You are on page 1of 292

CORNELL

UNIVERSITY
LIBRARY

BOUGHT WITH THE INCOME


OF THE SAGE

ENDOWMENT

rUND GIVEN IN 189I BY


HENRY WILLIAMS SAGE
MATHlSMATICS

Cornell University Library

QC

321.C32 1921

Introduction to the mathematical theory

3 1924 001 599 806

The
tine

original of

tiiis

book

is in

Cornell University Library.

There are no known copyright

restrictions in

the United States on the use of the

text.

http://www.archive.org/details/cu31924001599806

ON TO THE MATHEMATICAL THE


:!ONDUCTION OP HEAT IN SOLID

MACMILLAN AND
LONDON

CO., Limited

BOMBAY CALCUTTA
MELBOURNE

MADRAS

THE MACMILLAN COMPANY


NEW YORK

BOSTON CHICAGO
SAN FRANCISCO

DALLAS

THE MACMILLAN

CO.

OF CANADA,

TORONTO

Ltd.

INTRODUCTION TO THE

MATHEMATICAL THEORY
OF THE CONDUCTION OF
HEAT IN SOLIDS

BY

H.

S!

CARSLAW

ScD. (Cambridge), D.Sc. (Glasgow), F.R.S.E.


PROFESSOR OF MATHEMATICS IN THE UNIVERSITY OF SYDNEV
FORMERLY FELLOW OF EMMANUEL COLLEGE, CAMBRIDGE, AND LECTURER IN MATHEMATICS
IN THE UNIVERSITY OF GLASGOW

SECOND EDITION, COMPLETELY REVISED

MACMILLAN AND
ST.

CO.,

LIMITED

MARTIN'S STREET, LONDON


1921

COPYRIGHT
First Published, 1906.

PRINTED IN GREAT BRITAIN

PEEFACE
This volume complekss the new edition

of iny book on Fourier's


and the Mathematical Theory of the Conduction
The original work was first published in 1906 and has now
of Heat.
for some time been out of print.
The first volume of the new
edition appeared towards the middle of this year, and deals with
the Theory of Infinite Series and Integrals, with special reference
to Fourier's Series and Integrals.
This formed a completely new
work with the title Fourier's Series and Integrals. The second
volume is devoted wholly to the Mathematical Theory of the
Conduction of Heat in Solids. This part of the book has also been
completely rewritten and much enlarged. It now includes a discussion of all the important boundary problems associated with the
Equation of Conduction. The treatment of these questions,

Series

and

Integrals

especially in the later chapters, should be of use to those interested

in the application of

modern

analysis to the solution of the differ-

mathematical physics.
the Differential Equation of Conduction

ential equations of

In Chapter

I.

and some general theorems

is

obtained

as to its solution are established. Chapter

The next two chapters are devoted


The principal changes made in these chapters are
cpnnected -with the more exact treatment of the Infinite Series and
Chapters V. and VI.,
Integrals which enter into the solutions.
which deal with Two-Dimensional Problems and Flow of Heat in
II.

deals with Fourier's Eing.

to Linear Flow.

a Kectangular Parallelepiped, differ


chapters in the

little

from the corresponding

first edition.

Chapter VII. deals with the Circular Cylinder, Chapter VIII. with
the Sphere and Cone, Chapter IX. with Sources and Sinks, and

Chapter X. with Green's Functions.


additional matter.

These chapters contain much

PREFACE

VI

Chapters XI. and XII. are quite new.


"

The former

is

entitled

of

Contour Integrals in the Solution of the Equation of

Conduction."

Bromwich's recent work has directed attention

The Use

to the " operational

method

" of Heaviside.

It

is

questions examined in this chapter could be solved

But

to justify the operational

integration,

by me,

and the

method we must

true that

by

rely

all

the

that method.

upon contour

between the method developed


chapter, and the operational method is

chief difference

as illustrated in this

that I prefer in each case to turn to the standard path in the plane
of the

complex variable instead of using ^ kind

of

mathematical

shorthand.

In the

last chapter

Chapter XII.a sketch

is

given of the use

of Integral Equations in the solution of the Equation of Conduction.

This second volume could not have appeared so soon after the first
had I not been privileged to spend this year on leave of absence
from the University of Sydney in my old College at Cambridge.
For the f acihties so fuUy granted to me there I take this opportunity
of expressing

my

heartfelt thanks.

Emmanuel College,
Cambridge,

October, 1921.

CONTENTS
THE MATHEMATICAL THEORY OF THE CONDUCTION OF
HEAT IN SOLIDS

CHAPTER

THE DIFFERENTIAL EQUATION OF THE MATHEMATICAL THEORY


OF THE. CONDUCTION OF HEAT
SECTION

PAGE

1.

Introductory

2.

Conductivity

3.

Flow

of

Heat

across

an Isothermal Surface

4.

Flow

of

Heat

across

any Surface

5.

The Equation

6.

The Transformation

7.

Initial

8.

The Solution

9.

Simplification of the

j
-

4
g

of Conduction

of Coordinates

jq

and Boundary Conditions is

12

Unique

14

Problem

Ig

CHAPTER

II

FOURIER'S RING
10.

Introductory

11.

The Equation

12. Variable

20
of Conduction in the

Temperature

20

13.

Steady Temperature

14.

Neumann's Ring Method


Emissivity

Ring

22
25
for Determining the Conductivity
'

and
26

CONTENTS

CHAPTER

III

LINEAR FLOW OP HEAT. INFINITE AND SEffl-INFINITE


SOLID AND BOD
i'KGE

15.

29

Introductory

16-17.

The

Infinite

29

SoUd

18.

The

Semi-Infinite

19.

The

Infinite or Semi-Infinite

20.

Conductivity Experiments upon Bars.

SoUd

Rod

21-22. Conductivity Experiments


23. Semi-Infinite Solid.

Initial

of the

at Surface into

Medium

Solid.

Surface into

27. Terrestrial
28.

The Age

Variable Temperature

46

Temperature a Harmonic Function

Initial

...

Temperature Constant.
at Zero

Temperature Zero.

Medium whose Temperature

Time

Radiation

50

Radiation at
is

a Function

53

Temperature
Earth

of the

41

Surface Tempera-

47

Initial

of the

39.
-

25. Semi-Infinite Solid.

26. Semi-Infinite

Time

Sxirface

Time

Steady Temperature

upon Bars.

33
38

Temperature Zero.

ture a Function of the


24. Semi-Infinite Solid.

53

57

CHAPTER

IV

LINEAR FLOW OF HEAT. SOLID BOUNDED BY TWO PARALLEL


PLANES.
29. Introductory

30-31. Finite

FINITE ROD.

Rod.

Initial

Temperature

61

Ends at

fix).

Zero.

Radiation

Rod. Ends at Fixed Temperatures.


Temperature
-

32. Finite

33. Finite

Rod.

f{x).

Radiation

Rod. Ends at Temperatures


Temperature /(.r). No Radiation

Neumann's Bar Method


Emissivity

for

Initial

<t>i{t)
-

66

Temperature

and

61

Steady
-

...

34. Finite

35.

Radiation.

Ends at Fixed Temperatures.

No

No

(fiiit).

g^

Initial
.

gS

Determining the Conductivity and


"

70

CONTENTS

ix

SECTION

KAGB

Kod. Radiation at Ends into Medium at Zero.


Temperature f{x). No Radiation at Surface

36. Finite

Initial

74

37. Application of this Solution to the Determination of the Con-

ductivity
38.

and Emissivity

79

Equation of Conduction in a Wire heated by a Constant Electric


Current

81

39-40. Determination of Blpctrioal

and Thermal Conductivities

83

CHAPTEK V
two-dimbnsionaIj problems
41.

Introductory

42-43. Infinite Rectangular Solid.


44-45.

The Use

88

Steady Temperature

...
...

Conjugate Eunctions in Problems of Steady


.
Temperature
of

and Sinks

46. Sources
47. Variable

91

102

in Steady Temperature

Temperature

89

103

CHAPTER VI
FLOW OF HEAT

A RECTANGULAR PARALLELEPIPED

IN

105

48. Introductory

49-50. Steady Temperature


51-52. Variable

105

108

Temperature

CHAPTER
FLOW OF HEAT

VII

IN A CIRCULAR CYLINDER
113

53. Introductory
64. Infinite Cylinder.

Steady Temperature

114

55. Infinite Cylinder.

Variable Temperature

114

66.

The

Integrals

["

rJ{ar)J{^r)dr

and

rJ^{ar)dr

117

57. Applications of these Integrals


58. Semi-Infinite Cylinder.

Steady Temperature

59. Semi-Infinite Cylinder.

Variable Temperature

60. Finite Cylinder.

116

JO

Jo

Initial

Temperature f(r,

0, z).

121

122
Surface at Zero

124

CONTENTS
PAGE
-

125

62. General Problems on the Cylinder

127

Radiation

61. Finite Cylinder.

131

Determination of the Conductivity from Cylinders

63.

CHAPTER

VIII

FLOW OF HEAT IN A SPHERE AND CONE


Radiation at the Surface of a Sphere into

65.

Initial

135

64. Introductory

Medium

at Zero.

Temperature /(r).

136

Determination of the Conductivity and Emissivity from Spheres

66.

67. Surface of the


68. Solid

Initial

Temperature /(r,

bounded by a Sphere and Cone.

/{', d>

The Cone.

69.

Sphere kept at Zero.

Sizrface at

4>)-

Initial

Initial

Zero

Temperature f{r,

140

4>)

141

Temperature
144

6,

d,

Surface at Zero

(j)).

147

CHAPTER IX
THE USE OP SOURCES AND SINKS IN CASES OF VARIABLE
TEMPERATURE
70. Instantaneous Point

Source

71. Spherical Surface Source

72. Instantaneous

Line Source

73. Instantaneous

Plane Source

74.

79.

151

152

153

Doublets

75-78.

149

156

The Method

of

Images

158

Sommerfeld's Extension of the Method of Images

166

CHAPTER X
THE USE OF GREEN'S FUNCTIONS IN THE SOLUTION OF
THE EQUATION OF CONDUCTION
80. Introductory
81. Linear

Flow.

Initial

82.

The Same

Semi-Infinite Solid

bounded by

Temperature f(x)

at Zero

Source at x' at t=Q.

Solid.
-

x=0

Radiation into
-

at

X69

4>{t).
.

172

Medium
.

173

CONTENTS
The Same Solid.

83.

ture

xi

Radiation into Medium at

cjb

Initial

().

Tempera-

/(a;)

176

84. Finite Solid.

Source at

=0.

x' at

Boundaries a;=0 and

x^a

at Zero

The Same

85.

I77

Source at x' at =0.

Solid.

Medium

Radiation into

at Zero

180

86.

Two-Dimensional Problems

87.

Three-Dimensional Problems

88. Infinite Cylinder.

Medium
The Wedge

90.

Initial

Surface at Zero

f{r, d).

Temperature

184

Radiation into

/(r,, 5).

at Zero

of

187

any Angle

91. Infinite Cylinder.


Initial

183

Temperature

Initial

89. Infinite Cylinder.

182

189

Surface

Planes 6=0, 5 = 00 at Zero.

r=a and

Temperature f{r, d)

194

Extensions of the Previous Results

92.

93. Sphere.

Initial

Temperature

196

J{r, d,

Surface at Zero

<f>).

198

CHAPTER XI
THE USE OP CONTOUR INTEGRALS IN THE SOLUTION OP
THE EQUATION OP CONDUCTION
94.

Introductory

95. Semi-Infinite

End

Rod.

at v^

Initial

96.

The Same Rod.

End

97.

The Same Rod.

Radiation at

at a cos

Temperature Zero

Initial

oit.

End

201

into

202

Temperature Zero

Medium

at v

203

Initial

Temperature Zero
98.

The Same Rod.


Initial

205

Radiation at

End

into

Medium

Temperature Zero.

Rod of two Different Materials.


Temperature Zero
Ends

at Zero

and

v.

206

End at

99. Semi-Infinite

100-101. Pinite Rod.

at a cos at.

Initial

206
Initial

Temperature

Zero
102.

210

The Same Rod.


Zero

103.

Ends

The Same Rod.

at Zero

....

and

Ends at Zero and a cos

Ct.

cui.

Initial

Initial

Zero
104.

One End

at v^.

Initial

at Zero.

211

Temperature
-

The Same Rod.

Medium

Temperature

212

Radiation at the other into

Temperature Zero

212

CONTENTS

xii

PAGK

SKCTION

Rod

105-106. Finite

and

tig.

107. Sphere.

Initial

108-109. Sphere

Temperature Zero.

two Different

of

The Ends at Zero

Surface at

Materials.

Initial

Some Problems on

'

217

Temperature
-

218
222

the Cylinder

CHAPTER

213

Surface at Uo

Zero.
1 10.

two Different Materials.


Initial Temperature Zero
of

XII

INTEGRAL EQUATIONS AND THE EQUATION OP CONDUCTION


225

111. Introductory

112-114. Integral Equations


115. Fourier's

and Linear Plow

of

Heat

Ring

227

232

116-117. Two-Dimensional Problems


118. Three-Dimensional

Problems

234
237

Examples on the Condttction of Heat

238

Appendix

I.

248

Appendix

II.

Note on Bessel's Function


Bibliography

List op Authors

Quoted

General Index

265
267

volume the author's book Fourier's


and Integrals (2nd Ed.), 1921, will be

[In this
Series

250

referred to as F.8.]

CHAPTER

THE DIFFERENTIAL EQUATION OF THE MATHEMATICAL


THEORY OF THE CONDUCTION OF HEAT
^^
1.

Introductory.

When

different parts of

a body are at different temperatures,

heat flows from the hotter to the colder.

rod

Consider the metal

ABCD,

FIG.

1.

and suppose it is heated at the end A from some external source.


For some time the temperature of the rod gradually rises, the parts
near A being heated first, but no change takes place at CD till BC
has had its temperature raised. Ultimately, if the end A is heated
long enough, it is found that a steady state of temperature is reached,
in which, while the temperature may vary from poiat to point,
it

remains the same at each point as the time changes.


This transference of heat from the hotter portions of a body

It must be distinguished
is called Conduction of Heat.
from Convection, on the one hand, and Eadiation, on the other.
In Convection the transference of heat is due to the motion of the

to the colder

heated body

itself,

as, for

example, when the different parts of a

liquid are at different temperatures, currents are


of which a uniform temperature

hotter

body

reached.

produced by means
In Eadiation the

and the colder body gains


some intervening medium.

loses heat

process occurring in

is

it

by means

of a

2. Conductivity.

The Mathematical Theory of the Conduction of Heat may be


said to be founded upon a hypothesis suggested by the following
experiment

THE DIFFERENTIAL EQUATION OF THE

metal plate

an extent

is

given,

bounded by two

parallel planes of

such

that, so far as points well in the centre of the planes

may

are concerned, these bounding surfaces

The two planes

be supposed

infinite.

are kept at different temperatures, the difference

not being so great as to cause any sensible change in the properties


of the soUd.

For example, the upper surface

temperature of melting

upon

it,

by a supply

ice

and the lower at a

may

be kept at the

pounded

of

fixed temperature

packed

ice

by having

a stream

of warm water continually flowing over it. When these conditions


have endured for a sufficient time the temperature of the different
points of the solid settles down towards its steady value, and at
points well removed from the ends the temperature will remain the
same along planes parallel to the surfaces of the plate.
Consider the part of the solid bounded by an imaginary cylinder
of cross-section

This cylinder

is

S whose axis

is

normal to the surface

of the plate.

supposed so far in the centre of the plate that no

flow of heat takes place across

its

generating lines.

Let the tem-

perature of the lower surface be o C. and of the upper Vi C. {V(,>V]},


and let the thickness of the plate be d centimetres. The results of

experiments upon different metals suggest that when the steady

been reached, the quantity Q of heat


which flows up thrpugh the plate in t seconds over the surface S
state of temperature has

is

equal to

T;r,

o_.
'

d
where

is

a constant, caUed the Thermal Conductivity of the

substance, depending

upon the material

of

which

it is

made.

other words, the flow of heat between these two surfaces

is

In
pro-

portional ta the difference of temperature of the surfaces.

This result must not be regarded as proved by these experiments.


They suggest the law rather than verify it. The more exact verification

is

to be found in the agreement of experiment with calculations

obtained from the mathematical theory based on the assumption


of the truth of this law.

Strictly speaking, the conductivity

is

not constant for the

same substance, but depends upon the temperature.

However,

when the range

K may be

neglected,

and

of temperature

is

Umited, this change in

in the ordinary mathematical theory

it is

that the conductivity does not vary with the temperature.

assumed

A nearer

MATHEMATICAL THEORY OP CONDUCTION OF HEAT

approximation to the actual state may be obtaiaed by making


a liaear function of the temperature v,

K==K^{l+av),

e.g.,

where a

is

small.

It is important to notice the dimensions


of K, as it is frequently
necessary to change the units of length, mass and time
in terms

which

of

it is

stated.

Z=

Smce

K-'^i)

St'

dimensions will depend upon those of QI{voVj}.


The unit of heat is taken as that quantity which will raise unit
mass of water 1 C. The dimensions of Q/{vo~Vj) are then simply
its

[M], since the unit of heat varies jointly as the unit of mass and
the value of the degree.

[K]^^^^

It follows that

On

'mm

the c.G.s. system the unit of heat

which

will raise 1

gramme

of

water

is

the Calory, the quantity

C*

measure heat by the work necessary to prothe dynamical unit in this system would be the erg. The
relation between the calory and this unit is given to a sufficient
If it is desired to

duce

it,

approximation by the equation


1

calory =4'2x 10' ergs,

and the numerical value


be 4'2xl0' times

will

of

K, when heat

its

value

when

is

measured

this

in calories,

dynamical unit

is

employed.!
is the British Thermal Unit (b.t.u.), i.e. the
pound of water at its maximum density (39 P. ) by 1 F.

''Another unit sometimes used


quantitj' required to raise

B.T.U.

=252 '0

oal.

f Experiments show that the amount of heat required to raise 1 gramme of


water 1 are not quite the same at different temperatures, and in an exact definition of the calory the temperature of the water would need to be specified.
It is
usual to take for this specified temperature 15 C, and the calory will then be the

quantity of heat required to raise 1


this 15 calory we have the equation
1

gramme

of

water from 15 C. to 16 C.

For

calory =4-184 x 10^ ergs.

See Kaye and Laby, Tables of Physical and Chemical Constants {4th Ed.),

p. 5.

THE DIFFERENTIAL EQUATION OF THE

In the fundamental experiment from which our definition of the


conductivity

and

of

is

derived, the solid

supposed to be homogeneous

is

such a material that, when a point within

hqat spreads out equally well in

all directions.

it is

Such a

heated, the
solid is said

to *be isotropic, as opposed to crystalline and non-isotropic soUds,

in which certain directions are more favourable for the conduction


of heat

than others.

There are also heterogeneous

solids, in

which

the conditions of conduction vary from point to point as well as in


direction at each point.

Theory
3.

of

In this book we shall examine only the

Conduction in Homogeneous Isotropic

The Flow

of

Solids.

Heat across an Isothermal Surface.

Consider an isotropic soUd with a distribution of temperature at

the time

given by
'v=f{x, y,

We may

z, t).

suppose a surface described in the soUd, such that at

every point upon

it

the temperature at this instant

is

the same,

Such a surface is called the Isothermal Surface for the


temperature F, and it may be looked upon as separating the
parts of the body which are hotter than 7 from the parts which
We may imagine the isothermals drawn for
are cooler than F.
this instant for difierent degrees and fractions of a degree.
These
surfaces may be formed in any way, but no two isothermals can
cut each other, since no part of the body can have two temperatures
The solid is thus pictured as divided up into
at the same time.
thin shells by its isothermals. Heat is flowing from one shell to
say F.

another, this flow of heat being along the normals to the surfaces,
as

no transference

of heat takes place along the surfaces of equal

temperature.

GeneraUsing the result of

2 we- take as our fundamental hypothesis

for the Mathematical Theory of the Conduction of Heat that the rale
at

which heat crosses from

the inside to the outside of

an

isothermal

surface per unit area per unit time is equal to

dn
where v

is the

temperature of the surface,

K the

Thermal Conductivity

of the substance, and =- denotes differentiation along the outward'

drawn normal

to the surface.

MATHEMATICAL THEORY OF CONDUCTION OP HEAT

As a particular case, when the isothermals are planes perpendicular


to the axis of x, the rate of flow of heat per unit area per unit time
's

~^ -^ in the direction of the positive axis of x.

as X increases, this rate wiU be positive.

the rate will be negative,

meamng

If is decreasing

If v increases as

that the

flow' of

x increases,

heat

is

in the

direction of the negative axis of x.

4.

The Flow

o!

Heat across any Surface.

We

have stated in the preceding article that we assume that


4he rate of flow of heat across an isothermal at a point P is

on
per unit area per unit time,

or, in

the language of differentials,

on

dS being an element

of the isothermal surrounding the point P.

We proceed to obtain an

analogous expression for the rate at which

heat flows across any surface, not necessarily isothermal, per unit
area per unit time at any point P.

We shall denote this rate of flow by /. The value of / will


depend upon the position of the point, the direction of the
normal to the surface at that point, and the time. We shall now
show that, if the values of/ are given for three mutually perpendicular
planes meeting at a point, its value for any other plane through
the point may be written down.
Consider the elementary tetrahedron PABC, whose three faces
PBC, PCA,

PAB

are parallel to the coordinate planes, while the

perpendicular to the face


cosines (A,

yu, v),

and

is

ABC from the

of length p.

point

then the
Let the area of ABC be A
PAB are respectively XA, /xA and i^A.
;

P has

the direction-

(Fig. 2.)

areas of

PBC, PCA and

we denote the rates of flow for the elementary areas PBC,


PCA, PAB and ABC by f^, fy, /, and /, the rate at which heat
is gained by the tetrahedron is ultimately given by
If

THE DIFFERENTIAL EQUATION OF THE

However,

if

of gain of heat

and p are the

specific

heat and density, this rat

equal to

is

dv

Proceeding to the limit when jj^O, this expression becomes zero


and/a.,/^,/j and/ become the rates of flow at the point P acros
2

Fig. 2.

planes parallel to the coordinate planes

the

perpendicular to which

we have

is

and a plane through JF


Thu
/x, v).

in the direction (A,

x/.+m/. +./.=/

Now, according

to

our fundamental hypothesis, the rate

flow of heat across an isothermal surface per unit area per uni

time

is

equal to the product of the conductivity and the rate

diminution of the temperature in the direction of the normal t


the surface. Let P be a point upon the isothermal, and the nomas
at

the axis of

through P-

z,

the axes of x and y being in the tangent plaD

Then /a; and/

are both zero, since

along the surface.


Therefore

/="/

dv
az

V^'"

no flow takes

plac

MATHEMATICAL THEORY OF CONDUCTION OF HEAT


3

where

denotes differentiation in the direction


^
dh
dv

dv

dv

dv

dv

(\,

dv

fx,

v),

since

Thus

the rate of flow of heat at a point across any surface


inside to the outside per unit a/rea per unit time is

from

the

-K-,
dn
where

^ denotes

differentiation along the

outward-drawn

nmmal

to

the surface at the point.


5.

The Equation

of Conduction.

Consider the case of a homogeneous isotropic sohd heated in

any way and "then allowed to


point P{x,

the

y, z) will

The temperature v at the


x, y, z and t, and

cool.

be a continuous function of

first differential coefficients of

v will also be continuous.

Consider an element of volume of the sohd at the point P, namely,


the rectangular parallelepiped with this point as centre,

and

being parallel to the coordinate axes,

its

edges

of lengths 2dx, 2dy

and

2dz.

Let
is

ABCD

and A'B'G'D' be the

perpendicular.

Then the

where f^

ABCD(xdx)

is

of

flowing into

will ultimately

is

the rate of flow at

be

is

across the corresponding plane.

flowing out across the face

is

given by

idydz{f,+^^dx).
Thus the

rate of gain of heat

from these two faces

8dx dy
Similarly from the others

we

-Sdxdydz^
But

Uydz{f,-'^dx

Similarly the rate at which heat

A'B'G'D'

which the axis

rate at which heat

the parallelepiped over the face

gi^^^^y

faces to

this element of

volume

dz

is

equal' to

^.

obtain

and
is

-Sdxdydz^.

gaining heat at the rate of

8dx dy dzcp

~r-

THE DIFFERENTIAL EQUATION OP THE

Therefore

we have

But we have seen that

and

K is independent of x,

y and

z.

Therefore our equation becomes

/c=

where

cp

The constant k was called by Kelvin the DifEusivity of the suband by Clerk-Maxwell its Thermometric Conductivity.

stance,

The dimensions
conductivity A'

of the diffusivity k are obtained at once

(of. p. 3).

Since c above

from those of the

the ratio of the quantity of heat

is

required to raise unit mass of the substance 1 C. to the quantity required


to raise unit mass of water 1C., it is of zero dimensions in mass, length
and time. Also the dimensions of the density p are [Jf]/[i'].
Thus we have
rrii

It follows that

the units of length and time are the foot and year

if

instead of the centimetre and second, the value of k for these units will

have to be multiplied by
system.

(30'48)'''/3-1557

x 10^ to reduce

it

to the.c.G.s.

(Cf. p. 58.)

By some

of the early writers

who

did not employ the

o.o.s.

system, the

thermal unit was taken as the amount of heat which would raise unit
volume of water 1 C.

Let

of these units

be required to raise unit volume of the substance

1 C.

Then the equation

where

is

the conductivity in terms of the

It is clear that

form

of conduction takes the

Kjc

in

this

system

new

is

unit.

equal to the diffusivity Kjpc

discussed above.

On

the other hand,

when

the thermal unit

is

the amount of heat required

to raise unit volume of water 1C., the numerical value of the. conductivity

with that obtained when the unit is the amount required to


mass of water 1 C, unless the linear unit is the centimetre.

will not agree


raise unit

MATHEMATICAL THEORY OP CONDUCTION OP HEAT


If

the solid

is

isotropic,

but not homogeneous, the equation for

V becomes

In the case of Steady Temperature, when the temperature does


not vary with the time, the equation becomes that of Potential.
Also if at the point P{x, y, z) there exists a source of heat supplying
in the time dt the quantity

Adt

of heat per unit

volume, the

equation becomes

^'^=a^(^aJ+a^(^a^)+a^(^a-J+^Such a condition

is

realised

when conduction

a wire along which an electric current


is

is

takes place along

flowing, since this current

generating heat in accordance with Joule's Law.


These results

that

when

^,

may also

ij,

^,

tinuous functions of

be obtained by the application

of Green's

Theorem,*

as well as their first differential ooefSoients, are conx,

y and

z,

inside

a closed

surface,

j\%+mr, +nC) dS =///(!+! +1) dx dy &,


(I,

m, n) being the direction-cosines of the outward-drawn normal, and the


and throughout its volume.
Suppose any such surface drawn lying wholly inside the given soM.
The rate at which heat flows out across the element dS of the surface is

integrations being taken over the surface,

m+mf,+nf,)dS.
Therefore the total rate of gain of heat within the surface

is

-jjilL+rnfy+nf,)dS.

But

this rate of gain of heat

may

also be expressed

by

fff{op^^dxdydz,
the integration being taken through the region bounded by this surface.

Jjjcp'^^dx dy dz + fj{lfj, .+ /; +/.) dS =0.

Thus

Therefore by Green's Theorem

this holds whatever closed surface we consider, provided


within the solid and no source of heat exists within it.

and

* Cf.

Lamb, Hydrodynamics,

42.

it lies

wholly

THE DIFFERENTIAL EQUATION OF THE

10
Apply

this result to the

element surrounding the point P(x,

y, z),

and we

obtain the equation


"'

"P-dt^-dx^-dy^-dz
as before.
6.

The Transformation

These equations

may

of Coordinates.

be easily transformed into other systems

orthogonal coordinates, the most useful being the Spherical

of

Polar System, in which the position of the point

by

its

distance r from the origin,

and the CyUndrical System,


the polar coordinates

and the coordinate

r,

in

Q of

which
its

and

latitude 0,

its

its

position

is

is

determined

its

azimuth 0,

determined by

projection on the plane of x, y,

z.

These are special cases of the general system of orthogonal coordinates, in

which the position of a point

given by the iater-

is

section of the three orthogonal surfaces,

^=const.,

We

proceed to show

how

>;=const.,

f=const.

this transformation

may most

easily

be

affected.

Consider the element of volume bounded by the surfaces id^,


and let A'B'C'D' and
be the faces i+d^.

ABCD

ridr,, f eZf,

Let

ds^=X''di^+/ui.''d,,^+v^d^^

be the equation giving the length of the elementary arc joining


(^,
f) and {i+d^, rj+drj, C+d^).
Then the area of the section of the ^ surface through P{^,
cut off by the surfaces ijd>i, ^d^ is given by

the points

ri,

4/xi/ dt]

and the

r/,

d^,

rate at which heat flows across this section per unit time
iflV

dr]

f)

is

d^fi,

across the surface ^.


f{ being the rate of flow of heat at
Therefore the rate at which heat flows into the element across

the face

ABCD is ultimately
4 [/^"/^-^(fxvfi) dij

and the

dr,

d^,

rate at which heat flows out across the face A'B'C'D'

^{f^vh+YMh)di)dr,d^.

is

MATHEMATICAL THEORY OF CONDUCTION OF HEAT


Hence the

total rate of gain of heat

The other

faces give respectively

from these two faces

is

-8 |(XmA) didr, d^.

-8 1 (.A/,) didr, dl

Inserting the values of/^,/, and/f, namely,

f__Kdv

^^~

f__Kdv

f__Kdv
J-'~

di'

^^~

9'?'

'"

H'

and equating the expression we thus obtain to


SXfiv

d^ d>] d^cp ^,

we have

which reduces to
dv

when

is

constant,

and

as usual

we have

written

k=
cp

Spherical Polar Coordinates.

In this system

=r sin
y=r sin

a;

z=r

61

cos ^,

d sin

(p,

cos 6,

ds^=dr^+rHe^+r'sm^9d<l>^.

and

Therefore the equation for v becomes


dv
K[d (
TrT^ldrV

which

may

dv\.

d (

dr) +ii5^ 30

3d\

V'

dH^
d<t,^J

be written

dv_
where fi=cos

-dH.28v,ld(-.

Bv\

0.

Cylindrical Coordinates.

In this system

and

doJ "^sin^ 6

x=r

cos 6,

y=r

sin 6,

dsa=rfrH'''c?0'+'^z'-

d'vl

11

THE DIFFERENTIAL EQUATION OF THE

12

Therefore the equation for v becomes

dv_K
'dt~r

may

which

drV drr dd\r ddr dz\

be written
dv

j^K

7. Initial

Before

"r..2 Ofl2"r

and Boundary Conditions.

we can proceed

which wiU express the

to the mathematical discussion of the

it is

necessary to determine the fofmulae

Initial

and Boundary Conditions which the

problems of Conduction,

These are partly the direct expression of


the results of experiment and partly the mathematical statement

temperature

of hypotheses

We

I.

satisfies.

founded upon these

assume that

in the interior of the solid v

function of x, y, z and
differential coefficient

difEerential coefficients

results.

and that

and of the
with regard to x, y and z.

with regard to

is

a continuous

this holds also of


t

first

the

first

and second

Initial Conditions.

II.

The temperature throughout the body is supposed given arbitrarily


at the instant which
If this arbitrary

of our

we

problem which

to this value.

take as the origin of the time coordinate -i.

function

is

continuous,

shall, as

In other words,

if

we

require to find a solution

converges to zero, also converge


the

initial

temperature

is

given by

our solution of the equation,

must be such that

Lt

()

=f{x,

y, z)

t->-0

at all points of the sohd.


If

the

initial distribution is

discontinuous at points or surfaces,

and

in this

must disappear after ever so short a time,


case our solution must converge to the value given by

the

initial

temperature at

these discontinuities

all

points where this distribution

is

continuous.

Boundary or Surface Conditions.


The Surface of Separation of two Media of
ductivities Ki and K^.
III.

(A)

Different Con-

MATHEMATICAL THEORY OF CONDUCTION OF HEAT


Let ^1 and v^ denote the temperatures in the two media.
it is

assumed that at the surface

13

Then

of separation the temperatures are

the same.

Suppose an element
tion,
off

of area

and that an element

lengths

quantity

of

dS taken upon the surface of separavolume is constructed by measuring

along the normals over this area into both media, the

being an infinitesimal of a lower order than the linear

dimensions of dS.

FIG.

Then the

rate at

which heat

is

3.

gained by this element of volume

from the flow over the surface will ultimately be

the differentiations being taken along the normals from the common surface into each medium, the contribution from the ends
being negligible.

Equating

Ci,

Cg

this to the expression

being the specific heats and

media, and proceeding to the Hmit

and

pi,

p^ the densities of the two

when

vanishes,

we have

'yi=2.

as the conditions at the surface of separation of the two substances.


(B) When radiation takes place at the surface of the soUd iato

a gas at the temperature

v^^,

it

is

assumed, and the assumption

THE DIFFERENTIAL EQUATION OF THE

is

suggested by experiment, that the loss of heat per unit area per

unit time

is

proportional to the difference of the temperatures

of the surface

and the

H{vVq), where

and

solid

its

is

In other words this

gas.

This quantity

surface.

loss of

heat

is

constant associated with the state of the

a,

Exterior Conductivity, and

is

called the Bmissivity or

foimd to vary considerably with

it is

the temperature and the state of the surface, so that in experiments

on conduction

it is

best, as far as possible,

always to reduce the

heat by radiation at the surface to the magnitude of a small

loss of

by treating the surface with a suitable material.


The conditions at the surface follow in the same way as above,
and we have
g^
correction

^+h{v-Vo)=0,

H/K=h, and

where

the differentiation

is

taken along the outward-

drawn normal.
(C)

There are other possible surface conditions.

may be

The boundary

kept at a constant temperature, or at a temperature which

and with the time or the


boundary may be rendered impervious to heat. The analytical
varies with the position of the point

expressions for these cases are obvious.


In the mathematical treatment of the question these surface and
conditions are not regarded as conditions which v
itself
is

They

or at the instant t=0.

must

satisfy

on

ttie

initial

surface

are taken as limiting conditions, and

it

required in the one case that our solution shall converge to the given surface

or initial value, and in the other case that the differential coefficients in the
limit as

8.

we approach the

The Solution

We

shall

surface shall satisfy the corresponding conditions.

Problem

of the

now show

is

Unique.

that the general problem of conduction,

based upon the equation of conduction and these


conditions, admits of only one solution.
If possible, let there

initial

be two independent solutions

equations

~,=kVH
v=f(x,
v=(p{x,

Let

in the solid,

y, z) for
y, z,

V=ViV2.

t)

t=0

in the solid,

at the surface.

and surface

v^, v^ of

the

MATHEMATICAL THEORY OF CONDUCTION OP HEAT


Then

15

satisfies

dV

-^=kV^V in

the solid,

F=0 for t=0 in the solid,


F=0 at the surface.
We shall prove that V must be zero everywhere

in the solid.

Consider the volume integral

the integration being taken through the

solid.

=k{{{vWW dx dy dz.
But by Green's Theorem,

\\v ~^dS={{{vVWdx dy dz

the integrals being taken over the surface and through the volume
of the solid.

Therefore

Since

is

zero

over the surface, the

first

integral vanishes,

and we obtain

-'M()"-f)"+>*Therefore

THE DIFFERENTIAL EQUATION OF THE

16

shows that there can be only one solution

similar discussion

for the

problem with the other Boundary Conditions and for the

case of Steady Temperature. To prove that the equations must


have a solution is another matter. Their physical interpretation
requires that this be true
the mathematical demonstration of
:

such Existence Theorems belongs to Pure Analysis.


9. Simplification of the

When

General Problem of Conduction.

the surface conditions do not vary with the time,

we may

reduce the general problem to depend upon two simpler cases,

one of these being a case of Steady Temperature.

For example, when we have to

satisfy

=k:V^, through the soUd,

3v

'"=fi^> y> 2) initially,

and

=c/)(a;, y, z)

we may put

v=u-\-w,

where m

is

a function of

at the surface,

x, y, z only,

\I^u=Q through the

and
and

u=(j>{x, y,

is

a function of

x, y, z

z)

and

and

at the surface
t,

z)u

solid,

initially,

w=0 at the surface.

and

The

y,

such that

-=k'W^w through the


wf{x,

satisfies

solid,

first

is

a case of Steady Temperature, and the second

is

case of Variable Temperature with zero surface temperature.

The case

medium whose temperature does


be treated in the same way.
the surface temperature varies with the time, or when
of

Radiation into a

not vary with the time

When

may

radiation takes place at the surface into a

medium whose tem-

perature varies with the time, three different methods

employed.

two

The

first is

may

be

due to Duhamel, who showed that these

cases could be reduced to those of constant surface temperature

or radiation into a

medium

method corresponds

at constant temperature.

The second

to the use of Green's Function in the Theory

MATHEMATICAL THEORY OF CONDUCTION OF HEAT


of Potential.
integrals.

At

we

I.

z,

in which the initial temperature


is <f>{x, y, z,

t),

is

When

X), then the solution of the

problem

the surface temperature

Q^F{x,y,z, X, t-\)d\.

is zero from t=ao


to t=0,
from
t=0
X)
to t=t, we may say that the initial
zero and the surface temperature is 0(a;, y, z, X),

the surface temperature

<p{x, y, z,

temperature

is

so that the temperature at the time

v=F{x,

when

Therefore

t=X and

and

is zero,

given by

^=J

(p(x, y, z,

Also when the


t=\-^d\ and 0(a;,

<=X

X) from

X,

given by

t is

when t>0.

t),

y,z, X,

to t=t,

is

zero from

i= oo

X)

z,

< X), when t>\.

is zero from = oo
from t=\-\-d\ to t=t, we have

y, z, X,

Hence when the

to

we have

surface temperature
y,

v=F{x,

<p(x, y, z,

y,

z,

the surface temperature

v=F{x,

tX,

If v=F{x, y, z, \, t) represents the temferature at {x, y, z) at the


t
a solid in which the initial temperature is zero, while its

surface temperature is <p{x, y,

and

Duhamel's method, which

shall refer only to

depends upon the following theorem

time

involves the use of contour

tliird

XL)

Ch.

this stage

The

Ch. X.)

(Cf.

(Cf.

17

to

t~\~dX), when t>X+dX.

surface temperature

zero from

is

t=co

to

X)-from i!=X to t=X-\-dX, and zero from tX-\-dX

we have

to t=t,

v=F{x,

y, z, X,

tX)F{x,

y, z, X,

tX~dX),

or ultimately

v=-^F(x,

y,

z,X, tX)dX.

{t>X)

In this way, by breaking up the interval t0 to t=t into these


small intervals, and then
find the

solution

^(a;, y, z, t) in

of

results thus obtained,

the form

v^j
* Cf. J.
C.C.H.

summing the

ec.

we

the problem for the surface temperature

^^F{x, y,

z,

polytech., Paris, 14,

X,

t-X)dX.

Cah. 22,

p. 20, 1833.

THE DIFFERENTIAL EQUATION OF THE

18

The corresponding theorem


follows

If v=F(x,

II.

the

for

case

radiation

of

is

the time

y,

z,

\,

t)

represents the temperature at {x, y, z) at

in a solid in which the initial temperature is zero, while

radiation takes place at

surface into a

its

medium

at (p{x, y,

z, \),

then the solution of the problem in which the initial temperature

and

zero,

as

the temperature of the

medium

is

^{x, y,

is

z, t),

is

given by

j^F(x,y,z,\,t-X)d\.
=J

When
into

the surface temperature, or the temperature of the

which radiation takes

place, does

but changes only with the time, these results


slightly simpler

III.

time
is

form as follows

medium

not vary from point to pointi

may

be stated in a

If v=F(x, y, z, t) represents the temperature at (x, y, z) at the


in a solid in which the initial temperature is zero, while its surface

kept at temperature unity [or, in the case of radiation, while radiation

takes place into

a medium

at temperature unity],

then the solution

of the problem when the surface is kept at temperature


the case of radiation, while radiation takes place into

temperature

(j>{t)],

is

[or, in

at

given by

=!' 4,{X)j^F{x,

Now

(p(t)

a medium

y, z,

t-\)d\.

the general problem with varying surface temperature

requires the solution of the equations

dv

=kVH through the solid,


=/(;, y,

z) initially,

v=(j){x, y,

z, t)

""^

at the surface.

Put v=u-{-w, where

=/cV% through the


M=0

solid,

initially,

u=(p(x,

y, z, t)

at the surface

MATHEMATICAL THEORY OF CONDUCTION OF HEAT


and

-^=(cV%
w=f{x,

w=0
The equations

for

through the

solid,

y, z) initially,

at the surface.

u we have

in their simplest form.

19

just discussed.

Those for

Hence Duhamel's Theorem

are

simplifies this

problem and reduces it to the case of surface temperature independent


of the time.

CHAPTER

II

FOURIER'S RING
10. Introductory.

From

symmetry

reasons of

ture or the form of the soUd,


for the temperature

may

be simplified,

in the initial distribution of tempera-

it will

often

happen that the equations

which we have obtained in the previous chapter


and that one, and sometimes two, of the co-

ordinates disappear from these equations.

For example,

dealing with a sphere in which the initial temperature

is

if

we

are

a function

only of the distance r from the centre, and the surface conditions

same

are the

over the sphere, the temperature will depend

all

Similarly, if the soUd is bounded by two parallel


and x=a, and if the initial temperature is a function
of X only, and the surfaces are kept at constant temperatures, the
isothermals will remain planes parallel to the bounding planes,
and the teniperature will depend only upon x and t. Further,
in the case of an infinite cylinder whose generating hues are

only upon r and

planes,

t.

x=0

the axis

parallel

to

same

aU points on

at

of

z,

when the

initial

distribution

lines parallel to this axis,

is

the

and the boundary

same nature, the temperature will depend


and wiU be the same at points in the cylinder

conditions are of the

only upon

which
11.

One

lie

x,

y and

on hues

t,

parallel to the axis.

The Equation

of Conduction in Fourier's Ring.

of the simplest

duction of Heat,

and most suggestive problems

in the Con-

when the temperature depends only upon

coordinate and the time,

one

Problem of the Ring. This


problem is also of special interest, as it was the first to which
Fourier applied his mathematical theory, and for which the results
of his mathematical iavestigation were compared with the facts of
is

Fourier's

experiment.*
* Fourier,

TMorie analytique de
20

la chuleur, Ch. II.

and IV.

FOURIER'S RING

For simplicity we shall suppose the ring to be formed by tt


its normal cross-section, a circle of small radiu!
about an axis perpendicular to the plane of the ring, though th
revolution of

investigation will also apply to

any curved bar of small cross


which forms a closed curve with no loops.
The cross-section is supposed so small that the temperatui
may be regarded as the same at all points of the section. Th
section, the axis of

and the problem


temperature at any point in the ring when
initial distribution is given,

by

to cool

radiation and conduction, or

the surface

We

is

is

it

to determine th

has been allowe

by conduction

alone,

whe

impervious to heat.

choose the length x from a fixed point on the

circle passin

through the centres of the normal sections as the coordinate definin


circle.
We examine the movemen
volume contained between the sectior
ah and a '6' at distances x and x-\-dx from the origin, the area of th
cross-section being o) and the perimeter p.
The rate at which heat flows into this element over the face a

the position of a point on this


of heat in

is

an element

of

equal to

dv

rate at which

and the

Hence ultimately the


given by

it

flows out over a '6'

is

rate of gain of heat due to the

two ends

92^

K^^codx.,

The

rate at which heat

of the element

where

is

is

is

being lost by radiation at the surfac

H{vVf,)pdx,.

the emissivity

and the

total rate of gain of heat

therefore ultimately

(K^o,-j)H{v-Voy)dx.
But,

if

c is the specific heat

this rate of gain of heat is

and p the density

also ultimately equal to

dv

cp^.w dx.
,,

Therefore

dv

dH

a",
a7=
cp dx^

dt

Hp.

^(^-'"o)-

cpw

of the substance

FOURIER'S RING

22

^=X

Writine;

~=k,

and

cp

cpuj

we have

^^-\{v-Va).

g-=/c

Wlieii the surface

and

form by substituting

be reduced to

this

is

g^^

g^

Also the case in which there

zero,

rendered impervious to heat

is

the equation becomes

radiation

is

^^^

may

-\-ue->'K

the external temperature o being constant.


is not constant, a similar discussioD
It win be noticed that when

leads to the equation

dv

12.

d(dv\

Cp ax\

at

0x1

The Variable Temperature

Hp,

cpoo

of the Ring.

Consider the distribution of temperature in such a homogeneous


isotropic ring of unit radius,

when there is no radiation at the surface,

and the

is

initial

temperature

an arbitrary continuous

satisfying Dirichlet's Conditions

In this problem

we

shall

(cf.

fimction/(a;),

F.S., 93),* while /(-7r)=/(7r).

suppose this arbitrary function continuous.

In the other cases of Linear

Flow

of

Heat the

difficulties

introduced

by discontinuities in the initial temperature will be examined, f


The equations for the temperature are the following
:

dv

,^

d^v

(2)

(3)

(?)
\dx/x= =(?)
\dxJx
the third condition simply expressing the fact that the temperature

and the flow

of heat

must be continuous at the point given by

a;=7r, or the opposite end of the diameter through the

origin.

Let the Fourier's Series for /(a;) be

o+(%
* In this

''OS

a;+6i sin x)-\-{a^ cos Ix-^-h^ sin

2a3)

...

volume the author's hook Fourier's Series and Integrals (2nd

1921, will he referred to as F.8.


t Cf. e.g. 17. 30, 31.

Ed.),

FOURIER'S RING

23

so that

1 C"

0'n=1

and

hn= -

fix')

COS nx' dx',

/(a;')

sin nx' dx'.

Consider the function v defined

ao+(ai cos x-\-\

by the

infinite series

sin a;)e-'"+(a2 cos 2a;+62 sin 2a;)e-''2'+

...

00

2(*flCos?ia;+6sinwa;)e-'"'-^'rt=0

It

is

clear that each

equation

term

of this series satisfies the difierential

and the conditions (3), and that if we were deaUng


with the sum of a finite number of terms, the sum would also satisfy
these conditions. In the case of an infinite series we have seen *
that we must proceed with more caution.
Since f{x) is bounded,! there is a positive number
such that
|/(a;)|<il4' in (-x, k).
It follows that |aoi<'^= k|<2M and
(1)

|6|

<2M for all values of n.

Therefore

(a cos

nx +6

where

But the
is

sin

series

wa;)e-'"'^'

<4Me-'"'^',

= o<0.

<

^e"""^'"

convergent, and

its

terms are independent both of x and

therefore the series


00

2(n cos nx-\-bn sLQ nx)e-''^^'^

=o
is

uniformly convergent for any interval of

regarded as a function of
tg

t,

it is

x,

when t>0, and

uniformly convergent when t^tf)>0,

being any positive number.

The function v defined by this series is thus a continuous function


X and a contiauous function of i in these intervals. (Of. F.S., 68.)
It is easy to show that the series we obtain by term by term
differentiation of v with respect to x and t are also uniformly convergent in these intervals of x and t respectively. Thus these series
of

represent the diSerential coefficients of the function


71.)
* Cf. F.S., Ch. V.

t Of- P-'S'-. 24, 31.

v.

(Of.

F.8.,

FOURIER'S RING

24

-^

= V)cw^(a cos nx-\-b sin a;)e~''"^',

^-"2

= 2'f'*^(*n COS nx+b sin wx)e-""^',

Hence

/(

''^

n=0

dv_

and
Thus

satisfies

i;

Now

d^v

the differential equation

(1).

the series which defines v has been shown to be uniformly

convergent for any interval of x, when t>0, and thus to be continuous

m such an interval,

and the same holds

But the values we obtain


the same.

and

Therefore.

^t

similarly

It

when we

for v

^^ (^)_

jj.

^
dv

of the series for

substitute

xir

are

(^^^

(|)= i*^(S)-

would have been more correct to state the conditions

our problem in this form, since

we

the value of these functions for

a;=7r

are not so

much

(3) of

concerned with

as with their limits

when

tends to 7r.

We
initial

have now to examine whether the function v satisfies the


conditions (2). For this purpose we must use the extension

of Abel's

Theorem

(cf.

that the series for v

F.S., 73,

is

since

we have only proved


when i5i<o>0, and

uniformly convergent

without further examination

when t=0

I.),

we could not

use the fact that v=f{x)

as equivalent to the initial condition (2), which

t^a*

is

reaUy

Lt {v)=f{x).
t-s-O

In the extension of Abel's Theorem above referred


that
is

when

a convergent series whose

sum

is

to,

we saw

A, then the

series

0(O=aoe-'+aie-''i*+...,

where

= ao<ai<... and 0<t,

is

also a convergent series,

Lt <p(t)=ao+ai+ ...=A.
(->0

Let us apply this theorem to the

series

v=ao-\-(aiCos x-{-bj^sia.x)e~''^+ ...

and

FOURIER'S RING
Since
fies

we have assumed

Dirichlet's

25

that the continuous function f{x) satis-tt^k^tt, and that

Conditions in the interval

fM=f{T!-), we know that the


cos

<'So+('fi

converges to the value f{x)

series

x+bi

sin

a;)

..;

m the whole iaterval

-tt

= ^ x.
a;

(Of.

F.S., 95.)

Thus our

series

converges

when t=0, and our theorem

tells

us that

U{v)=f(x).
Therefore

=2(^ ^'^^ nx+b sia wa;)e-''"^'


11=0

satisfies all

the conditions of the problem.

13. The Steady Temperature of the Ring.


Let the section of the ring at a;=7r be maintained at a constant
temperature V until the flow of heat has become stationary along

the ring, radiation taking place into a


perature, which

we take

The equations

and

medium

at a constant tem-

as the zero of our scale.

for v are

(1)

^ju^v=0, where iJ.^-^,'{ir<x<ir)

(2)

v=F

at a;=7r,

(3)

j-=0

at

x=0,

the last equation being required by the symmetry of the distribution


of temperature.

The general

solution of (1)

is

v=A cosh(/;ia;-|-a),
A

and a being arbitrary constants, and


by the solution

it is

clear that all the con-

ditions are satisfied

cosh ax

v=VTT cosnxr^'
/uTT

which thus expresses the

method

final state of

temperature in the

of determining the conductivity

result.*
* Of. 32.

is

ring.

founded upon

this

FOURIER'S RING

26

Let the temperatures be

and

and

in the ring,

ajj

any three points

v^ at

^rH,=^^+^l^_^^_3=2

Then

cosh

and

and

Vi, v^

cCi,

x-^

This result

is

let

for three such points this ratio

confirmed by experiment, and was


Putting {Vi-j-V3)/v2='2n, say,

cosh^uZ,

fix^
is

first

constant.

pointed out by

we have

Fourier.*

e'^=n-i-'\/(n^l),

lJ^=log(n+^/{n^-l)).

and thus

we have two rings of equal' perimeter, cross-section and


and temperature observations are taken at three points,
as described above, we would obtain the ratio of their conductivities
in the form
_ logK-f VK^-1
If,

then,

emissivity,

))Y

2%

and 2w2 being the values of (Vi+Vsj/v^ in the two substances.


The disadvantage of this method is the uncertain character

of

the emissivity.
14.

Neumann's King Method


and Emissivity.

of obtaining the Values of the Con-

ductivity

Suppose the

ring, as in 13,

heat has become steady.


the ring

is

heated at a;=7r until the flow

The source

of heat is

allowed to cool, radiation taking place into a

constant temperature, which

we

shall take as zero.

time from the instant at which the source of heat


equations for v are as follows
(1)

cosh yUX
.

/'dv\

Fourier,

loc. cit,

is

_(dv\

107-110.

\,

=o.-.a.s
{t>0)

and

u=a/-.
\/c

at

removed, the

-Tr<x<ir)

/ULTT

X=
^, k=
cpw
cp

where

(*>^'

-"S^-

<^)

medium

Measuring the

Jt"^" dx^"^'"'

of

then removed, and

FOUBIER'S RING

27

Putting v=e-''% these equations give


(4)

(5)
(6)

(>0)

But, using the Cosine Series for

5?^^

namely,

cosh /xx
cosh

/xa;

_ 2| tanh

cosh ^TT

/^tt

"

our solution for v f oUows at once, and

^tanh

cos MTT

TT

7re-'^'

TT

is

given by

^COSWtt

os+Zj

Vm

L2/x2

+*g

coswxe-

icn'^f

After a considerable time has passed, the convergency of this

becomes very rapid owing to the presence of the factor e- ''''""''.


Neglecting the terms after the first two, we have the equations,

series

Vq+i>^=

tanh

mtt

e"'^',

fjLTT

iVfX

-Vo+V^=f

-tanhyU7re-<''+^^',

(M^+l)7r"

connecting the temperatures at

x=0

and

a;=7r.

This method requires the observations of the temperature


the ring

is

cooling at the points x='0

and x=ir.

when

These observations

should be taken at equidistant intervals after a sufl&cient time has

passed to allow our approximations to hold.


the observed values of

are satisfied,

If these conditions

log(t;,d='yo) will

he on two

straight lines.

Let

Vo-\-v^=ai
Vf,-\-v^=a2

when
when

t=ti,
i=<2.

Then
and

The mean
of X.

of a set of such observations

wiU thus give the value

FOURIER'S RING

2$

In the

CHAPTER

III

THE LINEAR FLOW OF HEAT.


THE INFINITE AND SEMI-INFINITE SOLID AND EOD
15. Introductory.

In this chapter we shall examine the different problems where


the isothermal surfaces are planes parallel to a;=0 and the, flow
of heat
X.

is

linear, the

lines

of flow being parallel to the axis of

It will be seen that the results

we

obtain in this

way

also

serve for the flow of heat along straight rods of small crosssection

when

there

is

no radiation at the

surface.

After obtaining the solution for the Infinite SoUd,


to examine, in detail, the

many

we proceed

important problems of Linear Flow

of Heat in the Semi-Infinite Solid, or the solid which is bounded


by the plane a;=0 and extends to infinity in the direction of x
positive.

Various applications of these results in obtaining the

values of the Conductivity will be noticed.

The corresponding

problems in the case of the Finite Solid bounded by the planes


a;=0 and x=a will be treated in the next chapter.
16.

The

Infinite Solid.

In the theoretical case where the solid


initial

temperature

is

given by the equation


i)=/(aj),

the equation of conduction reduces to

is

unbounded and the

THE LINEAR FLOW OF HEAT

30

du

Since

^57-=

also

an

- fi
4'H

2rf^

a;2

tB

^rr,

i*'

integral.

particular integrals

also

is

an

integral,

sum

v=^

which

is

of

is

convergent.

f{x+2V{Kt)i)e-^'di.

In the limit when t^>-0,f{x-\-2\/(Kt)^)=f{x),

and it
given by

continuous
is

any number

a;'=aj+2V(rf)f>

find that

integral

of

and thus

the equation, assuming that this integral

Putting

we

^
<,

4:k2*

Further, the equation being linear, the

satisfies

*'

a particular integral of the differential equation.

is

Therefore
is

-:^

je
4/ci-

tC

;^-s=
2

9a;'

this expression

X^

92jf

-^
4'-|

2P

and

=6

is

assumed that the

if

this function

is

Uniiting value of this

equal tof(x).

Therefore the temperature in the Infinite Solid at time

t,

due

to the initial temperature

v=f(x),
''

^^^'^ ^^

"=2vfe^)E/(^')

The corresponding

CO
*^*^

^=

results for

^''"^dx'.

two and three dimensions are

00

(2^(TKf)) j

] f^'^'V''^')^

*"'

dx'dy'dz'.

00 CO CO

Since
Jo

e-"'''cos26a;(ia;=^e'^,*
2a

* Cf. J./S., p. 195, Ex. 13,

and Gibson, Treatise on

the Calculus (2nd. Ed.), p. 469.

AND

INFINITE
and therefore

AND EOD

SEMI-INFINITE SOLID

*^<

cos a {x' x) da =ir7r7K

''''

Jo

i>J\Kt)

we may transform

expression for

tlie

poo

i)

into

poo

dx'\
TtJ -00

f{x') cos

a(a;'

x)

-""''

(?a,

Jo

a form which would be suggested by Fourier's Integral for


(<
j";"
namely,
.]^
^1 ^<^\ f{x')cosa(x'x)dx'.
TT Jo
J -m

The above

17.

Solid

is

/(.t),

the form in which Laplace's solution for the Infinite

There are several points in the argument which


if the discussion is to be at all rigorous.*

usually presented.

is

31

obviously require fuUer treatment


I.

We

shall assume, in the first place, that the arbitrary function f{x)

bounded for aU values


in any given interval.

is

Let

of

v{x,

This integral

is

|/(ir)l<Jf, forallvaluesof

a; {e.gr.

t)

=iriT-T\

r fi^'W^^

convergent when <>0, and

a;

Then

t.

be a point at which the arbitrary function


to the positive

number

sponds a positive number

e,

dx'.

is

(Cf. F.8., 86).

continuous.

chosen as small as we please, there corre-

such that

ij

\f(x')-f(x)\<l, when
1

and integrable

can be differentiated under

it

the sign of integration, both with regard to x and

Let

a;}

Denote these integrals by

/j, I^

r"

fx+T)

/"-^-i?

and

-x\^7].

\x'
\

_(x'-xf

I^.

yi7
JTTJ-cc

Therefore

Now

7]

is

/-i)/2v(')
'""" ""'

\Ii\<^f
known

since the integral

it

follows that

we can

""'

du.

choose, t^ so that

e-"'dM<jTjje, when 0<i^i,

e-"

du converges.

* For the bearing of this work on the representation of continuous functions


series of polynomials, reference may be made to Borel, Lemons sur lea fonctiona

by a

de va/riables redles, p. 50, Paris, 1905.

THE LINEAR FLOW OF HEAT

32
Thus

0<t^ti.

^liei

\h\<h'
we can choose

Similarly,

so that

t^

|^sl<ie.

when 0<t^ta.

Further,

fl

i^-x)''

4-

^TT

a:)'

(a:'

Z^(TrKt) Jx - 7)

sJtt J - 7)/2^/((cO

=?IM (

/-J..

e-"'

du -

\Jo

e-"' dji)

-'W(0

'

Therefore

But we can choose

i^

so that

^14^
Vtt

_J-_

Also

^VM -/W|e

Therefore

But
if

is

-fix)

t)

the smallest of
\v(x,

I)

t-y,

t^

^a''<jV/

''''

e-"'(i

<^e, when >0.


when 0^^2.

^2 -/(a;)|<ij

v{x,

Thus,

e''du<y, when 0<<^/2.

J7,/V(<0

=h+(h -m) + hand

ig,

we have

when 0<<^t.

-f{x)\<e,

In other words, we have shown that

Uv{x,t)=fix),
when X

is

any point at which f{x) is continuoiis,and the function has been assumed
all valves qfx and integrable in any given interval.

bounded for

n.

It will be found

by a

similar

argument that

Lt v{x,t)=l{f(x+Q)+f(x-Qi)},

when the

limits /{a;

+0) and/(a;-0)

exist,

and the function

is

subject to

the same conditions as before.


III.

Let /(a;) be continuous in the interval

interval.

Then the number

X such that

a^x^fi.

-q

referred to

(Of. F.S., 31,

With some obvious verbal changes


uniformly to /(a;) in the interval

(o, y8)

(a, (3)

above

and

also at the ends of this

will serve for all values of

Theorem L)

in
as

(I.) it will
i->-0.

be seen that

v{x,

t)

tends

INFINITE

AND SEMI-INFINITE SOLID AND ROD

33

In other words, we have

when 0<i^T,

\v{x,t)-f{x)\<e,

the same t serving for

The theorem

IV.

all

values of x in

established in

(I.) is

(a, /8).

also true,

when

f{x) does not satisfy all the conditions there imposed

For example, itf{x) =x%

B"*

0-7?

it is

:'^e

JV

not bounded for

^'=^1

*'"

all

the arbitrary function

upon

it.

values of

x.

{x+2^{Kt)u)H-^du
u'e-"''du.

JttJp:'-x)''

It follows that

when

<->-0, =7;

*"'

x'^e

dx' has the

hmit

x^.

l^JyKKt) J_

Further,

it

will

be seen that when


fix) =af,+aix

we

have

also

Lt

may

V. It

that

+ ... +ax",

v{x,

t)

=f{x).*

be noted that in the above discussion

f(x')dx' converges.

It

is

not

difficult to

f{x)

is

hounded and integrable in any given

verges, v{x,

function

18.

is

has not been assumed

show, as in (L), that when

interval,

and

\f{x')\dx' con-

t) has the limit f(x) (or- J[/(a;+0)+/(a;-0)]) as i-^0, when


the
continuous at x or has an ordinary discontinuity there.

The

Let the

it

Semi-Infinite Solid.
solid

be bounded by the plane a;=0 and extend to

in the direction of

x positive, the

initial

infinity

temperature being given

by v=f{x), and the plane x=0 being kept at zero temperature.


The solution of this problem may be deduced from that of the
Infinite Solid.

We

suppose the solid continued on the negative side of the plane

35=0, and the initial temperature at x' (a;'>0) to be -~f{x'), the


initial

plane

temperature at

x=0

will

x'

remain at

being

/(aj').

With

this distribution the

zero.

Then we have
1

/f"

U-x'p

':=27M (Jo^(^>"^

fO

(x-x')2

dx'+\j-fi-x'))e-^dx'),

this reduces to

If"/

'

Per a more general discussion


C.C.H.

(x+xT \

i':-^')^

see Goursat, Oours d' Analyse, T. III., 543, 1915.

THE LINEAR FLOW OF HEAT

34
It

is

clear that this value of v satisfies all the conditions of the

problem

whose bounding plane

of the Semi-Infinite Solid

is

kept

at zero temperature.

When
may be
and

x'

the

initial

simplified

temperature

by

= x+2y^(Kt)^

a constant, V, this expression

is

substituting

x'=x+2y^(Kt)^

We thus obtain

in the first part,

in the second.

e'^dP

yTTJo

The definite
we write

integrals of this type

have been tabulated,* and

if

rx

yTT Jo
the solution of the problem of the Semi-Infinite Solid, whose surface
is

kept at zero temperature, the

initial

temperature being V,

is

With the aid of the tables for these functions we can find the
time which must elapse before the temperature at a depth x has
fallen to

Since,

a given fraction
if

and from the

tables for 0{x)

From

say ^of

its original

value.

v=^V, we must have

..

it

follows that

=477 approximately.

calculations based

upon the values of k for silver and


it would take J second for the temat a depth of 1 cm. in silver, and that

bismuth, Weber f states that


peratures to

fall

by one

half

* The first table of these integrals was published by Enoke in a paper on the
Method of Least Squares " in the Berlin Astronomisches Jahrbiich for 1834,
giving the values of 6 (a;) for x=Q to x=2 at intervals of 01 computed to seven
"
decimal places. De Morgan extended this to a; = 3 in his " Essay on Probabilities
has
fifteen
A
new
table,
to
places,
from
of
0001
x=0 to a; =3 at intervals
(1838).

"

been published by Burgess in his paper "

On

the Definite Integrals -r-

with extended Tables of Values," Edinburgh, Trans. R. Soc, 39,


j

Weber-Biemann, Die partidlen

Phyaik, Bd.

II. (2

Aufl

),

37,

Differential-gleicJmngen

Braunschweig, 1912.

e~''dt

p. 257, 1899.

der

mathematischcn

AND

INFINITE

SEMI-INFINITE SOLID

AND ROD

35

it would take 8 minutes


wMle the times required for'
change at a depth of 1 metre would be about 1 hour and 1|

in bismutli
siich a

months respectively
It will

two substances.

in the

be noticed that the expression for the temperature may-

be transformed as in

v= TT

=-

16 into
f{x') [cos a {x'

dx''

Jo

dx'\

-x)- cos a {x' + x)]

e-""'"

da

Jo

ttJo

ax e~'"^'' da,

f{x') sin ax' sin

Jo'

a form suggested by Fourier's Integral

f{x)=
TT

da\ f(x') sin ax' sin ax dx'.

Jo

Jo

Prove that when the boundary


and the initial temperature is zero,
Es. 1.

= l --r-

x=0

rxl2V(Kt)

e i^d^.

n/tJo

Ex.

8.

kept at temperature unity

is

Prove that when the boundary a;=0

is

impervious to heat, the

solution takes the form

=
1"

dx'

Jo

The curves

for different values of

and approach the

line

may

Aa

eS^'dP

As

increases these curves get flatter

gets smaller

and smaller they approach

^j^en a;>0,

of the curve for

x=Oj

=0

and the

is

line

the origin
I.

x>0j

case in which the semi-inflnite solid

perature and the surface a;=0

given by

and

^^y^

and the limiting form

we take the

= -i-

be drawn.

v=0.

theline

If

f(x')oosax' coH axe-'^'^'da.

Jo

is

is initially at zero temkept at unit temperature, the solution is

r^lWM

,.

m this case

We

are indebted to Professor A. Stanley Mackenzie for permission to

reproduce the curves, Figs.

4, 6, 9,

and

10, given in his

Equations pertaining to the Propagation of Heat in an


Philadelphia, Pa., Proc. Amer. Phil. Soc, 41, 1902).

On Some
Medium "

paper "
Infinite

36

THE LINEAR FLOW OF HEAT

INFINITE

AND SEMI-INFINITE SOLID AND ROD

37

THE LINEAR FLOW OF HEAT

38
19.

The

Bod.

Infinite or Semi-Infinite

The problems
cross-section

may

Fourier's Ring, the cross-section, of the rod

the same as that at

of

all

is

points of the section

of small

in the case of

supposed so small

may

be considered

its centre.

Suppose the rod to lie along the axis of x, and consider the element
volume bounded by the sections at P{x) and P' {x-\-dx).

The

As

be solved in the same way.

that the temperature at

Eod

Infinite or Semi-Infinite

for the

rate at which heat flows into this element over the face at

is

dv

,,

where

oo is

the area of the cross-section of the rod.

Similarly, the rate at

Hence ultimately the


two faces is

which heat flows across the face at P'

is

rate of gain of heat in the element from these


32^

ooK ^-z^x.

The

rate at which heat

is lost

by

radiation at the surface

is

H{vVg)pdx,
where p is the perimeter of the cross-section and
of the medium.

Vg is the temperature

Also the total rate of gain of heat in the element


dv

is

ultimately

(ccp -- ax.

Ihus we have

which becomes

on putting

dv

dt

cp ox^

-^,=

3/

Hp ('"

dH

-59

cpu)

~'^ 3~2

-=X

~^

('"

and

the surface of the rod

~''^o)

=:.
Cp

Cpu)

When

^n);

is

rendered impervious to heat, so

that no radiation takes place, the equation for the temperature


takes the form

g^

32^

dt^^dx^'

and the problems on the

distribution of temperature in

an

Infinite

AND

INFINITE
or Semi-Infinite

Rod

AND ROD

SEMI-INFINITE SOLID

are reduced to those of Linear

Flow

39
in

an

Infinite or Semi-Infinite Solid.

When

radiation takes place into a

ture, this

may be

medium

taken as the zero of our

becomes

a,t

constant tempera-

scale,

and the equation

ga^

g^

^=k^,
at
dx^

which reduces to

on substituting v=ue-'^K

Thus the problem

is

reduced to that of Linear Plow already

examined.
If the material of the

rod

not homogeneous,

is

it is

clear that

the temperature equation becomes

dt

cp dx V

dx)

30. Conductivity Experiments

"''

cpoo

upon Bars.

The fundamental experiment described


our definition of conductivity

is

derived,

Steady Temperature.
in

2,

from

determination of the conductivities of different substances.

mathematical theory

Rod

of the

which

has been used in the

The

Conduction of Heat in a Semi-Infinite

has also been employed in finding the conductivity and

emissivity.

We

shall

refer

in

this

article

to

experiments in

which the Steady Temperature is used.


A straight bar of small cross-section of the material to be tested
is taken and heated at one end till the temperature becomes steady.
If the bar is long enough, the temperature of the further end is
practically unaffected by the source of heat and remains the same
The
as that of the surrounding medium, which is taken as zero.
circumstances of the experiment are thus represented by the
equations
/c-=-^ Xu=0,

and

= Fata;=0,

v=Oata;=co,

v^yg-vWK)x

gives the temperature at the distance x from the end which

kept at the temperature

is

long enough for the steady state of

temperature to be reached. It will be noticed that this requires


that the rod should be of such a length I that VCVf ) ^ is very large.

THE LINEAR FLOW OF HEAT

40

Bars of different metals of the same dimensions are used and


the surfaces are varnished in the same way, so that the value of
the emissivity will be the same for each.

In this case temperature

observations in the rods give the values of X/k, and thus the ratios
of the conductivities are obtained.

Despretz,

Wiedemann and Franz,

The experiments of Ingenhausen,


upon this method, and

are based

work may be found in Text-books of Physics.*


However the power to radiate heat is one which it is hard to
regulate, and for this reason these experiments are not of such value
as others which we shall discuss later, in which the conductivity
descriptions of their

found directly and without reference to the value of the emissivity.

is

It

is

to be noticed also that these experiments only give the relative

values of the conductivities.

The classical experiments of Forbes f afford a method of obtaining


the absolute value of the conductivity of metals. Forbes also
employed the Bar Method and used the SteadyState

of Temperature.

His method consists of two essentially distinct sets of observations.


In the

first,

the steady state of temperature of a long bar of wrought

long and IJ sq. in. in cross-section) was considered.


The bar was heated at one end till the temperature had become
steady, and it was of sufficient length to allow the end furthest
iron (8

ft.

from the source of heat to keep the temperature

medium.

The

the heated end

of the surrounding

rate of flow of heat across the section distant x from


is

given by
dv

OX
uo

being the area of the cross-section.

This must be the same as the rate at which heat is being lost
by radiation at the surface of the bar from this section to the end.

Forbes determined the value of

mometers placed at

dv

from the readings

different points along the bar,

and

of ther-

his

work

is

thus independent of the mathematical solution which the other


experimenters employed.

* Cf. Poynting and Thomson, Text-book of Physics


Heat (6th Ed.), p. 96 et seq.
Preston, Theory of Heat (3rd Ed.), 296-299; Winkelmann, Handbuch der Physik
(2. Aufl.),

Bd.

III., p.

450

et seq.

Edinburgh, Trans. S. Soc., 28,

p. 133, 1864.

INFINITE

AND

SEMI-INFINITE SOLID

AND ROD

41

The second

set of observations was designed to give the rate


was being lost at the surface.
For this purpose he employed another bar of the same material,
and exactly similar to the first, except that its length was only
20 inches. This bar he heated uniformly and placed to cool alongside the other, which had now cooled, so that the circumstances
of the radiation might be the same.
The rate at which heat was

at which heat

radiated

by

since for

any element dx

this bar at different temperatures could


it

be obtained,

would be equal to
dv

wcp ax -^
But the value
is

of -=- corresponding to

any

definite

temperature

given by continued observations of the temperature at any point.

Thus the quantity

of

heat lost per second by every part of the

larger bar in the first part of the experiment could be determined.

Therefore

is

given

by the equation

dv

dv J
Bv

In this work Forbes employed graphical methods, 3- on the


left-hand side of this equation being obtained from the curve of

the temperature given

by the observations on the long

bar,

and

the integral on the right-hand side being obtained as the area of

a curve plotted from the temperature observations on the second


bar.

By these means he found the

value of the conductivity at different

bar at different temperatures, and he showed that


the conductivity of iron decreases with rise of temperature. His
observations have been repeated by different physicists, and they
occupy an important position among the methods for determining
sections of the

the value of the conductivity of metals.*


21. Conductivity Experiments

Variable Temperature.

upon Bars

{continued).

Angstrom's Method.t

In the preceding article we have shown

how the Steady Tempera-

ture of a long metal rod of small cross-section


* Cf.

Poynting and Thomson,

t Ann.

loc. cit., p.

Physih, Leipzig, 114, p. 513, 1861

may

be employed

98; Winkelmann, he. dt,


;

123, p. 628, 1864.

p. 454.

THE LINEAR FLOW OF HEAT

42

obtaining the conductivity of the substance.

in

The

variable

temperature has also been used, in the case in which one end of the
bar is subjected to periodic variations of temperature, which cause
heat waves to travel

down the

The conductivity is

bar.

calculated

from the march of these waves. Angstrom was the first to employ
this method, and his work is of exceptional interest both from the
neatness of the mathematical discussion and the novelty of his
experimental method.

Hagstrom *

same

later discussed the

prp-

blem, assuming that the conductivity and emissivity vary with the

temperature

Neumann and Weber f extended

the method to the

case of a short bar, both ends of which undergo periodic changes


of temperature.
.

Angstrom employed long bars

x=0 was

subjected to

alternately heated
of cold

of small cross-section.

by a current

of

steam and cooled by a current


this has gone on for some

When

water for equal intervals.

down

time, the temperature in the bar will ultimately settle


periodic state, independent of the initial distribution.
periodic state which
results depend.

Angstrom

The bar

is

The end

temperature, being

periodic changes of

investigates

It is this

and upon which

allowed to radiate into a

medium

a constant temperature, taken as the zero of the experiment.


before,

it is

to a

his

at

As

supposed of such small cross-section that the temperature

over the section

may

be taken as that at the centre, and

of such

length that the temperature at the further end remains unafEected

by the

alterations at x=0, so that in the mathematical treatment


supposed unlimited in this direction.
The equation for the temperature we have already obtained in
the form
-,
,
it is

The

solution will be periodic with the

of the

up
^

temperature at x=0, and

terms

of

it

may

same

F COS n(et-{-Q Bin nwt,


^

/->

0=-=-.

wnere

* Hagstrom, Stockholm, Vet.-Ah. Ofvera., 48, 1891.

t See

35.

period.

as that

thus be supposed built

AND

INFINITE
The

quantities

and

SEMI-INFINITE SOLID

will

AND ROD

43

be functions of x which satisfy

d^ X -^_ nu) p
dx^~K^~~ K ^'
since these results follow

and cos nait to zero

from equating the

Thus we have

(^^_a2yp^^4p=0,

where

a^=-

and

R^=
K

Therefore

coefficients of sinwoii

in the temperature equation.

P=24e-

V cos

(gr/a; e)+^V^ cos (,g'x~e'),

where

and A, A',
Since

e, e'

are arbitrary constants.

vanishes

when a;=oo

equation becomes

it

follows that

A'=0, and our

p=Ae-V cos (gn'x-e),

from which we obtain

Q=Ae-V sin (g^'xe).


P cos nwt-\-Q sin ntoi
Ae~V cos (wcoi g'/a; +e),

Thus the term


becomes

and

v=Aoe-!'^+Aie-Si'' cos (t? 5'i'a;+ei)


-f^a^-"*^ cos (^ojtg^'x+e^),

with the same notation as above.


It is clear that

9oV{^/i^)

and that the mean temperature

while

is

given by

^"=V{^+V&+1?

THE LINEAR FLOW OF HEAT

44

In Angstrom's experiments the heating and cooling effects were

The period

carried out for intervals of 12 minutes each.

of the

temperature in the rod was thus 24 minutes.

oscillations of the

The temperature

at a fixed point was then taken after the lapse

of a considerable

time from the beginning of the experiment, at

way

the temperature curve for

intervals of 1 minute,

and

that point obtained.

This curve should be periodic and of the

in this

same period.

By

analysing this curve

we may obtain the

coefficients in the

expression for the temperature, written in the form,

5o+5iCos(co+/3j)+2Cos(2(<+^2)+....
Similar observations for another point give the coefficients in the

expression

Co +6'i cos
for the

(ftjf

+yi) +C2 cos(2a)f +72) +

temperature there.

Comparing these with the expression for

v,

namely,

v=Age-!'"+Aie-''^^ C0& {ootgj'x-\-ei)

we

. .

= eP^
,r= -i^
^ic-fi^'^

see that

(-"^j

...

- ^i >

Ci

and that
If

^iyi^^Oxi^i^i)-

the distance between the points be

I,

we obtain from

the

formula.

,_W7r
9n9n

the result

log^i-logC'i //3i-yA _ 7^
I

Therefore

By

)~Kr

'T{(3,-y,){\ogB,-\ogC,y

The conductivity
vity.

ji'

is

thus determined independently of the emissi-

changes, the values obtained for k should not vary.

made such changes, and his results confirmed


by his earlier experiments.
When is known, X can be found at once.
IT

X
Angstrom

altering the nature of the surface of the bar so that

the values given

INFINITE

AND SEMI-INFINITE SOLID AND ROD

22. Conductivity Experiments

upon Bars

45

(continued).

Variable Temperature.

In the mathematical problem, where the semi-infinite rod is


initially at zero temperature and the end a;=0 is kept at temperature
unity, the temperature at time t is given by
e-^'di

=1_It

(xjis/lKt)

e-Pdi.

(Cf. 18,

Ex.

1.)

might appear that this solution would afford a means of deterK, since from the observed temperature at any point x-^ at

mining

the time

if,

the table of values of the function

Q(x)=^[\-^'dE
would give the value

and thus

of

T^I2-\/(Ktj),

would be known. The difficulty in using this method


depends upon the fact that the end of the bar, in this case x=0,
is

generally heated

Now

perature.

end

of the

by

a current of water at the given constant tem-

experiment has shown that

it is

not true that the

bar immediately attains the temperature of the

fluid

and thus the mathematical statement of the conditions of the


experiment can only be accepted as an approximation. However
it has been shown to be a fair approximation, and is still accepted
as a means of determining the thermal constants.*
Kirchhoff and Hansemann,f who first discussed this case, made
the assumption that the temperature at x=0 would be given by
C-\-(p{t), where C was a constant and <p(t) a function of the time
which was to be taken indefinitely small. The value of C was
to be determined by temperature observations in the immediate
neighbourhood of the heated end, and was not assumed to be equal
to the temperature of the fluid by means of which the heat was
suppHed.

Another method of treating the same problem has been developed,


and a series of experiments devised and carried out in the Berlin
Physikalische Institut has proved its power. The assumption of
* Cf.

Cf.

Ann. Physih, Leipzig


Ann.Physih, Leipzig

(N.F.), 66, p. 207, 1898.


(N.F.), 9, p. 1, 1880

(N.F.), 13, p. 406, 1881.

THE LINEAR FLOW OF HEAT

46

a sudden change at
is

x=0

to the temperature of the heating hquid

avoided by considering the alteration of the temperature with

the time at two points Xj and

in the bar.

a; 2

may be

solution of the equation of conduction

It

is

shown that

obtained which

give the observed temperatures at these two points.


is

x=0 are

followed.

Two

make

it

mate

solution

This

is

suit the
is

x=0, v=l

actual

distinct hues of treatment are

In one the approximate solution

condition that at

The

only used to obtain a suitable mathe-

matical form for the solution.

is

This solution

then available for the evaluation of the conductivity.

conditions at the end

and then

is

derived from the

this solution is

changed to

In the second the approxi-

observed temperatures.

derived from the condition that at

x=0,

ov
ir-

ox

= ~C.

taken to represent the facts of the case when the end

heated, not

by the flow

will

of water,

x=0

but by radiation from a plate

platinum kept at white heat and supposed to convey to the


end of the bar a constant supply of heat. For the details of these
methods we must refer to the'papers noted below.*
of

23. Semi-Infinite

Temperature

We

Solid.

Temperature Zero.

Initial

have seen in

9 that,

with the time, the solution

when the

surface temperature varies

may be deduced, by Duhamel's

from the case in which this temperature

Now,

Surface at

^(t).

in the Semi-Infinite Solid,

is

Theorem,

constant.

where v has to

satisfy

dv_ dH

v=0 when
!;=1

and>
the solution

is

t=0,

x=0,

at

given by
2

C-'l''

VttJo

V'
Ann. Physik, Leipzig (4. F.), 3, p. 43, 1900; Giebe, Diss., Berlin,
D. physik. Ges., p. 60, 1903
Hobaon and Diesselhorst, " Warmeleitung," Enc. d. math. Wiss., Bd. V., Tl. I., pp. 224-227, 1905.
* Gruneisen,

1903

Verh.

AND

INFINITE

SEMI-INFINITE SOLID

v=0
and
is

47

for t=0,

x=0,

v=<p(t) at

the solution

AND ROD

given by

where

^(a;,

<_x) =_?_["

g-^di.

V''''Ja:/2v'(it(<-A))

In this case

-^) =

I^F^ix,

-y^--''-)|2V(.(^-X))
e

*(-A).

2-v/(7r/c(i-X)8)

Therefore the solution of our problem

is
x2

4ic((- A)

<p{X)-~^d\.

< X=-^

we have

Kf/.

*'=-^

and
In this form

x^2 \
-,

'

e-''^<^u.

clear that our solution satisfies the differential

it is

equation and the

0U 1
/

initial

and boundary

34. Semi-Infinite Solid.

conditions.

Surface Temperature a Harmonic Func-

tion of the Time.


If

the surface temperature in the Semi-Infinite SoHd bounded

by the plane

x=0

is

given

by

v=A

cos{oDte),

an arbitrary function v=f(x), we


problem by putting v=u+w, where

temperature

is

du_

dhi

M=0 initially,

u=A cos {wte)

and
* Cf.

9,

III.

at

x=0

and the

may

initial

solve the

THE LINEAR FLOW OF HEAT

48

dw
w=/(a;)

The value

of

d^w

u we have found

=^r

in 23 to be

Lit -^\-e)e->^'d^,

cos

=^cos(ft)< e)
-1

But

it is

yx

se

i^dix

2^"

sin;j
JxI-UkI)

(1)

'^M

* that

r cos .^
=

*Let

cos

^sin(ft)< e)

known
2

w;=Oata;=0

and

initially,

-J{l>j^^o.U.y

.^

d=

and

f.-'^oos^-jcia;

e-*'sin

=, da;.

and

TTi-'^'^l

2a;2

C"

Thus

sin x^da;

a?

^!=2i>.
da?

Similarly

it

can be shown that

Therefore

-T~i~ ~2m.

-j-n

dor

tt=e~''(4 cos a +

and

Now |m|<|

7}

+ 4M=0,

sin a)

+ e''(^ 'cos a + 5' sin a).

c~''-fZx<i;^7r for every real

a.

Jo

Therefore A' and B' must vanish.

M = e~''(j4 cosa + 5sina)

Thus

i)=e~''(^ sino- Booaa).


a-0, 2i = JV"" and v=().
A^^^l-w and B = 0.

and

But when
Therefore

Thus we have shown that


I

c""^oos=2^^ = J/^yTre'^coBa

-^^sin^2(Zx = J;v'7re

"sin

a.

u and are continuous functions of a, and the


the above argument are justifiable. (Cf. F.8., 84, 86.)

The

integrals

?'

differentiations in

AND

INFINITE
when

Therefore

by those

(1)

t is

of (2),

so great that
is

AND ROD

SEMI-INFINITE SOLID

we can

49

replace the integrals of

given by the equation

u=Ae~ -J i^-^'' COS, (<^t-J{^yFurther,

we know from
'^=d~77
=-TTJO

and, as

when a

sufficient

;\

18 that
*'

fix')[B

ix' \j{x') sin

4t

]dx'

ax sin ax' e-'"^^'-da,

increases, this expression gets smaller

and

smaller.

Thus

time has passed to allow the distribution of tem-

perature to become purely periodic, and the influence of the


distribution has passed away, the temperature

is

initial

given by

l;=4e-^/(ii)^ cos(c-^(g)cc-e).
This result might have been obtained directly, as in the discussion of
Angstrom's method. If we assume that suflSoient time has passed to allow
the temperature throughout the solid to become periodic, it must be given

by terms
where

of the type

P and Q are

functions of x only.

Then, substituting

we

v=P cos (ot+Q sin


'dv

m the equation
,,

p cos at+Q sin at,

,.

37

~"

'S~2'

have, on equating the coefficients of cos

das'

(ut

and

sin

(at,

w/k =j.^

Put

d*P

and we have
which gives

oyt

.d'v

+/x*P =0,

P=A'e- i^lV^ cos (/xa;/V2 + e') + ^'V'^/'/z cos

((ua;/^2

+ "),

where A', A", e', and e'' are arbitrary constants to be determined by the
and boundary conditions.
But when k^oo P must not be infinite.

initial

Therefore

and
Also since

THE LINEAR FLOW OF HEAT

50
The

conditions at

a;

=0 show

that

A'=A,
c'=,

v = Ae''^^^''

and

(wt-^J( ~jx-

cos

as before.

When

the surface temperature

of period 27r/a),

(j>{t),

Series for

is

a periodic function of the time

we can obtain the

solution

by using the Fourier's

<p(t)

(j){t)

=ao+(ai cos wt-\-bi sin wt)-\-(a^ cos toi+ftg ^i^^ wt)-\-...


=Aq-\-Ai cos (a) ei)+^2 cos (2a)< 2) + ...

With

this value of ^(i)

we

+^26 " V(s


25. Semi-Infinite Solid.

at

Zero Temperature.

When

the

initial

"

have, from the above discussion,

cos

(2cot-J(^^ x-e^ +etc.

Radiation at the Surface into a Medium

Temperature Constant.*

Initial

temperature

equations for v are as follows

is

constant and equal to

Vg,

dv

d^v

v=Vg when <=0,


dv

^+hv=0
T

when x=0.
1

Let

(p=v-

Then we have

dv

h dx

-^=kj^,
(p=VQ when t=0,

0=0 when x=0.


Therefore, from 18,

yTTJo

and

it

will
*

be noticed that, when


The

a;->oo

case of initial temperature f(x)

^{x,
is

t)

has the limit

treated in

83.

Vg.

the

INFINITE
To determine

Thus

AND

SEMI-INFINITE SOLID

we have

51

the equation

=Ce''^-Ae''*f
J

on integrating

AND ROD

(/)(^,

t)e-''^d^,

00

this equation in the usual

way.

/OO

v=Ce''^+h\

Therefore

rf)(a;+^, i)e-'"'rf,

Jo

on putting f =a;+>;.

But

as a;-oo

^{x,

t)

C must be

follows that

Hence the

has the limit

solution of our problem

v=h\

o,

and

as v

must be

finite, it

zero.
is

given by

(j>{x+ri,t)e-^-^dri

2vJi

f(*+i)/2v'(ico

e-^'^du

LJo

V'^rJo

Therefore
2v^
s/tt

Jo

Jo

VttJo

VlTTKfjJo

VttJo

ViTrKt)

Jo

In the second integral put


x+,,+2hKt=2'^/{Kt)u.

Then

V"J

V 'T

J (x+2kKt)l2^(.Kt)

VttJo

Vtt

VJo

Olx)

=-^ f e-'^^du*

Jo

'

Therefore

where, as in

18,

V'^Jo

^___
* Cf. Kirohhoff,

Bd. IV., pp. 25-27


Bouaainesq, Theorie analytique de
165-167, Paria, 1903; Weber-Riemaim, loc. cit., Bd. JI., 38.

toe. cit.,

la chaleur, T. 11.,

THE LINEAR FLOW OF HEAT

52

Put x=0, and we have


V
=e'^"'''{l-Q(hV{Kt))}
o

z
_2_

f"

\/-K

We

proceed to find

tlie

surface temperature after a considerable

time has passed since coohng began.


It can easily

be proved, by repeated integration by parts, that

This series does not converge, since the ratio of the w"' term to
the

(n
does not remain less than unity, as n increases. Howif we take n terms of the series
the remainder namely,
l)*"''

ever,

1.3...2w-ire-'=
is

less

than the

w"'

term, since

P
We
up

_ ^ f" ix

_^idx

can thus stop at any term, and take the sum

of the terms

to this term as an approximation for the function, the error being

less in

absolute value than the last term

If in this

way we

V-rryhViKt)

and choose

where

e is

we have

retained.

take

2{hV(d)f

so great that

any

error in taking

positive quantity taken as small as

we

please, the

for the temperature at the surface will be less in value than

e.

INFINITE

AND SEMI-INFINITE SOLID AND ROD

26. Semi-Infinite Solid.

at Temperature f (t).

In

this

Radiation at the Surface into a

Initial

53

Medium

Temperature Zero.

problem the temperature v has to satisfy

dv_

d^v

dv
^-\-hv=hf{t)

at a;=0,

v=0 when
Proceeding as in the

last article,

Then we have the

i=0.

put
dv

following equations to determine

Tl^"

dx^'

c/)=f{t)

^=0

(j>

x=0,

at

when

i!=0.

These equations have already been discussed in 23, and we

have seen that


2

Hence, as in

x^ \

25,

27. Terrestrial Temperature.

Observations of the temperature at points near the surface of the


earth have been carried out at a large

number

stations in difierent parts of the world for

have estabHshed the existence

results

of

of meteorological

many

two

years.

distinct

These

phenomena

of terrestrial temperature.

The

first is

the heat
tures

of

that the variations of the surface temperature from

by day
points

to the cold

at

by night do not

affect the

a depth of more than 3-4

feet,

tempera-

while

the

yearly changes from the cold of winter to the heat of summer


may be observed up to a depth of 60-70 feet. Below that depth
the temperature remains practically constant from day to day

and

is

surface.

not

subject to alterations

due to the changes at the

In other words, the heat waves due to the changes of

THE LINEAR FLOW OF HEAT

54

the temperature at the surface die


to a depth of

more than 60-70

ferred to the earth oscillates in

away

before they penetrate

and the heat which is thus transthe upper crust, and while it proceeds

feet,

inwards at certain seasons of the year, at others

it

ascends and

radiates into space at the surface.

However,

we

after

pass the limit at which the temperature

is

and reach the depths at which


it remains constant from day to day and year to year, there is
a marked increase in the temperature as we descend. The temperatures observed at a great number of points at considerable depth
and at many different stations leave no doubt upon this phenomenon.
affected

surface changes,

been observed near the equator as well as in the temperate


rate of increase varies with different places

It has

and although the

zone,

and

by these

much

is

greater in the neighbourhood of active volcanoes or

thermal springs,

it

may

roughly be taken as about 1 F. for every

50 feet of descent at depths up to about one mile.* This rise of


temperature was ascribed, both by Fourier and Laplace, to the
high

temperature of the earth, this supply of heat being

initial

gradually diffused outwards, and


at the centre of the earth.

still

to a great extent preserved

Such an assumption does not require

that the rate of increase of temperature should be uniform as


we continue to descend, and other physical phenomena show
that

the

interior

of

the

earth cannot be a

mass

of

molten

rock.

periodic changes in the temperature near the surface have

The

been used by writers from Fourier and Poisson onwards

in the

determination of the conductivity of the Earth, these determinations

becoming
in the

owing to the growth


which thermometric observations have

of increasing value in recent years

number

of stations at

been made.
Since these daily and annual variations of surface temperature
are noticeable only at points comparatively near the Earth's surface,

the problem

may

be simplified by neglecting the curvature of the

Earth and supposing the surface to be the plane x=0, which is


subjected to a periodic change of temperature. This problem has

Some

now regard the average as more nearly 1 F. for every 60 feet


even 1F. for every 70 feet.
Cf. Sollas, The Age of the Earth and

authorities

of descent, or

other Geological Studies, 1905.

INFINITE AND SEMI-INFINITE SOLID AND ROD


been discussed in

24,*

and the temperature

55

at the depth x below

the surface was found to be

A^+A^e-^iT>

cos

(c-^() *-ei)

+^2e~^^(^)'' cos (^2toi-

when the

surface temperature
V

It

is

JQcc-eaj-f

...

is

=^0+^1 cos (mi ei) -\-A^ COS (2&)i ea) +

. .

true that the temperature at the surface of the Earth depends

not only upon the time, but upon the position of the place of observation,

and that the constants

tion of these points

but,

A^^,

if

Ai,

...

will

be functions

of

the posi-

a comparatively small portion of the

is considered, the temperature over this may still be supposed


dependent only on the time, and the general principles under which
the periodical surface changes are transmitted into the interior and

surface

there die

away

will

be fully illustrated by the solution in this form.


partial wave is propagated

Thus the theory shows that each

that the amplitudes. of the waves


more rapidly than those of greater
have a more rapid alteration of phase

with unaltered period inwards


of

shorter period diminish

period
while,

and that they also


on the other hand, the velocity
;

of their propagation is sinaller

in the ratio of the square roots of the periodic times.

that the periodical variation takes a simpler form as

where the partial waves


negligible, so that after

It follows

we

descend,

become more rapidly


a certain depth the principal wave with
of smaller period

the largest period and greatest amplitude will alone be found

become negligible
have become constant. The depth at
which the amplitude of the yearly variation is e.g. 0-1 will be about
19 times greater than that at which the corresponding amplitude
while at a

still

greater depth this will also- have

and the temperature

will

for the daily variation will occur, since

'V(5)~'''VCr,
gives the ratio of the depths

and T'=365T.

This result,

it will

be seen, agrees with the temperature observations which have


shown that while the daily variation is not noticeable after a depth
*See

also Bousainesq,

Paris (Ser.

2), 39,

1915.

loc. cit,

T.

I.,

pp. 210-228

and papers

in Bui.

sci.

math.,

THE LINEAR FLOW OF HEAT

56

of 3 or 4 feet, the annual variation

60 or 70

may

be traced to a depth

of

feet.

These features of the problem were

whom

Poisson, to

all

this discussion is due.

noticed by Fourier and


The simplest application

of the solution to the determination of the Conductivity of the

at different places on

its

Earth

surface will be found in Kelvin's paper on

The Reduction of Observations of Underground Temperature." *


The substance of the Earth is taken to be a homogeneous mass of
such rock as we have on the surface at the place of observation, and

the values obtained for the conductivity are to a very considerable

extent affected by the nature of the soil or rock in which the thermometers are imbedded. The data are the temperature observations
at places on the same vertical and at different depths, these observations extending over a considerable

number

of years.

memoir, Forbes' Edinburgh Observations for a period

In Kelvin's
of 18 years

were employed.
These observations allow the mean temperature curve for a
year to be drawn, and

harmonic components to be obtained.

its

In this way we find for the depths at x^ and


Vi

and

V2 in

X2,

the temperatures

the form,

v,=A^'+AC cos(^ t-e{)+A^' cos(^ - e,") + ...


But according to the solution

of 24,

v=A,+A,e-^i^)^'ooB

we have

(^ t~^[^) x-e)

+A2e-M>oos{';t-^(^)x-.:)
Therefore

we should have
4

'

"

A:=A^e--l^>\
A"=A,e-^(!^)^

Edinburgh, Trans. R. Soc.

AND SEMI-INFINITE SOLID AND ROD

INFINITE
lo

Thus

The

g^'-log^/ _e/-e'_

results of the calculations of the

57

l/riTA

mean temperature

curves

at different depths give values for Aq, Aq", ... , which vary only
to a very slight extent.
These agree with the theoretical result

mean temperature due to the surface changes should not


vary as we descend.
that the

The first harmonic term, or the annual variation, is the largest,


and observations based upon it will therefore be most trustworthy.
Kelvin found that there was almost complete agreement between
the values of

log

^/- log ^/'


!l!z:^',

and

the two expressions which should each be equal to

^/{^r|K),

the

unit of time being the year.

From

these results the value of

k,

or Kjcp, was obtained for the

material at the place of observation.


Calculations were also made of the second harmonic amplitude and epoch.
In this case the different temperature curves for the different years gave fundamental differences in the coefficients for the semi-annual period. These
discrepancies and others in the case of the higher harmonics are not to be
wondered at, as the actual state of affairs is not the ideal one which has been

postulated with regard to the periodical variation of the temperature and

the material of the Earth.

28.

The Age

of the Earth.

We have seen in 27 that after the limits at which the temperature


is
is

by the

affected

surface changes are passed, a

marked

increase

observed as the depth increases, and that this temperature gradient

has been taken, in ordinary circumstances, as about 1 F. for every


50 feet of descent up to a depth of about one mile. That this
gradient might be used to obtain a rough estimate of the time that

has elapsed since the Earth i^egan to cool from


was remarked by Fourier himseK.*

its

molten

state,

In the problem, as simplified by him for mathematical treatment.


* "Extrait d'un Memoire sur le refroidisaement seoulaire du globe terrestre,''
Also (Euvres de
BvM. des sciences par la Societe philoma&ique de Paris, 1820.

Fourier, T. II.

(cf.

p.

284).

THE LINEAR FLOW OF HEAT

58

the curvature of the Earth

supposed constant.

The

neglected and the conductivity

is

surface

medium

radiation takes place into a

at temperature zero

taken

temperature when cooling began


constant and equal to

v^,.

that for large values of

is

approximately

He

(k)

taken as the plane x=0, and

is

as

the

time

the

t=0

obtained the result given in

is

25

the temperature gradient near the surface

VQl'\/{TrKt).

Kelvin * took the simpler problem of the semi-infinite sohd


bounded by x=0, the boundary being kept at zero temperature, the
initial temperature being constant and equal to Vq.
We have seen
in 18 that the temperature at the depth x at the time t is given by
fJ;/2^/(()'

v=-f-\
3v

^-=

Hence

When

ax

is

small and

V
,," ^.
^/(TrKt)

t is

c'm.
becomes approximately

large, this

S/ilTKt)'

as in Fourier's problem.

With the value of k used in Kelvin's paper (cf. loc. cit., 15)
namely 400 the units of length and time being the foot and year,f
we have
9^
__^

""

dx-divt'
Taking =7000F.
and <=10^ we have

as a suitable temperature for melting rock,

g^

ax

50-6

Thus at x=Q, the rate of increase of temperature is 1 in


and this temperature gradient will hold for about the first 10^
At a depth of 4 x 10^ feet, we obtain

9v_J_

_^

aa;~50-6^
or about ^rs

o^

^ degree per foot


dv

50-6,
feet.

'

while at 8

x 10^

feet

we have

_^

9x^50^^

'

or about -j^Vxr "^ ^ degree per foot.

Since the temperature gradients are inversely proportional to


*
j-

The Secular Cooling of the Earth," Edinburgh,


With c.G.s. units this value of k will be -0118.
'

'

Trans. R. Soc, 28,


Cf. p. 8.

p. 157, 1864.

INFINITE

AND

the square roots of the times,

if

will

be

I, 1,

and 2 per foot correspond

-Jjy

at the

respectively.

AND ROD

SEMI-INFINITE SOLID
the gradient

same depth at 4 x 10* years

is -yV
;

59

at 10^ years,

it

while gradients of

to 160,000, 40,000,

and 10,000 years

"We are thus led to the result that, with the approxi-

mation which our statement

problem

of the

affords, for the last

96,000,000 years the rate of increase of temperature underground

has gradually diminished from about J^ of a degree Fahrenheit


per foot to about J^, and that the time which has been required
for the transition from a melting state to that in which the present
gradient holds will be 10^ years.

The assumption

of a higher initial

temperature, 10,000 F., an extremely high estimate, would increase


the term required to 200,000,000 years.

Even

allowing for effects

of higher temperature in altering the conductivities

and

specific

heats of rocks, Kelvin held that this investigation justified the

statement that the consolidation of the Earth, and the time from

which coohng commenced, could not have taken place less than
20 million of years ago, or we should now have a more rapid increase
of temperature as we descend, nor could it have taken place more
than 400 milUon years ago, or we should not have so much as is
required for the smallest value obtained at present from the temperature observations.

This assumption of temperatures of from 7000 F. to 10,000 F.

he recognised* to be a high estimate for the tempetature of molten


rock, but he adopted it, as he was most anxious not to under-estimate
the Age of the Earth, and his wish was to give the largest possible
limits

rather

than the smallest.

Lat6r experiments upon the

behaviour of rocks under high temperatures led him to believe that


these temperatures are much higher than those required for a
typical basalt of the primitive character,

be

a fairer estimate.

and that 1200

C.

would

This change from 7000 F. to 1200 C. would

and he seems
to have been somewhat of the opinion of Kingf that we have no
warrant in this argument for extending the Earth's age beyond
reduce his estimate of 10^ years to a little less than

10^,

24 million of years.

The

limits of the

attracted

much

Age

* Cf. Nature, 59, p. 438, 1895


I Amer.

J. Sci.,

of the

Earth given by Kelvin in 1864

attention, for the geologists then, as now,


;

also Phil.

Newhaven, Conn.,

Mag., London

45, 1893.

demanded

(Ser. 5), 47, p. 66, 1899.

THE LINEAR PLOW OF HEAT

60
a

much

longer period of time for the coohng from the molten state,

arguments being based on the

their

and

visible processes

Since Kelvin's pronouncement

of stratification.

much

effects

discussion

has taken place between the physicists on the one hand, with the
estimates based on the temperature gradient as only one of their

methods

of attacldng the problem,

and the

geologists

on the other *

And

as Kelvin continued to attach much weight to the estimates


from the observed temperature gradient, the simple mathematical

problem treated above has become

classical.

However the

dis-

covery of radioactivity towards the beginning of the twentieth

century has not only afforded


of the

Age

of the

Earth

it

new methods

of attacking the

may be said to have

problem

definitely closed the

controversy as to the reliability or otherwise of the results obtained


by Kelvin's and other allied methods. [ His paper " On the Secular
Cooling of the Earth " has

But

it is

somewhat

now

only a historical interest.

surprising that physicists attached so

much

importance to conclusions where the assumptions made were so


reaching and significant.
tical solution of the

far-

There can be no doubt that the mathema-

heat problems involved, based on hypotheses just

as credible, even before the discovery of radioactivity,

would have

given results widely different from those which the geologists were

urged to accept as the only answer to the question at

issue. J

* Cf Woodward (i) "The Mathematical Theories of the Earth," American


Association for the Advancement of Science {Toronto), 1889; (ii) "The Century's
Progress in Applied Mathematics," Bull. Amer. Math. Soc, 6, p. 147, 1900.
.

t See Rutherford's works, Radio- Activity (2nd Ed.), 271, 1905, and Radioactive
Substances and their Radiations, 258-260, 1913 ; also a little volume in Harper's

Library of Living Thought, entitled The Age of the Earth by A. Holmes (1913).
At the British Association Meeting in Edinburgh in 1921 a discussion on " The
Age of the Earth " was opened by Lord Bayleigh, who, as R. J. Strutt, did most
important work on the amount of radium in the earth's crust and its internal
heat.
The abstract of his address and the contributions by J. W. Gregory and

Eddington to the discussion will be found in the British Association Reports (Edinburgh), 1921. Bayleigh's conclusion is that " radioactive methods of estimation
indicate a moderate multiple of 1,000 million years as the possible and probable
duration of the earth's crust as suitable for the habitation of living beings, and
that no other considerations from the side of physios or astronomy afford any
definite presumption against this estimate."
His address in full appears in Nature,
No. 2713, October 27th, 1921.
t There is an interesting series of papers by Perry in Nature, 51, 1895, the aim
of which was to show that other possible internal conditions would give enormously
greater ages than physicists had been inclined to alldw.
This was before the
radioactive properties of minerals entered into the discussion.

made important

contributions to the discussion.

Vol. IL, Ch. v., entitled.

these questions

we

Mathematics and
Chapter

shall return in

the

Age of

XL

(Cf.

(Cf.

Heaviside

also

Electromagnetic Theory,

the Earth, 1899.

99, 109.)

To some

of

CHAPTER IV
THE LINEAR FLOW OF HEAT. SOLID BOUNDED BY
TWO PARALLEL PLANES. FINITE ROD
S9. Introductory.

In the

last chapter

we have examined the

Linear Flow of Heat when the solid

and

is

unbounded

different cases of

bounded by the plane, a; =0


x positive, and we have seen

is

in the direction of

that the problems of the Semi-Infinite

Rod

are reduced to the

same fundamental differential equation. In this


chapter we shall examine the corresponding problems when the
range of x is limited to the interval Q = x=l, and we shall also see
integration of the

that the mathematical discussion of these problems


in the evaluation of the Conductivity

may

be used

and Emissivity.

In the
have the
rods of so great a length that, when they are heated at one end only,
the other end remains unaffected by the change of temperature
due to- the application of the source of heat, and the rod may be
treated in the mathematical statement of the problem as unlimited
practical

work

of the laboratory it is always possible to

Yet some

in that direction.

of the

most trustworthy methods

of

obtaining the conductivity are founded upon the mathematical


discussion of the flow of heat in a bar of short length, both ends

being subjected to definite conditions of temperature.

Ends

30. Finite Bod.

ture f(x).

No

at Zero Temperature.

Let the origin be taken at one end


of the rod be

Tempera-

of the rod,

and

let

the length

I.

Then the problem


dv

is

reduced to the solution of the equations


d^v

,.

,_,

,,

m=^B^' (^<^<^)
v

and

Initial

Radiation at the Surface.

= 0, when x =

v=f{x), when

anda;

t=0
61

(1)

Z,

(2)
(3)

THE LINEAR FLOW OF HEAT

62
If

the

initial distribution

were

v=An&va.~Y-x,
it is

v=^sin-5-a;e"''"ia~'

clear that

would

satisfy all the conditions of the problem.

Let us suppose that the

temperature, /(x),

initial

is

a bounded

function satisfying Dirichlet's Conditions {F.S., 93) in the interval


(0,0-

by the

Consider the function v defined

.A

infinite series

=Vasm^-a;e -r^.
.

n-w

la

where

This

e"

I-

series,

is

=y

Its
(4)

*<

f{x') sin -^ x' dx!.

owing to the presence

of the

convergency factor

uniformly convergent for any interval of

and, regarded as a function of

t,

it is

x,

when <>

uniformly convergent when

^^0 > 0) 'o being any positive number. (Cf. 12.)


Thus the function v, defined by the series (4), is a continuous
function of x, and a continuous function of t, in these intervals.*
It is easy to show that the series obtained by term by term
differentiation of (4) with respect to x and t are also uniformly
convergent in these intervals of x and t respectively. Thus they
f

are equal to the differential coefficients of the function

Hence

v.

BOUNDED BY TWO PARALLEL PLANES

SOLID

Since the series


interval
of

a;

63

imiformly convergent with respect to x in the

is

O^x^l, when i>0,

represents a continuous function

it

iQ this interval.

Thus
Lt i;=the value

of the

sum

of the series

when a;=0

of the

sum

of the series

when x = l

= 0,
and

Lt 'y=the value

=0.
Hence the Boundary. Conditions are satisfied.
With regard to the Initial Conditions, we may again use the
extension of Abel's Theorem contained in F.S., 73 I.
We have assumed that f{x) is bounded and satisfies Dirichlet's
Conditions in

(0, I).

Therefore the Sine Series for f(x),


2x3;
+ a2sm-T

ttx
ffljSm-=.

converges, and
f(x)

is

its

sum

\-

...

f(x) at every point between

is

continuous, and |{/(a;+0)+/(a;

0)}

at

all

aad I where

other points.*

(Cf. F.S., 98.)

It follows

that

when

Lt

v=

(->-0

from the extension


defined

is

Lt 2j"'n
I->-0

by

sm--^-

(i),

xe

'

of Abel's

Theorem

referred to

above

we have
1-2'

"

=f{x) at a poiat of continuity


=i-{/(a'+0)+/(a'-0)} at

all

other points.

Thus we have shown that if the initial temperature satisfies


and is continuous from a; = to x=l, while
=
the
function
defined by (4) t satisfies all the conditions
0,
(0)
=f{l)
y
Dirichlet's Conditions,

of the problem.
If

by

the

initial

temperature has discontinuities, the function defined

(4) at these points tends to \{f{x-{-0)-\-f{x0)} as <->0.

*lt f(x) is bounded and satisfies Dirichlet's Conditions,


93 that it can only have ordinary discontinuities.
f

it

follows from F.8.

This can be written as

v=yj
since the series

/( )j

under the integral

sm-^^c

sm-pxe

i-i

\dx,

uniformly convergent.

If

(F.S.,

70.)

THE LINEAR FLOW OF HEAT

64
is

taken small enough, v

/(x

+ O),

from f{x0)

bridge the gap

will

and the temperature curve

^{f(x+0)+f{x~0)}.
It must be remembered that the physical problem,
stated

rod

it

to

will pass close to the point

as

we have

for discontinuity, either at the ends of the rod or in the

itself, is

an ideal one.

In nature there cannot be a discontinuity

in the temperature in the rod initially.

we must assume that

In the physical problem

a sudden change of temperature takes place

at the instant from which our observations are measured, in the

immediate neighbourhood of the point of discontinuity or the ends,


if

they are points of discontinuity.

is

thus smoothed over.

we have obtained

The

The gap

in the temperature

solution of the mathematical problem

these conditions, and

satisfies

it

may

be taken

as

representing the physical problem in this modified aspect.

Some

31.

may

further remarks

made

be

as to the Lt

v,

and the way

in

f-<-0

which the fmiction v


I.

defined

(x, t)

by

We know that the Sine Series for f{x),

previous section,

is

the Initial Conditions.

(4) satisfies

under the conditions stated in the

uniformly convergent in an interval

tinuous in that interval and at

its

ends.

(a, ji), if

f(x)

is

con-

{F.8., 107.)

from F.8., 73, L, that v(x,t) converges uniformly to /(a;) as i->-0


In other words, given the arbitrary positive number e, there
a positive number x such that

It follows

in that interval.
exists

\v(x,t) -f{x)\

the

same t serving

Let x be a point within

(a, /3).

Then there

number

is

a positive

<

\f{x) -/(a:)|

Take the rectangle

in the {x,

t)

< <V,

when

/3).

ij

such that

when

\x -Xo\

rj.

x-r]'^x^x+ri,

a'^T-7j< x^+7]-^(3.

where
(a-, t)

be any point of this rectangle.

Then

v{x,

Therefore

t)

\v{x,t)

-f(x,) ={v(x,t) -f(x)] +{f{x) -/(a;)}.

-f(x,)\^\v{x,t)- f(x)\ +

<
<
Thus

tends to f{x),
this rectangle towards (x, 0).

On

e,

plane given by

0<iT,
Let

<

for all points in (a,

v{x,

t)

the oth^r hand,

if

2e,

\f(x) -f(x,)\

when

when the point

x is a point in

/(x -0) exist, and are different from f{xa),

0< ifrr.
(x,

t)

moves along any path

(i<x<l
ill

at which

that can be said

/(aj^
is

in

+ 0) and

that v(x,t)

SOLID BOUNDED

BY TWO PARALLEL PLANES

tends t@ jy(ir + 0) +f{x -,0)} as


the lijtifc x=Xa towards the point

t^O, when the point

65

moves along

{x,t)

{Xo,t).

II. /n the argument of


30 it is assumed that f{x) is bounded and satisfies
DiricJ iet's Conditions in
(0, I), so that we can replace it by. the Sine Series.

But ' jur physical jjituition tells us that there must be a solution for our problem
confespondiiMg to any conceivable initial distribution of temperature in the
roc 1, and, in particular, for

any

a (continuous fumction need not

distribution which

However

continuous.

is

satisfy Dirichlet's Conditions,

and

in fact it

known that there are continuous functions whose Fourier's Series diverge
at an! infinite number of points of the given interval.*

I'i

Fej6r's T%Boiem{F.8.,

101)

and Bromwich's Theorem

the i^athematioal demonstration that the series

proHlem, when all that is assumed as to the


is bibunded and integrable in the interval
'V'he

a;

that

it

much from that given above.


bounded and integrable and
2

r'

/.,/,.

mr

is clear that a<


for every positive integer
jhumber independent of x and n.
if J

^anSm^xe'"

Also the series

and the various


or

series

if

w,

M being some positive

',

a;

is

i.f

proof does not differ

?5ince f{x) is

to

the solution of our

temperature f{x)

initial

(J.S., 73,11.) furnish

(4) is

obtained

by term by term

differentiation with regard

converge uniformly through the region O'^x'Sl,

t^t>0, when

an arbitrary positive number.


Thus the differential equation and the boundary conditions are

t is

satisfied,

as before.

Further,

by

Fej6r's

Theorem, the

Series for f{x) converges to f{x),

if

series of Arithmetic

the function

is

Means

the Sine

for

continuous at that point.

Then, by Bromwich's Theorem, with the theorem given in F.S.,

73, V.,

Lt v(x,t)=f(x),
->-o

and

v{x,t) approaches the limit /(*) uniformly as i->0,

interval in
If

is

differing

which f{x)

is

when x

lies

in

an

continuous.
/(a;), /(a;, 0) existing but
Theorem and Bromwich's Theorem, show that

-an ordinary point of discontinuity of

from

f{x), Fej^r's

Uvix,t) = i{f{x + 0)+f{x-Q)}.

The remarks

in

I.

as to the

way

in which v{z,

t)

tends to

its

Umit, apply

also to these cases.

*Cf. Pej6r, Ann.


t Of.

aci.

Ec. norm., Paris (S6r. 3), 28, p. 63, 1911.

Moore, Bull. Amer. Math. Soc, 85, p. 269, 1919.

THE LINEAR FLOW OF HEAT

66
32.

Finite

Rod.

Radiation

at

the

Surface.

Ends at Fixed

Temperatures.
Steady Temperature.

Wlien the surface is not impervious to heat and the tempers ,ture
medium is taken as zero, the equation for the temperatuiip is

of the

dv_K
dt

dH>

Hp

cp dx^

cpw

with the notation of 19.

The observation

Steady Temperature in such a bar, when

of the

ends are kept at constant temperatures Vi and V^, is oiie of


the earliest methods of obtaining the relative values of the con-

its

ductivities of different solids.


If

we put fx,^=HplKo}, we have the equations

and our

solution

is

v=Vi,

when a;=0,

v=Vfi,

when x=l,

given by

+5

where

Vi=A

and

72=^6''' +6-"'.

FisinhM(?-a;)+Fasinh/^a;

^.

Ti^^g

sinh fjd

Let the temperatures be


Xg,

v^, 2

where
"^3

nj

"^2

and

"^2

v^ at

the points

x^, x^

and

111
j

Then

-i-31_?=2cosh/xa=2M, say.
"2

Hence

e'^n+'\/{n^l),

a result independent of Vi and Fg.


perimeter, cross-section,

and

y\Kj

Ml

*Cf. 13; also Preston,

For two bars

emissivity,

it

logCwi+VK"-!))'

loc. cit.,

of the

follows that

296-299.

same

SOLID

BOUNDED BY TWO PARALLEL PLANES


Ends

Rod.

33. Finite

No

perature (x).

at

Fixed

Temperatures.

Initial

67

Tem-

Radiation at the Surface.

In this case we have the equations

dH

dv

,.

,,

Tr^w^' (o<^<o.

v = V2,

when x=0,

v=f{x),

when t=0.

and

As

in 9,

we reduce

v-i^,

when x =

l,

this to a case of steady temperature,

a case where the ends are kept at zero temperature.

Put

v^u4-w.

'

where u and

satisfy the following equations

g=0,

(0<x<l)

u=Vi, when 33=0,

u=V2, when x=l,

dw

d^w

',,

,_

Tt^'W^' (<*<^)

and

w=0, when x=0 and x=l,


w=f(x)u, when t=0.
"We find at once that ru^Vi-f{v2Vi)xll,

and

it

follows from 30 that

nir

w=>,asm-j-a!e

where

-^%
i^

[/(')-(^i+(^2"i) j)

=7:J

sm -^x

ax.

Thus
v^;,

V =i'i+K-i)

S.^UaCoswTT
!;,
^

J+^ S

"

^^"^

+j^sin-j-oi;e

'

nv

-^'^t
"^^

f(x)Bm-j-x ax.

and

THE LINEAR FLOW OF HEAT

68

When

radiation takes place at the surface, the equations

5-5
~T=
at
cp ox'

=,

v=V2,

and
are solved

{0<x<l)

-V,
cpfii

-when

.5^

= 0,

when x = l,

v=f{x),-when

= 0,

v=e''''\

by substituting

v=Hpjcpio,

where

and then using the

results of this

34. Finite Rod.

Temperature

f (x).

Ends

No

and the next

section.

at Temperatures ^^(t)

and

(i>^{i).

In

Radiation at the Surface.

In this case we have the equations

('<"<^)

=^B^^'

and

v=(j)-i{t),

when

v=(p2{t),

when x=l,

v=f{x),

when t=0.

a;=0,

Following the general method given at the close of

9,

put

v=u4-w,

u=0, when x=0 and x=l,


uf(x), when t=0,

and

(^<^<^)

ar='^a^'

w=^i(t), when x=0,

w=02(f), when xl,

w=0, when t~0.


The value

of

follows from 30,

u = j2^6

I'

and

is

given by

sin-j-x\ f{x)sinj-x ax.

To obtain w we may use Duhamel's Theorem


solution for the surface temperatures

<Pj{t)

that for the surface temperatures v^ and

Vj.

and

(9),

tp^it) is

where

derived

BOUNDED BY TWO PARALLEL M^ANES

SOLID

69

In this case the temperature at time t, when the temperature


through the rod at t=\ is zero, and the ends are kept at ^i(A) and
02(A) from t=\ to t=t, is given by

+ 0,(X) [I + ^ 2 ^ COS ^Tre-''^^'-'' sin '^


Hence, when the surface temperatures are

(p^if)

!^].

and 02 (i), we

obtain

[0i(X)|ji(a;, -X)+02(A)|^2(, t-\)]d\,

">=

F^{x,

i-X)=l-f-- V-e-"^<'-^>sin^a;,

^^(cc,

t~\)=%+-y]~ cos WTre-''^'^'-"^ sin '^ x.

where

Thus

^^-P

-ie-''-i^sin^a;] e''^^(0i(A)-(-l)"02(X))c?X.

Therefore, finally,

i>=j2je

.^"^t
i"

JiTT

sm-j-a;

f'
1

,,

/(oc)

?l7r

sm-j-a;

aas

+Tfo e''^^'(<^i(X)-(-l)"02(X))iX
This solution

may also be obtained by the method used by Stokes in different

Potential Problems.
If

we assume

that v can be expanded in Fourier's Sine Series,


Z, a>n sin ',

is

a function of

given by

where

7'(a;, t)

is

X,

'

vlnf,

t)

sm -=- a;

JO

aa;

the temperature at x at the time

Then, integrating by parts twice,


/

vh;', i)8iE
) sin -y-

y dx"

JO

=^[^0,

t)-(-irvil,

0]-^J' l^in^f .'d.'.

THE LINEAR PLOW OF HEAT

70

But by our hypothesis,

and thus the

is

_!!LE!a,.

Therefore

we have

Therefore

we have
nV.

')>

expansion of

^r-^,

which

is

equal to

+ ?^(0,(O-(-l)"<^,(O).

to determine o

from the equation

nV, ""',
= Ge-'~p-' + ^e-''-ir' f^e'lF\4,,{X)-{-l)H,{X))dX,

where 0
But,

="(0)

= v{l,t),

coefficient of sin -=-a; in. the

given by

4>i(f)
4>,{t)

is

(,

a constant yet to be determined.

vf{x).

initially,

C = f

Thus we must have

f(x')sia-^x'dx'.
Jo

tr

Hence
,

TiTT

" = - 2sm-5-a:e
t
1

-"-,8-

'

nir
f /(a;')
J.,
sm
a;'
,,

./o

'

'

+ ?^ I sm^a; e -"T^

Neumann's Bar Method

35.

^f-

da;

"

J^ e

T^ N<^i(A)

1)<^,( A))

dA/

of obtaining the Conductivity and

Emissivity.

In

Ms

paper, Tiber das Wdrmeleitungsvermngen von Eisen und

Neusilber,'\

Weber

describes a series of experiments which he con-

ducted on a method suggested by


idea of this

method

is

Neumann

in his lectures.

The

the same as in that of Angstrom, but in this

case both ends of the rod are subjected to periodical changes of

temperature, so that the mathematical solution required

The end

the preceding article.

perature

4=0
'

to

t=2T

Vi,

t=T.
and

is

Then

while

of the rod

AB

is

is

that of

kept at tem-

kept at temperature v^ for the interval


is

kept at

V2

and

this is repeated indefinitely.

at

v^

When

from
this

t=T

surface temperatures has gone on for a sufficient time, the

tribution of temperature in the bar approaches

two

to

series of
dis-

limiting states,

which continually repeat themselves, the one belonging to the even,


* Cf. Mollison, Mess. Math., Cambridge, 10, p. 170, 1881.

f Ann. PJiysik, Leipzig, 146,

p. 257, 1872.

SOLID BOUNDED BY

TWO PARALLEL PLANES

71

and the other to the odd period.

Both

of these are independent of

initial distribution of temperature, and this may be


taken as zero throughout the rod.
In 33, 34 we have seen that the temperature at time i in a
rod of length I whose ends are kept at ^^(i) and (f>2,{t), while the initial

the arbitrary

temperature of the rod

medium

is zero,

at zero temperature,

K=K/cp

where

is

and radiation takes place

into a

given by

and

= Hplcpa},

In Neumann's Problem

and

r being zero or

^i{t)=vi

when 2rT<t<{2r+l)T,

<l>i{t)=v^

when {2r+l)T<t<{2r+2)T,

4>2{t)=V2

when 2rT<t<{2r+l)T,

^j(j)=i,i

when (2r+l)2'<<(2r+2)r,

any

positive integer.

Thus at the time t=2rT+t',

{0<t'

Jr

\Jo

where

i'n='c /a

Therefore

<T)

JzrT
-V.

=^S e"^"*- (i-(-l)"^2) sin^a;


{-l + (Hh(-l)")(}q:|S)+^"^'^^^'j

and

the value of
v=

when

becomes very great, this expression


v at the time i! in the even period becomes

in the Hmit,

for

S^K-(-i^'2)sin'^^[i-(i+(-i)")j:g^].

THE LINEAR FLOW OF HEAT

72

In the odd period we obtain, in the same way, for the approxii
value of

V,

Thus at the time

from the commencement

one of the

of

periods,

'"=-72-

K X Sin
^2)

2K7r,

.^^Sw+l

'

and at the time

^=

(2i

4K7r,
72-

x-^

(i '"2) Zj

2K7r,

+ l)7r
^

/'2n+l

from the commencement

j-cc

one of the odd

of

peri

sin-^- X
*

2}l7r

,^2n+l

'

(2w+l)7r

Fin+l

''

+ l^K-^2)S^'^-r^Tq:i=^^'
where we have dropped the accent from t' and simplified the
These two expressions may be stiU further simplified, since
sinh yua; sinh

fj.

n
P 2^ ^.00

(lx)

47r-^
"

2 sinh ^i

"Inir

V4mV +M

sin

se

72

and
sinh//x+sinh//(; a;) 27r^
2w + l
2sinh^Z
-"P"4>(2w+l)V
ij.^=vIk,

we have

sinh/icc+sinhufi!
r^

o
u
2sinh/.Z

siDh/ja;+Binh

smce

/jl

a;)

^kit^ n

2w7r

sm -^x=U,
^=-T2-2j
?2
I
^^^^^

(lx)

2K7r^2w+l

P2='<^j2

i'2n+i=f

say,
J'

(2i+l)7r

+"

(2w+l)V

and

+'^

P
Putting

(2w-f-l)7r

'''
,

1-"-

"^

SOLID BOUNDED BY

TWO PARALLEL PLANES


we have

Therefore, for the even period,

and

73

for the odd period,

(2)

We

shall

now show how from

obtain the values of

C7=0

Since

the equations

(1)

and

(2)

we can

H and K.
at

x=U
"^

F=s2 coshin 7'

and

fjul

and the other terms vanish, the temperature

v^i,

at the middle

point of the bar, remains constant and equal to

V1+V2

2cosh|^r

V =

Therefore

where

]^g (+V(''-l)>

2a

simple

unknown

way

^
-

of obtaining another relation

quantities

is

is

between the two

to take the difference of the temperatures

at x=l/6 and x~2l/3 at any instant.

which w

(3)

For these points the terms

in

a multiple of 2 or 3 disappear from the series in the

expression for the difference of the temperatures, and this series


is

so rapidly convergent that

we may

neglect the term for

n=5

and

those which follow.

Thus, with this approximation, the difference of the temperatures


at these two points at the time

after the beginning of

one of the

periods takes the form

M-NerPi,
P ='^-72' +

where

and M,

"^^

do not vary with t during the interval.


Let the differences of the temperatures at these points at the
times <i, 2, i+)8, and i2+/3 be d^, d^, d-l, and d^.

Then

d-^d^ =N{e-P^e-'^^),

d^ -d^ =N{e-P<^-e-Pi')e-i'^.

THE LINEAR FLOW OF HEAT

74

d.

d-y

Therefore

e^P

di'd^"

and

(^^+.)^=logK-d,)-logK'-ci,')

These two equations


Rod.

36. Finite

Temperature.

(3)

and

are sufficient to determine

K and H.

Medium

at Zero

(4)

Radiation .at Ends into

Temperature

Initial

(4)

No

f(x).

Radiation

at

the

Surface.

In this case the equations for the temperature are

Jt=^W- ("<^<^)'
fit)

~^^+hv =

^+hv=0
md

at

x=0,

(2)

at

x=l,

(3)

v=f{x),-wh.ent=0

The expression

e""" *{A cos

(4)

aa;+B sin ax)

latisfies (1).

It also satisfies (2)

and

(3),

provided that

-aB+hA=Q,
md

a{B cos alA

From

these

sin

al)+h(B sin al-\-A

cos aZ)=0.

A B

=t
n
a

we obtain

*^i^^=^^

^nd

(5)

Hence the expression

(cosax+-

-feaH

sin aic)e~

and (3), where A is an arbitrary constant and a


any root other than zero of the equation
2ha
,
satisfies (1), (2),

tan OS =5

To form an
is

rb-

idea of the distribution of the real roots of

(5); it

only necessary to note that they correspond to the abscissae

the

common

points of the curves

''=t^
where we have put al=^.

^'^'^

is

^ hi
''=M-:r'

of

SOLID BOUNDED
The second
centre

of these curves

at the origin

is

BY TWO PARALLEL PLANES

a rectangular hyperbola, whose

is

and whose asymptotes are

^=0

and

,=l^:

hyperbola and the cotangent curve are drawn,

If this

from the

figure that the positive roots lie one in

intervals (0,

tt),

27r),

(tt,

75

...

and the negative

it is

clear

each of the

roots are equal in

absolute value to the positive ones. Also there are no repeated roots.

Further,

Since

it is

clear that (5) cannot

have a pure imaginary root

ib.

we would have

tanh?&+pq-p=0,
which

is

impossible as both terms are of the same sign.

Also we shall see later* that

cannot have an imaginary root

it

form aib therefore its roots are all real.


Let us assume that /(a;) can be developed in an

of the

f(x)=A,X,+l,X,+...

where

Z=cosaa;H

infinite series
(6)

sin aa;,

On

a being the

Then the

nth. positive root of (5).

solution of our problem

is

=i;4^e^'

(7)

and the question of the


validity of this solution will be referred to again on p. 182, but
if we assume that such an expansion exists, and that we may
integrate the series term by term, the value of the coefficients may
be obtained in a similar way to that in which the coefficients in

The

possibility of the expansion (6)

Fourier's Series, with similar assimiptions,

may

This depends upon the fact that

X,J[dx=0,
Jo

which we

shall

now

* Cf. p. 78.

prove.

{m=\=n)

be found.

THE LINEAR FLOW OF HEAT

76

Since

and

{<Mr?-an^)'^X^Jx=^\[{x,>" A^i
dx'

1^ dx

dx^

(ice

Jo"

dXr
+hXr=0, when x=0,
dx

But

dX,
-'rhXr=0, when x=l,
dx

and

may

whatever positive integer r

Thus
and,

d^X

when

{a

be.

J- a2) X^ndx=
[

0,

m is not equal to n,
Jo

To obtain the value

of

X^dx, we note that

Jo

Thus

...|>...=-[x.t]>(tr-

But

aXn=an cos anX-\-h sin a,ja;,

and

-j^=^an sin

a,ia;+/i cos

tta;.

Therefore

and

an'\'Xn'dx

+ ^' (^^Jdx={an' + h')l

But we have seen that

.fr'^-i;(t)-=-[--s-

dx

SOLID
Therefore

But

BOUNDED BY TWO PARALLEL PLANES


2arJ'\'Xr,^dx=l(an^+h^)~[xJ^'J

_^+^;s:^=0, when x=0,

and

^j^JtX^^O, when

^^=

and

x=l.

^^^^

^^''

'^=0-

jZ^+a)'=HA..

But

and

dx

Z^=l, both when

Therefore

x=0 and

fz^<^.=..("^+/y+^^

Hence,

if

a;=?.

-^

^Cln

Jo

we assume the possibility of the expansion and


by term, we have

integrate term

aA

Xn^dx=\ f{x)Xdx
Jo

Jo

Thus

We

'

z/j^"=-AZ2, when x=l,

Therefore

we may

77

stated above (p. 75) that the equation


,

2ah

tanaf=-r

rh

cannot have an imaginary root of the form aib.

that

THE LINEAR FLOW OF HEAT

78
If this

were possible, we would have two conjugate roots aib,

and these would give the two expressions

a sm aX,
sin

X==cosax-\

Z' =cos

a'x-\

where

a=a+ib

and

a'=aih.

Now we have

a'x,

seen that for any two unequal roots of

(5),

Xm^^dx=0,
and

this applies also to

X, X',

so that

[^XX'dx=0.
Jo

But dividing

X into its real and imaginary parts, we have


X=R+iS
X'=R-iS,

and

we would have

so that

(R^-\-S^)

dx=0,

Jo

which is impossible.
Thus we see that (5) has only

real roots.

If radiation takes place at x =0


i

and

2,

and x=l into media at temperatures


the problem can be reduced to the above as usual by
v=u-\-w,

putting

where m

is

a function of x only which satisfies the equations

=0, (0<x<l)
T
and

and

-J-

w is

\-h{uVj}=0, when x=0,

+h{uV2)=0, when x=l

a function of x and

\-Tmo^,

which

when

satisfies

a;=0,

Yhw=^, when x=l,

and

w=f{x)u, when t0.

the equations

SOLID

BOUNDED BY TWO PARALLEL PLANES

The problems where one end

of the

rod

is

kept at a constant

temperature, and radiation takes place at the other end, or

one end

is

rendered impervious to heat,

79

may

when

be treated in the

same way.*
37. Application of this Solution to the Determination of the Conductivity

and Emissivity.t

In the case of radiation at the surface of the rod into a medium


at temperature zero, the solution may be deduced at once from

and

given by

that of

36,

A and

X,i having the values of that article,

is

and

being equal to

Hplcpw.

Neumann showed

may

that this result

be used in determining

His method requires the


and ,, when x=0 and x=l.

the values of the thermal constants.

measurement

of the temperatures

v,,

Now

Z=:cosaa;H

sinaa;,

an

and thus X=l, when x0.


Also

we have

when x=l.
Since

seen that

We proceed to
tan a,il

Xn^=l
determine the sign of Z.

2anh
an'-h^'

On ^+h^

THE LINEAR FLOW OF HEAT

80

As ^n

increases with w,

if < is

chosen large enough,

a close approximation by using only the

first

we

shall obtain

term in each

of these

series.

On

this understanding

and

^{v^Vi)-=Aze-^^K

In Neumann's experiment he

heated one end of the bar

first

by a flame, and then allowed the bar to cool by radiation. After


some time he began to take observations of Dq*^; ^^ equal intervals.
These readings showed when the temperatures began to obey the
law given above. By this means the constants /3i and ^^ ^'^e found
and thus two equations are obtained from which the conductivity and emissivity may be determined. However, as the
values of a^ and a^ involve h, this calculation has to proceed by
successive approximations and is somewhat complicated.
A simpler method is obtained by observing also the temperature
;

at the middle point of the bar.

When

a;=JZ,

sm

Z = cos
But

tan anl=

It follows that

- a/A. when n

^anl-\

is

tan ^aj

is

5
t->

equal to /i/a

when n

even.

x=hh

Thus, for

Z=cos JanM IH
\

cos f '

=0, when n
g

Thus

^aj>.

vu=Ai
'

when n
is

g~^8^

7+^3

cos fait

odd,

is

even.

Pit

tanlaZ)

an

7-

i
"cos^agl

and to our approximation

-^=cosW.

is

odd, aijd to

SOLID
From

BOUNDED BY TWO PARALLEL PLANES


we

this result

find aj

then h follows from the equation

tan -101?=

and

ag

from

tan ^ag?

Also

=y

^i=Kai^+v,
K and

give

v,

so that the values of the conductivity

may

81

and the emissivity

then be found.

38. Equation of Conduction in a

Thin Wire heated by an Electric

Current of Constant Strength.^

The equation for the temperature in a thin wire along which


an electric current of constant strength is flowing was given by
Verdet in 1872.t For some time httle use was made of this method
'

of heating the metal, although it has several obvious advantages.

In the

first place,

the electrical measurenients can be

made with

found possible to arrange the experiments


so that the difference of temperature along the wire wiU be small.
The error due to the neglect of the change of the Electrical Conducsuch accuracy that

and

it is

Thermal Conductivity,
Further, it is of importance that the same method
is thus avoided.
of heating the wire should be employed in the cases when the
temperatures to be examined are widely different, and that the
two conductivities electrical and thermal should be obtained by
tivity,

also,

though not so marked,

of the

simultaneous experiments.

By

in this form, the question raised

* Verdet,

TMorie Mdcaniqite de

using the equation of conduction

by Wiedemaim and Franz, t

as to

la 0/ialeur, T. II., p. 197, 1872.

t Alternating currents have also been used in this connection.


papers may be referred to

Several important

Cranz, Zs. Math., Leipzig, 34, p. 92, 1889.


Ebeling, Ann. Physik, Leipzig

(4. F.),

Weinreioh, Za. Math., Leipzig, 63,

27, p. 391, 1908.

p. 1, 1914.

The last-named memoir contains a valuable account of the literature on this


and the variation of the temperature over the section as well as along the

subject,

length of the wire

is

taken into consideration.

J Ann. Physik, Leipzig,


C.C.H.

89, p. 497, 1853.

THE LINEAR FLOW OF HEAT

82

the ratio of the thermal and electrical conductivities, has beer

again examined, and

it

has been shown that the ratio

nearly constant as was at

We

is

not

sc

supposed.

first

Equation of Conduction, and then show


how the Steady Temperature and the Variable Temperature in
such a wire have been used to determine the Electrical and Thermal
shall first find the

Conductivity of the metal.

Let the wire be of length

I,

and

strength of the current, and

K,
and

let

conductivity, specific heat, density,

c,

p,

and

H be its thermal

emissivity.

Let

the electrical conductivity,

o-

be the
i.e.

the

reciprocal of the resistance per unit cross-section per unit length.

Consider the element of the wire contained between the sections


distant x

The

and

x-\-dx

from one end.

rate of gain of heat in this element

over the sections at x and x-\-dx


Koo
ft)

from the flow

of heat

ultimately

is

-T-j

dx,

being the area of the cross-section of the wire.

The

rate at which heat

is

lost at the surface of the element

is

H{vVg)pdx,

being the perimeter of the cross-section, and Vg the temperature

of the surrounding

The

medium.

rate of gain of heat due to the current i

The

total rate of gain of heat

This must be equal to

and therefore the equation

Writing

dx.

therefore

dv

wcp-^dx,
of conduction

dv

_K dH

dt

Cp dx^

is

Up.
"'
Cpw (-y-^o)-

k=, \=-^,
cp

this equation

is

is

Cpw

and

Cpcio^cr'

a=

^,

cpufa-

becomes -^=K^-i~\{vv^-\-a.

SOLID BOUNDED BY
39.

TWO PARALLEL PLANES

The Steady Temperature.

83

Kohlrausch's Method of obtaining

the ratio of the Electrical and Thermal Conductivities.

Kohlrausch* has. shown how the Steady Temperature may be


employed in finding the ratio of the electrical and thermal conductivities.

The ends
peratures.

of the wire are

kept at as nearly as possible equal tem-

The surface

supposed rendered impervious to heat,

and the current

i is

is

supposed to have been flowing long enough

to allow the steady rate of temperature to have been reached.

In this case the equation of conduction becomes

K (Pv
cp dx^
^

'

dx^

cpcoV

o-ft)^

Let u be the potential at the section

Then

But

x.

i^ wcr^rdx

THE LINEAR FLOW OF HEAT

84

When

the temperatures at the ends of the wire are kept the same,

the distribution of temperature in the wire will be symmetrical

about its middle point. Let the points x^ and ajg J)e at equal distances
from the middle point x^ on either side.

Then

^i='y3

% M2=M2~'^3-

and
Therefore

we have from

(3)

^(^2-^l)=2K-w2)^
and we have thus obtained a simple method of determining the
K/a of the Thermal and Electrical Conductivities,

value of the ratio

involving only the reading of the difference of the temperatures

and potentials at two points

40.

The

when the

of the wire,

regulated that the temperature of the wire

is

current

is

so

steady.

The Variable Temperature.


variable temperature of a wire along which a constant

electrical current is

surface, has also

The

constants.

flowing, while radiation takes place at the

been used in determining the thermal and

We have found the equation of


dv

electrical

due to Straneo.*
conduction ( 38) in the form

following investigation

is

8^v

r=fg^2-X(-o)+a.

k=, X=
Cp

where

Cpw

Suppose the temperature

a=

and

of the

5-.

Cpw

medium

or

into which radiation

takes place to be zero, and that

< <

a;
when < =0, (0
'y=0,whenaj=0 and x=l,

V =0,

are the initial

To

and boundary

conditions.

we proceed

integrate the equation of conduction,

to break

up the problem

of Variable

* Straneo,

See also
Weinreieh,

Z)

as usual

into one of Steady Temperature and one

Temperature.

Roma, Rend. Ace. Lincei (Ser.


Sohanfelberger, Ann. Physih,

loc. cit.

5), 7,

Sem.

Leipzig,

ii.,

(4.

1898.

F.), 7, p. 589,

1902

and

SOLID

BOUNDED BY TWO PARALLEL PLANES

Pit

v=u+w,

where u

is

independent of the time and

M=0
and

85

w is

a function of x and

dw

dho

at

x=0

w=0
of

is

which

w==n
The value

x=0

at

at

the equations

and x==ll
satisfies

the equations

x=U

and

t=0

obtained immediately in the form

_,

_sinh /ux+sioh /x(Z re)"]

r^

where

sinhjuZ

and

iu.=J{\Ik)

This function

satisfies

may

J'

b=a/\.

be expanded in the Sine Series

2jAsm-^x,
where An=0, when n is an even

when M
With

is

an odd

u the

solution of the equations for

follows

we have
r

w=~2jAnSm^xe
1

and

integer.

this value of

immediately, and

integer,

nv

'-

.n,
-^

>

"

or

46-

y^

(2tt-l)7r

-[<^^^%x]*

Therefore

_,

r,

_ sinh fxx+ sinh fx{l x)l

siahfjil

In applying this solution we note that the


in the series for

diminish rapidly, and

coefficients of the

when x=lj3

terms

or 21/3 the

86

THE LINEAR FLOW OF HEAT

second term in

Hence

zero.

is

value of V at these points

is

to a close approximation the

given by

Let observations be made at the point x=^l at the times

and

ig,

t^, l,

i^ ij=3 ^2=^.

where

Let the temperatures be

v^, v^,

and

v^ respectively.

IS^=('F+^)^
.-.='^'4^('"<-*"'"-"^-'''"'')

-^

Also take the value of the steady temperature at the middle point
of the wire, viz.,

^=Ki-c-^J

(^)

These three equations determine the values of

from

(1)

we

A,

and

a-.

this value of

we have an equation
from (3) we find b.

giving the value of

=r^>

-W

b=l.

(5)

M=v/(V)
/m

and A, and
and (5).

of K
(4)

fx,

/j.

But

Therefore,

For

find the value of k-j^-{-\.

Inserting this in (2)

and with

/c,

being known,
b being

(6)

we have from

known, the value

of

(1)
tr

and

(6)

the values

follows at once from

In the actual experimental work this process was reversed.

The wire was

first

heated by the current

was observed that


The current was then
the ends and the surrounding
till it

the steady temperature had been attained.


cut

off

and the wire allowed to

cool,

medium being still kept at the given constant temperature.


With this arrangement the steady temperature u is given
before by
_, /, sinh /xx+sinh fi(lx) \

sinh/u?

/'

as

SOLID BOUNDED BY
and the variable temperature v
dv

TWO PARALLEL PLANES


is

determined by the equations

dH

v=0, when

and
Thus

87

x=Q and

x=l,

v=u, when f=0.

+A

7rV(2w-l){(2}i-l)V+ZV}^^^

Then, as before, we find

The values

of

k,

A,

and

o-

follow as above.

^^

CHAPTER V
TWO-DIMENSIONAL PROBLEMS
41. Introductory.

two chapters we have been examining different cases


Flow of Heat. In these the temperature has been dependent only upon the time and upon one geometrical coordinate. Such
In the

last

of Linear

problems

may

be referred to as one-dimensional.

We

proceed

to the discussion of cases in which the flow of heat takes place in


parallel planes.

If these planes are

taken parallel to the xy plane,

the temperature will depend only upon x and


of

steady temperature, or upon

We

with variable temperature.

x, y,

and

t,

y, if
if

it

we

is

a case

are dealing

speak of these problems as two-

dimensional.

The

first

problem in the Conduction

Fourier in his treatise,


Infinite Solid

is

of

Heat discussed

in detail

by

that of the Steady Temperature in the

bounded by the planes a;=|7r, y=0, and extending

The boundaries a;= ^7r are


to infinity in the direction y positive.
kept at zero temperature, and the base y=0 at temperature unity.
His discussion led him to the expansion of unity

in

the interval

Jtt < < ^tt in the series


a;

4
{cos

xi cos 3a;+i cos 5x ...},*

TT

and he then proceeded to consider the question of the development


of an arbitrary function in trigonometrical series, and obtained
the expansion now known as Fourier's Series. He was thus able

The

{cos a;-^ cos

series

3a;

+ icos6a;

-...[

7r

may

be obtained in the ordinary

way

f{x)

as the Cosine Series for f(,x),

= l,

(0<x<iw)
{iTr<x<ir)

f{x)=-l.
88

when

TWO-DIMENSIONAL PROBLEMS

89

to give the distribution of temperature in this solid,

when the

fease

kept at the temperature v=f{x),f(x) beiag an arbitrary function


of X, while the faces x=^Tr are kept as before at zero.

is

48. Infinite Rectangular Solid.

Steady Temperature.
Instead of takiag Fourier's Problem in the form which he adopted,
we shall take the soHd as bounded by the planes x=0 and a!=7r,
which are kept at zero temperature, and the plane y=0, which is

We assume that the function /(a;)


bounded and satisfies Dirichlet's Condition (F.S. 93) ia (0, tt).
The equations for the temperature will thus be as follows

kept at the temperature v=f{x).


is

v=0, when a;=0 and


and

when y=0.

v=f{x),

Also

we have

a;=7r,
(0

< <

tt)

a;

Lt {v)=0.

Starting with the Sine Series for /(a;),


! sin x-\-a2 sin 2a;+...

where

"Jo
let

f{x') sia nx' dx',

us examine the function v defined by the equation

v=aje~^
Since /(a;)

is

sin

x+a^e-^" sin

bounded and integrable


2

an = ~

2a;+...

in (0,

tt)

and

f""

f{x')smnx'dx',

ttJo

then

|a|

a\ sin nx e-"

Also
yo being

< 2M,

where

\f{x)\

< 2Me

-"'',

<,M

when

in (0,
2/

tt).

= > 0,
/o

an arbitrary positive number.


CO

Now the

2^""^

series

is

terms are independent of x and

convergent and

its

Thus the

v=aie-^smx-\-a2e-'^'>sm2x+

series

regarded as a function of
of X,

when y >

convergent when

x, is

...

(1)

uniformly convergent for any interval

and, regarded as a function of

y^yo'^0.

y.

y, it is

uniformly

TWO-DIMENSIONAL PROBLEMS

90

TLe same

is

true of the series obtained

tion of (1) with respect to

x and y

|^= - S'*^**"^"""'"' ^ "^

Therefore

,^=

and

Y\n^ane-"vsinnx.

+ a^^=^-

Hence
a^2

Further, since (1)

and the sum


of as

a;

by term by term differentia-

in these intervals.

is

= a;^

uniformly convergent in the interval

of the series vanishes

approaches these values

when a;=0 and

tt,

a;=7r, the limit

y being positive. Thus


x=7r are satisfied.
bounded and satisfies Dirichlet's
is

zero,

the boundary conditions at the faces a;=0 and

We

have assumed that f(x)

Conditions in the interval

is

(0, tt).

Therefore the Sine Series


sin x-\-a2 sin 2x+...

tti

converges, and

which /(cc)

is

It follows

its

sum

is

and tt at
f{x) at every point between
all other points.

continuous, and^{/(a;+0)+/(a; 0)}at

from F.S.

73, 1.* that, if

Lt v==f(x) at a point

is

defined

by the

of continuity

=2{/{+0)+/(ic-0)} at aU other

points.

v='^ae-"''smnx

Thus
is

series (1),

the solution of our problem.

This

may

be written

f(x')y](e-"vsinnxsmnx')dx',

Tfjo

since the series under the integral


Ex.

If

the solid

is

uniformly convergent.

bounded by the planes x=0, x=ai and y=h, which

is

are kept at zero temperature, and 2/=0, which

show that

1.

MT ,,

(b -y)
smh
=- y -^^^^-^^ sm
smh
6

is

kept at temperature

/{a;)f

fix)

sm

x dz

'"

and

which three of the boundaries


and the fourth at an arbitrary temperature.
Series and Spherical, Cylindrical and Ellipsoidal

discuss the other three oases in each of

are kept at zero temperature


Cf.

Byerly, Fourier's

Harmonics, pp. 102-104.


*

The argument

of W.8. 73, II. also applies to this question.

Cf.

above,

31.

TWO-DIMENSIONAL PROBLEMS

91

3ase at Temperature Unity.


In the solution of 42, put/(a;)=l, and we have
43.

v=~2^1
V coswTT e-"!' sin nx.
Therefore

^D=e-!'

sin

x+^e-^y

sin

=the imaginary part

_
~

3a;+ ...
of (e-!'+"'+^e-^!'+^^*+...)
1,

l-)-e'(^+3')

2 ^^ 1 e'C^+'y)
l

e-^y+2ie-ysuxx \
-2e-ycosx+e-^y)

IM\
_^/sina;\

1,

Vsinh y/

=^an-if?5^^

Thus

TT

The conjugate function*

to

\smh M/

2,tan~^ (

IS
) ^

i,

/ l+2e-J^cosa;+e-^n

1,

4 ^ U-2e-!' cos x+e-^y)


1,

/cosh y+cos
Vcosh / cos

a; \

a;/

It follows that the lines of flow are given

by

coshw+cosa;
v-^-i
=constant,
cosh?/ cos a;
,

these being orthogonal to the isothermals'


sin
-r-T-

= constant,
a;

smhy

44.

of Conjugate Functions in Problems of Steady

The Use

perature.

Let

^,

);

be real functions of x and y such that


i+iri==f{x+iy)=f{z).

"f

For the

definition

and properties

Cf. Fourier, loc. cit., 205.

of conjugate funotiona see 44.

Tem-

TWO-DIMENSIONAL PROBLEMS

92

Then

^,

are called conjugate functions of x

;;

and

y.

Also

we have

|=-^|,

Therefore

(1)

l^=?f
dx
It follows

that the curves

(2)'

dy

^=constant and ?= constant

are

orthogonal,

Agam,smce

B^^^^d^y'
^5~>
5~s=3
ay^ ax ay

and
it

follows that

and

similarly

Further,

if

t;

is

iJ+32=0'

(3)

^+9^=*^

^*^

a function of x and

such that

g+5=
^5-^+55=0-

we can show that

9a;^
-,

J or

<^>

ay^

dv

dv d^

003

of

8v dn

5-=5> 5^+5-5-

we have

oic

ax

orj

and

dx^~d^\dx)

'^

didr,dxdx'^dr,^\dx) '^ di dx^^ dr, dx^'

Similarly

dy^^di^\dy)

Adding these two

we

d$dr,dydy'^dr,Ady)
results,

and using

(1),

'^ d^ dy^'^ d^ dy^'


(2),

(3),

see that

dx^'^dy^

Thus,

if

we can obtain "a

solution of the equation

9p+9?~

'

(4),

and

(5),

TWO-DIMENSIONAL PROBLEMS

93

boundary conditions at the curves

satisfying certain

this solution in the

^r/

plane

may

be transferred to the xy plane,

the boundaries being the curves in the

ccy

plane which correspond

by the transformation
i+iri=f{x+iy)
to the curves

f=fi,

while the temperatures at these boun-

etc.,

daries correspond to the temperatures at the boundaries in the


f;?

plane.

Suppose that we have the case of the rectangle in the

and that

at

o=fi{r,)

^17

plane

(h<'7<V2)

f =^1,

=/2W at^=f2> {ni<n<Vi)


V=F,{i)

The

Bit

r,=r,

solution of this problem

is

{ii<i<i,)

obtained by breaking

it

up

into

four cases, in each of which three of the boundaries are kept at


zero temperature.

In this

way we

find

sinh ^T<^''~f)

^"'-'^'^

('72

V
where a,

a', 6,1,

:^i,

and

&,/

ft

W^('?2-'?l)

are the coefficients in the Sine Series,

\_V

/2(.)

(2-^1)

^^('?-'?i)

= S'>n'^^^),

^.(a=s^>in'g^^.

TWO-DIMENSIONAL PROBLEMS

94

Substituting for f

from the relation

>;

we have the temperature

in the region bounded by the curves which


correspond to f =fi, etc., these curves being kept at the temperatures
corresponding to/i(ij), etc.
45. Applications of this Method.*
I. The Sector of a Circle.
Consider the transformation

^=-

In this case

and the sector

in the

^t]

;y=-log(-

0,

a and angle a corresponds to the region

of radius

plane.

Thus the equations

de-^dr,^

'''

v=0, when

(0<^)

f=0

and

w=l, when ij0,

lead to

a~2+3~2=0

and ^=7r,

the sector,

0"^^^

v=0, when 0=0 and 0=a,


and

v=\, when r=a.

These equations in (f i;) we have already discussed in 43 when


dealing with the Infinite Eectangular Solid, and their solution is
,

=- tan-i
TT

\smh

;;/

Therefore the temperature in the sector

sm-

v=-

given

by

tan~^

TT

sinh

* Cf.

is

- log -

Mathieu, Gours de Physique MatMmatique, Ch.

III.

TWO-DIMENSIONAL PROBLEMS

95

the boundary r=a had been kept at v=f(d), the problem


would have reduced to the solution of the equations
If

v=0, when
and
and

^=0 and

^=ir,

^=^-^1, when ^=0,


this

II.

has been discussed in 42.

The

Circle.

Consider the transformation

x+iy

^-\-ir,= -ilog-

Then
and the

i=6,
circle

r=a

^=logp

corresponds to the region

0<v, 0<i<27r
of the

^Tj

plane.

Thus the equations

g|2+9^a=0 over

this region,

^^^

v=f(^), when ri=0,


lead to

gj+g^=0 in the

circle,

v=f(6), when r=a.

The

solution of (1)

is

'

|
J

given by

v='^ e-'"'(a cos wf +6 sin w^),


where

^^0=9-

fii')^i'>

=-r7(ncos<'d^',
TTJO

and

b=^\jii')smni'di'.
TT JO

^2)

TWO-DIMENSIONAL PROBLEMS

96

Thus we have

1 e-^i

C^^

=2^ Jo l-2e-"cos

(^-^')

+ e-2/(^')'^^'-

Therefore the temperature in the circle

is

given by

,,=J-[7(0')-2

/ ^" ''a2-2arcos
(0-0') +-^
a^
27rJo
)(,

III.

Twp

Concentric Circles.

may

This

'

be obtained from the same transformation.

solution can readily be found in the

The

form

(acosw^+6sinn|)
'siahw(v2 >;i)
,

y^

>^ sinhn(n i)

+7
where

-^-^:

..

Mf + 6'sm ni),

(a cos

/i(^)

= S (cosM^+6sinw^),

/2()

= 2 cosn^H-6'sinn^),

are the Fourier's Series for/i(^) and/2(^) in the interval

IV.

Two

to 2t.

Intersecting or Non-Intersecting Circles.

Consider the transformation

^+ir,=\og

x^l-iy
xl
i

Then
where

r^

^=logJ,
and

rj are

v=0,-92,

the distances from the points

to the point P{x, y),

and

6^, 6^ are the angles

with the positive direction of the axis of

A (1,

AP

0),

and

B{1, 0),
BP make

x.

Thus ^=constant represents the system of coaxal circles with


A, B as limiting points, and >;=constant represents the system of
circles passing through A, B, these two sets of curves, as in all
cases of conjugate functions, being orthogonal.
With this notajiion
the xy plane is given by 7r<)j<7r and oo <i$<iKi
the lower
= tt; the lines Ax and Bx' are ;?=0; and the
side of BA is
;

;j

TWO-DIMENSIONAL PROBLEMS
upper side of
the point

BA

f=

Also

is .)=7r.

is

the point

97

^=+00 and B
,

is

00.
J'

We proceed to apply this transformation to several cases in which


the region in the ay plane

as in Fig.

bounded by

and

(0

^=0

iiiKiKii)

and r,=T,

over ^=^2> '?=0, and ri=Tr,

=/(/) over

Then we 1.
have

where

v= -^
> a

V
,

<*==-

is

<,,<7r)

6.

mi.

^=^1.

^Vr

sinhw(A ^)
-^,

sin

nri,

"sinhw(^a-^i)

f{ri')^va.nij' dn'.

easy to extend this solution to the case where


v=fi{ri)

orei i=ii,

(0<>i<7r)

'v=Mv)

over ^=^2,

(0

v^F,(i) over ,=0,

v=F,{i) over
(ii)

arcs of these circles,

i=ii and f=^2,

v=0

Let

It

is

Consider the region bounded by

(i)

r,=7r.

< <
.;

tt)

{ii<i<i,)
ii^<i<i,)

Consider the region bounded by the two complete circles

=^i and ^=^2

surrounding the limiting point A.

Let v=fi{ri) over f=fi, and


solution

is

v=0

over

=^2.

obviously

siDhndz-i)

(acosm)j+6sinw>;),
=s
f sinhniis-ii)
C.O.H.

Then the

TWO-DIMENSIONAL PROBLEMS

98

where

o=oJiTT J

fi{n')d>i')
IT

1 f"

=-

ttJ

fi(ri')C0Snri'd,i',

-w

1 ["

b=-\
TTJ

Similarly,

when

v=f2(ri)

over

^=^2

^ siiihw(f A)

have
where

fi(>i)smnrid,i'.

-n

a'

and

in the interval

b' are

tt

to

v=0

^.nd

i,

over f=fi,
,>

the coefficients in the Fourier's Series for/2(


tt.

Adding these two results we have the solution for the case
the circles ^^ and ^2 at temperatures /^(ij) and/2(r;).
It is clear that if /i(;) and/2(>?) are constant and equal to % and
respectively, we have only to solve the equations

v=Vi, when i==ii,


v=V2, when ^=^2-

The

solution

is

v=Vi,

{j^) +^2 (1^^^)

ri8.

(iii)

1.

Consider the region bounded by

.)='I2,

(0<^<>)
(0<^<co)

^=0,

W<n<

')=ii,

as in Fig.

7.

V2)

TWO-DIMENSIONAL PROBLEMS
(a)

Let

w=/i()) over

v=0
v=0
Then the

solution

(j8)

{rii<rj<r,^)

over

j=ji,

(0

over

>)=j/2.

(0

<^<oo)
<^<oo)

clearly

is

V ae-'^

=
where

f=0,

sin

/''~''^\

wtt,

["""^^. WTr
f/iC,,') sin
)2 )lJr,i
('72 f!!)

a=
Again,

09

cZ^'.

(Cf. -P.S., 98, (7).)

let

i;=0

over ri=^^,

(0<^<co)
(0<<oo)

u=0

over |=0.

(0

=/2(^) over

,)=;,i,

< |<

oo

Suppose in the iirst place that the point A {io6 ) is excluded


by a circle ^=a, where a is a very large positive quantity, and
that the part of this circle included in the region is kept at zero
temperature.

Then we would have


ainh ~(ri,^-r,)

sin

an

Sinh- (;;2 j?l)


where

ar:

= ~\' m')s\n'^i'M'.
(X

To obtain the

Jo

solufion of our problem,

Proceeding as in F.8.

we must

let

a tend to

oo

118, in the case of Fourier's Integral, put

a = \

and

= AA.
a

This result might have been deduced from Fourier's Sine Integral

iovMi), namely,

-rcZXsinX^r/2(f)sinXrdf.
Tjo

(y) In the

same way,
'^

Jo
if

^faii)' ""^hen

v=

0,

v = 0,

r,

,,2,

when =
when ^ = 0,
ri

t]j^,

(0

< ^<

oo

(0<f<oo)
{ri^<v<l2)

TWO-DIMENSIONAL PROBLEMS

100

i"^M'?-'?i)
^ = - f" dX

we have
(S)

By

adding

(a), (/3),

and

^i^,

we

(y),

^^ f"y

(^^j

.j^

xf df
when

find the solution

the

three surfaces are kept at v=fi{ri), v^f^i^), and ^=/s(f).


(iv)

Consider the region bounded by

(-<<^<')
= ii
'v=fi{i) oyer
'?

Let

= %

yi

=/2(^) over

and

Then we

find, as

>?

;;2-

above,

V. Confocal Ellipses or Hyperbolas.


Consider the transformation

.x + iy
^ + 1^ = cosh-^
.

x+iy ccoah{^+i^).

or

Then

a;

= ccosh^cos>;,

2/

= c sinh ^ sin

-4.+^^
sinh^^

and

?;,

cosh^f

3/'

cos^ij

sin^j?

Thus the curves f= constant, and ;;= constant, are a set


ellipses and hyperbolas, and the xy plane is given by w

of confocal

< < t and

-< ^
of

r]

<

00

the lower part of the xy plane having negative values

and the upper part

(i)

Two

i;

positive values.

Confocal Ellipses.

Consider the region bounded

by i=^i and ^=^2-

v=f^(ti) over

Let

^=fi,

v^fiM over i=ii.


Then, as above,
^=

^sinh%(^o #)

^""^^''+^^^"^'')

?
+ S sinh^(|,4V"" cos, + 6 sm,),
sinhn(g-|y

TWO-DIMENSIONAL PROBLEMS
where

and

a, b, a',

for/i((j)

the coefficients in the Fourier's Series

6' are

and/2(v) in the interval

Two

(ii)

-101

v to

tt.

Semi-Ellipses and the part of the major axis between

In this case the region

bounded by

is

i=i^ and i=i^


= 7r.
r,=0 and

(0<;?<7r)

iii<i<Q

,,

Let

^=/i('?) over f=fi,

'"=M>]) over ^=^2,

=0

and
It

is

over ij=0 and

clear that the solution

V
^
+ Vein
V
,

where a and

a' are

;?

= 7r.

is

sinh 11(^2-^1)
.

sin nri

sinh 71(^-^1)

^^r

7^^

H^,

'sinhw(f2-^i)

the coefficients in the Sine Series for/i()j) and

/2W.
(iii)

Semi-ElUpse.

In this case the region

^=0
^=0

is

bounded by

and i=i^,
and r,=-7r.

Let v=f{ri) over i=ii, and

Then

temperature.

^=

where a
(iv)

In

is

(0

< <

,r)

;?

(0<f <fi).
the major axis be at zero

let

sinh n^
^
2sm^,^-j^,
.

the coefficient in the Sine Series for /(;;).

Complete Ellipse.

we have

this case

to satisfy

3^2"^ a^a

^ =/(?),

(-7r< >j<7r)j

'

when ^= ^1

- TT <

;;

<

tt)

Also there must be no discontinuity in the temperature or the


flow of heat as

we

cross the

major axis or pass along

All these conditions are satisfied

cosh nP

by the
,

it.

expression

sinh

n^

^=S(^^5^'='^''+^^iHh#,^^"S'

TWO-DIMENSIONAL PROBLEMS

102

where a and b are the

in the interval

tt

to

(v) Quadrilateral

coefficients in the Fourier's Series for /(>/)

tt.

bounded by

the

Arcs of two Confocal Ellipses

and Hyperbolas.
and the

solution

In the cases of steady motion considered in the preceding

articles,

This reduces to the rectangle in the

^i;

plane,

follows.

46. Sources

and Sinks in Steady Temperature.

the supply of heat which maintains the steady temperature

is

applied

On the analogy of the flow of electric

at the boundaries of the solid.

when the

currents along thin conducting plates,

current

is

conveyed

and withdrawn by another, we may


imagine the steady flow of heat in two dimensions or the flow of
heat in thin plates to be produced by the introduction of a quantity
of heat at one or more points and its withdrawal at others.
These
points may be called Sources and Sinks of Heat.
to the plate

by one

electrode,

In this

case,

we

point at which heat

describe a small circle of radius r round a


is

steadily flowing into the plate, in the limit

the flow of heat out through this circle must be equal to the flow
in at the source.

Hence the

must take the form


where m

is

solution of our equation

a solution which remains

the source, and

m is

+ u,

log r

finite as

the point approaches

the quantity of heat introduced there per unit

time.

Similarly at a sink the part of v which tends to infinity

must be

tfl

equal to

^^ log

r.

Consider the part of the xy plane for which y>Q, and let the
boundary y = Ohe kept at temperature zero, while there is a source
of strength

m at (0, y^).

Then we have

= -p^ log ^ T
4.K-7r

since this expression satisfies


It

win be seen that

L
^x^ + {y-yY'
all

the conditions of the problem.

this solution is obtained

by putting a

sink

TWO-DIMENSIONAL PROBLEMS

at (0,

which

/o))

will

balance the source at

103

Indeed

(0, y^).

the use of Sources and Sinks in two-dimensional problems of Steady

Temperature

is

exactly analogous to the use of Images in Electro-

and Hydrodynamics, and the reader

statics

is

referred to the dis-

cussions in the books on these subjects for further illustrations.


It is clear that the

functions

is

Z=X

where

and
(j>,

method

(f>

transformation by conjugate

+iY

Since

if

=f(x+iy),

Zo=Xo+iYo

=f{xo+iyo),

Zo'=Xo-iYo

=f{xoiyo),

looked upon as a function of

Also

of

appUcable in this connection.

also

{x, y), satisfies

7=0

in the

(xq, y^),

which

vanishes at the boundary which corresponds to

XY plane,

and

is

infinite as

h-^^ log r at the point


/ATT

corresponds to the point {Xq, Yq) in the

XY plane.

47. Variable Temperature.

In this case the equation of conduction

dv_
and we obtain the solution as

the

initial

(d^v

in 16 in the

If

form

temperature being
^ =/('.

y)-

bounded by /=0 kept at temperature zero and the


temperature is v=f{x, y), when - cxj <a;<oo and 0<2/, we

the soUd

initial

is

dH\

is

have
-1

v=

(>

T^A

To obtain

(<

f,y')\e

this solution

beyond the plane

y=0

{x-iif+(v-y'f

-e

(j;-xT+(.v+y') \

we may suppose

and that there

is

]dxdy.

the solid continued

a symmetrical initial

104

TWO-DIMENSIONAL PROBLEMS

distribution of temperature in the

added region y

<

which

will

cause the plane y


always to keep zero temperature: that is,
we take the initial temperature at {x',y'), (y'
0), to be equal in

>

absolute value but opposite in sign, to that at

The question

it,

y').

tempefature at the boundary y=0,


will be referred to in 74, 86 and 87.

of a variable

or of radiation across

(a;',

CHAPTEE VI
THE FLOW OF HEAT IN A EECTANGULAE
PAEALLELEPIPED
48. Introductory.

Several of the methods of finding the conductivity of solids which

we have

discussed in the previous chapters cannot be applied to

poor conductors, since the amount of heat


the bar

by

along the bar, and the emissivity


that

it is

lost at the surface of

radiation becomes large compared with that conducted

best to have

it,

when

is

such an uncertain quantity

possible,

reduced to the rank of a

With poor conductors this would be impossible


in the bar methods.
But for the cube, the sphere, and the cylinder
the mathematical problem can be solved and its solution applied
to the evaluation of the thermal constants. In this chapter we shall
small correction.

discuss the case of the rectangular parallelepiped.

perature the solution

much

is

For steady tem-

given by a rather complicated

series,

without

practical value, but for different problems of variable tem-

perature

we obtain

results

immediately applicable to experimental

investigation.
49. Steady Temperature.

Consider the solid bounded by the planes x=0,


y==zb

z=0, z=c.

The equation

for the temperature

We take the following surface conditions

v=Vi, when x=0,


v=Vi, when x=a,

and

the other faces at zero.


105

x=a
is

y=0,

IN RECTANGULAR PARALLELEPIPED

FLOW OF HEAT

106
It

is

clear that the expression


t),
-i^

sinhi{a a;)+i'osi!ih?a;
'
,

sinh La
satisfies all

5-^ sin ntrz


.

the conditions provided that

Therefore the solution of the problem


CO

CO

%\ sinh

sm^-sin

sinh^c

11'

Y" y] A^^ sin

provided that

given by

is

(a x) +1)2' sinh ^^j^.^ mTry ^.^ mrz

^m,T
^=S
S-^"'."
i=l n = l

r.

be

imrv

sin

-^

sin
^
"

= 1.*
^

tn-Try

^y'^mSinr-^ = i)

But

a =

when

(1 coswitt).

Therefore

sin^-^i^sin^^2lll!

_16'^ ^Visinhi(tt a;)+t^2sinhia;


where
50. Steady

When

(2g+l)

{2p+l)

_ {2p+lY

i^

&

iEhTS

''"^^oj^'o

'

(2q+lY

Temperature (cowiwMed).

radiation takes place at the faces

and the faces x=0,

we can obtain

x=a

y=0, y=b, z=0, z=c,

are kept at temperatures

v^, v^,

as before,

a similar solution of the problem.

In this case the surface conditions are

v^Vi when

51-^1' = 0,

when

^-\-}bv=0,

when z=0;

az
*

a;=0, and

For a rigorous treatment

t/

= 0;

v=v^ when x=a]

= 0,

\-hv

\-hv=^0,

when y = h;
when z=c.

oz

of this question it

would be necessary to justify


The same remark
v.

the term by term differentiation of the double series for


applies to the other problems discussed in this chapter,

Cf. 12, 30' above, and Moore: "On Convergence Factors in Double Series,
and the Double Fourier's Series," Trans. Amer. Math. Soc., 14, p. 99, 1913.
Also the same author's paper in Bull. Amer. Math. 80c., 25, p. 274, 1919.

FLOW OF HEAT

IN RECTANGULAR PARALLELEPIPED

107

The expression

~rx) +Vo

Vi sinh. lla

.h.

sinh. Ix /

(cos mi/ H

^^

-^

sinmy

,h.
amnz)\

\/
\

cos nz-]

satisfies

the equation of conduction provided that

Also

satisfies

it

values of

I,

the surface conditions at

and those at the other

faces,

x=0 and x=a


ii

for all

m, n are the positive

roots of
,

(cf.

2'mh

tanmo=

2nh

and

5-^

tanic=

36).

Now we

have seen in

assumptions as to

36, that, with- certain

the possibility of the expansion of an arbitrary function in a series

whose terms are

of the type

Yr=coa mry-\

sm

niry,

the coefficients in the expansion

f(y)=A,Y,+A,Y,+...
are given

But

^r= ^^^,^j^:;^^_^^f^ y{y) Y,.dy.

by

Yrdy=\

Jo

smmryjdy

(cosmry-\

^m^ h= ^

= (sinm^feH

(l

mr\

tan

Also,

'

***

mr

CQsmrb)

h
tan|m2j-+i^=
(cf.

m.2r+l

37).

Hence

f*
I

r,.(^2/

= 0,

when

r is

even;

Jo

m/=,

4^

VI

when

r is

^^

odd

Similar results follow for the corresponcjing expression Z^.

FLOW OF HEAT

108

RECTANGULAR PARALLELEPIPED

IN

Therefore the solution of our problem

is

v ^i '^^"^^ l(ax) + V2 sinh Ix


~ 1fi*2V
sinh la
^0 ^0
where
It

is

i^=2j,+i+W25+i-

do not lend themselves to numerical


and that they are not suitable for the evaluation of the

clear that these solutions

calculation,

thermal constants.

No

51. Variable Temperature.

Radiation at the Surface.

Let the solid be the rectangular parallelepiped of the


sections,
X,

y and

we have

the

and

z,

of

Then

the faces being kept at zero temperature.

all

two

last

temperature being an arbitrary function

initial

to satisfy the equations,

=kVH

-^

in the solid;

(1)

=0, when x=Q, y=0, z=0\


and

v=f{x,

y, z),

,.

x=a, y=b, z=cj

'

wheia.t=0

(3)

The expression
^^

and

satisfies (1)

sm a; sm -?- V
^

'

"

mr
sin

(2).

Extending Fourier's Sine Series to the case oif(x,


have
f^^' y'

'^)

^SS

= ""''
f

("

,.,

/(a;',

a;

=S
l

y\

,v

z')

-^ysm
z
O

sin

't

sm

we would

y, z),

a;'

WITT

n-K

sm -^y %in

z ix

T
J
dy dz
,

=X

"

"

Itt

rrnr

.__

mr

S ^Am,nii^^-xam-yysm~z,sa:y.
a

m=l n=l

solution of our problem will be given

by

^=S2SAm.nSm-a;sm-g-3/sm ^e

^-

' '

"

Ai__n being the definite integral

f"

f" ,,

iTT

mw

mr

,"1

"

Hence the

Z> sm a

l=lm = ln = l

CP

IJ

FLOW OF HEAT
When /(a;,

IN RECTANGULAR PARALLELEPIPED
constant and equal to

y, z) is

109

v^,

Therefore, in this case,

^=0^l:2i:{-(2;'+l)fsin(2,+l)!^sin(2.+l)^
"0
\'"
C
3j=09=0r=0

">

=
(

ism

NTT/

sin-^sm
1

Sttx

try

vz

c'

-KlT'(-,+l-.+-^t

Tra;

StTW

-n-Z

+7:Sin
sm-T-^sin
e
3
a
6
c

Bab

6!>

^^

+jrSin sm-i^sm^
1

6
,

(2r+l)' \

(2a+l)'

(2;j+l)(2j+l)(2r+l)

0.

3 / (2p+l)'
\
a

-icn'(-.+ l;+\)t
\'

''

'

"''

'^>

Thus, for large values of

t,

the series

is

rapidly converging, and

to a close approximation
TTX
Try
V

- Sin sin T^sm


= f4iy
a
W/
%
.

Also,

we note that

_(-i^4.i+l)(

-rrz

^^

''

"

'

at the pouit
(^a, ^0, |c),

the 2nd, 3rd, and 4th terms disappear, so that

e
TT

is

a good approximation to the value

values of

In applying
are equal,

of v at this point for large

t.

this result it is usual to take a cube, so that a, b

and
24 V3
TT

-3.r

and

110

FLOW OF HEAT

That

sufficient

IN RECTANGULAR PARALLELEPIPED

time has passed to allow this to represent the

state of the temperature

may

be tested by seeing

if

the temperature

readings at this point follow the exponential curve.


noticed that the 5th, 6th,

It will be

and 7th terms are

2473

-27^I

and that these should be negHgible.


If Vi

and

V2 are the readings of the

temperature at

ij

and

t^,

the

equation

=e3/c-^(<2-'l)
"

Vl

gives the value of

k.

52. Variable

Temperature (continued).

When

is

there

medium

Radiation at the

Surface.'^

radiation at the faces of the parallelepiped into

at zero temperature,

we have

to satisfy the equations,

^=KVh> in the solid

(1)

ot

-^
ox

when x=a;

(2)

_p.jf.hv=0, when y=0,

^+hv==0, when y = b;

(3)

^+hv=0,

^+hv=0, when z=c;

(4)

ax

when 2=0,

oz

oz

and

+ hv=0,

i^-t-hv0, when a;=0,

v=f(x,y,z), when t=0

It is clear that the expression

where

Z=cos

ax-]

r=cos/32/+o

Z=cos

yz-i

sin ax,

sin

sin yz,

y
* Cf. Fourier, loc.

cit.,

;8jr,

Ch. VIII.

(5)

FLOW OP HEAT
and

a,

/3,

IN RECTANGULAR PARALLELEPIPED

111

are positive roots of

2ah

tan aa

2/3A

tan/36=

satisfies (1), (2), (3)

and

(4).

.(6)

(Cf. 36.)

Hence assuming the possibility of the expansion of the given


arbitrary function in a triple series whose terms are of this nature,

we have

as the solution of our problem

where

.4^, p, y is

fi

the coefficient of the term in

and the summation

is

roots of the equations

taken over the

(6).

In the case when the

initial

temperature

to Vg, this expression simplifies, as in 50,

__

If

may

is

<

4,313

CO

X^Y^Zy in the expansion


number of positive

infinite

CO

V V ^^

is

constant and equal

and we have

t;-

-^ap+i

-"-aa+i

so great that in each of these series the terms after the

be neglected, we have

In applying this solution to the case of the cube, we have

where

and

Zi=cos

aiX-\

yi=cos

ai/H

Zi=cos

aiZ-\

sin aiX,

sin oiy,

sin a^z.

first

112

FLOW OF HEAT

Therefore,

if v-y

IN

RECTANGULAR PARALLELEPIPED

and V2 are the temperatures at any point at t^ and t^,

Hence, k having been found by other means, aj

But

tania,a=".

Thus we have the value

of h.

may

be obtained.

CHAPTER

VII

THE FLOW OF HEAT IN A CIECULAR CYLINDER


53. Introductory.

We

have seen in

6 that the equation of conduction, when

expressed in cylindrical co-ordinates, becomes

If

a circular cylinder whose axis coincides with the axis of z

is

and boundary conditions are independent

of

the temperature will be a function of

heated,

and the

initial

the coordinates 6 and

and

only,

and

z,

this equation reduces to

/d^v

Idv

dt~'^ \dr^

r dr

dv

In this case the flow of heat takes place in planes perpendicular


to the axis,

When

and the Hnes of flow are radial.


initial and boundary conditions do not contain

the

z,

the flow of heat again takes place in planes perpendicular to the


axis,

and the equation


dv

when the

of conduction reduces to

/B^v

dv

9V

and boundary conditions do not contain


6, the flow of heat takes place in planes through the axis, and the
equation of conduction becomes
Again,

initial

dv

_ /dH

3^

S^i"'
I

dt~'''\d^^^rd'r'^M.
O.CH.

113

FLOW OF HEAT

114

54. Infinite Cylinder.


If

the solid

IN

A CIRCULAR CYLINDER

Steady Temperature.

a hollow circular cylinder, whose inner and outer

is

and r^, and the surfaces are kept at the constant


temperatures v^ and V2, the equations for the temperature become
radii are r^

(Pv

= v^, when
= V2, when

and

Ti^us

Idv

^ ^

= r^,

r=

?'2-

K-i^i) logr+^i logr^-Vi logr^

logr^-logri
If

the cyUnder

is

solid

and heat

ciding with its axis,

is

suppHed by an

electric current

through a uniform straight wire

of constant strength passing

coin-

and the heating has gone on long enough

for

the steady state of temperature to be attained, the rate of flow


of heat out

through any concentric cylinder

at which heat

Thus,

if

ff

by the

is

equal to the rate

is

supplied to the wire

is

the heat supplied per unit length per second, and

current.

K the Thermal Conductivity,


or
Therefore on integration

Vi

and

V2

we have

being the temperatures at

rj

and

H=PR,

But

I being the strength of the current and

Thus

r=a

R the resistance.

27rK{Vi~V2)=I^Rlog^-^.*

55. Infinite Cylinder.

Let the

r^.

Variable Temperature.

temperature be given by v=f{r) and let the surface


be kept at a constant temperature, which may be taken as
initial

zero.f
*Cf. Niven, London, Proc. B. Soc, 76
(a), p. 433, 1905.

(a), p. 34,

1905;

Lees, London, Phil

Trans. B. Soc, 204


t

If

the constant surface temperature

zero temperature

by putting v=Vg+w.

is Vq,

we may reduce

this to the case

o\

FLOW OF HEAT
The equations

IN A CIRCULAR CYLINDER

115

for v are as follows

dv

/d^v

v=0,
and

,ldv\

if.^^.

when r=a,
when t=0.
where m is a function

v=f{r),

If we put
must have

=e~'"'''M,

d^u

du

of r only,

then we

^+r^ + ""=^'
,

which

is

As the

Bessel's equation of order zero.

solution of the second kind

is infinite

at

r=0, the particular

integral of the temperature equation suitable for our problem

is

v==AJo(ar)e-'"''i,

where Jo{x)

To

is

Bessel's function of order zero of the first

satisfy the

boundary condition a must be a root


Jo{aa)

kind*

of

= 0.

known

that this equation has no imaginary roots or repeated


roots, and that it has an infinite number of real positive roots
It

is

"i. oa, 03....

Also to each positive root a there corresponds a negative root


If f(r)

can be expanded in the

a.f

series

AiJo{air)+A2Jo(a2r)+ ...
the problem will be satisfied by
,

the conditions of

CO

Assuming

and
by term, we can obtain the

for the present the possibility of the expansion J

that the series can be integrated term


values of the coefiicients

by the help

of the

two important

definite

integrals to be discussed in the next section,


*

For information as to the Bessel's functions see Appendix

I.

Gray and Mathews,

Bessel Functions, Ch. V., 1895 ; Watson, Theory


This important work has been passing through
of Bessel Functions, Ch. XV., 1922.
own. I am indebted to Professor Watson for
thejpress at the same time as

f Of.

my

the references to

it

here and in the pages which follow.

J For a discussion of the possibility of expanding an arbitrary function in a


Hobson, London, Proc. Math. Soc. (Ser. 2), 7, 1909

series of Bessel's functions, see

Moore, Trans. Amer. Math. Sac, 10, 1909


Proc. Math. Soc. (Ser.

2), 18, 1920.

12, 1911

The subject

and

21, 1920

is also

Young, London,

treated in Dini, Serie di

Fourier, pp. 246-269, 1880, and Ford, Studies in Divergent Series

Ch.V.. 1916.

and Summabilify,

FLOW OF HEAT IN A OIRCULAB CYLINDER

116

The Integrals

56.

rrJ,,(ar)J(l3r)dr

and

Putting u=J(ar) and v=J,{^r),

^\{J^(ar)fdr.*

we have from Bessel's

equation,

rl(4)+("-?)='''

dv-y

r du

and

this vanishes

dv~\''

when

aJ(^a)J'(aa)-;8J-(aa)J'(/3a)=0,

where
'^"'^"'^^^\5r'^"^'^Vr-a

Thus when a and

two

are

/3

difEerent positive roots of

(i)

J(aa)=0,

or

(ii)

J'(aa)=0,

or

(iii)

we have

aJ n

{aa)

J%J(ar)JJ/3r)dr=0.

M = 0.

Again, since

''V'/^fe?

dud/

du

drV

dr

rfr

Therefore
rfj-V

and

dr

by

parts and

it

dr

follows that
/du\'^

laM ru^d/r= r''\^\


The convergence

shall be greater than


shall

n^-^ O,

+ aV -5dr

"j?w*+[('jT--]:=-

Integrate

,.(1)

+ 'hJn{aa) = 0,

)De

of the integrals

1.

when

+(aV-2)M2

requires that the real part of n

In the applications of these integrals in the text we

dealing with n real and not less than zero.

FLOW OF HEAT IN A CIRCULAR CYLINDER

117

Therefore

rr{J(aT)fdr=^^,[a'a'(J^{aa)f+{a'a'-nWn{aam
-^a

Jo

=|'[(J^'M)^+(i-^,)(^M)^]Thus

when a

(i)

a root of J{aa)

is

= 0,*

r{J{ar)fc!^=Uj'{aa)f;
Jc
(ii)

when a

is

a root of J'(aa)

= 0,

nf.2

iM 2

(2)

|%(J(ar))2cfo-=|'(l-^,j(J(aa))^

and

when a

(iii)

is

a root of aJ'(aa)+/iJ(aa)=0,

Jo-'"
Applications of these Integrals.

57. Infinite Cylinder {continued).

We may

apply these results to the case of the Infinite Cylinder,

assuming the possibility of the expansions needed in each


and that they may be integrated term by term.t
I.

r=a

Surface

at temperature zero.

case,

Initial temperature v=f(r).

In this case we take

f(r)=AiJo(air)+A2Jo(a2r)+...

where

ai, ag

...

are the positive roots of

Jo{aa)=0.

AA''r{Jo{ar))^dr=

Thus

Jo

rf{r)Jo{a^r)dv,

Jo

since

rJ a{a^r)J oia^r) dr=Q.

Jo

But
9
Therefore i>= -5
a^
*It
are

is

known

all real

rf{r)Jo{a^r)d^

0,
I

that the roots of this equation and the others in

and not repeated.

Cf.

Watson,

loc. cit., 15.

(ii)

and

(iii)

23, 15. 25.

treatment of some of the problems named in this section


f For a more rigorous
and the next, reference may be made to Moore's papers in Trans. Amer. Math.
Soc. 10. 1909: 12. 1911: and 14, 1913.

FLOW OF HEAT

118

ra

Surface

II.

A CIRCULAR CYLINDER

IN

impervious

Initial temperature

to heat.

v=f(i

In this case we take


/(r)=^iJo(ai')+AJo(2') + "-

where

oj, ag,

Jo'(aa)

Then, since

= 0.

r[J o{ar,r)Yd/r=-^{Jn{arfl)Y,

rf{r)Jo{ar)d^

we have

>

are the positive roots of

= ^S -"'' f

^oM-

"'

III. Radiation at surface

(j(,)).

r=a

into a

medium

at zero temperatur

Initial temperature v=f{r).

In this case we take


/(r)

where

ai, oa,

= ^iJo(ai-)+^2'^o(a2-) + aJg (aa)

Then, since

>

are the positive roots of

Jo

+ hJ^iaa) = 0.
a^

r{Jo{ar))^dr=^-i^{h^+aJ)(Jo{aa))^
-n

aj\

rf(r)Jo{anr)dr

we have
IV.

^=^2 S^-'- y;^2)(j(^,)). ^o(a.r).


Surface r=aat zero temperature. Initial temperature

v =f{r,

In this case the equation of conduction becomes


1 dH''
fd^v Idv
'^Kdr^'^rdr'^r^dd^J'

dv

dt

and the expression

e -"''/ {ar)(^

is

periodic in

Now

cosn0+5

n being taken

satisfies this equation,

with period

sinm6*)

integral as the temperatu

27r.

take the Fourier's Series for f{r,

6),

namely,

CO

f(r, 6)

where

=2

('*'

1 f"
= -I f (r,

&,,=

-T
"J

and

^ ^^

f(r,

6)

+K sin

'^^).

cos nddO,

6)smnede,

IT

ao=^Wfir,6)de.

(n^l)

FLOW OF HEAT
These

coefficients are functions of

of Bessel's functions of the

where

IN A CIRCULAR CYLINDER

ai, a^,

...

as,

Expand them

r.

wth order,

119

in the series

e.g.

are the positive roots of

J(aa)=0.

Then we have

^-'==

'='*^'^"('-)'-^'-^^'
7ra''(.7j(,))4 r_/^'' ^^

Thus we obtain our

^'=2

solution in the form

(^. ^^

V. Radiation at surface

nd+Bn^s sinn6)J(agr)e~'"'t.

r=a

into

a medium at zero temperature.

Initial temperature

v=f{r, d).
In this case we take the Fourier's Series for /(r,
00

/(".

S)

=S

('^^ *^ **^

6),

namely,

+ ^ '*^)'

11=0

as in IV.

The

coefficients are functions of

r.

Expand them

of Bessel's functions of the ith order, e.g.

in the series

00

where

ai, a^, are

the positive roots of the equation

aJ'(aa)+AJ(aa)=0.

Then we have
2

f{r,Q)Jo{asr)rdrde,

^os=--^r-^~rmTT-nAvA'' f

A,=

f"
7
^^Jt
r-"/( S)cosnejJasr)TdTde,
7ra^[as'+h^~){Jn(asa))^'' ^

'^'^

r r-"f{r, 6) sin neJ {asr)r dr dO,

Jo

^a'^(as^+h^-^^Wn(asa))'-

FLOW OF HEAT

120

IN A CIRCULAR CYLINDER

yuj,y
^=2 '^{^n,s'^osnd+Bn^sSiD.nd)Jn{asr)e-

and

.lin s UV.U

VI. Surface

r=a

.i^n g uiij, ,!,>,;.,

.1/.,

Initial temperature

at zero.

v=f{r,

6, z).

In this case we have

dt~''\dr^'^rdr'^r^de^'^dz^)'

and

sin
is

?^az

e-''('''+'''>*J(ur)*'P^w9

a particular

sin

integral.

Now expand f{r,

6, z)

in the Fourier's Series


(a cos nd-\-bn sin nd)

?l=0

The
F{r,

and 6 are functions

coefficients a

z)

and

Expand these functions


by the positive roots of

2 ^"

(^) "^n

(^')

((,

{z)

-co

= -\

da\

(pr,

J XI

^n{z)=-\ da\
"Jo

Thus we get our solution

2S

/'

and ^(z), namely,

/ao

"TJo

"=

denoted by

Finally, take the Fourier's Integrals for <p{z)

"

z,

= 0,

F (, 2) =

let

and

in the series of Bessel's functions given

J(f^a)

and

of r

ff(r, z).

O)

cos a {/3-z) d^,

xlr{/3)cosa{l3-z)dl3.

J CO

in the

form

coo

e -=(''^+''^) J(^r) [^(/3)

cos nd+y}r^{^)

smnQ}

Xcosa(;8 z)c?a(^/3,
bhe

summation with regard to

jj.

being over the positive roots of

J(/xa)=0.
VII. Surface

of

r=a at v=F{6,

z).

we reduce this
steady temperature, by putting

As shown

in 9,

Initial

temperatwe v=f{r, 0, z).


and a problem

to the case V.,

V=U-\-V),

FLOW OF HEAT
and choosing

IN A CIRCULAR CYLINDER

a function of

u,

u=F{d,
and

w to

d,

z),

and

-=- =/c Vm),

satisfy

w=f{r,

find

z only, to satisfy

when r=a,j

6,

z)u, when t=0,

w=0,
To

121

u we expand

when r=a.

l'(9, z), as

above, in the Fourier's Series

CO

^{anCosnd+bamn6),
11=0

a and 6 being in this case functions of

and

<pn(z)

and

z,

which we denote by

Then we take the Fourier's


and u is given by the equation

\/r(z).

^/^(z),

Integrals for <p{z)

XCOSa{^z)dad^.

To

find

w we

proceed as in VI. above.

58. Semi-Infinite Cylinder.

Let the axis


let its

of the

Steady Temperature.

cyHnder as before

base be the plane z=0.

We

along the axis of z and

lie

shall exapaine first of all the

when the base is kept at temperature


medium at zero temperature
at the surface r=a. When we put/(r)='Uo the solution of this problem will correspond to the exact discussion of the Flow of Heat
case of steady temperature

v=f{r) and radiation takes place into a

in a

Rod which

has been treated in

the cross-section

is

20,

with the assumption that

so small that the temperature over

it

may

be

However, when radiation


centre.
takes place at the surface, there must be a flow of heat outwards
from the middle of the rod, and the assumption of Linear Flow
taken as equal to that at

of

Heat

affairs.

its

serves only as an approximation to the actual state of

This approximation

is

admissible

when the

emissivity

Bar
Methods of determining the Conductivity are employed only for
good conductors. We have seen in 52 that in the case of poor
is

poor and the conductivity good.*

Cf.

Peek, Phil. Mag., London (Ser.

For

6), 4,

this reason the

1902.

FLOW OF HEAT

122

IK A CIRCULAR CYLINDER

conductors experiments have been conducted on cubical blocks

and in 63, 66 we shall find that cylinders and


for the same purpose.
employed
be

of the substance,

spheres

may

In this case of steady temperature the equations for v are

<<'<^<^)

IMhw'-"'
v=f{r),
dv
^^-\-hv
or

= 0,

when z=0,

(2)

when r=a

(3)

The expression
is

e~<^JQ(ar)

a particular integral of

and

(1)

satisfies (3) if

aJo(aa)-\-hJo{aa)=0

Thus we take

/(r)=^i<7o(ai'')+-42<7o(2'")

(4)

+ "-

>

00

and

=24e--.Vo(ar).
1

When

V the first

is

small, the roots a^, a^,

term in

Further,

if

ai^a^

may

and we may take

for

be neglected, we have

J{aja)

Therefore, from

increase rapidly,

...

this expansion.*

=l

and

J'{aja)= -^a^a.

- ^a^^a

(4),

or

+h=0,

.=vit will be noticed that this requires

^1 to

this

approximation the value

Thus we have
and

ah to be small, and that we obtain for


is constant and equal to 7.

F when/(r)

= Ve ^

'

this agrees with the solution of 20.

59. Semi-Infinite Cylinder.

We

shall

examine

first

Variable Temperature.

the problem of the Semi-Infinite Cylinder,

given that

v=f(r,

*Cf. Watson,
10, p. .S97, 1909.

6, z) initially,

v=Q, when r=a and z=0.

and
loc.

cit.,

15.23,

15.

32;

Moore, Trans. Amer. Math. Soc,

FLOW OF HEAT

A CIRCULAR CYLINDER

IN

123

In this case we start with the expression


e-(a^+^)

/^)
J^^
'

which

cos

^g ^.^

sin

-=kV^v,

satisfies

at

and

also the surface conditions

Now

expand /(/,

6, z)

when z=0 and r=a,

if

J(^a) = 0.

in the Fourier's Series

CO

nd+bn sin m0),

(n cos

a^ and 6 being functions of r and

Then expand J(r, z) and 6r(r,


given by the positive roots of
and take Fourier's Sine

and 6r{r,

z,

denoted by Fn

z)

in the series of Bess el's functions

(r, z)

Integrals for the coefficients ^,i(z)

and

z).

y{/n{z)

terms in these expansions.


In this way we find the solution in the form

of the

^,='1
TT

2 i; r
|u,

re-('+^")*J(Mr)[0(;8) cosi9
m=0 Jo Jo
X sin a/3 sin az da d^,

the summation with regard to

being over the positive roots of

/n

JnifJ^O-)

If

the temperature at the surface


v=Xi('''' ^)<

and

= X2(^'

^))

v =f(r, 6,

if

z),

+ V.(;8) sinn0]

is

=0-

given by

when 2=0,
when r=a,

when t=0,

up the problem into two, the one being a case


and the other a case of variable temperature. In
the steady temperature problem we have

we proceed

as in

9 to break

of steady temperature,

u=Xi{r,
ii,

= Xii9>

6),
^)'

when 2=0,
when r=a.

Then we put M = Mi + tta, and choose

Mj, u^ to satisfy

V^Mi =0,
Ui=Xi{r,

and

d),

when=0,

V%2=0,
U2=0, whenz=0,
Ui=Xi{z, d)-Ui, when r=a.

FLOW OF HEAT

124

A CIRCULAR CYLINDER

IN

Thus we have*

and

%=-

where

sin Az ."...

d\

cos m^

[4> (a)

R = J(r^+r'^ -^rf cos{6 - d"))

and

+ i/'(a) sin nd] sin Aa (ia,


and ^ (a) are the coefficients in

<^(2)

the Fourier's Series


CO

S ((f cos Ji9

+ 6 sin S

11=0

for

'Xal^.

variable temperature problem

The

former part of this

article,

^)-iis

the same as that discussed in the

the initial temperature being

fir,d,z)-u.
Surface at Zero Temperat

60. Finite Cylinder.

perature

tem-

f(r, 6, z).

The equations

for the temperature are

{0<r<a,-l<z<l),

_J=V2,

(1)

v=f(r,d,z), when t=0,

v=0, when r=a and

and

(2)

z=l

(3)

The expression
--c(x'+^V
e

satisfies (1)

and

*'

(3), if

Now expand /(r,

''

-COS

J
sin
Jn{fxr)-^nd

m is any integer and

6, z)

mir

^{z +
.

,.

//

is

l)

a root of

in the Fourier's Series

CO

(n cos

nd + b^ sin nd),

and 6 being functions of r and z, denoted by Fn{r, z) and Gnir, z).


Then expand Fn{r, z) and (?(, 2) in the series of Bessel's functions
given by the positive roots of i/(^a)=0, and take the Sine Series,

a,j

whose terms are the sines

of multiples of

^ (z+l),

for the coefficients

in this series.

In this way we find the solution in the form

v=^

S S e""^"

^''

^^

JJiuLr)sin'^ {z+l)iA^,mnCosn9
+S^,^.sinw9),

Heine, Handbuch der KugelfimcUonen

(2.

Aufl.),

Bd.

II.,

57.

FLOW OF HEAT IN A CIRCULAR CYLINDER

125

where

sin'^^(z+Z)&r

2J"-/(M-)<^rj'

and a

similar expression holds for

the temperature at the surface

V If

^ =Xi(^>

^))

is

cos n9f(r,

6,

z)dd

B^^m,n*
given by

when r =a,

^ = X2(''>^)i whenic!=Z,

when z=

'=X3{'. 6),

we

-I,

have, as before, to consider the steady temperature problem

where

'*=Xi(^'^))
'*

and

= Xa('''

0)'

'M=X3{''' ^)>

This

may

whenr=a,
when z =
l,

when z=

-Z.

be solved by taking

satisfies V ^Mj =0, has the given value at r =a, and


with similar conditions for the others.

where Mj
In

this

way we

sin-5j-(z

+;)(^ cosTOg +-B,sin?tg),

where Am,n and 5, are determined by the expansion of Xi(^>


Also

M,
"

=2
jn

I ^"^^
o t^^
smn//x(

.4^,,

''. (i"*")

(^^'. "

cs '^^

+ S''.

2)

as above.

'^)'

n=o

where the summation in

and

= Z,

obtain

2 E

Mi=

zero at 2

is

/* is

over the positive roots of

B^, are determined by the expansion of

Xal*". 6)-

Similarly

%=2 I
where

4^1, ,

?i5j4^^J(/*-)(4M,<'osw6l+Bp.,sinjie),

5n, are determined by the expansion of

61. Finite Cylinder.

XsC*". S).

Radiation.

in the cylinder of 60, radiation takes place at all the surfaces


into a medium at zero temperature, and there is an arbitrary initial
If,

* Of. Heine,

loc. cit.,

Bd.

II.,

81

when 71=

0,

the expression

is

to be halved.

FLOW OF HEAT IN A CIRCULAR CYLINDER

126

temperature v =f(r,
temperature

we have the

6, z),

following equations for the

_
Jt='^'''
dv

(1)

when<=0,

v=f{r,9,z),

^+hv=0,

whenr=a,

(3)

^-\-hv=0,

when z=Z,

(4)

dv

when2=

dv

^-\-hv=0,

and

Put

v=u+w,

and

where

u,

u=Uf(r,

e,

w=^{f{r,

6,

satisfy (1), (3), (4),

sin

satisfies (1)

and

and

(5)

(5),

when t=0,

-z)},

when t=0.

6,

and

expression

Xz^^^nd JAfxr) e- (^+'^') '.

sm

(3), if

"

it satisfies (4)

and

is

/*

//j;(/xa)

Further,

z)-f(r, 6-z)},
z)+f{r,

To determine u we take the

This

(2)

'

a root of

+ W(/xa) =

(5), if

is

a root of

X cosAZ+AsinXZ=0
If

we then expand the

(6)

(7)

function

\{^{r, e, z)-f{r, 6,

-2)}

in the series,
CO

S S n=0
S (A.;a,nCOsn9+Bx,^,nSinni9)sinX2:<7"(M^),
X

is

given

ft

by the equation
CO

"=S 2 S

(^,^,cosw0+B^,^,sinn9)sinX2J(Mr)e-''<'^"+'''",

the summations with regard to


roots of (6)

and

To determine

fj.

and X being taken over the positive

(7) respectively.

w,

we take the
cos X2

expression

''"^^ J(M>-)e-(^"+'''>',

sm

FLOW OF HEAT IN A CIRCULAR CYLINDER


and the

127

X's are the positive roots of

h cos XZ X sin Xl=0.

The function
is

then expanded in a

given

z)+f(r,

{f{r, 6,

-z)}

6,

whose terms are

series

of this type,

and

w is

by the equation *
00

tL

n=0

More General Problems on the

63.

The methods

Cylinder.

may

of the preceding sections

be used in dealing

with the hollow cyHnder, or the solid in which the bounding surface

formed by the cylinder (or hollow cyhnder), two planes through


the axis, and one or two planes perpendicular to the axis.
It will be sufficient to give here the solutions of the following
problems of this kind
is

I.

Hollow Cylinder.

Infinite

Initial

Surfaces

r=a and rb

kept at zero.

Temperature f{r).

In this case we have


dv

/d^v

ldv\

,,.

J = \W^+-rFr)

If

we put

^^^

= e~''-''^u, where u depends on

r only,

the equation for

is

Idu

dhi

or Bessel's equation of order zero.

As the range

of r does not extend to the origin, Bessel's functions

of the second kind are not excluded.

function r

Appendix

(cf .

F, where

I., 2), it is

Zrw=J+*r

(cf.

Instead of introducing the


better to take the function

Appendix

I.,

4),

since

H^^^z)

vanishes at infinity in the upper part of the z-plane.

U,(ar)=J,{ar)H^Hah)-J,{ah)H<'\ar),

Let

where H^^^r)

is

Then JJo(ah)=0 and Uo(aa)


^* ^

(2)

this solution of Bessel's equation of order zero.


is

also zero, provided that a is a

Jo{aa)H,^Hab)-J,(ab)H,^^^{aa)

This equation has no imaginary or repeated roots, and


'Cf. Heine, he.

oil.,

Bd. IL,

83.

(3)
it

has

FLOW OF HEAT IN A CIRCULAR CYLINDER

128

an

infinite

number

To every

positive root a there

\''rU,{ar)U,{l3r)dr=^0,

(4)

of real roots.

a*

corresponds a negative root


Further,

where

a, /8

we

have, as in 56>

two

are

And

different positive roots of (3).

l/^o^(-)^-ifrf")X

But

*^X^r-^{H.'iab)j^^JM)-JM)j^^Hw}
'

TT

J,{z) |ff<!'(2)-^<"(.)

since

*The equation f7o(ao)=0

The
p.

is

the same as

linown

real roots of this equation are

242

I Jo(^)=|^,-t

Gray and Mathews,

[ot.

loc. cit.,

(vi)].
I

To show

we

that there are no pure imaginary roots,

that

,rr(z)=2J(z)(logji/2+7)

Appendix

see from

I.,

2,

+4

+ (^/2)2-^(z/2)^+
(2!)'

...

^^Y^ix)

Thus

y=

'Joiix)
is real,

when x

is

Ai=

real.

i=7?fc) {^(^)
[of.

Therefore y

is

^'(^) -^o'(^) r.(.)}^^

Watson,

loc. cit.,

a continually increasing function of

It follows that

1./

Jo (to)

-TV-/

3. 63

(1)].

x.

=0

/o(')

cannot have any pure imaginary root.


Again,

when

a,

Thus

we know

j3

that

are different roots of i7o(ao)=0.

U^{aa) =

cannot

36, p. 78).
The equation U(aa)

have

an

imaginary

root

of

(of.

fUf. Watson,

=Q

loc. cit.,

can be treated in the same way.

%S. 63

(1).

the

form

XsV

FLOW OF HEAT

IN A CIRCULAR CYLINDER

129

Also

Jo(aa)_ go'"() _

But
foUows that

It

fr
\

Therefore

we have

Assuming

that/(>')

= -
^)
ar /r=a

pTI"

can be expanded in the

series

AiUo{air)+A2Uo{a2r)+...

and that the


(4) and (5),

series

can be integrated term by term, we have, from

[rf(r)U,(a^T)dr
A

ia
Ja

Thus we are

led to the solution of our problem in the

the summation being taken over the positive roots of


II.

0=00

Infinite

Cylinder.

kept at zero.

The surface r=a and

form

(3).

the planes

6=0,

Initial temperature f(r, 6).

In this case we have

dv_

/d^v

dv

1 d^v\

dt\~''\di^'^rdr'^?dd'y'

and
is

Jmir (ar) sin

" -KaH
-3- Qe-

a particalar integral of this equation.

Also the conditions at r=a,


that

0=0 and 6=60 ^'^^ satisfied,

m is a positive integer and a is a root of


^m:T(aa)=0.

C.CH.

provided

FLOW OF HEAT IN A CIRCULAR CYLINDER

130

Expand/(r,

the Sine Series

6) in

Va,sin-^-t',
Oo

the coefficient a^ being a function of

Then expand F^i^)

say F^{r).

r,

by

in the series of Bessel's functions given

positive roots of

the

J,^ (aa) =0.


So

In this

way we

are brought to the solution of our problem in the

form
"o

m=l

where
2
Aa,

/(>, 9) sin

4/

Ml IT

'^

/(' ^)

^u 2

-^ dJm;{ar)r dr dd
^

rJUar)dr^'>-'

eo\

= ^2fl

wiTT

f" r*"

w.=fi

-^ OJmAar) r dr dd,

sill

the summation in a being over the positive roots of

J^aa)=0.
0

The

solution for the

deduced from the above by letting

the surface

r=a

a^

oo

The ends z=l,

III. Finite Cylinder.

and

9=0 and 6=0o

wedge given by

Cf

69, 90.

the planes

6=0, 6=6o,

Initial temperature/ {r, 6,

kept at zero.

can be

z).

In this case we. have

dv_ /3^

dt~''\dF^'^r

and a particular

dv

1 d^v

integral of this equation


*'

''

d'^v\

JrW' W^MJ'

JnAXr) sm-s-d

is

Sin

-^{z+l).

da

Also the conditions at the surface are satisfied


integers,

and A

Expand/(r,

is

a root of

6, z) in

J^(Xa)=0.

the Sine Series

2^asm-3-e,
1

a being a function of r

and

z,

C'O

say ^(r,

z).

if

m, n are

positive

FLOW OF HEAT IN A CIRCULAR CYLINDER


Then expand
of

The

2j(z+0-

say

Fn{r,

z) in

131

the Sine Series whose terms are multiples

be functions of

coefficients of this series will

r,

J'.H.nCr).

Finally expand

by the

Ji, (r) in

positive roots of

the series of Bessel's functions given

j^ (xa) =o.
h

In this way we are led to the solution of our problem in the form

where

^x,

m,

nOo

a?/)

,t

/^

wi

wn rjm

C^

f{r, 9, z) sin

Vinr

-^ (z+l) Bm.^-dJn,r{\'r)rdrdB dz.

63. Determination of the Conductivity

from

Cylinders.

The results of the last sections can be reduced to a simpler form


when the initial temperature is constant. We proceed to examine
three cases which lend themselves to experiihental investigation.
I.

Initial temperature

v=Vo.

Radiation at r=a,

z=l

into

mediitm, at zero.*

In this case we take the expression


cos

which

is

'\zJo(iuir)e-''^>'^+i^'^^,

a particular integral of the equation

=kW
ot
This

satisfies

the surface conditions,

equation

h cos

and

IX

if

is

a positive root of the

X sin ZX =0

fxisa. positive root of iJiJo'{jua)+}iJo(iUia)=0.

Now, assuming

as before the possibihty of the expansions.

1=^1 cos Xi2:+^2 cos X2Z+...

l=BiJoW)+BMf,^r) + ...

we may obtain the values

of these coefficients

by

integration.

* Weber, H. P., Ann. Physik, Leipzig (N.F.), 10, p. Ill, 1880.

FLOW OF HEAT IN A CIRCULAR CYLINDER

132

To determine

Ai, A^,

and

We have

we

...

show that

fiist

cosXzcosA2(i!z=0

cos''

cos \^z cos \z dz

A 2 dz

(m^n),
"

{cos{X^+\)z+'cos(\^~\)z)dz
|-J

2(Xm X)

2{Xm+^)

cosXZcosXZ
(^m^

= 0,
Also

^n

> 7
%
^
(XmtanXj-XtanX0
7.

J.

XtanX?=A.

since

(l+cos2Xz)dz

cos2Xz<?z=|l

Therefore, on multiplying

by cos Xz and

integrating,

4sinXn?

"~2Xj+sin2Xr
To determine

B^, B^,

...

we have

Ctt

("a

^n\ fJi{nr?')dr=\ rJa{fir)dr.


Jo

But

we have

Also

^
d^

since

Jo (m')

;:

"^o (m')

y"^<^o

(/.)

seen ( 56) that

Jo

Therefore

+ ^

=0,

rJo{fXnr)dr=-^ ^^[r^Jo{fij-)jdr

^2j

we have

'o

'i/^n

B"

2A
a(A2+//^)Jo(/a)'

we have

FLOW OF HEAT IN A CIRCULAR CYLINDER


Thus we have the

133

solution in the form

v=Vo{Aj^cos\iZe-''^''^+...){BjJo{fXir)e-'""'t+...),
\i, X2>

and

fix,

being the positive roots of the equations

n^,

hcoslX\sm.lX=0
Further,
time,

if

and

readings after a considerable

we have approximately
v=v^-iB-i^ cos

Two

and

/xJo'(m*)+^o(/"<'^)=0.

we take temperature

observations

may

v^, v^,

at the

same point at times

t^, t^,

give

be determined, h being supposed known.

r=a

Surface

Radiation at z.= l into a medium

v=Vq.

Initial temperature

II.

at zero.

XizJo(/"i')e"''^'^i''*''''''^'.

kept at zero.*

In this case we have, as before,


v=Va{AiCosXiZe-'''"'t+...){BjJo{fXjr)e-''i''^+...),

where

Ai,

A 2,

...

are the positive roots of

h cos

and

are

fxi, yua,

l\

X sin IX =0,

the positive roots of Jo{fjia)=0.


.

Also

4sinAZ

2XZ+sin2Xj

5= -

and

2
.

,,

,-

Approximating as before,

v=VoAiBi cos XizJo{iuiir)e-''<>'^'+i'^'>',


and from two observations at the same point we obtain the value
of K, h being supposed known.
III. Initial temperature

perature.

In

v=Vq.

Radiation at z=-{-l and

this case

we

z=l

Surface

r=a

into

kept at zero tem-

a medium

at zero.'\

start with the particular integral


sinA(z+2)Jo(M')e-''(''+'''>'.

At

Wiss., p. 472, 1880.

Weber, H.

Beglinger, Berlin, Verh. Ver. Oewerbfi., 45, 1896.

Ithaca, N.

F., Berlin, SitzBer.

Y., 10, p. 297, 1900.

See also Hall, Physic. Rev.,

134

The

FLOW OF HEAT

IN

A CIRCULAR CYLINDER

surface conditions are satisfied

equations

if

X and

^u

are given

by the

x cos 'm+h sin 2X1=0

and

nJ^{fxa)-{-hJ^{ij.a)=Q.

Thus we have the solution

in the

form

=Vo(^isinXi(2!+0e-''V4-...)(5iJo(Mi'-)e-'"'''"+.-.).

where

A,,=

4(l-cos2Xj)
4XZ sin4Xr

B=.

2h

'a(h^+/uLr,^)Jo(lu.a)'

andXi; Xa, ... mi> /2> are the positive roots of the above equations.
Approximating as before,
,

CHAPTER

VIII

THE FLOW OF HEAT IN A SPHERE AND CONE


64. Introductory.

We

have seen in

6 that the equation of conduction,

when

ex-

pressed in spherical polar coordinates, becomes

dv_ fdH
In the case

of

3t;

d (

Flow

of

Heat

dv\

in the Sphere,

d^v\

when the

initial

and

surface conditions are such that the isothermal surfaces are con-

and the temperature thus depends only upon the


and t, this equation becomes

centric spheres,

coordinates r

On

putting u=vr,

we have

du_

d^u

Steady Temperature.
If
fa,

the solid

is

a hollow sphere, of inner radius

we have

r^

and outer radius

u=m,

where

d^^^'
u=v^r-i, at

and

r=ri,

u=Vir^ at r=r2,
the temperatures of the inner and outer surfaces being v^ and
Therefore

-=

^ ^

Y^
rir^-rj)
,

135

v^.

136

FLOW OF HEAT

A SPHERE AND CONE

IN

Variable Temperature.

Let the sphere be of radius


given by v=f{r).
o,

the equations for

initial

temperature be

kept at a constant temperature

is

are as follows

(0<r<a)

a^='^a^'

and

and the

a,

the surface

If

M=0,

when r=0,

u=av^,

when r=a,

u=rf{r),

when <=0.

These are the same as the equations we obtaiaed for the case

of

whose ends are kept at temperatures zero and av^,


its initial temperature being rf{r).
The problem of the symmetrical
distribution of heat in a sphere of radius a is thus mathematically
the same as that of the flow of heat in a rod of length a.
The case of a hoUow sphere can be discussed in the same way.
a rod of length

65. Sphere.

a,

Radiation at the Surface

Zero Temperature.

Initial

Temperature

r=a

into

a Medium

at

f (i).

In this case the equations for v are as follows

dv

w
dv

and
Puttirig

\-hv

= 0,

when r=a,

(2)

when t=0

v=f{r),

(3)

u=vr, we have

Tt^^d^'

(^<^<^)

di

g^ +

u=0, when r=0,

(5)

('i--)tt=0, when r=a,

(6)

and

u=rf{r), when

t=0

(7)

The problem is thus reduced to that of the flow of heat in a rod,


one end being kept at zero temperature, while at the other end
radiation takes place into a

Proceeding as in

36,

we

medium

at zero.

consider the expression


g-KoH gin

Qj._

FLOW OF HEAT
This
if

and

satisfies (4)

(5)

IN A SPHERE

whatever a

be,

and

137

it satisfies (6)

a root of the equation

is

aacos aa-\-(ahl)

To

may

AND GONE

sin

aa=0

(8)

find the nature of the roots of this equation put

and we

see that they correspond to the abscissae of the

common

points of the curves

>

where 'p=dh\

The
f x)

1.

roots are thus symmetrically situated on the axis of

regard to the origin.


(0,

=tan^
and >?=-.
*
P

(tt,

<^ <; 00

Itt)
,

...

they

lie

When 1

< <
j?

0,

they

lie

and approach |(2w l)Tr as w


in the intervals (^tt,

approach |(2w 1) tt as n increases

FlO.

tt)

increases

(f tt, 27r)

But

J(^)=^+(aA-l) tanh^=0
F'(/ii)

when
,

and

roots.

8.

Further there can be no pure imaginary root of


where ju is real.

Also

...

and there are no repeated

(8).

ossible, let aa=ifjL,

Then

^ with

in the intervals

= 1 +{ah-l) sech Va^-l>-l.

and

F{0)=0.

For,, if

FLOW OF HEAT

138

IN A SPHERE

Therefore F'i/n) never vanishes, and

F(iul)

AND CONE

has no other

real

root than zero.

no imaginary roots of the form \ifx.


a=\+i/jL be such a root.
Then a' = \ifA must also be a root.

Also, there can be

For,

if

Then

possible, let

F=sin

since

F'

^+^^=0

satisfy

ehave

(a^-a'^)

But F, F'

satisfy (6),

ar,

= sin a'r,
and

^+a'^F'=0,

FFrf

dr

d/r

and therefore
ar

ar Jo

r Vrdr=0.

Thus

Jo

sin(X+i^)rsin(X i/x)r(Zr = 0,

Thus

and

this requires that

(sin^ \r cosh^ far

+ cos^ Xr sinh^ ^r

(ir

= 0,

Jo

which

is

impossible.

We

have thus shown that


are infinite in number.
Let

(8)

has only real roots, and that these

ai, aa,

...

denote the positive roots in ascending order.


CO

Then the

y]4e-''"-'^sinar

series

satisfies (4), (5),

and

(6),

and

'rf{r)

will satisfy (7)

if

= ^Aamar.*
1

Assuming the possibiUty of this expansion and that the series


may be integrated term by term, we can obtain the values of these
coefficients as in 36.
*

On

the possibility of this expansion, see Ford,

loc. cit, p.

144.

FLOW OF HEAT

IN A SPHERE

AND CONE

139

We first show that

fo sin a^ sin ar =
J

{m=f=n)

c?r

a^an^+{ahlf

Jo

These follow at once by integration.

sin(a_,

cos

and

a^a cos

sin(a + a)a

a_(X

^^ (a tan a^a a^ tan

a^ and a are roots

this vanishes, since

aa_

tan
a

Also

For

aja

of

a
1

s.m.^anrdr=^\

ah'

(1

cos2ar)c?r

_a
J
and

2tanaa

sin2aa

s
sin2act=:p-7-T
l+tan^amO!.

2anCi{ahl)

sm^ardr=^

Therefore
J^

"

^^.j+^ah-l^

Hence, assuming that


r/"(r)

aA

we have

=^1 sin air+A^ sin a2r+


sin^ ar

. .

dr=\ rf{r)smardr

rf{r)sm.ardr.
^n=g
2
!/
IN
"
a aW+aft(aft
l)Jo

and

I.

Then the

solution of our problem

is

00

M=^^e-''""'* sin

a^r,

which gives
^

ar,f^aW+fflA(aA-l)VJo

'^^

>

att)

FLOW OF HEAT

140

DT

A SPHERE AND CONE

This solution has been appKed to the problem of Terrestrial

Temperature, the

temperature being constant,

initial

and the

The solution of the problem of


SoUd with arbitrary initial distritemperature may also be deduced from the result of

radius of the sphere very great.


radiation in the Semi-Infinite

bution of

this article*
66. Application

to

Determination of the Conductivities

the

of

Poor Conductors.

The expression we have

just obtained for the temperature in

a sphere cooling by radiation at the surface converges so rapidly


that

when a

sufficient

calculation,

time has passed the terms after the

first

may

This gives an expression suitable for mmierical

be neglected.

and

has been apphed in different experiments where

it

the initial temperature of the sphere

is

constant.

For example, a ball of the material to be tested is immersed in


a bath at a constant temperature F for a sufficient time to allow
the whole baU to acquire the temperature of the bath. It is then
removed and allowed to cool by radiation in a medium at constant
temperature. After the cooling has gone on for a certain time,
observations of the temperature are taken. In one set of experiments
these readings are for the temperatures at the centre and the surface.
In another set of experiments the temperature at the centre alone is
required.

With the

notation, of 65,

A
=
Hence,

and

if

e"*"''''

sin aiT to our approximation.

a=the temperature

at

r=a at the

o=the temperature

at

r=0

time

at the time

t,

sin a,a

(1)

Also
Kai^

Thus

(!^^=i.=e'"'i'<'2-'i).

is

given by

(2)

and

oj

by

(1),

(2)

remembering that

< aoj <

tt.

K is obtained.

*Cf. Biemann,
Weber-Riemann,

Partielle. Differentialgleichungen,
loc. cit.,

Bd.

II., 55.

69-70, Braunschweig, 1869;

FLOW OF HEAT

Further, the original equation for


ffla

AND CONE

IN A SPHERE

141

namely

a,

cos aa-\-(ahl) sin

aa=0,

h*

gives the value of

Ayrton and Perry used the second method ia determining the


The temperature at the centre at time

Conductivity of Stone.f
t

is

approximately

g_^^2,_

r sin a^r dr
I

But

A,=v,f^

Therefore

rm.

Thus

^)

^gj;, /sin a,a-a,a cos a,a\


! \a^a WO. a^a COS a^aj

= n2'U sinaiffl
^ aM cos a-,a e-""'^
aiffl sin a^a cos a^a
=Ne-^\

~,

say.

The value of n is obtained by two observations of the temperature,


and n being known, the value of N may be found. Also a table of
the values of the expression

since a; cos
a; sin a; cos

35
a;

from the known value of N.


But TO=/cai^, and thus the conductivity

will give oi

Surface

67. Sphere.

p^aturef(r,

6,

r=a

is

determined.

at Zero Temperature.

cdH

.'2-dv

'

9 /

v=:f{r,e,<p),

v=0,
Put
write

v=e-'"^'^u,

ju

where

tt

is

3t^\

d^v \

when t=0,
when r=a

a function of

r,

(2)
(3)

6 and

<p

only,

(1),

Weber, R., Zurich, VieHdjahrsch. Natf.


Of. Phil.

Mag., London

(Ser. 5), 5, 1878.

= cos 0.

Then we have from

Tem-

(j,).

Ih this case the equations for v are as follows

dv_

Initial

Ges., 23, p. 209, 1878.

and

FLOW OF HEAT

142

Now

IN

A SPHERE AND CONE

the Zonal Harmonic P((u),

when n

is

a positive integer,

is

and

it

the coefficient of A" in the expansion of {l2/jLh-{-h^)~-,


satisfies

Legendre's equation,

m
Also

w={l-ij.YD^PM

satisfies

m'2

n*
w = 0,'

-M^

D for

where we have written

-j-

It follows that the expression

oLLX

will satisfy (4)

provided that R

dr^

dr

is

a function of r only, and

r^

72=(ar)"V+j(ar),

This leads us to

the solution J_(+j)(ar) being inadmissible, as


to infinity as

it

would make R^ tend

r>0.

We are thus brought to the following solution of

(1)

=e^(ar)-^/+i(ar)(l-M=)'^Z'P(/.) ""^^mcp,

(5)

m and n being positive integers.


The condition

at the surface

is satisfied

by

(5), if

is

a root of

J+,(aaHO
If,

as before,

we assume that/(r,

whose terms are

of the

6,

(j>)

(6)

can be expanded in a

series

form
"*

O Oft
oIXl

and that

this series

can be integrated term by term, we can find the

coefficients in the expansion.

*Byerly,

loc. cit., p.

196

Analysis (3rd Ed.), 15.

The second

(11),

Boston, 1893; Whittaker and Watson,

Modem

5, 1920.

solution of this equation has an infinity at

unsuitable for our problem.

ir,

and

is

thus

FLOW OF HEAT

IN A SPHERE

AND CONE

143

For, let
<

/(r, 0, ^)

=2
a.

CO

SS

(^)"V+i(ar)(l-M^)^2)'"P(M)

m =0 n=0

the summation in a being over the positive roots of

(6).

Then we have
I

f(r, 6,

(j>)

cos

Also

we know *

and

m<j) d(p

that

(l-;u2)i)P(Ai)-DP'(//)rf/x=0,

w^w'.

I'

Therefore

)_^(l-M^)^DP(/x)(^/x

f{r, 6,

<ji)

cos mi>d,p

_
Finally,
fa

from

56,

27r

we

obtain
r^'T

{l-^YB^P^{^) dfj.

f^Jn+i (ar) dr
JO

fl

{n+m)\

-1

f{r, 9,

cj>)

cos

m0 d<p

Jo
ira'a-i {n+m)\
-^..2n + l (w-m)P'^+^-^

In these results tt must be replaced by 27r when


Also Ba,m,n can be found in the same way.

Thus we are led to the


CO

^^

m = 0.

solution of our problem in the form

CO

M=0=0

{ ^a, m,

COS

nKj)

+ 5, , sin mcj)

},

the constants Aa,m,n and 5,m, being determined as above, the

summation

in a being over the positive roots of the equation

J+s(aa)=0.
* Cf Byerly, loc.
.

cit.

106

Whittaker and Watson,

loc. cit.

(3rd Ed. ), 15. 61.

FLOW OF HEAT

144

A SPHERE AND CONE

IN

been shown in 9 that, when the surface temperature is not zero,


may be reduced to a case of steady temperature, and a case of
variable temperature with zero surface temperature.
It has

the problem

When the surface of the sphere is kept at v = F{d, <^), the steady temperature,
or potential probjem,

given

is

by

'Vhi = n, through the sphere,

u = F{d,

4>),

at the surface,

and we have*

m=^|(2m + 1)Q"_^ sine'de'j^

^^Fid', <t>')P{cosy)dct>',

where the ordinary notation of Spherical Harmonics is employed.


The variable temperature problem will then be given by

w=f(r,

6, <i>)-u,

jo=0,

And

when < = 0,
when r=a.

the solution of the problem with which

we

started

is

v=v, + w.

The corresponding

may

questions,

when

radiation takes place at the surface,

be treated in the same way.t

68.

The Part of the Sphere r = a cut out by the Cone 6 = do.

Temperature Zero.

Initial

Temperature

f (r, 6,

Surface

<f>).

In this case the equations for v are as follows

where yu=cos0,

v=0, when r=a and when d=0Q,


v=fir,

d, <p),

Proceeding as in

(1)

'.

(2)

when<=0

67,

we

(3)

are led to the following solution of

(1)

=e'*()-* J+j(ar) P--(^)"'m^,

(4)

Bill

where

P,;;'"(^) is

equation

the generalized Legendre's function J given by the

* Cf.

Byerly,

loc. cit,

114.

t Of. Heine, he.

cit.,

Bd.

II.,

J The best treatment of the generalized Legendre's functions


a paper by Barnes, Q.J. MatK, London, 89, p. 97, 1908.
It is obvious that PZ'^.i (m) is the same as P"'"(/u).

84, 85.

will be found in

FLOW OF HEAT IN A SPHERE AND CONE

145

Also, in this case,

m is zero or a positive integer,


and

is

a root greater than

is

a positive root of

-J

of Pn'^duo) =0,

J^+i (aa)

= 0.

The surface conditions (2) are satisfied by the value of v in (4),


and there is no infinity within the solid.
If, as before, we assume that the ftmction/(r, 6, (j>) can be expanded
in a series whose terms are of the form
olU.

'

and that

can be integrated term by term, we can find the

this series

coefficients in the expansion.

For, let
CO

m=0 n

X {^, , cos m^f) + B, ,, n sin m^}


n being over the roots greater than -1

the summation in

p-"(^d)=0, and that

in

being over

the

positive

roots

of
of

J+j(aa)=0.
Then we have

an

Jo

Also
are

it is

two

known

that,*

when

different roots greater

m is any positive number and n, n'

than -

1 of

P^(^o) =0,

J Mo

fJPr(.)N.=--(^)
U=P-'^M

*Let

and

5^ P.--(.o)^P.-(.o)
'

= PW.

Then we have

|((l-.=)g)H-(n(.I)-j^.)=0,

Therefore

{n'-n){n'+n+l)

uu'd/j,

Jim,

=/:{-s(<'-'-'i)-"i("-'''f)}*
= n 1 _ ^2) j '
C.O.H.

^ _ M ^1 T
K

[Note contimxd, p. 146.

FLOW OF HEAT

146

A SPHERE AND CONE

IN

Thus we have
/.2ir

(1

P-'" (m) dfi

f(r, d,

cos

<p)

m^ d^

Jo

Ji^

from

Finally,

56,

we obtain

ri

fa r^J+j(ar)*
"

("2:7

A,,

In these results

f{r, e,<p) cos m<}>d<l>

Pr"(M)^M
-"o

^^"

2(2n+l)

^ must

be replaced by

Further Ba,m,n follows in the

27r,

when

^''<'^^'^+j('*)^

m = 0.

same way.

Thus we are brought to the solution

^=

"

^'"'M^o

of our

problem in the form

S m=0
S i;e-'''''(ar)-*J+s(ar)P-(M)
n

X (4,,,cosm9!)+5a,m,sinm0),
the constants Aa,m,n> Ba,m,n being determined as above.
If

the solid consists of the part of the sphere

and the planes <^=0,

(j>=tj)a,

we expand/{r,

the series

o.

on the same

0, (p) in

r=a

cut out by the cone d

m=\ n

9o

lines as above.

It follows that,

when

n, n'

are

two

different roots of P^'"(mi))=0 greater

than -i,
'(M)P:,"'(M)d;" = 0.

d
Also

J^(f,T'"(.))

'^>^=-2iC^]t
_

when

=8 at

the surfaces being kept at zero temperature,

/'7'"(Mo)

= 0-

(l-Mo)

>Ap-m,

FLOW OF HEAT

IN A SPHERE

The Cone.
The result of 68, when the
and the cone 6 = 6o> ^^'^ ^^ W^

AND CONE

147

69.

r'*J+j(a/)dr'

Jo

solid is

bounded by the sphere

^s follows

=a

/(/,0',9f>')cosm(<^-f)c?</,', (1)

(P.r-'U'))''^/

Jo

Jmj

the summation in wbeingovertherootsgreaterthan 4 of P^(/,(o)=0,


that in a over the positive roots of J+j(aa)=0, and

But when

or a positive integer.

being zero

=0, the result has to be divided

by 2.
The temperature problem for the cone 6 = Q^, when the surface is
kept at zero, and the initial temperature is/(r, 6, </>), can be deduced
from (1) by letting a^^oo
.

Now it is known*
J+j

(affl)

that,

/(

when a
j

is

cos [aa

very large,

{n -\- 1) ~j approximately,

and thus
J'n+i (aa)

If a,

= A/f

+ Aa

(aa{n+ 1) ^)

sin
J

approximately.

when a

are consecutive roots of J+i(aa)=0,

is

very large, we have the approximations

aAa=7r,
3/-.,

Using these

results,

2a

.\2

"j

and remembering that the summation

definite integral, the solution of our

will become a
problem for the cone

is

as follows

^ V-

in a

temperature

(2w+i)p-'"(//)

.,j

..

1 2\s=0
T>-ml..
T>-ml.. \-'0
..*(l-,0V=0|^p-.(^^)^p-.(^^).
<*

1/

f"

"'

fZTr

fl

Jn
* Cf.

'

Watson,

f{r\8',<l>')cosm{cp-<p')d,^',

(P-0'''))^c?/

/*J+i(ar')c?/

Jo

Jmo

loc. cit.,

7. 21

(i)

Gray and Mathews, he.

cit.,

p. 40.

FLOW OF HEAT

48
le

summation

in

lacing

IN A SPHERE

AND CONE

taken over the roots greater than

| of

'"{/"o)=0.

m=0,

In the term corresponding to


splaced

by

as noted above,

be

27r.

If the solid is the part of the

cone 6 = 0o cut

= <^

68 applies.

v must

the remark at the end of

The general case when the surface

m be treated in the same way.

is

The

off

by the planes

<^

and

given by

results are

more complicated

as Bessel's

mctions of both kinds, and the generalized Legendre's functions of both


nds,

must now be taken

into account.

CHAPTER IX
THE USE OF SOUECES AND SINKS IN CASES OF
VAEIABLE TEMPERATURE*
70. Instantaneous Point Source.

Suppose that a sphere of radius a at temperature Y is placed at


in an infinite solid of the same material at zero temperature

i=0
and

If
t

is

left

to cool.

M=w, we know

from

64 that the temperature v at the time

given by the equations

3w
dt

The

solution

d'^u

"dr^'

when t=0,

u=0,

when f=0, >,

M = 0,

when r=0.

is

thus

known

( 18) to

(r+/)'

4/ct

)dr'
^ y

'^\
Jo

2rV(7rK)

Expanding the integrand


3

Va-

_rr'

r'i

in powers of

we obtain the approximate

=-

be

fr-r')'

v=

small,

0<r:

u=Vr,

r',

and assuming that a

is

solution

_r.

^)4)-

67r*/c

LetQ=i-7raW.
This method is due to Kelvin. Of. "Compendium of the Fourier Matjiematios for the Conduction of Heat in Solids," Mathematical and Physical Papers,
Vol. II., p. 41. See also, Hobson, London, Proc. Math. Soc, 19, p. 279, 1889 ;
Rayleigh, Phil. Mag., London (Ser.

6), 22, p.

149

381, 1911.

THE USE OF SOURCES AND SINKS

150

Then

this result

Now

can be written

the radius of the sphere tend to zero, Q remaining


and we are led to the following solution of the equation of

let

constant,

conduction

-Tl

'"''

(^)

"=^(2j(7rKt)r^

where r^=x^-{-y^+z^.

As

i-i-O this value of v tends to zero

origin,

where

everywhere except at the

if we integrate
we obtain Q.
temperature given by (1) is said

becomes

it

Also

infinite.

the infinite solid at any time

(t

The distribution of
an Instantaneous Point Source of Strength Q at the
an Instantaneous Point Source of Strength Q placed
(x', y', z'), we have in the same way
to

to be due

origin.

For

at the point

(x-xT+(y-yy+(.z-z'r

v=

v through

> 0)

{2J{7rKt)f

This

is

the fundamental solution of the equation of conduction in

three dimensions.
is

The quantity

of

heat concentrated at the source

Qpc.

We

might have started with a cube whose edges are of length h, instead
If this cube is placed in the infinite solid as above
and left to cool, we would have ( 16)
of the sphere of radius a.

rV'

fi"

(a!-') '+(i/-i/')'+( a-s')'

the centre of the cube being at the origin

and

its

edges parallel to the axes.

This leads to the approximate solution,

where Q = h?V.
Also letting ft->0

we have

as above.

For
from

Q
to

write
t.

<j>{t')dt'

in (1),

change

into [tt')

and integrate

IN CASES OP VARIABLE

TEMPERATURE

151

Then, we have another solution of the equation of conduction;


namely,

v=

,;

,^

.,3

cl>{t')^irdt'.

This distribution of temperature


Point Source of Strength
If

(/)

said to be due to a Continuous

is

from t=0 onwards.


q, we have

(j){t)

constant and equal to

(t) is

ng

4(i-j')

5-ai

=V
Let t^-oo and

4''

on putting

rfr,

=^t.

this reduces to v=q/4:7rKr, a steady

distribution where a constant supply of heat

continually intro-

is

duced at the origin and spreads outwards in the

temperature

infinite solid.

71. Spherical Surface Source.

Again

let

placed at

a spherical shell a<Cr<ia-\-h at temperature

t=0

in

temperature and

an

at the time

y
xi=

ra+h
r

will

('+''') '\

let h-^0,

*"'

)d,r'.

remaining constant.

obtain another solution of the equation of conduction,

v=

^^(e
STrariTTKty^

is

**

Put Q=^Tr((a+hf -a^)V, and

This

be given by

(> '')^

f'\e

2r{TrKtYJ"-^

en we

be

left to cool.

The temperature

'"^

same material at zero

infinite solid of the

*"'

-e

""

).

'

the temperature due to an Instantaneous Spherical Surface

Source of Strength Q.*


The temperature due to a Constant Spherical Surface Source of
Strength q will be obtained by writing q dt' for Q, changing t into
{t

t'),

and integrating from

to

t.

* This solution can also be obtained by distributing the Instantaneous Point


Sources of 70 uniformly over the surface of the sphere and evaluating the surface
integral.

THE USB OF SOURCES AND SINKS

152

Then we have

T= n

where

thj

m2=(r a)^

/.

//

For the steady temperature problem,

when r>a, and

v = g/4irKa,

when

let i->oo,

and

n^=('+a)2.

and we obtain

v=q/4:irKr,

?<.

72. Instantaneous Line Source.

Let the whole of the xy plane be

initially at zero

temperature

except a square with centre at the origin and edges of length h

temperature of this square being

parallel to the axes, the initial

constant and equal to F.


This two-dimensional problem has for

nh

its

solution (47)

(a:-j;')'+(y-,v')'

fJA

dx'dy'.

TT/Cf J-i/iJ-i/i,
i-TTKti

Now

let h->0,

while FA^ remains constant and equal to Q.

Then, in the hmit.

This

is

the temperature due to an Instantaneous Line Somce

coinciding with the axis of z of strength Q.

When

cides with the parallel to the axis of z through

v=^e
This

and

is

it is

Strength

the source coin-

(x', y', 0),

we have
(1)

the fundamental solution for two-dimensional problems,


usually referred to as due to an Instantaneous Source of

at

(a;',

y').

It will be noticed that the quantity of heat

along the source per unit length

Using polar coordinates,

(1)
r\

is

may

Qpc.

be written

r'+)-" 2rr'coB(9 y)

IN CASES OF VARIABLE
and

this reduces, with the aid of a

TEMPERATURE

153

well-known integral in Bessel's

functions,* to

^=^6Vo(aB)rfa,
where

(2)

R^=r^+r'^-2rr' cos{d-d).

The

solution for the Instantaneous Cylindrical Surface Source of


Strength Q corresponding to the Spherical Surface Sowce of 71 can

be

by

obtained

Line Sources uniformly over the

distributing

r=a.

cylinder

In this way the temperature at the time


from the axis is given by
*

dd'

e""*,

Sir^Kt"

at a point distant r

Jo
riar~\

^ -e'^^J
47rrf

''L2;rfJ-

73. Instantaneous Plane Source.

Again

let the infinite solid,

^= i^,

be

at

initially

initially at the

except the

zero

between the planes

slice

temperature,

portion

this

being

constant temperature F.

Then we have

( 16)

dx'.
'(7rKi)J_j,i'
2V(7

Now

let

A->0, while

Then, in the
This
Strength

is

Vh remains

limit,

ff
Z^yrrKt)

over the plane

x=Q.

If

the source coincides with the

-<^=^
ixt

The quantity

the fundamental solution for Unear flow.

heat per unit area concentrated on the plane

For the case


this solution as

The diagrams
of

(1)

we have

Q
is

e~'^

the temperature due to an Instantaneous Plane Source of

plane x=x',

This

constant and equal to Q.

is

of flow of heat along a rod it

is

usual to refer to

due to a Point Sowce at the point


in Kgs.

9,

of

Qpc.

x'.

10 illustrate graphically the distribution

temperature due to this Instantaneous Point Source.


*Cf. Gray and Mathews,

loc. cU., p.

77 (158),

The dotted

154

THE USE OF SOURCES AND SINKS

IN CASES OF VARIABLE

TEMPERATURE

VALUES OF

'

ITia.

W.

155

10.

The temperature at the point (x, y, 0) due to a uniform distribution of heat


on the plane a; =0 at the time t =6 could also be obtained by using the polar
element of area in that plane.
In this way we have

-^

./
rJa
-(2V(Ti))^.

a;''+i/'+p'-2j/pcos9

rii,

a^'+r

Thus

:j

4rt

pdpdd.

Jo
ry>

!>&'

iKt

dp.

(2)

THE USE OF SOURCES AND SINKS

156
It follows

from

and

(1)

(2)

that

one of Weber's integrals in Bessel's functions, and a physical interby considering a more general
distribution of heat along the plane y =0.*
This

is

pretation of the other integrals can be obtained

74. Doublets.

We

have seen that


(x~xy+(v-i/)'+(.^-'/)'

Q
is

4Kt

a solution of the equation of conduction

dv_
It

is

rd^v

d^v

d^v

oUows that
_ ix-x')'Mv-yyH^-^r-

dv

Q{x-x')

dx'

UmjinKt))^

also a solution.

This can be obtained by combining with a source of strength Q'


at (x'+dx',
dx'^>0,

y', z')

a sinkf of strength

For the temperature at


is

Q'

at

y,' z'), letting

{x,'

and putting Lt (Q'dx')=Q.


{x,

y,

z)

due to the source and sink

given by
r

ry

___

^x-x'y+{y-y)'+{^-z'y -

(x-x'-da:')'+(y-y)'+iz-zy

v=,

iKt

Thus
Q'

_ (3;-a;')'+(t/-i/')'+(2-z')'

2{x-x')dx' -{dx.V

V
Q'gr^

(a;-a;')'+(i/-i/T+(g-0'

Proceeding to the limit,

we obtain

'0= ^^

where

When

.'.

V)

**,

'R^=-{x-x'Y+{y-y'Y-\-{z-z'f.

* Of. Bayleigh, loc.

+ higher powers of dii.

''^

-:{x'3i)e

cit.,

p.

384

Gray and Mathews,

the strength of a source as defined in

70

is

loc. cit.,

p.

negative,

78 (160).

it ia

called a sink:

IN CASES OF VARIABLE

TEMPERATURE

157

Tlie distribution of temperature in (1) is said to be

due to an

instantaneous point doublet of strength

at (x', y',

z'),

with

its

axis

parallel to the axis of x.

Corresponding definitions apply to the doublet in linear flow


point

x',

and the

at the

whose axis is parallel to the

line doublet at (x', y')

axis of X in two-dimensional problems.

The temperature at the time t due


will

to the

two doublets

just

named

be

-^^=^

.Qix-x')

Q(x-x')
v=^
2,T^e

and

- (^-fr+l-yr

The extension to the continuous


strength

is

obvious.

tinuous doublet of strength


flow, is given

doublet of variable or constant

For example, the temperature due to a concp {t)

at the point

and

in the case of lineai?

by

47rVJo^'
Substitute

x',

{t-t'Y

xx' = 1ic{tt'ya, and we have

~7^}

v=r-\

^~-'4>(t

'o=T-\

e-'-'^fi

--j

V, when

-)(?a,

a;>a;',

when a!<

k'.

2V(0

Thus
and

.,,o=Mre-^c?a =

%'-o= ^5

>

0,
Thus, in the case of linear flow in the semi-infinite solid a;
when
placing
temperature
by
at
kept

0,
can be
the plane cc
^ ()

>

a continuous doublet of strength

2/(:^(<)

at a;=0.

[Cf. 23.]

In two-dimensional problems the boundary 2/=0 can be kept at


temperature f(x, t) by placing a continuous doublet at (a;', 0) with its
axis parallel to the y-zxss, and strength 2/(r/(a;', t) dx', and integrating
along thejaxis of x. A corresponding result holds for the three-

THE USE OF SOURCES AND SINKS

158

dimensional case
f(y'.

when the plane c=0

to be kept at temperature

is

z', t).*

The Method of Images.


The method of images, which plays

75.

Mathematical Theory of Electricity,

so important a part in the


peculiarly adapted to the

is

solution of problems in Conduction of Heat,

boimded by

when

the solid

and these are kept at zero temperature.

planes,

imagine the solid to be continued in

all

is

We

directions without limit,

and we then obtain, by a suitable distribution of sources and sinks,


a temperature function vanishing on the boundaries, with the
required singularities (sources, sinks, etc.) in the solid. The distribution of sources and sinks outside the solid is in this case determined by taking images of the original distribution in the solid.
We shall see in next chapter that the temperature due to a single
source in the given solid, when its boundary is kept at zero, is of
considerable importance in the solution of the general problem
of conduction for that soUd.

We

proceed to discuss different types of problems whose solution

can be obtained by the use of sources and sinks in this way.


Linear Flow.
I.

Semi-Infinite

Solid

a;

>

Temperature at the Boundary x

0.

Initial

Consider the source of strength

may

take the

initial

Temperature f{x).

= 0.

Zerio-

_^
/(a;')

tZa;'

at the plane

x'.

We

temperature as due to a distribution of these

sources along the positive axis of x.

With the source f(x')dx' at x', we associate the


two give zero temperature at x=0.

sink f(x')dx'

at -x', as these

Hence
II.

rA

1^=25-77

fix')\e

*"

Finite solid bounded by the planes

Temperature f{x).

Bounding Planes kept

*"

}dx'.

a;=0 and x=a.

Initial

at zero.

x', we have to take the images


and x=a, a source and a sink

Starting with the sovace f(x')dx' at


of this source in the planes

x=0

alternating so that the boundaries

may

way we have

x'+2na and

x'-\-2na,

sources at the points

where n

is

zero or

be kept at zero.

any positive or negative

* This statement can be verified

from the

In

this

sinks at the points


integer.

results of 86, 87.

IN CASES OF VARIABLE
Thus we have

We

TEMPERATURE

159

finally

have already in 30 obtained another expression

this case,

->,sin

may

which

for v in

namely,

a;e

be written, when

/(wlsin

>

a;aa;,

0,

~rf{x'}y]smxsm'^x'e~''^'dx'
ajo

We

(2)

proceed to show that these solutions are identical.

may be proved by the properties


aid of the following theorem

of the Theta-functions * or

This

with the

an even function of x which can he expanded, as also


f(x2na), in a Fourier's Series of Cosines ofmultvphs of Trxja, then
If f{x)

is

2/(a'+2na) = -f f{x)dx+^'^co&'^x{ f(x')cos'^x'dx',


provided the integrals are convergent and the series converges.

Since

/()

=-

f{x')dx'+-y]cosx\ f{x')cosx'dx',
"'1

"^Jo

and

f{x+ 2na) = -

Jo

f /(a;' + 2na) dx'

<^Jo

+-ycos
a

aV

f(x'+2na)cos

a;

Jo'"'

=l\ f{x')dx'+-y]coB'^x\
and

/(a;-2jia)='

/(a;Va''+-S'^o^

aj(2,-i)a

it

x'dx'

f{x')cos'^x'dx',

f{x')co&--x'
V^
a
J(2-i)

dx',

follows that

V /(x+ 2na) =i Vf[x') dx'+lf^ cos

'^'^

x [fix') cos*^ a;' cZx'.

* Of. Poinoar^, Theorie de la propagation de la Ghaleur, p. 91

Watson,

loc. cit.

(3rd Ed.), p. 475.

Whittaker and

THE USE OF SOURCES AND SINKS

160

Let/(a;)

=e

(a;+2ma)'

00

and we have

*"*,

a:''

/.oo

=-

rfa;'

aJo
^

^^^-^

\_

<

+ -Vcos
a

x''

e"

a;

cos

Jo

xe
+ 2>,cos
a

"

jc'c^a;'

\.

Therefore
(x-x:+ina)'
{x-x:+inay

S^=^^* p[1 + 22 cos '^(x-c.')e-''^']

2e

m't'.I

II

.(3)

and
(a;+a:'+2g )'

--^^

tt

Using

and

(3)

+ 2Vcos

L_

the solution

(4)

.\-^
^Tx/
v=~\
f(x)y,e

(x+x)e

"

(4)

{*

_|

reduces to

(1)

-''^'

sm

"

a;sin

x ax

n-ir

,,

,
.

Temperature zero. Boundary x =


Boundary x=a kept at zero.
Starting with the Continuous Doublet of strength 2k0i() at
x=0, which would keep x = a at temperature (pi{t) if the solid
extended to infinity, we have to take an equal doublet at x = 2a
to keep the plane a; =a at zero
and so on.
Thus we have doublets of strength 2K(pi(t) at the points 2na,
n being zero, or any positive or negative integer.
III.

The Same

Solid.

kept at temperature

Initial

(pi(t).

Therefore

^=o^r-A ^^^(i^+^na)e- ^^^t-n)dt'.


A

may

corresponding result

boundary

a:

these solutions

is

we

76. Application of the

of

Q
*

of

Images to Fourier's Ring.

Source of strength

< <

tt,

at

x0.

Initial

zero.

Consider the problem of the rod


strength

^1 addition

34.

Method

Ring of Unit Radius.

Temperature

<^4f)t ^'^^

are led to another form for the expression for the

temperature in the problem of

I.

be obtained for the case when the

kept at zero and a;=a at

tt

a;

at a;=0, and no flow of heat across

Cf footnote
.

p.

30.

its

with a source of
boundaries.

The

IN CASES OF VARIABLE
solution

is

TEMPERATURE

obtained by putting equal sources at the points 2mr, n

being any integer, positive or negative;

given by

It

is

and the temperature

is

fa+2^)a

clear that all the conditions for the temperature in the ring

are satisfied
II.

161

by

'

this solution.

Ring of Unit Radius.

Initial Temperature v=f(x).

Since, as above, the temperature

due to a source

of strength

^Ji-KKt)
for the initial distribution /(a;)

Using

75

=^ r

(3), this

we have
/

fir

"I

reduces to

/(*')[1+2S

cosji(a;-a;')e-'''*] dx',

as in 12.*
77. Application of the

- 00

<z<

Kj.

Source at x'(x'

Method of Images
is K^, and when

to the Infinite SoUd.

the conductivity

< x < oo

^->S.<->)

a)

t=-S'

(-<)

(^)

10=0,

(3)

Vi=W2,

when

-='
^'^^S=^i' -^and Lt Vi=0, except at x~x', where

v-i

2V(W)^
Ltvs=0.

is

to take the

(^>

form

'"''
(^

< 0)

(=0

* Cf. Niven, London, Proc. S. Sac, 76 (A), p. 41, 1905.


C.C.H.

is

> 0).

We require a solution of the equations

and

When

the conductivity

^'^
(6)

THE USE OF SOURCES AND SINKS

162

and

Differentiating (3)

(4)

with respect to

and using

(1)

and

(2),

we have

at .i=0,

3^-yi

3'i'2

AiKi-g^ = yi2K2 g^,

and,

Now, using Maclaurin's Theorem,

we have

2..

= (l+a)|l,(V(^),.)"(0)

-(-4M-V(:i).^rO)o'
a=

where

-'^^f-^

This holds for any integral of

(1), (2), (3)

Take

= re

first

the expression

Vi

and

*"!*

(4).

.'

(7)

v'

which

satisfies (1).

Then we have
,

Mki/kj) -')'

(a;-^/(.^/.l)a!')^

= i+i!;eNow

it is

clear that this

is

u^t

,_

(V(K,Mx+!^y

+^e~

not the solution

we

(a!+V(K;/Ki)a;')'
'^'4

(8)

require, as the second term

involves a source at the point sUk^Ik-^x' in the part of the solid where

00 < a; <
we would

We

0.

are able to obtain the true solution

by considering what

obtain from a sink at s]{kJk-^x', such as this second term

requires, in the

Exactly as in

same way as above.


(8)

we

have, corresponding to the solution of

(2),

V'

So that in the case of


/,

^^-^^'^~

(x+VMKi)x')'

""

(^>
'

IN CASES OF VARIABLE TEMPERATURE


we have
i\

(jg+a; )-

/I

i\ /I

163

(xx')''

-=-o-j)(&>'^-(^-;)G-^>-^
Adding the

solutions

(7), (8), (9),

_
(a--l)'i\

_/,
-(

and

"2

which
in

^'i'

(,x+x'f

l-a^

*"!'

(a!--^Wici)a;'P

= o~77

we have

(10),

(x~x' )'

and

(^)

^''^'
'

satisfy all the conditions of the

the strength of the source at

problem except for a numerical factor

x'.

We obtain the actual solution

by dividing both by

(a+l)V(^Ki)
'

2a

and thus have


(j:+a;')^

"^=2^0^'

+;^i2-;/^^)

'^''

^^^^''^
'

_(--vWs)0!
*2<

2a

"2 = ,;
1 + a

and

^tTi

(^

r\

2;y/(irKi)

< 0)
'

Thus the temperature at any time t in the part a; > 0, is the same as
solid had been of that material and another source of strength

if

the

whole

had been placed at x': and the temperature in the part a; < is the same
as if the whole solid had been of that material and a source of strength
2^1 Kg
had been placed at J{kJki) x'.*
78. Applications of

the

Images in Two or Three

Method

of

x> 0.

Initial

Dimensions.
I.

Semi-Infinite

Boundary

x=0

Solid,

Temperature f{x,

y).

Tceft at zero.

Starting with the line source of strength /(a;', y')dx' dy' at

we must take an equal

{x', y'),

sink at {x', y') to satisfy the condition

at the boundary.

Hghcg
*"'

=4^J/^'J_/(a:',2/'){e

-e

*'

Weber, OoUingen,
;
Bd. II., 40.

* Cf. Sommerfeld, Math. Ann., Leipzig, 45, p. 266, 1894

Nachr. Ges. Wiss., p. 722, 1893

Weber-Riemann,

loc. cit.,

]dy'.

THE USE OP SOURCES AND SINKS

164
II.

Semi-Infinite Solid,

x=0

Boundary

x>

Initial

0.

Temperature f(x,

Starting with the point source of strength


at

{x', y', z'),

y, z).

kept at zero.

we take an equal

sink at {~x',

/(a;', y', z')dx' dy' dz'

y', z'),

since these give

zero temperature at a;=0.

Hence

^-(27(^r

LL

f^''

'''

^0{e---e--}c^.'^.'e^.',

R^^{x-x'Y+{y~y'Y+{z-z'Y
R'^={x+x'Y+{y-y'Y+{z-z'Y.

where

and

The Wedge of Angle Trim, where m is any positive integer.


The two and three-dimensional problems given above in I. and II.
are special cases of the wedge of angle 7r/m, where m is any positive
III.

integer.

We shall now treat this problem, confining ourselves to the

two-dimensional case of a hue source at the point

[x', y'),

the edge

wedge coinciding with the axis of z. The three-dimensional


case of a poiut source at (x', y', z'), and the extension to the general
problem of an. arbitrary initial temperature offer no difficulty.
Taking cylindrical coordinates, the surface of the wedge is supposed given by the planes 0=0 and 9=Tr/m; these planes are to
be kept at zero temperature.
Within the wedge we have <6 < Trim.
Let the source be placed at the point Pq whose coordinates are
of the

(a, a).

Let the

circle

and d=Trjm at

through Pq

and

^^^

(Pig. 11).

its

centre at the origin cut

6=0

IN CASES OF VARIABLE TEMPERATURE


Then the angles AOP, FOB and AOB
^=Trjma and y=7r/m.

^ and

are a,

165
y,

where

Start with the unit source at Pq.

To

of

Pq

in

04

To balance the
of

Pj in

05

i.e.

To balance the
of

Pj

OA we

give zero temperature at

image

in

04

i.e.

i.e.

put a unit sinh at P^, the

at (a).

sink at Pj, in

OB, we put a

sowrce at Pg, the image

at (a+2y).

source at Pg in
at

(a+2y)

OA, we put a
and so on.

sinfc

In this way we have the set of images P^, P^,

at Pg, the image

where

PP2=P,P4=-=2y,

'

P^P^=P,P,=...=2y.

(a+2(m l)y).

Also Pam-i

lies

Thus

PoP2_i+2/3=2a+2(m-l)y+2/3=2,r.

at

Therefore Pgm-i coincides with the image of


set of

images

is

in

OB, and the

closed, P2,,_i being the last.

Also these sources and sinks give, with the source at Pq, zero

temperature over the planes

The temperature

at

{r,

0)

0=0 and

due to
2m-l

v=

d=Tr/m.
system is

this

(-ir^..'

(1)

s=

where

v^ is

the temperature due to a unit source at P^ in the infinite

solid.

But we have seen


a unit source at

(r',

in 72 (2) that the temperature at

(r,

Q')\s

R^=r^+r'^-2rr' co& [d-d').

where

Using Neumann's expansion *

this

may

be written as

^
or

fAe-"^'*

S J{Xr)J^{Xr') cosn{d-6')

if-y;cosn(e-6l')
itTf trSi

* Cf.

Xe-'^'t

J (Xr) J (Xr') dX.

Jo

Gray and Mathews,

dX,

loc. cit., p.

27

(69').

6)

due to

THE USE OF SOURCES AND SINKS

166
It follows

from
00

-1

v=^^

that

(1)

2 S

{cosn{9

a2sy)cosn{9+a+2sy))

^Tf n=-> s=0

r\e-'''tJn{\r)

JJ\a)

d\.

...(2)

Jo

When n is not a multiple of m the series in s has zero


and when it is a multiple of m, its sum is equal to
2m sin nO
Thus, from

(2),

fj

CO

{a, a)

in the form

/CO

Ssin
^sin^a Ae-^^* Jp. (Xr) J^,, (Xa) ciA
?;= p^i
y Jo
y
y
7^
where, as above,
For

tlie

sin na.

wedge 6 = 0, 0=irjm,

in the

sum

solution of our. problem of the unit

we obtain the

source at the point

for its

we have

written

three-dimensional case,

we

(3)

for n-jm.

start witli the expression

r+r" 2rr'cos(9 8')-Kz 2')'

(2^(7rK0)^

corresponding to the unit source at

(r',

<:').

ff,

Proceeding on the same lines as above,

we

obtain the solution of our

problem in the form

17^

sin

6^

sin

"yJ(7rK<)p=:

79.

^y

Jp{Xr') dk.
f Xe-<^->^Vp.(X.r)
-V

-z-

Sommerfeld's Extension of the Method of Images.

The method of images as used above for the wedge of angle tt/w,
m is any positive integer, fails when the angle is mrjm, where
TO, n are both positive integers, prim.e to each other.
For example, when the angle is a right angle, and the given
where

source

is

at Po(r',

6'),

a sink at Pi(/,

and

< < ^tt, the images are as follows

where

0)

a source at

a sink at P^ir', nd').

But when the angle

is

27r/3

P-^ir',

tt+O')

(Cf. Fig. 11.)

and the given source

is

at

IN CASES OF VARIABLE TEMPERATURE


^o(*"'j 9'),

where

<

5'<

the successive

2ir/3,

167

images are

follows

a sink at Pi(/, -d')

and

p.

a source at P^(r',

a sink at Pg^/,

-y^-^')

a sinkatP5(^r',

-^-0').

^+0')

a source at P4(r',
(Fig. 12.)

^+d')

as

THE USE OF SOURCES AND SINKS

168

problem

after applying it to the heat

bounded by the planes

of a source in the region

6=0

and 6=27r, Sommerfeld by its aid


gave the first exact solution of the diffraction of waves by a semiinfinite plane {e.g. 6=0).
But a simpler method of treating these
questions both from the equation of conduction of heat and for the
other partial difierential equations of mathematical physics has now
been evolved* For this reason it will be sufi&cient here only to
give references to Sommerfeld's and other papers in which the
Riemann's Surface idea is used.t
We return to the problem of the wedge in
obtained ia 78, III.
true for the

wedge

of

(3) for

any

the angle

w/m

90 and the solution


will

be found to be

angle.

*See below 90 and papers by the imthoT

in

London, Proc. Math.

Soc. (Ser. 2),

1910 and, 18, 1920.

8,

(1) Math. Ann., Leipzig, 45, p. 274, 1894, and 47, p. 317, 1896
London, Proc. Math. Soc, 28, p. 395, 1897. Sohwarzsoliild, Math. Ann.,
Leipzig, 55, p. 177, 1892.
Carslaw, (1) London, Proc. Math. Soc, 80, p. 121,

t Sommerfeld,

(2)

1899

(2)

London, Rep.

Brit. Ass., p.

Phil. Soc, 18, p. 277, 1900.-

644, 1900.

Hobson, Cambridge, Trans.

CHAPTER X
THE USE OF GEEEN'S FUNCTIONS IN THE SOLUTION
OF THE EQUATION OF CONDUCTION
80. Introductory.

The use of Green's functions in the theory of potential is well


The function is most conveniently defined for the closed
surface S as the potential which vanishes over the surface, and is
infinite as 1/r, when r is zero, at the point P{x', y', z') inside the
surface.
If this solution of the equation VH=0 is denoted by
G{P), the solution with no infinity inside S and an arbitrary value
F over the surface is given by
known.

'd

S^G(P)VdS,

4'!rJJ(

5- denoting differentiation along the outward drawn normal.*

We

proceed to show

how

a similar function

may

be employed

with advantage in the mathematical theory of the conduction


heat.

In this case we shall take

at \x, y, z) at the time

due

to

of

the Green's function as the temperature

an instantaneous point source of

unity generated at the point P{x',

y',

z')

at

strength

the time t, the solid

being initially at zero temperature, and the surface being kept at zero
temperature.

This solution

may

be written

u=F{x,
and u

satisfies

y, z, x', y', z',

{t

> t)

the equation

*Cf. Clerk-Maxwell, Electricity


Slectricity

tr),

and Magnetism,

and Magnetism, Vol.

p. 290.

169

I.,

97

(5),

and Webster,

THE USE OF GREEN'S FUNCTIONS IN THE

170

However, since

only enters in the form (< t),

we have

also

Ot

Further, Lt(M)
(x', y', z'),

at

all

points inside S, except at the point

where the solution takes the form


(3:-a;')i!+(y-0'+(z-z')2

1
;

Finally, at the surface

/S,

4/t(t-T)

M=0.

{t <it.)

Let V be the temperature at the time


surface temperature

Then v

(j>

{x, y, z, t)

in this solid

due to the

and the initial temperature/(!r, y, z).

the equations

satisfies

j=kVH, (>0)
v=f{x,

and

y, z) initially, inside

v=(j>{x, y,

Also, since the time

interval for

t,

when

S,

>

0.

former equations

of our

lies

within the

we have
^^ = kVH,
v=(p{x,

Therefore

at S,

z, t)

=- iuv) =m=-

OT

Ot

{T<t)

y, z, t) at'the surface.

[mV^ vV^u],
+ v^5-=k
Ot

and

a"

^""^

t^a;

% &1

c?T

[(mV^i;

[
1

- v^hi) dx dy dzA dr,

the triple integration being taken throughout the

any positive number

less

than

t,

as small as

we

solid,

and

being

please.

Interchanging the order of integration on the left-hand side


this equation

and applying Green's Theorem to the right-hand

we have
{uv)^^t-tdx dy

dz

{uv)r^odx dy dz

='ir[iKi-"a)^^:

of

side,

SOLUTION OF THE EQUATION OF CONDUCTION


w^liere

171

denotes differentiation along the inward-drawn normal,

and we have used the condition that u vanishes at the surface.


Now take the limit as e tends to zero. The left-hand side gives
M<MjjMT={-o(^a'!y
the

(?z

JJ \{u)r=o{v)r=odxdydz,

integral being taken through

an element of volume including


where the function u becomes infinite at
t=T, the second integral being taken through the solid
and
first

the point P{x',

y',

z'),

[vp]i

stands for the value of v at the point P(x',

time

t.

But

since

{x','y', z')

is

the temperature at the time

y',

z')

at the

due to a source at

at the time r,
\\\{'^)T=t-odx dy

dz=\,

and we have
M=jJJ(w)r-o()T=o(^a:c?2/(^z+/cl

rj|?;g^t?<S

dr

=JJJ(M)T=o/(a;, y,z)dxdydz-\-K\ \^^(ji{x,y,


as the temperature at

tribution /(a;, y, z)

x', y', z'

at the time

z,

T)^ds\dT

due to the

and the surface temperature

(j)(x,

y,

(1)

initial dis-

z, t).*

In the case of Kadiation at the surface, the Green's function u


is

taken as the temperature at

{x, y, z) at

point source of strength unit generated at


taking place at the surface into a

The temperature at
distribution

/(a;, y, z)

* This discussion is

Dissertation,

time

due

{x', y', z')

medium

to

an instantaneous

at time r, radiation

at zero temperature.

P(x', y', z') at the time t due to an initial


and radiation at the surface into a medium

due to Minnigerode, and was published in his Gottingen

Uber die Wdrmeleitung in KrystaUen, Gottingen, 1862.

Ann.

Cf.

also

Toscane, 10, p. 143, 1868, Pisa; (2) Ann. Mat.,


Milano, 1, p. 373, 1868 ; (3) Mem. Soc. Italiana delle Scienze (Ser. 3), 1, p. 373,
1868, Firenze ; (4) Collectanea Mathematica inedita in Memoriam Domenici Chelini,
Betti, (1)

delle Universitd

Sommerfeld, Math. Ann., Leipzig, 45, p. 274, 1894. WeberII., 52, and papers by the author, (1) Phil. Mag., London
(Ser. 6), 4, p. 162, 1902 ; (2) Edinburgh, Proc. Math. Soc., 21, p. 40, 1903 ; (3)
London, Proc. Math. Soc. (Ser. 2), 8, p. 365, 1910. See also for appUoation to
Uber die Partielle-Differentialgleichung
the equation {V^ + k')u = 0, Pookels,
{V^ + k')u = 0, T1. IV. 4, Leipzig, 1891; Sohwarzschild, Math. Ann., Leipzig,
p. 238, 1881, Milano.

Riemann,

loe. eit.,

55, p. 177, 1902.

Bd.

THE USE OP GREEN'S FUNCTIONS IN THE

172

at temperature

(j>(x,

y, z,

t)

follows from

We find in the

that given above.

a discussion similar to

end

bp](=JJj(M)T=o/(a;, y, z)dxdydz-\-'hK^

UJ**^^*'

2^'

^>

'r)d^\dT

=\\Y'^)-'=of{x,y,z)dxdydz-\-K^ [Jj(g^).0(a:.2/.2.T)rf/SjciT,
since at the surface

(2)

= hu,

^^
drii

and our result takes the same form as in (1).


The solution of the general problems in conduction

of

heat

which the temperature

in

is

and

(2)

can be obtained at once.

infinity of order

(1)

(2) will

this

-.

-^
-

*',

'inrKl

change the equations which correspond to

be

[vp],=j{u)r=of(3>)dx+KJ

and

and

^'''

tt

With

respectively.

and

'

-E.

h
ik e
2^(7r/c5)

results

Instead of an

.k

{2V(,r/cJ)F'

we have

solid

required.

In the case of linear or two-dimensional flow of heat


similar to (1)

thus

is

reduced to the determination of the Green's function for the

du
<t>{T)^dT,

[pJ(=jJ(M)^=o/(a;, /)<?a;t^y+KJ

[j0(a;,

(3)

2/,

T)g^&J*r,

(4)

where the integration with regard to s is along the bounding arc.


In the articles which follow we shall employ these theorems in
solving various problems in the conduction of heat.
Some of them
have already been discussed by other methods, the possibility
of

the expansion of the arbitrary function in the form required in

This assumption will not

the solution being assumed.

now

be

necessary.
81. Linear

Flow

Temperature

Semi-Infinite Solid bounded by x=0.

of Heat.

Boundary kept

at Temperature ^(t).
In this case the Green's function, or the temperature at {x,

Initial

at the time

f (x).

due to the unit source at

J-

{x', y', z')

at the time t,

^-^e Mt-T)_e-4(-r)L

2V(' K{tr))

y, z)

is

SOLUTION OF THE EQUATION OP CONDUCTION


It follows

from

80

the temperature at

with a slight change in the notation, that

(3),

{x, y, z)

at the time
(g-ic')'

is

given by

(a;+it')' ^

17.

4/ti

g 4(-t)
(^r,

(-t)*

which agrees with

The Same

82.

into a

We
solid,

Medium

23.

Source at

Solid.

x'

t=0.

at

Radiation at

x=0

at Zero.

start with the solution for a unit source at x' in the infinite

namely

1
/CO

cos2hxdx=i^ e

Since
it is clear

^*

la

Jo

that
e-^e><{-!B')(;a.

(1)

'~27rJ_

2J(-,rKt)

Consider the integral

J_

L-KaHeMx-x')ga

Ztt J

> 0)

over the path (P) in the a-plane of Kg.

The path (P)

in the

14, this

path being

a - plane

Tig. 14.

chosen so that at infinity on the right the argument of a

and

Jtt,

and on the

left

between

Itt

and

lies

between

tt.

Fib. 15.

Complete

this

path as in Fig.

15.

The

integral vanishes over

the circular arcs joining the path (P) to the points oo on the real
* Cf.iJootnote, p. 30.

THE USE OF GREEN'S FUNCTIONS IN THE

174
axis,

and there are no poles

of the integrand inside this closed

circuit.

from Cauchy's Theorem, that

It follows,

g-Ka.'teia(z-X')^^^_J_L-^a.Heia.(x~X')^a,

(2)

the second integral being taken over the path (P).

Also

we have

the second integral being taken over the path (P).


It is convenient in the following

>

x',

and

(3)

when x

<

x',

Now
is

argument to take

for the transformation of

(2)

when

is in-

(1).

r- Ae-'-'V'^

a solution of the differential equation p:=f v^) when

dependent
If

(P),

of

X and

we choose
and add

it

properly, integrate this solution over the path

to (2)

function satisfying

The choice

of

t.

all
is

and

(3)

above,

we

a temperature

find

the conditions of our problem.

indicated

by the equation

when a!=0.
This equation leads to

A=

? e'"*'.

Then we have
'0=

~^ fe

'%'(:-^')rfa+7i-

Itt i

f?-t^ e-'"''e*(^+^')da,

...(4)

2'!rjnia

over the path (P), the positive or negative sign being chosen
according as

This

may

a;

5 x'-

be written

If\e-'"^'i{eiMx-x')j^pia(x+x'))da V^
+^ttJ

v= ~~

f,ia(x+x')

ttJ

e-'"''

+
,

.,

da,

(5)

ltl

the integrals being taken over the path (P) and the positive or

negative sign being chosen according as

We shall now
of

our problem.

show that

a)

5 x'.

this value of v satisfies all the conditions

SOLUTION OF THE EQUATION OF CONDUCTION

175

From the way in which it has been built up, we see that it satisfies
the equation of conduction and the surface condition at x=0.
Also, from

The

first

what has been

said above, the

term corresponds to the source at

first

x',

part of

(5)

reduces to

the second to an equal

source at x'.

We

have thus only to show that the second part


when <->-0 and x is positive.
Assuming that the integral

of (5) vanishes

in the limit

gia.(,X+x')

0-m2
'

a continuous function of

is

da, over the

for

path

t=0* we have

da, over the

(P),

(6)

only to establish that

path (P)

(7)

zero.

is

Take the closed circuit of Fig. 16, consisting of the path


and the part of a circle, centre at the origin, lying above the path

(P),

(P).

FlO. 16.

There are no poles

of the integrand of (7) in this circuit,

the integral over the whole


infinity,

is

zero.

But when the

and thus

radius tends to

the integral over the circular arc vanishes, x+x' being

positive.

Thus the integral (7) over the path (P) vanishes.


we have obtained in (5) the required temperature.

Therefore

The continuity

on the

the variable

is

when t&ff, is not difficult to establish somewhat


But the discussion of infinite integrals, when
somewhat outside the range of this book.

of this integral,

lines of F.S.,

Chapter VI.

complex,

lies

THE USE OF GREEN'S FUNCTIONS IN THE

176

Replacing the integrals over the path (P)


integrals over the real axis,

v=-

by the

equivalent

we have

+ cos a {x+x')

cos a(a; a;')

e-'"'''M

in (5)

237p(^ <^o^ ('"+ ^')

a sin a {x+x') vda

(8)

Therefore

ax+h sm
Iif,,(acosaa!+^sin

2 f"

V!;=-!

(a cos
e-""''-

aa;)(acosaa;'
ax)
(a cos ax
s-r-rs

sin ax
ax') ^
+ hi^sin
-da

,
,.,

Again,

[e-cos^ci^=^

[ e-sina^.i^=^,.

and

It follows that

2hC' _

The

last

Ji cos

e-'-'^n e-"(cosa{x + x'+i)di]da

= ^f

e-^fff e-'"'''cosa(a;+a;'+^)rfa]<Zf

(8)

we have
e'

,.^^^,

+e

-2A

e-^e

term in

sinks extending

(10) represents the

from x' to

83. The Same Solid.


Medium at Temperature
The Green's function,

^^J.

(10)

temperature due to a Une

of

oo .*

Initial

Temperature

f (x).

Radiation into a

^(t).

or the temperature at

due to the unit source at


The

a{x+x') asin a{x+x') ^

= ^j

Thus from
^;=

{x', y', z')

{x, y, z)

at the time

at the time t, radiation taking

was first given by Bryan (Cambridge, Proc. Phil. Soc,


See also Bryan, London, Proc. Math. 8oc., 22, p. 424, 1891. The treatment in the text is taken from the author's paper, Edinburgh, Proc. Math. Soc,
*

7, 1891).

21, 1902.

solution in (10)

SOLUTION OF THE EQUATION OF CONDUCTION


place at the boundary
82 (10) in the

x=0

into a

medium

177

at zero, follows

from

form

Jo

2V(7r/c(-T))L

''J

Thus from 80 (3), for the general problem with initial temperature
f{x) and the medium at temperature (p{t), the temperature at
(a;,

y, z)

at the time

=27ib)l

is

given

by

K "''+' "-4'

*-

"

-'fJ/M*']
(1)

Let the initial temperature be unity and the medium at


Then, from (1), we have
C

v=^

,,

(X-xy

\e

2V(tk)Jo L

(+')'

+e

..

f"

-2A

(+!!/

+ '

zero.

l
d^dx'.
-I

Jo

Now
r'^

dx'\
^{tTKt) Jo

Jo

^^^^^=
^

AJn

-^x
V(tkOJo

<f

*^*~'^' r

Jn

LJ _xi
x+$

-i

2\/(k)

Thus we have
1

J-n-i

if"

if"

^/TJ_^_

f_.,^^

84. Finite Solid.

-M-'-5^

>/('rKOJ

-^-^^IF,^,

Source at x' at t=0.

a. in 25.

Boundaries

x=0 and x=a

at Zero.

To obtain the Green's function when


surface, we start with the solution
1

''-2VM)L

(^-')'

there

is

no radiation at the

Cx+xy

THE USE OF GREEN'S FUNCTIONS IN THE

178

which

and

the condition for a source at x' at the time i=0,

satisfies

also the

This

may

boundary condition at a;=0.


be replaced as in 82 by

the integrals being taken over the standard path (P) of Fig. 14
also

when x >

x'

we choose

the positive sign

when x

<

x',

the

negative sign.

Thus we have the transformation


i'n

= 1 f e~'"''*sinaa;'e'"''(?a

(a!>a;)

[x

(1)

g- '''* sin ax e'^'da,

<

x')

^7^J

the integrals being taken over the path (P).

Vt=

Let

\Ae~'"^'''SLnaxda,

over the path (P), and choose the constant


at

a;

= a is satisfied by

Then

v,

so that the condition

v=Vo+Vi.

= \e'

^sinaa;smaa;

.,
(2)

sm aa

^7rJ

Consider the solution

= + Vi

V(,

smaa

iw]

^_^f

.(3)

^^..sma^sinoja-^^^

sm aa

^7^J

The value of v given in (3) satisfies the equation of conduction and


when a;=0 and x=a.
We shall now show that it satisfies the initial condition for a

vanishes

source at

x'.

We

have seen that LtQ has the required value.

We

t->0

have thus only to show that Lt

1^1

= 0,

where

v^ is

given in

(2).

tt-O

Assuming that the


,, sin
f e-KaH

integral

aa sin ax
._

e'-^da,

.^

over the path

(P).

(4)

SOLUTION OF THE EQUATION OF CONDUCTION


a continuous function of

is

for <=0,*

sm aX sin ax
sm aa

we have only

179

to establish that

over the path (P),

e^"'da,

(5)

zero.

is

This follows, as before, from the closed circuit of Fig. 16, since
there are no poles of the integrand of (5) in this circuit, and the
integral over the whole vanishes.
But when the radius tends to
infinity,

the integral over the circular arc vanishes, provided that

x-\-x'2a

path (P)

is

is

negative.

integral (5) over the complete

Thus the

zero.

The solution (3) can be reduced to an infinite series as follows


Take the path (Q) of Fig. 17 formed by the path (P), the image of
this path on the real axis, and the circular arcs (dotted in the
:

The path (Q) in the a-plane


FIQ. 17.

The dotted part

diagram) joining the ends of these two curves.

Also the integrand in

this circuit gives zero in the limit.

odd function of a.
Thus we have from
1

v=

;rT-

(3) is

of

an

(3)

j,smaa;
e-'"^'

sma(a a;)
^^
-.

-da,

smafis

27rJ

^
,<
^x^a)
,

(x

.(6)

1
ii= --^

sinaa;sina(ffl a;') J
-da.
e-""*
^^

-.

sm aa

2*Tr J

the integrals

The

now

,n

(0

being taken over the complete path

poles of the integrand are at a

2 "

v=~ >,sm
a-^

for

integrals of (6).

See footnote, p. 175.

a a;sm nwa x

w,r

(Q).

= tt/k, 2Tr/a,

Thus, by Cauchy's Theorem, we obtain

the two

xx

-.r^t
'

etc.

THE USE OF GREEN'S FUNCTIONS IN THE


[ence the Green's function for this case

->,sm a;sm
a
a^

the temperature at x at the time

t,

is

',

when the

initial

temperature

and the surfaces x=0 and x=a are kept at


follows from 80 (3) in the form

'(x),
!),

y=->,sm

aV

sm

a;

Jo

xe

"

^^(t)

and

f(x)dx

+^2nsin^xf[^,(T)-(-l)^.(T)]e-"">'"-^'dT.*
The Same

5.
[

x=a

lere

into a

we

Source at

Solid.

Medium

x' at

t=0.

(-')'
4k

-e
''''~2V(7r/cO

^_

L-KaHgi^{x-x;)^f^^

(a;>a;')

= -j^fe-'''e-^''(*-="')(Za,

{x<:x')

integrals being taken over the standard

Ne

associate with

Vi

IT

path (P)

of I"ig. 14.

this another solution,

= ^\e-'"''t{Ae^<'+Be-^'''') da,

the path (P), and determine

and

so that the boundary

iditions

+ ox
7S

when x=0. and x = a,

\-hv=0,

by v=Vq-\-Vi.
way we obtain

satisfied

[n this

x=0

start with the solution


1

'

Radiation at

at Zero.

(h sin

A = (h+ia)-

a(ax')-\-a cos a (a a;')

(i^^ a^)

^sinax^+gcosax-

B= ih+ia)
(A^

a.

34.

sinaa+2a^cosaa

^)

sin

aa + 2aA cos aa

'

^,

SOLUTION OF THE EQUATION OF CONDUCTION


Substituting these values for

f e-""''

(^ sin

ggi'

+ cos

ttJ

(i%*

when x'>x'

and,

and

S in

%, we have

finally

(^ sin a (g
+ a cos a (a
a^) sin aa+2aA cos aci
ga;')

a;)

when x -< x' we interchange


,

a;

181

and

a;) )

,^.

"'

x' in this

expression, the integral being taken over the path (P).

From
satisfies

the

way

which

in

(1)

has been built up,

we know

that

it

the equation of conduction and the boundary conditions.

We shall now show that it satisfies the initial conditions for

a source

at x'

As

v^

corresponds to the source at

x' ,

we have

only to show

that Lt 'Ui=0.

This foUows as in

84,

by introducing the path

of Fig. 16, for

the

integrals vanish over the circular arc in the limit, provided that

c+a;'>0,
cc+a;' 2a<0,

and

and both of these conditions are satisfied.


Also we know that the roots of the equation
(A^ a^)sinaffi+2aAcos aa =

and not repeated. (Cf. 36.)


The solution in (1) can now be reduced to an infinite series by using
x'
the path (Q) of Fig. 17. For we have from (1), when x
are all real

>

_^^,j(Asinaa;'+acosaa;')(Asina(a

+ acosa(a
a;)

(A2~a^)sinaa-f2aAcosaa

2i7rJ

the integral being taken over the path

a;))

,,

,
'

(Q).

Then, by Cauchy's Theorem, we have


^T-v

{Asinaa;'+acosaa;')(i^sinaa;+acosaa;)

^=2S-'^

a(/.^+a^)

+ 2^

,,
'

(^^

the summation being taken over the positive roots of the equation

(A^ a^)

sin

aa+2aA cos aa=0.

The result in (3) holds both when


Hence the Green's function is
_

a;

>

a;'

and when x

< x'

(Asinaa3'+acosaa;')(^sinaa;+acosaa;)

'

THE USE OF GREEN'S FUNCTIONS IN THE

182

The solution for an arbitrary initial temperature f{x) follows at


and we obtain the temperature at x at the time t, for the case
when the medium is at zero, in the form

once,

.,-0 ^%u.'\ K^ .-.aH (^ sin aai'+a cos ax')


]o^^

'^

{h sin

This admits of term by term integration, and

ow

.,

ax+ a cos ax) j^,


"'^-

a{h^+a^)+2h

(Asinaa;

acosaa;)

.,

,7

,,

may

be written
a 7/

/,

^^>

,->

(Cf. 36.)

may be noted that the result given in (5) leads to the expansion

It

Ex.

sin ax'

+ a cos ax') dx'.

The same Solid. Source at x'. The boundary x = kept


The boundary x = a radiating into medium at zero.

1.

at zero

temperature.

+
a^

2f

w = 22~'"' sinouKsiuco;'

Result:

where the summation

is

A"

t-t-

=7^,

aa^ + h(l+/ia)'

taken over the positive roots of the equation


a cos aa + h sin aa = 0.

Ex.

The same SoUd.

2.

Boundary conditions

as above.

Initial

tempera-

ture f{x).

Result

>

= 2 2 e"""

J,
'

a
,

Ex.

beat.

+
,,,,,,,
+ h{\+ha)
a^

A"

sin oa;

aa?'

f sin (ud fix') dm',

''

Jo

'

The same Sohd. Source at x'. The boundary a; = impervious


The boundary x=a radiating into a medium at zero temperature.

3.

y=2

Result

2 e""""

a^ + h^

a,
,

i /T^

>

cos

ax

cos

to

aa:',

where the summation extends over the positive roots of

a sin acs - A cos


Ex.

4.

The same

Solid.

(MS

= 0.

Boundary conditions as above.

Initial

tempera-

ture f(x).

Result:

= 2 S e"""

cos a.f

86.
I.

,,,,,,

aa' + A{l+ha)jo
,

coBax'f(x')dx'.

Two-Dimensional Problems.*

Semi-Infinite

Boundary

y=0

Solid y

kept at F{x,

>

*Cf. Hobson, London, Proc. Math. 80c.


[iven in 86, 87.

Initial

0.

Temperature f{x,

y).

t).

19, pp. 286, 293, 1889, for

the result

SOLUTION OF THE EQUATION OF CONDUCTION


In this case the Green's function
U=-,

\dJi
It follows
is

from

80

(4)

(a;-a:r+(i/+i/')' \

-e~

^'t'-^)

j-^~^[e

\dyJy=o

is

{x-iO'+iy-yV

183

*''<-^'

iTTKHt-rV^
that the temperature at

{x,

y) at the

time

given by
("

r"

fJie~Sam^oUd.

II.

a;~a!')'+(y+i/')^ -i

a'-a!')'+(i/-i/')

Radiation at

y=0 into Medium at F(x,


may be deduced from

t).

In this case the Green's function


in the

82

form
_
'

hii'

{x-afy + jy+y')'

{x-x')'+{v-y')'
4/c(-t)

iKljt-T)

_|_g

47r/i:( t)L

(a;-a;')'+(y+i/'+i))'

fto

and

Thus the temperature at

{x,

zero, is given

by

temperature

is

y) at the time

t,

when the

initial

87. Three-Dimensional Problems.


I.

Semi-Infinite Solid

Boundary
In

x=0

a;

kept at F{y,

>>

0.

Initial

Temperature f(x,

z).

y,

z, t).

this case the Green's function is

u^^

-Ae

*^(*--)

-e

* This result proves the truth of the statement at the

tinuous doublets.

M*-.)

end

of

74 about con-

THE USE OF GREEN'S FUNCTIONS IN THE

184

Thus the temperature at


/

at the time

JoJ-a)J-oo
/Same

**

The Green's function

u=

-\e

is

dx' dy' dz'

a;'+(y-i/')'+(g-g')'

i^(t^)

drdy'dz'.*

//__\*

Radiation at

/SoZtd.

given by

(3;+a;')'+(i/-y')'+(8-8')' -|

^Jy^ZjT)^

II.

is

iW

#-00

(a:-a;')'+(y-yr+(z-z')'

(x, y, z)

4K(t-T)

-e

x=0 into Medium

at

F{y,

z, t).

in this case

(x-a;')=+(i/-i/y+(z-z')'
i(-r)

(a;+a;')'

+e

+ (i/-y')'+(g-8')'
M-t)

87r^K"(^ t)(+'+ fl'+(a/-y')'+(8- 8')'

-2Ay

Thus the teniperature at


temperature
h

is

[<

{x, y, z)

zero, is given

f"

f"

88. Infinite Cylinder

f^^J-

at the time

t,

when the

t\

It

(a:+i)'+(i/-i/')'+(z-g')'

(j_^)f

r=a.

Initial

Temperature

f(r, d).

Temperature Zero.

To

obtain

tlie

Green's function for this case,

we

start with

r'+r''-2rr' cos (9-9')


1

n=

We transform this,
-I

OO

^TT-co

initial

by

rFM

8^*^4JoJ-J-Jo

^(-^>

e-'^e

-.

-.

int

4,r/c

as in 78 (III.)j into
/.OT

Jo

*Tlie footnote on page 183 also applies to this result.

Surface

SOLUTION OF THE EQUATION OF CONDUCTION

186

But
J

=J

ae-''''tJn(ar')HJ'\ar)da.

J 00

And

proceeding as in

82, this

ae-'=-"J(a/)J(ar)c?a

may

be replaced by
'

= -jfae-'"''*J(a/)ir'"(ar)c?a

(r>r')
-

= -ijae J (ar) ZfJ^' (a/) da,

<

(r

the integrals being taken over the standard path (P) of Fig.

r')

14.

Thus we have
"

Vo=-^^cosn{d-d')
when

>

r',

and we interchange

\ae-'"--'tJ{a7')HJ'^{ar)da,

and

r',

when r

<

r',

(1)

the integrals

being taken over the path (P).

To

r=a, we associate with


taken over the same path (P), where

satisfy the condition at

solution Vi

another

CO

Vi=-r-

Ug

2 cosn{d-d')

*7r -a,

and we choose the constant


Then we have

\Aae-'"''tJn{ar')Jn(ar).da,
J

so that

Vq+^i

shall

be zero when r=a.

^i=^S cos^(0-0') jae-V(ar')J(ar)^|Me?.

...(2)

Consider the solution

=_i
47r

2
f^

cos ^(0^

0')

ae -j4^J
Jn{aa)
'

'J

X {ff/'M</M--^/'(a)'^M}

When X is real and positive,

Thus

we know from Appendix

<"(a;)-e"''^?^'(-:)=2/(;).

I.,

da,

4 that

(3)

THE USE OF GREEN'S FUNCTIONS IN THE

186

when

>

the integral being taken over the path (P)

r',

when

being interchanged,

<

and

r, r'

r'.

The value of v given in (3) satisfies the equation of conduction,


and vanishes when r=a.
We shall now show that it satisfies the initial condition for a
Hne source at (r', 6'). This requires that Lt'yi=0, where v^ is
given in

But

From

'^

(2).

84 by introducing the path of Fig.

this follows as in

the approximations for

of the 2-plane,* it will

the

over

and

16.

HJ^^{z) in the upper part

be seen that

J(arV(ar)gJ"(aa) j

vanishes

t/(z)

circular

arc

the

in

provided

limit,

r+r' 2a <[ 0, a condition which is satisfied.


Also we know that the roots of J^{aa)=0

are

all real

that

and not

repeated.

The

solution in (3) can

now be reduced

to an infinite series by

using the path (Q) of Fig. 17.

For we

may

_ J_

replace the term

Le-'.H'IJ^ {HJ'^{ar)J,,{aa)-HJ'^{aa}JUar)}da

over the path (P),

by

half this integral over the path (Q).

Using Cauchy's Theorem,

2a^

this

term becomes

aer

J'(aa)

the summation being taken over the positive roots of J(aa)=0.

But

it is

Thus we

known

that t

find for the temperature at

source at (/,

(r,

6) at

the time

t,

due to the

%'),

.=;^.|S-(.--)e-.-.^-^<y^
and

this holds

"Of.

,4,

when r%r'

Appendix

I.,

4, 5.

;J-Cf.

Watson,

loc. cit.,

3. 63.

(1).

SOLUTION OF THE EQUATION OF CONDUCTION


The Green's function

When
at

(f,

the initial temperature


at the time

6)

When

for this case

therefore,

is f{r),

we

obtain the temperature

form

in the

t,

is,

187

the initial temperature is/(r,

6),

we have

If we assume that this series may be integrated term by


we have, for the coefficient of J(ar) cos nd the expression

and

for

rr

r'f(r', e')Jn{o-r')

n=0 the result must be

divided

term,

cos nd' dr' dff

by two.

Thus we are led to the series for/(r) and/(-, 0) of 57, 1, and


and they occur here as the limiting values of v when t^^O.
89. Infinite Cylinder r =a.
at Zero.

Initial

Temperature

Radiation at the Surface into a

it

Medium

f (r, Q).

Starting with the expression for the line source at

transform

IV.,

(r',

6'),

we

as before into

%=-TS coan{e-d') J{ae-'"^'tJ^{ar'}H^'\ar)da


^^^

(r>r')|
{ ...(1)

= --^f]cosn{e-9') faeV(ar)F'"(arVa,

(r

<

r),

the integrals being taken over the standard path (P) of Fig. 14.

We
^^1

then obtain the solution

= ^2 cos (^-<?')

fae V(ar) J(ar')

xl_i)
(aJ'(aa)

over the path (P), and

we prove

+ ^Jaa))

ida

...(2)

that ^='yo+%> which satisfies the

THE USE OF GREEN'S FUNCTIONS IN THE

188

dv

surface condition ^

\-hv=0, also satisfies the initial condition for a

or
source at

{r', 0').

Thus we are led to the solution

= -. S COS n{9-9')

fae-"""' J(a/)
J

"ItT -do

^HJ{ar)[aJ:(aa)+hJ^{aa)]

when
r

<

r >

r',

the integral being taken over the path

we interchange

r'

The

and

{Q) of Fig. 17, for

(P)-

and when

an

infinite series

we know* that the

by using

roots of the equation

aJ'(aa)+hJ{aa)=0,

(4)

and not repeated.


The coefficient of cos n{66') becomes

are

all real

d
* -^-v

1.

2^

"^

'

/^

"^

'

d(aa)

aaJ"

ff'"(aa)+M("(aa)

{aa)

+ {l+ha)J' [aa)

the summation being taken over the positive roots of

(4).

But

J(aa)+At/(aa)=0.

ajy

and

Also

we

a-j^-. HJ{aa) + hH^'^iaa) =

Therefore

jB^-^,

find that

Therefore the coefiicient of cosn{d

J-,
Tra

2 a2

d') becomes

J(ar).7(a/)
^

the summation being taken over the positive roots of


* Cf. footnote, p. 117.

-^^^

r'.

solution in (3) can be reduced to

the path

'^"'

aJ:(aa)+hJ,(aa)

(4).

SOLUTION OF THE EQUATION OP CONDUCTION


Hence the temperature
(r', 6'),

when

zero, is given

at

(r,

6) at

the time

due to a source at

radiation takes place at the surface into a

medium

by

(^Ha^-^)(J(a))^
and

this holds

When

and

the

when

initial

r'.

temperature

is/(j-),

or/(r, 6),

189

we have

at

190

THE USE OF GREEN'S FUNCTIONS IN THE

Introducing the complex variable

a, this

becomes

2-cos(a-e)_
4kJ

sda.

\'

over any closed path in the a-plane


=9'
and no other singularity of the integrand.
enclosing the point a
It is clear that singularities enter only from the poles a=2m7r+6',
and the infinities of e^rr'cos(^-e)yz^
On putting a=a+ib, we see that when 6=ao, cos (a 9)
must be negative or the integrand would be infinite. Hence in
deforming the path to 6= oo we must take care to have a in such
a region that cos (a 6) must be negative. The shaded portions of

the integral being taken

The path (A) in the a-plane


Via. 18.

Fig. 18 represent such parts of the a-plane,

the circuit round a =6'

may be deformed

and taking

\09'\<i'!r,

into that given in the

new path being composed

of two symmetrical curved


and two rectilinear parts, drawn at a
distance 27r from each other.
The integrals over these straight
hnes cut each other out, owing to the periodicity of the integrand
and the fact that they are described in opposite directions. We are
left with the two curved portions, which we refer to as the path (A)
figure, this

parts extending to infinity,

in the a-plane for this value of 6.

In connection with the problem of the wedge of angle mr/m,

was used to obtain another

this solution

of period 2mr, with only one singularity in that

SOLUTION OF THE EQUATION OF CONDUCTION


and to the second solution the method

range,

of images

was

191

applied.

(Cf.

79, and the author's paper in Proc. London Math. Sac, 30, 1899.)

Consider the expression


lira

_g

So

bhe integral being taken over the path {A), corresponding to the

current coordinate d.
(i)

This expression

is

a solution of

the equation of conduction,

since every element of the integrand satisfies this equation.

(ii) It is periodic in 6 and of period


change in d, e.g. from
29o.
to
9
0, simply translates the path along the real axis of a, and
the term e^rr'ooaia.-e'npM jg unaltered. Further, if the change in d is

equal to 2ndo (n being any positive or negative integer), the other


factor of the integrand also remains unaltered.
(iii) It vanishes when t->0 in the interval
6o<^6
when r-^' and d^>0', where it takes the form

y-(-r'

_ 2rr' cos (B

e_

<i6o, except

ff)-

!!

To prove this, we have only to note that the path (A) can be
changed into the two straight lines of Fig. 18, together with the
small circuits surrounding such of the poles as he in the interval
{d

TT,

Q-[-Tr).

The

integrals over these straight lines vanish in the

limit as i->0, since these lie in the

We

are thus

^o'C^'^C^o

^iid

circuits all vanish

source at
(iv)

^o<^<^o)

when

i->-0,

tlis

integrals

poles

round

and

if

these

except one which corresponds to the

(r', 6').

When

the factor
II.

shaded portions of the diagram.

with the circuits round the

left

The

r^-co

the expression vanishes

owing to the presence

of

e-'-V4.

solution given in (2) can be transformed into a series

involving Bessel's functions

by the

aid of the contour integral which

defines the Bessel's function of the first kind.*

*Cf. Whittalier and Watson,

loc. cit.

(3rd Ed.), p. 363.

THE USE OF GREEN'S FUNCTIONS IN THE

192

We

start with

r>+
*'!

^cos(a_e)

e*

fiKt

rst^a

lira

over the path {A).

Putting a

d^a,

tTr0'

_e

*0

becomes

this

So

|r!coaa'

v=-

'

iit(e+a.')

^TTKdfjt

da'
fay

over the path (A') of Fig. 19

77ie

path

^>4) /n

the a'-plane

Fig. 19.
iir(9+a')

In expanding the term

fafl'

upper and lower parts

of the

the positive imaginary infinity in the a'-plane

gi"''*"

we must proceed
path

M-(9+a')

differently for the

{A').

From

the upper part

we obtain
(

since at

vanishes and the series


.

is

convergent.

Also from the lower part

we have

SOLUTION OF THE EQUATION OP CONDUCTION

Now

we change the

if

integral to the

sign of a' in the lower part,

upper part, and have,

we

193

bring this

finally.

over the upper portion of the path {A').

Now put u'~rr'e-^'I^Kt. In the M-plane, the path proceeds


from the negative end of the real axis on the lower side, makes a
circuit round the origin, and ends at the negative end of the real
axis

on the upper

side, as in Fig. 20.

FIG. 20,

Also the integral

IS

replaced by

2::<,o8a' ?i5-'
e

da

Je^"'

At}) V

^^^

'

du

over this path in the M-plane.

Thus

it is

replaced

by

-2.e -J
on using the contour integral
Hence we have
_ r'+r"

2^).

for <7(2).

(.CO

aeV(ar)J(ar')^=2^e

But

_r'+r"

""

/irr'\*

e-J*V(^27j

for

any value

of

for

which the

real part is greater

than 1.

A,

p. 184 (2).
*Cf. Gray and Mathews, loc. pit., p. 78 (161): Nielsen, loc.
This formula for a positive integer n was first given by Hankel (Math. Ann.,
Leipzig, 8, p. 470, 1875) and for the general case by Sonine (ibid., 16, p. 40,

1880).

CC.H.

N,

THE USE OF GREEN'S FUNCTIONS IN THE

194

Therefore our solution, with period 29o, can be written

^=?^icos'^{d-d')rae-''^VAcr)JnA^r)da

^fo^

III.

Now

let 6' lie

found above in

Jo

(fo

by

(3)

^^^ denote the

0o>

solution

v{6').

This has the value required


s

< <

within

(3)

To

by a source

at the points

(r',

d'2sdQ),

being zero or any positive integer.

hj v(d'), the corresponding


v=v(e')~v{~d')

Similarly denote

Then

solution for (9').

=2i(cosf(^-^')-cos^(.+r))
/CO

ae-"^VUar)Jn^iar')da
'o

all

satisfies

0<^6

the

So

(4)

temperature in the wedge

conditions for the

<^9q, the unit line source being placed at

and the planes being kept at zero.


The solution given in (4) can be written

in the

(r',

6')

at !!=0,

form

=5-2sin^(9sin^e'| ae V^(ar)J,(arVa
^0

In

78, III.,

we found

angle Tr/m, where


to be true for a

We
If

we

Wo

t7o

(5)

1^1^

Jo

this solution

by images for the wedge of


The result is now seen

m is any positive integer.

wedge

of

have been dealing in

any

angle.

this section with a line source through

start with a point source at

{r', $', z')

(r',

d').

and take

(z-zfy

-^

we obtain by a

similar

r'+r"-2rr'aos(e-e')

argument the solution

V = -^j^
r Ssin;2-0sm-5-e'
foV^TKt; 1
^0
Co
t7o

91.

8=do

We

Infinite Cylinder.

kept at Zero.

ae-"* V(a-) J^^(a/) da.

The Surface

Initial

for the point source in the form

r=a

Temperature

and the Planes d=Q,

f(r, 6).

start with the solution of 90, corresponding to the line

source at

{r', 9'),

namely,

sin^Ssin^e'
^o=3-S
fo
fo
Jo
^0
1

ae-'"'V{ar')Jn(ar)da.
:

SOLUTION OF THE EQUATION OF CONDUCTION


As we have seen

^0= -5-

S ^9 ^
sin

may

in 88, this
sin

0'

be written

ae V.(a/)fl^i(ar)cZa

= -rS sinT Q sin^ 9'fae-V.(ar)ffi':(a/)c?a,

(r

>

(r

< /)

the integrals being taken over the path (P) of Fig. 14.
This leads, as in 88, to the solution for the source at

the solid bounded

195

(r',

r')

Q') in

by r=a and 9=^0,0 = 0^:

=-z-Ssm-3-0sm-5-0
Oo^

^0

^0

X
when
r

<

r',

>

r',

we

the integral being taken over the path (P)

interchange r and

and when

r'.

This solution can be reduced to an infinite series as before by


introducing the path (Q) of Fig. 17.
Then we obtain
(1)

mr

2w.

TS S ^

Jni,{-r)Jn,r{ar')Bn{m)

nir

^ sin

^"

d'ae-^-^

^f

...(3)

So

the summation being taken over the positive roots of

But, as in

Thus

(3)

86

ir (aa) J^ (aa)

J^(aa)=0

= --

reduces to
J^{ar)J,J,ar')

which holds when r % r'.

The solution for the initial temperature /(r, 6) follows at once


from (4) by integration, and we have, for this case,
'

4,

fa

fii

'JnAar)Jn(ary

sm -pr- 9 sm
^0

[Of. 62, II.].

-pr-

""

'

[J'nAca)J

dr'd9'

THE USE OF GREEN'S FUNCTIONS IN THE

196
The

following results can be obtained by the method of the preceding


In each case the surface of the solid is kept at temperature zero,
and a unit source is placed at < =0 at a point of the solid.
92.

section.

The Solid bounded internally by the Cylinder r=a. Line Source


Here the temperature at (r, 9) at the time t is given by
I.

iirri

'

at

(r', $').

H^^\aa)

J-'

X {J(ar)H<')(aa) -./(aa)ff<(ar)}rfa,

when r<

We interchange r and r', when r

r'.

The Solid bounded

II.

>r'.

internally by the Cylinder

r=a, and

the

Planes 0=0,

d=da- Line Source at (>', 8').


In this case we are dealing with the region
'

Then we have
Hn{ar )
t/O

wo

f/0

J -a

tt'"

X |./Mr(ar)H^(oa)-/n^(aa)ff^(a?')|(ia,
I.

when

r<r'.

III.

In

We interchange r

^0

^0

and

r',

when

^0

The Solid bounded by the Cylinders ra and r=b.

this case

we

^0

r>r',

Line Source

at

(r', 0'),

are dealing with the region

a^r^b, 0^e^2ir.
Then we have
{/(aa)ffm(ar')

- ^(ar')fl<(<i)}

K(aa)H/l(a&)-J-(a6W(ao)}
when r>r'. We interchange r and r', when r<r'.
The summation in a is taken over the positive roots

of

J{aa)HW{ab) - J(a&)ffm(aa) =0.

The above

result reduces to

.=^i2cosM0-^)a%-..

V^(alf-tva)

where U{ar) =J(ar)ff)(a6) ^ J(a6)rW(ar).


IV. The Solid bounded by the Cylinders
this case

we

62, I.)

r-a and r=b, and

6I=(9.

In

(a.

^"('-)^"("-'>'

are dealing with the region

the

Planes 6=0,

SOLUTION OF THE EQUATION OF CONDUCTION

197

Then we have
c

!''

(1)

X
Co

2, Sin -ga
"o

So

flo

"\

- JiuA<M')Hrm{oa)

J Jr^aa)EnA,o.ll')

/^7i;(a6)frjw(ar)-./B,r(ar)fl'jw(oi6)].

& Sin -g- ffae"o

J-

{ ^w(atl)fl'j-(a&)
*

fl[)

- J'jta-(a6)fl'n7r(aa)
flo

6o

when r>r'. We interchange r and r', when r<r'.


The summation in a is taken over the positive roots

>

flo

of

Jnn(aa)HnAb) - JrJiab) Hn^(aa) =0.


flo

The above

flo

flo

flo

result reduces to
^7r(afls)

=^'
""

9'a%-'*
1 2 sin ^ d sin ^
""
"

'

''

where

^a

{7^(^) j;'^(ar'),

4^(a6)-4u(<w)

Um,{a,r)=J^{ar)HnJiab)-J^(ab)Hn;r{ar).
flo

flo

flo

flo

flo

6=0

V. The Wedge bounded by the Planes z=0, z=:h,

Source at

In

flT

and d=9a-

Point

(r', 6', z').

this case

we

are deaUng with the region

Then we have
v=T-7r

2j

2/6

"

Wom=l=l

sin -T- K sin -T- z

"

sm 7p

t*

sin Tjt7o

C^o

Jo

ae-'"^'tJn{a.r)Jnn{ar')da.
-^
-57

VI. The Solid bounded by the Cylinder r=a, and


and 6=6(1. Point Source at (r', 6', z').
In this case we are deaUng with the region

the

6=0

Planer z =0, z=h,

O^r^a, 0^6^60, O^z^h.


Then we have
v=-ivr7v

2)

2e

a%6o ^uSil:!

"

sin -T- sin -T- sin -g- &


h
h
60

sm -g60

JnA.ar)J'nJ,al')

xe-^o-t-^

'

[f'nM)J
the summation in a being taken over the positive roots of J^(oia)=0.
(Of. 62, III.)

9S"

THE USE OF GREEN'S FUNCTIONS IN THE

198

The Sphere r=a.

93.

Initial

Temperature

f(r,

9, ^).

Surface

Temperature Zero.
In this case we start with

--

1
"^0

li^^tWi

19.

where R^=r^-\-r'^2rr' cos y, with the usual notation, y being the


angle between the radii to {r, 9, <p) and (r', 9', (j>').

v.=

Then

But from the expansion


we have

.... e^"'

of e""""* in a series of

Zonal Harmonics,*

e-"--V(y)?<^+').'ii''".()'-<-^>Thus

o=gi^^;7(;^E
f"

But

^* Jn.,i^;)Pnioos y).

(2^^+!)

*"*'

/it'r'\

ae V+i(ar)J+i(a/)rfa = -^-p^.^J+j^-2^j^t.

J^

Hence

""=

4^^/)

But, as in

aeV+i(ar)J+i(a/)cZa.

(2^+l)-P^(cos

...(1)

y)J^

88,

ae-''''Vni.i{ar)Jn+i{a/)da=-Aae-'"'Vn+i{ar')Hn+i{ar)da,
J

over the path (P) of Fig. 14,

when r

<

when

>

r'

and we interchange

Therefore

we have

o= -j^^^^S (2n+l)P(cos y)|ae-V+i(a/)ffL'+i(ar)(Za,

when

>

r, r'

r'.

r',

the integral being taken over the path (P), and

being interchanged

when

<

(2)

r, r'

r'.

Now let
CO

2
= S7rV(r/)
S (2 + l)P(cos

over the path (P), and choose


*Cf. Heine,

loc. cit.,

Bd.

I., p.

4 aeV+i(ar)J.H ('') t^",

y)J

so that

82

(14).

Vq-\-v-^

vanishes

fCf. footnote,

when r=a.
p. 193.

SOLUTION OF THE EQUATION OF CONDUCTION

199

^=:^iM.

This leads to

Jn+i{aa)

Also

we have

i^Jn+i{o'Ci)H'n+i{ar)-Jn+i{ar)HllU{aa}}da,

when

/>*'; while

we

interchange r and

can be shown, as before, that

It"

conditions for a source at

(r', 9',

r'

when

<

(3)

r'.

this value of v satisfies all the

(j>')

at

=0

in the sphere.

This solution can be reduced to an infinite series by introducing


the path {Q) of Fig. 17.
Then we obtain

,^_^VV(2n+l)PJcosy)ae--' '^"+^^'')'^y^(f)5+*("")
the summation in a being taken over the positive roots of J+ j(aa)

But, as in

88,

Then from

(4),

= 0.

H^+i (aa) J+ j (aa) =

we have

the summation in a being as above.


If

at

the initial temperature of the sphere

(r,

6,

(f))

at the time

follows

from

(5)

is/(r, 6, 96),

by

the temperature

integration

and we have

X e-'* Jnii^^iip ,'i gin e'dr'dd'd<t>'.


^
[J+i(aa)]'

Remembering that

THE USE OF GREEN'S FUNCTIONS

200
it will

in

be seen that this solution agrees with the result obtained

67.

However,

in the discussion in this section,

that the arbitrary function /(r, 9,

(p)

we have not assumed

can be expanded in a

terms of the form


(ar)-*J+j(ar) (1

- ^.'fD-^PM !! m<l>

series of

CHAPTEE XI
THE USE OF CONTOUE INTEGRALS IN THE SOLUTION
OF THE EQUATION OF CONDUCTION
94. Introductory.

In the previous chapter we have obtained the Green's functions


by integrating suitable solutions along a path in the
plane of the complex variable. The same method can be applied
in other cases, and, indeed, it is the simplest and most direct way
in several cases

of solving

many

problems of conduction.

In this chapter we shall

some problems already solved by elementary methods,


and to others which, so far, have not been solved at all, or have
only been treated by Heaviside's " operational method."* In this
class may be mentioned the problem of the semi-infinite rod composed
of two materials, the end kept at a constant temperature, the initial
temperature of the whole being zero, and the corresponding problems
for the finite rod and the sphere.
The methods lased in the solution
of these three problems give equally satisfactory results when the
apply

it

to

surface temperature varies with the time, or radiation takes place


into a

medium

at a constant or varying temperature.

* Heaviside's " operational

method " may be said to be simply a kind of shortThe formulae can be established by the use of the contour integrals employed
in the following pages. And the results are confirmed in this chapter. But his
work is hard to follow, and it may safely be said that he makes little attempt to
Indeed the real justification of his method
justify the steps in Ms argument.
seems to depend upon some such use of contour integrals as will be found below.
Reference should be made to Heaviside's Electromagnetic Theory, Vol. II.,
Chapter V., and to his paper in London, Proc. B. Sac, 52, p. 604, 1893 also to
hand.

Bromwich's papers, London, Proc. Math. 8oc. (Ser. 2), 15, p. 401, 1917 Phil. Mag.,
London (Ser. 6), 87, p. 407, 1919 and Cambridge, Proc. Phil. Soc, 20, p. 411, 1921.
The method employed in this chapter was given by the author in his paper in
See also Cambridge, Proc. Phil.
Phil. Mag., London (Ser. 6), 89, p. 603, 1920.
Soc, 20, p. 399, 1921.
;

201

THE USE OF CONTOUR INTEGRALS IN THE

202

LINEAE FLOW.
95. Semi-Infinite

perature

Vo-

Bod

(x

>

End x=0 kept

0).

and

^tt,

infinity in

and on the

left

tt.

This value of v

satisfies

the equation

x=0 and

It also satisfies the conditions at


initial

(1)

In this path at

over the standard path (P) of Fig. 14.

the right the argument of a Ues between

For the

Tem-

da

Consider the integral


?;=-. le'"*
ITTj
a

between f tt and

at Constant

Temperature Zero.

Initial

condition put

i=0

in (1),

da, over the


iirj

t=0.

and we have

path (P)

of Fig. 14."

Consider this integral over the closed circuit of Fig. 16, consisting
of the

path (P) and the part

above the path

of

circle,

centre at the origin, lying

There are no poles of the integrand

(P).

inside

and therefore the integral over the whole vanishes.


But when the radius of the circle tends to infinity, the integral over
the circular arc vanishes, x being positive. It follows that the
integral over the complete path (P) vanishes when x is positive.
For the boundary condition put x=0 in (1), and we have

this circuit,

iTTj

This

is

da, over the path (P).

da, over the path (Q) of Fig 17,

equal to

2^
2)
since the integrand

is

an odd function

of a,

and the

For a more exact discussion, we should consider Lt

(),

integrals over

when

is

a given

not

difficult

a;

->o

positive
to

number, and Lt

show that the value

when

a given positive number.

It

is

of v given in (1) is a continuous function of

t,

{v),

is

X being a given positive number, and a continuous function


t

of x,

when J^O,
when x^Q,

being a given positive number.

The same remark

applies to the discussion of the

throughout this chapter.

boundary and initial conditions

SOLUTION OF THE EQUATION OF CONDUCTION


the circular arcs at infinity (dotted in the diagram) vanish,
is

203

when

positive.

It follows

by Cauchy's Theorem that, when a; =0,

Thus the value

is

equal to

Do-

of v given in (1) satisfies all the conditions of our

problem.

But the path


of Fig. 21,

(P) can be replaced in (1)

by the

with the semi-circle enclosing the origiu

and,

by

letting

r\.

-00

straight path

o
FlO. 21.

the radius of this circle tend to zero,


in the

form

y=l

f
1

ttJo

Now

it is

known

we

obtain our solution

sinaa; ,
e-"-'
da
a

,>

(2)

that
cos

e-"''''

*"'

bxdx=i^e
2a

and that we can integrate under the


Ex.

p. 195,

sign of integration.

(Cf. F.S.,

13.)

re-'-'^^^dx='^ f e'^'db.

Thus

Jo

2a Jo

a;

It foUowB from (2) that


X

v=l

'

rwut)

f-

e-^'dl

(Cf. p.

35,

Ex.

1.)

s/tJo

96. Semi-Infinite

a cos

wt.

Initial

Rod

(x

>

End x=0 kept

0).

at

Temperature

Temperature Zero.

Consider the integral

\^^-''-^^

(^)

over the standard path (P) of Fig. 14.*


This value of v

ment

satisfies

the equation

as in 95 shows that

it

In

this case the

e~ '-x-;

path (P)

ia

>

^^^ ^^^ same argu-

also satisfies the condition at t=0.

For the condition at the end

^ =i^^^

of the rod,

WT-;

we take the

integral

over the path (P).

to lie above the points

a/(")*^

^""^

vi^)^^"

THE USE OF CONTOUR INTEGRALS IN THE

204
This

is

equal to
-icaH

a^da
.

2i7r 1'

and the poles

.,

over the path IQ) of Fig.

of the integrand are at a/( -je^^", a/(

17.

- je^^*"

FIG. 22.

It

follows

by Cauchy's Theorem

that,

when x=0, we have

v=acos(ji}t.

Thus the value

of v

given in

(1) satisfies all

the conditions of our

problem.

But the path


Fig. 23

(P) can be replaced in (1)

from right to

left,

with the small

VG)^''

^^^

by the

straight path of

circles enclosing the points

VG)^'^^-

o
FIQ. 23.

We thus

It

obtain our solution in the form

can be shown that

and from

this equality the solution in (2)

as found in 23.

can be reduced to

SOLUTION OF THE EQUATION OF CONDUCTION


If

the temperature of the end a;=0

solution

ao,

is

kept at

ada

Saw

we

flssintui,

205

start with the

-J'

and the

real

form

of the

-J(?)o'

97. Semi-Infinite

answer

is

// < \

Bod (x>

Temperature

at Constant

Here we take

0).

Radiation at

Initial

Vg.

o"'"

x=0

into a

Medium

Temperature Zero.

u=^le"'*T
h%a
iTT J

(1)

over the standard path (P) of Fig. 14.


It follows as in 95 that all the conditions of the
satisfied

by

this value of v, the condition at

problem are

a;=0 now being

-g|+A(-'"o)=0-

We may
path

and by

of Fig. 21,

to zero

we

IliVa f "
TT

J,

A sin

ax-\-a. cos ax,

/''"..
the solution in

(2)

e-nsmaa;ffia;

-
= hsmax+acosax
e-,
.
^
.

can be reduced to

v^^'Te-h^du-r
^Jtt Jo

as obtained in

For we have

and from

this

we

at

(^

a;

=0

e-v'dr,,

x+ii.

follows directly

=^0]]^^^'

from the value

*^ P^*^

'^^^

obtain the approximate value

3w_
p. 52.

'J

23.

The gradient

on

da

A^+a^

Jo

using the integral


I

as

straight

letting the radius of the semi-circle tend

obtaia our solution in the form

"

By

by the

replace the path (P) in-(l), as in 95,

"_/!

l_l

of in (1).

^^'l'

,.

THE USE OF CONTOUR INTEGRALS IN THE

206

98. Semi-Infinite
at

Rod

(x

Temperature acoswt.

>

Radiation at

0).

v=\e^^^

Here we take

x=0

into

a Medium

Temperature Zero.

Initial

"-3-;

(1)

57-!,

over the standard path (P) of Fig. 14.*


It follows as before that all the conditions of the
satisfied

by

this value of v, the condition at

problem are

x=0 now being

^\-h{vacoswt) = 0.
dv

The

solution given in (1) reduces to

..^sinax+acosaa;

2ha['^

^^+a2

Jo

,r

d^da

o^+coVk"'

y=tan-^^|^^.

where
For radiation at x =

into a

v=

medium

--haw

at a sin (ot,

start with the integral

ada

^e-"""'

e* T

we

r:

27-2

over the path (P).

The

solution reduces to

ha

V(?>sin(.*-V(^).-y)

V{('W)^S}
,,fesinaa;+acosaa;

ihao) f

h'+a^

Kir }
KTT

ada
a*+Ui'lK''

where y has the same value as above.


99. Semi-Infinite

Rod

from x=0 to
Constant Temperature Vg.

Material

(x> a).
Initial

With the usual notation,


a a; 0, and v^, K^, c^,

< <

Also

"

let

p.

x= a

x=0

of one

x= a

kept at

to

End

Temperature Zero.

let v^,

K^,

p^ refer to

i^i=Ki/cipi

The footnote on

From

Infinity of another.

and

203 applies here

Cj,
a;

>

p^ refer to the interval


0.

Kg=Z2/c2/02.

also.

SOLUTION OP THE EQUATION OF CONDUCTION


Then the equations to be solved are

W='^^a^'-<^<o=

(1)

(3)

a < <

Vi=0, when t=0,

It

where

a;

when

x> 0.

(3')

c=0,

KiP^=K2^
ox.
ox

when

a;=0.

Vi=(Aje?^+ B^ er^-^) e-"'"'*,

and

and

(5),

(!').

provided that

A^=^(l+a)A^,

where

(^)

a.

V2=0, when =0,

(5)

also satisfy (4)

Thus

Vi=V2,

ij.=sJ{kiIk^, satisfy (1)

They

a;

(4)

clear that

is

9f='^^^'^>^
when

Vj^=Vf

(2)

as follows

207

}
5i=4(l-cr)^2,

(T^nKJK-^^JiKrp^pJK^Cip-^.

Introducing the path (P) of Fig. 14 and choosing a suitable


value for A^

(this

value

is

indicated

solutions
L

0-) e

+ (1

by
0-)

(2)),

are led to the

e-^ e-"!"'*

(H-o-)e-''+(l-o-)e'

a^d

we

'^2=^"](i+^),-,.+(i_^),.-

^.

^^^

(7)

the integrals in both cases being taken over the standard path (P).

The expressions in (6) and (7) satisfy the differential equations


(1) and (!'), and the conditions at a;=0 given in (4) and (5).
They also satisfy the remaining conditions (2), (3) and (3').

For
is

equal, as before, to

da over the path (P)

v^.

Also the roots of the equation


(1

are given

by

0-)

e-+ (1

0-)

e'=

THE USE OF CONTOUR INTEGRALS IN THE

508

We may therefore use Fig.


ranish,

We

16 as before, and

we

see that Vj

and

v^

when t=0.

shall

now

simplify the solution of our problem

we have

)btained in the form of a contour integral.


I.

o-

>

1.

Put o-=coth
(l

^,

e.

+ o-)e' +{l (r)e^^ _


+ o-)e-*^+(l-(r)e*"^

sin

Thus

?',
'

= ^

"

^
^

"

r^ -

^-7

iir J

sm.

(ax iO)

sin(aa+t0)'

(l

da, over the

(aa+i6) a

path

(P),

i>or f s^^Jax-iO)

sin(aa;+^g) \e~'"''''^^

tTrJo lsiii(aa+*0)

sin(aa i0)ja

2^)osinh2g

sina(a;+a)

TT

e-"!""'^^

20 cos 2aa

Jo cosh

,g,

Similarly,

1),=

-.

TT

T^TfT
sm (aa+id)

da, over the

path (f ),

+ a)
1 + tanh 0) sin a
(1 tanh 6) sin a {fxxa)
(^la;

cosh20 cos2aa

I"

ej"'""'

.q>

<r=l.

II.

Then

%=2| e*<*+"^^

da,

over the path (P).

These reduce to
^^1=^0

^1 e-"'"'*
Jo

'-da

TT
/CO

_2?;,^"

e--i'(Z^*

(10)

~^''^'/

(11)

WW
2v,

,nd

"2=T^f

III.
ry,

< <
o-

1.

Put o-=tanh d.

+ g-)e^ +(l-g-)e-^ _ cos(aa;-tg)


(l + <7)e-*'+(l-o-)e^''
cos(aa+i0)'
(l

Using the integral

^'"^^-^"^

I''

^^='^

^''''f.

as in 95,

SOLUTION OF THE EQUATION OF CONDUCTION


Also our solutions

(6)

and

Jo cosh

TT

reduce to

(7)

sinag+a)

2j;osinh2gr

209

ej^^-^

29+ COS 2aa

^^

and
{'"

(l+tanh9)sina(jua;+a)

+ (l-tanli6')sina(//a;-a)e-''"''' ^^
cosh29+cos2aa

,,,
^^^^

This problem was discussed by Heaviside (loc. cit., p. 16), the gradient
a; = - a being required for the question of the Age of the Earth.
This
gradient follows at once from our solutions (6) and (7).

at

We

have

= -?

^^-M

)^

'r-^r- e - "i"'*

da,

over the path (P).

cr

^'^^

+l

L.=y ;4|SS->-^.
= -?^f

over the path (P),

A+2 2^"cos2aa]e-i'^=ia
BW

,2(7>

=
Hence

-^7^/l +2 f^e "".).

for large values of

we have approximately

/BpA

Pq

1+fe

'

^J(IrKlt) \K-fiiPi)

When

the surface

is

kept at zero and the initial temperature of the whole


temperature gradient, when x=-a, will be

solid is Vo, it is clear that the

minus the above.


But the gradient in Kelvin's classical treatment
( 28 above) was found to be vJ{iTKjt).

of the

Age

of the

Earth

This modification of the problem, allowing for greater conductivity and


capacity for heat in the interior than in the outer skin,

makes the interval

required for subsidence to the same gradient {K^c^p^j^K^Cip^ times as great


as before.

With the data adopted by Perry and Heaviside,* and


Thus Kelvin's estimate

{K^^2)j{K^c^Pi) is nearly 450.


be increased to 45 x 10' years.
* Of. p,

c.c.H.

221 and the footnote on

p. 60.

the above notation,


of 10' years

would

THE USE OF CONTOUR INTEGRALS IN THE

210

Rod

100.

Length

of

peratures Zero and


It

is

clear

/.

The Ends

Vg, respectively.

from the argument


v

x=0

and x=? kept

at

Temperature Zero.

Initial

of 95 that

Va f sinacc e-"""'
= ^\
da,

,_,

(1)

*7r J

Tem-

sm at

J-

over the standard path (P) of Pig. 14

the conditions of

satisfies all

our problem.

From

v=^

-.

2^7^J

And

we

this solution

obtain, as before,
da, over the path
(Q)
\-o/ of Fig.
ro 17.

smaJr a

by Cauchy's Theorem, we have

finally,

n.^TT-

v=v

>

~-\
I

TT^

'

' Sin
vp
sm-T-a;e

j,

-t-,

Another form of the solution can be obtained as follows

from

'a

etc.

(1),

'

ITT

.(2)

since the poles of the integrand are at 0,

We have,

'

*"

=^^\

{sina((2-l)?+a;)-sina{(2-l)?-x)}-~''''ia

(2tc+1);+!1:

= 7-2:
^/7^

J (2>-l);-a;

e-f',^A
t

(Cf.
^

75,
s

'

III.)/

(3)
i
\

2^''(()

A similar treatment of (I) leads to two expressions for ;^.


For example, we have
vo /e-K"'*

('dv\

=T- ^
hg\ox/x=o 7rj SinaiJ

^,

=f[l + 2S(-l)e-"i'']
'

The footnote on

p.

da, over the path


(P),
f
/>
\

208 applies here

also.

(4)

SOLUTION OF THE EQUATION OF CONDUCTION


And

(1)
\OX/x =

JJLof^
-e^lal
TT

211

^.uda

='L\
TT
^0

cos('iLn-\)a.le-"-.Hda

2
,,

-'2:ii)'"

,>2e

(5)

of Length I.
The Ends x=0 and x=^ kept at Temperaand Zero, respectively. Initial Temperature Zero.

Rod

101.
tures Vo

In this case the solution

obviously

is

^^^fsin(Z-^)e-^=*
*7r

sin al

over the path (P) of Fig. 14.

And

this leads, as before, to

The remarks

form of the

as to a second

solution,

and the gradient t^, are

equally applicable to this case, which, of course, could be deduced from the

preceding by a change of origin.

102.

Bod

ture Zero

of

and

x=0 and

s.=l kept at Tempera-

Initial Temperature Zero.


from the argument of 95 that
the problem are satisfied by

In this case
conditions of

The Ends

Length I.

Ct, respectively.
it is

clear

Cain ax

all

the

e-""'*

da,

(1)

iKTT J sin al a*

over the path (P) of Fig. 14.

Again we

may

write (1) in the form

V= jr^

fsin ax
;

2iKTr J

sm atf

over the path (Q) of Fig. 17.


And, finally, by Cauchy's Theorem,

e-"'''*

-s

da,

a"*

we have

^=-;^r^J+^s^^^^^T^^
When x=lia
in the

kept at temperature

same way.

Ct',

W,

etc.,

'

(2)

the results can be obtained

THE USE OF CONTOUR INTEGRALS IN THE

212
103.

Rod of Length
and a cos

tures Zero

In this case
satisfied

it is

I.

x=0

The Ends

and 7i=l kept

Initial

wt, respectively.

clear that all the conditions of the

by

at

Tempera-

Temperature Zero.

problem are

fsinaa;
da.

tTrJsinai!

(1)

a*+ft>V/c2

over the path (P) of Fig. 14.*

We may

write (1) in the form

fsin ax

a?

da

27r J sin al

over the path {Q) of Fig. 17.


And, finally, by Cauchy's Theorem,

we have

r
sin
\%tat

sin^

sin

s-

V()<'+*>'
,

cosh

2/iZ

2a.^,

,.

a/(s)(i-,-'-p-*

WTT

cos 2/11 [{cos fi{xl) cosh fj.{x-\-l)

cos in{x+l) cosh iul(x1)} cos wt


{sin (a; sinh (a;+
sin (x+l) sinh (a; Z)}sin wt]
/x

/i

-,

I)

Z)

//

yu

mr

+'-^i,i- -l)"sin -r-x


Z
w*7rVZ*+toV
where

/m==J{u}/2K).

Similar results

may

be obtained for the case when x = l

ia

kept at a

sin

lot,

and when the temperatures at a;=0 and x = l are interchanged.

Rod of Length ;. The End x=0 kept at Zero. Radiation at x=^


into a Medium at Constant Temperature Vo. Initial Temperature Zero.
Here we have to solve the equations
104.

^=0,

SOLUTION OF THE EQUATION OF CONDUCTION

Starting from the solution

wien 33=0, we

by

are led

aA

sin axe-""'* of (1),

to choose

(3)

213

which vanishes

so that

sm al~Vo)=0,

cos al-\-h{A

4=

^^rr
sm al

a cos M-\-h

Introducing the path (P) of Fig. 14,

we

obtain the solution of

our problem in the form


hVg f
sin ax
e-""'*
v=^\
da,
a cos al-\-h sm alf,

iTT

,_,
(5)

the integral being taken over the path (P).


It

is

show that the

easy, as in 36, to

a cos

roots of the equation

aZ+A sin al=0

(6)

and not repeated. They are symand may be denoted by 0, aj, a^, etc.
(5) reduces, by using the path (Q) of

are infinite in number, all real,

metrical about the origin,

The
Fig. 17

solution given in

and Cauchy's Theorem, to

{-^ |2V(

v=hv
"

l^"-

Xl^-hl^^^^

''''''

^>

v/("'+^')

h(l+M) +

^''^'^nX

the summation being taken over the positive roots of


If radiation takes place at
Vi

and

V2,

respectively,

we

a;

(6).

= and x = l into media at constant temperatures

are led to the solution in the form

v = l- [{Ai
I sin ax +

over the path (P), where

H \
,7.
1'-^''

ctn

B cos cue)

do-i

A and B are determined by the equations


-Aa+hB = hvi,
)

A[a

cos al

+ hsin

aJ]

+B[-

sin al

+h cos oWj^hUi.)

of radiation into a medium at Gt, Ct^, etc., or a cos (ot, a sin lot,
can be treated in the same way.
,
For example, if radiation takes place a,t x = l into a medium at temperature
is kept at zero, our solution is given by
Ct, and the end a; =

The case

etc.,

v=

--.

iKir j

a cos al +h sin al a^

over the path (P).

In each case we obtain the value of u in the form of an

infinite series

by

taking the path {Q) and considering the poles of the integrand.

of Length b composed of two different Materials. The


and x=b kept at Zero and Constant Temperature Vo,
Initial Temperature Zero.
respectively.

105.

Ends

As

Bod

x=0

in 99,

we

let

%, K^,

Cj,

pi refer to the

first

part of the rod

THE USE OF CONTOUR INTEGRALS IN THE

214

< <

(0

a;

and

a),

2>

*^^

K2,

c^,

p^ to the second (a

and

Ki=K,!cipi

Then the equations

(2)

Vi=0, when a;=0:

(3)

Vi=0, when <=0,

af='^^a^'

^2=%

0<a;<a:

%=%

a;

we

Also

b).

K2=KJc^p2.

to be solved are as follows

War='^^a^' o<^<^

(4)

< <
:

a<.;<6.

when x=b,

...(D
(2')

^2=0, when =0, a<a;<6.

(3')

when a3=a.

It is clear that

Vi=Ai

sin aa;

V2={A2
where

e""!""*,

sin //a(a; a)+S2 sin

and

/ji=J(Ki/K2), satisfy (1)

They

also satisfy (4)

and

(5),

^tt(6 a;)}e-''i''''',

(!').

provided that

Ai

sin

aa=B2 sin /jLa(ba),

K^Ai

cos

aa=K2ix{A2B2

cosi/xa(6-a)).J
,

Thus we take
A2={cr cos aa+sin aa cot /xa(6 a))^i,

where

sin aa

^~ s,m ij.a{ba)

^'

cr=KJK2iii=J{KiCj^pJK2C2p2).

Introducing the path (P) of Fig. 14, and choosing a suitable value
for ^1,

we

are led to the solutions


o

Vo

e-''"'*

ain ma (x
- [ f
= Vq^
^sin
i-TT

where

sin aa

a)

^g (6 a;) e""'"'*
F{a) amiua(ba)) a

sin aa sin

fjLa{b a)

i''{a)=cr cos

..

"!

"'

'"^

'

aa sin //a(6 a)+sin aa cos iua{ba),

the integrals being taken over the path (P).

The value
__v f

q-

of V2

given in

(7)

reduces to

cos aa sin jua (a; a)

2--J
over the path (P).

+sin aa cos //g (a; a)

j(^j

e""!"'* ,
^a.

(8)

SOLUTION OF THE EQUATION OF CONDUCTION

215

The expressions in (6) and (8) satisfy the differential equations


(1) and (1'), and the conditions at x=0 and x=a given by (2), (4)
and (5).
Further, putting

x=b

in (8),

we have
{"--''"'da,

we know that

over the path (P), and

this is equal to v^.

We shall prove below that the roots


F{a)=a- cos aa

sin

are infinite in number,

of the equation

/xa(6 a)+sin aa cos fji.a(ba)=0

(9)

and not repeated, and it is clear that


an equal and opposite negative root.

all real,

to each positive root there

is

Then using Fig. 15, as before, it will be seen that the values of v^
and v^ given by (6) and (8) satisfy the initial conditions (3) and (3').
Finally, the solution is obtained as an infinite series.
For we have, from (6) and (8),
Va fsin

Vq

ax

e-'i"'*

Ccrcosaasin fxa{x a) sin aa cos fxaix a) e~'''^^'*^


Fja)
"a ^"'

'''~2i^}

over the path (Q) of Fig.

Hence

17.

r-+2V

Vi=Vo\-7j

^^^

j^,,

".

(10

(rC0Saasin^a(a; a)

'''-''''>

f (rfi{x~a)+a

^.^ +sin graces /xa(a;-a)

FM

W(6-a) + a+^4>

the summation being taken over the positive roots of

When

e-'i""''!

/'

^^^>

(9).

the conditions at the surface are of the form discussed in

the previous pages, the method of this section can be appUed with
success to the solution of the problem.
106. It remains to discuss the roots of the equation F(a) = 0.

the

common

These are

roots, if any, of

sint
inaa=0,\
= 0,J
a) =
sin /ia(6>-a)

and the

is

From

o-cotaa + cot/ia(6-a)=0

roots of

The equations
rational and

the curves

'

(2)

have common roots other than zero only if fj.{b - a)ja


iji(b a)/a is small, these values of aa will be large.

(1) will
if

?/=o-cot.r,

y=

(b-a)

catii.-

X,

THE USE OF CONTOUR INTEGRALS IN THE

216
it is

clear that there are

an

position can be determined.

number of real roots of (2), and their


They are symmetrically placed with regard to

infinite

the origin and they are not repeated.


^Further (2) cannot

have a pure imaginary


(T

coth

ari

+ coth

root, m; say, since

-a)

fir]{b

cannot be zero.

We

now show

shall

that there are no imaginary roots of (2) of the

form ^iij.

Consider the function

defined as foUows

<x <,a,

!f7i =sin ax,

ad
,,
V
rsmaaib-x),
U, = ~.
TT
"^
sw. ij.a[b -a)
sin

_,

where a

is

a root oi F{a)

''^^-+/i''2{7a

= 0,

= 0, when a;=0 ;
Ui = Ui, when x=a.

Also

f/i

And
ITurther,

I.

= 0.

^^+a^U^ = 0,

Then we have

^
^
a<x<b,

C/a

0<x<a,

(12)

a<x<b

(13)

= 0, when

j;

= 6.

when x = a,

dx

dz

"

sui;u.a(6-a)

= 0,

since

{^1 cos aa sin fia{b - a)

H-ii'j/i

sin

aa cos /xa(6 - a)]

Ki = a-iLK^.

Let a and fi be two different roots of J'(a) = 0.


Let ?7i, Ui have the values given above, and let Fj, Fa be the corresponding
expressions when /3 is put in the place of a. Then we have from (12) and (13),

^(a^-^^)l''u^V^dx +

(a^-13')

^ f{U,"V^-U.J^")dx =

rUiVidx+

f{,U^"V^- UiVi")dx=0

Therefore
(a^-/?")

(14)

(TfX J a

J,

l^r U,V, dxl


[ fViFi dx +

= fVsF/'C^fJ-Ja

U,"V^)d3;+

fViFi"Jo

But we have seen that

dU,_

dU,
ax
ax
dV,_j. dV,

\
I

^jjgjj

^^^_

^i"Fi)d^

(15)

SOLUTION OF THE EQUATION OF CONDUCTION

217

It follows that

(a'-l3'')U''u^Vida;+!^l''u^V,clx\ =

(16)

It is clear from (16) that a, fi cannot be of the form ^ir],


and Ui, Fa would be conjugate imaginaries and

since

l/j,

Fj

rU^Vidx+M' U^V^dx
would be

positive.

FLOW OF HEAT
107.

Wheu

the

IN

A SPHEEE.

and surface conditions

initial

in a

homogeneous

sphere are such that the isothermal surfaces are concentric spheres,

we have

seen that the equations for the temperature can be reduced

to those for a rod whose length

equal to the radius ( 64). When


is kept at the

is

and the surface

the initial temperature

is

constant temperature

the equations for v are as follows

v^,

zero

v=0, when t=0,


v=Vq, when r=a.
Thus, from

100

we have
Vnttf sin ar e""'

vr=-^

da,

-.
I

iir J sin

aa

over the standard path (P) of Fig. 14.


Also

(^)

+%=

cot aa

= _ M e

"*

I"

2toa
TT

approximately when

?a

''

da, over the path (P),

-^2

rg-..(^^^4M
"

Jo

is large.

e^^"- e-"-"' da

V rcos2miae
X

Jo

'cZa

FLOW OF HEAT IN A SPHERE

218

And when

the initial temperature of the sphere

Vq

is

and the

surface temperature zero, we have the approximation

\drJr=a
Let

<i

and

t^

sli-wKty

be the times required for subsidence to a certain temperature

gradient at the sui'face in the plane problem (28) and this spherical

problem.

Then we have
Therefore

= ix + r.

Let

<2

Then

t=~ Ji'^Kh))

But the gradient

approximately.

of 1 degree in 50 feet adopted

by Kelvin

is

the same as

degree in 2743 cm.

Also a=6-38 x lO^ cm. and =4000


It follows that

:r

= ?:r:*
29

ti

108. Sphere of radius b

r=0

to

r=a

of

C.

composed oi two

one: from

r=a

to

kept at Constant Temperature Vq

r=b

different Materials.

of another.

Initial

Temperature

Surface

From

r=b

Zero.-j-

As in 105, let v^, K^, c^, pi refer to the part from -=0
and V2, K^, Cg, /02 to that from r=a to r=b.
Also let Ki=Ki/cipi and jcj =^2/^2/02Then the equations to be solved are as follows

to r=a,

(2)
(3)

^1=0, when i=0,

V2=Vq,

whenr=6.

0<r<a:

2=0, when f=0, a<.r<h.

(4)

Vi=V2, when r=a.

(5)

K-^

ji = K2-^, when r=a.

*This agrees with Heaviside's

result, loc.

cit.,

p. 14.

Heaviside solved this problem by his "operational method" (ef. loc. cit.,
19), but he has not published his investigation.
See also the papers by

p.

...(3')

Bromwioh and the author {Cambridge,

Proc. Phil. Soc, 20) referred to on p. 201.

FLOW OF HEAT
a putting

219

Vjr=Ui and V2r=U2, these reduce to

W^"^^'

0<'-<:

Mi=0, when <=0,

a<r<b

l/='^^l^'

Mi=0, when r=0:

=6%

u^

0<r<a:

(6'

^^en r=6

M2=0, when =0,

(7'

a<r<6.

Ui=U2, when r=a.

(9)

K^{a-^~u^=K^{a-^~Ui), when

(10)

It

A SPHERE

IN

r=ffi.

clear that

is

Ui=Ai

sin ar

e-*'""*,

%=(^2 sin ^a(r a) + 2 sin //a(6 )) e~"i-'*,


where m =\/('i/f 2)1 satisfy

They

^1

(6)

and

also satisfy (9)

sin

and

(10),

(6').

provided that

aa =^2 sin

//a

(6 a),

KiA^[aacosaa sin aa]=K2[iJiaa{A2 B 2 cos fxa{b a)

J-

B2 sin fjia(b a)].)

Therefore

aaA2BA

aa cos /j.a{b a) -i

sin/xa(6 )
= (T(aa cos aa sin aa) A^,

where

K^=K2na.

Thus we take
cr

cos aa sin /* (6

a) + sin aa cos /naiba)


,

'- sin a sin


//cr

MOO
Sin ^a(o a)

sin aa

(0 a)
ua /L
\

____/(

sm ^a(o a)
Introducing the path (P) of Fig. 14, and choosing a suitable
vralue for A-^

'^

ivhere

we

are led to the solutions

bvg f sin ar

e-"'"'*

ITT J If (a)

,, .

_bVo f/ sin//a(r g) sinaa sin/xa(6 r) \ _-"'"'*,


"'"'
"^~i,r J lsin;ua(6-a)^ -F'(a) sin^a(6-a)J a
F{a) = (T cos aa sin //a (6 a) + sin aa cos //a (6 a)
1

//cr

sin

fxaa
bhe integrals being taken over the path (P).

aa sin /*a(6 a).

,^.

^^^'

(8'

FLOW OF HEAT

220

The value

of
,cr

IN A SPHERE

given in (12) reduces to

u^,

cos aa sin

/xa(ra)+ sin aa cos ij.a{ra)


^-Sin

-^

aa am fxair a)

,,

over the path (P).

The values

and Wg given in (11) and (13) satisfy all the


For, from the way in which they have
been built up, they obviously satisfy (6) and (6'), and the conditions
(7), (9) and (10), which hold when r=0 and ra.
Further, putting r=h in (13), we have
of

conditions of our problem.

V da,
Ja
and we know this is equal
I

iTT

over the path (P),

can be proved just as on


i' (a) =0- cos aa sin ^ua (6 a)

It

p.

+ sin aa cos fia (ba)


H

are infinite in number,

root there

jnaa

sinaasinua(6
a)=0
'^

and not repeated, and to each

all real

this to

be the case,

it

follows

and

For we have, from


bv f sin ar

16, as before,

bva

is

obtained as an infinite

(11)

and

series.,

'

(13),

^"'

cos aa sin fxa (r a)


T

^
2iTrJ
iirJ

I
I

the integrals

initial con-

e-*'"''

"^=2-i7jj>)a
(7

'

(8').

Finally the solution

'

positive

root.

from Fig.

that the values of u^ and u^ in (11) and (13) satisfy the


ditions (8)

(14)

'

an equal and opposite negative

is

Assuming

to bv.

215 that the roots of the equation

now

+ sin aa cos fia (r a)

uaa
F(a)

sin

aa sin ^
ua (r
a)I
^

^,,
g-Ha.t
'

being taken over the path {Q) of Fig. 17.

Hence
1

-P

(a)

an

"2=^0+261)0
trcosaasin^a(r a) + 8inaacos^a(r a)
I

xS

1 MO sinaasinua(r
^'(aj

a)

(15)

FLOW OF HEAT
fcte

A SPHERE

IN

summation being taken over the

221

positive roots of the equa-

tion (14).

When the surface is kept at a temperature Gt, C(\ etc., or Ccoso)*, etc., or
radiation takes place at the surface into a medium ait a constant temperature,
or at one of these just named, the method of this section can be applied with
success to the solution of the problem.

109.

When

the sphere of 108 has

surface kept at zero temperature

its

md the initial temperature of the whole

solid is v^, the

temperatures in the

inner and outer parts follow from 108 (16) and are given respectively
"'

= _ ?^ V

f J"(a)a

cos

o-

"''

by

sin ar e^"'"""'

aa sin

)x,aj^r

W^)

The summation

- a) + sin a^a cos jLia [r - a)

is

a-

taken over the positive roots of the equation J'(a) =

[108(14)].

o-

cos aa cos /xa(6

a)-\

sin

aa cos iJ.a(b - a)

-sinaffisin/xa(6-a) _-^,^h

_o,,,,
~^ " y

/"()

The equation F{a) =

will be satisfied

sin acf

smfj.a(b

by the common

= 0,

roots,

if

any, of

"j

a)=0,j

and by the roots of

^=0

crcotaa + cot/Lia(6-a)H

is

(2)

have common roots other than zero only if /i(6 a)la


is small, these values of aa will be large.
With the data adopted by Perry and Heaviside in c.g.s. units, and the

The equations
rational and
:

ibove notation

(1) will

if /*(&

a)/a

a = 6-38xl08,

b-a=4xl0\

Ki = -47,
Ci/3i

^j,--=

= 2-86,

Ca/32

Vo^^x'W,

-00595,

= -507,

Ki=^ = -1643,

/<2=^=-0117,

,j,=J{K,IK,) = 3-U2,

T=J{(K,c,p,)l(K,c,p,))=.21-l,

Also the equation

(2)

above becomes

211cota' + cot(2'35xl0~'j;)-

20*8

= 0,

where aa=a;.

pr=79.

FLOW OF HEAT

222
The

root

first

roots approach

= 2'9871

the second root

[^\ when r =

In the Age of the Earth problem


1 degree in 50 ft. or 1 degree

cot

a-,a

''^"
2743 =

(cf. 28),

in

'

cot ua,'^(b-a)-]
'^

LiaM

required.

is

cot raai1^(b

-a)/

FLOW OP HEAT

'""'''

This

is

about one-fifth of the result

IN A CIRCULAR CYLINDER.

The method employed


upon

and

problem.

in the preceding

applicable to the case of the circular cylinder,


V depends only

.^l-Mcr

in 99 for the corresponding plane

110.

the hiter

the time of cooling to the

2743 cm.

.
^ ,, ,
,, a)-\
aacos(ic'aja+ij,{b-a)cosec'iJ.ai{o

gives for the time in years 9'02x 10'.

found

= 5'980; and

found* that the equation

o-

.i'2

b.

gradient

It will be

is

Also t is large.
take the first term as a good enough approximation for the

Sir, Air, etc.

Thus we may
gradient

is .ri

A SPHERE

IN

sections

is

also

when the temperature

t.

In this case the equation of conduction becomes


dv

{
dr^

We

shall refer only

Cylinder, as with a

I.

Vo.

very briefly to certain problems for the

little

practice these can be easily solved.

Circular Cylinder of radius


Initial

r dr

a.

Surface at Constant Temperature

Temperature Zero.

It is clear that

,^^[JoMr'' da,
over the standard path (P) of Fig. 14,
of the problem, for the zeros of Jf,(z) are

repeated,! and

when t=0 we can use

(1)

satisfies all

known

the conditions

to be real

Fig. 16 as before.

*Cf. Proc. Gamb. Phil. Soc, 20, p. 404, 1921.


t Cf. Watson, he. cit., 15. 21, 15. 25.

and not

FLOW OF HEAT IN A CIRCULAR CYLINDER


The

223

solution given in (1) can be put in the form

'

2*tJ Jo(aa) a
over the path (Q) of Fig. 17.
And, finally, we obtain

the summation being taken over the positive roots of Jo{aa)==0.


(Cf.

57,

I.)

The Same

II.

Surface Temperature Ct.

Solid.

Here our solution

is

v=.

da,

-o

';

(1)

over the path (P) of Fig. 14.


This reduces to

the summation being as before.


For certain applications* the mean temperature over the cylinder
required,

or

2irrvdr,

-s

This could be obtained from


the latter method,
the

we

vrdr.

a' Jo

Tra^Jo

By

is

i.e.

or more directly, at once from

(2),

(1).

see that

mean temperature =2

f vrdr

integrating over the path (P).

a
f rJ{ar)dr=--J'(aa).

But

Jo

ct

=
Thus the mean temperature
^

2(7
-;

/'e-''
/

-j-

iKira] a*

= C
-.

/e-""'*
I

iKTraJa*

J'(aa)
-^ (da, over the path (P),
</o(aa)

J^{a.a) ,
,,
r "ai over the

-^

J^iaa)

-^L'-s^^+^f
* Cf.

Bromwioh,

Phil.

Mag., London

temperature over a sphere follows from

(Ser. 6),

^*

87,

p.

,,

path
'^

._,

(O),
^

'

J413,

102 in the same way.

1919.

The mean

224
III.

FLOW OF HEAT IN A CIRCULAR CYLINDER


The Same

Solid.

Here our solution

Surface Temperature

C cos ait.

is

over the path (P) of Fig. 14,* and this can be reduced, as before, to

an

infinite series.

Same

IV. The

Radiation at the Surface into a

Solid.

Constant Temperature

Here our solution

Medium

at

v^.

is

,=^ r_

JoM

r-'^^^,

over the path (P) of Fig. 14, which can be reduced to an infinite
series, as before.

The footnote on

p.

203 appliea here

also.

CHAPTER

XII

INTEGEAL EQUATIONS AND THE EQUATION OF


CONDUCTION

111. Introductory.

An integral equation* is one which involves an unknown function


under the sign of integration. The equation
<p{x)

= x[''K{x,i),p{i)di

(1)

J a
is

The function (f>{x) is


The known function K{x, ^) is called the

called a homogeneous integral equation.

the

unknown

function.

hernel (or nucleus) of the equation.

function of

{x,

are concerned.

f) in the region

It will usually be a continuous

a~x^b,

More general conditions

= ^^b

for

with which

we

K{x, ^) are referred

to in the works cited in the preceding footnote.

In the theory of integral equations


continuous function which

zero of a certain function

D (X)

The

D(\) =

roots of the equation

*For a

shown that the only


^{x)=0, when X is not a

it is

satisfies (1) is

asspciated with this equation.


are caUed the characteristic

discussion of the elementary theory of integral equations, reference

may be made to

BScher's Introduction

Tracts in Mathematics, No.

Whittaker

10).

to the

There

Study of Integral Equations (Cambridge


a short chapter on the subject in

is

and Watson's Modern Analysis, and a much

Goursat's Cours

d' Analyse, T.

treatment in

fuller

III.

For the applications, the following works will be found specially useful
Horn, Einfilhrung in die Theorie der partiellen Differentialgleiehungen, Leipzig,
(Sammlung Schubert, LX.)
1910.
Kneser, Die IrUegralgleichungen und ihre Anwendungen in der mathematiachen
Physik, Braxmsohweig, 1911.

Heywood

et Fr^chet, L'Equation de Fredholm

et

ses Applications

la

Physique

Mathematique, Paris, 1912.


Hilbert, Qrundzuge einer allgemeinen Theorie der linearen Integralgleichungen,
Leipzig, 1912.

Vivanti, Elementi della teoria delle equazioni integrali lineari, Milano, 1916.
c.c.H.

225

INTEGRAL EQUATIONS AND

226

When

K{x, f) is a symmetric function


numbers are known to be all real.

numbers of this kernel.


of X, ^, the characteristic

Let

be a continuous function, not identically zero, satisfying

<p{x)

(1) for

a value of X which

D (X)==0.

a root of

is

It is said to be

a characteristic function corresponding to this characteristic number


Also

A.

it is

said to be normalised

if

(p^dx

1.

Ja

It

is

known

that

if

^^(x) corresponds to X^ and

(pnix)

to

another characteristic number X for a symmetric kernel, then


I
Jo

(j>ni<t>n^^

= ^-

Functions for which

<ji{x)\lr{x)dx=0 are said

Jff

to be orthogonal,

and they are normalised orthogonal functions


\{r^(x)dx=l.

0^rfa!=

in addition,
Ja

if,

Ja

In the case of

finite

symmetric kernels there

is

an upper

limit to

number of orthogonal characteristic functions which correspond


the same value of X and every other characteristic function, for

the
to

that value of X,

There

will

is

linearly

dependent on these orthogonal functions.

thus be a set of characteristic functions


cp^ix),

<Pi(x),

finite or infinite in

number

as the case

...

may

be, orthogonal to each

other and normalised, such that every characteristic function of


this

symmetric kernel

them.

Such a system

is
is

linearly

dependent upon a

finite

number

of

spoken of as a complete orthogonal system

of normalised characteristic functions of this kernel.

With regard
Let ^i(x),

to such a system

<j>2,(x),

...

we have the following theorem

be a complete orthogonal system of normalised

characteristic functions for the

homogeneous integral equation with finite

symmetric kernel

rb

<p(x)=\\ K{x,i)<t>{i)di.
Ja

and

Xi, X2,

...

V
is

the corresponding characteristic numbers.

uniformly convergent in the region a

sum

is

K{x,

^) at every point at

If

the series

which

= x^b, a = ^^b,

then

its

the kernel is continuous.

Of this theorem we shall have to make frequent use in the applications of integral equations of this type to the solution of problems
in the conduction of heat.

THE EQUATION OF CONDUCTION


112. Integral Equations and Linear

Consider

now

Flow

227

o Heat.

the temperature problem for the rod of length

the ends kept at zero, the

I,

temperature being the arbitrary

initial

function f(x).

Then we have the equations


3-r-g^.'

0<x<l,

v=0, when x =

and x l,

(1)

v=f{x), when i=0.


Putting v=e~''^*(j>{x),

we have

^+^^-'
0=0, when

a;==0

Thus X = i^7r7?^ and choosing

and x=l.

(2)

so that

(t>^{x)dx

= \,

the

Jo

normalised function 0
Also the functions

is

Ji-jj&irx-^x, n being a positive integer.

are orthogonal, since

0m0<?a;=l, m=l=n.

Jo

Now

there

is

a continuous function K{x,

f),

which

satisfies

the

equation for steady temperature {v"=0), and the same boundary


conditions, while its differential coefficient with regard to x, denoted

by

K'(x,i),

in such

continuous except at a;=^, where

is

away

that

K'{x, ^)

=1.

it is

This function

discontinuous
is

Green's function for the steady temperature equation.*

called the
It

is.

the

steady temperature due to a constant source of a certain strength


at the point x=^, and in the case given above it is clear that
all

the conditions are satisfied

by

K{x,i)=x(l-ill),

=i(i-x/i),

i<x<i.

F{x)^<f,{x)K'{x, i)-<t>'{x)K{x,

Let

where

0<x<i,

(f>{x) is

given by

^^^
'

i),

(2).

The Green's functions employed

in the application of Integral Equations

to the Conduction of Heat must not be confused with those to which the
term was applied in Chapter X.

same

INTEGRAL EQUATIONS AND

228

Then F(x)
tinuous in

discontinuous

is

when x = ^, but

Also

F'{x)

= \K{x,^)<^{x),

except

\F'{x)dx = \\K {x,

Therefore

Jo

But

it is

otherwise con-

(0, I).

F'{x)

when x=^.
(a;)

i)

dx.

"0

={

F'{x)dx+ [ V(a;) dx

"If-"

Jj+o

^F{x)\

since

F{0)^F{1)=0.

It follows that
cj>

ii)

oT'^-K

\k'(x,

Jf+0

Thus

{'k{x, i)<p{x) dx.


Jo

<j>{i)==\\'K(x,i)<j,(x)dx,
Jo

and

since K{x, ^)

is

a sjTmnetric function of
<p{x)

x, f,

we have

= x\'K(x,i)<j>(i)di

(4)

Jo

Thus

the functio7is

<j>

of (2) occur as characteristic functions of

[x)

homogeneous integral equation (4).


The converse is also true. Every continuous solution
integral equation (4) satisfies the equations (2).
the

We

start with

(a;)

=X f

(p(x) of

the

K{x, i)<p{^) d^,

Jo

where K{x, ^)

is

given by

(3).

Since K'(x, ^) is discontinuous when x


^, we cannot differentiate
under the sign of integration, if we rely only upon the theorem

proved in F.S.

78.

But

it is

easy to extend that theorem to such

a case as this by the following method

Let K(x, ) be given by


fix,

Then
*

f{x,

(3)

i)=K(x, i)-=x{l-ill), when x<i


=K(x, ^) + {x^), whena;>^.

i)=x(l ^jl), and

For a disouBsion

:*

above, and let

this function is continuous in (x, ^)'

1902

38, 1904.

under the sign of integration,


and Mess. Math., Cambridge, 31,

of the question of differentiation

see Hardy, Q. J. Math., London, 38, p. 66, 1901,

THE EQUATION OF CONDUCTION


O^x^l, O^^^l.

in the region

Also

is

229

continuous in this

region.

Further,

Jo

Jo

Therefore

-^\[f,i>ii)di

(5)

we cannot differentiate (5) under the sign of integration if


upon the theorem of F.S., 78, since the integrand is
discontinuous when x=^. But we can extend that theorem to
such a case as this in a similar way
Again,

we

rely only

Let

Then

g{x, i)

g{x, i)

and

= K'{x, i) = l-ijl,
= K'{x,i)+\,

= lill

i>'{x)

=\

^"(x)

\'g{x,

= \\

i)^{i)

Finally, since
(j)

defined

K(x,

by

since

^) vanishes

(4) also

5^=0.

when

vanish when

homogeneous integral equation

In

d^-X f%(a di.

A<pi^)di-'\<p{x)

Thus we have established that


the

Jo

= X^{x),
tions

x<i,\

in the given region,

Jo

Therefore

when

when a;>f

x=Q
x=0

and x=l, the funcand x=l.

all the characteristic

functions of

(i) are solutions of the equations (2).

this question the characteristic niunbers are n^tr^jP,

where
complete
number,
and
a
normalised
orthogonal
n is any positive
system of characteristic functions is <j)i{x), (j>i{x), ... where
,

INTEGRAL EQUATIONS AND

230
T>

21

BXLt the series


is

~^

^1
^X ^ sm mr
.

n-TT.

a;

sm-y- f

uniformly convergent in the region Q


It follows

for

K{x,

= x^l,

0^^=l.

from the theorem enunciated at the end

f) as given in (3)

of 107 that

we have

Xix, ^) = f]in(^lMil
A

(6)

113.

flow of

The

real difi&culty in the solution of the

problem

of linear

112 consists in establishing the possibility of the expansion

of the arbitrary function in the appropriate series.

in (6) above, it

equal to the

From the result

can be shown that the arbitrary function f{x) is


of an infinite series of these characteristic

sum

functions, provided that

f(x)=\'K(x,i)^ii)di
Jo

and i/r(a;) is bounded and integrable


For we are given that

IT

and

in (0, l).*

O^x^l,

this series is uniformly convergent in the region

It follows that

fix)

[^

Jo

Also

^"('^)^"(^)

y.

\j(r,).pMdn^f^ <t>M

^{i)di

K{n,i)i.(i)di\d,

=\l^{i)[\[K{r,,i)ci>Mdn'\di
(F.S.,179)

=\[^ii)[\lK(i,r,)<j>{r,)d,'^di
{K(^,

n)

being symmetric)

An Jo
*It f{x)
continuous.

satisfies this condition,

it

follows from

F.8. 77,

II.

that

it

is

THE EQUATION OF CONDUCTION


Thus when. /(a;) can be put

and

i/r(a;) is

in the

bounded and integrable

.fip)

= ^(i'n<l>n{x),

231

form

in

(0, 1),

we have

where a=\f{x')cl>n{x')dx'.
Jo

Also the solution of the temperature problem of

112

is

v=^a,<j>^{x)e-^''K
1

The other problems


the same way.
114.

I.

Suppose

place at

x=l

the

into

for the rod of length

x=Q

end

a medium

can be treated in

kept at zero and that radiation takes

at zero.

Then we have the equations

=0,

when x=0,

dv
-^-\-hv=0,

when x=l,

'

v=f(x),

Putting

v=e-''^^(j){x),

when <=0.

we have

with the same conditions when

The Green's function K{x,

x=0 and xl.

^)

is

as follows:

K(x,i)=(l-^^x, 0<x<i,

Also the characteristic numbers are X=a^, where a

is

any

positive

root of h tan al-\-a0.

The normalised

characteristic functiojis are

^(^)=;7^PUp!l)i^^The

series

V
1

}9n\e)
^n

^-^j^

-^^

(Cf.

65 and Ex.

2, p. 182.)

found to be uniformly con-

INTEGRAL EQUATIONS AND

232

vergent in the region

O^x^l, O^^^l, and we


"-n

Suppose radiation takes place

II.

have

at both ends

into

a medium

at zero.

Then the conditions

at the ends are

-= \-hv=0, when x=0,


OX
-

u^^=0,

when x=l.

dx

In this case

(l+h^)(l+h{l-x))
2h+lh^
Also X

= a^,

where a

The normalised
0(a:)=^/(

is

any positive root

i<x<i.
of

{a^ h^) tan a.l=2ah.

characteristic functions are

J ^" ^^^ a;+A sin

(^^2^^2)
:i^+h^)l+2hJ
;+2

(Cf. 36.)

ax).

Also

In these two cases the possibility of the expansion of the arbitrary


function in the required series,

when

it is

subject to the condition

stated at the beginning of 113, follows in the

same way.

The

solution of the temperature problem can then be written down.

When

h=

this discussion fails.

For a treatment

no heat escapes at the ends, reference


above, Ch.

I., p.

may

of the

problem when

be made to Kneser's work cited

19.

115. Fourier's Bing.

For the ring

when there is radiation


we have the equations

of unit radius, treated in 12,

at the surface into a

medium

dv
dt

at zero,

dh)
zr-hH,
" dx^

-7r<a;<7r,

=Vx = .
(1)

Kdx/^.

\dxJ.

v=f(x), when i=0,


where- 6

is

a given number.

THE EQUATION OF CONDUCTION


Putting ;=e-(f +'"')

^(o;),

233

we have

S+^^=0'
with the same conditions when x= and x=ir.

(2)

Tr

Thus we have ;u=0 and /u=w2, and the corresponding normalised


functions are

,.

and r- cos

It will be noticed that to the

Now

there

is

a continuous function K{x, ^), which

conditions at a;=
is

tt

and

tt,

while

satisfies

the

(Kv"bh=0), and the same

its differential coefficient

continuous except at x=^, where

way that

there correspond two

- sin nx.

steady temperature

for

two.

number n^

orthogonal functions -j- cos nx and

equation

-r- sin

inx,

it is

K'{x, ^)

discontinuous in such a

=1.

\k'(x, ^)1

the Green's Function for the equation v"c^=0,


as follows

This function

where c^=h^lK

is

Tzi

coshc(a; ^+7r)

A^

^^("'^)=

2sinhcl

<-<f=|
^

'

j,

^coshc(-^+^
b ^
2smhc7r
follows as in 112 that, when

'

Also

it

equations

<l>{x)

is

given by the

(2),

<p{x)=\\'' K{x,i)<j,{i)di,

(4)

J IT

where X=yu4-c^

And

all

equations

the characteristic functions of

(4)

are solutions of the

(2).

The special feature of this question is that, to each characteristic


munber other than \-=c\ there correspond two orthogonal characteristic

functions
e.g.

'\=c^-{-n^ has

Thus the theorem at the end


,

- coswa; and
of

-r-sinwa;.

110 in this case leads to

^ cos

^ cosw(a;

Jia;

cos w^+,sin m; sin

^)

M^

INTEGRAL EQUATIONS AND

234

provided that this

series is

TT^cc^TT, tt^^^tt;

uniformly convergent in the region

a condition which

clearly satisfied.

is

The remarks in the previous section about the possibility of


expanding an arbitrary function in the appropriate series apply
also to this case.
Thus we obtain the Fourier's Series for f{x),
under the condition stated above, for the interval ( t, tt) and the
solution of the problem of Fourier's ring.
116. Two-Dimensional Problems.

The

two and
by
a Green's function similar to that which is

solution of the general problems of conduction in

three dimensions can be

the introduction of

made

to depend

used in the theory of potential

work

is

upon

integral equations

but the rigorous treatment of this

harder than that in the previous sections, since the kernel

of the integral equation has

integration takes
integrals,

place,

and the

series

an

infinity in the region in

Space does not

are double or triple.

permit of more than a slight sketch of the method


first

the two-dimensional case where the boundary

Here we have to solve the equations

Jt='\d^^+dr^'

kept at zero.

is

on the curve G,

we put

and we take

v=f{x), when <=0, inside the curve


before,

msideacurveC,

v=0,

As

which

the integrals being surface or volume

(1)
I

CJ

and we have

v=e~"^*0(a;, y),

9-^+1^+^^=^'^^'^"^'
0=0, on
The Green's function
is

G.

(2)
j

which we shall denote by K(x,y;

x', y')

the solution of the equation 3-2+^-2"*^' which vanishes on the

curve

C and

is finite

and continuous, as

also its first

and second
where

differential coefficients, inside G, except at the point {x', y'),


it

becomes

infinite as

(logr)/27r

steady temperature at the point

when

[x,

of a certain strength at the point

y)

r->-0.

This function

due to a constant

(a;',

y'),

is

the

line source

the boundary curve

being kept at zero.

We now

apply Green's Theorem in two dimensions to the region

THE EQUATION OF CONDUCTION


bounded by the curve C and a small circle
(x', y'), tke functions employed being K{x,
y
(t>{x,

235

V whose centre
x', y')

is

at

and a function

y) satisfying the equations (2).

Then we have
\\

{cl>V^K-KV^)dxdy=^{^

H-^f^) *'

(3)

where =- denotes differentiation along the normal drawn outward

from the region of integration, and the integrals on the right are
taken roimd the curve
and the circle V.
But V^K=0 and V2^+X^=0 in the region through which
integration takes place

while K{x, y;
point {x', y').

Thus from

and

(3),

0,

x', y') is infinite

as

both vanish on the curve

(logr)/27r,

on letting the radius

we have

when r^O,

of the circle

at the

V tend to zero,

pC

<t>{x',y')=M\K{x,y; x',y')4,{x,y)dxdy,
the double integral

now

(4)

being taken through the region bounded

byC.

Theorem in two dimensions to


by the curve C and small -circles V, T" with

similar application of Green's

the region bounded


centres at

K(x, y;

{x',

x',y')

y') and {x", y"), the functions employed being


and K{x, y\ x", y"), shows that

K(x',

y';

x", y")

= K(x",

y";

Thus the Green's function K{x,y; x, y)


of the two pairs {x, y) and (x', y').
It follows from (4) that
0(a;, ii/)=A
j

[7^(0:;, 2/;

is

x', y').

a symmetric function

x', y')4>{x',

y')dx'dy',

(5)

the double integral being taken through the region bounded by C,

and [x, y) being any point in this region.


Thus the numbers A and the fimctions
of equations (2) enter
as the characteristic numbers and characteristic functions of the
homogeneous integral equation (5) with a symmetric kernel.
Also the converse is true. The theory of potential shows that
when {x, y) is defined by (5), we have
f/>

(j)

and

4>

=0

on

G.

INTEGRAL EQUATIONS AND

236

The characteristic functions corresponding to different characternumbers are orthogonal, and they are to be normalised by

istic

arranging that

rr

\\<p^{x,y)dxdy=l,
the integral being taken through the given region.

The question
the

initial

functions,

istic

of the

expansion of the arbitrary function defining

temperature in a complete

series of

orthogonal character-

and the corresponding expansion

of the Green's

function, ofEer greater difficulty than in the case of one-dimensional

For a full discussion of these topics reference must be


made to the works dealing with integral equations.
The argument of this section appUes equally well to the case
problems.

when

117.

by ^+A=0, both

problem and the definition


(i)

at zero, the condition at

dv

the boimdary being replaced


of the

medium

radiation takes place into a

in the statement

of the Green's function.

Consider the temperature problem for the rectangle


.

x = bA

= 0,

y = 0,

i/

= sin -p

,v

= c,j

the sides kept at zero.

Here we have

<^

where m, n are positive

sin

(7Th

71,

be given by

<^ will

i/,

and

integers,

The normalised function

mir

rnr

if(.,.

y;

"

.,/,y) = -- E S'

assuming that the general theorem of


(ii)

For the cylinder

= a,

d)=r-

cl>,Ur,

6)=

"

107 applies to this case,

cos

y\h
and'

'

the surface kept at zero,


J

4>m.n{i;

mir
nmimr
mr
..i-T-.rsm
^ sin -r- a; sin

Also

'^o('^.or)

7,
''''^''^(".

we have

')*)
_

when n

> 0,

THE EQUATION OF CONDUCTION


The

numbers A, being a^

characteristic

where

a, is

237
the with positive

root of the equation J{aa) = 0.

Also

K(^,

^',y) =

^i

^.(a..o>-)^.(a.,o/)
.

For a fuller discussion of these and other similar problems reference


be made to Kneser's book, cited above.

may

118. Thiee-Dimensional Problems.

The work

of

116 can be extended to three dimensions by using

Vhi=0, which vanishes on the


and continuous, as also its first and

as Green's function the solution of

surface of the soUd,

second differential
(x,

y', z'),

and

is finite

coefficients, inside the solid,

where it becomes infinite as

l/47rr

except at the point

when r-*0. This function

the steady temperature due to a constant source of a certain

is

strength at the point

[x', y' , z'),

the surface of the sohd being kept

at temperature zero.

-^=kVH

The equation
is

V^^ + A^=0

reduced to

by

the substitution

v=e-''''*<p(x, y, z).

And we have
(j>{x,

y,

z)=\ \S\k(x,

y, z\

x\

the integral being taken through the

y',

z')^{x', y', z')dx'dy'dz',

solid.

EXAMPLES ON THE CONDUCTION OP HEAT.*


An

bounded by the plane x=0, across which it radiates


The solid is initially heated throughout to the uniform
temperature v. Find the temperature at any point at any subsequent time,
and prove that at the time the temperature at the surface is given by
1.

into a

infinite solid is

medium at zero.

^ re-=-2;m^/(/t0dM.
JiiTJO
2. If

V,

the temperature of an infinite solid has different uniform values

on opposite

the temperature

is

sides of a given plane, prove that at

any subsequent time

given by the expression,


X

X being measured from the plane towards the side where the temperature was
initially F.

A uniform

its length at two different


Find the temperature of any point of it at any
subsequent time, the surface being so protected that there is no gain or loss
For example, take an iron bar 50 cm. long. The
of heat from without.

3.

temperatures

Vq

bar

given with the two halves of

is

and

v^.

is -16 (water being standard subthermal capacity per unit volume is -875. Prove that at
1400 seconds from the beginning the temperature at either end will be

thermal conductivity of iron (c.G.s, units)


stance)

4.

zero.

and

its

bar of length
If initially the

is

heated so that

temperature
V

is

?{<

jj'

two ends are at the temperature

= cx(l -x)
P
^

show that the temperature at the time

Sce-H(
=
^e

its

given by

TTx

'

at

any point

--j!2~

sm-p + -e

'

is

given by

Sttx
sm-p
+
.

These problems are mostly taken from the Examination Papers for the CamSome of them have already been published in
Turner's Examples on Heat and Electricity.
*

bridge Mathematical Tripos.

238

EXAMPLES ON THE CONDUCTION OF HEAT


5.

One end

an

of

infinite

rod

is

there being surface radiation.

239

kept for a long time at temperature

v,

part whose extremities are distant b and

b+l from this end is then cut from the rod and kept from loss or gain of heat.
Show that the temperature at time t at a point distance x from the end of the
part

is
I

where a

is

One

-eai

,,

?, 1

-COS 7re ' "n-cosnirx

a quantity depending upon the material of the rod.

x=c

infinite slab is kept at temperature zero.


The
=0, is kept up to the time t at pt, p being a conAfter the time t it is kept at a constant temperature; -Find expressions
stant.
for the temperature at any time, and show that if i is so great that e-^(t-r)
6.

face

of

an

temperature of the other,

may be

neglected, the total quantity of heat which has passed across unit area

of the surface

where

7.

s is

a;

up

to the time

is

uniform cylindrical bar, of length

is

the conductivity, and

and small

cross section, is kept

the capacity for heat per unit volume,

at a constant temperature v^ at one end and placed in a

medium

at temperature

temperature at a distance x from the end in the steady state


is v^^"', prove that half the radius of the bar and the ratio of the conductivity
to the emissivity are each equal to a-l.
zero.

If the

When the steady state is attained the sides of the bar are coated with an
adiathermanous substance, but its further end is left unaltered, the nearer
end being still kept at the temperature iv Prove that the distribution of
temperature at the time t is
w = 'o +

2 Om sin mxe<^'^,

EXAMPLES ON THE CONDUCTION OF HEAT

240

values so long that the distrihution of temperature in the solid

is

purely

periodic

X =0

{i)

(iii)

(iv)

10.

An

a,t

=a +b sin pt

x=la,tzeTO.

x=0 a,tv =a+b sin pt: x=l, impervious to heat.


x=0 &ndx =1 a.tv =a+b sin. pt.
x=0 a,t V =a+b sin pt: x=l aX v=a -b sin pt.*

(ii)

infinite isotropic solid is

across that plane into a

medium

after a time so great that

all

bounded by an

infinite

plane and radiates

A cos(A* +;8).

at temperature

Prove that

traces of the initial distribution of temperature

throughout the solid have disappeared, the temperature at a distance x from


the boundary

is

-^^^-^^^,-^oos(Xt-^x+l3-.),

e=tan-^^,

where

^^ = 2^,

and h, k have the usual meanings.


Find the corresponding formula when the temperature
11.

uniform rod of length

I,

of the

medium iaf{t).

cross-section 8, perimeter p, conductivity

K and emissivity H,

capacity for heat 0, density D, and electrical resistance


B, is placed in a medium at temperature zero, and has one end heated to
temperature ^o> the other end being kept at zero until the temperature is

An

steady.

electrical current of strength

is

now

passed along the rod

from the cold to the hot end. Show that when the temperature has again
become steady, the rise of temperature due to the current at a point distant
X from the cold end is

cosh^gg-

PBS,

lo-e,

(^^^^)%inhz(g/'
where

cr is

the electrical conductivity of heat and

Two

uniform plates of the

o- ^j- is

neglected.

UiX

12.

and one

slips

same substance and thickness a are in

over the other with constant velocity

v,

contact,

the friction per unit

The outside surface of one is impervious to heat, and that of


kept at zero temperature. Show that at any time t their temperatures at a distance x from the impervious surface are given by
area being F.
the other

is

S^o.. ,,e"'
e=|^(a
^(a + vj2+l

i5^'^cos(2n
cosf2M-l-nl
+ l)^)
IB"'"^^'

(2n-H)V'C't

Wo,/

e=^{2a-x + 2A2+ie
where

J is

is

the conductivity, c

is

le-'^'

cos(2+l)|^j,

the thermal capacity per unit volume, and

the mechanical equivalent of heat.

* Cf. Kirsch,

Die Bewegung der Wiirme in den Cylinderwandungen der Dampf-

machine, p. 68, Leipzig, 1886.

EXAMPLES ON THE CONDUCTION OP HEAT

241

13. An infinite homogeneous slab whose bounding planes are x=a is


placed between two media, one beyond the plane x= -a, a,t which the temperature is v^-mH, and the other beyond the plane x=+a, a,t which the
temperature is zero. Show that if the ratio of the emissivity to the con-

ductivity

is

slab at time

n tan
t

is

What would
of the
14.

and k has the usual meaning, the temperature within the

/3,

given by

= v^e-i^t sin p coseo Zift-na) cos {n(x -a)+ fi).

be the temperature at any point of the slab if the temperatures


sides of the slab were v^-i^oH and v^e- '!'* ?

media at the two

A sohd is bounded by two infinite parallel planes.

the radiation from

its surfaces,

Taking into account

show that the temperature at any

internal

point wiU be given by

2^e-'W* cos(^^^^)^^i^^).
I

A=tan-i

where
l^

A'=tan-i

V are the emissivities of the two faces, supposed unequal, and 2a

The

thickness of the solid.

origin

is

the

taken midway between the faces, and

is

m is determined by the equation


ma= WTT + 2A + X'
n being any
15.

'

integer.

ring of uniform small section so coated as to have everywhere the


is made partly of brass and partly of iron, and one of the
kept at a constant temperature, while the whole cools in air.

same emissivity
junctions

is

Determine the ratio of the lengths of the two parts when the coolest place
is

at the other junction.


16.

thin circular ring whose surface

by a continuous source
time t is given by

is

of heat of strength Q.

impermeable to heat is heated


Show that the temperature at

12AK TrAKfn^
4.TAK
= the conductivity,

2TcAa-

where

if
cr

=the thermal capacity per unit volume,

A =the
c

s =an
(f)

and

A,cr

radius,

arc of the ring measured from the heated spot,

=the angle subtended by

* Cf. Niven,

CCH.

cross section of ring,

mean

=its

s at the centre,

5-.*

Lmdon,

Proc. B. Soc. (A), 76, p. 42, 1905.

EXAMPLES ON THE CONDUCTION OF HEAT

242

Any

17.

which makes an angle 6 with a fixed


is maintained at temperature
Show that when the motion of heat is steady

point, the radius through

radius of the edge of a circular disc of radius a,

f(d), where /(tt

+ 6)= -f(d).

the temperature at

(r,

d)\s
f%//,

arfa" -r^) cos <i;

ji

,,

the disc being supposed not to radiate heat.

uj,

be kept at the temperature v-i, and the circumprove that the temperature at any point is

"i

2,

18. If the diameter of

ference at temperature

circle

is

'

ir~

and that the isothermal


diameter which

2ay
- "i)s^tan~i
"
x'+y'-a"
^

{'"2

"I

lines

are circles passing through the ends of the

taken as axis of

x,

a being the radius of the

circle.

two directions and is bounded by two


meet at right angles in A. Both edges are at temperature zero, except a portion j4B of one edge, which is kept at temperature unity.
Prove that the temperature at any point P is
19.

plate extends to infinity in

straight edges which

-tan-'f^)

or

1{lAPB -lAPC),

where a^, b^ are the semi-axes of the ellipse and a^,


can be drawn through P, having A as centre and

BA

ies in

b^ of

the h3rperbola, which

as focus,

and where

AG =AB.

produced, so that

a sector of radius a and angle a the radii be maintained at the


Vi and the circumference at the temperature v^, verify that the
temperature at any point of the section wiU be
20. If in

temperature

-"'-"^

g^Han-

\ .?^=tan-4

\2arsin^y

!!r!!!Z^
,T_<,a

plane area is bounded by a semi-ellipse and its axis major. The


boundary is maintained at the uniform temperature unity and the
straight boundary at the temperature zero.
Prove that the temperature
at any point within the area is
21.

elliptic

4/sinh</>
^

-(

TrVsmha

IsmhSA
\
.
smg^ +o
o sm3g+... /'
),
3smh3a
.

where

c cosh <^, c sinh <^ ; c cos 6, c sin d are the semi-axes of the ellipse and
hyperbola through the given point, confocal with the boundary, and a is
the value of <j> at the boundary.

22.

circular cylinder of infinite length

is

divided into four compart-

ments by planes through the axes at right angles to each other.

Measuring

EXAMPLES ON THE CONDUCTION OF HEAT

243

6 from one of the planes, the temperatures cff the successive quadrants of
the surface are maintained at the respective values

r sin
is

temperature at a point
the unit of length, prove that
TTV

T sin 6,

Toosd,

61,

If V is the

inside,

Tcoad.

and

if

the radius of the cylinder

TTT

jT=-2-(sine+cose)+r(oos^-sine) -r'(cos30 +sin3e)

+^(cos5e-sin5e)-^(cos7e+sin7e)
+ ^{eos9e -sin9e)

-~ (cos 11(9 +sinll(9)

+ etc.
23. If

slit

be

made

and extending

origin

in the plane along the line 6 =0,

commencing at the
and if both sides

indefinitely in the positive direction,

of the sHt be maintained at zero temperature during the diffusion, prove that

'

4^KtJo
is
t

I.

the temperature at time

=0

at

due to a

generated at

line source of strength q

(r', 6').

Show

also

how

to obtain the corresponding expression for the portion of

the infinite plane bounded

by two straight edges inohned at an angle 2ir/(2ro + 1),

m being integral, the edges being both maintained at zero.


24. A conducting sphere initially at zero temperature has
at a constant temperature c for a given time, after

Find the temperature at any time in the second

which

its

it is

surface kept
kept at zero.

stage.

v^ is surrounded by an
and of initial temperature
zero.
Prove that the temperature at distance r from the centre of the sphere
at the time t is given by

25.

infinite

sphere of radius a with

medium

initial

same material

of the

^ttLJ r~a

temperature

as the sphere

2V(*)

26. A uniform sphere of radius a is at a uniform temperature v^, and is


surrounded by a spherical shell of thickness a at zero. The whole is left to
cool in a medium at zero. Prove that

4
^cr

where the values of

<t

sin era - acr cos

ua
iaa - sin icra

are given

sin

a-r

by
2a-a

tan2o-o=:;

l-2fta

Also consider the case

when the two substances have different conductivities.

EXAMPLES ON THE CONDUCTION OF HEAT

244

27. A homogeneous solid bounded by two concentric spheres of radii a


and 2a respectively has its inner surface coated with a layer of a substance
impervious to heat. The solid is raised to the temperature Vf, and left to

cool in a

medium

v^^A^e-'^H

Prove that
where A

is

at zero temperature.

a root of
- 2ah

(1

Show how A\ may be

A sphere

28.

A cos A(r -a) +sin A(r -a)

+ 2\'^a^)s\n\a = (\ +2ah)kaaos.\a.

found.

of radius c

is

symmetrically heated so that

ture at a distance r from the centre is/(r). It

medium

is

its initial

tempera-

then allowed to cool by radiation

Prove that if the sphere is very small so that powers


can be neglected in comparison with unity, except just at
the beginning of the coohng, the temperature becomes approximately prointo a
of ^c

above the

at zero.
first

_iiM.

portional to

sin

hr
29.
r

The

initial

from the centre

nt of a sphere of radius
temperature at a point

sinh

and the sphere

is

c at

a distance

is

(-3
I

Kr

surrounded by a medium at temperature zero.

ratio of the emissivity

and the conductivity and

Aj, A^,

...

If

is

the

are the roots of

the equation

cA cos cA +(Ac -

1) sin

cX 0,

prove that the subsequent temperature at the time


equal to

2(^6-1)

within the sphere

is

| ^^_,_,,^,, sin A^rsin A^c


X^'c^+hcihc-iy

30.

sphere of radius c has initial temperature

l-l

2r

and radiation takes place at

A^

is

its

surface into a

medium

at zero.

Show

that

if

a root of the equation

ck cos cA + (fee the temperature at the time

is

1) sin

cA =0,

given by

4l|sinA,. (^-^)^>;;-A. ^_,,,.


where

P=c^k^' +{hc~iy,
Q=c^X^'+hc{hc-l).

The

temperature at any point of a sphere exposed in an


medium at temperature zero is given by
31.

initial

1^,

irr

Vc sin~T-e
2c

1-;'^>

infinite

EXAMPLES ON THE CONDUCTION OF HEAT


c being the radius
if

and

r the distance of the point

from the

centre.

245

Show that,

a root of the equation

is

cX cos cA +{hc -

1) sin

cX =0,

the temperature at the same point after a time t is


"
1
-25^6""'^'"'' sin A^r,

Vm^iX^H^ + (he - \Y}\yic -\)[\ +^') - X^H^hc- 3)

+2eX^c^{X^'^o^ +{hc -

(AV+7ic{Ac-l)){l+(|-A^cy}{l+(| + A^c)'}
33.

ture

A homogeneous

that the temperature at time

equation

(ah - 2) (ma cot

the external
33.

is

heated so that the

initial

tempera-

=1 (cos 6 sin mr - mr cos mr).

is o

Show

sphere of radius a

medium being

is Wje -""*,

??ia

where

ro is

a root of the

1) =m,V,

at zero temperature.

solid globe of metal, radius a, conductivity k^, thermal capacity per

Cj, is surrounded by one of another metal, outer radius 6, conand capacity k^ and Cj, the whole radiating into a medium at zero.
Prove that, if at any time the temperatures of the two metals are represented by
sinmr

unit volume
ductivity

u= mr
v=

sinmacosnir -a)

Ki
-t

mr

Ko

/ki

\k^

m coswiasin.{r-a)
mn

,'\ sin?raasinw(r

viTMr

-a)
'

m-^=n''=0,

where

the temperatures at any subsequent time will be

and determine the equation connecting n with the emissivity

of the outer

metal.

surrounded by a concentric non-conducting spherical


is filled with hot liquid. kept at a uniform
that if v is the temperature at a point in
Prove
agitation.
by
temperature
the sphere, and v', the temperature of the Uquid at any time, then v, v' are
34,

surface,

solid sphere is

and the space between

of the forms

v=c+2A
.

sin nr

,.

6-"^%

m'

siana

,,

EXAMPLES ON THE CONDUCTION OF HEAT

246
where

and n

-^r- is
is

the ratio of the heat capacity of the sphere to that of the liquid,

any root

of the equation

-( 1 +ha
35. In

an

and

tan xa.

conductor made of uniform material an instantaneous

infinite

spherical source of strength

radius a

-^,

is

generated over the surface of a sphere of

to diffuse through the conductor.

left

Prove that

generated over equal elements of the spherical surface

is

if

the heat

everj'Tvhere the

same, the subsequent temperature at a point distant r from the centre


of the sphere is

Q
^^

1
c

TT^'^^

sinh

dS

36. If over everj' element

[x, y, i)

dS

an

of a spherical surface in

being a soUd spherical harmonic of degree n ;


of the element

infinite solid

Y{x', y', z')dS of heat are generated (Y(x', y',

sources of strength

point

or
2k<

J{TrKt) 4:irar

x', y', z'

z')

being the coordinates

referred to axes through the centre), prove that, at the

distant r from the centre at time


a^"+-

-^

t,

2^^^Kt)"n

/smhu\

d"

It

ar

u = ^-.

where

The surface temperature of a sphere of thermometric conductivity k


made to vary according to the law v =S cos ct, where S^ is a surface har-

37.
is

monic of degree
the interior

is

Prove that the consequent fluctuation of temperature in

to.

given by the formula

^ [P{r)P{a) +Q(r)Q{a)]ooB(rt+[P{r)Q{a)

-Q(r)P{a)-]am,7t /rY

\aj

where

P(r)

&

(St

=1
2

2to

+3

2?i

+5

8. 2to

+3

...

2to

+9

2.2ra+3

Examine and
large

38.

2.4.6.2to+3...2)+7

interpret the forms which the result assumes (1)

compared with

^,

and

(2)

when n

is

large

when

ct

is

compared with cr

A homogeneous soMd sphere has half its surface, viz. the portion bounded

by a great

circle,

maintained at temperature unity, while the other half is


Find the steady motion of heat. Show also that the

maintained at zero.

mean

of the temperatures at

distant from the centre

is

two points on the same diameter and equally

the same for

all

diameters.

EXAMPLES ON THE CONDUCTION OP HEAT

247

Let the sphere while in

this state of temperature be enclosed in a closely


envelope impervious to heat and left to itself. Show how to find the
temperature at any point after any time t, and when t is very great, sho\v
fitting

that the temperature at a point whose coordinates are


Al
i+A(arooa(rr
^

^e

,C0SS
-siacrr)

(r,

6) is

Ktr't
""%

where cr is a constant depending on the radius of the sphere.


determine this constant.
39.

line source of strength

is

approximately

Show how

to

instantaneously generated along the

an infinitely long circular cylinder at the time t=0. The temperature V was everywhere previously zero, and the temperature of the boundary
r =a is maintained at zero.
Prove that at any subsequent instant
axis of

where k

is

the thermometrio conductivity, and the quantities mj are the

positive roots of the equation

Jo{m,a) =0.

Prove that when a

is

made

infinite the

above expression assumes the form

and by comparison with an independent solution


definite integral.

of the problem, evaluate the

APPENDIX

NOTE ON BESSEL'S FUNCTION.


1.

The

Bessel's function J{z) is given

T,.s'
"^

and
is

is

thus defined for

used.

all

when n

J(z),

is

by the equation

(-ir(^/2)"+^'-

~rioT{r + l)T{n + r + iy
values of n,

when

the general

Gamma

Function*

not an integer, will be many- valued, but

it

will

be made single-valued by restricting the complex variable to a complete


e.g. by taking |argj| <Tr.
For other values of the argument we use the equation

revolution

2.

When n

not an integer, J{z) and J-{z) are independent solutions

is

of Bessel's equation, but J(0)=(-1)"J^(2;),

For a second

"^

the limit being taken


It is

when n

solution, available for all values of n,

knownf

'

when n

sin
is

integral.

we choose

mr

an integer.

that with this definition Y^iz)

that Y(z) and Y(z) are given

is

is infinite

when

= 0, and

by

rrY{z)^2J,{z){log(zl2)+y) + {zl2y-llJ^(zl2y + 1+^+11^ (ZI2Y-...

7rY{z)=JJ,z){2log(zl2) + 2y-

The

Tmr^- E

mr^] - "s

{n-r--l)\
(z/2)-"+'-

Bessel's function of the second

kind defined above and denoted by


used by Weber.t It is taken as the standard function of
the second kind by Nielsen, but he uses tlie notation Y"(z) and calls it
Neumann's function. As a matter of fact, Neumann (K.) in his Theorie der
3.

Y{z) was

first

Bessel'schen Functionen (Leipzig, 1867)

was concerned only with

positive

Whittaker and Watson, loc. cil. (3rd Ed.), Oh. XII.


jWhittaker and Watson, loc. cit. {3rd Ed.), p. 372;

3. 51(3), 3. 52(3).

XMath. Ann.,

Leipzig,

Nielsen, loc.

cil., p.

6, p.

148, 1873.

10.

248

Watson,

loc.

cit.,

APPENDIX
integral values of n,

the notation Y"{z),

where 7

is

is

249

and

his function of the second kind, for which he used


connected with our r(2) by the relation

(1+5 + 5 +

Lt

Euler's Constant,* namely

logra).

.-H

The function we denote by Y^^z) is the same as Gray and Mathews'


Y(2).t It should be noticed that Schafheitlin's Y(z) is minus ours J
also that Gray and Mathews' Y(z) is Neumann's function Y'^{z).
:

In

4.

many

questions the functions denoted by ff'"(z) and H^^^'iz) are

They are defined by the simple

very u3eful.

Hr:"{z)=J(z) + iY(z),

They may be described as


must find a permanent place
of the simple

and

relations

H^^{z)=J{z)-iY{z).

Bessel's functions of the third kind,

formulae for |arg 2|

<w

approximately in the upper part of the 2-plane, when

l^l

approximately in the lower part of the 2-plane, when

|s| is

5.

Since, in the

and they

in the treatment of Bessel's functions because

upper part of the 2-plane when

|2|

is lai'ge

large.

is large,

||

we have the

approximation **
"^

'

^/(2ir2)

it follows that at infinity in the upper part of the 2-plane J^ (2) is infinite,
H^^\z) vanishes, and J {az)Hj^^\bz) vanishes, when a and 6 are real and

positive,

a<b.

*Cf. Whittaker and Watson,

loc. cit.

(3rd Ed.), p. 235.

Gray and Mathews,

Schaf heitlin, Theorie der BesseUchen FunMimien,

loc, cit., p. 64, (131).

Nielsen, loc. cit., p. 16.


II

Of.

Watson,

Whittaker and Watson,

loc. cit., 3.

loc. cit.

p. 44, Leipzig, 1908.


(1).

(3d Ed.), pp. 368, 371

Watson,

loc. cit.,

7.2(1).
** Whittaker and Watson,

7.21(1).

loc.

cit.,

(3rd Ed.),

p.

368; Watson,

loc.

cit.,

APPENDIX

II

BIBLIOGRAPHY*
Thu Conduction of Heat
TREATISES.
Theorie analytique de la chaleur. Paris, 1822.
English translation (Freeman), Cambridge, 1872. German trans-

FouBiER.

lation (Weinstein), Berlin, 1884.

PoissoN. Theorie mathematique de la chaleur. Paris, 1835. Supplement, Paris, 1837.


Lam^. Legons sur les fonotions inverses des transoendantes et les
surfaces isothermes.

RiBMANN.

Paris, 1857.

Partielle Differentia] gleichimgen

und deren Anwendungen

[See
1869.
also
physikalisohen Fragen. Braunschweig,
Weber, Die partidkn Differenticd-gleichungen der mathematischen Physilc,
This work is also luiown as the
(2 Aufl.), Braunschweig, 1910-12.
It is cited
fifth edition of Riemann's lectures referred to above.

auf

in the text as

Weber-Riemann.]

Dbonke.

Einleitung in die analytische Theorie der Warmeverbreitung.


(Nach. A. Beer u. J. Pliicker.) Leipzig, 1882.
KiBSCH. Die Bewegung der Warme in den Cylinderwandungen der

Dampfmafichine. Leipzig, 1886.


D'Amicis. Introduzione alia teoria matematica della propagazione
Torino, 1891.
Th6orie analytique

del calore.

PonsTCARB.

de la propagation

de

la

chaleur.

Paris, 1895.

Weber.' See above luider Riemann. 1900-1.


BoussiNESQ. Theorie analytique de la chaleur. Paris, 1901-3.
BuRKHARDT. Entwicklungen
oscillirenden
Punotionen.
naoh
Jahresber. D. Math. Ver., Leipzig, Bd. X., Heft. II., 1901.
Cabslaw. Fourier's Series and Integrals and the Mathematical Theory
of the Conduction of Heat.
London, 1906.
*

The abbreviated

titles

for the Journals are taken

Catalogue of Scientific Literature.

250

from the International

APPENDIX

II

251

Ingbesoll and Zobbll.

The Mathematical Theory of Heat ConducBoston, 1913.


In many other works considerable space is given to the Mathematical
Theory of the -Conduotion of Heat. The following may be specially
tion.

mentioned
BiOT.

Traite de Physique, T. IV.

Heine.
Bd.

Paris, 1814.

Theorie der KugeJfunctionen und der verwandten Functionen,


II. (2 Aufl.).

Berlin, 1881.

Hblmhoitz.

Vorlesungen iiber theoretische Physik, Bd. VI. Theorie


der Warme (herausgegeben von Richarz), Leipzig, 1903.
KmoHHOFT.
Vorlesungen iiber mathematischen Physik, Bd. IV.
Theorie der Warme. Leipzig, 1891.
Mathieu. Cours de Physique Math6matique. Paris, 1873.
PocKELS.
tJber die Differentialgleichung Au + k^u-0 und deren
Auftreten in der mathematischen Physik. Leipzig, 1891.
Also the articles on " Heat " in the Em. Brit. (9th Ed.), 1878, by
Kelvin, and Supplement, 1902, by Callendar, and by Sommerfeld in
Enc. d. math. Wiss., Bd. II., 1904, entitled " Randwerthaufgaben in die
Theorie der partiellen Differentialgleichungen."

MEMOIRS AND PAPERS.


FOTJBIEE.

(See F.S., p. 303.)


PoissoN. (i) Sur la distribution de la chaleur dans les corps solides.
(Three papers.) J. ec. poly tech., Paris, 12, 1823.

Sur la distribution de la chaleur dans un anneau homogene et


d'une epaisseur constante, lorsque la temperature du lieu oil il est
place varie d'un point a un autre. Connaissance des Temps, 1826.
(iii) Sur la temperature des differents points de la terre particulierement pres de sa surface. Connaissance des Temps, 1827.
Pagani. (i) Sur I'integration complete de I'equation du mouvement de
la chaleur dans lane barre prismatique d'lone petite epaisseur.
Quetelet, Correspondance Math^matique, 3, 1827.
(ii) Sur le mouvement de la chaleur dans une sphere composee de
(ii)

deux parties h^t^rogfenes et concentriques. Quetelet, Correspondance Mathematique, 4, 1828.


DiBiCHLET. Solution d'une question relative k la theorie mathematique
de la chaleur. J. Math., Berlin, 5, 1830.
J. Math., Berlin, 7, 1831.
LiBEl.' Sur la theorie de la chaleur.
(i) Sur les equations g6nerales de la propagation de la chaleur
DtxhameIi.
dans les corps solides dont la conducibilit6 n'est pas la meme dans
tous les sens.
(ii)

Sixr la

les

J. ec. polytech., Paris, 13, 1832.

m^thode g6n6rale

relative

au mouvement de

la chaleur

dans

corps solides plonges dans les milieux dont la temperature varie

avec

les

temps.

J. 6c. polytech., Paris, 14, 1833.

APPE2SfDIX

252

II

la ohaleur dans les polyedros et


(i) Sur la propagation de
principalement dans le prisme triangulaire legulier. J. ec. polyteoh.,

LamiJ.

Paris, 14, 1833.

Sur les surfaces isothermes dans les corps solides en 6quilibre de


temperature. Ann. chim. phys., Paris, 53, 1833.
PoissoN. Thdorie math6matique de la chaleur. J. Math., Berlin, 12,
(ii)

1834.

Lamk.

Sur I'equilibre des temperatures dans

cylindrique.

J.

math., Paris,

1,

les

corps solides de forme

1836.

LiouviLLB. Siir le developpement des fonctions ou parties des f onctions


en series dont les divers termes sent assujettis a satisfaire k une
equation differentielle du second ordre contenant un parametre
(Also 2, 1837 and 3, 1838.)
variable.
J. math., Paris, 1, 1836.
Stubm. Sur les Equations differentielles lin^aires du second ordre. J.
math., Paris, 1, 1836.
LamiS.
Sur les surfaces isothermes dans les corps solides homogenes
en equilibre de temperature. J. math., Paris, 8, 1837.
DuHAMEL. Sur les phenomenes thermo-mecaniques. J. ec. polytech.,
Paris, 15, 1837.

LiouviLLB ET Stuem. Extrait d'un memoire sur le developpement des


fonctions en series dont les difterents termes sont assujettis k satisfaire k une equation difterentielle lineaire, contenant un paramfetre
math., Paris, 2, 1837.
temperatures de la partie solide du globe, de I'atmosphere, et du lieu de I'espace oti la Terre se trouve actuellement.
Paris, C.R. Acad, sci., 4, 1837.
LAMfi.
(i) Sur les surfaces isothermes dans les corps solides homogenes
en equilibre de temperature. Paris, Mem. Sav. Strangers, 5, 1838.
(ii) Sur la propagation de la chaleur en poly^dres.
Paris, Mem. Sav.
variable.

PoissoN.

Sur

Strangers,

J.

les

5,

1838.

LiouviLLE. Observations sur un mtooire de M. Libri relatif a la theorie


de la chaleur. J. math., Paris, 3, 1838.
DuHAMBL. Sur les surfaces isothermes dans les corps solides dont la
conducibilite n'est pas la meme dans tons les sens.
J. math.
Paris, 4, 1839.

Sur I'equilibre des temperatures dans un ellipsoide homogene


Paris, C.R. Acad, sci., 8, 1839.
(ii) Sur les axes des surfaces isothermes du second degre consideres
comine des fonctions de la temperature. J. math., Paris, 4, 1839.
(iii) Sur I'equilibre des temperatures dans un ellipsoide k trois axes
inegaux. J. math., Paris, 4, 1839.
(iv) Sur I'equilibre des temperatures dans les corps solides homogenes
de forme ellipsoidale, ooncernant particuUerement les ellipsoides de
revolution.
J. math., Paris, 4, 1839.
Peolet. Sur la determination des coefficients de conducibilite des
metaiix par la chaleur. Ann. chim. phys., Paris (Ser. 3), 2, 1841.

Lame,

(i)

et solide.

APPENDIX

253

II

Heine.

tHber einige Aufgaben, welche auf partielle Differentialgleichungen fiihren. J. Math., Berlin, 26, 1843.
Kelvin, (i) Note on a passage in Foiorier's Heat. Cambridge and

Dublin Math. Journal, 3, 1843.


On the Uniform Motion of Heat in homogeneous Solid Bodies and
its Connection with the Mathematical Theory of Electricity.
Cambridge and Dublin Math. Journal, 3, 1843.
(iii) On the Linear Motion of Heat.
Cambridge and Dublin Math.
(ii)

Journal,

3,

1843.

Cambridge and
Surfaces.
Dublin Math. Journal, 3, 1843. (Cf. also 4, 1844.)
J. math.,
LAMfi.
(i) Sur les surfaces orthogonales et isothermes.

Note on Orthogonal Isothermal

(iv)

Paris, 8, 1843.

Sur

(ii)

la

m6thode de recherche des surfaces isothermes.

J.

math.,

Paris, 8, 1843.

Kelvin,

(i)

On

4,

Motion of Heat referred to


Cambridge and Dublin Math. Journal,

the Equations of the

Curvilinear Coordinates.
1844.

Note on some points in the Theory of Heat. Cambridge and


Dublin Math. Journal, 4, 1844.
Hbaen. On the permanent state of heat in a thin uniform wire of any
form, acted on by two sources of heat of equal intensity at its
(ii)

extremities.

LiouviLLE.

Phil. Mag.,

London

(Ser. 3), 29, 1846.

Lettres sur diverses questions d'analyse et de physique

math^matique. J. math., Paris, 11, 1846.


Stokes. On the Critical Values of the Sums of Periodic Series. Cambridge, Trans. Phil. Soc, 8, 1847.
Bonnet. Sur quelques cas particuliers d'equilibre de temperature
dans les corps dont la conductibilitd varie avec la position et direcParis, C.R. Acad, sci., 27, 1848.
tion.
DuHAMBL. Sur la propagation de la ohaleur dans les crystaux. Paris,
J. 6c. polytech., Paris, 19, 1848.

LlouviLLE.

Sur r^quation aux differences

r^quilibre de la chaleur dans


Paris, 13, 1848.

Amslbk.

un

partielles

qui concerne
J. math.,

corps h6terogene.

Zur Theorie der Anziehung und der Warme.

J.

Math.,

Berlin, 42, 1851.

Stokes. On the Conduction of Heat in Crystals. Cambridge and


Dublin Math. Journal, 6, 1861.
Amsleb. iiber die Gesetze der Warmeleitimg im Innern fester Korper
unter Berucksichtigung der durch ungleichformige Erwarmung
erzeugten Spannung.

Angstrom.

J.

Math., Berlin, 42, 1862.

Neue Methode das Warmeleitimgsvermogen der Korper

zu bestimmen. Ann. Physik, Leipzig, 114, 1861.


EvEBETT. On a Method of reducing Observations of Underground
Temperature, with Applications to the Monthly Mean Temperatures

APPENDIX

254

II

of Underground Thermometers at the Royal

Observatory, Edinburgh.

Edinburgh, Trans. R. Soc, 22, 1861.


Kelvin. On the Reduction of Observations on Underground Temperature, with Applications to Professor Forbes' Edinburgh Observations and the continued Calton Hill Series. Edinburgh, Trans.
R. Soc, 22, 1861.

MiNNiGERODE.

tjber die

Warmeleitung in Krystallen.

Diss.

Gottin-

gen, 1862.

Neumann,

F.

Experiences sur la conducibilite calorifique des solides.

Ann. chim. phys., Paris,

66, 1862.

Gleichgewicht der Warme und das der


Elektrioitat in einem Korper welcher von zwei nicht coneentrischen
Kugelflachen begrenzt wird. J. Math., Berlin, 62, 1863.
Angstrom. Nachtrag zu dem Aufsatze Neue Methode das WarmeAnn. Physik,
leitungsvermogen der Korper zu bestimmen.

Neumann,

C.

tJber

das

Leipzig, 123, 1864.

Investigation of an Expression for the Mean Temperature


a Stratum of Soil in terms of the Time of Year. Edinburgh,
Trans. R. Soc, 23, 1864.
Forbes. An Experimental Enquiry into the Laws of the Conduction
Pt. I.
Edinburgh, Trans. R. Soc, 23, 1864.
of Heat in Bars.
Kelvin. On the Secular Cooling of the Earth. Edinbiu-gh, Trans.
R. Soc, 23, 1864.
Betti. Sopra la determinazione delle temperature variabili d'un
cilindro.
Pisa, Ann. delle Universita Toscane (Ser. 2), 10, 1867.
BoussiNBSQ. Sur un nouvel ellipsoide qui joue un grand role dans la
th^orie de la chaleur.
Paris, C.R. Acad, soi., 65, 1867.
Forbes. An Experimental Enquiry into the Laws of the Conduction
Pt. II.
of Heat in Bars.
Edinburgh, Trans. R. Soc, 24,

Everett.
of

1867.

Betti. (i) Sopra la determinazione delle temperature di una lastra


terminata. Ann. mat., Milano (Ser. 2), 1, 1868.
(ii.) Sopra la determinazione delle temperature variabili di una lastra
terminata, quando la conducibilite non 6 la stessa in tutte le
direzioni.
Nuovo Cimento, Pisa, 28, 1868.
Sopra la determinazione delle temperature nei corpi solidi
(iii)
omogenei. Firenze, Mem. della Soc. delle Sci. (Ser. 3), 1, 1868.
Christofpel. Sul problema delle temperature stazionarie e la rappresentazione di ima data superficie. Ann. mat., Milano (Ser. 2),
1,

1868.

Frosch.

Uber den Temperaturzustand eines von zwei nicht coneen-

trischen Kugelflachen eingeschlossen Korpers.


13, 1868.

Zs. Math., Leipzig,

(Cf. also 17, 1872.)

BoussiNESQ. Sur les surfaces isothermes et sur les courants de chaleur


dans les milieux homogtees chauftds en un de leur points. J.
math., Paris (S6r.

2), 14,

1869.

APPENDIX

II

255

Mathieu.

Sur le mouvement de la temperature dans les corps renfermfe entre deux oylindres circulaires exoentriques et dans les
eylindres lemniscatiques.

J. math., Paris (S6r. 2), 14, 1869.


tJber die Integration der partiellen Differential-eleichung

Weber, H.

Math. Ann., Leipzig,

3x''^dy~^'^

1,

1869.

SCHLAFLI. OberdiepartiellenDifferential-gIeichung^=K|^.

J.

Math.,

Berlin, 72, 1870.

Von dbk Muhll.


Ann., Leipzig,

Knake.

tJber
2,

den stationaren Temperatiirzustand.

Math.

1870.

Warmebewegung in einem von zwei parallelen


begrenzten Korper dessen Begrenzungen mit einem Gase
in Beriihrung stehen.
Diss.
Halle, 1871.
Tiber die lineare

Wanden

Weber, H.

F.

Uber ein Problem der Warmetheorie.

Zurich, Viertel-

jahrseh. Natf. Ges., 16, 1871.

Sur I'int^gration des Equations aux differences partielles de


physique math^matique. J. math., Paris (S^r. 2), 17, 1872.
Webeb, H. tJber das Warmeleitungsvermogen von Eisen und Neusilber.
Ann. Physik, Leipzig, 146, 1872.
Baeb. tJber die Bewegung der Warme in einer homogenen Kugel.
Mathietj.
la

Diss.

Halle, 1875.

Forbes. On the Thermal Conductivity of Ice, and a New Method of


Determining the Conductivity of Different Substances. Edinburgh, Proc. R. Soc, 8, 1875.
Langeb. Uber die Warmeleitung in einer homogenen Kugel. Hab.
Schrift, Jena, 1876.

On Angstrom's Method for the Conductivity in Bars. Edinburgh, Proc. R. Soc, 8, 1875.
PuBSEE. An Application of Elliptic Functions to a Problem in the
Distribution of Heat in a Rectangular Lamina. Mess. Math.,
Tait.

Cambridge,

5,

1877.

Aybton and Pebby.


London
Kelvin.

Conductivity of Stone.

Problems relating to Underground Temperature.

London

Phil. Mag.,

Phil Mag.,

(Ser.'5), 5, 1878.

On a Method of measuring the Absolute Thermal

Lodge.
of

On the Heat

(Ser. 5), 5, 1878.

Crystals,

and other rare Substances.

Phil.

Conductivity
Mag., London

(Ser. 6), 6, 1878.

Mathieu.

iStude des solutions simples des Equations aux differences

partielles
5,

Tait.

de la physique mathematique.

J.

math., Paris (S6r.

3),

1879.

Thermal and Electric Conductivity.

Edinburgh, Trans. R. Soc,

28, 1879.

Sur les problemes des temperatures stationaires.


math., Paris (S6r. 3), 6, 1880.

BoussiNESQ.

J.

APPENDIX

256

KmcHHOFF UND Hansemann.

IT

tJber die Leitungsfahigkeit des Eisens

Warme. Ann. Physik, Leipzig (N. Folge), 9, 1880.


On Heat Conduction in Ellipsoids of Revolution. London,

fur die

NivEN.

Soc, 171, 1880.

Phil. Trans. R.

Webbb, H. F. Die Beziehungen zwischen dem Warmeleitungsvermogen und dem elektrischen Leitungsvermogen der Metalle.
Ak. Wiss., 1880.
propagazione del calore.

Berlin, SitzBer.

Betti.

Sopra

la

Chelini Coll. mat., Milano,

1881.

KiRCHHOFF UND
Metalle

fiir

Hansemann.

Warme und

tJber

die

Elektrioitat.

Leitungsfahigkeiten der
Ann. Physik, Leipzig

(N. Folge), 13, 1881.

LoBBEBG.

tJber

Warmeleitung in einem System von Cylindern.

Ann.

Physik, Leipzig (N. Folge), 14, 1881.


LoBENZ. tJber das Leitungsvermogen der Metalle fiir Warme und Elektricitat.
Ann. Physik, Leipzig (N. Folge), 13, 1881.
MoLLisoN. Note on Conduction of Heat. Mess. Math., Cambridge,
10, 1881.

Tait. Note on Thermal Conductivity and on the Effects of Temperature Changes of Specific Heat and Conductivity on the Propagation
Phil. Mag., London (Ser. 5), 12, 1881.
of Plane Heat Waves.
Resax. Commentaire k la th^orie analytique de la ohaleur de Fourier.
J.

math., Paris (S6r.

3), 8,

1882.

BOTTOMLEY. On the permanent Temperature of Conductors through


which an Electric Ciurent is passing, and on Surface Conductivity
London,
or Emissivity. With a note by Sir William Thomson.
Proc. B. Soc, 37. 1884.
FuDZiSAWA. Uber eine in die Warmeleitungstheorie auftretende nach
den Wurzeln einer transcendenten Gleichung fortschreitende

imendliche Reihe. Diss. Strassburg, 1886.


Intorno ad alcuni problemi di propagazione del calore.
Bologna, Mem. Ace. sci. (Ser. 4), 8, 1887. (See also Nuovo Cimento,
Pisa (Ser. 3), 23-25, 1888.)
Habnack. Zur Theorie der Warmeleitung in festen Korpern. Zs.
Math., Leipzig, 32, 1887.
Poinoabe. Sur la theorie de la chaleiu'. Paris, C.R. Acad, sci., 104, 1887.
WooDWABD. On the free Cooling of a Homogeneous Sphere of Initial
Uniform Temperature in a Medium which maintains a Constant
Surface Temperature. Ann. Math., Camb. Mass., 3, 1887.
Poinoab. Sm' la thdorie de la chaleur. Paris, C.R. Acad, sci., 107,

Beltbami.

1888.
(i) On the Conditioned Cooling and the Contraction of
a Homogeneous Sphere. Ann. Math., Camb. Mass., 4, 1888.
(ii) On the Diffusion of Heat in a HomogeneoTos Rectangular Mass,
with special reference to Bars used as Standards of Length. Ann.
Math., Camb. Mass., 4, 1888.

WooDWABD.

APPENDIX
HoBSON.

(i)

On

a Radiation Problem.

II

257

Cambridge, Proc.

Phil.

Soc

1889.

Synthetical Solutions in
Proc. Math. Soc, 19, 1889.

(ii)

On

Lees.

the

Law

in the Analytical

of Cooling

Theory

the Conduction of

and

its

of Heat.

Heat.

London,

Bearing on certain Equations

Phil. Mag.,

London

(Ser. 5), 28,

1889.

Chwolson. (i) tJber einem Fall von variabler Temperaturvertheilung


in einem Stabe.
Exner's Repertorium, 26, 1890.
(ii)
'Uber die Abhangigkeit der Warmeleitungsfahigkeit von der
Temperatur. St. Petersburg, M6m. Acad. Sci., 37, 1890.
Stbpan. "Dber einige Probleme der Theorie der Warmeleitung. Wien,
SitzBer. Ak. Wiss., 98, 1890.
Beyan. An Application of the Method of Images to the Conduction of
Heat. London, Proc. Math. Soc, 22, 1891.

Chwolson.

Uber die Vertheilung der Warme in einer einseitig


bestrahlten sohwarzen Kugel. St. Petersburg, Mem. Acad. Sci.,
38, 1891.

HA(SsmOM.
Vergleichende Untersuchungen iiber die Methode von
Angstrom und Neumann zur Bestimmung der Warmeleitung der
Korper. Stockholm, Vet. Ak. Ofvers, 48, 1891.
LiNDB.
(i) Methode zur Bestimmung der Warmeleitungsvermogen in
einer Kugel.
Exner's Repertorium, 27, 1891.
(ii) Uber
die Temperaturbestimmung eines Drahtes wenn durch
denselben ein galvanischer Strom durchfliesst.
Exner's Repertorium, 27, 1891.

Appbll.

Sut I'equation

3% 3z
~
^-^ = 0,

et la theorie

de la ohaleur.

math., Paris (S6r. 4), 8, 1892.


Bbill. a property of the Equation of the Conduction of Heat.
Math., Cambridge, 21, 1892.
Bryan. Note on a Problem in the Linear Conduction of Heat.
bridge, Proc. Phil. Soc,

7,

J.

Mess.

Cam-

1892.

Uber einige partioulare Losungen der Differentialgleichiing


Warmeleitung in einem Kreiscylinder und deren Anwendimgen. Wien, SitzBer Ak. Wiss., 102, 1893.
Lees. On the Thermal Conductivity of Crystals and other Bad
Conductors. London, Phil. Trans. R. Soc, 183, 1893.
Wbbbb, H. Uber den Temperatur-Ausgleich zwischen zwei sich
KoBAiiB.
fiir

die

beriihrenden heterogenen Korpern.

Gottingen, Nachr. Ges. Wiss.,

1893.

Chwolson.

Zwei Warmeleitungs -probleme.

Ann.

Physik,

Leipzig

(N. Folge), 51, 1894.

CzERMAK. Uber die Temperatittvertheilung eines diinnen Drahtes, der


von einem constanten Strome durchflossen wird. Wien, SitzBer.
Ak. Wiss.j 103, 1894.

APPENDIX

258
PoiNCARK.
Rend.

Sur

les

IT

Equations de la physique math6matique.

Palermo,

Circ. mat., 8, 1894.

Zur analytischeu Theorie der Warmeleitung. Math.


Ann., Leipzig, 45, 1894.
Lacour. Sur r^quation de la chaleur. Ann. Fao. Sci., Toulouse, 9,

SoMMERFELD.

1895.

Perry. On the Age of the Earth. Nature, London,


SoMiGLiANA. Sul problema della temperatura nell'
mat., Milano (Ser.

2), 24,

51, 1895.
ellissoide.

Ann.

1896.

Eine neue Methode zur Untersuchung der Warmeleitung in


Gottingen, Nachr. Ges. Wiss., 1896.
BouLANGER. SuT I'^quation de la propagation de la chaleur. Paris,
Bui. soc. math., 25, 1897.
Laukicblla. Sulle temperature stazionarie. Palermo, Rend. Circ.
VoiGT.

Krystallen.

mat., 11, 1897.

Sur I'int^gration des equations de la chaleur. Ann. sci. iSc.


norm., Paris (Ser. 3), 14, 1897. (Also, 15, 1898.)
VoiGT. Bestimmimg relativer Warmeleitungsfahigkeiten nach der
isothermen Methode. Gottingen, Nachr. Ges. Wiss., 1897.
AscoLi. Sulla determinazione della temperatura e dei coefficienti di
Nuovo Gimento,
conduttivita interna ed esterna di un conduttore.

Le Roy.

Pisa (Ser.

Carslaw.

4), 7,

1898.

Some Multiform

Solutions

of

the

Partial

Differential

Equations of Physical Mathematics and their Applications. London,


Proc. Math. Soc, 30, 1898.
Lauricella. Sulla propagazione del calore. Torino, Atti Ace. sci.,
33, 1898.

Lees.

On

the Thermal Conductivities of Single and Mixed Solids and


London, Phil. Trans. R. Soc,

their Variations with Temperature.

191, 1898.

Methode zur Bestimmung der Warmeleitung


Ann. Physik, Leipzig (N. Folge), 66, 1898.
Stekloff. (i) Sur le probleme de refroidissement d'une barre h6t6rogene.
Paris, C.R. Acad, sci., 126, 1898.
(ii) Sur un probleme de la th6orie analytique de la chaleur.
Paris,
C.R. Acad. sci.. 126, 1898.
Straneo. (i) Sulla temperatura di un conduttore lineare bimetallico.
Roma, Rend. Aco. Lincei (Ser. 5), 7, 1898.
ScHULZB.

tjber

eine

fester Korper.

(ii)

Sulla determinazione simultanea delle oonducibilit4 termiche ed

electrice dei metalli

Lincei (Ser.

KoHLRAUSCH.

5), 7,

a differenti temperature.

Roma, Rend. Ace

1898.

t)ber den stationaren Temperaturzustand eines von


einem elektrischen Strome erwarmten Leiters. Berlin, SitzBer.
Ak. Wiss., 1899.
Pierce and Willson. On the Thermal Conductivities of certain Poor
Conductors. Boston, Mass., Proc. Amer. Acad. Arts Sci., 34, 1899.

APPENDIX

259

II

VoiGT. tJber ein von Herrn Fr. Kohlrausch aufgestelltes Problem der
Warmelehre. Gottingen, Naohr. Ges. Wiss., 1899.
BoTJSSlNESQ. (i) Reduction de certains problemes d'^chauffement ou de
refroidissement par rayonnement au cas plus simple de reohauffement ou du refroidissement des mgmes corps par contact ; 6chaufEement d'un mur d'^paisseur ind6finie. Paris, C.R. Acad, sci., 130,
1900.

Probleme du refroidissement de la oroute terrestre, traits au meme


point de vue que I'a fait Fourier, mais par une m6thode d'int6gration beaucoup plus simple. Paris, C.R. Acad, sci., 130, 1900.

(ii)

Probleme du refroidisseraent d'un mur par rayonnement ramen6


au cas plus simple oil le refroidissement aurait lieu par contact.

(iii)

Paris, C.R. Acad,

sci.,

130, 1900.

Echauffement permanent mais in^gal par rayonnement, d'un


mur d'^paisseiu- ind^finie, ramen6 au cas d'un Echauffement analogue
par contact. Paris, C.R. Acad, sci., 131, 1900.
(v) Probleme de I'echauffement permanent d'une sphere par rayonnement, ramen6 au probleme plus simple de I'echauffement de la
meme sphere par contact. Paris, C.R. Acad, sci., 131, 1900.
Cotton. Mouvement de la chaleur sur la surface d'un t6traedre dont
les aretes oppos6es sont egales.
Ann. Fac. Sci., Toulouse (S6r. 2),
(iv)

2,

1900.

Tiber die Bestimmung des metaUischen Warmeleitvermogens und iiber sein Verhaltniss zur elektrischen Leitfahigkeit.
Ann. Physik, Leipzig (4. Folge), 3, 1900.
Hall. Concerning Thermal Conductivity in Iron. Physic. Rev.,

Gbttneisen.

Ithaca, N.Y., 10, 1900.


KoHiiBAXTSCH. tJber den stationaren Temperaturzustand eines elektrischgeheisten Leiters.
Ann. Physik, Leipzig (4. Folge), 1,
1900.

Peibce.

On

London

the Thermal Conductivity of Vulcanite.

Phil.

Mag.,

(Ser. 5), 49, 1900.

Sur I'equilibre calorifique d'tme surface fermee rayomiant


Paris, C.R. Acad, sci., 130, 1900.
Bull.
(ii) Sur quelques problemes relatifs
k I'Equation ^u = k^u.
Sci. Math, de France, 38, 1900.
STEKLOFi". Le probleme des temperatures stationaires. Paris, C.R.
Acad, sci., 131, 1900.
BoTJSSiNESQ. Probleme de la dissipation en tous sens de la chaleur
dans un mur 6pais k face rayonnement. Paris, C.R. Acad, sci.,
PlOABD.

(i)

au dehors.

133, 1901.

DouGALL. Note on the Application of Complex Integration to the


Equation of Conduction of Heat, with a special application to Dr.
Edinburgh, Proc. Math. Soc, 19, 1901.
of a Sphere in a Mass of well-stirred
Edinburgh, Proc. Math. Soc, 19, 1901.

Peddie's Problem.

Peddie.

Note on the Cooling

Liquid.
O.C.H.

r2

APPENDIX

260

Probleme du refroidissement d'une barre h6t6rog6ne. Ann.


Toulouse (Ser. 2), 3, 1901.
a Problem in Conduction of Heat. Phil. Mag., London

Stekloff.
Fac.

II

Sci.,

Cabslaw.

(Ser. 6), 4, 1902.

CesXbo.

Sur un probleme de propagazione de

(i)

la chaleur.

Bruxelles,

Bui. Acad, roy., 1902.


(ii) Intorno ad una limitazione di constanti nella teoria analitica
Napoli, Rend. Ace. sci. (Ser. 3), 8, 1902.
del calore.

Lauricella.

delle

Sull 'integrazione

Roma, Mem.

del calore.

Soc.

XL.

equazioni

della propagazione

(Ser. 3), 12, 1902.

Mackenzie. On some Equations pertaining to the Propagation of


Heat in an Infinite Medium. Philadelphia, Pa., Proc. Amer.
Phil.

Soc,

41, 1902.

The Steady Temperatures

Peck.

of

a Thin Rod.

Phil. Mag.,

London

(Ser. 6), 4, 1902.

Schaufelbbrgbb.

Warmeleitungsfahigkeit

aus

dem

bestinamt,

und

Kupfers,

des

imd variablen Temperaturzustand

stazionaren

Warmefluss in einer durch Kiihlwasser bespiilten Endflaohe eines


Warmeleiters. Ann. Physik, Leipzig (4. Folge), 7, 1902.
Stekloff. Sur les problemes fondamentaux de la physique mathematique. Ann. sci. ific. norm., Paris (S6r. 3), 19, 1902.
Carslaw. The Use of Green's Functions in the Mathematical Theory
of the Conduction of Heat. Edinburgh, Proc. Math. Soc, 21, 1903.
GiBBE.
tjber die Bestimmung des Warmeleitungsvermogen bei
tiefen Temperaturen.
Berlin, Verh. D. Physik. Ges., 15, 1903.
Prasad. Constitution of Matter and Analytical Theories of Heat.
Gottingen, Abh. Ges. Wiss., 2, 1903.
SoMiGLiAJSTA. Intomo ad un problema di distribuzione termica. Milano,
Rend. 1st. lomb. (Ser. 2), 36, 1903.
BoussiNESQ. Sur la unicite de la solution simple fondamentale et sur
I'expression asjrmptotique des temperatures dans le problfeme du
'

refroidissement.

Hecht.

F. E.

fahigkeit

und

Bui.

sci.

ihre Durchfiihrung

Leipzig

VOLTBEBA.

math., Paris (S6r.

1904.

2), 28,

Neumanns Methode zur Bestimmung der Warmeleitungsschlecht leitender Korper in Kugel- und Wiirfelform,

(4.

an Marnior, Glas,

etc.

Ann. Physik.

Folge), 14, 1904.

Sur

les

Equations

diff6rentielles

du

tjrpe

parabolique.

Paris, C.R. Acad, sci., 139, 1904.

BtTHL.

Sur I'approximation des fonctions par des polynomes dans

aux d6riv6es partielles ;


application au probleme de I'^tat initial in physique inath6matique.
Paris, C.R. Acad, sci., 140, 1905.
Glage. F. E. Neumanns Methode zur Bestimmung der Warmeleitses rapports aveo la thdorie des Equations

fahigkeit gut leitender Korper in Stab- und Ringform, und ihre


Durchfiihrung an Eisen, Stahl, etc. Ann. Physik, Leipzig (4. Folge),
18, 1905.

APPENDIX

II

261

Lees.

The Effects of Temperature and Pressure on the Thermal


Conductivities of Bodies. Part I. The Efiect of Temperature

on the Thermal Conductivity of some


London, Phil. Trans. R. Soc, 204, 1905.

Electrical

Insulators.

On a Method of finding the Conductivity for Heat.


Proc. R. Soc, 76, 190S.

NiVBN.

Zabemba.

Solution

gte^rale

du probleme de

London,

Krakow,

Fourier.

Bull. Intern. Acad., 1905.

PiCAED.

Sur quelques problemes de physique math6matique se


M. Fredholm. Paris, C.R. Acad sci., 142,

(i)

rafetachant a I'equation de

1906.

Sur quelques applications de I'equation fonctionelle de M.


Fredholm. Palermo, Rend. Circ. mat., 23, 1906.

(ii)

Thoma.

Warmeleitungsproblem bei wellig begrenzten Oberflache und

Anwendung

dessen

auf Tunnelbauten.

Freiburg u. Br.,

Diss.

1906.

Holmgren.
1907.

Sur I'equation

^-j=^.

Paris, C.R. Acad,

sci.,

145,

(See also 146, 1908.)

KoijnGSBEBGEE TJND DiscH.

Bestimmimg der Veranderlichkeit des


von Fourier und experimentelle Anwendung auf Waimeleitung von Isolatoren. Ann.
Koefficienten der Differentialgleichung

Physik, Leipzig
Levi.

Stdl'

(4.

Folge), 23, 1907.

equazione del calore.

Roma, Rend. Ace.

Lincei (Ser.

5),

16, 1907.

Mylleb-Lebbdefp. Die Theorie der Integralgleichungen in Anwendung auf einige Reihenentwickelungen (Warmeleitung in
einem Kreiscylinder). Math. Ann., Leipzig, 64, 1907.
Suir equazione della propagazione di calore in un
Lincei (Ser. 5), 16, 1907.

PicciATTi.

fllo.

Roma, Rend. Ace.

Un

Steklofp.

probleme d'analyse intinaement

refroidissement d'une barre h6t6rogene.

116

au probleme du

Paris, C.R. Acad,

sci.,

144, 1907.

Tbdonb.

Sul

ellissoide

problema dell' equilibrio delle temperature in


a tre assi disuguali. Palermo, Rend. Circ. mat.,

un
24,

1907.

Ebbling. tJber den Temperaturverlauf in wechselstromdurchflossenen


Drahten. Ann. Physik, Leipzig (4. Folge), 27, 1908.
Sur I'equation de la propagation de
Matem., Stockholm, 4, 1908.

Holmgeen.

Xaueicella.

Applicazione

della

del raffreddamento dei corpi.


1908.

la chaleur.

di Fredholm
Ann. mat., Milano

teoria

al

Arkiv.

problema

(Ser.

3),

14,

APPENDIX

262

II

Temperature and Pressure on the Thermal


Part II. The Effects of low Temperatures on the Thermal and Electrical Conductivities of certain
London, Phil. Trans.
approximately pure Metals and Alloys.
R. Soc, 208, 1908. (Also Proc. R. Soc, 80.)

The

Lees.

Effects of

Conductivities of SoUds.

Levi,
(ii)

Sul problema di Fourier.

(i)

Sur l'6quation

^=

MYLLEB-LEBEDErr.

P*"^'

f^-

tjber die

einer parabolischen

Torino, Atti Ace.

^^-

sci.,

43, 1908.

^^'^- ^^-' ^^^' ^^^-

Anwendung der

Randwertaufgabe.

Integralgleichungen in

Math. Ann., Leipzig,

66,

1908.

Labose.

(i)

Sur les solutions particuliferes de I'^quation W^~"^ = 0*


Acad, sci., 148, 1909.
Paris, C.R. Acad.
probldme de rarnulle de Fourier.

Paris, C.R.
(ii)

Sur

sci.,

le

148, 1909.

Sur
Acad,

(iii)

problema de I'armille aveo deux ruptures.

le

sci.,

Paris, C.R.

148, 1909.

PiCAED. (i) Quelques remarques sur les Equations int6grales de premiere espece et sur certains problemes de physique math^matique.
Paris, C.R. Acad, sci., 148, 1909.
(ii) Sur une Equation aux d6riv6es partielles du second ordre, relative
a une surface fermte, correspondant a un 6quilibre calorifique.
Ann sci. &. norm., Paris, 26, 1909.

The Green's Fimction for a Wedge of any Angle and


other Problems in the Conduction of Heat. London, Proc. Math.

Carslaw.

Soc. (Ser. 2), 8, 1910.

Lees.

On

the Shapes of the Isogeotherms londer Mountain Ranges


London, Proc. R. Soc, 83, 1910.

in radioactive Districts.

PiCABD. Sur un th6oreme g6n6ral relatif aux Equations integrates


de premiere espece et sur quelques problemes de physique math6matique. Palermo, Rend. Circ. mat., 29, 1910.

Mabcolongo.

Suir equazione della propagazione del calore nei corpi


Napoli, Rend. Ace. sc. (Ser. 3), 17, 1911.

cristallizzati.

Raylbigh.

Problems in the Conduction of Heat.

PhU. Mag., London

(Ser. 6), 22, 1911.

SucHY.
(4.

Warmestrahlung und Warmeleitimg.

Ann. Physik, Leipzig

Folge), 36, 1911.

Carslaw.

problem in the linear flow of heat discussed from the


point of view of the theory of integral equations. Edinburgh,
Proc. Math. Soc, 30, 1912.

Silla.

Sulla propagazione

del

calore.

Roma, Rend. Aco.

Lincei

(Ser. 5), 21, 1912.

VoLTEEEA.

Sulla temperatura nell' intorno .delle montagne.

Cimento, Pisa

(Ser. 6), 4, 1912.

Nuovo

APPENDIX
Webeb.

263

II

den Eindeutigkeitsbeweis in der Theorie der WarmeHeinrich Weber Festschrift, 1912.

tJber

leitung.

MitANKOwiTSCH. t5TDer ein Problem der Warmeleitung


Anwendiing auf die Thfeorie des solaren Klimas.

tind dessen
Zs.

Math.,

Leipzig, 62, 1913.

PiCABD.

Application de la theorie des equations int^grales k certains


problteies de la th6orie analytique de la chaleur dans I'hypothese
d'un saut brusque de temperature k la surface de separation des
corps en contact. Paris, C.R. Acad, sci., 156, 1913.

SoMiGLiANA

(e
Veboblli).
Siilla
temperatura nei grandi trafori

previsione
alpini.

matematica

Torino,

Mem.

della

Aco.

sci.

(Ser. 2), 63, 1913.

Stbkloit.

Sur

(Also Veroelli, Torino, Atti Ace. sc, 48, 1913.)


certaines questions d'analyse qui se rattachent k

plusieurs probl^mes de la physique math6matique.

Mem. Acad.

sci. (Ser. 8), 31,

St.

Petersburg'

1913.

Veboblii. Sulla determinazione dei coeflficienti di conduttivita


termica mediante il raffreddamento di sfere. Nuovo Cimento,
Pisa (Ser.

Wbinbeich.

6), 6, 1913.

tJber

Drahten,

den Temperaturverlauf in stromdiurchflossenen


im Fall von Wechselstrom. Zs. Math.,

besonders

Leipzig, 63, 1914.

BoussiNBSQ.

(i) Calcul correct de I'influence de rin6galit6 elimaterique


sur la Vitesse d'accroissement des temperatures terrestres avec
la profondetir sous le sol.
Bui. sci. math., Paris (S^r. 2), 39, 1915.

(ii)

Sur

le

probleme du refroidissement de

la croute terrestre con-

sider6 k la maniere et suivant les idees de Fomier.

Paris (Ser.

2), 39,

Bui.

sci.

rnath.,

1915.

On the non-stationary state of heat in an' ellipsoid. Bull.


Calcutta Math. Soc, 8, 1915.
Laudien. Entwicklung willkiirlicher Funktionen bei einem thermoDatta.

elastischen Problem.

AiCHi.

(i)

On

J.

Math., Berlin, 148, 1918.


^6=hW. Proc. Phys.-math. Soc,

Picard's solution of

Japan (Ser. 3), 1, 1919.


Heat Distribution on a, radiating plane. Proc. Phys.-math^ Soc,
Japan (Ser. 3), 1, 1919.
Bbomwich. Examples of operational methods of solving problems in
the conduction of heat. Phil. Mag., London (Ser. 6), 37, 1919.
McLeod. On the lags of thermometers with spherical and cylindrical
bulbs in a mediiom whosS temperature is changing at a constant
rate.
Phil. Mag., London (Ser. 6), 37, 1919.
AiCHi.
(i) Heat Distribution on a radiating plane, and especially when
the boundary is circular. Proc Phys.-math. Soc, Japan (Ser. 3),
2, No. 2 and No. 6, 1920.
convection of heat by the uniform
(ii) On the two-dimensional
current of steam. Proc Phys.-math. Soc, Japan (Ser. 3), 2,
(ii)

No.

7,

1920.

APPENDIX

264
Cabslaw.

On Bromwich's method

II

of solving

problems in the con-

London (Ser. 6), 39, 1920.


Hayashi. On Picard's solution of Au=k-u.
Science Reports of the
Tokuhu Imperial University, Sendai, Japan (Ser. 1), 9, 1920.
Bromwich. Symbolical methods in the theory of conduction of heat.
duction of heat.

Phil. Mag.,

Cambridge, Proc.

Phil.

Cabslaw.

Soc,

20, 1921.

The

cooling of a solid sphere with a concentric core of a


different material.
Cambridge, Proc. Phil. Soc, 20, 1921.

Owen.
is

On the lag of a thermometer in a medium whose temperature


a linear function of the time. London, Proc. Math. Soc. (Ser. 2),

19, 1921.

LIST OF

AUTHOES QUOTED

The mumbers
Angstrom, 41, 42, 44,
Ayrton, 141.

refer to pages.

Hagstrom,

70.

42.

HaU, 133.
Hansemann,
Hardy, 228.

Barnes, 144.
Beglinger, 133.

45.

Heaviside,60, 201, 209, 218, 221.


Heine, 124, 125, 126, 144, 198.
Heywood and Freohet, 225.

Betti, 171.

B6oher, 225.
Borel, 31.

Hilbert, 225.
Hobson, 46, 115, 149, 168, 182.

Boussiuesqi 51, 55.


Bromwioh, 201, 218, 223.

Bryan, 176.

Holmes, 60.
Horn, 225.

Burgess, 34.
Byerly, 90, 142, 143, 144.

Ingenhaiisen, 40.

Carslaw, 168, 171, 176, 189, 201, 218,

Kaye and Laby,

222
Clerk-MaxweU,

Kelvin,

209,

218.

8, 169.

King, 59.

Granz, 81.

Ozermak,

3.

8, 56, 57, 58, 59, 60, 149,

Kirchhoff, 28, 45, 51.


Kirsch, 240.
Kneser, 225, 232, 237.

83.

De Morgan,

34.

Kohlrausch, 83.

Despretz, 40.
Diesselhorst, 46.
Dini, 115.

Lamb,

Duhamel, 17.
Duncan, 83.

Laplace, 31, 54.


Lees, 114.

Ebeling, 81.

EddJngton, 60.
Encke, 34.
Sejer, 65.
Forbes, 40, 41, 56.
Eord, 115, 138.
Fourier, 20, 26, 54, 56, 57, 58, 88, 91,
110.
Pranz, 40, 81.
^

Gibson, 30.
Giebe, 46.
Goursat, 33, 225.

Gray and Mathews,


156, 165, 193.

115, 128, 147, 153,

9.

Mackenzie, 35.
Mathieu, 94.
Minnigerode, 171.
MoUison, 70.
Moore, 65, 106, 115, 117, 122.

Neumann,

F., 26, 42, 70, 71, 79, 80.


Nielsen, 193.
Niven, 114, 161, 241.

Peck, 121.
Perry, 60, 141, 209, 221.
Pookels, 171.
Poincar^, 159.
Poisson, 54, 56.

Poynting and Thomson, 40,

41.

Preston, 40, 65.

Gregory (J. W.), 60.


Gruneisen, 46.

Rayleigh, Lord

265

(J.

W.

Strutt), 149, 156.

LIST OF

266
Rayleigh, Lord (B.

J. Strutt), 60.

Riemann,

140.
Rutherford, 60.

Schaufelberger, 84.
Sohwarzsohild, 168, 171.
SoUas, 54.

Sommerfeld, 163, 166, 167, 168, 171.


Sonine, 193.
Stokes, 69.
Straneo, 84.

Turner, 238.
Verdet, 81.
Vivanti, 225.

AUTHORS
Watson, 115, 117, 122, 128, 147, 186,
222.

Weber, H., 34, 69, 70, 163.


H. F., 131, 133.
R., 141.

Weber-Riemann,

34, 51, 140, 163, 171.

Webster, 169.
Weinreich, 81, 83, 84.
Whittaker and Watson, 142, 143, 159,
191, 225.

Wiedemann, 40, 81.


Winkelmann, 40, 41.
Woodward, 60.
Young, 115.

GENERAL INDEX
The numbers

Age

refer to pages.

of the Earth : Kelvin's classical treatment, 57


activity, 60 ; later discussion, 209, 218, 221.

British

Calory.

Thermal Unit.

effect of discovery of radio-

See under Quantity of Heat.

See under Quantity of Heat.

Conductivity.

See under Conduction of Heat.

Conduction of Heat

Definition of, 1 ; Definition of Conductivity {K), 2


:
Definition
of Diffusivity or Thermometric Conductivity (k), 8 ; Definition of Emissivity
(H), 14.
Equation of Conduction, in Cartesian Coordinates, 8 ; in Spherical
Polar Coordinates, 11 ; in Cylindrical Coordinates, 12; Initial and Boundary
Conditions, 12.
Methods of obtaining the Value of the Conductivity (K) and
;

the Emissivity (H)


Foi Good Conductors
Despretz, 40, 66 ; Forbes, 40 ; Giebe, 46 ;
:

Bar Methods

Angstrom, 41
Hagstrom, 42
Neumann, 70, 79 Wiede-

Gruneisen, 46

Ingenhausen, 40 ; KirohhofE and Hansemann, 45


mann and Franz, 40. King Methods Fourier, 25 Neumann, 26. Electric
Duncan, 83
Cranz, 81 ; Czermak, 83
Ebeling, 81
Current Methods
Sohaufelberger, 84
Straneo, 84
Verdet, 81 ;
Kohlrausch, 83
Lees, 84
Weinreich, 81, 83, 84. For Poor Conductors: Cube, 110, 112; Cylinder:
Beglinger, 133; Hall, 133; Weber, H. F., 131, 133: Sphere: Ayrton and
Perry, 141
Weber, K., 141.
Conjugate Functions, Their use in solving two-dimensional problems, 91.
;

Contour Integrals, Their use in solving the equation of conduction (Chapter XI.).
Convection of Heat,

1.

Cylindrical Surface Source, 153.

DifEusivity or

See under Conduction of Heat.

Thermometric Conductivity.

Dimensions, of Conductivity (K), 3

of Diffusivity

Diriohlet's Conditions, referred to first

on

p. 22.

(), 8.

For

definition, see F.S., 93.

Doublets, 156.

Duhamel's Theorem, 17
Emissivity.

See under Conduction

of

Heat.

Two-Dimensional Flow,
Flow of Heat : Linear Flow, Chapters III. and IV.
In a Circular
In a Rectangular Parallelepiped, Chapter VI.
Chapter V.
In a Wedge,
In a Sphere and Cone, Chapter VIII.
Cylinder, Chapter VII.
;

164, 189.

Fourier's Ring, Chapter II

Temperature due to a source at a point


267

in, 160.

GENERAL INDEX

268

Green's Functions, Their use in solving the equation of conduction, Chapter X.


Heaviside's Operational Method, 201.

Images, The method

of, 158.

See also under Sommerfeld's Extension,

Integral Ecinations, Their use in solving the equation of conduction, Chapter

Isothermal Surfaces,
Quantity of Heat

Radiation of Heat,

XII

4.

the Calory, 3

the British Thermal Unit (b.t.u.),

3.

1.

Riemann's Surface and Space, 167.


Sommerfeld's Extension of the Method of Images, 166.
Sources and Sinks, In two-dimensional flow (steady temperature), 102
in cases of variable temperature, Chapter IX.
Spherical Surface Source, 151.

Terrestrial Temperature, 53.

Thermometric Conductivity.

See under Conduction of Heat.

Their use