PURE AND APPLIED MATHEMATICS

A Program of Monographs, Textbooks, and Lecture Notes

Vector and Tensor Analysis



Second Edition, Revised and Expanded




Eutiquio C. Young






Department of Mathematics
Florida State University
Tallahassee, Florida








Library of Congress Cataloging-in-Publication Data




Young, Eutiquio C.
Vector and tensor analysis / Eutiquio C. Young -- 2nd ed., rev.
and expanded.
p. cm. -- (Monographs and textbooks in pure and applied
mathematics; 172)
Includes bibliographical references and index.
ISBN 0-8247-8789-7 (alk. paper)
1. Vector analysis. 2. Calculus of tensors. I. Title.
II. Series.
QA433.Y67 1992
515'.63--dc20 92-33741
CIP





This book is printed on acid-free paper.




Copyright 1993 by MARCEL DEKKER, INC. All Rights Reserved.






Neither this book nor any part may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopying, microfilming, and recording,
or by any information storage and retrieval system, without permission in writing from
the publisher.






MARCEL DEKKER, INC.
270 Madison Avenue, New York, New York 10016






Current printing (last digit):
10 9 8 7 6 5 4 3 2 1







PRINTED IN THE UNITED STATES OF AMERICA

Preface to the Second Edition




In this new edition we have tried to maintain the objective of the first edition, namely, to
acquaint students with the fundamental concepts of vector and tensor analysis together
with some of their physical applications and geometrical interpretations, and to enable
students to attain some degree of proficiency in the manipulation and application of the
mechanics and techniques of the subject. We have tried to retain the qualities and
features of the previous edition, placing great emphasis on intuitive understanding and
development of basic techniques and computational skills.






In this edition each chapter has been rewritten and certain chapters have been
reorganized. For example, in Chapter 3 the section on directional derivatives of vector
fields has been deleted, the section on transformation of rectangular cartesian coordinate
systems, together with the invariance of the gradient, divergence and the curl has been
incorporated in the discussion of tensors. In Chapter 4 the section on test for
independence of path has been combined with the section on path independence. In each
chapter we have expanded discussions and provided more examples and figures to
demonstrate computational techniques as well as to help clarify concepts. Whenever it is
helpful we have introduced subtitles in each section to alert students to discussion of
new topics. Throughout the book, we have written statements of definitions and
theorems in boldface letters for easy identification.






The author will appreciate receiving information about any errors or suggestions for the
improvement of this book. The author also wishes to thank Miss Deirdre Griese,
Production Editor, and her staff for assistance rendered in the revision of this book.





EUTIQUIO C. YOUNG

Preface to the First Edition




This book is intended for an introductory course in vector and tensor analysis. In writing
the book, the author's objective has been to acquaint the students with the various
fundamental concepts of vector and tensor analysis together with some of their
corresponding physical and geometric interpretations, as well as to enable the students to
attain some degree of proficiency in the manipulation and application of the mechanics
and techniques of the subject.






Throughout the book, we place great emphasis on intuitive understanding as well as
geometric and physical illustrations. To help achieve this end, we have included a great
number of examples drawn from the physical sciences, such as mechanics, fluid
dynamics, and electromagnetic theory, although prior knowledge of these subjects is not
assumed. We stress the development of basic techniques and computational skills and
deliberately de-emphasize highly complex proofs. Teaching experience at this level
suggests that highly technical proofs of theorems are difficult for students and serve little
purpose toward understanding the significance and implications of the theorems. Thus
we have presented the classical integral theorems of Green, Gauss, and Stokes only
intuitively and in the simplest geometric setting. At the end of practically every section,
there are exercises of varying degree of difficulty to test students' comprehension of the
subject matter presented and to make the students proficient in the basic computation
and techniques of the subject.







The book contains more than enough material for a one-year or two-quarter course at the
junior or senior level or even at the beginning graduate level for physical sciences
majors. Omitting Secs. 3.9 through 3.12, Chaps. 1 through 4 can serve as material for a
one-semester course in vector analysis, or for a one-quarter course with further deletion
of topics depending on the interest of the class. Preceded by Secs. 3.9 and 3.11, the
material of Chaps. 5 and 6 can then be used for a second-semester or a one-quarter
course in tensor analysis.

As a prerequisite for a course based on this book, the students must be familiar with the
usual topics covered in a traditional elementary calculus course. Specifically, the students
must know the basic rules of differentiation and integration, such as the chain rule,
integration by parts, and iterated integration of multiple integrals. Although a knowledge
of matrix algebra would be helpful, this is not an essential prerequisite. The book
requires only the bare rudiments of this subject, and they are summarized in the text.




The author wishes to thank his colleagues Professor Steven L. Blumsack, Wolfgang
Heil, David Lovelady, and Kenneth P. Yanosko for reviewing portions of the manuscript
and offering valuable comments and suggestions, and Professors Chiu Yeung Chan and
Christopher K. W. Tam for testing the material on tensors in their classes during the
developmental stage of the book. Last but not least, the author acknowledges with
gratitude the assistance rendered by the production and editorial department of the
publisher.





EUTIQUIO C. YOUNG


Contents




Preface to the Second Edition






v






Preface to the First Edition






vii





Chapter 1
Vector Algebra







1







1.1 Introduction 1.2 Definition of a Vector 1.3 Geometric
Representation of a Vector 1.4 Addition and Scalar Multiplication
1.5 Some Applications in Geometry 1.6 Scalar Product 1.7 Vector
Product 1.8 Lines and Planes in Space 1.9 Scalar and Vector Triple
Products








Chapter 2
Differential Calculus of Vector Functions of One Variable





75





2.1 Vector Functions of a Real Variable 2.2 Algebra of Vector
Functions 2.3 Limit, Continuity, and Derivatives 2.4 Space Curves
and Tangent Vectors 2.5 Arc Length as a Parameter 2.6 Simple
Geometry of Curves 2.7 Torsion and Frenet-Serret Formulas 2.8
Applications to Curvilinear Motions 2.9 Curvilinear Motion in Polar
Coordinates 2.10 Cylindrical and Spherical Coordinates








Chapter 3
Differential Calculus of Scalar and Vector Fields





147





3.1 Scalar and Vector Fields 3.2 Algebra of Vector Fields 3.3
Directional Derivative of a Scalar Field 3.4 Gradient of a Scalar
Field 3.5 Divergence of a







Vector Field 3.6 Curl of a Vector Field 3.7 Other Properties of the
Divergence and the Curl 3.8 Curvilinear Coordinate Systems 3.9
Gradient, Divergence, and Curl in Orthogonal Curvilinear Coordinate
Systems







Chapter 4
Integral Calculus of Scalar and Vector Fields





207





4.1 Line Integrals of Scalar Fields 4.2 Line Integrals of Vector Fields
4.3 Properties of Line Integrals 4.4 Line Integrals Independent of
Path 4.5 Green's Theorem in the Plane 4.6 Parametric Representation
of Surfaces 4.7 Surface Area 4.8 Surface Integrals 4.9 The
Divergence Theorem 4.10 Applications of the Divergence Theorem
4.11 Stokes' Theorem 4.12 Some Applications of Stokes' Theorem








Chapter 5
Tensors in Rectangular Cartesian Coordinate Systems





307





5.1 Introduction 5.2 Notation and Summation Convention 5.3
Transformations of Rectangular Cartesian Coordinate Systems 5.4
Transformation Law for Vectors 5.5 Cartesian Tensors 5.6 Stress
Tensor 5.7 Algebra of Cartesian Tensors 5.8 Principal Axes of
Second Order Tensors 5.9 Differentiation of Cartesian Tensor Fields
5.10 Strain Tensor








Chapter 6
Tensors in General Coordinates







373




6.1 Oblique Cartesian Coordinates 6.2 Reciprocal Basis;
Transformations of Oblique Coordinate Systems 6.3 Tensors in
Oblique Cartesian Coordinate Systems 6.4 Algebra of Tensors in
Oblique Coordinates 6.5 The Metric Tensor 6.6 Transformations of



Curvilinear Coordinates 6.7 General Tensors 6.8 Covariant
Derivative of a Vector 6.9 Transformation of Christoffel Symbols
6.10 Covariant Derivative of Tensors 6.11 Gradient, Divergence,
Laplacian, and Curl in General Coordinates 6.12 Equations of
Motion of a Particle






Solutions to Selected Problems




469




Index






495



VECTOR ALGEBRA
1.1 INTRODUCTION
In the study of physics and engineering, we encounter many
important quantities which can be described by the specification of
their magnitude alone in terms of some appropriate units. For ex-
ample, the volume of a cube can be described by the number of cubic
inches or cubic centimeters, and the temperature at a particular time
of a day can be described by giving the number of degrees on a Fah-
renheit scale or a Centigrade scale. Such quantities characterized by
the fact that they have magnitude only are called scalar quantities.
Scalar quantities are represented by real numbers, and they are also
called scalars. On the other hand, there are many other physical
quantities such as displacement, force, velocity, and acceleration
which cannot be described by the specification of their magnitude
alone. These quantities possess not only magnitude but also direc-
tion so that a complete description of any such quantity must specify
these two pieces of information. Thus when a weatherman reports
the wind velocity on a particular day, he specifies not only the speed
of the wind (magnitude of the wind velocity), but also the direction
from which the wind is blowing. Such quantities characterized by
having magnitude and direction are called vector quantities.
Just as we use real numbers or scalars to represent and mani-
pulate scalar quantities, so we use the mathematical entities called
vectors to represent and manipulate vector quantities. Thus, in a
sense, vectors can be thought of as generalized numbers. The study
of the representation of vectors, the algebra and calculus of vectors,
together with some of their various applications constitute the
subject matter of vector analysis.
Chapter 1
Scalars and vectors are hardly sufficient to treat the class of
quantities that are of interest in physics, engineering, and applied
mathematics. In fact, there are quantities of a more complicated
structure whose description requires more than knowledge of a
magnitude and a direction. For example, to describe a quantity such
as stress, we need to give a force (described by a vector) and a surface
on which the force acts. Such a quantity can be described and repre-
sented only by the more sophisticated mathematical entity called
tensor. As we shall see later, scalars and vectors are actually special
cases of tensors. This is in accord with an important characteristic of
mathematics in which concepts of more general nature do incorpo-
rate as particular cases those from which the concepts originated in
the first place.
For the most part in this book, we shall study vectors and ten-
sors in the familiar two- and three-dimensional Euclidean spaces.
Although in many instances the concepts and results we obtain can
be immediately extended to higher dimensional spaces, we shall do
so only in few cases.
In order to distinguish vectors from scalars, we will use bold-
faced letters, A, B, . . , a, b, . . , to denote vectors, and lower case
letters, A, B, . . , a, b, . . , to denote scalars or real numbers. Tensors
will be represented by their so called components.
1.2 DEFINITION OF A VECTOR
A vector may be defined in essentially three different ways:
geometrically, analytically, and axiomatically. The geometric defini-
tion of a vector makes use of the notion of a directed line segment
or an arrow. A line segment determined by two points P and Q is
said to be directed if one of the points, say is designated as the
initial point and the other, Q, the terminal point. The directed line
segment so obtained is then denoted by and it is represented
graphically by drawing an arrow from to Q (Fig. 1.1). We denote
the length of by Two directed line segments PQ and
are said to be equal, and we write = RS , i f they have the same
length and the same direction. Graphically, this means that the
arrows representing and RS are parallel, have the same length
and are pointing in the same direction.
Geometrically, then, a vector is defined as the collection of all
directed line segments or arrows having the same length and direc-
tion. (Such a collection is also called an equivalence class of directed
Vector Algebra
Fig. 1.1 Directed line segments.
line segments, and any two members of the class are said to be
equivalent.) The common length of the arrows represents the
magnitude of the vector, and the arrowhead indicates the direction
of the vector. Further, for the purpose of representation, any one of
the arrows in the collection can be used to represent the vector. For
example, the collection of arrows shown in Fig. 1.1 defines a vector
A, and anyone of the arrows can be used to represent the vector A.
Under this definition, the algebraic operations on vectors are intro-
duced and studied geometrically, making maximum use of our
geometrical intuition. One advantage of this approach is that it does
not require the introduction of any particular coordinate system
and, hence, one can concentrate on the intrinsic vector relations.
However, this approach is rather cumbersome and inefficient for
computational purposes.
In the analytic approach, a vector in three-dimensional space is
defined as an ordered triple of real numbers relative to a
given coordinate system. The real numbers , are called the
components of the vector. As we shall see later, once a coordinate
system is introduced, these components arise naturally from the
geometric description of a vector. Likewise, in the two-dimensional
space, the xy-plane, a vector is defined as an ordered pair of real
numbers Algebraic operations on vectors are then per-
Chapter 1
formed through the components of the vectors. By far this approach
is the most convenient for theoretical and computational considera-
tions, and it is the approach that we will adopt in this book.
However, it should be kept in mind that the components of a vector
are dependent on the coordinate system used. This means that
whenever we change the coordinate system in our discussion, the
components of the vector will change, although the vector itself
remains the same. For example, the components of a vector in the
cylindrical coordinate system will be different from its components
in the spherical coordinate system.
Lastly, the axiomatic point of view treats a vector simply as an
undefined entity of an abstract algebraic system called a linear vector
space. In such a system, the vectors are required to satisfy certain
axioms with respect to two algebraic operations that are undefined
concepts. As we shall see later, the sets of axioms for a linear vector
space are precisely the properties satisfied by vectors with respect to
the vector operations of addition and multiplication by scalars as
developed by either the geometric or the analytic approach.
In this chapter, we shall develop the algebra of vectors on the
basis of the analytic definition of a vector. We shall use directed line
segments or arrows to represent vectors geometrically and to give
goemetric interpretations of our results. Accordingly, we assume a
coordinate system in our three-dimensional space. As is customary,
we assume a right-handed rectangular cartesian coordinate system
(x, y, z). The student may recall that such a coordinate system con-
sists of three straight lines that are perpendicular to each other at a
common point called the ori gn The lines are designated
Fig. 1.2 Rectangular cartesian coordinate system.
Vector Algebra 5
as the x-, y-, and z-coordinate axes, and a definite direction on each
axis is chosen as the positive direction. The coordinate system is
then said to be right-handed if the so-called 'right-hand rule' holds;
that is, i f the index finger and the center finger of the right hand are
made to point along the positive x- and y-axes, respectively, the
thumb points along the positive z-axis. If we delete the z-axis, we
obtain the corresponding cartesian coordinate system in two dimen-
sional space, known as the xy-plane. We now define a three dimen-
sional vector with respect to a right-handed cartesian coordinate
system.
Definition 1. A three-dimensional vector A is an ordered
triple of real numbers , written as
is called the first component the second component, and
the third component of the vector.
A vector whose components are all zero is called the zero
vector and is denoted by that is, = 0, 01. The negative of a
vector A, denoted by - A, is defined as the vector whose compo-
nents are the negative of the components of A. That is, if A =
then - A = [- , -a
3
].
Definition 2. The magnitude of a vector A = [a,, ,
denoted by A is the real number defined by
It is clear that and that = if and only if A = If
= I, we call the vector A a unit vector. In the sequel, unless
stated otherwise, we shall assume our vectors to be non-zero.
1.3 GEOMETRIC REPRESENTATION OF A VECTOR
Let A = be a given vector and let be the point
with the coordinates , Let us examine the directed line
segment or arrow that is drawn from the origin to the point as
6 Chapter 1
shown in Fig. 1.3. From the figure we see that the length of the
arrow is equal to + which is the magnitude of the
vector A as defined in (1.1). The direction of the arrow OP can be
described analytically by the three numbers cos a, cos and cos y,
called the direction cosines of the directed line segment. The angles
a , and known as the direction angles, are the angles made by
the directed line segment the positive x-, y-, and z-coordi-
nate axes, respectively. Now, from Fig. 1.3, it is clear that
= - - - , = y =
I
Thus the direction cosines of proportional to the compo-
nents of the vector. In this manner, the length and direction of the
directed line segment OP correspond to the mapi t ude and direc-
tion of the vector A. From this discussion it follows that every
three dimensional vector is associated with a point in space. Con-
versely, every point in space determines a three dimensional vector
whose components are the coordinates of the point. The same is
true of two dimensional vectors and the points in the xy-plane.
Notice that by (1.1) the direction cosines (1.2) satisfy the impor-
tant relation
Fig. 1.3 Geometric representation of a vector.
Vector Algebra
Thus the vector = is a unit vector. This
implies that every nonzero vector can be converted into a unit
vector. The process of converting a nonzero vector into a unit
vector is known as the normalization process. A vector that has
been converted into a unit vector is said to have been normalized.
We should point out that the representation of a vector by a
directed line segment does not depend on the initial point from
which the directed line segment is drawn. In other words, a vector
can also be represented by an arrow drawn from an arbitrary point in
space so long as the arrow has the same length or magnitude and
the same direction as the vector. For example, consider the vector
A = a,] which is represented geometrically by the arrow OP
in Fig. 1.3. Let Q be an arbitrary point with the coordinates (x,,
) and let R be the point with the coordinates + + +
see Fig. 1.4. Then the directed line segment drawn from the
point Q to the point R also represents geometrically the vector A.
Fig. 1.4 Representation of a vector from an arbitrary point.
Chapter 1
This follows immediately from the fact that the directed line seg-
ment 1.4) and the directed line segment 1.3) have
the same length and the same direction cosines.
From the above discussion, it follows that if a directed line
segment PQ represents a vector A, where PQ is drawn from the
point P to the point Q whose coordinates are , , ) and
respectively, then the components of the vector A are given
Similarly, in the xy-plane (a two dimensional space), a vector
A = can be represented geometrically by a directed line seg-
ment drawn from the origin to the point P with the coordinates
as shown in Fig. 1.5. The direction of the vector is uniquely
defined by the angle = It is clear that
where = +
Fig. 1.5 Vector in the plane.
Vector Algebra 9
Example 1. Find the components and the magnitude of the
vector A that is represented by the directed line segment where
P: (3, and Q: (-1, 1, 4).
Solution: Let denote the components of A. By (1.4)
we have
a , = - 1 - 3 = - 4 , a 2 =1 - ( - 2 ) =3 , a 3 = 4 - 2 = 2
Hence, by (1.1)
Example 2. Find the direction cosine of a directed line seg-
ment which represents the vector A = [2, -1,2 1.
Solution: The magnitude of the vector is equal to
Hence, by (1.2), the direction cosines of a directed line segment
representing the vector are
Example 3. If A = [-I, 31, what is the length and the direction
of an arrow which represents the vector?
Solution: The length of an arrow representing the vector is
equal to the magnitude of the vector , which is given by
The direction is determined by the angle 8 = arctan (31-1) = 108. 43~
measured counterclockwise from the positive x-axis.
Example 4. A vector A is represented by the directed line
segment PQ , where P: (2, -1, 3) and Q: (-1, -2, 4). Find the compo-
nents and the magnitude of the vector, the direction cosines of the
line segment, and the unit vector in the same direction as A.
Solution: Let A = [al , a2, a, 1. By (1.4) we have
al = -1 - 2 = -3, a2= -2 - (-1) = -1, a 3 = 4 - 3 = 1
Chapter 1
Hence, I A I = [(-312 + + l2l1I2 = dl 1 and, by (1.2),
cos a = -3 /dl 1 , cos (3 = -1 / dl 1, cos y = 1 /dl 1
Thus u = [- 3/d11, -1 /d11, 1 /dl11 is the unit vector in the same
direction as A.
1.1 EXERCISES
In each of Problems 1 through 6, find the components and the
magnitude of the vector that is represented by the directed line
segment PQ whose initial point P and terminal point Q are given.
In each of Problems 7 through 13, find the coordinates of the
initial point P or the the terminal point Q of PQ which represents
the given vector.
A = [-2,3], P : (1, -2).
A = [4,-51, Q: (-1,l).
A = [1/2,2/3], P: (1/2,3).
A = [2,1,4 1, P : (1, -1, l).
A = [-l,3, -21, P : (2, -1, -3).
A = [2, -3, -41, Q : (3, -1,4).
A = [3,1, -21, Q: (-2,1,3).
If A = [-I, -2, 21, find the direction cosines of an arrow represen-
ting the vector.
Repeat Problem 14 for A = [I, - 3,2].
If a plane vector A is represented by an arrow whose length is
5 units and whose direction is defined by H = 150 O, find the
components of the vector.
A wind is blowing in the southwesterly direction at 15 mph.
Find the x- and the y-component of the wind velocity.
Vector Algebra 11
18. Show that the direction cosines defined by equation (7.2) satisfy
the relation
19. If A = [2, -1, -21, find the unit vector that points in the same
direction as A; in the opposite direction to A. How many
vectors are there in either direction?
20. An aircraft takes off from an airport at the air speed of
100 mph at an angle of inclination of 15 deg. How high from
the ground and how far from the airport (on the ground) is the
aircraft 10 min after takeoff?
1.4 ADDITION AND SCALAR MULTIPLICATION
We now begin the study of the various algebraic operations
which can be performed on vectors. First, we introduce the notion
of equality between two vectors.
Fig. 1.6 Equality of vectors.
Chapter 1
Definition 1. The vectors A = [al, a2 , a3 ] and B = PI, b2 , b3]
are said to be equal, written A = B, if and only if
a l =b l , a2=b2, a3=b3 (1 ~ 6 )
Clearly, two vectors that are equal can be represented geometri-
cally either by a single directed line segment or by two directed line
segments originating from two distinct points and having the same
length and the same direction (see Fig. 1.6).
We now introduce the concept of vector addition.
Definition 2. The sum of two vectors A = [al, a2, a?] and B =
[bl , b2 , b3], denoted by A + B, is the vector
It follows from the definition (1.7) and the corresponding
properties of real numbers that addition of vectors satisfies the com-
mutative law
and the associative law
for any vectors A , B , C . Further, for every vector A , we have
and
A + ( - A) =O
A student who has been exposed to the concept of a group in
modern algebra will undoubtedly recognize that the properties (1.8)
through (1.11) are precisely the axioms satisfied by the elements of a
commutative group with respect to the binary operation "+". For
this reason we say that the set of vectors forms a commutative
group with respect to the operation of vector addition. This con-
stitutes part of the characterization of a linear vector space.
Vector Algebra 13
Using the idea of the negative of a vector, we can introduce the
operation of subtraction or the difference between two vectors. We
define the difference A - B of two vectors A and B to be the sum of
the vectors A and - B. That is, if A = [al, a2, a3] and B = [bl , b2 , b,],
t hen
A- B = A + (-B) = [a1-bl , a2-b2, a, -b3]
(1.12)
The addition and subtraction of vectors defined above can be
demonstrated geometrically using directed line segments or arrows.
For convenience we show this for vectors in the xy-plane. Refer-
ring to Fig. 1.7, let the arrows OP and OQrepresent the vectors A =
[al, a,] and B = [bl, b,], respectively, as determined by the points P
and Q. Now consider the parallelogram OPRQ with OP and OQ as
two of its sides. Since opposite sides of a parallelogram are equal in
length and are parallel, it follows that OP = QR and OQ= PR..
Hence QR also represents the vector A and PR also represents the
vector B. Moreover, since OA = al , 08 = bl , we see that OC = OA
+ OB = al + bl. Likewise, since OD = b2, OE = a,, we have OF = OD
0 B A C
Fig. 1.7 Parallelogram law of addition of vectors.
Chapter 1
+ OE = b2 + a2 Hence, the point R has the coordinates (al + bl , a2 -t
b2) and, therefore, the arrow ORrepresents the vector sum A + B.
Because OR is the diagonal of the parallelogram formed by the vec-
tors A and B, the rule for the addition of vectors is sometimes
called the yarnllelogranz law of addition.
The subtraction of two vectors is illustrated geometrically in
Fig. 1.8. Notice that in the parallelogram formed by the geometric
vectors, one of the diagonals represents the sum of the vectors,
while the other represents the difference.
Next we introduce the operation of multiplication of vectors
by real numbers or scalars.
Definition 3. Let nl be a real number and let A = [a, , a2, a3 ]
be a vector. The product of nz and A, denoted by nz A, is the vector
defined by
nz A = [nz a, , nz a2, nl a3] (1.1 3)
The vector m A is sometimes called a scalar multiple of the
vector A. This term should not be confused with the scalar product
of two vectors, which is an entirely different operation (Sec. 1.6).
Geometrically, if ZQ represents a vector A and B = rn A, then
Fig. 1.8 Difference of two vectors.
Vector Algebra
Fig. 1.9 Scalar multiples of a vector.
an arrow representing the vector B will have length equal to Im I
times the length of PQ; it will point in the same direction as PQ
if nz is positive and in the opposite direction i f nz is negative, as
shown in Fig. 1.9. From the definition (1.13) and the known proper-
ties of real numbers, it is easily verified that for any vectors A, B
and for any scalars nz , n , we have
nz (TI A) = (nz 12 )A (1.14)
(nr +t.z)A = nzA + zzA (1 .I 6)
These properties together with those listed in (1.8) through (1.11)
constitute the axioms that are satisfied by the elements of a linear
vector space with respect to the two binary operations of "addition"
and "multiplication by scalars". An element of such a space is then
called a vector.
Having introduced the concepts of addition and scalar multi-
plication of vectors, we now observe that every vector A = [al, a2,
a,] can be written as
Chapter 1
Since each of the vectors [I, 0, 01, [O, 1, 01 and [O, 0, 11 has magnitude
equal to 1, they are unit vectors. We denote these unit vectors by i ,
j , k , respectively; thus,
Henceforth we write every vector A = [al, a2, a3] in the form
and call (1.19) an analytic representation of the vector A.
The set of vectors (i , j , k) is called the natural basis for the
three dimensional vector space. Geometrically, the unit vectors i, j,
k are represented by arrows of unit length along the positive coor-
dinate axes as shown in Fig. 1.10. Assuming that the components of
the vector are all positive, we notice that the vectors al i , a2j , a3 k
: I
: I
: I : I
: I
: 3
z3 k
A = z, i + z2j + z3k
+
Y
X
Fig. 1.10 Analytic representation of a vector.
Vector Algebra
are parallel to the coordinate axes and they form the edges of a
rectangular parallelepiped of which A is the diagonal.
Position Vector of a Point
Let P be a point with the coordinates (x, y, z). We call the
vector R = x i + y j + z k the position vector of the point. Geometri-
cally, the position vector of a point is the directed line segment
drawn from the origin to the point I?
Example 1. Let A = i - 2j + 2k and B = 2i + 4j - 5k Find the
linear combination 2A + B and (1 /2)A - 2B.
Solution: We have
(1/2)A - 28 = (1/2)(i - 2j + 2k) - 2(2i + 4j - 5k)
= (-7/ 2)i -9j + I l k
Example 2. Let A = - 4i +2j , B = 2 i + j , and C= 2 i + 3j.
Find the scalars m and n such that C = m A + n B.
Solution: We have
Equating the corresponding components, we obtain the system of
equations
-4m +2n = 2, 2m + n = 3
for the scalars m and n. Solving for m and n, we find m = 112
and n = 2. Thus
(1/2)A + 2B = C
Example 3. Let A, B and C be three distinct points on a straight
line and let A, B and C denote their respective position vector
relative to a fixed point 0 (Fig.l.11). If C is the midpoint of the line
segment AB, show that C = (A + B )/2.
Solution: In Fig. 1.1 1 we note that the arrow AB represents
the vector B - A. Since C is the midpoint of the line segment AB, it
Chapter 1
Fig. 1.11 Position vector of the midpoint of a line segment.
follows that the arrow AC represents the vector (B - A) 12. By the
law of vector addition, we know that OC = OA + AC . Thus we
have C = A + ( B- A) / 2 = (A + B)/2.
Physical Applications
As mentioned previously, physical quantities that have both
magnitude and direction are represented and studied by using
vectors. In the following examples, we consider some physical
problems to illustrate the applications of vectors.
Example 4. Suppose ship A is cruising in the easterly direc-
tion at the speed of 15 h / h r while ship B is speeding at 30 km/hr
in the northeast direction. Assuming that both ships started from
the same place at the same time, find the rate at which the two ships
are separating.
Solution: Taking the x-axis in the easterly direction and the y-
axis in the northerly direction (Fig. 1.12), we see that the velocity of
the ship A is represented by the vector A = 15i and the velocity of
the ship B by the vector
B = 30(cos 45Oi + sin 45Oj) = 15\12(i + j)
Vector Algebra
Ship B A
A = 1 5 i Ship A
Fig. 1.12 Rate of displacement between two ships.
The displacement of the ship B relative to the ship A is represented
by the vector D = B - A. Hence the rate at which the ships are sepa-
rating is equal to the manitude of the displacement vector D. Since
D = B - A = 15(\12 - I)i + 15\/2j, it follows that
Example 5. A body weighing 100 kg is suspended from two
ropes as shown in Fig. 1.13. Find the tension in the ropes.
Solution: Let the tension in the ropes be denoted by F1 = ali +
a2j and F = bli + b2j. The weight of the body is represented by the
vector W = -100j. From elementary mechanics, since the system is
in equilibrium, the sum of all the forces must be equal to zero.
Thus F1 + F2 + W = (al + bl)i + (a2 + b2 - 100) = 0. This implies
Chapter 1
Fig. 1.13 Tension in ropes from which a weight is suspended.
Hence, al = - bl and a2 + b2 = 100. By symmetry, we know that
IF, I = I F2 1 , which implies that a2 = b2. Therefore, a2 = b2 = 50.
Since al = 1 F1 I cos 45' and a2 = I F1 1 sin 459 it follows that 1 F1 I = / F2) =
5Od2 and al = - bl = 50. Thus F1 = 50(i + j) and F2= 50(-i + j). Notice
that F1 + F2= 100j.
mph
from
Example 6. An aircraft is cruising at an air speed of 25Od2
in the northeast direction. If a wind of 25 mph is blowing
the west (tailwind), what is the effective speed of the aircraft
relative to the ground? In what direction is the aircraft headed as a
result of the tailwind? Assume that all motions are taking place in
a plane.
Solution: If there was no tailwind, the aircraft would remain
on its original course flying in the northeast direction. As a result
of the tailwind, the aircraft is also moving toward the east at 25
mph. Thus the effective velocity of the aircraft is the sum of its
Vector Algebra 21
cruising velocity and the tailwind velocity. Let A denote the velo-
city of the aircraft and let W denote the velocity of the tailwind.
Referring to Fig. 1.14, we note that
A = 250\/2(cos 45Oi + sin 45Oj) = 250(i + j), W = 25i
Hence the effective velocity of the aircraft is
and the effective speed is 1 V I = [(275)" ( 250) ~] ' / ~ = 25 d221 mph.
The direction in which the aircraft is headed is given by the angle
0 = arctan(2501275) = 42.27 deg.
Fig. 1.14 Effective velocity of an aircraft with tailwind.
1.2 EXERCISES
1. Prove the properties (1.8) through (1.11) of vector addition.
2. Prove the properties (1.14) through (1.17) of scalar multiplica-
22 Chapter 1
tion.
If A = 2 i - j + 3k a n d B=3 i +2 j - 4 k , find (a) 2 A+ B,
(b) -3A + 2B, (c) I A + B 1 , (d) J A - B 1.
If A = i - j andB =2i - 2j +3k, f i nd (a) 4A-2B,
@) 2A-3B, (c) M A + B) I / (d) IAI + IBI.
If A = 3i - j + 4k and B = 2i + j, find the vector X so
that 2A + 3X = 5B.
If A = -i + 2k and B = 2i + j, find the vector X so that
-2X + 3B = 2A.
Let A = 3i - j + 4kand B = 2i + k. If C = 4i - 2j + 7k, find the
scalars nz and 1.2 so that C = nz A + nB.
Let A = 4i - 3j - 6kand B = 2i + 5j - 7k. If C = 8 i - 19j - 4k,
find the scalars nz and n so that C = nz A + 1.2 B.
If A = i - j + 2k, B = j - k, C = 2i + k, and D = 3i + 2j - k, find
the scalars a , b , and c so that D = aA + bB + c C.
Repeat Problem 9 if A = 2i - j + 4k, B = i + 2j - 3k, C = 2i - 3j
- 3k, and D = 3i - 6j + 4k.
Showt hat (a) IA + B I s I AI + I B I , ( b ) IA - B I = I B - AI .
Interpret the inequality (a). When does equality hold?
Show that ( A - B I 2 I ( A ( - ( B I I. When does equality hold?
Let A be a constant vector and let B be a vector whose
magnitude is constant but whose direction varies. Determine
the maximum and the minimum value of I A - B I.
An airplane is cruising at an air speed of 250 rnph in the
easterly direction. If a wind of 50 rnph is blowing from the
southwest, what is the effective velocity of the plane and what
is its course?
A helicopter has an air speed of 100 mph. A wind of 30 rnph is
blowing from the southeast. In what direction should the pilot
fly the helicopter in order for the plane to be moving north?
A light plane is flying at an air speed of 150 rnph in the N 600
E direction. If the tail wind is 50 rnph due north, what is the
effective velocity of the plane and what is its course?
At noon the location and the velocity of ships A and B are
sighted from a lighthouse (assumed at the origin) to be RA =
3(i + j) (miles), VA = 4i + 3j (rnph) and RB = 8i + 3j, VB = 2i +
3j, respectively. If the ships continue in their current course,
show that the ships will collide. Find the time and the loca-
tion where the collision occurs.
Vector Algebra 23
18. At noon an aircraft takes off from an airport with air speed of
100 mph at an angle of inclination of 15 deg. Ten minutes later,
another aircraft takes off from the same airport at the same
point with air speed of 110 mph at 20 deg. How far apart are
the aircrafts at 12:20 PM?
19. Three forces F1, F2 and F3 act in the direction of the sides AB,
BC and CA, respectively, of an equilateral triangle ABC. If
I F2 1 =2 1 F1 I and I F3 I = 3 1 F1 I , find the resultant of the forces.
Show that the resultant force is parallel to the bisector of the
angle at the vertex A.
20. A body weighing 100 kg is hanging from two wires that are
attached to two pegs 8 m apart on a ceiling as shown in Fig. (i)
below. Find the tension in each of the wires.
Fig. (i )
21. Suppose A, B and C are three vectors that are represented by
the directed line segments AB, BC and CA of a triangle ABC.
(a) Show that A + B + C = 0. (b) Show that the sum of the
vectors represented by BA, CA and 2(BC) has a magnitude
equal to 3(BD), where D is a trisecting point on the side AC .
22. A man crossing a river 1500 ft wide rows a boat at the speed of
150 ft/min. Assuming that the man always has the boat
pointed perpendicular to the opposite bank and that there is a
current of 30 ft/min, what is the velocity of the boat relative to
an observer on the ground? How long will it take the man to
cross the river and how far downstream on the opposite bank
will he land?
Chapter 1
1.5 SOME APPLICATIONS IN GEOMETRY
The fact that vectors are mathematical entities independent of
any frame of reference makes vector algebra a convenient tool for
proving some theorems and establishing certain identities in geo-
metry. In this section we consider some examples to illustrate the
technique and procedure.
Example 1. Prove that the diagonals of every parallelogram
bisect each other.
Solution: Let OACB be a given parallelogram, and let P and Q
be the midpoints of the diagonals OC and BA, respectively (Fig. 1.15).
Denote the directed line segments OA and OB by A and B. Then by
the parallelogram law of vector addition, the directed line segment
OC represents the vector sum A + B. Since P is the midpoint of OC,
OP represents the vector (A + B)/2. On the other hand, the directed
line segment BA represents the vector A - B, and so BQ represents
the vector (A - B)/2. Now since OQ = OB + BQ, we have OQ = B +
(A - B)/2 = (A + B)/2 = OP, which implies that P and Q are the same
point. Thus the diagonals OC and BA bisect each other.
Fig. 1.15 Diagonals of a parallelogram bisect each other.
Example 2. Let A, B, and C denote the position vectors of the
points A, B, and C, respectively, relative to a fixed point 0, where A,
B, and C are collinear (Fig. 1.16). If the point C divides the line seg-
ment AB in the ratio m:~z (both nz and n not equal to zero), show
that C = (nA +m B ) / ( m + n).
Solution: Since AC : CB = nz: n, it follows that CB =
Vector Algebra
Fig. 1.16 Position vector of a dividing point of a line segment.
(n /nz )AC . But AC represents the vector C - A and CB represents
the vector B - C. Hence B - C = (n /nz )(C - A). Solving for C, we
obtain the desired result.
When nr = 12, C is the midpoint of AB for which we obtain C =
(A+B)/2. This agrees with the result obtained in Example 3, See. 1.4.
Example 3. Prove that the medians of a triangle intersect at a
point which trisects the medians.
Solution: Let A, B, and C denote the vertices of a triangle, and
let A, B, and C denote their respective position vectors relative to a
fixed point 0 (Fig. 1.17). Denote by L, M, and N the respective mid-
points of the sides BC, CA, and AB, and let L, M, and N denote their
position vectors. Then we have
Now let P, Q, and R denote the points which divide the respective
medians AL, BM, and CN in the ratio 2 : 1 (trisecting points). Then,
according to the result of Example 2, we have
26 Chapter 1
This implies that the points T: Q, and R coincide since they have the
same position vector.
Fig. 1.17 Intersection point of the medians of a triangle.
Example 4. Let A, B, C, and D be four points in space with
possition vectors A, B, C, and D, respectively, relative to a fixed
point 0 (Figure 1.18). I f 3A - 2B + C - 2D = 0, show that the points
A, B, C, and D are coplanar. Determine the point at which the line
segments AC and BD intersect and the ratio in which the point
divides the line segments.
Solution: If we can show that the line segments AC and BD do
intersect, then it will follow that the points A, B, C, and D are
coplanar. In Fig. 1.18 we note that the point P is a point on AC if and
only if OP is a linear combination of A and C; likewise, Q is a point
on BD if and only if OQ is a linear combination of B and D. Now
from the given equation, we obtain 3A + C = 2(B + D) or (3A + C)/4
= (B + D)/2. This indicates a point common to the line segment AC
and BD. Thus the line segments intersect and, therefore, the
points A, B, C, and D are coplanar.
Now (3A + C)/4 denotes a point on AC which divides the line
segment in the ratio 1 : 3. On the other hand, (B + D)/ 2 denotes the
midpoint of BD.
Vector Algebra
Fig. 1.18 Coplanar points.
1.3 EXERCISES
1. Prove that if the midpoints of the consecutive sides of a
quadrilateral are joined by line segments, then the resulting
quadrilateral is a parallelogram.
2. Let Q, and R be the respective midpoint of the sides AB, BC,
and CA of a triangle ABC. Show that the quadrilateral APQR
is a parallelogram.
3. Prove that the line segment joining the midpoints of two sides
of a triangle is parallel to and is equal in length to half the
length of the third side.
4. Show that the line segment joining the midpoints of the two
non-parallel sides of a trapezoid is equal in length to half the
sum of the length of the parallel sides and is parallel to them.
5. Show that the line which joins one vertex of a parallelogram
to the midpoint of an opposite side trisects the diagonal.
6. Let A , B , C , and D denote the position vectors of the points A,
B, C, and D, respectively, relative to a reference point 0. If
B - A = 2(D - C), show that the line segments AD and BC
(extended if necessary) meet in a point which trisects these
lines.
28 Chapter 1
The points P and Q divide the sides AC and BC of a triangle
ABC in the ratio h/ ( l - h) and k/ ( l - k), respectively. If PQ =
m AB, for some scalar m , show that h = k = 1 - rn .
Let AB and AC represent the vectors A and B, respectively,
such that [ A I = ( B I . Show that a point P is on the bisector of
the angle CAB if and only if there exists a scalar rn such that
AP = m (A + B).
Show that the line segments which join the midpoints of the
opposite sides of a quadrilateral bisect each other.
Consider a triangle ABC. Let D be a point between A and B,
and let E be a point between B and C, each dividing the sides of
the triangle in the ratio 2 : 1. Find the ratio in which the point
of intersection of AE and CD divides the line segments.
Let D and E be points on the sides AC and BC of a triangle such
that the line segments DE is parallel to AB. If the line segments
AE and BD meet at a point P, show that the line CP bisects AB.
A line from a vertex of a triangle bisects the opposite side. It is
bisected by a similar line issuing from another vertex. In what
ratio does the latter line intersect the opposite side?
Le t A=OA, B=OB, C=OC, a n d D=OD, whereA, B, C, and
D are four given points, and 0 is any fixed point. If A - B =
C - D, show that the points A, B, C, and D form the vertices of
a parallelogram.
Let A, B, C, D, and E denote the vectors represented by the
arrows from the center of a regular pentagon to the vertices.
Express the sides of the pentagon in terms of the vectors and
s h o wt h a t A+ B + C + D+ E = O .
1.6 SCALAR PRODUCT
There are two important operations on vectors which corre-
spond to the operation of multiplication in algebra. One of these
operations results in a scalar and the other results in a vector. We
consider first the operation that results in a scalar.
Definition 1. Let A = al i + a2 j + a, k and B = bl i + b2 j + b, k
be two vectors. The scalar product of A and B, denoted by A. B, is
the scalar (real number) given by
Vector Algebra
The scalar product of two vectors is also known as the dot product
(for obvious reason) or the inner product.
From the definition (1.20), it is easy to verify that for any
vectors A, B, C and for any scalar rn , we have
(a) A* B = B - A (commutative law)
(b) A. (B + C ) = A- B + A. C (distributive law) (1.21)
(c)
nz (A. B ) = (nz A). B = A- (nz B)
(d) A . A = I A I ~
These properties are analogous to the properties satisfied by real
numbers with respect to ordinary multiplication.
Geometric Interpretation of the Scalar Product of Vectors
Fig. 1.19 Angle between two vectors.
30 Chapter 1
The scalar product of two vectors has an interesting and im-
portant geometric interpretation. Before we present this interpreta-
tion, we first introduce the notion of angle between two vectors.
Let A and B be two vectors, and let them be represented by the
arrows OA and OB as shown in Fig. 1 .I 9. By the angle between the
vectors A and B, we mean the angle 8, 0 < 8 < n, formed by the line
segmens OA and OB. It follows from this definition that if B = rn A,
then the angle between A and B is zero i f rn > 0 and 8 = n if m < 0.
Two vectors are said to be parallel i f and only if the angle between
them is 0 or n. On the other hand, two vectors are said to be ortho-
gonal or perpendicular if and only if the angle between them is n/2.
These concepts are of course geometrically evident.
Now let A and B be two vectors represented by the arrows OA
and OB (Fig. 1.20). By the parallelogram law of vector addition, we
know that BA represents the vector A - B. Applying the law of
cosines to the triangle OAB, we find
.:.2
----.---.----.----
a .
a . . .
I :
_^^_--_3
X
Fig. 1.20 The scalar product of two vectors.
Vector Algebra
where 8 is the angle between A and B. In terms of the magnitude of
the vectors A, B, and A - B, this becomes
On the other hand, by the properties of scalar product, we have
Comparing (1.22) and (1.23), we obtain
Fig. 1.21 Component of a vector along another vector.
Chapter 1
Thus the scalar product of two vectors is equal to the product of
their magnitudes and the cosine of the angle between then.
If we examine the plane formed by the vectors A and B in
Fig. 1.21, we see that the quantity I B I cos 8 is simply the orthogonal
projection of the magnitude of B onto the vector A. We call it the
component of B along A; it is positive, zero, or negative according
as 0 < 8 < n/ 2, 0 = 0, or n/ 2 < 0 < n. Likewise, the quantity [ A I cos 8
is called the component of A along the vector B. Thus the scalar
product of A and B may also be stated as the product of IA I and
the component of B along A, or the product of ( B I and the compo-
nent of A along B.
From (1.24) we obtain a formula for calculating the angle bet-
ween two vectors, namely,
Example 1. Let A = 2i - j + k and B = i + 2j - 2k. Find the
component of A along B, and the component of B along A.
Solution: By definition the component of A along B is given
by 1 A I cos 8. By (1.24) this is equal to
IAIcos8 =A. B/ l BI
Since
A.B = 2(1) + (-1)2 + I(-2) = -2,
1B I = 3
it follows that
I AI cos8= A*B/ l BI =-213
Similarly, since I A 1 = 46, the component of B along A is given by
Example 2. Find the angle between the vectors A = 2i - j - 2k
and B = i + 2j + 2k.
Solution: By (1.20) we have A* B = 2(1) + (-1)2 + (-2)2 = -4.
Since 1 A I = 3 and I B 1 = 3, it follows from (1.25) that cos 8 = -419.
Vector Algebra
Thus the angle between the vectors is
0 = arccos (-419) = 116'23
From the formula (1.24) we see immediately that two nonzero
vectors A and B are orthogonal if and only if their scalar product
is zero. Indeed, if the vectors are orthogonal, then the angle 8
between them is x/2. Hence cos 0 = 0 and by (1.24), A. B = 0. Con-
versly, if A* B = 0 and A and B are nonzero vectors, then by (1.24),
cos 0 = 0. This implies that 8 = x/2, which says that the vectors are
orthogonal. We state this important result as a theorem.
Theorem 1. Two nonzero vectors A and B are orthogonal if
and only if A. B = 0.
For the unit base vectors i , j , and k, we have
The set of base vectors (i , j , k) is said to form an orthonormal basis.
34 Chapter 1
Example 3. Prove that if a and b are the two legs of a right
triangle, and c is its hypothenuse, then a
2
+ b
2
= c
2
. (This is the well
known Pythagorean theorem in plane geometry.)
Solution: Let the vectors A and B represent the legs of the
right triangle as shown in Fig. 1.22. Then A + B represents the
hypothenuse. Since A is perpendicular to B, A' B = 0. Hence
Example 3. Let A = 2i + j - 2k and C = i + 2j + 3k. Express
the vector C as the sum of a vector that is parallel to A and a
vector that is orthogonal to A.
Solution: A vector that is parallel to A is of the form nr A, for
some scalar m . Let B be a vector that is orthogonal to A. We wish
to find nr and B so that C = nr A + B. Let us take the scalar product of
C and A. Since B is orthogonal to A, we obtain
Since A o C = 2(1) + l(2) + (-2)3 = - 2 and \ A I = 3, we find
TH = A . c / I A ~ ~ =- 219
Hence
B = C - nzA = (13i + 20j + 23k)/9.
It is readily seen that A. B = 0 and
Work Done By a Force
A simple physical problem which makes use of the scalar
product of vectors is the problem of calculating the amount of work
done by a force when it displaces an object through a certain dis-
tance. Suppose that a constant force F acts on an object and
Vector Algebra
Fig. 1.23 Work done by a force through a displacement.
causes the object to be moved through a distance represented by the
displacement vector D (Fig. 1.23). In physics, the work done by the
force is defined as the product of the displacement and the compo-
nent of the force along the displacement. Hence if 8 is the angle bet-
ween the force and the displacement, then the component of the
force along the displacement is equal to IF lcos 8. Thus the work
done by the force is given by
Example 5. If a constant force F = 2i - j + 3k (in kg), acting on
an object, moves the object through a displacement D =3i + 2j + 4k
(in meters), then the work done by the force is equal to
Work = F- D = 2(3) + (-1)2 + 3(4) = 16 kg-m.
1.4 EXERCISES
In each of Problems 1 through 4, find the orthogonal projec-
tion of IA I along B, and determine the cosine of the angle between
the vectors.
Chapter 1
A= 2 i - 2 j + k, B= - i - 2 j + 2 k .
A = i - 2 j + 3k, B= 3 i + j - 2k.
A = 3i - 5j + 4k, B = 4i + 3j - 2k.
A = 2 i + 3j-6k, B = i - 2 j + 4k.
Express the vector A in Problem 1 as the sum of two vectors,
one parallel to and the other orthogonal to B.
Express the vector B of Problem 4 as the sum of two vectors,
one parallel to and the other orthogonal to A.
Show that the vectors A = i - 2j + k, B = 3i + j - k, C = i + 4j +
7k are mutually orthogonal.
Let A = i + 2j + k and B = 2i + j - k Find a vector that is
orthogonal to both A and B. Is such a vector uniquely deter-
mined?
Let A = i + 3j - 4k and B = 2i - 3j + 5k. Find the value of nz
so that A + nz B is orthogonal to (a) A, (b) B.
Let A = 3i - j + 5k and B = 2i - 4j - 3k. Find the value of nz
so that m A + B is orthogonal to (a) A, (b) B.
Le t A= i + 3 j + k , B= 2 i - j + 2 k , a n d C= - 3 i + 2 j - k .
Find the values of the scalars nz and n so that the vector
C - m A - n B is orthogonal to both A and B.
Repeat Problem 11 if A = 3i - 2j + 5k, B = 2i + j - k, and
C = - i + 4 j - 6 k
Let A= i - 2 j + k , B = i +j - 2k, and C= 2 i - j + k . Findall
the unit vectors that are linear combinations of A and B and
that are orthogonal to C.
A constant force F = 3i + j - 2k moves an object through a
displacement D = - 2i + 2j + k. What is the work done by the
force?
A constant force F = 2i + 5j - 3k moves an object through
consecutive displacements Dl = 3i - 2j and % = i + 4j - 2k.
Find the total work done by the force.
Let A = cos a i + sin a j and B = cos f3 i + sin f3 j. By talung
the dot product of A and B, derive the trigonometric identity
cos (a - f3) = cos a cos f3 + sin a sin P
Show that a set of orthogonal nonzero vectors (A1, Az, . . , A,)
is linearly independent.
Vector Algebra
18. Let
Show that the set (A, B, C) is an orthonormal basis.
19. Prove the Cauchy-Schwarz inequality I A* B 1 5 1 A I I B I .
20. Prove the triangle inequality I A + B I 5 I A I + I B I .
(Hint: Consider l A + B 1 = (A+B)* (A+B) and use the result
of Problem 19.)
21. Deduce I A - B I 2 I I A I - I B I I from the triangle inequality.
22. Show that
(a) I A + B I ~ + I A - B I ~ = 2 1 ~ 1 ~ + 2 1 ~ 1 ~
(b) I A + B I ~ - I A - B I ~ =4A*B
23. Prove that the vectors A and B are orthogonal if and only if
I A + B I = I A - B I . What is the geometric interpretation of
this?
24. Let A and B be two vectors and set m = I A I, n = I B I .
(a) Show that the vectors nA + m B and nA - mB are orthogo-
nal. (b) If C = (nA + m B)/(m +n), show that the angle between
A and C is the same as the angle between B and C.
25. Using the scalar product, prove that a triangle inscribed in a
semicircle with one of its sides along the diameter is a right
triangle.
26. As in Problem 25, prove that the line joining the vertex of an
isosceles triangle to the midpoint of its base is perpendicular
to the base.
27. Prove Appolonius theorem which states that i n a triangle
ABC,
where D is the midpoint of BC.
1.7 VECTOR PRODUCT
In Sec. 1.6 we defined a product ot two vectors that always
yields a scalar or a real number. There is another type of product of
two vectors in which the result is another vector. The resulting
new vector has the distinct property of being orthogonal to the two
38 Chapter 1
given vectors. This product is called the vector yroduct or cross
product and it is defined as follows:
Definition 1. The vector product of two vectors A = al i + a2 j
+ a3 k and B = bl i + b2 j + b3 k , denoted by A x B (read A cross B), is
the vector defined by
Because of its notation, this product is sometimes called the cross
product of the vectors A and B. In contrast to the term inner product
for the dot product of two vectors, the product (1.26) is also known
as the outer product of A and B.
Example 1. If A = 3i - 2j + 4k and B = 2i + 3j + k , then
At first glance, it may appear that the definition (1.26) is too
complicated to remember. Actually it is rather easy as all we need to
remember is the first component of the vector product, a2b3 -
We observe that the first term 3b3is the product of the second com-
ponent a2 of A and the third component b3 of B, while the second
term a+2 is just the reverse, being the product of the third compo-
nent of A and the second component of B. Thus the first component
involves the subscripts 2 and 3, pertaining to the components of A
and B. Knowing the first component, the second component is then
obtained by cyclic permutation of 1, 2, 3; that is, the second compo-
nent involves the subscripts 3, 1, and finally, the third component
involves the subscripts 1, 2. Thus, the components of the vector
product A x B are:
Subscripts: 2 3 3 1 1 2
Vector Algebra 39
A more convenient formula for the cross product (2.26) which
makes use of the notion of a determinant of a matrix is given in
(1.29). For the sake of those who may not be familiar with the
concept of a matrix and its determinant, we present briefly the basic
ideas below.
Matrices and Determinants
We use capital letters A, B, . . . to denote matrices, and I A ( or
det A to denote the determinant of a matrix A.
Definition 2. A 2 x 2 (read two by two) matrix A is an array of
(real) numbers al , a2, bl , b2 written as
The numbers a l , a2 are called the elements of the first row, and bl,
bz the elements of the second row. On the other hand, the numbers
al, bl are called the elements of the first column, and ;tl, b2 the ele-
ments of the second column.
The determinant of a 2 x 2 matrix A, denoted by I A I, is a real
number defined by the formula
This is called a determinant of the second order.
For example:
40 Chapter 1
From the definition (1.27), it is easily seen that if we inter-
change the rows or the columns of the matrix, the determinant of
the resulting matrix changes sign. For example, i f we interchange
the rows, we see that
If we interchange the columns, we obtain
Next, we define a 3 x 3 matrix A as an array of nine numbers
ai, bit ci (i = 1, 2, 3), written as
The numbers are again arranged in three rows and three columns.
The determinant of a 3 x 3 matrix is defined in terms of
determinants of the second order and it is given by the formula
This is called a determinant of the third order .
Vector Algebra 41
Without some mnemonic device, formula (1.28) would be
difficult to memorize. The rule to remember here is that the right-
hand side of (1.28) consists of the algebraic sum of products of an
element ai (1 5 i s 3) of the first row and the determinant of a 2 x 2
matrix obtained when the elements of the first row and the ele-
ments of the i-th column (the column in which ai lies) are deleted.
The s i p s of the terms alternate beginning with a plus sign in the
first term. Thus, the second order determinant in the second term
on the right-hand side of (1.28) is the determinant of the resulting
matrix when the elements of the first row and the elements of the
second column are deleted:
With the formula (1.28) we say that the determinant is glven
by expansion by the elements of the first row. The second order
determinants appearing in (1.28) together with the algebraic sign are
called the cofnctors of the elements in the first row. That is, the
determinants
are the cofactor of the elements al , a2, a3, respectively.
For example:
Chapter 1
Just like the second order determinant, the determinant of a 3
x 3 matrix also changes sign whenever any two rows or any two
columns of the matrix are interchanged. We leave this to the reader
to verify. For example, if we interchange the first and the third
columns of the last example above, we obtain
Other important properties of determinants are given in the exer-
cises.
Formula for the Vector Product A x B
Now let us return to the definition of vector product. From
the formulas (1.26) and (1.27), we observe that
Comparing this with the formula (1.28), we deduce the following
alternative representation for the vector product
For example, if A = 2 i + j + 3k and B = i - 2 j + k , then
A x B =
i j k
al a2 a3
b1 b2 b3
Vector Algebra
From the definition of vector product, it can be easily shown
that for any vectors A, B, C and for any scalar nz, the following
properties hold:
A x B =
(a) A x B = - (B x A)
(b) A x (B + C) = (A x B) + (A x C)
(1.30)
(c) m (A x B) = (nz A) x B = A x (nz B)
(d) A x A = O
These properties correspond to those listed in equation (1.21) for the
scalar product. Unlike the scalar product, however, (1.30a) implies
that vector product is not commutative. This property means that
the vectors A x B and B x A are equal in magnitude but opposite in
direction. Thus the order of the factors in a vector product makes a
difference. This property should be kept in mind as this may well be
the student's first encounter with an operation bearing the name of
"product" which is not commutative.
Property (1.30d) says that the cross product of a vector with
itself results in the zero vector, even though the vector itself may
not be the zero vector. This is quite in contrast to the scalar product
of a vector with itself, that is, A-A = / A 12, which is zero if and only
if A is the zero vector.
i j k
2 1 3
1 -2 1
Geometric Interpretation of A x B
=
1 3 2 3 2 1
As in the case of the scalar product, the vector product A x B
also has an important geometric interpretation. First, we show that
the vector A x B is orthogonal to both the vectors A and B. From
the definition of scalar product, we see that
44 Chapter 1
and similarly, B* (A x B) = 0. Thus A x B is orthogonal to both the
vectors A and B. Further, when the vectors are represented geome-
trically by arrows, the direction of the vector A x B is related to the
direction of A and B by the right-hand rule (Fig. 1.24). That is, if the
index finger of the right hand points along the vector A and the
center finger points along the vector B, then the thumb points in the
direction of A x B (assuming that the thumb is perpendicular to the
two fingers). If the vectors A and B are orthogonal, that is, A-B = 0,
then the vectors A, B, A x B, in that order, form a right-handed tri-
ple of vectors that are mutually orthogonal.
For the unit vectors i, j, k, it follows that
i x i = O, j x j = O, k x k = O
i x j = k , j x k = i , k x i = j
which of course can also be verified directly from the formula (1.29).
Next, we consider the magnitude of the vector A x B. For this
purpose, we need the following identity
This identity can be established directly by showing that the two
expressions
I A x B 1 = (a2b3 - a3b2 )2 + ( apl - albd2 + (a1b2 - a&1)2
and
I A ~ ~ I B ~ ~ - ( A B ) ~ = ( a t + a 2 + a32)(b12+ b 2 + b33
are indeed equal. (See also Example 5, Sec. 1.9.) Now let 0 denote
the angle between A and B, and let us write A- B = I A ( I B I cos 8.
Then the identity (1.31) becomes
from which we obtain
Vector Algebra
Fig. 1.24 Geometric interpretation of the vector product.
This gives the magnitude of the vector A x B.
Example 2. Find all the unit vectors that are orthogonal to
the vectors A = i + 2j - k and B = 2i - 2j + k.
Solution: We know that the vector
is orthogonal to the vectors A and B. By normalizing this, we obtain
the unit vectors
u = + (j + 2k)/\/5
Example 3. Let A = i + 2j - 3k and B = 3i - j + 2k. Find
A x B and verify the formula (1.32).
Solution: By (1.29) we have
i j k
-3
3 - 1 2
2 -3 1 -3 1 2
= I 1 2 l i - 13 2 l j +13 -1 l k
Chapter 1
= i - 1l j - 7k
Therefore
IA x B I = [l + (-11)~+ (-7)2]112 = dl 71
Since
and A. B = 1(3) + 2(-1) + (-3)2 = - 5, we find
cos 0 = A. B/ IA I ( B 1 = -5/14
Hence
sin 8 = [I - cos
2
= [I - 25/ 1 96]lI2 = dl 71 / 14
and so
I A I lB I sin 0 = (d14)(d14)(d171/14) = dl71
as we wish to verify.
Referring to Figure 1.24, we see that IB (sin 0 represents the
altitude of the parallelogram formed by the vectors A and B, and
IA I represents its base. Since the area of a parallelogram is equal to
the product of its base and its altitude, it follows from (1.32) that the
magnitude of the vector A x B represents numerically the area of the
parallelogram formed by the vectors A and B. This has an important
geometric application as we shall see later.
In Sec. 1.6 we saw that two nonzero vectors A and B are ortho-
gonal if and only if A* B = 0. If the vectors are parallel so that the
angle between them is either 0 or x, then it follows from (1.32) that
IA xB I = ] A ( IB I si n8 = 0. This implies that A x B = 0, the zero
vector. Conversely, if A x B = 0, then I A x B I = 0. Since A + 0, B + 0,
we infer from (1.32) that sin 0 = 0. Hence 8 = 0 or 0 = x, which
means that the vectors are parallel. We state this result as a theo-
rem.
Theorem 1. Two nonzero vectors A and B are parallel if and
only if A x B = 0.
Example 4. Let A = 3i + 2j - 6k and B = i - 2j + 2k. Find the
area of the triangle formed by the vectors A, B, and A - B.
Solution: For any vectors A and B, we know that ( A x B I
Vector Algebra 47
represents the area of the parallelogram formed by the vectors A
and B. Since the area of the triangle formed by the vectors A, B and
A - B is one half that of the parallelogram (Fig. 1.25), it follows that
area of triangle = (1 /2) ( A x B (
Now
A x B = (3i + 2j - 6k) x (i -2j + 2k) = - 8i - 12j - 8k
and
I A x B I = [(-B)~ + (-12)~+ (-8)q1I2 = 4d17
Hence the area of the triangle is 2d17 square units.
Fig. 1.25 Area of a triangle formed by the vectors A, B, and A - B.
Example 5. Find the area of the triangle with vertices at the
points P: (1,2, 3), Q: (-1, 3,2), and R: (3, -1,2).
Solution: Let A and B denote the vectors represented by the
directed line segments PQ and PR, respectively. Then
A = - 2 i t - j - k and B = 2 i - 3 j - k
Taking the cross product of these vectors, we find
48 Chapter 1
Since I A x B I = 4d3, the area of the triangle is 2d3 square units.
Some Applications of the Vector Product
As an example of the use of vector product, let us consider the
motion of a particle P that is revolving at an angular speed w about a
fixed axis (Fig. 1.26). By convention the angular velocity of the par-
ticle, which is a vector quantity, is represented by a vector w whose
magnitude is equal to w and whose direction is along the axis of
rotation as determined by the right-hand rule. (Let the fingers of the
right hand point in the direction of rotation. Then w points in the
direction of the thumb along the axis of rotation). Now let R denote
the position vector of the particle relative to son~e fixed point on the
axis of rotation. The velocity V of the particle is then given by
In fact, from the definition of vector product, we know that V is
orthogonal to both w and R. This implies that V is tangential to the
path of motion of the particle. Further, by (1.32), we see that
Fig. 1.26 Velocity vector of a particle revolving around an axis.
Vector Algebra 49
where r is the radial distance of the particle from the axis. This is
the basic formula in elementary physics which relates the tangential
speed z1 of the particle with its angular speed w.
Example 6. A particle is revolving about the z-axis on a circle
of radius 2 cm on the plane z = 5 cm at the rate of 3 revolutions per
second. What is the velocity vector and the tangential speed?
Solution: Since the z-axis is represented by the unit vector k,
the angular velocity is given by w = wk = 3k. Let R denote the
position vector of the particle on the plane z = 5. Then R = x i + y j
+ 5 k, where x + y2 = 4. Therefore
and the tangential speed is
As another example of a physical problem where vector pro-
duct is used, consider a force F that acts on a particle at a point I!
Denote by R the position vector of P relative to a fixed point 0 as
shown in Fig. 1.27. The moment or torque of the force F about the
point 0 is defined as the product of the magnitude of the force and
the distance from 0 of the line of action of the force, that is, referring
to Fig. 1.27,
Since cr = 8 - n/2, where 8 is the angle between F and R, we see that
which is the magnitude of the vector product of F and R. We call
the vector
M o = R x F (1.34)
the moment vector or torque vector of a force F that acts at a point P,
where R is the position vector of P relative to a reference point 0.
Notice that the vector Mo points in the same direction as the vector
Chapter 1
Fig. 1.27 Moment vector M,.
representing the angular velocity of the point P as caused by the
force F in accordance with the right-hand rule.
Example 7. A force F = -2i + j + 3k (in lbs) acts on a body
located at the point (-3, 2, 4) (distance in ft). Find the torque vector
about the origin and the amount of torque caused by the force.
Solution: By (1.34) the torque vector about the origin is given by
Hence the amount of torque is I M, I = d6 ft-lb.
M , = R x F = - 3 2 4
1.5 EXERCISES
1. Find the determinant of each of the following matrices:
i j k
-2 1 3
= 2 i + j + k
Vector Algebra
2.
Find the determinant of each of the following matrices:
3. Show that the determinant of a 3 x 3 matrix changes sign
whenever any two rows or two columns are interchanged.
4. Show that if a row or a column of a matrix is multiplied by a
constant k, then the determinant of the resulting matrix is k
times that of the original matrix.
5. From the results of Problems 3 and 4, deduce that if the
elements of any two rows or two columns of a 3 x 3 matrix are
proportional, then the determinant of the matrix is zero.
6. Verify each of the following identities:
(a)
7. Prove the properties listed in (1.30).
8. Show in two different ways that the vectors A and B are
a 2 a 3
bl + d l b2+ d2 b3+ d 3
1 C2 3
(4
a , + kb, a2 t kb2 a ? + kb3
bl b2 b3
1 2 C3
=
=
bl b2 b3
C2 C3
a 2 a 3
b, b2 b,
c1 c2 C3
+
a 3
d l d2 d,
c2 c3
Chapter 1
parallel:
(a) A = - i + 2j - 3k,
B = 2i - 4j + 6k
(b) A= 3 i +6j - 9k, B = i +2 j - 3 k
Find a unit vector that is orthogonal to the given two vectors:
(a) A = i - 2 j +3 k , B = 2 i + j - k
(b) A= 3 i - j +6 k , B = i + 4 j + k
I f A= 2 i + j - k , B = - i +3 j + 4k, C=i - 3j + 5k, compute
the vectors (a) A x B, (b) B x C, (c) A x B. C, (d) A- B x C,
(e) A x (B x C), (f) (A x B ) x C.
What do you notice about
A0 Bx C a n d Ax B* C; Ax ( Bx C) and ( AxB) xC?
If A = 3i - 6j + 5k, B = 2i - j + 4k, C = i + j - k, compute
(a) A x B, (b) B x C, (c) (A + B ) x (A - B ), (d) (A x B ) x
(B x C ), (e) A x (B x C ), and express this vector as a linear
combination of B and C.
Use the vector product to compute the area of the triangle with
the given vertices:
(a) P: (1,0, 3), Q: ( 12, -1, R: (-2,1, 3)
(b) P: ( - 2 , - 3 , Q: (1, 2, -1, R: (4,3,-3)
(c) P: (4,-2,3), Q: (-3,1,1), R: (1,1,1)
(d) I ' (-1 - 3 , Q: ( 22, - 1, R: (-3,2, -2)
A force F = 3i - 2j + 5k acts on a particle at (4, -2, 2). Find the
moment or torque of the force about (a) the orign; (b) the
point (1, 2, 1).
A force F = i + 4j - 3k acts on a particle at (-3, 1, 5). Find the
torque about (a) the point (1, 0,2) ; (b) the point (-1,2, -3).
A particle at (2, 5, 0) is acted upon by the forces F = - i + 3j and
G = -2i - j. Find the total torque about (a) the origin;
(b) the point (1, 1, 0).
A particle has an angular speed of 2 rad/sec and its axis of
rotation passes through the points (1, 0, 2) and (2, 3, -1). Find
the velocity of the particle when it is located at the point (3, 4,
5) .
A particle has an angular speed of 4 rad/sec with its axis of
rotation along the vector i + j + k Find the velocity and the
speed of the particle when it is located at the point (-1, 2, 3).
If three vertices of a parallelogram are P: (-1, 1, I), Q: (2, 0, I),
R : (0, 2, 3), find all the possible points D which can be the
fourth vertex of the parallelogram.
Let A and B be two linearly independent vectors.
(a) Show that the angle between A and C = (A x B) - A lies
between n/ 2 and n.
Vector Algebra 53
(b) Show that the angle between B and D = (A x B) + B lies
between 0 and n/2.
20. Let A and B be two orthonormal vectors.
(a) Show that the set (A, B, A x B) forms an orthonormal
basis.
(b) Show that (A x B) x A = B, (A x B) x B = - A, and give the
geometric interpretation.
21. Let A = i - 2 j + k a n d C = i + j + k
(a) Find a vector B such that A x B = C. Is the vector B uni-
quely determined?
(b)
Find a vector B such that A x B = C and A* B = 1. Is this
vector uniquely determined?
22. Prove that IA x B ( = ( A I (B ( if and only if A and B are ortho-
gonal.
23. If A + B + C = 0, show that A x B = B x C = C x A.
24. If A + B + C - R = O and A - B + C - 3R=O, show that Bx R
=O, Ax R=Rx C, a n d Ax B= Bx C.
25. If A x B = C and A x C = B, show that B = 0 and C = 0.
26.
Show that
(a) if A x B = 0 and A-B = 0, then A = 0 or B = 0;
(b) i f A # 0, A x B = Ax C, a n d A. B=BoC, t henB=C.
1.8 LINES AND PLANES IN SPACE
In Sec. 1.5 we saw how vector arithmetic could be used to solve
some elementary problems in geometry without reference to any
coordinate system. Here we consider applications of the vector
operations of scalar product and vector product in the study of lines
and planes in space.
Equations of Lines
We begin with the problem of finding an equation of a line. A
line is completely determined if we know any two points on the line
or if we know a point on the line and the orientation or direction of
the line. The direction of a line can be described by means of the
direction cosines cos a, cos fJ and cos y of the line. Recall that a, fJ
and y are the angles that the line makes with the positive x-, y- and
z-axes, respectively. Equivalently, the direction of a line can also be
described by a vector to which the line is parallel, in which case, the
Chapter 1
components of the vector becomes the direction numbers of the
line. So suppose we wish to find an equation of a line that passes
through the point Po: (x, y, z, ) and is parallel to the vector A = al i
+ a2j + a 3 k Let R = x i + y j + z k denote the position vector of an
arbitrary point P: (x, y, z) on the line (Fig. 1.28). I f R, = x, i + yo j +
z, k denotes the position vector of the point Po, then the vector R -
R, is parallel to the vector A. Hence there exists a scalar t such that
R - R, = tA. Thus the position vector R of an arbitrary point on the
line is gven by
R= R, + t A, ( - m < t < a ) (1. 35)
This is a vector equation of the line. The scalar t is called a parame-
ter. As the parameter ranges from - m to m , the vector R traces the
line from one end to the other with t = 0 corresponding to the point
"0.
To obtain the equation of the line in cartesian coordinates x, y,
z, as we know it in anal
y
tic geometry, we equate the corresponding
Fig. 1.28 Equation of a line.
Vector Algebra
components of the vectors in (1.35). We obtain
These are called the parametric equations of the line. If we elimi-
nate the parameter t from the equations (1.36), we obtain the sym-
metric form of the equation of the line:
We point out that equation (1.37) can also be derived from
another vector equation of the line, namely, the equation
Notice that the numbers al, a2, a3 appearing in equations (1.36)
and (1.37) are the components of the vector A. They are called the
direction numbers of the line. Notice also that any vector parallel to
A may be used in place of A in (1.35). This means that any numbers
proportional to al , a2, a, may also be used as direction numbers of
the line.
Example 1. Find a vector equation of the line passing through
the points P: (1, -2, 1) and Q: (3, 1, 1). What are the corresponding
parametric equations and the equations in symmetric form?
Solution: We note that the line is parallel to the vector
determined by the directed line segment PQ . Let R, = i - 2j + k and
let R = x i + y j + z k denote the position vector of an arbitrary point
(x, y, z) on the line. Then by (1.35) we have
By equating the corresponding components of the vector equation,
Chapter 1
we obtain the parametric equations
By eliminating the parameter t, we obtain the equation in symme-
tric form
The line lies on the plane z = 1.
Example 2. Find the parametric equations of the line that
passes through the point (-1, 3, -2) and is perpendicular to the vec-
tors A = 3i + 4j + k and B = i + 2j.
Solution: Since the line is perpendicular to the vectors A and
B, it is parallel to the vector product of A and B. Since
it follows that the parametric equations of the line are
Example 3. Find the point of intersection of the two lines
defined by the vector equations
and determine the angle between them .
Solution: We need to find a value of the parameter t and a
value of the parameter t" such that R1 = Rz . If no such values exist,
then the lines do not intersect. Now equating the corresponding
components of the equation R1 = RZ, we find
Vector Algebra 57
Solving for f and t*, we find t = 1 and t* = 1. Thus the lines inter-
sect at the point (2, 0, 3).
The angle between the lines is equal to the angle between the
two vectors to which the lines are respectively parallel. From the
vector equations of the lines, we see that these two vectors are
A l = i + j + k and A2 = - i + 3 j + 2 k
Let 8 denote the angle between these vectors. Since
A,-A2 = 4 , I A1 1 = 43, and I A2 I = 414
we find
cos 8 = Al* A2 / I Al I I A2 I = 4/(43 414) = 4/442
Hence the angle between the lines is equal to 9 = arccos (4/d42) =
51.9O.
Now suppose we wish to find the distance from a point P : (xl ,
yl , zl) to a line defined by the vector equation (1.35). Using vector
method, this problem, which otherwise would involve long and
tedious computations, can be easily tackled as follows. Referring to
Fig. 1.29, we see that the distance d of the point P from the line is
Fig 1.29 Distance of a point from a line.
58 Chapter 1
given by d = IP$ 1 sin 0, where 0 is the angle between the given
line and the directed line segment P$. Notice that Po can be chosen
as any point on the given line. Since the line is parallel to the vec-
tor A, we have
I P g x A I = (P$IIAI si n0
Solving for I P$ I sin 8, we obtain the formula
Example 4. Find the distance of the point P: (2, 3, 6) from the
line defined by the vector equation
Solution: We choose the point Po: (3, -1, 4) on the line
corresponding to t = 0. Then POP = (2 - 3)i t (3 - (-1))j + (6 - 4)k
= - i + 4j + 2k and hence
Since 1 POP x A 1 = d77 and 1 A I = d6 , it follows from (1.39)
Example 5. Find the distance of the line defined by (1.35) from
the origin, assuming that it does not pass through the origin.
Solution: Consider the equation of the line R = Ro + tA ,
- 03 < t < 03 . We seek a point on the line whose position vector is
perpendicular to the line. The distance of this point from the origin
is the answer to the problem. Let R* denote the position vector of
such a point and suppose t = to is the value of the parameter which
corresponds to the point (Fig. 1.30). Then R* = Ro t toA. Since R* is
perpendicular to the line, it is perpendicular to the vector A. Hence,
taking the dot product, we have
Vector Algebra
which gives
t o = - A* Ro/A*A
Therefore, the position vector of the point we are seeking is
R* = R, - [A* %/A0A] A
and so the distance of the line from the origin is given by
d = IR*I
J
Fig. 1.30 Distance of a line from the origin.
Equations of Planes
We know that a plane is uniquely determined by three non-
collinear points or by a point and a line (or a vector) that is perpen-
dicular to the plane. If three non-collinear points, for example, P, Q,
and R are given, we can readily find a vector that is perpendicular to
the plane determined by the three points. The vector is simply the
60 Chapter 1
cross product of any two of the vectors represented by the directed
line segments Pa, QR, and XP. Thus it is sufficient to consider only
the case where we are gven a point and a vector. So let Po: (x,, yo,
z,) be a gven point and A = ali + azj + a3ka given vector. We
wish to find an equation of the plane that passes through the point
P, and is perpendicular to the vector A. Let P: (x, y, z) be an arbi-
trary point on the plane. Then the vector represented by the direc-
ted line segment P$ = (x - xo) i + (y - yo) j + (z -zo) k lies on the
plane, and since A is perpendicular to the plane, I POP( is orthogonal
to the vector A (Fig. 1.31). Therefore, AWP,P = 0, which yields the
equation
al(x - X& + a2(y - yo) + a3(z - zo) = (1.40)
This equation may be written in the form
Fig. 1.31 Equation of a plane.
Vector Algebra
where d = - (al xo + a2 yo + a3z, ) is a constant.
We observe that the coefficients in (1.40) and (1.41) are simply
the con~ponents of the vector A that is perpendicular to the plane.
The vector A is called a normal vector to the plane.
Example 5. Find the equation of the plane that passes
through the point (2, -3, 5) and is perpendicular to the line deter-
mined by the points (3, 1, 2) and (-2, 3, 4).
Solution: First we find the direction numbers of the line that
is determined by the points (3, 1, 2) and (-2, 3, 4). They are given by
Hence by (1.40) the equation of the plane is gven by
Example 6. Find the equation of the plane that is determined
by the points P: (- 2, 1,3), Q: (1,2, - I), and R: (- 3, - 2, 1).
Solution: Let A and B denote the vectors represented by QP
and QR, respectively. Then
Since these vectors both lie on the plane in question, it follows that
is a normal vector to the plane. Hence, using (- 2, 1, 3) as a fixed
point, we find
14(x + 2) - lO(y-1) + 8(z - 3) = 0
62 Chapter 1
The student should verify that the same equation is obtained
when either the point (1, 2, - 1) or the point (- 3, - 2, 1) is used for a
fixed point. Alternatively, the equation of the plane may be derived
by substituting the coordinates of the points P, Q, and R in (1.41),
thus obtaining a system of three equations which can be solved for
the unknowns al , a> a3.
We now derive a formula analogous to (1.39) for the distance
from a gven point to a given plane. Let P: (xl, yl,zl) be a given point
and let the plane be defined by the equation (1.40):
A. ( R- R,) = 0
We note that the vector A is perpendicular to the plane and the
plane passes through the point
Po: (x,,y,, ,z,) whose position vector
is R,. Referring to Fig. 1.32, we see that the distant from the point P
to the plane is given by d = I P$ I cos 0, where 0 is the angle
between the normal vector A and the vector represented by POP.
Fig. 1.32 Distance from a point to a plane.
Vector Algebra
Since A .P,P = IA 1 IPoP I cos 0, we obtain the formula
Example 7. Find the distance from the point (2, 4, 5) to the
plane
2x- y +2z + 6 = 0
Solution: First we pick a point Po on the plane. This is easily
done by assigning (arbitrary) values to any two of the variables x, y,
z and then solving for the remaining variable. If we set x = -1, y = 2,
we find z = -1. Thus the point Po: (- 1, 2, - I ) lies on the plane, and
P$ represents the vector
From the equation of the plane, we deduce the normal vector
A = 2i - j + 2 k Hence by (1.42) we find
1.6 EXERCISES
1. Find a set of parametric equations of the line that passes
through point (-1, 3, 2) and is parallel to the vector A = 2i - 3j
+ 5 k
2. Find the symmetric equations of the line that passes through
the points (3, -1, 4) and (-2, 3, 5).
3. Find the parametric equations of the line that passes through
the point (1, - 2, 1) and is perpendicular to the plane 2x -3y + 42
= 5.
4. Find the parametric equations of the line that passes through
the point (3, 4, 6), perpendicular to the vector i + 2j - k and
parallel to the plane 2x - 3y + 5z + 4 = 0.
5. Find an equation of the line of intersection of the planes x - 2y
+ ~ - 3 = O a n d 2 ~ + 3 y - 5 z - 4 = 0 .
64 Chapter 1
6. Find the point of intersection of the lines defined by
and
R2=2i + j + 2 k + ( i - 3j + k)t *.
What is the cosine of the angle between these lines?
7. Find the point of intersection of the lines represented by
x - 2 - y + 4 - 2 - 3
- - - - . - -
x - 2 y - 5 2 - 3
- - and - - - -
1 2 1 2 1 -2-
and determine the cosine of the angle between them.
Find the point of intersection of the line
and the plane 3x + 2y - 22 - 5 = 0.
Find an equation of the plane that passes through the point
(- 2, 3, 1) and contains the line R = i + 2j + k + ( 3i - j + 2k)t .
Find an equation of the plane that passes through the points
(1, - 1, I), (- 2, 3, 4), and (- 3, -2, 1).
Find an equation of the plane that contains the two lines given
in Problem 6.
Find an equation of the plane that contains the two lines given
in Problem 7.
Find an equation of the plane that passes through the point
(- 1, 2, 3) and is perpendicular to the line of intersection of the
planes x + 2y - 32 + 4 = 0 and 3x - 2y + 52 - 2 = 0.
Find the distance of the point (2, - 2, 3) from the line
Find the distance of the line R = 2i - j + k + (3i - j +4k)t
from the origin.
Find the distance of the plane 3x - 2y + 52 - 2 = 0 from the
origin.
Vector Algebra
17. What is the distance of the point (1,3,2) from the plane given
in Problem 16?
18. Find the distance between the line that joins the points (1, -2,3)
and (2,4, -1) and the line that joins the points (3,4, -2) and (2,
-1,5).
19. What is the distance from the point (1, -2, l ) to the plane deter-
mined by the three points (2, -2, -I), (3, - 4, -3), and (-1,2, O)?
20. Find the distance between the skew lines
x = - 1 + 4 t , y = 2 - 3 t , z = - 3 + 2 t
and
~ = 1 + 3 t , y = - 3 + t , z = - 2 t ( - - < t < = )
21. Find the center and the radius of the sphere that is tangent to
the planes x - 2y + z = 4, x - 2y + z = 10 and whose center lies on
the planes x - 2y = 2 and y - z = 3. Determine the points at
which the sphere is tangent to the given planes?
1.9 SCALAR AND VECTOR TRIPLE PRODUCTS
In this section we study the results of applying the operations
of scalar and/ or vector product on three vectors. Apart from the
or- der in which the vectors appear, the three types of product for
three vectors A , B , C are of the form
(i) (A B)C , (ii) A (B x C) , (iii) A x (B x C)
The first type is merely the product of the scalar A B and the
vector C, hence it is a scalar multiple of C. The second type is the
scalar product of A and B x C, which results in a scalar. This
product is called the scalar triple product of the vectors in the order
they appear, and it is denoted by (ABC). Thus
[Notice that in writing A B x C, there is no ambiguity as to which
operation is to be performed first, since the alternate form (A B ) x
C is meaningless. Why?]
If A= a l i + a 2 j + a 3 k , B = b l i + b 2 j + b 3 k and C = c l i + c 2
j + c3 k, then in terms of the components of the vectors, we have
Chapter 1
Thus the scalar triple product (ABC) is simply the determinant of
the matrix whose elements in the first, second and third rows are
the components of the vectors A, B, and C, respectively.
A- Bx C = A-
Example 1. Find the scalar triple product (ABC) if A = 3i + j -
2 k , B= 2 i - j - 3 k , a n d C= 4 i + 2 j + k.
Solution: By (1.44) we find
i j k
b, b2 b3
c1 c2
Since the determinant of a matrix changes sign whenever any
two of its rows are interchanged, we see that if we interchange in
(1.44) the first row and the second row, and then the second row and
the third row, the determinant remains the same. Thus, we obtain
If we repeat the process on the the scalar triple product B- C x A , we
obtain
Therefore, the scalar triple product of three vectors has the property
Vector Algebra 67
Thus in the expression A B x C, the two operations " . " and " x "
are interchangeable, that is, A B x C = A x B . C.
The scalar triple product (1.44) provides a convenient way for
determining whether or not three given vectors are coplanar; that
is, the vectors can be represented by arrows lying on the same plane.
The method is contained in the following theorem.
Theorem 1. Three vectors A, B, C are coplanar i f and only if
(ABC) = 0.
Proof: Clearly, the theorem is trivial if one of the vectors is
the zero vector. So suppose (ABC) = 0 and that none of the vectors
is 0. By (1.45) it follows that one of the vectors must be orthogonal
to the cross product of the other two vectors. Suppose A B x C = 0.
Then A is orthogonal to B x C. Since B x C itself is orthogonal to B
and C, we deduce that A lies on the plane determined by B and C.
This means that the vectors are coplanar.
Conversely, if the vectors are coplanar, then we can express
one of the vectors as a linear combination of the other two. Suppose
A is a linear combination of B and C. Since B x C is orthogonal to B
and C, it is also orthogonal to A, and so A . B x C = 0.
Geometric Interpretaion of Scalar Triple Product
It is interesting to note that the scalar triple product of three
nonzero vectors represents in absolute value the volume of the
parallelepiped determined by the vectors. To see this, let us consi-
der the parallelepiped formed by the vectors A, 8, and C as shown in
Fig. 1.33. From the definition of scalar product, we have
where 0 is the angle between the vectors A and B x C. Now from
the definition of vector product, we know that
1 B x C I represents
the area of the parallelogram formed by the vectors B and C.
Further, the quantity h = [ A 1 cos0 represents numerically the alti-
tude of the parallelepiped relative to the base formed by the vectors
B and C. Since the volume of a parallelepiped is equal to the pro-
duct of a base area and its altitude, it follows that the numerical
Chapter 1
B x C
Fig 1.33 Geometric interpretation of the scalar triple product.
value of (1.44) represents the volume of the parallelepiped deter-
mined by the vectors.
It is geometrically evident that (ABC) is positive or negative
according to whether the vectors A, B, C form a right-handed or a
left-handed triple. When (ABC) # 0, the vectors are said to be linear-
ly independent; otherwise, the vectors are said to be linearly depen-
dent. It follows that three nonzero vectors are linearly dependent if
and only i f they are coplanar. In the case of two vectors, linear de-
pendence implies that the vectors are scalar multiple of each other.
Hence two nonzero vectors are linearly independent if and only if
the vectors are not parallel.
Example 2. Let P : (1, -2, 3), Q : (2, 1, -2), R : (-2,1, -I), and
S : (2, 2, 3)
be four given points. Find the volume of the parallele-
piped formed by PQ, PR, and PS.
Solution: Let A, B, and C denote the vectors represented by
P(2, PR , and PS , respectively. Then we have
Vector Algebra
Thus the volume of the parallelepiped is equal to
[Note that if (ABC) turns out to be negative, we simply take the abso-
lute value for the volume.]
Example 3. Show that the four points P : (2, 4, 3), Q : (-2, 1, 4),
R : (1, 2, 5), and S : (-5, 0, 3) are coplanar.
Solution: The four points are coplanar if and only if the paral-
lelepiped determined by the points degenerate into a planar figure,
in which case the volume is then zero. So let A, B, and C denote the
vectors represented by PQ, PR, and PS . We have
A = 4i - j + k, B = - i - 2j + 2k, and C = - 7i - 4j . Since
it follows that the points are coplanar.
Finally, let us consider the product A x (B x C). This is called a
zlccfor triple product . From the definition we know that this vector
is orthogonal to the vectors A and B x C. Since B x C is in turn
orthogonal to the vectors B and C, it follows that A x (B x C ), B, and
C are coplanar or linearly dependent (Fig. 1.34). Hence, we can ex-
press the vector A x (B x C ) in terms of B and C, that is,
for some scalars nz and n. We will show that
so that
A x (B x C ) = (CeA)B - (BeA)C (1.46)
This formula is sometimes referred to as the "CAB minus BAC"
formula for the vector triple product A x (B x C) .
Chapter 1
Fig. 1.34 Vector triple product.
We establish the formula (1.46) by showing that the correspon-
ding components of the vectors on opposite sides of the equation are
equal. So suppose A = a l i + a2j + a 3 k B = b l i + b2j + b 3 k and C =
cl i + c2j + c3 k. We note that
Thus the first component of the vector A x (B x C ) is given by
On the other hand, the first component of the vector on the right-
hand side of (1.46) is given by
Vector Algebra
which agrees with the first component of the vector A x (B x C). The
corresponding second and third components of the vectors can also
be shown to be equal in exactly the same manner, thus establishing
the formula (1.46).
Since A x (B x C) = - (B x C) x A , it follows that
which shows that the vector triple product (A x B) x C is also a linear
combination of the vectors inside the parentheses.
Example 4. Verify the formula (1.46) for the vectors
A = i - j t 2k, B = i - 2 k a n d C= - 2 i + j - k .
Solution: We have
hence
On the other hand,
A. C = I(-2) + (-1)1 + 2(-1) = - 5
A. B = 1(1) + (-1)O + 2(-2) = - 3
so that
which equals the vector A x (B x C), thus verifying the formula
(1.46).
Chapter 1
Example 5. Prove the identity
for any vectors A and B.
Solution: We observe that
Set D= A x B and consider the scalar triple product D. (A x B). By
the property of scalar triple product, we see that
Since D x A = (A x B) x A = (A.A)B - (A- B)A , it follows that
1.7 EXERCISES
1. Compute the scalar triple product (ABC) if
(a) A = 2 i + j , B = - i + j + 2 k , C = i + j + 2 k
(b) A = - i + j + 2 k , B = i + 2 j , C=2 i - j + 4 k
(c) A= 2 i + j + 3 k , B= - 3 i + 2 k , C= 2 i - j + 4 k
(d) A= 3 i + j + 2 k , B= 2 i + 5 k , C = i + 6 j + 3 k
2. Find the volume of the parallelepiped determined by the
vectors i + j , j + k , i + k.
3. Find the volume of the parallelepiped determined by the
line segments P(Z, PR, and PS, where P: (1, 2, O), Q: (3, 5, O),
R: (4, 3, O), and S: (- 1, - 1,2).
4. Repeat Problem 3 if the points are P: (2, 0, 2), Q: (4, 3, -I),
R: (0,2, 0) and S: (3, 1, 4).
5. Showthatthevectors A= 2 i - j + 4 k f B= i + 2 j - 3 k , a n d C=
- 4i + 2j - 8k are linearly dependent and determine which
of the vectors is a linear combination of the other two.
6. Show that A = i + j , B = j + k , C = i + k are linearly indepen-
dent and express D = 4i - j - k as a linear combination of these
vectors.
Vector Algebra 73
Repeat Problem 6 if A = 2i + j + k , B = i + 2j + k , C = 3i +
j + 2 k , a n d D= - i + 5 j .
Show that the vectors A = i - 2j + 2k and B = 2i + j - 3k
are linearly independent, and express D = 4i - 3j + k as a
linear combination of these vectors.
Repeat Problem 8 if A = 2i + 2j - 3k, B = i - 2j + k , and
D = 7i - 2j - 3k.
Express A = 3i + 5j as a linear combination of B = - i + j
and C=3 i + j.
Express A = -9i - 2j as a linear combination of B = i + 2j and
C = 3i - 2j.
Show that A = 2 i + j - 3 k , B = i + 2j + k , and C = 2i + 3j -
2k are linearly independent.
Let U = 2 A - B + C , V = A + B - 2 C , a n d W = A - 2 B +
3C, where A = i - 2 j + 3 k , B = 2 i + j + 2 k , C = - i + j + k
Show that the vectors U, V, W are linearly independent.
L e t U = A - B , V = B + 2 C , a n d W= 2 A - B + C , w h e r e
A = i + j - k , B = 3 i - j + 2 k , C=i - 2 j +3 k . Sh o w t hat U, V, and
W are linearly independent.
Le t A= i - 2 j - k , B= 2 i - j + 3 k , a n d C= i + 2 j + 4 k Determine
in two ways (a) A x (B x C) and (b) (A x B) x C.
Prove that (A x B) x (C x D) = (ABD)C - (ABC)D = (ACD)B
- (BCD)A.
Prove that (A x B). (B x C) x (C x A) = (ABC)~.
Prove that Ax ( Bx C) + Bx ( Cx A) + C x( AxB) =O.
Prove that D x (A x B). (A x C) = (ABC)(Aa D).
Show that A x (B x C ) = (A x B) x C if and only if B x (C x A)
= 0 .
Prove the Lagrange identity
(A x B) (CxD) = (A -C)(B - D) - (A. D)(B . C)
(Hint: Consider A [B x (C x D)].)
DIFFERENTIAL CALCULUS OF VECTOR
FUNCTIONS OF ONE VARIABLE
In this chapter we shall extend the vector algebra developed in
the previous chapter to vector functions of a real variable. We shall
introduce the idea of a derivative of a vector function and consider
its geometrical as well as physical significance. Beginning with Sec.
2.8 and through Sec. 2.10, we shall discuss curvilinear motion of a
particle in various coordinate systems.
2.1 VECTOR FUNCTIONS OF A REAL VARIABLE
Let us first recall the definition of a real-valued (scalar) func-
tion in a domain D. We note that a domain D is an open set of
points in which any two points can be joined by line segments lying
entirely in the set. Examples are the set of points inside a sphere or a
parallelepiped in three dimensional space, the set of points inside a
circle or a rectangle in two dimensional space, and the set of points
in the interval a < x < b in one dimension. Now we recall that a
real-valued function defined in a domain D is a mapping or a rule f
that assigns to each point P in D a unique real number f(P) from a set
R. We call D the domain of definition of f and R the range of f. In a
similar manner, we define a vector-valued function or simply a
vector function in a domain D as a rule F that assigns to each point
P in D a unique vector denoted by F(P). In other words, a function
defined in a domain D whose range is a set of vectors is called a
vector function. In this chapter, we are concerned with vector
functions defined in an interval which may contain one or both
Chapter 2
endpoints or which may be infinite. Such functions are called vector
functions of a real variable. The real variable is usually denoted by t,
which indicates time in many applications. We denote vector func-
tions by bold-faced letters F, G, etc., and the value of a function F at
t by F(t).
Following standard notations, we denote the open interval a <
t < b by (a, b) and the closed interval a s t s b by [a, b]. The semi-
infinite interval - 0 < t s a is denoted by (-0, a] and the interval a s t
< a by [a, 0) .
Now let F be a vector function defined in the interval [a, b].
For each t E: [a, b], F(t) is a vector and hence it is uniquely determined
by its three components, say, fl(t), f2(t) and f3(t), with respect to a
given rectangular cartesian coordinate system. Thus we can asso-
ciate to each vector function F three scalar functions fl(t), f2(t) and
f3(t), all defined in the same interval as F, whose values at t consti-
tute the components of F(t). Therefore, we can represent each vector
function F as
where the scalar functions fl(t), f2(t) and f,(t) are called the compo-
nents of F(t). This provides an analytic representation of a vector
function of one variable. The representation (2.1) is sometimes also
written as F(t) = [fl(t), f2(t), f3(t)].
Example 1. The equation
of the line that passes through the point R,, and is parallel to the
vector A is a vector function of the parameter t. For each value of t,
the function defines a vector which is the position vector of a point
on the line.
Example 2. The vector function
maps each value of t, 0 s t < 2x, to a vector in the xy-plane. If we
Vector Functions of One Variable
represent the vector by an arrow from the origin, then as t ranges
over [O, 2m), the arrow traces the unit circle (Fig. 2.1). In this case the
parameter t denotes the polar angle that the vector makes with the
positive x-axis.
Fig. 2.1 The unit circle.
2.2 ALGEBRA OF VECTOR FUNCTIONS
We now extend the vector algebra developed in Chapter 1 to
vector functions. We do this by expressing the vector functions in
terms of their components and then performing the operations on
the components. First, let F be a vector function defined in [a, b],
and let fl , f2, f3 denote its components. Then, for each t E [a, b], the
magnitude of F(t) is given by
This is a scalar function of t in a s t 5 b. For some vector functions,
however, the magnitude may be a constant. For example, the mag-
nitude of the vector function F(t) = a(cos t i + sin t j) is given by
IF(t) I = a, which is a constant .
Now let F(t) = fl(t)i + f2(t)j + f,(t)k and G(t) = gl(t)i + g2(t)j +
g3(t)k, a s t 5 b, be two vector functions, and h(t) a scalar function
all defined in the same interval [a, b]. Then the sum of F and G, the
Chapter 2
h scalar multiple of F, the scalar and vector products of F and G are
defined as follows:
and
Notice that the sum F + G, the scalar multiple hF, and the vector
product F x G are all vector valued, whereas the scalar product
Fa G is real valued.
The concepts of linear dependence and orthogonality of vec-
tors can likewise be carried over to vector functions. For example,
we say that the functions F, G, H defined in the same interval [a, b]
are linearly dependent if
for all t E [a, b]. Otherwise, the functions are said to be linearly inde-
pendent in [a, b]. Similarly, we say that the functions F and G are
orthogonal in [a, b] if F(t). G(t) = 0 for all t E [a, b].
Example 1. If F(t) = t i + (1 - t
2
) j + cos t k and G(t) = sin t i +
In t j + t
2
k for t > 0, then we have
F(t) + G(t) = (t + sin t)i + (1 - t
2
+ In t)j + (cos t + t?k
F(t)- G(t) = t sin t + (1 - t? In t + t
2
cos t
Vector Functions of One Variable
and
= [t
2
(1 - t
2
) - (cos t) In t]i + (sin t cos t - t3)j + [t In t - (1 - t
2
)sin t]k
G(t)=
The magnitude of F and G are given by
t 1- t L cost
sin t In t t
2
2 4 2
IF(t)I = \ 1 - t + t + cos t
and
- --
2 2 4
IG(t)I = \ sin t + In t + t
Example 2. Show that the vector functions F(t) = a(cos t i +
sin t j) and G(t) = b(sin t i - cos t j), 0 s t < 2 x , are orthogonal.
Solution: We show that the scalar product of F(t) and G(t) is
zero. We have
F(t). G(t) = ab(cos t sin t - sin t cos t) = 0
for all t in [O, 2x1. Thus the vector functions are orthogonal.
2.1 EXERCISES
1. Let F ( t ) = t i + ( l + t ? ) j - s i n t k a n d G( t ) = l n ( l + t ) i - e t j + k
for t 2 0. Find (a) 2F - G , (b) Fa G , (c) F x G , (d) ( F I and ( G 1 .
2. Let F(t) = (3t - l ) i + (t
2
- 2)j + t k and G(t) = 2e
t
i - cos t j + t
2
k
for t r 0. Find (a) 3F - 2G, (b) F- G, (c) F x G , (d) IF + GI .
3. Let F(t) = t cos 2t i + t sin 2t j + et k (t z 0). Find a scalar func-
tion h such that ( h(t)F(t) ( = I for all t 2 0.
4. Repeat Problem 3 i f F(t) = e-t cos t i + e-t sin t j t et k (t r 0).
5. Let F(t) = a (cos t i + sin t j) + bt k and G(t) = - sin t i + cos t j
(0 s t < 2n ), where a and b are constants. Show that F and G
are orthogonal for all t in [0, 2x). Find a unit vector function
80 Chapter 2
that is parallel to F x G for 0 s t < 2x.
6. Repeat Problem 5 if F(t) = (cos t - sin t) i + (sin t + cos t) j + k
and G(t) = - (cos t + sin t) i + (cos t - sin t) j ( 0 < t < 2n ).
7. Let F(t) = t i + t
2
j + 9 k , G(t) = sin t i + cos t j +e
t
k, a nd
H(t) = (t + 1) i - 2t j + et k , t 2 0. Verify that (a) F(t)* G(t) x H(t)
= F(t) x G(t) H(t) and (b) F(t) x [G(t) x H(t)] = g(t) G(t)
- h(t) H(t), where g(t) = F(t). H(t) and h(t) = F(t)- G(t) for all t.
8. Repeat Problem 7 if F(t) = t
2
i - I ? j , G(t) = et i + cosh t j , and
H(t) = t i - In t k (t > O), and express [F(t) x G(t)] x H(t) as a
linear combination of F(t) and G(t) for each t > 0.
9. Let U(8) = cos 8 i + sin 8 j and V(@) = cos @ i + sin @ j, where
both 8 and @ range over [ 0, 2n ). (a) Show that U and V are
both unit vector functions; (b) by considering the dot product
U* V, obtain the addition formula
2.3 LIMIT, CONTINUITY, AND DERIVATIVES
The definitions of limit, continuity and derivatives for vector
functions bear strong resemblance to the corresponding definitions
for scalar functions. As a matter of fact, we may introduce the defi-
nitions of limit, continuity and derivative via the components of a
vector function since these components are real valued functions.
Consequently, many of the theorems on scalar functions concerning
limits, continuity and derivatives can be carried over to vector func-
tions in an almost straightforward fashion. We begin with the
notion of a limit.
Limit of a function
Definition 1. Let F be a vector function defined in a 5 t < b, to
a point in this interval, and A a constant vector. We say that the
limit of F(t) as t approaches to is A, written as
lim F(t) = A
t --> t
(2.7)
0
if and only if for any given number E > 0, there is a number 6 > 0
Vector Functions of One Variable
such that
I F(t) - A I < E whenever 0 < 1 t - to ( < b
(2.8)
This definition is analogous to the corresponding definition of
limit for a scalar function. Note, however, that (2.8) refers to the
magnitude of the vector F(t) - A, whereas in the scalar case it refers
to the absolute value of the difference between the scalar function
and its limit. Geometrically, the definition implies that relative to a
fixed point 0 the terminal point of the geometric vector F(t) lies
inside the sphere of radius E and center at the terminal point of A
whenever 0 < I t - to 1 < 6 (Fig. 2.2).
In terms of components, if F(t) = f,(t) i + f2(t) j + f,(t) k and A =
al i + a2 j + a, k , then (2.7) is equivalent to
lim fi(t) = ai (i = 1, 2, 3) (2.9)
t --> to
Thus we have
I I I
t o- 6 t o t o+ b
*
Fig. 2.2 Limit of F(t) as t t ends t o to.
82 Chapter 2
if and only if (2.9) holds. As a result of this, many of the properties
of limits for scalar functions also hold for vector functions. We state
these properties as a theorem.
Theorem 1. Let F and G be vector functions and h a scalar
function with a common interval of definition. Suppose that
lim F(t) = A , lim G(t) = B , lini h(t )=nz
t --> to t --> to t --> to
where A and B are constant vectors and nz is a scalar. Then we have
(a) lim [F(t) + G(t)] = A + B
t --> to
(b) lim F(t) G(t) = A. B
t --> to
( 4 lim F( t ) x G( t ) =Ax B
t --> to
(d) lim h(t)F(t) = nzA
t --> t,,
These properties can, of course, be proved from the definition
(2.7) and (2.8) or, more easily, by expressing the vector functions in
terms of their components and then using the equivalent definition
(2.9).
Continuous Vector Functions
Definition 2. A vector function F is said to be continuous at a
point to in its interval of definition if and only if
lirn F(t) = F(t,)
t --> t
0
If (2.12) holds for all points in its interval of definition, then F
is said to be continuous in that interval. It follows from (2.10) that a
vector function is continuous in an interval if and only if its
components are continuous in that interval. Moreover, from Theo-
rem 1, we see that if F, G and h are continuous in a common inter-
Vector Functions of One Variable 83
val of definition, then so are the functions F + G, F . G, F x G and hF.
Example 1. The following vector functions are continuous in
the indicated interval:
(a)
F(t) = A sint t + B cos t , 0 I t < 2x, A and B are constant
vectors.
@) G( t ) =( l / t ) i + ( t 2- l ) j + e
t
k , t > O
(c) H(t) = e-t sin t i + t In t j + cos 2t k , t > 0
Derivative of Vector Functions
Definition 3. Let F be a vector function defined in [a, b]. The
derivative of F at a point t in [a, b], denoted by F1(t), is defined as the
limit
F(t + At) - F(t)
F
f
( t ) = lim
A t - 0 At
provided the limit exists. I f Ff(t) exists for all t in (a,b), then we say
that F is differentiable in (a, b).
In terms of components, i f F(t) = f,(t) i + f2(t) j + f3(t) k then
(2.13) may be written as
f,(t + At) - fJt) f2(t + At) - f2(t)
F' (t) = lim i +
At -s 0
At
i
Therefore, i f the components f , , f 2, f3 are all differentiable, then we
have
Thus a vector function F is differentiable in an interval i f and only if
84 Chapter 2
its components are all differentiable in that interval. Since a scalar
function is continuous whenever it is differentiable, it follows that
F(t) is continuous in its interval of definition whenever F
t
(t) exists
in that interval.
The rules for differentiating vector functions are similar to
those for scalar functions. We state these rules as a theorem.
Theorem 2. Let F , G, and h be differentiable functions in (a,
b). Then, in the same interval, we have
(a)
[F(t) + G(t)ll = F1(t) + G' (t)
(b) [h(t)F(t)ll = h(t)F
1
(t) + h
t
(t)F(t) (2.1 5)
(c)
[F(t)* G(t)]' = Ft (t). G(t) + F (t). G1 (t)
(d)
[F(t) x G(t)ll = F1(t) x G(t) + F(t) x G' (t)
These rules are easily established by expressing the vectors in
terms of their components and then applying the rules for differen-
tiating scalar functions. Observe that although (2.15d) is similar to
the ordinary formula for the derivative of product of scalar func-
tions, the order of the factors must be maintained since cross pro-
duct of vectors is not commutative.
Formula (2. 15~) leads to an important result which is used a
great deal in our later discussion. We state this result as a theorem.
Theorem 3. Let F be a vector function that is differentiable
in (a, b). If F(t) has constant magnitude for all t in (a, b) (that is,
IF(t) I = const.), then F(t) and F1(t) are orthogonal in (a, b).
Proof: To prove the theorem, we need to show that F(t)- F
t
(t) =
0 for all t in (a, b). Now, since 1 F(t) I is constant in (a, b), we have
I F(t) 1 = F(t) F(t) = constant, for all t in (a, b). Talung the derivative
of both sides of the equation, we find by (2.15~)
for all t in (a, b). This implies that F(t) and F1(t) are orthogonal in
(a, b).
As an example, consider the function F(t) = cos t i + sin t j,
0 s t < 271. Since I F(t) I = 1 for all t in [0, 2x), F(t) and F1(t) are ortho-
gonal on [0, 2n). Indeed, we find F1(t) = - sin t i + cos t j so that F(t).
F
t
(t) = 0 readily follows.
Vector Functions of One Variable
Example 2. Let F(t) = In t i + cos t k and G(t) = t
2
j + et k for
t > 0. Find the derivative of F(t). G(t) in two different ways.
Solution: First, we observe that
~ ( t ) . G(t) = et cos t
Therefore,
[F(t ) G(t)lt = et cos t - et sin t
Alternatively, by (2.15c), we also find
[ F(t).G(t)]' = [(1 /t) i - sin t k] [ t
2
j + et k]
+ [In t i + cos t k] . [2t j + et k]
= et (cos t - sin t)
Example 3. Let F(t) = t2i + et k and G(t) = t i + sin t j + k Find
the derivative of F(t) x G(t) in two different ways.
Solution: First, the cross product of F(t) and G(t) is given by
[F(t) x G(t)lt = (- et sin t - et cos t)i + (t et + et - 2t)j
+ (2t sin t + t
2
cos t)k
F( t ) xG( t ) =
On the other hand, by (2.15d) we find
Hence
i j k
t 2 0 e t
t s i n t 1
and
t 2
=- e s i n t i + ( t e t - t S j + t s i n t k
Ft ( t ) xG( t ) =
i j k
2t 0 e t
t sin t 1
t t
= - e s i n t i + ( t e - 2 t ) j + 2 t s i n t k
t t 2
= - e c o s t i + e j + t c o s t k F( t ) xGt ( t ) =
i j k
t 2 0 et
1 cost 0
86 Chapter 2
The sum of these last two expressions gives the derivative of F(t) x
G(t) which agrees with the preceding result.
2.2 EXERCISES
In Problems 1 through 4, let F(t) = A sin t + B cos t (0 5 t < 2n ),
where A = i - j and B = i + j. Evaluate the following limits.
1. lim F(t)
t ---> n/ 3
4. lim ( tan t)F(t)
t ---> X/3
In Problems 5 through 8, let F(t) = t i + t
2
j + t 3k and
G(t) = 2t i - j + 3t" (t t 0). Evaluate the indicated limits.
5. lim I F(t) + G(t) 1
t --> 1
7. lirn F(t) x G(t)
t --> 1
8. lim 1 F(t) x G(t) 1
t -->l
9. In each case, find the derivative of F(t). G(t) and F(t) x G(t):
(a) F(t) = t i + j - 2t
2
k and G(t) = sin t i + cos t j + t k (t s 0)
@) F(t) = i - t2j + t k and G(t) = t i t (1 + t) j + (t
2
+ 1) k (t r 0).
(c)
F(t) = (i - j + k ) sin t + (2i - j - k) cos t and
G(t) = (2i + j)e
t
- (j + 2k) In t (t > 0)
10. Let F(t) = a(cos wt i + sin wt j) + b k (0 s t < 2x/w), where a,
b and o are positive constants. Show that F satisfies Theo-
rem 3.
Vector Functions of One Variable 87
11. Let F(t) = a sin 2t i + a(1 + cos 2t) j + 2a sin t k (0 s t < n ).
Find the derivative of
I F(t)- F '(t) I .
12. Let
Show that the angle between F(t) and F1(t) is independent of t.
13. Let F(t) = Ae + Be- Wt, where A and B are constant nonzero
vectors and w is a scalar. Show that F(t) satisfies the equation
FH(t) - 0 2 ~ ( t ) = 0.
14. If F(t) is a vector function which is twice differentiable, show
that [F(t) x F1(t)]' = F(t) x F "(t).
15. Let F(t) be a differentiable vector function, and set F(t) =
IF(t) 1 . Show that F1(t). F(t) = F1(t)F(t).
6 I f G(t) = F(t). Ft(t) x FU(t), show that G'(t) = F(t). F1(t) x FU' (t ).
17. Let F(t) = cos wt i + sin wt j (0 5 2n/w). Show that F(t) satisfies
the equation F"(t) + w2 ~ ( t ) = 0.
18. Let F(t) = A + B f(t), where A and B are constant vectors and f
is a twice differentiable scalar function. Show that F1(t) x FM(t)
= 0.
19. Let R(t) = A cos wt + B sin wt, where A, B, co are constant.
Show that R(t) x R1(t) = o A x B.
20. Let r(t )=R(t )/ IR(t )I. Showt hat r x d r = (R xdR) / I RI 2.
21. Show that if R(t) x Rt(t) = 0, then R(t) has a constant direction.
22. Show that i f F1(t) = w x F(t) and G'(t) = w x G(t) for a s t s b,
then [F(t) x G(t)]' = w x [F(t) x G(t)] , where w is a constant
vector.
2.4 SPACE CURVES AND TANGENT VECTORS
The study of curves and surfaces in space is greatly facilitated
by the use of vector calculus. Beginning in this section, we consi-
der some of the basic geometric properties of space curves and their
applications in the study of curvilinear motions.
Chapter 2
Equations of Space Curves
We recall that a curve in space may be defined as a set of points
(x, y, z) determined by three equations of the form
The functions x(t), y(t), and z(t) are assumed to be continuous func-
tions of the real variable t in some interval [a, b]. As in the case of
the equation of a straight line, the equations (2.16) are called parame-
tric equations of the curve, and t is called the parameter. To obtain a
vector equation for the curve, we consider the position vector R(t) of
each point on the curve corresponding to the parameter t. Since the
components of R(t) are precisely the coordinates of the point, it
follows that
R( t ) =x ( t ) i +y ( t ) j +z ( t ) k , ( a s t s b ) (2.1 7)
Thus the vector equation of a space curve is a vector function of a
Fig. 2.3 A space curve.
Vector Functions of One Variable
real variable t defined in some interval a s t s b .
Conversely, suppose R(t) is a vector function that is defined
and continuous in [a, b]. If we regard R(t) as the position vector of a
point for each t E [a. b], then as t ranges over [a, b] the set of points
determined by R(t) describes a curve in space (Fig. 2.3).
Therefore, just as a scalar function of one variable, y = f(x), (c s
x s d) may be represented graphically by a plane curve, so a vector
function R = R(t) (a s t s b) of a single variable may be represented
graphically by a space curve. Of course, i f z = 0 identically, then R(t)
= ~ ( t ) i + y(t) j represents a curve on the xy-plane. By eliminating the
parameter t from the equations x = x(t), y = y(t), it may be possible to
obtain an equation of the curve in the form y = f(x).
For example, the vector function
R(t) = (1 + a cos t)i + (1 + a sin t)j , ( 0 5 t < 2n)
represents a circle of radius a and center at the point (1, I ) (Fig. 2.4).
If we write the equation in the form
R(t) = (i + j) + a(cos t i + sin t j)
= R, + R* (t)
we see that the position vector of each point on the circle is the
vector sum of the position vector R, of the center of the circle (rela-
tive to the origin) and the position vector R*(t) of the point relative
+ sin t
) = R,
7
0
Fig. 2.4 Vector equation of a circle.
90 Chapter 2
to the center of the circle. By eliminating the parameter t from the
parametric equations
we obtain
(x - 112 + (y - 1) 2 = a2
which is the standard equation of the circle in cartesian coordi-
nates.
Orientation of Space Curves
A curve that is represented in parametric or vector equation
can be given one of two possible directions with respect to the para-
meter in a natural way. There is the direction in which the curve is
traced as the parameter t increases from a to b in the interval a 5 t 5
b. It is customary to call this direction the positive direction on the
curve; the opposite direction is then called the negative direction.
A curve on which a direction has been designated is said to be
oriented. Graphically, we denote the orientation of a curve by an
arrowhead on the curve. Thus, for example, the circle in Fig. 2.4 has
a counterclockwise orientation. This means that the positive direc-
tion is counterclockwise, which is the direction in which the circle is
traced as t increases from 0 to 2 ~ .
We should point out here that the parametric representation of
a space curve is not unique. This means that it is possible, if we so
desire, to change the parameter t in (2.16) or (2.17) to another para-
meter, say s, by setting t = f(s) where f is any differentiable function
such that f ' + 0, without changing the curve itself. The condition
f'+ 0 ensures the existence of the inverse function f of f so that the
correspondence between the two parameters s and t is one-to-one. If
f is an increasing function of s, that is, f ' > 0, then under the change of
parameters the original orientation of the curve is preserved. On the
other hand, if f ' < 0, the orientation of the curve will be reversed. For
example, if we set t = -s in the equation for the circle shown in
Fig. 2.4, the orientation of the circle becomes clockwise.
Tangent Vector on a Space Curve
We now consider the geometric significance of the derivative
Vector Functions of One Variable 91
of a vector function with respect to the curve it represents. Let C be
a space curve represented by the equation
where the functions x(t), y(t), and z(t) are continuously differentiable
in a c t c b. From the definition in Sec. 2.3, we recall that
d lim AR(t)
%R(t) = At-->0 At
where we have set AR(t) = R(t + At) - R(t). By assumption, this limit
exists. Let P and Q denote the points on the curve whose position
vectors are given by R(t) and R(t + At), respectively (Fig. 2.5). Then,
as At --> 0 the point Q tends to the point P along the curve and the
line passing through P and Q approaches a limiting position at P
which is the position of the tangent line to the curve at P. Hence, in
the limit, the vector represented by the derivative (2.19) (assuming
R'(t) + 0) is tangent to the curve at P and points in the (positive)
direction in which the curve is traced as t increases (see Fig. 2.5).
Thus the derivative of a vector function is related to the notion of
Fig. 2.5 The tangent vector.
92 Chapter 2
tangency as in the case of a scalar function.
We call the vector represented by dR(t)/dt a tangent vector to
the curve at the point P corresponding to a value of the parameter t.
The vector
is called the unit tangent vector. In terms of its components, this
unit tangent vector is gven by
x' (t) i + y' (t) j + z' (t) k
t = - - - - (2.21)
2 2 2
I' x' (t) + y' (t) + z' (t)
Knowing the tangent vector at a point of a curve, it is then easy
to write down the equation of the tangent line to the curve at that
point by the method of Sec. 1.8. In fact, if to is the value of the para-
meter that corresponds to the point Po on the curve, then a vector
equation for the tangent line at Po is given by
(see Fig. 2.6 ), where R
H
denotes the position vector of an arbitrary
point on the tangent line. The corresponding parametric equation
are
where x0 = ~ ( t , ) , yo = y(t, ), and z,, = z(t, ), - co < t < co .
Example 1. Describe the curve defined by the equation
R(t) = et cost i + et sin t j , t r 0
and find an equation of the tangent line to the curve at the point
corresponding to t = x/ 4 .
Vector Functions of One Variable
Fig. 2.6 Tangent line to a curve.
Solution: The curve is a plane curve whose parametric equa-
tions are given by
x = et cos t , y = e
t
s i n t , t r O
For each value of t, the distance from the origin of the correspond-
ing point on the curve is equal to
Thus as t increases the point (x, y) on the curve recedes from the
origin. Further, since the sine and cosine functions are periodic, the
point (x, y) goes around the origin in the counterclockwise direc-
tion. Therefore, as t increases the point (x, y) goes around the origin
with increasing radial distance. The curve traced by the points is
shown in Fig. 2.7. It is called a spiral.
Calculating R
t
(t) at t = n / 4 , we find
Rt(n/4) = enl 4 [(cos x/ 4 - sin n/ 4) i + (sin n/ 4 + cos n/ 4) j] = \12enl4 j
94 Chapter 2
Hence the tangent vector at t = x/ 4 is parallel to the unit vector j
and so the tangent line to the curve at that point has the equation
R*(t) = R,,+ t j,
( - m < t < m )
where R,, = R(n14) = 42 eenI4( i + j)/2.
Fig. 2.7 The spiral r = e
t
.
Example 2. Describe the curve defined by the vector equation
R(t) = a cos t i + a sin t j + bt k, ( t 2 0 )
where a and b are positive constants, and find the unit tangent
vector to the curve for any t > 0.
Solution: The parametric equations of the curve are gven by
x = a cos t, y = a sin t, z = bt,
(t r 0)
Vector Functions of One Variable
Since
we see that the curve lies on the lateral surface of a circular cylinder
of radius a whose axis is the z-axis. Moreover, as t increases the
coordinate z increases since b > 0. Therefore, as t increases, the curve
spirals upward around the cylinder in the counterclockwise direc-
tion as viewed from atop the positive z-axis (Fig. 2.8). The curve is
called a circular helix.
A tangent vector to the curve for any value of t (t > 0) is given
"Y
Fig. 2.8 A circular helix.
R'(t) = - a sin t i + a cos t j + bk
Chapter 2
To obtain the unit tangent vector, we normalize R'(t) by dividing it
by its magnitude, namely, I R'(t) I = (a
2
+ b2)1/2. Hence we have
We close this section with a few remarks concerning the types
of curves that we shall be dealing with in this book. We say that a
space curve represented by R(t), a < t < b, is smooth if R(t) has con-
tinuous derivative in (a, b) such that R'(t) + 0. Geometrically, this
means that at each point on the curve there is a unique tangent
vector which turns continuously as it moves along the curve with
the parameter t. Examples of smooth curves are lines, circles, para-
bolas, spirals and helixes. We say that a space curve is piecewise
smooth if R(t) has piecewise continuous non-zero derivative in (a,
b). This means that the interval [a, b] can be divided into subinter-
vals a = to < tl < . . < t, = b such that in each of the subintervals
(ti-l, ti), 1 5 i 5 n, R(t) has continuous derivative different from 0
and that the left-hand and right-hand limits of R'(t) at t = ti exist
(jump discontinuity ). Thus a piecewise smooth curve is composed
of pieces of curves each of which is smooth. For example, the sides
of a rectangle constitute a piecewise smooth (closed) curve.
We say that a curve defined by R(t), a 5 t < b, is closed if R(a) =
R(b). If, in addition, each point on the curve corresponds to one and
only one value of the parameter t (other than t = a and t = b), then
the curve is called a simple closed curve. Roughly speaking, a sim-
ple closed curve is one that does not cross itself. Unless otherwise
stated, all curves discussed in this book are assumed to be piecewise
smooth, and all closed curves are assumed simple.
2.5 ARC LENGTH AS A PARAMETER
Let C be a smooth space curve represented by
Vector Functions of One Variable
Denote by s(t) the arc length of the curve from the point corres-
ponding to t = a to a point corresponding to an arbitrary value of t
(Fig. 2.9). Then, as shown in elementary calculus, the arc length is
given by the integral
Since the integrand in (2.23) is precisely the ma~mitude of dR(t)/dt,
we can write (2.23) as
This gives us the relationship between the arc length s of the curve
and the parameter t. By the Fundamental Theorem of calculus, it
follows that
dt
(t) = -- (t) > 0
ds "",:
Thus the arc length s is an increasing function of the parameter t.
Equation (2.25) implies that
Hence if we define the vector differential dR by the equation
dR = d x i + d y j + d z k
then we have
The differential ds is called the element of arc length.
Chapter 2
Fig. 2.9 Arc length of a curve.
The study of the various properties of a space curve is greatly facili-
tated when the curve is represented by parametric equations whose
parameter is the arc length s of the curve measured from some ini-
tial point on the curve. Indeed, i f we take one end of the curve as
the initial point (s = 0), then each point P on the curve is uniquely
determined by the arc length of the curve measured from the initial
point to the point I? With arc length as a parameter, we immedi-
ately see from (2.25) that
dR (s)
1 ds 1; '
Hence the vector dR(s)/ds is a unit tangent vector. Therefore, when
a curve is represented by an equation R = R(s) with its arc length s,
0 s s s L, as a parameter, the unit tangent vector is given by
Vector Functions of One Variable
For example, the vector equation
R(t) = r (cos t i + sin t j), (0 s t < 2n)
represents a circle of radius r with center at the origin. The tangent
vector is given by
Rt(t) = r (- sin t i + cos t j)
so that 1 R1(t) 1 = r. Therefore, by (2.24) the arc len@h of the circle
corresponding to a value of t is given by
t
s r = r t , 0 s t < 2 n
Hence, if we replace t by sl r, we obtain the equation
R(s) = r[cos ( sl r ) i + sin (s/ r) j]
with the arc length s, 0 s s s 2nr, as a parameter. The unit tangent
vector is given by
R1(s) = -si n ( sl r ) i + cos (s/ r) j
Example 1. Find the arc length of the curve represented by
R(t) = et(cos t i + sin t j) , (0 s t 5 n)
What is the parametric equation in terms of its arc length?
Solution: We have
R1(t) = et(cos t - sin t) i + e
t
(sin t + cos t) j
so that
IR'(t) I = 42 et
Hence by (2.24) the arc length of the curve is
Chapter 2
The arc length from the initial point (1, 0) (t = 0) to an arbitrary
point (arbitrary value of t) on the curve is given by
Hence if we wish to represent the curve in terms of its arc length as
a parameter, we have to express the parameter t in terms of the arc
length s. In other words, we need to solve for t in terms of s from
the above relation. We readily find
t = ln[s/d2 + 11
so that
R(s) = (s/& + I){COS [ln(s/d2 + I)] i + sin [ln(s/d2 + l ) ] j)
Example 2. Represent the helix R(t) = a(cos t i + sin t j) +
bt k , 0 s t s n , in terms of its arc length as a parameter.
Solution: We have
Rt(t) = a(- sin t i + cos t j) + bk
so that I Rt(t) I = [a
2
+ b2]1/2. Hence
Replacing t by s/[a
2
+ b2]1/2 in the vector equation of the helix, we
obtain the desired equation.
In general, it is often very difficult to obtain the arc length s in
terms of a parameter t in the sense that the integral (2.24) may not
always be expressible by elementary functions. And even when this
is possible, it may still be difficult to solve for t in terms of s. The
artifice of using arc length as the parameter is, however, very
convenient in the theoretical development of geometric properties
of curves as we shall see in the next section.
Vector Functions of One Variable
2.3 EXERCISES
1. Obtain the representation of the spiral
R(t) = et(cos t i + sin t j)
( t r 0)
in terms of the new parameter r , where r = et , and show that
the orientation of the curve is preserved.
2. Obtain another representation of the curve
in terms of the parameter T, where r = 1 - t
2
. What happens
to the orientation of the curve?
3. Show that the curve defined by
R(t) = a(cos t i + sin t j) + a(1 + sin t )k
(t r 0)
is a plane curve. (Hint: Show that it is the intersection of a
cylinder and a plane.)
4. Showt hat t hecur vedef i nedby
R(t) = t(cos t i + sin t j + k)
(t 2 0)
lies on a cone. This curve is called a conical helix.
In each of Problems 5 through 10, describe the curve repre-
sented by R(t), and find the unit tangent vector to the curve at the
point corresponding to the given value of t.
5. R(t) = t(cos t i + sin t j) (t r 0); t = n/ 6
6. R(t) = a cos t i + b sin t j , a and b are constants (0 s t < 2n);
t = n/ 4
7. ~ ( t ) = ( t ~ + l ) i + t j (t rO); t = 2
8. R(t) = cost i + sin t j + (1 + sin t )k ( 0 s t t 24; t = n / 6
9. R( t ) =2( t - s i nt ) i +2( 1- cos t ) j ( t r O) ; t =n / 3
10. R(t) = e 2 i + e-2j + (1 + t
2
)k ( t r 0); t = 1
11. Determine the arc length of the curve represented by
R(t) = et(cos t i + sin t j + k)
(t r 0)
as a function of t. Express the equation of the curve in terms
102 Chapter 2
of the arc length s.
12. Determine the arc length of the curve represented by
R(t) = t(cos t i + sin t j + k)
(t r 0)
as a function of t.
13. Represent the curve defined by
R(t) = (sin t - t cos t)i + (cos t + t sin t)j + (t
2
/2)k (t r 0)
in terms of the arc length of the curve.
14. Find the length of the curve defined by
R(t) = sin t i + t j + (1 - cos t )k
for 0 r t < 2n.
25. Find the length of the circular helix
R(t) = a(cos t i + sin t j) + bt k
from t = 0 to t = 2x.
16. Find the expression of the arc length ds for the curve given in
(a) Problem 3, (b) Problem 4.
17. Let C be a curve in the xy-plane represented in polar coordi-
nates by the equation
r = f(0),
(x = r cos 8, y = r sin 8)
where f is a continuously differentiable function. Show that
ds
2
= [f
2
(0) + f'2(8)]d82.
2.6 SIMPLE GEOMETRY OF CURVES
We consider a space curve that is parametrized by R = R(s),
where s denotes arc length of the curve, 0 r s r L. We assume that
R(s) has continuous derivatives up to the second order in the inter-
val (0, L). As we saw in the preceding section, the unit tangent
vector to the curve is given by t = dR(s)/ds. Now let us consider how
the unit vector t changes as it moves along the curve. Since t has
constant magnitude, only its direction can change. The rate of this
change is measured by the derivative dt / ds. Since the magnitude of
t is constant, we know from Theorem 3 (Sec. 2.3) that t and dt / ds are
orthogonal; hence, t dt / ds = 0.
Curvature and Principal Normal Vector
If dt / ds = 0 , then t is a constant vector (does not change in
magnitude and direction); hence, the curve is a straight line. On the
other hand, if dt / ds + 0, we can write
Vector Functions of One Variable
103
where n is a unit vector in the same direction as dt / ds and, hence,
orthogonal to t. It follows from (2.28) that
and so
The quantity ~ ( s ) 2 0 defined by (2.29) is called the curz~cl t urc
of the curve at the point corresponding to the parameter s. It is a
measure of the rate at which the unit tangent vector t changes its
direction at a point. A large value of K at a point implies that the
curve is "sharply curved" at that point, whereas a small value of K
means that the curve is almost "straight". Thus for a straight line K
= 0 as we have observed above.
When K z 0, its reciprocal p = 1 / ~ , is called the radius of curzJa-
t u r f . This is the radius of the circle, called the circle of curvature,
that best fits the curve at the point determined by s.
The unit vector n defined by (2.30) is called the principle nor-
nznl vector to the curve at the point corresponding to the parameter
s. The center of the circle of curvature lies on the line along the
principal normal vector.
In order to get an idea of the direction of n on the curve, let us
examine geometrically the definition of dtlds, namely,
dt t(s + AS) - t(s)
- ( s ) = lim
ds
AS + 0
As
Referring to Fig. 2. 10, we notice that when As > 0 the vector At =
t(s + As) - t(s) is directed toward the concave side of the curve. Since
As > 0, the same is true of the limit of AtlAs as As tends to zero. For
As < 0, the vector At is directed toward the convex side of the curve;
however, since As < 0, the vector AtlAs points in the opposite direc-
Chapter 2
As)
Fig. 2.10 The principal normal vector.
tion to At, and so is the limit AtlAs (see Fig. 2.10 and also Example
3). Therefore, at each point on the curve, the principal normal vec-
tor n points toward the concave side of the curve.
Example 1. Consider the circle of radius a described by
R(s) = a(cos (s/a) i t sin (s/a) j) (0 s s I 2na)
We have t(s) = R1(s) = - sin (s/ a) i + cos (s/a) j, so that
t'(s) = (1 /a)[- cos (s/a) i - sin (s/a) j]
Since tl(s) = Kn , we deduce that K = 1 / a and n = - cos (s/a)i -
sin (s/a)j. Clearly n = - R(s) /a, which shows that the principal nor-
mal vector points toward the center as shown in Fig. 2.11. It follows
that the circle of curvature coincides with the circle itself.
Vector Functions of One Variable
Fig. 2.11 The vectors t and n on a circle.
Osculating Plane and the Binormal Vector
We observe that at each point on a curve, the two unit vectors
t and n determine a plane containing that point. This plane is
known as the osculating yl n~z t ~. It is the plane that best fits the curve
at that point and thus contains the circle of curvature. For a plane
curve (not a straight line), the osculating plane coincides with the
plane on which the curve lies. In general, however, the osculating
Fig. 2.12 Osculating planes.
Chapter 2
plane varies from point to point on the curve as shown in Fig. 2.12.
Since t and n are orthogonal unit vectors, it follows that I bl = It x n 1
= 1. Moreover, since b is orthogonal to both the vectors t and n, it
is orthogonal to the osculating plane. We call the unit vector b the
birlorvlzal vector to the curve. From the definition of cross product,
it follows that the three vectors t , n , b , in that order, form a right-
handed triple of orthonormal vectors varying in direction from
point to point on the curve. The vectors are sometimes referred to
as a moving trihedral (Fig. 2.13).
For the plane curve in Example 1, we observe that t(s) =
- sin (s/ a) i + cos (s/a) j and n(s) = - cos s / a) i - sin (s/a) j. Hence the
binormal vector b(s) is given by
which is indeed orthogonal to the xy-plane on which the circle lies.
Fig. 2.13 The moving trihedral.
Some Important Formulas
When a curve is represented by R = R(t), where t is an arbitrary
parameter, its unit tangent vector t is given by t(t) = R1(t)/ I R1(t) I .
By the chain rule, the derivative of t with respect to arc length s is
Vector Functions of One Variable
dt dt dt - dt/dt tf (t)
- -
d ~ ( ~ ) = d t d s ds/dt - R f ( t ) ,
Thus the vector t'(t) has the same direction as tf(s) and, therefore,
the principal normal vector n may also be given by the formula
tf (t)
n(t) = It' (t) I
This shows that the formula (2.30) for the principal normal vector is
valid no matter what parameter is used in representing the curve.
From the definition (2.29) of the curvature we deduce from
(2.32) that
Another formula for the curvature, which may son~etimes be more
convenient to use, is given by (see Problem 10)
~ ( t ) =
I Rf (t) x R" (t) I
(2.34)
I R' (t) I
Example 2. Determine the curvature, the principal normal
and the binormal vector for the circular helix
R(t) = a(cos t i + sin t j) + bt k
(t s 0)
Solution: Since Rt(t) = a(- sin t i + cos t j) + b k , we have
Taking the derivative, we find
- a c o s t i - a s i n t j
tf (t) = - -
- - - - -
i - >
Chapter 2
and so I tt(t) I = a(a
2
+ b2)-'I2. Hence by (2.33) we obtain
Thus the helix has a constant curvature as to be expected.
By (2.30) we find
n = - c o s t i - s i nt j
which is directed toward the axis (z-axis) of the helix. Taking the
cross product of t and n, we obtain the binormal vector
Fig. 2.14 A moving trihedral on a helix.
1
b = t x n =
2
\ a + b 2
i j k
-a sin t a cos t b
- cost - s i n t o
Vector Functions of One Variable
The trihedral (t , n , b ) on the helix are shown in Fig. 2.14
Curvature for a Plane Curve
Consider a curve in the xy-plane. Since t is a unit vector, we
may write it in the form
t (t) = cos 0(t) i + sin 0(t) j
where H(t) denotes the angle of inclination of the tangent vector at
the point corresponding to t (Fig. 2.15). Then, differentiating with
respect to t, we obtain
t'(t) = [- sin 0 (t) i + cos O(t) j]0' (t)
Hence 1 t'(t) 1 = ( 0'(t) 1, so that 1 t'(t) I is a measure of the absolute
0
Fig. 2.15
K (t) = I ds/dO I on a plane curve .
110 Chapter 2
value of the rate of change of the angle of inclination of the tangent
vector. If O1(t) > 0, then O(t) is increasing with t and hence the princi-
pal normal vector is given by n(t) = - sin 8(t) i + cos O(t) j. If O(t) < 0
so that 8 is decreasing with t, then according to (2.30) we have n(t) =
sin O(t)i - cos O(t)j. In either case, we see that n is directed toward
the concave side of the curve.
Now by (2.33) the curvature K is given by K = lO'(t) 1 / IR1(t) I .
Since
9' (t) =
dO ds dO
= IRt(t)l
ds dt ds
it follows that
Thus for a plane curve the curvature is a measure of the absolute
value of the rate of change of 8, the angle of inclination of the tan-
gent vector, with respect to arc length.
Example 3. Find the curvature of the curve represented by
R(t) = a[(t - sin t)i + (1 - cos t)j] 0 s t s 2n, a > 0
Solution: We use the formula (2.34). First, we have
R
t
(t) = a[(l - cos t) i + sin t j]
R"(t) = a[sin t i + cos t j]
so that
Therefore
I R'(t) x R1'(t) I = a
2
(cos t - 1)
R' ( t ) xRM( t ) =
and so, by (2.34), we find
i j
a( 1- cost ) a s i n t 0
a sin t a cos t 0
2
= a ( c o s t - l ) k
Vector Functions of One Variable
2
a (1 - cos t)
K =
- -
1
I R l (t) 13
2 ~ 2 ( I - cos t) a
2.4 EXERCISES
In each of Problems 1 through 5, find the (a) curvature, (b)
principal normal vector, and (c) the binormal vector of the space
curve represented by the given vector equation.
R( t ) = t 2 i + ( 1 + t ) j + & t k ( t z O) .
R(t) = et cos t i + et sin t j + et k ( t r 0 ).
R( t ) = t c o s t i + t s i n t j + t k ( t r O) .
R(t) = cosh t i + sinh t j + t k ( t r 0 ).
R(t) = (sin t - t cos t) i + (cos t + t sin t) j + t
2
/2 k ( t r 0 ).
Find the radius of curvature of the curve described by
R(t) = cos t i + sin t j + cos t k (0 5 t < 2n )
and determine the center of the circle of curvature when t =
n / 4 .
Find the center and the radius of the circle of curvature of the
curve described by
R(t) = (t + sin t) i + (1 - cos t ) j + 4 sin (t / 2) k when t = n/ 2.
Find the curvatuve of the curve described by
R( t ) =e t i +e mt j + d 2 t k ( t t 0 ) .
In each case, find an equation of the osculating plane to the
given curve at the indicated point:
(a) the curve given in Problen~ 2 at t = n/4;
(b) the curve given in Problem 3 at t = n.
Let the curve C be represented by R(t) (a s t s b) where R(t) is
twice continuously differentiable. Show that the curvature
may be given by
[Hint: Differentiate Rf(t) = t (ds/ dt ) with respect to s and use
(2.28).]
112 Chapter 2
11. Referring to Problem 10, show that also
I I R' (t) 1 1 R" (t) 1 2- ( R' (t) - R" (t))21
[Hint: Differentiate t = Rf(t )/ I R1(t) I and use (2.33).]
12. If C is a plane curve described by R(t) = x(t) i + y(t) j,
(a) show that the formula in Problem 10 reduces to
(b) If C is described by y = f(x) , show that
13. Let C be a plane curve described by R(s) = x(s) i + y(s) j, where
s denotes arc length. Show that
and
14. Referring to Problem 10, show that b and n are given by
R' (t) x R" (t)
b =
I R' (t) x R" (t) 1
and
W' ( t ) ~ R" (91 x R' (t)
n = --
K R' (t) 4
-
- -- P' - (t)x R" (01 x R' ( 9
1 R' (t) x R" (t)ll R' (t) ;
Vector Functions of One Variable
2.7 TORSION AND FRENET-SERRET FORMULAS
We have seen that the curvature of a curve is related to the
rate of change of the unit tangent vector as it moves along the
curve. There is another property of a curve, called torsion, that is
related to the rate at which the binormal vector b changes its
direction as it moves along the curve with the other two unit vec-
tors t and n. This rate of change is measured by the derivative
dblds. Now since b has constant magnitude, we know that db/ ds is
orthogonal to b. By differentiating the equation b - t = 0, we obtain
since b dt l ds = b . Kn = 0 in view of (2.28). Therefore, the vector
db/ ds is orthogonal to both b and t .
If db/ds = 0, then b is a constant vector, and hence the curve
is a plane curve. On the other hand, i f dbl ds i. 0 , then because it is
orthogonal to t and b, it is parallel to the principal normal vector
n. Hence we can write
db
(s) = - t(s) n(s)
(2.35)
ds
The scalar factor t(s) is called the torsio~z of the curve at the point
corresponding to the parameter s. The torsion of a curve is a meas-
ure of the rate at which the binormal vector or the osculating plane
rotates about the tangent vector as it moves along the curve. The
negative sign in (2.35) is introduced so that the torsion will come
out positive when the right-handed triple t, n, b rotates in a right-
handed sense about the tangent vector as it moves in the positive
direction on the curve.
Just as the curvature of a straight line is zero, we deduce from
(2.35) that the torsion of a plane curve is zero.
Equations (2.28) and (2.35) are the results of differentiating the
unit vectors t and n with respect to arc length. To complete the ana-
lysis of the orthonormal set of vectors (t, n, b), we now consider the
derivative of n. Since n = b x t , we have
dn db dt
= -- x t + b x - -
ds ds ds
Chapter 2
Using (2.28) and (2.35), this becomes
since n x t = - b and b x n = - t. Therefore, collecting our results, we
have the system of equations
satisfied by the unit vectors t , n , b. These equations are known as
the Frenet-Serret formulas. They play a central role in the study of
differential geometry of curves.
When the curve is represented by R(t), where t is any para-
meter, a convenient formula for the torsion is given by
T(t) =
R' (t) x R' ' (t) R' ' ' (t)
(2.37)
If we substitute (2.34) for the curvature K, (2.37) becomes
R' (t) x R" (t) . R' " (t)
T(t) = - - -
-
I R' ( ~) x R" (t)12
To derive the formula (2.37), we note that
dR ds ds
R' (t) = ;is
j = t dt
Vector Functions of One Variable
= ( + K ( - nt + r b )
2
d s d s d3s
+ ~n + t -
dt dt
2
dt "
where we have repeatedly used the formulas (2.36). Taking the cross
product of R' and R", we obtain
3
R' (t) x R" (t) = K (t x n)
Talung the dot product of this expression with R"', noting that b is
orthogonal to t and n , we obtain
6
R' (t) R' (t) x R" (t). R' " (t) =
from which (2.37) follows in view of (2.25).
Example 1. Determine the torsion of the helix described in
Example 2, Sec. 2.6.
Solution: From Example 2, Sec. 2.6, we have
bsi n t i - b c o s t j + a k
b(t) = -
2 2
\ a + b
Chapter 2
and
Hence
db d b d t b c o s t i + b s i n t j 1
- -
-
- -
ds d t d s ' 2 2
\ a + b L' a
2
+ b2
since n = - cos t i - sin t j. Therefore, the torsion of the helix is
which is a constant. This means that the osculating plane rotates
clockwise about the tangent vector at a uniform rate as it climbs up
the helix. The student should verify that
sin t i - c o s t j dn
z b - ~ f = - - - - - - - - - - . -
2 2 ds
l a + b
Example 2. Determine the torsion of the curve represented by
R(t) = a(t - sin t) i + a(l - cos t) j + b t k
( t 2 0)
where a and b are positive constants.
Solution: We use the formula (2.38) to determine the torsion.
Differentiating R three times, we find
R'(t) = a(1 - cos t) i + a sin t j + b k
R"(t) = a sint i + a cos t j
R"'(t) = a cos t i - a sin t j
Vector Functions of One Variable
Hence
and so
Since
R' (t) x R" (t) =
it follows from (2.38) that
i j k
a(1 - cos t) a sin t b
a s i n t acos t 0
2.5 EXERCISES
In each of Problems 1 through 5, determine the torsion of the
given curve.
1. The curve described in Problem I, Exercise 2.4.
2. The curve described in Problem 2, Exercise 2.4.
3. The curve described in Problem 3, Exercise 2.4.
4. The curve described in Problem 4, Exercise 2.4.
5. The curve described in Problem 5, Exercise 2.4.
6. Find the torsion of the curve described by
where a is a positive constant.
7. Show that the torsion of the curve described by
R ( t ) = t i + ( l + l / t ) j + ( l / t - t ) k ( t r O )
is zero.
Chapter 2
8. Let C be a curve represented by R(t), where R(t) is three times
continuously differentiable for t 2 0. Show that the torsion at
any point is given by
K =
( R' (t) R" (t) R' ' ' (t) )
I Rt (t ) x R"(t) l 2
where the numerator denotes the scalar triple product of R',
R" and R"'. (Hint: Differentiate R"(t) and use the Frenet-
Serret formula.)
9. Show that if C is represented by R(s) where s denotes arc
length, then
(R' (s) R" (s) R ' ' ' (s))
K = ~Rt' (s)I, r =
1 R' (s) ~2
10. Find the curvature and the torsion for the curve C represented
by
R( t ) = ( 3 t - t " i + 3 t 2 j + ( 3 t + t 3 ) k ( t 2 O)
11. Write the Frenet-Serret formulas for the curve
R(t) = (sin t - t cos t) i + (cos t + t sin t) j + t
2
/2 k ( t 2 0 )
12. Write the Frenet-Serret formulas for the curve
2.8 APPLICATIONS TO CURVILINEAR MOTIONS
Let R(t) denote the position vector of a moving particle at time
t relative to a rectangular coordinate system. As t varies over a time
interval, the particle traces out a curve in space. This curve is called
the path or trajectory of the particle (Fig. 2.16). We shall show that
the physical concepts of velocity, speed, and acceleration of the
particle can be defined in terms of the derivatives of the position
vector R(t).
Vector Functions of One Variable
v( t ) = R' (t )
Fig. 2.16 Curvilinear motion of a particle.
Velocity Vector
Suppose that a particle is at a point P at time t and at a point Q
at time t + At. Then, as shown in Fig. 2.16, AR = R(t + At) - R(t)
represents the displacement of the particle from P to Q during the
time interval At. The instantaneous velocity of the particle at time
t, which we denote by v(t), is defined as the limit
lim AR
v (t) =
dR
At --> 0 At
(t) = di (t)
We call the vector function (2.39) simply the velocity of the particle
at time t. Thus the time derivative of the position vector of a mov-
ing particle represents the velocity o f the particle.
Now let s denote the distance traveled by the particle along its
path. By the chain rule we have
Chapter 2
where t is the unit tangent vector to the trajectory of the particle.
This shows that the velocity vector v is directed along the tangent
vector, that is, the velocity vector is tangential to the path of motion.
Moreover, from
we see that the magnitude of the velocity vector represents the
rate of change of arc length with respect to time, which is precisely
the rate at which the particle is moving along its path. The scalar
v(t) is called the speed of the particle at time t.
If R(t) = x(t)i + y(t)j + z(t)k (a s t s b), then the velocity (2.40) is
given by
so that the speed at time t is
- -
2 2 2 d s
v (t) = Y x' (t) + y' (t) + z' (t)
=
d t
Acceleration Vector
The velocity of a moving particle changes when there is
change in the speed or the direction of motion. The rate of change
of the velocity is called the acceleration, denoted by a(t). Thus we
have
7
dv d k
a(t) = -(t) = - - (t)
dt
dt
2
In terms of its components, the acceleration vector is given by
2 2 2
d z
a(t) = X(t) i + :(t) j + (t ) k
dt
2
dt dt
Vector Functions of One Variable
so that its magnitude is
Unlike the velocity vector, the accelertation vector is not tangential
to the path of motion. In fact, we shall show later that a(t) can be
expressed as a linear combination of the tangent vector t and the
principal normal vector n.
Example 1. Consider the motion of a particle described by
R(t) = a (cos ot i + sin ot j) , t 2 0
where a and o are constants. Find the velocity, speed, and accele-
ration of the particle as a function of time t, t r 0.
Solution: If we let 8 = ot denote the polar angle which the po-
sition vector R makes with the x-axis, we see that as t increases the
particle is moving counterclockwise around the origin on the circle
of radius a. The velocity vector is given by
v(t) = R'(t) = ao(- sin a t i + cos o t j)
so that the speed is
which is constant. The scalar o is called the angular speed of the par-
ticle, expressed in either revolutions or radians per unit time. It is
the rate at which the particle revolves around the origin on the
circle.
The corresponding acceleration vector is given by
a(t) = vt(t) = - ao2(cos o t i + sin ot j) = - 0 2 ~ ( t )
which also has constant magnitude I a(t) I = am2. This shows that the
acceleration vector is opposite in direction to the position vector;
hence it is directed toward the origin, as shown in Fig. 2.17. This
acceleration is called the centripetal or normal acceleration.
Chapter 2
Fig. 2.17 Circular motion.
Tangential and Normal Components of Acceleration
We now show that the acceleration vector can be expressed as a
sum of two vectors, one parallel to the unit tangent vector t and the
other parallel to the principal normal n. In view of (2.40) we see
that a can be written as
Since dt / ds = Kn and v = ds/dt, this becomes
Thus we have resolved the acceleration vector as the sum of two
orthogonal vectors, one of magnitude
Vector Functions of One Variable
directed along the tangent vector t , and the other of magnitude
directed along the principal normal vector n, as shown in Fig. 2.18.
The scalars (2.48) and (2.49) are called the tangential and the normal
components of the acceleration, respectively. It is clears that
Fig. 2.18 Tangential and normal components of acceleration.
124 Chapter 2
From the formulas (2.48) and (2.49), we see that if a particle is
moving around a curve at a constant speed (v = const.), the tangen-
tial component of the acceleration is zero so that the only accelera-
tion is the normal acceleration (see Example 1). On the other hand,
if the particle is moving along a straight line so that the curvature is
zero, then the only acceleration present is the tangential accelera-
tion. These are of course intuitively evident.
Example 2. Consider the motion of a particle described by
R(t) = (sin t - t cos t) i + (cos t + t sin t) j
(t r 0)
Find the velocity, the acceleration, and the tangential and normal
components of the acceleration.
Solution: The velocity is p e n by
v(t) = Rt(t) = t sint i + t cos t j
so that the speed is
The acceleration is gven by
a(t) = (sin t + t cos t) i + (cos t - t sin t) j
and its magnitude is
2 2 ' 2
a = V (sin t + t cos t) + (cos t - t sin t) = 1' I + t
By (2.48) the tangential component of the acceleration is equal to
hence, by (2.50) the normal component is given by
Vector Functions of One Variable
125
Example 3. The position vector of a particle at time t is given
by R(t). If R(t) x a(t) = 0, where a(t) is the acceleration, show that
the motion takes place on a plane.
Solution: It is sufficient to show that the torsion of the trajec-
tory of the particle is zero, since it implies that the trajectory is a
plane curve. Now, by assumption, we have
Differentiating this, we find
Taking the dot product of the last equation with R"'(t), we obtain
By (2.37) , it follows that z = 0; hence, the trajectory lies on a plane.
An Energy Theorem in Mechanics
We conclude this section with the derivation of an important
theorem in mechanics. Let m denote the mass of a particle in mo-
tion whose position vector is given by R(t). According to Newton's
second law of motion, the force F acting on the particle at time t is
given by
Thus the force and the acceleration have the same direction. Now
since
Chapter 2
we have by (2.51)
The quantity (1 /2)mv
2
(t) in (2.52) is known as the kinetic
energy of the particle at time t. If TI and T2 denote the kinetic ener-
gy of the particle at time tl and t2 (tl < t j , respectively, then from
(2.52) the change in kinetic energy during the time interval t2- tl is
given by
This is known as the energy theorem in mechanics.
For instance, let m denote the mass of the particle in Example
1. Since the acceleration is given by a = - w2R(t), the force acting on
the particle is equal to F = - rnw2R. This force is directed toward the
center of the circular path and it is called the cerztriyctal force. The
force -F which is directed away from the center is called the cerz-
trifugal force. This is the force that tends to throw a car off its track
when it makes a turn or changes its direction. The kinetic energy of
the particle is given by
T = (1 /2) m
which is constant. Thus a particle in circular motion at constant an-
gular speed o has constant energy.
2.6 EXERCISES
In each of Problems 1 through 5, the position vector R(t) of a
moving particle at time t is given. Determine the velocity, speed,
and acceleration of the particle at the indicated time.
Vector Functions of One Variable
~ ( t ) = 3 t i + 3t
2
j +2 t 3 k , t = 1.
R(t) = a(sinh t - t) i + a(cosh t - 1) j , t = In 2.
R( t ) = t c o s t i + t s i n t j + t
2
/ 2 k , t = x .
R(t) = a(l + cos t) i + b sin t j + a cos t k , t = x/3.
R(t) = cosh t cos t i + cosh t sin t j + sinh t k , t = x.
Consider the motion of a particle on the helix
R(t) = a cos wt i + a sin wt j + bwt k (t r 0)
where a, b, and co are constants.
(a) Show that the velocity and acceleration both have constant
magnitude.
(b) Show that the velocity and acceleration are orthogonal.
(c) Express the acceleration in terms of the principal normal
vector to the curve.
The motion of a particle in space is described by the equation
R(t) = t i + sin t j + (1 - cos t) k
(t r 0 )
Find the normal and tangential components of the accelera-
tion. Express the acceleration in terms of t and n .
The motion of a particle in space is described by the equation
R(t) = (t + sin t) i + (1 - cos t ) j +4cos (t / 2) k
(t r 0 )
Find the acceleration and express it in terms of t and n .
A particle moves in space according to the equation
R(t) = a(1 + cos t ) i + a sin t j + bt k
( t r 0 )
Find the tangential and normal components of the accelera-
tion.
The position vector of a moving particle is given by
R(t) = a(1 + cos t) i + a sin t j + a(1 + cos t) k
( t r 0)
(a)
Show that the path of motion is an ellipse, and find the
equation of the plane on which the ellipse lies.
(b) What are the tangential and normal components of the
acceleration?
A particle is moving around a circle such that its distance s in
feet from a fixed point on the circle is given by s(t) = t
2
+ 1,
Chapter 2
where t is time in seconds. If the magnitude of the accelera-
tion is (16t4 + 4)lI2 ft/sec, find (a) the tangential and normal
components of the acceleration, (b) the radius of the circle.
12. The position vector of a moving particle is given by
R(t) = A cos o t + B sin o t
where A , B , and o are constants. (a) Find the velocity of the
particle. (b) Show that the acceleration is directed toward the
origin.
13. Show that if a particle moves in space with constant speed,
then the acceleration vector is orthogonal to the velocity vec-
tor.
14. Show that if the velocity and acceleration of a moving particle
have constant magnitude, then its path has constant curvature.
15. Show that if the force acting on a particle is at all times ortho-
gonal to the direction of motion, then the speed is constant.
16. The position vector of a moving particle is given by R(t).
(a) Show that R(t) x a = 0 if and only if R(t) x v = A , where
A is a constant vector and v and a denote the velocity and
acceleration vectors, respectively. (b) Show that if R(t) x v = A ,
then the particle lies on a plane.
2.9 CURVILINEAR MOTION IN POLAR COORDINATES
For a particle whose motion takes place on a plane, its trajec-
tory is described by the equation
with respect to a rectangular cartesian coordinate system. Some-
times, however, it is more convenient to study the motion using
polar coordinates rather than cartesian coordinates. In this section,
we consider this approach and derive the corresponding formulas
for the velocity and acceleration.
Since the cartesian coordinates (x, y) are related to the polar
coordinates (r, 8) by the equations
Vector Functions of One Variable
Fig. 2.19 The unit vectors u, and uo i n polar coordinates.
we may express (2.54) in polar coordinates as
R(t) = r(t) [cos 8(t) i + sin 8(t) j]
= r(t)u ,(8 0)) (2.55)
where
ur(O (t)) = cos 8 (t) i + sin 8 (t) j (2.56)
The quantities r(t) and H(t) denote the radial distance and the polar
angle of the particle at time t (Fig. 2.19).
It is clear that u, is a unit vector and has the same direction as
the position vector R. By Theorem 3 (Sec. 2.3) , it follows that
dur
u 0 =
= - sin 8 i + cos 8 j (2.57)
do
is orthogonal to u,. The vector (2.57) is also a unit vector and it
points in the directon of increasing H (see. Fig. 2.19). Further, by dif-
ferentiating (2.57), we find
Chapter 2
The formulas (2.57) and (2.58) imply that the process of differentia-
ting either one of the unit vectors u r or ug is equivalent to rotating
that vector through 90 in the counterclockwise direction. As we
shall see in this discussion, the unit vectors u r and uo play the same
important role in polar coordinates as the unit vectors i and j in
cartesian coordinates.
Let us now determine the expressions for the velocity and the
acceleration of a moving particle described by (2.55) in polar coordi-
nates. Since 0 is a function of time or the parameter t, so is the unit
vector u,. Hence by the chain rule, we have
which, in view of (2.57), becomes
This is the formula for the velocity in polar coordinates. We see
from (2.59) that the velocity is the vector sum of two orthogonal
vectors, one parallel to u, and the other parallel to uO. The scalar
factors dr / dt and rd8/dt in (2.59) are called the radial and the trans-
verse components of the velocity, respectively (see Fig. 2.20).
Since u, and uo are orthogonal unit vectors, it follows that
and so the speed is given by
Vector Functions of One Variable
0 I
Fig. 2.20 Velocity vector in polar coordinates.
If the particle is moving around a circle with center at the ori-
gin, then r = const. and so dr / dt = 0. In such a case, the velocity
(2.59) reduces simply to v = r (do /dt)u,. The scalar quantity w = dR/dt
is, of course, the angular speed.
To obtain an expression for the acceleration, we differentiate
(2.59) and use the relations (2.57) and (2.58). We find
Thus the acceleration vector is also resolved into its radial compo-
nent (r"- ret 2) and transverse component (re" + 2r'Ht), as shown in
Chapter 2
Fig. 2.21 Acceleration vector in polar coordinates.
Fig. 2.21. Unlike the velocity vector, however, the acceleration vec-
tor is no longer tangential to the path of motion, as we observed in
Sec. 2.8. The term 2r'B'ug in (2.61) is sometimes called the Coriolis
acceleration.
If a particle is moving around a circle at uniform speed, so that
dr/ dt = 0 and dB/dt = const., then d28/dt2 = 0 and, hence, the
acceleration (2.61) reduces to
which is the normal acceleration obtained in Example 1, Sec. 2.8.
Example 1. A particle is moving from the center to the rim
along a spoke of a wheel that is rotating at a constant angular speed
o. Its position vector is given by R(t) = tu,. Find the velocity and
the acceleration of the particle as a function of time.
Solution: By differentiating the vector function R, we obtain
Vector Functions of One Variable
the velocity
The velocity is composed of the velocity u, of the particle relative to
the wheel and the additional velocity t(dH/dt)uo due to the rotation
of the wheel.
By differentiating v, we obtain the acceleration
The Coriolis acceleration is given by 2(d8 /dt)uo which points in the
direction of rotation. The term -t(d0 ~ d t ) ~ u , is obviously the centri-
petal acceleration directed toward the center of the wheel.
Kepler's Law on Planetary Motion
Equation (2.61) leads to an important result when the motion
is such that the acceleration is radial, that is, when the transverse
component of the acceleration vanishes. This is the case with regard
to the motion of the planets around the sun. Here we derive the
so-called Kepler's second law on planetary motion. The law states
that the position vector froni the sun to a planet sweeps out area at a
constant rate. In other words, as the planet moves around the sun,
its position vector sweeps out equal area during an equal interval of
time.
Let us assume that the mass of the sun is M and the mass of a
planet is m. Let R = r u, denote the position vector from the sun to
the planet, and let F denote the gravitational force with which the
planet is attracted to the sun (Fig. 2.22). According to Newton's
Chapter 2
Orbit of planet
Sun
Fig. 2.22 Planetary motion.
universal law of gravitation, we have
where G is the gravitational constant. By Newton's second law of
motion this force is equal to the mass of the planet times its accele-
ration a, that is, F = ma. Hence we deduce
which is a radial acceleration directed toward the sun (because of the
negative sign). Since R and a are parallel, it follows that R x a = 0.
From Example 3, Sec. 2.8, we conclude that the orbit of the planet
lies on a plane.
Now let A(t) denote the area swept out by the position vector
of the planet from a certain time t = a to a subsequent time t (see Fig.
2.22), and let Oo and 0 denote the corresponding polar angles. From
elementary calculus, we know that
Vector Functions of One Variable
0 (1)
A (t) = 11 --
r.9
2
0 ,(a)
The rate of change of this area is given by
( t ) ] d0 1 2de
2 d8
r d$ = r dt
Since the acceleration of the planet is radial, the transversal com-
ponent of its acceleration vanishes. Hence from (2.61) we have
which implies that
2 do
r - = const.
dt
Therefore,
1 ad0
At ( t ) = r
2 dt
= const.
which says that the area swept out by the position vector of the pla-
net changes at a constant rate. This is Kepler's second law of plane-
tary motion.
2.7 EXERCISES
In each of Problems 1 through 6, the position vector of a
moving particle is given. Find the velocity and acceleration of the
particle at time t, and point out the Coriolis acceleration, if any.
1.
R(t) = cosh ot (cos t i + sin t j), w positive constant.
2.
R(t) = t
2
(cos wt i + sin wt j), w > 0 constant.
3.
R(t) = r(t)(cos 3t i + sin 3t j), where r(t) = a(1 - cos 3t), a > 0
Chapter 2
constant.
R(t) = r(t) u, , where r(t) = a(1 + sin t), 0 = 1 - e-t.
R(t) = r(t)(cos wt i + sin wt j) , where r(t) = a/ ( l + cos t), a > 0.
R(t) = r(t)(cos t i + sin t j ) , where r(t) =3a/[2(2 + cos t)], a > 0.
Let R (t) = r(t) u, . Express d 3 ~ /dt3 in terms of the unit vectors
u, and ug.
A particle moves on a plane with constant angular speed w
about the origin. If the acceleration increases at a rate propor-
tional to the vector u,, (a) find the transverse component of
d 3 ~ / d t ? @) Show that d%/dt2 = rw2/3.
A person weighing 160 lb walks at the rate of 2 ft/sec toward
the edge of a merry-go-round that is rotating with constant
angular speed of n / 5 radlsec. Find the centrifugal and the
Coriolis forces experienced by the man when he is 5 ft from the
center.
From formula (2.62), deduce that the force acting on the parti-
cle of mass m has two components, F, along u,and Fg along uo.
Thus show that the niotion of the particle is governed by the
quations
(i) F, = m
d 2r
dt
d 28
(ii) F o = m r + 2 m
dr d8
dt
dt dt
(a) From Eq. (ii) of Problem 10, show that
The quantity mr
2
(d0/dt) is called the angular momentum of
the particle.
(b) Hence, if the force acting on the particle is always parallel
to the radial vector u,, show that the angular momentum is
constant.
Vector Functions of One Variable 137
12. A particle is rotating counterclockwise around the ori gn of the
xy-plane with angular speed w.
(a) I f w = wkdenotes the angular velocity, show that
v = R1(t) = w x R(t), where R(t) = r(t) u,
(b) Show that a = w' x R(t) - w2 ~ ( t ) , where w' is the angular
aceleration.
(c) Thus deduce that if the angular velocity is constant, the
acceleration consists only of the normal acceleration.
2.10 CYLINDRICAL AND SPHERICAL COORDINATES
In studying motions in space it is oftentimes more convenient
to use noncartesian coordinate systems. The two most widely used
coordinate systems of this kind are the cylindrical and the spherical
coordinate systems. In this section we derive the expressions for the
velocity and the acceleration in these two coordinate systems.
Cylindrical Coordinates
The equations relating the rectanplar coordinates (x, y, z) and
the cylindrical coordinates (r, 8, z) of a point are given by
x = r c o s 8 , y = r s i n 8 , z = z
where r 2 0, 0 5 8 < 2n, - co < z < 03. The inverse transformation is
given by
r = (x2+ y2)1/2, 0 = arctan(y/x), z = z
Notice that the coordinates (r, 8) coincides with the polar coordi-
nates of the point (x, y). Referring to Fig. 2.23, we see that the posi-
tion vector of a point in cylindrical coordinates is given by
where u, = cos 8 i + sin 8 j. By taking the first order and the second
order derivatives of the vector function (2.62), we obtain the velocity
and the acceleration. Since the first term in (2.62) coincides with
(2.55), we find from (2.59) and (2.61)
Chapter 2
Fig. 2.23
z
The unit vectors u, , uo , k in cylindrical coordinates.
and
Example 1. Find the velocity and the acceleration of a moving
particle whose position vector is given by
R(t) = a(cos wt i + sin wt j) + b sin wt k
(t 2 0)
Solution: We note that
u, = cos wt i + sin wt j and ue = - sin wt i + cos wt j
where 8 = wt. Hence the velocity is given by
v(t) = R'(t) = am(- sin wt i + cos wt j) + bw cos cot k
= aw uo + bw cos wt k
Vector Functions of One Variable
so that the speed is
v(t) = w(a2 + b
2
cos
2
The acceleration is given by
a(t) = vl(t) = - aw2(cos wt i + sin wt j) - bw2 sin wt k
= - w2 R(t)
which shows that the acceleration is directed toward the origin.
Spherical Coordinates
The spherical coordinates (p, @, 8) of a point P are related to its
cartesian coordinates (x, y, z) by the equations
x = p sin cp cos 8, y = p sin cp sin 8, z = p cos $
where p is the distance of the point from the origin, 9 is the angle
measured from the positive z-axis to the line segment OP, 0 s cp s n,
and 8 is the polar angle of the point (x, y, 0) on the xy-plane, 0 s 8 <
271 (Fig. 2.24). Thus the position vector of a moving particle in
spherical coordinates is given by
R(t) = p sin $ cos C) i + p sin $ sin 8 j + p cos + k
= pup(@, 8) (2.65)
where
up($, 8) = sin cp cos 8 i + sin cp sin 8 j + cos cp k (2.66)
is the unit vector having the same direction as R.
In order to obtain the expressions for the velocity and the acce-
leration in spherical coordinates, we need to determine two other
unit vectors pointing in the direction of increasing @ and 8. First, let
us differentiate the unit vector up($, 8) with respect to $ and set u
cp =
du,/d$. We find
Chapter 2
Fig.
Y
b
2.24 The unit vectors u,, , u$ , uo i n spherical coordinates.
U$ = cos @ c o d i + cos $ sin 0 j - sin @ k (2.67)
This is a unit vector orthogonal to up(@, 8 ) and points in the direc-
tion of increasing $. Next we differentiate up($, A) with respect to 0
to obtain
dup($, 8)/d0 = - sin $ sin 0 i + sin $ cos 0 j
We know this is orthogonal to up($, 0) and, by direct calculation, it is
also orthogonal to uq However, it is not a unit vector. We there-
fore normalize it to obtain
u8 = - s i n0i + cos 0j
This is the same unit vector we encountered in polar coordinates.
The three mutually orthogonal unit vectors uV uq and ue,
shown in Fig. 2.24, play similar role in spherical coordinates as the
unit vectors i , j , and k in rectangular cartesian coordinates. They
are related to one another by the following formulas, which can be
easily verified:
Vector Functions of One Variable
du du9
=s i n quo, - = cos (Pug
ae de
Using these formulas, we now determine the expressions for the
velocity and the acceleration in spherical coordinates. For the velo-
city, we differentiate (2.65) to obtain
Differentiating (2.69) we obtain the expression for the acceleration,
de
+ "4" + : t [p i: sin "1 u0
+ dt [ d ,, dt
Chapter 2
By the formulas in (2.68), this becomes
2
dt dt
sin 0 + p ' sin + + 2p
dt dt
COS @
1
(2.70)
dt
2
We observe that both the velocity and the acceleration are expressed
in terms of their components along the unit vectors u,,, u+ and ug.
Example 1.
A particle is moving at a constant angular speed y
along a meridian of a sphere of radius a. I f the plane of the meri-
dian makes an angle 8 = n/ 6 with the x-axis, find the velocity and
the acceleration of the particle at any time t.
Solution: Since the particle is revolving at constant angular
speed y, the angle traversed in time t measured from the positive
z-axis is equal to @ = yt . Therefore, using spherical coordinates, the
position vector of the particle is given by
R(t) = a sin yt cos ( ~ 1 6 ) i + a sinyt sin (x/6) j + a cos yt k
= as i nyt u, + a c os yt k
where u, = cos (n/6) i + sin (n/6) j (see Fig. 2.25). Hence the velocity
is
v(t) = R'(t) = ay cos yt u, - ay sin yt k
Vector Functions of One Variable
and the acceleration is
a(t) = v'(t) = - ay
2
sin yt u, - ay
2
cos yt k = - y2 R(t)
It follows that the acceleration is directed toward the center of the
sphere.
Fig. 2.25 Motion on a sphere.
Example 2. Referring to the problem in Example 1, suppose
the sphere is also rotating at a constant angular speed w about the
z-axis, what is the velocity and the acceleration of the particle at any
time t?
Solution: Since the plane of the meridian rotates with the
sphere, the unit vector u, is no longer constant but is of the form
u, = cos wt i + sin wt j
where 8 = wt is the angle measured from the positive x-axis. Notice
that the x-axis together with the other coordinate axes is fixed in
space. By differentiating the position vector
R(t) = a sin yt u, + a cos yt k
Chapter 2
we obtain the velocity
v(t) = a y cos yt u, + a w sin yt (dur/dO) - a y sin yt k
= a y cos yt u, + a w sin yt ug - a y sin yt k
The term aw sin yt ue arises from the rotation of the sphere.
Differentiating v(t), we obtain the acceleration
2
a(t) = - ay sin yt u, + awy cos yt (du,/dR) + awy cos yt ut,
+ w sin yt (duo /dB) - ay cos yt k
2 2
= - ay sin yt u, + 2awy cos yt uR - aw sin yt u, - ay 2cos yt k
Since the sum of the first and the last terms is equal to - y2 R ,
the acceleration can be written as
2 2
a(t) = - y R + 2awy cos yt uu - aw sin yt u,
Comparing this with the result in Example 1, we see that the first
term is the radial acceleration directed toward the center of the
sphere which is due to the motion of the particle on the meridian.
2
The last term, aw sin yt u , is the acceleration directed toward the
z-axis and is due to the rotation of the sphere. The second term,
2awy cos ytu@ is the Coriolis acceleration, which results from the
interaction of the rotation of the sphere and the motion of the par-
ticle on the sphere.
2.8 EXERCISES
1. A particle moves in space so that its position at time t has
cylindrical coordinates r = 2, 8 = 3t, z = 2t. Find the velocity
and acceleration of the particle at time t.
2. A particle moves in space so that its position at time t has
cylindrical coordinates r = t, 8 = t, z = 2t. The particle traces out
Vector Functions of One Variable
a curve called a conical helix. Find the velocity and accelera-
tion of the particle at time t.
3. Referring to Problem 2, find a formula for the angle between
the velocity vector and the generator of the cone at each point
on the curve.
4. A particle moves in space so that its position at time t has
cylindrical coordinates r = sin t, 8 = t, z = t .
(a) Show that the curve traced out by the particle lies on the
cylinder 4x
2
+ 4(y - 1 1 2 ) ~ = 1.
(b) Find the velocity and acceleration of the particle at time t.
(c) Determine a formula for the angle between the velocity
vector and the unit vector k .
5. A particle moves in space at constant angular speed w along
the circle of intersection of the plane z = 1 with the sphere
of radius 2 and center at the origin.
(a) Find the formulas for the velocity and acceleration of the
particle at time t. Assume the particle is moving counter-
clockwise as viewed atop the positive z-axis.
(b) Suppose the sphere is also rotating at a constant angular
speed y about the z-axis. Find the velocity and acceleration if
the sphere rotates in the same direction as the particle; in the
opposite direction.
6. Referring to the problem in Example 1, suppose the particle is
moving along the meridian at angular speed y = t (so that $ =
t
2
). Find the transverse and radial components of the accelera-
tion as function of t.
7. Referring to the problem in Example 2, suppose the sphere is
rotating at a variable speed such that 8 = 9. Find the accelera-
tion and determine the Coriolis acceleration.
8. A particle moves at constant angular speed a along the circle
of intersection of the plane y = z and the sphere of radius a
(center at the origin). Show that the position vector of the
particle at time t can be expressed as
R(t) = a c os a t i + a sin a t j"
where j* = cos (3~14) i + sin (3~14) j and a t is the angle made
byR(t ) with the x-axis. What are the formulas for the velocity
146 Chapter 2
and acceleration?
9. Referring to Problem 8, suppose the sphere is now rotating
about the x-axis at constant angular speed y. Find the formula
for the position vector of the particle and determine the velo-
city and acceleration at time t.
10. Establish the formulas given in (2.70).
11. Show that the scalar coefficient of uo in (2.70) can be written as
12. A particle slides from (a, 0, 0) down a frictionless circular helix
R(0) = a cos 0 i + a sin 0 j + b0 k, 0 r 0, a, b > 0 with constant
angular speed dO/dt = x rad/sec. Find the velocity of the par-
ticle and the distance (assumed in meters) it has traveled at t =
2 sec. (Assume the positive z-axis to be pointed downward.)
13. A free falling body starting from rest at (a, 0, 0) slides down a
frictionless circular helix R(0) = a cos 8 i + a sin 8 j + be k, 8 2 0,
a, b > 0 at the speed v = \/(2gb0), where g is the gravitational
acceleration (32 ft / sec
2
or 980 cm / sec
2
).
(a) Find 0 as a function of time t r 0. Note that
(b) Find the distance traversed by the body along the helix at
t = 2 sec. (Assume the positive z-axis to be pointed downward.)
DIFFERENTIAL CALCULUS OF SCALAR
AND VECTOR FIELDS
In this chapter, we shall st udy the differential calculus of scalar
and vector fields. We shall generalize the idea of a derivative to
that of a directional derivative and consider the various spatial rates
of change which lead to the important concepts of the gradient of a
scalar field and the divergence and the curl of a vector field. These
concepts have significant applications in both geometrical and phy-
sical situ a t' ions.
3.1 SCALAR AND VECTOR FIELDS
Let D be a domain in the three-dimensional space, and let f be a
scalar function defined in D. Then at each point P: (x, y, z) in D, f
assigns a unique real number f(x, y, z) [or f(P) for brevity], which is
the value of f at that point. The domain D, together with the corres-
ponding values of f at each point in D, is called a scalar field. The
functio~i f is said to define the scalar field in D. Oftentimes, how-
ever, when the domain is clear from the context, the scalar field is
simply identified with the scalar function. Thus, for example, i f at
each point of the atmosphere there is assigned a real number T(P)
which represents the temperature at then T is a scalar field (also
called a temperature field). Other examples of scalar fields are the
density of air in the atmosphere, the pressure in a body of fluid, and
the gravitational potential in space.
Occasionally, we also consider scalar fields in a plane. Such a
field is, of course, defined by a scalar function of t wo independent
variables. A somewhat sophisticated though rather interesting
Chapter 3
example of a two-dimensional scalar field is that of a "mine field."
As we know, a mine field refers to a body of water or a piece of land
where mines are placed at randomly chosen points. Such a field can
be defined analytically by a function having the value 1 at each point
where there is a mine and zero everywhere else. The field is, of
course, not continuous.
On the other hand, if at each point P of a domain D there is
assigned a unique vector F(P), then the domain together with the
corresponding vectors at each point constitutes a vector field. The
vector function F is said to define the vector field in D. In this
context, we see that a vector function of a real variable defines a vec-
tor field in an interval. As in the case of a scalar field, a vector field
is frequently identified with the vector function defining the field.
Thus, for example, if at each point P of the atmosphere we assign a
vector v (P) which represents the wind velocity at P, then v defines a
vector field known as the velocity field. Other examples of vector
fields are the electric field surrounding a power transmission line,
the magnetic field around a magnet bar, the velocity field of a fluid
in motion, and the gravitational field surrounding the earth.
A vector field may be described graphically by drawing the
arrows that represent the vectors at each point in the domain. As an
example, the velocity field defined by F = -yi/(x2+y2) + xj /(x
2
+y2) on
the xy-plane has the graphic representation shown in Fig. 3.1 for
points lying on the circles of radii 1 and 2. The field represents a
rotational flow about the origin. It approximates, for example, the
velocity field of water draining in a bathtab.
Let f define a scalar field in a domain D. The set of points in D
at which f assumes the same value will, in general, describe a sur-
face in space. Such a surface is called a level surface of f and has the
equation f(x, y, z) = c for some constant c. For example, if f = x
2
+ y2
+ z
2
, then the level surfaces are concentric spheres defined by x
2
+ y?
+ z = c. In many physical problems, the level surfaces are usually
called by special names depending on the nature of the scalar field
defined by f. For instance, if f denotes the temperature or pressure in
the atmosphere, then the corresponding level surfaces are known as
isothermal or isobaric surfaces. If f denotes the gravitational poten-
tential about the earth, then the level surfaces are called equipoten-
tial surfaces. These concepts are of great importance in meteorology
and physics.
In the two-dimensional case, the set of points at which a scalar
field f has the same value describes a plane curve called a level
Differential Calculus of Fields
Fig. 3.1 Rotating velocity field.
curve or a contour. It has the representation f(x,y) = c. In many
applications, it is sometimes necessary to consider level curves on a
level surface of a scalar field. For example, suppose f(x, y, z) repre-
sents the pressure field of the atmosphere where z denotes the
altitude in some suitable scale. The isobaric surfaces S: f(x,y,z) = c
consists of points in the atmosphere where the pressure is equal to
the constant c. If we set z = k in the equation of an isobaric surface,
we then obtain a contour on the surface defined by f(x, y, k) = c.
This represents the contour of constant pressure at the altitude z = k.
This contour is known as an isobar in meteorology, and it appears
prominently in all weather maps.
Example 1. Discuss the level surfaces for the scalar field
defined by f(x, y, z) = 2x
2
+ y2 + z. Sketch the level surface for c = 1
and the contour at z = 1 /2.
Solution: The level surfaces are defined by 2x
2
+ y2 + z = c,
which is a family of elliptic paraboloid with vertices on the z-axis.
For c = 1 the elliptic paraboloid has vertex at (0, 0, 1) as shown in
Chapter 3
Fi . 3.2. The contour at z = 1 / 2 is an ellipse whose equation is 4x
2
t
5
2y = 1.
Fig. 3.2 A contour on a level surface.
3.2 ALGEBRA OF VECTOR FIELDS
Like vector functions of a single real variable, a vector field F
can also be expressed analytically in the form
where P, Q, and R are scalar fields defined in the same domain D.
These scalar fields are called the components of the vector field F.
By performing the operations on the components, the usual opera-
tors of vector algebra can be applied to vector fields in exactly the
same way it was done in Chapter 2 for vector functions of a single
variable.
Indeed, suppose F(x, y, z) = P(x, y, z)i + Q(x, y, z)j + R(x, y,
z)k and G(x, y, z) = U(x, y, z)i + Vx, y, z)j + W(x, y, z)k are two vector
fields, and f is .a scalar field all defined in the same domain D. Then
Differential Calculus of Fields
we have
(ii) (f F)(x, y, z) = f(x, y, z) [P(x, y, z)i + Q(xr y, z)j + R(xr y, z)kl
(iii) (Fa G)(x, y, z) = P(x, y, z)U(x, y, z) + Q(x, y, z)V(xr y, z) +
R(x, yr z)W(x, y, z) (3.2)
The sum F + G and the products f F and F x G are again vector
fields, whereas the dot product Fa G is a scalar field.
The notion of orthogonality and linear dependence can also be
extended to vector fields in a straightforward manner. Thus, for ex-
ample, the two vector fields F and G are orthogonal in D if and only
if Fa G = 0 identically in D.
The basic concepts of limit and continuity can also be carried
over directly to vector fields. Suppose F is a vector field defined in a
domain D and A is a constant vector. Let R,, denote the position
vector of a point (x,, y,, z,,) in D. Then we say that
(iv) (F x G)(x, y, z) =
lim
F(x, y, z) = A
R --> R,
i j k
P(x, yr z) Q(x, y, z) R(x, Y, z)
U (xr yr Z) V (xr y, Z) W (xf Y, z)
if for any number E > 0, there is a number b > 0 such that
whenever I R - R, I < h. In this definition, it is not required that F be
defined at the point (x, y,, z,,) itself.
The inequality I R - R, I i b is equivalent to
152 Chapter 3
which describes the interior of a sphere of radius 8 and center at (x,
y, zJ.
Thus, geometrically, the definition (3.3) means that for every
E > 0 there is a sphere about (x, y, z,) of radius 6 such that for each
point (x, y, z) inside the sphere the magnitude of the difference
between F(x, y, z) and A is less than E .
In terms of components, if F = Fli + F j + F+ and A = ali +
a j + a+, the definition (3.3) implies that
lim Fi (x, y, z) = ai (3.4)
(XI Yr Z) --> (x, Yy zo)
for i = 1,2, 3.
The following theorem can be easily established from the defi-
ni tion.
Theorem 1. Let F and G be vector fields such that
lim F(x, y, z) = A and lim G(x, y, z) = B
R --> R, R --> R,,
Then we have
(a) lim [F(x, y, z)+G(x, y, z)]= A + B
R --> R,
(b) lim nzF(x,y,z) =nzA
R --> R,
(c) lim F(x, y, z) G(x,y,z) = A0 B
R --> R,
(d)
lim
F(x, y, z) x G(x, y, z) = A x B
R --> R,
(e) 1im I F(x,y,z) I = I A l
R --> R,
Thus, for example, if F = x2i + y2j + zzk and G = xyi + yzj +xzkr
t hen
lim
F(x, y, 4. G(x, y, z)
(x,y,z) --> (1,2,-1)
Differential Calculus of Fields
A vector field F is said to be continuous at a point (xo , yo , zo )
if and only if
The definition requires that the function F must be defined at (xo ,
yo, z,). If F is continuous at each point of a domain D, then we say
that F is continuous in D. From (3.4) it follows that a vector field F
is continuous in D if and only if each of its components is conti-
nuous in D.
From Theorem 1, we see that if F and G are continuous in a
domain D, then so are the sum F + G, the scalar multiple hF, the
scalar product Fe G, and the cross product F x G.
3.1 EXERCISES
For each of the given scalar fields, sketch the level surface f(x,
y, z) = c corresponding to the given value of the constant c.
For each of the given scalar fields, sketch the contour f(x,y) = c
corresonding to the given value of the constant c.
5. f(x, y)= xy, c=- 2, - 1, 1, 2.
6. f ( ~ , y ) = x ~ - ~ ~ , C = -1, -2, 0, 1, 2.
7. f(x, y)= arctan(y/x), c = - l , O, l .
8. f(x, y)= e
X
cosy, c = O, l .
9. Sketch the isobars on the isobaric surface z = (4 - x
2
- y2)112 at
the altitudes z = 112, 1, 312.
154 Chapter 3
9. Sketch the isobars on the isobaric surface z = (4 - x
2
- y2)1/2 at
the altitudes z = 1/ 2, 1, 3/ 2.
10. Sketch the isotherms on the isothermal surface z = 5 - 2x
2
- 3y2
corresponding to z = 1,2,3.
11. Sketch the vector field defined by each of the following:
(a) F = xi + yj;
(b) F = x
2
i + yj
(c) F = yi - xj;
(d) F = xi + yj + zk
12. Let F = As i ny+Bc os x, whe r e A = i + j and B = i - j .
Evaluate each of the following:
( 4 l i mo ( + )F(xry)
(x, y) --> ( n / l n/ 4)
13. Let F = x cos z i + x sin z j + y k. Evaluate
( 4 lini F(x, z)
(x, y, z) --> (I, 2, n/ 27
(b) Show that F has a constant magnitude on the cylinder
x
2
+ y2 = a2, where a is a constant.
(c) show that dF/dx , dFldy , dFldz are mutually orthogonal.
14. Let F = s i n x i + c o s y j + I n z k ( zt O) , and G= e Xc o s y i +
e
X
s i nyj . Find (a) F + G; (b) F-G; (c) FxG.
15. Let F = e Xc o s z i +e
X
s i n z j + k and G =e Xc o s z i +e Xs i n z j
- eB k.
(a) Show that F and G are orthogonal for all (x, y, z).
(b) Show that the vector field F x G is parallel to the xy-plane.
3.3 DIRECTIONAL DERIVATIVE OF A SCALAR FIELD
Let f be a scalar field defined in a domain D, and suppose that
f is differentiable in D. We know that the first partial derivatives
df / ax, df lay, and df /dz of f represent the rate of change of f along the
x, y, and z coordinate axes, respectively. In many applications, it is
often necessary to know the rate of change of f in an arbitrary direc-
tion. This requires a more general concept of a derivative known as
directional derivative.
Differential Calculus of Fields 155
The definition of a directional derivative is very much like
the definition of an ordinary derivative in differential calculus.
Let
R, denote the position vector of a point PO: (xu y,, z,) in a domain D
and let the unit vector
u = cos t x i + cos fi j + cos y k
specify a direction at Po (Fig. 3.3). We note that points on the line
segment issuing from Po in the direction of u are given by
where the parameter s denotes the distance of the point from Po.
We define the derivative of f at Po in the direction of the vector u
as the limit
f(x,,+ s cos ( x , yo+ s cos p, z,,+ s cos y) - f(x,,, yo, z,,)
= l i m
s - 0
S
If we set
g(" = f (x,, 4- s cos u, y, + s cos p, z, + s cos y)
(3.8)
then the definition (3.7) becomes
Now differentiating (3.8) with respect to s and then setting s = 0, we
obtain
From (3.6) we find
Chapter 3
X
Fig . 3.3 Directional derivative.
dx
--
dz
= c o s a , dy = cos g, = cosy
ds ds ds
Thus the directional derivative of f at PC, in the direction of the vec-
tor u is given by the formula
When a = 0, @ = n / 2 , and y = n / 2 , so that u = i , we see that
(3.9) reduces to af(P,)/dx. Similarly, when u = j and u = k, (3.9)
yields df(P,) / dy and df(P,) / dz, respectively. Thus the partial deriva-
tives of f are actually directional derivatives in the positive direction
of the coordinate axes.
Example 1. Find the directional derivative of f = 2x
2
+ xy +
yz2 at (1, -1,2) in the direction of the vector A = i - 2j t 2k.
Solution: We have
Differential Calculus of Fields
so that at (1, -1,2)
a f a f a f
- = 3 , -- = 5 , - - - = -4
ax ay dz
Since I A I = 3, a unit vector in the direction of A is u = (1 /3)A.
Hence by (3.9) we have
Example 2. What is the directional derivative of f(x,y) = x?
+ 2 In y at (1, 1) in the direction of the line making 300 with the
positive x-axis?
Solution: At (1,l) we have
df ( l , l ) / ax = 2 and df ( l , l ) / dy = 3
A unit vector in the direction of the given line has the form
u = cos 30" i + sin 30 j
Hence, the directional derivative is equal to
We observe that the function defined in (3.8) is actually the
value of f at points on the line segment from Po and extending in
the direction of the unit vector u. Suppose we now replace the line
segment by a smooth curve C passing through P,,. Let the curve be
represented by the vector equation
where s denotes arc length of the curve. Then on C the function f
Chapter 3
assumes the value
g(s) = f [x(s), y(s), z(s)l
Differentiating this with respect to s, we find by the chain rule
where P is the point corresponding to the parameter s. Since the
coefficients dx/ ds, dy /ds, and dz/ ds in (3.10) are precisely the compo-
nents of the unit tangent vector t = dR/ ds to the curve at T: it fol-
lows that (3.10) is the directional derivative of f at P on the curve C
in the direction of the tangent vector. This derivative is called the
tangential derivative of f on the curve. It represents the rate of
change of f on the curve with respect to the arc length.
Exam le 3. Find the tangential derivative of f = xy on the
rt
circle x
2
+ y = a
2
oriented counterclockwise.
Solution: A vector equation that orients the circle in the
counterclockwise direction is given by
R(t) = a cos t i + a sin t j
(t 2 0)
Hence, the unit tangent vector to the circle is
R' (t)
= R(t) ~
= - sin t i + cos t j
so that cos tr = - sin t and cos (3 = cos t. Therefore, by (3.10) the tan-
gential derivative of f on the circle is given by
df I ds = y(- sin t) + x cos t
= y( - l / aJ + x(x/ a)
= ( x - y )/ a
Note that if we express f as a function of s on the circle (s = at),
we have
f = a
2
sin (s/ a) cos (s/ a) = (a
2
/2)[ sin 2(s/a)]
Differential Calculus of Fields
Differentiating this with respect to s, we find
df l ds = a cos
2
(s/ a) = a [cos (s/ a) - sin 2(s/a)] = (x
2
- y2)/a
which agrees with the previous result.
3.2 EXERCISES
In each of Problems 1 through 8, find the directional derivative
of the given scalar field at the given point in the direction of the
given vector.
f(x, y) = x
2
- y2, (2, -1) ; A = i + j.
f ( ~ , ~ ) = x ~ - 2 ~ ~ + 3 ~ ~ , ( - 1, 2) ; A = - i + 2 j .
f(x,y) = e
X
cos
( l r n/ 3) ; A = i - j .
1
f(x, y) = x" 3x y - xy + y2, (2,2); A = 2i + j.
f ( x , y , z ) =x y +y z +z x , (1, 1, -1); A = 2 i + j - 2 k .
f ( xr y, z) =xyeZ+z2, ( - 1, 2, 0) ; A = - i + 2 j +2k.
f(x, y, z) = x sin z - y cos z , (2, 3, n/ 4) ; A = 3i + 2j - 2k.
f(x,y, z) = ln(x2+ y2- z ) , (2,2,-1); A = i + j - k.
Show that the directional derivative of a scalar field f(x, y)
in the direction of a line malung an angle 8 with the positive
x-axis is given by (see Example 2)
Determine the value of 8 for which df/ ds is maximum and
determine that maximum value. (Note that the point P is
fixed .)
10. Apply the result of Problem 9 to find the maximum value of
the directional derivative of f(x, y) = x/($ + y2 ) at the point (1,
I), and determine the value of 8 in which the maximuni
value occurs.
11. Repeat Problem 10 for f(x, y) = x
2
- xy + y2.
12. The function f has at (1, -1) a directional derivative equal to d2
in the direction toward (3, I), and dl 0 in the direction toward
(0, 2). Find the value of df l dx and df /dy at (1, -1) . Determine
160 Chapter 3
the derivative of f at (1, -1) in the direction toward (2, 3).
13. Find the tangential derivative of f(x, y) = x)Z - ex in y at (d3 , 2)
on the curve x = 2 cos t, y = 4 sin t (t r 0).
14. Find the tangential derivative of f(x, y) = x
2
+ xy + )2 at any
point on the curve
15. Let f be a nonconstant scalar field that is twice continuously
differentiable on the xy-plane.
(a) Show that the tangential derivative of f is zero on the
curve described by f(x, y) = c, where c is a constant.
(b)
Show that the direction in which the directional derivative
of f is maximum is orthogonal to the curve f(x, y) = c.
16. Find the tangential derivative of f(x, y, z) = x
2
- 2xy + z
2
at
(d3, l , n / 3 ) on the helix x =2 s i n t, y = 2 c o s t, z = t ( t r 0).
17. Find the tangential derivative of f(x, y, z) = ex y2z on the
curve R(t) = (t - l ) i - t2j + 2tk when t = 1.
3.4 GRADIENT OF A SCALAR FIELD
Let f be a scalar field defined in a domain D, and suppose that f
is differentiable in D. We have seen that the directorial derivative
of f at a point P in the direction of a unit vector u = cos ai + cos Pj +
cos yk is given by
d f a f a f a f
(P) = - ( P) cosa + (P)COSP + ( P ) c o s y
ds ax ay dz
If we introduce the vector
a f a f a f
grad f(P) = ; -(P) i + ;-(P) j + az(P) k
dx dy
whose components are the partial derivatives of f at P, then the ex-
pression (3.1 1) is simply the dot product of grad f(P) and u, that is,
d f
( P ) = grad f(P). u
ds
Differential Calculus of Fields 161
The vector field, grad f, is called the gradient of the scalar field f. It is
one of the basic concepts in vector analysis that is of great impor-
tance in application.
The gradient of a scalar field f is also commonly written as Vf,
where V denotes the vector differential operator
In other words, grad f can be interpreted as the result of applying the
differential operator V to f. The operator V plays a similar role in
vector analysis as the operator D = d/ dx in calculus. Indeed, using
the operator V together with the familiar rules on differentiation,
we can easily establish the following properties:
Theorem 1. Let f and g be scalar fields which are differentia-
ble in a domain D. Then we have
(a) V( f +g) =Vf + Vg
(b)
V(cf) = cVf, where c is any constant (3.1 5)
(c) V(fg)=fVg+gVf
(d) V(f/g) = (gVf - fvg)/?, provided g # 0 in D.
For example, to prove part (c), we proceed as follows:
The operator V has many other applications, and we shall come
across them in our discussion later.
Chapter 3
Geometric Properties of Grad f
We now turn our attention to some geometric properties of
the gradient. Supoose that Vf(P) z 0, and let 8 denote the angle be-
tween Vf(P) and the unit vector u. Then, from the geometric prop-
erty of scalar product, we have
= I Vf(P) I cos 8 (3.1 6)
This shows that the directional derivative of a scalar field f at a point
P in the direction of a unit vector u is simply the component of the
gradient vector Vf(P) at P along u. This component is positive or
negative according as 0 .: 0 < x 12 or n 12 < 8 I n ( see Fig. 3.4 ).
By examining equation (3.16), we see that the directional deri-
vative is maximum when cos 8 = 1, that is, when u is in the same
direction as Vf(P). The maximum value is given by IVf(P) 1 . Since
(Vf(P) 1 z 0, unless f is identically zero, it follows that f is increas-
ing in the direction of the vector Vf(P). In other words, at a point T:
the scalar field f experiences its nlaximum rate of increase in the
direction of the gradient vector Vf(P).
Fig. 3.4 The gradient vector Vf (P).
Differential Calculus of Fields
Example 1. Let f(x, y, z) = x
2
y + z + 1. Determine the
direction in which the directional derivative of f at (2, 1, 3) is max-
mum, and find that maxiniuni value. What is the minimum value
and in what direction does it occur?
Solution: The maximum value of the directional derivative
of f at (2, 1, 3) occurs in the direction of the gradient vector Vf(2, 1, 3).
Since
grad f = 2xyi + (x
2
+ 2yz)j + y%
we find Vf(2,1,3) = 4i + 10j + k. Thus the maximum value of the
directional derivative is I Vf(2,1,3) 1 = dl 17.
The niininium value of the directional derivative occurs in
the opposite direction to the gradient vector Vf(2, 1, 3), that is, in the
direction - 4i - I Oj - k and the minimum value is - dl 17.
Example 2. Suppose the tem erature distribution in a metal
!?
ball is defined by T(x,y,z) = a(x
2
+ y + z
2
), where a is a positive
constant. What is the direction of maximum cooling?
Solution: The maximum rate of increase of the temperature
occurs in the direction of the vector
grad 7' = 2a(xi + yj + zk) = 2aR
where R is the position vector of the point (x, y, z). Hence, niaxi-
mum cooling occurs in the direction opposite to grad T, that is, in
the direction - R, which is toward the origin.
The gradient vector has another geometric significance. Sup-
pose S is a level surface of a scalar field f which is differentiable in a
domain D. Let Po be a point on this surface, and consider a curve C
which lies on S and passes through the point Po (Fig. 3.5). Assume
that the curve is described by the vector equation R(s) = x(s)i + y(s)j +
z(s)k, where s denotes arc length such that s = so corresponds to the
point Po. Since C lies on S, we have f(x(s), y(s), z(s)) = c, where c is a
constant (in fact, c = f(Po)). Hence, taking the derivative of f with
respect to s by the chain rule, and setting s = s, we find
Chapter 3
This implies that the gradient vector Vf (Po) is orthogonal to the
tangent vector dR/ ds at Po (see Fig. 3. 5). Since (3.17) holds for all
curves on S passing through the point Po , we conclude that Vf (Po)
is orthogonal to the tangent vectors of all these curves. These tan-
gent vectors at Po determine a plane which is called the tangent
plane to the surface at P,, If Vf (Po) * 0, then Vf (Po) is normal to the
tangent plane at Po. Therefore, given a level surface f(x, y, z) = c, we
conclude that the gradient of f, Vf(P), is orthogonal to the surface at
the point P on the surface. The vector
is then a unit normal vector to the surface.
If R* denotes the position vector of a point on the tangent
plane and R, is the position vector of the point P, then an equation
of the tangent plane is given by
(R* - R,).Vf (Po) = 0
In cartesian coordinates, this becomes
Fig 3.5 Grad f normal to the surface f(x, y, z) = c.
Differential Calculus of Fields
If a level surface is closed so that it encloses a domain (e. g., a
sphere), then the unit normal vector pointing away from the en-
closed domain is called the outward unit normal vector on the
surface. The directional derivative of a scalar field on a level surface
in the direction of the outward unit normal vector is called the
outzuard 110rnzal d e r i z ~a t i z ~~ of the scalar field on the surface. Thus,
i f the surface has the representation f(x, y, z) = c so that its outward
unit normal vector is n = Vf/ lVf 1 , then by definition the outward
normal derivative of a scalar field g on the surface is given by
Example 3. Find the outward unit normal vector on the
sphere x
2
+ y2 + z
2
= 9, and determine the equation of the tangent
plane at the point (2, 1, 2).
Solution: Let f(x, y, z) = x
2
+ y2 + z
2
. We find Vf = 2xi + 2yj +
2zk so that the outward unit normal vector on the sphere is
This is in the same direction as the position vector R of the point
(x, y, z). Since Vf(2, 1,2) = 4i + 2j + 4k, the equation of the tangent
plane is given by
Example 4.
Find the outward normal derivative of g(x, y, z) =
xy + y z + r x on thesurface x 2 s Y2 + z = 1 .
Solution: Here we take the outward normal vector to be
pointing away from the domain bounded by the paraboloid as
shown in Fig. 3.6. Letting f(x, y, z) = x
2
+ y2 + z, we find
166
so that
Chapter 3
Thus, since Vg = (y + z)i + (z + x)j + (x + y)k, the outward normal
derivative of g on the surface is given by
In particular, at the point (-1 12, -1 12, 112) on the surface, we find
The negative sign indicates that g is decreasing along the outward
normal vector.
Fig. 3.6 Outward normal vector on a paraboloid.
3.3 EXERCISES
1. Find the gradient vector at the given point for the scalar fields
defined by the following functions:
Differential Calculus of Fields
(a) f(x,y) = x
2
cos y + y , (1,O)
(b) f(x, y) = x sin y + y cos x , (x/4, x/ 4 )
(c) f ( x, y) = arct an(y/ x), (1,-1)
(d) f ( xf y, z) =x 2 + y~ + z
2
, (1,-1, I )
(e) f ( ~ , y , z ) = l n ( x - + ~ ~ - 2 z ~ ) , (2,-2,l)
(f) f ( x, y, z) =xye
X
+ y l n z , (-1, -1, 1)
Find the maximum value of the directional derivative of each
of the scalar fields at the given point:
(a) f(x, y) = x 2 y + ( ~ - y ) ~ , (1r1)
(b) f(x, y) = ln (x2+ 2y2) , (-1,1)
(c) f (x, y , z) = x2 + 2y2 - 3z
2
, (2, - 1, 1)
(d) f(x, y , z) = z arctan (y / x) , (-1,1,2)
Find the points (x, y) on the unit circle x
2
+ y2 = I at which the
tangential derivative of f(x, y) = 2x2 + y2 is (a) maximum;
(b) minimum. What are these values?
Find the outward normal derivative of each of the following
scalar fields at the given point on the given surface:
(a) f(x, y , z) = x
2
+ yz + z at (-2,2, 1) on the sphere
x'+ Y2+ z2= 9
@) f(x, y , z) = xye
X
+ y In z at (-1, -1, 1) on the ellipsoid 2x
2
+
2v2 + 3z
2
= 7.
Find i he directional derivative of f(x, y , z) = x
2
+ y2 + z2 on the
surface x
2
+ y2 = 1 - z along the normal vector which makes an
acute angle with the positive z-axis.
Find the outward normal derivative of f(x, y , z) = xy + yz + zx
on the facesof the cubex = + 1, y = 2 1, z = + 1.
Find the cartesian equation of the tangent plane to the surface
x
2
+ y2 = z - 1 at the point (-1, 1, 3).
Find a tangent vector to the curve of intersection of the plane y
- z + 2 = 0 and the cylinder x
2
+ y2 = 4 at the point (0,2,4).
Find a tangent vector to the curve of intersection of the given
surfaces at the indicated point, and determine the angle at
which the corresponding tangent planes to the surfaces meet at
that oint:
5
(a) x + y2+ z
2
= 9 and y2+ z 2= x + 3 at (2, -1,2)
@) x
2
+z
2
=4 and y 2 + z 2 = 4 at (\ 12, -d2, d2)
Find the value of b so that at each point on the curve of inter-
section of the two spheres
( x - I ) ~ + y2+ z
2
= 1, x
2
+ ( ~ - b ) ~ + z
2
= 4
the two corresponding tangent planes will be perpendicular.
Chapter 3
11. Find the vertex of the cone x
2
+ y2 = (z - a12 so that at each
point of its intersection with the paraboloid x
2
+ y2 = z the two
corresponding tangent planes will be perpendicular.
12. Find an equation of the tangent line to the curve of intersec-
tion of the surfaces x
2
y2 + 2x + z
3
= 16 and 3x
2
+ y2 - 2z = 9 at
the point (2,1,2).
13. Show that the sphere x
2
+ y2 + z
2
- 8x - By - 62 + 24 = 0 and the
ellipsoid x
2
+ 3y2 + 2z
2
= 9 are orthogonal at the point (2,1,1).
14. Let u = 1 /2 ln(x
2
+ y2) and v = arctan (y/x). Show that the an-
gle between Vu and Vv is a constant.
15. L e t f ( x , y , z ) = ~ ~ + ~ ~ + z ~ , a n d l e t u = u l i + u 2 j + u 3 k a n d v =
vli + v2j + v3k be two constant unit vectors. If @(u) denotes
the directional derivative of f in the direction u , show that
the directional derivative of @ in the direction v is equal to
2u0 v.
16. Let f be a differentiable function of u, where u = @ (x, y, z), @
being a differentiable scalar field. Show that
17. Let f be a differentiable function of r, where r = d(x2 + y2 +z
2
).
(a) Show that Vf(r) x R = 0, where R denotes the position vec-
tor of (x, y, z).
(b) Find Vr
n
, where n is a positive integer.
(c) If Vf is always a scalar multiple of R, show that f = f(r).
18. Let f(x, y, z) = (R x A)* (R x B), where A and B are constant
vectors. Show that
Vf =B x(R x A) + A x( RxB)
19. Prove the rest of Theorem 1.
3.5 DIVERGENCE OF A VECTOR FIELD
There are two basic concepts that arise in connection with the
spatial rate of change of a vector field F, namely, the divergence of
F, denoted by div F, and the curl of F, denoted by curl F. Although
these concepts can be defined in terms of their physical interpreta-
tions, they can also be defined purely on analytical ground in terms
of the del operator
Differential Calculus of Fields 169
analogous to the definition of the gradient of a scalar field. In this
section we shall first introduce the analytic definition of the diver-
gence of a vector field and then consider its physical interpretation.
The definition of the curl of a vector field will be considered in the
next section.
So let F (x, y, z) = P(x, y, z)i + Q(x, y, z)j + R(x, y, z)k be a vec-
tor field defined in a domain D, where Q, and R have continuous
first order derivatives. The divergence of F, denoted by div F, is the
scalar field defined by
I f we treat the del operator V formally as a vector and take the dot
product of V and F, we find
Thus,
div F = V' F
It should be noted, however, that since V is actually not a vector,
he dot product of V and F is not commutative, that is, F . V z
V - F. As a matter of fact, the symbol F - V represents the operator
which we shall discuss later.
From the definition and by the rules on differentiation, the
following properties of the divergence can be easily verified:
(a) div (F + G) = div F + div G
(3.20)
(b) div (f F) = f div F + grad f . F
170 Chapter 3
where f is any differentiable scalar field. In terms of the del operator
V, these can be written as
If F = Vf = (df /ax) i + (df / dy) j + (df / dz) k, for some scalar field f,
t hen
a af a af a af
div F = V-Vf = ax (ax) + ay(ay) + az (az)
This expression is known as the Laplacian of the function f, denoted
byv2f. Thus, v2f = div(grad f). The Laplace equation v2f = 0 is one of
the important partial differential equations of mathematical physics.
Example 1. Determine the divergence of (a) F = x2 y i + eYz j
+ x s i nz k and (b)F = gradf, where f =y 2 z + z
2
x.
Solution: (a) We have
= 2xy+z e Y+xc os z
(b) Since
we have
a 2 a a 2
div F = ( Z ) + (2yz)+ d z ( 2 ~ ~ + y ) = 2 x + 2 z
ax ay
Example 2. Let f(x, y,z) = x 2 y z and F(x, y, z) = x z i + y2j
+ x% k. Verify property (b) of (3.20).
Solution: We note that Vf = 2xyz i + x% j + x? k and div F = z
Differential Calculus of Fields
+ 2y + x
2
. Hence
f div F + Vf F = ( X ~ ~ Z ) ( Z + 2y + x
2
) + (2xyz)(xz) + (x
2
z)y2
On the other hand, we find
div (f F ) = 3x2yz2 + 3x2y2z + 2x4yz
thus verifying property (b) of (3.20).
Physical Application of the Divergence
The concepts of the divergence and the curl of vector fields are
of fundamental importance in the study of fluid flow. Here we con-
sider the significance of the divergence in a fluid flowing in a
domain of our space. Let V(x, y, z, t) denote the velocity of the fluid
at the point (x, y, z) at time t, and let p(x, y, z, t) denote the corres-
ponding density (mass per unit volume). Then the vector field F(x,
y, Z, t) = p(x, y, Z, t)V(x, y, z, t) represents the rate of flow (mass per
unit area per unit time) of the fluid. This is called the flux density of
the fluid. If the fluid has no sources, that is, points at which fluid is
being introduced, or if it has no sinks, that is, points at which fluid is
being carried away, then it can be shown (Sec. 4.10) that the diver-
gence of F satisfies the equation
div F = div (pv) = - +/at (3.21)
This equation is known as the cont i nui t y equation .
Equation (3.21) says that the divergence of the flux density is
related to the rate of change of the density of the fluid. Hence, if
div F > 0 so that +/at < 0, then the density p is diminishing, which
means that the fluid is expanding. On the other hand, if div F < 0
so that 3p/at > 0, then the density is increasing, which means that
the fluid is being compressed. If the flow is steady, that is, the den-
Chapter 3
sity does not change with time, then (3.27) becomes
div (pV) = 0 (3.22)
If the fluid is incompressible, that is, 6 ) = const., of which water is a
fair example, then (3.21) simplifies to
div V = 0 (3.23)
Therefore, for an incompressible fluid flow without source or sink
the divergence of its velocity field vanishes. Equation (3.23) is
known as the co?zditio~z of i~zconryressibility .
In the case of electrostatic fields, the sources and sinks are elec-
tric charges. If a single electric charge of magnitude q (in appropriate
units; for example, coulombs) is located at the origin, then the elec-
trostatic field at a point (x, y, z) is given by (assume vacuum me-
di um)
where R is the position vector of the point (x, y, z) and r = I R I . This
can be written as E = (q/?)u,, where ur = R/ r is a unit vector. By
direct differentiation, we find
so that
div E = l x ] + [ I + I Z]
dxr 3 dyr 3 a z r 3
Differential Calculus of Fields
for all (x, y, z) + (0, 0, 0). Thus the divergence of an electrostatic field
vanishes at each point where no souce or sink is present.
If there are N electric charges of magnitude qi located at the
points (xi, yi, zi), i = 1, . . , N, then the electric field at a point (x, y, z)
is given by the sum
where Ri denotes the position vector of the point (xi, yi, zi ) , i = 1, . .
, N. Again, at each point (x, y, z) # (xi , yi , zi), i = 1, . . , N, where
there is no electric charge present, it is easily verified that div E = 0.
A vector field F which satisfies div F = 0 is called a sol e?zoi da 1
field. A vector field F for which there exists a scalar field f such that
F = grad f is called a conservative field. In such a case, the scalar
field f is called a potential function of F. Note that i f f is a potential
function of a conservative field F, so is f + C, where C is any cons-
tant. This is clear since grad (f + C) = grad f = F. Thus, the potential
function of a conservative field is unique only up to an additive
constant. I f a conservative field is also solenoidal, then its potential
function satisfies the Laplace equation, that is,
div F = div(grad f) = v Z ~ = 0
Later we shall present a method for determining whether or
not a given vector field is conservative. In the next example we
illustrate a method for finding a potential function for a vector field
that is conservative.
Example 3. Determine a potential function for the conserva-
tive field
F = (yz + e
X
cosy) i + (xz - e
X
si ny) j + (xy + 22) k
Solution: Let f be a potential function of F. Since Vf = F, we
equate the components of Vf and F to obtain the three equations
fx = yz + ex cos y , f = xz - ex sin y , f z = xy + 22
Y
To deternine f, we first integrate the first equation partially with res-
174 Chapter 3
pect to x (treating all other variables as constants). We find
f = xyz + eXcos y + h(y, z)
where the constant of integration h depends on y and z. Next, we
differentiate f with respect to y and equate the result with the second
equation above. We find
This implies h - 0, so that h = h(z). Finally, differentiating f with
Y -
respect to z and equating it with the third equation, we obtain
Hence h
f
(z) = 22 or h(z) = 9 + C. Therefore, the potential function
is given by
f = x y z + e X c o s y + z 2 + ~
Example 4. Find a potential function of the electrostatic field
which is known to be a conservative field.
Solution: As in the preceding example, we determine the
potential function f from the three equations
Integrating the first equation with resspect to x , we find
If we differentiate this with respect to y and equate the result with
the second equation above, we find that h (y,z) = 0. Hence h depends
Y
Differential Calculus of Fields 175
only on z, that is, h = h(z). Finally, differentiating f with respect to z
and using the third equation, we find hl(z) = 0. Thereore, h is a con-
stant and so the potential function is given by
On physical ground it is cust on~ary to take C = 0 so that f = q/r.
EXERCISES
Find the divergence of each of the following vector fields.
F(x, y) = e
X
(cos yi + sin yj)
~ ( x , y) = -x/(x
2
+ y3i + y/(x2 + y9j
F(x, y, z) = (x2 + y2)i + zexYj
F(x, y, z) = (1 + y2)i + xe
Z
j
F(x, y, z) = xe
Z
i + ye
X
j + z e k
F(x, y, z) = x sin yi + y sin (xz)j + cos(e
z
)k
F(x, y, z) = (xi + yj + zk)/(x
2
+ y2 + z2)lI2
Prove the properties given in (3.20).
Let u and v be twice continuously differentiable functions.
Show that v2 (uv) = uv2v + 2Vu. Vv + vv2 u.
If F = r
n
R, where R = xi + yj + zk and r = IR I , find div F
and show that grad (div F ) = n(n + 3)rnm2 R.
Find the gradient of the divergence of F = xyi + yzj + zxk.
Let R1 and R2 denote the position vectors of the fixed points
(x,, y,, z,) and (x2, y2, z2), respectively, and let R be the position
vector of a variable point (x, y, z). If F = R - R, and G = R - R2,
show that V(F- G) = F + G and V' (F x G) = 0.
Let f be a differentiable scalar field, and let F be the vector
field
Show that F = R x V f , F . R = O and F - V f =O.
Chapter 3
For each of the following conservative vector fields, determine
a potential function $.
14.
F = (xi + yj)/(x2+y2)
15. F = ( - y i + X ' ) / ( X ~ + ~ ~ )
1
16. F = (2xy + z )i + (2yz + x
2
)j + (2xz + y3k
17. F = e
X
c o s z i + 2 y j - e
X
s i n z k
18. F = (eY + z
2
cos x)i + (xeY+ 2yz in z)j + (y2 + y2 in z + 22 sin x)k
3.6 CURL OF A VECTOR FIELD
Let F(x, y, z) = P(x, y, z)i + Q(x, y, z)j + R(x, y, z)k be a vector
field defined in a domain D, where Q, and R have continuous
first order derivatives. The curl of F, written as curl F, is the vector
field defined by the equation
This equation can be easily remembered as the expansion of the
determinant
by the elements of the first row, where each product such as d/dy
and R is to be interpreted as the partial derivative dR/dy. Now since
(3.25) resembles the forn~ula for the cross product of two vectors
with d/ dx, d/dy, d/dz being the "components" of the "vector" V, we
can write
curl F = V x F (3.26)
curl F =
Thus we see that the gradient of a scalar field f, the divergence and
the curl of a vector field F can be interpreted formally as the result
of taklng the scalar multiple Vf, the dot product V'F and the cross
i j k
d d d
~ J X dy az
P Q R
Differential Calculus of Fields 177
product V x F, respectively, of the del operator V and the fields f and
F.
It follows from the definition and the rules on differentiation
that the curl satisfies the following properties:
(a) curl (F + G) = curl F + curl G
(3.27)
(b) curl (f F) = f curl F + grad f x F
where f is a differentiable scalar field. The student should compare
these properties with those listed in (3.20) for the divergence. In
terms of the del operator V, these properties can be written as
Vx ( F + G) = Vx F + V x G
Vx ( f F) = f V x F + Vf x F
Example 1. Find the curl of the vector field
F = xy2i + x sin (yz)j + z
2
eJ'k
Solution: We have
= [z2eY-xycos (yz)]i + (0)j + [sin (yz) -2xylk
= [z
2
eY- xy cos (yz)]i + [sin (yz) - 2xyIk
curl F =
Example 2. Show that the curl of F = (-yi + xj)/ (x
2
+ y2)
vanishes for all (x, y) + (0, 0).
Solution: Since F is independent of the variable z and its third
component is zero (R= 0), we readily find
i j k
a a a
ax a y a z
2 2~
xy x sin (yz) z e
Chapter 3
Alternatively, we can perform the above calculation by using
(3.27%) with G = - y i + x j and f = 1 / (x
2
+ y2). In this case we see that
curl G = 2k, g a d f = - (2x i + 2y j)/(x
2
+ y32 and
grad f x G = [-2(x
2
+ y2)/(x2+ y212]k = [-2/(x
2
+ y2)]k
Thus
c u r l ( f G) = f c u r l G+g r a d f x G
= [2 / (x
2
t y2)]k - [2 / (x
2
+ y2)] k = 0
Example 3. Let A be a constant vector and R the position
vector of the point (x, y, z). Show that curl (A x R) = 2A .
Solution: Let al, a2 a, denote the components of the vector A.
Since
Ax R = ( a2z- a3y) i + (a3x - al z) j + (sly- a2x)k
we find
curl (A x R) =
I
Physical Application of t he Curl
Roughly speaking the curl can be interpreted in fluid motion
as a measure of the tendency of the velocity field to cause a rotation
at a point. For example, let us consider the motion of a fluid that is
in pure rotation about a fixed axis with an angular speed m. As we
Differential Calculus of Fields 179
saw in Sec.l.7, the rotation can be described by a vector w of magni-
tude w pointing in the direction of the axis of rotation determined by
the right-hand rule. Assuming that the rotation is about the z-axis,
we then have w = wk. Hence the velocity field V is given by
V = w k x R = w ( - y i + x j )
Now let us take the curl of the velocity field V. We find
Thus the curl of the velocity field of a fluid in pure rotation about
curl V =
the z-axiz is twice the angular velocity w . Moreover, we have div V
= 0, so that the velocity field is solenoidal.
The graphical view of the velocity field V as shown in Fig. 3.7
suggests that we may describe the velocity field geometrically as
"curling".
i j k
a d d
dx ay az
- w y wx 0
I...,,.
I . . . ,
q7Z.,
4
Fig. 3.7 Graphic view of the velocity field V = w (- y i + x j).
180 Chapter 3
In contrast suppose the velocity field of a fluid flow is described
byV = (xi + yj + z k) / r =R/ r , where r = \/(x2+ 42 + z3. Then, since
curl R = 0, grad (1 / r ) = - ~ / r ' and R x R = 0, we find
curl V = (1 / r) curl R + grad (1 / r) x R = 0
Thus the vector field does not "curl". However, if we take its diver-
gence, we find
The graphical view of the velocity field V is shown in Fig. 3.8, which
suggests that we may describe the field geometrically as "diverging".
Fig. 3.8 Divergent vector field.
3.5 EXERCISES
In each of Problems 1 through 6, find the curl of the given
vector field.
1.
F(x, y) = e
x
(sin yi + cos yj)
2. F(x, y)= ( - y i + ~ j ) / ( x ~ + ~ ? l / ~
3. F(x, y, z)= ~ ~ ~ z ~ ( i + j + k)
4.
F(x, y, z) = (y z
2
- 2x)i + (X z2 + 2y)j + (2xy + x
2
)k
Differential Calculus of Fields
and
F(x, y, Z) = xyz% + 2xyq - x2 yzk
F(x, y, z) = (4xy - z
3
)i + 2x3 - 3xz%
Prove the formulas in (3.27).
In each of Problems 8 through 11, let F(x, y, z) = yi + zj + xk
G(x, y, z) = x
2
zi + y2 xj + z
2
yk.
Calculate div(F x G).
Calculate curl (F x G).
Calculate F x curl G.
Calculate curl F x curl G.
Let f be a differentiable function of r, where r = (x
2
+ y2 +
z ~ ) ' / ~ . Show that curl (f(r)R) = 0 , where R is the position vec-
tor of (x, y, z).
Suppose that u and v are differentiable scalar fields satisfying
a functional relation f(u,v) = 0. Show that Vu x Vv = 0, assum-
ing that f is differentiable.
Let F (x, y, z) = x2yz i - x y j + xyz
3
k. Calculate div (curl F).
Let F (x, y, z) = 2xyz i + x z j + x
2
yk. Show that curl F = 0,
and thus find all scalar fields f so that grad f = F.
Show that curl [(A. R)R] = A x R for any constant vector A.
Show that curl [(A- R)B] = A x B for any constant vectors A
and B.
3.7 OTHER PROPERTIES OF THE DIVERGENCE
AND THE CURL
Let f be a scalar field with continuous second-order deriva-
tives. We have already seen that
so that if f is a harmonic function, then v2f = 0. On the other, if we
take the curl of the gradient of f, we find
Chapter 3
Since the second-order derivatives of f are continuous, the mixed
derivatives are independent of the order of differentiation, that is,
f = f etc. Therefore, we have
ZY YZ '
curl (grad f) = V x Vf = 0 (3.29)
for any twice continuously differentiable scalar field f. Thus the gra-
dient of a scalar field is irrotational.
Equation (3.28) implies that if a vector F is conservative, so that
F = Vf, for some scalar field f, then curl F = 0. Hence a necessary
condition for a vector field to be conservative is that its curl must
vanish. For a special type of domain, it turns out that the condition
curl F = 0 is also a sufficient condition for F to be a conservative
field, that is, there exists a scalar field f such that grad f = F. A more
detailed discussion of this topic is given in the next chapter.
A counterpart of equation (3.28) for the divergence is that if F
is twice continuously differentiable, then
div (curl F) = V' (V x F )
Thus the curl of a vector field is solenoidal. Under certain condi-
tion on the domain where F is defined (for example, a spherical
domain), it turns out that a solenoidal field is the curl of some vec-
Differential Calculus of Fields 183
tor field (see Problem 1). That is, suppose div F = 0, say in a sphe-
rical domain D, then there exists a vector field V in D such that F =
curl V. Moreover, it can be shown that V is determined up to a
term Vf, where f is an arbitrary differentiable function (see Problem
3). Therefore, for certain types of domains a solenoidal field is the
curl of some vector field just as a conservative field is the gradient
of some scalar field.
In (3.28) and (3.29) we calculated the divergence and the curl of
the vector field grad f. In (3.30) we have the divergence of the vector
field curl F. To complete the analysis, we now compute the curl of
V x F, that is,
curl (curl F ) = V x (V x F )
Although this can be evaluated strictly from the definition of a curl,
we introduce here a method which makes use of the algebraic prop-
erties of the vector operations of dot product, cross product and vec-
tor triple product. The method consists in treating the del operator
V as though it were a vector and applying known vector identities.
In doing this, however, we have to make sure that the operator V
applies only to the factor it is supposed to operate on and that it
produces the appropriate kind of field (a scalar or a vector field).
We illustrate the procedure by considering the expression V x
V x F) bearing in mind that this results in a vector field. By apply-
ing the vector identity
A x (B x C) = (A. C)B - (A. B)C
= B (A. C) - (A. B)C
identifying A and B as V, we obtain
V x V x F) = V(Vm F ) - VmV)F = grad (div F ) - V% (3.31)
Here V ~ F denotes the Laplacian of F which is given by
V ~ F = v2(Pi + Qj + Rk) = (v2p)i + ( v 2 ~ ) j + ( v ~ R ) ~
Notice that V(V . F ) # (V . F )V since V is not a vector. The student
should verify this result from the definition of the curl of a vector
field (see Problem 4).
184 Chapter 3
We now list a few other important properties of the gradient,
divergence, and curl, expressing them in terms of the differential
operator V:
(a)
V(F* G) = (F* V)G + (G.V)F + F x (V x G) + G x (V x F)
(b) Ve ( Fx G) = G V x F - F g Vx G (3.32)
(c) V x (F x G) = (G. V)F - (Fa V)G + (V' G )F - (V. F )G
where F and G are differentiable vector fields.
These identities can be proved by the same method we used in
establishing the identity (3.31). For example, to prove (3.32b) we
write
where the symbol VF means that the operator V applies only to the
factor F, and Gi s to be treated as a constant. A similar interpreta-
tion holds for the symbol VG. This is in accordance with the rule for
differentiating product of functions. Now, to calculate the terms on
the right-hand side of (3.33), we make judicious use of the vector
identity
A- Bx C = B a C x A = C - A x B (3.34)
Since VF is to operate only on the factor F, we adopt the last mem-
ber of (3.34) to obtain
VF * ( F x G) = G. Vx F
Similarly, we have
VG. ( FxG) = - VG. ( Gx F) =- F. Vx G
Substituting these on the right-hand side of (3.33), we obtain the
identity (3.32b). The student should try to prove the rest of (3.32) by
following the same procedure.
3.6 EXERCISES
1. Let V = xi + 2yj - 3zk. Verify that div V = 0 for all (x, y, z)
and hence find a vector field F so that curl F = V.
[Hint:
Assume F(x, y, z) = P(x, y, z)i + Q(x, y, z)j and then solve the
Differential Calculus of Fields
system Q, = - x, P, = 2y, Q - Py = - 32.1
2. Repeat Problem 1 when V = yzi + zxj + xyk.
3. Let V be a given vector field for which div V = 0 in a spheri-
cal domain D. If F is a vector field such that curl F = V, show
that F is determined up to a term Vf, where f is a differentia-
ble scalar field; that is, i f curl F1 = curl F2 = V, then F, = F2 + Vf.
4. Verify the formula (3.31) from the definition of the curl.
5. Using the vector identity for A x (B x C), show that
(a) VF(Fe G ) = G x V x F) + (Go V)F
(b) VFx ( FxG) =(G. V)F - ( VSF) G
6. Prove part (a) of (3.32). [ Hint : Use part (a) of Problem 5.1
7. Prove part (c) of (3.32). [ Hint : Use part (b) of Problem 5.1
8. Show that V' (Vf x Vg) = 0 .
9. Show that (F. V)F = (1 /2)V(Fe F) - F x (V x F)
10. Show that div(f grad g - g grad f) = f~~~ - g~2f
11. If f(x, y, z) = (R x A). (R x B), where R is the position vector
of (x, y, z) and A, B are constant vectors, show that
grad f = B x (R x A) + A x (R x B)
12. Compute V(F. G x H).
13. Compute V' [F x (G x H)].
14. Compute V x [F x (G x H)].
3.8 CURVILINEAR COORDINATE SYSTEMS
So far we have studied the various field properties - gradient,
divergence, and curl - and their analytic representations in a rectan-
gular cartesian coordinate system. In many applications of vector
analysis, however, it often becomes necessary to use coordinate sys-
tems other than the cartesian. In such situations, the representa-
tions of the gradient of a scalar field, the divergence and the curl of a
vector field take on completely different forms. In this section we
shall discuss the idea of a curvilinear coordinate system in yrepa-
ration for the task of deriving the corresponding expressions of the
gradient, the divergence, and the curl in such a coordinate system.
Chapter 3
Let ul , u2, u3 denote new coordinates, and suppose that they
are related to the cartesian coordinates x, y, z by the equations
u1= e,(x, y, 4 , u, = O2(x, y, z) 1 u3 = O,(x, y, z)
(3.35)
or briefly, u, = Qi(x, y, z), i = 1, 2, 3. We assume that the functions
$(x, y, z), 1 5 i s 3, have continuous first-order derivatives in some
domain D of the xyz-space, and that at some point PC,: (x,, yo, z,) in
D the condition
holds. The determinant on the r-ight-hand side of (3.36) is called the
Jacobian of the functions gl, 4, @,
The condition (3.36) ensures that in some neighborhood of the
point P, it is possible to determine x, y, z in terms of the coordi-
nates ul , u2, u3; that is, there exist functions f , g , h such that
where f, g, h are defined in a neighborhood of the point (uol, uo2,
uo3) determined from (3.35) by the point ( x , y , z ) Moreover,
in that neighborhood f, g and h also have continuous first deriva-
tives for which
at (uol, uo2, uo3) The functions f, g , and h define the inverse
transformation of (3.35). It is important to note that the Jacobians
(3.36) and (3.38) satisfy the relation (see Problem 4)
Differential Calculus of Fields 187
Now let P be any point in D with the coordinates (x, y, z), and
let the numbers u,, u2 u3 be determined by (3.35). We call the or-
dered triple of numbers (u,, u2 u?) the curvilinear coordinates of the
point l? The equations given in (3.35) are called the coordinate
transformation, and they are said to define a curvilinear coordinate
system in D. It follows from (3.39) that the Jacobian of a coordinate
transformation is the reciprocal of the Jacobian of its inverse.
Example 1. Consider the transformation from the rectangular
cartesian coordinates (x, y) on a plane to the polar coordinates (r, H)
defined by
2 2
r = \ x + y
X
H = arccos = arcsin
Y
2 2 2 2
x + y \ x + y
[Thus H is the unique angle 0 i 0 < 2n such that cos 0 = x/ d(x2+ y2)
and sin H = y/\i(x2 + y?]. The transformation is defined for all (x, y)
except at (0, 0) where 0 is not defined. The Jacobian of this transfor-
mation is given by
Hence, except at the origin, the transformation has an inverse, and it
is given by
x=r cos C) , y = r s i n 0
This inverse transformation is defined, however, for all (r, 0).
By direct calculation, we find
Chapter 3
cos 8 - r sin 8
sin 8 r cos8
It is clear that
in accord with (3.39).
Coordinate Surfaces and Coordinate Curves
Let us now examine the nature of a curvilinear coordinate
system from a geometric point of view. Suppose P, is a point in
space with curvilinear coordinates (uol, uo2 uo3) as defined by the
coordinate transformation (3.35). Then the equations
define three surfaces in space each of which passes through the
point P,, (Fig. 3.9). These surfaces correspond to the three coordinate
planes x = x,, y = y, , z = z, in the cartesian coordinate system, where
(x, y, z,) are the cartesian coordinates of Po. Recall that by the trans-
formation (3.35), we have uoi = #(x, y, z,), i = 1, 2, 3. Accordingly,
we call the three surfaces described by (3.40) the coordinate surfaces
intersecting at the point Po. The curves of intersection of these
coordinate surfaces correspond to the coordinate axes in cartesian
coordinates. Thus, for example, the surfaces q2(x, y, z) = uo2 and
$?(x, y, z) = uo3 intersect in the curve on which only the coordinate
ul varies. Hence we call this curve the ul-coordinate curve or sim-
ply the ?-curve. The u2- and u3-curves are defined similarly and
they are shown in Fig. 3.9.
Using the inverse transformation (3.37), the position vector of
a point in curvilinear coordinates now has the representation
Differential Calculus of Fields
Fig. 3.9 Curvilinear coordinate surfaces and curves.
If we set u2 = c2 and u1 = c3, where %, c, are constants, then the re-
sulting vector function R(ulr 5 c.$ represents the ul-curve in which
the two coordinate surfaces q2(x, y, Z) = c2, +,(x, y, Z ) = c1 intersect.
On this curve ul is the parameter. It follows that the derivative
dR/dul represents the tangent vector to this curve. Likewise, we
have dR /du2 and dR/du, representing the tangent vectors to the u2-
and u3-curves, respectively (see Fig. 3.9).
Since the determinant of a matrix is the same as the determi-
nant of its transpose, it follows from the definition of the Jacobian
(3.36) and that of the scalar triple product that
Chapt er 3
Hence, at each point where (3.42) is not zero the three tangent vec-
tors dRldul, dR/du2, dR/du, to the coordinate curves are linearly
independent and thus form a basis. Every vector or vector field at
each point can then be represented in terms of this basis. Unlike the
unit vectors i, j, k, however, this new basis varies from point to
point in space. For this reason we call (dR/dul, dR/du2 dR/ duS a
local basis.
Ort hogonal Curvilinear Coordinates a nd Arc Lengt h
Henceforth, we assume that the coordinates ul , u2, u, are so
labeled that the base vectors dR/ du, , dR/ du2, dR /du3 , in that order,
form a right-handed triple.
This implies that the Jacobian d(x, y, z) /
d(ul, u2, u? ) is positive in a domain D* of the (ul, u2, u3)-space.
A curvilinear coordinate system defined by (3.35) or (3.37) is
said to be ortlzogorzrrl if at each point the base vectors dR/dul,
dR/ du2, dR/du, are mutually orthogonal. This means that although
the vectors may vary in magnitude and direction from point to
point, they remain orthogonal to each other at each point. When
this is the case, the formulas for the gradient, the divergence, and
the curl appear very much like their representations in rectangular
cartesian coordinate system. We shall derive these representations
in the next section. As we shall see the polar coordinates in t wo
dimensions and the cylindrical and spherical coordinates in three
dimensions are all orthogonal curvilinear coordinate systems.
We conclude this section by deriving the formula for arc
length in a curvilinear coordinate system. Let R(ul, u2, u3) denote
the position vector of a point in curvilinear coordinates. Then by
definition and by the chain rule, we have
Differential Calculus of Fields
where
The quantities g i j may be written in the matrix form
This is known as the metric of the space with respect to the curvilin-
ear coc)rdinates ul, u2 , u3.
I f the coordinate system is orthogonal,
then (3.44) reduces to
where hi = 1 dR / dui 1, i = 1, 2, 3.
In this case formula (3.43) simplifies
to the form
ds
2
= (h, du, )2+ ( h 2 d ~ 2 ) 2 + ( h 3 d ~ 3 ) 2 (3.46)
This reduces to the familiar form ds2 = dx
2
+ d
y
2
+ dz
2
in rectangu-
lar cartesian coordinates.
Example 2. The transformation relating the cylindrical coordi-
nates r, H, z to the rectangular cartesian coordinates x, y, z is defined
by the equations
r = 4 x 2 + y2
192 Chapter 3
The transformation is defined for a11 (x, y, z) except for points on the
z-axis where 8 is not defined. By the same calculation as in Example
1, we find
The inverse transformation is given by
x = r c o s 8 , y = r s i n 8 , z = z
which is valid for all r r 0 , fl , z.
The coordinate surfaces intersecting at a point (x,, y, z,) con-
sist of a cylinder of radius r,, = \1(x02 + y(?) with axis along the z-axis, a
plane containing the z-axis and making an angle Oo = cos -' (x, /rO) =
sin
( yo /rO) with the xz coordinate plane, and the plane z = z,. The
coordinate curves comprise of a circle and two straight lines as
shown in Fig. 3.10.
The position vector in cylindrical coordinates assumes the
form
so that the tangent vector to the coordinate curves are
aR
= cos 8 i t sin 8 j ,
dR dR
= - r sin 0 i t r cos 0 j , - - - = k
d r d o dz
these vectors are mutually orthogonal for all r 2 0 and 8. Therefore,
the cylindrical coordinate system is an orthogonal curvilinear coor-
dinate system.
Letting ul = r, u2= 8, and u3 = z, it follows that h, = 1, h2 = r,
h,= I so that from (3.46) we have
Differential Calculus of Fields
Fig . 3.10 Cylindrical coordinate system.
This is the expression for arc length in cylindrical coordinate system.
Obviously, when we drop the third variable z from the above
discussion, we obtain the polar coordinate system. It follows that the
polar coordinate system is a two dimensional curvilinear coordinate
system.
Example 3. The spherical coordinates ul = r, u2 = 9, u3 = 8 are
defined by the equations
X
0 = arccos = arcsin
J' -
j x2+ y2) l n j X
2
+ y2j ' I 2
Chapter 3
(r r 0, 0 s 9 < n , 0 5 8 < 2x ). Notice that the coordinate 8 is defined
in the same way as in the polar coordinate system. By a straight-
forward differentiation and calculation of the determinant, we find
2
r \ x2+ y2
r sin $
The student is asked to verify this result in Prob. 5. Hence, except for
points on the z-axis, the transformation has an inverse given by
x = r sin $ cos 8, y = r sin $ sin 8 , z = r cos $
valid for r r 0, 0 5 $ 5 x, 0 5 8 < 2x. From (3.39) or by direct calcula-
tion, we find
The coordinate surfaces intersecting at a point (x,, y,, z,,) con-
sist of a sphere of radius ro = \1(x,; + y(; + z,: ) with center at the
origin, a "cone" with vertex at the origin, axis along the z-axis, and
generating angle $,, = arccos (z,, /r,), and the plane containing the
z-axis making an angle
H (, = arccos [x,, /(x,2 + y,,2)1/2] = arcsin [ yO/(x,,2 + ~ ( , 2 ) ~ / ~ ]
Differential Calculus of Fields
Fig. 3.11 Coordinates surfaces i n spherical coordinates.
with the xz-coordinate plane (see Fig. 3.11). In this connection we
must broaden the interpretation of the word "cone" to include the
xy-plane for which 9 = n/ 2 and the cones with generating angles cp
greater that n/ 2.
The position vector in spherical coordinates is given by
R (r, 9, 8 ) = r sin $ cos 8 i -t r sin @ sin 8 j + r cos $ k
Thus the tangent vectors to the coordinate curves are
aR
- = sin q c o s 8 i + s i n @s i n @j + cos $k
dr
dR
= r cos qcos 8 i + r cosqsi n 8 j - r sin $ k
d9
dR
= - r s i n @si n 8 i + r s i n q c o s ~ j
ae
196 Chapter 3
It is readily verified that these vectors are mutually orthogonal so
that the spherical coordinate system is an orthogonal curvilinear
coordinate system.
2, g ?? = r sin 9 and hence By (3.44), we find g l l = 1, g = r
. .
the formula (3.43) for arc length in spherical coordinates is given by
3.7 EXERCISES
1. By observing that the determinant of a matrix is the same as
the determinant of its transpose, show that the Jacobian d(x, y,
z)/d(ul, u2, US is equal to the scalar triple product of dR/dul,
dR/du2, dR/du,, that is,
2. Show that
3. Recall that the product of two n x n matrices A = (aij) and
B = (b..) is defined as the matrix AB = (c..), where
'J '1
Thus show that
Differential Calculus of Fields
[Hint: 6.. = dui /du, = (du, /dx)(ax/au,) + ( a~, / ay) ( dy/ d~. ) +
'1 1
( dui / az) ( az/ a~, ) 1
4. Deduce the relation (3.39) from the result of Problem 3.
5. Verify the calculation of the Jacobian d(r, @,8)/d(x, y, z) in
Example 3.
6.
The transformation relating the cartesian coordinates x, y, z to
the parabolic cylindrical coordinates u, v, z is given by the
equations
(a) Determine the curves on the xy-plane (z = 0) which corres-
pond to u = const.; and v = const. In particular, what are the
curves which correspond to u = 0; v = 0; u = v; u = - v?
(b) Find the tangent vectors dR /du, dR /dv, dR /dz, and show
that the new coordinate system is orthogonal.
(c) What is the expression of the arc length in the new coordi-
nate system?
7. The transformation relating the cartesian coordinates x, y, z to
the elliptic cylindrical coordinates u, v, z is given by
x = a cosh u cos v, y = a sinh u sin v, z = z
(u 2 0, 0 I; v < 2n, a > 0 const.)
(a) Show that in the xy-plane a curve u = const. represents an
ellipse, while a curve v = const. represents half of one branch
of an hyperbola.
(b)
Sketch the curves on the xy-plane conrresponding to the
values u = 0; v = 0 and v = n ; v = n / 2 and v = 3n/2.
(c)
Verify that the new coordinate system is orthogonal.
(d)
Show that the arc length in the new coordinates is given
by
ds
2
= a
2
(cosh
2
u - cos2 v)(du2 + dv2) + dz2
8.
The transformation relating the cartesian coordinates x, y, z to
the parabolic coordinates u, v, 8 is gven by
( u 2 0 , v 2 0 ). Verify that the parabolic coordinate system is
Chapter 3
orthogonal and find the expression for the arc length in that
coordinate system.
9. Let u, v, w denote the new coordinates defined by the equa-
tions
x = u v , y =( u 2 +v 2 ) / 2 , z = w
Find the Jacobian of u, v, w and show that the new coordinate
system is not orthogonal.
10. Consider the new coordinates u, v, w defined by
(a) Find the inverse transformation.
(b)
Show that the coordinate curves are straight lines.
(c)
Show that the coordinate system (u, v, w) is not orthogo-
nal. We call this an oblique cartesian coordinate system.
(d)
Show that the u, v, w - coordinate axes are left handed.
(e)
Find the expression of the arc length in the coordinates
U, v, w.
3.9 GRADIENT, DIVERGENCE AND CURL IN
ORTHOGONAL CURVILINEAR COORDINATE
SYSTEMS
In this section we derive the expressions of the various vector
concepts in an orthogonal curvilinear coordinate system. Let ul, u2,
u3 denote such a coordinate system defined by the transformation
Ui =qi ( xf y, z) ( i =l , 2, 3) (3.47)
and let
denote the inverse transformation. As in the preceding section, we
assume the same differentiability conditions on both the functions
(3.47) and (3.48). By relabeling the coordinates, i f necessary, we may
assume that the Jacobian d(x, y, z)/d(ul, u2,u,) is positive so that the
base vectors dR/dul, dR /du2, dR ldu,, in that order, form a right-
Differential Calculus of Fields 199
handed triple. To obtain an orthonormal basis, we normalize the
base vectors by multiplying each vector by the reciprocal of its mag-
nitude. We denote these unit base vectors by
where h, = I dR/dui I . It follows that u = u x u k, where if j, k are
.I
cyclic permutations of 1, 2, 3; that is,
Gradient of a scalar field
Now suppose f is a differentiable scalar field given in a rectan-
gular cartesian coordinate system x, y, z. Under the transformation
of coordinates (3.48), f becomes a function of u,, u2, u, The total dif-
ferential o f f is given by
Let
be the expression of grad f in the new coordinate system. We shall
determine the components f,, f2, f, Since df = Vf- dR and by (3.49),
= ul hl du, + u2h2du2 + u,h3du,
we find
df = Vf dR = f, h, du, + f2h2 du, + f,h3 du,
Comparing this with (3.51 ), we deduce
Chapter 3
Substituting these in (3.52), we finally have
This is the representation of the gradient of a scalar field in an or-
thogonal curvilinear coordinate system (ul, u2, u3).
From (3.53) we deduce the expression
for the del operator V. Thus we see that the b~adi ent of f may again
be interpreted as the result of applying the operator (3.54) to the
function f.
Example 1. Find the expression of V and deterniine the
gradient of f = xyz in cylindrical coordinates.
Solution: In cylindrical coordinates the function becomes
f = I-% sin 0 cos 8. Let ul = r, u2 = 0, and u, = z. Then, by (3.54),
we have
where (see Example 2, Sec. 3.8)
u l = c 0s 0i + sin Hj , u2 = - s i n0 i + cosH j , u 3 = k
Applying V to the given scalar field f, we find
Vf = 2rz sin 0 cos Oul + rz(cos
2
H - sin
2
0 )u2
+ r sin 0 cos 0u3
Differential Calculus of Fields 201
It is instructive to note that i f we substitute the expressions of
u,, u2, u, in the above formula for Vf and collect the terms invol-
ving i , j , k , we find
Vf = r z s i n e i + r z c o s e j +r 2 s i n 0 c o s Hk
= yzi + xzj + xyk
which is precisely the gradient of f in rectangular coordinates. This
shows the invariant property of Vf under any coordinate transfor-
mation.
Divergence of a Vector Field
Next we derive the expression of the divergence of a vector
field. First, we observe that by applying the del operator (3.53) to the
functions defined in (3.47) and noting that dui/du. = 6 . where 6.. is
1 1 ~ ' '1
the Kronecker delta, we obtain
Since the curl of the gradient of a scalar field is zero, it follows that
Moreover, by (3.32b), (3.50) and (3.53), we have
where i, j, k assume the cyclic permutations 123, 231, 312.
Now let
Chapter 3
be a vector in the orthogonal curvilinear coordinate system ui with
basis ui, i = 1, 2, 3. By definition, we have
div F = V- F = V' (Fl u, ) + V' (F2 u2) + V' (F3 u d
(3.57)
To calculate each of the terms on the right-hand side of (3.57), we
write
for ijk = 123, 231, 312. By the property of divergence, we find
But from (3.52) we have
~1 a ~2 a
v ( Fi hj h,) =
u3 d
h l dul ( F i h i h k ) f h 2 a u , ( F i h j h k ) + h3"3 ( Fi h j h k )
Since the vectors ul, u2 , u 3 are orthonormal, (3.58) yields
for ijk = 123, 231, 312 . Substituting this in (3.57) for i = 1, 2, 3, we
finally obtain the formula
Differential Calculus of Fields 203
This expresses the divergence of a vector field in an orthogonal
curvilinear coordinate system.
In particular, if F = Vf, where Vf is given by (3.52), then we
have
v2f = div grad f (3.60)
This is the expression of the Laplacian of a scalar field in an ortho-
gonal curvilinear coordinate system.
Example 2. Find the divergence in cylindrical coordinates of
the vector field F = r 3ul + r
2
sin H u2 + z
2
u,
Solution: In cylindrical coordinates we have hl = 1, h2 = r, and
h3 = 1. Hence by (3.59) we have
If we change back to cartesian coordinates, this result becomes
which is precisely the divergence of F when F is expressed in carte-
sian form.
Chapter 3
Curl of a Vector Field
Finally, we derive the expression for the curl of a vector field F
which we assume in the form (3.56). By defintion, we have
To calculate each of the terms on the right-hand side of (3.61), we
write
i
F i u i = ( F i h i ) - ( i =1, 2, 3)
hi
Then by the property of the curl and by (3.54), we find
For i = 1, 2, 3 , we know that
and from (3.50) we have ul = u2 x uy u2 = u3 x u p u, = ul x u2 Hence,
by taking the cross product in (3.62) for i = 1, 2, 3, we find
Differential Calculus of Fields
and
Substituting these results on the right-hand side of (3.61), we finally
obtain
curl F =
3
This can be written in the more convenient form
1
curl F =
h l h 2 h 3
which resembles the corresponding form in rectangular cartesian
coordinate system.
Chapter 3
EXERCISES
Find the expressions of v2f and V x F in cylindrical coordinate
system.
Find the expressions of v2f, V* F, and V x F in spherical coor-
dinates.
Find Vf and v2f in cylindrical and spherical coordinates when
f(x, y, z) = xy + yz + zx; @) f(x, y, z) = x
2
+ y2+ z2.
Consider the unit base vectors u, = cos 0 i + sin 0 j ,
u2 = - sin 0 i + cos 0 j , u3 = k in cylindrical coordinates.
Treating the equations as algebraic system, solve for the unit
vectors i , j , k in terms of the unit vectors u,, u2, U, Thus
express the vector field F = zi - 2xj + 2yk in cylindrical coor-
dinates .
Find the divergence and the curl in cylindrical coordinates of
the vector field given in Problem 4.
Repeat Problem 5 when F = yi + z2j + (x
2
+ y2)k.
(a)
From Example 3, Sec. 3.9, determine the unit base vectors
u, = ul, U$ = ua uo = u3 in the spherical coordinate system.
@) Determine i, j, k in terms of u , ug uu.
Express the vector field F = zi + 2yj + 3zk in spherical coor-
dinate system, and find the divergence and curl in that coordi-
nate system.
Repeat Problem 8 when F = zi + 2yj + 3zk.
Find the expressions of Vf, V' F , and V x F in the parabolic
cylindrical coordinate system given in Problem 6, Exercise 3.7.
Find the expressions of Vf, V' F, and V x F in the parabolic
cordinate system given in Problem 8, Exercise 3.7.
Using the formulas (3.53), (3.59), and (3.63), show that
(a) div curl F = 0 and (b) curl grad f = 0 in any orthogonal
curvilinear coordinate system.
INTEGRAL CALCULUS OF SCALAR
AND VECTOR FIELDS
In this chapter we shall study line and surface integrals of
scalar and vector fields. These integrals are the natural generaliza-
tion of ordinary single and multiple integrals. In the ordinary
single integral
the function f is defined in the interval [a, b] of the x-axis, while in
the double integral
f is defined in the domian D of the xy-plane. In contrast, in a line
integral (the term curvilinear integral would seem more appro-
priate), the function is defined on a space curve and the integration
is performed with respect to arc length of the curve. In a surface
integral the function is defined on a surface and the integration is
performed with respect to the area of the surface. In actual compu-
tation of these integrals, however, they are generally converted to an
ordinary single or multiple integrals where the integration is per-
formed in the usual manner.
For this reason, the student is presumed to be already familiar
with the usual definition of ordinary single and multiple integrals as
well as with the techniques for evaluating such integrals.
Chapter 4
4.1 LINE INTEGRALS OF SCALAR FIELDS
Let f be a scalar field defined in a domain D of the xyz-space
and let R(t) = x(t)i + y(t)j + z(t)k, a 5 t < b, represent a space curve C
that lies in D (Fig. 4.1). We assume that C is a smooth curve and that
f is continuous on C. From Sec. 2.5 we know that the arc length of C
is given by
s(t) = [ 'In1(?) I di ( a < t < b )
so that ds = I R'(t) I dt. Thus we define the line integral of f on C with
respect to arc length as follows:
Definition 1: The line integral of f on C with respect to arc
length, denoted by jc f ds, is the integral
where
2
I R' (t) I = -)c(t,
Fig. 4.1 Integral of f on a curve C.
Integral Calculus of Fields 209
Notice that when f = 1, the integral (4.1) gives the length of the
curve C. Also, the value of the integral (4.1) depends only on the
function and the geometry of the curve C. It does not depend on the
parametrization of the curve C. This means that if R = R(t), a 5 t i b,
and R* = R*(r), a < z 5 p, are two different representations of the
curve C such that they orient the curve in the same direction, that is,
R(a) = R*(a) and R(b) = R*(P), then
Example 1. Evaluate the line integral Ic fds of f(x,y) = x + 2y
on the straight line y = 2x from the origin to the point (1,2).
Solution: We represent the straight line by the equation
R(t) = ti + 2tj, 0 5 t 5 1. Then I Rt(t) I = 45 so that
Example 2. Evaluate the line integral IC xyz ds, where C is the
helix R(t) = cos t i + sin t j + tk, 0 < t 12n:.
Solution: First we note that R'(t) = - sin t i + cos t j + k so that
I R'(t) I = 42. Thus
xyz ds = \ 2 t cos t sin t dt
6'"
2" 271
2
= \ 21 i si n2t l o - \22 1, sin t dt
Chapter 4
Applications of Line Integrals
Line integrals occur in many problems in mathematical phy-
sics. For example, if f represents a force that moves a body along the
curve C and that f remains tangential to the curve at each point, then
the line integral Ic fds represents the work done by the force in
moving the body along the curve. We shall consider this point in
more detail in the next section. In mechanics the calculation of the
mass of a body, the center of gravity, and the moment of inertias of
geometric figures lying on curves, all make use of line integrals. For
example, if a piece of wire is in the shape of a space curve C and its
density (mass per unit length) is given by p(x, y, z), then the total
mass of the wire is given by the line integral
The center of mass of the wire refers to the point whose coordinates
(xc, yc, zc) are given by the equations
all involving line integrals. The moment of inertia of the wire with
respect to a line L is given by the line integral
where 6 denotes the perpendicular distance of each point (x, y, z) on
the wire from the line L. Thus, for example, if L is the z-axis, then
6 2 = x 2 + y 2 sothat
2 2
On the other hand, if f(x, y, z) denotes the temperature of the
wire at the point (x, y, z), then the integral
Integral Calculus of Fields
where L denotes the length of the wire, represents the average tem-
perature of the wire. We consider a few examples below.
Example 3. A piece of wire is in the shape of a parabola
y = x
2
, - 2 < x < 2, and its density is given by p(x,y) = I x I . Find the
mass of the wire.
Solution: Using x as a parameter, we re present the parabola by
the vector equation R(x) = xi + x2j, -2 < x 5 2. Then
and so the mass is given by
Example 4. A iece of wire is in the shape of a semicircle of
4'
radius a, y = 'i (a
2
- x ), - a 5 x 5 a, and its density is given by p (x, y)
= y. Find the center of mass of the wire, and its moment of inertia
with respect to the x-axis; the y axis.
Solution: Let us represent the semicircle by the equation
R(8) = a(cos 8 i + sin 8 j),
0 5 8 5 x
Then, ds = a de and so the mass of the wire is given by
Chapter 4
2 2
M = i r ( a sin 0) a do = a - cos 0) I: =,a
Let (xc , yc) denote the center of mass of the wire. By symmetry
we deduce that xc = 0, (this can of course be verified by actual calcu-
lation). The ordinate y, satisfies the equation
MI, = in (a sin o)(a sin0)a do = a 'in sin do
Hence, yc = (n: a3/M) = I[ a/4 and so the center of mass is at the
point (0, n: a/4). The moment of inertia of the wire with respect to
the x-axis is given by the integral
2
(a sin 0) (a sin 0) a d0
3 2
= a4[nsin o dO = a sin o (I - cos 8) do
The moment of inertia with respect to the y-axis is given by
2
(a cos 0) (a sin 0) a d0
3
4 C OS O 2 4
[- 3 ]:=P
Integral Calculus of Fields
Line Integrals with Respect to x, y, z
In a similar manner we define the line integral of f on a space
curve C: R(t) = x(t)i + y(t)j + z(t)k, a < t I b, with respect to the
variables x, y, and z as the ordinary integrals
Notice that when f(x, y, z) = 1, these integrals give x(b) - x(a), y(b) -
y(a), and z(b) - z(a), which are the projections of the space curve on
the x-, y-, and z-coordinate axes, respectively. As with integrals with
respect to arc length, the integrals (4.2) do not depend on the para-
metrization of the curve C.
Example 5. Evaluate the line integrals of f(x, y, z) = xyz with
respect to x, y, and z along the curve R(t) = cos t i + sin t j + tk , 0 < t
< I t .
Solution: We have
jc xyz dx = I, t cos t sin t (- sin t) dt
n
3 310=;
= I[-t sin t - cos t + cos t
3 3
kc xyz dy = I, t cos t sin t (cos t) dt
3 1 3 n:
=I[-tcos 3 t + s i n t - 3 sin t]:=3
214
and
Chapter 4
2 1 sin 2t =
I;yz dr = Ln t cos t sin t dt =
Example 6. Calculate the line integrals 1, (x
2
+ y)dx and
Ic (x2 + y)dy from (0,O) to (4,2) along the parabola y = 'ix (0 5 x 5 4).
Solution: Using x as a parameter, we find
and
If we use y as a parameter, we find
and
This verifies the fact that the line integral along the parabola does
not dependent on the manner we parametrize the curve.
Geometric Interpretation of Line Integrals in the Plane
The student ma recall from elementary calculus that when
X
f(x) t O the integral 1, f(x)dx represents the area bounded by the
curve defined by y = f(x) and the lines x = a, x = b. It is interesting to
note that an analogous geometric interpretation also holds for line
integrals along curves in the xy-plane. Indeed, suppose f(x, y) is a
non-negative function defined on a curve C in the xy-plane. Then
Integral Calculus of Fields 215
we may regard z = f(x, y) as the height of a vertical cylindrical sur-
face at each point (x, y) on the curve C (see Fig. 4.2); that is, z = f ( ~ , y)
defines a "fence" whose base is the curve C. Then, the line integral
Ic f(x, y)ds represents the area of the "fence" along C whose height
Fig. 4.2 Geometric interpretation of a line integral.
at each point (x, y) is iven by z = f(x, y). In this context, the inte-
grals Jc- f(x, y)dx andTc f(x, y)dy represent the area of the projection
of the "fence" on the xz- and yz-coordinate planes, respectively, pro-
vided there is no overlapping in the projection. As a simple illus-
tration, we consider the following example.
Example 7. Find the area of the fence whose base is the line
y = 2x, 0 I x 5 1 (ft), and whose height is defined by z = xy (ft), see
Fig. 4.3. Verify the area of the projection of the fence on the coordi-
nate planes by using the "area cosine principle".
Solution: Using x as a parameter, we have ds = d(l + y' 2) dx
= 45 dx, so that the area is
Chapter 4
Fig. 4.3 Area of projections of a surface on xz- and yz-
coordinate planes.
On the other hand, the area of the projection on the xz-plane is
and the area of the projection on the yz-plane is
Now by the area cosine p rinciple, the area of the projection on
the xz-coordinate plane is given by the product of the area and the
cosine of the angle a which the line makes with the x-axis. Like-
wise, the area of the projection on the yz-coordinate plane is the pro-
duct of the area and the cosine of the angle p which the line makes
with the y-axis. From Fig. 4.3 we see that cos a = 1/45 and cos
=
2/45. Thus we have
Integral Calculus of Fields
which agree with the results given by the integrals.
Example 8. Find the surface area of a wall that is erected on
the curve x
2
+ 4y2 = 1, x > 0 , y 2 0, whose height at each point (x, y)
is given by f(x, y) = xy. Assume the measurement to be in meters.
Solution: Let us parametrize the curve by setting x = 2 cos 0,
y = sin 0 (0 0 5 n/2). Then we have
R(O ) = 2 cos 0 i + sin 0 j , R'(0) = - 2 sin 0 i + cos 0 j
so that
Hence the surface area of the wall is equal to
7112 2
{c~Yds =
2 cos0 sin 0 \ 3 sin 0 + i d0
The area of the projection of the wall on the xz-plane is given
by JC xy dx. To ensure a positive area, we need to integrate in the
direction of increasing values of x. Thus we find
Icxy dx = JnD
2 cos 0 sin 0 d(2 cos 0)
0
-
3 O 4
- 1,
-4 cos0 sin$ d0 = - 4 sin 0 I n n =3
3
The area of the projection on the yz-plane is given by
Chapter 4
4.1 EXERCISES
In each of Problems 1 through 5, calculate the line integral
along the given curve.
2
1.
( (x - yz) ds, C : R(t) = a cos t i + a sin t j (0 < t < 2 ~ )
. C
traced counterclockwise.
2.
l c ( X - y) ds, C : quadrilateral with vertices at (1, O), (0, I),
(-1, O), (0, -1) traced counterclockwise.
3. l c (Z + 1) ds, C: R(t) = t cos t i + t sin t j + t k (0 < t 5 K)
from (0, 0, 0) to (- K, 0, n) .
r
4. 1 (xy - z) ds, C : R(t) = a cos t i + a sin t j + bt k (0 < t < 3rr/4)
. C
from (a, 0,O) to (- Y'2a/2, d2a/2,3n b/4).
from (2,0,O) to (0,2,2)
6. A piece of wire is in the shape of an equilateral triangle with
vertices at (1,0, O), (0,1, O), (0,0,1). If the density is given by
f(x, y, z) = x + y + z, find the mass and the center of mass of the
wire.
7. The density of a ring of wire with radius a is defined by p =
a/ n + I x I . Find (a) the mass of the wire, (b) the center of mass,
(c) the moment of inertia with respect to the x-axis, (d) the
moment of inertia with respect to the y-axis.
8. Find the average temperature of a piece of wire that is in the
shape of a curve defined by x = t , y = 45 t
2
/2, z = 2t (0 < t 5 1)
if the temperature at each point (x, y, z) is given by f(x, y, z) =
YZ.
Integral Calculus of Fields
9. Repeat Problem 8 for a piece of wire in the shape of a curve
defined by x = t cos t, y = t sin t, z = 2t (0 5 t I n) whose tempe-
rature is given by f(x, y, z) = z.
In each of Problems 10 through 15, calculate the line integral
along the given curve.
2
10.
(x - y$ dx along the parabola y = x - x /i from (0.0) to (4.0).
11. 1 (x - y\ dy along the curve of Problem 10.
. C'
12.
i >x2- $) dx around a drclc of radius a traccdcoonterdockwisr.
L L
2 2 X
13. />x + y ) dy around the ellipse + y2 = 1 traced
a b
counterclockwise.
2
14. / (xy + z ) dz along the helix x = 2 cos t, y = 2 sin t, z = 2t
. (1
from (2, 0,O) to (- 42, 42, 3n/2).
15. I&xz + yz + xy ) dy around the intersection of the cylinder
x
2
+ y2 = 1 and the plane y + z = 1 in the clockwise direction
as viewed from the origin.
16. Let C1 and C2 be two simple smooth curves represented by
the equations
where f(a) = g(a), f(b) = g(b) , and f(x) < g(x) for a r x < b . Let
C denote the closed curve composed of C, and C2 as shown
in Fig. 4.4. Show that the integral jC y dx gives the area of
the domain bounded by C.
Chapter 4
0 x = a x = b
Fig. 4.4 Area of a domain bounded by a closed curve.
4.2 LINE INTEGRALS OF VECTOR FIELDS
We now extend the definition of a line integral to vector fields
To motivate the definition, we consider the notion of work in phy-
sics. Suppose F represents a force field in space and C represents
the smooth path of a particle moving under the influence of F. We
want to calculate the amount of work done by the force in moving
the particle along C. If F is a constant force and C is a line segment
represented by the vector D, then, by definition, the work done by F
in moving the particle through the displacement D is given by
Work = F* D
More generally, if C is a space curve defined by R(t) = x(t)i + y(t)j +
z(t)k (a I t I b), we calculate the work done by F by decomposing C
into a finite number of paths Ci and determining the work done by F
on each path, and then taking the sum. So let us divide the interval
[a, b] into n subintervals with endpoints
Integral Calculus of Fields
Fig. 4.4 Work done by a force field F along a curve C.
Let ARi = R(ti) - R(timl) and denote the length of the arc between
R(ti-l) and R(ti) by Asi (Fig. 4.5). When n is large, the arc length As is
small and may be approximated by I ARi. Further, the force F is
approximately constant on Ci, say it is equal to Fi = F [x(ki), y( ti),
z( ti)], for some ki, ti-1 5 ti 5 ti (1 5 i 2 n). Then the work done by Fi
in moving the particle through Asi is approximately equal to
Hence the work done by F in moving the particle along the curve C
is approximately equal to the sum
Chapter 4
Now by the mean value theorem of differential calculus, we can
write
ARi = R(ti) - R(ti-,) = R'(qi) Ati
where ti-, 5 qi 5 ti and Ati = ti - ti-l. Thus we have
n n
Work = x Fi. ARi= 2 Fi . Rf(qi )At i
i = l i =l
This approximation improves as the partition of [a, b] is made finer
by making n larger. In fact, the exact amount of work done is
obtained when we take the limit of the sum as 1.2 tends to infinity.
From our study of calculus, we know that this limit is precisely-the
integral of the scalar field F[x(t), y(t), z(t)] Rf(t) over the interval [a,
b], that is,
n
W ork = lim F i. R' (q i)At =
F[x(t), y(t), z(t)]. R' (t) dt
~ - - > o o i = l
We call this integral the line integral of F on the curve R(t) = x(t)i t
y(t)j + z(t)k (a 5 t 5 b). Thus, we define the line integral of a vector
field on a curve as follows:
Definition 1. Let F(x, y, z) = P(x, y, z)i + Q(x, y, z)j + R(x, y, z)k
be a vector field defined and continuous in a domain D, and let C
be a smooth curve in D represented by R(t) = x(t)i + y(t)j + z(t)k, a < t
< b. The line integral of F on C, denoted by IC F-dR, is the integral
Notice that the integral (4.3) is taken along the positive direc-
tion on the curve C. If the curve is traversed in the opposite
direction, the integral changes sign. Further, as in the case of a line
integral of a scalar field, the integral (4.3) does not depend on the
parametrization of the curve C. That is,
Ic F dR = [ F [ ~( t ) , y(t), z(t )] R (t) dt
Integral Calculus of Fields
F[x(z), y(t), z(z)]. R*' (z)dt =
F. dR*
k
where R = R(t), a 5 t < b, and R* = R*(t), a < t 5 p, are two different
representations of the curve maintaining the same orientation.
If F(x, y, z) = P(x, y, z)i + Q(x, y, z)j + R(x, y, z)k, the integral
(4.3) can also be written as
Notice that the functions P, Q, and R are integrated along the curve
Cwith respect to the variables x, y, and z, respectively, in accord-
ance with the definition given in the preceding section.
Example 1. Calculate the line integral of F(x, y) = x
2
yi +
(x
2
+ y)j from the origin to the point (2,4) along the parabola y = x
2
.
Solution: We represent the parabola by R(t) = ti + t2j, 0 5 t r 2.
Then on the curve we have
so that
3
lC F d R =l(;(t4+ 4t ) dt = 112 5
Alternatively, if we write
then, using x as the parameter, we find
Chapter 4
Thus if F represents a force field that moves a particle from the ori-
gin to the point (2,4) along the parabola, the value of the integral (in
appropriate units, e. g., ft-lb) represents the amount of work done by
the force on the particle.
Example 2. Calculate the line integral of F(x, y, z) = yi - xj + zk
along the helix C: R(t) = cos t i + sin t j + tk , 0 I t 5 n.
Solution: We note that R'(t) = - sin t i + cos t j + k so that on
the helix we have
F[cos t, sin t, t] R'(t) = - sin t - cos t + t = - 1 + t
Hence,
Other Notations
We recall that the unit tangent vector on a curve defined by
R(t), a I t I b, is given by
Thus the line intgeral (4.3) can also be written as
b b
j C ~ . d n = [ F I R' (t ) R~ (t) I i ~ y t ) d t = [ ~ + t ~ ~ y t ) d t
Integral Calculus of Fields
This says that the integral j C ~ dR is simply the integral of the tan-
gential component of the vector field F along the curve C. If F
represents a force field, then indeed only the tangential component
of the force along the curve contributes to the work done by F.
When C is a closed curve, it is customary to write the integral
in the form
The integration around the closed curve C is taken in the positive
direction which is, by convention, counterclockwise.
Example 3. Evaluate the line integral of F(x, y) = x
2
i + j
around the circle x
2
+ y2 = 4.
Solution: We represent the circle in polar coordinates
x = 2 cos 8, y = 2 sin 8, 0 1 8 < 2n
Then we have
Example 4. Evaluate the integral of F (x, y, z) = (x + y2)i +
(x + z)j + xyk from the origin to the point (1, -1, 1) along (a the
(0 5 t 1 1).
1
straight line joining the points; (b) the curve R(t) = t i - t
2
j + t k
Solution: (a) Let us represent the line segment that joins the
origin and the point (1, -1,l) by the parametric equations
226
Then
Chapter 4
(b) Along the curve C: R(t) = t i - t
2
j + t3 k, 0 2 t 5 1, we have
This example shows that the value of the line integral from (0,
0, 0) to (1, -1, 1) depends on the curve that joins the points. This is
true in general. Later we shall study vector fields whose line inte-
grals from one point to another are independent of the curve
joining the points.
Example 5. A particle of mass m is moving along a curve R
= R(t) under the action of a force F = F(t) (t > 0). If v(t) denotes the
velocity of the particle at time t, show that
where v(t) = I v(t) I .
Solution: According to Newton's second law of motion, we
have
Integral Calculus of Fields
Then
hence
which yields the desired result.
The quantity (1/2)mv
2
(t) is known as the kinetic energy of the
particle at time t. Thus the formula says that the work done by F du-
ring the time interval [t, , t2] is equal to t he change i n kinetic
energy of the particle. This is a basic law in mechanics.
4.3 PROPERTIES OF LINE INTEGRALS
We have defined line integrals of scalar and vector fields in
terms of ordinary integrals of a function of a single variable. As a
result of this line integrals satisfy many algebraic properties which
are similar to those satisfied by ordinary integrals. For example,
suppose f and g are scalar fields and F and G are vector fields, all
defined and continuous in a domain containing the smooth curve
C: R = R(t) (a 2 t 2 b). Then for any constants c, and c2, we have
and
This property is known as the linearity property of line integrals.
Next, suppose the curve C is comprised of two connected
curves C, : R = R,(t) (a 2 t < b) and C2 : R = R2 (t) (c 5 t < d). Then
Chapter 4
and
This is known as the additive property of line integrals. It corres-
ponds to the familiar formula
of ordinary integrals.
More generally, if C is composed of 12 smooth curves Cl, C2, . ,
C,, as shown in Fig. 4.6 for n = 5, then
and
Fig. 4.6 Line integral along a piecewise smooth curve.
Integral Calculus of Fields 229
We observe that each of the integrals on the right-hand side of (4.9)
and (4.10) may be evaluated using different parametrization so long
as each parametrization of a curve preserves the original orientation
of the curve.
Example 1. Evaluate the integral jC xdx - zdy + 2ydz, where
C consists of the line segments from the origin to the point (1,1, 0)
and from (1,1,O) to the point (1,1,2) as shown in Fig. 4.7.
Solution: Referring to Fig. 4.7, we note that on C1, x = y, z = 0,
so that dx = dy, dz = 0, and on C2, x = 1, y = 1 so that dx = 0, dy = 0.
Hence we have
Fig. 4.7 Integrals along line segements.
Example 2. Evaluate the line integral
2 2
I =/ c( x2+y) dx+( y +z ) d y +( z +x) dz
230 Chapter 4
around the closed curve C consisting of the line segments C1: x + z =
1(O I x i 1, y = 0) and C2: x + y = 1 ( O i y i I, z = 0), and the quarter
circle C?: y2 + z
2
= 1 (y 2 0, z 2 0) in the direction indicated in Fig.
4.8.
Solution: We calculate the line integral on each of the line seg-
ments and on the quarter circle. On C, we have y = 0, z =1 - x so that
dy = O and dz = dx. Hence, using x as the parameter, we find
On C2, using y as the parameter, we have x = 1 - y, z = 0, so that dx =
- dy, dz = 0. Hence
Finally, on C3 where x = 0, we set y = cos 0, z = sin 0 (0 1 0 17t / 2).
Then
Integral Calculus of Fields
Fig. 4.8 Integration around a closed curve.
Thus the line integral around the closed curve is equal to the sum
Example 3. Evaluate the line integral jc ydx - xdy + zdz
-
around the curve of intersection of the cylinder x
2
+ y2 = a
2
and the
plane z - y = a taken in the counterclockwise direction as shown in
Fig. 4.9.
Solution: First, we need to find a parametrization of the curve
of intersection of the cylinder and the plane. We note that every
point on the cylinder x
2
+ y2 = a
2
may be represented in cylindrical
coordinates by the equations
Since the curve of intersection consists of points lying on the cylin-
der whose third coordinate z is given by z = a + y, we have
Chapter 4
Fig. 4.9 Integration along an ellipse.
a parametric representation of the curve of intersection. Observe
that the curve is an ellipse. Thus we have
~ ; y d x - x d y + z d z =
[a sin 8 (- a sin 0) - a cos 8(a cos 8) + a(l + sin 8)(a cos 8)]d8
Integral Calculus of Fields
4.2 EXERCISES
In each of Problems 1 through 8, compute the line integral of
the vector field F along the curve described.
F(x, y) = (x
2
- y)i + ( y2 - x)j from the origin to the point (1, 2)
along the parabola y = 42 x.
F(x, y) = (x2 + y2)i - xyj from the point (1, 1) to (-2, 4) along
the parabola y = x
2
.
F(x, y) = 2xyi + (x
2
- y2)j from (2, 0) to (0, 2) along the first
quadrant arc of x
2
+ y2 = 4.
F(x, y) = (y2 - 2xy)i + ( x
2
+ 2xy)j from (-1, 0) to (1,O) along the
upper semi-circle of radius 1.
F(x, y, z) = y(x - z)i + z(y - x)j + x(z - y)k from (1, 1, 1) to (2, -2,
3) along the line segment joining the points.
F(x, y, z) = (x - z)i + (y - x)j + (z - y)k from the origin to (1, 1 , l )
along the curve R(t) = ti + t
2
j + t
3
k (0 < t 5 1).
F(x, y, z) = (y2 + z)i + (z
2
+ x)j + (x
2
+ y)k from (l,O, 0) to (O,1,
0) to (0, 0, l ) along the line segments joining the points.
F(x, y, z) = (y - 1)i + ( x + 1)j + 2zk from (0, a, 0) to (a, 0, IT)
along the helix R(t) = a(sin t)i + a(cos t)j + 2tk (0 _< t 5 7c/2).
Evaluate the line integral
in each of the two cases:
(a) when C is the sides of the square I x I = 1, I y I = 1 oriented
counterclockwise.
(b) when C is the unit circle x
2
+ y2 = 1 oriented counterclock-
wise.
Evaluate the integral
I -
(x + y) dx - (x - y) dy
2 2
+ Y
around the circle x
2
+ y2 = a
2
in the counterclockwise direc-
tion.
11. Evaluate the integral jc ydx - xdy + zdz around the intersec-
-
tion of the cylinder x
2
+ y2 = 1 and the plane y + z = 1 in the
clockwise direction as viewed from the origin.
234 Chapter 4
Evaluate the integral ydx - y(x - 1)dy + y2zdz around the
intersection of the hemisphere x
2
+ y2 + z
2
= 4 (z 2 0) and the
cylinder (x - 1)2 + y2 = 1 in the clockwise direction as viewed
from the origin.
Calculate the work done by the force field F(x, y) = (x
2
- y2 )i +
2xyj in moving a particle once around the (a sides of the
square I x I =I , I y I =I , (b) unit circle x
2
+ y J = I , both in the
counterclockwise direction.
A force field is given by F(x, y, z) = (y - z)i + (z - x)j - (x - y)k.
Find the work done by this force field in movin q a pyticle
once around the intersection of the hemisphere x + y + z
2
= 4
(z > 0) and the cylinder x
2
+ (y - 112 = 1 traced in the clockwise
direction as viewed from the origin.
A particle moves along a path C: R(t) = sin t i + cos t j + t k (t 2
0) under the influence of a force F (x, y, z) = (y - l )i + zj +xk.
Suppose the particle lectves its path and moves along the tan-
gent at t = n/2, calculate the work done by the force on the
particle in the time interval [n/2, n .
1
Evaluate IC f ds, where f(x, y, z) = x - y + z and C is the broken
line segments joining the origin to the point (1, 1, 0) and (1, 1,
0) to the point (1,1,2).
Evaluate I, (xy - z)ds from (1,0,0) to (0,0, 1) along the line seg-
ment x + y =1 joining (1, 0, 0) to (0, 1, 0) and then along the
quarter circle y2 + z
2
= 1 (y > 0, z > 0) joining (0, 1, 0) to ( 0, 0,
1).
Let R(t) (a < t I b) describe a smooth curve C, and let t = $(z) (a
5 z 5 p), where 4, is a differentiable function such that 4,' > 0. If
$( a) = a and $(p) = b, show that R"(z) = R(@(r)) describes the
same curve and preserves the orientation of the curve.
HINT: Show that R'(t) points in the same direction as R"'(z).
Thus establish the validity of the integral
Let R(t) (a I t I b) describes a smooth curve C, and suppose
that F is parallel to the tangent vector at each point on C. Show
that
Integral Calculus of Fields
20. Show that
where M = max I F I on C and L is the length of the curve.
4.4 LINE INTEGRALS INDEPENDENT OF PATH
Let F(x, y, z) be a vector field that is defined and continuous in
a domain D, and let A and B be any two points in D. The line inte-
gral of F from the point A to the point B will generally depend on
the curve C that joins the points. If the value of the integral from A
to B remains the same however we choose the curve that connects
the points, then we say that the integral of F is independent of path
in D. The integral depends only on the points A and B. In such a
case, we write the integral as
An immediate result arising from an integral being indepen-
dent of path is contained in the following theorem.
Theorem 1. Let F be continuous in a domain D. The integral
of F is independent of path in D if and only if
for any piecewise smooth simple closed curve C in D.
Proof: We note that a simple closed curve is a closed curve
that does not cross itself. So suppose that the integral of F around
any simple closed curve in D is zero. Let A and B be any two points
in D, and let C1 and C2 be any two curves that join A to B (Fig. 4.10).
Then we can form a closed curve C consisting of C1 from A to B and
- C2 from B to A, opposite in direction to C2. Since the line integral
around C is zero, we then have
Chapter 4
Fig. 4.10 Integral around a closed curve.
This implies
Thus the line integral of F from A to B along C1 is the same as that
along CZ. Therefore, I F* dR is independent of path in D.
On the other hand, suppose I F* dR is independent of path in
D. If C is a simple closed curve in D, we can choose two points, say
A and B, on C and split C into two parts: C from A to B and C2 from
B to A. Then we have
Integral Calculus of Fields
Caution: Notice that the theorem requires that the integral jc F* dR
must be zero on every closed curve in D in order for the integral to
be independent of path in D. In other words, the line integral may
be zero for a particular closed curve or even for infinitely many
closed curves, and still the integral may not be independent of path
(see Example 5). For this reason, Theorem 1 is only useful for estab-
lishing that an integral is not independent of path by showing that
Jc F. dR t 0 for a certain closed curve C.
Necessary Condition for Independence of Path -
Conservative Field
We now discuss a necessary and a sufficient condition for the
line integral of a vector field to be independent of path. First, we
show that if the line integral of F is independent of path in a do-
main D, then F must be a conservative field; that is, there exists in D
a scalar field @, called a potential function, such that F = V@. Thus
F being a conservative field is a necessary condition for F* dR to be
independent of path in D. To prove this, we define a function 9 by
the line integral of F from any fixed point (x,, y,,, z,, ) to an arbitrary
point (x, y, z) in D, that is,
We shall show that V@ = F, so that the function (4.11) is a potential
function for F.
Indeed, since the integral is independent of path, we may
choose a path from (x, , y,,, z,,) to (x, y, z) to consist of a smooth
curve from (x,, y,,, z,, ) to (x,, y, z ) and the line segment from (xl, y,
z) to (x, y, z) as indicated in Fig. 4.11. By the additive property of
line integrals, we have
Chapter 4
Fig. 4.11 An integral independent of path.
Notice that along the line segment joining (xl, y, z) and (x, y, z), the
coordinates y and z remain fixed so that dy = 0, and dz = 0. Now,
differentiating $(x, y, z) with respect to x, we obtain
Similarly, if we integrate along a smooth curve from (x,,, y,,, z,,) to (x,
yl, z) and then along a line segment from (x, yl, z) to (x, y, z), we
obtain
Differentiating this with respect to y, we find
Integral Calculus of Fields
The final result
is established in the same way. Thus we have shown that VQ, = F, so
that F is a conservative field.
Since V x VQ, = curl (grad Q,) = 0, an equivalent necessary con-
di t i on for I F* dR to be independent of path is curl F = 0. Therefore,
if curl F + 0, we can immediately conclude that the line integral of F
is path dependent.
Sufficient Condition - Conservative Field and Connectivity
The converse of the preceding result is not true unless we
make some restriction about the domain D in which curl F = 0. In
other words, we may have curl F = 0 in D and yet the line integral
of F is not independent of path in D (see Examples 1, 2 below). In
order for the converse to hold, we require that D be s i mpl y connect -
e d. By this we mean that D has the property that whenever a simple
closed curve lies in D, all its interior also lies in D. Thus, for exam-
ple, a domain bounded by a circle or a polygon is simply connect-
ed, whereas a domain bounded by two concentric circles is not.
However, the domain bounded by two concentric spheres is simply
connected, whereas the interior of a torus (think of a donut) is not
(see Fig. 4. 12).
So suppose F = VQ,, or equivalently, curl F = 0 in a domain D,
where D is simply connected. Let A and B be any two points in D
and let C: R(t) = x(t) i + y(t) j + ~ ( t ) , a 5 t 2 b, be a piecewise smooth
curve joining A and B. We will show that the line integral I F* dR
from A to B does not depend on C. Now on the curve C we have
Hence
Chapter 4
Simply connected Doubly connected
(a) Plane domains
Simply connected Doubly connected
(b) Domains in 3-dimensional space
Fig. 4.12 Simply and doubly connnected domains.
Thus the value of the line integral depends only on the value of 0 at
the points A and B. This shows that F* dR is independent of path
in D.
We summarize the results obtained above in the following
theorem.
Theorem 2. Let F = Pi + Qj + Rk be a continuous vector field
in a domain D, and let A, B be any two points in D. If the line inte-
Integral Calculus of Fields
gral
is independent of the path joining A and B, then curl F = 0 or, equi-
valently, F is a conservative field, F = VQ. Conversely, if curl F = 0
in D, and D is simply connected, then the line integral of F in D is
independent of path, and
for any two points A and B in Dl where Q is a function such that
VQ = F.
Example 1. Consider the vector field
It is easily verified that F(x, y) = VQ, where Q = arctan (y/x), so that F
is a conservative field in the domain D = {(x, y) I (x, y) + (0, 0)).
However, since the origin has to be excluded, the domain has a
"hole" and, therefore, it is not simply connected. Indeed let C be any
closed curve C that encloses the origin. Then we see that
Hence, by Theorem 1, the integral I F* dR is not independent of path
in D. However, in any simply connected domain that does not con-
tain the origin, Ic F* dR = 0 for every closed curve C in D, which
shows that the integral is independent of path. For example, if we
introduce a "cut" in the xy-plane by deleting the entire non-positive
x-axis, then the resulting domain D* becomes simply connected (no
closed curve can now enclose the origin), and so I F. dR = dQ will
vanish on every closed curve in D".
Example 2. Consider next the vector field
Chapter 4
A ain it is easily verified that F(x, y) = VQ, where $ = In r, r
2
= x
2
+
F
y . Thus F is a conservative field in the doubly connected domain D
= { (x, y) I (x, y) # (0,O) ). For any two points A and B in D, we find
B B
/AB
F d R =IA V $ ~ R =lA d[ln r]
where r~ are rg denote the respective distance of the points A and B
from the origin. Thus the value of the line integral depends only
on the points A and B. Hence the line integral is independent of
path although the domain {(x, y) I (x, y) + (0, 0)) is not simply con-
nected.
This example shows that in Theorem 2 the condition that D be
simply connected is merely a sufficient condition and not a neces-
sary one.
Example 3. Let F = VQ, where Q = x
2
y + yz + z
2
. Calculate
the line integral of F from (1, -1, l ) to (2,1,3) along any path.
Solution: Since F is a conservative field, its line integral de-
pends only on the endpoints (1, - 1, l ) and (2,1,3). Hence, we have
Example 4. Given that F(x, y, z) = (sin y + z) i + (x cos y + e
Z
) j
+ (x + ye
Z
) k is a conservative vector field. Calculate the integral
Solution: Since the vector field is conservative, the integral is
independent of path. So we may choose any path from the origin to
the point (1, K, 2). We choose the path consisting of the line seg-
ments joining (0, 0, 0) to (1, 0, 0), (1,0,O) to (1, K, 0), and finally (1, K ,
Integral Calculus of Fields 243
0) to (1, a , 2). On each of these line segments the integral reduces to
just one term. Thus we obtain
Example 5. Show that the line integral of F = xy i + y j is
zero for any circle with center at the origin, but that F is not a con-
servative vector field.
Solution: Consider any circle of radius r with center at the ori-
gin and represent i t by the parametric equations
Then we find
j; F d R = ~ c x y d x + y d y
On the other hand, if we take C to be the sides of the square
with vertices at (0, O), (1, 0), (1, I), and (0, I), taken counterclockwise,
we find
Therefore, by Theorem 1, the integral is not independent of path
and hence F is not a conservative field.
Example 6. Show that the integral of F = (2xy + z
2
) i + x
2
I+
2xz k around the intersection of the plane x = y and the sphere x +
v2 + z2 = a2 (great circle) is zero.
~ol uti o>: To obtain a parametric representation of the inter-
section of the plane x = y and the sphere, we note that the plane is
defined in spherical coordinates by 8 = a /4, and the sphere is de-
fined by r = a. Hence their intersection has the parametric represen-
tation
Chapter 4 244
Therefore,
+ (2'2(\i a sin cos Q (- a sin $1 d~
0
\ 2 3 2 3 2
=12'(\2 2 a3cos$+ 4 a sin ~ c o s . * a sin
Principle of Conservation of Energy
Suppose we have a continuous conservative force field F
having $ as its potential function. Then, according to Theorem 2, the
work done by F in moving a particle from a point A to a point B is
independent of the path joining A and B, and it is given by
But according to Example 5 of Sec. 4.2, this work is also equal to the
change in the kinetic energy of the yarticle at the points A and 8.
Therefore, if we denote by K(P) the kinetic energy of the yarticle at
the point P, then
The scalar quantity - $ (P) is called the potential energy of the
Integral Calculus of Fields
particle at the point P. Thus equation (4.17) says that in a conserva-
tive force field the sum of the kinetic and potential energies of a
particle moving in that field is constant. This is the statement of the
well known principle of conservation of energy in mechanics. For
this reason a force field in which this principle holds is called a con-
servative field. In this context, Theorem 2 implies that in a conser-
vative field when a particle is moved around a closed curve back to
its starting point, the net work done is zero.
A very important example of a conservative field is, of course,
the gravitational force field considered in Example 4 of Sec. 3.6.
4.3 EXERCISES
1. Let F be a continuous vector field with continuous first deri-
vative in a domain D. If the line integral of F is independent
of path, show that curl F = 0 at each point in D.
2. Using the result of Problem 1, verify that for each of the
following vector fields the line integral is not independent of
path.
(a) F(x, y) = (x
2
+ y)i + (y2 - x)j
(b)
F(x,y) = (y + xcosy)i + ( x+ ysi nx)j
(c)
F(x, y, z) = (x
2
+ sin y + z)i + (y cos y + l ) j + (z
2
- xy)k
(d)
F(x, y, z) = (2xy - z
2
)i + ( y2 + x cos z)j + x
2
zk
In each of Problems 3 through 9, the given vector field is a con-
servative vector field. Verify that curl F = 0 and determine a poten-
tial function for each of them.
F(x, y) = [x/(x2 + y
2
)li + [y/(x2 + y
2
)I j
F(x, y) = (y2 cos x + 2xeY )i + (2y sin x + x2ey )j
F(x, y, z) = (y z
2
eXY - y sin x)i + (cos x + x z
2
eXY ) j + (2zeXY + 1)k
F(x, y, z) = (ye
Z
- z)i + (xe
Z
+ 2y)j + (xye
Z
- x)k
F(x, y, z) = (2xz - ~ e - ~ ) i + (cos (z - 1) + e-x)j + [x
2
- y sin (z - l )]k
F(x,y,z) = (xi + yj + zk)/(x
2
+ y2 + z
2
) , ( x, y, z) + (0,0,0)
F(x, y, z) = (3e
Z
+ 2xy)i + (x
2
+ z sin y)j + (3xe
Z
- cos y)k
Show that the line integral I, F* dR of the vector field F(x, y) =
(x
2
+ xy)i + (x
2
+ y2)j is zero when C is (a) any circle with
center at the origin; (b) the sides of the square I x I + I y I = 1;
(c) the sides of the triangle with vertices at (-a, 0), (0, b), (a, 0).
Can you generalize this result for arbitrary C? Is F a conserva-
tive field? Does this contradict Theorem 1 or Theorem 2?
Show that for the vector field F(x, y) = (x + y2 )i + (xy - l)j,
IC F* dR = 0 when (a) C is any circle with center on the x-axis,
Chapter 4
(b) C is the sides of the square with vertices at (0, O), (1, -I), (2,
O), (1,l). Is F is a conservative field?
12. Show that F = (2xy + z3 )i + (x
2
+ 2y)j + 3xz
2
k is a conservative
field and calculate its line integral from (0, 1, 1) to (1, 2, 3)
along any p ath.
13. Show that F = (2xy + e
Z
cos x)i + (x
2
+ z)j + (e
z
sin x + y)k is
a conservative field and calculate its line integral from the
origin to the point (n/2,4, 1) along any path.
14. Evaluate the integral jc F. dR, where F = yi + zj + xk and C
is the helix R(t) = cos t i + sin t j + tk, 0 < t 1 n.
15. Evaluate the integral
along the line segment joining the points.
4.5 GREEN'S THEOREM IN THE PLANE
For a vector field defined in a domain of the xy-plane, there is
a fundamental theorem which converts its line integral around a
simple closed curve into a double integral over the domain en-
closed by the curve. The theorem is known as Green's theorem, in
honor of the English mathematician G. Green of the 19th century.
The theorem is one of the great theorems of mathematical analysis.
Its generalizations to three dimensions provide the foundation for
important theorems in applied mathematics. In this section we state
the theorem in its general form and prove it for special domains
having certain properties.
Theorem 1. Let D be a domain in the xy-plane bounded by a
piecewise smooth simple closed curve C. If P, Q and their first par-
tial derivatives are continuous in D and on C, then
where the line integral around C is taken in the positive (counter-
clockwise) direction.
Proof: We observe that it is sufficient to establish the follow-
Integral Calculus of Fields
ing two identities:
and
since their sum leads to the formula (4.16). We establish these iden-
tities in the special case where the boundary C of D can be split into
two parts with two different representations. We assume that C can
be represented by:
C1: y = f(x), C2: y = g(x), f(x) < g(x),
a 5 x r b
and by
C*l: x = h(y), C* 2: x = k(y), h(y) < k(y), c 5 y s d
as shown in Fig. 4.13.
To prove (4.17) we consider the decomposition of C as shown
in Fig. 4.13(a). Evaluating the double integral over D, we find
Fig. 4.13 Special domains for Green's theorem in the plane.
Chapter 4
Similarly, decomposing C into the two curves shown in Fig.
4.13(b), and performing the double integral (4.18) over D, we find
Combining these results, we obtain (4.16).
For a domain of the type shown in Fig. 4.14, we divide it into
two domains Dl and D2 (Fig. 4.14(a)) to prove (4.17), and into D",
and D*2 (Fig. 4.14(b)) to prove (4.18) by introducing a "cut". In each
case we see that the boundary of the subdomains can be treated in
the same way as the boundary of the domain in Fig. 4.13. For exam-
Integral Calculus of Fields 249
yle, to establish (4.17) we perform the double integral of 3P/dy over
Dl and D2 and add the results. When the integrals along the bound-
aries are combined, the line integral along the "cut" C C2 cancels
out since it is raced in opposite direction. The integrals along the
other parts of the boundary add up to the integral along the bound-
ary C of the domain. On the other hand, the sum of the double inte-
grals over Dl and D2 is exactly the double integral over the entire
domain. Thus (4.17) holds for the domain considered. To establish
(4.18), we divide the domain as indicated in Fig. 4.14(b). The bound-
ary of each of the subdomains can be treated as the domain of Fig.
4.13(b). Integrating 3Q/3x over D*l and D** and adding the results,
we obtain (4.18). Thus we have established Green's thoerem for the
domain of the ty ye shown in Fig. 4.14.
This technique can be used in extending Green' s theorem to
more general type of domains that are frequently encountered in
ay plications.
Fig. 4.14 Other types of domain for Green's theorem.
Example 1. By Green's theorem, evaluate the integral
Chapter 4
where C is the unit circle x
2
+ y2 = 1 traced in the counterclockwise
direction.
Solution: Here P(x,y) = x + y and Q(x,y) = 2x, so that 3P/ay
= 1 and 3Q/ax = 2. Hence, we have
Example 2. Calculate the line integral
where C is the sides of the squre with vertices at (1, 0), (0, I), (-1, 0),
(0, -I), traced in the counterclockwise direction.
Solution: By Green's theorem, we have
This integral can be calculated by iterated integration in the usual
manner, but we can deduce its value by noting that the integral is
equal to Ayc, where A is the area of the square and (xc, yc) is the
centroid of the area. By symmetry, we know that the centroid is at
the origin. Therefore yc = 0 and so the integral is zero.
Example 3. Calculate the line integral
where C is the ellipse x
2
+ 4y2 = 4, traced in the counterclockwise
direction.
Solution: By Green's theorem, we have
= area of the ellipse
= n: ab
Integral Calculus of Fields
where a and b are the major and minor axes of the ellipse. From the
equation of the ellipse, we deduce a = 2, b = 1. Thus
Example 4. Show that
where C is any piecewise smooth simple closed curve that encloses
the origin (see Fig. 4.15). (Cf. Example 1, Sec. 4.4).
Solution: Since the domain enclosed by C contains the origin
at which the integrand is not defined, we exclude the origin by de-
leting the interior of a circle C' of radius E, 0 < E < 1, about the
origin. The doubly connected domain D* is now bounded outside
by C, traced in the counterclockwise (positive) direction, and inside
by C' , traced in the clockwise direction. Now in the domain D*
and so, by Green's theorem, we have
- y d x + xdy
= 0
2 2
Fig. 4.15 Green's theorem in a doubly connected domain.
Chapter 4
Hence we have
- y d x + xdy
1, x
2
+; 2 + Y 2
Note that the circle C' is traced in the clockwise direction. Setting
x =ECOSB, y =&s i n e . wefind
b 2 2 2
- y d x + xdy E (sin 0 + cos B)d0
=I,
= 271
-I x
2
+; E 2
Tangential and Normal Forms of Green's Theorem
We can write Green's theorem in different forms by making
use of the unit tangent vector and the unit outward normal vector
on the curve. For this purpose, let F(x,y) = P(x, y)i + Q(x, y)j, and let
the curve C be represented by R(s) = x(s)i + y(s)j, where s denotes arc
length of C, s 2 0. Since the unit tangent vector on the curve is given
by
dx dy
t = i +
ds ds
and since
equation (4.16) can be written in vector form as
We note that F. t represents the tangential component of F on C. For
this reason we refer to (4.19) as the tangential form of Green's theo-
rem.
Similarly, we note that the outward normal vector n on the
curve is given by
Integral Calculus of Fields
so that
Hence, applying Green's theorem, we obtain
We refer to this as the normal form of Green's theorem. Since the
integrand on the right-hand of (4.20) is the divergence of F, we can
write (4.20) as
In particular, when F = Vu, this yields the integral formula
which relates the integral of the Laplacian of u in a domain with the
line integral of the normal derivative of u on the boundary of
the domain. Other important integral formulas that are derived
from (4.21) are asked for in the Exercises. These formulas have im-
portant applications for many physical p roblems involving the
Laplace operator v2. Both formulas (4.19) and (4.21) can be extended
to vector fields in three dimensions as we shall see in later sections.
4.4 EXERCISES
In each of Problems 1 through 6, evaluate the line integral by
using Green's theorem.
) dx + 2xy dy , where C is the square bounded by the
lines x = O, x = 2 , y = O, y = 2 .
Chapter 4
2 2 2
:( 2xy2dx + x y dy , where C is the ellipse 4x + y = 4.
2 Y
(x cos x - e
Y
) dx - (y + xe ) dy , where C is any yiecewise
k
smooth simvle closed curve.
2
1;-xiy dx + xy dy , where C consists of the upper semicircle
y = d(a2 - x
2
) (- a 5 x 5 a) and the diameter on the x-axis.
2
(cos x - 3y) dx + (x + ?e
y
) dy , where C is the circle of
radius 2 and center at (1,2).
3 3
/; (3x3- y3) dx + (x + ?y ) dy , where C is the unit circle
with center at the origin.
Verify Green's theorem for the line integral
where C is the boundary of the region in the first quadrant
bounded by x = 1, y = x
3
and the x-axis.
Show that
is zero on any circle C with center on the y-axis. Is the integral
independent of path?
Let u(x, y) = x
2
+ 2xy - Y2 be defined in a simply connected
domain bounded by a smooth simple closed curve C. Show
that
I'
Let u(x, y) = x2 - 2xy + y2. Evaluate Ic (3u/3n)ds where C
is the unit circle with center at the origin.
Evaluate the integral Ic (au/Sn)ds, where u(x, y) = x
3
- y3 and
C is the ellipse ( ~ / a ) ~ + (y /b)2 = 1.
Integral Calculus of Fields 255
12. Evaluate the integral i, u(au/an)ds, where u(x, y) = (x + y) 2
and C is the unit circle with center at the origin.
13. Let
Evaluate the integral ic F* dR around the circle of radius 2
and center at the origin.
14. Let
and
where (x, y) t (0, I), (-1, 0). Evaluate 1, Pdx + Qdy where C is
the circle x
2
+ y2 = 6.
15. Derive the following integral formulas from (4.21):
16. Derive the integral formula
from (4.16).
Chapter 4
4.6 PARAMETRIC REPRESENTATION OF SURFACES
A surface S may be defined as a set of points (x, y, z) satisfying
an equation of the form
f(x, y, z) = 0
(4.23)
where f is a continuous function in some domain D. If we regard f
as a scalar field, then the surface S is just a level surface of f. We call
(4.23) an implicit representation of the surface. If it is possible to
solve for one of the variables in terms of the other two, say, z in
terms of x and y, then the surface may also be represented by
We call this an explicit representation of the surface. For example,
the unit sphere is defined implicitly by the equation x
2
+ y2 + z
2
- 1
= 0. If we solve for z, it can also be represented explicitly by the two
equations
2 2 2 2
z = - \ 1 - x - y and z = \ 1 - x - y
The first equation describes the lower hemisphere, while the second
describes the upp er hemisphere.
For many theoretical considerations, however, it is more con-
venient to represent a surface by yet another method, namely, para-
metric representation. This method is the counterpart of the para-
metric representation of a space curve by means of three equations
involving a single parameter. Parametrically, then, a surface S is
represented by equations of the form
X = x(u, v), y = Y(U, v), z = z(u, v) (4.25)
where u and v are two parameters which range over some domain
D* in the uv-plane. The functions on the right-hand side of (4.25)
are assumed to be continuous in D*. As (u, v) ranges over D*, the
point (x(u,v), y(u,v), z(u,v)) traces the surface (Fig. 4.16). In this
sense, the surface may be regarded as the image of the domain D*
under the mapping defined by (4.25).
The parametric equations in (4.29) can be combined into a
single vector equation
Integral Calculus of Fields
Fig. 4.16 Parametric representation of a surface.
where R(u, v) denotes the position vector of the point on the surface
corresponding to the point (u, v) in D*. For a surface defined by an
equation of the form (4.24), the variables x, y play the role of the pa-
rameters u, v, so that its vector equation is given by
As in the case of space curves, the parametric representation of
a surface is not unique. This means that we may replace the para-
meters u, v in (4.25) by other parameters, say, 5. q, where 5, q are
related to u, v by the equations
Here it is necessary to require that the Jacobian S(k.q)/a(u,v) + 0 in
D* in order to ensure the existence of the inverse transformation
This guarantees that the transformation between the (u,v)-plane
and the (5, q)-plane is one to one.
Chapter 4
Fig. 4.17 Parametric representation of a sphere.
Example 1. The sphere of radius a and center at the origin can
be represented p arametrically by the equations
x = a sin u cos v, y = a sin u sin v, z = a cos u
where 0 5 u I n, and 0 < v < 2n (see Fig. 4.17). The upper hemis-
phere is defined by 0 2 u < n/2, 0 5 v < 2n, and the lower hemis-
phere is defined by n/2 < u I n, 0 I v < 2n. The equations x = acos v,
y = a sin v, z = 0, when u = n/2, represents the equator of the sphere.
Example 2. The paraboloid z = x
2
+ y2 can be represented
para- metrically by the equations
where (u, v) ranges over the whole uv-plane. An alternative repre-
sentation is given by
where u 2 0 and 0 5 v < 2n. For a fixed value of z, the equations
describe a circle on the surface, and for a fixed value of v the equa-
tions represent a parabola, which is the intersection of the plane y =
x tan v and the paraboloid (Fig. 4.18).
Integral Calculus of Fields
Fig. 4. 18 Parametric representation of a paraboloid.
Tangent Plane and Normal Vector on a Surface
Let S be a surface represented parametrically by
where the functions x(u,v), y(u,v), z(u,v) are continuous and have
continuous first order derivatives in a domain D" of the uv-plane.
Let us consider the position vector
of a point on S. If we set v = v,,, then R(u,v,) = x(u,v,,)i + ~ ( ~ , v , , ) j +
z(u,v,)k represents a curve on the surface with u as the parameter.
Its tangent vector is given by dR(u,v,,)/du. Similarly, if we set u =
u,,, the equation R(u,,,v) = x(u,,,v)i + y(u,,,v)j + z(u,,,v)k represents
another curve on the surface with its tangent vector given by r)R(u,,,
v)/dv (see Fig. 4.19). The two tangent vectors to the curves at the
point R(u,,,v,,) determine a plane, which is the tangent plane to the
surface at R(u,,v,,) A unit vector normal to the tangent plane is
given by
Chapter 4
Fig.
v-curve
4.19 The unit normal vector n and the tangent plane.
It is worth noting that if the surface is defined by z = z(x, y) so that
its vector equation is given by
R(x, y) = xi + y j + z(x, y)k
then
This is the gradient of f(x, y, z), where f(x, y, z) = z - z(x, y), which we
know is orthogonal to the surface f(x, y, z) = 0.
Example 3. Find a normal vector to the surface represented
by the equation
R(u, v) = u(cos v i + sin v j) + (1 - u
2
)k, u 2 0, 0 5 v < 27t
Integral Calculus of Fields
Solution: First, we determine 3R/3u and aR/av:
3R
= cos v i + sin v j - 2u k,
aR
= u(- sin v i + cos v j )
au av
Hence a normal vector to the surface is given by
3R 3R 2
n = x = 2u (cos v i + sin v j ) + uk
au 3v
Smooth, Piecewise Smooth and Orientable Surfaces
Suppose that S is a surface represented by an equation of the
form (4.26). We say that S is smooth if 3R/3u and 3R/3v are con-
tinuous and the normal vector 3R/3u x 3R/3v is not zero at any
point on S. Geometrically, this means that at each point on the sur-
face there is a normal vector which changes continuously as it moves
over the surface - thus a smooth surface has no corners or edges. If a
surface is not smooth but consists of a finite number of surfaces each
of which is smooth, then the surface is said to be piecewise smooth.
For example, the surface defined by a sphere is smooth and the faces
of a cube constitute a piecewise smooth surface.
Let S be a smooth surface. Then at any point on S we can
choose a unit normal vector n. Obviously, there are two choices and
they are opposite in direction. If n can be chosen at each point on S
in such a manner that it varies continuously as it moves about the
surface, then S is said to be ori t wt uhl c or t wo- s i dt 7d. The side of S
on which n emerges is called the "positive" side of the surface. If S
is a closed surface, then by convention n is chosen to point outward.
Practically all of the quadric surfaces encountered in calculus are
orientable surfaces.
It may appear surprising that not all surfaces are orientable. A
well known example of a non-orientable surface is the so-called
Mobius strip, which has only one side (Fig. 4.20). This surface may
be formed by taking a long rectangular strip of paper, giving it one
twist, and then joining the ends together. It is readily seen in this
case that when a normal vector n at a point P on the surface is
moved continuously around the strip, its direction reverses when it
returns to the point P. Thus, the vector n cannot be chosen so as
to vary continuously over the entire surface. In other words, the
surface is "one-sided" or nonorientable.
Chapter 4
Fig. 4.20 Mobius strip.
If a smooth surface S is bounded by a simple closed curve C
we may associate with the orientation of S a positive direction on C.
To achieve this, imagine that an object is moving along the boundary
C on the positive side of the surface. The direction in which the
object must move on C in order that the surface will at all time lie to
its left is designated as the positive direction on C (Fig. 4.21). Using
this convention, a yiecewise smooth surface may also be oriented by
choosing the normal vector on each smooth portion of S in such a
way that along a common boundary B of two portions of S, the posi-
Fig. 4.21 Orientation of a surface with respect to its boundary.
Integral Calculus of Fields
tive direction of B relative to one portion is opposite to the positive
direction of B relative to the other portion (see Fig. 4.21). In the
sequel we will consider only piecewise smooth orientable surfaces.
4.5 EXERCISES
In each of Problems 1 through 7, describe the surface reyre-
sented by the given parametric equations, and determine the coor-
dinate curves on the surface. Eliminate the parameters u, v and
obtain an implicit reprentation of the surface.
1. x =a c o s v, y = a s i n v , z = u ( O<v <2 n , u 2 0 )
2. x = LI cos V, y = u sin v, z = u
2
(u 2 0, O 5 v < 271)
3. x = u cos \i, y = u sin v, z = au (a > 0, u 2 0, 0 I v < 271)
4. x = a sin u cos v, y = b sin u sin v, z = c cos u (0 I u < n, 0 I v
5 27C)
5. x = a ( u + v ) , y = b ( u - v ) , z = u v ( - - < u < - , - - < v < - )
6. x = a sin u cosh v, y = b cos u cosh v, z = c sinh v (0 < u < 27~,
- - < v < - )
7. x = au cosh v, y = bu sinh v, z = u (- - < u < m, - m < v < - )
In each of Problems 8 through 14, find a parametric represen-
tation of the given surface.
8. The plane ax + by + cz + d = 0.
9. The parabolic cylinder z = x
2
.
10. The elliptic paraboloid
2 2
X
+ = 1 - z
2
a b2
11. The ellipsoid
a b c
12. The hyperboloid of one sheet
Chapter 4
13. The hyperboloid of two sheets
2 2 2
- y - z = l
2
a b2 c
2
14. The hyperbolic paraboloid
2 2
- Y -
z
-
a2 b
2
In each of Problems 15 through 19, determine a unit normal
vector to the given surface, and find an equation of the tangent
plane to the surface at the given point.
15. R(u, v) = u cos v i + u sin v j + u
2
k,
(1, -1,2).
16. R(u, v) = u cos v i + u sin v j + uk, (d3,1,2).
17. R(u, v) = 'i2(sin u cos v)i + 2d2(sin 11 sin v)j + (43 cos u)k, (1/2,
1,3/2).
18. R(u, v) = 2(u + v)i + (u - v)j + uvk, (2, -3, -2).
19. R(u, v) = 2(sin u cosh v)i + 3(cos u cosh v)j + (sinh v)k,
(1,343/2,0).
20. Let S be a smooth surface represented by R(u, v) , (u, v) E D",
and let r be a smooth curve in D" described by u = f(t), v =
g(t) (a 5 t 5 b). Then R*(t) = R(f(t), g(t)) represents a smooth
curve C on S. Show that the tangent vector dRS(t)/dt to the
curve C is a linear combination of the tangent vectors 3R/3u
and 3R/av to the coordinate curves on S, and thus deduce that
dR*(t)/dt is orthogonal to 3R/3u x 3R/3v.
4.7 SURFACE AREA
We know that the area of a plane region can be determined by
calculating a double integral. In this section we derive a double
integral formula for calculating the area of a surface. The idea of
area of a surface will later be used in the study of surface integrals of
scalar and vector fields. So let S be a smooth surface represented by
the equation
where (u, v) ranges over a domain D" in the uv-plane. To facilitate
the derivation of an expression for an element of area on the surface,
we regard S as the image of the domain D* under the transformation
Integral Calculus of Fields 265
(4.28), and consider the image of a rectangular region ABCD in D*
with vertices at (u, v), (u+Au, v), (u+Au, v+Av), (u, v+Av), see Fig.
4.22. Under the transformation (4.28) the sides of the rectangle are
mapped onto coordinate curves on S so that the rectangle itself is
mapped onto a curvilinear parallelogram, denoted by AS, with the
corresponding vertices A', B' , C' , D'. We assert that the area of AS is
approximately equal to the area of a parallelogram whose sides are
L ~ A ~ and l E I A v
To show this, we imagine the parameter u as representing
time when v is held fixed. Then 3R/3u is a velocity vector, and the
term 13R/3u I Au represents the distance traveled by the point A'
along the u-coordinate curve during the time interval Au. Now,
when Au is small, I dR/du I Au is approximately equal to the length
of the side A'B' of AS. Similarly, for small Av, 1 dR/dv I Av is ap-
proximately equal to the length of the side A'D'. Therefore, the area
of AS is a yyroximately equal to the area of the parallelogram whose
sides are equal to the magnitude of the vectors (aR/Ju)Au and
(aR/dv)Av, for small Au and Av.
We know that the magnitude of the cross product of two vec-
tors represents the area of the parallelogram formed by the vectors.
Therefore, the area of the parallelogram formed by (3R/du)Au and
(aR/av)Av is equal to
v
A . [
- - - - -
- - - - - --.-
"+
A
? v *p
I : u
0 u u+Au 0
Fig. 4.22 Area of a surface.
Chapter 4
and so we have
area of AS = aR
aR AUAV
a u xa Vi
Thus we define the area of a surface as follows:
Definition 1. Let S be a smooth surface represented by R =
R(u, v), where R(u, v) is co~~t i nuousl y differentiable in D". The area
of the surface S is given by the double integral
A(area of surface S) =
The formula
is called the element of surface area on S.
Example 1. Verify the surface area 4na2 of a sphere of radius
a using the formula (4.29).
Solution: From the representation of a sphere
R(u, v) = a(sin u cos v)i + a(sin u sin v)j + a cos uk ,
3R/au = a(cos u cos v)i + a(cos u sin v) j - a sin uk
3R/3v = - a(sin u sin v)i + a(sin u cos v)j
and so
2
~ ~ x ~ f l ~ = a sin u
Hence the area of the sphere is
2n n 2 2
Area = il il a sin u du dv = I n a
Integral Calculus of Fields
Example 2. Find the surface area of the circular cone
Solution: We have
3R/3u = cos v i + sin v j + k , 3R/3v = - u sin v i + u cos v j
and so
SR 3R
x = - uc os v i - u s i n v j + u k
3u 3v
Hence
so that the area is given by
Alternate Formulas for Surface Area
If the surface is represented by an equation of the form z = z(x,
y), where (x, y) ranges over a domain D on the xy-plane, we may use
x and v as the parameters and write
R(x, y) = xi + yj + z(x, y)k
Then
so that
Chapter 4
Hence an alternative formula for the surface area is
so that
Notice that the integral (4.31) is to be integrated over the domain D
which is the projection of S onto the xy-plane.
The formula (4.32) has an interesting and important applica-
tion. Recall that a unit normal vector on the surface is given by
If y denotes the angle between n and k, then
cosy = n .k=l/d-
Hence, by (4.32), we have
and so (4.31 ) can be written as
Equation (4.33) says that the projection (cos y dS) of the ele-
men of surface area dS on the xy-plane is precisely the element of
area (dA = dx dy) on that plane (Fig. 4.23). Thus if S lies on a plane
on which cos y > 0, then, by (4.33), the area is given by
area of S = (sec y )(area of D)
where D is the projection of S on the xy-plane. This result is some-
times called the "area cosine principle."
If the surface S can also be represented by an equation of the
form y = y(x, z), (x, z) E Dl, or x = ~ ( y , z), (y, Z) E Dl', then its surface
Integral Calculus of Fields
dA = cos y dS
Fig. 4.23 Area cosine principle.
area can also be calculated from the formula
A =I ID sec a dx dz
Notice that D' and DM are the projections of the surface S on the xz-
and yz-y lanes, respectively.
Example 3. Find the surface area of the paraboloid z = x
2
+
y2 that lies inside the cylinder x
2
+ y2 = a*.
Solution: The projection on the xy-plane of the surface in
question is defined by D = ( (x, y) I x
2
+ y2 5 a
2
). From the equation
of the paraboloid, we find
Hence, by (4.34), we have
Chapter 4
Introducing polar coordinates, we obtain
Example 4. Find the surface area of the portion of the hemi-
sphere
2 2 2
z = \ 4a - x - y
that is cut off by the cylinder (x - a)2 + y = a 2.
Solution: From the equation of the hemisphere, we find
zx = - x/z , zy = - y/z. Hence
where D = {(x, y) I (x - a)2 + y 5 a 1. In polar coordinates, we
observe that the circle bounding the domain D is described by the
equation
r = 2a cos 0 (0 5 0 5 n ).
Thus we obtain
2 2
= 2 a i n [- 2a sin 0 + 2a] d0 = - 8a + 4na
4.6 EXERCISES
1. Using the area cosine principle, find the area of the triangle
with vertices at (1,0, 0), (0, 1, 0), (0, 0, l ) .
2. Find the area of the portion of the plane x + y + z = a that lies
Integral Calculus of Fields 271
inside the cylinder x
2
+ y2 = a
2
.
Find the area of the portion of the plane y = z that lies inside
the cylinder x
2
+ y2 - 2ay = 0.
Find the surface area of the surface described by
R(u, v) = (u cos v)i + (u sin v)j + (1 - u
2
)k (u t 0 , 0 5 v c 27r)
above the xy-plane.
Find the surface area of the portion of the surface
corresponding to 0 5 u 5 da, 0 5 v < db.
Find the surface area of the portion of the sphere x
2
+ y2 + z
2
=
a
2
lying inside the cylinder x
2
+ y2 = ay, where a z 0.
Find the surface area of the portion of the cylinder x
2
+ z
2
= a
2
that is cut off by the cylinder y2 + z
2
= a
2
.
Find the surface area of the surface described by the equation
R(u, v) = u cos v i + LL sin v j + v k, where 0 < u 5 a, 0 < v 1 2 7 ~ .
Find the surface area of the portion of the s here x
2
+ y2 + z
2
=
a
2
that is cut off by the circular cone x
2
+ yY= z
2
.
Find the surface area of the portion of the cone x
2
+ y2 = z
2
(z 2 0) that is cut off by the plane 22 = y + 1.
Find the surface area of the ortion of the cone x
2
+ y2 = z
2
Y
lying inside the cylinder x + y2 = 2ay (a > 0).
Find the surface area of S: R(u, v) = ui + vj + uvk that lies
inside the cylinder x
2
+ y2 = a2.
Find the surface area of the portion of the cylinder x
2
+ z
2
= a
2
in the first octant between y = 0 and y = x.
Given the sphere x
2
+ y 2 + z
2
= 4a2 and the paraboloid x
2
+ y2
= a(z + a), where a > 0. (a) Find the surface area of the sphere
that lies inside the paraboloid. (b) Find the surface area of the
paraboloid that lies inside the sphere.
Find the area of the surface defined by
Find the area of the portion of the surface z = xy that is inside
the cylinder x
2
+ y2 = 1.
Find the area of the portion of the surface z = 2(xR/2 + y0/ 2)/ 3
that lies above the domain bounded by the lines x = 0, y = 1
and x = 3y.
A torus S can be represented by the parametric equations
Chapter 4
( 0 5 $ < 2n, 0 5 8 < 2n), R > 1 (R is fixed). Show that the
surface area of the torus is given by 4n R. NOTE : The torus
has an inner radius of R - 1 and an outer radius of R + 1. For
fixed 8, the equations describe a cross section of the torus with
$ serving as the polar angle.
4.8 SURFACE INTEGRALS
We now study the integration of a scalar or a vector field on a
surface. Such an integral is called a surface integral. A surface inte-
gral is a natural generalization of the concept of a double integral on
a plane domain. In fact, just as we calculate line integrals by reduc-
ing them to ordinary single integrals, we also calculate surface inte-
grals by transforming them into double integrals. We first define the
surface integral of a scalar field.
Surface Integral of a Scalar Field
Definition 1. Let f be a scalar field defined and continuous in
a domain D. Let S be a smooth surface in D represented by R(u, v)
= x(u,v)i + y(u,v)j + z(u,v)kf where R(u, v) is continuously differen-
tiable in a domain D" of the uv-plane. The surface integral of f on S,
denoted by ijS f dS, is the double integral
Notice that when f = 1, the integral (4.37) gives the surface area of S
as defined in Sec. 4.7.
If S is piecewise smooth, we define the surface integral of f on
S as the sum of the integrals over the pieces of smooth surfaces com-
prising S.
For a surface that is represented by an equation of the form z =
z(x, y), where (x, y) ranges over the projection D of the surface on the
xy-plane, the integral (4.37) can be written as
Integral Calculus of Fields
where
Similarly, if the surface can be represented by the equation x =
x(y, z) or y = y(x, z), the integral (4.37) may also be written as
where
D' and D
M
are the projections of the surface on the yz- and xz-coor-
dinate planes, respectively.
Surface integrals of scalar fields occur in many physical pro-
blems. For example, suppose we have a thin sheet of material in the
shape of a surface S whose density at each point (x, y, z) is given by
p(x, y, z). Then the mass of the material is given by the surface inte-
gral
Further, the center of mass (xc, y,, zc) of the material are given by
the following integrals:
where M is the mass of the material. Finally, the moment of inertia
of the material about a line L is given by the surface integral
Chapter 4
where G(x,y,z) denotes the distance of each p i n t (x, y, z) from the
line L.
Example 1. Find the surface integral of f(x,y,z) = xy + z on the
upper half of a sphere of radius a.
Solution: We represent the surface by
R(u, v) = a( sin u cos v i + sin u sin v j + cos u k ),
OI u I n / 2 , O<v <2 n . Then
dR/3u = a(cos u cos v i + cos u sin v j - sin u k)
and
dR/dv = a(- sin u sin v i + sin u cos v j)
so that
IdR/du x aR/dv I = a
2
sin u
Thus
2
[(a sin u cos v)(a sin u sin v) + a cos u]a sin u du dv
3
cos v sin v + a cos u sin u) du dv
Example 2. Evaluate the surface integral of f(x, y ) = x
2
+ y2
on the upper half of the unit sphere using (4.38).
Solution: We note that the projection D of the surface on the
xy-plane is the disk x
2
+ y2 5 1. Now from the equation of the unit
sphere, $(x, y, z) = x
2
+ y2 + z
2
= 1, we find the outward unit normal
vector
Thus, sec y = l / z = 1/ (1 - x
2
- y2
and so, by (4.38), we have
Integral Calculus of Fields
Example 3. The density of a hemispherical shell of radius a
is given by p = z. Find the moment of inertia about the z-axis.
Solution: The outward unit normal vector on the hemisphere
is given by n = (xi + y j + zk)/ a. Thus the moment of inertia about
the z-axis is equal to
where D is the disk x2 + y2 5 a
2
. Since sec y = a/ z on the surface, we
find, using polar coordinates,
Surface Integral of a Vector Field
We now consider the definition of an integral of a vector field
on a surface. To motivate the definition, we consider the calculation
of the amount of fluid that flows across a surface S submerged in a
fluid flowing in a domain D. Let V(x, y, z) denote the velocity and
p(x, y, z) the density of the fluid. Then the rate of flow (mass per
unit area per unit time) of the fluid is given by F = pV. This is called
the flux density of the fluid. If S is a plane surface and F is constant
in D, then clearly the mass of fluid flowing across the plane surface
is given by (F* n)A, where A denotes the area of the surface and n is
the unit normal vector on the side of S from which the fluid emer-
ges (Fig. 4.24).
To calculate the amount of fluid flowing across an arbitrary
smooth surface S, we subdivide the surface into n smaller surfaces
Si and calculate the amount of fluid flowing across each of the sur-
faces Si (1 5 i 5 n). So let ASi denote the surface area of Si, ni the unit
normal vector on Si, and Fi the flux density on Si. When n is suf-
ficiently large so that each Si is very small, ni and Fi are nearly con-
Chapter 4
Fig. 4.24 Mass of fluid flowing across a surface S.
stant on Si. Hence the mass of fluid flowing across Si is approximate-
ly equal to (Fie ni)ASi, and so the amount of fluid flowing across the
entire surface S is approximately equal to the sum Xi=," ( Fi ni )AS
i
.
The limit of this sum as IT increases leads to the double integral
which is precisely the surface integral of the scalar field F. n on S.
We call this integral the sif rf ace i nt egral of F on tile sir rf uce S.
Notice that the scalar field F* n represents the normal compo-
nent of the fluid flow on the surface. Thus if the surface is tangential
to the flow or the flow is tangential to the surface, then F n = 0,
which means that there is no fluid flowing across the surface.
Now if the surface S is represented by R = R(u, v), we know
that the unit normal vector on S is given by
so that
Integral Calculus of Fields 277
Therefore, we define the surface integral of F on S as follows:
Definition 2. Let F be a vector field defined and continuous
in a domain D, and let S be a smooth orientable surface in D repre-
sented by R = R(u, v), where R(u, v) is continuously differentiable in
D*. The surface integral of F on S, denoted by II Fa n dS, is the
double integral
If S is piecewise smooth, we define the surface integral (4.41) as
the sum of the integrals over the smooth surfaces comprising the
surface S. When S is a closed surface, it is customary to take n to be
the outward unit normal vector.
Other Notations of Surface Integrals
If F = Pi + Qj + Rk and we express the unit normal vector n
on S in terms of its direction cosines, n = cos a i + cos p j + cos y k,
then the integral (4.41) can be written in the form
This corresponds to the notation jC Pdx + Qdy + Rdz for the line
integral ,fc F* dR. Further, if the surface is represented by z = z(x, y),
(x, y) E D, SO that dS = sec y dx dy, then the integral (4.41) can also be
written in the form
Corresponding formulas can also be obtained using the relation dS
= sec p dy dz or dS = sec a dz dx, whenever the surface S can be
represented by y = y(x, z) or x = x(y, z).
278 Chapter 4
Example 4. Calculate the surface integral of the vector field
F(x, y, z) = xi + yj + 3zk on the surface S: R(u, v) = a(cos u i + sin u j)
+vk, OI u<. n, O<v <n .
Solution: On the surface we have
r3R 3R
x = a(cos u i + sin u j )
r3u 3v
and F = a(cos u i + sin u j) + 3vk. I-Ience
an an 2 2 2
F. x =( a cosu) + (asin u) = a
3u 3u
so that
Alternatively, we note that the surface can also be represented by the
equation
2 2
y = \ a - x ( - a < x < a , O l z l n )
The unit normal vector on the surface is given by n = (x i + y j)/a, so
that
F*n = (x
2
+ y2) / a
The projection of the surface on the xz-coordinate plane is the rec-
tangular domain D: - a 5 x 2 a, 0 5 z < n . Using the relation dS =
sec p dx dz = (a/y) dx dz, we thus obtain
Example 5. Calculate the surface inte ral of F(x, y, z) = xi + yj
5
+ zk on the hemisphere S: z =d( a2 - x
2
- y ), (x
2
+ y2 < a
2
), where
the unit normal vector n points away from the origin.
Solution: The unit normal vector on the surface is given by n =
(xi + yj + zk)/a, so that F n = (x2 + y2 + z
2
)/a. Hence, on the surface,
Integral Calculus of Fields
F* n = a, and so we have
where we made use of the fact that the surface area of the hemis-
phere is 27r a
2
.
Example 6. Evaluate the integral F* n dS, where F (x, y, z)
-
= x
2
i + y2 j + z
2
k and S consists of the faces of the unit cube 0 2 x 2
l , Ol y < l , 0 5 ~ 1 1 .
Solution: We calculate the surface integrals on the faces S, , S2
and S3 where x = 1, y = 1, and z =1, and on which the unit normal
vectors are i , j , k , respectively. We find
since the area of each of the faces is equal to 1. The surface integrals
on the corresponding opposite faces, where x = 0, y = 0, and z = 0,
are zero since F* n = 0 on each of those faces. Therefore, the surface
integral is equal to
Heat Flux Across a Surface
Surface integrals of vector fields also arise in the study of heat
conduction in solid bodies. Let U(x, y, z) denote the temperature at
a point (x, y, z) of a heat conducting body. According to the law of
thermodynamics, heat energy flows in the direction of decreasing
temperature (that is, from points of higher temperature to points of
280 Chapter 4
lower temperature) at a rate proportional to the temperature gra-
dient VU. Since VU points in the direction where U increases, it
follows that the amount of heat energy flowing across a unit surface
area is given by F = -kVU, where k denotes the thermal conductivity
of the body. This is known as the heat flux. Therefore the total flux
across the surface S of the body is given by jjs F* n dS.
Example 7. Sup pose that the temperature at each point inside
of a ball of radius a is given by U(x, y, z) = a
2
- x
2
- y2 - z
2
. What is
the heat flux across the surface of the ball, assuming k = I ?
Solution: The outward unit normal vector on the ball is n =
(xi + yJ + zk)/a, and F = - VU = 2( xi + yj + zk). Hence F n = 2(x
2
+
y2 + z ) / a which reduces to F* n = 2a on the surface. Therefore the
heat flux across the surface of the ball is equal to
4.7 EXERCISES
In each of Problems 1 through 5, evaluate the integral Ifs f dS
for the given scalar field f and surface S.
1.
f(x, y, z) = xyz , where S is the triangle with vertices (2, 0, O), (0,
2, O), (O,l, li
2.
f(x, y, z) = x + z
2
, where S is the lateral surface of the cylinder
x
2
+ y2 = a2 0 5 z Sb).
3. f(x, y, z) = b , where S is the upper hemisphere of radius a.
4.
f(x, y, z) = xz , where S is the paraboloid x
2
+ y2 = 1 - z above
the xy-plane.
5. f(x, y, z) = x + y + z, where S is the sphere of radius a. (Use the
symmetry of the problem.)
6. Let S be the triangle with vertices (1, 0, O), (0, 1, O), (0, 0, 1).
Argue by symmetry that
Thus, with very little computation, show that
Integral Calculus of Fields
1 \ 3
M = (area of triangle) =
3 6
7. Let S be the sphere of radius a. As in Problem 6, show that
and verify that I = (1/3)a2 (area of sphere) = 4n a
4
/3 .
8. A homogeneous cylindrical shell has radius a and altitude h.
Find the moment of inertia about its axis.
9. A material in the shape of a thin paraboloidal shell x
2
+ y2 = z
(0 5 z 5 1) has density given by f(x, y, z) = 1 + z. Find the total
mass of the shell.
10. A surface made of certain metal is in the shape of a thin
hemisphere z = ( a
2
- x
2
- y2 ( x
2
+ y2 5 a* ). If its density
is given by f(x, y, z) = z, find the moment of inertia about the
z-axis.
11. A certain material is in the shape of a thin cylindrical shell x
2
+ y2 = 4 (0 5 z 5 2), If the density is given by f(x, y, z) = I x I +
I y I , find the mass of the material.
12. Find the average temperature of the material of Problem 9 if
the temperature at each point (x, y, z) is given by T(x, y, z) = z.
In each of Problems 13 through 19, compute the integral
jjS F* n dS for the given vector field F and surface S, where n is the
normal vector on S with positive z-component.
F(x, y, z) = xi + yj + z k, where S is the triangle with vertices at
(LO, 01, (0, L O), (0, 0, l ) .
F(x, y, z) = yi + zj + xk, where S is the surface of Problem 1.
F(x, y, z) = xi + yj + z k, where S is the paraboloid x
2
+ y2 =
1 - z above the xy-plane.
F x, y, z) = x
2
i + yj + zk , where S is the cylindrical surface x
2
+
1
z = a
2
( O<y <a , zr O) .
F(x, y, z) = yi - zj + x
2
k , where S is the cylinder z
2
= 4y in the
first octant bounded by x = 1 and y = 4.
F(x, y, z) = xi + yj - 2zk, where S is the upper hemisphere of
radius a.
F(x, y, z) = zi + yj + x2 k , where S is the portion of the parabo-
loid y2 + z
2
= x in the first octant bounded by x = 1.
Compute IIs F n dS, where F(x, y, z) = x
2
i + y2 j + z
2
k and S is
the faces of the cube x = + 1, y = + 1, z = + 1. The vector n is
outward.
282 Chapter 4
Repeat Problem 20 when F(x, y, z) = xi + yj + zk .
Compute II F n dS, where F(x, y, z) = xi - yj + z
2
k, S is the
entire surface of the cylinder x
2
+ y2 = 4 (0 5 z 5 2), and n is the
outward unit normal vector.
Compute IIs F . n dS, where F(x, y, z) = zi + (x + y)j - xk , S is
the surface in the first octant consisting of x
2
+ z
2
= 9, x = 0,
y = 0, z = 0, y = 4 and n points away from the origin.
Let the temperature at each point (x, y, z) in space be given by
U(x, y, z) = x
2
+ y2. Find the heat flux across the upper hemi-
shpere of radius a.
Find the heat flux across the surface x
2
+ y2 = 1- z above the
xy-plane, if the temperature is given by U(x, y, z) = y + z.
The flux density of a fluid flow is given by F = - yi + xj + z k.
Find the total flux across the sphere of radius R.
The flux density of a fluid flow is given by F = xi + yj. Find
the rate of fluid flowing across the surcface of the paraboloid
x
2
+ y2 = z (0 5 z 5 4).
Evaluate curl F* n dS, where F = yzi + y 2 j + z
2
k, S is the
lower hemisphere of radius 2, and n is outward.
Evaluate 11 curl F. n dS, where F = yi - xi + xyzk, S is the
surface x
2
+ y2 + 2z2 = 1 (z 2 0).
Let S be a smooth surface defined by R = R(u, v), (u, v) E D*.
Suppose s, t are new parameters related to u, v by the equation
s = s(u, v), t = t(u, v) such that for each (u, v) F D* there corres-
ponds one and only one (s, t) E D"* and vice versa. Set R*(s, t) =
R(u, v), when u and v are expressed in terms of s and t. Show
that R, x q, and R*, x R*, have the same direction if a(s, t)/r)(u,
v) > 0, and opposite direction if 3(s, t)/3(u, v) < 0. Thus verify
that if 3(s, t)/3(u, v) > 0, then
Show that the surface integral il f dS is independent of the
parametric representation of the surface.
Let F = Pi + Qj + Rk and suppose it is represented by z = z(x,
y), (x, y) E D, where D is the projection of S on the xy-plane.
When n has positive z-component, show that
Integral Calculus of Fields
Obtain the corresponding formulas when S has the representa-
tion y =y(x, z) and x = x(y, z).
4.9 THE DIVERGENCE (GAUSS') THEOREM
We recall that Green' s theorem expresses a relationship bet-
ween a line integral on a simple closed curve in the plane and a dou-
ble integral taken over the plane domain bounded by the curve. The
extension of Green's theorem to three-dimensional space leads to the
important divergence theorem or Gauss' theorem. The theorem
relates a surface integral on a closed surface to a volume integral (or
triple integral) over the domain bounded by the surface. We state
the theorem as follows.
THEOREM 1. Let D be a domain in the xyz-space bounded
by a piecewise smooth closed surface S, and let n denote the out-
ward unit normal vector on S. If F is continuously differentiable in a
domain containing D, then
Proof: Let F(x, y, z) = P(x, y, z) i + Q(x, y, z) j + R(x, y, z) k and
let n = cos a i + cos p j + cos y k. Then (4.44) can be written as
Thus the theorem will be proved if we can show that
284 Chapter 4
As in the proof of Green's theorem, we establish these identi-
ties in the special case when the domain D has a non-degenerate
projection on the xy-, yz-, and xz-coordinate planes so that corres-
ponding to each projection, the bounding surface S can be split into
two surfaces S, and S2. We show such a domain D in Fig. 4.25 whose
projection on the yz-coodinate plane is denoted by T. We establish
the first identity in (4.46) for this domain. Let S1 and S2 be the two
parts of the surface and suppose they are represented by the equa-
tions
5,: x=f ( y, z) , s,: x = g(y, z),
where f(y, z) 5 g(y, z), (y, z) E T. By writing the triple integral of
3P/3x in iterated form and then integrating with respect to x first, we
find
We will show that the integral on the right-hand side of (4.47) is
equivalent to the surface integral & Pcos a dS. First we note that
( L Pcos a d~ = ( J, P cos a -+ 1 L 2 p cos a
(4.48)
Now on S2 we have cos a > 0, so that cos a dS = dy dz; hence
On S1 we have cos a < 0, so that cos a dS = - dy dz, and so
Substituting these in (4.48), it follows from (4.47) that
Integral Calculus of Fields
Fig .4.25 The Divergence Theorem
The other two identities can be established in exactly the same
way by considering the projection of the domain on the xz- and on
the xy-coordina te planes.
We have seen the usefulness of the Green's theorem in calcu-
lating certain line integrals along closed curves. Likewise the diver-
gence theorem can be used to evaluate certain surface integrals over
closed surfaces. We illustrate this in the following examples.
Example 1. Evaluate the surface integral F* ndS, where F(x,
y, z) = x
2
i + y2 j + z
2
k and S consists of the sides of the unit cube
0 5 x l l , O< y 5 l , 0 5 z 5 1 .
Solution: We note that div F = 2(x + y + z). Hence by the diver-
gence theorem, we have
Chapter 4
where D is the cubical domain. Since the centroid of the cube is at
(1/2, 1/2, 1/2) and the volume of the cube is 1, we conclude that
Therefore
Example 2. Evaluate the surface integral JIs Fa ndS, where F(x,
y, z) = xi + yj + zk and S is the sphere x
2
+ y2 + z
2
= a
2
.
Solution: Since div F = 3, we have by the divergence theorem,
where we noted that the volume of the sphere is 4n a
3
/3.
Example 3. Evaluate JJs F. ndS, where F(x, y, z) = sin y i + exj
+ z
2
k and S consists of the plane z = 0 and the upper hemishpere
Solution: By the divergence theorem, we have
To calculate the triple integral, we introduce cylindrical coordinates
to obtain
? ?
Integral Calculus of Fields
Therefore,
4.10 APPLICATIONS OF THE DIVERGENCE THEOREM
The divergence theorem is used in many physical problems to
derive mathematical equations satisfied by the physical entities con-
sidered. Before we show some of these derivations, let us see how
the theorem provides a natural definition of divergence that is inde-
pendent of the choice of coordinate system. Let Dr denote the do-
main bounded by a sphere Sr of radius r and center at P, and let F be
a vector field that is continuous including its first derivatives in Dr.
By the mean value theorem for integrals, we have
1 lDrdiv F dV = V(r) div F(Q)
where V(r) denote the volume of the sphere and Q is a point in Dr.
Notice that we purposely did not use dxdydz for the element of vol-
ume since we are not necessarily dealing with rectangular cartesian
coordinates. Now by the divergence theorem, we have
div F(Q) = F n d S
V(r)
As r goes to zero, it is clear that V(r) also tends to zero and the point
Q approaches the center P of the sphere. Thus, in taking the limit as
r tends to zero, we obtain
lim 1
div F(Q) = -- ,,V( r )
F n dS
r
This formula is sometimes used to define the divergence of a
vector field F, where Sr is taken to be an arbitrary piecewise smooth
and closed surface and the limit is taken as the volume shrinks to the
288 Chapter 4
point P. Such a definition clearly does not depend on the coordi-
nate system used, which shows that the concept of a divergence is an
intrinsic property of a vector field. On the other hand, if we intro-
duce a rectangular cartesian coordinate and choose S to be a cube
with center at (x, y, z), it can be shown that the limit (4.50) leads to the
familiar expression
where P, Q, and R are the components of F with respect to the coor-
dinate system.
The Continuity Equation
We now consider some applications of the divergence theorem
in fluid dynamics. In Sec. 4.8 we saw that for the flux density of a
fluid flow, F = pV, the integral Ij F. n dS represents the total flux
across the surface S, that is, the total mass of fluid flowing through
the surface (in the normal direction) per unit time. Hence, by (4.50),
we can say that the divergence of F at a point P is simply the rate of
outward flux per unit volume. Equivalently, we can also say that
it is the rate of change of mass at P per unit volume. Hence, if
div F > 0, then fluid is flowing into the region (presence of a source),
and if div F < 0, fluid is being drained out from the region (presence
of a sink).
Now consider a volume D of the fluid enclosed by a smooth
surface S (Fig. 4.26). The total mass of fluid enclosed by S is given
by the volume integral
The mass changes at the rate given by
Assuming that there is no point in the fiuid where mass is being
introduced (a source) or taken away (a sink), the rate at which the
mass is changing in D must be equal to the rate at which it is flowing
out through the surface S. Therefore,
Integral Calculus of Fields
Fig. 4.26 Mass of fluid in a domain D enclosed by a
surface S.
where the negative sign indicates that the density is decreasing.
Applying the divergence theorem on the left-hand side of (4.51), we
obtain
We assume that pV is continuously differentiable so that the inte-
grand in (4.52) is continuous. Then, since the integral holds for an
arbitrary volume D, we conclude that the integrand must vanish
identically, that is,
dP
div (pV)+-=o
at
(4.53)
Indeed, if the integrand were positive (or negative) at some point in
D, then because it is continuous it would remain positive (or nega-
ive) in a neighborhood of that point. Hence by choosing that neigh-
borhood for D in (4.52), we would have a non-zero integral, thus
contradicting (4.52). In fluid dynamics, equation (4.53) is known as
the cont i nui t y equat i on. It is actually a statement about conserva-
tion of mass.
Chapter 4
The Heat Equation
Next we show how the divergence theorem is used to derive
the so-called heat equation of mathematical physics. Let B be a heat
conducting body in space and let us consider the transfer or flow of
heat energy in an arbitrary volume D in B. Let U(P,t) denote the
temperature at a point in D at time t. According to the law of ther-
modynamics, heat energy flows in the direction of decreasing tem-
perature (i.e., from hot to cold) at a rate that is proportional to the
gradient of the temperature, i.e. VU. Hence, denoting the velocity of
heat flow by v and the thermal conductivity of the body by K
(assumed constant), we have v = - KVU, where the negative sign
indicates that v is opposite in direction to VU. Now the amount of
heat energy flowing across the boundary S of D is given by
By the divergence theorem, this is equal to
2
= - K / / iD V UdV
where v2U = div(grad U) is the Laplacian of U. By the law of conser-
vation of energy, the total heat energy flowing out of the volume D
across its boundary must be equal to the rate at which the heat ener-
gy inside the body decreases. Since the total heat energy inside the
body D is given by the volume integral
where c is the specific heat and p is the density of the body, both
assumed to be constant, it follows that
Integral Calculus of Fields
Hence, equating (4.54) and (4.55), we obtain
from which we again conclude that
The constant K = K/cp in (4.57) is called the coefficient of diffusivity.
Equation (4.57) is known as the heat equation. It is one of the prinii-
pal partial differential equations of mathematics physics.
If the temperature of the body is in equilibrium (steady-stsate),
then U is independent of time and so 3U/3t = 0. In such a case equa-
tion (4.57) becomes
v 2 u = 0 (4.58)
This is called Laylace's equation, which is another important partial
differential equations of mathematical physics.
Gauss's Law
We recall that the electrostatic field at a point (x, y, z) induced
by the presence of a point charge q at the origin is given by
By direct calculation, we see that div E = 0 at all points (x, y, z) + (0,
0, 0). Hence for any closed surface S not enclosing the origin, the
divergence theorem shows that jjj ,-, div E dV = 11 E . n dS = 0.
Thus the net electric flux across any surface S which does not en-
close the origin is zero. When S is a closed surface which encloses
the origin, the divergence theorem cannot be applied since inside
the domain enclosed by S, E is not defined at the origin. In this
case, we will show that the net flux across the surface S is 4nq, that is,
Chapter 4
In order to apply the divergence theorem in the domain boun-
ded by the surface S, we enclose the origin in a small sphere E of
radius E such that C lies inside S (Fig. 4.27). Then applying the diver-
gence theorem in the domain D* that is bounded by S and C, we
obtain
Hence
The surface integral on the sphere C can be calculated. On C
we note that the unit normal vector n is directed toward the origin
and is given by n = - R/E, where R is the position vector of points on
C. Thus on C we have
and so
J
Fig. 4.27 Gauss's law.
Integral Calculus of Fields 293
Therefore, the total electric flux across a closed surface S is equal to
47r times the charge q if the charge lies inside S, and it is zero if the
charge lies outside of S. This result is known as Cuuss' s luzo.
4.8 EXERCISES
In each of Problems 1 through 6, use the divergence theorem
to evaluate the surface integral F n dS for the given vector field F
and surface S, and verify the result by integrating the surface inte-
gral directly.
1. F(x, y, z) = xyi + z
2
j + 2yzk. S is the surface of the cube
0 5 x, y, z 21.
2.
F(x, y, z) = y2 i + yzj + xzk, S is the tetrahedron bounded by x
=O, y =O, z =O, a n d x + y + z = l .
3. F(x, y, z) = xzi + yxj + zyk, S is the surface x
2
+ y2 = 4, z = 0,
y + z = 2 .
4. F x, y, z) = xi + yj + (z - l )k, S is the surface formed by x
2
+
1
y = (z -2)2, z = 0, and z = 1.
5. F(x, y, z) = xyi + y2 j + yzk, S consists of an upper hemisphere
of radius 2 and the bottom plane z = 0.
6. F(x, y, z) = x% + xZyj + x
2
zk, S is formed by the cylinder x
2
+ y2
=1 and the planes z = 0, z = 4.
7. Let F be a vector field that is continuously differentiable and
satisfies div F = 0 in a domain D, except at a point P. Let S be
any closed, piecewise smooth surface enclosing P and let S' be
a small sphere with center at P and lying entirely within S.
Show that
8.
Let F(x, y, z) = (xi + yj + zk)/r
n
, where r = (x
2
+ y2 + z
2
and
n is a positive integer.
(a) Show that div F = - (11 - 3)/ r
n
.
(b) Evaluate the integral f l s F- ndS for n = 1, n = 2 when S is
the sphere x
2
+ y2 + z
2
= a
2
. Can you use the divergence theo-
rem? Explain.
9. Let F be the vector field given in Problem 8 when n = 3, so that
div F = 0. Let S be the surface of the sphere x
2
+ y2 + z
2
= a
2
.
(a) Show that
Chaapter 4
Does this contradict the divergence theorem? Explain.
(b) If S* is any piecewise, smooth closed surface enclosing the
origin, show that
(c) What is the value of jj F a n dS if S does not enclose the
origin?
10.
If u is a harmonic function, show that div (grad u) = I grad u 1 2.
Hence, evaluate the surface integral ijq u(du/dn)dS on the
spherex
2
+ y 2 + z 2 = a
2
when(a) u = x + y + z , (b) u = x
2
+ y2
+ z
2
. (Note that au/dn = Vu* n.)
11. Ev a l ~a t e / ~~( a u / d n ) d So v e r t h e s p h e r e x ~+ y 2 +z 2 =a 2 wh e n
u(x, y, z) = x2 + y2 - z.
12. Evaluate jjS (du/dn)dS over the cube I x I = 1, I y I = 1, I z I = 1,
when u(x, y, z) = x3 + y2 + z.
13. Let u and v be twice cc7ntinuously differentiable functions in a
domain D bounded by a piecewise smooth closed surface S.
Derive the following formulas from the divergence theorem.
2
(a) ( 1 u: : dS = 11 ( D ( ~ u ~ v +u V v) dxdydz
2 2
b ) \ ( s ( u ~ ~ - v ~ ~ ) J s = j ~ D ( ~ v ~ - ~ ~ ~ ) d ~ d y d ~
14. If u is a harmonic function, show that
2
(b) ( 1 u;: dS = ( ( jD I V U dxdydz
Integral Calculus of Fields 295
where S is a piecewise smooth surface bounding the domain D.
15. Let u be a harmonic function in a domain D that contains the
origin and is bounded by a smooth closed surface S. Let v =
l / r , where r = ( x
2
+ yZ + z
2
Show that
16. Evaluate the surface integral
when u(x, y, z) = x
2
- y2 + z + 1 and S: x2 + + ~2 = 4.
17. Show that
for any domain D that is bounded by a smooth surface S.
18. Let R = x i + y j + z k , r = I R I and n a positive integer.
Derive from the divergence theorem the following integrals:
(a) / L r " R . n d ~ = / l / ~ ( n + i ) r ~ d ~
Hint: For part (b) set F = A r
n
, where A is an arbitrary con-
stant vector, and note that jjs Ar
n
. n dS = A- jj5 r
n
n dS
and jjj I, div(Arn)dv = A* jjj DVrn dV.
19. Let f and G be continuously differentiable scalar and vector
fields in the whole space, respectively. Using the divergence
theorem, prove that
Chapter 4
Hint: In part (a), set F = A x G, where A is an arbitrary con-
stant vector and note that
In part (b), set F = f A.
4.11 STOKES' THEOREM
We recall that Green's theorem in the plane can be written in
the form
where F(x, y) = P(x, y)i + Q(x, y)j and D is a plane domain bounded
by a simple closed curve C. The extension of Green's theorem to
three dimensional vector fields leads to what is known as Stokes'
theorem. We establish this theorem in this section and consider
some of its applications.
Theorem 1. (Stokes' theorem) Let S be a piecewise smooth
orientable surface bounded by a piecewise smooth simple closed
curve C. If F(x, y, z) = P(x, y, z)i + Q(x, y, z)j + R(x, y, z)k is contin-
uously differentiable in a domain containing S and C, then
1 / D ( c u r l ~ ) n d ~ d y = I,: : F d R
where the integration along C is in the positive direction relative to
the unit normal vector n on S (Fig. 4.28).
In terms of the components, equation (4.59) can be written as
Integral Calculus of Fields
where n = cos a i + cos p j + cos y k is the unit normal vector on S.
Hence, to prove the theorem it suffices to show that
since the sum of these integrals yields (4.60). We prove only the first
equality in (4.61) since the other two can be established in exactly the
same way.
Fig. 4.28 Stokes's theorem
298 Chapter 4
Let the surface S be represented parametrically by the vector
equation
where the function x(u,v), y(u,v), z(u,v) are assumed to be twice con-
tinuously differentiable in a domain D*of the uv-plane (see Fig.
4.28). We assume that the orientation of the boundary C of the sur-
face corresponds to the positive direction of the boundary C* of D*
through equation (4.62), and that the positive unit normal vector n
on the surface relative to the orientation of the boundary C is given
by
Since
it follows that
Hence we have
where we have noted that dS = I R, x Rv I du dv.
Expanding the integrand on the right-hand side of (4.63), we
obtain
Integral Calculus of Fields
If we add and subtract the term (3P/dx)(3~/3u)(dx/3v) on the right-
hand side of (4.64), and note that
and
we see that
Therefore, by Green's theorem, the right-hand side of (4.63) gives
Now as (u, v) ranges on C*, the p i n t [x(u, v), y(u, v), z(u, v)] ranges
on C. Moreover, since x= x(u, v), we have
ax ax
dx = du + dv
au av
Hence we obtain
Chapter 4
which proves the first identity in (4.61).
We recall that I , F* dR is the line integral along C of the tan-
gential component of F on C, while IIs F- n dS is the surface integral
over S of the normal component of F on S. Thus Stokes' theorem
says that the integral of the normal component of curl F on an
orienentable surface S is equal to the line integral of the tangential
component of F along the boundary of S. In fluid dynamics when F
is a velocity field, lc F* dR represents the circulation of the fluid
around the closed curve C.
Example 1. Verify Stokes' theorem for F(x, y, z) = y2i + xy j +
xzk, where S: z = a
2
- x
2
- y2, z 2 0.
Solution: We choose the unit normal vector on S to be one
with positive z-component so that the orientation of the boundary
C: x
2
+ y2 = a2 (z = 0) of S is counterclockwise. Then we have
so that dS = sec y dx dy = [I + 4(x
2
+ y2 )] dx dy. Since curl F =
- zj - yk, we have
j I (curl F ) n
= -
2 2 2
a [2y(a - x - y ) + yldx dy
where D is the circular domain of radius a. Introducing polar coor-
dinates, we obtain
2 2
Iff [ [2r sin 9 (a - r ) + r sin 91 r dr dB = 0
On the other hand, evaluating the line integral of F on the
boundary x
2
+ y2 = a
2
(z = 0), we find
Integral Calculus of Fields
2K
2
= a x (- sin 0 + 2cos 0 sin 0) dB = 0
thus verifying the theorem.
Example 2. By Stokes' theorem, evaluate jjS (curl F)* ndS,
where
F (x, y, z) = (1 - z)yi + z exj + (x sin z)k
and
S: z = ( a2- x2 - y2
Solution: Taking the unit normal vector n on S to be outward,
we find
= I
(1 - z)y dx + ze
X
dy + x sin z dz
where C: x
2
+ y2 = a
2
is the boundary of S oriented counterclock-
wise. Since z = 0 on C, the line integral on C reduces to
& 2
$ c ~ - d ~ = $ c y d x =- a x sin 0d0
4.12 SOME APPLICATIONS OF STOKES' THEOREM
In Theorem 2 of Sec. 4.4 we proved that if curl F = 0 in a sim-
ply connected domain D, then the line integral of F is independent
of path in D. This fact now readily follows from Stokes' theorem.
Indeed, if curl F = 0 in a simply connected domain D, then for any
piecewise smooth simple closed curve C in D, application of Stokes'
theorem gives
302 Chapter 4
Here S is any piecewise smooth surface bounded by C. Thus the line
integral of F vanishes on every closed curve in D, and this implies
independence of path for I F* dR in D.
We have seen that the divergence theorem can be used to give
an alternative definition of the divergence of a vector field. Like-
wise, Stokes' theorem provides us an alternative definition of the
curl of a vector field F. To see this, let S, denote the circular disk
bounded by the circle C, of radius r and center at a point P. By the
mean value theorem for integrals, there exists a point Q in S, such
that
A(r) curl F(Q).n = I Is, (curl F ) n
Here n is a unit normal vector on S, and A(r) is the area of S, (see
Fig. 4.29). By Stokes' theorem, we obtain
curl F(Q).n =
Now as we let r tend to zero, the point Q approaches the point P,
and so in the limit we obtain
curl F(Q).n =
This defines the component of curl F at the point P on the surface
Sr in the direction of the normal vector n regardless of the coor-
Fig. 4.29 Curl of a vector field.
Integral Calculus of Fields 303
dinate system used. However, if we let n be the unit vectors i , j, k
successively, then (4.67) will give the three components of curl F in
the rectangular cartesian coordinate system.
As in the discussion of divergence, let us consider the flux den-
sity F = pV of a fluid flow with velocity field V and density p. We
observe that lC F. dR is the line integral of the tangential component
of F on C. Its value is called the circulation of F around C. Thus
equation (4.67) says that the component of curl F along the unit vec-
tor n is equal to the circulation per unit area of F around the bound-
ary of a surface that is perpendicular to n. If the circulation is zero
at every point, then curl F = 0, which means that the flow is irrota-
tional. Therefore, we call n vector field F for which curl F = 0 an
irrotational vector field.
It follows from (4.67) that the value of curl F(P). n will be
maximum when n is in the same direction as curl F. This means that
the circulation motion of the fluid is greatest on a surface that is
perpendicular to curl F. In fluid dynamics, curl F is called the vor-
ticity vector.
Example 1. Consider a fluid with constant density p that ro-
tates around the z-axis with angular speed o). Then the angular velo-
city w is represented by w = o)k and the velocity is given by V = w x
R = o( - yi + xj). Thus the flux density is F = pw(- yi + xj), for which
we find
curl F = 2pwk
and, therefore
curl F . k = 2po
On the other hand, the circulation of F around the circle x
2
+ y2
= r
2
is given by
6"
2
p o ( - y d x + x d y ) = por d0=2npor
Dividing this by the area of the circle, we find
2npo r2/ n r
2
= curl F k = 2po
in agreement with (4.67).
Lastly, we consider the application of Stokes' theorem in the
study of electromagnetism. A basic law of electromagnetic theory is
that if E and H represent the electric and magnetic fields at time t,
Chapter 4
then curl E = - 3H/3t. Hence, by Stokes' theorem,
I ; E. ~ R = ( curl Een d~ = -1 Q ;ifen ds
S
where C is a closed circuit and S is any smooth surface bounded
by C. This equation is known as Furudu!yls Iuru. The integral jc E
dR represents the electromotive force (voltage) around the circuit
C, while & H n dS represents the magnetic flux across the surface
S. Thus Faraday's law says that the electromotive force around a
closed circuit is equal to the negative of the rate of change of the
magnetic flux through the circuit.
It is interesting to note that the curl can also be defined as
curl F (P) = lim - J I s : n x ~ ) d ~
,--+ 0 V(r)
where V(r) denotes the volume of the sphere Sr of radius r. This
formula can be established by the use of the divergence theorem.
Note that the integrand in (4.69) is vector-valued and so the integral
must be considered componentwise.
4.9 EXERCISES
In each of Problems 1 through 4, evaluate the surface integral
jjS curl F* n dS by using Stokes' theorem, and verify the result by
integrating the integral directly. The unit normal vector n has a non-
negative z-component.
1.
F(x, y, z) = xzi - yj + xyk, where S is the triangle with vertices
(LO, O), (O,l, O), (O,O, 1).
2. F(x, y, z) = yi + zj + xk, where S is the ortion of the plane
Y
y + z = 2 intercepted by the cylinder x + y2 = 4.
3.
F(x, y, z) = zi - xzj + k, where S is the portion of the parabo-
2-
loid x
2
+ y - z for 0 5 z 5 1.
Integral Calculus of Fields
F(x, y, z) = yzi - xz' + xyk, where S is the hemisphere
z = (
- x2 - y2 ) I / &
Evaluate % ydx + zdy + dz, where C is the intersection of
the cylinder x
2
+ y2 = 2ax and the plane x = z oriented counter-
clockwise as viewed from a oint high on the positive z-axis.
Y
Evaluate Ic (y2 + z
2
)dx + (x + z
2
)dy + ( x
2
+ y2)dz, where C is
the triangle with vertices (1, 1, O), (0, 1, I ) , (1, 0, 1) oriented
counterclockwise as viewed from the point (1,1, 1).
Evaluate k- y2 dx + xydy + xzdz, where C is the intersection of
-
the hemisphere x
2
+ y2 + z
2
= 4x (z > 0) and the cylinder x
2
+
y2 = 2x oriented counterclockwise as viewed from a point high
on the positive z-axis.
Evaluate lc xzdx + yzdy + y2dz, where C is the intersection of
-
the surface z = x y and the cylinder x
2
+ y2 = a 2 oriented
counterclockwise as viewed from the point (0, 0,2a).
Evaluate IC (bz - cy) dx + (cx - az) dy + (ay - bx) dz, where C is
the intersection of the cylinder x
2
+ y2 = R~ and the plane ax +
by + cz = d oriented counterclockwise as viewed from a point
high on the positive z-axis.
Show that
where u and v are differentiable functions of x, y and D is a
simply connected domain bounded by a smooth closed curve
C.
By Stokes' theorem, show that
where C is a smooth simple closed curve bounding a smooth
surface S. Hint: Let F = fA, where A is an arbitrary constant
vector.
Show that
R x k
ficdrR = i fi > j f;
= / I,, 3 dx dy
r
for any simple closed curve C that does not enclose the origin
on the xy-plane.
306 Chapter 4
13. This problem shows how the Stokes' theorem may be
extended to a surface with a finite number of holes. Let
and let S be the portion of the hemisphere z = (a
2
- x
2
- y2
without the piece cut off by the cylinder x
2
+ y2 = e2, where E
is sufficiently small ( E < a).
(a) Show that curl F = 0 for all points (x, y, z) except those on
the z-axis.
(b) Show that curl FmdS = 0, and thus deduce that jC F0dR
= jr F0dR where C: x
2
+ y2 = a
2
and r is the intersection of the
hemisphere and the cylinder.
(c) Evaluate the integral IC F. dR.
TENSORS IN RECTANGULAR
CARTESIAN COORDINATE SYSTEMS
5.1 INTRODUCTION
Up to this stage, we have been concerned with vector analysis.
Among other things, we have studied the algebra of vectors, the
various field properties - gradient, divergence, and curl - together
with their representations in orthogonal curvilinear coordinate sys-
tems, and the various integral theorems involving the divergence
and the curl. Having learned this material, a student might now
assume that all quantities in physics and applied mathematics can
be described in terms of scalars and vectors. Unfortunately, this is
not the case. There are many physical quantities of more complex in
nature that cannot be described or represented by scalars or vectors.
Examples are the stress at a point of a solid body due to internal
forces, the deformation of an arbitrary element of volume of an elas-
tic body, and even the simple concept of moments of inertia. These
quantities can be described and represented adequately only by the
more sophisticated mathematical entities called tensors. As we shall
see later, scalars and vectors are actually special cases of tensors. In
this and the last chapters, we shall study the principal properties of
tensors and consider some of their elementary applications.
The concept of a tensor, like that of a vector, is an invariant
concept. By this we mean that it is an entity that does not depend on
any frame of reference or coordinate system. However, just as a vec-
tor can be represented by its so-called components when referred to
a particular coordinate system, a tensor can also be so represented.
The components of a tensor may be constants or functions. Whether
308 Chapter 5
a given set of constants and/or functions really represent a tensor
depends on the manner by which the members of the set transform
under a change of coordinate system. The situation is identical with
that of vectors. Relative to a given rectangular cartesian coordinate
system, we know that a vector is uniquely determined by its compo-
nents. With respect to a different coordinate system, the same vector
will have different components, and the new components are related
to the old components in a definite manner depending on the trans-
formation of the coordinate systems. Thus the key property of a ten-
sor is the manner by which its components transform under a change
of coordinate system.
Tensors that are represented in rectangular cartesian coordi-
nate system are commonly called cartesian tensors. In the study of
such tensors, one is interested in the way the components transform
from one rectangular cartesian coordinate system to another. There-
fore, the kinds of coordinate transformations involved in cartesian
tensors are transformations of rectangular coordinate systems. On
the other hand, tensors that are represented in general curvilinear
coordinate systems are called general tensors or simply tensors.
Clearly, since cartesian coordinate systems are special cases of gen-
eral coordinate systems, cartesian tensors are special cases of general
tensors.
In this book, we shall present separate discussion of cartesian
and general tensors. Our reason for doing this is twofold. First, a
student who is interested only in learning cartesian tensors can do
so without having to sieve out these results from a mass of infor-
mation pertaining to general tensors. Second, most students find
the transition from vectors to general tensors rather abrupt and thus
encounter considerable difficulty. By first studying cartesian ten-
sors, it is hoped that the student will have acquired sufficient under-
standing, appreciation and familiarity with some aspects of tensors
so that by the time the student studies general tensors, he or she will
find the transition quite natural.
5.2 NOTATION AND SUMMATION CONVENTION
It is sometimes said that tensor analysis is a study of (tensorial)
notations. There is a certain degree of truth in this statement. Be-
cause tensors are complex entities that can have a large number of
components, it is necessary, indeed imperative, to introduce certain
notations and convention to facilitate the representation and mani-
pulation of tensors. This section is designed to introduce the student
to these notations and convention.
Cartesian Tensors 309
Instead of the usual notation x, y, z for rectangular cartesian
coordinates, we now adopt the notation xl, x2 , x3 or briefly xi, with
the understanding that the subscript or index i assumes the values 1,
2, 3. Likewise, we denote the orthonormal basis i , j , k , by il , i 2, i3
respectively, or simply by ii for brevity. With this notation we can
now write the position vector of a point with coordinates xi in the
form
using the summation sign Z. Likewise the analytic representation of
a vector A with components ai can be written as
To achieve further simplification in our notation, we now drop
the summation sign and adopt the so-called Einstein summation
convention. The convention stipulates that whenever an index or a
subscript is repeated in a term, the term must be summed with
respect to that index for all admissible values of the index. In our
case, since we are dealing only with three dimensional space, the
admissible values of an index are 1, 2, 3. Thus, for example, the
summations (5.1) and (5.2) will be written simply as
and
where, in both cases, summation from 1 to 3 with respect to the
repeated index i is implied. Other examples are given below.
Example 1. The term aii implies the sum al l + +
Example 2. The term aibi implies the sum albl + a2b2 + a3b3.
Example 3. The expression aibj - ajbi does not imply any sum-
mation because none of the indices is repeated in the same term.
Chapter 5
The expression stands for the nine quantities
c.. = a.b, - a.b.
(i, j = 1, 2, 3)
!I I J J I
Example 4. The term aijbjk implies summation with respect to
the repeated index j from 1 to 3. Thus,
Example 5. Let @(xl , x2 , x? ) be a scalar field. Then
Example 6. Let @ (x, , x2, x? ) be a scalar field, where xi = xi(t)
I
i = l , 2, 3. Then
dQ(xi) - 3 @d x i
-
dt axi dt
5.1 EXERCISES
1. Write out in full each of the following, where tii, is the Kro-
necker delta.
(a) 8. . a- , (b) 6 a b , (c) a- x x , (d) aik bk, = Fij ,
1J J ?I 1 J 11 1 J
(e) s = a, . b. . ,
'J 1J
( f ) gij = (axi /auk)(ax, /auk)
2. Let @ be a differentiable function of xi (i = 1, 2, 3). If xi = fi(t),
where fi(t) has continuous derivative in (a, b), find d@ / dt and
express it without using the summation symbol.
Cartesian Tensors 311
3. Let $ be a twice continuously differentiable function of xi in a
domain D, and suppose xi = gi(ul , u2, u3 ) where gi is twice
continuously differentiable function in D* for i = 1, 2, 3. Find
a$ /hi and a2$ /auiauj using the summation convention.
4. Let $ be a function of xi and suppose xi = ail x; (i = 1, 2, 3) .
Find a$ /axi for i = l , 2, 3.
5.3 TRANSFORMATIONS OF RECTANGULAR CARTESIAN
COORDINATE SYSTEMS
In this section we study transformations of rectangular carte-
sian coordinate systems. Let xi (referred to as the old system) and x*i
(referred to as the new system) be two rectangular cartesian coor-
dinate systems with the corresponding base vectors ii and i*i (1 s i 5
3). For convenience we assume that the two coordinate systems are
both right-handed and that they have a common origin (see Fig.5.1).
Consider a point P whose coordinates with respect to the old and
the new coordinate systems are xi and x*,, repectively. Then R = xi ii
is the position vector of the point with respect to the old system, and
R* = x * ~ iei the position vector with respect to the new system. Since
Fig. 5.1 Transformation of cartesian coordinate systems.
Chapter 5
R = R" (see Fig. 5.1), we have
Notice that we have used different letters for the indices to indicate
their being independent of each other. Now to obtain the relation-
ship between the coordinates xi and x*~, we may solve (5.3) for x * ~ in
terms of xi or solve for xi in terms of x*~. To solve for x* in terms of
xi , we take the dot product of both sides of equation (5.3) with i*k
(1 1 k < 3). Since
we find
or
x*. = a, . x.
I ?1 J (i = 1, 2, 3) (5.4)
where we define
%.i = i*.. i .
1 J
(i, j = 1, 2, 3) (5.5)
The symbol 6ii/ which has the value 1 whenever the subscripts are
identical andzero value otherwise, is called the Kronecker delta.
Equation (5.4) defines the transformation from the rectangular
cartesian coordinate system xi to the rectangular cartesian coordi-
nate system x*~. Since the coordinate systems are assumed to be both
right-handed, we observe that by rotation the axes of the new coor-
dinate system can be brought to coincide with the axes of the old
coordinate system such that the xi - axis coincides with the x*, - axis
for i = 1, 2, 3. See Fig. 5.2(a). For this reason, the transformation (5.4)
is said to define a rotutiorl of coordinate axes.
If the transformation (5.4) carries a right-handed system into a
left-handed system (or vice versa), the transformation will consist of
a rotation and a rrf l rst i on. That is, when the corresponding coordi-
nate axes are made to coincide with each other by rotation, one of
the axes in one system will be opposite to its corresponding axis in
the other system, and hence it has to be reflected. For example, the
transforma tion
x * ~ = X2, x * ~ =- XI , x * ~ =-XO
defines a left-handed coordinate system x * ~ (assuming xi to be right-
Cartesian Tensors
(a) Rotation of axes
(b) Rotation about x3 - axis followed by reflection.
Fig. 5.2 Rotation and reflection of coordinate axes.
handed). The corresponding axes can be made to coincide with each
other by rotation about the x3-axis followed by reflection of the x3- or
314 Chapter 5
x * ~ - axis, as shown in Fig. 5.2(b).
From (5.5) and the geometric property of dot product, it fol-
lows that
a,j = i*i. i - = I i*i I I i I cos 0 = cos 0. .
(i, j = 1, 2, 3)
.I J 1J '.I
so that the quantity a, is the cosine of the angle 0,, between the xri -
axis and the x. - axis for if j = 1, 2,3.
J
On the other hand, if we take the dot product of ik with both
sides of (5.3), noting that ii . i = 6ik, we find
which, by (5.5), yields
x. = a . . x*.
(i = 1, 2, 3)
1 J 1 J (5.6)
This is the inverse of the coordinate transformation (5.4).
The relationship between the quantities ni, ( = i*i. i.i) i n (5.4)
and (rji (= i*,. ii) in (5.6) is easily seen when we substitute (5.6) for x
in (5.4) to obtain the identity
which implies that
Oc1.i "kj = 'ik
Similarly, if we substitute (5.4) for x * ~ in (5.6), we obtain
which implies
Ti "jk = 'ik
Matrix Representations of Coordinate Transformations
We can understand better the meaning and significance of
equations (5.4), (5.6), (5.7) and (5.8) if we express them i n matrix form.
First we observe that the coefficients of the transformation (5.4) can
be represented as a 3 x 3 matrix
Cartesian Tensors
called the f rurrsforrlzut ion wut ri x . Notice that the first subscrip t of
an element aii in (5.9) indicates the row and the second subscript the
column of the matrix in which the element is located. For example,
the subscripts of the element a2, indicate second row and third
column, the position in which the element is located. It is conve-
nient to regard the rows in (5.9) as row vectors and the columns as
column vectors. Thus, by (5.7) we see that the rows of the matrix
(5.9) forms an orthonormal set of row vectors. Similarly, by (5.8) the
columns form an orthonormal set of column vectors.
Likewise, the matrix corresponding to the inverse transforma-
tion (5.6) is given in (5.10). Notice that the rows and columns of this
matrix are precisely the columns and rows of the matrix (5.9). The
matrix (5.10) is called the transpose of (5.9). and is denoted by A ~ . In
general, the transpose of a matrix is the matrix obtained when the
rows are arranged as columns or the columns arranged as rows.
Thus
To be able to represent the transformations (5.4), (5.6) in matrix
form, we need to introduce the concept of matrix multiplication.
Let B = (b,,) and C = (c,.,) be two 3 x 3 matrices. The product BC of
the two matrices is defined as the 3 x 3 matrix D = (diL,), whose ele-
ments are given by
d . = b c . + biz c2
+ bi3 c, , = b, c,,
(if j = 1,2/ 3)
!I 11 1.1
Notice that dii is the element in the i-th row and j-th c o l u m~ ~ of the
matrix D. It is simply the scalar product of the i-th row vector of B
and the j-th column vector of C. Thus the i-th row of D (1 2 i 5 3)
316 Chapter 5
consists of the scalar products of the i-th row vector of B and the co-
lumn vectors of C.
For exam y le, if
1 0 2 0 2 1
B =I2 1 -11 and C = 1 1 2 [
0 3 0 3 1 -1
then
and
Notice that BC # CB, which shows that matrix multiplication is not
commutative. Also, if X is a matrix consisting only of one column (a
column matrix) with elements xl , x2, x3, then
It follows from the definition that as long as the row vectors of
B and the column vectors of C belong to the same vector space, the
product BC is always defined regardless of the number of rows in
and the number of columns in C. Indeed, if the matrix B has I H rows
and the matrix C has 12 columns, then the product BC will have 111
rows and vr columns - an x 11 matrix.
Using the definition, i t can be easily shown that matrix multi-
plication obeys the familiar associative and distributive laws of
ordnary multiplication o f numbers provided the products are
defined. Thus, if A, B, and C are 3 x 3 matrices, then
(a) A(BC) = (AB)C
(b) A(B + C) = AB + AC (5.11)
(c)
(B + C)A = BA + CA
Now let X denote the column matrix whose elements are xl,
x2, x3, and similarly for X*. Then the transformation (5.4) can be
written in the matrix form
Cartesian Tensors
Notice that the matrix multiplication AX entails taking the scalar
product of the row vectors of A with the column vector X, resulting
in a column vector X*. Similarly, the inverse transformation (5.6) can
be written in the form
where AT is the transpose of A. If we substitute (5.12) for X* in
(5.13), or if we substitute (5.13) for X in (5.12), we obtain the result
where I is the identity matrix
Equation (5.14) expresses the properties (5.7) and (5.8) in matrix
form. It means that AT is the inverse of the matrix A, or A is the
inverse of the matrix A
T
. If we denote the inverse of the matrix A by
A-l, then we have
A-' = AT and (AT)-' = A
A matrix whose transpose is its own inverse is called an ort t l o-
(yonu1 matrix. It follows that the matrix of a transformation that
represents pure rotation is an orthogonal matrix.
From the above results we see that the rows of the matrix (5.9)
are precisely the components of the unit vectors i*i ( i = 1, 2, 3 ) with
respect to the base vectors ii ( i = 1, 2, 3), that is,
i*. = a i.
(i = 1, 2, 3)
'J .I
(5.15)
Likewise, the columns of A are the components of the unit vectors
ii (i = 1, 2, 3) with respect to the base vectors i*i (i = 1, 2, 3), that is,
Notice that equations (5.15) and (5.16) are analogous to equations
Chapter 5
(5.4) and (5.6) which define the coordinate transformations.
Finally, since we assume that both coordinate systems are
right-handed, it follows from the definition of scalar triple product
of vectors that
i*, i*, - x i*, = det A = 1
and
. .
1 , . i 2 x i 3 = d e t ~ ' = l
On the other hand, if the t ra~~sformat i on (5.12) defines a left-handed
coordinate system x*~, then
We summarize our results as a theorem.
Theorem 1. Let (5.12) represents a transformation of rectan-
gular cartesian coordinates. Then A is an orthogonal matrix such
that det A = 1 if the transformation preserves orientation of the axes
(pure rotation), or det A = -1 if the trar~sformation reverses the
orientation (rotation and/or reflection).
Example 1. The unit base vectors i*i of a new coordinate
system x * ~ are given by
Find the equations (5.4) and (5.6) of the coordinate transformation.
Solution: In view of the resemblance of (5.4) and (5.15), we im-
mediately deduce the transformation
Thus the matrix of the transformation is given by
Cartesian Tensors 319
Since iXl. i*2 x i*s = det A = 1, the new coordinate system is right-
handed. The inverse transformation is given by
5.2 EXERCISES
In the following problems, the xi coordinate system is assumed
to be right-handed.
1. What is the equation of the transformation in which each point
(x,, xz , xS ) assumes the new coordinates ( x3, - x,, x2 )? Is the
new coordinate system right-handed or left-handed?
2. The new coordinate system x * ~ is obtained by rotating the xi
coordinate system about the x3-axis through an angle 8 (coun-
terclockwise). What is the equation of the transforma tion?
3. The new coordinate system x * ~ is obtained by rotating the x
i
-
axes so that the positive xl, x2, x, axes are in the same direction
as the positive xX3, xY,, x * ~ axes, respectively. Find the equa-
tion of the transformation and determine whether the x * ~ coor-
dinate system is right-handed or left-handed.
4. The new base vectors of a xXi coordinate system are given by
Find the equations of the coordinate transformation and deter-
mine whether the new coor di ~~at e system is right-handed or
left-handed.
5. Repeat Problem 4 when
iX3 = (i, + 2 i2 - i? ) / d 6
Chapter 5
6. Show that the transformation xei = a,, x,, where
represents a rotation of the coordinate axes. Is the new coordi-
nate system right-handed or left-handed?
7. Suppose that the x * ~ coordinate system is also right-handed.
Then it is true that i*, = i'; x i *k, where i, j, k are cyclic permu-
tations of 1, 2,3. By using the relations (5.15) and (5.16), obtain
the relations
a,r = a. a - a. a
JP kq Jq kp
where i, j, k and p, q, r are cyclic permutations of 1, 2, 3.
8. Obtain the correspondi~~g relations as in Problem 7 if the x*
coordinate system is left-handed.
5.4 TRANSFORMATION LAW FOR VECTORS
As a first step toward the definition of cartesian tensors, let us
examine how the components of a vector transform under a change
of coordinate system. To avoid unnecessary repeti tion of termino-
logies, let us agree that whenever we speak of a coordinate trans-
formation in this chapter, we mean a transformation from one rectan-
gular cartesian coordinate system into another rectangular cartesian
coordinate system. Thus, the equations of our coordinate transfor-
mations are precisely those derived in Sec. 5.3, namely,
where a - = i*.. i. (i, j = 1, 2,3) satisfy the relation
?I ' J
Now consider a vector A and sup pose that its components
with respect to the coordinate systems xi and x * ~ are ai and a*i, res-
Cartesian Tensors
pectively. Then we have
To determine the relationship between the components a.i and axi,
we take the dot product of i"k with both sides of (5.17) to obtain
a*. 6. = (i*k. i. )a
I ~k J .I
which leads to
ax*,
a*, = a. . a . =
IJ J a x . a ~ (5.18)
J
On the other hand, if we take the dot product of ik with both sides of
(5.17), we obtain
dx
a . = a . . a . = a*.
1 1 1 .I ax*. J
.I
Equation (5.18) or its inverse (5.19) demonstrates how the compo-
nents of a vector transform when we change coordinate system. We
call this the transformation law for vectors. Thus, in terms of the
transformation law (5.18) or (5.19), we may redefine a vector as
follows:
Definition 1. A vector is a quantity consisting of three compo-
nents a, ( i = 1, 2, 3) which transform under a change of coordi-
nates according to the law (5.18) or (5.19), where ai and a*i are the
components of the vector with respect to the coordinate systems xi
and x * ~ , respectively.
This definition casts vectors in the setting of cartesian tensors.
In fact, Definition 1 is precisely the definition of a cartesian tensor of
order 1 (or rank 1). In the context of this definition, a scalar becomes
a quantity that remains invariant (does not change in numerical
value) under any coordinate transformation. In tensor language,
scalars are tensors of order (or rank) zero.
Example 1. Show that the Euclidean distance between t wo
points in space is invariant (a scalar) with respect to any coordinate
transfor ma tion.
Chapter 5
Solution: The Euclidean distance between any two points in
space with the coordinates xi and yi is given by
d2 = (xi - yi )(xi - yi ) = (xi - yi )2
(summation on i)
Under a tranformation of coordinates, the new coordinates of the
points are given by
x*. = a.. x.
I lj J and y'i = aik Yk
Hence, in the new coordinate system, we have
Since
a. . ai k = 6. = {
1 when j = k
1.1 J k 0 when j t k
it follows that
Example 2. Show that the dot product o f two vectors is
invariant under any coordinate trans for ma tion.
Solution: Let A = ai ii and B = b., i.i be two vectors. Then
For this to be numerically the same in any other coordinate system,
we need to show that aibi = a*,b*.i, where a*, ad b*, are the compo-
nents of the vectors in the new coordinate system x*, defined by the
transformation x * ~ = a,, x.~. Now, according to the transformation law
for vectors, we have
a*, = aip ap,
b*i = aiq bq
Hence,
a*ib*i = a. a. a b = 6 a b = a b
' P ' 4 P 4 Pq P q I ' P
Cartesian Tensors
Example 3. Given the transformation of coordinates
x*, = ai, x, , where
If a vector A has the components [I, 2, -11 with respect to the xi-coor-
dinate system, find its components in the x * ~ - system.
Solution: According to the transformation law for vectors, we
have
Notice that by definition the elements in the rows of the Ina-
trix (al-, ) are the components of the unit vectors i*i (i = 1, 2, 3). Thus
the con~ponents a*i are just the dot product of the vector A with the
base vectors i*i (i = 1, 2,:i).
Example 4. Let ai and bi be the components of two nonzero
vectors A and B , respecti\~ely, in a rectangular car t esi a~~ coordinate
system xi, and set A s I3 = ci i, , where ci = a.b - akb, , i, j, k being the
J k
cyclic permutations of I., 2 . 3 Let a*, , b*i , and cqi be the respective
components of the vectors A , B , and A x B in the new coordinate
system x*, defined by x * ~ = u . x-, where
1.1 I
324 Chapter 5
Show that = - (a*jb*k - a*kb*.i), where i, j, k are cyclic permuta-
tions of 1, 2, 3.
Solution: Under the given coordinate transformation, the
new components of the vectors A , B , A x B are given by
aql = a l j a, = a,, = azi ai = - a l ,
a*, = a,, a, = a2
. .
b*, = a l j b, = b, , be2 = a 2, bj = - b, , b*3 = a 3, bj = b2
cx1 = a 11. c. - - c3 = alb2 - a2bl , c * ~ = a 21 C, = - c1 = - (a2b3 - a,b2)
c * ~ = a . c - ~2 = 9 bl - a1 b3
,J .I-
Thus we see that
This shows that the transform of the components ci = ajbk - akb.i
differ from a>bxk - a*kb*j by the factor (-1).
Notice that in Example 4, det (a*. ) = -1 which means that the
' 1
new coordinate system x * ~ is left-handed (as we assume that the
coordinate system xi is right-handed). This example illustrates the
fact that, in general, the components of the cross product of two
vectors do not transform according to (5.18) or (5.1 9). Therefore, the
cross product of two vectors is not a vector in the sense of Definition
1. It is called a pseudovector or an axial vector.
5.3 EXERCISES
1.
Consider the trajectory of a moving particle xi = xi(t) (t 2 O) ,
where t denotes time. Define
Show that vi (i = 1, 2, 3) are component s of a vector (t he
velocity vector).
Cartesian Tensors 325
2. Let a transformation from a coordinate system xi to a coordi-
nate system x * ~ be defined by x * ~ = a - x , where
!I J
(a) If A = - i, + 2i2 - ?i3 , find its components in the x * ~ coor-
dina te system.
(b) If B* = 2i*, + 2i
X
2
- 3i*3, find its components in the xi coor-
dinate system.
(c) Find the scalar product of the two vectors A and B*. (Does
it matter which components you used in taking the scalar
product?)
(d) Determine the vector product of A and B* in both coordi-
nate systems and show that the components are related by
the transformation law c*i = a,, c.,, where ci denote the com-
ponents in the x
i
-system and c'i the components in the xXi-
system.
3. Consider a transformation defined by x*, = a,., xi , where
( a) Show that the t ransformat i on changes a ri ght -handed
coordinate system into a left-handed coordinate system x Xi (or
vice versa).
(b) Find the compone~~t s of the vector A = 6il + 12i2 + hi3 in
the x*i - coordinate system.
(c) Find the components of the vector B* = 43 i*, + 42 iX2 +
.ih i*? in the xi - coordinate system.
(d) Using the components ci of A x B in the x
i
-coordinate
system, determine the components of A x B in the x
X
i
- coordi-
nate system by the transformation law
= ail c.;.
Chapter 5
4. Let ai and bi denote the components of the vectors A and B,
respectively.
(a) Show that the components of the vector product A x B can
be written as ci = a.ibk - akb.;, where if j, k are cyclic permuta-
tions of 1, 2, 3 (that is, ijk = 123,231,312).
(b) Under a coordinate transformation x * ~ = ai.i xj, show that
(c) From the fact that iX = w. i and i*k = akq i qf show that
J I P P
( d) Thus if both coordinate systems have the same orientation
(both right-handed or left-handed), show that
where p, q, r are also cyclic permutations of 1, 2, 3. On the
other hand, the coordinate systems have opposite orientation,
show that
i*i = - ( a. a
J P kq - a i q akP
(e) Since i*i = air i, deduce that a. a - a a = * a i r , where
.IP kq 19 kl3
the plus or minus sign prevails according as the transforma-
tion preserves or reverses the orientation of the coordinate sys-
tems.
(f) Finally, from (b) show that c*, = A air c,, where A = det(aii)
= i 1. [A quantity consisting of three components which trans-
form according to the law given in ( f ) is called a pseudotensor
of order one or axial vector.]
Cartesian Tensors
5.5 CARTESIAN TENSORS
111the preceding section we introduced the definition of carte-
sian tensors of order 0 and l , which are essentially scalars and vec-
ors. Wc now consider cartesian tensors of higher order. The order of
a tensor has to do with the number of compor~ents of the tensor. In
three-dimensional space, a tensor is said to be of order or rank r if it
has components. Thus scalars and vectors are of order 0 and 1
since they have 3" = 1 and 3' = 3 components, respectively.
To motivate the definition of tensors of higher order, we consi-
der a simple physical example of a quantity that is described by a
cartesian tensor of order 2. L,et a particle of mass m be located at
the point (xl , x2, x, ) The so-called moment of inertia of the particle
is a set of nine numbers defined by
I = m ( 6 xkxk - x,xj)
(i, j = 1, 2,3)
1.1 1.1
(5.20)
When i = j, this yields
These are called the moments of inertia of the particle about the xl -,
x2 -, X? - axes, respectively, and are the types of moments of inertia
usually st udi ed in a st andard calculus course. When i + j, (5.20)
gives
These are called the p rod uc ts of inertia.
Now let us exanline what happens to the moment of inertia of
the particle when we introduce a different coordinate system. Let
x * ~ = a ,i X, define a new coordinate system and let (x* , , x** , x*? ) be
the new coordinates of the particle. Then, by defintion, the moment
of inertia of the particle in the new coordinate system x*, is given by
I*.. = rn(8*.. x*
1.1
kx*k - x * ~ x * ~ )
(i, j = 1, 2, 3)
!I
(5.21)
Chapter 5
To relate these quantities with the I,, (i, j = 1, 2,3), we first observe
that under the coordinate transformation, we have
(invariance of Euclidean distance), and
that is, = 6,; ( because the Kronecker delta has the same value in
any coordinate system). Substituting these identities in (5.21), we
obtain
Equation (5.22) describes the manner by which the moment of
inertia of a particle transforms under a change of coordinate system.
Quantities that transform according to the formula (5.22) are called
cartesian tensors of order two. This is the classical definition of ten-
sors of order 2 in terms of compor~ents relative to various coordi-
nate systems. Thus the quantities I,, given in (5.20) define a second
order tensor. It is called the moment of inertia tensor, and the quan-
tities Iii are the components of this tensor. The components may be
viewed as constituting a representa tion of the tensor with respect to
a given coordinate system.
We now formally define a second order tensor.
Definition 2. (Tensor of Order Two) A tensor of order two is
a quantity consisting of 32 components a (i, j = 1, 2, 3), which trans-
1.1
form under a change of coordinate system x*, = a,,; x, according to
the law
a*.. = a. a. a
1.1 11' . 19W
(if j = 1, 2, 3)
(5.23)
where a, - and a*.. are the components of the tensor with respect to
1.1 IJ
the coordinate systems xi and x*~, respectively.
Cartesian Tensors 329
The transformation law (5.23) can be inverted if we multiply
both sides of the equation by air a,, and sum with respect to i and j.
We obtain
a,, = a a a* - -
1r .Is 1.1
(5.24)
in which we used the relation aij aik = 6jk.
It is convenient that a s e c o ~ ~ d order tensor can be written in a
matrix form
Because of this, certain algebraic operations i n v o l v i ~~g vectors and
tensors of order two can be calculated by matrix algebra and the
results expressed in matrix form. As a matter of fact, if we set
then the product biq = a a represents the element in the i-th row
IP lxl
and q-th column of the matrix product AA = R, while the product
biqctiq = alpajqaFs represents the element in the i-th row and j-th
column of t he matrix product E3AT = A A A ~ . Therefore, the trans-
formation law (5.23) can be written in the matrix form
Since = A
T
, the inverse transformation law (5.21) has the matrix
form
A = A ~ A " A (5.26)
It should be noted that matrix notation fails for tensors of
higher order.
Example 1. A tri~rial example of a second order tensor is the
Kronecker delta tiij which, by definition, is independent of any coor-
dinate system, that is, 6ij = 6*,,. Hence, we can write
330 Chapter 5
since c t c t = tiij. It is clear that the matrix representation of the
1P JP
Kronecker delta is the identity matrix, that is, (6,, ) = I.
Example 2. Let ai and bj be the components of two vectors.
The nine quantities cij = a, b.; (i, j = 1, 2, 3) form the components of a
second order tensor, known as the outer product of the two vectors.
To prove that (c,,) is indeed a second order tensor, we need to
show that it transforms according to the law (5.23). Now under a
change of coordinates, we know that the componentt. a , , bj transform
according to the laws:
Hence
c*.. = a*. b? = a. a. a b = a a. c
IJ 1 .I IP J9 P P 1P J9 IT
in accordance with (5.23).
Example 3. Given the second order tensor
Find the components of this tensor in the coordinate system x * ~
defined by x * ~ = ai, xi , where
( a ) = -1 0 o
I:: :' d l
Solution: The new components are given by
a*, . = a a. a
(i, j = 1, 2, 3)
1.1 'I' J q I T
Thus
a*ll = a l p al q am = "13 "13 "3 = =I
a*l' 2 = a ~ q a l y = "13 "21 "31 =
"'13 =
=
"13 "32 '32 = -
Cartesian Tensors
Hence
The student should check this result by using the formula (5.25).
Example 4. Let a body of unit mass be located at a point P:
(1, 1, 0). By (5.20) the components of the moment of inertia tensor are
I,, = 1, IZ2 = 1, I,, =
II2 = = -1 I I?:, = I?2 = O f '1.3 I31 = '
or, in matrix form
Now suppose we rotate the coordinate system about the x,-axis so
that the x,-axis passes through the point P. Let us designate this axis
as x*, and the others as x', - and x*, (see Fig. 5.3). Obviously, i *, = ( i ,
+ i 2)/ d?, i*, = (- i1 + i, ) / v2, i', = i, , so that the transformation effec-
ting thtl rotation is defined by the matrix
(Remember a,, = i *i . i ; . ) Thus, with respect to the x * ~ - coordinate
Chapter 5
Fig. 5.3 Principal axes of a tensor.
system, the moment of inertia tensor becomes
as is readily verified by using (5.25). Notice that in the x * ~- coordi-
nate system, the point P has the coordinates (d2, 0,O).
Therefore, in the ~*~- cvor di nat e system all the product inertias
are zero, and the moments of inertia about the the x * ~ -, xX2- and xX3 -
axes are I*11 = 0, I*22 = 2 and I*33 = 2, respectively. The axes x * ~
with respect to which Iij = 0 for all i t j are called the pui nci pul uut7s
of the tensor. This topic is discussed further in Sec. 5.8.
Example 5. Suppose that a certain quantity has components
defined by al l = xlx2 , al = x2 2, a2, = x, 2, a22 = xlx2 with res-
pect to a rectangular cartesian coordinate system xl, x2 on a plane. If
the quantity were a (plane) tensor of order two, its components in
any other rectangular cartesian coordinate system xei would be
Cartesian Tensors
given by
However, under a rotation of axes,
x * ~ = x1 cos 8 + x2 sin 8
x * ~ = - x1 sin 8 + x2 cos 8
we see that
aell = ali al, a,, = x1 x1 cos2 8 + x2 2 sin 8 cos 0
+ x1
sin 8 cos 8 + xl x2 sin
2
8
= xl x2 + (xl 2 + x2 ) sin 8 cos 8
while, on the other hand,
x * ~ x * ~ = ( xl cos 8 + x2 sin 8)( - xl sin 8 + x2 cos 8)
= xl x2 ( cos
2
8 - sin
2
8 ) + ( x2 - xl 2 ) sin 8 cos 8
Thus, a*ll + x * ~ x * ~ , and, therefore, the quantity is not a tensor. In
other words, the quantity given here cannot represent any physical
entity.
The definition of a cartesian tensor of order r (r > 2) is a
straight-forward extension of Definition 2.
Definition 3. A cartesian tensor of order r (r > 2) is a quantity
consisting of 3rcomponents ai I i 2 . . i , which transform under a change
of coordinates according to the law
where a* i l i 2 , , i r and ai
. are the components of the tensor with
1 2 . . . ' ,
respect to the xXi and xi coordinate systems, respectively.
Thus, for example, a tensor of order 3 will have 27 compo-
nents aijk (i, j, k = 1, 2, 3) which transform according to the law
a*.. = a a. a a
11k IF I9 k r W
Chapter 5
under a change of coordinates x * ~ = aij xj , i = 1, 2, 3.
Example 6. Let ai, bi, and ci (i = 1, 2, 3) be the components of
three nonzero vectors. Then the 27 quantities
d - = ai b.i ck
(i, j, k = 1, 2, 3)
1.l k
form the components of a third order tensor known as the outer pro-
duct of the vectors. To see this, we first note that under a change of
coordinates, we have
a*i = ai,pP , b*. = ajqbq, c * ~ = akrc,
.I
Hence
which shows that the quantities dijk indeed transform as components
of a third order tensor.
5.6 STRESS TENSOR
We consider another simple example of a physical quantity
that is described by a second order tensor. When a solid body is
under the influence of a system of forces and remains in equili-
brium, there is a tendency for the shape of the body to be distorted
or deformed. A solid body in such a state is said to be under strain.
The resulting internal forces in a body that is in a state of strain give
rise to a stress at each point of the body. The stress at a point is mea-
sured as force per unit area and depends not only on the force but
also on the orientation of the surface relative to the force. To illus-
trate this, consider a uniform thin cylindrical rod of cross-sectional
area a that is being stretched by a force IF (Fig. 5.4). Consider a plane
section whose unit normal vector n makes an angle 0 with the axis
of the rod. By the area cosine principle, the area of this plane section
i s equal to A = a sec 0. Thus the stress vector across the plane
section is equal to
S = F/A = (F/a) cos 8 (5.28)
Cartesian Tensors 335
It is clear from (5.28) that the stress vector depends on F as well as on
the orientation of the plane, i.e., n or 0. In fact, we see that the mag-
nitude of S is maximum when 0 = 0, and it is zero when 0 = x/2.
The stress vector S on the plane area can be resolved into two
Fig. 5.4 Stress in a thin rod.
components, one perpendicular to the plane section (along the unit
normal vector n) denoted by s,,, and the other parallel to the plane
section (perpendicular to n), denoted by st. The component s, along
n is called the tensile or normal stress, while the component st tan-
gential to the plane section is called the shearing stress.
To examine the stress at a point P in a solid body under strain,
we set up a coordinate system x, , x2, x3 and consider an arbitrary
element of volume (a small rectangular parallelepiped) of the body
with P as one of its vertices (see Fig. 5.5). Denote the length of the
edges of the element by Ax , , Ax2, Ax3 as shown in Fig. 5.5. Clearly
there will be stresses on each face o f the element. Let Si denote the
stress vector on the face that is perpendicular to the vector i i , and let
S
f
i
denote the stress vector o n the opposite face (perpendicular to
i i ) . Since the body is in equilibrium, the sum of the forces (stress
times area) acting on the element must be zero. Thus we have
Since the quantities Axl, AxZ, Ax3 are arbitrary, the equation implies
that S
f
1
= - S,, S'? = - S2, StO = - S3. Hence the stress vectors on oppo-
site faces of the element are equal in magnitude but opposite in
direction. Moreover, since (5.29) holds for any element of the body
Chapter 5
Fig. 5.5 Stress tensor.
at P, we conclude that the stress vectors Si (1 5 i 5 3) represent the
stress at P acting on surface areas that are normal to the unit vectors
ii (1 5 i 5 3), respectively.
Now each of the stress vectors S (1 2 i 2 3) can be represented
by its components sil , si 2, si3 , SO that Si = s . - i (i = 1, 2, 3). The
'J J
quantities sll , s22, s33 are the respective normal components of the
stress vectors S1 , S2, S3 , and they are called the tensile or normal
stresses; the quantities sij for i t j are called the shearing stresses.
Thus, at each point, the stress is represented by the nine quantities sij
(i, j = 1,2,3), which can be written in the matrix form
However, the shearing stresses sij, i t j, are not altogether indepen-
dent of one another. In fact, since the element of volume Axl, Ax2,
Ax3 is also in equilibrium with respect to rotation, the sum of the
moments of force or torques about any axis must be zero. So if we
consider the net torque about the edge through P parallel to the x3 -
axis (see Fig. 5.5), we obtain
Cartesian Tensors
The stresses sl , s22 ,
, s~~ are balanced by equal and op posite
stresses on the opposite faces of the element, and s13, s2,,
have
zero moment arm as they are directed along the x,-axis. Thus we
obtain,
512 = Sz1
In a similar manner, by considering the net torque about the edges
parallel to the xl - and x2 - axes, we find
"2 = =32 S31 = S13
Therefore, s. . = s. . (i, j = 1, 2, 3) which means that the matrix (5.30) is
'J J I
symmetric.
Next we show that the quantities s,, do transform according to
the transformation law
s*. = a. a. s
1.1 IP J9 W
(5.31)
where eii (i, j = 1, 2, 3) represent the stress with respect to a new
coordinate system x*,. First, we show that the components of the
stress vector S, across a surface with unit normal vector n = a,liii are
given by s,,, = Si n = si~,u,, ( i = 1, 2, 3), so that S,, = silii ,. We consider
an infinitesimal tetrahedron with P as a vertex and with the slant face
perpendicular to n as shown in Fig. 5.6. Let AA denote the area of
the slant face. Then, by the area cosine principle, the areas of the
sides with normal vectors i,, i2, and i, are given by cr,,,AA, AA,
and a,; AA, respectively. Now since the element is in equilibrium,
the force on the slant face must be equal to the sum of the forces on
the three faces of the tetrahedron. Thus, we have
S,, AA = S1 AA + S2 a,12 AA + S, a,,, AA
Equating the respective components, we obtain
Chapter 5
Fig. 5.6 Stress vector across an arbitrary surface area.
Since s. . = s. . we may write this in matrix form as
IJ J I '
Now let iei be the base vectors of a new coordinate system x * ~
and let S*i denote the stress vector across a surface that is yerpen-
dicular to i * i , i = 1,2,3. Then, by (5.32a), we have
On the other hand, if seij denote the components of Ssi with respect
to the new coordinate system x * ~ , then, S*i = s*ij i*,. Thus we have
Cartesian Tensors 339
We solve for s*, ~ by taking the dot product of both sides of (5.33)
with i*,,, noting that i'i. i*k = !ik. Since cxiq = i;- iry we finally
obtain
S* . = a. a. s
'.I 'P .I9 P9 (5 34)
Thus the quantities si., transform as components of a second order
tensor. This tensor is known as the stress tensor.
Example 1. The stress tensor at a point P is given by
What is the stress vector on a surface normal to the unit vector
at the point? What is the normal stress and the shearing stress on
such a surface?
Solution: According to (5.32a), the components of the stress
vector S, 011 the surface normal to the vector n are given by
The normal stress, that is, the stress along the direction of n , is given
by
3
L 1 2 I 0
s,, = S,; n = bnl + Sn2- sn3=
3 9
Chapter 5
Thus the shearing stress is
Example 2. For stress on a plane, the compo~~ent s of a stress
tensor are given by
where s12 = Then, by (5.32a), the components s,,, and sl12 of the
stress vector across a plane normal to n = cos 8 il + sin 0 i2 are given
by
slll = sl l cos 0 + s~~ sin 0
sn2 = Sl 2 cos 8 + s~~ sin 8
Thus the normal stress on such a plane is equal to
s = S n = s, , ~ cos 8 + sd sin 8
I1 I1
= sI1 cos
2
0 + 2 Sl2 sin 8 cos 0 + s22 sin
2
8
and the shearing stress is given by
For examvle, if
then the components of the stress across a plane normal to the unit
vector n = (\13/2) i1 + (1 /2) i2 are
Thus the normal stress across such a plane is
Cartesian Tensors
and the shearing stress is given by
Example 3. Consider a stress tensor given by
( s,, [ :]
1 1 0
If we change the coordinate system through the transformation
then in the x*~-coordinate system the stress tensor becomes
This can, of course, be verified by using the transformation law
s*.. = a a. s
'1 'P J9 P9
For instance, we find
Chapter 5
since a12 = 0, s,, = 0, and 33 = 0. Thus in the x * ~ coordinate system,
the stress consists only of normal stresses as all shearing stresses are
zero.
5.4 EXERCISES
1. The equation of a quadric surface with center at the origin of a
coordinate system xi is given by a x x = 1. If the equation of
'J 1 J
the surface under rotation of axes is a*,j x * ~ x*., = 1, show that
the cofficients a,, form components of a second order tensor.
2. Verify the result of Problem 1 for the conic
under the rotation of axes
x* = x cos0 + y sin 0
y" = - x sin 0 + y cos 0
3. A body of mass M is located at the point (2, -2, 3). Find the
components of its moment of inertia tensor.
4. The moment of inertia of a body of mass m about an axis in
the direction of a unit vector u = ui ii is given by
1 = I.. U. U.
11 IJ I J
where I,, is the moment of inertia defined in Sec. 5.5. Show
that I, = m[(xkxk - (xiui)*]. What is the moment of inertia of
the body in Problem 3 about the line xl = x2 (x3 = O)?
5. The components of the moment of inertia of a certain body
with respect to a coordinate system xi are given by
Find the moment of inertia of the body about a line in the
direction of the vector A = il - i2 + i 3.
6. Find the moment of inertia of a body of mass M located at (2,1,
3) about an axis in the direction of the vector A = 2il - i2 + 2i3 .
Cartesian Tensors 343
7. The moment of inertia tensor of a certain body is given by
Show that in the coordinate system x * ~ defined by x * ~ = aijxj,
where
I*,, = - 2 , I*,, = 1, I*,, = 4.
8. A stress tensor has components given by
Find I*,, , I*,, , and I*,, in the coordinate system x * ~ defined
by x " ~ = a . x. , where
'J J
9. Verify that the quantities
are components of a second order plane tensor.
Chapter 5
10. Show that the quantities
are not components of a tensor.
11. The stress tensor at a point has components given by
Find the stress vector across an area normal to the vector A =
i, - i2 + i3. What are the normal and the shearing stresses
across such an area?
12. Find the normal and shearing stresses across an area normal
to A = 2i1 + i2 - 2i, at a point where the stress tensor is defined
by
1 2
13. The stress tensor at a point on the x, x2 -plane has components
given by
Find the normal and the shearing stresses across a plane whose
normal is A = i, + 2 i 2.
14. The components of a stress tensor are given by
What are the components of the tensor in the coordinate
system x * ~ where the positive axis coincides with the posi-
Cartesian Tensors 345
tive x3-axis, and the positive x*~-, axes lie on the xlx2 plane
and make angles of x/3 and -x/6, respectively, with the posi-
tive xl-axis.
15. Let al, a2, a3 be the components of a vector, and define
Show that
under any rotation of axes x * ~ = a- x. such that det ( aij ) = 1.
1J J
16. Let A = (a
ij
) be a matrix whose elements aij are components of a
second order tensor, and let B be a column matrix whose elem-
ents bi are components of a vector. Show that the elements of
the matrix product AB are components of a vector.
17. Write down the transformation law and its inverse for a fourth
order tensor.
18. Show that if the corresponding components of two tensors of
order r are equal in one particular rectangular cartesian coor-
dinate system, then they are equal in all rectangular cartesian
coordinate systems.
5.7 ALGEBRA OF CARTESIAN TENSORS
Tensors possess much the same algebraic properties as vectors
with respect to various operations that combine tensors to produce
other tensors. These operations are defined as follows.
Definition 1. The sum of two tensors of order r is a tensor of
the same order whose components consist of the sum of the corres-
ponding components of the two tensors.
For example, if aii and bl, are the components of two second
order tensors, then their sum is a tensor of second order whose com-
346 Chapter 5
ponents are given by aij + b,,. It is easy to see that cij = a,, + bi, trans-
form according to the law
c*.. = a a. c
1J 1P Jq P9
where c*,, = a*ij + b*ij. In fact, since ail and bij are components of
tensors, we know that
Hence
c*.. = a*,, + b*,, = a. a. (a + bpq) = a. a. c
1J 1P 34 P9 1P 19 P1
Definition 2 (Outer Product.) The outer product (or simply
product) of two tensors of order r and s is a tensor of order r + s
whose components consist of the products of the various compo-
nents of the tensors.
For example, let a,, and bh be the components of a second
order and a third order tensor. Then the outer product of these two
tensors is a tensor of order five whose components consist of cijh =
a.. bh (if j, k, m, n = 1, 2, 3) . The tensorial character of this outer
1J
product follows from those of a,, and b-. In fact, since a*,, = aip
ajqam and b*hn = ahamSa,bst , we have
Definition 3 (Contraction.) The operation of setting two of the
indices in the components of a tensor of order r 2 2 equal and then
summing with respect to the repeated index is called contraction.
For example, consider the components a,, of a second order
tensor. The contraction of this tensor yields the scalar
which is just the trace of the matrix (a,,). For a third order tensor
Cartesian Tensors
with components aijk, there are three possible contractions, namely,
It can be shown (see below) that each of these contractions results in
a vector. In fact, the cc>ntraction of a tensor reduces the order of the
tensor by 2. We state this as a theorem.
Theorem 1 (Reduction of order by contraction.) The contrac-
tion of a tensor of order r > 2 results in a tensor of order r - 2.
We illustrate the proof of Theorem 1 for a third order tensor
aijk, where the contraction is with resyect to the indices i and j. First
we note that under a coordir~ate transformation, we have
aei.jk = a a n a
IP .lq kr pqr
Setting i = j (summation is implied), we find
a*. . =a . a. a a
I I ~ ~p ~q kr pqr
= 6,qab apqr
= a a
kr PPr
since a n = tiF9. Hence if we set c * ~ = a*iik (sum on i) and cr = appr
'P '9
(sum on p), then we have
c * ~ = ah Cr
which shows the tensor character of c,of order one (a vector).
It is evident that contraction of a tensor can be repeated with
respect to the various indices until the tensor reduces to a vector or
a scalar depending on whether the original order of the tensor is odd
or even.
Definition 4 (Inner Product.) An inner product of two tensors
is a contraction of the outer product with respect to two indices,
each belonging to a component of the tensors.
348 Chapter 5
For example, the components of the outer product of two vec-
tors A = ai ii and B = bi ii are given by cij = sib,. By contraction, we
obtain the inner product
which is the dot or scalar product of the two vectors. On the other
hand, for two second order tensors with components a,, and bh,,
there are four possible inner products, namely,
c...
= a,, b,,,, , c-. . = a,, b . (sum on j)
lJJln 1~ k~ k~
By Theorem 1, each of these inner products gives rise to a second
order tensor.
Quotient Rule
Thus far we have seen that the components of a tensor trans-
form according to definite laws under rotation of the coordinate
axes. Now suppose we are given a set of 3' quantities
How do we know that the quantities are components of a tensor
without actually going to the trouble of checking how they trans-
form under a change of coordinates? It turns out that there is a rule,
known as the qzlotit2\z t r ~l l t ' , which enables us to determine whether
or not the given quantities do form a tensor. We state the rule as a
theorem and demonstrate its proof for r = 2.
Theorem 2. If the outer or inner product of the set of quanti-
ties (5.35) and an arbitrary tensor of any order yields a tensor of
appropriate order, then the quantities are components of a tensor of
order r.
Let the given quantities be Xij and suppose that the outer pro-
duct X,, ak with an arbitrary vector ak is a tensor of order 3. Set bijk
= Xi, ak and let X*,, represent the quantities in a new coordinate
Cartesian Tensors 349
system . Since bijk are components of a third order tensor, we have
which implies
X*, = aip ajp XW
Thus the quantities Xi, are components of a second order tensor.
Example 1. Let the components of a second order tensor be
given by
and let the components ai of a vector be 1, -2,2. Find the inner pro-
duct t,aj and aiti,.
Solution: Set bi = ti, a, . Then we have
Next, let cj = ai ti,. Then
c, = ai ti, = l(2) + (-2)(1) + 2(0) = 0
c2 = ai ti2 = 1(-1) + (-2)2 + 2(-3) = -11
c3 = ai ti3 = l(0) + (-2)2 + 2(1) = -2
Example 2. Let si, denote the components of a stress tensor at a
point. Then the stress vector S*, across an area with normal vector
n = ni ii is the inner product of the stress tensor and the vector n,
that is, s*,, = sij n, = sji n, , since s.. = sji.
1J
Chapter 5
Symmetric and Anti-Symmetric Tensors
Many tensors that arise in physical problems possess certain
simplifying properties. For example, we have seen that the compo-
nents s of a stress tensor satisfy the condition sij = sji for i, j = 1, 2,3.
?I
We call such a tensor s t j n l r l ~ i ~ f r i ~ - t i > ~ s ( ~ r . A second order tensor
whose components satisfy the condition air = -aji iscalled an nnt i -
synl nl t 7t ri c t ens or . It is clear that for such a tensor, aii = 0 (no sum
on i) for i = 1, 2, 3. Thus a second order anti-symmetric tensor has
the matrix representation
Such a matrix is called a skew-symmetric matrix.
The definition of symmetric or anti-symmetric tensor can be
extended as well to tensors of higher order. For example, a third
order tensor with components ai i k is said to be sljrllrrrt.tric or m t i -
synr, l rrf ri c with respect to the firit two indices according as aiik = ajik
or a = -ajik. It is said to be symmetric or anti-symmetric with res-
I J ~
pect to the last two indices if
= a. . or a = - a.
I ~ J I J ~ 1k1.
It is important to remember that if a tensor is symmetric or
anti-symmetric in one coordinate system, it remains symmetric or
anti-symmetric in all other coordinate systems. This can, of course,
be verified through the transformation law governing the tensor.
For example, suppose aij = aji in a coordinate system xi. Under a
transformation of coordinates x * ~ = aij xj , we see that
a*.. = aipajq aw = a. a. a
= a. a. a = a*..
11 I P J 9 W J 9 I P 9P J 1
Thus the tensor is also symmetric in the x * ~ coordinate system.
We observe that every tensor of order r 2 2 can be expressed as
a sum of a symmetric and an anti-symmetric tensors of the same
order. For example, for a third order tensor with components aijk,
we set
Cartesian Tensors 351
Then, clearly, aijk = s. . I J ~ + t I J ~ and silk = sjik , t I J ~ = - t ~ l k , so that
s.. and t . . are components of a symmetric and an anti-symmetric
IJ k ?I k
tensors of third order, respectively.
5.5 EXERCISES
1. Let the components of a second order tensor be given by
and let A = 2il - 3i2 + 2io, B = - i, + 2i2 - 2i,. Find the inner
product (a) t . , a , ( h) t a , (c) tijaibj , where ai , b, denote the
1J J 1J I
components of A and B , respectively.
2. Given the components of second order tensors
find the inner product (a) ai,bjk, (b) aikbjk, (c) "jbik,
(dl aijbij-
3.
Let a, , a2 , a3 be given three numbers.
(a) If aibi is a scalar for an arbitrary vector bi , show that ai (1 5 i
< 3) must be componentsof a vector.
(b) If aib, are components of a second order tensor for any vec-
tor bi , show that ai (1 5 i 5 3) must be components of a vector.
4. Let Xijk (i, j, k = 1.2, 3) be given quantities having the property
that for arbitrary vectors A , B , C with components ai , bi , ci ,
the inner product Xijk aibjck is a scalar. Show that Xijk must be
components of a third order tensor.
5. Let T denote a second order tensor whose components are
given by
Chapter 5
Find the respective components s,, and ti, of the symmetric
and the anti-symmetric tensors whose sum is the tensor T.
6. Let a,, (i, j = 1, 2, 3) denote the components of an anti-symme-
tric tensor. Show that for any numbers bl , b2, b3, it is true
that a,, bibj = 0.
7. Let a,, and b,, (i, j = 1, 2, 3) denote the components of a symme-
tric and an anti-symmetric tensors, respectively. Show that
a . . b = 0.
1J 1J
8. Let aiik (i, j, k = 1, 2, 3) denote the components of a third order
tensor. If the tensor is symmetric with respect to the indices j,
k, show that it is symmetric with respect to the same indices in
any other coordinate system.
9. Let X.- (i, j =1, 2, 3) be given quantities. If Xi, aiaj is always a
11
scalar for components ai of an arbitrary vector, show that (Xi, +
X,, ) are components of a second order tensor. Hence, if X =
'J
Xji , deduce that Xi, are components of a symmetric tensor of
order two.
10. Let X. . (i, j, k = 1, 2, 3) be given quantities. If the inner product
IJ k
X b = c, (1 < i < 3) forms components of a vector, where b,,
l ~ k ~k
are components of an arbitrary symmetric tensor, show that
(Xijk + Xik,) form components of a tensor. Hence, if Xijk =
Xik, , deduce that Xljk are components of a tensor.
11. Referring to Problem 10, i f the quantities Xijkbik = C, (1 5 i 5 3)
form components of a vector, where biJ are components of an
arbitrary anti-symmetric tensor, show that (Xijk - Xik,) are
components of a tensor. Hence, if Xijk = - Xikj, deduce that Xi,,
are components of a tensor.
5.8 PRINCIPAL AXES OF SECOND ORDER TENSORS
In Example 4 of Sec. 5.5, we saw that by introducing an appro-
priate coordinate system, the components I,, (i t j) of the moment of
inertia tensor become zero, thus leaving only the moments of inertia
Ill, 122, 103 with respect to the coordinate axes. Likewise, in Exam-
ple 3, Sec. 5.6, the shearing stress components s,, (i t j ) vanish under
an appropriate coordinate transformation so that the components of
Cartesian Tensors 353
the tensor consisted only of normal stresses along the coordinate
axes. The coordinate axes with respect to which the components of a
second order tensor vanish are called the j ~r i r z ci / 7~1 U X ~ ~ S of the
tensor. In this section, we study the problem of determining the
principal axes of a given second order tensor.
Let tii (i, j = 1.2.3) denote the components of a second order
tensor T, and let ai (i = 1, 2,3) denote the components of a vector A.
We know that the inner product of T and A (with respect to the
second index of ti,) is a vector with components given by bi = ti,ai
(i = 1, 2, 3). It turns out that the principal axes of the tensor T are
determined by those vectors A whose inner product with the tensor
are parallel to the vectors themselves, that is,
where h is a scalar. Such vectors (nonzero), if they exist, are called
the eigenvectors of the tensor, and their directions are called the
principal directions. The principal axes of the tensor are simply the
axes determined by the principal directions. The values of the scalar
h that satisfy (5.36) are called the eigenvalues of the tensor. As it
turns out, the eigenvalues are precisely the components t
e
l l
, t*22, t*33
of the tensor with respect to the principal axes.
Therefore, to determine the principal axes of the tensor T, we
must solve the system of equations (5.36) for the eigenvectors corres-
ponding to each eigenvalue. Now the system of equations defined
by (5.36) consists of
for the components al , a2 , a3 of an eigenvector. Since we are inter-
ested in nonzero vectors, i t follows from linear algebra that the
determinant of the coefficients of the system (5.37) must vanish, that
is.
354 Chapter 5
This determinant leads to a cubic equation in h , known as the char-
acteristic equation of the tensor. Thus the eigenvalues of the tensor
are simply the zeros or roots of the characteristic equation. For each
eigenvalue h , the system (5.37) can be solved for the components al,
a2, an of the corresponding eigenvector, which is non-zero. Clearly,
if A is an eigenvector, so is cA for any constant c. Thus an eigencrec-
tor of a tensor is unique up to a constant factor. In linear algebra the
problem of solving for h and the quantities al , a*, n7 from the system
(5.37) is known as an r>ixtprl z l r 1 1 2 1 t ~ 17 r ohl t v~l .
Eigenvalue Problems for Symmetric Tensors.
Since most second order tensors encountered in practice are
symmetric, we shall confine our study of the eigenvalue problem
(5.37) to symmetric tensors. For a symmetric tensor, we first show
that the eigenvalues are all real. Indeed, suppose h is an eigenvalue
which might be complex, and let ai (1 5 i < 3) denote the components
of the corresponding eigenvector A. Then
Multiplying this equation by the conjugate aei of ai and summing
with respect to i, we obtain
a*. t. . a. =
a*. a.
I 1J 1 I I
(5.39)
If we take the complex conjugate of both sides of this equation,
noting that the quantities ti.; are real, we find
a.t.. a*. = A* a. a*.
1 '.I .I I I
(5.40)
Since the tensor is symmetric, we can show that the left-hand sides
of (5.39) and (5.40) are identical. Indeed, since tij = tji , we have
a*. t.. a. =a*. t. . a. = a . t.. a*. = a . t..a*.
1 ! 1 J J . 1 1 1 1 J l J 1 l J J
Therefore,
ha*i ai =h*a i a *i or ( h - h * ) a i a * i =O
Since a*i ai > 0, this implies h - h* = 0, which means that h is a real
number.
Cartesian Tensors 355
Another important characteristic of a symmetric tensor is that
its eigenvectors corresponding to distinct eigenvalues are orthogo-
nal. This lends the principal axes of a symmetric tensor a natural
axes for a new rectangular cartesian coordinate system. To prove the
orthogonality property, let a, and bi denote the components of two
eigenvectors A and B corresponding to the two distinct eigenvalues
h and p , respectively. Then
If we multiply the first eq~lation by bi, the second equation by ai, and
then sum with respect to i, we obtain
Since t = tji , by re-arranging the factors and changing indices, we
I I
find
h a , b = b. t. . a. = a.t..b. = a. t. . b. = a.t.. b. = p a. b.
1 I 1 1 I . I J l J 1 I J 1 I 1 1 1 I 1 1
therefore,
(h - p) nibi = 0
Since A t p , we conclude that aibi = A B = 0, which says that the
eigenvectors are orthogonal.
With symmetric tensors, even when the eigenvalues are not all
distinct, we can still determine three eigenvectors that are mutually
orthogonal. We shall not prove this fact here but shall simply illus-
trate it by an example (see Example 2).
Finally, we show that when we introduce a new coordinate
system x * ~ with axes along the eigenvectors (ordered so as to be
right-handed), the new components Vij of the tensor satisfy the rela-
tion Vii = O for i t j with t*,, , t* 22 , t*33 assuming the values of the
three eigenvalues h, , h2, h3. So let n")k ( k = 1, 2, 3) denote the com-
ponents of the normalized eigenvector i*i = A, / I Ai I (i = 1, 2, 3) that
corresponds to the eigenvalue hi (i = 1, 2, 3) , that is, i*i = n(i)kik. We
assume that the normalized eigenvectors i*; are so labelled that they
form a right-handed triple. Then the transformation from the xi co-
Chapter 5
ordinate system to the new x*~-coordinate system with base vectors
i*i is given by
x*. = a x.
where a.. = i*.. i . = n(i).
1 1 J J ' ?I 1 J J
Thus the row elements in the transformation matrix A = (aii) are the
comyonents of the normalized eigenvectors. With respect to the
new coordinate system x * ~ , we now show that the components of
the tensor tij assume the following values:
t*,, = hl , t*,, = h2, t",, = h, , t * = 0 for i + j (5.41)
'1
By the transformation law, we have
From (5.36) we have
tpl = hj nQ)p (no sum on j )
so that (5.42) becomes
This yields the relations given in (5.41).
Example 1. Consider the symmetric tensor
The characteristic equation is given by
and so the eigenvalues are h = 0, 1, 3. For h = 0, the system of equa-
tions (5.37) becomes
Cartesian Tensors
al + a 2 = 0 , a] +2 a 2 +a 3 =0 , a 2 + a 3 = 0
from which we find
a, = - a2,
a3 = - a2
Choosing a2 = -1, we obtain the eigenvector Al = il - i2 + i,.
Next, when h = 1, the system of equations (5.37) reduces to
a 2 = 0 , al + a 2 + a 3 = 0
By choosing al = 1, we obtain the corresponding eigenvector A2 = il
- i,. Finally, when h = 3, the system of equations becomes
- 2al + a2 = 0, ill - a2 + a3 = 0, a2 - 2a, = 0
from which we find
a2 = 2al , a l = a,
Setting al = 1, we obtain the corresponding eigenvector A3 = il +
2i2 + i, .
It is easily verified that the three eigenvectors are mutually
orthogonal. Moreover, since the scalar triple product ( A, A, - A3 ) is
positive, i t follows that the eigenvectors form a right-handed triple.
(Note that in case the eigenvectors do not form a right-handed triple,
we need only change the order of any two of them and re-label the
vectors accordingly.) Thus the principal axes of the tensor are along
the direction of the eigenvectors. The transformation from the rec-
tangular cartesian coordinate xi to the new coordinate system x ' ~
determined by the principal axes is defined by x * ~ = ail x, , where
aii = i*i . ii. Thus, taking the dot products of the base vectors, the
transformation is represented by the matrix
Now, with respect to the principal axes, the components of the given
tensor are represented by the matrix
358 Chapter 5
It is instructive for the student t o verify the elements t*,, either
from the transformation law t*,, = a. a. t or from the matrix pro-
1P J9 P9
duct
(t*,) = (aip) (tpq) (ajqlT
Example 2. Consider the symmetric tensor
Its characteristic equation is given by
and so the eigenvalues are h = 3, 3, 7 with h = 3 being a root of mul-
tiplicity 2. Corrresponding to h = 3, the system of equations (5.37)
gives
a2+ \ 3 a , = 0 , \ . 3 a , +3 a 3 = O
which are identical. Hence a2 = - '1.3 a,, so that al and a3 are arbitra-
ry constants. By choosing these constants judiciously, we can obtain
two orthogonal eigenvectors corresponding to the eigenvalue h = 3.
In fact, if we choose al = 1, a2 = 0, a3 = 0, we obtain Al = i,. By
choosing al = 0, a2 = 4.3, a, = - 1, we obtain a second eigenvector
A2 = 43 i2 - iO, which is orthogonal to Al .
Finally, when h = 7, we solve the system of equations
- 4al = o , - 3a2 + &a3 = 0 , &a2 - a3 = o
for which a solution is given by al = 0, a2 = 1, a3 = 4.3. Thus A3 =
i2 + d3i3 is an eigenvector corresponding to h = 7. Clearly this vector
is orthogonal to both Al and A*. These three eigenvectors are the
principal axes of the given tensor. With respect to these principal
axes, the tensor assumes the values
t*,, = t*,, = 3 , t,, = 7 , t * = O for i + j
'I
Cartesian Tensors 359
These values can be readily verified from the transformation law
t *. = a. a t where c r = i*i i and i*i = Ai / I A ,. In fact, we find
1J IP 19 W' IJ I
5.6 EXERCISES
1. The components of a moment of inertia tensor are given by
(a) Find the eigenvalues and the principal axes of the tensor.
(b) Express the tensor with respect to its principal axes.
(c) Write the equation of the transformation relating the coor-
dinates xi with that determined by the principal axes.
(d) Verify the results in (b) by using the transformation law for
second order tensor.
2. Repeat Problem 1 for the moment of inertia tensor
3. The components of a stress tensor are given by
Determine the coordinate system x * ~ in which the shearing
stresses vanish. What are the components of the stress tensor
in this new coordinate system?
Chapter 5
4. Repeat Problem 3 for the stress tensor
5. Find the normal stresses of the stress tensor
with respect to its principal axes.
6. The coefficients ai-i in the equation a,,xix-i =1 (a,, = aji) of a qua-
dric surface formt he components of a second order symme-
tric tensor (see Problem 1, Exercise 5.4). Show that with
respect to the principal axes of the tensor, the equation of the
quadric surface becomes
(In this case, the principal axes are also called the principal
axes of the quadric surface.)
7. Express the equation
of a quadric surface with respect to its principal axes. Can you
determine what type of a surface it represents?
8. A quadric surface is defined by the equation
Find the equation of the surface with respect to its principal
axes. (The surface is an elliptic hyperboloid of one sheet.)
9. Show that the quadric surface defined by
is an elliptic cylinder.
Cartesian Tensors
5.9 DIFFERENTIATION OF CARTESIAN TENSOR FIELDS
We recall that a scalar or a vector field consists of points of a
given domain together with the corresponding scalar or vector at
each of those points. Similarly, if at each point of a domain there is
defined a tensor, then we have a tensor field. For example, an elastic
body that is under the influence of an external force gives rise to a
second order tensor field since at each of its interior points there is a
stress tensor. Thus a tensor field, say of order two, is characterized
by a set of tensor components aij(xl, x2, x3) or briefly aij(xk), which
depend on the coordinates xi (1 5 i 5 3). Sometimes the components
of a tensor field may also depend on a parameter that is independent
of the space coordinates xi. For example, the components may de-
pend on a parameter t which represents time, thus leading to what
is called an unsteady tensor field.
We observe that a tensor field of zero order is simply a scalar
field. Now if @(xk) denotes a scalar field, then under a transforma-
tion of coordinates x * ~ = a x (or xj = aij xqi ), we have
!I 1
On the other hand, a tensor field of order one is simply a vector
field. If ai(xp) denote the components of a vector field, then under a
coordinate transfor ma tion, we have
Transformation laws for tensors of higher order can be written in a
similar manner.
It is evident that all the algebraic operations we have dis-
cussed before pertaining to (constant) tensors also hold for tensor
fields.
Tensor Gradients
Henceforth, we shall assume that the components of a tensor
field can be differentiated as often as needed and that the derivatives
are continuous in the domain of definition. Let us first examine
how the partial derivatives a@ (xi)/axj of a scalar field transform
under a change of coordinates. From (5.43) and by the chain rule, we
find
Chapter 5
since a ~ . / a x * ~ = a , in view of the inverse transformation x = a . . ~ * ~ .
J IJ J IJ
Hence, if we set a*i = a @* ( x * ) / a ~ * ~ and ai = 3@(x,)/axi, (5.45) becomes
J
This shows that the partial derivatives i3@(xj)/3xi of a scalar field
transform as components of a vector. Indeed, this vector field is
grad @ ( xi), which is the gradient of @ ( x,). Thus grad @ ( xj) is a tensor
field of order one. Further, since ii = a i*i, it follows from (5.45)
1J
that
This shows that the representation of grad @(x.;) is invariant with res-
pect to transformation of coordinates.
Next, we consider the derivatives of a vector field and exam-
ine how they transform under a change of coordinates. The compo-
nents of a vector field transform according to (5.44). Hence, by the
chain rule, we find
Setting t*i, = Wi/dx*, and tP1 = dap/dxq we conclude from (5.48) that
the partial derivatives ti, form components of a second order tensor
field. Similarly, if aii(xk) denote the components of a second order
tensor field, the partial derivatives i3aij/axk = tiik become the cornpo-
nents of a third order tensor field. In general, it can be shown by
induction that the partial derivatives of a tensor field of order r
results in a tensor field of order r + 1. We call the partial derivatives
of a tensor field of arbitrary order the l rnsor , yrudi rnt . Thus the ten-
sor gradient of a vector field is a second order tensor. We state this
result as a theorem.
Theorem 1. The tensor gradient of a cartesian tensor field of
order r is a tensor field of order r + 7 .
Cartesian Tensors 363
Caution: It should be noted that Theorem 1 is true only for
car tesian tensor fields because for such tensors all the admissible co-
ordinate transformations are of the form x * ~ = o . , ~ . , where the coeffi-
IJ J
cients matrix (aii) is an orthogonal constant matrix. The transforma-
tion equations aie linear and they represent rotations of the coordi-
nate axes.
It is clear from Theorem 1 that differentiation of a tensor field
may be repeated as often as permissible, and i t will always yield a
tensor field of one order higher.
Divergence of a Tensor Field
Now consider the components of the tensor gradient of a vec-
tor field dai/i)xj. By contraction we obtain
which is precisely the divergence of the vector field. Further, from
the transformation formula (5.48), we find
Equation (5.49) shows that the divergence of a vector field is invar-
iant under transformation of coordinates or, equivalently, that the
divergence of a vector field is a scalar.
Likewise, contraction with respect to the indices j, k in the ten-
sor gradient of a second order tensor aaii/axk leads to the compo-
nents bi = daij/dxj of a vector field. In fact, from the transformation
law
we obtain
which shows that the quantities bi indeed form components of a
Chapter 5
vector field. Similarly, the contraction with respect to the indices i,
k yields components of a vector field b, = 3aij/3xi. The quantities
aaij/3xi and 3aij/3x, are called the components of the divergence of
the tensor field aii with respect to the indices i and j, respectively. In
the general case,we define the divergence of a tensor field with res-
pect to any one of its indices as the contraction of the corresponding
tensor gradient with respect to that index. It follows from Theorem 1
above and Theorem 1 of Sec. 5.7 that the divergence of a tensor field
of order r yields a tensor field of order r - 1.
Example 1. Let al = x2x3 , a2 = xl2x3, a3 = x ~ x ~ ~ be the compo-
nents of a vector field in a domain D. Then the components of the
gradient of this vector field are given by
The divergence is given by
hi
= 2x2x3
3x
Example 2. Let the components of a second order tensor field
be given by
Cartesian Tensors 365
Then the components of the divergence with respect to the index i
are given by the formula
which yields
These are components of a vector field.
The divergence with respect to the second index j is given by
which gives
Notice that the two divergences are not the same.
5.10 STRAIN TENSOR
As a physical example of an application of the discussion in
the preceding section, we present here the concept of strain which
arises in the mechanics of continuous media. When a solid body is
acted on by external forces which are in equilibrium, its configura-
tion exhibits certain deformation. The deformation is usually des-
cribed in terms of the change in the relative distance between two
infinitely close points before and after deformation. This gives rise
to what is called a strain tensor. To describe this, suppose P and Q
are two points which are very close to each other in a body subjected
to external forces in equilibrium. Let the position vectors of these
points with respect to a rectangular cartesian coordinate system be
denoted by R and R + AR (Fig. 5.7) before deformation. Suppose
that after deformation the points P and Q have moved to points P'
and Q' , respectively. Let U(R) denote the displacement of P to P'
and U(R + AR) the displacement of Q to Q'. Notice that the displace-
ment vector U depends on the position of the point since different
points in the body will in general experience different displace-
ments. Then the position vectors of P' and Q' are given by
Chapter 5
OP' = R + U(R) and OQ' = (R + AR) + U(R + AR)
so that the relative position vector between P' and Q' is
AR' = OQf - OP' = (R + AR) + U(R + AR) - (R + U(R))
=AR + U(R+ AR) - U(R)
Thus the change in the relative position vectors between P, Q and
P', Q' is represented by
AU = AR' - AR = U(R + AR) - U(R) (5.52)
Now let R = xiii , AR = Axi ii , AR' = Axf i ii , and U(R) = u.(x.)ii ,
I J
where ui are continuously differentiable. Let us consider the i-th
component of (5.52), Aui = Ax q i - Axi = u. (x + Ax.) - u.(x.). Since the
1 J J I J
points P, Q are sufficiently close, the quantities Axi are very small.
Hence, by the mean value theorem of differential calculus, we can
Fig. 5.7 Strain tensor.
Cartesian Tensors
write
au ,
Aui = u, (x + A X ) - u( x, ) = Axj (sum 01-1 j )
.I
(5.53)
' J ax,
Thus the components of the change in the relative position vectors of
the points before and after deformation are linear combinations of
the Axj ( j = 1, 2, 3). In differential form, this can be written as
a x , a x2 3x3
>
The quantities
i)u
T. . = -
(i, j = 1,2,3)
'J ox.
.I
are components of the tensor gradient of the displacement vector U .
By Theorem 1, this tensor gradient is a second order tensor field
and it is called the di spl ucri ~l erzt t ri zsor. We shall see below that
the symmetric part of the displacement tensor corresponds to the
st rui n torzsor .
We begin by noting that the strain at the point P in the direc-
tion of AR is defined as the ratio
S =
IAR' I - IAR I
IAR I
If we assmue that I AU I is much smaller than either I AR I or I AR' I ,
then the change in the distance I AR' I - I AR I may be approximated
by the projection of AU on AR (see Fig. 5.8), that is
A R ' - A R 1 = A U .
AR
IAR I
Chapter 5
Fig. 5.8 Projection of AU on AR.
Hence (5.55) can be written as
AU. AR
S =
- -
1
AuiAxi
I AR I I A R I ~
In view of (5.53), this becomes
S =
1
AuiAxi =
1 3ui
Axi Axj
I AR l 2 IAR 1
where ni = Axi/ I AR I (i = 1, 2, 3) are the direction cosines of AR. Thus
the strain is a quadratic function of the direction cosines of AR with
coefficients which are the components of the displacement tensor
(5.54).
In matrix form, the strain (5.56) can be written as
where
Cartesian Tensors 369
The quantities S,, are the components of a second order tensor field
called the st rai n t ms o r . It is clear that (S,,) is just the symmetric part
of the displacement tensor when (5.54) is written as
( T ) = (S,,) + (Kij)
1J
where the components of the skew-symmetric part (K,,) are given by
The skew-symmetric tensor field (K,,) contributes to the defor-
mation of the body in a different way. ~ o s e e what kind of deforma-
tion it causes, let us associate a vector w = oiii with the three inde-
pendent components K12, KZ0 , K31 of the skew-symmetric tensor by
defining
(5.59)
where i, j, k are cyclic permutations of 1, 2, 3. Then, from (5.51), we
have
Aui = S - Ax + K.- Ax. = S.. Ax. - % Ax.
'J J 1J J 1J J J
(5.60)
Now the term - mk Ax, in (5.60) is precisely the i-th component of the
cross product w x AR . This product represents the displacement
due to an infinitesimal rotation represented by w (see Fig. 5.9). We
therefore conclude that the term K,, Ax, in (5.60) corresponds to a
deformation due to an infinitesimal rigid-body rotation about an
axis passing through P and parallel to w.
From the foregoing discussion, it follows that if the displace-
ment tensor is zero, that is, du,/ax, = 0 for i, j = 1, 2, 3, then (5.52)
implies AR = AR', which means that the deformation consists only of
translation. Geometrically, this means that the configuration PP'Q'Q
in Fig. 5.7 is a parallelogram. On the other hand, if the strain tensor
(Sij) is zero, then from (5.60) we obtain
Ax' . = =Ax. - O) Ax.
1 k J
Chapter 5
Fig. 5.9 Deformation due to an infinitesimal rotation.
which shows that the deformation consists of a translation and a rota-
tion.
It is interesting to note that by definition (5.59), we have
w = wi ii = (1/2) curl U
where U is the displacement vector of the point P. Hence, if curl U
= 0, then there is no rotation and so the deformation is purely due to
translation and strain.
5.7 EXERCISES
1. Verify that the quantities
are components of a second order plane tensor.
Cartesian Tesnors
2. Show that the quantities
are not components of a tensor.
3. Show that the curl of a vector field is a pseudotensor. (Hint:
Let a, denote the components of a vector field F and write
curl F = Piii, where Pi = 3ak/dx - da./ax,, and i, j , k are cyclic
J J
permutations of 1 , 2 , 3 . Show that
r, s = 1, 2, 3 and then proceed as in Problem 4, Exercise 5.3.)
4. Find the tensor gradient of each of the following:
(a) F(xl, ~ 2 , ~3 ) = x1 sin x3 il - x2 exp(x,) i2 + x2 x3 i3
(b) F(xl, ~ 2 , ~3 ) = (x, - ~2 )il + x3 ln(1 + x12 ) i2 - x12 cos x3 i3
5. Find all the divergences of each of the following:
6. Let ai.; denote the components of a second order tensor field.
Prove that .)ai,/3xk are components of a third order tensor
field.
Chapter 5
7.
Let @ be a scalar field and ai components of a vector field.
Prove that 3(@ai)/3xj are components of a second order tensor
field.
8. Find the symmetric and the skew-symmetric part of the tensor
gradient of
F(xl, X2 , X 3 ) = (xI x2 - x3*) il + x2 exp(x,) i2 + sin (x, x3 ) i,
9. Let aijk denote the components of a third order tensor field.
Write the symmetric and the skew-symmetric part of the
divergence of the tensor with respect to the last index.
10. Let wk be as defined in (5.59). Show that awk/axk = 0 (summa-
tion with respect to k).
11. The most general linear relationship between the stress tensor
(sij) and the strain tensor (Sij) is given by
s. . = h..
ljlun Slun
where hijhn are components of a fourth order tensor called the
elastic tensor. From the symmetric property of the stress and
strain tensors, show that Aijh = hjlk,,, and hijkIn = hijmk. If, in
addition, hijkln = Aklnij, show that the elastic tensor has 21 com-
ponents.
TENSORS IN GENERAL COORDINATES
In this chapter we shall discuss tensors in general curvilinear
coordinate systems. Such tensors are called general tensors or sim-
ply tensors. Obviously, cartesian tensors are special cases of general
tensors since cartesian coordinate systems are particular types of
coordinate systems. As a preview of what prevails in the theory of
general tensors and to make the transition from cartesian to general
tensors somewhat smooth, we shall first discuss tensors in oblique
cartesian coordinate systems. Oblique cartesian coordinate systems
are systems in which the coordinate axes are straight lines that are
not necessarily perpendicular to each other.
6.1 OBLIQUE CARTESIAN COORDINATES
Suppose we change from a rectangular cartesian basis defined
by the unit vectors ii to a new basis defined by the vectors ei (1 2 i S
3), where the new basis is not necessarily orthonormal. Then we can
construct an oblique cartesian coordinate system by taking three
straight lines in the direction of the base vectors ei and intersecting
at a common point 0, called the origin (Fig. 6.1). The straight lines
are the oblique coordinate axes. For reasons that will become evi-
dent later, we designate the coordinate axis along the vector ei by x1
(1
i < 3). using superscript. We reserve the notation xi for rectan-
gular cartesian coordinate system. Now let P be a point in space
and let R denote its position vector. Then, the oblique coordinates
of P with respect to the basis ei are the ordered triple of numbers xl,
x
2
, x
3
such that
Chapter 6
Fig. 6. 1 Oblique cartesian coordinate system.
The coordinate system is said to be right-handed or left-handed
according as the base vectors el, e2, e3, in that order, form a right-
handed or left-handed triple.
It should be pointed out that the coordinates of a point in an
oblique coordinate system depend on the base vectors ei. In other
words, if we choose a different set of base vectors along the same
coordinate axes, the coordinates of the point will change. More-
over, unlike rectangular coordinates, the oblique coordinates (x
l
, x
2
,
x
3
) of a point do not necessarily indicate distances measured along
the coordinate axes. For instance, referring to Fig. 6.1, we observe
that OA = xle,, so that I x' I = I OA I / I el I . Hence, unless el is a unit
vector, I x1 I does not represent the length of the line segment OA.
We pause here to say a few words about subscripts and super-
scripts with respect to the summation convention. The summation
convention now applies to repeated indices of which one is a sub-
script and the other a superscript. Using this convention, the sum
General Tensors
(6.1) can be written simply as R = x' ei. (A reason for using super-
script to denote the coordinates of points in oblique coordinate sys
tems is now a p parent.) As another example, the ex pression
d = g.. xi xJ
'J
implies two summations, one with respect to the index i and the
other with respect to the index j. The summation yields
Notice that we have used parentheses to distinguish between super-
scripts and exponentiation.
For repeated indices that both appear as subscripts or super-
scripts, the summation convention applies provided one of the in-
dices appears in a numerator term and the other in a denominator
term. For example,
i
i 3u1 3~ 1 a ul 2 a U' 3
du = . dx
J
= dx + dx + dx
ax
J
1 2
3~ ax ax"
where the s ummat i o~~ is with respect to the repeated index j, which
appears both times as a superscript in a term in the numerator and a
term in the denominator.
Example 1. Let (3, 4) be the rectangular cartesian coordinates
of a point on a plane with unit vectors il, i2. Find the coordinates of
the point in an oblique coordinate system determined by the basis
e, = 3il, e2 = il + 2i2 (Fig. 6.2).
Solution: The position vector of the given point with respect
to the basis il, i2 i s R = 3i l + 4i2. Let ( x
l
, x ) be the oblique
coordinates of the point. Then, by definition,
Substituting for the basis el, e2 and collecting similar terms, we find
Chapter 6
Fig. 6.2 Oblique coordinates of a point on a plane.
This implies that
3x
1
+ x
2
= 3 and 2x
2
= 4
and so x1 = 1/3 and x
2
= 2. Thus in terms of the basis el, e2, the
position vector of the point is given by R = (1/3)el + 2e2.
Example 2. Let (2,2, -1) be the coordinates of a point with res-
pect to a rectangular cartesian coordinate system. Find its coordi-
nates in an oblique coordinate system defined by the basis
Solution: Let (xl, x
2
, x
3
) be the oblique coordinates of the
point. Then, by definition,
Substituting for the base vectors el , e2, e3, and equating the corres-
ponding coefficients of il, i2, i3, we find
which yields x1 = 7, x
2
= - 5, x
3
= - 1.
General Tensors
6.2 RECIPROCAL BASIS; TRANSFORMATIONS OF
OBLIQUE COORDINATE SYSTEMS
Let x1 denote an oblique coordinate system determined by a
basis ei and suppose x*' is another oblique coordinate system deter-
mined by another basis e*i. We assume that the two systems have a
common origin (see Fig. 6.3). We wish to obtain the equations that
relate these two oblique coordinate systems. For this purpose, we
need to introduce the notion of a Y P C ~ ~ I Y O L ' U ~ bus i s of a given basis.
Fig. 6.3 Oblique cartesian coordinate systems.
Reciprocal Basis
Definition 1. The reciprocal basis of a given basis ei is a set of
vectors el that satisfies the condition
(Notice that the Kronecker delta is now denoted by 6$. )
From the definition we see that for distinct values of i, j, k, the
Chapter 6
vector e
i
is orthogonal to the vectors e and ek, and conversely, the
J
vector ei is orthogonal to the vectors d and e
k
. Further, by the prop-
erty of dot product, it follows that
ei . el = l ei I I el I cos Oi = 1
This means that the angle Oi between the vectors ei and e
i
is an acute
angle. In the two-dimensional case, the situation is illustrated in
Fig. 6.4.
Fig. 6.4 Reciprocal bases on the plane.
We now derive
a given set of base
ctor el . Since el is
a formula for determining the reciprocal basis
vectors ei, i = 1, 2, 3. Let us consider first the
orthogonal to both the vectors e2 and e 3, it
must be parallel to the cross product e2 x eo . Hence, there is a scalar
rn such that
el = nr (e2 x e3)
Taking the dot product with el, we get
General Tensors
Hence we find
and so we have
By following the same procedure, we also find
These results can be represented simply by the formula
where i, j, k are cyclic permutations of 1, 2, 3, and
Notice that the quantity V represents in absolute value the volume
of the parallelepiped formed by the vectors el, e2, e,. V is positive
or negative according as the vectors el, e2, e3 form a right-handed or
left-handed triples. In the sequel, we assume that V > 0 so that our
coordinate system is right-handed.
It is clear that the reciprocal basis given by (6.3) is uniquely
determined. Moreover, if we determine the reciprocal basis of the
vectors e
i
, we are led back to the basis e, (i = 1, 2, 3). In fact, it can be
readily verified that
where i, j, k assume cyclic permutations of 1, 2, 3, and VV' = 1. It
follows that when a basis consists of orthonormal vectors, the reci-
procal basis is the basis itself. For example, the reciprocal basis of
. . .
1, , i2 , 13 is again i , , i2 , i, .
Chapter 6
Example 1. Find the reciprocal basis of the base vectors
Solution: First we find
Hence, by (6.3), we have
Example 2. Find the reciprocal basis of the base vectors
el = 2il + i 2, e2 = 2i1 + 3i2
Solution: Since we are dealing with plane vectors, the formula
(6.3) can not be applied. However, the definition of a reciprocal
basis remains valid. So let us assume that
el = ail + bi2
Then, applying the conditions el* el = 1 and el* e2 = 0, we obtain
2 a +b =1 , 2 a +3 b =0
from which we find a = 3/4, b = -1/2. Therefore,
General Tensors 381
Similarly, by assuming e
2
= ail + bi2 and applying the conditions
e2. el = 0, e2. e2 = 1, we find
These are the reciprocal base vectors corresponding to the given
basis.
Equations of Transformations
We now return to the problem of finding the relationship be-
tween the two oblique coordinate systems x
i
and x * ~ defined by the
respective bases ei and e*,. Consider a point in space and denote its
coordinates with respect to the two systems by (xl, x
2
, x3 ) and (x*',
x' ~, xJ ). Since we assume a common origin, the position vector of
the point with respect to the two bases is given by
(Remember the summation convention.) Let e*' denote the recipro-
cal basis corresponding to the basis e*i. Taking the dot product of
both sides of (6.6) with e*l, noting that e*i-e*J = 6,1, we find
where we have set
Equation (6.7) with (6.8) defines the transformation from the oblique
coordinate system x
i
to the oblique coordinate system x*' as deter-
mined by the bases ei and eqi.
On the other hand, by taking the dot product of both sides of
(6.6) with the reciprocal basis el, we find
. .
x1 = (el- e*,)x*J = a*'. x*J
(i = 1, 2,3)
J J (6.9)
where
ml. = e 1. e*.
J J (1, j = 1,2,3)
(6.10)
Equation (6.9) with (6.10) is the inverse of the transformation (6.7).
382 Chapter 6
Thus the coordinate transformations from the x
i
system to the x*'
system and vice versa are given by equations (6.7) and (6.9), respec-
tively. These transformations coincide with those obtained in the
preceding chapter when the bases consist of orthonormal vectors.
Example 3. Find the coordinate transformation that relate the
oblique coordinate systems x
i
and xXi defined by the respective bases
el = il + 2i2, e2 = i1 + 2i2 + i3, e3 = - i2 + i3
and
e*, = il + i 2, e*2 = i2 + 2i3 , e*; = 2il + i2 - i3
Solution: First, we find the reciprocal bases corresponding to
the given bases. By (6.3), since V = el. e2 x e3 = 1, we find
Similarly, since V' = e*l. e*2 x e*3 = 1, we find
Hence, by ( 629, we have
Therefore, the coordinate transformation from the x
i
- to the xqi -
coordinate system is defined by
The inverse transformation is given by
General Tensors
which can be verified by calculating the coefficients from (6.10).
Properties of Transformation Matrix
The linear transformation (6.7) can be written in the matrix
form
X* = A*X (6.11)
where the coefficient matrix A* is given by
X* = [x"', x**, xe3 lT and X = [x', x
2
, x3 lT. Notice that the superscripts
and the subscripts denote rows and columns, respectively. Now
from (6.8) we readily deduce the relations
,*l - - a*'.; eJ ,
ei = a*J. e*.
' J
(6.13)
These imply that the elements in the i-th row of (6.12) are simply
the components of the reciprocal base vector e*' (i = 1, 2, 3) with res-
pect to the basis e I. Likewise the elements in the j-th column are
the components of the base krector ej ( j = 1, 2, 3) with respect to the
basis e*'.
Similarly, the inverse transformation (6.9) can be written in
matrix form as
X = A*X* (6.14)
where
Chapter 6
Here we observe that the elements in the i-th row of the matrix (6.15)
are components of the vector e1 with respect to the reciprocal basis
e*l and the elements in the j-th column are components of the vector
eej with respect to the basis ei . That is,
Now from (6.13), (6.16) and the definition of reciprocal basis,
we see that
Since the right-hand side of the above equation is the scalar product
of the i-th row of A* and the j-th column of A*, and hi j is the Kronec-
ker delta function, we conclude that the product A*A* is the identity
matrix. Thus, A*A* = I. Likewise, we have
which implies that A*A* = I. Therefore, the matrices defined in
(6.12) and (6.15) are inverses of each other, that is,
Thus, in Example 3, the matrix of the transformation is given by
General Tensors
and the matrix of the inverse transformation is
It is readily verified that
6.1 EXERCISES
1. Let (al , a2, a?) and (bl, b
2
, b
3
) denote the ordered triples (-2, 1,
3) and (1,2,3), respectively. Write down the elements repre-
sented by aibj. What is the term a
i
b
i
' !
2. Let atj denote the element in the i-th row and j-th column of a
matrix (a;). What does ail represent?
3. Let uJ (1 5 j < 3) be some functions of the variables xl, x
2
, x3
that are differentiable. Write down the expressions
, .
4. If u1 = alj XJ (1 5 i <3), where a
i
j
are all constants, show that
5. If a' = (3ui/3x~)bJ and a
i
and bJ are functions of the x
i
's, find
r3ai/3xk.
6. Let R be a twice continuously differentiable function of x
i
(1 5 i 5 3), and set Ai = 3R/3x
1
. Show that 3Ai/3xj = aAj/dxi for
all i, j.
Chapter 6
In each of Problems 7 through 11, the coordinates of a point
Tlven.
with respect to a rectangular cartesian coordinate system are b'
Determine the coordinates of the point in the oblique cartesian coor-
dinate system defined by the given basis.
7. P: (1, 2) ; el = 2il - i 2, e2 = 3il + i2.
8. P: (-2,2) ; el = il + i 2, e2 = -il + 2i2.
9. P: (2, -1,2) ; el = il + i,, e2 = i2 + i 3, e, = il + i,.
10.
P: (3, - 1 2) ; el = i , e2 = il + i 2,
e3 = i l + i2 + i,.
11.
P: (2,3, -2) ; el = il + i 2,
e2 = i1 + 2i2 + i 3, e, = 2il - i2 - i3.
12. Derive the formula (6.5) and show that VV' = 1.
13. Show that if a basis is orthonormal, its reciprocal basis coin-
cides with itself.
14. Find the reciprocal basis of el = il, e2 = il + i 2, e3 = il + i2 + iO.
15. Find the basis of which the reciprocal basis is: el = i2 + i,, e
2
=
i1 + i 3, e" i l + i 2.
16. Find the reciprocal basis of el = i l + i 2, e2 = i l + 2i2 + i j , e3 =
2il - i2 - i, .
In each of Problems 17 through 22, find the equation of the
transformation relating the coordinate systems x1 and x*' defined by
the given bases, and determine whether the transformation pre-
serves or reverses the orientation of the coordinate axes. Assume a
common origin for both system.
. .
17. el = i l + 1, + 13, e2 = - i 2, e3 = - i l + i3 ; e*l = il + i 2, e*2 = i2 +
i, ,
e*3 = i, + i3.
18. el = 2il + i,, e2 = - il + i2 ; eX1 = i l + 2 4 , e** = -2il + i2 .
. . . .
19. el = l2 - 1 3 , e2 = - i l + i 2, e3 = i1 + l2 + 13 ; exl = il , e*2 = i1 +
2i3, e*3 = i1 + 2i2 .
20. el = 2il , e2 = i1 + 2i2 ; e*l = 2il + i 2, e*2 = i 2.
21. el = 2il + i,, e2 = il + 2i2 + i 3, e3 = 3il - i2 + 2i3 ; e*, = - i1 + 2i,
+ i, , es2 = il + i2 + i 3, e*3 = 2il - 3i2 + 2i 0.
22. el = 2i2 - i 3, e2 = i l + 3i2 - i3 , e3 = 2il - i2 + i3 ; e*, = il + 2i2 +
i3 , e*2 = - i, + 2i2 + iO , eX3 = il - 3i2 - 2i0 .
General Tensors
6.3 TENSORS IN OBLIQUE CARTESIAN COORDINATE
SYSTEMS
As an introduction to the study of tensors in general curvi-
linear coordinate systems, we now discuss tensors in oblique carte-
sian coordinate systems. As in our discussion of cartesian tensors,
we first consider vectors in oblique cartesian coordinate systems and
examine how their components transform under a change of the
coordinates.
Tensors of Order One - Vectors
So suppose x
i
is an oblique cartesian coordinate system deter-
mined by a given basis ei (1 5 i r 3). We know that corresponding to
the given basis there exists another basis el (1 < i 5 3), called the
reciprocal basis. Hence every vector may be represented in terms of
either the given basis ei or the reciprocal basis el. This means that
every nonzero vector has two distinct sets of components. To dis-
tinguish these sets of components and to conform with the sum-
mation convection, we denote the comyonents of a vector A # 0
with respect to the reciprocal basis e' by ai, and the components
with respect to the basis e, by al. Thus we have
To determine the components ai and a' from (6.17), we take the dot
product of A with ei and with e
i
, using the defnition of reciprocal
basis. We find
A. e. = (a. el ) * e. = a.
I I 1 1
(6.18)
and
. .
A* el = (a1 ei ) a el = a1
We call the components a, and a1 (1 5 i 5 3) the ~ o z ~ ~ ~ r i i ~ n t and the
contruz~uriu,?t components of the vector A, respectively.
For example, suppose
(ai > 0, a' > 0, i = 1, 2.) Then, by the parallelogram law of vector
addition, the vector forms the diagonal of the parallelogram whose
Chapter 6
Fig. 6.5 Covariant and contravariant components of a vector.
adjacent sides are represented by the vectors alel and a
2
e
2
, involving
the contravariant components al, a
2
, as shown in Fig. 6.5. It is also
the diagonal of the parallelogram whose sides are represented by
a
l
e
1
and a
2
e
2
, involving the covariant components.
Actually, the terms covariant and contravariant have to do
with the manner in which the components of the vector transform
under a change of oblique coordinate systems. To see how the com-
ponents transform, suppose x*' is another oblique coordinate sys-
tem determined by a basis e
x
i
(1 5 i 5 3). Let a
x
i
and a*' denote the
components of the vector A with respect to the bases e*' and e*i,
respectively. Then we have
and
A = a*. e*' = aj
A = a*' e*. = aJ e
j
Now if we take the dot product of e*k with both sides of (6.20),
noting that e*'. exk = 6*ik, we find
General Tensors
where we noted that e*i e J = aJi . This is the same form as the
relationship between e
x
i
and ei as shown in (6.16). Thus the compo-
nents a, transform "in the same fashion" as the base vectors ei (1 5 i <
3), and for this reason they are called couuri unt components. The
base vectors ei are sometimes also called covuri ur~t basis. Similarly,
by taking the dot product of both sides of (6.21) with e*k, we obtain
This is the same form as the relationship in (6.16) between the reci-
procal bases ex' and el, and so the components a' (1 5 i 5 3) transform
"in the opposite fashion" to the base vectors ei. Accordingly, the
components a' are called cont ruuuri unt components of the vector.
The inverse transformation of (6.22) can be found by taking the
dot product of both sides of (6.20) with ej. We find
where eX1* e = ax',. Likewise, the inverse of the transformation (6.24)
J
can be obtained by taking the dot product of both sides of (6.21)
with d. It is given by
where a di = e ~ .
It should be remembered, however, that the two sets of com-
ponents represent the same vector referred to two different bases of a
given coordinate system.
Definition 1. A t ensor of ordrr ont p is a vector with covariant
components ai or contravariant components a, (1 5 i 5 3) with respect
to an oblique cartesian coordinate system x1 defined by a basis ei
(1 5 i < 3). Under a change of coordinates, the components transform
according to the laws (6.22) and (6.23) or (6.24) and (6.25).
Example 1. Consider an oblique coordinate system x
i
defined
by the basis
. . . .
el = l1 + 12, e2 = i2 - 13 , e3 = il + 2i3
Find the covariant and the contravariant components of the vector
A = 2i, + i2 - 3i3.
Chapter 6
Solution: By (6.18) the covariant components are given by
These are the components of the vector with respect to the recipro-
cal basis el (1 5 i 53). Since el. e2 x e3 = 1, the reciprocal base vectors
are given by
e1 = e2 x e3 = (i2 - i3) x ( il + 2i3) = 2il - i2 - i3
e
2
= e3 x el = (il + 2i3) x ( il + i2) = - 2i1 + 2i2 + i3
e
3
= el x e2 = (il + i2) x ( i2 - i,) = - il + i2 + i3
Thus, A = 3e1 + 4e
2
- 4e
1
. Next, by (6.19), the contravariant compo-
nents of A are
a' = A* e1 = (2il + i2 - 3i3 ) . (2il - i2 - i3 ) = 6
hence, A = 6el - 5e2 - 4e3.
In practice, certain vector quantities are naturally represented
by vectors whose components transform according to (6.22) or (6.23).
Accordingly, the vectors are called covariant vectors or contravar-
iant vectors. However, as we shall see later, the covariant and con-
travariant components of a vector are related; that is, it is possible to
switch from covariant to contravariant components or vice versa.
This property also holds for general tensors.
Example 2. Let Q (x
i
) be a scalar field in a rectangular cartesian
coordinate system x
i
and let x * ~ = n*ijx.i define an oblique coordinate
system x*~. Verify that 3Q/ axi (i = 1, 2, 3) are covariant components
of the gradient of Q.
Solution: We note that in rectangular coordinate system, we
have ei = ii = el, and so
Now under the given transformation, we find by the chain rule
General Tensors
which is of the form
ai = a*Ji a*.
(i = 1, 2, 3)
J
where
a ' = . , a* . =
I 1
a@
a x ' ax*
This is the transformation law (6.24) for the covariant components of
a vector. Further, since e * ~ = a*Ji e
i
, we see that
O@ e i = a@ ax*' i
grad@ = e
i
ax ax*-' ax'
This shows that the natural expression for grad @ is in terms of its
covariant components. Hence we say that grad Q, is a covariant vec-
tor field.
Example 3. Let R(t) = x
i
(t)e
i
, t t 0 be the position vector of an
object moving in space with respect to an oblique coordinate system
with basis ei. Then the velocity of the object is given by
If we change to another coordinate system defined by X*J = x
k
with basis e*i, the components d ~ * ~ ( t ) / d t of the velocity vector rela-
tive to the x*' coordinate system are related to dx
i
(t)/dt by the equa-
tion
dx* '(t) - - dx* dx
J
(t) = a* i . dx
J
(t)
dt dxj dt dt
This is the transformation law (6.23) for the contravariant compo-
nents of a vector. Further, since a*lj e*i = e., we have
J
Chapter 6
Thus the natural expression for velocity is in terms of contravar-
iant components. Hence we say that velocity is a contravariant
vector.
Example 4. Let A = a
i
e
i
= a
i
e
i
and B = b
i
e
i
= b
i
e
i
be two vec-
tors represented by their covariant as well as contravariant compo-
nents. Then the scalar product of A and B can be obtained in four
different ways, namely
k B = (a
i
ei)* (b. 4) = a' bi
. J
= (ai ei)* (aJ e,) = ai b1
= (aiei ) a (bJ e,) = ai bjei e.
J
= (ai e
i
) (b, d ) = ai bjei* eJ
It appears that the natural expression for the scalar product of
two vectors involves the covariant components of one vector and
the contravariant components of the other. In fact, under a coordi-
nate transformaion, we see that
since a*Ipa *iq = 6,,(l Likewise, we have
noting that a e i P a*lq = 6*" This shows again the invariant property
of the scalar product. The other two expressions involving both
covariant or both contravavariant components require the sets of
scalar products e i. e or ei+ e ~ . These sets define what is called the
J
metric tensor which is discussed in a later section.
General Tensors
Tensors of Higher Order
To define tensors of higher order, we need to introduce a more
general kind of notation that is universal for general . . tensors. First,
from a given coordinate transformation x*' = a*'j xJ or its inverse
x = x * ~, we observe that
Accordingly, the transformation laws for the covariant (6.22) and the
contravariant (6.23) components of a vector (tensor of order one) can
be written equivalently as
(Recall the summation convention with respect to repeated indices
that are both subscrip ts or superscripts.)
Definition 2. A tensor of order two is a quantity consisting of
32 (nine) covariant component a - or 3? contravariant components air
IJ
which transform under a change of coordinates according to the law
As we shall see later on, these are the same transformation
laws that govern general tensors of order two when x1 and x* l denote
general curvilinear coordinates. Notice that in cartesian coordinate
systems (rectangular or oblique), the quantities dxP/dx*q and dx*Q/ dxq
are constants because the coordinate transformations involve only
linear equations. In general curvilinear coordinate systems, these
quantities are no longer constants. This is a distinct difference be-
tween cartesian tensors (in rectangular or oblique coordinate sys-
tem) and general tensors.
394 Chapter 6
Example 5. The components of a second order symmetric car-
tesian tensor are given by
2 -1 0
(a,) = -1 0 3
Find the covariant and the contravariant components of the tensor
in the oblique coordinate system x
i
defined by the basis
(See Example 1.)
Solution: We note that for cartesian tensors, ei = el = ii (1 5 i 5
3) and a,, = ail. Hence, from Example 1, we have
so that a*Ij = e". e, is defined by the matrix
The inverse matrix defined by a*lj = el- e * is given by
J
Hence, by the formula a*,, = a * , P a *,q am, we find
aXll = a r l P a r l q aw = 0 , a*12=a*21 =ae1Paw2qa ~9 = - 4
aX13 = a*31 = a*1"*7q aP9 = 7, a*22 = ~ x ~ P a*2q am = - 4
= = a *2P am = 1, a*33 = m3P a* q a = 10
3 Pq
or, in matrix form
General Tensors
0
(a* ,,I = [-4
7
The contravariant componel~ts are given by the formula
J = a*lp a*Jq am
Thus we find
or, in matrix form
Notice that both (a*' ~) and (a*ii) are symmetric matrices; thus we see
that the symmetric property of the tensor is invariant under a coor-
dinate transformation.
Example 6. Let ai and bi be the covariant components of two
vectors, and define c = a b- (i, j = 1, 2, 3). Show that cij are compo-
'J 1 J
nents of a second order tensor.
Solution: The components ai and bj transform according to the
law (6.22). Hence, under a change of coordinate system, we have
which is in accord with the transformation law (6.27) for the covar-
iant components of a second order tensor.
396 General Tensors
Definition 3. A t t Jnsor of order r is a quantity consisting of 3'
covariant components
a . ( i I , i 2 ,..., i r =1, 2, 3)
1 1 l 2 . . . I,.
or 3' contravariant components
which transform under a change of coordinates according to the law
Although we shall not be dealing too much with it, we should
mention that there are also so-called rtz i x ~ d t ~ n s o r s whose compo-
nents have covariant as well as contravariant indices. For example, if
we take the outer product of two vectors using the covariant com-
ponents of one vector and the contravariant components of the other,
say, c$ = aibj, we obtain components of a second order mixed tensor.
The components transform according to the law
where c*? = a*ib*~. As another example, the outer product of a sec-
ond order tensor with components ail and a vector with components
b
k
leads to third order mixed tensor whose components are given by
c
k
= a..bk. The components of this tensor are governed by the trans-
1J 1J
formation law
where P i j k = a * bxk
1J .
General Tensors
6.4 ALGEBRA OF TENSORS IN OBLIQUE COORDINATES
With slight modifications, tensors in nonrectangular cartesian
coordinate systems possess much the same algebraic properties as
cartesian tensors. The modifications are necessitated by the diffe-
rence in the transformation laws governing covariant and contravar-
iant components of tensors. For instance, the sum of two tensors will
be defined only if the tensors have the same order and the compo-
nents have the same character, that is, the number of covariant (sub-
scripts) or contravariant (superscripts) indices must match. For ex-
ample, if ajk and bq are the components of two second order tensors
both having two covariant indices, then the sum of the two tensors is
a tensor whose components are given by cjk = aj, + bjk (j, k = 1, 2, 3).
On the other hand, the outer product of tensors are always defined
regardless of the order and character of the individual components.
The order and character of the product is determined in an obvious
manner. For example, the outer product of two tensors with respec-
tive components aij and b
h
is a tensor whose components are given
by cij- = a
ij
b
h
, which are of the mixed type. The verification of
these facts are similar to those given for cartesian tensors in the pre-
vious chapter.
One aspect of the algebra of cartesian tensors that needs certain
modification is the operation of contraction and, as a result, the
operation of inner product. In order for the contraction of two
tensors to be valid, the contraction must be performed over a pair of
indices, one of which must be a subscript (covariant index) and the
other a superscript (contravariant index). Under this condition,
Theorem 1 of Sec. 5.7 continues to hold. For example, let us consi-
der the contraction of a third order mixed tensor with the compo-
nents aljk Contraction with respect to the first subscript yields bj =
a',, while contraction with respect to the second subscript gives ci =
aJij. Both of these contractions lead to tensors of the first order or
vectors. In fact, from the transformation law
and the identity
Chapter 6
we see that
i q ax
r
p - ax
r
p
a* ik = Sp
k a pr
ax*
where b5 = a*iij. This shows that the quantities br = aPpr (r = 1, 2,3)
are components of a vector. A similar proof can also be shown for
the quantities ci = di j .
Therefore, the inner product of two tensors of order r and s is a
tensor of order r + s - 2, provided, of course, that the contraction is
performed over a pair of indices one of which is a subscript and the
other a superscript.
Theorem 2 of Sec. 5.7 (quotient rule) has its counterpart for
noncartesian tensors. That is, if
are quantities which have the property that their outer or inner pro-
duct with an arbitrary tensor yields a nonzero tensor of an appro-
priate order, then the quantities are components of a mixed tensor
of order k, where k = m + n.
To indicate the method of proof, let us suppose that TIjk are
quantities which have the property that for an arbitrary vector with
components a
k
, the inner product ~~~~a~ = b
i
j
results in components
b
i
j
of a mixed tensor of order two. Let T*;, denote the value of the
quantities T\k in the coordinate system x*'. For an arbitrary vector
a
k
, let a*k be the corres onding components in the coordinate sys-
tem x*~, and set T*$ a9 = h*lj Since b*lj are components of a mixed
tensor, they satisfy the transformation law
General Tensors
Likewise, we have
Hence
so that
t k
- -
a x*Pa xt k a X * P ~ x ax
Tt ka = a* r ~ s k
rlxS ax*' ax
S
ax* ax*
t k
a X x P a x ax
(T*q -
T , ~ )a* = o
ax
S
ax* ax*
Since this is true for an arbitrary vector, we conclude that the term
inside the parentheses must vanish. Thus the quantities TIjk trans-
form as components of a third order mixed tensor.
It should be noted that all the algebraic properties satisfied by
tensors in oblique coordinate systems are also satisfied by general
tensors. For this reason, we shall not present a separate discussion
of the algebra of general tensors.
Example 1. Suppose the quantities Tij have the property that
Ti, aib, is a scalar in any coordinate system, where a
i
and bJ are com-
ponents of two arbitrary vectors. Show that Ti, are components of a
second order tensor.
Solution: Let TI,, denote the values of Ti, in the coordinate
system x*~. For any vectors with components a
i
and W, we have
~*. . a*~b*. J = T apbq,
(a scalar)
1J P9
Since
a xP .
a = , a*' ,
ax* ax* '
Chapter 6
we find
. . 9
ax
P
i ax ,*j
T*, a* bJ = T~~ a*
ax* a x*j
Hence
P 9
(T* ij - T Pq ax ax )a+,*J=o
ax* 'ax* J
from which we conclude that the term inside the parentheses must
be zero. Thus we obtain
T*. . =
axP ax
q
1J jTw
ax* 'ax*
which shows that the quantities Tij indeed transform as components
of a second order tensor.
Example 2. Let a' and bi be the contravariant and covariant
components of two vectors A and B, respectively. Then the quanti-
ties cia, = a
i
b
j
form as components of a second order mixed tensor,
which is the outer product of A and B. In fact, since
we see that
ax*'
a* I - - a , b * ; = - - b q ax
q
axP ax*
General Tensors 401
The contraction of yields c ' . ~ = albi = A-B, which is the inner pro-
duct of the vectors.
6.2 EXERCISES
1. Deduce from (6.17), (6.18), (6.19) the expression
A = ( A ei)el = (A* el)ei
and then, by replacing A by e*' and e*,, obtain the results (6.13)
and (6.16).
2. In each of the following cases, find the covariant and contra-
variant components of the given vector in the oblique coor-
dinate system defined by the given basis e, (i = 1,2,3).
. .
(a) A = i, + 2i2 ; el = -il + 2i2, e2 = i1 + i2 .
(b) A = -2i, + 3i2 ; el = i, + 2i2, e2 = - il + i 2.
(c) A = i1 + 2i2 - 3i3 ; el = il + i2 , e2 = i2 + i3 , e3 = il + i3 .
. . . .
(d) A = 3il - i2 + 2i0 ; el = i1 - i 2, e2 = i2 - 10, e3 = il + i?
3. A vector field F has components F1 = x1x2, F2 = ~3x2, F3 = x1x3
with respect to a rectangular cartesian coordinate system x ;.
Find its components in the oblique coordinate system x1 de-
fined by the basis ei given in Problem 2(c).
4. Repeat Problem 3 for the vector field F = x12 i, + x22 i2 + x32 i3 ,
where the oblique coordinate system is defined by the basis ei
of Problem 2(d).
5. Find the covariant and contravariant components of the carte-
in the oblique coordinate system defined by the basis
el = i, + i,,
e2 = i l + 2i2 + i3,
e, = - i, + i2 + i3.
6.
Find the covariant and contravariant components of the carte-
Chapter 6
sian tensor
in the oblique coordinate system defined by the basis
7. Write the transformation law for a third order (a) covariant
tensor, (b) contravariant tensor, (c) mixed tensor with two
covariant indices.
8. Write the transformation law for a fifth order mixed tensor
having two covariant and three contravariant indices.
9.
Let A = (aii), A* = (a*ii ), A* = (a*;), = 3x1/axJ, with super-
scripts indicating rows. Show that the transformation law
a*. . =
ax
P
ax
q
IJ j a ~ q
ax* I ax*
is equivalent to the matrix multiplication A* = A * ~ A A*.
10 Obtain the corresponding matrix formula for the tranforma-
tion law
i j
ij ax* ax* pq
a* = a
a xP ax
q
11. If aii and blanll are components of two tensors, show that the
quantities cijkllul = a b k mn are components of a mixed tensor,
'J
covariant of order two and contravariant of order three.
12. If aii and bkm are components of two tensors, show that cik =
aijbjk are components of mixed tensor of order two.
13. Suppose that the quantities Xi W have the property that for arbi-
trary tensor with components aJ, the outer product bijM =
J
XiWajrare components of a tensor. Show that XiP4 are compo-
nents of a mixed tensor.
General Tensors
6.5 THE METRIC TENSOR
In Sec. 3.11 we introduce the notion of a metric in connection
with the expression for arc length in a curvilinear coordinate sys-
tem. The metric is defined by the quantities g,, given by
aR aR
i ' (i, j = 1,2,3)
au au
j
where u
i
denotes the curvilinear coordinates. Now in an oblique
coordinate system x
i
defined by a basis ei , we have R = x
i
ei so that
aR 3R
-
- = ei (i, j = 1,2,3)
- i - i
Thus, in an oblique coordinate system, the quantities gij are given by
pi, = e i ej
(i, j = 1, 2, 3)
(6.34)
If the base vectors ei (1 5 i 5 3) are unit vectors, then gij = cos 8,,.
where 8,, denotes the angle between the x
i
- and XJ - coordinate axes.
In particular, if the coordinate system is rectangular, so that the basis
is an orthogonormal set, then gu = 6,,.
The quantities gii defined in (6.33) form the covariant compo-
nents of a second order tensor. In fact, from the transformation law
(6.16) of the basis and (6.26), we see that
g* ij = e* i.e> =
a x P ax
q
axP ax
q
e . e = -
i j P 9 j ' ~ q
ax* ax* ax* ' ax*
Thus the quantities g,, transform as covariant components of a se-
cond order tensor. This tensor is called the r w t ri c tvrzsor of space
with respect to the coordinate system XI. Since g = g. , , the metric
' J J'
tensor is symmetric.
Using the reciprocal basis e
i
, we can likewise define the quanti-
ties gij by the equation
g i ~ = 4 (i, j = 1, 2,3) (6.36)
Chapter 6
From the transformation law of the reciprocal basis, it follows that
Thus the quantities gij are the contravariant components of a second
order tensor. This tensor is precisely the metric tensor whose covar-
iant components are the quantities gi,.
We note that the g,, and gij satisfy the important relation
To see this, we observe that the metric tensor relates the basis ei to its
reciprocal basis ei according to the formula
e. = g.. er
1 ?I
This follows from the identity A = (A. ej)d when we replace A by e,.
Now, if we take the dot product of both sides of (6.39) with e
k
, noting
that ei. e
k
= Sik, we find
Sik = g . d . ek = g..
1J ,J gJk
which establishes (6.38).
The relation (6.38) implies that the matrix ( g*~) is the inverse of
the matrix (g,,), and vice versa. This enables us to relate the cofac-
tors of the elements of (gi,) with the elements of (gij). Indeed, from
linear algebra, we recall that if A = (aii) is a nonsingular matrix, its
inverse is given by the formula
-1 adj A
A = - -
IA I
where adj A denotes the adjoint of A. The adjoint of A is defined as
the transpose of the matrix (A,,), that is, adj A = ( A, , ) ~, where A,, is
the cofactor of the element a,,. NOW since (g,,) is symmetric, its ad-
joint matrix is also symmetric because the cofactor of g,, is the same
as the cofactor of gji. Hence, if we denote the cofactor of g,, by G'J
and the determinant of (g,,) by g, then from the above formula, we
General Tensors 405
have
G'J = g gij = cofactor of gij (6.40)
This result is required in Sec. 6.11.
Example 1. Let el = il + i 2, e2 = i2 + i3, e3 = il + i3. Find the
quantities gij and giJ, and show that (gij)(gi~) is the identity matrix.
Solution: By (6.34) we find
Next, we determine the reciprocal basis e
i
(i = 1, 2, 3). Since elo e2 x
e3 = 2, we find
e1 = (e2 x e3)/2 = (il + i2 - i3)/2
e
2
= (el x e1)/2 = (- il + i2 + i3)/2
e" (el x e2)/2 = (il - i2 + i3)/2
Therefore, by (6.36), we have
It follows that
Relationships Between Covariant and
Contravariant Components
The covariant and contravariant components of a tensor are
related to each other through the metric tensor. Here we derive the
relationship for tensors of the first order or vectors. Let us consider
a vector A = ajd = ajej represented by its covariant as well as contra-
406 Chapter 6
variant components in a coordinate system defined by the basis ei
(1 2 i 13) . Taking the dot product of the vector with the base vector
e,, noting that el* ek = 6d, we find
In matrix notation this can be written as
Similarly, by taking the dot product of the vector with the recipro-
cal base vector el, we find
Thus we can shift from covariant to contravariant components or
vice versa through the metric tensor. Notice that both (6.41) and
(6.42) are inner product of the vector with the metric tensor. In a
rectangular cartesian coordinate, we have g.. = 6,,, so that a1 = ai.
'J
Hence for cartesian tensors there is no distinction between covariant
and contravariant components.
The relationships among the components a', a', a*', a*' of a
vector under a transformation of coordinates are best depicted in the
following schematic diagram:
x1 - coordinate system
I
i
x* - coordinate system
.................................
General Tensors
Similar relationships also hold between the covariant and con:
travariant components of tensors of higher order. For example, if alJ
are the contravariant components of a second order tensor, then its
covariant components are given by
To see this, let aIij denote the components of the tensor in rectangu-
lar cartesian coordinates xi . Then we have
axp axq
a. . = , , a' pq (sum on p and q)
1.l
3x1 axJ
Since a'm = atF4 for cartesian tensors and
;)xp axq kin
a' W =
a (sum on k and m)
k m
ax ax
it follows that
axp 3xpaxq axq k ~ n
a .. = a (sum on p and q) (6.45)
' i k j 111
3x 3x ax ax
Now from (6.16) and (6.26), we have
ifxp.
ei = -
l p (sum on p)
so that by (6.34)
axp axp
gij = ei - ej =
(sum on p)
ax
i
axJ
Thus (6.45) becomes
a IJ = gkgjn,ahn
which is (6.44). Conversely, if aij are known, then the contravariant
Chapter 6
components of the tensor are given by
Both (6.44) and (6.46) involve taking the inner product of the tensor
twice with the metric tensor.
It follows from (6.38) that
so that by (6.46) we conclude that the quantities giJ are the contravar-
iant components of the metric tensor.
The inner product of the metric tensor with tensors of order at
least two can lead to mixed tensors. For example, the inner product
of the metric tensor with a second order tensor with components aii
and ail leads to a mixed tensor with components defined by
and
In general the components ai.k and a'.k are not identical (see Prob-
lem 8).
From (6.47) and (6.48) we observe that taking the inner product
of a tensor with the metric tensor has the effect of lowering or
raising the index with respect to which the summation is performed
according to whether the covariant components gij or the contravar-
iant components glJ are involved in the product. The spot from
which an index has been lowered or raised is indicated by a dot. As
further example, let us consider the inner product of the metric ten-
sor with a fourth-order mixed tensor with components aiiW. The
inner product leads to four mixed tensors of the fifth order with
their components given by
General Tensors 409
We conclude this section by deriving the relationship between
the determinant g of the metric tensor matrix (g,,) and the quantity
V = elo e2 x e3. By (6.39) we note that el = gl j d, e2 = g2,d, e3 = g3,d
so that
V= e l e 2 x e 3 = g l i g 2 g e i *dxe k
J 3k
(6.49)
[Note that in (6.49) summations are to be performed with respect to
the repeated indices i, j, and k.] Now since V' = el* e
2
x e
3
has the
same value under cyclic permutations of 1, 2, 3, and since ei. d x e
k
= 0 whenever two of the indices are identical, it follows that
But V' = 1/V; therefore, we have V* = g or
Example 2. Let A = aiei = a
i
e
i
and B = b
i
e
i
= b
i
ei. In terms of
the covariant components, the scalar product of these vectors is
given by
A* B = (a
i
e
i
) (bj$) = gi ~ai bj
which involves the contravariant components of the metric tensor.
In terms of the contravariant components, the scalar product is given
by
A- B = ( (bJe.) = g. ai b~
J 'J
involving the covariant components of the metric tensor.
Example 3. Express the vector product of two vectors in terms
of their covariant and contravariant components, assuming a right-
handed coordinate system.
410 Chapter 6
Solution: First, suppose A = a
i
e
i
and B = b,ej. Then, by defini-
tion, we have
where i, j take on the values 1, 2; 2,3; 3 , l only. Since
it follows that
A x B = ~ ( a ' b ~ - a,bi)ek
where i, j, k are cyclic permutations of 1, 2, 3. Thus we find
A x B = v[(a2 b3 - a3 b2)e1 + (a%l - a' b3)e2 + (a
1
b2 - a2 bl)$]
Next, suppose A = a,e
i
and B = bjd. Then, by a similar argu-
ment, we find
= V a
Example 4. The covariant components of a second order
tensor with respect to the basis el = il + i 2, e2 = i2 + i 3, e3 = il + i3
(see Example 1) are given by
1 2 3
e e e
1 2 3
a a
b' b
2
b3
(a,,) = -1 2 0
[: : -11
Find the components ail, ai- and
General Tensors
Solution: From Example 1, we have
Hence, by (6.46), we find
The student may check these results by performing the matrix
multiy lication
(a1.i) = (gil ) ( ai j ) ( gi ~ )T
Next, by (6.47), we find
a, 1 = gljaj1 = 5/ 4, a1.2 = gljd2 = -3/4, a1.3 = gljd3 = 1/ 4
a,.' =ggajl =- 5/ 4, a 2 . ? =g 2 . a ~2 =7 / 4 , a2.3 =R2j d3=- 1/ 4
J
a,.' = gT.aJ1 = -1 / 4 , a , ? = g3ja~2 = 7/4 ,
= gIjaj3 = -5/4
J
and by (6.48), we have
a',1 = gljajl = 7/4,
= Rlja
= -1, al.3 = glja- = 1
j2 3 3
a',, = g2jajl = -5/4, a2 . - = g2jaj2 = I, a2,3 = g2Jni3 = O
a"l = g-?iajl = -1/4, a3.? = g3Ja
= 1, a3 - 7J
j2
.3 - gL aj3 = -1
Chapter 6
6.3 EXERCISES
1. Let
relative to the basis el = i2 + i 3, e2 = il + i3, e3 = il + i2 + i3.
Find the components ail, ai ek and ai-k of the tensor.
2. Let
1 3
x o x
(a,,) =[: 2 : j
Find the components aiJ, ai ak and ai.k of the tensor, relative to
the basis
el = i2 + i3,
e2 = il + i2 + 2i3, e3 = il + i2 + i3.
3. Suppose
relative to the basis of Problem 1. Find the components a,,, ai ek
and ai-k of the tensor.
4. The metric tensor g,, of an oblique coordinate system deter-
mined by a basis ei is given by
(a) What is the distance between the two points whose oblique
coordinates are given by (2, 1, -3) and (4, -3, -I)?
General Tensors
(b) Find the scalar product of the two vectors A = 2el + e2 - 2e3
and B = - el + 3e2 + 2e3.
(c) Express the vector product of A and B in terms of the reci-
procal basis el.
The metric tensor of an oblique coordinate system is given by
with respect to a basis ei. If A = 3el - e2 + 2e3 and B = 2el + 3e2
+ 4e3, find A* B and A x B. (Express in terms of the reciprocal
basis el.)
Show that = ai. k [see (6.47) and (6.48)] if and only if aji = aij
for all i, j.
Write down the formula relating the components aijk and ailk
of a third order tensor.
Show that
g. ir a i k = g~~gk4aipl
Show that
g..dk = gkJa.. and pila. = gkjaij
1J 1J I k
Verify the result V = dg in (6.51) for the basis el = i, + i2 - 2i3,
e2 = 2il - i2 + i3, e3 = i1 - 2i2 + iR.
Obtain the formula e
i
= gijej.
Using the result of Problem 11, show that V' = dG, where
V' = el. e
2
x e
3
and G is the determinant of the matrix (g,'i).
Thus deduce that gC = 1.
6.6 TRANSFORMATIONS OF CURVILINEAR
COORDINATES
We recall from Sec. 3.8 that a curvilinear coordinates system u1
(notice the shift to the use of superscript) is defined by a set of equa-
tions
Chapter 6
where the functions u
i
(x
l
, x2, x3) are single-valued and continuous-
ly differentiable in a domain D with cartesian coordinates x, , x2, x3.
It is assumed that the Jacobian 3(u1, u
2
, u
3
)/d(xl , x2, x3 ) t 0 at each
point in D, so that the inverse of (6.52) exists in a neighborhood of
every point in D. Let
denote the inverse transformation of (6.52). Naturally, since the
inverse of (6.53) leads back to (6.52), the Jacobian a( xl , x2 , x3 )/r)(ul,
u
2
, u
3
) of the transformation (6.53) is also different from zero in a
corresponding domain of the coordinates ul. As a matter of fact, the
Jacobian of the transformations (6.52) and (6.53) satisfy the impor-
tant relation
This is easily verified by the chain rule of partial differentiation
Local and Reciprocal Bases
So suppose we have a curvilinear coordinate system in our
space defined by (6.52) and let the inverse transformation be given
by (6.53). Then at each point P in space we can introduce a local
basis consisting of the tangent vectors to the u'-coordinate curves
passing through P (see Fig. 6.6). The base vectors are given by
where R is the position vector of the point P, and the x
k
(u
i
) are the
inverse functions defined in (6.53). Thus the metric tensor at the
General Tensors
point P has the components
which are functions of the coordinates of the point P.
Corresponding to the local basis (6.56), we can also define the
local reciprocal basis e
i
in exactly the same way as in Sec. 6.2. Hence
they are determined by the formula
for i = 1, 2, 3 where i, j, k are cyclic permutations of 1, 2, 3. Alterna-
tively, the local reciprocal base vectors el are more conveniently
determined by the formula
Fig. 6.6 Lc ~c a l base vectors.
Chapter 6
where the u
i
are the transformation functions defined in (6.52). That
(6.59) provides the local reciprocal basis follows readily from (6.55)
since
Now we know that vui is a normal vector to the coordinate
surface u
i
(x
l
, x2 , x3 ) = ci, i = 1, 2,3. Thus we see that while the local
basis consists of the tangent vectors to the coordinate curves at the
point P, the local reciprocal basis consists of the normal vectors to
the coordinate surfaces at P (Fig. 6.7). Therefore, if the curvilinear
coordinate system is orthogonal, then for each value of the index i,
the vectors ei and e
i
are parallel and have the same direction, so that
I ei I I el I = 1 for i = l , 2, 3.
It is instructive to note that the Jacobian of the transformation
(6.52) can be written as
1 2 3
a ( ~ , u , U ) 1 2 3
=Vu .Vu x v u
x*, 9)
Fig. 6.7 Local and reciprocal base vectors.
General Tensors
while the Jacobian of the inverse transformation (6.53) can be writ-
ten as
(Cf. Probs. 1 and 2, Exer. 3.7.) In Sec. 6.2 we established that VV' = 1,
thus we have again verified (6.55) from another viewpoint.
Example 1. The position vector of a point in the cylindrical
coordinate system is given by
Let us set u1 = r, u2 = 8, u3 = z, noting that i = i l , j = i 2, k = i3. Then
the local basis consists of the vectors
so that the metric tensor is given by
The reciprocal basis consists of the vectors
Chapter 6
2 X2 X ~ . sin 0 cos 0 .
e = V0 = - i l + 2 i 2 = -
2 r i l + '2
r r
3
e = Vz = i 3
and so
Coordinates Transformation
Now suppose u*' is another curvilinear coordinate system de-
fined by the transformatio~i equations
where the functions uqi(xl , x2, x3) possess the same properties as
those in (6.52). If we substitute (6.53) for xi in (6.62), we obtain the
transformation equations relating the two curvilinear coordinate
systems u1 and u*'. Suppose that these equations are given by
In y articular cases, these equations might be relating a rectangular
and an oblique coordinate system; a rectangular and a cylindrical or
spherical coordinate system; or a cylindrical and a spherical coordi-
nate system (see Prob. 8). Since
i i
au* - au* axk
-
3u j dxks ui
it follows from the definition of matrix multiplication that
General Tensors 419
This is the analogue of the chain rule of partial differentiation for
Jacobians. In particular, this shows that the Jacobian of the transfor-
mation (6.63) also differs from zero; hence, the inverse transforma-
tion equations, say
exist. Of course, these equations can also be determined by substitu-
ting the inverse of (6.62) for xi in (6.52).
Now let e*' denote the local basis corresponding to the cur-
vilinear coordinates u*' defined by (6.62). We wish to obtain the
relations between the local bases ei and e
i
of the two coordinate sys-
tems u
i
and u*'. From the definition (6.56) and by the chain rule, we
find
where the coefficients ;IuJ/;Iu*l are to be calculated from (6.65). The
inverse relation is given by
Similarly, the relations between the reciprocal bases e
i
and e*' can
be obtained by using the formula (6.59) along with the chain rule.
We find
The inverse of (6.68) is given by
i
i du e,j
e =
au* j
Chapter 6
[Compare these equations with (6.13) and (6.16).]
Example 2. Find the relationship between the cylindrical
coordinates
X, = r cos 8, x2 = r sin 8, x3 = z
and the elliptic cylindrical coordinates
xl = a cosh u cos v, x2 = a sinh u sin v, x3 = w
Calculate the Jacobian 3(r, 0, z)/3(u, v, w).
Solution: We note that
hence
Further,
so that
a sinh u sin v
t an0 = X 2 = = tanh u tan v
X,
a cosh u cos v
8 = arctan(tanh u tan v)
and z = w. Since
3r 30 3z 3z
= 0 , = 0 , = O f and = O
aw 3w au 3v
it follows that
d(r, 8, z) - 3r 38 3r 38
-
3(u, V, W)
3~ av - av au
2 2 2 2
- -
a(sinh u sec v + sech u sec v)
2 2 2 2
\, cosh u - sin v (1 + tanh u tan v)
6.4 EXERCISES
In each of Problems 1 through 5, the transformation equations
relating the rectangular cartesian coordinates x i and a curvilinear
General Tensors 421
coordinate system ui are given. Find the local basis ei, the local reci-
procal basis e
i
, and the covariant and contravariant components of
the metric tensor.
1. The spherical coordinate system:
x1 = r sin $ cos 8, x2 = r sin $ sin 8, x3 = r cos $
where u1 = r, u
2
= $, and u
3
= 8.
2. The yarabolic cylindrical coordinate system:
where u1 = v, u
2
= w, and u
3
= z.
3. The parabolic coordinate system:
where u1 = v, u
2
= w, andu3 =8.
4. The ellip tic cylindrical coordinate system:
xl = acoshvcos w, x2 = asi nhvsi nw, x3 = z
where u1 = v, u
2
= w, and u3 = z (a is a constant).
5. The curvilinear coordinate system:
6. Find the transformation equations relating the cylindrical
coordinates r, 8, z and the parabolic cylindrical coordinates v,
w, z. Calculate the Jacobian a(r, 8, z)/a(v, w, z).
7. Find the transformation equations relating the cylindrical co-
ordinates r, 8, z and the parabolic coordinates v, w, $. Calcu-
late the Jacobian 3(r, 8, z)/a(v, w, $).
8. (a) Find the transformation equations relating the cylindrical
coordinates r, 8, z and the spherical coordinates R, $, 8. Also
find the inverse transformation.
(b) Let u1 = r, u
2
= 8, u3 = z and u'l = R, u * ~ = @, ue3 = 8.
Evaluate aui / du*~ and au*' /au~ and represent them in matrix
form.
9. Let ei denote the local basis in cylindrical coordinate system.
Using the results of Prob. 8(b) and (6,66), obtain the local basis
Chapter 6
e*' in spherical coordinates. Check your results with that of
Problem 1.
6.7 GENERAL TENSORS
We now discuss tensors in general curvilinear coordinate sys-
tems. Such tensors are called general tensors or simply tensors. As
we will see here, the theory of general tensors is analogous to that of
tensors in an oblique coordinate system restricted to a point. For in
a general coordinate system, although we no longer have a fixed
basis that is valid for the whole space, it is possible as we see in Sec.
6.6 to set up at each point a local coordinate system with the tan-
gent vectors to the coordinate curves as the base vectors and the
normal vectors to the coordinate surfaces as the reciprocal base vec-
tors. Relative to these bases the entire edifice of tensor algebra that
we have developed for oblique caartesian coordinate systems can be
carried over.
First, suppose A is a vector with covariant and contravariant
components ai and a' at a point in space with respect to a local basis
ei of a given curvilinear coordinate system u
i
. Let a*i and a*' denote
the corresponding components of the vector at the same point with
respect to another curvilinear coordinate system u*' with basis e*i,
where LL' ~ = u * ~ (ul., u2, u3 ) ( i = 1, 2, 3). Then at the point in question,
we have A = a*ie*' = ajd, and so
Substituting the formula (6.66) for e*i and noting that el- ei = 8Ji, we
finally obtain the transformation law
a*. =
du'
a .
;Iu*' '
(i = 1, 2, 3) for the covariant components. Similarly, from the iden-
tity A = a*ie"i = de.3 we find
J
which, by (6.68), leads to the transformation law
General Tensors
(i = 1, 2, 3) for the contravariant components of the vector. These are
the laws governing the components of a tensor of the first order (a
vector).
For tensors of higher order, the transformation laws are pre-
cisely those given in Sec. 6.3 with respect to oblique coordinate sys-
tems. We need simply to replace the variables x
i
and x*' by ui and
u*', res pectively, to indicate general coordinate system. Thus, for
instance, a tensor of order 2 consists of 32 covariant components aii
:J
or 32 contravariant component ail or mixed components ai l and a'.,,
which transform under a change of coordinates according to the
laws:
(i) aei, =
au
P
auq
i j apq
du* au*
i au* du*j pq
(ii) a* = a
auP 5uq
.j au
P
au*' .q
(iii) a*i = a
i q P
au* au
i a u * i a u q p
(iv) a* . j =
ja .q
auP au*
An important example of a second order tensor is the metric
tensor (gij) defined by (6.57). Indeed, from the definition and the
formula (6.66), we see that
Chapter 6
so that the quantities g,, transform as covariant components of the
metric tensor.
From the results of our discussion in Sec. 6.5, we can verify
that at each point of the space the various components a,,, aij, ai.J,
and aiej of a second order tensor are related to one another by the
metric tensor at that point.
Example 1. Let $ be a scalar field in a rectangular cartesian
coordinate xi. If we change to general coordinates u
i
= u
i
(x
l
, x2, x3 )
(i = 1,2,3), the gradient of $ becomes
a$. a$ auJ . a$ ej
V$ = l i = 1. =
a x, 3 u ~ a xi '
au'
since (3uJ/3xi)ii = VUJ = d . This means that the natural expression
for V$ is in terms of the local reciprocal basis el; hence, the terms
3$ /3u1 are the covariant components of V$. In fact, under a change
of coordinates, we see that
a0 - a$ auJ
-
au*i auJ duei
which agrees with (6.70) for the covariant components of a vector.
For this reason, V$ is called a covariant vector field. Of course, V$
has contravariant components as well, given by gi~(3$ /3uJ).
It is worthwhile to note that the expression V$ = (I)@ /3u')ei is
invariant under any coordinate transformation. In fact, by (6.68), we
see that
Thus, in cylindrical coordinates r, 8, z, we have
34, 1 a$ 2 a$ 3
V$= e + e + e
3r 38 3z
where el = cos 8 il + sin 8 i 2, e
2
= (- sin 8 il + cos 8 i2 )/r, 2 = i3, as
derived in Example 1, Sec. 6.6.
General Tensors
Example 2. Find the gradient of @ = x1x3 + x2 in the curvilin-
ear coordinates v, w, z, where
Xl = vw, x2 = (v
2
+ w
2
)/2, x3 = z
Solution: From the position vector R = uv il + [(v
2
+ w
2
)/2] i2
+ z i3 with u1 = v, u
2
= w, and d = z, we find
and el. e2 x e3 = w2 - v2. Hence, the reciprocal base vectors are given
by
In the new coordinates v, w, z, we have 6 = vwz + (v
2
+ w
2
)/ 2.
Hence,
Vq = %
36 e 2+ % e3
av aw 3~
= (wz + v)el + (vz + w)e
2
+ vwe3
It is instructive to observe that if we substitute the expressions
of the vectors el, e
2
, e3 in V@ and collect the coefficients of il , i 2, iO ,
we find
V@ = x3il + i2 + xl i3
which is precisely the gradient of + in rectangular cartesian coordi-
nates.
Chapter 6
Example 3. Consider the coordinate transformation
u*' = ui(ul, u
2
, u
3
) (i = 1, 2, 3). Taking the differential, we find
i au*i
du* = du '
3u
j
By (6.71) this shows that the quantities dul transform as contravar-
iant components of a vector. This vector is exactly the differential
dR of the position vector of R. Note that
Thus dR is sometimes called a contravariant vector. It follows that
the inner product of VQ, and dR is a scalar:
Example 4. The Kronecker delta 69 is a mixed tensor of sec-
ond order (or, more precisely, forms the mixed components of a
second order tensor). Indeed, under any transformations of coordi-
nates u1 to u*' , we know that
Hence,
Further, we see that the Kronecker delta has the same components
in all coordinate systems. Such tensors are called i sot ropi c tensors.
Notice that since the metric tensor is symmetric, that is, gij = gji
and g i ~ = g ~ i , it follows that
General Tensors
Example 5. The components of a second order cartesian tensor
are given by
Find the covariant, contravariant and mixed components of the ten-
sor in cylindrical coordinates.
Solution: The transforma tion equations relating the cylindri-
cal coordinates r, 0, z and the rectangular cartesian coordinates xl ,
x2, x3 are given by
xl = r cos 0 , x2 = r sin 0 , x3 = z
and the inverse transformation are given by
Let us set xi = u
i
(1 5 i 5 3) and r = uxl, 0 = u' ~, z = u
x3
. Then the
different values of 3ui/au*i and duxi/dUj are represented by the
following matrices (the index i indicating row):
cos 0 -r sin 0 0
( : 3[ s : 0 r r . 0 :]
cos 0 sin 0
au
Hence, by [6.72(i)], and dropping all zero terms, we find
Chapter 6
au
L
au
L
2
a* 12 =
a22 = sin 8(r cos 8)(r sin 8)
au* au*
3 3
= r cos 8 sin 8
1 3
au au 2
a*17 = a = (COS 8)(r cos 8)z = rz cos 8
au* ' au* -
2 2
3u au 2
a*21 =
2 1
aZ2 = (r cos @)(sin 8)(r sin 8)
au* au*
3 3
= r cos8 sin 8
2 2
au au 2 2 4 2 2
a*2L =
2
2a2 = (r cos8) (r sin 8) = r sin 8 cos 8
au* au*
1 3
au au 2
a*, =
2
a 13 = (-r sin 8) (r cos 8)z = -r z cos 8 sin 8
au* au*'
3 1
au au
" * M = a.il = (cos 8) (r sin 8)z = rz cos 8 sin 8
au*33u*
3 1
au au 2 2
as2 = = (-r sin 8) (r sin 8)z = -r z sin 8
au* " u'
On the other hand, by [(6.72(ii)], we find (note t hat aij = aij)
1 1
11 au* au* z
a* =
2 2 2 4
2 2
a = (sine) (r sin 8) = r sin 8
au au
General Tensors
12 au*1au*2 22
a* =
cos 0 2
2 2
a = ( si ne) (r sin 0)
au au
r
= r cos 0 sin30
1 3
13 au* au* 13
a* =
2
a = cos 0 (r cos 0 )z = rz cos 0
1 3
au au
= r cos 0 sin30
2
n a u*2a u*2a n=( c or e )
a* =
2 2 2
(r sin 8) = cos 0 sin 0
2 2
au 3~
2
n au* aU*3 13 sin
a* = a = ( - - ) ( r c os 8) z =- z c os 0s i n0
1 3
au au
r
3 2
3 au* au* 31
a* =
sin 0 2
a = (- ) (r sin 8)z = - z sin 0
3 1
au au
r
The components ~ * I J may also be obtained from the formula (6.46).
Next, to find the components P i s k and a*'.k, we use the rela-
tions
g*, a *~kf = R* i ~ a;k
430 Chapter 6
from (4.47) and (4.48). The quantities g*,, and g*l~ of the metric tensor
in cylindrical coordinates are obtained in Sec. 6.6, Example 2. Thus
we find
a*,.k = g * agk = geii a*ik
IJ
so that
for k = 1, 2, 3. Similarly, we have
= g*!' a*jk = g*ll a*ik
so that
for k = 1,2,3.
6.5 EXERCISES
1. Show that the quantities
3 - I - J
i j rlu rlu
g" = v u v u = C
k = l
axk a xk
transform as contravariant components of the metric tensor.
2. From the definitions of gij and gij, verify that
3. The components of a cartesian vector field are given by
Find the covariant and contravariant components of the vector
in (a) parabolic cylindrical coordinates, (b) spherical coordi-
nates, (c) elliptic cylindrical coordinates. [Note: Use formula
(6,42) for determining a" and the fact that if the matrix (g,,) is
diagonal, i.e., g,, = 0 for i t j, then (giJ) consists only of l / gi i
along the diagonal.]
General Tensors 431
4. The components of a cartesian tensor are given by al = xlx2,
+ = 2x2 - ( x ~ ) ~ , a3 = (xl)I. Find the covariant and contravar-
iant components of the vector in (a) parabolic coordinates,
(b) spherical coordinates.
5. The components of a cartesian tensor field of order two are
given by
Find the covariant, contravariant, and mixed components of
the tensor in (a) cylindrical coordinates, (b) spherical cocxdi-
nates, (c) arabolic coordinates.
t:
6.
Let 4 = xl- + x22 + x32 . Find the gradient of Q in (a) parabolic
coordinates, (b) spherical coordinates.
7. Find the gradient of @ = xlx2 + ~ 2 x 3 + x1x3 in (a) parabolic cy-
lindrical coordinates, (b) spherical coordinates.
8. Find the covariant and contravariant components of the carte-
sian tensor
in cylindrical coordinates and spherical coordinates.
9. Let aljk denote the components of a third order tensor in a
coordinate system u'. Show that if
aijk = aikj
(or aijk = - aikj )
then
= a"kj
(or a*ljk = - a*lkj )
in any coordinate system u*'.
10. Let bgk denote the components of a third order tensor in a
coordinate system u
l
. Show that if
then
Chapter 6
11. Show that a\k = ajik does not imply = a*jil..
(The last three problems show that the symmetric and skew-
symmetric property of a cartesian tensor is also true for general
tensors provided that the indices to be interchanged are either
both covariant or both contravariant.)
6.8 COVARIANT DERIVATIVE OF A VECTOR
Let A be a vector field with components ai in a rectangular car-
tesian coordinate xi with orthonormal basis ii (1 5 i r 3). The par-
tial derivative of A with respect to xk (1 < k 5 3) is given by
since the basis vectors ii are constants. Thus, for each value of the
index k, the derivative 3A/dxk is a vector field with components
3ai/dxk. In fact, as we saw in Sec. 5.9, the nine quantities aai/axk (i, k
= 1, 2, 3) are the components of a second order tensor called the ten-
sor gradient of A. It is clear from (6.73) that
Likewise, if A is a vector field with covariant component ai or
contravariant components a
i
in an oblique coordinate system x
i
with
the basis ei and reciprocal basis e
i
, its derivative with respect to x
k
(k
= 1, 2, 3) is given by
As in the case of a cartesian tensor, it can be shown that the quantities
General Tensors
aai/axk and aai/axk are the covariant and mixed components of a
second order tensor (Problem 5). We deduce from (6.75) that
Now let us consider a vector field A in a general coordinate
system ul. At each point in space we can write
where ai and a' are the covariant and contravariant components of
A, and ei and e
i
are the local and reciprocal bases. Taking the partial
derivative with respect to u
k
, we find
The derivatives aei/aui and aei/aui are no longer zero since the base
vectors now depend on the coordinates u
i
of the point. Therefore,
the rate of change of A involves not only the rate of change of its
components (aai/aui, aai/aui ), but also the rate of change of the local
bases (aei/aui, aei/aui). In analogy to (6.76), the two kinds of com-
ponents of the derivative a ~ / a u ~ , which we denote by ai, and a
i
, k ,
are defined by
and
Chapter 6
respectively. The comma in the notation ai, or ai, k indicates diffe-
rentiation with respect to the coordinate u
k
. We will show that the
quantities ai, or a', k are components of a second order general ten-
sor called the covariant derivative of the vector A. More precisely,
we will show that ai, !, are the covariant components of the covariant
derivative of A = aiel, while a', k are the mixed components of the
covariant derivative of A when it is represented by its contravariant
components, i.e., A = a
i
e
i
. It is clear that in rectangular cartesian
coordinates the covariant derivative of a vector field is just the ordi-
nary derivative (6. 73).
Christoffel Symbols
To facilitate the calculation of the covariant derivative of vec-
tor fields and tensors in general, we now introduce the so-called
Christoffel symbols. First, let [if jk] denote the dot product of ei and
3ej / auk, that is
This notation is called the Clr ristoffel s!j~rzbol of the first kind. If we
multiply both sides of (6. 80) by gri and take the sum with respect to i,
noting that gri ei = erf we obtain
ri ri aej
g [i, jk]= g ei e
k
The notation on the right-hand side of (6.81) is called the Chri st of f rl
synr bol of the second kind. Since gli = g ~ k = tiik, it follows that
Thus the Christoffel symbols are related to each other by the metric
tensor. Moreover, if we assume that the coordinate transformation
General Tensors 435
equations are twice continuously differentiable, then we have
Hence the Christoffel symbols [i, jk] and li j k} are both symmetric
with respect to the indices j, k, that is, [i, jk] = [i, kj] and l i j k] = {i kj}.
However, the Christoffel symbols do not transform as components of
a tensor, as we shall see later.
Now from the fact that ei. e~ = ~ i , it follows that
so that
If we multiply both sides of (6.84) by gk and sum with respect to i,
noting that girei = e
r
, we find
Thus we obtain the various dot products of the bases ei and e~ with
their derivatives aei/auk and &,/auk.
Using (6.84) and (6.81), we can therefore write (6.78) and (6.79)
as
and
Chapter 6
From the definition (6.78) and (6.79), we observe that
ri 3A ri ;)A r r
g ai , k = k . ( ~ e i ) = . e =a , k
k
(6.88)
au au
and, conversely,
These relations are to be expected since ai,k and a' k are the (covaria-
iant and mixed) components of the covariant derivative of the same
vector A.
The results of the various dot products of the bases ei and d
with their derivatives 3ei/3uk and 3d/3uk are shown in the follow-
ing schematic diagram:
where the double-headed arrow points out the two vectors whose
dot product is equal to the Christoffel symbol written beside the
arrow.
Example 1. Find the values of the Christoffel symbols [i, jk]
and {i jk) in cylindrical coordinates.
Solution: Set u1 = r, u
2
= 8, and u3 = z. From Example 1 of
Sec. 6.6, we have
General Tensors 437
el = cos 0 il + sin 0 i 2, e2 = - r sin 0 i1 + r cos 0 i 2, e, = i3
Hence we find
Therefore, by (6.80), we see that among the 27 possible values of
[i, jk], only three are nonzero and they are given by
From (6.81) it follows that
Example 2. The covariant components of a vector field A in
cylindrical coordinates u1 = r, u
2
= 0, and u
3
= z are given by
al = 2 sin 0, a2 = r cos 0, a3 = z
Find the convariant and mixed components of the covariant deriva-
tive of A.
Solution: We calculate a,, k, a2, k, ag, for k = 1, 2, 3. From
Example 1, we recall that
Chapter 6
and li . } = 0, otherwise. Now since al = 2 sin 0, which depends
~k
only on 0, we find
a*, 1 = 0, a1,3 = 0
Similarly, we find
1
a2, = 2 - {221) a2 = cos 0 - ( ) ( r cos 8) = 0
r
These are the convariant components of the covariant derivative of
the vector.
To find the mixed components, we use the relation (6.88).
Since
gn = 1, g22 = l / r
2
, g33 = 1, and gij = 0 for i + j
we obtain
1 22
a , 2 = al, 2, a2,1 = g a2,1 = 0
a
2
, 2 = (l/r
2
) a2, = (sin @)/I-,
a3, 3 = a3,
while all the other components are zero.
Example 3. Determine the Christoffel symbols for the gene-
ral coordinates given in Example 2, Sec. 6.7.
Solution: The general coordinates are defined by the equa-
tions
XI = vw, X2 = (v2 + w2 )/ 2, x3 = z
Setting u1 = v, u
2
= w, u
3
= z, we find from Example 2, Sec. 6.7
General Tensors 439
and
with g.- = 0 and g i ~ = 0 for the rest of the values of the indices i, j .
'J
Therefore, by (6.80), we have
and [i, jk] = 0 for all the other values of i, j, k.
It follows from (6.81) that
and {I jkl = 0 for all the other values of i, j, k.
6.9 TRANSFORMATION OF CHRISTOFFEL SYMBOLS
We observe that the Christoffel symbols [i, jk] and, l i j k} are zero
440 Chapter 6
in all cartesian (rectangular or oblique) coordinate system since in
any such system the base vectors are constant. However, as we saw
in Example 1, Sec. 6.8, they do not all vanish in cylindrical coordi-
nates. This implies that the Christoffel symbols cannot possibly be
components of a tensor because a tensor that is zero in any particu-
lar coordinate system must be zero in all coordinate systems. By
considering the transformation laws for [i, jk] and {l ik}, we can also
see the non-tensorial character of these symbols. ~e' i l l ust r at e this
for the Christoffel symbol of the second kind, {l jk}.
Under a transformation of coordinates, say from u1 to u*' de-
fined by equations of the form (6.63), we have
Because ~f the presence of the second term in (6.90), the Christoffel
symbol {l ik} does not transform as components of a tensor.
using the transformation law (6.90), we can now show that the
quantities ai, and a', k transform as covariant and mixed compo-
nents of a second order general tensor, which is called the covariant
derivative of the vector field A. Indeed, under a change of coordi-
nates, we know that
and, hence, by (6.90) we have
General Tensors
Since (3u*~/au~)(au"/au*~) = ti;, this equation simplifies to
Therefore, the quantities ai, are components of a second order ten-
sor.
It follows from (6.88) and the tensorial character of the inner
product of two tensors that the quantities aitk are mixed components
of the same second order tensor which is the covariant derivative of
the vector field A. Of course, this fact can also be verified by exhib-
iting the transformation law for the quantities aifk. This is requested
in the exercises (see Prob. 13).
To obtain further insi ht into the significance of the compo-
nents ai, and aitk of aA/J$, we derive formulas for J A / ~ U ~ that
resemble those given in (6.75). First, from (6.84), we deduce that
This says that the quantities - {J ik} are the expansion coefficients of
deJ/3uk with respect to the reciprocal basis e
i
. Subsitituting this in
(6. 77a), we find
Chapter 6
in view of (6.86). Similarly, from (6.81) we deduce
Substituting this in (6.7713) leads to the formula
Equations (6.92) and (6.93) imply that the quantities ai, and aiYk are
precisely the components of 3 ~ / r ) u ~ with respect to the bases e
i
and
ei , respectively.
We conclude this section with the derivation of the relation of
the Christoffel symbols with the metric tensor. Using the fact that
aei/auk = &,/hi , we see that
But
General Tensors
therefore, we obtain
By (6.81) we further obtain
Hence, if the gii are constants, the Christoffel symbols vanish iden-
tically, and thus the covariant derivative of a vector field reduces to
the ordinary derivative. This is certainly the case when we are in a
cartesian coordinate system.
Using (6.94) we readily obtain
= [i, jk] + 0, ik]
k
which, in view of (6.82), can be written as
We can also derive a formula analogous to (6.97) for the partial
derivatives of the contravariant components gi, of the metric tensor.
First, from the identity gij g ~ k = 62, we obtain by differentiation
Now multiply both sides of this equation by and note that
Chapter 6
gij g
n
= 6jr. We obtain
where we have made use of (6.96). By (6.81) this can be written as
or, changing indices,
6.6 EXERCISES
1. If A = xlx2 il + ~3x2 i2 + xlx3 i3, find the derivatives of A with
respect to xl, x2, x3 , and write down the matrix (3ai/3xj).
2. Repeat Problem 1 when
A = xl cos x2 i , + x2 sin x3 i2 + x3 exp (x3) i3
3. Let A = x2x3 el + x2 cos x1 e2 - (1 + x
3
) e? , a contravariant vec-
tor in an oblique coordinate system x
i
with basis ei. Find the
derivatives of A with respect to the coordinates x
i
(1 < i r 3)
and write down the matrix (3ai/dxj).
4. Repeat Problem 3 for the covariant vector A = sin x
2
e1 +
x1 exp(x2) e
2
+ x
2
x3 e3
5. Let ai and a1 denote the covariant and the contravariant com-
ponents of a vector field in an oblique cartesian coordinate sys-
tem. Show that the quantities 3ai/3xk and 3ai/dxk are the co-
variant and mixed components of a second order tensor.
In each of Problems 6 through 9, determine the Christoffel
symbols [i, jk] and {l jk} in the given coordinate system.
6. The spherical coordinates u1 = r, u
2
= @, u
3
= 8, where
General Tensors
x1 = r sin Q cos 8, x2 = r sin Q sin 8, x3 = r cos Q.
The parabolic cylindrical coordinates u1 = v, u
2
= w, u" z,
where
XI = (v
2
- w2)/2, X* = VW, X3 = z.
The parabolic coordinates u1 = v, u
2
= w, u3 = 8, where
xl = vw cos 8, x2 = vw sin 8, x3 = (v
2
- w2)/2.
The general coordinates u1 = u, u
2
= v , u
3
= w, where
XI = uv + 1, x2 = vw - 1, X? = (u2 + w2)/2.
The covariant components of a vector field A in spherical coor-
dinates are given by
al = r sin
2
Q cos
2
8, a2 = r
2
sin Q cos Q cos
2
8,
" = r si n+ cos Q cos 8.
Find the components ai, and a', k of the covariant derivative
of A.
The covariant components of a vector field A in parabolic cy-
lindrical coordinates are given by
Find the components a,, and a', k of the covariant derivative
of A.
The covariant components of a vector field A in parabolic co-
ordinates are given by
al = w cos 8 + v, a2 = v cos 8 - w, a3 = - vw sin 8.
Find the covariant and mixed components of the covariant
derivative of A.
Show that the mixed components a
i
, k of the covariant deri-
vative of a vector A transform under any changes of coordi-
nates according to the law
Show that the Christoffel symbol of the first kind transforms
under any changes of coordinates according to the law
Chapter 6
15. Deduce from (6.80) and (6.84) the formulas
16. Show that
agJk = [i, j k] - F, i j ]
-
k
au aul
17. Show that in any orthogonal coordinate system (gij = Of if i t j),
the Christoffel symbol {iik} vanishes whenever if j, k have
distinct values.
18. Noting that the covariant derivative of a scalar field is precise-
ly the ordinary derivative, show that
where ai and b
i
are components of two vector fields. (Hint:
Differentiate; then add and subtract the term ai{ijk}bJ, observ-
ing that aili jk}bJ = a,{~ r}b'.)
6.10 COVARIANT DERIVATIVE OF TENSORS
The concept of covariant derivative of a vector field can be ex-
tended to tensors of higher order. First, we observe that if @ is a sca-
lar field, its covariant derivative is simply the ordinary derivative,
that is,
This is to be expected since the changes in the base vectors ei and el
General Tensors
have no effect on the field itself. Thus the covariant derivative of @
is precisely the covariant vector grad +. Next, let us consider a sec-
ond order tensor and suppose that a, . , ail, and ai.j are its covariant,
'J
contravariant, and mixed components, respectively. Then the cor-
responding components of its covariant derivative are given by
It can be readily shown that these quantities transform as compo-
nents of a third order mixed tensor under any transformations of
coordinates. In particular, the quantities transform as covariant
components, the as mixed components with two contravariant
indices, and the ai-jrk as mixed components with two covariant indi-
ces.
From the above definition it should be evident how the com-
ponents of the covariant derivative of a tensor of order r may be
defined. It consists of r +l terms, the first term being the partial
derivative of the components of the tensor with respect to the coor-
dinate, say u
k
. The remaining r terms are sums of quantities resem-
bling components of inner products of the tensor and the Christoffel
symbols of the second kind with the differentiation index k appear-
ing as one of the subscripts in the Christoffel symbol. Each index in
the components of the tensor that is replaced by a summation index
gets transferred to the unoccupied spot of the character (subscript or
superscript) in the Christoffel symbol. Finally, each of the remain-
ing r terms has a plus sign i f the summation is with respect to a con-
travariant index (superscript) and a minus sign if it is with respect to
a covariant index (subscript) of the components of the tensor. It can
be shown that the covariant derivative of a tensor of order r is a
tensor of order r + 1, where the increase in the order occurs as a
covariant index.
Chapter 6
Example 1. If ai.Jnl are the mixed components of a third order
tensor, then
. , aa; '"
-
ai,k - p 3u k - {.' 1 k }a;i111 + { r k j }.I "" + {m}a;Jr r i
are the mixed components of the covariant derivative of the tensor.
Example 2. Consider the mixed tensor defined by the Kronec-
ker delta 6;. By (101) we see that
Thus the mixed tensor defined by the Kronecker delta behaves like a
constant with respect to covariant differentiation.
It is easy to show that the rules for covariant differentiation of
sums and products of tensors are identical with those used in ordi-
nary differentiation. The verification of the case involving sums is
trivial. We illustrate the proof of the case involving products with
two particular examples. First, consider the components tij = a.b- of
1 J
a second order tensor, which is the outer product of two vectors
with components ai and bj. Then, in accordance with (6.99), we have
at
ti,, k = k
3~ - {;>trj - {jrk)tir
Expanding the first term and rearranging terms, we find
General Tensors
which is the result we desired.
Next, let us consider the components ai = ti,bj of the inner pro-
duct of a second order tensor and a vector. We find
But from the expressions of ti,, and b ~ , k, we have
and
Substituting these in the previous equation, we observe that all the
terms involving the Christoffel symbol cancel out, yielding the
desired result
(tij bJ ), k = tij, k tr' + ti, w, k
Ricci's Theorem
We conclude this section with the verification of the fact that
the metric tensor behaves like a constant with respect to covariant
differentiation, that is, gij, = 0. This is known as Ri cci ' s theorem.
Indeed, by definition, we find
450 Chapter 6
In view of (6.97), the right-hand side of this equation vanishes iden-
tically. Thus, we have
gij, k = 0.
Similarly, by (6.100) and (6.98), we obtain
As a consequence of Ricci's theorem, we see that the compo-
nents gij or giJ of the metric tensor can be regarded as constant fac-
tors with respect to covariant differentiation. For example, we have
and
6.11 GRADIENT, DIVERGENCE, LAPLACIAN, AND CURL IN
GENERAL COORDINATES
The results obtained in the preceding sections provide us with
tools for expressing a number of physical laws in forms that are va-
lid in any coordinate system. We illustrate this by expressing the
familiar concepts of gradient, divergence, Laplacian, and curl in ten-
sorial forms. First, we recall from Sec. 6.7 (Example 1) that the gra-
dient of a scalar field $ in general coordinates u' is given by
From this we deduce that the del operator V has the form
Now let A be a vector field represented by its contravariant compo-
nents a'. Then, by definition, the divergence of A is given by
General Tensors
i 3A
d i v A= V. A= e .
a u
i
But from (6.93) we have
3A j
= a . e .
i I ' J
au
hence (6.103) becomes
This says that the divergence of A is the contraction of the mixed
components of the covariant derivative of A. In rectangular carte-
sian coordinates (a
1
= ai, u' = xi), this reduces to the familiar form
div A = 3ai/dxi, which is the contraction of the tensor gradient
dai/dxj.
Writing the term al,i explicitly, we find
i
div A = " i + {ii)ak
3u
This can be written more concisely by calculating the Christoffel
symbol {I i k} . In fact, since g
ir
= gri, we have
By adding the two equations
and
we find
452 Chapter 6
Now consider the determinant g of the metric tensor (gij). Ex-
panding by the elements of the i-th row (i = 1, 2, 3), we find
g = g 11 Gil + g , , ~ i 2 + gi3ci3
(i = l , 2, 3)
= gi, Gir (sum on r for fixed i; no sum on i)
where cir is the cofactor of the element gir of the matrix (gij). Differ-
entiating this with respect to gij, noting that 3gi, /3gij = 6i and
G"agii = 0, we find
Thus, by the chain rule, we have
Since G~ J = g gij, it follows that
and so (6.106) becomes
Substituting this in (6.104), we finally obtain
div A = + {i t ) a k
aU '
General Tensors 453
This is the general expression of the divergence of a vector field A in
any coordinate system u
i
in terms of the contravariant components
a'. Because a
i
= giJ a,, the expression (6.108) can also be written in
terms of the covariant components ai as
To derive the expression of the Laplacian in general coordi-
nates ui, we observe that grad $ = e
i
a$ /aui, so that a$ /aui are the
covariant components of grad $. Therefore, from (6.109) we obtain
This is the expression of the Laplacian of a scalar field $ in general
coordinate system ul.
Finally, to obtain the formula for the curl of a vector field in a
general coordinate system u
i
, suppose A = a
i
e
i
where we assume
that the local basis forms a right-handed triple at each point. Then,
by definition and by (6.92), we have
i aA
curlA = Vx A = e X- T
aU '
i j i j
= e x(ajIie ) = a j i ( e x e )
But from (6.5) and (6.51), we note that
where i, j, k are cyclic permutations of 1, 2, 3. (Note that if the local
basis is left-handed, a minus sign appears in this equation.) There-
fore, (6.111) becomes
1
curl A = ---(a. . - a . ) e k
J,' ' J
(6.112)
Since {k ij) = {k
it follows that
Chapter 6
and so (6.112) simplifies to the form
curl A =
This formula may be conveniently written in the determinant form
which resembles the formula in rectangular cartesian coordinates.
curl A =
\ g
Orthogonal Coordinate Systems
el e 2 e3
1 d 3 a
1 - 2 - 3
3u Ou clu
a1 a2 a3
It is instructive to note that when the general coordinate system
is orthogonal, that is, the base vectors ei (1 2 i 5 3) are orthogonal at
every point, the various formulas we have derived above reduce to
the corresponding formulas we obtained in Sec. 3.9. To see this, we
must express the vector field in terms of its so-called physical com-
ponents. These are simply the components of the vector field when
referred to a normalized basis. First, we observe that when the local
basis ei forms an orthogonal set, the reciprocal basis el is given by
Thus if @ is a differentiable scalar field, then by (6.102) its gradient
becomes
General Tensors
in terms of the unit base vectors ui = ei /hi (i = 1, 2, 3) . The quanti-
ties
'I' , i = I, 2,3 (no sum on i)
h i . i
r l u
are called the physical components of the gradient vector VQ. It is
obvious that (6.115) coincides with (3.53).
Next, in orthogonal coordinates we note that gij = hi2 (no sum
on i ) and g,, = 0 for i t. j , so that g = det (gij) = (hlh2h3)2 or dg =
hlh2h3. Now consider the vector field A = a' e,. In terms of the
normalized basis ui = ei / hi (I 2 i 5 3), we can write this vector field
as
Thus the physical components of the vector field A are given by at i
= a" = a
i
h
i
(no sum on i) for i = 1, 2, 3, so that a' = al ' / hi . Hence, in
terms of the physical components, we obtain from (6.108) the formu-
1 a
1
div A =
This coincides with (3.59) of Sec. 3.9. If A = VQ, so that at i = ( I / hi )
3@ /dul, then (6.116) yields the Laplacian
456 Chapter 6
in accord with (3.60) of Sec. 3.9.
Finally, we consider A = aiel. By (6.114) we observe that
where a
t
i
= a' ' are the physical components defined by a t i = ai /hi
(i = 1, 2, 3) . Thus, in terms of the physical components a', and the
normalized basis u,, we have ai = a
i
t
h
i
and ei = h,u,. Substituting
these in (6.113), we obtain
which coincides with (3.61) of Sec. 3.9.
1
curl A =
h l h 2 h 3
6.7 EXERCISES
h1u1 h 2 ~ 2 h 3 ~ 3
3 3 3
1 2
du au au
3
a t l h l a' 2h2 "' 3h3
1. Let a,,,,, a',"', aij,, denote the various kinds of components
of a third order tensor. Write down the formulas for the cor-
responding components of the covariant derivative of the
tensor.
2. Write down the formula for the components of the covariant
derivative of a fourth order mixed tensor defined by the com-
ponents alJFq.
3. Verify that the components a ,j,k and ailrk transform as comyo-
nents of a third order tensor.
4. Verify the following formulas:
(a) (a,, + b,j),k = a,,, + b.- I J, k
(b) (0 ai),k = @ ai,k + ai(3@ /auk), where f is a differentiable
scalar field.
(c) (a" bj),k = aiJfk bj +
b, J f k
(d) (&ar. j),k = Rirar.j, k
General Tensors
5. Show that
aij, k = R ipg jq aPq ,k
and
a*,, k = g ' pgJqam,
6. Show that
(ai,bln), k= a,,, bm + aij blll,k
and
( a b c,,,), = ai, bj cm + a b
cIn + a b. cIn,
1 J I J, k 1 J
7.
Show that (6.106) can be written as
8. Suppose that a coordinate system u
i
is such that gij = 0 for i t j,
show that
(a) (jij} = - 1 %,,
'gii au '
where the summation convention is not applied and i + j .
Hint: Use (6.105).
9. Prove that . , . .
( ~ ~ ~ a ' b ) ) , ~ = air kbl + a' bi, k
10. Prove that the covariant derivative of the magnitude of a vec-
tor field A is given by
Hint: Consider I A l = (gij aiaj)'/* .
Chapter 6
11. The contravariant components of a vector field A i n spheri-
cal coordinates u1 = r, u
2
= Q, u
3
= 0 are given by
a' = r
2
sin
2
@ sin 20, a2 = sin 29 cos 0, a3 = r cos 2Q.
Find the physical component, the divergence, and the curl of
the vector.
12. The covariant components of a vector field A in spherical
coordinates u1 = r, u
2
= Q, u3 = 0 are given by
a, = r
2
sin
2
241 sin 0, a2 = sin
2
@, a" r ccos @.
Find the physical cc)mponents, the divergence, and the curl of
the vector.
6.12 EQUATIONS OF MOTION OF A PARTICLE
In this last section, we illustrate an application of covariant
differentiation by analyzing the motion of a particle. The particle is
assumed to be moving in the three dimensional Euclidean space and
that its position at a given time t is referred to some curvilinear coor-
dinates ul.
So let the curvilinear coordinates of the particle at time t be
given by u
i
= u
i
(t) (i = 1, 2, 3), where the functions u
i
(t are twice
1
continuously differentiable for t > 0. Let R = R(ul, u
2
, u- ) denote the
position vector of the particle at time t. Then the velocity of the par-
ticle is given by
where
j
v (t) =
du '(t)
dt
and ej = aR/duJ ( j = 1, 2, 3) is the local basis. The acceleration is given
by
General Tensors
But from (6.93) we note that
hence (6.120) becomes
where
The Christoffel symbols appearing in (6.122) must be calculated
from the metric tensor associated with the coordinate system ul.
The expression on the right-hand side of (6.122) is commonly
written as 6vJ/6t. Thus we can write
dv - 6vJ
-
dt 6t
ej
where fivJ/6t = vJtk du
k
/ dt . This is known as the absolute or intrin-
sic derivative of the velocity vector v = vjei. The process of intrinsic
differentiation can be extended to tensors of higher order (see Pro-
blem 1).
Now let m be the mass of the particle and F = F J ~ ~ the force act-
ing on the particle. By Newton's second law of motion, we have
F J ~ , = ma~e,, which implies that
In rectangular cartesian coordinates, the right-hand side reduces to
460 Chapter 6
m(d
2
x/dt
2
), and FJ denotes the physical components of the force. By
definition, the amount of work done by the force in displacing the
particle through a displacement dR is given by the scalar product
Thus the work done by the force in displacing the particle from a
point PI to a point P2 is given by the line integral
Since 9 = gijvivJ is a scalar, it is clear that d@ / dt = 69 /6t. Moreover,
it can be shown (Problem 1) that
Hence (6.124) becomes
where
1 i j l 2
T = rng. . v v = ml vl
2 'J 2
is the kinetic energy of the particle. Therefore, the work done by the
force in displacing the particle from PI to P2 is equal to the change
General Tensors 461
in the kinetic energy of the particle as it moves from PI to P2. (Cf.
Example 5, Sec. 4.3.)
Example 1. Find the expression for the kinetic energy and the
contravariant components of the acceleration vector in cylindrical
coordinates.
Solution: We recall that in cylindrical coordinates u1 = r, u
2
=
0, u
3
= z, the covariant components of the metric tensor are
g = r
2
, g = 1,
gij = 0 for i t j
Hence, by (6.127), the kinetic energy is given by
where the dot indicates derivative with respect to time t.
Next, in cylindrical coordinates the only nonzero Christoffel
svmbols are
Therefore, by (6.122), the contravariant components of the accelera-
tion vector are
1 ..
. 2 2 .. 2 , .
3 . .
a = r - r e . a = @ + r e , a = z
r
If the equations expressing r, 8, and z as functions of time t are
known, then the kinetic energy and the components of the accelera-
tion vector can be calculated explicitly from the formulas given
above.
Example 2. Find the expression for the kinetic energy and the
contravariant components of the acceleration vector in the general
coordinates v, w, z defined by the equations
Xl = vw, x* = (v
2
+ w
2
) / 2 , XJ = z
(Cf. Example 3, Sec. 6.8).
Solution: From Example 3 of Sec. 6.8, we recall that
462
and
Chapter 6
with gij = 0, {' i k) = 0 for all other values of i, j , k. Hence, by (6.127),
we find
and, by (6.122), we have
1 - . 1 . 2 . . . 2
a = v +
2 2
(vv - 2wvw+vw )
v -W
2 . . 1 . 2 . . . 2
a = w -
2 2
(wv - 2vvw + ww )
v -W
The equations of motion (6.123) can be given in another form
phrased in terms of the kinetic energy of the particle. We consider
the expression (6.127) for kinetic energy
Differentiating this with respect to u
i
and u
i
, treating u
i
and u
i
as
independent variables, we find
General Tensors
and
Differentiating (6.129) with respect to t by the chain rule, we obtain
If we subtract from this the expression (6.128) and note that
we find
In view of (6.122) this can be written as
where Fi denote the covariant components of the force field. These
equations are known as LuCqr un~a1s equations of motion.
Now if F is a conservative field, then we know there exists a
scalar field $ such that F = VQ. so that Fi = 3Q/3u1 (1 5 i < 3). It is
common among physicists to set V = - $ and call V the potential
energy of the particle. Doing this, Lagrange's equations (6.130)
Chapter 6
become
Since the potential energy V depends only on the coordinates ul, we
can write the equations in the form
where L = T - V. The function L is known as the Lugrungian func-
tion.
Example 3. Find the equations of motion of a pendulum bob
(mass m) that is suspended from an extensible string, neglecting the
potential energy stored in the string.
Solution: Let us introduce spherical coordinates as shown in
Fig. 6.6, and let u1 = r, u
2
= $, u
3
= 0. Then the components of the
metric tensor are given by
811 = 1, g22 =r
2
, g33=r 2si n2$, g . = O 1~ when i # j
Hence, the expression for the kinetic energy is given by
1 1 . j . 2 2 e 2 2 2 o 2
T = mg. . uu = I m(r + r $ + r sin $8 )
2 ' J 2
Let us choose the origin as the point where the potential energy
is zero. Then we have
V = - mgr cos $
Thus (6.131) leads to the following equations:
. 2 . 2
r - r @ - r 0 sin + =gcos$
" . . 2
r @+2r $- r e si n@cos$ =- gs i n$
n
d T L 2
--(r 0 sin $) = 0
dt
General Tensors
Fig. 6.6 Motion of a pendulum bob.
Now if the motion is confined to a plane, then de / dt = 0 and
the equations of motion simplify to
. 2 "
r - r@ = g cos @, r@ + 2r@ = - g sin @
If, in addition, the string is unstretchable so that dr / dt = 0, then we
obtain
- g
$ + - - sin @ = 0
r
which is the familiar equation of the simple pendulum. For small
angles of oscillation, the equation is linearized by replacing sin @ by
@, in which case the motion becomes simple harmonic. For large
oscillation, the solution of the equation is given in terms of elliptic
functions.
6.8 EXERCISES
1. Let aij, a", denote the covariant, contravariant, and mixed
components of a second order tensor, respectively. The corres-
ponding components of the intrinsic derivative of the tensor
are defined as follows:
Chapter 6
Show that
(a) 6gij/6t = 0, 6gi ~/ 6t = 0, where g,, and g ' ~ are the covariant
and contravariant components of a metric tensor;
(b) 6/6t(gij a
i
b ~ ) = gij[ai (6b~/ 6t ) + b~ (6ai/6t)], and thus deduce
the equation (6.125).
2.
Let al(ul, u
2
, u3, t) be the component of a vector field. Show
that
i
6ai - 3a + a i duJ
-
3t , J dt
and that 6ai/6t (1 < i 2 3) transform as contravariant compo-
nents of a vector field.
3. Find the expression of the kinetic and the covariant and contra-
variant components of the acceleration in a (a) parabolic coor-
dinate system, (b) parabolic cylindrical coordinate system.
4. Show from Lagrange's equations that if a particle is not subject-
ed to the influence of forces, then its trajectory is a straight
line given by xi = ait + bi (1 < i 2 3), where the ai and bi are
constants and the xi are rectangular cartesian coordinates.
5. The kinetic and potential energy of a particle moving in a con-
stant gravitational field are given by
1 . .
T = mxi xi and V = m p 3
2
where x,, x2, ~3 are rectangular cartesian coordinates with the
positive x3-axis directed upward. Find the equations of mo-
tion of the particle.
6. Prove that if a particle moves so that its velocity is constant in
magnitude, then its acceleration vector is either orthogonal to
the velocity vector or i t is zero. (Hint: Compute the intrinsic
derivative of I v 1 * = gvi vJ. )
1J
7. A particle slides in a frictionless tube which rotates in a hori-
zontal plane at a constant angular speed o. If the only horizon-
tal force is the reaction F of the tube on the particle, find the
equations of motion of the particle. Hint: Introduce polar
General Tensors
coordinates and note that
SOLUTIONS TO SELECTED
PROBLEMS
EXERCISES 1.1
1. [3,-31, IPQ 1 =3d2 2. [4,2], 1 PQ 1 = 245
3. [-2, -41, 1 PQ 1 = 245 4. [3,-2,1], IPQ I = d l 4
5. [-I, -6, -31, 1 PQ 1 = 446 6. [2,1,2], 1 PQ 1 = 3
7. Q:(-1, l) 8. P: (-5,6)
9. Q:(1,2/3) 10. Q: (3,0,5)
11. Q:(l, 2, -5) 12. P: (1,2,8)
13. P: (-5,0,5)
14. cos a = -1/3, cos P = -2/3, cos y = 2/ 3
15. cos a = 1/d14, cos P = -3/414, cos y = 2/d14
16. [-543/2,5/2] 17. [- 15./2/2, -1542/2]
19. u = [2/3, -1/3, -2/3] the same direction as A
20. 4.31 miles high and 16 miles from the airport
EXERCISES 1.2
3. (a) 7 i +2 k , (b) 7j -17k, (c) I A+BI =d27
(d)
I A - B I = 459
4. (a) 4 A- 2 B=- 6 k , (b) 2A- 3B=- 4i +4j - 9k
(c) 12( A+B) l =6d3, (d) IAI + IBI =4 2 +d 1 7
5. X = [4/3,7/3,-8/31 6. X = [4,3/2, -21
7. m = 2, n = -1 8. m = 3, n = -2
9. a=- 1, b = 1 , c = 2 10. a = 1, b = -1, c = 1
11. I A- BI = I-(B-A)I = I B- AI , I A+BI = IAI + IBI
when the vectors are in the same direction.
12. 1 A - B I = I I A I - I B I I when the vectors are in the same
direction.
13. Max I A- BI = IAI + IBI, min I A- BI = I IAI - IBI I
14. v = (250 +2542)i + 25d2jf 83 deg. east of north or N 83O E
15. N 12.25O E 16. 75433 +125j, N 46.1 E
17. 2:30 PM, 16.7 miles from the lighthouse.
18. 28.96 miles apart
470 Solutions to Selected Problems
19. Let F1 = Fli . Then F2 = F1(-i +d3j) and F3 = 3F1(-i - d3j)/2
so that F1 + F2 + F3 = (-3i - d 3 j ) ~ ~ / 2 .
20. Tension in shorter wire 50.13 kg, in longer wire 50 kg
22. v = 30i + 150j, t = 9.8 min, 294 ft downstream
EXERCISES 1.3
2. Let the sides AB, BC, and CA be represented by the vectors
A, B, and C, respectively. Then PQ = PB + BQ = (A + B)/2
= -(1/2)C = AR which shows that PQ and AR are parallel
and equal in length.
4. Consider a trapezoid ABCD, and let P and Q denote the
respective midpoints of the nonparallel sides BC and DA.
Since AB + BP + PQ + QA = 0, it follows that QP = AB
+ (1/2) BC + (1/2)DA = AB + (1/2)(BC + DA) = AB +
(1/2)(-AB - CD) = (1/2)(AB + DC) .
6. The hypothesis gives (B + 2C)/3 = (A + 2D)/3, which says
that the line segments AD and BC intersect. The point of in-
tersection divides the line segments in the ratio 2 : 1.
9. Let A, B, C, and D denote the position vectors of the respec-
tive vertices A, B, C, and D of a quadrilateral ABCD. Then
the position vectors of the midpoints of the sides AB, BC,
CD and DA are given by P = (A + B)/2, Q = (B + C)/2, R =
(C + D)/2, S = ( D + A)/2, respectively. Thus it follows that
(P + R)/2 = (Q + S)/2, which is what needs to be proved.
10. The point divides AE in the ratio 6 : 1 and divides CD in the
ratio 3 : 4.
12. The point divides BC in the ratio 2 : 1.
EXERCISES 1.4
4/3, cos 8 = 4/9 2. -5/d14, cos 8 = -5/14
-11/d29 , cos 8 = -11/5d58 4. -28/421, cos 8 = -4/d21
A = (4/9)B + C , where C = (1/9)[22, -10,1]
B = (-4/7)A + C, where C = [15/7, -2/7,4/7]
i - j + k 9. (a) m = 26/27, (b) m = 27/38
(a) m = 1/7, (b) m = 29/5, 11. m = 2/7, n = -8/7
u = + (2i + j - 3k)/d14, 14. work = - 6 , 15. work = 24
I A + B I = (A + B)* (A + B) = A* A + 2A* B + BOB = A*A + Be
B=( A- B) * ( A- B) = I A - B I ~ ifandonlyif Ao B=0 .
Show that A* C = B* C . 25. Write AD = (AB + BC)/2 and
BC = AC - AB , then consider AD *AD and BC* BC .
Solutions t o Selected Problems
EXERCISES 1.5
1. (a) -3, (b) -10, (c) 2,2. (a) 10, (b) -2, (c) abc, (d) 0
9.
(a) (-i + 7j + 5k)/1/75 ,
(b) (-25i + 3j + 13k)/d803
10. (a) 7i - 7j + 7 k , (b) 27i + 9j , (c) 63, (d) 63, (e) 9i - 27j - 9k
(f) -14i - 28j - 14k
11. (a) -19i - 2j + 9 k , (b) -3i + 6j + 3 k , (c) 38i + 4j - 18k
(d) -60i + 30j - 108k , (e) -4% - 24j
12 (a) 7, (b) 419, (c) 2413, (d) (d819)/2
13. (a) -6i - 14j - 2k, (b) -1% - 12j + 6k
14. (a) -15i - 9j - 17k, (b) -2% + 2j - 7k
15. (a) 19k, (b) 14k
16. v = (42i - 18j - 4k)/d19, speed = l v l
17. v = (4i - 16j + 12k)/d3, speed = l v I , 18. (3, 1, 3), (1, -1, -I),
(-3,3,3), 19. (a) A. C < 0, (b) B* D > 0
21.
(a) B = (1 + t)i - (1 + 2t)j + t k for arbitrary value of t.
(b) B = (2i - j - k)/ 3
EXERCISES 1.6
1.
x = -1 + 2t, y = 3 - 3t, z = 2 + 5t (-- < t < -)
2. ( x- 3) / 5=( y+l ) / ( - 4) =( z - 4) / ( - 1) or ( x+2) / 5=( y- 3) / ( - 4)
= ( Z - 5)/(-1)
3.
x = 1 + 2t, y = -2 - 3t, z = 1 + 4t (-00 < t < -)
4. x - 3 = - (y - 4) = - (z - 6) , (z is arbitrary).
5. x = 17/ 7+ z, y = - 2 / 7 + z , 6. (3,-2,3), COSO = 8/1/66
7 (8,8,9), cos 0 = 1/6/3, 8. (5, - 4, 1)
9. x + 3 y - 7 = 0 ,
10. 3x - 12y + 19z - 34 = 0
11. x = z , 12. x - z + l = O ,
13. 2~ - 7y - 42 + 28 = 0, 14. d(3/2)
15. d(35/26), 16. 2/1/38,
17. 5/d38, 18. 21/1/494
19. 10/d65, 20. 23/1/381
21, center at (18,8,5), radius = d(3/2)
EXERCISES 1.7
1. (a) 4 (b) -22 (c) 29 (d) -67, 2. 2 , 3. 14, 4. 30
6. D= 2 A- 3 B+ 2 C, 7. D = A + 3 B - 2 C , 8. D = 2 A + B
9. D = 2A + 3 B, 10. A = 3 B + 2 C , 11. A= - 3 B- 2 C
15. A x (B x C) = -15i + 5j - 25k,
16. Consider (A x B) x (C x D)
and expand with respect to C and D.
17. Set U = A x B and apply (1.45) and (1.46).
19. Apply (1.45).
472 Solutions t o Selected Problems
EXERCISES 2.1
1. (a) [2t - ln(1 + t)]i + [2(1 + t
2
) + e
t
]j - (2 sin t + 1) k
(b) t i n (1 + t) - et (l + t
2
) - sin t, (c) [(I + t
2
) - et sin t]i - [t +
(sin t) ln(1 + t)]j - [te
t
+ (1 + t
2
)1n(l + t)]k, (d) I F(t) I = [ I + 3t
2
+
t
4
+ sin
2
t11j2 , I G(t ) I = [ln2(1 + t) + e2t + 1 1
2. @) 2e
t
(3t - 1) - (t
2
- 2)cos t + 9, (c) (t
4
- 2t
2
+ t cos t)i + (2te
t
- 3$
+ t
2
)j - [(3t - 1) cos t + 2t
2
et - 4e
t
]k, (d) [(2e
t
+ 3t - 1)2 + (t
2
- 2 -
cos t)2 + t
2
(t + I ) ~ ] ' / ~ , 3. h(t) = [t
2
+ e
2
t]-1/2
4. h(t) = [e-2t + t
2
] 5. (-b cos t i - b sin t j + ak)/(a
2
+ b2)-lI2
6. [(sin t - cos t)i - (cos t + sin t)j + 2k]/ d6,
8. (F x G) x H = t - ? ~ - t e t ~
EXERCISES 2.2
I. [(I + 43) /2]i + (1 - d3)j , 2. 62, 3. 43,
4. (3 + 43) /2i + (63 - 3)/2j
5
5. 5 , 6 . 4 , 7. 4 i - j - 3 k ,
8. 626, 9. (a) t cos t - 6t , (1 + 4t cos t - 2t
2
sin t)i - (2t + 4t sin t +
2t
2
cos t)j - t sin t k , (b) 2 - 2t, -(4$ + 4t + 1)i + (1 + 3t
2
)k
11. 0 , 12. F(t)* Ft(t) = 0 (8 = 9o0), 13. Fn(t) = a 2 ~ ( t )
15. Differentiate [ ~ ( t ) ] ~ = F(t)- F(t), 20. Note that if r = R(t)/ I R(t) I,
then d r = d R/ I RI - R ~ I R I / I R ~
EXERCISES 2.3
1. R(r) = r cos (In r )i + r sin (In r ) r 2 I), R'(t) = r K( r )
2. R( r ) = d ( l - r)i + (1 - r ) j + (1 - r ) l / l k ( r 5 14, Ki t ) = - 2t K(r)
3. ~ 2 + ~ 2 = a ~ , z = a + y , 4. x
2
+ y = z
5. t = [(3& - n/2)i + (3 + d3n/2)j]/d(36 + n
2
)
6. t = (-ai + bj)/6(a
2
+ b
2
) , 7. t = (4i + j)/d17
8. t = (-i + J 3j + J3k)/ d7 , 9. t = (i + J3j)/2
10. t = (e
2
i - e-2 j + k)/d(e
4
+ em4 + 1)
11. s = d3(et - I), t = In (I + s/ 63)
12. s = t[d(t
2
+ 2)] / 2 + l n [t/ 42 + (t
2
+ 2)l l 2/ 2]
13. s = 42 t2/2 , 14. s = 2d2n, 15. s = 2nd(a2 + b
2
)
16. (a) d s = ad(1 + cos t) dt , (b) ds = d(t2 + 2)dt
EXERCISES 2.4
1.
K = (t + 1 ) - ~ / ~ / 2 , n = [i - (t/2)j - (43t/2)k]/d(l + t
2
), b = (d3/2)j
- (1/2)k
2.
K = d2e- /3, n = [(- sin t - cos t)i + (cos t - sin t)j]/d2,
b = [(sin t - cos t)i - (sin t + cos t)j + 2k]/d6
Solutions to Selected Problems
4.
K = (1/2) sech t, n = sech t i - tanh t k
b = (-tanh t i + j - sech t k)/d2
5. K = 1/2t, n = cos t i - sin t j, b = (sin t i + cos t j - k)/d2
6. K = d2/(1 + sin
2
tp12, p = 1 / ~ . At t = x/4, the center of the
circle of curvature is given by (d2/8, -d2/4, d2/8)
7. Center at (a/2 - 3,5,2d2) with radius p = 442
8. K = dsech t , 9. (a) d2y - 22 + e"14 = 0
EXERCISES 2.5
r = 0 , 2. r = et / 3, 3. r = (t
2
+ 6)/(t
4
+ 5t
2
+ 8)
r = 1 /(sinh
2
t + cosh
2
t + 1) = 1/(2cosh
2
t)
r = - 1/(2t) , 6. r = 2a/(2a
2
t
2
+ I ) ~ , 10. K = r = 1/ [3(l + t212]
.
dt/ds = 1/(2t)n, n = cos t i - sin t j, dn/ds = - (1/2t)t - (1/2t)b,
t =( s i nt i +c os t j +k) / d2, b =( s i n t i +cost j - k) / d2,
db/ds = 1/(2t)n
12.
t = (cos t i + j +si n t k) / d2, dt / ds=( 1/ 2) nf n=- s i n t i + cos t k
dn/ds =(-1/2)[t + b], b = (cos t i - j +si n tk)/d2
db/ds = (1 /2)n
EXERCISES 2.6
v = 3 i + 6 j + 6 k , v = 9 , a =6j +12k
v = (a/4)(i + 39, v = (a/4)d10, a = (a/4)(3i + 5j)
v = -i - aj + ak , v = d(2a2 + I ) , a = ai - 2j + k
v = - (a/2)d3i + (b/2)j - (a/2)d3k, a = -(a/2)i - (b/2)d3j - (a/2)k
v = - sinh n: i - cosh n: j + cosh n: k , a = (sinh n:)(- 2j + k)
a = ao12n, n = - cos o t i - sin o t j
a = n , n = - s i n t i + c o s t j
a = [I + cos
2
(t/2)]'12n, n = [- sin t i + cos t j - cos (t/2) k]/ I n l
a = an, n = - (cos t i + sin t j)
a, = a sin t cos t/d(l + sin
2
t), a, = \12a/d(l + sin
2
t)
a , =2, a, = 8t
2
, p = 1/2
If v = (ds/dt)t = const then a = ( d ~ / d t ) ~ dt / ds = ( d ~ / d t ) ~ m .
Thus v* a = 0.
If I v I = c and I a I = a are constants, then v =ct so that a = c2m.
Since I a I = a, it follows that C ~ K = a or K = a/ c2 = constant.
The force is given by F = ma = mR"(t) and the direction of
motion is indicated by the velocity vector R'(t). Then F* R' =
mR"(t). R'(t) = 0 implies (d/dt)[R' (t) R1(t)] = (d/dt)v
2
= 0, which
means v = const.
RxR" = O i f andonlyif ( RxR1) ' = ( Rxv) ' = 0.
Solutions to Selected Problems
EXERCISES 2.7
v = w si nhwr u, + cos hot u, , a =( w2 - l ) c o s h wt u , +
2 w sinh wt ue
v = 2t u, + wt2 u,, a = (2 - w2 t
2
)u, + 40)tu,
v = r f ur + 3r u,, a = (rf' - 9r )u, + 6rfu,, where r' = 3a sin 3t,
r" = 9a cos 3t
v = r' u, + e-tr u, , a = (r" - re-2t)ur + (2e-t r' - e%-)u,
where r' = a cos t, r" = -a sin t
v = r' u, + r o u, , a = (r" - r w
2
)u
r
+ 2r ' o u,, where
r' = a sin t / ( l + cos t ) 2, r" = a(2 - cos t ) / ( l + cos t12
d3R/d$ = (I"' - 3r"ef2 - 3ref0")ur + (3r"e' + 3rf0" - ref3 + r8"' )ue
Let R(t) = r(t) ur(0). Then v = r' u, + rwu, , a = (r" - r w2) u, +
2 wrfue. Since (d/ dt )a = ku, , it follows that the transversal com-
ponent of (d/dt)a must be zero, that is, (r" - r w2) ou, + 2 wr"ue
= (3 wr" - r o)"u, = 0. Thus r" = r w
2
/3.
Centrifugal force = n
2
, Coriolis force = 4tr
EXERCISES 2.8
v = u, + t uH + 2k, a = - tu, + 2ue
cos e = [5/ (5 + t
2
)]' "
(b) v = cos tu, + sin tu, + k , a = - 2 sin tu, + 2 cos tuH
(c) 0 = 45"
(a) v = ~ ~ C I I U ~ , a = - d302 U,
(b) v = d3(w + y ) ~ , , a = - d3(w + y)2u
(c) v = + d3(w + y ) ~ , . a = d3(0 + Y ) 2 ~ r
a, = - 4at2 sin t
2
+ 2a cos t
2
, at = - 4at
2
cos t
2
+ 2a sin t
2
v = a y cos yt ur +2at s i nyt u, - ay s i nyt k
a = - y2 R - 4at
2
sin y t u, + (2a sin y t + 4aty cos y t)u,
Coriolis acceleration = 2a sin y t + 4aty cos y t
v = -aa sin a t i + a a cos a t j*, a = - a2 R
v = - a a s i n a t i + ( a a c os a t s i nyt + ay s i nat cos yt)j
+ ( aa cos a t cos yt - ay sin a t sin y t)k,
a = - a
2
R + 2aay cos at(cos y t j - sin y t k) - ay2 sin a t j*
v = n(ai + bk), s = 21r(a2 + b2 )' /2
0 = gbt2/[2( a2 + b
2
)] , s = 2g,b/(a2 + b
2
)'I2
Solutions to Selected Problems 475
EXERCISES 3.1
1. Planes 2. Paraboloids 3. An ellipsoid 4. Hyperboloid of
two sheets 5. Hyperbolas with axes x = y and x = - y
6. Hyperbolas with axes x = 0 and y = 0
12. (a) (d2/2 + 1/2)i + (1 - u'2/2)j (b) 42 (c) (d2a/2)i
13. (a) j + 2k (b) I F I = a
14. (a) (sin x + e
X
cos y)i + (cos y + e
X
sin y)j + In z k
(b) e
x
(sin x + sin y)cos y
(c) - e
X
cos y in z i + e
X
cos y in z j + e
x
(sin x sin y - cos
2
y)k
EXERCISES 3.2
1. 342, 2. 34./5/5, 3. (1+\13)e/(242), 4. -38/45, 5. -4/3
6. - 8/ 3, 7. -9/./34, 8. d3/3, 10. 1/2, 8 = - 90 deg.
11. 42, 8 = 45 deg. 12. a f / a x = - I, 3f/3y = 3; 11/417
13. 20 + evf7 (1n 2 - d3), 14. 4t3 - 3t
2
+ 10t - 2
16. 243 + 4 + 2a/ 3, 17. 12
EXERCISES 3.3
(a) 2i + j (b)(l/./2)[(1 - l r / 4) i +( 1 +lr/4)j] (c) (i + j)/2
(d) 2i + j + k (e) (2/3)(i - j - k) (f) (-1 /e)im1k
(a) d5 (b) 2d5/3
(c) 2.117 (d) (32 + 9a ) /2/4
(-1/\12,l/d2) and (1/d2, - l / ~' 2) ; 1
(a) 14/3
(b) (2/e - 3)/d17
2(1 + x
2
+ y2)/[4(x2 + y') + 1]'/2
d f / d n = y + z o n x = 1 a nd- ( y+z ) o n x = - 1
df/dn = z + x on y = 1 and - (z + x) on y = - 1
d f / d n =x +y o n z = l a n d - ( x + y ) o n z = - 1
2 x - 2 y + z + 1 = 0 8. t = i
(a) 2 j + k , cos8=-8u' 21/63 (b) i - j - k , c o s 8 =1 / 2
b = 2 2 11. a = 3/4 15. (a) Vf(r) = fl(r)R/r (b) Vr
n
= nrn-2 R
EXERCISES 3.4
1. 2e
X
cos y 2. 2(x
2
- y2)/ (x2 + y2)2 3. x(2 + ze
X
Y)
4. 0 5. ex + eY + e
Z
6. sin y + sin (xz) - e
Z
sin e
Z
7. 2/(x
2
+ y2 + 10.
11. g r a d d i v F = i + j + k
15. @( x, y) =ar ct an( y/ x) +C
17. $(x, y, z) = e
X
cos z + y2 + C
18. $(x, y, z) = xeY + zy2 ln z + z
2
sin x
476 Solutions to Selected Problems
EXERCISES 3.5
1. 0 2. l/d(x2+y2)
3. (x2z3 - 3x2yz2) i + (3x2yz2 - 2xyz3 j + (2xyz3 - x2z3) k
J
6. 0 8. -2xy
2
- 2yz2 - 2x z
1
4. 2x(1- z) i + 2(yz - x - y)' 5. - x z i + 4xyz j +(2y3 - xz
2
) k
9. (3xy2 - x3 + 2y2z 2)-* 1 - (Y " 2xz2 - 3 z2) j + (3x2z - z3 + 2x2y) k
Y
10. (y2z-x"i +(xz 9 ) j +( x 2 y - z - ) k
11. ( ~ ~ - y ~ ) i + ( ~ ~ - z ~ ) j + ( z ~ - x ~ ) k 14. 0
15. + (x, y, z) = x yz + C
EXERCISES 3.6
1. 2yz i - xz j + rad +, 2 any differentiable function of x, y, z.
5
2.
(xz
2
/2)i + (x y/ 2- yz /2)j
12. F x (He V)G - F x (Go V)H - V. G(F x H) + V* H(F x G)
+ (F* V)(G x H) + (G x H) x (V x F) + [(G x H)* V]F
13. V(F- H)* G + (F* H)V* G - V(F. G). H - (F* G)V* H
14. V(F* H) x G + (F* H)V x G - V(F* G) x H - (F* G)V x H
EXERCISES 3.7
6.
(a) u = const. defines a parabola opening to the left
v = const. defines a parabola opening to the right
u = 0 corresponds to the nonpositive x-axis
u = v corresponds to the nonnegative y-axis
u = -v corresponds to the nonyositive y-axis
(b) R, = ui + vj , Rv = -vi + u j , R, = k
(c) ds
2
= (u
2
+ v
2
)(du
2
+ dv
2
) + dz
2
8. ds
2
= (u
2
+ v2)(du2 + dv
2
) + u
2
v2 de2
9. a ( ~ , V, w) / ~( x, y, Z) = I / ( u ~ - ~ 2 )
10. (a) x = (u + v + w)/2, y = (-u + v + w)/2, z = (u + v - w)/2
(b) aR/au* aR/3v x 3R/aw = - 1/2
(e) ds
2
= 3(du
2
+ dv
2
+ dw
2
)/4 + (du dv + dv dw - dw du)/2
EXERCISES 3.8
Solutions to Selected Problems
2
2. v i =--J-[r2sin 2 $ 1 + s i n ) + (
r sin @
38 sin @ 38
1 3 2
V. F =
2 j a r sin 9 F,/ + q. sin mi d + $r F?~, ]
r sin @ 30
F = Flul + F2u2 + F3u3, u1 = sin @ cos 8 i + sin @ sin 8 j + cos @ k,
1
cur 1 F = -------.
2
r si n@
3. (a) In cylindrical coordinates, Vf = (2r sin 0 cos @ + z sin 8
+ z cos 8)ul + [r(cos
2
8 - sin
2
8 ) + z(cos 8 - sin 8 )]u2
u, r u 2 r s i n @u g
3/3r 3/3@ 3/38
F, rF2 r s i n@F3
+ r(sin 8 + cos 8)u3, V f = 0
(b) In spherical coordinates, V f = 2r el , V f = 6
4. i = cos0 ul - si n8 u2, j = si n8 ul + cos8 u2, k = u 3
F = ( Z cos 8 - 2r sin 8 cos 8)ul - (z sin 8 + 2r cos2 8)u2
+ 2r sin 8 u3
5. div F = 0, curl F = (2 cos 8 + sin 8)ul + (cos 8 - 2 sin 8)u2 - 2u3
6. div F = 0, curl F = - 2z cos 8 ul + 2(z sin 8 - r) u2 - u3
7. i = sin Q cos 8 u, + cos @ cos 8 u$ - sin 8 ue ,
j = sin @ sin 8 u, + cos @ sin 8 ue - cos 8 ue,
k = cos @ u, - sin @ ue
8. F = (r sin
2
@ sin 0 cos 8 - 2r sin @ cos @ sin 8 +
r sin @ cos @ cos 8 )u, +
Solutions to Selected Problems
(r sin $ cos $ sin 0 cos 8 - 2r cos $ sin 0 - r sin
2
@ cos 8 )u+ -
(r sin + sin
2
0 + 2r cos @ cos 8 )ue . V. F = 0
V x F = (2sin$ cos8 - si n@ sin0 -cos@)u, + (2cos$ cos8 -
cos $ sin 0 + sin $)u - (2sin 8 + cos 8)u,
9. F = (r sin $ cos @ cos 0 + 2r sin
2
$ sin
2
8 + 3r cos
2
$ )u, +
( r cos
2
$ cos0 +2r s i n@ cos@ sin
2
0 - 3r si n+ cos+) u +
(2r sin $ sin 0 cos 8 - r cos + sin 6) ue
V . F = 5 , Vx F = s i n @s i n Ou , + c o s $ s i n 8 u + + c o s 8 ~ ~
Sol ut i ons t o Selected Probl ems 479
F = F , u , + F2u2+ F3u3, U, = ( v c o s 0 i + v s i n 0 j + uk) / d( u2+ v
2
)
u2 = ( u cos 0 i + u si n 0 j - vk) / d( u2 + v
2
), u3 = - si n 0 i + cos 0 j
EXERCISES 4.1
1. 0, 2. 242,
3. d(2 + x2'4[(2 + x2) / 3 + n/ 2] + ln[d(2 + x2) + x] - (2/3)d2 - (In 2)2
4. d(a2 + b-)(a
2
/4 - 9bx2/32) , 5. 8(22/2 - 1) / 3
6. (a) 342, (b) (1 /3, 1/ 3, 1/ 3)
7. (a) ha2, (b) (0,O), (c) 7a4/ 3, (d) l l a 4 / 3
8. (4/3)(\12 + 1)/[d10 + 45 i n (1 + d2)]
9. (2/3)[(x
2
+ 5)3/2 - 5.151, 10. 88/ 15, 11. - 8 / 3
12. O 13. 0 , 14. (2 + 9x"/)d2, 15. x
EXERCISES 4.2
1. 1 , 2. 18/ 5, 3. - 8/ 3, 4. 0 , 5. 137/ 6,
6. -1/60 , 7. 1 / 3 , 8. -2a + x2
9. (a) 2 x , (b) 2 x , 10. - 2x , 11. 2 , 12. - x
13. (a) 0, (b) 0 , 14. - 2 x , 15. x/ 2 - 3x ' / 8,
16. 2 - d2/ 6 17. -1 + d2/6
EXERCISES 4.3
3. (1/2)ln (x
2
+ y2) + C, 4. y2 sin x + x
2
eY + C
5. z
2
eXY + y cos x + z + C,
6. x y e
Z
- x z + y 2 + C
7. x
2
z + ye-X + y cos(z - 1) + C, 8. (1/2) l n (x
2
+ y2 + z
2
)
9. 3xe
Z
+x2y - z cos y + C, 10. F is not a conservative field
12. 32, 13. x2 + e + 4, 4 -2 - 2 , 15. (9 + In 5) / 2
480 Solutions to Selected Problems
EXERCISES 4.4
1. 16, 2. 0 , 3. 0 , 4. n a
4
/ 4, 5. 2071
6. 3n/ 2, 10. 4n, 11. 0 , 12. 6n, 13. 0
14. 47t, 15. (a) Let F = v Vu i n (4.21)
EXERCISES 4.5
1. x
2
+ y2 = a2 (z 2 O), a cylinder, 2. x
2
+ y2 = z , a paraboloid
3. a
2
(x
2
+ y2) = z
2
, a n u right circular cone
4. (x/a)' + (y/bf + ( z / c ) ~ = 1, a n ellipsoid
5. ( ~ / a ) ~ - (y/ b) = 42 , a hyperbolic paraboloid
6. ( ~ / a ) ~ + (y/bf - (z/c12 = 1 , a n elliptic hyperboloid of one sheet
7. ( ~ / a ) ~ - (y/ b) = z
2
, hyperboloid of t wo sheets
8. x = u , y = v , z = ( d - a u - b v ) / c , ( I u I <-, I v I < - )
9. x = u , y = v , z = u 2 , ( l u l <-, l vl <-)
10. x = au cos v, y = bu sin v, z = 1 - u2 ( U 2 0, 0 5 v 5 2n)
11. x = a sin u cos v, y = b sin u sin v, z = c cos u
( OI u < n , O< v < 2 n )
12. x = a cos u cosh v, y = b si n u cosh v, z = c si nh v ( I v I < -,
0 I u I 27t)
13. x = + a cosh u, y = b cos v si nh u, z = c si n v sinh u ( I u I < -,
O I v I 2 n )
14. x = a(u +v)/2, y = b(u - v)/2, z = cuv ( 1 u l < m, I v I < -)
15. 2x - 2y - z = 2 , 16. 43x + y - 2 z = O
17. 2 x + y + 4 z = 8 , 18. x + 6 ~ - 4 z + 8 = 0
19. 3~ + 43y - 12 = 0
EXERCISES 4.6
1. d3/ 2, 2 . d 3 n a 2 , 3 . d 2 n a 2 , 4. (545 - 1)n/ 6
5. 2d(2ab)(a + b) / 3, 6. 2a2(n - 2) , 7. 8a
2
8. n[ad(l + a
2
) + In (a + d( l + a2))] , 9. 4n a2(1 - 1/42)
10. (2/3)d 2/3) n
I
11. 242na2
12. 2n [(a + l ) ~ / $ - 1]/ 3 , 13. a
2
14. (a) 4n a2[2 - ((7/2) - \113/2)' 12] , (b) n a2[(2d13 + 3)0/2 - 11/ 6
EXERCISES 4.7
1. 2d3/15, 2. a%n + 2ab3n/3, 3. 2n a4/ 3
4. 0 , 5. 0 , 8. 2n $h, 9. 545n/4 - 3n/20
10. (a) na 5/ 2, (b) n a
5
/ 4, 11. 64
12. (2545 + 1)/(75d5 - 9) , 13. 1 / 2 , 14. 2
Solutions to Selected Problems
EXERCISES 4.8
1. 3/ 2, 2. 1/ 24, 3. 6 a , 4. 7 a , 5. 0
6. 5 a , 8. n = 1 4a a0 ; n = 2, 16a a
5
/5
5
10. (a) 4a a3, (b) 16a a / 5, 11. 16a 2 / 3 , 12. 16, 16. 1
EXERCISES 4.9
1. 116, 2. -BE, 3. -2.11, 4 . 0 , 5 . 0
6. 0 , 7. 0 , 8 -a a4/ 2, 9. 2aR2(a2 + b
2
+ c
2
)/c
10. (c) 2a
EXERCISES 5.1
I. (a) 6 4 . = &, a1 + 6i2a2 + tji3a3 (1 = 1, 2, 3) or (al, a2, a3)
1J J
(b) al bl + a2b2 + a3b3
(c) aI1xl2 + (al2 + a21)xlx2 + a2Zxz2 + (a13 + a31)xlx3 + (a23 +
a32)x2x3 + a 3 3 ~ 3 ~
(d) ailblj + ai2b2, + a,b, = 6,, (i, j = l , 2, 3)
(e) s = al l bl l + a12b12 + a13b13 + a21b21 + a22b22 + a23b23 +
a31b31 + a32b32 + a33b33
a xi a xj Sxi Sxj a xi a xj
+ ---------- + --------
(f) gii=m au2au2 au3au3
(i, j = 1, 2, 3)
482 Solutions to Selected Problems
EXERCISES 5.2
1. x*, = xO , x*, = - xl , x * ~ = X, , left-handed
2. x*, = X, cos 6 + x2 sin 6 , x*, = - X, sin 6 + x, cos 6 , x * ~ = x3
3. x*, = X, , x*, = X, , x*, = X, , right-handed
4. x*, = (x1 + x2 + x7 )/43, xS2 = (2x1 - x2 - x3 )/46, x", = (x2 - x3 )/42,
EXERCISES 5.3
2. (a) a*, = 1 - &/ 2 , a*, = 1/ 2 + h13/4, a*, = -1/4 - 3d3/2
(b) bl = -3/4 + d3/ 2, bl = 5/ 2 + 3.13/4, bl = 1 - 3J3/ 2
(c) 443 + 15/4
3. (b) [4.13, - 342,5461 , (c) [3, 1, 11
EXERCISES 5.4
3. I,, = 13M, I,, = 13M, I,, =8M, I,, = I,, =4M, I,,= I,, = 6M,
In = I,, = - 6 M,
4. 17M, 5. (a + b +c)/ 3, 6. 5M
8. I*,, = 11/4, = (75 + 4d3)/16, I*77 . . = - (39 + 443)/16
11. (5.13, - 4.13,143), S, = 10/3, st = d(26)/3
12. (4/3, - 1/3, - 4/3), sn = 5/3, st ='18/3
13. (8/d5, - 1/45), sn = 6/5, st =d(289)/5
14. s * ~ = (S1 + 3S2)/4, s*,* = 0, s*i3 = '13(Sl - s2)/4,
= 0,
s*,, = SO , s *, ~ = 0 , S*3, = , s*-J, = 0 , s * ~ = (35, + S2)/4
EXERCISES 5.5
1- (a) [- 1,8,151, (b) [- 5,1101, (4 3
, where C ij = ai& ki
4 10 1
(b) (ci j )= 0 -13
7 , wher e Ci j =ai &j k
[o -2 11
, wher e c, j = aki bkj , ( d) -2
Solutions to Selected Problems
EXERCISES 5.6
1. (a) h = 0, [2, -1, -11 /d6; h =2, [0, 1, -11 /d2; h = 3, [I, 1, 1] /1/3
(b) (I*,) = 0 2 0
I:: :: 'I]
(c) x * ~ = ail x, , where
2 / \ 6 - 1 / \ 6 - 1 / \ 6
0 1/ 12 4 / 1 2
1 / \ 3 1/ 13 1/ 13 1
2. (a) h = 1 , [0, -1, 1]/ d2; h = 2 , [1,0,0]; h = 5 , [0,1,1]/Y'2
3. h = I , [I, 1,-1]/d3; h =- 2, [I,-1,0]/1/2; h =5, [O,1,1]/1/2
4. h = - 3 , -3, [al , a2, a3]and [ bl , b2, b, Jwher e2al - 2a 2+a 3=0,
2bl - 2b2 +b3 = 0, and the vectors are chosen so that they are
orthogonal; for example, [I, 1, 0]/ d2 and [I, -1,4]/d18;
h = 6, [2, -2, 1]/ 3
5. sX1, = 0, s*,, LL = [ l + d(1 + 4(a2 + b
2
))]/2,
s*,, = [ I - 4(1 + 4(a
2
+ b
2
))]/2
7.
xq12 + ~ ' ~ 2 + 4 ~ * ~ 2 = 1, ellipsoid (the variables may be inter-
chan ed)
8. 3x*, p+ 3XXz2 + 9 ~ * ~ 2 = 1, the variables may be interchanged
9. x*~' + hx*22 = 1, the variables may be interchanged
484 Solutions to Selected Problems
EXERCISES 5.7
5. (a) cl = 4x1, c2 = 3x2, c3 = 4x3 , where ci =3sji/axj (i = 1, 2, 3) ;
c1 = 3x1 + x3, c2 = 4x2, c3 = 4x0, wher e ci =3sij/3xj (i = 1, 2, 3)
(b) cl = -1, c2 = 0, c3 = 1, where ci = a t / 3 x (i = 1, 2, 3);
1J J
c1 = 1, c2 = 0, c3 = x2, where ci =3t../axj (i = 1, 2, 3)
J'
(b)
EXERCISES 6.1
-
1
2x1 x3
2
1 + X l
-2x1 cos x1
>
2. Tr aceof t hemat r i x.
Solutions to Selected Problems
i 2 i
aa 3 u b.i + ~u abJ
5.
-
-
(i, k = 1, 2, 3)
j k
ax
k
axkaxJ JX ax
7. (-1, I), 8. (-2/3,4/3), 9. (-1/2, -1/2,5/2)
10. (4, -3,2), 11. (17/2, -7/2, -3/2),
14. e1 = [I, -1, 01, e
2
= [O, 1, -11, e
3
= [O, 0, 11
15. el = [-1/2,1/2,1/2], e2 = [1/2, -1/2,1/2], e, = [1/2,1/2,-1/21
16. e1 = [-I, 3,,-5]/2, e
2
= [I, -1,3]/2, e
3
= [I, -1,1]/2
17. x
i
= cxi*,x*~ , where
The transformation reverses the orientation of the coordinate
system.
18. x
i
= cx,'.,x*~ , where
19. x*; = cti*.xj , where
J
i
. d e t ( a x j ) =- 3 / 4
112 112 112
20. x*' = cx~*~x~ , where
I
coordinate system i s left-handed
Solutions to Selected Problems
22. x * ~ = cti*.xj, where
.I
EXERCISES 6.2
2. (a) a*1 = 3, aX2 = 3; a*' = 1 /1 I a*2 = 4/ 3
(b) a*l = 3, a*2 = 5; a*' = 1 /3, a*2 = 7/ 3
(c) a*' = 3, a*2 = -1, a*3 = -2;
= 3 I ,*2 = -1, ,3*3 = -2
(d) a*, = 4, a*2 = -3, a*3 = 5; a*' = 1, a*2 = 0, a*" 2
3. a*' = (xlx2 + ~ 3 x 2 - x1x3)/2, aX2 = (- ~ 1 x 2 + ~ 3 x 2 + xlx3)/2,
= (x1x2 - x3x2 + x1x3)/2, where xl = xX1 + x*?, x2 = x*' + x * ~ ,
X3 = x*2 + xx3
4.
a*' = (x12 - x22 - x32)/2, ae2 = (x12 + x22 - x32)/2,
(x12 + x22 + x32)/2, where xl = x* l + x*O, x2 = - x*' + x**,
X3 = - X*2 + X*3
x1
(a* ) = - x 1 - x 2 - x 3
2
lJ
- x2
where xl = - x*' + x * ~ , x2 = x*' + xe2 , x3 = - xs2 - x * ~
Solutions to Selected Problems
10. A* = A*A A *' , where A* = (a*'~) and A* = (a;')
EXERCISES 6.3
The index i denotes rows.
488 Solutions to Selected Problems
4. (a) 4.10, (b) -17, (c) 16e1 - 4e
2
+ 14e3
EXERCISES 6.4
e2 = [r cos Q cos 0, r cos @ sin 0, -r sin Q]
e? = [- r sin @ sin 0, r sin @ cos 0, 01
el = el , e
2
= e
2
/ r
2
, e
3
= e3/(r si n@) 2
el = [W cos 0, w sin 0, v], e2 = [v cos 0, v sin 0, -w]
e3 = [-vw sin 8, vw cos 0,0],
Solutions to Selected Problems
4. e l = [a sinh v cos w, a cosh v sin w, 0 ]
e2 = [- a cosh v sin w, a sinh v cos w, 0 1, e3 = [I, 0,0]
1 2 2 2 2 2
e = el/[a (sinh v + sin
2
w)] , e = e2/[a (sinh v + sin
2
w)]
2 2
a (sinh v + sin2 w) 0
(gij ) = o
2 2
a (sinh v + sin2 w)
0 0 1
5. el = [u2, ul , 01, e2 = [ ul , u2, 01, e3 = [O, 0, 11
e1 = [u
2
, -ul, 0] /v, e
2
= [- u
l
, u
2
, 0] /v, e" e3 , where
v = ( u2)2 - ( u')2
6.
r = ( v
2
+ w
2
)/2, 0 =arctan[2vw/(v
2
-w
2
)], z = z
3(r, 8 , ~ ) = 2
a ( ~ , w, Z)
2 2 2
7. r =v w, 0 = 0 , z = ( V - w5/ 2; a(r' O' z) = v + w
a h , w, 0)
8.
R~ = r
2
+ z
2
, 0 =0, Q = arctan(r/z); r = R sin Q, 0 = 0, z = R cos Q
EXERCISES 6.5
3.
(a) a*' = (v
2
- w2)v2w/2 + vw2z, a'2 = -(v
2
- w2)w2v/2 + v2wz,
a*, = (v
2
- w2)z/2 ; a*' = a*' /(v
2
+ w
2
), a*2 = a ~ ~ / ( v ~ + w
2
),
a*3 = a*3
Solutions to Selected Problems
(b) a*l = (x12x2 + ~22x3 + x1x32)/r
a*2 = (x1x?x3 cos 0 + ~ 2 x 3 ~ sin 0 - ~ 1 x 3 r sin Q
a*, = -x1x22 + xlx2x3 , where x1 = r sin Q cos 0
x2 = r sin @ sin 0, x3 = r cos Q
a * ~ - - a*l , a*2 = a*l /r
2
, a*3 = a*72 /(r
2
sin Q )
(c) a*1 = axlx2 sinh v cco w + axxg2 cosh v sin w
aY2 = - axlx2 cosh v sin w + ax3x2 sinh v cos w
a*3 = - xlx2, where xl = a cosh v cos w, x2 = asinh v sin w ,
x3 = z
4. aX1 = x1x2 w cos 0 + (2x, - - x32)w sin 0 + x12v
aX2 = x1x2 v cos 0 + (2x7 - - x32)v sin 0 - x12w
a*? = - xlx2 vw sin 0 + (2x2 - x32)vw cos 0 , where xl = vw cos 0
x2 = vw sin 0, x3 = (v
2
- w2)/2
a*' = a*l /(v2 + w2), a*2 = aL2/(v2 + w2), a*3 = a*?/(v2w2)
1) sin 0 cos 0
0
r(x2 - x 1) sin 0 cos 0
(b) a*ll = (x13 + x27 + x33)/r2
a*12 = a*21 = x12 x3 cos 0 / r + x22 x3 sin 8 /r - x2 x32 /(r sin 8)
a*13 =
= XI X2 ( X2 - xl)/r
a*22 = x, x32 cos
2
0 + x2 x32 sin
2
0 + x22 x3/sin 0
Solutions to Selected Problems 491
"'23 = a*32 = x x2 x3( sin 0 - cso 0), a*,, = x x~~ - x2x12
2 2 2
a* 12/r ae13/(r sin 0) I
4 2
a*??/(r sin 8)
(c)
aXl 1 = w2(x1cos2 8 + x2 sin2 0) + x3v2
ae12 =
= vw(xlcos2 8 + x2 sin
2
0 - x3)
=
= vw
2
(x? - x l ) sin 8 cos 0
a*22 = v2(xlcos2 0 + x2 sin
2
0) + x,w
2
a*23 = a*32 = wv
2
(x2 - x1) sin 0 cos 0
a*30 = v2w2(x1 sin * 0 + x2 cos 0), where xl = vw cos 0.
x2 = vw sin 0. x, = (v2 - w
2
)/2
, d l = a*l / (v2 + w ~ ) ~ , a*21 = a*2l /(v2 + w2)?
$12 = a*12/(v2 + w2)?, a*22 = a* 2 2 / ( ~ 2 + w' ) ~
,a31 = a* l o/ [ v 2 2 2 w ( V + w2)] , a*32 = a**,/[ \'*w2(v2 + w2)]
a*" = a*33/ ( v4~4)
6. (a) V@ = v (v
2
+ w2)e1 + w (v
2
+ w
2
)e
2
+, where el , e
2
are given
in answer to Prob. 3,, Exer. 6.4.
(b) VQ =2r e
l
, where el = [sin @ cos 0, sin Q sin 8, cos Q]
7. V@ = [(3v2w - w3)/2 + (w + v)z]e
l
+ [( v3 - 3vw
2
v )/ 2 + (v - w)z]e
2
+ [(v2 - w
2
)/ 2 + wv]e3 , where el , ex, e
3
are given in answer to
Prob. 2, Exer. 6.4.
EXERCISES 6.6
1.
dA/dxl = [x,, 0, x,], ;1A/0x2 = [xl, x,, 01, 3A/i3x7 = [0, x2, xl]
492 Solutions to Selected Problems
cos x2 -XI sin x2
0 sin x3 x2 cos x3
Vx1 0 eX1 1
I
0
3
X
2 1 1
3. -x si n x cos x
0 0 - 1
2
cos X
6. [I, 221 = -r, [2,12] = [2, 211 = r, [I, 331 = - r si n
2
9
[2. 331 = - r
2
sin 9 cos @, [3,13] = [3,31] = r sin
2
@
[3,23] = [3,32 = r
2
sin @ cos @
i 2 3
(''221 = - r , { 121 = { 211 = {3 131 = {
= l / r
33} = - r sin
2
Q, (3 231 = ( 3 32) = cos @ / si n @
{2 33} = - cos si n 9 ; otherwise, zero.
7. [ 1, 11] =[ 2, 12] =[ 2, 21] =- [ 1, 22] =v
[I, 121 = 1,211 = -[2, 111 = 2,221 = w
L 2
1 2
I
{1111={ 121={ 211=- { 221 =v / ( v 2 +w2 )
{I 12} = { 21) = - {
= w/ ( v2 + w
2
) = {2 22}; otherwise, zero.
8.
[I, 111 = [2, 121 = [2,21] = - [I, 221 = v
[I, 121 = [I, 211 = [2,22] = - [2, Ill = w
[1,33] = - vw
2
, [2,33] = - v
2
w
[3, 131 = 3,311 = vw2 , [3, 231 = [3,32] = v
2
w
L 2 1
111 = 1
= { 211 = - 1 221 = v/ (v2 + w2)
1 2 2
1l 121={ 21}= { 221=- { l 1 1 = w/ ( ~ 2 + w2 )
3 1
{3131= { 31) =- 1 331 = 1 / v
{3 23) = {3 32} = - {2 33) = 1 / w ; otherwise, zero.
9. [I, 111 = [I, 331 = [2,21] = [2,12] = u
[I, 121 = [I, 211 = [3, 231 = [3,32] = v
[2,23] = [2,32] = [3,11] = [3,33] = w
Solutions to Selected Problems
1 2
1 = { 33) = { 211 = {2 121 = u / b 2 + w2)
121 = 21) = w2/[v(u2 + w2)1
2
{2
= { 331 = - v/(u2 + w2)
3
23) = { 12} = - uw/[v(u2 f w2)]
2 3
{ 2 3 2 ~ = ( 23}={ 1 1 ) = { 3 3 3 ) = w/ ( u 2 + ~ 2 )
3
{3 23) = { 32} = u2/ [v(u2 + w2)] ; otherwise, zero.
10. al,, = sin
2
@ cos
2
0, al,2 = r si n$ cos $ cos
2
0
"1.7 = - 2r sin
2
$ cos 0 sin 0 - sin $ cos $ cos 0
= r sin $ cos $ cos
2
0. a2,2 = r
2
cos
2
$ cos
2
0
a2,3 = - (1/2)r
2
sin 29 sin 20 - r cos
2
$ cos 0 . a3,1 = 0
a3,2 = r sin
2
@ cos 0, a? = - (1/2)r sin 29 sin 9 + r
2
sin
2
9 cos
2
0
' I C
a?, 2 = r- a2, , a2, 3 = r- 2 a2,3, a3.1 =al I 1/(r2 sin'+)
a" 2 = a3,2/(r2 sin
2
$), a3,3 = a? I. ?/(r2 sin
2
9)
11. a , = w
2
, a1,2 = vw
2
, a1,3 = 0
~ I ~ , ~ = V W , a 2 , 2 = ~ 2 , a2,3 =
a3.1 = 0 , a3,2 = 0 , a q 3 = 0 . I-
d , =aj , /(v
2
+ w
2
) for i = l , 2, 3; j = l , 2
d , = a. . = 0, otherwise.
J l '
12.
a- . = 0, a', . = 0 for all i, j = 1, 2, 3
1. J J
EXERCISES 6.7
11. (a) [r
2
sin
2
$ sin 20. r3 sin 2$ cos 0. r
2
sin $ cos 2$]
(b) 4r sin
2
$ sin 20 + 2r
2
(2 cos
2
$ - sin
2
Q)cos 0
I
e 1 e2
(c) L 3 /3r a/3$
r sin $ 2 2 4 2 3 2
r sin $ sin 20 r sin 2$ cos 0 r sin $ coss 2@
12.
(a) [r
2
sin 2Q sin 0. r sin 9. cos 9 /sin 91
(b) 4r sin 2Q sin 0 + 3 sin $ cos $
494 Solutions to Selected Problems
I
e 1 e2
(c) 2 3/3r 3/39 3/30
r sin9 2 2 2
r s i n2qs i n0 r s i n 9 r cos 9
EXERCISES 6.8
2 2 . 2 2 2. 2
3. (a) ' m(v +wj ( v + w ) + ' mv w 0
2 2
Index



A





Acceleration, 19





centripetal, 121




Coriolis, 130, 131




in cylindrical coordinates, 136



in general coordinates, 459




normal component of, 121




in polar coordinates, 129




in spherical coordinates, 139



tangential component of, 122



Algebra






of matrices, 316





of tensors, 345, 397




of vectors, 10-16




of vector fields, 150




of vector functions, 77




Appolonius theorem, 37





Arc length of curves, 96, 190





in curvilinear coordinates, 191



in cylindrical coordinates, 192



in spherical coordinates, 196



Area cosine principle, 268





B







Basis





in curvilinear coordinates, 199



in cylindrical coordinates, 136



local, 190




natural, 15




orthonormal, 33




in polar coordinates, 128




reciprocal, 377-379




in spherical coordinates, 138



C





Cartesian coordinates





rectangular, 4





oblique, 373





Cauchy-Schwarz inequality, 37




Center of mass, 210





Characteristic equation, 353





Christoffel symbols, 434





Conservation of energy, 244





Conservative fields, 173, 237





Continuity equation, 288-289





Contraction of tensors, 346, 397




Contravariant components, 434




of metric tensors, 403-404




of tensors, 393-396





of vectors, 387





Coordinate





curvilinear, 185




elliptic cylindrical, 197




parabolic, 197




parabolic cylindrical, 197




Coordinate curves, 188





Coordinate surfaces, 188





Covariant components, 434





of metric tensors, 403




of tensors, 393-395




of vectors, 387






Covariant derivative





of tensors, 446-447




of vectors, 432-433

Cross product (see vector product)



Curl of vector fields, 176




in general coordinates, 456




in orthogonal curvilinear coordinates, 203-205



Curvature, 102, 106-109, 112



radius of, 102




Curves, 87-89




closed, 96




piecewise smooth, 96




simple closed, 96




smooth, 96




Curvilinear motion, 117




in cylindrical coordinates, 135




in spherical coordinates, 137



D





Derivative




directional, 155




intrinsic, 459




normal, 165




tangential, 158




of tensor fields, 361-362




of vector functions, 83




Determinant, 39-40, 51




Directed line segment, 2





Direction cosines, 5





Directional derivative, 155




Distance of a point from




a line, 58




a plane, 62




Divergence, 168, 287




in curvilinear coordinates, 203



in general coordinates, 459-453



properties, 169




of tensor fields, 363




Divergence (Gauss) theorem, 283



Domain





doubly connected, 240





simply connected, 239




Dot product (see Scalar product)



E





Eigenvalue, 353




Eigenvector, 353




Electrostatic field, 172




Energy




kinetic, 124, 227, 460




theorem, 125




Equivalence clsss, 2




F






Faraday's law, 304





Flux density, 171




across a surface, 276




Frenet-Serret formulas, 112



G





Gauss' law, 291-292




Gauss' theorem (see Divergence theorem)



Gradient of a scalar field, 160



in general coordinates, 450




geometric properties, 163




in orthogonal curvilinear coordinates, 198-199



Green's theorem, 246





normal form, 253





tangential form, 252




H





Heat flux, 279



equation, 289




Helix, 94




I





Inner product, 347




Incompressible flow, 171




Irrotational vector field, 303



J





Jacobian, 186




K






Kepler's law, 131



Kinetic energy, 124, 227, 460



Kronecker delta, 312, 377




L





Lagrange equation, 463




Lagrangian function, 464




Lagrange identity, 73




Laplace equation, 170, 291




in general coordinates, 453




in orthogonal curvilinear coordinates, 203



Left-handed coordinates, 318



Level curves, 149




surfaces, 149





Limits





properties of, 152, 153




of vector fields, 80, 151




Linear dependence, 78




Line, 53




equation of a, 55




Line integrals




geometric interpretation, 215



independent of path, 235-239



properties of, 227-228




of scalar fields, 208




of vector fields, 220-222





M





Magnitude of a vector, 5, 7-8



of vector functions, 77




Mass, 273




center of, 273




Matrix, 315-317, 329




inverse of a, 317




properties of, 316




of transformation, 383




transposed of a, 315




Metric, 191




Metric tensor, 403





Mobius strip, 262





Moment of inertia, 210, 273, 327



Moment vector, 49




N





Newton's second law, 124, 459



O





Orientation of curves, 90




Orthogonal curvilinear coordinates, 190



Orthogonal projection, 32




Orthogonal vectors, 32, 355



Orthonormal basis, 33




Osculating plane, 104





Outer product, 346





Outward normal vector, 165



P





Parallel vectors, 46




Parallelogram law of addition, 13



Parametric equations of curves, 87



Physical components, 455




Plane, 53




equation of a, 60




Potential function, 173




Principal axes, 332, 353




Principal normal vector, 102



Principle of conservation of energy, 244




Pseudo-vector, 326




Q





Quotient rule, 389




R





Radius of curvature, 102




Reflection of coordinates, 313



Ricci's theorem, 449




Right-handed coordinates, 318



Rotation of coordinates, 312



S





Scalar multiple, 13-14





Scalar product, 28





properties of, 29




Scalar triple product, 65




Stokes' theorem, 296




applications of, 302-304




Strain tensor, 369




Stress tensor, 339

Surface



area, 266-267




orientable, 261




parametric representation of, 256



smooth, 261




Surface integrals




of scalar fields, 272-273




of vector fields, 276




T





Tangent line, 92




Tangent plane, 164, 259




Tangent vector, 91





Tensor, 327, 387, 392, 422





anti-symmetric, 350




gradient, 361




symmetric, 350




Torque, 49, 52




Torsion, 112




Torus, 271




in polar coordinates, 128




Transformation




of cartesian coordinates. 312



of Christoffel symbols, 440




of oblique coordinates, 381





of general coordinates, 422





Transformation laws




for cartesian tensors, 321, 328-333



for general tensors, 422-423




Triangle inequality, 21, 37




Trihedral, 105



Triple product



scalar, 67




vector, 67




U





Unit tangent vector, 91, 98




Unit vector, 5





V





Vector




analytic representation of, 15



binormal, 104




definition of, 2-5




geometric representation of, 5



position. 16




principal normal, 102




pseudo, 326




Vector product, 38, 42




Velocity, 118




in cylindrical coordinates, 136



in spherical coordinates, 139




W





Work, 34-35, 221-222