
0-321056
0-640807

10-995574

0-380389

14-137167

0-605721

0-404260
0-588128

1-570796
4-712389

7-853982

17-27S760

4-

20-420352

23-561945
26-703538
29-845130

+
+

0-417922
O-577121

0-427036

32-986723
36-128316
39-269908

0-569413

0-433666

42-411501
45-553093

+
+

0-563631
0-438767
0-559088

48-694686

0-442848

dt

TlTT

'

3-I4I593

6-283185
9-424778
12-566371
15-707963

0-713972
0-343415

0-628940

0-387969

0-600361

18-849556

2I-99II49
25-132741

+
+

0-408301
0-584942
0-420516

28-274334

0-374937

31-415927

0-428877

34-557519

0-567822

37-699112
40-840704

0-435059

0-562398

43-982297
47-123890

+
+

0-439868
0-558096

50-265482

0-443747

746

TABLES

Table

VI.

OF

Functions

BESSEL

of

FUNCTIONS

equalorder

and

argument

748

TABLES

Table

VII.

Zeros,

BESSEL

OF

j^,
"

y^,

"

ji,

FUNCTIONS

Vi,

of
n,

J^ {x), J\ (.r),Ji (x), J\ {x)

TABLES

Table

VII.

OF

Zeros, j.;,,
"

y.^, ,,

BESSEL

j^,
,

.h.

5-i35"J223

"

y^,

of
"

J^ (.r); Y^ (x), Jg (x), Y^ {x)

Ja

3-3842418
6-7938074

749

FUNCTIONS

6-3801 619
9-7610231

J's,

"

4-5270247

8-4172441
11-6198412

10-0234780

13-0152007

11-3964667

13-2099868
16-3789666

16-2234640

147959518
17-9598195

14-6230726
17-8184543

21-1169971

19-5390400

24-2701123

27-420573()

30-5092045

22-6939559
25-8456137
28-9950804

32-0648524

20-9972845
24-1662357
27-3287998
30-4869896

33-7165195

32-1430023

35-2186707

33-6420494

10

36-8628565
40-0084467
43-1534538

35-2897939

38-3704724

36-7947910

II

38-4357335
41-5810149

41-5207197

12

44-7257771

15

46-2979967
49-4421641

44-6697431
47-8177857

39-9457672
43-0953675
46-2438744

47-8701227

50-9650299

49-3914980

15

16

52-5860235

51-0141287

16

55-7296271
58-8730158

54-1578545

19

62-0162224
65-1592732

54-III6I56
57-2576516
60-4032241
63-5484022
66-6932417

52-5383976

17
18

58-8304911
61-9758587
65-I2o86l2

66-7307471
69-8736034

69-8377884

68-2655491

21

72-9820804

71-4099642

2 2

3
4

9
lo

II

12

13
14

20

57-3013461
60-4446401
63-5877658

19-4094148
22-5827295
25-7481667
28-9083508

8-0975538

^3

74-5541409

77-7302971

73-0163509
76-1590031

76-1261492

^4

79-2700214

25

8o'8728269

79-3015713

82-4137195

77-6981084
80-8418910

26

84-0152867
87-1576839

85-5572629
88-7006678
91-8439487
94-9871177
98-1301857

2 2

28

90-3000252

29

93-4423160

82-4440651
85-5864927
88-7288612
91-8711766

30

96-5845614

95-0134441

32

102-8689327

98-1556685
101-2978536

33

106-01

104-4400031

35

10655
109-1531673
112-2952406

36

^15-4372877

37

118-5793107

38

27

13

55*6846964

68-3021898
71-4449899
74-5876882

21

18
19
20

24
25

83-9855095

26

S7-1289817
90-2723230

27
28

93-4155465
96-5586637

30

29

99-7016848
102-8446186
105-9874728

3i

IIO-70I6I97
113-8443033

109-1302542

34

112-2729691

i5

116-9869284

1 15-4156229
118-5582204

36

121-7213115

113-8662672
117-0083021
120-1503138

38

39

124-8632917

123-2923041

126-4144954

40

128-0052530

126-4342746

129-5569276

121-7007659
124-8432635
127-9857167

99-7267657

34

107-5821201
110-7242073

IOI-273I62I

104-4160552
107-5588722

120-1294994
123-2720205

37

39
40

750

TABLES

Table

VII.

BESSEL

OF

Zeros, J4,
"

y^,

"

J5.

,, ,

yr^,"

FUNCTIONS

of
,

J^ [x),Y^ {x),J, {x),Y^ {x)

TABLES

Table

VII.

The

last

two

BESSEL

Zeros, ji,3,",?/i;3,",

Sn,dn,
[Note.

OF

of

functions

v'3- Wvi{x)iiOfiZO"-

751

FUNCTIONS

of

Jy.^ {x), Yy,.i{x);with

zeros,

J_i,3 {X) + J"3 {X), J_i,3 (X-) e/i;3{X)


-

arc

eciual to

l\,3{x)smio''\,

^3.

120"
:./i,,3(.r)cos

I'j3(.r)8in i20"[

respectively.]

)l

2-9025862
6-0327471
9-1705067
12-3101938
15-4506490

d.,

M::ui"

i, n

4-4657883
7-6012412

2-3834466
5-610x956
8-6473577

16-7402x28

11-7868429

13-8803575

14-9272068

1-3530196

0-8477186
3-9441020

7-0782997
10-2x69407
13-3569532

18-5914863

17-02x0330

18-0679953

21-7325412

20-1619929

21-2090210

16-4975630
19-6384856

23-303x228
20-4443623

24-3501925

22-7795923

29-5856767

27-4914601
30-632794X

25-9208x65

ID

24-8737314
28-0150117
3I-I563549

29-0621201

10

II

34-2977437

32-7270444

33-7741762

12

37-4391666

35-8684514

36-9155941

32-2034801
35-34488x3

X2

40-5806158
43-7220857

39-0098884
42-15x3485

40-0570394

14
15

46-8635719

45-2928269

46-3399899

16

50-0050715

17
18

53-1465821

48-4343202
51-5758256

56-2881019

54-7173410

19

59-4296294
62-5711634

57-8588648
61-0003956

65-7x27030
68-8542475
71-9957961

64-1419325
67-2834747

75-1373484
78-2789040

79-8496829
82-99x2426

29

81-4204625
84-5620234
87-7035867
90-8451519

30

8
9

13

20

21

22

23
24
25

II

38-4863138
41-6277704
44-7692461

13

49-48x4874
52-0229964
55-7645147
58-90604x0
62-0475740

47-9I0737X

16

51-0522406

17

54-1937545

18

43-1985061

14
15

57-3352769

19

60-4768067

20

65-1891x27

63-6183^27

21

68-3306564

66-7598840

70-42502x3

71-4722044

73-56^5718
76-7081259

77-7553112

69-9014299
73-0429798
76-1845333

74-6137562

-y

24

79-3260899
82-4676492
85-6092109

26

89-2743691

80-8968692
84-0384298
87-X799926
90-3215576

88-7507749

29

93-9867191

92-4159353

93-4631244

91-8923408

30

31

97-1282878

95-5575032

100-2698581

95-0339085
98-1754777

33

103-4114297

101-8406437

34

104-9822160

35

106-5530025
109-6945765

96-6046929
99-7462629
102-8878343
106-0294070
109-X709808

31

32

3^

112-83615x6

1 1

37

115-9777275

38
39

1x9-1x93044
122-2608821

40

125-4024605

26

27
28

86-1328048

98-6990728

108-1237894
1-2653639
114-4069394

112-3125557

117-5485159

iiS-5957082
121-7372858
124-8788641

120-6900931
123-83x6712

115-4541315

27
28

32

101-3170485
104-4586205
107-6001938

33

IXO-741768X
113-8833435

36

1x7-0249x97

38

120-1664969
123-308074S

39

34
35

37

-I"

752

TABLES

Table

VIII.

OF

BESSEL

Integralsof

FUNCTIONS

functions

of order
Maxima

zero

and

Jo (i)dt

"

Jo.

minima

T i\ (t)dt

and

of

ii

J,{i)di

"

For values of
from tlieformulae

between

and

16, the integrals


may

2-4048256

0-7352208

5-5200781
8-6537279

0-3344230

-I-

II-79I5344
14-9309177

+
+

o-f-)34o842
0-3845594
0-6028269

18-0710640
6366

0-4064156

2 1 -2 1 1

0-5864441

24-3524715
27-4934791

0-4192836

30-6346065

0-5759911
0-4279931

33-7758202

0-5685888

36-9170984
40-0584258

+
+

0-4343856
0-5629957

43-1997917

0-4393331

46-3411884

0-5585784

49-4826099

0-4433085

be calculated

with

the

help of Table

{oi." 10-74)

+Ji{x)H.o{x)},ij''^yo(i)dt
i7rx{Y"(x)Uo'i.v) Y,{x)Uoix)}j^'f,U)dl=i^x\J,{,)llo
{""-")

BIBLIOGIIAPHY*
A.
ADAMOFF,
On the asymjitoticrepresentationof the cylinderfunctions J^ (z)and J^'{z)for large
values of the modulus
Inst, polyt.1906, pp. 239"265.
Ann.
of z. Petersburg,
[Jahrbuch
iiher die Fortschritte der Math. 1907, pp. 492"493.]

Note

on

Math.
Phijs.

AICHI, K.
the Function
A'^ {x),the Solution of the Modified
Soc. ofJapan, (3) ii. (1920),pp. 8"19.

Roots

and Bessel Functions

of the Xeumann

Equation.

I'roc.

J. R.

AIREY,
The

Bessel's

(Dec. 29, 1910). Proc. Phys.Soc.

xxiii.

(1911),pp. 219"224.
The Vibrations of Circular Plates and their Relation to Bessel Functions (Feb.15,1911).
Proc. Phys. Soc. xxiii.
(1911),pp. 225"232.
Functions.
and their Relation to Bessel and Neumann
The
Oscillations of Chains
Phil. Mag. (6)xxi. (1911),pp. 736"742.
Tables of Neumann
Functions
Y^ix). Phil. Mag. (6) xxii. (1911),pp. 658
6^"(.i')and
"

663.
The

Asymptotic expansionsof

(3) XX.

(1913),pp.

The

Bessel and

other functions.

Archio

der Math,

und

Phys.

240"244.

Vibrations

Shells.
Cylindersand Cylindrical

of

der Math,

Archiv

und

Phys. (3)

(1913),pp. 289"294.

XX.

Tables of the Neumann


Math, und Phys. (3) xxii.
Bessel and Neumann

functions

of the second

Bessel functions

or

(1914),pp.

.30"43.

Functions

of

Equal

and

Order

kind.

Phil.

Argument.

Archiv

der

Mag. (6) xxxi.

(1916),pp.

520"528.
Roots of Bessel and

The

(1916),pp.

Functions

Neuma,nn

of

High

Order.

Phil.

Mag. (6) xxxii.

7"14.

Bessel Functions

of

Equal

Order

and

Phil.

Argument.

Mag. (6)xxxii.

(1916),
pp.

237

"

238.

Numerical

The

Calculation

of the

Roots

derivative Jn{x). Phil. Mag. (6) xxxiv.


The

Addition

(6) xxxvi.

Theorem

(1918),pp.

of the Bessel

of the

Bessel

Function

its first

J^ix) and

(1917),pp. 189"195.
Functions

of Zero

and

Unit

Orders,

Phil.

Mag.

234"242.

of Electric Waves
and its Applicationto the Problem
The Lommel-Weber
Q. Function
A
Soc.
Proc.
xciv.
Thin Anchor
(1918),
7,
Royal
(Dec.
1917).
pp. .307"314.
Ring
to problems of Elastic
of small Fractional Order and their application
Bessel Functions
Phil. Mag. (6)xli. (1921),pp. 200"205.
Stability.

on

B.
AIRY, SIR GEORGE
with Circular Apei-ture(Nov. 24, 1834). Trans.
On the Diffraction of an Object-glass
Camb. Phil. Soc. v. (1835),pp. 283"291.
26,
On the Intensityof Light in the neighbourhoodof a Caustic (May 2, 1836 ; March
1838). Trans. Camb. Phil. Soc. vi. (1838),pp. 379"402.
On the Diffraction of an Annular
Aperture (Dec.4, 1840). Phil. Mag. (3)xviii. (1841),

"" 10.
Supplement to a Paper,On
(March 24, 1848). Trans. Camb.

pp.

of Light in the neighbom-hoodof


the Intensity
Phil. Soc. viii. (1849),pp. 595"599.

Caustic

M.
AKIMOFF,
variables (July 10, 1916 ; June 25, 1917 ;
de Fourier-Bessel a plusieurs
Transcendantes
clxv.
(1917), pp. 23"25,
Dec. 24, 1917). Comptes Rendus, CLXiii.
(1916),pp. 26"29;
1100"1103.

In

the

case

of

iiher die Fortschritte


W.

B. F.

few inaccessible memoirs, references


der Matli. or elsewhere.

are

given

to

abstracts in the Jahrbuch

48

754

FUNCTIONS

BESSEL

OF

THEORY

Equation

dx-'^x dx
(June 16, 1898).
On

1899).

Proc.

S.

W.

ALDIS,
of the

for the Solution

Tables

Royal Soc.

y-^xv^
203"223.

(1899), pp.

LXiv.

Expansion of
(Dec.20, 1886).

in terms

Functions

313"320.

E.

1911).
imaghuiren Arguments (Leipzig,

Funktionen

der Besselschen

(1888),pp.

xxxiii.

ANDING,
Tafeln
Sechsstellige

Jn{x ^li)(June 15,

P.
ALEXANDER,
and Allied Functions
of Linear, Cylindric,Spherical

Edinburgh Royal Soc.

Trans.

and

G(){x\Oi{x)

the numerical
computation of the functions
Proc. Royal Soc. Lxvi.
(1900),pp. 32"43.

C. T.*

ANGER,

Untersuchungen liber die Function // mit Anwendungen auf das Kepler'scheProblem.


29.
der Naturforschendender Ges. in Danzig.,
v. (1855),
Neueste SchrifteM
pp. 1
"

The
pp. 1

"

V. A.
ANISIMOV,
Warsaw
Riccati's
of
form
Proceedingsof
equation.
generalised
33. [Jahrhuch iiber die Fortschritte der Math.
1896, p. 256.]
P. E.

APPELL,
Sur
de

I'inversion
Kepler et des

pp.

419"423.

approchee de certaines
de

fonctions

Bessel

integralesreelles et sur I'extension de I'equation


(1915),
(April 6, 1915). Comptes Rendiis, CLX.
L.

AUTONNE,
la nature

Sur

Rendus,
Sur
cxxviii.

De

xcvi.

des

1896,
University,

de
integrales
alg^briquesde I'equation

Riccati

(May 7, 1883). Comptes

(1883),pp. 1354"1356.

de Riccati (Feb.13, 1899). ComptesRendus,


integrales
algebriquesde I'dquation
(1899),pp. 410"412.
BACH, D.
I'integration
par les series de I'equation
1 dy
d^y n
les

"

dx"^
sci. de VEcole

Ann.

norm.

sup.

(2)iii. (1874),pp.

47

G.

F.

BAEHR,
les racines

Sur

des

dx

Equations I

68.

"

W.

cos(a;cosw)c?")=0

et

vii.
(1872),pp.
(April,1872). Archives Neerlandaises,

BALL,

Ableitung einigerFormeln
Astr. Nach.

cxxviii.

(1891),col.

"

(.r
coso))sin^

wrfw =0

351

"

358.

der Bessel'schen

Functionen

(June 6, 1891).

4.

E.

BARNES,
On

cos

L. DE.

der Theorie

aus

j
jo

JO

W.

the

cients.
homogeneous linear difference equation of the second order with linear coeffi71.
(1905),pp. 52
Messenger,xxxiv.
defined by simple types of Hypergeometric Series (March 12, 1906).
Functions
"

On
Trans.
The
Series

Camb.

Phil. Soc.

(1908),pp. 253"279.

xx.

asymptoticExpansion of IntegralFunctions defined by generalised


Hypergeometric
(Dec.3, 1906). Proc. London Math. Soc. (2) v. (1907),pp. 59"116.
A.

BASSET,
On

method

(May 10, 1886).


On
about

of

findingthe potentialsof

Proc.

Camb.

Phil. Soc.

v.

B.

circular discs bv

(1886),pp.

the Potentials of the sui-faces formed


their axes
(Oct.25, 1886). Proc. Camb.

by

the

means

of Bessel's functions

425"443.

revolution

Phil. Soc. vi.

of

Lima9ons

and

Cardioids

(1889),pp. 2"19.
A Treatise on Hydrodynamics(2 vols.)
(Cambridge, 1888).
On the Radial Vibrations
of a Cylindrical
Elastic Shell (Dec. 12, 1889). Proc. London
Math. Soc. XXI.
(1891),pp. 53"58.
On a Class of Definite Integralsconnected
with Bessel's Functions
(Nov. 13, 1893).
Proc. Camb.
Phil. Soc. viii. (1895),pp. 122"128.
*

See also under

Bourget and

Cauchy.

756

des Theils

Untersuchung
Sonne

entsteht

der

die

Ahh.

Storungen

siir

^^

/ y^^dye r
I'intdgrale
J

Comptes Rendus,

xii.

der

berechnen

zu

der

Bewegung

der

I.
[Abhandiungen,

(Sept.24, 1836).

J. P. M.

priseen

tre des limites arbitraires

(May 24, 1841).

(1841),pp. 958"962.
M.

BOCHER,
85"90.
On

Kometeu

aus

1"52.

a,

of the

Bessel's functions

On

[1826],
pp.

1824

i. (1875),
[^Abhandiungen,
pp. 29"54.]

1"48.

BINET,
Note

welcher

planetarischenStorungen

(Jan. 29, 1824). Berliner

(1875),pp. 84"109.]
Beitrag zu den Methoden
Astr. Nach. xiv. (1837),col.

FUNCTIONS

BESSEL

OF

THEORY

second

kind

Annals

(Jan.1892).

of Math.

(1892),pp.

vi.

of Bessel's functions with pure imaginary index (Feb. 11, 1892).


applications
of Math. vi. (1892),pp. 137"160*.
of Bessel's functions
to the roots
and their application
certain methods
of Sturm
some

Annals
On

3Iath. Soc. III. (1897),pp. 205"213.


(Feb. 1897). Bidletin American
An
elementaryproof that Bessel's fimctions of the zeroth order have an infinite number
Math. Soc. v. (1899),
of real roots (Feb.25, 1899). Bulletin American
pp. 385"388.
linear difterential equationsof the second order (Feb.4, 1901). Bulletin
Non-oscillatory
Math.

American

Soc.

vii.

333"340.

(1901),pp.

P. E.
BOHMER,
ijber die Zylinderfunktionen(Nov. 26, 1913). Sitz. der Berliner Math. Ges.

xiii.

(1913),

pp. 30"36.

BOHREN,

A.

Tiber das Airysche Integral


t (Oct.6, 1902). Bern
BOOLE,
On

the transformation

On

of Definite

1902, pp.
Mittheilungen,

236"239,

G.

Integrals.Camb.

Math.

Journal,ill. (1843),pp.

216

"

224.

generalmethod

pp. 225"282.
A Treatise

Note

p. 283.
Mdmoire
sur
celeste. Journal
Memoirs

sur

sci. de VEcole

de

formule

M.

de Cauchy et leur
les nombres
de Math. (2)vi. (1861),pp. 33
vibratoire

le mouvement

norm.

sup.

ill.

(1866),pp.

Fourier-Besselschen

asymptotischer Darstellungen

55

des
"

"

1844,

(1854),

54.

membranes

circulaires

(June 5, 1865).

Ann.

95.

J. R.
deren

und

zur
Anwendung
Auffindung
der linearen
mit
Integi-alen
Diff'ereutialgleichungeu
Polyt.Inst. Nach. 1902, nos. 1, 2. [Jahrbuchiiber die

von

577.]

"

BRASSINNE,

E.

diverses

(sic).Mem.
Sur des
Math. XVI.

de Ricatti
equationsdifterentielles du premier ordre analoguesk I'equation
de VAcad.
R. des Sci. de Toidouse,(3) iv. (1848),pp. 234
236.
de Riccati. Journal
equations difterentielles qui se rattachent a I'equation
de
(1851),pp. 255"256.
"

W.

BRENKE,
Summation
Bidl. American

On

Soc.

k divers probl5mes de mccanique


application

Funktionen

Koefiizienten.
Warschau
rationalen
Fortschritte der Math. 1903, pp. 575

Sur

of the Royal

J.
BOURGET,
Anger (Aug. 7, 1854). Comptes Rendtis,xxxix.

BRAJTZEW,
tjber die

Trans,

Differential
Equations(London, 1872).

on

une

sur

analysis(Jan.18, 1844). Phil.

in

Certain

(1908),pp.

of

series of Bessel's functions


Math. Soc. xvi.
(1910),pp. 225
a

"

Development

by

C.
means

of

an

integral(Nov. 27, 1909).

230.

P. W.
BRIDGEMAN,
in Bessel's Functions
(July 22,

1908). Phil. Mag. (6) xvi.

947"948.
"

See also ibid. p. 136.


t This paper, which might have
the author does not give the formula

been

mentioned

"6-4 (2).

in "

6-4, contains

the formula

"6-4 (1); but

757

BIBLIOGRAPHY

H.

BRUNS,
Ueber
.

die

(1883),col.

des Heliometer-Objectives
Beugungsfigur
(Oct. 15,

BRYAN,
On

the

1882). Astr. Nach.

civ.

1"8.

waves

on

viscous

G.

H.

rotatingcylinder (June 4, 1888).

Proc.

Camb.

Phil. Soc.

vi.

(1889),pp. 248"264.
Wave

Motion

and

Bessel's Functions.

Nature^ lxxx.

BURKHARDT,
TrigonometrischeReihen
1 (Leipzig,
1904"16), pp.

II.

und

H.

F. K.

Integrale (bisetwa

Wins.

S.

Combinations

(May

(1913),pp. 294"297.

xxv.

de Bessel.

Ber, Bei and Allied Functions

of the

C.

CAILLER,
les fonctions

Sur

L.

819"1354.

of Certain

30, 1913). Proc. Phys. Soc.

309.

1850). Encyklopddie der Math.

BUTTERWORTH,
the Evahiation

On

(1909),p.

Archives

des Sci.

{Soc.Helvetique),
(4) xiv.

(1902),pp. 347

"

350.

Note

sur

operationanalytiqueet son applicationaux fonctions de Bessel (March,


(1902 5),pp. 295
phys. et d^histoire nuturelle de Geneve,xxxiv.

vine

1904). Mem.

de la Soc. de

"

"

368.
0.

CALLANDREAU,
Calcul

des transcendantes

pour

les

(2) XIV.

(i)'
r, (Sf

"^"^ ,

1.2..."

grandes valeurs
(1890),pp.

de Bessel

de

au

l.C"

(I)'

"""J'
l)'^1.2.(?i+l)(7i
+ 2)

de series

moyen

semiconvergentes.Bulletin

des Sci. Math.

110"114.

Sur le calcul des polynomes X^ (cos6) de


des Sci. Math. (2) xv. (1891),pp. 121"124.

della serie che

Bulletin

n.

F.

CARLINI,
Ricerche sulla convergenza

les grandes valeurs de

Legendre pour

alia soluzione del

serva

prohlema

di

Keplero*

(Milan,1817).
Some
and

Multiform

H. S.
CARSLAW,
of the Partial Differential

Solutions

Equations of PhysicalMathematics
161.
Apphcations (Nov. 10, 1898). Proc. London Math. Soc. xxx. (1899),pp. 121
Green's function for a wedge of any angle and other Problems
in the Conduction

their
The

of Heat

"

(Oct. 30, 1909). Proc. London

The

Scatteringof

Sound

Waves

The Green's function for the


Proc. London
Math. Soc. (2) xiii.
The

Theory of the

Math.

Soc.

by a Cone.
equation V-u
of Heat

Note

COS

I
I'integrale

sur

./0

'

ft3j

"

(London, 1921).
E.

CATALAN,
'^

147.
Lxxv.
(1914),pp. 133
(April28, 1913; March 20, 1916).
(2) xvi. (1917),pp. 84"93.

Ann.

+ k-u^O

(1914),pp. 236"257

Conduction

(2) viii. (1910),pp. 365"374.

Math.

C.

doc

(Feb.1840).

^r-

Journal

de Math.

(1840),pp. 110"114.

v.

(1 +X-)'"'

de la Soc. R. des Sci. de Liege,(2) xii. (1885),pp. 26


31.]
de Riccati (March 4, 1871). Bulletin de I'Acad. R. de Belgique,(2) xxxr.
I'equation
(1871),pp. 68"73.
Note
sur
I'equationxy" + ky' xy 0 par M. C. Le Paige (Rapport de M. Catalan).
Bulletin de I'Acad. R. de Belgique,(2) XLI. (1876),pp. 935
939.
^Applicationd'une formule de Jacobi (Nov. 1868). 3Iem. de la Soc. R. des Sci. de Li^ge,
Mem.
[Reprinted,

"

Sur

"

"

(2) XII. (1885),pp.

312"316.

CAUCHY,
Rdsume
des

d'un memoire
sur
(Lu k I'Acad. de

limites

pp. 48"112.
Memoire
pp. 587"588.
*

Translated

Werke,

vn.

la

mecanique

A.

celeste

L.
et

sur

un

calcul

nouveau

appele calcul

Turin, Oct. 11, 1831). Exercices d^Analyse, ii. (Paris,1841),


[Oeuvres,(2) xii. (1916),pp. 48"112.]
des series (Nov. 11, 1839). Comptes Rendus, ix. (1839),
la convergence
sur
(1) iv. (1884),pp. 518"520.]
[Oeuvres,
into

German

by Jacobi,

(1891),pp. 189"245.]

Astr.

Nach.

xxx.

(1850),col.

197

"

2.54.

[Ges. Math.

758

nouvelles

Considerations

la theorie

sur

FUNCTIONS

BESSEL

OF

THEORY

des

(April20, 1840). Comptes Rendiis,x. (1840),pp.

suites et

sur

les lois de leur convergence

(1) v. (1885),
[Oeuvres,
pp.

640"656.

180"

198.]
numerique des coefficients que rensimple et gdndralepour la determination
(Sept.14, 1840). ComptesRe7idus,xr.
developpement de la fonction perturbatrice
(1) v. (1885),pp. 288"310.]
[Oeuvres,
(1840),pp. 453"475.
Note sur le developpement de la fonction jjerturbatrice
(Sept.21,1840). ComptesRendus,
XI.
(1)v. (1885),pp. 311" 321.]
[Oeuvres,
(1840),pp. 501"511.
le calcul des inegalites
h simplifier
Methodes
periodiqueset seculaires des
propres
101.
des planetes(Jan. 11, 1841). Comptes Rendus, xil.
mouvements
(1841), pp. 84
(1) vi. (1888),pp. 16"34.]
[Oeuvres,
le developpement de la fonction perturbatrice
Note sur
transcendante
une
que renferme
relative au
1841).
(Oct.
4,
ComptesRendus, xiii. (1841),pp. 682 687.
systeme planetaire
[Oeuvres,
(1)vi. (1888),pp. 341"346.]
dans le
Note sur
la siibstitution des anomalies
escentriquesaux anomalies moyennes,
(Oct.25, 1841). ComptesRendus, xiii. (1841),
developpement de la fonction perturbatrice
(1) VI. (1888),pp. 3.54"359.]
[Oeuvres,
pp. 850"854.
des mouvements
Nouveau
le calcul des inegalites
Memoire
sur
planetaires(April 8,
(1) viii. (1893),pp. 168"
(1844), pp. 625"643.
1844). Comptes Rendus, xviii.
[Oeuvres,
Methode

ferme

le

"

"

188.

Sur

la transformation

leurs
fonctions iraplicites
et sur
en
isotropiques,
moyennes
trigonometriques(May 22, 1854). Comptes Rendus, xxxviii.
(1854),pp. 910"913.
(1) xii. (1900),pp. 148"151.]
[Oeiivres,
les mouvements
Sur la transformation
des variables qui determineut
d'une plan^teou
d'une comete
le developpement de ses fonctions
fonction explicite
du temps, et sur
meme
en
series couvergentes (June 5, 1854). Comptes Rendus, xxxviii.
en
993.
(1854),pp. 990
(1) xii. (1900),pp. 160"164.]
[Oeuvres,
Sur la resolution des equations et sur
le developpement de leurs racines en series con1107. [Oeuwes,(1)
vergentes (June 26, 1854). ComptesRendus, xxxviii.
(1854),pp. 1104
XII.
(1900),pp. 167"170.]
Sur
de M. Anger et sur
d'autres formules
formule
une
analogues (July 15, 1854).
135. [Oeuvres,
(1) xii. (1900),pp. 171
ComptesRendus, xxxix.
(1854),pp. 129
177.]

developpements

en

des

series

"

"

"

"

CAYLEY,
quelques formules du calcul integral.Journal de Math. xii. (1847),pp. 231
240.
[Collected
Papers,I. (1889),pp. 309"316.]
On
Eiccati's equation (Sept.29, 1868). Phil. Mag. (4) xxxvi.
(1868), pp. 348"351.
[Collected
Papers,vii. (1894),pp. 9"12.]
Note
the integrationof certain differential equations by series. Messenger(Old
on
v. (1869),pp. 77"82.
Series),
Papers, viii. (1895),pp. 458"462.]
[Collected
Proc. London
Math.
Soc*
v.
(1874), pp. 123"124.
[Collected
Papers,ix. (1896),pp.
A.

Sur

"

CHALLIS,
Extension

(1866),pp.

of the Solution

of Eiccati's

H.

51"53.

CHAPMAN,
On

W.

Equation (Oct.5, 1864). QuarterlyJournal,vii.

the

generaltheory of summability
QuarterlyJournal,XLlii. (1911),pp. 1 52.

S.

with

to
applications

Fourier's

and

other

series.

"

CHESSIN,
Note

the

General

Solution

of Bessel's

A.

S.

American
Journal of Math, xvi,
(1894),pp. 186"187.
On the expression of Bessel's Functions
in Form
of Definite Integrals.Johns Hopkins
Univ. Circulars,
21.
xiv.
(1895),pp. 20
Note ou Cauchy's Numbers.
Annals
of Math. x. (1896),pp. 1 2.
On
the relation between
and
Bessel's Functions
Cauchy's numbers
(July 1, 1898).
Annals
of Math. xii. (1899),pp. 170"174.
on

Equation.

"

"

On

relations

between

Bessel functions of the first and of the second


kind (Oct.
^ci. of St Louis,xii. (1902),
pp. 99"108.
de Bessel avec
second membre
I'equation
(Oct. 27, 1902). Comptes Rendus, cxxxv.

some

20, 1902). Tra7u. Acad.


Sur

(1902),pp. 678"679.
Sur

classe d'equationsdiflerentielles rt'dactibles


une
1903). Comptes Rendus, cxxxvi.
(1903),pp. 1124"1126.
*

See under

Lord

Rayleigh.

I'equationde

Bessel

(May 11,

759

BIBLIOGRAPHY

CHREE,
Longitudinalvibrations

of

circular bar.

C.

QuarterlyJournal, xxi.

298.
(1886),pp. 287
the Coefficients in certain Series of Bessel's Functions.
Phil. Mag. (6)xvii.
(1909),
pp. 329"331.
E. B.
CHRISTOFFEL,
Zur Abhandlung : "Ueber
die Ziihler und
Nenner
Kettender Niiherungswertevon
briichen"
(March, 18G0). Journal far Math. LViii, (1861),
pag. 231 des vorigen Bandes*
a

"

Ou

pp. 90"92.
DifFrazione

M.
CINELLI,
curve.
aperture fatte sopra sujierfici

per

II Nuovo

Cimento,(4) i. (1895),

141"155.

pp.

CLEBSCH,
Ueber

die Reflexion

einer

an

R.

F.

A.
Journal

Kugelflache(Oct. 30, 1861).

fur Math.

(1863),

LXi.

pp. 195"262.
W.

CLIFFORD,
On

t. Mathematical

Bessel's Functions

K.

Papers (London, 1882), pp.

of the Second

order.

Bessel's functions

of the Second

order.

QuarterlyJournal.,XX. (1885),pp.
QuarterlyJournal,xxi. (1886),pp.

COCKLE,

SIR

Equations of the Second Order (Dec. 24, 1861


May 21, 1862). Messenger(Old Series),i. (1862),pp. 118"124, 164"173,
COTTEH,

250

Method

Jan.

S sin

of Rodrigues' Theorem
proof
^
"

expansions derived

some

pp. 11

from

'^

it

15, 1862;

J. R.
of

of

nx=-

similar

L.

CRAWFORD,
A

192.

"

241"247.

Solving Legendre'sand Bessel's Equations and others


A (1909),pp. 157"161.
type (May 27, 1907). Proc. R. Irish Acad. % xxvii.
New

260.

"

183

JAMES.

I,inear Difierential

On

349.

"

C. V.

COATES,
Bessel's functions

346

"

"

"

"

"

~.

"

sin^""'

-5-

and

.b...(2n~\) \ii\nxdxj

I .A

Proc.

(Dec. 13, 1901).

Edinburgh

Math.

Soc.

xx.

(1902),

15.

"

L.

CRELIER,

quelques proprietesdes

Sur
continues

di Mat.

(June, 1895). Ann.

tirees de

Besseliennes

fonctions

(1896), pp.

(2) xxiv.

la theorie

des

fractions

Bern,
[Dissertation,

131"163.

1895.]
la fonction

Sur

[1898],pp.

besseliennes
(Sept.6, 1897;
0{x) et .S'"(.'r)
(1897),pp. 421"423, 860"863.

les fonctions

Sur

Rendus,

cxxv.

CURTIS,
On
and

the

Dublin

of
integration
Math.

and

Linear

Journal,

IX.

Partial

(1854),pp.

A. H.
Differential

272

"

CURZON,
of the

Generalisations

(Nov. 10, 1913).

Proc. London

Math.

Soc.

E.

H.
and

functions

Hermite

(2)

p20).
On

generalisationof Neumann's

Math.

Soc.

under

Heine.

See

xi.

J Not Trans. Camb.


1008, p. 383.
" See footnote t on
which
II This
paper,
and with various
Press.

(1921),pp.

xiii.

J.

their

with

connexion

Soc.

Bessel

functions

(1914),pp. 417"440.
A.

Expansion

in

Series of Bessel

(Feb.29,

Functions

(Jan.6, 1921).

Bulletin

221"230.

t Clifford died March


Phil.

29, 1897). Comptes

Equations (Nov. 24, 1854). Camb.

Bulletin Calcutta Math. Soc. xi. (1921),pp. 23"34.


extension
of Sonine's
an
Integral in Bessel Fuuctions|l

Calcutta

Nov.

290.

DATTA,
On

1897
Mittheilungen,

^S'"(.i-)
(Dec. 1896). Bern

de ii"= espece

Besselienne

61"96.

xxi.,

as

stated

in the Jahrbuch

8, 1879.

iiber die Fortschritte

der Math.

p. 27.

deals

with

related

the

forna, given in " 18-46 (10),of


published after Chapter xni had

corrected

integrals,was

Nicholson's
been

tegral,
in-

passed for

FUNCTIONS

BESSEL

OF

THEORY

760

P.
DEBYE,
fiir grosse Werte
fiir die Zylinderfunktionen
des Index
(Dec. 1908). Math.
Werte
veranderliche

des

Naherungsformelu
unbeschrankt
pp. 535"558.

und
Semikonvergente Entwickelungen fiirdie Zylinderfunktionen
no.
XL.
(1910),
Miinchener
Sitzungsherichte,
(Feb?
5, 1910).
Komplexe

.,

ihre

1917). Proc. Royal

of Yibrations
B
Soc. lxxxix.

5.

Sponge Spicules(May 11,

U.
delle

analitiche
rappresentazioni

altere

de la Soc.

Ann.

573"587.

(1917),pp.
DINI,

Se7-ie di Fourier

of Certain

Form

the

upon

ins

J. W.

XICHOLSOX,

AND

A.

DENDY,
the Influence

Ausdehnung

Integrationde I'equationde Bessel sous forme finie (Jan. 26, 1905).


Sci. de Bruxelles,xsix.
(1^'^partie)(1905),pp. 140"143.
On

(1909),

Lxvii.

C. J.

POUSSIN,

YALLEE

LA

DE

und

Arguments

Ann.

funzionidi

variabile reale

una

(Pisa,1880).
A.
DINXIK,
tJber die Dar.stellung
einer willkiirlichenFuuktion
Inst. {Engineering Section),1911, no.
1, pp. 83"85.
Math.

Reihe.
Kief.Polyt.
die Fortschritte der

der

und

Funktionen

J^^^

Funktionen

^^^l, ^"

Besselscben

Archiv

J^a.

Math,

der

und

Phys. (3)

(1911),pp. 337"338.

XVIII.

der Besselscben

Tafeln

(1913),pp.
Tafeln

XXI.

Bessel'sche

[Jakrbuchuher

1911, p. 492.]

Tafelu

XX.

durch

der

On
The

Funktionen

J^

{xi)und J^

.""

in Bessel's Functions.

Phys. (3)
und

Phys.

8.

"

(1903),p.

8.

F.

W.

Equation of Laplace'sFunctions,etc. (Dec. 11, 1856).


57.
(1857),pp. 43

Phil. Tram,

of the Royal

"

determination

(March 9, 1900).
Theorem

Die Theorie

(1871),pp.

of Green's

Proc.
Proc.

function

by

Math.

Soc.

Edinburgh

of Sonine

(Dec. 13, 1918).

in Bessel

J.

Edinburgh Math.

of

means

with

Soc.

or
SphericalHarmonics
Cylindrical

(1900),pp.

xviii.

Functions

two

(1919),pp.

xxxvii.

33"83.

Extensions

to

SphericalHarmonics

33"47.

P. D. G.
BOIS
DU
REYMOXD,
der Fourier'schen
(June 26, 1871). Math.
lutegraleund Formeln

Ann.

iv.

362"390.

S.

EARNSHAW,
Partial differential
equations.An
them (London, 1871).

essay

towards

ELLIS,

On

und

der Math,

(1903),pp.

Messenger,

DOUGALL,

On
Math.

{xi).Ai'chiv

xxxil.

DOXKIX,

Phys. (3)

226"227.

expansion of

Tbe

A. C.
DIXON,
Messenger,xxxil.
property of Bessel's Functions.

On tbe
Soc. cxLvii.

und

der Math,

t/^, und J'^g-.Archiv

Funktionen

Besselscben

(1914),pp.
a

der Math,

Archiv

"^"s-

238"240.

(1913),pp. 324"326.
Tafeln der Besselscben

(3)xxil.

s,

an

"method

nexo
entirely

of integrating

L.

R.

the

and

Integrationof certain Differential Equations (Nov. 1840


Journal,11. (1841),pp. 169"177, 193"201.
the Method
of Least Squares (March 4, 1844). Trails. Camb.

Feb. 1841). Camb.

Phil. Soc.

viii.

(1849),

pp. 204"219.

EMDE,
Zur
der

F.*
Bessel'schen

der
Berechnung der reellen NuUstellen
Math, und Phys. (3) xxiv.
(1916),pp. 239"250.

A.

ENXEPER,
Ueber

ein bestimmtes

Integral.Math.

Ann.

EPSTEIX,
Die vier

mit
Rechnungsoperationen

Einleitung(Bern, 1894,
845"846.]

58

Zylinderfunktionen.Archiv

vi.

(1873),pp. 360

S. S.
Functionen

"

365.

nebst einer
der Math.
Uber
die
Fortschritte
pp.),[Jahrbuch
*

See

Bessel'schen

also under

Jahnke.

geschichtlichen
1893

"

94, pp.

761

BIBLIOGRAPHY

W.

ERMAKOFF,
Ueber

die

Cylinderfunctionen
(May, 1872).

Math.

ESCHERICH,
Zur

Bessel'schen

G.

Ann.

639"640.

(1872),pp.

v.

VON.
Math,

Diflferential-Gleichung.
Monatsheftefur

und

Phys.

iii.

(1892),

p. 142.

tjber eine Nalierungsformel.MonatsheftefiirMath,

de M.

pp. 1"10.
Recherches

Euler

;\ M. de la

Phi/s.ill. (1892),p.

234.

L.

EULER,
Lettre

uiid

Grange (Jan.1, 1760).

Misc.

Taurinensia, ii. (1760

61),

"

de I'equatiou
rintegration

sur

ddz

ddz

h dz

dx'^

dt"
Misc.

XX

Taurinensia,III. (1762"65), pp. 60"91.


integratione
aequationum differentialium.Novi

De

[1763],
pp.
De

aequationis dy-\-av"/dx Kv''"'dx. Novi

resolutione

243"260.
InstitrUionum
De

[1784],
pp.

Calculi

Coinm.

JVovi

tympanorum.

ii. (Petersburg,
1769).
Integralis,
funis libere suspensi.Acta

Acad.

ab

chordarum

"

Petrop.v.

pondere oriunda.

earum

Analysis facilis aequatiouem Riecatianam


per fractionem
R. des Sci. de St. Pe'tersbourg,
vi. (1818),pp. 12
"

pars

(1781)

Nieutv

in series of the

resolvendi.

continuam

(1921),no.

Mem.

in der
"

Nahe

einer

248.

H.

it"t-i^}
integral/ e""*

Fysik,xv.

Petrop.v.

29.

J. H. M.
FALKENHAGEN,
Integraleeiner Riccati'schen Gleichung
WisL-unde,(2) Vl. (1905),pp. 209
Archiefvoor

FAXEN,

Expansion

Acad.

Acta

der

Verhalten

das

singularenStelle.

^^^^

(April14, 1920). Arkiv for Mat.

13.

M.

FELDBLUM,

satisfies it.
theory of Riccati's equation and applicationsof the function which
1898, nos. 5 and 7 ; 1899, no. 4. [Jahrbuchiiber die Fortschritte der
University,

The
Warsaw
Math.

ix. (1762
Pctroj).

178"190.

de I'Acad.

Astr. och

61)

157"177.

(1781) [1784],pp.

Ueber

"

x.
(1764) [1766],pp.
Petro-p.

Acad.

minimis

perturbationemotus

Petrop.viil. (1760

Acad.

Coium.

oscillationibus

De

Acad.

Comm.

154"169.

63) [1764],pp. 154"169.


Je motu
vibratorio

pars

dx

1898, pp. 279"280.]


K. DE.

FERIET,

hypercylindriques(June 13, 1921). Comptes Rendus, CLXXii.

les fonctions

Sur

(1921),

pp. 1464"1466.
J. C.

FIELDS,
A
VI.

of

method

solvingRiccati's Equation (April8, 1886).

(1886"87), p.

Analogous

Solutions
Johns

Hopkins

Univ.

to

Riccati's of

Circidars,vi. (1886

Equations
"

New

Mode

of

Expressing Solutions

On

the
Soc.

expansion of

the

(2) iv. (1906),pp.

Univ. Circulars

real variable

(1903),pp.

(May
y-^=T'"_y

19, 1886).

of

Operators

G.

Laplace'sEquation

in Terms

193"213.

(May 10, 1906).

Proc. London

396"430.

W.

B.

of diflferentiating
term-by-term
possibility

function of one
Math. Soc. IV.

HopHns

30.

Phil.

FORD,
On

L. N.
of

"

Mag. (6) vi. (1903),pp.


polynomials in series of functions

involving Bessel Functions.


Math.

of the form

29

87), pp.

FILON,
On

Johns

29.

in terms

178"184.

of Bessel

the

functions

developments for an
(June, 1902). Trans.

arbitrary
American

FUNCTIONS

BESSEL

OF

THEORY

762

A.

FORSYTH,
On

: in
lijieardifferentialequations

-^"^

^^

R.

that
particular

satisfied by the series


'

'^"""

+ l

+ l.f.f

1.2.y.y

ye

Journal,XIX. (1883),
i^^. 292" S37.
Quarterlv
and
order as products,
of Bessel functions of positive
The expression
129"149.
h. (1921),
fractions. Messenger,
rational
of
pp.
sums
as
powers
La

de
analytique

Theorie

",.,""

of their

inverse

J. B. J.
FOURIER,
bridge,
by A. Freeman, Cam(Paris,1822). [Translated

la Chaleur

1878.]
A.

FREEMAN,
Note on
Proc. Camb.

equationallied to Jq{z)

the value of the least root of an


Phil. Soc. iii. (1880),pp. 375"377.

(April19, 18S0).

FRESNEL, A. J.
la diffraction de la lumifere [July29, 1818 ; crowned 1819].Mem.
Memoire
sur
i. (1866),
[Oeuvres,
R. des Sci. v. (1821"22),pp. 339"476.
pp. 247"382.]

de VAcad.

G.

FRULLANI,

Sopra la dipendenzafra i differenziali delle funzioni


1818). 2Ie7n. soc. ital.(Modena),xviil. (1820),pp. 458"517.

definiti (Feb.4,
gliIntegrali

FUSS, P. H.
celebres geotnetres
du
et physique de quelqtces
mathe'matique
Correspondance

xviii^'""

II. (Petersburg,
lb43).
siecle*,

E. G.
circular disc and

GALLOP,
distribution of

The

the
on
electricity

Journal,xxi. (1886),pp.

the

bowl. Quarterli/
spherical

229"256.
A.
Funktionen

GASSER,
Ueber die NuUstellen der Besselschen
Ges. in Bern, 1904, pp. 92"135.

der Naturf.
(July,1 904).Mittheilungen

L.
zweiter Art

GEGENBAUER,
liber die Bessel'schen

Note

Lxv.
herichte,

(2)(1872),pp.

Functionen

Zur Theorie der Bessel'schen Functionen


Lxvi.
berichte,

(2)(1872),pp.

Note iiber bestimmte


pp. 202"204.
tjber die Functionen
pp. 357

"

(Feb.8, 1872). Wiener Sitzungs-

33"35.

zweiter Art

(July4, 1872).

Wiener

Sitzungs-

220"223.

(Feb.6,1873).
Integrate

Wiener

Lxvii.
(2)(1873),
Sitzungsherichte,

Lxviii.
(2)(1874),
AV" (June 13, 1873). Wiener Sitzungsberichte,

367.

Lxx.
tJber die Bessel'schen Functionen (March 19, 1874). Wiener Sitzungsberichte,
(2)
(1875),pp. 6"16.
tJber einige bestimmte
lxx.
(2)
Integrale(June 18, 1874). Wiener Sitzungsberichte,
(1875),pp. 433"443.
tJber einigebestimmte
Lxxii.
(2)
Integrale
(June 17, 1875). Wiener Sitzungsberichte,
(1876),pp. 343"354.
Tiber die Bessel'schen Functionen (June 22, 1876). Wiener Sitzungsberichte,
LXXiv.
(2)
(1877),pp. 124"130.

Zur

Theorie

der

Bessel'schen

Functionen

(Jan. 18, 1877).

Wiener

Sitzungsberichte,

(2)(1877),pp. 218"222.
tJberdie Functionen (7/(.r)
Lxxv.
(April12,1877). Wiener Sitzungsberichte,
(2)(1877),

Lxxy.

pp. 891"905.
Das Additionstheorem
e"^ nach
Functionen
welche bei der Entwicklungvon
deijenigen
den Naheruugsnennernreguliirer
Kettenbriiche auftreten. Wiener Sitzungsberichte,
Lxxxv,

(2)(1882),pp.
Uber

491"502.

die Bessel'schen Functionen


(1884),pp. 975"1003.
*

Lxxxviii.
(Oct.11, 1883). Wiener Sitzungsberichte,

(2)

This work contains a number


of letters from Nicholas Bernoulli (theyounger) to Goldbach,
Bernoulli's solution of Eiccati's equation is to be fonnd.
The reader should notice that,in
Daniel Bernoulli's letters to Goldbach
[ibid.
pp. 254, 256, 259),the equation described in the table
of contents
Eiccati's equation is reallythe linear equation; Eiccati's equation is mentioued
as
on
p. 260.

in which

FUNCTIONS

BESSEL

OF

THEORY

764

Example illustrative of a point in the solution of ditlerential equations by


Messenger,viii. (1879),pp. 20"23.
involving repeateddifferentiations (May 19, 1879). Proc.
On a symbolic theorem

series.
Camb.

Phil. Soc. III. (1880),pp. 269"271.


and on
Definite Integrals
which
some
On Riccati's Equation and its Transformations
the Royal Soc. 172 (1881),pp. 759"828.
Trans,
Phil.
of
them
1881).
16,
(June
satisfy
[Proc.Royal Soc. xxxii. (1881),p. 444.]
P.

GORDAN,
See under

Hermite.

J. H.
GRAF,
bei Bessel'schen Functionen
der Argumente
Subtraction
(1893),pp. 136"144.
einer Anwendung (March, 1893). Math. Ann. XLiii.
erster
Art, insbesondere
Ueber einigeEigenschaftender Bessel'schen Function

fur
Zeitschrift

grosses Argument.

Relations
di Mat.

entre

fiii-ein

115"120.

von

(1895),pp. 45"65.

(2) xxiii.

Ableitung der
Distanz

(1893),pp.

xxxviii.

zweiter Ordnung mit liuearen


linearenDifierentialgleichuugen
von
zweiter Ordnung denen gewisse bestimmte
Differentialgleichungen
(March, 1894). Math. Ann. XLV. (1894),pp. 235"262.
fraction continue (May, 1894).
et une
la fonction Besselienne de l""*^
espfece

wie

so

Integralegenugeu
Ann.

Math,

Auflosung

Beitrage zmCoeificienten

nebst

und

die Addition

Ueber

darstellt

Formeln

fur die Bessel'schen

bei welchen

Functionen

(Aug. 4, 1896). Verhandlungen

der

das

Oes.
Schweiz-Naturf.

Argument

ein

(1896),

Lxxix.

pp. 59"62.

Einleitungin

die Theorie

der Besselschen

Funktionen.

Von

J. H.

Graf

und

(2 Hefte ; Bern, 1898, 1900).


zweiter Ordnung denen
Diflfereutialgleichungen
Beitrag zur Auflosung von
Integralegeniigen (May, 1902). Math. Ann. lvi. (1903),pp. 423"444.

Treatise

on

Bessel Functions.

By

GRAY,

A.*

Andrew

Gray

SIR

GREENHILL,

A.

and

G. B. Mathews

E. Gubler
stimmte
gewisse be-

(London, 1895).

GEORGE.

298.
Equation and Bessel's Equation. QuarterlyJournal, xvi. (1879),
pp. 294
of the Strata in the Theory of the
On the Differential Equation of the Ellipticities
Figure of the Earth (April8, 1880). QuarterlyJournal, xvii. (1880),pp. 203"207.
Determination
that a vertical pole or
of the greatest height consistent with stability
be made, and of the greatest heightto which a tree of givenproportionscan grow
mast
can
(Feb.7, 1881). Proc. Camb. Phil. Soc. iv. (1883),pp. 65"73.
The Bessel-Cliftbrd Function
cvii.
(1919),p. 334.
(March 14, 1919). Engineering,
The Bessel-Cliftbrd Function
and its applications
(Aug. 11,1919). Phil. Mag. (6)xxxviir.
(1919),pp. 501"528.
J. A.
GRUNERT,
On

Riccati's

"

(^i-iH"
Beweis
der Math,

der
und

Gleichung

22y-j

ITJ^

Phys. iv. (1844),pp.

sin?".r

=(-l)'"U.3...(2i-l)

"

fur3

r"

Archiv

cos.r.

104"109.

GUBLER,
Die

"

E.+

Darstellung der allgemeinen Bessel'schen

Function

durch

(Sept.1888). Zurich Vierteljahrsschrift,


xxxiii.
(1888),pp. 130"172.
Verwandlung einer hypergeometrischenReihe im Anschluss an

bestimmte
das

Integrale

Integral

/"oo

J''{.v)e-'"'.v'-'^dx.

Inaugural-dissertation,
Zurich, 1894 (38 pp.). [Grafand Gubler, Einleitung
in die
der Besselschen Funktionen, ii. (Bern,1900),
pp. 110"135, 156.]
Ueber ein discontinuierliches Integrale
(Dec. 1895). Math. Ann. xlviii.
(1897),pp.
Beweis
Ueber

einer Formel
bestimmte
XLVii.

Bemerkungen

des Herrn

Sonine

Integralemit

(Dec. 1896). Math. Ann.

Bessel'schen

Functionen

Theorie
37"48.

(1897),pp. 583"584.
jahrsschrift,
(Oct,1902). Zurich Viertelxlix.

(1902),pp. 422"428.
iiber

S.
GUNTHER,
Cylinderfunctionen.Archiv

der

Math,

pp. 292"297.
*
See also under
t See also under

Sir

Joseph John

Graf.

Thomson.

und

Phus.

lvi.

(1874),

BIBLIOGRAPHY

The

in

employment of
its develoijmentinto
Sur

H. F.
GWYTHER,
to prove
Schlomilch's
geometrical construction
series,and
definite integral. Messenger,
107.
xxxiii.
(1904),})p. 97
J.
HADAMARD,

a
a

(1908),pp.

xxxvi.

to aid

"

I'expression
asymptotique de

France,

765

77

la fonction

de

Bessel.

de la Soc. Math,

Balletin

de

85.

"

E.
HAENTZSCHEL,
die functionentheoretischen
Zusammenhang zwischen den Lame'-schen,Laplace'schen und Bessel'schen
Functionen.
33.
Zeitschrift
(1886),pp. 55
fur Math. xxxi.
Ueber
die Fourier
Bessel'sche Transcendente
(Nov. 20, 1887). Zeitschrift
fur Math.
Ueber

"

(1888),pp.

XXXIII.

Studien

185"186.

M.
HAFEN,
dcr Potentialtheorie.

iiber einigeProbleme

HAGUE,
A

Note

the

on

(1917),pp.

XXIX.

The

of the

Graphs

Bessel

Math.

Ann.

Besselian

of

Functions

IntegralOi'der.

The

Function.

pp. 467"501.
Bestimmte
Integralemit
Reihen
Die Fourier'schen

(1875),pp.

471

und

264"321.

Multipleand

Definite

H.

HANKEL,

Cylinderfunctionenerster

VIII.

Phys.Soc.

(1843),pp.

xix.

the Calculation

of the Numerical
Values of a certain class of
Integrals(Sept.29, 1857). Phil. Mag. (4) xiv. (1857),pp. 375"382.

Ann.

Proc.

A.
HALL,
Analyst, i. (1874),pp. 81"84.

FluctuatingFunctions

Die

537.

"

211"214.

ROWAN*.
SIR
WILLIAM
HAMILTON,
(June 22, 1840). Trans. R. Irish Acad.

On
On

(1910),pp. 517

Lxix.

B.

zweiter

Art

(Dec. 15, 1868).

Math.

Ami.

(1869),

i.

+. Math. Ann.
viii.
(1875),pp. 453
Cylinderfunctionen
uud IntegralefiirCylinderfunctionen
t (May 16, 1869).

470.

"

Math.

494.

"

P. A.

HANSEN,

Storungen in Ellipsen von


beliebigerExcentricitat und
Ermittelung der absoluten
Neigung, i. Schriftender Sternivarte Seeberg(Gotha,1843). [Memoire sur la determination
d'uue excentricite et d'une inclinaison queldes perturbationsabsolues dans les ellipses
Traduit de TAIlemand
Par M. Hansen.
(Paris,1845).]
conques.
par M. Victor Mauvais
Entwickelung des Pi'oducts einer Potenz des Radius Vectors mit den Sinus oder Cosinus
der wahren
Anomalie
eines Vielfachen
in Reihen.
281.
LeipzigerAhh. ii. (1855),pp. 181
"

HANUMANTA
On

certain definite

C. V.

RAO,
(1918),pp.

integral.Messenger,XLVii.

G.

HARDY,
theorems

General
xxxii.

(1901),pp.

Notes
Trans.
On
pp. 445
On

the

Phil. Soc.

the

xxi.

166.
xxxv.
calculus,xviii. Messenger,
(1906),pp. 158
integral
Theory of Divergent Series and Integrals (May 18, 1908).
"

(1912),pp.

1"48.

IntegralEquation (Feb. 20, 1909).

an

London

Math.

expressedin

terms

Proc.

Soc. (2)

vii.

(1909),

472.

"

certain definite

Messenger,

xxxviii.

certain

On

in

researches

Camb.

applications.QuarterlyJournal,

some

369"384.

pointsin

some

on

Further

137.

"

H.

: with
integration

in contour

134

Notes

definite

values

can

be

of Bessel's functions.

129"132.

integralsconsidered

by Airy

and

Stokes.

Quarterb/ Journal,xli.

226"240.

(1910),pp.

some

on

NMes

whose
integrals

(1909),pp.

on

some

51.
Messenger, XL. (1911),})p. 44
pointsin the integralcalculus,xxvii.
(1913), pp.
Messenger, XLII.
points in the integralcalculus,xxxv.
"

89"93.

On

the

(1915),pp.
On

expressionof

Dirichlet's divisor

pp. 1"25.
Notes

on

number

as

the

sum

of two

QuarterlyJournal, XLVi.

squares.

263"283.

some

Proc. London

problem (April22, 1915).

pointsin

the

calculus,xlvii.
integral

Messenger,xlviii.

is published in Sir
Bessel functions
A letter by Hamilton
on
V65.
i. (Cambridge, 191)7),
pp. 131
ScientificCurreiipoiidence,
These memoirs
died Aug. 29, 1873.
were
t Hankel
composed from
*

Math.

G.

G.

Soc. (2)xv.

(1918),pp.

(1916),
81

Stokes, Memoir

"

88.
and

"

papers.

materials found

among

his

FUNCTIONS

BESSEL

OF

THEORY

766

C. J.

HAEGREA.VE,

RoijalSoc. 1848,
Riccati's

On

Diflferential

of Linear

the Solution

On

pp.

Equation (April4, 1865). QuarterlyJournal,

Asymptotic

an

J. A.
of Math,

HARRIS,
Journal

American

Functions.

Integrationder Differentialgleichung
^2 +
der Math,

Archiv

Coordinaten.

Analysis

T.

Propagation

Math.

integralfor
Ctc

IntegralsI

OOS

esp [x.

of p

IsotropicSpace

Dimensions

T.

function.
cylindrical

Neumann's

[Jahrbuchiiber die

86"90.

H.
in

Soc. (2) 11. (1904),pp. 122"137.

HAYASHI,
definite

the

parabolische

K.

HAVELOCK,
Wave

of
16, 1904). Proc. London

(1912),pp.

420.

Riemann.

Mathematical

"

(1896),pp. 170"199.

HATTENDORF,
See under

391

^=^'^/^'^^^'elliptischeund

Phys. (2) xiv.

und

(1912),pp.

xxxiv.

J. H.

HARTENSTEIN,

On

Mag. (6)

"

Harmonic

On

Phil.

41"62.

(1889),pp.

(March

in Bessel's Series.

Theorem

A.*
HARNACK,
durch
die Fourier-Bessel'schen
Darstellung einer willkiirlichen Function
191
xxxis.
214; Math. Ann.
Berichte,
(1887),
12,
1887).
Leipziger
pp.
(Dec.

Functioneu

On

(1866),pp. 256"258.

vii.

die

Ueber

XXXV.

the

R.

HARGREAVES,
xsxvi.

of

31"54.

and
Diffraction Problem
191"199.
(1918),pp.

Phil. Trans,

Equations (June 10, 1847).

Fortschritte der Math.

xxiii.
Nyt Tidsskrift,

1912, p.

555.]

Tohoku

Math.

COS

p6\

qOdO (Dec. 1920).

Journal,

xx.

107"114.

(1922),pp.

0.

HEAVISIDE,
Electrical

Papers, i., 11. (London, 1892).


(Dec.15, 1892
Operators in PhysicalMathematics

On

(1893),pp. 504"529;

Lll.

Liv.

(1893),pp.

June

8, 1893). Proc. Royal Soc.

105"143.

Mectro7nagneticTheoryf, 11.,iii. (London, 1899, 1912).


H.

HEINE,
Ueber
Journal

die Ziihler und

fUr

Math.

Nenxier

der

(1860),pp.

LVii.

Die Fourier-Besselsche

Function

E.

Naherungswerthe

Kettenbruchen

von

J (Sept.1859).

231"247.

(June,1868).

Journal

fur Math.

(1869),pp.

Lxix.

128

"

141.

Handhuch

der

Kugelfunctionen:Theorie

und

Anwendungen (2 Bande) (Berlin,1878,

1881).
Sm'

la transcendante

E,^. Ann.

C.
HERMITE,
(2) iii. (1870),p.

di Mat.

83

".

Extrait d'une lettre de Monsieur


Ch. Hennite
k Monsieur
Gordan
Paul
Journal fur Math, lxxvi.
(1873),pp. 303"311.
Extrait d'une lettre a M. E. Jahnke
(Nov. 25, 1900). Archiv der Math,
I.

(1901),pp.

(June 9, 1873).
und

Phys.(3)

20"21.

HERTZ,

H.

tjber die Induktion in rotierenden Kugelu. Dissertation,


Berlin,March
We7-ke,I. (Leipzig,
1895), pp. 37
134.]

15, 1880.

\Cles.

"

Uberdas

(3)XXII.
*

elasticher Flatten. Ann. der Physikund


Gleichgewichtschwimmender
449"455.
1895), pp. 288"294.]
{Ges.Werke, I. (Leipzig,

Harnack

t This

Cheinie,

(1884),pp.
died

work

in and

April 3, 1888.

consists

after

of

1894, with

series of articles first published in The


numerous

J See also under Christofl'el.


statement
" This note contains
a
from Poisson's
integral.

Electrician, Nature

and

where
else-

additions.
of Carlini's

formula, which

Hermite

apparently derived

BIBLIOGRAPHY

767

HERZ,

Bemerkungen

Theorie

zur

der

N.
Functionen

Bessel'schen

(1884),col. 17"28.
Note, betreffend die Entwicklung dei- storenden
CVII.
(1884),col. 429"432.
HILB, E.

(Sept.21, 1883).

Astr. Nach.

(Mar. 30, 1884).

Astr. Nach.

cvii.

Ziir Theorie

Kriifte

der

Eiitwicklungenwillkiirlicher Fuiiktionen nach Eigenfunktionen(Sept.


i. (1918),pp. 58"69.
Zeitschrift,
v.
Laplacesche Reihe (March
15, 1919; Nov. 17, 1919). Math. Zdtschrift,

11, 1917). Math.

tjber die

(1919),pp.

17"25

(1920),pp.

; viii.

79"90.

G. J. D.

HILL,
De

radicibus

ubi a, b, c

rationalibus

fiiuctiones

aequationisRiccatianae
c^y + a + by + cy^=0,
rationales ipsiu.s
.v (May 24, 1840).

Journal

fur Math.

(1843),

xxv.

pp. 22"37.
of

Systems

SphericalHarmonics

E. W.*
HOBSON,
(June 11, 1891). Proc. London

Math.

Soc.

(1891),

xxii.

pp. 431"449.
of a certain Surface- Integral,and its application
to the
On the Evaluation
Math.
in Series, of the Potential of Ellipsoids
(Jan. 12, 1893). Fi-oc. London

Expansion,
Soc.

xxiv.

80"96.

(1893),pp.

BessePs
with
On
Functions, and Relations connecting them
Hyper-spherical and
Math. Soc. xxv.
75.
(Dec. 14, 1893). Pi-oc. London
(1894),pp. 49
SphericalHarmonics
On the most
generalsolution of given Degree of Laplace'sEquation (May 9, 1895).
"

Math.

Proc. London
Note on

Soc. xxvi.
(1895),pp. 492
properties of Bessel's Fimctions

some

(1897),pp.

xxviii.

494.

"

(Jan.14, 1897).

Proc. London

of a function
On the representation
by series of Bessel's
Proc. London
Math. Soc. (2) vii. (1909),pp. 359"388.

functions

Ueber

lineare

lii.

(1899),pp.

die Nullstellen

der

veriiuderlichen

einem

mit
Difterentialgleichungen

Ami.

der Math,

und

Parameter

(Dec. 30,

Math.

xxxiii.

340"362.

A.

HURWITZ,
Ueber

Soc.

(Dec. 10, 1908).

L. UND
A. J. W.
HOPE,
SOMMERFELD,
tJber komplexe Integraldarstelluugen
der Zylinderfunktioneu.Archiv
Phys. (3) xviii. (1911), pp. 1"16.
J.
HORN,

1898). Math.

Math.

370"375.

Bessel'schen

Function

(June 2, 1888).

Ann.

(1889),pp. 246"266.
Ueber
II.

die

(1890),pp.

Wurzeln

einiger transcendenten

[Jahrbuch iiher die

25"31.

Gleichungen. Hamburger Mittheilungen^


1890, p. 115.]

Fortschritte der Math.


J.

HYMERS,
Treatise

on

and
Equations.,
Differential

on

the Calculus

of Finite inferences
(Cambridge

1839).
W. VON.
IGNATOWSKY,
tJber ilie Reihenentwicklungen mit Zylinderfunktioneu(May 13, 1911). Archiv der
Math, und Phys. (3) xviii.
(1911),pp. 322"327.
tJber Reihen mit Zylinderfunktioneunach dem Vielfachen des Argumentes (Dec.1913).
Archiv der Math, und Phys. (3) xxill.
(1915),pp. 193"219.
F.

ISELI,
Die

Riccati'sche

schrifteder Math.

Bern,
Gleichung. Dissertation,

Tables
of the Bessel Functi(ms
1904). Manchester Memoirs, XLViii.

XXI.

(1903),pp.
On

Royal

iiber die Fort(42 pp.). [Jahrbxich

J. G.
ISHERWOOD,
for pure imaginary values

(1903

"

4),no.

JACKSON,
Generalised

1909

1909, p. 369.1

forms

of the

series of Bessel

of the argument

(April26,

19.

F. H.
and Legendre. Proc.

Edinburgh

Math.

Soc.

65"72.

Generalised
functions of
Soc. xli. (1905),
pp. 1"28.
*

Legendre

and

See also under

Bessel

(Nov. 16, 1903).

Gegenbauer.

Trans.

Edinburgh

expansion of

Neumann's

of
generalization

FUNCTIONS

BESSEL

OF

THEORY

768

arbitraryfunction

an

in

series of Bessel

Math. Soc. (2) i. (1904),pp. 361"366.


(Nov. 26, 1903). Proc. London
of the Bessel-Function
Generalisation
to
(March 21, 1904). Trans.
a
Theorems
relating
Edinburgh Royal Soc. XLI. (1905),pp. 105"118.
of Lommel
Note on a theorem
(May 13, 1904). Proc. Edinburgh Math. Soc. xxii. (1904),
functions

pp. 80"85.
to Bessel's and
of Basic numbers
Legendre'sfunctions (June 2, 1904;
The application
Math. Soc. (2) ii. (1905),pp. 192"220;
Dec. 1, 1904 ; Jan. IS, 1905). Proc. London
(2) iii.

(1905),pp. 1"20,

Royal

21"23.

complete solution

The

Soc.

xxv.

Theorems

of the diflerential

equationfor J^nj(July4, 1904). Proc. Edinburgh

(1904),pp. 273"276.
relating to a Generalization

Edinburgh Royal

Soc.

xli.

of Bessel's

W.

JACKSON,

prism

integraliumdefinitorum

transformationis

(1836),pp.

of finite

angle (Nov. 11, 1903).

(July 9, 1835).

Werke, vi. (1891),pp.


[Oes.Math.

1"26.

Trans.

C. G. J.*

JACOBI,
XV.

(Feb. 20, 1905).

H.

the diffraction of lightproduced by an opaque


Math. Soc. (2) I. (1904),pp. 393"414.
Proc. London
On

Formula

Function

(1905),pp. 399"408.

86"

Journal

118.]"Formule

filrMath.

pour

la transformation

Journcd
de Math. i. (1836),pp. 195
classe d'integralesdefinies,
196.
einer Berechnung der grossen
Versuch
Ungleichheit des Saturns nach einer strengen
Entwicklung. Astr. Nach.xxxin.
(1849),col. 65"80, 81"94.
[Ges.Math. Werke, vii. (1891),
d'une

"

145"174.]

pp.

sehr entfernter Glieder in der Entwickelung der


tjber die annahernde
Bestimmung
Methode
mung
BestimCoordinaten, nebst einer Ausdehnung der Za/3^ace'schen
zur
elliptischen
Astr. Xach. xxviii.
Zahlen.
der Functionen
270.
Math.
(1849),col. 257
grosser
[6-'es.
Werke,vii. (1891),pp. 175"188.]
"

P. R. E.t
JAHNKE,
Von E. Jahnke
und F. Emde
Funktionentafeln mit Formeln und Kurven.
(Leipzig,
1909).
in der elektromagnetischenStrahlungstheorieauftretende,
tJber einige,
bestimmte
In267.
tegrale.Archiv der Math, imd Phys. (3) xxiii. (1914),pp. 264
"

E.

JAMET,
Sur

les equations

228.
(Ajaccio)(1901),pp. 207
les equations anharmoniques. Ann.

XXX.
gaise,

Sur

V.

anharmoniques (Sept.13, 1901). Comptes Rendus

de I'Assoc. Fran-

"

de

la Fac.

des Sci. de

Marseille,xii. (1902),

pp. 1"21.

JEKHOWSKY,

B.

Les fonctions de Bessel de


d'une variable (Feb. 28, 1916).
Sur

la fonction

Sci. math.
Sur

plusieursvariables exprimees par des fonctions de Bessel


Comptes Rendus, CLXii. (1916),pp. 318"319.
generatrice des fonctions de Bessel k plusieursvariables. Bulletin des

(2) xli.

(1917),pp.

les fonctions

(1921),pp.

58"60.

de Bessel

deux

variables

JOHNSON,
On

(May 30, 1921). ComptesRendus,

CLXXii,

1331"1332.

the Differential

W.

W.

Equation

g+y2+Py+(2=o.
Annals

of Math.

iii.

(1887),pp.

112"115.

JOLLIFFE,
The

expansion of the

functions.

Messenger,XLV.

of

square

(1916),p.

Bessel

XXVIII.

les fonctions

de

in the form

of

series of Bessel

deuxifeme

V.

A.

espfece(Dec. 1893).

Archives

Neerlandaises,

Archives

Neerlandaises,

(1895),pp. 221"225.

Sur les ondes


XXVIII.

de Bessel

E.

16.

JULIUS,
Sur

A.

function

lumineuses

spheriqueset cylindriques(Dec. 1893).

(1895),pp. 226"244.
*

See also under

Carlini.

See

also under

Hermite.

769

BIBLIOGRAPHY

A.

KALAHXK,

Tiber die Wurzeln


mid gewisser aus
ihnen gebildeter
GleicheinigerZylinderfuuktionen
filrMath, unci I'hys.liv. (1907),pp. 55"86.
ungen (March 1, 1906). Zeitsclirift
W.*

KAPTEYN,
Nouvelles

formules

representer la Ibnctiou

pour

Bulletin
J,^_-^{x).

des Sci. Math.

(2)

(1892),pp. 41"44.

XVI.

Over BESSEL'sche
Functien
(Marcli12, 1892). Nieuw
pp. 116"127.
Recherches
les fonctions de Fourier- Bessel. Ann.
sur

(1893),pp.

Archiefvoor

xx.
Jfis/fcurtrfe,
(1893),

sci. de I'Ecole

norm.

(3)

sup.

x.

91"120.

Sur quelques integralesdefinies con tenant


de" fonctions de Bessel.
Archives Neerlandaises,
(2) vi. (1901),pp. 103"116.
A definite integral
containingBessel'.s functions (June 29, 1901). Pi-oc. Section of Sci.,
K. Akcid.
Sur un

Wet.

van

te

Amsterdam,

developpement de

M.

102

(1902),pp.

iv.

Neumann.

Nieuw

103.

"

Archiefvoor

Wiskunde, (2)

vi.

(1905),

und

Phys.

pp. 49"55.
iiber Bessel'sche

Einige Bemerkungen
(1903),pp.

XIV.

Functionen.

MonatsheftefilrMath,

275"282.

connected
definite integrals
The values of some
with Bessel functions (Nov.
Proc. Section of Sci.,K. Akad.
Wet. te Amsterdam, vii. (1905),pp. 375
376.
van
On a series of Bessel functions (Dec. 24, 1904). Proc. Section of Sci.,
K. Akad.

26, 1904).

"

te

Amsterdam,
definite

Wet. te
On

(1905),pp.

vii.

integralof

Amsterdam,
an

Sur

"

(Sept.30, 1905). Proc. Section of Sci.,K. Akad.

Kummer

an

Wet.

van

500.

(1906),pp.

viii.

expansion of

(1906),pp.

XXXV.

494

350

van

357.

"

arbitraryfuiictiou

in

series of Bessel

functions.

Messenger,

122"125.

la sommation

d'une

serie iufinie. Nieuw

Archiefvoor

Wiskunde, (2) vii. (1907),pp.

20"25.

quotientof two successive Bessel Functions (Dec. 30, 1905 ; Jan. 27, 1906). Proc.
Wet. te Amsterdam, viii. (1906),pp. 547"549, 640
642.
of Sci.,K. Akad. van
Sur le quotientde deux fonctions besseliennes
successives. Archives Ne'erla ndaises,
(2)
(1906),pp. 149"168.
Me'm. de la Soc. R. des Sci. de Liege,(3) vi.
Recherches
les fonctions cylindriques.
sur

The
Section

XI.

"

(1906),no.
Sur

5.

le calcul

numerique

de la serie
1
2
.

(a2+ ^V)2^
Mem.
On
Akad.

de la Soc. It. des Sci. de Liege,(3) vi. (1906),no. 9.


relations between
Bessel's Functions
some
(Feb.24,
Wet. te Amsterdam, xiv.
969.
van
(1912),pp. 962

K,
1912). Proc. Section of Sci.,

"

LORD.
KELVIN,
of any depth or in a dispersive
single impulse in water
(Feb. 3, 1887). Proc. Royat Soc. xui.
(1887),p. 80; Phil. Mag. (5) xxiil. (1887),
and Phys. Papers,iv. (1910),pp. 303"306]
[3fath.
pp. 252"255.
and Ponderable
Matter
(Jan. 10, 1889).Journal of Inst, of Electrical
Ether, Electricity
and Phys. Papers,III. (1890),pp. 484
5 15. J
Engineers,xviii. (1890),pp. 4 37. \_Math.
On
medium

the

waves

produced by

"

"

S.

KEPINSKI,
tjber die Differentialgleichung
/"

3^2

m-\-\

dx^

dz

i^^z

ex

_r^

(Jan. 1905). Math.

Ann.

lxi.

(1906),pp.

8"

397"406.

Integrationder Ditterentialgleichung

a^i_i (^_ ''"


Bull. int. de VAcad.

des Sci. de

^ ct
^e
Cracovie,
1905, pp.
*

W.

B. F.

See also under

198

"

205.

Gegeubauer.
49

FUNCTIONS

BESSEL

OF

THEORY

770

G.

KIRCHHOFF,

(June,1853).

eines

Journal

(1854), pp.

Theorie

Zur

Magnetismus
filrMath, xlviii.

Ueber

oder unterirdischen
Teleder Elektricitait in unterseeischen
611.
Berliner Monatsherichte,1877, pp. 598
versinderlichem
von
Querschnitt
Transversalschwingungen eines Stabes
der

Bewegung

"

1879, pp.
(Oct.27, 1879). Berliner Monatsberichte,
(3)X. (1880),pp. 501"512.
Stroiaungen in einem

die elektrischen

Ueber

pp. 519

1883,
Sitzimgsberichte,

Amsterdam,
Wet. te

341

(1906),pp.
of

integraltheorem

An
van

viir.

Amsterdam,

"

Physik U7id Chemie,


Berliner

J. C.

of Sci.,K.

Froc. Section

Ahad.

Wet.

van

350.

Gegenbauer (Feb. 27, 1909).


(1909),pp. 749

xi.

der

Ann.

828;

"

Kreiscylinder(April26, 1883).

KLUYVER,
te

815

524.

"

problem (Sept.30, 1905).


probability

local

Eisen

348"376.

(Oct. 29, 1877).


graphendrahten
die

weichem

unbegreuzten Cylindersvon

inducirten

den

Ueber

Section

Proc.

of Sci.,K.

Akad.

755.

"

J. C. C. A.

KNESER,

Funktionen
in Reihen die nach Bessel'schen
Entwickluug willkiii'licherFuuktionen
fortschreiteu (July,1903). Archiv der Math, und Phys. (3) vii. (1903), pp. 123"133.
und die Darstellung willkiirlicher Funktionen
der Integralgleichungen
Die Theorie
524.
der mathematischen
Physik. Math. Ann. LXiii. (1907),pp. 477
Die

in

"

K.

KNOCKENHAUER,
die Oerter

Ueber

Beobachtungeu.

Ann.

J.

KONIG,
Ann.

Darstellung
(1872),pp. 310"340.

V.

von

Functionen

durch

unendliche

Bessel'schen

Reiheneutwicklung nach

Ueber

Fresnel'schen

"

die

Ueber

W.

des gebeugten Lichtes nach den


und Minima
110.
der Physik und Chemie, (2) xli. (1837),pp. 103

der Maxima

Reihen

(Sept.1871). Math.

(1880). Math.

Functionen

Ann.

xvii.

(1880),pp. 85"86.
W.
KOESTLER,
Beitragezu Reiheneutwickelungennach Besselsche Zylinderfunktionen.Dissertation,
Bern, 1907 (110 pp.). \Jahrbuch itber die Fortschritte der Math. 1908, p. 535.]

KOPPE,
an
Ausbreitung einer Erschtitterung

Die

Grenzfall

neuen

der Bessel'schen

M.
der Wellenmaschine

Functionen.

Programm

(28 pp.). [Jahrbuchuber die Fortschritte der Math.

1899

KUMMER,
de I'equation
de
I'integration

Sur
XII.

Riecati

E.
par des

darstellhar

(96). Andreas
1899, pp. 420
"

durch

einen

Realgymn. Berlin,

421.]

E.
definies.
integrales

Journal

fur Math.

(1834),pp. 144"147.
Ueber

die

hypergeometrischeReihe
a(a + l)^(/3+ l)
a(a+l)(a
a^
'^Ly""^ 1.2.y(y+l) "^ ^
1 2 3
Journal fur Math. xv. (1836),pp. 39"83, 127"172.
De integralibus
quibusdam detinitis et seriebus
Math. XVII.
(1837),pp. 228"242.
.

Note

sur

+
.

2)^(/3+l)(/3-^2)
''^'^"'
(y+ 1) (y+ 2)

infinitis

Journal
I'integrationde Tequation --j-^"x"Ky.

(April,1837).

fiirMath.

Journal

fur

(1839), pp.

xix.

286"288.
d^^v
Sur

de I'equation
I'integration
-r^

Sur

le

On
XIII.

the

On

de Math.

iv.

of

Viscous

H.
LAMB,
Spheroid (Nov. 10, 1881).

(1839),pp.

des Sci. de

Proc.

390"391.

Berlin,xxv.

London

Math.

Soc.

51"66.

the Vibrations

(1882),pp.

Journal

de

Oscillations

(1882),pp.

.y.

J. L. DE.
LAGRANGE,
Kepler (Nov. 1, 1770). Eist. de V Acad. R.
204"233.
in.
[Oeuvres,
(1869),pp. 113"138.]

probl5me

(1769) [1771],
pp.

x^

189-212.

of

an

Elastic

Sphere (May 11, 1882).

Proc.

London

Math.

Soc.

xill.

OF

THEORY

772

FUNCTIONS

BESSEL

P.
LINDNER,
mit
komplexenZeigerzu den Besselschen
der
Ableitungen
Die Beziehungeu
begrenzten
und ihren Nullstellen (Nov.29, 1911). Sitz.der Berliner Math. Ges. xi. (1911),
Funktionen
pp. 3"5.
A.
LINDSTEDT,
Ann. der Physikund
1882).
(
Aug.
Integrale

Zur Theorie der Fresnel'scheu


XVII.
(1882),
pp. 720"725.

Chemie,(3)

J.
LIOUVILLE,
des transcendantes et sur I'impossibilite
d'exprimerdes racines de
Sur la classification
des coefficients (June 8, 1835). Journal
fonctious finies explicites
en
certaines equations
de Math. II. (1837),pp. 56"105; iii. (1838),pp. 523"547.
d'une classe d'Equationsdifierentiellesdu second ordre en quantites
Sur I'integration
de
; Journal
finies explicites
(Oct.28, 1839). Comptes Eendus, ix. (1839),pp. 527"530
Math. IV. (1839),
pp. 423"456.
de Riccati (Nov. 9, 1840). Comptes Rendus, xi.
Remarques nouvelles sur I'equation

(1840),p. 729; Journal de Math.


Sur

I cos
I'intdgrale

Sur

une

i {u

vi.

(1841),
pp.

sin u) du.

.v

fur Math.

LVi.

der Differentialgleichung
Integral
^ +

vi.

(1841),pp. 69"73.

189"196.
R.

LOBATTO,
Sur

de Math.

36.

^4-/=0 (July14, 1858).Journal

-"

(1859),
pp.

(1841),
p.

vi.

R. 0. S.

LIPSCHITZ,
ein

de Math.

Journal

(March,1841). Journal

formule de M. Jacobi

Ueber

1"13.

des equations
I'integration
d^

d^

^
d.V'
(April,
1837).Journal fiirMath. xvii.

Note on the
pp. 494"498.

'

i/

dx^

(1837),
pp.

363"371.

LODGE, A.
SemiconvergentSeries of J" {.v).British

E. C. J.
LOMMEL,
Beitragezur Theorie der Beugung des Lichts.
(1861),
pp. 385"419.
Methode
ziuBerechnung einer Transcendenten.

(1861),pp.

Association

VON.
Archiv

der

Archiv

Report,York,1906,

Math,

der Math

und
und

Phys. xxxvi.
Phys. xxxvii.

349"360.

linearer Diflferentialgleichungen.
Archiv
Zur Integration
101"126.
(1863),
pp.
Studienilher die BesseV schen Functionen (Leipzig,
1868).
der Gleichung^"*
Integration
Ueber der Auwendung der
Math, und Phys.xv.
schriftfUr

0*.
.^^^^-^
+y
=

Math.

Bessel'seheu Functionen

(1870),pp.

Ann.

der

ii.

Math,

und

Phys. xl.

(1870),pp. 624"63.5.

in der Theorie

der

Beugung.

Zeit-

141"169.

Zur Theorie der Bessel'schen Functionen (Dec.1870). Math. Ann. iii. (1871),
pp. 475
487.
Zur Theorie der Bessel'schen Functionen
(Jan. 1871).Math. Ann. iv. (1871),pp. 103
116.
Ueber eine mit den Bessel'schen Functionen
verwandte
Function (Aug. 1875). Math.
"

"

Ann.

IX. (1876),
pp. 425"444.
Zur Theorie der Bessel'schen Functionen

(Oct.1878). Math.

Ann.

xiv.

(1879),pp. 510

"

(Oct.1879). Math.

Ann.

xvi.

(1880),pp.

"

536.

Zur Theorie der Bessel'schen Functionen

183

208.

Die Beugung-serscheinungen
einer kreisruuden Oeffuungund eines kreisrunden Schirmchens theoretisch und experimentell
bearbeitet. Miinchener Ahh. xv. (1884"86),pp. 233
328.
Die
Miinchener
Abh. xv.
Beugiuigserscheinungen
geradlinigbegrenzterSchirme.
"

(1884"86),pp.
*

The

531"663.

headmgs

of the pages

of tins paper

are

"

Zur

Theorie

der Bessel'schen

Functionen."

773

BIBLIOGRAPHY

LORENZ,
Sur

L.

le

des fonctions arbitraires au


de fonctions donnees.
Ti(hdevelopjiement
moyen
skriftfor Mathematik,1876, pp. 12!)" 144. [OeuvresScientifiques,
ii. (1899),pp. 495"513.]
Tlieorie de la dispersion.
Ann. der Physikund Chemie, (3)xx. (1883),
21. \Oeuvres
pp. 1
"

i. (1898),pp. 371"396.]
Scientijiques,

Sur la lurniere refiechie et refractee


par line
herncs Selskabs Skrifter,
(6)vi. (1890),pp. 1"62.

sphere transparente.

Videnska-

K. Danske

i. (1898),pp.
[Oeuvres
Scientifiques,

405"

A. E. H.
LOVE,
Tlie Scatteringof Electric Waves
by a Dielectric Sphere (Feb.9, 1899). Proc. London
Math. Sac. xxx.
(1899),pp. 308"321.
The
Transmission
the Surface
of the Earth
of Electric Waves
over
(Dec. 19, 1914).
Phil
Trans, of the Royal Soc. ccxv.
A (1915),
pp. 105"131.

H.

MACDONALD,
Note

Bessel

on

Functions

pp. 110"115.
Zeroes of the Bessel

London

Math.

Soc.

(1898),

xxix.

Functions

Math. Soc.
(April 7, 1898 ; Jan. 12, 1899). Proc. London
(1899),pp. 165"179.
for the Bessel Functions
Math.
(April5, 1900). Proc. London

(1898),pp. 575"584;

XXIX.

M.*

Proc.

(Nov. 11, 1897).

xxx.

The Addition
Theorem
Soc. XXXII.
(1901),pp. 152"157.
Some
The

Fourier's

Applicationsof

(1903),pp.

XXXV.

Bending

of Electric

1903). Proc. Royal Soc.


Note

on

Proc. London

Theorem

I'Dec.

11, 1902). Proc. London

Soc.

Math.

428"443.

Waves

round

Conducting Obstacle

(1903),pp. 251"258;

lxxi.

Lxxii.

(Jan. 21, 1903


(1904),pp. 59"68.

the evaluation of a certain integral


containingBessel's functions
Math. Soc. (2) vil. (1909),pp. 142"149.

May 12,

(Dec.6, 1908).

The Diffraction of Electric Waves


Obstacle (Aug. 13, 1909).
round a Perfectly
Reflecting
Phd.
Trans, of the Royal Soc. ccx.
A (1910),pp. 113"144.
The Transmission
around
the Earth's Surface (Jan. 10, 1914). P7-oc.
of Electric Waves
Soc.

Roy id

Formulae

xc.

(1914),pp.

Proc. London

1914).

Math.

[Extractfrom a
(1914),pp. 239"240.
A

50"61.

Harmonic
Sjjherical

for the

letter to

class of diffraction

when
l-/i is a small quantity(Feb.6,
/'"~"'(/li)
(2) Xlil. (1914),pp. 220"221.
Prof. Carslaw, Oct. 17, 1912.] Proc. London
Math. Soc. (2)xiii.

Soc.

Proc. London

problems (Jan. 14, 1915).

pp. 410"427.
The Transmission
of Electric Waves
Royal Soc. xcii. A (1916),pp. 433"437.

around

the Earth's

the

(1894),pp.
On

Descending Series

of Both

for Bessel's Functions

of the Bessel

the Roots

requind

(2)xiv.

(1915),
P7'oc.

(May 23, 1916).

Kinds.

of Math.

Anncds

viii.

57"61.

pp. 23"30.
for
The Ex[)ression
as

Surface

Soc.

J.

McMAHON,
On

Math.

in certain

and

Rational

Cases

certain

Related

Polynomial

in

Functions.

Annals

of Math.

Series of Bessel functions


Proc. American
Assoc.

of Dirichlet's Problem.

ix.

(1895),

of nth. Order

1900, 2)p. 42

"

43.

The

Modified

Function

Bessel

T. M.
MacROBERT,
A'" (2) (Feb. 13, 1920). Proc.

Edinburgh

(1920),pp. 10"19.
and the Fourier-Bessel
Asymptotic Expressions for the Bessel Functions
(1921),pp. 13"20.
(Jai".^4,1921). Proc. Edinburgh Math. Soc. xxxix.

Math.

Soc.

xxxviii.

G.

MAGGI,
Sulla storia delle funzioni cilindriche.
vol.

IV.

Expansion

A.

Atti della R. Accad.

dei

Lincei,Ser.

(Transunti),

(1880),pp. 259"263.

Sopra
(1880),pp.

un

di elettrostatica

problema

(June 3, 1880).

Rend,

del R. 1st. Lombardo^ (2)xm.

384"390.

MALMSTEN,

/
0

COS
"

C. J.

(JjVCajTj

^Jr^
(l+A'T

K. Svenska

"

V. Akad.

See also under

Handlingar, LXii.

Gegeubauer.

(1841),pp.

65

"

74.

diffdrentielle
I'equation

De

(1850),pp.

XXXIX.

Theorfemes

sur

FUNCTIONS

BESSEL

OF

THEORY

774

y^"+"^
+ ^^'",y

(March 18, 1849).

Journal

fiirMath.

108"115.

de I'Equation
-j\+
I'integration

(^^''^+

)V-

T2

Camh.

and

Dublin

Journal, v. (1850),
pp. 180"182.

Math.

C.
certaiues fonctions

MALTEZOS,
Sur la chute des corps dans le vide et sur
Annales de Math. (4)xi. (1902),
pp. 197"204.

pnmi gradus (Bologna,1707).


aequationumdifferentialium

tJber die Ausbreitung


1911). Ann. der Physikund

W.

H.

MAECH,

(1912),pp.

Chemie, (4) xxxvii.

ciliudriche di

funzioni

suUe

MAESHALL,
Method

New

of

prima specie(April6, 1889). Napoli Ren-

96"99.

diconto,(2) iii. (1889),pp.


On

Computing

of Math. (2) XI. (1910),pp.

the Roots

W.
of Bessel's Functions

Annals

(Oct.1909).

153"160.

G.

MATHEWS,
See under

50.

"

E.

MAECOLONGO,
teoremi

29

Erdkugel (Oct. 21,

der

Telegraphieauf

der drahtloseu

der Wellen

Alcuui

Nouvelles

G.

MANFREDIUS,
eonstructione

De

transcendantes.

B.

Gray.
of

the Focus

Eiccati's

Equation.

D.

Telescope(Feb.22, 1897). Proc.

L. W.
Math.

MEECH,

Integrationof

H.

E.

MAYALL,
the Diffraction Pattern near
Phil. Sac. IX. (1898),pp. 259"269.
On

of

Annals

i.

(1886),pp.

97

"

103

; lii.

Camb.

(1887),

pp. 47"49.
F. G.

MEHLEE,

Vertheilungder statischen Electricitat in einem von zwei kugelkalottenbe150.


grenzten Korjier(July,1867). Journal fiirMath, lxviii. (1868),pp. 134
imd ihre AnUeber eine mit den Kugel- und Cylinderfunctionen
verwandte
Function
die
der
in
Theorie
Jahreshericht,
wendung
Electi-icitatsvertheilung.
\_Elhing
1870.]Math.
die

Ueber

"

Ann.

(1881),pp.

xviii.

Ueber
function*
Notiz
und

die

v.

willkiirlicheu

Function

Variablen

zweier

Cylinder-

durch

(Dec. 1, 1871). Math.

Ann. v. (1872),pp. 135"140.


iiber die Dirichlet'schen
fiir die
IntegKilausdrticke

iiber eine

Ann.

161"194.

Darstellungeiner

KugelfunctionenP" (cos5)
analogelutegralformfiir die Cylinderfunction
J{x)\ (Dec. 2, 1871). Math.

(1872),pp.

141"144.

D.

MEISSEL,
Iserlohn

Programm,

1862.

F. E.

JVoiovellesAnnales
[Nielsen,

de Math.

(4) ii. (1902),pp. 396

"

410.]
Tafel der Bessel'schen
Functionen
Berliner Abh. 1888.
{Math. Abh. i.)
Ueber
die Besselscheu
Functionen

[Jahrbuchiiber die Fortschritte


EinigeEntwickeluugen die
Nach.

cxxvii.

Beitragzur

JfP und

^-=0

Ij} von

bis ^"=15-5

(Nov. 8, 1888).

Kiel, 1890.
Jj.^.Programm, Oberrealschitle,

der Math. 1890, pp. 521


Bessel'schen
/-Functionen
"

522.]
betreffend

(May 3, 1891).

Astr.

(1891),col. 359"362.
Theorie

(1891),col.

cxxviii.

/^.ound

der

allgemeinenBessel'schen

Function

(June 11, 1891).

Astr. Nach.

145"154.

AbgekiirzteTafel der Bessel'schen Functionen


I^^ (July 10, 1891). Astr.
(1891),col. 154"155.
Beitrag zur Theorie der Bessel'schen Functionen
(Oct.7, 1891). Astr.

Nach.

cxxviii.

Na^h.

cxxviii.

(1891),col. 435"438.
*

The

t The

headings of the pages


headingsof the pages

of this paper
of this paper

are
are

"Ueber
"Notiz

die

Cyhnderfunction J (x)."
und
P" (cos3^)

iiber die Functionen

(a;)."

775

BIBLIOGRAPHY

Neue

Entwickelungen
(1892),col.

cxxix.

Weitere

Entwickelungen

(1892),col.

cxxx.

Uber

iiber die Bessel'schen


iiber die Bessel'schen
der

Maxima

Bessel'schon

MELLIN,
einer einheitlichen

(June, 1909).

Math.

Ann.

Theorie

Resistance

of

des Sci. de

(1876),pp. 167

"

MOORE,
the roots

on

The

of Bessel

summabilit}^of the

1908). Trans. American


On

the

hypergeometriscbenFunktionen

J. H.

Soc.

Math.

convergence
Math. Soc. xii.

(1898),pp. 106"123.

Mag. (5)xlv.

H.

-j^

(March 6, 1876). Me'm. de V Acad,


ax^"'j/

189.

C. N.

Annals

developments

rhil.

of Math.

in Bessel

(2) ix. (1908),pp.

functions

with

On

(1909),pp. 391"435.
developments in Bes.sel functions
(1911),pp. 181"206.
x.

of the

(Oct.30, 1909).

function whose
development in Bessel's functions is non-summable
Math. Soc. xxiv.
(1918),pp. 145
(Sept.4, 1916). Bulletin American
summability of the developments in Bessel's functions (Sept. 4, 1917).
Math. Soc. xxi.
156.
(1920),pp. 107

the

"

of
149.
Trans.

"

The

Value

of the

1900). iVature,LXiii.

MORTON,
Cylinder Function of the
(1901),p.

W.

B.

Second

for Small

Kind

Arguments (Oct.25,

29.

R.
MURPHY,
general propertiesof definite integrals
(May 24, 1830). Trans.
(1830),pp. 429"443.
the

Camb.

Phil. Soc.

W.

MYLLER-LEBEDEFF,
See under

LebedefF.

NAGAOKA,
DiflFraction Phenomena
Coll.

162.

"

applications
(Sept.11,

orders

American

On

156

continuous

certain

in.

functions.

uniform

Tra7is. American

der

und

Ship* (Aug. 9, 1897).

Toulouse,(7) viii.

478.]

"

(1910),pp. 305"337.

Lxvnr.

difF^rentielle
de I'equation
I'integration

Note

Proyramm, Oberrealschule,
1892, pp. 476

HJ.

R.

MOLINS,
Sur

Nach.

Axtr.

(May 2, 1892).

Functionen.

der Gamma

MICHELL,
Wave

The

Functionen

(11 pp.). [Jahrbuch iiber die Fortschritte der Math.

Abriss

Astr. Nach.

(Jan. 25, 1892).

363"368.

die Absoluten

Kiel, 1892

Functionen

281"284.

ofSci.Imp.

Univ.

H.

produced by an Aperture on
Japan, iv. (1891),pp. 301"322.
NEUMANN,

CARL

Curved

GOTTFRIED

Surface.

Joiwnal

of the

t.

eines homogenen
AllgemeineLbsung des Prohl ems iiberden station'dren Tempcratxirzustand
welcher von
ziuei nicht concentrischen Kugelfi'dchen
begremt vnrd (Halle,1862).
Kiirpers,
das Gleichgewichtder Warme
und das der Electricitiit in einem
Ueber
Korper, welcher
zwei licht concentrischen
von
Kugelfliichen
begreuztwird (1862). Journal fiirMath. Lxn.

(1863), pp.

36"49.

nach den Besselscheu Functionen


gegebenerFunctionen
Entwickelung beliebig
(1867),pp. 310"314.
(March 28, 1867). Journal fiirMath, lxvii.
Theorie der Kugelfunctionen
Ein Analogon zur
Functionen.
Theorie der BesseVschen
1867).
(Leipzig,
nach Quadrateu und Prothicten der FourierUeber die Entwicklungen einer Function
256.
Functionen
Bessel'schen
{Math.
(Nov. 1869.) LeipzigerBerichte,xxi. (1869),pp. 221
Ann. III. (1871),pp. 581-610.]
Math. Ann. ii. (1870),
Ueber
und Quadrate der Bessel'schen Functionen.
Producte
p. 192.
iiberdie nach Kreis-,Kugel- und Cylinder
Entwickelungen
functionenfortschreitenden
1881).
(Leipzig,
iiber
Ai^= F insbesondere
Ueber
Integraleder Diflerentialgleichung
gewisseparticulare
1886).
die Entwickelung dieser particularenIntegralenach
8,
(
March
K\igelfunctionen
LeipzigerBerichte,xxxvili.
(1886),pp. 75 82.
Ueber

die

"

"

This

paper

contains

t See also under

some

Schlafli.

Tables

of Bessel

functions

wliich

were

computed by

B. A. Smith.

OF

THEORY

776

FUNCTIONS

BESSEL

F. E.

NEUMANN,

Halberstadt,1909 (25 pp.).


Dissertation,
Verallgemeinerungder Zylinderfunktionen.
Math.
der
1909, p. 575.]
[Jahrhuchiiber die Fortschritte
Eiiie

J. W.*

NICHOLSON,

Asymptotic Expansions

The

of Bessel

Functions

of

High

Order.

Phil.

Mag. (6) xiv.

(1907),pp. 697"707.
Functions

Bessel

On

pp. 271"279.
On the Inductance

Equal Ai-gument

of

Parallel Wires

of Two

and

Order.

Phil.

(June 12, 1908).

Mag. (6)

xvi.

(1908),

Phil.

Mag. (6) xvii.

(1909),

Phil.

Mag. (6)xviir.

(1909),

pp. 255"275.

Airy'sIntegralto

of

the Relation

On

the Bessel

Functions.

17.
pp. 6
Phil. Mag. (6)xix.
The Asymptotic Expansions of Bessel Functions.
round a Large Sphere,i. Phil.
of Electric Waves
On the^Bending
"

516

pp.

537.

"

The

(1910),pp. 228"249.
(1910),

Mag. (6) xix.

Approximate

Calculation

of Bessel

Functions

of

Imaginary Argument.

Phil.

Mag.

(1910),pp. 938"943.

(6) XX.

The Scattering of Light by a Large Conducting Sphere (March 10, 1910). Proc. London
Math. Soc. (2) IX. (1911),pp. 67"80.
Notes on Bessel Functions.
Journal, xlii. (1911),pp. 216"224,
Quarterlt/
The productsof Bessel functions.
QuarterlyJournal, XLlii. (1912),pp. 78"100.
Math.
obstacle (Dec. 14, 1911). Proc. London
The pressure of radiation on a cylindrical

(2) XI. (1913),pp. 104"126.

Soc.

The

Vibrations

Lateral
A

xciii.

of

Generalisation
Problem

of Variable

of Bars

Section

(1922),pp.

in the

theorem

Soc.

Royal

329.
due to Sonine. Quarterly
Journal, XLViii. (1920),pp. 321
Conduction
(Sept.29, 1921). Proc. Royal Soc. c. A
"

Theory

of Heat

NICOLAS,
fonctions

(2) xi.

sup.

Sur

Proc.

226"240.

des

Etude
norm.

(May 11, 1917).

(1917),pp. 506"519.

(premiere et
(1882),supplement,pp. 3 90.

J.
deuxifeme

sci. de

Ann.
espfece).

,^

I'Ecole

"

produitde

le

Fourier

de

deux

fonctions

N.t
NIELSEN,
cylindriques(July 30, 1898).

Math.

Ann.

Lii.

(1899),

pp. 228"242.
ix.
Udviklingerefter Cylinderfunktioner
(Aug. 28, 1898). Nyt Tidsskrift,

(1898),pp.

73"83.
Sur

le

developpement

de

zero

fonctions

en

cylindriques(March 25, 1899).

Math.

Ann.

(1899),pp. 582"587.

Lll.

x. B (1899),pp. 73"81.
FlertydigeUdviklingerefter Cylinderfunktioner.
Nyt Tidsskrift,
Note sur
les developpementsschloemilchiens
serie de fonctions cylindriques
en
(Nov. 15,
Videnskabernes
665.
1899). OversigtK. Danske
1899, pp. 661
Selskabs,
Note supplementaire relative aux
developpements schloemilchiens en .soriede fonctions
60.
K. Danske
Videnskabernes
(June 23, 1900). Oversigt
Selskabs, 1900, pp. 55
cylindriques
Sur une
classe de polynomes qui se presentent dans la theorie des fonctions
driques
cylin(Feb.26, 1900; Jan. 8, 1901). Ann. di Mat. (3)v. (1901),pp. 17"31
; (3)vi. (1901),
"

"

pp. 331"340.
Evaluation

nouvelle des integrales


foncindefinies et des series infinies contenant
une
cyUndrique (May 23, 1900). Ann. di Mat. (3) vi. (1901),pp. 43"115.
Sur une
classe de series infinies analogues a celles de Schlomilch
selon les fonctions
(Aug. 17, 1900). Ann. di Mat. (3) vi. (1901),pp. 301"329.
cylindriques
Recherche
les series,
des fonctions
C. Neumann
et
sur
cylindriquesdues k MM.
W. Kapteyn. An7i. sci.de I'Ecole norm.
75.
(1901),pp. 39
sup. (3) xviii.
Note sur
la convergence
d'une serie neumannienne
de fonctions cylindriques(Feb. 10,
tion

"

1901). Math.
Recherches

Ann.

Lv.

(1902),pp.
classe

493"496.

de

series infinies analogue a celle de M. W.


Kapteyn
Videnskabernes
146.
(April13, 1901). OversigtK. Danske
Selskabs,1901, pp. 127
Sur les series defactorielles(
160.
Jan. 20,1902). Comptes i2e?i(^?w,cxxxiv.(1902),pp.
157"
Theorie nouvelle des series asymptotiques obtenues
les fonctions cylindriques
et
pour
Videnskabernes
K. Danske
Selskabs,
pour les fonctions analogues(March 16, 1902). Oversigt
sur

une

"

1902, pp. 117"177.


*

See also under

Datta

and

under

Dandy.

f See also under

Souine.

777

BIBLIOGRAPHY

Equations differentielles lineaires obtenues pour le produit de deux


driques. Nouvelles Annales de Math. (4) ii. (1902),pp. 39G" 410.
Handhuch
der Theorie der Cylinderfwiktionen
1904).
(Leipzig,
Note sur
las series de fonctions bernoulliennes
(Dec.14, 1903). Math.

fonctions

Ann.

cylin-

(1904),

lix.

pp. 103"109.
102.
une
iutegraledefinie (June 5, 1904). Math. Ann. LIX. (1904),i^p. 89
quelques proprietesnouvelles des fonctions cylindriques(April 1, 1906). Atti

Sur

"

Sur

delta

R. Accad.

del Lincei,(5) xv. (1906),pp. 490"497.


d'Abel (Sept.17,1906).
Recherches
d'une identity integrale
sur
quelquesgeneralisations
Danske
Videnskabernes
Selskabs Hkrifter.,
(7) v. (1910),pp. 1 37.
146.
Sur les series des fonctions cylindriques.
Journ. fiirMath, cxxxii.
(1907),i^p. 127

K.

"

"

Sur
di Mat.

spheriques(Dec. 20, 1906).

des fonctions

quelques proprietesfondanientales

(1908),pp.
(3)
tjber dieVerallgemeiuerungeinigervon F. und C. Neumann
Reihenentwickelungen
Zylinderfunktionen fortschreitenden
Berichte,LXi. (1909),pp. 33"61.
Uber

das

Produkt

Schwingungen
Zeitschrift
fiirMath. xxv.

einer

Saite

deren

zur

Differentialgleichungen.
Zeitschrift
fiir

des

Math.

Cues.

iiber die

und

Kugel-

Leop. {Halle),1888, pp.

eine

stetigeFunction

der

der

Zeit

ist.

Bessel'schen

(1880),pp. 65"71.
R.

Cyliuderfunctionen(June 2, 1886).

Noca

Acta

Acad.

1"48.

G.

OLTRAMARE,
Note

xix.

Sitzungsberichte,

allgemeineu Integrals

xxv.

OLBRICHT,
Studien

Phys.

F.

Spannung

(1880),pp. 44"48.
numerischen
Berechnung

und

Munchener

(July4, 1908).

NIEMOLLER,
Ueber

gegebenen nach Kugel- und


(Jan. 26, 1908). Leipziger

Zylinderfunktionen.MonatsheftefiirMath,

zweier

(1908),pp. 164"170.
tJber den Legendre-Besselschen
Kettenbruch
xxxviii.
(1908),pp. 85"88.

Formeln

Ann.

69"90.

xiv.

Tintegrale
'O'

sur

/.
0

ComptesRendus

de V Assoc.

cos

yx

ax.

{d^+ b'^x^f
(1895),part

xxiv.
Frangaise,

i.

p. 182; part

ii.

pp. 167-171.

ONO,
On

the First Root

(6)XLli.

(1921),pp.

A.
of Fractional

of Bessel Functions

On

the

Carnb. Phil. Soc.

93"100.
Extensions
XXVII.

of Fourier's and
A

Mat.

Mag.

April 3,

(May 16, 1898;

xvii.

(1900), pp.

(Dec. 14, 1908). Proc.

Theorems

the Bessel-Fourier

x.

R. Irish

(1910), pp. 205"248.


C. W.

OSEEN,
Neue

Series

(1899),pp. 171"199, 283"290.


product J^{x)J,,{x)(May 15, 1899). Proc. C'amb. Phil. Soc.

1899). Trans.

Phil.

(April2, 1921).

McF.

W.

ORR,

Divergent (or Semicouvergent) Hypergeometric

On

Acad.

Order

1020"1021.

Losung

des

Ditiraktionsproblems
(Jan. 10, 1912).

Sommerfeldschen

Astr. och Fi/sik,vii.

(1912),no.

OTTI,
Bessel'scher
-^genschaften

Arkiv

for

40.

Funktionen

H.

ii^erArt.

1898
Mittheilungen,

Bern

[1899],pp.

61"96.

PAOLI,
definiti (Oct.8, 1827).
Sopra gl'integrali
Sci. {Alodena),
xx.
(1828),pp. 161"182.
dell' equazione
SuU' integrazione

(Oct.8, 1827). Mem.


188.

di Mat.

di Fis. della

Mem.

P.
di Mat.

Soc. Italiana

di Fis. della Soc. Italiana

delle Sci.

xx.

(1828), pp.

delle

183"

FUNCTIONS

BESSEL

OF

THEORY

778

M.

PAESEVAL,
les series et sur
ordre
lineaires du second
partielles
Memoire

sur

I'lnst. par
presente's

Mem.

1799).

divers savans,

differential

some

differences

d'une

equation aux
(Le 16 germinal,an
(1805),pp. 639"648.

7"

coefficiens constans
i.

127"131.
Walk.
of the Random
The Problem
A mathematical
theory of random

April 5,

K.

PEARSON,
of

solution

the

On

A.

complete
I'integration

equations by

Bessel's

functions.

ix.
Messenger,

(1880),pp.

Series,

Biometric

Nature,

Memoirs^

O.

B.

PEIRCE,
Willson.

PERES,
Bessel

de

les fouctions

Sur

Research

(1906).

iii.

See under

(1905),pp. 294, 342.

Lxxii.

migration. Drapers Company

de

J.

plusieursvariables (August 9, 1915).

Comptus Reiidiis,

(1915),pp. 168"170.

CLXi.

0.
PERRON,
iJber die Kettenbruchentwicklung des Quotienten zweier Bessel'schen
xxxvii.
(1907),pp. 483"504.
(Dec. 7, 1907). Mii7ichener Sitzungsherichte,
J.

PETZVAL,

Gleichungen,I. (Vienna,1851).
Differential-

Integrationder linearen

des

E.

C.

PICARD,

Applicationde la theorie
gauches. Ann. sci. de VEcole

complexes lineaires

serie per funzioni

des

h,I'etude

sup. (2) vi. (1877),pp. 329

norm.

"

surfaces

des

et

courbes

366.

S.

PINCHERLE,

Sopra alcuni sviluppiin


(4) III. (1881),pp.

Functionen

analitiche

(Jan.12, 1882). Bologna Memorie,

151"180.

teoremi
sopra glisviluppiin serie per funzioni analitiche (March 23, 1882).
del R. 1st. Lomhardo, (2) xv. (1882), pp. 224"225.
Delia trasformazione
di Laplace e di alcune sue
(Feb. 27, 1887). Bologna
applicazioni

Alcuni
Rend,

Memorie, (4) viii.

(1887),pp.

125"143.

PLANA,
Note

G.

A.

A.

d-y

0 (Jan.20, 1822).
de I'equation
I'integration
T-^+5'^''"y

sur

delle Sci. di

Torino, xxvi.

Recherches

(1821),i:.p.519"538.

de la loi de pesanteur des


la decouverte
Soleil et sur la theorie de leur mouvement
elliptique
(June 20, 1847). Mem.
delle Sci. di Torino, (2) x. (1849),
249"332.
pp.

analytiquessur

POCHHAMMER,
Ueber
Math.

Ann.

die

lineare

vers
planfetes

le

della R. Accad.

L.

zweiter
Differentialgleichung

(1890),pp.

xxxvi.

della R. Accad

Mem.

Ordnung

mit

linearen

Coefficienten.

84"96.

Ueber
zweiter Ordnung mit
einigebesondere Falle dcr linearen Differentialgleichung
linearen Coefficienten (Sept.1890). Math. Ann.
xxxviii.
(1891),pp. 225"246.
Ueber
eine binomische
lineare Differentialgleichung
n^^^ Ordnung (Sept.1890). Math.
Ann. XXXVIII.
(1891),pp. 247"262.
Ueber
die Differentialgleichung
der allgemeinereni^-Reihe (Jan. 1891). Math.
Ann.

(1891),pp. 586-597.
eine specielle
lineare Differentialgleichung
2**"^Ordnung mit linearen Coefficienten
(May, 1891). Math. Ann. XLi. (1893),pp. 174"178.
Ueber die Differentialgleichungen
der Reihen jF(p,o- ; x) und JF(p,
"t, t ; .r)(June, 1891).
Math. Ann. xli. (1893),
pp. 197"218.

XXXVIII.

Ueber

Memoire

(1813),pp.

sur

les

POISSON,
integralesdefinies. Journal

S. D.
de

V^cole Poly technique,ix.

(cahier 16)

215"246.

Sur une
nouvelle manifere d'exprimerles coordonnees
des plan^tesdans le mouvement
Connaissance
des Terns,1825 [1822],
elliptique.
pp. 379"385.
Memoire
sur
des Equations
differences partielles.
Journal
lineaires aux
I'integration
VEcole Polytechnique,
xii.
(cahier19) (1823),pp. 215"248.

de

OF

THEORY

780

FUNCTIONS

BESSEL

Phil. Mag. (6) viii. (1904),


Dimensions.
in two
On the Open Organ Pipe Problem
Papers,v. (1912),pp. 206"211.]
481"487.
[Scientific
Phil. Mag.
of the Ratio of the Electrical Units.
On the Experimental Determination
v. (1912),pp. 330"340.]
97"108.
Papers,
[Scientific
XII.
(1906),
pp.
(6)
Phil. Mag. (6)
On the Light dispersedfrom Fine Lines ruled upon ReflectingSurfaces.

pp.

Papers,v. (1912),pp. 410"418.]


[Scientific
Whispering Gallery.Phil. Mag. (6) xx. (1910), pp. 1001"1004.
Papers,v. (1912),pp. 617"620.]
[Scientific
Phil.
as
appliedto the Vibrations of a Circular Membrane.
Note on Bessel's Functions
vi.
53"58.
(1920),
1"5.]
Papers,
[Scientific
pp.
Mag. (6) xxi. (1911),pp.
Phil. Mag.
in Bessels Functions.
Theorem
of Schlomilch's
On a PhysicalInter^jretation
Papers,vi. (1920),pp. 22"25.]
[Scientific
(6) XXI. (1911),pp. 567"571.
Phil. Mag. (6) xxii.
of Heat.
(1911), pp. 381"396.
in the Conduction
Problems
Papers,vi. (1920),pp. 51"64.]
[Scientific
(Jan. 11, 1912). Proc. Royal
On "thepropagationof Waves
through a stratified Medium
A (1912),pp. 207"226.
Papers,vi. (1920),pp. 111"120.]
Soc. Lxxxvi.
[Scientific
A
Electrical Vibrations on a thin Anchor
Ring (June 27,1912). Proc. Royal Soc. Lxxxvii.
vi.
193"202.
111"120.]
(1920),
Papers,
[Scientific
pp.
(1912),pp.
to the Whispering Gallery
of High Order
Further Applicationsof Bessel's Functions
Phil. Mag. (6) xxvii.
Papers, vi.
and Allied Problems.
[Scientific
(1914), pp. 100"109.
211"219.]
(1920),pp.
Phil. Mag:{") xxviii.
of Viscous Fluid Motion.
(1914),
Further remarks on the Stability
609"619.
vi.
(1920),
[Scientific
Papers,
pp. 266"275.]
pp.
of a Viscous
of the simple Shearing Motion
On the Stability
IncompressibleFluid.
Phil. Mag. (6) xxx.
(1915),pp. 329"338.
[Scientific
Papers,vi. (1920),pp. 341"349.]
when
is great and 6 has any value (April27, 1916).
On Legendre'sFunction*
n
Pn\."),
Proc. Royal Soc. xcii. A (1916),pp. 433"437.
Papers,vi. (1920),pp. 393"397.]
[Scie^itific
the higher Temperature
On Convection
Currents
in a horizontal Layer of Fluid, when
546.
side. Phil. Mag. (6) xxxii.
the under
is on
[Scientific
Papers, vi.
(1916), pp. 529
(1920),pp. 432"446.]
and of Random
On the Problem
of Random
Vibrations
Flights in one, two, or three
Phil. Mag. (6) xxxvii.
347.
dimensions.
(1919), pp. 321
Papers, vi. (1920),
[Scientific
(1907),pp. 350"359.

XIV

Problem

The

of the

"

"

604"626.]

pp.

A.

REIXECK,

tion,
DissertaKugelfunktionen und Besselschen Funktionen.
Math.
ilber
Fortschritte
der
die
1907,
(Halle,1907). (72 pp.) [Jahrbuch
p. 495.]

Die Verwandtschaft
Bern

zwischen

RICCATI,
Animadversatioues

COUNT

J. F.

in aequationes diflerentiales secundi

Supplementa,viii. (1724),pp. 66
quae Lipsiaepublicantur
Com.
Jacobi
Riccati Appendix ad Animadversatioues
secundi

gradus,editus
Sectione

mentorum,

ii.

Theorie

Chemie, (2) xcv.

der

gradus.

Eruditorum

Actorum

73.

in aequationes differentiales
Tomo
Eruditorum
viii.
SuppleLipsiaepublicantur
quae
510.
Acta
Ervditoi-um
puhlicataLipsiae,1723, pp. 502

in Actis
p. 66.

"

RIEMANN,
Zur

"

Nobili'schen

(1855),pp.

G.

F. B.

Farbenringe (March 28, 1855).

Ann.

Physik und

der

130"139.

Partielle Differentialgleichungen
und deren
Bernhard
Riemann.
Fiir den Druck
bearbeitet

Fragen. Von
Anwendung axif physikalische
und herausgegebenen von
Karl Hattendorf

(Brunswick,1876).
ROHRS,
Sphericaland
Soc.

V.

CylindricMotion

J. H.
Fluid

in Viscous

(May 14, 1874).

P.
RUDSKI,
eine Klasse transcendenter
Gleichungen. Prace Mat. Fiz.
[Jahrbuchiiber die Fortschritte der Math. 1892, pp. 107"108.]
Note sur
la situation des racines des fonctions transcendantes
Ueber

1891). Mem.

de la Soc. R. des Sci. de

Likge,(2) xviii.
RUSSELL,

The

Proc. London

Math.

(1874),pp. 125"139.

Effective Resistance
the Ber and Bei and

and

Inductance

Allied

of

Functions

(1895),no.

A.
Concentric

iii.

(1892),pp. 69

Jn+^{a;) 0 (May 10,


=

Phil.

Mag. (6)xvii.

pp. 524"552.
*

This

is the function

which

other

writers

denote

by

the

81.

3.

Main, and Methods

(Jan.22, 1909).

"

symbol P"(cos 6).

puting
of Com-

(1909),

781

BIBLIOGRAPHY

RUTGERS,
Over

bepaaldeintegraalI

die

pp. 368"373.
Over eene
pp.

reeks

88"90.
Over reeksen

waarin

Besselsche

van

Besselsche

e-'^z^-'^dz.

Besselsche

met

functies

J. G.

Functies.

functies

Archiefvoor

Wiskunde,(2) vr. (1905),

Nieuw

Archief

Wiskunde, (2) vii. (1907),

daarmede

en

voorkomen.

Nieuw

Nieuw

voor

samenliangendebepaaldeintegralen,

Archief

Wiskunde, (2) vii. (1907),pp.

voor

164"181.

Sur
VII.

les functions

cylindriquesde premiere espcice.

(1907),pp. 385"405.
Over
eenige toepassingen der
Besselsche

naar

functies.

Fourier'sche

Nieuw

Arcldef voor

ontwikkeling van
Wiskunde,

Archiefvoor

W.

RYBCZYNSKI,
Uber

Nieuw

Ausbreitung der Wellen der drahtlosen Telegraphicauf


der Physik und Chemie, (4) XLi. (1913),pp. 191"208.
5, 1913). Ann.
SASAKI,
On

the

of the

roots

"

(, "^

equation
equation "-f

pp.

380.

"

der

Erdkugel (March

,"
0.
Jn {fca)

Tohoku

"

Math.

Journal, v. (1914),

Jn {kr)

375

VON.

die

willekeurigcfuuctie

een

(1909),pp.

viii.

Wiskunde, (2)

45"47.

H.

SAVIDGE,
Tables

of the

and

ber

bei and

Computation (Nov. 12, 1909).

and

ker

Phil.

G.

kei Functions

Mag. (6) xix.

SCHAFHEITLIN,
Ueber

die

(Feb. 1887).
Ueber
Anyi.

Math.

Ann.

(1887),pp.

xxx.

further

durch

der hypergeometrischen
Integraldarstellung

bestimmtes

ein

Integral

Reihen

(Jan. 1888). Math.

156.

(1888),p.

Nullstellen

for their

157"178.

und eine
Ueber
die Gauss-fiche. imd Bessel-ache DifFerentialgleichuug
der letzteren (Feb. 16, 1894). Journal fiirMath. cxiv.
(1895),pp. 31"44.
Die

formulae

P.*

Darstellungder hypergeometrische Reihe

einc

XXXI.

with

(1910),pp. 49"58.

der Besseliich.e\iFunctioneu.

Journal

Math,

fur

cxxii.

Integralform

ueue

(1900),pp. 299

"

321.

Uber

in.

der

die Nullstellen

der Math,

und

Besselschen

Phys. (3) i. (1901),pp.

tjber den Verlauf der Besselschen


(1904),pp. 83"85.

zweiter

Art

(Jan. 10, 1901). Archiv

(May 18, 1904). Berliner Sitzungsberichte,

Funktionen

Funktionen

Lage der Nullstellen der Besselschen


93.
v. (1906),pp. 82
Sitzunffsberichte,

Die
Berliner

zweiter

Art

(June 27, 1906).

"

tJber den Verlauf


Math.

Funktionen

133"137.

Theorie

Die

der

Funktionen

der Besselschen

Vereinigung,xvi.

(1907),pp. 272

zweiter

279.
Funktionen.
Von

Besselschen

Art.

Jahresbericht

der Deutschen

"

Paul

Schafheitlin

(Leipzig und

Berlin,

Phys. Schrf ten fiirIngenieureund Studierende,4).


und den Besselschen
Funktionen
Beziehungen zwischen dem Integrallogarithmus
67.
viii.
(1909),pp. 62
24, 1909). Berliner Sitzungsberichte,
Funktionen
fiirdie Besselschen
Die semikonvergenten Reihen
(Sept.1, 1909).
129.
Math.
der Deutschen
xix.
(1910),
Vereinigung,
pp. 120
1908;

Math.

(Feb.

"

bericht
Jahres-

"

W.
coefficients in the

SCHEIBXER,
On

the

asymptotic values

radiusyector
IV.

(1^6),
pp.
Ueber

die

according
177"182.

to

of the

of any

power

der Coefficienten

in den

nacli der mittleren

Entwickclungen (May 31, 1856). LeipzigerBerichte,vni.


xvil.
(1880),pp. 545"560.]
H.

SCHERK,
tJber die
lie Integration der

Oleic
Gleichung

See

also under

Anomalie

(1856),pp.

40

"

(a+ ^.v)i/.Journal

pp. 92"97
*

of the

Journal,

mean

Werthe
asymptotische

vorgenommeuen
64. [Math. Ann.

development

anomaly (March, 1856). Gould's Astronomical


[Math. Ann. xvii. (1880),i)p. 531"544.]
the

Sonine.

fiir Math.

x.

(1833),

FUNCTIONS

BESSEL

OF

THEORY

782

L.
SCHLAFLI,
definiti che giovano ad esprimerela soluzione generale
Sulle relazioni tra diversi integral!
della equazionedi Riccati. A/m. di Mat. (2) i. (1868),pp. 232"242.
Neumann's
Herrn
Untersuchungen iiber die Bessel'schen
zu
Einige Bemerkungen
Math.
Ann.
iii.
[With note by C. N.
(1871),
Functionen
pp. 134"149.
(May 4, 1870).
p.

149.]
Sopra
Sull'

(Aug. 1871).

Ueber

piu generaleed applicataalia

forma

assoluto

(2) vi. (1875),pp. 1"20.


Convergenz der Entwicklung einer

die

di

funzione

una

arbitraren

"^" Os-^^),
J" (/So^),
wo
(/3i,f),
vorstellen
17,1876). Math.
(Jan.
GleichungJ^^(i3)

der

Functionen

/*

h constante

(Oct. 4,

sur

la variation

des

0.

nach

fix)

den

"""

X.
d'une

arbitraires

constantes

Function

die positiven
^i, ^2, ^s,
Ann.
x. (1876),pp. 137"142.

"".,

SCHLOMILCH,
Note

funzione

199"205.

(2) v. (1873),pp.

di Mat.

Bessel'schen
Wm-zeln

recato

di Mat.

Ann.

luugo le qualiil valore

delle linee

uso

Ann.

1872).

di Jacobi

teorema

un

cilindrica

integraledefinie.

Journal

fur

(1846),pp. 268"280.
1848).
Studien.^il. (Leipzig,
Analijtische

Math.

XXXIII.

Zeitschrift
fur Math,

Function.

die Bessel'schen

Ueber

und

Phys. ii. (1857),pp.

137"

165.
J. J.
SCHONHOLZER,
Integralemit Hiilfe von Veranderungen
Bern, 1877). [Graf and Gubler, Einleitung in die
tionsweges{Dissertation,
BesseVschen Funktionen, ii. (Bern,1900).]
Ueber

die

Auswerthung

bestimmte

G.

SCHOTT,
Radiation
Electromagnetic

Beugung

und

Integra-

Theorie

der

A.

(Cambridge, 1912).
K.

SCHWARZSCHILD,
Die

des

Polarisation

des

Lichts

durch

einen

Spalt. Math.

Ann.

LV.

(1902),

pp. 177"247.
F.

SCHWERD,
Die

Beugimgserscheinungen
aus

den

M.

der
Fundamentalgesetzen

Undulationstheorie

heim,
(Mann-

1835).
SEARLE,
the

propagationof
(1908),pp. 51"66.

On
XXXIX.

waves

in

an

SEGAR,
On

the roots

of certain

continuants.

J. H.

atmosphere

Journal,
varying density.Quarterly

W.

H.

Messenger,xxii.

SERRET,
Note

C.
of

(1893),pp.

sur

IX.

(1844),pp.

the

Reflection

181.

(1843),pp. 1 27.
difierentielles k indices
viii.

458

"

"

475; Journal

de

193-216.

SHARPE,
On

"

J. A.

quelques formules de calcul integral.Journal de Math.


Memoire
d'une equationdifierentielle h.I'aide des
sur
I'integration
quelconques (Sept.4, 1843). Comptes Rendus, xvii.
(1843), pp.
Math.

171

of Sound

the

Surface

H.

J.

Paraboloid
a
(Nov. 1874). Quarterly
(1877),pp. 1 8.
On a diflferentialequation. Messenger,x. (1881),
44.
xi. (1882),pp. 41
pp. 174"185;
On a transcendental
differential equation. Messenger,XI. (1882),
63.
pp. 56
On a differential equation. Messenger,xiil. (1884),
79.
pp. 66
Note on Legendre'scoefficients. QuarterlyJournal,xxiv.
386.
(1890),pp. 383
On The Reflection of Sound
at a Paraboloid
(June 20, 1899). Froc. Camb. Phil. Soc. x.

Journal, xv.

at

of

"

"

"

"

"

(1900),pp.

101"136.

W.
F.
SHEPPARD,
expressionsof a function of a singlevariable
Quarterly
Journal,xxiii. (1889),pp. 223"260.
On

some

SIACCI, F.
di una
integrazione
equazionedifferenziale e
1901). Napoli Rendiconto,(3)vii. (1901),pp. 139"143.
Sulla

suUa

in terms

of Bessel's functions.

equazionedi

Riccati

(April13,

783

BIBLIOGRAPHY

SIBIRANI,
r

Sopra

equazione di

Fortschritte der Math.

Riccati.

Riv.

mat.
fis.

F.

die

Integraleeiner

iiicht homogeneii

Luisenschule,Berlin, 1890,

gramm,

[Jahrbuchiiher die

220.

"

1909, p. 369.]
SIETifON,

Ueber

(1909),pp. 216

xix.

22

P.

zweiter Ordimng.
Diflerentialgleichung

[Jahrbuchiiher die

pp.

Pro-

Fortschritte der Ma.th.

1890,

340"342.]

pp.

Table

E. A.*
SMITH,
Fq and J'j. Messenger.,
xxvi.

of Bessel's Functions

Arched

Dams.

Proc. American

Soc.

CLARA

SMITH,
theorem

of Abel

Bessel's functions

and

its

Mat.

Math.

O.

Soc.

of

function

in

(1907),pp.

viii.

of

terms

92

106.

"

sphdriqueset cylindriques.Giornale

Functionen

A.

J.

in der mathematischen

di

W.f

Physik. Dissertation,
Konigsberg,

(75 pp.). [Jahrbuchilber die Fortschritte der Math.

1891

425.

"

365"373.

SOMMERFELD,
willkiirlichen

375

A.

les fonctions

entre
integrales

(2) Xll. (1905),pp.

Die

development

(June 8, 1906). Trans. American

relations

101.

"

E.

the

applicationto

SMITH,

Quelques

(1897),pp. 98
(1920),pp.

of Civil Engineers,XLVi.

519"523.]

1891, pp.

Zur analytischenTheorie der Warmeleitung. Math. Ann. xlv,


Mathematische
Theorie der Diffraction (Summer, 1895). Math.

(1894),pp.
Ann.

263

277.

"

(1896),pp.

xlvii.

317"374.

tjber die Ausbreitung der Wellen der drahtlosen Telegraphic(Jan.15, 1909). Ann.
736.
Chemie, (4) xxviii.
(1909),pp. 665

Physik und

Greensche

Die

Vereinigung,

xxi.

Funktion

der

(1912),pp.

309

Schwingungsgleichung. Jahreshericht
"

the

Collection
pp.

resolution

Recherches
series

into

the Moscow

Math.

Sur

Math.

N.

J.

infinite series

(Jan. 17;29, 1870).

Mathematical

[MaTeMaTIIieCKia
v. (1870),
C6opHIIKt],

Soc.

323"382.

les fonctions

sur

(Aug. 1879). Math.


les fonctions

[Note on
k

of functions

publishedby

271"302,

der Devtscher

353.

SONINE,
On

der

"

paper by
les fonctions
a

Ann.

le

cylindriqueset

developpement

(1880),pp. 1"80.
cylindriques(Oct.24, 1887). Math.

des fonctions continues

en

xvi.

Ann.

xxx.

(1887), pp.

582

"

583.

Schafheitlin.]

Sur
cylindriques.(Extrait d'une Lettre adressee
Copenhague, May 6, 1904.) Math. Ann. lix. (1904),pp. 529"552.

SPITZER,

M.

Niels

Nielsen,

S.

Zeitschrift
filrMath,

der Differentialgleichung
Integi-al
.vy"-y=Q.

und

Phys. II. (1857),

pp. 165"170.
in unendliche
Entwickelung von e^^'^'^^-^
244"246.
(1858),pp.
Kettenbruches
Darstellungdes unendlichen

Reihen.

Zeitschrift
fur

Math,

und

Phys. in.

2.r + H
2x +

'i

+
^
2:f + 5 +
2:r + 74-...
,

in

Archiv der Math, und Phys. xxx.


geschlossenerForm.
X?eber die Integrationder Differentialgleichung

Lvii.

of the

Varying

also under

Michell.

cases

some

"

334.

(1860),pp. 82"87.
STEARN,

On

331

-''

dx'"-

(1859). Journal filrMath.

(1858),pp.

Motion

of

H.
a

Viscous

Fluid.

Quarterh/ Journal,

90"104.

(1880),pp.
*

See

+ See also under

Hopf.

xvii.

FUNCTIONS

BESSEL

OF

THEORY

784

L.

STEINER,

Beugungserscheinungdurch

Intensitiits-verhaltnisse der

(June 19, 1893). Math,

Berichte

Naturwiss.

und

eine

Ungam,

aus

xi.

kreisformigeOffnung
(1892"93) [1894],pp

362"373.
the Solution

On

Joxvrnal,xviil.
Qiutrterly

Equation

330

(1882),pp.

345.

"

A.

STEPHENSON,
extension

An

77.
pp. 70
A more

E.

A.

STEINTHAL,
of the

of the Fourier

method

sine series.

expansionin

of

Messenger,xxxiii.

(1904),

"

general case

of

expansion

series.

in sine

Messenger,xxxiii.

(1904),pp. 178"

182.

Phil.

(June,1907).

in Bessel's Functions

Expansion

On

(1907), pp.

Mag. (6)xiv.

547"

549.

M.

STERN,
Journal

Recherches
III.

quelques

sur

363

(1846),pp.

series

A.

Methode

Sturm'schen

der

die Anwendung
fur Math, xxxiil.

Ueber

transcendente

auf

Gleichungen.

365.

"

T. J.
STIELTJES,
semi-convergentes. Ann.

sci. de VEcole

norm.

sup.

(3)

(1886),pp. 201"258.
of

(March 11, 1850).


Papers,ii. (1883),pp. 329"357.]

Phil.

On

the

Calculation
Camh.
Trans.

Efl:'ectof the Internal


Camh.
Phil. Soc.

Trans.

1850).
(1901),pp. 1"141.]
On

the

GEORGE

SIR

STOKES,
numerical

the

On

Soc.

Friction

GABRIEL.
of

Class

ix.

Definite

(1856), pp.

of Fluids

on

Integralsand Infinite
166"187.
[Math, and

the Motion

Discontinuityof ArbitraryConstants which


Phil. Soc. X. (1864),
pp.

(May 11, 1857). Trans. Camh.


IV.
(1904),pp. 77"109.]

appear
106"128.

in

Phys.

of Pendukims

(1856), pp. [8]"[106J.[Math, and

ix.

Series

(Dec. 9,
Phys. Papers,ill.

Divergent Developments

[Math,and Phys.Papers,

which
the Discontinuityof ArbitraryConstants
on
appear in
(May 25, 1868). Trans. Camb. Phil. Soc. XI. (1871),pp. 412"425.
[Math,and Phys. Papers,iv. (1904),pp. 283"298.]
from a Vibrating Body to a surrounding Gas*
of Vibration
the Communication
On
463.
(1868)[1869],
[Math.
pp. 447
(June 18, 1868). Phil. Trans, of the Royal Soc. CLVlii.
and Phys. Papers,iv. (1904),pp. 299"324.]
Smith's
Prize Examination
Papers, Feb. 1853 and Jan. 29, 1867. [Math, and Phys.
Papers, v. (1905),pp. 319, 347.]
which
of Arbitrary Constants
as
the Determination
Note
on
Multipliersof
appear

Supplement

to

paper

Divergent Developments

"

Semi-convergent Series (June 3, 1889). Proc. Camb.

Phil. Soc.

[Math, ayid Phys. Papers,v. (1905),pp. 221"225.]


that appear
On the Discontinuity of ArbitraryConstants
Series t (April23, 1902). Acta Math. xxvi.
(1902),pp.
Papers,v. (1905),pp. 283"287.]
J. W.
STRUTT,
See Rayleigh(Lord).
H.
STRUVE,

as

vi.

(1889), pp.

Multipliersof

393"397.

Fernrohren
auf Lichtscheiben
den Einfluss der Diflraction an
de VAcad.
(7) xxx. (1882),no. 8.
hnp. des Sci. de St. Petershourg,
Beitrag zur Theorie der Diflraction an Fernrohren
(Aug. 1882). Ann.

Ueber

362"366.

vergent
Semi-con-

[Math,and Phys.

(May 25, 1882).

Mem.

Chemie,(3) xvii.

(1882),pp.

STURM,
les

Sur
Math.
*

I.

Equations difierentielles

An

abstract

composed

Physil-und

J. C. F

lineaires du

second

ordre

(Sept.28, 1833).

Journal

de

106"186.

(1836),pp.

will be found

t This, the last

der

1008"1016.

in Proc.

published paper

for the Abel

centenary

Royal

which

volume.

Soc. xvi.
(1868),pp. 470"471.
Stokes wrote, contains
a summary

of his researches

785

BIBLIOGRAPHY

P. J.
SUCHAR,
Sur les equationsdifferentielles lineaires reciproquesdu
Bidl. de la Soc. Math, de France,xxxii.
(1904),pp. 103"116.
A.
SVANBERG,
Nova
integralibusdefinitis disquisitiones.

De

second

ordre

(Nov. 18, 1903).

F.
Acta

Soc. Sci. Upsala,


y
t^

R.

(1832)
h
\

x.

pp. 231"288.
T.

TAKEUCHI,
On

integralj e''"''"'^'^
smqd, conqBdO (May 17, 1920).

the

Tohoku

Math.

Journal,

(1920),pp. 295"296.

XVIII.

Bestimmte

L.
THEISINGER,
Integrale.Monatshcfte
fur Math, und Phys. xxiv.

1912"13, pp. 8"10.

Reilie.
[Int.Catalogneof Scl. Lit. xiii.
SIR

THOMSON,

/CO
Recent

Researches

dx.

Seminar

JOSEPH

Jena,

JOHN.

(1882),pp. 377"381.

Magnetism (Oxford,1893). [Review by

and
Electricity

zu

(1913),p. 100.]

QuarterlyJournal,xvin.

2^iT5-T|:jT
in

^,

"

Nature, XLix.

Berichte der Math.

der Bessel-Fourierschen

COHSX

328"346.

J.

THOMAE,
Die Brauchbarkeit

(1913),pp.

Gray

(1894),pp. 357"359.]

THOMSON,
See Kelvin

W.

(Lord).
I.

TODHUNTER,

ElementaryTreatise

on

and

Functions

Functions,Lame's
Laplace's

BesseVs Functions

(London, 1875).
TURRIERE,

Uue

E.
la serie consideree par Airy dans la diflfraction des
Annales de Math. (4) ix. (1909),pp. 433"441.

applicationgeometrique de
circulaires. Nouvelles

ouvertures

F.

UNFERDINGER,
tjber die beiden
LVii.
berichte,

{nx-co^x)dx
Integrale ie^^^^'^^^

(1868),pp.

(April16, 1868).

G. C. A.
VALEWINK,
Haarlem,
asymptotischeontwikkelingen(Dissertation,
Fortschritteder Math. 1905, p. 328.]

On
pp. 75

the
"

Roots

Sitzungs-

611"620.

Over
die

Wiener

of Bessel-

and

E. B.
VAN
VLECK,
P-functious.
American

pp.

Journal

138). [Jahrhuch Uher

of Math.

xix.

(1897),

85.

VERDET,
i. (Paris,
1869).
Legons d'Optique
Physique,

E.

E.
ordinaires du

VESSIOT,
Sur

quelques equationsdifferentielles

des Sci. de
/^

Sopra

alcime

questionidi

second

ordre.

A nnales de la Fac.

6.

Toulouse,ix. (1895),no.

V.
VOLTERRA,
di integralidefiniti. Ann.

inversione

di Mat.

(2)xxv.

(1897),

pp. 139"178.
G.
VORONOI,
et ses applications
a la sonuuation
fonction transcendante
459"534.
xxi.
207"268,
Ann. sci.de VEcole norm.
(1904),
(3)
pp.
sup.
des somiues
Sur le develo^jpementj
a I'aide des fonctions cylindriqucs,
Sur

une

+ rn'^ est

oh

forme

2/ {pni^+ ^qmn + rn^)


positiveh coefficients

pm'^+ 2qmn
Kongressesin Heidelberg(1904),pp.
W.

B. F.

une

241

"

entiers.

de

quelques series.
doubles

Verh. des dritten Int.

245.
50

FUNCTIONS

BESSEL

OF

THEORY

786

C.

WAGNER,
Beitrage

Entwicklung

zur

Function

der Bossel'schen

Art.

i.

Bern

1894,
Mittheilungcn,

pp. 204"266.
tjber die

Integraledurch Bessel'sche Funktionen


Darstellung einiger bestimmten
B
ern
1895, pp. 115"119;
1896, pp. 53"60.
1895).
Mittheilungen,
5,
Aug.
(June 12, 1895;
G.

WALKEE,
formulae

Some

transformingthe originof

for

of Bessel's functions.

Messenger,

(1896),pp. 76"80.

XXV.

W.

G.

WALKER,
of electromagnetic
waves
scattering

The
pp.

T.

reference

by

(1900),

sphere. QuarterlyJournal,xxxi.

36"49.

J.

WALKER,
IVie

Theory of Light (Cambridge, 1904).


Analytical
G.

WALLENBURG,
hoherer
DifFerentialgleichungen

Riccati^sche

Ueber

Ordnung.

filrMath.

Journal

cxxi.

(1900),pp. 196"199.
deren allgemeinesIntegraleine lineare gebrochene Function
Die Diflerentialgleichungeu
217.
ist. Journal filrMath. cxxi.
der willklirlichen Constanten
(1900),pp. 210
differentielle de Riccati du second ordre (Dec.14, 1903). Comptes Rendus,
Sur I'equation
"

1033"1035.

(1903),pp.

cxxxvii.

WANGERIN,

A.

Encyclopddieder Math.
Cylinderfunktionenoder Bessel'sche Funktionen.
des Set. Math.
Tome
757.
Teil 1 (Leipzig,1904
16), pp. 742
[Encyclope'die
229), translation by A. Lambert.]
(Pariset Leipzig,1914, pp. 209

Wiss. Bd.

Ii.

vok

n.

"

"

"

G.

WATSON,
On

certain

pp. 50"55.
Bessel
XVI.

difference
and

Functions

equation of
Kapteyn

the second

Series

N.
order.

QuarterlyJournal,

(April 26, 1916).

Proc.

London

(1910),

xli.

Math.

Soc.

(2)

(1917),pp.

150"174.
Functions
of Equal Order

and Argument (June 17, 1916). Phil. Mag. (6)xxxir.


Bessel
(1916),pp. 232"237.
157.
Simple types of Kapteyn Series. Messenger,xlvi.
(1917),pp. 150
Bessel functions of equal order and argument (Nov. 13, 1916). Proc. Carnh. Phil. Soc.
"

XIX.

(1918),pp.

42

48.

"

The limits of applicability


of the Principleof StationaryPhase
Camb.
Phil. Soc. XIX.
(1918),pp. 49"55.
Bessel functions
of large order (June 14, 1917). Proc. Camb.

Proc.

(Nov. 22, 1916).


Phil.

Soc.

(1918),

xix.

pp. 96"110.
The

Zeros

of Bessel

Functions

(Aug. 17, 1917).

Royal Soc.

Proc.

xciv.

(1918),pp.

190"206.

Bessel
370.
The
A

of

Diffraction

Equal

of Electric

Order

and

Waves

Argument.

by

the Earth

Phil.

Mag. (6)xxxv.

(May 29, 1918).

(1918),
pp.

Proc.

Royal

364

Soc.

"

xcv.

(1919),pp. 83"99.
The

xcv.

Functions

On

Transmission

of Electric Waves

round

the Earth

(Jan. 13, 1919). Proc. Royal Soc.

(1919),pp. 546"563.
Nielsen's

The

functional

of Lommel's

zeros

53.
equations. Messenger,XLViii.
(1919),pp. 49
polynomials(May 15, 1919). Proc. Londoii Math.
"

Soc.

(2)xix.

(1921),pp. 266"272.
The

expansionof

(1904),pp.
On
Trans,

an

WEBB,
arbitraryfunction in

H.
a

A.

series of Bessel functions.

Messenger,xxxiii.

55-58.

the Convergence of Infinite Series


of the Royal Soc. cciv. A (1904),pp.

Ueber

einigebestimmte

Ueber

die

1868). Math.

of

Analytic

Functions

(Nov. 10, 1904).

H.
WEBER,
Integrale(Jan. 1868).

Journal

filrMath.

Integrationder partiellenDiffereutialgleichung:+
Ann.

i.

(1869),pp.

1"36.

Phil.

481"497.

LXix.

(1869),pp.

+ /?^^m
=

(July,

788

OF

THEORY

WORMS
Note

de

I'integration

sur

Math.

(3)

(1878),

iv.

dx

177"186.

pp.

W.

YOUNG,
On

1911).
On
pp.

integrals

infinite

Quarterh/
series

of

involving

Journal,
Bessel

P.

ROMILLY,

DE

I'e'quation

dx^
de

Journal

FUNCTIONS

BESSEL

XLiir.

functions

H.

generalisation

(1911),
(Dec.

pp.

6,

161

1917).

the

of

sine

and

cosine

functions

(Oct.

177.
"

Proc.

London

Math.

Soc.

(2)

xviii.

(1920)

163"200.

I.

ZELINSKIJ,
On
Naturalists'

the

integration
Soc.

Imp.

of
Univ.

Riccati's

Kazan,

equation
viii.

(1890),

(Jan.
pp.

I.

1890).

27,
337

342.
"

Proc.

Phys.

Math.

Section,

6,

INDEX
[^Thenumbers

J",, 243, 244,


An,Q;

263

529
"S^:,.n{z),
^n.. (0,571
6

a,

^",

refer to the pages

on

inhich the

symbolsare

j"F,{a"oi,,...
,oi^\p"p,,...,

584

^,,,(0,283
S4o{cc,
t); 598

SYMBOLS

OF

defined.^

(^),308

J.

3).(^),61

...,p,; ^),100

F(*-).639
646
F,(a,-),

j,507

"M, /9;7, V; I V), "'^71


/.(^),55
/" (z),483
526
/(^)j^,
536
f{2)R,

",./;,"",485
479
i.,n,

j.n,576

K,

559

A^n (^),78 [seealso p. 65]

K,(z),78 [seealso
/C, (z),46

Bm("2), 247
Bn; M, (t),293

(?,21

^0 (^),597

Gn (z),64

(^),529
^''^..

K, (^),78

81
(x),bei (a;),
81
ber^(a;),
bei^(a;),

a),640
(a,-,
G, (a;,
a),648
276
gn(t),

Ln(z),71

0,,

5'n,.(0284
303
^^,,(^),

ber

554

1 ; 232
:

65

ker

Cn"

(z),50

"^.(^X 82
W

129
(zip),

c, 508

(z),kei (^),81

Ii.(4

329

Z,^,,111

C"(^),658
C{n(a),S20

(2t)n,288

H,o77
ir,w (^),i7.'='
(z),73

il/,6

H,(^),328
her (^),hei (z),81
81
her, (^),hei^(z),

M{a;),mi
ii\ 471, 475,

..

i),,598

//,

74
i)"(2),

/.(^),77

13

; 505

514

324
iV_,,,.,",
661
iV^(^),

0,197

En{z), 56
Bin (a),320

J,(z),10,16
Jn{z\ 6, 14,
J-n {z\ 16

E;5),308

/^,",10

E,6

e, e,,

111

258
i^(6'),

J?'(a,0, 90

F{e, x), 253


77
P,(a;),

On(t),271, 272
19

30
J'^rC^),
J,(^),38
J(/i,a?),49

J{z\p),129
326
^n.fc(^),
J{z;

v,

k),327

0_, (0, 276


""" (0- 569
7^547
P., 227

Pn(^), 156
P,"'^-'
(^),295
P.-'^(^),51

F,Acoscf"p),129

p.

65]

790

THEORY

OF

BESSEL

FUNCTIONS

LIST
[The numbers

referto

OF

the pages.

AUTHORS
References are
pp.

Abel, N. H., 68, 616,


Adamoff, A., 196
Aichi, K., 80

321, 322, 654, 659

S.,65, 655, 656, 657, 658, 662, 663,

664

Alexander, P., 579

Anding, E., 313, 657


Anger, C. T.,21, 22, 308, 309, 310, 311,
371

479

Barnes, E. W., 83, 100, 102, 104, 105, 156,


190, 192, 195, 196, 220, 221, 340, 351, 357,
409

367, 383, 387, 402,


Basset, A. B., 76, 77, 78, 80, 172, 173, ISO,
385, 388, 395, 425, 426, 454
Bateman, H., 130, 131, 367, 370, 372, 373,
376, 379, 380, 389, 406, 417, 437, 456, 530,
533, 535
Bauer, G., 50, 128, 368, 370
Beltrami, E., 51, 358, 361, 374, 386, 389, 390,
(1700-1782), 2, 3, 4, 9, 85,

86, 87, 88, 111, 123, 478, 576


Bernoulli,James (1654-1705),1, 2,3,88, 90,

Bruns, H., 327


Bryan, G. H., 127, 480
Burkhardt, H. F. K. L., 236,

280

S.,305

Cailler,C., 44, 149, 386, 395, 415, 437, 455,


536, 537
Callandreau, 0., 196, 208, 387
Cantor, G. F. L. P.,637
Carhni, F., 6, 7, 194, 225, 226, 227, 249, 255,
268, 572
Carslaw, H. S., 177, 366, 395, 499, 500, 507,
509, 583

Catalan,E. C, 21, 22, 27, 96, 173, 188


Cauchy, (Baron) A. L., 7, 15, 16, 21, 150,
183, 230, 231, 232, 233, 247, 249, 259, 309,
319, 324, 449, 545, 554, 557, 579
Cayley, A., 88, 90, 96, 102, 103, 109, 188,

502*^
91

Chapman, S.,621
Chessin, A. S.,135, 175, 325, 346, 382
Chree, G, 6, 597
E. B., 154
Christofiel,

92

Bernoulli,John

^louUi,

Brassinne, E., 91
Brenke, W. C, 23
Bridgeman, P. W., 597
Bromwich, T. J. Fa., 8, 11, 44, 68, 156, 187,
189, 191, 202, 203, 214, 230, 231, 234, 279,
302, 349, 360, 385, 391, 393, 399, 574, 575,

Challis,IT. W.,

391, 500, 579, 621

ulli,Daniel

bibliography/,

Brajtzew,J. Ti.,169

Burnside,W.

L., 94

Bach, D., 96, 102, 1 10


Bachmann, P., 197

Berne

in the

601
312

Ani^imov, V. A., 92

Baehr, G. F. W.,
Ball,L. de, 157

to entries

659, 660

516, 659, 660

Autonue,

given

Borel,E., 8, 281, 536


Bourget,J.,6, 324, 325, 326, 484, 485, 517,

621

Airy, Sir George B., 188, 189, 229, 249, 320,

Appell,P. E., 146,

not

753"788.]

Airey,J. R., 65, 142, 214, 247, 319, 502, 505,

Aldis, W.

QUOTED

1, 2, 3

(1667-1748),
Nicholas, (1687-1759),2

Bernoulli,Nicholas, (1695-1726),2
Bessel,F. W., 1, 3, 4, 5, 9, 13, 14, 15, 18, 19,
21, 24, 25, 38, 84, 140, 148, 153, 160, 295,
308, 478, 551, 554, 654
Binet, J. P. M., 183
Bocher, M., 46, 57, 64, 376, 494, 495, 517
Bohmer, P. E., 142
Boole, G., 27,47, 110,627

Chrystal,G., 102, 288, 295


Cinelh, M., 633
Clebsch, R. F. A., 359, 363
Cliftbrd,W. K., 90, 91
Coates, C. v., 173, 180, 313, 622
Cotter, J. R., 41
Crawford, L., 27
Crelier,L.,286, 287, 288, 295, 300, 301, 302
Curtia,A. H., 96, 1 10
(hirzon,11. E. J.,395

D'Alembert, J. le Eond,
Dandelin,

BESSEL

OF

THEORY

792

L. von, 50, 51, 129, 138, 151,


274, 283, 284, 290, 293, 351, 362, 363, 365,
366, 367, 368, 369, 370, 373, 378, 379, 383,

Gegenbauer,

G. P., 503

Dai-boux, J. G., 233


Darwin, C. G., 437

384, 385, 386, 389, 390, 391, 393, 395, 396,

Debye, P., 225, 235, 237, 240, 241, 247, 249,


516

De

250, 251, 255, 262, 263, 268, 513,


la Valine Poussin, Ch. J.,53, 160,

Morgan, A., 188, 190


Dendv, A., 107
Dini, U., 10, 577, 578, 597, 600, 616,

189

651

Dinnik, A., 579, 659, 660

Dirichlet,P. G. Lejeune,157, 230, 406, 581,

C, 35, 480, 481,


Dixon,
Donkin, W. F., 109
Dougall,J.,65, 411

482

470

Earnshaw, S., 108


Ellis,R. L., 95, 109, 110, 173
Encke, J. F., 342
Enestrom, G., 92

Gifford,E., 662
Gilbert,L. P., 545, 548, 549
Giuliani,G., 155, 156, 324, 326, 327
Glaisher,J. W. L.,89, 96, 102, 103, 108, 109,

J.

B., 120

503

Graf, J. H., 32, 64, 75, 145, 153, 160, 165,


175, 197, 215, 227, 286, 287, 290, 295, 296,
299, 301, 302, 303, 341, 344, 345, 359, 360,
362, 398, 498, 502, 583
Gray, A., 64, 65, 78, 194, 206, 454, 480, 655

Greenhill,Sir A. George, 91, 96

Enneper, A., 173


Epstein,S. S., 145,
Escherich,

A., 197

656, 657, 658, 660


Green, G., 124

F., 248

Gregory, Duncan
Farquharson, 391
Gregory, Walter, 224
Grunert, J. A., 27
Gubler, E., 32, 64, 145, 160, 165, 177, 197,

290

455

G. von,

119

Graeffe,C. H., 502,

Reymond, P. D. G., 183, 455,


Duhamel, J. M. C, 38, 49, 59, 68, 227

ErmakoflF,W.,

G.

Goursat,Edouard

Bois

Emde,

Genocchi, A.,

140, 171, 173, 183, 664


Gordan, P. A., 55

623

A.

398, 406, 407, 413, 414, 415, 418, 426, 430,


438, 439, 480, 508, 517, 522, 524, 525, 579
Gibson,

De

Du

FUNCTIONS

165

Euler, L., 3, 4, 5, 6, 24, 49, 53, 60, 62,87,88,


92, 93, 123, 133, 183, 410, 498, 500, 501,
503, 576, 659

215, 227, 286, 287, 301, 329, 341, 351, 398,


408, 410, 426, 498, 502, 583
Gunther, S.,153
R. F., 621, 636
Gvvyther,

Falkenhagen, J. H. M.,
Fejer,L., 610
Feldblum, M., 92
Fields,J. C., 110

94

Hadamard, J.,204, 205, 527


Haentzschel, E., 71, 96, 159

Filon, L. N. G., 51, 578, 622, 623, 625, 629


Ford, W. B., 578, 605
Forsyth, A. R., 42, 57, 107, 109, 117, 346,
499

358, 400,
Fourier, (Baron) J. B. Joseph, 4, 9, 10, 13,
22, 84, 135, 449, 450, 454, 455, 456, 478,
482, 483, 501, 576, 577, 578,

616

501

Freeman, A.,
Frenet, F., 27
Fresnel, A. J.,544, 545
Frobenius, F. G., 57
Frullani,G., 14, 19

Hafen, M., 389


Hague, B., 656
Hall, A.,

Rowan, 12, 195, 655


Hankel, H., 10, 38, 57, 58, 61, 62, 63, 65, 73,
75, 76, 77, 160, 163, 164, 165, 167, 175,
195, 196, 203, 206, 208, 211, 384, 386, 390,
393, 395, 424, 427, 428, 429, 430, 434, 453,
454, 456, 457, 458, 459, 462, 464, 465, 471,
488, 513, 514, 577, 579, 581, 582, 633
Hansen, P. A., 14, 20, 30, 31, 37, 152, 154,
155, 158, 195, 292, 406, 655,
Rao, C. V., 437

Gasser, A., 509, 517


Gauss, C. F. (Johann FriedrichCarl),
191, 506

656

Hanumanta

Hardy,
Gallop,E. G., 405, 421, 422
Gaskin, T,, 109

15

Hamilton, Sir William

G.

H., 8, 111, 180, 183, 188, 189, 200,

309, 320, 321, 322, 324, 373, 382, 386, 395,


406, 421, 422, 437, 441, 442, 463, 464, 542,
546, 547, 573, 575, 579, 606, 615, 621
Hargreave, C. J.,88, 170, 171

LIST

OF

AUTHORS

Hargreavcs,11.,538
Harnack, A., 577
Harris, J. A., 15
Havelc.ck,T. H., 125, 171
Hayashi, T., 165
Heaviside, 0., 64, 65, 203, 367, 385, 387, 388,
393, 395, 410, 426
Heine, H. E., 4, 56, 65, 66, 84, 154, 155, 156,
157, 181, 358, 363, 365
Hermite, C, 55, 477
Hertz, H., 80, 81

Herz, N., 554,

555

Hill,C. J. D., 94

Hobson,

W., 10, 33, 54, 58, 125, 128, 129,


149, 172, 174, 280, 353, 363, 369, 385, 386,
387, 480, 485, 578, 586, 591 602
Hopf, L., 178, 406
Horn, J.,225, 526
Hurwitz, A., 9, 297, 302, 303, 304, 305, 306,
E.

307, 483, 484

44

Jackson, William

Hartas, 177
Jacobi,C. G. J.,6, 8, 14, 21, 22, 25, 26, 27,

Kalahne, A., 505, 507, 660


Kapteyu, W., 35, 183, 200, 268, 279, 281,
282, 292, 351, 370, 373, 376, 380, 386, 404,
413, 498, 499, 531, 532, 533, 535, 536, 538,
551, 559, 560, 562, 565, 566, 568, 569,
570

(Sir

Thomson),

Landau, E. G. H., 197


Laplace,P. S. de, 6, 7, 8,9,53, 280, 395, 421,
450

Largeteau,C. L., 501


Laurent,Paul Mathieu Hermann, 157
Lam'cnt, Pierre Alphonse, 100
Lebedeff,Wera Myller-,99
Lebesgue,Henri, 457
Lebesgue, Victor Amddee, 110, 123
Lefort,F., 6

Legendre,A. M., 52, 90, 183, 204, 485, 557


G. F. A. (Marquis de St Mesme),
I'Hospital,
134

Leibniz,G. W., 1, 2, 3
Le Paige,C.,96

Liouville,J., 27, 28, 87, 111, 112, 116, 117,


119, 120, 123
R. O. S.,11, 12, 195, 200, 206, 331,
Lipschitz,
339, 384, 386, 390, 633
N., 503
Lobatsclievsky,
Lobatto, R., 49, 90
Lodge, A., 224, 229
Lommel, E. C. J. von,

28, 29, 84, 195, 379, 555, 572

Jamet, E. Y., 94
Johnson, W. W., 92
A. E., 528, 529
Jollifte,
Y.
Julius, A., 65, 200

William

Lambert, J. H, 485
Lamd, G., 96, 159

Lindstedt,A., 545, 661

343

Kelvin

Lacroix,S. F., 27
Lagrange, J. L. de, 6, 27, 28, 551
Lamb, H., 56, 96, 385, 416, 475, 502

Lerch, M., 382, 393, 433, 434, 617


Lindner, P.,484

Ignatowsky, W. von, 365


Isherwood,J. (".,657, 658, 664
Jackson,Dunham,
Jackson, Frank Hilton, 43,

793

QUOTED

Lord, 81,

124, 203, 225, 229, 230, 233, 248, 654,

13, 21, 23, 25, 30, 34,


38, 43, 45, 46, 47, 49, 53, 59, 62, 65, 66,
71, 73, 76, 77, 96,97,98,99, 106,107, 132,
133, 135, 136, 140, 142, 143, 145, 148, 151,
152, 154, 196, 200, 294, 295, 296, 297, 298,
299, 300, 303, 308, 315, 328, 341, 345, 348,
350, 364, 374, 406, 478, 479, 482, 529, 531,
537, 538, 539, 540, 542, 543, 544, 545, 546,
548, 549, 550, 576, 654, 655, 656, 658, 660,

661, 664
Lorenz, L.,57, 96, 224, 229, 500
Love, A. E. H., 56, 226, 417, 449, 654,

662

658

78, 79, 80, 158, 170, 171,


Macdonald, H. I\l.,
225, 229, 233, 365, 377, 385, 386, 389, 395,

Kepinski,S.,99
Kepler,J.,551, 552

i^rchhoff,
G., 107, 196, 203,
Kluyver, J. C, 367, 419,
J. C. C.

Kneser,
Knockenhauer,
Konig,

389, 578, 616

420

A., 499, 578, 583


W., 545

K.

J.,354, 523

Koppe, M.,
Kummer,

247

E.

E., 49, 90, 101, 102, 104, 105,

148, 185, 190, 191, 196, 203, 394, 409

396, 412, 413, 439, 440, 482, 509, 511


Maclean, j\L,658
McMahon, J.,64, 195, 200, 505, 507, 581

MacRobert, T. M., 197


Maggi, G. A., 13
Mabnst^i, C. J.,99, 173,

IS.J,187, 188, 196,

203

Manfredius, G.,

92
50"5

THEORY

794

OF

BESSEL

March, H. W., 56, 225, 449


Marcolongo, R., 135
Marshall, W., 505, 506
Mathews, G. B., 64, 65, 78, 194, 206, 454,
480, 655, 656, 657, 658, 660

Maxwell, J. Clerk, 125


H. D., 550

FUNCTIONS

Pearson, Karl, 98, 99, 419,


Peirce,B. 0., 501, 660, 664

421

Perfes,J.,44
Perron, O., 154
Petzval,J.,49
Phragm^n, E., 358

Mayall,R.

Picard,C. E.,93, 94

Mehler, F. G., 65, 155, 157, 169, 170, 180,


183,425, 431, 455, 475, 476
Meissel,D. F. E., 7, 145, 204, 226, 227, 229,
232, 233, 234, 247, 391, 521, 557, 558, 561,

Pincherle,S.,190, 196, 271, 274, 386, 526,

564, 572, 655, 656, 658, 660, 662


Mellin, R. Hj.,190, 196
M. G., 83, 497
Mittag-Leffler,
Molins, H., 106
Moore, C. N., 479, 578, 579, 597, 649
Morton, W. B., 65, 66

Murphy, R., 91, 156,

157

W. {seeLebedeflF)
Myller-Lebedeff,

Nagaoka, H., 340, 633, 634


16, 19, 22, 23, 30,
Neumann, Carl Gottfried,
31, 32, 33, 34, 36, 37, 46, 59, 60, 65, 66, 67,

68, 69, 70, 71, 73, 128, 143, 150, 151, 155,
271, 273, 274, 276, 277, 278, 280, 281, 284,
286, 290, 291, 292, 345, 358, 359, 361, 363,
365, 386, 418, 424, 440, 441, 453, 455, 456,
470, 471, 473, 474, 475, 476, 522, 523, 524,
525

Neumann,

Friedrich

E., 154

F. W., 663, 664


Newton, Sir Isaac, 120
Nicholson, J. W., 107, 108, 145, 146, 149,
150, 189, 226, 229, 231, 247, 248, 249, 250,

Newman,

252, 262, 329, 332, 413, 415, 425, 426, 431,


440, 441, 446, 448, 505, 656

Nicolas, J.,77, 84
Nielsen, N., 24, 44, 49, 64, 73, 74, 77, 82, 83,
132, 142, 145, 148, 149, 154, 169, 224, 297,

528

Plana, G. A. A., 10, 38, 42, 45, 49, 53, 95, 96,
99, 195, 554
Plummer, H. C, 270, 552, 555
L., 100, 101, 297, 346,
Pochhammer,

410

Pocklington,H. C, 537
Poincare, J. Henri, 236
Poisson,S. D., 6,9, 10, 11, 12, 13, 24, 25, 38,
47, 49, 52, 67, 68, 69, 73, 95, 96, 160, 173,
183, 185, 186, 187, 194, 195, 308, 369, 477,
501

Porter, M. B., 299, 477, 480, 485, 515, 517


Preece, C. T., 27
Puiseux,v., 559

Raffy,L., 94
Ramanujan, S.,382,

449

Rawson, R., 91

Rayleigh(J.W. Strutt),Lord, 50, 55, 56, 74,


95, 137, 155, 157, 189, 230, 231, 233, 331,
333, 374, 389, 395, 419, 421, 477, 502, 510,
511, 616, 618, 660
Riccati,(Count) J. F., 1,2,3,85,86,87, 88,94
Riemann, G. F. B., 80, 158, 172, 203, 229,
235, 427, 457, 486, 623, 637,

649

Riesz,M., 606, 614


Rodrigues,O., 27
Rohrs, J. H., 10
Rudski, P.,477, 508
Russell,A., 81, 82, 204
Rutgers,J. G., 373, 374, 375, 376, 380,
Rybczynski, W. von, 56, 225, 449

579

298, 299, 315, 350, 355, 357, 359, 392, 405,


455, 465, 522, 523, 525, 526, 527, 528, 571,

Sasaki, S., 507

572, 574, 597, 622, 629, 636

Savidge,H. G., 82, 204, 658


Schafheitlin,
P.,64, 137, 142, 168, 169, 207,

NiemoUer,

F., 57, 68,

195

215, 373, 391, 392, 398, 401, 402, 406, 408,

Olbricht,R., 158,

481

Oltraraare,G., 173

Orr,W. McF., 145, 146,206, 224, 454,455, 579


Otti,H., 71, 274, 286, 341

421, 447, 477, 479, 482, 485, 487, 489, 490,


491, 492, 493, 494, 508, 510, 543

Scheibner,W., 6
L., 10, 14, 27, 28, 30, 32, 33, 63, 64,
Schlafli,
65, 67, 72, 79, 90, 91, 143, 145, 151, 160,

Panton, A. W.,

305

Paoli,P.,53, 95, 186


Parseval, M. A., 9, 21, 24, 68, 105, 229, 358,
359, 384

171, 174, 175, 176, 178, 179, 181, 185, 195,


196, 215, 216, 228, 253, 274, 276, 278, 284,
285, 286, 288, 289, 290, 341, 342, 344, 345,
508, 577, 579, 581, 582, 583, 585

INDEX

GENERAL

[7^henumbers
Addition

358-372

theorems,

(Chapterxi);

the

referto

jmges.]

coefficients of order

for Bessel

zero,

128, 359; for Bessel

the first kind

(Gegenbauer'stype),362, 367;
of the first kind (Graf'stype),130, 143, 359 ; for Bessel functions
or cylinder
kind (Gegenbauer'stype),363 ; for Bessel functions or cylinderfunctions of any
of any
functions
of two
functions
kind (Graf'stype),143, 361 ; for hemi-cylindrical
functions, 354; for Lommel's
rived
289; integrals devariables, 543; for Scblafli's function r"(;), 344; for Schlafli's ijolynomial,
of, 128, 130, 361, 363, 366; specialand degenerate forms
from, 367 ; phvsicalsignificance
coefficients of order
for Bessel functions

29; for Bessel

",

of

functions

of, 366, 368


188 ; expressedin terms
of Bessel functions of order one-third, 192 ; generalisedby
Hardy, 320; Hardy's expressions for the generalisedintegralin terms of the functions of Bessel,

Airy's integral,
Anger and
Analytic

Weber,

321

to tables

; references

Anger's function
Jvi"),308; connexion
by, 312; integrals
expressed in terms
Airy's integi-al
(generalised)
by, 321;
largeargument and order, asymptotic
to Bessel

Approximations

function, 310; differential equation satisfied


formulae
for, 311; representation of
with largeargument, asymptoticexpansion of, 313; with
expansion of, 316
with

Weber's

of, 312;

zero

(extensions due

to

of largenumbers
to functions
largedegree, 65, 155, 157, 158 ; to remainders

regions),248;
of
of

series of

Method

of

positiveterms,

steepest

8.

also Asymptotic

See

of

Asymptotic

Bessel

function

asymptotic expansions,
Method

of

; to the

213

stationary

sum

phase and

(ind

series

Kapteyn

"/*;"/Schlomilch

series

series
series

(forcomplex variables);
(forreal variables)

defined, 40

approximations to

expansions,

204; for Bessel

in

expansions,

descents

Arbitrary functions, expansions of, see Neumann


Dini
series, Fourier-Bessel
series, Neumann
Argument

recurrence

with largeargument, 10, 12 ; to Bessel functions


Meissel),226, 227, 232, 247, 521 ; (in transitional
tions
(Darboux),233; (Laplace),421; to Legendre func-

coefficients of order

6, 7
largeorder (Carlini),

of

of, 659

asymptoticexpansions of Bessel functions, 200

with

associated

of numbers

theory

coefficients of order

remainders

in, 213; conversion

into

convergent series,

with

largeargument, 10, 12, 194; for Bessel functions


of arbitrary
order with largeargument,
194-224
(Chapter vii); (functionsof the first and second
of the third kind),196; (functions of the third kind by Barnes' methods),
kinds),199 ; (functions
220 ; (functions
of the third kind by Schlafli's methods), 215; (functions
with imaginary argument),
202; for Bessel functions with order and argument
both large,225-270
(Chapterviii);
(ordergreater than argument),241 ; (orderless than argument), 244 ; (ordernearlyequal to argument),
of
245; (ordernot nearlyequal to argument, both being complex),262; for combinations
and products of Bessel
of largeargument,
functions
squares
221, 448; for Fresnel's
integrals,
545 ; for functions of Anger and Weber
(ofarbitraryorder with largeargument), 313 ; (with order
and argument
both large),
functions
of two
316; for I,omniel's functions, 351; for Lommel's
variables,549; for Struve's function (ofarbitraryorder with largeargument), 332; (with order
and argument
both large),
333 ; for Thomson's
and bei (z),
203 ; for Whittaker's
functions, ber (,:)
zero

function, 340;

magnitude of remainders
in, 206, 211, 213, 236, 314, 332, 352, 449; sign of
in, 206, 207, 209, 215, 315, 333, 449.
See also Approximations

remainders
Basic

appliedto

numbers

Bateman's

type

Ber (:).
(-:),

Bei

Bernoullian

See

Thomson's

polynomials

coefficient

functions, 43

integral,379,

of order

382

functions

associated

(Daniel)solution

Bernoulli's
Bessel

Bessel

of definite

with

Poisson's

of Eiccati's

zero,

equation, 85, 89, 123

Jy (z),3, 4;

of),5, 12, 59, 60; expressed as limit


uniform
heavy chain and, 3, 4 ; Parseval's
expansion of, 10, 12, 194;
Bessels

differential

zeros

49
integral,

differential

tion
equation satisfied by, 4, 5 ; (generalsoluLegendre function, 65, 155, 157; oscillations of a
representing,9 ; with largeargim^ent,asympintegral
totic

of

of, 4,

5.

See also

Bessel

functions

coefficients,Bessel

and

equation

coefficients
J" (z),5, 6, 13, 14-37 (Chaptern); addition theorem
for, 29; Bessel's integi-al
for,19; expansion in power series of, 15; generating function of, 14, 22, 23; inequalities
satisfied
by, 16, 31, 268; notations for, 13, 14; order of, 14; (negative),
formulae
for, 17;
16; recurrence

Bessel

square of, 32 ; tables of (oforders 0 and


order and argument), 664, 746; tables
coefficient

of order

zero,

Bessel's

1) 662, 666-697
,

of

differential

(oforder n),664,

730-732

(references to), 654, 655, 656, 658.


equation

and

Bessel

See

(withequal
also

Bessel

functions

(Chapter in) ; argument of, defined, 40; differential equations of order


satisfied by, 106 ; expressed as limits of Lam^
functions, 159 ; expressed
P-functions, 158 ; historyof, 1-13 (Chapteri); (compiledby Maggi and by Wagner),

functions, 38-84

Bessel

higher than

the second

as

limits of

13

; indefinite

terms

integralscontaining, 132-138 ; order of, defined, 38, 58, 63, 67, 70 ; rank of, defined,
129 ; relations between the various kinds of, 74 ; representation of
cylinderfunctions in
of, 82; solutions of difference equations in terms
of, 83, 355; solutions of Laplace's

GENERAL

797

INDEX

equation containing,83, 124 ; solutions of the equation of wave


temi
relations connecting, 300 ; with negativeargument, 75.
and
foUou-int)nitri,.",
Cylinder functions

motions
See

also

containing,123
the two

; tbree-

preceding"iud

ten

Bessel

functions of the first kind, -/,.


for, 143, 359, 362, 363, 367, 368 ;
(:),38; addition theorems
Barnes'
type of integralrepresenting,190 ; Bessel's type of integialrepresenting,176 ; cut in
satisfied by, 38 ; expansion of, in
plane to render uniform, 45 ; differential equation (Bessel's)
ascendingseries, 40 ; expansion of, in descendingseries, xee Asymptotic expansions ; expressed

generalisedhypcrgeometricfunction, 100, 101 ; expressed as the limit of a hypergeometric


154 ; expressed
as the limit of a Legendre function, 156; (physical
significanceof),155;
of, 45 ; generalisations
i)o]ynomial,302 ; functional ])roperties
expressed as the limit of a Lommel
satisfied by, 49, 255, 259, 270, 406 ; infinite intogi-als
of, 43, 44, 308-357 ; Inequalities
containing.
Chapter xiii, paxdm ; of complex order, 46 ; of order u "\-\, 41 ; (expressedin finite tei-ms),52,
55 ; (notations
for),55, 80 ; Poisson's integralrepresenting,47, 48 ; (modificationsof),161, 163,
of Bessel
functions
164, 169, 170 ; products of, see Products
expressed as a
; quotient of two,
as

function,

continued

fraction, 153, 154,

303

formulae

; recurrence

for, 45,

294

; relations

with

Lommel's

representedbv integralscontaining Legendre functions, 173, 174; symbolic


formulae
for, 170 ; tables of (of orders 0 and 1), 662, 666-697 ; (of order n), 664, 730-732 ; (of
of computing), 664;
order n-\-\).664, 740-745; (of order ^),664. 714-729; (oforder -^, method
to),
(with equal order and argument), 664, 746; (zeros of),664, 748-751; tables of (references
654, 655. 656, 658, 659, 660 ; Weierstrassian
product representing,497 ; with largeargument, xee
polynomial, 297;

expansions

Asymptotic

of, itee

; zeros

Zeros

of Bessel

functions

second
kind, ""(-) (afterHankel),57, 63; G" {z)(afterHeine), 65; r''"(s)
67; ir(~)(afterWeber-Hchlatli,the canonical form),63; addition theorems
for, 144, 361, 365, 368; Bessel's type of integralrepresenting,177; component parts of, 71, 72,
satisfied by,
63; differential equation (Bessel's)
840; continuity of [qua function of their order),
59, 63 ; expansion of, in ascendingseries,59, 60, 61, 69, 72 ; expansion of, in descendingseries,
nee
containing functions of the first kind, 5,
Asymptotic
expansions ; expressedas an integral
133, 382, 433 ; infinite integrals
containing, 385, 387, 393, 394, 424, 425, 426, 428, 429, 430, 433;
Poisson's type of integralrepresenting,68, 73, 165; (modifications
of),169, 170; products of,
221. 441, 446; (asymptoticexpansions of), 221, 448;
149; (rej)reRented
by infinite integrals),
formulae
for, 66, 71 ; representedby integi-als
recurrence
containing Legendre functions, 174 ;
symbolicformulae for, 170 ; tables of (oforders 0 and 1),662, 666-697 ; (oforder "),664, 732of computing), 664 ; (with equal order
735 ; (of order h), 664, 714-729
J, method
; (of order
and argument), 664, 747 ; (zerosof),748-751
to, 655, 656, 658 ; with large
; tables of, references
of, t^ee Zeros
of
xee
expansions ; with negative argument, 75 ; zeros
argument,
Asymptotic
See also Neumann's
Bessel
functions.
polsoiomial

Bessel

functions

of the

(afterNeumann),

third
the
kind, Hv'^' (z),Hi.'-' (z),73; Barnes'
integi'als
representing,192
representing,178; Poisson's type of integral
representing,166; (modiintegiul
fications
of),168, 169. 170; representedby integralscontaining Legendre functions, 174;
664, 714symbolic formulae for, 170 ; tables of (of orders 0 and 1), 662, 666-697 ; (oforder ^,),
729; tables of (references
to),657; with large argument, asymptotic expansions of, 199, 210,
and order, asymptoticexpansions of, 244, 245, 262; with negative
215; with large argument

Bessel

functions

Bessel's

type

of

of

argument, 75
Bessel

448,

functions
515

747 ; tables of
Bessel

functions

of a
sta]iility
751.
659

Of

and

(references
to)
,

whose

argument

order

equal, approximationsto, 229, 231, 232, 259, 260,


representing,258 ; tables of, 658, 664, 746,
integrals

are

658

is

fraction.

Of

orders

"i

190;
(and Airy's integi-al),

(and

the

tables of, 664, 714-729


of,
to), 659 ; zeros
; tables of (references
"?, tables of (references
to),659. Of orders "^, "1/, tables of (references
to),
of (references
fractional orders, tables of zeros
to), 502, 660. See also Bessel

vertical

Of orders
small

functions

order

whose

asymptotic expansions of, 245

whose

order

pole),96

is

"

(n + ^)

(Chapterviii)
large, 225-270
; asymptotic expansionsof, 241, 244,
(elementary)
approximation to, 6, 7 ; (extended by Meissel),226, 227 ; Horn's
of stationaryphaseapplied
approximation to, 225 ; Laplace's approximation to, 7, 8, 9 ; method
of steepest descents
applied to, 237 ; miscellaneous
to, 232 ; method
propertiesof, 252-261 ;
See
talDles of (referenceto),658; transitional formulae
of, 513, 516, 517, 518.
for, 248; zeros
"^Iso Bessel functions
order
and
are
wnose
argument
equal
Bessel

245,

Bessel

functions
262

whose

order

is

order

is

Carlini's

functions

whose

55, 80; tables of. 664, 740-745;


Bessel

notations

for,

argument,
imaginary
I,.{^),
Ki"{z),Ki(c), 77, 78; differential equation
representing(of Bessel's type),181 ; (ofPoisson's
type),79, 171. 172;
integrals
property of, 446 ; of order =" [iif A),
equivalenceof various types),185-188 ; monotonic

functions

satisfied

in finite terms, 52
^n ,- ^),10, 52, 80 ; expressible
tables of (references
to),658, 659

with

by, 77

(proofof
80 ; recun-ence
formulae, 79 ; tables of lof orders 0 and 1),663, 698-713 ; (oforder i), 664, 714664, 736, 737-739; tables of (references to),657, 658; with large
729 ; (of various
integralorders),
of, 511 ; (computation of), 512; (referencesto),
argument, asvmptotic expansions of, 202; zeros
660
Bessel's

differential

for functions

with

equation,

1,

19 ;

38 ;
(generalised),

imaginary argument,

77 ; fundamental

for functions

system

of order

of solutions

5, 12, 59, 60;


of. 42, 75 ; has no

zero,

798

FUNCTIONS

BESSEL

OF

THEORY

the functions satisfyingit


and only when
when
; soluble in finite terms
^, 52, 119; solution of. in ascending series, 39, 40, 57, 59-61 ; solution of, in
descending series"see Asymptotic expansions ; symbolicsolution of, 41 ; transfoi-mations of, 94,
117
integral,
algebraic
of order

are

97.

n +

also

See

and

coefl"cients

Bessel

Bessel

functions

and extensions
of, see
integral representing Bessel coefficients,19, 21 ; generalisations
and
Weber's
of,
function; modifications
Anger's function, Bourget's function, Bruns' function
tion
to represent Bessel functions of arbitraryorder,175, 176, 177, 178, 181 ; Theisinger'stransformaSee
177 ; used to obtain asymptoticexpansions,215.
of, 184 ; used in theoryof diffraction,
also Parseval's
integral

Bessel's

Bounds,

Inequalities

see

upper,

Bourget's

function

J"

^.

(z),326

; differential

equationsatisfied by, 327

formulae

recurrence

for,

326

Bruns'

function

Carlini's

(z;v, k),327
for Bessel

approximation

Cauchy's numbers
Cayley's solution

324;

jV_"_i.
,",
of Riccati's

Chain, oscillations of

largeorder, 6, 7;

formulae

by Meissel,226, 227

extended

for, 325

equation, 88

uniform

of

functions
recurrence

heavy,3, 4, 576

Eiccati

Cognate

equations, 91
functions
variables, expansions of arbitrary

Complex

functions

Composition

of Bessel

of the second

functions

kind

of

and

series

of, see Kapteyn

Bessel functions, 483 ; of Bessel


Complex
zeros
Lommers
polynomials, 306
of

with

imaginary argument,

order,
integral

503;
(Stokes'),

zeros

phase, Sehlafli's method

Constant

of, 216

of arbitrary(Stokes'
discontinuity
phenomenon), 201, 203, 238,
with Sehlafli's polynomial,288
Continuants, connected

Constants,

Continued

representingquotients of Bessel functions,153

fractions

336

of, 154, 303

; convergence

of asymptoticexpansionsinto, 204

conversion
series, Hadamard's
integral for Sehlafli's polynomial,288.

Convergent
Crelier's

511 ; of

340

500, 502;
of Bessel functions by various methods
(Graeffe's),
(Sturm's, for the smallest zero),516. See also Zeros of Bessel functions
of

Computation

series

Neumann

See

also Neumann's

integral

for

Neumann's

polynomial
of solutions

Cross-ratio
of

Cube
Cut

of Eiccati's equation, 94
function,expansion of, 149

Bessel

for definition of Bessel

necessary

motion
(circular),

Cylinder

of heat

functions, 45, 77

in, 9, 10, 576, 577

with Bessel
4, 82, 480; addition theorems, 143, 361, 365; connexion
'^^(^),
83 ; rank of, 129; solutions of differentialequations of order
originof the name,
higher than the second by, 106 ; three-teiTn relations connecting, 300. See also Bessel functions

Cylinder functions,

functions,83
and

Hemi-cylindrical

Darboux'

See

Definite

functions

of

functions

under

by geometrical methods, 374, 376, 378


also

Infinite

largenumbers, 233
the integralsign, 373-382
; the Ramanujan-Hardy method

Difference

representing
specialfunctions,see Bessel functions
polynomials,294 ; Wronskian,
with linear coefficients)solved by means
equations (linear
representingLommel's

Functional

equations

and

of

evaluation,

expansions, 605

coefficients,fractional,
107,

and

Integrals

42, 76, 77
of Bessel

functions,83.

See

formulae

Recurrence

of Fourier-Bessel
Differentiability
Differential

(Chapter xii);

integrals

integrals

Determinants,
also

approximatingto

integrals, containing Bessel

evaluated
382.

of

method

Definite

functions

; of

specialSchlomilch

series,635

125

Riccati
to the generalised
linear of the second order, equivalent
(ordinary),
oscillation
of
106
of
order
solved
Bessel
than
the
second
functions,
by
higher
;
;
solutions of, 518 ; satisfied by the productof two Bessel functions, 145, 146 ; solved by elementary
transcendants, 112; symbolic solutions of, 41, 108. See also under the names
of special
functions a7id
equations,such as Bessel's differential equation, and under the names
of various

Differential

equations

equation, 92

polynomials satisfyingdifferential
equations,such as Anger's function
solution of by an integi-al
equations (partial),
containing Bessel functions, 99

Differential

Laplace's equation

and

Diffraction, theoryof, connected


Schlomilch
series,633 ; with
Diffusion
Dini

of salts

expansion,

in
580.

equation

Wave-motions,
with

See also Dini

see

also

177; with
188; with Bessel's type of integral,
Airy's integral,

Struve's

and
liquid,

of

functions, 417

infinite
series

containing Bessel functions,


integi-als

437

799

INDEX

GENERAL

tion
series, 577, 580, 596-005, 615-617
(Chapterxviii),651-653 ; expansion of an arbitraryfuncof a real variable into. 580, 600 ; methods
of complex variables applied
of theory of functions

Dlni

theorem, analogue of,


lemma,
to, 596, 602; Kiemann-Lcbesgue
analogue of, 599; Riemann's
("49; summabilityof, 601, 615 ; uniformity of convergence
of,601, 604 ; uniqueness of, 616, 651 ;
value at end
Dirichlefs

of range,

602

factor, 406

discontinuous
of

Discontinuity

factor

Discontinuous

integrals, 398, 402, 406, 408, 411, 415,

Domain

A'

Du

Raymond's

Bois

Electric

(Kapteyn's),559

421

diagram of, 270

integrals with

integrandsexpressedin
oscillatoiy

of Bessel

terais

functions, 183

56, 226, 449

waves,

radiation, 551, 556

Electromagnetic
Elementary

336

(Stokes'phenomenon), 201, 203, 238,


405
406; (Weber's),
(Dirichlet's),

constants
:ul)itniry

Discontinuous

definition

transcendants,

of. 111

; order

of, 111

of differential

; solution

equationsby,

112

Equal
of

order

and

Eulers

of Bessel

functions

; tables referred

406 ; (Weber's),
405
(Dirichlet's),
productsof Weievstrassian, 497

analogue of, for Fourier-Bessel

Fej^r'stheorem,

Bessel

terms,

as

with, 231, 232, 258, 260

; tables

of, 746, 747

; tables

equation, 87

function, tables of, 698-713

Factors, discontinuous

not

functions

664

of Eiccati's

solution

Exponential

Finite

Bessel

argument,

to),658,
(references

to, 663, 064


; Neumann's

e"

or

2),22

expression

expansions, 610

of other orders
; Bessel functions
; solutions of Eiccati's equation in, 85, 86, 89 ; the solution of Eiccati's
and their limit, 123
possibleexcept in Daniel Bernoulli's cases
of order

functions

(n+ i) expressed in, 52

"

119
expressible,

so

equation in, not

problem of random,

Flights,

419

580.

Fourier-Bessel

expansion,

Fourier-Bessel

functions,

4,

Fourier-Bessel

integrals,

^^ee

See

alao

Fourier-Bessel

series

84

Multiple

infinite

integrals

function
of a
(Chapterxviii),649-651 ; expansion of an arbitrary
real variable into, 576, 580; Fejer'stheorem, analogue of, 610; Kneser-Sommerfeld
expansion
of a combination
of Bessel functions
of theory of functions of complex variables
into, 499 ; methods
appliedto, 582, 607 ; order of magnitude of terms in (Sheppard'stheorem),595 ; Eiemanntheorem, analogue of, 649 ; summability of, 578,
Lebesgue lemma, analogue of, 589 ; Eiemann's
series, 576-617

Fourier-Bessel

of, 593, 594 ;


606, 613 ; term-by-term differentiation of, 578, 605 ; uniformity of convergence
015 ; uniformityof summability of,612 ; uniqueness of,616, 649 ; value at end of
(near origin),
594, 603

range,
Fractional

differential

107,

coeCacients,

125

Fresnel's integrals, 544 ; asymptotic expansion of, 545 ; tables of, 744, 745 ; tables
and minima
of, 745 ; tables of (references
to),660, 661, 664
Functional

equations

Functions

of

8, 421. See
of steepest
Fundc^mental

deliningcylinderfunctions,

by Nielsen,
generalised
;

355

approximations due
of

stationary phase

to
and

Laplace,
Method

descents
of solutions

system

Gallop's discontinuous
Gamma

82 ;

approximations due to Darboux, 233


large numbers,
ahi" Approximations,
Asymptotic expansions, Method

of maxima

infinite

of Bessel's

integrals,

differential

equation, 42, 75, 78

421

functions, representationof Bessel functions by integrals


containing,190, 192, 221 ; applications
of asymptotic expansions, 220, 223 ; applications
to determination
to evaluation of infinite

integrals,383, 434, 436


functions by integrals
functions, representationof Lommel's
containing,351
of asymptotic expansions, 352
to determination

jBamma

Gegenbauers

addition

Gegenbauer's

discontinuous

Gegenbauer's

function

Gegenbauer's

polynomial

satistied
fornxulae

by,

283

theorem

Cj
;

for Bessel

infinite

applications

functions,362, 363, 367

integrals, 415,

418

{z),50, 129, 363, 365, 367, 308, 369, 378, 407


tion
A"^v{t),283 ; contour integralscontaining, 284, 524; differential equafunction, 351 ; recurrence
equivalence with specialforms of Lommel's

for, 283

293, 525
Gegenbauer's polynomial I'n.n.i'
(t),
Gegenbauer's representation of Jt,(z)by a double integralresembling Poisson's
Gegenbauer's
type ot delinite integral,378
Generalised

hypergeometric

functions,

-fee

Hypergeometric

functions

51
integral.

(generalised)

800
integrals

Generalised

Giuliani's function,

addition

Group

velocity, 229
of

Hankers

functions, 485

406
J^i^),31 ; generalised,
of Aiiy's integral),
320;
(
generalisations
Ei,^{a),
Si,^(a),
Ci,^{a),

upper

for

Anger and Weber, 321,

of Bessel,

of functions

terms

502

functions, 359, 361

integrals, 384, 386, 389, 390, 393, 395, 424, 427, 428, 434
hound

functions

Hardy's

for Bessel

of Bessel

zeros

function

calculating
zeros, 500,

theorem

infinite

Hansen's

548,
sr-r Bourget's

of

Grafs

Growth

functions of
Hardy's integrals representingLommel's
382
definite
of
method
Hardy's
integrals,
evaluating

of, 9, 10, 450, 576, 577, 616


S" (~), defined,353

conduction

Heat,

Hemi-cylindrical

functions

; addition

353

theorem

variables,546

two

expressedin

of order

zero,

expressedas

Bessel

functions,154
in terms
of, 100, 101;
expi-essed
; Sharpe'sdifferential equation

102

by, 105
Bessel

argument,

Struve's

of the function

terms

functions
Hypergeometric
90, 100; Bessel functions
(generalised),
notations for, 100 ; relations between
(Rummer's formulae),101,

Imaginary

expressedin

322

for, 354

functions, limitingforms

Hypergeometric

solved

282

549

Gilbert's integrals,

GraefiFe's method

188, 441, 463, 464


(withimpliedexponentialfactor),
coefficients,14, 22, 23 ; of Neumann's
polynomials,281,

of Bessel

function

Generating

FUNCTIONS

BESSEL

OF

THEORY

functions

with,

functions

Bessel

.'iee

functions

with

imaginary

argument

with, 329, 332

sign, 132-138, 350, 581


integrals containingBessel functions under the integral
of, 744, 745, 752 ; tables of (references
to),660, 661, 664

Indefinite

; tables

Inequalities satisfied by Bessel functions, 16, 31, 49, 255, 259, 268, 406; by Neumann's
nomial,
polyof Bessel functions,485, 489,
273, 282 ; by Struve's function, 328, 337, 417 ; by zeros
490, 492, 494, 515, 516, 521
Infinite integrals
383

;
under

of
Eamanujan's type (integrals
the

of

names

421

of

Integrals, expressed in

terms

of

; values

with

polynomialsof

the

Definite

functions

of Lommel's

of, deduced

integrals

Bessel

of Bessel

zeros

functions

addition

from

and

Neumann

(uul Infinite

and

infinite

integral

cylinderfunctions,4, 478, 481, 494,

495

variables,540; expressed in terms of


312; Fresnel's,544, 545, 660, 661, 664, 744, 745; Gilbert's,

Anger and Weber,

548, 549

integrals,e.p. Lipschitz-Hankel

various

Infinity of the number


the functions

continuous
the integral
sign,383-449 (Chapterxrn); dis441 ; methods
of evaluating,
described,
generalised,
449.
See also
functions with respect to their oi'der),

under

containing Bessel functions

398, 402, 406, 408, 411, 415,

of two

theorems,

Gegenbauer under

367 ; with oscillatory


183 ;
integrands,
the integral
sign, 277, 285. See also

integrals

Interference, 229
Interlacing of

Jacobi's

of Bessel

zeros

Irrationalityof

to

Kapteyn's

6, 26

cos

and

of

cylinderfunctions,479, 480,

connecting sin"j5 with the {n

transformation

respect

functions

481

90, 485

tt,

erroneouslyattributed

K, 559

domain

to

coefficient of sin2"-i^ with


l)thdififerential
Eodrigues,27 ; various proofs of, 27, 28
-

diagram of, 270

Kapteyn series, 6, 13, 551-575


(Chapterxvii); connexion with Kepler'sproblem,551 ; expansions
into, derived from Kepler's problem, 554, 555 ; expansion of an arbitraryanalyticfunction into,
570 ; fundamental
expansions into, 557, 559, 561, 564, 566, 568, 571 ; Kapteyn's domain
A', of
outside and on the boundary of
of convergence
of, 559 ; (diagramof),270 ; nature
convergence
K, 574 ; second kind of, 572

Kapteyn's polynomial
Kapteyn's

Kepler's problem,
Kinds

of Bessel

6, 551, 554

functions,

Kneser-Sommerfeld

Kummer's
Lam6

568; expressed in
"" (t),
380
integral,

terms

polynomial,569

of Neumann's

type of definite

formulae

; Bessel's

40
(first)

expansion

of

solution

of,

13 ;

solution of, 6
Lagi-ange's
73
; (third)

(second)58, 63, 64, 65, 67

combination

of Bessel

functions

as

Fourier-Bessel

series,499

connecting generalisedhypergeometricfunctions,101, 102

limitingforms expressedas Bessel functions,159


generalsolution due to Parseval, 9 ; generalsolution due to Whittaker,
involvingBessel functions, 83, 124; used to obtain addition theorems for Bessel

functions,

Laplace's equation,
solutions
127

124 ;
tions,
func-

of

methods

Laplace's

approximatingto

(tho

Large

mctliods

numbers,

order,

Weber's

Anger's

mv

approximation

of

"'(*/

Approximations

functions

of

large numbers,

8,

421

280, 395

Laplace's transformation,
Large

801

INDEX

GENERAL

Asymptotic
Bessel

function,

to functions

of

(Darboux),283

8,
(Laplace),

421.

See

expansions
fimctions

whose

order

is large,

function

Struve's

mid

function

Lebesgue's lemma,

Riemann-Lebesgue

-fee

lemma

notation
for, 1.56; integialscontaining,50, 173, 174, 339, 475; limits
Legendre functions, Ikunes'
of, expressed as Busscl functions, 65, 155, 157 ; (physical
significanceof),155 ; of largedegree,
of, 387.
approximationsto, 15H ; relation between two kinds of, 174 ; Whipple's transformation
See also Gegenbauer's
function
C'"''
{z)

389
integral, 384 ; generalised,
in finite terms, 350; integrals
of expression
345, 347; cases
'^^,v(z),
S^,i,(z),
cases
formulae, 348 ; special
expressible
by the polynomials
representing,346, 350 ; recurrence
with ^"c
odd negative integer,
of Gegenbauer, Neumann
and Schliitii,
an
350; specialcases
348 ; with largeargument, asymptotic expansion of, 351
Lommel's
functions
of two
variables,
?v(w, z), V:,{w,z),537, 538; addition formulae for,543;
formulae, 539 ; special
540, 546 ; reciprocationformulae, 542 ; recurrence
integi-als
representing,
infinite

LipscMtz-Hankel

Lommel's

functions

of, 581, 752; tables of 752; tables referred to, 660; with large argument, asymptotic expansions
of, 549
notation
Lommel's polynomial Rm, v [z),
294, 295 ; differentialequationsatisfied by,297 ; Hurwitz'
303 ; limit of, expressedas a Bessel function, 302 ; of negative order, R-m,v [z],
299;
(]m,v{z),
in Hurwitz'
formulae
formulae, 298; recurrence
notation,303; relations with Bessel
recurrence
of, 304, 305, 306
functions, 295, 297, 302; three-term relations connecting, 300, 301; zeros
ease

in

of remainders

Magnitudes

asymptoticexpansions,206, 211, 213, 236, 314, 332, 352, 449


table of, 745; of integrals
of Bessel
integrals,

functions, 488; of Fresnel's

of Bessel

Maxima

tions,
func-

table of, 752


Mean

anomaly,

series of, 6, 13, 552, 554, 556


of an orbit in trigonometrical
Legendre functions, limitingform expressedas Poisson's
representing

expansions of elements

Mebler-DiricMet

integral

157
integral,
Mehler-Sonine

Bessel functions, 169,


representing

integrals

approximations

Meissel's

to

Bessel

functions

of

170

large order, 226, 227, 232, 247, 521

types

of

Kapteyn series,557, 561, 564, 566


of

vibrations

Membrane,

circular, 5, 576, 618

; vibrations

of

510
sectorial,

216
(Schlalli's),
of stationary phase, 225, 229 ; appliedto Bessel functions, 231, 233
Method
235 ; appliedto Bessel functions, 237, 241, 244, 245,
Method
of steepest descents,
functions of Anger and Weber, 316 ; appliedto Struve's function,333; connexion
Method

of constant

method

of

approximation,

of Bessel

Minima

phase

262 ;
with

appliedto
Laplace's

421

functions, 488

; of Fresnel's

table of, 745


integrals,

; of

of Bessel
integi-als

tions,
func-

752
446
{v),257 ; of Jv {v)and JJ {v),260 ; of A'^(.i),
{vx)lJi,
453, 470 ; (generalised
by Neumann,
Multiple infinite integrals, 450-476 (Chapterxiv); investigated
by Orr),455; (modifiedby Weber), 468; Riemannby Hankel), 453, 456, 465; (generalised
Lebesgue lemmas, analoguesof, 457, 471 ; Weber's type of, 450
Monotonic

properties of J^

(Chapterxvi);expansion of an arbitraryanalytic function into, 523;


30, 31, 36, 69, 71, 151 ; Laurent's expansion,analogue of, 524;
(special
series),
Pincherle's theorem
the singularities
of, 526; specialseries,18, 23, 25, 33, 34, 35, 12.^. 130,
on
138, 139, 140, 527, 581 ; Webb-Kapteyn
(realvariable)
theoryof, 533. See also Addition theorems
series, 522-537

Neumann

525
generalised,

and

Lommel's

Neumann's

functions

factor

f,i (

^x-Neumann'sintegral

of two
1

for J

or

variables

2), 22

J {z),32

; for Neumann's

polynomial,278,

280

569 ; connected
271, 272, 273 ; connected with Kapteyn's polynomial.
polynomial 0" {t),
with Neumann's
polynomialt2"{t),292 ; connected with Schliifli'spolynomial,285, 286 ; contour
containing,277 ; differential equation satisfied by, 276 ; expressedin terms of Lommel's
integrals
functions, 350 ; formerly called a Bessel function of the second kind, 67, 273 ; generalised
by
generating function of, 281, 282; inequalities
Gegenbauer, see Gegenbauer's polynomial .(",i/(0;
satisfied by, 273, 282 ; infinite integrals
containing,433 ; Neumann's
representing,
integral
formulae
for, 274
278, 280 ; of negative order defined, 276 ; recurrence
Neumann's
polynomial (2,^
polynomial
(f),290, 291; expressedas integralcontaining Neumami's
292 ; Gegenbauer's generalisation
of, see Gegenbauer's polynomial Ihi.^. (t); recurrence
0" (t),
Neumann's

formula
Nicholson's

for, 292
infinite

integrals, 431, 441

Nielsen-Hankel

functions,

Null-functions,

Lerch's

functions

Bessel

st'f

BESSEL

OF

THEORY

802

theorem

FUNCTIONS

tliird kind

of the

382
representing,
integrals

on

representedby Schlomilch

series,

634, 636, 642, 647

analytictheoryof, associated with asymptotic expansionsof


fonnulae
for, 325
Cauchy's,324 ; recurrence

Numbers,

Numbers,
of

Order

equations,

differential

Ordinary

with regardto, 449


integrals

equations

equations, 518

heavy chains, 3, 4, 576


with, expressedin terms of
Keymond's integrals

Bois

Oscillatory integrands, Du

functions,200

; of uniform

5, 510, 576, 618

of membranes,

Oscillations

Differential

see

of linear differential

of solutions

Oscillation

defined,38, 58, 63, 67, 70

function

Bessel

Bessel

Bessel

functions,

183

expressedas Bessel functions, 158


Jo (^)i9i 21 ; modifications of, 21
integral representing

P-functions, limitingforms
Parseval's
Partial

equations,

differential

Phase, method

Phase, Schlafli's method


Pincherle's

see

Differential

equations

of, 225, 229; appliedto Bessel functions,231, 233


stationary,generalprinciples

of

theorem

of

constant, 216

integral for Bessel coefficients,12, 24,

Poisson's

defined

of functions
singularities

on

by Neumann

25 ; for Bessel

series, 526

functions, 47, 48, 49

(generalised
by

for Bessel functions of imaginary argument, 80 ; for


for Legendre
Bessel functions
of the second kind, 68, 73 ; limit of the Mehler-Dirichlet
integral
to representBessel functions of any order
into contour
functions as, 157 ; transformation
integrals
(ofthe first kind),161, 163, 164 ; (of the second kind),165 ; (of the third kind),166, 167; (with

Gegenbauer),50

(symbolicform of),50

of the

171, 172; transformations


integral nmJ Struve's function

imaginary argument),
also Parseval's
Polar

change of

coordinates,

used

Bessel

to express

Probl^me

de

of

moments

to obtain

of, used

axes

functions

as

168, 169,
integrals,

contour

transformations

of

Legendre functions,

limits of

170.

51, 374, 376,


integi'als,

See

378 ;

155

464
Stieltjes,

expansion of, 130, 370 ;


functions, 30, 31, 32, 82, 146, 147, 148, 149; Bateman's
31, 150, 221, 438,
expansions of arbitraryfunctions into series of, 525, 572; integrals
representing,
439, 440, 441, 445, 446, 448 ; series of, 30, 151, 152 ; with largeargument, asymptoticexpansions
of, 221, 448

Products

of Bessel

Products

of Weierstrassian

Quotient of Bessel
Radius

of

vector

an

factors,Bessel functions

functions

expressedas

continued

with

functions

anomaly, 6, 13, 552, 553, 554

mean

respectto their order, 449

382
evaluatingdefinite integrals,
419
flights,problem of,
of Bessel functions and cylinderfunctions, 129
Rank
Real
variables, expansions of arbitraryfunctions
of,
series (Webb-Kapteyn theory),
and
Neumann
Schlomilch

Ramanujan's

497

fraction, 153, 154, 303

series of the
orbit,
expansionastrigonometrical

integrals of Bessel

Ramanujan's

expressedas,

of

method

Random

Reality

of

of Bessel

zeros

Reciprocation

for

formulae

Recurrence

functions, 482, 483,

for Lommel's

formulae

functions

Dini

see

series,

Fourier-Beasel

series,

series

511
of two

variables,542

Anger's functions,311

of the first kind, 45 ; for Bessel functions


third kind, 74 ; for Bessel functions with

; for Bessel coefficients,17 ; for Bessel functions


of the second kind, 66, 71 ; for Bessel functions of the
imaginaryargument, 79 ; for Bourget'sfunctions, 326 ;

for Cauchy's numbers, 325 ; for cylinderfunctions, 82 ; for Gegenbauer's polynomials,283 ; for
Lommel's
functions, 348 ; for Lommel's
functions of two variables,
539 ; for Lommel's
nomials,
poly274 ; for Neumann's
298, 303 ; for Neumann's
polynomials0,^(t),
polynomialsii,^
(t),
283; for Schlafli's functions,71, 342, 343; for Schlafli's polynomials,285; for Struve's functions,
329 ; for Weber's
functions, 311 ; for Whittaker's
See also Functional
functions, 339.

equations, Hemi-cylindrical functions


functions, Cailler's,536

mid

Three-term

relations

Reduced

in asymptotic
expansions, magnitudes of, 206, 211, 236, 314, 332, 352
207, 209, 215, 315, 333 ; Stieltjes'
approximations to, 213

Remainders

Repetition
Riccati's

of

of Bessel

zeros

functions

and

differential

with

Bessel's

equation, 1, 90; equation

85;
(D. Bernoulli's),
exhausted
by D. Bernoulli's formula and its limit,123 ; solutions by various
88 ; (Euler),
2, 85, 89 ; (Cayley),
87 ; (Schlafli),
90 ; solved by
(D.Bernoulli),
Riccati's

by James

Bernoulli,1

differential

equation

signsof,206,

of, 479
cylinderfunctions, impossibility

equation, 1, 2, 85-94; connexion


cognate to, 91; limitingform of, 86; soluble cases

series

; transformations

of

soluble

cases

of,

mathematicians
of infinite

means

of, 86

generalised, 3, 92, 94

; cross-ratio

of

solutions,94

equivalence

804

FUNCTIONS

BESSEL

OF

THEORY

of Bessel coefficients (of orders 0 and 1), 662, 666-697;


(of order n),664, 730-732; (with
equal order and argument),664, 746 ; of Bessel functions of the tirstkind (oforders n + i, n
i),
664, 740-741 ; (oforder ^),664, 714-729 ; of Bessel functions of the second kind (oforders 0 and
664, 732-735 ; (of order A), 664, 714-729 ; (withequal order and
1),662, 666-697 ; (oforder ?(),
argument), 664, 747 ; of Bessel functions of the third kind (oforders 0 and 1), 662, 666-697; (of
of Bessel functions with imaginary argument
order ^), 664, 714-729;
(oforders 0 and 1),663,
698-713 ; (oforder )i),
664, 736, 737-739 ; (oforder i),664, 714-729 ; of c^',
663, 698-713 ; of Fresnel's

Tables

664,
integrals,

of
; of integrals
1),663, 666-697

744-745
0 and

(of orders
order

and

of order

Bessel
; of

functions of order zero, 664, 752 ; of Struve's functions


of Bessel coetJicients and functions of integral

zeros

i, 664, 748-751

of Bessel coefficients and functions


derivable from
functions (oforders " + i,
n
i),658, 659 ; (oforders "i,
"1), 659; (of orders "i, "|), 659; of Bessel functions of the second kind, 655, 656,658; of
Bessel functions of the third kind, 657 ; of Bessel functions with imaginary argument, 657, 658 ; of
of Bessel functions and Struve's functions,661 ; of Lommel's
Fresnel's integrals,
661 ; of integrals
functions ber .r and bei x, etc., 658 ; of zeros
functions of two variables,660 ; of Thomson's
of

659;
to)of Airy'sintegral,
(references

Tables

; of Bessel

654, 655, 656, 658

them,

Bessel coefficients,
functions and associated functions, 659, 660
Theisinger'sintegral
representationof Bessel functions,184; of Struve's and Weber's

functions,

338

Thomson's

(SirWiiliam) functions, berz, bei", 81


to tables

81 ; references

of, 658

with

; connexion

; squares

and

Bessel

productsof, 82,

asymptotic expansions of, 203


relations
connecting Bessel functions,cylinderfunctions

Three-term

functions,81

148 ; with

and

sations,
generali-

largeargument,

Lommel's

polynomials,

300, 301

elementary,definition of. 111

Transcendants,

by,

; order

of, 111

; solutions

of differentialequations

112

regions associated

Transitional

Uniformity

Upper

bounds,

see

Wave-motions,
obtain

and

Dini

of, associated

equation

addition

of

largeorder, 248
series,593, 594

; of

Kapteyn series,

; of

series,616, 649, 651

Schlomilch

series,643, 647

Inequalities

fluid,motion

Viscous

functions

series, 632

of Fourier-Bessel

Uniqueness

Bessel

of Dini series,601 ; of Fourier-Bessel

of convergence

575 ; of Schlomilch

with

of,

theorems

with

189
Airy'sintegi-al,

generalsolutions, 125

for Bessel

generalisedto

dimensions, 128

; used

to

functions,129

electric,56, 226, 446

Waves,
Weber's
Weber's

408,

of stationary
; on water, and the method
phase, 229
factor, 405
(H.)discontinuous
(H.)infinite integrals, 391, 393, 395, 396; (discontinuous
types of),398, 402, 405, 406,

411

with Anger'sfunction,310 ; connexion


308 ; connexion
with Struve's
(H. F.)function
E,/(z),
function, 336 ; differential equation satisfied by, 312 ; integrals
currence
expressedin terms of, 312 ; reformi^lae for, 311 ; representationof Airy's integral
by, 321 ; tables of,see
(generalised)
Struve's function
for, 338 ; with largeargument, asymptotic expansionof,
; Theisinger's
integi-al
313 ; with large argument
and order, asymptotic expansion of, 316

Weber's

products, expression for Bessel

Weierstrassian

Whipple's

transformation

Whittaker's

function

339 ; with
Wronskian

functions

as, 497

of

W^

Legendre functions, 387


{z),339 ; differential equation

largeargument,
determinant,

satisfied

by, 339

recurrence

foi-mulae

for,

asymptotic expansion of, 340

42, 76, 77

of Bessel functions, 477-521


methods
(Chapterxv);computationof (various
of),142, 500, 502,
connected
503, 516 ; inequalities
with, limits of, rates of growth of, 485, 489, 490, 491, 494, 507,
of (Bourget's
of, 4, 478 ; interlacing
513, 516, 518 ; infinity
of, 479, 480, 481 ; non-coincidence
484 ; non-repetitionof, 479 ; number
of, in a stripof arbitrarywidth, 495; reality
hypothesis),
of,
482, 483; tables of, 664, 748-751 ; tables of (referencesto),659 ; values of, 4, 5, 512, 516 ; with
imaginary argument, 511 ; with unrestrictedly
largeorder, 513, 516

Zeros

Zeros

of Lommel's

polynomials

Zeros

of Struve's

function, 479

PRINTED

IN

ENGLAND

(reality
of),304, 305,

BY

J.

B.

PEACE,

M.A.,

AT

306

THE

CAMBRIDGE

UNIVERSITY

PRESS

Você também pode gostar