G. E. Hay - Vector and Tensor Analysis
G. E. Hay - Vector and Tensor Analysis
G. E. Hay - Vector and Tensor Analysis
ANALYSIS
by
G. E. HAY
Associate Professor of Mathematics
University of Michigan
NEW YORK
Copyright© 1953 by Dover Publications, Inc.
All rights reserved under Pan American and In
ternational Copyright Conventions.
Pase
CHAPTER I. ELEMENTARY OPERATIONS
1. Definitions 1
2. Addition of vectors . . . . . . . 2
3. Multiplication of a vector by a scalar 4·
4. Subtraction of vectors . . . . . . 6
5. Linear functions . . . . . . . . 6
6. Rectangular cartesian coordinates 7
7. The scalar product . . . . 10
8. The vector product 11
9. Multiple products of vectors 15
10. Moment of a vector about a point 18
11. Moment of a vector about a directed line 20
12. Differentiation with respect to a scalar variable . 22
13. Integration with respect to a scalar variable 25
14. Linear vector differential equations 26
Problems . . . . . . . . . . . . . . . 28
15. Introduction . . . . . . . . . 34
16. Some theorms of plane geometry 34
v
Page
Differential Geometry
26. Introduction . . . . . . . 51
27. The principal triad . . . . 52
28. The Serret-Frenet formulas 53
29. Curvature and torsion 55
Problems . . . . . . . . 58
Motion of a Particle
30. Kinematics of a particle . . . . . . . . . . . . . . . . 62
31. Newton's laws . . . . . . . . . . . . . . . . . . . . 66
32. Motion of a particle acted upon by a force which is a given
function of the time . . . 68
33. Simple harmonic motion . 69
34. Central orbits . . . . . . 70
vi
Page
nates . . 124
Problems .. . . . . . . . 127
CHAPTER v. INTEGRATION
Vll
Page
viii
VECTOR AND TENSOR ANALYSIS
CHAPTER I
ELEMENTARY OPERATIONS
Figure 1
be interpreted as denoting vector character. This notation for vectors
is somewhat cumbersome. Hence when convenient we shall use a
simpler notation which consists in denoting a vector by a single symbol
in bold-faced type. Thus, the vector in Figure 1 might be denoted by
the symbol a. In this book no mathematical symbols will be printed
in bold-faced type except those denoting vectors.*
The magnitude of a vector is a scalar which is never negative. The
magnitude of a vector PQ will be denoted by either PQ or IPQI- Simi
larly, the magnitude of a vector a will be denoted by either a or !al.
Two vectors are said to be equal if they have the same magnitudes
and the same directions. To denote the equality of two vectors the
usual sign is employed. Hence, if a and b are equal vectors, we write
a=b.
p a Q
Figure 2
2
its origin at Q. Its terminus falls at a point R. The sum a+b is the
vector PR, and we write
a+b =PR.
a+h = h+a.
Proof. Let a and b be the two vectors shown in Figure 2. Then
a Q
Figure 3
(a+b)+c = a+(b+c).
Proof. Let us construct the polygon in Figure 4 having the vectors
a, b, c as consecutive sides. The corners of this polygon are labelled P,
Q, R and S. It then appears that
(a+b)+c ="PR+c
=PS,
a+(h+c) = a+Qs
=PS.
Hence the theorem is true.
According to Theorem 2 the sum of three vectors a, b, and c is
3
independent of the order in which they are added. Hence we can
write a+b+c without ambiguity.
b R
,� s Figure 4
//
Figure 5
4
(3.1) (m+n)a = ma+na,
(3.2) m(a+b) = ma+mb.
Proof of (3.1). If m+n is positive, both sides of (3.1) represent
a vector with magnitude (m+n)a and pointing in the same direction
as a. If m+n is negative, both sides of (3.1) represent a vector with
magnitude lm+nla and pointing in the direction opposite to a.
Figure 7
The two triangles PQ,R and STU are similar. Corresponding sides
are then proportional, the constant of proportionality being m. Thus
(3.4) mPR=SU.
Since PR and SU have the same directions, and since m is positive,
then mPR =SU. Substitution in both sides of this equation from
(3.3) yields (3.2).
Now, let m be negative. Then Figure 7 is replaced by Figure 8.
Equations (3.3) apply in this case also. The triangles PQ,R and STU
are again similar, but the constant of proportionality is Im I, so JmlPR
=SU. Since PR and SU have opposite directions and m is negative,
5
T ma s
Figure 8
a-b = a+(-b),
where the vector -b is as defined in the previous section. Figure 9
shows two vectors a and b, and also their difference. a- b.
Figure 9
6
k-37'
p.
Q
a
Figure JO
(5.1) c =PQ+QR.
But PQ is parallel to a, and QR is parallel to b. Thus there exist
scalars m and n such that
c = ma+nb.
Theorem 2. If a, b and c are any three vectors not all parallel to
a single plane, and if d is any other vector, then d can be expressed
as a linear function of a, b and c.
d =ma+nb+pc.
7
with the positive x2 axis. Otherwise the coordinates are "left-handed".
In Vector Analysis it is highly desirable to use the same orientation
always, for certain basic formulas are changed by a change in orien
tation. In this book we shall follow the usual practise of using right
handed coordinates throughout. Figure 11 contains the axes of such
a set of coordinates.
T--
J
a,
___ p....:-
::___
_L __ ,,______
_ �
/ Q. /R
/ /
----as ___ ___,,,
____ x,
o___._;,.
x,
Figure 11
8
parallelepiped whose edges have lengths la11, Ja21 and ja31. Hence the
magnitude a of the vector a is given by the relation
(6.1) a= Va 12+a22+a32.
From the figure it also appears that
(6.2) a= PQ+Q.R+Rs.
Now the vector PQ is i1• Because of the definitions of tzi
parallel to
and of the product of a scalar by a vector, we then have the ielation
PQ, a1i 1• Similarly QR = aJ2 and RS
= a3ia. Substitution in (6.2)
=
(6. 3)
a+b = a1i1+��+aaia+b1i1+b*+baia·
Now the sum of a number of vectors is independent of the order in
which the vectors are added, by Theorem 1 of§ 2. Hence we may
write the above equation in the form
a+b = aii1+b1i1+a2�+b2�+aaia+baia·
By the theorem in§ 3 we may then write this in the form
a+b= (a1+b1)i1+(�+b2)�+(aa+iha)ia.
Hence the components of a+b are ai+b1, a2+b2 and aa+b3• This
proves the theorem when two vectors are added. The proof is similar
when more than two vectors are added.
9
7. The scalar product. Let us consider two vectors a and b with
magnitudes a and b, respectively. Let oc be the smallest nonnegative
angle between a and b. as shown in Figure 12. Then 0° <ex. < 180°.
Figure 12
The scalar ab cos ex. arises quite frequently, and hence it is convenient
to give it a name. It is called the scalar product of a and b. It is also
denoted by the symbols a· b, and hence we have
cos oc = a1 bi +� b 2+ aa ha.
a b a b a b
a·h =h·a.
Proof. Because of (7.2), we have
a· b a1h1+a2b2+a3b3,
=
h·a = b1a1+b2a2+h3aa.
10
Since a1b1 = b1a1, etc., the truth of the theorem follows immediately.
Theorem 2. The scalar product is distributive· that is
' '
a·(b+c) =a·h+a·c.
a·h=0.
a·h= a·c,
it does not necessarily follow that b =c. For this relation can be
written in the form a· (b - c) = 0, and hence it can be said only that
at least one of the following is true: a= 0; b=c; a is perpendicular
to the vec tor b- c.
We note the following expressions, in wh i ch a is any vector and
i1, i:i and ia are the unit vectors introduced in § 6:
a·a= a 2 ,
i1. i1= 1, i1. iii=0, i1. ia = 0,
(7.3) i2·i1=0, iii· i:i= 1, iii·ia=0,
i3-i1=0, i a· i:i=0, ia·ia=l.
8. The vector product. Let us again consider two vec tors a and b,
.the smallest nonnegative angle between then being denoted by oc, as
shown in Figure 12. Then 0° < < 180°. The vector product of a
oc
11
(i) c is perpendicular to both a and b;
(ii) the direction of c is that indicated by the thumb of the right
hand when the fingers point in the sense of the rotation a from
the direction of a to the direction of b;
(iii) c = ab sin IX,
Figure 13
(8.1) c = axb.
= jaxbl.
Figure U
12
We shall now determine the components of the vector product c
c1 = K(a2b3-a3b2),
(8.3) c2 = K(a3b1 -a1b3),
c3 = K(a1b2-a2b1).
Now c2 = c12+c22+c32• Hence
c2 = K2 [(a2b3-a3b2)2+ (a3b1-a1b3)2+(a1b2-a2b1)2]
= K2(a12( b22+ba2)+a22(ba2+b12)+aa2(b12+b22)
-2(a2b2a3b3+a3b3a1b1+a1b1a2b2)].
The first term inside the square brackets can be written in the form
a12(b 2-b12). If the second and third terms are treated similarly it is
found that
c2 = K2[(a12+a22+a32) b 2-(a1b1+a2b2+aab3)2]
= K2[a 2b 2- (ab cos rx.)2]
= K 2a 2b 2 ( 1 - cos 2 ex)
= K 2a2 b 2 sin2 ex.
But by condition (iii) above, c2 a2b 2 sin2 ex. Thus K
= ± 1. If =
these two values of K are inserted in (8.3) two vectors c result with
the same magnitude but pointing in opposite directions. Only one of
these vectors satisfies condition (ii) above. Now both values of K are
numerical, and are hence independent of a and b. Thus the same
value of K will satisfy condition (ii) for all vectors a and b. Hence it
13
is only necessary to find K for any one special case in which c can be
found directly and with ease from conditions (i) - (iii) above. If we
take a = i1 and b �, it is found from these conditions that c = ia.
=
i1 � is
bxa = b1 b2 b3
a1 � aa
Since this determinant differs from the determinant in (8.6) only
in that two rows are interchanged, the two determinants differ only
in sign. Hence (8. 7) is true. The truth of this theorem can also be
seen easily by examining the three conditions which define the vector
product. According to these conditions the effect of interchanging the
order of a and b is only to reverse the direction of the vector product.
Theorem 3. The vector product is distributive; that is,
axa = 0,
i1Xi1 = 0, i1 Xi2 = i3, i1Xia = -i2,
(8.9) i2Xi1 = -ia, �x� =0, i2Xia = i1,
i3xi1 = �' i3 Xi2 = -i1, iaxia = o.
If a and b are parallel, then
axb = 0.
Also, if it is given that a X b = 0, then at least one of the fo llowing
must be true : a = 0; b = 0; a is parallel to b. Similarly, if
axb =axe,
then aX (b- c) = 0 and at least one of the following must be true:
a = O; b = c; a is paral le l to b-c.
15
Theorem I. The permutation theorem for scalar triple products. If the
vectors in a scalar triple product are subjected to an odd number of
permutations, the value of this product is changed only in sign; and
if the number of permutations is even the value of the product is not
changed.
Proof. A permutation of the vectors in a scalar triple product is
defined as the interchange of any two vectors which appear in the
product. From (9.1) it appears that a single permutation produces an
interchange of two rows in the determinant. Since such an interchange
of rows results in a change of sign only, the truth of the theorem is
established.
Because of this theorem we have
(9.2) V= ja·(bxc)j,
where the vertical lines here denote the absolute value.
Proof. Figure 15 shows the parallelepiped. Let d = b X c. Then
7
d
b
Figure 15
16
where h is the altitude of the parallelepiped. But d is perpendicular
to the base, and if fJ is the angle between a and d, then h =a I cos fJ j.
(The absolute value signs are necessary here, since fJ lies in the range
0° < fJ < 180° and hence cos fJ may be negative.) Thus
V =ladcos.BI
= la·d I
= I a. (h x c) 1.
The expression
ax (hxc)
is a vector, and is called a vector triple product of a, h and c. Let us write
d =b X c, e =a x d.
Then e is equal to the vector triple product a x (b X c). By (8.5)
we have
e1 =a2d3-a3d2
=a2(b1c2-b2c1)-a3(b3c1-b1c3)
=bi( a2c2+aaca)-c1(a2b2+aaba)·
Because of (7.2), this can be written in the form
(9.3) a x ( bx c) = b(a·c)-c(a·b).
This is a rather important identity. It will be used frequently.
We note that the right side of (9.3) is a vector in the plane of b
and c. This is to be expected, since the vector a x (b X c) is perpen
dicular to the vector b x c which is itself perpendicular to the plane
of band c.
Let us now consider the expression
(axb) x (cxd).
17
It is a vector. If we regard it as a vector triple product of a x b,
c and d, then b y (9.3),
(axb)x(cxd) =-(cxd)x(axb).
If we regard the right side of this equation as the vector triple pro
duct of c X d, a and b, then by (9.3),
(axh)·(cxd).
It is a scalar. If we consider it as the scalar triple product of a x b,
c and d, and subject these three vectors to two permutations, then
according to Theorem I of§ 9, we have
(axb)·(cxd) = c-[dx(axb)].
If the vector triple product on the right-hand side of this equation is
expanded by the identity in (9.3), we obtain
(axb)·(cxd) = (c·a)(d·b)-(c·b)(d·a),
or in a form more easily recalled,
ah, a· (b c), a X (b c) .
· ·
18
"
l 0
Figure 16
P = xxa.
Theorem 1. If P is the moment of a about a point 0, then
p =pa,
where p is the perpendicular distance from 0 to the line of action of a.
Proof. Now P = xa sin oc, where oc is the angle between x and a.
But p =x sin oc. Hence P = pa.
x a x a
0
Figure 17
19
tion, and with origins X and X', as shown in Figure 17. Let 0 be any
point, and let OX = x, OX' = x'. Let P and P' be the moments of
a and a' about 0. Then
P' =
(x+xx') xa'
= xxa'+xx'xa'.
Since XX' is parallel to a',XX'xa'= 0. Hence, since a'= a we
have finally
P'= xxa= P.
11. Moment of a vector about a directed line. Each line defines two
directions which are opposite. A line is said to be directed when one of
these directions is labelled the positive direction and the other the nega
tive direction.
Let us consider a directed line L, and let b denote a unit vector
pointing in the positive direction of the line, as shown in Figure 18. We
Figure 18
Figure 19
let P' be the moment of a about O', and let Q; be the corresponding
moment of a about L. Then
Q,=b·(xxa), Q;=b·(x'xa),
where x' is as shown. But x' = O"O+x. Thus
Q; =h·[(O"O+x) xa]
= h· (0"0Xa)+Q,.
Since b and 0"0 have the same line of action L, then b ( 0"0 X :x) = 0
·
21
12. Differentiation with respect to a scalar variable. Let u be a scalar
variable. If there is a value of a vector a corresponding to each value
of the scalar u, a is said to be a function of u. When it is desired to in
dicate such a correspondence, we write a(u).
Let us consider a general value of the scalar u and the corresponding
vector a(u). Let the vector OP in Figure 20 denote this vector. We
Figure 20
From the figure it is s�en that D.a = PQ,. Since D. u is a scalar, the
A. a
vector - has the same d.irect1on
. - The vector
as PQ,.
AU
1. . A.a
im1t-
11....... o A. u
is the rate of change of a with respect to u. It is also called the deriva
da 1. . A.a
- = im1t-•
du 11.u ..... o Au
22
In precisely the same way, we define the derivative with respect to u
d da d2a
du(du)
or
du2 •
d db da
(12.4) (a b) a
du+ du
X X X b.
du =
23
and (3.2). Because of the law exemplified by (3.1) we can then write
( 12.6) in the form
and because of the law exemplified by (3.2), we can then write (12.7)
in the form
A(ma) = ma+mAa+Am a+Am Aa - ma
= m Aa+Am a+Am Aa.
Ifboth sides ofthis equation are divided by Au, and ifAu is then made
to approach zero, ( 12.2) results.
Proof of(l2.3). When u increases by an amount Au, the change in
a·b is
A(a·b) = (a+Aa)-(b+Ab)-a·b.
If both sides ofthis equation are divided by the scalar ilu, we have
- a1 du
24
Now a1, a2 and a3 are scalar functions of u. Also il> i2 and i3 are unit
vectors pointing in the directions of the positive coordinate axis. If they
are the same for all values of u, then
O'
du du du
and so
da da1 da2 + da3
+
_ •
2
• • •
du - di; 11 du 1 du 13
a(u) = a1i1+a2i2+a3i3•
(13.1)
;u.r a(u) du =
25
Proof. We have
J (pu+q)du = �pu2+qu+c,
Jp cos u du = p sin u+c .
.
14. Linear vector differential equations. The equation
dn dn·I
(14.1)
(Po du n + Pi + ·· +Pn-1 dd +Pn) X =a,
dun-1
·
u
in which a and Po, p1, • • ·, Pn are given functions of the scalar u and x
26
c1, c2, c3 , · , en being arbitrary constant vectors and )'1, y2, y3, ... , y" being
· ·
a= a1i1+a2i2+a3i3,
:X: = X1i1 +X2i 2 + X3i3'
F[:x:] = F[xJi1+F[x2]i2+F[x3]i3.
where the c's are arbitrary constants. Let us multiply these three equa
tions by i1 , i 2 and i3, respectively, and then add. The result can be
written in the form
(14.4) x=Y+A,
where
Y = CiY1+CzY2+CaJa+ · · · +c.y.,
A= A1i1+AJ2+A3i3,
the vectors Ci, c2, c3, , c. being arbitrary constant vectors. Equa
· · ·
tion (14.4) gives the general solution of Equation (14.1). We note that
Y is the general solution of the homogeneous equation F[ x] = 0, and
that A is a particular solution of Equation ( 14.1). The particular solu
tion A can be found by procedures very similar to those used to find
particular solutions of linear scalar differential equations. This is de
monstrated below.
27
As an example, let us find the general solution of the differential
equation
(14.5)
where p and q are constant vectors. We must first find two linearly
independent solutions of the equation
d2y - dy � .
=0
-
(14.6)
du2 du
The auxiliary equation of this differential equation is
m2-m-2 =0.
It has roots -1, 2, whence the required solutions of (14.6) are e·• and
e2". Thus
Y = c1e·•+c2e2".
Problems
28
magnitudes are 1, 2, 3 and 2, and their directions are east, northeast,
north and northwest, respectively. Construct these vectors.
2. If a, b and c are defined as in Problem 1, construct the vectors
(a+b) +c, (b+a) +c, c+(a+b), and by measuring their magnitudes
and directions verify that they are equal.
3. If a and b are defined as in Problem 1, contruct the vectors a+2b,
2a+b, 3a-b, -2a-2b.
4. If a, b and c are defined as in Problem 1, express each of these
vectors as a linear function of the other two, determining the coeffi
cients graphically to two decimal places in each case.
5. Given that
a+2b = m, 2a-b = n,
29
13. Given that
a=i1+2i2+i3,
b 2i1+i2,
=
common origin. Show that the line joining their terminuses is parallel
to the x1x2 plane, and find its length.
17. Show that the vectors a=i1+4i2+3i3, b=4i1+2i2-4i3 are
perpendicular.
18. If a, b and c are as defined in Problem 13, find a· (bxc),
(bxa)·.c , ax(bxc), (axb)·(axc), (axb)x(axc).
19. If the vectors drawn from the origin to three points A, Band C
are respectively equal tothe three vectors a, b and c defined in Prob
lem 13, find a unit vector n perpendicular to the plane ABC. Hence
find the distance from the origin to this plane.
ax(bxc)+hx(cxa)+cx(axb) =0.
21. Show th at
30
25. Show that
du du2
33. If a is a function of a variable u, show that
31
34. Given that the unit vectors il> i and i 3 are independent of a
2
variable u, evaluate J a du when
i1+2ui +8u3i3,
2
(i) a=
2i1 + 2i2
(iii)
•
4+ u 2
. a =
4-u2
36. Find the vector x(u) in each of the following cases, given that
p, q and r are constant vectors:
dx
p u2
+qe '
2
du
(l') - =
•
d2x .
p cos u+q sm u,
du2
(ii) =
d2x .
(iii) (p sm u-q cos u) xr.
du2
=
d2x dx
4 4x
du2+ du +
(ii) = 0'
d2x dx
5x
du+
(iii) - 2 0,
du2
=
d"x 2
6 d3x + l l d x 6 dx O.
du3
(iv) =
du' du2 du
_ _
32
dx. -3x p(3u2+1),
( i)
=
du
d2x 4
(ii) - x 16p cos 2u,
du2
=
5
.
d2x + dx
(iii) 2 u
6 pe + q sm u.
du2 du = -
du
(1.1') d2x dx + 18
3d 2p e u
qu2 '
du2- u=
d2x dx
(iii) u2 du2 - u du 3x
-
6p. =
33
CHAPTER II
APPLICATIONS TO GEOMETRY
34
c
Figure 21
(16.2) d = OA+AD.
Since D lies on the line AC there exists a scalar v such that
AD =vie.
Hence we can write (16.2) in the form
(1 6.3 ) d = a+vAC.
We now equate the above two expressions given ford in (16.1) and
( 16. 3 ) , obtaining
(16.4) a+vAC = u b.
The next step is to express all vectors in this equation as linear functions
of any two vectors in the plane, say a and c. From the figure we see
that
AC =-a+c, b = a+c,
whence (16.4) becomes
(1-u-v) a= (u-v) c.
35
Since a and c do not have the same line of action it then follows that
1-u-v=0, u-v=0.
point of OB.
We have now proved that the point of intersection D of the diago·
nals is the middle point of one of these diagonals. From symmetry, D
must also be the middle point of the other diagonal.
Theorem 2. The medians of a triangle meet in a single point which
trisects each of them.
Proof. Let us consider the triangle OAB in Figure 22. The points
Figure 22
C, D and E are the middle points of the sides, and F is the point of
intersection of the medians AD and BE. We must prove that F is a
point of trisection of each of the three medians AD, BE and OC.
For convenience we denote the vectors drawn from 0 to the points
A, B, C, D,-E and F by a, b, c, d, e and f, respectively. Now Flies on
(16.5) DF = uDA
36
Similarly, since Flies on the median BE; we hav e
f = OE+vEB,
where v is some scalar. We now equate these two expressions for f,
obtaining
(16.6)
The next step is to express all vectors in this equation as linear func
tions of any two vectors in the plane, say a and h> F ro m Figure 22 we
see that
t-u - tv = 0, !-!u-v = 0 .
37
x,
Figure 23
x = x1i1+x2i2+x3i3•
The quantities x1, x2 and x3 are also the coordinates of the point X.
We shall use the letters A, B, C, · · · to denote specific points in space,
and shall denote the position-vectors of these points by a, b, c, · · · . The
component of these vectors will be denoted in the usual way by the
symbols (a1, a2, a3), (b1, b2, b3), (c1, c2, c3 ), · • ·• We note that these
quantities are also the coordinates of the points A, B, C, · · · .
38
B
Figure 24
AC CB
( 18.1) -=-·
m n
If we now denote the position-vectors of A, B and C by a, b and c,
respectively, then
AC = c - a , CB = b-c,
and ( 18 .1) can then be written in the form
n(c-a) =m(b-c).
Solving this equation for c, we obtain
=;= mb+na.
(l8.2) c
m+n
This formula expresses the position-vector c of the desired point C in
terms of the known quantities a, b, m and n.
In books on Analytic Geometry, formulas are usually given which
express the coordinates of C in terms of m, n and the coordinates of
A and B. It should be noted that (18.2) is entirely equivalent to these
formulas, for these formulas can be deduced from (18.2) simply by
equating the components of the left side of (18.2) to the components of
the right side of (18.2).
19. The distance between two points. Let us suppose that A and Bare
two given points, and that it is desired to find the distance d between
A and B in terms of the position-vectors a and b of A and B. Figure
25 illustrates the problem. Now
39
But AB = b -a . Thus
Figure 25
20. The area of a triangle. Let us suppose that A, B and C are three
given points, and that it is desired to find the area D.00c of the triangle
ABC in terms of the position-vectors a, b and c of A, B and C. Figure
26 illustraties the problem.
C D
A B
Figure 26
(20.1)
40
where
(20.2) cp =ABXAC.
Now AB = b - a, AC = c - a. Thus
cp =
(b - a) x ( c - a) .
This simplifies to
(20.3) cp = hxc+cxa+axb.
The required area of the triangle is thus given by (20.1), qi being the
magnitude of the vector given by (20.3).
A property of the vector cp will now be recorded, for future use.
Since this vector is equal to ABX AC we conclude that the vector cp is
perpendicular to the plane of the triangle ABC, and its direction is that indicated
by the thumb of the right hand when thefingers are placed to indicate the direction
of the passage around the trianglefrom A to B to C.
21. The equation of a plane. There are several ways in which a plane
can be specified. For example, three points which are on the plane and
do not lie on a single straight line can be given, or a line in the plane
and a point on the plane but not on the line can be given. In each of
several such cases we sh�ll now deduce the equation which must be
satisfied by the position-vector x of every point X on the plane. This
equation will be referred to simply as the equation of the plane. In books
on analytic geometry the equation of a plane usually appears as an
equation which involves scalars only, and is satisfied only by the co
ordinates of points on the plane. We shall refer to this latter equation
as the cartesianform of the equation of the plane.
(i) To find the equation of the plane through a given point and perpendicular
to a given vector. Let A be the given point and b be the given vector.
Figure 27 illustrates the problem, the plane P being the plane in
question.
Let X be a general point on P, and let a and x denote the position
vectors of A and X, respectively. Now AX is perpendicular to b. Thus
AX·b =0.
41
But AX = x - a, whence it follows that
(21.1) ( x-a) ·b = 0 .
This is the desired equation of the plane P.
..,
Figure 27
42
cp = bxc+cxa+axb
is perpendicular to the plane P. Hence, by Problem (i) above the
equation of P is
(21.2) (x-a)·cp =0.
x,
Figure 28
(22.1) p = Kb .
OD+DE+EA+AO = 0,
or
(22.2) d+Kb+EA-a = 0.
Thus
(a-d)·b
K = ,
b2
44
and substitution for Kin Equation ( 22.1) then yields
(a-d)·b
(22.4) p= b2 b.
a·b
(22.5)
p=b2b.
(ii) To find the vector-perpendicular from a point D to a plane P through three
given points. Let A, B and C be the three given points, with position
vectors a, band c, respectively. Figure 30 illustrates the problem. It
p in terms of a, b, c and d.
is desired to find the vector-perpendicular
A c
Figure 30
q>
= bxc+cxa+axb
is perpendicular to the plane P. Hence we may regard P as the plane
through the given point A and perpendicular to the given vector q>.
Thus, from Equation (22.4) it follows that the required vector-perpen
dicular is given by the relation
45
(a-d)· q>
p= -- 2-'4>·
cp
a· (b X c ) - d· q>
(22.6) p =
cp2 q>.
23. The equation of a line. There are several ways in which a line in
space can be specified. For example, two points on the line can: be
given, or two planes through the line can be given. In each of several
such cases we shall now deduce the equation which must be satisfied
by the position-vector x of every point X on the line. This equation will
be referred to simply as the equation of the line.
(i) To .find the equation of the line through a given point and parallel .to a
given vector. Let A be the given point, with position-vector a, and let b
be the given vector. Figure 31 illustrates the problem, L being the line
in question.
Figure 31
AX X b = 0.
46
But AX = x-a, whence it follows that
(23.1) (x-a)xb = 0 .
(ii) To.find the equation of the line through two given points. LetA and B be
the given points, with position-vectors a and b, respectively. Figure 32
Figure 32
illustrates the problem, L being the line in quest:J.on. Now Lis parallel
to the vector AB, and AB = b - a. Thus, by Problem (i) above, the
desired equation of L is
(iii) To find the equation of the line through a given point and perpendicular
to two given vectors. Let A be the given point with position-vector a, and
let b and c be the given vectors . Figure 33 illustrates the problem, L
being the line in question. Now Lis parallel to the vector bX c. Hence,
by Problem \i) above, the desired equation of Lis
(iv)To find the equation of the line through a given point and perpendicular
to the plane through three given points. Let A be the given point on the line,
and let B, C and D he the given points on the plane. Figure 34 illustrates
47
L
Figure 33
the problem, L and P being the line and plane in question. We denote
the position-vectors of A, B, C and Din the usual manner. Let us con-
sider the vector <p given by the relation
<p = cxd+dxb+bxc.
L
/1
Figure 34
48
(23.4) (x-a)Xcp=O.
24. The equation of a sphere. Let S be a sphere of radius a with
center at a point C, as shown in Figure 35. If X is general point on
the sphere S, then
CX·CX = iCXi2 = a2•
But CX= x - c. Thus
x
�-----C
49
Thus
DX·EX= (x-d)·(x-e) =X·X-X·(d+e)+d·e.
But d+e=O, and if a denotes the radius of the sphere then x·x = aa,
DX·EX =
0,
and so DX is perpendicular to EX.
x
J)
Figure 36
Figure 37
50
the equation of the plane P which touches S at a given point X.
Let Y be a general point on P. We denote the position-vectqrs of the
various points in the usual manner. From the figure it follows that
CY=CX+XY,
or
y-c=cx+xY.
( x - c) (y - c)
· = a2•
Differential Geometry
26. Introduction. We shall consider here only a small portion of the
differential geometry of curves in space. Rectangular cartesian co-or
dinates x1, x2 and x3 are introduced, with originat a pointO. The quan
tities x1, x2 and x3 denote the coordinates of a general point X with
position-vector x. If i1, i2 and i3 are unit vectors in the directions of
the positive coordinate axes, then as before,
(26.1) x = x1i1+xJ2+x3i3•
A curve consists of the set of points the position-vectors of which satisfy
the relation
x = x(u),
where is a function of a scalar parameter u. We shall consider only
x(u)
those parts of the curve which are free of singularities of all kinds.
If the set of points comprising a curve all lie in a single plane, the
curve is said to be a plane curve. If this set of points does not lie in a
single plane, the curve is said to be a skew curve.
It is convenient to choose as the scalar parameter the lengths of the
arc of the curve measured from some fixed point A. The quantity s is
51
positive for points on one side of A, and negative for points on the
other side of A. The equation of the curve may then take the form
x = x(s).
The derivatives with respect to s of the function x s
( ) will be denoted by
' " '"
x , x , x , etc.
0
Figure 38
way by the symbols j1', j2' and j3'. They are shown in Figure 38, and
are defined by the conditions:
increasing;
(ii) j2 lies. in the plane of the vectors j1 and j1', and makes an acute
angle withj1';
1
(iii) j3 is such that the vectors j1, j2 and j3 form a right-handed triad •
1. At points on the curve where j1' is equal to zero, these conditions are not suf.
ficient or a unique determination of j2 and j3• We exclude such points from con
/
sideration here.
52
The straight line through the point X and parallel to j2 is called the
pr incipal normal to the curve . The straight line through X and parallel
to j3 is called the binormal to the curve. The vectors j1, j2 and j3 are
called the unit t angent vector, unit normal vector, and unit binormal vector,
respectively. The triad formed by these vectors is called the principal
triad. The plane through X and perpendicular to j1 is called the normal
p lane. The plane through X and perpendicular to j3 is called the
o sculating plane.
b.x
x'= limit
l!.s ..... 0 b.s
Now
limit l
b.xl
= l.
1!.s _,. 0 b.s
Thus x' is a unit vector. Further, the vector fl.x/b.s lies along XY, and
its direction then tends to that of j1 as fl.s tends to zero. Since j1 is a unit
vector, we can then write
(28.1)
The vectors j 1 ', j2' and j3' can each be expressed as� linear func
tion of any three non-coplaner vectors. In particular, they can be ex
pressed as linear functions of the vectors j1, j2 and j3, and we then
have relations of the form
j1 '=a 1 d1 +a12j2+a1aja,
'
(28.2) j2 =a2d1 +a22j2+a23j3,
'
ja =aad1 +aa2j2+aaaja,
where the scalar coefficients are functions of the parameter s. Since
the vectors j1, j2 and j3 are orthogonal unit vectors, they satisfy the
relations
53
j1-j1= 1, jl -j2= 0, jl -j3= 0'
(28.3) j2-jl= 0, jd2= 1, j2-ja= 0'
j3 -jl = 0, ja-j2= 0, ja-ja= 1.
We differentiate with respect to s the first equation in the first line
of (28.3). This yields the relation
jl ·j1' +j1' -jl= 0.
Since in a scalar product the order in which the vectors appear is im
material, we can interchange the vectors in the second scalar product.
It then follows that
jd1'= 0.
Ifwe substitute here for j1' from the first equation in (28.2), and then
make use ofEquations (28.3),we find thata11=0. Similarly a22=a33=0,
and we may write
(28.4)
We now differentiate with respect to s the second relation in the first
line of (28.3). This yields
jl ·j2' +j1' -j2= 0.
If we substitute here for j1' and j2' from the first two equations in
(28.2), and then make use of Equations (28.3), we find that a12+a21=0.
Similarly we can find two similar relations, and we have altogether
(28.5) · a12+a21=0, a23+a32=0 , a31+a13=0.
So far, only conditions (i) and (iii) above have been used. By con
dition (ii) the vectorj1' is to be in the plane of j1 and j2. This can be
true only if
(28.6) a13= 0.
By condition (ii), the vector j1' is to make an acute angle with j2. If
this angle is denoted by oc, then cos oc must be positive. But
/j1'/ cos oc=jd1'.
Ifwe substitute here for j1' from Equations (28.2) and then use Equa
tions (28.3) we find that
54
Thus
(28. 7)
We now define two quantities x an d -r by the relati ons
(28.8)
Then, by (28. 7) it follows that
(28.9) x > 0,
and because of Equations (28.4), (28.5), (28.6) and (28.8), we can now
express Equations (28.2) in the form
ji' = xj2,
(28.10) j2' = -i:ia xj1,
-
ja' = - -i:i2·
/::,.s-+OLis
where Li6 is the angle between the tangents to the curve Cat the points
X and Yin Figure 38. Thus x is the rate at which the tangent at the
point X ro tates as X moves alon g the curve. The reciproc al of x is
called the radius of curvature, and will be denoted by the symbol p.
1.
't" =
lffilt-,
Li<I>
.
a,-+ o Lis
where Li<I> is the angle between the binormals to the curve C at the
55
points X and Y in Figure 38. Thus -r is the rate at which the unit bi
normal at the point X rotates as X moves along the curve. The re
ciprocal of -r is called the radius of torsion, and will be denoted by er.
(29.1)
Substitution from the second of these relations for j1' in the first of the
Serret-Frenet formulas then yields the equation
" =
(29.2) x xj2.
We now multiply each side of this equation scalarly by itself, obtaining
x2 = x " x ".
·
We now substitute for j2' from the second Serret-Frenet formula, ob
taining
=
jl [X2'Tj1 +x3ja]
·
= x2-r.
Thus
I
(29.5) -r
= 2 x'· ( x"xx"' ) .
x
56
Substitution from these relations in Equations (29.3) and (29.4) then
yields
(29.6) V;-" 2
x =
l +x2----;-;;+x3 ,,2 '
I
X1 x2' X3 I
(29.7) ff ff
x3 If
x
't'=-
x2
xl 2
fff x Ill Ill
X1 2 X3 I
Since x can now be found, we can obtain the unit tangent vector j1
and the unit normal vector j2 by use of Equations (29.1) and (29.2).
The unit binormal vector j3 can then be found easily, since it is equal
to j1 xj2 We have the collected results
•
1 1
(29.8)
•
J2 = x x"
•
- Ja = - x' X x" .
x
Let us now find the equation of the tangent to the curve at the point
X. If Y is a general point on this tangent, the desired equation is easily
seen to be·
(y-x) X j1 = 0.
(y-x) · j1 = 0.
(29.12) (y-x)·x'=O.
In the same way we can find the equation of the osculating plane in the
form
(29.13) (y-x)·(x'xxff) =0.
57
Problems
1. Prove that the line joining the middle points of any two sides of a
triangle is parallel to the third side, and is equal in length to one half
the length of the third side.
. 2. Prove that the lines joining the middle points of the sides of a
oA+oB+oc = on+oE+oF.
oA+oB+oc+on = 40E.
point. Hint: the sum of unit vectors along two sides lies along the bi
sector of the contained angle.
10. Prove that the perpendicular bisectors of the sides of a triangle
meet in a point.
11. If 0 is a point in space and ABC is a triangle with sides oflengths
l, m and n, then
l OA+m OB+n OC = ( l+m + n) OD,
where D is the center of the inscribed circle.
12. If ABC is a given triangle, the middle points of the sides BC, CA
58
and AB are denoted by D, E and F respectively, G is the point of inter
section of the perpendiculars from the vertices to the opposite sides,
and His the center of the circumscribed circle, prove that
GA+GB+GC= 2 GH.
and is parallel to each of two given vectors b and c. Derive the equa
tion of this plane in the form
( x - a) (b x c) =
· 0.
59
with position-vector c, its line of action along the perpendicular from
C to L, and its terminus on L. Show that
(a-c)·b
p =a-c- -- -2--b.
b
where c= bxb'.
26. Prove that if the torsion of a curve is equal to zero, the curve is
a plane curve.
- 27. Prove that
-"- ' +
+( x II -x3 - XT" +
x1111 = -
3 I•
l<. l<. J1 )Jz (2 x
I
" XT
' '
)Ja .
28. If the position-vector x of a general point on a curve is given as
a function of a parameter t, and if primes denote differentiations with
respect to t prove that
1 1
x = -V x"'· x" -s"2 ' -r = -
z x'·(x"Xx""') '
s'z x s,a
'
• x . x' X x" ·
J1 = Ja=�
s' '
x = a cos t i1+asn
i t i 2+a t cot oc i3,
where a an d oc constants, and t is a parameter. Find p, a and the
are
pr i ncipa l triad. Answer: p = a co sec2 oc, a 2a cosec 2oc, j1 = sin oc
=
(-i1 sin t+i2 cos t+i3 cot oc), j2 = -i1 cost- i2sn
i t, j3 =cos oc (i1 s i n t
- i2 cos t+i3 tan oc ) .
30. The position-vector x of a general point on a curve is given by
the relation
x = a(3t-t3)i1 +3at2i2+a(3t+t3)i3,
where a is a constant and t is a parameter. Find p, a and the princi pal
triad. Answer: p =a= 3a yS,v2j1 = y·1(oc i1+�i2+yi3), j2 = y·1(-f'i1
1 2
+oc i2), V2 j3 = y- (-oc i1 -� i2+y i3), where oc = 1 - t , � = 2t,
y = 1 +t •
2
x = a[(t-sint)i1+(1-cost)i2+ti3],
where a is a constant and 't is a parameter. Find p and cr. Answer:
2
p=a oca/2, f'-i1 ; cr =-a�' where oc = 3 - 2 cos t, � = 2 - 2 cos t
+ cos2 t.
61
CHAPTER III
Motion of a particle
30. Kinematics of a particle. The phrase ,,kinematics of a particle"
refers to that portion of the study of the motion of a particle which is
not concerned with the forces producing the motion, but is concerned
rather with the mathematical concepts useful in describing the motion.
Let us consider a moving particle. It is necessary to introduce a
,,frame of reference" relative to which the motion of the particle can
be measured. For a frame of reference we take a rigid body. Such a
body is one having the property that the distances between all pairs
of particles in it do not vary with the time. We then introduce a set of
rectangular cartesian coordinate axes fixed in the frame of reference.
Figure 39 shows these axes and the associated unit vectors i1, i2 and i3•
x
••
Figure 39
62
Let the curve C in this figure be the path of the particle, and let the
point X denote the position of the particle at time t. The vector OX
is the position-vector of the particle. We denote this vector also by x.
(i) The components of the velocity and acceleration in the directions of rec
tangular cartesian coordinate axes. Let x1, x2, x3 denote the rectangular
cartesian coordinates of the point X in Figure 39. Then
x = x1i 1+x2i2+x3i3•
If we now adopt the convention that a single superimposed dot de
notes a first time derivative, and a pair of superimposed dots denotes
a second time derivative, then
dx . . • . • •
V = di = X11 1+x212+x313 ,
dv .. ..
• . • . •
a = di = x111 +x212+x313 •
(ii) The comp onents of the velocity and acceleration in the directions of the
principal triad of the curve traced out by the particle. The curve C in Figure
39 is the path of the particle. Let j1, j2 and j3 denote the principal
triad at the general point X on C. The principal triad was discussed
in§ 27. Ifs denotes the arc length of C, then from Equations (28.1)
and (28.10) we have
( 30 . 3 )
63
where x is the curvature of C. Now
dx dx.
v=-=- s
dt ds '
and because of the first equation in (30.�) we then have
(30.4) v = s jl.
Thus the velociry of the particle is directed along the tangent to its path, and the
speed is v= s .
Because of (30.4) we have
dv .. •. dj1
a= di= SJ1+s di"
But because of the second equation in (30.3) we have
and hence
(30.5) .ij1 +xs2j •
a
2
=
Thus the acceleration a lies in the osculating plane of C. Also, the components
of a in the directions ofthe tangent, normal and binormal are
.. . dv v2
(30. 6 ) s=v=v ' xs2
. = xv2= - ' 0'
ds p
64
x
Figure 40
If i1, i2 and i3 are the usual unit vectors associated with the rec-
tangular cartesian coordinate axes in Figure 40, then
65
(30.8)
dv
dt = rk1+(r6+r6)k2+x3k3
.. . · ·· ..
a=
. 1 d .
(30.12) x, x6; x x62 - (x26) .
'x dt
- -
66
pended from a standard spring at a standard place in the earth's
gravitational field, produces a standard deflection of the spring.
Hence we can assign a numerical value to the mass of any body.
We now introduce the laws governing the motion of a particle.
These laws, which were first stated by Isaac Newton and are called
Newton's laws, are as follows:
(i) Every particle continues in a state of rest or uniform motion in a
straight line unless compelled by some external force to change that
·
state.
(ii) The product of the mass and acceleration of a particle is pro
portional to the force applied to the particle, and the acceleration is in
the same direction as the force.
(iii) When two particles exert forces on each other, the forces have
the same magnitudes and act in opposite directions along the line
joining the two particles.
In the second law, the acceleration of the particle enters. This accele
ration depends on the frame of reference employed. It thus appears
that Newton's second law cannot apply in all frames of reference.
Those frames of reference in which this law. does apply are called
.Newtonian frames of reference. A frame of reference fixed with respect to
the stars is Newtonian, and in making an accurate study of any motion
such a frame of reference should be used. However, for many problems
we may consider the earth as a Newtonian frame of reference, when
effects due to the motion of the earth are negligible.
Let us now consider a particle of mass m acted upon by a force F.
Let a denote the acceleration of the particle relative to a Newtonian
frame of reference. Then according to Newton's second law
F =kma,
(31.1) F =ma.
There are three such systems of units in general use. These are indi
cated in Table 1, together with abbreviations commonly used for these
67
units. Thus, for example, when a force of one pdl. acts on a particle
with a mass of one lb., the acceleration of the particle is one ft./sec.2•
The systems of units in the second and third columns of Table l are
called foot-pound-second sytems, or simply f.p.s. systems. The sy
stem of units in the fourth column is called the centimeter-gram
second system, or simply the c.g.s. system.
f.p.s. c.g.s.
The lb. wt. is the force exerted on a· mass of one lb. by the earth's
gravitational field. If G denotes the acceleration due to gravity, ex-
·
1 slug =G lb.
dv
(31.2) F=m
dt '
F = 12 p+q cost,
68
where p and q are given constant vectors. Because of the first equa
tion in (31.2) we then have
dv
m dt = 12p+q cost,
whence
mv = j(l2p+q cos t) dt.
We now carry out this inte gration in the manner outlined in § 13,
obtaining
r = mv0, s = mx0+q.
d2x
-kx = m '
dt2
or
d2x k
-+-x =O.
dt2 m
69
This is a differential equation of the type cons i dered in§ 14. Ac co rding
to the procedure demonstrated there, the general solution of this dif
ferential equation is
cos
i/I
;;, t + c2
. i/I
sm ;;, t ,
x = c1
V V
where c
1 and c2 are arbitrary constant vectors. These arbitrary con
stant vectors can be found if the initial values of x and v are known.
c2 = V�vo,
whence we have
x = x0 cos
l /k
t+v0 i /;;, sm
. i /k t
V ;;, Vk V;.
It will be noted that x is a linear function of Xo and v0; hence it follows
that the motion of the particle is confined to the plane P containing
the given vectors Xo and v0 • This result could h ave been anticipated ,
for the force F acting on the particle has no component perpendic
ular to the plane P.
F =F(x).
The equation of motion is
F =ma.
70
nates are -F, 0. Also , the components of a in these directions are given
in Equation (30.12). Hence we have
(34.1) -F=m(x-x62),
md ·
(34.3) y = l/x.
Then from (34.2) we have
ey-2 = const. = h
whence
(34.4) e = hy2•
Then
·
= --. y-2 dye· hdy
·
x - y-:y
2
=
de =
- de '
d2y . d2y
(34.5) x = -h e = -h2.J2
de2 de2•
By substitution in (34.1) for x, e and x from (34.3), (34.4) and (34.5),
we finally obtain
(34.6)
Now Fis a function ofy alone. Once the form of this function has been
assigned , we can find the path of the particle by solvin g Equation
(34.6).
Let us now consider the special case when F varies inversely as the
square of x. Then we can write
F = ymy2,
where y is a constant, and Equation (34.6) becomes
71
d2y y
=
de2+Y h2·
The general solution of this equation, expressed in terms of x, is
1 Y
(34.7) + cos e +c·2 sm e,
x = h2 c1 ·
(35.1) m = :E mj.
j= 1
We denote the coordinates of the particle of mass mi by the symbols
(xi1, xi2, xi3). The position-vector xi of this particle then satisfies the
relation
xi xi1 i1 + x i2 i2+xi3 i3 CJ = 1, 2,
= · · · , .N).
We have then a set of 3.N scalars xik (j 1, 1, 3, = · · · , .N; k. = 1, 2, 3)
which denotes the coordinates of the particles.
The center of mass of the system of particles is defined to be the
point C with position-vector xc determined by the equation
N
(35.2) m xc = Z: mi Xj.
j =I
72
then be replaced by integrations. Thus, if p is the density of matter
in the body, V is the region occupied by the body, dV is the volume
of an element of the body and x is the position-vector of a point in
dV, then
(35.3) m = j p dV,
v
first consider a single particle of mass m. Let l denote the length of the
perpendicular from the particle to a line L, as shown in Figure 41.
m
L�
Figure 41
The moment of inertia of the particle about the line Lis defined to be
the scalar I given by the relation
I= m 12•
Let p and q denote the lengths of the perpendiculars from the particle
to two perpendicular planes P and Q,, as shown in Figure 42. The
product of inertia of the particle with respect to these two planes is
defined to be the scalar K given by the relation
K = mpq.
73
[
p
I
p
Figure 12
f1 = � Tllj (X2j2+X�3) ,
j= I
N
(36. l) f2 = � m; (X2ja+X2ji) ,
j = I
N
la = � m; (x2;i +x2i2) .
j= I
The products of inertia of this system of particles with respect to the
three coordinate planes, taken in pairs, are denoted by K1, K2 and K3•
It is easily seen that
N
K1 = 1: Tllj Xh Xi3 '
j = I
N
(36.2) K2 = � mi Xfa xh ,
i= I
N
K3 = � mj Xh xh .
j=l
74
Ofcourse, ifthe system of particles forms a rigid body, the summations
in Equations (36.1) and (36.2) must be replaced by integrations.
If 11, 12 , 13, K1, K2 a nd Ka are known, the moment of inertia I of
the system about any line L through the origin can be found easily.
Figure 43
}/
1 :E mifi2.
j =I
=
Pi = xi sin 6 = Jb X xi j ,
75
N N
2 2 2
+ b3 � mi (xi1 +xi2 ) - 2 b2b3 � mi xi2 xi3
i= 1 i= 1
N N
It will be noted that b1, b2 and b3, which are the components of the
unit vector b on the line L, are also the direction cosines of L. Equa
tion (36.3) is the desired equation which permits a simple determi
nation of the moment of inertia I of a system about any line L through
the origin, once 11, 12, 13., K1, K2 and Ka have been found.
If it happens that K 1 = K2 = K3 = 0, the coordinate axes are said
to be principal axes o f inertia at the point 0. It can be pr.oved that at
1
every point there is at least one set of principal axes of inertia. In
many cases, principal axes of inertia can be deduced readily by con
siderations of symmetry of the system of particles. For example, at
the center of a rectangular parallelepiped the principal axes of inertia
are parallel to the edges of the body.
We shall now state without proof two theorems the proofs of which
are very simple and may be found in almost any text book on calculus.
The theorem ef perpendicular axes. If a system of particles lies entirely
in a plane P, the moment of inertia of the system with respect to a line
L perpendicular to the plane P is equal to the sum of the moments of
inertia of the system with respect to any two perpendicular lines
intersecting L and lying in P.
The theorem f parallel axes.
o The moment of inertia I of a system of
particles about a line L satisfies the relation
I= l'+m l ,
2
where!' is the moment of inertia of the system about a line L' parallel
to L and through the center of mass of the system, m is the total mass
76
of the system, and l is the perpendicular dista nce between L and L'.
In most books of mathematical tables there are listed the moments
of inertia of many bodies with respect to certain axes associated with
these bodies. By the use of such tables together with Equation (36.3)
and the above two theo rems , it is fre q uently possible to determine
rapidly the moment of inertia of a body with respect to any given ·line.
Figure 1:4
Let 6 denote the angle between w and x, and let p d enote the length
of the perpendicular from X to the line of action of w. The displace
ment dx of the poin t X in time dt has the following prop erties :
77
when the fingers are placed to indicate the sense of the rotation
6 from w to x;
(iii) its magnitude is pwdt, which is equal to xwdt sin 6, x being the
magnitude of the vector x.
In view of the definition in § 8 of the vector product of two vectors,
it then appears that dx =w X x dt . Thus
dx
=wxx.
dt
(37.1) v =wxx.
(37.2) v = u+w X x.
We shall now prove that angular velocity obeys the law of vector
addition. Let us consider a body which is rotating simultaneously
about two lines L and L' which intersect at a point 0 fixed in a frame
of reference S. These angular velocities can be represented by the
arrows w and w' in Figure 45. Let X be a general point in the body,
Figure 1:5
(37.3) v = wxx+w'xx.
To complete the proof we must show that (37.3) can be written in the
form v =w" xx, where w" is an arrow obtained by the application
78
of the law of vector addition to the arrows w and w'. Even though
angular velocity has not been assumed to satisfy the law of vector
addition, Equation (8.5) may be applied to the two products in (37.3)
to yield
V1 = W2 X3 - Ul3 x2+w'2 X3 - cu'3 X2
= (w2+w'2) x3 - (w3+w'3) x2,
and two similar expressions for v2 and v3• Hence we can write
v=w"xx,
wX i1 , w X i2 , wX i3 •
Hence
di3 •
(38. l) dt = WXla.
79
a = a1 i1 +a2 i2 +a3 i3,
and the time derivative of a, relative to S, is then
(38.3)
Equation (38.2) expresses da/dt as the sum of two parts. The part
oa/ot is the time derivative of a relative to the moving coordinate
system. The part w Xa is the time derivative of a relative to S when
a is fixed relative to the moving coordinate system.
When the origin of the coordinate system is not at rest relative to S
but has a velocity u, Equations (38.1) still hold, and hence also does
Equation (38.2).
Figure 46
80
M=mv.
m
dxe = £ mi dxi.
dt j=I dt
But
dXc dx;
=Ve, =v;
dt dt ,
81
with a line of action which passes through 0. The velocity relative to
S oftheJ-th particle in the body is then
Vj =CJ> X Xj,
and by Equation (39.1) the angular momentum of the system about
0 is then
N
(39.3) h = � mj xj X (w X xj) .
j=l
Because of the identity (9.3), we can then write
N
h � mj [ w x l - x/ xj · w) ] .
j= l
=
Now let us introduce coordinate axes with origin at the point 0 fixed
in S. The directions of these coordinate axes need not be fixed in S.
As before we denote the coordinates of the J-th particle by (xj1, xj2,
xi3) . The component h1 of h then has the value
N
2 2 2
h1 = � mi [w1 (xi1 +xi2 +xi3 ) - xi1 (xi1w1+xi2w2+xi 3w3)]
i=I
N N N
2
= cu1.� mi (xj22+xia ) - w2.� mi Xii Xiz - wa.� mi Xia xii
J=l J=l J=l
= l1w1 - Kaw2 - K2w3'
where 11, K2 and K3 are moments and products of inertia defined in
§ 36. There are similar expressions for h2 and h 3 • We have finally
vi = vc+wxxj.
82
Hence the angular momentum h of the system about the center of
mass C has the value
N
h =_I: mi xix(vc+wxxi)
J= I
(
= . � mi xi Xvc+
lJ=
) � mi xix (wxxi).
j= I
By Equation (35.2) the first sum is equal to mxc. Since the origin of
the coordinate system and the center of mass C of the body coincide,
xc = 0. Thus
N
h =jI:th-x-
J 1 x(wxx1-).
=l
The right side of this equation is the same as the right side of Equation
(39.3). Hence in the present case the components ofh are also given
by Equations (39.4).
We have then the important result: Equations (39.4) may be usedfor
the determination of the components of the angular momentum h of a rigid
body about either a fixed point 0 in the body or the center of massC of the body.
In the two cases the origin of the coordinates is at 0 and C, respectively, the
directions of the coordinate axes being quite general. Equations (39.4)
cannot be used in the case of the angular momentum of a rigid body
about a moving point which is not the center of mass of the body.
the double sum in Equation ( 40.2) vanishes, and we can then write
(40.2) in the form
dM
(40.3) F '
dt
=
the system, and Ve is the velocity of the center of mass of the system.
Thus Equation (40.3) can be written in the form
dvc
(40.4) m F
_
dt - .
41. The motion of a rigid bo4v with a fixed point. Let us now con
sider a system of particles which constitutes a rigid body with a point
0 fixed relative to a Newtonian frame ofreference.
Theorem I. The rate of change of the angular momentum of the
body about 0 is equal to the total moment about 0 of the external
forces.
Proof. Let us introduce coordinates with origin at 0. Then
· dx;'
(41.l) V
J
dt
=
84
where vi, xi and t have the usual meanings. By Equation (39.1), t h e
angular momentum h o f the body about the fixed point O is
N
h=
j
� mj X;XV;,
l
and so
dh
( 41.2) -=A+B '
dt
where
N N
h· h·
A = L m; -d 1Xv ; , B L m; xi x -1
i =I
= •
j= I f dt
A= L m ;v;XV; = 0.
j= I
where
N
(41.3) G = L X·XF·
i=I 1 J'
N N
H = � L XjXFjk•
j=l k=I
It will be recalled that Fi is the external force acting on the j-th par
ticle and Fik is the internal force exerted on the j-th particle by k-th
particle. We note that G is the sum of the moments about 0 of the
external forces. Now
N N N N
H =L L x ;XF;k = � .L xkxFk;·
j= l k=I k=I ;=l
Thus
N N
( 41.4) 2 H =L L (:x;xF;k+xkxFk;).
i=I k= 1
85
But Fk; = -F;k. Thus (41.4) becomes
N N
2H = k � ( x-;- xk ) x Fi".
j= 1 k=l
Since the lines of action of the vectors Xj - xk and Fik coincide, their
vector product vanishes. Hence H = 0, and B = G, so Equation
(41.2) reduces to the form
(41.5)
(41.6) h1 =
f1<.u1' h2 = f2W2' ha = l3<.u3'
where li. 12, 13 are the moments of inertia of the body about the coor
dinate axes, and w1, <.u2, w3 are the components of the angular velo
city w of the body about 0.
In most cases the coordinate axes will be fixed in the body and will
hence have an angular velocity w about 0. However, in a few special
cases when the body has a certain symmetry it will be found possible
and desirable to choose coordinate axes not fixed in the body. To
include such special cases we denote the angular velocity of the axes
about 0 by S2, which may or may not differ from w. According to
Equation (38.3) we then have
or
�� �� +S2xh
=
(41.7)
86
l:"rom this equation we can read off the components of the vector
dh/dt. According to Equation (41.5) these components are equal to
the components of G. Hence we have the equations
11<»1 - l2(.t)2na+la(.t)an2 = Gu
(41.8) 12<»2 - la(.t)an1 +11 (.t)1na = G2,
la<»a - l1(.t)1n2+l2(.t)2n1 = Ga.
In the case when the coordinate axes are fixed in the rigid body,
then Q =wand so (12.8) reduce to the form
G' = � x;Xm;gk
; -1
= ( f X;)
j= l
m; xg k.
87
where m is the total mass and xc is the position-vector of the center of
mass. Thus
G' = mxcXgk = XcX (mgk).
But xcX (mgk) is the moment about 0 of a single force mgk equal
to the resultant of the gravity forces and acting at the center of mass
C of the system. This completes the proof.
Example I. A sphere of radius a is placed on a rough plane which makes
an angle cc with the horizontal, and is then released. Find the distance the
sphere moves down the plane in time t.
Figure 4 7 shows the configuration of the system at a general time t.
mgj
Figure 47
88
by gravity on all the p articles of the sphere may be replaced by a single
force mgj acting at C, as shown, where j is a unit vector . The reaction
of the plane is a force which may be resolved into a force N i3 normal
to the plane and a force -T i1 along the plane, as shown. The moment
G of the external forces about 0 is given by the relation
G = ocxmgj,
since the moments ofN and T about 0 are equal to zero . But
Thus
G = ai3 X mg (i1 sin IX - i3 cos IX)
= mgai2 sin IX,
so
(41.11)
w = <I> i2,
where the superimposed dot denotes differentiation with respect to t.
Thus
The moments of inertia of the sphere about the coordinate axes are
Thus
ii> 5g sin IX
-
_
7a '
and two integrations then yield
89
_ 5g sin oc 2
<I> t
- 14a '
-since <I> = <I> = 0 when t = 0.
If z is the distance the sphere has rolled down the plane in time t,
then z = a<I> and we have
Figure 48
plate as the origin of the coordinate system. We also choose the unit
vectors i1 and i2 parallel to edges of the plate, as shown. The unit
vector i3 is then perpendicular to the plane of the plate. The directions
of these three vectors are principal directions of inertia of the plate
at 0. The moments of inertia of the plate about the coordinate axes
are
2 2 2
(41.15) /1 = !ma22, 12 = }ma1 , /3 = }m(a1 +a2 ),
where m is the mass of the plate, and 2a1 and 2a2 are the lengths of
the edges.
90
The plate has an angular velocity w the line of action of which is
the diagonal AC and the magnitude of which is the given constant
w. Ifoc is the angle between w and i1, then
(41.16)
and
w = wi1 cosoc+wi2sinoc.
Thus
(41.17)
Since the coordinate axes are fixed in the body, Euler's Equati ons
(41.9) apply. We substitute in these equations from (41.15) and
(41.17), obtaining the relations
G1 =0, G2 =0, Ga =tm(a12-a22)w2sinoccosoc.
Thus the moment G about 0 of the external forces is normal to the
plate and rotates with it. Hence the forces exerted on the shaft by the
bearings must be in the plane of the plate . Let us denote these forces
byRand -R, as shown in Figure 48. We must then have
2cR =Ga
= .l mw2 a1 a 2 ( - a22) .
a12
6 c(a12+a 2)
2
By Newton's third law (§ 31), the forces exerted on the right and
left bearings are -R and R, with magnitudes R given in Equation
(41.18) above.
91
axis of symmetry is large. For example, the disc and shaft in Figure
49 constitute a gyroscope. ··
Figure 1:9
In Figure 50 the line OA is the axis of the gyroscope, the fixed point
being at 0 and the center of mass being at C. We introduce a fixed
unit vector j pointing up from 0, and a set of moving orthogonal
Figure 50
G = (l i3) X (-mgj).
92
But
w = sia+pj.
Because of Equation (41.19) we then have
Q =Pj
=p (i1 sin 6+ia cos 6).
Thus
(41.23)
We now substitute in Euler's equations (41.8) from Equations
(41.20), (41.21), (41.22) and (41.23) to obtain the relations
93
gyroscope is vertical, and the gyroscope is said to be "sleeping". If
0 is not equal to zero, then Equations (41.24) and (41.26) yield
p = constant, s =constant,
laP la
The quantities p and e may be observed readily. The corresponding
spin s may be computed by means of this relation. If the precession
is small, we note from Equation (41.28) that the spin is large and has
the approximate value
[mg
S=-·
la
42. The general motion of a rigid body. We now consider a rigid body
moving in a general manner. It may or may not have a fixed point.
The motion of its mass center can be determined from Theorem 2 of
§ 40, which applies to the motion of any system of particles. This
theorem yields
d c
( 42.1) m v =
F'
dt
where m is the total mass of the body, vc is the velocity of its center of
mass, and F is the sum of the external forces acting on the body.
Integration of ( 42. l) gives the position-vector xc of the center of mass
C of the body as a function of the time t.
94
To do this, we choose the origin 0 of the coordinate system at the
center of mass C of the body. We then consider the body as having
a velocity of translation vc plus an angular velocity w with a line of
action through C. The velocity vi of the j-th particle is then given by
the relations
dx·
(42.2) /
Vj =vc+ d = vc+w X Xj •
h = � m·X·XV·
11
i=l J>
dh
(42.3) -=A+B
dt '
where
N N
dx· dv·
(42.4) A=� m· -�xv·
d J ' B =.� mixi X -d 1 .
i=I J t ;=I t
From Equation (42.2) we then have
N
A=� m·(V·
J J
- vc) Xv·J
j=l
N N
=j� m·V·XV·-V
11 I cX j�l m·V·
1 1
=l =
=0-vcXM,
(42.5)
d
d � = G .
We have thus the result: the rate of change of the angular momentum of
a body about its center of mass is equal to the total moment of the external
forces about the center of mass.
95
We have placed the origin of the coordinate system at the center of
mass C of the body. If we choose the coordinate axes to coincide with
principal axes of inertia of the body at C, then just as in § 41 we obtain
Equations (41.6) and finally Euler's Equations (41.8) from which we
can find the unknown quantities w1, w2 and w3 which characterize the
rotation of the body about its center of mass.
In conclusion, it should be noted particularly that the equatio n
dh/dt = G can be used only in the two following cases: (i) the body
has a fixed point and the origin is at this fixed point; (ii) the origin
is at the center of mass of the body.
Example I. A g yroscope with a constant spin is carried along a horizontal
circular path at a cons.tant speed, with its axis tangent to the path of its center
of mass. Find the forces exerted on the axle of the gyroscope by the bearings in
which the axle turns, neglecting gravity.
We choose the center of mass of the gyroscope as the origin 0 of the
coordinate system. The path of G is shown in Figure 51; it is a circle
Figure 51
96
27t
+2 11
• •
w =
Sia
7t aI v
-
v
-11+s13.
• •
=
Thus
(42.6) <.u3 = s •
v.
Q =
;z11
whence
(42.7)
Also
(42.8)
sv
G1 = 0 , G2 = -13 - , G3 = 0 .
a
2bR = 13 �
a
whence
Problems
l. A particle moves on the curve x2 = h tan kxl> x3 = 0, where h
and k are constants. The x2 component of the velocity is constant.
Find the acceleration.
97
2. A particle moves with constant speed. Prove that its acceleration
is perpendicular to its velocity.
3. A particle moves on an elliptical path with constant speed. At
what points is the magnitude of its acceleration (i) a maximum, (ii)
a minimum?
4. A particle moves in space. Find the components of its velocity and
acceleration along the parametric lines of spherical polar coordinates.
5. A particle moves in space. Its position-vector :x: relative to the
origin of a fixed set of rectangular cartesian coordinate axes is given in
terms of the time t by the relation
where h is a constant and i1, i2 and i3 are the usual unit vectors in the
directions of the coordinate axes. Find the components of the veloc
ity and acceleration in the directions of (i) the coordinate axes men
tioned above, (ii) the principal triad of the path of the particle, (iii)
the parametric lines of spherical polar coordinates. Find the speed
and the magnitude of the acceleration.
6. A particle describes a rhumb line on a sphere in such a way that
its longitude increases uniformly. Prove that the resultant acceleration
varies as the cosine of the latitude, and that its direction makes with
the inner normal an angle equal to the latitude.
7. Two forces A and B act at a point. If 11. is the angle between
their lines of action, prove that the magnitude of the resultant R is
given by the relation
2 2
R2 = A +B +2AB cos 11..
98
10. A force F acts on a particle of mass m. Find the magnitude of
the acceleration, given that (i) F = 6 poundals, m
= 3 lb., (ii) F =
6 lb. wt.,
m.= 3 lb., (v) F = 5 dynes, m = 10 gm.
where p and q are constant vectors and t is the time. Find the veloc
ity v and position-vector x of the particle in terms oft, given that v
= 0 and x = 0 when t = 0.
12. A particle of mass m is acted upon by two forces P and Q..
The force P acts in the direction of the x1 axis. The force Q. makes
angles of 45° with the axes of x and
2
= p sin kt and Q, =
x3• Also P
q cos kt, where p, q and k are constants and t is the time. At time
t = 0 the particle has coordinates (b, 0, 0) and is moving towards the
origin with a speed p/mk. Find the position-vector x of the particle.
Prove that the particle moves on an ellipse, and ftnd the center and
lengths of the axes of the ellipse.
13. A particle of mass m moves under the action of a force pe-q1
and a resistance -lv, where p is a constant vector, q and l are positive
constants, t is the time, and vis the velocity ofthe particle. Prove that
1
Xoo - x0 = Tq (p+ m q u)
where u is the velocity when t = 0, and Xo and Xoo are respectively the
position-vectors of the particle when t = 0 and when t becomes in
finite. Is the above result true when l = mq?
14. A particle ofmass m moves under the action of a force p cos qt
- kx , where p is a constant vector, q and k are positive constants,
t is the time, and x is the position-vector of the particle relative to a
99
distance from 0 to the particle. At time t = 0 the particle is at a point
Band has a velocity of magnitude u in a direction perpendicular to the
line OB. Prove that the orbit is (i) an ellipse if bu2 < 2y, (ii) a parab
ola if bu2 = 2y , (iii) an hyperbola if bu2 > 2y , where b = OB.
16. Find the moment of inertia of a circular disk of mass m and
radius a about (i) the axis of the disk, (ii) a diameter of the disk.
[Answer: (i) tma2; (ii) !ma2.J
17. Using the result of Problem 16, find the moment of inertia of
a circular cylinder of mass m, length 2! and radius a about (i) the
axis of the cylinder, (ii) a generator of the cylinder, (iii) a line
through the center of the cylinder perpendicular to its axis, (iv) a
diameter of one end of the cylinder. [Answer: (i) tma2 ; (ii) !ma2;
(iii) 1\ m (4l2+3a2); (iv) -(2 m (16l2+3a2).]
18. A circular cylinder has a mass
. m, length 2l and radius a. Rec
tangular cartesian coordinates are introduced, with origin 0 at the
center of the cylinder, and the x3 axis coinciding with the axis of the
cylinder. Two particles each of mass m' are attached to the cylinder
at the points (0, a, l) and (0, -a, -l). Find the moments and products
of inertia 11, 12, 13, K1, K2, and K3 for the system consisting of the
cylinder and the two particles.
19. A circular disk of mass m and radius a spins with angular speed
w about a line through its center 0, making an angle ex with its axis.
Find the angular momentum of the disk about 0.
20. For the system of masses in Problem 18, find the angular
momentum about the point 0 when the system has an angular speed
w about (i) the x1 axis, (ii) the x2 axis, (iii) the x3 axis.
21. A circular cylinder of mass m, length 2l and radius a turns
freely about its axis which is horizontal. A light inextensible cord is
wrapped around the cylinder several times. A constant force F is
applied to the end of the cord. If the cylinder starts from rest at time
t = 0, show that at time tit has turned through the angle Ft2/ma.
22. The circular cylinder of Problem 21 is again mounted with its
axis horizontal, and has a light inextensible cord wrapped around it.
A body with a mass m' is attached to the end of the cord. If the cylinder
starts from rest at time t = 0 show that at time t the cylinder has
100
2
mgt
turned t hrough the ang1e a( + ')' where g is the acce1erat10n
· d ue
m 2m
·
to gravity.
constant angle (1. with the axis of the disk. The disk turns in two
smooth bearings each at a distance c from the point 0. Find the reac
tions of the bearings.
26. A uniform rod of length 2l is free to turn about an axis L
perpendicular to it and through �cs center. The center of the rod
moves at constant speed v around a circular track of radius a, the
axis L being always tangent to the track. Deduce the equations of
motion of system.
101
CHAPTER IV
PARTIAL DIFFERENTIATION
43. Scalar and vector fields. Let Vdenote a region in space, and let
X be a general point in V. Let x1, x2, x3 denote the rectangular cart
esian coordinates of X, and let x denote the position-vector of X. Then
x= x1i1+x2i2+x3i3•
Let us now consider the case when there is associated with each
point in the region Va value of a scalar f Then we write f = f(x1,
x2, x3), or more compactly
f =f( x) ·
The values off associated with all the points in V constitute a scalar
field.
Let us now consider the case when there is associated with each
point in the region V a value of a vector a. Then we write a = a ( x1,
x2, x3), or more compactly
a=a x ( ).
The values ofa associated with the points in Vconstitute a vector field.
It is frequently necessary to consider scalar and vector fields which
vary with a parameter, such as the time t. In such cases we write
from some fixed point Q on C. It was seen in§ 28 that the vector :
is a unit vector tangent to C in the direction of s increasing. This
vector is shown in Figure 52.
102
i,
Figure 52
ds
Let us now consider all curves through X having the same tangent at
X as Chas there. For each of these curves the value at X of the right
side of Equation ( 44. l) is the same. Thus at each point X there is
" d () d• •
(44.2)-
•
'f
uX1
(.d .
\l = 11 ""'} + 12 ""'} + 13 "l
UX2 UX3
J .d)1
or
'f . J.f • ()j . J.f
\l =11T+12T + 1a""'l"
uX1 VX2 UX3
103
The expression \!J is often called the gradient off, and is denoted
by gradf.
We note that the right side of Equation (44.1) is equal to the scalar
product of\Jjand d x/ds, by Equation (7.2). Thus
dx
dj \J -,
( 44.3) = f
ds
-
ds
or \
(44.4) dj = \J+·t
ds J '
-
by Equation (44.4).
(45.1)
Then
3
• Jf
( 45.2) \Jf="f. r=l
·
dx,
1,- .
= \Jf +\Jg.
Theorem 2. If f is a function of a single scalar field u, then
\Jf=4f\Ju.
du
(45.4)
so
105
dj du
f=�
\7
r=I
j,
du dX,
{)u
=dj � i,
du r-1 Jx,
= dj 'Vu.
du
Theorem 3. Iff is a function of n scalar fields u1, u2, • • • , u. , then
(45.5)
Hence
f Jj Ju,
'VJ= � r-1
i,
s=I Ju, Jx,
.
/
n
Jj
= r., -::;-- \Ju,.
s= I C/Us
3
{)
f\lg =f2.. i, -i'g
C/Xr
r-1
{) {)
= 11 f C/�
g + .
12
!J g +.
13
f� g
C/
•
�
C/X3
·
Xl X2
106
Theorem 4. If f and g are scalar fields, then
(45.6) 'V (Jg) = f'J g+gVf.
Proof. We have
3
0
V(fg) =,
:1 i,. ox, (Jg)
� i, fog + g
= 1
( of)
OX,
r= dX,
= f 'Vg + g'Jf.
(·
This operator has the obvious meaning
(46 . 1 ) a· "'
v =
•
( a1J.1+a2J.2+a3J.3)
• •
•
0
ox1
• 0
OX2
•
s of s - oh
(46.2) (a·'V)f=l: a,-, (a·'V) b = l: a,-·
r= I OX, r=I OX,
We note that
(a· V)f-:- a· Vj.
The operator ax V caJ:l be .considered similarly. We have
_
3) "' • • • (· () . 0 + . 0 )
(46 . ax v = ( a111+a212+a313) X 11 -::;- + 12 T 13 -::;- ,
0X1 0X2 OX3
or
or
107
(46.5) ax\/ =
dX1 dX2 X3 d
The operator a X \/ is a vector operator. It can be applied to scalar
fields. Thus, ifj is a scalar field, then
i 1 i 2 ia
(46.6) (ax \/)f =
a
Jj
1 J.a2 Jaa
We note that
dX1 Jxf� Jxf. 3
(ax \/)f = ax \If.
In writing the expressions (a·\/)j, (a·\/) b and (a X \/) j, one must
exercise care in the matter of the order in which the symbols appear,
since the operator \/ and all operators constructed from it operate
only on ·the functions on their immediate right. Thus, for example,
1
+ +
·
r=l
=
r=I
=
r=I
But
i,·b = b, (r = 1, 2, 3).
Thus
(46.9)
108
The expression \l · b is often called the divergence of b, and 1s
written div b.
We also have
'7 d
(·11 -;:;-- + 12 + 13 )xb . d ()
xb
•
(46.10)
•
v -;i- -::;--
OX 1 oX 2
=
oX3 .
Thus we may write
(46.11) 'Vxb =
(3Li., -;;-d) xb L i,.x -::;-Jb = 3 -::;- (i,xb).
=
3
L
()
r=I ox, r=I OX1 r=I OX1
But
i1xb = i1X (b1i1+b2i2+b3i3)
= h2i3 - h3i2
by Equations (8.9). Similarly we have
(46.12)
(46.14) \l ·
a+'V · b,
(a+b) = \l ·
_ � (, -r=I
da, + Jb,
dX, dx,
)
= 'V·a+'V·b.
109
Proof of Equation (46.15). From (46.13) we have
i1 i2 i3
() {) {)
\7x (a + b ) = dX2 dX3
dX1
ai+b1 a2+b2 a b
a + 3
i1 i2 is
() () ()
dX 1 dX2 Jx3
a1 a2 as
i1 i2 ia
() () ()
+ d X1 d X2 d Xs
h1 h2 b3
= \7xa+'Vxb.
(46.17) (a X \7) b =
· ( a2 ffx3 - a3 JxJ (i1 b) ·
(
+ a 3 !.__a
dX1 dX3 1 !..._) ( i-·b)
-�
But
i1·b = h1, i2·b = h2,
Figure 53·
quantities will be denoted by x1', x2', x3' and i1', i2', i3'. Figure 53
shows the coordinate axes and the associated unit vectors.
We now introduce the two operators
"' • J • J d
-::i- + 12 -::i- + 13 -::i- '
•
v = 11
UX1 UX2 UX3
J () ()
"''
v = 11
• ' � +.12 , � +.13 , -::i-;
UX1 UXz UX3
•
Letf be a scalar field, and let b be a vector field. We shall now con
sider proofs of the relations
111
field, the resultant field is likely to depend on the particular choice
of coordinates. However, Equations (47.1), (47.2) and (47. 3) state
that this is not the case. It is this property of del which is termed
"invariance of the operator del''.
Before proceeding to a proof of the above three formulas, we shall
develop some preliminary formulas. Let us consider the three angles
which the x1' axis makes with the axes of the system S. These are the
angles ell, 612 and 013 in Figure 53. We denote their cosines by all,
a12 and a13• These three quantities are the components of i1' relative
to the system S, and so
3
There are similar expressions for i2' and i3'. The three expressions
can be written in the compact form
i1 = � a,1 i,' .
s=I
Proceeding similarly for i2 and i3, we obtain the three equations
3
(4 7 . 5 ) i, = � asr i/ (r = 1, 2, 3).
s=I
Throughout the remainder of this section we will adopt the conven
tion that latin subscripts range over the values 1, 2 and 3, as in the
above equations.
We now introduce a set of nine quantities called the Kronecker
delta. This set is denoted by the symbol �'', and is defined by the
equation
;,rs = 1 if r = s
(47.6)
= 0 ifr =I= s.
112
Thus a11 = a22 = a33 = I and a23 = a31 = a12 = a32 = a13 = a21 = 0.
Since i1 , i2 and i3 are unit orthogonal vectors, they satisfy the nine
relations (7 3). These nine relations can now be written compactly
.
in the form
(47.8)
u=I
asu
3 3
(47 9 )
. a,,= L: L: a,, a,. a,. = L: a,, l: asu a,•.
l=lu=l l=l u=I
(47.10)
The nine equations in (47. I.I) are called the orthogonality conditions.
We could also run through the above derivation of ( 4 7 .11) but with
the roles of the primed and unprimed quantities interchanged. This
would entail substituting in Equations (47.7) from (47.5). In this
way we would obtain the orthogonality conditions in the form
113
(47.12)
3 3
= L: � b, a,, i,'.
s=l r=l
'
(47.16) x,' = L a,, x,, x, = I: a,, x, .
r= 1 r= 1
114
Jx', Jx,
(47.17) -::;---- = a,,' = a,, .
ox, Jx',
We are now in a position to turn to proofs of Equations (4 7 .1), ( 4 7 .2)
and (47.3).
Proofof Equation (47.1). This equation reads
\lf = \lf.
Now
{)
(47.18) \Jf= r=I
� �
i', dX r ,
and
Jf = � Jf Jx, .
Because of this relation and (47.4) we can write (47.18) in the form
Jf Jx, .
(47.19) \lj = � � � a,, i,
r=I s= I I=I dX1 dX ,r
But, from Equations (47.17) we have
Jx,
-::;---- = a,,.
ux'r
Thus (47.19) becomes
3 3 3
, Jj
j = "'I;, "'I;, L a,, a,11,
•
\1 """l
r=I s=I t=I OXt
3 3
Jf 3
= � "'I;, 1, L a,, a,1
•
-::i- •
1, -::i- a,,
s=I t=I OXt
3
3 Jj
=L 1, L -::i- a,1
•
•
and so
The truth of Equation (47.1) also follows from Theorem 2 of§ 44,
as this theorem states that \Jf points in the direction in which the
directional derivative ofjhas a maximum value, and that the magni
tude of \Jf is equal to this maximum value. This theorem then im
plies that both the direction and magnitude of \Jf are independent
of the coordinate system.
Proof of Equation ( 4 7 .2). This equation reads
\J'·b= \J·b.
Now
3 Jb'
\J'·b =I: -'.
s=l Jx',
= � �
s=l r=l
Osr (}b,
dx',
()b
3 3 3
\J' b= I: I: I: a,, a,1 T
t=l
·
3 3 Jb 3
� � � I: a,, a,1
1=1 UXp=l
= •
r-1
116
\7'·b = � � Jb, a,,
r=l t=l Jx,
=
� Jb,
,
r=I dX,
by a repetition of the arguments leading up to Equations (47 .10) and
(47.20). This completes the proof.
ProofofEquation (47.3). This is left as an exercise for the reader
(Problem 13 at the end ofthe present chapter).
(48.9) \7·x=3,
(48.10) \7Xx= 0.
(48.11) (a ·\7)x =a.
Direct proofs of all eleven of these formulas follow similar lines. We
shall present here only proofs of (48.1), (48.3), (48.5) and (48.8).
The proofs of the remaining formulas are left as exercises for the
reader (Problems 14, 15 and 16 at the end of the present chapter).
Proof of Equation (48.1). Because of Equation (46.8), which gives
some equivalent forms for \7 b, we have
·
117
s
d
V·(Ja) :E i,·- (fa)
r=l
=
dX,
Jf
� i, . ff da +
\ d X,
=
r=I dXr
a)
3 Ja 3 {)f
= f'Z 1,·- + 'Z 1,
• •
·a
r=I dX,
-·
r=l d X,
= f(V·a)+(Vf) ·a.
Proof of Equation (48.3). Because of Equation (46.8) we have
3 {)
(� X b ) =r� i,· a X b)
\J •
I dX, (
3 Ja 3 Jb
='Z i,· -xb + 'Z i,· ax-
( ) ( )
Jx, r=I
·
r=l d X,
3 Ja' 3 {)b
V·(axb) ='Z h· i,x- - 'Z a · i,x-
( ) ( )
r=1 dX, r=I dX,
3 Ja 3 Jb
h :E i,x--a·"Z
= · ' i,x-
,_, dx, r=I dX,
b·(Vxa)-a·(Vxb).
=
We now apply to the right side of this equation the identity (9.3)
for vector triple products, obtaining
3 Jb 3 Jb
ax(V xb) = 'Z i, a·-::;- - 'Z -::;- (a·i,)
(48.12)
( )
r=l dX, r=I uX,
3· Jb 3 {)b
'Z 1 a·- - 'Z a
( )
Jx, r=I 'dX,
= -
r-1 '
3 db
= 'Z i, a·- - (a· V) b.
( )
r=l Jx,
118
Similarly, by an interchange ofa and b we have
(48.13) b x ( \7 x a) =, � i (�: b)
1 , ·. - (b · \7) a.
= \7 (a·b) - (a·\7)b- (b \7 ) a.
If the last two terms here are taken to the left side of the equation,
(48.5) results.
In addition to this manner of proving formulas (48. 1 )-(48.1 1 ),
there is another manner of proof which applies to some of these for
mulas, and which is quite expeditious. We mention it here because it
affords one an opportunity to acquire additional facility in manipu
lations involving the operator del. This manner of proof consists in
applying to expressions involving del the permutation theorem for
scalar triple products (Theorem 1 of§ 9), and the identity (9.3) for
vector triple products which is
scalar or vector field must never be moved from one side of del to the
other, since del operates only on those quantities on its right. Also,
when such application is made, the resulting expression must have the
same scalar or vector character as the original expression.
For example, let us prove Equation (48.8) in this way. Ifwe apply
the identity (48. 1 4) formally to \7 x (\7 xa), we obtain the difference
oftwo·terms. The first term must be one ofthe following:
(48. 15) · ,
\7 (\7 a) \7 (a·\7), (\7·a) \7, (a·\7) \7.
119
we find that the term corresponding to the second term on the right
side of(48.14)is ('V \J ) a . Thus
·
\JX(\Ja) = \J ('V·a)-('V·'V)a.
We note that the operator \J \J satisfies the relation
·
I dZ1 d Z1 dZ1
I
dX1 dX 2 dX3
dZ 2 dZ2 Jz2
( 49.2) I'=
dX1 d X2 dX3
We shall consider only the case when I' does not vanish anywhere
in V, so that Equations ( 49.1) may be solved 1 for x1, x2 and x3 to
yield the relations
121
larly, we have parametric lines of z2 and z3• For example, in the case
of cylindrical Coordinates r, 6, X3, the parametric lines Of r, 6 and x3
k,
.c,,
Figure 54
where dx1, dx2 and dx3 are the infinitesimal differences between the
rectangular cartesian coordinates of the two points. From Equations
·
(49.3) we have
122
(r = 1, 2, 3).
(49.5)
where h1, h2 and h3 are known positive functions of z1, z2 and Z3•
The right side of (49 .5) is called the fundamental
quadratic fomt, or the
metric form.
Les s1, s2 and s3 denote the arc lengths of the three parametric
lines in Figure 54. From (49.5) we then have
(49.6)
(i) 'Vz1, being normal to the surface z1 =constant, has the same
direction as k1 ,
(ii) !'Vz11 is equal to the maximum value of the directional deri
tive dz1/ds,
(iii) this maximum value arises when tl?-e directional derivative is
taken in the direction ofk1, and is hence equal to dz1/ds1•
Because of (ii), (iii) and Equation (49.6) it follows that !'Vz1l = 1/h1;
and because of (i) we then have 'V z1 =k1/h1. ' rvations
Similar obse
regarding z2 and z3 then permit us to write
(49. 7)
k1 (k2 xk3)
· = h1h2h3 'VZ1 ('V z� X 'Vzs) •
·
But the left side of this equation is equal to one since k1, k2 and k3
form a right-handed orthogonal triad of unit vectors. Thus
123
(49.8)
Jj ()j ()j
\!f = -:;;---- \! Z1 + -;:;---- \! Z2 + -:;;---- \! Za ·
uz1 uz2 UZ3
( 50.1)
or
3 1 ()j
(50.2) \lf= � h- -;:;-- k,.
r-1 r u<;,
(50.3)
124
(50.4) \7 b· = \7 (b1h2ha \7z2 x V'za) + \7 · (b2h3h1 V'za x \7z1)
·
For the first of the three terms on the right side of (50.4) we can write,
because of relation (48.1),
(50.5) \7 •
(h1h2ha \7 Z2X\7Z)
s = \7(b1h�3) ( \7z2 X \7 z3)
•
() () ()
\} (h1h2hs) -::;-- (h1 h 2ha) \} Z1 + -;:;-- (h1h2hs) \7Z2 + -::;-- ( h1h2hs) \l Zs
uz1
=
UZ2
•
uza
Thus, since 'Vzl> \7z2 and 'Vz3 are mutually perpendicular, we obtain
()
'V (h1h2ha) · ( \7 Z2X\7 Za) -::;-- (h1h2h3) \l Z1 ( \7 Z2 X \7 Z3)
uz1
= ·
This relation and two similar relations involving the second and third
terms on the right side of Equation (50.4) then permit us to write
(50.4)
(50.8)
in the form
'V· b
l [()
-:;--- ( h1h2 h3) + ()
-::;-- (b2h3h1) + ()
-::;--
J
( b3h1h2) .
h1h 2h3
=
125
or
For the first term on the right side we can then write
[
and by Equation (45.5) we have
d d
\7 (b1h1) X \7 .<:1 = ( b1h1 )\7 z1 + �
�
oz1 oz2 (b1h1) \7 Z2
\lz1x'Vz1 = 0,
1 1
\lz2x \lz1 = - k2 xk1 - IT ka,
h2h1 2 1
=
1 1
\7 .<:a X \7 Z1 = ks X k1 2
hah1 =
hsh1 k.
Thu s Equat ion (50.10) reduces to
k2 J ks J
\lx(b1h1 \lz1) � (b1h)
1 - - o � (b1h1)·
h3-h1 oz a h2 h1 z2
=
This relation and two similar relations invo lving the second and third
terms on the right side of Equation (50.9) permit us to write Equation
[
(50.9) in the form
k1 d d
(50.11) \lx b= hJza
k2 [
Jz2 (bahs)- J.<:3 (b2 h2) J
J
d d
ha Jza (b1h1) - rh:.1 (bahs)
[
+ h1
k3 J . . J
126
+
h1h2 2 - dz2 (b1 h1)
Jz1 (b:)z) J ·
We note that if the curvilinear coordinates happen to be rectangular
cartesian coordinates, then (50.ll) reduces to (46.12), as expected.
Problems
2
1. If f = (x1) x2+(x2)2x3 - x1x2x3, find the directional derivative
off at the p oint A(l,-4,8) in the direction of the position-vector
a of A.
2. Iff = x1
(7tx2) +x3 tan (7tx2), find the directional derivative
sin
of/at the point A(l, 0, -2) in the direction of the vector drawn from
A to the point B ( 3, -3 , 4) .
3. Find a unit vector normal to the surface x�3 - x3x1+x1x2 - 1 0 =
\J.x" = nX'"2x,
where n is a constant.
7. Find \}r and 76, where r and a are the usual plane polar coor
dinates. Also, find the magnitudes and directions of \}r and 76.
8. If f r3 - cos2 a, where r and a are plane polar coordinates,
=
find \}fin terms of r, a and the unit vectors i1 and i2 associated with
the corresponding rectangular cartesian coordinates.
9. If/ and g are scalar fields, prove that
10. Ifj =
2
(X1) +xa V(X1) +(x2) 2 and g
2 X1X2X3, = find at the point
A (3, 4, 5) the expressions \}(jg) and 'V (fig). Note Theorem 4 of
§ 45, and Problem 9 above.
11. Iff X1X�3' a
= X1i1 - X2i2 and b
= X3X1i2 - X1XJ3' compute
=
the following: (i) (a·'V)/, (ii) (a·'V)b, (iii) (ax'V)/, (iv) (ax'V)· b ,
127
(v) (ax\7)Xb, (vi) 'V·b, (vii)'Vxb, (viii)a·('Vxb).
12. Let S and S' be two rectangular cartesian coordinate systems.
The axes of S can be moved into coincidence with the axes of S' by
a positive rotation of radians about the x3 axis, followed by a
i-7t
second positive rotation of i-7t about the bisector of the angle between
the positive axes of x1 and x2• ( i) Express the coordinates of S' in terms
of the coordinates of S, and conversely. (ii) If J = x2x3+x3x1 and
b =( x1 + x2) i1 +( x1 - x2) i2+ x3i3, express f and b in terms of quan
tities pertaining to the system S'.
13. Prove Equation (47.3).
14. Starting from Equations (46.8) and (46.11), verify Equatiors
(48.2), (48.4) , (48.6), (48.7), (48.8), (48.9), (48.10) and (48.11).
15. Using the identity (48.14), verify Equations (48.4) and (48.5).
16. Using the permutation theorem for scalar triple products,
verify Equation ( 48. 7).
17. If a is a constant vector, prove that \7(a· x) =a.
18. Prove that ax (\7X x) 0. =
128
27. If r, 6, q:i are spherical polar coordinates, describe the paramet
ric surfaces and lines, and show that
(ds)2 = (dr)2+r2(d6 )2+r2 sin26 (dq:i)2.
28. Prove that for the transformation from rectangular cartesian
coordinates to orthogonal curvilinear coordinates z1, z2, z3 for which
the metric form is as given in Equation (49.5), the Jacobian I satisfies
the relation h1h2 h31 = 1.
29. Express Equation (50.11) in terms of a determinant.
30. Write out the expressions \Jf, \J ·band \J xb in the case of
cylindrical coordinates r, 6, z .
31. Write out the expressions \Jf, \J.b and \J x b in the case of
spherical polar coordinates.
32. By setting b =\Jf in Equation (50.8), deduce an expression
for \72.f in terms of general orthogonal curvilinear coordinates.
33. Show that, in cylindrical coordinates r, 6, z, we have
J2j 1 J2j J2j 1 w
\7 + + +
Jr2 ;:2 J6 2 Jz2 -;:- Jr
=
2.f ·
,
Jk
JO2 =-k1,
dkq:i2=k3 cos 0v,
J
!
d 3=0, =0, ��3 ��3
-k1 sin·e -k 2 cos e.
=
129
CHAPTER v
lNTEGRATION
x = x1(u)i1+x2 (u)i2+x3(u)i3,
where x is the position-vector of a general point X on the curve, u
is a parameter with the range o: < u < �, and for this range of u the
functions x1(u), x2(u) and x3(u) are continuous with continuous first
derivatives. A curve which consists of a finite number of regular arcs
joined end to end, and which does not intersect itself, is called a
regular curve. Throughout this chapter we shall consider only regular
curves, and shall refer to them simply as curves.
Let us consider a curve C with terminal points A and B, as shown in
Figure 55. Letf (x1, x2, x3) be a function which is single valued and
-�ft Q,,
K,
Q,
Figun 55
ment Qp_1Qp) (p 1, 2,
= , N) is denoted· by lisp. Let Xp be a point
· · ·
on the arc QP-1QP, and let its coordinates be (xp1, xP 2, xp3). The line
integral off over C is then defined to be
N
(51.1) lim
N-+ooP=I
1: J (xp1, Xp2 , xp3) lisp= J J ds.
C
Asp-+- 0
This limit is independent of the manner in which the curve C is divided
into parts, since f is continuous and single valued on C. If f = 1
130
everywhere on C, then Equation (51.1) defines the arc length of C.
Let X be a general point on the curve C, as shown in Figure 56.
Figure 56
Let s denote the arc length of C measured from the end A of C. The
vector dx/ds was seen in § 28 to be a unit vector tangent to C in the
direction of s increasing. Denoting this vector by t, we have
dx
(51.2)
t ds.
=
Let b ( x1, x2, x3) be a vector field defined over C. The orthogonal
projection of b on the unit tangent vector t is called the tangential
component ofb. Ifwe denote it by b1, we have
(51.3) b, = h·t.
c c
(51.5)
Thus
c c c
131
( 5 1.8 ) J bxt ds J b xdx =
c c
Thus
dx = i1dx1+i2dx2 (ii-1-�i2) du,
=
(ii) On C we have
2 3
x2 = 2x1 , J = (x1) +8(x1) ,
132
(i) We have
c 0
]
From we have
[
(ii) (51.9)
dx1 dx3
b . dx X2 + (xa+x1)2+x1 d
dx2 x2.
=
dx2
But on C we have x1 =tx2, x3 =0, so
4
B fo.o. •)
Figure 57
from the point A to the point B, the curve C being a portion of the
intersection ofthe cylinder (x1)2+(x2)2 =a2 and the plane x1+x3 =a.
133
Now
But on C we have
2
(x2) = a2 - ·(x1)2, X3 =a -X1
dx
-=-2 X1 dxa
- = -1.
dx1 X2 ) dx1
·
Thus
0
Figure 58
134
N
(52.1) lim
N->ooP=I
� f(xp1,xp2,Xpa) !:l.Sp= f fdS.
tl.Sp
.... O
S
This limit is independent of the manner in which S is divided into
parts, since f is continuous and single valued throughout the region
S. If f = 1, then Equation (52.1) yields the surface area of S.
Let us suppose that Sis a regular surface element, as shown in Figure
59. To evaluate the surface integral in Equation (52.1), we let X be
"
.,,,.----5'---- -- ......
,,
c·-----
Figure 59
dS' = n3 dS.
Let the equatio.n of the surface S be
(52.2) X3 - g(x1, X2) = 0..
Denoting the left side of this equati.on by G, we have by Theorems 2
and 3 of§ 44,
J G. JG. JG.
n v Gj = "'
ln v G = -;-- 11 + -::;- 12 + ·::i- 1 3
u v
X1 UX
X2 3
Jx3 • dX3. •
=
- T
1 1 - :I l2+1a.
vx1 vx 2
135
Thus
( 52.3)
We then have
To evaluate the integral on the right side, we may write dS' dx1
dx , and then x1 -and
=
(52. 7) dS =ttdS,
so Equation (52.6) may take the form
s s
136
(52.9) j b dS = i1 f b1dS+i2.f b2dS+ia f b3dS,
s s s s
.,
Figure 60
s s
But on S, we have
x3 = 2 - 2x1-2x2,
2 -
f = (x1) +2x2+(2 - 2x1 - 2x2) 1
= ( x1 - 1)2.
137
Thus, ifS' denotes the projection of Son the x1x2 plane, we have
I 1-x1
dS 3dx1 dx2•
=
Thus
J b.dS = 2 J (x2+x3)dx1 dx2•
s s
v x,...
;. .@
�Y,
Figure 61
( 53 . 1 ) lim
N-+ooP-1
L f(xpuXp2,xp3) �VP= jJ dV.
v
AVp-+0
138
This limit is independent of the manner in whi ch V is divided into
parts, sincef is single valued and continuous in V.
We may also consider triple integrals with integrands whi ch are
vectors. Thus, if b is a vector field which is single valued and contin
uous in V, we have, following the definition of integration of vectors
in§ 13,
To evalute triple integrals, one may divide the region V into ele
ments by means of three systems of planes parallel to the coordinate
planes. The value of the integral is then found by the performance of
three integrations with respect to the rectangular cartesian coordinates
x1, x2 and x3. Or we may divide V using parametric surfaces of a
Problems
1. If x is the prosition-vector of a general point on a circle C of
radius a, and t is the unit tangent vector to C, evaluate Jt · dx.
c
2 2
2. If f = (x1) - (x2) , evaluate the line integral off along the line
x1 +2x2 = 2 from the point A(O,l) to the point B(2,0).
3. Evaluate the line integral in Problem 2 when the curve C consists
of (i) the two line segments AD and DB, where D has coordinates
(1,1), (ii) the two line segments AO and OB, where 0 is the origin.
4. If f = 8 lx1 9, evaluate the line integral off along the curve
-
2
(x2 ) = (x1)3 from the origin to the point (1,1).
5. If f = x2x3+x3x1+x1x2, evaluate the line-·integral ofjfrom the
origin 0 to the point B(l,2,3) along the path consisting of (i) the line
segment OB; (ii) the three line segments OD, DE and EB, where D
and E have coordinates (1,0,0) and (1,2,0), respectively.
139
11. If f = x1 +x2+x3, evaluate the line integral of J along the curve
x =a cos u i1+a sin u i2+a u cot IX i 3, (O<u<in:),
where a and ix are constants.
7. Ifb (x1 - x2)i1 +xJ.2, evaluate the line integral of b along the
=
following curves: (i) the path in Problem 2 above, (ii) the two
paths in Problem 3 above, (iii) the path in Problem 4 above.
2 2
8. If b 2x1x.
= J 1 + [( x1 ) - (x2) ]i2, evaluate the line integral of b
from the point A(O,O) to the pointB(l,l) along the following curves:
2
(i) x2 =x1, (ii). (x1) x2, (iii) x1 = (x2)2•
=
2
14. If b = x1i1 +(x2) i3, evaluate the surface integral ofb over the
region Sin Problem 13, the origin being on the negative side of S.
15. If f (x1)2+ (x2)2, evaluate its surface integral over the region
=
140
c
.0'-----1>--- x,
Fig;re 62
Figure 62, and let b be a vector field continuous and with continuous
first derivatives in the region S. Then
s c
in the case when C can be cut by a line parallel to the x2 axis in two
points at most. Figure 63 illustrates the situation. On C there are
two points D and E where the tangent to C is parallel to the x2 axis.
Let d and e be the abscissas of D and E, respectively. These points
divide C into two parts C' and C". At a general point X(x1, x2) in S
we introduce an element of area lying in a strip parallel to the x2 axis,
the left edge of the strip cutting C' and C" a t the points X' ( x1, x2')
and X"(x1, x2"), as shown. Then
141
e x2"
e e
d d
d '
e d
Let us now consider the case when C can be cut by a line parallel to
the x2 axis in more than two points, such as the case of the curve C
in Figure 62. Here we have only to join the points F and G where
there are tangents parallel to the x2 axis by a curve K which is contained
E
D X\x,.x,)
Figurt 63
142
K are required. And we can do likewise when Sis multiply connected,
that is, S has holes in it and C then consists of several isolated parts.
In an analogous fashion we can prove that
(54.4)
f (�!: �! �!:)
·
(55.2) + +
.
dV = f (h11t1 +h2n2+b3n3) dS .
v s
in the case when S can be cut by a line parallel to the x3 axis in two
points at most. Figure 64 illustrates the situation, T being the projec
tion of S on the x1x2 plane. On S there is a curve C consisting of points
where the tangent plane to S is parallel to the x3 axis. The curve C
cuts S into two portions S' and S". At a point X(x1, x2, x3) in V we
introduce an element of volume lying in a prism parallel to the x3
axis, the vertical line through X meeting S' and S" at the points X'
(x1, x2, x3') and X" (x1, x2, x;'), as shown. Thus
143
,'
x
(55.4)
s·
S"
o. ------- --- x,
---------- .......
... T
Figure 64
Let n' be the unit outer normal vector at X1, and let dS' be the area of
of the element cut from S' by the vertical prism. Let us define and
"
n
Let us now consider the case when S can be cut by a vertical line in
more than two points. In such cases we can always divide V into a
144
V2, • • • can be cut by a vertical line in two points at most. The above
proof of (55.3) then applies to the regions V1, V2, If we apply
,
• • • •
(55.3) to V1 V2, • • • and add, the surface integrals over Kl> K2, · · •
f \7 . ( \7 j) d V f \7ID dS'
=
v s
or
and df/dn is the directional derivative ofjin the direction of the outer
normal to the surface S.
56. The symmetric form of green's theorem. Letf and g be scalar fields
with continuous second derivatives in a closed region V bounded by
a surface S. We may then apply Green's theorem as stated in Equa
tion (55.1), but with the vector b replaced by f \lg. This yields
(56.1) j \7
v
· (f 'Vg) dV = jf \lg·n dS.
s
But
v · (f 'Vg) =f (\7 · \l)g+'V I \lg
= f \7 2g + v I v g,
145
(56.2) f (f\12g+\lf \lg) dV = ff �! dS.
v s
s c
( 5 7 .2)
dbs db2') + · (�b Jb3) + n3 (�b2 �bi)]
J [n1 (dX2 n2 1 _ _ dS
s dx3, _
in the case when S is a regular surface element and the positive side of
146
S is that side on which the unit normal vector n points in the direction
of increasing x3• Figure 65 illustrates the situation, and shows the
x,
·r--+----1 1---+--- .,
X'
--
.
. C'
S'
Figure 65
unit tangent vector t of C and also the region S' in the x1x2 plane into
which S projects.
Now
Let the equation of the surface Sbe x3 = g(x1, x2). Then on S, we have
b1(X1,X2,X3(X1,X2)] = C1(X1,X2),
dC1 db1 + (}bl dXa.
(57.5)
_
_
= -l1+l2,
where 11 and 12 denote the two integrals on the right side of this
equation.
Let us consider 11• We have n·i3 dS = n3 dS = dS', where dS' is
the projection of dS on the x1x2 plane. Since c1 is a function of x1 and
x2 only, we can then write
147
11 = - f ;:1 2
dS.
'
s'
then have
(57.7) 11 =
f C1(X1, X2)dx1 f b1[X1,X2,X3(X1h)]dx1 f b1dX1·
=
=
C' C' C
Hence
and so
(57.8) 12
f n· y-
dX db1
dS.
T
=
X2 X3
S
dX
n·- =0.
dX2
Thus Equation (57.8) yields 12 = 0, and from Equations (57.6) and
(57.7) we can then conclude that Equation (57.1) is true.
When the positive si.de of S is chosen so that the unit normal vector
n points in the direction of decreasing x3, the proof of Equation (57.3)
is similar to the above, the only differences in the proofs being that
in the present case n3 is negative and the direction of integration around
the curve C is opposite that in the above proof.
When the surface S is not a regular surface element, we divide it
into a number of regular surface elements S1, S,
2 • • • by a number of
curves L1, L2, • • • • The above proof of (57.3) then applies to the
regions S1, S2, • • • • Ifwe apply (57.3) to these regions, and add, the
line integrals over L1, L,
2 · · • cancel, andwe hence establish Equation
(57.3) for the entire region S.
148
In a manner similar to the above we can prove that
v s
(58.5) j(nx\J)fdS=jtfds,
s c
149
(58.6) j (nx\l)xbdS=J txbds.
s c
But we have
\7 (Jc)
. = \7 f c,
150
The six integration formulas ( 58.1)-(58.6) may be written com
pactly in the form
where T can denote a scalar field or a vector field, and the asterisk has
the following meanings: if T is a scalar field, it denotes the multipli
cation of a vector and a scalar; and if T denotes a vector field, it
denotes either scalar or vector multiplication. Thus, for example,
if T denotes a vector field b and the asterisk denotes scalar multipli
cation, then Equation (58.10) becomes (58.2).
(59.1) 'VXb=O.
Let cp be any scalar field with continuous second derivatives; and
let us write b = \Jcp. Then
\JXb=:VX\Jcp =0,
so a vector b defined as the gradient of a scalar field is irrotational.
We shall now show that an irrotational vector field b has the fol
lowing properties :
(i) Its integral around every reducible circuit in V vanishes.
(ii) When V is simply connected, b is the gradient of a scalar
field.
To verify the first property, we consider a general circuit in V
which is reducible, that is, it can be contracted to a point without
leaving V. Let S be a surface entirely in V and bounded by C. If we
assume that b has continuous first derivatives, then Stokes's theorem
(5 7 .1) yields
fb-tds = J n·('Vxb) dS = o
c s
by Equation (59.1).
151
To verify the second property, we let X be a general point in V,
and let X0 be a given point. We also let C' and C" be any two paths
m V from X0 to X, as shown in Fi gu re 66. Bec ause of property (i)
x
x,
Figure 66
above, the line integral of b from X0 to X is the same for paths C' and
C" and hence has the same value for all paths in V from X0 to X. Thus ,
if we write
x
(59.2) cp = J h·dx,
x,
60. Solenoidal vectors. A vector field b(x1, x2, x3) is said to be sole
noidal in a regi on V in space if everywhere in V we have
(60.1) \7·h=0.
152
Let q> be any vector field with continuous second derivatives, and
let us write b =\7 X q>. Then
\l·b = \7·(\!Xq>) =0.
We shall now show that, ifb is any solenoidal vector field, there exists
a vector field q> such that b = \7 X q>.
To prove the proposition, we must solve the scalar equations
dcp3 - dcp2'
(60.2) b1 -
- dX2 dX3
dcpl - dcp3'
(60.3) b2 =
dx1
-
dX3
dcpz - dcp1'
(60.4) b3
- dx1 dx2
for cp1, cp2 and cp3, where b1, b 2 and b3 are given functions subject to
the condition
(60.5)
Let us choose Cfii =0. Then we have from Equations (60.3) and
(60.4) by partial integrations with respect to x1,
(60.6)
a,
(60.7)
a,
b1
= - x1
x1
Jx2 Jx3 dx2 dx3
a,
153
This equation is satisfied if we choose h = 0,
<p1 = 0, Cfl2 =
.,
fba(X1, X2, X3) dx1,
<p3 =
-
a1
fb2(X1, X2, X3) dx1+ fb1(a1, X2, X3) dx2,
a:a
where all integrations are partial integrations, and a1 and a2 are con
stants. The function <p is called a vector potential function.
In the above proof, several arbitrary selections have been made.
This indicates that a given solenoidal vector field b does not possess a
unique vector potential function. In order to see this more clearly,
we let <p be one vector potential function corresponding to the sole
noidal vector field b, and let] be any scalar field. Then
Problems
1. Let C be a closed curve in the x1x2 plane. Prove that the area
A of the region S enclosed by C is given by the relation
154
S, and n is the unit outer normal vector to S, prove that the volume
V of the region enclosed by S is given by the relation
V=t J n·xdS.
s
J n·(Vxb) dS = 0.
s
2
6. If b = a1 (x1) i1+a2(x2)2i2+a3(x3)2i3, where a1, a2 and a3 are
constants, evaluate the surface integral of b over the sphere through
the origin with center at the point A (a1, a2, a3).
7. If b = (x1)2i1+x1x2i2+x3i3, evaluate the surface integral of b
over the cube bounded by the planes x1 = 2, x2 = 2, x3 = 2 and the
coordinate planes.
8. If b [(x1)2 - x2] i1+[2(x1)2+3x2] i2 - 2 x1x3 i3,
= evaluate the
surface integral of b over the sphere S with center at the point E
(1, 0, 2 ) and passing through the point F(3, -2, 1).
9. If C is any closed curve, prove that dx = 0. J
c
155
the fingers of the right hand when the thumb points in the direction
of the positive x3 axis.
11. Prove Equations (58.5) and (58.6).
12. A vector field b has continuous first derivatives in a closed
region V. On the bounding surface S of V, b is normal to S. Prove
that
j'VxbdV=O.
v
s v v
X1i1+XJ.2
2
(x1) + (x2)2
-
dX 1 dX 2 Q, 3
':lX
156
CHAPTER VI
TENSOR ANALYSIS
(62.1) z"=J'(z1,z2,<:3),
where the three functions J' (r = 1, 2, 3) are single valued and
differentiable for some range of values of z1, z2 and <;'3. These equa
tions represent a coordinate transformation to new curvilinear coordi
nates z'1, z'2 and z'3• The Jacobian I' of a transformation such as
157
this one was defined in § 49. We express it conveniently here by writing
(62.2)
the ranges for rand s being 1, 2, 3. We assume that I' does not vanish,
whence, as mentioned in § 49, Equations (62.1) may be solved to
yield
(62.3)
3 () ,,
(62.4) dz'' � 3- dz' (r 1, 2, 3).
{} '
= =
z
s=I
ifr = s
(62.6)
= 0 ifr. =F s.
158
We note the identity
Jz' Jz''
(62.7) ll1 ,
r
=
Jz'' Jzt
which is true because its right side is equal to Jz'/Jz', which, because
of the independence of the coordinates z', satisfies the relations
if r = t,
i f r =I= t .
159
() ,,
(63.1) A''=__£_ A'
Jz' '
where the partial derivatives are evaluated at X.
Because of Equation (62 .5) we see that the quantities dZ: are the
components of a contravariant vector. Also, we shall see later on that
when the transformation is from rectangular Cartesian coordinates to
rectangular cartesian coordinates, the components of a vector as defi
ned in§ 6 have the law of transformation (63.1) and are hence also
the components of a contravariant vector.
We shall now define contra variant tensors. Let A" be a set of nine
quantities associated with the point X with curvilinear coordinates
;!, and let A" transform into A'" when the coordinates are transfor
med to z". We then have the definition: a set qf quantities A" is called
the components of a contravariant tensor of the second order if they transform
according to the equation
') fr d Is
Jz' Jz•
Contravariant tensors of higher order are defined analogously. Contra
variant vectors are often called contravariant tensors of the first order.
(}zS
A, =
I
(64.1) '' A,.
Jz
As an example, let us consider a function <l>(z1, z2, z3). We have
whence we note that the three expressions J<I> / Jz' are the components
of a covariant vector.
The definition of a covariant tensor of order two is as follows:
160
a set of nine quantities A,, is said to be the components of a covariant tensor
65. Mixed tensors. Invariants. We can also define tensors whose law
of transformation involves a combination of both contravariant and
covariant properties. Such tensors are called mixed tensors. Thus,
for example, a set of twenry-seven quantities A:1 is said to be the components
of a mixed tensor of the third order with one contravariant suffix and two co
variant suffixes �f they transform according to the equation
(65.1)
(65.2)
161
Now
Jz" ' Jzi ' Jz' S' + Jz' s'
-
- '
Jz'' s" Jz'' 81 + Jz'' ' Jz''
.
When t is given the values 1, 2 and 3, the right side of this relation
reduces respectively to the expressions
Jz1 Jz•
' '
Jz'• Jz'•
Thus we may write
Jz'' Jz'
·
Jz' Jz'·
But this is equal to the left side of Equation (65.2) because of Equation
(62.7).
(66. l)
It is easily proved that the sum of two tensors is a tensor of the same
order and type as the two tensors added.
There are two products of tensors, called the outer product and the
inner product. The outer product of two tensors is defined to be the
set of quantities obtained by multiplication of each component of the
first tensor by each component of the second tensor. Thus, for exam
ple, the outer product of the tensors A; and B;, is the set of 243 quanti
ties c;:. �iven by the relations
(66.2)
162
suffix at the same level on both sides of the equation. When this
convention if followed, it is easily proved that the outer product of
two tensors has the tensor character indicated by the number and
positions of its suffixes. Thus, for ex ample, in Equation (66.2), c;:,, is
a mixed tensor of the fifth order, with two contravariant suffixes and
three covariant suffixes.
We now introduce an operation called contra ction. It consists in
identifying a superscript and a subscript of a tensor. Thus, for example,
if from the tensor A:, we form the set of quantities B, defined by the
relation
(66.3)
C,, = A:S,,.
Of course the inner product of two tensors is not unique. Since outer
multiplication and contraction of tensors both yield tensors, the inner
multiplication of tensors also yields tensors.
163
(67 .1) A;,=0.
''
Let z be any other coordinate system, and let A';, denote the com
ponents of this tensor for the coordinate system z". We must show
that
(67 .2)
By the laws of tensor transformation we have
- 0,
A;,=B;,.
from coordinates z" to z'": then for the overall transformation from
the coordinates z' to z"' the set of quantities is a tensor of this same
order and type. It is easily proved that a tensor of any order or type
is transitive.
164
We shall now prove that A" is a contravariant tensor of the second
order. Since A"X, is a contravariant vector, we have
(} tr
(68.1) '
A'"X s -
- __£
(};::_' A'"X u•
(};::_'V I
(68.2) X u =
(};::_" Xv•
We now rewrite Equation (68.1), replacing the dummy suffix s on
the left side by v, and substituting for X. on the right side from (68.2),
to obtain the relation
or
()
Since X, is arbitrary, so is X'0• Hence the expressions in the brackets
in Equations (68.3) vanish, whence we conclude that A" is a contra
variant tensor of the second order.
Example 2. Let A,, be a set of quantities such that A,,X'X' is
an invariant, where X' is an arbitrary contravariant vector. We shall
now prove that A,, is a covariant tensor of the second order provided
it is symmetric in all coordinate systems, that is, provided we have
for every coordinate system relations of the form A,, = A,,. We
proceed much as in Example 1. Since A,,X'X' is an invarint, we have
Since X' is a contravariant vector we can express the X' and X' in
this equation in terms of X'', obtaining
() ()
A' X''X''
TS -
-
A TS ;::,' ;::,'U X''X'" '
;::, ' ' (};::_'
()
We now replace the dummy suffixes rand son the left side by t and u,
(68.4)
165
where
(}z' (}z' A
A tu - " ,,.
I
(68.5) b1u
J t Jz'
= z'
Since X' is arbitrary, so are the three quantities X". Hence the coef
ficients in Equation (68.6) must vanish, which leads to the equation
(68.7)
From (68.5) we then have
' Jz' az: Jz' azs
A, u A A + '" ' A
t + "' = ;Jz'' ;Jz '• ,, Jz Jz ' ,,.
In the last term on the right side, we interchange the dummy suffixes
r and s, obtaining the relation
Jz J
(68.8) A,tu + A'ut = '' Z:" (Ars + Asr)
Jz' Jz'
•
Since we are given that A,, = A,,, A'1• = A '"" then Equation (68.8)
reduces to the form
Jz' JZ:
A 1u VI ""'ilU A,,,
I
=
uz uz .
166
1
(69.2) (ds)2
b11(dz )2+b (d<:,")2+ b,,(dz')"+b.,dz2dz3
=
22
+b0,dz3dz'+b1 dz'dz2,
2
where b., are six functions of z', z', z'. We now define nirie quantities
g., by the relations
(69.3)
th•• =g•• g,., t b,, g., g,,, th1 g,. g 1.
2 2
= = = = =
70. The conjugate tensor. Let g denote the determinant whose ele
ments are the components of the metric tensor. Then
(70.1) g = jg.,j.
In the expanded form of this determinant, the coefficient of any one
element g., is called the cofactor of g.,. We denote it by the symbol
f:::.". We note in passing that the minor of g,, i s equal to
(-1)'+' A.".
We shall now prove that
167
determinant lg,,j , the subscript r varies over the rows, while the sub
script s varies over the columns. We now consider Equation (70.2)
when s = t = 1. The right side is equal to g. The left side is
11"
(70.4) g" = -·
g
Then Equations (70.2) and (70.3) yield the expressions
(70.5)
gsrX' = Y,,
168
then Y, is an arbitrary covariant vector. Hence we have
Br ,,rt B B�s =
gst Brt) B" = ,,rt,,su B,•.
•S = 5 ts> 0 0
169
of writing no two suffixes in any one vertical line was introduced solely
to ensure this property.
(72.1) A = v'g,)'A'.
From this relation it readily follows that
(72.3) A = v(A1)2+(A2)2+(A')2•
Let X' and Y' be two unit vectors at a point. The angle e between
them is defined by the relation
(73.1) t =f'(s),
the arc lengths of the curve being used as the parameter. Because of
Equation (62.5) we see that dt/ds are the components of a contra
variant vector. We denote it by p', so
, dt
(73.2) p =
ds
-·
170
m n dzm dz" (ds)• 1.
P2 = gmnP 'P = gmn dS Ji = (ds)' =
Thus p' is a unit vector. It is called the unit tangent vector of the
curve C.
A geodesic may be defined as the curve of shortest length joining
two points. In our three-dimensional space, the geodesics are straight
lines. If we consider surfaces, which are of course two-dimensional
spaces, the geodesics are not necessarily straight lines. For example,
in the case of a spherical surface, the geodesics are the great circles,
that is, those circles on the sphere whose centers coincide with the
center of the sphere.
Let X and Y be two points. The distance Lbetween them, measured
along some curve, is given by the line integral
y y y
where
(73.4
According to the Calculus of Variations, Lis an extremum if the path
joining X and Y is such that
(73.5)
d ( CJw\ Jw
ds CJp1- Jz' o.
=
dW .-m n m<-n
gmn o,'P + gmnP ,o,
Jp
' =
dzm
dW =
Jgmn pmpn
·
Jz Jz
Hence equations (73.5) become
(73.6) g ds
rn
dp" + Jgrn m n
P 'P
Jzm
- .!.
2
Jgmn m n
Jz'
P 'P = 0.
Jzm Jz•
=
Substituting for one half of the middle term in Equation (73.6) from
Equation (73. 7), we then obtain the relation
where
" dn"
+ F�.pmp•
(73.11) g grn ds = 0,
where
"ds-ds-ds• .
_
172
Thus we may write the differential equations of a georl.esic in the form
(73.14)
(74.1)
and
(74.2)
But
dz' Jz: dz'P
ds = Jz'P Ts'
d'z' Jz' d'z'P J2Z: dz'P dz'q
ds' = Jz'P ds'- + {}z'P Jz'q Ts dS ·
d'z''
(
dz'' J'z' Jz'' J zm J z"
ds' + Jz' J z'P J z'q + Pm. Jz J.z'P Jz'q Ts dS
=0
·
) dz'P dz' q
173
75. Absolute differentiation. Ap be a covariant vector defn
Let i ed
over a curve C with equations z f'(u), u being a parameter on C.
=
(75.1)
ou du "' ' du·
We shall now prove that oAp/ou is a covariant vector. We have
0A1p A1p F'' '' dz'q.
(75.2 ) - d A -
ou du pq du
Now
(75.3)
dA1p }_ (}z
_
, A
(
du - du (}z'P '
)
Jz dA, o'z' dz'q
A
=
Jz'P du + Jz P (}z'q (h; ,·
'
Also, because of Equation (74.3), the second term on the right side of
Equation (75.2) satisfies the relation
(75.4)
O
OU - •
174
In this case the components A, satisfy first order differential equations,
and can hence be assigned arbitrarily at any one point on C. When
the space is three dimensional and the coordinates are rectangular
cartesian, Equations (75.5) reduce to dA,/du = 0, so that A, are
constant along C.
If A' is a contravariant vector defined on a curve C, its absolute
derivative along C is
(75.6)
dA' dB,
.!:_ (A'B,) B, + A'
du du du
=
dA' d
= B + A' F':!. Bm i'
du ' m du
-
-
( m
dA + pm ' dz:'
A B
m )
du '" du .
Let Am", A�. and Am• be any second order tensors. Their absolute
derivatives along a curve C are
'i5Amn dAmn + pm fm dz9 + " d
pq A du F Amp �
=
(75.7)
'i5u du "" du '
175
M�� = dA':'. + F"' A" dzq _ p A"' dzq
(75.8)
15u du pq du .n nq . p du ,
13A�st _ dA�st + F' A" dzq -F" A' dzq -F" A' dzq.
pq ·St sq .pt t
13u q
.sp
- dS dt dt dt
(75.10)
The proofs of these are left as exercises for the reader (Problem 28 at
the end of this chapter).
The rule for the absolute derivative of a product of two tensors is the
same as for the ordinary derivative of a product. Thus, for example
176
76. Covariant derivatives. Let A, be a covariant vector defined over
some region V in space, and let C be a curve in V. If u is a parameter
on C, then
oAp
ou
_
-
dAP_p
du pqA
dzq
' du -
_ ( JAp _ F, A dz�.
\Jzq pq ' du
)
The expression on the right side is a covariant vector, and dzq /du is
a contravariant vector which may be assigned arbitrarily. Hence the
coefficient of dzq /du is a covariant tensor of the second order. We
denote it by A* and call it the covariant derivative of Ap, so we have
( 77.l ) A rim=
JA
J
�, - FP Ap•
rm
177
Since A,1111 is a covariant tensor of the second order, then
J
(77.2) Arimn = (Ar1m)1n = Jz" (Ar1m)-FrnAq1m-F�nAr1q•
(77.3)
where
J
R,.rmn zm F'rn- Jz" F'rm + FPrn F'pm - FPrm F'pn•
J
(77.4)
J
=
From Equation (77.3) it follows by the tests for tensor character that
R�,.,,.. has the tensor character indicated by its suffixes. It is called the
mixed curvature tensor.
The covariant curvature tensor is
(77.5)
This tensor plays an important role in mathematical physics. For
any coordinate system in three dimensional space, this tensor vanishes,
since it vanishes for rectangular cartesian coordinates. On the other
hand, if for any one coordinate system on a surface this tensor vanishes,
then there exists for this surface a curvilinear coordinate system such
that
g,, = 1 if r = s
= 0 if r=s.
a,,x',+a',,
'
(78.1) x , = a,,x,+ a,, x, =
178
where a,, are the constants considered in§ 47 satisfying the orthogon
ality conditions
(78.2)
Also, a, and a', are constants such that a',= -a,, a,. Just as in §47,
we have the relations
Jx', Jx,
(78.3) a
Jx' ,, Jx',
=
= ·
I= \�::,! = Ja,,J.
(78.4)
sol=±l.
gmn = amn' g =
1' gmn = 8m.,
so that the raising or lowering of a suffix does not change the values
of the components.
Because of this theorem, when dealing with cartesian tensors we
do not need both superscripts and subscripts, so subscripts will be used
exclusively.
In§ 6 we introduced the orthogonal projections of a vector on rec
tangular cartesian coordinate axes, calling these projections the com
ponents of a vector. Throughout the rest of this book we shall refer
to these as the physical components of a vector for rectangular cartesian
coordinates.
Theorem 2. The physical components of a vector for rectangular
cartesian coordinates constitute a cartesian tensor of the first order.
179
Proof. The nine constants a,, in Equations (78. l) are the cosines
of the angles between the axes of two rectangular cartesian coordinate
systems. Hence, if b is a vector with physical components b, and b',
for these two systems of coordinates, then just as in § 47 we have
b,' = asr b,
= �x�
ux,
b,.
This is the Jaw of transformation ofa cartesian tensor of the first order,
so the proof is complete.
When the coordinates are rectangular cartesian,. the Christoffel
symbols vanish, and so the absolute derivative becomes the ordinary
derivative and the covariant derivative becomes the partial derivative.
For example, 'ilb,/au becomes db,/du, and b,1, reduces to Jbrf Jx,.
Thus we conclude that the directional derivatives of the physical
components of a vector for rectangular cartesian coordinates constitute
a cartesian tensor of the first order, and the nine partial derivatives
of the physical components of a vector constitute a cartesian tensor
of the second order.
79. Oriented cartesian tensors. Equation (78 .4) shows that the Jaco
bian I of a transformation from one set of rectangular cartesian coor
dinates to another is equal to· either plus one or minus one. The former
case arises when the transformation is between coordinates whose
axes have the same orientation (both right-handed or both left
handed), and the latter case arises when the orientations are opposite.
A set of quantities is said to constitute an oriented cartesian tensor if
it is a cartesian tensor when I= 1 and is not a cartesian tensor when
I= -1.
In two-dimensional problems, suffixes have the range 1, 2. For
such problems we introduce a permutation �bol crs defined as follows:
In the three dimensional case, the permutation symbol c,,1 has the defi
nition
180
erst = 0 if two suffixes are equal
(79.2) = 1 if (rst) is an even permutation of ( 123)
= -1 if (rst) is an oud permutation of ( 123) ,
a single permutation of (rst) being an interchange of any two of r, s
and t, and an even or odd permutation meaning an even or odd
number of single permutations. Thus, for example,
C123 = C 312 = 1, -1 , Cn3 = Q
C321 = •
In Theorem 4 of the next section we shall see that both of these per
mutation symbols are oriented cartesian tensors.
We can now express in tensor notation many of the formulas and
equations of the earlier chapters of this book. Thus, the scalar and
vector products of two vectors a and b are respectively
a, b,, erst a, b,.
We note that this vector product is an oriented cartesian tensor. The
scalar triple product a· (b X c ) is
erst a, b, c1•
The expressions \Jf, \J b and \J
· Xb become
Jj Jb, Jbt
Jx, ' Jx,' Crst Jx,'
respectively . The differentiation formulas (48.1)-(48.11) may also be
expressed in this notation; when this is done (Problem 36 at the end
of this chapter), some of the formulas become trivial, and the truth of
others follows at once from the identity
( 79 .3) Crmn Crst = ams ant - amt a.,,
the proof of which is left to the reader (Problem 34 at the end of this
chapter).
() ,, A' JW
A''= 3-- '
Jz'
W being an integer. Relative contravariant tensors of higher order
are defined analogously, as are relative covariant tensors, relative
mixed tensors, and relative invariants. For example, A�,, is a relative
mixed tensor of weight W if it has the law of transformation
Jz' Jz•
(80.1) g , = 1.g,,,. I =
J ''
Jz Jz'
' g,.
I
= I ;;:.1 I::�./
. ,gvw'
·
l'g,
which completes the proof.
=
182
Let us now introduce a covariant permutation symbol c,,, and a
"'
contravariant permutation symbol c , both defined numerically
just as in Equation (79.2).
Let us denote the right side ofthis equation by 1-1 <l>,,1• Then
(80.3)
But the expression on the right side here is just the expansion of the
determinant /, so <1>123 =I. In a similar manner, we see that each
component of <I>,,, is equal to a determinant . If the suffixes on any
one component are subjected to a single permutation, two rows or
columns of the corresponding determinant are interchanged, and so
the sign of the component is changed. We thus have the results:
(i) <I>,,, = 0 if two suffixes are equal, (ii) <l>,,1 = I if (rst) is an even
permutation of (123), (iii)<1>,,1 = - I if (rst) is an odd permutation of
·
183
positive. We can then construct the absolute permutation symbols as
follows:
rsl c"'
(80.4) "I),,, =
-v'g c,,,, =
"1) v-g·
Because of Theorems I, 3 and 4 above, we see that 'I),,, is an absolute
covariant tensor, while 'IJ"' is an absolute contravariant tensor.
Let A::: denote a general relative tensor ofweight W. Then giwA:::
is an absolute tensor. The covariant derivative Of A::: is defined to be
... !W ( -rW ···)
A ... l A
_
(80.5)
r -
g g 1r · . . .
We note that A:::1, has the same weight as A:::. The absolute deri
vatives of relative tensors are defined analogously. It ea� be proved
that both the covariant and absolute derivatives of relative tensors
obey the same product rule as do ordinary derivatives.
oz'
It is easily proved that B, g,,B'. = The quantities B' and B, describe
the vector bin a certain manner, and have the tensor character indi
cated by their suffixes.
In § 72 of the present chapter we defined abstractly the magnitude
of a contravariant vector. According to this definition, the magnitude
of B' is
B Vg,,B' B'.
=
Since the right side of this equation is an invariant, and since �,, is
184
Thus the abstract definition of magnitude of a vector given in § 72
agrees with the definition of magnitude given in § 1. In a similar
fashion, the abstract definition given in § 72 for the angle between
two vectors agrees with our physical notions of angle.
Let us consider a unit vector having components ')( for a curvili
i
near coordinate system . . The physical component of B' in the direc
tion of )..' is defined to be the invariant
B'
cos a g,, )..'.
T
=
(81.4)
Thus
B ,1> = g,,B'·\:> = B').,,>, = B1h,.
In a similar dashion we get B(•) and Beal, so that the physical compo- ·
185
(81.5)
T(rs) = T"'")...(r)m)...(s)n'
)...f,; being the three unit vectors in the directions of the parametric
lines of the coordinates. When the coordinates ;! are orthogonal, then
)...<;l is as given in Equation (81.3), and there are two similar relations
for )...<;) and )...(;l. We then get for Trrs; the expressions
186
� �
(821)
. v, a =
= dt' , dt.
If f, is the force acting bri the particle and m is its mass, then by
Newton's second law we have
(82.2) ma, = f, .
Let ;! be curvilinear coordinates of the particle, V', A' and F' being
contravariant component of velocity, acceleration and force for this
coordinate system. We have by definition
(82.4) A' =
l'i V'
St
= dV' + F'
dt
mn
dt
"
ym dz .
We could define F' in terms off,, but it is easier to use the relation
To check these we note that they are tensor equations, and are true
when the coordinates are rectangular cartesian since in this case they
reduce to (82.2). Hence they are true for all coordinate systems.
The mathematical theory of elasticity. Using rectangular cartesian
coordinates x,, we have the displacement u,, the strain components
e,, and the stress components T,,. For the determination of these,
(-. )
we have the equations
Ju, Ju,
- + -
1
(82.7) e,, _
-2 '
dx, Jx,
__!!_!_!!!__ + J'es J•e,m + J•e,.
Jx,Jxm ,
m =
(82.8)
dX,JXm Jx,dXn Jx,Jxn
187
(82.9)
JT,, , J•u,
(82.IO) X=
Jx, + p ,
Jt
together with certain boundary conditions. In the above, k,smn are
elastic constants, X, is the external force per unit volume, p i s the
density and t is the time. If the body is at rest, the right side of (82.10)
vanishes. If the body is isotropic, that is, it has no preferred directions
elastically, Equation (82.9) reduces to
dp d
(82.19) Jt + dx,
(pv,) = O,
dV, �Vr , _l Jp
(82.20) + V = X,
dt dx, p dX, ,
where p is the density, t is the time, v, is the velocity, X, is the external
force per unit volume, and p is the pressure. We convert these to
curvilinear coordinates in a manner analogous to that used for elas
ticity, obtaining the equations
(82.21) �� + (pv')1r = 0,
(82.22)
Problems
1. Prove that
(A,,+Asr)lt = 2A,,z'z'.
2. If A = A,,z'z', where A,, are constants, prove that
JA J'A
Jz = (A,, + Asr)z', Jz ;}z'
: = Ars + Asr•
189
{J'z'' Jz'' Jz'' Jz'w
d'z"
J�Jz1 = - Jz" J� Jz1 Jz'' Jz'w
6. Let b, denote the covariant components of a vector for rectangu
lar cartesian coordinates. Find the covariant components of this
vector for cylindrical coordinates r, 6, z in terms of b, and r , 6, z.
z
8. Prove that the sum of two tensors A;1 and B;, is a tensor.
9. Prove that the outer product of two tensors A; and B;, is a tensor.
10. If A;, is a tensor, prove that A�, is a tensor.
11. In three-dimensional space, how many different expressions are
represented by A:B;qCi'? When each such expression is written out
explicitly, how many terms does it contain?
12. Prove that the tensor A;' is transitive.
13. Let A; be a set of nine quantities such that A;x: is an invariant,
where x; is an arbitrary mixed tensor of the second order. Prove that
A; is a mixed tensor of the second order.
14. Let A,,1 be a set of quantities such that A,,1X1 is a covariant
tensor of the second order, where X1 is an arbitrary contravariant
vector. Establish the tensor character of A,,1•
15. If g,, = 0 for # r s, prove that
1 ' g•• = 1 ' 1
gll = -
t• = - ' -
gu g,, goa
g .. = tl =g" =0.
16. Find the components of g" for cylindrical coordinates r, 6, z.
17. Find the components of g" for spherical polar coordinates
r, 6, q>.
1
18. Let z and z' be plane oblique cartesian coordinates whose
axes have an angle � between them. For these coordinates find g,,
and g'.
19. Prove that g,,g" 3. =
Fsrs J d
vg {) z' g.
• 1_
= v
191
32. If f is an invariant, use the results of Problem 31 to evaluate
\J) in terms of cylindrical coordinates r, 6, z, comparing the result
aX(bXc) = b(a·C)-'c(a·b),
and prove it by use of Equation (79.3).
36. Express the eleven equations (48.1)-(48.11) in cartesian tensor
form, a:nd then verify them.
37. If A" is an absolute tensor, prove that the determinant IA"I is a
tensor.
40. If A' is a relative tensor of weight one, prove that
JA'
A'.1r ''
Jz
=
192
dr
( d6 d<p) (dr 2d6 d<p
' '
dt dt dt ' dt
'r
2
s
dt' r m
• 2
6Tt) '
(�;, r:�, r sin 6 �;) ·
JE JH
\1 x H= k-, \1 x E =-µ-,
dt dt
'V·E = 0, 'V·H=O,
where His the magnetic intensity, k is the dielectric constant, Eis the
electric intensity, 1i. is the permeability and t is the time. Express these
equations in tensor form for general curvilinear coordinates i.
193