Jeffreys-CartesianTensors Text PDF

10$ (50p)
net in U.K.
$1.65
in U.S.A.
CARTESIAN
TENSORS
m
Z 1
i/l
V)
CARTESIAN TENSORS
BY
HAROLD JEFFREYS
M. A. D. S c, F. R. S.
CAMBRIDGE
AT THE UNIVERSITY PRESS
1969
Publiahed by tho Syudioa of tho Cambridge Univorsity Press
Bentley Houso, 200 Euston Rond, London N. W.
PREFACE
1
American Branch: 32 East 57th Street, New York, N. Y. 10022
Standard 13ook Xumbers: It is widoly felt that wlien the equations of mathematical
521 05423 clothbound
physics aro written out in f uli Cartesian form tho structural
521 09191 8 paperback
simphcity of the formulae is often hidden by the mechanical
labour of writing out every term explicitly. Atteuipts have
been made to reduce this labour by one form or another of
veetor algebra but it has always seemed to me that this
;
method both introduces new dimeulties and is insufficiently

general. Thus the produet of two veetors, in veetor lan-
First published 1031 guage, means one of two tliings, either the scalar or the
Roprinted 1952 1957 1961
veetor produet, and it is not physically obvious why just
1903 1965 1969
these funetions of the veetors should arise and no others.
The use of tensor notation, with the summation con-
vention, carries out as great a simphfication of the writing
as does veetor notation. The notation has actually at-
traeted attention owing to its applications in the theory of
relativity, but for ordinary purposes two great abbrevia-
tions maybe made. We use rectangular Cartesian axes;
the result that the distinetion between covariant and
is
contravariant veetors disappears, and with it the torms

arising from curvature of the surfaces of reference. The
formidable charaeter of most of the formulae of the theory
of relativity is absent from the formulae of tensors referred
to Cartesian axes. The tensor method is a necessity for re-
lativity; for applications in dynamics, electricity, elas-
ticity, and hydrodynamics a great convenience.
it is
It is found that the scalar and veetor produets are not

the only funetions of two veetors that arise, though the
Printed in Great Britain
m the Univoraity Printing House, Cambridge theory provides reasons why they are important in many
(Brooko Crutohley, Univoraity Printor) applications. There is also a symmotrical produet, which
ordinary veetor notation is completely unable to express.
In tensor notation it arises naturally as a symmetrical
VI
PREFACE
tensor of the socond ordor.The systom of momonts and
products of inertia of a rigid body constitutes such a
tensor; so do the strcss components and the strain com-
ponents in an
NOTE
elastic solid.
The presont method, like vector notation, of use prin-
is
Since this book was written most of the material in it
cipally in proving general theorems. In concrete applica- has been incorporated in Chapters 2 and 3 of Methods of
tions there is usually some asymmetry about the coor-
Mathematical Physics, by my wife, Bertha Swirles Jeffreys,
dinates that makes it necessary to abandon the tensor
and myself. The present reprint has been made partly
form at some stage in the work. It has been said that
because of a continuing deniand for a treatment of Car-
vector equations are like a pocket map, and it has been re-
tesian tensors by themsclves partly because some results,
;
plied that a pocket map has to be taken out of the pocket
notably on the thermodynamics of an elastic solid and the
and unf olded before it is of any use. The same applies to the
circulation in viscous flow, are not givcn in textbooks of
tensor method, and for the same reason; but it has the
great advantage that the special subjects.
not a new notation, but a concise
it is
way of writing the ordinary notation, so that the unfolding

HAROLD JEFFREYS
Januari/, 1952
can be carried out more conveniently when occasion arises.
What is usually called Statics
is treated in Chapter v,
after Dynamics. I consider tbis to be the proper order, be-

cause Statics is a special case of Dynamics, and many of its
formulae have physical significance for reasons explaincd NOTE ON THE SEVENTH IMPRESSION
in Dynamics. The customary reversal of the order is due,
I believe, to the fact that an introduction to mechanics has
In the present reprint a fow small changes have been made.
to be given at schools before the students have received
In Ex. 3, p. 15, the components are given explicitly these
any training in calculus; but this need not infiuence ;
students working for a university examination.

have been found useful in developing tho strain energy for
It Bhould perhaps be stated that the object of this work an elastic sphere under rotation. Omissions have been
is to illustrate the use of tensor methods; it does not claim
corrected on pp. 81 and 82. As these do not affect the
to give a complete theory of all the subjects touched, re- later work they appear to have givon readers no trouble.
ference for which must be made to the Standard text-books. HAROLD JEFFREYS
I must express my gratitude to Mr M. H. A. Newman, December, 19C8
Miss L. M. Swain, Dr S. Goldstein, and Dr Bertha Swirles
for assistance at various stages in the work, and to the
staff of the University Press for their care in the printing.
HAROLD JEFFREYS
8eplember 1031
CONTENTS CHA P T E R I
CARTESIAN TENSORS
Chap. I. Cartesian Tensore page 1
If we have two sets of rectangular axes (0x, Oy, Oz),

II. Geometrical Applications 16
(0x', Oy', Oz') at the same origin, the coordinates of a
III. Particle Dynamics 24 point P with respect to the second set are given in terms
of the coordinates with respect to the first set by the
IV. Dynamics of Rigid Bodies 29
equationa
V. Equivalence of Systems of Forces 42 x' = liX + m^y + n^z
y' = li x + m^y + w,a (1).
VI. Continuous Systems 48
z' = laX + rr^y + rt^z
VII. Isotropic Tonsors 66

The quantities {lt , m,, 74 , .... n^) aro the cosines of the
VIII. Elasticity 71 anglos between the varions axes ; thus ii is the cosine of the
angle between the axes 0x' and 0x, n^ is the cosine of the
IX. Hydrodynamics 83
angle between Oy' and Oz, and so on. It follows that the
Index 93 coordinates (a;, y, z) can be expressed in terms of (x' , y', z')
by the relations
x = ltx' + ky' + l^z' 'j
y = nâ:' + rn^y' + rn^z' <-

(2).
z = nja;' + j y' + n^z'
We can shorten the writing of 1 ) and (2) considerably (
by a change of notation. Instead of (a;, y, z) let us write

fa, *2 a^) aQd instead of (x' y', z') write (as/, x2 a*,'). We
',
t
can now say that the coordinates with respect to the first
set of axes are x where i may be 1, 2, or 3; and thoso with
( ,
respect to the second set are x/, where,;' may be 1, 2, or 3.

Then in (1) each coordinate x/ is expressed as the sum of
three terms depending on the three x Each x is associated (
.
t
with the cosine of the angle betwoon the diroetion of that

2 CARTESIAN TENSORS CARTESIAN TENSORS 3
x increasing and that of x/ increasing. Let us donote

t this the left it is to be given in turn all the values 1,2,3, and the
cosine by a if Then we have, for all values of j,
. resulting equation is asserted in each case. In each such
xi = ais x + <hi xz + equation the right side is the sum of three terras, obtained
i ChiZi
by givingj or i the values 1, 2, 3in turn and adding. Wher-
= S a x, tl (3).
i-1. 2. 3 ever such a summation occurs a suffix is repoated in the
Conversely (2) can be written expression for the general term; where thore isa summa-
'
tion for all values ofj the general term, such as a if u t con- ',
z, = S a (S x t '
(4),
y-i.2.3 tainsj twice. We make it a regular convention that, unless
the a ti having the same value as in (3), for the same values the contrary is stated, whenever a suffix is repeated it is
of i and j, because it is in both casos the cosino of the angle to be given all possible values and that the terms are to be
between the directions of x ( and x/ increasing. added for all. Thus we write (5) as simply
In mathematical physics we often have to deal with sets u, = a (i Ui (9).

of three quantities in relation to a set of axes, of the general
the summation sign being automatically understood by
form u t (that is, v^, v^, w,), and such that in relation to a
our convention. Then (9) really means three equations,
different set of axes the corresponding quantities arc
with tlu-ee terms on the right of each, but we can by means
(,', u/, u 3 '), which satisfy the relations
of our conventions express all of the twelve terms com-
w/ = 2 dijUi (5)
1-1,2,3 pactly by the single equation (9).
There are single quantities, such as mass and distance,
and u =
( 2 <)/ (6).
i-1. i. 3 that are the same for all sets of axes. These are called
Such sets of three quantities are called tensors of the first tensors of zero order, or scalars.
order, or vectors. The individual j.Bj may be oalled the Consider now two tensors of the first order, u and v k { .
components of the tensors. (When we write "a tensor m<" we mean of course a tensor
Clearly if we multiply all of the u t and u,' by the same of the first order whose components are u^, w,, j. This is
quantity m we get another piece of shorthand.) Suppose each component of
mu/ = 2 a (wm,) the one multiplied by each component of the other; then
(7),
fl. 9.3 we obtain a set of nine quantities expressed by u vk where t ,
so that mut is another tensor of the first order. each ofi and k is independen tly given all the values 1, 2, 3.
Again, if we have two tensors of the first order, u, and The components of u,, v k with rospect to the other set of
v ( we shall have
, axes are u/, v ' say and t
;
u/ + v/ - 2 a iS (u t + vt ) (8), u/v,' = (a ti Ui) (atl vk )

i li 2. .*
so that u ( + vt
a tensor of the first order.
is
= ai,akiUiVk (10).
We notice that each of the equations (3) to (8) is really The suffixes i and k are repeated on the right. Thus (10)
a set of three equations; where the suffix i orj appears on reprcsents nine cquations, each with nine terms on the
right. Each term on the right is the product of two factors, If we interchangej and Z in (11), we get
one of the form a it akl depending only on the orientation of
,
tv,/ = auauw* (12).
the axes, and the other of the form u ( vk representing the
,
products of the components referred to the original axes.

But on the right i and k are "dummy sufBxes"; that is,
In this way the various u/ v,' can be obtained in terms of

they are to be given and the results
all possible values
the original u,vk But products of two vectors are far from
.
addcd. It is unimportant which of them we call i and which
k; we may therefore interchange them and get
being the only quantities satisfying this rule. In general a
set of nine quantities w ik referred to a set of axes, and w,/ = a kl a ti wki = ai,a kl tok , (13).
transformed to another set by the rule
Thus wkt transforms according to the same rule as w lk and
w,i = aâ^w^ (H). therefore is another tensor of the second order. The im-
portance of this is that if we know the set of quantities
is called a tensor of the second order.
arranged
VVe go on similarly to construct and define tensors
may
'n w u>,
of the third, fourth, and higher orders. Thus a set of quan-
v. ;
u>2 , w,. (14)

tities that transforms like zt O)k Xn K9 ... is called a tensor of
VW,, w..
order n, whore n is the number of factors in this product. Ufe
When wo say that a certain set of quantities is a tensor to be a tensor of the second order, then the arrangement
ofany order n, we mean that we have ways of specifying
compononts with respect to any set of axes, and that the h ^81 31 \
its
w,2 'm %1 (15)
components with regard to any two different sets of axes
are related according to the rule appropriate to tensors of
wl3 1
W.a *l*
MJ33

that order, and in particular to the products of the coor- isanother tensor of the second order. Therefore the sets
dinates with ra factors. For instance, if we say that u ( is a -i- w,.,) and (w (k
(w ik wki ) are tensors of the second order.
tensor of order 1, we are not simply defining w/ as meaning The first of these has the property that it is unaltered by
a (i u ( We are supposing both that u/ has a meaning, such
. interchanging i and k, and is therefore called a symmetrical
as a displacement or a velocity, with reference to the axes tensor. all its components reversed in sign
The second has
of z,', and that the value of each component is equal to when and k are interchanged, and is called an antisym-
i
a ti u Thus the statement that any set of quantities is a
t
.
metrical tensor. Clearly in an antisymmetrical tensor the
tensor is not a mere convention, but a statement capable "leading diagonal" components, i.e. those with i and k
of test and therefore nceding proof. In (7) and (8), for equal, are all zero. Also, since
example, our data are that m, and u/ are the components - wkl
w = i (w + ) + J (w (16),
of a vector with regard to two different sets of axes. We
)
prove that the sets of quantities obtained by multiplying we can consider any tensor of the second order as the sum
both by the same quantity are related according to the of symmetrical and antisymmetrical parts.
vector rule and therefore the products are vectors.
; The gradient of a scalar is a vector. For if U is a scalar,
CARTESIAN TEKSORS CARTESIAM TENSORS 7
6
formation is that 8,,' = 1 if j = and 8' = l, if j # l, It
its gradient is dUjdx, or dU/dzf according to the set of
follows that the set of quantities
axe8. But
dU dx, dU /1 <K
~ U (17),
dx/ dx/ ~dx, " dx,
10 (21)
so that the gradients transforin according to

the vector \0 1/
rule. Similarly the gradient of a

vector is a tensor of
istransforraed into itself by the rule (11) and thorefore ia
order 2. For if u, and u/ are the components of a vector a tensor of the second order.
with respect to two sets of axes, If u, is a vector and we form the product S ik u m we have ,
dxk du/ 3 a tensor of order 3. But now put m = k and add for all
du/

.
sx r^dxr ^ ak, {a " Ui)x

values of k. Since 8 lk
different from zero is
=
that for k
except for k
= i, and
= i, the only term
this is ut . Hence
= Oii^kl
CXi.
(18),
8 a tt* = u, (22).
This operation therefore replaces the sufnx k by i. The

so that the rule of transformation ia as in (11).
tensor 8 rt can therefore be called the substitulion tensor.
Since z, is a vector, it follows that dx i jdxk is a
tensor of
In the tensor w^ let us put k = i, and in accordance
order. But dx,jdx is unity if i = k and zero
if
the second k
with our convention add for all values of i. Then the
i ^ k. Hence the sot of quantities 8 <fcl such that
corresponding quantity u^' is got by putting l=j and
8n = 22 = 33 = 1>
summing; but
8 = 8I3 = 8^ = 8, = 831 = 8^ = O, w,,' = a tl a k ,w tk = 8 tt w ft
constitutes a tensor of the second order. We can prove this = w (23).
dircctly; for if we apply (11), 8,,' in the new system of

Thus wtransforms into itself and therefore is a scalar.
tl
ooordinates should be given by This operation of putting two suffixes in a tensor equal
8,,' = a it ak Ak (
19 )- and adding accordingly is known as contraction. In general
gives a new tensor, whose order ia less by 2 than that of
k has to take all values 1, 2, 3. But if k # i, it
The suffix
instance we contract the tonsor
zero. If k = the original tensor. If for
O, and the corresponding term is
i,
8 ik is
with regard to A UiVk , we obtain
8 rt = 1, and the result of the summation is
u v, =u vl + Wjt>2 + 3 w3 (24),
8' = a tl a n (20). { l
x/ with which is the scalar product of u ( and v k .
But the a a are the direction cosines of the axis of

Similarly the tensor uivk yields the 8ymmetrical and
regard to
regard to the x ( and the a are those of ,' with
,
antisymmetrical tensors ( t +tt w t ) and (u i vk uk v ). We
Hence a,^, is the cosine of the angle between x/ and
fc i
*,..
maycall these the 8y?nmetricala.ndantisymmelrical products
z,', and is equal to 1 if the axes are
identical and to if they
of u t and vk .
are perpendicular. It foUows that the result of the trans-

CARTESIAN TENSORS CARTESIAN TENSORS 9
8
The tensor ou kjdx gives similarly, on contraction, a scalar

We notice that the second and third of (27) both contain
{
(o,,, aa , Oj,). We may therefore solve thom for tho ratios
3 , 9t 3s
= 3i + 3a^
(25), of these quantities. Thus
dx t 8x, dfy
gives also a3 i
which is known as the divergence of ; while it e* 21
-4
O23O32 ^33 a 22 a 3312 a l3 fl 32 *13 fl 22 "2312
symmetrical and antisymmetrical tensors
(28),
+ dx- ** Substituting in the first of (26) we get
te< k sr^- say.
= k{ail - -
especially in the l (aâst a^dn) + o, (0330,2 a l3 o w )
The former has important apphcations,
theory of elasticity and hydrodynamics the ;
latter is + 3! (0,3022 - OÔjj)}
known as the curi or rotation of u ( The vanishing of the . -k an a 03, (29).

of a scalar.
curi is the condition that u, may be the gradient 0,2 Oag Ojj
considerations can be extended to any
Ali the above
13 23 fl33
number of dimensions. In n dimensions a tensor of order r

Also
has n' components. A tensor of order 2, in particular, has
antisymmetrical, the n diagonal & {(023032 - 0330.22)* + (033012 - O1SO32)
2
+ (aian- <hz<hi)*) = 1
n 2 components. If it is
(30).
components are zero, and the others are equal and opposite But we have a general identity
tensor of order 2 has
in paira. Hence an antisymmetrical
\n (n - 1) independent components. If n = 1, 2, 3, 4,... in
(o + b' + c 2
)
+ i' 2 + c' 2 - (aa' +
(o' 2
+ ec')* ) W
It happens that in three = (bc' - cb') 2 + (ca' - ac') 2 + (ab' - ba') 2 (31).
turn, this number is 0, 1, 3, 6
dimensions the number of numerically indopendent com- Hence
ponents of an antisymmetrical tensor of the second order is
& 2 [(w 2 + o*3 8 + asa
2
K2 2 + OM* +
) o3, 2 )
equal to the number of components of a vector. Actually it
- (u + <hi<ha + )*] = 1 (32).
can be proved that with any vector we can associate an anti-
But on account of the second and third and the first
of (26)
symmetrical tensor of the second order, and conversely.
of (27) the expression in brackets is unity, and therefore
This is not true in any number of dimensions othor than 3.
Since the a are the direction cosines with respect to the s, k = 1 (33).
of three perpendicular linos, they are connected by six re-

For any given transformation the determinant in (29) is
lations o,,* + aa* + a
2
= 1
]
therefore equal to 1. Evidently its sign is reversed if we
interchango any two of the suffixes j, for this interchanges
o, 2 + 0^ + 0^=1 (20),
two rows of the determinant so that the sign is a mattor of ;
2
o 13 + <ha* + a
33
r>*
:
the numbering of the axes. If we start with a rigid frame
a Vi a i3 + a 2i a a + aâ^ j attached to the axes x t and rotate it continuously till it
,
0| :1 o,, + "a^ + a 333i = (27). is attached to the axes z,', all the a (l vary continuously and
therefore the detorminant cannot change from + to

OaOn + a 3i a 32 =
1 1
an an -t- .
a jct
or from 1 to + 1. If then x goes to x \ x% to x2 x x

', and a^
Those relations are of course identical with those ex-
to xa', the dctorminant is initially
pressod in the usual notation of solid geometry by
1
= 1 (34), ^= m 3 - m^;
2 ?j = m2 l, - n^; n^ = m - l2 3 ^
(40).
1 Now suppose that u t is a veetor, and consider the set of
and therefore quantities

&= - 1 (35), '=/0 3 ~i\ (41)-
-% i
and the determinant forraed by the o is always + 1. (
\ Itj -m, 0/J
If we have a frame of axes (x l xi x3 ) we can turn it by a
continuous movement so as to bring ar, along the old a^, Apply ( 1 1 ) to this, taking i to be the number of the row and
xt along the old a:3 and , a^, along the old xx . In this case we k that of the column. We see that v^ enters as
- ma.
and as %-
have Its coefficient in w,{ is therefore OyOg, 03^,,
x t = xi '
, xa ' = x3 , x3 ' = Xi (36), and in all
and wn = ~ asi <hi) u \ + - <ht<hi) ui
(<h) a3i (3>On
Ou=0, aa = 1, o,, = 0; a12 = 0, an = 0, 0^ = l;l + (o ,ai ,-o2J o ,)3

1 1 (42).
(37).
Oia-1, O0, Oa-0 J This is obviously zero ilj = l. If j # and if
l the other axis
'
perpondicular to */ and a:,' is xn and jinjin is a cychc
The determinant of the a (j is therefore ,
order, the quantities in braekets are equal to (oln , ain , <%).

1 = 1 (38), This is immediately succeeds j in the order. If l
true if l
1 precedes j by one place the signs are reversed. Hence if
1 j = 1 and l = 2, or j = 2 and l = 'S, or if j = 3 and l = 1,
Any rotation of the axes that does not alter the

wn = a in Uf = un (43),
as before.
cychc interchange of suffixes 1 2 3 1 2 3 1 ... therefore and in the alternative case
leaves the determinant equal to unity, and therefore so
long as we always use right-handed or always left-handed
wH = - < (44).
axes the determinant of the a tt is + 1.

Thus wn = '
/ ,' - t (45),
With this restrietion (

- a' u,'
all = an hi ( ~~
'hl^w'i tt21 = 13 a32
a33 fl is;
(39),
Oj! = OÔj, and is of the same form as (41). Thus with any veetor we
0230,2
can associate an antisymmetrical tensor of the socond order.
and therefore every direetion cosine is equal to ite first
Conversely with any antisymmetrical tensor of the second
minor in the determinant. order we can associate a veetor.
so that the antisymmetrical tensor associated with a vector

We can proceed alternatively by considering the set of
can actually be obtained from it by multiplying by e ikm
quantities e, km defincd by the condition that if any two
,
of i, k, m
are equal the corresponding component is 0; if and contracting.
i, k, m are all unequal and in cyclic order, the component Again, suppose that we are given a tensor of the second
is + if the order is not cyclic, the component is 1. order w^ and that we form a vector um by multiplying by
1 ;
Let us see whether this is a tensor of the third order. K so, e ikm and contracting twice. We have, if m= 3,
we should have S = w"!* = 123 'U + 213 'SI = Wit ~ WH
/In ~ a H a kl a mn e ikm Thus um = e ikm w ik (51).
If w ik w symmetrical, this evidently gives zero. If it is

- (hi a u aan - <hi a*i ai-> - au<hi<hn (
46 )-
antisymmetrical the components of um are numerically
Now if, for instance, j = l, the right side is clearly zero and
twice those of w ik .
em = 0. If j, n are all unequal, the expression is

'
l,
On account of the intimate relation botween the vector
<hi an <hi and the antisymmetrical tensor we donote
shall habitually
(h, an a3I (47), the tensor w (k of (41) by u ik so that ,
Oln n a 3n U = 8g= 33 = 0; i2= 3, 23 = 1 Sl = 2;j .^

which is equal to 1 if jin are in cyclic order and to - 1 if = 3. M = 1. 13 =
2
'
Jl
not. Hence the set of quantities e ikm is transformed into It willalways be seen at once whether the vector or the
itself by the rule for transforming tensors of order 3, and tensor is intended, since the former has one and the latter
therefore constitutes a tensor of order 3. This is called the
two suffixes.
alternating tensor.
If we have any three vectors w,, v { , w it and consider the
Now consider the product e tkm u v , where p is a vector.
scalar e lkm u vk w m we see that
t ,
This is a tensor of the fourth order. If we contract it by

putting p = m
and summing we get a second order tensor Cikm u t
v* wm = 12'3 + V-tV 3W x + al^tt,
w = e ikm u mand k = 2, the only value of m that
. If = 1
- ttgi;,^ - u9 vi wx - j^Wj
ik
makes e 12m different f rom zero is 3, and then e ikm = + 1 Hence .

Ki s 3 (53),
%= O, (48). V, v2 V,
If i and k = 1, m is 3; but 213

= 2 is the reverse of cyclic w x
wz w3
order and % L Hence
so that we have a concise way of writing the determinant
fa- - U, (40).
formed by the components of three vectors. If any two of
Similarly we find that the elements of w ik are
the vectors are parallel this scalar vanishes.
/0 w3 -j\ In associating a vector with an antisymmetrical tensor
(-, vA (50), of order 2 a sign convention clearly arises. make the We
\ u* u, / positive signs in (41) he one place to the right of the leading
diagonal. If then we havo two vectors u, and vk their anti- , If i = m, k = p, the component is + 1, unless i also = k,
symmotrical product is u vk u k v iy and in tho associatod

t i = p, k = m, the component is
unless i also = k, 1 ,
vector we give the positive sign to w ik when k follows in i = k or m = p, the component is 0.
the cyclic order. Hence the components of this vector are These rcsults apply also to the tensor
taken to be 8 Im S tp S /V 8 km ,
- - V^Vg, t^V, - j V,). and therefore

(UiVg 3 g, 3!
!*.< = &tmhp - 8 ip 8ft (55).
This vector is porpendicular to both the original vectors ;

f or
i (ua v3 - u*,v2 ) + ut (U^ - ttiV8 ) + s (ihvt - ,,) = 0, EXAMPLES

1. If tif, u/, uk are the components of a vector with rogard to
'
v, (ihv3 - u^Vt) + v2 (vz vl - MjVa) + va (m,Uj - u,^) = 0.
three sets of axes, provo that the valuos of uk " are the samo as
We call it the vector product of u ( and vk and can save
,
would be obtained by transforming Urat from ut to w/ and then
from ut' to uk ".
writing by denoting it by [u, v] m .
= 'a*** "
Similarly with the antisymmetrical tensor ^ ^
'
we
2.
3.
Prove that
Show that
8 3: -
associate a vector so as to leave the sign unaltered when 'okc'tkm

" 8|8 6t Scm Sa ,S im Sck + iak S bm Scl
k follows i in the cyclic order. Thus the components are &ak&*t&cm + 8<mAf5c* ~ &tm*t>k&cf
/du _dut duy _duz du3 _du1 \ , * 4. Provo that ,<*, - 28, n ; t, km t, km - 6.
? y
54 h
\dxt dx3 '
oxa 8as,' a! dxj Prove that - -
5. <*. ,rt P *.(,..
This is often called "curi u." 6. Prove that if u vk wm are vectors,

t , ,
The Tensor e lk,e mvl Since this tensor is the product of

.
t* [. film - {<) - "m (V-.)-

two third order tensors, once summed, it is a tensor of
m [ Om " ftm u i'* wmt
the fourth order, t, k, m, p being arbitrarily assignable.
Evidently if i = k or m = p, the corresponding component
7. If A (u) u m
is zero. ti a "a
If i = m, the contribution from any value of s is zero n s "ss
unless also k = p, and then prove that
A (u) -
e lfk , mn uu, m Ub,t
e ikt = e mt = *>
. Mim M*n " clmn A (U),
and the component is + 1. 6A (U) - taHtanttdttjmUfe,.
If = p, then no value of s givos a contribution unless 8. Use Ex. 7 to prove the rule for the multiplication of deter-
= m. Then one of e ik , and e miu is + 1 and the other - 1,
jfc minants
<ta , m A A = 6A (u) A = 6A {uv),
and the component is 1. Hence the components of the (m) () (..)
tensor are as follows. where ()&> - u u v ip-

GEOMETRICAL APPLICATIONS 17
the distance between x, and y t is given by
(yt - xt )
% = (sm (
- rl,)*
OH A PT E R II = s* + r2 - 2rsl l m i
(10).
GEOMETRICAL APPLICATIONS But this quantity is also equal to s 2 + ra - 2rs cos 6, whcre
6 is the anglo between the lines. Honce the angle between
The displacement from any point to any other obviously two intersocting lines is given by
constitutes a vector. The distance between the points is a =
cos 8 Z,m, (11).
scalar. x ( y, are the coordinates
If ,
of the points and r the
distance between them, If two have the same direction cosines they are said
lines
- to be parallel. If two lines do not intersect we can take a

r 2
= (y, x ( )* (1),
line through any point on one of them parallel to the other;
the square on the right indicating the scalar product of the then this line is inclined to the first at an angle given by ( 1 1 ).
vector into itself Also the quantities (y, x,)jr constitute
.
We can then use (11) to detormine a unique quantity
a vector. associated with any two lines, which we may call their
If we take a fixed point a,- and consider points given by
inclination, whether they intersect or not.
xt = at +lr t (2), If we have a line given by (2) and y, is a point outside it,
where r is a variable scalar and the lt are constants such the line joining a, and y, subtends a right angle at x ( if
that
[Vi - *) = (x, - a,) 2 + (y s - *,)
lt* = 1 (3),
(x, - a<)* = r2 (4),

- rH? + (y t -a,- rl t )*
so that r is the distance of x ( from a,. If we take anothcr = (y< - <)

a - 2W, (y { - at) + 2rHS (12),
point y i such that and therefore

y, = a, + lt 8 (5), r = lt ( Vi - a,) (13).
(y, - <r,) = s*
2
(6),
This gives the projection of the displacement y (
a, on
(y, - x,Y ={s- r) (7),
the line. The foot of the perpondicular is
and theref ore the distances between a ,x ( and y, are such
+ l<r = a + -
i ,
a, l ( lk (y* ak ).
that the sum of two of them is equal to the third. Thus the
t
points are on a straight line; and (2) givcs the equations Evidently r in (13) will be the same for all points y< such
of the line in teras of the parameter r. The Z, are the that i,y ( is constant. Hcnce
direction cosines of the line. = (H)
hy< s
If we take two lines through a, given by
x,= a, + lj (8), represents a plane perpondicular to the line.
we take two intersocting lines given by (8) and (9), we
y t
= a, + m 4
a (0),
If
18 GEOMETRICAL APPLICATIONS GEOMETRICAL APPLICATIONS 19
can find the equation of the plane containing them as For any point on (8),
follows. If this plane is
n,z t =p (15),
CikmVklm = tt-m (* + h*) L = (*m *J m (25).
This a constant vcctor for all values of r, and may there-

is
this cquation must be satisfled by x, and ?/, for all values of
fore be considered as a property of the line. We denote it
r and a. Honce
n at = p by l,'. Then we have six properties of the line given by
t (16),
M =0
TOiTO, =
(17),
lu l '. These are coordinates of the line. They are connected
(
by t wo relations,
(18),
and from (15) and (16),

V= 1 (26),
,(z4 -o,)-0 (19).

M/ ~ **** -0 (27).
The l/ have a geometrical interpretation. Thus if we con-

Then (17), (18), (19) are three horaogeneous equations in
sider the plane
the n it and can be consistent only if
ar ,
2 -x i =l3
l
1
'
(28),
em (z< - a,) km m = (20).
this plane passes through the Also if x3 = at x3 = <%,
line. ,
This is the equation of the required plane. Also the n are

t (28) is satisfied for all values of xlt and therefore if we take
proportional to
a through a parallel
line t to the xl axis, (28) represents the
eikmhmm IhVh ~ k m i> k 1n\ ~ k mn> tin - k m i) (21).
= plane through (8) and this line. Two such planes determine
But n, 2 = 1 (22),
the line, and therefore the Z, and l/ together determine the
line.
(kh ~ h) + (Zm, - Z,m 3 )* + (^m, - Z2 m,) 2
2
we have two non-intersecting given by

= ft 2 V + V) 2
I-
K
+ a 2 + m) - (^m, + Jjm, + Z3 ? 3 )*
If
x, = a, +
lines
= - cos 2
1
rl { (29),
= sin*0 (23). yt m.
pt + sm, (30),
the line y, = fa + rl {
Thus ain 6 n, = 6 m ^m m (24).
(31)
The ambiguity in sign corrosponds to passes through fa and is parallel to (29). The plane including
a general ono in
spccifying the parameter r of a point on a line. If r in (8) (30) and (31) is, by (20),
is taken negative, we get a point on the line on the opposite -

'om (z* Pi) hm m = (32).
side of a from those given by positive values of r. But if
(
This therefore represents a plane through (30) parallel to
we reverse both r and the l we still keep l, 2 = 1, and we
t
(29). The plane through (29) parallel to (30) is
still have the same point. We may take either direction
along a line to be that of r increasing; if we reverse the m (z< - <) hm m = (33).

direction the signs of all the lt are reversed for the same The distance between these planes is the projection of the
point.
line joining any two points on them upon a line perpon-
dicular to both. If a lino perpendicular to both has direc- If we have a line given by
tion cosines n,, the shortest distance d botwccn the lines
x( = a, + rl t (39),
is therefore given by
and f} t
a point not on the line, let us suppose the point j8,-
is
d sin 6 = (fi t a ( ) n f sin d turned through an angle 9 about the line. The foot of the
= (fr
- a t) e ikm lk mm normal from /3, to the line is given by
= {etk m Ptkm m - eucmathm,,,} (34). r = l, Wi - ,) (40),
and therefore the displacemont from the foot of the normal

But e ikm p mm = -
i
mk '; e ikm a t lk = l
m
'
(35),
to /3< is equal to
and therefore - - IA
ft {, + hh (& - a k )} = lk * (p t - a() (ft - ak )
d sin d = ( lt m mt
'
{
l t '),
80 that, apart from the ambiguity in sign, (41).
d sin 6 = l( m + t
'
m,Z,' (36).
The magnitudo of this displacement, p, is given by
Thus the shortest distance is directly expressible in terms P = (A -

2
<*) - {h ifii - *)}
a
of the coordinates of the two lines. = lk

(ft
-
a y - l lk (/3, - a,) (& -
t (
ak )
Now consider two intersecting lines = i % W, ~ 1) ~ h (ft - *)}
2
(42),
x{ = at + rit; Vt = a, + sm, (37). the J being needed because in the double summation each
pair of values of the suffixes would occur twice.
The area of the triangle ormed by a ,x i ,y is \rs sin 0.
f i l
The plane through and the line is
The projections of these points on the plane a^ = are /3,
(0, a;;, a z ), (0, Xi, x3 ), (0, y y3

2 ) and
, form a triangle whose nt (z, - at) = 0,
area is subject to n,lt = 0,
1 2 3 = i 1 2 <*3 n< (fr - at) = 0,
1 xt Xs Z2r Zgr
and is therefore
m,a WI3S fam fe - ) h (A - am) = (43),

1 y-2 y3
= - while the n
are proportional to e lkm lk
t (/? m - a m ) and there-
Irs (Za ?/i3 /gwt,)
fore equal to e ikm lk (&
am )/p. -
= rsa, sin (38). If now we turn jS, through an angle 6, it receives a dis-
Thus the projections of a triangle, and therefore of any placement p (1 cos 0) along the normal to (39), and a
plane area, on the coordinate planes are in the ratios of the displacement p sin 6 along the perpendicular to the plane
direction cosines of the normal to the planes. A piano area (43). If it goes to y,, we have therefore
can therefore be treated as a vector whose components are yt -p = - t (1 - cos 6) lk {lk (^ - a t)
- l( (& - ak )}
proportional to the direction cosines of the normal. einec itm -a m
lk tfm ) (44).
If 6 is a small angle and we neglect i

, the displacement is
for k v* i; if k = i, the coefficient is (1 cos 0) (1 Z,^.)-
siinply
Thus
y, - /J, = ettra i* (&,,
- am ) (45),
6 = (1 - cos 0) (J,4 - 8,*) - sin c (k (52),
where c It is the antisymmetrical tensor corresponding to

and in particular
lm namely
,
Yl- &" i* (A - a) - *3* (A - a,)} (46), /0 k -h\

and from the additive form of this equation we see that the l -k k J
(53).
displacement is the sum of those given by separate small \ 1. -4 0/

rotations l( B about axes through a t parallel to the co- Thus the displacement is represented in general by b ik fi k ,
ordinate axes. Conversely, displacements due to small where b ik is a tensor of the second order, expressed as the
rotations about axes through a point can be added vectori- sum of symmetrical and antisymmetrical tensors. The anti-
ally as were applied to the system in its original
if all symmetrical part is seen to be of the first order of magnitudo
position, and give the same total displacement as if they in d and the symmetrical part of the second order.
were compounded into a single rotation about an axis by
the vector rule. We still have, however, to establish a sign
convention. We decide that 8 is to be taken positive if a EXAMPLES
turn about the axis of Xg is from x x towards a:,. This would
1, Givcn that the general quadric aurface is
make
S B lA lk x,xk + B iX( + C = 0,
y,-A=-0(A-*); y,-A = 0(A-i) (47)
prove that the locus of the mid-points of parallel chords is a plone,
with ll = lt = 0, l^ = Hence 1 . and fiiul the condition for this piano to be perpondicular to the
chords.
7 ~i A = <*mW Wm ~ m) (48).
If we write 2. Find the condition that the line
lk 9 = mk ,
l{ Xf +p =
yi -&-** (A -*) (49). may touch the C|uadric S.
For instance, if am = 0, we have

Vi - A = B> 2 /?3 - ra3 A; y2 - A= ro 3
A- bj,03 ;
y3 ~A= A ~ * A (50).
For finite rotations we return to (44) or (41), keeping the

positive sign in the second term. We may transfer the
origin to a t to save writing. The coefncient of j8t in y, /?<
is
b* = (1 - cos 0) l<lk - sin ee ikm l m (51),
PARTICLE DYNAMICS 25
is actually measured by the acceleration. If we have the

three acceleration components of the particle 1 due to the
C H A PT E R 111 particle 2, we can find the acceleration in the direction of

z/ by the formula (4), and the form (1) will become
PARTICLE DYNAMICS mx/ = LZ/ (5),
The essence of particle dynamics is that the second de-

provided that we define X/ by the rule
rivatives of tbe coordinates of a particle with regard to the X,' = aX, (6).
time are equal to functions of the position and velocity of The meaning of the force in any direction requires de-
the particle with reference to neighbouring particles. The finition inany case: if we define it by its relation to the
relations therefore provide a set of differential equations acceleration component in that direction, it follows auto-
to determino the coordinates. The equations for any matically by (6) that force is a vector. It follows that the
particle can be put in the form sum any number of forces, obtained by adding the
of
mx,= 'LX l (1),

components separately, is also a vector. But the practical
importance of the idea of force arises equally from the fact
where m is the mass of the particle, X ( are the forces due
that in many cases the force components are known once
to the other particles, and the summation is for all the
for all from experience as functions of the coordinates and
other particles*. matter of experiment that thi8
It is a
volocities. The form (6) then says, as a general principle,
form holds when the axes are a certain type of Cartesian
that the forces in any direction are additive.
axes, which we call non-rotating and unaccelerated, or in
If we form the contracted or scalar product of mx, and
brief dynamical.
EJf, by the vector x t we get
Ifwe take a different set of non-rotating axcs at the
,
same origin we have mi( t - ZX,a-, (7),

x/ = ax { (2),
the left side of which is
j (im.-c, 2 ). (The square of course
and since the axes are not rotating the a lt are constants.
Hence by differentiation implies the product of x f by x t and therefore the summation
= for i = 1, 2, 3.) We write
*/ ati*t (3),
x,' = a a xt (4),
T = tm&f = \m (x, + s
*2 * + i3 2 ) (8),
and therefore velocity and acceleration are vectors. and call T the kinetic energy of the particle. Then by in-
The force X, on the particle duc to some other particle tegration with regard to the time from ^ to tf, we get
is measured by the contribution to mx due to the other
[pr-sf x
t
particle; that is, the part of mx, that would disappear if t ttt dt (9).
the other particle was removed to an infinite distance. It

But in any short interval of time dt, x t dt is the inercaso
* Cf. Jeffreya, Scicntific Ip/crence, Chapter vni. for a fuller analysis.
of x,, namely dx ( . Hence the right side is equal to
3 ,CT
26 PARTICLE DYNAMICS PARTICLE DYNAMICS 27
If U exists, then in all possible small displacements of

2 X (
dx teken from the
t
initial to the final position of the
the system
particle. We call the scalar X (
dx, the tvork done on the 2EAV&, = dU (14),
particle by the force Z,- in a small displacement dx (
. Then where the dx it 3n in number for n particles, are all in-
S X,dx is the total work done by all the forces on the dependent. The coefficient of dx for any one particle is (
{
S-J, the total force in the direetion of x on that particle.
,
particle during the motion. We have thereforc the scalar
Hence the force on a particle in the direetion of x ( is
t
relation that in any motion of a particle

dU/dx,, which is of course a veetor.
Increase of kinetic energy = Work done on the particle W<8 can also take the scalar produet of (1) by any set of
(10). small quantities 8x,- whatever that constitute a veetor.
In a system of particles we may add up the equations ( 10) Then
for all the particles. We now take T, the kinetic energy of mx,8xi = I,X i Sx i (15)
the whole system, as the sum of those of all the particles. is a scalar relation; but as the Sa:, are arbitrary we can
Thcn we get equate their coefficients and regenerate the equation (1).
If we now integrate with regard to / from t to < x we get
\tY' = [i $*,*]* = 2zJ''A\<ic, (11),

f \nxtSxidt = f
' I.X Sx i i
dt (16).
Jh ih
the double summation implying summation for all pairs of The left side, on integration by parts, gives
particles the first
; summation is producing
for the particles
r T* r'
1
d
the forees X and the socond for the particles acted on. It mi,SxJ mXijSx,dt (17).
may happen that the

{
X are all funetions of the coordinates

t
t
But we can consider the x t + Sx ( as coordinates of a par-
alone, and not of their volocities, and that provided the
ticle in a motion differing slightly from the actual one that ;
initial and final valucs of the x t are the same the integral is
is, at a given value of t, the coordinates are x t + Sx, instead
the same however the x may vary in the interval. If so,
{
of x (
. Then
the integral is the difference between the values of a certain
fnnetion U for the initial and final positions of the system;
d ~
8x ' = d ,
X +
. . rf
dt dt( '
SXi)
-dt Xi
we call the system conservalive and U the wmh-funetion. = xi + 8x,-x i
Then (11) becomes
= 8x { (18),
Hl-K (,2) '

since x { +
of the varied coordinate x -f- Sx
Sx
simply the varied velocity or rate of change
Then
(
is
T V=
t (
or constant (13).
f' d P'
mx -,x dl=\ mitSZjdt
tho equation of conservation of energy. The quanti-
i l
This
|
is
ties on both sides are scalars. U is often donoted by V l
and called the potential energy. \

h (\mx?) dt +O (Z t )* (19).
- lu
28 PARTICLE DYNAMICS
Then for every particle, to the first order,
f'' {8 (tm&t*) + ZX,8xi) dt = Im^Sa:.! '

(20).
C HA PTER IV
We can now add up thcse equations for all the particles of
the system. If U exists we can espress the result in the
DYNAMICS OF RIGID BODIES
form
A rigid body
one such that whenever it is displaced the
is
f' S (7* + U) dt = S r^S*,]'' (21), distance between any two particles of it is unaltered. Since
three particles A, B, C are in a straight line if the sum of two
or, sincethe hmits are not varied, we can move the 8 out- of the distances AB, BG, C A is equal to the third, it follows
side the integral. If the variation is such that the initial
that straight lines are unaltered by displacements of a rigid
and final positions of the system are unaltered, Sx { = body. Since when A,B,C are not in a straight line the angles
when l = t
u or l lt and we have of the triangle ABC are determinate funetions of the three
h sides, it followsthat all angles are unaltered by displace-
s[ (T+ U)dt =
lu ments a rigid body. If three lines meeting at a point and
of
fixed in the body are mutually perpendicular before dis-
to the first order in the variations of the path. This is
placement, they are still perpendicular after the displace-
Hamilton's principle.
ment.
The equations of dynamics in the form mii = X ( are
EXAMPLES true with respect to dynamical (that is, non-rotating and
1. Tf the Cartesian coordinates of ovory particle of a system are unaccelerated) axes. Let any particle Q of a body have
known functioiis of a sot of generalized coordinates q r , prove that coordinates with reference to dynamical axes at O. Then
a;,
(d /dT\ ari. let the body be displaced in any way, and let the particle
have the new coordinates z/, We require the relation
dZ between x and x ( Consider a particle of the body, P say,
(
'
.
lX i
&el ~l.X l
'Bq r ,
whose coordinates before and after the displacement are
d
where the summation convention is also understood on the right o, and a/. Put
side. Deduce Lagrango's cquations for the case where the q r are
all independent. *< = a< + yr, x/ = a,' + y/ (1).
2. The equations of motion of a particle are m = i

X t
fci, Then y and y/ are the coordinates
t
of Q with respect to
where k is constant. Prove that axes at P parallel to the dynamical axos before and after
the displacement. Also if we imagine the original axes at
2T = - xi X i + f, Qmft + (m?). P to be specified by the particles on them, these particles
Hence show that for a system in periodio motion, or in one slowly in the new position still specify a set of rectangular axcs,
changing its state, on an average over a long time, 2T 2x i X i .
with respect to which the coordinates of Q are still y t . If
30 DYNAMICS OF RIGID BODIES DYNAMICS OF RIGID BODIES 31
where are the direetion cosines of the axis of rotation.

the cosine of the angle between the yk axis in its new posi-
Z,
tion and the y t axis in its old one is a ik we have theref ore ,
Thus the a ik are determined in torms of this axis and the
angle of rotation. Conversely, if we take the symmetrical
y/ = o.ikVk ( 2 ).
and antisymmetrical parts,
whence x/ = o/ + a ik (xk - ak ) (3).
(l,lk - 8 (k ) (1 - = cos 6) J (b <k + bu) (10),
The displacement of Q is
a,' -a + a - (4). J (M. - c*.) lm sin 6 = J (b (k - bu) (11).

t lk yk y,
We can prove that thore ia a straight line of particles in the But for a given m, with ikm in cychc order,
body such that the displacement of Q is the same as that
*<m - <*m = - 2 (12).
of P. For such a particle we should have
- = 5 ). Thus l, n sin = - J {b u
- bkl ) (13),
a ( kVk ~yt= (o<* s ) Vk (
and these three homogeneous equations in yk aro consistent and

provided
sin 9 = - 621 ) + (6a - 632 ) + - 6 13 )*} (14).
II
o* -8* = 11 (6), J {(612 (63,
that is, a
'n - 1 <hi
= (-') Now suppose that the displacement is small. Then a ik
have nearly their values for zero displacement, that
a- 1 fl23
is,
S tt . The direetion cosines of the yk and y, axes in their new

3 38 - '
positions are a ik and a n and thus, if k = l,
,
Now the determinant || a ik ||

is unity, and each element in
it is equal to its first minor. If we expand (7) we get a ik a,i = (15),
II ft II
~ {(U 22 - Ol2 au) + (022033 ~ 0*3032) and, if k = l, oa = (, 1 (16).
+ (O33O11 - OsiOis)} + (<hi + 0,4 + 033) -1 = (8),

But in (15) for i = k, a ik is nearly 1and a n small f or i = l, ;
since the terms all cancel. Hence (5) have an infinite a tl is nearly 1 and a ik small for i ; not equal to k or l, both
number of solutions, ail proportional, and the points there- a ik and a n are small. Hence to the first order
fore he on a straight line.
If we take any two planes through this line, the angle
o- + Oi., = (17).
between them is the samo after displacement as before, If k = l, then f or i ^ k or / both terms of ( 1 6) are small of the
and therefore all planes are rotated through the same angle. second order, and therefore for i = k or l a lk = 1 + a ,
Thus any displacement of a rigid body is equivalent to a second order quantity. Thus to the first order
displacement of a particle of it combined with a rotation
o* = 8<k - 6 (18),
about a line through that particle. If the angle of rotation
is 9, we have, by comparing (5) with (52) of Chapter n, where b ik an antisymmetrical tensor.
is
b ik = 8 The displacementof Q is 8a, -f- a ik yk y u where 8a is t
= (M* - 8) the displacement of P, and is a first order small quantity.

(1-cosfl) -<*,*, Bin (9),
Now let X, be the total force acting on the particle at Q. sider the displacements that take place in a short time
In the displaccmcnt the work it does is the scalar interval Si, and put in the limit
8W = X ( (8a, + a ik yk - yt) 8af = u,8t; ba = (oÛ (24),
=X t (8a, - b lk yk ) and call u t the velocity of P and tn tk the angular velocity

of the body. Then the velocity of Q
=X t Sa t - J6, t (X iyk + XkVi ) - \b ik (X iVk - X kVi )
is
(19). x = u
i i
-w ik yk (25).
When i and k are interchanged, b lk and X X k y, are

Now consider the centroid 0, with coordinates 7:
t de-
t yk
fined by
reversed in sign, while X t yk + X k y ( is unaltered. Hence
(2m) x = Xmx
t t (26).
8W = X i 8a i + i (y Xt - y k X ( {) b ik (20).
usually assumed without proof that
It is G is fixed in the
Now write y( Xk - yk X = L a t (21), body, though this is not obvious. But suppose that the
80 that an antisymmetrical tensor. The second term
Z/ tt is
particle atG with coordinates $i in the original position of
the body goes to G' when the body is displaced, its new
in (20) is the sum of nine terms, of which three are always
coordinates are ,', given by
zero and the others equal in pairs. If we replace i tt and
L lk by the associated vectors, we have 5/ = a/ + a tk fa - ak ) (27),
S W = XM + Lm b m (22),
and the coordinates of the new centroid G" are B/, given by
where the Sa, and b are all indepondont of one another (Sm) 5-,.* = Znw;/ (28).
and the same for all particles of the body. Honco if we add We have to show that G' and G" coincide. We have
for all particles of the body (Sm) J," = Sm {a/ + a -
ik (xk ak )} (29),
8W =&X )8a i i +(ZLm )b n (23), and therefore
and the total work done in any given small displacement is (Zm) (x/ -x t
') = Sm {o/ + a lk (xk - a k )}
determinate if we sum up the forces acting by the six
- (Zm) {a/ + a lk (.7t - a k )}
expressions SX<, SZ/. Further, the contributions to
2Z Si m from the internal reactions are zero. This follows = Zma xk 'Lma xk
ik ik
( ,
at once if these reactions consist of equal and opposite = (30).

forces between pairs of particles along the line joining Thus the particle originally at G is displaced to the new
them, and also has the justification that it leads to correct position of the centroid, and therefore the centroid is fixed
results. Then we may restrict EX<, L m to the contribu- in the body.
tions from the external forces. This is d'Alemberl's Prin-
Now return to the equations of motion of a particle of
ciple.
the body
In the limiting case of continuous motion, we may con- rruii = X, (31).
By addition we form from these the equation8 the independent 8a { and b m of (23) must all
efficionts of
Stflg, = SX, (32), vanish. But these coefficients are the 2Z, and SL m , and
therefore the x ( and h ik do not vary with the tiine. The
and by cross-raultiplication followed by addition for all
vanishing of the work done by the external forces in six
the particles of the body,
independent small displacements is therefore a necessary
Sm (x xk - xk x\) =S (x,X k - xk X t) (33).
(
and sufficient condition for equilibrium.
On the right sides of (32) and (33) the contributions from The equations (38) may be put in another form. If we
the internal reactions are zero, by d'Alembert's principle. consider any moving origin O', not necessarily fixed in the
Also body, with coordinates a t ,
we can write for the coordinates
ZmZi = ZmZi = jTi (Lm) x, = Mx t (34), of a particle with respect to O',
dti dt
= xi - a{
where M = 2m (35),
y, (39).
Also If any vector associated with the particle, such as its

the total mass of the body. if
velocity, momentum, or acceloration, or the force acting
A tt = Zm (x f xk xk xi ) (36),
on it, has components u we may form the antisymmetrical
t ,
we can reduce (32) and (33) to tensor y u k yk u ( and call it the moment of the vector
t
Mx = ZX
{ t (37), about O'. From (32) and (33) we can form the equations
h* = 2Z"* (38).
Swi {XiXk xk Xi) (a t "L7iik ak 2.mx ()
dt
=S {x i X k - xk X - t) (o,SZt - o fc
SZ () (40),
These are the fundamental equations of rigid dynamics.
The three independent h tk are expressible in terms of the where only the external forces make any contribution to
three independent <o lk by (36), and we have therefore six the right side. But by (39) this is cquivalent to
differential equations for the ic and a>a , which determine

t
2m (yiXk - yk x ( ) =S (y t Xk - yk X,) (41).
them in terms of the initial conditions if the external forces
are given. The motion of the body is therefore determinate. Therefore the sum of the moments of the mass-accelera-
The principle of virtual worlc follows imraodiately. For tion products about any origin is equal to the sum of the
if a body is initially at rest, .t< and ha are zero, and the moments of the external forces about that origin.
condition that they may remain zero is that ZA, and ~L ik If we denote the moment of momentum, or angular
shall vanish. But this implies, by (23), that in any small momentum, about O' by h ik we have ,
displacement of the body the work done by the external -

7t w' = Sm (y,k yk xi) (42),
forces a small quantity of the second order in the dis-
is and
placements. Converscly, if there are six independent
possible small displacements such that the external forces g Kk = S (yiXk - yk x\) + Z.m {y t xk - yk x t)
do no work, to the first order, in any of them, the co- (43).

;

The second term may be written where B (m is the symmetrical tensor defined by
Zm {(i & t ) xk (xk uk) (} = 2m (ct,it u k x ) t B im = -Lmy iym (49),
= -M (atSk - a k Z,) (44). and depends only on the massea and positions of the par-
Tbus (41) is equivalent to
ticles of the body. In the three-dimensional case, with i
h lk '
+ M (a t u:k - a k x,) = 2 (y,X k - yk X,) (45).
and k unequal, m must be equal in turn to i, k, and the
dt other value. Then, for instance,
In many important cases the second teusor on tho left ^12' = (Buiii + Btt<iz + B.a u> i3 )
vanishes identically. This is clearly true if O' is fixed, when
(âia, 4- B yi a> a + B^w^)
the time derivatives of the a, are zero when the centroid ;
is fixed; and, if both are moving, if the vectors d t and w t

= (B& + Bu ) <o lt -6,30*23 B.a w 31 (50).
are parallel, that is, if the velocity of the moving origin 0' If we replace the antisymmetrical tensors by the associated
is parallel to that of the centroid. The most important case vectors this takes the form
is where the moving origin is identical with the centroid, h3 ' = (B u + Bn w, jBu a;, Bj3(Uj
)
when the last condition holds automatically. These terms
also disappear if the moving origin
an instantaneous is
= 5 + -#22 + -^33) "3 ~ (^13^1 + B23"2 +
( 11 833(1)3) (51),
centre of rotation always at the same distance from the and in general
centroid. They vanish for a sphere or circular cylinder hj = A<m o>i (52),
rolling down an inclined plane, but not for a rolling elhptic where A im is the symmetrical tensor given by
cylinder.
Am = Ba-Sfn B im
If the moving origin is the centroid,
= (Smjfc) 8 <m - Vmy ym (53).
hj = 2m {y, fo + yk - yk ) (.r, + y t )}
t
It evidently corresponds to the system of moments and

= 2m (y yk - yk y ( t) (46),
products of inertia given in ordinary treatises on dynamics
since Zmy = Smyt = 0, { A u An , , moments of inertia A, B, C,
A33 are the ordinary
by the definition of the centroid. Thus the angular mo- but A^, 3l , 12 are A
equal and A
opposite to the ordinary
menta about the centroid are expressed completely in products of inertia F, O, H.
terms of positions and motions relativo to the centroid, and It should be noticed that this reduction is characteristic
the formulao for them have the same form as those for the of three dimensions; in a higher number of dimensions
total angular momenta with reference to a fixed origin. there is no analogous simplification of the form (48).
By (25) y< =- a> im y m (47),
The equations of motion of a rigid body then take the
form
m being here a dummy suffix. Substituting in (46) we have
MS = t 2Z, (54),
h* = - Sm (ai km y ym
i
- w im yk ym )
= B*m<im - Blm"km (48), ^ h = SD it

'
ft
' ; or
I k m = *L m
' '
(55),
where L lk ia the moment of the external forces about the The kinetic energy of the body ia given by
centroid.
2T = 2ww,
It may happen that one point of the body ia fixod. In
that caao we may take the origin at that point. Then in = Sra (x, - a), m y m )*
addition to the known external forces there ia a reaction = Smî* + I,myk y m <o tk a> lm
at the origin which can required be found from (37); but
if
the reaction has no moment about the origin and the = Ms * + Bkm w t ik co lm (63).
motion is given by (38). But in this case In three dimensions (o ik oi im is zero unless i is differcnt from
x, = - u> im x m (50),
both k and m. If i = 1, k = 2, m= 3, or if i = 1, k = 3,
m = 2, a>l( c0 U = - w 2 a>3 . If i = 1, k = m= 2, o>M = 3

* 2
,
and we find, by a process analogous to the last one,
and has coefficient B&. Thus a>3 2 entcrs with a coofficient
bn = *m<0 (m - D tm <kn> (57), 5j2 + Bn or ^33.

, But <o 2 o>3 has a coefficient IB^ or
where D tm = I,mx x m -4m + ^m- Thus in all

i (58).
2T = M5* + A km w k a> m (64).
In terms of the associated vector,
When a point of the body is kept fixed,
(59),
2T=Ckm a> k alm (65).
where C (n - (S>** 2
) S, ra - Smx,ar (GO).
We notice that the linear and angular momenta take the
The C im correspond to the moments and products of inertia forms
about the origin. Thoy can be exprcssed in terms of those dT dT
M^t = -_ . ; h t' = 5 (* constant),
about the centroid for ; 0x1 cw (
C (m = {Sm (x k + y k y-} S tm - Sro fa + y{) [Wm + y,n ) (66)

= (Sn?) 8 in - Zmx xn t
for origin at a point of the body held fixed.

+ 4* (61),
In one respect the forms (55) and (62) are inconvenient.
the terms linear in the y's vanishing by the definition of the They involve the tensora A im and C (m which depend on ,
centroid. the Xj and j/ and therefore in general change as the body

(
The relevant equations of motion then take the form rotates. It is moro convenient to use such axes that the
relevant tensor in the particular problem is constant. To
ftK-ZLn (62), achieve this the axes must rotate, and then are no longer
dynamical axes. Supposo then that we have a set of axes
where the L m are the moments of the external forces about x/ rotating in any way, and that their direetion cosines
the origin, the reaction at the origin making no contri- with respect to the dynamical axes x, are a (j . Ali the usual
bution.
tensor relations hold for transformations from the x/
System to the x system and vice versa; but the a if while
t , If for instance the components of a displacement along the
specified for each instant, are now functions of the time. x/ axes are u/, the formula (73) gives the components of
If u, is any vector, we have the velocity along these axes if u/ are the components of
;
u/ = a ti u ( ; u, = a,jU/ (67). velocity, (73) gives the components of accoloration if u/

;
are the components of angular momentum, (73) gives the

Then j-' is another vector; and components of their rates of change with reference to
dynamical axes, and these are equal to the components of
du. the moments of the forccs acting.
(68).
The component of this vector in the x,' direction is

EXAMPLE
du/ da l}
,
^s
*
A +
-/, /
f/i
(69).
Prove that
The first term of this is
du/ du,' - A ivtU{ .
s
b
"~dt ~ dt
(70).
Also ~~ is the x, velocity of a point at unit distance along
the x/ axis, that is, a point with coordinates a tj (t' = 1, 2, 3)

with reference to tlie dynamical axes. This velocity is
Oik aki> where 6 ik is the antisymmetrical tensor exprcssing
the rotation of the rigid frame consisting of the moving
axes. Then
da-/]
a " Ul
/
= ~ .
n
a " u > e " ai<i
dt
= - 0,,'u,' (71),
and the required component is
du,'
-e./u/ (72).
dt
If we use instead of 6,,' the associated vector, the three

components become
(V - ,'*,' + u,'e,', u/ - u 3 'e,' + w,

s' - i'0/ + u x 'el ') (73).
|CI
EQUIVALENCE OF SYSTEMS OF FORCES 43
If we take the associated vector, its component in the

direction m m is
c n a pt e n v R (mJJ - e < km mm a t lk ) =R (w,!/ + Z./) (8),
EQUIVALENCE OF SYSTEMS where m/ are the other coordinates of the line through a,
OF FORCES in the direction ?,. We notice that
m,7/ + Iftn/ = dsinfl (9),

An X< acting at a point x of a rigid body
external force t
produces dynamical effects sumined u p in the vector Xi where 6 is the angle between the two lines and d the length
and the antisymmotrical tensor x t X k xk X { A force has .
of their common perpendicular. We may call Rd sin 9 the
a line of action that is, if we take its resultant R given by
;
moment of the force about the line (m,, m,').
By d'Alembert's principle, the motion of a body is un-
R = 2
X,* (1)
altered if to the forces acting on we add two equal and
it
we can dofine a direction l{ by opposite foi'ces acting at the same point, or, by (3), along
the same line of action.
X = BI,
t
(2).
Now consider a pair of equal and opposite forces X and (
By convention R is always taken positive. a force X t If X acting at points a, and 6,. They cloarly make no
t
acts at the point x{ + rl lt where r is arbitrary, we have contribution to LX,. Their contribution to T.L lk is
(x t + rl<) X t
- (xk + rlk ) X, = xt Xk - xk X t (3). (a ( Xk - ak X ( )
- (b,X k - bk X t)
Thus the dynamical effects are the sarae if the force X f = (a, - b() Xk - (a k - bk ) X t (10).
acts at any point of a line through x with direction cosines (
Since a, and 6, by any motion of the
are equally affected
proportional to X
This line is called the line of action of
(.
origin, the contribution of such a pair of forces applied to
the force, and the force can be said to act along it.
definite particles to both 2X< and 2L lt is independent of
If a force has magnitudo R and acts at x in the direction t
the position of the origin. Such a pair is called a couple,
l, , we have
x = t m t (4),
and its contribution to "ZL ik is called the moment of the
couple.
L = R (x
lk tl
k - xk l t) (5),
If the vectors a, b t and X, are both perpendicular to
or Lm = Rl m
'
(6), a with direction cosines n, the components of Z-m are
line ,
the components of a vector along this line; while the forces

in the notation of the coordinatos of a line. Thus the X,
act in the same piano perpendicular to the line. This plaue
and L m are the products of the resultant into the six
is called the plane of tlte couple, and the line aii axis of the
coordinates of the line of action.
couple. Evidently equal couples in parallel planes are
The moment of the force ( about a point a t is X equivalent. The magnitudo of this vector is Rd, where
(Xi
- a t) Xk - (xk - ak ) X =Rt (l ik ' -a t lk + ak l ( ) (7). i? 2 = AY and d is the perpendicular distance of a t from a
44 EQUIVALENCE OF SYSTEMS OF FORCES EQUIVALENCE OF SYSTEMS OF FORCES 45
lino through 6, parallel to A',; and its compononts can so that p is the same for all adinissible values of a,. With
accordingly be written Rdn m . this value of p, (16) represents three planes with a line in
Any systeni of forces is equivalont to a force at an common. If we change a, to a, + oX i , where a is a scalar,
arbitrary poiiit a, together with a couple. each For witb. the left side of (16) is increased by
A', acting at x wo can associato a pair of forces A', at a t
t
.
- eikm aX Xk =0 t (18),
Then the system is equivalent to 1jX, at a, together with
and therefore if o, is one point on the line, all points on the
a set of couples whose total momont is
line through a, parallel to X ( satisfy the conditions. This
Ma = S {(*, - a() Xk - (xk - a k ) X,} (11). line is the central axis of tho system the system is equiva- ;
lent to a force R along the central axis and a couple O

But this is equivalent to a single couple ; for we have only
about it. This expressos the system as a wreneh. If we take
to raake n m proportional to the Mm and Rd equal to their the central axis as one of the coordinate axes, we have,
resultant.
If L are the moments about the origin,
since A', 2 and XM ( t are scalara,
ik
B* =X* (19),
L tt = 2 (x t Xk - xk X { ) (12),
t
OR = X M = pX ( t t
*
(20),
and if for brevity we roplace Z A', by simply X it
and therefore G, R, and p are determined.
M lt
= La - (a ( X k - ak X t) (13), The system can also be reduced to a couple parallel to
or Mm = Lm - e ikm a,Xk (14).

a proassigned plane together with a force. For if S is the
couple, and w, are the direetion cosines of the normal to the
Evidently A', 2 is a scalar and independent of a, . Also plane, and if the force acts through a,, we have, for the
Xm Mm = X m Lm - eikm a t Xk X m (15).
moments about a,
The first term is a scalar and independent of a, . The second

Mm = Lm - a X e lkm t k (21),
is identically zero. Hence X, 2 and XM i i are scalar in- and also M m = Sn m (22).
variants. We have three equations to determine the o, and S. Again

The system equivalent to a force
is A', and a couple M ( there are a single infinity of solutions. But if we take the
at a, . These veetors are parallel if scalar produet by A',, , we get
Lm - flkmai X k = pX (16), Sn m X m = L m X m e ikm a t Xk Xm

where p a scalar length called the pileh. These give three
is
= GR (23),
linear relationsbotween the three a, and the piteh, and we so that S is determined
provided n m m is not zero, that is, X
therefore expect a single infinity of solutions. But if we provided the resultant force is not parallel to the plane.
take the scalar produet of (10) by X m we have Then the equations
X m Lm = PX m *
(H), Lm - eikn, ai Xk = Sn m (24)
46 EQUIVALENCE OF SYSTEMS OF FORCES EQUIVALENCE OF SYSTEMS OF FORCES 47
dctermino a line parallcl to X which ia the line of action so that we have in general just enough equations. Two such
iy
lines are conjugate lines. Clearly any line intersecting two
of tho force.
are the direction cosines of a line through a it the conjugate lines is a null line.
If lt
moment of the system about this line is It can be shown easily that
Mm = - Xk S (X + L,Z,.) = LfS.t
l/ (31),
m Lm eikm a i
l
m
l l l m (25)
T* = X? - 2Sl X + S a- (32),
= l m Lm + lk 'Xk (26),
i i
whence (28) and (29) determine the coordinates of the

where lk ' are the other coordinates of the line. If this
second line explicitly.
moment vanishes the line is called a null line of the system.
If b ( is another point on it,
An (b m - 0) = ikmai Xk (bm - a.) (27),

EXAMPLES
1. A is reduced to a force at P together with
system of forcos
which shows that b t hes in a definite plane through a Ali .
f
a couple ; P
chosen so that the couple is parallel to a given
is
null lines through a point therefore lie in one plane. This plane. Show that the locus of P is a straight line parallel to the
plane is called the null plane of the point. central axis.
Ali null lines in a plane pass through a point. For the 2. Show that
system can in general be reduced to a couple in the plane
and a force whose line of action intersects the plane in one
point. Then the system has no moment about any line in is a point on tho central axis.
the plane through this point, which is the null point of the
3. Two systems of forces are given by (Xit Lf ),(Y it X,). Show
plane. that XK
i
+ y ( L ( is invariant.
i
Any system is equivalent to two forces, one of which can

be made to act along a given line. For let the lines of
action pass through a, and b t , and have direction cosines lt
and mi and let the magnitudes

, of the forces be S and T.
Then we have six equations,
X = Slt +
t Tmt (28),
Lm = e ikm (a { Slk + biTm,)

= Sl m ' + Tmm '
(29).
Tho coordinates of the first line bcing given, these are six
linear ecjuations to determine the six coordinates of the
second line and S and T. But we have also
m *=
t 1; mm =t t
'
(30),
CONTINUOUS SYSTEMS 49
integral is ovcr S; and the sonso of the normal is such that

if the contour is described in the positive senso about any
axis x t li is taken positive whcn the normal in the direc-
CH A PT E R VI , is
tion of x, increasing.
CONTINUOUS SYSTEMS The gravitational potential of a distribution of matter is
given by
In problems involving volume and surface integrals we find
itconveniont to donote elements of volume and surface by
v=rz (6),
dr and dS respectively both are always taken positive.

;
where m is an element of mass, r is its distance from the
They are of course scalars. The direction cosines of the point where V is to be found, and / is a constant equal to
normal to an element of surface are usually denoted by Z, 6-66 x 10~ 8 when m
and r are measured in grams and centi-
in most cases the normal is drawn outwards from the region metresandthe time in seconds. When the mass is distributcd
under consideration. continuously over a surface or through a volume, in must be
Green's Lemma takes the form replaced by adS orpdr respectively, where a and p are called
the surface and volume densities. The work in displacing
m.' ffm AS (1).
a mass m' through a small distance is m'dV.
and we have the corollary, if The electrostatic potential of a system of point charges
u = dV/dx (2),
is given by
( (
where V a e
is scalar, F = /2 (7),
s
on
(3),
where e is a typical element of charge. If e is in electro-
8 9 static units of charge and r in centimetres,/is + 1. The work
where
dx *~dx1 *
+ dx + dx * *
(4), in a small displacement of a charge e' in the field is e'dV.
i i 3
The usual relations follow, that in free space in both
and dj&n denotes differentiation along the outward normal.
cases
Stokes's Theorem takes the form
V2 F = (8),
and in space occupied by matter of finite density

J.-Mft(-)
F-- 4nfp (9).
Also we have Gauss's Theorem

\\l<e<kmdu m /dxk dS (5).
J
~ dS = - 477/S'm or - tefL'e (10),
The integral on the round a closed contour C. On
left is
the right l, is the direction cosine of the normal to any where the summation is over all the masses or charges
element dS of a surface S that fills up the contour; the within the closed surface <S. In crossing a surface where
60 CONTINUOUS SYSTEMS CONTINUOUS SYSTEMS 51
there is a finite surface density a, dVjdn has a finite dis- where the first integral is through all regions where p is
continuity nfo. finite and the second over all surfaces where a is finite ; the
All theso relations are scalar in form. For a gravitational first integral may therefore be taken through all space,
field the force on a sinall particle of mass m' is excluding the surfaces where there is a surface density. It
can be shown that with proper precautions about the de-

m,'X = m'dVldx t
(H),
finition of V the restriction p = g gives no trouble provided
t
while for an electrostatic field the force on a small charge V ia everywhere finite.
e' ia Now if we consider the integral through all space except
e'Z, = - eW/a*, (12). thin laminae surrounding the surfaces where a is finite,
The vectors X
are called the inlensities of gravitational
(
and electric force rcspectivcly. The analogy in form be- ^\\\ **$ V d* V
dXi*
A
tween (6) and (7) is constructed for mathematical con-

vcnience; the difference in sign in (11) and (12) embodies
the physical difference that whereas two positive masses
attract, two positive charges repel.
In a gravitating system we may construct a work-
where the dn in the first integral is out from the region of
function
integration and therefore towards the surface where a is
W = sa/!"'"' (13),
finite. On the two sidcs of such a surface the values of V
'M
difler by an indefinitely small amount, and for the two
where the summation is over all pairs of particles mw dV
and rw is the distanco between ra and m The case where .
sides together - dS = bnfodS, by Gauss's Theorem.
p = q is excluded. Then the force on the pth particle is
Hence the first integral in (15) is equal to <7FdS
d W/fai), The potential at m due to the other particles is
.
J
V p = %'fmjr, the accent indicating that the case where taken over the surface and cancels the second term in (14).
q = p is excluded from the summation and the function ;
Hence
Em, V v = 2W, since each pair of particles is counted twice
in this double summation. The function Em p V v therefore '-^Hl'*
plays the part of a work-function. Similarly in electro- through all space. In conscquence of this form we may say
statics the function JSe P F P plays the part of the work- that the gravitational work-function, or the electrostatic
function with its sign reversed, that is, of a potential energy. potential energy, is R /87rf per unit volume,
2
where R is the
These results may
be generalized to the case of continuous resultant of the appropriate intensity vector.
distribution3; thus we can replace these functions by When the properties of the medium vary from place to
F is no longer of the form Z/e/r, and V 2 V is no longer
place,
W=ij\\ P Vdr+lâVdS (14),
zero. But a potential still exists; if a small charge e' is
52 CONTINUOTJS SYSTEMS CONTINUOUS SYSTEMS 63
nioved from a point P to

a point Q, the work done is still The first integral, applicd to all space except thin regions to
the same whatever the route, and may be denoted by cut off surfaces where thcre are surface densities, gives
e' (V
p- VQ ). The treatment is suggested by the fact that
two similar condensers with the plates at the same po-
tentials, but with air between the plates of one and
another
material between those of the other, have charges in a ratio
K depending only on the media. dVjdn on the outside of
/dV\ s
the condenser being small compared with its value be-
tween the plates, we infer that the charge per unit area f or
--M**-**^87rf)!}"VdxJ
nfi) dn
the same distribution of F is related to the discontinuity in The surface integral cancels the integral \ f aFrfS; and
KdVI&n, where K depends on the material. This suggests hence
in turn that Gauss's Theorem must be replaced by
W K ~
d
dS = - 47r/Se (17),
so that the energy can be considered equal to KR /8nf per
l
unit volume.
where the summation is for all charges inside S. Then Magnetism may be treated similarly, starting from the
applying this to the two sides of any surface we have assumption of volume and surface distributions of mag-
netic pole strength, subject to the condition that the total
K 7n]=~^ (18) "
pole strength in any solid is zero; or we may regard the

[
ultimate magnetic unit as the doublet, which explains the
and applying it to a region with a finite volume density we
need for the restrietion involved in the former method of
have
treatment. The potential at x, due to a doublet of strength
-MJJ>HJJ (*)* <1B >- M at the origin with its axis in the direetion A< is
V = yMX x /r
i i (24),
and as this must hold for all such regions,
and the doublet strength per unit volume in a solid is /
if
d_ ldV\ (20).
in the direetion A, we can introduce the intensity of mag-
Cj \ cxj netization at f,, the veetor A t = /A,, and say that the
potential at x {
These equations are all homogcneous in V, p, a; hence is
the potential due to any set of charges is proportional to V= A^J'd- (25),

the charges if all are altcred in the same ratio. Using this
principle we can show by the usual method that the energy where A, corresponds to the point , and dr = d$t dt dgt-
of a distribution is y is a constaut. The magnetic foree in freo space is
P VdT+
\-LVe= \ (26).
\\\
l frvdS (21).
64 CONTINUOUS SYSTEMS CONTINUOUS SYSTEMS 65
also contributes nothing to X,; but if the cylinder is

We may write it
of disc-like form it contributes 4ary\J to X (.Hence if we

'Ir
tako V for the complcte body and define a, by
a, = - dV (29),
dx (
dA< 1 . a, is the value taken by X t in a thin cylindrical cavity

(27).
JJ<- parallel to the intensity of magnetization. The force in a
flat cylindrical cavity with its generators in this direction is
The potential is therefore equivalent to that due to a dis-
tribution of magnetic poles lf A per unit area over the
t
a{ = a( + 4:TryX l I = a( + bnyAi (30).
boundary and - dAi/d t per unit volume through the in-

Evidently a and a are both vectors the former is called
t t
;
terior.
the magnetic force and the latter the magnetic induction.
Witliin a solid special treatment is needed. To define V
The theory of susceptibility and permeability may then be
or the force at z, when s, is within a solid, we must imagine
developed as usual. Also (27) shows that V is continuous
,
a small cavity made about x { the intensity of magnetiza-

,
=
BV/dn has a discontinuity
across a boundary; but l, a,
tion everywhere remaining as before, and consider V and
4-rTyliA,; whence J,a, is continuous across a boundary.
a, within it; then the values of V and a, at x ( are defined to
The mutual potential energy of two doublets M and M'
be the limits of those in the cavity when the dimensions of
at x, and as/, oriented in directions X and A/, is (
the cavity become indefinitely small. This process leads to
little difficulty in gravitational and electrostatic problems, ?V
(31),
but in magnetism the limit of the force is found to depend
on the shape and orientation of the cavity. The force in
where V is the potential at x t ' due to the magnet at x ( ;
this
the cavity can be written
gives
(28),
W = yMM%'^,(x k ^r^)
where V is givon by (27) in the first integral the normal is
;
'~
~
inwards towards the cavity. The contributions to X ( from = yMM' j-''*-' - 3 ^* Xt ' Xk Xl x<
\ (32)
the volume integral and the outer boundary are of the same
form as for gravitation, and give no trouble. If the cavity /' ' (cos e 3 cos 9 cos 9') (33),
is a cylinder in the dircction of the intensity of magnetiza- r3
tion, lt A , is zero over the sides and equal to /, the resultant
where c is the angle between the axes of the magnets and
intensity of magnetization, on the ends. Such a surface
9 and 9' are the angles made by the axes with the line
density over the ends in the limit contributes nothing to V ;
joining the centres.
if the radius of the cylinder is small compared to its length
CONTINUOUS SYSTEMS 67
66 CONTINUOUS SYSTEMS
second magnet is tnrned through a small angle
If the Sî
we have
about a line with direction cosines n { ,
by (49) of
W-camihMfy
Chapter n. Hence, by (32),
(
34 )>
[
i c
i
pox
<fo,+
tJc
X,dx f t + \
JC
n.du, (4).
The last integral is rjw, 2 ], which always vanishes because

8W =
U SV when we move round the contour we come back to the
same point, where the velocity has its original value. Also
x ~ if X is the gradient of a single-valued
potential, as when
= yMM' * h -t
3 cos 6 <
?<} eikm n k X m ', t
(the commonest case),

the bodily forces are due to gravity
80 that the couplo about a line parallel to the axis of xk is

X,dx, is the change of this potential round the contourand
^
f
-Zcose -L^}
X
M k
= eikm X m '{x t (35). is zero. Again, a function of p only, as in an incom-
if p is
pressible liquid or a gas at uniform temperature, the first
Hydrostatics and Classical Hydrodynamics. The internal integral vanishes and

reaction in a fluid across an element of surface dS is a
pressure pdS normal to that surface. If the density is p,
the bodily force per unit mass X { ,
and the velocity of the If then ever zero around a circuit it remains so per-
2 is
fluid at x, is u t the
, acceleration of the fluid is found, by manently. This is true if the fluid is initially at rest and is
set in motion by solids moving in it, and in various
other
considering a small parallelepiped, to be given by
cases of importance. But the vanishing of Q for all
circuits
du ( velocity potential <f>
= is the condition for the existence of a
p dl
fr* (!)
such that
If u t
is given in the Eulerian way as a function of the
coordinates x t and the time t, the operator d/dl, giving the
n this case we can rewrite the equations of motion
in
rate of change of any element associated with a given I
particle of the fluid, is equivalent to the form

ldp x
Uk (-),
d d d_ cl
'
dxk
dl~dt
+ Uk
dxk
(2).
and multiplying by dx and adding we havot
If we consider the circulation O. around any closed

| W dx i + uk i dx
lk
l
=-- lP dx< + X
pox,
i
dxt (8),
circuit C in the fluid and moving with it, defined by dtdx( * ox,
du k
~ dx dx =
Z*<j>
2 = u ( dx ( (3),
since
dxk
~ dx (9).
k t t
Jc JCT
3
CONTINUOUS SYSTEMS CONTINUOUS SYSTEMS 69
1
This shows that for all contomporaneous variations and a solution of (4) is
d<j> d
+ toh?='-\
- -
-+ U+ constant (10),
(C),
di j p
where x ( and A at
where X =^
t
dx
(11); <f>
is to be evaluated at f<;
t
dr = d&d&df, ("),
ut 2 is tho sqnare of the resultant velocifcy q. The constant of
integration not necessarily the same at all instants and
is
and f is the distance from , to xt .
therefore may be a function of the time. Hence we have
n
If =
U
dF -n = dFm
(S),
the Bernoulli integral
'
dxk -dx-m ^dx-k
U + F(t) where F a veetor such that
I +**--/? ( is
(12).
Tho change of mass within a given small paral-

ratc of (0),
lelepiped dx l dx2 dx3 is equal and opposite to the rate of

outflow; hence g
p g
.
we have
du t
- = e ikm 3 =
d*Fm
(10).
dT = {pU )dT (13),
di ~dx { '
Thon also
and we have the equation of continuity duk du { _ du k
' m ~ dx't~ dxk
~ eum
dlCi
dp d
S
. .
fisW (14),
- d dF,
e,km ek1l
dp _ du t dx'i 'dxv
or (15).
dt~~ P dx, d"-b\
t'tm e plcs
and Curi. We sometimes dx t dxv
Veciors with given Divergence
have to find a veetor u, such that d*F,
du f
(1),
dx ( VF,
= s. Bt
dv k du (
S = Z&L 'dx t dx.
O),: (2),
dx t dx k
where A is a given scalar and a> lk a given antisymmetrical ~ dx dxt dx i *

m
tensor. We want particular integrals of these equations.
= - V*F m (ii)
dd>
Evidently if (3),
by (9).
where any Thus if F = di (12),

9!. is scalar, \\\t
(4), we shall have
k _du1
aM_8u, = duk du ( _ du { _
dxt dxk
(5),
~^;~ WttJ 2x,-
(13),
CONTINUOUS SYSTEMS COXTINUOUS SYSTEMS 61
60
Thus m/ is the gradient of a scalar <j>' satisfying Laplace's
provided (9) is satisfiod. But
equation. There is no such scalar that makes the velocity
finite everywhcre, including at an infinite distance, except

a linear function of the coordinates. Hence u t
'
is the same
(14),
everywhere.
If there are solid boundaries or free surfaccs at a finite
since <o is a fimction of ,- alone and r a function of distance u- may not be constant. the region where
If
Xf - Applying Grccn's Theorem to all space exccpt a vorticity is present does not extend to a boundary, there is
f,.
small sphere about x we get t ,
no contribution to from points outside this region; we
(14)
therefore take 14) through a region largo enough to contain
1 2^m j
(
~IH\ 2&0B+T
r 4?r
Lim
r 0$m
(15),
the whole of the vorticity. Then the surface integral in (15)
must also be taken over the boundary of this region, but
still and (17) is still a solution. But (17) may not
vanishes,
since .Fm has tho forra of a gravitation potcntial and
satisfy the boundary conditions, and then we must add to
cF ldXi that of a gravitational force. But the first integral
m
when the sphere becomes very small, u, an irrotational solution chosen to make the whole velocity
vanishes in the limit
satisfy them.
and the integrand in the second is zero provided the coin-
In electromagnetism u t
may
be the magnetic intensity
ponents u { exist, for
and w m the electric current across unit surface in a piano
,
ean (10). of x m constant.
dxm ~dxn \dx dxj dx dx m
i
In many cases m, has no curi outside a limited region of
t
If then we are given a scalar A and a vector co, n such that very small eross-seetion. This is often true in the motion of
its divergence is zero, a solution of (1) and (2) is
a real fluid, when the region may be called a vortex fila-
d* dF dFk ment, and in magnetism, when the region is a wire carrying
U,
_
~ + 'dxn
,
(I"),
dx( k m an electric current. The former statement may be expresscd
where <j> and Fn are given by (6) and (12) and ihu are also by saying that the motion is irrotational outside the
in cyclic order. vortex filament the latter says that magnetic forees due to
;
This analysis has two practical applications. In hydro- electric currents have a potcntial. In either case the integral
dynamics u, is the velocity and w m is twice the vorticity,
denoted by 2f m in Chapter ix. Here the divergence of the Q= u dx t
t
(20),
ia
velocityand the vorticity may be given through all space,
and the velocity (17) satisfics the conditions. If the actual taken around a elosed circuit, is zero if the circuit can be
velocity is v( , we may put filled up bya surface not cutting the filament or the wire,
v, = u, + u/ (18), and has a constant value for all circuits that cannot be so
conditions are mutually exclusive, and
and then
du t
'
du k
*
du/
(19).
filled up. The two
must be a elosed circuit.
dx. '
dx< oxk therefore the critical region itself
62 CONTINUOUS SYSTEMS CONTINUOUS SYSTEMS
In each case A vanishes. Then = 0, aud the component
(f>
and
F m is given by
*ikm ds = for i = 1 or 2
Fm = l-
^dr
r
(21) ra
and = 27T for i = 3.
taken through thc critical region. But if we consider the =
Also for unit current 3 the magnetic foree,
, 2tt.
contribution from an element between two planes sopa-
Hence, in this case, by (25),
rated by d$ m and call the element of surface in a plane
,
parallel to these dS, wc have

tr-dUdS (22),
and in general, if the current is /, fl = 477/, and
(fit- 4 l ds
|f Wmf
/S = |,^, = Q (23),
e ikm (26)
taken around the boundary of *- **

where the line integral is f iKtn dL (27).
the filanient. Hence -'J
Fm =
O. [d
(24),
This may be transformed by Stokes's Theorem into an
477 J r integral over a surface with the wire as its boundary thus ;
the integral being taken around the length of the filanient.

t ff; 9 - dS
Also b
O. ( 3 (dj^_J_[îl
rf m 3
u< = [
fak i r dxn J r j
;
Q
in"" /* --/}JL(S,.S-8S) 5
^ ?j
(!)^
iA k)lm
>ls
< 25)
t\** ~r?
where ds is an element of length of the filament and lm a

direetion cosine. --'Jj^&i-
In hydrodynamics Q is the cireulation around the fila-
ment. In electromagnetism the unit current is such that But /m - . . (-) is the magnetic foree at ar< due to
if it flows in a cirele of radius 1 cm. it produces magnetic a doublet of unit strength at , with its axis in the direetion
foree 2tt at the centre. If we take the cirele to be in the plane lm. Hence the foree at any point is equivalcnt to that due
of x3 constant, with its centre at the origin, we have to a distribution of doublets over the elosing surface, with
ds = dd; r = 1; a^ = a^ = a^ = 0; intensity / per unit area and direeted normally to the sur-
face. Such a distribution constitutes a magnetic shell of
& = cos 6, & = Bin d, ia = 0; l1 = -8uxd, J = cos 0, l,=0
strength /.
CONTINUOUS SYSTEMS CONTINUOUS SYSTEMS 65
64
customary to assert as a fundamental postulate that

It is
where N is now the normal magnetic foree due to the whole
is equivalent form may be further estended
of the rest of the field. This
the magnetic foree due to an electric current
this course un- to express the mutual energy of two electric currents.
to that of a magnetic shell, but I think
nature and From this equation we can infer, as is done in the Standard
desirable. The magnetic shell does not exist in
therefore impossible. Further, NdS varies with
direct experimental test is works of Jeans and Livens, that if 1 1 the
though it gives the same does not give the same
foree, it
potential the magnetic

;
potential due to a magnetic shell time, the variation generates an induced e. M. F. -g \\NdS
with a discontinuity at the
is a single-valued f unetion
finite
in the circuit.
shell,while that due to an electric current is a cyclic If we have two circuits carrying currents I and J, their
funetion with no discontinuity escept at the wire itself. mutual influence is expressed by the statement that their
Mutual Energy of Electric Circuit and Magnetic Field. mutual energy is of the form (32), where N is taken to be
A magnetic pole m at x t
is under a foree mu t ;
but the foree near f due to the current J points on the latter
< ;
circuit can be taken to be given by x Then (

.
mUi = ml
teJ)
lm
dZm \r)
dS (29) W = IJM (33),
- the magnetic foree normal to the

where M-W 1 "- dS (34),
and l m S- (-) is
and w, is the magnetic foree at , due to unit current in the

surface at due to the pole. Hence -
J j
l m g-^- Q dS second
a current,
circuit. But if F ( is the veetor potential due to such
is the total flux aeross the

elosing surface, or through the
circuit, of the magnetic foree due to the polo. Also 'd V ia u, = e ikm
gg- (35),
the magnetic potential due to the current,
dV jr- ij^j^gds (36)

(30),
(37),
and therefore the correct forees are given by taking
taken around the circuit. But
where N
-'H
the normal magnetic foree due to a unit mag-
is
NdS (31),
M*
taken around the second circuit; and therefore
"
dxm
(38),
netic pole. This can now be generalized; we say

that in
general the mutual potential of a current and a W

set of M -W- (39),
magnetic poles is
taken around both circuits. This gives the requircd form
W -- I il NdS (32),
for the coefficicnt of mutual induetion of two circuits.
ISOTROPIC TENSORS 67
If i and k are unequal, take i = 1, k = 2. Since c = &&=

we have
c, 2 w a+ CjjMm + cô,, + cû^ = (3),
CH A PT E R VII
and therefore
ISOTROPIC TENSORS uat = u =fi;ua = un (4).
By symmetry u ik is therefore if i ^ k, while ^11 = = u.

A tcnsor is components retain the
called isotropic if its
If i and 4 are both 1, we have
same values however the axes are rotated. We have
alrcady obtained three examplcs, namely S it e lkm and , ,
C1SM21 + c 13 M3, 4- c,j12 + CaUa = (5),
e i"i mpi- wliich is satisfied since every term vanishes. Hence the
There are no isotropic tensors of the first order. For if only isotropic tensor of order 2 is a scalar multiple of S rt .
u was such a tensor, let us give the axes a small rotation

t
If u ikm is an isotropic tensor of the third order,
expressed by the antisyrnmetrical tensor c ik Then in the
.
'a = [B a - e\ (8 kl - cH )
(S mn - c mn ) u Hn (1),
new system
u/ = - c) u, = u,- e ti n,
and therefore, for all values of i, k, m,
(8 (1),
and this can be equal to w, only if c ti u lkm + ckl u iim + cml u lki = (2).
c (l u, = (2) Take = k = 1. Then

for all admissible values of the c i} . Thus ca uilm + ci33im + Cuîam + Cu^lSm
+ cml tt, u + Wttj,, + cÛua = (3).

+ C^Mg+ Cjs3= o|
CaMj (3).
Ca,Mj + cÛt + c^tig = O)

Now put m= 2 so that cm2 = 0. Then
Bnt Cn = ca = Cn = 0, while c^,^,^ are independent and U2 + 122 = Hiu

c j
and opposite to the components obtained by inter-
uji 1
312 + 132 = 11 (4).
changing suffixes. Hence (3) can be satisfied only if 11 - o
Ui = a = s = (4),
From the last equation, and by symmetry, u tkm = if two
and therefore there no isotropic tensor of the first order
is
of i, k, m are equal and the third unequal. Then by the
other than zero. first, u ikm is also zero if all of i, k, m are equal; and the
If u ik is an isotropic tensor of the second order, second shows that
'a = (8
- c u) (
8 ~ u) un ftm = **m
= c8fci tt /i - <ha&u*n If in (3) we put m= 1, evcry term vanishes, so that (3)

= u ik - cuik - c kl u it (1) holds.
to the first order, for all values of i and k. Hcnce Now in (2) if i, k, m are all dificrent, ,,,, is zero unless
Ci,u ik + cki u u = (2). j = t, and then c = 0. Hence (2) holds.

tj It follows that the
68 ISOTROPIC TENSORS ISOTROPIC TENSORS 69
multiples by interchange of suffixes, all components

only isotropic tensors of order 3 are scalar and therefore, of
of ,-*. .
class (b) are zero. Also, from the coefficicnt of c^,
,
lf Uikmv is an isotropic tensor of order 4, we have,
3112 + ISU + U32 = (10).
similarly,
+ C u U m9 + C m ,U (klp + C vl U ikmi = (1). But in (3) the last term vanishes and we infer
CMmp
and
There are only three possible values for i, k, m, p, 2123 + 1223 = (11),
therefore at least two of them must be
equal. We may con-
and therefore
aider separately the cases where (a) two
are equal and the 1312 + 3112 - (12),
three equal, (c) two equal and the whence, by (10),

other two unequal, (6)
1132 = (13).
other two equal, (d) all four equal.
In case (a), take i = k = 1, m= 2, p= 3. Then Thus all components of class (a) are also zero.
12 2123 + ^ "8123 + C,21223 + C 131328 The cocfficient of c lt in (8) gives

+ Caûia + ûw + Cai^m + ^lm = (
2 )-
Uli = 2112 + 1212 + 1122 (14),
C&,
Hence, by the antisymmetrical property of
so that the components of class ('/) are expressible in terms
+ ,EB = - 1113 0)
of the three types of class (c).
(3),
3123 + 1323
- ,, =
11*1 0) No further information is got by transforming com-
= ponents of classes (c) and (d). Thusif = k= l,m = p=2,
1133 - Mlia (
4 )- t'
replacing i or k by j will give a zero component unlcss j is

Other instances of case (a) can be obtained by inter- equal to 1 and then the faetor cu or c kj is zero, and the
;
changing suilixes tliat are not already equal, and by turning relation holds automatically. Similar considerations apply
the axes so as to bring 3 into the position of 1, 1 into that
if all of i, k, m, p are equal.
of 2, and 2 into that of 3. Thus (4) gives
We may denote the components of type (5) by A, those
1133 = 1122 = 2233 = 2211 = 3322 = 3311 ( 5 )" of type (6) by + v, and those of type (7) by y. - v. Then
/j.
And also (14) gives

1111 = 2222 = 3333 = * + 2 /* (
15 )'
1313 = 1212
= 2323 = 2121 = 3232 = 3131 (^)>
There appear therefore to be three independent isotropic
= - = U im = asa = 1 (7).
3113 2112
tensors of order 4, obtained by taking each of A, /x, v in
In case (6), take i = k = m= 1, p= 2. turn equal to 1 and the others to zero.
+ + C121212 + C 131313
C 3112 In the A tensor, u itmv = 1 if i = k and = p, and in all m
Cl22112
= S )' other cases is zero. It is therefore eqnivalent to 8 8mp ,
+ CuU22 + Cl3U32 + C2l"llll +

/fc
C231113 (
wliich is obviously a tensor of order 4, being the produet of
The last term shows that two tensors of order 2.
ni3 = (9), In the /x tensor, u ikmp = 1 if i = m, k = p, or if i = p,
.
70 ISOTROPIC TENSORS
k = m, and i ^ k. If also i = k, the component is 2. Other
coniponents are zero. Tiris can be written
iltnp = S (m 8kv + 8 iv St,, (16), CnA PTER VIII

and is obviously a tensor of order 4. ELASTICITY
In the v tensor, u ikmj> = 1 if i = m, k = p, and = 1 if
i = p, k = m, and in all other cascs is zero. If also i = k, In anelastic solid, as in a fluid, the distanco between any
u ikmp is zero. In tiris case, therefore, two partieles of the body usually varies with the timo.
= S, Sjtp - 8,p
1
")
The body, howcver, has an equilibrium configuration that
ftm 84,,, (
could persist if the external forees were zero or constant.

Tiris can also be written
We may take this as a Standard of reference. If a
M<*mj. = 6tmrt (
18 )> partiele actually at x would be at x, u, in the Standard
t
= = 1. But configuration, we call u the displacement at x in praetice

for if i = = = unless j 2 and then ;
1, ifc 3, e lik t t
=1 m= - = 3, the squares of the u can usually be neglected. Evidently

then e m , p if 1, p= 3, 1 if 7 p=\, and f
= Thus m, is a veetor.
otherwise 0.
At a point x + y ( where y , is suiall, the displacement is
i3i3= li "i33i =- 1 ( 19 )>
( (
< + *><> where

with corrc8ponding values for the other components. du,
Evidently (17) and (18) represent a tensor of order 4. It

*-* (1)
has already appeared in Chapters i and vi.

= (e tk -^ )Vkk (2),
The general isotropic tensor of order 4 is therefore
where e ik and ik are the symmetrical and antisymmetrical
A8 tt 8 ra + /x (8 im S kv + S,,^) + v(8 im 8kp - 8^) (20),
tensors
where A, /x, v are scalars.
*(g+> .-<S-S) < 3 >-
If e lk is zero at x ( the displacement has the same form

,
EXAMPLE as that due to a general displacement u, together with a

Prove that
rotation expressed by the tensor lk Also, if e a is every-
8rt8 mo w rt '."io
.
where zero,
m 8. p + 8,p8*j w rt = wmp +
(S, i rm ,
(8,-A - 8, p 8 m ) w - w mP - >p IB . =
fon * dxm \dx t cxj
2
dx,\dx m oxk J - dxk \cxm ox,/
>e,. 3e
(4).
dx t
dx k

72 ELASTICITY ELASTICITY 73
the rotation the same the e lk vanish except e^ = e^ = e.
Hence the a are constants and
is all Similarly ejj and e
the neces- can be assigned independently.
everywhere. The vanishing of the e ik is therefore
In an elastic solid the internal foree aeross an element of
sary and sufficient condition for a pure
rotation.
due to the If we consider surface is in general inelined to the surface. If the area of
Now consider the part of v f
e ik .
the element is dS, the foree aeross it must be specified by

the auadric surface
= r2 5 ). three components of the form pidS parallel to the axes;
**yiy* (
surface pass where n may be regarded as indicating the normal to the

where r is a constant chosen so as to raake the
cosincs pro- surface. If we consider a small parallelepiped with edges
through yt , the normal at y, has direction
The displacement dxv dxa dx3 centred at (z, zt x3 ), the foree aeross the face
portional to e ik yk and therefore to v,.
, , , , ,
to this of area dz^dz, centred at x, +

due to e ik is therefore parallel to the normal at y, <&, (p n ,p lit p 13 ) dr^dZi,
is
direction of the radius where the p u are evaluated at x, + idz^. The foree on the
quadric. This coincides with the
principal axis. There are therefore opposite face is - (p, pn p l3 ) dx3 dx3 evaluated at
veetor to y, if y, is on a ,
relative displaoement due - c u

three direetions such that the xi <&i. and the total is - ete,da:2 </a:3 In general the
.
and these direetions are

to the e* is in the direction of yt ,
& foree in the x { direction due to the stress aeross the faces
mutually perpendicular. If we take new axes lt &,
along them, the quadric reduces to of xk constant is ^- dr^fda^, and we take account of
*
%'& + '&* + <*'& = r < 6)
all faces by using the summation convention. If the
and all terms e,,' with j * l The displacement in
are zero. acceleration of the element has components
/, and the
the & direction is now en '^, so that aU distances in this density is p, the raass is pdx t dx t dx3 while if the bodily ;
the ratio to 1 + ej,'. The dis- foree aeting has components X

direction are increascd in t per unit mass, the equa-
1
placement due to the then the resultant of three

e lk is tions of motion are
homogeneous strains parallel to three orthogonal axes.
We see therefore that the displacement in any small
combination of
*-fe+* (9).
neighbourhood can be represented as the The system of quantities p ik constitutes a symmetrical

right angles. The latter
a rotation with three extensions at tensor of the second order. To prove this, consider first a
element of the
cxpross the changes of size and shape of an plane whose normal has direction cosines a, interseeting
solid. this roason e is called the strain tensor and f*
For tk
lines through x< parallel to the coordinate axes at short
the rotation tensor. Evidently e ik has six independent distances from x i ; thus a small tetrahedron is formed,
components. For we can have whose sides are of order l, say. Let the area of the sloping
face be dS; then those of the others are a dS. Consider
! = ey lt vt = v3 = i'), u
now the rate of change of momentum of the matter within
making the e lk zero except e = e; similarly
all
e^ and e^ tliis tetrahedron. Evidently the contributions from the ac-
may exist indepcndcntly of the others. Also, if celeration and X, are of the order of the volume, that is,
3
t>! = 0, vt = ez, v3 = ey (3), of l . The foree aeross dS has magnitude p ti dS. That aeross
"
j"
, ,
74 ELAST1CITY ELASTICITY 75
the faco of x k constant is - p tiines the area of the Now consider the energy interchange between the small
is,-paw <*S. But dS is of order 1. Hence we parallelepiped and its surroundings. The stresses across the
face, that
^^ face of area ds^da^ centred at a^ + \dxx are doing work on
( P -a wPtl )0(*) = 0 3 )
('). the element at a rate (pkX u k ) dx2 dx3 and the contribution
,
is indeftnitely small we have fol the strcss from the two opposite faces is
and hence if l
across a plane normal to a (l at x,
t 11 )*
Ph = %j Pu
Now consider three perpendicular directions with direction Tliu8 in all the stresses are doing work at a rate
=
1, 2, 3). The force per
unit area across a
cosincs a (j
plane perpendicular to one of these axes, in the
direction
oxk ^ Kt "
of x { is givcn by p Resolving this along the direction of
.
,
The esternal forees are doing work at a rate pX,v,dT. The
one of the new axes x, we have, therefore, kinetic energy of the element is bpitfdr, and is increasing
Pti = akl aptl = aa*i2><* (
12 )
at a rate putfidr. (We consider the actual specimen of
and k; which is precisely the law of matter occupying the element of volume dr at time t thus
by intcrchanging i ;
transfonnation of a socond order tensor. its mass is pdr, and if we keep to the same piece of matter
at .r,-, with
Consider again a small parallelcpiped centred at time t + dt the mass is we considercd the
unaltered. If
axes, and form the equa- energy within a givcn element of volume we
edges parallel to the coordinate variation of
about its centre. The contributions from should have to allow for the variation of p and the faet that
tion of moments
Xk are of order at most, where the edges have the matter moving out of the element is taking its energy
f k and
l*
lengths of order l. The moment about an axis parallel to x3 with it.) The rate of performance of work on the element
of the stress p, in the face x2 = constant and parallel to x, therefore exceeds the rate of inercase of kinetic energy by
the product of p 2l into the area of the face
and the dis-
is
3
tance of the face from the centre; that is,
to order l ,
jj (T'*.' 1 *) + pX<>h ~ P>'if<\ dr

makes an equal contnbu-
Jparftf.r/zjrf*,. The opposite face
tion. The stress p 12 in the face x = constant and parallel = + P*A ~ *i (^ + P*)} ^,
{ixk <*"*<)
i
direction. We have,
to Zi tends to turn in the opposite
therefore, ^ _ ^ ^^^ _ Q ^ (l8)>
by the equations of motion,
we must
and therefore when we make l indefinitely small
-MSD* (16).
have . >
This work is stored up as internal energy of the element of
.
(l4) '
Pu = Pu
and in general volume. Evidently from its form it is a scalar.
= Pu (ld)
Ptt '
In any clastic solid the internal energy is a definite
symmetrical. funetion of the state of the solid. In any change of stato
so that the tensor p,k is
ELASTICITY ELASTICITY 77
70
only on and 8Q,'0 are perfect diSerentials. Hence if we rcplace the
the increase of internal energy therefore depends
passage six independent e ik by er we can write
the initial and final states and not on the method
of
the 8E=
from one to the other. Now we have seen that six of (p r + q r ) 8e r + P c8d (21),
acquires displace-
e ik are independent, and if the element
ments 8u, in time 8t the corresponding increase of internal
-Y-WJ M (22),
energy is pki 8 (l- A dr. There is an apparent asymmetry and
and k are equal or unequal. Thus p n dr has

according as i
coefficient 8 (8tt,/3ri) = Sen but i = 1, * = 2

,
contributes (P. + 1.) =
de.
(Pr + 1r) i
Z0
(Pr + 7,) = ^ 0*) (23),
Pil 8 (dujdzj,
and = 2, k = 1 contributes pn 8 (dujdtj,
the two together giving 2pa Se lt But this is the same as
.
+ PuShn and the whole contribution from the

p a Se u It follows at once that kept constant Sp 8e r and
process an if is
changes of strain is p ik 8e ik dr. Also during the
r
Lqr 8e, are perfect differentials. Also

amount of heat 8Qdr may be absorbed. If then Edr is
the
internal onergy of the element,
8E = p ik 8e ik + 8Q (17).
and therefore
Since Ea definite funetion of the state of the system,
is
the
and aix of the 8e ft are independent and determine qr - V 'A'
c0
(26).
on changes of the temperature

other three, 8E depends the
=
If 80 we can write
and of the six independent e ik and has a definite value
in
whatever order these changes come about. But p r 8e = r

SIP (27),
where
2F - c f 2c r e r + ce r e, + O (e s )
(28).
\dejJcQ-o \deuJiQ-o
The c ,c T , c may be funetions of 0. Then if we retain
If the absolute temperature and a certain amount of

is 0, only terms in W up to order e1 ,
linear dimen-
heat 8Q is absorbed without chango of any
to 8Q by the rule
pr = cT + ce, (29).
sion, the rise of temperature is related
The the stresses that would remain if the
c r reprosent
8Q = P c80 (19).
strains e r were removed without change of temperature.
where c is the specific heat at constant strain. If

there is In most practical cases the original state is one of uniform
also a change of strain, since 8Q and 86 are scalars, we must temperature and no stress, so that c, = 0. If there is a rise
,mVe of temperature 0' under no stress, an element will acquire
8Q = q lk 8e ik + P c80 (20), displacements
,-a0'y t (30),
order. But hE
where the q ik constitute a tensor of the second
78 ELASTIC1TY ELASTICITY 79
where a lk is a second order tensor expressing the thcrmal simply; then a is the coefficient of linear expansion and
expansion. Thus Ptt - S (30).
e* = I (a + ) ' = fr*0' (31), The second order terms in W 2 constitute a scalar; and
we have
wliere /? lX is. a syrametrical tensor; and
3p = c il.mp (37),
Pik = Ca + c ik mp pmp 0'
_
(32), ce
whcre c rt _, is a fourth order tensor.

.
a tonsor of order 4. If it is isotropic it must be of the
But by hypothosis this deformation takes place undcr no form (20) of Chapter vn. Then the linear terms in p lk givo
stress and thcrcfore p ik = 0. This detcrmines c lk and our ,
Pa = c..mi.emp (33)
formula for the stress ia = ^S +
lfc
8 mp C mp -r /i (S,' m 8fcp 8<p8m) c mp
Ptt = c, mil (e mp - p mB 0') (33). + v{S im 8kp - SjpS^Jc,,

The coefficient c lkmp is the coefficient of e ik e mp in W. = AS rt e ram + /a (B im e mk + 8, p ctB ) + v (S tm e mk 8 ip ekp )
Since there arc six independent e r there are twenty-one = + eki + -
MWb + H
A8 lt e fflm + y. (e ) v (e lk ek{ )
possible terms in a quadratic form ce r e,, and therefore
(39).
there are twenty-one coefficients c, t mp They clcarly form .
a tensor of order 4 such a tensor in general would have

;
the last term vanishing since e ik is symmetrical.
cighty-one componcnts, but this satisfies the symmetry Then
relations that it is unaltcred if we interchange i and k, or 2Wt = p e tk ik (40)
m and p, or i and k together with m and p togethcr. = Aee mm + 2 tie lk e lk
From (26) and (33),
= AA+ 2/ t(e u + e B , + ^ s*+ 2e + 2%, + 2e12 *) (41),
where A = e = dw,/3.c, rt (42)
The scalars A and /* represent properties of the material.

-6e '38 ( c *,mp) + OCfk.mvP (34) Uoth are positive.
We can also write
if d' is small. The second term does not involve the e lk ;
2 \\\ = (A + 2M ) A* + 4M e?n * e31 * + e *
( I- 12
the first is small of the first order in the e ik .
Jlany solids are isotropic; that is, they have the samc - fen ~ eê,, - e,!^) (43).
properties in all directions. This applies to vitreous (glassy)
This appears to differ from the form in Lovo's Elasikily,
solids and to mixturcs of crystals oriented at random. In
1906, p. 100, but the present e ik differ from Love's strain
that caso a uniform rise of temperature in an element
componeuts. My e,, is the same as his, namely 3, ,'S.e, but ;
gives an equal cxpansion in all directions and my e^j is only half his, so that his assemblage of strain
vt = a6 y, (35) componcnts is not a tensor.
80 ELASTICITY ELASTICITY 81
If all werc cqual to one another and thcrcfore to

the e,, If 8Q = 0, so that no heat is lost or gained by conduetion,
A, wc should have pcBB = - qik 8elk = - 2ka68 ik 8eik
= - 3ifca08A (M),
2W = 2 (A + |f) A 2
(44),
and therefore, if the strain takes place adiabatically,
P,*=(A+f/i )A (-*); =0 (i* A)
In general wc write
Pik (45).
0< = -Â pc
(52),
A + f/*
= k (40), Qk*a*0\
7
and ?- (A + ^-) 8A + 2 M e (53).
and call k the bulk-modulus. pc
Thus an adiabatic disturbance the constant A is
in in-
2ir2 =(A + ,*)A 2
ereased above its value for a Standard disturbance to A',
+ 2/x (e + e* 2 + e^ + 2e^ + 2^ + 2e,/) where

(54),
- M (u
2
+ ea 2 + ess
2
+ 2 eiaa3 + Se^n + 2e I1 ew ) pc
The bulk-modulus k therefore also
= (A+3A 2 while /t is unaltered. is
increased to k', where

+ |f* (eu* + a* + ^* - e4a - sau - eua)
*-* + !**
pc
(55).
+ 4 M (e^ + ea, + e,,*)
2 2
In a simple thormal expansion at zero stress the absorp-

-a+t/)A
tion of heat 8Q is equal to pc v 80, where c v is called the
+ i/* {( - + fe - e)* + (n - ea )
ess)
2
specific heat at zero stress, and is the specific heat mca-
+ ee* + 6e3, 8 + 6e } 2
(47). sured in ordinary experiments. Then
12
pcv 86= 8Q = P c80 + q ik 8e ik

The coefficient of /t. vanishcs if and only if the strain is
a symmetrical expansion, and may therefore be called the

= P c80+ 3ka88 8e ik lll
distortional strain energy. = pc8d + 3ka9 3a80. .
If we allow for variations in temperature, 2

9&a 0\ k'
Thus ,-(* + (56).
p ik = A (A - 30') 8 a + 2/x (e lk - a6'8 ik )
pc J' k
The equations of motion at constant temperature, if the

= {AA-(3A+2 At )0'}8 <t +2 / xc tt (48)
properties A and p. are uniform, can be written
and >*--% (49). = A+ + pX(

Pfl (A8 '*
2,te '* ) (57)
Every term in p tk is of the first order in the displacements

but c6'/dB - 1 and tliercfore gives risc to a constant term.
This term is
- |- {(A + A} + pVu, + pX, (58).
q u ~e.8kaSa
/*)
(50).
a
7CT
82 ELASTICITY
If the changes are adiabatic A must be rcplaced by
A'.
any heat conduction, the absorption
If there is of heat C H A PT E R IX
per unit time by the elcment of volume dr is
HYDRODYNAMICS
In comparison with a typical elastic solid, a real fluid shows
where K is the thermal conductivity. Then tho equation a great resemblance and a fundamental difference. The
of heat conduction is force per unit area across an element of surface parallel to
a coordinate plane again constitutes a symmetrical tensor
2, and for the same reasons. The equations of
(59). of order
If we write motion are still (9) of Chapter vui, and also the rate of
e tk = a8 ik 6' + e ik
'
(GO), performance of work on an element of volume dr still has
the form (16). The difference is that the internal energy in
so that e ik is the strain due to the stresscs,
a fluid does not depend directly on how much it has bcen
(61), deformed. However the fluid is moved about and stirred
up, providcd it returns to its initial position, density, and
and the equation becomes temperature, the initial and final internal encrgies aro
equal. The deformation, however great it may be, makes
(62). no contribution the stresscs do work on each element, and
;
thereby supply energy, but this is removed in rostoring the

The elemont dr originally liad volume original temperature. If energy of deformation existed in
d (xt - i ,xt-ut ,x3 -v VL )
dT
a fluid, all particles of it would have a tendency to return
d(x1 ,xt ,x3 ) spontaneously to their original positions when stresses are
removed, and thcy have none. Accordingly, whilo the rate
of performance of work on the element of volume dr in time
\dxl oxs cx3 i )
p ki du jdx k dr, where the symbols have the same
dt is still i
so that its density was p (1 + A). Henco

meanings as in elasticity we can no longer assert f rom this the
,
dA = _\dP (64).
equation (18) of Chapter vui, because the change of internal
di " pdt energy is not determinate when the changes of the e lk are
given. The e ik may be as great as we like, but the energy
This is the equation of continuity.
does not increase indefinitely apart f rom changes in density
and temperature and the fluid moves in the same way under
;
the same external forces whatever its previous history. We

may say that an elastic solid has a memory a fluid has none.
;
HYDRODYNAMICS 85
84 HYDRODYNAMICS
energy chonge in a symmetrical expansion. If the density
The stresses in a fluid are rolated, not to the total de- dimensions in
increases by 8p, there is a contraetion in all
formation, but to the rate of increase of deformation. Ia work
the ratio $pIp an^ * ne stresses do
the former notation these have components
de ' k
or i (
d ' l<
-
+ i*\
ft.*-f* (7)-
At the same time there may an absorption or gencration

The velocities now appear, instead of the displacements
of heat; then the energy change is
frora an initial conflguration, and we now denote tho
velocities, instead of the displacements, by u f . 8E = p P + 8 Q (8),
We also write
and the heat absorbed may be written
e =
(du du t \ - _,/3*_^\ m 8Q = MBP + pcSO
t
l{dx k +
(1) '
dz )' ^"Hac, BxJ
(9),
where M
f
unknown and
is c is the specific heat at constant
80 that e ik denotes the rate of increase of strain and
now
usually called the vorticily. volume. Then
ftt the local angular velocity, ,,-(* + m)*> + pcU (10),
Now we say that p ik is linearly related to the new e ik ; and
therefore
Pik = *** + C ik,mp e mp (")
B
e
=
1>
8p + P 9 (11) '
where P ik is a syinmetrical tensor of order 2 that possibly and the condition that these quantities are perfect differen-
does not vanish with the and c a >mjl is a tensor of order 4.

e ik ,
.
tials gives
an isotropic tensor. 3 (p\
Further, in a fluid at rest the stress is
the forrn pS ik
M = -6,dO\ (12),
Hence P ik is isotropic and must be of , P )
where p is a scalar, and c ik%mp must be of the form (20)

(13).
of Chapter vu. On carrying out the summation with regard
to m and p the term in v disappears and we have It appears also that p must be a function ot p and d
^ alone, for a given material; it does not involve dpfit* We
p ik = - p8 ik + ik e mm + 2p.e lk (3),
where A and p. are scalars, at present undetermined. may call p the pressure.
Evidently p and A are not both necessary to preserve the * This araounts to snying that there ia no diBsipution of energy in a
symmetrical compression or expansion. Tliis is truo in a gas on tho older
form, and we may introduce the further eonvention
kinetio theory; but Enskog has shown (Kungl. Svenska Akad. Eandlingar,
p (l = - 3p (4),
03, no. 4, 1922, p. 18) that p can with greater accuracy bo given by
so that - p is the mean of the three p lk with equal suffixes. P= M '+i-

This gives
- = - where i; is a "seeond " ooofficient of visoosity. But n/jj is only of tho order
3p 3p + 3Ae mra + 2pe (( (5),
of tho square of the ratio of the volume of the moleoules themsclvca to the
and therefore X fp (
6 )- whole volume of tho gas.
function of the In a linuid the coefficient of dp/dt, if it cxists, is within the cxperimentul
If we consider the internal energy as a
error; an analogous statement is true for iiuperfcctly olastio solids.
density p and temperature d alone, we may
consider the
86 HYDRODYNAMICS HYDRODYNAMICS 87
The stress components niay now be written notation for total diffcrentials, since this particular one
PH =~(P + iMCmm) 8 + 2 M C l(t (14), isthe only total differential that occurs in hydrodynamics.
where The acceleration components are given similarly by

p a function of the density and temperature and
is
expresses a property of the fluid. We call p. the coefficienl

/x
of viscosity. For a uniform fluid we have the equations of

''- * s (21) -
motion Consider now the circulation Q about a closed contour.

PU = g- J- {P + Me mm We have
) 8 + M (g + a-)(
=- (P + i^ mm + ) *V, + m g- e
dXi
-^(P-l/*A) + f*V*< (15),
where we now write the scalar The integral

f
u ( du always vanishes;
t
X t dx, vanishes
e mm =A (16). for bodily forces derivcd from a potential. Then
In time dt the outflow from a volume element dxx dx2 dx^

is {3 (pu t )ldx i ) drdt. The mass within the element therefore
decreases at a rate and we have the equation
{3 (pu t )/dx t} dr,
of continuity
(H).
or 'g"*
do du (
.
(18).
Idip+i^-iJcPIJLi&jdz,
C3
d |
.'C P -k
Here . denotes differentiation with regard to the time
dx Scpox k
following a given particlo of the fluid, so that '
are ) c p
given by If /t = theorem of classical

this reduces to the circulation
dx, hydrodynamics. In many
problems of real fluida p. is
= u. (19).
dt small and constant, and A small. Then the integrals
and
<Z _ 3 ,dX(
+
d_ 3
+ 9_ -<(uA), [ -^du k , and - /xdA are zero or products
dxt ~
(20). f f
dt
~~
dt di di *' dx, JcP JcP dx k JcP
of two small auantities, and can be ignored. - dp is zero
most works on hydrodynaraics this operator is denoted
In
JcP
by DjDt, but I see no reason for departing from the usual if p is a function of p, which is truo in many cases, though
|
88 HYDRODYNAMTCS HYDRODYNAMICS 89
exceptions arise whon the tem[>crature or tho composition change is an actual one, the term (27) contributesthep8p/p,
varies from place to place. In the coinmonest case, to while (28) contributes to SQ. So long as tho initial and final
considerable accuracy, states are given, it is immaterial whether &Q represents
heat conducted into the element, absorbed from radiation,
vlk**** (24)> or generated chemically within it, or mechanical energy

dissipated into heat by viscosity.
and circulation arises only from variations of vorticity in
We notice that O can vanish if , and only if,
the neighbourhood of the contour. It follows that in a
fluid originally at rest or in irrotational motion circulation eu e22 C33,
can arise only through the diffusion of vorticity inwards

(29),
from a boundary.
The work done on the element dr per unit time exceeds so that the deformation represents a symmetrical expres-
the rate of increase of kinetic energy by sion or contraction.
If we consider any finite volume, the rate of dissipation
Pki g^" T = Pik^lk<lT (25), within it is
and Piknik = - {p + /*A) S lk e tk + 2fie ik e ik <t>dr= i&+2pe lk e ik )dr (30).

%f
= - (P + SM) A + 2 M e <*e* (26).
du k du {
But tikCik = iktk + 33, dx k
(31),
But (27),
and
and if a function of p (including a constant as a par-
p is
is a dilTerential with regard to the time

ticular oase) this
and yields on the whole zero if the original density is evcr
(32),
recovered. This term may thereforo be considered to give
the increase of internal energy due to compression. The where l k is a direction cosine of the outward normal to the
remainder may be written boundary. But
4> = - + 2pe tk e n
s/tA*
- S/* {(a - a
+ (c - t*)* + (c - e^)*}
<fea)
+ 4^ (e^ + ,, + e,,*) (28). = ^2l k u i $ ik dS+^j d^dS

Thus O is analogous in form to tho distortional strain
2
energy of elasticity ((47) of Chapter vm), It is csscntially 7-
= ^2 eikm v,l k m dS + i dS (33),
positive, andthcrcfore represents work doneon the fluidand ((
continually storcd as hiternal energy. In (8), therefore, if the
whcre d/dn denotes differentiation along the outward
HYDRODYNAMICS 91
90 HYDRODYNAMICS
change of pressure we define a coeflicient of volume
normal, q is the resultant velocity, and m is the vorticity
espansion a by
vector associated with g tlc Also .
ldp
= "a (37).
'p dO
JJR*-JJK*-fJJ$-t)* Then for small changes of temperature and pressure

(34),
and in all *,-,(?-*) (38).
dT = M
JJ
W dS + 4p.
|J
e lkm u ht m dS
t
Also in a free expansion under constant pressure
JJj 8Q = M8P + pchd
-{*(!)
d
M 8*
^
This forra is useful in such a problem as that of waves where the partial differentiation is to be carried out at
on deep water, where the viscosity is small; if it were constant density. But this makes
absent we should have a permanent oseillation in a normal dp
ak (40),
mode. The vorticity is nogligible except within a distance de
from the bottom of order (vjy)^, where y.= vp and 27r/y is k
SQ = p
f )
the period of the motion. Then in the second integral m and therefore \c + a2 - dt 89 (41),
is zero at the free surface and , is zero at the bottom, so

so that c j,, the specific heat at constant pressure, is given by
that this integral vanishes. A is negligible everywhere.
Near the bottom g ik contains a factor proportional to the aH-6
cv = c + (42).
velocity that would exist there on the classical theory, and
thus can be made indefinitoly small for deep water. Hence This form is analogous to that of (56) of Chapter vni, the
the important term in (35) is present a being the coefficient of volume expansion, and
the previous one that of linear expansion.
JJS* In an adiabatic change SQ 0; then
bottom, so that need only 8

and even this vanishes at the it
-akd -?+pc80=Q (43),
be estimated at the free surface.
If a portion of the fluid is compressed without change of and therefore a9Sp = (pc + a 2 kO) 89
temperature, we define the bulk-modulus k by a 2 lc9\ al-0
= ( ~pc)
8p,
ldp 1
= (36),
p dp k a*k6\ 8P
so that (44).
pc ) p
and if it expands under change of temperature without
92 HYDRODYNAMICS I NDEX
Tlius tho bulk-inodulus for adiabatic changes ia
Lagrango's cquations, 28
Aceelcration, 24
Angular momentum, 35
(45). Aren, plauo, as veotor, 20 Magnotism, 53
\ pc I c Axes, ohange of, 1 Moment of foreo, 43
dynamical, 24 of veetor, 35
The equation of heat conduction needs to be modified rotating, 40
Kuil planes and poinls, 10
to allow for heat generated internally. In time 8t, per Axis, coutral, 45
volume dr, we have Centroid, 33 Parallel, 17

Change of axcs, 1 Totential, 49
Circulation, 5G, S7
Couplo, 43 Kigid bodies, 29
= P c80 + akBASt (40), Curi, 14, 58 Rotating axe.s, 40
Rotation, 21, 30
D'Alenibert'e principle, 32
whence dxj
(47). Dielectrics, 52 Seatar produet, 7
fa, \ Direction cosines, 10 Sealam, 3
1,
Dissipation, 88 Specific heat, 81, 91

Divergcnco, 8, 58 Stokes's theorem, 48
Strain tensor, 72
EXAMPLES Elastioity, 71 Stresa tensor, 73
Electromagnetism, 61 Suffixea, dummy, 5
Obtain tho oquations of motion in torma of the stress
1. Eleotrostatics, 50 Summation conveution. 3
componcnta by considering the momentum of a finite volume of Energy, kinetic, 25
any form and applying Groon's theorem. potcntial, 26 Tunsors, 2
Equation of continuity, 58, G2 antisymmetrical, 5, 8
2. Similarly obtain equation (25). 66
isotropic,
Force, 25 Bymmotrical, 5
linc of aetion of, 42 8 ft ,6
A. 12
Gauas's theorem, 49
Gravitation, 40 Veotor produet, 7, 14
Gruen'a theorem, 48 Vootors, 2
with given divergenco and eurl,
liamilton'a principle, 28 58
Heat conduction, 82 Velocity, 24
Hydrodynamics, 83 Virtual work, 34
Hydrostatics anil clossical hydro- Visoosity, 86
dynamics, 56 Vorticity, 60
Incrtia, momentsand produetsof, 37 Work funetion, 20, 50

Iaotropic solids, 78 Work, virtual, 34
Wreneh, 45
Kinetic energy, 25
SCIENCE: MATHEMATICS CAM 191
CARTESIAN TENSORS
SIR HAROLD JEFFREYS
The structural simpliclty of equations in mathematical physics In

Cartesian form ii often hidden by the labour of writing out every
term explicitly. The u$e of tensor notation with the summation
convention overcomes the difficulty; and it is the object of thii work
to illustrate the use of *uch methods.
an attcmpt to provlde a shorthand superior to the vcctor nota-

It Is
tion.The book It well wrltten and printod, and, presupposlng a falr
knowledge of mathematlcs, easy to f4low. Ojfferd Mogmtne
The appllcatlons o* the theory to mcchanlcs, Includlng elastlclty and

hydrodynamlcs, are glvcn, and the book U provlded wlth that neces-
sity for the true student, oxamplcs to be worked by the reader. It
should do mueh to contrtbute to a wlder knowledge of the subiect,
and Is an excellcnt Introductlon to the more advanced treatlscs.
Procasdingt of tht Phytleal Soctety (Votum* 4i)
CAMBRIDGE UNIVERSITY PRESS

Bentley House, 200 Euston Road, London, N.W.1
American Brancht J2 East 57th Street, New York, M.Y. 10022
Standard Book No. 521 09191 8
PRINTEO IN GREAT BRITAIN

Jeffreys-CartesianTensors Text PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jeffreys-CartesianTensors Text PDF

Uploaded by

Copyright:

Available Formats

10$ (50p)

American Branch: 32 East 57th Street, New York, N. Y. 10022

method both introduces new dimeulties and is insufficiently

contravariant veetors disappears, and with it the torms

It is found that the scalar and veetor produets are not

way of writing the ordinary notation, so that the unfolding

after Dynamics. I consider tbis to be the proper order, be-

students working for a university examination.

If we have two sets of rectangular axes (0x, Oy, Oz),

VII. Isotropic Tonsors 66

x = ltx' + ky' + l^z' 'j

y = n^a:' + rn^y' + rn^z' <-

z = nja;' + j y' + n^z'

We can shorten the writing of 1 ) and (2) considerably (

by a change of notation. Instead of (a;, y, z) let us write

respect to the second set are x/, where,;' may be 1, 2, or 3.

with the cosine of the angle betwoon the diroetion of that

x increasing and that of x/ increasing. Let us donote

In mathematical physics we often have to deal with sets u, = a (i Ui (9).

u/ + v/ - 2 a iS (u t + vt ) (8), u/v,' = (a ti Ui) (atl vk )

products of the components referred to the original axes.

In this way the various u/ v,' can be obtained in terms of

u>2 , w,. (14)

so that the gradients transforin according to

rule. Similarly the gradient of a

sx r^dxr ^ ak, {a " Ui)x

This operation therefore replaces the sufnx k by i. The

dircctly; for if we apply (11), 8,,' in the new system of

x/ with which is the scalar product of u ( and v k .

But the a a are the direction cosines of the axis of

are perpendicular. It foUows that the result of the trans-

The tensor ou kjdx gives similarly, on contraction, a scalar

known as the curi or rotation of u ( The vanishing of the . -k an a 03, (29).

number of dimensions. In n dimensions a tensor of order r

of three perpendicular linos, they are connected by six re-

therefore the detorminant cannot change from + to

or from 1 to + 1. If then x goes to x \ x% to x2 x x

1 Now suppose that u t is a veetor, and consider the set of

and therefore quantities

Ou=0, aa = 1, o,, = 0; a12 = 0, an = 0, 0^ = l;l + (o ,ai ,-o2J o ,)3

order, the quantities in braekets are equal to (oln , ain , <%).

1 precedes j by one place the signs are reversed. Hence if

1 j = 1 and l = 2, or j = 2 and l = 'S, or if j = 3 and l = 1,

Any rotation of the axes that does not alter the

axes the determinant of the a tt is + 1.

With this restrietion (

so that the antisymmetrical tensor associated with a vector

we should have S = w"!* = 123 'U + 213 'SI = Wit ~ WH

/In ~ a H a kl a mn e ikm Thus um = e ikm w ik (51).

If w ik w symmetrical, this evidently gives zero. If it is

em = 0. If j, n are all unequal, the expression is

(h, an a3I (47), the tensor w (k of (41) by u ik so that ,

Oln n a 3n U = 8g= 33 = 0; i2= 3, 23 = 1 Sl = 2;j .^

This is a tensor of the fourth order. If we contract it by

makes e 12m different f rom zero is 3, and then e ikm = + 1 Hence .

If i and k = 1, m is 3; but 213

symmotrical product is u vk u k v iy and in tho associatod

vector we give the positive sign to w ik when k follows in i = k or m = p, the component is 0.

- - V^Vg, t^V, - j V,). and therefore

This vector is porpendicular to both the original vectors ;

i (ua v3 - u*,v2 ) + ut (U^ - ttiV8 ) + s (ihvt - ,,) = 0, EXAMPLES

associate a vector so as to leave the sign unaltered when 'okc'tkm

This is often called "curi u." 6. Prove that if u vk wm are vectors,

Yl- &" i* (A - a) - 3 (A - a,)} (46), /0 k -h\

\tY' = [i $,]* = 2zJ''A\<ic, (11),