Treatise On Bessel Functions and Their Applications To Physics

hail
AOi
77
Library
Cornell University
functions andjheir
^^nBesse.
on
A
treatise
Bessw.iu;^,,,,,,,,,,!!^^^
Cornell University
Library
The
original of this
book
is in
the Cornell University Library.
There are no known copyright
restrictions in
the United States on the use of the
text.
http://www.archive.org/details/cu31924001120892
A TREATISE ON
BESSEL FUNCTIONS
AND
THEIR APPLICATIONS TO PHYSICS.
A TBEATISE ON
BBSSEL FUNCTIONS
AND
THEIR APPLICATIONS TO PHYSICS.
BY
ANDREW
GRAY,
M.A.,
PROFESSOR OF PHYSIOS IN THE UNIVERSITY COLLEGE OF NORTH WALEB
AND
G.
B.
MATHEWS,
M.A.,
FELLOW OF ST JOHN'S COLLEGE, CAMBRIDGE

PROFESSOR OF MATHEMATICS IN THE UNIVERSITY COLLEGE OF NORTH WALES.
" And as for the Mixt Matliematikes I may onely make this prediction, that
disclosed."
there cannot faile to bee more kindes of them, as Nature growes furder
Bacon.
Hotrtfon
MACMILLAN AND
AND NEW
YOKE.
1895
[All Eights reserved.]
CO.
A%53
PRINTED BY
J.
&
0. P. CLAY,
AT THE UNIVERSITY PRESS.
PREFACE.
This book has been written
in view of the great and growing
importance of the Bessel functions in almost every branch of
mathematical physics
convenient form so
and
much
principal object
its
is
to supply in a
of the theory of the functions as is
necessary for their practical application, and to illustrate their
use by a selection of physical problems, worked out in some

detail.
Some
readers
may be
contain a needless
inclined to think that the earlier chapters
amount of tedious
remembered that the
analysis; but
it
must be
properties of the Bessel functions are not
without an interest of their own on purely mathematical grounds,
and that they afford excellent
illustrations of the
more recent
theory of differential equations, and of the theory of a complex

variable.
And
even from the purely physical point of view
impossible to say that an analytical formula

purposes;
it
may
is
useless for practical
be so now, but experience has repeatedly shown
that the most abstract analysis
may
unexpectedly prove to be of
the highest importance in mathematical physics.

fact it will
it is
be found that
little, if
As a matter
of
any, of the analytical theory
included in the present work has failed to be of some use or other

in
the later chapters
and we are
anything superfluous has been
so
far
inserted, that
from thinking that
we
could almost wish
that space would have allowed of a more extended treatment,

especially in the chapters on the
complex theory and on definite
integrals.
With
regard to that part of the book which deals with physical
applications, our
aim has been
to avoid, on the one hand, waste of
PREFACE.
VI
time and space in the discussion of
any pretension of writing an
trivialities,
and, on the other,
elaborate physical treatise.
"We have
endeavoured to choose problems of real importance which naturally
and to
require the use of the Bessel functions,
treat
them
in
considerable detail, so as to bring out clearly the direct physical
One
significance of the analysis employed.
has been that the chapter on diffraction
we hope
long; but
may
that this section
attention in this country to the valuable
result of this course
proportionately rather
is
more general
attract
and interesting
results
contained in Lommel's memoirs, from which the substance of that
chapter
is
mainly derived.
much pleasure that we acknowledge

encouragement we have received while composing
It
We
is
with
are indebted to Lord Kelvin and Professor J. J.
permission to
make
free use of their researches
electrical oscillations respectively; to Professor
on
the help and

this treatise.
Thomson
fluid
for
motion and
A. Lodge for copies
of the British Association tables from which our tables IV., V., VI.,
have been extracted; and to the Berlin Academy of Sciences and
Dr
Meissel for permission to reprint the tables of J and
appeared in the Abhandlungen for 1888.
Dr
the former in manuscript;
which
II.
and
McMahon
has
generously placed at our disposal the materials for Tables

III.,
Meissel has also very
and Professor
J.
very kindly communicated to us his formulae for the roots of
Jn (#) =
and other transcendental equations.
also especially
due to
the proof sheets.
Mr
Finally
we wish
workmen of the Cambridge University

The bibliographical list on pp. 289
list
are
to acknowledge our sense of
the accuracy with which the text has been set
as anything but a
Our thanks
G. A. Gibson, M.A., for his care in reading
up
in type
by the
Press.
291 must not be regarded
of treatises and
memoirs which have been
consulted during the composition of this work.
CONTENTS.
CHAPTER
I.
Introductory.
Bernoulli's problem of the
solid
cylinder,
equation,
2;
Bessel's
oscillating chain, 1
conduction of heat in a
astronomical problem, 3;
Bessel's differential
5.
CHAPTER
II.
Solution of the Differential Equation.

Solution by series
when n
or zero, 8; definition of
explicit expression of
is
Jn (x),
Yn (x),
not an integer, 7
;,
and when n
is
an integer
11; elementary properties, 13; definition and
14;
expressions for J- n (x) and
Tn (x)
as in-
tegrals, 15.
CHAPTER
III.
Functions of Integral Order. Expansions in Series of

Bessel Functions.
x
+co
_1
Proof of the theorem exp-(*-< )= S
Bessel's expressions of
Jn (x)
Jn (x)t";
with
corollaries,
17;
as a definite integral, 18; transformation of
power-series into series of Bessel functions, 19;
Neumann's expression
for
Tn
the addition theorem, 24; Neumann's extension thereof, 25; expan, 23;
sions in series of squares and products of Bessel functions, 29 ; Schlomilch's
theorem, 30.
CHAPTER
IV.
Semiconvergent Expansions.
Solution of Bessel's equation by successive approximation, 34;
new
ex-
pression for J(x) as a definite integral, 38; the semiconvergent series for
Jn (x) and
(x),
40 ; numerioal value of log 2 - y, 41
table of
Jk+i (x),
42.
CONTENTS.
Vlll
CHAPTER
V.
The Zeroes of the Bessel Functions.

Bessel's proof that
Jn {x)=Q
calculation of the roots, 46
has an
Stokes's
infinite
number of
and McMahon's
real roots, 44;
[See also
formulae, 49.
p. 241.]
CHAPTER
VI.
Fourier-Bessel Expansions.
Bayleigh's application of Green's
theorem, 51
value of the integral
n (Kr)Jn (\r)rdr, 53; application to the conduction of heat, 54 j formulae

/;o
for Fourier-Bessel expressions, 55, 56.
CHAPTER
VII.
Complex Theory.
Hankel's integrals,
5965
proof of the semiconvergent
the functions
In
Kn
66
8.
Lipschitz's
series, 69.
CHAPTER
VIII.
Definite Integrals involving Bessel Functions.
CHAPTER
IX.
The Relation of the Bessel Functions to Spherical

Harmonics.
CHAPTER
X.
Vibrations of Membranes.
Equation of motion, 95
of the result, 97 ; case of
solution by Bessel functions, 96 ; interpretation

an annular membrane, 99.
;
CHAPTER XL
Hydrodynamics.
Eotational motion in a cylinder, 101 oscillations of a cylindrical vortex
about a state of steady motion, 104; hollow irrotational vortex, 106; vortex
surrounded by liquid moving irrotationally, 109; waves in a tank, 110;
;
rotating basin, 113; motion of a viscous liquid, 116; Stokes's problem of the
cylindrical pendulum, 118.
CONTENTS.
CHAPTER
IX
XII.
Steady Plow of Electricity oe of Heat

Media.
Differential equations, 122; potential
due to
in
Uniform Isotropic
electrified circular disk, 124;
flow of electricity from disk-source in infinite conducting
medium, 126;
problem of Nobili's
rings, 128 flow in cylindrical conductor with source and sink at extremities
of axis, 131; flow in infinite medium with one plane face separated from infinite
plate electrode by film of slightly conducting material, 132; case in which
the medium has another parallel plane face in which the second electrode is
situated, 136; further limitation of conductor by cylindrical surface, 139;
flow in cylindrical conductor with electrodes on the same generating line,
flow in infinite
medium between two
parallel planes:
139.
CHAPTER
XIII.
Propagation of Electromagnetic Waves along Wires.

Equations of electromagnetic field, 142 modification for symmetry round an
the equations, 146; solution adapted to long cylindrical
conductor in insulating medium, 147; case of slow signalling along a cable,
;
axis, 144; solution of
150; electric and magnetic forces in this case, 151; expansion of

in ascending powers of
sc,
153;
approximative values of
xJ^te)^
Jo(ti)/Jl(rj)
(#)
and
ô(v)/^'o(v) f r large values of 17, 155; case of more rapid oscillations, 157;
effective resistance and effective self-inductance of cable, 157 ; Hertz's solution
for a time-periodic electric doublet, 161
Hertz's solution for a wire, 162.
CHAPTER
XIV.
Diffraction of Light.
Diffraction through a circular orifice, 165; calculation of intensity of
Ulumination, 166 ; expression of intensity by Bessel functions, 167 discussion

of Lommel's U, V functions and their relations to Bessel functions, 170
;
application of results to Fraunhofer's diffraction phenomena, 178; graphical
expression of intensity, 179; application to Fresnel's more general case, 181;

maxima and minima of illumination,
graphical discussion, 182; positions of
when the orifice is replaced by an opaque circular disk, 192;

maxima and minima, 194 1 intensity due to infinite linear source,
184; diffraction
positions of
198; Struve's discussion of this case, 201; application to determination of

space-penetrating power of telescope, 205; diffraction through narrow slit
with parallel edges, 208 ; expression of Fresnel's integrals in terms of Bessel

functions, 209.
'
CONTENTS.
CHAPTER
XV.
Miscellaneous Applications.
Small vibrations of a gas, and variable flow of heat in a sphere, 212;
a solid cylinder, 218;
Bernoulli's problem for a chain of varying density, 221 ; differential equation
reducible to Bessel's standard form, 222.
stability of vertical wire, 215; torsional vibrations of
Note on the Second Solution of Bessel's Equation which

vanishes at Infinity
223
Examples
226
Formula for Roots of Bessel and Related Functions
Explanation of Tables
245
Tables
Table
I.
Table
II.
Table
III.
...
J {x), -Jx (x), for a;=0, -01, ..., 15-5

Jn (x) for x=0, 1, ..., 24, and all integral values
of n for which Jn {x) is not less than 10 -18
The first fifty roots of J1 (x) = 0, with the corre.
sponding
maximum
(x*Ji) for
Table V.
J,
(a:)
for
Table VI.
In (x)
for
Table IV.
241
x=0,
x=0,
n=0,
or
minimum
-2, ...,
-01, ...,
values of
</
.r)
266
280
281
5-1
11,
247
247
282
and
a:
= 0,
Bibliography
-2, ...,
285
Graph of J and Jx
289
.
CORRIGENDUM.
The analysis on p. 39
But it holds good if n
for
Jn
satisfies
formula
is
is
invalid if
n>\.
between - J and \ and the semiconvergent expression

(asymptotically) the relations (16) and (19) on p. 13, so that the
lies
applicable for all real values of n.
CHAPTER
I.
INTRODUCTORY.
Bessel's functions, like so
many
others, first presented
selves in connexion with physical investigations
upon a discussion
therefore, before entering
it
them-
may be
well,
of their properties, to
give a brief account of the three independent problems which led

to their, introduction into analysis.
The
of these
first
is
the problem of the small oscillations of a
uniform heavy flexible chain, fixed at the upper end, and
when
the lower,
its position
Then
may be
It is
assumed that each element
regarded as oscillating in a horizontal straight
m is the mass of
if
free at
disturbed, in a vertical plane, from
of stable equilibrium.
of the string
line.
it is slightly
the chain per unit of length,
the
length of the chain, y the horizontal displacement, at time t, of an

element of the chain whose distance from the point of suspension
is w,
we
and
find,
T+dT are
if T,
by
the tensions at the ends of the element,
resolving horizontally,
mdx
M = U T i) dx
'
d?y
Now,
m dy\
T = mg (I a)
d2v
andhence
If
we
which y = ue
= g(l -
d?y
,-,
are adopting,
dy
*)j%-9fc
%),
and consider a mode of vibration

u being a function of z, we shall have
write z for
nti
(I
d2 u
dz2
G. M.
we
to the degree of approximation
du
dz
ji
for
INTRODUCTORY.
Let us put k2
n'/g,
and assume a solution of the form
= 1arzr
+(r + 1) ra^z^ +
2(202+3.2.0^+
then
+ (r + 1) ar+1zT +
+ (o1 + 2a +
+ arzr +
) = 0,
+ K*(a + a z +
and therefore
i + a = 0,
4a + ! = 0.
u=a
1-
+az+az +
2
!i
2ir
(r
=a
so that
=a
k 2z
+ 1 ) ar+1 +
2
-+
-^-
2 ar
,^-32
2 2 32 4 2
(k, 2),
say.
The
series
<
(, z), as will
of a Bessel function;
it is
be seen presently,
absolutely convergent, and therefore
arithmetically intelligible, for all finite values of k
The
fact that the
upper end of the chain
by the condition
<f>(ic,l)
which,
when
(k, I)
is
given,
is
same
is
and
z.
fixed is expressed
= 0,
a transcendental equation to find
to the
thing, n.
periods of the normal vibrations of the type considered.
be shown analytically hereafter that the equation

always an infinite number of real roots
number
infinite
of
possible
</>
(k, I)
the chain
It will
This
may
but arguments of this kind, however specious, are
oscillations of
v.);
all.
a uniform chain were considered by Daniel
Bernoulli and Euler (Comm. Act. Petr.

t.
be
flexibility of
always untrustworthy, and in fact do not prove anything at
Petr.
has
so that there will he an
normal vibrations.
thought intuitively evident, on account of the perfect
The
k, or,
In other words, the equation

expresses the influence of the physical data upon the
which comes
<j)
a special case
is
tt. vi, vii,
and Acta Acad.
the next appearance of a Bessel function
is
in Fourier's
Thdorie Analytique de la Chaleur (Chap. VI.) in connexion with
the motion of heat in a solid cylinder.

It is supposed that a circular cylinder of infinite length
heated in such a
way
that the temperature at any point within
is
it
INTRODUCTORY.
I.]
depends only upon the distance of that point from the axis of the
cylinder.
The cylinder is then placed in a medium which is kept
at zero temperature
and
required to find the distribution of
it is
temperature in the cylinder after the lapse of a time
Let v be the temperature, at time

then v is a function of x and
t,
t.
x from the
at a distance
Take a portion of the

cylinder of unit length, and consider that part of it which is
bounded by cylindrical surfaces, coaxial with the given cylinder,
is the conductivity of the cylinder,
and of radii x, x + Ax. If
the excess of the amount of heat which enters the part considered
above that which leaves it in the interval (t, t + dt) is
axis
t.
dH = 2-jtK (x
or, say,
The volume
and G the
of the part
^ + ^j
is 2-irxdx,
specific heat, the rise of
CD
where
dxdt.
so that if
D is
temperature
the density,
dv
is
= jji
<&>
Iirxdx Kidt = dH.
ot
Hence, by comparison of the two values of dH,
Fourier writes k for

function of x only
K/CD, and assumes
if
du
xdx +
5-
we put | = g, we
is
u being a
substantially the
Bernoulli, except that
n
tu
k
find there is a solution
.-Ad-S^s + J*u'/-a^
2 2 .4 2
which
= ue~nt
this leads to the differential equation
d2u
-,
dx2
and now,
&+
2 2 .4 2 .6 2
V
/
same function as that obtained by
we have
\ga? instead of
i?z.
to
condition leads to a transcendental equation
detail.
in
problem
find g ; but this is not the place to consider the
of the functions
discovery
Bessel was originally led to the
connected
problem
which bear his name by the investigation of a
The boundary
with
motion,
elliptic
r
which
may
be stated as
follows.
12
INTRODUCTORY.
be a point on an ellipse, of which A A' is the major
Draw the ordinate NPQ meetaxis, S a focus, and C the centre.
ing the auxiliary circle in Q, and join CQ, SP, SQ.
Let
is
Then in the language of astronomy, the eccentric anomaly of P

the number of radians in the angle ACQ, or, which is the same
thing,
it is
<f>,
where
ACQ
area of sector
_
'
AQA'
area of semicircle
It is found convenient to introduce
a quantity called the mean
anomaly, defined by the relation

area of elliptic sector
ASP
APA'
area of semi-ellipse
'
'
(By Kepler's second law of planetary motion,

to the time of passage from
fx.
to P, supposing that
is
proportional
is
the centre
of attraction.)
Now by
orthogonal projection
ASP
area of
area of
APA' = area of ASQ area of AQA'

= (ACQ-CSQ):AQA'
= (a - |ea sin
^7ra
= {4>e sin
it,
:
</>
<j>)
<f))
where
e is the eccentricity.
Moreover,
if ft
and
periodic function of
is
fi, e,
p = < e sin
relation
/u.
Hence
a multiple of
fx,
it.
vary while
<\>
are connected
by the
3
</>.
remains constant,
which vanishes at
We may
and A'
<f>
that
is,
/a is
when
therefore assume
00
<j>
fi = 2-4 r sin
r/j,,
and the
mined.
coefficients
Ar
are functions of e
which have to be
deter-
INTRODUCTORY.
I.J
Differentiating
4 with
respect to
%rA r cos
and
by
therefore, multiplying
îrrA r
j
J
Now $ =
when
/*
and integrating,
r/j,
-^-
y cos
cos 974 d/i
ru. tZu.
/<
= 0,
we have
= -^ 1,
rfj.
cos
/*,
and
<
= ir when
changing the independent variable from p to
\^rA r =
/*
</>,
7r
we
so that
by
obtain
cos r/t d<,
Jo
Jo
and
.4,.
which
.4,.
cos
We
is
e sin $)
Ar
<)
d$,
d<,
as a definite integral.
The
Ar
e,
be obtained directly from the
fact,
shall not, however, follow
but merely show that
which
e sin
can be expressed in a series of positive powers of
and the expansion may, in

-integral.
r((f>
expression for
is Bessel's
function
cos r (0
satisfies
up the
investigation here,
a linear differential equation
analogous to those of Bernoulli and Fourier.
Write x
for
e,
and put
7rr
cos r(d>
x sin
$)
d<\>
Jo
then, after partial integration of -7- with respect to
^+=
,1 du
d?u
â
=ru
2
(*
,,
COSr( *-* Sm<
J
I
cos
tj)
cos
r((j>
find that
^
,.
x sin <) d</>
{(1
#cos$)-l}cosr($-
JO
'
= - ('r - \u
r(6
v:r
x sin rf>)]
^ /J o + u
it
r Tsin
L
a?
z
we
fir
= -r2u
&>
<,
+-:+*H)" a
a;
a;
sin
$)d$
INTRODUCTORY.
If
we put rx =
z,
becomes
this
du
2
and
this is
what
is
du
now
r2 \
/.,
considered to be the standard form of
Bessel's equation.
If in Fourier's equation,
1 du
x dx
d?u
I
dx2
we put
du
di2
The
is
is
A-
the transformed equation
which
nu
(
du
+ zdz +U ='
a special case of Bessel's standard form with r

differential equation
is,
for
many
reasons, the
= 0.
most con-
venient foundation upon which to base the theory of the functions
we
shall therefore define a Bessel function to
be a solution of the
d2u
da?
du
/,
i--U=a
+-T-+
xl
x dx
2
In the general theory no restriction

of n
is
is
placed upon the value
the most important case for physical applications
zero or a positive integer.
analytical theory presents
some
Moreover when n
spscial features
is
is
when n
integral the
so that for both
reasons this case will have to be considered separately.
CHAPTER
II.
SOLUTION OF THE DIFFERENTIAL EQUATION.

j
If
we denote the
may be
operation x
by
-=-
&, the differential equation
written in the form
Wy + (x 2
Assume
that there
ft
y=
0.
a solution of the form
is
=x
(a
+ a x + a. x + ...)
2
we
then, if
substitute this expression in the left-hand side of the
differential equation,
and observe that %xm = mxm the
result
is
oo
(r 3
ra )
a xT +{(r
+ iy- n?} a xr+1 + 2 [{(r + s) - w

2
2
}
as
+ as_J xr+s
The equation will be formally satisfied if the coefficient of every

power of x in this expression can be made to vanish. Now there
is no loss of generality in supposing that a is not zero, hence the
first condition to be satisfied is
r'
n? = 0,
r
or
= + n.
In*general, neither of these values of r will
vanish*; consequently
be
c^
= 0, and all
make
(r
+ 1) -
the as with odd suffixes
w2
must
zero.
*
An
exception occurs
when n=i, r= -; but this does not require separate

we still have the distinct solutions y x and y 2 with
discussion, since in this case
n=i
The only
peeuliarity
is
that = - \ leads to both of these solutions.
8
If
we take r =
we have
n,
and
a4
=-
2(2+2)'
a2
4(2m +
2.4.(2m
4)
+ 2)(2w + 4)'
solution of the differential equation
is
there-
by putting
fore obtained
y = yi
a2
A formal
so on.
+ s)as + as_ = 0,
(2m
and hence
[II.
= a,asn
(1
a?
2 (2 W
2)
as
+ 2) (2n +
4 (2n
'"
4)
(~y^
+ 2)(2ra + 4)...(2n + 2s)
putting r = n we obtain the
".
l
/'
"2.4...2s.(2ra
In a similar way by
formal
solution
x2
y,
- do*-"
x*_
+ 2(2w _2) +
^1
4 (2n
.
- 2) (2n -
4)
N
- '
'
which, as might be anticipated, only differs from y l by the change

of n into n.
If
is
any
complex quantity, except a real integer,

which occur in y 1 and y2 are absolutely con-
finite real or
the infinite series
vergent and intelligible for

fact ultimately
all finite
/yi4
/y>2
2 .4
the rapid convergence of which
The
values of x: each series in
behaves like
ratio of
yx
to
y%
is
,716
~~
2
.
is
42 6 2
"
'
obvious.
not constant; hence (with the same
reservation) the general solution of the differential equation
y = Ay + By
1
and
If
B being arbitrary
n=
0,
is
2,
constants.
the integrals yy and y 2 are identical
if
is
a positive
integer y 2 becomes unintelligible, on account of the coefficients in

the series becoming infinite.
Similarly when n is a negative
integer y^
is still
available,
In each of these
but yx
is
unintelligible.
cases, therefore, it is necessary to discover
and since n appears only in the form of a square

in the differential equation, it will be sufficient to suppose that n
second integral
is
zero or a positive integer.
.'
II.
In accordance with the general theory of linear

we assume a solution
differential
equations,
= (a + 6
log x) x~ n
+ (a + b
x
log x) ar m+1
. .
00
= arS (a + b
s
Xs
log x)
then, observing that
(xm log x)
= w?xm log x + 2mxm
and making a few easy reductions, we
^y + (x
find that this form of
y gives
-n?)y = - 2nb x~n
+ 1) (ai + b, log x) + (- 2n + 2) 6,} *-" +1
{(- 2n
00
+2
(- 2n
{s
+ s) (a +
s
bs log x)
+ (a _ + 6 _
s
log
as)
2n + 2s)
+ (-
b}
srn+s
The expression y
will be a formal solution of the differential

equation if the coefficient of every term a; 4"* or x~ n+s log x on the
made
right-hand side of this identity can be

that this
may be
the case
to vanish.
In order
be found that the following con-
it will
ditions are necessary*:

(i)
The
(ii)
All the as with odd suffixes
(iii)
The
and so
on,
coefficients b
coefficient
up
&,
bu b2
is
...
iam-i
must
must
vanish.
all
vanish.
indeterminate;
2(2-2)'
a-2
-
4 (2n - 4)
(2ra
- 2) (2w - 4)
to
Oo
3-2-
The
(iv)
"271+2
first
a.
is
...
2"
indeterminate
'
~2n
2(2n+2)
\8
-^-J
The reader
a few of the
(2n-2)(2-2)(2w-4)
_ ^2712
+28
coefficient
2 .4...
_a
2.2n(2
+ 2)'
&2JI-2
2.4...2.2n(2n
2)...(2n
2)'
<'e>0'>
v
'
perhaps follow the argument more easily if he will write down

terms of the sum, and also a few in the neighbourhood of s = 2it.
will
10
and
the coefficients b m+1
all
b.2n+3
with odd suffixes must
etc.
[il.
vanish.
Finally
(v)
<W
2M+4
Can
2 (2m
a*n.
m+2
+ 2)
2 (2m
2)
[1
+ 2m
-~2(2n + 2) + 2(2n + 2) (2
(2ft + 8)
dm+a
a "* H
~~4(2n + 4) 4 (2m + 4) te+4
b
+ 2j m
'
~2
4 (2m
.
+ 2) (2m + 4) ~ 2
(i
i2
and, in general,
~2
when
+ 2) (2m + 4)
(2ra
+I+
2mT2
+ 2w i_U
+ 4r2n
'
> 0,
(") s
"
'
ttwi.-i-w
Ml+3S
+ 2) (2n + 4).
(2m
4. .2s
.
All the coefficients which
expressed in terms of two of
.(2m
+ 2s)
may
do not vanish
them
if
we
therefore be
choose am and b m
for
these two, y assumes the form
V = <Wi
where
y,
^^
= d"|l -
+ &22/2>
g.
4(2n+ 2 )(^ + 4)
7
f
(the solution previously obtained), and
a2
y*
2(2n
- 22""lw! ^-
1)!
+
2)
r"
a4
2.4..(2w + 2)(2?i+4)
2l2n^2 )
"
log
a;
+ 2.4.(2,-2)(2,-4) +
a;"-
2
1
-I
(2.4.6...
(1
2{2n
2
4 (2w
.
'
+ 2)
|2
2) (2n
t 2k
2m-2) J
+
111
2
+ 4~) J2 ^ 4
2.4...2s(2w + 2)(2re
"*"
2+ 2
1
"*"
+ 4)...(2w +
2m
+ 4]
2s)
\2s
2m + 2s/
II.]
the
The
characteristic properties of the integral
sum
of
11
y2 are that
it is
log x and a convergent series, proceeding by ascend-
which only a limited number of negative

powers of x occur, and the coefficient of xn is zero. It becomes
infinite, when x = 0, after the manner of a;
for any other finite
value of x it is finite and calculable, but not one-valued, on
account of the logarithm which it involves.
ing powers of
in
x,
The general
solution of the differential equation
= Ay + By2>
and
value
being arbitrary constants.
When
its
is
n = 0,y2 does not involve any negative powers of

simply written
x,
and
may be more
-S + 2^--) l0g
'
xA
1\
/,
+S-H)^+(
i+
x6
1\
3 J 22 4a 6 3
.
It is found convenient, for reasons which will appear as we

proceed, to take as the fundamental integrals, when n is a positive
of these by 2"
yx and y3 but the quotients
n (x),
special integrals will be denoted by Jn (x) and
integer, not
x
2(2rH-2)~ "2.4.(2tt
fi
W-2Tîf
=
These
!.
so that
x%
r/%_
+ 2)(2rc + 4)
'")i
LJ_5
IO
o^.s^Ji + s)!
(where, as usual,
is
interpreted to
Wn (x) = Jn (x) log x - J2'-
(n
mean
and
1),
- 1)! r +
^jf
xn
l+4
+ 0a; +
(_)-i
a;
2-5(n-3)!a;-'
also
2!
4/1
71 -1- 28
J2^!F^)!
? ks
\)
+ 2n-+2s)\-
/ (x) may be extended to include the case

function Iln, which
is not integral by means of Gauss's
denoted by T (n + 1). When n is a positive integer, Tin is
The
when n
is
'
'l
~2
+, ,+ 2(n-l)!J
definition of
the same as n\;
its
general value
is
the limit,
when k
of
1 .2. 3...K
r,
,
N
n(">)
=
(,l
+ 1)(n + 2 )...(r +
l
/c
is infinite,
12
The
lire is intelligible
function
except
when
values of
re,
infinite
by a convention
and
finite for all real finite
when
a negative integer,
is
re
[il.
lire
becomes
be well
which are expressed by the formulae
we
define 110 to be
to recall the properties of lire
It will
1.
= rail (re -1)

1) = w cosec nir
II(i) = iVir
lire
II
( re) II (re
13
>
I
Forsyth, Treatise on Differential Equations (1885) p. 196
(Cf.
Gauss, Werke
p. 144.)
ill.
The function lire may also be interpreted when n is complex;

but we shall not require this extension of its meaning.
Jn (x)
This being premised, the general definition of
by the
and
given
is
relation
/ "(*)
in like manner, if
j
If n
is
re
2+ 2 *IIsII(rc
H**n+2s
14'
a positive integer, the function J_
is
ing, does not exist; it
I4
not a positive integer,
(x)
+ s)
properly speak-
(as),
however, convenient in this case to adopt
is,
the convention* expressed by the formula
J^n
When
write
to
= (-)" Jn (x),
[re
integral].
the argument x remains the same throughout
Jn instead
x by accents
By
(x)
;.
of
Jn (x), and
thus
J'
we may
indicate differentiation with respect
will
mean
^p-^
and
so on.
differentiating the general expression for
Jn (x) we
find
that
j,
"
_ ( -)"
~
_re
~x
/-(n+e)
may
when
be proved that
c is
the convention
is
infinitesimal,
,l
+
1
"
+ 2s) xn +-
nsn(re +
_re
* It
(re
2 n+2s
_|
s)
(-)*fl^ +
2+ 2s - 1 11
(s
5
-1
1) II
(re
+ s)
xn+2s+1
(-)
2+ 2s + i risn(re
J_ B is defined as in
and n a positive integer,
if
(14')
+ s + l)'
the limiting value of
is precisely
necessary in order to secure the continuity of
Jn
- )Vn
so that
II.]
(on writing
or,
which
is
+1
for s)
that
is,
the same thing,
= ~ Jn J'n
"n+x
+ 2s
Again, writing
\S
gM+SSX
T
FT TT /
n+*
nsn(
+ s -i)
t
=\
(/-i
Jn+i)-
l6'
+ s) + s, we
2+ 2s
n(s-i)n(n+s)
1
be found that with the convention
all
values of
It is
n.
have
\8 ,+2S
oo
On+
a,
2+
It will
true for
in the form (
"^
"
13
these formulae are
worth while to notice the special
result
J',
Jn+1
If
is
obtained
-J
18
x.
eliminated by combining 16 and
*>
is
and
similarly
20
the formula
= "nr-x ~ Jn
by
19
eliminating, J'n
it will
be found that
2W
T
T
-C\
"
~ "n
+ "n+x
T
Jnx
The formulae 16
17,
20
are very important, and are continually
required in applications.
It follows from 17 that
w: = 2,/;_ -2 /;
1
+1
= \yn2 ~ "n) ~ \" n ~ "n+v
Jn2 &Jn
and
it
may be proved by
"I"
+2j
induction that
2V ^ = Jn8 ~ s"n8+v H
"ns+i
+ ()
m+s>
the coefficients being those of the binomial theorem for the expo-
nent
s.
14
The analogous
W, which may be obtained by a
formulas for
precisely similar, although
2n
n
-
more
[il.
tedious, process, are
W^
Jn+1
l
21
'
II.]
There
way
another
is
of deducing a second solution of the
differential equation which deserves notice.
= uJn (x), where

observing that Jn (x) is
In that equation
new dependent variable

an integral, we find that
put y
is
^()+^()+^<-))5-o
dx~\ 1
-f-\
x
du
dx
or
and
is
2/
is
*-
Jtn t(x)\
then,
dx
= A Jn (x)j
^jij^j + BJn (x)
32
the complete solution of the equation.
By properly choosing
to
2J'n {x)
xJ\ (x) -5- =

dx
whence
15
the constants
make y identical with J_

an integer.
(x) or
A and B it must be
Yn (x)
according as n
Taking the general value of Jn (x), n not being an
f-WT-,= -2
so that
and
^!*x^r-
The leading term
of
The value of S
integer,
we
- 1)Ax-n+
obtain
r^rr
n(n-l)II(-n)
will
not or
J^ (x) is n ,< ~", and by making this
agree with the preceding,
A =
*
is
m- nnTl(n-l)x-+...
2n - lu(n
AJ
possible
2
sinn7r.
33
ir
depend upon the lower limit of the integral

r*
dx
I' xJl (x)
16
[il.
Jn (x) and differentiate with

when y = /"_ (00) we have
Divide both sides of 32 by
to
then in the case
(J-ri\
2 sin WIT
dx\Jn )
JLJ-n -
or
With the help
of 16
irxfl
JUJn =
and 19
sin nir.
an
If
integer.
we
to the form
sin Wit.
35
ttx
This suggests a similar formula involving

is
34
may be reduced
this
JnJ-n+r + J-Jn-x
when n
respect
Neumann
functions,
write
U = X (j n+l I n i/X+i)
we
find with the help of 16, 19, 26,
27 that
+ Jn+1 (nTn - *F+1 ) - Jn (xYn - n + 1 FB+1 )

n+i (?W m ~" S m+i)
-*
=
identically.
Hence m
the explicit forms of
unity.
is
We
is
Jn
independent of x
and
Returning to
it is
and by making use
of
easily found that the value of
are thus led to the curious result that
"n+i*
is
Yn
32,
we
n~ J n* n+\ ~ ~
find,
3^
by comparison with
30, that,
when
a positive integer,
c
,-Jnj
As
dx
xJ ,
+ BJn
in the other case, the value of
limit of the integral.
37
depends upon the lower
CHAPTER
III.
FUNCTIONS OF INTEGRAL ORDER. EXPANSIONS IN SERIES OF

BESSEL FUNCTIONS.
Throughout
contrary
is
this chapter
definition of the Bessel functions,
is
a positive integer.
pression for the Bessel function of the

J-(^)
which
may be
=2
hence we conclude that

(t
jr
ntb order
2 +2Ss!(ft+s)!
-J
xn+s
ex-
(_ xy
2 n +.(n+s)l
'
2*771'
Jn (x) is the coefficient of
according to powers of
exp ^
(t
- r ) = exp j
1
exp
7 2'.W;
and the
The
is
written in the form

oc
of exp
be supposed, unless the
will
it
expressed, that the parameter n, which occurs in the
n
coefficient of t , obtained
t.
In
n in the expansion
fact,
-^-~
2 s s!
.
by putting r
= n,
is
In the same way the coefficient of trn in the expansion

n
() Jn (x), or J-n(x), so that we have identically
exp|(*-r ) = XJn (a;)^.

1
G. M.
is
38
00
FUNCTIONS OF INTEGRAL ORDER.
18
[ill.
The absolute value of Jn+1 (x)/Jn (x) decreases without limit

when n becomes infinite: hence the series on the right hand is
absolutely convergent for all finite values of x and
t.
= $", then the identity becomes

gtesin * _
j (^ + 2iJ (%) sin + 2 J (x) cos 20
+ 2iJ (x) sin 30 + 2J (x) cos 40 +
Suppose that
39
. . .
and hence
cos {x sin
</>_)
sin (x sin 0)
= J (x) + 2Jz (x) cos 20 + 2J4 (x) cos 40 4 2J (x) sin + 2 J3 (x) sin 30 + 2J5 (x)smo<f> + ...
. .
Change
into ^
= Jo (*) 2/ (a;) cos 20 + 2 J" (a;) cos 40

sin (a; cos 0) = 2.^ () cos
2J (x) cos 30 + 2J (x) cos 50
5
These formulae are true

integrating from
to
43
by cos
n<f>
and
we obtain
it,
cos 7i0 cos (x sin 0)
. . .
for all finite values of 0.
of these four formulae
first
42
. .
Multiplying the
41
thus
cos (x cos 0)
40
d0 = irJn (x),
if
is
even (or
zero),
/,
= 0,
or,
if
odd,
is
in a single formula,
/* cos w0 cos (x
I
'0
Jo
sin 0)
d<j>
=
^
(1
+ ( 1)]
{1
- (-
J" (*).
44
Similarly
f*
I
By
sin
n0
sin
(a;
7T
sin 0) c0
sin 0)
= 7rJ" (a;),
-,
1)} J"
(a;).
45
addition,
cos (iz0
a;
d<j)
46
which holds good
for all positive integral values of n.
be remembered that an integral of this form presented

connexion with Bessel's astronomical problem in fact we
arrived at the result that if
It will
itself in
fi
e sin 0,
00
then
= fi + %A r sin rfi,
1
where
Ar =
2
f"'
cos r (0
- e sin 0) d<j>.
EXPANSIONS IN SERIES OP BESSEL FUNCTIONS.
III.]
We now see
A r may
that
19
be written in the form
2
A r = -Jr(re),
and in
this notation
$ = fi +. 2
It is
2 cos n)-> -
+ (_y
43,
+ ~J
let
l)
(2cos
. .
^_2s+
be applied to the identities 42 and

left hand be expanded according
the expressions on the
to powers of
x cos
then by equating the
cf> :
coefficients of cos"
on
(f>
both sides we find

.*
= 2
{jB
(n
+ (n + 2) Jm+2 + ( + *)(" +
6)
(ra
2) (w
+ 1)
-
3!
,
^-j
which holds good

0.
jn+4
(ra+2s)(w+s-l)( + s-2)...0? + l)
n=
i/ji+6
!)
The
first
</+!!
for all positive integral values of n,
+ .-r,
and
47
also for
three cases are
= Jo + 2J + 2/ + ... + 2JW + ...

x = 2J + 6J3 + 10J + ... + 2 (2s + 1) J"28+1 +
a? = 2 (47, + 16 J, + 36 J +
+ 4s 7 +
1
...
In confirmation of these results

series in brackets in
2c,
47
is
. . .
it
2g
. .
.).
should be observed that the
absolutely convergent; for
if
we write
Jn+ia we have
of
c8+1 j; t+28+2
ct
Jn+i8
+ s)oo^
l)(w
4(s +
+ 2s)(, + 2s+l)' 1 _
(n
2(2w + 4s + 6)
this decreases
without limit when
"'
...
or
and
(2ra
+ 4s + 2)
s increases indefinitely.
22
it
20
Moreover
it
if
[ill.
Sh denotes the sum
can be proved by induction that
.n.b h -x
or,
which
is
*^
'
2^(n + r- + A)!r!/i!
the same thing,
2 n .n\Sh = xn
(n
+ h)lxn+lh+
2*M(n + 2A"+l)!(A+l)l
~2
Since the series in brackets

the right hand
may be made
h large enough
is
(A
+ 2)(w + 2A+2).2!
convergent, the expression on
as near to
moreover the
+"
xn as we please by taking
series
8=00
s=h+l
becomes ultimately infinitesimal when h increases
indefinitely;
therefore the relation
I ce Jn+v
2.*I
is
true and arithmetically intelligible for
Now
suppose
we have an
all finite
values of
x.
infinite series
00
2 ago? = a +a x +
1
a^x*
+ ...,
then on substituting for each power of x

functions,
its
expression in Bessel
and rearranging the terms, we obtain the expression

00
2 b Js = b J +
ftiJj
where
=a
61
= 2a
&2
1,
+ b#T +
a
= 8a + 2a
2
. .
.,
...,
and, in general,
6
2s
-4' + 2H^D-%
+ 2 (s-l)(s-2)'
\
2!
^- +
I
...l
+ 2 (s-l)(s-2)(s-3)'
3! ^
6
48
4
the
sum within
as s is
brackets ending with a term in a^ or according
odd or even.
EXPANSIONS IN SERIES OF BESSEL FUNCTIONS.
III.]
If the series
"tagsc*
21
and XbgJs are both absolutely convergent,
we may put
= 'ZbgJg,
l,a^f
49
and the arithmetical truth of this relation may be

method similar to that employed above.
An
important special case
Cauchy's
first test
when the
is
of convergence
that
verified
by a
series ta^nf satisfies
to say,
is
when
for all
positive integral values of s above a certain limit,
cue
where k
a definitely assigned proper fraction.
is
limit of bgJs/bs^ J"g _i is a^x/a^.!,
and the
In this case the
series 26gJg is absolutely
convergent.
Cases in which #/_!

examined separately.
If
we
ultimately equal to unity have to be
is
are assured of the possibility of an expansion such as
may be found by any method
that here considered, the coefficients
which proves convenient.
As an
illustration of this, let us
assume
F = Jo log x + %cn Jn
Then
Y'
= J' log x + ^ + %c n J'n
30
and hence
= F +-
+F
Y'
'
CO
-Q + Xcft +
2j:
Therefore
by
8.
Now
Xrfc nJn =
lj-.
2#J' = 2a;Ji
+ JJ
by repeated application of 20
xJ = 4/3 xJ
= 4Ja -8J + xJ =
= 2(2J -4 / +6J -...)
t
<
22
2n2cJ =
Hence
and
4J + 6J
4 (2 Ja
[ill.
.),
finally
Y = J \ogx + 4iUj
<)
--^Ji +-J
Differentiate both sides with regard to x,
-...
so
j.
and apply
\J, 18, 29;
thus
that
is,
Now we
have
by 40
identically,
l=J +2J + 2J
and therefore
-y-
f-
- Ja
1,1
+
<b
I"
/i
or 47,
+...
+ - J* +
2.3
274
. .
,/,
2.5
476
Consequently
6(fe + 3)
(4s
+ 2)(4s+4)
2 (4s +5)
,74 +s
(4a
/
+ 4)(4s + 6)* * M+ -
Similarly
o
^"0
- -^
F^log*-
TT
2*1 \
7-1
|4Jt
12.6
8J^_
_^8^6.10
4.
which may be transformed into

Y,
= J,\ogx---^J
+6
--J
(0-A) + 9 (3A5-^)
}'
S2
III.]
23
or again into
2s(2s
Neumann
(2s-l)(2s +
2)
^(2s-l)(2s +
iy ts+
2s(2s
l)
+ 2)
^+ -
has given a general formula for F which
55,
may be
written in the form
-s
may be proved by
This
There
is
v
j
'
(n
+ s)
" +J>
induction with the help of 20 and 28.
way
another interesting
of expressing
Tn
If
we
write
Tn = 2"-
(-!)!-
2 "~3 (rc- 2 )
apnw
*"(n-3)t
-^
2!
oji-i1 /
i\t' 57
2"(ft-l)!
:
a polynomial which has already appeared in the expressions
Wn and Yn
Wn = Jn log x - Tn + 2 n+2
7i
for
then we have
+4
n(n
{w
+ l)
+ 2}
/ +2
,
n+6 w(n+l)(n + 2)
+,-..31
6>+T)l
+2J^
f
r^
+ 2,
n(, + l).
2s(/i+s)(.
(,
+ ,-l)
s!
and the corresponding expression
^J^
for
can at once be written down.
Other
illustrations of these expansions will
examples at the end of the book.
be found in the
24
number
great
for Bessel functions
and interesting
of valuable
nected with a proposition which
may be
some of these
will
[ill.
results are con-
called the addition theorem
now be
given.
By 38 we have
+
ijn (u + v).t- = exp &1 (t-\

-erp|(-l)exp|(t-I)
= jn (u)tn+-jn (v)tn
00
CD
Multiply out, and equate the coefficients of the powers of
both sides
+ v)= J
Ji (u + v) = J
(u
J
(u) J
(u)
(v)
(v)
- 1J-, (u) J^ (v) + 2J
(u)
+ J! (u) Jo (v)
- J (u)J (v)-J {u)J
2
and
on
thus
(v)
(v)
. .
...,
in general
n
Jn (u + v) = 2,J, (w) Jn+ 2 (-)
(v)
Jn+g () + Jn+S () J
J, (u)
59
+ ...,
60
+ J (u) Jâ (v) + ...+Jn (u) J (v)

+ J (u) Jn+S (v) + J" (u) Jn+i (v)+ ...
+ Jn+% (u) J (v) + Jn+i (u)J (v)+ ....
60'
()}
Observing that
if n is odd
Jn (- v) = (-) nJ(v), we
find that
Jn (u + v) + Jn (u - v)} = /, (u) Jn_! (v)

+ J (u) J"_ (v) +
+ Jn (u) Jo (v)
- Ji (u) Jn+1 (v) - J (u) Jn+S (v)- ...
3
. . .
+
and
if
is
J+2
O) J
{V)
+ Jn+i (u) J
(v)
even
{J* (u
+ v)+ Jn (u - v)} = J
(u)
Jn (v)
III.
25
Similarly
if
is
odd
2 {^ (
+ )- Jn (u - v)} = / (u) Jn (v)
+ Ja () J_ ()+.+ J^ (M ) j; ()
+ J" (m) Jn+i (y) + J (u) Jn+i ()+...
- ^n+i () /i
- n+a ()/() ...,.
2
(i>)
and
if
t/"
even
is
{/ (it
+ v) - Jn (u - v)} = J
(u) J"_! (v)
+ J (u) J_ 0) + + J_i (m) Jj (d)

- J"j (w) /+! () - J (u) Jn+a (v)- ...
3
. ,
-7n+i()/i(w)- Jn+,(w) J,()-
By
we
=x
putting u
obtain
from
and
= yi,
i;
where x and y are
these formulae expressions
imaginary parts of
Jn {x + yi).
....
6l'
real quantities,
the real and
for
be noticed that Jn (yi)

as a real or purely imaginary quantity
It should
immediately presents itself

according as n is even or odd.
We will
now
theorem which
By
38
consider a remarkable extension of the addition
is
due to Neumann*.
we have
exp
(kt
- i) = iknJn (x) tn
i(*-5)-T('-9 + I(*-i)i
therefore
pS( w -5)"
eip
+
that
knJn (x)
is,
s(*-i)-
= eH
62
then
+00
irsinfl
S/(/-e w)<n
* Strictly speaking,
^('-7)'
= * (*"*) J (fcc) *.
00
Put x = r, k
8ip
= e"
Neumann
generalisation immediately suggests
+oo
2#"Jn (r)P.
63
only considers the case when n = 0; but the

itself.
26
Equating the
coefficients of tn ,
[ill.
we have
Jn (re M) = en6iJn (r) - HÂ e ^+^Jn+1 (r)
= Hn + VA
where
64
= Jn (r) cos nd + r sin
6 sin (n + 1) 6 Jn+1 (r)
r2 sin2 e
2!
_. . r
. .
,
cas(ra-|-2)0/ n+3 (7-)
r sin3 #
3
+ 3)^J'+3 (r) + ...,
sin(m
3!
and
t\
= J" (r) sin nd r sin
cos (n
r 2 sin2 5
-sin (n
2!
1)
64'
Jn+1 (r)
+ 2) 6 Jn+!i (r)
+ ^coB(n + 3)W)H) (r) + ....

As a
special case, let 6
= 7T
thus
yH
J (n) = in I J* (r)
c,
64"
+ rJn+i (r) +
2j
Jn +%
^:
(r)
+ g,
<7 +3 (r)
65
-f
Returning to 63 let us put r, 6 successively equal to

7 and multiply the results together thus
b,
ft
and
Jn (6eP*)
QO
n /" (ce*) w
00
Now
^(6sing+CBin y) +qo
+oo
00
00
the left-hand expression
is
equal to
+00
2J B (6e'K +ce*0*,
00
and if the right-hand side is expanded according to powers of t,

the coefficient of tn gives an expression for Jn (beP* + ce**) in the
form
Jn (&e"* + ce?0 =C - Cj, (b sin ft + c sin 7)

C '(iisin)3 + (!si]i7)i
i
>Jk
JJ J
III.]
27
where
Go
= e^Jn (b) / (c) + e{*- <+** J^b) J (c) +... + ^r*J (b)Jn (c)
- t" +1 e-WJn+i (b) J> (c) + e{<+ *-n}' m+2 (6) J (c) x
>
- e+1
and
in like
>
Jn+1 (c) + e{ <+
r*}* J, (6)
manner
C1}
for
is
we
If
is
. .
68
too complicated for practical
shall only consider in detail the case
Moreover we
a real quantity.
that n
r-*)*Jt (b)Jn+i (c)
2)
(72 , etc.
Since, however, this formula
purposes,
. .
when
suppose in the
shall
btf 1
first
+ ctf
instance
= 0.
we put
+ ce^ =
be^
a,
a real quantity, we have
a2 =
+ c cos y) + (b sin /3 + c sin y)

s
(6 cos fi
= & + 26ccos(/3-y) + c
2
and
b sin
also
Let us put
/8
7=a
/3
+ c sin 7 = 0.
then the general formula 67 becomes in
this special case
Jo (Jb 2
26c cos a
+c)=J
2
+2
(b)
= J (6) J
(6)
(c)
(c)
(c) cos
- 2 Jj (b) Ji (c) cos a

-
2a
+ 25 (-)
. .
s (6)
J (c) cos sa.

8
69
If
we change
a into
a,
tr
the formula becomes

00
Jo (x/6
- 26c cos a + c ) = J (6) J

2
+ 22 J, (6) J
(c)
(c) cos sa,
69'
and
this is
By way
Neumann's
result already referred to.
of verification, put a
then we are brought back to
the addition formulae.
Suppose
Jo
=o
(jW+tf)
>
*^ en
we âve
= J (b) J (c) - 2 J (b) J
and hence, by supposing
(6
(c)
+2
(6)
(c)
. . .
70
= b,
2
V2) = J (6)
- 2J2
(6)
2J^ (b)
28
=c
In 69 and 69' suppose b

J (26 cos
[ill,
thus
= J?(6) - 2 J? (6) cos a + 2 J| (6) cos 2a -
I)
. .
72
Jo (26 sin
In order
n = 1,
=J
I)
(6) + 2J?(b) cosa + 2J
to obtain
another special case of 67
four quantities
b, c,
ft
/3
let
c sin
us suppose that
7 = 0.
and
7),
(/3
Since the
are connected by this single relation, there
are three independent quantities at our disposal.

b,
+ ...
(&) cos 2a
retaining the condition 6 sin
still
for which, as before,
we
Let us choose
Then
will write a.
a has the meaning already assigned to
may
it, it
be
if
verified,
geometrically or otherwise, that
a& = b +
ce oi ,
Now when
aJ
= l,
ae**
= c + be~ ai
the formula 67 gives
= atP {J, (b) J (c) + e~" J (b) J (c)

- e ai J (b) J, (c) + e " J, (6) J2 (c) -
(a)
- e- J, (6) J
2
Substitute b
ce
and equate the
out,
a J, (a)
{&
= {bJ, (b) J
+ [bJ
= 6J
(6)
(c)
(c)
(c)
for ae^ f
+ e-
+ cJ
(b)
aJ
side,
(a);
. .
.j.
multiply
thus
J, (c)}
(c)
bJ^b) J2 (c)
c J (6)
" J, (6) J, (c) -
on the right-hand
real part of the result to
- bJ (b) J,
(b) J, (c)
(b)
ai
(c)
- cJ, (b) J
- cJ
(c)
+ cJi (b) J
(c)}
cos a
(c)}
cos 2a
(b) J, (c)
+ cJ
(b)
(c)
+ S (-) {6 [Jn+1 (6) - J_x (6)] J (c)

1
+ c [Jn+i (c) - Jn-j (c)] J (6)} cos na.

This result
may
also
73
be obtained from 69 by applying to both
sides the operation
and
in fact this operation affords the easiest
the formulas for n
2, 3, 4, etc.
means of obtaining
III.]
29
may be obtained from 73 like those deduced

For example, putting b=c, we find, after a little re-
Special results
from 69.
duction, that
(2b cos
|)
= 2 J, (6) J, (b) cos | - 2 J, (b) J, (b) cos 3a

2
+
This
may
respect to
(6)J3 (6)cos^-....
74
be deduced from 72 by differentiating with
also
b.
In 72 and 74 write &
powers of cos =
and
2,72
and equate
= Jl (w)
rf
=4
expand both
for b,
coefficients
%J{ (x)
-r
{(*)
sides according to
thus
+ 2J2 ()-...
+ 4J?(e)+ 9J?() +...},
in general
+ 2)...(2n + *-l);
+ 2(n + s)(2n + l)(2n
-J'n+s + ..j,
^71 ,
75
while for the odd powers of x
2JJ + QJ J + 10J J +
^ = 16
JiJ,
5 JaJs
+ 14/3/4 +
. . .
2 (2s + 1)
J JS+1 + ...
S
. .
s( g
and
-1
a;
l)(2s
g
+ l)
"'H-lTfi
in general
= 2 m-
x.
/o
w! (ra-1)! \jVr- 1 Jn
x^r
+ (2n + l)Jn J^.
2n(2n
SI
l) ,
2n(2n + 2s-l)(2n + l)(2n

H
5)(2n
+ 2)...(2n + s-2)
"n+s l"n+a
76
30
By means of 75 and j6 the

bJl +
to
"Za^ may be transformed into
series
JJ + bJl + bJJz +
...
= a h = 2^,
63 = 16^ + 6^,
when
and
&x
[ill.
= 4aa + 2a,
64= 64a4 + 16a + 2a
b2
so on.
The arguments employed already

prove that we may write
00
0O
00
la^f = %(bw Jl +
and that
b w+l
J Js+1
s
77
),
this is arithmetically intelligible so long as
inside the circle for
This result
We
in a similar case will suffice
will
is
which Za,p?
due to
also
x remains
absolutely convergent.
is
Neumann
(Leipzig Berickte 1869).
conclude this chapter by a proof of Schlomilch's
theorem, that under certain conditions, which will have to be

examined, any function f(x) can be expanded in the form
f{x)
= rd, + OiJo (oc) + a^Jo (2x) +
where
and, if
n>
...+ an J (nx)
^/^\
|/ '{/(0) +
...,
78
*>
0,
ffl
=-
tJo
To prove
this,
we
2
u cos null
shall require
j\7j =
JT1 /(nu sin <f>)
d)
2s
+ 1 s!(s
+ l)!
have
'
therefore
2 2s
M+1
M25 + 1
2s!
s!(s+l)!"1.3.5...(2s+l)
_g (-)"w
+
(-)
(2s
M+I
uw+1 _
2)!
- cos nu
nu
"
=
III.]
Now
if
we assume
/O) = 2 a + a Jo(oo)+ ... + anJ

i
">
we
shall
31
(nsc)
...
have
= - a.J^x) - 2a J, (2x) - ... - na^^nx)
f'(x)
Write u sin $
for
between the limits
and integrate both sides with regard
ar,
and
<
thus
'
n.
/!
oo
/a
/' (w sin
to
= - 2 na
d<
<)
Jo
J^ (nw sin A) dd>
Jo
_i ^(cosrew
_j
1)
;
and therefore
mi
(u sin
'
(j))dij>
'
JO
% an cos
nu
%an
= San cosnM+
Ka
-/(0)J
So
Hence
7T
2
or,
which
f"
o-/(0)=-J
/"
jttj
/'(
sin
^)cty
dti
the same thing,
is
IT
and,
when
re
>
\
0,
<*n
On
r f/() +. / v (^ sin ^> # } du,
putting sin
cos wit |l /' (w
<J>
given by Schlomilch
%,
but
has not yet been proved.
mination of the
sion 78
and
also
sin
<)
c& > du.
we obtain the coefficients in the form

it must be observed that the theorem
All that has been effected
coefficients,
is
the deter-
assuming the possibility of the expan-
assuming that the result of differentiating the ex-
pansion term by term
is
fix).
FUNCTIONS OP INTEGRAL ORDER.
32
[ill.
In order to verify the result a posteriori, let the coefficients a

a2) etc. have the values above assigned to them, and let us
a,,
write
=s
ifr(x)
= /() + - f u
+ (hJoix) + a^J,, (2x) +
+J
0") cos
F(u) =
where
u+
/' (w sin
t 2a
<)
cos
...
2it
. .
F(u) du,
d$.
Jo
n
n,
We
have
J" (w#)
&
cos (nx sin
x cos
Now
if
a;
lies
</>
nw du
V^-w
Jo
on putting x sin
<) d</>
Jo
= .
and
between
ir,
we
are entitled to assume an
expansion
6 (it)
= C + Cj cos w + c
left
cos 2i*
. .
.,
(of w )-* from it =

to m = #, and equal to
u = x to u = ir; the isolated values m = 0, a;, ?r being
2
0(w) being equal to

zero from
out of consideration.
By
the usual process,

7tc
Jo
1
jreton
a (u) du
Jo
V#2 -u2
=H
2
u
III.]
This
of 9 (u)
tion of
is
is
the series which appears in
known
for the
y}r
(%),
whole range from
an isolated discontinuity when u
= x, we
it
sin
<f>
= ^,u cos
17
with the excep-
ir
have
sin <j>)udud(f>
_
p /'(w
2 f*
Va2 -
ttJo Jo
Put
and since the value

to
V je2 w2
it J o
_ ..q.
33
then this becomes
V^-p-ij
ttJo Jo
=/(0) + {/(*)-/(0)}
This shows that Schlb'milch's expansion
between
lies
and
it,
is valid,
provided that
and that the double integral
fMML
//:Vtf2 taken over the quadrant of a
P+ * =
t?
f-
circle
bounded by = 0,
admits of reduction to = {/()
certainly, the case, for instance, if
over the whole quadrant.
G. M.
2
7?
f(x)
/(0)}
is finite
t]
= 0,
which
is
and continuous
CHAPTER
IV.
SEMICONVERGENT EXPANSIONS.
The
power-series
Yn Jn log x
have been
which
obtained
are convergent
Jn
for
for all finite values of x,
and
but they
become practically useless for numerical calculation when the

modulus of * is even moderately large it is therefore desirable to
find expressions which approximate to the true values of Jn and Yn
when x is large, and which admit of easy computation. The expressions which we shall actually obtain are of the form
;
Jn
= a / P sin x + Q cos x\
V TTX
fl
'
Yn = a/ - {R sin x + S cos x],

where P,
Q,
R,
are series, in general infinite, proceeding accord-
ing to descending powers of
x.
It will
appear that these
series are
ultimately divergent, and the sense in which the equations just

written are to be understood
is
that
by taking a
of terms of the expressions on the right
we
number
when x is large,
The approximate
suitable
obtain,
the approximate numerical values of / and n

value of J when x is large was discussed by Poisson*, but not in a
.
very detailed or satisfactory manner; in this chapter we shall

follow the method of Stokes -f-, which is important as being capable
of application to a large
number
of functions of a kind which
frequently occurs in physical investigations.
In a later chapter
the question will be discussed in a different manner, depending on
the theory of a complex variable.
Sur
la distribution de la chaleur
dans Us corps
solides.
Journ. de l'Eoole
Polyt. oah. 19 (1823), p. 349.
t On
the numerical calculation of
(Camb. Phil. Trans,

ii.
ix.
(1856
a class of definite integrals and infinite series

read March 11, 1850), p. 166 or Collected Papers,
;
p. 329).
On
the effect of the internal friction of fluids
on the motion of pendulums (Camb.
Phil. Trans, ix. (1856; read Deo. 9, 1850), p.

[8]).
B
SEMICONVEUGENT EXPANSIONS.
IV.J
In Bessel's differential equation put
be found that u
Now when
is
small
and
is
if,
Jn {x) =
35
ux~$; then
it will
a solution of
large compared with n, the value of (n a
is
after the analogy of the process
J)/*
employed
in the
expansion of an implicit function defined by an algebraical equation f(u, x)
equation,
= 0, we
omit the term (m2 J) uja? in the differential
we obtain
d?u
d*J
of which the complete solution
where A,
=u =A
1
We
are constants.
+ U = '
is
sin
+ B cos x,
are justified
a posteriori
in regard-
ing this as an approximate solution of 81 because this value of u

does in fact make (n 2 \)ujx^ small in comparison both with u and
d?u
...
Let us now try to obtain a closer approximation by putting
+ ) cos x,
u=Uz= (A + J sin x+ (
A Au B
where
<Pu
B^ are constants.
n-i\
/_
(n*-j)A 1
2B1 -(n'-i)A
right becomes comparable with
The expression on the
we assume
This value of u gives
2A = -(n*-l)B
2B1= (n -l)A
1
x~s
if
The value
u2
=A
of
2 thus becomes
jsin
+ ^2aT C0S 4 + B
cos
~^r
sin
*}
and we3 have

rf
w2
da?
= ^Zl^=l) (- A, cos + B
+ ( i _ ZÎU
2
Zdf
/
\
sin *).
a;
be observed that we thus obtain not only an approximate solution when x is large, but an exact solution when n =
It will
or
+f
We
shall return to this second point presently.
32
36
[iV.
Let us now assume
u = u2 +
then
it will
be found that
(72,2
IS.)
(J. 2 sin
a;
+ B2 cos #).
we put
then,
If,
+ Bi cos x
sin
^ _
2
(n-j) (-|)
274
~~
"'
"'
2~74
the value of u becomes
u = us =
A
,
+ i>
n2~
-i*
f
jsm
a; +
cos
jr- sin
a;
icos
a;
(n 2
2x
-i)(w2 -|)
4a;
sin
a;^
cos x\
J
and we have
Proceeding in this way, we find by induction that,
we put
B Vr
ur = AJJr +
tf-l4
.
(n
U
cosa;-'
r = smx+
a
where
if
82
-i)(n2 -l)
?/
'^
sina;-...
2 4a?
2a;
/
B
w'-g)...(n'-(r-iy)
r-1
+ ( '-i)(
***[* +
2.4.6...(2r-2)af-i
K r = cosa;
^siaa;-2a;
^-V-;2
2
cosa;
V'
83
+ ...
4a;
tf
r-1
/
| ).,.(tf-(r-ty)
C S
+ ( -i)(.\" +
2.4.6...(2r-2)ar-i
V'
SEMICONVEKGENT EXPANSIONS.
IV.]
CPU,
,,
'
37
n*-\\
(tf-j)(n-f)...(n'-(r-4 )')f ,
r-1
/
A
L sin
sin(.+
2.4.6...(2r-2)af+1
.
+ oosâ;+^-5 ttU.
The expression ur
is
7T
85
therefore an approximate solution so long
as
(n'-j)(n-f)...(n-(r-}))
2.4.6...(2r-2)a;'-+ 1
is
small
and u r
*s
is
a closer approximation than u r
numerically
less
so long as
than 2{rl)x.
It is to be observed that the closeness of the approximation has

been estimated throughout by the numerical smallness of the ex-
pression
di%.
n?
- \\
u rt
the value of which
is zero for an exact solution

that it is not to
be inferred that the expression above given for ur reflects in any
adequate way the functional properties of an exact solution and
;
finally,
that as r
is
taken larger and larger the expression ur
more and more (in every sense) from a proper solution.

These considerations, however, do not debar us from employing the
expressions u r up to a certain point for approximate numerical
deviates
calculation;
and the analysis we have employed shows the exact

ur is the approximate value of a solution, and the
sense in which
degree of the approximation obtained.
by putting
It is convenient to alter the notation
A
A
C
x+
= G cos (a x),
sin x = C sin (a x),
cos
a being new constants:

Al + Bl, which is independent
and
C* =
B
cos x B
sin
this
of
x.
is
legitimate,
86
because
Then we have
ur = C{Pr cos (a x) + Qr sin (a- x)},
87
38
[iV.
where
2
(4rc
r
-l)(4re2 -9)
(8a;) 2
1.2
+
4w2
yr__ 8^
and
(4w
- 1) (4w - 9) (4w - 25)(4n - 49)
-' 88
1.2.3.4(8)
(8*y
2
1) (4w
2
9) (4n
- 25)
1.2.3(to)s
Qr between them
r,
(4w 2
+ '"
contain r terms, involving
89
x,
or1 or2,...
,
_r+1 respectively.
i
What
remains to be done
is
to determine the constants
and
m^ - * may
be an approximation to the special solutions

of Bessel's equation which are denoted by Jn and t/_ or, when n
a so that
by
is integral,
For
Jn
Yn
and
we
this purpose
may
positive, it
xn
^ = ^-
To show
this,
2n
U(n- ))o
we observe
^^=l
/_ \S
since s
and
(2s)
^
5
n(2s)
therefore
and
y.j.
a positive integer
cos (x cos
</>)
sin-
JQ
real
Sinm * ^
ens 28
2s
{f>
^ sin-^tty
'
n(2s)n(^+s)
(2s-
1) (2s
- 3).
.1
^ = -~^
IIs'
2 2S IIs
2 2s
d+ = Vrfl
*)
yvn (w - i)
^^^
.
-" / (),
This proves the proposition.
14.
Now
consider the integral

7T
cos (x cos
sin
2
</>)
< d<f>
=2
Jo
and put
and
n(n-i)n(-i)
is
= 2 n (s)
g&S
cos(cos0)sin 2
Now
+ \ is
that
_|/
7^
by
COS {X C0S
Tir
/tt
Jn in
expression for
Supposing that n
be proved that
1
Jn
new
shall require a
the form of a definite integral.
Jo
cos
$=
fj,
cos (x cos
<)
sin2
</>
dtp,
IV -]
SEMICONVKRGENT EXPANSIONS.
then
and, on
writing
integral
becomes
cos
x cos fix + sin x sin fix
U=
where
=2
J7,
dd>
cos
Uj,
39
x+
cos(cos0),
for
the
TJ2 sin x,
cos /c (2/*
- /*)-* du,
sin /b
- /i )-s d/i.
.'o
=2
f/a
(2/*
In the integral
then
F= (% jfl -^-Y
where
By
^ put fix = t:
it
assumes the form
-lli-icos*d.
breaking up the interval from
x into the
to
2r
('I)'
(Y'T)-(
(!'t)'
the greatest multiple of
is finite,
term in
ir
Ti7r
contained in
however large x may be
'
x, it
')'
intervals
wliere(7'- :l)7ris
may be proved
that
hence the most important
is
xn+if
P-l cob tdt.
This represents the true value more and more nearly the larger
at the same time approaches the value
x becomes, and
n+i
%m+i
2 * + rr /
f *-*
cos
_ td*
^, =
_ 2"+*
in m
II (-*)
/
op"'
^+jJJoo
(supposing that
of
a;
2w
+ 1) *
is positive).
In the same way it may be proved that the approximate value

x is very large and positive is
2
U when
2 n+ *,
(n A)
3-,
II
2n
.
1N sin (
i
+ 1 ) 7T
t-
40
and hence,
to the
same degree of approximation,
xn
^=2V^nTî)
=v
Comparing
[IV.
^lC0Sa,+
cos
f-^4
we
this with 87
UlSina!}
91
see that
-^-
<>-</{
and hence that
^()- Av/7TrP cos
f(ft,
1),r
(2W
Qsini
+ 1) 7T
92
...
D . "^(W-l)(4-9)
WlfchP=1
2!(L) 2
+
s
n _4n -l
(4rc
(4n,
2
1) (4ra
2
9) (4ra
4!
2
(8a?)
-l)(4?i2 -9)(4i2
- 25) (4k - 49)

2
-25)
3!(8a?) s
8
For the sake of reference
it is
the corresponding expression for

result of Hankel's (Math.
Ann.
convenient to give in this place
Yn which
I.
may
be deduced from a
(1869), pp. 471, 494) to be con-
sidered later on.

If
by
we
write 7 for Euler's constant, which
- ^(0)
or
-IT(0)
-r-
11(0),
if
log 2
is
56649 01532 86060
the natural logarithm of
log 2
otherwise denoted
and the value of which
7 = _Tjr(0) = -57721
and
is
2,
is
65...
93
the value of which
is
-69314 71805 59945 30941 72...,
then the approximate value of
Yn
when x is a
is,
94
large real positive
quantity,
Yn (x) = (log 2 - i) Jn (x) +
*J
[Q
cos
pV
P sin i-
*}
95
IV.]
the values of
Jn (x),
P, and
Q on
41
the right hand being those given
above.
Thus
to the first degree of approximation
7.-v<*.-,,..|Ss+!-.}
I
ir
v^
= (l<>g 2 - 7) Jn +
The value
of log
I +
2 7 is,
~ Jn
ii+i
96
twenty-two places,
to
15156 58412 44881
It is interesting to confirm these results
J n+i n
!i
log2-7 = -11593
tion
+ l)7r
((2n,-
sin
07.
97
by means of the
=~
~**"
The approximate formulae for Jn and Yn hold good

x with a large modulus, provided the real part
values of
positive,
and that value of
a;
is
kJx is
for all
of
is
taken which passes continuously
when x
into a real positive value
when
rela-
is real
and
The
positive.
case
a pure imaginary will have to be considered separately
in a subsequent chapter.
We
n
will
now
briefly consider the case, already alluded to,
the half of an odd integer.
is
By
Jn (x), we
the general definition of
T-*
X~i
_
^r-i n (- i)
have
X*
~2TT + 2.4.i.3
COS X,
7TX
and
Ji
= ,
X2
J
{ 1
-Z /1X
2*n(i)t
/2a?
sin
Hence and by means
sJ7 +
2.3
x_
'
/ 2
X*
2.4.3.5
sm#.
of the relation
2n
Jn+l
~ "n ~ " n
when
42
we may
[IV.
Jk+i where k is any positive

functions thus obtained are of impor-
calculate the expression for
or negative integer.
The
tance in certain physical applications, so that the following short

table
may be
2n
useful.
IV.]
differential equations will observe
43
"
that the " indicial equation
corresponding to
<Pu
dx*
1 du
x dx
r2
is
so that
T+
when
2ra is
re
=o
i--Au
x-
n =
2
an odd integer we have the case when the
roots
of the indicial equation, although not themselves integral, differ
an integer
and
it is this
circumstance that gives
ceptional character of the solutions.

Differential Equations, vol.
I.
(Cf. Craig,
(1889), chap. 5.)
rise to
by
the ex-
Theory of Linear
CHAPTER
V.
THE ZEROES OF THE BESSEL FUNCTIONS.
For
the purpose of realising the general behaviour of a transit is important to discover, if possible, the
cendental function
values of the independent variable which cause the function to
vanish or become infinite or to assume a

It has already
value.
finite
and continuous
maximum
or
minimum
been observed that the function
for all finite values of
we
a;;
will
Jn (x)
now
is
proceed
to investigate the zeroes of the function, that is to say, the values
of
x which make Jn (x)
that n
is
important case
On
= 0.
It will
be supposed in the
first
place
a positive integer, or zero, this being by far the most

for physical applications.
account not only of
ness and simplicity,
its historical interest,
we reproduce
Jn (x) =
the theorem that the equation
but of
its direct-
here Bessel's original proof of
has an infinite number of
real roots *
It has
been shown
(see
46 above) that
w
(x)
= - cos (x sin
t Jo
= 2- P
I
ir
m' a positive proper fraction
.
,
Vl-*
where
m is a
Vl-
cos
ttJo
positive integer, and
then
ttJo
dd>
dt
/ a
cos (xt)
TrJo
Suppose that x =
d>)
V{(2m
+ m') - v%'T
2
Berlin Abhandlungeft.(1824), Art. 14.
'
THE ZEROES OP THE BESSEL FUNCTIONS.
V.]
dv
7TV
Now
cos
+ m') - a
2
2 V{(2m
and hence, by taking a
I 2
+ m') - {h + uf}
2
2 J V{(2m
= h-l,b = h +
and putting A, succes(2m -1), we obtain, on writing fi for
5,...
1, 3,
C S
J^ft
sively equal to
45
l,
2m + m',
H
J<>-;/:
7TO
W=
Si!
V/x
-(1+m)
V/i2
-(3 + w) 2
(-1)"
^
V/t2
m
,2,
+ - (- 1l)m
cos
/'
The
1
2 V/u2 -(2m
-(A + m)
Sm
U,
therefore positive
m
ff
-(A + M)2
du
7Ttt
ITU
cos
T2 vV - (2m +
2
evidently positive.
VM- (A -)}**'
moreover
'
I.
fs
i.
is
=-.
+ M)
du
2 V/* 8
and
-=-
J,
cZm
integral
ttu
/:
du
iru
'
>.
7T
-(2m-l+)
Therefore
/"
w) 2
has been reduced to the
(a;)
form
Jo(aO
where
Mj,
w2
...
= - Wi +!-*+
um are a
u1 <u2 <u3
Therefore
+(-)"*,
series of positive quantities

...
<Mm
and
(x) is positive or negative according as
is
even
and consequently as x increases from kir to (k + 1) ir where

k is any positive integer (or zero) J (x) changes sign, and must
therefore vanish for some value of a; in the interval.
or odd;
This proves that the equation

of real positive roots
(x)
has an infinite number
the negative roots are equal and opposite to
the positive roots.

It has been
value of
shown
in the last chapter that the asymptotic
(x) is
\/^x
C0S
X
{
-l)'
THE ZEROES OP THE BESSEL FUNCTIONS.
46
so that the large roots of J
where k
To
k=9
is
are approximately given
(as)
[v.
by
= (k + f) IT,
a large positive or negative integer.
give an example of the degree of the approximation, suppose
then
(*
+ f)ir = 30-6305...,
the true value of the corresponding root being

306346....
It has
been proved that
JlW ~
now between every two
function J' (x)
successive
must vanish
'
dx
of
roots
(x) the
derived
and therefore between

there must be at least one
at least once,
J (x)
every pair of adjacent roots of
Thus the equation J"x (x) = has also an infinite

number of real roots we infer from the asymptotic value of Jx (x)
that the large roots are approximately given by
root of Jj (x).
x = (k
where
k,
Now
as before,
let
is
n be any
any large
+ |) v,
integer.
positive integer,
(!)
Jn (x) = Rn
4
then
it
may be
verified
Rn+1 vanishes
~?
'
from 16 or io that
n+1
so that
and put
when
"~
R, is a
dz
'
maximum
or a
minimum.
We
between any two consecutive roots of Rn =

there must be at least one root of R n+1 =
and hence, by induction, that the equation Jn (x) =
has an infinite number of real
infer, as before, that
roots.
The
proved,
Jn (x) having thus been

remains for us to devise a method of calculating them.
existence of the real roots of

it
We will begin by explaining a process which, although of little or

no practical value, is very interesting theoretically. Let and Rn
have the meanings just given to them, and put
V.]
where /,(()
is
the
rest of the series.
47
sum of the first s terms of Bn and

Then when s is very large s (f) is
,
(f) is the
<j> s
<j>
very small
and therefore, if /3 is a root of Rn fs (@)

be very small for a sufficiently large value of s. For
for all finite values of ,
must
also
large values of
equationyj(f)
and even
for
s,
be real roots of the algebraical

which are approximations to the roots of Rn =0;
therefore, there will
moderate values of
s,
we may expect
to
obtain
approximations from /s (f ) = 0, but we must be careful to see that

<s (/3) is small as well as/ (/3) = 0.
If the real roots of the equations
f=
s
are plotted
off, it
will
be found that they ultimately fall into groups or clusters, each

cluster " condensing " in the neighbourhood of a root of
n = 0.
The points belonging to a particular cluster are, of course, derived
= 0.
from different equations fs
As an
illustration,
suppose n
then
tt
1 2 .2 S .3 2
1 2 .2 2 .3 2 .4 2
fc2
"
ff
"l>.2 2
and the equations fs
"'
are
-1 =
^-4^+4 =
f-9f + 36-36 =
4
and
-16f + 144 - 576^ + 576 = 0,

2
so on.
The
real roots of these are, in order,

1
2,
1-42999...
1-44678..., 5-42...
and we already see the beginning of a condensation

If we put
sion 142999, 1-44678.
?4 = 1-44678,
we
x=
find
and the
2-405...;
least positive root of J {x)
xx
so that x
= #i
is,
= 2*4048...
to three places of decimals.
in fact,
in the succes-
48
The
best practical
method of
calculating the roots
[v.
is
that of
Stokes*, which depends upon the semiconvergent expression
To
fix
the ideas, suppose n
for
Then we have approximately
0.
M) = \/^ P cos (* - i) + Q sin (* - l)\
g2
P = i-^^ +
J2
where
1 2 .3 2 .5 2
y ~8^
Jf =
then
Now
Jo
3!(8#)
it is
VP + Q
2
not
we
2
,
1 2 .3 2 .5 2 .7 2 .9 2
5!(8) 6
^ = tan-
to find
We
large that
we
an approximate value
for
VP + Q
2
are justified in calculating

;
-^
when x
that
and Q
and Q were convergent series

same kind of meaning as the
derived.
""
4!(8^
difficult to see
use a few terms of
(x),
!p
52
P = Jfcos^>, Q=Msin^;
Put
may
g2
J2
is so
and tan -1 ~
the results, of course, will have the

series
and Q from which they
are
thus obtain the semiconvergent expressions
M=l-'
16a
5*
2
+l
512a4
_/l
33
3417
\
* = tan (8a-512^ + 1638l^--J
.
The value
as if
of
(x) is (approximately)
which vanishes when
x = (k l)v
yjf,
k being any integer.

Write, for the moment,
Camb.
Phil. Trans, ix. (1856), p. 182.
98
THE ZEROES OF THE BESSEL FUNCTIONS,
V.]
49
then we have to solve the transcendental

equation
-0 + t M i-(
.MlT
*-JL
+
512^^
\8x
on the supposition that
(f>
x=A
Assume
16384a; 6
and
a;
"7)
are both large.
+ ~ + + -2.+
then, with the help of Gregory's series,

,
* + i + Js + + F
and therefore
= 1-
Substituting for
its
</>
- = (* - i) +
or,
31
= - -
value (k
27r 2
(4J
etc.
) ir, we
j)
have
6?r 4 (4fc
finally
_!)
reducing to decimals,
^=
tt~
.g-
D+
1.
"050661
-053041
4A-1
(4A-l)
The corresponding formula
^-z.
v- k +
'
ok
25
'151982
"262051
(4&-l)"~""
for the roots of Ji
-015399
(a;)
is
is
-245270
-^h^ + (ikTTf-(4k + iy +
The same method
ioo
--"
I0
applicable to Bessel functions of higher
orders.
The general formula

X~a
for the
m-1
&th root of
4 (to- 1) (7m
3(8a)
8a
Jn () =
is
-31)
2
_ 32 (to - 1) (83m - 982m + 3779) + "
15 (8a)6
where
G.M.
a=
\ir (2n
4th),
m = 4re
2
.
ioo"
50
[v.
This formula is due to Prof. MMahon, and was kindly communicated to the authors by Lord Rayleigh. It has been worked
out independently by
reasonable doubt of
Mr W.
St B. Griffith, so that there
its correctness.
may be remarked
It
Stokes gives the incorrect value "245835
for
is
no
that
the numerator of
the last term on the right-hand side of 101'; the error has some-
how
arisen in the reduction of
H79/57T6 which
,
is
the exact value,
to a decimal.
The values
of the roots
may
also be obtained
by interpolation
from a table of the functions, provided the tabular difference
is
sufficiently small.
The reader
will find at the
end of the book a graph of the
functions J (x) and Jj (x) extending over a sufficient

show how they behave when x is comparatively large.
interval to
seems probable that between every pair of successive real

Jn (x) there is exactly one real root of Jn+l (x). It does
not appear that this has been strictly proved; there must in any
case be an odd number of roots in the interval.
It
roots of
With regard
to values of
n which are not integral, it will be

n is real the equation
sufficient for the present to state that if
Jn (x) =
has an infinite
number
the roots of the equation are real.
of real roots
and
if
n>
1, all
CHAPTER
VI.
FOURIER-BESSEL EXPANSIONS.
One
of the most natural ways in which the Bessel functions
present themselves
This has, in
now
will
fact,
in connexion with the theory of the potential.
is
already appeared in the introduction
(p. 3);
we
consider this part of the theory in some detail, adopting,
in the main, the
method
of Lord Rayleigh (Phil.
Mag. November
1872).
If
we use
V < = 0,
s
cylindrical coordinates
which must be
6, z,
r,
by a
satisfied
Laplace's equation
potential function
<,
be-
comes
+r
9r s
Assume
where k is a
and z. Then
</>
+r
dd"
dz"
= ne'*" cos nd,
102
and u
real positive quantity,

<j>
is
independent of 6
will satisfy Laplace's equation if
d?u
du
ns N
A Jn (kt) + BJ-n (Kr)\

u = AJn (,cr) + B7n(Kr) J
u=
whence
or
according as n
Br
is
not or
is
104
an integer.
Consider the two particular solutions

(
yfr
= e~
lcz
= e~
cos n6Jn
(icr)
105
>,z
cos
nOJn (\r) J
42
52
FOTJRIER-BESSEL EXPANSIONS.
V$=
2
then since
= 0,
and VS/r
[VI.
Green's theorem gives
the surface integration being taken over the boundary of a closed

space throughout which <f> and i/r are finite, continuous, and one-
valued
and ^~ denote the space

rates of
r
as usual, J*-
av
of
</>
and
yfr
av
along the outward normal to the element dS.
suppose that n
First,
Then we may
positive.
variation
the cylinder r
= a,
is
an integer, and that and X are
integrate over the surface bounded by
and the planes z
0, z
=+
oo
When z = 0,
<j)
= cos ndJn (kt), - = X cos n6Jn (Xr),
and the part of the integral on the left-hand side of 106 which
is derived from the flat circular end bounded by z = 0, r = a, is
[** fa
Xj
Jo io
rdddrcos*n9Jn (/cr)Jn (\r)
fa
= Tr\l Jn (/cr)Jn (Xr)rdr.

Jo
When z = +
oo
</>
^-
vanishes,
and nothing
is
contributed to
the integral.
For the curved surface r
-' z
= a,
cos nOJn (ko),
^ = Xe~
Kz
cos
so that the corresponding part of the integral
Xjn (m) J'n (Xa)
f%* f">
is
ad0dze-(K +V z cos^nd
JO Jo
Working out the other

fa
nQJ'n (Xa),
side of 106 in a similar way,

-\
/ (kv) Jn (\r) rdr + ^-
*C "T" Ai
=k
obtain
Jn (m) J'% (Xa)
Jn ( Kr) Jn (Xr) rdr +

/
we
J'n (tea) Jn (Xa)

-\-
A*
'
VI -J
53
or, finally
- \)
Jo
Jn (kt) Jn (Xr) r dr
The very important
conclusion follows that
if
k and
are
different,
Jn (icr)Jn (Xr)rdr = 0,
108
J o
provided that
XJn (xa) J'n (Xa) The
condition 108'
kJ^ (ko) Jn (Xa)
is satisfied,
among
X are different roots of
Jn (ax) = 0,
if k,
(ii)
if
they are different roots of J'n (ax)
(iii)
if
they are different roots of
and
108'
0.
other ways,
(i)
AxJ'n (ax) + BJn (ax) =

where
are independent of x.
= 0,
0,
All these cases occur in
physical applications.
In the formula 107 put k
make h
X + h,
divide both sides by
we
decrease indefinitely; then
find,
h,
and
with the help of
Taylor's theorem, that

a
\
Jo
Jl(Xr)r dr
= ~r {Xa [J'n (\a)] - Jn (Xa) J'n (Xa) - XaJn (Xa) J

2
Reducing
this
(Xa )}.
with the help of the differential equation, we
obtain finally
V(Xr) rdr =
j
| {<C(M) + (l -^) J2(Xo)J
109
When n is not an integer, we may still apply Green's theorem,

provided that in addition to the cylindrical surface already considered we construct a diaphragm extending from the axis to
the circumference in the plane
boundary,
first for
0,
= 0,
and then
and consider
for
= 2ir.
this as a double
54
When
= 0,
and when 9 =
(f)
[VI.
2tt,
Gil?
= e " KZ cos 2mrJn (r),
-^-
th
-Az sin
2mrJn (\r)
therefore the additional part contributed to the left-hand side of
106
is
n sin
2n7r cos 2?wr
J" (-r)
Jo Jo
Now
this is
Jn (\r) drd^.
symmetrical in k and \; therefore the same ex-
pression will occur on the right-hand side of io6, and consequently
the formula 107 remains true for
all real
same way the formulae 108 and 109 are true
To show the
values of
n.
In the
for all values of n.
application of these results,
we
will
employ the
function
problem in the conduction of heat.

Consider the solid cylinder bounded by the surfaces r = 1, z = 0,
to obtain the solution of a
=+
30
medium
and suppose that
convex surface
its
is
Then when the
of temperature zero.
become steady, the temperature

must satisfy the equation
surrounded by a
flow of heat has
at any point in the cylinder
V F=0
2
and moreover, when r \,
dV
or
where k is the conductivity of the material of the

h is what Fourier calls the " external conductivity."
If we put V=<f>, the first condition
second will also be satisfied, if
\kJ^ (\)
+ hJ
(\)
= 0.
is
cylinder, and
satisfied;
and the
10
Suppose, then, that \ is any root of this equation, and suppose,

moreover, that the base of the cylinder is permanently heated so
V
VL]
55
that the temperature at a distance r from the centre is

Then the temperature at any point within the cylinder is
because this
(\r).
the conditions of the problem.
satisfies all
The equation no has an infinite number of real

that we can construct a more general function
roots
X^X,,
etc. so
= !-4
in
e- A s J(\ sr)
and
the temperature of the same cylinder
this will represent
when
subject to the
same
at any point of the base
conditions, except that the temperature
is
<j>
now given by
= 2A J
8
112
(\ 8 r).
l)
Now
there does not appear to be any physical objection to
supposing an arbitrary distribution of temperature over the base

of the
cylinder, provided
from point to point and
may
the temperature varies continuously
everywhere
is
finite.
In particular we
suppose the distribution symmetrical about the centre, and
put
<h=f(r)
where f(r)
is
any function of r which
continuous from r
to r
= 1.
is
one-valued, finite and
The question
is
whether this
function can be reduced, for the range considered, to the form
expressed by
1 1 2.
Assuming that
this is so,
f(r)
it
we can
in the form of definite integrals;
follows
by 107, 109, and

(VO/(r-) rdr =
at once obtain the coefficients

for if
= -ZA J
8
we put
(\ sr)
no that
A f J* (\ r) rdr
s
Jo
Jo
and therefore
2/c
X2
113
FOURIER-BESSEL EXPANSIONS!
56
Whenever the transformation
$=
'ZA s e-
3 is legitimate, the function
1 1
*J
{\ s r)
gives the temperature at any point
convex surface, as before,
[VI.
115
of the
cylinder,
temperature, and the circular base
is
when
medium
surrounded by a
is
its
of zero
permanently heated accord-
ing to the law

</>o
the coefficients
much more
=/(*)
= 24,J
far);
being given by the formula
14.
general form of potential function
is
obtained
by putting
= 2 (A
<
where the summation

If
we
cos
n6
+ B sin n6) e~^Jn (Kr)
refers to
to integral values and take for
Jn (X) = 0,
the positive roots of
we have a
-
when 6 is changed
and also when r=l.
potential which remains unaltered
which vanishes when
when
16
n and X independently.
restrict the quantities
the quantities
z= + <x>,
into 6
+ 2ir,
The value
is
= 2 (A
cos
n0
+ B sin
n6)
Jn (\r).
1 1
The
function
</>
may be
interpreted as the. temperature at any
when the
point in a solid cylinder
flow of heat is steady, the

convex surface maintained at a constant temperature zero, and
the base of the cylinder heated according to the law expressed
by 117. We are led to inquire whether an arbitrary function
f(r, 8), subject only to the conditions of being finite, one- valued,
and continuous over the
circle
member
of the right-hand
= 1,
can be reduced to the form
of 117.
Whenever this reduction is possible, it

Thus if, with a more complete
coefficients.
f(r, 6) = 22 (A
we
nj ,
cos
nO +
Bn
,
_
is
easy to obtain the

'
we have
notation,
sin nff)
Jn (\ r)
g
find successively
2ir
f(r, 0) cos nO dd
= LitA
a> s
Jn (X
sr)
/,
and
fir, 6) cos
n8Jn (Xsr) rdddr = irA n
Jl (\r) rdr
Jo
2 " (^)
-^,
1 1
VI.]
by 109;
so that
AiS =
and
57
f 1 [to
f (r, 6) cos n6Jn (\ r)rd0dr;
f''iri~\\
same way
in the
n8'
r1
TrJ'*(\)
r^
-f( r >
e)
&mnGJn{âr)rdBdr
Other physical problems may be constructed which suggest

and 118; some of these
are given in the Examples.
analytical expansions analogous to 113
We
do not propose to discuss the validity of the expansions
way would
and require a dis-
obtained in this chapter; to do so in a satisfactory
many
involve a great
amount
proportionate
discussions to
delicate considerations,
And
of space.
after
all,
the value of these
the practical physicist, in the present stage of
applied mathematics,
is
not very great
for the difficulties of the
amount
must be applied to an arbitrary function in
may admit of expansion in the required way, and
analytical investigation are usually connected with the
of restriction which
order that
in
the
it
physical
these
applications
restrictions
generally
are
from the nature of the case. Such a work as Fourier's

Th4orie Analytique de la Chaleur is sufficient to show that the
instinct of a competent physicist preserves him from mistakes
satisfied
in analysis, even
plicated
when he employs
and peculiar
functions of the most com-
description.
For further information the reader

KugeJfunctionen, 2nd edition, Vol.
II.
is
p. 210,
referred
to
Heine's
and to a paper by
the same author, entitled Einige Anwendungen der Residuen39.*

t. 89 (1880), pp. 19
rechnung von Gauchy, Crelle,
As an example
of the conclusions to which
Dr Heine
is
led,
consider the expression on the right-hand side of 113, where the

coefficients
are determined by 114; then if the infinite series
%A J (\ r)
s
is
uniformly convergent,
its
value is/(r).
Similar considerations apply to the expansion 118, and others

There are a few elementary arguments, which,
of the same kind.
* See also the papers by du Bois-Eeymond (Math. Ann.,
vi. 146),
and Hankel
(ibid. viii. 471).
iv. 362),
Weber
(ibid.
58
although not amounting to a demonstration,
[VI.
may
help to explain
the possibility of these expansions in some simple cases.
Suppose, for instance, that
and continuous
finite,
<f>
a function which
is
= 1,
over the circle r
all
is
one-valued,
and admits
of an
absolutely convergent expansion
$ = Ug + Mi + M2 +
where us
in
= Ug,
a homogeneous rational integral function of degree
is
Cartesian coordinates
= r sin 0,
GO
. . .
x,
Then by putting x = r
y.
cos
s
6,
this is transformed into

<
= v + v r +
x
CO
r2
t>2
. . .
Svgr8
where
By
0,
an integral homogeneous function of cos
v is
and
sin
0.
expressing v 8 in terms of sines and cosines of multiples of
and rearranging the terms, $ may be reduced to the form
= p +
= 2 (p
(/i
s
+ <y
cos
cos s0
o-s
sin 0)
+ (p2 cos 20 + <r2 sin 20) +
where ps and a s are, generally speaking,

each beginning with a term in r*.
Now
\1; \2
if
J* {\f), Js
are the roots of
etc.
(*2")> etc.
. .
sin s0)
infinite power-series in
(\)
= 0,
the series
each begin with a term in r3 and

,
= A -? Je (V) + AfJ (V) +
if
r,
for
we put
. .
may be possible to determine the constants Af, A 2 etc. so as

to make the coefficients of the same power of r on both sides agree
If the result of carrying out
to any extent that may be desired.
s)
it
this comparison indefinitely leads to a convergent series
A^Js (V) + AfJs (Xar) +

we may
and in
. . .
legitimately write
like
p s = tAWJs (\ mr),
arrive at a valid formula
manner we may
and then
<p
= 2 [cos s9%A% J
which
is
equivalent to
1 1
(\s r)
+ sin s0!B% J (V)]

s
wi
8.
It will be understood that this does not in
any way amount
to
a proof of the proposition: but it shows how, in a particular

case of the kind considered, the expansion may be regarded as a
straightforward algebraical transformation verifiable a posteriori.
CHAPTEK
VII.
COMPLEX THEORY.
Many
properties of the Bessel functions
by means
illustrated
may be
proved or
of the theory of a complex variable, as ex-
plained, for instance, in Forsyth's Theory of Functions.

few
of the most obvious of these applications will be given in the
present chapter.
To avoid unnecessary
that n + ^ is
throughout
contrary
complication,
expressly stated,
is
and
real
it
will
positive,
that the
be supposed
and, unless the
real part
of
is
also
positive.
Then, as already proved
Jn (a,,)
(p. 38),
xn
^ cos ^ sinM ^'
cos
2VwU(-i) i o
Now
["cos (x cos
<f>)
sin 2 " 0cZ<
=
=
cos
from
r+i
xn\
and this
to
1 to
- 1*)"-* alt
xti
(l-t*)n-*dt
is
+ 1. We may
taken along the axis of

therefore write
(l-^) n-^dz = 2 n ^TTll(n-\)Jn {x),
119
remain true for any finite path of integration from

which is reconcilable with the simple straight path.
will
+1
ea,d
(1
+
(
on the supposition that the integral

real quantities
(art)
COMPLEX THEORY.
60
We are
thus naturally led to consider the function
u = xn
e xli
where
[VII.
are independent of
a, y8
(l-
z*)
n -i dz
20
a.
x.
It will be found that
du
x dx
d?u
dx*
/..
2N
a?)
f {x(l- z
= xn~
(2w
1) iz) (F* (1
- )-*
dz
J a
= - ix*-
therefore
{eF* (1
eF
when
=a
and
(1
also
n,
+ 1, 1 and
where 4
k+
cci,
we
Under the
restrictions
the admissible values of a and

is
assigning these special values to
paths of integration,
if
2 )" + i
2
when z = ft.
imposed upon x and
be expressible in the form
Each
is
are
number
of solutions in
solution must, of course,
AJn + BJ^,
termination of the constants
/8
any finite real quantity. By

a and /3 and choosing different
obtain a large
the form of definite integrals.
is
dz
**)+*}
be a solution of Bessel's equation
will
zi
vanishes
or
AJn + BTn
the de-
not always easy, and in fact this
the principal difficulty that has to be overcome.
must be carefully borne in mind that, in calculating the

value of u for any particular path, the function (1 ^ 2 ) B_ * must
The value of u is not deterbe taken to vary continuously.
minate until we fix the value of (1 z*)'1 at some one point
If U is one value of u for a particular path, the
of the path.
general value for the same (or an equivalent) path is
It
M==e (2m-l)fc
where k
is
any
real integer.
Consider the function
ux = xn
we may choose
1
to
0,
*"
\
zi
eF
(1
n -
dz
121
as the path of integration a straight line from
followed by a straight line from
to
oot.
Then
if
COMPLEX THEORY.
VII.]
denotes a real variable, and
u1 = xn \\
e*
we take
if
- )-i dt + if
K (1
- 2 )"-*
2
that value of (1
when z = 0,
+1
which reduces to
61
+ )"-* dt
e~ xt (1
IT
= xn
cos
sin
(a:
cos m <bdd>
tf>)
Jo
IT
ixn
sin (x sin <) cos 2"
<f>d(j>
e~* ^""fr cosh 2"
c/>
d^l
For convenience,
let
us write
CB =2V*rII(n-i);
122
xn
then
cos
(a;
sin
and
if
<f>)
# d</> = |0/n
cos 2"
we put
r2
s"
J
sin (x sin
</>)
cos 2"
d<
Jo
-as"
I
e-x
^ h ^>c6sh m ^d(j) = ^Cn Tn
123
Jo
is
also a solution of Bessel's equation,
Ui
and one value of
u^ is
= -^Cn (Jn + iTn );
124
y^
the general value being
where k
is
any
0(271-1) fori
U^
25
real integer.
In the same way,

Ui
if
= xn
Vzi
(1
taken along a straight line from

line from
to ooi, one value of w 2 is
and the general value
- z'y-i dz,
1 to 0, followed
U,
= ^Cn (J n -iTn)
M2
= e P-l|*rt
is
J72
126
by a straight
127
COMPLEX THEORY.
62
Now
[VII;
consider the integral
= xn
us
*e* (1
z*)
taken along a path inclosing the points
throughout the integration

value, or modulus, of
\z\
>
where
1,
- z*) n -l = e^-D"
and therefore one value of m3
and + 1, and such that

means the absolute
- ( - i) z-* +...},
{z-1
is
may put
given by
integral on the right-hand
side
may
pv**.
be evaluated by
taking the path along the axis of imaginary quantities from

to
ei,
then round the origin, and then from
value of the result

tfn-w
g-xt âi-zs-i fa
eniri (e4""
= (-)" 2ieim
'ri
Therefore one value of

u, = a* 4 "* sin
ei
to +ooi.
+ xi
One
is
= (- )
which
28
\z\
,^'r (-/<-*)<>(-+*)
The
z.
Then, by the binomial theorem, we

(1
n -l dz,
tfrai fin-y,
- 1)
g-xt pn-as-i
e~ xt P*-*-'
<ft
'0
Jo
II(2n-2s-l)
sin 2?wr -
u-6 is
given by
2mr 2 (g T*X"-t)-("-* + i)
n (2ji - 2s - 1) -.
Observing that
(-j)(-t)...(n- + })IT(2n-2-l)
ns
1
"a
28
n(2w-i)n(w-a-i)
n (, - 1) n (s)
and that
II (n
1) II (s
w) = ir cosec (w s) ir = ( )* 7r cosec imt,
'
COMPLEX THEORY.
VII.]
63
we have
fi,
= 4* cos , ^< 2w -\> J

n (m - 1)
2+i tt*
:
n(-n-i)
C_V
<-^
/it-+m
2 s8
n (s) n (s - w)
/
/_(*)
on reduction.
The general value

we have the
is
us
= e ~1)kni u
and by putting k
{!!n
=-
special value
^3
= FT7<--*)'
I T\ J-n ()
= 2GH COS W7T /_ (*),
129
and the general value
been convenient to suppose that \z\ >1 throughout the

but the value of u3 will not be altered if we suppose
the integration taken along the path ABGDEFGH represented in
It has
integration
the figure below.
&
-a
If the argument of (1 2^)" -J is taken to be zero along DE, we

must take it to be (2m 1) it along AB, BG, and + (2n 1) ir
along FG, GH. Hence we have one value of w s given by
,"+1
u3 = -er 2"- 1 "

'
U + e" "-1 "

2
'
E/i
+ of
1
1
= Cn [Je- "' (Jn - iTn) + \e (Jn +

2
= 2Cn cos nir {Jn cos nir Tm sin mr).
e*
M (1
i"Tn )
- t )n ~ i dt
i
+ /]
1
30
COMPLEX THEORY.
64
Since n
is,
is real,
this
with 2Gn cos nir J_ n
must agree with the real value

Hence we infer that
J-n = Jn cos
The formula
nir
still
is
that
131
an integer, 131 becomes an
To
not really indeterminate.
is
by which Tn was defined, is intelligible only

x is positive we may, however, if we
a definition of T for all values of x. Then 123
remains valid so long as the integrals are
Tm
w3
123,
regard 131 as
When n
of
of
T sin W7r.
so long as the real part of

like,
[VJI,
finite.
but the value
identity,
find its real value, write
n e for n, e being a small positive quantity, and n being supposed

an integer then
;
</_(_,,)
- (-)" cos
Divide both sides by ( )n
The
ejrJn- e
= (-)" sin 67rT^
and make
vanish
expression on the right-hand side
may be
thus
written in the
form
{xy+<
"- 1 (-)"+*
{2)
n()
(A* +
en(-n + e + s)\2)
(a>\
\2)
,_
v
K }
/(e)
e
fxY
\2)
'
where
^ (e)
n(m + s)n(s+6)l2J ~U(s)U(n+s-e)\2)
'
Let us use the notation
,*<*> = |; logn(,) =
then the limit of/(e)/e
when
e is zero is
|f
IP
'
VII -J
COMPLEX THEORY.
Again
II
65
(p. 12),
(- n
+ e + s)
II (n
e - s 1) = ir cosec (n- e-s)ir

Sin67T
therefore
Lt ell (- n
+ e + s) =
e=o
Finally, then,
when n
is
(-)"
II (n
s])
a positive integer,
(ir^nwnW^
Comparing
is
this
I-^ (s) -^ (n+s) }-
with the formula 30
14)
(p.
we
I33
when
see that,
a positive integer,
= 2 Yn - (log 2 + ^ (0)) /],
ttT
or,
21og
{
with the notation of
p. 40,
and on substituting
for
its
34
value
as a definite integral,
rn = (log2- 7)J- + |Tn

=
2
2V-n(n-i) rg ~
7)
C S
Jo
^ C ^ ainm * ^
S
ir
+ 7T
sin
sin
(a;
<j>)
cos 2"
d<
Jo
er xsinh -t>cosh. m ^d<f)l.
-ir
135
When
a?
is
a very large positive quantity, the
on the right-hand side

Stokes's
method*
to
approximate value of
may be
neglected
the other two integrals,
Yn
given on
p.
last
integral
and then, by applying
we obtain the
40 above.
Jn
and T as
when n
is positive.
This course is practically adopted by Hankel, whose
memoir in the first volume of the Mathematische Annalen we
have been following in this chapter.
Hankel writes Tn for a
There would be a certain advantage in taking
the fundamental solutions of Bessel's equation in
all
cases
* Or, preferably, that of Lipschitz, explained later on, pp. 69
G. M.
71.
5
COMPLEX THEORY.
66
function
we may
for all values of
[VII.
moment, by
denote, for the
Yn
this is defined
n by the formula
\
J " 008B,r
Fn =2,rt~'
sin 27r
136
>nm
7re'
-T
costwt
(see 131 above).
Since, however,
adopted,
fined.
we
It
Neumann's notation
shall continue to use
may be remarked
that,
Tn
now being
is
generally
in the sense previously de-
when n
is
a positive integer,
Yn = TTTn =2{Yn -(log2-y)Jn
137
}.
Hitherto the real part of x has been supposed positive
now suppose that x isjijpure imaginary.

tion we put x = ti, it becomes
d?u
dt*
By
du
dt
',
ri
I
-(l
+ ) = 0.
proceeding as in Chap.
we
If in Bessel's equa-
will
II. it is easily
138
found that, when n
is
not an integer, the equation has two independent solutions / (t),
I^n (t), defined by

/()
= *-"J (*)= 2
2' i+2s n (s) n
o + s)
n+28
I-n(t)
n
i J-(it)
one of which vanishes when
When n
is
= -Z 2-+
(
= 0,
2s
139
Ii(s)n(-n + sV
while the other becomes
an integer we have the solution In as
there will be another solution obtained
by taking the
infinite.
before,
Yn (it). But it is found that both these functions (and

manner In and /_) become infinite when t = <x> and it
i-
portant for certain applications to
vanishes
when
is infinite.
the equation 123, by which
discover
is
im-
a solution which
expression on the left-hand side of that equation

n
in like
We shall effect this by returning to

Tn was originally defined. If in the
becomes.
i tr
and
real part of
{-
Un + iVn
),
we put x = ti,
it
COMPLEX THEORY.
VII.]
Un =
where
cos
(t
sinh
cosh 2
(/>)
67
d<h,
<f>
Jo
IT
Vn =
sinh
sin 0) cos 2"
(t
d</>
Jo
The
but
if
(t
sinh <)cosh 2n </> d(/>.
Un =
whether n is positive or negative

and it may be verified that
infinite
Un is
negative
1 is
is
Vn is
function
2n
sin
Jo
cos
finite,
sinh
(t
<)
cosh 2"
<f>
dd>
Jo
Now
a solution of Bessel's transformed equation 138.
this does not give us
positive
which
although
what we want, namely a solution when n
shall vanish
when
is infinite, it
suggests that
is
we
should try the function
r cos
*
which
is
obtained from
Un by
when t =
tion obviously vanishes

satisfies
(t
00
and
it is
it is
n.
This func-
easily verified that it
the solution
we
require.
be found convenient to write
reducing,
<f>)
changing n into
the equation 138, so that
It will
sinh
^cosh^
when n
is
(t
sinh 0)
an integer, to
-y .. 3
Then / and
r cos
2 îrf^
...
5 ...(8
d(f>
.-^/; -s*
Kn are solutions of 138
which are available
for all
and which are the most convenient solutions

to take as the fundamental ones, whether or not n is an integer.
positive values of n,
As
in Chapter II.
it
may be
proved that
jln+ In+i
r=h
--I
141
In
/;
= /,
52
COMPLEX THEORY.
[VII.
and that the functions

n are connected by relations of precisely
the same form. It is in order to preserve this analogy that it has
been thought desirable to modify the notation proposed in Basset's
Hydrodynamics, vol. 11. p. 19. The function here called
n (t) is
Kn
2 n times Mr. Basset's
By applying Stokes's method to the

we obtain the semiconvergent expansion
u ~ Aet
8T~ +
differential equation 138
2U8if
(4ra2
- 1) (4w2 - 9) (4ra - 25)

2
3!(8<)
t-i II +
e
l
+ Be-*
~+
"
^'-^ + (*"' -*)(*' -9)

2\(8ty
8*
(4w2
2
1) (4rc
- 9) (4w - 25)
2
:V.(8t)
which terminates, and gives an exact solution, when n is half an

odd integer. This may be used for approximate calculation when
t is
large.
When u = Kn
A somewhat troubleit is evident that A = 0.

some and not very satisfactory process, suggested by a formula in
Laurent's TraiU d 'Analyse, t. in. p. 255, leads to the conclusion
B = Vf 7r
that
Kn
cos nir, so that the semiconvergent expression for
is
Hr
^ = V27
COS
^- e"
f
t
1
l
The corresponding
n
~M
2irt
(4w2
-l)
J+
Sr
(4w2 -l)(4w2 -9)
2!(802
expression for /
8t
"}
I42
is
2!(8*)
')'
There does not appear to be much reason to doubt the
43
correct-
ness of these results, since the formula for / works out very
fairly,
even
for
comparatively small values of
t,
and the
formulae
are consistent, as they should be, with
In+1 Kn - InKn+ i = - cos

The formula 30
real part of i
-"
(p.
im.
14) leads us to conclude that
Yn (it), Kn (t)
must
agree,
up
if
Rn (0 is the
to a numerical factor,
with
i2 w
144
(0-(log2- 7 )/.();
COMPLEX THEORY-
VII.]
by Stokes's
this is confirmed
6'9
Gamh. Phil. Trans.,
investigation,
vol.
ix. p. [38].
It
would be easy to multiply these applications of the complex

we will conclude this chapter by giving
theory to any extent;
an alternative proof of the semiconvergent expansion of Jn obtained in Chap. IV. above.

The method is that of Lipschitz
(Crelle LVL, p. 189), who works out in detail the case when n = 0,
and indicates how the general case may be

Let us suppose that n
and the
treated.
real part of
x are both
Consider the integral
positive.
= Jexzi (1 - z )n~
2
dz
taken, in the positive direction, along the contour of the rect-
angle whose vertices are at the points corresponding to the

quantities
Then the
total value of the integral is zero,
it
as the
we
+1,
1,
sum
+ hi, 1+hi,
where h
and positive.
and by expressing
is real
of four parts arising from the sides of the rectangle,
obtain the relation
0=i V* (1 _
'
-i
where
t is,
The
n_1 dt
x M+t)i {1
{
V(
Jo
* )
1+if)i
+ ij V' ^"*
1
{1
- (1 + ity]"-* dt
_ (hi + t y}n-i dt
{1
- (- 1+ ity} n~
dt,
throughout, a real variable.
first
integral
integral vanishes
is
Gnx'nJn and when A=oo,

,
the third
hence
Cn ar*Jn (*) = ie~ix
- y*
j {lit +
Jo
r
3 )""*
- dt
(-
lit
~ xt
+ i2)"_i e
dt
45
Jo
IT
The argument
vanishes
of 2it
+ t must
2
be taken to be ^ wnen
hence
(2it
+ t )n~i = 2"-* e*
2
-11
(2
/
"
*-*
1
[
\"- J
2i)
'
COMPLEX THEORY.
*70
where
n ~ h is real,
and the argument of
(2n + 1) w
^-4
+ 5-.
Applying a similar transformation to (
2it
+ e-^
Now
put xt = f
e^'r
e-**
when t = 0.
vanishes
+ i2)"-
and putting
*,
= *.
we have
/
[VII.
ft
+J
\
,l
~J
d
n-* (l
dt.
2"-)
then
2-"+
By
C^ JB () = e**
J
Maclaurin's theorem,
we may
'
+ J-)*
e-i -* (l
dg
write
+ lX~ = + ( n -&% + ( - iK^zJ).?

h
(1
2iW
"
2! (2m) 2
2ta>
"
" "
t
(s-1)!
\2isc)
\2ix) \
2^
^-
may
be ex-
panded in a similar way. Substitute these expressions

preceding equation, and make use of the formula
in the
where #
is
some proper
fraction,
f QO
e -|-H-
df
Jo
then,
if
we
"*"
"
and
= n (n - i + s)
further observe that
Cn =2Virn(n-i),
we
obtain
VT- /"
(a;)
= C0S
^ + "2^
Sm ^~
cos
21(20?
^+-
(to s terms)
+ R,
H&
COMPLEX THEORY.
VII.]
VI
where
2U(n-^R = (n -^^-^ "(
Each
of the integrals on the right
hand is increased in absolute
the quantities under the integral sign are replaced by
value
if
their
moduli
moreover,
further increased
when
> n ,
the values are
still
by replacing
-*(+) -*('-
each by unity.
A fortiori,
if
we put x = a,
|
VW
IIsII(?i-)
'
fa-)(rc-f)-(W -s +
The expression on the
in the (s
+ l)th
cos
term of the
(2a) s
'
right
value of the coefficient of cos
n(n + -j)
j)
nn(-4)
Jo
yfr
hand
is
or sin
precisely the absolute
$,
as the case
may
be,
series
^ + L_^ sln ^_V__^__^ cos t+ -
so that the semiconvergence of this expansion,
representation of
a/
-5-
Jn (x)
and
its
are fully established.
approximate
CHAPTER
VIII.
DEFINITE INTEGRALS INVOLVING BESSEL FUNCTIONS.
Many
definite integrals involving Bessel functions
have been
evaluated by different mathematicians, more especially by Weber,

Sonine, Hankel, and
shall give
In the present chapter we
Gegenbauer.
a selection of these integrals, arranged in as natural an
order as the circumstances seem to admit.
By
the formula 42
(p.
18) or 90
1
J (bx)
=7T
Suppose that b
Others
will
be found
end of the book.
in the examples at the
is real,
(p.
38)
we have
[*
cos (bx cos
<t>) d(f>.
Jo
and
let
a be a real positive quantity
then
I
e~ ax
(bx)
dx = -
fJo
Jo
e_ax cos (bx cos
dxl
cf>) d<f>.
Jo
upon a and b, we may change

the order of integration on the right hand, and make use of the
Under the
conditions imposed
formula
e
-mx cog nx dx
m
.
Jo
thus
e- aa
(bx)dx
adcj>
=-
tJo
Jo
a-
+ cos $
2
1
"
Jo?
This result will
still
vided that the integral
be true
is
for
+6
complex values of a and
Now
convergent.
a = al
+ a i,
3
= b + bî,
x
if
b pro-
DEFINITE INTEGRALS.
VIII.]
the expression e~ ax
(boa),
when x
e -aix cos Jj
73
very large, behaves like
is
fog.
P(x),
*Jx
where
is
P (x) is
convergent
a trigonometrical function of x
i>
and under
we
this condition
\b 2
(bx)dx
Va2 + 6 2
b is real
1
= -7===,
Va + 6
147
being taken which reduces to a when
and
positive,
and as a
have
Jo
When
hence the integral
shall still
r e- aa J
that value of
if
6=0.
we may put a = 0, and thus obtain

dx
(bx)
= j-
148
special case
Jo (*)
dx
= 1.
149
A
.'O
I
Again
let b
be real and
being real and positive
If b 2
> a?
positive,
and put
ai instead of a,
thus
2
the positive value of V&
- a must
2
be taken;
if
62
<
a?
we must put
V&3 - a? = +
We
Va2 -
b\
thus obtain Weber's results
cos
ax dx =
62
I
(6a;)
sin
>a
2
,
150
axdx =
Jo
Jo (bx) cos a#
d* =
Jo
I
Jo
a2 >
si
a# cfc
Jo (&#) sin
=
4a? -b\
62
ISI

74
DEFINITE INTEGRALS.
[VIII.
a2 = 62 the integrals become divergent; the reason

that they ultimately behave like
When
dx
/""
....
cos (a
*
instead of like
- x) dx
-
V*
Jo
The formula 147 with which we started

(Grelle lvi.) another due to the same author
;
rx~^ J
where
is
due
to Lipschitz
is
n (.-i)n(-^)ng-i)
(ax)
dx =
sin rmr, 152
m is a positive proper fraction.
To prove
this
-1
a;"
we
observe that
Jo (ax) dx
dx
*m_1 cos (ax cos d>)
and on changing the order of integration

211
(.-l)c -3iram
11
(m
drf>,
Jo
""Jo
Jo
this
becomes
fi
m
J
JO cos
(j>
mi
1) cos
-=2" _1 _m+l
2
which
f
Jo
(\-t)-^dt
7ra"
Now
n(-^)n(5-l)-^cosec?f
and
II (
(see p. 12)
11
is
- g J = /yV
thus the expression reduces to
(m- l^cos-.- VtIII
m g
wwr
+ l^Tr/m ,\
II I -llsin
.
7TCI"
2".
DEFINITE INTEGRALS.
VIII.]
that
is,
75
to
'-)"(-'-r)(|-').sin mir,
27r
i
s
am
the value given above.
The result may be written

make use of the formula
in various other forms
thus
if
we
1 +m
r
n(m-i
a i) n(ii ^)cos^ = 2-v rn(^-i)
it
becomes
x-1 J (ax) dx =
sm
3=
Jo
- -ad 1
"'"(I- )
153
n(-=)
reducing, as might have been expected, to lja
We
will
obtained by
when
m = 1.
next consider a group of integrals which have been
Weber
(Crelle lxix.)
by means of a very ingenious
analysis.
Let
Fbea
function which
of space,
and which
one- valued, finite

direction,
and continuous,
throughout the whole
also satisfies the equation
F+m F=0.
Using polar coordinates

equation
is
any
as well as its space-flux in
r,
6,
<f>,
and putting cos0
is
r+i r+rr
Let
Vd/j,
d<f>
then, observing that
/j,,
this
'
DEFINITE INTEGRALS.
76
because
-=-r
and ^ are one-valued, we have

0/JL
0(f>
r2 dr \
or,
which
is
[vill.
or)
the same thing,

32
~ (ra>) + mVsi =
whence
= (A sin mr + B cos mr),
co
A and B are independent

must have 5 = 0, and
where
of
is
ca
the value of
clear that, if
it is
of
sin
w when
r=
is
If
r.
= a>
o>
where
eo is
finite
when r=0 we
mr
,
mr
Now
0.
the value of
from the
definition
V when r = 0,
r+l r+w
and therefore
Now
47rF
w=
Vo
dfid(j)
sin
consider
F as a function
= 4>ttV
mr
1
r
of rectangular coordinates a,
and put
moreover
let
us write
J ao J 00 J 00
Then
if
we
introduce polar coordinates by writing
a x = r sin 6 cos
we have
y = r sin d sin $,
Z = r COS 0,
JO
where
<J>'
is
<,
COS
r^-^drf
fJU,
Vcfyicty,
J -1 J -IT
the transformed expression for
4>.
54
b, c
DEFINITE INTEGRALS.
VIII.
By
the preceding theorem this

A
77
is
,'00
H = <& re"^"- sin mr dr

m "Jo
t^
m
-
= 4
 (a,
2
e-^!(-^) + (M/>+
6, c)
=-*)'}
rf
a d& dc
J co
=
Suppose now that
$ is
independent of
j
J
00
e -pHc-z)* dc
e -pn> dt
J 00
4p<J)(a;,
156
y,5).
then since
= ^jp
we have
r+oo
+ <
* (a, 6) e-^ {(-)S+(6-3/)=} rfa db = - e

2
J 00 J
00
the equation satisfied by
<E>
OT 2
*p2 3> (, y),
=J
being
(mr),
=a +6
x=y =
where
r*
and suppose that
2
,
= r cos
#,
= r sin 0,
then the formula becomes, after integrating with respect to

f
re-***!,, (mr)
dr = ^- e
generally,
<
we
57
we may put
In particular
More
0,
_i2?
*p\
158
*P"
by putting
= / (mr-) (.4
# = p cos yS,
cos
+ B sin 0),
y = p sin /3,
nd
obtain
6 -pV f
r-e-p
* Jn ( mr ) dr
J0
I
J
"e^V-** ^'^
cos
n0 +
B sin nd) d8
-w
e _^1 Jn (mp) (A cos nft + Bsmn/3).

7T
4/p*
159
I
DEFINITE INTEGRALS.
78
In
this formula
[VIII.
put
A = l, B =
/3
i,
then the integral with respect to 6
= i7r;
is
r+n
g2p 2p- sin e+n&i
^Q
J IT
= 21
cos (2ip 1pr sin 6
nd) d6
Jo
= 2irinIn (2p*pr).
The formula thus becomes,
both sides by
after substitution,
division of
Cre-v** Jn (mr) In (2fpr) dr
^P e~^+pV J (m
more symmetrically, putting \
for
Jo
or,
and
2mne^ ^,
m, and
fi
for
2p2p,
j'rtr^ Jn (Xr) In (fir) dr = A. e~*& Jn

or again, changing
into
/x.
which does not
i/i,
P ),
g^)
160
affect the conver-
gence of the integral,
re~^ Jn (Xr) Jn {pr) dr =ê~^~
Jo
By making
(i
we
infinitesimal,
î^jn(X
(g
161
J.
obtain the additional result
r)<fr
= -^L_ e -S.
162
In
all
these formulae the real part of
p must
2
be positive
in
order to secure the convergence of the integrals.
After this singularly beautiful analysis,
Weber
evaluate the integral
r^-n
-1
Jn {Xr)dr
Jo
which
is
convergent so long as
<
It
is
known
< n + .
that
1
r*n+*- q
If"
n(n-hq)Jo
rm-J?
p-rtc fj
'
proceeds to
DEFINITE INTEGRALS.
VIII.]
79
therefore
/QO
Now
it
if
xn~î rn+1
dr
r-r
$q)Jo
Jo
we do
is
(I. c. p. 230) that the value of the

not affected by changing the order of integra-
this,
and make use of
rg-"-1 Jn (\r)
162,
we
2 +i
n(n-ig).Jo
2+i
n (w - g)
= n + 1, we
^x^^dx
e-1^
2?
2--i >
putting q
obtain
dr=
A,"
By
Jn Ckr) e- * dx.
may be shown
double integral
tion
pen
*oo
-j
ri-n-iJn (\r)dr= Tf
.ii(n
Jo
wi?- 1
dy
n(k-i)a
n(-i?)
obtain
1
Jn (at) dr = and by putting q
we
= n,
164
A.
/,
which
is
permissible so long as n
is
not zero,
find
J
1
-^^-dr = -.
165
/;
Weber's results have been independently confirmed and general-
by Gegenbauer and Sonine, especially by the latter, whose

elaborate memoir {Math. Ann. xvi.) contains a large number of
very elegant and remarkable formulae. Some of these may be
verified by special artifices adapted to the particular cases considered; but to do this would convey no idea of the author's
general method of procedure, which is based upon the theory of
complex integration. The formulas obtained are connected by a
very close chain of deduction, so that an intelligible account of
them would involve the reproduction of a great part of the memoir,
and for this we have not space. We are therefore reluctantly
compelled to content ourselves with referring the reader to M.
ised
Sonine's original memoir.
We will
now
to that which
is
consider a remarkable formula which
generally
be stated as follows
known
is
as Fourier's Theorem.
analogous
It
may
'
DEFINITE INTEGRALS.
80
If
</>
(r) is
a function which
is finite
[VIII.
and continuous for
all
values of r between ike limits p and q (p and q being real and

positive), and ifnis any real positive integer, then
(Xp)Jn CKr)dP
f dxf\<}>(p)J
{^J^/or <
l66
very instructive, although perhaps not altogether rigorous,

of discussing this integral is given in Basset's Hydrody-
method
namics,
vol.
II.
15
p.
we
begin by giving this analysis in a
will
slightly modified form.
Taking cylindrical coordinates p, 0, z let us suppose that we

a thin material lamina bounded by the
have in the plane z =
circles p = q and p=p; suppose, moreover, that each element
attracts according to the law of gravitation, and that the surface
density at any point
is
expressed by the formula
= <p(p) cos nd.
<r
Then the
potential of the lamina at

[p f
9+M
+r +p
2
{z
n&dp d&
- 2rp cos (& - 0)}*
0'-0 = V
Put
r
+ p 2rp cos v = R
2
(r, 0, z) is
(p) cos
p<\>
=/:/:
any point
2
;
then
2w
P&
J, Jo
,a ?
Now when
*)
cos
n & cos nr ~
>
(z 2
'
is
(*"
nr
i\
dp dv
positive
'-"
z=
n@ s
P^(p)cosnv dpdv
(dm=fo
(see p. 72)
s* n
+R )*
MXE)dx
hence the potential on the positive side of the plane
is
F,
/"2t
rp
= cos 710
J q
By Neumann's
m
e~ Xz p<f> (p) cos nn
dn\
dp
(XR) dX.
formula
(p.
27)
we have
GO
J (XR)
J",,
(Xr)
(Xp)
+ 2 2 J, (Xr) J
(Xp) cos sv,
67
VII I-]
DEFINITE INTEGRALS.
and hence, if we assume
that
81
remains the same if we change the
order of integration,
F,
= cos n0\
dk\ dp
J q
= 2tt cos n8
By making
e-^ p<j>
(p) cos nrj
(\R)
dr>
.'
dX J* dpe'^ p<f>
(p)
Jn (\r) Jn (\p).
a similar assumption we find that the potential on
the negative side of the plane z
7 = 2tt cos ndf d\f

J
J
is
dpe^ p(p)Jn (\r)Jn (\p).
p and
the value of the expression on
q,
= 4nr<j> (r) cos n6
in all other cases it is zero, excepb possibly
we admit the
if
168
is
4tto-
Thus,
can be found by
when z = 0, we have
\dz~~dz) z the
and
when r q
validity of the process
been deduced from 167, the proposition immediately
With regard
to the case
when r = q
or
r=p.
or
by which 168 has
it
follows.
may
be observed
that just as the equation
is
really connected with the fact that the attraction of a uniform
circular disc of surface density
an amount
4nr<r
as
we
a-
is
altered
(algebraically)
one side to a point close to the centre on the other, so we

infer that
when r = q
or
by
pass from a point close to the centre on
may
p
\ dz
because the expression on the
dz
)ss=o
left
is
ultimately the difference
in the attraction, normal to its plane, of a uniform semicircle
as
we
pass from a point close to the middle point of
G. M.
its
bounding
6
82
DEFINITE INTEGRALS.
diameter on one side of
Thus we
other*.
[nil
to the corresponding point on the
it
are led to the conclusion that
io
\P
dXj
Jo
(p)
<f>
r=q
if
This
what the analogy
is
Jn (Xp) Jn (Xr) dp = %<p (r)l

or p.
of the ordinary Fourier series would
lead us to expect.
The assumptions involved
in the foregoing analysis are not
very easy to justify, especially the
one: we will therefore
first
give the outline of a more satisfactory demonstration, derived,
memoirs of Hankel (Math. Ann. vin. 471)
in substance, from the
and du Bois-Reymond
(Grelle lxix. 82).

rh
u=
rP
dX
Jr
Jo
XpJn (Xr) Jn (Xp) dp
where n is a positive integer, and

with p > r > 0.
If
is
change in u
increased by a small
r,
p,
h are positive
amount
quantities,
dp, the corresponding
is
rh
du =
dX {Xp Jn (Xr) Jn (Xp)
dp}
Jo
= pdp
rh
XJn (rX) Jn (pX) dX

Jo
= î{rMph)J:(rh) - P J'n (ph)Jn (rh)}

by
107.
Now
suppose that h
is
very large, and
(2n
+ l)ir
let
4
then we
may put
Jn (ph)
= /\/^p cos ( - ph),
Jn(ph)-\/h sin (a
* This
curvature
argument applies
is
continuous.
to
~ph),
any disc at a point on the circumference where
the
DEFINITE INTEGRALS.
VIII.]
and similarly
Jn (rh),
for
J^ (rh)
83
therefore ultimately
when h
is
very large,
du
"
dp
_p
irr
=' p*dp
j=
(sin
(p
p sin (a ph) cos (a rh)}

sin (2a (p + r) h)\
r)h
pr
f sin
(p
p+r
- r) h
(
P _r
7T7-M
Hence the value
i}
+7
'
'
infinite is given
is
- r)h
û=u\r sin(p
P~r P
7,
_y +1 cos (p + r)h
u when h
of
ph) sin (a - rh)
{r cos (a
by
dp
= oo (Jr
j.,
p cos (p
+ r)h
In virtue of two theorems due to Dirichlet

T
psin(jo
h = xJ
r)h
pr
p -^smhx
.,,
'
= r*.
2
pcos()o
vr
Ti
Lt
and,
+ r)A p*
,
'
,
= TLt
dp
.
p +r cos/w;
(*
- r)*,,eta
= 0;
7rr*I7"= Jot*
consequently
That
and therefore Z7= J.
d\
JO
In the same way

r
it
to say
Xp Jn (Xr) Jn (Xp) dp
I
.'
is
may be shown
It has
It
may
< q<r
if
(Xp) dp
= .
J q
been shown by Hankel that
even when q
that
rr
dX\ Xp Jn (Xr) Jn
JO
= J.
this last formula holds
= 0.
good
be thought paradoxical that the value of the integral
dXJ Xp
Jn (Xr) Jn (Xp) dp,
62
DEFINITE INTEGRALS.
84
which vanishes when
The
limit p.
p = r,
[VIII.
should be independent of the upper
statement
fact is that this
true only on the sup-
is
p and r are separated by a finite interval, however

small that may be. Thus it may be verified that if the positive
position that
proper fraction e be so determined that
sin
/,
i
Lt
then
fr
dXl
+T
h
XpJn (Xr)Jn (Xp)dp-. h
Jr
h=a>Jo
ax = \tt
whole integral
so that the value of the
obtained by confining
is
the variation of p to the infinitesimal range
(r,
+ eirjh).
It should be observed that the expression

rr+e
rh
Xp Jn (Xr)
dX\
Jo
Jn (Xp) dp
J r
when h becomes
infinite and e infinitesimal has no particular

meaning unless a relation is assigned connecting the ways in
which h and e respectively become infinite and infinitesimal.
It
may
be worth noticing that these results are analogous
to
the reduction of the effective portion of a plane wave of light to

that of a part of the
first
For convenience,
let
Huygens
zone.
us write
f Xp Jn (Xr) Jn (Xp) dX =
(p, h);
then
it
follows from
what has just been proved that

(1
Lt
/*
<&(p,h)dp
=h
h = aoJ q
(.0
In the
first
iiq<r<p,
r = q or r = p,
r<q
or
r>p.
case of this proposition the inequalities
q<r<p
must be understood to mean that i

q and p r are finite positive
quantities and in like manner with regard to the other inequali;
ties
which occur in
Now
this connexion.
suppose that f(p) is any function of p which throughout

(q, p) remains finite and continuous and is always
the interval
increasing or always decreasing as p goes from q to p.
lemma due
to
du Bois-Reymond
Then by
85
DEFINITE INTEGRALS.
VIII.]
f/OO * (P.
dp =f(q)
h)
J q
f*
Jq
(p,
h) dp
P
+ {f(p)-f(q)}l
J
(p,h)d<l>,
p.
where
some quantity between q and
p. is
p.
Suppose that r is outside the interval (q, p); then, when h

infinite, both the integrals on the right hand vanish, and
becomes
therefore
Lt
ft=oo
fp
I
.'
f(p)(p,h)dp
[r>p
or
r<q].
169
Next suppose r = q
then by the same lemma
JVOO *<! *) <*P =/(") * (p, A) dp +
{/(p) -/(r)}
* (p, h) dp
and therefore
Lt (7(p)
* (p,
h)
dp
= i/(r) + Jf
/(p) -/(r)},
A=oo J r
where
is
certainly finite, because
it is
separated from r by a finite interval, and
Now
by
169, if
t is
zero if
it is
p, is
when p
ultimately
= r.
a positive quantity greater than p,

t
Lt
and hence, by adding the

Lt ff(p)
<E>
(p,
f{p)^{p,h)dP =
last
h)
two formulae,
dp
Since the expression on the
= i/(r) + M {/(p) -f(r)}.

left
same must be true of that on the

be
zero.
0,
hand is independent of
and consequently
p, the
M must
right,
Therefore
P
Lt
provided r
It
{p,h)dp = \f(r)
that r and
may be proved
in the
are separated by a finite interval.
same way that
Lt r/(p)<P(p,h)dp
7t=oo J q
when
170
exceeds q by a finite quantity.
= ^f(r)
170'
DEFINITE INTEGRALS.
86
Hence,
finally,
/(r)if ? <r<p,
Lt
[VIII.
dp
Xpf(p)Jn (\r)Jn (Xp)dX=\\f(r)
r=qorr=p,
r < q or r >^,
171
and, as in the
other case,
we may,
if
we
like,
change the order of
integration.
been supposed that f(p) continually increases or conbut it may be

(q, p)
shown, as was done by Dirichlet in the analogous case of the
It has
tinually decreases throughout the interval
Fourier integrals, that the formula

is
removed.
The
function f(p)
is still
valid if this restriction
may even
present- any finite
number of isolated discontinuities in the interval, and

become infinite for isolated values of p provided that
it
may
Jq
I"
'9
is finite.
It is
validity
upon the formula

of the
Chap. VI.
7 1 that
Hankel bases
his proof of the
Fourier-Bessel Expansions briefly discussed in
CHAPTER
IX.
THE RELATION OF THE BESSEL FUNCTIONS TO SPHERICAL

HARMONICS.
There
is
a remarkable connexion between the Bessel functions
and spherical harmonics which appears to have been first disp. 134, and Math. Ann. v.
His results have since been developed by
covered by Mehler (Grelle lxviii. (1868),

(1872), pp. 135, 141).
Heine, Hobson and others.

If,
it is
as usual,
known
Pn (x) denotes
the zonal harmonic of the nth order,
that
Pn (cos 8) = cos"
jl
n ~ V)
- n^
tan
n(zl )^-8)(-_) tm, g
= xjn, and suppose that n becomes

Put
mains finite then since
infinite while
Lt
cos
noo
and
lit
ft
it follows
-=
n
X
n tan -
= QO
x,
"*
that
LtP^cos-j
This result
may
also
irP n (cos 0)
-!--+_-
...
be obtained from the formula
|{cos
Jo
+ ism
cos
_...|.
n
</>}
d;
re-
SPHERICAL HARMONICS.
88
the limit,
for
when n
cos
x
-+
n
x
n
sin - cos
[iX.
of
is infinite,
is
rf>
the same as that of

ix cos
and
this is
e*"
008
<j>\
+ ~ir~ )
'
so that
jrhtPn (cos^)
re
tocos *
cty
= irj
{x),
and hence
Lt Pn ( cos - j
= Jo (x)
as before.
Another known theorem (Heine, Kugelfunctionen
I.,
p.
165)
is
that
Q(cosh0)
and
if
-J^
(coshfl+sinhflcosh^+i-
we put 6 = xjn, and proceed
as before,
we
find
+M
Lt Qn (cosh
|)
e-* c sh * cty
| f
=K
{x)
173
(see p. 67).
Changing x into
*'#,
we
obtain
UPn (cosh 3 = /(),
174
Lt Q (cos -) =
(log
nj
2-y)J
(x)
-Y
(x)
(Kugelf.i.,^. 184,245*).
Thus every theorem
in zonal harmonics
may be
expected to
yield a corresponding theorem in Bessel functions of zero order
may, of course, happen in particular cases that the resulting

theorem is a mere identity, or presents itself in an indeterminate
form which has to be evaluated.
it
There
is
a misprint on
p.
245
in line 8 read
instead of - C.
IX.]
89
Heine has shown that the Bessel functions of higher orders
may be
regarded as limiting cases of the associated functions
which he denotes by P$ (x).
we
shall write
Then
For the sake of a consistent notation
P* (x) instead
(Kugelf.
I.
p.
of Heine's PI
{).
207)
7rP?t (cos0)
.. H(n
+ s)' TL(n
v
= 2"
a v
s)
w
f
II (2)
(cos
K
+ 1 sin
,.
n
cos <b)
T dd>,
r
T/ cos s6
,
and therefore
TrLtP^fcos^)
fg.
=00
n (. + ,) n (. - .)
n(2)
Now when
is
>C03 , cos
jj
very large
we have asymptotically
n (t) = J2w e-' i'+i,

and therefore
to the
same degree
of approximation,
when n
is
large,
n (n - s) n (n + s) _ 2 n
lire- (n
~~ IT (2m)
s) n+s+i
-2
2?r 2-"+i e "
(n - s)"-+i
nm+i
hence we infer that
n (2)
.=. U/mr
=- V
=e
n/
.e~*=l.
Consequently
Lt i-?L P (cos - H
nj)
= [Jn-ir
\
=-
008 *
cos
s<f> d<j>
irJo
= tV,(),
which
is
Heine's formula
In the same way

p.
it
(I.
c. I.,
175
p. 232).
may be
inferred from the formula
(I.e.,
223)
n./nhtoV t cosn > -
n(2 + l)
2n (r? + s) n (n - s) J
f (cosh
cosh^cfo
sinh
cosh ^)"+
90
[iX.
that
J1 ^
it {&*
c sh
-ly*** c sh s + *+
1)}
= (-)Ks (x).
By changing x
and
(log 2
In Chap. VII.
functions
Is (x)
obtain similar formulae for
- 7) Js (x) - Ys (x).
has been supposed that the argument of the
it
Is (x),
we may
into ix
176
(x) is a real quantity; in fact the functions
were expressly introduced to meet the difficulties arising from

the values of Js (x), Ys (x) when a; is a pure imaginary.
In the Kugelfunctionen Heine practically defines his
(x)
by
the formula
(x
Oi)
= (- i)
if,
cosh
= a (sin a + i cos a),
where x + Oi
hence
&IJOsh *
to avoid confusion,
and use the
symbol
latter
= (-yKs (-x-0i),\
%ir<a.<\ir;
we write Ht s
for
function previously denoted by
K {x) = (-)
scf> d<f)
it,
(aO for Heine's
77
K (x)
s
an extended definition of the
we may
write
00811 *
cosh sd>
f e-*
d<f>
Jo
178
with
By
= a (cos a + i sin a), \ir< a< ^w,
putting
K {-x) = (-yK {x)

to K (x) in
cases except when
we give a meaning
and if we
imaginary
K
we
179
all
define
(ti),
where
t is
real,
(ti)
= \ [Ks {ti + 0) + Ks (ti - 0)},
(ti)
= i {(log 2 - 7 ) Js (t) - Ys (t)\.
a;
is
a pure
by the formula
find that
K
From our
present point of view, the most proper course
take as the standard solutions of Bessel's equation, not
80
is to
Jn and Yn
but Jn and i^Kn (ix) as above defined. The function i-^Kîx)

has in fact already appeared in Chap. VII.; if, when iftif maae*
IX.]
one-valued as explained above, we
91
Qn (x),
call it
Yn (x)
then
also
becomes a one-valued function denned by
Yn (x)=(log2- y )Jn (x)-Gn (x)
181
with a discontinuity along the axis of real quantities expressed by

the formula
Yn
{t
Ot)
-7n (t- Ot) = -iriJn
(t).
For any point on the axis of real quantities
Yn (t) = (log 2 - 7 Jn (t) - On

)
The reader
Kn
Heine's
it
Gn (x)
will observe that, in fact,
We
(x).
Kn (x) in
have employed
conformity with the usage
now
(t).
is
a different sense in
generally current in England
must be admitted that Heine's notation
82
identical with
but
on the whole, the
is,
preferable one.
To Heine is also due the following excellent

methods of this chapter.
It is
known
with a proper determination of the signs of
that,
the radicals (see Kugelf.
Pn {xy - Vai^T. Vy
I.
p. 49),
^Tcos $) = 2 (-yas P (x)P*n (y) cos sd>,

s
where the summation
refers to
_
a ~
b,
and
">]>""
1
3.5 (2n-l)j.
(n-s)l(n + s)l
suppose that
x=
where
s,
183
{1.3. 5...(2-1)} 3
Now
illustration of the
b
cos -
y = cos -
c are finite real quantities,
V'a? \=i sin -
Then
and that n
\A/ 2
very large.
is
i = i sin -
and
xu
*
*Ja?
1 VwJ
2
1 cos
<f>
= cos -
cos
= 1 b
= cos an
-+
n
sin
26c cos
2n2
sin - cos
^+ c +h
d>
...
<f>
92
to the second order,
[IX.
if
a=
/6
26c cos
<\>
+ &.
Making n increase indefinitely, and employing

Jo
JO? <v \
26c cos
Proceeding as on
_ T
p. 89,
(2w !)
c )
=2A J
(6)
(c) cos
s<f>,
iwr
^'".t.
^ -
we have
=U tl.3.5...(2-l)} ^
where
<f>
183, 175,
7Z7T
i*. 2- ( iy
fl.3.5...(2n-l)j* rr
2""
( + )! (-)!
we
find
2
/2^"e- 2 't (2w)2"+*} W7r
_
- TM
{N
_
~
+*}
2 {Vê-
and similarly
(2n!ywr
I A -T,t
* *
2*(n!)(+*)!(n-)!
_..
'
thus finally we arrive at Neumann's formula
(V6 2 - 26c cos
+ c ) = Jo (6) J, (c) + 22 J
2
</>
(6)
(c) cos
s<j>.
84
85
86
In the same way, from the expansion

(
+ i) Q. {*</ - </(?- l)(y-l) cos <} = I' i (y) Q (*) coss<^
(Kugelf.
(Jb*
I.
333) we derive the formula
p.
- 26c
cos
+c) =
2
</>
& (6)
(c)
+ 22 G
(6)
(c) cos
s<f>,
in which
it is
supposed that
By combining
F (Jb*
- 26c cos
186, 184,
6,
are real and that 6
and 181 we
+c)=7
2
(6)
J (c)
>
c.
infer that
+ 22
(6)
(c) cos s<,
187
a result originally given by
Neumann
(Bessel'sche
Functionen,
p. 65).
We
shall not proceed any further with the analytical part of

the theory; for the extension of spherical harmonics and Bessel
IX.]
functions to a ^-dimensional geometry the reader
93
may
consult
Heine (Kugelf. I. pp. 449 479) and Hobson {Proc. L. M. S. xxu.

p. 431, and xxv. p. 49), while for the solution of ordinary differential equations by means of Bessel functions he may be referred
to Lommel's treatise, and the papers by the same author in the
Maihematische Annalen. (On Riccati's equation, in particular, see
Glaisher in Phil. Trans. 1881, and Greenhill, Quart. Jour, of Math.
vol. XVI.)
CHAPTER
X.
VIBRATIONS OF MEMBRANES.
One
of the simplest applications of the Bessel functions occurs
in the theory of the transverse vibrations
By
membrane.
the term membrane
we
of a plane circular
understand a
shall
thin,
perfectly flexible, material lamina, of uniform density throughout
and we
shall suppose that it is
tension
by means
maintained in a state of uniform
of suitable constraints applied at one or more
closed boundaries, all situated in the
membrane
librium,
is
same
When
plane.
and then
left to itself, it will
execute small
oscillations,
the nature of which
we
assumptions made
the purpose of simplifying the analysis.
We
for
shall proceed to consider,
under certain
shall attend only to the transverse vibrations,
and assume
that the tension remains unaltered during the motion

if
its
represents the plane which contains the
undisturbed position, and

z
defines
the form of the
supposed that
may be
the
slightly displaced from its position of stable equi-
d(f>/dai
and
moreover
membrane
in
if
= (f>(x,
y)
membrane
d<j>jdy
at any instant,
will
it
be
are so small that their squares
neglected.
Let or be the mass of the

Tds be the tension across a
anywhere upon the membrane;
of area, which for simplicity we
let
curvature.
Then
if
membrane per
unit of area, and
straight line of length ds drawn
moreover
may
let
dS be an
elemeat
suppose bounded by
lines of
rlt r2 are the principal radii of curvature, the
applied force on the element
is
\r,
rj
X.]
and
of action
its line
along the normal to the element.
is
clearness, suppose that the
direction of the axis of z
95
element
then the equation of motion
For
concave to the positive
is
is
where yfr is the small angle which the inward-drawn normal makes
with the axis of z.
Now, neglecting squares

1_
and
of small quantities,
_d*z_
cos
d*z_
=1
yjr
hence the equation of motion becomes
3% _
td>z_
W~~
\dx*
with
av
dtf
fl-
it
remains to find a solution
of
sufficiently general to
and boundary conditions
satisfy the
initial
and
dz/dt
may have
z=
0, for all
values of
prescribed values
t,
these
when
= 0,
are that z
and that
at points on the fixed boundaries of
the membrane.
By changing from rectangular to

I may be transformed into
cylindrical coordinates the
equation
ldz
d*z_ ^(d*z
&z\
Now suppose that the membrane is circular, and bounded

by the. circle r = a; then we have to find a solution of 3 so as
to satisfy the initial conditions, and such that z = 0, when r = a,
for all values of
t.
Assume
where u
is
independent of
= u cos pt,
then putting
P.=K
>
u has
to satisfy the equation
3r2
r dr
r 2 d(P
96
and
if
we
further assume that
u=v
where v
is
1 dv
d?v
dr2
a positive integer
From
r dr
(*-$v-o.
be sufficient for our present purpose to suppose that n

this being so, the solution of 8 is
It will
cos n0,
a function of r only, this will be a solution provided
that
is
[X.
= AJn (icr) + BYn (Kr).
the conditions of the problem v must be finite when
hence
5 = 0, and
z
we have a
solution of 3 in the form
= AJn (icr) cos n0 cos pt

= A Jn (r) cos nd cos kcL
In order that the boundary condition
may be
satisfied,
we
must have
Jn (a) = 0,
10
and this is a transcendental equation to find

proved in Chap. V. that this equation has an
of real roots
k.
It has been
number
infinite
k2 ks etc.; to each of these corresponds a normal

,
vibration of the type
9.
The
initial conditions
which result
in
this particular type of vibration and no others are that when

t
= 0,
z
By
= AJn (kv) cos n6,
n the values
assigning to
0, 1, 2, etc.
value of n the associated quantities
Jn (i) = 0, we
z
k%'\
k^, k \
are enabled to construct the
etc.
derived from
more general
+ Bm sin nd cos /4 ct
+ Gm cos n6 sin 4 ct + Bns sin nd sin 4 ct) Jn (4
A m Bm Gns Dm denote arbitrary constants.
= 2 (A m cos n6 cos 4
n)
solution
M)
ct
n)
m)
where
and taking with each
K|
r)>
' :
If the initial configuration
is
defined by
*=/(r,
0)
we must have
f(r, 0)
= 2, (A ns cos
n0
+ Bns sin n0) / (<,*" r),
12
X.]
and whenever f(r,
admits of an expansion of this form the
ff)
A m Bm
coefficients
are determined as in Chap. V. (pp. 56, 57)
in the form of definite integrals.
venience, instead of
writing k s
fact,
for con-
1ls
In
B)
*ej
2 ""
A =
97
Tra-Jn{Ks a)
dti
n8Jn ( Ksr) rdr,
f(r, 8) cos
Jo
J o
13
S =
Since
so that
dS
TrT/ a)\
ira^J,,(K
s
Jn (ic a) = 0,
>
Jo
J o
it
f ( r d) sin n(>Jn ( K*r) rdr
follows from the formula 19
we may put Jn2_i (g<x)
for
J (k o)
(p.
13) that
in the expressions for
fi-nst -Ens'
membrane
If the
are
all
zero.
If,
starts
from
the coefficients G^, i)m
rest,
however, we suppose, for the sake of greater
generality, that the initial motion is defined

"
by the equation
A=
<Kn0),
\dtJo
we must have
(r,
8)
2|,">
(Cm cos n8 +
from which the coefficients
From
Dm sin nd) Jn (4B r),
14
>
Gm D^ may
,
be determined.
the nature of the case the functions f(r,
8),
<f>
are one-valued, finite, and continuous, and are periodic in
period being
ih like
or an aliquot part of 2ir; thus f(r,
2-ir
manner
(f>
(r, 8),
f (r,
8)
may be expanded
=a +
! cos
+ 6,
sin 8
in' the
+b
6,
6)
the
and
form
+
sin 28 + ...
a2 cos 28
8),
(r,
. .
the quantities a 8 b being functions of r. The possibility of expanding these functions in series of the form "2,A,Jn (ic,r) has
been already considered in Chaps. VI. and VIII.
,
In order to realise more clearly the character of the solution

thus obtained, let us return to the normal oscillation corre-
sponding to
z=
ks being the sth root

G.
m.
Jn (Ksr) cos nd cos K ct,

of Jn ( s a) = 0.
98
[x.
Each element of the membrane executes a simple harmonic

oscillation of period
27T_27T
Kg
KgO
la_
J.
and of amplitude
Jn
The amplitude
at rest,
cos
(icsr)
116.
vanishes, and the element accordingly remains
if
= 0,
cos nd = 0.
Jn (Ksr)
or if
The
first
equation
not only
satisfied,
is
at the boundary, but also
when r = a,
that
is
when
r a,
= a,
...
r=
Ks
Kg
Kg
consequently there exists a series of
(s
1)
nodal circles concentric
with the fixed boundary.

equation, cos n6 = 0,
The second
ff-JL
ft-Ê
2n' ~
2n'
therefore there
membrane
is
is satisfied
fl
when
_ (4W - 1) 7T
2n
a system of n nodal diameters dividing the
into 2w equal segments every one of which vibrates
same way. It should be observed, however, that

any particular instant two adjacent segments are in opposite
in precisely the
at
phases.
The normal
vibration considered is a possible form of
tion not only for the complete circle but also for a
bounded by portions of the nodal
circles
oscilla-
membrane
and nodal diameters.
It is instructive to notice the dimensions of the quantities
which occur in the equations.

a-
The dimensional formula
for
is
that for
hence by 2 that of c
Since ks
and
[<r]
= [ML-3];
[T]
= [MT-2 ]
[c]
= [LT-i].
T is
is
is
found from
Jn ( a) = 0,
s
Ks a
is
an abstract number,
X.]
Thus the period
27r//cs c
comes
99
out, as of course it should, of
dimension [T].
If
we
write
the period
Ksa, so that
for
/x s
may be
the sth root of
is
fi s
Jn (x) = 0,
written in the form
la
2-n-a
f^M
%f
l6
Mrvr^VT'
where
clearly
is
the mass of the whole membrane.
how the
period
This shows very
increased by increasing the mass of
is
the membrane, or diminishing the tension to which
it is
sub-
jected.
As a
particular case, suppose
smallest root of
J (x) = 0;
vibration which
is
quency
n = 0, and
let /^
= 2*4048,
the
then we have the gravest mode of

symmetrical about the centre, and its fre-
is
2^\/j=y x
678389
Thus, for instance, if a circular membrane 10 cm. in diameter

and weighing "006 grm. per square cm. vibrates in its gravest
mode with a frequency
220, corresponding to the standard
adopted by Lord Rayleigh, the tension
,
,
25-7T
T is
x -6784
x 006/
determined by
= 220.
whence
220
\2
T =(^)
X 157r
-
= 4956
in dynes per centimetre, approximately.

of force this is about 50
per
In gravitational units
grams per centimetre,
or,
roughly, 3"4
lb.
foot.
In the case of an annular membrane bounded by the circles

and r = b, the normal type of vibration will generally inThus if we put
volve both Bessel and Neumann functions.
r
=a
[J
{Jn
(ica)
n Ur)
= A\-^
i
.
Y
Y
this will correspond to a possible
is
determined so as to
.
n (icr)\
^f-y 1-Y cos rid cos
n {Ka))
mode
17
icct,
of vibration provided that
satisfy
Jn (ko) Yn (Kb) - Jn (Kb) Yn (kcl) = 0.
72
100
It
may be
Yn that
[X.
inferred from the asymptotic values of
this equation has
an
infinite
number of
real roots
Jn
;
and
and it
seems probable that the solution

z
= 22
is sufficiently
{A cos nd
+ B sin
n6\
&^\ ~ 54^1
general to meet the case
cos K ct
when the membrane
io
starts
from rest in the configuration defined by
z=f{r,6).
Assuming that this is so, the coefficients A, B can be expressed in the form of definite integrals by a method precisely
similar to that explained in Chap. VI.
Thus if we write
u
it will
_ Jn (icr) _ Yn (kt)
Jn (ica)
F(/ra)'
be found that
d6
JO
f "d0
J0
f(r, 6) ur cos
nddr = LA
Ja
20
f(r, 6)
ur sin nddr
= LB,
Ja
where
21
This value of
L may
perhaps be reducible to a simpler form
in virtue of the condition
Jn {kcl) Yn (xb) - Jn (icb) Yn (ko) = 0.

For a more detailed treatment of the subject of
reader is referred to Riemann's Partielle
gleichungen and Lord Rayleigh's Theory of Sound.
the
this chapter
Differential-
CHAPTER
XI.
HYDRODYNAMICS.
In Chapter VI.
<f)
it
has been shown that the expression
= 2 (A cos n6 + Bsin nff) e'*2 Jn (kr)
satisfies Laplace's
equation
V < = 0,
2
and some physical applications of this result have been already

considered.
In the theory of fluid motion
may be interpreted
<j!>
as a velocity-potential defining a form of steady irrotational motion
of an incompressible fluid, and
we have
We
is
a proper form to
assume when
to deal with cylindrical boundaries.

shall not stay to discuss
any of the
special problems thus
suggested, but proceed to consider some in which the
procedure
is less
method
of
obvious.
Let there be a mass of incompressible fluid of unit density

in such a way that the path of each element lies in a
moving
plane containing the axis of

equal to
to,
z,
and that the molecular rotation
is
the axis of rotation for any element being perpen-
dicular to the plane which contains its path.
Then, taking cylindrical coordinates

noting by
u,
v the
component
r, 0,
z as usual, and de-
velocities along r
and
parallel to
the axis of z respectively,
fr(ur)
a
and
+ ^(vr) = 0,
du
5-
or
oz
dv
= 2g>.
HYDRODYNAMICS.
102
Equation
where
i|r
is
[XI.
may put
shows that we
Stokes's current function
thus equation 2 becomes
'
dr \r dr J
When
the motion
is
dz \r dz
steady,
yjr
is
a function of r and z
and
if
we put
=m +v
2
<f
2
,
so that 5 is the resultant velocity, the
dynamical equations may
|+|:<W-?-*|
whence
The
it
follows that co/r
simplest hypothesis
must be expressible as a function

a>
where f
is
a constant
i^.
= #r,
on this assumption, 4 becomes
3r \r dr )
Now
of
is
dz \r dz J
the ordinary differential equation
dr \r dr )
is satisfied
by
i
s
X = lr + Ar + B,
where
and
are arbitrary constants
and
= X + P r cos nz
V
where p
is
if
we assume
>
a function of r only, we find from 7 that

2
the solution of which
is
dr-
f3H* + ?)' =
'
= C^/j (717-) + D^! (nr).
Finally, then,
$-
= Kt" + At* + B + r2
{GJ,
(nr)
+ DnK, (nr)}
cos nz,
1
HYDRODYNAMICS.
XI.]
103
where the values of n and of the other constants have to be

determined so as to meet the requirements of the boundary
conditions.
Suppose, for instance, that the fluid
fills
the finite space in-
by the cylinders r = a, r = b and the planes z

the boundary conditions are
closed
=+
Then
h.
^=
cz
when r = a
or
b,
for all values of z
and
|fc.
or
when z =
h, for all
values of r such that
atrtb.
One way
of satisfying these conditions
is
to
make
yfr
constant
and equal to zero at every point on the boundary.

we put
Now
if
+-KC )<--*>-wa$>-!$}=5
this is
when
of the right form,
= b,
for
r = a.
It vanishes
provided the values of n are chosen so as to satisfy
ij (na)
and, finally,
and vanishes
it
vanishes
(nb)
(nb)
when
z=h
gi(nr)
),
-i(r
(na)
if
the coefficients
Gn
are
determined so that
il^wr)
rU()
for all values of r
Si(na)J
> r > b.
possibility of this expansion, the coefficients
are found in the usual
The
yy/
b)l
such that
Assuming the
)(
a)(X
way by
integration.
stream-lines are defined by

ty
= const.,
8=
const.,
so that the outermost particles of fluid remain, throughout the
motion, in contact with the containing vessel.
(The above solution was given in the Mathematical Tripos,

Jan. 1884.)
104
HYDRODYNAMICS.
[xi.
Some very interesting results have been obtained by Lord

Kelvin (Phil. Mag. (5) x. (1880), p. 155) in connexion with the
oscillations of a cylindrical vortex about a state of steady motion.
Adopting the fluzional notation to denote complete differentiation
with respect to the time, the dynamical equations of motion, and
the equation of continuity, are in cylindrical coordinates
_dp
J _
dr
rd6~
-dz
t,
r,
is
We
Tr
o-t
+z
or
xdz
dd
dr
dz
'
12
dz
.dz
+ e dd +Z dz'
dr
3 (r0)
dz
or
rod
dz
to
obtain a possible state of steady motion by supposing that
r
this
.dz
+r
+ad
dr
be understood that r, 0, z are treated as functions

while r, 6, z are supposed expressed as explicit functions
0, z, t; and the density of the liquid is taken to be unity.
It
of
3r
tra + r
dt
dpL= dz
of
dt
makes the
= 0,
= io
= 0,
(a,
constant),
resultant velocity
U = cor,
while the pressure
II
14
is
= Jto'rdr = n + i<V,
being a constant depending on the boundary conditions.
Now assume as a solution of 12 and 13

r = p cos mz sin (nt s0), r0=U+u cos mz cos (nt sd),)
z = w sin mz sin (nt s0),
p = II + vs cos mz cos (nt s0),\
where
s is
a real integer, m, n are constants, and
functions of r which are
small
in
p, u, w, vr are
comparison with U.
Then
substituting in 12 and 13 and neglecting squares and products of
small quantities,
we
obtain the approximate equations

d'ST
dr
nt
= (n ~ sm > p ~ 2o)U
.
'
= (n seo) u + 2o>p,
m-sr
dp
dr
= (n sco) w,
o
su
=
T+-+-+mw
r
r
,
0.
18
HYDRODYNAMICS.
XI.J
From
we
equations 17
105
obtain
dw
2sa>
(n-Sft>)|(n-sa>)~
w)
m {4o> (n sta)
2
{4ft>
19
(- sto) 120,-j3
s(n- sm)
dw
(W SO))
and on substituting these expressions in
we
find, after
little
reduction,
\dw (m
JJ2 + rrfr
-T+ (
ctr
d?w
- (n - swf\ s
---2\ W =
\i
(n scof
r
2
(4a> 2
If the quantity
{4w 2 (n
(n
is positive, let it
Then
2
.
be called k2
s<u)
if it is negative, let it
be denoted by
in the first case
w = AJ
+ BYS (kv),
(r)
and
8u)2
in the second case
w = CI
(\r)
+ DK
22
(Xr).
The constants must be determined by appropriate

For instance, suppose the
boundary conditions.
initial or
fluid to occupy,
during the steady motion, the whole interior of the cylinder r = a.

Then in order that, in the disturbed motion, w may be everywhere
small
necessary that
it is
To
fix
B=
in
the ideas, suppose m, n,

4o>
D=
20 and
s, &>
> (n - sw)
assigned,
in 21.
and that
2
;
then by 19 and 21
A
"
By
(n
- sw)
if
- sco) kJ', (r) - -^- Js (*r)
m {4<b
(n sw)
16 the corresponding radial velocity
and
\(n
is
= a,
is
= p cos mz sin (nt sO)
the value of p
radius, for r
23
when r=a, the
initial velocity
is
cos
mz sin s0.
along the
HYDRODYNAMICS.
106
Now
may have any
(small) constant value
this is prescribed, the constant
by
[XI.
is
determined,
supposing that
its
value being,
23,
{4a)"
(n
(n sa>) mp
2
sco) \(n sa>) kJ' (ko)

s
(.
(tea)
ic'po/m
24
kJ'' (ko)
Of
course, the
There
is
-.
(n
r J. (KO)
v
'
sa>)a
other initial component velocities and the
must be adjusted
initial pressure
equations 16
so as to be consistent with the
19.
no
difficulty in realising the general nature of the
disturbance represented by the equations 16;
it
evidently travels
round the axis of the cylinder with constant angular velocity
When
is
given,
we can obtain a very
n/s.
general solution by
compounding the different disturbances of the type considered

which arise when we take different values of m,n,s; according to
Lord Kelvin it is possible to construct in this way the solution for
" any arbitrary distribution of the generative disturbance over the
cylindric surface, and for any arbitrary periodic function of the
time."
The general
solution involves both the
Another case of steady motion

vortex in a fixed cylindrical tube.
is
7- (9
is
If a
is
functions.
obtained by putting
= c,
where c is a constant; the velocity-potential

at any point is
U = -.
r
the
that of a hollow irrotational
This
= 0, i=0,
J and
is c0,
and the
velocity
25
the radius of the free surface, the pressure for the un-
disturbed motion
is
HYDRODYNAMICS.
XI.]
107
Putting these values of

and II in equations 16 and proceeding as before, we find for the approximate equations corresponding to 17 and 18
dur
2cu
cs\
S'sr
cs\
mzT
dp
dr
ldw
w = AI
(mr)
w,
by
that
is,
is
^v
s
+ BK
(mr),
30
is
defined by r =
w=A
Thus
we must have r =
when r = b.
Is (mr)
\I's
still
b,
by 16 and 29
dr
(mb)
K (mr))
s
31
K's (mb)]-
to express the condition that
p=
point on the free surface for the disturbed motion.
must
arbitrary constants.
dw
have
(nP
r dr
If the fixed boundary
We
differential equation satisfied
Consequently
= -âz
m dr
dHu
dr2
B are
m\
dw
and therefore the
28
/
-in
mr
p
when r = b
su
sw
Hence
where A,
IT
at every
To do
this
find a first approximation to the form of the free surface.
the steady motion, the coordinates
r,
we
In
z of a particle of fluid
remain invariable and

8
whence
In the disturbed motion r does not

value r, and if we take the equation
r
0=~.
differ
= p cos mz sin (nt sd)
much from
its
mean
HYDRODYNAMICS.
108
we obtain a
mean
[XI.
approximation by putting 6
first
value p and neglecting the variation of z
r = p cos
mz sin n
ct
giving p
its
thus
cs\,
-^
and therefore
r
=r
cos m.s cos (nt
s6).
32
cs
Putting r
p
= a, and
writing p a for the corresponding value of
the approximate equation of the free surface

r
u cos mz cos (nt s&).
=a
33
sc
Now
is
;
2
by 16 and 26
p=
II
and the condition
= J c2
~ c 2 (
p=U
- +
J
gives,
to-
cos m;z cos (nt
- s6)
with the help of 33,
HYDRODYNAMICS.
XI.]
This
may
109
be regarded as an equation to find n when the other
quantities are given.
we
If
write
c
a2
(the angular velocity at the free surface in the steady motion),
and
at-
_,
ma
K' {ma) \
K'g (mb)]
J'u{ma)
\I's
(mb)
K (ma)
Is {ma)
37
K's {mb)]'
\l',{mb)
the roots of the equation 35 are given by
n=
*/N).
(s
38
an abstract number, which is positive whenever a, b, m

> a. Thus the steady motion is stable in relation
to disturbances of the type here considered.
This might have
been anticipated, from other considerations.
iV
is
are real and b
The
values
interpretation of 38
to, s
that corresponding to each set of
is
there are two oscillations of the type 16, travelling
with angular velocities
(l +
^)
-(l-^)
and
respectively about the axis of the vortex.
A special
case worth noticing
when
is
00
In this case
we
must put
w = AK
and 37 reduces
(mr)
to
(ma)
N = ma^
K (ma){.
K',
The
third
cylindrical
case
core
considered by
rotating
like
liquid which extends to infinity
no
slip at
the interface between
the radius of the core,
is
that
of
and moves
it and the
irrotationally,
core.
Thus
if
with
is
we have
TJ
= a>r when
tt =
U
for the
Lord Kelvin
a solid body and surrounded by
aa?
undisturbed motion.
r<
a,~\
u
^ a,
when
r>
39
HYDRODYNAMICS.
110
[XI.
For the disturbed motion we start as before with equations

and by precisely the same analysis we find
w = AJ (icr) when
w = BK (mr) when
tf
At the
both
...
With
,
2
interface
< a,'
> a,
-(n-s)
rr
(n sa>f
{4ft)
40
L,
p,
Now by
sides.
16,
and
must have the same value on
ts
17 and 27
it
same when those of w agree

be satisfied are, by 23 and 29,
are the
follows that the values of

;
hence the two conditions to
A Js (ko) = BK
(ma),
(ko) >
and
A (n sea)
(n
son)
m [4ta
Ia
kJ's (ko)
'
.
2
(n
- say)
.
= - BK'sK
S (ma).
'
42
*
Eliminating A/B, we obtain, on reduction,
which
or,
is
2sw
i?aK', (ma)
icaJ's (ko)
mK
(a)
(ma)
_ ft
sa>
'
the same thing,
/v-
K?aK',0ma)
micaJ'Aica)
J s (ica)
s (ma)
a transcendental equation to find k when the other constants are
When
given.
is
known, n
n
= (o
is
[s
V
given by
2m
.-
\
I
Jif+mV
For a proof that the equation 43 has an infinite number of

and for a more complete discussion of the three problems
real roots,
in question, the reader
is
referred to the original paper above
cited.
Since the expression for i involves the factor sin mz, we may,
if
we
like,
suppose that the planes z
boundaries of the
We
will
now
and z = ir/m
are fixed
fluid.
consider the irrotational wave-motion of homo-
geneous liquid contained in a cylindrical tank of radius a and
depth
z
The upper surface

when undisturbed.
h.
is
supposed
free,
and in the plane
HYDRODYNAMICS.
XI.]
The
velocity-potential
(j>
must
Ill
satisfy the equation
~
"*"
Br2
and
also the
r dr
2
r-
"*"
30 a
44
fo3
boundary conditions
-i =
when
oz
= h,
45
^- =
when r = a.
or
These conditions are
all fulfilled if
= AJn (r) sin n0 cosh
4>
provided that k
we assume
k {z
+ h) cos mt,
46
chosen so that
is
Jn(ica)=0.
If gravity
for
47
the only force acting,
is
we
have, as the condition
free surface,
when z =
neglecting small quantities of the second order
0,
therefore
m?
cosh kU
m = gic tanh
or
The equations
46, 47,
tank,
interior of the
may
be one-valued.
B
of roots 1"',
,
jj
',
etc.,
49
/e/i.
49 give a form of
a normal type of oscillation
<f>
+ gic sinh ich = 0,
when the
<
corresponding to
liquid occupies the whole
n must be a whole number in order that

The equation 47 has an infinite number
so that for each value of n we may write,
more generally,
= 2vl Jn {dpr) cosh /4
M)
<f>
(z
h) cos
mfH sin nd,
50
and by compounding the solutions which arise from different

integral values of n we obtain an expression for
which contains
a doubly infinite number of terms. Moreover instead of the single
<fr
trigonometrical factor
A cos mt sin nd
in the typical
(A
cos
term we
mt + B
where A, B,
C,
may put
sin nit) sin
nd + (C cos mt + D sin mt)
are arbitrary constants.
cos nd,
HYDRODYNAMICS.
112
As a simple
illustration, let us
<f>
then
as usual,
if,
= %AJ
we
(tcr)
write
tj
start
from
rest.
+ K) sin mt
cosh k{z
when
(/cr)
= 0,
Integrating 51
sinh
mt,
ich sin
the liquid must be supposed
to
we have
with regard to
possible initial form of the free surface

71
level,
~2,kA
this vanishes
= 0, and put
take n
for the elevation of the free surface at
any moment above the mean
and since
[XI.
t,
denned by
2 A sinh KhJ (at),
C2
the summation referring to the roots of
J '(a) = 0.
By
the methods of Chap. VI. the solution
may be
to suit a prescribed form of initial free surface defined
adapted
by the
equation
*?=/(*).
It will
if,
be observed that in 49
ich is
in the special case last considered,
root of
J '(X) = 0,
specially
V Xg
interesting
diaphragm, whose thickness
case
coth
is
defined
and
a
is
occurs
may be
X, is
^.
when a
rigid vertical
neglected, extends from the
axis of the tank to its circumference.
diaphragm
so that
the period of the corresponding oscillation
?5 =2
an abstract number
we put a = \,
If the position of the
by 6 = 0, we must have,
in addition to the
other conditions,
?f
dp
when
= 0.
This excludes some, but not all, of the normal oscillations

which are possible in the absence of the barrier but besides
those which can be retained, we have a new set which are ob;
tained by supposing
rc
where h
is
any integer.
= Jc+ \,
Thus
in the simplest case,
when k = 0,
we may put
<
= AJh {icr) cos jr cosh k{z + K) cos mt,
HYDRODYNAMICS.
XI.]
or,
which
is
113
the same thing,

a
cos ^ cosh
$ = Ar~i sin (at)
tc
+ h) cos mt,
(z
with the conditions

tan tea ica=
m = gic tanh
2
0,\
"
'
kIi J
The equation
tan x x
number of real roots, and to each of these corresponds an oscillation of the type represented by 53.
has an infinite
More
generally, if
where A
is
<f>
we put
(2fc+l)7T
[*<W\
2g
any integer, the function
= (A cos mt + B sin m) J" (r) cosh k(z + h) sin ?i0,
with the conditions 47 and 49 as before, defines a normal type of

oscillation in a tank of depth h bounded by the cylinder r = a and
the planes 6 = + a.
Similar considerations apply to the vibrations of a circular
membrane with one
radius fixed, and of a
membrane
in the shape
of a sector of a circle (see Eayleigh's Theory of Sound,
Another instructive problem, due to Lord Kelvin

may be stated as follows.
I.
p. 277).
(Phil.
Mag.
(5) x. (1880), p. 109),
circular basin, containing heavy homogeneous liquid, rotates

with uniform angular velocity a> about the vertical through its
centre
it is
required to investigate the oscillations of the liquid
on the assumptions that the motion of each particle

nearly horizontal, and only deviates slightly from what
if
is infinitely
it
would be
and
the liquid and basin together rotated like a rigid body
further that the velocity
same
always equal for particles in the
vertical.
The legitimacy
that, if
is
a.
constant,
G. M.
of these assumptions
the radius of the basin,
g and that the

with
is
We
m 2a
is
is
secured
greatest depth of the liquid
is
we suppose
small in comparison
shall suppose, for simplicity, that the
and equal
if
small in comparison with
mean depth
to h.
is
HYDRODYNAMICS.
114
[XI.
Let the motion be referred to horizontal rectangular axes

which meet on the axis of rotation, and are rigidly connected with
Then
the basin.
if u,
v are the component velocities, parallel to
these axes, of a particle whose coordinates are
x, y,
the approxi-
mate equations of motion are

du
ldp
)
'
p dx
55
p oy
ot
+z
If h
is
the depth of the liquid in the vertical through the
point considered, the equation of continuity
is
56
while the condition for a free surface leads to the equations
dp
dx
_
=
"
57
_ dz
dyWdydp
If
we eliminate p from
we obtain
55 by
means of 57 and change
to polar
coordinates,
du
dz
dr
58
dv
dl
where
u,
+ 2wu-
now denote the component
vector and perpendicular to
The equation
of continuity, in the
,
From
/du
dv
w\
\dr
rdd
rr)t
the equations 58
velocities along the radius
it.
we
new
notation,
dz
^r =
is
0.
59
dt
obtain
dz_
rdd
60
+*-)
3Z
&z
dr
rdOdt
d^ + 4a) \
2
(9
and eliminating
u, v
we
HYDRODYNAMICS.
XI.]
obtain a differential equation in
z,
115
which, after reduction,
is
found
to be
fe
+4
y-
'
03
The work now proceeds as in other similar cases already conThus for instance in the simplest case, that of an open
circular pond with a vertical bank, we take m to be a real integer,
sidered.
and put
z
The boundary
= Jm (kt) cos (m9 nt).
64
condition
when r = a
u=
gives, for the determination of n, the equation
2mo>
If
a)
is
Jm (ko) - nica <Tm (a) = 0.
small in comparison with gh,

2
/t
65
we have approximately
=-r
gh
and 65 becomes
In the general case
60 are
satisfied
it will
be found that the equations 58 and
by putting
u=U sin (m0 - nt),

_
dS
= Fcos (m9 - nt),

2me\
66
67
n3
-4w
dr
r J)
82
HYDRODYNAMICS.
116
By assuming
for the solution of
[XI.
62
Z=AJm (*r)+BYm (Kr)

we
which
obtain a value for z
circular
may be
adapted to the case of a
with a circular island in the middle.
pond
It should be remarked that the problem was suggested to Lord

Kelvin by Laplace's dynamical theory of the tides the solution is
applicable to waves in a shallow lake or inland sea, if we put
o> = 7 sin \ j being the earth's angular velocity, and \ the latitude
:
of the lake or sea, which
is
supposed to be of comparatively small
dimensions.
We
will
conclude the chapter with a brief account of the
application of Bessel fu actions to the two-dimensional motion of a
II.
244, that if
p.
may be shown, as in Basset's Hydrodynamics,

we suppose the liquid to be of unit density, and
It
viscous liquid.
that no forces act, the equations of motion are
2 dv\
!+"("-?
r dd)
'
68
If
-vjr
is
-=-^
+ C^
rdO
dp
uv
-/*(*
if
~r*
+r
dii
dd
the current function,
U
and
~rd0'
djr
'
dr
we put
/**-?* = *
the equations of motion
v*-
dp
dx
du
dr
r dd
dt
dp
3%
dv
rdO
dr
dt
uv
r
If
may
69
70
the squares and products of the velocities are neglected,
HYDRODYNAMICS.
XI.J
117
Hence, eliminating p,
dr\ dr)
which
or,
is
rd6*~
'
the same thing,
V x = 0.
2
72
comparatively simple solution
may be
constructed by sup-
posing that
%=
and
^ is a function of r only.
d*y
if
we
dV
mi
write
one value of M?
a,
[(l-i) K r}.
obtain a real function for
= (a + /3i) e
&m
is
V = AJ
i|r
fM
m_
We
This leads to
r dr
dr-
and
= yremti
yjr
where
0,
mti
t|t
by putting
- i) at} + (a - i) c"*"/, {(1 + i) kt},
{(1
being any real constants.
scribed to be
am sin mt when r =
ckb sin
mt
,X
Suppose the velocity

a; then
is
73
pre-
\3j- /r=o
/c
(1
- i) (a + @i) emti J"
+ (1 + i) (a - /ft) e-"
1 '1
{(1
'./;
{(1
- z) a}
+*) ko],
and therefore
(1 -i) (a + /8t)./i{(l -*)
m] = -
^,
so that
aa>(l-t).7"t {(l-i)*r}
Wi{(l-i)*a}
ffla>(l+*V,f(l+t) r}
4K^k{(l
+ *)a}
In order to obtain this explicitly in a real form,
let
K
"
us write
74
118
HYDRODYNAMICS.
and Q being
real functions
[XI.
then
yfr = \aas {(P + Qi) (cos mt + i sin mt)

+ (P Qi) (cos mt i sin mt)}
= âa>(P cos mtQ sin mt),
The boundary condition may be

liquid to
is
fill
move with angular
with
with
it
velocity
a>
sin
a,
the
which
mt about
its
the particles of liquid which are in contact
it.
(This example
is
Wednesday
Tripos,
taken from the paper set in the Mathematical
afternoon, Jan.
3,
1883.)
very important application of the theory
memoir
Stokes's
"
On
is
contained in
the effect of the internal friction of
on the motion of pendulums" (Camb. Phil. Trans.,

for
75
by supposing
the interior of an infinite cylinder of radius
constrained to
axis, carrying
realised
vol.
fluids
IX.):
the details of the investigation the reader should consult
the original paper, but
we
shall
endeavour to give an
outline
of the analysis.
The
practical
problem
that of taking into account the
is
viscosity of the air in considering the small oscillations, under
the action of gravity, of a cylindrical pendulum.

simplify the analysis,
we begin by supposing
cylinder of radius
infinite
that
In order
to
we have
an
surrounded by viscous liquid
a,
of
and we proceed to construct

a possible state of two-dimensional motion in which the cylinder
moves to and fro along the initial line 8 = in such a way that its
velocity V at any instant is expressed by the formula
density
p, also
extending to infinity
V = erf + c e-* vn
11
ti
S)
where v
= fi/p,
fi
76
being the coefficient of viscosity
jugate complex constants, and
c,
c are
n, n are con-
conjugate complex constants
of small absolute value.
The current
function
i|r
must vanish at
infinity,
and
satisfy the
equation
and, in addition, the boundary conditions
It
00
when r = a.
^Va cos6, d-f=Vsme,

or
78
HYDRODYNAMICS.
XI.]
Now
*=
if
119
we assume
[***"
B
{7 + X W} + <r# { + 5oXo (r) J]
part of this expression, namely the
sum
sin 6
of the first
79
and third
terms, satisfies the equation
V Y = 0,
and the remaining part
%%
provided the functions
dr3
satisfies
are chosen so that
r dr
d\
dr
These equations are
satisfied
=K
by
{(l+i)*Jnr},
Put
(1
(1
+ i) V?i = ^j
- i) \Mo = \>
then
X
Xo
where
P and Q
=K (\r)=P +
=K (\ r)=P-iQ$>
iQ,)
1
1
are real functions of
The boundary
80
!)
r.
conditions are satisfied if
-^+Bx
'(a)
= c;
whence
x( a)+ ax'(a)
81
X () + ax' ()
aad
J.
are obtained from these
by changing
into
-i
HYDRODYNAMICS.
120
[XI.
The equations 79, 80, 8 1 may be regarded as giving the motion

when the cylinder is constrained to move according to
the law expressed by 76.
of the fluid
By
proceeding as in Basset's Hydrodynamics,
shown that the
II.
280
it
may
be
resistance to the motion of the cylinder, arising
from the viscosity of the surrounding
liquid,
amounts, per unit
length of the cylinder, to
Z = 2-Trpvnia (Le2mti - L^**),
82
where
L = ~-BX (a)
_ ca\ax(a)-Zx(a )}
X
and
+ ax(a)
o.
*3
'
conjugate to L.
is
Let
must
(a)
<t
be the density of the cylinder: then the force which
act at time
upon each unit length of
tain the prescribed motion,
F=ir<ro?
it,
in order to main-
is
~+Z
dt
= 2iriva (Ne2mti - N^-^ ),

1
84
with
iV=
i^
nc
x( a) + ax'( a )
=the conjugate quantity.
85
Now let us suppose that we have a pendulum consisting of

a heavy cylindrical bob suspended by a fine wire and making
small oscillations in air under the action of gravity.
We
shall
assume that when the amplitude of the oscillation is sufficiently

small, and the period sufficiently great, the motion will be approximately of the same type as that which has just been worked
out for an infinite cylinder; so that
placement of the bob at time
from
if
is
the horizontal
mean
its
position,
we
dis-
shall
have
The
force arising
= V = ce*mti + c e~ "m
!
ti
.
from gravity which acts upon the bob
unit of length, and to the
first
order of small quantities,
-TT^-p)^.^,
86
<'
is,
per
HYDRODYNAMICS.
XI.]
where
is
121
the distance of the centre of mass of the pendulum from
the point of suspension!
Equating this to the value of
F given
above,
we have the
con-
ditional equation
%vl (N'e2mH
- N e-imai) + {a--p)g% = 0,
87
which must hold at every instant, and may therefore be differentiated with regard to the time.
Doing this, and substituting
for g its value in
{-
terms of the time, we obtain
invHN + (<r-p) gc]
which
is satisfied
which
is
{- kn^lN^ + (<r-p) gc ] e-2"nK
identically if
(a-
or,
imti
= 0,
we put
p) go = 4mvHN,
the same thing,
(ZZE>9 = 4 ( a +
i
"*>)-y) p ) ,.,
x( a ) + ax
()
88
This, with ^ (a) defined by 80 above, is an equation to find n

which must be solved by approximation: since the motion is
must have a positive

when p is very small in
actually retarded, the proper value of n
imaginary part.
As might be
comparison with
a,
expected,
4,vV = g/l
approximately.
The
% and
by the
together with the equations 86 and 87.
constants c and c are determined
initial
values of
CHAPTER
XII.
STEADY FLOW OF ELECTRICITY OR OF HEAT IN UNIFORM

ISOTROPIC MEDIA.
Chapter VII.
pansions, contains
above, which deals with

all
that
Fourier-Bessel Ex-
required for the application of
is
Bessel Functions to problems regarding the distribution of potential;
but
discussion
it
may
be advisable to supplement that theoretical
by a few examples
fully
worked
We
out.
a few cases of electric flow of some physical
take here
Other
problems with notes as to their solution in certain cases will
interest.
be found in the collection of Examples at the end of the book.

In the discussions in this chapter we speak of the flow as electric
but the problems solved
may
be regarded as problems in the
theory of the steady flux of heat or incompressible fluid moving

irrotationally, or
even of the distribution of potential and
an electrostatic
field.
stood.
The
potential
The method
the flux
in
of translation
of
is
electricity
force in
well under-
becomes the
temperature in the thermal analogue, while the conductivities and

strength of source (or sink) involve no change of nomenclature;
the potential in the flux theory and that in the electrostatic theory
coincide, the sources and sinks in the former become positive and
negative charges in the latter, while specific inductive capacity
takes the place of conductivity.
If
V be
the potential, then in
all
the problems here considered
the differential equation which holds throughout the
dx*
or in cylindrical coordinates
df
dz*
'
r, 6, z,
d*V
15F
dr>
r dr
r 2 d&>
d*V
d*V
dz a
medium
is
XII.]
STEADY FLOW OF ELECTRICITY OR OF HEAT, ETC.
At the
ductivities
surface of separation of two media of

klt &2 the condition which holds is
where n 1
n^
123
different con-
denote normals drawn from a point of the surface
and V1} V2 are the potentials in the

two media infinitely near that point.
If one of the media is
=
an insulator, so that say &2 0, the equation of condition is
into the respective media,
dn
Let us define a source or sink as a place where
electricity is
drawn off from the medium, and consider the electricity

delivered or drawn off uniformly over a small spherical electrode
led into or
of perfectly conducting substance
medium
the bounding surface.

or
(of radius
r) buried in the
at a distance great in comparison with r from any part of
withdraw a
Let
be kept at potential V, and deliver

S per unit of time, then since
it
quantity
total
V = constant/r,
3
S
The quantity on the
right
is
4,-rrkr
half the resistance between a source
and a sink thus buried in the medium and kept at a
difference of
potential IV.
on the surface (supposed of continuous

the electrode must be considered as a
will be double the former amount.
resistance
hemisphere, and the
If the
electrode
curvature) of the
is
medium
In this case
I8
When
made
is
therefore in the
infinitely small
two cases just
*
2-wkr
we must have rV
finite,
and
specified
Ltr7= 4nrk
and 7 are the conditions to be fulfilled

in the problems which we now proceed to give examples of.
Those we here choose are taken from a very instructive paper
Equations
I,
2,
3,
STEADY FLOW OF ELECTRICITY OR OF HEAT
124
by Weber ("Ueber Bessel'sche Functionen und

auf die Theorie der elektrischen Strome,"
ihre.
[XII.
Anwendung
Grelle, Bd. 75, 1873),
and
some changes in notation to suit that adopted

in the present treatise, and the addition of some explanatory
are given with only
analysis.
We
shall prove first the following proposition.
If
be the
potential due to a circular disk of radius ry on which there
is
charge of electricity in equilibrium unaffected by the action of

electricity external to the disk, then if z be taken along the axis
of the disk,
and the origin
at the centre,
V = r
e ^shx(\ ri )J
IT
(Xr)^,
Jo
where the upper sign
is
to be taken for positive values of z and
the lower for negative values, and
In the
we can prove
the potential at the disk.
c is
place this expression for
first
that
we
the proper value of the electric density

solution.
By
46, p. 18 above, if e
e _e * sin (Xrj) / (Xr)
=-
= 1-
>
satisfies
shall
if
have verified the
0,
/""
dd
cos (Xr sin 0)
d0
er eK sin \rj cos (\r sin 6)
dX
Jo
A,
since changing the order of integration is permissible here.
/7\
e""**
sin (XTi) cos (\r sin 8)
/,
This can be written
= [
2.
e-<* [sin {X (r,
But we know that
if
/
Jo
+ r sin 0)} + sin
{X (r,
>
e
-ax gm x
^x
a2 + l"
then
and gives
d\
X
e~ eK sin (Xrj)
""Jo
when z = 0,
reduces to a constant
it
A.
- r sin 0)}]

IN
XII.]
'
UNIFORM ISOTROPIC MEDIA.
125
Multiplying this equation by da and integrating both sides from
a=
to
a=
>
oo (e
we obtain
0),
e~' x
smx
/,o
Thus the
= 2 tan
dx
x
IT
integral considered has the value
~ 2 tan_1 K+rlm 8 "
-1
tan ~
-rdn0
rx
'
and
e~' K sin
dQ
Jo
a,
= -pr h
2
d# tan-1
2j
The
e
(Xr^ cos (Xr sin 0)
jo
integral on the left
including
Hence
0.
>
n + rsmO
is
we
if
dd tan"
2J
r,
convergent for
all
r sin 8
10
positive values of
evaluate the equivalent expression on

e we shall obtain the
when z = 0. In doing
the right for a very small positive value of

value of the integral on the right of 8
this there
is
no
difficulty if
>
r-j
but
r;
of the last integral in io, for which 8
< r, the element

rjr, is dd, and the
rx
if
-1
= sin
integral requires discussion.
The
The second
small.
sought
is
Jo
-jr-
rx
tan-1
ri
d6(%- tan_
f
J
is
~ rsing
dd - 2
d0
rx
since e is not zero.
element of the second
-If
2i(
be very
r sin 8\
tan-
first
integral
when 8 =
^^
<tf.
.,
sin *
sin-
\irdd except just
so that the integral
\2
Jo
Each element of the

right
when rx > r,
Now
7r /2.
^- = 2
-rsm8a
-2
also vanishes
in that case
f'd8 tan-1
of the two integrals on the right vanishes if
first
down on the
when it vanishes,
just written
sin
-1
(rjr),
Similarly except just at the beginning each

is
îrd8.
Hence
dd tan-1
?-j
- r sin 8a
it
for r
sin
-1
>r
7*1
7T
=2
126
and
STEADY FLOW OP ELECTRICITY OR OF HEAT
[xil.
IN
XII.]
as small as
we
127
please in comparison with the total flow elsewhere
by increasing the radius of the
The
total flow
disk.
from the disk to the medium
S = -2k
Putting in this
for
dVjdz
is
thus
rndV
p*
J'o
its
~rdrdtj>.
we get
value
JO
= 8c/fov
Thus the amount supplied by each
time is 4>ckrl and we have
side
disk per unit
of the
8
on the bounding surface of a conductor
take place only from one face to the conducting
If the disk is laid
the flow will

mass, and
has only half of

c
its
value in the other case.
Then
16
'
Ahrx
holds all over the surface

dV/dn =
except at the disk-electrode, and of course 2 holds within the
At any point of the disk distant r from the centre
conductor.
In
this case the condition
_3F = 2c
dz
tv
\_
vV r2
a
^kr, Vr2 r2
'
We can now find the resistance of the conducting mass between

two such conducting electrodes, a source and a sink, placed anywhere on the surface at such a distance apart that the streamlines from or to either of them are not in its neighbourhood
disturbed by the position of the other.
The whole current up to
the disk by which the current enters is S, and we have seen that
For distinction let the potentials
c is the potential of that disk.
of the source and sink disks be denoted by Cj, c 2 then if R be
the resistance between them
;
If the wires leading the current
up
to
and away from the
electrodes have resistances p lt p 2 , and have'their farther extremities
128
(at the generator or battery)
F,-c, = %,
Fi
We
for
-V = Sp
3
F (cj c2) = S (jOj + p2

2
E = |{F -7
and
ct
lt
2,
the
falls of
and along the outgoing
potential along the inleading electrode,

are
so that
V V
at potentials
[XII.
-/S( o 1 +
j
),
18
o 2 )]-
Another expression for the resistance can be found as follows.

have seen that the potential at the source-disk is clt also that
conduction from one side of the disk
S = 4e &r
For the sink-disk the outward current in
manner
like
is
S = 4<cjcr
2.
S=2k (dr-j c r.
Hence
But
and
-R
also
c^ = 02^2,
From
a
S
stt
2k (cin
r,+r2
This result
is
we
infer that
first disk,
1/4^
1/4^
the
is
the part due to
of great importance, for
of calculating an inferior limit to
made on the
cfa)
the latter form of the result
part of the resistance due to the
the second.
so that
-r,
means
2 ).
it
gives a
the correction to be
resistance of a cylindrical wire in consequence of
its
being joined to a large mass of metal.
From
this
problem we can proceed to another which
with that of Nobili's rings solved

conductor
is
bounded by two
first
by Biemann.
parallel planes z
= a,
is identical
An
infinite
and two
electrodes are applied to these planes, so that their centres
the axis of
z.
disk
lie in
It is required to find the potential at each point of
the conductor and the resistance between the electrodes.
From
the distribution of potential the stream-lines can of course be
found
also.
IN UNIFORM ISOTROPIC MEDIA.
XII.]
The
solution
must
fulfil
the following conditions
W+; ^+ W =
dV
~-
0, for
-a<z< + a,
*
tor
z=a,r>r
n
= v,
dV
129
1,
According to the
last condition
the current
supposed to flow
is
along the axis in the direction of z decreasing.
The
first
condition
V=(
is satisfied
by assuming
{^(X)^' + f(X)e-Kz }J (Xr)dX
20
l>
Jo
where
<f>
i/r
(A,),
(X) are arbitrary functions of
integral convergent
and
fulfil
X which render the
the other necessary conditions.
Without loss of generality V may be supposed zero when z =

and hence we must put <f>(\.)=yjr (A.). Thus 20 becomes
F=fJo
With regard
2cj>
(X) sinh (Xz) J" (Xr) dX.
21
two conditions, by
to the other
0,
150, 151, p. 73,
above,
I
sin (ATi)
dX = 0, when r > r
(Xr)
Jo
1
when
Hence
if
r<r
we take
a
2<j>
(X) cosh (Xa)
._
^
both conditions will be
X=
sin
sin (X^),
Xrt
47T&7 1 cosh (Xa)

,
satisfied.
The
solution of the problem
therefore
8
V =MFj
Tr
G. M.
sinh (Xz)
..
,.
dX
^Xa) Sin(Xri)Jo(Xr) ^9
is
STEADY FLOW OF ELECTRICITY OK OF HEAT
130
From
we can
this
resistance
easily obtain
an approximation to the
For we have from 23
R between the
electrodes.
7
d c = TT/CTi
tanh (Xa) J (Xr) sin
(\ri)
Jo
=
;
[XII.
A.
A>
(Xn
sin
\j*!
24
If the second term in brackets be neglected in the last expression,
we have
to a first approximation, since
by
the integral
1 1
is
equal
to tt/2,
~ ^2 _ *
8 ~2&V
fit
This of course could have been obtained at once from 19 by

= r2 To obtain a nearer approximation the
simply putting ra
expression on the right in 24
and
may be expanded
in powers of r
If terms of the order r^/a? and upwards be neglected,
Tj.
the result
is
log2
1
R= 2kr
irka
we may put \r
If the electrodes are extremely small
for sin Xrj
and we obtain from 23
V- s
/;=&->*-
27rA;Jo cosh(Xa)
This expression applies to the space between the two planes
z= +
Hence expanding by
a.
uk
sinh (Xz)
cosh (Xa)\
Fourier's
2X a
^
a 1
method we obtain
nir
;
Ti
/mry
nirz
sm "5" sm -52a
Hence
5 a
irka
Now
it will
wr
wr f
2a
be proved,
p.
/"
X+ V2aJ
^
200 below, that

sin (fa;)
/.
Hence putting 2
f"
Jo
-ir
for (nir/2a) 2
(Xr-)XdX
* 2 +X2
XdX
/wry
(Xr)
,
if
x be
positive
df
we have
_2 r"_df_ f X sin (fo-X) dX

+X
TrJi Vf^lJo
2
2;

IN
XII.]
But
it
can be shown that according as r
>
r Xsin(gr\)d\
?r
since f and
r-
*2 +
Jo
Substituting in 28
K
+X
*a
Jo
Thus
we
tt
or
131
<
0,
29
are both here positive
\2
Ji
Vf ^!
2
30
^!
obtain
= -f-Ssm-s-sm
trka
2a
31
V\2 1
which agrees with the solution of the problem given by Riemann

(Werke, p. 58, or Pogg. Ann. Bd. 95, March, 1855).
mass instead of being infinite be a circular

and radius c, bounded by non-conducting matter,
the problem becomes more complicated. To solve it in this case a
part
must be added to V fulfilling the following conditions
If the conducting
cylinder of axis z
(!)
=0
-W + r~dF + ~te
W=
(2)
a7
dV dV =
0,
h o
we
r<C - a<z< + a
'
0, for
-=
(3)
If
f r
'
for
>
s=a,
r=
c,
a < z < + a.
write
'e-^dX
v
'
Ji
,,
V\ -l
2
ferK"\d\\
Ji V\ -l
2
32
Vi-x
and denote by Zj
stituted for
r,
(c),
(c),
&c. the
same quantities with
the conditions stated are found to be
For consider the
fulfilled
sub-
by
series
b1
Sirz
irz
sm^
+ b sm^ +
r
.
34
92
^
132
The
the
which are functions of
coefficients b u b s , ...,
taken as to
known
we have
i-Wl-\
'
'
expanded in powers of r and integrated
is
becomes
it
infinite
Thus
solutions
V\ -
Ji
If the first
can be so
fulfilled.
the general differential equation
of which there are two
that
r,
the differential equation in the medium.
three conditions are
first
By
fulfil
[xil.
We
= 0.
for
it is
found
therefore take as the
solution of 35
r+i er^dX
""^J-ix/l^v'
where #
is
a constant to be determined by the remaining equation
We
of condition, (3).
obtain in the notation of 32
P =
and the
total potential at
F+
r4i
trka
sin
-S
any point
sin
we were to put r = 0,
we should obtain c c2
If in 37
integrals
wrr
36
is
>
'
<r>
2a
{c)
^raim sm T
<>
> <
r>
.
Af2 (c)
a,
and could evaluate
37
J/
the
the difference of potential now
existing for the given total flow S.
This divided by
S would give
the resistance.
When r =
0,
(r)
= 7r, so
that the change in the resistance due
to the limitation of the flow to the finite cylinder is
^ M,
(c)
(c)
ka
ka
if
a
- be very small.
n M
Hence the
-11
resistance is approximately
R=
We
A plane
is
now
log 2
2&Ti
Tr&a
&a
-^
pass on to another problem also considered by Weber.
metal plate which
may be
regarded as of
infinite extent,
separated from a conductor of relatively smaller conductivity by
For example, this

an electrode of metal from a con-
a thin stratum of slightly conducting material.
may be a
film of gas separating
IN
XII.]
ducting liquid as in cases of polarization in

calculate the resistance for the case in
small and
is
133
cells.
We
shall
which the electrode
applied at a point within the conducting, mass.
is
Take
the axis of z along the line through the point electrode perpendicular to the metal plate, and the origin on the surface of the
Thus the point
conductor close to the plate.

at the point z
a,r =
electrode
is
applied
0.
We further suppose that there is a difference of potential w

between the surface of the conductor and the metal plate on the
This will give a slope of potential through
if 8 be the film thickness.
If the con-
other side of the film.
the film of amount w/8

ductivity of the film be
fe, the resistance for unit of area will be S/^,

and thus the flow per unit of area across the film is w&a /8. This
must be equal to the rate at which electricity is conducted up to
the surface of the conductor from within, which is kdV/dz.
Thus
if w be the positive difference between the plate and the conductor
surface, the condition holds
when z = 0,
where
h=
Sk/ki.
Let p, p', be the distances of any point z, r from the electrode

and from its image in the surface respectively. Then the differential
equation and the other conditions laid down are satisfied by
v=
provided that
k(rP -}h
39
'
the equations
fulfils
,87
-h-K- + w =
oz
at the surface, and
33
throughout the conductor.

solution
is
dw
+ rTr +
we should have had
W=
first
term on the right
The
if
d2
'
is
the
the film had not existed, the second
the increased potential at each point in consequence of the rise
in crossing the film from the plate.
A value of w which
satisfies
w=\
Jo
40
e-* z d> (\)
given by
is
(Kr)
d\
41
134
where
(\) is
,
p'i
p*
z+ a
dw
4?irk
p' 3
dz
dw
+ j-2)t
dz
'
surface condition becomes having regard
to 41,
^(^-r (i+Ax) * {x)/o(xr)dx=a

But
a the equation
differentiating with respect to

,00
e- aK JJ\r)d\
Jo
we
42
(a2
+ r2)i
get
e -<*J
(\r)\d\=
Jo
(a2
+ r )8
2
43
This substituted in 42 gives
^-fcST + AX'
Hence
Now
J. (Ar)
|~
e~^+^ t dt =
J z+a
MX
+ hX
so that
we have
hS
-t
ATTK
a '"
dte-n
J z+a
h
Z
-Pf
te-Ht
J ?2
27rAp'
e-hM J (\r)Xd\
2ttM
q
e
W
I
I
I
^
/
^v
44
'
IN
XII.]
by integration by
point
z,
Thus we obtain
parts.
for
135
the potential at any
4,irk\p
2-rrkh
p'J
~TI
",
Take a new variable f given by

ht
= f + z + a,
and the solution becomes
4-rrk
p'J
\p
The meaning
J(+ z + af + r*
2-rrkh ]
of this solution
is
that the introduction of the
non-conducting film renders the distribution of potential that
which would exist
for the
same
total flow 8,
were there a com-
bination of two equal positive sources of strength

electrode
and
Sji-rrk,
at the
image, with a linear source extending along the
its
image
axis of z from the
to
oo
and of intensity
J-e
2wkh
per unit of length, at distance from the point

If the conducting
enough pp'
= r, and
mass be of small thickness then nearly

z + a = a. Thus we obtain
2irk{r
if
as
we suppose
(z
If h/r be small
of
(z + a).
+ a)/h may
V hH* + r3)
be neglected.
we can expand (hW + r3)'*
in ascending powers
by the binomial theorem and integrate term by term.
We
thus get
Jo'JhW +
Wo
r*
2.4... 2n
\rj
= l%(-l)"{l.S...(2n-l)Y(f
by which the value of the integral may be calculated
too small.
Hence
if
r be very great
2irk r 3
'
47
if
r be not
'
136
[xil.
or the potential at a great distance from the electrode varies in-
versely as the cube of the distance.
We may
which the conducting

bounded by two parallel infinite planes, the metal plate
z 0, and the plane z = a, and the source is a disk electrode of
radius r, with its centre on the axis of z applied to the latter. As
before a feebly conducting film is supposed to exist between the
metal plate and the conducting substance.
mass
solve similarly the problem in
is
We
simply add a quantity w, as before, to the distribution of
potential which could have existed if there
by the solution
out above
had been no film. Thus

stratum with disk electrode worked
we have
Tr
F=
The
for the infinite
sinh Xz
..
Jô^ sin(Xri)/
potential
must
fulfil
dw
Tr-
(Xr
dX
>T + w
48
the conditions
= A0,
tor z
=a
oz
(since the flow from the source-electrode is supposed unaffected
by w)
h
dV
-=
oz
w = 0,
for
= 0,
besides of course the differential equation for points within the
medium.
The
condition
first
w=
is
satisfied if
2 cosh \{z
we take
a)$ (X) J
(Xr) dX.
Jo
Also when z
= 0,
w=_
\
.
zwfcrtJo cosh (Xa)
oz
- 2A
sinh (Xa)
(Xr)'
dX
X0 (X) J (Xr) dX
Jo
2 /cosh (Xa)
</>
(\) Jo (Xr)
dX
= 0.
d>
(X)
-^
sin (Xrt )
isirkrx
2Trkrl J
cosh (Xa) {cosh (Xa)
+ AX sinh (Xa)}
cosn
) SU1 (^ri) ô PÔ
cosh (Xa) {cosh (Xa) + AX sinh (Xa
M*
49
'
'
'
XII.]
137
so that
TJ.
smh(\z) + h\cosh(\z)
cosh(Xa)+ftXsmh(Xa)
27rr1 Jo
Va the
we denote by
If
,. .dX
v
^
'
potential at the disk electrode
,
S f sinh (Xa) + hX cosh (\a)
,.
Sin(Xr )J
'a=5-T
^ smhu/\
k (Xa)
k \ + hX
cosh
(Xa)
27TAT! J
T7.
._
o"X
(Xr')---,
and
if
the area of the electrode be very small
S fsinh(Xa) + AXcosh (Xa) T

5= % J o cosh(Xa) + AX~sinh(Xa") ^ ^r) "*"
~'2^
.
This
is,
we have
is
the difference of potential between the electrodes, that
Comparing it with the difference

same flow through the stratum of the con-
the disk and the metal plate.
of potential for the
ductor without the plate, that

potential
which
given by 23
for
is
with half the total difference of
the two electrodes at distance 2a,
is
5y
tanh (Xa) J (Xr) dX,
Attic j
we
see that
it
exceeds the latter by
Sh
2irk J
The
XJ (Xr)dX
+ hX sinh (Xa)}
cosh (Xa) {cosh (Xa)
resistance of the
compound stratum now considered
is
therefore
R= J:
4Ar
ib2 +
irka
XJ (Xr)dX
jLr cosh (Xa) {cosh
(Xa) + hX sinh (Xa)}
2irk J
Since the resistance
is
between the plate and the
electrode,
taken as of very small radius, J (Xr) = Jo (0) nearly, and

so we put unity for J (Xr) in the expansion just found.
The last
which
term
is
is
ductor,
the resistance of the film between the plate and the con-
and
in the case of a liquid in a voltaic cell, kept from
complete contact with the plate by the disengagement of gas,
is
approximate value,
if
the apparent resistance of polarization.

a/h
is
Its
capable of being taken as infinitely small,
is
h
log2irka ft a
1
The value
of
in 50 can be expanded in a trigonometrical
series so as to enable
comparisons of the value of
to be
made

138
A suitable form
for different values of r.
to
assume
[XII.
is
V=tb cosn^-.
53
llL
From
this
by the equation
oz
which holds
for z
= 0, we
obtain the condition
cotfi
= fi-.
54
This transcendental equation has an infinite number of positive

roots
which are the values of
To
we put
find the expansion
p.
cosh
+ h\
j-
sinh (\z)'
;
Multiplying both sides by cos

to
Ckz)
:
,\
hX sinh (\a)
cosh (\a)
{/u.
(z
= zcu cos
a)ja)
z a
.
/j,
and integrating from
a we obtain from the left-hand side

a?X
2
/*
+\W
and from the right
by the relation
sin
/,
2uA
/,
ha
Thus
54.
ca
2aX
a?
+ Ay
Hence
6
"
rc
2tt/,:
r
^
The expansion
jST
-rica
is
^
Z
+ AV)
5a(a'
rt
fy. (\r) \<fl

M + â
* ~ wJfc (rf + Aa + Ay) J
thus
z-aT e * dk
+ hy
C SfJ
a } 1 V\-l
a! + ha + hW
a*
bb
'
by 30 above.
is
The first root of 54 is smaller the greater A is, the second root
always greater than ir. Thus if r be fairly great the first term
of the series just written
down
will suffice for V.
Hence
has a considerable value at a distance from the axis.
for z
=a
IN
XII.]
139
The solution can be modified by a like process to that used

above to suit the case of a cylinder of finite radius c. We have
in this case a function
to add to
which fulfils the conditions
V
dV
-g =
0, for z
dV
= a, h-^-=
V,
W+ W =
and
hTr:=c
'
for z
0,
'
The reader may
the general differential equation.
satisfies
verify that if in the quantities Lîr), ...,^(6),
be replaced by
...,
of 32 above, k
fija,
V+V
S
irka
s
tf
As a
M^cj^^-Mê)^)
a? + hW
+ ha + h /*
2
final
z-a
^
(c)
b6
and very instructive example of the use of Fourierwe take the problem of the flow of electricity in
Bessel expansions
a right cylindrical conductor when the electrodes are placed on the
same generating
line of the cylindrical surface, at equal distances
We
from the middle cross-section of the cylinder.

sketch the solution, leaving the reader to
fill
in
shall
merely
the details of
calculation.
The
differential equation to
be
satisfied
by the potential in
this case is
3T2
+r
+r
dr
~
2
30 2
57
dz*
If the electrodes be supposed to be small equal rectangular
having their sides parallel to generating lines and ends of

the cylinder, and the radius be unity, the surface conditions to be
disks,
satisfied are
summed up
-=
in the equations
0,
for*=a, gp--*,
where
-<f><

= + cfor

= 0,
e< +
<j)
+/3<^</8+S
_ 8>,>-08 + S)
/
The
for all other points.
distances of the centres of the electrodes from the central
cross-section are here
axis
by
their breadth
is
(ft
2<j>,
+ ^8)
and the angle subtended at the
while the height of the cylinder
is
2a.
STEADY FLOW OF ELECTRICITY OE OF HEAT
140
We
have
conditions.
to find an expression for
first
which
[xil.
fulfils
these
This can be obtained by Fourier's method and the
result is
<&
4c
=
7r
{ d>
=-
(2m+l)ir //0
.J
2^~*
(2m+l)7r
2^ ^ + 8 )} sm
008
ft"
2a
assume
rr
and the
2m +
(2m+l)7r_
cos
Now
oo
+ 2 Z - sin nd>T cos nd} 2
(2m
+1)77-2
^
2S^.^ m,Vr (r) sm

= -cv
cos?.0,
57 will be satisfied
58
ty(r) be a
if
function of r which satisfies the equation
B?+rd-r-p + {-2a-V\ U
= -
Hence we put
f (r) = Jn (i
To complete the
irrj
2a
= Jn (ix).
solution the constant
59
A mi n must
be chosen
as to ensure the fulfilment of the surface condition.

is
so
This clearly
done by writing
.
Am
_ 4c
=
2 sin
^ 2m +1
n(j>
nty' (1)
f
C0S
(2m +
in which
To
when n = 0,
<f>
is
l)7r
2a
P/3 ~ CS
(2
+ 1 )?
2a
to be put instead of 2 sin
find the effect of
making the
/o_L
^+
n<j>/n.
electrodes very small
substitute
(2m+l)7r,
a
2sm
.
2a
for the cosines in
the value
1tN
& + &>**
A m>
sin n< sin
and
{(2m
(2m +
<f>8
4a
l)7r s
*
for
+ 1) irS/4ia}
(2m +
Remembering that
c<j>8 is finite,
4c<8/7r 2
1) 7r/4a
and therefore putting
=l,
j;\l
S)
|
we
XII.]
we get the
solution
a
7=^-2
2a
T,
7T
VM
where
(2m
v
a
-f^-(cosi0sm
en
+ l)<n-/3
2a
Y(l)
B=0n.=o
For an
141
= l,
e!=
e2
=e =
...
= e=
sm
...
(2m +
1) 71-2
=-^
2a
= 2.
we can obtain the
infinitely long cylinder
60
solution by
putting in 60
^=aX
a
(2TO
J-
1K = X,
2a
and replacing summation by integration.
V= J2e cos nd
e
as before being
1,
Thus we obtain
"
.'. sin (X/8) sin (Xs) d\,
I
Jo *A*/ B {i\)
and
61
all
the others
2.
The reader may verify as another example that if the electrodes

be applied at the central cross-section at points for which 6 = a,
the potential
T7-
is
given by
r"
a
ov
=S
22 v
2 sm a sm no+
2
-
in
"('"'
T,
mirz
m sm na sm wt* cos a
/ 1
x
( .T/wrN
If a be infinitely small the second part of this expression vanishes
and the
first
term can be written
V- II
l-2rcos(a + +
W^l-2rcos(a-(9)-|-r
g)
r-*
!!,
which agrees with an expression given by Kirchhoff (Pogg. Ann.

Bd. 64, 1845) for the potential at any part of a circular disk with a
source and a sink in its circumference.
The reader may

electric flow, for
by Weber (Crelle, Bd.

some more complicated problems of
refer to another paper
76, 1873) for the solution of
example a conducting cylinder covered with a

two electrodes
coaxial shell of relatively badly conducting fluid, the
being in the fluid and core respectively
and a
cylindrical core
covered with a coaxial cylinder of material of conductivity comparable with that of the core.
Each of Weber's papers contains a very valuable introductory

most part with definite integrals involving
analysis dealing for the
Bessel
Functions.
Several of his results are included in the
Examples at the end of
this volume.
CHAPTER
XIII.
PROPAGATION OF ELECTROMAGNETIC WAVES ALONG WIRES.
The
field were first given by

been used in a somewhat
by Hertz and by Heaviside in their
equations of the electromagnetic
Maxwell
They have
in 1865*.
modified form with great effect
since
researches on the propagation of electromagnetic waves.
by these writers is important as showing

reciprocal relation which exists between the electric and
magnetic force, and enables the auxiliary function called
modification used
The
the
the
the
vector-potential to be dispensed with in most investigations of

this nature.
If P, Q, M,
a,
y denote the components of
/8,
magnetic forces in a medium of conductivity

capacity
k,
electric
and
k, electric inductive
and magnetic inductive capacity
/i,
the equations
referred to are
+ 4<7rdt) F
4>7r\dy
dz)
4sir\dx
dy.
4<7r
dt
and
fi
da
l_(dR_dQ\\
4nr
dy
dz)
d
^JL( d2L_ J*\

JL d =
~
Vol.
On
ii.
4nr dt
4nr\dz
dx)
47T dt
4nr
\dx
dy)
the Electromagnetic Field, Phil.
Chap. xx.
Tram. 1865,
Electricity
and Magnetism,
PROPAGATION OF ELECTROMAGNETIC WAVES.
XIII.]
From
these
may
be derived the equations
dP
dx
dQ
dx
The
first
point x,
143
afi
dy
+ dz~
dy
dz
'
is no electrification at the
and the second that the magnetic force, being purely
of these expresses that there
y, z,
inductive, fulfils the solenoidal condition at every point, except
of course at the origin of the disturbance.
At the surface of separation between two media the normal

components of the magnetic induction, and the tangential components of the magnetic force, are continuous. The tangential
components of electric force are also continuous.
From
the equations given above the equations of propagation
of an electromagnetic
Q and
(1)
R by means
and
(4),
we
wave can be
of the
first
at once derived.
of (2), the second
Eliminating
and the third of
get
4^ + ^-=Va,
.
32a
3a
and similarly two equations of the same form for and

are the equations of propagation of magnetic force.
By
a like process
we
7.
These
obtain the equations of propagation of
electric force
&c.
&c.
Now
for
the case of propagation with a straight wire as guide
in an isotropic
medium, the
every instant.
Therefore there
symmetrical round the wire at

no component of electric force at
field is
is
right angles to a plane coinciding with the axis.
From
this it
by the equations connecting the forces, that the magnetic

force at any point in a plane coinciding with the axis is at right
The lines, of magnetic force are therefore
angles to that plane.
circles round the wire as axis.
follows
Thus we may choose the axis of x as the axis of symmetry,

and consider only two components of electric force, one P, parallel
to the axis, and another R, from the axis in a plane passing through
it.
We
shall denote the distance of the point considered from the
144
PROPAGATION OF ELECTROMAGNETIC
[XIII.
by x, and its distance from the axis by p, and

magnetic force at the same point the symbol H,
origin along the axis

shall use for the
which
will thus correspond to the
From
and 2 we get
t,
and
2.
our special case the equations
for
^kP +
of equations
^
aR
7
4nrkR
+ ^r
,
--
(pH)
'
= dH
~
ox
at
^S_dR_dP
11
Eliminating
first
dp
da>
H and R from these equations we find
differential equation satisfied
H and P
Eliminating
dt
we
by
for the
see that
satisfy a slightly different equation,
must be taken
so as to
namely,
aiR d*R d*R

ldR 1
+K
*^k BR
di
*W = w + oy + pdp--p> R
,
Finally,
we
same way that

same as n.
easily find in the
differential equation precisely the
In dealing with the problem we

wire has a certain finite radius, and
satisfies
shall suppose at first that the
surrounded at a distance by
is
a coaxial conducting tube which
may
infinity in the radial direction.
There
be supposed to extend
will
to
therefore be three
regions of the field to be considered, the wire, the outside conduct-
ing tube, and the space between them.
The
differential equations
found above are perfectly general and apply, with proper values of
the quantities
Taking
suppose
k,
first
The appropriate
= 0,
we
shall
with a perfectly insulating isotropic substance.

differential equations are therefore obtained
in io
If the electric
respect to x and
each region.
the space between the two conductors
it filled
putting k
k, to
p,,
t,
and
and magnetic
each
by
1 1.
will
forces be simply periodic with

be of the form
{mx~nt)i
f(p)
WAVES ALONG
XIII.]
P = ue {mx~
Let
where
u, v
WIRES.
nt
>
145
i
,
denote the values of f(p) for these two quantities.
Substituting in 10, remembering that k

dht,
d?
ldu
+ -pdp- {m
= 0, we
-^ )^ =
a
find
12
0-
The quantity 2 - /c^n a is in general complex since mi includes

a real factor which gives the alteration of amplitude with distance
travelled by the wave along the wire.
On the other hand n is
essentially real being
If the
2-rr
times the frequency of the vibration.
wave were not controlled by the wire we should have in

m2 k/jlu 2 = 0. The velocity of propagation of an
the dielectric
electromagnetic disturbance in a
according to theory JI/k/i
been proved to be that of

If
we denote
medium
of capacities
k,
is
p,
this velocity has, for air at least,
light.
w*
is
and
m ic/m* by p
3 2 it
which
ldu
and write f
for ppi,
becomes
+u=0,
I3
the differential equation of the Bessel function of zero
order / ().
In precisely the same way we get from
&v
ldv
1 1
the equation
1\
A,
the differential equation of the Bessel function of order
1,
namely
MS)-
An
equation of the same form as 14
is
obtained in a similar
manner for H.
Turning now to the conductors we suppose that in them k
k.
In ordinary conductors x/k
10 -17 in order of magnitude, so that
we may
is
neglect the displace-
ment currents represented by the second terms on the

equations
1.
substituting
We
left
in
thus obtain the proper differential equations by
- iirkiini for p\
We
shall denote this
by
q*
write the equations

G, M.
is
about
10
and
146
&u
drj
du
7)
dt)
[xill.
_-
^ + i^ + fl_I^ =
7]dr)
drj*
where
77
= qpi = pi
16
if)
Qeirkimi.
Two values of q and 77 will be required, oue for the wire and
the other for the outer conductor we shall denote these by qlt r) lt
;
2 , r) 2 ,
respectively.
The general
solution of 13
u = aJ
where a and
as
is
we have
+ bY
(g,
The
same form, with
77
p. 11,
17
(),
6 are arbitrary constants to
conditions of the problem.
seen above,
be determined to
suit the
solution of 15 has of course the
substituted for
For very large values of 77 both Jo (77) and F (77) become infinite,
and we have to choose the arbitrary constants so that u may
vanish in the outer conductor when p, and therefore 77, is very
great.
It
is
clear from the value of
for large values of
Yn (x)
given at
p.
40 ahove
the argument, that this condition can only be
fulfilled if
where y is Euler's constant,

of the argument is positive.
b (7
it
- log 2),
being understood that the real part
Again, for small values of the argument
We
becomes very
great,
0,
thus get for the value of
in the three regions, the wire,
so that in the wire
we must put
since
77
there vanishes.
the dielectric, and the outer conductor, the equations

18
P = AJ (v )e^x-nt
19
P = {BJ (f) + CF ()} "*> \
20
ei*-"" \
P = JD{(y- log 2) J
+Y
A, B, C, D are constants to be determined by means of
>
<)
t (77 2 )
in
which
(77 2 )}
the conditions which hold at the surfaces of separation between

the adjacent regions.
WAVES ALONG WIRES.
XIII.]
From
P we
the value of
can easily obtain the component

at
all the quantities are periodic 8
right angles to the axis.
and 9 may
147
Since
(iirk
Kni) R = miH
iiniH = miR ^r
dp
Eliminating
first IT,
7-,
iS
rr
then R, between these equations we obtain
_
ra 2
Thus
we
if
mi-
=
2
m fiicn
. dP
4>irfikni
21
'
dp
(4nrk Kni)
dP
&TTft,kni fjLKn dp
22
'
we put k = 0, and write p*

remembering that ppi= %,
in the dielectric
get from 19,
-rr
~~
icn
dP
3f
for
m? /wms
ff/V)
= - -^
A7J() + CTtff)} ei-^-f *.
24
In the wire on the other hand where

q x pi
we have
22
971,
[g
=m
2
-l^-^^W*-"^,
27 =
L.
Lastly in the outer conductor
47i>i&m],
we
25
^/'(Tiê *-"" \
1
26
have, writing, as at p. 91,
(va) for
_{( 7 -log2)J-
( 1?a
)+7 (r
?2 )} )
R= dP = _m ])G Me{rm_
,
q2
We
now
0%
nt)
27
q2
introduce the boundary conditions, namely that the
tangential electric force and the tangential magnetic force are
continuous.
magnetic
From
force,
the latter condition
being
dielectric, are so also in
circles
it
follows that the lines of
round the axis of the wire in the
the wire and also in the outer conductor.
102
148
These conditions expressed
BJ
K1l
and
for
(Z)
the surface of the wire give
for
+ CY
[XIII.
(Z)
= AJ
(vi),
ÂJi( Vl
BJl(& + CY^)} =
for
29
),\
=a
+ OF. ()=* -2)0.(1*),
BJ, (0
(f;), F () for p = Oi> an d p = <h by

and
eliminating the four constants
2
ô(?)ii ô(?)i> Jo{%\,
equations
just written down, we find
means
of
the
.4,2?, (7, 2) by
Denoting the values of
F (g)
4?rA;1piJ (7?1 )
/CTfr / fa) J' (g^

- Knq G {-q ) /' ()
47T^ pt/o (^ ) Fo(g)! - itWg! J (%) TpCg)!
4flr^ptG (vd F,(& - ? Go(?? ) F '(?)
(g) t
4nrk2 piGo(r} 2 ) 7o()s
'
Considering
that
we
Kyi is
small.
small,
is
'
long waves of low frequency and remembering
V is the velocity of light in the dielectric,

m nearly, and the real part of m 47r /\
1/F2 where
see that
where A
first
reduces to
Thus
the wave-length.
if
ax
is
is
not large pok
is
very
Also if a 2 the radius of the insulating cylinder, is moderately

,
pa2
is also
small.
Now when d,
a2
are small the approximate values of the
functions at the cylindrical boundaries are
MZ)=h
Jo(v)=l,
^(?) = -i
F (f) = logf,
<W = ~K
Using these values
for the
y"o(?)
and
ff.(ij)
= |,
= -log^,
,'(1,)
--1.
functions in equation 31,
putting for brevity $, x, f r the ratios </ ('?i)/2o'( 17i)> GoOfrVGoOfc).

and <*!, a2 for 4^, 4tt&2 and neglecting terms involving the factors
,
aâ2
<b.a in comparison with others involving the factor aâ^, we
find after a little reduction

icn
1
a2
<7i
laiOi
(ft
a2 a2
*<ftff 2
aiOjOiao
>
WAVES ALONG WIRES.
XIII.]
In
2
|
is
1
cases which occur in practice
all
approximately 4w/*&.
Further
may be assumed
it
ten
149
= n/fiV*, so
term within the brackets in the preceding expression
n2
87r/iF
The
first
47r/i&ini
<f>
It-irk^
and
for small values of a^,
the modulus of the

vibrations
is
VS
first
very great,
a.2>
is
- a?
**
two terms within the brackets
of the other
V/^/a, Og2
2
that
that the last
is
i V
~ V2 tJM^cta'
__ 1
firyin
the values of
<f>,
<f>
x> are large.
Hence
term, unless the frequency, njîr, of the

is
large in comparison with that of the
third term.
The same thing can be proved of the second term
and the third. Hence the third term, on the supposition of low
frequency and small values of c^, a 2 may be neglected in comEquation 32 thus reduces to
parison with the first and second.
,
V^yN
1-i/Vjjj*
0*1
Let now the frequency be

Then we have
X=
and
W^)
so small that
=q
q^
is
very small.
it is clear that the second term of 33 bears to the first only a

very small ratio unless a 2 be very great indeed. In this case then
we may neglect the second term in comparison with the first, and
and
we get
I
n> =
p
fiV* 2ttoA
L_
34
a,'
1q
or,
since
= m icpri* = m2 - / F
2
tf*
2
,
n
m = VH// 1 + Zirfikina,
2
TI.
-~
^
1
o
2
\
\
I
35
150
The modulus
[xill.
of the second term in the brackets
is
great in
comparison with unity, and hence if we take only the imaginary

part of 2 as given by 35 we shall get a value of m, the real part of
which is great in comparison with that which we should obtain if

we used only the real part, that is we shall make the first
Thus we write instead
approximation to m which 35 affords.
of 35
2 n
Og
o,
But
if
r be the resistance of the wire, and c the capacity of the
cable, each
taken per unit of length,

r
(where k
is
= I/ttoA,
= */(2 log ffls/oj)
taken in electromagnetic
units),
and we have
= nr Nnrc,
30
V2
taking the positive sign.

This corresponds to a wave travelling with velocity V2n./vVc,
and having
its
amplitude damped down to 1/e of
its initial
amount
in travelling a distance V2/Vmrc.
The other root of m2 would give a wave travelling with the

same speed but in the opposite direction, and with increasing
amplitude.
We
It is therefore left out of account.
have thus fallen upon the ordinary case of slow
signalling
along a submarine or telephone cable, in which the electromagnetic

induction
by a
may be
neglected, and the result agrees with that found
direct solution of this simple case of the general problem.
The
velocity of phase propagation being proportional to the
square root of the frequency of vibration, the higher notes of a

piece of music would be transmitted faster than the lower, and
the harmony might
from this cause.
damped out with
if
the distance were great enough be disturbed
Further these higher notes are more
rapidly
distance travelled than the lower, and hence the
relative strengths of the notes of the piece
would be
higher notes being weakened relatively to the lower.
altered, the
WAVES ALONG WIRES.
XIII.]
We
can
now
find the electric
motive intensity in the wire
is
P = AJ
where q1 = q1 pi, the
But
the wire.
thin,
Jo (%) = 1,
if
151
and magnetic
forces.
The
electro-
given by
~
(ttl )e^m nt ^
>
'
denoting that p is less than the radius of

the wire is, as we here suppose it to be, very
suffix
P is the same over any crossHence if 7 denote the total current in the
the plane x = 0, when t = 0, we have
and the value of
section of the wire.
wire at
and therefore
Hence
realizing
we obtain
P = y re
The
which
cos
(W
-y-oc-nt).
radial electromotive intensity in the wire
is
38
given by 25,
is
- AJfa) = - yorfa
But
Hi
/1
~ J 1-i pr ,
nr'
2V2
R=
SO that
Again
realizing
we
R = fcy rp
R therefore vanishes
intensity
is
ry
orp nj^gimx-nt)
Jnrce
-Jvc? x
2
1 nrc
cos
a/-~-
at the axis of the wire,
there along the axis.
The magnetic
39
find
the surface the ratio of
its
Elsewhere
00
nt ir\
-r-J
40
and the electromotive

R is sensible, and at
amplitude to that of P
force in the wire is given
is
ax '/nrc.
by the equation [26
above]
H=- ^^ AJ{(
Vl )
<->,
which by what has gone before reduces to
H=- y
t
pe^ mx
-nt i
>
41
152
The
realized form of this is
H= - , Vope
cos
By
a well-known theorem
is
- ntj
-g-fl!
(y
we ought
2irpH=
where 7
[XIII.
42
to have numerically
^ 7,
4tt
For 7 we have
the total current at any cross-section.
here the equation

,
=7
e<
ma!
-n
*>
43
which when compared with 42 obviously
Hence the
tion numerically.
We
0p
shall
now
the required rela-
fulfils
results are so far verified.
calculate the forces in the dielectric.
Putting
p from the axis in

we find by the approximate
for the electromotive intensity at distance
the plane x
= 0,
and at time
values of the functions given at
0,
148 above
p.
B + G\og(ppi) = P
But
at the surface of the wire
P =
0p
09 .
<y r,
where r denotes
Thus
as
before the resistance of the wire per unit length.
B + C log (pctai) = y
r.
Hence subtracting the former equation we

-Pop
find
= 7r-0 log-.
P
can be found from 29 by putting (by 18)
nating B.
A = y r and
elimi-
Thus we obtain
C=2yofj,crV
very approximately.
Hence
,--
P = y r (l - 2fic V
of which the realized form
log
<ffla!-
""
44
is
P - 7.r (1-2^7* log )r^' cos
^/^ x -nt).
4S
WAVES ALONG WIRES.
XIII.]
From
and
23, 44,
36',
since
p=m
153
nearly,
we get
R = (l+i)fiVjoX / !(->*,
V
46
n p
or retaining only the real part

7T\
Arc
/
T72
AC ! -n/'Ô!
v
E = O2^F
cos^-g-fls-nf
+
7oy--e
T>
47
^J.
Finally from 24 and 36
we have
ff=-2ê<ma!-"*>,
48
or
_ff=-2ê v
Thus the solution
is
cos
completed
a/ -9-
as
nt)
49
for slow vibrations in a cable of
small radius.
we have followed with certain modifications the analysis

Thomson, as set forth in his Recent Researches in
Electricity and Magnetism (the Supplementary Volume to his
Edition of Maxwell's Electricity and Magnetism). To that work
So
far
of Prof. J. J.
the reader
may
refer for details of other applications to Elec-
trical Oscillations.
Reference should also be
made
Heaviside's important memoirs on the same

Papers, Vols. I. and 11. passim.
We
shall
now
obtain an expansion of
to
Mr
Oliver
subject, Electrical
xJ (x)/Jq(x)
in ascending
powers of x which will be of use in the discussion of the effective

resistance and self-inductance in the case of a cable carrying
rapidly alternating currents, and which
applications
when^, q^,
Denoting the function

u,
we have by the
also useful in
other
ocJ (<v)/Jq(w) (or, for brevity,
xJ /Jg) by
relations proved at p. 13 above

*J
tt=#-j7
Hence
is
qâ^ are not very small.
*J
= JByl/ = Z+Xyj-
ldu 1
~j~ =udx x
"
Jo
J{
J(,
Jrx
+T
=!-* +
t&-<)
_2
u
x
u'
J,
<' 2 \
SO
PROPAGATION OP ELECTROMAGNETIC
154
[XIII. V;
Therefore
i
Now
guided by the value in brackets in 50 we assume
m= 2
Then
x2
+t +ax +a
i
x6 +
. .
x (| + 4a4 a3 + 6a6 a?+ ...)=a;2
+ (^ + ai x*+
(-2+^ + .-.J
...J
Multiplying these expressions out and equating coefficients we

find
= -2a + &,
6a = -2a + a
8og = 2a + \a + a
lOojo = la^ + a + 2a
4
4,
2
4 ,
111
Hence
ai _
4a4
o
2 .3'
os5
'
a6
13
15 3 .5'""
a6
i
6
,09
o> as as,
022 .5'
e> a 10
2 9 .3'
2 9 .3
2 .3
Thus we have
X
/( _
^,Jl_
+
+ _il_ +
5
Jl(x)~
This expansion
2 .3
2 .3
may
^
9
2 .3 .5
+2
13a;1 "
15
+ ""
.3 3 .5
be converted into a continued
52
.
fraction,
the successive convergents of which will give the value of the

function to any desired degree of accuracy.
be
verified
by the
x
reader,
J<,(x)
The
result,
which may
is
x1
a?
x2
J&)--* + 4ZW^8Z--
S3
Using 52 we obtain
iqiai
Now
_ + 2T3"2
Jîqiai )-~^
approximately qf =
i-n-fi^^ni so
îai J^^)-"t
2^
+l
\
T3
--
that
2 9 .3 2 .5
273
"
'
+
54
WAVES ALONG WIRES.
XIII.]
The
following table of values

3
4 7r/A1 Jia1 &l

,
is
given by Prof.
155
J. J.
Thomson
156
Jn (ix)=in In (x)
prove, from the relation
semiconvergent expansion 143,
p. 68,
(139, p.
[XIII.
66),
and the
that approximately
'
Jo(v)~
when 7] = ix, and a? is a complex quantity of which the
positive, and of which the modulus is large.
Again
it is
part positive,
worth noticing
we may
write by
that, for
p.
real part is
any value of x with
real
90
Gn {ix) = in Kn (x)
i"\/ -<*>* mr.e-"
by
142, p. 68, if the
modulus of a; be very
55
large.
Similarly
G'n (ix)
= i^K'n (x) = i>" \K^{x) -
v cos
**
approximately,
if
GUix)
ft
x have a very large modulus.
Gn {ix) _
of which
{rj)jG'a {r))
=-
i,
is
(*)}
- 1 ) 7r
e_a;
ss'
Thus we get
56
a particular case.
In the same circumstances
Jn (ix) = in In (x)
2^'
57
and further
J'n {ix)
when the modulus
= i^ iln^(x) - 1 In (x)\
of
is
very great.
Tp, =
already stated above for n =
0.]
Thus we have the

i
result
58'
WAVES ALONG WIRES.
XIII.]
As a
157
may work out the case of

both q1 a1 and q1 al are very large. Here
further illustration the reader
oscillations so rapid that
so that by 33
Thus
K2
K2 J
SWi^
V?xi
and approximately
m=n
The
velocity
JObm
of
travelled, while the
of its original value,
W aA
propagation
amplitude
aAJ
thus V,
is
59
oi
and the distance
diminishing to the fraction 1/e

the product of n by the reciprocal of the
is
is
coefficient of i
within the brackets.
slow, since the
imaginary part of
The damping
is
much
of
in this case is
smaller modulus
a2 -k.2 be small compared with a^,

by sea water, the outside
conductor will mainly control the damping, and nothing will be
gained by using copper in preference to an inferior metal.
than the real part.
Here
if
as in the case of a cable surrounded
We
shall
now
calculate the current density at different distances
from the axis in a wire carrying a simply periodic current, and the
effective resistance
conductor.
and self-inductance of a given length
Everything
of the
supposed symmetrical about the axis of
is
the wire.
By
we have
for the axially directed electromotive intensity
at a point in the wire distant p (= n/iq) from the axis
P = AJ
(r) 1 )ei
mx - nt
>
6o
This multiplied by k lt the conductivity of the wire, gives an expression for the current density parallel to the axis of the wire at
distance p from the axis.

If the value of
P at
the surface of the wire be denoted by
Pa = AJ (Vl )e^
^
(p
= i).
x-nt
>
i
,
tti ,
6i
158
The magnetic
force at the surface is
=-
ai
Â J
m*-**>\
'(vi)e<
= Oj). Therefore if T be the total

47rr = 27ra i7 0l and so
r = ^^' AJ
current in the wire
(p
we have
'
The
[XIII.
electromotive intensity
e*-^ \
62
the resultant parallel to the
is
axis of the impressed and induced electromotive intensities.

To
solve the problem proposed we must separate the part impressed
by subtracting from
electromotive force
Now
the induced part.
the same
is
the impressed
over any cross-section of the
all
wire at a given instant, and will therefore be determined

find
tensity
But
the surface.
it for
if
we
since the induced electromotive in-
due to any part of the current
directly proportional to
is
its
time-rate of variation, the induced electromotive intensity at the

surface must be directly proportional to the time-rate of variation
Hence by 61
of the whole current in the wire.
if
denote the
impressed electromotive intensity
E-AT=P
=a
where A't (A'
constant)
is
put
parallel to the axis at the surface.
E = A U(%) - n
Putting r
for
^^
12
180
r2
{-* +
/il
Or taking the impressed

two ends of a length
we have
K_jK
V
.
the induced intensity
for
Thus
A'Ji( Vl)\ !-> <
the resistance (= 1/TraA) of unit length of the
wire and using the expansion above,
1*
we get
+ -)
since qf
= 47r/A
&1 ra,
U 48^~ + 8640^~~-J} r
difference of potential
64
between the
of the wire the resistance of which
1^_J^/^
& + -)
is
12
...
R>
180
(I
u^nH*
13
ufn'P
\)
WAVES ALONG WIRES.
XIII.]
159
If we. denote the series in brackets in the

respectively
by
R', L'
first 'and
second terms
we get
V=R'V + L't.
Thus R' and
of the length
It
66
I! are the effective resistance and self-inductance
of the wire.
remains to determine the constant A'.
If there be no dis-
placement current in the dielectric comparable with that in the

wire, a supposition sufficiently nearly in accordance with the fact
and the return current be capable of
for all practical purposes,
being regarded as in a highly conducting skin on the outside of the

dielectric, so that there is no magnetic force outside, we can find
A' in the following manner. The inductive electromotive force
per unit length in the conductor at any point is then equal to the
rate of variation of the surface integral of magnetic force taken
per unit length in the dielectric at that place.
Now,
if
there
is
no displacement current,
will be in circles round the axis of
the wire, and will be inversely as the radius of the circle at any
point, since
2-irrH
Thus
if
= - 47rr.
be the magnetic force at distance r from the axis of
the wire
f^Hrdr = - ^2^2 AJ^ ( Vl )
and
But
this last expression
given by 62.
is
^ e<"*>
log
by what has been stated above
67
*.
is
AT, and
Thus we obtain
^'
Z-log- +
and
If
and
a;
= -21og^
^- g^ + _^-...).
i
J (at V*) be denoted by X - Yi, BX/dx,

= 2 V/^w/r, then by 63 we easily find
t
dYjdy by X',
68
Y',
xXYj-XJ
2
X'2 +
F' 2
'
160
aa
Lti = 2ll ^ +
=17
xlr
XX' +YY'
X*+Y*
and L are
2n
a form in which the values of R'
[XIII.
7
'
easily calculated for
any given values of x and n from the Table of

the end of the book.
(x Vi) given at
Equation 65 shows the effect of /^ on R' and II at different

If however the frequency be very great, we must
frequencies.
put in 63
(yi)/Jo (vi)
R'
Thus
i-
We nnd for this case
= J^MR, L'^^^-IA'....
in the limit R'
is
indefinitely great,
and L' reduces
constant term
IA'.
infinitely thin
stratum at the surface of the wire.
The current
now
is
71
to the
insensible except in an
The problem of electrical oscillations has been treated somewhat differently by Hertz in his various memoirs written in
connection with his very remarkable experimental researches*.
He
an unlimited dielectric
iti
and magnetic disturbances from a vibrator
consisting of two equal plates or balls connected by a straight
discussed
medium, of
first
the propagation,
electric
wire with a spark-gap in the middle, and, secondly, the propagation
same medium of disturbances generated by such a vibrator

guided by a long straight wire. The action of the vibrator simply
in the
consisted in a flow of electricity alternately from one plate or ball
up by an initially impressed difference of potential

between the two conductors.
to the other, set
Taking the simple case first as an introduction to the second,

which we wish to give some account of here, we may take the
vibrator as an electric doublet, that is as consisting electrically of
two equal and opposite point-charges at an infinitesimal distance
apart, and having the line joining them along the axis of x, and
the origin midway between them. It is clear in this case that
everything is symmetrical about the axis of x, that the electric
through the axis, and that the
round the wire.
forces lie in planes
force are circles
The equations
* See Hertz's
lines of
magnetic
of motion are those given on p. 142 above.
Untersuchungen
iiber die
By
Ausbreitung der elektrischen Kraft,
A. Barth, Leipzig, 1892; or Electric Waves (the English Translation of the same
work, by Mr D. E. Jones), Macmillan and Co., London, 1893.
J.
'
'
WAVES ALONG WIRES.
XIII.]
symmetry the component a of magnetic

and the equation
161
force in the
medium
is
zero,
dy
two components.
holds, connecting the other

fidz
ydy
is
dz
This shows that
a complete differential of some function of
Hertz's notation
we take
P
The equations
dt
In
?2
~dydf
become then
dP ^
d_
"
(dm
dt\dy*
9Q =
dzdt'
of motion
"
y, z.
this function as dlljdt, so that
a 2 n\
dz')'
dtdxdy'
dt
dm
<m_
dt
dtdocdz
which declare that the quantities

K
df
dz'
"y
'
dxdy
KK +
dxdz
The propagation of waves in the medium

we suppose each of these quantities
have the value zero. Thus we assume as the fundamental
are independent of
t.
therefore will not be affected if

to
equations
/dm
kP=
(w
dms
~d*)'
dxdy
3*11
dxdz'
Using these
in the equations of
dz \ dt 2
(*n
dy V
dt
which show that the quantity

G. M.
Kfi
magnetic
force, 2,
we obtain
J_v.n)-o,
k/jl
in brackets is a function of
:>-
11
and
162
[XIII.
Thus we write
only.
f-^n=/(M).
we may put f(x,
It is easy to see that
the electric and magnetic

tion
without affecting
we have supposed
solution adapted to the vibrator

II
where r
is
t)
and the equation of propaga-
is
<E>
fields,
the
is
= sin
nt),
74
the distance of the point considered from the origin, and
maximum moment
From
(mi
is
of the electric doublet.
and magnetic forces are found

In cylindrical coordinates x, p,
the equation
this solution the electric
by differentiation.
becomes
dm _ j^
di?
/<m
Kfi
dm
an\
dp
dp)
2+
7S
Here p 2 = y 2 + z*, and hence if we put

now P and B for the axial and radial components of electric force,
we must in calculating them from II use the formulae
since II
is
independent of
0.
kF =
76
We
take the meridian plane as plane of
magnetic force
identical with
H which
/3.
is
x, z, so
that the
at right angles to the meridian plane
is
Thus
*--%
The fully worked out
results of this solution are very interesting,
any applications of Bessel functions, we

detail.
We have referred to them
inasmuch as the case of the propagation of waves along a wire, for
the solution of which the use of Bessel functions is requisite, may
be very instructively compared with this simple case, from which
it may be regarded as built up.
but, as they do not involve
do not consider them in
In the problem of the wire we have II at each point of

the medium close to the surface of the conductor a simple
WAVES ALONG WIRES.
XIII.]
163
harmonic function of the distance of the point from a chosen origin.

We shall suppose that the wire is very thin and lies along
the axis of x, and is infinitely extended in at least one way so that
there is no reflection to be taken into account.
Hence
at any point just outside the surface

II
If
we
of
= A sin (mx nt + e).
we exclude any damping out
see that
of the wave or change of form
cannot involve x or
t; it is
therefore a function
Thus
p.
II
(nix
=f(p) sin
nt + e).
78
Substitution in the differential equation which holds for the

gives for /the equation
medium
^+J|-(
Here
shall
ra
/m2
2
ic/j,
79
dp*
by
is satisfied
therefore instead of 79
pdp
(ipp)
PJ
and by G(ipp) where pp
is real.
latter solution only is applicable outside the wire, as /must be
zero at infinity.
We
where
G and
Now by
IT
have therefore in the insulating medium
= 2GG
II
(ipp) sin
( - nt +
80
e),
e are constants.
140, p.
= 2(7
67 above, this solution

cos (pp sinh
-j
Putting p sinh
<j>
= fj,we
<f>)
dtp
may
sin
be written
(mx
nt
n=GI
e).
get
n = 2C C -îL d^. sin (mx- nt +

or
We
positive, that is
is
than that of free propaga-
is less
We have
tion in the dielectric.
This
= 0.
by p2 and suppose that p*
that the velocity of propagation
The
the square of the velocity of propagation.
is
nt- n
denote
m2 - w
^L d?
sin
e),
(mx-nt +
81
e).
112
PROPAGATION OF ELECTROMAGNETIC WAVES.
164
may
This result
p. 162,
from which
When ipp
be compared with that obtained above,

of course capable of being derived.
74,
it is
small,
is
[XIII.
we have
e
Go(w) = -log -|*,

so that neglecting the
imaginary part we have
Go
Hence
p = 0,
the solution
may
+ lgf)-
=-2C( 1+ log 2) sin (mx-nt+e).
that
is if
the velocity of propagation
is
82
that of light,
is
II
as
(7
at the surface of the wire

II
If
(w) = -
easily
C log p sin (ma nt+
e),
83
be verified by solving directly for this particular
case.
In
up
all
cases the
wave
at
any instant in the wire may be divided
into half wave-lengths, such that for each lines of force start out
from the wire and return in closed curves which do not intersect,
and are symmetrically arranged round the wire. The direction of
the force in the curves
is
reversed for each successive half-wave.
When p = 0, the electric force, as may very easily be seen, is

normal to the wire, and each curve then consists of a pair of
parallel lines, one passing out straight to infinity, the other
returning to the wire.
CHAPTER
XIV.
DIFFRACTION OF LIGHT.
Case of Symmetry round an Axis.
The problem
here considered
is
the diffraction produced by a
small circular opening in a screen on which
falls light propagated

waves from a point source. We take as the axis
of symmetry the line drawn from the source to the centre of
the opening; and it is required to find the intensity of illumination
in spherical
at
any point
of a plane screen parallel to the plane of the
opening, and at a fixed distance from the latter.
Let the distance of any point of the edge of the
orifice
from the
and consider the portion of the wave-front of radius a

which fills the orifice. If the angular polar distance of an element
of this part of the wave-fronb be 8, and its longitude be <, the area
Putting f for
of the element may be written a 2 sin 6d&d<f>.
the distance of this element from the point, P, of the screen
at which the illumination is to be found, regarding the element as
a secondary source of light, and using the ordinary fundamental
source be
a,
formula,
we
an ether
particle)
obtain for the disturbance (displacement or velocity of
produced at
P by this source
a sin ddddd)
z.
where
m = 2w/\,
n=
2ir/T,
sin
(m
the expression
..
nt),
X and T being the
length and period
of the wave.
Thus,
1,
if
the angular polar distance of the edge of the orifice be
the whole disturbance at
r T ?\
Wo
Jo
sin
is
sin
(mf -
nt) ddd<j>.
+
166
Let f be the distance of

the distance of the screen
from the element

of the
is
illumination
from the axis of symmetry, and
from
the spherical wave of radius
that the distance of the screen
a, so
that the longitude of the point
P is
= {b + a (1 - cos
Since 6
Because of the symmetry
we may suppose without
from the element to
the nearest point or pole of
cos 0) + b.
(1
[XIV.
is zero.
loss of generality
Then the
distance
given by
2
6>)}
+ (a sin
cos
2
</>)
+ (? sin
3
<.
small this reduces to
is
= 6 + 4a (a + b) sin - - 2af sin

2
#+
cos
2
,
or
,
2a(q
+ 6)
?
2b'
If
sin 2
now we
^=
write p for a sin

/4a 2,so that
or a0,
0,
2
/3
Hence finally if the opening be of so small radius r, and P

be so near the axis that we may substitute 1/6 for the factor
l/>
we
obtain for the total disturbance the expression
ixioVo
sin
+a
&-j ^ ^p')- nt}pdp^
Separating now those terms of the argument within the large

brackets which do not depend upon p from the others, and denoting
them by -us, so that
we may
write the expression in the form

1
2w
f
or
rr (C sin
ST
+ S cos ot),
'
XIV.]
167
where
cos
T (iSF p*~l p cos *) ****'
sin
/o
8=
The
foil
intensity of illumination at
1
and
"*
5'
cos
*)
P is thus proportional
8.
of Bessel Functions.
C we
Changing the order of integration in
G=
T CS?
008
II {Jo*
Now
G and
p*
have
p cos +) d +} pdp
considering the inner integral and writing

2tt
+b
2tt
T"2^ = **'
Ti'
we have
/2tt
COS (î|r
* COS
Jo
= COS ^l|r
<f>)
d(f>
COS (* COS
<)
d$,
cos
dcj>
= 0.
Jo
since
f2ir
sin î|r
sin
(a;
cf>)
JO
But
fir
(2?r
cos J^r
cos (x cos
<j))d<f>
= 2 cos |i/r
Jo
cos
p.
(as
cos #)
jo
= 2w cos ^yjr
by 44,
(%),
18 above.
Hence
^ = S- yi
cos
where z denotes the value of x when p
This
Lommel * depending
can be done by the following process due to
upon the properties
to
CO2 + S2 )
only remains to calculate the integrals
it
'^
(*')
dx
>
= r.
* Abh. d. k. Bayer. Akad. d. Wissenech. xv. 1886.
d$
p
168
we can show that

6 2 \2 f*
Similarly
S = s
sin |-f
xJ
These integrals can be expanded in
[XIV.
(x) dx.
series of Bessel Functions
in the following manner.

First multiplying 19, p. 13 above, by
n
x and rearranging we obtain
,
*./"_! (x)
= nxn~*Jn (x) + xnJ'n (x),
and hence by integration
xnJn..1 (x)dx
= xnJn (x).
Jo
Integrating by parts and using this result

cos ^jt
The same
xJ
(x)
process
dx
= cos ^yfr xJ
.
may now be
(x)
Thus putting \y
for
the
when x = z, and writing
\-ty
+1
U = y-J
get
repeated on the integral of the
second term on the right and so on.

value of
we
(z)
- g)V, (*)+...= S (- 1)
(f
Jm+1 (*),
&,=g)V,(*)-g)V (5)+...=s(-i)"(5"
4
we obtain
finally,
putting 47ra V 8/6 a \ a for
H,
^- + .(4
z'\
The values
series
U U
of
G and $
can thus be found by evaluating the
the given value of
This can be done easily

z.
1
2
by the numerical tables of Bessel Functions given at the end of
,
for
this volume.
The series Ult U2 proceed by ascending powers of yjz. Series

proceeding by ascending powers of zjy can easily be found by
a process similar to that used above.
We begin by performing
the partial integration
first
upon cos
^yjr
xdx, and then continuing
DIFFKACTION OF LIGHT.
XIV.]
the process,
making use
of the equation
^ (ar Jn
which
is
169
= - ar Jn+1 (x),
(w))
in fact the relation
stated in 16, p. 13, above.
Thus remembering that

62
we have
a+
as the first step in the process
G=
b 2 X2
Jo () COS
//.a;
#C&e
Jo
6 2 A.2
1
r- sin pz*
27r? 2 ,,*"-'62 \2
1
sin
/u-^
2^(2/14
pi
J (z) + zr"^'
2/*
.
-J
(x)sm
llx2
'
2
.
J" (z)
v 7
11 j
Ji (2) cos
/u.2
1
.
xdoo [
Proceeding in this way we obtain
[^{'.-e)"* w+
+
8l^
yl2y
22
sin
2.v
Ay
...}
+ -}_
cos Ay
T7.
\v
^.v
where
2
Fo=j (^)-g) j (^)+...=2(-irg)

2
r = f J, (,) x
g)
w+
Vm (^
n+1
. . .
= 2 (-
1)
sin
J +1 (*)
(~f
Similarly we obtain
22
cos Ay
\y
2y
Ay
\y
ir
170
Comparing 4 and
[XIV.
with 6 and 8 we get
y = z + V
1
cos \y
+ U
sin
|y
sin
sin
U* cos y
sin
cos
\y
- V
y -
cos
cos \y,
sin ^y,
which give
Ux+ V = Smi{y +
x
^y
- U,+ V = cos \(y+-)

Squaring 4 and
and 6 and 8 we obtain equivalent expressions

point P on the screen;
for the intensity of illumination at the
thus
if Trr*
=1
*k-{
+ 7 + vl-*r.aM i(y +
'.
-27,8ini(y+0.
The
calculation of these
U and V functions by
means
y
10
of tables
of Bessel Functions will be facilitated by taking advantage of

certain properties
which they
We
possess.
follow
Lommel
in the
following short discussion of these properties, adopting however a
somewhat
different analysis.
Consider the more general functions

+2
un = g)" Jn 0) - g)" Jn+2 () +

= 2(-l)*g)
...
Jn+ip (z),
11
+1
Vn
n (z)-(^J
=(^fj
Jn+2 (z)+...
= s (- iy
(|)
where n may be anypositive or negative

First
of all
is
it
all
values of y and
we
get
z.
clear
Now
l-=Jl(z)
-/^co,
12
integer.
that the series are convergent for

if
in 75 p. 29 above
2Jl(z)
we put
2Jl(z)+....
x=n = 0,
XIV.]
Heuce we
see that since
Functions must be
(z)
<
than 1/^2.
less
171
each of the other Bessel
It follows that if
series for
Un is
more convergent than the geometric
and
<
Vn
if
z/y
1,
y/z<
more convergent than the geometric
is
the
series
series
z \n+ap
y)
It is therefore
more convenient
in the latter to use

If
Vn
in the former case to use
Un
purposes of calculation.
for
=z
= J ~J + J -...
=
==F ,/ - / + J -...
U=V
ETI
But putting
in 42
and 43,
p.
18 above,
= 0, we
<f>
find
= J (z) 2J (z) + 2J4 (z) ...

=
sin z
2 J (z) - 2J (z) + 2J {z)~ ...
cos z
Therefore
when
=y
U = V = {Jo (z) + cos z],
U = V = \ sin 2,
= V = \ {J (z) cos z],
[7"
2
and generally
C^ = Vm =
{J (*)
-"
-
(- 1)+* / (*),
p=
(-1)
t^ 971+1
+ cos z\
Bin*- 2
rgn-Ul
2+l
Returning now to
1 1
and
(-
we
Un + Un+i =
13
l)n+P
Jw+i
(*)
easily find
(^J
Jn {z),
14
Z n
Vn +Vn+2 = ( -) jn (z)\
and therefore
Un+2 = y Vn + Vn+2
z*(Un +
Also since J_ n (z)
the second and
first
= ( 1) Jn (z), we
of
successively,
find,
and
putting
also z
Un + Un+ = (- 1) ( F_ + F_n+a
Fm + y +a = (-l)(tU + /_n+2
,
15
).
=y
),
).
for
n in
172
Differentiating
1 1,
we
[XIV.
find
+ (f)V;w-g)"*V+ ,(2 ) + ....

Using in the second
line of this result the relation [16, p. 13
above]
71
JL{ z)=-Jn(z)-Jn+i{z)
a
we
get
= --
16
#+i-
This gives by successive differentiation the equation

~
m
3m 2
d Un _ _
_ z d^
Un+1
m~* n+1
m-1
&STwe
Similarly
[19, p.
ydz^=i
dz
I7
obtain by differentiating 12 and using the relation
13 above]
J'n
00 =
>/n-l
4
00 - T
z
O0>
9_F_
y
dz
and therefore
dm
Again
differentiating the first of 9,
But by 16 and 18
this
-^
Differentiating again
_
or
Vn _ m-1 d^_ v
g^m-2
S ^m
z 3- 1
g^ri Vi-
we get
becomes
+F = cos \{y +
we
-V
obtain
SJTa
9_F
a*
a*
tf 3
+F_ = -
_ sin
=
"
.
?/
^
2/
by 16 and 18
1
ini(V +
i).
19
=
XIV.J
By
repeating this process
we
clear that
it is
173
Um+1 + F_m+1 = (- 1) sin 4 (y +
shall obtain
"A
20
- Um+z + V-m = (-
i (y +
we put n =
equations
these
If in
1)" cos
0,
-)
we
fall
back upon
9.
Putting in 9 the values of the functions as given in the defining

equations
<
s <-
1 1
and
we
obtain the theorems
^ Kin + sn ^
* {*
1)
21
2" +2
2 (- iy
sin
j(f )
gf
/*
(*)
= ^0 (*) - cos I (y +
which include the equations
= 2/j (z) - 2J
cos z= J (z) 2 J"
sin z
+ 2J
(z) + 2J
(z)
(z)-...
(z)
as particular cases, those namely, for which
By
Taylor's theorem
y=
...
z.
we have
Calculating the successive differential coefficients by means

of 16,
and rearranging the terms we obtain
TJn (y,z
h)=Un
"
Similarly
g^~
U n+1 f
t/+2-.--
n+J"
pi(2y)"
we can prove that
Vn (y,z + h) = t
p\
(2y)1
Vn^.
23
These expansions are highly convergent and permit of easy

calculation of
Un+S!
...
Un (y, z + h), Vn (y, z + h).
V^,
F_2
...
functions
Un Vn
Un+1 Un+2>
,
by using 16
and then deducing the others by
Un+1 V^,
and 18 to calculate
The
can be found from
successive applications of 14.
Differentiating 11
and 12 with respect
to y,
and using
resulting expressions the relation 20, p. 13 above, namely,
nJn (z) = \z Jn_
(z)
\z
Jn+l (z),
in the
'
174
we
[XIV.
find
24
Now
if
it
be a function of y we have
^P
Using
this
= (--l)-9^+2m2/^- + ^
1
theorem we find by successive differentiation
+ (m-l)m(^-^=r
*
dym
dy m+l
+f
-i-^ -J,
dy
+ (-l)(-a^r+i-9^rjIf
we
26
consider y as a function of z then
dz
If
25
dz
dy dz
= cz
"-
dz
% (c Z7"_i
Un+1 j
by 16 and 24 above. By successive differentiation, and application

of this result we obtain
and generally
PP,_
TO
(m-l)...(m-j) + l)
,r
XIV.]
Similarly
it
175
can be shown that
_ n 4- n
28
The calculation of the differential coefficients can be carried

out by these formulae with the assistance of 14 which now
become
Un +Un+a = cnJn (z)A
We conclude this analytical discussion with some theorems in

which definite integrals involving Bessel Functions are expressed
in terms of the U and V functions.
By
above we have
6
p cos
G= <jp\
i'f xJ
2
\2
fcrp,.
Now
let
= zu,
x) dx
then
x
\^ = \y~ = \yu\
therefore since z*
= 4:7r ^'r /\-b1
G = 2-irr2
Similarly
we
cos (\yu*)
uJ
(zu) du.
30
obtain
8 = 277T2
sin (\yv?)
uJ
(zu) du.
But equations 4 and
5 give
7rr2
Ccoa
y + Ssm%y = ^ Ui,
Csm^y- 8 cos hy=TZ U
*>
and these by 30 and 31 give the equations

I
Jo
Jq (zu) cos \y (1
m ) udu = - U
u") udu = - Ua
(zu)
sin
\y (1
[XIV.
176
we
Differentiating with respect to z
j;
if)
=-
J, (*) cos \y (1
Jx
-)
get, since
(*),
v?du
=-
-^
and similarly
Ji (zm) sin
Now
- u?)
(1
u*du
= Us
we assume
if
C1
~1
J^-x (t). cob Jy(l -)"**
1 fz\ n
-y
and
i7
making use
differentiate,
&
33
of the relation
n1
we
easily obtain
1
Jn (zu) cos \y (1 - vf) u^du = - g)
CT +1 .
for
Thus
n + 1.
if
holds for
the theorem 33 hold for any integral value of n it holds

But as we have seen above it holds for n = 1, it therefore
integral values of n.
all
we obtain
Similarly
1 iz\ n
J^(zu)smy(l-u*)tin- du = -( 7
1
The values
f
of
C in
T
z2
= -sin
cosiy 2 .M /o(2w)cfo
v
>
'
Similarly those of
Jo
34
6 and 30 give
.
io
Un
y \yi
Jo
~2
7-
If instead of
radius of the
in 8
7
sin
Aw
^y
cos
T,
35
ty
and
T,
M pr
\y
give
&
cos ^7
2y
T7-
sin
\y
Tr
/c
u we use the variable p (= wr) where r is the

and write y = kr2 z = Ir, we have instead of
orifice,
35.36
!
Jo
(fy>)
(fy>)
cos (P/3 2 ) pdp

.
rr
I
sin
= t sin
I
(pp ) pdp = cos

2
=r
I
1
+ t sin (i&r ) F - t cos (pr ) V
2
^-^
cos
(Pr2)
- ^1 sin (^r ) V,
2
XIV.]
If
now r be made infinite while

and we have
177
and k do not vanish,
and
Vi vanish,
p) cos
J"
(lp) sin (fy> 2 )
Jo
f
J^
(iV) -P dP = l
ô
shl
2k
37
= ^ cos ^,
-pdp
formulae which will be found useful in

special cases of
by successive
We
what follows. They are

more general theorems which can easily be obtained
differentiation.
come now to the application of these results to the problem

Of this problem there are two cases which may be
stated above.
distinguished, (1) that in which y
The
vanish.
first
case
much
mena, and has received

specially
= 0,
which y does not

that of Fraunhofer's diffraction pheno-
is
(2) that in
We
attention.
and afterwards pass on
here,
to
shall consider it
the more general
case (2).
When
= 0,
either
= 00
and
former case the wave incident on the
on which the light from the
parallel screen
great distance from the

points out,
00
realised
is
orifice.
a=
or
orifice
is
b.
In the
and the
plane,
orifice falls is at
when the
a very
Lommel
This arrangement, as
phenomena are
and collimator of
interference
observed with a spectrometer, the telescope
which are adjusted for parallel rays.

The orifice is placed
between the collimator and the telescope at right angles to the
parallel
beam produced by the
When
a
is
a=
negative the orifice
converging
with
b, a may
its
to
former.
be either positive or negative.

is
When
to be supposed illuminated by light
the point-source, and the screen
is
there situated
plane at right angles to the axis of symmetry.
This can
be realised at once by producing a converging beam of light

by means of a convex lens, and then introducing the orifice
between the lens and the screen, which now coincides with the
focal
plane of the lens.
When a
falls
on the
is positive,
orifice,
with
and therefore
its
b negative, the light-wave
front convex towards the direction
The interference is then to be considered as

of propagation.
produced on a screen passing through the source, and at right
angles to the line joining the source with the centre of the orifice.
G.
M.
12
178
[XIV.
This case can be virtually realised by receiving the light from the
opening by an eye focused on the source.

is
then produced on the retina.
The
diffraction pattern
Or, a convex lens
may
be placed
at a greater distance from the source than the principal focal
distance of the lens, so as to receive the light after having passed
the
and the screen
orifice,
in the focal plane of the lens.
The screen may be examined by the naked eye

magnifying
If a magnifying lens is used the
lens.
or through a
arrangement
is
equivalent to a telescope focused upon the point-source, with the
opening in front of the object-glass.
This
is
Fraunhofer's arrange-
ment and we shall obtain the theory of the phenomena observed

by him if we put y = in the above theoretical investigation.
;
Putting y =
so that, writing
in
M*
3,
for
we have
C2 + S*, we
obtain
if ={Ji (*)}"
Airy gave*
38
the same quantity the expression, in the present
for
notation,
z*
which
is
z*
+
274 2.4.4.6
Z*
2.4.4.6.6.8
"".
simply the quantity on the right of 38.
By means
of Tables of Bessel Functions the value of
M can
be found with the greatest ease, by simply doubling the value of
and dividing the result by the argument.

shown graphically in the adjoining diagram.
Ji for any given argument,
The
result is
The maxima
J-L {z)jz is
of light intensity are at those points for which
maximum
points for which
(z)
or a
minimum.
The minima
= 0. Now when Jx (z)/z
minimum
a
{^(*)} = o.
Camb. Phil. Trans,
p. 283, 1834.
is
are those
maximum
or a
XIV.]
But
^)-^> = -^<4
so that the condition
becomes
which
(20, p.
13 above)
is
(z)
= 0,
equivalent to
179
180
is
(m
+ 1) ir,
and of
Hence
of the root.
(z)
= 0,
is
(m + f)
[XIV.
where
77-,
for great values of z
m is the number
the difference between
the values of z for successive maxima or minima is approximately

it, and the difference for a zero and the next following maximum
is \tt.
The
The
rings are thus ultimately equidistant.
difference of path of the rays from opposite extremities
of a diameter of the orifice to the point
2r/b or \zfir.
The
distance in wave-lengths
is
is
2r tan-1
%/b,
therefore
that
is
zjir.
following Table gives the values of z corresponding to
maximum and
col. 2.
The
Col.
zero values of 2z~1
(z),
which are contained in
3 contains the corresponding values of
1
.
181
XIV.]
The whole
light received, within a circle of radius z is pro-
portional to
I*
M*zdz
= 4 P r*J\ (z) dz.
But
j^J'M-j^j'^z).
Hence
f M*zdz = 2(1- J% (z) - J\ (*)}.
39
If z is made infinite the expression in the brackets becomes 1.

Hence, as has been pointed out by Lord Rayleigh* the fraction of
the total illumination outside any value of z
But
at a dark ring
outside any dark ring
the successive roots of

"047,
. .
.
so that
= 0,
(z)
J\
is
(z)
is
J\
(z)
+ J\ (z).
so that the fraction of the whole light

(z).
= 0,
The
values of this fraction for
are approximately "161, "090, '062,
more than -^ of the whole
light is received within
the second dark ring.
In the more general case of diffraction, contemplated by Fresnel,
is
and
not zero, and
ETj,
we have
can be calculated by the formulas given above from the
Tables of Bessel Functions at the end of the present volume,

equation 3 being used
and V,
if
if
> y, and
9,
with the expansions of
The maximum and minimum values

Table below
for the values of
stated.
of
We
and of
are given in the
also give here
diagrams showing the forms of the intensity curve

values of
< y.
y.
The curves
are
for the
drawn with values of z
some
same
as abscissas,
as ordinates.
* Phil. Mag., March, 1881;
Edition, p. 433.
or
'Wave Theory
of Light,' Encyc.
Brit.,
9th
182
[XIV.
XIV.J
y
2-649454
+ 0-067178
3-831706
7-015587
+ 0-068485
+ 0068964
+ 0045384
10173467
-0017711
4-431978
183
y
o
- 0-010782
0004513
0004806
+
+ 0-048204
0007931
0-004756
0-076624
0-002637
184
are values of z which give a maximum or
The values
of
are obtained
2U /y,
1
by
[XIV.
minimum
of illumination.
2lT2/y which correspond to these values of
interpolation from those of
JTi,
i72 or
V V
z,
for
the values of z for which Tables of Bessel Functions are available.
The
formulae of interpolation are 22, 23 above.
The maxima and minima which
arise
through the vanishing of
Supposing
are found in a similar manner.
the roots of JJ2 the tabular value of

,
(z)
it is
required to find
nearest to a zero value
which causes U2 to vanish is found

by means of the expression on the right of 22 equated to zero, with
2 put for n, that is from the equation
is
+h
taken, and the value of z
+
frw^^-^w+n
,
TT
(2z
Since the series

retained
is
h)
1 h? (2z
+ Kf
(fry
*w -TT
43
very convergent only a few terms need be
and the value of h (2z
+ h)/2y
found,
and therefore that
ofh.
Values of z which render

2 CTi/y for the
Z72
0,
same arguments are
being thus found, those of
calculated.
The squares
of
which correspond to the roots of U2 = 0.

Elaborate Tables, each accompanied by a graphical representation
these are the values of
of the results, are given
by Lommel
in his
memoir.
The
short
Tables with the illustrative diagrams given above will serve as

a specimen.
We
conclude the discussion of this case of diffraction with
an account of an interesting graphical method of finding, for

different values of y, the values of z which give maxima or minima.
This is shown in the next diagram. The axis of ordinates is that
of y, and the axis of abscissse that of z.
Lines parallel to the axis
of y are drawn for the values of z which satisfy Jz (z) = 0.
These
are called the lines
U^y*
(z)
= 0.
On
the same diagram are drawn
= 0.
These are transcendental curves having

double points on the axis z = 0, as will be seen from the short
the curves
discussion below.
Let now the edge of a sheet of paper be kept parallel to the

axis of z and be moved along the diagram from bottom to top. It
will intersect all
the curves.
The
distances from the axis of y

its positions to the points of
along the edge of the paper in any of

intersection are values of
z, for
the value of y for that position,
which
185
XIV.]
satisfy 41
value of y give
and are therefore values of z which with that
maximum
ioit
or
minimum
values of
2
.
186
[XIV.
When 2 = 0, that is where the curve meets the axis of y,

V = 1, V1 = 0, so that by g,
= sin \y,
U = 1 cos \y.
Thus C7j = 0, whenever cos ^y = 1, that is when \y = 2m-7r, or
y = 4mnr. The curve U = therefore meets the axis of y at every
J7"i
multiple of
But
4tt.
this value of
y makes
and likewise
^.z/yl^J^-U^z/y}
zero, so that
^? =
The value
which the curve meets the axis of
a double point.
is
If
z'
be current coordinates the equation of the pair of
tangents at a double point
It
of dy/dz is therefore indeterminate at the points on the
axis of y, that is each point in
~ =o
is
is
very easy to verify by differentiation and use of the properties
of the functions that
when
**
dz*
and y
dydz
= 4m7r,
'
By'
*'
so that the equation of the tangents reduces to

(j/
- im-n-y = 1z">.
Thus the equations of the tangents are

y'
and their
4wnr = \/Zz',
that
4mw = \J2z',
inclinations to the axis of z are given
tan
is
y'
by
$ = JZ,
= + 5444'8"-2.
Where the curve meets the
axis of z
equivalent to the two equations y*
1/2=0 splits into two straight

and the curve represented by
3
2/
= 0,
we may regard
U^/y 2
= 0,
lines coincident
U=
2
as
so that the curve
with the axis of
z,
XIV.]
We have
Hence
by
187
3,
the last curve
for
I flu)
when y = 0.
-U
dy
dz\y*
dz~
V
/l^N
By
VjTV
Thus the branches
_00,
of the curve
U /y =
2
cut the
axis of abscissae at right angles, as shown in the diagram on p. 185

above. We shall see below that this intersection takes place at
= 0= 0.
Thus the curves
points satisfying the equation J"2 {%)
when y =
0,
see,
(z)
Ui/y",
be seen from the curves in the diagram that the value
It will
of dy/dz
touch the curves
is
negative so long at least as y
< z,
that
is,
as
we
shall
within the region of the curve corresponding to the geometrical
But
shadow.
at points along a line Ji{z)
= 0,
2*
dy
and
is
positive so long as z
> y,
that
Hence no
the geometrical shadow.
is also
within the region of
intersection of a line Jj.{z)
with the other curves can exist in the region of the diagram corre-
sponding to the geometrical shadow.
To
settle
where the maxima and minima are we have to
calculate d*M*/dz\
Now
-W-~ 2 \y)dz
{-
=2
Thus considering
maximum
On the
are
or a
(?)
first
|i Jx Ut
-J U + Ji (Ji - Ci)}
points
upon the
minimum
according as
when Z72 =
minima according
or
JU
t
is
0,
we have
as Ji7"s zjy {= Ji
negative or positive, or as
J?<or>-0iJi.
if
positive or negative.
the points on the curves CT2
other hand,
maxima
lines Ji
46
(Jx
UiZJy)} is
188
we
Calculating d"M*/dz 3
and a curve
see that this does not vanish for points
(z) = 0, U = 0, that is wherever a line

U = 0, intersect there is a point of inflexion
satisfying the equations
Ji (z)
[XIV.
drawn with M" as ordinates and values
of the curve of intensity,
of
z as abscissae.
by the statement above
It follows
as to the inclination of the
curve within the region corresponding to the geometrical shadow,

that within that region there can be no point of inflexion on the
intensity- curve.
Also, as can easily be verified, there are points of inflexion of
has a
we can
see
the intensity-curve, wherever the curve UJy*
minimum
or
maximum
ordinate.
Eeferring
now
to the diagram, p. 185,
how
to in-
where there are maxima and minima. For pass

is crossed.
along a line Jx = until a branch of the curve UJy* =
Here clearly Ut changes sign, while JQ (z) does not. Thus JQ {z) U%
dicate the points
changes sign, and so
all
points of a portion of a curve
= 0,
(z)
intercepted between branches of the other curve, give maxima, or

give minima, according to the
number
of branches of the latter
which have been crossed to reach that portion

Ji(z)
from the axis of
z.
portion, positive for the second,
ings being
If
we
(z)
and
is
by proceeding along
negative for the
number
so on, the
first
of cross-
0, 1, 2,
pass along a curve U^jy 1
Ji (z) changes sign, but not so
and cross Jt (z)

by 14 when
for
= 0,
then
= 0,
(z)
and U1 is a maximum or a minimum, since t7"2 = 0.

lt
must be further noticed that when for a branch of the
U = U
3
But
it
curve
C/2/y
the value of dy/dz
changes sign while
and because of
minimum.
At
U = 0,
2
(z)
is zero,
does not
9 UJdz
= 0,
that
also for
vanish
first
U z/y = 0,
s
U = 0,
so that Z7X is
those which give minima,
above of using the diagram
The
when
maximum
or a
these points therefore d^M^jdz3 changes sign, and
hence they also separate regions of the curve
maxima from
is
is
UJy2 =
when the
which give
process described
carried out.
three successive differential coefficients of
when z=0, and y=4mnr,
that
is
all
at the double points,
XIV.]
and as there, as the reader may
verify,
aw _3
a
dz*
~2roV'
the douhle points are places of
The
189
regions of the curves
minimum
UJy =
2
M\
(zero) value of
can now, starting from the
maxima
minima when the diagram is used in the manner described
above.
To mark regions which give minima they are ruled heavy
double points, be easily identified as regions which give
or
in the
diagram
the other regions, which give maxima, are ruled
light.
Thus the
minimum
first
regions from the double points to a
maximum or
J (z) = Q,
of the curve, or to a point of crossing of
whichever comes
first,
are ruled heavy, then the region from that
U "changes
point to the next point at which
sign
is
ruled light, and
so on.
The lower
regions of the curves Ji (s)
to the points of
meeting with
= 0,
U /y = 0, are
2
from the axis of z

the next
ruled heavy
regions, from the first points of crossing to the second, light,
Thus the whole diagram
so on alternately.
and
is filled in.
As we have seen
U,=
F - cos ($y +^Q = /.(*)- g)' J
Hence
(z)+
as y increases in comparison with
z,
...
-cos (\y+
the equation
U = J(z)-cos$y
a
more and more nearly
holds.
The reader may
verify that the
curve Jo (^) cos \y = meets the axis of y at the same points as

the exact curve, and has there the same double tangents.
On
the other hand,
if
y be made smaller
in comparison with
z,
then by 3 we have more and more nearly
so that the branches of
to the lines
p.
= 0.
(z)
U /y' =
a
187 above.
The value
of
2
,
approach more and more nearly
Thus we
namely
(2V
verify the statement
made
at
190
with increasing z and stationary
y,
that
Hence
of the rays, approaches zero.
is
[XIV.
with increasing obliquity
at a great distance from the
geometrical image of the orifice the illumination
is
practically
zero.
Consider a line drawn in the diagram to
y=cz.
line
have the equation y = -
the equation
fulfil
making the same angle with the
axis of
y would
Let us consider, the intensities
z.
for
points on these two lines.
Since yjz
=c
for the first line,
then for any point on that
line
e
/j
= cJ c J +
3
...
U = c*J -&J +
i
= -COS
47
...
+ J -- J, + - J - ...
|i*(c
For the other
line
we
-)J
have, accenting the functions for dis-
tinction,
U[1
1
1
= -J -=i J,+
l
= sin hz
(c
...
+ -H-(cJ" -c J3 +c J
s
-...
),
47'
m=lj*-^J*+
=-
cos
Uz (c +
+ Jo - c / + c / 2
-J!
Therefore
1+
tf
[r;
= sinji*(c +
i)J,
48
Now
if
the radius of the geometrical shadow be f

t;
and
= (a + b)r/a,
then
XIV.
191
be the distance of a point of the illuminated area upon

= z/c we have evidently
If '
the other line y
As
special
y y=
of
cases of these
or the axis of
lines
and y
z,
we have z = 0,
= z. The last
or the axis
is
dotted in
the diagram, and by the result just stated corresponds to the edge
of the geometrical
The
shadow.
intensities for points along the first line are the intensities
at the axis of
symmetry
for different radii of the orifice, or
constant radius for different values of
from the
Those
orifice.
b,
points along the
for
with
the distance of the screen
second line are
the intensities for the case of Fraunhofer, already fully considered.
In the
first
case
we have by
= 0,
20, since z
U = sin \y, U =2 sin
\y
so that
M'~(?)\u;+ui)
This
is
the expression for the intensity at a point of a screen,
produced by diffraction through a narrow

denoting
where a
1-rra^\\f,
is
slit,
\y in that case
the half breadth of the
the distance of the illuminated point from the
slit,
and the
distance of the point from the geometrical image of the slit on the
screen.
Thus Tables, which have been prepared for the calculation
slit,
of the brightness in the latter case, are available also for calculating
the brightness at the centre of the geometrical image of the circular

orifice.
The
intensity
is
zero excluded), that
treme and central

a
maximum when
when \y = mir (m being any whole number,

when the difference of path between the exrays is a whole number of wave-lengths. It is
zero
is
tan \y
= \y,
+b A
-=-^- r2
(ira
or
tan
U-
2ab
ira
+b
-=-=- ra.
X 2ab
192
Some
values are given in the following Table.
= 0.
hf
[XIV.
XIV.J
wave being allowed
of the
193
to pass unimpeded.
original expressions, obtained at p.
167 above
Going back
for
see that for the total effect of the uninterrupted
by
to the
the intensity,
we
wave we have
and 2
= 27r
(?,
f
I
2
Jo (Ip) cos (P/3 ) pdp
P\
= 7r -v
sin
gT
5i
'
30
I'
Sx =2ttJ J
{lp) sin
(V)
pdp
= 7T t cos ^r
Thus we get
"---$'-
if,
170 above,
as at p.
be, as
7rr
be taken as unity.
leads to the expression l/(a
it
This
+ 6)
We
the point in which the axis meets the screen.
statement,
made on the
last page, that the
at the centre of the geometrical
is
as
it
ought to
for the intensity at
thus verify the
maximum
image of the
illumination
orifice is four
times
that due to the uninterrupted wave.
might.be objected that the original expressions obtained,

which are here extended to the whole wave-front, had reference
only to a small part of the wave-front, namely that filling the
It
It is to
orifice.
be observed however that the
elements of the wave-front, which
effects of those
at a distance from the axis,
lie
are very small compared with those of the elements near the axis,
and so the integrals can be extended as above without
To
on
p.
opaque disk we have simply
find the illumination with the
to subtract from the values of Goo,
169
for
the
orifice.
error.
Sx
the values of
Thus denoting the
Gr Sr
,
differences
given
by Cu
Sj
we get
2
C = Ooo-Cr = --(F sin^- FiCos^)

1
S3
S,
= -&=
?(7; cos
y
M\ =
and
G. M.
Q\V\ +
& +7,
sin iy)
V\),
54
13
194
[XIV.
Comparing these with the expressions on p. 170 for M* we

same except that now Z7a is replaced hy Fj
and U2 by V
see that they are the
.
= 0,
If z
that
is if
the point considered he at the centre of the
V = 1, V = 0, and
geometrical shadow,
MIT'
that
is
the brightness there
opaque disk did not

first
exist.
56
always the same, exactly, as
is
This
is
if
the
the well-known theoretical result
pointed out by Poisson, and since verified by experiment.
M\
For any given values of y and z

U1 U2 by the equations g
those of
is easily
calculated from
V = cos i(j/ +
^J
+ U
lt
F^sini^ + )-^.
J
A valuable
set of
numerical Tables of
M\
all fully illustrated
by
curves will be found in Lommel's memoir.
When
continually increased in value the equations
is
V = cos \{y+^j,
V =sm\(y + Z
more and more nearly

approach zero.
hold, since,
The value
of
by n,
and
thus becomes
4</y
?72 continually
2
at a great dis-
tance from the shadow, as in the uninterrupted wave.
As
mum.
before
we can
find the conditions for a
Differentiating,
and reducing by 18 and
The maxima and minima have

7,(*)
The
maximum
place therefore
= 0,
or F
obtain
when
= 0.
roots of these equations are the values of z
M> ()_
we
14,
or mini-
yJIi-n
which
satisfy
XIV.J
and
or
are, therefore,
minimum.
book; those of
similar to,
is
roots of
V =
(z)
make
(z)
and
maximum
are given at the end of this
can be found by a formula of interpolation

in the same way as, 43 above.
and obtained
The tangent
of z
values of z which
The
195
of the inclination of the curves 7"
to the axis
given according to 24 by
ay
dz
y
s
r+
r-
'
(i)
By
using in this the values
V_,=-^-(i)V, + ()V,-...,
we
see that if
dv
= 00
-~-
00
values of y parallel to the axis
that
,of y.
the curves are for great
is
Also since
the asymptotes of these curves are the lines

J.()
= 0,
A table
drawn
parallel to the axis of y.
tion
given at the end of this book, and as has been seen above
(p.
is
of the roots of this equa-
46) their large values are given approximately by the formula

(to
+ f ) ir.
Writing now
iK>V
'-(f)
and making the values of

disappear,
that
is
and we are
left
y,
V +-=
z small, the terms after the
first all
with
+ z1 =(2m+l)iry.
59
132
196
[XIV.
We
This equation represents a circle passing through the origin.

infer that the branches of the curve
V=
become near the
origin
arcs of circles all touching the axis of z at the origin.
The curves V =
23
7JT22
are
shown
in the adjoining diagram,
and
'
XIV.]
Where both
(z)
but not so that of -^-~-
and
V=
Hence
the value of -^-r2 vanishes,

oz
at such points the curves of in-
tensity have points of inflexion, but there are
As
197
no
others.
in the other case, points of inflexion of the intensity curve
can only exist outside the shadow region of the diagram.
V_ = - Vu
= 0,
Ji(z)
For since
58 becomes
dy
which
-T- is
is
positive if y
<
negative
z,
the line
(z)
with
for
V = 0,
except
> z.
But by the diagram
when y <
z.
intersection of
Thus the
state-
comparing further the case of the disk
orifice, let
us contrast the intensity along a line
with that along the line y
the second
Hence there can be no
Lastly, for the sake of
y = cz
if
proved.
is
with that of the
/z\-
positive everywhere.
ment just made
dz
line,
we can
= z/c.
Accenting the quantities
easily prove that
Vl+V2 + V'* + r? = Ul + U\ + U'* + U? + 2J (z) cos %z(c + ).
61
Now we
and
for
have
for the orifice
the disk
Thus by 61
& (M{ - M*) + \
(#/
- M'*) =
ja
/ (*) cos \z (c
+ -)
62
198
The shadow region

therefore
1,
by the
so that 62
line
is
that for which y
y = z.
On
this line
[XIV.
> z, and
M=M',
is
bounded
and
M = M[
1
becomes
2 (Ml
- M% =1 = I Jt 0) cos *.
z
from the diagram that as y increases the number of

fall within the shadow also increases.
It is clear
dark rings which
The reader must
refer for further information on these cases of

Lommel's paper, which contains, as we have indicated,
a wealth of numerical and graphical results of great value. The
discussion given above is in great part an account of this memoir,
with deviations here and there from the original in the proofs of
various theorems, and making use of the properties of Bessel
diffraction to
functions established in the earlier chapters of this book.
The same volume of the Ablmndlungen der Konigl. Bayer.

Akademie der Wissenschaften contains another most elaborate
memoir by Lommel on the diffraction of a screen bounded by
straight edges, in which the analysis
that used in the

find space for
first
is
in
many respects
We
paper, and given above.
similar to
can only here
some particular applications therein made of Bessel
Functions to the calculation of Fresnel's integrals, and a few other

results.
From the result obtained above for Fraunhofer's interference

phenomena, namely that the intensity of illumination is pro4
portional to
Jl (z), the source of light being a point, we can find
when the source is a

uniform straight line arrangement of independent point-sources.
the intensity at any point of the screen
Let the circular
orifice
be the opening of the object-glass of the
telescope which in Fraunhofer's experiments
on the source of
the telescope
plane of the
light.
If the source
we may suppose with

orifice is at right
is
is
supposed focused
at a great distance from
sufficient accuracy that the
angles to the ray coming from any
point of the linear source.
Let rectangular axes of
rj
be drawn on the screen, and
the line of sources be parallel to the axis of
f=
0.
A little
77
and
let
in the plane
consideration shows that the illumination at any
J
XIV.]
199
upon f and (constant
point of the screen must depend
factors
omitted) be represented by
dr).
/
Jo
But
if
r be the radius of the object-glass, and
"
the distance of
the point considered from the axis of the telescope,
= -^?=/*say,
^ =?2 _|,= f!_ ?
and
TT
Hence
an
M
if y
= /* p.
The
integral
is
This integral
for
may be
/*
Vs
-v
therefore
if
suitable
dz
z dz
=z
=
Jl(z) dz
Z
*Jz-
transformed in various ways into a form
The
numerical calculation.
process here adopted
depends on the properties of Bessel functions, and is due to

Dr H. Struve*. Another method of obtaining the same result
be found in Lord Rayleigh's Wave Theory of Lightf.
will
Before, however,
three
we can give
lemmas on which
The
first is
Struve's analysis
we have
to prove
his process depends.
a theorem of Neumann's and
is
expressed by the
equation
IT
2 f2
Jl(*)
By
72, p.
J
But
= -l Jm (2z sin a) da.

7T
'n
TT.'o
63
28 above, we have
(2c sin
I)
= J? (c) + 2J? (c) cos a + 2 J
(2c sin |j
-J
1
=-
cos \2c sin | sin
f*
{Jo (2c sin
</>)
2a'.+
d$
<
+2
(c) cos
(2c sin
<f>)
cos a
+ 2 J, (2c sin <) cos 2a +

* Wied. Ann. 16, (1882), p. 1008.
+ Encyc. Brit. 9th Ed.,
. .
.}
p. 433,
d$
200
by
[XIV.
we
Identifying terms in the two equations
40, p. 18 above.
obtain
1 t"
Jl
(c)
= 7T
J ( 2g sm ft #>
2 fi"
or
if
Jn
we
(*)
=~
write z for c and a for
^m ( 2 * Sln )
7TJ0
Thus the
</>.
first
lemma
is estab-
lished.
The second lemma

,
the equation
is
2 f
00
sin (xz)
V^
2
7T J
2
tt
>'0.
If
Pn denote the
zonal harmonic of the ?ith order, then
it is
theorem of Dirichlet's that
"
fl
Pm (cos0) =
*
r.
/in
cos Ad>
2y cos w<
r
cos 5)
ov2(cos
^<f>
sin
+ f
j e
w(/>
sin A<
2y cos nd>
Y
V2
(cos #
cos
/"'
JoV2(cos</> cos 0)
\<f>
V2 (cos
cos
6
sin
,.
d<f>
<j))
w/>
cos 0)
Subtracting the second expression on the right from the

we obtain
f'
oobQ*
-j)0
cos#)
./ov2(cos<
For # and
put 0/ra,
equation becomes
</>/,
<f>
gm(.-i)0
<J
and
cos
V 2 (cos
first
let
<f>)
n be made very great
the
last
cos
(f>d<j>
00
Z*
Jo\/6 -(f) ~Je
The quantity on the
Write #z
for 0,
left is
and u*
""
&
which
is
the second
lemma
sin
</><&
V<fr>
-d
îrJ
for 6,
\
xdz
(6), (see p.
for d<
/""sin (xz)
J v
'
Vz2
stated above.
and 65 becomes
dz
V*
32 above).
Hence
XIV.]
The
third
lemma
I
is
expressed by the equation
{z sin a) sin
ada =
Using the general
66
JO
definition (p. 12 above) of a Bessel function
we get
of integral order
^J
(*sina)sinada=2
J-
201
sin*+W
(1I
Jo
_
~y
z
_^
~
(zy
(-) 8
2 2s (lis) 2
n(2s +
(n s y\2)
_sing
i)~
n(2s +
z
(-) 8
i)
'
which was to be proved.
Returning now to the integral
r Jl(z)dz
JvZ V.Z V
2
let
us denote
by
it
We
Z.
Z=
But by
(2z sin a)
and by 45,
p.
Z NZ*
V2 JO
first
lemma
(2z sin a) da.
13
20, p.
have by the
-^=
TJ
= ^ S*n
{</i
(2z sin a)
+ Js (2z sin a)],
18
sin a)
= -1
J, (2s sin a)
= -1
/i (2,z
/"*
sin (2s sin a sin /3) sin ft dft,

fir
I
sin (2s sin a sin #) sin 3/3 dj8,
so that
!/*"
y
Z=
^
a
smadotlf/(sin /3
7T./0
But
if
*o\ d/3
Jfl
+ sin 3/3)
,
Jo
f sin (2* sin sin
ff) cfc
, 2
VZ V*
<
we put 2 sin a sin ft = x we get by the second lemma

f
sin (2z sin a sin ft) dz
ir T
Hence
1
=5-
f*
^7T J
(sin
^ + sin 3/3) d/3
f J*
J"
JO
(2w sin a sin
ft)
sin acta,

202
[XIV.
which by the third lemma becomes
Z=-\
(sin
sin 3/3)
2v sin
7tJ
2 /*"
=
Let now
sm /3)
sin (2
cos2
^' d@
67
/3 d/3.
be a function defined by the equation
(z)
HJz) = -[
68
sin (s sin 6) dd.
7rJo
Expanding
and integrating we obtain
sin (z sin 0)
^> = !{*-i^ + id^5*-"}Now
69
be another function defined by
let flj (2)
H (z)=rH
(z)zdz,
Jo
1
then by the series in 69

2
We
shall
z3
now prove
Hi (z)=
rl*
sin (z sin 6) cos 2
dd.
71
Jo
71"
by differentiating that
Multiplying by zdz and integrating
Now
that
2z2
It can be verified
z7
zs
2s
we
d#
find
(a)
by 68
H-^ =
dz
cos (s
sin 0)' sin dad.
v
tt J
Hence
(z)
=
i"
2z
=
it
Az
=
w
il
r^
cos (^ sin 6) sin 0c
Jo
cos (z sin 0)| sin 0d9
/""
sin2 (\z sin 0) sin
Jo
d0.
74
XIV.]
It
may be
It is clear
203
noted that every element of this integral
from the form of H (z) given in the
equations just written that
approximates when z
(z)
is positive.
of the three
first
is
large to
2z/tt.
by parts we obtain
Integrating 74
(z)
we write 2v
sin (z sin 0) cos 2 Odd,
JO
term vanishes at both
since the integrated
If
and
for z
6 the equation becomes
/3 for
H (2v) = 2
if
limits.
sin (2v sin /3) cos 2 /3d/3.
75
Jo
Hence
Hi(2t0
by
67.
(z) are
It is to
1
= --
sin .(2v sin /3) cos 2 /3d/3
TV Jf
4 3
= Z,
be observed that the functions here denoted by
the same as Lord Rayleigh's K(z),
the Theory of Sound, 302, to which the reader

further
details.
function
(z)
(z)
the relation between the two functions

1
The value
by the
We
of
H (z) can
series in 70,
(z)
must then have recourse

Chap.
to that established in
The
series
is
referred for
= zJQ
is
(z).
is
large this series
is
is
not too great
not convenient.
to a semicpnvergent series, similar
above for the Bessel functions.
rv.
can be found easily by the method of Lipschitz, already
used in Chap. vn.
The
following
is
a brief outline of the
process*.
By
(z),
differs
be calculated when z
but when z
however from the

1
If we denote the latter by >! (z),
The function
used by Struve.
discussed in
{z)
the definitions of the functions
J (z)-iH (z) = -
V W
fesin
7T J
we have
e^vz
T J Vl - V
*
dv.
See Theory of Sound, 302.
204
Now
[XIV.
take the integral

C
e-Zw dw
Vl + w2
which w = u + iv) round the rectangle, the angular points of

which are 0, h, h + i, i where h is real and positive. This integral
is zero, and if h = oo it gives after some reduction
(in
zJo
JoVl-j)2
dp
z'
^M-lr*
Expanding the binomials and integrating, making use of the
theorem
Jo
and equating the
real part of the
î-jrH
imaginary part to
namely
(z) as in
(z)
Chap.
we
{z),
IV.
(z)
2
.
S 2z~ 5
-I
3 2 5*z- 7 +
. .
IT
+ sj
where
and
written for
From
73 and
^ {P
and the
and
= - (z- - 2T + 1
1
^ttJ
result to
get the expansions required,
sin (*
.)
-frr)-Q cos (z - |tt)},
have the values stated on
p.
48 above
76
(z
being
x).
this the value of
(z)
at once found
is
by the
relation
is
H, 0) = -
(*
+ r-i - %z~ + l
s
2
.
3 2 5z- 5
.
- ...)
7T
2z
7T
cos (z
v
^7r) ) 1 (l -4)(3 -4)

,
'
(l 2
1.2. (8s)2
- 4) (3 - 4) (o - 4) (7 - 4)
2
1.2.3.4.(8^)*
/2z
P-4
8*
(l 2
- 4) (3 2
4) (5
1.2.3!(8^)
- 4)
"
"*
77
be noticed that
is nowhere zero, and that
1 (2v)
has maxima and minima values at points satisfying the
It is to
205
XIV.]
(2i>)/w
equation
H (2v) = ^H (2v)-SH (2v )^
dv
v*
tf
The corresponding values
of v are therefore the roots of
4>v*H (2v)-Sff1 (2v)
= 0.
Now
let there be two parallel and equally luminous linewhose images in the focal plane are at a distance apart
It is of great importance to compare the inTr/fi,, say.
sources,
v//i
image of either
tensity at the
line
with the intensity halfway
In this way can be determined the minimum

distance apart at which the luminous lines may be placed and
We shall take the image of
still be separated by the telescope.
between them.
one as corresponding to v = 0, and that of the other as correspond= 7r. Thus the intensity at any distance corresponding to
ing to v
is
B, (2v)
proportional to
Putting
^ fli(2)
we have by JO
i
The
two
2 2u 2
^ = ir3 -
.3.5
2V
+ l '.3 .5
!
ratio of the intensity of illumination
.7
""'
midway between the
lines to that at either is therefore
2Z(tt)
Z(o)+iw
This has been calculated by Lord Rayleigh (to
comparison is due) with the following results
Z(0) =
-3333,
Z(ir)
so that
2L
intensity
4| per cent,
less
is
this
= -0164, (**) = -1671,
Z(0) + Z(tt)
The
whom
=-955.
79
therefore, for the distance stated, only about
than at the image of either
line.
206
[XIV.
Now
27T7-
which gives
Since b
is
2r'
the focal length of the object-glass, the two lines are,
by this result, at an angular distance apart equal to that subtended

by the wave-length of light at a distance equal to the diameter of
Two
the object-glass.
lines unless at
a greater angular distance
could therefore hardly be separated.
This result shows that the resolving (or as

the space-penetrating)
called
power
of a
it
is
sometimes
telescope is directly
proportional to the diameter of the object-glass.
By
multiplying
2
7T
by
that
(idl; ,
is
by
H,
(2v)
(2vf
and integrating from f
dv,
= <*
to
= + oo we
get an expression which, to a constant factor, represents the whole

illumination received by the screen from a single luminous point
the image of which
mode
in
which
at the centre of the focal plane.
is
H (2v)j^ was
1
as the illumination received
obtained,
by the
it
plainly
latter point
Or,
may be
by the
regarded
from an
infinite
uniformly illuminated area in front of the object-glass.

If the integral
is
taken from f to
oo it will
represent on the
same scale, the illumination received by the same point from

an area bounded by the straight line parallel to rj corresponding to
the constant value of f. The point will be at a distance f from
the edge of the geometrical shadow, and will be inside or outside
the shadow according as ( is positive or negative.
We
have by 71
Jo
(2)
irJo
= ij
Jo
cos*/3d/3
o
= |.
Now
fH
(2*0
f S (2v)
1
f H, (2v)
XIV.]
The second term on the right caa be

Hence we get
207
calculated
by means
of the ascending series 70.

'
( 2v )
_ ""
2V
v
f
l2
J V
+
This multiplied by 4/tt
32
2V
80
l3
32
52
the expression given by Struve for
is
the intensity produced by a uniform plane source, the image
of which extends from v to
Struve's result
we
write
/(+)=*
Hence
from
00
if
to
00
when
f
For the sake of agreement with
v is positive
(2v) s
2v
81
/ be
the illumination when the plane source extends
we have
00
!(+) + J(-t>)=/=l.
This states that the intensities at two points equally distant
from the edge of the geometrical shadow, but on opposite sides of
it,
The
are together equal to the full intensity.
edge of the shadow
The reader may

series gives
is
verify that
when
is
great the semi-convergent
approximately
fuj/l,
+
W
7TU
The
intensity at the
therefore half the full intensity.
12W
cos(2i>
1
27T
+ |7r)
following Table (abridged from Struve's paper) gives the
intensity within the geometrical shadow at a distance f = bXvj2irr

from the edge, and therefore enables the enlargement of the
image produced by the
diffraction
estimated.
27rr
of
the object-glass to
be
208
We
now
shall
consider very briefly the theory of diffraction
of light passing through a narrow
We
[XIV.
shall suppose that the diffraction
every plane at right angles to the
bounded by
slit
parallel edges.
may be taken
as the
so that the
problem
slit,
same
in
one
is
Let a then be the radius of a circular

wave that has just reached the gap, and consider an element of the
in only two dimensions.
Let also
wave-front in the gap.

pole so that
be the distance of
distance from the source
its
is
a+
b,
P from
the
ds the length of
the element of the wave, and 8 the retardation of the secondary
wave (that is the difference between the distances of P from the

element and from the pole). The disturbance at P produced will
be proportional to
cos
2-n- (
If the distance of the
b,
The disturbance
COS 27T
The
jp
be
2ab
we get
%ttv2 for 2ttS/\,
ds.
element from the pole be s, and

then it is very easy to show that
b
Writing as usual
yp -J
ab\
at P. is therefore
-A = COS ^TTV* COS 2-7T yp + SU1 ^7Tl>2 sin 27T yp
intensity of illumination
due
to the element is therefore
constant, being proportional to

cos 2
fyiriP
+ sin
\irv\
where
,
The whole
intensity
is
2(a + b)
ao\
thus proportional to
cos \irv 2 dv
.
/sin îrif.
dv
the integrals being taken over the whole arc of the wave at
the
slit.
XIV.]
The problem
is
thus reduced to quadratures, and
We
evaluate the integrals.
C=\
are
Various
known
=1
cos îrv^dv,
Let
sin \irv^dv.
Jo
as Fresnel's integrals.
methods of calculating these integrals have been
devised; but the simplest of

lation is
remains to
it
shall write
.'o
G and 8
209
all for
purposes of numerical calcu-
by means of Bessel functions, when Tables are

îrv*
= s,
available.
then
V=\
cos zdz = h
\/
V 7TZ
J-i ( z ) dz
"
%z
>
.'
S = \\/wyzdz=\'ji (B)dz.
Ji
83
J n
Let us now consider the Bessel function integrals on the

Using the relation
right.
we have
J. i ( Z ) = 2Ji(Z) +
-2J' (z)
h
Ji (z)
22
+ 2JL(z) + ... + 2JJn +i ( z) + -U> <
Thus we obtain
1
I*
J
J- 1S 00 d = Ji
z)
+ -/(*) + . +J4n+i(
_ 2~ z ) +
2
I I
'0
J* 0) dz
By taking (4n + 3)/2 sufficiently great the integral on

Thus we get
of 84 may be made as small as we please.
Similarly
we
84
the right
find
8 = ^j'ji (z)dz = Ji (z) + Ji (z) + J^(z) + ....

These
series are convergent,
86
and give the numerical value of
the integrals to any degree of accuracy from Tables of Bessel

functions of order (2n + 1)/2, by simple addition of the values of the
G. M.
**
210
[XIV.
successive alternate functions for the given argument.
are apparently due
memoir
Lommel, and are stated
to
referred to above, p. 198.
C=\
Jo
He
(z)
dz
The
series
the second
gives also the series
J_ h 0) dz = V2 (P cos
S=$f J
in
\z
+ Q sin
\z)
87
= V2 (P sin \z-Q cos %z)
88
where
P = Ji (^)-J4 (^) + Jf (^)-...
Q=J,(\z)-J^z) + J^(\z)-....
The proof is
the reader.
left to
C and S
of v by
were expressed long ago in series of ascending powers

Knochenhauer, and in terms of definite integrals by
Gilbert.
From
when
the latter semi-convergent series suitable for use
v is large are obtainable
at p. 203 above.
by a process similar to that sketched

however to pursue the matter
It is not necessary
here.
The very elegant construction shown

known as Cornu's spiral,
shows graphically how the
in
the diagram, which
is
G +S
2
value of
The
!i
"Y
varies.
abscissae of the curve
are values of
and the
ordi-
nates values of S.
It can
be shown that the
distance along the curve from
the origin to any point
is
the
value of v for that point, that
the inclination of the tangent

to the axis of abscissae
is
^7ru 2,
and that the curvature there
As
v varies from
more and more
The
closely
is tro.
and from
round the poles
to oo
to
and B.
oo
the curve
is
wrapped
origin of the curve corresponds to the pole of the point
considered, so that if vl7 v a correspond to the distances from the
pole to the edges of the

v u v2
on the
spiral
slit,
and draw
we have
only to
the chord.
The
mark the two
points
square of the length
XIV.]
211
of this chord will represent the intensity of illumination at the

point.
to
The square
any point v
is
l
4
of the length of the chord from the origin
the value of
.
C +S
2
2
,
that
is
of
,74 (*)&}+ i{Jy4 d*
As v varies it will be seen that the value of this sum oscillates

more and more rapidly while approaching more and more nearly
to the value \.
14
9!
CHAPTER
XV.
MISCELLANEOUS APPLICATIONS.
of
In this concluding chapter we propose to give a short account

some special applications of the Bessel functions which, although
not so
difficult as
those already considered, appear too important
or too interesting to be passed over entirely or simply placed in the

collection of examples.
We
begin with the equation
will
^=aV
2
Ob
which occurs in various physical problems, such as the small

vibrations of a gas, or the variable flow of heat in a solid sphere.
Using polar coordinates, u
a function of
is
o
du
9 m
[
- 1 n Ka
r 2 2 T
2r 7T- +
+
^
sin
dd
8r
dr
6
1*1
du
a?
dt
Assume, as a particular
(
n du\
sin &
dd)
\
t,
r, 0,
$, such that
T sin
d<f>\
solution,
u=e"M*vSn
where v
is
a function of r only, and
8n is a surface spherical
harmonic
of order n, so that
1
sin
d(.
80 V
dSn\
J^
dd
Then after
a
substitution in
d*Sn
sin 2 6
dp
,._
appears that v must satisfy the
(2), it
equation
and now
if
d?v
2dv
dr 2
r dr
f
'
n(n
we put
v=r~
+ l))
r2
['"
w,
'
XV.]
we
find that
the equation
satisfies
<Pw
213
dw
(n
+ 1) )
2
Hence
w = AJn+i {r) + BJ_ n _ h (*r),

and
finally
= .-**
is
is
+i
()
+ U/.^ ()}
a particular solution of the equation
or
In practice n
2.
a whole number, and by a proper determination of the constants
A,
n, k,
is
{^
B the
function
adapted in the usual way to suit the particular conditions of
the problem.
pp. 176
(See Eiemann's Partielle Differentialgleichungen,

and Rayleigh's Theory of Sound, chap. XVII.)
If in the above
constant,
we suppose n =
the function
Sn
reduces to a
and
Jj+l (icr)
(see p. 42)
=\
sin kv
thus with a simplified notation,

tt
Vd
K
which
0,
189
may be adapted
sin/w
we have a
solution
'
to the following problem (Math. Tripos,
1886):
"
uniform homogeneous sphere of radius b
temperature V, and
is
substance of thickness b at temperature zero.

left
to cool in a
time
t,
is
at uniform
surrounded by a spherical shell of the same
medium
at temperature zero.
The whole
the temperature at a point distant r from the centre
_ v ^ 4 sin xb Kb cos Kb sin kt

k
4>Kb
sin 46
is
Prove that, after
- K wt
is
'
'
214
[XV.
where the values of k are given by the equation

tan
h being the
2/cb
of the
ratio
2 Kb
1-2A6'
surface
conductivity to the
conductivity."
Here the conditions
to be satisfied are
V= V from r =
r=b
V=0
= b,
...r = 2b,
to r
...
when
and
Br
when r =
2b, for all
values of
Now, assuming a
t.
solution of the form

siaicr
V=2A
the last condition
is satisfied if
3 /sin
ii r\
3r \
when r = 2b
that
e- KVt
A
h
+
.
sin kt
r
=0,
if
is,
k cos 26
siD
~^6
h sin
2xb
2/c6
+ ~~26
4P"
'
leading to
tan 2/eo
- 2Kb
as above stated.
Proceeding as in Chap. VI. above,
sin2 Krdr
Jo
that
is,
,,
we
infer that
Vr
sin icrdr,
rsb
rib
Jo
sin46\
z
fb
=
_
.
"
r sin /OT* r
I
6 cos k&\
/sin 6
\
and hence
.
_4
~k'
sin 6
& cos
K.b
4<Kb- sin 4*6
which agrees with the result above given.
"
internal
XV.]
215
Returning to the solution given by equation 4 above, we may

observe that when a2 is real and positive, the solution is applicable
when
there is a "damping" of the phenomenon considered,

problem just discussed. When there is a forced vibration
imposed on the system, as when a spherical bell vibrates in air, we
must take a? to be a pure imaginary + ia/, so as to obtain a
to cases
as in the
is then 27r/a.
An illustraend of the book.
consider two problems suggested by
The period
time-periodic solution.
tion of this will be found at the
We
will
now proceed
to
the theory of elasticity.
The
first is
that of the stability of an isotropic circular cylinder
of small cross-section held in a vertical position with its lower
clamped and upper end

It is
a matter of
end
free.
common
observation that a comparatively
short piece of steel wire,
such as a knitting-needle,
placed vertically with
lower end clamped in a vice
its
would be impossible to keep

piece of the same wire.
To find the greatest length
vertical in the
is
stable
;
when
whereas
it
same way a very long
consistent with stability,
we
consider
the possibility of a position of equilibrium which only deviates

slightly
from the
vertical.
be the weight of the wire per unit of length, /3 its

Then if x is the height of any point on the
flexural rigidity.
wire above the clamped end, and y its horizontal displacement
Let
from the vertical through that end, we obtain by taking moments

for the part of the wire above (x, y)
being the whole length of the wire.

Differentiate with respect to
or
then
'
216
Now
[XV.
put
on = r*.
then
da;2
dr2
and the transformed equation

dr 2
and now,
is
+ 3rdr
'
= r z,
3
be found that
di*
Hence
+ rdr +
\9P
9rV
if
2
_4w
*
it
9/3^
we put
if
^>
it will
drj
9/3'
follows that
When
a;
Z,
that
= AJ^ (*r) + BJ_ 4 (r).
is,
when r =
0,
we must have
da;
'
whence
r*^ =
<2r
when r =
that
'
is,
if
idz
_a
or
3r^ + .
Now
forms
the initial terms of /|(r) and J_^(kv) are of the
XV.]
J|
and
it is
(**)
= ar* + /8r* +
217
....
only the second of these that satisfies
when r = 0.
= 0.
Therefore J.
Again, when a; = 0, that is, when r = & ^, and therefore z,

must be zero. Hence, in order that the assumed form of equilibrium may be possible,
/_ 4 (^) = 0.
The
least value
of
obtained from this equation gives the
when
length of the wire
critical
in the vertical position
and
it first
if
is
shows signs of instability

than this, the vertical
less
position will be stable.

It is
is
found that the least root of
approximately l 88: so that the

-
critical
length
is
about
(?)'
or
1-996 yjjjw,
approximately.
To the degree
or in terms of
ft
of approximation adopted
and W, the whole weight of the

I
determining the
is
critical
convenient
wire will be stable
if
if
wire,
= J80/W = 2-83 J /3/F.
Of the two formulae given the

second
we may put
first
is
the proper one for
length for a given kind of wire;
we wish
to
know whether a given
placed in a vertical position with
the
piece of
its
lower
end clamped.
(See Greenhill, Proc. Camb. Phil. Soc.
Math. Theory of Elasticity,
II.
p. 297.)
IV.
1881, and Love,
218
[XV.
As another simple illustration derived from the theory of

we will give, after Pochhammer and Love (I.e. p. 115), a
elasticity,
discussion of the torsional vibration of an isotropic solid
short
circular cylinder of radius
If
u, v,
(r, 6,
c.
any point of the
z) are the coordinates of
cylinder,
and
the corresponding displacements, the equations of motion
for small vibrations are
3i
dr
30
dz
3ot,
dnr s
'W^ + ^TTe-^li + ^hr

w
9A
d2
lu, 3
-^ 5- (rw 3 )
r or
2u
-
r du
Sot,
where
1 3 (ru)
ldv
dw
7*30
3.3
3r
'
3 (n;)
2OTl=
')
rlsl""
du
2 5r, =
dw
dz~~ dr'
The
r\
3 (rv)
du
~dr~
~dd
stresses across a cylindrical surface
P = \A +
P =
er
Fsr
We proceed to
F is
du
dr'
.
rdd
/du
IJ
'\dz
form of
\r.
dw
dr~
= 0,
= 0, i>=Fei(v* +3,,
a function of r only, and
The
L(l
dr
to obtain a periodic vibration with

is real.
are
construct a particular solution of the type
where
= constant
2/j,
1 du
/x
7,
are constants.
In order
no damping, we suppose that p
torsional character of the oscillation is clear from the
u, v, w.
XV.]
If
we
put, for the
219
moment,
we have
A = 0,
2w2 =
0,
Zd(rV)
and the equations of motion reduce to two identities and
- ppvz = - rfvz
Thus
and since
V must
+î(l^r (rV))
satisfy the equation
d?V
ldV
[pf-(if
dr2
r dr
V must be finite
I)
r2
/j,
when r =
V=AJ
0,
the proper solution
is
( K r)
where
K2
=^p -y\
2
/*
If the curved surface of the cylinder

Prr, Per, Pzr
identically
= c.
must vanish when r
P^
is free,
will vanish if
im-"'
when r = c:
that
is, if
kcJ[ (kc)
or,
which
is
the same thing,
(kc)
0,
if
kcJ3 (kc) = 0,
(see pp. 13, 179).
If k
= 0,
the differential equation to find
d?V
dr
of which the solution
ldV
1 rr
y
r dr
r1
is
V=Ar +
then the stresses
Now P
V is
= o.
and
zr
vanish
220
or, for
[XV.
our present purpose
V=Ar.
This leads to a solution of the original problem in the shape
u = 0,
v
w = 0,
= Are
HyZ+pt)
,
with
and
in particular
u
we have
= r ^cos
Ain
when n =
0, 1, 2,
But we may
nirz
as a special case
mrt
Tl^^TVj
. . .
and
is
also take for
any
is
real constant,
constant,
nirt
//i\
+ BnSin T\/p)>
constant.
a,
tc
J
of which there
III
any one of the
(kc)
real roots of
an infinite number. If ks is any one of these, p

and 7 determined, as a real or pure imaginary
by the equation
we have a
solution
Unless some further conditions are assigned,

whatever value of
tc s
is
taken
that
is
is arbitrary,
to say, vibrations of any
period are possible.
When
the period
is 2ir/p
the axis of the cylinder
the velocity of propagation parallel to
is
P_
ffv>
which is approximately equal to Jfijp so long as ^k\

compared with pp\
If
is
pp
fiK
is
now a damping,
negative the type of vibration

of the vibration as
we go
is
is
altered
small
there
in one direction along
the axis of the cylinder.

Special solutions
conditions
may be
thus for instance
constructed to suit special boundary

if
we put
XV.]
= 2 (A^ cos pt + Bm sin pt) J^ (ic r) sin =-
s,
when
m is a real integer,
and length
21,
ks any root of
mode
= 0,
(ks c)
and
of vibration for a cylinder of radius
the circular ends of which are glued to fixed
parallel planes, the
The doubly
<>
this gives a possible

c
221
curved surface of the cylinder being
number
infinite
determined by suitable
4 m B^
of constants
initial conditions.
For the discussion of the extensional and
flexural vibrations
the reader should consult Love's treatise already referred
memoir
the
Many
of
left free.
have to be
Pochhammer,
to,
and
Crelle, lxxxi. p. 324.
other illustrations of the use of Bessel functions in the
theory of elasticity will be found in recent memoirs by Chree,
Lamb, Love, Rayleigh and
To
others.
we have begun, with the oscillations of a

us modify Bernoulli's problem by supposing that the
conclude, as
chain, let
density at any point of the chain varies as the nth power of
its
distance from the lower end.
Proceeding as on
equation of motion
but measuring x from the
p. 1,
is
xn&V
df
and
if
d (9*n+ 'dy\
dx \n + 1 dx)
'
we put
y = u cos
where u
is
a function of
x,
qx
2irpt,
we have
du
dhi,
+ ldx
y dx
or
d*u
-j-o2
dx
n+
x
du
k+ u = n0,
dx
x
where
k= 2irpJ(n + \)jg.
Assuming
u
= a + a^x + a^x* +
a
...
free end, the
222
we
find that the differential equation is satisfied
K*X
K*X
f,
n+
by
2(ra+l)(ro
+ 2)
2.3(n+l)(n + 2)(n+3)
or,
which
[XV.
the same thing, by
is
UÂx'^J^KX1
).
Finally, therefore,
y
Professor
problem
problem
= Ax-^Jn UTrPA
Greenhill,
to
whom
+ 1)x
this
\ cos 2irpt.
extension of Bernoulli's
due, remarks that to realise the conditions
is
of the
we should take, instead of the chain, a blind

a very large number of small uniform horizontal rods,
practically,
composed of
the shape of the blind being defined by the curves
with x positive.
Thus n =
1 gives a triangular blind,
and
so on.
In connection with the reduction of the differential equation

h
it may be pointed out here, that, if yr~ be substituted
at p. 216,
for u,
and x^kt1
for r, the differential equation of the
form
can be reduced to the standard form

d?y
\dy
dx^xdx +
so that it
is
(
\
!_^
xV y
'
integrable by Bessel functions.
This gives a general rule for the transformation in cases in
which
it is
not so obvious as in the case considered above.
that case
u = p, X = I, n=%,
(See
p.
233 below
fi
l,
for other examples.)
k 2 = 4w/9/8.
In
NOTE.
As
Yn (x), Jn {x)
the determination of the coefficients of
in that
second solution of the general differential equation which vanishes at

infinity is
not without
difficulty,
and the solution
is
of great importance
for physical applications, the following explanation of
ment
of the
problem
Bd. 75, 1873)
(Crette,
may
It has been proved, pp. 59, 60 above, that
such that the real part of zi

z*
Weber's
treat-
not be superfluous.
if
z be complex,
and
negative
is
e*M(l-\y-ld\ = 2 n JvIL(n-%)Jn (z).
Further the investigation given in Chapter
VII.
above shows that
the differential equation
d 2w
dw
v?\
/,
(-5)
is satisfied
also
by taking
,=*/;
For by differentiation
it will
**(X -!)""*<&.
be found that the quantity on the
left
in the differential equation reduces to
-izn - 1 {e
which vanishes when
\= 1,
izk
(\
-l) n+i }^
and when
\= oo
These two definite integral solutions of the differential equation

seem to be due to Riemann, who gave them in his memoir on Nobili's
Rings referred to on
In the second
p.
128 above.
of these solutions,
on the supposition that the path
of integration is along the axis of real quantity from 1 to oo
the real
part of iz must be negative, so that, either z must be essentially
complex, or
n< 0.
The
solution then in the absence of fulfilment of
the latter condition does not hold for real values of

solution which holds for both
real,
z.
We
can find a
and complex values of z as
follows.
224
NOTE.
Let a new path of integration from +

We have
to 0,
and from
to oo i be
chosen.
Now,
moment
for the
in the second integral
Since X
the
is real in
second part
also,
be real, = x, say, and we have putting

where is real
/x.
(l-X*)*"*^ +/""-* (l+^)"-*dJ.
to
= (_a,)./J
let z
for A,
fii
first
integral
use X as variable in the
and we have
w = (- xf j fV"* (1
Real part of
we may
+ A 2 )"~*dX
- Psin(aX)(l-Xs f~ J dxl.
Thus
if
1
u = xn
it is
for
nn(a!X)(l-X ) "*dX,
{ f
/""""*
+ X*)*"*^,
(1
It will hold also
a solution of the general differential equation.
complex values of x provided that the real part of x
is positive.
We now write
1.3.5.^(2^-1)
where
A m Bn
follows.
= A * T* <& + BnJn (X)
This can be done as
are constants to be determined.
Multiplying both sides by x n and putting x

,
on the
for the second term
'
= 0, we have
zero
and therefore
right,
A
tU.s.%~i) =^ ^toThe
x=
0.
term on the right of 5 multiplied by xn vanishes when

Expanding the quantity under the sign of integration in the
first
second term by Taylor's theorem
xn
C e'
xK
"- 1
JX
we
obtain the expression
^^ X
2"- 3
...
+ X*"
^}
d\,
where
(0
<6<
1).
(2n-l)(2n-3)...3
2- 1 (?i-l)!
rr(2 w )
limit
XV
X2 ^- 1
Integrated this takes the form
and the
6\i
when x =
x
of
2n-ir (2n- l)
2
ai"it is
2 "" 1
a;
therefore -
"!''
(2ra),
that
is
- (2w -
1)
!.
NOTE.
225
from the explicit form of Yn (x), 30, p. 14 above, we see that

Ya (x) when x = is - 2"" 1 ( - 1) !. Hence finally
Now
the limit of xn
2 l.8...(h-l) - l.l...(h-l) -^
*" M
tx
(2n-l)!
^ =
or
"Writing m for
d u
what we have
we
2t
"(- 1
,
1.
called u,
S"- 1
r1
Integrating by parts
we have
_t
2-i
get
u.
S-iii-K^-sTrC-w-a-n^*
dx x
-..',-
jo
+
"2w-
*.n+l
'
c"
so
+1
xdxxn
Thus wm+1 is derived from without any change of coeflBcient, and

comparing with the case of n =
(see Examples 15. ..19, p. 229
below) we get
B = y log 2.
Thus
for the values
differential
equation
may be
u.=
where
G.
C is
M.
of
specified
the solution of the general
written by 6 and
7,
C{r.(<B) + ( y - log 2)
./.(*)},
a constant.
15
EXAMPLES.
as on p.
If,
1.
4,
we put
/x.
<j>
e sin
<j>,
prove that
de
and hence obtain

Prove
2.
SP
1-**4>
130
?4>
t
e de
Bessel's expression for
that, in the
problem of
=Q
9/x
'
in terms of
<j>
elliptic
fi.
motion, the radius vector
given by the equation
is
a_ _1
r
-e
+ 2 {Jx
cos
</>
(e)
cos
jj.
(2e) cos 2/t
...},
(see p. 19).
Verify the following expansions
3.
e = JQ (x) + i {( + >A?Tl
(i)
(ii)
cosh nx
=J
sinh
= 22
race
where
<j>
cos wa;
(iii)
sin
where
<j>
cosh
(x)
+ ( -
+ 25 cosh s$Ja
sinh
V^+T)
(x),
J,
(a;).
[s
[s =
s<,7, (a),
2, 4,
. . .
1, 3, ...]
= sinh -1 n.
=J
(x)
nx = 22
_1
w,
is
+ 23
(-)*
(8_1)
(-)
cosh
cosh
s<j>Js
s<f>Je (x),
(as),
being supposed greater than
8
Y = J3 logx-- ---lJ -5j
?
[s
2, 4,
. .
[s
1, 3,
. .
.]
.]
1.
(iv)
3 ( 4* +
1)
2 (2*+ 1) (2+2)
J4,,+1
4S + 3
~4(2 S + 3)
4s+3+
""
s-1 A - J
'
EXAMPLES.
4.
Prove that
1=J*+
J\
2J\ + 2J\+...,
2 (JtJz +
2 J,JS - J\ = 2 (JJ, +
5.
UJt.
227
Show
J^ + J J +...),
2
JVJ, + JJ, +
...).
that
(JT?) = 2jT? {J, (b) J> (c) - 3 Js (b) Js (c) +
6.
5 J, (b)
(c)~... }.
Prove that
fi
A (t)W-2.nir
fi+
6
c-'i
2w+5
2ra + 3
&
2(2w + 2)'
e
2.3(2+2)'
(2n+5)(2n+7)
2.3.4.(2rc+2)(2w +
4)
(2w + 7) (2w + 9)
2.3.4.5.(2ra+2)(2ra + 4)
Verify the following
7.
dynamics, vol. II.)
[
_
K W-)o
,
JT (ax) cos
bxdx =
results
(taken from
aJ
da
(ax)
1+a2
2
Jir (a
2)
*+
}
Hydro-
Basset's
'
"*,
"0
-CO
I
e~
mK
(bx)
dx=(- a3)
=
f" Z (ax)
(bx)
(a2 - J 2)
dx = (ai +
f e~ m {J (bx)f dx = 2*-[F (k) denotes the

8.
first
/z,2
tan" 1 ^
complete
^Z*!
_i tanh" 1
by i F {b (a
(a 2
""
b")~
+ 45a)-l F \ 26
> a,
< ,
(a2
},
+ 46 2)~*}.
elliptic integral to
modulus k]
Prove that
2
(x)
=-
[
I
cos
(a;
cosh 6) d6,
""Jo
[e-^Y,(bx)dxê-^K
j\- a3*' K
(bx)
dx =
,&
JT
Q,
(|L)
(Basset.)
152
EXAMPLES.
228
[The following Examples
(9
19) are taken from Weber's paper in
Bd. 75, 1873.]
Crelle,
Prove that
9.
=_
&oosfl
cose dS.
Prove that
10.
/CO
J"x (x)
dx =
(ax)
>
=1)
(a
1,
1)
Jo
=h
= 0,
(a 2 >l).
[Substitute the value of J^ (x) from Ex.

respect to
9,
and integrate
first
with
a;.]
Prove that
11.
^J dX= T
K sin(Xx)
e-'
6
ek
e'
Hence show by integrating
sin (\x)
first
dX j"
sin (Xr)
./ (r)
ir.
with respect to X on the right and
having regard to 151, p. 73, that

.
r sin
(a;)
=-
(Aas)
v
dX,
and therefore
ie
(
^Mdx=r**Mdx.
h jlTx*
12.
A Jx*-i
Prove that
log x
(x)
dx = -(y + log
2),
Jo
y=
where
~
I
x
e~ logxdx, (seep.
40 above).
Jo
[By the
last example,
y.00
I
_ea!
log kJ. (x) dx =
Jo
and the second

theorems.
Or,
integral on
Jo
e"
ra
log a; sin
(Xai)
dx,
the right can be evaluated by
00
of log
known
x given by the equation
e -wc
du
Jo
first
/.CO
I
1 Jo
more simply, the value

loga;=
may be
J\
--=
^X -
substituted in the given integral,
and the integration performed
with respect to x then with respect to
w.J
EXAMPLES.
229
Prove that
13.
f
Jo
(x)dx = \og2-y.
[Use Neumann's series for Y p. 22 above, and the theorem, 149,

with the result of the preceding example.]
,
p.
73,
Establish the equations
14.
rsin(aX)
sin (aX)
'
w
.
Jo
r ô
Jo
[
(Xx)dX
(Xx)
dX _ sinh&a
sinh ha ["
*"
Jo
e-**
~~kh Jv-i
(aX)J { X,)
,00
dx = coshka
* 2 + A.2
J1
(C)
./o
WTx*
(x
p-kx\
> a)
dX
J\*-\
^=2- T -^.(tfa)
A'
'^^Mdx--^ M
(x
< as).
ik x)
[Use the theorems

/cosa;X
Jo
15.
dx =
Tx>
/Xsin(a:X)
-to
it
~l^Jr dx
'
00
s: Jl-X
which z may be
real or imaginary,
isX
dX
Jx*-i
but
always such that the real
is
the Bessel function of zero order.
If (see Ex. 15)
2 f"
e"*
-dX,
IT
Jx
Jx*-i
prove that for z real and positive (=
x,
say)
2 f" cos(a:\)
/H=-J j={dK + iJ
,.
7>
(x),
and for
both satisfy the differential equation of
part, if any, of iz is negative)
16.
-^
tt
Prove that the definite integrals

dX,
(in
a real
and negative (= ..
x),
/'
cos(a;\)
EXAMPLES.
230
17.
If
we
write, as
from Ex. 15 we are entitled to
f(z)=A Y (z)+J}J
f{~ x)~f (x) AiriJ
prove that
A=
and therefore
do,
(z),
(as),
2
IT
by the theorem
of Ex. 16.
[Change from +x to x along a semicircle round the origin and

have regard to the term J (x) log x in Y (as).
18.
Prove that
I
/(as)
dx = i,
Jo
and hence by Ex. 13 and the theorem
149, p. 73, that
EXAMPLES.
21.
Prove that
V=
if
-1
27T
djx.
J
then
231
e-f* cos
Xv cos /xvJ
(jxts)
dv,
EXAMPLES.
232
Prove that
25.
^^ = Jl-3Jl+5J!-....
(Lommel.)
Prove that
26.
if
denote
-j-
then
D m {x~ in Jn (} = (-
X~ Hn+m) Jn+m (Jx),
D{x*"Jn (Jx)} = $) m xi(n
- m)
Jn - m (Jx).
(Lommel.)
Prove that the equation
27.
*r + *r +F =o
dy*
da?
is satisfied
by
F= (4 m cos <^ + Pm sin mf) (- 2p)

where x
= p cos
Show
<,
also that
{p
- {p
a)
lp
(/,)
(j
2 ( /,)
0*)
lp W + ..j
Show
0*)
by
.}'
for
= {J (JT^)f,
{Jm (vl
/a )}
2
2
.
that the equation
cPE
dx2
is satisfied
and obtain a corresponding expression

28.
^^
= p sin <.
2dR
R _ n{n+l)
x dx
a?
'
either of the series
(- l) n xn
1.3...(2m+1)
+ 1)
1_ _
a?
\(
l.(2w+3)
1.(2m+3)
2
1
(-l)1.3...(2n-l)
a;"
+1
(2 + 3) (2m +
1.
5)
...},
rf
(2m- 1)2
a*
1.2. (2m-
1)
(2m-
3)
Express un as a Bessel function; and show that
M-(+D = _(+!).

EXAMPLES.
233
Verify the following solutions of differential equations by
29.
means
of Bessel functions
If
(i)
+i
Jam+i
m *1 + V
da?
>
then
2m+l
y=
when
x~~~-
2to
% Cp {J-m-i
^...a^ are the
a,,,
(2upN/a3)
Wm+i (2ap Jx)},
am+1 =+i, and C
roots of
C1}
...
C^
are
arbitrary constants.
If
(ii)
2n-\ dy
d?y
dx2
dx
+y~
'
then
y = x[AJn (x) + J- n
(x)].
If
(iii)
a?Pi- (2np -l)x^- + Ffx^y = 0,

then
n
B
y = x* [AJn (yx ) + BJ- n (yxf )]
>
If
(iv)
a?
+ (2a - 2j3n +
1)
^+
{a (a
- 2) + /?V^} y = 0,
then
2/
(v)
(a
= ^"- a
Deduce from
(iv)
[^n
that
yx
p
)
+ ^_( ya
)].
if
form of Biccati's equation)
A
-
ofl
/A
T
Oft
and solve
2
<fc
(Lommel.)
EXAMPLES.
234
Prove that
30.
if
is
any integral
d2u
-=-;
when
where
is
a, b
a function of
x,
and
of
.
_.
+ Xu = 0,
if
are constants, then the complete integral of
{* + [i-('-i)^}ir = o
is
(Lommel.)
Prove that the solution of Riccati's equation
31.
x -- ay + by* = cxp
ax
can be made to depend upon the solution of Bessel's equation
w
^d~
+ r-12
dw
dr
or
(AV
+ /7
v
n?)
w = A0,
where n = ajp.
be attached to the lowest end of a

uniform flexible chain hanging vertically, then the displacement at a
point of the chain distant s from the fixed end is, for the small
If
32.
oscillations
a bead of mass
about the vertical,
(A n cos nt +
Bn sin nt)
7W
where
Vn = {nJMY (n/3JM) - JmgY, (npJM)} J (n^^-ms)
-{nJMJ (nfiJM)-*/^^ (npJM))Y
ix
(np*l ,x-ms),
being the total mass of the chain and bead, and
2/iJmg, where
values of
33.
m is the mass of unit length
of the chain.
/8
denoting
How
are the
to be determined 1
Assuming that
(x)
vanishes
when x = 24, show
that in
a V-shaped estuary 53 fathoms (10,000 -f 32-2 ft.) deep, which communicates with the ocean, there will be no semi-diurnal tide at about
300 miles from the end of the estuary.

(See p. 113
et seq.)
'
EXAMPLES.
The
34.
temperature of a homogeneous solid sphere of
initial
by
radius a is given
= Ar'*
prove that at time
its
cos 6 (sin
temperature
mr mr cos mr)
is
u = v e~ mm
provided that in
235
a root of the equation
is
(ah
2k) (ma cot ma 1) = m?a?k,
h being the internal and surface conductivities, and the surrounding
k,
medium being
at zero temperature.
(Weber.)
A spherical bell of radius c
35.
is
vibrating in such a
the normal component of the velocity at any point of
Sn cos k at,
where
Sn
manner that
its
surface
is
a spherical surface harmonic of degree n, and a

the velocity of transmission of vibrations through the surrounding
is
is
Prove that the velocity potential at any point outside the
air.
bell at
a distance r from the centre, due to the disturbance propagated in the

air outwards, is the real part of the expression
_ t e*W-r+c)
r
(1
fn(ikr)Sn
(ike) ikef
+ ikc)fn
(ike)
where
Pn denoting the
zonal harmonic of degree n.
Show that the

except when n=\.
resultant pressure of the air on the bell is zero
manner as a rigid body about a

which is at a given distance from a large
perfectly rigid obstacle whose surface is plane ; determine the motion
sphere
is
vibrating in a given
position of equilibrium
at
any point in the

'
A sector of
36.
two
air.
an
flexible
membrane which
so that the velocity at
centre as origin
line,
c
infinitely long circular cylinder is
rigid planes inclined at
are
r, 6, is
and the
is
an angle 2 a, and
is
forced to perform small normal oscillations,
any
point,
whose coordinates, referred to the
bisector of the angle of the sector as initial
qr p cos p6 cos net, where
pa = iir,
the velocity of propagation of plane waves in
time
t,
cylinder
bounded by
by a
closed at one end
the velocity potential at any point
being an integer and

air.
(r, 6, z)
Prove
that, at
of the air in the
is
20730*
cos p6
**
2t
k
../l,,,,
(nW-p^Jpfoa)
cos net
kz
{e '
'
or sin kz\,
'
EXAMPLES.
236
where J'v
gives the requisite values of ri, a being the radius

and where k is a real quantity given by the equation
2
the upper and lower sign before k corresponding to the
(ria)
of the cylinder,
n' 2
= n* k2
first
and second term in the bracket
respectively.
A given mass of air is at rest in a circular cylinder of radius c
37.
under the action of a constant force to the
axis.
Show
that
if
the
any sub-
force suddenly cease to act, then the velocity function at
sequent time varies as
where a
(kc)
38.
circular
right
which
is
liquid at time
air, the summation extends to all

and the square of the condensation
0,
cylinder of radius a
its axis.
is filled
with viscous
and made to rotate with uniform

Prove that the velocity of the
is
2C S
different values of
Show
initially at rest,
angular velocity a about
where the
neglected.
liquid,
JAkr)
the velocity of sound in
is
values of k satisfying
is
also that
if
+WT
XJ^Xa)
'
X are the roots
of the equation
(Xa)
= 0.
the cylinder were surrounded by viscous liquid
the solution of the problem might be obtained from the definite

integral
a
f
Jo
dx
<
Xutj> (u)
(Xu)
(\r) du,
Jo
by properly determining
<
(w) so as to satisfy the
boundary
conditions.
39.
In two-dimensional motion of a viscous fluid, symmetrical
with respect to the axis r 0, a general form of the current function is
where
An n
,
40.
are arbitrary complex quantities.
(Of. p. 116.)
right circular cylindrical cavity
whose radius is a is made

prove that the frequency p of the electrical
oscillations about the distribution of electricity where the surface
density is proportional to cos s$, is given by the equation
in
an
infinite
conductor
j;(pa/v)
where v
= 0,
the velocity of propagation of electromagnetic action through

the dielectric inside the cavity.
is
EXAMPLES.
237
Prove that the current function due to a fine circular vortex,

and strength m, may be expressed in the form
41.
of radius c
mra
Jo
,'o
the upper or lower sign being taken according as
z-z'
is
negative or
positive.
42.
magnetic pole of strength
placed in front of an iron
is
and thickness c if m be the origin

rectangular coordinates x, y, and x be perpendicular and y parallel
the plate, show that O the potential behind the plate is given by
plate of magnetic permeability

of
to
fx.
the equation
p
where
43.
l-p2e-
Jo
i/r
V at
2c(
'
=
+
right circular cylinder of radius
forwards with velocity

prove that
'
a containing
right angles to its axis,
is
air,
moving
suddenly stopped
the velocity potential inside the cylinder at a point
distant r from the axis,
and where the radius makes an angle 8 with

is given by the equation
the direction in which the cylinder was moving,
= - 2 Fcos 6
\1i
J (kt)'
cos nat,
JT )\(KO)
x
where a
all
is
the velocity of sound in
and the summation
air,
values of k which satisfy the equation
44.
Prove that
if
J-l (ko)
is
taken for
= 0.
the opening of the object-glass of the telescope
in the diffraction problem considered at p. 178 above be ring-shaped
the intensity of illumination produced by a single point-source at any

point of the focal plane
is
proportional to
{J,{z)- P
4
(1
if
z = 2Trr/\f,
where
-p*f
is
V z)f
a3
the outer radius,
pR
the inner radius of the
opening, r the distance of the point illuminated from the geometrical
image of the source,
andy
the focal length of the object-glass.
45.
Prove that the integral of the expression in the preceding
example taken for a line-source involves the evaluation of an integral
of the
form
Jo
x J a? i 2
EXAMPLES.
238
Show
46.
that
J*n (ax)JnK
n (bx) =
K
'
!+ sin d>dd>.
Y V
a?
+ b*-2ab
cos
<p
Hence prove that
47.
[" J,IV(ax)
J,1V(bx)
ab f* sin ( v
.
Ja? + 5 2 - 2ab cos <)
,
-^-sm^dtp.
a? + b 2 - 2ab cos <f,
TzJo
-!-dx
xJx
Jz
7=
2"jTrIL(n-l)Jo
'
-*
(Struve.)
48.
solid
sphere
isotropic
and
is
strained symmetrically in the
show
u be the strain at distance r from the centre, k and n the
bulk and rigidity moduli, and p the density, the equation of motion is
radial direction
that
then
is
left to
perform radial oscillations
if
k + ^n
2
/d*u
/d
u
dt?
VSr
4 8w\
du\
dr)
'
with the surface condition

(k
+ &n)
+ 3k- =
0.
Prove that the complete solution subject to the condition stated
=2
+2
J".
Vl
(%)
^
|
(4r>
| 0&.)
sin c
+ J'
|-3 (4P
is
cos c.t)
sin Cp
+ 5; cos
c,*)
where
cp being the pth. root of the equation
(i
Malfoj^ + n^M-o,
\rf
= <?pa*/(k + |m)] which
p.
53 above], prove that,
Show
that for the motion specified
1>~
if
p = B'p =
and [using
the initial values of ru, ru be
(a
p
49.
holds at the surface r = a of the sphere.
(Ve)?r
<p (r),
\j/
108,
(r),
A '
JP*
{J
%
p )fr dr
(r,
Obtain the equation of motion of a simple pendulum of
variable length in the form
jd 2 6
-dldO
EXAMPLES.
239
and show that if I = a + bt, where a and b are constants, the equation of
motion for the small oscillations may be written
dhi
+ U = >
cW
where
u = W,
u by means
Solve the equation in

that
when
bjsjga
is
m being arbitrary
of Bessel functions,
-isHv S-)
/
constants.
The problem
(See Lecornu, G. R. Jan. 15, 1894.
the swaying of a heavy
body
let
If the functions C" (cos
50.
(1
and prove
small,
-'( i
p and
x = gljb*.
down by a
<j>)
are defined by the identity
+ a2 )-" = 5
-2a cos
suggested by
is
crane.)
C8n (cos<)as
s=0
prove that
Jn (J a
(a
+6
+
a
2
ft
- 2ab
cos
<)
-2a6cos^
= 2n
(n
1)
+ .)
%^> % C?
(cos *).
(Gegenbauer.)
Prove that
51.
,,
if
n>m >
,.,
&
(a
-J 2 )"-'"- 1
1
if
a>
and that the value
of the integral is zero if a
< b.
(Sonine.)
52.
Ifm>-
Jm (ax) Jm (bx) Jm (ex) x

[(a
~m
dx
+ b + c)(a+b-c)(b + c-a)(G + a-b)]

3- 1
n (m - ) am6rocm
Jit 2
this is
+a
6,
a+
'
provided that 6 + c-a,
m-i
are
not the case the value of the integral
all positive
and that
if
is zero.
(Sonine.)

240
EXAMPLES.
53.
Prove that
_
sin (u
+ r)
(Sonine and Hobson.)

54.
(i)
Verify the following expansions

e
rcose
(r sin 0)
^)
(r sin 0) 1
=2
-.
Pn (cos 9),
^.
tt
with the notation of example 50.
(Hobson, iVoc. L. M.
55.
S.
xxv.)
Prove that
V/ ""Jo
Jw
(r sin 0)
J\ \\ 'j..
sin 0d0 =
"+
y
,
vk (r
sin 0) sin+*0
d =
(ibid.)
FORMULA FOR CALCULATION OF THE

ROOTS OF BESSEL FUNCTIONS.
In a paper, shortly to be published,
"On
the roots of the Bessel
and certain related functions," for a MS. copy of which we are
MMahon
debted to the author, Professor J.
important results, the
The sth
(i)
which has already been given in
first of
root, in
(as)
4(m-l)(7m-31)
- w _fl_!!Ld
32 (m
part.
order of magnitude, of the equation
Jn
;,
1)
(83m2 - 982m + 3779)_

15
64 (m -
1)
(8/3)
(6949m3 - 153855m2 + L85743m - 6277237)

10 5(8) 7
where
(ii)
/3
The
= ^7r(2w + 4s
1),
m = 4w
2
.
sth root, in order of magnitude, of the equation
*n
=y-
m+3
4 (7m2 +
"8T
82m -
9)
3 (8y)
32 (83m3 + 2075)n2 - 3039m + 3537)
I5(8yf
y = i"" (2 +
where
(iii)
The
4s
1),
s-
m = in".
m+7
_
X- y -~%y
(si
1S
4 (7m 2 +
154m +
95)
WJEyf
32 (83m 3 + 3535m 2 + 3561m + 6133)
15(8# i
where, as above,
y = \ir (2n + is +
G.
M.
in-
obtains the following
m = 4n
1),
16
FORMULA FOR CALCULATION OF THE
242
(iv)
The
Yn (x) + (y-\og2)Jn (x) =

is
given by the series for
in
if
(i)
/?
- \t be therein substituted
for p.
(v)
The
*-{Yn (x) + {y- log 2)Jn (x)}=0

is
given by the series for s#> in
(ii) if
y - |ir be therein substituted for y.
[The y in the expression here differentiated is of course Euler's constant,

and is not to be confounded with the y in the expression for the root.]
The
(vi)
Gn (px) _
Gn (x)
where -
Gn (x) = Yn (x)
where
J^
_
8=
a
()
Yn (x)
Jn (x), or,
Yn (px) _
Jn (*)
Jn (P)
+ (y - log
p = ^zl
P
8p
which
2)
the same,
is
'
4(m-l)(m-25)( P3 -l)
<7
v
3(8p) (p-l)
32(^-l)K-114m + 1073)(p'-l) ,m-to.
r~
p
3
5(8p )(p-l)
(vii)
The

U)
J'n (x)
is
J'n {f,x)
given by the same formula as in
_m+3
p
(vi),
_4(m
8T'
p>i
but with
+ 46m-63)(p3 -l)
3
3(8p) (p-l)
_ 32(m + 185m -2053m+

2
>
1899)(p -l)
5(8p)=(p-l)
[Of course here also the
functions
may be
used instead of the
functions without altering the equation.]

(viii)
The
=
dx
{x
*w
g{(p*r iJn(p*)}
0,
p>l,
ROOTS OF BESSEL FUNCTIONS.

is
also given
by the formula in
_m+7
p~
>
8P
243
but with
(vi),
_ 4(m2 4-70m-199)(p3 -l)

q~
H&pf(p-i)
3
_ 32 (m + 245ot2 - 3693m + 4471) (pB
5(8p)'(p-l)
[As before the
may
functions
1.
Examples
the equation in
of
the equation
of this equation
2.
7 functions.]
useful.
and elsewhere above.
equivalent to
is
tan x =
which occurs in
are found in all kinds of
(i)
physical applications, see pp. 56, 96, 178, 219,
When n =
1)
here replace the
The following notes on these equations may be
'
many problems
x,
The
(see pp. 113, 191, above).
can therefore be calculated by the formula in
The equation
of
which the roots are given in
importance for physical applications, for example
(ii) is
it
roots
(i).
also of great
gives the
wave
lengths of the vibrations of a fluid within a right cylindrical envelope.

It expresses the condition that there is
pp.
no motion
of the gas across the
[See Lord Rayleigh's Theory of Sound, Vol. n.
cylindrical boundary.
265269.]
When n = \,
the equation
is
equivalent to
tan x
and when w = f,
it is
2x,
equivalent to
3a;
tan * =
3-^'
and other equivalent equations can be obtained by means of the Table

on
p.
3.
42 above.
The
roots of the equation given in
problem of waves in a
The equation
motion must
is
fluid
and x =
where a
ku,
J,
the equation
is
the radius.
The
roots there-
equivalent to
tan x =
given also by the equation in
is
is
(See Lord Rayleigh's Theory of
k.
H., p. 231, et seq.)
When n =
equation
are required for the
the expression of the surface condition which the
fulfil,
fore give the possible values of
Sound, Vol.
(iii)
contained within a rigid spherical envelope.
(i)
x,
when n = f.
Again when n = f the
equivalent to
tan x =
2x
,
-a2
'
which gives the spherical nodes of a gas vibrating within a spherical

envelope.
162
BOOTS OF BESSEL FUNCTIONS.
244
4.
The
many
roots of the equation in (vi) are required for
problems, for example the problem of the cooling
of a
physical
body bounded by
two coaxial right cylindrical surfaces, or the vibrations of an annular

membrane. (See p. 99 above.) The values of x and px are those of
The roots of the
Ka, Kb, where a, b are the internal and external radii.
equation thus give the possible values of k for the problem.
5.
The
roots of the equation in (vii) are required for the determin-
ation of the
two
wave lengths
of the vibrations of a fluid contained between
coaxial right cylindrical surfaces.
As
for this annular space.
where
a, b
before,
It
is
the proper extension of (hi)
x and px are the values
of Ka, k&,
are the internal and external radii.
In (viii) the equation given is derived from the conditions

6.
which must hold at the internal and external surfaces of a fluid vibra-
two concentric and fixed spherical surfaces.

and px are as before those of Ka, xb, where a, b are the
ting in the space between
The values
of x
The
internal and external radii.
roots thus give the possible values
of k for the problem.

7.
If for low values of s the formulae for the roots are
not very convergent,
it
may be
any of them
preferable to interpolate the values
from Tables of the numerical values of the functions,
if
these are avail-
able.
In conclusion,
as calculated
it
may
be stated that the ten
by Dr Meissel and given
first
roots of
are
= 240482 55577
= 5-52007 81103
ks = 8-65372 79129
&!
k,,
K
ks
k6
h,
=11-79153 44391
= 14-93091 77086
18-07106 39679
=21-21163 66299
ka = 24-35247 15308
k s =27-49347 91320
*,= 30-63460 64684
while, for larger values of n,
h = (n~l)* + *i8 - Aj8 + A 8 - h^ +h

5
where
(x)
= 0,
in the paper referred to below,
. .
Log/H = 8-59976 01403

Log 2 = 7-41558 08514
Log s = 6-90532 68488
Log 4 = 6-78108 01829
Log hs = 6-92939 63062
and LogAj means the common logarithm of
h^ increased
by
10.
EXPLANATION OF THE TABLES.

Dr
Table
I. is
tionen I
und
for 1888.
We are indebted to
a reprint of
Meissel's " Tafel der Bessel'schen Func-
/," originally published in the Berlin
Dr
Abliandlwngen
Academy
Meissel and the Berlin
only change that has been
74 and
of
made
is
(x)
and
(x) instead
Three obvious misprints in the column of arguments
I.
have been corrected; and the value of
from '3932... to
Dr
to write
of
The
Sciences for permission to include this table in the present work.
(l
71)
has been altered
from
3922... in accordance with a communication
Meissel.
Table II.
MS. very kindly
derived from an unpublished
is
by
at our disposal
its
author,
Dr
for
It gives,
Meissel.
placed
positive
n and x, all the values of Jn (x), from x = 1 to

The table may be used, among
are not less than 10~ 18
integral values of
a;
= 24, which
Jn (x) when
other purposes, for the calculation of
Thus
if
x=y+
h,
lies
between two consecutive integers
y,
is
not integral.
y+\ we may
put
and then
A2
Jn (x) = Jn {y) + hJi{y) + j
j;(y) +
...
We
n (y)
+h \z J (y) - J
(y)}
take this opportunity of referring to two papers on the Bessel

Dr Meissel contained in the annual reports on the Ober-
functions by
Realschule at Kiel for the years
188990 and 18912.
shown, among other things, that, when x

value of
for
which the function
Jn (x)
is
given, there
It
is
is
there
a special
changes sign for the last time
from negative to positive; that the function then increases to

absolute maximum, and then diminishes as n increases, with ever
its
in-
creasing rapidity.
Table III., which

to,
gives the first
is
taken from the
the papers just referred
first of
50 roots of the equation
(x)
= 0,
with ^he corre-
EXPLANATION OP THE TABLES.
246
sponding values of
values of
(x)
(x),
which
are, of course,
maximum
or
minimum
according as they are positive or negative.
Tables IV., V., and VI. are extracted from the Reports of the
and 1893.
The Association
was thought too long to reprint, so the
British Association for the years 1889

table corresponding to V.
tabular difference has been taken to be -01 instead of -001.
These
any special explanation the functions / are the

same as those denoted by that symbol in the present work.
tables do not require
Table
X
I.
247
248
Tabt.e
J.W
I.
{continued).
-Ji()
-/,(*)
o-8o
0-846287352750 - 0-368842046094
o-8
0-842579716344 -0-372680644052
21
0-666137120084 - 0-500829672641
0-82
22
0-661116273214
23
-0-503333567025
0-656070571706 -0-505800572628
0-84
0-838833832154 -0-376491556779
0-835049978414 -0-380274508136
0-831228436109 - 0-384029224303
24
0-651000385275
-0-508230524394
0-85
0-827369488950
-0-387755433798
25
0-645906085271
-0-510623260320
o-86
26
0-640788044651
-0-512978621467
27
0-635646637944 -0-515296451971
o-88
0-823473423352 -0-391452867506
0-819540528409 -0-395121258696
0-815571095868 -0-398760343044
28
0-630482241224 -0-517576599061
0-89
0-811565420110 - 0-402369858653
29
0-625295232074 -0-519818913063
0-90
0-807523798123
-0-405949546079
30
0-620085989562
- 0-522023247415
0-91
0-803446529473
o-7993339i6288
-0-409499148347
-0-413018410976
31
0-614854894203
- 0-524189458680
32
0-609602327933
-0-526317406556
33
0-604328674074
-0-528406953885
0-94
0-795186263226 -0-416507081996
0-791003877452 -0-419964911971
34
0-599034317304 -0-530457966666
o-9S
0-786787068613
-0-423391654020
35
0-593719643626 -0-532470314063
0-96
0-782536148813
-0-426787063833
36
0-588385040333
0-97
0-778251432583
-0-430150899695
37
0-583030895983
-0-536378505258
0-98
0-773933236862 - 0-433482922506
0-769581880965 -0-436782895795
38
0-577657600358
- 0-538274103303
'39
0-572265544440 -0-540130544481
00
0-765197686558
40
0-566855120374 -0-541947713931
'01
0-760780977632
'41
0-561426721439 -0-543725500014
42
'55598o742oi4 -o-545463794323
43
o-55o5i7577543
'04
0-751851323654 - 0-449657657556
O747339037965 - 0-452793929666
44
0-545037624510 -0-548821490179
1-05
0-74279SS56434 -0-455896789778
45
0-539541280398
ix>6
0-738221214269 - 0-458966020374
46
1-07
0-733616348841
-0-462001406715
47
0-534028943664 -0-552019999133
0-528501013700 -O-553559322039
ro8
0-728981299655
- 0-465002736858
48
1-09
0-724316408322 -0-467969801675
49
0-522957890804 -0-555058570983
o-5i7399976i46 -0-556517660374.
no
0-719622018528
- 0-470902394866
50
0-511827671736
0-714898476008 '-0-473800312980
0-710146128520 -0-476663355426
51
0-506241380391
viz
52
-0-559315034582
0-500641505700 - 0-560653164677
1-13
0-705365325811
0-495028451994 -0-561950825786
0-700556419592
-0-479491324496
- 0-482284025373
53
1-14
54
0-489402624312
- 0-563207948806
viS
0-695719763505
-0-485041266154
55
0-483764428365
-0-564424467949
1-16
0-690855713099 -0-487762857858
56
0-478114270507 - 0-565600320742
1-17
0-685964625798 - 0-490448614448
0-681046860871 - 0-493098352841
57
0-472452557702
-0-566735448033
i-i8
58
0-466779697485
1-19
0-676102779403 -0-495711892924
59
0-461096097935
-0-567829793994
-0-568883306126
0-671132744264 -0-498289057567
i-6o
0-455402167639
- 0-569895935262
083
0-87
0-92
0-93
0-99
I
1-02
1-03
I
ii
-0-440050585745
-0-443285761209
0756332080477 -0-446488193730
0-671132744264 -0-498289057567
- 0-534443868418
-0-547162491686
-0-550440691132
-0-557936507910
Table
To (a)
I.
249
(continued).
JoW
-M*)
-M*)
i-6o
0-455402167639
-0-569895935262
2-00
0-223890779141
-0-576724807757
i-6i
2-01
0-218126821326
- 0-576060090955
2-02
0-212369710458
i-6 3
0-449698315660 (-0-570867635566
0-44398495 1 500 - 0-57 1 798364542
0-438262485071 '-0-572688083032
2-03
0-206619845483
-0-575355433450
-0-574610928248
1-64
0-432531326660
-0-5735367552I7
2-04
0-200877624399
-0-573826671543
r62
1-65
0-426791886896
-0-574344348624
2-05
0-195143444226
-0-573002762707
i-66
0-421044576715 -0-575110834122
0-415289807326 -0-575836185927
0-409527990183 -0-576520381599
o 403759S36945 -0-577163402048
2-06
0-189417700977
-0-572139304279
2-07
0-183700789621
-0-571236401957
2-08
0-177993104055
2-09
0-172295037073
-0-570294164587
-0-569312704151
0-166606980332
-0-568292135757
2-12
0-160929324324 -0-567232577628
0-155262458341 -0-566134151091
1-67
r68
1-69
170
0-397984859446
-0-577765231529
2-10
171
2-II
i'73
0392204369660 -0-578325857645
0-386418479668 -0-578845271345
0-380627601627 -0-579323466925
2-13
0-149606770449
- 0-564996980564
174
0-374832147732
- 0-579760442028
2-14
0-143962647452
-0-563821193544
175
0-369032530185
-0-580156197639
2-15
0-138330474865
- 0-562606920596
176
0-363229161163
-0-580510738087
2-16
0-132710636881
177
178
179
0-357422452782
-0-580824071043
0-351612817064 -0-581096207515
0-345800665906 -0-581327161851
2-17
0-127103516344
-0-561354295339
- 0-560063454436
2-18
0-121509494713
2-19
o-i
15928952037
-0-558734537577
-0-557367687469
r8o
0-339986411043
-0-581516951731
2-20
0-110362266922
-0-555963049819
r8i
0-334170464016
-0-581665598167
2-21
0-104809816503
1-82
0-328353236143
-0-581773125501
2-22
0-099271976413
-0-554520773326
-0-553041009659
r8 3
0-322535138478 -0-581839561397
0-316716581784 -0-581864936842
2-23
2-24
0-093749120752
0-088241622061
-0-551523913451
1-84
1-85
0-310897976496
-0-581849286141
2-25
0-082749851289
-0-548378356647
r86
0-305079732690
2-26
0-077274177765
-0-546750219981
2-27
0-071814969172
-0-545085398603
i-88
-0-581792646910
0-299262260050 - 0-581695060074
0-293445967833 -0-581556569863
2-28
0-066372591512
-0-543384061721
1-89
0-286631264839
-0-581377223803
2-29
0-060947409082
-0-541646381412
1-90
0-281818559374 -0-581157072713
2-30
0-055539784446 - 0-539872532604
rgi
0-276008259222
0-270200771606
-0-580896170703
2-31
0-050150078400 -0-538062693065
3'32
0-044778649952
-0-536217043381
193
-0-580594575158
0-264396503162 - 0-580252346743
2-33
0-039425856288
-0-534335766941
1-94
0-258595859901
-0-579869549389
2-34
0-034092052749
-0-532419049921
i-9S
0-252799247180
-0-579446250290
2-35
0-028777592796
-0-530467081267
1-96
0-247007069667
-0-578982519892
2-36
0-023482827990
-0-528480052675
1-97
0-241219731308
-0-578478431892
2-37
0-018208107961
-0-526458158577
0-235437635298
0-229661184046
-0-577934063221
2-38
0-012953780380
-0-524401596119
-0-577349494047
2-39
0-007720190934
-0-522310565146
0-223890779141
-0-576724807757
2-40
0-002507683297
-0-520185268182
172
1-87
1-92
1-98
1-99
-0-549969642278
250
X
Table
I.
(continued).
Table
X
I.
(continued).
251
252
Table
X
I.
(continued).
253
254
256
258
Table
X
I.
(continued).
259
Table
260
I.
(continued).
-J,{x)
M*)
Jl(*)
io 40
-0-243371750714 +0-055472761849
o-8o
-0-203201967112
+ 0-142166568299
10-41
-0-242805134273 +0-057849210087
o-8i
-0-242214793214 + 0-060217647828
10-43 - 0-241600808767 +0-062577850293
10-44 - 0-240963264405 +0-064929593703
0-82
-0-201770826005
- 0-200320840603
+ 0-144058996415
+ ' I4S93S3988I2
io -42
0-84
-0-198852172014 +0-147795605727
-0-197364983034 +0-149639449122
0-83
i'4S
- 0-240302245833
+ 0-067272655308
0-85
-0-195859438131
10-46
-0-239617840978 + 0-069606813400
o-86
-o- 194335703428 +0-153277381926
-0-238910139979 +0-071931847339
10-48 -0-238179235177 +0-074247537568
10-47
+0-151466762702
087 -OI92793946683 + 0-155071144022

o-88
10-49
-0-237425221101
+ 0-076553665638
0-89
-0-191234337275 +0-156847888004
-0-189657046181 +0-158607454682
10-50
-0-236648194462
+ 0-078850014227
0-90
-0-188062245963 +0-160349686681
10-51
-0-235848254136 + 0-081136367158
-0-235025501155 + 0-083412509421
10-53 -0-234180038696 + 0-085678227191
10-54 -0-233311972068 + 0-087933307849
0-91
10-52
0-92
-0-186450110748 + 0-162074428448
-0-184820816208 +0-163781526274
0-93 -0-183174539542 +0-165470828298
+ 0-167142184528
0-94
-0-181511459461
+ 0-090177540002
0-95
-0-179831756165 +0-168795446850
+ 0-092410713500
-0-230573231989 + 0-094632619458
10-58 -0-229615843992 + 0-096843050272
10-59 - 0-228636409922 + 0099041799642
0-96
-0-178135611325
-0-176423208066
10-55
-0-232421408701
10-56
-0-231508458131
10-57
0-97
0-98
+ 0-170430469041
+ 0-172047106783
0-99
-0-174694730946 +0-173645217675
-0-172950365937 + 0-175224661243
+ o-ioi 228662586
i-oo
-0-171190300407 +0-176785298957
+ o- 103403435462
-0-225567019886 +0-105565915987
1063 - 0-224500600296 +0-107715903254
10-64 -0-223412744130 +0-109853197747
I'OI
io'6o
io-6i
- 0-227635047621
-0-226611876971
10-62
10-65
-0-222303579310 +0-111977601366
-0-169414723099 +0-178326994235
-0-167623824113 +0-179849612465
1-03 -0-165817794883 + 0-181353021005
I-04 -0-163996828161 + 0-182837089204
-
02
-05
-0-162161117996 +0-184301688406
1066 -0-221173235728 + 0-114088917441
1-06
-0-160310859712 +0-185746691967
- 0-220021845238 +0-116186950748
io-68 -0-218849541635 +0-118271507531
10-69 -0-217656460650 +0-120342395515
1-07
-0-158446249891
10-70
- 0-216442739924 +0-122399423927
I'lO
10-71
-0-215208519001
i-ii
10-67
+0-124442403513
10-72
-0-213953939309 +0-126471146550
1073 -0-212679144146 +0-128485466871
1074 -0-211384278663 +0-130485179874
1075 -0-210069489850 +0-132470102543

10-76
+0-187171975260
-0-156567486350 +0-188577415689
1-09 -0-154674768122 +0-189962892696
1-08
-0-152768295436 +0-191328287775
-0-150848269694 +0-192673484480
-0-148914893455 +0-193998368432
113 -0-146968370410 +0-195302827334
I
'12
-14
-0-145008905360 +0-196586750976
15
-0-143036704202 +0-197850031243
-0-208734926518 +0-134440053463
10-77 - 0-207380739286 +0-136394852837
1078 - 0-206007080560 +0-138334322500
116 -0-141051973900 + 0-199092562127
1079 -0-204614104523 +0-140258285937
io-8o
-0-203201967112 +0-142166568299
1-17
-19
-0-139054922470 +0-200314239736
-0-137045758956 + 0-201514962299
-0-135024693407 +0-202694630176
11-20
-0-132991936860 +0-203853145865
1-18
Table
J
{x)
I.
261
(continued).
JW
Ji()
-<M*)
20
-0-132991936860
+ 0-203853145865
n-6o -0-044615674094 + 0-232000474620
'21
-0-130947701315
+ 0-204990414012
ii-6i
'22
-0-128892199715
11-62
II -24
+ 0-206106341416
-0-126825645926 + 0-207200837037
-0-124748254710 + 0-208273812006
11-25
-0-122660241711
+ 0-209325179625
11-65
1
1
1
11-23
- 0-04229447730I
-0-039971051364
11-63 - 0-037645628720
11-64 -0-035318441806
+ 0-232235010376
+ 0-232446303109
+ 0-232634351719
+ 0-232799157379
- 0-032989723038
+ 0-232940723529
11-26
-0-120561823424
n-66 - 0-030659704782 + 0-233059055883
11-27
-0-118453217184
11-67
+ 0-210354855380
+ 0-211362756947
n-28 -0-116334641133 + 0-212348804193
11-29 -0-114206314208 + 0-213312919188
-0-028328619340
+ 0-233154162418
u-68 - 0-025996698919 + 0-233226053376

11-69
-0-023664175616
+ 0-233274741260
+ 0-233300240831
-0-112068456110
+ 0-214255026208
11-70
-0-021331281388
11-31
-0-109921287289
+ 0-215175051739
11-71
-0-018998248037 + 0-233302569105
11-32
-0-107765028918
11-72
"33 -0-105599902872
"34 -0-103426131706
+ 0-216072924488
+ 0-216948575381
+ 0-217801937572
"73
"74
-0-016665307180 + 0-233281745349
- 0-014332690232 + 0-233237791079
-0-012000628381 + 0-233170730054
-0-101243938632
+ 0-218632946448
"75
- 0-009669352567
-0-099053547496 + 0-219441539632
n'37 -0-096855182759 + 0-220227656988
11-38 - 0-094649069469 + 0220991240623
11-76
- 0-007339093458 + 0-232967393973
-0-005010081428 + 0-232831177619
11-39
-0-092435433245 + 0-221732234896
"79
11-40
-0-090214500248
+ 0-222450586415
n-8o + 0-001967173307 + 0-232284734267
-41
-0-087986497163
+ 0-223146244045
11-81
11-42
-0-085751651176
11-82
11-84
+ 0-004288899920 + 0-232056778820
+ 0-006608232761 + 0-231805986948
+ 0-008924943683 + 0-231532402401
+ 0-011238804987 + 0-231236071121
+ O-0I3549589443 + 0-230917041237
11-30
"'35
11-36
"77
1 1
+ 0-233080588274
78 - 0-002682546537 + 0-232671971904
0-000356718505 + 0-232489811743
11-44
-0-081262341601
+ 0-223819158911
+ 0-224469284397
+ 0-225096576153
11-45
- 0-079008334679
+ 0-225700992096
11-85
11-46
-0-076748398145
+ 0-226282492413
n-86 + 0-015857070317 + 0-230575363062
11-47
-0-074482761342
11-87
11-48
-0-072211653982
11-49
-0-069935306115
+ 0-226841039560
+ 0-227376598268
+ 0-227889135543
11-89
+ 0-02275743I9I6 + 0-229414974489
11-50
-0-067653948112
+ 0-228378620665
11-90
+ 0-025049441700 + 0-228983249662
11-51
- 0-065367810637 + 0228845025194
11-91
11-52
11-92
11-94
+ 0-027337022362
+ 0-029619950574
+ 0-031898003653
+ 0-034I70959578
+ 0-036438597013 + 0-226490458847
"43 -0-083510189950
"83
S3
-0-063077124631
-0-060782121280
11-54
-0-058483032003
+ 0-229288322968
+ 0-229708490101
+ 0230105504990
11-55
-0-056180088419
+ 0-230479348310
11-95
"
+ 0-0l8l6l02I385 + 0-230211089083
n-88 + 0-020461216961 + 0-229824273953
11-93
+ 0-228529160587
+ 0-228052770520
+ 0-227554144849
+ 0-227033351083
11-97
+ 0-231742699216
+ 0-038700695332
+ 0-040957034634
11-98 + O-043207395768
u-99 + 0-04545I560353
+ O225925539874
+ 0-225338667993
+ 0-2247299I9I24
+ 0-22409937I266
+ 0-232000474620
12-00
+ 0-04768931O797 + 0-223447104491
11-56
-0-053873522332
11-57
-0-051563565704
11-58
- 0-049250450632
+ 0-230830003018
+ 0-231157454348
+ 0-231461689817
11-59
-0-046934409328
11 -6o
-0-044615674094
11-96
Table
262
J.W
I.
{continued).
-Jl(*)
12 'OO
+0-047689310797 +0-223447104491
2-40
12-01
+ 0-049920430320 + 0-222773200930
+0-052144702973 + 0-222077744768
2-41
(x)
+0-129561026518
-Ji(a)
+ 0-180710246883
I2 04
+0-054361913660 + 0-221360822234
+0-056571848157 +0-220622521586
+0-131360894344 + 0179260532985
+ 0-133146181728 + 0-177794184461
2-43 +0-134916723111 + 0-176311359192
2-44 +0-136672354521 + 0-174812216550
12-05
+0-058774293132 +0-219862933107
2-45
I2'06
+ 0-060969036167 +0-219082149091
2-46
12-02
12-03
-
+0-063155865777 +0-218280263834
I2 o8 +0-065334571427 +0-217457373624
12-09 + 0-067504943560 + 0-216613576726
I2X>7
-
I2-IO
+ 0-069666773607 +0-215748973377
+0-071819854013 + 0-214863665770
+0-073963978255 +0-213957758045
12-13 + 0-076098940860 +0-213031356277
I2'I4 +0-078224537427 +0-212084568463
12-11
12-12
2-42
+0-140138239554 + 0-171765624000
+0-141848173298 + 0-170218500152
2-48 +0-143542557339 + 0-168655711017
2-49 +0-145221235856 + 0-167077423179
2-50
2-54
+0-153373709704
+ 0-158959430343
+0-154954977468
+ 0-157291725265
+ 0-080340564642 + 0-211117504511
2-55
2-56
12-18
+ 0-082446820302 +0-210130276228
+0-084543103331 + 0-209122997309
+ 0-086629213798 + 0-208095783320
12-19
+0-088704952938 +0-207048751691
+ 0-205982021700
+0-156519496560 + 0-155609725554
+0-158067124921 +0-153913608430
2-58 + 0-159597722266 + 0-152203552365
2-59 + 0-161111150104 + 0-150479737058
2-57
2-60 +0-162607271746
+ 0-148742343422
+0-164085952318
+ 0-146991553564
12-21
2"6l
12-22
2-62
+0-092824528115 +0-204895714458
+0-094867972612 + 0-203789952902
12-23 + 0-096900262741 +0-202664861776
12-24 +0-098921205837
+ 0201520567620
+ 0-146884054700 + 0-165483804615
251 + 0-148530861410 + 0-163875024675

2-52 + 0-150161505225 + 0-162251254066
2 53 +0-151775837096 + 0-160612664833
12-16
12-20 +0-090770123171
+ 0-173296917383
2-47
12-15
12-17
+0-138412913587
+0-165547058774 + 0-145227550765
2-63 +0-166990459905 + 0-143450519461
2-64 +0-168416026353 + 0-141660645228
12-25
+ 0-100930610511 +0-200357198756
2-65
+ 0-169823630622 + 0-139858114759
12-26
+ 0-102928286663 +0-199174885273
2.66
+0-104914045507 +0197973759015
12-28 + o- 106887699579 +0-196753953565
12-29 +0-108849062765 +0-195515604234
2-67
+0-171213147086
+ 0-172584452006
2-68
12-30 +0-110797950308 +0-194258848041
2-70 +0-176587888562
I2-3I
+0-112734178832 +0-192983823702
12-32 +0-114657566356 +0-191690671617
271 +0-177885147930 + 0-128787741891
12-33
2 73
12-27
+0-116567932311 +0-190379533851
12-34 +0-118465097559
+0-189050554121
2-69
+ 0-138043115846
+ 0-136215837361
+0-173937423535 + 0-134376469238
+0-175271941729 + 0-132525202454
2-72
+0-179163605667 + 0-126901935099
+0-180423149549 + 0-125005003575
2 74 + 0-181663669309 + 0-123097143211
12-35
+0-120348884405 +0-187703877780
2 75
+0-182885056640
12-36
+ 0-122219116616 +0-186339651802
276
277
278
279
+0-184087205211
+0-124075619437 +0-184958024768
12-38 +0-125918219608 +0-183559146848
12-39 +0-127746745377 +0-182143169785
12-40 +0-129561026518 +0-180710246883
2-80 +0-188701354781
12-37
+ 0-130662229004
+ 0-121178550823
+0-119249424132
+0-185270010670 +0-117309961743
+0-186433370658 +0-115360363124
+0-187577184813 + 0-113400828590
+ 0-111431559278
Table
M*)
-Jx
I.
263
(continued).
(x)
-Ji(a)
(x)
I2'8o
+0-188701354781
+ 0-III43I559278
13-20
+ 0-216685922259 + 0-027066702765
I2-8I
+0-189805784222
+ 0-I09452757I29
+ OTO7464624869
+ 0-105467365986
13-21
12-82
12-84
+ 0-192999696401 + 0-103461184712
13-24
+ 0-216945650832
+ 0-217183496687
+ 0-217399452738
+ 0-217593514066
12-85
+0-194024240934
+ O-IOI4462860OI
13-25
+ 0-217765677921 + 0-016121474234
12-86
12-89
+ 0-099422875508
+ 0-09739II5957I
+0-196976389945 + O-0953SI345I87
+0-197919671360 + 0-093303639994
13-26
13-29
+ 0-217915943717
+ 0-218044313033
+ 0-218150789610
+ 0-218235379352
12-90
+0-198842437136
0-091248252250
13-30
+ 0-218298090319 + 0-005177480555
12-91
+0-199744611493 + 0-089185390809
0-087115265106
+ 0-200626120738
12-93 +0-201486893280 + 0-085038085131
12-94 +0-202326859628 + 0-082954061409
13-31
12-92
13-32
13-34
+ 0-218338932728
+ 0-218357918950
+ 0-218355063505
+ 0-218330383064
+0-190890378823
12-83 +0-191955046298
12-87
12-88
+0-195028593748
+ 0-196012670759
13-22
13-23
13-27
13-28
13-33
+ 0-024878857605
+ 0-022690195350
+ 0-020500932874
+ 0-018311286951
+ 0-013931711237
+ 0-011742214308
+ 0-009553199615
+ 0-007364883118
+ 0-002991207414
+ 0-000806278917
-0-001377090000
-0-003558684713
12-95
+0-203145952399
+ 0-080863404982
13-35
+ 0-218283896439 - 0-005738290927
12-96
+ 0203944106324 + 0-078766327385
13-36
+ 0-218215624587 -0-007915694697
12-97
+0-204721258250
+ 0-076663040627
13-37
12-98
+0-205477347147
12-99 +0-206212314114
+ 0-074553757I68
+ 0-072438689899
13-38
13-39
+ 0-218125590599 - 0-010090682449
+ 0-218013819702 -0-012263041002
+ 0-217880339252 0-014432557586
13-00
+0-206926102377
+ O-O703I8052I22
13-40
+ 0-217725178732 -0-016599019864
13-01
+0-207618657300 + 0-068192057526
+0-208289926385 + O-06606092OI68
i3 03 +0-208939859276 + 0-063924854454
13-04 +0-209568407762 + 0-06I784075III
13-41
13-42
13-43
13-44
+ 0-217548369742
+ 0-217349946004
+ 0-217129943348
+ 0-216888399712
0-018762215954
13-02
-0-027378117768
13-05
+0-210175525783
+ 0-059638797173
13-45
+ 0-216625355135
13-06
0-020921934445
0-023077964423
-0-025230095486
-
+ 0-210761169428 + 0-057489235957
13-46
+0-216340851750
-0-029521821957
13-07
+ 0-055335607039
13-47
-0-031660999316
+ 0-053178126239
13-48
13-09
+ 0-212388847348 + 0-051017009592
13-49
+ 0-216034933785
+ 0-215707647547
+ 0-215359041426
13-10
+ 0-2I2888I97522 + 0-048852473334
13-50
+ 0-214989165880
13-11
+ 0-046684733877
+ 0-044514007788
+ OO423405 1 1 767
+ 0-040164462629
13-51
13-52
+ 0-214598073436
+ 0-214185818679
13-53
+0-213752458244
I3-I4
+ 0-213365886137
+ 0-213821882244
+ 0-214256157060
+ 0-214668683969
13-54
+ 0-213298050815
-0-046491085613
I3-I5
+ 0-215059438525 + 0-037986077278
13-55
+0-212822657111
13-16
+ O-O35805572692
+ 0-033623165893
+ 0-031439073935
13-56
+ 0212326339882
-0-050675756773
13-57
13-58
+ 0-211809163903 -0-052758325976
+ 0-211271195961 -0-054834115851
13-19
+ 0-215428398451
+ 0-215775543638
+ 0-216100856151
+ 0-2I6404320223
+ O-0292535I3878
13-59
+ 0-210712504851
13-20
+ 0-2I6685922259 + 0-027066702765
13-60
+ 0-210133161369 -0-058964557249
+0-211325296943
13-08 + 02I 1867868729
13-12
I3-I3
I3-I7
13-18
-0-033795441703
-0035924941590
0-038049292086
-0-040168286951
-0-042281720622
o 044389388228
0-048586609352
0-056902926099
Table
264
M*)
I.
(continued).
M*)
-Ji()
13-60
+0-210133161369 -0-058964557249
4-00 +0-171073476110
13-61
+0-209533238299 -0-061018810678
+0-208912810407 - 0-063065488629
4-01
13-62
3
'63
3
1
'64
+0-208271954434 -0-065104394233
+0-207610749084 -0-067135331522
-Ji(*)
-0-133375154699
+0-169731671331
-0-134983384921
+0-168373856986 -0-136577042971
4-03 +0-167000179537 -0-138155981458
4-02
4-04 +0-165610786908
-0-139720054543
-0-069158105453
4-05
+0-164205828478 -0-141 2691 1 7950
-0-071172521923
+0-205505853079 -0073178387788
i 3 -68
+0-204764076220 -0-075175510884
13-69 +0-204002372641 -0-077163700040
4-06
+0-162785455058
4-07
+0-161349818877
-0-144321646527
4-08
+0-159899073571
-0-145824831084
4-09
+0-158433374159 -0-147312444762
1370 +0-203220832633 -0-079142765100
4-10 +0-156952877033
1371 +0-202419548383 -0-081112516941

1372 +0-201598613965 -0-083072767489
1373 +0-200758125328 - 0-085023329736
13-74 +0-199898180285 -0-086964017760
4-1
13-65
13-66
+0-206929275015
+0-206227614833
13-67
13-75
-0-142803028980
-0-148784351297
+0-155457739939 -0-150240416070
+0-153948121961 -0-151680506109
4-13 +0-152424183503 -0-153104490110
4-14 +0-150886086277 -0-154512238442
4-12
+0-199018878503 - 0-088894646742
4-15
+0-149333993280 -0-155903623164
1376
1377
1378
1379
+0-198120321493 -0-090815032981
4-16
+0-197202612595 -0-092724993914
+0-196265856970 -0-094624348132
+0-195310161589 -0-096512915397
4-17
+0-147768068780 -0-157278518033
+0-146188478301 -0-158636798515
4-18 +0-144595388601 -0-159978341800
419 +0-142988967659 -0-161303026807
13-80
+0-194335635216 -0-098390516658
4-20 +0-141369384657
+0-193342388402 -0-100256974070
+0-192330533469 -O-IO2II2III0O8
13-83 +0-191300184501 -OTO3955752084
13-81
4-21
13-82
4-22
13-84 +0-190251457328
-0-105787723166
-0-162610734200
+0-139736809960 -0-163901346396
+0-138091415099 -0-165174747575
4-23 +0-136433372759 -0-166430823692
4-24 +0-134762856750 0-167669462485
13-85
+0-189184469514 -0-107607851391
4-25
13-86
+0-188099340348 -0109415965181
+0-186996190826 -0-111211894262
4-26
+0-133080042002 -0-168890553486
13-89
+0-185875143642 -0-112995469678
+0-184736323171 -0-114766523805
+0-131385104536 -0-170093988031
+0-129678221452 -0-171279659270
4-28 +0-127959570912 -0-172447462171
4-29 +0-126229332114 -0-173597293538
13-90
+0-183579855458
4-3 +0-124487685284 -0-174729052013
13-87
13-88
-0-116524890369
+0-182405868205 -o- 1 1 827040446

+0-181214490755 -0-120002902550
I3-93 +0-180005854081 -0-121722222501
4-27
13-91
4"3i
13-92
4-32
-0-123428203590
+0-122734811649 -0-175842638087
+0-120970893423 -0-176937954108
4-33 +0-119196113786 -0-178014904291
4'34 +0-117410656869 -0-179073394724
+0-177537335004 -0-125120686515
435 +0-115614707731 -0-180113333378
I3-94 +0-178780090769
I3-95
13-96
+0-176277722558 -0-126799513414
I3-97 +0-175001390777 -0-128464527879
13-98
I3-99
+0-173708478559 -0-130115574971
+0-172399126347 -0-131752501232
14-00 +0-171073476110
-0-133375154699
4-36
+0-113808452342 -0-181134630112
+0-111992077563 -0-182137196684
4-38 +0-110165771130 -0-183120946756
4-39 +0-108329721631 -0-184085795902
4-37
14-40
+0-106484118490 -0-185031661615
Table
J
(x)
I.
JqW
-M*)
+ 0-106484118490
-0-185031661615
14-80
+0-104629151946
+ 0-102765013033
43 + 0-100891893564
14-44 + 0-099009986107
14-42
-0-185958463314
-0-186866122350
14-81
-0-187754562014
14-40
14-41
265
(continued).
-0-188623707542
-JiW
+ 0-027082314586 -0-206595567180
+0-025015737179 -0-206716471994
+ 0-022948053986 -0-206816724913
14-83 + 0-020879471508 -0-206896329814
14-84 +0-018810196197 - 0-206955292607
14-82
I4'45
+ 0-097119483970 -0-189473486119
14-85
+0-016740434436 -0-206993621235
14-46
-0-190303826889
14-86
+ 0-014670392520 -0-207011325670
-0-191114660960
14-87
-0-191905921406
14-88
+ 0012600276630
+ 0-010530292822
14-49
+ 0-095220581177
+ 0-093313472454
+ 0-091398353204
+ 0-089475419488
-0-192677543276
14-89
-0-206984911980
+0-008460646998 - 0-206940823925
14-50
+ 0-087544868010 -0-193429463596
14-90
+ 0-006391544891 -0-206876171810
14-51
+ 0-085606896092
+ 0-083661701655
+ 0-081709483202
+ 0-079750439794
-0-194161621377
~ 0-194873957618
14-91
+ 0-004323192042 -0206790975716
-0-195566415311
14-93
-0-196238939443
+0-002255793783 -0-206685257736
+0-000189555214 -0-206559041974
14-94 -0-001875318817 -0-206412354539
+0-077784771035
-0-196891477005
14-95
-0-003938623732 - 0-206245223541
14-56
+ 0-075812677046 -0-197523976991
14-96
-0-006000155243 - 0-206057679091
H'97 - 0-008059709376 -0-205849753289
I4-59
+ 0-073834358450 -0-198136390405
+ 0-071850016350 -0-198728670261
+ 0069859852307 -0-199300771592
14-60
+0-067864068323 -0-199852651447
15-00
- 0-014224472827
14-61
15-01
-0-016274084604 -0-204814948148
-0018320704486 - 0-204505666588
14-63
+ 0-065862866820 -0-200384268898
+ 0-063856450617 -0-200895585039
+ 0-061845022913 -0-201386562994
14-64
+0-059828787267
-0-201857167913
15-04
14-65
+0-057807947575
-0-202307366980
1505 - 0-024440598094 -0-203457124785
14-66
+ 0-055782708050 -0-202737129411
+ 0-053753273205 - 0-203146426455
15-06
15-07
-0-026473238057 - 0-203067538060
- 0-028501882349 -0-202657999596
+0-051719847828 -0-203535231400
+0-049682636966 -0-203903519571
15-08
-0030526331722 -0-202228563094
15-09
-0-032546387470 -0-201779284182
1470 +0-047641845902 -0-204251268330
15-10
-0-034561851456 -0-201310220408
+0-045597680133 - 0-204578457081
+ 0-043550345355 -0-204885067267
15-11
15-12
-0-036572526126 -0-200821431239
-0-038578214533 -0-200312978045
1473 + 0-041500047438 -0-205171082373

1474 + 0-039446992407 -0-205436487924
15-13
-0-040578720351
15-14
-0-042573847897 -0-199237334565
+0-037391386420 -0-205681271486
15-15
-0-044563402147 -0-198670276496
14-76
+0-035333435752 - 0-205905422669
+ 0-033273346769 -0-206108933120
1478 + 0-031211325913 -0-206291796530
479 +0-029147579677 - 0-206454008627
15-16
I4-77
15-17
+ 0-027082314586 -0-206595567180
15-20
i4 47
14-48
14-52
I4'S3
I4"54
-
i4 55
i4 57
14-58
14-62
14-67
14-68
14-69
1471
I4-72
14-75
14-80
-0-207008417910
14-92
14-98
-0-010117082484 -0-205621480228
1499 -0-012172071276 - 0-205372895984
15-02
15-03
-0-205104038614
-0-020364130779 -0-204176237900
- 0-022404162240 - 0-203826708006
-0-199784924098
-0-046547188761 -0-198083818818
- 0-048525014094 -0-197478032331
15-18 - 0-050496685220 -0-196852989694
15-19 - 0-052462609949 -0-196208765420
- 0-054420796844 -0-195545435866
266
X
Table
I.
(continued).
Table
II. (continued).
267
268
Table
II.
(conty/meô
^33
Table
II.
(continued).
269
270
Table
II.
(continued).
Table
II.
{continued).
271
272
Table
II.
(continued).
table
n
II.
(Qontinwd).
273
274
Table
II.
(continued).
Table
11
II. (continued).
275
276
Table
II.
(continued).
Table
Jn (2l)
3i
+ 0-00036
+ 0-00014
32
+ 0-00005 30368 13766 205
33
+ o ooooi
34
82263 iooii 863

26858 96763 539
89501 07095 370
65206 65676 388
35
21644 29380 552
36
06940 98925 453

02153 383 6 3 859
00647 I2 45i 956
00188 59081 313
37
38
39
40
+ o-
41
00003 92245 45'

00001 02103 682
42
43
25893 439
06402 159
44
45
01544
00363
00083
00018
46
47
48
49
So
5i
O0O53 35564 35i

00014 66878 120
+ o-
385
716
679
818
00004 139
891
52
188
53
039
008
54
55
002
II.
(continued).
277
278
Table
II.
{continued).
Table
II. (continued).
279
Table
280
The
first fifty
roots of
maximum
or
(x)
III.
= 0,
minimum
with the corresponding
values of
(%).
Table IV.
281
282
Table
X
V. (continued).
283
284
Table VI.
285
286
Table VI.
(continued).
Table VI.
(continued).
287
288
Table VI.
(continued).
BIBLIOGEAPHY.
{Other references will be found in the
TREATISES.
I.
Riemann.
Neumann
text.)
Partielle Differentialgleichungen.
Theorie der Bessel'schen Functionen.
(C).
Ueber die nach Kreis-, Kugel-, und Cylinderfunctionen fortschreitenden Entwickelungen.
,,
Lommel (E.). Studien iiber die Bessel'schen Functionen.

Heine (E.). Handbuch der Kugelfunctionen.
Todhunter (I.). Treatise on Laplace's, Lamp's, and Bessel's
Functions.
Treatise on Fourier Series and Spherical, Cylindrical,
Byerley.
and
Ellipsoidal Harmonics.
Theorie Analytique de la Chaleur.
Fourier.
Ratleigh.
Theory of Sound.
Treatise on Hydrodynamics, Vol. n.
Basset.
II.
Theoremata
Bernoulli (Dan.).
flexili
vi.
Euler.
(Ibid. vii. 99
Sur
de
corporum
oscillationibus
connexorum,
etc.
filo
(Comm. Ac. Petrop.
108.)
Demonstratio theorematum,
Poisson.
MEMOIRS.
and Acta Acad. Petrop.
la distribution
etc. (ibid.
v. pt.
de la Chaleur dans
i.
vn. 162).
pp. 157, 178.)
les
corps
solides
(Journ. de l'Ecole Polyt. cap. 19).
Bessel.
Jacobi.
Untersuchung des Theils der planetarischen Storungen, etc.

(Berlin Abh. 1824).
Formula transformationis integralium definitorum, etc. (Crelle
xv. 13).
Hansen.
Ennittelung der absoluten Storungen in Ellipsen, etc.

(Schriften der Stern warte Seeburg (Gotha, 1843)).
Hamilton (W.
R.).
On
pt.
i.
Fluctuating Functions (Irish Acad. Trans.
xix. (1843) p. 264).
Hargreaves.
On
a general method of Integration,
etc. (Phil.
Trans.
1848).
G.M.
19
290
BIBLIOGRAPHY.
Anger.
Untersuchungen
Schlomilch.
Ueber die
Phys.
ii.
iiber die
Function /*,
etc.
(Danzig, 1855).
Bessel'sche Function (Zeitsch. d.
Math.
u.
137).
Ueber die Nobili'schen Farbenringe (Pogg. Ann. xcv.).

Lipschitz.
Ueber die Bessel'sche Transcendente I. (Crelle lvi.).
Weber (H.). Ueber einige bestimmte Integrale. (Crelle lxix.).
Ueber die Bessel'schen Functionen u. ihre Anwendung,
,,
Riemann.
etc. (ibid. lxxv.).
Ueber eine Darstellung
willkiirlicher
Functionen durch
Ann. VI.).
stationaren Stromungen der Elektricitat in
Bessel'sche Functionen (Math.
Ueber
die
Cylindern
Heine
(E.).
Hankel
(ibid, lxxvi.).
Die Fourier-Bessel'sche Function (Crelle

Die Cylinderfunctionen erster
(H.).
Ann. i.).
Bestimmte Integrale mit
,,
u.
lxix.).
zweiter
Art (Math.
Cylinderfunctionen
(ibid.
VIII.).
Die Fourier'schen Beihen
functionen
Neumann
(C).
u.
Integrale fur Cylinder-
(ibid. viii.).
Entwickelung einer Function nach
Produkten
der
Quadraten u.
Functionen
Fourier-Bessel'schen
(Leipzig Ber. 1869).
Lommel.
deichung
Integration der
durch Bessel'sche Functionen (Math. Ann. II.).

Zur Theorie der Bessel'schen Functionen (ibid. III., IV.).
Ueber eine mit den Bessel'schen Functionen verwandte
Function
(ibid. ix.).
Zur Theorie der Bessel'schen Functionen (ibid, xiv., xvi.).

Ueber die Anwendung der Bessel'schen Functionen in der
Theorie der Beugung (Zeitsch. d. Math. u. Phys. xv.).

Schlafli.
Einige Bemerkungen zu Herrn Neumann's Untersuchungen iiber die Bessel'schen Functionen (Math. Ann. in.).
Ueber die Convergenz der Entwickelung einer arbitraren
Function zweier Variabeln nach den Bessel'schen Functionen, etc. (Math. Ann. x.).
Sopra un Teorema di Jacobi, etc. (Brioschi Ann. (2) v.).
Sull' uso delle linee lungo le quali il valore assoluto di una
,,
funzione e costante (Ann. di Mat. (2) VI.).

Die Fourier'schen Integrale
(P.).
du Bois-Reymond
(Math. Ann.
iv.).
u.
Formeln
BIBLIOGRAPHY.
Mehler.
Ueber die Darstellung einer
291
willkiirlichen Function zweier
Variabeln durch Cylinderfunctionen (Math. Ann.

Notiz fiber die Functionen
Sonine
Recherches sur
(N.).
Ann.
Gegenbauee
(L.).
Pn (cos#)
und J (x)
v.).
(ibid.).
les fonctions cylindriques, etc.
(Math.
xvi.).
Various papers in the Wiener Berichte, Vols, lxv.,

LXVI., LXIX., LXXII., LXXIV., XCV.
Strutt [Lord Rayleigh]. Notes on Bessel Functions (Phil. Mag. 1872).

Rayleigh. On the relation between the functions of Laplace and
Bessel (Proc. L.
Ueber
Pochhammer.
M.
Schwingungen,
Glaisher
(J.
Greenhill.
W.
On
L.).
On
On Height
kleiner
etc. (Crelle lxxxi.).
Riccati's Equation (Phil. Trans. 1881).
Riccati's
Equation and Bessel's Equation (Quart.
Journ. of Math.
S. ix.).
Fortpflanzungsgeschwindigkeit
die
xvi.).
consistent with Stability (Oamb. Phil. Soc.
Proc. iv. 1881).
Bryan.
Chree.
Stability of a Plane Plate, etc. (Proc. L.
The equations
M.
S. xxn.).
in polar and
Camb. Phil. Soc. xiv. p. 250).
Systems of Spherical Harmonics (Proc. L. M. S. xxn. p. 431).
On Bessel's Functions, and relations connecting them with
Hyperspherical and Spherical Harmonics (ibid. xxv. p.
of
an
isotropic
elastic
solid
cylindrical coordinates (Proc.
Hobson.
49).
Thomson
(J. J.).
Electrical
(Proc. L.
McMahon
(J.).
On
Oscillations
M.
in Cylindrical
Conductors
S. xvii. p. 310).
the roots of the Bessel and certain related func-
tions (Annals of Math., Jan. 1895).
eg
c8
^3
a
o
a
o
o
a
ra
s
o
KB
o
W
a,
o a
.a
a
2
a
cS
m
it
Pi
a>
^1
<B
c3
ra
Cambridge: pbinted by
J.
& 0.
F.
clay, at
the university fbess.

Treatise On Bessel Functions and Their Applications To Physics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Treatise On Bessel Functions and Their Applications To Physics

Uploaded by

Copyright:

Available Formats

hail

the Cornell University Library.

There are no known copyright

the United States on the use of the

THEIR APPLICATIONS TO PHYSICS.

THEIR APPLICATIONS TO PHYSICS.

PROFESSOR OF PHYSIOS IN THE UNIVERSITY COLLEGE OF NORTH WALEB

FELLOW OF ST JOHN'S COLLEGE, CAMBRIDGE

AT THE UNIVERSITY PRESS.

in view of the great and growing

importance of the Bessel functions in almost every branch of

of the theory of the functions as is

necessary for their practical application, and to illustrate their

use by a selection of physical problems, worked out in some

inclined to think that the earlier chapters

remembered that the

properties of the Bessel functions are not

without an interest of their own on purely mathematical grounds,

and that they afford excellent

theory of differential equations, and of the theory of a complex

even from the purely physical point of view

impossible to say that an analytical formula

useless for practical

be so now, but experience has repeatedly shown

that the most abstract analysis

the highest importance in mathematical physics.

any, of the analytical theory

included in the present work has failed to be of some use or other

the later chapters

anything superfluous has been

from thinking that

could almost wish

that space would have allowed of a more extended treatment,

complex theory and on definite

regard to that part of the book which deals with physical

aim has been

to avoid, on the one hand, waste of

time and space in the discussion of

any pretension of writing an

and, on the other,

elaborate physical treatise.

endeavoured to choose problems of real importance which naturally

require the use of the Bessel functions,

considerable detail, so as to bring out clearly the direct physical

significance of the analysis employed.

has been that the chapter on diffraction

that this section

attention in this country to the valuable

result of this course

contained in Lommel's memoirs, from which the substance of that

much pleasure that we acknowledge

are indebted to Lord Kelvin and Professor J. J.

free use of their researches

electrical oscillations respectively; to Professor

the help and

A. Lodge for copies

have been extracted; and to the Berlin Academy of Sciences and

Meissel for permission to reprint the tables of J and

appeared in the Abhandlungen for 1888.

the former in manuscript;

generously placed at our disposal the materials for Tables

Meissel has also very

very kindly communicated to us his formulae for the roots of