This action might not be possible to undo. Are you sure you want to continue?
,
,
an informal text
h. m. schey
,
and all that
on vector calculus third edition
Copynght © 1997, 1992, 1973 by W W Norton & Company, Inc
All rights reserved
Pnnted in the United States of Amenca
The text of this book is composed in Times Roman with the display set in Optima Composition by University Graphics, Inc
Library of Congress CataloginginPublication Data Schey. H. M. (Harry Montz), 1930
Div, grad. curl. and all that an informal text on vector calculus I H M Schey 3rd ed
p cm
Includes bibliographical references (pp 16061 and in x
I Vector analysis I Title QA433 S28 1996
515' 63dc20
964942
ISBN 0393969975 (pbk )
W W Norton & Company. Inc. 500 Fifth Avenue, New York, NY lOllO http IIweb wwnorton com
W W Norton & Company, Ltd, 10 Coptic Street, London
I 2 3 4 5 6 7 890
Zusammengestohlen aus Verschiedenem diesem und jenem Ludwig van Beethoven
Gontents
Preface
IX
Chapter I Introduction, Vector Functions, and
Electrostatics 1
Introduction 1
Vector Functions 2
Electrostatics 5
Problems 9
Chapter 1/ Surface Integrals and the
Divergence 11
Gauss' Law 1 I
The Unit Normal Vector 12
Definition of Surface Integrals 17
Evaluating Surface Integrals 21
Flux 29
Using Gauss' Law to Find the Field 32
The Divergence 36
The Divergence in Cylindrical and Spherical
Coordinates 41
The Del Notation 43
The Divergence Theorem 44
Two Simple Applications of the Divergence
Theorem 49
Problems 52
Chapter III Line Integrals and the Curl 63
Work and Line Integrals 63
Contents
Line Integrals Involving Vector Functions Path Independence
The Curl
66 70 75
The Curl in Cylindrical and Spherical Coordinates 82
The Meaning of the Curl 85
Differential Form of the Circulation Law 90
Stokes' Theorem 92
An Apphcatlon of Stokes' Theorem 98
Stokes' Theorem and Simply Connected Regions 100
Path Independence and the Curl 101
Problems 102
Chapter IV The Gradient 114
Line Integrals and the Gradient 114
Finding the Electrostatic Field 121
Using Laplace's Equation 123
Directional Derivatives and the Gradient 130
Geometric Significance of the Gradient 137
The Gradient in Cylindrical and Spherical
Coordinates 140
Problems 143
Solutions to Problems 156
Bibliography 160
Index 162
Vlll
Preface to the Third Edition
If if ain 't broke, don 't fix if
Anonymous
This new edition constitutes a finetuning of its predecessor. Several new problems have been added, two other problems awkwardly worded in the earlier editions have been revised. and a diagram has been corrected. The major change involves replacing the operators div, grad. and curl by the appropriate expressions using the V operator, to hring the text into conformity with mod
em notational practice. I have, however, resisted retitling the book V· , V, V X, and All That.
I wish to express my gratitude to Richard Liu, Stephen Nettel, and Sally Seidel for their useful reviews of the previous edition. I take particular pleasure in thanking those of my readers who
over the years have been good enough to send me comments, criticisms, and suggestions which have contributed significantly to the quality of the text.
Chapter I
Introduction, Vector Functions,
and Electrostatics
One lesson. Nature. let me learn oj thee.
Matthew Arnold
Introduction
In this text the subject of vector calculus is presented in the context of simple electrostatics. We follow this procedure for two reasons. First, much of vector calculus was invented for use in electromagnetic theory and is ideally suited to it. This presentation will therefore show what vector calculus is and at the same time give you an idea of what it's for. Second, we have a deepseated conviction that mathematics in any case some mathe
1
maticsis best discussed in a context which is not exclusively mathematical. Thus, we will soft pedal mathematical rigor. which
Introduction,
we thillk is all obstacle to lealJlillg this subject 011 a first exposUie
Vector Functions, and Electrostatics
to it, and appeal as much as possible to physical and geometric intuition.
Now, if you want to learn vector calculus but know little or nothing about electrostatics, you needn't be put off by our approach; no very great knowledge of physics is required to read
and understand this text. Only the simplest features of electrostatics are involved, and these are presented in a few pages near the beginning: It should 1I0t be all impediment to anyone: III fact,
the entire discussion is based on a search for a convenient method of finding the electrostatic field given the distribution of electric charges which produce it. This is the thread which runs through, and unifies, our presentation, so that as a bare minimum all you really need do is take our word for the fact that the electric field
is an important enough quantity to warrant spending some time and effort in setting up a general method for calculating it. In the process, we hope you will learn the elements of vector calculus.
Having said what you do not need to know, we must now say what you do need to know. To begin with, you should, of course,
be fluent in elementary calculus. Beyond that you should know how to work with functions of several variables, partial derivatives, and multiple (double and triple) integrals. I Finally, you must know something about vectors. This, however, is a subject of which too many writers and teachers have made heavy weather. What you should know about it can be listed quickly' definition of vector, addition and subtraction of vectors, multiplication of vectors by scalars. dot and cross products, and finally,
resolution of vectors into components. An hour's time with any reasonable text on the subject should provide you with all you need to know of it to follow this text.
Vector Functions
One of the most important quantitites we deal with in the study of electricity is the electric field. and much of our presentation will make use of this quantity. Since the electric field is an example of what we call a vector function. we begin our discussion with a brief resume of the function concept.
A function of one variable. generally written y = f(x), is a rule
2
1 Differential equations are used in one section of this text The section is not essential and may be omitted if the mathematics i<; too fnghtening
Vector Functions which tells us how to associate two lIumbels x alld y; givell x,
the function tells us how to determine the associated value of y. Thus, for example, if y = I(x) = x2  2, then we calculate y by squaring x and then subtracting 2. So, if x = 3,
Functions of more than one variable are also rules for associatillg sets of numbers: FOI example, a fUllctioll of thlee variables
designated w = Ftx, y, z) tells how to assign a value to w given x, y, and z. It is helpful to view this concept geometrically; taking (x, y, z) to be the Cartesian coordinates of a point in space, the function F(x, y, z) tells us how to associate a number with each point. As an illustration, a function T(x, y, z) might give the
temperature at any point (x, y, z) in a room.
The functions so far discussed are scalar functions; the result of "plugging" x in I(x) is the scalar y = [tx). The result of "plugging" the three numbers x, y, and z in T(x, y, z) is the temperature, a scalar. The generalization to vector functions is
straightforward. A vector function (in three dimensions) is a rule which tells us how to associate a vector with each point (x, y, z). An example is the velocity of a fluid. Designating this function vex, y, z), it specifies the speed of the fluid as well as the direction of flow at the point (x, y, z). In general, a vector function F(x, y, z) specifies a magnitude and a direction at every point (x, y, z) in some region of space. We can picture a vector function as a collection of arrows (Figure II), one for each point (x, y, z).
z
y
x
Figure 11
The direction of the arrow at any point is the direction specified by the vectOi functiorr, alld its lellgth is plOpOi (iollal to the mag
3
nitude of the function.
A vector function, like any vector, can be resolved into com
Vector Functions, and Electrostatics
Figure 12
ponents, as in Figure [2. Letting i, j, and k be unit vectors along
the X, y, and zaxes, respectively, we wnte
F(x, y, z) = iFxCx, y, z) + jF)(x, y, z) + kF:(x, y, z).
The three quantities Fx' F). and Fz' which are themselves scalar functions of x, y, and z, are the three Cartesian components of
the vector function F(x, y, z) in some coordinate system.'
An example of a vector function (in two dimensions for simplicity) is provided by
F(x, y) = ix + jy,
which is illustrated in Figure [3. You probably recognize this
, /
~ t »
( ( * x
)
It' ~
./ ,
~
Figure 13 4
2 Some wnters use subscripts to indicate the partial derivative. for example. F, = aFlax. We shall not adopt such notation here, our subscripts will always denote the vector component.
Electrostatics
functioll as the positioll vectOi r: Each arrow in the figUie is ill
the radial direction (that is, directed along a line emanating from the origin) and has a length equal to its distance from the origin.' A second example,
iy + jx
G(x, y) = v' 2 2 '
X + Y
is shown in Figure [4. You should verify for yourself that for this vector function all the arrows are in the tangential direction
x
Figure 14
(that is, each is tangent to a circle centered at the origin) and all have the same length.
Electrostatics
We shall base our discussion of electrostatics on three experimental facts. The first of these facts is the existence of electric charge itself. There are two kinds of charge, positive and negative. and every material body contains electric charge," although
J Note that hy convention an arrow is drawn with its tail, not iff hearl, at the
5
point where the vector function is evaluated
4 Purists will point out that neutrons, neutral pi mesons, neutrinos. and the like do not contai n charge.
Introduction, Vector Functions, and Electrostatics
Figure 15
often the positIve and IIegative chmges me pleseIlt III equal
amounts so that there is zero net charge.
The second fact is called Coulomb's law, after the French physicist who discovered it. This law states that the electrostatic force between two charged particles (a) is proportional to the product of their charges, (b) is inversely proportional to the
square of the distance between them, and (c) acts along the line joining them. Thus, if qo and q are the charges of two particles a distance r apart (Figure 15), then the force on qo due to q is
F = K qqo A
2 u, r
where fi is a unit vector (that is, a vector a length 1) pointing from q to qo, and K is a constant of proportionality. In this text we'll use rationalized MKS units. In that system length, mass, and time are measured in meters, kilograms. and seconds. respectively, and electric charge in coulombs. With this choice of units
K = (1/41TEo), where the constant Eo, called the permittivity of free space, has the value 8.854 x 1012 coulombs/ per newtonmeters", and Coulomb's law reads
6
(II)
You should convince yourself that the familiar rule "like charges repel, unlike charges attract" is built into this formula.
The third and last fact is called the principle of superposition.
If F 1 is the force exerted on qo by ql when there are no other chmges nearby; aIld F2 is the fOice exerted 011 qo by q2 whell
there are no other charges nearby. then the pnnciple of superposition says that the net force exerted on qo by ql and q2 when
Electrostatics
they are both plesellt is the veetOi sum F I + F 2' This is a deepel
statement than it appears at first glance. It says not merely that electrostatic forces add vectorially (all forces add vectorially), but that the force between two charged particles is not modified by the presence of other charged particles.
We now introduce a vector function of position which will play
a leading role in our discussion. It is the electrostatic field, denoted E(r) and defined by the equation Err) = F(r)lqo, or F(r) qoE(r). That is, the eleetlOstatic field is the fmee pel ullit
charge. From Equation (11) we have
F(r) I a :
E(r) =  =  2' u.
qo 41TEo r
(12)
This is the electrostatic field at r due to the charge q.
A natural extension of these ideas is the following. Suppose we have a group of N charges with ql situated at r., q2 at r2, ••• , qN at r N' Then the electrostatic force these charges exert on a charge qo situated at r is
(13)
where 0, is the unit vector pointing from r, to r. From Equation (13) we have
(14)
This is the electrostatic field at r = ix + jy + kz produced by the charges ql at r, (/ = l , 2, ... , N). Equation (14) says that the field due to a glOup of ehruges is the veetOi sum of the fields
each produces alone. That is, superposition holds for fields as well as forces. You may think of the region of space in the vicinity of a charge or group of charges as "pervaded" by an electrostatic field; the net electrostatic force exerted by those charges on a charge q at a point r is then qE(r).
You may be a bit mystified about our bothering to introduce a new vector function, the electrostatic field, which differs in an app3lently tlivial way nom the eleetlostatic force: Thele are two
7
major reasons for doing this. First, in electrostatics we are interested in the effect that a given set of charges produces on another
Introduction,
set. This plOblem call be cOllvellielltly divided illto two parts by
Vector Functions, and Electrostatics
introducing the electrostatic field, for then we can (a) calculate the field due to a given distribution of charges without worrying about the effect these charges have on other charges in the vicinity and (b) calculate the effect a given field has on charges placed in it without worrying about the distribution of charges that pro
duced the field. In this book we will be concerned with the first of these.
The second reason rOl intIOducing the electr ostatic Held is
more basic. It turns out that all classical electromagnetic theory can be codified in terms of four equations, called Maxwell's equations, which relate fields (electric and magnetic) to each other and to the charges and currents which produce them. Thus, electromagnetism is afield theory and the electric field ultimately
plays a role and assumes an importance which far transcends its simple elementary definition as • 'force per unit charge."
Very often it is convenient to treat a distribution of electric charge as if it were continuous. To do this, we proceed as follows. Suppose in some region of space of volume d V the total electric
charge is dQ. We define the average charge density in d Vas
_ dQ P.1V = d V .
([5)
Using this, we can define the charge density at the point (x, y, z), denoted p(x, y, z), by taking the limit of P.1V as d V
shrinks down about the point (x, y, z):
p(x. y, z) =
lim
dQ dV
lim P.1V.
(16)
about (X.).Z)
about (A >.Z)
The electric charge in some region of volume V can then be expressed as the triple integral of p(x. y, z) over the volume V; that is.
Q = I I Iv p(x, y, z) dV.
8
In much the same way it follows that for a continuous distribution of charges, Equation ([4) is replaced by
Problems
E(r) = _1_ f f f p(r')u(r') dV'.
41TEo v [r  r ']'
(17)
PROBLEMS
II Using arrows of the proper magnitude and direction, sketch each of the following vector functions.
(a) iy + jx. (e) jx.
(b) (i + j)/Y2 . (f) (iy + jx)/V x2 + y2 , (x, y) oF (0, 0).
(c) ix  jy. (g) iy + jxy
(dj iy (h)i+jy
/2 Using arrows, sketch the electric field of a unit positive charge situated at the origin. [Note. You may simplify the problem by con
fining your sketch to one of the coordinate planes. Does it matter which plane you choose?]
/3 (a) Write a formula for a vector function in two dimensions which is in the positive radial direction and whose magnitude is 1.
(b) Write a formula for a vector function in two dimensions whose direction makes an angle of 45° with the xaxis and whose magnitude at any point (x, y) is (x + y)2.
(c) Write a formula for a vector function in two dimensions whose direction is tangential (in the sense of the example on page 5) and whose magnitude ar any point (x, y) is equal to its distance from the ongin
(d) Wnte a formula for a vector function in three dimensions which is in the positive radial direction and whose magnitude is
I.
/4 An object moves in the xyplane in such a way that its position vector r is given as a function of time I by
r = ia cos WI + jb sin wt,
where a, b, and ware constants.
(a) How far is the object from the origin at any time t?
(b) Find the object's velocity and acceleration as functions of time.
(c) Show thai the object moves on the elliptical path
9
/5 A charge + 1 is situated at the point (1, 0, 0) and a charge 1 is situated at the point (1, 0, 0). Find the electric field of these two charges at an arbitrary point (0, y, 0) on the yaxis.
Introduction, Vector Functions, and Electrostatics
I 6 Instead of u!iing arrows to represent vector functions (as in Prob
lems 11 and 12), we sometimes use families of curves called field lines. A curve y = y(x) is a field line of the vector function F(x. y) if at each point (xo, Yo) on the curve, F(xo, Yo) is tangent to the curve (see the figure).
x
(a) Show that the field lines y = y(x) of a vector function F(x, y) = iF,(x. y) + jF,(x. y)
are solutions of the differential equation
dy F,(x, y)
= ___,;,_;,__;._
dx F,(x, y)
(b) Determine the field lines of each of the functions of Problem II Draw rhe field lines and compare with the arrow diagrams of Problem 1 1
10
Chapter /I
Surface Integrals
and the Divergence
Oh, could I jiow like thee. and make thy stream
Mv great example .
Sir John Denham
Gauss'Law
Since the electrostatic field is so important a quantity in electrostatics, it follows that we need some convenient way to find it, given a set of charges. At first glance it might appear that we solved this problem before we even stated it, for, after all, do not Equations (14) and ([7) provide us with a means of finding E? The answer is, in general, no. Unless there are very few charges in the system and/or they are arranged simply or very symmetrically, the sum in Equation (14) and the integral in Equation
1 1
(17) are usually prohibitively difficultand frequently impossibleto perform. Thus, these two equations provide what is
Surface Integrals
usually only a "formal" solution' to the problem; not a plactical
and the Divergence
one, and we must cast about for some other way to calculate the field E.
In the course of this casting about, we come inevitably to that remarkable relation known as Gauss' law. We say "inevitably" because it is hard to think of any other expression in elementary
electricity and magnetism containing the electric field [apart, of course, from Equations (14) and (17), which we have already rejected]. Gauss' law is
f r E· 0 dS = !L .
Js Eo
(111 )
If you don't understand this equation, don't panic. The lefthand side of this equation is an example of what is called a surface integral, an important concept in vector calculus and one that is probably new to you. The integrand of this integral is the dot product of the electric field and the quantity 0, which is called a
• 'Unit normal vector" and is probably also unfamlhar. We are about to discuss both surface integrals and unit normal vectors in excruciating detail, and one of our main reasons for quoting Gauss' law at this point in our narrative is to motivate this discussion.
We won't stop here to derive Gauss' law, since the derivation wouldn't mean much to you until you have read the next few sections. Then you can consult almost any text on electricity and
magnetism for the gory detaIls. And If you can contain yourself, wait until we've discussed the divergence theorem (pages 44 52), after which you will be able to derive Gauss' law easily (see Problem 1127).
The Unit Normal Vector
One of the factors in the integrand in Gauss' law [Equation (III)] is a quantity designated 0 and called the unit normal vector. This quantity is part of the integrand in most if not all of the surface integrals we'll encounter; furthermore, as we'll see, it plays an important role in the evaluation of surface integrals even
12
I The word "formal" in this context is a euphemism for' 'useless ..
The V"it .NlJmutJ when .it does IIDt appear explicitly. Thus,. before discussing surVector fnceIntegrals themselves .. we'Il dispose of the questions of bow this, vector function is defined and calculated.
Theword "normal' ~ in the present context means, loose l y speaking, uperpendicular. I. Thus, a ",ector N nonnal to the xyplane is dearly one parallel te the zaxis (Fi.gure III.). while a
vector normal to a sPherical surface must be,m dle radial directi()n (Figure 112). To give a precise defimition of a vector norma] to
N
Figure 112
a surface, consider am arbitrary surface S as Shown in Figure .11 3 .. Construct two nOJloollinearvectors II andvtangenl to S at some point P. A vector N which isperpendi.cular to bothu and.
13 vat P is, by definit~on .. normal to Sat P. Now. as we know. the
Surface Integrals
N
and the Divergence
Figure 113
vector product of u and v has precisely this property; it is perpendicular to both u and v. Thus, we may write N = u X v. To make this a unit vector (that is, one whose length is I) is simple: we just divide N by its magnitude N. Thus,
fi
Nux v
 :~
N [u X vi
is a unit vector normal to S at P.
To find an expression for ft, we consider some surface S given
by the equation z = j(x, y); see Figure 114. Following the pro
z
z =/(x, y)
y
, ,
I
I
I
I
I
I II
II
Figure 114
cedure suggested by the above discussion, we'll find two vectors u and v whose cross product will yield the required normal vector ft. For this purpose let's construct a plane through a point P on S and parallel to the xzplane, as shown in Figure 114. This plane intersects the surface S in a curve C. Vie construct the vector u
14
tangent to Cat P and having an xcomponent of arbitrary length
Th 8· ... t
tc; The z=comporrent of u is (ajlax)u:x, ill this explessioll we use
the fact that the slope of u is, by construction, the same as that
z
c
x
Figure IIS
of the surface S in the xdirection (see Figure 115). Thus,
u = iu" + k (:~) u, = [i + k (:~) ] u.,
(112)
To find v, the second of our two vectors, we pass another plane through the point P on S, but in this case parallel to the yzplane
(Figure 116). It intersects S in a curve C', and the vector v can
z
.I I I I I I
: 'tI I
,
I I I I I I I I
I
y
I I
________________ J
Figure 116
now be constructed tangent to C' at P with a ycomponent of arbitrary length vy• Arguing as above. we have
(113)
Using the two vectors u and v as given in Equations (II 2) and
15
(113), we now construct their cross product. The result,
Surface Integrals
and the Divergence
is a vector, which as we stated above. is normal to S at P. To make a unit vector of this. we divide it by its magnitude to get
n(x, Y. z)
u X V
. al . al 1  J  + k
ax ay
(11=4)
[u x vi
This, then. is the unit vector normal to the surface z = I(x, y) at the point (x. y, z) on the surface.2 Note that it is independent of
the two arbitrary quantities u, and v y.
A couple of examples may be in order here. First a trivial one:
What is the unit vector normal to the xyplane? The answer. of course, is k (see Figure III). Let's see how Equation (114) provides us with this answer. The equation of the xyplane is
z = I(x, y) = 0,
whence we have the profound observations
apax = 0
and
apay = o.
Substituting these in Equation (IJ4), we get fi = krv'1 = k, as advertised.
As a second example, consider the sphere of radius I centered at the origin (Figure 112). Its upper hemisphere is given by
whence
1 The uniqueness of our result [Equation (114)] may be questioned on two counts The first of these is a sign ambiguity If 6 is a unit normal vector, so is 6 The matter of which sign to use is discussed below. The second question anses from the fact that the two tangent vectors u and v used in determining 6 are rather special, since each is parallel to one of the coordinate planes Would
16
we get the same result using two arbitrary tangent vectors? This issue is considered in Problem IV26. where it is shown that 6 as given by Equation (114) is, apart from sign, indeed unique
Dejinititm of Smfttce Integrals
iJf x
,_.:;:: 
ax z
and.
fjf y
 ........
Using these in Equation (U4) lleads to
ix jy . ++k
·'+··+kz
fi =.~z==z='. == _ LX . AY. . .•.. " , = ix +" + kz
Vx2 + )12 + z:l: . JY.
where we have used the equarion of the unit sphere Xl + yZ + .z? :: I. This is" as expected. a vector in lli,e .radial, direction (see Figure U2). To show that its length is I,. we observe that i\ ·8 := x2 + y + Z2 = I.
With the matter of ttle unmt normal veetor now disposed of~ we turn to our next task, a discussion of surface integrals.
Definition of Surface Integrals
We now define fhe surfaeeintegeal of the normal component of a vectos funct~o.n F(x. )' ... z), This quantity is denoted. by
(U5)
and as you can see. Gauss" law [Equati()n (UI)) is expressed in terms of just such an integral. Let z = f(x. y) bethe equation of some surface, We "II consider a Ii.mited. portion of this surface, which we designate S (see Figure 117}. .. Our fi.fSt step in fo.rmu·
17 Figure 07
Surface Integrals
lating the definition of the sUiface illteglal (II 5) is to approxi=
and the Divergence
mate S by a polyhedron consisting of N plane faces each of which is tangent to S at some point. Figure 118 shows how this approx
z
y
Figure 118
imating polyhedron might look for an octant of a spherical shell
We concentrate our attention on one of these plane faces, say the lth one (Figure 119) Let its area be denoted ilSI and let
Figure 119
(XI' YI' 21) be the coordinates of the point at which the face is tangent to the surface S. We evaluate the function F at this point and then form its dot product with fi{. the unit vector normal to the lth face. The resulting quantity, F(x{, Yt. 2{) • fit. is then multiplied by the area 6.S{ of the face to give
18
We carry out this same process for each of the N faces of the
Definition of appI ox. imating polyhedloll and then fOlIII the sum ovel all lv'
Surface Integrals faces:
N
2: F(x/, y/, z/) • ii/ 6.5/.
/=1
The surface integral (115) is defined as the limit of this sum as the number of faces, N, approaches infinity and the area of each
face approaches zero. ~ Thus,
I L F' il dS = li';'! #. F(x,. y,. z.) • il, ss,
each ~Sr0
(116)
If we want to cross all the t's and dot all the i's; this integral, strictly speaking, should be written
rr
J J~ F(x, y, z) • ii(x, y, z) d5
since both F and ii are in general functions of position. We prefer, and where possible will use, the less cluttered notation f f sF' ii d5 with the arguments of the functions understood.
The surface 5 over which we integrate a surface integral can be one of two kinds: closed or open. A closed surface, such as a spherical shell, divides space into two parts, an inside and an
outside, and to get from inside to outside, you must go through the surface. An open surface, such as a flat piece of paper, does not have this property; it is possible to get from one side of the sheet to the other without going through it. The definition of surface integrals given in Equation (116) applies equally well to
both closed and open surfaces. However, the surface Integral IS not welldefined until we specify which of the two possible directions of the normal we are to use (see Figure 1110). In the case of an open surface, the direction must be given as part of the statement of the problem. In the case of a closed surface, there is a gentlemen's agreement which specifies the direction once
l The statement' 'each t1S,  0" is not quite precise The area of a rectangular patch, for example. might tend to zero becallse its width decreases while its
19
length remains hxed. This would not be acceptable Here and elsewhere we must interpret "each t1S,  0" to mean that all linear dimensions of the lth patch tend to zero
and the Divergence
Figure 1110
andforall: the normal is chosen so that it points outward from the volume enclosed by the surface.
The integral in Gauss' law [Equation (III)] is taken over a closed surface. Gauss' law, in fact, says that the surface integral of the normal component of the electric field over a closed surface
is equal [0 the total (net) charge enclosed by the surface,divided by Eo. Below (pages 3236 and Problems IIII, 1112, and 11 13) we'll see how, when the charges are arranged neatly and symmetrically, Gauss' law can be used to determine the electric field. But the thrust of our whole discussion will be to subject Gauss' law to a series of harrowing adventures which eventually
transform it into an expression useful for finding E even when we don't have symmetry to help us.
Sometimes we encounter surface integrals which are a little simpler than the kind we've just defined, although basically they are almost the same. These are surface integrals of the form
LL G(x, y, z) as.
(117)
where the integrand G(x, y. z) is a given scalar function rather than the dot product of two vector functions as in (115) and (11 6). We go about defining this kind of surface integral much as we did above: we approximate 5 by a polyhedron, form the product Gtx; YI' Z/) nSI' sum over all faces, and then take the limit.
I Is G(x, Y, z) d5 =
N
lim 2: ou; v, ZI) 6,5,.
(118)
N+co 1= I
each .:lS,O
As an example of this kind of surface integral, suppose we have a surface of negligible thickness with surface density (that is, IIIass pel unit area) <T(x, y, z). and we wish to determine its total
20
mass. Approximating this surface by a polyhedron as above. we recognize that <T(x1• Yh Z/) tl51 is approximately the mass of the
Evaluating lth face of the polyhedloll alld that
Surface Integrals
N
2: cr(x" v, z,) 6,S,
'=1
is approximately the mass of the entire surface. Taking the limit
lim
N r r
N+co
each ~s,o
we get the total mass of the surface.
An example of an even simpler surface integral of this kind is
I Is dS.
This integral is taken as the definition of the surface area of S.
Evaluating Surface Integrals
Now that we have defined surface integrals, we must develop methods to evaluate them, and that will be our task here. For simplicity we'll deal with surface integrals of the form (II7), where the integrand is a given scalar function, rather than the
slightly more complicated form (115). There will be no loss of generality in doing this for all our results can be made to apply to integrals of the form (115) just by replacing G(x, y, z) everywhere by F(x, y, z) • fi,
To evaluate the integral
f Is G(x, y, z) dS
over a portion S of the surface z = f(x, y) (see Figure 1111), we go back to the definition of the surface integral [Equation (11 8) J. Our strategy will be to relate !lS, to the area !lR, of its projection on the xyplane, as shown in Figure 1112. Doing so, as we'll see, will enable us to express the surface integral over Sin
21
terms of an ordinary double integral over R, which is the projection of S on the xyplane, as shown in Figure IIII.
Surface Integrals llnd the .Divergen,ce
22
figure 11 m 1
figure 1.112
G.
I I I
I "
I I I
tID
Remating flSI toM, is not dmfficuIt if we recall .tlat .as, (like the area of any plane surface) can be approximated to any desired degree of accuracy by a set of rectangles as shown in Figure n~ 13. For this reason we need only find the relation between the
figure 11 m.3
r r ......__~,
, ~ki

~
,
I
I !i
, ,
I
I
_iJ
l I I
I
......... area of a rectangle and its projection on the xypIane. Thus,. con
Surface Integrals
Figure II14
sider a rectangle so oriented that one pair of its sides is parallel to the xyplane (Figure 1114). If we call the lengths of these sides a, it's clear that their projections on the xyplane also have length a. But the other pair of sides, of length b. have projections of length b I, and in general, band b' are not equal Thus, to
relate the area of the rectangle ab to the area of its projection ab', we must express b in terms of b'. This is easy to do, for if 6 is the angle shown in Figure 1114, we have b = b'/cos 6, and so
ab' ab = . cos 6
If we let fi denote the unit vector normal to our rectangle, then we can readily convince ourselves that cos 6 = fi • k where k, as always, is the unit vector of the zdirection. Thus,
ab' ab = . fiok
Since the area Jl.Sr can be approxi mated with arbitrary accuracy by such rectangles, it follows that
23
Jl.S{ = ,.. , Dr· k
Surface Integrals
where, of course, iii is the unit vectOi IIormal to the lth plane
and the Divergence
surface.
We can now rewrite the definition of the surface integral [Equation (118)] as
rr
lim L G(xl, YI' Z/) ;:kl ,
N+ao 1= I nl •
each 6.R,.O
N Il.R
J J s G(x, y, z) as =
(119)
where the statement "each Il.SI ~ 0" has been replaced by the equivalent but now more appropriate "each 6.RI ~ 0." We are now obviously well on the road to rewriting the surface integral over S as a double integral over R. In fact,
lim ~ G(xl• YI' ZI) 6.R
L.J ... k 1
N+QO 1= I nl •
each 6.R,.O
= If G(x. y, z) dx d (1[10)
R ii(x. y, z) • k y,
where flex, Y. z) is the unit vector normal to the surface S at the point (x, Y, z), This is a double integral over R even though it does not quite look like one. What appears to spoil it is that nasty z in G and ii; a double integral over a region in the xyplane clearly has no business containing any z·s. But the zdependence is spurious because (x, y, z) are the coordinates of a point on S, and so Z = I(x, y). Hence, at the expense of making the integral
look even fiercer than it already does in Equation (II 10), we can eliminate the apparent zdependence of the integrand and write
If G[x, y, f(x, y)] dx d
... y.
R n[x, y, f(x, y)] • k
(IIII)
The faint of heart can take courage; in most cases this integrand reduces quickly to something much simpler and pleasanter lookinga fact we will demonstrate by example below. At this point we introduce the expression for the unit normal vector [Equation (114)]. We find
fi • k = r====:======::
"\1"1 + (itfliJX)2 + (itfliJy)2 '
24
and so Equation (IIII) becomes
EmlruJtinc Sur.foce Integrals
f L G(x, y"Z) dS ;: f III G[x. y" f(x" y)]
, ('" ,(Jf)':! (af)"'!
• .1 + 'ax" + ". iJy . dx dy. (1112)
Thus" the surface il1ltegral of G(x, y, z) over the surface S bas been expressed as a double integral of a messy looking fUnW()D over the region R~ the projection of S .ill the xyplane. As we remarked sblM; in practice the integral is usually much less ghastly than it appears written ()U1! in Equations (IIII) or (D 12). YllJuwiD see this in the example we now give.
Let's compute the surface integral
where S is the octant of the sphe.re of ,radius • centered at the origin as shown in FigureUIS. The pro}eCllioo of SOil the xy
y
x
F~gure n~.1,5
plane (that is. R) is the area: enclosed by the quarte.r circle, The equation of S is Xl + y2: + Z2 = 1, or
z ·=",,=,/(x, y)=""= +V1  .x2 )11 •
It foU()ws then that
and
25 so that
Sllrjttcelnt.grols and the Divergent:e
I + (.. af)·· .. · 2 t ( ..•.. ··.€J.f ) •.. :2 = ..
ax· iJy
where we have used x2 + y2 + 1_2= I. Hence,
Substituting for z in terms of x and y"we get
This ism ordinary double integrel, and yOLl should verify that its value is fi/6. [Suggestion: Ccnverl to polar ceerdinates; x ~ T cos 0" and. y = r sin Et The integration is then trivial.]
It should be ern.phasized. thatlhe foregoing discussion was based. enthe assumption that the surface S is described by an equation of the form z = J(x, y); in such a situation a surface imegral is converted imo a double integraW. over a. region in the Jryplane. But it may happen that a given surface is more ccn"leniently described by an equation of.tIe form y = g(xlI z) as in Figure 1I16(a.).lfthis is so, then
26 Figure 1I16(a)
EJ'BlIl4ling I r
Surface Integrals I,,' Js G(x, y •. z) tIS
= If. G[x. g(x. d. z) iI 1 (:!)' + (:~r Ox tk.
where R is a region in the xzplane. Similarly" ifwe have a. surface described by x =1 h(y~ z), as m Figure D16(b), thenwe lise
I Is G(x, y,z) tIS
= n G[h(y. ~j. y. z] I + (::)' + (::r dy dt.
where.R in dlis case Is a region in the yzplane. Finany, a surface
, ,
"
.Figulle U.16.(b)
may have several parts" and it may then be~enien1l to project differeJ1Jt parts on different eeordinate planes.
To cvafuate surface integrals of the form I(US), that is,
we merely replace G by F • Ii ill Equation (U~ 1.2) to gel
fL,·, F • Ii. dS = I" I., ,., .F· ft l + (,' iJ,if,.' ')'.,:2 + (,,' ',a .. ,if, ,',,) •• z Ox dy.
S 'R ' ax ' it),
If we now use Equation (114) to write this out in detail, we find
27 th.at tile squareroot factor cancels and. we gel
Surface Integrals Clllilthe Divergence
28
fl ... F· R tIS = f'/. •.. ,I. F.J[x. y.~. y») a.if.
s w'· ax
We leave ,iI to the reader to write OOWl'l. the anai(),gous f<mnulas when the surface S is given by y = g(x,z) or x = h(y. z), whi,ch mnst be projected onto regions in the Xl and yz~planes. respectively.
This last equation (Equation !(II13)] is enough to make Sirong men weep, but,. as before" in most calculations it quickly reduces to something quite tame. For example, suppose we wish 10 calculale Jls p .. j) as wbereF(x. Y~z)=iz  jy + kx and S is the portion of the plane
x + 2y 1 2,t = 2
bounded by .he coordinate planes, that is" the triangle reclining gracefully in Figure U17(a) .. The normal vector il is chosen so
Rgure II11(a)
that iltl'oimts away from the origin as shown ill Figure UI,7(a.), and well II, project S onto die xyplme. We have
x
z=f(x..y)= lZYt
and so
af
_. =1
ax .,
0/ _
 L iJy
We also have
, Jr
F,.; = Z = I  "2  y~
F,. = y~
F =x
, ;:"
Hence
= Ii {[  (I  i  Y)] (t) + y(I) + x} dxdy = Ii (~ i +0 dxdy.
The region ,R, over which. the integral ruus1t be taken is shown in Figure U17(b). The problem has thus been re<lnoed to the oom
1
Figure 1117(b)
pulation of a radl.er simple double integral, and yvu should carry mit tb.e imtegrati.on yuurself (tb,e answer isl).
Flux
An, integral, of the Iype
f fs F(x. Y. z) • n as
(11.=14)
is sometimes called the uflux ofF.H Thus Gauss' Iaw [Equat~on (III) I states that the flux of the elecaroseatie field over some closed sLlrface is the en.closed charge divided. by f4I.
29 It is llsefuI in obtaining a geometrical feelm.g for seme aspects
Surface Integrals
of vectOi calculus to undel stand the significance of the wmd flux
and the Divergence
(Latin for "flow") used in this context. For this purpose let us consider a fluid of density p moving with velocity v. We ask for the total mass of fluid which crosses an area AS perpendicular to the direction of flow in a time /j.t. Clearly all the fluid in the cylinder of length v /j.t with the patch /j.S as base will cross 6.S
in the interval /j.t (Figure 1118). The volume of this cylinder is
Figure 1118
v 6.t 6.S, and it contains a total mass pv 6.1 6.5. Dividing out the
6.t will give the rate of flow. Thus,
(Rate of flOW) = v ss through /j.S p .
Now let us consider a somewhat more complicated case in which the area ilS is not perpendicular to the direction of flow (Figure II 19). The volume containing the material which will
C)
)
Figure 1119
flow through /j.S in time /j.t is now just the volume of the little skewed cylinder shown in the diagram. The volume is v 6.t /j.S cos 6, where 6 is the angle between the velocity vector valid ft, the unit vectOi lIormal to 6.S alld poilltillg outward hom
30
the skewed cylinder. But v cos 6 = v· ft. So, multiplying by p and dividing by /j.t, we get
Flux
(Rate. • • •.. Q. '.f.". fl .. OW).· = 'If' Ii AS.
through /!S " p .
Finally. consider a surface S ,in some region of spaee containiflgn.owi~g matter (Figure n"::ID). Approximate d1,e surface by a
Figure 1120
polyhedrm,. By the above argument. ttle rate at which, matte,r Hows through lli.e l(h face of tNs polybedroois approximately
H~ of coarse, (Xl" Ylt Zt) are lli.e coordinaaes of the point on the Ml. face at whiCh it is tangent to S. and ft, is the unit veetee normal to tbe lth face .. Summing over a1l the faces and taking the limit,. we get
(Rate d flOW)" f ( .,
through S •• = J. p(x. y •. z)v(x. y,. z) • ft dS.
If S happens to be a closed surface and there is a net rate of flow (Jut of the volume it encloses. then. Y()U can convince yourself tl1at this imegral will be positive. and if there is a net rate of 1:10w in, the integraIwill be negative ..
H in this last equation we put
F(Jt; y, ,z)  p(Jt; y, z)v(x, y~z),
the integral is seen to be formally identical with tllall, ill Equatj,m (U14). For this reason any inte,gral m llie form (U ..... l.4) is called "the flux of F over the surface S.'· even when the ftm.oti()ftF is
31 not the product of a density and a velocity! The reason fer stress
Surface Integrals
iug this poiut about flux is that, misuomer though it may be, it
and the Divergence
nonetheless gives a good geometric or physical picture of Gauss' law: The electric field "flows" out of a surface enclosing charge, and the "amount" of this "flow" is proportional to the net charge enclosed. Warning: This is not to be taken literally; the electric field is not flowing in the sense in which fluid flows. It
is merely picturesque language intended to aid us in understanding the physics in Gauss' law.
Using Gauss' Law to Find the Field
Having rejected the two expressions for E [Equations (14) and (I 7)], we find that the only candidate left for providing us with
a good general method for calculating the field is Gauss' law. At first glance it does not appear to be a very likely candidate because, unlike Equations (14) and (17), it is not an explicit expression for E. That is, it does not say' 'E equals something." Rather, it says "The flux of E (the surface integral of the normal
component of E) equals something." Thus, to use Gauss' law, we must' 'disentangle" E from its surroundings. Despite this, there are situations in which Gauss' law can be used to find the field as an example will now show.
Consider a point charge q placed at the origin of a coordinate system. Symmetry considerations tell us two things about its electric field: (1) It must be in the radial direction (that is, it must point directly toward, or directly away from, the origin), and (2)
it must have the same magnitude at all points on the surface of a sphere centered at the origin. Stating this in symbols, we have E = erE(r), where er = rlr is a unit vector in the radial direction. Thus, Gauss' law becomes
f L E(r)er • fi dS = qlEo·
If, for the surface S, we now choose a spherical shell of radius r centered at the origin, a lillie thought will convince you that fi = e so that fi • e = I and we get
r' r
f Is E(r) dS  qlEo·
32
This integral is trivial to perform if we recognize that r is a con
Ust (j ,
smgauss Law to Find the Field
stant over the spherical surface S. This means that E(r) is also a
constant on S and we ger'
f L E(r) dS = E(r) f L dS = 41TrE(r) = q/Eo,
whence
q E(r) = 41TEo r
and
33
'" e,. q
E(r) = e,E(r) =  ..:1 ' 41TEo I
in agreement with Equation (12),
We can see from this example how heavily we depend on sym
metry when using Gauss' law to obtain the field. In fact, to use Gauss' law in the form given in Equation (III) requires even more symmetry and simplicity than Equations (14) and (17), The blunt truth is that this form of the law yields the electric field in a grand total of three situations (and combinations thereof): (I) a spherically symmetric distribution of charge (of which the point charge considered above is a special case), (2) an infinitely long cylindncally symmetric distribution (including the case of
an infinitely long uniformly distributed line of charge), and (3) an infinite slab of charge (including as a special case an infinite uniformly charged plane)." The real value of Equation (III) is that it can be twisted and beaten into a more useful form.
What is it about Equation (III) that makes it difficult to find E? To answer this question. suppose we are doing a numerical
calculation on a computer and wish to evaluate f f s E • fi dS. The standard procedure for dealing with integrals numerically is to approximate them as sums, a rather obvious thing to do since an integral, after all, is the limit of a sum. Thus, suppose we divide the surface S into, say, 10 patches. We then have as an approximation to Equation (III)
4 Shortcuts like this often make jt possihle to evaluate surface integrals without
using all the paraphernalia we discussed above. Further examples are gi ven in Problem 1110.
~ Examples of these are given in Problems IIII. 1112. and 11 I3
Surface Integrals and the Divergence
10
L EI • 01 ASI ==: q/EO'
;=1
where EI is the value of E, and 01 is the unit normal, somewhere on the Ith patch. There is little or no hope of finding E from this:
it is one equation in the 10 unknowns E I' E2, ••• , E 10' Furthermore, it is probably not very accurate. To improve the accuracy, we might make 100 subdivisions rather than just 10 to get
100
L EI • 01 ASI ==: q/Eo·
/""1
Much more accurate! And much more hopeless, too, because this
is one equation in 100 unknowns. Even more accurate (and more hopeless) is
f Is E • 0 dS = q/Eo,
which is one equation in infinitely many unknowns. These unknowns are, of course, the values of E • 0 at everyone of the infinitely many points of the surface S.6
We have now isolated the trouble with Equation (III): it involves an entire surface and therefore the value of E . 0 at infinitely many points. If, somehow, we could deal with the "flux at a single point" (whatever that may mean!) rather than the flux through a surface, perhaps then Gauss' law might yield some
thing tractable. How might we arrange this? For simplicity let us surround some point P by a set of concentric spherical shells S I' S2' S3' and so on (Figure 1121), and calculate the flux <1>1' <1>2' <1>3' and so on. through each shell. We might then attempt to define the "flux at the point P" as the Jimiting value approached
by the sequence of fluxes calculated this way over smaller and smaller shells centered at P.
This sounds good; it has a heartening • 'mathematical" ring to it. Unfortunately, it does not work because (assuming the charge density is finite everywhere) the sequence of fluxes. calculated as described above, approaches zero for any point P. This is fairly obvious since it is merely the statement that the flux through a
34
b The reason Gauss' law yields the expression for the field of a point charge examined above is that symmetry in that case shows the infinitely many unknowns are all equal. This turns Gauss' law into one equation in one unknown.
Using Gauss' Low to FilUl the Field
G
surface tends to zero as the surface shrinks to apoint, Since cur objective was to find a way to determine: tl1.e flux. at a point, and thereby learn someti1ing about the ti.eld at that point, and since we get zero at any point IlO matter what the field there may be, we have obvi.ously net obtained what we want.
It. is useful to give a physicist's roughand .... ready proof of the fact that lli.eH.ux goes to zero as the surface shrinks down to 11 point, for even though this fact: may be ()bvious, the proof will suggest how to puU this ches'tnut out of lli.e 'Ire. Por dlis purpose we n6te that if P4V denotes the average density of electric charge [Equationl_:5] if!. some region orwlume fj, Vt tll.en the ttl.a1 charge in A V is PAil a V. Thus Gauss' law [Equatiml. (U1U may be written
(1115)
where. as indi.cated in Figure, 1I~ the surface integral is taken
s
figure 1122
over the surfa.ce Swhiob encloses the vehnne AV. From this
35 exprlessioJ\l [Equation 0115)] we can see the validity of our
Surface Integrals and the Divergence
assertioIl: As S ~ 0, the ellclosed volume av must, of course,
also approach zero. Thus, the flux also tends to zero and the assertion is proved.' We have not only given a proof, but (and this is the point) we can now isolate a quantity which does not vanish as S ~ O. Dividing Equation (1115) by A V, we get
This expression, awkward and unappealing though it may be, is nonetheless close to what we are after, even though it still involves an integral of E over an entire surface. For if we now take the limit as S shrinks to zero about some point in A V whose coordinates are (x, Y. z), then. as we see from Equation (I 6), the
average density PIlV approaches p(x, y, z). the density at (x, y, z), and we get
lim
_1_ f r E· ft dS = p(x, y, z)/Eo. av Js
(1116)
aboul (x.y,zl
This expression is admittedly downright hideous and whether it will be of any practical use whatever depends on our being able to pound the lefthand side into a form which looks and acts at least halfcivilized. We tum to this task now.
The Divergence
Let us consider the surface integral of some arbitrary vectorfunction F(x, y, z):
We shall be interested in the ratio of this integral to the volume enclosed by the surface S as the volume shrinks to zero about some point, for that is exactly the type of quantity which appears in Equation (1116). This limit is important enough to warrant a special name and notation. It is called the divergence ofF and is designated div F. Thus.
36
7 This line of reasoning and the conclusion must be altered if the system contains point charges
37
divF ;;;;;;.
_!_. . f ( F· Ii liS. ~V . Js
(1117)
!lim
AIlO ~ (.r,..\.d
This quantity is clearly a scalar, Furthermore, i.t will. im general. have different values at differelltpoints. (x. y. z), Thus th.c diver~ gen.re of a.vectee ftOle.ion is a scalar /lmCtion of positiOll ..
Equation (U16) cam now be written
div E = P!Eo.
AI this stage.,howev«~ our fancy new notatlon bas only a cosmetie value., .he.lping to beautify an ugl.y equatioo.. W.heth.er it bas any practical value as well.is the matte.[ taken up in the following discus:siol1. in whiCh we actuaUy cill.culaJe tb.e limit of l:lhe ratio of fiux to encl.()sedvolume and find tbat it can. be expressed reasonably simply in terms of certain par1!iaJ derivatmves. Before lum.m.g to this calculation .. however; il.'·S worth mentioning that if we take our new tem1inology IiteraUy" we camin'erprel. EQUation (1I:18) to mean that the field .... d.iveqes'''' from a. point •. and how much it diverges. so to speak" depends on how much charge dlere is at: that point as represented by die density there,
Our next order of business is to find the reasenably simple expression fur the: divergence of a vector function promised above, Thus, eonsider a small rectangular paraUelepriped.8 with edges of length i:u, .6y. and llz paraille) to the eeordinate axes (Fig1lllre ll23) .. Let the point at the ,center of the little cuboid have
~ :) •
y
s.
HgureU23
(I HencefMhwe'U refer to this as a "cuboid," a madetlp, term thM Itllkes tess time and space. Ihan lhe sesquipedalian '''rectangular parallelepiped·'
Surface Integrals and the Divergence
cOOidillates (x, y, z): We calculate the surface illtegl al of F ovel
the surface of the cuboid by regarding the integral as a sum of six terms, one for each cuboid face. We begin by considering the face marked S I in the figure. We want
Now it is clear the unit vector normal to this face and pointing outward from the enclosed volume is i. Thus, since F • i = Fx' the above integral is
(i F,,(x, y, z) as.
By assumption the cuboid is small (eventually we shall take the limit as it shrinks to zero). We can therefore calculate this integral approximately as F" evaluated at the center of the face S I multiplied by the area of the face.? The coordinates of the center of
S I are (x + ~/2, y, z), Thus,
I L, F,(x, y. z) dS ~ F, (x + a; . y, z) Ay lJ.z.
(1119)
The same kind of reasoning applied to the opposite face S2 [whose outward normal IS j and whose center is at
(x Axl2, y, z)] leads to
38
(1120)
Adding together the contributions of these two faces [Equations (1119) and (1120)], we get
g The rationale behind this is as follows: There is a mean value theorem. which telIs us that the integral of F. over SI is equal to the area of SI multiplied by the function evaluated somewhere on SI' Since SI is small. the point where we should e\laluate F. and the point where we do evaluate it (thai is, the center) must be
close together. and F. must have nearly the same value at the two points. Hence what our procedure gives us is a good approximation to the value of the integral. Furthermore. as the cuboid shnnks to zero, the two points get closer and closer so that in the limit our result [Equation (1122)] will be exact.
The Divergence
= [F, (X + t.; , y, Z)  F, (X  t.; , y, Z) ] Ay Az
F.( x + t.; , y, Z)  F, (X  t.; , y, Z)
Ax Ax Ay Az.
Recognizing that Ax Ay AZ = AV, the volume of the cuboid, we have
We now must take the limit of this as AV approaches zero. 10 But. of course, as A V goes to zero, so do each of the sides of the cuboid. Thus, on the righthand side of Equation (II21) we can
write lim6,.. o in place of lim6,v __ o and we find
I r r
F, (x + t.; , y, Z) (x  t.; , y, Z) aF,
= lim  =
..
evaluated at (x, y. z). This last equality follows from the definition of the partial derivative. It should come as no surprise that the other two pairs of faces of the cuboid contribute aF/ay and aFJaz. Thus,
I r r ,., sr, aF, aFz
lim  F • n dS   + __... +  .
39
10 Note that we have postponed calculating the contnbutions from the other four faces of the cuboid.
Surface Integrals
The limit 011 the lefthand side of this last equation, as we have
and the Divergence
already remarked, is the divergence ofF [Equation (1117)]. Thus we have just demonstrated that
(1122)
It can be shown that this result is independent of the shape of the
volume used to obtain it (see Problem 1117).
Using Equation (1122) to find the divergence of a vector function is a straightforward matter, but we'll give an example just for the record. Consider the function
F(x, y, z) = ix2 + jxy + kyz.
We have
aF~
ax
2x,
___.!.  x
ay •
aE
and
yo az
Thus.
div F = 2x + x + Y = 3x + y.
Returning now to the electrostatic field. we combine Equations
(1118) and (1122) to get
ee, aE) et:
 + +  = p/Eo·
ax ay az
(1123)
This equation. which is much more general than our derivation of it suggests, is one of Maxwell's equations and is completely equivalent to Gauss'law [Equation (III)]. It is sometimes called the • 'differential form" of Gauss' law.
We have now arrived at our goal (almost') for we have related a property of the electrostatic field at a point (that is, its divergence) to a known quantity (the charge density) at that point. In
40
all fairness it should be said that Equation (1123) can in a sense be regarded as a single (differential) equation in three unknowns. (Ex. E" EJ and for this reason is not often used in this form to find the field. It turns out, however, that the three components of
The DiFBrgsllcB in Cylindrie4Z1 and Spherical Coordinates
E can be related to each other very elegantly; when we develop that. fieiati.on!iliip, we shaU return to this qmesocn of finding a eenveniera means .of calculating E.
The Divergence in Cylindrical and Spherical Coo.rdinates
One often sees Eguatioo (n~22) given as the definiticn of the divergence of the vector function F. While tbis is certainly acceptable, we much prefer to define the div;ergence as thelirnit of flux to volume as stated in Equation (111.6). Equation (1122) is then mer1el.y the fo:rm the divergence takes in Cartesian. eoordinates. In. other coordinate systems .it I.oots qui.te different. For example, in cylindrical coordinates tI1.e I:lmct~on F bas three components, which yen will not be shocked to learnare designated .FT, .Fe. and F:_ [see Figure Il24(a)ll. To obtain the divergen.oo
F.
.1
A I I
; __ ._~~)F,
" ....
~Fr
y
Figure n....:24(a)
of F in icylindrical. coordinates.we consider the "'cylindrical cuboid' , shown in Figure 024(b) with volume AV = r A,. A6 Az
41 Figure U2.4(b)
Surface Integrals and the Divergence
and center at the point tr; 8, ,,) .. ' Tire flux ofF tillough the face
marked" I" is
= F, (r + ~ , 6, z) (r + ~) L\6 L\~
while through the face marked "2" it is
 F,(r ~,6,Z) (r ~r)L\6L\z,
Adding these two results and dividing by the volume AV of the cuboid, we find
42
which in the limit as Ar (and therefore AV) approaches zero becomes
II Note that in the Cartesian case (Figure 1123) each face of the cuboid is given
by an equation of the form x = constant, Y = constant, or Z = constant In the same way each face of the surface in Figure 1124(b) is given by an equation of the form r = constant, 6 = constant, or Z = constant.
The DellvTotation Arguillg in all allalogous way for the othel foUl faces (see Prob
lem 1118), we arrive finally at the expression for the divergence in cylindrical coordinates:
. 1 a 1 aFe aF.
div F =   (rFr) +   + _' .
r ar r as az
(1124)
In spherical coordinates where the components of F are Fr.
Fa. and r; (see Figure 1125) sImIlar reasonmg (see Problem
y
x
Figure 1125
1121) leads to the expression
1 a a 1 aF
div F = "2  (r2Fr) + .  (sin 9 Fa) + . 9 ~q,.
r ar r SIn S as r SIn a
(1125)
The Del Notation
There is a special notation in terms of which the divergence may be written. There would be little or no reason for introducing it if it served only to provide another way of writing "div," but as we shall soon see. it has considerable usefulness in vector calculus.
Let us define a quantity designated V (read "del") by the following rather peculiar looking equation:
43
V.a .a ka
=I+J+ .
ax ay az
Surface Integrals and the Divergence
If we take the dot product of V and some vector function F
v . F = ( i ~ + j ~ + k ~ ) . (iF + jF + kF.)
ax ay az x \ 
Now we interpret the' 'product" of alax and F, as a partial derivati ve; that is,
There are similar equations forthe two other "products" (aJay)Fy and (alaz)Fz• With this convention we recognize V· F (' 'del dot F") as the same as div F, and henceforth, to conform with mod
em notational practice, we shall always use V· F to indicate the divergence. Thus, Equations (1118) and (1123) will be written
V • E = plea.
Mathematicians call a symbol like V an operator. When we "operate" with V by dotting it into a vector function we get the
dIvergence of that functIon, as we have Just seen. In subsequent discussions we shall introduce three other quantities (gradient, curl, and Laplacian) all of which are operators and all of which can be written in terms of V.
The Divergence Theorem
For the remainder of this chapter we digress from the mainstream of our narrative to discuss a famous theorem which asserts a remarkable connection between surface integrals and volume integrals. Although this relation may be suggested by the work we have done in electrostatics, the theorem is a mathematical statement holding under quite general circumstances. It is inde
44
pendent of any physics and is applicable in many different places.
De Divergellce ThefJl'mI
45
It is called the divergence theorem and sometimes Gauss" theerem (not to be ,confused with Gauss t law).
We shamI noC. give amathematicamIy rigorous proof of the diveegenee theorem; such a proof is given in many texts in advanced calctllus. Instead we present here Mothelr physicist" s roughandfle3dy proof. Thus, CUDsider a closed surface S. Subdivide the vohime V enclosed by S arbitrarily into N subvolumes, one of which is si10wn in Figure 0.",;26 (drawn as 81 cube for conven
f~gure n~26
ienee), We begin our proof by assening that the ftux of an al1bitrary vector fUllwon F(x, y, x) through the surface S equals the sum of tire fluxes thrm.l.gh the surfaces of each of the subvolumes:
(1126)
Here 5, is the surface which encloses the subvohsne A V,. To establish. Equation (U26). consider two adjacent subvohones (Figure D~21). Let their common face be denoted. Sfi}" The :flux
Surface Integrals
tlnough the subvolume mruked "I" ill FigUi e II 27 includes, of
and the Divergence
course, a contribution from So, which is
Here 0 I is a unit vector normal to the face So, and by our usual convention, it points outward from subvolume 1. The flux
through the subvolume marked "2" also includes a contribution from So:
The vector O2 is a unit normal which points outward from subvolume 2. Clearly 0 I =  O2, Thus, in forming the sum in Equation (1126), we shall include, among other things, the pair of terms
I I.,o F • 0, dS + I Iso F • O2 dS =
I Iso F • 0, dS  I Lo F • 0, dS = O.
We see that these terms cancel each other and there is no net contribution to the sum in Equation (1126) due to the face So.
In fact this sort of cancellation will obviously occur for any subvolume surface which is common to two adjacent subvolurnes. But all subvolume surfaces are common to two adjacent subvolurnes except those which are part of the original ("outer") surface S. Hence the only terms in the sum in Equation (1126) which survive come from those subvolume surfaces which, taken
together, constitute the surface S. This establishes the validity of Equation (1126).
We now rewrite Equation (1126) in the following curious fashion:
I ( F· 0 dS = £ [i I ( F· 0 dS] ~ VI' (1127)
J s 1= I .£.l VI J s,
46
This clearly alters nothing since we have just multiplied and divided each term of the sum by ~ VI' the subvolume enclosed by
The Divergence
the surface SI' We call 1I0W imagille paltitionillg the 01 igillal vol
Theorem
ume V into an ever larger number of smaller and smaller subvolumes. In other words, we take the limit of the sum in Equation (1127) as the number of subdivisions tends to infinity and each 6. VI tends to zero. We recognize that the limit of the quantity in square brackets in Equation (1127) is, by definition, (V • F)/. that
is, the divergence of F evaluated at the point about which a VI is shrinking, Thus, for each a VI very small, Equation (1127) becomes
f Is F· fi dS = ~ (V· F), d V,.
(1128)
Further, in the limit. this sum is, again by definition. the triple
mtegral of V • F over the volume enclosed by S:
lim
~ (V • F), A V, = f n V • F dV.
(1129)
N oo
each .<1VtO
Putting together Equations (1126) through (1129), we arrive at our result:
(1130)
This is the divergence theorem. In words it says that the flux of
a vector function through some closed surface equals the triple integral of the divergence of that function over the volume enclosed by the surface.
The major reason the proof given above is not rigorous is that a triple integral is defined as the limit of a sum of the form
L g(XI, YI, ZI) 6. VI'
I
where the function g is welldefined. In Equation (1127), however, the quantity multiplying the volume element 6. VI in each term of the sum is not a welldefined function in this sense. That is, as 6. VI tends to zero the quantity in the square brackets changes; it can be identified as the divergence of F only in the
47
limit. A careful. rigorous treatment would show that Equation (1130) is valid if F (that is. F; Fy. and FJ is continuous and
Sruface Integrals and tie DWergence
differeJ1tiabl~ and .its :first derivatives are continuous in V and. 00 s.
Now let~siUustrate the divergence: theorem" Since endless pages of hideous integrals w.ill not serve our purpose~ we"Uuse a simple example. Let
F(x~ y. z) = it I jy + kz
and choose fo.r S the surface shown. in. Figure U28 .• oonsistimg
.R
y
Figure 1128
of the hemisphedcal shell of radi.us I amld tileregi()n R m tbe ~ plane enclosed by the W1it circle, On th.ebemispllere we bave Iii = it + .iY tkz. so that fi • F = x2 + y2 + Z 2 =1. Thus" Oll d1Je hemisphere,
where the lase. equality follows from the fact chat the integral 'is merely the surface area of the unit hemisphere .. On the region R we have ft = ~k so that fi ·.F =  z, Hence, 011 R,
because z = 0 everywhere CD R: ThuS1 there is no contribution
48 to the sunace integral from the circular region R and
Two Simple
I' I'
Applications of the Divergence Theorem
J J5 F • fi dS = 2'1T.
Next we find by a trivial calculation that V • F = 3. It follows then that
I I Iv V • F dV = 3 I I Iv dV =
2'1T
3  = 2'1T
3 '
where we usi the fact that the volume of the unit hemisphere is 2'1T/3. Since the surface and volume integrals are equal, this illustrates Equation (1130).
Two Simple Applications of the Divergence Theorem
As one example of the use of the divergence theorem we give an alternative derivation of Equation (1118), the analysis of which led us to the divergence theorem itself. In other words, this is
how easy it would have been if we had known the divergence theorem to begin with!
We start with Gauss' law in the form
(1131 )
Next we apply the divergence theorem to the surface integral in
the above equation to get
(1132)
Thus, combining Equations (1131) and 1132), we find
If Iv V • E dV = :. If Iv p sv.
In general, if two volume integrals are equal, it is not necessarily true that their integrands are equal. It might be that the integrals are equal only over the particular volume of integration V, and
49
by integrating over a different volume, we would wreck the equality. In the present case, however, this is not true because
Surface' ,lntlJgm1s and the DiryeTgence
Gauss' law holds for any arbitrary volume V,l and we canrI.ot, upset the lequalUy by Changing the volume, But this can be so omy .if the integrands are equal Bence.
v . E =: p/Eo·,
which. should look familiar!
Another ex.ampl.e of the use of lli.e div!ergence theorem. is the following. Suppose tbatin some region of space """stuff"' (matter. electri.c charge, anything) is movimg (Figure II~29). Let .lIe den.
s
Fiigure 029
sity of this stuff at any point. (x" y,. z) and any time r be p(X,. Ylf to' r) and let its velocity be v(x, y~ Z,. t). Further, suppose this stuff is conserved; that. is, it. is neillie.[ created nor destroyed. Conoentrating o:n some arbitrary volume V in space, we ask:
What is the rate at. which the amount of stuff in this volume Is changing? At any t~me t the amount of stuff in Y is
and the .rate at which it is changing is
d ff·I.· I " ff·/. iJp
'~, . I I •. ' p{Xl' y. z, t) dV ~,." . I .... tW.
dt I \" II' • vat
(To be able to move the derivative under the integral sigm this
50 way requtres tI1at aplat be continuous . .)
Two Simple
Next we I ecall flom all earliel discussion that the 1 ate at which
Applications of the Divergence Theorem
stuff flows through a surface S is
We then assert that the rate at which the amount of stuff in V is changing is equal to the rate at which it is flowing through the ellciosillg surface S, ilt'equatioll form this statement reads
I I Iv ~ dV =  I L pv • Ii dS.
There are two features about this equation that require discussion:
1. The negative sign must be included because the surface integral as defined is positive for a net flow out of the volume, but a net flow out means the amount of stuff in the volume is decreasing.
2. This equation states that the amount of stuff in V can change only as a result of stuff flowing across the boundary S. If stuff were being created or destroyed in V, terms would have to be included in the equation to reflect that fact. The absence of any such terms is thus an expression of the conservation of the stuff.
Now, finally, let us apply the divergence theorem. We find
Hence,
I I!v : dV =  I I Iv V • (pv) dV.
Arguing as we did above that V is an arbitrary volume, we can then say
51
ap _ at
V . (pv).
(1133)
Surface Integrals
Usually we define the current density J
pv and write Equation
and the Divergence
(1131) as
ap
 + V· J = O. at
An equation of this type is referred to as a continuity equation and is, as we have seen, an expression of a conservation law (see
Problems III20, 11121, and IV21 ). Besides playing an important role in electromagnetic theory, it is a basic equation both in hydrodynamics and diffusion theory. Finally. considerations similar to those which led to the continuity equation are involved in the analysis of heat flow.
PROBLEMS
III Find a unit vector fi normal to each of the following surfaces.
(a) z = 2  x  y. (c) z = (1  X2)112.
(b) z = (x2 + y2)112. (d) z = x2 + y2.
1I2 (a) Show that the unit vector normal to the plane
ax + by + cz = d
is given by
(b) Explain in geometnc terms why this expression forft is independent of the constant d.
1I3 Derive expressions for the unit normal vector for surfaces given by y = g(x, z) and by x = h(y, z). Use each to redenve the expression for the normal to the plane given in Problem II2.
II~ In each of the following use Equation 1112 to evaluate the surface integral f f. G(x. y. z)dS.
(a) G(x. y, z) = z,
where S is the portion of the plane x + y + Z = I in the first octant.
(b) Gix, y, z) = 1 + 4(xl + I)'
where S is the portion of the paraboloid z = xl + y2 between
52
z  0 and z  I.
(c) G(x, y, z) = (I  xl  y2)1I2.
where S is the hemisphere z = (l  xl  y2)112.
Problems
11 5 In each of the following use Equation II 13 to evaluate the surface
integral I Is F • n dS.
(a) F(x. y. z) = ix  kz,
where S is the portion of the plane x + y + 2z = 2 in the first octant.
(b) F(x, v, z) = Lx + jy + kz,
where S IS the hemisphere z  Va2  Xl  i (c) F(x. y. z) = jy + k,
where S is the portion of the paraboloid z =  x2  i
above the xyplane.
116 The distnbution of mass on the hemispherical shell z = (R2  x2  y2)112
is given by
where <To is a constant. Find an expression in terms of <To and R for the total mass of the shell.
//7 Find the moment of inertia about the zaxis of the hemisphencal shell of Problem 116.
//8 An electrostatic field is given by
E = A(iyz + jxz + kxy).
where A is a constant. Use Gauss' law to find the total charge enclosed by the surface shown in the figure consisting of S [1 the hemisphere
z = (R2  x2  y2)112,
and S2' its circular base in the xyplane.
z
//9 An electrostatic field is given by E = A(Lx + jy). where A is a constant. Use Gauss' law to find the total charge enclosed by the surface shown in the figure consisting of S I' the curved portion of the
53
halfcylInder z  (r2  y2)ln of length h; S2 and S~. the two semicircular plane end pieces; and S4' the rectangular portion of the xyplane. Express your results in terms of A, r, and h
Surface Integrals ondthe ,Divergence
54
IIH) II: sometimes happens that sunacein:tegrals!canbe evaluated withoot using the longwind.ed. procedures olllt1m,ed in the text. 1'I'y evallUlting JI sF" ii dS for each of the following; tl:Jink a bit and avcid a lot ofwork~
(a) iF = Ix + jy + kz.,
s~ me three squares eaeh of side b a'S shown in '1he6gure.
(b) F = (ix + jy) ,JIm (xl! + y'').
S.the ~nnder (btcl,mtingi£hetop and bottom) of radius R and heigb.t h shown .in the, ,figure.
y
(c) IF ""'" (it + jy +ki)e~2+'y+;z\
S. we surface of the sphere of radius 11. ce.ntered at the origin as shown ~nlhe figure.
(d) F = IE(:d~ where E(x) .is a£i arbitrary scalar function of 11:.
S. the surfaee of the cube, of s.ide, .b shown in ilhe, .figul\e
IIII (a.) Use Gauss' law and symmem:ry to find the e'teclooSla1tic field as. a. function of position for an infinite llIliform plane, of charge, Let: the ohrruge lie jn t'he yz~plane and. denote the charge per unit area by <r
(b) Repeat part (a) for an infinite slab of clIarge parallel to the yZ~:p~al1e. Whose density is given by
, . {I~. b < x < b.
IPU) =1 O. Ixl> h.
where P,d) a.nd b are rons.tanu
(c) Repeat part (b) with p(x) = p(lej""li~.
55
111.2 (a.) Use Gauss' law and symme1l:ry to find the electrostatic field. as a. funoniJoDi of position fer aninfinit:e un~form line of' cllarge.. Le1I the charge lie al.ong the zaxis and denote the charge per unit Ienglh by ~
(b) Repeat pad (a) for an infinite cy'~inder of charge wboseaxis
Surface Integrals
coincides with the z=axis and whose density is given in cylin
and the Divergence
drical coordinates by
p(r) = {po, r < b. 0, r ~ b.
where Po and b are constants.
(c) Repeat part (b) with per) = poerib.
ll13 (a) Use Gauss' law and symmetry to find the electrostatic field as a function of position for the spherically symmetric charge distribution whose density is given in spherical coordinates by
p(r) = {po, r < b,
0, r ~ b.
where Po and b are constants.
(b) Repeat part (a) for per) = poe· rib. (c) Repeat part (a) for
eO' r < b,
p(r) = PI' b '5 r < 2b,
0. r ~ 2b.
How must Po and PI be related so that the field will be zero for r > 2b? What is the total charge of this distribution under these circumstances?
11/4 Calculate the divergence of each of the following functions using Equation (1122)·
(a) ixl + jy2 + kz2.
(b) iyz + jxz + kxy.
(c) ie +x + je  ~ + ke  !.
(d) i  3j + kz2•
(e) (ixy + jx2)/(x2 + i), (f) kYx2 + y2 .
(x, y) =1= (0. 0).
(g) ix + jy + kz.
(h) ( iy + jx)I\I~X::2 +y=2 ,
(x. y) =1= (0, 0).
1I15 (a) Calculate IIs F' fi dS for the function in Problem II14(a) over the surface of a cube of side s whose center is at (xo• Yo. zo) and whose faces are parallel to the coordinate planes.
(b) Divide the above result by the volume of the cube and calculate the limit of the quotient as s + O. Compare your result
56
wlth the dlvergence found 10 Problem 1114(a)
(c) Repeat parts (a) and (b) for the function of Problem 1114(b) and (c).
Problems
I1 16 (a) Calculate the divelgence of the function
F(x, y, z) = if(x) + jf(y) + kf(2z)
and show that it is zero at the point (c. c, c/2). (b) Calculate the divergence of
G(x, y, z) iJ(y, z) + jg(x, z) + kh(x, y).
1/17 In the text we obtained the result
V·F=~+:::..:.....l.+___3.
ax ay az
by integrating over the surface of a small rectangular parallelepiped. As an example of the fact that this result is independent of the surface. redenve it using the wedgeshaped surface shown in the figure.
z
y
/IIB (a) Let i, j, and k be unit vectors in Cartesian coordinates and e,. ea, and ez be unit vectors in cylindrical coordinates. Show that
. " e " . e
I = e, cos  ee sm,
j = e, sin 6 + ee cos 6,
k = e ..
(b) Rewrite the function in Problem II14(e) in cylindrical coordinates and compute its divergence, using Equation (II 24). Con
vert your result back to Cartesian coordinates and compare with the answer obtained in Problem II14(e),
(c) Repeat part (b) for the function of Problem 1114(0,
/119 (a) Let i, j, and k be unit vectors in Cartesian coordinates and en ee, and eq, be unit vectors in spherical coordinates. Show that
i = e, sin 6 cos ~ + ee cos 6 cos ~  eq, sin ~.
j e, sin 6 sin ~ + @o cos 6 sin ~ + eq, cos ~.
57
k = e, cos 6  ee sin 6.
Surface Integrals and the Divergence
(H· It. ...... d...· f" d k
dinates and compute its divergence using Equation (II25). Convert your result back to Cartesian coordinates and compare with the answer obtained in Problem I114(g).
(c) Repeat part (b) for the function of Problem II14(h).
//20 In cylindrical coordinates the divergence of F is given by
1 a 1 sr, st:
V • F =   (rFr) +   + _ .
r Br r as az
In the text (pages 4142) we derived the first term of this expression. Proceeding the same way, obtain the other two terms.
//21 Repeat Problem 1120 to obtain the divergence in spherical coordinates by carrying out the surface integral over the surface of the volume shown in the figure and thereby obtaining the expression
1 a 1 a 1 aF
V • F =   (r2F) +  (sin S Fo) + _____!!! •
r2 ar r r sin S as r sin S a<f>
z
y
1122 Consider a vector function of the form
F(r) = erl(r),
58
where er = (ix + jy + kz)/r is the unit vector in the radial direction, r = (x2 + y2 + Z2)1f2, andf(r) is a differentiable scalar function. Using the results of Problem II2It determine f(r) so thai V • F = o. A vector function whose divergence is zero is said to be solenoidal.
Problems
fiF' dS = fffv. FdV
in each of me following cases. (a)F = h ... jy + kz.
s. the, SilIrfllCe, oflhe cube of side b shown ill (he figure,
(b) F = e,r + e~lt r = ix +jy.
S, [ihe surface oflihe quarter C)i~inder (radius R. beigbJl h) shown in the figure
If
(c) F = e.r".
r = ix + jy +kz_
S •. the surface of the ~p.here of radiu.s It centered at the origin as shown in the figure.
59
Surface Integrals and the Divergence
1I 24 (a) One of Maxwell's equations states that V • B
0, whele B
is any magnetic field. Show that
for any closed surface 5
(b) Determine the flux of a uniform magnetic field B through the curved surface of a right circular cone (radius R, height h)
oriented so that B is normal to the base of the cone as shown in the figure. (A uniform field is one which has the same magnitude and direction everywhere.)
/125 Use the divergence theorem to show that
[1 Ii as = 0,
v v
where 5 is a closed surface and Ii the unit vector normal to the surface 5.
1/26 (a) Use the divergence theorem to show that
~ f L Ii . r d5 = V,
where 5 is a closed surface enclosing a region of volume V, Ii is a unit vector normal to the surface 5, and r = ix + jy + kz.
(b) Use the expression given in (a) to find the volume of:
(i) a rectangular parallelepiped with sides a, b. c.
(ii) a right circular cone with height h and base radius R.
60
[Him: The calculation is very simple with the cone onented as shown in the figure]
(iii) a sphere of radius R.
Prublems
.1121 (a) Consider a vector fUl'lclioo with dlepropedY V .. F = o everywhere 011 two ci1.osed surfaces 5, and ~ and in the vol,umeV endosed by them (see the figure). Show mat the flux. of Fthrough Slequa~s the flux of .F Ihrough S2,. In calcul!ating t1teftuxes, choose the direction of the nonnals as indicated by the arrows ~n the figur1e.
(b), G.iven the electroswticfield of a point cl:m.rge q situated at r :;;: Ot
_ .~I~ lj ...
E •. le", 4'l'fE.tI r
where r'" = Xl + y2 + Z!. show by direct adcu1ation that
fer all r ¢ O.
(c) Prove Gauss" law fOT the field of a single point charge g~ven ~n (b). [Hint: It is easy to calculate IbeHux of E over a sphere centered at r = O.J
(d) How would youexlend this proof to CO"ef the case of an arbitrary charge distribution?
Sur/we Integrals 6nd the Dwe,rgence
62
1I~2B (a) Show by direct calcu:mation that the diveJg'el'loe Iheorem does not hold f'Or
F(r. e, 41) == .~, •
r<
wilh S the surface of a sphere of .. mimi, R ,vmkred at the orig~n. and V the enclosed, volume Why does, the theorem fail?
(b) VerIfy by direct calculatien thaI. the divergwve Ihoorem does hold fm Che function If of port (8) w~en S is the surface SI of a splter1e of radius R I plus the surface Sz of a :sphere of radius Rp both centered at Ihe origin .• and V islhe volume enclosed by SI and s;
(c) In general, whatl1eslrio1tion musl be placed on a surface, S so that the, dive,rgelltle theorem will ,held for the fmelion of pal1.(a)?
Chapter III
Line Integrals and the Curl
To err from the right path is common to Mankind.
Sophocles
Work and Line Integrals
We lelll3Jked above that the diffelelltial form of Gauss' law,
Equations (1118) and (1123), although it fulfills our goal of relating a property of the electric field (its divergence) at a point to a known quantity (the charge density) at the same point, nonetheless falls short of providing a convenient way to find E. The reason is that V • E = p/Eo is (or seems to be) a single differential equation in three unknowns (Ex. Ey. Ez). But there is another feature of electrostatic fields which has not yet played an explicit IOle in OUi discussion and which will yield a relationship among
63
the components of E. It will thus provide us with the crucial last step in obtaining a useful way to calculate fields. In the process
line Integrals
of examillillg this question, we shall ellCoulltel some of the most
and the Curl
important topics in vector calculus.
The property of electrostatic fields that we shall now begin to discuss is intimately bound up with the question of work and energy. You no doubt recall the elementary definition of work as force times distance. Thus, in one dimension. if a force F(x) acts
from x = a to x = b, the work done is, by definition,
rb
L F(x) dx.
To be able to handle more general situations, we must now introduce the concept of the line integral.
Suppose we have a curve C in three dimensions (Figure 1111)
and suppose the curve is dIrected. By this we mean that we put
z
Figure 1111
an arrow on the curve and say "This is the positive direction." Let s be the arc length measured along the curve from some arbitrary point on it with s = S I at a point P I and S = S 2 at P 2' Suppose further that we have a function ft», Y, z) defined everywhele 011 C. Now let us subdivide the pOI (ioll of C betweell PI
and P 2 arbitrarily into N sections. Figure 1111 shows an example of such a subdivision for N = 4. Next.join successive subdivision points by chords, a typical one of which, say the lth, has length AsI• Now evaluate f(x, y, z) at (XI' YI. ZI), which is any point on the lth subdivision of the curve, and form the product f(xl, YI' ZI) AsI· Doing this for each of the N segments of C, we form the sum
64
n
2: f(xl, YI' ZI) AsI·
1=1
Work and line
By definition, the Jille illteglal ofj(x, y, z) alollg the curve C is
Integrals
the limit of this sum as the number of subdivisions N approaches infinity and the length of each chord approaches zero:
{, lex, y, z) ds =
.,~
N
lim 2: I(x/, y/, z/) As, .
N ...... OO
To e\laluate the line integral, we need to know the path C.
Usually the most convenient way to specify this path is parametrically in terms of the arc length parameter s. Thus, we write x = x(s), y = y(s), and z = z(s). In such a situation the line integral can be reduced to an ordinary definite integral:
r ft»; y, z) ds = [51 ![x(s), yes), z(s)] ds.
Je 51
An example of a line integral will be helpful here. For simplicity let us work in two dimensions and evaluate
Ie (x + y) ds,
where C is the straight line from the origin to the point whose coordinates are (I, I) (Figure 1112). If (x, y) are the coordinates
x
Figure 1112
of any point P on C and if s is the arc length measured from the origin, then x = slv2 and y = slv2. Hence, x + y = 2slv2 = v2s. Thus,
65
Ie (x + y) ds = v2 J.v1 s ds = v2.
.. • r _W
U. I) 11\
x
LAlle: .n"(;6(U~
and the Curl
Figure 1113
Let us integrate this same function (x + y) from (0, 0) to (l , 1) along another path as shown in Figure 1113. Here we break the
integration into two parts. one along eland the second along C 2' On C 1 we have x = sand y = O. Thus, on C I' X + Y = s, and so
Ier (x + y) ds = LI S ds = i.
Along C 2, x = 1 and y = s [note that the arc length on this segment of the path is measured from the point (1,0)]. It follows then that
f (x + y) ds = (I (l + s) ds = l
C2 Jo
Adding the results for the two segments, we find
Ic (x + y) ds
= f (x + y) ds + f (x + y) ds = ! + ! = 2.
Cr C"
The lesson to be learned is this: the value of a line integral can (indeed, usually does) depend on the path of integration.
Line Integrals Involving Vector Functions
66
Although the above discussion tells us what a line integral is, the kind of line integral we must deal with here has a feature not yet
Line Integrals mentioned. You williecall that we iJltloduced OUI discussioJl of
Involving Vector line integrals with the concept of work. Work, in the most ele
Functions mentary sense, is force times displacement. That this needs elaboration becomes clear when we recognize that both force and displacement are vectors.
Thus, consider some path C in three dimensions (Figure 111
4). Let us suppose that under the action of a force an object moves
Figure 1114
on this path from S I to s 2' At any point P on the curve let the force acting be designated f(x, y, z). The component of f which does work is, by definition, only that one ~hich acts along the curve, that is, the tangential component. Let t denote a unit vector which is tangent to the curve at P.I Then the work done by the
force m moving the object from S I to S 2 along the curve C IS
w = Ie f(x, y, z) • ids,
where it is understood, of course, that the integration begins at s = S I and ends at s = S2' The new feature of this integral is that the integrand is the dot product of two vector functions. To be able to handle such a line integral, we must know how to find i, and it is to this problem that we now turn.
Consider an arbitrary curve C (Figure 1115) parametrized by its arc length. At some point s on the curve we have x = xes), y = y(s), and z = z(s). At another point s + ~s we have x + ~x
67
~ ~
I t is a function of x. y. and z and should really be wntten t(x, y, z) We write
simply t to avoid complicating the notation.
y
and the Curl
Figure IlI5
= x(s + AS), y + Ay = y(s + AS), and z + Az = z(s + As).
Thus, the chord joining the two points on the curve directed from the first to the second is the vector Ar == iAx + jAy + kAz, where
Ax = x(s + As)  x(s), Ay yes + As) y(s),
AZ = z(s + As)  z(s).
If we now divide this vector by As, we get
Taking the limit of this as As approaches zero yields
dx . dy dz
i+J+k
ds ds ds '
A
alJd we assert that this is t. To begilJ with, it's clem that as ds
~ 0, the vector Ar becomes tangent to the curve at s. Further, in the limit As ~ 0, we see that IArl ~ As. Hence, in the limit the magnitude of this quantity is I. It follows then that we can make the identification
A • dx . dy k dz
t(s) = I  + J  +  .
ds ds ds
68
If we return now to the expression for work Wand use this
Line IntegralS Involving Vector Functions
formula for i. we find
f [. d.x • dy dZ]
W = f(x, y. z)· I  + J  + k  ds
e ds ds ds
r
= ]e (Ix dx + I, dy + f~ dz).
This is a formal expression, often, to carry out the integration, it
is useful to restore the ds as the following example illustrates. Consider
f(x, y, z) = iy  jx
and the path shown in Figure 1116(a). To calculate f e (f· t) ds
y
x
69
Figure 11I6(a)
in this case, we break the path C into three parts, Ct. C2, and C3 as shown. Sincef_ = O. we have
= Ie y d.x  x dy.
Now, on Ct. y = 0 and dy = 0, so there is no contribution to the
mtegral. SImIlarly, on C3 we have x  a and ax  0, and again the result is zero. Thus, the only contribution to the integral over
line Integrals and the Curl
C can come ftom C2• Restoriug the ds, we have
f (y dx  x dY) ds.
c ds ds
But (I x)/s cos 45°
1'''\ q and y's
r~ I
[Figure 1116(b)]. Thus,
s dx
x=l~=
v2 ds
I
_
v2
O<s<v2.
s dy 1
y=v2~ds=v2
Hence, the integral is
1 1../2
=  v2 0 ds =  1.
y
I Iy I
I
x
Ix
Figure III6(b)
Path Independence
III a liue iuteglai the path of integration is one of the ingredients
70
which determines the very function we integrate. It isn't remarkable, then, that the value of the integral can depend on the path
Path
of'trrtegration: What is .enunkable is that. nllde. some conditions;
Independence
the value of the integral does not depend on the path!
We show how this path independence comes about in the case of the Coulomb force. Let a charge qo be fixed at the origin and let another charge q be situated at (x, y, z) (Figure 1117). The
z u
(x,y,z)L
;'
;'
r ;'
;'
6 /
;'
I' Y
qo .I'
I
1
.....
.... I
.... I
....
.... .... 1 Figure 1I17
Coulomb force on q is
F = _1_ qqo Q
41TEo r2 '
(IIII)
where r = (X2 + y2 + Z2)112 is the distance between the two charges and Q is a unit vector pointing from qo to q. With this arrangement Q is clearly in the radial direction. Even more clearly, the radial vector r is in the radial direction.
Thus, we have Q = rlr = (ix + jy + kz)lr. and so
F = ss« _ix_+_.;:j~y_+_k_z
41TEo r3
Thus.
A
F . t ds = F; dx + F\ dy + F, dz
qqo x dx + y dy + z dz
41TEo r
The trick now is to use the relationship
71
line Integrals
Takillg diffelelltials ill this equatioll alld dividing by a factOi of
and the Curl
2 yields
x dx + y dy + z dz = r dr,
A qqo r dr qqo dr
F·tds
Suppose now that the charge q moves from a point P I at a distance r, from the origin to a point P 2 at a distance r2, over some path C connecting the two points (Figure 1118). Then
Figure 1118
Notice that to get this result, we haven't had to specify C in any way whatever; we'd get the same answer for any path connecting PI and P2• This, of course, proves that the line integral
Ie F· i ds
with F given by Equation (IIII) is pathindependent, but the result, so far, has been established only for the Coulomb force on q due to a single charge qo [Equation (1111)]. If there are
72
many charges ql' q2' ...• qN' then the total force on q IS F, + F 2 + ... + F N where F, is the Coulomb force on q due to the lth
Path
charge q; Helice,
Independence
( F· t ds = ( F • i ds + . . . + ( F • t ds.
Je Je I Je N
Now the discussion given above shows that each term of this sum
is pathindependent: hence, so is the sum itself. (All this, of course, is merely an application of the superposition principle.)
To phrase thIs result 10 terms of the field requires one last tnvlal step: Since F = qE, it follows that q feE· i ds is pathindepen
...
dent, whence feE· t ds is also. Strange to say, it is this fact that
will enable us eventually to convert V • E = p/Eo into a more useful equation.
If you examine the foregoing discussion carefully, you'll see
that the fact that the Coulomb force varies inversely as the square of r has nothing whatever to do with the path independence of the line integral. The path independence rests solely on two properties of the Coulomb force: (l) It depends only on the distance between the two particles. and (2) it acts along the line joining
them. Any force F with these two properties is called a central force, and f e F· ids is independent of path for any central force.' One further step pertaining to path independence can be taken here. If
Ie F· i ds
is independent of path, then
where, as indicated in Figure 1119. CI and C2 are two different arbitrary paths connecting the two points PI and P2 and directed as shown in the figure. Now if instead of integrating along CI from PI to P2, we go the other way, we simply change the sign
2 Our having illustrated path independence with a central force may give the erroneous impression that only central forces have pathindependent line integrals That is certainly not true, many functions which are not central forces have
73
pathindependent line integrals. Later we'll develop a simple cntenon for identifying such functions.
and the Curl
Figure 1119
of the line integral; that is,
f F . i ds =  ( F· t ds,
c, Jc,
where ••  c ," merely indicates that the integration is to be carried out along C, from P 2 to P r Thus
( F· i ds =  f F . i ds
Jcz c,
or
But"  C, + C 2" is just the closed loop from PI to P 2 and back, as shown in Figure 111] 0 Thus, if f F • i ds is independent of
Figure 11110
path, then
f F . i ds 0,
74
where tj is the standard notation for a line integral around a closed
The Curl
The Curl
path. It follows that if E is all elect.ostatic field, we call write
f E· i ds = o.
(1112)
The term "circulation" is often given to the path integral around
a closed curve of the tangential component of a vector function. Thus we have demonstrated that the circulation of the electro
static field is zero. In what follows we'll call this the circulation law.
If we are given some vector function F(x, y, z) and asked' 'Could this be an electrostatic field?" we can, in principle, provide an answer. If
t F . i ds "* 0
over even one path, then F cannot be an electrostatic field. If
f F . i ds = 0
75
over every closed path, then F can (but does not have to) be an electrostatic field.
Clearly this criterion is not easy to apply since we must be sure the circulation of F is zero over all possible paths. To develop a more useful criterion, we proceed much as we did in dealing with Gauss' law, Which, like the cit culation law, is an exp. ession
involving an integral over the electric field. Gauss' law is more useful in the differential form [Equations (1118) and (1123)] obtained by considering the ratio of flux to volume for ever decreasing surfaces. We now treat the circulation law in the same spirit and attempt to find the differential form of Equation (111 2). To stress the generality of our analysis and results, we deal with an arbitrary function F(x, y, z) and specialize to E(x, y, z) at a late. stage in the development:
Let us consider the circulation of F over a small rectangle parallel to the xyplane, with sides Ax and Ay and with the point
line Integrals
(x. y, z) at the celltel [FigUie III 11 (a)]. As shown ill FigUie
and the Curl
IIIII (b), we carry out the path integration in a counterclockwise direction looking down at the xyplane. The line integral is broken up into four parts: CB (bottom), CR (right), CT (top), and CL
z
(x,Y,:)
I 6y
I :
I I I Y
____J I __
I I I
I L..,
I / I I
I ~ I "
I I I I
~ I
Figure IIIII (a)
)'
T
./
......
cL ,II • 6YI j\ cR
Ax
....
,.
cB x Figure IIIII (b)
(left), Since the rectangle is small (eventually we shall take the limit as it shrinks down to zero), we'll approximate the integral over each segment by F • t evaluated at the center of the segment,
multiplied by the length of the segment.' Taking CB first, we have
= F, (x, y  11; , z) I1x. (III3a)
76
J Reread footnote 9 of Chapter II and then give an argument in support of this approxi mali on
The Curl
Dve. CT we fiud
r F. i ds = r r, dx
JCT JCT
=  F, (x. Y + ; , z) lu. (III 3b)
The negative sign here is required by the fact that
and dxlds = ( over C; Adding Equation (1I13a) and (1I13b), we find
F, (x. Y + ¥ ' Z)  F, (x, Y  ¥ ' Z)
=  Ay Ax Ay.
The factor Ax Ay is clearly the area AS of the rectangle. Thus,
 J (F, t) ds
AS Cr+CB
I r
(1114)
F, (x. Y + A; . Z)  F, (x. y  A; . z)
  
Exactly the same sort of analysis applied to the left and right sides of the rectangle (CL and CR) results in
77
F, (x + ¥ ' y, z)  F, (x  ¥ ' y, Z)
(1115)

Ax
Line Integrals
Adding Equations (1114) and (III 5) and taking the limit as 6.5
and the Curl
shrinks down about a point (x, y, z) (in which case ~ and Ay ~ 0 as well). we get
. I ~ " dF, aFx
hm  F . t ds = ~   •
As.u AS ~ a r ay
(1116)
about r r, v.z)
where !f1 is our semicomical notation meaning the circulation
around the little rectangle.
You may wonder about the generality and uniqueness of this result since it is obtained using a path of integration which is special in two ways: first, it is a rectangle, and second, it is parallel to the xyplane. If the path were not a rectangle, but a plane
curve of arbitrary shape, it would not affect our result (see Problems 1112 and III30). But our result definitely does depend on the special orientation of the path of integration. The choice of orientation made above clearly suggests two others, and they are shown in Figure III12(a) and (b) along with the result of
;~J: _
". I L
t'/:~ :    
: lJ~
".
r /
r .:
~
/
y
aFx az
aFt ax
Figure III 12( a)
y
aFz _ aFy
78
Figure III12(b)
The Curt
calculating
lim.'. }s'. J, F • i,th
.e.so ~ Rf
!!bout (:1:'),.1:)
for each.
Each of these three paths is named. in honor of the vector normal to Ihe enclosed area. The convention we use is this: Trace the curve C sothat the enclosed area is always to the left [Frgure IUI3{a)]. Then choose the normal so that, it, points "up" in the
Figure IIIB(a)
d· . t.. .,... Tl.: .'. .
IreCti.on snown m tile pteture, ",nilS coavennoa ,IS sometimes
called the nght~halld. rule, for ,if the nghthand is ori.ented so that thcfingers cud in the dirlocUo.n fn which the curve is traced" the thumb~ extended, points ,in the directio,fl of the normal [Figure IU13(b)]. Using the righthand rule, we have the foUowing:
79 figure IIIB(b)
Une Integrals
CaIculatmg hm~s+o ~ F • ids/aS
and the Curl
. . of oFy
for a path whose normal IS I, we get _z  ~ ,
oy oz
. oFx oFz
for a path whose normal is J, we get _  _ ,
B: ax
(l1I7a)
of er,
for a path whose normal is k, we get :..:..J:  _ •
ox oy
It turns out that these three quantities are the Cartesian compo
nents of a vector. To thIs vector we give the name "curl of F," which we write curl F. Thus, we have
I F  . (OFz OF,) . (OFx OFz)
cur  I   ____._ + J   _
oy oz oz ox
+ k (~  OFx). (l1I7b)
ox oy
This expression is often (indeed, usually) given as the definition of the curl, but we prefer to regard it as merely the form of the curl in Cartesian coordinates. We shall define the curl as the limit of circulation to area as the area tends to zero. To be precise, let
~c F· ids be the circulation of F about some path whose normal
n
is ft as shown in Figure 11114. Then by definition
R • curl F =
] i "
 F· t ds.
AS en
lim
~s+o
(1118)
By taking ft successively equal to i, j, and k, we get back the results given in Equation (11I7b). Since this limit Will, in general, have different values for different points (x, y, z), the curl of F is a vector Junction of position.' Note incidentally that although in our work we always assumed that the area enclosed
4 The word rotation (abbreviated "rot," amusingly enough) was once used for
80
what we now call the curl Though the term has long since dropped out of use, a related one survives' If curl F = 0, the function F is said to be irrotational
Figure 11114
by the path of integration was a plane, this need not be the case. Since the curl is defined in terms of a limit in which the enclosed surface shrinks to zero about some point, in the final stages of
this limiting process the enclosed surface is infinitesimally close to a plane, and all our considerations apply.
Since it is undoubtedly beyond the powers of a mere mortal to remember the expression given above for curl F in Cartesian coordinates [Equation (lII7b)], it is fortunate that there is a mne
monic device to fall back on. If the threebythree determinant
i
j
k
ajax aJay aJaz
F~
is expanded (most conveniently in minors of the first row) and if certain "products" are interpreted as partial derivatives [for example, (aJax)F1 = aFJax], the result will be identical with the one given in Equation (IllTb)." Thus, the anguish of remembering the form of curl F in Cartesian coordinates can be replaced
by the pain of remembering how to expand a threebythree determinant. Chacun a son gout.
As an example of calculating the curl, consider the vector function
F(x, y, z) = irz + jyz  ky2.
5 A mathematician would object to this since, strictly speaking, a determinant
81
cannot contain either vectors or operators. We aren't doing any senous damage, however, because our "determinant" is merely a memory aid.
Line Integrals and the Curl
We have
j k curl F = a/ax a/ay a/az
= i(2y  y) + j(x  0) + k(O  0)
= 3iy + jx.
You may have noticed that the curl operator can be written in terms of the del notation we introduced earlier. You can verify for yourself that
curl F = V X F,
which is read "del cross F." Henceforth, we shall always use V X F to indicate the curl.
The Curl in Cylindrical and Spherical Coordinates
To obtain the form of V X F in other coordinate systems. we proceed as we did above in finding the Cartesian form, merely modifying the paths of integration appropriately. As an example, using the path shown in Figure 1I[15(a) will yield the zcornponent of V X F in cylindncal coordinates." Note that we trace
J;.>.
6fJ "
y
Figure III15(a)
6 In deriving the Cartesian form of yo X F, each segment of each path of inte
82
gration (see Figures IIIII and 11112) was of the form x = constant, y = constant, or z = constant Similarly, in deriving the cylindrical form, each segment of each path is of the form r = constant, e = constant, or z = constant
The Curl in
Cylindrical and Spherical Coordinates
the CUI ve in accOidance with the righthand lule given in the
previous section. Viewing the path from above [as we do in
y
x
Figure lIlI5(b)
...
Figure III15(b)], the line integral of F(r, e, z) • t along the
segment of path marked "1" is
fe, F • i ds = F, (r. 6  ~6 , z) t:..r,
while along segment "3" it is
fe, F • i ds =  F, (r. 6 + ~6 . z) t:..r,
The area enclosed by the path is r Ar Ae. so
In the limit as Ar and Ae tend to zero, this becomes
~ ! aF, r ae
evaluated at the point (r, e, z).
Along segment "2" we find
83
L F' i ds = F. (r + t:..; ,6, z) (r + ~) M.
line Integrals
aJld aloJlg segmeJlt "4"
and the Curl
L F· i ds = F. (r  'Y ,0, z) (r  'Y) M.
1 r "
 F· t ds
Aa [ ( Ar) ( Ar )
= r Ar Aa r + 2 Fa r + 2 ' a, z
In the limit this becomes (llr)(a/ar)(rFa) evaluated at (r, a, z), Hence,
. 1 f" 1 a I er,
(V x F)z = hm  F . t ds =   (rFa)    .
.1s __ o AS r ar r aa
Paths for finding the r and acomponents of V x F are shown in Figures III15(c) and (d), respectively. You are asked to obtain
z
AD
x
Figure III15(c)
these two components yourself in Problem III 8. For complete
84
ness we give all three components of V x F in cylindrical
....
..........
y
the Curl
Figure III15(d)
coordinates:
I et: er, (V x F), = ; as  az '
I a I er,
(V x F). =   (rFe)   _ .
 r ar r as
The three components of curl F in spherical coordinates (see Problem 1119) are as follows:
r sm S as
I a
. (sin e F4»
r sin S a<1> '
(V x F),
I aF, I a
(V x F)e = .     (rF<l»'
r srn S a<1> r a r
(V X F)<l> =   (rFe)   _' .
r ar r as
The Meaning of the Curl
The preceding discussion may leave you with the feeling that knowing how to define and calculate the curl of some vector function is a far cry from knowing what it is. The fact that the
85 curl has something to do with a line integral around a closed path
and the Curl
_
/'
I f ,
x
I
\ I
\ I
\ I
....... ....
..... ..
Figure 11116
(indeed, the word "curl" itself) may suggest to you that it somehow has to do with things rotating, swirling, or curling around. By means of a few examples taken from fluid motion, we'll try to make these vague impressions a little clearer.
Suppose water IS HowlOg 10 Circular paths, somethmg lIke the water draining from a bathtub. A small volume of the water at a point (x, y) at time I has coordinates x = r cos wt, y = r sin wI, where 00 is the constant angular velocity of the water (Figure 11116).7 Thus, its velocity at (x, y) is
v = i(dxldl) + j(dyldl) = rw[  i sin sot + j cos WI] = w( iy + jx).
This expression gives what is called the velocity field of the water; it tells us the velocity of the water at any point (x, y). Your intuition probably tells you that, because the motion is circular, this velocity must have a nonzero curl. In fact, as you can show very easily,
v x v = 2kw.
This result should seem quite reasonable because it says that curl of the velocity is proportional to the angular velocity of the swirling water. We see that V x v is a vector perpendicular to the
7 This is not a realistic description of water draining from a tub since rotating
86
water shears tangentially and its angular velocity will therefore vary with r The crude description we use here is adequate for our purposes and has the virtue of being simple
The Meaning of pialle of motioll alld ill the positive z=dn ectioll [FigUi e III 17(a)].
the Curl
z
curl II
Figure III17(a)
If the water were rotating around in the other direction, the curl
ofv would then be in the negative zdirection [Figure III17(b)].
z
y
Figure IIl17(b)
Note that this is consistent with the righthand rule (see page 79),
If we were to put a small paddle wheel in the water, it would commence spinning because the impinging water would exert a net torque on the paddles (Figure 11118). Furthermore, the pad
87
Figure 11118
line Integrals
die wheel would lotate with its axis poillting ill the diIectioIl of
and the Curl
the curl.
Now consider a different velocity field, namely,
where Vo and h. are constants. Water with such a velocity field would have a flow pattern as indicated in Figure III 19. The
velocity at all points is in the positive ydirection, and its magnitude (indicated by the length of the arrows) varies with y. Since you see only straight line flow here without any rotational motion, you would probably guess that V x v = 0 in this case, and you would be right, as a simple calculation shows. There
would be no net torque on a paddle wheel placed anywhere in this flow pattern, and as a consequence, it would not spin.!
t t t t f
I~ ~ J\ I~ 1\
x
t
t
t
Figure 11119
Our last example is trickier than the two given above and shows that intuition can lead you astray if you're not careful. Let
a velocity field be given by
As in the previous example, the velocity in this case is everywhere in the ydirection, but now it varies with x, not y (Figure 11120). Here as in the above example you see no evidence of
rotational motion and you might guess that V x v = 0 once again.
88
8 If V x v = 0, the flow is said to be irrotationaJ. Compare with footnote 4.
t t
the Curl
x
Figure 11120
But as you should show for yourself,
2x "2
V X V =  k v  e  rll\
o ~2 •
A small paddle wheel placed in this flow pattern would spin, even
though the water is everywhere moving in the same direction. The reason this happens is that the velocity of the water varies with x, so that it strikes one of the paddles (P in Figure 11121)
p'
1
Figure 11121
with greater velocity than the other (P '), Thus, there will be a net torque. In more mathematical terms, the line integral of v • i around a small rectangle (Figure 11122) will be different from zero, for while
89
r v • i ds = r v· i ds = 0,
Jbonom JIOP
r • r • \I
Llne • nlel$' I,U"
and the Curl ~
,
,~ 1\ fly
....
(x, y) ,.
llx
x Figure 11122
the contributions from the other two sides are
f v· i ds = v\(x + Ax) dy
nght
and
r v > i ds = v\(x) ~y.
Jlef!
These do not cancel because v\ (x) "* v \ (x + dx). Incidentally, you should try to explain to your own satisfaction why in this example V x v is in the negative (positive) zdirection when x is positive (negative) and why V x v = 0 at x = o.
Differential Form of the Circulation Law
The curl is defined to be the limit of circulation to area, Thus,
1 f "
fi ' V x E = lim A E • t ds,
.:ls ..... O ~S (
where fi is a unit vector normal to the surface enclosed by C at the point about which the curve shrinks to zero. But if E is an electrostatic field, then
90
fc E· t ds = 0
Differential Form fOi allY path C. It follows that
of the Circulation
Law fi . V X E = O.
Since the curve C is arbitrary. we can arrange matters so that fi
is a unit vector pointing in any direction we choose. Thus,
takmg Ii  i, we have (V x E),  O~ taking fi = J. we have (V x E)\ = O~ taking fi = k, we have (V x E)~ = O.
Thus. all three Cartesian components of V x E vanish, and we
can conclude that for an electrostatic field,
V x E = O.
This is the longsoughtafter differential form of the circulation
law. We are now in a position to give an alternative and much more tractable answer to the question "Can a given vector function Ft.r, y, z) be an electrostatic field?" The answer is:
If V x F = 0, then F can be an electrostatic field, and if V x F =1= 0, then F cannot be an electrostatic field.
This is clearly a much more convenient criterion to apply than our earlier one (page 76), which required us to determine the line integral of F over all closed paths! To see how it works, let us do several examples.
Example I. Could F = K(iy + jx) be an electrostatic field? (K
is a constant.) Here we have
I VXF K
91
= 0 ~ V x F = O.
Line Integrals lind the Curl
Answer: Yes.
Example 2. Could F = K(iy ~ jx) be an electrostatic field? In this case
I _ . (... ax .ay) . .. '
 V x F """ k.·' '_. __ .... = 2k:) V x ,F =2kK.
K • ax oy'.
Answer.~No
From these examples we can see how easy this criterion is to apply.
Stokesl Theorem
[for the f\emaindelr of this ebapserwe digress from our presentanon to discuss another famous theorem, one strongly eeminiseera of the divergence theorem. and yet, as we'U see, quite differeet from it. This theorem, named fOlr the mathematician Stokes, relates a line integral around a dosed path to a surface integra.l over what is called a oopping surface of the path, so the first item on Ollr agenda is to define this term. Suppose we have a closed curve C. as shown in Figure 11I23(a)I, and imagine that it is made
of wire. Now let us suppose we attach 310 elastic: membrane to the wire as indicated inPigure UI23(b). This membrane is a
92
capping surface of the curve C .. ,Any oCher surface which C31ii be formed by stretching the membrane is a~so a capping surface; all examplels shown in Fi.gllre lU23(c).. Fi.gure IU24 shows four
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.