# -10

-5
0
5
10
-10
-5
0
5
10
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
z
x
y
z
Vector
Calculus
Michael Corral
Vector Calculus
Michael Corral
Schoolcraft College
Michael Corral is an Adjunct Faculty member of the Department of Mathematics
at Schoolcraft College. He received a B.A. in Mathematics from the University
of California at Berkeley, and received an M.A. in Mathematics and an M.S. in
Industrial & Operations Engineering from the University of Michigan.
This text was typeset in L
A
T
E
X2
ε
with the KOMA-Script bundle, using the GNU
Emacs text editor on a Fedora Linux system. The graphics were created using
MetaPost, PGF, and Gnuplot.
Permission is granted to copy, distribute and/or modify this document under the terms
of the GNU Free Documentation License, Version 1.2 or any later version published
by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts,
and no Back-Cover Texts. A copy of the license is included in the section entitled
Preface
This book covers calculus in two and three variables. It is suitable for a one-semester
course, normally known as “Vector Calculus”, “Multivariable Calculus”, or simply
“Calculus III”. The prerequisites are the standard courses in single-variable calculus
(a.k.a. Calculus I and II).
I have tried to be somewhat rigorous about proving results. But while it is impor-
tant for students to see full-blown proofs - since that is how mathematics works - too
much rigor and emphasis on proofs can impede the flow of learning for the vast ma-
jority of the audience at this level. If I were to rate the level of rigor in the book on a
scale of 1 to 10, with 1 being completely informal and 10 being completely rigorous, I
would rate it as a 5.
There are 420 exercises throughout the text, which in my experience are more than
enough for a semester course in this subject. There are exercises at the end of each
section, divided into three categories: A, B and C. The A exercises are mostly of a
routine computational nature, the B exercises are slightly more involved, and the C
exercises usually require some effort or insight to solve. A crude way of describing A,
B and C would be “Easy”, “Moderate” and “Challenging”, respectively. However, many
of the B exercises are easy and not all the C exercises are difficult.
There are a few exercises that require the student to write his or her own com-
puter program to solve some numerical approximation problems (e.g. the Monte Carlo
method for approximating multiple integrals, in Section 3.4). The code samples in the
text are in the Java programming language, hopefully with enough comments so that
the reader can figure out what is being done even without knowing Java. Those exer-
cises do not mandate the use of Java, so students are free to implement the solutions
using the language of their choice. While it would have been simple to use a script-
ing language like Python, and perhaps even easier with a functional programming
language (such as Haskell or Scheme), Java was chosen due to its ubiquity, relatively
clear syntax, and easy availability for multiple platforms.
Answers and hints to most odd-numbered and some even-numbered exercises are
provided in Appendix A. Appendix B contains a proof of the right-hand rule for the
cross product, which seems to have virtually disappeared from calculus texts over
the last few decades. Appendix C contains a brief tutorial on Gnuplot for graphing
functions of two variables.
This book is released under the GNU Free Documentation License (GFDL), which
allows others to not only copy and distribute the book but also to modify it. For more
details, see the included copy of the GFDL. So that there is no ambiguity on this
iii
iv Preface
matter, anyone can make as many copies of this book as desired and distribute it
as desired, without needing my permission. The PDF version will always be freely
available to the public at no cost (go to http://www.mecmath.net). Feel free to
contact me at mcorral@schoolcraft.edu for any questions on this or any other
matter involving the book (e.g. comments, suggestions, corrections, etc). I welcome
Finally, I would like to thank my students in Math 240 for being the guinea pigs
for the initial draft of this book, and for finding the numerous errors and typos it
contained.
January 2008 MICHAEL CORRAL
Contents
Preface iii
1 Vectors in Euclidean Space 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Vector Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Dot Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4 Cross Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5 Lines and Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.6 Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.7 Curvilinear Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.8 Vector-Valued Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
1.9 Arc Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2 Functions of Several Variables 65
2.1 Functions of Two or Three Variables . . . . . . . . . . . . . . . . . . . . . 65
2.2 Partial Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.3 Tangent Plane to a Surface . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.4 Directional Derivatives and the Gradient . . . . . . . . . . . . . . . . . . 78
2.5 Maxima and Minima . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
2.6 Unconstrained Optimization: Numerical Methods . . . . . . . . . . . . . 89
2.7 Constrained Optimization: Lagrange Multipliers . . . . . . . . . . . . . . 96
3 Multiple Integrals 101
3.1 Double Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.2 Double Integrals Over a General Region . . . . . . . . . . . . . . . . . . . 105
3.3 Triple Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.4 Numerical Approximation of Multiple Integrals . . . . . . . . . . . . . . 113
3.5 Change of Variables in Multiple Integrals . . . . . . . . . . . . . . . . . . 117
3.6 Application: Center of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.7 Application: Probability and Expected Value . . . . . . . . . . . . . . . . 128
4 Line and Surface Integrals 135
4.1 Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.2 Properties of Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.3 Green’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
v
vi Contents
4.4 Surface Integrals and the Divergence Theorem . . . . . . . . . . . . . . . 156
4.5 Stokes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
4.6 Gradient, Divergence, Curl and Laplacian . . . . . . . . . . . . . . . . . . 177
Bibliography 187
Appendix A: Answers and Hints to Selected Exercises 189
Appendix B: Proof of the Right-Hand Rule for the Cross Product 192
Appendix C: 3D Graphing with Gnuplot 196
History 209
Index 210
1 Vectors in Euclidean Space
1.1 Introduction
In single-variable calculus, the functions that one encounters are functions of a vari-
able (usually x or t) that varies over some subset of the real number line (which we
denote by ). For such a function, say, y = f (x), the graph of the function f con-
sists of the points (x, y) = (x, f (x)). These points lie in the Euclidean plane, which,
in the Cartesian or rectangular coordinate system, consists of all ordered pairs of
real numbers (a, b). We use the word “Euclidean” to denote a system in which all the
usual rules of Euclidean geometry hold. We denote the Euclidean plane by
2
; the
“2” represents the number of dimensions of the plane. The Euclidean plane has two
perpendicular coordinate axes: the x-axis and the y-axis.
In vector (or multivariable) calculus, we will deal with functions of two or three vari-
ables (usually x, y or x, y, z, respectively). The graph of a function of two variables, say,
z = f (x, y), lies in Euclidean space, which in the Cartesian coordinate systemconsists
of all ordered triples of real numbers (a, b, c). Since Euclidean space is 3-dimensional,
we denote it by
3
. The graph of f consists of the points (x, y, z) = (x, y, f (x, y)). The
3-dimensional coordinate system of Euclidean space can be represented on a flat sur-
face, such as this page or a blackboard, only by giving the illusion of three dimensions,
in the manner shown in Figure 1.1.1. Euclidean space has three mutually perpendic-
ular coordinate axes (x, y and z), and three mutually perpendicular coordinate planes:
the xy-plane, yz-plane and xz-plane (see Figure 1.1.2).
x
y
z
0
P(a, b, c)
a
b
c
Figure 1.1.1
x
y
z
0
yz-plane
xy-plane
xz-plane
Figure 1.1.2
1
2 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
The coordinate system shown in Figure 1.1.1 is known as a right-handed coordi-
nate system, because it is possible, using the right hand, to point the index finger in
the positive direction of the x-axis, the middle finger in the positive direction of the
y-axis, and the thumb in the positive direction of the z-axis, as in Figure 1.1.3.
Figure 1.1.3 Right-handed coordinate system
An equivalent way of defining a right-handed system is if you can point your thumb
upwards in the positive z-axis direction while using the remaining four fingers to
rotate the x-axis towards the y-axis. Doing the same thing with the left hand is what
defines a left-handed coordinate system. Notice that switching the x- and y-axes
in a right-handed system results in a left-handed system, and that rotating either
type of system does not change its “handedness”. Throughout the book we will use a
right-handed system.
For functions of three variables, the graphs exist in 4-dimensional space (i.e.
4
),
which we can not see in our 3-dimensional space, let alone simulate in 2-dimensional
space. So we can only think of 4-dimensional space abstractly. For an entertaining
discussion of this subject, see the book by ABBOTT.
1
So far, we have discussed the position of an object in 2-dimensional or 3-dimensional
space. But what about something such as the velocity of the object, or its acceleration?
Or the gravitational force acting on the object? These phenomena all seem to involve
motion and direction in some way. This is where the idea of a vector comes in.
1
One thing you will learn is why a 4-dimensional creature would be able to reach inside an egg and
remove the yolk without cracking the shell!
1.1 Introduction 3
You have already dealt with velocity and acceleration in single-variable calculus.
For example, for motion along a straight line, if y = f (t) gives the displacement of
an object after time t, then dy/dt = f

(t) is the velocity of the object at time t. The
derivative f

(t) is just a number, which is positive if the object is moving in an agreed-
upon “positive” direction, and negative if it moves in the opposite of that direction. So
you can think of that number, which was called the velocity of the object, as having
two components: a magnitude, indicated by a nonnegative number, preceded by a
direction, indicated by a plus or minus symbol (representing motion in the positive
direction or the negative direction, respectively), i.e. f

(t) = ±a for some number a ≥ 0.
Then a is the magnitude of the velocity (normally called the speed of the object), and
the ± represents the direction of the velocity (though the + is usually omitted for the
positive direction).
For motion along a straight line, i.e. in a 1-dimensional space, the velocities are
also contained in that 1-dimensional space, since they are just numbers. For general
motion along a curve in 2- or 3-dimensional space, however, velocity will need to be
represented by a multidimensional object which should have both a magnitude and a
direction. A geometric object which has those features is an arrow, which in elemen-
tary geometry is called a “directed line segment”. This is the motivation for how we
will define a vector.
Definition 1.1. A (nonzero) vector is a directed line segment drawn from a point P
(called its initial point) to a point Q (called its terminal point), with P and Q being
distinct points. The vector is denoted by
−−→
PQ. Its magnitude is the length of the line
segment, denoted by
¸
¸
¸
−−→
PQ
¸
¸
¸, and its direction is the same as that of the directed line
segment. The zero vector is just a point, and it is denoted by 0.
To indicate the direction of a vector, we draw an arrow from its initial point to its
terminal point. We will often denote a vector by a single bold-faced letter (e.g. v) and
use the terms “magnitude” and “length” interchangeably. Note that our definition
could apply to systems with any number of dimensions (see Figure 1.1.4 (a)-(c)).
0
x P Q R S
−−→
PQ
−−→
RS
(a) One dimension
x
y
0
P
Q
R
S

P
Q
−−→
RS
v
(b) Two dimensions
x
y
z
0
P
Q
R
S

P
Q
−−→
R
S
v
(c) Three dimensions
Figure 1.1.4 Vectors in different dimensions
4 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
A few things need to be noted about the zero vector. Our motivation for what a
vector is included the notions of magnitude and direction. What is the magnitude of
the zero vector? We define it to be zero, i.e. 0 = 0. This agrees with the definition of
the zero vector as just a point, which has zero length. What about the direction of the
zero vector? A single point really has no well-defined direction. Notice that we were
careful to only define the direction of a nonzero vector, which is well-defined since the
initial and terminal points are distinct. Not everyone agrees on the direction of the
zero vector. Some contend that the zero vector has arbitrary direction (i.e. can take
any direction), some say that it has indeterminate direction (i.e. the direction can not
be determined), while others say that it has no direction. Our definition of the zero
vector, however, does not require it to have a direction, and we will leave it at that.
2
Now that we know what a vector is, we need a way of determining when two vectors
are equal. This leads us to the following definition.
Definition 1.2. Two nonzero vectors are equal if they have the same magnitude and
the same direction. Any vector with zero magnitude is equal to the zero vector.
By this definition, vectors with the same magnitude and direction but with different
initial points would be equal. For example, in Figure 1.1.5 the vectors u, v and w all
have the same magnitude

5 (by the Pythagorean Theorem). And we see that u and
w are parallel, since they lie on lines having the same slope
1
2
, and they point in the
same direction. So u = w, even though they have different initial points. We also see
that v is parallel to u but points in the opposite direction. So u v.
1
2
3
4
1 2 3 4
x
y
0
u
v
w
Figure 1.1.5
So we can see that there are an infinite number of vectors for a given magnitude
and direction, those vectors all being equal and differing only by their initial and
terminal points. Is there a single vector which we can choose to represent all those
equal vectors? The answer is yes, and is suggested by the vector w in Figure 1.1.5.
2
In the subject of linear algebra there is a more abstract way of defining a vector where the concept of
“direction” is not really used. See ANTON and RORRES.
1.1 Introduction 5
Unless otherwise indicated, when speaking of “the vector” with a given magnitude
and direction, we will mean the one whose initial point is at the origin of the
coordinate system.
Thinking of vectors as starting from the origin provides a way of dealing with vec-
tors in a standard way, since every coordinate system has an origin. But there will
be times when it is convenient to consider a different initial point for a vector (for
example, when adding vectors, which we will do in the next section).
Another advantage of using the origin as the initial point is that it provides an easy
correspondence between a vector and its terminal point.
Example 1.1. Let v be the vector in
3
whose initial point is at the origin and whose
terminal point is (3, 4, 5). Though the point (3, 4, 5) and the vector v are different ob-
jects, it is convenient to write v = (3, 4, 5). When doing this, it is understood that the
initial point of v is at the origin (0, 0, 0) and the terminal point is (3, 4, 5).
x
y
z
0
P(3, 4, 5)
(a) The point (3,4,5)
x
y
z
0
v = (3, 4, 5)
(b) The vector (3,4,5)
Figure 1.1.6 Correspondence between points and vectors
Unless otherwise stated, when we refer to vectors as v = (a, b) in
2
or v = (a, b, c)
in
3
, we mean vectors in Cartesian coordinates starting at the origin. Also, we will
write the zero vector 0 in
2
and
3
as (0, 0) and (0, 0, 0), respectively.
The point-vector correspondence provides an easy way to check if two vectors are
equal, without having to determine their magnitude and direction. Similar to seeing
if two points are the same, you are now seeing if the terminal points of vectors starting
at the origin are the same. For each vector, find the (unique!) vector it equals whose
initial point is the origin. Then compare the coordinates of the terminal points of
these “new” vectors: if those coordinates are the same, then the original vectors are
equal. To get the “new” vectors starting at the origin, you translate each vector to
start at the origin by subtracting the coordinates of the original initial point from the
original terminal point. The resulting point will be the terminal point of the “new”
vector whose initial point is the origin. Do this for each original vector then compare.
6 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Example 1.2. Consider the vectors
−−→
PQand
−−→
RS in
3
, where P = (2, 1, 5), Q = (3, 5, 7), R =
(1, −3, −2) and S = (2, 1, 0). Does
−−→
PQ =
−−→
RS ?
Solution: The vector
−−→
PQ is equal to the vector v with initial point (0, 0, 0) and terminal
point Q − P = (3, 5, 7) − (2, 1, 5) = (3 − 2, 5 − 1, 7 − 5) = (1, 4, 2).
Similarly,
−−→
RS is equal to the vector w with initial point (0, 0, 0) and terminal point
S − R = (2, 1, 0) − (1, −3, −2) = (2 − 1, 1 − (−3), 0 − (−2)) = (1, 4, 2).
So
−−→
PQ = v = (1, 4, 2) and
−−→
RS = w = (1, 4, 2).

−−→
PQ =
−−→
RS
y
z
x
0
−−→
P
Q
−−→
R
S
Translate
−−→
PQ to v
Translate
−−→
RS to w
P
(2, 1, 5)
Q
(3, 5, 7)
R
(1, −3, −2)
S
(2, 1, 0)
(1, 4, 2)
v = w
Figure 1.1.7
Recall the distance formula for points in the Euclidean plane:
For points P = (x
1
, y
1
), Q = (x
2
, y
2
) in
2
, the distance d between P and Q is:
d =

(x
2
− x
1
)
2
+ (y
2
− y
1
)
2
(1.1)
By this formula, we have the following result:
For a vector
−−→
PQ in
2
with initial point P = (x
1
, y
1
) and terminal point
Q = (x
2
, y
2
), the magnitude of
−−→
PQ is:
¸
¸
¸
−−→
PQ
¸
¸
¸ =

(x
2
− x
1
)
2
+ (y
2
− y
1
)
2
(1.2)
1.1 Introduction 7
Finding the magnitude of a vector v = (a, b) in
2
is a special case of formula (1.2)
with P = (0, 0) and Q = (a, b) :
For a vector v = (a, b) in
2
, the magnitude of v is:
v =

a
2
+ b
2
(1.3)
To calculate the magnitude of vectors in
3
, we need a distance formula for points
in Euclidean space (we will postpone the proof until the next section):
Theorem 1.1. The distance d between points P = (x
1
, y
1
, z
1
) and Q = (x
2
, y
2
, z
2
) in
3
is:
d =

(x
2
− x
1
)
2
+ (y
2
− y
1
)
2
+ (z
2
− z
1
)
2
(1.4)
The proof will use the following result:
Theorem 1.2. For a vector v = (a, b, c) in
3
, the magnitude of v is:
v =

a
2
+ b
2
+ c
2
(1.5)
Proof: There are four cases to consider:
Case 1: a = b = c = 0. Then v = 0, so v = 0 =

0
2
+ 0
2
+ 0
2
=

a
2
+ b
2
+ c
2
.
Case 2: exactly two of a, b, c are 0. Without loss of generality, we assume that a =
b = 0 and c 0 (the other two possibilities are handled in a similar manner). Then
v = (0, 0, c), which is a vector of length |c| along the z-axis. So v = |c| =

c
2
=

0
2
+ 0
2
+ c
2
=

a
2
+ b
2
+ c
2
.
Case 3: exactly one of a, b, c is 0. Without loss of generality, we assume that a = 0,
b 0 and c 0 (the other two possibilities are handled in a similar manner). Then
v = (0, b, c), which is a vector in the yz-plane, so by the Pythagorean Theorem we have
v =

b
2
+ c
2
=

0
2
+ b
2
+ c
2
=

a
2
+ b
2
+ c
2
.
x
y
z
0
a
Q(a, b, c)
S
P
R
b
c
v
Figure 1.1.8
Case 4: none of a, b, c are 0. Without loss of generality, we can
assume that a, b, c are all positive (the other seven possibil-
ities are handled in a similar manner). Consider the points
P = (0, 0, 0), Q = (a, b, c), R = (a, b, 0), and S = (a, 0, 0), as shown
in Figure 1.1.8. Applying the Pythagorean Theorem to the
right triangle △PS R gives |PR|
2
= a
2
+b
2
. A second application
of the Pythagorean Theorem, this time to the right triangle
△PQR, gives v = |PQ| =

|PR|
2
+ |QR|
2
=

a
2
+ b
2
+ c
2
.
This proves the theorem. QED
8 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Example 1.3. Calculate the following:
(a) The magnitude of the vector
−−→
PQ in
2
with P = (−1, 2) and Q = (5, 5).
Solution: By formula (1.2),
¸
¸
¸
−−→
PQ
¸
¸
¸ =

(5 − (−1))
2
+ (5 − 2)
2
=

36 + 9 =

45 = 3

5.
(b) The magnitude of the vector v = (8, 3) in
2
.
Solution: By formula (1.3), v =

8
2
+ 3
2
=

73.
(c) The distance between the points P = (2, −1, 4) and Q = (4, 2, −3) in
2
.
Solution: By formula (1.4), the distance d =

(4 − 2)
2
+ (2 − (−1))
2
+ (−3 − 4)
2
=

4 + 9 + 49 =

62.
(d) The magnitude of the vector v = (5, 8, −2) in
3
.
Solution: By formula (1.5), v =

5
2
+ 8
2
+ (−2)
2
=

25 + 64 + 4 =

93.

¨
Exercises
A
1. Calculate the magnitudes of the following vectors:
(a) v = (2, −1) (b) v = (2, −1, 0) (c) v = (3, 2, −2) (d) v = (0, 0, 1) (e) v = (6, 4, −4)
2. For the points P = (1, −1, 1), Q = (2, −2, 2), R = (2, 0, 1), S = (3, −1, 2), does
−−→
PQ =
−−→
RS ?
3. For the points P = (0, 0, 0), Q = (1, 3, 2), R = (1, 0, 1), S = (2, 3, 4), does
−−→
PQ =
−−→
RS ?
B
4. Let v = (1, 0, 0) and w = (a, 0, 0) be vectors in
3
. Show that w = |a| v.
5. Let v = (a, b, c) and w = (3a, 3b, 3c) be vectors in
3
. Show that w = 3 v.
C
x
y
z
0
P(x
1
, y
1
, z
1
)
Q(x
2
, y
2
, z
2
)
R(x
2
, y
2
, z
1
)
S (x
1
, y
1
, 0)
T(x
2
, y
2
, 0)
U(x
2
, y
1
, 0)
Figure 1.1.9
6. Though we will see a simple proof of Theorem 1.1
in the next section, it is possible to prove it using
methods similar to those in the proof of Theorem
1.2. Prove the special case of Theorem 1.1 where
the points P = (x
1
, y
1
, z
1
) and Q = (x
2
, y
2
, z
2
) satisfy
the following conditions:
x
2
> x
1
> 0, y
2
> y
1
> 0, and z
2
> z
1
> 0.
(Hint: Think of Case 4 in the proof of Theorem
1.2, and consider Figure 1.1.9.)
1.2 Vector Algebra 9
1.2 Vector Algebra
Now that we know what vectors are, we can start to perform some of the usual al-
gebraic operations on them (e.g. addition, subtraction). Before doing that, we will
introduce the notion of a scalar.
Definition 1.3. A scalar is a quantity that can be represented by a single number.
For our purposes, scalars will always be real numbers.
3
Examples of scalar quanti-
ties are mass, electric charge, and speed (not velocity).
4
We can now define scalar
multiplication of a vector.
Definition 1.4. For a scalar k and a nonzero vector v, the scalar multiple of v by k,
denoted by kv, is the vector whose magnitude is |k| v, points in the same direction
as v if k > 0, points in the opposite direction as v if k < 0, and is the zero vector 0 if
k = 0. For the zero vector 0, we define k0 = 0 for any scalar k.
Two vectors v and w are parallel (denoted by v w) if one is a scalar multiple of
the other. You can think of scalar multiplication of a vector as stretching or shrinking
the vector, and as flipping the vector in the opposite direction if the scalar is a negative
number (see Figure 1.2.1).
v
2v 3v 0.5v
−v
−2v
Figure 1.2.1
Recall that translating a nonzero vector means that the initial point of the vector
is changed but the magnitude and direction are preserved. We are now ready to define
the sum of two vectors.
Definition 1.5. The sum of vectors v and w, denoted by v + w, is obtained by trans-
lating w so that its initial point is at the terminal point of v; the initial point of v + w
is the initial point of v, and its terminal point is the new terminal point of w.
3
The term scalar was invented by 19
th
century Irish mathematician, physicist and astronomer William
Rowan Hamilton, to convey the sense of something that could be represented by a point on a scale or
graduated ruler. The word vector comes from Latin, where it means “carrier”.
4
An alternate definition of scalars and vectors, used in physics, is that under certain types of coordinate
transformations (e.g. rotations), a quantity that is not affected is a scalar, while a quantity that is
affected (in a certain way) is a vector. See MARION for details.
10 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Intuitively, adding w to v means tacking on w to the end of v (see Figure 1.2.2).
v
w
(a) Vectors v and w
v
w
(b) Translate w to the end of v
v
w
v + w
(c) The sum v + w
Figure 1.2.2 Adding vectors v and w
Notice that our definition is valid for the zero vector (which is just a point, and
hence can be translated), and so we see that v + 0 = v = 0 + v for any vector v. In
particular, 0 + 0 = 0. Also, it is easy to see that v + (−v) = 0, as we would expect. In
general, since the scalar multiple −v = −1 v is a well-defined vector, we can define
vector subtraction as follows: v − w = v + (−w). See Figure 1.2.3.
v
w
(a) Vectors v and w
v
−w
(b) Translate −w to the end of v
v
−w
v − w
(c) The difference v − w
Figure 1.2.3 Subtracting vectors v and w
Figure 1.2.4 shows the use of “geometric proofs” of various laws of vector algebra,
that is, it uses laws from elementary geometry to prove statements about vectors. For
example, (a) shows that v + w = w + v for any vectors v, w. And (c) shows how you
can think of v − w as the vector that is tacked on to the end of w to add up to v.
v
v
w w
w+ v
v + w
−w
w
v − w
v − w
v
(b) Subtract vectors
v
w
v + w
v − w
Figure 1.2.4 “Geometric” vector algebra
Notice that we have temporarily abandoned the practice of starting vectors at the
origin. In fact, we have not even mentioned coordinates in this section so far. Since we
will deal mostly with Cartesian coordinates in this book, the following two theorems
are useful for performing vector algebra on vectors in
2
and
3
starting at the origin.
1.2 Vector Algebra 11
Theorem 1.3. Let v = (v
1
, v
2
), w = (w
1
, w
2
) be vectors in
2
, and let k be a scalar. Then
(a) kv = (kv
1
, kv
2
)
(b) v + w = (v
1
+ w
1
, v
2
+ w
2
)
Proof: (a) Without loss of generality, we assume that v
1
, v
2
> 0 (the other possibilities
are handled in a similar manner). If k = 0 then kv = 0v = 0 = (0, 0) = (0v
1
, 0v
2
) =
(kv
1
, kv
2
), which is what we needed to show. If k 0, then (kv
1
, kv
2
) lies on a line with
slope
kv
2
kv
1
=
v
2
v
1
, which is the same as the slope of the line on which v (and hence kv) lies,
and (kv
1
, kv
2
) points in the same direction on that line as kv. Also, by formula (1.3) the
magnitude of (kv
1
, kv
2
) is

(kv
1
)
2
+ (kv
2
)
2
=

k
2
v
2
1
+ k
2
v
2
2
=

k
2
(v
2
1
+ v
2
2
) = |k|

v
2
1
+ v
2
2
=
|k| v. So kv and (kv
1
, kv
2
) have the same magnitude and direction. This proves (a).
x
y
0
w
2
v
2
w
1
v
1
v
1
+ w
1
v
2
+ w
2
w
2
w
1
v
v
w
w
v + w
Figure 1.2.5
(b) Without loss of generality, we assume that
v
1
, v
2
, w
1
, w
2
> 0 (the other possibilities are han-
dled in a similar manner). From Figure 1.2.5,
we see that when translating w to start at
the end of v, the new terminal point of w is
(v
1
+ w
1
, v
2
+ w
2
), so by the definition of v + w
this must be the terminal point of v+w. This
proves (b). QED
Theorem 1.4. Let v = (v
1
, v
2
, v
3
), w = (w
1
, w
2
, w
3
) be vectors in
3
, let k be a scalar. Then
(a) kv = (kv
1
, kv
2
, kv
3
)
(b) v + w = (v
1
+ w
1
, v
2
+ w
2
, v
3
+ w
3
)
The following theorem summarizes the basic laws of vector algebra.
Theorem 1.5. For any vectors u, v, w, and scalars k, l, we have
(a) v + w = w+ v Commutative Law
(b) u + (v + w) = (u + v) + w Associative Law
(c) v + 0 = v = 0 + v Additive Identity
(d) v + (−v) = 0 Additive Inverse
(e) k(lv) = (kl)v Associative Law
(f) k(v + w) = kv + kw Distributive Law
(g) (k + l)v = kv + lv Distributive Law
Proof: (a) We already presented a geometric proof of this in Figure 1.2.4(a).
(b) To illustrate the difference between analytic proofs and geometric proofs in vector
algebra, we will present both types here. For the analytic proof, we will use vectors
in
3
(the proof for
2
is similar).
12 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Let u = (u
1
, u
2
, u
3
), v = (v
1
, v
2
, v
3
), w = (w
1
, w
2
, w
3
) be vectors in
3
. Then
u + (v + w) = (u
1
, u
2
, u
3
) + ((v
1
, v
2
, v
3
) + (w
1
, w
2
, w
3
))
= (u
1
, u
2
, u
3
) + (v
1
+ w
1
, v
2
+ w
2
, v
3
+ w
3
) by Theorem 1.4(b)
= (u
1
+ (v
1
+ w
1
), u
2
+ (v
2
+ w
2
), u
3
+ (v
3
+ w
3
)) by Theorem 1.4(b)
= ((u
1
+ v
1
) + w
1
, (u
2
+ v
2
) + w
2
, (u
3
+ v
3
) + w
3
) by properties of real numbers
= (u
1
+ v
1
, u
2
+ v
2
, u
3
+ v
3
) + (w
1
, w
2
, w
3
) by Theorem 1.4(b)
= (u + v) + w
This completes the analytic proof of (b). Figure 1.2.6 provides the geometric proof.
u
v
w
u + v
v + w
u + (v + w) = (u + v) + w
Figure 1.2.6 Associative Law for vector addition
(c) We already discussed this on p.10.
(d) We already discussed this on p.10.
(e) We will prove this for a vector v = (v
1
, v
2
, v
3
) in
3
(the proof for
2
is similar):
k(lv) = k(lv
1
, lv
2
, lv
3
) by Theorem 1.4(a)
= (klv
1
, klv
2
, klv
3
) by Theorem 1.4(a)
= (kl)(v
1
, v
2
, v
3
) by Theorem 1.4(a)
= (kl)v
(f) and (g): Left as exercises for the reader. QED
A unit vector is a vector with magnitude 1. Notice that for any nonzero vector v,
the vector
v
v
is a unit vector which points in the same direction as v, since
1
v
> 0
and
¸
¸
¸
v
v
¸
¸
¸ =
v
v
= 1. Dividing a nonzero vector v by v is often called normalizing v.
There are specific unit vectors which we will often use, called the basis vectors:
i = (1, 0, 0), j = (0, 1, 0), and k = (0, 0, 1) in
3
; i = (1, 0) and j = (0, 1) in
2
.
These are useful for several reasons: they are mutually perpendicular, since they lie
on distinct coordinate axes; they are all unit vectors: i = j = k = 1; every vector
can be written as a unique scalar combination of the basis vectors: v = (a, b) = a i + b j
in
2
, v = (a, b, c) = a i + b j + c k in
3
. See Figure 1.2.7.
1.2 Vector Algebra 13
1
2
1 2
x
y
0
i
j
(a)
2
x
y
0
ai
bj
v = (a, b)
(b) v = a i + b j
1
2
1 2
1
2
x
y
z
0
i
j
k
(c)
3
x
y
z
0
ai
bj
ck
v = (a, b, c)
(d) v = a i + b j + c k
Figure 1.2.7 Basis vectors in different dimensions
When a vector v = (a, b, c) is written as v = a i+b j+c k, we say that v is in component
form, and that a, b, and c are the i, j, and k components, respectively, of v. We have:
v = v
1
i + v
2
j + v
3
k, k a scalar =⇒ kv = kv
1
i + kv
2
j + kv
3
k
v = v
1
i + v
2
j + v
3
k, w = w
1
i + w
2
j + w
3
k =⇒ v + w = (v
1
+ w
1
)i + (v
2
+ w
2
)j + (v
3
+ w
3
)k
v = v
1
i + v
2
j + v
3
k =⇒ v =

v
2
1
+ v
2
2
+ v
2
3
Example 1.4. Let v = (2, 1, −1) and w = (3, −4, 2) in
3
.
(a) Find v − w.
Solution: v − w = (2 − 3, 1 − (−4) − 1 − 2) = (−1, 5, −3)
(b) Find 3v + 2w.
Solution: 3v + 2w = (6, 3, −3) + (6, −8, 4) = (12, −5, 1)
(c) Write v and w in component form.
Solution: v = 2 i + j − k, w = 3 i − 4 j + 2 k
(d) Find the vector u such that u + v = w.
Solution: By Theorem 1.5, u = w−v = −(v−w) = −(−1, 5, −3) = (1, −5, 3), by part(a).
(e) Find the vector u such that u + v + w = 0.
Solution: By Theorem 1.5, u = −w− v = −(3, −4, 2) − (2, 1, −1) = (−5, 3, −1).
(f) Find the vector u such that 2u + i − 2 j = k.
Solution: 2u = −i + 2 j + k =⇒ u = −
1
2
i + j +
1
2
k
(g) Find the unit vector
v
v
.
Solution:
v
v
=
1

2
2
+1
2
+(−1)
2
(2, 1, −1) =

2

6
,
1

6
,
−1

6

14 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
We can now easily prove Theorem 1.1 from the previous section. The distance d
between two points P = (x
1
, y
1
, z
1
) and Q = (x
2
, y
2
, z
2
) in
3
is the same as the length
of the vector w − v, where the vectors v and w are defined as v = (x
1
, y
1
, z
1
) and w =
(x
2
, y
2
, z
2
) (see Figure 1.2.8). So since w− v = (x
2
− x
1
, y
2
− y
1
, z
2
− z
1
), then d = w− v =

(x
2
− x
1
)
2
+ (y
2
− y
1
)
2
+ (z
2
− z
1
)
2
by Theorem 1.2.
x
y
z
0
P(x
1
, y
1
, z
1
)
Q(x
2
, y
2
, z
2
)
v
w
w− v
Figure 1.2.8 Proof of Theorem 1.2: d = w− v

¨
Exercises
A
1. Let v = (−1, 5, −2) and w = (3, 1, 1).
(a) Find v − w. (b) Find v + w. (c) Find
v
v
. (d) Find
¸
¸
¸
1
2
(v − w)
¸
¸
¸.
(e) Find
¸
¸
¸
1
2
(v + w)
¸
¸
¸. (f) Find −2 v + 4 w. (g) Find v − 2 w.
(h) Find the vector u such that u + v + w = i.
(i) Find the vector u such that u + v + w = 2 j + k.
(j) Is there a scalar m such that m(v + 2 w) = k? If so, find it.
2. For the vectors v and w from Exercise 1, is v − w = v − w? If not, which
quantity is larger?
3. For the vectors v and w from Exercise 1, is v + w = v + w? If not, which
quantity is larger?
B
4. Prove Theorem 1.5(f) for
3
. 5. Prove Theorem 1.5(g) for
3
.
C
6. We know that every vector in
3
can be written as a scalar combination of the
vectors i, j, and k. Can every vector in
3
be written as a scalar combination of
just i and j, i.e. for any vector v in
3
, are there scalars m, n such that v = mi + n j?
1.3 Dot Product 15
1.3 Dot Product
You may have noticed that while we did define multiplication of a vector by a scalar in
the previous section on vector algebra, we did not define multiplication of a vector by
a vector. We will now see one type of multiplication of vectors, called the dot product.
Definition 1.6. Let v = (v
1
, v
2
, v
3
) and w = (w
1
, w
2
, w
3
) be vectors in
3
.
The dot product of v and w, denoted by v··· w, is given by:
v··· w = v
1
w
1
+ v
2
w
2
+ v
3
w
3
(1.6)
Similarly, for vectors v = (v
1
, v
2
) and w = (w
1
, w
2
) in
2
, the dot product is:
v··· w = v
1
w
1
+ v
2
w
2
(1.7)
Notice that the dot product of two vectors is a scalar, not a vector. So the associative
law that holds for multiplication of numbers and for addition of vectors (see Theorem
1.5(b),(e)), does not hold for the dot product of vectors. Why? Because for vectors u, v,
w, the dot product u··· v is a scalar, and so (u··· v) ··· w is not defined since the left side of
that dot product (the part in parentheses) is a scalar and not a vector.
For vectors v = v
1
i + v
2
j + v
3
k and w = w
1
i + w
2
j + w
3
k in component form, the dot
product is still v··· w = v
1
w
1
+ v
2
w
2
+ v
3
w
3
.
Also notice that we defined the dot product in an analytic way, i.e. by referencing
vector coordinates. There is a geometric way of defining the dot product, which we
will now develop as a consequence of the analytic definition.
Definition 1.7. The angle between two nonzero vectors with the same initial point
is the smallest angle between them.
We do not define the angle between the zero vector and any other vector. Any two
nonzero vectors with the same initial point have two angles between them: θ and
360

− θ. We will always choose the smallest nonnegative angle θ between them, so
that 0

≤ θ ≤ 180

. See Figure 1.3.1.
θ
360

− θ
(a) 0

< θ < 180

θ
360

− θ
(b) θ = 180

θ
360

− θ
(c) θ = 0

Figure 1.3.1 Angle between vectors
We can now take a more geometric view of the dot product by establishing a rela-
tionship between the dot product of two vectors and the angle between them.
16 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Theorem1.6. Let v, wbe nonzero vectors, and let θ be the angle between them. Then
cos θ =
v··· w
v w
(1.8)
Proof: We will prove the theorem for vectors in
3
(the proof for
2
is similar). Let
v = (v
1
, v
2
, v
3
) and w = (w
1
, w
2
, w
3
). By the Law of Cosines (see Figure 1.3.2), we have
v − w
2
= v
2
+ w
2
− 2 v w cos θ (1.9)
(note that equation (1.9) holds even for the “degenerate” cases θ = 0

and 180

).
θ
x
y
z
0
v
w
v − w
Figure 1.3.2
Since v − w = (v
1
− w
1
, v
2
− w
2
, v
3
− w
3
), expanding v − w
2
in equation (1.9) gives
v
2
+ w
2
− 2 v w cos θ = (v
1
− w
1
)
2
+ (v
2
− w
2
)
2
+ (v
3
− w
3
)
2
= (v
2
1
− 2v
1
w
1
+ w
2
1
) + (v
2
2
− 2v
2
w
2
+ w
2
2
) + (v
2
3
− 2v
3
w
3
+ w
2
3
)
= (v
2
1
+ v
2
2
+ v
2
3
) + (w
2
1
+ w
2
2
+ w
2
3
) − 2(v
1
w
1
+ v
2
w
2
+ v
3
w
3
)
= v
2
+ w
2
− 2(v··· w) , so
−2 v w cos θ = −2(v··· w) , so since v 0 and w 0 then
cos θ =
v··· w
v w
, since v > 0 and w > 0. QED
Example 1.5. Find the angle θ between the vectors v = (2, 1, −1) and w = (3, −4, 1).
Solution: Since v··· w = (2)(3) + (1)(−4) + (−1)(1) = 1, v =

6, and w =

26, then
cos θ =
v··· w
v w
=
1

6

26
=
1
2

39
≈ 0.08 =⇒ θ = 85.41

Two nonzero vectors are perpendicular if the angle between them is 90

. Since
cos 90

= 0, we have the following important corollary to Theorem 1.6:
Corollary 1.7. Two nonzero vectors v and ware perpendicular if and only if v··· w = 0.
We will write v ⊥ w to indicate that v and w are perpendicular.
1.3 Dot Product 17
Since cos θ > 0 for 0

≤ θ < 90

and cos θ < 0 for 90

< θ ≤ 180

, we also have:
Corollary 1.8. If θ is the angle between nonzero vectors v and w, then
v··· w is

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
> 0 for 0

≤ θ < 90

0 for θ = 90

< 0 for 90

< θ ≤ 180

By Corollary 1.8, the dot product can be thought of as a way of telling if the angle
between two vectors is acute, obtuse, or a right angle, depending on whether the dot
product is positive, negative, or zero, respectively. See Figure 1.3.3.
0

≤ θ < 90

v
w
(a) v··· w > 0
90

< θ ≤ 180

v
w
(b) v··· w < 0
θ = 90

v
w
(c) v··· w = 0
Figure 1.3.3 Sign of the dot product & angle between vectors
Example 1.6. Are the vectors v = (−1, 5, −2) and w = (3, 1, 1) perpendicular?
Solution: Yes, v ⊥ w since v··· w = (−1)(3) + (5)(1) + (−2)(1) = 0.
The following theorem summarizes the basic properties of the dot product.
Theorem 1.9. For any vectors u, v, w, and scalar k, we have
(a) v··· w = w··· v Commutative Law
(b) (kv) ··· w = v··· (kw) = k(v··· w) Associative Law
(c) v··· 0 = 0 = 0 ··· v
(d) u··· (v + w) = u··· v + u··· w Distributive Law
(e) (u + v) ··· w = u··· w+ v··· w Distributive Law
(f) |v··· w| ≤ v w Cauchy-Schwarz Inequality
5
Proof: The proofs of parts (a)-(e) are straightforward applications of the definition of
the dot product, and are left to the reader as exercises. We will prove part (f).
(f) If either v = 0 or w = 0, then v ··· w = 0 by part (c), and so the inequality holds
trivially. So assume that v and w are nonzero vectors. Then by Theorem 1.6,
v··· w = cos θ v w , so
|v··· w| = |cos θ| v w , so
|v··· w| ≤ v w since |cos θ| ≤ 1. QED
5
Also known as the Cauchy-Schwarz-Buniakovski Inequality.
18 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Using Theorem 1.9, we see that if u ··· v = 0 and u ··· w = 0, then u ··· (kv + lw) =
k(u··· v) + l(u··· w) = k(0) + l(0) = 0 for all scalars k, l. Thus, we have the following fact:
If u ⊥ v and u ⊥ w, then u ⊥ (kv + lw) for all scalars k, l.
For vectors v and w, the collection of all scalar combinations kv + lw is called the
span of v and w. If nonzero vectors v and w are parallel, then their span is a line; if
they are not parallel, then their span is a plane. So what we showed above is that a
vector which is perpendicular to two other vectors is also perpendicular to their span.
The dot product can be used to derive properties of the magnitudes of vectors, the
most important of which is the Triangle Inequality, as given in the following theorem:
Theorem 1.10. For any vectors v, w, we have
(a) v
2
= v··· v
(b) v + w ≤ v + w Triangle Inequality
(c) v − w ≥ v − w
Proof: (a) Left as an exercise for the reader.
(b) By part (a) and Theorem 1.9, we have
v + w
2
= (v + w) ··· (v + w) = v··· v + v··· w+ w··· v + w··· w
= v
2
+ 2(v··· w) + w
2
, so since a ≤ |a| for any real number a, we have
≤ v
2
+ 2 |v··· w| + w
2
, so by Theorem 1.9(f) we have
≤ v
2
+ 2 v w + w
2
= (v + w)
2
and so
v + w ≤ v + w after taking square roots of both sides, which proves (b).
(c) Since v = w+ (v − w), then v = w+ (v − w) ≤ w + v − w by the Triangle
Inequality, so subtracting w from both sides gives v − w ≤ v − w. QED
v
w
v + w
Figure 1.3.4
The Triangle Inequality gets its name from the fact that in any tri-
angle, no one side is longer than the sum of the lengths of the other two
sides (see Figure 1.3.4). Another way of saying this is with the familiar
statement “the shortest distance between two points is a straight line.”

¨
Exercises
A
1. Let v = (5, 1, −2) and w = (4, −4, 3). Calculate v··· w.
2. Let v = −3 i − 2 j − k and w = 6 i + 4 j + 2 k. Calculate v··· w.
For Exercises 3-8, find the angle θ between the vectors v and w.
1.3 Dot Product 19
3. v = (5, 1, −2), w = (4, −4, 3) 4. v = (7, 2, −10), w = (2, 6, 4)
5. v = (2, 1, 4), w = (1, −2, 0) 6. v = (4, 2, −1), w = (8, 4, −2)
7. v = −i + 2 j + k, w = −3 i + 6 j + 3 k 8. v = i, w = 3 i + 2 j + 4k
9. Let v = (8, 4, 3) and w = (−2, 1, 4). Is v ⊥ w? Justify your answer.
10. Let v = (6, 0, 4) and w = (0, 2, −1). Is v ⊥ w? Justify your answer.
11. For v, w from Exercise 5, verify the Cauchy-Schwarz Inequality |v··· w| ≤ v w.
12. For v, w from Exercise 6, verify the Cauchy-Schwarz Inequality |v··· w| ≤ v w.
13. For v, w from Exercise 5, verify the Triangle Inequality v + w ≤ v + w.
14. For v, w from Exercise 6, verify the Triangle Inequality v + w ≤ v + w.
B
Note: Consider only vectors in
3
for Exercises 15-25.
15. Prove Theorem 1.9(a). 16. Prove Theorem 1.9(b).
17. Prove Theorem 1.9(c). 18. Prove Theorem 1.9(d).
19. Prove Theorem 1.9(e). 20. Prove Theorem 1.10(a).
21. Prove or give a counterexample: If u··· v = u··· w, then v = w.
C
22. Prove or give a counterexample: If v··· w = 0 for all v, then w = 0.
23. Prove or give a counterexample: If u··· v = u··· w for all u, then v = w.
24. Prove that

v − w

≤ v − w for all v, w.
L
w
v
u
Figure 1.3.5
25. For nonzero vectors v and w, the projection of v onto w
(sometimes written as pro j
w
v) is the vector u along the
same line L as w whose terminal point is obtained by drop-
ping a perpendicular line from the terminal point of v to L
(see Figure 1.3.5). Show that
u =
|v··· w|
w
.
(Hint: Consider the angle between v and w.)
26. Let α, β, and γ be the angles between a nonzero vector v in
3
and the vectors i, j,
and k, respectively. Show that cos
2
α + cos
2
β + cos
2
γ = 1.
(Note: α, β, γ are often called the direction angles of v, and cos α, cos β, cos γ are
called the direction cosines.)
20 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
1.4 Cross Product
In Section 1.3 we defined the dot product, which gave a way of multiplying two vectors.
The resulting product, however, was a scalar, not a vector. In this section we will
define a product of two vectors that does result in another vector. This product, called
the cross product, is only defined for vectors in
3
. The definition may appear strange
and lacking motivation, but we will see the geometric basis for it shortly.
Definition 1.8. Let v = (v
1
, v
2
, v
3
) and w = (w
1
, w
2
, w
3
) be vectors in
3
. The cross
product of v and w, denoted by v××× w, is the vector in
3
given by:
v××× w = (v
2
w
3
− v
3
w
2
, v
3
w
1
− v
1
w
3
, v
1
w
2
− v
2
w
1
) (1.10)
1
1
1
x
y
z
0
i
j
k = i ××× j
Figure 1.4.1
Example 1.7. Find i ××× j.
Solution: Since i = (1, 0, 0) and j = (0, 1, 0), then
i ××× j = ((0)(0) − (0)(1), (0)(0) − (1)(0), (1)(1) − (0)(0))
= (0, 0, 1)
= k
Similarly it can be shown that j ××× k = i and k××× i = j.
In the above example, the cross product of the given vectors was perpendicular to
both those vectors. It turns out that this will always be the case.
Theorem 1.11. If the cross product v ××× w of two nonzero vectors v and w is also a
nonzero vector, then it is perpendicular to both v and w.
Proof: We will show that (v××× w) ··· v = 0:
(v××× w) ··· v = (v
2
w
3
− v
3
w
2
, v
3
w
1
− v
1
w
3
, v
1
w
2
− v
2
w
1
) ··· (v
1
, v
2
, v
3
)
= v
2
w
3
v
1
− v
3
w
2
v
1
+ v
3
w
1
v
2
− v
1
w
3
v
2
+ v
1
w
2
v
3
− v
2
w
1
v
3
= v
1
v
2
w
3
− v
1
v
2
w
3
+ w
1
v
2
v
3
− w
1
v
2
v
3
+ v
1
w
2
v
3
− v
1
w
2
v
3
= 0 , after rearranging the terms.
∴ v××× w ⊥ v by Corollary 1.7.
The proof that v××× w ⊥ w is similar. QED
As a consequence of the above theorem and Theorem 1.9, we have the following:
Corollary 1.12. If the cross product v ××× w of two nonzero vectors v and w is also a
nonzero vector, then it is perpendicular to the span of v and w.
1.4 Cross Product 21
The span of any two nonzero, nonparallel vectors v, w in
3
is a plane P, so the
above corollary shows that v ××× w is perpendicular to that plane. As shown in Figure
1.4.2, there are two possible directions for v×××w, one the opposite of the other. It turns
out (see Appendix B) that the direction of v ××× w is given by the right-hand rule, that
is, the vectors v, w, v ××× w form a right-handed system. Recall from Section 1.1 that
this means that you can point your thumb upwards in the direction of v ××× w while
rotating v towards w with the remaining four fingers.
x
y
z
0
θ
v
w
v××× w
−v××× w
P
Figure 1.4.2 Direction of v××× w
We will now derive a formula for the magnitude of v××× w, for nonzero vectors v, w:
v××× w
2
= (v
2
w
3
− v
3
w
2
)
2
+ (v
3
w
1
− v
1
w
3
)
2
+ (v
1
w
2
− v
2
w
1
)
2
= v
2
2
w
2
3
− 2v
2
w
2
v
3
w
3
+ v
2
3
w
2
2
+ v
2
3
w
2
1
− 2v
1
w
1
v
3
w
3
+ v
2
1
w
2
3
+ v
2
1
w
2
2
− 2v
1
w
1
v
2
w
2
+ v
2
2
w
2
1
= v
2
1
(w
2
2
+ w
2
3
) + v
2
2
(w
2
1
+ w
2
3
) + v
2
3
(w
2
1
+ w
2
2
) − 2(v
1
w
1
v
2
w
2
+ v
1
w
1
v
3
w
3
+ v
2
w
2
v
3
w
3
)
and now adding and subtracting v
2
1
w
2
1
, v
2
2
w
2
2
, and v
2
3
w
2
3
on the right side gives
= v
2
1
(w
2
1
+ w
2
2
+ w
2
3
) + v
2
2
(w
2
1
+ w
2
2
+ w
2
3
) + v
2
3
(w
2
1
+ w
2
2
+ w
2
3
)
− (v
2
1
w
2
1
+ v
2
2
w
2
2
+ v
2
3
w
2
3
+ 2(v
1
w
1
v
2
w
2
+ v
1
w
1
v
3
w
3
+ v
2
w
2
v
3
w
3
))
= (v
2
1
+ v
2
2
+ v
2
3
)(w
2
1
+ w
2
2
+ w
2
3
)
− ((v
1
w
1
)
2
+ (v
2
w
2
)
2
+ (v
3
w
3
)
2
+ 2(v
1
w
1
)(v
2
w
2
) + 2(v
1
w
1
)(v
3
w
3
) + 2(v
2
w
2
)(v
3
w
3
))
so using (a + b + c)
2
= a
2
+ b
2
+ c
2
+ 2ab + 2ac + 2bc for the subtracted term gives
= (v
2
1
+ v
2
2
+ v
2
3
)(w
2
1
+ w
2
2
+ w
2
3
) − (v
1
w
1
+ v
2
w
2
+ v
3
w
3
)
2
= v
2
w
2
− (v··· w)
2
= v
2
w
2

1 −
(v··· w)
2
v
2
w
2

, since v > 0 and w > 0, so by Theorem 1.6
= v
2
w
2
(1 − cos
2
θ) , where θ is the angle between v and w, so
v××× w
2
= v
2
w
2
sin
2
θ , and since 0

≤ θ ≤ 180

, then sin θ ≥ 0, so we have:
22 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
If θ is the angle between nonzero vectors v and w in
3
, then
v××× w = v w sin θ (1.11)
It may seem strange to bother with the above formula, when the magnitude of the
cross product can be calculated directly, like for any other vector. The formula is more
useful for its applications in geometry, as in the following example.
Example 1.8. Let △PQR and PQRS be a triangle and parallelogram, respectively, as
shown in Figure 1.4.3.
b
h h
θ θ
P P
Q Q R R
S S
v
w
Figure 1.4.3
Think of the triangle as existing in
3
, and identify the sides QR and QP with vectors
v and w, respectively, in
3
. Let θ be the angle between v and w. The area A
PQR
of
△PQR is
1
2
bh, where b is the base of the triangle and h is the height. So we see that
b = v and h = w sin θ
A
PQR
=
1
2
v w sin θ
=
1
2
v××× w
So since the area A
PQRS
of the parallelogram PQRS is twice the area of the triangle
△PQR, then
A
PQRS
= v w sin θ
By the discussion in Example 1.8, we have proved the following theorem:
Theorem 1.13. Area of triangles and parallelograms
(a) The area A of a triangle with adjacent sides v, w (as vectors in
3
) is:
A =
1
2
v××× w
(b) The area A of a parallelogram with adjacent sides v, w (as vectors in
3
) is:
A = v××× w
1.4 Cross Product 23
It may seem at first glance that since the formulas derived in Example 1.8 were
for the adjacent sides QP and QR only, then the more general statements in Theorem
1.13 that the formulas hold for any adjacent sides are not justified. We would get a
different formula for the area if we had picked PQ and PR as the adjacent sides, but it
can be shown (see Exercise 26) that the different formulas would yield the same value,
so the choice of adjacent sides indeed does not matter, and Theorem 1.13 is valid.
Theorem 1.13 makes it simpler to calculate the area of a triangle in 3-dimensional
space than by using traditional geometric methods.
Example 1.9. Calculate the area of the triangle △PQR, where P = (2, 4, −7), Q =
(3, 7, 18), and R = (−5, 12, 8).
y
z
x
0
v
w
R(−5, 12, 8)
Q(3, 7, 18)
P(2, 4, −7)
Figure 1.4.4
Solution: Let v =
−−→
PQ and w =
−−→
PR, as in Figure 1.4.4.
Then v = (3, 7, 18) −(2, 4, −7) = (1, 3, 25) and w = (−5, 12, 8) −
(2, 4, −7) = (−7, 8, 15), so the area A of the triangle △PQR is
A =
1
2
v××× w =
1
2
(1, 3, 25) ××× (−7, 8, 15)
=
1
2
¸
¸
¸((3)(15) − (25)(8), (25)(−7) − (1)(15), (1)(8) − (3)(−7))
¸
¸
¸
=
1
2
¸
¸
¸(−155, −190, 29)
¸
¸
¸
=
1
2

(−155)
2
+ (−190)
2
+ 29
2
=
1
2

60966
A ≈ 123.46
Example 1.10. Calculate the area of the parallelogram PQRS , where P = (1, 1), Q =
(2, 3), R = (5, 4), and S = (4, 2).
x
y
0
1
2
3
4
1 2 3 4 5
P
Q
R
S
v
w
Figure 1.4.5
Solution: Let v =
−−→
S P and w =
−−→
S R, as in Figure 1.4.5. Then
v = (1, 1) − (4, 2) = (−3, −1) and w = (5, 4) − (4, 2) = (1, 2).
But these are vectors in
2
, and the cross product is only
defined for vectors in
3
. However,
2
can be thought of
as the subset of
3
such that the z-coordinate is always 0.
So we can write v = (−3, −1, 0) and w = (1, 2, 0). Then the
area A of PQRS is
A = v××× w =
¸
¸
¸(−3, −1, 0) ××× (1, 2, 0)
¸
¸
¸
=
¸
¸
¸((−1)(0) − (0)(2), (0)(1) − (−3)(0), (−3)(2) − (−1)(1))
¸
¸
¸
=
¸
¸
¸(0, 0, −5)
¸
¸
¸
A = 5
24 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
The following theorem summarizes the basic properties of the cross product.
Theorem 1.14. For any vectors u, v, w in
3
, and scalar k, we have
(a) v××× w = −w××× v Anticommutative Law
(b) u××× (v + w) = u××× v + u××× w Distributive Law
(c) (u + v) ××× w = u××× w+ v××× w Distributive Law
(d) (kv) ××× w = v××× (kw) = k(v××× w) Associative Law
(e) v××× 0 = 0 = 0 ××× v
(f) v××× v = 0
(g) v××× w = 0 if and only if v w
Proof: The proofs of properties (b)-(f) are straightforward. We will prove parts (a)
and (g) and leave the rest to the reader as exercises.
x
y
z
0
v
w
v××× w
w××× v
Figure 1.4.6
(a) By the definition of the cross product and scalar multi-
plication, we have:
v××× w = (v
2
w
3
− v
3
w
2
, v
3
w
1
− v
1
w
3
, v
1
w
2
− v
2
w
1
)
= −(v
3
w
2
− v
2
w
3
, v
1
w
3
− v
3
w
1
, v
2
w
1
− v
1
w
2
)
= −(w
2
v
3
− w
3
v
2
, w
3
v
1
− w
1
v
3
, w
1
v
2
− w
2
v
1
)
= −w××× v
Note that this says that v ××× w and w××× v have the same
magnitude but opposite direction (see Figure 1.4.6).
(g) If either v or w is 0 then v×××w = 0 by part (e), and either v = 0 = 0wor w = 0 = 0v,
so v and w are scalar multiples, i.e. they are parallel.
If both v and w are nonzero, and θ is the angle between them, then by formula
(1.11), v ××× w = 0 if and only if v w sin θ = 0, which is true if and only if sin θ = 0
(since v > 0 and w > 0). So since 0

≤ θ ≤ 180

, then sin θ = 0 if and only if θ = 0

or 180

. But the angle between v and w is 0

or 180

if and only if v w. QED
Example 1.11. Adding to Example 1.7, we have
i ××× j = k j ××× k = i k××× i = j
j ××× i = −k k××× j = −i i ××× k = −j
i ××× i = j ××× j = k××× k = 0
Recall from geometry that a parallelepiped is a 3-dimensional solid with 6 faces, all
of which are parallelograms.
6
6
An equivalent definition of a parallelepiped is: the collection of all scalar combinations k
1
v
1
+k
2
v
2
+k
3
v
3
of some vectors v
1
, v
2
, v
3
in
3
, where 0 ≤ k
1
, k
2
, k
3
≤ 1.
1.4 Cross Product 25
Example 1.12. Volume of a parallelepiped: Let the vectors u, v, w in
3
represent
adjacent sides of a parallelepiped P, as in Figure 1.4.7. Show that the volume of P is
the scalar triple product u··· (v××× w).
h
θ
u
w
v
v××× w
Figure 1.4.7 Parallelepiped P
Solution: Recall that the volume of a par-
allelepiped is the area A of the base paral-
lelogram times the height h. By Theorem
1.13(b), the area A of the base parallelogram
is v×××w. And we can see that since v×××w is
perpendicular to the base parallelogram de-
termined by v and w, then the height h is
u cos θ, where θ is the angle between u and
v××× w. By Theorem 1.6 we know that
cos θ =
u··· (v××× w)
u v××× w
. Hence,
vol(P) = Ah
= v××× w
u u··· (v××× w)
u v××× w
= u··· (v××× w)
In Example 1.12 the height h of the parallelepiped is u cos θ, and not −u cos θ,
because the vector u is on the same side of the base parallelogram’s plane as the vector
v×××w(so that cos θ > 0). Since the volume is the same no matter which base and height
we use, then repeating the same steps using the base determined by u and v (since w
is on the same side of that base’s plane as u××× v), the volume is w··· (u××× v). Repeating
this with the base determined by w and u, we have the following result:
For any vectors u, v, w in
3
,
u··· (v××× w) = w··· (u××× v) = v··· (w××× u) (1.12)
(Note that the equalities hold trivially if any of the vectors are 0.)
Since v ××× w = −w××× v for any vectors v, w in
3
, then picking the wrong order for
the three adjacent sides in the scalar triple product in formula (1.12) will give you the
negative of the volume of the parallelepiped. So taking the absolute value of the scalar
triple product for any order of the three adjacent sides will always give the volume:
Theorem 1.15. If vectors u, v, w in
3
represent any three adjacent sides of a paral-
lelepiped, then the volume of the parallelepiped is |u··· (v××× w)|.
Another type of triple product is the vector triple product u××× (v ××× w). The proof of
the following theorem is left as an exercise for the reader:
26 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Theorem 1.16. For any vectors u, v, w in
3
,
u××× (v××× w) = (u··· w)v − (u··· v)w (1.13)
An examination of the formula in Theorem 1.16 gives some idea of the geometry of
the vector triple product. By the right side of formula (1.13), we see that u××× (v ××× w)
is a scalar combination of v and w, and hence lies in the plane containing v and w
(i.e. u ××× (v ××× w), v and w are coplanar). This makes sense since, by Theorem 1.11,
u ××× (v ××× w) is perpendicular to both u and v ××× w. In particular, being perpendicular
to v××× w means that u××× (v××× w) lies in the plane containing v and w, since that plane
is itself perpendicular to v ××× w. But then how is u××× (v ××× w) also perpendicular to u,
which could be any vector? The following example may help to see how this works.
Example 1.13. Find u××× (v××× w) for u = (1, 2, 4), v = (2, 2, 0), w = (1, 3, 0).
Solution: Since u··· v = 6 and u··· w = 7, then
u××× (v××× w) = (u··· w)v − (u··· v)w
= 7 (2, 2, 0) − 6 (1, 3, 0) = (14, 14, 0) − (6, 18, 0)
= (8, −4, 0)
Note that v and w lie in the xy-plane, and that u×××(v×××w) also lies in that plane. Also,
u××× (v××× w) is perpendicular to both u and v××× w = (0, 0, 4) (see Figure 1.4.8).
y
z
x
0
u
v
w
v ××× w
u ××× (v ××× w)
Figure 1.4.8
For vectors v = v
1
i + v
2
j + v
3
k and w = w
1
i + w
2
j + w
3
k in component form, the cross
product is written as: v ××× w = (v
2
w
3
− v
3
w
2
)i + (v
3
w
1
− v
1
w
3
)j + (v
1
w
2
− v
2
w
1
)k. It is often
easier to use the component form for the cross product, because it can be represented
as a determinant. We will not go too deeply into the theory of determinants
7
; we will
just cover what is essential for our purposes.
7
See ANTON and RORRES for a fuller development.
1.4 Cross Product 27
A 2 × 2 matrix is an array of two rows and two columns of scalars, written as
,
a b
c d
¸
or
¸
a b
c d

where a, b, c, d are scalars. The determinant of such a matrix, written as

a b
c d

or det
,
a b
c d
¸
,
is the scalar defined by the following formula:

a b
c d

It may help to remember this formula as being the product of the scalars on the down-
ward diagonal minus the product of the scalars on the upward diagonal.
Example 1.14.

1 2
3 4

= (1)(4) − (2)(3) = 4 − 6 = −2
A 3 × 3 matrix is an array of three rows and three columns of scalars, written as
,
¸
¸
¸
¸
¸
¸
¸
¸
¸
a
1
a
2
a
3
b
1
b
2
b
3
c
1
c
2
c
3
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
or

¸
¸
¸
¸
¸
¸
¸
¸
¸
a
1
a
2
a
3
b
1
b
2
b
3
c
1
c
2
c
3

¸
¸
¸
¸
¸
¸
¸
¸
¸
,
and its determinant is given by the formula:

a
1
a
2
a
3
b
1
b
2
b
3
c
1
c
2
c
3

= a
1

b
2
b
3
c
2
c
3

− a
2

b
1
b
3
c
1
c
3

+ a
3

b
1
b
2
c
1
c
2

(1.14)
One way to remember the above formula is the following: multiply each scalar in the
first row by the determinant of the 2 × 2 matrix that remains after removing the row
and column that contain that scalar, then sum those products up, putting alternating
plus and minus signs in front of each (starting with a plus).
Example 1.15.

1 0 2
4 −1 3
1 0 2

= 1

−1 3
0 2

− 0

4 3
1 2

+ 2

4 −1
1 0

= 1(−2 − 0) − 0(8 − 3) + 2(0 + 1) = 0
28 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
We defined the determinant as a scalar, derived from algebraic operations on scalar
entries in a matrix. However, if we put three vectors in the first row of a 3 × 3 matrix,
then the definition still makes sense, since we would be performing scalar multiplica-
tion on those three vectors (they would be multiplied by the 2 × 2 scalar determinants
as before). This gives us a determinant that is now a vector, and lets us write the
cross product of v = v
1
i + v
2
j + v
3
k and w = w
1
i + w
2
j + w
3
k as a determinant:
v××× w =

i j k
v
1
v
2
v
3
w
1
w
2
w
3

=

v
2
v
3
w
2
w
3

i −

v
1
v
3
w
1
w
3

j +

v
1
v
2
w
1
w
2

k
= (v
2
w
3
− v
3
w
2
)i + (v
3
w
1
− v
1
w
3
)j + (v
1
w
2
− v
2
w
1
)k
Example 1.16. Let v = 4 i − j + 3 k and w = i + 2 k. Then
v××× w =

i j k
4 −1 3
1 0 2

=

−1 3
0 2

i −

4 3
1 2

j +

4 −1
1 0

k = −2 i − 5 j + k
The scalar triple product can also be written as a determinant. In fact, by Example
1.12, the following theorem provides an alternate definition of the determinant of a
3 × 3 matrix as the volume (or negative volume) of a parallelepiped whose adjacent
sides are the rows of the matrix. The proof is left as an exercise for the reader.
Theorem 1.17. For any vectors u = (u
1
, u
2
, u
3
), v = (v
1
, v
2
, v
3
), w = (w
1
, w
2
, w
3
) in
3
:
u··· (v××× w) =

u
1
u
2
u
3
v
1
v
2
v
3
w
1
w
2
w
3

(1.15)
Example 1.17. Find the volume of the parallelepiped with adjacent sides u = (2, 1, 3),
v = (−1, 3, 2), w = (1, 1, −2) (see Figure 1.4.9).
y
z
x
0
u
v
w
Figure 1.4.9 P
Solution: By Theorem 1.15, the volume of the parallelepiped
P is the absolute value of the scalar triple product of the three
adjacent sides (in any order). By Theorem 1.17,
u··· (v××× w) =

2 1 3
−1 3 2
1 1 −2

= 2

3 2
1 −2

− 1

−1 2
1 −2

+ 3

−1 3
1 1

= 2(−8) − 1(0) + 3(−4) = −28, so
vol(P) = |−28| = 28.
1.4 Cross Product 29
Interchanging the dot and cross products can be useful in proving vector identities:
Example 1.18. Prove: (u××× v) ··· (w××× z) =

u··· w u··· z
v··· w v··· z

for all vectors u, v, w, z in
3
.
Solution: Let x = u××× v. Then
(u××× v) ··· (w××× z) = x··· (w××× z)
= w··· (z ××× x) (by formula (1.12))
= w··· (z ××× (u××× v))
= w··· ((z ··· v)u − (z ··· u)v) (by Theorem 1.16)
= (z ··· v)(w··· u) − (z ··· u)(w··· v)
= (u··· w)(v··· z) − (u··· z)(v··· w) (by commutativity of the dot product).
=

u··· w u··· z
v··· w v··· z

¨
Exercises
A
For Exercises 1-6, calculate v××× w.
1. v = (5, 1, −2), w = (4, −4, 3) 2. v = (7, 2, −10), w = (2, 6, 4)
3. v = (2, 1, 4), w = (1, −2, 0) 4. v = (1, 3, 2), w = (7, 2, −10)
5. v = −i + 2 j + k, w = −3 i + 6 j + 3 k 6. v = i, w = 3 i + 2 j + 4k
For Exercises 7-8, calculate the area of the triangle △PQR.
7. P = (5, 1, −2), Q = (4, −4, 3), R = (2, 4, 0) 8. P = (4, 0, 2), Q = (2, 1, 5), R = (−1, 0, −1)
For Exercises 9-10, calculate the area of the parallelogram PQRS .
9. P = (2, 1, 3), Q = (1, 4, 5), R = (2, 5, 3), S = (3, 2, 1)
10. P = (−2, −2), Q = (1, 4), R = (6, 6), S = (3, 0)
For Exercises 11-12, find the volume of the parallelepiped with adjacent sides u, v, w.
11. u = (1, 1, 3), v = (2, 1, 4), w = (5, 1, −2) 12. u = (1, 3, 2), v = (7, 2, −10), w = (1, 0, 1)
For Exercises 13-14, calculate u··· (v××× w) and u××× (v××× w).
13. u = (1, 1, 1), v = (3, 0, 2), w = (2, 2, 2) 14. u = (1, 0, 2), v = (−1, 0, 3), w = (2, 0, −2)
15. Calculate (u××× v) ··· (w××× z) for u = (1, 1, 1), v = (3, 0, 2), w = (2, 2, 2), z = (2, 1, 4).
30 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
B
16. If v and w are unit vectors in
3
, under what condition(s) would v ××× w also be a
unit vector in
3
17. Show that if v××× w = 0 for all w in
3
, then v = 0.
18. Prove Theorem 1.14(b). 19. Prove Theorem 1.14(c).
20. Prove Theorem 1.14(d). 21. Prove Theorem 1.14(e).
22. Prove Theorem 1.14(f). 23. Prove Theorem 1.16.
24. Prove Theorem 1.17. (Hint: Expand both sides of the equation.)
25. Prove the following for all vectors v, w in
3
:
(a) v××× w
2
+ |v··· w|
2
= v
2
w
2
(b) If v··· w = 0 and v××× w = 0, then v = 0 or w = 0.
C
26. Prove that in Example 1.8 the formula for the area of the triangle △PQR yields the
same value no matter which two adjacent sides are chosen. To do this, show that
1
2
u××× (−w) =
1
2
v ××× w, where u = PR, −w = PQ, and v = QR, w = QP as before.
Similarly, show that
1
2
(−u) ××× (−v) =
1
2
v××× w, where −u = RP and −v = RQ.
27. Consider the vector equation a××× x = b in
3
, where a 0. Show that:
(a) a··· b = 0
(b) x =
b××× a
a
2
+ ka is a solution to the equation, for any scalar k
28. Prove the Jacobi identity: u××× (v××× w) + v××× (w××× u) + w××× (u××× v) = 0
29. Show that u, v, w lie in the same plane in
3
if and only if u··· (v××× w) = 0.
30. For all vectors u, v, w, z in
3
, show that
(u××× v) ××× (w××× z) = (z ··· (u××× v))w− (w··· (u××× v))z
and that
(u××× v) ××× (w××× z) = (u··· (w××× z))v − (v··· (w××× z))u
Why do both equations make sense geometrically?
1.5 Lines and Planes 31
1.5 Lines and Planes
Now that we know how to perform some operations on vectors, we can start to deal
with some familiar geometric objects, like lines and planes, in the language of vectors.
The reason for doing this is simple: using vectors makes it easier to study objects in
3-dimensional Euclidean space. We will first consider lines.
Line through a point, parallel to a vector
Let P = (x
0
, y
0
, z
0
) be a point in
3
, let v = (a, b, c) be a nonzero vector, and let L be the
line through P which is parallel to v (see Figure 1.5.1).
x
y
z
0
L
t > 0
t < 0
P(x
0
, y
0
, z
0
)
r
v
tv
r + tv
r + tv
Figure 1.5.1
Let r = (x
0
, y
0
, z
0
) be the vector pointing from the origin to P. Since multiplying the
vector v by a scalar t lengthens or shrinks v while preserving its direction if t > 0, and
reversing its direction if t < 0, then we see from Figure 1.5.1 that every point on the
line L can be obtained by adding the vector tv to the vector r for some scalar t. That
is, as t varies over all real numbers, the vector r +tv will point to every point on L. We
can summarize the vector representation of L as follows:
For a point P = (x
0
, y
0
, z
0
) and nonzero vector v in
3
, the line L through P parallel
to v is given by
r + tv, for − ∞ < t < ∞ (1.16)
where r = (x
0
, y
0
, z
0
) is the vector pointing to P.
Note that we used the correspondence between a vector and its terminal point.
Since v = (a, b, c), then the terminal point of the vector r + tv is (x
0
+ at, y
0
+ bt, z
0
+ ct).
We then get the parametric representation of L with the parameter t:
For a point P = (x
0
, y
0
, z
0
) and nonzero vector v = (a, b, c) in
3
, the line L through P
parallel to v consists of all points (x, y, z) given by
x = x
0
+ at, y = y
0
+ bt, z = z
0
+ ct, for − ∞ < t < ∞ (1.17)
Note that in both representations we get the point P on L by letting t = 0.
32 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
In formula (1.17), if a 0, then we can solve for the parameter t: t = (x − x
0
)/a. We
can also solve for t in terms of y and in terms of z if neither b nor c, respectively, is
zero: t = (y−y
0
)/b and t = (z −z
0
)/c. These three values all equal the same value t, so we
can write the following system of equalities, called the symmetric representation of L:
For a point P = (x
0
, y
0
, z
0
) and vector v = (a, b, c) in
3
with a, b and c all nonzero, the
line L through P parallel to v consists of all points (x, y, z) given by the equations
x − x
0
a
=
y − y
0
b
=
z − z
0
c
(1.18)
x
y
z
0
x = x
0
x
0
L
Figure 1.5.2
What if, say, a = 0 in the above scenario? We can not divide
by zero, but we do know that x = x
0
+ at, and so x = x
0
+ 0t = x
0
.
Then the symmetric representation of L would be:
x = x
0
,
y − y
0
b
=
z − z
0
c
(1.19)
Note that this says that the line L lies in the plane x = x
0
, which
is parallel to the yz-plane (see Figure 1.5.2). Similar equations
can be derived for the cases when b = 0 or c = 0.
You may have noticed that the vector representation of L in formula (1.16) is more
compact than the parametric and symmetric formulas. That is an advantage of using
vector notation. Technically, though, the vector representation gives us the vectors
whose terminal points make up the line L, not just L itself. So you have to remem-
ber to identify the vectors r + tv with their terminal points. On the other hand, the
parametric representation always gives just the points on L and nothing else.
Example 1.19. Write the line L through the point P = (2, 3, 5) and parallel to the
vector v = (4, −1, 6), in the following forms: (a) vector, (b) parametric, (c) symmetric.
Lastly: (d) find two points on L distinct from P.
Solution: (a) Let r = (2, 3, 5). Then by formula (1.16), L is given by:
r + tv = (2, 3, 5) + t(4, −1, 6), for − ∞ < t < ∞
(b) L consists of the points (x, y, z) such that
x = 2 + 4t, y = 3 − t, z = 5 + 6t, for − ∞ < t < ∞
(c) L consists of the points (x, y, z) such that
x − 2
4
=
y − 3
−1
=
z − 5
6
(d) Letting t = 1 and t = 2 in part(b) yields the points (6, 2, 11) and (10, 1, 17) on L.
1.5 Lines and Planes 33
Line through two points
x
y
z
0
L
P
1
(x
1
, y
1
, z
1
)
P
2
(x
2
, y
2
, z
2
)
r
1
r
2
r
2
− r
1
r
1
+ t(r
2
− r
1
)
Figure 1.5.3
Let P
1
= (x
1
, y
1
, z
1
) and P
2
= (x
2
, y
2
, z
2
) be distinct
points in
3
, and let L be the line through P
1
and
P
2
. Let r
1
= (x
1
, y
1
, z
1
) and r
2
= (x
2
, y
2
, z
2
) be the vectors
pointing to P
1
and P
2
, respectively. Then as we can
see from Figure 1.5.3, r
2
− r
1
is the vector from P
1
to
P
2
. So if we multiply the vector r
2
− r
1
by a scalar t
and add it to the vector r
1
, we will get the entire line
L as t varies over all real numbers. The following is
a summary of the vector, parametric, and symmetric
forms for the line L:
Let P
1
= (x
1
, y
1
, z
1
), P
2
= (x
2
, y
2
, z
2
) be distinct points in
3
, and let r
1
= (x
1
, y
1
, z
1
),
r
2
= (x
2
, y
2
, z
2
). Then the line L through P
1
and P
2
has the following representations:
Vector:
r
1
+ t(r
2
− r
1
) , for − ∞ < t < ∞ (1.20)
Parametric:
x = x
1
+ (x
2
− x
1
)t, y = y
1
+ (y
2
− y
1
)t, z = z
1
+ (z
2
− z
1
)t, for − ∞ < t < ∞ (1.21)
Symmetric:
x − x
1
x
2
− x
1
=
y − y
1
y
2
− y
1
=
z − z
1
z
2
− z
1
(if x
1
x
2
, y
1
y
2
, and z
1
z
2
) (1.22)
Example 1.20. Write the line L through the points P
1
= (−3, 1, −4) and P
2
= (4, 4, −6) in
parametric form.
Solution: By formula (1.21), L consists of the points (x, y, z) such that
x = −3 + 7t, y = 1 + 3t, z = −4 − 2t, for − ∞ < t < ∞
Distance between a point and a line
θ L
v
w d
Q
P
Figure 1.5.4
Let L be a line in
3
in vector form as r + tv (for −∞ < t < ∞),
and let P be a point not on L. The distance d from P to L is the
length of the line segment from P to L which is perpendicular
to L (see Figure 1.5.4). Pick a point Q on L, and let w be the
vector from Q to P. If θ is the angle between w and v, then
d = w sin θ. So since v××× w = v w sin θ and v 0, then:
d =
v××× w
v
(1.23)
34 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Example 1.21. Find the distance d from the point P = (1, 1, 1) to the line L in Example
1.20.
Solution: From Example 1.20, we see that we can represent L in vector form as: r+tv,
for r = (−3, 1, −4) and v = (7, 3, −2). Since the point Q = (−3, 1, −4) is on L, then for
w =
−−→
QP = (1, 1, 1) − (−3, 1, −4) = (4, 0, 5), we have:
v××× w =

i j k
7 3 −2
4 0 5

=

3 −2
0 5

i −

7 −2
4 5

j +

7 3
4 0

k = 15 i − 43 j − 12 k , so
d =
v××× w
v
=
¸
¸
¸15 i − 43 j − 12 k
¸
¸
¸
¸
¸
¸(7, 3, −2)
¸
¸
¸
=

15
2
+ (−43)
2
+ (−12)
2

7
2
+ 3
2
+ (−2)
2
=

2218

62
= 5.98
It is clear that two lines L
1
and L
2
, represented in vector form as r
1
+ sv
1
and r
2
+tv
2
,
respectively, are parallel (denoted as L
1
L
2
) if v
1
and v
2
are parallel. Also, L
1
and L
2
are perpendicular (denoted as L
1
⊥ L
2
) if v
1
and v
2
are perpendicular.
x
y
z
0
L
1
L
2
Figure 1.5.5
In 2-dimensional space, two lines are either identical, parallel, or
they intersect. In 3-dimensional space, there is an additional possi-
bility: two lines can be skew, that is, they do not intersect but they
are not parallel. However, even though they are not parallel, skew
lines are on parallel planes (see Figure 1.5.5).
To determine whether two lines in
3
intersect, it is often easier
to use the parametric representation of the lines. In this case, you
should use different parameter variables (usually s and t) for the lines, since the val-
ues of the parameters may not be the same at the point of intersection. Setting the
two (x, y, z) triples equal will result in a system of 3 equations in 2 unknowns (s and t).
Example 1.22. Find the point of intersection (if any) of the following lines:
x + 1
3
=
y − 2
2
=
z − 1
−1
and x + 3 =
y − 8
−3
=
z + 3
2
Solution: First we write the lines in parametric form, with parameters s and t:
x = −1 + 3s, y = 2 + 2s, z = 1 − s and x = −3 + t, y = 8 − 3t, z = −3 + 2t
The lines intersect when (−1 + 3s, 2 + 2s, 1 − s) = (−3 + t, 8 − 3t, −3 + 2t) for some s, t:
−1 + 3s = −3 + t : ⇒ t = 2 + 3s
2 + 2s = 8 − 3t : ⇒ 2 + 2s = 8 − 3(2 + 3s) = 2 − 9s ⇒ 2s = −9s ⇒ s = 0 ⇒ t = 2 + 3(0) = 2
1 − s = −3 + 2t : 1 − 0 = −3 + 2(2) ⇒ 1 = 1 (Note that we had to check this.)
Letting s = 0 in the equations for the first line, or letting t = 2 in the equations for the
second line, gives the point of intersection (−1, 2, 1).
1.5 Lines and Planes 35
We will now consider planes in 3-dimensional Euclidean space.
Plane through a point, perpendicular to a vector
Let P be a plane in
3
, and suppose it contains a point P
0
= (x
0
, y
0
, z
0
). Let n = (a, b, c)
be a nonzero vector which is perpendicular to the plane P. Such a vector is called a
normal vector (or just a normal) to the plane. Now let (x, y, z) be any point in the
plane P. Then the vector r = (x − x
0
, y − y
0
, z − z
0
) lies in the plane P (see Figure 1.5.6).
So if r 0, then r ⊥ n and hence n··· r = 0. And if r = 0 then we still have n··· r = 0.
(x
0
, y
0
, z
0
) (x, y, z)
n
r
Figure 1.5.6 The plane P
Conversely, if (x, y, z) is any point in
3
such that r = (x − x
0
, y − y
0
, z − z
0
) 0 and
n··· r = 0, then r ⊥ n and so (x, y, z) lies in P. This proves the following theorem:
Theorem 1.18. Let P be a plane in
3
, let (x
0
, y
0
, z
0
) be a point in P, and let n = (a, b, c)
be a nonzero vector which is perpendicular to P. Then P consists of the points (x, y, z)
satisfying the vector equation:
n··· r = 0 (1.24)
where r = (x − x
0
, y − y
0
, z − z
0
), or equivalently:
a(x − x
0
) + b(y − y
0
) + c(z − z
0
) = 0 (1.25)
The above equation is called the point-normal form of the plane P.
Example 1.23. Find the equation of the plane P containing the point (−3, 1, 3) and
perpendicular to the vector n = (2, 4, 8).
Solution: By formula (1.25), the plane P consists of all points (x, y, z) such that:
2(x + 3) + 4(y − 1) + 8(z − 3) = 0
If we multiply out the terms in formula (1.25) and combine the constant terms, we
get an equation of the plane in normal form:
ax + by + cz + d = 0 (1.26)
For example, the normal form of the plane in Example 1.23 is 2x + 4y + 8z − 22 = 0.
36 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Plane containing three noncollinear points
In 2-dimensional and 3-dimensional space, two points determine a line. Two points
do not determine a plane in
3
. In fact, three collinear points (i.e. all on the same
line) do not determine a plane; an infinite number of planes would contain the line
on which those three points lie. However, three noncollinear points do determine a
plane. For if Q, R and S are noncollinear points in
3
, then
−−→
QR and
−−→
QS are nonzero
vectors which are not parallel (by noncollinearity), and so their cross product
−−→
QR×××
−−→
QS
is perpendicular to both
−−→
QR and
−−→
QS . So
−−→
QR and
−−→
QS (and hence Q, R and S ) lie in the
plane through the point Q with normal vector n =
−−→
QR×××
−−→
QS (see Figure 1.5.7).
Q
R
S
n =
−−→
QR×××
−−→
QS
−−→
QR
−−→
QS
Figure 1.5.7 Noncollinear points Q, R, S
Example 1.24. Find the equation of the plane P containing the points (2, 1, 3), (1, −1, 2)
and (3, 2, 1).
Solution: Let Q = (2, 1, 3), R = (1, −1, 2) and S = (3, 2, 1). Then for the vectors
−−→
QR =
(−1, −2, −1) and
−−→
QS = (1, 1, −2), the plane P has a normal vector
n =
−−→
QR×××
−−→
QS = (−1, −2, −1) ××× (1, 1, −2) = (5, −3, 1)
So using formula (1.25) with the point Q (we could also use R or S ), the plane P consists
of all points (x, y, z) such that:
5(x − 2) − 3(y − 1) + (z − 3) = 0
or in normal form,
5x − 3y + z − 10 = 0
We mentioned earlier that skew lines in
3
lie on separate, parallel planes. So
two skew lines do not determine a plane. But two (nonidentical) lines which either
intersect or are parallel do determine a plane. In both cases, to find the equation of
the plane that contains those two lines, simply pick from the two lines a total of three
noncollinear points (i.e. one point from one line and two points from the other), then
use the technique above, as in Example 1.24, to write the equation. We will leave
examples of this as exercises for the reader.
1.5 Lines and Planes 37
Distance between a point and a plane
The distance between a point in
3
and a plane is the length of the line segment
from that point to the plane which is perpendicular to the plane. The following theo-
rem gives a formula for that distance.
Theorem 1.19. Let Q = (x
0
, y
0
, z
0
) be a point in
3
, and let P be a plane with normal
form ax + by + cz + d = 0 that does not contain Q. Then the distance D from Q to P is:
D =
|ax
0
+ by
0
+ cz
0
+ d|

a
2
+ b
2
+ c
2
(1.27)
Proof: Let R = (x, y, z) be any point in the plane P (so that ax + by + cz + d = 0) and
let r =
−−→
RQ = (x
0
− x, y
0
− y, z
0
− z). Then r 0 since Q does not lie in P. From the
normal form equation for P, we know that n = (a, b, c) is a normal vector for P. Now,
any plane divides
3
into two disjoint parts. Assume that n points toward the side
of P where the point Q is located. Place n so that its initial point is at R, and let θ be
the angle between r and n. Then 0

< θ < 90

, so cos θ > 0. Thus, the distance D is
cos θ r = |cos θ| r (see Figure 1.5.8).
Q
R
n
r D
θ
D
P
Figure 1.5.8
By Theorem 1.6 in Section 1.3, we know that cos θ =
n··· r
n r
, so
D = |cos θ| r =

n··· r

n r
r =

n··· r

n
=
|a(x
0
− x) + b(y
0
− y) + c(z
0
− z)|

a
2
+ b
2
+ c
2
=
|ax
0
+ by
0
+ cz
0
− (ax + by + cz)|

a
2
+ b
2
+ c
2
=
|ax
0
+ by
0
+ cz
0
− (−d)|

a
2
+ b
2
+ c
2
=
|ax
0
+ by
0
+ cz
0
+ d|

a
2
+ b
2
+ c
2
If n points away from the side of P where the point Q is located, then 90

< θ < 180

and so cos θ < 0. The distance D is then |cos θ| r, and thus repeating the same
argument as above still gives the same result. QED
Example 1.25. Find the distance D from (2, 4, −5) to the plane from Example 1.24.
Solution: Recall that the plane is given by 5x − 3y + z − 10 = 0. So
D =
|5(2) − 3(4) + 1(−5) − 10|

5
2
+ (−3)
2
+ 1
2
=
|−17|

35
=
17

35
≈ 2.87
38 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Line of intersection of two planes
L
Figure 1.5.9
Note that two planes are parallel if they have normal vectors
that are parallel, and the planes are perpendicular if their normal
vectors are perpendicular. If two planes do intersect, they do so in
a line (see Figure 1.5.9). Suppose that two planes P
1
and P
2
with
normal vectors n
1
and n
2
, respectively, intersect in a line L. Since
n
1
××× n
2
⊥ n
1
, then n
1
××× n
2
is parallel to the plane P
1
. Likewise,
n
1
××× n
2
⊥ n
2
means that n
1
××× n
2
is also parallel to P
2
. Thus, n
1
××× n
2
is parallel to the intersection of P
1
and P
2
, i.e. n
1
××× n
2
is parallel to L. Thus, we can
write L in the following vector form:
L : r + t(n
1
××× n
2
) , for − ∞ < t < ∞ (1.28)
where r is any vector pointing to a point belonging to both planes. To find a point in
both planes, find a common solution (x, y, z) to the two normal form equations of the
planes. This can often be made easier by setting one of the coordinate variables to
zero, which leaves you to solve two equations in just two unknowns.
Example 1.26. Find the line of intersection L of the planes 5x − 3y + z − 10 = 0 and
2x + 4y − z + 3 = 0.
Solution: The plane 5x − 3y + z − 10 = 0 has normal vector n
1
= (5, −3, 1) and the
plane 2x +4y −z +3 = 0 has normal vector n
2
= (2, 4, −1). Since n
1
and n
2
are not scalar
multiples, then the two planes are not parallel and hence will intersect. A point (x, y, z)
on both planes will satisfy the following system of two equations in three unknowns:
5x − 3y + z − 10 = 0
2x + 4y − z + 3 = 0
Set x = 0 (why is that a good choice?). Then the above equations are reduced to:
−3y + z − 10 = 0
4y − z + 3 = 0
The second equation gives z = 4y + 3, substituting that into the first equation gives
y = 7. Then z = 31, and so the point (0, 7, 31) is on L. Since n
1
×××n
2
= (−1, 7, 26), then L is
given by:
r + t(n
1
××× n
2
) = (0, 7, 31) + t(−1, 7, 26), for − ∞ < t < ∞
or in parametric form:
x = −t, y = 7 + 7t, z = 31 + 26t, for − ∞ < t < ∞
1.5 Lines and Planes 39

¨
Exercises
A
For Exercises 1-4, write the line L through the point P and parallel to the vector v in
the following forms: (a) vector, (b) parametric, and (c) symmetric.
1. P = (2, 3, −2), v = (5, 4, −3) 2. P = (3, −1, 2), v = (2, 8, 1)
3. P = (2, 1, 3), v = (1, 0, 1) 4. P = (0, 0, 0), v = (7, 2, −10)
For Exercises 5-6, write the line L through the points P
1
and P
2
in parametric form.
5. P
1
= (1, −2, −3), P
2
= (3, 5, 5) 6. P
1
= (4, 1, 5), P
2
= (−2, 1, 3)
For Exercises 7-8, find the distance d from the point P to the line L.
7. P = (1, −1, −1), L : x = −2 − 2t, y = 4t, z = 7 + t
8. P = (0, 0, 0), L : x = 3 + 2t, y = 4 + 3t, z = 5 + 4t
For Exercises 9-10, find the point of intersection (if any) of the given lines.
9. x = 7 + 3s, y = −4 − 3s, z = −7 − 5s and x = 1 + 6t, y = 2 + t, z = 3 − 2t
10.
x − 6
4
= y + 3 = z and
x − 11
3
=
y − 14
−6
=
z + 9
2
For Exercises 11-12, write the normal form of the plane P containing the point Q and
perpendicular to the vector n.
11. Q = (5, 1, −2), n = (4, −4, 3) 12. Q = (6, −2, 0), n = (2, 6, 4)
For Exercises 13-14, write the normal form of the plane containing the given points.
13. (1, 0, 3), (1, 2, −1), (6, 1, 6) 14. (−3, 1, −3), (4, −4, 3), (0, 0, 1)
15. Write the normal form of the plane containing the lines from Exercise 9.
16. Write the normal form of the plane containing the lines from Exercise 10.
For Exercises 17-18, find the distance D from the point Q to the plane P.
17. Q = (4, 1, 2), P : 3x − y − 5z + 8 = 0 18. Q = (0, 2, 0), P : −5x + 2y − 7z + 1 = 0
For Exercises 19-20, find the line of intersection (if any) of the given planes.
19. x + 3y + 2z − 6 = 0, 2x − y + z + 2 = 0 20. 3x + y − 5z = 0, x + 2y + z + 4 = 0
B
21. Find the point(s) of intersection (if any) of the line
x − 6
4
= y + 3 = z with the plane
x +3y +2z −6 = 0. (Hint: Put the equations of the line into the equation of the plane.)
40 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
1.6 Surfaces
In the previous section we discussed planes in Euclidean space. A plane is an exam-
ple of a surface, which we will define informally
8
as the solution set of the equation
F(x, y, z) = 0 in
3
, for some real-valued function F. For example, a plane given by
ax+by+cz+d = 0 is the solution set of F(x, y, z) = 0 for the function F(x, y, z) = ax+by+cz+d.
Surfaces are 2-dimensional. The plane is the simplest surface, since it is “flat”. In this
section we will look at some surfaces that are more complex, the most important of
which are the sphere and the cylinder.
Definition 1.9. A sphere S is the set of all points (x, y, z) in
3
which are a fixed
distance r (called the radius) from a fixed point P
0
= (x
0
, y
0
, z
0
) (called the center of
the sphere):
S = { (x, y, z) : (x − x
0
)
2
+ (y − y
0
)
2
+ (z − z
0
)
2
= r
2
} (1.29)
Using vector notation, this can be written in the equivalent form:
S = { x : x − x
0
= r } (1.30)
where x = (x, y, z) and x
0
= (x
0
, y
0
, z
0
) are vectors.
Figure 1.6.1 illustrates the vectorial approach to spheres.
y
z
x
0
x = r
x
(a) radius r, center (0, 0, 0)
y
z
x
0
x − x
0
= r
x
x
0
x − x
0
(x
0
, y
0
, z
0
)
0
, y
0
, z
0
)
Figure 1.6.1 Spheres in
3
Note in Figure 1.6.1(a) that the intersection of the sphere with the xy-plane is a
circle of radius r (i.e. a great circle, given by x
2
+ y
2
= r
2
as a subset of
2
). Similarly
for the intersections with the xz-plane and the yz-plane. In general, a plane intersects
a sphere either at a single point or in a circle.
8
See O’NEILL for a deeper and more rigorous discussion of surfaces.
1.6 Surfaces 41
Example 1.27. Find the intersection of the sphere x
2
+ y
2
+ z
2
= 169 with the plane
z = 12.
y
z
x
0
z = 12
Figure 1.6.2
Solution: The sphere is centered at the origin and has

169, so it does intersect the plane z = 12.
Putting z = 12 into the equation of the sphere gives
x
2
+ y
2
+ 12
2
= 169
x
2
+ y
2
= 169 − 144 = 25 = 5
2
which is a circle of radius 5 centered at (0, 0, 12), parallel
to the xy-plane (see Figure 1.6.2).
If the equation in formula (1.29) is multiplied out, we get an equation of the form:
x
2
+ y
2
+ z
2
+ ax + by + cz + d = 0 (1.31)
for some constants a, b, c and d. Conversely, an equation of this form may describe a
sphere, which can be determined by completing the square for the x, y and z variables.
Example 1.28. Is 2x
2
+ 2y
2
+ 2z
2
− 8x + 4y − 16z + 10 = 0 the equation of a sphere?
Solution: Dividing both sides of the equation by 2 gives
x
2
+ y
2
+ z
2
− 4x + 2y − 8z + 5 = 0
(x
2
− 4x + 4) + (y
2
+ 2y + 1) + (z
2
− 8z + 16) + 5 − 4 − 1 − 16 = 0
(x − 2)
2
+ (y + 1)
2
+ (z − 4)
2
= 16
which is a sphere of radius 4 centered at (2, −1, 4).
Example 1.29. Find the points(s) of intersection (if any) of the sphere from Example
1.28 and the line x = 3 + t, y = 1 + 2t, z = 3 − t.
Solution: Put the equations of the line into the equation of the sphere, which was
(x − 2)
2
+ (y + 1)
2
+ (z − 4)
2
= 16, and solve for t:
(3 + t − 2)
2
+ (1 + 2t + 1)
2
+ (3 − t − 4)
2
= 16
(t + 1)
2
+ (2t + 2)
2
+ (−t − 1)
2
= 16
6t
2
+ 12t − 10 = 0
The quadratic formula gives the solutions t = −1 ±
4

6
. Putting those two values into
the equations of the line gives the following two points of intersection:
¸
2 +
4

6
, −1 +
8

6
, 4 −
4

6

and
¸
2 −
4

6
, −1 −
8

6
, 4 +
4

6

42 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
If two spheres intersect, they do so either at a single point or in a circle.
Example 1.30. Find the intersection (if any) of the spheres x
2
+ y
2
+ z
2
= 25 and x
2
+
y
2
+ (z − 2)
2
= 16.
Solution: For any point (x, y, z) on both spheres, we see that
x
2
+ y
2
+ z
2
= 25 ⇒ x
2
+ y
2
= 25 − z
2
, and
x
2
+ y
2
+ (z − 2)
2
= 16 ⇒ x
2
+ y
2
= 16 − (z − 2)
2
, so
16 − (z − 2)
2
= 25 − z
2
⇒ 4z − 4 = 9 ⇒ z = 13/4
⇒ x
2
+ y
2
= 25 − (13/4)
2
= 231/16
∴ The intersection is the circle x
2
+ y
2
=
231
16

231
4
≈ 3.8 centered at (0, 0,
13
4
).
The cylinders that we will consider are right circular cylinders. These are cylinders
obtained by moving a line L along a circle C in
3
in a way so that L is always per-
pendicular to the plane containing C. We will only consider the cases where the plane
containing C is parallel to one of the three coordinate planes (see Figure 1.6.3).
y
z
x
0
r
(a) x
2
+ y
2
= r
2
, any z
y
z
x
0
r
(b) x
2
+ z
2
= r
2
, any y
y
z
x
0
r
(c) y
2
+ z
2
= r
2
, any x
Figure 1.6.3 Cylinders in
3
For example, the equation of a cylinder whose base circle C lies in the xy-plane and
is centered at (a, b, 0) and has radius r is
(x − a)
2
+ (y − b)
2
= r
2
, (1.32)
where the value of the z coordinate is unrestricted. Similar equations can be written
when the base circle lies in one of the other coordinate planes. A plane intersects a
right circular cylinder in a circle, ellipse, or one or two lines, depending on whether
that plane is parallel, oblique
9
, or perpendicular, respectively, to the plane containing
C. The intersection of a surface with a plane is called the trace of the surface.
9
i.e. at an angle strictly between 0

and 90

.
1.6 Surfaces 43
The equations of spheres and cylinders are examples of second-degree equations in

3
, i.e. equations of the form
Ax
2
+ By
2
+ Cz
2
+ Dxy + Exz + Fyz + Gx + Hy + Iz + J = 0 (1.33)
for some constants A, B, . . . , J. If the above equation is not that of a sphere, cylinder,
plane, line or point, then the resulting surface is called a quadric surface.
y
z
x
0
a
b
c
Figure 1.6.4 Ellipsoid
One type of quadric surface is the ellipsoid,
given by an equation of the form:
x
2
a
2
+
y
2
b
2
+
z
2
c
2
= 1 (1.34)
In the case where a = b = c, this is just a sphere.
In general, an ellipsoid is egg-shaped (think of
an ellipse rotated around its major axis). Its
traces in the coordinate planes are ellipses.
Two other types of quadric surfaces are the hyperboloid of one sheet, given by
an equation of the form:
x
2
a
2
+
y
2
b
2

z
2
c
2
= 1 (1.35)
and the hyperboloid of two sheets, whose equation has the form:
x
2
a
2

y
2
b
2

z
2
c
2
= 1 (1.36)
y
z
x
0
Figure 1.6.5 Hyperboloid of one sheet
y
z
x
0
Figure 1.6.6 Hyperboloid of two sheets
44 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
For the hyperboloid of one sheet, the trace in any plane parallel to the xy-plane is
an ellipse. The traces in the planes parallel to the xz- or yz-planes are hyperbolas (see
Figure 1.6.5), except for the special cases x = ±a and y = ±b; in those planes the traces
are pairs of intersecting lines (see Exercise 8).
For the hyperboloid of two sheets, the trace in any plane parallel to the xy- or xz-
plane is a hyperbola (see Figure 1.6.6). There is no trace in the yz-plane. In any plane
parallel to the yz-plane for which | x| > |a|, the trace is an ellipse.
y
z
x
0
Figure 1.6.7 Paraboloid
The elliptic paraboloid is another type of quadric sur-
face, whose equation has the form:
x
2
a
2
+
y
2
b
2
=
z
c
(1.37)
The traces in planes parallel to the xy-plane are ellipses,
though in the xy-plane itself the trace is a single point. The
traces in planes parallel to the xz- or yz-planes are parabo-
las. Figure 1.6.7 shows the case where c > 0. When c < 0 the
surface is turned downward. In the case where a = b, the
surface is called a paraboloid of revolution, which is often
used as a reflecting surface, e.g. in vehicle headlights.
10
A more complicated quadric surface is the hyperbolic paraboloid, given by:
x
2
a
2

y
2
b
2
=
z
c
(1.38)
-10
-5
0
5
10
-10
-5
0
5
10
-100
-50
0
50
100
z
x
y
z
Figure 1.6.8 Hyperbolic paraboloid
10
For a discussion of this see pp. 157-158 in HECHT.
1.6 Surfaces 45
The hyperbolic paraboloid can be tricky to draw; using graphing software on a com-
puter can make it easier. For example, Figure 1.6.8 was created using the free Gnuplot
package (see Appendix C). It shows the graph of the hyperbolic paraboloid z = y
2
− x
2
,
which is the special case where a = b = 1 and c = −1 in equation (1.38). The mesh lines
on the surface are the traces in planes parallel to the coordinate planes. So we see
that the traces in planes parallel to the xz-plane are parabolas pointing upward, while
the traces in planes parallel to the yz-plane are parabolas pointing downward. Also,
notice that the traces in planes parallel to the xy-plane are hyperbolas, though in the
xy-plane itself the trace is a pair of intersecting lines through the origin. This is true
in general when c < 0 in equation (1.38). When c > 0, the surface would be similar to
that in Figure 1.6.8, only rotated 90

around the z-axis and the nature of the traces in
planes parallel to the xz- or yz-planes would be reversed.
y
z
x
0
Figure 1.6.9 Elliptic cone
The last type of quadric surface that we will consider is
the elliptic cone, which has an equation of the form:
x
2
a
2
+
y
2
b
2

z
2
c
2
= 0 (1.39)
The traces in planes parallel to the xy-plane are ellipses,
except in the xy-plane itself where the trace is a single
point. The traces in planes parallel to the xz- or yz-planes
are hyperbolas, except in the xz- and yz-planes themselves
where the traces are pairs of intersecting lines.
Notice that every point on the elliptic cone is on a line
which lies entirely on the surface; in Figure 1.6.9 these
lines all go through the origin. This makes the elliptic
cone an example of a ruled surface. The cylinder is also a ruled surface.
What may not be as obvious is that both the hyperboloid of one sheet and the hy-
perbolic paraboloid are ruled surfaces. In fact, on both surfaces there are two lines
through each point on the surface (see Exercises 11-12). Such surfaces are called
doubly ruled surfaces, and the pairs of lines are called a regulus.
It is clear that for each of the six types of quadric surfaces that we discussed, the
surface can be translated away from the origin (e.g. by replacing x
2
by (x − x
0
)
2
in
its equation). It can be proved
11
that every quadric surface can be translated and/or
rotated so that its equation matches one of the six types that we described. For ex-
ample, z = 2xy is a case of equation (1.33) with “mixed” variables, e.g. with D 0 so
that we get an xy term. This equation does not match any of the types we considered.
However, by rotating the x- and y-axes by 45

in the xy-plane by means of the coor-
dinate transformation x = (x

− y

)/

2, y = (x

+ y

)/

2, z = z

, then z = 2xy becomes
the hyperbolic paraboloid z

= (x

)
2
− (y

)
2
in the (x

, y

, z

) coordinate system. That is,
z = 2xy is a hyperbolic paraboloid as in equation (1.38), but rotated 45

in the xy-plane.
11
See Ch. 7 in POGORELOV.
46 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

¨
Exercises
A
For Exercises 1-4, determine if the given equation describes a sphere. If so, find its
1. x
2
+ y
2
+ z
2
− 4x − 6y − 10z + 37 = 0 2. x
2
+ y
2
+ z
2
+ 2x − 2y − 8z + 19 = 0
3. 2x
2
+ 2y
2
+ 2z
2
+ 4x + 4y + 4z − 44 = 0 4. x
2
+ y
2
− z
2
+ 12x + 2y − 4z + 32 = 0
5. Find the point(s) of intersection of the sphere (x − 3)
2
+ (y + 1)
2
+ (z − 3)
2
= 9 and the
line x = −1 + 2t, y = −2 − 3t, z = 3 + t.
B
6. Find the intersection of the spheres x
2
+y
2
+z
2
= 9 and (x −4)
2
+(y +2)
2
+(z −4)
2
= 9.
7. Find the intersection of the sphere x
2
+ y
2
+ z
2
= 9 and the cylinder x
2
+ y
2
= 4.
8. Find the trace of the hyperboloid of one sheet
x
2
a
2
+
y
2
b
2

z
2
c
2
= 1 in the plane x = a, and
the trace in the plane y = b.
9. Find the trace of the hyperbolic paraboloid
x
2
a
2

y
2
b
2
=
z
c
in the xy-plane.
C
10. It can be shown that any four noncoplanar points (i.e. points that do not lie in
the same plane) determine a sphere.
12
Find the equation of the sphere that passes
through the points (0, 0, 0), (0, 0, 2), (1, −4, 3) and (0, −1, 3). (Hint: Equation (1.31))
11. Show that the hyperboloid of one sheet is a doubly ruled surface, i.e. each point
on the surface is on two lines lying entirely on the surface. (Hint: Write equation
(1.35) as
x
2
a
2

z
2
c
2
= 1−
y
2
b
2
, factor each side. Recall that two planes intersect in a line.)
12. Show that the hyperbolic paraboloid is a doubly ruled surface. (Hint: Exercise 11)
y
z
x
0
(0, 0, 2)
(x, y, 0)
(a, b, c)
1
S
Figure 1.6.10
13. Let S be the sphere with radius 1 centered at (0, 0, 1),
and let S

be S without the “north pole” point (0, 0, 2).
Let (a, b, c) be an arbitrary point on S

. Then the line
passing through (0, 0, 2) and (a, b, c) intersects the xy-
plane at some point (x, y, 0), as in Figure 1.6.10. Find
this point (x, y, 0) in terms of a, b and c.
(Note: Every point in the xy-plane can be matched
with a point on S

, and vice versa, in this manner.
This method is called stereographic projection, which
essentially identifies all of
2
with a “punctured” sphere.)
12
See WELCHONS and KRICKENBERGER, p. 160, for a proof.
1.7 Curvilinear Coordinates 47
1.7 Curvilinear Coordinates
x
y
z
0
(x, y, z)
x
y
z
Figure 1.7.1
The Cartesian coordinates of a point (x, y, z) are determined
by following a family of straight paths from the origin: first
along the x-axis, then parallel to the y-axis, then parallel to
the z-axis, as in Figure 1.7.1. In curvilinear coordinate sys-
tems, these paths can be curved. The two types of curvilinear
coordinates which we will consider are cylindrical and spher-
ical coordinates. Instead of referencing a point in terms of
sides of a rectangular parallelepiped, as with Cartesian co-
ordinates, we will think of the point as lying on a cylinder or sphere. Cylindrical
coordinates are often used when there is symmetry around the z-axis, while spherical
coordinates are useful when there is symmetry about the origin.
Let P = (x, y, z) be a point in Cartesian coordinates in
3
, and let P
0
= (x, y, 0) be the
projection of P upon the xy-plane. Treating (x, y) as a point in
2
, let (r, θ) be its polar
coordinates (see Figure 1.7.2). Let ρ be the length of the line segment from the origin
to P, and let φ be the angle between that line segment and the positive z-axis (see
Figure 1.7.3). φ is called the zenith angle. Then the cylindrical coordinates (r, θ, z)
and the spherical coordinates (ρ, θ, φ) of P(x, y, z) are defined as follows:
x
y
z
0
P(x, y, z)
P
0
(x, y, 0)
θ x
y
z
r
Figure 1.7.2
Cylindrical coordinates
Cylindrical coordinates (r, θ, z):
x = r cos θ r =

x
2
+ y
2
y = r sin θ θ = tan
−1

y
x

z = z z = z
where 0 ≤ θ ≤ π if y ≥ 0 and π < θ < 2π if y < 0
x
y
z
0
P(x, y, z)
P
0
(x, y, 0)
θ x
y
z
ρ
φ
Figure 1.7.3
Spherical coordinates
Spherical coordinates (ρ, θ, φ):
x = ρ sin φ cos θ ρ =

x
2
+ y
2
+ z
2
y = ρ sin φ sin θ θ = tan
−1

y
x

z = ρ cos φ φ = cos
−1

z

x
2
+y
2
+z
2

where 0 ≤ θ ≤ π if y ≥ 0 and π < θ < 2π if y < 0
Both θ and φ are measured in radians. Note that r ≥ 0, 0 ≤ θ < 2π, ρ ≥ 0 and 0 ≤ φ ≤ π.
Also, θ is undefined when (x, y) = (0, 0), and φ is undefined when (x, y, z) = (0, 0, 0).
48 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Example 1.31. Convert the point (−2, −2, 1) from Cartesian coordinates to (a) cylin-
drical and (b) spherical coordinates.
Solution: (a) r =

(−2)
2
+ (−2)
2
= 2

2, θ = tan
−1

−2
−2

= tan
−1
(1) =

4
, since y = −2 < 0.
∴ (r, θ, z) =

2

2,

4
, 1

(b) ρ =

(−2)
2
+ (−2)
2
+ 1
2
=

9 = 3, φ = cos
−1

1
3

∴ (ρ, θ, φ) =

3,

4
, 1.23

For cylindrical coordinates (r, θ, z), and constants r
0
, θ
0
and z
0
, we see from Figure
1.7.4 that the surface r = r
0
is a cylinder of radius r
0
centered along the z-axis, the
surface θ = θ
0
is a half-plane emanating from the z-axis, and the surface z = z
0
is a
plane parallel to the xy-plane.
y
z
x
0
r
0
(a) r = r
0
y
z
x
0
θ
0
(b) θ = θ
0
y
z
x
0
z
0
(c) z = z
0
Figure 1.7.4 Cylindrical coordinate surfaces
For spherical coordinates (ρ, θ, φ), and constants ρ
0
, θ
0
and φ
0
, we see from Figure
1.7.5 that the surface ρ = ρ
0
is a sphere of radius ρ
0
centered at the origin, the surface
θ = θ
0
is a half-plane emanating from the z-axis, and the surface φ = φ
0
is a circular
cone whose vertex is at the origin.
y
z
x
0
ρ
0
(a) ρ = ρ
0
y
z
x
0
θ
0
(b) θ = θ
0
y
z
x
0
φ
0
(c) φ = φ
0
Figure 1.7.5 Spherical coordinate surfaces
Figures 1.7.4(a) and 1.7.5(a) show how these coordinate systems got their names.
1.7 Curvilinear Coordinates 49
Sometimes the equation of a surface in Cartesian coordinates can be transformed
into a simpler equation in some other coordinate system, as in the following example.
Example 1.32. Write the equation of the cylinder x
2
+y
2
= 4 in cylindrical coordinates.
Solution: Since r =

x
2
+ y
2
, then the equation in cylindrical coordinates is r = 2.
Using spherical coordinates to write the equation of a sphere does not necessarily
make the equation simpler, if the sphere is not centered at the origin.
Example 1.33. Write the equation (x − 2)
2
+ (y − 1)
2
+ z
2
= 9 in spherical coordinates.
Solution: Multiplying the equation out gives
x
2
+ y
2
+ z
2
− 4x − 2y + 5 = 9 , so we get
ρ
2
− 4ρ sin φ cos θ − 2ρ sin φ sin θ − 4 = 0 , or
ρ
2
− 2 sin φ (2 cos θ − sin θ ) ρ − 4 = 0
after combining terms. Note that this actually makes it more difficult to figure out
what the surface is, as opposed to the Cartesian equation where you could immedi-
ately identify the surface as a sphere of radius 3 centered at (2, 1, 0).
Example 1.34. Describe the surface given by θ = z in cylindrical coordinates.
Solution: This surface is called a helicoid. As the (vertical) z coordinate increases,
so does the angle θ, while the radius r is unrestricted. So this sweeps out a (ruled!)
surface shaped like a spiral staircase, where the spiral has an inﬁnite radius. Figure
1.7.6 shows a section of this surface restricted to 0 ≤ z ≤ 4π and 0 ≤ r ≤ 2.
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0
2
4
6
8
10
12
14
z
x
y
z
Figure 1.7.6 Helicoid θ = z
50 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

¨
Exercises
A
For Exercises 1-4, find the (a) cylindrical and (b) spherical coordinates of the point
whose Cartesian coordinates are given.
1. (2, 2

3, −1) 2. (−5, 5, 6) 3. (

21, −

7, 0) 4. (0,

2, 2)
For Exercises 5-7, write the given equation in (a) cylindrical and (b) spherical coordi-
nates.
5. x
2
+ y
2
+ z
2
= 25 6. x
2
+ y
2
= 2y 7. x
2
+ y
2
+ 9z
2
= 36
B
8. Describe the intersection of the surfaces whose equations in spherical coordinates
are θ =
π
2
and φ =
π
4
.
9. Show that for a 0, the equation ρ = 2a sin φ cos θ in spherical coordinates describes
a sphere centered at (a, 0, 0) with radius |a|.
C
10. Let P = (a, θ, φ) be a point in spherical coordinates, with a > 0 and 0 < φ < π. Then
P lies on the sphere ρ = a. Since 0 < φ < π, the line segment from the origin to P
can be extended to intersect the cylinder given by r = a (in cylindrical coordinates).
Find the cylindrical coordinates of that point of intersection.
11. Let P
1
and P
2
be points whose spherical coordinates are (ρ
1
, θ
1
, φ
1
) and (ρ
2
, θ
2
, φ
2
),
respectively. Let v
1
be the vector from the origin to P
1
, and let v
2
be the vector from
the origin to P
2
. For the angle γ between v
1
and v
2
, show that
cos γ = cos φ
1
cos φ
2
+ sin φ
1
sin φ
2
cos( θ
2
− θ
1
).
This formula is used in electrodynamics to prove the addition theorem for spherical
harmonics, which provides a general expression for the electrostatic potential at a
point due to a unit charge. See pp. 100-102 in JACKSON.
12. Show that the distance d between the points P
1
and P
2
with cylindrical coordinates
(r
1
, θ
1
, z
1
) and (r
2
, θ
2
, z
2
), respectively, is
d =

r
2
1
+ r
2
2
− 2r
1
r
2
cos( θ
2
− θ
1
) + (z
2
− z
1
)
2
.
13. Show that the distance d between the points P
1
and P
2
with spherical coordinates

1
, θ
1
, φ
1
) and (ρ
2
, θ
2
, φ
2
), respectively, is
d =

ρ
2
1
+ ρ
2
2
− 2ρ
1
ρ
2
[sin φ
1
sin φ
2
cos( θ
2
− θ
1
) + cos φ
1
cos φ
2
] .
1.8 Vector-Valued Functions 51
1.8 Vector-Valued Functions
Now that we are familiar with vectors and their operations, we can begin discussing
functions whose values are vectors.
Definition 1.10. A vector-valued function of a real variable is a rule that asso-
ciates a vector f(t) with a real number t, where t is in some subset D of
1
(called the
domain of f). We write f : D →
3
to denote that f is a mapping of D into
3
.
For example, f(t) = ti + t
2
j + t
3
k is a vector-valued function in
3
, defined for all real
numbers t. We would write f : →
3
. At t = 1 the value of the function is the vector
i + j + k, which in Cartesian coordinates has the terminal point (1, 1, 1).
A vector-valued function of a real variable can be written in component form as
f(t) = f
1
(t)i + f
2
(t)j + f
3
(t)k
or in the form
f(t) = ( f
1
(t), f
2
(t), f
3
(t))
for some real-valued functions f
1
(t), f
2
(t), f
3
(t), called the component functions of f. The
first form is often used when emphasizing that f(t) is a vector, and the second form is
useful when considering just the terminal points of the vectors. By identifying vectors
with their terminal points, a curve in space can be written as a vector-valued function.
y
z
x
0
f(0)
f(2π)
Figure 1.8.1
Example 1.35. Define f : →
3
by f(t) = (cos t, sin t, t).
This is the equation of a helix (see Figure 1.8.1). As the value
of t increases, the terminal points of f(t) trace out a curve spi-
raling upward. For each t, the x- and y-coordinates of f(t) are
x = cos t and y = sin t, so
x
2
+ y
2
= cos
2
t + sin
2
t = 1.
Thus, the curve lies on the surface of the right circular cylin-
der x
2
+ y
2
= 1.
It may help to think of vector-valued functions of a real variable in
3
as a general-
ization of the parametric functions in
2
which you learned about in single-variable
calculus. Much of the theory of real-valued functions of a single real variable can be
applied to vector-valued functions of a real variable. Since each of the three compo-
nent functions are real-valued, it will sometimes be the case that results from single-
variable calculus can simply be applied to each of the component functions to yield
a similar result for the vector-valued function. However, there are times when such
generalizations do not hold (see Exercise 13). The concept of a limit, though, can be
extended naturally to vector-valued functions, as in the following definition.
52 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Definition 1.11. Let f(t) be a vector-valued function, let a be a real number and let c
be a vector. Then we say that the limit of f(t) as t approaches a equals c, written as
lim
t→a
f(t) = c, if lim
t→a
f(t) − c = 0. If f(t) = ( f
1
(t), f
2
(t), f
3
(t)), then
lim
t→a
f(t) =

lim
t→a
f
1
(t), lim
t→a
f
2
(t), lim
t→a
f
3
(t)

provided that all three limits on the right side exist.
The above definition shows that continuity and the derivative of vector-valued func-
tions can also be defined in terms of its component functions.
Definition 1.12. Let f(t) = ( f
1
(t), f
2
(t), f
3
(t)) be a vector-valued function, and let a be a
real number in its domain. Then f(t) is continuous at a if lim
t→a
f(t) = f(a). Equivalently,
f(t) is continuous at a if and only if f
1
(t), f
2
(t), and f
3
(t) are continuous at a.
The derivative of f(t) at a, denoted by f

(a) or
df
dt
(a), is the limit
f

(a) = lim
h→0
f(a + h) − f(a)
h
if that limit exists. Equivalently, f

(a) = ( f
1

(a), f
2

(a), f
3

(a)), if the component deriva-
tives exist. We say that f(t) is differentiable at a if f

(a) exists.
Recall that the derivative of a real-valued function of a single variable is a real
number, representing the slope of the tangent line to the graph of the function at a
point. Similarly, the derivative of a vector-valued function is a tangent vector to the
curve in space which the function represents, and it lies on the tangent line to the
curve (see Figure 1.8.2).
y
z
x
0
L
f(t)
f

(a)
f(a)
f(a + h)
f
(
a
+
h
)

f
(
a
)
Figure 1.8.2 Tangent vector f

(a) and tangent line L = f(a) + sf

(a)
Example 1.36. Let f(t) = (cos t, sin t, t). Then f

(t) = (−sin t, cos t, 1) for all t. The tangent
line L to the curve at f(2π) = (1, 0, 2π) is L = f(2π) + s f

(2π) = (1, 0, 2π) + s(0, 1, 1), or in
parametric form: x = 1, y = s, z = 2π + s for −∞ < s < ∞.
1.8 Vector-Valued Functions 53
A scalar function is a real-valued function. Note that if u(t) is a scalar function
and f(t) is a vector-valued function, then their product, defined by (u f)(t) = u(t) f(t) for
all t, is a vector-valued function (since the product of a scalar with a vector is a vector).
The basic properties of derivatives of vector-valued functions are summarized in the
following theorem.
Theorem 1.20. Let f(t) and g(t) be differentiable vector-valued functions, let u(t)
be a differentiable scalar function, let k be a scalar, and let c be a constant vector. Then
(a)
d
dt
(c) = 0
(b)
d
dt
(kf) = k
df
dt
(c)
d
dt
(f + g) =
df
dt
+
dg
dt
(d)
d
dt
(f − g) =
df
dt

dg
dt
(e)
d
dt
(u f) =
du
dt
f + u
df
dt
(f)
d
dt
(f ··· g) =
df
dt
··· g + f ···
dg
dt
(g)
d
dt
(f ××× g) =
df
dt
××× g + f ×××
dg
dt
Proof: The proofs of parts (a)-(e) follow easily by differentiating the component func-
tions and using the rules for derivatives from single-variable calculus. We will prove
part (f), and leave the proof of part (g) as an exercise for the reader.
(f) Write f(t) = ( f
1
(t), f
2
(t), f
3
(t)) and g(t) = (g
1
(t), g
2
(t), g
3
(t)), where the component func-
tions f
1
(t), f
2
(t), f
3
(t), g
1
(t), g
2
(t), g
3
(t) are all differentiable real-valued functions. Then
d
dt
(f(t) ··· g(t)) =
d
dt
( f
1
(t) g
1
(t) + f
2
(t) g
2
(t) + f
3
(t) g
3
(t))
=
d
dt
( f
1
(t) g
1
(t)) +
d
dt
( f
2
(t) g
2
(t)) +
d
dt
( f
3
(t) g
3
(t))
=
d f
1
dt
(t) g
1
(t) + f
1
(t)
dg
1
dt
(t) +
d f
2
dt
(t) g
2
(t) + f
2
(t)
dg
2
dt
(t) +
d f
3
dt
(t) g
3
(t) + f
3
(t)
dg
3
dt
(t)
=

d f
1
dt
(t),
d f
2
dt
(t),
d f
3
dt
(t)

··· (g
1
(t), g
2
(t), g
3
(t))
+ ( f
1
(t), f
2
(t), f
3
(t)) ···

dg
1
dt
(t),
dg
2
dt
(t),
dg
3
dt
(t)

=
df
dt
(t) ··· g(t) + f(t) ···
dg
dt
(t) for all t. QED
54 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Example 1.37. Suppose f(t) is differentiable. Find the derivative of f(t).
Solution: Since f(t) is a real-valued function of t, then by the Chain Rule for real-
valued functions, we know that
d
dt
f(t)
2
= 2f(t)
d
dt
f(t).
But f(t)
2
= f(t) ··· f(t), so
d
dt
f(t)
2
=
d
dt
(f(t) ··· f(t)). Hence, we have
2f(t)
d
dt
f(t) =
d
dt
(f(t) ··· f(t)) = f

(t) ··· f(t) + f(t) ··· f

(t) by Theorem 1.20(f), so
= 2f

(t) ··· f(t) , so if f(t) 0 then
d
dt
f(t) =
f

(t) ··· f(t)
f(t)
.
We know that f(t) is constant if and only if
d
dt
f(t) = 0 for all t. Also, f(t) ⊥ f

(t) if
and only if f

(t) ··· f(t) = 0. Thus, the above example shows this important fact:
If f(t) 0, then f(t) is constant if and only if f(t) ⊥ f

(t) for all t.
This means that if a curve lies completely on a sphere (or circle) centered at the origin,
then the tangent vector f

(t) is always perpendicular to the position vector f(t).
Example 1.38. The spherical spiral f(t) =

cos t

1 + a
2
t
2
,
sin t

1 + a
2
t
2
,
−at

1 + a
2
t
2

, for a 0.
Figure 1.8.3 shows the graph of the curve when a = 0.2. In the exercises, the reader
will be asked to show that this curve lies on the sphere x
2
+ y
2
+ z
2
= 1 and to verify
directly that f

(t) ··· f(t) = 0 for all t.
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
z
x
y
z
Figure 1.8.3 Spherical spiral with a = 0.2
1.8 Vector-Valued Functions 55
Just as in single-variable calculus, higher-order derivatives of vector-valued func-
tions are obtained by repeatedly differentiating the (first) derivative of the function:
f
′′
(t) =
d
dt
f

(t) , f
′′′
(t) =
d
dt
f
′′
(t) , . . . ,
d
n
f
dt
n
=
d
dt

d
n−1
f
dt
n−1

(for n = 2, 3, 4, . . .)
We can use vector-valued functions to represent physical quantities, such as veloc-
ity, acceleration, force, momentum, etc. For example, let the real variable t represent
time elapsed from some initial time (t = 0), and suppose that an object of constant
mass m is subjected to some force so that it moves in space, with its position (x, y, z) at
time t a function of t. That is, x = x(t), y = y(t), z = z(t) for some real-valued functions
x(t), y(t), z(t). Call r(t) = (x(t), y(t), z(t)) the position vector of the object. We can define
various physical quantities associated with the object as follows:
13
position: r(t) = (x(t), y(t), z(t))
velocity: v(t) = ˙ r(t) = r

(t) =
dr
dt
= (x

(t), y

(t), z

(t))
acceleration: a(t) = ˙ v(t) = v

(t) =
dv
dt
= ¨ r(t) = r
′′
(t) =
d
2
r
dt
2
= (x
′′
(t), y
′′
(t), z
′′
(t))
momentum: p(t) = mv(t)
force: F(t) = ˙ p(t) = p

(t) =
dp
dt
(Newton’s Second Law of Motion)
The magnitude v(t) of the velocity vector is called the speed of the object. Note that
since the mass m is a constant, the force equation becomes the familiar F(t) = ma(t).
Example 1.39. Let r(t) = (5 cos t, 3 sin t, 4 sin t) be the position vector of an object at time
t ≥ 0. Find its (a) velocity and (b) acceleration vectors.
Solution: (a) v(t) = ˙ r(t) = (−5 sin t, 3 cos t, 4 cos t)
(b) a(t) = ˙ v(t) = (−5 cos t, −3 sin t, −4 sin t)
Note that r(t) =

25 cos
2
t + 25 sin
2
t = 5 for all t, so by Example 1.37 we know that
r(t) ··· ˙ r(t) = 0 for all t (which we can verify from part (a)). In fact, v(t) = 5 for all t
also. And not only does r(t) lie on the sphere of radius 5 centered at the origin, but
perhaps not so obvious is that it lies completely within a circle of radius 5 centered at
the origin. Also, note that a(t) = −r(t). It turns out (see Exercise 16) that whenever an
object moves in a circle with constant speed, the acceleration vector will point in the
opposite direction of the position vector (i.e. towards the center of the circle).
13
We will often use the older dot notation for derivatives when physics is involved.
56 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Recall from Section 1.5 that if r
1
, r
2
are position vectors to distinct points then
r
1
+t(r
2
−r
1
) represents a line through those two points as t varies over all real numbers.
That vector sum can be written as (1 − t)r
1
+ tr
2
. So the function l(t) = (1 − t)r
1
+ tr
2
is
a line through the terminal points of r
1
and r
2
, and when t is restricted to the interval
[0, 1] it is the line segment between the points, with l(0) = r
1
and l(1) = r
2
.
In general, a function of the form f(t) = (a
1
t + b
1
, a
2
t + b
2
, a
3
t + b
3
) represents a line in

3
. A function of the form f(t) = (a
1
t
2
+ b
1
t + c
1
, a
2
t
2
+ b
2
t + c
2
, a
3
t
2
+ b
3
t + c
3
) represents a
(possibly degenerate) parabola in
3
.
Example 1.40. Bézier curves are used in Computer Aided Design (CAD) to approx-
imate the shape of a polygonal path in space (called the Bézier polygon or control
polygon). For instance, given three points (or position vectors) b
0
, b
1
, b
2
in
3
, define
b
1
0
(t) = (1 − t)b
0
+ tb
1
b
1
1
(t) = (1 − t)b
1
+ tb
2
b
2
0
(t) = (1 − t)b
1
0
(t) + tb
1
1
(t)
= (1 − t)
2
b
0
+ 2t(1 − t)b
1
+ t
2
b
2
for all real t. For t in the interval [0, 1], we see that b
1
0
(t) is the line segment between
b
0
and b
1
, and b
1
1
(t) is the line segment between b
1
and b
2
. The function b
2
0
(t) is the
Bézier curve for the points b
0
, b
1
, b
2
. Note from the last formula that the curve is a
parabola that goes through b
0
(when t = 0) and b
2
(when t = 1).
As an example, let b
0
= (0, 0, 0), b
1
= (1, 2, 3), and b
2
= (4, 5, 2). Then the explicit
formula for the Bézier curve is b
2
0
(t) = (2t +2t
2
, 4t +t
2
, 6t −4t
2
), as shown in Figure 1.8.4,
where the line segments are b
1
0
(t) and b
1
1
(t), and the curve is b
2
0
(t).
0
0.5
1
1.5
2
2.5
3
3.5
4
0
1
2
3
4
5
0
0.5
1
1.5
2
2.5
3
z
(0,0,0)
(1,2,3)
(4,5,2)
x
y
z
Figure 1.8.4 Bézier curve approximation for three points
1.8 Vector-Valued Functions 57
In general, the polygonal path determined by n ≥ 3 noncollinear points in
3
can be
used to define the Bézier curve recursively by a process called repeated linear interpo-
lation. This curve will be a vector-valued function whose components are polynomials
of degree n−1, and its formula is given by de Casteljau’s algorithm.
14
In the exercises,
the reader will be given the algorithm for the case of n = 4 points and asked to write
the explicit formula for the Bézier curve for the four points shown in Figure 1.8.5.
0
0.5
1
1.5
2
2.5
3
3.5
4
0
1
2
3
4
5
0
0.5
1
1.5
2
z
(0,0,0)
(0,1,1)
(2,3,0)
(4,5,2)
x
y
z
Figure 1.8.5 Bézier curve approximation for four points

¨
Exercises
A
For Exercises 1-4, calculate f

(t) and find the tangent line at f(0).
1. f(t) = (t + 1, t
2
+ 1, t
3
+ 1) 2. f(t) = (e
t
+ 1, e
2t
+ 1, e
t
2
+ 1)
3. f(t) = (cos 2t, sin 2t, t) 4. f(t) = (sin 2t, 2 sin
2
t, 2 cos t)
For Exercises 5-6, find the velocity v(t) and acceleration a(t) of an object with the given
position vector r(t).
5. r(t) = (t, t − sin t, 1 − cos t) 6. r(t) = (3 cos t, 2 sin t, 1)
B
7. Let f(t) =

cos t

1 + a
2
t
2
,
sin t

1 + a
2
t
2
,
−at

1 + a
2
t
2

, with a 0.
(a) Show that f(t) = 1 for all t.
(b) Show directly that f

(t) ··· f(t) = 0 for all t.
8. If f

(t) = 0 for all t in some interval (a, b), show that f(t) is a constant vector in (a, b).
14
See pp. 27-30 in FARIN.
58 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
9. For a constant vector c 0, the function f(t) = tc represents a line parallel to c.
(a) What kind of curve does g(t) = t
3
c represent? Explain.
(b) What kind of curve does h(t) = e
t
c represent? Explain.
(c) Compare f

(0) and g

(0). Given your answer to part (a), how do you explain the
difference in the two derivatives?
10. Show that
d
dt

f ×××
df
dt

= f ×××
d
2
f
dt
2
.
11. Let a particle of (constant) mass m have position vector r(t), velocity v(t), accelera-
tion a(t) and momentum p(t) at time t. The angular momentum L(t) of the particle
with respect to the origin at time t is defined as L(t) = r(t) ××× p(t). If F(t) is the force
acting on the particle at time t, then define the torque N(t) acting on the particle
with respect to the origin as N(t) = r(t) ××× F(t). Show that L

(t) = N(t).
12. Show that
d
dt
(f ··· (g ××× h)) =
df
dt
··· (g ××× h) + f ···

dg
dt
××× h

+ f ···

g ×××
dh
dt

.
13. The Mean Value Theorem does not hold for vector-valued functions: Show that
for f(t) = (cos t, sin t, t), there is no t in the interval (0, 2π) such that
f

(t) =
f(2π) − f(0)
2π − 0
.
C
14. The Bézier curve b
3
0
(t) for four noncollinear points b
0
, b
1
, b
2
, b
3
in
3
is defined by
the following algorithm (going from the left column to the right):
b
1
0
(t) = (1 − t)b
0
+ tb
1
b
2
0
(t) = (1 − t)b
1
0
(t) + tb
1
1
(t) b
3
0
(t) = (1 − t)b
2
0
(t) + tb
2
1
(t)
b
1
1
(t) = (1 − t)b
1
+ tb
2
b
2
1
(t) = (1 − t)b
1
1
(t) + tb
1
2
(t)
b
1
2
(t) = (1 − t)b
2
+ tb
3
(a) Show that b
3
0
(t) = (1 − t)
3
b
0
+ 3t(1 − t)
2
b
1
+ 3t
2
(1 − t)b
2
+ t
3
b
3
.
(b) Write the explicit formula (as in Example 1.40) for the Bézier curve for the
points b
0
= (0, 0, 0), b
1
= (0, 1, 1), b
2
= (2, 3, 0), b
3
= (4, 5, 2).
15. Let r(t) be the position vector for a particle moving in
3
. Show that
d
dt
(r ××× (v ××× r)) = r
2
a + (r ··· v)v − (v
2
+ r ··· a)r.
16. Let r(t) be the position vector in
3
for a particle that moves with constant speed
c > 0 in a circle of radius a > 0 in the xy-plane. Show that a(t) points in the opposite
direction as r(t) for all t. (Hint: Use Example 1.37 to show that r(t) ⊥ v(t) and
a(t) ⊥ v(t), and hence a(t) r(t).)
17. Prove Theorem 1.20(g).
1.9 Arc Length 59
1.9 Arc Length
Let r(t) = (x(t), y(t), z(t)) be the position vector of an object moving in
3
. Since v(t) is
the speed of the object at time t, it seems natural to define the distance s traveled by
the object from time t = a to t = b as the definite integral
s =

b
a
v(t) dt =

b
a

x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt , (1.40)
which is analogous to the case from single-variable calculus for parametric functions
in
2
. This is indeed how we will define the distance traveled and, in general, the arc
length of a curve in
3
.
Definition 1.13. Let f(t) = (x(t), y(t), z(t)) be a curve in
3
whose domain includes the
interval [a, b]. Suppose that in the interval (a, b) the first derivative of each component
function x(t), y(t) and z(t) exists and is continuous, and that no section of the curve is
repeated. Then the arc length L of the curve from t = a to t = b is
L =

b
a
f

(t) dt =

b
a

x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt (1.41)
A real-valued function whose first derivative is continuous is called continuously
differentiable (or a C
1
function), and a function whose derivatives of all orders are
continuous is called smooth (or a C

function). All the functions we will consider will
be smooth. A smooth curve f(t) is one whose derivative f

(t) is never the zero vector
and whose component functions are all smooth.
Note that we did not prove that the formula in the above definition actually gives the
length of a section of a curve. A rigorous proof requires dealing with some subtleties,
normally glossed over in calculus texts, which are beyond the scope of this book.
15
Example 1.41. Find the length L of the helix f(t) = (cos t, sin t, t) from t = 0 to t = 2π.
Solution: By formula (1.41), we have
L =

0

(−sin t)
2
+ (cos t)
2
+ 1
2
dt =

0

sin
2
t + cos
2
t + 1 dt =

0

2 dt
=

2(2π − 0) = 2

Similar to the case in
2
, if there are values of t in the interval [a, b] where the
derivative of a component function is not continuous then it is often possible to parti-
tion [a, b] into subintervals where all the component functions are continuously differ-
entiable (except at the endpoints, which can be ignored). The sum of the arc lengths
over the subintervals will be the arc length over [a, b].
15
In particular, Duhamel’s principle is needed. See the proof in TAYLOR and MANN, § 14.2 and § 18.2.
60 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Notice that the curve traced out by the function f(t) = (cos t, sin t, t) from Example
1.41 is also traced out by the function g(t) = (cos 2t, sin 2t, 2t). For example, over the
interval [0, π], g(t) traces out the same section of the curve as f(t) does over the interval
[0, 2π]. Intuitively, this says that g(t) traces the curve twice as fast as f(t). This makes
sense since, viewing the functions as position vectors and their derivatives as velocity
vectors, the speeds of f(t) and g(t) are f

(t) =

2 and g

(t) = 2

2, respectively. We
say that g(t) and f(t) are different parametrizations of the same curve.
Definition 1.14. Let C be a smooth curve in
3
represented by a function f(t) defined
on an interval [a, b], and let α : [c, d] → [a, b] be a smooth one-to-one mapping of an
interval [c, d] onto [a, b]. Then the function g : [c, d] →
3
defined by g(s) = f(α(s)) is a
parametrization of C with parameter s. If α is strictly increasing on [c, d] then we
say that g(s) is equivalent to f(t).
s t f(t)
[c, d] [a, b]

3
α f
g(s) = f(α(s)) = f(t)
Note that the differentiability of g(s) follows from a version of the Chain Rule for
vector-valued functions (the proof is left as an exercise):
Theorem 1.21. Chain Rule: If f(t) is a differentiable vector-valued function of t, and
t = α(s) is a differentiable scalar function of s, then f(s) = f(α(s)) is a differentiable
vector-valued function of s, and
df
ds
=
df
dt
dt
ds
(1.42)
for any s where the composite function f(α(s)) is defined.
Example 1.42. The following are all equivalent parametrizations of the same curve:
f(t) = (cos t, sin t, t) for t in [0, 2π]
g(s) = (cos 2s, sin 2s, 2s) for s in [0, π]
h(s) = (cos 2πs, sin 2πs, 2πs) for s in [0, 1]
To see that g(s) is equivalent to f(t), define α : [0, π] → [0, 2π] by α(s) = 2s. Then α
is smooth, one-to-one, maps [0, π] onto [0, 2π], and is strictly increasing (since α

(s) =
2 > 0 for all s). Likewise, defining α : [0, 1] → [0, 2π] by α(s) = 2πs shows that h(s) is
equivalent to f(t).
1.9 Arc Length 61
A curve can have many parametrizations, with different speeds, so which one is the
best to use? In some situations the arc length parametrization can be useful. The
idea behind this is to replace the parameter t, for any given smooth parametrization
f(t) defined on [a, b], by the parameter s given by
s = s(t) =

t
a
f

(u) du. (1.43)
In terms of motion along a curve, s is the distance traveled along the curve after
time t has elapsed. So the new parameter will be distance instead of time. There
is a natural correspondence between s and t: from a starting point on the curve, the
distance traveled along the curve (in one direction) is uniquely determined by the
amount of time elapsed, and vice versa.
Since s is the arc length of the curve over the interval [a, t] for each t in [a, b], then
it is a function of t. By the Fundamental Theorem of Calculus, its derivative is
s

(t) =
ds
dt
=
d
dt

t
a
f

(u) du = f

(t) for all t in [a, b].
Since f(t) is smooth, then f

(t) > 0 for all t in [a, b]. Thus s

(t) > 0 and hence s(t) is
strictly increasing on the interval [a, b]. Recall that this means that s is a one-to-one
mapping of the interval [a, b] onto the interval [s(a), s(b)]. But we see that
s(a) =

a
a
f

(u) du = 0 and s(b) =

b
a
f

(u) du = L = arc length from t = a to t = b
s t
[0, L] [a, b]
α(s)
s(t)
Figure 1.9.1 t = α(s)
So the function s : [a, b] → [0, L] is a one-to-one, differen-
tiable mapping onto the interval [0, L]. From single-variable
calculus, we know that this means that there exists an in-
verse function α : [0, L] → [a, b] that is differentiable and the
inverse of s : [a, b] → [0, L]. That is, for each t in [a, b] there
is a unique s in [0, L] such that s = s(t) and t = α(s). And we
know that the derivative of α is
α

(s) =
1
s

(α(s))
=
1
f

(α(s))
So define the arc length parametrization f : [0, L] →
3
by
f(s) = f(α(s)) for all s in [0, L].
Then f(s) is smooth, by the Chain Rule. In fact, f(s) has unit speed:
f

(s) = f

(α(s)) α

(s) by the Chain Rule, so
= f

(α(s))
1
f

(α(s))
, so
f

(s) = 1 for all s in [0, L].
So the arc length parametrization traverses the curve at a “normal” rate.
62 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
In practice, parametrizing a curve f(t) by arc length requires you to evaluate the
integral s =

t
a
f

(u) du in some closed form (as a function of t) so that you could then
solve for t in terms of s. If that can be done, you would then substitute the expression
for t in terms of s (which we called α(s)) into the formula for f(t) to get f(s).
Example 1.43. Parametrize the helix f(t) = (cos t, sin t, t), for t in [0, 2π], by arc length.
Solution: By Example 1.41 and formula (1.43), we have
s =

t
0
f

(u) du =

t
0

2 du =

2 t for all t in [0, 2π].
So we can solve for t in terms of s: t = α(s) =
s

2
.
∴ f(s) =

cos
s

2
, sin
s

2
,
s

2

for all s in [0, 2

2π]. Note that f

(s) = 1.
Arc length plays an important role when discussing curvature and moving frame
fields, in the field of mathematics known as differential geometry.
16
The methods in-
volve using an arc length parametrization, which often leads to an integral that is ei-
ther difficult or impossible to evaluate in a simple closed form. The simple integral in
Example 1.43 is the exception, not the norm. In general, arc length parametrizations
are more useful for theoretical purposes than for practical computations.
17
Curvature
and moving frame fields can be defined without using arc length, which makes their
computation much easier, and these definitions can be shown to be equivalent to those
using arc length. We will leave this to the exercises.
The arc length for curves given in other coordinate systems can also be calculated:
Theorem 1.22. Suppose that r = r(t), θ = θ(t) and z = z(t) are the cylindrical coor-
dinates of a curve f(t), for t in [a, b]. Then the arc length L of the curve over [a, b]
is
L =

b
a

r

(t)
2
+ r(t)
2
θ

(t)
2
+ z

(t)
2
dt (1.44)
Proof: The Cartesian coordinates (x(t), y(t), z(t)) of a point on the curve are given by
x(t) = r(t) cos θ(t), y(t) = r(t) sin θ(t), z(t) = z(t)
so differentiating the above expressions for x(t) and y(t) with respect to t gives
x

(t) = r

(t) cos θ(t) − r(t)θ

(t) sin θ(t), y

(t) = r

(t) sin θ(t) + r(t)θ

(t) cos θ(t)
16
See O’NEILL for an introduction to elementary differential geometry.
17
For example, the usual parametrizations of Bézier curves, which we discussed in Section 1.8, are
polynomial functions in
3
. This makes their computation relatively simple, which, in CAD, is desirable.
But their arc length parametrizations are not only not polynomials, they are in fact usually impossible
to calculate at all.
1.9 Arc Length 63
and so
x

(t)
2
+ y

(t)
2
= (r

(t) cos θ(t) − r(t)θ

(t) sin θ(t))
2
+ (r

(t) sin θ(t) + r(t)θ

(t) cos θ(t))
2
= r

(t)
2
(cos
2
θ + sin
2
θ) + r(t)
2
θ

(t)
2
(cos
2
θ + sin
2
θ)
− 2r

(t)r(t)θ

(t) cos θ sin θ + 2r

(t)r(t)θ

(t) cos θ sin θ
= r

(t)
2
+ r(t)
2
θ

(t)
2
, and so
L =

b
a

x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt
=

b
a

r

(t)
2
+ r(t)
2
θ

(t)
2
+ z

(t)
2
dt QED
Example 1.44. Find the arc length L of the curve whose cylindrical coordinates are
r = e
t
, θ = t and z = e
t
, for t over the interval [0, 1].
Solution: Since r

(t) = e
t
, θ

(t) = 1 and z

(t) = e
t
, then
L =

1
0

r

(t)
2
+ r(t)
2
θ

(t)
2
+ z

(t)
2
dt
=

1
0

e
2t
+ e
2t
(1) + e
2t
dt
=

1
0
e
t

3 dt =

3(e − 1)

¨
Exercises
A
For Exercises 1-3, calculate the arc length of f(t) over the given interval.
1. f(t) = (3 cos 2t, 3 sin 2t, 3t) on [0, π/2]
2. f(t) = ((t
2
+ 1) cos t, (t
2
+ 1) sin t, 2

2t) on [0, 1]
3. f(t) = (2 cos 3t, 2 sin 3t, 2t
3/2
) on [0, 1]
4. Parametrize the curve from Exercise 1 by arc length.
5. Parametrize the curve from Exercise 3 by arc length.
B
6. Let f(t) be a differentiable curve such that f(t) 0 for all t. Show that
d
dt

¸
¸
¸
¸
¸
¸
f(t)
¸
¸
¸f(t)
¸
¸
¸

¸
¸
¸
¸
¸
¸
=
f(t) ××× (f

(t) ××× f(t))
f(t)
3
.
64 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
Exercises 7-9 develop the moving frame field T, N, B at a point on a curve.
7. Let f(t) be a smooth curve such that f

(t) 0 for all t. Then we can define the unit
tangent vector T by
T(t) =
f

(t)
f

(t)
.
Show that
T

(t) =
f

(t) ××× (f
′′
(t) ××× f

(t))
f

(t)
3
.
8. Continuing Exercise 7, assume that f

(t) and f
′′
(t) are not parallel. Then T

(t) 0
so we can define the unit principal normal vector N by
N(t) =
T

(t)
T

(t)
.
Show that
N(t) =
f

(t) ××× (f
′′
(t) ××× f

(t))
f

(t) f
′′
(t) ××× f

(t)
.
9. Continuing Exercise 8, the unit binormal vector B is defined by
B(t) = T(t) ××× N(t).
Show that
B(t) =
f

(t) ××× f
′′
(t)
f

(t) ××× f
′′
(t)
.
Note: The vectors T(t), N(t) and B(t) form a right-handed system of mutually per-
pendicular unit vectors (called orthonormal vectors) at each point on the curve f(t).
10. Continuing Exercise 9, the curvature κ is defined by
κ(t) =
T

(t)
f

(t)
=
f

(t) ××× (f
′′
(t) ××× f

(t))
f

(t)
4
.
Show that
κ(t) =
f

(t) ××× f
′′
(t)
f

(t)
3
and that T

(t) = f

(t) κ(t) N(t).
Note: κ(t) gives a sense of how “curved” the curve f(t) is at each point.
11. Find T, N, B and κ at each point of the helix f(t) = (cos t, sin t, t).
12. Show that the arc length L of a curve whose spherical coordinates are ρ = ρ(t),
θ = θ(t) and φ = φ(t) for t in an interval [a, b] is
L =

b
a

ρ

(t)
2
+ (ρ(t)
2
sin
2
φ(t)) θ

(t)
2
+ ρ(t)
2
φ

(t)
2
dt.
2 Functions of Several Variables
2.1 Functions of Two or Three Variables
In Section 1.8 we discussed vector-valued functions of a single real variable. We will
now examine real-valued functions of a point (or vector) in
2
or
3
. For the most part
these functions will be defined on sets of points in
2
, but there will be times when we
will use points in
3
, and there will also be times when it will be convenient to think
of the points as vectors (or terminal points of vectors).
A real-valued function f defined on a subset D of
2
is a rule that assigns to
each point (x, y) in D a real number f (x, y). The largest possible set D in
2
on which
f is defined is called the domain of f , and the range of f is the set of all real num-
bers f (x, y) as (x, y) varies over the domain D. A similar definition holds for functions
f (x, y, z) defined on points (x, y, z) in
3
.
Example 2.1. The domain of the function
f (x, y) = xy
is all of
2
, and the range of f is all of .
Example 2.2. The domain of the function
f (x, y) =
1
x − y
is all of
2
except the points (x, y) for which x = y. That is, the domain is the set
D = {(x, y) : x y}. The range of f is all real numbers except 0.
Example 2.3. The domain of the function
f (x, y) =

1 − x
2
− y
2
is the set D = {(x, y) : x
2
+ y
2
≤ 1}, since the quantity inside the square root is nonneg-
ative if and only if 1 − (x
2
+ y
2
) ≥ 0. We see that D consists of all points on and inside
the unit circle in
2
(D is sometimes called the closed unit disk). The range of f is the
interval [0, 1] in .
65
66 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
Example 2.4. The domain of the function
f (x, y, z) = e
x+y−z
is all of
3
, and the range of f is all positive real numbers.
A function f (x, y) defined in
2
is often written as z = f (x, y), as was mentioned in
Section 1.1, so that the graph of f (x, y) is the set {(x, y, z) : z = f (x, y)} in
3
. So we see
that this graph is a surface in
3
, since it satisfies an equation of the form F(x, y, z) = 0
(namely, F(x, y, z) = f (x, y) − z). The traces of this surface in the planes z = c, where
c varies over , are called the level curves of the function. Equivalently, the level
curves are the solution sets of the equations f (x, y) = c, for c in . Level curves are
often projected onto the xy-plane to give an idea of the various “elevation” levels of the
surface (as is done in topography).
Example 2.5. The graph of the function
f (x, y) =
sin

x
2
+ y
2

x
2
+ y
2
is shown below. Note that the level curves (shown both on the surface and projected
onto the xy-plane) are groups of concentric circles.
-10
-5
0
5
10
-10
-5
0
5
10
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
z
x
y
z
Figure 2.1.1 The function f (x, y) =
sin

x
2
+y
2

x
2
+y
2
You may be wondering what happens to the function in Example 2.5 at the point
(x, y) = (0, 0), since both the numerator and denominator are 0 at that point. The
function is not defined at (0, 0), but the limit of the function exists (and equals 1) as
(x, y) approaches (0, 0). We will now state explicitly what is meant by the limit of a
function of two variables.
2.1 Functions of Two or Three Variables 67
Definition 2.1. Let (a, b) be a point in
2
, and let f (x, y) be a real-valued function
defined on some set containing (a, b) (but not necessarily defined at (a, b) itself). Then
we say that the limit of f (x, y) equals L as (x, y) approaches (a, b), written as
lim
(x,y)→(a,b)
f (x, y) = L , (2.1)
if given any ǫ > 0, there exists a δ > 0 such that
| f (x, y) − L| < ǫ whenever 0 <

(x − a)
2
+ (y − b)
2
< δ.
A similar definition can be made for functions of three variables. The idea behind
the above definition is that the values of f (x, y) can get arbitrarily close to L (i.e. within
ǫ of L) if we pick (x, y) sufficiently close to (a, b) (i.e. inside a circle centered at (a, b) with
If you recall the “epsilon-delta” proofs of limits of real-valued functions of a single
variable, you may remember how awkward they can be, and how they can usually
only be done easily for simple functions. In general, the multivariable cases are at
least equally awkward to go through, so we will not bother with such proofs. Instead,
we will simply state that when the function f (x, y) is given by a single formula and is
defined at the point (a, b) (e.g. is not some indeterminate form like 0/0) then you can
just substitute (x, y) = (a, b) into the formula for f (x, y) to find the limit.
Example 2.6.
lim
(x,y)→(1,2)
xy
x
2
+ y
2
=
(1)(2)
1
2
+ 2
2
=
2
5
since f (x, y) =
xy
x
2
+y
2
is properly defined at the point (1, 2).
The major difference between limits in one variable and limits in two or more vari-
ables has to do with how a point is approached. In the single-variable case, the state-
ment “x → a” means that x gets closer to the value a from two possible directions
along the real number line (see Figure 2.1.2(a)). In two dimensions, however, (x, y) can
approach a point (a, b) along an infinite number of paths (see Figure 2.1.2(b)).
0
x a
x x
(a) x → a in
x
y
0
(a, b)
(b) (x, y) → (a, b) in
2
Figure 2.1.2 “Approaching” a point in different dimensions
68 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
Example 2.7.
lim
(x,y)→(0,0)
xy
x
2
+ y
2
does not exist
Note that we can not simply substitute (x, y) = (0, 0) into the function, since doing so
gives an indeterminate form 0/0. To show that the limit does not exist, we will show
that the function approaches different values as (x, y) approaches (0, 0) along different
paths in
2
. To see this, suppose that (x, y) → (0, 0) along the positive x-axis, so that
y = 0 along that path. Then
f (x, y) =
xy
x
2
+ y
2
=
x0
x
2
+ 0
2
= 0
along that path (since x > 0 in the denominator). But if (x, y) → (0, 0) along the straight
line y = x through the origin, for x > 0, then we see that
f (x, y) =
xy
x
2
+ y
2
=
x
2
x
2
+ x
2
=
1
2
which means that f (x, y) approaches different values as (x, y) → (0, 0) along different
paths. Hence the limit does not exist.
Limits of real-valued multivariable functions obey the same algebraic rules as in
the single-variable case, as shown in the following theorem, which we state without
proof.
Theorem 2.1. Suppose that lim
(x,y)→(a,b)
f (x, y) and lim
(x,y)→(a,b)
g(x, y) both exist, and that k is
some scalar. Then:
(a) lim
(x,y)→(a,b)
[ f (x, y) ± g(x, y)] =
,
lim
(x,y)→(a,b)
f (x, y)
,
±
,
lim
(x,y)→(a,b)
g(x, y)
,
(b) lim
(x,y)→(a,b)
k f (x, y) = k
,
lim
(x,y)→(a,b)
f (x, y)
,
(c) lim
(x,y)→(a,b)
[ f (x, y)g(x, y)] =
,
lim
(x,y)→(a,b)
f (x, y)
,,
lim
(x,y)→(a,b)
g(x, y)
,
(d) lim
(x,y)→(a,b)
f (x, y)
g(x, y)
=
lim
(x,y)→(a,b)
f (x, y)
lim
(x,y)→(a,b)
g(x, y)
if lim
(x,y)→(a,b)
g(x, y) 0
(e) If | f (x, y) − L| ≤ g(x, y) for all (x, y) and if lim
(x,y)→(a,b)
g(x, y) = 0, then lim
(x,y)→(a,b)
f (x, y) = L.
Note that in part (e), it suffices to have | f (x, y) − L| ≤ g(x, y) for all (x, y) “sufficiently
close” to (a, b) (but excluding (a, b) itself).
2.1 Functions of Two or Three Variables 69
Example 2.8. Show that
lim
(x,y)→(0,0)
y
4
x
2
+ y
2
= 0.
Since substituting (x, y) = (0, 0) into the function gives the indeterminate form 0/0,
we need an alternate method for evaluating this limit. We will use Theorem 2.1(e).
First, notice that y
4
=

y
2

4
and so 0 ≤ y
4

x
2
+ y
2

4
for all (x, y). But

x
2
+ y
2

4
=
(x
2
+ y
2
)
2
. Thus, for all (x, y) (0, 0) we have

y
4
x
2
+ y
2

(x
2
+ y
2
)
2
x
2
+ y
2
= x
2
+ y
2
→ 0 as (x, y) → (0, 0).
Therefore lim
(x,y)→(0,0)
y
4
x
2
+ y
2
= 0.
Continuity can be defined similarly as in the single-variable case.
Definition 2.2. A real-valued function f (x, y) with domain D in
2
is continuous
at the point (a, b) in D if lim
(x,y)→(a,b)
f (x, y) = f (a, b). We say that f (x, y) is a continuous
function if it is continuous at every point in its domain D.
Unless indicated otherwise, you can assume that all the functions we deal with
are continuous. In fact, we can modify the function from Example 2.8 so that it is
continuous on all of
2
.
Example 2.9. Define a function f (x, y) on all of
2
as follows:
f (x, y) =

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
0 if (x, y) = (0, 0)
y
4
x
2
+ y
2
if (x, y) (0, 0)
Then f (x, y) is well-defined for all (x, y) in
2
(i.e. there are no indeterminate forms for
any (x, y)), and we see that
lim
(x,y)→(a,b)
f (x, y) =
b
4
a
2
+ b
2
= f (a, b) for (a, b) (0, 0).
So since
lim
(x,y)→(0,0)
f (x, y) = 0 = f (0, 0) by Example 2.8,
then f (x, y) is continuous on all of
2
.
70 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES

¨
Exercises
A
For Exercises 1-6, state the domain and range of the given function.
1. f (x, y) = x
2
+ y
2
− 1 2. f (x, y) =
1
x
2
+ y
2
3. f (x, y) =

x
2
+ y
2
− 4 4. f (x, y) =
x
2
+ 1
y
5. f (x, y, z) = sin(xyz) 6. f (x, y, z) =

(x − 1)(yz − 1)
For Exercises 7-18, evaluate the given limit.
7. lim
(x,y)→(0,0)
cos(xy) 8. lim
(x,y)→(0,0)
e
xy
9. lim
(x,y)→(0,0)
x
2
− y
2
x
2
+ y
2
10. lim
(x,y)→(0,0)
xy
2
x
2
+ y
4
11. lim
(x,y)→(1,−1)
x
2
− 2xy + y
2
x − y
12. lim
(x,y)→(0,0)
xy
2
x
2
+ y
2
13. lim
(x,y)→(1,1)
x
2
− y
2
x − y
14. lim
(x,y)→(0,0)
x
2
− 2xy + y
2
x − y
15. lim
(x,y)→(0,0)
y
4
sin(xy)
x
2
+ y
2
16. lim
(x,y)→(0,0)
(x
2
+ y
2
) cos
¸
1
xy

17. lim
(x,y)→(0,0)
x
y
18. lim
(x,y)→(0,0)
cos
¸
1
xy

B
19. Show that f (x, y) =
1
2πσ
2
e
−(x
2
+y
2
)/2σ
2
, for σ > 0, is constant on the circle of radius
r > 0 centered at the origin. This function is called a Gaussian blur, and is used as
a filter in image processing software to produce a “blurred” effect.
20. Suppose that f (x, y) ≤ f (y, x) for all (x, y) in
2
. Show that f (x, y) = f (y, x) for all
(x, y) in
2
.
21. Use the substitution r =

x
2
+ y
2
to show that
lim
(x,y)→(0,0)
sin

x
2
+ y
2

x
2
+ y
2
= 1 .
(Hint: You will need to use L’Hôpital’s Rule for single-variable limits.)
C
22. Prove Theorem 2.1(a) in the case of addition. (Hint: Use Definition 2.1.)
23. Prove Theorem 2.1(b).
2.2 Partial Derivatives 71
2.2 Partial Derivatives
Now that we have an idea of what functions of several variables are, and what a limit
of such a function is, we can start to develop an idea of a derivative of a function of
two or more variables. We will start with the notion of a partial derivative.
Definition 2.3. Let f (x, y) be a real-valued function with domain D in
2
, and let
(a, b) be a point in D. Then the partial derivative of f at (a, b) with respect to x,
denoted by
∂f
∂x
(a, b), is defined as
∂f
∂x
(a, b) = lim
h→0
f (a + h, b) − f (a, b)
h
(2.2)
and the partial derivative of f at (a, b) with respect to y, denoted by
∂f
∂y
(a, b), is
defined as
∂f
∂y
(a, b) = lim
h→0
f (a, b + h) − f (a, b)
h
. (2.3)
Note: The symbol ∂ is pronounced “del”.
1
Recall that the derivative of a function f (x) can be interpreted as the rate of change
of that function in the (positive) x direction. From the definitions above, we can see
that the partial derivative of a function f (x, y) with respect to x or y is the rate of
change of f (x, y) in the (positive) x or y direction, respectively. What this means is
that the partial derivative of a function f (x, y) with respect to x can be calculated by
treating the y variable as a constant, and then simply differentiating f (x, y) as if it were
a function of x alone, using the usual rules from single-variable calculus. Likewise,
the partial derivative of f (x, y) with respect to y is obtained by treating the x variable
as a constant and then differentiating f (x, y) as if it were a function of y alone.
Example 2.10. Find
∂f
∂x
(x, y) and
∂f
∂y
(x, y) for the function f (x, y) = x
2
y + y
3
.
Solution: Treating y as a constant and differentiating f (x, y) with respect to x gives
∂f
∂x
(x, y) = 2xy
and treating x as a constant and differentiating f (x, y) with respect to y gives
∂f
∂y
(x, y) = x
2
+ 3y
2
.
1
It is not a Greek letter. The symbol was first used by the mathematicians A. Clairaut and L. Euler
around 1740, to distinguish it from the letter d used for the “usual” derivative.
72 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
We will often simply write
∂f
∂x
and
∂f
∂y
∂f
∂x
(x, y) and
∂f
∂y
(x, y).
Example 2.11. Find
∂f
∂x
and
∂f
∂y
for the function f (x, y) =
sin(xy
2
)
x
2
+ 1
.
Solution: Treating y as a constant and differentiating f (x, y) with respect to x gives
∂f
∂x
=
(x
2
+ 1)(y
2
cos(xy
2
)) − (2x) sin(xy
2
)
(x
2
+ 1)
2
and treating x as a constant and differentiating f (x, y) with respect to y gives
∂f
∂y
=
2xy cos(xy
2
)
x
2
+ 1
.
Since both
∂f
∂x
and
∂f
∂y
are themselves functions of x and y, we can take their partial
derivatives with respect to x and y. This yields the higher-order partial derivatives:

2
f
∂x
2
=

∂x

∂f
∂x

2
f
∂y
2
=

∂y

∂f
∂y

2
f
∂y ∂x
=

∂y

∂f
∂x

2
f
∂x ∂y
=

∂x

∂f
∂y

3
f
∂x
3
=

∂x

2
f
∂x
2

3
f
∂y
3
=

∂y

2
f
∂y
2

3
f
∂y ∂x
2
=

∂y

2
f
∂x
2

3
f
∂x ∂y
2
=

∂x

2
f
∂y
2

3
f
∂y
2
∂x
=

∂y

2
f
∂y ∂x

3
f
∂x
2
∂y
=

∂x

2
f
∂x ∂y

3
f
∂x ∂y ∂x
=

∂x

2
f
∂y ∂x

3
f
∂y ∂x ∂y
=

∂y

2
f
∂x ∂y

.
.
.
Example 2.12. Find the partial derivatives
∂f
∂x
,
∂f
∂y
,

2
f
∂x
2
,

2
f
∂y
2
,

2
f
∂y ∂x
and

2
f
∂x ∂y
for the
function f (x, y) = e
x
2
y
+ xy
3
.
2.2 Partial Derivatives 73
Solution: Proceeding as before, we have
∂f
∂x
= 2xye
x
2
y
+ y
3
∂f
∂y
= x
2
e
x
2
y
+ 3xy
2

2
f
∂x
2
=

∂x
(2xye
x
2
y
+ y
3
)

2
f
∂y
2
=

∂y
(x
2
e
x
2
y
+ 3xy
2
)
= 2ye
x
2
y
+ 4x
2
y
2
e
x
2
y
= x
4
e
x
2
y
+ 6xy

2
f
∂y ∂x
=

∂y
(2xye
x
2
y
+ y
3
)

2
f
∂x ∂y
=

∂x
(x
2
e
x
2
y
+ 3xy
2
)
= 2xe
x
2
y
+ 2x
3
ye
x
2
y
+ 3y
2
= 2xe
x
2
y
+ 2x
3
ye
x
2
y
+ 3y
2
Higher-order partial derivatives that are taken with respect to different variables,
such as

2
f
∂y ∂x
and

2
f
∂x ∂y
, are called mixed partial derivatives. Notice in the above
example that

2
f
∂y ∂x
=

2
f
∂x ∂y
. It turns that this will usually be the case. Specifically,
whenever both

2
f
∂y ∂x
and

2
f
∂x ∂y
are continuous at a point (a, b), then they are equal at that
point.
2
All the functions we will deal with will have continuous partial derivatives of
all orders, so you can assume in the remainder of the text that

2
f
∂y ∂x
=

2
f
∂x ∂y
for all (x, y) in the domain of f .
In other words, it doesn’t matter in which order you take partial derivatives. This
applies even to mixed partial derivatives of order 3 or higher.
The notation for partial derivatives varies. All of the following are equivalent:
∂f
∂x
: f
x
(x, y) , f
1
(x, y) , D
x
(x, y) , D
1
(x, y)
∂f
∂y
: f
y
(x, y) , f
2
(x, y) , D
y
(x, y) , D
2
(x, y)

2
f
∂x
2
: f
xx
(x, y) , f
11
(x, y) , D
xx
(x, y) , D
11
(x, y)

2
f
∂y
2
: f
yy
(x, y) , f
22
(x, y) , D
yy
(x, y) , D
22
(x, y)

2
f
∂y ∂x
: f
xy
(x, y) , f
12
(x, y) , D
xy
(x, y) , D
12
(x, y)

2
f
∂x ∂y
: f
yx
(x, y) , f
21
(x, y) , D
yx
(x, y) , D
21
(x, y)
2
See pp. 214-216 in TAYLOR and MANN for a proof.
74 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES

¨
Exercises
A
For Exercises 1-16, find
∂f
∂x
and
∂f
∂y
.
1. f (x, y) = x
2
+ y
2
2. f (x, y) = cos(x + y)
3. f (x, y) =

x
2
+ y + 4
4. f (x, y) =
x + 1
y + 1
5. f (x, y) = e
xy
+ xy 6. f (x, y) = x
2
− y
2
+ 6xy + 4x − 8y + 2
7. f (x, y) = x
4
8. f (x, y) = x + 2y
9. f (x, y) =

x
2
+ y
2
10. f (x, y) = sin(x + y)
11. f (x, y) =
3

x
2
+ y + 4
12. f (x, y) =
xy + 1
x + y
13. f (x, y) = e
−(x
2
+y
2
)
14. f (x, y) = ln(xy)
15. f (x, y) = sin(xy) 16. f (x, y) = tan(x + y)
For Exercises 17-26, find

2
f
∂x
2
,

2
f
∂y
2
and

2
f
∂y ∂x
(use Exercises 1-8, 14, 15).
17. f (x, y) = x
2
+ y
2
18. f (x, y) = cos(x + y)
19. f (x, y) =

x
2
+ y + 4
20. f (x, y) =
x + 1
y + 1
21. f (x, y) = e
xy
+ xy 22. f (x, y) = x
2
− y
2
+ 6xy + 4x − 8y + 2
23. f (x, y) = x
4
24. f (x, y) = x + 2y
25. f (x, y) = ln(xy) 26. f (x, y) = sin(xy)
B
27. Show that the function f (x, y) = sin(x + y) + cos(x − y) satisfies the wave equation

2
f
∂x
2

2
f
∂y
2
= 0 .
The wave equation is an example of a partial differential equation.
28. Let u and v be twice-differentiable functions of a single variable, and let c 0
be a constant. Show that f (x, y) = u(x + cy) + v(x − cy) is a solution of the general
one-dimensional wave equation
3

2
f
∂x
2

1
c
2

2
f
∂y
2
= 0 .
3
Conversely, it turns out that any solution must be of this form. See Ch. 1 in WEINBERGER.
2.3 Tangent Plane to a Surface 75
2.3 Tangent Plane to a Surface
In the previous section we mentioned that the partial derivatives
∂f
∂x
and
∂f
∂y
can be
thought of as the rate of change of a function z = f (x, y) in the positive x and y direc-
tions, respectively. Recall that the derivative
dy
dx
of a function y = f (x) has a geometric
meaning, namely as the slope of the tangent line to the graph of f at the point (x, f (x))
in
2
. There is a similar geometric meaning to the partial derivatives
∂f
∂x
and
∂f
∂y
of
a function z = f (x, y): given a point (a, b) in the domain D of f (x, y), the trace of the
surface described by z = f (x, y) in the plane y = b is a curve in
3
through the point
(a, b, f (a, b)), and the slope of the tangent line L
x
to that curve at that point is
∂f
∂x
(a, b).
Similarly,
∂f
∂y
(a, b) is the slope of the tangent line L
y
to the trace of the surface z = f (x, y)
in the plane x = a (see Figure 2.3.1).
y
z
x
0
(a, b)
D
L
x
b
(a, b, f (a, b))
slope =
∂f
∂x
(a, b)
z = f (x, y)
(a) Tangent line L
x
in the plane y = b
y
z
x
0
(a, b)
D
L
y
a
(a, b, f (a, b))
slope =
∂f
∂y
(a, b)
z = f (x, y)
(b) Tangent line L
y
in the plane x = a
Figure 2.3.1 Partial derivatives as slopes
Since the derivative
dy
dx
of a function y = f (x) is used to find the tangent line to the
graph of f (which is a curve in
2
), you might expect that partial derivatives can be
used to define a tangent plane to the graph of a surface z = f (x, y). This indeed turns
out to be the case. First, we need a definition of a tangent plane. The intuitive idea is
that a tangent plane “just touches” a surface at a point. The formal definition mimics
the intuitive notion of a tangent line to a curve.
Definition 2.4. Let z = f (x, y) be the equation of a surface S in
3
, and let P = (a, b, c)
be a point on S . Let T be a plane which contains the point P, and let Q = (x, y, z)
represent a generic point on the surface S . If the (acute) angle between the vector
−−→
PQ and the plane T approaches zero as the point Q approaches P along the surface S ,
then we call T the tangent plane to S at P.
Note that since two lines in
3
determine a plane, then the two tangent lines to the
surface z = f (x, y) in the x and y directions described in Figure 2.3.1 are contained in
76 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
the tangent plane at that point, if the tangent plane exists at that point. The existence
of those two tangent lines does not by itself guarantee the existence of the tangent
plane. It is possible that if we take the trace of the surface in the plane x − y = 0
(which makes a 45

angle with the positive x-axis), the resulting curve in that plane
may have a tangent line which is not in the plane determined by the other two tangent
lines, or it may not have a tangent line at all at that point. Luckily, it turns out
4
that
if
∂f
∂x
and
∂f
∂y
exist in a region around a point (a, b) and are continuous at (a, b) then the
tangent plane to the surface z = f (x, y) will exist at the point (a, b, f (a, b)). In this text,
those conditions will always hold.
y
z
x
0
(a, b, f (a, b))
z = f (x, y)
T
L
x
L
y
Figure 2.3.2 Tangent plane
Suppose that we want an equation of the tangent plane
T to the surface z = f (x, y) at a point (a, b, f (a, b)). Let L
x
and L
y
be the tangent lines to the traces of the surface
in the planes y = b and x = a, respectively (as in Figure
2.3.2), and suppose that the conditions for T to exist do
hold. Then the equation for T is
A(x − a) + B(y − b) + C(z − f (a, b)) = 0 (2.4)
where n = (A, B, C) is a normal vector to the plane T.
Since T contains the lines L
x
and L
y
, then all we need are vectors v
x
and v
y
that are
parallel to L
x
and L
y
, respectively, and then let n = v
x
××× v
y
.
x
z
0
v
x
= (1, 0,
∂f
∂x
(a, b))
∂f
∂x
(a, b)
1
Figure 2.3.3
Since the slope of L
x
is
∂f
∂x
(a, b), then the vector v
x
= (1, 0,
∂f
∂x
(a, b)) is
parallel to L
x
(since v
x
lies in the xz-plane and lies in a line with
slope
∂f
∂x
(a,b)
1
=
∂f
∂x
(a, b). See Figure 2.3.3). Similarly, the vector
v
y
= (0, 1,
∂f
∂y
(a, b)) is parallel to L
y
. Hence, the vector
n = v
x
××× v
y
=

i j k
1 0
∂f
∂x
(a, b)
0 1
∂f
∂y
(a, b)

= −
∂f
∂x
(a, b) i −
∂f
∂y
(a, b) j + k
is normal to the plane T. Thus the equation of T is

∂f
∂x
(a, b) (x − a) −
∂f
∂y
(a, b) (y − b) + z − f (a, b) = 0 . (2.5)
Multiplying both sides by −1, we have the following result:
The equation of the tangent plane to the surface z = f (x, y) at the point (a, b, f (a, b))
is
∂f
∂x
(a, b) (x − a) +
∂f
∂y
(a, b) (y − b) − z + f (a, b) = 0 (2.6)
4
See TAYLOR and MANN, § 6.4.
2.3 Tangent Plane to a Surface 77
Example 2.13. Find the equation of the tangent plane to the surface z = x
2
+y
2
at the
point (1, 2, 5).
Solution: For the function f (x, y) = x
2
+y
2
, we have
∂f
∂x
= 2x and
∂f
∂y
= 2y, so the equation
of the tangent plane at the point (1, 2, 5) is
2(1)(x − 1) + 2(2)(y − 2) − z + 5 = 0 , or
2x + 4y − z − 5 = 0 .
In a similar fashion, it can be shown that if a surface is defined implicitly by an
equation of the form F(x, y, z) = 0, then the tangent plane to the surface at a point
(a, b, c) is given by the equation
∂F
∂x
(a, b, c) (x − a) +
∂F
∂y
(a, b, c) (y − b) +
∂F
∂z
(a, b, c) (z − c) = 0 . (2.7)
Note that formula (2.6) is the special case of formula (2.7) where F(x, y, z) = f (x, y) − z.
Example 2.14. Find the equation of the tangent plane to the surface x
2
+y
2
+z
2
= 9 at
the point (2, 2, −1).
Solution: For the function F(x, y, z) = x
2
+ y
2
+ z
2
− 9, we have
∂F
∂x
= 2x,
∂F
∂y
= 2y, and
∂F
∂z
= 2z, so the equation of the tangent plane at (2, 2, −1) is
2(2)(x − 2) + 2(2)(y − 2) + 2(−1)(z + 1) = 0 , or
2x + 2y − z − 9 = 0 .

¨
Exercises
A
For Exercises 1-6, find the equation of the tangent plane to the surface z = f (x, y) at
the point P.
1. f (x, y) = x
2
+ y
3
, P = (1, 1, 2) 2. f (x, y) = xy, P = (1, −1, −1)
3. f (x, y) = x
2
y, P = (−1, 1, 1) 4. f (x, y) = xe
y
, P = (1, 0, 1)
5. f (x, y) = x + 2y, P = (2, 1, 4) 6. f (x, y) =

x
2
+ y
2
, P = (3, 4, 5)
For Exercises 7-10, find the equation of the tangent plane to the given surface at the
point P.
7.
x
2
4
+
y
2
9
+
z
2
16
= 1, P =

1, 2,
2

11
3

8. x
2
+ y
2
+ z
2
= 9, P = (0, 0, 3)
9. x
2
+ y
2
− z
2
= 0, P = (3, 4, 5) 10. x
2
+ y
2
= 4, P = (

3, 1, 0)
78 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
2.4 Directional Derivatives and the Gradient
For a function z = f (x, y), we learned that the partial derivatives
∂f
∂x
and
∂f
∂y
represent
the (instantaneous) rate of change of f in the positive x and y directions, respectively.
What about other directions? It turns out that we can find the rate of change in any
direction using a more general type of derivative called a directional derivative.
Definition 2.5. Let f (x, y) be a real-valued function with domain D in
2
, and let (a, b)
be a point in D. Let v be a unit vector in
2
. Then the directional derivative of f
at (a, b) in the direction of v, denoted by D
v
f (a, b), is defined as
D
v
f (a, b) = lim
h→0
f ((a, b) + hv) − f (a, b)
h
(2.8)
Notice in the definition that we seem to be treating the point (a, b) as a vector, since
we are adding the vector hv to it. But this is just the usual idea of identifying vectors
with their terminal points, which the reader should be used to by now. If we were to
write the vector v as v = (v
1
, v
2
), then
D
v
f (a, b) = lim
h→0
f (a + hv
1
, b + hv
2
) − f (a, b)
h
. (2.9)
From this we can immediately recognize that the partial derivatives
∂f
∂x
and
∂f
∂y
are
special cases of the directional derivative with v = i = (1, 0) and v = j = (0, 1), respec-
tively. That is,
∂f
∂x
= D
i
f and
∂f
∂y
= D
j
f . Since there are many vectors with the same
direction, we use a unit vector in the definition, as that represents a “standard” vector
for a given direction.
If f (x, y) has continuous partial derivatives
∂f
∂x
and
∂f
∂y
(which will always be the case
in this text), then there is a simple formula for the directional derivative:
Theorem 2.2. Let f (x, y) be a real-valued function with domain D in
2
such that the
partial derivatives
∂f
∂x
and
∂f
∂y
exist and are continuous in D. Let (a, b) be a point in D,
and let v = (v
1
, v
2
) be a unit vector in
2
. Then
D
v
f (a, b) = v
1
∂f
∂x
(a, b) + v
2
∂f
∂y
(a, b) . (2.10)
Proof: Note that if v = i = (1, 0) then the above formula reduces to D
v
f (a, b) =
∂f
∂x
(a, b),
which we know is true since D
i
f =
∂f
∂x
, as we noted earlier. Similarly, for v = j = (0, 1)
the formula reduces to D
v
f (a, b) =
∂f
∂y
(a, b), which is true since D
j
f =
∂f
∂y
. So since
i = (1, 0) and j = (0, 1) are the only unit vectors in
2
with a zero component, then we
2.4 Directional Derivatives and the Gradient 79
need only show the formula holds for unit vectors v = (v
1
, v
2
) with v
1
0 and v
2
0.
So fix such a vector v and fix a number h 0. Then
f (a +hv
1
, b +hv
2
) − f (a, b) = f (a +hv
1
, b +hv
2
) − f (a +hv
1
, b) + f (a +hv
1
, b) − f (a, b) . (2.11)
Since h 0 and v
2
0, then hv
2
0 and thus any number c between b and b + hv
2
can be written as c = b + αhv
2
for some number 0 < α < 1. So since the function
f (a + hv
1
, y) is a real-valued function of y (since a + hv
1
is a fixed number), then the
Mean Value Theorem from single-variable calculus can be applied to the function
g(y) = f (a + hv
1
, y) on the interval [b, b + hv
2
] (or [b + hv
2
, b] if one of h or v
2
is negative)
to find a number 0 < α < 1 such that
∂f
∂y
(a + hv
1
, b + αhv
2
) = g

(b + αhv
2
) =
g(b + hv
2
) − g(b)
b + hv
2
− b
=
f (a + hv
1
, b + hv
2
) − f (a + hv
1
, b)
hv
2
and so
f (a + hv
1
, b + hv
2
) − f (a + hv
1
, b) = hv
2
∂f
∂y
(a + hv
1
, b + αhv
2
) .
By a similar argument, there exists a number 0 < β < 1 such that
f (a + hv
1
, b) − f (a, b) = hv
1
∂f
∂x
(a + βhv
1
, b) .
Thus, by equation (2.11), we have
f (a + hv
1
, b + hv
2
) − f (a, b)
h
=
hv
2
∂f
∂y
(a + hv
1
, b + αhv
2
) + hv
1
∂f
∂x
(a + βhv
1
, b)
h
= v
2
∂f
∂y
(a + hv
1
, b + αhv
2
) + v
1
∂f
∂x
(a + βhv
1
, b)
so by formula (2.9) we have
D
v
f (a, b) = lim
h→0
f (a + hv
1
, b + hv
2
) − f (a, b)
h
= lim
h→0
,
v
2
∂f
∂y
(a + hv
1
, b + αhv
2
) + v
1
∂f
∂x
(a + βhv
1
, b)
¸
= v
2
∂f
∂y
(a, b) + v
1
∂f
∂x
(a, b) by the continuity of
∂f
∂x
and
∂f
∂y
, so
D
v
f (a, b) = v
1
∂f
∂x
(a, b) + v
2
∂f
∂y
(a, b)
after reversing the order of summation. QED
Note that D
v
f (a, b) = v···

∂f
∂x
(a, b),
∂f
∂y
(a, b)

. The second vector has a special name:
80 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
Definition 2.6. For a real-valued function f (x, y), the gradient of f , denoted by ∇f ,
is the vector
∇f =

∂f
∂x
,
∂f
∂y

(2.12)
in
2
. For a real-valued function f (x, y, z), the gradient is the vector
∇f =

∂f
∂x
,
∂f
∂y
,
∂f
∂z

(2.13)
in
3
. The symbol ∇ is pronounced “del”.
5
Corollary 2.3. D
v
f = v··· ∇f
Example 2.15. Find the directional derivative of f (x, y) = xy
2
+ x
3
y at the point (1, 2) in
the direction of v =

1

2
,
1

2

.
Solution: We see that ∇f = (y
2
+ 3x
2
y, 2xy + x
3
), so
D
v
f (1, 2) = v··· ∇f (1, 2) =

1

2
,
1

2

··· (2
2
+ 3(1)
2
(2), 2(1)(2) + 1
3
) =
15

2
A real-valued function z = f (x, y) whose partial derivatives
∂f
∂x
and
∂f
∂y
exist and are
continuous is called continuously differentiable. Assume that f (x, y) is such a function
and that ∇f 0. Let c be a real number in the range of f and let v be a unit vector in

2
which is tangent to the level curve f (x, y) = c (see Figure 2.4.1).
x
y
0
v ∇f
f (x, y) = c
Figure 2.4.1
5
Sometimes the notation grad( f ) is used instead of ∇f .
2.4 Directional Derivatives and the Gradient 81
The value of f (x, y) is constant along a level curve, so since v is a tangent vector to
this curve, then the rate of change of f in the direction of v is 0, i.e. D
v
f = 0. But we
know that D
v
f = v ··· ∇f = v ∇f cos θ, where θ is the angle between v and ∇f . So
since v = 1 then D
v
f = ∇f cos θ. So since ∇f 0 then D
v
f = 0 ⇒ cos θ = 0 ⇒ θ = 90

.
In other words, ∇f ⊥ v, which means that ∇f is normal to the level curve.
In general, for any unit vector v in
2
, we still have D
v
f = ∇f cos θ, where θ is the
angle between v and ∇f . At a fixed point (x, y) the length ∇f is fixed, and the value
of D
v
f then varies as θ varies. The largest value that D
v
f can take is when cos θ = 1
(θ = 0

), while the smallest value occurs when cos θ = −1 (θ = 180

). In other words, the
value of the function f increases the fastest in the direction of ∇f (since θ = 0

in that
case), and the value of f decreases the fastest in the direction of −∇f (since θ = 180

in that case). We have thus proved the following theorem:
Theorem 2.4. Let f (x, y) be a continuously differentiable real-valued function, with
∇f 0. Then:
(a) The gradient ∇f is normal to any level curve f (x, y) = c.
(b) The value of f (x, y) increases the fastest in the direction of ∇f .
(c) The value of f (x, y) decreases the fastest in the direction of −∇f .
Example 2.16. In which direction does the function f (x, y) = xy
2
+ x
3
y increase the
fastest from the point (1, 2)? In which direction does it decrease the fastest?
Solution: Since ∇f = (y
2
+ 3x
2
y, 2xy + x
3
), then ∇f (1, 2) = (10, 5) 0. A unit vector in
that direction is v =
∇f
∇f
=

2

5
,
1

5

. Thus, f increases the fastest in the direction of

2

5
,
1

5

and decreases the fastest in the direction of

−2

5
,
−1

5

.
Though we proved Theorem 2.4 for functions of two variables, a similar argument
can be used to show that it also applies to functions of three or more variables. Like-
wise, the directional derivative in the three-dimensional case can also be defined by
the formula D
v
f = v··· ∇f .
Example 2.17. The temperature T of a solid is given by the function T(x, y, z) = e
−x
+
e
−2y
+e
4z
, where x, y, z are space coordinates relative to the center of the solid. In which
direction from the point (1, 1, 1) will the temperature decrease the fastest?
Solution: Since ∇f = (−e
−x
, −2e
−2y
, 4e
4z
), then the temperature will decrease the fastest
in the direction of −∇f (1, 1, 1) = (e
−1
, 2e
−2
, −4e
4
).
82 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES

¨
Exercises
A
For Exercises 1-10, compute the gradient ∇f .
1. f (x, y) = x
2
+ y
2
− 1 2. f (x, y) =
1
x
2
+ y
2
3. f (x, y) =

x
2
+ y
2
+ 4 4. f (x, y) = x
2
e
y
5. f (x, y) = ln(xy) 6. f (x, y) = 2x + 5y
7. f (x, y, z) = sin(xyz) 8. f (x, y, z) = x
2
e
yz
9. f (x, y, z) = x
2
+ y
2
+ z
2
10. f (x, y, z) =

x
2
+ y
2
+ z
2
For Exercises 11-14, find the directional derivative of f at the point P in the direction
of v =

1

2
,
1

2

.
11. f (x, y) = x
2
+ y
2
− 1, P = (1, 1) 12. f (x, y) =
1
x
2
+ y
2
, P = (1, 1)
13. f (x, y) =

x
2
+ y
2
+ 4, P = (1, 1) 14. f (x, y) = x
2
e
y
, P = (1, 1)
For Exercises 15-16, find the directional derivative of f at the point P in the direction
of v =

1

3
,
1

3
,
1

3

.
15. f (x, y, z) = sin(xyz), P = (1, 1, 1) 16. f (x, y, z) = x
2
e
yz
, P = (1, 1, 1)
17. Repeat Example 2.16 at the point (2, 3).
18. Repeat Example 2.17 at the point (3, 1, 2).
B
For Exercises 19-26, let f (x, y) and g(x, y) be continuously differentiable real-valued
functions, let c be a constant, and let v be a unit vector in
2
. Show that:
19. ∇(c f ) = c ∇f 20. ∇( f + g) = ∇f + ∇g
21. ∇( f g) = f ∇g + g ∇f 22. ∇( f /g) =
g ∇f − f ∇g
g
2
if g(x, y) 0
23. D
−v
f = −D
v
f 24. D
v
(c f ) = c D
v
f
25. D
v
( f + g) = D
v
f + D
v
g 26. D
v
( f g) = f D
v
g + g D
v
f
27. The function r(x, y) =

x
2
+ y
2
is the length of the position vector r = x i + y j for
each point (x, y) in
2
. Show that ∇r =
1
r
r when (x, y) (0, 0), and that ∇(r
2
) = 2 r.
2.5 Maxima and Minima 83
2.5 Maxima and Minima
The gradient can be used to find extreme points of real-valued functions of several
variables, that is, points where the function has a local maximum or local minimum.
We will consider only functions of two variables; functions of three or more variables
require methods using linear algebra.
Definition 2.7. Let f (x, y) be a real-valued function, and let (a, b) be a point in the
domain of f . We say that f has a local maximum at (a, b) if f (x, y) ≤ f (a, b) for all
(x, y) inside some disk of positive radius centered at (a, b), i.e. there is some sufficiently
small r > 0 such that f (x, y) ≤ f (a, b) for all (x, y) for which (x − a)
2
+ (y − b)
2
< r
2
.
Likewise, we say that f has a local minimum at (a, b) if f (x, y) ≥ f (a, b) for all (x, y)
inside some disk of positive radius centered at (a, b).
If f (x, y) ≤ f (a, b) for all (x, y) in the domain of f , then f has a global maximum at
(a, b). If f (x, y) ≥ f (a, b) for all (x, y) in the domain of f , then f has a global minimum
at (a, b).
Suppose that (a, b) is a local maximum point for f (x, y), and that the first-order
partial derivatives of f exist at (a, b). We know that f (a, b) is the largest value of
f (x, y) as (x, y) goes in all directions from the point (a, b), in some sufficiently small
disk centered at (a, b). In particular, f (a, b) is the largest value of f in the x direction
(around the point (a, b)), that is, the single-variable function g(x) = f (x, b) has a local
maximum at x = a. So we know that g

(a) = 0. Since g

(x) =
∂f
∂x
(x, b), then
∂f
∂x
(a, b) = 0.
Similarly, f (a, b) is the largest value of f near (a, b) in the y direction and so
∂f
∂y
(a, b) = 0.
We thus have the following theorem:
Theorem 2.5. Let f (x, y) be a real-valued function such that both
∂f
∂x
(a, b) and
∂f
∂y
(a, b)
exist. Then a necessary condition for f (x, y) to have a local maximum or minimum at
(a, b) is that ∇f (a, b) = 0.
Note: Theorem 2.5 can be extended to apply to functions of three or more variables.
A point (a, b) where ∇f (a, b) = 0 is called a critical point for the function f (x, y). So
given a function f (x, y), to find the critical points of f you have to solve the equations
∂f
∂x
(x, y) = 0 and
∂f
∂y
(x, y) = 0 simultaneously for (x, y). Similar to the single-variable case,
the necessary condition that ∇f (a, b) = 0 is not always sufficient to guarantee that a
critical point is a local maximum or minimum.
Example 2.18. The function f (x, y) = xy has a critical point at (0, 0):
∂f
∂x
= y = 0 ⇒ y = 0,
and
∂f
∂y
= x = 0 ⇒ x = 0, so (0, 0) is the only critical point. But clearly f does not have a
local maximum or minimum at (0, 0) since any disk around (0, 0) contains points (x, y)
where the values of x and y have the same sign (so that f (x, y) = xy > 0 = f (0, 0))
and different signs (so that f (x, y) = xy < 0 = f (0, 0)). In fact, along the path y = x
84 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
in
2
, f (x, y) = x
2
, which has a local minimum at (0, 0), while along the path y = −x
we have f (x, y) = −x
2
, which has a local maximum at (0, 0). So (0, 0) is an example of
a saddle point, i.e. it is a local maximum in one direction and a local minimum in
another direction. The graph of f (x, y) is shown in Figure 2.5.1, which is a hyperbolic
paraboloid.
-10
-5
0
5
10
-10
-5
0
5
10
-100
-50
0
50
100
z
x
y
z
Figure 2.5.1 f (x, y) = xy, saddle point at (0, 0)
The following theorem gives sufficient conditions for a critical point to be a local
maximum or minimum of a smooth function (i.e. a function whose partial derivatives
of all orders exist and are continuous), which we will not prove here.
6
Theorem 2.6. Let f (x, y) be a smooth real-valued function, with a critical point at
(a, b) (i.e. ∇f (a, b) = 0). Define
D =

2
f
∂x
2
(a, b)

2
f
∂y
2
(a, b) −

2
f
∂y ∂x
(a, b)

2
Then
(a) if D > 0 and

2
f
∂x
2
(a, b) > 0, then f has a local minimum at (a, b)
(b) if D > 0 and

2
f
∂x
2
(a, b) < 0, then f has a local maximum at (a, b)
(c) if D < 0, then f has neither a local minimum nor a local maximum at (a, b)
(d) if D = 0, then the test fails.
6
See TAYLOR and MANN, § 7.6.
2.5 Maxima and Minima 85
If condition (c) holds, then (a, b) is a saddle point. Note that the assumption that
f (x, y) is smooth means that
D =

2
f
∂x
2
(a, b)

2
f
∂y ∂x
(a, b)

2
f
∂x ∂y
(a, b)

2
f
∂y
2
(a, b)

since

2
f
∂y ∂x
=

2
f
∂x ∂y
. Also, if D > 0 then

2
f
∂x
2
(a, b)

2
f
∂y
2
(a, b) = D +

2
f
∂y ∂x
(a, b)

2
> 0, and so

2
f
∂x
2
(a, b) and

2
f
∂y
2
(a, b) have the same sign. This means that in parts (a) and (b) of the
theorem one can replace

2
f
∂x
2
(a, b) by

2
f
∂y
2
(a, b) if desired.
Example 2.19. Find all local maxima and minima of f (x, y) = x
2
+ xy + y
2
− 3x.
Solution: First find the critical points, i.e. where ∇f = 0. Since
∂f
∂x
= 2x + y − 3 and
∂f
∂y
= x + 2y
then the critical points (x, y) are the common solutions of the equations
2x + y − 3 = 0
x + 2y = 0
which has the unique solution (x, y) = (2, −1). So (2, −1) is the only critical point.
To use Theorem 2.6, we need the second-order partial derivatives:

2
f
∂x
2
= 2 ,

2
f
∂y
2
= 2 ,

2
f
∂y ∂x
= 1
and so
D =

2
f
∂x
2
(2, −1)

2
f
∂y
2
(2, −1) −

2
f
∂y ∂x
(2, −1)

2
= (2)(2) − 1
2
= 3 > 0
and

2
f
∂x
2
(2, −1) = 2 > 0. Thus, (2, −1) is a local minimum.
Example 2.20. Find all local maxima and minima of f (x, y) = xy − x
3
− y
2
.
Solution: First find the critical points, i.e. where ∇f = 0. Since
∂f
∂x
= y − 3x
2
and
∂f
∂y
= x − 2y
86 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
then the critical points (x, y) are the common solutions of the equations
y − 3x
2
= 0
x − 2y = 0
The first equation yields y = 3x
2
, substituting that into the second equation yields
x − 6x
2
= 0, which has the solutions x = 0 and x =
1
6
. So x = 0 ⇒ y = 3(0) = 0 and
x =
1
6
⇒ y = 3

1
6

2
=
1
12
.
So the critical points are (x, y) = (0, 0) and (x, y) =

1
6
,
1
12

.
To use Theorem 2.6, we need the second-order partial derivatives:

2
f
∂x
2
= −6x ,

2
f
∂y
2
= −2 ,

2
f
∂y ∂x
= 1
So
D =

2
f
∂x
2
(0, 0)

2
f
∂y
2
(0, 0) −

2
f
∂y ∂x
(0, 0)

2
= (−6(0))(−2) − 1
2
= −1 < 0
and thus (0, 0) is a saddle point. Also,
D =

2
f
∂x
2

1
6
,
1
12

2
f
∂y
2

1
6
,
1
12

2
f
∂y ∂x

1
6
,
1
12

2
= (−6
1
6

)(−2) − 1
2
= 1 > 0
and

2
f
∂x
2

1
6
,
1
12

= −1 < 0. Thus,

1
6
,
1
12

is a local maximum.
Example 2.21. Find all local maxima and minima of f (x, y) = (x − 2)
4
+ (x − 2y)
2
.
Solution: First find the critical points, i.e. where ∇f = 0. Since
∂f
∂x
= 4(x − 2)
3
+ 2(x − 2y) and
∂f
∂y
= −4(x − 2y)
then the critical points (x, y) are the common solutions of the equations
4(x − 2)
3
+ 2(x − 2y) = 0
−4(x − 2y) = 0
The second equation yields x = 2y, substituting that into the first equation yields
4(2y − 2)
3
= 0, which has the solution y = 1, and so x = 2(1) = 2. Thus, (2, 1) is the only
critical point.
To use Theorem 2.6, we need the second-order partial derivatives:

2
f
∂x
2
= 12(x − 2)
2
+ 2 ,

2
f
∂y
2
= 8 ,

2
f
∂y ∂x
= −4
2.5 Maxima and Minima 87
So
D =

2
f
∂x
2
(2, 1)

2
f
∂y
2
(2, 1) −

2
f
∂y ∂x
(2, 1)

2
= (2)(8) − (−4)
2
= 0
and so the test fails. What can be done in this situation? Sometimes it is possible
to examine the function to see directly the nature of a critical point. In our case, we
see that f (x, y) ≥ 0 for all (x, y), since f (x, y) is the sum of fourth and second powers
of numbers and hence must be nonnegative. But we also see that f (2, 1) = 0. Thus
f (x, y) ≥ 0 = f (2, 1) for all (x, y), and hence (2, 1) is in fact a global minimum for f .
Example 2.22. Find all local maxima and minima of f (x, y) = (x
2
+ y
2
)e
−(x
2
+y
2
)
.
Solution: First find the critical points, i.e. where ∇f = 0. Since
∂f
∂x
= 2x(1 − (x
2
+ y
2
))e
−(x
2
+y
2
)
∂f
∂y
= 2y(1 − (x
2
+ y
2
))e
−(x
2
+y
2
)
then the critical points are (0, 0) and all points (x, y) on the unit circle x
2
+ y
2
= 1.
To use Theorem 2.6, we need the second-order partial derivatives:

2
f
∂x
2
= 2[1 − (x
2
+ y
2
) − 2x
2
− 2x
2
(1 − (x
2
+ y
2
))]e
−(x
2
+y
2
)

2
f
∂y
2
= 2[1 − (x
2
+ y
2
) − 2y
2
− 2y
2
(1 − (x
2
+ y
2
))]e
−(x
2
+y
2
)

2
f
∂y ∂x
= −4xy[2 − (x
2
+ y
2
)]e
−(x
2
+y
2
)
At (0, 0), we have D = 4 > 0 and

2
f
∂x
2
(0, 0) = 2 > 0, so (0, 0) is a local minimum. However,
for points (x, y) on the unit circle x
2
+ y
2
= 1, we have
D = (−4x
2
e
−1
)(−4y
2
e
−1
) − (−4xye
−1
)
2
= 0
and so the test fails. If we look at the graph of f (x, y), as shown in Figure 2.5.2, it
looks like we might have a local maximum for (x, y) on the unit circle x
2
+ y
2
= 1. If we
switch to using polar coordinates (r, θ) instead of (x, y) in
2
, where r
2
= x
2
+ y
2
, then
we see that we can write f (x, y) as a function g(r) of the variable r alone: g(r) = r
2
e
−r
2
.
Then g

(r) = 2r(1 − r
2
)e
−r
2
, so it has a critical point at r = 1, and we can check that
g
′′
(1) = −4e
−1
< 0, so the Second Derivative Test from single-variable calculus says
that r = 1 is a local maximum. But r = 1 corresponds to the unit circle x
2
+ y
2
= 1.
Thus, the points (x, y) on the unit circle x
2
+ y
2
= 1 are local maximum points for f .
88 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
-3
-2
-1
0
1
2
3
-3
-2
-1
0
1
2
3
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
z
x
y
z
Figure 2.5.2 f (x, y) = (x
2
+ y
2
)e
−(x
2
+y
2
)

¨
Exercises
A
For Exercises 1-10, find all local maxima and minima of the function f (x, y).
1. f (x, y) = x
3
− 3x + y
2
2. f (x, y) = x
3
− 12x + y
2
+ 8y
3. f (x, y) = x
3
− 3x + y
3
− 3y 4. f (x, y) = x
3
+ 3x
2
+ y
3
− 3y
2
5. f (x, y) = 2x
3
+ 6xy + 3y
2
6. f (x, y) = 2x
3
− 6xy + y
2
7. f (x, y) =

x
2
+ y
2
8. f (x, y) = x + 2y
9. f (x, y) = 4x
2
− 4xy + 2y
2
+ 10x − 6y 10. f (x, y) = −4x
2
+ 4xy − 2y
2
+ 16x − 12y
B
11. For a rectangular solid of volume 1000 cubic meters, find the dimensions that will
minimize the surface area. (Hint: Use the volume condition to write the surface
area as a function of just two variables.)
12. Prove that if (a, b) is a local maximum or local minimum point for a smooth func-
tion f (x, y), then the tangent plane to the surface z = f (x, y) at the point (a, b, f (a, b))
is parallel to the xy-plane. (Hint: Use Theorem 2.5.)
C
13. Find three positive numbers x, y, z whose sum is 10 such that x
2
y
2
z is a maximum.
2.6 Unconstrained Optimization: Numerical Methods 89
2.6 Unconstrained Optimization: Numerical Methods
The types of problems that we solved in the previous section were examples of uncon-
strained optimization problems. That is, we tried to find local (and perhaps even
global) maximum and minimum points of real-valued functions f (x, y), where the
points (x, y) could be any points in the domain of f . The method we used required
us to find the critical points of f , which meant having to solve the equation ∇f = 0,
which in general is a system of two equations in two unknowns (x and y). While this
was relatively simple for the examples we did, in general this will not be the case. If
the equations involve polynomials in x and y of degree three or higher, or complicated
expressions involving trigonometric, exponential, or logarithmic functions, then solv-
ing even one such equation, let alone two, could be impossible by elementary means.
7
For example, if one of the equations that had to be solved was
x
3
+ 9x − 2 = 0 ,
you may have a hard time getting the exact solutions. Trial and error would not help
much, especially since the only real solution
8
turns out to be
3

28 + 1−
3

28 − 1. In a
situation such as this, the only choice may be to find a solution using some numerical
method which gives a sequence of numbers which converge to the actual solution. For
example, Newton’s method for solving equations f (x) = 0, which you probably learned
in single-variable calculus. In this section we will describe another method of Newton
for finding critical points of real-valued functions of two variables.
Let f (x, y) be a smooth real-valued function, and define
D(x, y) =

2
f
∂x
2
(x, y)

2
f
∂y
2
(x, y) −

2
f
∂y ∂x
(x, y)

2
.
Newton’s algorithm: Pick an initial point (x
0
, y
0
). For n = 0, 1, 2, 3, . . . , define:
x
n+1
= x
n

2
f
∂y
2
(x
n
, y
n
)

2
f
∂x ∂y
(x
n
, y
n
)
∂f
∂y
(x
n
, y
n
)
∂f
∂x
(x
n
, y
n
)

D(x
n
, y
n
)
, y
n+1
= y
n

2
f
∂x
2
(x
n
, y
n
)

2
f
∂x ∂y
(x
n
, y
n
)
∂f
∂x
(x
n
, y
n
)
∂f
∂y
(x
n
, y
n
)

D(x
n
, y
n
)
(2.14)
Then the sequence of points (x
n
, y
n
)

n=1
converges to a critical point. If there are several
critical points, then you will have to try different initial points to find them.
7
This is also a problem for the equivalent method (the Second Derivative Test) in single-variable calcu-
lus, though one that is not usually emphasized.
8
There are also two nonreal, complex number solutions. Cubic polynomial equations in one variable
can be solved using Cardan’s formulas, which are not quite as simple as the familiar quadratic formula.
See USPENSKY for more details. There are formulas for solving polynomial equations of degree 4, but
it can be proved that there is no general formula for solving equations for polynomials of degree five or
higher.
90 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
Example 2.23. Find all local maxima and minima of f (x, y) = x
3
− xy − x + xy
3
− y
4
.
Solution: First calculate the necessary partial derivatives:
∂f
∂x
= 3x
2
− y − 1 + y
3
,
∂f
∂y
= −x + 3xy
2
− 4y
3

2
f
∂x
2
= 6x ,

2
f
∂y
2
= 6xy − 12y
2
,

2
f
∂y ∂x
= −1 + 3y
2
Notice that solving ∇f = 0 would involve solving two third-degree polynomial equa-
tions in x and y, which in this case can not be done easily.
We need to pick an initial point (x
0
, y
0
) for our algorithm. Looking at the graph of
z = f (x, y) over a large region may help (see Figure 2.6.1 below), though it may be hard
to tell where the critical points are.
-20
-15
-10
-5
0
5
10
15
20
-20
-15
-10
-5
0
5
10
15
20
-350000
-300000
-250000
-200000
-150000
-100000
-50000
0
50000
z
x
y
z
Figure 2.6.1 f (x, y) = x
3
− xy − x + xy
3
− y
4
for −20 ≤ x ≤ 20 and −20 ≤ y ≤ 20
Notice in the formulas (2.14) that we divide by D, so we should pick an initial point
where D is not zero. And we can see that D(0, 0) = (0)(0) − (−1)
2
= −1 0, so take
(0, 0) as our initial point. Since it may take a large number of iterations of Newton’s
algorithm to be sure that we are close enough to the actual critical point, and since
the computations are quite tedious, we will let a computer do the computing. For this,
we will write a simple program, using the Java programming language, which will
take a given initial point as a parameter and then perform 100 iterations of Newton’s
algorithm. In each iteration the new point will be printed, so that we can see if there
is convergence. The full code is shown in Listing 2.1.
2.6 Unconstrained Optimization: Numerical Methods 91
//Program t o f i nd the c r i t i c al poi nt s of f ( x , y)=x^3−xy−x+xy^3−y^4
public class newton {
public static void main( Stri ng [ ] args ) {
//Get the i ni t i al poi nt ( x , y ) as command−l i ne parameters
double x = Double . parseDouble ( args [ 0 ] ) ; //I ni t i al x value
double y = Double . parseDouble ( args [ 1 ] ) ; //I ni t i al y value
System. out . pri nt l n ( " I ni t i al poi nt : ( " + x + " , " + y + " ) " ) ;
//Go through 100 i t e r at i ons of Newton ’ s algorithm
for ( int n=1; n<=100; n++) {
double D = fxx ( x , y) ∗ fyy ( x , y ) − Math. pow( fxy ( x , y ) , 2 ) ;
double xn = x ; double yn = y ; //The current x and y values
i f (D == 0) { //We can not di vi de by 0
System. out . pri nt l n ( " Error : D = 0 at i t er at i on n = " + n ) ;
System. exi t ( 0 ) ; //End the program
} else { //Cal cul ate the new values f or x and y
x = xn − ( fyy ( xn , yn) ∗ f x ( xn , yn) − fxy ( xn , yn) ∗ f y ( xn , yn ) ) / D;
y = yn − ( fxx ( xn , yn) ∗ f y ( xn , yn) − fxy ( xn , yn) ∗ f x ( xn , yn ) ) / D;
System. out . pri nt l n ( "n = " + n + " : ( " + x + " , " + y + " ) " ) ;
}
}
}
//Below are the parts s pe c i f i c t o the f unct i on f
//The f i r s t part i al deri vat i ve of f wrt x : 3x^2−y−1+y^3
public static double f x ( double x , double y ) {
return 3∗Math. pow( x , 2) − y − 1 + Math. pow( y , 3 ) ;
}
//The f i r s t part i al deri vat i ve of f wrt y : −x+3xy^2−4y^3
public static double f y ( double x , double y ) {
return −x + 3∗x∗Math. pow( y , 2) − 4∗Math. pow( y , 3 ) ;
}
//The second part i al deri vat i ve of f wrt x : 6x
public static double fxx ( double x , double y ) {
return 6∗x ;
}
//The second part i al deri vat i ve of f wrt y : 6xy−12y^2
public static double fyy ( double x , double y ) {
return 6∗x∗y − 12∗Math. pow( y , 2 ) ;
}
//The mixed second part i al deri vat i ve of f wrt x and y : −1+3y^2
public static double fxy ( double x , double y ) {
return −1 + 3∗Math. pow( y , 2 ) ;
}
}
Listing 2.1 Program listing for newton.java
92 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
To use this program, you should first save the code in Listing 2.1 in a plain text
file called newton.java. You will need the Java Development Kit
9
to compile the
code. In the directory where newton.java is saved, run this command at a command
prompt to compile the code: javac newton.java
Then run the program with the initial point (0, 0) with this command:
java newton 0 0
Below is the output of the program using (0, 0) as the initial point, truncated to show
the first 10 lines and the last 5 lines:
java newton 0 0
Initial point: (0.0,0.0)
n = 1: (0.0,-1.0)
n = 2: (1.0,-0.5)
n = 3: (0.6065857885615251,-0.44194107452339687)
n = 4: (0.484506572966545,-0.405341511995805)
n = 5: (0.47123972682634485,-0.3966334583092305)
n = 6: (0.47113558510349535,-0.39636450001936047)
n = 7: (0.4711356343449705,-0.3963643379632247)
n = 8: (0.4711356343449874,-0.39636433796318005)
n = 9: (0.4711356343449874,-0.39636433796318005)
n = 10: (0.4711356343449874,-0.39636433796318005)
...
n = 96: (0.4711356343449874,-0.39636433796318005)
n = 97: (0.4711356343449874,-0.39636433796318005)
n = 98: (0.4711356343449874,-0.39636433796318005)
n = 99: (0.4711356343449874,-0.39636433796318005)
n = 100: (0.4711356343449874,-0.39636433796318005)
As you can see, we appear to have converged fairly quickly (after only 8 iterations)
to what appears to be an actual critical point (up to Java’s level of precision), namely
the point (0.4711356343449874, −0.39636433796318005). It is easy to confirm that ∇f = 0
at this point, either by evaluating
∂f
∂x
and
∂f
∂y
at the point ourselves or by modifying our
program to also print the values of the partial derivatives at the point. It turns out
that both partial derivatives are indeed close enough to zero to be considered zero:
∂f
∂x
(0.4711356343449874, −0.39636433796318005) = 4.85722573273506 × 10
−17
∂f
∂y
(0.4711356343449874, −0.39636433796318005) = −8.326672684688674 × 10
−17
We also have D(0.4711356343449874, −0.39636433796318005) = −8.776075636032301 < 0,
so by Theorem 2.6 we know that (0.4711356343449874, −0.39636433796318005) is a sad-
dle point.
9
2.6 Unconstrained Optimization: Numerical Methods 93
Since ∇f consists of cubic polynomials, it seems likely that there may be three criti-
cal points. The computer program makes experimenting with other initial points easy,
and trying different values does indeed lead to different sequences which converge:
java newton -1 -1
Initial point: (-1.0,-1.0)
n = 1: (-0.5,-0.5)
n = 2: (-0.49295774647887325,-0.08450704225352113)
n = 3: (-0.1855674752461383,-1.2047647348546167)
n = 4: (-0.4540060574531383,-0.8643989895639324)
n = 5: (-0.3672160534444,-0.5426077421319053)
n = 6: (-0.4794622222856417,-0.24529117721011612)
n = 7: (0.11570743992954591,-2.4319791238981274)
n = 8: (-0.05837851765533317,-1.6536079835854451)
n = 9: (-0.129841298650007,-1.121516233310142)
n = 10: (-1.004453014967208,-0.9206128022529645)
n = 11: (-0.5161209914612475,-0.4176293491131443)
n = 12: (-0.5788664043863884,0.2918236503332734)
n = 13: (-0.6985177124230715,0.49848120123515316)
n = 14: (-0.6733618916578702,0.4345777963475479)
n = 15: (-0.6704392913413444,0.4252025996474051)
n = 16: (-0.6703832679150286,0.4250147307973365)
n = 17: (-0.6703832459238701,0.42501465652421205)
n = 18: (-0.6703832459238667,0.4250146565242004)
n = 19: (-0.6703832459238667,0.42501465652420045)
n = 20: (-0.6703832459238667,0.42501465652420045)
...
n = 98: (-0.6703832459238667,0.42501465652420045)
n = 99: (-0.6703832459238667,0.42501465652420045)
n = 100: (-0.6703832459238667,0.42501465652420045)
Again, it is easy to confirm that both
∂f
∂x
and
∂f
∂y
vanish at the point
(−0.6703832459238667, 0.42501465652420045), which means it is a critical point. And
D(−0.6703832459238667, 0.42501465652420045) = 15.3853578526055 > 0

2
f
∂x
2
(−0.6703832459238667, 0.42501465652420045) = −4.0222994755432 < 0
so we know that (−0.6703832459238667, 0.42501465652420045) is a local maximum. An
idea of what the graph of f looks like near that point is shown in Figure 2.6.2, which
does suggest a local maximum around that point.
Finally, running the computer program with the initial point (−5, −5) yields the crit-
ical point (−7.540962756992551, −5.595509445899435), with D < 0 at that point, which
94 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
-1
-0.8
-0.6
-0.4
-0.2
0
0
0.2
0.4
0.6
0.8
1
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
z
x
y
z
.
(-0.67,0.42,0.57)
Figure 2.6.2 f (x, y) = x
3
− xy − x + xy
3
− y
4
for −1 ≤ x ≤ 0 and 0 ≤ y ≤ 1
We can summarize our findings for the function f (x, y) = x
3
− xy − x + xy
3
− y
4
:
(−0.6703832459238667, 0.42501465652420045) : local maximum
The derivation of Newton’s algorithm, and the proof that it converges (given a “rea-
sonable” choice for the initial point) requires techniques beyond the scope of this text.
See RALSTON and RABINOWITZ for more detail and for discussion of other numerical
methods. Our description of Newton’s algorithm is the special two-variable case of a
more general algorithm that can be applied to functions of n ≥ 2 variables.
In the case of functions which have a global maximum or minimum, Newton’s algo-
rithm can be used to find those points. In general, global maxima and minima tend
to be more interesting than local versions, at least in practical applications. A maxi-
mization problem can always be turned into a minimization problem (why?), so a large
number of methods have been developed to find the global minimum of functions of
any number of variables. This field of study is called nonlinear programming. Many of
these methods are based on the steepest descent technique, which is based on an idea
that we discussed in Section 2.4. Recall that the negative gradient −∇f gives the di-
rection of the fastest rate of decrease of a function f . The crux of the steepest descent
idea, then, is that starting from some initial point, you move a certain amount in the
2.6 Unconstrained Optimization: Numerical Methods 95
direction of −∇f at that point. Wherever that takes you becomes your new point, and
you then just keep repeating that procedure until eventually (hopefully) you reach the
point where f has its smallest value. There is a “pure” steepest descent method, and a
multitude of variations on it that improve the rate of convergence, ease of calculation,
etc. In fact, Newton’s algorithm can be interpreted as a modified steepest descent
method. For more discussion of this, and of nonlinear programming in general, see
BAZARAA, SHERALI and SHETTY.

¨
Exercises
C
1. Recall Example 2.21 from the previous section, where we showed that the point
(2, 1) was a global minimum for the function f (x, y) = (x − 2)
4
+ (x − 2y)
2
. Notice
that our computer program can be modified fairly easily to use this function (just
change the return values in the fx, fy, fxx, fyy and fxy function definitions to use the
appropriate partial derivative). Either modify that program or write one of your
own in a programming language of your choice to show that Newton’s algorithm
does lead to the point (2, 1). First use the initial point (0, 3), then use the initial
point (3, 2), and compare the results. Make sure that your program attempts to do
100 iterations of the algorithm. Did anything strange happen when your program
ran? If so, how do you explain it? (Hint: Something strange should happen.)
2. There is a version of Newton’s algorithm for solving a system of two equations
f
1
(x, y) = 0 and f
2
(x, y) = 0 ,
where f
1
(x, y) and f
2
(x, y) are smooth real-valued functions:
Pick an initial point (x
0
, y
0
). For n = 0, 1, 2, 3, . . . , define:
x
n+1
= x
n

f
1
(x
n
, y
n
) f
2
(x
n
, y
n
)
∂f
1
∂y
(x
n
, y
n
)
∂f
2
∂y
(x
n
, y
n
)

D(x
n
, y
n
)
, y
n+1
= y
n
+

f
1
(x
n
, y
n
) f
2
(x
n
, y
n
)
∂f
1
∂x
(x
n
, y
n
)
∂f
2
∂x
(x
n
, y
n
)

D(x
n
, y
n
)
, where
D(x
n
, y
n
) =
∂f
1
∂x
(x
n
, y
n
)
∂f
2
∂y
(x
n
, y
n
) −
∂f
1
∂y
(x
n
, y
n
)
∂f
2
∂x
(x
n
, y
n
) .
Then the sequence of points (x
n
, y
n
)

n=1
converges to a solution. Write a computer
program that uses this algorithm to find approximate solutions to the system of
equations
f
1
(x, y) = sin(xy) − x − y = 0 and f
2
(x, y) = e
2x
− 2x + 3y = 0 .
Show that you get two different solutions when using (0, 0) and (1, 1) for the initial
point (x
0
, y
0
).
96 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
2.7 Constrained Optimization: Lagrange Multipliers
In Sections 2.5 and 2.6 we were concerned with finding maxima and minima of func-
tions without any constraints on the variables (other than being in the domain of the
function). What would we do if there were constraints on the variables? The following
example illustrates a simple case of this type of problem.
Example 2.24. For a rectangle whose perimeter is 20 m, find the dimensions that will
maximize the area.
Solution: The area A of a rectangle with width x and height y is A = xy. The perimeter
P of the rectangle is then given by the formula P = 2x +2y. Since we are given that the
perimeter P = 20, this problem can be stated as:
Maximize : f (x, y) = xy
given : 2x + 2y = 20
The reader is probably familiar with a simple method, using single-variable calculus,
for solving this problem. Since we must have 2x + 2y = 20, then we can solve for, say,
y in terms of x using that equation. This gives y = 10 − x, which we then substitute
into f to get f (x, y) = xy = x(10 − x) = 10x − x
2
. This is now a function of x alone, so we
now just have to maximize the function f (x) = 10x − x
2
on the interval [0, 10]. Since
f

(x) = 10−2x = 0 ⇒ x = 5 and f
′′
(5) = −2 < 0, then the Second Derivative Test tells us
that x = 5 is a local maximum for f , and hence x = 5 must be the global maximum on
the interval [0, 10] (since f = 0 at the endpoints of the interval). So since y = 10 − x = 5,
then the maximum area occurs for a rectangle whose width and height both are 5 m.
Notice in the above example that the ease of the solution depended on being able to
solve for one variable in terms of the other in the equation 2x + 2y = 20. But what if
that were not possible (which is often the case)? In this section we will use a general
method, called the Lagrange multiplier method
10
, for solving constrained optimization
problems:
Maximize (or minimize) : f (x, y) (or f (x, y, z))
given : g(x, y) = c (or g(x, y, z) = c) for some constant c
The equation g(x, y) = c is called the constraint equation, and we say that x and y are
constrained by g(x, y) = c. Points (x, y) which are maxima or minima of f (x, y) with the
condition that they satisfy the constraint equation g(x, y) = c are called constrained
maximum or constrained minimum points, respectively. Similar definitions hold for
functions of three variables.
The Lagrange multiplier method for solving such problems can now be stated:
10
Named after the French mathematician Joseph Louis Lagrange (1736-1813).
2.7 Constrained Optimization: Lagrange Multipliers 97
Theorem2.7. Let f (x, y) and g(x, y) be smooth functions, and suppose that c is a scalar
constant such that ∇g(x, y) 0 for all (x, y) that satisfy the equation g(x, y) = c. Then to
solve the constrained optimization problem
Maximize (or minimize) : f (x, y)
given : g(x, y) = c ,
find the points (x, y) that solve the equation ∇f (x, y) = λ∇g(x, y) for some constant λ
(the number λ is called the Lagrange multiplier). If there is a constrained maximum
or minimum, then it must be such a point.
A rigorous proof of the above theorem requires use of the Implicit Function The-
orem, which is beyond the scope of this text.
11
Note that the theorem only gives a
necessary condition for a point to be a constrained maximum or minimum. Whether a
point (x, y) that satisfies ∇f (x, y) = λ∇g(x, y) for some λ actually is a constrained max-
imum or minimum can sometimes be determined by the nature of the problem itself.
For instance, in Example 2.24 it was clear that there had to be a global maximum.
So how can you tell when a point that satisfies the condition in Theorem 2.7 really is
a constrained maximum or minimum? The answer is that it depends on the constraint
function g(x, y), together with any implicit constraints. It can be shown
12
that if the
constraint equation g(x, y) = c (plus any hidden constraints) describes a bounded set
B in
2
, then the constrained maximum or minimum of f (x, y) will occur either at a
point (x, y) satisfying ∇f (x, y) = λ∇g(x, y) or at a “boundary” point of the set B.
In Example 2.24 the constraint equation 2x + 2y = 20 describes a line in
2
, which
by itself is not bounded. However, there are “hidden” constraints, due to the nature
of the problem, namely 0 ≤ x, y ≤ 10, which cause that line to be restricted to a line
segment in
2
(including the endpoints of that line segment), which is bounded.
Example 2.25. For a rectangle whose perimeter is 20 m, use the Lagrange multiplier
method to find the dimensions that will maximize the area.
Solution: As we saw in Example 2.24, with x and y representing the width and height,
respectively, of the rectangle, this problem can be stated as:
Maximize : f (x, y) = xy
given : g(x, y) = 2x + 2y = 20
Then solving the equation ∇f (x, y) = λ∇g(x, y) for some λ means solving the equations
11
See TAYLOR and MANN, § 6.8 for more detail.
12
Again, see TAYLOR and MANN.
98 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
∂f
∂x
= λ
∂g
∂x
and
∂f
∂y
= λ
∂g
∂y
, namely:
y = 2λ ,
x = 2λ
The general idea is to solve for λ in both equations, then set those expressions equal
(since they both equal λ) to solve for x and y. Doing this we get
y
2
= λ =
x
2
⇒ x = y ,
so now substitute either of the expressions for x or y into the constraint equation to
solve for x and y:
20 = g(x, y) = 2x + 2y = 2x + 2x = 4x ⇒ x = 5 ⇒ y = 5
There must be a maximum area, since the minimum area is 0 and f (5, 5) = 25 >
0, so the point (5, 5) that we found (called a constrained critical point) must be the
constrained maximum.
∴ The maximum area occurs for a rectangle whose width and height both are 5 m.
Example 2.26. Find the points on the circle x
2
+ y
2
= 80 which are closest to and
farthest from the point (1, 2).
Solution: The distance d from any point (x, y) to the point (1, 2) is
d =

(x − 1)
2
+ (y − 2)
2
,
and minimizing the distance is equivalent to minimizing the square of the distance.
Thus the problem can be stated as:
Maximize (and minimize) : f (x, y) = (x − 1)
2
+ (y − 2)
2
given : g(x, y) = x
2
+ y
2
= 80
Solving ∇f (x, y) = λ∇g(x, y) means solving the following equations:
2(x − 1) = 2λx ,
2(y − 2) = 2λy
Note that x 0 since otherwise we would get −2 = 0 in the first equation. Similarly,
y 0. So we can solve both equations for λ as follows:
x − 1
x
= λ =
y − 2
y
⇒ xy − y = xy − 2x ⇒ y = 2x
2.7 Constrained Optimization: Lagrange Multipliers 99
x
y
0
(4, 8)
(1, 2)
(−4, −8)
x
2
+ y
2
= 80
Figure 2.7.1
Substituting this into g(x, y) = x
2
+ y
2
= 80 yields 5x
2
= 80,
so x = ±4. So the two constrained critical points are (4, 8)
and (−4, −8). Since f (4, 8) = 45 and f (−4, −8) = 125, and since
there must be points on the circle closest to and farthest
from (1, 2), then it must be the case that (4, 8) is the point
on the circle closest to (1, 2) and (−4, −8) is the farthest from
(1, 2) (see Figure 2.7.1).
Notice that since the constraint equation x
2
+ y
2
= 80 de-
scribes a circle, which is a bounded set in
2
, then we were
guaranteed that the constrained critical points we found
were indeed the constrained maximum and minimum.
The Lagrange multiplier method can be extended to functions of three variables.
Example 2.27.
Maximize (and minimize) : f (x, y, z) = x + z
given : g(x, y, z) = x
2
+ y
2
+ z
2
= 1
Solution: Solve the equation ∇f (x, y, z) = λ∇g(x, y, z):
1 = 2λx
0 = 2λy
1 = 2λz
The first equation implies λ 0 (otherwise we would have 1 = 0), so we can divide
by λ in the second equation to get y = 0 and we can divide by λ in the first and
third equations to get x =
1

= z. Substituting these expressions into the constraint
equation g(x, y, z) = x
2
+ y
2
+ z
2
= 1 yields the constrained critical points

1

2
, 0,
1

2

and

−1

2
, 0,
−1

2

. Since f

1

2
, 0,
1

2

> f

−1

2
, 0,
−1

2

, and since the constraint equation
x
2
+ y
2
+ z
2
= 1 describes a sphere (which is bounded) in
3
, then

1

2
, 0,
1

2

is the
constrained maximum point and

−1

2
, 0,
−1

2

is the constrained minimum point.
So far we have not attached any significance to the value of the Lagrange multiplier
λ. We needed λ only to find the constrained critical points, but made no use of its value.
It turns out that λ gives an approximation of the change in the value of the function
f (x, y) that we wish to maximize or minimize, when the constant c in the constraint
equation g(x, y) = c is changed by 1.
100 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
For example, in Example 2.25 we showed that the constrained optimization problem
Maximize : f (x, y) = xy
given : g(x, y) = 2x + 2y = 20
had the solution (x, y) = (5, 5), and that λ = x/2 = y/2. Thus, λ = 2.5. In a similar
fashion we could show that the constrained optimization problem
Maximize : f (x, y) = xy
given : g(x, y) = 2x + 2y = 21
has the solution (x, y) = (5.25, 5.25). So we see that the value of f (x, y) at the constrained
maximum increased from f (5, 5) = 25 to f (5.25, 5.25) = 27.5625, i.e. it increased by
2.5625 when we increased the value of c in the constraint equation g(x, y) = c from
c = 20 to c = 21. Notice that λ = 2.5 is close to 2.5625, that is,
λ ≈ ∆f = f (new max. pt) − f (old max. pt) .
Finally, note that solving the equation ∇f (x, y) = λ∇g(x, y) means having to solve a
system of two (possibly nonlinear) equations in three unknowns, which as we have
seen before, may not be possible to do. And the 3-variable case can get even more
complicated. All of this somewhat restricts the usefulness of Lagrange’s method to
relatively simple functions. Luckily there are many numerical methods for solving
constrained optimization problems, though we will not discuss them here.
13

¨
Exercises
A
1. Find the constrained maxima and minima of f (x, y) = 2x + y given that x
2
+ y
2
= 4.
2. Find the constrained maxima and minima of f (x, y) = xy given that x
2
+ 3y
2
= 6.
3. Find the points on the circle x
2
+y
2
= 100 which are closest to and farthest from the
point (2, 3).
B
4. Find the constrained maxima and minima of f (x, y, z) = x + y
2
+ 2z given that 4x
2
+
9y
2
− 36z
2
= 36.
5. Find the volume of the largest rectangular parallelepiped that can be inscribed in
the ellipsoid
x
2
a
2
+
y
2
b
2
+
z
2
c
2
= 1 .
13
See BAZARAA, SHERALI and SHETTY.
3 Multiple Integrals
3.1 Double Integrals
In single-variable calculus, differentiation and integration are thought of as inverse
operations. For instance, to integrate a function f (x) it is necessary to find the an-
tiderivative of f , that is, another function F(x) whose derivative is f (x). Is there a
similar way of defining integration of real-valued functions of two or more variables?
The answer is yes, as we will see shortly. Recall also that the definite integral of a
nonnegative function f (x) ≥ 0 represented the area “under” the curve y = f (x). As
we will now see, the double integral of a nonnegative real-valued function f (x, y) ≥ 0
represents the volume “under” the surface z = f (x, y).
Let f (x, y) be a continuous function such that f (x, y) ≥ 0 for all (x, y) on the rectangle
R = {(x, y) : a ≤ x ≤ b, c ≤ y ≤ d} in
2
. We will often write this as R = [a, b] × [c, d].
For any number x∗ in the interval [a, b], slice the surface z = f (x, y) with the plane
x = x∗ parallel to the yz-plane. Then the trace of the surface in that plane is the curve
f (x∗, y), where x∗ is fixed and only y varies. The area A under that curve (i.e. the area
of the region between the curve and the xy-plane) as y varies over the interval [c, d]
then depends only on the value of x∗. So using the variable x instead of x∗, let A(x) be
that area (see Figure 3.1.1).
y
z
x
0 A(x)
R
a
x
b
c d
z = f (x, y)
Figure 3.1.1 The area A(x) varies with x
Then A(x) =

d
c
f (x, y) dy since we are treating x as fixed, and only y varies. This
makes sense since for a fixed x the function f (x, y) is a continuous function of y over
the interval [c, d], so we know that the area under the curve is the definite integral.
101
102 CHAPTER 3. MULTIPLE INTEGRALS
The area A(x) is a function of x, so by the “slice” or cross-section method from single-
variable calculus we know that the volume V of the solid under the surface z = f (x, y)
but above the xy-plane over the rectangle R is the integral over [a, b] of that cross-
sectional area A(x):
V =

b
a
A(x) dx =

b
a
,
d
c
f (x, y) dy
¸
dx (3.1)
We will always refer to this volume as “the volume under the surface”. The above
expression uses what are called iterated integrals. First the function f (x, y) is inte-
grated as a function of y, treating the variable x as a constant (this is called integrat-
ing with respect to y). That is what occurs in the “inner” integral between the square
brackets in equation (3.1). This is the first iterated integral. Once that integration
is performed, the result is then an expression involving only x, which can then be
integrated with respect to x. That is what occurs in the “outer” integral above (the sec-
ond iterated integral). The final result is then a number (the volume). This process
of going through two iterations of integrals is called double integration, and the last
expression in equation (3.1) is called a double integral.
Notice that integrating f (x, y) with respect to y is the inverse operation of taking the
partial derivative of f (x, y) with respect to y. Also, we could just as easily have taken
the area of cross-sections under the surface which were parallel to the xz-plane, which
would then depend only on the variable y, so that the volume V would be
V =

d
c
,
b
a
f (x, y) dx
¸
dy . (3.2)
It turns out that in general
1
the order of the iterated integrals does not matter. Also,
we will usually discard the brackets and simply write
V =

d
c

b
a
f (x, y) dx dy , (3.3)
where it is understood that the fact that dx is written before dy means that the func-
tion f (x, y) is first integrated with respect to x using the “inner” limits of integration a
and b, and then the resulting function is integrated with respect to y using the “outer”
limits of integration c and d. This order of integration can be changed if it is more
convenient.
Example 3.1. Find the volume V under the plane z = 8x + 6y over the rectangle R =
[0, 1] × [0, 2].
1
due to Fubini’s Theorem. See Ch. 18 in TAYLOR and MANN.
3.1 Double Integrals 103
Solution: We see that f (x, y) = 8x + 6y ≥ 0 for 0 ≤ x ≤ 1 and 0 ≤ y ≤ 2, so:
V =

2
0

1
0
(8x + 6y) dx dy
=

2
0

4x
2
+ 6xy

x=1
x=0

dy
=

2
0
(4 + 6y) dy
= 4y + 3y
2

2
0
= 20
Suppose we had switched the order of integration. We can verify that we still get the
V =

1
0

2
0
(8x + 6y) dy dx
=

1
0

8xy + 3y
2

y=2
y=0

dx
=

1
0
(16x + 12) dx
= 8x
2
+ 12x

1
0
= 20
Example 3.2. Find the volume V under the surface z = e
x+y
over the rectangle R =
[2, 3] × [1, 2].
Solution: We know that f (x, y) = e
x+y
> 0 for all (x, y), so
V =

2
1

3
2
e
x+y
dx dy
=

2
1

e
x+y

x=3
x=2

dy
=

2
1
(e
y+3
− e
y+2
) dy
= e
y+3
− e
y+2

2
1
= e
5
− e
4
− (e
4
− e
3
) = e
5
− 2e
4
+ e
3
Recall that for a general function f (x), the integral

b
a
f (x) dx represents the differ-
ence of the area below the curve y = f (x) but above the x-axis when f (x) ≥ 0, and the
104 CHAPTER 3. MULTIPLE INTEGRALS
area above the curve but below the x-axis when f (x) ≤ 0. Similarly, the double inte-
gral of any continuous function f (x, y) represents the difference of the volume below
the surface z = f (x, y) but above the xy-plane when f (x, y) ≥ 0, and the volume above
the surface but below the xy-plane when f (x, y) ≤ 0. Thus, our method of double inte-
gration by means of iterated integrals can be used to evaluate the double integral of
any continuous function over a rectangle, regardless of whether f (x, y) ≥ 0 or not.
Example 3.3. Evaluate

0

π
0
sin(x + y) dx dy.
Solution: Note that f (x, y) = sin(x + y) is both positive and negative over the rectangle
[0, π] × [0, 2π]. We can still evaluate the double integral:

0

π
0
sin(x + y) dx dy =

0

−cos(x + y)

x=π
x=0

dy
=

0
(−cos(y + π) + cos y) dy
= −sin(y + π) + sin y

0
= −sin 3π + sin 2π − (−sin π + sin 0)
= 0

¨
Exercises
A
For Exercises 1-4, find the volume under the surface z = f (x, y) over the rectangle R.
1. f (x, y) = 4xy, R = [0, 1] × [0, 1] 2. f (x, y) = e
x+y
, R = [0, 1] × [−1, 1]
3. f (x, y) = x
3
+ y
2
, R = [0, 1] × [0, 1] 4. f (x, y) = x
4
+ xy + y
3
, R = [1, 2] × [0, 2]
For Exercises 5-12, evaluate the given double integral.
5.

1
0

2
1
(1 − y)x
2
dx dy 6.

1
0

2
0
x(x + y) dx dy
7.

2
0

1
0
(x + 2) dx dy 8.

2
−1

1
−1
x(xy + sin x) dx dy
9.

π/2
0

1
0
xy cos(x
2
y) dx dy 10.

π
0

π/2
0
sin x cos(y − π) dx dy
11.

2
0

4
1
xy dx dy 12.

1
−1

2
−1
1 dx dy
13. Let M be a constant. Show that

d
c

b
a
M dx dy = M(d − c)(b − a).
3.2 Double Integrals Over a General Region 105
3.2 Double Integrals Over a General Region
In the previous section we got an idea of what a double integral over a rectangle
represents. We can now define the double integral of a real-valued function f (x, y)
over more general regions in
2
.
Suppose that we have a region R in the xy-plane that is bounded on the left by
the vertical line x = a, bounded on the right by the vertical line x = b (where a < b),
bounded belowby a curve y = g
1
(x), and bounded above by a curve y = g
2
(x), as in Figure
3.2.1(a). We will assume that g
1
(x) and g
2
(x) do not intersect on the open interval (a, b)
(they could intersect at the endpoints x = a and x = b, though).
x
y
0
y = g
2
(x)
y = g
1
(x)
R
a
b
(a) Vertical slice:

b
a

g
2
(x)
g
1
(x)
f (x, y) dy dx
x
y
0
x = h
1
(y)
x = h
2
(y)
R
c
d
(b) Horizontal slice:

d
c

h
2
(y)
h
1
(y)
f (x, y) dx dy
Figure 3.2.1 Double integral over a nonrectangular region R
Then using the slice method from the previous section, the double integral of a
real-valued function f (x, y) over the region R, denoted by

R
f (x, y) dA, is given by

R
f (x, y) dA =

b
a
,
g
2
(x)
g
1
(x)
f (x, y) dy
¸
dx (3.4)
This means that we take vertical slices in the region R between the curves y = g
1
(x)
and y = g
2
(x). The symbol dA is sometimes called an area element or infinitesimal,
with the A signifying area. Note that f (x, y) is first integrated with respect to y, with
functions of x as the limits of integration. This makes sense since the result of the
first iterated integral will have to be a function of x alone, which then allows us to
take the second iterated integral with respect to x.
Similarly, if we have a region R in the xy-plane that is bounded on the left by a curve
x = h
1
(y), bounded on the right by a curve x = h
2
(y), bounded below by the horizontal
106 CHAPTER 3. MULTIPLE INTEGRALS
line y = c, and bounded above by the horizontal line y = d (where c < d), as in Figure
3.2.1(b) (assuming that h
1
(y) and h
2
(y) do not intersect on the open interval (c, d)), then
taking horizontal slices gives

R
f (x, y) dA =

d
c
,
h
2
(y)
h
1
(y)
f (x, y) dx
¸
dy (3.5)
Notice that these definitions include the case when the region R is a rectangle.
Also, if f (x, y) ≥ 0 for all (x, y) in the region R, then

R
f (x, y) dA is the volume under the
surface z = f (x, y) over the region R.
Example 3.4. Find the volume V under the plane z = 8x+6y over the region R = {(x, y) :
0 ≤ x ≤ 1, 0 ≤ y ≤ 2x
2
}.
x
y
0
y = 2x
2
R
1
Figure 3.2.2
Solution: The region R is shown in Figure 3.2.2. Using vertical slices
we get:
V =

R
(8x + 6y) dA
=

1
0
,
¸
¸
¸
¸
¸
¸

2x
2
0
(8x + 6y) dy
¸
¸
¸
¸
¸
¸
¸
dx
=

1
0
¸
8xy + 3y
2

y=2x
2
y=0

dx
=

1
0
(16x
3
+ 12x
4
) dx
= 4x
4
+
12
5
x
5

1
0
= 4 +
12
5
=
32
5
= 6.4
x
y
0
2
x =

y/2
R
1
Figure 3.2.3
We get the same answer using horizontal slices (see Figure 3.2.3):
V =

R
(8x + 6y) dA
=

2
0
,
¸
¸
¸
¸
¸
¸

1

y/2
(8x + 6y) dx
¸
¸
¸
¸
¸
¸
¸
dy
=

2
0
¸
4x
2
+ 6xy

x=1
x=

y/2

dy
=

2
0
(4 + 6y − (2y +
6

2
y

y )) dy =

2
0
(4 + 4y − 3

2y
3/2
) dy
= 4y + 2y
2

6

2
5
y
5/2

2
0
= 8 + 8 −
6

2

32
5
= 16 −
48
5
=
32
5
= 6.4
3.2 Double Integrals Over a General Region 107
Example 3.5. Find the volume V of the solid bounded by the three coordinate planes
and the plane 2x + y + 4z = 4.
y
z
x
0
(0, 4, 0)
(0, 0, 1)
(2, 0, 0)
2x + y + 4z = 4
(a)
x
y
0
y = −2x + 4
R
2
4
(b)
Figure 3.2.4
Solution: The solid is shown in Figure 3.2.4(a) with a typical vertical slice. The volume
V is given by

R
f (x, y) dA, where f (x, y) = z =
1
4
(4 − 2x − y) and the region R, shown in
Figure 3.2.4(b), is R = {(x, y) : 0 ≤ x ≤ 2, 0 ≤ y ≤ −2x +4}. Using vertical slices in R gives
V =

R
1
4
(4 − 2x − y) dA
=

2
0
,
−2x+4
0
1
4
(4 − 2x − y) dy
¸
dx
=

2
0

1
8
(4 − 2x − y)
2

y=−2x+4
y=0

dx
=

2
0
1
8
(4 − 2x)
2
dx
= −
1
48
(4 − 2x)
3

2
0
=
64
48
=
4
3
For a general region R, which may not be one of the types of regions we have consid-
ered so far, the double integral

R
f (x, y) dA is defined as follows. Assume that f (x, y)
is a nonnegative real-valued function and that R is a bounded region in
2
, so it can
be enclosed in some rectangle [a, b] × [c, d]. Then divide that rectangle into a grid of
subrectangles. Only consider the subrectangles that are enclosed completely within
the region R, as shown by the shaded subrectangles in Figure 3.2.5(a). In any such
subrectangle [x
i
, x
i+1
] ×[y
j
, y
j+1
], pick a point (x
i∗
, y
j∗
). Then the volume under the surface
z = f (x, y) over that subrectangle is approximately f (x
i∗
, y
j∗
) ∆x
i
∆y
j
, where ∆x
i
= x
i+1
− x
i
,
108 CHAPTER 3. MULTIPLE INTEGRALS
∆y
j
= y
j+1
−y
j
, and f (x
i∗
, y
j∗
) is the height and ∆x
i
∆y
j
is the base area of a parallelepiped,
as shown in Figure 3.2.5(b). Then the total volume under the surface is approximately
the sum of the volumes of all such parallelepipeds, namely
¸
j
¸
i
f (x
i∗
, y
j∗
) ∆x
i
∆y
j
, (3.6)
where the summation occurs over the indices of the subrectangles inside R. If we
take smaller and smaller subrectangles, so that the length of the largest diagonal of
the subrectangles goes to 0, then the subrectangles begin to fill more and more of
the region R, and so the above sum approaches the actual volume under the surface
z = f (x, y) over the region R. We then define

R
f (x, y) dA as the limit of that double
summation (the limit is taken over all subdivisions of the rectangle [a, b] ×[c, d] as the
largest diagonal of the subrectangles goes to 0).
x
y
0
d
c
y
j
y
j+1
a
b
x
i
x
i+1
(x
i∗
, y
j∗
)
(a) Subrectangles inside the region R
y
z
x
0
R
x
i
x
i+1
y
j
y
j+1
z = f (x, y)
∆y
j
∆x
i
(x
i∗
, y
j∗
)
f (x
i∗
, y
j∗
)
(b) Parallelepiped over a subrectan-
gle, with volume f (x
i∗
, y
j∗
) ∆x
i
∆y
j
Figure 3.2.5 Double integral over a general region R
A similar definition can be made for a function f (x, y) that is not necessarily always
nonnegative: just replace each mention of volume by the negative volume in the de-
scription above when f (x, y) < 0. In the case of a region of the type shown in Figure
3.2.1, using the definition of the Riemann integral from single-variable calculus, our
definition of

R
f (x, y) dA reduces to a sequence of two iterated integrals.
Finally, the region R does not have to be bounded. We can evaluate improper double
integrals (i.e. over an unbounded region, or over a region which contains points where
the function f (x, y) is not defined) as a sequence of iterated improper single-variable
integrals.
3.2 Double Integrals Over a General Region 109
Example 3.6. Evaluate

1

1/x
2
0
2y dy dx.
Solution:

1

1/x
2
0
2y dy dx =

1
¸
y
2

y=1/x
2
y=0

dx
=

1
x
−4
dx = −
1
3
x
−3

1
= 0 − (−
1
3
) =
1
3

¨
Exercises
A
For Exercises 1-6, evaluate the given double integral.
1.

1
0

1

x
24x
2
y dy dx
2.

π
0

y
0
sin x dx dy
3.

2
1

ln x
0
4x dy dx 4.

2
0

2y
0
e
y
2
dx dy
5.

π/2
0

y
0
cos x sin y dx dy
6.

0

0
xye
−(x
2
+y
2
)
dx dy
7.

2
0

y
0
1 dx dy
8.

1
0

x
2
0
2 dy dx
9. Find the volume V of the solid bounded by the three coordinate planes and the
plane x + y + z = 1.
10. Find the volume V of the solid bounded by the three coordinate planes and the
plane 3x + 2y + 5z = 6.
B
11. Explain why the double integral

R
1 dA gives the area of the region R. For sim-
plicity, you can assume that R is a region of the type shown in Figure 3.2.1(a).
C
b
c
a
Figure 3.2.6
12. Prove that the volume of a tetrahedron with mutually per-
pendicular adjacent sides of lengths a, b, and c, as in Figure
3.2.6, is
abc
6
. (Hint: Mimic Example 3.5, and recall from
Section 1.5 how three noncollinear points determine a plane.)
13. Show how Exercise 12 can be used to solve Exercise 10.
110 CHAPTER 3. MULTIPLE INTEGRALS
3.3 Triple Integrals
Our definition of a double integral of a real-valued function f (x, y) over a region R in

2
can be extended to define a triple integral of a real-valued function f (x, y, z) over
a solid S in
3
. We simply proceed as before: the solid S can be enclosed in some
rectangular parallelepiped, which is then divided into subparallelepipeds. In each
subparallelepiped inside S , with sides of lengths ∆x, ∆y and ∆z, pick a point (x

, y

, z

).
Then define the triple integral of f (x, y, z) over S , denoted by

S
f (x, y, z) dV, by

S
f (x, y, z) dV = lim
¸ ¸ ¸
f (x

, y

, z

) ∆x ∆y ∆z , (3.7)
where the limit is over all divisions of the rectangular parallelepiped enclosing S into
subparallelepipeds whose largest diagonal is going to 0, and the triple summation
is over all the subparallelepipeds inside S . It can be shown that this limit does not
depend on the choice of the rectangular parallelepiped enclosing S . The symbol dV is
often called the volume element.
Physically, what does the triple integral represent? We saw that a double integral
could be thought of as the volume under a two-dimensional surface. It turns out that
the triple integral simply generalizes this idea: it can be thought of as representing
the hypervolume under a three-dimensional hypersurface w = f (x, y, z) whose graph
lies in
4
. In general, the word “volume” is often used as a general term to signify the
same concept for any n-dimensional object (e.g. length in
1
, area in
2
). It may be
hard to get a grasp on the concept of the “volume” of a four-dimensional object, but at
least we now know how to calculate that volume!
In the case where S is a rectangular parallelepiped [x
1
, x
2
] × [y
1
, y
2
] × [z
1
, z
2
], that is,
S = {(x, y, z) : x
1
≤ x ≤ x
2
, y
1
≤ y ≤ y
2
, z
1
≤ z ≤ z
2
}, the triple integral is a sequence of
three iterated integrals, namely

S
f (x, y, z) dV =

z
2
z
1

y
2
y
1

x
2
x
1
f (x, y, z) dx dy dz , (3.8)
where the order of integration does not matter. This is the simplest case.
A more complicated case is where S is a solid which is bounded below by a surface
z = g
1
(x, y), bounded above by a surface z = g
2
(x, y), y is bounded between two curves
h
1
(x) and h
2
(x), and x varies between a and b. Then

S
f (x, y, z) dV =

b
a

h
2
(x)
h
1
(x)

g
2
(x,y)
g
1
(x,y)
f (x, y, z) dz dy dx . (3.9)
Notice in this case that the first iterated integral will result in a function of x and y
(since its limits of integration are functions of x and y), which then leaves you with a
3.3 Triple Integrals 111
double integral of a type that we learned how to evaluate in Section 3.2. There are, of
course, many variations on this case (for example, changing the roles of the variables
x, y, z), so as you can probably tell, triple integrals can be quite tricky. At this point,
just learning how to evaluate a triple integral, regardless of what it represents, is the
most important thing. We will see some other ways in which triple integrals are used
later in the text.
Example 3.7. Evaluate

3
0

2
0

1
0
(xy + z) dx dy dz.
Solution:

3
0

2
0

1
0
(xy + z) dx dy dz =

3
0

2
0

1
2
x
2
y + xz

x=1
x=0

dy dz
=

3
0

2
0

1
2
y + z

dy dz
=

3
0

1
4
y
2
+ yz

y=2
y=0

dz
=

3
0
(1 + 2z) dz
= z + z
2

3
0
= 12
Example 3.8. Evaluate

1
0

1−x
0

2−x−y
0
(x + y + z) dz dy dx.
Solution:

1
0

1−x
0

2−x−y
0
(x + y + z) dz dy dx =

1
0

1−x
0

(x + y)z +
1
2
z
2

z=2−x−y
z=0

dy dx
=

1
0

1−x
0

(x + y)(2 − x − y) +
1
2
(2 − x − y)
2

dy dx
=

1
0

1−x
0

2 −
1
2
x
2
− xy −
1
2
y
2

dy dx
=

1
0

2y −
1
2
x
2
y − xy −
1
2
xy
2

1
6
y
3

y=1−x
y=0

dx
=

1
0

11
6
− 2x +
1
6
x
3

dx
=
11
6
x − x
2
+
1
24
x
4

1
0
=
7
8
112 CHAPTER 3. MULTIPLE INTEGRALS
Note that the volume V of a solid in
3
is given by
V =

S
1 dV . (3.10)
Since the function being integrated is the constant 1, then the above triple integral
reduces to a double integral of the types that we considered in the previous section
if the solid is bounded above by some surface z = f (x, y) and bounded below by the
xy-plane z = 0. There are many other possibilities. For example, the solid could be
bounded below and above by surfaces z = g
1
(x, y) and z = g
2
(x, y), respectively, with y
bounded between two curves h
1
(x) and h
2
(x), and x varies between a and b. Then
V =

S
1 dV =

b
a

h
2
(x)
h
1
(x)

g
2
(x,y)
g
1
(x,y)
1 dz dy dx =

b
a

h
2
(x)
h
1
(x)
(g
2
(x, y) − g
1
(x, y)) dy dx
just like in equation (3.9). See Exercise 10 for an example.

¨
Exercises
A
For Exercises 1-8, evaluate the given triple integral.
1.

3
0

2
0

1
0
xyz dx dy dz 2.

1
0

x
0

y
0
xyz dz dy dx
3.

π
0

x
0

xy
0
x
2
sin z dz dy dx
4.

1
0

z
0

y
0
ze
y
2
dx dy dz
5.

e
1

y
0

1/y
0
x
2
z dx dz dy
6.

2
1

y
2
0

z
2
0
yz dx dz dy
7.

2
1

4
2

3
0
1 dx dy dz 8.

1
0

1−x
0

1−x−y
0
1 dz dy dx
9. Let M be a constant. Show that

z
2
z
1

y
2
y
1

x
2
x
1
M dx dy dz = M(z
2
− z
1
)(y
2
− y
1
)(x
2
− x
1
).
B
10. Find the volume V of the solid S bounded by the three coordinate planes, bounded
above by the plane x + y + z = 2, and bounded below by the plane z = x + y.
C
11. Show that

b
a

z
a

y
a
f (x) dx dy dz =

b
a
(b−x)
2
2
f (x) dx. (Hint: Think of how changing
the order of integration in the triple integral changes the limits of integration.)
3.4 Numerical Approximation of Multiple Integrals 113
3.4 Numerical Approximation of Multiple Integrals
As you have seen, calculating multiple integrals is tricky even for simple functions
and regions. For complicated functions, it may not be possible to evaluate one of the
iterated integrals in a simple closed form. Luckily there are numerical methods for
approximating the value of a multiple integral. The method we will discuss is called
the Monte Carlo method. The idea behind it is based on the concept of the average
value of a function, which you learned in single-variable calculus. Recall that for a
continuous function f (x), the average value
¯
f of f over an interval [a, b] is defined as
¯
f =
1
b − a

b
a
f (x) dx . (3.11)
The quantity b − a is the length of the interval [a, b], which can be thought of as the
“volume” of the interval. Applying the same reasoning to functions of two or three
variables, we define the average value of f (x, y) over a region R to be
¯
f =
1
A(R)

R
f (x, y) dA , (3.12)
where A(R) is the area of the region R, and we define the average value of f (x, y, z)
over a solid S to be
¯
f =
1
V(S )

S
f (x, y, z) dV , (3.13)
where V(S ) is the volume of the solid S . Thus, for example, we have

R
f (x, y) dA = A(R)
¯
f . (3.14)
The average value of f (x, y) over R can be thought of as representing the sum of all the
values of f divided by the number of points in R. Unfortunately there are an infinite
number (in fact, uncountably many) points in any region, i.e. they can not be listed in
a discrete sequence. But what if we took a very large number N of random points in
the region R (which can be generated by a computer) and then took the average of the
values of f for those points, and used that average as the value of
¯
f ? This is exactly
what the Monte Carlo method does. So in formula (3.14) the approximation we get is

R
f (x, y) dA ≈ A(R)
¯
f ± A(R)

f
2
− (
¯
f )
2
N
, (3.15)
where
¯
f =
¸
N
i=1
f (x
i
, y
i
)
N
and f
2
=
¸
N
i=1
( f (x
i
, y
i
))
2
N
, (3.16)
114 CHAPTER 3. MULTIPLE INTEGRALS
with the sums taken over the N random points (x
1
, y
1
), . . ., (x
N
, y
N
). The ± “error term”
in formula (3.15) does not really provide hard bounds on the approximation. It repre-
sents a single standard deviation from the expected value of the integral. That is, it
provides a likely bound on the error. Due to its use of random points, the Monte Carlo
method is an example of a probabilistic method (as opposed to deterministic methods
such as Newton’s method, which use a specific formula for generating points).
For example, we can use formula (3.15) to approximate the volume V under the
plane z = 8x + 6y over the rectangle R = [0, 1] × [0, 2]. In Example 3.1 in Section 3.1,
we showed that the actual volume is 20. Below is a code listing (montecarlo.java) for
a Java program that calculates the volume, using a number of points N that is passed
on the command line as a parameter.
//Program t o approximate the double i nt egral of f ( x , y)=8x+6y
//over the rect angl e [ 0 , 1] x [ 0 , 2] .
public class montecarlo {
public static void main( Stri ng [ ] args ) {
//Get the number N of random poi nt s as a command−l i ne parameter
int N = Integer . parseInt ( args [ 0 ] ) ;
double x = 0; //x−coordi nat e of a random poi nt
double y = 0; //y−coordi nat e of a random poi nt
double f = 0. 0; //Value of f at a random poi nt
double mf = 0. 0; //Mean of the values of f
double mf2 = 0. 0; //Mean of the values of f ^2
for ( int i =0; i <N; i ++) { //Get the random coordi nat es
x = Math. random( ) ; //x i s between 0 and 1
y = 2 ∗ Math. random( ) ; //y i s between 0 and 2
f = 8∗x + 6∗y ; //Value of the f unct i on
mf = mf + f ; //Add t o the sum of the f values
mf2 = mf2 + f ∗ f ; //Add t o the sum of the f ^2 values
}
mf = mf /N; //Compute the mean of the f values
mf2 = mf2/N; //Compute the mean of the f ^2 values
System. out . pri nt l n ( "N = " + N + " : i nt egral = " + vol ( ) ∗ mf + " +/ − "
+ vol ( ) ∗Math. sqrt ( ( mf2 − Math. pow( mf , 2 ) ) / N) ) ; //Print the r e s ul t
}
//The volume of the rect angl e [ 0 , 1] x [ 0 , 2]
public static double vol ( ) {
return 1∗2;
}
}
Listing 3.1 Program listing for montecarlo.java
The results of running this program with various numbers of random points (e.g.
java montecarlo 100) are shown below:
3.4 Numerical Approximation of Multiple Integrals 115
N = 10: 19.36543087722646 +/- 2.7346060413546147
N = 100: 21.334419561385353 +/- 0.7547037194998519
N = 1000: 19.807662237526227 +/- 0.26701709691370235
N = 10000: 20.080975812043256 +/- 0.08378816229769506
N = 100000: 20.009403854556716 +/- 0.026346782289498317
N = 1000000: 20.000866994982314 +/- 0.008321168748642816
As you can see, the approximation is fairly good. As N → ∞, it can be shown that the
Monte Carlo approximation converges to the actual volume (on the order of O(

N), in
computational complexity terminology).
In the above example the region R was a rectangle. To use the Monte Carlo method
for a nonrectangular (bounded) region R, only a slight modification is needed. Pick a
rectangle
˜
R that encloses R, and generate random points in that rectangle as before.
Then use those points in the calculation of
¯
f only if they are inside R. There is no need
to calculate the area of R for formula (3.15) in this case, since the exclusion of points
not inside R allows you to use the area of the rectangle
˜
For instance, in Example 3.4 we showed that the volume under the surface z = 8x+6y
over the nonrectangular region R = {(x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 2x
2
} is 6.4. Since the
rectangle
˜
R = [0, 1] × [0, 2] contains R, we can use the same program as before, with
the only change being a check to see if y < 2x
2
for a random point (x, y) in [0, 1] × [0, 2].
Listing 3.2 below contains the code (montecarlo2.java):
//Program t o approximate the double i nt egral of f ( x , y)=8x+6y over the
//regi on bounded by x=0, x=1, y=0, and y=2x^2
public class montecarlo2 {
public static void main( Stri ng [ ] args ) {
//Get the number N of random poi nt s as a command−l i ne parameter
int N = Integer . parseInt ( args [ 0 ] ) ;
double x = 0; //x−coordi nat e of a random poi nt
double y = 0; //y−coordi nat e of a random poi nt
double f = 0. 0; //Value of f at a random poi nt
double mf = 0. 0; //Mean of the values of f
double mf2 = 0. 0; //Mean of the values of f ^2
for ( int i =0; i <N; i ++) { //Get the random coordi nat es
x = Math. random( ) ; //x i s between 0 and 1
y = 2 ∗ Math. random( ) ; //y i s between 0 and 2
i f ( y < 2∗Math. pow( x , 2 ) ) { //The poi nt i s in the regi on
f = 8∗x + 6∗y ; //Value of the f unct i on
mf = mf + f ; //Add t o the sum of the f values
mf2 = mf2 + f ∗ f ; //Add t o the sum of the f ^2 values
}
}
mf = mf /N; //Compute the mean of the f values
mf2 = mf2/N; //Compute the mean of the f ^2 values
System. out . pri nt l n ( "N = " + N + " : i nt egral = " + vol ( ) ∗ mf +
116 CHAPTER 3. MULTIPLE INTEGRALS
" +/ − " + vol ( ) ∗Math. sqrt ( ( mf2 − Math. pow( mf , 2 ) ) / N) ) ;
}
//The volume of the rect angl e [ 0 , 1] x [ 0 , 2]
public static double vol ( ) {
return 1∗2;
}
}
Listing 3.2 Program listing for montecarlo2.java
The results of running the program with various numbers of random points (e.g.
java montecarlo2 1000) are shown below:
N = 10: integral = 6.95747529014894 +/- 2.9185131565120592
N = 100: integral = 6.3149056229650355 +/- 0.9549009662159909
N = 1000: integral = 6.477032813858756 +/- 0.31916837260973624
N = 10000: integral = 6.349975080015089 +/- 0.10040086346895105
N = 100000: integral = 6.440184132811864 +/- 0.03200476870881392
N = 1000000: integral = 6.417050897922222 +/- 0.01009454409789472
To use the Monte Carlo method to evaluate triple integrals, you will need to gen-
erate random triples (x, y, z) in a parallelepiped, instead of random pairs (x, y) in a
rectangle, and use the volume of the parallelepiped instead of the area of a rectan-
gle in formula (3.15) (see Exercise 2). For a more detailed discussion of numerical
integration methods, see PRESS et al.

¨
Exercises
C
1. Write a program that uses the Monte Carlo method to approximate the double
integral

R
e
xy
dA, where R = [0, 1] × [0, 1]. Show the program output for N =
10, 100, 1000, 10000, 100000 and 1000000 random points.
2. Write a program that uses the Monte Carlo method to approximate the triple in-
tegral

S
e
xyz
dV, where S = [0, 1] × [0, 1] × [0, 1]. Show the program output for
N = 10, 100, 1000, 10000, 100000 and 1000000 random points.
3. Repeat Exercise 1 with the region R = {(x, y) : −1 ≤ x ≤ 1, 0 ≤ y ≤ x
2
}.
4. Repeat Exercise 2 with the solid S = {(x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 − x − y}.
5. Use the Monte Carlo method to approximate the volume of a sphere of radius 1.
6. Use the Monte Carlo method to approximate the volume of the ellipsoid
x
2
9
+
y
2
4
+
z
2
1
=
1.
3.5 Change of Variables in Multiple Integrals 117
3.5 Change of Variables in Multiple Integrals
Given the difficulty of evaluating multiple integrals, the reader may be wondering if
it is possible to simplify those integrals using a suitable substitution for the variables.
The answer is yes, though it is a bit more complicated than the substitution method
which you learned in single-variable calculus.
Recall that if you are given, for example, the definite integral

2
1
x
3

x
2
− 1 dx ,
then you would make the substitution
u = x
2
− 1 ⇒ x
2
= u + 1
du = 2x dx
which changes the limits of integration
x = 1 ⇒ u = 0
x = 2 ⇒ u = 3
so that we get

2
1
x
3

x
2
− 1 dx =

2
1
1
2
x
2
· 2x

x
2
− 1 dx
=

3
0
1
2
(u + 1)

u du
=
1
2

3
0

u
3/2
+ u
1/2

du , which can be easily integrated to give
=
14

3
5
.
Let us take a different look at what happened when we did that substitution, which
will give some motivation for how substitution works in multiple integrals. First, we
let u = x
2
− 1. On the interval of integration [1, 2], the function x → x
2
− 1 is strictly
increasing (and maps [1, 2] onto [0, 3]) and hence has an inverse function (defined on
the interval [0, 3]). That is, on [0, 3] we can define x as a function of u, namely
x = g(u) =

u + 1 .
Then substituting that expression for x into the function f (x) = x
3

x
2
− 1 gives
f (x) = f (g(u)) = (u + 1)
3/2

u ,
118 CHAPTER 3. MULTIPLE INTEGRALS
and we see that
dx
du
= g

(u) ⇒ dx = g

(u) du
dx =
1
2
(u + 1)
−1/2
du ,
so since
g(0) = 1 ⇒ 0 = g
−1
(1)
g(3) = 2 ⇒ 3 = g
−1
(2)
then performing the substitution as we did earlier gives

2
1
f (x) dx =

2
1
x
3

x
2
− 1 dx
=

3
0
1
2
(u + 1)

u du , which can be written as
=

3
0
(u + 1)
3/2

u ·
1
2
(u + 1)
−1/2
du , which means

2
1
f (x) dx =

g
−1
(2)
g
−1
(1)
f (g(u)) g

(u) du .
In general, if x = g(u) is a one-to-one, differentiable function from an interval [c, d]
(which you can think of as being on the “u-axis”) onto an interval [a, b] (on the x-axis),
which means that g

(u) 0 on the interval (c, d), so that a = g(c) and b = g(d), then
c = g
−1
(a) and d = g
−1
(b), and

b
a
f (x) dx =

g
−1
(b)
g
−1
(a)
f (g(u)) g

(u) du . (3.17)
This is called the change of variable formula for integrals of single-variable functions,
and it is what you were implicitly using when doing integration by substitution. This
formula turns out to be a special case of a more general formula which can be used
to evaluate multiple integrals. We will state the formulas for double and triple inte-
grals involving real-valued functions of two and three variables, respectively. We will
assume that all the functions involved are continuously differentiable and that the re-
gions and solids involved all have “reasonable” boundaries. The proof of the following
theorem is beyond the scope of the text.
2
2
See TAYLOR and MANN, § 15.32 and § 15.62 for all the details.
3.5 Change of Variables in Multiple Integrals 119
Theorem 3.1. Change of Variables Formula for Multiple Integrals
Let x = x(u, v) and y = y(u, v) define a one-to-one mapping of a region R

in the uv-plane
onto a region R in the xy-plane such that the determinant
J(u, v) =

∂x
∂u
∂x
∂v
∂y
∂u
∂y
∂v

(3.18)
is never 0 in R

. Then

R
f (x, y) dA(x, y) =

R

f (x(u, v), y(u, v)) | J(u, v)| dA(u, v) . (3.19)
We use the notation dA(x, y) and dA(u, v) to denote the area element in the (x, y) and
(u, v) coordinates, respectively.
Similarly, if x = x(u, v, w), y = y(u, v, w) and z = z(u, v, w) define a one-to-one mapping
of a solid S

in uvw-space onto a solid S in xyz-space such that the determinant
J(u, v, w) =

∂x
∂u
∂x
∂v
∂x
∂w
∂y
∂u
∂y
∂v
∂y
∂w
∂z
∂u
∂z
∂v
∂z
∂w

(3.20)
is never 0 in S

, then

S
f (x, y, z) dV(x, y, z) =

S

f (x(u, v, w), y(u, v, w), z(u, v, w)) | J(u, v, w)| dV(u, v, w) .
(3.21)
The determinant J(u, v) in formula (3.18) is called the Jacobian of x and y with
respect to u and v, and is sometimes written as
J(u, v) =
∂(x, y)
∂(u, v)
. (3.22)
Similarly, the Jacobian J(u, v, w) of three variables is sometimes written as
J(u, v, w) =
∂(x, y, z)
∂(u, v, w)
. (3.23)
Notice that formula (3.19) is saying that dA(x, y) = | J(u, v)| dA(u, v), which you can think
of as a two-variable version of the relation dx = g

(u) du in the single-variable case.
The following example shows how the change of variables formula is used.
120 CHAPTER 3. MULTIPLE INTEGRALS
Example 3.9. Evaluate

R
e
x−y
x+y
dA, where R = {(x, y) : x ≥ 0, y ≥ 0, x + y ≤ 1}.
Solution: First, note that evaluating this double integral without using substitution
is probably impossible, at least in a closed form. By looking at the numerator and
denominator of the exponent of e, we will try the substitution u = x −y and v = x +y. To
use the change of variables formula (3.19), we need to write both x and y in terms of u
and v. So solving for x and y gives x =
1
2
(u + v) and y =
1
2
(v − u). In Figure 3.5.1 below,
we see how the mapping x = x(u, v) =
1
2
(u + v), y = y(u, v) =
1
2
(v − u) maps the region R

onto R in a one-to-one manner.
x
y
0
x + y = 1
1
1
R
u
v
0
1
−1 1
R

u = v u = −v
x =
1
2
(u + v)
y =
1
2
(v − u)
Figure 3.5.1 The regions R and R

Now we see that
J(u, v) =

∂x
∂u
∂x
∂v
∂y
∂u
∂y
∂v

=

1
2
1
2

1
2
1
2

=
1
2
⇒ | J(u, v)| =

1
2

=
1
2
,
so using horizontal slices in R

, we have

R
e
x−y
x+y
dA =

R

f (x(u, v), y(u, v)) | J(u, v)| dA
=

1
0

v
−v
e
u
v
1
2
du dv
=

1
0

v
2
e
u
v

u=v
u=−v

dv
=

1
0
v
2
(e − e
−1
) dv
=
v
2
4
(e − e
−1
)

1
0
=
1
4
¸
e −
1
e

=
e
2
− 1
4e
3.5 Change of Variables in Multiple Integrals 121
The change of variables formula can be used to evaluate double integrals in polar
coordinates. Letting
x = x(r, θ) = r cos θ and y = y(r, θ) = r sin θ ,
we have
J(u, v) =

∂x
∂r
∂x
∂θ
∂y
∂r
∂y
∂θ

=

cos θ −r sin θ
sin θ r cos θ

= r cos
2
θ + r sin
2
θ = r ⇒ | J(u, v)| = |r| = r ,
so we have the following formula:
Double Integral in Polar Coordinates

R
f (x, y) dx dy =

R

f (r cos θ, r sin θ) r dr dθ , (3.24)
where the mapping x = r cos θ, y = r sin θ maps the region R

in the rθ-plane onto the
region R in the xy-plane in a one-to-one manner.
Example 3.10. Find the volume V inside the paraboloid z = x
2
+ y
2
for 0 ≤ z ≤ 1.
y
z
x
0
x
2
+ y
2
= 1
1
Figure 3.5.2 z = x
2
+ y
2
Solution: Using vertical slices, we see that
V =

R
(1 − z) dA =

R
(1 − (x
2
+ y
2
)) dA ,
where R = {(x, y) : x
2
+ y
2
≤ 1} is the unit disk in
2
(see Figure 3.5.2). In polar coordinates (r, θ) we know
that x
2
+ y
2
= r
2
and that the unit disk R is the set
R

= {(r, θ) : 0 ≤ r ≤ 1, 0 ≤ θ ≤ 2π}. Thus,
V =

0

1
0
(1 − r
2
) r dr dθ
=

0

1
0
(r − r
3
) dr dθ
=

0

r
2
2

r
4
4

r=1
r=0

=

0
1
4

=
π
2
122 CHAPTER 3. MULTIPLE INTEGRALS
Example 3.11. Find the volume V inside the cone z =

x
2
+ y
2
for 0 ≤ z ≤ 1.
y
z
x
0
x
2
+ y
2
= 1
1
Figure 3.5.3 z =

x
2
+ y
2
Solution: Using vertical slices, we see that
V =

R
(1 − z) dA =

R
¸
1 −

x
2
+ y
2

dA ,
where R = {(x, y) : x
2
+ y
2
≤ 1} is the unit disk in
2
(see Figure 3.5.3). In polar coordinates (r, θ) we know
that

x
2
+ y
2
= r and that the unit disk R is the set
R

= {(r, θ) : 0 ≤ r ≤ 1, 0 ≤ θ ≤ 2π}. Thus,
V =

0

1
0
(1 − r) r dr dθ
=

0

1
0
(r − r
2
) dr dθ
=

0

r
2
2

r
3
3

r=1
r=0

=

0
1
6

=
π
3
In a similar fashion, it can be shown (see Exercises 5-6) that triple integrals in
cylindrical and spherical coordinates take the following forms:
Triple Integral in Cylindrical Coordinates

S
f (x, y, z) dx dy dz =

S

f (r cos θ, r sin θ, z) r dr dθ dz , (3.25)
where the mapping x = r cos θ, y = r sin θ, z = z maps the solid S

in rθz-space onto
the solid S in xyz-space in a one-to-one manner.
Triple Integral in Spherical Coordinates

S
f (x, y, z) dx dy dz =

S

f (ρ sin φ cos θ, ρ sin φ sin θ, ρ cos φ) ρ
2
sin φ dρ dφ dθ ,
(3.26)
where the mapping x = ρ sin φ cos θ, y = ρ sin φ sin θ, z = ρ cos φ maps the solid S

in
ρφθ-space onto the solid S in xyz-space in a one-to-one manner.
3.5 Change of Variables in Multiple Integrals 123
Example 3.12. For a > 0, find the volume V inside the sphere S = x
2
+ y
2
+ z
2
= a
2
.
Solution: We see that S is the set ρ = a in spherical coordinates, so
V =

S
1 dV =

0

π
0

a
0
1 ρ
2
sin φ dρ dφ dθ
=

0

π
0
¸
ρ
3
3

ρ=a
ρ=0

sin φ dφ dθ =

0

π
0
a
3
3
sin φ dφ dθ
=

0
¸

a
3
3
cos φ

φ=π
φ=0

dθ =

0
2a
3
3
dθ =
4πa
3
3
.

¨
Exercises
A
1. Find the volume V inside the paraboloid z = x
2
+ y
2
for 0 ≤ z ≤ 4.
2. Find the volume V inside the cone z =

x
2
+ y
2
for 0 ≤ z ≤ 3.
B
3. Find the volume V of the solid inside both x
2
+ y
2
+ z
2
= 4 and x
2
+ y
2
= 1.
4. Find the volume V inside both the sphere x
2
+ y
2
+ z
2
= 1 and the cone z =

x
2
+ y
2
.
5. Prove formula (3.25). 6. Prove formula (3.26).
7. Evaluate

R
sin

x+y
2

cos

x−y
2

dA, where R is the triangle with vertices (0, 0), (2, 0)
and (1, 1). (Hint: Use the change of variables u = (x + y)/2, v = (x − y)/2.)
8. Find the volume of the solid bounded by z = x
2
+ y
2
and z
2
= 4(x
2
+ y
2
).
9. Find the volume inside the elliptic cylinder
x
2
a
2
+
y
2
b
2
= 1 for 0 ≤ z ≤ 2.
C
10. Show that the volume inside the ellipsoid
x
2
a
2
+
y
2
b
2
+
z
2
c
2
= 1 is
4πabc
3
. (Hint: Use the
change of variables x = au, y = bv, z = cw, then consider Example 3.12.)
11. Show that the Beta function, defined by
B(x, y) =

1
0
t
x−1
(1 − t)
y−1
dt , for x > 0, y > 0,
satisfies the relation B(y, x) = B(x, y) for x > 0, y > 0.
12. Using the substitution t = u/(u + 1), show that the Beta function can be written as
B(x, y) =

0
u
x−1
(u + 1)
x+y
du , for x > 0, y > 0.
124 CHAPTER 3. MULTIPLE INTEGRALS
3.6 Application: Center of Mass
x
y
0
y = f (x)
R
( ¯ x, ¯ y)
a
b
Figure 3.6.1 Center of mass of R
Recall from single-variable calculus that for a re-
gion R = {(x, y) : a ≤ x ≤ b, 0 ≤ y ≤ f (x)} in
2
that rep-
resents a thin, flat plate (see Figure 3.6.1), where
f (x) is a continuous function on [a, b], the center of
mass of R has coordinates ( ¯ x, ¯ y) given by
¯ x =
M
y
M
and ¯ y =
M
x
M
,
where
M
x
=

b
a
( f (x))
2
2
dx , M
y
=

b
a
x f (x) dx , M =

b
a
f (x) dx , (3.27)
assuming that R has uniform density, i.e the mass of R is uniformly distributed over
the region. In this case the area M of the region is considered the mass of R (the
density is constant, and taken as 1 for simplicity).
In the general case where the density of a region (or lamina) R is a continuous
function δ = δ(x, y) of the coordinates (x, y) of points inside R (where R can be any
region in
2
) the coordinates ( ¯ x, ¯ y) of the center of mass of R are given by
¯ x =
M
y
M
and ¯ y =
M
x
M
, (3.28)
where
M
y
=

R
xδ(x, y) dA , M
x
=

R
yδ(x, y) dA , M =

R
δ(x, y) dA , (3.29)
The quantities M
x
and M
y
are called the moments (or first moments) of the region R
about the x-axis and y-axis, respectively. The quantity M is the mass of the region R.
To see this, think of taking a small rectangle inside R with dimensions ∆x and ∆y close
to 0. The mass of that rectangle is approximately δ(x

, y

)∆x ∆y, for some point (x

, y

)
in that rectangle. Then the mass of R is the limit of the sums of the masses of all such
rectangles inside R as the diagonals of the rectangles approach 0, which is the double
integral

R
δ(x, y) dA.
Note that the formulas in (3.27) represent a special case when δ(x, y) = 1 throughout
R in the formulas in (3.29).
Example 3.13. Find the center of mass of the region R = {(x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 2x
2
},
if the density function at (x, y) is δ(x, y) = x + y.
3.6 Application: Center of Mass 125
x
y
0
y = 2x
2
R
1
Figure 3.6.2
Solution: The region R is shown in Figure 3.6.2. We have
M =

R
δ(x, y) dA
=

1
0

2x
2
0
(x + y) dy dx
=

1
0

¸
¸
¸
¸
¸
¸
¸
xy +
y
2
2

y=2x
2
y=0

¸
¸
¸
¸
¸
¸
¸
dx
=

1
0
(2x
3
+ 2x
4
) dx
=
x
4
2
+
2x
5
5

1
0
=
9
10
and
M
x
=

R
yδ(x, y) dA M
y
=

R
xδ(x, y) dA
=

1
0

2x
2
0
y(x + y) dy dx =

1
0

2x
2
0
x(x + y) dy dx
=

1
0

¸
¸
¸
¸
¸
¸
¸
xy
2
2
+
y
3
3

y=2x
2
y=0

¸
¸
¸
¸
¸
¸
¸
dx =

1
0

¸
¸
¸
¸
¸
¸
¸
x
2
y +
xy
2
2

y=2x
2
y=0

¸
¸
¸
¸
¸
¸
¸
dx
=

1
0
(2x
5
+
8x
6
3
) dx =

1
0
(2x
4
+ 2x
5
) dx
=
x
6
3
+
8x
7
21

1
0
=
5
7
=
2x
5
5
+
x
6
3

1
0
=
11
15
,
so the center of mass ( ¯ x, ¯ y) is given by
¯ x =
M
y
M
=
11/15
9/10
=
22
27
, ¯ y =
M
x
M
=
5/7
9/10
=
50
63
.
Note how this center of mass is a little further towards the upper corner of the region
R than when the density is uniform(it is easy to use the formulas in (3.27) to showthat
( ¯ x, ¯ y) =

3
4
,
3
5

in that case). This makes sense since the density function δ(x, y) = x + y
increases as (x, y) approaches that upper corner, where there is quite a bit of area.
In the special case where the density function δ(x, y) is a constant function on the
region R, the center of mass ( ¯ x, ¯ y) is called the centroid of R.
126 CHAPTER 3. MULTIPLE INTEGRALS
The formulas for the center of mass of a region in
2
can be generalized to a solid S
in
3
. Let S be a solid with a continuous mass density function δ(x, y, z) at any point
(x, y, z) in S . Then the center of mass of S has coordinates ( ¯ x, ¯ y, ¯ z), where
¯ x =
M
yz
M
, ¯ y =
M
xz
M
, ¯ z =
M
xy
M
, (3.30)
where
M
yz
=

S
xδ(x, y, z) dV , M
xz
=

S
yδ(x, y, z) dV , M
xy
=

S
zδ(x, y, z) dV , (3.31)
M =

S
δ(x, y, z) dV . (3.32)
In this case, M
yz
, M
xz
and M
xy
are called the moments (or first moments) of S around
the yz-plane, xz-plane and xy-plane, respectively. Also, M is the mass of S .
Example 3.14. Find the center of mass of the solid S = {(x, y, z) : z ≥ 0, x
2
+y
2
+z
2
≤ a
2
},
if the density function at (x, y, z) is δ(x, y, z) = 1.
y
z
x
0 a
( ¯ x, ¯ y, ¯ z)
a
Figure 3.6.3
Solution: The solid S is just the upper hemisphere inside the
sphere of radius a centered at the origin (see Figure 3.6.3). So
since the density function is a constant and S is symmetric
about the z-axis, then it is clear that ¯ x = 0 and ¯ y = 0, so we
need only find ¯ z. We have
M =

S
δ(x, y, z) dV =

S
1 dV = Volume(S ).
But since the volume of S is half the volume of the sphere of
radius a, which we know by Example 3.12 is
4πa
3
3
, then M =
2πa
3
3
. And
M
xy
=

S
zδ(x, y, z) dV
=

S
z dV , which in spherical coordinates is
=

0

π/2
0

a
0
(ρ cos φ) ρ
2
sin φ dρ dφ dθ
=

0

π/2
0
sin φ cos φ
¸
a
0
ρ
3

dφ dθ
=

0

π/2
0
a
4
4
sin φ cos φ dφ dθ
3.6 Application: Center of Mass 127
M
xy
=

0

π/2
0
a
4
8
sin 2φ dφ dθ (since sin 2φ = 2 sin φ cos φ)
=

0

a
4
16
cos 2φ

φ=π/2
φ=0

=

0
a
4
8

=
πa
4
4
,
so
¯ z =
M
xy
M
=
πa
4
4
2πa
3
3
=
3a
8
.
Thus, the center of mass of S is ( ¯ x, ¯ y, ¯ z) =

0, 0,
3a
8

.

¨
Exercises
A
For Exercises 1-5, find the center of mass of the region R with the given density func-
tion δ(x, y).
1. R = {(x, y) : 0 ≤ x ≤ 2, 0 ≤ y ≤ 4 }, δ(x, y) = 2y
2. R = {(x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ x
2
}, δ(x, y) = x + y
3. R = {(x, y) : y ≥ 0, x
2
+ y
2
≤ a
2
}, δ(x, y) = 1
4. R = {(x, y) : y ≥ 0, x ≥ 0, 1 ≤ x
2
+ y
2
≤ 4 }, δ(x, y) =

x
2
+ y
2
5. R = {(x, y) : y ≥ 0, x
2
+ y
2
≤ 1 }, δ(x, y) = y
B
For Exercises 6-10, find the center of mass of the solid S with the given density func-
tion δ(x, y, z).
6. S = {(x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 }, δ(x, y, z) = xyz
7. S = {(x, y, z) : z ≥ 0, x
2
+ y
2
+ z
2
≤ a
2
}, δ(x, y, z) = x
2
+ y
2
+ z
2
8. S = {(x, y, z) : x ≥ 0, y ≥ 0, z ≥ 0, x
2
+ y
2
+ z
2
≤ a
2
}, δ(x, y, z) = 1
9. S = {(x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 }, δ(x, y, z) = x
2
+ y
2
+ z
2
10. S = {(x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 − x − y}, δ(x, y, z) = 1
128 CHAPTER 3. MULTIPLE INTEGRALS
3.7 Application: Probability and Expected Value
In this section we will briefly discuss some applications of multiple integrals in the
field of probability theory. In particular we will see ways in which multiple integrals
can be used to calculate probabilities and expected values.
Probability
Suppose that you have a standard six-sided (fair) die, and you let a variable X
represent the value rolled. Then the probability of rolling a 3, written as P(X = 3),
is
1
6
, since there are six sides on the die and each one is equally likely to be rolled,
and hence in particular the 3 has a one out of six chance of being rolled. Likewise the
probability of rolling at most a 3, written as P(X ≤ 3), is
3
6
=
1
2
, since of the six numbers
on the die, there are three equally likely numbers (1, 2, and 3) that are less than or
equal to 3. Note that P(X ≤ 3) = P(X = 1) + P(X = 2) + P(X = 3). We call X a discrete
random variable on the sample space (or probability space) Ω consisting of all possible
outcomes. In our case, Ω = {1, 2, 3, 4, 5, 6}. An event A is a subset of the sample space.
For example, in the case of the die, the event X ≤ 3 is the set {1, 2, 3}.
Now let X be a variable representing a random real number in the interval (0, 1).
Note that the set of all real numbers between 0 and 1 is not a discrete (or countable)
set of values, i.e. it can not be put into a one-to-one correspondence with the set of
positive integers.
3
In this case, for any real number x in (0, 1), it makes no sense
to consider P(X = x) since it must be 0 (why?). Instead, we consider the probability
P(X ≤ x), which is given by P(X ≤ x) = x. The reasoning is this: the interval (0, 1) has
length 1, and for x in (0, 1) the interval (0, x) has length x. So since X represents a
random number in (0, 1), and hence is uniformly distributed over (0, 1), then
P(X ≤ x) =
length of (0, x)
length of (0, 1)
=
x
1
= x .
We call X a continuous random variable on the sample space Ω = (0, 1). An event A is
a subset of the sample space. For example, in our case the event X ≤ x is the set (0, x).
In the case of a discrete random variable, we saw how the probability of an event
was the sum of the probabilities of the individual outcomes comprising that event (e.g.
P(X ≤ 3) = P(X = 1) + P(X = 2) + P(X = 3) in the die example). For a continuous random
variable, the probability of an event will instead be the integral of a function, which
we will now describe.
Let X be a continuous real-valued random variable on a sample space Ω in . For
3
For a proof see p. 9-10 in KAMKE, E., Theory of Sets, New York: Dover, 1950.
3.7 Application: Probability and Expected Value 129
simplicity, let Ω = (a, b). Define the distribution function F of X as
F(x) = P(X ≤ x) , for −∞ < x < ∞ (3.33)
=

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
1, for x ≥ b
P(X ≤ x), for a < x < b
0, for x ≤ a .
(3.34)
Suppose that there is a nonnegative, continuous real-valued function f on such that
F(x) =

x
−∞
f (y) dy , for −∞ < x < ∞ , (3.35)
and

−∞
f (x) dx = 1 . (3.36)
Then we call f the probability density function (or p.d.f. for short) for X. We thus have
P(X ≤ x) =

x
a
f (y) dy , for a < x < b . (3.37)
Also, by the Fundamental Theorem of Calculus, we have
F

(x) = f (x) , for −∞ < x < ∞. (3.38)
Example 3.15. Let X represent a randomly selected real number in the interval (0, 1).
We say that X has the uniform distribution on (0, 1), with distribution function
F(x) = P(X ≤ x) =

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
1, for x ≥ 1
x, for 0 < x < 1
0, for x ≤ 0 ,
(3.39)
and probability density function
f (x) = F

(x) =

¸
¸
¸
¸
¸
¸
1, for 0 < x < 1
0, elsewhere.
(3.40)
In general, if X represents a randomly selected real number in an interval (a, b), then
X has the uniform distribution function
F(x) = P(X ≤ x) =

¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
1, for x ≥ b
x
b−a
, for a < x < b
0, for x ≤ a ,
(3.41)
and probability density function
f (x) = F

(x) =

¸
¸
¸
¸
¸
¸
1
b−a
, for a < x < b
0, elsewhere.
(3.42)
130 CHAPTER 3. MULTIPLE INTEGRALS
Example 3.16. A famous distribution function is given by the standard normal dis-
tribution, whose probability density function f is
f (x) =
1

e
−x
2
/2
, for −∞ < x < ∞. (3.43)
This is often called a “bell curve”, and is used widely in statistics. Since we are claim-
ing that f is a p.d.f., we should have

−∞
1

e
−x
2
/2
dx = 1 (3.44)
by formula (3.36), which is equivalent to

−∞
e
−x
2
/2
dx =

2π . (3.45)
We can use a double integral in polar coordinates to verify this integral. First,

−∞

−∞
e
−(x
2
+y
2
)/2
dx dy =

−∞
e
−y
2
/2
¸

−∞
e
−x
2
/2
dx

dy
=
¸

−∞
e
−x
2
/2
dx
¸

−∞
e
−y
2
/2
dy

=
¸

−∞
e
−x
2
/2
dx

2
since the same function is being integrated twice in the middle equation, just with
different variables. But using polar coordinates, we see that

−∞

−∞
e
−(x
2
+y
2
)/2
dx dy =

0

0
e
−r
2
/2
r dr dθ
=

0
¸
−e
−r
2
/2

r=∞
r=0

=

0
(0 − (−e
0
)) dθ =

0
1 dθ = 2π ,
and so
¸

−∞
e
−x
2
/2
dx

2
= 2π , and hence

−∞
e
−x
2
/2
dx =

2π .
3.7 Application: Probability and Expected Value 131
In addition to individual random variables, we can consider jointly distributed ran-
dom variables. For this, we will let X, Y and Z be three real-valued continuous random
variables defined on the same sample space Ω in (the discussion for two random
variables is similar). Then the joint distribution function F of X, Y and Z is given by
F(x, y, z) = P(X ≤ x, Y ≤ y, Z ≤ z) , for −∞ < x, y, z < ∞. (3.46)
If there is a nonnegative, continuous real-valued function f on
3
such that
F(x, y, z) =

z
−∞

y
−∞

x
−∞
f (u, v, w) du dv dw , for −∞ < x, y, z < ∞ (3.47)
and

−∞

−∞

−∞
f (x, y, z) dx dy dz = 1 , (3.48)
then we call f the joint probability density function (or joint p.d.f. for short) for X, Y
and Z. In general, for a
1
< b
1
, a
2
< b
2
, a
3
< b
3
, we have
P(a
1
< X ≤ b
1
, a
2
< Y ≤ b
2
, a
3
< Z ≤ b
3
) =

b
3
a
3

b
2
a
2

b
1
a
1
f (x, y, z) dx dy dz , (3.49)
with the ≤ and < symbols interchangeable in any combination. A triple integral, then,
can be thought of as representing a probability (for a function f which is a p.d.f.).
Example 3.17. Let a, b, and c be real numbers selected randomly from the interval
(0, 1). What is the probability that the equation ax
2
+ bx + c = 0 has at least one real
solution x?
a
c
0
c =
1
4a
1
1
1
4
R
1
R
2
Figure 3.7.1 Region
R = R
1
∪ R
2
Solution: We know by the quadratic formula that there is at
least one real solution if b
2
− 4ac ≥ 0. So we need to calculate
P(b
2
− 4ac ≥ 0). We will use three jointly distributed random
variables to do this. First, since 0 < a, b, c < 1, we have
b
2
− 4ac ≥ 0 ⇔ 0 < 4ac ≤ b
2
< 1 ⇔ 0 < 2

a

c ≤ b < 1 ,
where the last relation holds for all 0 < a, c < 1 such that
0 < 4ac < 1 ⇔ 0 < c <
1
4a
.
Considering a, b and c as real variables, the region R in the ac-plane where the above
relation holds is given by R = {(a, c) : 0 < a < 1, 0 < c < 1, 0 < c <
1
4a
}, which we can see
is a union of two regions R
1
and R
2
, as in Figure 3.7.1 above.
Now let X, Y and Z be continuous random variables, each representing a randomly
selected real number from the interval (0, 1) (think of X, Y and Z representing a, b
and c, respectively). Then, similar to how we showed that f (x) = 1 is the p.d.f. of the
132 CHAPTER 3. MULTIPLE INTEGRALS
uniform distribution on (0, 1), it can be shown that f (x, y, z) = 1 for x, y, z in (0, 1)
(0 elsewhere) is the joint p.d.f. of X, Y and Z. Now,
P(b
2
− 4ac ≥ 0) = P((a, c) ∈ R, 2

a

c ≤ b < 1) ,
so this probability is the triple integral of f (a, b, c) = 1 as b varies from 2

a

c to 1 and
as (a, c) varies over the region R. Since R can be divided into two regions R
1
and R
2
,
then the required triple integral can be split into a sum of two triple integrals, using
vertical slices in R:
P(b
2
− 4ac ≥ 0) =

1/4
0

1
0
..............
R
1

1
2

a

c
1 db dc da +

1
1/4

1/4a
0
..................
R
2

1
2

a

c
1 db dc da
=

1/4
0

1
0
(1 − 2

a

c) dc da +

1
1/4

1/4a
0
(1 − 2

a

c) dc da
=

1/4
0

c −
4
3

a c
3/2

c=1
c=0

da +

1
1/4

c −
4
3

a c
3/2

c=1/4a
c=0

da
=

1/4
0

1 −
4
3

a

da +

1
1/4
1
12a
da
= a −
8
9
a
3/2

1/4
0
+
1
12
ln a

1
1/4
=
¸
1
4

1
9

+
¸
0 −
1
12
ln
1
4

=
5
36
+
1
12
ln 4
P(b
2
− 4ac ≥ 0) =
5 + 3 ln 4
36
≈ 0.2544
In other words, the equation ax
2
+ bx + c = 0 has about a 25% chance of being solved!
Expected Value
The expected value EX of a random variable X can be thought of as the “average”
value of X as it varies over its sample space. If X is a discrete random variable, then
EX =
¸
x
x P(X = x) , (3.50)
with the sum being taken over all elements x of the sample space. For example, if X
represents the number rolled on a six-sided die, then
EX =
6
¸
x=1
x P(X = x) =
6
¸
x=1
x
1
6
= 3.5 (3.51)
is the expected value of X, which is the average of the integers 1 − 6.
3.7 Application: Probability and Expected Value 133
If X is a real-valued continuous random variable with p.d.f. f , then
EX =

−∞
x f (x) dx . (3.52)
For example, if X has the uniform distribution on the interval (0, 1), then its p.d.f. is
f (x) =

¸
¸
¸
¸
¸
¸
1, for 0 < x < 1
0, elsewhere,
(3.53)
and so
EX =

−∞
x f (x) dx =

1
0
x dx =
1
2
. (3.54)
For a pair of jointly distributed, real-valued continuous random variables X and Y
with joint p.d.f. f (x, y), the expected values of X and Y are given by
EX =

−∞

−∞
x f (x, y) dx dy and EY =

−∞

−∞
y f (x, y) dx dy , (3.55)
respectively.
Example 3.18. If you were to pick n > 2 random real numbers from the interval (0, 1),
what are the expected values for the smallest and largest of those numbers?
Solution: Let U
1
, . . . , U
n
be n continuous random variables, each representing a ran-
domly selected real number from (0, 1), i.e. each has the uniform distribution on (0, 1).
Define random variables X and Y by
X = min(U
1
, . . . , U
n
) and Y = max(U
1
, . . . , U
n
) .
Then it can be shown
4
that the joint p.d.f. of X and Y is
f (x, y) =

¸
¸
¸
¸
¸
¸
n(n − 1)(y − x)
n−2
, for 0 ≤ x ≤ y ≤ 1
0, elsewhere.
(3.56)
Thus, the expected value of X is
EX =

1
0

1
x
n(n − 1)x(y − x)
n−2
dy dx
=

1
0

nx(y − x)
n−1

y=1
y=x

dx
=

1
0
nx(1 − x)
n−1
dx , so integration by parts yields
= −x(1 − x)
n

1
n + 1
(1 − x)
n+1

1
0
EX =
1
n + 1
,
4
See Ch. 6 in HOEL, PORT and STONE.
134 CHAPTER 3. MULTIPLE INTEGRALS
and similarly (see Exercise 3) it can be shown that
EY =

1
0

y
0
n(n − 1)y(y − x)
n−2
dx dy =
n
n + 1
.
So, for example, if you were to repeatedly take samples of n = 3 random real numbers
from (0, 1), and each time store the minimum and maximum values in the sample,
then the average of the minimums would approach
1
4
and the average of the max-
imums would approach
3
4
as the number of samples grows. It would be relatively
simple (see Exercise 4) to write a computer program to test this.

¨
Exercises
B
1. Evaluate the integral

−∞
e
−x
2
dx using anything you have learned so far.
2. For σ > 0 and µ > 0, evaluate

−∞
1
σ

e
−(x−µ)
2
/2σ
2
dx.
3. Show that EY =
n
n+1
in Example 3.18
C
4. Write a computer program (in the language of your choice) that verifies the results
in Example 3.18 for the case n = 3 by taking large numbers of samples.
5. Repeat Exercise 4 for the case when n = 4.
6. For continuous random variables X, Y with joint p.d.f. f (x, y), define the second
moments E(X
2
) and E(Y
2
) by
E(X
2
) =

−∞

−∞
x
2
f (x, y) dx dy and E(Y
2
) =

−∞

−∞
y
2
f (x, y) dx dy ,
and the variances Var(X) and Var(Y) by
Var(X) = E(X
2
) − (EX)
2
and Var(Y) = E(Y
2
) − (EY)
2
.
Find Var(X) and Var(Y) for X and Y as in Example 3.18.
7. Continuing Exercise 6, the correlation ρ between X and Y is defined as
ρ =
E(XY) − (EX)(EY)

Var(X) Var(Y)
,
where E(XY) =

−∞

−∞
xy f (x, y) dx dy. Find ρ for X and Y as in Example 3.18.
(Note: The quantity E(XY) − (EX)(EY) is called the covariance of X and Y.)
8. In Example 3.17 would the answer change if the interval (0, 100) is used instead of
(0, 1)? Explain.
4 Line and Surface Integrals
4.1 Line Integrals
In single-variable calculus you learned how to integrate a real-valued function f (x)
over an interval [a, b] in
1
. This integral (usually called a Riemann integral) can be
thought of as an integral over a path in
1
, since an interval (or collection of intervals)
is really the only kind of “path” in
1
. You may also recall that if f (x) represented the
force applied along the x-axis to an object at position x in [a, b], then the work W done
in moving that object from position x = a to x = b was defined as the integral:
W =

b
a
f (x) dx
In this section, we will see how to define the integral of a function (either real-
valued or vector-valued) of two variables over a general path (i.e. a curve) in
2
.
This definition will be motivated by the physical notion of work. We will begin with
real-valued functions of two variables.
In physics, the intuitive idea of work is that
Work = Force × Distance .
Suppose that we want to find the total amount W of work done in moving an object
along a curve C in
2
with a smooth parametrization x = x(t), y = y(t), a ≤ t ≤ b, with
a force f (x, y) which varies with the position (x, y) of the object and is applied in the
direction of motion along C (see Figure 4.1.1 below).
x
y
0
C
t = a
t = b
∆s
i

∆x
i
2
+ ∆y
i
2
t = t
i
t = t
i+1
∆y
i
∆x
i
Figure 4.1.1 Curve C : x = x(t), y = y(t) for t in [a, b]
We will assume for now that the function f (x, y) is continuous and real-valued, so
we only consider the magnitude of the force. Partition the interval [a, b] as follows:
a = t
0
< t
1
< t
2
< · · · < t
n−1
< t
n
= b , for some integer n ≥ 2
135
136 CHAPTER 4. LINE AND SURFACE INTEGRALS
As we can see from Figure 4.1.1, over a typical subinterval [t
i
, t
i+1
] the distance ∆s
i
traveled along the curve is approximately

∆x
i
2
+ ∆y
i
2
, by the Pythagorean Theorem.
Thus, if the subinterval is small enough then the work done in moving the object along
that piece of the curve is approximately
Force × Distance ≈ f (x
i∗
, y
i∗
)

∆x
i
2
+ ∆y
i
2
, (4.1)
where (x
i∗
, y
i∗
) = (x(t
i
∗), y(t
i
∗)) for some t
i
∗ in [t
i
, t
i+1
], and so
W ≈
n−1
¸
i=0
f (x
i∗
, y
i∗
)

∆x
i
2
+ ∆y
i
2
(4.2)
is approximately the total amount of work done over the entire curve. But since

∆x
i
2
+ ∆y
i
2
=

¸
∆x
i
∆t
i

2
+
¸
∆y
i
∆t
i

2
∆t
i
,
where ∆t
i
= t
i+1
− t
i
, then
W ≈
n−1
¸
i=0
f (x
i∗
, y
i∗
)

¸
∆x
i
∆t
i

2
+
¸
∆y
i
∆t
i

2
∆t
i
. (4.3)
Taking the limit of that sum as the length of the largest subinterval goes to 0, the sum
over all subintervals becomes the integral from t = a to t = b,
∆x
i
∆t
i
and
∆y
i
∆t
i
become x

(t)
and y

(t), respectively, and f (x
i∗
, y
i∗
) becomes f (x(t), y(t)), so that
W =

b
a
f (x(t), y(t))

x

(t)
2
+ y

(t)
2
dt . (4.4)
The integral on the right side of the above equation gives us our idea of how to
define, for any real-valued function f (x, y), the integral of f (x, y) along the curve C,
called a line integral:
Definition 4.1. For a real-valued function f (x, y) and a curve C in
2
, parametrized
by x = x(t), y = y(t), a ≤ t ≤ b, the line integral of f (x, y) along C with respect to arc
length s is

C
f (x, y) ds =

b
a
f (x(t), y(t))

x

(t)
2
+ y

(t)
2
dt . (4.5)
The symbol ds is the differential of the arc length function
s = s(t) =

t
a

x

(u)
2
+ y

(u)
2
du , (4.6)
4.1 Line Integrals 137
which you may recognize fromSection 1.9 as the length of the curve C over the interval
[a, t], for all t in [a, b]. That is,
ds = s

(t) dt =

x

(t)
2
+ y

(t)
2
dt , (4.7)
by the Fundamental Theorem of Calculus.
For a general real-valued function f (x, y), what does the line integral

C
f (x, y) ds
represent? The preceding discussion of ds gives us a clue. You can think of differen-
tials as infinitesimal lengths. So if you think of f (x, y) as the height of a picket fence
along C, then f (x, y) ds can be thought of as approximately the area of a section of that
fence over some infinitesimally small section of the curve, and thus the line integral

C
f (x, y) ds is the total area of that picket fence (see Figure 4.1.2).
x
y
0
C
ds
f (x, y)
Figure 4.1.2 Area of shaded rectangle = height × width ≈ f (x, y) ds
Example 4.1. Use a line integral to show that the lateral surface area A of a right
circular cylinder of radius r and height h is 2πrh.
y
z
x
0
r
h = f (x, y)
C : x
2
+ y
2
= r
2
Figure 4.1.3
Solution: We will use the right circular cylinder with base cir-
cle C given by x
2
+ y
2
= r
2
and with height h in the positive z
direction (see Figure 4.1.3). Parametrize C as follows:
x = x(t) = r cos t , y = y(t) = r sin t , 0 ≤ t ≤ 2π
Let f (x, y) = h for all (x, y). Then
A =

C
f (x, y) ds =

b
a
f (x(t), y(t))

x

(t)
2
+ y

(t)
2
dt
=

0
h

(−r sin t)
2
+ (r cos t)
2
dt
= h

0
r

sin
2
t + cos
2
t dt
= rh

0
1 dt = 2πrh
138 CHAPTER 4. LINE AND SURFACE INTEGRALS
Note in Example 4.1 that if we had traversed the circle C twice, i.e. let t vary from
0 to 4π, then we would have gotten an area of 4πrh, i.e. twice the desired area, even
though the curve itself is still the same (namely, a circle of radius r). Also, notice
that we traversed the circle in the counter-clockwise direction. If we had gone in the
clockwise direction, using the parametrization
x = x(t) = r cos(2π − t) , y = y(t) = r sin(2π − t) , 0 ≤ t ≤ 2π , (4.8)
then it is easy to verify (see Exercise 12) that the value of the line integral is un-
changed.
In general, it can be shown (see Exercise 15) that reversing the direction in which
a curve C is traversed leaves

C
f (x, y) ds unchanged, for any f (x, y). If a curve C has a
parametrization x = x(t), y = y(t), a ≤ t ≤ b, then denote by −C the same curve as C but
traversed in the opposite direction. Then −C is parametrized by
x = x(a + b − t) , y = y(a + b − t) , a ≤ t ≤ b , (4.9)
and we have

C
f (x, y) ds =

−C
f (x, y) ds . (4.10)
Notice that our definition of the line integral was with respect to the arc length
parameter s. We can also define

C
f (x, y) dx =

b
a
f (x(t), y(t)) x

(t) dt (4.11)
as the line integral of f (x, y) along C with respect to x, and

C
f (x, y) dy =

b
a
f (x(t), y(t)) y

(t) dt (4.12)
as the line integral of f (x, y) along C with respect to y.
In the derivation of the formula for a line integral, we used the idea of work as force
multiplied by distance. However, we know that force is actually a vector. So it would
be helpful to develop a vector form for a line integral. For this, suppose that we have
a function f(x, y) defined on
2
by
f(x, y) = P(x, y) i + Q(x, y) j
for some continuous real-valued functions P(x, y) and Q(x, y) on
2
. Such a function f
is called a vector field on
2
. It is defined at points in
2
, and its values are vectors
in
2
. For a curve C with a smooth parametrization x = x(t), y = y(t), a ≤ t ≤ b, let
r(t) = x(t) i + y(t) j
4.1 Line Integrals 139
be the position vector for a point (x(t), y(t)) on C. Then r

(t) = x

(t) i + y

(t) j and so

C
P(x, y) dx +

C
Q(x, y) dy =

b
a
P(x(t), y(t)) x

(t) dt +

b
a
Q(x(t), y(t)) y

(t) dt
=

b
a
(P(x(t), y(t)) x

(t) + Q(x(t), y(t)) y

(t)) dt
=

b
a
f(x(t), y(t)) ··· r

(t) dt
by definition of f(x, y). Notice that the function f(x(t), y(t)) ··· r

(t) is a real-valued func-
tion on [a, b], so the last integral on the right looks somewhat similar to our earlier
definition of a line integral. This leads us to the following definition:
Definition 4.2. For a vector field f(x, y) = P(x, y) i + Q(x, y) j and a curve C with a
smooth parametrization x = x(t), y = y(t), a ≤ t ≤ b, the line integral of f along C is

C
f ··· dr =

C
P(x, y) dx +

C
Q(x, y) dy (4.13)
=

b
a
f(x(t), y(t)) ··· r

(t) dt , (4.14)
where r(t) = x(t) i + y(t) j is the position vector for points on C.
We use the notation dr = r

(t) dt = dx i + dy j to denote the differential of the vector-
valued function r. The line integral in Definition 4.2 is often called a line integral of
a vector field to distinguish it from the line integral in Definition 4.1 which is called a
line integral of a scalar field. For convenience we will often write

C
P(x, y) dx +

C
Q(x, y) dy =

C
P(x, y) dx + Q(x, y) dy ,
where it is understood that the line integral along C is being applied to both P and
Q. The quantity P(x, y) dx + Q(x, y) dy is known as a differential form. For a real-
valued function F(x, y), the differential of F is dF =
∂F
∂x
dx +
∂F
∂y
dy. A differential form
P(x, y) dx + Q(x, y) dy is called exact if it equals dF for some function F(x, y).
Recall that if the points on a curve C have position vector r(t) = x(t) i+y(t) j, then r

(t)
is a tangent vector to C at the point (x(t), y(t)) in the direction of increasing t (which we
call the direction of C). Since C is a smooth curve, then r

(t) 0 on [a, b] and hence
T(t) =
r

(t)
¸
¸
¸r

(t)
¸
¸
¸
is the unit tangent vector to C at (x(t), y(t)). Putting Definitions 4.1 and 4.2 together
we get the following theorem:
140 CHAPTER 4. LINE AND SURFACE INTEGRALS
Theorem 4.1. For a vector field f(x, y) = P(x, y) i +Q(x, y) j and a curve C with a smooth
parametrization x = x(t), y = y(t), a ≤ t ≤ b and position vector r(t) = x(t) i + y(t) j,

C
f ··· dr =

C
f ··· Tds , (4.15)
where T(t) =
r

(t)
r

(t)
is the unit tangent vector to C at (x(t), y(t)).
If the vector field f(x, y) represents the force moving an object along a curve C, then
the work W done by this force is
W =

C
f ··· Tds =

C
f ··· dr . (4.16)
Example 4.2. Evaluate

C
(x
2
+ y
2
) dx + 2xy dy, where:
(a) C : x = t , y = 2t , 0 ≤ t ≤ 1
(b) C : x = t , y = 2t
2
, 0 ≤ t ≤ 1
x
y
0
(1, 2)
2
1
Figure 4.1.4
Solution: Figure 4.1.4 shows both curves.
(a) Since x

(t) = 1 and y

(t) = 2, then

C
(x
2
+ y
2
) dx + 2xy dy =

1
0

(x(t)
2
+ y(t)
2
)x

(t) + 2x(t)y(t) y

(t)

dt
=

1
0

(t
2
+ 4t
2
)(1) + 2t(2t)(2)

dt
=

1
0
13t
2
dt
=
13t
3
3

1
0
=
13
3
(b) Since x

(t) = 1 and y

(t) = 4t, then

C
(x
2
+ y
2
) dx + 2xy dy =

1
0

(x(t)
2
+ y(t)
2
)x

(t) + 2x(t)y(t) y

(t)

dt
=

1
0

(t
2
+ 4t
4
)(1) + 2t(2t
2
)(4t)

dt
=

1
0
(t
2
+ 20t
4
) dt
=
t
3
3
+ 4t
5

1
0
=
1
3
+ 4 =
13
3
4.1 Line Integrals 141
So in both cases, if the vector field f(x, y) = (x
2
+ y
2
) i + 2xy j represents the force
moving an object from (0, 0) to (1, 2) along the given curve C, then the work done is
13
3
.
This may lead you to think that work (and more generally, the line integral of a vector
field) is independent of the path taken. However, as we will see in the next section,
this is not always the case.
Although we defined line integrals over a single smooth curve, if C is a piecewise
smooth curve, that is
C = C
1
∪ C
2
∪ . . . ∪ C
n
is the union of smooth curves C
1
, . . . , C
n
, then we can define

C
f ··· dr =

C
1
f ··· dr
1
+

C
2
f ··· dr
2
+ . . . +

C
n
f ··· dr
n
where each r
i
is the position vector of the curve C
i
.
Example 4.3. Evaluate

C
(x
2
+y
2
) dx +2xy dy, where C is the polygonal path from (0, 0)
to (0, 2) to (1, 2).
x
y
0
(1, 2) 2
1
C
1
C
2
Figure 4.1.5
Solution: Write C = C
1
∪C
2
, where C
1
is the curve given by x = 0, y = t,
0 ≤ t ≤ 2 and C
2
is the curve given by x = t, y = 2, 0 ≤ t ≤ 1 (see Figure
4.1.5). Then

C
(x
2
+ y
2
) dx + 2xy dy =

C
1
(x
2
+ y
2
) dx + 2xy dy
+

C
2
(x
2
+ y
2
) dx + 2xy dy
=

2
0

(0
2
+ t
2
)(0) + 2(0)t(1)

dt +

1
0

(t
2
+ 4)(1) + 2t(2)(0)

dt
=

2
0
0 dt +

1
0
(t
2
+ 4) dt
=
t
3
3
+ 4t

1
0
=
1
3
+ 4 =
13
3
Line integral notation varies quite a bit. For example, in physics it is common to
see the notation

b
a
f ··· dl, where it is understood that the limits of integration a and
b are for the underlying parameter t of the curve, and the letter l signifies length.
Also, the formulation

C
f ··· Tds from Theorem 4.1 is often preferred in physics since it
emphasizes the idea of integrating the tangential component f···T of f in the direction of
T (i.e. in the direction of C), which is a useful physical interpretation of line integrals.
142 CHAPTER 4. LINE AND SURFACE INTEGRALS

¨
Exercises
A
For Exercises 1-4, calculate

C
f (x, y) ds for the given function f (x, y) and curve C.
1. f (x, y) = xy; C : x = cos t, y = sin t, 0 ≤ t ≤ π/2
2. f (x, y) =
x
x
2
+ 1
; C : x = t, y = 0, 0 ≤ t ≤ 1
3. f (x, y) = 2x + y; C: polygonal path from (0, 0) to (3, 0) to (3, 2)
4. f (x, y) = x + y
2
; C: path from (2, 0) counterclockwise along the circle x
2
+ y
2
= 4 to
the point (−2, 0) and then back to (2, 0) along the x-axis
5. Use a line integral to find the lateral surface area of the part of the cylinder
x
2
+ y
2
= 4 below the plane x + 2y + z = 6 and above the xy-plane.
For Exercises 6-11, calculate

C
f ··· dr for the given vector field f(x, y) and curve C.
6. f(x, y) = i − j; C : x = 3t, y = 2t, 0 ≤ t ≤ 1
7. f(x, y) = y i − x j; C : x = cos t, y = sin t, 0 ≤ t ≤ 2π
8. f(x, y) = x i + y j; C : x = cos t, y = sin t, 0 ≤ t ≤ 2π
9. f(x, y) = (x
2
− y) i + (x − y
2
) j; C : x = cos t, y = sin t, 0 ≤ t ≤ 2π
10. f(x, y) = xy
2
i + xy
3
j; C : the polygonal path from (0, 0) to (1, 0) to (0, 1) to (0, 0)
11. f(x, y) = (x
2
+ y
2
) i; C : x = 2 + cos t, y = sin t, 0 ≤ t ≤ 2π
B
12. Verify that the value of the line integral in Example 4.1 is unchanged using the
parametrization of the circle C given in formulas (4.8).
13. Show that if f ⊥ r

(t) at each point r(t) along a smooth curve C, then

C
f ··· dr = 0.
14. Showthat if f points in the same direction as r

(t) at each point r(t) along a smooth
curve C, then

C
f ··· dr =

C
f ds.
C
15. Prove that

C
f (x, y) ds =

−C
f (x, y) ds. (Hint: Use formulas (4.9).)
16. Let C be a smooth curve with arc length L, and suppose that f(x, y) = P(x, y) i +
Q(x, y) j is a vector field such that f(x, y) ≤ M for all (x, y) on C. Show that

C
f ··· dr

≤ ML. (Hint: Recall that

b
a
g(x) dx

b
a
|g(x)| dx for Riemann integrals.)
17. Prove that the Riemann integral

b
a
f (x) dx is a special case of a line integral.
4.2 Properties of Line Integrals 143
4.2 Properties of Line Integrals
We know from the previous section that for line integrals of real-valued functions
(scalar fields), reversing the direction in which the integral is taken along a curve
does not change the value of the line integral:

C
f (x, y) ds =

−C
f (x, y) ds (4.17)
For line integrals of vector fields, however, the value does change. To see this, let
f(x, y) = P(x, y) i + Q(x, y) j be a vector field, with P and Q continuously differentiable
functions. Let C be a smooth curve parametrized by x = x(t), y = y(t), a ≤ t ≤ b,
with position vector r(t) = x(t) i + y(t) j (we will usually abbreviate this by saying that
C : r(t) = x(t) i + y(t) j is a smooth curve). We know that the curve −C traversed in the
opposite direction is parametrized by x = x(a + b − t), y = y(a + b − t), a ≤ t ≤ b. Then

−C
P(x, y) dx =

b
a
P(x(a + b − t), y(a + b − t))
d
dt
(x(a + b − t)) dt
=

b
a
P(x(a + b − t), y(a + b − t)) (−x

(a + b − t)) dt (by the Chain Rule)
=

a
b
P(x(u), y(u)) (−x

(u)) (−du) (by letting u = a + b − t)
=

a
b
P(x(u), y(u)) x

(u) du
= −

b
a
P(x(u), y(u)) x

(u) du , since

a
b
= −

b
a
, so

−C
P(x, y) dx = −

C
P(x, y) dx
since we are just using a different letter (u) for the line integral along C. A similar
argument shows that

−C
Q(x, y) dy = −

C
Q(x, y) dy ,
and hence

−C
f ··· dr =

−C
P(x, y) dx +

−C
Q(x, y) dy
= −

C
P(x, y) dx + −

C
Q(x, y) dy
= −
¸
C
P(x, y) dx +

C
Q(x, y) dy

−C
f ··· dr = −

C
f ··· dr . (4.18)
144 CHAPTER 4. LINE AND SURFACE INTEGRALS
The above formula can be interpreted in terms of the work done by a force f(x, y)
(treated as a vector) moving an object along a curve C: the total work performed
moving the object along C from its initial point to its terminal point, and then back
to the initial point moving backwards along the same path, is zero. This is because
when force is considered as a vector, direction is accounted for.
The preceding discussion shows the importance of always taking the direction of
the curve into account when using line integrals of vector fields. For this reason, the
curves in line integrals are sometimes referred to as directed curves or oriented curves.
Recall that our definition of a line integral required that we have a parametrization
x = x(t), y = y(t), a ≤ t ≤ b for the curve C. But as we know, any curve has infinitely
many parametrizations. So could we get a different value for a line integral using
some other parametrization of C, say, x = ˜ x(u), y = ˜ y(u), c ≤ u ≤ d ? If so, this would
mean that our definition is not well-defined. Luckily, it turns out that the value of a
line integral of a vector field is unchanged as long as the direction of the curve C is
preserved by whatever parametrization is chosen:
Theorem 4.2. Let f(x, y) = P(x, y) i + Q(x, y) j be a vector field, and let C be a smooth
curve parametrized by x = x(t), y = y(t), a ≤ t ≤ b. Suppose that t = α(u) for c ≤ u ≤ d,
such that a = α(c), b = α(d), and α

(u) > 0 on the open interval (c, d) (i.e. α(u) is strictly
increasing on [c, d]). Then

C
f··· dr has the same value for the parametrizations x = x(t),
y = y(t), a ≤ t ≤ b and x = ˜ x(u) = x(α(u)), y = ˜ y(u) = y(α(u)), c ≤ u ≤ d.
Proof: Since α(u) is strictly increasing and maps [c, d] onto [a, b], then we know that
t = α(u) has an inverse function u = α
−1
(t) defined on [a, b] such that c = α
−1
(a),
d = α
−1
(b), and
du
dt
=
1
α

(u)
. Also, dt = α

(u) du, and by the Chain Rule
˜ x

(u) =
d ˜ x
du
=
d
du
(x(α(u))) =
dx
dt
dt
du
= x

(t) α

(u) ⇒ x

(t) =
˜ x

(u)
α

(u)
so making the susbstitution t = α(u) gives

b
a
P(x(t), y(t)) x

(t) dt =

α
−1
(b)
α
−1
(a)
P(x(α(u)), y(α(u)))
˜ x

(u)
α

(u)

(u) du)
=

d
c
P( ˜ x(u), ˜ y(u)) ˜ x

(u) du ,
which shows that

C
P(x, y) dx has the same value for both parametrizations. A simi-
lar argument shows that

C
Q(x, y) dy has the same value for both parametrizations,
and hence

C
f ··· dr has the same value. QED
Notice that the condition α

(u) > 0 in Theorem 4.2 means that the two parametriza-
tions move along C in the same direction. That was not the case with the “reverse”
parametrization for −C: for u = a + b − t we have t = α(u) = a + b − u ⇒ α

(u) = −1 < 0.
4.2 Properties of Line Integrals 145
Example 4.4. Evaluate the line integral

C
(x
2
+y
2
) dx+2xy dy fromExample 4.2, Section
4.1, along the curve C : x = t, y = 2t
2
, 0 ≤ t ≤ 1, where t = sin u for 0 ≤ u ≤ π/2.
Solution: First, we notice that 0 = sin 0, 1 = sin(π/2), and
dt
du
= cos u > 0 on (0, π/2). So
by Theorem 4.2 we know that if C is parametrized by
x = sin u , y = 2 sin
2
u , 0 ≤ u ≤ π/2
then

C
(x
2
+ y
2
) dx + 2xy dy should have the same value as we found in Example 4.2,
namely
13
3
. And we can indeed verify this:

C
(x
2
+ y
2
) dx + 2xy dy =

π/2
0

(sin
2
u + (2 sin
2
u)
2
) cos u + 2(sin u)(2 sin
2
u)4 sin u cos u

du
=

π/2
0

sin
2
u + 20 sin
4
u

cos u du
=
sin
3
u
3
+ 4 sin
5
u

π/2
0
=
1
3
+ 4 =
13
3
In other words, the line integral is unchanged whether t or u is the parameter for C.
By a closed curve, we mean a curve C whose initial point and terminal point are
the same, i.e. for C : x = x(t), y = y(t), a ≤ t ≤ b, we have (x(a), y(a)) = (x(b), y(b)).

C
t = a t = b
(a) Closed

C
t = a
t = b
(b) Not closed
Figure 4.2.1 Closed vs nonclosed curves
A simple closed curve is a closed curve which does not intersect itself. Note that
any closed curve can be regarded as a union of simple closed curves (think of the loops
in a figure eight). We use the special notation

C
f (x, y) ds and

C
f ··· dr
to denote line integrals of scalar and vector fields, respectively, along closed curves.
In some older texts you may see the notation

or

to indicate a line integral
traversing a closed curve in a counterclockwise or clockwise direction, respectively.
146 CHAPTER 4. LINE AND SURFACE INTEGRALS
So far, the examples we have seen of line integrals (e.g. Example 4.2) have had the
same value for different curves joining the initial point to the terminal point. That
is, the line integral has been independent of the path joining the two points. As we
mentioned before, this is not always the case. The following theorem gives a necessary
and sufficient condition for this path independence:
Theorem 4.3. In a region R, the line integral

C
f ··· dr is independent of the path
between any two points in R if and only if

C
f ··· dr = 0 for every closed curve C which is
contained in R.
Proof: Suppose that

C
f ··· dr = 0 for every closed curve C which is contained in R. Let
P
1
and P
2
be two distinct points in R. Let C
1
be a curve in R going from P
1
to P
2
, and
let C
2
be another curve in R going from P
1
to P
2
, as in Figure 4.2.2.

C
1
C
2
P
1
P
2
Figure 4.2.2
Then C = C
1
∪ −C
2
is a closed curve in R (from P
1
to P
1
), and so

C
f ··· dr = 0. Thus,
0 =

C
f ··· dr
=

C
1
f ··· dr +

−C
2
f ··· dr
=

C
1
f ··· dr −

C
2
f ··· dr , and so

C
1
f ··· dr =

C
2
f ··· dr. This proves path independence.
Conversely, suppose that the line integral

C
f ··· dr is independent of the path be-
tween any two points in R. Let C be a closed curve contained in R. Let P
1
and P
2
be
two distinct points on C. Let C
1
be a part of the curve C that goes from P
1
to P
2
, and
let C
2
be the remaining part of C that goes from P
1
to P
2
, again as in Figure 4.2.2.
Then by path independence we have

C
1
f ··· dr =

C
2
f ··· dr

C
1
f ··· dr −

C
2
f ··· dr = 0

C
1
f ··· dr +

−C
2
f ··· dr = 0 , so

C
f ··· dr = 0
since C = C
1
∪ −C
2
. QED
Clearly, the above theorem does not give a practical way to determine path inde-
4.2 Properties of Line Integrals 147
pendence, since it is impossible to check the line integrals around all possible closed
curves in a region. What it mostly does is give an idea of the way in which line inte-
grals behave, and how seemingly unrelated line integrals can be related (in this case,
a specific line integral between two points and all line integrals around closed curves).
For a more practical method for determining path independence, we first need a
version of the Chain Rule for multivariable functions:
Theorem 4.4. (Chain Rule) If z = f (x, y) is a continuously differentiable function
of x and y, and both x = x(t) and y = y(t) are differentiable functions of t, then z is a
differentiable function of t, and
dz
dt
=
∂z
∂x
dx
dt
+
∂z
∂y
dy
dt
(4.19)
at all points where the derivatives on the right are defined.
The proof is virtually identical to the proof of Theorem 2.2 from Section 2.4 (which
uses the Mean Value Theorem), so we omit it.
1
We will now use this Chain Rule to
prove the following sufficient condition for path independence of line integrals:
Theorem 4.5. Let f(x, y) = P(x, y) i + Q(x, y) j be a vector field in some region R, with
P and Q continuously differentiable functions on R. Let C be a smooth curve in R
parametrized by x = x(t), y = y(t), a ≤ t ≤ b. Suppose that there is a real-valued
function F(x, y) such that ∇F = f on R. Then

C
f ··· dr = F(B) − F(A) , (4.20)
where A = (x(a), y(a)) and B = (x(b), y(b)) are the endpoints of C. Thus, the line integral
is independent of the path between its endpoints, since it depends only on the values
of F at those endpoints.
Proof: By definition of

C
f ··· dr, we have

C
f ··· dr =

b
a

P(x(t), y(t)) x

(t) + Q(x(t), y(t)) y

(t)

dt
=

b
a
¸
∂F
∂x
dx
dt
+
∂F
∂y
dy
dt

dt (since ∇F = f ⇒
∂F
∂x
= P and
∂F
∂y
= Q)
=

b
a
F

(x(t), y(t)) dt (by the Chain Rule in Theorem 4.4)
= F(x(t), y(t))

b
a
= F(B) − F(A)
by the Fundamental Theorem of Calculus. QED
1
See TAYLOR and MANN, § 6.5.
148 CHAPTER 4. LINE AND SURFACE INTEGRALS
Theorem 4.5 can be thought of as the line integral version of the Fundamental
Theorem of Calculus. A real-valued function F(x, y) such that ∇F(x, y) = f(x, y) is called
a potential for f. A conservative vector field is one which has a potential.
Example 4.5. Recall from Examples 4.2 and 4.3 in Section 4.1 that the line integral

C
(x
2
+ y
2
) dx + 2xy dy was found to have the value
13
3
for three different curves C going
from the point (0, 0) to the point (1, 2). Use Theorem 4.5 to show that this line integral
is indeed path independent.
Solution: We need to find a real-valued function F(x, y) such that
∂F
∂x
= x
2
+ y
2
and
∂F
∂y
= 2xy .
Suppose that
∂F
∂x
= x
2
+y
2
, Then we must have F(x, y) =
1
3
x
3
+xy
2
+g(y) for some function
g(y). So
∂F
∂y
= 2xy + g

(y) satisfies the condition
∂F
∂y
= 2xy if g

(y) = 0, i.e. g(y) = K, where
K is a constant. Since any choice for K will do (why?), we pick K = 0. Thus, a potential
F(x, y) for f(x, y) = (x
2
+ y
2
) i + 2xy j exists, namely
F(x, y) =
1
3
x
3
+ xy
2
.
Hence the line integral

C
(x
2
+ y
2
) dx + 2xy dy is path independent.
Note that we can also verify that the value of the line integral of f along any curve C
going from (0, 0) to (1, 2) will always be
13
3
, since by Theorem 4.5

C
f ··· dr = F(1, 2) − F(0, 0) =
1
3
(1)
3
+ (1)(2)
2
− (0 + 0) =
1
3
+ 4 =
13
3
.
A consequence of Theorem 4.5 in the special case where C is a closed curve, so that
the endpoints A and B are the same point, is the following important corollary:
Corollary 4.6. If a vector field f has a potential in a region R, then

C
f ··· dr = 0 for
any closed curve C in R (i.e.

C
∇F ··· dr = 0 for any real-valued function F(x, y)).
Example 4.6. Evaluate

C
x dx + y dy for C : x = 2 cos t, y = 3 sin t, 0 ≤ t ≤ 2π.
Solution: The vector field f(x, y) = x i + y j has a potential F(x, y):
∂F
∂x
= x ⇒ F(x, y) =
1
2
x
2
+ g(y) , so
∂F
∂y
= y ⇒ g

(y) = y ⇒ g(y) =
1
2
y
2
+ K
4.2 Properties of Line Integrals 149
for any constant K, so F(x, y) =
1
2
x
2
+
1
2
y
2
is a potential for f(x, y). Thus,

C
x dx + y dy =

C
f ··· dr = 0
by Corollary 4.6, since the curve C is closed (it is the ellipse
x
2
4
+
y
2
9
= 1).

¨
Exercises
A
1. Evaluate

C
(x
2
+ y
2
) dx + 2xy dy for C : x = cos t, y = sin t, 0 ≤ t ≤ 2π.
2. Evaluate

C
(x
2
+ y
2
) dx + 2xy dy for C : x = cos t, y = sin t, 0 ≤ t ≤ π.
3. Is there a potential F(x, y) for f(x, y) = y i − x j? If so, find one.
4. Is there a potential F(x, y) for f(x, y) = x i − y j? If so, find one.
5. Is there a potential F(x, y) for f(x, y) = xy
2
i + x
3
y j? If so, find one.
B
6. Let f(x, y) and g(x, y) be vector fields, let a and b be constants, and let C be a curve
in
2
. Show that

C
(a f ± b g) ··· dr = a

C
f ··· dr ± b

C
g··· dr .
7. Let C be a curve whose arc length is L. Show that

C
1 ds = L.
8. Let f (x, y) and g(x, y) be continuously differentiable real-valued functions in a region
R. Show that

C
f ∇g ··· dr = −

C
g ∇f ··· dr
for any closed curve C in R. (Hint: Use Exercise 21 in Section 2.4.)
9. Let f(x, y) =
−y
x
2
+y
2
i +
x
x
2
+y
2
j for all (x, y) (0, 0), and C : x = cos t, y = sin t, 0 ≤ t ≤ 2π.
(a) Show that f = ∇F, for F(x, y) = tan
−1
(y/x).
(b) Show that

C
f ··· dr = 2π. Does this contradict Corollary 4.6? Explain.
C
10. Let g(x) and h(y) be differentiable functions, and let f(x, y) = h(y) i + g(x) j. Can f
have a potential F(x, y)? If so, find it. You may assume that F would be smooth.
(Hint: Consider the mixed partial derivatives of F.)
150 CHAPTER 4. LINE AND SURFACE INTEGRALS
4.3 Green’s Theorem
We will now see a way of evaluating the line integral of a smooth vector field around a
simple closed curve. A vector field f(x, y) = P(x, y) i+Q(x, y) j is smooth if its component
functions P(x, y) and Q(x, y) are smooth. We will use Green’s Theorem(sometimes called
Green’s Theorem in the plane) to relate the line integral around a closed curve with a
double integral over the region inside the curve:
Theorem 4.7. (Green’s Theorem) Let R be a region in
2
whose boundary is a
simple closed curve C which is piecewise smooth. Let f(x, y) = P(x, y) i + Q(x, y) j be a
smooth vector field defined on both R and C. Then

C
f ··· dr =

R
¸
∂Q
∂x

∂P
∂y

dA , (4.21)
where C is traversed so that R is always on the left side of C.
Proof: We will prove the theorem in the case for a simple region R, that is, where the
boundary curve C can be written as C = C
1
∪ C
2
in two distinct ways:
C
1
= the curve y = y
1
(x) from the point X
1
to the point X
2
(4.22)
C
2
= the curve y = y
2
(x) from the point X
2
to the point X
1
, (4.23)
where X
1
and X
2
are the points on C farthest to the left and right, respectively; and
C
1
= the curve x = x
1
(y) from the point Y
2
to the point Y
1
(4.24)
C
2
= the curve x = x
2
(y) from the point Y
1
to the point Y
2
, (4.25)
where Y
1
and Y
2
are the lowest and highest points, respectively, on C. See Figure
4.3.1.
x
y

y = y
2
(x)
y = y
1
(x)
x = x
2
(y)
x = x
1
(y)
Y
2
Y
1
X
2
X
1 R
C
a
b
d
c
Figure 4.3.1
Integrate P(x, y) around C using the representation C = C
1
∪ C
2
given by (4.23) and
4.3 Green’s Theorem 151
(4.24). Since y = y
1
(x) along C
1
(as x goes from a to b) and y = y
2
(x) along C
2
(as x goes
from b to a), as we see from Figure 4.3.1, then we have

C
P(x, y) dx =

C
1
P(x, y) dx +

C
2
P(x, y) dx
=

b
a
P(x, y
1
(x)) dx +

a
b
P(x, y
2
(x)) dx
=

b
a
P(x, y
1
(x)) dx −

b
a
P(x, y
2
(x)) dx
= −

b
a
(P(x, y
2
(x)) − P(x, y
1
(x))) dx
= −

b
a

P(x, y)

y=y
2
(x)
y=y
1
(x)

dx
= −

b
a

y
2
(x)
y
1
(x)
∂P(x, y)
∂y
dy dx (by the Fundamental Theorem of Calculus)
= −

R
∂P
∂y
dA .
Likewise, integrate Q(x, y) around C using the representation C = C
1
∪ C
2
given by
(4.25) and (4.26). Since x = x
1
(y) along C
1
(as y goes from d to c) and x = x
2
(y) along C
2
(as y goes from c to d), as we see from Figure 4.3.1, then we have

C
Q(x, y) dy =

C
1
Q(x, y) dy +

C
2
Q(x, y) dy
=

c
d
Q(x
1
(y), y) dy +

d
c
Q(x
2
(y), y) dy
= −

d
c
Q(x
1
(y), y) dy +

d
c
Q(x
2
(y), y) dy
=

d
c
(Q(x
2
(y), y) − Q(x
1
(y), y)) dy
=

d
c

Q(x, y)

x=x
2
(y)
x=x
1
(y)

dy
=

d
c

x
2
(y)
x
1
(y)
∂Q(x, y)
∂x
dx dy (by the Fundamental Theorem of Calculus)
=

R
∂Q
∂x
dA , and so
152 CHAPTER 4. LINE AND SURFACE INTEGRALS

C
f ··· dr =

C
P(x, y) dx +

C
Q(x, y) dy
= −

R
∂P
∂y
dA +

R
∂Q
∂x
dA
=

R
¸
∂Q
∂x

∂P
∂y

dA .
QED
Though we proved Green’s Theorem only for a simple region R, the theorem can also
be proved for more general regions (say, a union of simple regions).
2
Example 4.7. Evaluate

C
(x
2
+y
2
) dx+2xy dy, where C is the boundary (traversed coun-
terclockwise) of the region R = { (x, y) : 0 ≤ x ≤ 1, 2x
2
≤ y ≤ 2x }.
x
y
0
(1, 2)
2
1
C
Figure 4.3.2
Solution: R is the shaded region in Figure 4.3.2. By Green’s Theorem,
for P(x, y) = x
2
+ y
2
and Q(x, y) = 2xy, we have

C
(x
2
+ y
2
) dx + 2xy dy =

R
¸
∂Q
∂x

∂P
∂y

dA
=

R
(2y − 2y) dA =

R
0 dA = 0 .
We actually already knew that the answer was zero. Recall from Example 4.5 in
Section 4.2 that the vector field f(x, y) = (x
2
+ y
2
) i + 2xy j has a potential function
F(x, y) =
1
3
x
3
+ xy
2
, and so

C
f ··· dr = 0 by Corollary 4.6.
Example 4.8. Let f(x, y) = P(x, y) i + Q(x, y) j, where
P(x, y) =
−y
x
2
+ y
2
and Q(x, y) =
x
x
2
+ y
2
,
and let R = { (x, y) : 0 < x
2
+ y
2
≤ 1 }. For the boundary curve C : x
2
+ y
2
= 1, traversed
counterclockwise, it was shown in Exercise 9(b) in Section 4.2 that

C
f ··· dr = 2π. But
∂Q
∂x
=
y
2
− x
2
(x
2
+ y
2
)
2
=
∂P
∂y

R
¸
∂Q
∂x

∂P
∂y

dA =

R
0 dA = 0 .
This would seem to contradict Green’s Theorem. However, note that R is not the
entire region enclosed by C, since the point (0, 0) is not contained in R. That is, R has a
“hole” at the origin, so Green’s Theorem does not apply.
2
See TAYLOR and MANN, § 15.31 for a discussion of some of the difficulties involved when the boundary
curve is “complicated”.
4.3 Green’s Theorem 153
x
y
0
C
1
C
2
1
1
1/2
1/2
R

Figure 4.3.3 The annulus R
If we modify the region R to be the annulus
R = { (x, y) : 1/4 ≤ x
2
+ y
2
≤ 1 } (see Figure 4.3.3),
and take the “boundary” C of R to be C = C
1
∪ C
2
,
where C
1
is the unit circle x
2
+ y
2
= 1 traversed
counterclockwise and C
2
is the circle x
2
+ y
2
= 1/4
traversed clockwise, then it can be shown (see Ex-
ercise 8) that

C
f ··· dr = 0 .
We would still have

R

∂Q
∂x

∂P
∂y

dA = 0, so for this
R we would have

C
f ··· dr =

R
¸
∂Q
∂x

∂P
∂y

dA ,
which shows that Green’s Theorem holds for the annular region R.
It turns out that Green’s Theorem can be extended to multiply connected regions,
that is, regions like the annulus in Example 4.8, which have one or more regions cut
out from the interior, as opposed to discrete points being cut out. For such regions, the
“outer” boundary and the “inner” boundaries are traversed so that R is always on the
left side.
C
1
C
2
R
1
R
2

(a) Region R with one hole
C
1
C
2
C
3
R
1
R
2
◮ ◮
◭ ◭

(b) Region R with two holes
Figure 4.3.4 Multiply connected regions
The intuitive idea for why Green’s Theorem holds for multiply connected regions
is shown in Figure 4.3.4 above. The idea is to cut “slits” between the boundaries of a
multiply connected region R so that R is divided into subregions which do not have any
“holes”. For example, in Figure 4.3.4(a) the region R is the union of the regions R
1
and
R
2
, which are divided by the slits indicated by the dashed lines. Those slits are part
of the boundary of both R
1
and R
2
, and we traverse then in the manner indicated by
154 CHAPTER 4. LINE AND SURFACE INTEGRALS
the arrows. Notice that along each slit the boundary of R
1
is traversed in the opposite
direction as that of R
2
, which means that the line integrals of f along those slits cancel
each other out. Since R
1
and R
2
do not have holes in them, then Green’s Theorem holds
in each subregion, so that

bdy
of R
1
f ··· dr =

R
1
¸
∂Q
∂x

∂P
∂y

dA and

bdy
of R
2
f ··· dr =

R
2
¸
∂Q
∂x

∂P
∂y

dA .
But since the line integrals along the slits cancel out, we have

C
1
∪C
2
f ··· dr =

bdy
of R
1
f ··· dr +

bdy
of R
2
f ··· dr ,
and so

C
1
∪C
2
f ··· dr =

R
1
¸
∂Q
∂x

∂P
∂y

dA +

R
2
¸
∂Q
∂x

∂P
∂y

dA =

R
¸
∂Q
∂x

∂P
∂y

dA ,
which shows that Green’s Theorem holds in the region R. A similar argument shows
that the theorem holds in the region with two holes shown in Figure 4.3.4(b).
We know from Corollary 4.6 that when a smooth vector field f(x, y) = P(x, y) i+Q(x, y) j
on a region R (whose boundary is a piecewise smooth, simple closed curve C) has a
potential in R, then

C
f ··· dr = 0. And if the potential F(x, y) is smooth in R, then
∂F
∂x
= P
and
∂F
∂y
= Q, and so we know that

2
F
∂y ∂x
=

2
F
∂x ∂y

∂P
∂y
=
∂Q
∂x
in R.
Conversely, if
∂P
∂y
=
∂Q
∂x
in R then

C
f ··· dr =

R
¸
∂Q
∂x

∂P
∂y

dA =

R
0 dA = 0 .
For a simply connected region R (i.e. a region with no holes), the following can be
shown:
The following statements are equivalent for a simply connected region R in
2
:
(a) f(x, y) = P(x, y) i + Q(x, y) j has a smooth potential F(x, y) in R
(b)

C
f ··· dr is independent of the path for any curve C in R
(c)

C
f ··· dr = 0 for every simple closed curve C in R
(d)
∂P
∂y
=
∂Q
∂x
in R (in this case, the differential form Pdx + Qdy is exact)
4.3 Green’s Theorem 155

¨
Exercises
A
For Exercises 1-4, use Green’s Theorem to evaluate the given line integral around the
curve C, traversed counterclockwise.
1.

C
(x
2
− y
2
) dx + 2xy dy; C is the boundary of R = { (x, y) : 0 ≤ x ≤ 1, 2x
2
≤ y ≤ 2x }
2.

C
x
2
y dx + 2xy dy; C is the boundary of R = { (x, y) : 0 ≤ x ≤ 1, x
2
≤ y ≤ x }
3.

C
2y dx − 3x dy; C is the circle x
2
+ y
2
= 1
4.

C
(e
x
2
+ y
2
) dx + (e
y
2
+ x
2
) dy; C is the boundary of the triangle with vertices (0, 0),
(4, 0) and (0, 4)
5. Is there a potential F(x, y) for f(x, y) = (y
2
+ 3x
2
) i + 2xy j? If so, find one.
6. Is there a potential F(x, y) for f(x, y) = (x
3
cos(xy) + 2x sin(xy)) i + x
2
y cos(xy) j? If so,
find one.
7. Is there a potential F(x, y) for f(x, y) = (8xy + 3) i + 4(x
2
+ y) j? If so, find one.
8. Show that for any constants a, b and any closed simple curve C,

C
a dx + b dy = 0.
B
9. For the vector field f as in Example 4.8, show directly that

C
f ··· dr = 0, where C is
the boundary of the annulus R = { (x, y) : 1/4 ≤ x
2
+ y
2
≤ 1 } traversed so that R is
always on the left.
10. Evaluate

C
e
x
sin y dx + (y
3
+ e
x
cos y) dy, where C is the boundary of the rectangle
with vertices (1, −1), (1, 1), (−1, 1) and (−1, −1), traversed counterclockwise.
C
11. For a region R bounded by a simple closed curve C, show that the area A of R is
A = −

C
y dx =

C
x dy =
1
2

C
x dy − y dx ,
where C is traversed so that R is always on the left. (Hint: Use Green’s Theorem
and the fact that A =

R
1 dA.)
156 CHAPTER 4. LINE AND SURFACE INTEGRALS
4.4 Surface Integrals and the Divergence Theorem
In Section 4.1 we learned how to integrate along a curve. We will now learn how to
perform integration over a surface in
3
, such as a sphere or a paraboloid. Recall
from Section 1.8 how we identified points (x, y, z) on a curve C in
3
, parametrized by
x = x(t), y = y(t), z = z(t), a ≤ t ≤ b, with the terminal points of the position vector
r(t) = x(t)i + y(t)j + z(t)k for t in [a, b].
The idea behind a parametrization of a curve is that it “transforms” a subset of
1
(normally an interval [a, b]) into a curve in
2
or
3
(see Figure 4.4.1).

1
a
t
b
y
z
x
0
(x(a), y(a), z(a))
(x(t), y(t), z(t))
(x(b), y(b), z(b)) r(t)
C
x = x(t)
y = y(t)
z = z(t)
Figure 4.4.1 Parametrization of a curve C in
3
Similar to how we used a parametrization of a curve to define the line integral along
the curve, we will use a parametrization of a surface to define a surface integral. We
will use two variables, u and v, to parametrize a surface Σ in
3
: x = x(u, v), y = y(u, v),
z = z(u, v), for (u, v) in some region R in
2
(see Figure 4.4.2).
u
v
R

2
(u, v)
y
z
x
0
Σ
r(u, v)
x = x(u, v)
y = y(u, v)
z = z(u, v)
Figure 4.4.2 Parametrization of a surface Σ in
3
In this case, the position vector of a point on the surface Σ is given by the vector-
valued function
r(u, v) = x(u, v)i + y(u, v)j + z(u, v)k for (u, v) in R.
4.4 Surface Integrals and the Divergence Theorem 157
Since r(u, v) is a function of two variables, define the partial derivatives
∂r
∂u
and
∂r
∂v
for (u, v) in R by
∂r
∂u
(u, v) =
∂x
∂u
(u, v)i +
∂y
∂u
(u, v)j +
∂z
∂u
(u, v)k , and
∂r
∂v
(u, v) =
∂x
∂v
(u, v)i +
∂y
∂v
(u, v)j +
∂z
∂v
(u, v)k .
The parametrization of Σ can be thought of as “transforming” a region in
2
(in the
uv-plane) into a 2-dimensional surface in
3
. This parametrization of the surface is
sometimes called a patch, based on the idea of “patching” the region R onto Σ in the
grid-like manner shown in Figure 4.4.2.
In fact, those gridlines in R lead us to how we will define a surface integral over Σ.
Along the vertical gridlines in R, the variable u is constant. So those lines get mapped
to curves on Σ, and the variable u is constant along the position vector r(u, v). Thus, the
tangent vector to those curves at a point (u, v) is
∂r
∂v
. Similarly, the horizontal gridlines
in R get mapped to curves on Σ whose tangent vectors are
∂r
∂u
.
Now take a point (u, v) in R as, say, the lower left corner of one of the rectangular grid
sections in R, as shown in Figure 4.4.2. Suppose that this rectangle has a small width
and height of ∆u and ∆v, respectively. The corner points of that rectangle are (u, v),
(u +∆u, v), (u +∆u, v +∆v) and (u, v +∆v). So the area of that rectangle is A = ∆u ∆v. Then
that rectangle gets mapped by the parametrization onto some section of the surface
Σ which, for ∆u and ∆v small enough, will have a surface area (call it dσ) that is very
close to the area of the parallelogram which has adjacent sides r(u + ∆u, v) − r(u, v)
(corresponding to the line segment from (u, v) to (u + ∆u, v) in R) and r(u, v + ∆v) − r(u, v)
(corresponding to the line segment from (u, v) to (u, v + ∆v) in R). But by combining our
usual notion of a partial derivative (see Definition 2.3 in Section 2.2) with that of the
derivative of a vector-valued function (see Definition 1.12 in Section 1.8) applied to a
function of two variables, we have
∂r
∂u

r(u + ∆u, v) − r(u, v)
∆u
, and
∂r
∂v

r(u, v + ∆v) − r(u, v)
∆v
,
and so the surface area element dσ is approximately
¸
¸
¸(r(u + ∆u, v) − r(u, v)) ××× (r(u, v + ∆v) − r(u, v))
¸
¸
¸ ≈
¸
¸
¸
¸
¸
¸
(∆u
∂r
∂u
) ××× (∆v
∂r
∂v
)
¸
¸
¸
¸
¸
¸
=
¸
¸
¸
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸
¸
¸
¸
∆u ∆v
by Theorem 1.13 in Section 1.4. Thus, the total surface area S of Σ is approximately
the sum of all the quantities
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸ ∆u ∆v, summed over the rectangles in R. Taking
the limit of that sum as the diagonal of the largest rectangle goes to 0 gives
S =

R
¸
¸
¸
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸
¸
¸
¸
du dv . (4.26)
158 CHAPTER 4. LINE AND SURFACE INTEGRALS
We will write the double integral on the right using the special notation

Σ
dσ =

R
¸
¸
¸
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸
¸
¸
¸
du dv . (4.27)
This is a special case of a surface integral over the surface Σ, where the surface area
element dσ can be thought of as 1 dσ. Replacing 1 by a general real-valued function
f (x, y, z) defined in
3
, we have the following:
Definition 4.3. Let Σ be a surface in
3
parametrized by x = x(u, v), y = y(u, v),
z = z(u, v), for (u, v) in some region R in
2
. Let r(u, v) = x(u, v)i + y(u, v)j + z(u, v)k be the
position vector for any point on Σ, and let f (x, y, z) be a real-valued function defined on
some subset of
3
that contains Σ. The surface integral of f (x, y, z) over Σ is

Σ
f (x, y, z) dσ =

R
f (x(u, v), y(u, v), z(u, v))
¸
¸
¸
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸
¸
¸
¸
du dv . (4.28)
In particular, the surface area S of Σ is
S =

Σ
1 dσ . (4.29)
Example 4.9. A torus T is a surface obtained by revolving a circle of radius a in the
yz-plane around the z-axis, where the circle’s center is at a distance b from the z-axis
(0 < a < b), as in Figure 4.4.3. Find the surface area of T.
y
z
0
a
(y − b)
2
+ z
2
= a
2
u
b
(a) Circle in the yz-plane
x
y
z
v
a
(x,y,z)
(b) Torus T
Figure 4.4.3
Solution: For any point on the circle, the line segment from the center of the circle
to that point makes an angle u with the y-axis in the positive y direction (see Figure
4.4 Surface Integrals and the Divergence Theorem 159
4.4.3(a)). And as the circle revolves around the z-axis, the line segment from the origin
to the center of that circle sweeps out an angle v with the positive x-axis (see Figure
4.4.3(b)). Thus, the torus can be parametrized as:
x = (b + a cos u) cos v , y = (b + a cos u) sin v , z = a sin u , 0 ≤ u ≤ 2π , 0 ≤ v ≤ 2π
So for the position vector
r(u, v) = x(u, v)i + y(u, v)j + z(u, v)k
= (b + a cos u) cos v i + (b + a cos u) sin v j + a sin u k
we see that
∂r
∂u
= −a sin u cos v i − a sin u sin v j + a cos u k
∂r
∂v
= −(b + a cos u) sin v i + (b + a cos u) cos v j + 0k ,
and so computing the cross product gives
∂r
∂u
×××
∂r
∂v
= −a(b + a cos u) cos v cos u i − a(b + a cos u) sin v cos u j − a(b + a cos u) sin u k ,
which has magnitude
¸
¸
¸
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸
¸
¸
¸
= a(b + a cos u) .
Thus, the surface area of T is
S =

Σ
1 dσ
=

0

0
¸
¸
¸
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸
¸
¸
¸
du dv
=

0

0
a(b + a cos u) du dv
=

0

abu + a
2
sin u

u=2π
u=0

dv
=

0
2πab dv
= 4π
2
ab
Since
∂r
∂u
and
∂r
∂v
are tangent to the surface Σ (i.e. lie in the tangent plane to Σ at each
point on Σ), then their cross product
∂r
∂u
×××
∂r
∂v
is perpendicular to the tangent plane to
160 CHAPTER 4. LINE AND SURFACE INTEGRALS
the surface at each point of Σ. Thus,

Σ
f (x, y, z) dσ =

R
f (x(u, v), y(u, v), z(u, v)) n dσ ,
where n =
∂r
∂u
×××
∂r
∂v
. We say that n is a normal vector to Σ.
y
z
x
0
Figure 4.4.4
Recall that normal vectors to a plane can point in two
opposite directions. By an outward unit normal vector
to a surface Σ, we will mean the unit vector that is normal
to Σ and points away from the “top” (or “outer” part) of the
surface. This is a hazy definition, but the picture in Figure
4.4.4 gives a better idea of what outward normal vectors
look like, in the case of a sphere. With this idea in mind,
we make the following definition of a surface integral of a
3-dimensional vector field over a surface:
Definition 4.4. Let Σ be a surface in
3
and let f(x, y, z) = f
1
(x, y, z)i + f
2
(x, y, z)j +
f
3
(x, y, z)k be a vector field defined on some subset of
3
that contains Σ. The surface
integral of f over Σ is

Σ
f ··· dσ =

Σ
f ··· ndσ , (4.30)
where, at any point on Σ, n is the outward unit normal vector to Σ.
Note in the above definition that the dot product inside the integral on the right is
a real-valued function, and hence we can use Definition 4.3 to evaluate the integral.
Example 4.10. Evaluate the surface integral

Σ
f ··· dσ, where f(x, y, z) = yzi + xzj + xyk
and Σ is the part of the plane x +y +z = 1 with x ≥ 0, y ≥ 0, and z ≥ 0, with the outward
unit normal n pointing in the positive z direction (see Figure 4.4.5).
y
z
x
0
1
1
1
Σ
x + y + z = 1
n
Figure 4.4.5
Solution: Since the vector v = (1, 1, 1) is normal to the plane x + y +
z = 1 (why?), then dividing v by its length yields the outward unit
normal vector n =

1

3
,
1

3
,
1

3

. We now need to parametrize Σ. As
we can see from Figure 4.4.5, projecting Σ onto the xy-plane yields
a triangular region R = { (x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1 − x }. Thus, using
(u, v) instead of (x, y), we see that
x = u, y = v, z = 1 − (u + v), for 0 ≤ u ≤ 1, 0 ≤ v ≤ 1 − u
4.4 Surface Integrals and the Divergence Theorem 161
is a parametrization of Σ over R (since z = 1 − (x + y) on Σ). So on Σ,
f ··· n = (yz, xz, xy) ···
¸
1

3
,
1

3
,
1

3

=
1

3
(yz + xz + xy)
=
1

3
((x + y)z + xy) =
1

3
((u + v)(1 − (u + v)) + uv)
=
1

3
((u + v) − (u + v)
2
+ uv)
for (u, v) in R, and for r(u, v) = x(u, v)i + y(u, v)j + z(u, v)k = ui + vj + (1 − (u + v))k we have
∂r
∂u
×××
∂r
∂v
= (1, 0, −1) ××× (0, 1, −1) = (1, 1, 1) ⇒
¸
¸
¸
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸
¸
¸
¸
=

3 .
Thus, integrating over R using vertical slices (e.g. as indicated by the dashed line in
Figure 4.4.5) gives

Σ
f ··· dσ =

Σ
f ··· ndσ
=

R
(f(x(u, v), y(u, v), z(u, v)) ··· n)
¸
¸
¸
¸
¸
¸
∂r
∂u
×××
∂r
∂v
¸
¸
¸
¸
¸
¸
dv du
=

1
0

1−u
0
1

3
((u + v) − (u + v)
2
+ uv)

3 dv du
=

1
0

¸
¸
¸
¸
¸
¸
(u + v)
2
2

(u + v)
3
3
+
uv
2
2

v=1−u
v=0

¸
¸
¸
¸
¸
¸
du
=

1
0
¸
1
6
+
u
2

3u
2
2
+
5u
3
6

du
=
u
6
+
u
2
4

u
3
2
+
5u
4
24

1
0
=
1
8
.
Computing surface integrals can often be tedious, especially when the formula for
the outward unit normal vector at each point of Σ changes. The following theorem pro-
vides an easier way in the case when Σ is a closed surface, that is, when Σ encloses
a bounded solid in
3
. For example, spheres, cubes, and ellipsoids are closed surfaces,
but planes and paraboloids are not.
162 CHAPTER 4. LINE AND SURFACE INTEGRALS
Theorem 4.8. (Divergence Theorem) Let Σ be a closed surface in
3
which bounds
a solid S , and let f(x, y, z) = f
1
(x, y, z)i + f
2
(x, y, z)j + f
3
(x, y, z)k be a vector field defined on
some subset of
3
that contains Σ. Then

Σ
f ··· dσ =

S
div f dV , (4.31)
where
div f =
∂f
1
∂x
+
∂f
2
∂y
+
∂f
3
∂z
(4.32)
is called the divergence of f.
The proof of the Divergence Theoremis very similar to the proof of Green’s Theorem,
i.e. it is first proved for the simple case when the solid S is bounded above by one
surface, bounded below by another surface, and bounded laterally by one or more
surfaces. The proof can then be extended to more general solids.
3
Example 4.11. Evaluate

Σ
f ··· dσ, where f(x, y, z) = xi +yj +zk and Σ is the unit sphere
x
2
+ y
2
+ z
2
= 1.
Solution: We see that div f = 1 + 1 + 1 = 3, so

Σ
f ··· dσ =

S
div f dV =

S
3 dV
= 3

S
1 dV = 3 vol(S ) = 3 ·
4π(1)
3
3
= 4π .
In physical applications, the surface integral

Σ
f··· dσ is often referred to as the flux
of f through the surface Σ. For example, if f represents the velocity field of a fluid,
then the flux is the net quantity of fluid to flow through the surface Σ per unit time.
A positive flux means there is a net flow out of the surface (i.e. in the direction of the
outward unit normal vector n), while a negative flux indicates a net flow inward (in
the direction of −n).
The term divergence comes from interpreting div f as a measure of how much a
vector field “diverges” from a point. This is best seen by using another definition of
div f which is equivalent
4
to the definition given by formula (4.32). Namely, for a point
(x, y, z) in
3
,
div f(x, y, z) = lim
V→0
1
V

Σ
f ··· dσ , (4.33)
3
See TAYLOR and MANN, § 15.6 for the details.
4
See SCHEY, p. 36-39, for an intuitive discussion of this.
4.4 Surface Integrals and the Divergence Theorem 163
where V is the volume enclosed by a closed surface Σ around the point (x, y, z). In the
limit, V → 0 means that we take smaller and smaller closed surfaces around (x, y, z),
which means that the volumes they enclose are going to zero. It can be shown that this
limit is independent of the shapes of those surfaces. Notice that the limit being taken
is of the ratio of the flux through a surface to the volume enclosed by that surface,
which gives a rough measure of the flow “leaving” a point, as we mentioned. Vector
fields which have zero divergence are often called solenoidal fields.
The following theorem is a simple consequence of formula (4.33).
Theorem 4.9. If the flux of a vector field f is zero through every closed surface con-
taining a given point, then div f = 0 at that point.
Proof: By formula (4.33), at the given point (x, y, z) we have
div f(x, y, z) = lim
V→0
1
V

Σ
f ··· dσ for closed surfaces Σ containing (x, y, z), so
= lim
V→0
1
V
(0) by our assumption that the flux through each Σ is zero, so
= lim
V→0
0
= 0 . QED
Lastly, we note that sometimes the notation

Σ
f (x, y, z) dσ and

Σ
f ··· dσ
is used to denote surface integrals of scalar and vector fields, respectively, over closed
surfaces. Especially in physics texts, it is common to see simply

Σ

Σ
.

¨
Exercises
A
For Exercises 1-4, use the Divergence Theorem to evaluate the surface integral

Σ
f···dσ
of the given vector field f(x, y, z) over the surface Σ.
1. f(x, y, z) = xi + 2yj + 3zk, Σ : x
2
+ y
2
+ z
2
= 9
2. f(x, y, z) = xi + yj + zk, Σ : boundary of the solid cube S = { (x, y, z) : 0 ≤ x, y, z ≤ 1 }
3. f(x, y, z) = x
3
i + y
3
j + z
3
k, Σ : x
2
+ y
2
+ z
2
= 1
4. f(x, y, z) = 2i + 3j + 5k, Σ : x
2
+ y
2
+ z
2
= 1
164 CHAPTER 4. LINE AND SURFACE INTEGRALS
B
5. Show that the flux of any constant vector field through any closed surface is zero.
6. Evaluate the surface integral from Exercise 2 without using the Divergence Theo-
rem, i.e. using only Definition 4.3, as in Example 4.10. Note that there will be a
different outward unit normal vector to each of the six faces of the cube.
7. Evaluate the surface integral

Σ
f ··· dσ, where f(x, y, z) = x
2
i + xyj + zk and Σ is the
part of the plane 6x + 3y + 2z = 6 with x ≥ 0, y ≥ 0, and z ≥ 0, with the outward unit
normal n pointing in the positive z direction.
8. Use a surface integral to show that the surface area of a sphere of radius r is 4πr
2
.
(Hint: Use spherical coordinates to parametrize the sphere.)
9. Use a surface integral to show that the surface area of a right circular cone of
radius R and height h is πR

h
2
+ R
2
. (Hint: Use the parametrization x = r cos θ,
y = r sin θ, z =
h
R
r, for 0 ≤ r ≤ R and 0 ≤ θ ≤ 2π.)
10. The ellipsoid
x
2
a
2
+
y
2
b
2
+
z
2
c
2
= 1 can be parametrized using ellipsoidal coordinates
x = a sin φ cos θ , y = b sin φ sin θ , z = c cos φ , for 0 ≤ θ ≤ 2π and 0 ≤ φ ≤ π.
Show that the surface area S of the ellipsoid is
S =

π
0

0
sin φ

a
2
b
2
cos
2
φ + c
2
(a
2
sin
2
θ + b
2
cos
2
θ) sin
2
φ dθ dφ .
(Note: The above double integral can not be evaluated by elementary means. For
specific values of a, b and c it can be evaluated using numerical methods. An
alternative is to express the surface area in terms of elliptic integrals.
5
)
C
11. Use Definition 4.3 to prove that the surface area S over a region R in
2
of a
surface z = f (x, y) is given by the formula
S =

R

1 +

∂f
∂x

2
+

∂f
∂y

2
dA .
(Hint: Think of the parametrization of the surface.)
5
BOWMAN, F., Introduction to Elliptic Functions, with Applications, New York: Dover, 1961, § III.7.
4.5 Stokes’ Theorem 165
4.5 Stokes’ Theorem
So far the only types of line integrals which we have discussed are those along curves
in
2
. But the definitions and properties which were covered in Sections 4.1 and 4.2
can easily be extended to include functions of three variables, so that we can now
discuss line integrals along curves in
3
.
Definition 4.5. For a real-valued function f (x, y, z) and a curve C in
3
, parametrized
by x = x(t), y = y(t), z = z(t), a ≤ t ≤ b, the line integral of f (x, y, z) along C with
respect to arc length s is

C
f (x, y, z) ds =

b
a
f (x(t), y(t), z(t))

x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt . (4.34)
The line integral of f (x, y, z) along C with respect to x is

C
f (x, y, z) dx =

b
a
f (x(t), y(t), z(t)) x

(t) dt . (4.35)
The line integral of f (x, y, z) along C with respect to y is

C
f (x, y, z) dy =

b
a
f (x(t), y(t), z(t)) y

(t) dt . (4.36)
The line integral of f (x, y, z) along C with respect to z is

C
f (x, y, z) dz =

b
a
f (x(t), y(t), z(t)) z

(t) dt . (4.37)
Similar to the two-variable case, if f (x, y, z) ≥ 0 then the line integral

C
f (x, y, z) ds
can be thought of as the total area of the “picket fence” of height f (x, y, z) at each point
along the curve C in
3
.
Vector fields in
3
are defined in a similar fashion to those in
2
, which allows us
to define the line integral of a vector field along a curve in
3
.
Definition 4.6. For a vector field f(x, y, z) = P(x, y, z) i + Q(x, y, z) j + R(x, y, z) k and a
curve C in
3
with a smooth parametrization x = x(t), y = y(t), z = z(t), a ≤ t ≤ b, the
line integral of f along C is

C
f ··· dr =

C
P(x, y, z) dx +

C
Q(x, y, z) dy +

C
R(x, y, z) dz (4.38)
=

b
a
f(x(t), y(t), z(t)) ··· r

(t) dt , (4.39)
where r(t) = x(t) i + y(t) j + z(t) k is the position vector for points on C.
166 CHAPTER 4. LINE AND SURFACE INTEGRALS
Similar to the two-variable case, if f(x, y, z) represents the force applied to an object
at a point (x, y, z) then the line integral

C
f ··· dr represents the work done by that force
in moving the object along the curve C in
3
.
Some of the most important results we will need for line integrals in
3
are stated
below without proof (the proofs are similar to their two-variable equivalents).
Theorem4.10. For a vector field f(x, y, z) = P(x, y, z) i+Q(x, y, z) j+R(x, y, z) k and a curve
C with a smooth parametrization x = x(t), y = y(t), z = z(t), a ≤ t ≤ b and position vector
r(t) = x(t) i + y(t) j + z(t) k,

C
f ··· dr =

C
f ··· Tds , (4.40)
where T(t) =
r

(t)
r

(t)
is the unit tangent vector to C at (x(t), y(t), z(t)).
Theorem 4.11. (Chain Rule) If w = f (x, y, z) is a continuously differentiable function
of x, y, and z, and x = x(t), y = y(t) and z = z(t) are differentiable functions of t, then w is
a differentiable function of t, and
dw
dt
=
∂w
∂x
dx
dt
+
∂w
∂y
dy
dt
+
∂w
∂z
dz
dt
. (4.41)
Also, if x = x(t
1
, t
2
), y = y(t
1
, t
2
) and z = z(t
1
, t
2
) are continuously differentiable function of
(t
1
, t
2
), then
6
∂w
∂t
1
=
∂w
∂x
∂x
∂t
1
+
∂w
∂y
∂y
∂t
1
+
∂w
∂z
∂z
∂t
1
(4.42)
and
∂w
∂t
2
=
∂w
∂x
∂x
∂t
2
+
∂w
∂y
∂y
∂t
2
+
∂w
∂z
∂z
∂t
2
. (4.43)
Theorem 4.12. Let f(x, y, z) = P(x, y, z) i +Q(x, y, z) j +R(x, y, z) k be a vector field in some
solid S , with P, Q and R continuously differentiable functions on S . Let C be a smooth
curve in S parametrized by x = x(t), y = y(t), z = z(t), a ≤ t ≤ b. Suppose that there is a
real-valued function F(x, y, z) such that ∇F = f on S . Then

C
f ··· dr = F(B) − F(A) , (4.44)
where A = (x(a), y(a), z(a)) and B = (x(b), y(b), z(b)) are the endpoints of C.
Corollary 4.13. If a vector field f has a potential in a solid S , then

C
f··· dr = 0 for any
closed curve C in S (i.e.

C
∇F ··· dr = 0 for any real-valued function F(x, y, z)).
6
See TAYLOR and MANN, § 6.5 for a proof.
4.5 Stokes’ Theorem 167
Example 4.12. Let f (x, y, z) = z and let C be the curve in
3
parametrized by
x = t sin t , y = t cos t , z = t , 0 ≤ t ≤ 8π .
Evaluate

C
f (x, y, z) ds. (Note: C is called a conical helix. See Figure 4.5.1).
Solution: Since x

(t) = sin t + t cos t, y

(t) = cos t − t sin t, and z

(t) = 1, we have
x

(t)
2
+ y

(t)
2
+ z

(t)
2
= (sin
2
t + 2t sin t cos t + t
2
cos
2
t) + (cos
2
t − 2t sin t cos t + t
2
sin
2
t) + 1
= t
2
(sin
2
t + cos
2
t) + sin
2
t + cos
2
t + 1
= t
2
+ 2 ,
so since f (x(t), y(t), z(t)) = z(t) = t along the curve C, then

C
f (x, y, z) ds =

0
f (x(t), y(t), z(t))

x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt
=

0
t

t
2
+ 2 dt
=
¸
1
3
(t
2
+ 2)
3/2

0
=
1
3

(64π
2
+ 2)
3/2
− 2

2

.
-25
-20
-15
-10
-5
0
5
10
15
20
25
-25
-20
-15
-10
-5
0
5
10
15
20
25
30
0
5
10
15
20
25
30
z
x
y
z
t = 0
t = 8π
Figure 4.5.1 Conical helix C
Example 4.13. Let f(x, y, z) = x i + y j + 2z k be a vector field in
3
. Using the same
curve C from Example 4.12, evaluate

C
f ··· dr.
Solution: It is easy to see that F(x, y, z) =
x
2
2
+
y
2
2
+ z
2
is a potential for f(x, y, z) (i.e.
168 CHAPTER 4. LINE AND SURFACE INTEGRALS
∇F = f). So by Theorem 4.12 we know that

C
f ··· dr = F(B) − F(A) , where A = (x(0), y(0), z(0)) and B = (x(8π), y(8π), z(8π)), so
= F(8π sin 8π, 8π cos 8π, 8π) − F(0 sin 0, 0 cos 0, 0)
= F(0, 8π, 8π) − F(0, 0, 0)
= 0 +
(8π)
2
2
+ (8π)
2
− (0 + 0 + 0) = 96π
2
.
We will now discuss a generalization of Green’s Theorem in
2
to orientable surfaces
in
3
, called Stokes’ Theorem. A surface Σ in
3
is orientable if there is a continuous
vector field N in
3
such that N is nonzero and normal to Σ (i.e. perpendicular to the
tangent plane) at each point of Σ. We say that such an N is a normal vector field.
y
z
x
0
N
−N
Figure 4.5.2
For example, the unit sphere x
2
+y
2
+z
2
= 1 is orientable, since
the continuous vector field N(x, y, z) = x i+y j+z k is nonzero and
normal to the sphere at each point. In fact, −N(x, y, z) is another
normal vector field (see Figure 4.5.2). We see in this case that
N(x, y, z) is what we have called an outward normal vector, and
−N(x, y, z) is an inward normal vector. These “outward” and
“inward” normal vector fields on the sphere correspond to an
“outer” and “inner” side, respectively, of the sphere. That is,
we say that the sphere is a two-sided surface. Roughly, “two-
sided” means “orientable”. Other examples of two-sided, and
hence orientable, surfaces are cylinders, paraboloids, ellipsoids, and planes.
You may be wondering what kind of surface would not have two sides. An example
is the Möbius strip, which is constructed by taking a thin rectangle and connecting
its ends at the opposite corners, resulting in a “twisted” strip (see Figure 4.5.3).
A
B A
B
−→
(a) Connect A to A and B to B along the ends
A

A

(b) Not orientable
Figure 4.5.3 Möbius strip
If you imagine walking along a line down the center of the Möbius strip, as in Figure
4.5.3(b), then you arrive back at the same place from which you started but upside
down! That is, your orientation changed even though your motion was continuous
4.5 Stokes’ Theorem 169
along that center line. Informally, thinking of your vertical direction as a normal
vector field along the strip, there is a discontinuity at your starting point (and, in
fact, at every point) since your vertical direction takes two different values there. The
Möbius strip has only one side, and hence is nonorientable.
7
For an orientable surface Σ which has a boundary curve C, pick a unit normal vector
n such that if you walked along C with your head pointing in the direction of n, then
the surface would be on your left. We say in this situation that n is a positive unit
normal vector and that C is traversed n-positively. We can now state Stokes’ Theorem:
Theorem 4.14. (Stokes’ Theorem) Let Σ be an orientable surface in
3
whose
boundary is a simple closed curve C, and let f(x, y, z) = P(x, y, z)i + Q(x, y, z)j + R(x, y, z)k
be a smooth vector field defined on some subset of
3
that contains Σ. Then

C
f ··· dr =

Σ
(curl f ) ··· ndσ , (4.45)
where
curl f =
¸
∂R
∂y

∂Q
∂z

i +
¸
∂P
∂z

∂R
∂x

j +
¸
∂Q
∂x

∂P
∂y

k , (4.46)
n is a positive unit normal vector over Σ, and C is traversed n-positively.
Proof: As the general case is beyond the scope of this text, we will prove the theorem
only for the special case where Σ is the graph of z = z(x, y) for some smooth real-valued
function z(x, y), with (x, y) varying over a region D in
2
.
y
z
x
0
n
(x, y)
D
C
D
C
Σ : z = z(x, y)
Figure 4.5.4
Projecting Σ onto the xy-plane, we see that the
closed curve C (the boundary curve of Σ) projects onto
a closed curve C
D
which is the boundary curve of D
(see Figure 4.5.4). Assuming that C has a smooth
parametrization, its projection C
D
in the xy-plane
also has a smooth parametrization, say
C
D
: x = x(t) , y = y(t) , a ≤ t ≤ b ,
and so C can be parametrized (in
3
) as
C : x = x(t) , y = y(t) , z = z(x(t), y(t)) , a ≤ t ≤ b ,
since the curve C is part of the surface z = z(x, y). Now, by the Chain Rule (Theorem
4.4 in Section 4.2), for z = z(x(t), y(t)) as a function of t, we know that
z

(t) =
∂z
∂x
x

(t) +
∂z
∂y
y

(t) ,
7
For further discussion of orientability, see O’NEILL, § IV.7.
170 CHAPTER 4. LINE AND SURFACE INTEGRALS
and so

C
f ··· dr =

C
P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz
=

b
a
¸
P x

(t) + Qy

(t) + R
¸
∂z
∂x
x

(t) +
∂z
∂y
y

(t)

dt
=

b
a
¸¸
P + R
∂z
∂x

x

(t) +
¸
Q + R
∂z
∂y

y

(t)

dt
=

C
D
˜
P(x, y) dx +
˜
Q(x, y) dy ,
where
˜
P(x, y) = P(x, y, z(x, y)) + R(x, y, z(x, y))
∂z
∂x
(x, y) , and
˜
Q(x, y) = Q(x, y, z(x, y)) + R(x, y, z(x, y))
∂z
∂y
(x, y)
for (x, y) in D. Thus, by Green’s Theorem applied to the region D, we have

C
f ··· dr =

D
¸

˜
Q
∂x

˜
P
∂y

dA . (4.47)
Thus,

˜
Q
∂x
=

∂x
¸
Q(x, y, z(x, y)) + R(x, y, z(x, y))
∂z
∂y
(x, y)

, so by the Product Rule we get
=

∂x
(Q(x, y, z(x, y))) +
¸

∂x
R(x, y, z(x, y))

∂z
∂y
(x, y) + R(x, y, z(x, y))

∂x
¸
∂z
∂y
(x, y)

.
Now, by formula (4.42) in Theorem 4.11, we have

∂x
(Q(x, y, z(x, y))) =
∂Q
∂x
∂x
∂x
+
∂Q
∂y
∂y
∂x
+
∂Q
∂z
∂z
∂x
=
∂Q
∂x
· 1 +
∂Q
∂y
· 0 +
∂Q
∂z
∂z
∂x
=
∂Q
∂x
+
∂Q
∂z
∂z
∂x
.
Similarly,

∂x
(R(x, y, z(x, y))) =
∂R
∂x
+
∂R
∂z
∂z
∂x
.
4.5 Stokes’ Theorem 171
Thus,

˜
Q
∂x
=
∂Q
∂x
+
∂Q
∂z
∂z
∂x
+
¸
∂R
∂x
+
∂R
∂z
∂z
∂x

∂z
∂y
+ R(x, y, z(x, y))

2
z
∂x ∂y
=
∂Q
∂x
+
∂Q
∂z
∂z
∂x
+
∂R
∂x
∂z
∂y
+
∂R
∂z
∂z
∂x
∂z
∂y
+ R

2
z
∂x ∂y
.
In a similar fashion, we can calculate

˜
P
∂y
=
∂P
∂y
+
∂P
∂z
∂z
∂y
+
∂R
∂y
∂z
∂x
+
∂R
∂z
∂z
∂y
∂z
∂x
+ R

2
z
∂y ∂x
.
So subtracting gives

˜
Q
∂x

˜
P
∂y
=
¸
∂Q
∂z

∂R
∂y

∂z
∂x
+
¸
∂R
∂x

∂P
∂z

∂z
∂y
+
¸
∂Q
∂x

∂P
∂y

(4.48)
since

2
z
∂x ∂y
=

2
z
∂y ∂x
by the smoothness of z = z(x, y). Hence, by equation (4.47),

C
f ··· dr =

D
¸

¸
∂R
∂y

∂Q
∂z

∂z
∂x

¸
∂P
∂z

∂R
∂x

∂z
∂y
+
¸
∂Q
∂x

∂P
∂y

dA (4.49)
after factoring out a −1 from the terms in the first two products in equation (4.48).
Now, recall from Section 2.3 (see p.76) that the vector N = −
∂z
∂x
i −
∂z
∂y
j + k is normal
to the tangent plane to the surface z = z(x, y) at each point of Σ. Thus,
n =
N
¸
¸
¸N
¸
¸
¸
=

∂z
∂x
i −
∂z
∂y
j + k

1 +

∂z
∂x

2
+

∂z
∂y

2
is in fact a positive unit normal vector to Σ (see Figure 4.5.4). Hence, using the
parametrization r(x, y) = x i + y j + z(x, y) k, for (x, y) in D, of the surface Σ, we have
∂r
∂x
= i +
∂z
∂x
k and
∂r
∂y
= j +
∂z
∂y
k, and so
¸
¸
¸
∂r
∂x
×××
∂r
∂y
¸
¸
¸ =

1 +

∂z
∂x

2
+

∂z
∂y

2
. So we see that
using formula (4.46) for curl f, we have

Σ
(curl f) ··· ndσ =

D
(curl f ) ··· n
¸
¸
¸
¸
¸
¸
∂r
∂x
×××
∂r
∂y
¸
¸
¸
¸
¸
¸
dA
=

D
¸¸
∂R
∂y

∂Q
∂z

i +
¸
∂P
∂z

∂R
∂x

j +
¸
∂Q
∂x

∂P
∂y

k

···
¸

∂z
∂x
i −
∂z
∂y
j + k

dA
=

D
¸

¸
∂R
∂y

∂Q
∂z

∂z
∂x

¸
∂P
∂z

∂R
∂x

∂z
∂y
+
¸
∂Q
∂x

∂P
∂y

dA ,
which, upon comparing to equation (4.49), proves the Theorem. QED
172 CHAPTER 4. LINE AND SURFACE INTEGRALS
Note: The condition in Stokes’ Theorem that the surface Σ have a (continuously
varying) positive unit normal vector n and a boundary curve C traversed n-positively
can be expressed more precisely as follows: if r(t) is the position vector for C and
T(t) = r

(t)/r

(t) is the unit tangent vector to C, then the vectors T, n, T××× n form a
right-handed system.
Also, it should be noted that Stokes’ Theorem holds even when the boundary curve
C is piecewise smooth.
Example 4.14. Verify Stokes’ Theorem for f(x, y, z) = z i + x j + y k when Σ is the
paraboloid z = x
2
+ y
2
such that z ≤ 1 (see Figure 4.5.5).
y
z
x
0
n
C
Σ
1
Figure 4.5.5 z = x
2
+ y
2
Solution: The positive unit normal vector to the surface
z = z(x, y) = x
2
+ y
2
is
n =

∂z
∂x
i −
∂z
∂y
j + k

1 +

∂z
∂x

2
+

∂z
∂y

2
=
−2x i − 2y j + k

1 + 4x
2
+ 4y
2
,
and curl f = (1 − 0) i + (1 − 0) j + (1 − 0) k = i + j + k, so
(curl f ) ··· n = (−2x − 2y + 1)/

1 + 4x
2
+ 4y
2
.
Since Σ can be parametrized as r(x, y) = x i + y j + (x
2
+ y
2
) k
for (x, y) in the region D = { (x, y) : x
2
+ y
2
≤ 1 }, then

Σ
(curl f ) ··· ndσ =

D
(curl f ) ··· n
¸
¸
¸
¸
¸
¸
∂r
∂x
×××
∂r
∂y
¸
¸
¸
¸
¸
¸
dA
=

D
−2x − 2y + 1

1 + 4x
2
+ 4y
2

1 + 4x
2
+ 4y
2
dA
=

D
(−2x − 2y + 1) dA , so switching to polar coordinates gives
=

0

1
0
(−2r cos θ − 2r sin θ + 1)r dr dθ
=

0

1
0
(−2r
2
cos θ − 2r
2
sin θ + r) dr dθ
=

0

2r
3
3
cos θ −
2r
3
3
sin θ +
r
2
2

r=1
r=0

=

0

2
3
cos θ −
2
3
sin θ +
1
2

= −
2
3
sin θ +
2
3
cos θ +
1
2
θ

0
= π .
4.5 Stokes’ Theorem 173
The boundary curve C is the unit circle x
2
+ y
2
= 1 laying in the plane z = 1 (see
Figure 4.5.5), which can be parametrized as x = cos t, y = sin t, z = 1 for 0 ≤ t ≤ 2π. So

C
f ··· dr =

0
((1)(−sin t) + (cos t)(cos t) + (sin t)(0)) dt
=

0
¸
−sin t +
1 + cos 2t
2

dt
¸
here we used cos
2
t =
1 + cos 2t
2

= cos t +
t
2
+
sin 2t
4

0
= π .
So we see that

C
f ··· dr =

Σ
(curl f ) ··· ndσ, as predicted by Stokes’ Theorem.
The line integral in the preceding example was far simpler to calculate than the
surface integral, but this will not always be the case.
Example 4.15. Let Σ be the elliptic paraboloid z =
x
2
4
+
y
2
9
for z ≤ 1, and let C be its
boundary curve. Calculate

C
f ··· dr for f(x, y, z) = (9xz + 2y)i + (2x + y
2
)j + (−2y
2
+ 2z)k,
where C is traversed counterclockwise.
Solution: The surface is similar to the one in Example 4.14, except now the boundary
curve C is the ellipse
x
2
4
+
y
2
9
= 1 laying in the plane z = 1. In this case, using Stokes’
Theorem is easier than computing the line integral directly. As in Example 4.14, at
each point (x, y, z(x, y)) on the surface z = z(x, y) =
x
2
4
+
y
2
9
the vector
n =

∂z
∂x
i −
∂z
∂y
j + k

1 +

∂z
∂x

2
+

∂z
∂y

2
=

x
2
i −
2y
9
j + k

1 +
x
2
4
+
4y
2
9
,
is a positive unit normal vector to Σ. And calculating the curl of f gives
curl f = (−4y − 0)i + (9x − 0)j + (2 − 2)k = −4y i + 9x j + 0 k ,
so
(curl f ) ··· n =
(−4y)(−
x
2
) + (9x)(−
2y
9
) + (0)(1)

1 +
x
2
4
+
4y
2
9
=
2xy − 2xy + 0

1 +
x
2
4
+
4y
2
9
= 0 ,
and so by Stokes’ Theorem

C
f ··· dr =

Σ
(curl f ) ··· ndσ =

Σ
0 dσ = 0 .
174 CHAPTER 4. LINE AND SURFACE INTEGRALS
In physical applications, for a simple closed curve C the line integral

C
f··· dr is often
called the circulation of f around C. For example, if E represents the electrostatic
field due to a point charge, then it turns out
8
that curl E = 0, which means that the
circulation

C
E ··· dr = 0 by Stokes’ Theorem. Vector fields which have zero curl are
often called irrotational fields.
In fact, the term curl was created by the 19
th
century Scottish physicist James Clerk
Maxwell in his study of electromagnetism, where it is used extensively. In physics,
the curl is interpreted as a measure of circulation density. This is best seen by using
another definition of curl f which is equivalent
9
to the definition given by formula
(4.46). Namely, for a point (x, y, z) in
3
,
n··· (curl f )(x, y, z) = lim
S →0
1
S

C
f ··· dr , (4.50)
where S is the surface area of a surface Σ containing the point (x, y, z) and with a
simple closed boundary curve C and positive unit normal vector n at (x, y, z). In the
limit, think of the curve C shrinking to the point (x, y, z), which causes Σ, the surface it
bounds, to have smaller and smaller surface area. That ratio of circulation to surface
area in the limit is what makes the curl a rough measure of circulation density (i.e.
circulation per unit area).
x
y
0
f
Figure 4.5.6 Curl and rotation
An idea of how the curl of a vector field is related
to rotation is shown in Figure 4.5.6. Suppose we
have a vector field f(x, y, z) which is always parallel
to the xy-plane at each point (x, y, z) and that the vec-
tors grow larger the further the point (x, y, z) is from
the y-axis. For example, f(x, y, z) = (1 + x
2
) j. Think
of the vector field as representing the flow of wa-
ter, and imagine dropping two wheels with paddles
into that water flow, as in Figure 4.5.6. Since the
flow is stronger (i.e. the magnitude of f is larger) as
you move away from the y-axis, then such a wheel
would rotate counterclockwise if it were dropped to
the right of the y-axis, and it would rotate clockwise if it were dropped to the left of
the y-axis. In both cases the curl would be nonzero (curl f(x, y, z) = 2x k in our example)
and would obey the right-hand rule, that is, curl f(x, y, z) points in the direction of your
thumb as you cup your right hand in the direction of the rotation of the wheel. So
the curl points outward (in the positive z-direction) if x > 0 and points inward (in the
negative z-direction) if x < 0. Notice that if all the vectors had the same direction and
the same magnitude, then the wheels would not rotate and hence there would be no
curl (which is why such fields are called irrotational, meaning no rotation).
8
See Ch. 2 in REITZ, MILFORD and CHRISTY.
9
See SCHEY, p. 78-81, for the derivation.
4.5 Stokes’ Theorem 175
Finally, by Stokes’ Theorem, we know that if C is a simple closed curve in some solid
region S in
3
and if f(x, y, z) is a smooth vector field such that curl f = 0 in S , then

C
f ··· dr =

Σ
(curl f ) ··· ndσ =

Σ
0 ··· ndσ =

Σ
0 dσ = 0 ,
where Σ is any orientable surface inside S whose boundary is C (such a surface is
sometimes called a capping surface for C). So similar to the two-variable case, we
have a three-dimensional version of a result from Section 4.3, for solid regions in
3
which are simply connected (i.e. regions having no holes):
The following statements are equivalent for a simply connected solid region S in

3
:
(a) f(x, y, z) = P(x, y, z) i + Q(x, y, z) j + R(x, y, z) k has a smooth potential F(x, y, z) in S
(b)

C
f ··· dr is independent of the path for any curve C in S
(c)

C
f ··· dr = 0 for every simple closed curve C in S
(d)
∂R
∂y
=
∂Q
∂z
,
∂P
∂z
=
∂R
∂x
, and
∂Q
∂x
=
∂P
∂y
in S (i.e. curl f = 0 in S )
Part (d) is also a way of saying that the differential form Pdx + Qdy + Rdz is exact.
Example 4.16. Determine if the vector field f(x, y, z) = xyz i + xz j + xy k has a potential
in
3
.
Solution: Since
3
is simply connected, we just need to check whether curl f = 0
throughout
3
, that is,
∂R
∂y
=
∂Q
∂z
,
∂P
∂z
=
∂R
∂x
, and
∂Q
∂x
=
∂P
∂y
throughout
3
, where P(x, y, z) = xyz, Q(x, y, z) = xz, and R(x, y, z) = xy. But we see that
∂P
∂z
= xy ,
∂R
∂x
= y ⇒
∂P
∂z

∂R
∂x
for some (x, y, z) in
3
.
Thus, f(x, y, z) does not have a potential in
3
.

¨
Exercises
A
For Exercises 1-3, calculate

C
f (x, y, z) ds for the given function f (x, y, z) and curve C.
176 CHAPTER 4. LINE AND SURFACE INTEGRALS
1. f (x, y, z) = z; C : x = cos t, y = sin t, z = t, 0 ≤ t ≤ 2π
2. f (x, y, z) =
x
y
+ y + 2yz; C : x = t
2
, y = t, z = 1, 1 ≤ t ≤ 2
3. f (x, y, z) = z
2
; C : x = t sin t, y = t cos t, z =
2

2
3
t
3/2
, 0 ≤ t ≤ 1
For Exercises 4-9, calculate

C
f ··· dr for the given vector field f(x, y, z) and curve C.
4. f(x, y, z) = i − j + k; C : x = 3t, y = 2t, z = t, 0 ≤ t ≤ 1
5. f(x, y, z) = y i − x j + z k; C : x = cos t, y = sin t, z = t, 0 ≤ t ≤ 2π
6. f(x, y, z) = x i + y j + z k; C : x = cos t, y = sin t, z = 2, 0 ≤ t ≤ 2π
7. f(x, y, z) = (y − 2z) i + xy j + (2xz + y) k; C : x = t, y = 2t, z = t
2
− 1, 0 ≤ t ≤ 1
8. f(x, y, z) = yz i + xz j + xy k; C : the polygonal path from (0, 0, 0) to (1, 0, 0) to (1, 2, 0)
9. f(x, y, z) = xy i + (z − x) j + 2yz k; C : the polygonal path from (0, 0, 0) to (1, 0, 0) to
(1, 2, 0) to (1, 2, −2)
For Exercises 10-13, state whether or not the vector field f(x, y, z) has a potential in
3
(you do not need to find the potential itself).
10. f(x, y, z) = y i − x j + z k 11. f(x, y, z) = a i + b j + c k (a, b, c constant)
12. f(x, y, z) = (x + y) i + x j + z
2
k 13. f(x, y, z) = xy i − (x − yz
2
) j + y
2
z k
B
For Exercises 14-15, verify Stokes’ Theorem for the given vector field f(x, y, z) and
surface Σ.
14. f(x, y, z) = 2y i − x j + z k; Σ : x
2
+ y
2
+ z
2
= 1, z ≥ 0
15. f(x, y, z) = xy i + xz j + yz k; Σ : z = x
2
+ y
2
, z ≤ 1
16. Construct a Möbius strip from a piece of paper, then draw a line down its center
(like the dotted line in Figure 4.5.3(b)). Cut the Möbius strip along that center line
completely around the strip. How many surfaces does this result in? How would
you describe them? Are they orientable?
17. Use Gnuplot (see Appendix C) to plot the Möbius strip parametrized as:
r(u, v) = cos u (1 + v cos
u
2
) i + sin u (1 + v cos
u
2
) j + v sin
u
2
k , 0 ≤ u ≤ 2π , −
1
2
≤ v ≤
1
2
C
18. Let Σ be a closed surface and f(x, y, z) a smooth vector field. Show that

Σ
(curl f ) ··· ndσ = 0. (Hint: Split Σ in half.)
19. Show that Green’s Theorem is a special case of Stokes’ Theorem.
4.6 Gradient, Divergence, Curl and Laplacian 177
4.6 Gradient, Divergence, Curl and Laplacian
In this final section we will establish some relationships between the gradient, diver-
gence and curl, and we will also introduce a new quantity called the Laplacian. We
will then show how to write these quantities in cylindrical and spherical coordinates.
For a real-valued function f (x, y, z) on
3
, the gradient ∇f (x, y, z) is a vector-valued
function on
3
, that is, its value at a point (x, y, z) is the vector
∇f (x, y, z) =
¸
∂f
∂x
,
∂f
∂y
,
∂f
∂z

=
∂f
∂x
i +
∂f
∂y
j +
∂f
∂z
k
in
3
, where each of the partial derivatives is evaluated at the point (x, y, z). So in this
way, you can think of the symbol ∇ as being “applied” to a real-valued function f to
produce a vector ∇f .
It turns out that the divergence and curl can also be expressed in terms of the
symbol ∇. This is done by thinking of ∇ as a vector in
3
, namely
∇ =

∂x
i +

∂y
j +

∂z
k . (4.51)
Here, the symbols

∂x
,

∂y
and

∂z
are to be thought of as “partial derivative operators”
that will get “applied” to a real-valued function, say f (x, y, z), to produce the partial
derivatives
∂f
∂x
,
∂f
∂y
and
∂f
∂z
. For instance,

∂x
“applied” to f (x, y, z) produces
∂f
∂x
.
Is ∇ really a vector? Strictly speaking, no, since

∂x
,

∂y
and

∂z
are not actual numbers.
But it helps to think of ∇ as a vector, especially with the divergence and curl, as we
will soon see. The process of “applying”

∂x
,

∂y
,

∂z
to a real-valued function f (x, y, z) is
normally thought of as multiplying the quantities:
¸

∂x

( f ) =
∂f
∂x
,
¸

∂y

( f ) =
∂f
∂y
,
¸

∂z

( f ) =
∂f
∂z
For this reason, ∇ is often referred to as the “del operator”, since it “operates” on
functions.
For example, it is often convenient to write the divergence div f as ∇ ··· f, since for
a vector field f(x, y, z) = f
1
(x, y, z)i + f
2
(x, y, z)j + f
3
(x, y, z)k, the dot product of f with ∇
(thought of as a vector) makes sense:
∇··· f =
¸

∂x
i +

∂y
j +

∂z
k

···

f
1
(x, y, z)i + f
2
(x, y, z)j + f
3
(x, y, z)k

=
¸

∂x

( f
1
) +
¸

∂y

( f
2
) +
¸

∂z

( f
3
)
=
∂f
1
∂x
+
∂f
2
∂y
+
∂f
3
∂z
= div f
178 CHAPTER 4. LINE AND SURFACE INTEGRALS
We can also write curl f in terms of ∇, namely as ∇ ××× f, since for a vector field
f(x, y, z) = P(x, y, z)i + Q(x, y, z)j + R(x, y, z)k, we have:
∇××× f =

i j k

∂x

∂y

∂z
P(x, y, z) Q(x, y, z) R(x, y, z)

=
¸
∂R
∂y

∂Q
∂z

i −
¸
∂R
∂x

∂P
∂z

j +
¸
∂Q
∂x

∂P
∂y

k
=
¸
∂R
∂y

∂Q
∂z

i +
¸
∂P
∂z

∂R
∂x

j +
¸
∂Q
∂x

∂P
∂y

k
= curl f
For a real-valued function f (x, y, z), the gradient ∇f (x, y, z) =
∂f
∂x
i +
∂f
∂y
j +
∂f
∂z
k is a
vector field, so we can take its divergence:
div ∇f = ∇··· ∇f
=
¸

∂x
i +

∂y
j +

∂z
k

···
¸
∂f
∂x
i +
∂f
∂y
j +
∂f
∂z
k

=

∂x
¸
∂f
∂x

+

∂y
¸
∂f
∂y

+

∂z
¸
∂f
∂z

=

2
f
∂x
2
+

2
f
∂y
2
+

2
f
∂z
2
Note that this is a real-valued function, to which we will give a special name:
Definition 4.7. For a real-valued function f (x, y, z), the Laplacian of f , denoted by
∆f , is given by
∆f (x, y, z) = ∇··· ∇f =

2
f
∂x
2
+

2
f
∂y
2
+

2
f
∂z
2
. (4.52)
Often the notation ∇
2
f is used for the Laplacian instead of ∆f , using the convention

2
= ∇··· ∇.
Example 4.17. Let r(x, y, z) = x i + y j + z k be the position vector field on
3
. Then
r(x, y, z)
2
= r ··· r = x
2
+ y
2
+ z
2
is a real-valued function. Find
2
(b) the divergence of r
(c) the curl of r
(d) the Laplacian of r
2
4.6 Gradient, Divergence, Curl and Laplacian 179
Solution: (a) ∇r
2
= 2x i + 2y j + 2z k = 2 r
(b) ∇··· r =

∂x
(x) +

∂y
(y) +

∂z
(z) = 1 + 1 + 1 = 3
(c)
∇××× r =

i j k

∂x

∂y

∂z
x y z

= (0 − 0) i − (0 − 0) j + (0 − 0) k = 0
(d) ∆r
2
=

2
∂x
2
(x
2
+ y
2
+ z
2
) +

2
∂y
2
(x
2
+ y
2
+ z
2
) +

2
∂z
2
(x
2
+ y
2
+ z
2
) = 2 + 2 + 2 = 6
Note that we could have calculated ∆r
2
another way, using the ∇ notation along with
parts (a) and (b):
∆r
2
= ∇··· ∇r
2
= ∇··· 2 r = 2 ∇··· r = 2(3) = 6
Notice that in Example 4.17 if we take the curl of the gradient of r
2
we get
∇××× (∇r
2
) = ∇××× 2 r = 2 ∇××× r = 2 0 = 0 .
The following theorem shows that this will be the case in general:
Theorem 4.15. For any smooth real-valued function f (x, y, z), ∇××× (∇f ) = 0.
Proof: We see by the smoothness of f that
∇××× (∇f ) =

i j k

∂x

∂y

∂z
∂f
∂x
∂f
∂y
∂f
∂z

=
¸

2
f
∂y ∂z

2
f
∂z ∂y

i −
¸

2
f
∂x ∂z

2
f
∂z ∂x

j +
¸

2
f
∂x ∂y

2
f
∂y ∂x

k = 0 ,
since the mixed partial derivatives in each component are equal. QED
Corollary 4.16. If a vector field f(x, y, z) has a potential, then curl f = 0.
Another way of stating Theorem 4.15 is that gradients are irrotational. Also, notice
that in Example 4.17 if we take the divergence of the curl of r we trivially get
∇··· (∇××× r) = ∇··· 0 = 0 .
The following theorem shows that this will be the case in general:
Theorem 4.17. For any smooth vector field f(x, y, z), ∇··· (∇××× f) = 0.
The proof is straightforward and left as an exercise for the reader.
180 CHAPTER 4. LINE AND SURFACE INTEGRALS
Corollary 4.18. The flux of the curl of a smooth vector field f(x, y, z) through any
closed surface is zero.
Proof: Let Σ be a closed surface which bounds a solid S . The flux of ∇×××f through Σ is

Σ
(∇××× f ) ··· dσ =

S
∇··· (∇××× f ) dV (by the Divergence Theorem)
=

S
0 dV (by Theorem 4.17)
= 0 . QED
There is another method for proving Theorem 4.15 which can be useful, and is often
used in physics. Namely, if the surface integral

Σ
f (x, y, z) dσ = 0 for all surfaces Σ in
some solid region (usually all of
3
), then we must have f (x, y, z) = 0 throughout that
region. The proof is not trivial, and physicists do not usually bother to prove it. But
the result is true, and can also be applied to double and triple integrals.
For instance, to prove Theorem 4.15, assume that f (x, y, z) is a smooth real-valued
function on
3
. Let C be a simple closed curve in
3
and let Σ be any capping surface
for C (i.e. Σ is orientable and its boundary is C). Since ∇f is a vector field, then

Σ
(∇××× (∇f )) ··· ndσ =

C
∇f ··· dr by Stokes’ Theorem, so
= 0 by Corollary 4.13.
Since the choice of Σ was arbitrary, then we must have (∇×××(∇f ))··· n = 0 throughout
3
,
where n is any unit vector. Using i, j and k in place of n, we see that we must have
∇××× (∇f ) = 0 in
3
, which completes the proof.
Example 4.18. A system of electric charges has a charge density ρ(x, y, z) and produces
an electrostatic field E(x, y, z) at points (x, y, z) in space. Gauss’ Law states that

Σ
E··· dσ = 4π

S
ρ dV
for any closed surface Σ which encloses the charges, with S being the solid region
enclosed by Σ. Show that ∇··· E = 4πρ. This is one of Maxwell’s Equations.
10
10
In Gaussian (or CGS) units.
4.6 Gradient, Divergence, Curl and Laplacian 181
Solution: By the Divergence Theorem, we have

S
∇··· E dV =

Σ
E··· dσ
= 4π

S
ρ dV by Gauss’ Law, so combining the integrals gives

S
(∇··· E− 4πρ) dV = 0 , so
∇··· E− 4πρ = 0 since Σ and hence S was arbitrary, so
∇··· E = 4πρ .
Often (especially in physics) it is convenient to use other coordinate systems when
dealing with quantities such as the gradient, divergence, curl and Laplacian. We will
present the formulas for these in cylindrical and spherical coordinates.
Recall from Section 1.7 that a point (x, y, z) can be represented in cylindrical coordi-
nates (r, θ, z), where x = r cos θ, y = r sin θ, z = z. At each point (r, θ, z), let e
r
, e
θ
, e
z
be unit
vectors in the direction of increasing r, θ, z, respectively (see Figure 4.6.1). Then e
r
, e
θ
,
e
z
form an orthonormal set of vectors. Note, by the right-hand rule, that e
z
××× e
r
= e
θ
.
x
y
z
0
(x, y, z)
(x, y, 0)
θ x
y
z
r
e
r
e
θ
e
z
Figure 4.6.1
Orthonormal vectors e
r
, e
θ
, e
z
in cylindrical coordinates
x
y
z
0
(x, y, z)
(x, y, 0)
θ x
y
z
ρ
φ
e
ρ
e
θ
e
φ
Figure 4.6.2
Orthonormal vectors e
ρ
, e
θ
, e
φ
in spherical coordinates
Similarly, a point (x, y, z) can be represented in spherical coordinates (ρ, θ, φ), where
x = ρ sin φ cos θ, y = ρ sin φ sin θ, z = ρ cos φ. At each point (ρ, θ, φ), let e
ρ
, e
θ
, e
φ
be unit
vectors in the direction of increasing ρ, θ, φ, respectively (see Figure 4.6.2). Then the
vectors e
ρ
, e
θ
, e
φ
are orthonormal. By the right-hand rule, we see that e
θ
××× e
ρ
= e
φ
.
We can nowsummarize the expressions for the gradient, divergence, curl and Lapla-
cian in Cartesian, cylindrical and spherical coordinates in the following tables:
182 CHAPTER 4. LINE AND SURFACE INTEGRALS
Cartesian (x, y, z): Scalar function F; Vector field f = f
1
i + f
2
j + f
3
k
∂F
∂x
i +
∂F
∂y
j +
∂F
∂z
k
divergence : ∇··· f =
∂f
1
∂x
+
∂f
2
∂y
+
∂f
3
∂z
curl : ∇××× f =
¸
∂f
3
∂y

∂f
2
∂z

i +
¸
∂f
1
∂z

∂f
3
∂x

j +
¸
∂f
2
∂x

∂f
1
∂y

k
Laplacian : ∆F =

2
F
∂x
2
+

2
F
∂y
2
+

2
F
∂z
2
Cylindrical (r, θ, z): Scalar function F; Vector field f = f
r
e
r
+ f
θ
e
θ
+ f
z
e
z
∂F
∂r
e
r
+
1
r
∂F
∂θ
e
θ
+
∂F
∂z
e
z
divergence : ∇··· f =
1
r

∂r
(r f
r
) +
1
r
∂f
θ
∂θ
+
∂f
z
∂z
curl : ∇××× f =
¸
1
r
∂f
z
∂θ

∂f
θ
∂z

e
r
+
¸
∂f
r
∂z

∂f
z
∂r

e
θ
+
1
r
¸

∂r
(r f
θ
) −
∂f
r
∂θ

e
z
Laplacian : ∆F =
1
r

∂r
¸
r
∂F
∂r

+
1
r
2

2
F
∂θ
2
+

2
F
∂z
2
Spherical (ρ, θ, φ): Scalar function F; Vector field f = f
ρ
e
ρ
+ f
θ
e
θ
+ f
φ
e
φ
∂F
∂ρ
e
ρ
+
1
ρ sin φ
∂F
∂θ
e
θ
+
1
ρ
∂F
∂φ
e
φ
divergence : ∇··· f =
1
ρ
2

∂ρ

2
f
ρ
) +
1
ρ sin φ
∂f
θ
∂θ
+
1
ρ sin φ

∂φ
(sin φ f
θ
)
curl : ∇××× f =
1
ρ sin φ
¸

∂φ
(sin φ f
θ
) −
∂f
φ
∂θ

e
ρ
+
1
ρ
¸

∂ρ
(ρf
φ
) −
∂f
ρ
∂φ

e
θ
+
¸
1
ρ sin φ
∂f
ρ
∂θ

1
ρ

∂ρ
(ρf
θ
)

e
φ
Laplacian : ∆F =
1
ρ
2

∂ρ
¸
ρ
2
∂F
∂ρ

+
1
ρ
2
sin
2
φ

2
F
∂θ
2
+
1
ρ
2
sin φ

∂φ
¸
sin φ
∂F
∂φ

The derivation of the above formulas for cylindrical and spherical coordinates is
straightforward but extremely tedious. The basic idea is to take the Cartesian equiv-
alent of the quantity in question and to substitute into that formula using the appro-
priate coordinate transformation. As an example, we will derive the formula for the
4.6 Gradient, Divergence, Curl and Laplacian 183
Goal: Show that the gradient of a real-valued function F(ρ, θ, φ) in spherical coordi-
nates is:
∇F =
∂F
∂ρ
e
ρ
+
1
ρ sin φ
∂F
∂θ
e
θ
+
1
ρ
∂F
∂φ
e
φ
Idea: In the Cartesian gradient formula ∇F(x, y, z) =
∂F
∂x
i +
∂F
∂y
j +
∂F
∂z
k, put the Carte-
sian basis vectors i, j, k in terms of the spherical coordinate basis vectors e
ρ
, e
θ
, e
φ
and functions of ρ, θ and φ. Then put the partial derivatives
∂F
∂x
,
∂F
∂y
,
∂F
∂z
in terms of
∂F
∂ρ
,
∂F
∂θ
,
∂F
∂φ
and functions of ρ, θ and φ.
Step 1: Get formulas for e
ρ
, e
θ
, e
φ
in terms of i, j, k.
We can see from Figure 4.6.2 that the unit vector e
ρ
in the ρ direction at a general
point (ρ, θ, φ) is e
ρ
=
r
r
, where r = x i + y j + z k is the position vector of the point in
Cartesian coordinates. Thus,
e
ρ
=
r
r
=
x i + y j + z k

x
2
+ y
2
+ z
2
,
so using x = ρ sin φ cos θ, y = ρ sin φ sin θ, z = ρ cos φ, and ρ =

x
2
+ y
2
+ z
2
, we get:
e
ρ
= sin φ cos θ i + sin φ sin θ j + cos φ k
Now, since the angle θ is measured in the xy-plane, then the unit vector e
θ
in the θ
direction must be parallel to the xy-plane. That is, e
θ
is of the form a i + b j + 0 k. To
figure out what a and b are, note that since e
θ
⊥ e
ρ
, then in particular e
θ
⊥ e
ρ
when
e
ρ
is in the xy-plane. That occurs when the angle φ is π/2. Putting φ = π/2 into the
formula for e
ρ
gives e
ρ
= cos θ i + sin θ j + 0 k, and we see that a vector perpendicular
to that is −sin θ i + cos θ j + 0 k. Since this vector is also a unit vector and points in the
(positive) θ direction, it must be e
θ
:
e
θ
= −sin θ i + cos θ j + 0 k
Lastly, since e
φ
= e
θ
××× e
ρ
, we get:
e
φ
= cos φ cos θ i + cos φ sin θ j − sin φ k
Step 2: Use the three formulas from Step 1 to solve for i, j, k in terms of e
ρ
, e
θ
, e
φ
.
This comes down to solving a system of three equations in three unknowns. There
are many ways of doing this, but we will do it by combining the formulas for e
ρ
and
e
φ
to eliminate k, which will give us an equation involving just i and j. This, with the
formula for e
θ
, will then leave us with a system of two equations in two unknowns (i
and j), which we will use to solve first for j then for i. Lastly, we will solve for k.
First, note that
sin φ e
ρ
+ cos φ e
φ
= cos θ i + sin θ j
184 CHAPTER 4. LINE AND SURFACE INTEGRALS
so that
sin θ (sin φ e
ρ
+ cos φ e
φ
) + cos θ e
θ
= (sin
2
θ + cos
2
θ)j = j ,
and so:
j = sin φ sin θ e
ρ
+ cos θ e
θ
+ cos φ sin θ e
φ
Likewise, we see that
cos θ (sin φ e
ρ
+ cos φ e
φ
) − sin θ e
θ
= (cos
2
θ + sin
2
θ)i = i ,
and so:
i = sin φ cos θ e
ρ
− sin θ e
θ
+ cos φ cos θ e
φ
Lastly, we see that:
k = cos φ e
ρ
− sin φ e
φ
Step 3: Get formulas for
∂F
∂ρ
,
∂F
∂θ
,
∂F
∂φ
in terms of
∂F
∂x
,
∂F
∂y
,
∂F
∂z
.
By the Chain Rule, we have
∂F
∂ρ
=
∂F
∂x
∂x
∂ρ
+
∂F
∂y
∂y
∂ρ
+
∂F
∂z
∂z
∂ρ
,
∂F
∂θ
=
∂F
∂x
∂x
∂θ
+
∂F
∂y
∂y
∂θ
+
∂F
∂z
∂z
∂θ
,
∂F
∂φ
=
∂F
∂x
∂x
∂φ
+
∂F
∂y
∂y
∂φ
+
∂F
∂z
∂z
∂φ
,
which yields:
∂F
∂ρ
= sin φ cos θ
∂F
∂x
+ sin φ sin θ
∂F
∂y
+ cos φ
∂F
∂z
∂F
∂θ
= −ρ sin φ sin θ
∂F
∂x
+ ρ sin φ cos θ
∂F
∂y
∂F
∂φ
= ρ cos φ cos θ
∂F
∂x
+ ρ cos φ sin θ
∂F
∂y
− ρ sin φ
∂F
∂z
Step 4: Use the three formulas from Step 3 to solve for
∂F
∂x
,
∂F
∂y
,
∂F
∂z
in terms of
∂F
∂ρ
,
∂F
∂θ
,
∂F
∂φ
.
Again, this involves solving a system of three equations in three unknowns. Using
a similar process of elimination as in Step 2, we get:
∂F
∂x
=
1
ρ sin φ
¸
ρ sin
2
φ cos θ
∂F
∂ρ
− sin θ
∂F
∂θ
+ sin φ cos φ cos θ
∂F
∂φ

∂F
∂y
=
1
ρ sin φ
¸
ρ sin
2
φ sin θ
∂F
∂ρ
+ cos θ
∂F
∂θ
+ sin φ cos φ sin θ
∂F
∂φ

∂F
∂z
=
1
ρ
¸
ρ cos φ
∂F
∂ρ
− sin φ
∂F
∂φ

4.6 Gradient, Divergence, Curl and Laplacian 185
Step 5: Substitute the formulas for i, j, k from Step 2 and the formulas for
∂F
∂x
,
∂F
∂y
,
∂F
∂z
from Step 4 into the Cartesian gradient formula ∇F(x, y, z) =
∂F
∂x
i +
∂F
∂y
j +
∂F
∂z
k.
Doing this last step is perhaps the most tedious, since it involves simplifying 3 ×3 +
3 × 3 + 2 × 2 = 22 terms! Namely,
∇F =
1
ρ sin φ
¸
ρ sin
2
φ cos θ
∂F
∂ρ
− sin θ
∂F
∂θ
+ sin φ cos φ cos θ
∂F
∂φ

(sin φ cos θ e
ρ
− sin θ e
θ
+ cos φ cos θ e
φ
)
+
1
ρ sin φ
¸
ρ sin
2
φ sin θ
∂F
∂ρ
+ cos θ
∂F
∂θ
+ sin φ cos φ sin θ
∂F
∂φ

(sin φ sin θ e
ρ
+ cos θ e
θ
+ cos φ sin θ e
φ
)
+
1
ρ
¸
ρ cos φ
∂F
∂ρ
− sin φ
∂F
∂φ

(cos φ e
ρ
− sin φ e
φ
) ,
which we see has 8 terms involving e
ρ
, 6 terms involving e
θ
, and 8 terms involving e
φ
.
But the algebra is straightforward and yields the desired result:
∇F =
∂F
∂ρ
e
ρ
+
1
ρ sin φ
∂F
∂θ
e
θ
+
1
ρ
∂F
∂φ
e
φ

Example 4.19. In Example 4.17 we showed that ∇r
2
= 2 r and ∆r
2
= 6, where
r(x, y, z) = x i + y j + z k in Cartesian coordinates. Verify that we get the same answers
if we switch to spherical coordinates.
Solution: Since r
2
= x
2
+ y
2
+ z
2
= ρ
2
in spherical coordinates, let F(ρ, θ, φ) = ρ
2
(so
that F(ρ, θ, φ) = r
2
). The gradient of F in spherical coordinates is
∇F =
∂F
∂ρ
e
ρ
+
1
ρ sin φ
∂F
∂θ
e
θ
+
1
ρ
∂F
∂φ
e
φ
= 2ρ e
ρ
+
1
ρ sin φ
(0) e
θ
+
1
ρ
(0) e
φ
= 2ρ e
ρ
= 2ρ
r
r
, as we showed earlier, so
= 2ρ
r
ρ
= 2 r , as expected. And the Laplacian is
∆F =
1
ρ
2

∂ρ
¸
ρ
2
∂F
∂ρ

+
1
ρ
2
sin
2
φ

2
F
∂θ
2
+
1
ρ
2
sin φ

∂φ
¸
sin φ
∂F
∂φ

=
1
ρ
2

∂ρ

2
2ρ) +
1
ρ
2
sin φ
(0) +
1
ρ
2
sin φ

∂φ
(sin φ (0))
=
1
ρ
2

∂ρ
(2ρ
3
) + 0 + 0
=
1
ρ
2
(6ρ
2
) = 6 , as expected.
186 CHAPTER 4. LINE AND SURFACE INTEGRALS

¨
Exercises
A
For Exercises 1-6, find the Laplacian of the function f (x, y, z) in Cartesian coordinates.
1. f (x, y, z) = x + y + z 2. f (x, y, z) = x
5
3. f (x, y, z) = (x
2
+ y
2
+ z
2
)
3/2
4. f (x, y, z) = e
x+y+z
5. f (x, y, z) = x
3
+ y
3
+ z
3
6. f (x, y, z) = e
−x
2
−y
2
−z
2
7. Find the Laplacian of the function in Exercise 3 in spherical coordinates.
8. Find the Laplacian of the function in Exercise 6 in spherical coordinates.
9. Let f (x, y, z) =
z
x
2
+ y
2
in Cartesian coordinates. Find ∇f in cylindrical coordinates.
10. For f(r, θ, z) = r e
r
+ z sin θ e
θ
+ rz e
z
in cylindrical coordinates, find div f and curl f.
11. For f(ρ, θ, φ) = e
ρ
+ ρ cos θ e
θ
+ ρ e
φ
in spherical coordinates, find div f and curl f.
B
For Exercises 12-23, prove the given formula (r = r is the length of the position
vector field r(x, y, z) = x i + y j + z k).
12. ∇(1/r) = −r/r
3
13. ∆(1/r) = 0 14. ∇··· (r/r
3
) = 0 15. ∇(ln r) = r/r
2
16. div (F + G) = div F + div G 17. curl (F + G) = curl F + curl G
18. div ( f F) = f div F + F··· ∇f 19. div (F××× G) = G··· curl F − F··· curl G
20. div (∇f ××× ∇g) = 0 21. curl ( f F) = f curl F + (∇f ) ××× F
22. curl (curl F) = ∇(div F) − ∆F 23. ∆( f g) = f ∆g + g ∆ f + 2(∇f ··· ∇g)
C
24. Prove Theorem 4.17.
25. Derive the gradient formula in cylindrical coordinates: ∇F =
∂F
∂r
e
r
+
1
r
∂F
∂θ
e
θ
+
∂F
∂z
e
z
26. Use f = u ∇v in the Divergence Theorem to prove:
(a) Green’s first identity:

S
(u ∆v + (∇u) ··· (∇v)) dV =

Σ
(u ∇v) ··· dσ
(b) Green’s second identity:

S
(u ∆v − v ∆u) dV =

Σ
(u ∇v − v ∇u) ··· dσ
27. Suppose that ∆u = 0 (i.e. u is harmonic) over
3
. Define the normal derivative
∂u
∂n
of u over a closed surface Σ with outward unit normal vector n by
∂u
∂n
= D
n
u = n··· ∇u.
Show that

Σ
∂u
∂n
dσ = 0. (Hint: Use Green’s second identity.)
Bibliography
Abbott, E.A., Flatland, 7th edition. New York: Dover Publications, Inc., 1952
Classic tale about a creature living in a 2-dimensional world who encounters a higher-
dimensional creature, with lots of humor thrown in.
Anton, H. and C. Rorres, Elementary Linear Algebra: Applications Version, 8th edi-
tion. New York: John Wiley & Sons, 2000
Standard treatment of elementary linear algebra.
Bazaraa, M.S., H.D. Sherali and C.M. Shetty, Nonlinear Programming: Theory and
Algorithms, 2nd edition. New York: John Wiley & Sons, 1993
Thorough treatment of nonlinear optimization.
Farin, G., Curves and Surfaces for Computer Aided Geometric Design: A Practical
Guide, 2nd edition. San Diego, CA: Academic Press, 1990
An intermediate-level book on curve and surface design.
An intermediate-level book on optics, covering a wide range of topics.
Hoel, P.G., S.C. Port and C.J. Stone, Introduction to Probability Theory, Boston, MA:
Houghton Mifﬂin Co., 1971
An excellent introduction to elementary, calculus-based probability theory. Lots of good
exercises.
Jackson, J.D., Classical Electrodynamics, 2nd edition. New York: John Wiley & Sons,
1975
An advanced book on electromagnetism, famous for being intimidating. Most of the
mathematics will be understandable after reading the present book.
Marion, J.B., Classical Dynamics of Particles and Systems, 2nd edition. New York:
Standard intermediate-level treatment of classical mechanics. Very thorough.
O’Neill, B., Elementary Differential Geometry, New York: Academic Press, 1966
Intermediate-level book on differential geometry, with a modern approach based on dif-
ferential forms.
187
188 Bibliography
Pogorelov, A.V., Analytical Geometry, Moscow: Mir Publishers, 1980
An intermediate/advanced book on analytic geometry.
Press, W.H., S.A. Teukolsky, W.T. Vetterling and B.P. Flannery, Numerical Recipes
in FORTRAN: The Art of Scientific Computing, 2nd edition. Cambridge, UK:
Cambridge University Press, 1992
An excellent source of information on numerical methods for solving a wide variety of
problems. Though all the examples are in the FORTRAN programming language, the code
is clear enough to implement in the language of your choice.
Protter, M.H. and C.B. Morrey, Analytic Geometry, 2nd edition. Reading, MA:
Thorough treatment of elementary analytic geometry, with a rigor not found in most
recent books.
Ralston, A. and P. Rabinowitz, A First Course in Numerical Analysis, 2nd edition.
New York: McGraw-Hill, 1978
Standard treatment of elementary numerical analysis.
Reitz, J.R., F.J. Milford and R.W. Christy, Foundations of Electromagnetic Theory,
Intermediate text on electromagnetism.
Schey, H.M., Div, Grad, Curl, and All That: An Informal Text on Vector Calculus, New
York: W.W. Norton & Co., 1973
Very intuitive approach to the subject, from a physicist’s viewpoint. Highly recom-
mended.
Taylor, A.E. and W.R. Mann, Advanced Calculus, 2nd edition. New York: John Wiley
& Sons, 1972
Excellent treatment of n-dimensional calculus. A good book to study after the present
book. Many intriguing exercises.
Uspensky, J.V., Theory of Equations, New York: McGraw-Hill, 1948
A classic on the subject, discussing many interesting topics.
Weinberger, H.F., A First Course in Partial Differential Equations, New York: John
Wiley & Sons, 1965
A good introduction to the vast subject of partial differential equations.
Welchons, A.M. and W.R. Krickenberger, Solid Geometry, Boston, MA: Ginn & Co.,
1936
A very thorough treatment of 3-dimensional geometry from an elementary perspective,
includes many topics which (sadly) do not seem to be taught anymore.
Appendix A
Answers and Hints to Selected Exercises
Chapter 1
Section 1.1 (p. 8)
1. (a)

5 (b)

5 (c)

17 (d) 1
(e) 2

17 2. Yes 3. No
Section 1.2 (p. 14)
1. (a) (−4, 4, −3) (b) (2, 6, −1)
(c)

−1

30
,
5

30
,
−2

30

(d)

41
2
(e)

41
2
(f) (14, −6, 8) (g) (−7, 3, −4)
(h) (−1, −6, 1) (i) (−2, −4, 2) (j) No.
3. No. v + w is larger.
Section 1.3 (p. 18)
1. 10 3. 73.4

5. 90

7. 0

9. Yes, since v··· w = 0.
11. |v··· w| = 0 <

21

5 = v w
13. v+w =

26 <

21 +

5 = v +w
15. Hint: use Definition 1.6.
24. Hint: See Theorem 1.10(c).
Section 1.4 (p. 29)
1. (−5, −23, −24) 3. (8, 4, −5) 5. 0
7. 16.72 9. 4

5 11. 9 13. 0
and (8, −10, 2) 15. 14
Section 1.5 (p. 39)
1. (a) (2, 3, 2) + t(5, 4, −3) (b) x = 2 + 5t,
y = 3 + 4t, z = 2 − 3t (c)
x−2
5
=
y−3
4
=
z−2
−3
3. (a) (2, 1, 3) + t(1, 0, 1) (b) x = 2 + t,
y = 1, z = 3 + t (c) x − 2 = z − 3, y = 1
5. x = 1 + 2t, y = −2 + 7t, z = −3 + 8t
7. 7.65 9. (1, 2, 3)
11. 4x − 4y + 3z − 10 = 0
13. x − 2y − z + 2 = 0
15. 11x − 24y + 21z − 26 = 0 17. 9/

35
19. x = 5t, y = 2+3t, z = −7t 21. (10, −2, 1)
Section 1.6 (p. 46)
center: (−1, −1, −1) 5. No intersection.
7. circle x
2
+ y
2
= 4 in the planes z = ±

5
9. lines
x
a
=
y
b
, z = 0 and
x
a
= −
y
b
, z = 0
13.

2a
2−c
,
2b
2−c
, 0

Section 1.7 (p. 50)
1. (a) (4,
π
3
, −1) (b) (

17,
π
3
, 1.816)
3. (a) (2

7,
11π
6
, 0) (b) (2

7,
11π
6
,
π
2
)
5. (a) r
2
+ z
2
= 25 (b) ρ = 5
7. (a) r
2
+9z
2
= 36 (b) ρ
2
(1 +8 cos
2
φ) = 36
10. (a, θ, a cot φ) 12. Hint: Use the dis-
tance formula for Cartesian coordinates.
Section 1.8 (p. 57)
1. f

(t) = (1, 2t, 3t
2
); x = 1 + t, y = z = 1
3. f

(t) = (−2 sin 2t, 2 cos 2t, 1); x = 1,
y = 2t, z = t 5. v(t) = (1, 1 − cos t, sin t),
a(t) = (0, sin t, cos t)
9. (a) Line parallel to c (b) Half-line
189
190 Appendix A: Answers and Hints to Selected Exercises
parallel to c (c) Hint: Think of the
functions as position vectors.
15. Hint: Theorem 1.16
Section 1.9 (p. 63)
1.

5
2
3.
2
27
(13
3/2
− 8) 5. Replace
t by

27s+16
2

2/3
− 4

9 6. Hint: Use
Theorem 1.20(e), Example 1.37, and
Theorem 1.16 7. Hint: Use Exercise 6.
9. Hint: Use f

(t) = f(t)T, differ-
entiate that to get f
′′
(t), put those ex-
pressions into f

(t) ××× f
′′
(t), then write
T

(t) in terms of N(t). 11. T(t) =
1

2
(−sin t, cos t, 1), N(t) = (−cos t, −sin t, 0),
B(t) =
1

2
(sin t, −cos t, 1), κ(t) = 1/2
Chapter 2
Section 2.1 (p. 70)
1. domain:
2
, range: [−1, ∞) 3. domain:
{(x, y) : x
2
+ y
2
≥ 4}, range: [0, ∞)
5. domain:
3
, range: [−1, 1] 7. 1
9. does not exist 11. 2 13. 2 15. 0
17. does not exist
Section 2.2 (p. 74)
1.
∂f
∂x
= 2x,
∂f
∂y
= 2y 3.
∂f
∂x
= x(x
2
+y +4)
−1/2
,
∂f
∂y
=
1
2
(x
2
+ y + 4)
−1/2
5.
∂f
∂x
= ye
xy
+ y,
∂f
∂y
= xe
xy
+ x 7.
∂f
∂x
= 4x
3
,
∂f
∂y
= 0
9.
∂f
∂x
= x(x
2
+ y
2
)
−1/2
,
∂f
∂y
= y(x
2
+ y
2
)
−1/2
11.
∂f
∂x
=
2x
3
(x
2
+ y + 4)
−2/3
,
∂f
∂y
=
1
3
(x
2
+ y + 4)
−2/3
13.
∂f
∂x
= −2xe
−(x
2
+y
2
)
,
∂f
∂y
= −2ye
−(x
2
+y
2
)
15.
∂f
∂x
= y cos(xy),
∂f
∂y
= x cos(xy) 17.

2
f
∂x
2
= 2,

2
f
∂y
2
= 2,

2
f
∂x ∂y
= 0 19.

2
f
∂x
2
= (y + 4)(x
2
+ y + 4)
−3/2
,

2
f
∂y
2
= −
1
4
(x
2
+ y + 4)
−3/2
,

2
f
∂x ∂y
= −
1
2
x(x
2
+ y + 4)
−3/2
21.

2
f
∂x
2
= y
2
e
xy
,

2
f
∂y
2
= x
2
e
xy
,

2
f
∂x ∂y
= (1 + xy)e
xy
+ 1 23.

2
f
∂x
2
= 12x
2
,

2
f
∂y
2
= 0,

2
f
∂x ∂y
= 0 25.

2
f
∂x
2
= −x
−2
,

2
f
∂y
2
= −y
−2
,

2
f
∂x ∂y
= 0
Section 2.3 (p. 77)
1. 2x + 3y − z − 3 = 0 3. −2x + y − z − 2 = 0
5. x + 2y = z 7.
1
2
(x − 1) +
4
9
(y − 2) +

11
12
(z −
2

11
3
) = 0 9. 3x + 4y − 5z = 0
Section 2.4 (p. 82)
1. (2x, 2y) 3. (
x

x
2
+y
2
+4
,
y

x
2
+y
2
+4
)
5. (1/x, 1/y) 7. (yz cos(xyz), xz cos(xyz), xy cos(xyz))
9. (2x, 2y, 2z) 11. 2

2 13.
1

3
15.

3 cos(1) 17. increase: (45, 20),
decrease: (−45, −20)
Section 2.5 (p. 88)
1. local min. (1, 0); saddle pt. (−1, 0)
3. local min. (1, 1); local max. (−1, −1);
saddle pts. (1, −1), (−1, 1) 5. local min.
(1, −1); saddle pt. (0, 0) 7. local min. (0, 0)
9. local min. (−1, 1/2) 11. width = height
= depth=10 13. x = y = 4, z = 2
Section 2.6 (p. 95)
2. (x
0
, y
0
) = (0, 0) : → (0.2858, −0.3998);
(x
0
, y
0
) = (1, 1) : → (1.03256, −1.94037)
191
Section 2.7 (p. 100)
1. min.

−4

5
,
−2

5

; max.

4

5
,
2

5

3. min.

20

13
,
30

13

; max.

20

13
, −
30

13

4. min.

−9

5
, 0,
2

5

; max.

9
8
,

59
4
,
−1
4

5.
8abc
3

3
Chapter 3
Section 3.1 (p. 104)
1. 1 3.
7
12
5.
7
6
7. 5 9.
1
2
11. 15
Section 3.2 (p. 109)
1. 1 3. 8 ln 2 − 3 5.
π
4
6.
1
4
7. 2 9.
1
6
10.
6
5
Section 3.3 (p. 112)
1.
9
2
3. (2 cos(π
2
) + π
4
− 2)/4 5.
1
6
7. 6
10.
1
3
Section 3.4 (p. 116)
1. The values should converge to ≈ 1.318.
(Hint: In Java the exponential function
e
x
can be obtained with Math.exp(x).
Other languages have similar functions,
otherwise use e = 2.7182818284590455 in
2. ≈ 1.146 3. ≈ 0.705 4. ≈ 0.168
Section 3.5 (p. 123)
1. 8π 3.

3
(8 − 3
3/2
) 7. 1 −
sin 2
2
9. 2πab
Section 3.6 (p. 127)
1. (1, 8/3) 3. (0,
4a

) 5. (0, 3π/16)
7. (0, 0, 5a/12) 9. (7/12, 7/12, 7/12)
Section 3.7 (p. 134)
1.

π 2. 1 6. Both are
n
(n+1)
2
(n+2)
7.
1
n
Chapter 4
Section 4.1 (p. 142)
1. 1/2 3. 23 5. 24π 7. −2π 9. 2π
11. 4π
Section 4.2 (p. 149)
1. 0 3. No 4. Yes. F(x, y) =
x
2
2

y
2
2
5. No 9. (b) No. Hint: Think of how F is
defined. 10. Yes. F(x, y) = axy +bx +cy +d
Section 4.3 (p. 155)
1. 16/15 3. −5π 5. Yes. F(x, y) = xy
2
+ x
3
7. Yes. F(x, y) = 4x
2
y + 2y
2
+ 3x
Section 4.4 (p. 163)
1. 216π 2. 3 3. 12π/5 7. 15/4
Section 4.5 (p. 175)
1. 2

2 π
2
2. (17

17 − 5

5)/3 3. 2/5
4. 1 5. 2π(π − 1) 7. 67/15 9. 6
11. Yes 13. No 19. Hint: Think of
how a vector field f(x, y) = P(x, y) i+Q(x, y) j
in
2
can be extended in a natural way to
be a vector field in
3
.
Section 4.6 (p. 186)
1. 0 3. 12

x
2
+ y
2
+ z
2
5. 6(x + y + z)
7. 12ρ 8. (4ρ
2
− 6)e
−ρ
2
9. −
2z
r
3
e
r
+
1
r
2
e
z
11. div f =
2
ρ

sin θ
sin φ
+ cot φ;
curl f = cot φ cos θ e
ρ
+ 2e
θ
− 2 cos θ e
φ
25. Hint: Start by showing that e
r
=
cos θ i + sin θ j, e
θ
= −sin θ i + cos θ j, e
z
= k.
Appendix B
We will prove the right-hand rule for the cross product of two vectors in
3
.
For any vectors v and w in
3
, define a new vector, n(v, w), as follows:
1. If v and w are nonzero and not parallel, and θ is the angle between them, then
n(v, w) is the vector in
3
such that:
(a) the magnitude of n(v, w) is v w sin θ,
(b) n(v, w) is perpendicular to the plane containing v and w, and
(c) v, w, n(v, w) form a right-handed system.
2. If v and w are nonzero and parallel, then n(v, w) = 0.
3. If either v or w is 0, then n(v, w) = 0.
The goal is to show that n(v, w) = v ××× w for all v, w in
3
, which would prove the
right-hand rule for the cross product (by part 1(c) of our definition). To do this, we will
perform the following steps:
Step 1: Show that n(v, w) = v××× w if v and w are any two of the basis vectors i, j, k.
This was already shown in Example 1.11 in Section 1.4.
Step 2: Show that n(av, bw) = ab(v ××× w) for any scalars a, b if v and w are any two of
the basis vectors i, j, k.
If either a = 0 or b = 0 then n(av, bw) = 0 = ab(v ××× w), so the result holds. So assume
that a 0 and b 0. Let v and w be any two of the basis vectors i, j, k. For example,
we will show that the result holds for v = i and w = k (the other possibilities follow in
a similar fashion).
For av = ai and bw = bk, the angle θ between av and bwis 90

. Hence the magnitude
of n(av, bw), by definition, is ai bk sin 90

= |ab|. Also, by definition, n(av, bw) is
perpendicular to the plane containing ai and bk, namely, the xz-plane. Thus, n(av, bw)
must be a scalar multiple of j. Since its magnitude is |ab|, then n(av, bw) must be
either |ab|j or −|ab|j.
There are four possibilities for the combinations of signs for a and b. We will con-
sider the case when a > 0 and b > 0 (the other three possibilities are handled simi-
larly).
192
193
In this case, n(av, bw) must be either abj or −abj. Now, since i, j, k form a right-
handed system, then i, k, j form a left-handed system, and so i, k, −j form a right-
handed system. Thus, ai, bk, −abj form a right-handed system (since a > 0, b > 0, and
ab > 0). So since, by definition, ai, bk, n(ai, bk) form a right-handed system, and since
n(ai, bk) has to be either abj or −abj, this means that we must have n(ai, bk) = −abj.
But we know that ai ××× bk = ab(i ××× k) = ab(−j) = −abj. Therefore, n(ai, bk) = ab(i ××× k),
which is what we needed to show.
∴ n(av, bw) = ab(v××× w)
Step 3: Show that n(u, v + w) = n(u, v) + n(u, w) for any vectors u, v, w.
If u = 0 then the result holds trivially since n(u, v +w), n(u, v) and n(u, w) are all the
zero vector. If v = 0, then the result follows easily since n(u, v + w) = n(u, 0 + w) =
n(u, w) = 0+n(u, w) = n(u, 0) = n(u, w) = n(u, v) +n(u, w). A similar argument shows
that the result holds if w = 0.
So now assume that u, v and ware all nonzero vectors. We will describe a geometric
construction of n(u, v), which is shown in the figure below. Let P be a plane perpen-
dicular to u. Multiply the vector v by the positive scalar u, then project the vector
u v straight down onto the plane P. You can think of this projection vector (denoted
by pro j
P
u v) as the shadow of the vector u v on the plane P, with the light source
directly overhead the terminal point of u v. If θ is the angle between u and v, then
we see that pro j
P
u v has magnitude u v sin θ, which is the magnitude of n(u, v).
So rotating pro j
P
u v by 90

in a counter-clockwise direction in the plane P gives a
vector whose magnitude is the same as that of n(u, v) and which is perpendicular to
pro j
P
u v (and hence perpendicular to v). Since this vector is in P then it is also per-
pendicular to u. And we can see that u, v and this vector form a right-handed system.
Hence this vector must be n(u, v). Note that this holds even if u v, since in that case
θ = 0

and so sin θ = 0 which means that n(u, v) has magnitude 0, which is what we
would expect.
u
v
pro j
P
u v
u v
n(u, v)
θ
θ
P
Now apply this same geometric construction to get n(u, w) and n(u, v + w). Since
u (v + w) is the sum of the vectors u v and u w, then the projection vector
194 Appendix B: Proof of the Right-Hand Rule for the Cross Product
pro j
P
u (v + w) is the sum of the projection vectors pro j
P
u v and pro j
P
u w (to
see this, using the shadow analogy again and the parallelogram rule for vector addi-
tion, think of how projecting a parallelogram onto a plane gives you a parallelogram in
that plane). So then rotating all three projection vectors by 90

in a counter-clockwise
direction in the plane P preserves that sum (see the figure below), which means that
n(u, v + w) = n(u, v) + n(u, w).
u
v
w
v + w
u (v + w)
pro j
P
u v
pro j
P
u w
pro j
P
u (v + w)
u v
u w
n(u, v) n(u, w)
n(u, v + w)
θ
θ
P
Step 4: Show that n(w, v) = −n(v, w) for any vectors v, w.
If v and w are nonzero and parallel, or if either is 0, then n(w, v) = 0 = −n(v, w), so
the result holds. So assume that v and w are nonzero and not parallel. Then n(w, v)
has magnitude w v sin θ, which is the same as the magnitude of n(v, w), and hence
is the same as the magnitude of −n(v, w). By definition, n(v, w) is perpendicular to
the plane containing w and v, and hence so is −n(v, w). Also, v, w, n(v, w) form a
right-handed system, and so w, v, n(v, w) form a left-handed system, and hence w,
v, −n(v, w) form a right-handed system. Thus, we have shown that −n(v, w) is a vec-
tor with the same magnitude as n(w, v) and is perpendicular to the plane containing
w and v, and that w, v, −n(v, w) form a right-handed system. So by definition this
means that −n(v, w) must be n(w, v).
Step 5: Show that n(v, w) = v××× w for all vectors v, w.
Write v = v
1
i + v
2
j + v
3
k and w = w
1
i + w
2
j + w
3
k. Then by Steps 3 and 4, we have
195
n(v, w) = n(v
1
i + v
2
j + v
3
k, w
1
i + w
2
j + w
3
k)
= n(v
1
i + v
2
j + v
3
k, w
1
i) + n(v
1
i + v
2
j + v
3
k, w
2
j + w
3
k)
= n(v
1
i + v
2
j + v
3
k, w
1
i) + n(v
1
i + v
2
j + v
3
k, w
2
j) + n(v
1
i + v
2
j + v
3
k, w
3
k)
= −n(w
1
i, v
1
i + v
2
j + v
3
k) + −n(w
2
j, v
1
i + v
2
j + v
3
k) + −n(w
3
k, v
1
i + v
2
j + v
3
k).
We can use Steps 1 and 2 to evaluate the three terms on the right side of the last
equation above:
−n(w
1
i, v
1
i + v
2
j + v
3
k) = −n(w
1
i, v
1
i) + −n(w
1
i, v
2
j) + −n(w
1
i, v
3
k)
= −v
1
w
1
n(i, i) + −v
2
w
1
n(i, j) + −v
3
w
1
n(i, k)
= −v
1
w
1
(i ××× i) + −v
2
w
1
(i ××× j) + −v
3
w
1
(i ××× k)
= −v
1
w
1
0 + −v
2
w
1
k + −v
3
w
1
(−j)
−n(w
1
i, v
1
i + v
2
j + v
3
k) = −v
2
w
1
k + v
3
w
1
j
Similarly, we can calculate
−n(w
2
j, v
1
i + v
2
j + v
3
k) = v
1
w
2
k − v
3
w
2
i
and
−n(w
3
j, v
1
i + v
2
j + v
3
k) = −v
1
w
3
j + v
2
w
3
i .
Thus, putting it all together, we have
n(v, w) = −v
2
w
1
k + v
3
w
1
j + v
1
w
2
k − v
3
w
2
i − v
1
w
3
j + v
2
w
3
i
= (v
2
w
3
− v
3
w
2
)i + (v
3
w
1
− v
1
w
3
)j + (v
1
w
2
− v
2
w
1
)k
= v××× w by definition of the cross product.
∴ n(v, w) = v××× w for all vectors v, w.
So since v, w, n(v, w) form a right-handed system, then v, w, v ××× w form a right-
handed system, which completes the proof.
Appendix C
3D Graphing with Gnuplot
Gnuplot is a free, open-source software package for producing a variety of graphs.
Versions are available for many operating systems. Below is a very brief tutorial on
how to use Gnuplot to graph functions of several variables.
INSTALLATION
Zip file with a name such as gp420win32.zip, which is version 4.2.0. All the
examples we will discuss require at least version 4.2.0.
2. Install the downloaded file. For example, in Windows you would unzip the Zip file
you downloaded in Step 1 into some folder (use the “Use folder names” option if
extracting with WinZip).
RUNNING GNUPLOT
1. In Windows, run wgnuplot.exe fromthe folder (or bin folder) where you installed
Gnuplot. In Linux, just type gnuplot in a terminal window.
2. You should now get a Gnuplot terminal with a gnuplot> command prompt. In
Windows this will appear in a new window, while in Linux it will appear in the
terminal window where the gnuplot command was run. For Windows, if the font
is unreadable you can change it by right-clicking on the text part of the Gnuplot
window and selecting the “Choose Font..” option. For example, the font “Courier”,
style “Regular”, size “12” is usually a good choice (that choice can be saved for
future sessions by right-clicking in the Gnuplot window again and selecting the
option to update wgnuplot.ini).
3. At the gnuplot> command prompt you can now run graphing commands, which
we will now describe.
GRAPHING FUNCTIONS
The usual way to create 3D graphs in Gnuplot is with the splot command:
splot <range> <comma-separated list of functions>
196
197
For a function z = f (x, y), <range> is the range of x and y values (and optionally the
range of z values) over which to plot. To specify an x range and a y range, use an
expression of the form [a : b][c : d], for some numbers a < b and c < d. This will cause
the graph to be plotted for a ≤ x ≤ b and c ≤ y ≤ d.
Function definitions use the x and y variables in combination with mathematical op-
erators, listed below:
Symbol Operation Example Result
+ Addition 2 + 3 5
− Subtraction 3 − 2 1
* Multiplication 2*3 6
/ Division 4/2 2
** Power 2**3 2
3
= 8
exp(x) e
x
exp(2) e
2
log(x) ln x log(2) ln 2
sin(x) sin x sin(pi/2) 1
cos(x) cos x cos(pi) −1
tan(x) tan x tan(pi/4) 1
Example C.1. To graph the function z = 2x
2
+ y
2
from x = −1 to x = 1 and from y = −2
to y = 2, type this at the gnuplot> prompt:
splot [−1 : 1][−2 : 2] 2*x**2 + y**2
The result is shown below:
-1
-0.5
0
0.5
1
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0
1
2
3
4
5
6
7
2*(x**2) + y**2
198 Appendix C: 3D Graphing with Gnuplot
Note that we had to type 2*x**2 to multiply 2 times x
2
. For clarity, parentheses can
be used to make sure the operations are being performed in the correct order:
splot [−1 : 1][−2 : 2] 2*(x**2) + y**2
In the above example, to also plot the function z = e
x+y
on the same graph, put a
comma after the first function then append the new function:
splot [−1 : 1][−2 : 2] 2*(x**2) + y**2, exp(x+y)
By default, the x-axis and y-axis are not shown in the graph. To display the axes, use
this command before the splot command:
set zeroaxis
Also, by default the x- and y-axes are switched from their usual position. To show the
axes with the orientation which we have used throughout the text, use this command:
set view 60, 120, 1, 1
Also, to label the axes, use these commands:
set xlabel "x"
set ylabel "y"
set zlabel "z"
To show the level curves of the surface z = f (x, y) on both the surface and projected
onto the xy-plane, use this command:
set contour both
The default mesh size for the grid on the surface is 10 units. To get more of a col-
ored/shaded surface, increase the mesh size (to, say, 25) like this:
set isosamples 25
Putting all this together, we get the following graph with these commands:
set zeroaxis
set view 60, 120, 1, 1
set xlabel "x"
set ylabel "y"
set zlabel "z"
set contour both
set isosamples 25
splot [−1 : 1][−2 : 2] 2*(x**2) + y**2, exp(x+y)
199
-1
-0.5
0
0.5
1
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0
5
10
15
20
25
z
2*(x**2) + y**2
6
5
4
3
2
1
exp(x+y)
20
15
10
5
x
y
z
The numbers listed below the functions in the key in the upper right corner of the
graph are the “levels” of the level curves of the corresponding surface. That is, they
are the numbers c such that f (x, y) = c. If you do not want the function key displayed,
it can be turned off with this command: unset key
PARAMETRIC FUNCTIONS
Gnuplot has the ability to graph surfaces given in various parametric forms. For
example, for a surface parametrized in cylindrical coordinates
x = r cos θ , y = r sin θ , z = z
you would do the following:
set mapping cylindrical
set parametric
splot [a : b][c : d] v*cos(u),v*sin(u),f(u,v)
where the variable u represents θ, with a ≤ u ≤ b, the variable v represents r, with
c ≤ v ≤ d, and z = f (u, v) is some function of u and v.
Example C.2. The graph of the helicoid z = θ in Example 1.34 from Section 1.7 (p. 49)
was created using the following commands:
200 Appendix C: 3D Graphing with Gnuplot
set mapping cylindrical
set parametric
set view 60, 120, 1, 1
set xyplane 0
set xlabel "x"
set ylabel "y"
set zlabel "z"
unset key
set isosamples 15
splot [0 : 4*pi][0 : 2] v*cos(u),v*sin(u),u
The command set xyplane 0 moves the z-axis so that z = 0 aligns with the xy-plane
(which is not the default in Gnuplot). Looking at the graph, you will see that r varies
from 0 to 2, and θ varies from 0 to 4π.
PRINTING AND SAVING
In Windows, to print a graph from Gnuplot right-click on the titlebar of the graph’s
window, select “Options” and then the “Print..” option. To save a graph, say, as a PNG
file, go to the File menu on the main Gnuplot menubar, select “Output Device ...”, and
enter png in the Terminal type? textfield, hit OK. Then, in the File menu again, select
the “Output ...” option and enter a filename (say, graph.png) in the Output filename?
textfield, hit OK. Now run your splot command again and you should see a file called
graph.png in the current directory (usually the directory where wgnuplot.exe is lo-
cated, though you can change that setting using the “Change Directory ...” option in
In Linux, to save the graph as a file called graph.png, you would issue the following
commands:
set terminal png
set output ’graph.png’
and then run your splot command. There are many terminal types (which determine
the output format). Run the command set terminal to see all the possible types. In
Linux, the postscript terminal type is popular, since the print quality is high and
there are many PostScript viewers available.
To quit Gnuplot, type quit at the gnuplot> command prompt.
Version 1.2, November 2002
Copyright c 2000,2001,2002 Free Software Foundation, Inc.
51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
Everyone is permitted to copy and distribute verbatim copies of this license
document, but changing it is not allowed.
Preamble
The purpose of this License is to make a manual, textbook, or other functional
and useful document "free" in the sense of freedom: to assure everyone the effec-
tive freedom to copy and redistribute it, with or without modifying it, either com-
mercially or noncommercially. Secondarily, this License preserves for the author and
publisher a way to get credit for their work, while not being considered responsible
This License is a kind of "copyleft", which means that derivative works of the doc-
ument must themselves be free in the same sense. It complements the GNU General
because free software needs free documentation: a free program should come with
manuals providing the same freedoms that the software does. But this License is not
limited to software manuals; it can be used for any textual work, regardless of subject
matter or whether it is published as a printed book. We recommend this License
principally for works whose purpose is instruction or reference.
1. APPLICABILITY AND DEFINITIONS
This License applies to any manual or other work, in any medium, that contains
of this License. Such a notice grants a world-wide, royalty-free license, unlimited
in duration, to use that work under the conditions stated herein. The "Document",
below, refers to any such manual or work. Any member of the public is a licensee, and
is addressed as "you". You accept the license if you copy, modify or distribute the work
in a way requiring permission under copyright law.
201
A "Modified Version" of the Document means any work containing the Document
or a portion of it, either copied verbatim, or with modifications and/or translated into
another language.
A "Secondary Section" is a named appendix or a front-matter section of the Doc-
ument that deals exclusively with the relationship of the publishers or authors of the
Document to the Document’s overall subject (or to related matters) and contains noth-
ing that could fall directly within that overall subject. (Thus, if the Document is in
part a textbook of mathematics, a Secondary Section may not explain any mathemat-
ics.) The relationship could be a matter of historical connection with the subject or
with related matters, or of legal, commercial, philosophical, ethical or political posi-
tion regarding them.
The "Invariant Sections" are certain Secondary Sections whose titles are desig-
nated, as being those of Invariant Sections, in the notice that says that the Document
is released under this License. If a section does not fit the above definition of Sec-
ondary then it is not allowed to be designated as Invariant. The Document may con-
tain zero Invariant Sections. If the Document does not identify any Invariant Sections
then there are none.
The "Cover Texts" are certain short passages of text that are listed, as Front-Cover
Texts or Back-Cover Texts, in the notice that says that the Document is released under
this License. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may
be at most 25 words.
A "Transparent" copy of the Document means a machine-readable copy, repre-
sented in a format whose specification is available to the general public, that is suit-
able for revising the document straightforwardly with generic text editors or (for im-
ages composed of pixels) generic paint programs or (for drawings) some widely avail-
able drawing editor, and that is suitable for input to text formatters or for automatic
translation to a variety of formats suitable for input to text formatters. A copy made in
an otherwise Transparent file format whose markup, or absence of markup, has been
arranged to thwart or discourage subsequent modification by readers is not Transpar-
ent. An image format is not Transparent if used for any substantial amount of text.
A copy that is not "Transparent" is called "Opaque".
Examples of suitable formats for Transparent copies include plain ASCII without
markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly
available DTD, and standard-conforming simple HTML, PostScript or PDF designed
for human modification. Examples of transparent image formats include PNG, XCF
and JPG. Opaque formats include proprietary formats that can be read and edited
only by proprietary word processors, SGML or XML for which the DTDand/or process-
ing tools are not generally available, and the machine-generated HTML, PostScript
or PDF produced by some word processors for output purposes only.
The "Title Page" means, for a printed book, the title page itself, plus such following
pages as are needed to hold, legibly, the material this License requires to appear in the
title page. For works in formats which do not have any title page as such, "Title Page"
203
means the text near the most prominent appearance of the work’s title, preceding the
beginning of the body of the text.
A section "Entitled XYZ" means a named subunit of the Document whose title
either is precisely XYZ or contains XYZ in parentheses following text that translates
XYZ in another language. (Here XYZ stands for a specific section name mentioned be-
low, such as "Acknowledgments", "Dedications", "Endorsements", or "History".)
To "Preserve the Title" of such a section when you modify the Document means that
it remains a section "Entitled XYZ" according to this definition.
The Document may include Warranty Disclaimers next to the notice which states
that this License applies to the Document. These Warranty Disclaimers are consid-
ered to be included by reference in this License, but only as regards disclaiming war-
ranties: any other implication that these Warranty Disclaimers may have is void and
has no effect on the meaning of this License.
2. VERBATIM COPYING
You may copy and distribute the Document in any medium, either commercially or
notice saying this License applies to the Document are reproduced in all copies, and
that you add no other conditions whatsoever to those of this License. You may not
use technical measures to obstruct or control the reading or further copying of the
copies you make or distribute. However, you may accept compensation in exchange
for copies. If you distribute a large enough number of copies you must also follow the
conditions in section 3.
You may also lend copies, under the same conditions stated above, and you may
publicly display copies.
3. COPYING IN QUANTITY
If you publish printed copies (or copies in media that commonly have printed cov-
ers) of the Document, numbering more than 100, and the Document’s license notice
requires Cover Texts, you must enclose the copies in covers that carry, clearly and
legibly, all these Cover Texts: Front-Cover Texts on the front cover, and Back-Cover
Texts on the back cover. Both covers must also clearly and legibly identify you as the
publisher of these copies. The front cover must present the full title with all words
of the title equally prominent and visible. You may add other material on the covers
in addition. Copying with changes limited to the covers, as long as they preserve the
title of the Document and satisfy these conditions, can be treated as verbatim copying
in other respects.
If the required texts for either cover are too voluminous to fit legibly, you should put
the first ones listed (as many as fit reasonably) on the actual cover, and continue the
If you publish or distribute Opaque copies of the Document numbering more than
100, you must either include a machine-readable Transparent copy along with each
Opaque copy, or state in or with each Opaque copy a computer-network location from
network protocols a complete Transparent copy of the Document, free of added mate-
rial. If you use the latter option, you must take reasonably prudent steps, when you
begin distribution of Opaque copies in quantity, to ensure that this Transparent copy
will remain thus accessible at the stated location until at least one year after the last
time you distribute an Opaque copy (directly or through your agents or retailers) of
that edition to the public.
It is requested, but not required, that you contact the authors of the Document well
before redistributing any large number of copies, to give them a chance to provide you
with an updated version of the Document.
4. MODIFICATIONS
You may copy and distribute a Modified Version of the Document under the condi-
tions of sections 2 and 3 above, provided that you release the Modified Version under
precisely this License, with the Modified Version filling the role of the Document, thus
licensing distribution and modification of the Modified Version to whoever possesses
a copy of it. In addition, you must do these things in the Modified Version:
A. Use in the Title Page (and on the covers, if any) a title distinct from that of the
Document, and from those of previous versions (which should, if there were any,
be listed in the History section of the Document). You may use the same title as a
previous version if the original publisher of that version gives permission.
B. List on the Title Page, as authors, one or more persons or entities responsible for
authorship of the modifications in the Modified Version, together with at least five
of the principal authors of the Document (all of its principal authors, if it has fewer
than five), unless they release you from this requirement.
C. State on the Title page the name of the publisher of the Modified Version, as the
publisher.
D. Preserve all the copyright notices of the Document.
F. Include, immediately after the copyright notices, a license notice giving the public
permission to use the Modified Version under the terms of this License, in the form
205
G. Preserve in that license notice the full lists of Invariant Sections and required
Cover Texts given in the Document’s license notice.
H. Include an unaltered copy of this License.
I. Preserve the section Entitled "History", Preserve its Title, and add to it an item
stating at least the title, year, new authors, and publisher of the Modified Version
as given on the Title Page. If there is no section Entitled "History" in the Docu-
ment, create one stating the title, year, authors, and publisher of the Document as
given on its Title Page, then add an item describing the Modified Version as stated
in the previous sentence.
J. Preserve the network location, if any, given in the Document for public access to
a Transparent copy of the Document, and likewise the network locations given in
the Document for previous versions it was based on. These may be placed in the
"History" section. You may omit a network location for a work that was published
at least four years before the Document itself, or if the original publisher of the
version it refers to gives permission.
K. For any section Entitled "Acknowledgments" or "Dedications", Preserve the Title
of the section, and preserve in the section all the substance and tone of each of the
contributor acknowledgments and/or dedications given therein.
L. Preserve all the Invariant Sections of the Document, unaltered in their text and
in their titles. Section numbers or the equivalent are not considered part of the
section titles.
M. Delete any section Entitled "Endorsements". Such a section may not be included
in the Modified Version.
N. Do not retitle any existing section to be Entitled "Endorsements" or to conflict in
title with any Invariant Section.
O. Preserve any Warranty Disclaimers.
If the Modified Version includes new front-matter sections or appendices that qual-
ify as Secondary Sections and contain no material copied from the Document, you
may at your option designate some or all of these sections as invariant. To do this,
add their titles to the list of Invariant Sections in the Modified Version’s license notice.
These titles must be distinct from any other section titles.
You may add a section Entitled "Endorsements", provided it contains nothing but
endorsements of your Modified Version by various parties–for example, statements of
peer review or that the text has been approved by an organization as the authoritative
definition of a standard.
You may add a passage of up to five words as a Front-Cover Text, and a passage of up
to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified
Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be
includes a cover text for the same cover, previously added by you or by arrangement
made by the same entity you are acting on behalf of, you may not add another; but
you may replace the old one, on explicit permission from the previous publisher that
The author(s) and publisher(s) of the Document do not by this License give per-
mission to use their names for publicity for or to assert or imply endorsement of any
Modified Version.
5. COMBINING DOCUMENTS
You may combine the Document with other documents released under this License,
under the terms defined in section 4 above for modified versions, provided that you
include in the combination all of the Invariant Sections of all of the original docu-
ments, unmodified, and list them all as Invariant Sections of your combined work in
its license notice, and that you preserve all their Warranty Disclaimers.
The combined work need only contain one copy of this License, and multiple iden-
tical Invariant Sections may be replaced with a single copy. If there are multiple
Invariant Sections with the same name but different contents, make the title of each
such section unique by adding at the end of it, in parentheses, the name of the original
author or publisher of that section if known, or else a unique number. Make the same
adjustment to the section titles in the list of Invariant Sections in the license notice
of the combined work.
In the combination, you must combine any sections Entitled "History" in the vari-
ous original documents, forming one section Entitled "History"; likewise combine any
sections Entitled "Acknowledgments", and any sections Entitled "Dedications". You
must delete all sections Entitled "Endorsements".
6. COLLECTIONS OF DOCUMENTS
You may make a collection consisting of the Document and other documents re-
leased under this License, and replace the individual copies of this License in the
various documents with a single copy that is included in the collection, provided that
you follow the rules of this License for verbatim copying of each of the documents in
all other respects.
You may extract a single document from such a collection, and distribute it individ-
ually under this License, provided you insert a copy of this License into the extracted
document, and follow this License in all other respects regarding verbatim copying of
that document.
207
7. AGGREGATION WITH INDEPENDENT WORKS
A compilation of the Document or its derivatives with other separate and indepen-
dent documents or works, in or on a volume of a storage or distribution medium, is
called an "aggregate" if the copyright resulting from the compilation is not used to
limit the legal rights of the compilation’s users beyond what the individual works per-
mit. When the Document is included in an aggregate, this License does not apply to
the other works in the aggregate which are not themselves derivative works of the
Document.
If the Cover Text requirement of section 3 is applicable to these copies of the Doc-
ument, then if the Document is less than one half of the entire aggregate, the Doc-
ument’s Cover Texts may be placed on covers that bracket the Document within the
aggregate, or the electronic equivalent of covers if the Document is in electronic form.
Otherwise they must appear on printed covers that bracket the whole aggregate.
8. TRANSLATION
Translation is considered a kind of modification, so you may distribute translations
of the Document under the terms of section 4. Replacing Invariant Sections with
translations requires special permission from their copyright holders, but you may
include translations of some or all Invariant Sections in addition to the original ver-
sions of these Invariant Sections. You may include a translation of this License, and
all the license notices in the Document, and any Warranty Disclaimers, provided that
you also include the original English version of this License and the original versions
of those notices and disclaimers. In case of a disagreement between the translation
and the original version of this License or a notice or disclaimer, the original version
will prevail.
If a section in the Document is Entitled "Acknowledgments", "Dedications", or "His-
tory", the requirement (section 4) to Preserve its Title (section 1) will typically require
changing the actual title.
9. TERMINATION
You may not copy, modify, sublicense, or distribute the Document except as ex-
pressly provided for under this License. Any other attempt to copy, modify, sublicense
or distribute the Document is void, and will automatically terminate your rights un-
der this License. However, parties who have received copies, or rights, from you under
this License will not have their licenses terminated so long as such parties remain in
full compliance.
10. FUTURE REVISIONS OF THIS LICENSE
The Free Software Foundation may publish new, revised versions of the GNU Free
Documentation License from time to time. Such new versions will be similar in spirit
to the present version, but may differ in detail to address new problems or concerns.
See http://www.gnu.org/copyleft/.
Each version of the License is given a distinguishing version number. If the Doc-
ument specifies that a particular numbered version of this License "or any later ver-
sion" applies to it, you have the option of following the terms and conditions either of
that specified version or of any later version that has been published (not as a draft)
by the Free Software Foundation. If the Document does not specify a version number
of this License, you may choose any version ever published (not as a draft) by the Free
Software Foundation.
To use this License in a document you have written, include a copy of the License in
the document and put the following copyright and license notices just after the title
page:
and/or modify this document under the terms of the GNU Free Documenta-
Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-
Cover Texts. A copy of the license is included in the section entitled "GNU Free
If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, replace
the "with...Texts." line with this:
with the Invariant Sections being LIST THEIR TITLES, with the Front-Cover
Texts being LIST, and with the Back-Cover Texts being LIST.
If you have Invariant Sections without Cover Texts, or some other combination of
the three, merge those two alternatives to suit the situation.
If your document contains nontrivial examples of program code, we recommend
releasing these examples in parallel under your choice of free software license, such
as the GNU General Public License, to permit their use in free software.
History
This section contains the revision history of the book. For persons making modifications
to the book, please record the pertinent information here, following the format in the
first item below.
1. VERSION: 1.0
Date: 2008-01-04
Author(s): Michael Corral
Title: Vector Calculus
Modification(s): Initial version
209
Index
Symbols
D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
M
x
, M
y
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
M
xy
, M
xz
, M
yz
. . . . . . . . . . . . . . . . . . . . . . . 126
∆ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
¯ x. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
¯ y. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
¯ z . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
δ(x, y) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
∂(x, y, z)
∂(u, v, w)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
∂f
∂x
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

S
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

R
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

C
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136, 139
C
1
, C

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
∇ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80, 177

2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

Σ
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

C
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
∂. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
D
v
f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
e
r
, e
θ
, e
z
, e
ρ
, e
φ
. . . . . . . . . . . . . . . . . . . . . 181
dr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
i, j, k . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
A
acceleration. . . . . . . . . . . . . . . . . . . . . . . 2, 55
angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
annulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
arc length . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
area element . . . . . . . . . . . . . . . . . . . . . . . 105
average value . . . . . . . . . . . . . . . . . . . . . . 113
B
Bézier curve . . . . . . . . . . . . . . . . . . . . . . . . . 56
Beta function . . . . . . . . . . . . . . . . . . . . . . 123
C
capping surface . . . . . . . . . . . . . . . . . . . . 175
Cauchy-Schwarz Inequality . . . . . . . . . 17
center of mass. . . . . . . . . . . . . . . . . . . . . . 124
centroid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Chain Rule . . . . . . . . . . . . . . . . . . . . . 60, 147
change of variable. . . . . . . . . . . . . 117, 119
circulation . . . . . . . . . . . . . . . . . . . . . . . . . 174
closed curve . . . . . . . . . . . . . . . . . . . . . . . . 145
closed surface . . . . . . . . . . . . . . . . . . . . . . 161
collinear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
conical helix. . . . . . . . . . . . . . . . . . . . . . . . 167
conservative field . . . . . . . . . . . . . . . . . . 148
constrained critical point. . . . . . . . . . . . 96
continuity . . . . . . . . . . . . . . . . . . . . . . . 52, 69
continuously differentiable . . . . . . 59, 80
coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Cartesian. . . . . . . . . . . . . . . . . . . . . . . . . 1
curvilinear . . . . . . . . . . . . . . . . . . . . . . 47
cylindrical . . . . . . . . . . . . . . . . . . 47, 182
ellipsoidal . . . . . . . . . . . . . . . . . . . . . . 164
left-handed . . . . . . . . . . . . . . . . . . . . . . . 2
polar . . . . . . . . . . . . . . . . . . . . . . . 47, 121
rectangular . . . . . . . . . . . . . . . . . . . . . . . 1
210
Index 211
right-handed . . . . . . . . . . . . . . . . . . . . . 2
spherical . . . . . . . . . . . . . . . . . . . 47, 182
coplanar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
correlation . . . . . . . . . . . . . . . . . . . . . . . . . 134
covariance. . . . . . . . . . . . . . . . . . . . . . . . . . 134
critical point. . . . . . . . . . . . . . . . . . . . . . . . . 83
cross product . . . . . . . . . . . . . . . . . . . . . . . . 20
curl . . . . . . . . . . . . . . . . . . . . . . 169, 178, 182
curvature. . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
D
density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
derivative. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
directional. . . . . . . . . . . . . . . . . . . . . . . 78
mixed partial . . . . . . . . . . . . . . . . . . . . 73
partial . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
vector-valued function . . . . . . . . . . 52
determinant . . . . . . . . . . . . . . . . . . . . . . . . . 26
differential . . . . . . . . . . . . . . . . . . . . . . . . . 139
differential form . . . . . . . . . . . . . . . . . . . 139
directed curve. . . . . . . . . . . . . . . . . . . . . . 144
direction angles . . . . . . . . . . . . . . . . . . . . . 19
direction cosines. . . . . . . . . . . . . . . . . . . . . 19
directional derivative . . . . . . . . . . . . . . . 78
distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
between points. . . . . . . . . . . . . . . . . 6, 7
from point to line. . . . . . . . . . . . . . . . 33
point to plane. . . . . . . . . . . . 37, 41, 42
distribution function. . . . . . . . . . . . . . . 129
joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
normal . . . . . . . . . . . . . . . . . . . . . . . . . 130
divergence . . . . . . . . . . . . . . . 162, 177, 182
Divergence Theorem. . . . . . . . . . . . . . . 162
dot product . . . . . . . . . . . . . . . . . . . . . . . . . . 15
double integral . . . . . . . . . . . . . . . . 102, 105
polar coordinates . . . . . . . . . . . . . . 121
doubly ruled surface. . . . . . . . . . . . . . . . .45
E
ellipsoid . . . . . . . . . . . . . . . . . . . 43, 123, 164
elliptic cone . . . . . . . . . . . . . . . . . . . . . . . . . 45
elliptic paraboloid . . . . . . . . . . . . . . . . . . . 44
Euclidean space . . . . . . . . . . . . . . . . . . . . . . 1
exact differential form. . . 139, 154, 175
expected value . . . . . . . . . . . . . . . . . . . . . 132
extreme point . . . . . . . . . . . . . . . . . . . . . . . 83
F
flux. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
continuous . . . . . . . . . . . . . . . . . . . . . . 69
scalar. . . . . . . . . . . . . . . . . . . . . . . . . . . .53
vector-valued. . . . . . . . . . . . . . . . . . . . 51
G
Gaussian blur . . . . . . . . . . . . . . . . . . . . . . . 70
global maximum . . . . . . . . . . . . . . . . . . . . 83
global minimum. . . . . . . . . . . . . . . . . . . . . 83
gradient . . . . . . . . . . . . . . . . . . . . . . . . 80, 182
Green’s identities . . . . . . . . . . . . . . . . . . 186
Green’s Theorem. . . . . . . . . . . . . . . . . . . 150
H
harmonic . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
helicoid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
helix . . . . . . . . . . . . . . . . . . . . . . . . 51, 59, 167
hyperbolic paraboloid . . . . . . . . . . . . . . . 44
hyperboloid. . . . . . . . . . . . . . . . . . . . . . . . . . 43
one sheet . . . . . . . . . . . . . . . . . . . . . . . . 43
two sheets. . . . . . . . . . . . . . . . . . . . . . . 43
hypersurface . . . . . . . . . . . . . . . . . . . . . . . 110
hypervolume . . . . . . . . . . . . . . . . . . . . . . . 110
I
improper integral . . . . . . . . . . . . . . . . . . 108
integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
double. . . . . . . . . . . . . . . . . . . . . 102, 105
improper . . . . . . . . . . . . . . . . . . . . . . . 108
iterated . . . . . . . . . . . . . . . . . . . . . . . . 102
multiple. . . . . . . . . . . . . . . . . . . . . . . . 101
surface . . . . . . . . . . . . . . . . . . . . 156, 158
triple. . . . . . . . . . . . . . . . . . . . . . . . . . . 110
irrotational . . . . . . . . . . . . . . . . . . . . . . . . . 174
212 Index
iterated integral . . . . . . . . . . . . . . . . . . . 102
J
Jacobi identity . . . . . . . . . . . . . . . . . . . . . . 30
Jacobian . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
joint distribution. . . . . . . . . . . . . . . . . . . 131
L
Lagrange multiplier. . . . . . . . . . . . . . . . . 96
lamina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Laplacian . . . . . . . . . . . . . . . . . . . . . 178, 182
level curve. . . . . . . . . . . . . . . . . . . . . . . . . . . 66
limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
vector-valued function . . . . . . . . . . 52
line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
intersection of planes . . . . . . . . . . . 38
parallel . . . . . . . . . . . . . . . . . . . . . . . . . . 34
parametric representation . . . . . . 31
perpendicular . . . . . . . . . . . . . . . . . . . 34
skew. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
symmetric representation. . . . . . . 32
through two points . . . . . . . . . . . . . . 33
vector representation . . . . . . . . . . . 31
line integral . . . . . . . . . . . . . . . . . . . 136, 139
local maximum. . . . . . . . . . . . . . . . . . . . . . 83
local minimum . . . . . . . . . . . . . . . . . . . . . . 83
M
mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
matrix. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
mixed partial derivative. . . . . . . . . . . . . 73
Möbius strip . . . . . . . . . . . . . . . . . . . . . . . 168
moment . . . . . . . . . . . . . . . . . . . . . . . 124, 126
momentum. . . . . . . . . . . . . . . . . . . . . . . . . . 55
Monte Carlo method . . . . . . . . . . . . . . . 113
moving frame fields . . . . . . . . . . . . . . . . . 62
multiple integral . . . . . . . . . . . . . . . . . . . 101
multiply connected. . . . . . . . . . . . . . . . . 153
N
n-positive direction . . . . . . . . . . . . . . . . 169
Newton’s algorithm . . . . . . . . . . . . . . . . . 89
normal derivative . . . . . . . . . . . . . . . . . . 186
normal to a curve . . . . . . . . . . . . . . . . . . . 81
normal vector field. . . . . . . . . . . . . . . . . 168
O
orientable . . . . . . . . . . . . . . . . . . . . . . . . . . 168
orthonormal vectors . . . . . . . . . . . . . . . . . 64
outward normal . . . . . . . . . . . . . . . . . . . . 160
P
paraboloid. . . . . . . . . . . . . . . . . . . . . . . . . . . 44
elliptic. . . . . . . . . . . . . . . . . . . . . . . . . . . 44
hyperbolic . . . . . . . . . . . . . . . . . . . 44, 84
of revolution. . . . . . . . . . . . . . . . . . . . . 44
parallelepiped . . . . . . . . . . . . . . . . . . . . . . . 24
volume . . . . . . . . . . . . . . . . . . . . . . . . . . 25
parameter . . . . . . . . . . . . . . . . . . . . . . . 31, 60
parametrization. . . . . . . . . . . . . . . . . . . . . 60
partial derivative. . . . . . . . . . . . . . . . . . . . 71
partial differential equation. . . . . . . . . 74
path independence. . . . . . . 146, 154, 175
piecewise smooth curve . . . . . . . . . . . . 141
plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
coordinate . . . . . . . . . . . . . . . . . . . . . . . . 1
Euclidean. . . . . . . . . . . . . . . . . . . . . . . . . 1
in space . . . . . . . . . . . . . . . . . . . . . . . . . 35
line of intersection . . . . . . . . . . . . . . 38
normal form. . . . . . . . . . . . . . . . . . . . . 35
normal vector . . . . . . . . . . . . . . . . . . . 35
point-normal form. . . . . . . . . . . . . . . 35
tangent. . . . . . . . . . . . . . . . . . . . . . . . . . 75
through three points . . . . . . . . . . . . 36
position vector . . . . . . . . . . . . . . 54, 55, 139
potential . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
probability . . . . . . . . . . . . . . . . . . . . . . . . . 128
probability density function. . . . . . . . 129
projection. . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Q
quadric surface. . . . . . . . . . . . . . . . . . . . . . 43
R
random variable . . . . . . . . . . . . . . . . . . . 128
Riemann integral . . . . . . . . . . . . . . . . . . 135
Index 213
right-hand rule. . . . . . . . . . . . . . . . . 21, 192
ruled surface . . . . . . . . . . . . . . . . . . . . . . . . 45
S
saddle point . . . . . . . . . . . . . . . . . . . . . . . . . 85
sample space. . . . . . . . . . . . . . . . . . . . . . . 128
scalar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
combination. . . . . . . . . . . . . . . . . . . . . 12
scalar function . . . . . . . . . . . . . . . . . . . . . . 53
scalar triple product. . . . . . . . . . . . . . . . . 25
Second Derivative Test . . . . . . . . . . . . . . 84
second moment . . . . . . . . . . . . . . . . . . . . 134
second-degree equation . . . . . . . . . . . . . 43
simple closed curve . . . . . . . . . . . . . . . . 145
simply connected. . . . . . . . . . . . . . 154, 175
smooth function . . . . . . . . . . . . . . . . . 59, 84
solenoidal . . . . . . . . . . . . . . . . . . . . . . . . . . 163
span. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
spherical spiral . . . . . . . . . . . . . . . . . . . . . . 54
standard normal distribution . . . . . . 130
steepest descent . . . . . . . . . . . . . . . . . . . . . 95
stereographic projection. . . . . . . . . . . . . 46
Stokes’ Theorem . . . . . . . . . . . . . . 168, 169
surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
doubly ruled. . . . . . . . . . . . . . . . . . . . . 45
orientable . . . . . . . . . . . . . . . . . . . . . . 168
ruled . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
two-sided . . . . . . . . . . . . . . . . . . . . . . 168
surface integral . . . . . . . . . . . . . . . 156, 158
T
tangent plane . . . . . . . . . . . . . . . . . . . . . . . 75
torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
trace. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
triangle inequality . . . . . . . . . . . . . . . . . . 18
triple integral . . . . . . . . . . . . . . . . . . . . . . 110
cylindrical coordinates . . . . . . . . . 122
spherical coordinates . . . . . . . . . . 122
U
uniform density . . . . . . . . . . . . . . . . . . . . 124
uniform distribution . . . . . . . . . . . . . . . 129
uniformly distributed . . . . . . . . . . . . . . 128
unit disk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
V
variance. . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
addition . . . . . . . . . . . . . . . . . . . . . . . . . . 9
angle between. . . . . . . . . . . . . . . . . . . 15
basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
components . . . . . . . . . . . . . . . . . . . . . 13
direction. . . . . . . . . . . . . . . . . . . . . . . . . . 3
magnitude . . . . . . . . . . . . . . . . . . . . . 3, 7
normal . . . . . . . . . . . . . . . . . . . . . 35, 160
normalized . . . . . . . . . . . . . . . . . . . . . . 12
parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
perpendicular . . . . . . . . . . . . . . . 16, 17
positive unit normal . . . . . . . . . . . 169
principal normal N. . . . . . . . . . . . . . 64
scalar multiplication . . . . . . . . . . . . . 9
subtraction. . . . . . . . . . . . . . . . . . . . . . 10
tangent. . . . . . . . . . . . . . . . . . . . . . . . . . 52
translation. . . . . . . . . . . . . . . . . . . . . 5, 9
unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
unit binormal B. . . . . . . . . . . . . . . . . 64
unit tangent T . . . . . . . . . . . . . . . . . . 64
zero . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3, 4
vector field . . . . . . . . . . . . . . . . . . . . . . . . . 138
normal . . . . . . . . . . . . . . . . . . . . . . . . . 168
smooth. . . . . . . . . . . . . . . . . . . . . . . . . 150
vector triple product. . . . . . . . . . . . . . . . .25
velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 55
volume element . . . . . . . . . . . . . . . . . . . . 110
W
wave equation. . . . . . . . . . . . . . . . . . . . . . . 74
work . . . . . . . . . . . . . . . . . . . . . . . . . . 135, 166
Z
zenith angle . . . . . . . . . . . . . . . . . . . . . . . . . 47

Vector Calculus
Michael Corral
Schoolcraft College

About the author: Michael Corral is an Adjunct Faculty member of the Department of Mathematics at Schoolcraft College. He received a B.A. in Mathematics from the University of California at Berkeley, and received an M.A. in Mathematics and an M.S. in Industrial & Operations Engineering from the University of Michigan.
A This text was typeset in LTEX 2ε with the KOMA-Script bundle, using the GNU Emacs text editor on a Fedora Linux system. The graphics were created using MetaPost, PGF, and Gnuplot.

Copyright c 2008 Michael Corral. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled “GNU Free Documentation License”.

I would rate it as a 5. There are 420 exercises throughout the text. and perhaps even easier with a functional programming language (such as Haskell or Scheme). which allows others to not only copy and distribute the book but also to modify it. While it would have been simple to use a scripting language like Python. see the included copy of the GFDL. and the C exercises usually require some effort or insight to solve. “Moderate” and “Challenging”. The A exercises are mostly of a routine computational nature. However. with 1 being completely informal and 10 being completely rigorous.a. It is suitable for a one-semester course.k. So that there is no ambiguity on this iii . divided into three categories: A. The prerequisites are the standard courses in single-variable calculus (a. B and C would be “Easy”. There are exercises at the end of each section. Appendix B contains a proof of the right-hand rule for the cross product. Answers and hints to most odd-numbered and some even-numbered exercises are provided in Appendix A. respectively.Preface This book covers calculus in two and three variables.g. This book is released under the GNU Free Documentation License (GFDL). normally known as “Vector Calculus”. I have tried to be somewhat rigorous about proving results. Java was chosen due to its ubiquity. the B exercises are slightly more involved. For more details.4). relatively clear syntax. Calculus I and II). There are a few exercises that require the student to write his or her own computer program to solve some numerical approximation problems (e. Appendix C contains a brief tutorial on Gnuplot for graphing functions of two variables. B and C. The code samples in the text are in the Java programming language. If I were to rate the level of rigor in the book on a scale of 1 to 10. A crude way of describing A. the Monte Carlo method for approximating multiple integrals. or simply “Calculus III”.since that is how mathematics works . in Section 3. But while it is important for students to see full-blown proofs .too much rigor and emphasis on proofs can impede the flow of learning for the vast majority of the audience at this level. “Multivariable Calculus”. hopefully with enough comments so that the reader can figure out what is being done even without knowing Java. so students are free to implement the solutions using the language of their choice. many of the B exercises are easy and not all the C exercises are difficult. Those exercises do not mandate the use of Java. and easy availability for multiple platforms. which in my experience are more than enough for a semester course in this subject. which seems to have virtually disappeared from calculus texts over the last few decades.

comments. without needing my permission. and for finding the numerous errors and typos it contained. etc).g. Finally. suggestions. I would like to thank my students in Math 240 for being the guinea pigs for the initial draft of this book.edu for any questions on this or any other matter involving the book (e. January 2008 M ICHAEL C ORRAL . corrections. I welcome your input.net). anyone can make as many copies of this book as desired and distribute it as desired. The PDF version will always be freely available to the public at no cost (go to http://www.mecmath. Feel free to contact me at mcorral@schoolcraft.iv Preface matter.

. . 1. . . . . . . . .3 Triple Integrals . . . . . .3 Tangent Plane to a Surface . . . . . . . . . . . . . . . . . .Contents Preface 1 Vectors in Euclidean Space 1. . . . . . . . 1. . . . . . . . . . . . . . . . . . . . .1 Double Integrals . . . . . . . . . . . . . . 3. . . . . . . . . . . 2. . . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Unconstrained Optimization: Numerical Methods 2. . . . . . . . . . . . . . . . . .6 Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . 1. . . . . . . . . . .4 Directional Derivatives and the Gradient . . . .4 Numerical Approximation of Multiple Integrals 3. . . . . . . . . . . . . . . . . . . . . . . .5 Lines and Planes . . . . . . . . . .9 Arc Length . . .5 Change of Variables in Multiple Integrals . .2 Properties of Line Integrals . . . . . . . . . . . . . . . . . .5 Maxima and Minima . . . . . . . . . .1 Introduction . . . . . . . . . . . . . . . . . . .7 Application: Probability and Expected Value . . . . 150 v . . . . . . . . 143 4. . . . . . . . . .6 Application: Center of Mass . . . . . . . . . . . . . .4 Cross Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 Curvilinear Coordinates 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. . . 2. . . . . . . . . . . . .2 Partial Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . .3 Green’s Theorem . . . . . . . .3 Dot Product . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Line and Surface Integrals 135 4. . . . 2. . . . . . . . . . . . . . . . . . . . . . . . 3 Multiple Integrals 3. .1 Line Integrals . . . . . . . . . . . . . . .2 Double Integrals Over a General Region . . . . . . . . . . . .8 Vector-Valued Functions 1. . . . . . . . . . . . . 2 Functions of Several Variables 2. . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Vector Algebra . . . . . . . . . . . . . 1. . . . . . . iii 1 1 9 15 20 31 40 47 51 59 65 65 71 75 78 83 89 96 101 101 105 110 113 117 124 128 . 135 4. . .1 Functions of Two or Three Variables . . . . . . . . . .7 Constrained Optimization: Lagrange Multipliers . . . 3. . . . . . . . . . . . . . . . . . . . . . .

. 156 4. . . . . . . . . . . . . . . . . . . . . .4 Surface Integrals and the Divergence Theorem . . . .5 Stokes’ Theorem . . Divergence. . . . . . . . . . 177 Bibliography Appendix A: Answers and Hints to Selected Exercises Appendix B: Proof of the Right-Hand Rule for the Cross Product Appendix C: 3D Graphing with Gnuplot 187 189 192 196 201 209 210 GNU Free Documentation License History Index . . . . . . . . . . . . . . . . .6 Gradient. . . . Curl and Laplacian . . .vi Contents 4. . . . 165 4. . . . .

1.1. in the Cartesian or rectangular coordinate system. lies in Euclidean space. y). We use the word “Euclidean” to denote a system in which all the usual rules of Euclidean geometry hold. z. such as this page or a blackboard. b. y)). the functions that one encounters are functions of a variable (usually x or t) that varies over some subset of the real number line (which we denote by ). In vector (or multivariable) calculus. z = f (x. The Euclidean plane has two perpendicular coordinate axes: the x-axis and the y-axis. the graph of the function f consists of the points (x. y) = (x. y or x.1. z) = (x. respectively). y. we will deal with functions of two or three variables (usually x. y = f (x).2).1.1 Figure 1. c) b 0 a x x y xz-plane yz-plane z y 0 xy-plane Figure 1. y and z).1 Introduction In single-variable calculus. f (x. b). The graph of a function of two variables. Since Euclidean space is 3-dimensional. f (x)). c). The 3-dimensional coordinate system of Euclidean space can be represented on a flat surface. only by giving the illusion of three dimensions. in the manner shown in Figure 1. and three mutually perpendicular coordinate planes: the xy-plane. The graph of f consists of the points (x. We denote the Euclidean plane by 2 . y. yz-plane and xz-plane (see Figure 1. say. which.2 1 . which in the Cartesian coordinate system consists of all ordered triples of real numbers (a. z c P(a.1 Vectors in Euclidean Space 1. say. These points lie in the Euclidean plane. the “2” represents the number of dimensions of the plane. b.1. For such a function. consists of all ordered pairs of real numbers (a. y. Euclidean space has three mutually perpendicular coordinate axes (x. we denote it by 3 .

1.3. the middle finger in the positive direction of the y-axis. For functions of three variables.1. But what about something such as the velocity of the object. which we can not see in our 3-dimensional space. let alone simulate in 2-dimensional space. see the book by A BBOTT. For an entertaining discussion of this subject. So we can only think of 4-dimensional space abstractly.and y-axes in a right-handed system results in a left-handed system. the graphs exist in 4-dimensional space (i. using the right hand. and the thumb in the positive direction of the z-axis.e. 1 One thing you will learn is why a 4-dimensional creature would be able to reach inside an egg and remove the yolk without cracking the shell! . Doing the same thing with the left hand is what defines a left-handed coordinate system. as in Figure 1. or its acceleration? Or the gravitational force acting on the object? These phenomena all seem to involve motion and direction in some way. we have discussed the position of an object in 2-dimensional or 3-dimensional space.2 CHAPTER 1.3 Right-handed coordinate system An equivalent way of defining a right-handed system is if you can point your thumb upwards in the positive z-axis direction while using the remaining four fingers to rotate the x-axis towards the y-axis. 4 ).1 So far. to point the index finger in the positive direction of the x-axis. VECTORS IN EUCLIDEAN SPACE The coordinate system shown in Figure 1.1. and that rotating either type of system does not change its “handedness”. Notice that switching the x. Figure 1. because it is possible. This is where the idea of a vector comes in.1 is known as a right-handed coordinate system. Throughout the book we will use a right-handed system.

Then a is the magnitude of the velocity (normally called the speed of the object).1. which was called the velocity of the object. since they are just numbers. Note that our definition could apply to systems with any number of dimensions (see Figure 1. indicated by a plus or minus symbol (representing motion in the positive direction or the negative direction. and its direction is the same as that of the directed line segment. and it is denoted by 0.4 (a)-(c)). To indicate the direction of a vector. then dy/dt = f ′ (t) is the velocity of the object at time t. So you can think of that number. We will often denote a vector by a single bold-faced letter (e. which in elementary geometry is called a “directed line segment”. This is the motivation for how we will define a vector. For motion along a straight line. The zero vector is just a point. For example. Definition 1. denoted by PQ . i. A (nonzero) vector is a directed line segment drawn from a point P (called its initial point) to a point Q (called its terminal point). respectively).1. in a 1-dimensional space. for motion along a straight line. v) and use the terms “magnitude” and “length” interchangeably. as having two components: a magnitude. indicated by a nonnegative number.1 Introduction 3 You have already dealt with velocity and acceleration in single-variable calculus. and the ± represents the direction of the velocity (though the + is usually omitted for the positive direction). A geometric object which has those features is an arrow. and negative if it moves in the opposite of that direction. with P and Q being − − → distinct points.e.g.1.or 3-dimensional space. we draw an arrow from its initial point to its terminal point. velocity will need to be represented by a multidimensional object which should have both a magnitude and a direction. however. y z − −→Q P v − − → RS 0 R − − → PQ P Q x 0 P R − − → RS Q S x S x → − −S R R Q −→ − PQ y 0 P v S (a) One dimension (b) Two dimensions (c) Three dimensions Figure 1. The vector is denoted by PQ. The derivative f ′ (t) is just a number.4 Vectors in different dimensions . For general motion along a curve in 2. f ′ (t) = ±a for some number a ≥ 0. if y = f (t) gives the displacement of an object after time t. preceded by a direction. i. the velocities are also contained in that 1-dimensional space. which is positive if the object is moving in an agreedupon “positive” direction.e. Its magnitude is the length of the line − − → segment.1.

. This agrees with the definition of the zero vector as just a point. See A NTON and R ORRES.2 Now that we know what a vector is. Two nonzero vectors are equal if they have the same magnitude and the same direction. Definition 1. and we will leave it at that. some say that it has indeterminate direction (i. however. v and w all √ have the same magnitude 5 (by the Pythagorean Theorem). and they point in the 2 same direction. By this definition. Our motivation for what a vector is included the notions of magnitude and direction.e.e.5. Some contend that the zero vector has arbitrary direction (i. i. Is there a single vector which we can choose to represent all those equal vectors? The answer is yes. Our definition of the zero vector.e.5 the vectors u.4 CHAPTER 1. y 4 3 2 1 0 1 v w x 2 3 4 u Figure 1. even though they have different initial points. those vectors all being equal and differing only by their initial and terminal points. and is suggested by the vector w in Figure 1. vectors with the same magnitude and direction but with different initial points would be equal. does not require it to have a direction. the direction can not be determined).1.1.1. since they lie on lines having the same slope 1 . which is well-defined since the initial and terminal points are distinct. we need a way of determining when two vectors are equal. VECTORS IN EUCLIDEAN SPACE A few things need to be noted about the zero vector. What is the magnitude of the zero vector? We define it to be zero. So u v. which has zero length. can take any direction). We also see that v is parallel to u but points in the opposite direction. And we see that u and w are parallel. What about the direction of the zero vector? A single point really has no well-defined direction. in Figure 1. For example. Notice that we were careful to only define the direction of a nonzero vector.5 So we can see that there are an infinite number of vectors for a given magnitude and direction. 0 = 0. 2 In the subject of linear algebra there is a more abstract way of defining a vector where the concept of “direction” is not really used. This leads us to the following definition. So u = w. Any vector with zero magnitude is equal to the zero vector.2. Not everyone agrees on the direction of the zero vector. while others say that it has no direction.

0. when adding vectors. b.1. Also. . Do this for each original vector then compare. then the original vectors are equal. 4. 4. 5 Thinking of vectors as starting from the origin provides a way of dealing with vectors in a standard way. we will write the zero vector 0 in 2 and 3 as (0. 4.1. it is understood that the initial point of v is at the origin (0. respectively. since every coordinate system has an origin. 5).5) y 0 x (b) The vector (3. b) in 2 or v = (a. when speaking of “the vector” with a given magnitude and direction. 4. For each vector. When doing this. which we will do in the next section).1. Let v be the vector in 3 whose initial point is at the origin and whose terminal point is (3. The resulting point will be the terminal point of the “new” vector whose initial point is the origin. it is convenient to write v = (3. find the (unique!) vector it equals whose initial point is the origin. 4. 5) z v = (3.6 Correspondence between points and vectors Unless otherwise stated. The point-vector correspondence provides an easy way to check if two vectors are equal. when we refer to vectors as v = (a. 5) y 0 x (a) The point (3.1 Introduction Unless otherwise indicated. without having to determine their magnitude and direction.4. 0.5) Figure 1. But there will be times when it is convenient to consider a different initial point for a vector (for example. To get the “new” vectors starting at the origin. 4. Then compare the coordinates of the terminal points of these “new” vectors: if those coordinates are the same. you are now seeing if the terminal points of vectors starting at the origin are the same. 0).4. 5). Though the point (3. 5). we mean vectors in Cartesian coordinates starting at the origin. Example 1. 0) and (0. we will mean the one whose initial point is at the origin of the coordinate system. c) in 3 . 0) and the terminal point is (3. Another advantage of using the origin as the initial point is that it provides an easy correspondence between a vector and its terminal point. 5) and the vector v are different objects. you translate each vector to start at the origin by subtracting the coordinates of the original initial point from the original terminal point. Similar to seeing if two points are the same. z P(3.

1.6 CHAPTER 1. 0. 7). 1. y2 ) in d= 2. 0) and terminal point S − R = (2. 5. −3. − − → Similarly. 0) − (1. 2) and RS = w = (1. 1. where P = (2. 5. we have the following result: − − → For a vector PQ in 2 with initial point P = (x1 . the magnitude of PQ is: − − → PQ = (x2 − x1 )2 + (y2 − y1 )2 (1.1. 1 − (−3). y1 ) and terminal point − − → Q = (x2 .1) (x2 − x1 )2 + (y2 − y1 )2 By this formula. −3. 5) = (3 − 2. 4. 0 − (−2)) = (1. 4. 1. RS is equal to the vector w with initial point (0. 0. − − → − − → ∴ PQ = RS z → − −Q P P (2. VECTORS IN EUCLIDEAN SPACE − − → − − → Example 1.7 Recall the distance formula for points in the Euclidean plane: For points P = (x1 . 5). 0) and terminal point Q − P = (3. 7) − − → Translate PQ to v v=w (1. Q = (x2 . the distance d between P and Q is: (1. y1 ). 1. 5 − 1.2) . 4. y2 ). 0) y → − −S R x Figure 1. −2) = (2 − 1. 4. 7) − (2. R = − − → − − → (1. Does PQ = RS ? − − → Solution: The vector PQ is equal to the vector v with initial point (0. − − → − − → So PQ = v = (1. 5. Q = (3. 2). 0). 1. 7 − 5) = (1. −2) and S = (2. Consider the vectors PQ and RS in 3 . −3.2. 2). 2). −2) 0 S (2. 4. 2) − − → Translate RS to w R (1. 5) Q (3.

b. 0). y2 .1. z2 ) in d= (x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 3 is: (1.4) The proof will use the following result: Theorem 1.√ c). c) in v = 3. b. R = (a. c). This proves the theorem. 0. y1 . Then v = (0. b. this time to the right triangle √ △PQR. Q = (a. c are all positive (the other seven possibilities are handled in a similar manner). c).1. The distance d between points P = (x1 .1 Introduction Finding the magnitude of a vector v = (a. A second application of the Pythagorean Theorem.3) To calculate the magnitude of vectors in 3 . Then v = 0. b 0 and c 0 (the other two possibilities are handled in a similar manner). Case 4: none of a. Without loss of generality. we assume that a = 0. b.8 . 0) and Q = (a. we can assume that a. b.2. c is 0. gives v = |PQ| = |PR|2 + |QR|2 = a2 + b2 + c2 . z1 ) and Q = (x2 . b) : For a vector v = (a. so v = 0 = 02 + 02 + 02 = a2 + b2 + c2 . 0). b. Without loss of generality. which is a vector in the yz-plane. Applying the Pythagorean Theorem to the right triangle △PS R gives |PR|2 = a2 +b2 . Without loss of generality. b) in with P = (0. so by the Pythagorean Theorem we have b. 0. which is a vector of length |c| along the z-axis. √ √ v = b2 + c2 = 02 + b2 + c2 = a2 + b2 + c2 . and S = (a. 0). c) c y 0 b R a S x P Figure 1. So v = |c| = c2 = √ √ 02 + 02 + c2 = a2 + b2 + c2 . c are 0. c are 0.5) a2 + b2 + c2 Proof: There are four cases to consider: √ √ Case 1: a = b = c = 0. b) in 2. as shown in Figure 1. For a vector v = (a. QED z v Q(a.2) the magnitude of v is: v = a2 + b2 (1. √ Then v = (0. we assume that a = b = 0 and c 0 (the other two possibilities are handled in a similar manner). the magnitude of v is: (1. b. b. Case 2: exactly two of a.1. Case 3: exactly one of a.8. 0.1. 2 7 is a special case of formula (1. Consider the points P = (0. we need a distance formula for points in Euclidean space (we will postpone the proof until the next section): Theorem 1.

Show that w = |a| v . and consider Figure 1. 3. 2). 2) and Q = (5. z1 ) and Q = (x2 .8 CHAPTER 1. (Hint: Think of Case 4 in the proof of Theorem 1. y1 . 0) . Prove the special case of Theorem 1. and z2 > z1 > 0. −1. 0. 1). 0) x U(x2 . z1 ) R(x2 . √ √ Solution: By formula (1. 0.9 y C 6.  A Exercises ©  ¨ 1. 0) be vectors in 3. 5. 0. 5). 4). z1 ) 0 S (x1 . R = (1. v = 52 + 82 + (−2)2 = 25 + 64 + 4 = 93. √ √ Solution: By formula (1. Q = (1. y2 .1. z2 ) satisfy the following conditions: x2 > x1 > 0. R = (2. 0. 0) Figure 1. 2.1 where the points P = (x1 . 4. y2 . For the points P = (1. Let v = (a. y1 . −1.2. S = (2. z2 ) P(x1 . 3b. 0) (c) v = (3.5). it is possible to prove it using methods similar to those in the proof of Theorem 1. 0).3). Calculate the following: − − → (a) The magnitude of the vector PQ in − − → Solution: By formula (1.1. −2) (d) v = (0. −2) in 3 . y1 . 3c) be vectors in Show that w = 3 v . Q = (2. 1) (e) v = (6. b. Though we will see a simple proof of Theorem 1. −1) (b) v = (2. Solution: By formula (1.2).2. −1. y2 > y1 > 0.3. 0) and w = (a. (d) The magnitude of the vector v = (5. S = (3. −2. (b) The magnitude of the vector v = (8. y2 . c) and w = (3a.9. Calculate the magnitudes of the following vectors: (a) v = (2. 3) in 2 . Let v = (1. 0. 3. z Q(x2 . does PQ = RS ? − − → − − → 3. √ √ √ (5 − (−1))2 + (5 − 2)2 = 36 + 9 = 45 = 3 5.) T (x2 .4). −1. v = 82 + 32 = 73. 0. 2). 4) and Q = (4. 1). VECTORS IN EUCLIDEAN SPACE Example 1. 3. −3) in 2 . 8. 1). (c) The distance between the points P = (2. PQ = 2 with P = (−1. 2. 2). does PQ = RS ? B 4. For the points P = (0. the distance d = (4 − 2)2 + (2 − (−1))2 + (−3 − 4)2 = √ √ 4 + 9 + 49 = 62. y2 .1 in the next section. −4) − − → − − → 2. y1 .

electric charge. A scalar is a quantity that can be represented by a single number.g. Two vectors v and w are parallel (denoted by v w) if one is a scalar multiple of the other.2 Vector Algebra Now that we know what vectors are. and as flipping the vector in the opposite direction if the scalar is a negative number (see Figure 1.5. used in physics. 3 The term scalar was invented by 19th century Irish mathematician. where it means “carrier”. is that under certain types of coordinate transformations (e.5v −v −2v Figure 1. we can start to perform some of the usual algebraic operations on them (e. is obtained by translating w so that its initial point is at the terminal point of v. we will introduce the notion of a scalar. We are now ready to define the sum of two vectors.2. rotations). while a quantity that is affected (in a certain way) is a vector.3 Examples of scalar quantities are mass. we define k0 = 0 for any scalar k. the scalar multiple of v by k. points in the same direction as v if k > 0. and speed (not velocity). points in the opposite direction as v if k < 0.4. denoted by v + w. physicist and astronomer William Rowan Hamilton. For our purposes.g. For a scalar k and a nonzero vector v. and its terminal point is the new terminal point of w. The sum of vectors v and w. and is the zero vector 0 if k = 0. Definition 1. 4 An alternate definition of scalars and vectors. is the vector whose magnitude is |k| v . Definition 1. to convey the sense of something that could be represented by a point on a scale or graduated ruler.1 Recall that translating a nonzero vector means that the initial point of the vector is changed but the magnitude and direction are preserved. addition.2. Definition 1. a quantity that is not affected is a scalar. See M ARION for details.1.4 We can now define scalar multiplication of a vector. the initial point of v + w is the initial point of v. v 2v 3v 0. You can think of scalar multiplication of a vector as stretching or shrinking the vector.2 Vector Algebra 9 1. . subtraction).1).3. Before doing that. The word vector comes from Latin. denoted by kv. For the zero vector 0. scalars will always be real numbers.

that is. In fact. . See Figure 1.3 Subtracting vectors v and w Figure 1. since the scalar multiple −v = −1 v is a well-defined vector. we can define vector subtraction as follows: v − w = v + (−w).4 shows the use of “geometric proofs” of various laws of vector algebra. 0 + 0 = 0.2. w. (a) shows that v + w = w + v for any vectors v. it is easy to see that v + (−v) = 0. w w v (a) Vectors v and w v+w w v (c) The sum v + w v (b) Translate w to the end of v Figure 1.2.3.2.10 CHAPTER 1. v w w+v v+w v (a) Add vectors w w v−w −w v+w v v−w w v−w v (c) Combined add/subtract (b) Subtract vectors Figure 1. And (c) shows how you can think of v − w as the vector that is tacked on to the end of w to add up to v. and hence can be translated). Since we will deal mostly with Cartesian coordinates in this book. the following two theorems are useful for performing vector algebra on vectors in 2 and 3 starting at the origin. In general. v −w (b) Translate −w to the end of v w v (a) Vectors v and w v v−w −w (c) The difference v − w Figure 1. VECTORS IN EUCLIDEAN SPACE Intuitively. In particular. it uses laws from elementary geometry to prove statements about vectors.2 Adding vectors v and w Notice that our definition is valid for the zero vector (which is just a point. Also. we have not even mentioned coordinates in this section so far.2. adding w to v means tacking on w to the end of v (see Figure 1.2).2. as we would expect. For example.4 “Geometric” vector algebra Notice that we have temporarily abandoned the practice of starting vectors at the origin. and so we see that v + 0 = v = 0 + v for any vector v.2.

1.2 Vector Algebra Theorem 1.3. Let v = (v1 , v2 ), w = (w1 , w2 ) be vectors in (a) kv = (kv1 , kv2 ) (b) v + w = (v1 + w1 , v2 + w2 )
2,

11 and let k be a scalar. Then

Proof: (a) Without loss of generality, we assume that v1 , v2 > 0 (the other possibilities are handled in a similar manner). If k = 0 then kv = 0v = 0 = (0, 0) = (0v1 , 0v2 ) = (kv1 , kv2 ), which is what we needed to show. If k 0, then (kv1 , kv2 ) lies on a line with slope kv2 = v2 , which is the same as the slope of the line on which v (and hence kv) lies, kv1 v1 and (kv1 , kv2 ) points in the same direction on that line as kv. Also, by formula (1.3) the magnitude of (kv1 , kv2 ) is (kv1 )2 + (kv2 )2 = k2 v2 + k2 v2 = k2 (v2 + v2 ) = |k| v2 + v2 = 1 2 1 2 1 2 |k| v . So kv and (kv1 , kv2 ) have the same magnitude and direction. This proves (a).

(b) Without loss of generality, we assume that v2 + w2 y v v1 , v2 , w1 , w2 > 0 (the other possibilities are hanw2 w w2 dled in a similar manner). From Figure 1.2.5, w v+w we see that when translating w to start at v2 w1 v the end of v, the new terminal point of w is x (v1 + w1 , v2 + w2 ), so by the definition of v + w v1 + w 1 w1 v1 0 this must be the terminal point of v + w. This Figure 1.2.5 proves (b). QED Theorem 1.4. Let v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) be vectors in (a) kv = (kv1 , kv2 , kv3 ) (b) v + w = (v1 + w1 , v2 + w2 , v3 + w3 ) The following theorem summarizes the basic laws of vector algebra. Theorem 1.5. For any vectors u, v, w, and scalars k, l, we have (a) v + w = w + v Commutative Law (b) u + (v + w) = (u + v) + w (c) v + 0 = v = 0 + v (d) v + (−v) = 0 (e) k(lv) = (kl)v (f) k(v + w) = kv + kw (g) (k + l)v = kv + lv Associative Law Additive Identity Additive Inverse Associative Law Distributive Law Distributive Law
3,

let k be a scalar. Then

Proof: (a) We already presented a geometric proof of this in Figure 1.2.4(a). (b) To illustrate the difference between analytic proofs and geometric proofs in vector algebra, we will present both types here. For the analytic proof, we will use vectors in 3 (the proof for 2 is similar).

12

CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
3.

Let u = (u1 , u2 , u3 ), v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) be vectors in u + (v + w) = (u1 , u2 , u3 ) + ((v1 , v2 , v3 ) + (w1 , w2 , w3 )) = (u1 , u2 , u3 ) + (v1 + w1 , v2 + w2 , v3 + w3 )

Then

by Theorem 1.4(b)

= (u1 + (v1 + w1 ), u2 + (v2 + w2 ), u3 + (v3 + w3 )) by Theorem 1.4(b) = ((u1 + v1 ) + w1 , (u2 + v2 ) + w2 , (u3 + v3 ) + w3 ) by properties of real numbers = (u1 + v1 , u2 + v2 , u3 + v3 ) + (w1 , w2 , w3 ) = (u + v) + w This completes the analytic proof of (b). Figure 1.2.6 provides the geometric proof.
u + (v + w) = (u + v) + w

by Theorem 1.4(b)

v+w u u+v

w

Figure 1.2.6

v Associative Law for vector addition

(c) We already discussed this on p.10. (d) We already discussed this on p.10. (e) We will prove this for a vector v = (v1 , v2 , v3 ) in k(lv) = k(lv1 , lv2 , lv3 ) = (klv1 , klv2 , klv3 ) = (kl)(v1 , v2 , v3 ) = (kl)v (f) and (g): Left as exercises for the reader.
QED 3

(the proof for

2

is similar):

by Theorem 1.4(a) by Theorem 1.4(a) by Theorem 1.4(a)

A unit vector is a vector with magnitude 1. Notice that for any nonzero vector v, v 1 the vector v is a unit vector which points in the same direction as v, since v > 0
v and v = v = 1. Dividing a nonzero vector v by v is often called normalizing v. v There are specific unit vectors which we will often use, called the basis vectors: i = (1, 0, 0), j = (0, 1, 0), and k = (0, 0, 1) in 3 ; i = (1, 0) and j = (0, 1) in 2 . These are useful for several reasons: they are mutually perpendicular, since they lie on distinct coordinate axes; they are all unit vectors: i = j = k = 1; every vector can be written as a unique scalar combination of the basis vectors: v = (a, b) = a i + b j in 2 , v = (a, b, c) = a i + b j + c k in 3 . See Figure 1.2.7.

1.2 Vector Algebra
z z

13

2 y y 1 k

v = (a, b, c)

2 1 j 0

v = (a, b) bj

ck y 1 2 x
(c)
3

y ai 0 bj

x i 1
(a)
2

x 0 ai
(b) v = a i + b j

i 0 j 1 x 2

2

(d) v = a i + b j + c k

Figure 1.2.7

Basis vectors in different dimensions

When a vector v = (a, b, c) is written as v = a i+b j+c k, we say that v is in component form, and that a, b, and c are the i, j, and k components, respectively, of v. We have: v = v1 i + v2 j + v3 k, k a scalar =⇒ kv = kv1 i + kv2 j + kv3 k v = v1 i + v2 j + v3 k, w = w1 i + w2 j + w3 k =⇒ v + w = (v1 + w1 )i + (v2 + w2 )j + (v3 + w3 )k v = v1 i + v2 j + v3 k =⇒ v = Example 1.4. Let v = (2, 1, −1) and w = (3, −4, 2) in v2 + v2 + v2 1 2 3
3.

(a) Find v − w. Solution: v − w = (2 − 3, 1 − (−4) − 1 − 2) = (−1, 5, −3) (b) Find 3v + 2w. Solution: 3v + 2w = (6, 3, −3) + (6, −8, 4) = (12, −5, 1) (c) Write v and w in component form. Solution: v = 2 i + j − k, w = 3 i − 4 j + 2 k (d) Find the vector u such that u + v = w. Solution: By Theorem 1.5, u = w−v = −(v−w) = −(−1, 5, −3) = (1, −5, 3), by part(a). (e) Find the vector u such that u + v + w = 0. Solution: By Theorem 1.5, u = −w − v = −(3, −4, 2) − (2, 1, −1) = (−5, 3, −1). (f) Find the vector u such that 2u + i − 2 j = k. 1 Solution: 2u = −i + 2 j + k =⇒ u = − 2 i + j + 1 k 2 (g) Find the unit vector Solution:
v v v v

. (2, 1, −1) =
1 √ 2 √ , √ , −1 6 6 6

= √

1 22 +12 +(−1)2

14

CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

We can now easily prove Theorem 1.1 from the previous section. The distance d between two points P = (x1 , y1 , z1 ) and Q = (x2 , y2 , z2 ) in 3 is the same as the length of the vector w − v, where the vectors v and w are defined as v = (x1 , y1 , z1 ) and w = (x2 , y2 , z2 ) (see Figure 1.2.8). So since w − v = (x2 − x1 , y2 − y1 , z2 − z1 ), then d = w − v = (x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 by Theorem 1.2.
z P(x1 , y1 , z1 ) w−v v w 0 x Q(x2 , y2 , z2 ) y

Figure 1.2.8

Proof of Theorem 1.2: d = w − v 

A

¨

1. Let v = (−1, 5, −2) and w = (3, 1, 1). (a) Find v − w. (b) Find v + w. (e) Find
1 2 (v

(c) Find

v v

.

(d) Find

1 2 (v

+ w) .

(f) Find −2 v + 4 w.

(g) Find v − 2 w.

− w) .

(h) Find the vector u such that u + v + w = i. (i) Find the vector u such that u + v + w = 2 j + k. (j) Is there a scalar m such that m(v + 2 w) = k? If so, find it. 2. For the vectors v and w from Exercise 1, is v − w = v − w ? If not, which quantity is larger? 3. For the vectors v and w from Exercise 1, is v + w = v + w ? If not, which quantity is larger?

B
4. Prove Theorem 1.5(f) for
3.

5. Prove Theorem 1.5(g) for

3.

C
6. We know that every vector in 3 can be written as a scalar combination of the vectors i, j, and k. Can every vector in 3 be written as a scalar combination of just i and j, i.e. for any vector v in 3 , are there scalars m, n such that v = m i + n j? Justify your answer.

Let v = (v1 .3 Dot Product You may have noticed that while we did define multiplication of a vector by a scalar in the previous section on vector algebra. the dot product is still v · w = v1 w1 + v2 w2 + v3 w3 . v. w3 ) be vectors in The dot product of v and w. which we will now develop as a consequence of the analytic definition.e.1. and so (u · v) · w is not defined since the left side of that dot product (the part in parentheses) is a scalar and not a vector.3. Definition 1. w2 . w.3.6. θ θ 360 − θ (a) 0◦ < θ < 180◦ ◦ 360◦ − θ θ 360◦ − θ (b) θ = 180◦ (c) θ = 0◦ Figure 1. v2 ) and w = (w1 . v2 . We will now see one type of multiplication of vectors. See Figure 1. the dot product u · v is a scalar.6) the dot product is: (1. does not hold for the dot product of vectors. There is a geometric way of defining the dot product.(e)).5(b). We will always choose the smallest nonnegative angle θ between them. is given by: v · w = v1 w1 + v2 w2 + v3 w3 Similarly. not a vector. w2 ) in v · w = v1 w1 + v2 w2 2. The angle between two nonzero vectors with the same initial point is the smallest angle between them. So the associative law that holds for multiplication of numbers and for addition of vectors (see Theorem 1.1. For vectors v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k in component form. so that 0◦ ≤ θ ≤ 180◦ . . by referencing vector coordinates. i. Also notice that we defined the dot product in an analytic way.3 Dot Product 15 1. v3 ) and w = (w1 . we did not define multiplication of a vector by a vector. Any two nonzero vectors with the same initial point have two angles between them: θ and 360◦ − θ.1 Angle between vectors We can now take a more geometric view of the dot product by establishing a relationship between the dot product of two vectors and the angle between them. (1. for vectors v = (v1 . called the dot product. Definition 1. denoted by v · w. 3. We do not define the angle between the zero vector and any other vector. Why? Because for vectors u.7.7) Notice that the dot product of two vectors is a scalar.

9) gives + w 2 −2 v w cos θ = (v1 − w1 )2 + (v2 − w2 )2 + (v3 − w3 )2 −2 v + w 2 − 2(v · w) . then cos θ = 1 1 v· w = √ √ = √ ≈ 0. QED cos θ = v w = v = (v2 + v2 + v2 ) + (w2 + w2 + w2 ) − 2(v1 w1 + v2 w2 + v3 w3 ) 1 2 3 1 2 3 2 = (v2 − 2v1 w1 + w2 ) + (v2 − 2v2 w2 + w2 ) + (v2 − 2v3 w3 + w2 ) 1 1 2 2 3 3 Example 1. 1). Then cos θ = v· w v w (1. Since cos 90◦ = 0. w be nonzero vectors. √ √ Solution: Since v · w = (2)(3) + (1)(−4) + (−1)(1) = 1. so since v 0 and w 0 then v· w .6.9) holds even for the “degenerate” cases θ = 0◦ and 180◦ ). w2 . Find the angle θ between the vectors v = (2. v = 6. We will write v ⊥ w to indicate that v and w are perpendicular. we have v−w 2 = v 2 + w 2 −2 v w cos θ (1. v3 − w3 ). v2 . −1) and w = (3. .41◦ v w 6 26 2 39 Two nonzero vectors are perpendicular if the angle between them is 90◦ . and let θ be the angle between them.3. VECTORS IN EUCLIDEAN SPACE Theorem 1.08 =⇒ θ = 85. z v θ 0 x Figure 1.2 w v−w y Since v − w = (v1 − w1 .5.2).16 CHAPTER 1. 1.8) Proof: We will prove the theorem for vectors in 3 (the proof for 2 is similar). v3 ) and w = (w1 . v2 − w2 . Two nonzero vectors v and w are perpendicular if and only if v· w = 0.3.9) (note that equation (1. −4.6: · Corollary 1. Let v = (v1 .7. since v > 0 and w > 0. Let v. w3 ). so w cos θ = −2(v · w) . and w = 26. By the Law of Cosines (see Figure 1. expanding v − w v 2 2 in equation (1. we have the following important corollary to Theorem 1.

.8. 1) perpendicular? Solution: Yes. and are left to the reader as exercises. so |v · w| = |cos θ| v w . See Figure 1.3. or zero. The following theorem summarizes the basic properties of the dot product. and scalar k.3 Dot Product Since cos θ > 0 for 0◦ ≤ θ < 90◦ and cos θ < 0 for 90◦ < θ ≤ 180◦ . v. then  > 0 for 0◦ ≤ θ < 90◦      v · w is 0 for θ = 90◦     < 0 for 90◦ < θ ≤ 180◦ 17 By Corollary 1. negative. 1. Theorem 1.1.8. So assume that v and w are nonzero vectors. the dot product can be thought of as a way of telling if the angle between two vectors is acute. (f) If either v = 0 or w = 0. v ⊥ w since v · w = (−1)(3) + (5)(1) + (−2)(1) = 0. then v · w = 0 by part (c). v · w = cos θ v w . respectively. obtuse.6. we also have: Corollary 1. −2) and w = (3.6. we have (a) v · w = w · v Commutative Law (b) (kv) · w = v · (kw) = k(v · w) Associative Law (c) v · 0 = 0 = 0 · v (d) u · (v + w) = u · v + u · w Distributive Law Distributive Law Cauchy-Schwarz Inequality5 (e) (u + v) · w = u · w + v · w (f) |v · w| ≤ v w Proof: The proofs of parts (a)-(e) are straightforward applications of the definition of the dot product. or a right angle. and so the inequality holds trivially. We will prove part (f). so |v · w| ≤ v w since |cos θ| ≤ 1. For any vectors u. w 0◦ ≤ θ < 90◦ v (a) v · w > 0 w 90◦ < θ ≤ 180◦ v (b) v · w < 0 w θ = 90◦ v (c) v · w = 0 Figure 1. w.3 Sign of the dot product & angle between vectors Example 1. Then by Theorem 1.3. 5. If θ is the angle between nonzero vectors v and w.3. 5 QED Also known as the Cauchy-Schwarz-Buniakovski Inequality. depending on whether the dot product is positive.9. Are the vectors v = (−1.

l. Calculate v · w.4 A Exercises ©  1. 1. l.3. For vectors v and w. we have (a) v 2 = v · v (b) v + w ≤ v + w Triangle Inequality (c) v − w ≥ v − w Proof: (a) Left as an exercise for the reader. QED The Triangle Inequality gets its name from the fact that in any triangle.”  ¨ v+w w v Figure 1. 2. Thus. For Exercises 3-8. 3). find the angle θ between the vectors v and w. then their span is a plane. if they are not parallel.4). (c) Since v = w + (v − w). . so by Theorem 1. VECTORS IN EUCLIDEAN SPACE Using Theorem 1. so since a ≤ |a| for any real number a. as given in the following theorem: Theorem 1. the collection of all scalar combinations kv + lw is called the span of v and w.10. −4. then v = w + (v − w) ≤ w + v − w by the Triangle Inequality.9. we have v+w 2 = v ≤ v ≤ v = (v + w) · (v + w) = v · v + v · w + w · v + w · w 2 2 2 + 2 |v · w| + w +2 v + 2(v · w) + w 2 2 . no one side is longer than the sum of the lengths of the other two sides (see Figure 1. (b) By part (a) and Theorem 1.18 CHAPTER 1. then their span is a line. we see that if u · v = 0 and u · w = 0. Calculate v · w. Another way of saying this is with the familiar statement “the shortest distance between two points is a straight line.9(f) we have 2 w + w = ( v + w )2 and so v + w ≤ v + w after taking square roots of both sides. For any vectors v. then u ⊥ (kv + lw) for all scalars k. we have the following fact: If u ⊥ v and u ⊥ w. so subtracting w from both sides gives v − w ≤ v − w . −2) and w = (4. w. So what we showed above is that a vector which is perpendicular to two other vectors is also perpendicular to their span. The dot product can be used to derive properties of the magnitudes of vectors. we have . Let v = (5.3.9. then u · (kv + lw) = k(u · v) + l(u · w) = k(0) + l(0) = 0 for all scalars k. Let v = −3 i − 2 j − k and w = 6 i + 4 j + 2 k. If nonzero vectors v and w are parallel. which proves (b). the most important of which is the Triangle Inequality.

6. C 21. for Exercises 15-25. v = i.5). 0. w = 3 i + 2 j + 4k 19 9. v = (7. 2. 4. w from Exercise 5. w from Exercise 6.3. 11. Prove or give a counterexample: If v · w = 0 for all v. j. −10). Prove Theorem 1. 4) 6. v = − i + 2 j + k. For nonzero vectors v and w.9(d). v = (5. 19. Prove that v − w ≤ v − w for all v. β.9(b). Prove Theorem 1. respectively. (Note: α.3 Dot Product 3. Is v ⊥ w? Justify your answer. 16. 10. then w = 0. 23. 3) 5. 2. −2). B Note: Consider only vectors in 15.5 w 26. γ are often called the direction angles of v. 2. 14.) . −1). 0) 7. Show that cos2 α + cos2 β + cos2 γ = 1. 18. −4. the projection of v onto w (sometimes written as pro jw v) is the vector u along the same line L as w whose terminal point is obtained by dropping a perpendicular line from the terminal point of v to L (see Figure 1.9(a). cos β.) v L u Figure 1.10(a).9(c). cos γ are called the direction cosines. w. w (Hint: Consider the angle between v and w. w from Exercise 6. 4). Is v ⊥ w? Justify your answer. 1. 20. verify the Triangle Inequality v + w ≤ v + w . β. For v. verify the Cauchy-Schwarz Inequality |v · w| ≤ v 13. Show that |v · w| u = . −2. and γ be the angles between a nonzero vector v in 3 and the vectors i. 4). and k.9(e). 17.3. verify the Triangle Inequality v + w ≤ v + w . verify the Cauchy-Schwarz Inequality |v · w| ≤ v 12. −2) 8. Prove or give a counterexample: If u · v = u · w. w = (2. and cos α. w . v = (2. Let v = (8. 24. then v = w. w = −3 i + 6 j + 3 k 4. w = (1. For v. Let α. w from Exercise 5. 22. 25. Prove Theorem 1. w = (8. For v. Let v = (6. 1. 3) and w = (−2. 4. v = (4. 3 w . w = (4.1. 4) and w = (0. Prove Theorem 1. Prove Theorem 1. 1. then v = w. For v. Prove or give a counterexample: If u · v = u · w for all u. Prove Theorem 1. −1).

v3 w1 − v1 w3 . Let v = (v1 .11.1 In the above example. then it is perpendicular to the span of v and w.10) Example 1.4. 1) =k Similarly it can be shown that j × k = i and k × i = j. (1)(1) − (0)(0)) = (0. v1 w2 − v2 w1 ) · (v1 . not a vector. was a scalar.20 CHAPTER 1. denoted by v × w.4 Cross Product In Section 1. v2 .12. is only defined for vectors in 3 .8. v2 . Find i × j. however. 0. Definition 1.7. Solution: Since i = (1.9. In this section we will define a product of two vectors that does result in another vector. If the cross product v × w of two nonzero vectors v and w is also a nonzero vector. which gave a way of multiplying two vectors. Proof: We will show that (v × w) · v = 0: (v × w) · v = (v2 w3 − v3 w2 . (0)(0) − (1)(0). The definition may appear strange and lacking motivation. The resulting product. v3 ) and w = (w1 .7. w3 ) be vectors in product of v and w. It turns out that this will always be the case. The proof that v × w ⊥ w is similar. called the cross product. then i × j = ((0)(0) − (0)(1). 0) and j = (0. then it is perpendicular to both v and w. The cross (1. = v1 v2 w3 − v1 v2 w3 + w1 v2 v3 − w1 v2 v3 + v1 w2 v3 − v1 w2 v3 ∴ v × w ⊥ v by Corollary 1.3 we defined the dot product. w2 . This product. VECTORS IN EUCLIDEAN SPACE 1. v3 w1 − v1 w3 . v3 ) = v2 w3 v1 − v3 w2 v1 + v3 w1 v2 − v1 w3 v2 + v1 w2 v3 − v2 w1 v3 = 0 . 1. 0. after rearranging the terms. x 1 k = i× j 1 y i 0 j 1 Figure 1. the cross product of the given vectors was perpendicular to both those vectors. Theorem 1. If the cross product v × w of two nonzero vectors v and w is also a nonzero vector. 0). . QED As a consequence of the above theorem and Theorem 1. v1 w2 − v2 w1 ) z 3. is the vector in 3 given by: v × w = (v2 w3 − v3 w2 . but we will see the geometric basis for it shortly. we have the following: Corollary 1.

4 Cross Product 21 The span of any two nonzero. so the above corollary shows that v × w is perpendicular to that plane. since v > 0 and w > 0. for nonzero vectors v. w in 3 is a plane P. w: v× w 2 = (v2 w3 − v3 w2 )2 + (v3 w1 − v1 w3 )2 + (v1 w2 − v2 w1 )2 = v2 (w2 + w2 ) + v2 (w2 + w2 ) + v2 (w2 + w2 ) − 2(v1 w1 v2 w2 + v1 w1 v3 w3 + v2 w2 v3 w3 ) 1 2 3 2 1 3 3 1 2 and now adding and subtracting v2 w2 . w.2 We will now derive a formula for the magnitude of v × w.4. so 2 − (v · w)2 1− 2 sin2 θ . there are two possible directions for v× w. As shown in Figure × 1.6 v 2 w 2 w 2 (1 − cos2 θ) . so by Theorem 1. and v2 w2 on the right side gives 1 1 2 2 3 3 2 = v2 (w2 + w2 + w2 ) + v2 (w2 + w2 + w2 ) + v3 (w2 + w2 + w2 ) 2 3 1 2 3 2 1 2 3 1 1 2 − (v1 w2 + v2 w2 + v2 w2 + 2(v1 w1 v2 w2 + v1 w1 v3 w3 + v2 w2 v3 w3 )) 3 3 1 2 2 2 = v2 w2 − 2v2 w2 v3 w3 + v2 w2 + v2 w1 − 2v1 w1 v3 w3 + v2 w2 + v2 w2 − 2v1 w1 v2 w2 + v2 w2 2 3 3 2 3 1 3 1 2 2 1 2 = (v1 + v2 + v2 )(w2 + w2 + w2 ) 2 3 1 2 3 − ((v1 w1 )2 + (v2 w2 )2 + (v3 w3 )2 + 2(v1 w1 )(v2 w2 ) + 2(v1 w1 )(v3 w3 ) + 2(v2 w2 )(v3 w3 )) so using (a + b + c)2 = a2 + b2 + c2 + 2ab + 2ac + 2bc for the subtracted term gives 2 = (v1 + v2 + v2 )(w2 + w2 + w2 ) − (v1 w1 + v2 w2 + v3 w3 )2 2 3 1 2 3 2 = v w 2 = v = v v× w 2 2 2 2 w = v w (v · w)2 . then sin θ ≥ 0. and since 0◦ ≤ θ ≤ 180◦ .1. that is.1 that this means that you can point your thumb upwards in the direction of v × w while rotating v towards w with the remaining four fingers. It turns out (see Appendix B) that the direction of v × w is given by the right-hand rule. so we have: .4. v × w form a right-handed system. v2 w2 . nonparallel vectors v. where θ is the angle between v and w.2. the vectors v. z v θ w x 0 P v× w y −v × w Direction of v × w Figure 1. one the opposite of the other. Recall from Section 1.

as shown in Figure 1. So we see that b= v and h = w sin θ 1 v w sin θ 2 1 v× w = 2 of the parallelogram PQRS is twice the area of the triangle APQR = APQRS = v w sin θ So since the area APQRS △PQR. respectively.22 CHAPTER 1.4. in 3 . The area APQR of 1 △PQR is 2 bh.3.8.11) It may seem strange to bother with the above formula. Example 1. when the magnitude of the cross product can be calculated directly. and identify the sides QR and QP with vectors v and w.13. where b is the base of the triangle and h is the height.4. then (1. The formula is more useful for its applications in geometry. Area of triangles and parallelograms (a) The area A of a triangle with adjacent sides v. as in the following example.3 Q S w θ v R P h S Think of the triangle as existing in 3 . Let θ be the angle between v and w.8. w (as vectors in A= 1 v× w 2 3) 3) is: (b) The area A of a parallelogram with adjacent sides v. Let △PQR and PQRS be a triangle and parallelogram. then By the discussion in Example 1. P h θ Q b R Figure 1. VECTORS IN EUCLIDEAN SPACE If θ is the angle between nonzero vectors v and w in v× w = v w sin θ 3. respectively. w (as vectors in A = v× w is: . we have proved the following theorem: Theorem 1. like for any other vector.

0) × (1. 8. as in Figure 1. 12. 18) − (2. 0) = (0. −7) Figure 1. 2) = (1. 12.9.4. −7) = (−7. 4) − (4.13 makes it simpler to calculate the area of a triangle in 3-dimensional space than by using traditional geometric methods. 8.13 that the formulas hold for any adjacent sides are not justified. 2 S So we can write v = (−3. 3. 3. Theorem 1. 2. as in Figure 1. 0). 0) and w = (1. 4.4. (−3)(2) − (−1)(1)) 2 3 4 Figure 1. 8) − (2. We would get a different formula for the area if we had picked PQ and PR as the adjacent sides. Calculate the area of the parallelogram PQRS . 18) Then v = (3.4. −1.1. where P = (2.13 is valid. Q = (2. and Theorem 1. 7. and S = (4. 25) and w = (−5.8 were for the adjacent sides QP and QR only. 4. 7.4. Then R v = (1. 0. − − → − − → z Solution: Let v = PQ and w = PR. R = (5. 12. 8). Then the v 1 P area A of PQRS is x A = v × w = (−3. 4). 15) 2 2 1 ((3)(15) − (25)(8). (25)(−7) − (1)(15).5. −7). Example 1. − − → − − → y Solution: Let v = S P and w = S R. −1. −7) = (1. 18). 4. Q = (3. 3). 2 can be thought of as the subset of 3 such that the z-coordinate is always 0. 4 2 . 15). 2) = (−3. Calculate the area of the triangle △PQR.46 A= v 0 x P(2.4 Cross Product 23 It may seem at first glance that since the formulas derived in Example 1. −190. but it can be shown (see Exercise 26) that the different formulas would yield the same value. and R = (−5. 7.10. 8) 1 1 v× w = (1. then the more general statements in Theorem 1. 25) × (−7. and the cross product is only But these are vectors in Q w 3 defined for vectors in 3 . 2). (1)(8) − (3)(−7)) = 2 1 = (−155.4 w y Example 1. 2. −5) 0 1 = ((−1)(0) − (0)(2). 4.4. However. (0)(1) − (−3)(0). where P = (1. −1) and w = (5. 1) − (4. 2). so the area A of the triangle △PQR is R(−5. 1). Q(3. so the choice of adjacent sides indeed does not matter. 29) 2 1 √ 1 (−155)2 + (−190)2 + 292 = 60966 = 2 2 A ≈ 123.5 5 A=5 .

v1 w2 − v2 w1 ) z v v× w y w x 0 = −(v3 w2 − v2 w3 .14. So since 0◦ ≤ θ ≤ 180◦ . i.6). (a) By the definition of the cross product and scalar multiplication. w3 v1 − w1 v3 . v1 w3 − v3 w1 . Adding to Example 1. we have (a) v × w = −w × v Anticommutative Law (b) u × (v + w) = u × v + u × w Distributive Law (c) (u + v) × w = u × w + v × w Distributive Law (d) (kv) × w = v × (kw) = k(v × w) (e) v × 0 = 0 = 0 × v (f) v × v = 0 (g) v × w = 0 if and only if v w Associative Law Proof: The proofs of properties (b)-(f) are straightforward. Figure 1. v3 in 3 .e. We will prove parts (a) and (g) and leave the rest to the reader as exercises. For any vectors u.7. v. If both v and w are nonzero. so v and w are scalar multiples. and either v = 0 = 0w or w = 0 = 0v.6 × (g) If either v or w is 0 then v× w = 0 by part (e). and θ is the angle between them.11.4. But the angle between v and w is 0◦ or 180◦ if and only if v w.4. w in 3 . w1 v2 − w2 v1 ) = −w × v w× v Note that this says that v × w and w × v have the same magnitude but opposite direction (see Figure 1. we have: v × w = (v2 w3 − v3 w2 . k3 ≤ 1. which is true if and only if sin θ = 0 (since v > 0 and w > 0). Theorem 1. we have i× j = k j × i = −k j× k = i k × j = −i k× i = j i × k = −j i× i = j× j = k× k = 0 Recall from geometry that a parallelepiped is a 3-dimensional solid with 6 faces. . v2 .24 CHAPTER 1. QED Example 1. v2 w1 − v1 w2 ) = −(w2 v3 − w3 v2 . they are parallel. VECTORS IN EUCLIDEAN SPACE The following theorem summarizes the basic properties of the cross product. v × w = 0 if and only if v w sin θ = 0. all of which are parallelograms. then sin θ = 0 if and only if θ = 0◦ or 180◦ . then by formula (1. k2 . v3 w1 − v1 w3 . and scalar k.6 6 An equivalent definition of a parallelepiped is: the collection of all scalar combinations k1 v1 +k2 v2 +k3 v3 of some vectors v1 .11). where 0 ≤ k1 .

The proof of the following theorem is left as an exercise for the reader: .12 the height h of the parallelepiped is u cos θ. w in 3 represent adjacent sides of a parallelepiped P. where θ is the angle between u and v × w. w in 3 . By Theorem 1.15. And we can see that since v× w is θ perpendicular to the base parallelogram dew termined by v and w.13(b). the area A of the base parallelogram × × is v× w . If vectors u.1. u · (v × w) = w · (u × v) = v · (w × u) (Note that the equalities hold trivially if any of the vectors are 0. then the volume of the parallelepiped is |u · (v × w)|. Another type of triple product is the vector triple product u × (v × w). the volume is w · (u × v). So taking the absolute value of the scalar triple product for any order of the three adjacent sides will always give the volume: Theorem 1. Show that the volume of P is the scalar triple product u · (v × w). as in Figure 1. then picking the wrong order for the three adjacent sides in the scalar triple product in formula (1. because the vector u is on the same side of the base parallelogram’s plane as the vector × v× w (so that cos θ > 0). we have the following result: For any vectors u.12) will give you the negative of the volume of the parallelepiped.4.6 we know that u · (v × w) . Hence.7. w in 3 represent any three adjacent sides of a parallelepiped. Volume of a parallelepiped: Let the vectors u.12.4 Cross Product 25 Example 1. Solution: Recall that the volume of a par. v. u v× w vol(P) = A h u u · (v × w) = v× w u v× w = u · (v × w) cos θ = h v Parallelepiped P Figure 1.12) Since v × w = −w × v for any vectors v.v × w allelepiped is the area A of the base parallelogram times the height h. w in 3. By Theorem u 1. and not − u cos θ. v. then the height h is u cos θ. Since the volume is the same no matter which base and height we use.7 In Example 1. Repeating this with the base determined by w and u. v.) (1. then repeating the same steps using the base determined by u and v (since w is on the same side of that base’s plane as u × v).4.

because it can be represented as a determinant. For any vectors u. Find u × (v × w) for u = (1. 2.4. We will not go too deeply into the theory of determinants7 . 18.e. we see that u × (v × w) is a scalar combination of v and w. which could be any vector? The following example may help to see how this works. u × (v × w) is perpendicular to both u and v × w = (0. v.8). v× w z u y 0 u × (v × w) v w x Figure 1. 0. w = (1. 14. It is often easier to use the component form for the cross product.26 CHAPTER 1. 0) = 7 (2. u × (v × w) is perpendicular to both u and v × w. −4. By the right side of formula (1. then u × (v × w) = (u · w)v − (u · v)w = (8. 3.13. we will just cover what is essential for our purposes. 0) − 6 (1.4. by Theorem 1. the cross product is written as: v × w = (v2 w3 − v3 w2 )i + (v3 w1 − v1 w3 )j + (v1 w2 − v2 w1 )k. 7 See A NTON and R ORRES for a fuller development. 0) − (6. and that u × (v × w) also lies in that plane. In particular. Example 1. Also. VECTORS IN EUCLIDEAN SPACE 3. Theorem 1. 0) Note that v and w lie in the xy-plane. . and hence lies in the plane containing v and w (i. v and w are coplanar).13) An examination of the formula in Theorem 1. 2. 0) = (14. 0).16. 4). 4) (see Figure 1.8 For vectors v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k in component form.13).16 gives some idea of the geometry of the vector triple product. But then how is u × (v × w) also perpendicular to u. 0). w in u × (v × w) = (u · w)v − (u · v)w (1. v = (2.11. since that plane is itself perpendicular to v × w. Solution: Since u · v = 6 and u · w = 7. This makes sense since. 3. being perpendicular to v × w means that u × (v × w) lies in the plane containing v and w. u × (v × w). 2.

c d is the scalar defined by the following formula: a b = ad − bc c d It may help to remember this formula as being the product of the scalars on the downward diagonal minus the product of the scalars on the upward diagonal. written as a b c d or a b c d 27 where a.14) and its determinant is given by the formula: One way to remember the above formula is the following: multiply each scalar in the first row by the determinant of the 2 × 2 matrix that remains after removing the row and column that contain that scalar.4 Cross Product A 2 × 2 matrix is an array of two rows and two columns of scalars.1. Example 1. b. Example 1. d are scalars. 1 2 = (1)(4) − (2)(3) = 4 − 6 = −2 3 4 A 3 × 3 matrix is an array of three rows and three columns of scalars. then sum those products up. written as a b c d or det a b .15. written as     a1 a2 a3  a1 a2 a3                  b1 b2 b3  or b1 b2 b3  . The determinant of such a matrix.14. putting alternating plus and minus signs in front of each (starting with a plus). 1 0 2 −1 3 4 −1 3 = 1 0 2 1 0 2 − 0 4 3 1 2 + 2 4 −1 = 1(−2 − 0) − 0(8 − 3) + 2(0 + 1) = 0 1 0 . c.             c1 c2 c3 c1 c2 c3 a1 a2 a3 b b b b b b b1 b2 b3 = a1 2 3 − a2 1 3 + a3 1 2 c1 c2 c1 c3 c2 c3 c1 c2 c3 (1.

w2 . By Theorem 1. = 2(−8) − 1(0) + 3(−4) = −28.15) Example 1.16. 1. Let v = 4 i − j + 3 k and w = i + 2 k. 1. 2 1 3 2 u · (v × w) = −1 3 1 1 −2 3 2 =2 1 −2 −1 2 − 1 1 −2 −1 3 + 3 1 1 u 0 x w Figure 1. For any vectors u = (u1 .4.17. The proof is left as an exercise for the reader. Theorem 1. 3. v3 ). by Example 1. since we would be performing scalar multiplication on those three vectors (they would be multiplied by the 2 × 2 scalar determinants as before). v2 . w3 ) in u1 u2 u3 u · (v × w) = v1 v2 v3 w1 w2 w3 3: (1.9 P z v y vol(P) = |−28| = 28. −2) (see Figure 1. 3). 2). This gives us a determinant that is now a vector. u2 . However.17.4. u3 ). v = (v1 .12. derived from algebraic operations on scalar entries in a matrix. if we put three vectors in the first row of a 3 × 3 matrix. so . Find the volume of the parallelepiped with adjacent sides u = (2. v = (−1.9). VECTORS IN EUCLIDEAN SPACE We defined the determinant as a scalar. w = (w1 . Solution: By Theorem 1. then the definition still makes sense. w = (1. the following theorem provides an alternate definition of the determinant of a 3 × 3 matrix as the volume (or negative volume) of a parallelepiped whose adjacent sides are the rows of the matrix.17. Then i j k −1 3 × w = 4 −1 3 = v i − 0 2 1 0 2 4 3 j + 1 2 4 −1 k = −2 i − 5 j + k 1 0 The scalar triple product can also be written as a determinant. the volume of the parallelepiped P is the absolute value of the scalar triple product of the three adjacent sides (in any order).28 CHAPTER 1. In fact. and lets us write the cross product of v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k as a determinant: i j k v v3 v v3 v v2 × w = v1 v2 v3 = 2 i − 1 v j + 1 k w2 w3 w1 w3 w1 w2 w1 w2 w3 = (v2 w3 − v3 w2 )i + (v3 w1 − v1 w3 )j + (v1 w2 − v2 w1 )k Example 1.15.

2). 0) Exercises ©  ¨ 2. 2). 11.4 Cross Product 29 Interchanging the dot and cross products can be useful in proving vector identities: Example 1.16) = (z · v)(w · u) − (z · u)(w · v) = u·w u·z v· w v· z = (u · w)(v · z) − (u · z)(v · w) (by commutativity of the dot product).12)) = w · (z × (u × v)) = w · ((z · v)u − (z · u)v) (by Theorem 1. 1). P = (2. 1. w. 6). v = (2. −2). 9. 0. 0. 3). 1) For Exercises 13-14. . Q = (1. −1) For Exercises 9-10. v. (u × v) · (w × z) = x · (w × z) = w · (z × x) (by formula (1. calculate u · (v × w) and u × (v × w). 3. w = (4. 7.  A For Exercises 1-6. 1. −2). 2. w. find the volume of the parallelepiped with adjacent sides u. Prove: (u × v) · (w × z) = Solution: Let x = u × v. 5. v = (1. z = (2. 1). P = (4. w = (5. 3. v = (−1. v = (3. R = (2. w = (2. v = (5. −4. v = − i + 2 j + k. v = (3. 2). 1. w = 3 i + 2 j + 4k 5. 2). v = (7. −2) 15. u = (1. u = (1. w = (2. 3). 0. −2. u = (1. −10). 4. 0. 2. v = i. 1) 10. w = (2. 2) 14. 0. 2. 1. 3) 3. 1. −10). w = (2.18. P = (−2. z in v·w v·z 3. 0) 8. −2). 4). w = (1.1. −2) 12. 4). R = (2. 4) 4. 5). 2. S = (3. 1. Calculate (u × v) · (w × z) for u = (1. 2). 1. 4. v = (7. v. w = (1. −10) 6. Q = (2. 2. 2). 4). 3). w = −3 i + 6 j + 3 k For Exercises 7-8. Then u· w u· z for all vectors u. w = (7. S = (3. P = (5. 1. v = (2. 1. calculate the area of the parallelogram PQRS . 3). 2). 0) For Exercises 11-12. 5). 13. −4. 4). 6. 3). R = (6. calculate v × w. 0. 1. 2. Q = (1. u = (1. R = (−1. Q = (4. 1. calculate the area of the triangle △PQR. 0. 1. 0.

Show that u. VECTORS IN EUCLIDEAN SPACE B 16. w lie in the same plane in 30.16. 20.14(e). Prove Theorem 1. show that 1 1 2 u × (−w) = 2 v × w . −w = PQ. 27. 3. 1 1 Similarly. then v = 0 or w = 0.8 the formula for the area of the triangle △PQR yields the same value no matter which two adjacent sides are chosen. 19. and v = QR. Prove the Jacobi identity: u × (v × w) + v × (w × u) + w × (u × v) = 0 29. then v = 0. v. Prove that in Example 1. + |v · w|2 = v 2 w 2 C 26. Prove Theorem 1. for any scalar k a 2 28. 22. show that (u × v) × (w × z) = (z · (u × v))w − (w · (u × v))z and that (u × v) × (w × z) = (u · (w × z))v − (v · (w × z))u Why do both equations make sense geometrically? . w in (a) v × w 2 3: (b) If v · w = 0 and v × w = 0. where u = PR. show that 2 (−u) × (−v) = 2 v × w . w. 3 if and only if u · (v × w) = 0. Prove the following for all vectors v. 17. v. If v and w are unit vectors in 3 . under what condition(s) would v × w also be a unit vector in 3 ? Justify your answer.) 25. Prove Theorem 1.17. To do this. where a 0.14(c). Prove Theorem 1. where −u = RP and −v = RQ. Prove Theorem 1. (Hint: Expand both sides of the equation. Show that if v × w = 0 for all w in 18. 21. Prove Theorem 1.14(d). Show that: (a) a · b = 0 b× a (b) x = + ka is a solution to the equation.30 CHAPTER 1. w = QP as before. 24. z in 3.14(f). Consider the vector equation a × x = b in 3. 23. For all vectors u. Prove Theorem 1.14(b).

3. z0 ) be the vector pointing from the origin to P. y0 . y = y0 + bt. z0 ) and nonzero vector v = (a. We then get the parametric representation of L with the parameter t: For a point P = (x0 . . and reversing its direction if t < 0. z0 ) be a point in 3 . z0 ) and nonzero vector v in 3 . Note that we used the correspondence between a vector and its terminal point. b. c) be a nonzero vector. y0 .1 0 v tv r + tv t>0 y L Let r = (x0 . z0 ) r r + tv t<0 x Figure 1. then we see from Figure 1. for − ∞ < t < ∞ (1.5 Lines and Planes Now that we know how to perform some operations on vectors. let v = (a.1). y0 . and let L be the line through P which is parallel to v (see Figure 1. the line L through P parallel to v is given by r + tv. z0 + ct).1 that every point on the line L can be obtained by adding the vector tv to the vector r for some scalar t.1. z) given by x = x0 + at. y.5. y0 . then the terminal point of the vector r + tv is (x0 + at.5. z P(x0 . y0 + bt. as t varies over all real numbers. the line L through P z = z0 + ct. That is. like lines and planes.5. We can summarize the vector representation of L as follows: For a point P = (x0 .16) where r = (x0 . c) in parallel to v consists of all points (x. b. for − ∞ < t < ∞ (1. Line through a point. parallel to a vector Let P = (x0 . We will first consider lines. Since v = (a.17) Note that in both representations we get the point P on L by letting t = 0. b. the vector r + tv will point to every point on L. The reason for doing this is simple: using vectors makes it easier to study objects in 3-dimensional Euclidean space. in the language of vectors. c). y0 . z0 ) is the vector pointing to P. Since multiplying the vector v by a scalar t lengthens or shrinks v while preserving its direction if t > 0. we can start to deal with some familiar geometric objects. y0 .5 Lines and Planes 31 1.

though. but we do know that x = x0 + at. for − ∞ < t < ∞ (c) L consists of the points (x. say. then we can solve for the parameter t: t = (x − x0 )/a. the parametric representation always gives just the points on L and nothing else. 3. Lastly: (d) find two points on L distinct from P. c) in 3 with a. b and c all nonzero.16). 5). 11) and (10. You may have noticed that the vector representation of L in formula (1. Example 1. z = 5 + 6t.19. Solution: (a) Let r = (2. the vector representation gives us the vectors whose terminal points make up the line L. which x is parallel to the yz-plane (see Figure 1. y. Technically. 17) on L. for − ∞ < t < ∞ (b) L consists of the points (x.16) is more compact than the parametric and symmetric formulas. L is given by: r + tv = (2. y − y0 z − z0 = b c (1. z) given by the equations x − x0 y − y0 z − z0 = = a b c What if. 3. Similar equations Figure 1. −1. Then the symmetric representation of L would be: x = x0 . z) such that x−2 y−3 z−5 = = 4 −1 6 (d) Letting t = 1 and t = 2 in part(b) yields the points (6.2 can be derived for the cases when b = 0 or c = 0. z) such that x = 2 + 4t.5. . not just L itself. and so x = x0 + 0t = x0 . VECTORS IN EUCLIDEAN SPACE In formula (1. called the symmetric representation of L: For a point P = (x0 . y.32 CHAPTER 1. 6). respectively. 3.17). 5) and parallel to the vector v = (4. That is an advantage of using vector notation. Write the line L through the point P = (2. 5) + t(4. y0 . y = 3 − t. On the other hand. Then by formula (1. the line L through P parallel to v consists of all points (x. (b) parametric. b. z0 ) and vector v = (a. 6). −1. So you have to remember to identify the vectors r + tv with their terminal points.18) Note that this says that the line L lies in the plane x = x0 . a = 0 in the above scenario? We can not divide by zero.19) x0 z L y 0 x = x0 (1. 1. if a 0. These three values all equal the same value t. y. in the following forms: (a) vector. so we can write the following system of equalities. (c) symmetric.2). is zero: t = (y − y0 )/b and t = (z − z0 )/c.5. 2. We can also solve for t in terms of y and in terms of z if neither b nor c.

Then as we can see from Figure 1. So if we multiply the vector r2 − r1 by a scalar t and add it to the vector r1 . y2 . then: d= v× w v P w θ Q v Figure 1. z = −4 − 2t. So since v × w = v w sin θ and v 0. r2 = (x2 . y. y2 . z2 ) be the vectors pointing to P1 and P2 . and symmetric forms for the line L: z P2 (x2 .21) (1. respectively. Pick a point Q on L.1.20. y2 . Solution: By formula (1. for − ∞ < t < ∞ (1.5. Then the line L through P1 and P2 has the following representations: Vector: r1 + t(r2 − r1 ) . y1 y2 .21).5 Lines and Planes 33 Line through two points Let P1 = (x1 . and let P be a point not on L. y1 . and z1 z2 ) (1. 1. The following is a summary of the vector.3. then d = w sin θ.20) Example 1. L consists of the points (x. r2 − r1 is the vector from P1 to P2 . we will get the entire line L as t varies over all real numbers. z) such that x = −3 + 7t. Symmetric: x − x1 y − y1 z − z1 = = x2 − x1 y2 − y1 z2 − z1 (if x1 x2 . z1 ) and P2 = (x2 . y1 . Write the line L through the points P1 = (−3. z2 ) P1 (x1 .3 r2 r1 + t(r2 − r1 ) y Let P1 = (x1 .23) . and let w be the vector from Q to P. z2 ). y1 .5. Let r1 = (x1 .5. parametric.4). z2 ) be distinct points in 3 . 4.22) y = y1 + (y2 − y1 )t. z1 ) and r2 = (x2 . for − ∞ < t < ∞ Distance between a point and a line Let L be a line in 3 in vector form as r + tv (for −∞ < t < ∞).5. y1 . z1 ) r2 − r1 L r1 0 x Figure 1. y2 . and let L be the line through P1 and P2 .4 d L (1. and let r1 = (x1 . z2 ) be distinct points in 3 . y2 . z1 ). y = 1 + 3t. z1 ). y1 . z = z1 + (z2 − z1 )t. for − ∞ < t < ∞ Parametric: x = x1 + (x2 − x1 )t. −4) and P2 = (4. If θ is the angle between w and v. −6) in parametric form. The distance d from P to L is the length of the line segment from P to L which is perpendicular to L (see Figure 1. P2 = (x2 .

y.34 CHAPTER 1. 1. we see that we can represent L in vector form as: r + tv. it is often easier Figure 1. there is an additional possibility: two lines can be skew. It is clear that two lines L1 and L2 . However.) 1 − s = −3 + 2t : 1 − 0 = −3 + 2(2) ⇒ 1 = 1 Letting s = 0 in the equations for the first line. 3. 1. In 3-dimensional space. y = 8 − 3t. even though they are not parallel. Also. 1).98 62 15 i − 43 j − 12 k v× w = = d= v (7. . they do not intersect but they L2 y are not parallel. 8 − 3t. or L1 they intersect. Setting the two (x. Find the point of intersection (if any) of the following lines: x+1 y−2 z−1 = = 3 2 −1 and x+3= y−8 z+3 = −3 2 Solution: First we write the lines in parametric form. so 4 0 √ 2218 = √ = 5. that is. skew 0 lines are on parallel planes (see Figure 1. L1 and L2 are perpendicular (denoted as L1 ⊥ L2 ) if v1 and v2 are perpendicular. 2. 1 − s) = (−3 + t. gives the point of intersection (−1. since the values of the parameters may not be the same at the point of intersection. In this case.21. 5). 1. Example 1. 1) − (−3. VECTORS IN EUCLIDEAN SPACE Example 1. parallel. 1. or letting t = 2 in the equations for the second line. we have: i j k 3 −2 v × w = 7 3 −2 = i − 0 5 4 0 5 7 −2 j + 4 5 7 3 k = 15 i − 43 j − 12 k .20. z = −3 + 2t The lines intersect when (−1 + 3s. 1) to the line L in Example 1. t: −1 + 3s = −3 + t : ⇒ t = 2 + 3s 2 + 2s = 8 − 3t : ⇒ 2 + 2s = 8 − 3(2 + 3s) = 2 − 9s ⇒ 2s = −9s ⇒ s = 0 ⇒ t = 2 + 3(0) = 2 (Note that we had to check this. −2). you should use different parameter variables (usually s and t) for the lines. then for − − → w = QP = (1. are parallel (denoted as L1 L2 ) if v1 and v2 are parallel.5.20. for r = (−3.5 to use the parametric representation of the lines. 1. −3 + 2t) for some s. 3. 0. Since the point Q = (−3. z = 1 − s and x = −3 + t. respectively.5. −4) is on L. −4) = (4. Solution: From Example 1. x To determine whether two lines in 3 intersect.5). z) triples equal will result in a system of 3 equations in 2 unknowns (s and t). 2 + 2s. −2) 152 + (−43)2 + (−12)2 72 + 32 + (−2)2 z In 2-dimensional space. represented in vector form as r1 + sv1 and r2 + tv2 .22. two lines are either identical. y = 2 + 2s. with parameters s and t: x = −1 + 3s. −4) and v = (7. Find the distance d from the point P = (1.

b. Example 1. y.5 Lines and Planes We will now consider planes in 3-dimensional Euclidean space. c) be a nonzero vector which is perpendicular to the plane P. y0 . And if r = 0 then we still have n · r = 0.6 The plane P Conversely. Now let (x. we get an equation of the plane in normal form: ax + by + cz + d = 0 (1. z0 ) Figure 1. . then r ⊥ n and hence n · r = 0. z) lies in P. z) such that: 2(x + 3) + 4(y − 1) + 8(z − 3) = 0 If we multiply out the terms in formula (1. n r (x. z) be any point in the plane P. or equivalently: a(x − x0 ) + b(y − y0 ) + c(z − z0 ) = 0 The above equation is called the point-normal form of the plane P. z) is any point in 3 such that r = (x − x0 . y − y0 . 1.25) and combine the constant terms. Then the vector r = (x − x0 . the plane P consists of all points (x. Solution: By formula (1. Find the equation of the plane P containing the point (−3.5. 4. 35 Plane through a point. Then P consists of the points (x.6). b. Let P be a plane in 3 . if (x. c) be a nonzero vector which is perpendicular to P. let (x0 . y − y0 . z0 ). z) satisfying the vector equation: n· r = 0 (1. Such a vector is called a normal vector (or just a normal) to the plane. z − z0 ) lies in the plane P (see Figure 1. z − z0 ). y0 .1. y. the normal form of the plane in Example 1.18. y. 3) and perpendicular to the vector n = (2. 8). y.25). z − z0 ) 0 and n · r = 0. So if r 0. z) (x0 .24) where r = (x − x0 . then r ⊥ n and so (x. and let n = (a.25) For example. y. This proves the following theorem: Theorem 1.26) (1.23 is 2x + 4y + 8z − 22 = 0. y0 . z0 ) be a point in P. y. and suppose it contains a point P0 = (x0 . Let n = (a.5. y − y0 .23. perpendicular to a vector Let P be a plane in 3 .

5. 1.36 CHAPTER 1. as in Example 1. But two (nonidentical) lines which either intersect or are parallel do determine a plane.24. 2) and (3. 3). − − → −→ − n = QR × QS − − → QR R S Q −→ − QS Figure 1. In fact. −1) and QS = (1. 2. R. to write the equation. So two skew lines do not determine a plane. Two points do not determine a plane in 3 .e. (1. 2. −3. 1). three collinear points (i. −2). For if Q. 1. 3).e. parallel planes. and so their cross product QR × QS − − → −→ − − − → −→ − is perpendicular to both QR and QS . In both cases. two points determine a line. then use the technique above. 2) and S = (3. Then for the vectors QR = −→ − (−1.25) with the point Q (we could also use R or S ). R and S are noncollinear points in 3 . one point from one line and two points from the other). −2. the plane P has a normal vector − − → −→ − n = QR × QS = (−1. S Example 1. all on the same line) do not determine a plane. However. 1. 1. 5x − 3y + z − 10 = 0 We mentioned earlier that skew lines in 3 lie on separate. VECTORS IN EUCLIDEAN SPACE Plane containing three noncollinear points In 2-dimensional and 3-dimensional space. simply pick from the two lines a total of three noncollinear points (i. −2. the plane P consists of all points (x.7 Noncollinear points Q. −1) × (1. then QR and QS are nonzero − − −→ → − vectors which are not parallel (by noncollinearity). z) such that: 5(x − 2) − 3(y − 1) + (z − 3) = 0 or in normal form.5. R and S ) lie in the − − → −→ − plane through the point Q with normal vector n = QR × QS (see Figure 1. So QR and QS (and hence Q. 1) So using formula (1. We will leave examples of this as exercises for the reader. an infinite number of planes would contain the line on which those three points lie. − − → Solution: Let Q = (2. three noncollinear points do determine a − − → −→ − plane. −1.7). 1). R = (1. −2) = (5. −1.24. . to find the equation of the plane that contains those two lines. Find the equation of the plane P containing the points (2. y.

b. Find the distance D from (2. Thus. y. Place n so that its initial point is at R. y0 − y.5.8). any plane divides 3 into two disjoint parts. z0 − z). The following theorem gives a formula for that distance. so n r D = |cos θ| r = |a(x0 − x) + b(y0 − y) + c(z0 − z)| √ n r n a2 + b2 + c2 |ax0 + by0 + cz0 − (ax + by + cz)| |ax0 + by0 + cz0 − (−d)| |ax0 + by0 + cz0 + d| = = = √ √ √ a2 + b2 + c2 a2 + b2 + c2 a2 + b2 + c2 r = = If n points away from the side of P where the point Q is located. Assume that n points toward the side of P where the point Q is located. the distance D is cos θ r = |cos θ| r (see Figure 1.25. Now. z) be any point in the plane P (so that ax + by + cz + d = 0) and − − → let r = RQ = (x0 − x. so cos θ > 0. and thus repeating the same argument as above still gives the same result. c) is a normal vector for P. So . Then 0◦ < θ < 90◦ . n Q r D θ R Figure 1. Let Q = (x0 . −5) to the plane from Example 1. Theorem 1.1.87 35 35 Solution: Recall that the plane is given by 5x − 3y + z − 10 = 0. Then r 0 since Q does not lie in P.5. QED Example 1.19. we know that n = (a. then 90◦ < θ < 180◦ and so cos θ < 0. z0 ) be a point in 3 .6 in Section 1. From the normal form equation for P.5 Lines and Planes 37 Distance between a point and a plane The distance between a point in 3 and a plane is the length of the line segment from that point to the plane which is perpendicular to the plane. 4. and let θ be the angle between r and n. and let P be a plane with normal form ax + by + cz + d = 0 that does not contain Q. D= |5(2) − 3(4) + 1(−5) − 10| 52 + (−3)2 + 12 |−17| 17 = √ = √ ≈ 2.3. we know that cos θ = n· r n· r n· r .8 D P By Theorem 1.27) Proof: Let R = (x. Then the distance D from Q to P is: D= |ax0 + by0 + cz0 + d| √ a2 + b2 + c2 (1.24. The distance D is then |cos θ| r . y0 .

31) is on L. Solution: The plane 5x − 3y + z − 10 = 0 has normal vector n1 = (5. for − ∞ < t < ∞ (1. VECTORS IN EUCLIDEAN SPACE Line of intersection of two planes Note that two planes are parallel if they have normal vectors that are parallel. respectively. −1). Figure 1. 7. This can often be made easier by setting one of the coordinate variables to zero. i. z) to the two normal form equations of the planes. 1) and the plane 2x + 4y − z + 3 = 0 has normal vector n2 = (2. Thus. z) on both planes will satisfy the following system of two equations in three unknowns: 5x − 3y + z − 10 = 0 2x + 4y − z + 3 = 0 Set x = 0 (why is that a good choice?). and the planes are perpendicular if their normal vectors are perpendicular. y = 7 + 7t. 26). 4. −3. then n1 × n2 is parallel to the plane P1 .9). To find a point in both planes. If two planes do intersect. for − ∞ < t < ∞ . Example 1.e.26. we can write L in the following vector form: L : r + t(n1 × n2 ) . A point (x.9 n1 × n2 ⊥ n2 means that n1 × n2 is also parallel to P2 . n1 × n2 is parallel to the intersection of P1 and P2 . y. and so the point (0. Likewise. Suppose that two planes P1 and P2 with normal vectors n1 and n2 . which leaves you to solve two equations in just two unknowns. Find the line of intersection L of the planes 5x − 3y + z − 10 = 0 and 2x + 4y − z + 3 = 0. Since n1 and n2 are not scalar multiples. for − ∞ < t < ∞ or in parametric form: x = −t. Then z = 31. Since n1 × n2 ⊥ n1 . they do so in L a line (see Figure 1. find a common solution (x.5. 26).28) where r is any vector pointing to a point belonging to both planes. Thus.38 CHAPTER 1. 7. then the two planes are not parallel and hence will intersect. 7. y. z = 31 + 26t. Then the above equations are reduced to: −3y + z − 10 = 0 4y − z + 3 = 0 The second equation gives z = 4y + 3. Since n1 × n2 = (−1. n1 × n2 is parallel to L.5. substituting that into the first equation gives y = 7. then L is given by: r + t(n1 × n2 ) = (0. 31) + t(−1. 7. intersect in a line L.

−1). Write the normal form of the plane containing the lines from Exercise 10. −2). −3). write the normal form of the plane containing the given points. y = 2 + t. P2 = (−2. (−3. (4. v = (2. −3) 3. 2x − y + z + 2 = 0 18. P : 3x − y − 5z + 8 = 0 19. 5). P = (1. 12. 1) 2. 2). P = (2. 6) For Exercises 13-14. 5) For Exercises 7-8. −1. 2).) . −2). 0. 17. P1 = (1. 0). P = (0. 0. and (c) symmetric. v = (1. Q = (4. (b) parametric. z = 7 + t 8. 2. 0. P = (2. Find the point(s) of intersection (if any) of the line x−6 = y + 3 = z with the plane 4 x + 3y + 2z − 6 = 0. y = 4 + 3t. 1. −3). write the line L through the point P and parallel to the vector v in the following forms: (a) vector. 3). 16. 15. v = (5. P1 = (4. 1) 4. Q = (6. P : −5x + 2y − 7z + 1 = 0 20. (6. 0. (0. −2. y = −4 − 3s. n = (4. 5. (1. Q = (5. 6. 1.1. 3). x + 3y + 2z − 6 = 0. 1. 4) 14. write the line L through the points P1 and P2 in parametric form. −10) 6. 1) 11. 1. v = (7. n = (2. x = 1 + 6t. z = 5 + 4t For Exercises 9-10. 3x + y − 5z = 0. 1. 9. 1. −1. find the line of intersection (if any) of the given planes. 7. P = (0. y = 4t. find the distance d from the point P to the line L. find the point of intersection (if any) of the given lines. For Exercises 17-18. 3). B 21. 0. (1. −4. 3) For Exercises 5-6. 5. 2. z = −7 − 5s and 10. −4. P = (3. 4. 2. Write the normal form of the plane containing the lines from Exercise 9. find the distance D from the point Q to the plane P. Q = (0. 3. z = 3 − 2t x−6 x − 11 y − 14 z + 9 = y + 3 = z and = = 4 3 −6 2 For Exercises 11-12. P2 = (3.5 Lines and Planes  39 A Exercises ©  ¨ For Exercises 1-4. 0). 1. 0). 0). L : x = 3 + 2t. 3) 13. x + 2y + z + 4 = 0 For Exercises 19-20. write the normal form of the plane P containing the point Q and perpendicular to the vector n. −1). L : x = −2 − 2t. 8. 1. (Hint: Put the equations of the line into the equation of the plane. −2. x = 7 + 3s.

for some real-valued function F.6.6. Definition 1.6.e. Figure 1.9. z z (1. z) in 3 which are a fixed distance r (called the radius) from a fixed point P0 = (x0 . center (0.1(a) that the intersection of the sphere with the xy-plane is a circle of radius r (i. Surfaces are 2-dimensional.30) x =r x y 0 x − x0 = r x x − x0 (x0 . a plane intersects a sphere either at a single point or in a circle.40 CHAPTER 1. y0 . which we will define informally8 as the solution set of the equation F(x.1 illustrates the vectorial approach to spheres. y0 . given by x2 + y2 = r2 as a subset of 2 ). z) = 0 in 3 . a great circle. y. z) : (x − x0 )2 + (y − y0 )2 + (z − z0 )2 = r2 } (1. 0) 0 x (b) radius r. y0 . A sphere S is the set of all points (x. this can be written in the equivalent form: S = { x : x − x0 = r } where x = (x. z0 ) x0 y x (a) radius r. For example. y. center (x0 . y. In general. z) = 0 for the function F(x. the most important of which are the sphere and the cylinder. . y. VECTORS IN EUCLIDEAN SPACE 1. y. 0. y. a plane given by ax+by+cz+d = 0 is the solution set of F(x. since it is “flat”. z0 ) Figure 1. In this section we will look at some surfaces that are more complex. z) = ax+by+cz+d. y0 .29) Using vector notation. 8 See O’N EILL for a deeper and more rigorous discussion of surfaces. Similarly for the intersections with the xz-plane and the yz-plane.6 Surfaces In the previous section we discussed planes in Euclidean space. z0 ) (called the center of the sphere): S = { (x.1 Spheres in 3 Note in Figure 1. z) and x0 = (x0 . A plane is an example of a surface. The plane is the simplest surface. z0 ) are vectors.

which can be determined by completing the square for the x. Is 2x2 + 2y2 + 2z2 − 8x + 4y − 16z + 10 = 0 the equation of a sphere? x2 + y2 + z2 − 4x + 2y − 8z + 5 = 0 (x − 2)2 + (y + 1)2 + (z − 4)2 = 16 Solution: Dividing both sides of the equation by 2 gives (x2 − 4x + 4) + (y2 + 2y + 1) + (z2 − 8z + 16) + 5 − 4 − 1 − 16 = 0 which is a sphere of radius 4 centered at (2.29) is multiplied out. and solve for t: (3 + t − 2)2 + (1 + 2t + 1)2 + (3 − t − 4)2 = 16 (t + 1)2 + (2t + 2)2 + (−t − 1)2 = 16 6t2 + 12t − 10 = 0 4 The quadratic formula gives the solutions t = −1 ± √ . y = 1 + 2t. Solution: The sphere is centered at the origin and has √ radius 13 = 169. Putting those two values into 6 the equations of the line gives the following two points of intersection: 4 4 8 2 + √ . Conversely. −1 − √ .31) for some constants a. 0. −1.1.28. y and z variables. Example 1. Example 1. x Figure 1. c and d. Putting z = 12 into the equation of the sphere gives x2 + y2 + 122 = 169 x2 + y2 = 169 − 144 = 25 = 52 which is a circle of radius 5 centered at (0.6. which was (x − 2)2 + (y + 1)2 + (z − 4)2 = 16. 12). Find the intersection of the sphere x2 + y2 + z2 = 169 with the plane z = 12. 4 + √ 6 6 6 . an equation of this form may describe a sphere. 4).6. Solution: Put the equations of the line into the equation of the sphere. so it does intersect the plane z = 12. b. we get an equation of the form: x2 + y2 + z2 + ax + by + cz + d = 0 (1.6 Surfaces 41 Example 1.28 and the line x = 3 + t. Find the points(s) of intersection (if any) of the sphere from Example 1. −1 + √ . z = 3 − t. 4 − √ 6 6 6 and 8 4 4 2 − √ .29.2 z z = 12 y 0 If the equation in formula (1.27. parallel to the xy-plane (see Figure 1.2).

42

CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

If two spheres intersect, they do so either at a single point or in a circle. Example 1.30. Find the intersection (if any) of the spheres x2 + y2 + z2 = 25 and x2 + y2 + (z − 2)2 = 16. x2 + y2 + z2 = 25 16 − (z − 2)2 = 25 − z2 x2 + y2 + (z − 2)2 = 16 ⇒ ⇒ ⇒ ⇒ ∴ The intersection is the circle x2 + y2 = x2 + y2 = 25 − z2 , and ⇒

Solution: For any point (x, y, z) on both spheres, we see that

x2 + y2 = 16 − (z − 2)2 , so 4z − 4 = 9
2 2 231 16

z = 13/4

x + y = 25 − (13/4)2 = 231/16 of radius
√ 231 4

≈ 3.8 centered at (0, 0, 13 ). 4

The cylinders that we will consider are right circular cylinders. These are cylinders obtained by moving a line L along a circle C in 3 in a way so that L is always perpendicular to the plane containing C. We will only consider the cases where the plane containing C is parallel to one of the three coordinate planes (see Figure 1.6.3).
z r z r y 0 x
(a) x2 + y2 = r2 , any z

z

y x

r 0

y

0 x
(b) x2 + z2 = r2 , any y
3

(c) y2 + z2 = r2 , any x

Figure 1.6.3 Cylinders in

For example, the equation of a cylinder whose base circle C lies in the xy-plane and is centered at (a, b, 0) and has radius r is (x − a)2 + (y − b)2 = r2 , (1.32)

where the value of the z coordinate is unrestricted. Similar equations can be written when the base circle lies in one of the other coordinate planes. A plane intersects a right circular cylinder in a circle, ellipse, or one or two lines, depending on whether that plane is parallel, oblique9 , or perpendicular, respectively, to the plane containing C. The intersection of a surface with a plane is called the trace of the surface.
9

i.e. at an angle strictly between 0◦ and 90◦ .

1.6 Surfaces

43

3,

The equations of spheres and cylinders are examples of second-degree equations in i.e. equations of the form Ax2 + By2 + Cz2 + Dxy + Exz + Fyz + Gx + Hy + Iz + J = 0 (1.33)

for some constants A, B, . . . , J. If the above equation is not that of a sphere, cylinder, plane, line or point, then the resulting surface is called a quadric surface. One type of quadric surface is the ellipsoid, given by an equation of the form: x2 y2 z2 + + =1 a2 b2 c2 (1.34) a In the case where a = b = c, this is just a sphere. In general, an ellipsoid is egg-shaped (think of an ellipse rotated around its major axis). Its traces in the coordinate planes are ellipses.
0

c

z

y

b

x Figure 1.6.4

Ellipsoid

Two other types of quadric surfaces are the hyperboloid of one sheet, given by an equation of the form: x2 y2 z2 + − =1 (1.35) a2 b2 c2 and the hyperboloid of two sheets, whose equation has the form: x2 y2 z2 − − =1 a2 b2 c2
z z

(1.36)

y 0 0

y

x Figure 1.6.5 Hyperboloid of one sheet

x Figure 1.6.6 Hyperboloid of two sheets

44

CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

For the hyperboloid of one sheet, the trace in any plane parallel to the xy-plane is an ellipse. The traces in the planes parallel to the xz- or yz-planes are hyperbolas (see Figure 1.6.5), except for the special cases x = ±a and y = ±b; in those planes the traces are pairs of intersecting lines (see Exercise 8). For the hyperboloid of two sheets, the trace in any plane parallel to the xy- or xzplane is a hyperbola (see Figure 1.6.6). There is no trace in the yz-plane. In any plane parallel to the yz-plane for which | x| > |a|, the trace is an ellipse. The elliptic paraboloid is another type of quadric surface, whose equation has the form: x2 y2 z + 2 = 2 c a b (1.37)
z

The traces in planes parallel to the xy-plane are ellipses, though in the xy-plane itself the trace is a single point. The y traces in planes parallel to the xz- or yz-planes are parabo0 las. Figure 1.6.7 shows the case where c > 0. When c < 0 the x surface is turned downward. In the case where a = b, the surface is called a paraboloid of revolution, which is often Figure 1.6.7 Paraboloid used as a reflecting surface, e.g. in vehicle headlights.10 A more complicated quadric surface is the hyperbolic paraboloid, given by: x2 y2 z − 2 = 2 c a b (1.38)

100 50 0 -10 z -50 -5 -100-10 -5 0 y 5 10 10 5 0 x

Figure 1.6.8
10

Hyperbolic paraboloid

For a discussion of this see pp. 157-158 in H ECHT.

1.6 Surfaces

45

The hyperbolic paraboloid can be tricky to draw; using graphing software on a computer can make it easier. For example, Figure 1.6.8 was created using the free Gnuplot package (see Appendix C). It shows the graph of the hyperbolic paraboloid z = y2 − x2 , which is the special case where a = b = 1 and c = −1 in equation (1.38). The mesh lines on the surface are the traces in planes parallel to the coordinate planes. So we see that the traces in planes parallel to the xz-plane are parabolas pointing upward, while the traces in planes parallel to the yz-plane are parabolas pointing downward. Also, notice that the traces in planes parallel to the xy-plane are hyperbolas, though in the xy-plane itself the trace is a pair of intersecting lines through the origin. This is true in general when c < 0 in equation (1.38). When c > 0, the surface would be similar to that in Figure 1.6.8, only rotated 90◦ around the z-axis and the nature of the traces in planes parallel to the xz- or yz-planes would be reversed. The last type of quadric surface that we will consider is the elliptic cone, which has an equation of the form: x2 y2 z2 + − =0 a2 b2 c2 (1.39)
z

y The traces in planes parallel to the xy-plane are ellipses, 0 except in the xy-plane itself where the trace is a single point. The traces in planes parallel to the xz- or yz-planes are hyperbolas, except in the xz- and yz-planes themselves where the traces are pairs of intersecting lines. x Notice that every point on the elliptic cone is on a line which lies entirely on the surface; in Figure 1.6.9 these Figure 1.6.9 Elliptic cone lines all go through the origin. This makes the elliptic cone an example of a ruled surface. The cylinder is also a ruled surface. What may not be as obvious is that both the hyperboloid of one sheet and the hyperbolic paraboloid are ruled surfaces. In fact, on both surfaces there are two lines through each point on the surface (see Exercises 11-12). Such surfaces are called doubly ruled surfaces, and the pairs of lines are called a regulus. It is clear that for each of the six types of quadric surfaces that we discussed, the surface can be translated away from the origin (e.g. by replacing x2 by (x − x0 )2 in its equation). It can be proved11 that every quadric surface can be translated and/or rotated so that its equation matches one of the six types that we described. For example, z = 2xy is a case of equation (1.33) with “mixed” variables, e.g. with D 0 so that we get an xy term. This equation does not match any of the types we considered. However, by rotating the x- and y-axes by 45◦ in the xy-plane by means of the coor√ √ dinate transformation x = (x′ − y′ )/ 2, y = (x′ + y′ )/ 2, z = z′ , then z = 2xy becomes the hyperbolic paraboloid z′ = (x′ )2 − (y′ )2 in the (x′ , y′ , z′ ) coordinate system. That is, z = 2xy is a hyperbolic paraboloid as in equation (1.38), but rotated 45◦ in the xy-plane.
11

See Ch. 7 in P OGORELOV.

3). −4. 0. 3) and (0. b. 2x2 + 2y2 + 2z2 + 4x + 4y + 4z − 44 = 0 2. 7.35) as a2 − c2 = 1 − b2 . 0). Find the trace of the hyperboloid of one sheet the trace in the plane y = b. z = 3 + t. 160. 0.46 CHAPTER 1. Show that the hyperboloid of one sheet is a doubly ruled surface. Let S be the sphere with radius 1 centered at (0.6. x2 + y2 + z2 + 2x − 2y − 8z + 19 = 0 4. . 2). B 6. (1. and y2 b2 2 2 − = z c in the xy-plane. 0. Find the intersection of the sphere x2 + y2 + z2 = 9 and the cylinder x2 + y2 = 4. 2) (a. (0. 0) x Figure 1. Show that the hyperbolic paraboloid is a doubly ruled surface.e. 0). 2). Find the trace of the hyperbolic paraboloid x2 a2 x2 a2 y z + b2 − c2 = 1 in the plane x = a. b and c. 8. 1). y. C 10. points that do not lie in the same plane) determine a sphere. c) be an arbitrary point on S ∗ . y = −2 − 3t. c) 1 0 (x. 2) and (a.10. Then the line passing through (0. b. (Hint: Exercise 11) 13. If so. which essentially identifies all of 2 with a “punctured” sphere. c) intersects the xyplane at some point (x. Let (a. and vice versa. This method is called stereographic projection. 1. x2 + y2 − z2 + 12x + 2y − 4z + 32 = 0 5. Find the intersection of the spheres x2 + y2 + z2 = 9 and (x − 4)2 + (y + 2)2 + (z − 4)2 = 9. Find this point (x. factor each side. −1. 0. VECTORS IN EUCLIDEAN SPACE  A Exercises ©  ¨ For Exercises 1-4. b. find its radius and center. p.6. for a proof. x2 + y2 + z2 − 4x − 6y − 10z + 37 = 0 3. y.) 12 z (0. in this manner. 0. as in Figure 1. 0. Recall that two planes intersect in a line.) 12. (Hint: Equation (1. each point on the surface is on two lines lying entirely on the surface. It can be shown that any four noncoplanar points (i.31)) 11.12 Find the equation of the sphere that passes through the points (0. 0) in terms of a. and let S ∗ be S without the “north pole” point (0. y. (Hint: Write equation y2 z2 x2 (1. Find the point(s) of intersection of the sphere (x − 3)2 + (y + 1)2 + (z − 3)2 = 9 and the line x = −1 + 2t. (Note: Every point in the xy-plane can be matched with a point on S ∗ . 9.10 S y See W ELCHONS and K RICKENBERGER. i.e. determine if the given equation describes a sphere.

and let P0 = (x. z) are defined as follows: Cylindrical coordinates (r. and φ is undefined when (x. y) as a point in 2 . θ. z) = (0. Note that r ≥ 0.7.7. ρ ≥ 0 and 0 ≤ φ ≤ π. then parallel to y the z-axis. y. Cylindrical coordinates are often used when there is symmetry around the z-axis. z) are determined by following a family of straight paths from the origin: first z along the x-axis. Let ρ be the length of the line segment from the origin to P. y. 0 ≤ θ < 2π. φ): x = ρ sin φ cos θ y = ρ sin φ sin θ z = ρ cos φ ρ= x2 + y2 + z2 y x z x2 +y2 +z2 z ρ φ 0 x θ x y P(x. y. φ) of P(x. z) be a point in Cartesian coordinates in 3 . y. 0).7. as with Cartesian coordinates.1 sides of a rectangular parallelepiped. then parallel to the y-axis.2 Cylindrical coordinates Spherical coordinates (ρ.7. θ.3 Spherical coordinates Both θ and φ are measured in radians. y. z): x = r cos θ y = r sin θ z=z r= x2 + y2 y x z P(x. Instead of referencing a point in terms of Figure 1. while spherical coordinates are useful when there is symmetry about the origin.7 Curvilinear Coordinates z (x. θ is undefined when (x. θ) be its polar coordinates (see Figure 1. The two types of curvilinear coordinates which we will consider are cylindrical and sphery x ical coordinates.1. z) z y θ = tan−1 φ = cos−1 √ P0 (x. y. y) = (0.7. . 0) be the projection of P upon the xy-plane. In curvilinear coordinate sysx 0 tems. as in Figure 1. z) The Cartesian coordinates of a point (x. y. these paths can be curved. 0) where 0 ≤ θ ≤ π if y ≥ 0 and π < θ < 2π if y < 0 Figure 1.1. and let φ be the angle between that line segment and the positive z-axis (see Figure 1. we will think of the point as lying on a cylinder or sphere. z) and the spherical coordinates (ρ. Also. θ. Treating (x. θ.7. 0. y. let (r. y. 0). y. 0) where 0 ≤ θ ≤ π if y ≥ 0 and π < θ < 2π if y < 0 Figure 1.2). Let P = (x. Then the cylindrical coordinates (r.7 Curvilinear Coordinates 47 1.3). φ is called the zenith angle. z) z θ = tan−1 z=z 0 x θ x y y r P0 (x.

7. we see from Figure 1. For cylindrical coordinates (r. and the surface z = z0 is a plane parallel to the xy-plane. θ0 and z0 . θ.7.31. θ. θ.5 that the surface ρ = ρ0 is a sphere of radius ρ0 centered at the origin. 4 . 1 √ 1 (b) ρ = (−2)2 + (−2)2 + 12 = 9 = 3. since y = −2 < 0. Convert the point (−2. the surface θ = θ0 is a half-plane emanating from the z-axis.23 4 Example 1. φ) = 3.4 Cylindrical coordinate surfaces For spherical coordinates (ρ.7.5 Spherical coordinate surfaces Figures 1. 1) from Cartesian coordinates to (a) cylindrical and (b) spherical coordinates. we see from Figure 1. z ρ0 y 0 0 θ0 x (a) ρ = ρ0 z z y φ0 y 0 x (c) φ = φ0 x (b) θ = θ0 Figure 1. z) = 2 2. θ. and the surface φ = φ0 is a circular cone whose vertex is at the origin.7. z r0 y 0 0 θ0 x (a) r = r0 z z0 y z y 0 x (b) θ = θ0 x (c) z = z0 Figure 1. and constants ρ0 . 5π .23 radians. θ = tan−1 −2 = tan−1 (1) = 5π . 1.4(a) and 1.5(a) show how these coordinate systems got their names. the surface θ = θ0 is a half-plane emanating from the z-axis. z). .7. −2.4 that the surface r = r0 is a cylinder of radius r0 centered along the z-axis.7. θ0 and φ0 .48 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE ∴ (ρ. √ Solution: (a) r = (−2)2 + (−2)2 = 2 2. −2 4 √ 5π ∴ (r. φ). φ = cos−1 3 ≈ 1. and constants r0 .

Example 1.5 -1.6 shows a section of this surface restricted to 0 ≤ z ≤ 4π and 0 ≤ r ≤ 2. while the radius r is unrestricted.7 Curvilinear Coordinates 49 Sometimes the equation of a surface in Cartesian coordinates can be transformed into a simpler equation in some other coordinate system. 14 12 10 8 z 6 4 2 0 -2 -1 -0. so we get ρ2 − 4ρ sin φ cos θ − 2ρ sin φ sin θ − 4 = 0 . Solution: Multiplying the equation out gives x2 + y2 + z2 − 4x − 2y + 5 = 9 . Note that this actually makes it more difficult to figure out what the surface is. 0). So this sweeps out a (ruled!) surface shaped like a spiral staircase. Write the equation (x − 2)2 + (y − 1)2 + z2 = 9 in spherical coordinates. Describe the surface given by θ = z in cylindrical coordinates.33.34.5 1 1 1.7. Write the equation of the cylinder x2 +y2 = 4 in cylindrical coordinates.1.5 1. where the spiral has an inﬁnite radius. then the equation in cylindrical coordinates is r = 2. as in the following example. As the (vertical) z coordinate increases. 1. Example 1. so does the angle θ. if the sphere is not centered at the origin. Figure 1. or ρ2 − 2 sin φ (2 cos θ − sin θ ) ρ − 4 = 0 after combining terms.5 2 2 x -2 -1.5 -1 0 -0. Solution: Since r = x2 + y2 . Solution: This surface is called a helicoid.32.5 y 0 0. as opposed to the Cartesian equation where you could immediately identify the surface as a sphere of radius 3 centered at (2. Example 1. Using spherical coordinates to write the equation of a sphere does not necessarily make the equation simpler.5 0.5 Figure 1.6 Helicoid θ = z .7.

θ2 . Describe the intersection of the surfaces whose equations in spherical coordinates are θ = π and φ = π . 5. C 10. 1 2 . −1) For Exercises 5-7. find the (a) cylindrical and (b) spherical coordinates of the point whose Cartesian coordinates are given. Since 0 < φ < π. θ2 . Let v1 be the vector from the origin to P1 . This formula is used in electrodynamics to prove the addition theorem for spherical harmonics.50 CHAPTER 1. the line segment from the origin to P can be extended to intersect the cylinder given by r = a (in cylindrical coordinates). (−5. θ1 . Show that the distance d between the points P1 and P2 with cylindrical coordinates (r1 . VECTORS IN EUCLIDEAN SPACE  A Exercises ©  ¨ For Exercises 1-4. 0. 2) 1. write the given equation in (a) cylindrical and (b) spherical coordinates. θ1 . See pp. respectively. z1 ) and (r2 . φ1 ) and (ρ2 . 0) with radius |a|. 11. 5. Show that the distance d between the points P1 and P2 with spherical coordinates (ρ1 . (0. Show that for a 0. φ1 ) and (ρ2 . 6) 3. x2 + y2 + z2 = 25 6. 100-102 in J ACKSON. show that cos γ = cos φ1 cos φ2 + sin φ1 sin φ2 cos( θ2 − θ1 ). φ) be a point in spherical coordinates. ( 21. respectively. 12. which provides a general expression for the electrostatic potential at a point due to a unit charge. For the angle γ between v1 and v2 . respectively. Let P = (a. φ2 ). Then P lies on the sphere ρ = a. x2 + y2 + 9z2 = 36 B 8. 2 3. and let v2 be the vector from the origin to P2 . (2. the equation ρ = 2a sin φ cos θ in spherical coordinates describes a sphere centered at (a. x2 + y2 = 2y 7. is d= 2 2 r1 + r2 − 2r1 r2 cos( θ2 − θ1 ) + (z2 − z1 )2 . 2. Let P1 and P2 be points whose spherical coordinates are (ρ1 . 0) 4. 2 4 9. Find the cylindrical coordinates of that point of intersection. √ √ √ √ 2. φ2 ). θ1 . is d= ρ2 + ρ2 − 2ρ1 ρ2 [sin φ1 sin φ2 cos( θ2 − θ1 ) + cos φ1 cos φ2 ] . 13. θ2 . θ. z2 ). with a > 0 and 0 < φ < π. − 7.

For each t. sin t.8. f(t) = ti + t2 j + t3 k is a vector-valued function in 3 . so x + y = cos t + sin t = 1. a curve in space can be written as a vector-valued function. 1). f2 (t). By identifying vectors with their terminal points. The first form is often used when emphasizing that f(t) is a vector. the terminal points of f(t) trace out a curve spiraling upward. which in Cartesian coordinates has the terminal point (1. However. t). f3 (t). Definition 1. Since each of the three component functions are real-valued.1). Much of the theory of real-valued functions of a single real variable can be applied to vector-valued functions of a real variable. can be extended naturally to vector-valued functions. 1.and y-coordinates of f(t) are x = cos t and y = sin t. This is the equation of a helix (see Figure 1.8 Vector-Valued Functions 51 1. the x. defined for all real numbers t. the curve lies on the surface of the right circular cylinder x2 + y2 = 1. We write f : D → 3 to denote that f is a mapping of D into 3 . and the second form is useful when considering just the terminal points of the vectors. where t is in some subset D of 1 (called the domain of f). z Example 1. it will sometimes be the case that results from singlevariable calculus can simply be applied to each of the component functions to yield a similar result for the vector-valued function. though. f3 (t)) for some real-valued functions f1 (t). At t = 1 the value of the function is the vector i + j + k.8. As the value of t increases. as in the following definition. 2 2 2 2 f(2π) f(0) x y 0 Figure 1. For example. We would write f : → 3 . called the component functions of f. we can begin discussing functions whose values are vectors. The concept of a limit.10.1. . A vector-valued function of a real variable can be written in component form as f(t) = f1 (t)i + f2 (t)j + f3 (t)k or in the form f(t) = ( f1 (t). there are times when such generalizations do not hold (see Exercise 13).35. Define f : → 3 by f(t) = (cos t.1 It may help to think of vector-valued functions of a real variable in 3 as a generalization of the parametric functions in 2 which you learned about in single-variable calculus.8 Vector-Valued Functions Now that we are familiar with vectors and their operations. f2 (t). Thus. A vector-valued function of a real variable is a rule that associates a vector f(t) with a real number t.

t). lim f2 (t). 2π) is L = f(2π) + s f ′ (2π) = (1. f2 (t). Then we say that the limit of f(t) as t approaches a equals c.52 CHAPTER 1. if the component derivatives exist. f3 ′ (a)). 1). cos t. sin t. t→a f(t) is continuous at a if and only if f1 (t). The above definition shows that continuity and the derivative of vector-valued functions can also be defined in terms of its component functions. the derivative of a vector-valued function is a tangent vector to the curve in space which the function represents. 1.2 Tangent vector f ′ (a) and tangent line L = f(a) + sf ′ (a) Example 1. and let a be a real number in its domain. denoted by f ′ (a) or (a). then t→a t→a lim f(t) = lim f1 (t). Similarly. Let f(t) be a vector-valued function.8. f2 (t).11. Then f(t) is continuous at a if lim f(t) = f(a). Let f(t) = (cos t. and it lies on the tangent line to the curve (see Figure 1. and f3 (t) are continuous at a. f ′ (a) = ( f1 ′ (a). f3 (t)) be a vector-valued function. The tangent line L to the curve at f(2π) = (1.36. f2 ′ (a). z = 2π + s for −∞ < s < ∞. is the limit dt f ′ (a) = lim f(a + h) − f(a) h→0 h if that limit exists. Then f ′ (t) = (− sin t. 0. Definition 1. Equivalently. VECTORS IN EUCLIDEAN SPACE Definition 1. f3 (t)). Let f(t) = ( f1 (t). lim f3 (t) t→a t→a t→a t→a provided that all three limits on the right side exist. Equivalently.8.12. f2 (t). df The derivative of f(t) at a. 2π) + s(0. 0. 1) for all t. or in parametric form: x = 1. if lim f(t) − c = 0. If f(t) = ( f1 (t). representing the slope of the tangent line to the graph of the function at a point. let a be a real number and let c be a vector. written as lim f(t) = c. Recall that the derivative of a real-valued function of a single variable is a real number. . y = s. We say that f(t) is differentiable at a if f ′ (a) exists. z f(a) f(a f ′ (a) + h) − L f(t) y f(a ) f(a + h) 0 x Figure 1.2).

1. Then d d (f(t) · g(t)) = ( f1 (t) g1 (t) + f2 (t) g2 (t) + f3 (t) g3 (t)) dt dt d d d = ( f1 (t) g1 (t)) + ( f2 (t) g2 (t)) + ( f3 (t) g3 (t)) dt dt dt d f1 dg1 d f2 dg2 d f3 dg3 = (t) g1 (t) + f1 (t) (t) + (t) g2 (t) + f2 (t) (t) + (t) g3 (t) + f3 (t) (t) dt dt dt dt dt dt d f2 d f3 d f1 (t). (t). (t). f2 (t). f3 (t)) · (t). let u(t) be a differentiable scalar function. Then d (c) = 0 dt d df (b) (kf) = k dt dt df dg d + (c) (f + g) = dt dt dt d df dg (d) (f − g) = − dt dt dt du df d f+u (e) (u f) = dt dt dt d df dg (f) (f · g) = · g + f· dt dt dt df dg d × g+f× (g) (f × g) = dt dt dt (a) Proof: The proofs of parts (a)-(e) follow easily by differentiating the component functions and using the rules for derivatives from single-variable calculus. where the component functions f1 (t). and leave the proof of part (g) as an exercise for the reader.8 Vector-Valued Functions 53 A scalar function is a real-valued function. f2 (t). Let f(t) and g(t) be differentiable vector-valued functions. let k be a scalar. (f) Write f(t) = ( f1 (t). g2 (t). f3 (t). f3 (t)) and g(t) = (g1 (t). The basic properties of derivatives of vector-valued functions are summarized in the following theorem. g2 (t).20. (t) · (g1 (t). defined by (u f)(t) = u(t) f(t) for all t. g3 (t)) = dt dt dt dg2 dg3 dg1 + ( f1 (t). g1 (t). We will prove part (f). Theorem 1. g3 (t)). Note that if u(t) is a scalar function and f(t) is a vector-valued function. is a vector-valued function (since the product of a scalar with a vector is a vector). QED dt dt . (t) dt dt dt df dg = (t) · g(t) + f(t) · (t) for all t. f2 (t). g2 (t). g3 (t) are all differentiable real-valued functions. and let c be a constant vector. then their product.

8 -0. then f(t) is constant if and only if f(t) ⊥ f ′ (t) for all t. we have dt dt 2 f(t) d d f(t) = (f(t) · f(t)) = f ′ (t) · f(t) + f(t) · f ′ (t) by Theorem 1.3 Spherical spiral with a = 0.8. Find the derivative of f(t) .3 shows the graph of the curve when a = 0.6 -0.2 y 0 0. Hence. The spherical spiral f(t) = 0.6 0.2 0. the reader will be asked to show that this curve lies on the sphere x2 + y2 + z2 = 1 and to verify directly that f ′ (t) · f(t) = 0 for all t.2. dt f(t) d We know that f(t) is constant if and only if f(t) = 0 for all t.05 -0. Solution: Since f(t) is a real-valued function of t.4 0.15 0.2 -1 -0.2 . so f(t) 2 = (f(t) · f(t)).2 0.37.8 -1 -0.6 0. Suppose f(t) is differentiable.20(f). Also. Example 1. VECTORS IN EUCLIDEAN SPACE Example 1.4 -0. so if f(t) 0 then ′ (t) · f(t) f d f(t) = .8.54 CHAPTER 1. then the tangent vector f ′ (t) is always perpendicular to the position vector f(t).15 -0. dt dt d d But f(t) 2 = f(t) · f(t). f(t) ⊥ f ′ (t) if dt and only if f ′ (t) · f(t) = 0. √ .1 0.2 x 0 0. so dt dt = 2f ′ (t) · f(t) . the above example shows this important fact: If f(t) 0. cos t sin t −at .2 0.6 -0.4 0. This means that if a curve lies completely on a sphere (or circle) centered at the origin. then by the Chain Rule for reald d valued functions.05 0 z-0.8 -0.8 1 1 Figure 1. for a 0.1 -0. In the exercises.4 -0. Thus. √ . √ 1 + a2 t 2 1 + a2 t 2 1 + a2 t 2 Figure 1.38. we know that f(t) 2 = 2 f(t) f(t) .

1. y (t). acceleration. −3 sin t. z (t)) ˙ velocity: v(t) = r(t) = r ′ (t) = momentum: p(t) = mv(t) ˙ force: F(t) = p(t) = p ′ (t) = dp dt (Newton’s Second Law of Motion) The magnitude v(t) of the velocity vector is called the speed of the object.8 Vector-Valued Functions 55 Just as in single-variable calculus.37 we know that ˙ r(t) · r(t) = 0 for all t (which we can verify from part (a)). Also. In fact. 3 sin t. z) at time t a function of t. y(t). 4 sin t) be the position vector of an object at time t ≥ 0. −4 sin t) We will often use the older dot notation for derivatives when physics is involved. higher-order derivatives of vector-valued functions are obtained by repeatedly differentiating the (first) derivative of the function: f ′′ (t) = d ′ f (t) . dn f d dn−1 f = dtn dt dtn−1 (for n = 2. 4. 3. z(t). Let r(t) = (5 cos t. dt f ′′′ (t) = d ′′ f (t) . towards the center of the circle). with its position (x. Example 1. 3 cos t. y. y (t). . so by Example 1. Find its (a) velocity and (b) acceleration vectors. ˙ Solution: (a) v(t) = r(t) = (−5 sin t. It turns out (see Exercise 16) that whenever an object moves in a circle with constant speed. z (t)) dv ˙ acceleration: a(t) = v(t) = v ′ (t) = dt d2 r ¨ = r(t) = r ′′ (t) = 2 dt ′′ ′′ ′′ = (x (t). v(t) = 5 for all t also. the force equation becomes the familiar F(t) = ma(t). . x = x(t). z(t)) the position vector of the object.. That is. . and suppose that an object of constant mass m is subjected to some force so that it moves in space. such as velocity. note that a(t) = −r(t). the acceleration vector will point in the opposite direction of the position vector (i. momentum.) We can use vector-valued functions to represent physical quantities. z = z(t) for some real-valued functions x(t).. etc. but perhaps not so obvious is that it lies completely within a circle of radius 5 centered at the origin. dt . force. For example. . . z(t)) dr dt ′ ′ ′ = (x (t). 4 cos t) Note that r(t) = 25 cos2 t + 25 sin2 t = 5 for all t. Note that since the mass m is a constant.e.39. let the real variable t represent time elapsed from some initial time (t = 0). y(t). Call r(t) = (x(t). We can define various physical quantities associated with the object as follows:13 position: r(t) = (x(t). y = y(t). 13 ˙ (b) a(t) = v(t) = (−5 cos t. And not only does r(t) lie on the sphere of radius 5 centered at the origin. y(t).

4t + t2 . 0.3) 3 2. 0). 0 1 0 (1. as shown in Figure 1. 1] it is the line segment between the points. and the curve is b2 (t).2) (0. define b1 (t) = (1 − t)b0 + tb1 0 b1 (t) = (1 − t)b1 + tb2 1 b2 (t) = (1 − t)b1 (t) + tb1 (t) 1 0 0 = (1 − t)2 b0 + 2t(1 − t)b1 + t2 b2 for all real t.8.5 that if r1 .5 z 1 0. a3 t + b3 ) represents a line in 3 . b2 in 3 . VECTORS IN EUCLIDEAN SPACE Recall from Section 1.5 (4. b2 . 1].4 Bézier curve approximation for three points . 2. we see that b1 (t) is the line segment between 0 b0 and b1 . given three points (or position vectors) b0 . b1 = (1. So the function l(t) = (1 − t)r1 + tr2 is a line through the terminal points of r1 and r2 . Note from the last formula that the curve is a parabola that goes through b0 (when t = 0) and b2 (when t = 1).0. a function of the form f(t) = (a1 t + b1 . That vector sum can be written as (1 − t)r1 + tr2 . 0 where the line segments are b1 (t) and b1 (t). r2 are position vectors to distinct points then r1 +t(r2 −r1 ) represents a line through those two points as t varies over all real numbers. and b1 (t) is the line segment between b1 and b2 .5 x 1 1. with l(0) = r1 and l(1) = r2 . let b0 = (0. b1 . 3). and when t is restricted to the interval [0. Example 1.8. 2). 5. A function of the form f(t) = (a t2 + b t + c . Then the explicit formula for the Bézier curve is b2 (t) = (2t + 2t2 . a t2 + b t + c . For instance.40.5 1 2 y 3 4 5 4 3 3.5 2 1. As an example. The function b2 (t) is the 1 0 Bézier curve for the points b0 . and b2 = (4.5.2. b1 . a t2 + b t + c ) represents a 1 1 1 2 2 2 3 3 3 (possibly degenerate) parabola in 3 .0) 0 0. Bézier curves are used in Computer Aided Design (CAD) to approximate the shape of a polygonal path in space (called the Bézier polygon or control polygon).5 0 0 2 2. In general.4. 6t − 4t2 ). For t in the interval [0.5 Figure 1.56 CHAPTER 1. a2 t + b2 .

5. 27-30 in FARIN.1. √ . find the velocity v(t) and acceleration a(t) of an object with the given position vector r(t). b). 5. If f ′ (t) = 0 for all t in some interval (a. with a 2 t2 2 t2 1+a 1+a 1 + a2 t 2 (a) Show that f(t) = 1 for all t. . b). 2 cos t) 2 For Exercises 5-6. 2 sin t.1) 1. 2. (b) Show directly that f ′ (t) · f(t) = 0 for all t. f(t) = (sin 2t.0. and its formula is given by de Casteljau’s algorithm.3. r(t) = (3 cos t. f(t) = (t + 1.2) (0.5 4 5 4 Figure 1.0) 0 0 1 2 y 3 3 3.14 In the exercises.5 x 2 2. show that f(t) is a constant vector in (a. 1) B cos t sin t −at 7.8.0) 0. This curve will be a vector-valued function whose components are polynomials of degree n − 1. t2 + 1. f(t) = (et + 1. t3 + 1) 3. the polygonal path determined by n ≥ 3 noncollinear points in 3 can be used to define the Bézier curve recursively by a process called repeated linear interpolation.5 1 1.8. 2 sin2 t. et + 1) 4.1. calculate f ′ (t) and find the tangent line at f(0). r(t) = (t.5. √ . sin 2t.5 0 0. Let f(t) = √ . t − sin t.5 Bézier curve approximation for four points  ¨ A 1. f(t) = (cos 2t. 2 (4.5 (2. 1 − cos t) 6.8 Vector-Valued Functions 57 In general. the reader will be given the algorithm for the case of n = 4 points and asked to write the explicit formula for the Bézier curve for the four points shown in Figure 1. t) Exercises ©  For Exercises 1-4.5 1 z (0. 14 0. e2t + 1. See pp. 8.

dt dt dt dt 13. Show that d (r × (v × r)) = r 2 a + (r · v)v − ( v 2 + r · a)r. b2 = (2. b3 in 0 the following algorithm (going from the left column to the right): b1 (t) = (1 − t)b0 + tb1 0 b1 (t) = (1 − t)b1 + tb2 1 b2 (t) = (1 − t)b1 (t) + tb1 (t) 0 0 1 b2 (t) = (1 − t)b1 (t) + tb1 (t) 1 1 2 is defined by b3 (t) = (1 − t)b2 (t) + tb2 (t) 0 0 1 (a) Show that b3 (t) = (1 − t)3 b0 + 3t(1 − t)2 b1 + 3t2 (1 − t)b2 + t3 b3 . b3 = (4. 5. 2π − 0 3 C 14. The angular momentum L(t) of the particle with respect to the origin at time t is defined as L(t) = r(t) × p(t). 15. acceleration a(t) and momentum p(t) at time t. The Mean Value Theorem does not hold for vector-valued functions: Show that for f(t) = (cos t. Let a particle of (constant) mass m have position vector r(t). b1 = (0.58 CHAPTER 1. b1 (t) = (1 − t)b2 + tb3 2 . b2 . (Hint: Use Example 1. 0.40) for the Bézier curve for the points b0 = (0.37 to show that r(t) ⊥ v(t) and a(t) ⊥ v(t). Show that d df d2 f f× = f × 2. b1 . Show that a(t) points in the opposite direction as r(t) for all t. Let r(t) be the position vector in 3 for a particle that moves with constant speed c > 0 in a circle of radius a > 0 in the xy-plane. 1). 1. dt 16. 9. VECTORS IN EUCLIDEAN SPACE 0. there is no t in the interval (0. the function f(t) = tc represents a line parallel to c. 2π) such that f ′ (t) = f(2π) − f(0) . The Bézier curve b3 (t) for four noncollinear points b0 . 3. 2). Prove Theorem 1. 12.) 17. 0). how do you explain the difference in the two derivatives? 10. t). (c) Compare f ′ (0) and g ′ (0). Show that d dg df dh (f · (g × h)) = · (g × h) + f · × h + f· g × . then define the torque N(t) acting on the particle with respect to the origin as N(t) = r(t) × F(t). velocity v(t). For a constant vector c (a) What kind of curve does g(t) = t3 c represent? Explain. 0).20(g). 0 (b) Write the explicit formula (as in Example 1. sin t. and hence a(t) r(t). Given your answer to part (a). dt dt dt 11. (b) What kind of curve does h(t) = et c represent? Explain. Show that L ′ (t) = N(t). If F(t) is the force acting on the particle at time t. Let r(t) be the position vector for a particle moving in 3 .

Find the length L of the helix f(t) = (cos t.2. y(t) and z(t) exists and is continuous. which can be ignored).41) A real-valued function whose first derivative is continuous is called continuously differentiable (or a C1 function). normally glossed over in calculus texts. y(t). b] into subintervals where all the component functions are continuously differentiable (except at the endpoints. and that no section of the curve is repeated. Since v(t) is the speed of the object at time t. b) the first derivative of each component function x(t). z(t)) be the position vector of an object moving in 3 . (1.41).13. the arc length of a curve in 3 . and a function whose derivatives of all orders are continuous is called smooth (or a C∞ function). 15 In particular. A rigorous proof requires dealing with some subtleties. y(t).2 and § 18. This is indeed how we will define the distance traveled and.1. which are beyond the scope of this book. Duhamel’s principle is needed.15 Example 1. See the proof in T AYLOR and M ANN. z(t)) be a curve in 3 whose domain includes the interval [a. A smooth curve f(t) is one whose derivative f ′ (t) is never the zero vector and whose component functions are all smooth.40) which is analogous to the case from single-variable calculus for parametric functions in 2 . Suppose that in the interval (a. Then the arc length L of the curve from t = a to t = b is b b L= a f ′ (t) dt = a x ′ (t)2 + y ′ (t)2 + z ′ (t)2 dt (1. Definition 1. if there are values of t in the interval [a. . sin t. we have 2π 2π 2π L= (− sin t)2 + (cos t)2 + 12 dt = 0 √ √ = 2(2π − 0) = 2 2π sin2 t + cos2 t + 1 dt = 0 0 √ 2 dt Similar to the case in 2 .9 Arc Length 59 1. All the functions we will consider will be smooth. Let f(t) = (x(t). b]. The sum of the arc lengths over the subintervals will be the arc length over [a.41. b] where the derivative of a component function is not continuous then it is often possible to partition [a. b]. in general. § 14. Solution: By formula (1. Note that we did not prove that the formula in the above definition actually gives the length of a section of a curve.9 Arc Length Let r(t) = (x(t). it seems natural to define the distance s traveled by the object from time t = a to t = b as the definite integral b b s= a v(t) dt = a x ′ (t)2 + y ′ (t)2 + z ′ (t)2 dt . t) from t = 0 to t = 2π.

b] be a smooth one-to-one mapping of an interval [c. respectively.14.60 CHAPTER 1.21. 2π]. viewing the functions as position vectors and their derivatives as velocity √ √ vectors. Likewise. b]. one-to-one. g(t) traces out the same section of the curve as f(t) does over the interval [0. This makes sense since. 2π].41 is also traced out by the function g(t) = (cos 2t. 2s) for s in [0. sin 2s. Example 1. maps [0. d] onto [a. d] → 3 defined by g(s) = f(α(s)) is a parametrization of C with parameter s. 1] To see that g(s) is equivalent to f(t). If α is strictly increasing on [c. Let C be a smooth curve in 3 represented by a function f(t) defined on an interval [a. The following are all equivalent parametrizations of the same curve: f(t) = (cos t. Chain Rule: If f(t) is a differentiable vector-valued function of t. over the interval [0. s [c. define α : [0. Then the function g : [c. the speeds of f(t) and g(t) are f ′ (t) = 2 and g ′ (t) = 2 2. and t = α(s) is a differentiable scalar function of s. b] f(t) 3 f g(s) = f(α(s)) = f(t) Note that the differentiability of g(s) follows from a version of the Chain Rule for vector-valued functions (the proof is left as an exercise): Theorem 1. and is strictly increasing (since α ′ (s) = 2 > 0 for all s). and let α : [c. . t) for t in [0. sin 2t. and df df dt = (1. d] then we say that g(s) is equivalent to f(t). Definition 1. Then α is smooth. 2t). We say that g(t) and f(t) are different parametrizations of the same curve. then f(s) = f(α(s)) is a differentiable vector-valued function of s. this says that g(t) traces the curve twice as fast as f(t). π] h(s) = (cos 2πs. 2πs) for s in [0. b]. 2π] by α(s) = 2πs shows that h(s) is equivalent to f(t). 2π] g(s) = (cos 2s. For example. d] → [a. sin t. π] → [0. sin 2πs. π]. 2π] by α(s) = 2s. VECTORS IN EUCLIDEAN SPACE Notice that the curve traced out by the function f(t) = (cos t. t) from Example 1. Intuitively.42.42) ds dt ds for any s where the composite function f(α(s)) is defined. sin t. d] α t [a. 1] → [0. π] onto [0. defining α : [0.

9 Arc Length 61 A curve can have many parametrizations. differentiable mapping onto the interval [0. b] → [0. Thus s ′ (t) > 0 and hence s(t) is strictly increasing on the interval [a. L]. with different speeds. for any given smooth parametrization f(t) defined on [a. Since f(t) is smooth. so 1 = f ′ (α(s)) ′ . L]. L] t [a. by the Chain Rule. So the arc length parametrization traverses the curve at a “normal” rate. L] → 3 s [0. b] → [0. for each t in [a. b] s(t) Figure 1. t] for each t in [a. By the Fundamental Theorem of Calculus. That is. So the new parameter will be distance instead of time. L] → [a.1. the distance traveled along the curve (in one direction) is uniquely determined by the amount of time elapsed. b] that is differentiable and the inverse of s : [a. then it is a function of t. f(s) has unit speed: f ′ (s) = f ′ (α(s)) α ′ (s) by the Chain Rule. by the parameter s given by t s = s(t) = a f ′ (u) du. b] onto the interval [s(a). L].9. so f (α(s)) f ′ (s) = 1 for all s in [0. and vice versa. Then f(s) is smooth. then f ′ (t) > 0 for all t in [a. There is a natural correspondence between s and t: from a starting point on the curve. its derivative is s ′ (t) = ds d = dt dt t f ′ (u) du = f ′ (t) a for all t in [a. b]. From single-variable calculus. b]. Since s is the arc length of the curve over the interval [a. Recall that this means that s is a one-to-one mapping of the interval [a. L]. so which one is the best to use? In some situations the arc length parametrization can be useful. L] is a one-to-one. In fact. . b]. L] such that s = s(t) and t = α(s). And we know that the derivative of α is 1 1 = ′ α ′ (s) = ′ s (α(s)) f (α(s)) So define the arc length parametrization f : [0.1 t = α(s) by f(s) = f(α(s)) for all s in [0. s is the distance traveled along the curve after time t has elapsed. The idea behind this is to replace the parameter t. b]. But we see that a b s(a) = a f ′ (u) du = 0 and s(b) = a f ′ (u) du = L = arc length from t = a to t = b α(s) So the function s : [a. (1.43) In terms of motion along a curve. we know that this means that there exists an inverse function α : [0. b]. s(b)]. b] there is a unique s in [0.

t).43). for t in [a. sin t.43. The simple integral in Example 1. Solution: By Example 1. 2 √ s s s ∴ f(s) = cos √ . θ = θ(t) and z = z(t) are the cylindrical coordinates of a curve f(t). which. We will leave this to the exercises. y(t) = r(t) sin θ(t). s So we can solve for t in terms of s: t = α(s) = √ . sin √ . 2π]. In general. √ for all s in [0. z(t)) of a point on the curve are given by x(t) = r(t) cos θ(t). for t in [0. If that can be done.41 and formula (1. in the field of mathematics known as differential geometry. parametrizing a curve f(t) by arc length requires you to evaluate the t integral s = a f ′ (u) du in some closed form (as a function of t) so that you could then solve for t in terms of s. But their arc length parametrizations are not only not polynomials. you would then substitute the expression for t in terms of s (which we called α(s)) into the formula for f(t) to get f(s). they are in fact usually impossible to calculate at all. This makes their computation relatively simple. are polynomial functions in 3 . we have t t s= 0 f ′ (u) du = 0 √ √ 2 du = 2 t for all t in [0. Parametrize the helix f(t) = (cos t. by arc length. b] is b L= a r ′ (t)2 + r(t)2 θ ′ (t)2 + z ′ (t)2 dt (1.44) Proof: The Cartesian coordinates (x(t). not the norm. which often leads to an integral that is either difficult or impossible to evaluate in a simple closed form. in CAD. 2 2π]. and these definitions can be shown to be equivalent to those using arc length. which we discussed in Section 1. VECTORS IN EUCLIDEAN SPACE In practice. y(t).16 The methods involve using an arc length parametrization.22. 16 17 y ′ (t) = r ′ (t) sin θ(t) + r(t)θ ′ (t) cos θ(t) See O’N EILL for an introduction to elementary differential geometry.17 Curvature and moving frame fields can be defined without using arc length. which makes their computation much easier. b].62 CHAPTER 1. the usual parametrizations of Bézier curves.8. Example 1. 2 2 2 Arc length plays an important role when discussing curvature and moving frame fields. The arc length for curves given in other coordinate systems can also be calculated: Theorem 1. 2π]. Suppose that r = r(t). is desirable. . arc length parametrizations are more useful for theoretical purposes than for practical computations.43 is the exception. z(t) = z(t) so differentiating the above expressions for x(t) and y(t) with respect to t gives x ′ (t) = r ′ (t) cos θ(t) − r(t)θ ′ (t) sin θ(t). Then the arc length L of the curve over [a. Note that f ′ (s) = 1. For example.

1] 3. θ ′ (t) = 1 and z ′ (t) = et .44. f(t) = (2 cos 3t. 1. (t2 + 1) sin t. 1]. 3t) on [0. calculate the arc length of f(t) over the given interval. 3 sin 2t. 2 2t) on [0.    dt f(t)  f(t) 3 . f(t) = (3 cos 2t. Let f(t) be a differentiable curve such that f(t) 0 for all t.1. Find the arc length L of the curve whose cylindrical coordinates are r = et . Solution: Since r ′ (t) = et . 2t3/2 ) on [0. Parametrize the curve from Exercise 3 by arc length. B 6. Parametrize the curve from Exercise 1 by arc length. f(t) = ((t2 + 1) cos t. 5. π/2] √ 2. then 1 L= 0 1 r ′ (t)2 + r(t)2 θ ′ (t)2 + z ′ (t)2 dt e2t + e2t (1) + e2t dt = 0 1 = 0 √ √ et 3 dt = 3(e − 1)  A Exercises ©  ¨ For Exercises 1-3. for t over the interval [0. Show that   d  f(t)  f(t) × (f ′ (t) × f(t))    =   .9 Arc Length and so x ′ (t)2 + y ′ (t)2 = (r ′ (t) cos θ(t) − r(t)θ ′ (t) sin θ(t))2 + (r ′ (t) sin θ(t) + r(t)θ ′ (t) cos θ(t))2 = r ′ (t)2 (cos2 θ + sin2 θ) + r(t)2 θ ′ (t)2 (cos2 θ + sin2 θ) − 2r ′ (t)r(t)θ ′ (t) cos θ sin θ + 2r ′ (t)r(t)θ ′ (t) cos θ sin θ b 63 = r ′ (t)2 + r(t)2 θ ′ (t)2 . and so L= a b x ′ (t)2 + y ′ (t)2 + z ′ (t)2 dt r ′ (t)2 + r(t)2 θ ′ (t)2 + z ′ (t)2 dt QED = a Example 1. 2 sin 3t. 1] 4. θ = t and z = et .

t). the curvature κ is defined by κ(t) = Show that T ′ (t) f ′ (t) × (f ′′ (t) × f ′ (t)) . = ′ (t) f f ′ (t) 4 Show that f ′ (t) × f ′′ (t) and that T ′ (t) = f ′ (t) κ(t) N(t). f ′ (t) f ′′ (t) × f ′ (t) 9. Find T. Continuing Exercise 9. B and κ at each point of the helix f(t) = (cos t. sin t. κ(t) = 11. B at a point on a curve. Show that the arc length L of a curve whose spherical coordinates are ρ = ρ(t). Then we can define the unit f ′ (t) . Continuing Exercise 7. 7. 12. f ′ (t) 3 0 8. Let f(t) be a smooth curve such that f ′ (t) tangent vector T by T(t) = Show that T ′ (t) = 0 for all t. θ = θ(t) and φ = φ(t) for t in an interval [a. N. T ′ (t) f ′ (t) × (f ′′ (t) × f ′ (t)) . f ′ (t) × f ′′ (t) Note: The vectors T(t).64 CHAPTER 1. b] is b L= a ρ ′ (t)2 + (ρ(t)2 sin2 φ(t)) θ ′ (t)2 + ρ(t)2 φ ′ (t)2 dt. f ′ (t) f ′ (t) × (f ′′ (t) × f ′ (t)) . N(t) and B(t) form a right-handed system of mutually perpendicular unit vectors (called orthonormal vectors) at each point on the curve f(t). . VECTORS IN EUCLIDEAN SPACE Exercises 7-9 develop the moving frame field T. the unit binormal vector B is defined by B(t) = T(t) × N(t). B(t) = 10. Continuing Exercise 8. f ′ (t) × f ′′ (t) . N. Then T ′ (t) so we can define the unit principal normal vector N by N(t) = Show that N(t) = T ′ (t) . f ′ (t) 3 Note: κ(t) gives a sense of how “curved” the curve f(t) is at each point. assume that f ′ (t) and f ′′ (t) are not parallel.

y). We see that D consists of all points on and inside the unit circle in 2 (D is sometimes called the closed unit disk). y) = 1 − x2 − y2 is the set D = {(x. y) as (x. A real-valued function f defined on a subset D of 2 is a rule that assigns to each point (x. the domain is the set D = {(x. y. The domain of the function f (x. since the quantity inside the square root is nonnegative if and only if 1 − (x2 + y2 ) ≥ 0. The domain of the function f (x. 1] in . y) in D a real number f (x. The range of f is the interval [0.1 Functions of Two or Three Variables In Section 1. y) : x y}. but there will be times when we will use points in 3 .8 we discussed vector-valued functions of a single real variable. The largest possible set D in 2 on which f is defined is called the domain of f . y. y) varies over the domain D.3. The range of f is all real numbers except 0. Example 2. For the most part these functions will be defined on sets of points in 2 . That is. and the range of f is the set of all real numbers f (x. The domain of the function f (x. y) for which x = y.2. y) = 1 x−y is all of 2 except the points (x. z) defined on points (x. y) = xy is all of 2.1. Example 2. y) : x2 + y2 ≤ 1}. z) in 3 .2 Functions of Several Variables 2. and there will also be times when it will be convenient to think of the points as vectors (or terminal points of vectors). 65 . and the range of f is all of . We will now examine real-valued functions of a point (or vector) in 2 or 3 . Example 2. A similar definition holds for functions f (x.

1. where c varies over . the level curves are the solution sets of the equations f (x.5 at the point (x.1 The function f (x. We will now state explicitly what is meant by the limit of a function of two variables. 0). The traces of this surface in the planes z = c. y.2 -0. y) = c. F(x.2 0 -0. y) approaches (0. as was mentioned in Section 1. since it satisfies an equation of the form F(x. so that the graph of f (x. since both the numerator and denominator are 0 at that point. z) : z = f (x.5. y)} in 3 . 0). 0). The function is not defined at (0. Equivalently. The domain of the function f (x. y) is the set {(x. y) = sin x2 + y2 x2 + y2 is shown below. y) = (0. y) = √ 2 2 x +y You may be wondering what happens to the function in Example 2. y. FUNCTIONS OF SEVERAL VARIABLES Example 2.4 z 0.8 0. y) − z). Note that the level curves (shown both on the surface and projected onto the xy-plane) are groups of concentric circles. z) = f (x. z) = 0 (namely. but the limit of the function exists (and equals 1) as (x. So we see that this graph is a surface in 3 . 1 0.4 -5 -10 -5 y 0 0 5 5 10 10 x -10 √ sin x2 +y2 Figure 2. The graph of the function f (x.6 0. y. and the range of f is all positive real numbers. z) = e x+y−z is all of 3.1. Example 2.66 CHAPTER 2. y. are called the level curves of the function. . for c in . y).4. Level curves are often projected onto the xy-plane to give an idea of the various “elevation” levels of the surface (as is done in topography). y) defined in 2 is often written as z = f (x. A function f (x.

2. b) (i. In general. b) into the formula for f (x.e. The idea behind the above definition is that the values of f (x.2(a)). In two dimensions. the multivariable cases are at least equally awkward to go through.1. A similar definition can be made for functions of three variables. you may remember how awkward they can be. the statement “x → a” means that x gets closer to the value a from two possible directions along the real number line (see Figure 2. In the single-variable case. y x 0 a x x 0 (b) (x. b) in 2 (a. and how they can usually only be done easily for simple functions. y) is given by a single formula and is defined at the point (a. b) with some sufficiently small radius δ).1) if given any ǫ > 0.1 Functions of Two or Three Variables 67 Definition 2. y) = xy x2 +y2 is properly defined at the point (1. b) itself). b). however. y) equals L as (x. (x. and let f (x. y) → (a. within ǫ of L) if we pick (x. b) x (a) x → a in Figure 2. 2).b) lim f (x. y) can approach a point (a. (2. there exists a δ > 0 such that | f (x. y) sufficiently close to (a. y) can get arbitrarily close to L (i. y) = L . inside a circle centered at (a.1. b) (but not necessarily defined at (a. we will simply state that when the function f (x. Then we say that the limit of f (x. y) be a real-valued function defined on some set containing (a.6. b) along an infinite number of paths (see Figure 2.y)→(a.g. (x.e.1.1. b) (e. The major difference between limits in one variable and limits in two or more variables has to do with how a point is approached. written as (x. is not some indeterminate form like 0/0) then you can just substitute (x. Example 2. Let (a.2 “Approaching” a point in different dimensions .2) lim x2 xy (1)(2) 2 = 2 = 2 2 5 +y 1 +2 since f (x. so we will not bother with such proofs. b) be a point in 2 .y)→(1.2(b)). Instead. y) to find the limit. y) approaches (a. y) = (a. If you recall the “epsilon-delta” proofs of limits of real-valued functions of a single variable. y) − L| < ǫ whenever 0 < (x − a)2 + (y − b)2 < δ.

y) = 0.68 Example 2.y)→(a.b) lim lim g(x. y) approaches different values as (x. y) f (x.y)→(a. . we will show that the function approaches different values as (x. 0) along different paths. y) ± (x. Then f (x. y) lim g(x. y) for all (x.y)→(a. Suppose that some scalar. CHAPTER 2.y)→(a. which we state without proof.y)→(a.b) (x. y) ± g(x. y) (x. 0) along the positive x-axis. y) = L.b) (x.y)→(a. b) itself).y)→(a. 0) along different paths in 2 .b) lim [ f (x. y) = g(x. then we see that f (x.7.b) f (x. Then: (a) (b) (c) (x. then (x. y) both exist.y)→(a.y)→(a.b) Note that in part (e). y) = (0. y) approaches (0.b) lim f (x. y)g(x.y)→(a.y)→(0. 0) into the function. y)] = f (x. To see this.y)→(a. y) (x. y) = k lim (x. 0) along the straight line y = x through the origin. since doing so gives an indeterminate form 0/0. To show that the limit does not exist.1. as shown in the following theorem. y) lim (x. y) → (0.b) lim g(x.b) lim f (x. y) − L| ≤ g(x.b) lim (x.y)→(a.y)→(a.0) lim xy does not exist x2 + y2 Note that we can not simply substitute (x.y)→(a. suppose that (x. y)] = k f (x. y) → (0. y) → (0. But if (x.b) f (x. y) = x2 1 xy = 2 = 2 + y2 2 2 x x +x which means that f (x.b) lim g(x.b) lim g(x. y) − L| ≤ g(x. y) g(x. y) = xy x0 = 2 =0 x2 + y2 x + 02 along that path (since x > 0 in the denominator). and that k is lim lim [ f (x. so that y = 0 along that path. y) and if (x. y) for all (x. y) “sufficiently close” to (a.b) lim (x. FUNCTIONS OF SEVERAL VARIABLES (x.y)→(a. (e) If | f (x. for x > 0. y) and (x. Hence the limit does not exist. Theorem 2. y) if (x.y)→(a. it suffices to have | f (x. y) 0 lim f (x. b) (but excluding (a.b) (d) (x.b) (x. Limits of real-valued multivariable functions obey the same algebraic rules as in the single-variable case.

b) function if it is continuous at every point in its domain D. y) is a continuous (x.8. 2. for all (x. y) =  y    2  x + y2 if (x. b). y) = (0.e.8 so that it is continuous on all of 2 . b) in D if lim f (x. y) = f (a. y) = 2 (i. y) = (0. Thus.y)→(0.y)→(a.0) 69 lim y4 = 0. y)). Show that (x. 0). there are no indeterminate forms for (x. y) with domain D in 2 is continuous at the point (a. b) a2 + b2 (0. x2 + y2 x + y2 Therefore lim y4 = 0. We say that f (x. A real-valued function f (x. 0) by Example 2.1 Functions of Two or Three Variables Example 2.y)→(a.2.b) b4 = f (a.2. b) for (a. and we see that lim f (x. y) on all of 2 as follows:  0  if (x. notice that y4 = y2 4 and so 0 ≤ y4 ≤ x2 + y2 4 for all (x.1(e). then f (x. In fact. y) (0.8. Unless indicated otherwise. 0) we have y4 (x2 + y2 )2 ≤ 2 = x2 + y2 → 0 as (x. y) (0.0) Continuity can be defined similarly as in the single-variable case. we can modify the function from Example 2. y) in any (x. Define a function f (x. So since (x.y)→(0. We will use Theorem 2. x2 + y2 Since substituting (x. we need an alternate method for evaluating this limit. 0). 0)    4 f (x. x2 + y2 (x. y) is well-defined for all (x. y) is continuous on all of . Definition 2. y). 0) into the function gives the indeterminate form 0/0. 0) Then f (x. y) → (0. First. But x2 + y2 4 = (x2 + y2 )2 . Example 2. you can assume that all the functions we deal with are continuous.y)→(0.0) lim f (x. y) = 0 = f (0.9.

y)→(1.y)→(0. state the domain and range of the given function.) 23.0) 2. 18. y) = x2 + y2 − 1 3. y) = x2 + y2 − 4 Exercises ©  ¨ For Exercises 1-6. Prove Theorem 2.0) lim e xy xy2 x2 + y4 xy2 x2 + y2 x2 − 2xy + y2 x−y (x2 + y2 ) cos cos 1 xy x2 − y2 (x. y. 17.) C 22. y) = f (y. and is used as a filter in image processing software to produce a “blurred” effect.0) lim (x.1(a) in the case of addition.y)→(0. lim (x. (x. 2. (Hint: Use Definition 2.1. x) for all (x. 21. 2 2 2 20. f (x.0) lim (x. FUNCTIONS OF SEVERAL VARIABLES  A 1. y) in (x. x) for all lim sin x2 + y2 x2 + y2 =1.y)→(0.y)→(0.y)→(0. 15.1(b). Suppose that f (x.0) lim lim x2 − y2 x−y 4 sin(xy) y x2 + y2 x y (x. is constant on the circle of radius r > 0 centered at the origin. 14. .0) lim 1 xy (x. f (x.y)→(0. y) = x2 1 + y2 x2 + 1 y (x − 1)(yz − 1) 5. y) ≤ f (y. 10. f (x. (Hint: You will need to use L’Hôpital’s Rule for single-variable limits.y)→(0. evaluate the given limit. f (x.70 CHAPTER 2. y) in 2 .0) x2 + y2 x2 − 2xy + y2 11.y)→(1.0) lim lim (x. (x. z) = For Exercises 7-18. Use the substitution r = x2 + y2 to show that (x. 16. y) = 4.0) (x. Show that f (x. z) = sin(xyz) 6. (x.y)→(0. Prove Theorem 2.y)→(0.0) lim cos(xy) 8. y.y)→(0.−1) x−y lim 13.0) lim B 1 19. f (x.1) (x. Show that f (x. 7. f (x. This function is called a Gaussian blur. for σ > 0. 9.y)→(0. y) = 2πσ2 e−(x +y )/2σ . 12.

the partial derivative of f (x. Likewise. Example 2. y) in the (positive) x or y direction. h→0 ∂y h (2. b + h) − f (a. y) = x2 y + y3 . y) with respect to y gives ∂f (x. y) for the function f (x.1 Recall that the derivative of a function f (x) can be interpreted as the rate of change of that function in the (positive) x direction. b) (a. y) with respect to x or y is the rate of change of f (x. y) as if it were a function of x alone. we can start to develop an idea of a derivative of a function of two or more variables. . y) and (x.2. b).2 Partial Derivatives 71 2. y) with respect to x can be calculated by treating the y variable as a constant. y) = 2xy ∂x and treating x as a constant and differentiating f (x. ∂y 1 It is not a Greek letter. and then simply differentiating f (x. using the usual rules from single-variable calculus. We will start with the notion of a partial derivative.3. b) = lim . b) be a point in D. Clairaut and L. is ∂y (2. b). y) with respect to y is obtained by treating the x variable as a constant and then differentiating f (x. to distinguish it from the letter d used for the “usual” derivative. y) with respect to x gives ∂f (x. The symbol was first used by the mathematicians A. Definition 2. denoted by defined as ∂f f (a. b) with respect to y. From the definitions above. we can see that the partial derivative of a function f (x.3) Note: The symbol ∂ is pronounced “del”. b) − f (a. and let (a. is defined as ∂x f (a + h. Find ∂f ∂f (x. y) = x2 + 3y2 . y) as if it were a function of y alone. y) be a real-valued function with domain D in 2 . respectively. b) with respect to x. Euler around 1740. b) ∂f (a. ∂x ∂y Solution: Treating y as a constant and differentiating f (x.2) ∂f (a. What this means is that the partial derivative of a function f (x. b) = lim h→0 ∂x h and the partial derivative of f at (a.2 Partial Derivatives Now that we have an idea of what functions of several variables are. ∂f denoted by (a. Let f (x. Then the partial derivative of f at (a. and what a limit of such a function is.10.

we can take their partial ∂x ∂y derivatives with respect to x and y. 2 ∂2 f ∂ f ∂ f ∂2 f ∂2 f ∂2 f . 2. 2. and for the function f (x. y). This yields the higher-order partial derivatives: Since both ∂2 f ∂ ∂f = 2 ∂x ∂x ∂x 2f ∂ ∂f ∂ = ∂y ∂x ∂y ∂x ∂3 f ∂ ∂2 f = ∂x3 ∂x ∂x2 3f ∂ ∂2 f ∂ = ∂y ∂x2 ∂y ∂x2 ∂ ∂2 f ∂3 f = ∂y2 ∂x ∂y ∂y ∂x ∂ ∂2 f ∂3 f = ∂x ∂y ∂x ∂x ∂y ∂x . y) = 2 ∂x ∂y x +1 Solution: Treating y as a constant and differentiating f (x. Find ∂f ∂f sin(xy2 ) . y) with respect to y gives 2xy cos(xy2 ) ∂f = . . y) with respect to x gives ∂f (x2 + 1)(y2 cos(xy2 )) − (2x) sin(xy2 ) = ∂x (x2 + 1)2 and treating x as a constant and differentiating f (x. ∂y x2 + 1 ∂f ∂f and are themselves functions of x and y. y) and (x. y) = e x y + xy3 . ∂ ∂f ∂2 f = 2 ∂y ∂y ∂y 2f ∂ ∂ ∂f = ∂x ∂y ∂x ∂y ∂ ∂2 f ∂3 f = ∂y ∂y2 ∂y3 3f ∂ ∂2 f ∂ = ∂x ∂y2 ∂x ∂y2 ∂ ∂2 f ∂3 f = ∂x2 ∂y ∂x ∂x ∂y ∂3 f ∂ ∂2 f = ∂y ∂x ∂y ∂y ∂x ∂y Example 2. Find the partial derivatives function f (x.72 CHAPTER 2. . .12. ∂x ∂y ∂x ∂y We will often simply write Example 2. FUNCTIONS OF SEVERAL VARIABLES ∂f ∂f ∂f ∂f and instead of (x. and for the ∂x ∂y ∂x ∂y ∂y ∂x ∂x ∂y .11.

fx (x. we have 2 ∂f = 2xye x y + y3 ∂x 2 ∂ ∂2 f = (2xye x y + y3 ) 2 ∂x ∂x 2 ∂f = x2 e x y + 3xy2 ∂y 2 ∂ ∂2 f = (x2 e x y + 3xy2 ) 2 ∂y ∂y 73 = 2ye x y + 4x2 y2 e x 2 2y = x4 e x y + 6xy ∂2 f ∂ 2 x2 y = (x e + 3xy2 ) ∂x ∂y ∂x = 2xe x y + 2x3 ye x y + 3y2 2 2 2 2 ∂2 f ∂ = (2xye x y + y3 ) ∂y ∂x ∂y = 2xe x y + 2x3 ye x y + 3y2 2 2 Higher-order partial derivatives that are taken with respect to different variables. 214-216 in T AYLOR and M ANN for a proof. all orders. so you can assume in the remainder of the text that ∂2 f ∂2 f = for all (x. y) See pp. Dyy (x. Notice in the above example that whenever are continuous at a point (a. y) D22 (x. y) . are called mixed partial derivatives. y) . y) . ∂y ∂x ∂x ∂y In other words.2 Partial Derivatives Solution: Proceeding as before. y) . b). y) . f22 (x. it doesn’t matter in which order you take partial derivatives. y) . y) . y) D12 (x. ∂2 f and ∂x ∂y It turns that this will usually be the case. y) . y) D21 (x. f1 (x. fxx (x. fxy (x. y) . Specifically. y) D2 (x. Dy (x. This applies even to mixed partial derivatives of order 3 or higher. f12 (x. then they are equal at that 2 All the functions we will deal with will have continuous partial derivatives of point. fyx (x. f11 (x. y) in the domain of f . Dxx (x. y) . The notation for partial derivatives varies. y) . f2 (x. Dyx (x. D1 (x. y) . y) . y) . y) . y) . All of the following are equivalent: ∂f : ∂x ∂f : ∂y ∂2 f : ∂x2 ∂2 f : ∂y2 ∂2 f : ∂y ∂x ∂2 f : ∂x ∂y 2 ∂2 f ∂y ∂x ∂2 f both ∂y ∂x = ∂2 f ∂x ∂y . y) D11 (x. fy (x. f21 (x. Dxy (x. fyy (x. . Dx (x. ∂2 f ∂2 f such as ∂y ∂x and ∂x ∂y .2. y) . y) .

Let u and v be twice-differentiable functions of a single variable. y) = tan(x + y) ∂2 f ∂2 f . y) = x2 + y2 19. y) = xy + 1 x+y x2 + y + 4 2 +y2 ) 13. y) = sin(x + y) 12. f (x. f (x. f (x. find 17. f (x. y) = x2 + y + 4 ∂f ∂x Exercises ©  and ∂f ∂y . y) = x + 2y 10. f (x. f (x. y) = x2 + y + 4 and ∂2 f ∂y ∂x (use Exercises 1-8. y) = 11. y) = x4 25. f (x. y) = sin(xy) For Exercises 17-26. f (x. f (x. f (x. find 1. 18. y) = cos(x + y) 4. f (x. y) = x4 9. . y) = ln(xy) 16. f (x. f (x. f (x. y) = x2 + y2 3 6. 15). Show that the function f (x. f (x. y) = cos(x + y) 20. y) = sin(xy) B 27. f (x. f (x. y) = x + 2y 26. ¨ 2. f (x. 14. f (x. y) = x+1 y+1 5. 28.74 CHAPTER 2. y) = e xy + xy 23. y) = x2 − y2 + 6xy + 4x − 8y + 2 24. f (x. y) = e−(x 14. See Ch. ∂x2 ∂y2 The wave equation is an example of a partial differential equation. and let c 0 be a constant. f (x. Show that f (x. y) = u(x + cy) + v(x − cy) is a solution of the general one-dimensional wave equation3 ∂2 f 1 ∂2 f − 2 2 =0. it turns out that any solution must be of this form. f (x. f (x. y) = x2 − y2 + 6xy + 4x − 8y + 2 8. y) = e xy + xy 7. ∂x2 c ∂y 3 Conversely. y) = ln(xy) 22. 1 in W EINBERGER. f (x. FUNCTIONS OF SEVERAL VARIABLES  A For Exercises 1-16. y) = x+1 y+1 21. y) = x2 + y2 3. f (x. y) = sin(x + y) + cos(x − y) satisfies the wave equation ∂2 f ∂2 f − =0. f (x. ∂x2 ∂y2 15.

2. b). f (a. y) in the x and y directions described in Figure 2.1 are contained in . b) D y x (b) Tangent line Ly in the plane x = a Figure 2.3 Tangent Plane to a Surface In the previous section we mentioned that the partial derivatives ∂ f and ∂ f can be ∂x ∂y thought of as the rate of change of a function z = f (x. y) in the plane y = b is a curve in 3 through the point (a. Recall that the derivative dx of a function y = f (x) has a geometric meaning. and the slope of the tangent line L x to that curve at that point is ∂ f (a. the trace of the surface described by z = f (x. There is a similar geometric meaning to the partial derivatives ∂ f and ∂ f of ∂x ∂y a function z = f (x. y). b. b)). The formal definition mimics the intuitive notion of a tangent line to a curve. Definition 2. you might expect that partial derivatives can be used to define a tangent plane to the graph of a surface z = f (x. b)) slope = ∂f ∂x (a. First. c) be a point on S . ∂ f (a. The intuitive idea is that a tangent plane “just touches” a surface at a point. y) (a.1 Partial derivatives as slopes dy Since the derivative dx of a function y = f (x) is used to find the tangent line to the graph of f (which is a curve in 2 ). z) represent a generic point on the surface S . then the two tangent lines to the surface z = f (x. b) in the domain D of f (x. b.3 Tangent Plane to a Surface 75 2. we need a definition of a tangent plane. and let Q = (x.3. then we call T the tangent plane to S at P. Note that since two lines in 3 determine a plane. If the (acute) angle between the vector − − → PQ and the plane T approaches zero as the point Q approaches P along the surface S . f (a. b) z slope = ∂f ∂y (a. y) be the equation of a surface S in 3 . and let P = (a. b. z z = f (x. b) is the slope of the tangent line Ly to the trace of the surface z = f (x.3.3.1). This indeed turns out to be the case. f (a. Let T be a plane which contains the point P. y): given a point (a. y) (a. b. y). ∂x Similarly. Let z = f (x. namely as the slope of the tangent line to the graph of f at the point (x.4. y) in the positive x and y direcdy tions. b) z = f (x. y. y) ∂y in the plane x = a (see Figure 2. respectively. b) D (a) Tangent line L x in the plane y = b y 0 a x (a. f (x)) in 2 . b)) Ly Lx b 0 (a.

b)) is parallel to Ly . z vx = (1. Luckily. y) will exist at the point (a. then all we need are vectors vx and vy that are parallel to Lx and Ly .b) ∂f ∂x (a.2 Tangent plane Since the slope of Lx is ∂ f (a. or it may not have a tangent line at all at that point. C) is a normal vector to the plane T . Hence.3 = − ∂ f (a. those conditions will always hold. (2. The existence of those two tangent lines does not by itself guarantee the existence of the tangent plane. b)) is ∂x ∂x parallel to Lx (since vx lies in the xz-plane and lies in a line with slope ∂x 1 = ∂ f (a.3. 0. y) at a point (a.3. b) ∂f where n = (A. then the vector vx = (1. f (a. and then let n = vx × vy . b) then the ∂x ∂y tangent plane to the surface z = f (x. b) = 0 . y) at the point (a. B. b) i ∂x − ∂f ∂y (a. the resulting curve in that plane may have a tangent line which is not in the plane determined by the other two tangent lines. Since T contains the lines Lx and Ly .3. b) (y − b) − z + f (a. the vector ∂y i j n = vx × vy = 1 0 0 1 k ∂f ∂x (a. § 6. b) (x − a) − ∂f ∂y (a.2). b. respectively (as in Figure y 2.5) Multiplying both sides by −1. Similarly. b)).3. b). b)) L in the planes y = b and x = a. we have the following result: The equation of the tangent plane to the surface z = f (x. b)) ∂x (a. b)) = 0 (2. respectively. b) j +k is normal to the plane T . f (a. f (a. b. b) ∂f ∂y (a.4) 0 x Figure 2. FUNCTIONS OF SEVERAL VARIABLES the tangent plane at that point. b. ∂ f (a. it turns out4 that if ∂ f and ∂ f exist in a region around a point (a. b) and are continuous at (a. .76 CHAPTER 2. b)) is ∂f ∂f (2.4. See Figure 2. 0.3). the vector ∂x vy = (0. It is possible that if we take the trace of the surface in the plane x − y = 0 (which makes a 45◦ angle with the positive x-axis). Thus the equation of T is − ∂f ∂x (a. z Suppose that we want an equation of the tangent plane T to the surface z = f (x. and suppose that the conditions for T to exist do Lx hold. b). 1. In this text.6) ∂x (a. b) (y − b) + z − f (a. if the tangent plane exists at that point. ∂ f (a. y) and Ly be the tangent lines to the traces of the surface (a. Then the equation for T is T y A(x − a) + B(y − b) + C(z − f (a. b) = 0 4 See T AYLOR and M ANN. b)). f (a. b) (x − a) + ∂y (a. b. ∂ f (a. Let Lx z = f (x. b) x 0 1 Figure 2.

y) = x2 + y2 . Find the equation of the tangent plane to the surface x2 + y2 + z2 = 9 at the point (2. 1. 2. 2. we have of the tangent plane at the point (1. P = (−1. 4. 2. and 2(2)(x − 2) + 2(2)(y − 2) + 2(−1)(z + 1) = 0 . 2. 1. b. In a similar fashion. P = (1. f (x. y) = x2 + y2 .3 Tangent Plane to a Surface 77 Example 2. f (x. 4) 2. 2 √ 11 3 9. y. 1) 6. 0) .2. Example 2. P = (1. b. −1). z) = f (x. x2 + y2 − z2 = 0. P = (2. z) = 0. c) is given by the equation ∂F ∂x (a. P = ( 3. −1. y. it can be shown that if a surface is defined implicitly by an equation of the form F(x. 5) is ∂f ∂x = 2x and ∂f ∂y = 2y. P = (0. 5) 8. x2 4 + y2 9 + z2 16 = 1. 1. 1. 5).7) where F(x. 4. find the equation of the tangent plane to the surface z = f (x. 3) √ 10. y) = x2 y. ∂F ∂y = 2y. so the equation of the tangent plane at (2. P = 1. (2. Solution: For the function F(x. 2) 3. P = (3. x2 + y2 = 4. y) − z. b. 7.13. b.7) Note that formula (2. f (x. f (x. y. c) (x − a) + ∂F ∂y (a. 0. P = (3. Solution: For the function f (x. y) = x + 2y. or 2x + 4y − z − 5 = 0 . y) = x2 + y3 . c) (y − b) + ∂F ∂z (a. so the equation 2(1)(x − 1) + 2(2)(y − 2) − z + 5 = 0 . c) (z − c) = 0 . then the tangent plane to the surface at a point (a. y) = xy. 0. x2 + y2 + z2 = 9. y) = xey . 5) For Exercises 7-10. 1) 5. z) = x2 + y2 + z2 − 9. 1.14. 2. f (x. find the equation of the tangent plane to the given surface at the point P.6) is the special case of formula (2. −1) 4. or ¨ A Exercises ©  For Exercises 1-6. y) at the point P. we have ∂F ∂z = 2z. P = (1. −1) is 2x + 2y − z − 9 = 0 . Find the equation of the tangent plane to the surface z = x2 + y2 at the point (1. f (x.  ∂F ∂x = 2x.

b) = v1 ∂f ∂f (a. y). Then the directional derivative of f at (a. b) be a point in D. b) = lim f ((a. b) = ∂ f (a. v2 ). respectively. is defined as Dv f (a. FUNCTIONS OF SEVERAL VARIABLES 2. as that represents a “standard” vector for a given direction. b). b + hv2 ) − f (a. respectively. and let (a. Let (a. b) + v2 (a. Similarly. b) h (2. which the reader should be used to by now. 0) and j = (0. But this is just the usual idea of identifying vectors with their terminal points. So since ∂y ∂y i = (1.4 Directional Derivatives and the Gradient For a function z = f (x. ∂x ∂y and let v = (v1 . b) be a point in D. Since there are many vectors with the same ∂x ∂y direction. as we noted earlier. If f (x. b). b). h (2.9) h→0 From this we can immediately recognize that the partial derivatives ∂ f and ∂ f are ∂x ∂y special cases of the directional derivative with v = i = (1. y) be a real-valued function with domain D in 2 . 0) and v = j = (0. Then Dv f (a. b) as a vector. v2 ) be a unit vector in 2 . 1). ∂x ∂y (2. That is.8) h→0 Notice in the definition that we seem to be treating the point (a. which is true since Dj f = ∂ f . 0) then the above formula reduces to Dv f (a.2. ∂ f = Di f and ∂ f = Dj f . 1) are the only unit vectors in 2 with a zero component. What about other directions? It turns out that we can find the rate of change in any direction using a more general type of derivative called a directional derivative. If we were to write the vector v as v = (v1 . ∂x which we know is true since Di f = ∂ f . b) . b) + hv) − f (a.10) Proof: Note that if v = i = (1. then we . for v = j = (0. 1) ∂x the formula reduces to Dv f (a. b) . b) = ∂ f (a. Let v be a unit vector in 2 . b) in the direction of v. we use a unit vector in the definition. we learned that the partial derivatives ∂ f and ∂ f represent ∂x ∂y the (instantaneous) rate of change of f in the positive x and y directions. denoted by Dv f (a. since we are adding the vector hv to it. then Dv f (a. y) has continuous partial derivatives ∂ f and ∂ f (which will always be the case ∂x ∂y in this text). then there is a simple formula for the directional derivative: Theorem 2.5.78 CHAPTER 2. b) = lim f (a + hv1 . Definition 2. Let f (x. Let f (x. y) be a real-valued function with domain D in 2 such that the partial derivatives ∂ f and ∂ f exist and are continuous in D.

b + αhv2 ) = g ′ (b + αhv2 ) = = ∂y b + hv2 − b hv2 and so f (a + hv1 . b) hv2 ∂y (a + hv1 . b + hv2 ) − f (a. there exists a number 0 < β < 1 such that f (a + hv1 . b) . b) − f (a. b). we have ∂f ∂f f (a + hv1 . (2. b) (a + hv1 . b + αhv2 ) + hv1 ∂x (a + βhv1 . b) ∂y ∂x ∂f (a + βhv1 . b) . ∂x so by formula (2. b) = v1 (a.11) Since h 0 and v2 0. b) + f (a + hv1 . b) = f (a + hv1 . The second vector has a special name: . b] if one of h or v2 is negative) to find a number 0 < α < 1 such that ∂f g(b + hv2 ) − g(b) f (a + hv1 . b) + v2 (a. b + hv2 ) − f (a + hv1 . Note that Dv f (a. b) by the continuity of ∂y ∂x ∂x ∂y ∂f ∂f Dv f (a. y) on the interval [b. b) = v · ∂f ∂f ∂x (a. b + hv2 ) − f (a. v2 ) with v1 So fix such a vector v and fix a number h 0. b + hv2 ) − f (a. b + hv2 ) − f (a + hv1 . So since the function f (a + hv1 . b) = lim f (a + hv1 . b) h ∂f ∂f = lim v2 (a + hv1 . b + hv2 ] (or [b + hv2 .2. b) ∂x ∂y h→0 QED after reversing the order of summation. b + αhv2 ) + v1 (a + βhv1 . b + αhv2 ) + v1 (a + βhv1 . f (a + hv1 .4 Directional Derivatives and the Gradient need only show the formula holds for unit vectors v = (v1 .9) we have Dv f (a. ∂y (a. y) is a real-valued function of y (since a + hv1 is a fixed number). b) . Then 0 and v2 79 0. then the Mean Value Theorem from single-variable calculus can be applied to the function g(y) = f (a + hv1 . by equation (2. b) = h h ∂f ∂f = v2 (a + hv1 . b + hv2 ) − f (a + hv1 . b) − f (a. b + αhv2 ) . so = v2 (a. b) h→0 ∂y ∂x ∂f ∂f ∂f ∂f and . then hv2 0 and thus any number c between b and b + hv2 can be written as c = b + αhv2 for some number 0 < α < 1. b) = hv2 ∂f (a + hv1 .11). b) + v1 (a. b) = hv1 Thus. ∂y By a similar argument.

15. y) whose partial derivatives ∂ f and ∂ f exist and are ∂x ∂y continuous is called continuously differentiable. For a real-valued function f (x. y) = xy2 + x3 y at the point (1. Assume that f (x. y). 2) = 1 1 √ . FUNCTIONS OF SEVERAL VARIABLES Definition 2. .80 CHAPTER 2. The symbol ∇ is pronounced “del”. is the vector ∂f ∂f (2.3. so Dv f (1. the gradient is the vector ∇f = ∂f ∂f ∂f . Solution: We see that ∇ f = (y2 + 3x2 y. . y) is such a function and that ∇ f 0. y) = c x 0 Figure 2. ∂x ∂y ∂z (2.6. y v ∇f f (x. Dv f = v · ∇ f Example 2. For a real-valued function f (x. z). 2xy + x3 ).4. 2(1)(2) + 13 ) = 15 √ 2 A real-valued function z = f (x.13) in 3.12) ∇f = . 2) = v · ∇ f (1. √ 2 2 · (22 + 3(1)2 (2). √ 2 2 .5 Corollary 2. Find the directional derivative of f (x.4. y. Let c be a real number in the range of f and let v be a unit vector in 2 which is tangent to the level curve f (x.1 5 Sometimes the notation grad( f ) is used instead of ∇ f . the gradient of f . 2) in the direction of v = 1 1 √ .1). denoted by ∇ f . ∂x ∂y in 2. y) = c (see Figure 2.

16. 1. √ 5 5 and decreases the fastest in the direction of . z) = e−x + e−2y +e4z . . i.4 for functions of two variables.e. 2) = (10. Let f (x. the value of the function f increases the fastest in the direction of ∇ f (since θ = 0◦ in that case). In other words. So since v = 1 then Dv f = ∇ f cos θ. So since ∇ f 0 then Dv f = 0 ⇒ cos θ = 0 ⇒ θ = 90◦ . Dv f = 0. At a fixed point (x. 2)? In which direction does it decrease the fastest? that direction is v = 2 1 √ . the directional derivative in the three-dimensional case can also be defined by the formula Dv f = v · ∇ f . while the smallest value occurs when cos θ = −1 (θ = 180◦ ). y. then ∇ f (1. y) decreases the fastest in the direction of −∇ f . 1) = (e−1 . so since v is a tangent vector to this curve. then the temperature will decrease the fastest in the direction of −∇ f (1. (b) The value of f (x. A unit vector in = 1 2 √ . (c) The value of f (x. 2e−2 . y) = c. −4e4 ). The largest value that Dv f can take is when cos θ = 1 (θ = 0◦ ). 4e4z ). We have thus proved the following theorem: Theorem 2. But we know that Dv f = v · ∇ f = v ∇ f cos θ. a similar argument can be used to show that it also applies to functions of three or more variables. −2e−2y . y) be a continuously differentiable real-valued function.17. Example 2. Likewise.4. f increases the fastest in the direction of −2 −1 √ . then the rate of change of f in the direction of v is 0. In which direction does the function f (x. 1. In general. 5) ∇f ∇f 0. Then: (a) The gradient ∇ f is normal to any level curve f (x. where x. 1) will the temperature decrease the fastest? Solution: Since ∇ f = (−e−x .4 Directional Derivatives and the Gradient 81 The value of f (x. √ 5 5 Solution: Since ∇ f = (y2 + 3x2 y. In other words. Though we proved Theorem 2. which means that ∇ f is normal to the level curve. The temperature T of a solid is given by the function T (x.2. Example 2. and the value of Dv f then varies as θ varies. y. for any unit vector v in 2 . with ∇ f 0. 2xy + x3 ). In which direction from the point (1. where θ is the angle between v and ∇ f . √ 5 5 . Thus. and the value of f decreases the fastest in the direction of −∇ f (since θ = 180◦ in that case). we still have Dv f = ∇ f cos θ. where θ is the angle between v and ∇ f . z are space coordinates relative to the center of the solid. ∇ f ⊥ v. y) the length ∇ f is fixed. y) increases the fastest in the direction of ∇ f . y) is constant along a level curve. y) = xy2 + x3 y increase the fastest from the point (1.

f (x. f (x. 16. f (x. √ . let c be a constant. ∇(c f ) = c ∇ f 21. f (x. f (x. y) = x2 + y2 + 4 Exercises ©  ¨ For Exercises 1-10. 3). P = (1. 1. Repeat Example 2. y) = x2 + y2 + 4. y. r . find the directional derivative of f at the point P in the direction of v = 1 1 √ . B For Exercises 19-26. Show that: 19. √ 3 3 3 . y) and g(x. y. and that ∇(r2 ) = 2 r. f (x. y) = x2 ey . y) = x2 ey 6. 18. f (x. ∇( f + g) = ∇ f + ∇g 22. P = (1. 12. y. z) = sin(xyz) 9. f (x. let f (x. 1. f (x. y. f (x. 2. 2).82 CHAPTER 2. Dv (c f ) = c Dv f 26. √ 2 2 . y) = x2 + y2 − 1 3. y) in 20. y) = x2 + y2 − 1. z) = x2 eyz 10. f (x. 1) 15. 1) For Exercises 15-16. y. y) = each point (x.17 at the point (3. f (x. 1) + y2 11. FUNCTIONS OF SEVERAL VARIABLES  A 1. y) = ln(xy) 7. z) = x2 + y2 + z2 5. 0). compute the gradient ∇ f . y. 1) 17. D−v f = −Dv f 25. Dv ( f + g) = Dv f + Dv g 27. P = (1. 1) 13. 1. f (x. y) (0. f (x. y) be continuously differentiable real-valued functions. z) = x2 + y2 + z2 For Exercises 11-14. P = (1.16 at the point (2. f (x. z) = x2 eyz . y) g2 0 24. Dv ( f g) = f Dv g + g Dv f x2 + y2 is the length of the position vector r = x i + y j for 2 . y) = 2x + 5y 8. ∇( f g) = f ∇g + g ∇ f 23. z) = sin(xyz). P = (1. Repeat Example 2. P = (1. find the directional derivative of f at the point P in the direction of v = 1 1 1 √ . 1) 14. The function r(x. y) = x2 1 . f (x. ∇( f /g) = g ∇ f − f ∇g if g(x. Show that ∇r = 1 r when (x. and let v be a unit vector in 2 . y) = x2 1 + y2 4.

b) = 0 is not always sufficient to guarantee that a critical point is a local maximum or minimum. y) inside some disk of positive radius centered at (a. b). the single-variable function g(x) = f (x. y). y) ≥ f (a. Definition 2. y) ≤ f (a. b) = 0. and that the first-order partial derivatives of f exist at (a.5 Maxima and Minima The gradient can be used to find extreme points of real-valued functions of several variables. b)). b) for all (x. b). f (a. b) = 0 is called a critical point for the function f (x. b). the necessary condition that ∇ f (a. We know that f (a. b) = 0. We will consider only functions of two variables. b) where ∇ f (a. 0)). y) to have a local maximum or minimum at (a. Since g ′ (x) = ∂ f (x. ∂x and ∂ f = x = 0 ⇒ x = 0. y) in the domain of f . b). along the path y = x . b) for all (x. then f has a global maximum at (a. b) has a local maximum at x = a. y) inside some disk of positive radius centered at (a. then f has a global minimum at (a. and let (a. y) for which (x − a)2 + (y − b)2 < r2 . b) is a local maximum point for f (x. that is. Then a necessary condition for f (x. so (0. f (a. 0) is the only critical point. Let f (x. y) ≤ f (a. b) for all (x. y) = 0 simultaneously for (x. then ∂ f (a. y) = xy > 0 = f (0. 0)) and different signs (so that f (x. If f (x. 0) contains points (x.2. we say that f has a local minimum at (a. y) = xy has a critical point at (0.7. b).5 Maxima and Minima 83 2. b). b) is the largest value of f in the x direction (around the point (a. b) if f (x. y). b) is that ∇ f (a. Suppose that (a. functions of three or more variables require methods using linear algebra. Note: Theorem 2. So we know that g ′ (a) = 0. b) if f (x. The function f (x. We say that f has a local maximum at (a. y) goes in all directions from the point (a. that is. y) be a real-valued function such that both ∂ f (a. In fact. So given a function f (x. ∂x ∂x Similarly. b) ∂x ∂y exist. 0): ∂ f = y = 0 ⇒ y = 0. y) in the domain of f .e. y) be a real-valued function. Likewise. to find the critical points of f you have to solve the equations ∂f ∂f ∂x (x. i. Similar to the single-variable case. in some sufficiently small disk centered at (a. Let f (x. y) ≥ f (a.5 can be extended to apply to functions of three or more variables. y). In particular. b) for all (x. 0) since any disk around (0. b). b) be a point in the domain of f . b) and ∂ f (a. b). b) is the largest value of f near (a. there is some sufficiently small r > 0 such that f (x. y). ∂y We thus have the following theorem: Theorem 2. y) where the values of x and y have the same sign (so that f (x. b) = 0. y) as (x. If f (x. y) = 0 and ∂y (x. b) for all (x. b) is the largest value of f (x.5. b) in the y direction and so ∂ f (a. A point (a. points where the function has a local maximum or local minimum. Example 2. But clearly f does not have a ∂y local maximum or minimum at (0. y) ≤ f (a. y) = xy < 0 = f (0.18.

6. 0) The following theorem gives sufficient conditions for a critical point to be a local maximum or minimum of a smooth function (i. Let f (x. 100 50 0 z -50 -10 -100-10 -5 -5 y 0 0 5 5 10 10 x Figure 2. a function whose partial derivatives of all orders exist and are continuous). . while along the path y = −x we have f (x. which has a local minimum at (0. with a critical point at (a. ∇ f (a.5.1 f (x. The graph of f (x. f (x. 0) is an example of a saddle point. b) ∂x2 ∂2 f (a.6 Theorem 2. y) = xy. b) − ∂y ∂x ∂x2 ∂y 2 > 0.e. b) (a. y) be a smooth real-valued function.84 CHAPTER 2. then f has a local minimum at (a. b) ∂x2 ∂2 f ∂2 f ∂2 f (a. which has a local maximum at (0. 0). So (0. which we will not prove here. then f has neither a local minimum nor a local maximum at (a.1.e.6. b) (c) if D < 0. Define D= Then (a) if D > 0 and (b) if D > 0 and ∂2 f (a. then the test fails. b) < 0. then f has a local maximum at (a. which is a hyperbolic paraboloid. b) (i. y) = −x2 . b) = 0). i. saddle point at (0. FUNCTIONS OF SEVERAL VARIABLES in 2 . § 7. 0). y) is shown in Figure 2. it is a local maximum in one direction and a local minimum in another direction.5.e. b) (d) if D = 0. y) = x2 . b) 2 (a. 6 See T AYLOR and M ANN.

y) is smooth means that ∂2 f (a. Solution: First find the critical points. y) = (2. y) = x2 + xy + y2 − 3x. b) ∂x ∂y ∂2 f (a. −1) is the only critical point.19. i. −1) ∂x2 ∂2 f =2. where ∇ f = 0. b) ∂ 2 (a.6. y) = xy − x3 − y2 . .5 Maxima and Minima 85 If condition (c) holds. then (a. then the critical points (x. −1) is a local minimum. b) 2 D= since ∂2 f ∂x2 ∂2 f ∂y ∂x = ∂2 f ∂x ∂y . Thus. i. Find all local maxima and minima of f (x. Note that the assumption that f (x. Solution: First find the critical points. ∂x2 and so D= and ∂2 f (2.e. −1). −1) − (2. b) ∂y ∂x ∂2 f (a. ∂y2 ∂2 f =1 ∂y ∂x 2 ∂2 f ∂2 f ∂2 f (2. b) ∂x2 2f ∂2 f (a. b) and (a. where ∇ f = 0. we need the second-order partial derivatives: ∂2 f =2. Since ∂f = 2x + y − 3 ∂x and ∂f = x + 2y ∂y Example 2. −1) 2 (2. and so (a. b) ∂y2 theorem one can replace if desired. This means that in parts (a) and (b) of the by ∂2 f (a. b) is a saddle point.e.20. Find all local maxima and minima of f (x. if D > 0 then ∂2 f (a. b) ∂x2 ∂y > 0. b) have the same sign. b) ∂x2 ∂2 f (a. So (2.2. y) are the common solutions of the equations 2x + y − 3 = 0 x + 2y =0 which has the unique solution (x. b) ∂y2 = D+ ∂2 f ∂y ∂x (a. To use Theorem 2. Since ∂f = y − 3x2 ∂x and ∂f = x − 2y ∂y Example 2. −1) ∂y ∂x ∂x2 ∂y = (2)(2) − 12 = 3 > 0 = 2 > 0. (2. ∂2 f ∂y2 Also.

0) − ∂y ∂x ∂x2 ∂y = (−6(0))(−2) − 12 = −1 < 0 and thus (0. 12 is a local maximum. which has the solutions x = 0 and x = 6 . ∂x2 ∂2 f =8.6. ∂y2 2 ∂2 f =1 ∂y ∂x ∂2 f ∂2 f ∂2 f (0. 12 2 = (−6 1 )(−2) − 12 = 1 > 0 6 = −1 < 0. To use Theorem 2. To use Theorem 2. we need the second-order partial derivatives: ∂2 f = 12(x − 2)2 + 2 .21. y) are the common solutions of the equations y − 3x2 = 0 x − 2y = 0 The first equation yields y = 3x2 . 0) is a saddle point. we need the second-order partial derivatives: ∂2 f = −6x . y) = 6 . Solution: First find the critical points. y) = (x − 2)4 + (x − 2y)2 . 0) (0. 1) is the only critical point.86 CHAPTER 2. substituting that into the first equation yields 4(2y − 2)3 = 0. Also. (2. ∂x2 6 12 ∂2 f ∂x2 1 1 6 . 0) and (x. y) are the common solutions of the equations 4(x − 2)3 + 2(x − 2y) = 0 −4(x − 2y) = 0 The second equation yields x = 2y. 1 1 6 . 12 ∂2 f ∂y2 1 1 6 . FUNCTIONS OF SEVERAL VARIABLES then the critical points (x.e. D= and ∂2 f 1 1 . 12 − ∂2 f ∂y ∂x 1 1 6 . where ∇ f = 0. ∂x2 So D= ∂2 f = −2 . y) = (0. ∂f = −4(x − 2y) ∂y then the critical points (x. 12 . Thus. i. 6 6 1 1 So the critical points are (x.6. ∂y2 ∂2 f = −4 ∂y ∂x . and so x = 2(1) = 2. Thus. 0) 2 (0. Since ∂f = 4(x − 2)3 + 2(x − 2y) ∂x and Example 2. substituting that into the second equation yields 1 x − 6x2 = 0. which has the solution y = 1. Find all local maxima and minima of f (x. So x = 0 ⇒ y = 3(0) = 0 and 2 1 x = 1 ⇒ y = 3 1 = 12 .

1) ∂y ∂x ∂x2 ∂y 2 = (2)(8) − (−4)2 = 0 and so the test fails. y). y) ≥ 0 for all (x. y) ≥ 0 = f (2. where ∇ f = 0. we see that f (x. the points (x. But we also see that f (2. 1) is in fact a global minimum for f . y) in 2 . In our case. y) on the unit circle x2 + y2 = 1 are local maximum points for f . we have D = 4 > 0 and ∂ 2 (0. If we switch to using polar coordinates (r.e. so the Second Derivative Test from single-variable calculus says that r = 1 is a local maximum. we need the second-order partial derivatives: 2 2 ∂2 f = 2[1 − (x2 + y2 ) − 2x2 − 2x2 (1 − (x2 + y2 ))]e−(x +y ) 2 ∂x 2 2 ∂2 f = 2[1 − (x2 + y2 ) − 2y2 − 2y2 (1 − (x2 + y2 ))]e−(x +y ) 2 ∂y 2 2 ∂2 f = −4xy[2 − (x2 + y2 )]e−(x +y ) ∂y ∂x f At (0. 0). 1) − (2.5 Maxima and Minima So D= 87 ∂2 f ∂2 f ∂2 f (2. Thus. y) as a function g(r) of the variable r alone: g(r) = r2 e−r .6. it looks like we might have a local maximum for (x. 1) for all (x. y) on the unit circle x2 + y2 = 1. y) on the unit circle x2 + y2 = 1.22. then 2 we see that we can write f (x. Since 2 2 ∂f = 2x(1 − (x2 + y2 ))e−(x +y ) ∂x 2 2 ∂f = 2y(1 − (x2 + y2 ))e−(x +y ) ∂y 2 +y2 ) . θ) instead of (x. y). 2 Then g ′ (r) = 2r(1 − r2 )e−r . To use Theorem 2. However. Find all local maxima and minima of f (x.5. But r = 1 corresponds to the unit circle x2 + y2 = 1. y) = (x2 + y2 )e−(x Solution: First find the critical points. . y).2. as shown in Figure 2. 1) = 0. y) on the unit circle x2 + y2 = 1. and we can check that g ′′ (1) = −4e−1 < 0. If we look at the graph of f (x. and hence (2. 0) is a local minimum. 0) = 2 > 0. What can be done in this situation? Sometimes it is possible to examine the function to see directly the nature of a critical point. 0) and all points (x. Example 2. y) is the sum of fourth and second powers of numbers and hence must be nonnegative. ∂x for points (x. so (0. 1) 2 (2. where r2 = x2 + y2 .2. so it has a critical point at r = 1. i. since f (x. Thus f (x. we have 2 D = (−4x2 e−1 )(−4y2 e−1 ) − (−4xye−1 )2 = 0 and so the test fails. then the critical points are (0.

f (x. f (x. b) is a local maximum or local minimum point for a smooth function f (x.3 0.4 0. y) at the point (a.88 CHAPTER 2. y). f (x. f (x. y) = x2 + y2 Exercises ©  ¨ For Exercises 1-10.5.) 12.5. find all local maxima and minima of the function f (x. (Hint: Use the volume condition to write the surface area as a function of just two variables.2 f (x.05 0 -3 -2 -1 0 1 y 2 3 3 2 1 -3 -2 -1 0 x Figure 2. Prove that if (a. y) = x3 + 3x2 + y3 − 3y2 6. y) = (x2 + y2 )e−( x 2 +y2 )  A 1. y) = 2x3 − 6xy + y2 8. f (x. f (x. Find three positive numbers x. f (x.2 0. FUNCTIONS OF SEVERAL VARIABLES 0. b)) is parallel to the xy-plane. y) = x3 − 12x + y2 + 8y 4. . 2. f (x. y) = x3 − 3x + y3 − 3y 5.1 0. z whose sum is 10 such that x2 y2 z is a maximum. y) = x + 2y 10. b. then the tangent plane to the surface z = f (x. f (a. (Hint: Use Theorem 2. y) = 2x3 + 6xy + 3y2 7. y) = 4x2 − 4xy + 2y2 + 10x − 6y B 11. find the dimensions that will minimize the surface area.35 0.25 0. For a rectangular solid of volume 1000 cubic meters. f (x. f (x. y) = −4x2 + 4xy − 2y2 + 16x − 12y 9.15 z 0. y) = x3 − 3x + y2 3. y. y).) C 13.

7 For example. or complicated expressions involving trigonometric. y) be a smooth real-valued function. Cubic polynomial equations in one variable can be solved using Cardan’s formulas. y) could be any points in the domain of f . yn ) (2. The method we used required us to find the critical points of f . In a much. exponential. . complex number solutions. 8 There are also two nonreal.14) Then the sequence of points (xn . 1. which meant having to solve the equation ∇ f = 0. y) = ∂2 f ∂2 f ∂2 f (x. If the equations involve polynomials in x and y of degree three or higher. Let f (x. which you probably learned in single-variable calculus.6 Unconstrained Optimization: Numerical Methods 89 2. yn ) . though one that is not usually emphasized. If there are several n=1 critical points. the only choice may be to find a solution using some numerical method which gives a sequence of numbers which converge to the actual solution. especially since the only real solution8 turns out to be 28 + 1− situation such as this. . then you will have to try different initial points to find them. Newton’s method for solving equations f (x) = 0. yn )∞ converges to a critical point. y ) ∂y2 n n ∂f ∂y (xn . if one of the equations that had to be solved was x3 + 9x − 2 = 0 . y ) ∂x2 n n ∂f ∂x (xn . yn ) ∂f ∂y (xn . In this section we will describe another method of Newton for finding critical points of real-valued functions of two variables. yn ) ∂2 f ∂x ∂y (xn . you may have a hard time getting the exact solutions. yn ) ∂2 f (x . yn ) ∂f ∂x (xn . For n = 0. Newton’s algorithm: Pick an initial point (x0 . . where the points (x. yn+1 = yn − D(xn . which in general is a system of two equations in two unknowns (x and y). and define D(x. define: ∂2 f (x . in general this will not be the case. While this was relatively simple for the examples we did. could be impossible by elementary means. 7 . which are not quite as simple as the familiar quadratic formula. There are formulas for solving polynomial equations of degree 4. let alone two. Trial and error would not help 3 √ 3 √ 28 − 1. then solving even one such equation. 3. y). For example. See U SPENSKY for more details. . y) 2 (x. yn ) xn+1 = xn − D(xn . yn ) ∂2 f ∂x ∂y (xn .6 Unconstrained Optimization: Numerical Methods The types of problems that we solved in the previous section were examples of unconstrained optimization problems. This is also a problem for the equivalent method (the Second Derivative Test) in single-variable calculus. 2.2. but it can be proved that there is no general formula for solving equations for polynomials of degree five or higher. y0 ). y) − ∂y ∂x ∂x2 ∂y 2 . we tried to find local (and perhaps even global) maximum and minimum points of real-valued functions f (x. y) (x. That is. or logarithmic functions.

1. In each iteration the new point will be printed. though it may be hard to tell where the critical points are. we will let a computer do the computing.6. we will write a simple program. 0) as our initial point. Find all local maxima and minima of f (x. so that we can see if there is convergence. For this.90 CHAPTER 2. We need to pick an initial point (x0 . And we can see that D(0. y) = x3 − xy − x + xy3 − y4 . y0 ) for our algorithm. ∂y ∂x ∂x2 ∂y2 Notice that solving ∇ f = 0 would involve solving two third-degree polynomial equations in x and y. = 6xy − 12y2 . y) over a large region may help (see Figure 2. The full code is shown in Listing 2. which will take a given initial point as a parameter and then perform 100 iterations of Newton’s algorithm. and since the computations are quite tedious. Looking at the graph of z = f (x. Since it may take a large number of iterations of Newton’s algorithm to be sure that we are close enough to the actual critical point. using the Java programming language. .1 f (x. so take (0. FUNCTIONS OF SEVERAL VARIABLES Solution: First calculate the necessary partial derivatives: Example 2. = −x + 3xy2 − 4y3 ∂x ∂y ∂2 f ∂2 f ∂2 f = −1 + 3y2 = 6x .6. ∂f ∂f = 3x2 − y − 1 + y3 .23. y) = x3 − xy − x + xy3 − y4 for −20 ≤ x ≤ 20 and −20 ≤ y ≤ 20 Notice in the formulas (2.14) that we divide by D.1 below). so we should pick an initial point where D is not zero. which in this case can not be done easily. 0) = (0)(0) − (−1)2 = −1 0. z 50000 0 -50000 -100000 -150000 -200000 -250000 -300000 -350000-20 -20 -15 -10 -5 -15 -10 0 -5 5 0 5 10 10 15 15 20 20 x y Figure 2.

p r i n t l n ( " Error : D = 0 at i t e r a t i o n n = " + n ) . 2 ) . e x i t ( 0 ) . y ) − Math . y ) . // I n i t i a l y value System . y ) ∗ fyy ( x . System .java . parseDouble ( args [ 0 ] ) . parseDouble ( args [ 1 ] ) . 2 ) − 4∗Math . //Go through 100 i t e r a t i o n s o f Newton ’ s algorithm for ( int n=1. out . } //The mixed second p a r t i a l d e r i v a t i v e o f f wrt x and y : −1+3y^2 public s t a t i c double fxy ( double x . y = yn − ( fxx ( xn . double y ) { return 3∗Math . double y ) { return −1 + 3∗Math . yn ) ∗ f x ( xn . y )= x^3−xy−x+xy^3−y^4 public class newton { public s t a t i c void main ( String [ ] args ) { //Get the i n i t i a l p o i n t ( x . double yn = y . y ) as command− l i n e parameters double x = Double . yn ) ∗ f x ( xn . double y ) { return −x + 3∗x∗Math . //End the program } else { //C a l c u l a t e the new values f o r x and y x = xn − ( fyy ( xn . //The c u r r e n t x and y values i f (D == 0 ) { //We can not d i v i d e by 0 System .1 Program listing for newton. yn ) − fxy ( xn . } } } //Below are the p a r t s s p e c i f i c t o the f u n c t i o n f //The f i r s t p a r t i a l d e r i v a t i v e o f f wrt x : 3x^2−y−1+y^3 public s t a t i c double f x ( double x . pow ( y . pow ( y . 3 ) . yn ) ) / D. pow ( fxy ( x . } //The second p a r t i a l d e r i v a t i v e o f f wrt x : 6x public s t a t i c double fxx ( double x . yn ) ∗ f y ( xn . } //The second p a r t i a l d e r i v a t i v e o f f wrt y : 6xy −12y^2 public s t a t i c double fyy ( double x . 2 ) . yn ) − fxy ( xn . p r i n t l n ( "n = " + n + " : ( " + x + " . System . yn ) ) / D. // I n i t i a l x value double y = Double . 2 ) .6 Unconstrained Optimization: Numerical Methods 91 //Program t o f i n d the c r i t i c a l p o i n t s o f f ( x . p r i n t l n ( " I n i t i a l p o i n t : ( " + x + " . pow ( y . pow ( x . double y ) { return 6∗x∗y − 12∗Math . double y ) { return 6∗x . out . } } Listing 2.2. " + y + " ) " ) . n++) { double D = fxx ( x . yn ) ∗ f y ( xn . 2 ) − y − 1 + Math . pow ( y . double xn = x . out . " + y + " ) " ) . pow ( y . 3 ) . } //The f i r s t p a r t i a l d e r i v a t i v e o f f wrt y : −x+3xy^2−4y^3 public s t a t i c double f y ( double x . n<=100.

n = 96: (0.java is saved.-0.39636433796318005) n = 99: (0. 9 Available for free at http://java.4711356343449874.4711356343449705.4711356343449874.776075636032301 < 0.4711356343449874.39636433796318005) n = 10: (0. It is easy to confirm that ∇ f = 0 at this point.0. In the directory where newton. You will need the Java Development Kit9 to compile the code.-1. 0) with this command: java newton 0 0 Below is the output of the program using (0.47123972682634485.39636433796318005) n = 97: (0.6065857885615251.-0.4711356343449874. −0.0.484506572966545.4711356343449874.85722573273506 × 10−17 ∂x ∂f (0.3966334583092305) n = 6: (0.-0. namely the point (0.. you should first save the code in Listing 2.4711356343449874.0) n = 2: (1.-0.92 CHAPTER 2.-0.4711356343449874.3963643379632247) n = 8: (0.4711356343449874.4711356343449874.-0.39636433796318005) = −8.39636433796318005) is a saddle point.405341511995805) n = 5: (0. It turns out that both partial derivatives are indeed close enough to zero to be considered zero: ∂f (0.. −0.4711356343449874.-0.39636433796318005) n = 100: (0.-0. run this command at a command prompt to compile the code: javac newton.sun.39636433796318005) n = 98: (0.-0.47113558510349535. truncated to show the first 10 lines and the last 5 lines: java newton 0 0 Initial point: (0.39636433796318005). −0.6 we know that (0.39636433796318005) = −8.39636433796318005) As you can see.0. so by Theorem 2.39636450001936047) n = 7: (0.39636433796318005) n = 9: (0.-0. 0) as the initial point.326672684688674 × 10−17 ∂y We also have D(0.4711356343449874. either by evaluating ∂ f and ∂ f at the point ourselves or by modifying our ∂x ∂y program to also print the values of the partial derivatives at the point.0) n = 1: (0. FUNCTIONS OF SEVERAL VARIABLES To use this program.4711356343449874.4711356343449874. −0.44194107452339687) n = 4: (0.-0.1 in a plain text file called newton.39636433796318005) = 4.0.com/javase/downloads .java.-0.java Then run the program with the initial point (0.39636433796318005) .-0.5) n = 3: (0. we appear to have converged fairly quickly (after only 8 iterations) to what appears to be an actual critical point (up to Java’s level of precision). −0.-0.

6733618916578702. 0.-1.0. which means it is a critical point.6703832459238667.-0. −5) yields the critical point (−7.5788664043863884.0.1855674752461383.2.5) n = 2: (-0.6703832459238667.6703832459238667.129841298650007.42501465652421205) n = 18: (-0. An idea of what the graph of f looks like near that point is shown in Figure 2.42501465652420045) = −4.5.4250146565242004) n = 19: (-0.42501465652420045) .42501465652420045) Again.-2..6.0.6703832459238667.-1.-0.004453014967208.4176293491131443) n = 12: (-0.540962756992551.42501465652420045) n = 99: (-0.0.-1.-0.42501465652420045) = 15. and trying different values does indeed lead to different sequences which converge: java newton -1 -1 Initial point: (-1. it is easy to confirm that both ∂ f and ∂ f vanish at the point ∂x ∂y (−0.595509445899435). 0.0.5426077421319053) n = 6: (-0. which makes it a saddle point.6703832459238667.6703832459238701. −5.2918236503332734) n = 13: (-0..0.6704392913413444.0.6 Unconstrained Optimization: Numerical Methods 93 Since ∇ f consists of cubic polynomials.49848120123515316) n = 14: (-0.6703832679150286.6985177124230715.-1.4319791238981274) n = 8: (-0.24529117721011612) n = 7: (0.4252025996474051) n = 16: (-0.0. n = 98: (-0. Finally.0) n = 1: (-0. And D(−0. which does suggest a local maximum around that point.08450704225352113) n = 3: (-0.5161209914612475. running the computer program with the initial point (−5.8643989895639324) n = 5: (-0. it seems likely that there may be three critical points.0. The computer program makes experimenting with other initial points easy. .05837851765533317.-0.-0.-0.0222994755432 < 0 ∂x2 so we know that (−0.0.6703832459238667.0.4540060574531383.2047647348546167) n = 4: (-0.6703832459238667.11570743992954591.0.2. with D < 0 at that point.42501465652420045) n = 100: (-0.9206128022529645) n = 11: (-0.4794622222856417.42501465652420045) n = 20: (-0.3853578526055 > 0 ∂2 f (−0.0.3672160534444.-0.6703832459238667.4250147307973365) n = 17: (-0. 0.6703832459238667.6536079835854451) n = 9: (-0.4345777963475479) n = 15: (-0.42501465652420045) is a local maximum.121516233310142) n = 10: (-1.6703832459238667. 0.49295774647887325.42501465652420045).

2 0. then. y) = x3 − xy − x + xy3 − y4 : (−0.2 f (x.4 z -0. and the proof that it converges (given a “reasonable” choice for the initial point) requires techniques beyond the scope of this text.57) -1 -0. In the case of functions which have a global maximum or minimum. at least in practical applications.4711356343449874. you move a certain amount in the (0. In general. −0.6 0.4.39636433796318005) : saddle point .4 0.94 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES .67. This field of study is called nonlinear programming.6 y 0.8 -1 0 0. global maxima and minima tend to be more interesting than local versions.595509445899435) : saddle point The derivation of Newton’s algorithm. See R ALSTON and R ABINOWITZ for more detail and for discussion of other numerical methods. Recall that the negative gradient −∇ f gives the direction of the fastest rate of decrease of a function f .2 -0.4 0.0. A maximization problem can always be turned into a minimization problem (why?). Newton’s algorithm can be used to find those points. Many of these methods are based on the steepest descent technique.2 0 -0.6 -0. y) = x3 − xy − x + xy3 − y4 for −1 ≤ x ≤ 0 and 0 ≤ y ≤ 1 We can summarize our findings for the function f (x. 0. Our description of Newton’s algorithm is the special two-variable case of a more general algorithm that can be applied to functions of n ≥ 2 variables.8 (-0. which is based on an idea that we discussed in Section 2.4 -0. −5. The crux of the steepest descent idea.6 -0.2 1 0 x Figure 2.6703832459238667.42501465652420045) : local maximum (−7.540962756992551. 0.8 -0.6.42. is that starting from some initial point.0. so a large number of methods have been developed to find the global minimum of functions of any number of variables.

fxx.21 from the previous section. how do you explain it? (Hint: Something strange should happen. Either modify that program or write one of your own in a programming language of your choice to show that Newton’s algorithm does lead to the point (2. . yn ) ∂ f1 ∂x (xn . yn ) ∂ f2 ∂ f1 ∂ f2 ∂ f1 (xn . 3). .  ¨ C Exercises ©  1. then use the initial point (3. . yn ) f2 (xn . 2). For n = 0. Did anything strange happen when your program ran? If so. where f1 (x. yn ) f2 (xn . Recall Example 2. 3. ease of calculation. In fact. and you then just keep repeating that procedure until eventually (hopefully) you reach the point where f has its smallest value. Notice that our computer program can be modified fairly easily to use this function (just change the return values in the fx. Newton’s algorithm can be interpreted as a modified steepest descent method.6 Unconstrained Optimization: Numerical Methods 95 direction of −∇ f at that point. yn ) − (xn . y) and f2 (x. y) = 0 . There is a “pure” steepest descent method. y0 ). yn ) ∂ f2 ∂x (xn . and a multitude of variations on it that improve the rate of convergence. and compare the results. yn ) xn+1 = xn − ∂ f1 ∂y (xn . yn )∞ converges to a solution. 0) and (1.) 2. . fy. etc. 1) for the initial point (x0 . yn ) D(xn .2. yn ) ∂ f2 ∂y (xn . S HERALI and S HETTY. There is a version of Newton’s algorithm for solving a system of two equations f1 (x. fyy and fxy function definitions to use the appropriate partial derivative). yn ) . and of nonlinear programming in general. yn ) (xn . Wherever that takes you becomes your new point. yn ) = ∂x ∂y ∂y ∂x . Show that you get two different solutions when using (0. First use the initial point (0. Make sure that your program attempts to do 100 iterations of the algorithm. . yn ) . y0 ). yn ) (xn . 1) was a global minimum for the function f (x. 1. y) = sin(xy) − x − y = 0 and f2 (x. 2. yn+1 = yn + D(xn . y) = 0 and f2 (x. where Then the sequence of points (xn . D(xn . y) are smooth real-valued functions: Pick an initial point (x0 . define: f1 (xn . y) = (x − 2)4 + (x − 2y)2 . where we showed that the point (2. see B AZARAA. For more discussion of this. y) = e2x − 2x + 3y = 0 . yn ) f1 (xn . Write a computer n=1 program that uses this algorithm to find approximate solutions to the system of equations f1 (x. 1).

called the Lagrange multiplier method10 . So since y = 10 − x = 5. y in terms of x using that equation. This gives y = 10 − x. which we then substitute into f to get f (x. then we can solve for. then the Second Derivative Test tells us that x = 5 is a local maximum for f . . For a rectangle whose perimeter is 20 m. y) = c are called constrained maximum or constrained minimum points. Example 2. using single-variable calculus. Since we must have 2x + 2y = 20. 10] (since f = 0 at the endpoints of the interval). The perimeter P of the rectangle is then given by the formula P = 2x + 2y. y) with the condition that they satisfy the constraint equation g(x. This is now a function of x alone. Similar definitions hold for functions of three variables. respectively. and hence x = 5 must be the global maximum on the interval [0. y) = xy given : 2x + 2y = 20 The reader is probably familiar with a simple method. Notice in the above example that the ease of the solution depended on being able to solve for one variable in terms of the other in the equation 2x + 2y = 20. y) = c (or g(x. y. The Lagrange multiplier method for solving such problems can now be stated: 10 Named after the French mathematician Joseph Louis Lagrange (1736-1813).6 we were concerned with finding maxima and minima of functions without any constraints on the variables (other than being in the domain of the function). and we say that x and y are constrained by g(x. y) = xy = x(10 − x) = 10x − x2 .24. for solving this problem. find the dimensions that will maximize the area. Points (x. y) (or f (x. y) = c. this problem can be stated as: Maximize : f (x. say. FUNCTIONS OF SEVERAL VARIABLES 2. so we now just have to maximize the function f (x) = 10x − x2 on the interval [0. y. 10]. y) = c is called the constraint equation. But what if that were not possible (which is often the case)? In this section we will use a general method. z)) given : g(x.7 Constrained Optimization: Lagrange Multipliers In Sections 2. for solving constrained optimization problems: Maximize (or minimize) : f (x.5 and 2.96 CHAPTER 2. y) which are maxima or minima of f (x. What would we do if there were constraints on the variables? The following example illustrates a simple case of this type of problem. z) = c) for some constant c The equation g(x. Solution: The area A of a rectangle with width x and height y is A = xy. Since we are given that the perimeter P = 20. Since f ′ (x) = 10 − 2x = 0 ⇒ x = 5 and f ′′ (5) = −2 < 0. then the maximum area occurs for a rectangle whose width and height both are 5 m.

y) for some constant λ (the number λ is called the Lagrange multiplier).11 Note that the theorem only gives a necessary condition for a point to be a constrained maximum or minimum. y) = c .8 for more detail.25. y) or at a “boundary” point of the set B. . y) 0 for all (x. y ≤ 10. Again. y) = λ∇g(x. which cause that line to be restricted to a line segment in 2 (including the endpoints of that line segment). y) = xy given : g(x. of the rectangle. in Example 2. this problem can be stated as: Maximize : f (x. which is beyond the scope of this text. use the Lagrange multiplier method to find the dimensions that will maximize the area. Let f (x. Whether a point (x.24.7 really is a constrained maximum or minimum? The answer is that it depends on the constraint function g(x. y) satisfying ∇ f (x. For instance. find the points (x. y) and g(x. y) = λ∇g(x. y) that solve the equation ∇ f (x. y). A rigorous proof of the above theorem requires use of the Implicit Function Theorem. y) given : g(x. y) = c (plus any hidden constraints) describes a bounded set B in 2 . together with any implicit constraints. It can be shown12 that if the constraint equation g(x. and suppose that c is a scalar constant such that ∇g(x. y) that satisfy the equation g(x. y) for some λ means solving the equations 11 12 See T AYLOR and M ANN. with x and y representing the width and height. y) will occur either at a point (x. then it must be such a point. So how can you tell when a point that satisfies the condition in Theorem 2. Then to solve the constrained optimization problem Maximize (or minimize) : f (x. respectively. Solution: As we saw in Example 2. which is bounded. namely 0 ≤ x.2. y) for some λ actually is a constrained maximum or minimum can sometimes be determined by the nature of the problem itself. In Example 2. y) be smooth functions. However. y) = 2x + 2y = 20 Then solving the equation ∇ f (x. If there is a constrained maximum or minimum. y) = λ∇g(x. see T AYLOR and M ANN. which by itself is not bounded.7. y) = λ∇g(x. For a rectangle whose perimeter is 20 m. there are “hidden” constraints.7 Constrained Optimization: Lagrange Multipliers 97 Theorem 2.24 the constraint equation 2x + 2y = 20 describes a line in 2 . § 6. then the constrained maximum or minimum of f (x.24 it was clear that there had to be a global maximum. due to the nature of the problem. Example 2. y) = c. y) that satisfies ∇ f (x.

y) = x2 + y2 = 80 Solving ∇ f (x. Thus the problem can be stated as: Maximize (and minimize) : f (x. so now substitute either of the expressions for x or y into the constraint equation to solve for x and y: 20 = g(x. y 0. So we can solve both equations for λ as follows: x−1 y−2 =λ= x y ⇒ xy − y = xy − 2x ⇒ y = 2x . and minimizing the distance is equivalent to minimizing the square of the distance. Similarly. Find the points on the circle x2 + y2 = 80 which are closest to and farthest from the point (1. y) = 2x + 2y = 2x + 2x = 4x ⇒ x=5 ⇒ y=5 There must be a maximum area. FUNCTIONS OF SEVERAL VARIABLES ∂f ∂g ∂f ∂g =λ and = λ . 5) = 25 > 0. Example 2. Doing this we get x y =λ= 2 2 ⇒ x=y. ∴ The maximum area occurs for a rectangle whose width and height both are 5 m. namely: ∂x ∂x ∂y ∂y y = 2λ .26. y) = λ∇g(x. since the minimum area is 0 and f (5.98 CHAPTER 2. y) means solving the following equations: 2(x − 1) = 2λx . 2(y − 2) = 2λy Note that x 0 since otherwise we would get −2 = 0 in the first equation. x = 2λ The general idea is to solve for λ in both equations. y) to the point (1. then set those expressions equal (since they both equal λ) to solve for x and y. 5) that we found (called a constrained critical point) must be the constrained maximum. 2) is d= (x − 1)2 + (y − 2)2 . so the point (5. y) = (x − 1)2 + (y − 2)2 given : g(x. 2). Solution: The distance d from any point (x.

We needed λ only to find the constrained critical points. −8). 0. 0. So the two constrained critical points are (4. It turns out that λ gives an approximation of the change in the value of the function f (x. 0. y. Example 2. y) = x2 + y2 = 80 yields 5x2 = 80. but made no use of its value. z) = λ∇g(x. 8) = 45 and f (−4. 0. x2 + y2 + z2 = 1 describes a sphere (which is bounded) in constrained maximum point and −1 √ . and since there must be points on the circle closest to and farthest from (1. when the constant c in the constraint equation g(x. 8) (1. x2 + y2 = 80 y 99 (4.7. −1 √ 2 2 1 1 √ . Since f (4.27. y. so we can divide by λ in the second equation to get y = 0 and we can divide by λ in the first and 1 third equations to get x = 2λ = z. 2) 0 x (−4. y) that we wish to maximize or minimize. Substituting these expressions into the constraint equation g(x. So far we have not attached any significance to the value of the Lagrange multiplier λ. . √ 2 2 is the is the constrained minimum point. 2). y. √ 2 2 −1 √ . and since the constraint equation 3. 0. 2) and (−4. Maximize (and minimize) : f (x. so x = ±4. 8) and (−4. y. √ 2 2 1 1 √ . 8) is the point on the circle closest to (1. Since f > f . −8) = 125.2.7 Constrained Optimization: Lagrange Multipliers Substituting this into g(x.1 The Lagrange multiplier method can be extended to functions of three variables. which is a bounded set in 2 . Notice that since the constraint equation x2 + y2 = 80 describes a circle.7. 2) (see Figure 2. z): 1 = 2λx 0 = 2λy 1 = 2λz The first equation implies λ 0 (otherwise we would have 1 = 0). z) = x2 + y2 + z2 = 1 yields the constrained critical points and −1 √ . then it must be the case that (4. −8) is the farthest from (1. 0. y. z) = x + z given : g(x. then we were guaranteed that the constrained critical points we found were indeed the constrained maximum and minimum. z) = x2 + y2 + z2 = 1 Solution: Solve the equation ∇ f (x. −8) Figure 2. −1 √ 2 2 .1). y) = c is changed by 1. −1 √ 2 2 then 1 1 √ .

25 we showed that the constrained optimization problem Maximize : f (x.e. y) = 2x + 2y = 20 had the solution (x. Find the constrained maxima and minima of f (x. y) at the constrained maximum increased from f (5. Thus. note that solving the equation ∇ f (x. λ = 2. 3). Luckily there are many numerical methods for solving constrained optimization problems. Find the constrained maxima and minima of f (x. y) = xy given that x2 + 3y2 = 6.25) = 27. 3. y) = xy given : g(x. In a similar fashion we could show that the constrained optimization problem Maximize : f (x. z) = x + y2 + 2z given that 4x2 + 9y2 − 36z2 = 36. B 4.25.25.5 is close to 2. y) = c from c = 20 to c = 21. y) = λ∇g(x. 5.13  ¨ A Exercises ©  1.5625. that is. 5). y) = 2x + 2y = 21 has the solution (x.5. a2 b2 c2 13 See B AZARAA. 2. Find the points on the circle x2 + y2 = 100 which are closest to and farthest from the point (2. Finally. though we will not discuss them here. FUNCTIONS OF SEVERAL VARIABLES For example. Find the constrained maxima and minima of f (x. y) = (5. it increased by 2. And the 3-variable case can get even more complicated. 5.5625 when we increased the value of c in the constraint equation g(x.5625. y) = 2x + y given that x2 + y2 = 4. 5) = 25 to f (5. Notice that λ = 2. So we see that the value of f (x. y) means having to solve a system of two (possibly nonlinear) equations in three unknowns. y) = xy given : g(x.100 CHAPTER 2. in Example 2. pt) − f (old max. λ ≈ ∆ f = f (new max. . may not be possible to do. All of this somewhat restricts the usefulness of Lagrange’s method to relatively simple functions. y) = (5. and that λ = x/2 = y/2. pt) . 5.25). Find the volume of the largest rectangular parallelepiped that can be inscribed in the ellipsoid x2 y2 z2 + + =1. i. which as we have seen before. y. S HERALI and S HETTY.

Is there a similar way of defining integration of real-valued functions of two or more variables? The answer is yes. Then the trace of the surface in that plane is the curve f (x∗. For instance. y) : a ≤ x ≤ b. let A(x) be that area (see Figure 3. z z = f (x. d 101 . where x∗ is fixed and only y varies. d]. y) is a continuous function of y over the interval [c.1. slice the surface z = f (x. so we know that the area under the curve is the definite integral. The area A under that curve (i. the double integral of a nonnegative real-valued function f (x. to integrate a function f (x) it is necessary to find the antiderivative of f . y) ≥ 0 for all (x. As we will now see. y) be a continuous function such that f (x.1 The area A(x) varies with x Then A(x) = c f (x. y) with the plane x = x∗ parallel to the yz-plane.1 Double Integrals In single-variable calculus.3 Multiple Integrals 3. y) ≥ 0 represents the volume “under” the surface z = f (x. the area of the region between the curve and the xy-plane) as y varies over the interval [c. and only y varies. y) c a x b x R 0 A(x) d y Figure 3. Let f (x. For any number x∗ in the interval [a. d] then depends only on the value of x∗. This makes sense since for a fixed x the function f (x. b]. as we will see shortly. d]. that is. b] × [c. differentiation and integration are thought of as inverse operations.1). c ≤ y ≤ d} in 2 . y). Recall also that the definite integral of a nonnegative function f (x) ≥ 0 represented the area “under” the curve y = f (x). y). y) on the rectangle R = {(x. We will often write this as R = [a. another function F(x) whose derivative is f (x).1. y) dy since we are treating x as fixed.e. So using the variable x instead of x∗.

Find the volume V under the plane z = 8x + 6y over the rectangle R = [0. treating the variable x as a constant (this is called integrating with respect to y). y) dy dx (3. 18 in T AYLOR and M ANN. 1] × [0. which can then be integrated with respect to x. we will usually discard the brackets and simply write d b V= c a f (x. This order of integration can be changed if it is more convenient.102 CHAPTER 3. b] of that crosssectional area A(x): b b d V= a A(x) dx = a c f (x. 1 due to Fubini’s Theorem. That is what occurs in the “outer” integral above (the second iterated integral). and the last expression in equation (3. y) dx dy . Also.1).2) It turns out that in general1 the order of the iterated integrals does not matter. 2]. and then the resulting function is integrated with respect to y using the “outer” limits of integration c and d. The final result is then a number (the volume). Also. so that the volume V would be d b V= c a f (x. This process of going through two iterations of integrals is called double integration. That is what occurs in the “inner” integral between the square brackets in equation (3. y) is integrated as a function of y.1) is called a double integral. (3. Once that integration is performed. y) but above the xy-plane over the rectangle R is the integral over [a. we could just as easily have taken the area of cross-sections under the surface which were parallel to the xz-plane. which would then depend only on the variable y.3) where it is understood that the fact that dx is written before dy means that the function f (x. Notice that integrating f (x. y) dx dy . y) with respect to y.1) We will always refer to this volume as “the volume under the surface”. . See Ch. (3. First the function f (x. MULTIPLE INTEGRALS The area A(x) is a function of x. The above expression uses what are called iterated integrals.1. y) is first integrated with respect to x using the “inner” limits of integration a and b. This is the first iterated integral. the result is then an expression involving only x. Example 3. y) with respect to y is the inverse operation of taking the partial derivative of f (x. so by the “slice” or cross-section method from singlevariable calculus we know that the volume V of the solid under the surface z = f (x.

y) = 8x + 6y ≥ 0 for 0 ≤ x ≤ 1 and 0 ≤ y ≤ 2. 3] × [1. y). Solution: We know that f (x.2. y) = e x+y > 0 for all (x. We can verify that we still get the same answer: 1 2 V= 0 1 0 (8x + 6y) dy dx 8xy + 3y2 0 1 y=2 y=0 = = 0 dx (16x + 12) dx 1 0 = 8x2 + 12x = 20 Example 3.1 Double Integrals Solution: We see that f (x. and the b . 2]. Find the volume V under the surface z = e x+y over the rectangle R = [2. so: 2 1 103 V= 0 2 0 (8x + 6y) dx dy 4x2 + 6xy 0 2 x=1 x=0 = = 0 dy (4 + 6y) dy 2 0 = 4y + 3y2 = 20 Suppose we had switched the order of integration. the integral a f (x) dx represents the difference of the area below the curve y = f (x) but above the x-axis when f (x) ≥ 0.3. so 2 3 V= 1 2 2 e x+y dx dy e x+y 1 2 x=3 x=2 = = 1 dy (ey+3 − ey+2 ) dy − ey+2 4 2 1 4 =e y+3 5 = e − e − (e − e3 ) = e5 − 2e4 + e3 Recall that for a general function f (x).

2π]. y) represents the difference of the volume below the surface z = f (x.104 CHAPTER 3. 5. 2] 1 2 For Exercises 5-12. y) but above the xy-plane when f (x. 2π π Example 3. We can still evaluate the double integral: 2π 0 0 π 2π sin(x + y) dx dy = 0 2π − cos(x + y) x=π x=0 dy = 0 (− cos(y + π) + cos y) dy 2π 0 = − sin(y + π) + sin y =0 = − sin 3π + sin 2π − (− sin π + sin 0)  A 1. y) = e x+y . Let M be a constant. f (x. Solution: Note that f (x. Show that M dx dy = M(d − c)(b − a). 2. 12. R = [1. 0 2 0 1 x(x + y) dx dy x(xy + sin x) dx dy −1 π −1 0 1 2 π/2 7. . 0 2 0 4 xy cos(x2 y) dx dy xy dx dy 0 1 d b c a sin x cos(y − π) dx dy 1 dx dy 11. y) = sin(x + y) is both positive and negative over the rectangle [0. 0 (1 − y)x2 dx dy (x + 2) dx dy 6. y) = x3 + y2 . our method of double integration by means of iterated integrals can be used to evaluate the double integral of any continuous function over a rectangle. y) over the rectangle R. regardless of whether f (x. Similarly. 1] × [0. −1 −1 13. y) ≥ 0 or not. R = [0. 1] × [−1. f (x. 1] 4. 1] 1 2 1 2 1 Exercises ©  ¨ For Exercises 1-4. 0 1 9. Thus. π] × [0. 2] × [0. MULTIPLE INTEGRALS area above the curve but below the x-axis when f (x) ≤ 0. 1] 3. f (x. 0 π/2 0 8. R = [0. y) ≤ 0. y) = x4 + xy + y3 . y) = 4xy. evaluate the given double integral. and the volume above the surface but below the xy-plane when f (x. f (x.3. the double integral of any continuous function f (x. 1] × [0. Evaluate 0 0 sin(x + y) dx dy. R = [0. 10. find the volume under the surface z = f (x. y) ≥ 0.

is given by f (x. with the A signifying area. We will assume that g1 (x) and g2 (x) do not intersect on the open interval (a.2. denoted by R b g2 (x) f (x. if we have a region R in the xy-plane that is bounded on the left by a curve x = h1 (y). y) dy dx (3. bounded on the right by the vertical line x = b (where a < b). the double integral of a real-valued function f (x.2 Double Integrals Over a General Region 105 3. y) dA = R a g1 (x) f (x.2 Double Integrals Over a General Region In the previous section we got an idea of what a double integral over a rectangle represents.1(a).4) This means that we take vertical slices in the region R between the curves y = g1 (x) and y = g2 (x). y y = g2 (x) d y R c x 0 a b a g2 (x) g1 (x) x = h1 (y) x = h2 (y) y = g1 (x) b f (x. Suppose that we have a region R in the xy-plane that is bounded on the left by the vertical line x = a. and bounded above by a curve y = g2 (x). We can now define the double integral of a real-valued function f (x. b) (they could intersect at the endpoints x = a and x = b. y) over more general regions in 2 . Note that f (x. This makes sense since the result of the first iterated integral will have to be a function of x alone. y) over the region R.2. which then allows us to take the second iterated integral with respect to x. y) dy dx R 0 (b) Horizontal slice: d c h2 (y) h1 (y) x (a) Vertical slice: f (x. though). The symbol dA is sometimes called an area element or infinitesimal. y) dA. bounded on the right by a curve x = h2 (y). as in Figure 3. bounded below by a curve y = g1 (x). bounded below by the horizontal . y) dx dy Figure 3. with functions of x as the limits of integration. y) is first integrated with respect to y.3.1 Double integral over a nonrectangular region R Then using the slice method from the previous section. Similarly.

0 ≤ y ≤ 2x2 }.3 We get the same answer using horizontal slices (see Figure 3. f (x. then taking horizontal slices gives d h2 (y) f (x.1(b) (assuming that h1 (y) and h2 (y) do not intersect on the open interval (c. and bounded above by the horizontal line y = d (where c < d). y) dx dy (3. Find the volume V under the plane z = 8x+6y over the region R = {(x.4 y 2 x= y/2 R x 0 1 Figure 3.2.2.4. y) over the region R. y) dA is the volume under the Also. Example 3.2.2. y) dA = R c h1 (y) f (x.2.106 CHAPTER 3. as in Figure 3. y) : 0 ≤ x ≤ 1. y) ≥ 0 for all (x.2. Using vertical slices we get: V= R y y = 2x2 (8x + 6y) dA 1 R x 0 1 Figure 3. if f (x. d)). MULTIPLE INTEGRALS line y = c.3): V= R (8x + 6y) dA  2 1    √   2 2 = 0 y/2 = 0 4x2 + 6xy     dy (8x + 6y) dx   x=1 x= √ y/2 dy = 0 (4 + 6y − (2y + √ 6 √ √ y y )) dy 2 2 2 = = 4y + 2y2 − 6 2 5/2 5 y 0 =8+8− 0 √ √ 6 2 32 5 √ (4 + 4y − 3 2y3/2 ) dy = 16 − 48 5 = 32 5 = 6. then R surface z = f (x. Solution: The region R is shown in Figure 3. y) in the region R.5) Notice that these definitions include the case when the region R is a rectangle.2 = 0 1 = 0 1       2x2 0 8xy + 3y2     (8x + 6y) dy dx   y=2x2 y=0 dx = 0 (16x3 + 12x4 ) dx 1 12 5 5 x 0 = 4x4 + =4+ 12 5 = 32 5 = 6.4 .

y) dA is defined as follows. is R = {(x. Using vertical slices in R gives V= R 2 1 4 (4 − 2x − y) dA 1 4 (4 = 0 2 0 −2x+4 − 2x − y) dy dx y=−2x+4 y=0 = 0 2 − 1 (4 − 2x − y)2 8 1 8 (4 dx = 0 − 2x)2 dx 2 0 1 = − 48 (4 − 2x)3 = 64 48 = 4 3 For a general region R. y) : 0 ≤ x ≤ 2. Find the volume V of the solid bounded by the three coordinate planes and the plane 2x + y + 4z = 4. The volume V is given by f (x.3. xi+1 ] × [y j . as shown by the shaded subrectangles in Figure 3.4(a) with a typical vertical slice. 0) (a) y = −2x + 4 R x 0 (b) (0.2. y j∗ ) ∆xi ∆y j . y j∗ ).4(b). 4.5(a). y 4 z (0. .5. y) over that subrectangle is approximately f (xi∗ . pick a point (xi∗ . d]. 0 ≤ y ≤ −2x + 4}. shown in 4 R Figure 3.2 Double Integrals Over a General Region 107 Example 3. In any such subrectangle [xi . 0. Only consider the subrectangles that are enclosed completely within the region R. y j+1 ].2. y) dA. Assume that f (x. 0.2. 1) 2x + y + 4z = 4 y 0 x (2.2. Then divide that rectangle into a grid of subrectangles. so it can be enclosed in some rectangle [a. b] × [c. y) = z = 1 (4 − 2x − y) and the region R. where f (x. y) ered so far.4 Solution: The solid is shown in Figure 3. Then the volume under the surface z = f (x. the double integral R is a nonnegative real-valued function and that R is a bounded region in 2 . 0) 2 Figure 3. where ∆xi = xi+1 − xi . which may not be one of the types of regions we have considf (x.

y j∗ ) is the height and ∆xi ∆y j is the base area of a parallelepiped. MULTIPLE INTEGRALS ∆y j = y j+1 − y j . the region R does not have to be bounded. We then define R summation (the limit is taken over all subdivisions of the rectangle [a. b] × [c. then the subrectangles begin to fill more and more of the region R. so that the length of the largest diagonal of the subrectangles goes to 0. over an unbounded region.6) where the summation occurs over the indices of the subrectangles inside R. We can evaluate improper double integrals (i. and f (xi∗ .5 Double integral over a general region R A similar definition can be made for a function f (x. . namely f (xi∗ . using the definition of the Riemann integral from single-variable calculus. or over a region which contains points where the function f (x. If we take smaller and smaller subrectangles.2. y j∗ ) yj c 0 a (a) ∆y j z = f (x. d] as the largest diagonal of the subrectangles goes to 0). In the case of a region of the type shown in Figure 3. y j∗ ) R y x xi xi+1 b Subrectangles inside the region R xi+1 x (b) Parallelepiped over a subrectangle. y) f (xi∗ . y) dA reduces to a sequence of two iterated integrals. y) over the region R. y j∗ ) y j y j+1 0 xi (xi∗ . with volume f (xi∗ . y) is not defined) as a sequence of iterated improper single-variable integrals. y j∗ ) ∆xi ∆y j Figure 3. y) < 0.108 CHAPTER 3. definition of R Finally.5(b). y j∗ ) ∆xi ∆y j . Then the total volume under the surface is approximately the sum of the volumes of all such parallelepipeds. as shown in Figure 3.e.2.1. y d z ∆xi y j+1 (xi∗ . y) dA as the limit of that double z = f (x. y) that is not necessarily always nonnegative: just replace each mention of volume by the negative volume in the description above when f (x. our f (x.2. and so the above sum approaches the actual volume under the surface f (x. j i (3.

you can assume that R is a region of the type shown in Figure 3. Prove that the volume of a tetrahedron with mutually perpendicular adjacent sides of lengths a.6. 0 ey dx dy xye−(x 2 +y2 ) 2 5. Find the volume V of the solid bounded by the three coordinate planes and the plane 3x + 2y + 5z = 6. (Hint: Mimic Example 3. 1 π/2 0 y 4.1(a). b. 0 2 0 sin x dx dy 2y 0 ∞ 0 1 0 x2 ∞ x ln x 3. B 11. and c. y 1. Show how Exercise 12 can be used to solve Exercise 10.2 Double Integrals Over a General Region ∞ 1 0 1/x2 109 Example 3.5 how three noncollinear points determine a plane.2.6 b . 0 2 0 y cos x sin y dx dy 1 dx dy 0 0 6.2. 2 dy dx 0 9. C 12. Evaluate Solution: ∞ 1 0 1/x2 2y dy dx. 0 dx dy 7. Find the volume V of the solid bounded by the three coordinate planes and the plane x + y + z = 1. as in Figure 3. 0 24x2 y dy dx 4x dy dx 2. is abc .2.) 13. Explain why the double integral R 1 dA gives the area of the region R. evaluate the given double integral. For sim- plicity.6. 2y dy dx = 1 ∞ ∞ 1 y2 y=1/x2 y=0 dx ∞ 1 = 1 x−4 dx = − 3 x−3 = 0 − (− 1 ) = 3 1 3  A 1 1 √ 2 Exercises ©  π ¨ For Exercises 1-6. 8. and recall from 6 Section 1.5. c a Figure 3. 10.3.

y is bounded between two curves h1 (x) and h2 (x). In each subparallelepiped inside S . z) whose graph lies in 4 . Then define the triple integral of f (x. Physically. z) : x1 ≤ x ≤ x2 .3 Triple Integrals Our definition of a double integral of a real-valued function f (x. ∆y and ∆z.7) where the limit is over all divisions of the rectangular parallelepiped enclosing S into subparallelepipeds whose largest diagonal is going to 0. that is. We simply proceed as before: the solid S can be enclosed in some rectangular parallelepiped. and the triple summation is over all the subparallelepipeds inside S . y. bounded above by a surface z = g2 (x. y. A more complicated case is where S is a solid which is bounded below by a surface z = g1 (x. with sides of lengths ∆x.g. pick a point (x∗ . It can be shown that this limit does not depend on the choice of the rectangular parallelepiped enclosing S . z) dx dy dz . Then b h2 (x) h1 (x) g2 (x. z) dV = S a g1 (x. z1 ≤ z ≤ z2 }. y. what does the triple integral represent? We saw that a double integral could be thought of as the volume under a two-dimensional surface. length in 1 .y) f (x. z) over a solid S in 3 . z) over S . y). y. z) dV. which then leaves you with a . namely z2 y2 y1 x2 f (x. the triple integral is a sequence of three iterated integrals. y. (3. z) dV = S z1 x1 f (x. z) dV = lim S f (x∗ . y. z2 ].8) where the order of integration does not matter. denoted by f (x. y1 ≤ y ≤ y2 . It turns out that the triple integral simply generalizes this idea: it can be thought of as representing the hypervolume under a three-dimensional hypersurface w = f (x. x2 ] × [y1 . z∗ ). This is the simplest case. z∗ ) ∆x ∆y ∆z . In general. the word “volume” is often used as a general term to signify the same concept for any n-dimensional object (e. y∗ . which is then divided into subparallelepipeds. The symbol dV is often called the volume element.y) f (x. area in 2 ). y) over a region R in 2 can be extended to define a triple integral of a real-valued function f (x. z) dz dy dx . and x varies between a and b. S = {(x. (3.9) Notice in this case that the first iterated integral will result in a function of x and y (since its limits of integration are functions of x and y). (3. y2 ] × [z1 . y∗ . It may be hard to get a grasp on the concept of the “volume” of a four-dimensional object. y. but at least we now know how to calculate that volume! In the case where S is a rectangular parallelepiped [x1 .110 CHAPTER 3. y). by S f (x. y. y. MULTIPLE INTEGRALS 3. y.

8.3. 3 2 0 0 1 Example 3. z). changing the roles of the variables x. so as you can probably tell. of course. We will see some other ways in which triple integrals are used later in the text. Solution: 1 0 0 1−x 0 2−x−y 1 1−x 0 1 1−x 0 1 1−x 0 1 1 2 − 2 x2 − xy − 1 y2 dy dx 2 y=1−x y=0 1 (x + y)(2 − x − y) + 2 (2 − x − y)2 dy dx 1 (x + y)z + 2 z2 (x + y + z) dz dy dx = 0 z=2−x−y z=0 dy dx = 0 = 0 = 0 1 1 2y − 2 x2 y − xy − 1 xy2 − 1 y3 2 6 11 6 dx = 0 − 2x + 1 x3 dx 6 1 1 4 24 x 0 = 11 6 x − x2 + = 7 8 . y. regardless of what it represents.2. triple integrals can be quite tricky.3 Triple Integrals 111 double integral of a type that we learned how to evaluate in Section 3.7. just learning how to evaluate a triple integral. Evaluate 0 (x + y + z) dz dy dx. Evaluate 0 (xy + z) dx dy dz. Solution: 3 0 0 2 0 1 3 2 0 3 2 0 3 1 2 4y 3 1 2y 1 2 2x y (xy + z) dx dy dz = 0 + xz x=1 x=0 dy dz = 0 + z dy dz y=2 y=0 = 0 + yz dz = 0 (1 + 2z) dz 3 0 = z + z2 = 12 1 1−x 0 0 2−x−y Example 3. is the most important thing. many variations on this case (for example. At this point. There are.

112

CHAPTER 3. MULTIPLE INTEGRALS
3

Note that the volume V of a solid in V=

is given by 1 dV .
S

(3.10)

Since the function being integrated is the constant 1, then the above triple integral reduces to a double integral of the types that we considered in the previous section if the solid is bounded above by some surface z = f (x, y) and bounded below by the xy-plane z = 0. There are many other possibilities. For example, the solid could be bounded below and above by surfaces z = g1 (x, y) and z = g2 (x, y), respectively, with y bounded between two curves h1 (x) and h2 (x), and x varies between a and b. Then
b h2 (x) h1 (x) g2 (x,y) b h2 (x) h1 (x)

V=
S

1 dV =
a g1 (x,y)

1 dz dy dx =
a

(g2 (x, y) − g1 (x, y)) dy dx

just like in equation (3.9). See Exercise 10 for an example. 
¨

A
3 2 0 π x 0 e y 0 2 4 2 0 0 xy 1

1

For Exercises 1-8, evaluate the given triple integral.
x 0 1 z 0 2 y2 0 1 1−x 0 0 0 1−x−y 0 z2 0 y y

1.
0

xyz dx dy dz x2 sin z dz dy dx
0 0 1/y

2.
0

xyz dz dy dx zey dx dy dz
0
2

3. 5.
1 0 3

4. 6.
1

x2 z dx dz dy 1 dx dy dz
1

yz dx dz dy 1 dz dy dx
0

7.

8.
z2 y2 x2 z1 y1 x1

9. Let M be a constant. Show that

M dx dy dz = M(z2 − z1 )(y2 − y1 )(x2 − x1 ).

B
10. Find the volume V of the solid S bounded by the three coordinate planes, bounded above by the plane x + y + z = 2, and bounded below by the plane z = x + y.

C
b z a a y b

11. Show that
a

f (x) dx dy dz =
a

(b−x)2 2

f (x) dx. (Hint: Think of how changing

the order of integration in the triple integral changes the limits of integration.)

3.4 Numerical Approximation of Multiple Integrals

113

3.4 Numerical Approximation of Multiple Integrals
As you have seen, calculating multiple integrals is tricky even for simple functions and regions. For complicated functions, it may not be possible to evaluate one of the iterated integrals in a simple closed form. Luckily there are numerical methods for approximating the value of a multiple integral. The method we will discuss is called the Monte Carlo method. The idea behind it is based on the concept of the average value of a function, which you learned in single-variable calculus. Recall that for a continuous function f (x), the average value f¯ of f over an interval [a, b] is defined as f¯ = 1 b−a
b

f (x) dx .
a

(3.11)

The quantity b − a is the length of the interval [a, b], which can be thought of as the “volume” of the interval. Applying the same reasoning to functions of two or three variables, we define the average value of f (x, y) over a region R to be f¯ = 1 A(R)
R

f (x, y) dA ,

(3.12)

where A(R) is the area of the region R, and we define the average value of f (x, y, z) over a solid S to be 1 f (x, y, z) dV , (3.13) f¯ = V(S )
S

where V(S ) is the volume of the solid S . Thus, for example, we have f (x, y) dA = A(R) f¯ .
R

(3.14)

The average value of f (x, y) over R can be thought of as representing the sum of all the values of f divided by the number of points in R. Unfortunately there are an infinite number (in fact, uncountably many) points in any region, i.e. they can not be listed in a discrete sequence. But what if we took a very large number N of random points in the region R (which can be generated by a computer) and then took the average of the values of f for those points, and used that average as the value of f¯? This is exactly what the Monte Carlo method does. So in formula (3.14) the approximation we get is f (x, y) dA ≈ A(R) f¯ ± A(R)
R

f 2 − ( f¯)2 , N
N 2 i=1 ( f (xi , yi ))

(3.15)

where f¯ =

N i=1

f (xi , yi ) N

and

f2 =

N

,

(3.16)

114

CHAPTER 3. MULTIPLE INTEGRALS

with the sums taken over the N random points (x1 , y1 ), . . ., (xN , yN ). The ± “error term” in formula (3.15) does not really provide hard bounds on the approximation. It represents a single standard deviation from the expected value of the integral. That is, it provides a likely bound on the error. Due to its use of random points, the Monte Carlo method is an example of a probabilistic method (as opposed to deterministic methods such as Newton’s method, which use a specific formula for generating points). For example, we can use formula (3.15) to approximate the volume V under the plane z = 8x + 6y over the rectangle R = [0, 1] × [0, 2]. In Example 3.1 in Section 3.1, we showed that the actual volume is 20. Below is a code listing (montecarlo.java) for a Java program that calculates the volume, using a number of points N that is passed on the command line as a parameter.
//Program t o approximate the double i n t e g r a l o f f ( x , y )=8 x+6y //over the r e c t a n g l e [ 0 , 1 ] x [ 0 , 2 ] . public class montecarlo { public s t a t i c void main ( String [ ] args ) { //Get the number N o f random p o i n t s as a command− l i n e parameter int N = I n t e g e r . parseInt ( args [ 0 ] ) ; double x = 0 ; //x−c o o r d i n a t e o f a random p o i n t double y = 0 ; //y−c o o r d i n a t e o f a random p o i n t double f = 0 . 0 ; //Value o f f at a random p o i n t double mf = 0 . 0 ; //Mean o f the values o f f double mf2 = 0 . 0 ; //Mean o f the values o f f ^2 for ( int i =0; i <N; i ++) { //Get the random c o o r d i n a t e s x = Math . random ( ) ; //x i s between 0 and 1 y = 2 ∗ Math . random ( ) ; //y i s between 0 and 2 f = 8∗x + 6∗y ; //Value o f the f u n c t i o n mf = mf + f ; //Add t o the sum o f the f values mf2 = mf2 + f ∗ f ; //Add t o the sum o f the f ^2 values } mf = mf /N; //Compute the mean o f the f values mf2 = mf2 /N; //Compute the mean o f the f ^2 values System . out . p r i n t l n ( "N = " + N + " : i n t e g r a l = " + v o l ( ) ∗ mf + " +/ − " + v o l ( ) ∗ Math . s q r t ( ( mf2 − Math . pow ( mf , 2 ) ) /N ) ) ; //Print the r e s u l t } //The volume o f the r e c t a n g l e [ 0 , 1 ] x [ 0 , 2 ] public s t a t i c double v o l ( ) { return 1 ∗ 2 ; } } Listing 3.1 Program listing for montecarlo.java

The results of running this program with various numbers of random points (e.g. java montecarlo 100) are shown below:

3.4 Numerical Approximation of Multiple Integrals N N N N N N = = = = = = 10: 100: 1000: 10000: 100000: 1000000: 19.36543087722646 21.334419561385353 19.807662237526227 20.080975812043256 20.009403854556716 20.000866994982314 +/+/+/+/+/+/2.7346060413546147 0.7547037194998519 0.26701709691370235 0.08378816229769506 0.026346782289498317 0.008321168748642816

115

As you can see, the approximation is fairly good. As N → ∞, it can be shown that the √ Monte Carlo approximation converges to the actual volume (on the order of O( N), in computational complexity terminology). In the above example the region R was a rectangle. To use the Monte Carlo method for a nonrectangular (bounded) region R, only a slight modification is needed. Pick a ˜ rectangle R that encloses R, and generate random points in that rectangle as before. Then use those points in the calculation of f¯ only if they are inside R. There is no need to calculate the area of R for formula (3.15) in this case, since the exclusion of points ˜ not inside R allows you to use the area of the rectangle R instead, similar to before. For instance, in Example 3.4 we showed that the volume under the surface z = 8x+6y over the nonrectangular region R = {(x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 2x2 } is 6.4. Since the ˜ rectangle R = [0, 1] × [0, 2] contains R, we can use the same program as before, with the only change being a check to see if y < 2x2 for a random point (x, y) in [0, 1] × [0, 2]. Listing 3.2 below contains the code (montecarlo2.java):
//Program t o approximate the double i n t e g r a l o f f ( x , y )=8 x+6y over the //r e g i o n bounded by x =0 , x =1 , y =0 , and y=2x^2 public class montecarlo2 { public s t a t i c void main ( String [ ] args ) { //Get the number N o f random p o i n t s as a command− l i n e parameter int N = I n t e g e r . parseInt ( args [ 0 ] ) ; double x = 0 ; //x−c o o r d i n a t e o f a random p o i n t double y = 0 ; //y−c o o r d i n a t e o f a random p o i n t double f = 0 . 0 ; //Value o f f at a random p o i n t double mf = 0 . 0 ; //Mean o f the values o f f double mf2 = 0 . 0 ; //Mean o f the values o f f ^2 for ( int i =0; i <N; i ++) { //Get the random c o o r d i n a t e s x = Math . random ( ) ; //x i s between 0 and 1 y = 2 ∗ Math . random ( ) ; //y i s between 0 and 2 i f ( y < 2∗Math . pow ( x , 2 ) ) { //The p o i n t i s in the r e g i o n f = 8∗x + 6∗y ; //Value o f the f u n c t i o n mf = mf + f ; //Add t o the sum o f the f values mf2 = mf2 + f ∗ f ; //Add t o the sum o f the f ^2 values } } mf = mf /N; //Compute the mean o f the f values mf2 = mf2 /N; //Compute the mean o f the f ^2 values System . out . p r i n t l n ( "N = " + N + " : i n t e g r a l = " + v o l ( ) ∗ mf +

2. 3.0. y) in a rectangle.g.349975080015089 +/. Repeat Exercise 2 with the solid S = {(x. 6. java montecarlo2 1000) are shown below: N N N N N N = = = = = = 10: 100: 1000: 10000: 100000: 1000000: integral integral integral integral integral integral = = = = = = 6. 100000 and 1000000 random points. 1000. 100.  C Exercises ©  ¨ 1. 0 ≤ y ≤ x2 }. 1000. 1].15) (see Exercise 2).03200476870881392 6. 5. 2 ] public s t a t i c double v o l ( ) { return 1 ∗ 2 . 0 ≤ z ≤ 1 − x − y}.31916837260973624 6. 10000. Use the Monte Carlo method to approximate the volume of a sphere of radius 1.0. Show the program output for N = integral R 10.116 CHAPTER 3. 1 ] x [ 0 . and use the volume of the parallelepiped instead of the area of a rectangle in formula (3. 1] × [0.0. where S = [0.01009454409789472 To use the Monte Carlo method to evaluate triple integrals. see P RESS et al.417050897922222 +/. y. instead of random pairs (x. } //The volume o f the r e c t a n g l e [ 0 . y) : −1 ≤ x ≤ 1.0. where R = [0. 4.2.95747529014894 +/. 2 ) ) /N ) ) . y.0.3149056229650355 +/.9549009662159909 6.10040086346895105 6. 100000 and 1000000 random points. } } Listing 3. 1] × [0. z) in a parallelepiped. s q r t ( ( mf2 − Math . 100. z) : 0 ≤ x ≤ 1.440184132811864 +/. Write a program that uses the Monte Carlo method to approximate the double e xy dA.2 Program listing for montecarlo2. For a more detailed discussion of numerical integration methods.java The results of running the program with various numbers of random points (e. MULTIPLE INTEGRALS " +/ − " + v o l ( ) ∗ Math . Write a program that uses the Monte Carlo method to approximate the triple integral e xyz dV. x2 9 + y4 + z1 = 2 2 . Use the Monte Carlo method to approximate the volume of the ellipsoid 1. pow ( mf . 1].9185131565120592 6. 0 ≤ y ≤ 1. 1] × [0. Show the program output for S N = 10.477032813858756 +/. Repeat Exercise 1 with the region R = {(x. 10000. you will need to generate random triples (x.

on [0. The answer is yes. √ Then substituting that expression for x into the function f (x) = x3 x2 − 1 gives √ f (x) = f (g(u)) = (u + 1)3/2 u . for example. 3]). 2] onto [0. On the interval of integration [1. we let u = x2 − 1. 2]. the reader may be wondering if it is possible to simplify those integrals using a suitable substitution for the variables. which will give some motivation for how substitution works in multiple integrals.5 Change of Variables in Multiple Integrals Given the difficulty of evaluating multiple integrals.5 Change of Variables in Multiple Integrals 117 3. 3] we can define x as a function of u. the definite integral 2 1 x3 x2 − 1 dx . though it is a bit more complicated than the substitution method which you learned in single-variable calculus.3. First. Recall that if you are given. the function x → x2 − 1 is strictly increasing (and maps [1. namely √ x = g(u) = u + 1 . That is. . 0 √ 14 3 5 Let us take a different look at what happened when we did that substitution. which can be easily integrated to give . then you would make the substitution u = x2 − 1 ⇒ x2 = u + 1 du = 2x dx which changes the limits of integration x=2⇒u=3 so that we get 2 1 2 x=1⇒u=0 x3 x2 − 1 dx = = 1 3 0 1 2 2x 1 2 (u 3 · 2x x2 − 1 dx √ + 1) u du = = 1 2 u3/2 + u1/2 du . 3]) and hence has an inverse function (defined on the interval [0.

differentiable function from an interval [c. This formula turns out to be a special case of a more general formula which can be used to evaluate multiple integrals. if x = g(u) is a one-to-one.32 and § 15. and it is what you were implicitly using when doing integration by substitution. . so that a = g(c) and b = g(d). (3. We will assume that all the functions involved are continuously differentiable and that the regions and solids involved all have “reasonable” boundaries. g−1 (2) f (x) dx = 1 g−1 (1) In general. b] (on the x-axis). d). then c = g−1 (a) and d = g−1 (b).17) g−1 (a) This is called the change of variable formula for integrals of single-variable functions. MULTIPLE INTEGRALS dx = g ′ (u) ⇒ dx = g ′ (u) du du dx = 1 (u + 1)−1/2 du . which means that g ′ (u) 0 on the interval (c.62 for all the details. 2 so since g(0) = 1 ⇒ 0 = g−1 (1) g(3) = 2 ⇒ 3 = g−1 (2) then performing the substitution as we did earlier gives 2 2 f (x) dx = 1 1 3 x3 x2 − 1 dx 1 2 (u 3 = 0 √ + 1) u du . We will state the formulas for double and triple integrals involving real-valued functions of two and three variables. d] (which you can think of as being on the “u-axis”) onto an interval [a. which can be written as = 0 2 √ 1 (u + 1)3/2 u · 2 (u + 1)−1/2 du . The proof of the following theorem is beyond the scope of the text. which means f (g(u)) g ′ (u) du .2 2 See T AYLOR and M ANN. respectively. § 15.118 and we see that CHAPTER 3. and b g−1 (b) f (x) dx = a f (g(u)) g ′ (u) du .

z) = S S′ ∂x ∂v ∂y ∂v ∂z ∂v ∂x ∂w ∂y ∂w ∂z ∂w (3. v. respectively. w) (3. v)| dA(u. y. (3. y(u. ∂(u.18) f (x(u. the Jacobian J(u.5 Change of Variables in Multiple Integrals 119 Theorem 3. v. (3. w) = ∂u ∂z ∂u is never 0 in S ′ . w) of three variables is sometimes written as J(u. v) = ∂(x. w).1. v. v. y. w) define a one-to-one mapping of a solid S ′ in uvw-space onto a solid S in xyz-space such that the determinant ∂x ∂u ∂y J(u. y) dA(x. y(u. ∂(u. . y. Then f (x. The following example shows how the change of variables formula is used. y) and dA(u. v.18) is called the Jacobian of x and y with respect to u and v. y = y(u. z(u. v) define a one-to-one mapping of a region R′ in the uv-plane onto a region R in the xy-plane such that the determinant ∂x ∂u J(u. v. if x = x(u. v)| dA(u.19) is saying that dA(x. v) . Similarly. w) = ∂(x. and is sometimes written as J(u.21) The determinant J(u. which you can think of as a two-variable version of the relation dx = g ′ (u) du in the single-variable case. v. w)| dV(u. w). y) = R R′ ∂x ∂v ∂y ∂v (3. y) and (u. then f (x. w) and z = z(u.19) We use the notation dA(x. v. v. w) . v) (3. v)) | J(u. v. v) = ∂y ∂u is never 0 in R′ .20) f (x(u. v) and y = y(u. v) to denote the area element in the (x. Change of Variables Formula for Multiple Integrals Let x = x(u. w)) | J(u.3.22) Similarly. w). v. y) . v). y) = | J(u.23) Notice that formula (3. z) dV(x. v) in formula (3. v) coordinates. v. z) . v).

By looking at the numerator and denominator of the exponent of e. v)) | J(u.120 CHAPTER 3. v) = ∂y ∂u ∂x 1 2 ∂v = 1 ∂y −2 ∂v 1 2 1 2 = 1 1 1 = . Solution: First. y ≥ 0. v) = 2 (u + v).19). v)| dA v −v = 0 1 ev u 1 2 du dv dv = 0 1 u=v v u v 2 e u=−v v 2 (e = = 0 2 v − e−1 ) dv 1 0 4 (e − e−1 ) = 1 e2 − 1 1 e− = 4 e 4e . MULTIPLE INTEGRALS x−y Example 3. In Figure 3. we have x−y e x+y dA = R R′ 1 f (x(u. So solving for x and y gives x = 1 (u + v) and y = 2 (v − u). y = y(u. at least in a closed form. Evaluate R e x+y dA. v) = 1 (v − u) maps the region R′ 2 onto R in a one-to-one manner.5. y(u. where R = {(x. 2 1 we see how the mapping x = x(u.5. ⇒ | J(u.9.1 below. we need to write both x and y in terms of u 1 and v. v).1 x 1 x = 2 (u + v) v 1 R′ u = −v −1 The regions R and R′ 0 u=v u 1 1 2 (v y= − u) Now we see that ∂x ∂u J(u. y) : x ≥ 0. y 1 x+y=1 R 0 1 Figure 3. v)| = 2 2 2 so using horizontal slices in R′ . we will try the substitution u = x − y and v = x + y. x + y ≤ 1}. To use the change of variables formula (3. note that evaluating this double integral without using substitution is probably impossible.

θ) = r cos θ we have ∂x ∂r J(u.2). 1 where R = {(x. v) = ∂y ∂r ∂x cos θ −r sin θ ∂θ = r cos2 θ + r sin2 θ = r ⇒ | J(u. Example 3. y) : x2 + y2 ≤ 1} is the unit disk in 2 (see Figure 3. Letting x = x(r.3.5 Change of Variables in Multiple Integrals 121 The change of variables formula can be used to evaluate double integrals in polar coordinates. y) dx dy = R R′ f (r cos θ.5.2 z = x2 + y2 V= 0 (1 − r2 ) r dr dθ (r − r3 ) dr dθ − r=1 r4 4 r=0 = 0 = 0 2π dθ = = π 2 0 dθ . (3. Find the volume V inside the paraboloid z = x2 + y2 for 0 ≤ z ≤ 1. v)| = |r| = r . so we have the following formula: Double Integral in Polar Coordinates f (x. y = r sin θ maps the region R′ in the rθ-plane onto the region R in the xy-plane in a one-to-one manner. 2π 1 0 2π 1 0 2π r2 2 1 4 y 0 x Figure 3. r sin θ) r dr dθ . θ) = r sin θ . θ) we know that x2 + y2 = r2 and that the unit disk R is the set R′ = {(r. = ∂y sin θ r cos θ ∂θ and y = y(r. θ) : 0 ≤ r ≤ 1. 0 ≤ θ ≤ 2π}.24) where the mapping x = r cos θ.5.10. we see that V= R z x2 + y2 = 1 (1 − z) dA = R (1 − (x2 + y2 )) dA . In polar coordinates (r. Solution: Using vertical slices. Thus.

θ) we know that x2 + y2 = r and that the unit disk R is the set R′ = {(r.25) where the mapping x = r cos θ. z) r dr dθ dz .5. r sin θ. (3.3). θ) : 0 ≤ r ≤ 1. 0 ≤ θ ≤ 2π}. z x2 + y2 = 1 1 Example 3. z) dx dy dz = S S′ f (ρ sin φ cos θ. ρ sin φ sin θ. Triple Integral in Spherical Coordinates f (x. it can be shown (see Exercises 5-6) that triple integrals in cylindrical and spherical coordinates take the following forms: Triple Integral in Cylindrical Coordinates f (x. z) dx dy dz = S S′ f (r cos θ.5. In polar coordinates (r. z = z maps the solid S ′ in rθz-space onto the solid S in xyz-space in a one-to-one manner.26) where the mapping x = ρ sin φ cos θ. Thus. where R = {(x.11. MULTIPLE INTEGRALS x2 + y2 for 0 ≤ z ≤ 1. . Find the volume V inside the cone z = Solution: Using vertical slices. 2π 1 0 2π 1 0 2π r2 2 1 6 y 0 x Figure 3. y) : x2 + y2 ≤ 1} is the unit disk in 2 (see Figure 3. we see that V= R (1 − z) dA = R 1− x2 + y2 dA . y = r sin θ. y. ρ cos φ) ρ2 sin φ dρ dφ dθ .122 CHAPTER 3. y = ρ sin φ sin θ. y. (3. z = ρ cos φ maps the solid S ′ in ρφθ-space onto the solid S in xyz-space in a one-to-one manner.3 z = x2 + y2 V= 0 (1 − r) r dr dθ (r − r2 ) dr dθ − r=1 r3 3 r=0 = 0 = 0 2π dθ = = π 3 0 dθ In a similar fashion.

defined by 1 B(x. Using the substitution t = u/(u + 1). Solution: We see that S is the set ρ = a in spherical coordinates.) 8. y) for x > 0.3. Find the volume V inside the paraboloid z = x2 + y2 for 0 ≤ z ≤ 4. Show that the Beta function. Prove formula (3. show that the Beta function can be written as B(x. (u + 1) x+y for x > 0. satisfies the relation B(y. (2. v = (x − y)/2. 2 y z x 10. y > 0.12.25). 0). y > 0. Prove formula (3. find the volume V inside the sphere S = x2 + y2 + z2 = a2 . 3 = 0 − a3 3 cos φ dθ = 0 2a3 3 ¨ dθ =  A Exercises ©  1. then consider Example 3. 0) and (1.5 Change of Variables in Multiple Integrals Example 3. Find the volume of the solid bounded by z = x2 + y2 and z2 = 4(x2 + y2 ). (Hint: Use the change of variables u = (x + y)/2.26). y > 0. z = cw. Find the volume V inside both the sphere x2 + y2 + z2 = 1 and the cone z = 5. Find the volume V inside the cone z = B x2 + y2 for 0 ≤ z ≤ 3.) 11. 7. Find the volume inside the elliptic cylinder x2 a2 2 + y2 b2 2 C = 1 for 0 ≤ z ≤ 2. 4. 9. 1). 3. y = bv. where R is the triangle with vertices (0. (Hint: Use the 3 change of variables x = au. y) = 0 ∞ u x−1 du . Evaluate R x2 + y2 . 12.12. Find the volume V of the solid inside both x2 + y2 + z2 = 4 and x2 + y2 = 1. Show that the volume inside the ellipsoid a2 + b2 + c2 = 1 is 4πabc . . x−y 2 sin x+y 2 cos dA. so 2π π 0 0 2π π 0 a 123 V= S 2π 1 dV = 0 π 0 2π 1 ρ2 sin φ dρ dφ dθ ρ=a ρ=0 = 0 ρ3 3 sin φ dφ dθ = 0 φ=π φ=0 2π a3 sin φ dφ dθ 3 4πa3 . 2. for x > 0. y) = 0 t x−1 (1 − t)y−1 dt . x) = B(x. 6. For a > 0.

.1).28) xδ(x. M= R δ(x. y) given by ¯ ¯ x= ¯ where b y y = f (x) R ( x . y) of the center of mass of R are given by ¯ ¯ x= ¯ where My = R My M and y= ¯ Mx .29). Find the center of mass of the region R = {(x.27) assuming that R has uniform density. y) : 0 ≤ x ≤ 1. y∗ ) in that rectangle. 0 ≤ y ≤ f (x)} in 2 that represents a thin. which is the double integral δ(x.6. y) = x + y. Then the mass of R is the limit of the sums of the masses of all such rectangles inside R as the diagonals of the rectangles approach 0. y) dA . Mx = R yδ(x. y∗ )∆x ∆y. R Note that the formulas in (3. if the density function at (x. where f (x) is a continuous function on [a. y) is δ(x. The quantity M is the mass of the region R. y) dA . the center of mass of R has coordinates ( x.e the mass of R is uniformly distributed over the region. and taken as 1 for simplicity). y) : a ≤ x ≤ b.6.124 CHAPTER 3. In the general case where the density of a region (or lamina) R is a continuous function δ = δ(x. y) of the coordinates (x. M Figure 3. for some point (x∗ .1 Mx = a ( f (x))2 dx .13. (3. (3. MULTIPLE INTEGRALS 3. think of taking a small rectangle inside R with dimensions ∆x and ∆y close to 0. Example 3. y) of points inside R (where R can be any region in 2 ) the coordinates ( x. y) dA . M (3. b]. i. respectively. 2 b b My = a x f (x) dx .6 Application: Center of Mass Recall from single-variable calculus that for a region R = {(x.27) represent a special case when δ(x. In this case the area M of the region is considered the mass of R (the density is constant. M= a f (x) dx . y) ¯ ¯ 0 a b Center of mass of R x My M and y= ¯ Mx .29) The quantities M x and My are called the moments (or first moments) of the region R about the x-axis and y-axis. flat plate (see Figure 3. y) dA. The mass of that rectangle is approximately δ(x∗ . 0 ≤ y ≤ 2x2 }. To see this. y) = 1 throughout R in the formulas in (3.

M 9/10 27 y= ¯ Mx 5/7 50 = = . y) = 3 . y) dA (x + y) dy dx 0 0   y=2x2  1   y2    xy +  dx   =    2 y=0  0 = 1 = (2x3 + 2x4 ) dx 1 2x5 + = 2 5 and 0 4 x = 0 9 10 Mx = R 1 yδ(x. y) = x + y ¯ ¯ 4 increases as (x.6 Application: Center of Mass Solution: The region R is shown in Figure 3. y) approaches that upper corner.3. M 9/10 63 Note how this center of mass is a little further towards the upper corner of the region R than when the density is uniform (it is easy to use the formulas in (3.6.2.2 δ(x. 5 in that case).27) to show that 3 ( x. y) is called the centroid of R. y) dA 2x2 0 = = = y(x + y) dy dx   y=2x2  1 2   xy y3    dx    =     2 + 3 0 y=0 0 0 1 = (2x5 + 8x7 21 0 6 x 8x6 ) dx 3 = 5 7 = x(x + y) dy dx  2 1 2 y=2x   2     x y + xy  dx   =    2 y=0  0 0 1 (2x4 + 2x5 ) dx 1 0 1 0 3 + = 2x5 x6 + 5 3 = 0 11 . y) is a constant function on the region R. ¯ ¯ . This makes sense since the density function δ(x. where there is quite a bit of area. We have M= R 1 2x2 125 y y = 2x2 R x 0 1 Figure 3. In the special case where the density function δ(x. y) dA 2x2 My = R 1 xδ(x.6. the center of mass ( x. y) is given by ¯ ¯ x= ¯ My 11/15 22 = = . 15 so the center of mass ( x.

Then the center of mass of S has coordinates ( x. then M = 2πa .6.3 a y δ(x. z) at any point (x. y.30) xδ(x.32) M= In this case. M y= ¯ M xz . y. y. which in spherical coordinates is S 2π π/2 0 2π π/2 0 a a = = 0 (ρ cos φ) ρ2 sin φ dρ dφ dθ sin φ cos φ 0 2π 0 π/2 0 a4 4 0 = = 0 ρ3 dρ dφ dθ sin φ cos φ dφ dθ . M z= ¯ M xy .31) (3. y.3). Example 3. y. M xz = S yδ(x. y. y. y. M (3.12 is 4πa . M xz and M xy are called the moments (or first moments) of S around the yz-plane. Myz . So since the density function is a constant and S is symmetric about the z-axis. x2 + y2 + z2 ≤ a2 }. y. z) is δ(x. Solution: The solid S is just the upper hemisphere inside the sphere of radius a centered at the origin (see Figure 3. z) dV = S 1 dV = Volume(S ). respectively. (3. z) dV . And 3 3 M xy = S zδ(x. S M xy = S zδ(x.14. z) dV z dV . y.126 CHAPTER 3. z) dV . y.6. MULTIPLE INTEGRALS The formulas for the center of mass of a region in 2 can be generalized to a solid S in 3 . But since the volume of S is half the volume of the sphere of 3 3 radius a. Find the center of mass of the solid S = {(x. M is the mass of S . z) ¯ ¯ ¯ 0 x Figure 3. xz-plane and xy-plane. y. then it is clear that x = 0 and y = 0. z) dV . z). where ¯ ¯ ¯ x= ¯ where Myz = S Myz . so we ¯ ¯ need only find z. We have ¯ M= S a z ( x . y. δ(x. z) in S . if the density function at (x. which we know by Example 3. z) = 1. Also. z) : z ≥ 0. z) dV . Let S be a solid with a continuous mass density function δ(x.

6. 0 ≤ y ≤ 1. y. y) : 0 ≤ x ≤ 1. z) : z ≥ 0. 0. 8 Thus. R = {(x. y. find the center of mass of the solid S with the given density function δ(x. y) : y ≥ 0. y. S = {(x. y) = x + y 3. y. R = {(x. z) = 1 9. y. R = {(x. z) : 0 ≤ x ≤ 1. z) : x ≥ 0. y. 0 ≤ z ≤ 1 }. x2 + y2 ≤ 1 }. y) = 5. R = {(x. z) : 0 ≤ x ≤ 1. x2 + y2 ≤ a2 }. δ(x. z) = 0. y. y ≥ 0. the center of mass of S is ( x. δ(x. 0 ≤ y ≤ 4 }. y) = 2y 2. z). y) : y ≥ 0. z) = x2 + y2 + z2 8. y) = 1 4.6 Application: Center of Mass 2π π/2 0 2π 4 127 M xy = 0 a4 8 sin 2φ dφ dθ (since sin 2φ = 2 sin φ cos φ) φ=π/2 φ=0 = 0 2π a − 16 cos 2φ a4 8 dθ = 0 dθ = so πa4 . S = {(x. δ(x. δ(x. 0 ≤ y ≤ 1. R = {(x. z) = xyz 7. y) : 0 ≤ x ≤ 2. ¯ ¯ ¯ 8  A Exercises ©  ¨ For Exercises 1-5. δ(x. 1 ≤ x2 + y2 ≤ 4 }. 1. 4 M xy = M πa4 4 2πa3 3 z= ¯ = 3a . y). δ(x. y. S = {(x. 0 ≤ z ≤ 1 − x − y}. δ(x. find the center of mass of the region R with the given density function δ(x. 0 ≤ y ≤ x2 }. z ≥ 0. y. x2 + y2 + z2 ≤ a2 }. S = {(x. y) = y x2 + y2 B For Exercises 6-10. y) : y ≥ 0. z) = x2 + y2 + z2 10. x ≥ 0. y. δ(x. δ(x. δ(x. z) : 0 ≤ x ≤ 1. x2 + y2 + z2 ≤ a2 }.3. y. 0 ≤ y ≤ 1. y. 3a . S = {(x. 0 ≤ z ≤ 1 }. z) = 1 .

the probability of an event will instead be the integral of a function. and hence in particular the 3 has a one out of six chance of being rolled.. . 1). For example. 1) has length 1. 1 is 6 . Note that P(X ≤ 3) = P(X = 1) + P(X = 2) + P(X = 3). which is given by P(X ≤ x) = x. 2. Theory of Sets. i. In particular we will see ways in which multiple integrals can be used to calculate probabilities and expected values. and hence is uniformly distributed over (0.3 In this case. then length of (0. Probability Suppose that you have a standard six-sided (fair) die. E. 1).e. in our case the event X ≤ x is the set (0. Then the probability of rolling a 3. MULTIPLE INTEGRALS 3. 5. For For a proof see p. 2. An event A is a subset of the sample space. . For example. and 3) that are less than or equal to 3. we consider the probability P(X ≤ x). is 3 = 1 . We call X a discrete random variable on the sample space (or probability space) Ω consisting of all possible outcomes. 1). we saw how the probability of an event was the sum of the probabilities of the individual outcomes comprising that event (e. length of (0. and for x in (0. Instead. Ω = {1.7 Application: Probability and Expected Value In this section we will briefly discuss some applications of multiple integrals in the field of probability theory. Now let X be a variable representing a random real number in the interval (0. Likewise the probability of rolling at most a 3. x). 9-10 in K AMKE . since there are six sides on the die and each one is equally likely to be rolled. 6}. 1). Note that the set of all real numbers between 0 and 1 is not a discrete (or countable) set of values. there are three equally likely numbers (1. for any real number x in (0. written as P(X ≤ 3).128 CHAPTER 3. 3}. it can not be put into a one-to-one correspondence with the set of positive integers. New York: Dover. which we will now describe. x) x = =x. 4. since of the six numbers 6 2 on the die. 1) 1 P(X ≤ x) = In the case of a discrete random variable. 1950. So since X represents a random number in (0. it makes no sense to consider P(X = x) since it must be 0 (why?). 3. and you let a variable X represent the value rolled. An event A is a subset of the sample space. Let X be a continuous real-valued random variable on a sample space Ω in 3 We call X a continuous random variable on the sample space Ω = (0. in the case of the die. In our case. written as P(X = 3).g. the event X ≤ 3 is the set {1. x) has length x. 1). 1) the interval (0. The reasoning is this: the interval (0. 2. For a continuous random variable. P(X ≤ 3) = P(X = 1) + P(X = 2) + P(X = 3) in the die example).

let Ω = (a. 1).f. a for a < x < b .     = P(X ≤ x). We thus have P(X ≤ x) = f (y) dy . and probability density function  1. (3. f (x) dx = 1 .42) .41) F(x) = P(X ≤ x) =  b−a .39) F(x) = P(X ≤ x) =  x.   1 .  ′ (3. (3.35) Suppose that there is a nonnegative. Let X represent a randomly selected real number in the interval (0. continuous real-valued function f on F(x) = and f (y) dy . for 0 < x < 1   f (x) = F (x) =  0. for x ≥ 1      (3. for x ≤ a .   1.  for x ≥ b    x  (3. then X has the uniform distribution function  1.34) such that (3.40) and probability density function In general.33) (3.15. for x ≤ 0 . elsewhere. We say that X has the uniform distribution on (0. we have F ′ (x) = f (x) . for −∞ < x < ∞ . for a < x < b   f (x) = F (x) =  b−a 0. (3. (3.  elsewhere. for a < x < b     0.38) Example 3. Define the distribution function F of X as F(x) = P(X ≤ x) . for 0 < x < 1     0. if X represents a randomly selected real number in an interval (a.7 Application: Probability and Expected Value simplicity. b). b). for short) for X. with distribution function  1. ′ (3.36) Then we call f the probability density function (or p.d. for −∞ < x < ∞. 1). by the Fundamental Theorem of Calculus.     0. x −∞ ∞ −∞ x 129 for −∞ < x < ∞ for x ≥ b for a < x < b for x ≤ a .37) Also.3.

we should have ∞ −∞ 2 1 √ e−x /2 dx = 1 2π (3. A famous distribution function is given by the standard normal distribution. 2π for −∞ < x < ∞. ∞ −∞ ∞ −∞ e−(x 2 +y2 )/2 dx dy = = = ∞ e−y 2 /2 ∞ −∞ e−x 2 /2 dx dy 2 /2 −∞ ∞ −∞ ∞ −∞ e−x e−x 2 /2 dx 2 ∞ −∞ e−y dy 2 /2 dx since the same function is being integrated twice in the middle equation. e−x 2 /2 dx = −∞ . First.36).44) by formula (3.130 CHAPTER 3.16.43) This is often called a “bell curve”. But using polar coordinates. MULTIPLE INTEGRALS Example 3. and hence √ 2π .. whose probability density function f is 2 1 f (x) = √ e−x /2 . which is equivalent to ∞ −∞ e−x 2 /2 dx = √ 2π . just with different variables. (3. and is used widely in statistics. 0 and so ∞ −∞ 2 e−x ∞ 2 /2 dx = 2π .f. (3.d. Since we are claiming that f is a p. we see that ∞ −∞ ∞ −∞ e−(x 2 +y2 )/2 2π dx dy = 0 2π 0 ∞ e−r 2 /2 r dr dθ dθ r=∞ = 0 2π −e−r 2 /2 r=0 2π = 0 (0 − (−e0 )) dθ = 1 dθ = 2π .45) We can use a double integral in polar coordinates to verify this integral.

3.7 Application: Probability and Expected Value

131

In addition to individual random variables, we can consider jointly distributed random variables. For this, we will let X, Y and Z be three real-valued continuous random variables defined on the same sample space Ω in (the discussion for two random variables is similar). Then the joint distribution function F of X, Y and Z is given by F(x, y, z) = P(X ≤ x, Y ≤ y, Z ≤ z) ,
z y −∞ ∞ −∞ x

for −∞ < x, y, z < ∞.
3

(3.46)

If there is a nonnegative, continuous real-valued function f on F(x, y, z) =
−∞ −∞ ∞ −∞ ∞ −∞

such that (3.47)

f (u, v, w) du dv dw ,

for −∞ < x, y, z < ∞

and

f (x, y, z) dx dy dz = 1 ,

(3.48)

then we call f the joint probability density function (or joint p.d.f. for short) for X, Y and Z. In general, for a1 < b1 , a2 < b2 , a3 < b3 , we have
b3 b2 a2 b1

P(a1 < X ≤ b1 , a2 < Y ≤ b2 , a3 < Z ≤ b3 ) =

f (x, y, z) dx dy dz ,
a3 a1

(3.49)

with the ≤ and < symbols interchangeable in any combination. A triple integral, then, can be thought of as representing a probability (for a function f which is a p.d.f.). Example 3.17. Let a, b, and c be real numbers selected randomly from the interval (0, 1). What is the probability that the equation ax2 + bx + c = 0 has at least one real solution x? Solution: We know by the quadratic formula that there is at least one real solution if b2 − 4ac ≥ 0. So we need to calculate P(b2 − 4ac ≥ 0). We will use three jointly distributed random variables to do this. First, since 0 < a, b, c < 1, we have √ √ b2 − 4ac ≥ 0 ⇔ 0 < 4ac ≤ b2 < 1 ⇔ 0 < 2 a c ≤ b < 1 , where the last relation holds for all 0 < a, c < 1 such that 0 < 4ac < 1 ⇔ 0 < c < 1 . 4a
c 1 c= R1 0
1 4 1 4a

R2 1

a

Figure 3.7.1 Region R = R1 ∪ R2

Considering a, b and c as real variables, the region R in the ac-plane where the above 1 relation holds is given by R = {(a, c) : 0 < a < 1, 0 < c < 1, 0 < c < 4a }, which we can see is a union of two regions R1 and R2 , as in Figure 3.7.1 above. Now let X, Y and Z be continuous random variables, each representing a randomly selected real number from the interval (0, 1) (think of X, Y and Z representing a, b and c, respectively). Then, similar to how we showed that f (x) = 1 is the p.d.f. of the

132

CHAPTER 3. MULTIPLE INTEGRALS

uniform distribution on (0, 1), it can be shown that f (x, y, z) = 1 for x, y, z in (0, 1) (0 elsewhere) is the joint p.d.f. of X, Y and Z. Now, √ √ P(b2 − 4ac ≥ 0) = P((a, c) ∈ R, 2 a c ≤ b < 1) , √ √ so this probability is the triple integral of f (a, b, c) = 1 as b varies from 2 a c to 1 and as (a, c) varies over the region R. Since R can be divided into two regions R1 and R2 , then the required triple integral can be split into a sum of two triple integrals, using vertical slices in R:
1/4 1 0 R1 1/4 1 0 1/4 1 √ √ 2 a c 1 1/4a 0 1/4a 0 1 √ √ 1 db dc da 2 a c

P(b2 − 4ac ≥ 0) =

1 db dc da +
1/4 R2 1 1/4 1 1/4 1 1/4 1 1/4 1 12a

0

=
0

√ √ (1 − 2 a c) dc da +
c=1 c=0

√ √ (1 − 2 a c) dc da
c=1/4a c=0

=
0 1/4

√ c − 4 a c3/2 3 1−
4 3

da + da

c−

4 3

√ 3/2 ac

da

=
0

√ a da + + 1 ln a 12

8 = a − a3/2 9

1/4 0

1 1 1 1 1 5 = − ln + ln 4 + 0− = 4 9 12 4 36 12 5 + 3 ln 4 ≈ 0.2544 P(b2 − 4ac ≥ 0) = 36 In other words, the equation ax2 + bx + c = 0 has about a 25% chance of being solved!

Expected Value
The expected value EX of a random variable X can be thought of as the “average” value of X as it varies over its sample space. If X is a discrete random variable, then EX =
x

x P(X = x) ,

(3.50)

with the sum being taken over all elements x of the sample space. For example, if X represents the number rolled on a six-sided die, then
6 6

EX =
x=1

x P(X = x) =
x=1

x

1 = 3.5 6

(3.51)

is the expected value of X, which is the average of the integers 1 − 6.

3.7 Application: Probability and Expected Value If X is a real-valued continuous random variable with p.d.f. f , then EX =
∞ −∞

133

x f (x) dx .

(3.52)

and so

1 . (3.54) 2 −∞ 0 For a pair of jointly distributed, real-valued continuous random variables X and Y with joint p.d.f. f (x, y), the expected values of X and Y are given by EX = x f (x) dx = x dx = EX = respectively.
∞ −∞ ∞ −∞

For example, if X has the uniform distribution on the interval (0, 1), then its p.d.f. is  1, for 0 < x < 1   f (x) =  (3.53) 0, elsewhere, 
∞ 1

x f (x, y) dx dy and

EY =

∞ −∞

∞ −∞

y f (x, y) dx dy ,

(3.55)

Example 3.18. If you were to pick n > 2 random real numbers from the interval (0, 1), what are the expected values for the smallest and largest of those numbers? Solution: Let U1 , . . . , Un be n continuous random variables, each representing a randomly selected real number from (0, 1), i.e. each has the uniform distribution on (0, 1). Define random variables X and Y by X = min(U1 , . . . , Un ) and Y = max(U1 , . . . , Un ) .

Then it can be shown4 that the joint p.d.f. of X and Y is  n(n − 1)(y − x)n−2 , for 0 ≤ x ≤ y ≤ 1   f (x, y) =  0,  elsewhere. Thus, the expected value of X is
1 1

(3.56)

EX =
0 1 x

n(n − 1)x(y − x)n−2 dy dx
y=1 y=x

=
0 1

nx(y − x)n−1

dx

=
0

nx(1 − x)n−1 dx , so integration by parts yields 1 (1 − x)n+1 n+1
1 0

= −x(1 − x)n − 1 , EX = n+1
4

See Ch. 6 in H OEL, P ORT and S TONE.

134

CHAPTER 3. MULTIPLE INTEGRALS

and similarly (see Exercise 3) it can be shown that
1 y 0

EY =
0

n(n − 1)y(y − x)n−2 dx dy =

n . n+1

So, for example, if you were to repeatedly take samples of n = 3 random real numbers from (0, 1), and each time store the minimum and maximum values in the sample, 1 then the average of the minimums would approach 4 and the average of the max3 imums would approach 4 as the number of samples grows. It would be relatively simple (see Exercise 4) to write a computer program to test this. 

B
1. Evaluate the integral
∞ −x2 e −∞

¨

dx using anything you have learned so far.
2 2 ∞ 1 √ e−(x−µ) /2σ −∞ σ 2π

2. For σ > 0 and µ > 0, evaluate 3. Show that EY =
n n+1

dx.

in Example 3.18

C
4. Write a computer program (in the language of your choice) that verifies the results in Example 3.18 for the case n = 3 by taking large numbers of samples. 5. Repeat Exercise 4 for the case when n = 4. 6. For continuous random variables X, Y with joint p.d.f. f (x, y), define the second moments E(X 2 ) and E(Y 2 ) by E(X 2 ) =
∞ −∞ ∞ −∞

x2 f (x, y) dx dy and

E(Y 2 ) =

∞ −∞

∞ −∞

y2 f (x, y) dx dy ,

and the variances Var(X) and Var(Y) by Var(X) = E(X 2 ) − (EX)2 and Var(Y) = E(Y 2 ) − (EY)2 .

Find Var(X) and Var(Y) for X and Y as in Example 3.18. 7. Continuing Exercise 6, the correlation ρ between X and Y is defined as ρ=
∞ ∞

E(XY) − (EX)(EY) Var(X) Var(Y)

,

where E(XY) = −∞ −∞ xy f (x, y) dx dy. Find ρ for X and Y as in Example 3.18. (Note: The quantity E(XY) − (EX)(EY) is called the covariance of X and Y.) 8. In Example 3.17 would the answer change if the interval (0, 100) is used instead of (0, 1)? Explain.

y) which varies with the position (x. since an interval (or collection of intervals) is really the only kind of “path” in 1 . we will see how to define the integral of a function (either realvalued or vector-valued) of two variables over a general path (i. for some integer n ≥ 2 135 . y = y(t) for t in [a. a curve) in 2 .e. b].1 below). Suppose that we want to find the total amount W of work done in moving an object along a curve C in 2 with a smooth parametrization x = x(t). This definition will be motivated by the physical notion of work. y) of the object and is applied in the direction of motion along C (see Figure 4.1 Curve C : x = x(t). b] in 1 .1 Line Integrals In single-variable calculus you learned how to integrate a real-valued function f (x) over an interval [a. the intuitive idea of work is that Work = Force × Distance . so we only consider the magnitude of the force. b] We will assume for now that the function f (x.1. y = y(t). b] as follows: a = t0 < t1 < t2 < · · · < tn−1 < tn = b . This integral (usually called a Riemann integral) can be thought of as an integral over a path in 1 .1. with a force f (x. y) is continuous and real-valued. In physics. then the work W done in moving that object from position x = a to x = b was defined as the integral: b W= a f (x) dx In this section. y C t=a t = ti ∆yi ∆si ≈ ∆xi ∆xi 2 + ∆yi 2 t = ti+1 t=b x 0 Figure 4. a ≤ t ≤ b. Partition the interval [a. You may also recall that if f (x) represented the force applied along the x-axis to an object at position x in [a. We will begin with real-valued functions of two variables.4 Line and Surface Integrals 4.

y) along C with respect to arc length s is b f (x. y) ds = C a f (x(t). (4. y). y(ti ∗)) for some ti ∗ in [ti . the line integral of f (x. (4.1. ti+1 ] the distance ∆si traveled along the curve is approximately ∆xi 2 + ∆yi 2 . Thus. y) and a curve C in 2 . (4. yi∗ ) i=0 ∆xi ∆ti 2 + ∆yi ∆ti 2 ∆ti . parametrized by x = x(t).1. called a line integral: Definition 4.136 CHAPTER 4. (4. y) along the curve C. yi∗ ) ∆xi 2 + ∆yi 2 . y = y(t). and f (xi∗ .3) Taking the limit of that sum as the length of the largest subinterval goes to 0. by the Pythagorean Theorem. yi∗ ) i=0 ∆xi 2 + ∆yi 2 (4.6) . over a typical subinterval [ti . For a real-valued function f (x. LINE AND SURFACE INTEGRALS As we can see from Figure 4. then n−1 ∆xi ∆ti 2 + ∆yi ∆ti 2 ∆ti . the integral of f (x.2) is approximately the total amount of work done over the entire curve. ∆xii and ∆yii become x ′ (t) ∆t ∆t and y ′ (t). But since ∆xi 2 + ∆yi 2 = where ∆ti = ti+1 − ti . respectively. y(t)) x ′ (t)2 + y ′ (t)2 dt . ti+1 ]. y(t)) x ′ (t)2 + y ′ (t)2 dt . yi∗ ) becomes f (x(t). so that b W= a f (x(t). the sum over all subintervals becomes the integral from t = a to t = b.1) where (xi∗ . for any real-valued function f (x.1. if the subinterval is small enough then the work done in moving the object along that piece of the curve is approximately Force × Distance ≈ f (xi∗ . (4. y(t)).5) The symbol ds is the differential of the arc length function t s = s(t) = a x ′ (u)2 + y ′ (u)2 du .4) The integral on the right side of the above equation gives us our idea of how to define. and so n−1 W≈ f (xi∗ . a ≤ t ≤ b. yi∗ ) = (x(ti ∗). W≈ f (xi∗ .

Use a line integral to show that the lateral surface area A of a right circular cylinder of radius r and height h is 2πrh. C y f (x. y) y 0 x C : x2 + y2 = r2 Figure 4. y). (4. y). ds = s ′ (t) dt = x ′ (t)2 + y ′ (t)2 dt .2 Area of shaded rectangle = height × width ≈ f (x.1. y = y(t) = r sin t .4. y) = h for all (x. then f (x. y) ds = C 2π a f (x(t). You can think of differentials as infinitesimal lengths.1. y) ds can be thought of as approximately the area of a section of that fence over some infinitesimally small section of the curve. for all t in [a. Solution: We will use the right circular cylinder with base circle C given by x2 + y2 = r2 and with height h in the positive z direction (see Figure 4. So if you think of f (x. y(t)) x ′ (t)2 + y ′ (t)2 dt h (−r sin t)2 + (r cos t)2 dt 2π =h 0 r sin2 t + cos2 t dt 2π = rh 0 1 dt = 2πrh . y) as the height of a picket fence along C.1. y) ds is the total area of that picket fence (see Figure 4. That is.7) by the Fundamental Theorem of Calculus.3 Let f (x.3). t]. 0 ≤ t ≤ 2π z r h = f (x. y) ds represent? The preceding discussion of ds gives us a clue.9 as the length of the curve C over the interval [a. Parametrize C as follows: x = x(t) = r cos t .1.1. For a general real-valued function f (x. y) C ds x 0 Figure 4. Then b A= = 0 f (x. and thus the line integral f (x. y) ds Example 4.2).1 Line Integrals 137 which you may recognize from Section 1. b]. what does the line integral C f (x.

e. twice the desired area.1 that if we had traversed the circle C twice. even though the curve itself is still the same (namely. y = y(t). then denote by −C the same curve as C but traversed in the opposite direction. and b f (x. However.12) as the line integral of f (x. and its values are vectors in 2 .138 CHAPTER 4. we used the idea of work as force multiplied by distance. 0 ≤ t ≤ 2π . If a curve C has a parametrization x = x(t). y) on 2 . y) = P(x. For this. let t vary from 0 to 4π. a≤t≤b. y) and Q(x. So it would be helpful to develop a vector form for a line integral. then we would have gotten an area of 4πrh. (4. y) along C with respect to y. y) defined on 2 by f(x. using the parametrization x = x(t) = r cos(2π − t) . y) ds unchanged. We can also define b f (x. it can be shown (see Exercise 15) that reversing the direction in which a curve C is traversed leaves C f (x. y) dx = C a f (x(t). notice that we traversed the circle in the counter-clockwise direction. Then −C is parametrized by x = x(a + b − t) . i. It is defined at points in 2 . y(t)) x ′ (t) dt (4. (4. a circle of radius r). a ≤ t ≤ b.9) f (x. In the derivation of the formula for a line integral. for any f (x. i.11) as the line integral of f (x. If we had gone in the clockwise direction. LINE AND SURFACE INTEGRALS Note in Example 4. y = y(t). y) ds = C −C y = y(a + b − t) . (4. For a curve C with a smooth parametrization x = x(t). y = y(t) = r sin(2π − t) . y) ds . In general. y). y) dy = C a f (x(t). Such a function f is called a vector field on 2 .e. a ≤ t ≤ b.8) then it is easy to verify (see Exercise 12) that the value of the line integral is unchanged. we know that force is actually a vector. Also. let r(t) = x(t) i + y(t) j .10) Notice that our definition of the line integral was with respect to the arc length parameter s. y) j for some continuous real-valued functions P(x. suppose that we have a function f(x. y(t)) y ′ (t) dt (4. y) along C with respect to x. and we have f (x. y) i + Q(x.

so the last integral on the right looks somewhat similar to our earlier definition of a line integral. We use the notation dr = r ′ (t) dt = dx i + dy j to denote the differential of the vectorvalued function r. Notice that the function f(x(t). y = y(t). y) j and a curve C with a smooth parametrization x = x(t). y) dy is called exact if it equals dF for some function F(x.14) f(x(t). y) dy = C P(x. The quantity P(x. a ≤ t ≤ b. b] and hence T(t) = r ′ (t) r ′ (t) is the unit tangent vector to C at (x(t). For convenience we will often write P(x. y) = P(x. y). This leads us to the following definition: Definition 4. then r ′ (t) is a tangent vector to C at the point (x(t). y) dx + Q(x. the line integral of f along C is C f · dr = = P(x. y(t)) on C. the differential of F is dF = ∂F dx + ∂F dy. y) dy = a b P(x(t). Putting Definitions 4. y(t)) · r ′ (t) dt = a by definition of f(x. Recall that if the points on a curve C have position vector r(t) = x(t) i+y(t) j. where it is understood that the line integral along C is being applied to both P and Q. The line integral in Definition 4.13) (4. For a vector field f(x. y) dx + Q(x. where r(t) = x(t) i + y(t) j is the position vector for points on C. Since C is a smooth curve. y(t)) x ′ (t) dt + a Q(x(t). y) dx + C C Q(x. y) dx + C C Q(x. y). y(t)) y ′ (t) dt = a b (P(x(t). y(t)) in the direction of increasing t (which we call the direction of C). y) dy (4. y(t)) x ′ (t) + Q(x(t). y) dx + C b a C Q(x. Then r ′ (t) = x ′ (t) i + y ′ (t) j and so b b 139 P(x. b]. y) dx + Q(x. y). y(t)). y(t)) · r ′ (t) is a real-valued function on [a. y) dy .2 is often called a line integral of a vector field to distinguish it from the line integral in Definition 4. For a realvalued function F(x. then r ′ (t) 0 on [a.1 Line Integrals be the position vector for a point (x(t).2 together we get the following theorem: . y(t)) y ′ (t)) dt f(x(t).1 and 4. A differential form ∂x ∂y P(x. y(t)) · r ′ (t) dt . y) dy is known as a differential form.4. y) i + Q(x.1 which is called a line integral of a scalar field.2.

1.1. C f · dr = C f · T ds . y(t)). If the vector field f(x. then 1 (x2 + y2 ) dx + 2xy dy = C 0 1 (x(t)2 + y(t)2 )x ′ (t) + 2x(t)y(t) y ′ (t) dt x = 0 1 (t + 4t )(1) + 2t(2t)(2) dt 13t dt 0 2 2 2 0 1 Figure 4. y = 2t2 .140 CHAPTER 4.15) where T(t) = r ′ (t) r ′ (t) is the unit tangent vector to C at (x(t). (4. where: 0≤t≤1 0≤t≤1 y (1.1. y = y(t). 2) 2 Solution: Figure 4.4 = = (b) Since x ′ (t) = 1 and y ′ (t) 13t3 3 1 = 0 13 3 = 4t. (4.16) Example 4. C (x2 + y2 ) dx + 2xy dy.4 shows both curves. (b) C : x = t . Evaluate (a) C : x = t . LINE AND SURFACE INTEGRALS Theorem 4. then 1 (x2 + y2 ) dx + 2xy dy = C 0 1 (x(t)2 + y(t)2 )x ′ (t) + 2x(t)y(t) y ′ (t) dt (t2 + 4t4 )(1) + 2t(2t2 )(4t) dt 0 1 = = = 0 3 t (t2 + 20t4 ) dt 1 3 + 4t 5 0 = 1 13 +4= 3 3 . a ≤ t ≤ b and position vector r(t) = x(t) i + y(t) j. y) = P(x. then the work W done by this force is W= C f · T ds = C f · dr .2. y) i + Q(x. y) j and a curve C with a smooth parametrization x = x(t). (a) Since x ′ (t) = 1 and y ′ (t) = 2. y = 2t . For a vector field f(x. y) represents the force moving an object along a curve C.

the formulation C f · T ds from Theorem 4. if C is a piecewise smooth curve. C (x2 + y2 ) dx + 2xy dy. then the work done is 13 . as we will see in the next section.1 is often preferred in physics since it · emphasizes the idea of integrating the tangential component f· T of f in the direction of T (i. 2) along the given curve C. 2). 0) to (1. . the line integral of a vector field) is independent of the path taken. . if the vector field f(x. However. . 0 ≤ t ≤ 2 and C2 is the curve given by x = t.1. . where C1 is the curve given by x = 0. 2) to (1. and the letter l signifies length. Example 4. y) = (x2 + y2 ) i + 2xy j represents the force moving an object from (0.e.1 Line Integrals 141 So in both cases. where it is understood that the limits of integration a and b are for the underlying parameter t of the curve. .5). y = 2. + f · drn C C1 C2 Cn where each ri is the position vector of the curve Ci . . 2) (x2 + y2 ) dx + 2xy dy (x2 + y2 ) dx + 2xy dy C2 2 + = 0 (02 + t2 )(0) + 2(0)t(1) dt + 0 2 1 (t2 + 4)(1) + 2t(2)(0) dt = = 0 3 t 0 dt + 0 1 (t2 + 4) dt 13 1 +4= 3 3 3 + 4t 0 = Line integral notation varies quite a bit. Cn . this is not always the case. y = t. . Also. .5 1 Solution: Write C = C1 ∪ C2 .3. where C is the polygonal path from (0.1. 3 This may lead you to think that work (and more generally.4. ∪ Cn is the union of smooth curves C1 . which is a useful physical interpretation of line integrals. Although we defined line integrals over a single smooth curve. then we can define f · dr = f · dr1 + f · dr2 + . in physics it is common to b see the notation a f · dl. that is C = C1 ∪ C2 ∪ . 0 ≤ t ≤ 1 (see Figure 4. . For example. Evaluate to (0. Then (x2 + y2 ) dx + 2xy dy = C C1 (1. in the direction of C). 0) y 2 C2 C1 x 0 1 Figure 4.

10. 2) 4. y = sin t. y) and curve C. y) = x i + y j.) 16. f (x. 1.8). C: path from (2. C 15. 0) to (3. 9. y) j is a vector field such that f(x. calculate 6. 0) along the x-axis 5. 0) counterclockwise along the circle x2 + y2 = 4 to the point (−2. y) ds. f(x. f (x.9). 8. C : x = t. y) i + Q(x. 14. 0) to (3. 0 ≤ t ≤ 2π C : the polygonal path from (0. then C f · dr = 0. then C f · dr = C f ds. y = sin t. Use a line integral to find the lateral surface area of the part of the cylinder x2 + y2 = 4 below the plane x + 2y + z = 6 and above the xy-plane. C: polygonal path from (0. 13. 0 ≤ t ≤ 2π C : x = cos t. y = sin t. 0) and then back to (2. y) = 2x + y. C : x = 3t. f(x. f (x. f(x. . y) on C. y) = y i − x j. Prove that the Riemann integral b a f (x) dx is a special case of a line integral.) C 17. y) ≤ M for all (x. y) = xy2 i + xy3 j. Show that if f ⊥ r ′ (t) at each point r(t) along a smooth curve C. Show that if f points in the same direction as r ′ (t) at each point r(t) along a smooth curve C. y) and curve C. y) ds for the given function f (x. Prove that C f (x. f(x. For Exercises 6-11. (Hint: Use formulas (4. (Hint: Recall that a g(x) dx ≤ a |g(x)| dx for Riemann integrals. y = sin t. 0) to (1.142 CHAPTER 4. y) = xy. y) = (x2 + y2 ) i. y) = 2 x +1 3. y) = i − j. LINE AND SURFACE INTEGRALS  A For Exercises 1-4. f(x. 0 ≤ t ≤ 1 C : x = cos t. 0 ≤ t ≤ 1 2. y) = x + y2 . Verify that the value of the line integral in Example 4. f(x. 0) C : x = 2 + cos t. Let C be a smooth curve with arc length L. B 12. y = 0. y) = (x2 − y) i + (x − y2 ) j. 0 ≤ t ≤ 2π C : x = cos t. 0) to (0. 11. f (x. calculate C Exercises ©  ¨ f (x. y = 2t. 0 ≤ t ≤ π/2 x .1 is unchanged using the parametrization of the circle C given in formulas (4. y) ds = −C f (x. y = sin t. C : x = cos t. 1) to (0. C f · dr for the given vector field f(x. Show that b b f · dr ≤ ML. and suppose that f(x. 0 ≤ t ≤ 2π 7. y) = P(x.

y(u)) x ′ (u) du b a b =− −C P(x(u). reversing the direction in which the integral is taken along a curve does not change the value of the line integral: f (x. y) dx + − P(x. y) dx C since we are just using a different letter (u) for the line integral along C. y) dx + Q(x. y) ds = C −C f (x. (4. y) i + Q(x.2 Properties of Line Integrals We know from the previous section that for line integrals of real-valued functions (scalar fields). however. so a P(x. y) dy C C −C f · dr = − C f · dr . a ≤ t ≤ b. y(u)) x ′ (u) du . y) = P(x. y) dy C P(x. y) dy = − Q(x. y = y(t). Let C be a smooth curve parametrized by x = x(t). To see this. Then b P(x. let f(x. since a b =− .17) For line integrals of vector fields. y(a + b − t)) (−x ′ (a + b − t)) dt (by the Chain Rule) P(x(u). with P and Q continuously differentiable functions. a ≤ t ≤ b. C and hence −C f · dr = =− =− P(x. with position vector r(t) = x(t) i + y(t) j (we will usually abbreviate this by saying that C : r(t) = x(t) i + y(t) j is a smooth curve). the value does change. y) dx + −C C −C Q(x. y = y(a + b − t). A similar argument shows that −C Q(x. We know that the curve −C traversed in the opposite direction is parametrized by x = x(a + b − t). y) dx = −C a b P(x(a + b − t).2 Properties of Line Integrals 143 4. y) dy . y) dy Q(x. y) dx = − P(x. y) j be a vector field. y(u)) (−x ′ (u)) (−du) (by letting u = a + b − t) a = b = b P(x(u). y) ds (4. y(a + b − t)) d (x(a + b − t)) dt dt = a a P(x(a + b − t).4.18) .

y = y(t).e. y(t)) x (t) dt = a ′ α−1 (b) P(x(α(u)). But as we know. y) (treated as a vector) moving an object along a curve C: the total work performed moving the object along C from its initial point to its terminal point. say. c ≤ u ≤ d. and du = α ′1 . That was not the case with the “reverse” parametrization for −C: for u = a + b − t we have t = α(u) = a + b − u ⇒ α ′ (u) = −1 < 0. Recall that our definition of a line integral required that we have a parametrization x = x(t). a ≤ t ≤ b. d) (i. LINE AND SURFACE INTEGRALS The above formula can be interpreted in terms of the work done by a force f(x. ˜ ˜ ˜ which shows that C P(x. y) j be a vector field.2 means that the two parametrizations move along C in the same direction. y) dy has the same value for both parametrizations. b] such that c = α−1 (a). So could we get a different value for a line integral using ˜ ˜ some other parametrization of C. this would mean that our definition is not well-defined. dt = α ′ (u) du. d]). y = y(u). y = y(t). and then back to the initial point moving backwards along the same path. The preceding discussion shows the importance of always taking the direction of the curve into account when using line integrals of vector fields. c ≤ u ≤ d ? If so. b]. the curves in line integrals are sometimes referred to as directed curves or oriented curves. QED and hence C f · dr has the same value. ˜ ˜ Proof: Since α(u) is strictly increasing and maps [c. Also. y) i + Q(x. y(u)) x ′ (u) du . Notice that the condition α ′ (u) > 0 in Theorem 4. b = α(d). This is because when force is considered as a vector. For this reason.144 CHAPTER 4.2. direction is accounted for. α(u) is strictly · increasing on [c. such that a = α(c). y(α(u))) α−1 (a) d x ′ (u) ′ ˜ (α (u) du) α ′ (u) = c P( x(u). Luckily. is zero. Then C f· dr has the same value for the parametrizations x = x(t). and let C be a smooth curve parametrized by x = x(t). y = y(u) = y(α(u)). y = y(t). x = x(u). d = α−1 (b). Let f(x. A similar argument shows that C Q(x. and α ′ (u) > 0 on the open interval (c. any curve has infinitely many parametrizations. y) dx has the same value for both parametrizations. a ≤ t ≤ b for the curve C. Suppose that t = α(u) for c ≤ u ≤ d. a ≤ t ≤ b and x = x(u) = x(α(u)). . and by the Chain Rule dt (u) x ′ (u) = ˜ d dx dt dx ˜ = (x(α(u))) = = x ′ (t) α ′ (u) du du dt du ⇒ x ′ (t) = x ′ (u) ˜ α ′ (u) so making the susbstitution t = α(u) gives b P(x(t). it turns out that the value of a line integral of a vector field is unchanged as long as the direction of the curve C is preserved by whatever parametrization is chosen: Theorem 4. then we know that t = α(u) has an inverse function u = α−1 (t) defined on [a. d] onto [a. y) = P(x.

y = 2 sin2 u . Section 4.2 we know that if C is parametrized by x = sin u .1.4. a ≤ t ≤ b. namely 13 . We use the special notation f (x. π/2). along the curve C : x = t. i.2 Properties of Line Integrals 145 Example 4. In some older texts you may see the notation traversing a closed curve in a counterclockwise or clockwise direction. So 0 ≤ u ≤ π/2 then C (x2 + y2 ) dx + 2xy dy should have the same value as we found in Example 4. 1 = sin(π/2).2. Evaluate the line integral C (x2 +y2 ) dx+2xy dy from Example 4. ◮ t=a t=b t=a ◮ ◭ C (a) Closed t=b ◭ C (b) Not closed Figure 4. the line integral is unchanged whether t or u is the parameter for C.e. Solution: First. y(a)) = (x(b). we notice that 0 = sin 0. along closed curves. And we can indeed verify this: 3 π/2 (x2 + y2 ) dx + 2xy dy = C 0 π/2 (sin2 u + (2 sin2 u)2 ) cos u + 2(sin u)(2 sin2 u)4 sin u cos u du sin2 u + 20 sin4 u cos u du 0 π/2 = = sin3 u + 4 sin5 u 3 0 1 13 = +4= 3 3 In other words. y = 2t2 . .2. respectively. 0 ≤ t ≤ 1. we have (x(a). y) ds and C C f · dr or to indicate a line integral to denote line integrals of scalar and vector fields. and by Theorem 4. y(b)). y = y(t).1 Closed vs nonclosed curves A simple closed curve is a closed curve which does not intersect itself. we mean a curve C whose initial point and terminal point are the same. where t = sin u for 0 ≤ u ≤ π/2. for C : x = x(t). By a closed curve. respectively. dt du = cos u > 0 on (0.2.4. Note that any closed curve can be regarded as a union of simple closed curves (think of the loops in a figure eight).

the line integral has been independent of the path joining the two points. and let C2 be the remaining part of C that goes from P1 to P2 .2. Then by path independence we have f · dr = f · dr = 0 f · dr = 0 .g. the examples we have seen of line integrals (e. and so C f · dr = 0. Let C be a closed curve contained in R. Thus. Proof: Suppose that C f · dr = 0 for every closed curve C which is contained in R. and so f · dr = C2 f · dr.146 CHAPTER 4.3. That is. the line integral C f · dr is independent of the path between any two points in R if and only if C f · dr = 0 for every closed curve C which is contained in R. this is not always the case. and let C2 be another curve in R going from P1 to P2 . Example 4.2) have had the same value for different curves joining the initial point to the terminal point. Let C1 be a curve in R going from P1 to P2 . as in Figure 4. the above theorem does not give a practical way to determine path inde- .2. Then C = C1 ∪ −C2 is a closed curve in R (from P1 to P1 ). Conversely. so f · dr = 0 QED C1 C2 f · dr C1 f · dr − C2 C1 f · dr + −C2 C since C = C1 ∪ −C2 . As we mentioned before. LINE AND SURFACE INTEGRALS So far. In a region R.2. again as in Figure 4. suppose that the line integral C f · dr is independent of the path between any two points in R. Let C1 be a part of the curve C that goes from P1 to P2 . This proves path independence.2 = C1 C1 f · dr . 0= C C1 ◮ P1 P2 f · dr f · dr + f · dr − −C2 C2 = C1 f · dr ◮ C2 Figure 4. Clearly. Let P1 and P2 be two distinct points on C.2. The following theorem gives a necessary and sufficient condition for this path independence: Theorem 4. Let P1 and P2 be two distinct points in R.2.

5.4) = a = F(x(t).4 (which uses the Mean Value Theorem). y(a)) and B = (x(b). we have f · dr = = P(x(t). so we omit it.19) f · dr = F(B) − F(A) . Then C (4. y) is a continuously differentiable function of x and y. y) = P(x. (Chain Rule) If z = f (x.1 We will now use this Chain Rule to prove the following sufficient condition for path independence of line integrals: Theorem 4.4. What it mostly does is give an idea of the way in which line integrals behave. the line integral is independent of the path between its endpoints. 1 See T AYLOR and M ANN. (4. y) j be a vector field in some region R.5. Let f(x. y) i + Q(x.2 from Section 2. and how seemingly unrelated line integrals can be related (in this case. The proof is virtually identical to the proof of Theorem 2. Thus. and both x = x(t) and y = y(t) are differentiable functions of t. For a more practical method for determining path independence.2 Properties of Line Integrals 147 pendence. y(t)) x ′ (t) + Q(x(t). Proof: By definition of b C C f · dr.20) where A = (x(a).4. y(t)) dt (by the Chain Rule in Theorem 4. y(b)) are the endpoints of C. y = y(t). then z is a differentiable function of t. a specific line integral between two points and all line integrals around closed curves). Suppose that there is a real-valued function F(x. y(t)) b a = F(B) − F(A) QED by the Fundamental Theorem of Calculus. we first need a version of the Chain Rule for multivariable functions: Theorem 4. a ≤ t ≤ b. y(t)) y ′ (t) dt a b a b ∂F ∂F ∂F dx ∂F dy dt (since ∇F = f ⇒ + = P and = Q) ∂x dt ∂y dt ∂x ∂y F ′ (x(t). since it is impossible to check the line integrals around all possible closed curves in a region. since it depends only on the values of F at those endpoints. y) such that ∇F = f on R. Let C be a smooth curve in R parametrized by x = x(t). § 6. and dz ∂z dx ∂z dy = + dt ∂x dt ∂y dt at all points where the derivatives on the right are defined. with P and Q continuously differentiable functions on R. .

If a vector field f has a potential in a region R. then C f · dr = 0 for any closed curve C in R (i. y) = x i + y j has a potential F(x. where ∂y ∂y K is a constant. y) = 1 3 x + xy2 . C ∇F · dr = 0 for any real-valued function F(x. since by Theorem 4. Evaluate C x dx + y dy for C : x = 2 cos t. so ∂x 2 ∂F 1 = y ⇒ g ′ (y) = y ⇒ g(y) = y2 + K ∂y 2 . Note that we can also verify that the value of the line integral of f along any curve C going from (0. 2) − F(0. y) for f(x. a potential F(x. LINE AND SURFACE INTEGRALS Theorem 4. Since any choice for K will do (why?). 0) to the point (1. we pick K = 0. ∂y Suppose that ∂F = x2 + y2 . y)).3 in Section 4. 0) to (1.6. is the following important corollary: Corollary 4. g(y) = K.2 and 4. A real-valued function F(x. y) = 1 x3 + xy2 + g(y) for some function ∂x 3 g(y). Recall from Examples 4.1 that the line integral (x2 + y2 ) dx + 2xy dy was found to have the value 13 for three different curves C going 3 C from the point (0. y): 1 ∂F = x ⇒ F(x. Solution: We need to find a real-valued function F(x. 0 ≤ t ≤ 2π. 2) will always be 13 . y) = x2 + g(y) .e. 0) = (1)3 + (1)(2)2 − (0 + 0) = + 4 = 3 3 3 C A consequence of Theorem 4.5 3 1 13 1 .e. namely F(x. Thus. 3 Hence the line integral C (x2 + y2 ) dx + 2xy dy is path independent. y) = (x2 + y2 ) i + 2xy j exists. y = 3 sin t. f · dr = F(1.5 in the special case where C is a closed curve. Then we must have F(x. A conservative vector field is one which has a potential. y) such that ∇F(x. y) is called a potential for f.148 CHAPTER 4. Example 4. Use Theorem 4.5 can be thought of as the line integral version of the Fundamental Theorem of Calculus. Solution: The vector field f(x. so that the endpoints A and B are the same point. i. y) = f(x.5 to show that this line integral is indeed path independent. Example 4.5. y) such that ∂F = x2 + y2 ∂x and ∂F = 2xy . 2).6. So ∂F = 2xy + g ′ (y) satisfies the condition ∂F = 2xy if g ′ (y) = 0.

y) = f · dr = 0 x2 4 by Corollary 4. 0 ≤ t ≤ 2π. y) = −y x2 +y2 i+ x x2 +y2 j for all (x. Let f(x. (Hint: Consider the mixed partial derivatives of F. and let C be a curve in 2 .) 9. and let f(x. Can f have a potential F(x. Is there a potential F(x.6. B 6. A 1. 7. y) = tan−1 (y/x). and C : x = cos t. 0 ≤ t ≤ π. 2 2 x dx + y dy = C C 149 for any constant K. Show that C (a f ± b g) · dr = a C f · dr ± b C C g · dr . y = sin t. (b) Show that C f · dr = 2π. y)? If so. y) = y i − x j? If so. y) for f(x. y) (0. Evaluate C 3. 0 ≤ t ≤ 2π. y) = h(y) i + g(x) j. 4. Let f(x.2 Properties of Line Integrals 1 2 1 2 x + y is a potential for f(x. y) and g(x. 5. Evaluate C Exercises ©  (x2 + y2 ) dx + 2xy dy for C : x = cos t. find one. y) for f(x. (Hint: Use Exercise 21 in Section 2. 2.) . Let C be a curve whose arc length is L. so F(x. y = sin t. Does this contradict Corollary 4. y) be continuously differentiable real-valued functions in a region R. y) for f(x. You may assume that F would be smooth. Show that f ∇g · dr = − g ∇ f · dr C C for any closed curve C in R.4. 0). Let f (x. find one. y) be vector fields. 8. Show that 1 ds = L. for F(x. (a) Show that f = ∇F. y = sin t. y) = xy2 i + x3 y j? If so. y) and g(x.4. find it. (x2 + y2 ) dx + 2xy dy for C : x = cos t. since the curve C is closed (it is the ellipse  ¨ + y2 9 = 1). Thus.6? Explain. C 10. Let g(x) and h(y) be differentiable functions. y) = x i − y j? If so. y). let a and b be constants. find one. Is there a potential F(x. Is there a potential F(x.

A vector field f(x.3.23) where X1 and X2 are the points on C farthest to the left and right.150 CHAPTER 4. Then f · dr = R C ∂Q ∂P − dA . y) j is smooth if its component functions P(x. y) i + Q(x. (Green’s Theorem) Let R be a region in 2 whose boundary is a simple closed curve C which is piecewise smooth.1. y) are smooth.22) (4. y) = P(x.25) where Y1 and Y2 are the lowest and highest points. See Figure 4. ∂x ∂y (4. and (4. y) j be a smooth vector field defined on both R and C. C1 = the curve x = x1 (y) from the point Y2 to the point Y1 C2 = the curve x = x2 (y) from the point Y1 to the point Y2 . respectively.21) where C is traversed so that R is always on the left side of C. Let f(x. y) and Q(x.7. y) = P(x. Proof: We will prove the theorem in the case for a simple region R.23) and .3 Green’s Theorem We will now see a way of evaluating the line integral of a smooth vector field around a simple closed curve. y y = y2 (x) d ◭ x = x1 (y) X1 Y1 Y2 X2 x = x2 (y) R ◮ C x c y = y1 (x) a Figure 4. where the boundary curve C can be written as C = C1 ∪ C2 in two distinct ways: C1 = the curve y = y1 (x) from the point X1 to the point X2 C2 = the curve y = y2 (x) from the point X2 to the point X1 . (4. We will use Green’s Theorem (sometimes called Green’s Theorem in the plane) to relate the line integral around a closed curve with a double integral over the region inside the curve: Theorem 4.3. y) around C using the representation C = C1 ∪ C2 given by (4.24) (4. respectively. LINE AND SURFACE INTEGRALS 4.1 b Integrate P(x. on C. y) i+ Q(x. that is.

then we have P(x. y1 (x))) dx P(x. as we see from Figure 4. y2 (x)) dx P(x. Since y = y1 (x) along C1 (as x goes from a to b) and y = y2 (x) along C2 (as x goes from b to a). y) dy + c d c Q(x2 (y). y) dy dx (by the Fundamental Theorem of Calculus) ∂y ∂P dA . then we have Q(x.25) and (4. y) dy d =− = c Q(x1 (y). y) dx a = a b P(x. ∂y R Likewise. y) dy = C C1 c Q(x.3.3.3 Green’s Theorem 151 (4. y) around C using the representation C = C1 ∪ C2 given by (4.1. y) dy + c d Q(x2 (y). y) dy d = d Q(x1 (y). y) dy (Q(x2 (y). y) y=y2 (x) y=y1 (x) dx a b a y2 (x) y1 (x) ∂P(x.26).4. Since x = x1 (y) along C1 (as y goes from d to c) and x = x2 (y) along C2 (as y goes from c to d). y2 (x)) dx a = a P(x. y1 (x)) dx + b b P(x. y) − Q(x1 (y). y) x2 (y) x1 (y) x=x2 (y) x=x1 (y) dy = c ∂Q(x. y1 (x)) dx − b a b =− =− =− =− (P(x. y) dy + C2 Q(x. and so ∂x . as we see from Figure 4.1. y2 (x)) − P(x. y) dx + C2 P(x.24). y) dx dy (by the Fundamental Theorem of Calculus) ∂x = R ∂Q dA . y)) dy d = c d Q(x. y) dx = C C1 b P(x. integrate Q(x.

note that R is not the entire region enclosed by C. Evaluate C (x2 +y2 ) dx+2xy dy. . y) = x (x2 + y2 ) dx + 2xy dy = C R ∂Q ∂P − dA ∂x ∂y (2y − 2y) dA = R R C x 0 = 0 dA = 0 . y) = x2 x . 2 + y2 and Q(x. 2 See T AYLOR and M ANN.2 that C f · dr = 2π. y) : 0 < x2 + y2 ≤ 1 }. y) dy ∂Q dA ∂x R =− R ∂P dA + ∂y = R ∂Q ∂P dA .2 Example 4. y) = (x2 + y2 ) i + 2xy j has a potential function 1 F(x. By Green’s Theorem. That is. Example 4. y (1. y) = 3 x3 + xy2 .2. 2x2 ≤ y ≤ 2x }. y) = 2xy. 2) Solution: R is the shaded region in Figure 4. it was shown in Exercise 9(b) in Section 4. the theorem can also be proved for more general regions (say. But ∂P ∂Q y2 − x2 = = 2 ⇒ ∂x ∂y (x + y2 )2 ∂Q ∂P − dA = ∂x ∂y R R 0 dA = 0 . so Green’s Theorem does not apply. y) i + Q(x. y) : 0 ≤ x ≤ 1. since the point (0. LINE AND SURFACE INTEGRALS C f · dr = P(x. y) j. 0) is not contained in R. y) = x2 −y + y2 and Q(x.7.2 We actually already knew that the answer was zero.2 that the vector field f(x. y) dx + C C Q(x.8. + y2 and let R = { (x. Recall from Example 4.31 for a discussion of some of the difficulties involved when the boundary curve is “complicated”.6. we have 2 for P(x. This would seem to contradict Green’s Theorem. For the boundary curve C : x2 + y2 = 1.3. Let f(x. where C is the boundary (traversed counterclockwise) of the region R = { (x. a union of simple regions). However. − ∂x ∂y QED Though we proved Green’s Theorem only for a simple region R. 1 Figure 4.3. § 15. traversed counterclockwise. where P(x. y) = P(x.5 in Section 4.152 CHAPTER 4. and so C f · dr = 0 by Corollary 4. R has a “hole” at the origin.

C1 R1 C2 R1 C3 C2 ◮ ◭C ◮ ◮ ◭ R2 1 ◭ ◮ ◭ R2 ◮ (a) Region R with one hole (b) Region R with two holes Figure 4.3. The idea is to cut “slits” between the boundaries of a multiply connected region R so that R is divided into subregions which do not have any “holes”.4 above.3 Green’s Theorem If we modify the region R to be the annulus R = { (x.3.8. that is. and we traverse then in the manner indicated by ◭ ◮ .3.4(a) the region R is the union of the regions R1 and R2 . which are divided by the slits indicated by the dashed lines.4 Multiply connected regions The intuitive idea for why Green’s Theorem holds for multiply connected regions is shown in Figure 4. in Figure 4. It turns out that Green’s Theorem can be extended to multiply connected regions.4. where C1 is the unit circle x2 + y2 = 1 traversed counterclockwise and C2 is the circle x2 + y2 = 1/4 traversed clockwise. as opposed to discrete points being cut out. which have one or more regions cut out from the interior. y) : 1/4 ≤ x2 + y2 ≤ 1 } (see Figure 4.3). For such regions. regions like the annulus in Example 4. Those slits are part of the boundary of both R1 and R2 .3 The annulus R R we would have f · dr = R C ∂Q ∂P dA . so for this Figure 4. − ∂x ∂y which shows that Green’s Theorem holds for the annular region R. For example.3. ∂Q ∂x We would still have R − ∂P ∂y dA = 0. the “outer” boundary and the “inner” boundaries are traversed so that R is always on the left side. and take the “boundary” C of R to be C = C1 ∪ C2 . then it can be shown (see Exercise 8) that C 153 y 1 C1 ◭ R C2 0 1/2 1 x 1/2 f · dr = 0 .3.

We know from Corollary 4. the following can be shown: The following statements are equivalent for a simply connected region R in (a) f(x.154 CHAPTER 4. Since R1 and R2 do not have holes in them. A similar argument shows that the theorem holds in the region with two holes shown in Figure 4. y) j on a region R (whose boundary is a piecewise smooth. − ∂x ∂y R and so C1 ∪C2 f · dr = R1 ∂Q ∂P dA + − ∂x ∂y R2 ∂Q ∂P dA = − ∂x ∂y which shows that Green’s Theorem holds in the region R. y) i+Q(x. and so we know that ∂y ∂2 F ∂2 F ∂P ∂Q = ⇒ = in R. we have C1 ∪C2 f · dr = bdy of R1 f · dr + bdy of R2 f · dr . y) = P(x. y) i + Q(x. then Green’s Theorem holds in each subregion. y) = P(x. ∂Q ∂P dA . And if the potential F(x. if ∂P ∂y = ∂Q ∂x in R then f · dr = R C ∂Q ∂P dA = − ∂x ∂y R 0 dA = 0 .3. simple closed curve C) has a potential in R.4(b). y) j has a smooth potential F(x. ∂y ∂x ∂x ∂y ∂y ∂x Conversely. Notice that along each slit the boundary of R1 is traversed in the opposite direction as that of R2 .e. a region with no holes). LINE AND SURFACE INTEGRALS the arrows. then C f · dr = 0. − ∂x ∂y But since the line integrals along the slits cancel out. so that bdy of R1 f · dr = R1 ∂Q ∂P dA − ∂x ∂y and bdy of R2 f · dr = R2 ∂Q ∂P dA . y) is smooth in R. then ∂F = P ∂x and ∂F = Q. which means that the line integrals of f along those slits cancel each other out.6 that when a smooth vector field f(x. For a simply connected region R (i. y) in R (b) C 2: f · dr is independent of the path for any curve C in R f · dr = 0 for every simple closed curve C in R (in this case. the differential form P dx + Q dy is exact) (c) C (d) ∂P ∂Q = in R ∂y ∂x .

b and any closed simple curve C. where C is the boundary of the rectangle with vertices (1. 4) 5. y) = (y2 + 3x2 ) i + 2xy j? If so. For a region R bounded by a simple closed curve C. 0). B 9. y) = (x3 cos(xy) + 2x sin(xy)) i + x2 y cos(xy) j? If so. C 4. 10. show that the area A of R is A=− y dx = C C x dy = 1 2 C x dy − y dx . 7. y) : 0 ≤ x ≤ 1. find one. C 3. For the vector field f as in Example 4. Is there a potential F(x. y) for f(x. 2 2 2. Is there a potential F(x.3 Green’s Theorem  155 A Exercises ©  ¨ For Exercises 1-4. 1. Evaluate C e x sin y dx + (y3 + e x cos y) dy. y) for f(x. where C is traversed so that R is always on the left.) R . C a dx + b dy = 0. 2x2 ≤ y ≤ 2x } x2 y dx + 2xy dy. C is the boundary of R = { (x. use Green’s Theorem to evaluate the given line integral around the curve C. C 11. y) = (8xy + 3) i + 4(x2 + y) j? If so. traversed counterclockwise. 1). Show that for any constants a. C (4. find one.8. y) : 1/4 ≤ x2 + y2 ≤ 1 } traversed so that R is always on the left. (Hint: Use Green’s Theorem and the fact that A = 1 dA. C is the boundary of R = { (x. Is there a potential F(x. −1). C is the circle x2 + y2 = 1 (e x + y2 ) dx + (ey + x2 ) dy. y) for f(x. 1) and (−1. where C is the boundary of the annulus R = { (x. y) : 0 ≤ x ≤ 1. (1. find one. C (x2 − y2 ) dx + 2xy dy.4. traversed counterclockwise. x2 ≤ y ≤ x } 2y dx − 3x dy. 8. (−1. 0) and (0. −1). show directly that C f · dr = 0. 6. C is the boundary of the triangle with vertices (0.

v)k for (u. z = z(u. v) 0 u Figure 4.4. v).4. a ≤ t ≤ b. We will now learn how to perform integration over a surface in 3 . v) y Parametrization of a surface Σ in In this case. y(t). v) y = y(u.8 how we identified points (x.4 Surface Integrals and the Divergence Theorem In Section 4. u and v.1). The idea behind a parametrization of a curve is that it “transforms” a subset of (normally an interval [a. parametrized by x = x(t). v) = x(u. Recall from Section 1. with the terminal points of the position vector r(t) = x(t)i + y(t)j + z(t)k for t in [a. b]) into a curve in 2 or 3 (see Figure 4. v)i + y(u. v)j + z(u.1 we learned how to integrate along a curve. to parametrize a surface Σ in 3 : x = x(u. z = z(t). y = y(u.2 x 3 r(u. y. v) z = z(u. v 2 z Σ R (u. y = y(t). z(a)) x = x(t) y = y(t) z = z(t) 1 1 C r(t) 0 x (x(b).2). v) in R. the position vector of a point on the surface Σ is given by the vectorvalued function r(u. y(a). v). z(b)) y a t b Figure 4. y(b).156 CHAPTER 4. z (x(t). such as a sphere or a paraboloid. we will use a parametrization of a surface to define a surface integral. .4.4. v) x = x(u. b].1 Parametrization of a curve C in 3 Similar to how we used a parametrization of a curve to define the line integral along the curve. LINE AND SURFACE INTEGRALS 4. We will use two variables. z) on a curve C in 3 . z(t)) (x(a). for (u. v). v) in some region R in 2 (see Figure 4.

v) − r(u. and the variable u is constant along the position vector r(u. will have a surface area (call it dσ) that is very close to the area of the parallelogram which has adjacent sides r(u + ∆u. the horizontal gridlines ∂v in R get mapped to curves on Σ whose tangent vectors are ∂r . Similarly. the total surface area S of Σ is approximately the sum of all the quantities ∂r × ∂r ∆u ∆v. v) in R by ∂r (u. Along the vertical gridlines in R. v) in R) and r(u. Thus. v) ≈ . we have ∂r r(u + ∆u. ∂v ∂v ∂r ∂u 157 and ∂r ∂v The parametrization of Σ can be thought of as “transforming” a region in 2 (in the uv-plane) into a 2-dimensional surface in 3 . as shown in Figure 4.4. v) ≈ .4. v)k .2. and ∂u ∆u ∂r r(u. The corner points of that rectangle are (u.4. (u + ∆u. v).3 in Section 2. v + ∆v). v + ∆v) and (u. v) in R as. Thus. v) − r(u.8) applied to a function of two variables. This parametrization of the surface is sometimes called a patch. v + ∆v) − r(u. v) is a function of two variables. ∂u Now take a point (u.4 Surface Integrals and the Divergence Theorem Since r(u. So the area of that rectangle is A = ∆u ∆v.2. say. Taking ∂u ∂v the limit of that sum as the diagonal of the largest rectangle goes to 0 gives S = R ∂r ∂r × du dv . v) to (u. v) = ∂v ∂x (u. the variable u is constant. ∂u ∂v (4. respectively. ∂v ∆v and so the surface area element dσ is approximately (r(u + ∆u. v). for ∆u and ∆v small enough. Suppose that this rectangle has a small width and height of ∆u and ∆v. v + ∆v) − r(u. v)i + ∂v ∂y ∂z (u. v) to (u + ∆u. v + ∆v) in R). But by combining our usual notion of a partial derivative (see Definition 2. v)i + ∂u ∂x (u. the lower left corner of one of the rectangular grid sections in R. v)) × (r(u. v) (corresponding to the line segment from (u.12 in Section 1. v)j + (u. based on the idea of “patching” the region R onto Σ in the grid-like manner shown in Figure 4. v)) ≈ (∆u ∂r ∂r ∂r ∂r ) × (∆v ) = × ∆u ∆v ∂u ∂v ∂u ∂v by Theorem 1. v) − r(u.2) with that of the derivative of a vector-valued function (see Definition 1. v)k . v). define the partial derivatives for (u. v) (corresponding to the line segment from (u. summed over the rectangles in R.26) . In fact. those gridlines in R lead us to how we will define a surface integral over Σ. So those lines get mapped to curves on Σ. (u + ∆u. and ∂u ∂u ∂y ∂z (u. v) is ∂r . v)j + (u.4. Then that rectangle gets mapped by the parametrization onto some section of the surface Σ which. the tangent vector to those curves at a point (u. v) = ∂u ∂r (u. v + ∆v) − r(u.13 in Section 1.

z z (y − b)2 + z2 = a2 a 0 u y y v (x. the surface area S of Σ is S = Σ 1 dσ . v). the line segment from the center of the circle to that point makes an angle u with the y-axis in the positive y direction (see Figure . y = y(u. v). v) in some region R in 2 .29) Example 4. ∂u ∂v (4.4. ∂u ∂v (4. z = z(u. y. z(u. v). v)i + y(u. z) defined in 3 . y. y(u. (4.28) In particular. v) = x(u.y. v)j + z(u.27) This is a special case of a surface integral over the surface Σ. v). LINE AND SURFACE INTEGRALS We will write the double integral on the right using the special notation dσ = Σ R ∂r ∂r × du dv . A torus T is a surface obtained by revolving a circle of radius a in the yz-plane around the z-axis.3 Solution: For any point on the circle. Find the surface area of T . for (u. Replacing 1 by a general real-valued function f (x. and let f (x. where the circle’s center is at a distance b from the z-axis (0 < a < b).4. z) be a real-valued function defined on some subset of 3 that contains Σ. as in Figure 4. y.3. v). v)k be the position vector for any point on Σ.3. z) over Σ is f (x. z) dσ = Σ R f (x(u. Let r(u.158 CHAPTER 4. where the surface area element dσ can be thought of as 1 dσ. y. The surface integral of f (x.z) a b (a) Circle in the yz-plane x (b) Torus T Figure 4. we have the following: Definition 4. Let Σ be a surface in 3 parametrized by x = x(u.9. v)) ∂r ∂r × du dv .

4 Surface Integrals and the Divergence Theorem 159 4. v)k = (b + a cos u) cos v i + (b + a cos u) sin v j + a sin u k we see that ∂r = −a sin u cos v i − a sin u sin v j + a cos u k ∂u ∂r = −(b + a cos u) sin v i + (b + a cos u) cos v j + 0k . ∂u ∂v which has magnitude ∂r ∂r = a(b + a cos u) . then their cross product ∂r × ∂r is perpendicular to the tangent plane to ∂u ∂v .e. v) = x(u. 0 ≤ v ≤ 2π 1 dσ ∂r ∂r du dv × ∂u ∂v a(b + a cos u) du dv 0 2π 0 = 0 = = 0 2π abu + a2 sin u 2πab dv 0 u=2π u=0 dv = = 4π2 ab Since ∂r and ∂r are tangent to the surface Σ (i. v)i + y(u. So for the position vector r(u.4.3(a)). And as the circle revolves around the z-axis. × ∂u ∂v Thus. ∂v and so computing the cross product gives ∂r ∂r × = −a(b + a cos u) cos v cos u i − a(b + a cos u) sin v cos u j − a(b + a cos u) sin u k . the surface area of T is S = Σ 2π 2π 0 2π 2π y = (b + a cos u) sin v . z = a sin u .4. lie in the tangent plane to Σ at each ∂u ∂v point on Σ). v)j + z(u. 0 ≤ u ≤ 2π . Thus. the torus can be parametrized as: x = (b + a cos u) cos v .3(b)).4. the line segment from the origin to the center of that circle sweeps out an angle v with the positive x-axis (see Figure 4.

As we can see from Figure 4. using (u.4 Definition 4. z Recall that normal vectors to a plane can point in two opposite directions. 1. y) : 0 ≤ x ≤ 1. (4. n is the outward unit normal vector to Σ. f (x.5.4. By an outward unit normal vector to a surface Σ. y = v.10. projecting Σ onto the xy-plane yields a triangular region R = { (x. y(u. LINE AND SURFACE INTEGRALS the surface at each point of Σ.4. y ≥ 0. Example 4. and z ≥ 0. v). v)) n dσ .4. We say that n is a normal vector to Σ.5). y. z(u. z = 1 − (u + v). 0 ≤ y ≤ 1 − x }. With this idea in mind. This is a hazy definition. we will mean the unit vector that is normal to Σ and points away from the “top” (or “outer” part) of the surface. z) = yzi + xzj + xyk and Σ is the part of the plane x + y + z = 1 with x ≥ 0. in the case of a sphere. y). where n = ∂r ∂u × ∂r ∂v . z) = f1 (x. z)j + f3 (x. we see that x = u. We now need to parametrize Σ.30) where. z) dσ = Σ R f (x(u.160 CHAPTER 4. y. y. Let Σ be a surface in 3 and let f(x. y. where f(x. y. Thus. then dividing v by its length yields the outward unit normal vector n = 1 1 1 √ .4 gives a better idea of what outward normal vectors look like. we make the following definition of a surface integral of a 3-dimensional vector field over a surface: y 0 x Figure 4. for 0 ≤ u ≤ 1. and hence we can use Definition 4.3 to evaluate the integral. Thus. √ 3 3 3 z 1 Σ 0 1 x 1 x+y+z=1 Figure 4.4.4. Solution: Since the vector v = (1. but the picture in Figure 4. v). y. with the outward unit normal n pointing in the positive z direction (see Figure 4.5 n y . Evaluate the surface integral Σ f · dσ.4. Note in the above definition that the dot product inside the integral on the right is a real-valued function. 0 ≤ v ≤ 1 − u . z)i + f2 (x. v) instead of (x. √ . 1) is normal to the plane x + y + z = 1 (why?). z)k be a vector field defined on some subset of 3 that contains Σ. at any point on Σ. The surface integral of f over Σ is f · dσ = Σ Σ f · n dσ .

g. v). xz. √ . For example. as indicated by the dashed line in Figure 4. spheres. and for r(u. cubes. √ = √ (yz + xz + xy) 3 3 3 3 1 1 = √ ((x + y)z + xy) = √ ((u + v)(1 − (u + v)) + uv) 3 3 1 = √ ((u + v) − (u + v)2 + uv) 3 161 for (u. 1 1 1 1 f · n = (yz. −1) × (0. .4 Surface Integrals and the Divergence Theorem is a parametrization of Σ over R (since z = 1 − (x + y) on Σ). 1. z(u. and ellipsoids are closed surfaces. y(u.5) gives f · dσ = Σ Σ f · n dσ (f(x(u. v)) · n) R 1 1−u = = ∂r ∂r × dv du ∂u ∂v √ 1 √ ((u + v) − (u + v)2 + uv) 3 dv du 0 0 3  v=1−u  1 2  (u + v)3 uv2  (u + v)    du   = − +    2  3 2 0 v=0 1 = 0 1 u + − + 6 2 2 6 1 3u2 5u3 du u u2 u3 5u4 = + − + 6 4 2 24 = 0 1 . xy) · √ . v)j + z(u. but planes and paraboloids are not. v)i + y(u. The following theorem provides an easier way in the case when Σ is a closed surface. v)k = ui + vj + (1 − (u + v))k we have ∂r ∂r × = (1. that is. v). ∂u ∂v Thus. So on Σ. −1) = (1. especially when the formula for the outward unit normal vector at each point of Σ changes. integrating over R using vertical slices (e. when Σ encloses a bounded solid in 3 . 0. v) in R. 1.4. 8 Computing surface integrals can often be tedious. 1) ∂u ∂v ⇒ √ ∂r ∂r × = 3. v) = x(u.4.

∂ f1 ∂ f2 ∂ f 3 + + ∂x ∂y ∂z (4. for a point (x. z)k be a vector field defined on some subset of 3 that contains Σ. See S CHEY. Namely. y. (4.6 for the details. z) = xi + yj + zk and Σ is the unit sphere Solution: We see that div f = 1 + 1 + 1 = 3.31) where div f = is called the divergence of f. where f(x. z)i + f2 (x. z) in 3 .33) V→0 V Σ 3 4 See T AYLOR and M ANN. for an intuitive discussion of this. The term divergence comes from interpreting div f as a measure of how much a vector field “diverges” from a point. .11. (4. it is first proved for the simple case when the solid S is bounded above by one surface. z) = lim f · dσ .3 Example 4. and let f(x. LINE AND SURFACE INTEGRALS Theorem 4.e. y.162 CHAPTER 4.8. For example. while a negative flux indicates a net flow inward (in the direction of −n). if f represents the velocity field of a fluid. y.e. This is best seen by using another definition of div f which is equivalent4 to the definition given by formula (4. z)j + f3 (x. 1 div f(x. p. 3 =3 S 1 dV = 3 vol(S ) = 3 · In physical applications. § 15. so f · dσ = Σ S div f dV = S 3 dV 4π(1)3 = 4π . 36-39. The proof can then be extended to more general solids. then the flux is the net quantity of fluid to flow through the surface Σ per unit time.32) The proof of the Divergence Theorem is very similar to the proof of Green’s Theorem.32). i. bounded below by another surface. Σ f · dσ. the surface integral Σ f · dσ is often referred to as the flux of f through the surface Σ. y. (Divergence Theorem) Let Σ be a closed surface in 3 which bounds a solid S . y. y. A positive flux means there is a net flow out of the surface (i. y. and bounded laterally by one or more surfaces. Then f · dσ = Σ S div f dV . z) = f1 (x. in the direction of the outward unit normal vector n). Evaluate x2 + y2 + z2 = 1.

y. then div f = 0 at that point. Notice that the limit being taken is of the ratio of the flux through a surface to the volume enclosed by that surface. z ≤ 1 } 3. y. y. f(x. f(x. respectively. Vector fields which have zero divergence are often called solenoidal fields. z). at the given point (x. y. Especially in physics texts.33). as we mentioned. which means that the volumes they enclose are going to zero. z) = xi + yj + zk. z).9. The following theorem is a simple consequence of formula (4.4. Σ : boundary of the solid cube S = { (x. f(x. V → 0 means that we take smaller and smaller closed surfaces around (x.33). y. If the flux of a vector field f is zero through every closed surface containing a given point. z) over the surface Σ. z) dσ and Σ Σ QED f · dσ is used to denote surface integrals of scalar and vector fields. y. it is common to see simply instead of . z) : 0 ≤ x. It can be shown that this limit is independent of the shapes of those surfaces. Σ : x2 + y2 + z2 = 1 4. y. Σ : x2 + y2 + z2 = 9 2. z) = x3 i + y3 j + z3 k. which gives a rough measure of the flow “leaving” a point. Lastly. y. f(x. 1. Σ : x2 + y2 + z2 = 1 Σ · f· dσ . y. Σ Σ  A Exercises ©  ¨ For Exercises 1-4. we note that sometimes the notation f (x. y. z) = 2i + 3j + 5k. z) we have div f(x. use the Divergence Theorem to evaluate the surface integral of the given vector field f(x. Proof: By formula (4. z) = lim 1 V→0 V Σ f · dσ for closed surfaces Σ containing (x. z) = xi + 2yj + 3zk. so 1 (0) by our assumption that the flux through each Σ is zero. y. y. In the limit. Theorem 4.4 Surface Integrals and the Divergence Theorem 163 where V is the volume enclosed by a closed surface Σ around the point (x. so = lim V→0 V = lim 0 V→0 =0. y. z). over closed surfaces.

z = R r. An alternative is to express the surface area in terms of elliptic integrals. i. Evaluate the surface integral from Exercise 2 without using the Divergence Theorem.5 ) C 11. y. Use a surface integral to show that the surface area of a sphere of radius r is 4πr2 . The ellipsoid x2 a2 + y2 b2 + z2 c2 = 1 can be parametrized using ellipsoidal coordinates x = a sin φ cos θ . Show that the flux of any constant vector field through any closed surface is zero. for 0 ≤ θ ≤ 2π and 0 ≤ φ ≤ π. 1961. with Applications. 6. (Hint: Think of the parametrization of the surface. and z ≥ 0. (Hint: Use the parametrization x = r cos θ.. Use a surface integral to show that the surface area of a right circular cone of √ radius R and height h is πR h2 + R2 . Evaluate the surface integral Σ f · dσ. as in Example 4.) 5 B OWMAN. y = b sin φ sin θ .164 CHAPTER 4. where f(x. Introduction to Elliptic Functions. LINE AND SURFACE INTEGRALS B 5. Show that the surface area S of the ellipsoid is π 2π S = 0 0 sin φ a2 b2 cos2 φ + c2 (a2 sin2 θ + b2 cos2 θ) sin2 φ dθ dφ . y) is given by the formula S = R 2 of a 1+ ∂f 2 ∂x + ∂f 2 ∂y dA . . 7. (Note: The above double integral can not be evaluated by elementary means.7. for 0 ≤ r ≤ R and 0 ≤ θ ≤ 2π. (Hint: Use spherical coordinates to parametrize the sphere. New York: Dover.10. For specific values of a. h y = r sin θ. y ≥ 0. z) = x2 i + xyj + zk and Σ is the part of the plane 6x + 3y + 2z = 6 with x ≥ 0.3.) 9. using only Definition 4.3 to prove that the surface area S over a region R in surface z = f (x. b and c it can be evaluated using numerical methods. Use Definition 4. § III. 8. F. with the outward unit normal n pointing in the positive z direction.e. z = c cos φ .) 10. Note that there will be a different outward unit normal vector to each of the six faces of the cube.

a ≤ t ≤ b. a ≤ t ≤ b. z) along C with respect to y is b f (x. z) dy = C a f (x(t). y.5 Stokes’ Theorem 165 4. y. y(t).36) The line integral of f (x. z(t)) x ′ (t) dt . For a vector field f(x. Vector fields in 3 are defined in a similar fashion to those in 2 . (4. z(t)) · r ′ (t) dt . z) = P(x. z) and a curve C in 3 . y. (4. parametrized by x = x(t). z) along C with respect to arc length s is b f (x. y. y = y(t). y.5 Stokes’ Theorem So far the only types of line integrals which we have discussed are those along curves in 2 .1 and 4. z(t)) z ′ (t) dt . y. y(t). y(t). z) along C with respect to x is b f (x. y. y. z) k and a curve C in 3 with a smooth parametrization x = x(t). y.6. y. z(t)) x ′ (t)2 + y ′ (t)2 + z ′ (t)2 dt . (4. z) dz = C a f (x(t). But the definitions and properties which were covered in Sections 4. y. y. y. . y.35) The line integral of f (x. z) ≥ 0 then the line integral C f (x. z) dx = C a f (x(t). y. y.34) The line integral of f (x.37) Similar to the two-variable case. if f (x. z) dy + C R(x. z) along C with respect to z is b f (x. z(t)) y ′ (t) dt . y(t). z) j + R(x. y(t). the line integral of f (x.4. y. z) ds can be thought of as the total area of the “picket fence” of height f (x.39) C f(x(t). y. z = z(t). z) at each point along the curve C in 3 . (4. z) dx + C b C Q(x. y = y(t). so that we can now discuss line integrals along curves in 3 . z) dz (4. For a real-valued function f (x. the line integral of f along C is f · dr = = a P(x. z = z(t).5. z) i + Q(x. y. z) ds = C a f (x(t). Definition 4. Definition 4. which allows us to define the line integral of a vector field along a curve in 3 .2 can easily be extended to include functions of three variables.38) (4. where r(t) = x(t) i + y(t) j + z(t) k is the position vector for points on C.

y = y(t). If a vector field f has a potential in a solid S . z(b)) are the endpoints of C. y = y(t1 . y.10. then w is a differentiable function of t.166 CHAPTER 4.41) Also. Let C be a smooth curve in S parametrized by x = x(t). ∂t2 ∂x ∂t2 ∂y ∂t2 ∂z ∂t2 (4. if x = x(t1 . z) then the line integral C f · dr represents the work done by that force in moving the object along the curve C in 3 . and x = x(t). Then C f · dr = F(B) − F(A) .11. with P. z) i+Q(x. Q and R continuously differentiable functions on S . and z. (4. then C · f· dr = 0 for any closed curve C in S (i. For a vector field f(x. z(a)) and B = (x(b). and dw ∂w dx ∂w dy ∂w dz = + + . y(b). z) k be a vector field in some solid S . t2 ).43) Theorem 4. y(t). Corollary 4. z) j + R(x. a ≤ t ≤ b. y. z = z(t). dt ∂x dt ∂y dt ∂z dt (4. y.e.5 for a proof. z) = P(x. z)). z) j+R(x. Suppose that there is a real-valued function F(x. y = y(t). y.42) ∂t1 ∂x ∂t1 ∂y ∂t1 ∂z ∂t1 and ∂w ∂w ∂x ∂w ∂y ∂w ∂z = + + . (4.40) where T(t) = r ′ (t) r ′ (t) is the unit tangent vector to C at (x(t). z = z(t). LINE AND SURFACE INTEGRALS Similar to the two-variable case. z) i + Q(x. y. z) = P(x. z) k and a curve C with a smooth parametrization x = x(t).13. a ≤ t ≤ b and position vector r(t) = x(t) i + y(t) j + z(t) k. z(t)). then6 ∂w ∂w ∂x ∂w ∂y ∂w ∂z = + + (4. Let f(x. y(a). y. y. t2 ). y = y(t) and z = z(t) are differentiable functions of t. y. C 6 ∇F · dr = 0 for any real-valued function F(x. Theorem 4. y. See T AYLOR and M ANN. y. if f(x. § 6. y. y. . Theorem 4. C f · dr = C f · T ds . (Chain Rule) If w = f (x. t2 ) are continuously differentiable function of (t1 . z) such that ∇F = f on S . Some of the most important results we will need for line integrals in 3 are stated below without proof (the proofs are similar to their two-variable equivalents). y. z) is a continuously differentiable function of x. z) represents the force applied to an object at a point (x.12.44) where A = (x(a). y. t2 ) and z = z(t1 .

4. y. y. so since f (x(t). Let f (x. y.1 Conical helix C Example 4.12. we have x ′ (t)2 + y ′ (t)2 + z ′ (t)2 = (sin2 t + 2t sin t cos t + t2 cos2 t) + (cos2 t − 2t sin t cos t + t2 sin2 t) + 1 = t2 (sin2 t + cos2 t) + sin2 t + cos2 t + 1 = t2 + 2 . Let f(x. z) (i. y(t).e. y(t). y ′ (t) = cos t − t sin t. then 8π f (x. y.5 Stokes’ Theorem Example 4. y.1). z) ds. z) = x i + y j + 2z k be a vector field in curve C from Example 4. y. z) = z and let C be the curve in x = t sin t .5. . evaluate C f · dr. Solution: It is easy to see that F(x. Using the same + y2 2 + z2 is a potential for f(x.12. Evaluate C 3 167 parametrized by y = t cos t . z(t)) = z(t) = t along the curve C. 0 ≤ t ≤ 8π . and z ′ (t) = 1.5.13. z) ds = C 0 8π f (x(t). z=t. Solution: Since x ′ (t) = sin t + t cos t. See Figure 4. (Note: C is called a conical helix. f (x. z(t)) t 0 x ′ (t)2 + y ′ (t)2 + z ′ (t)2 dt = t2 + 2 dt 8π 1 = (t2 + 2)3/2 3 = 0 √ 1 (64π2 + 2)3/2 − 2 2 . z) = x2 2 3. 3 30 25 20 15 z 10 5 0-25 -20 -15 -10 -5 y -25 -20 -15 -10 x t = 8π t=0 5 0 -5 0 5 10 15 20 25 30 25 20 15 10 Figure 4.

z) = x i+y j+z k is nonzero and −N normal to the sphere at each point. and planes. LINE AND SURFACE INTEGRALS ∇F = f). 0) (8π)2 + (8π)2 − (0 + 0 + 0) = 96π2 . so = F(0. Other examples of two-sided. since N the continuous vector field N(x. paraboloids.168 CHAPTER 4. y(8π). as in Figure 4. A surface Σ in 3 is orientable if there is a continuous vector field N in 3 such that N is nonzero and normal to Σ (i.5. called Stokes’ Theorem. We see in this case that y N(x. We say that such an N is a normal vector field. “twosided” means “orientable”. In fact.5. x Figure 4.5. where A = (x(0). 8π) − F(0.2 we say that the sphere is a two-sided surface.3(b).2). which is constructed by taking a thin rectangle and connecting its ends at the opposite corners. You may be wondering what kind of surface would not have two sides. 8π cos 8π. z) is an inward normal vector. That is. respectively. z For example. surfaces are cylinders. ellipsoids. the unit sphere x2 +y2 +z2 = 1 is orientable. 0.e. z(8π)). 2 We will now discuss a generalization of Green’s Theorem in 2 to orientable surfaces in 3 . −N(x. These “outward” and “inward” normal vector fields on the sphere correspond to an “outer” and “inner” side.5. resulting in a “twisted” strip (see Figure 4. y. 8π) − F(0 sin 0. and hence orientable. y. of the sphere. So by Theorem 4. your orientation changed even though your motion was continuous A . An example is the Möbius strip. z) is another normal vector field (see Figure 4. 8π. z) is what we have called an outward normal vector. y. y. perpendicular to the tangent plane) at each point of Σ. Roughly. 0) =0+ = F(8π sin 8π. 0 cos 0.3). z(0)) and B = (x(8π).3 Möbius strip If you imagine walking along a line down the center of the Möbius strip. then you arrive back at the same place from which you started but upside down! That is.12 we know that C f · dr = F(B) − F(A) . A A B → → −→ B A (a) Connect A to A and B to B along the ends (b) Not orientable Figure 4. y(0). and 0 −N(x.5.

there is a discontinuity at your starting point (and.2).5. say C D : x = x(t) . see O’N EILL.5 Stokes’ Theorem 169 along that center line. z = z(x(t). y) n C y 0 x D (x. y). thinking of your vertical direction as a normal vector field along the strip. y). y) varying over a region D in 2 .46) n is a positive unit normal vector over Σ.5. a ≤ t ≤ b . y = y(t) . and hence is nonorientable.7. z)j + R(x. for z = z(x(t). y) for some smooth real-valued function z(x. y.4 in Section 4. y = y(t) . Then f · dr = Σ C (curl f ) · n dσ . pick a unit normal vector n such that if you walked along C with your head pointing in the direction of n.14.4). Projecting Σ onto the xy-plane. § IV. z) = P(x. a ≤ t ≤ b . ∂y ∂z ∂z ∂x ∂x ∂y (4.4. at every point) since your vertical direction takes two different values there. z)i + Q(x. then the surface would be on your left. we will prove the theorem only for the special case where Σ is the graph of z = z(x. The Möbius strip has only one side. z)k be a smooth vector field defined on some subset of 3 that contains Σ. Assuming that C has a smooth parametrization. with (x. we see that the closed curve C (the boundary curve of Σ) projects onto a closed curve C D which is the boundary curve of D (see Figure 4. Proof: As the general case is beyond the scope of this text. and C is traversed n-positively. y. y(t)) as a function of t. by the Chain Rule (Theorem 4. in fact.4 as C : x = x(t) . we know that z ′ (t) = 7 ∂z ′ ∂z ′ x (t) + y (t) . and let f(x. We can now state Stokes’ Theorem: Theorem 4. y) CD Figure 4. . (Stokes’ Theorem) Let Σ be an orientable surface in 3 whose boundary is a simple closed curve C. ∂x ∂y For further discussion of orientability. y. its projection C D in the xy-plane also has a smooth parametrization. (4. y.7 For an orientable surface Σ which has a boundary curve C. Informally. and so C can be parametrized (in 3) z Σ : z = z(x. y(t)) . Now.45) where curl f = ∂P ∂R ∂Q ∂P ∂R ∂Q − i+ − j+ − k. We say in this situation that n is a positive unit normal vector and that C is traversed n-positively. since the curve C is part of the surface z = z(x.

y)) (x. y) in D. z(x. y) dy . ∂x ∂x ∂z ∂x . y) ∂y for (x. z) dz C b P x ′ (t) + Q y ′ (t) + R a b ∂z ′ ∂z ′ x (t) + y (t) ∂x ∂y dt = a P+R ∂z ′ ∂z ′ x (t) + Q + R y (t) dt ∂x ∂y = CD ˜ ˜ P(x. z(x. y. y. y) + R(x. y))) = + + ∂x ∂x ∂x ∂y ∂x ∂z ∂x ∂Q ∂Q ∂z ∂Q ·1+ ·0+ = ∂x ∂y ∂z ∂x ∂Q ∂Q ∂z = + . y)) (x. ˜ ∂z ∂ ∂Q Q(x. y)) (x. so by the Product Rule we get = ∂x ∂x ∂y ∂ ∂z ∂ ∂z ∂ (Q(x. y) = Q(x. y) .47) Thus.42) in Theorem 4. y) . ∂R ∂R ∂z ∂ (R(x. we have f · dr = D C ˜ ˜ ∂Q ∂P − dA . LINE AND SURFACE INTEGRALS C f · dr = = P(x. y. y. y)) + R(x. z(x. z(x. by Green’s Theorem applied to the region D. z(x. y. y)) + R(x. and ∂x ∂z ˜ Q(x. y. y. ∂x ∂y (4. z(x. z(x. y)) (x. z) dy + R(x. z(x. y. ∂x ∂z ∂x Similarly. z(x. we have ∂ ∂Q ∂x ∂Q ∂y ∂Q ∂z (Q(x. y. y. y) . y) dx + Q(x. y. by formula (4. y))) = + .170 and so CHAPTER 4. y)) (x. = ∂x ∂x ∂y ∂x ∂y Now. where ∂z ˜ P(x. z(x. y))) + R(x. y) = P(x. y. Thus.11. z(x. y)) + R(x. z) dx + Q(x. y. y.

QED which. Hence.5. ∂y ∂y ∂z ∂y ∂y ∂x ∂z ∂y ∂x ∂y ∂x So subtracting gives ˜ ˜ ∂Q ∂P ∂Q ∂R ∂z ∂R ∂P ∂z ∂Q ∂P − = − + − + − ∂x ∂y ∂z ∂y ∂x ∂x ∂z ∂y ∂x ∂y since ∂2 z ∂x ∂y 171 (4. z(x. by equation (4. y) = x i + y j + z(x. y) k.48) = ∂2 z ∂y ∂x by the smoothness of z = z(x. and so ∂y using formula (4. y) at each point of Σ. ˜ ∂Q ∂Q ∂Q ∂z ∂2 z ∂R ∂R ∂z ∂z = + + + + R(x. y. using the parametrization r(x.4). N n= = N ∂z − ∂x i − ∂z ∂y j+k + ∂z 2 ∂y 1+ ∂z 2 ∂x is in fact a positive unit normal vector to Σ (see Figure 4. of the surface Σ. we have ∂z ∂z = i + ∂x k and ∂r = j + ∂y k. we can calculate ˜ ∂2 z ∂P ∂P ∂P ∂z ∂R ∂z ∂R ∂z ∂z = + + + +R . ∂x ∂z ∂x ∂x ∂y ∂z ∂x ∂y ∂x ∂y In a similar fashion.46) for curl f. recall from Section 2. . y) in D.49).5 Stokes’ Theorem Thus. upon comparing to equation (4. for (x. So we see that (curl f) · n dσ = Σ D (curl f ) · n ∂r ∂r × dA ∂x ∂y = D ∂z ∂z ∂P ∂R ∂Q ∂P ∂R ∂Q i+ j+ k · − i − j + k dA − − − ∂y ∂z ∂z ∂x ∂x ∂y ∂x ∂y − D = ∂P ∂R ∂z ∂Q ∂P ∂R ∂Q ∂z − − − + − ∂y ∂z ∂x ∂z ∂x ∂y ∂x ∂y dA . we have ∂r ∂x ∂r ∂x × ∂r ∂y = 1+ ∂z 2 ∂x + ∂z 2 ∂y . y). Hence. y)) ∂x ∂x ∂z ∂x ∂x ∂z ∂x ∂y ∂x ∂y 2z ∂Q ∂Q ∂z ∂R ∂z ∂R ∂z ∂z ∂ = + + + +R .4.49) after factoring out a −1 from the terms in the first two products in equation (4. Thus. − D C f · dr = ∂R ∂Q ∂z ∂P ∂R ∂z ∂Q ∂P − − − + − ∂y ∂z ∂x ∂z ∂x ∂y ∂x ∂y dA (4. proves the Theorem.3 (see p.47). ∂z ∂z Now.48).76) that the vector N = − ∂x i − ∂y j + k is normal to the tangent plane to the surface z = z(x.

14. z) = z i + x j + y k when Σ is the paraboloid z = x2 + y2 such that z ≤ 1 (see Figure 4. so switching to polar coordinates gives 2π 1 0 2π 1 0 2π = 0 (−2r cos θ − 2r sin θ + 1)r dr dθ (−2r2 cos θ − 2r2 sin θ + r) dr dθ 3 = 0 = 0 2π − 2r cos θ − 3 2r3 3 sin θ + 1 2 r=1 r2 2 r=0 dθ = 0 − 2 cos θ − 2 sin θ + 3 3 2π 0 dθ 2 = − 2 sin θ + 3 cos θ + 1 θ 3 2 =π. y. LINE AND SURFACE INTEGRALS Note: The condition in Stokes’ Theorem that the surface Σ have a (continuously varying) positive unit normal vector n and a boundary curve C traversed n-positively can be expressed more precisely as follows: if r(t) is the position vector for C and T(t) = r ′ (t)/ r ′ (t) is the unit tangent vector to C. then (curl f ) · n dσ = Σ D (curl f ) · n ∂r ∂r × dA ∂x ∂y 1 + 4x2 + 4y2 dA = D −2x − 2y + 1 1+ 4x2 + 4y2 = D (−2x − 2y + 1) dA .5). Verify Stokes’ Theorem for f(x. Solution: The positive unit normal vector to the surface z = z(x. n. it should be noted that Stokes’ Theorem holds even when the boundary curve C is piecewise smooth. y) = x i + y j + (x2 + y2 ) k for (x. Example 4. so (curl f ) · n = (−2x − 2y + 1)/ 1 + 4x2 + 4y2 . y) = x2 + y2 is n= ∂z − ∂x z C n 1 i− ∂z ∂y j+k + ∂z 2 ∂y 1+ ∂z 2 ∂x = −2x i − 2y j + k 1 + 4x2 + 4y2 . .172 CHAPTER 4. Σ y 0 x Figure 4.5. T × n form a right-handed system. y) in the region D = { (x. then the vectors T. Also.5.5 z = x2 + y2 Since Σ can be parametrized as r(x. y) : x2 + y2 ≤ 1 }. and curl f = (1 − 0) i + (1 − 0) j + (1 − 0) k = i + j + k.

as predicted by Stokes’ Theorem. Example 4.15.5. so (curl f ) · n = x (−4y)(− 2 ) + (9x)(− 2y ) + (0)(1) 9 1+ x2 4 + 4y2 9 = 2xy − 2xy + 0 1+ x2 4 + 4y2 9 =0. y. In this case. except now the boundary 2 2 curve C is the ellipse x + y9 = 1 laying in the plane z = 1. The line integral in the preceding example was far simpler to calculate than the surface integral. and so by Stokes’ Theorem f · dr = Σ C (curl f ) · n dσ = Σ 0 dσ = 0 .14. y)) on the surface z = z(x. y) = x + y9 the vector 4 n= ∂z − ∂x i − ∂z ∂y 2 2 j+k + ∂z 2 ∂y 1+ ∂z 2 ∂x = x −2 i − 2y 9 x2 4 j+k + 4y2 9 . but this will not always be the case. which can be parametrized as x = cos t. 2 4 0 − sin t + f · dr = Σ here we used cos2 t = 1 + cos 2t 2 So we see that C (curl f ) · n dσ. As in Example 4. where C is traversed counterclockwise. using Stokes’ 4 Theorem is easier than computing the line integral directly. Let Σ be the elliptic paraboloid z = x + y9 for z ≤ 1. y. 1+ is a positive unit normal vector to Σ. z) = (9xz + 2y)i + (2x + y2 )j + (−2y2 + 2z)k.5). Solution: The surface is similar to the one in Example 4. at 2 2 each point (x.14. z = 1 for 0 ≤ t ≤ 2π. . y = sin t. Calculate C f · dr for f(x. And calculating the curl of f gives curl f = (−4y − 0)i + (9x − 0)j + (2 − 2)k = −4y i + 9x j + 0 k .4. z(x. So 2π C f · dr = = ((1)(− sin t) + (cos t)(cos t) + (sin t)(0)) dt 0 2π 1 + cos 2t dt 2 0 t sin 2t 2π = cos t + + =π. and let C be its 4 boundary curve.5 Stokes’ Theorem 173 The boundary curve C is the unit circle x2 + y2 = 1 laying in the plane z = 1 (see Figure 4.

for a simple closed curve C the line integral C f· dr is often called the circulation of f around C. Suppose we have a vector field f(x. and it would rotate clockwise if it were dropped to the left of the y-axis. that is. Think of the vector field as representing the flow of water. then it turns out8 that curl E = 0. 2 in R EITZ. y. Notice that if all the vectors had the same direction and the same magnitude.e. 78-81. See S CHEY.5. In the limit. and imagine dropping two wheels with paddles x into that water flow. In fact. For example. That ratio of circulation to surface area in the limit is what makes the curl a rough measure of circulation density (i. Vector fields which have zero curl are often called irrotational fields. the magnitude of f is larger) as you move away from the y-axis. z) points in the direction of your thumb as you cup your right hand in the direction of the rotation of the wheel. Namely. which means that the circulation C E · dr = 0 by Stokes’ Theorem.6. z) is from the y-axis. z). z) = lim 1 S →0 S f · dr . p. M ILFORD and C HRISTY. n · (curl f )(x. For example. for the derivation. to have smaller and smaller surface area.174 CHAPTER 4. z) in 3 . y. where it is used extensively. the surface it bounds. y. 8 9 See Ch. z). as in Figure 4. z) = 2x k in our example) and would obey the right-hand rule. In physics. y.e. think of the curve C shrinking to the point (x. f(x. y. z) = (1 + x2 ) j. z) which is always parallel f to the xy-plane at each point (x.50) C where S is the surface area of a surface Σ containing the point (x. . This is best seen by using another definition of curl f which is equivalent9 to the definition given by formula (4.6. y. if E represents the electrostatic field due to a point charge. LINE AND SURFACE INTEGRALS · In physical applications. y. for a point (x.5. meaning no rotation). Since the 0 flow is stronger (i. the curl is interpreted as a measure of circulation density. So the curl points outward (in the positive z-direction) if x > 0 and points inward (in the negative z-direction) if x < 0. y. y. y An idea of how the curl of a vector field is related to rotation is shown in Figure 4. curl f(x. then such a wheel Figure 4. y. In both cases the curl would be nonzero (curl f(x. (4. y. z) and that the vectors grow larger the further the point (x. z) and with a simple closed boundary curve C and positive unit normal vector n at (x.6 Curl and rotation would rotate counterclockwise if it were dropped to the right of the y-axis. circulation per unit area). then the wheels would not rotate and hence there would be no curl (which is why such fields are called irrotational. which causes Σ. the term curl was created by the 19th century Scottish physicist James Clerk Maxwell in his study of electromagnetism.46).5.

y. So similar to the two-variable case. z) does not have a potential in  A For Exercises 1-3.e. calculate C Exercises ©  ¨ f (x. ∂y ∂z throughout 3. regions having no holes): The following statements are equivalent for a simply connected solid region S in 3: (a) f(x. z) = xy. z) in ∂x 3. we just need to check whether curl f = 0 throughout 3 . and R(x. y. z) ds for the given function f (x. y. then f · dr = Σ C (curl f ) · n dσ = Σ 0 · n dσ = Σ 0 dσ = 0 . y. by Stokes’ Theorem. Thus. Example 4. z) is a smooth vector field such that curl f = 0 in S . y. y. z) = xz. But we see that ∂R ∂P = xy . y.16. z) = xyz i + xz j + xy k has a potential in 3 . z) i + Q(x. z) = P(x. that is. 3 . z) k has a smooth potential F(x. z) in S (b) C f · dr is independent of the path for any curve C in S f · dr = 0 for every simple closed curve C in S (c) C (d) ∂Q ∂P ∂R ∂Q ∂P ∂R = . y. we have a three-dimensional version of a result from Section 4. y. ∂z ∂x and ∂Q ∂P = ∂x ∂y where P(x. z) = xyz. z) j + R(x. where Σ is any orientable surface inside S whose boundary is C (such a surface is sometimes called a capping surface for C). f(x. . for solid regions in 3 which are simply connected (i. ∂P ∂R = . Q(x.5 Stokes’ Theorem 175 Finally.e.3. =y ∂z ∂x ⇒ ∂P ∂z ∂R for some (x. we know that if C is a simple closed curve in some solid region S in 3 and if f(x. curl f = 0 in S ) ∂y ∂z ∂z ∂x ∂x ∂y Part (d) is also a way of saying that the differential form P dx + Q dy + R dz is exact. y. y.4. y. and = in S (i. ∂R ∂Q = . Determine if the vector field f(x. z) and curve C. Solution: Since 3 is simply connected. = . y. y.

f(x. z = t. 0. f(x. z ≤ 1 16. f(x. y. f(x. z) = (x + y) i + x j + z2 k 11. 1 ≤ t ≤ 2 y 3. 0 ≤ t ≤ 2π C : x = cos t. z) = i − j + k. y. Construct a Möbius strip from a piece of paper.3(b)). f(x. y. f(x. C : x = cos t. y = sin t. z) = y i − x j + z k 12. 0) to (1. 0. C : x = t sin t. f(x. y. f (x. . z) = x i + y j + z k. y = t cos t. Show that (curl f ) · n dσ = 0. z = 1. z) = yz i + xz j + xy k. z) = xy i − (x − yz2 ) j + y2 z k B For Exercises 14-15. y. y = 2t. Show that Green’s Theorem is a special case of Stokes’ Theorem. y. y. 0 ≤ t ≤ 1 f · dr for the given vector field f(x. 5. 10. y. z) and surface Σ. −2) For Exercises 10-13. C : x = 3t. c constant) 13. 0 ≤ t ≤ 2π x 2. f (x. z) = z2 . z) and curve C. z) = 2y i − x j + z k. 15. 2. C : x = t2 . f(x. (1. − 2 ≤ v ≤ 2 2 2 1 2 C 18. y = t. z = C √ 2 2 3/2 . y = sin t. 8. y. z = t. 0 ≤ t ≤ 2π C : x = t. f(x. z) has a potential in (you do not need to find the potential itself). z) = a i + b j + c k (a. C : x = cos t. state whether or not the vector field f(x. z = t2 − 1. z ≥ 0 Σ : z = x2 + y2 . y. z = 2. verify Stokes’ Theorem for the given vector field f(x. 2. z) a smooth vector field. 9. 0) to (1. 3 t 0≤t≤1 For Exercises 4-9. (Hint: Split Σ in half. Use Gnuplot (see Appendix C) to plot the Möbius strip parametrized as: 1 r(u. y. 0) to (1. v) = cos u (1 + v cos u ) i + sin u (1 + v cos u ) j + v sin u k . z) = + y + 2yz. y. f(x.) Σ 19. How many surfaces does this result in? How would you describe them? Are they orientable? 17. 0) C : the polygonal path from (0. Cut the Möbius strip along that center line completely around the strip. y. 0) to (1. z) = (y − 2z) i + xy j + (2xz + y) k. Let Σ be a closed surface and f(x. z) = xy i + (z − x) j + 2yz k. y. calculate 4. 0. z = t. y. 0) to 3 7.5. y. 6. f (x. f(x. y = sin t. 0. 2. 14. y. 0 ≤ t ≤ 1 C : the polygonal path from (0. Σ : x2 + y2 + z2 = 1. y. y. 0 ≤ u ≤ 2π . LINE AND SURFACE INTEGRALS 1. z) = xy i + xz j + yz k. y = 2t. f(x. z) = y i − x j + z k. z) = z. then draw a line down its center (like the dotted line in Figure 4. b.176 CHAPTER 4.

y. the symbols ∂x . z). z) produces ∂ f . the dot product of f with ∇ (thought of as a vector) makes sense: ∇· f = ∂ ∂ ∂ i+ j+ k · f1 (x. y. z) is a vector-valued function on 3 . you can think of the symbol ∇ as being “applied” to a real-valued function f to produce a vector ∇ f . We will then show how to write these quantities in cylindrical and spherical coordinates. For a real-valued function f (x. since for a vector field f(x. where each of the partial derivatives is evaluated at the point (x. But it helps to think of ∇ as a vector. z) = f1 (x. since ∂x . y. y.6 Gradient. as we ∂ ∂ ∂ will soon see. It turns out that the divergence and curl can also be expressed in terms of the symbol ∇. The process of “applying” ∂x . y. ∂z to a real-valued function f (x.4. the gradient ∇ f (x. ∂x ∂x ∂f ∂ (f) = . that is. y. For example.6 Gradient. say f (x. ∂y and ∂z are to be thought of as “partial derivative operators” that will get “applied” to a real-valued function. z)j + f3 (x. y. ∇ is often referred to as the “del operator”. Curl and Laplacian 177 4. y. z) is normally thought of as multiplying the quantities: ∂f ∂ (f) = . . ∂x ∂y ∂z (4. divergence and curl. z) is the vector ∇ f (x. z)j + f3 (x. z) on 3 . namely ∇= ∂ ∂ ∂ i+ j+ k. no. y. For instance. z)i + f2 (x. This is done by thinking of ∇ as a vector in 3 . y. So in this way. since it “operates” on functions. and we will also introduce a new quantity called the Laplacian. its value at a point (x. z)k. to produce the partial ∂ derivatives ∂ f . y. Curl and Laplacian In this final section we will establish some relationships between the gradient. ∂x “applied” to f (x. ∂x ∂y ∂z ∂x ∂ ∂ ∂ Is ∇ really a vector? Strictly speaking. z)k ∂x ∂y ∂z ∂ ∂ ∂ ( f1 ) + ( f2 ) + ( f3 ) = ∂x ∂y ∂z ∂ f1 ∂ f 2 ∂ f 3 = + + ∂x ∂y ∂z = div f . especially with the divergence and curl. y. Divergence. i+ j+ k ∂x ∂y ∂z ∂x ∂y ∂z in 3 .51) ∂ ∂ ∂ Here. z) = ∂f ∂f ∂f ∂f ∂f ∂f = . z)i + f2 (x. ∂y ∂y ∂f ∂ (f) = ∂z ∂z For this reason. ∂ f and ∂ f . z). ∂y . it is often convenient to write the divergence div f as ∇ · f. ∂y and ∂z are not actual numbers. y. y. Divergence. y.

y. Let r(x. y. z) = x i + y j + z k be the position vector field on r(x. namely as ∇ × f. we have: i j k ∂ ∂ ∂ ∇×f = ∂x ∂y ∂z P(x. y. y. z) = P(x.52) ∂x ∂y ∂z Often the notation ∇2 f is used for the Laplacian instead of ∆ f . z) 2 = r · r = x2 + y2 + z2 is a real-valued function. z) = vector field. z) R(x. so we can take its divergence: div ∇ f = ∇ · ∇ f = ∂f ∂x ∂f ∂y ∂f ∂z i+ j+ k is a ∂ ∂ ∂ i+ j+ k · ∂x ∂y ∂z ∂ ∂f ∂ ∂f = + + ∂x ∂x ∂y ∂y ∂2 f ∂2 f ∂2 f = 2 + 2 + 2 ∂x ∂y ∂z ∂f ∂f ∂f i+ j+ k ∂x ∂y ∂z ∂ ∂f ∂z ∂z Note that this is a real-valued function. y. z) Q(x. z). y. using the convention ∇2 = ∇ · ∇. z) = ∂R ∂P ∂Q ∂P ∂R ∂Q i − j + k − − − ∂y ∂z ∂x ∂z ∂x ∂y ∂R ∂Q ∂P ∂R ∂Q ∂P = i + j + k − − − ∂y ∂z ∂z ∂x ∂x ∂y = curl f For a real-valued function f (x. to which we will give a special name: Definition 4. the Laplacian of f . is given by ∂2 f ∂2 f ∂2 f ∆ f (x. Example 4. z)k. denoted by ∆ f . For a real-valued function f (x. Then (b) the divergence of r (c) the curl of r (d) the Laplacian of r 2 . Find (a) the gradient of r 2 3. z)j + R(x.17.7. the gradient ∇ f (x. z). y. z) = ∇ · ∇ f = 2 + 2 + 2 . y. y. (4. since for a vector field f(x. z)i + Q(x.178 CHAPTER 4. y. y. LINE AND SURFACE INTEGRALS We can also write curl f in terms of ∇. y. y.

Also.15 is that gradients are irrotational. Proof: We see by the smoothness of f that i j k ∂ ∂ ∂ ∇ × (∇ f ) = ∂x ∂y ∂z ∂f ∂f ∂f ∂x ∂y ∂z ∂2 f ∂2 f ∂2 f ∂2 f ∂2 f ∂2 f − − − i− j+ k=0. ∇ · (∇ × f) = 0. notice that in Example 4. z) has a potential. For any smooth real-valued function f (x. Divergence. The following theorem shows that this will be the case in general: Theorem 4. QED 2 we get . The following theorem shows that this will be the case in general: Theorem 4. = ∂y ∂z ∂z ∂y ∂x ∂z ∂z ∂x ∂x ∂y ∂y ∂x since the mixed partial derivatives in each component are equal. For any smooth vector field f(x. ∇ × (∇ f ) = 0. y. y. If a vector field f(x.4. Corollary 4. y. z). The proof is straightforward and left as an exercise for the reader.17. Curl and Laplacian Solution: (a) ∇ r (b) ∇ · r = (c) i ∂ ∇× r = ∂x x (d) ∆ r 2 ∂ ∂x (x) 2 179 = 2x i + 2y j + 2z k = 2 r + ∂ ∂z (z) + ∂ ∂y (y) =1+1+1=3 j ∂ ∂y y ∂2 (x2 ∂y2 k ∂ = (0 − 0) i − (0 − 0) j + (0 − 0) k = 0 ∂z z + y2 + z2 ) + ∂2 (x2 ∂z2 = ∂2 (x2 ∂x2 + y2 + z2 ) + + y2 + z2 ) = 2 + 2 + 2 = 6 Note that we could have calculated ∆ r 2 another way.16. z). Another way of stating Theorem 4.6 Gradient. then curl f = 0.17 if we take the curl of the gradient of r ∇ × (∇ r 2 ) = ∇ × 2 r = 2 ∇ × r = 2 0 = 0 .17 if we take the divergence of the curl of r we trivially get ∇ · (∇ × r) = ∇ · 0 = 0 . using the ∇ notation along with parts (a) and (b): ∆ r 2 = ∇ · ∇ r 2 = ∇ · 2 r = 2 ∇ · r = 2(3) = 6 Notice that in Example 4.15.

Proof: Let Σ be a closed surface which bounds a solid S . and physicists do not usually bother to prove it. LINE AND SURFACE INTEGRALS Corollary 4. y. (by Theorem 4. we see that we must have ∇ × (∇ f ) = 0 in 3 . if the surface integral f (x. The proof is not trivial. A system of electric charges has a charge density ρ(x. y. This is one of Maxwell’s Equations. y.17) QED There is another method for proving Theorem 4. z) is a smooth real-valued function on 3 . y. z) at points (x. The flux of the curl of a smooth vector field f(x. to prove Theorem 4.180 CHAPTER 4. z) in space. j and k in place of n.18. Since ∇ f is a vector field.10 10 In Gaussian (or CGS) units.15.18. then (∇ × (∇ f )) · n dσ = Σ C Σ ∇ f · dr by Stokes’ Theorem. and is often used in physics. z) and produces an electrostatic field E(x. where n is any unit vector. y. Using i. z) = 0 throughout that region. The flux of ∇ × f through Σ is (∇ × f ) · dσ = Σ S ∇ · (∇ × f ) dV 0 dV S (by the Divergence Theorem) = =0. then we must have (∇× (∇ f ))· n = 0 throughout 3 .e.13. Σ is orientable and its boundary is C). so = 0 by Corollary 4. z) dσ = 0 for all surfaces Σ in some solid region (usually all of 3 ).15 which can be useful. y. × · Since the choice of Σ was arbitrary. assume that f (x. Let C be a simple closed curve in 3 and let Σ be any capping surface for C (i. For instance. Gauss’ Law states that E · dσ = 4π Σ S ρ dV for any closed surface Σ which encloses the charges. Example 4. z) through any closed surface is zero. with S being the solid region enclosed by Σ. and can also be applied to double and triple integrals. y. which completes the proof. then we must have f (x. Namely. . Show that ∇ · E = 4πρ. But the result is true.

z) z 0 x θ x y r (x. respectively (see Figure 4. We will present the formulas for these in cylindrical and spherical coordinates. φ). ez be unit vectors in the direction of increasing r. respectively (see Figure 4. y. ez z (x. φ). eθ .6. Then er .6 Gradient. so combining the integrals gives ∇ · E − 4πρ = 0 since Σ and hence S was arbitrary. z. By the right-hand rule. y = ρ sin φ sin θ. 0) x eθ er y 0 x θ y z (x. y. Often (especially in physics) it is convenient to use other coordinate systems when dealing with quantities such as the gradient.6. let er . y.6. θ. z) can be represented in cylindrical coordinates (r. Note. θ.2). θ. y. by the right-hand rule. so ∇ · E = 4πρ . eφ in spherical coordinates Similarly. ez in cylindrical coordinates Figure 4.1 Orthonormal vectors er . z) can be represented in spherical coordinates (ρ. ez form an orthonormal set of vectors.7 that a point (x. y.1). so S by Gauss’ Law. y.2 Orthonormal vectors eρ .6. eθ . z). where x = r cos θ. eφ are orthonormal. eθ . At each point (ρ. y = r sin θ. φ. that ez × er = eθ . z = ρ cos φ. θ. cylindrical and spherical coordinates in the following tables: . At each point (r. curl and Laplacian. eθ . z = z. θ. divergence. we have ∇ · E dV = S Σ 181 E · dσ ρ dV S = 4π (∇ · E − 4πρ) dV = 0 . a point (x. curl and Laplacian in Cartesian. eθ . Curl and Laplacian Solution: By the Divergence Theorem. Then the vectors eρ . 0) Figure 4. divergence. eθ . z). Recall from Section 1. where x = ρ sin φ cos θ. we see that eθ × eρ = eφ . eφ be unit vectors in the direction of increasing ρ. θ. We can now summarize the expressions for the gradient. Divergence. z) φ ρ z eρ eθ eφ y (x.4. let eρ .

φ): Scalar function F. θ. y. we will derive the formula for the gradient in spherical coordinates. z): Scalar function F. . Vector field f = f1 i + f2 j + f3 k ∂F ∂F ∂F i+ j+ k ∂x ∂y ∂z ∂ f1 ∂ f2 ∂ f3 divergence : ∇ · f = + + ∂x ∂y ∂z ∂ f3 ∂ f2 ∂ f1 ∂ f3 ∂ f2 ∂ f1 − − − curl : ∇ × f = i+ j+ k ∂y ∂z ∂z ∂x ∂x ∂y ∂2 F ∂2 F ∂2 F Laplacian : ∆ F = 2 + 2 + 2 ∂x ∂y ∂z gradient : ∇F = Cylindrical (r. z): Scalar function F.182 CHAPTER 4. θ. Vector field f = fρ eρ + fθ eθ + fφ eφ 1 ∂F 1 ∂F ∂F eρ + eθ + eφ ∂ρ ρ sin φ ∂θ ρ ∂φ 1 ∂ 2 ∂ 1 ∂ fθ 1 divergence : ∇ · f = 2 (ρ fρ ) + + (sin φ fθ ) ρ sin φ ∂θ ρ sin φ ∂φ ρ ∂ρ ∂ fφ ∂ fρ 1 ∂ ∂ 1 eρ + eθ (sin φ fθ ) − (ρ fφ ) − curl : ∇ × f = ρ sin φ ∂φ ∂θ ρ ∂ρ ∂φ 1 ∂ fρ 1 ∂ + − (ρ fθ ) eφ ρ sin φ ∂θ ρ ∂ρ ∂2 F 1 ∂ 2 ∂F 1 ∂F ∂ 1 Laplacian : ∆ F = 2 ρ + sin φ + 2 2 sin2 φ ∂θ2 ∂ρ ∂φ ρ ∂ρ ρ sin φ ∂φ ρ gradient : ∇F = The derivation of the above formulas for cylindrical and spherical coordinates is straightforward but extremely tedious. The basic idea is to take the Cartesian equivalent of the quantity in question and to substitute into that formula using the appropriate coordinate transformation. Vector field f = fr er + fθ eθ + fz ez ∂F 1 ∂F ∂F er + eθ + ez ∂r r ∂θ ∂z 1 ∂ fθ ∂ f z 1 ∂ (r fr ) + + divergence : ∇ · f = r ∂r r ∂θ ∂z 1 ∂ ∂ fr 1 ∂ fz ∂ fθ ∂ fr ∂ f z curl : ∇ × f = er + eθ + ez − − (r fθ ) − r ∂θ ∂z ∂z ∂r r ∂r ∂θ 1 ∂ ∂F 1 ∂2 F ∂2 F Laplacian : ∆ F = r + 2 2 + 2 r ∂r ∂r r ∂θ ∂z gradient : ∇F = Spherical (ρ. LINE AND SURFACE INTEGRALS Cartesian (x. As an example.

θ. ∂F . eρ = r xi + yj + zk = . z = ρ cos φ. j. eθ .4.6. Thus. and ρ = eρ = sin φ cos θ i + sin φ sin θ j + cos φ k Now. y. θ and φ. note that since eθ ⊥ eρ . Then put the partial derivatives ∂F . eθ . j. Since this vector is also a unit vector and points in the (positive) θ direction. which will give us an equation involving just i and j. and we see that a vector perpendicular to that is − sin θ i + cos θ j + 0 k. r x2 + y2 + z2 x2 + y2 + z2 . k in terms of eρ . then the unit vector eθ in the θ direction must be parallel to the xy-plane. ∂F in terms of ∂F . eφ and functions of ρ.2 that the unit vector eρ in the ρ direction at a general r point (ρ. then in particular eθ ⊥ eρ when eρ is in the xy-plane. z) = ∂F i + ∂F j + ∂F k. since eφ = eθ × eρ . This comes down to solving a system of three equations in three unknowns. which we will use to solve first for j then for i. Lastly. it must be eθ : eθ = − sin θ i + cos θ j + 0 k Lastly. θ. k. First. There are many ways of doing this. ∂θ Step 1: Get formulas for eρ . eφ in terms of i. we will solve for k. θ and φ. but we will do it by combining the formulas for eρ and eφ to eliminate k. Curl and Laplacian 183 Goal: Show that the gradient of a real-valued function F(ρ. φ) is eρ = r .6 Gradient. eφ . since the angle θ is measured in the xy-plane. Putting φ = π/2 into the formula for eρ gives eρ = cos θ i + sin θ j + 0 k. note that sin φ eρ + cos φ eφ = cos θ i + sin θ j . Divergence. We can see from Figure 4. put the Carte∂x ∂y ∂z sian basis vectors i. we get: eφ = cos φ cos θ i + cos φ sin θ j − sin φ k Step 2: Use the three formulas from Step 1 to solve for i. with the formula for eθ . φ) in spherical coordinates is: 1 ∂F 1 ∂F ∂F eρ + eθ + eφ ∇F = ∂ρ ρ sin φ ∂θ ρ ∂φ Idea: In the Cartesian gradient formula ∇F(x. To figure out what a and b are. where r = x i + y j + z k is the position vector of the point in Cartesian coordinates. j. y = ρ sin φ sin θ. That occurs when the angle φ is π/2. we get: so using x = ρ sin φ cos θ. ∂x ∂y ∂z ∂ρ ∂F ∂F . eθ is of the form a i + b j + 0 k. That is. eθ . This. ∂φ and functions of ρ. will then leave us with a system of two equations in two unknowns (i and j). k in terms of the spherical coordinate basis vectors eρ .

we see that: k = cos φ eρ − sin φ eφ Step 3: Get formulas for ∂F . ∂F . Using a similar process of elimination as in Step 2. ∂ρ ∂x ∂ρ ∂y ∂ρ ∂z ∂ρ ∂F ∂F ∂x ∂F ∂y ∂F ∂z = + + . we see that cos θ (sin φ eρ + cos φ eφ ) − sin θ eθ = (cos2 θ + sin2 θ)i = i .184 so that CHAPTER 4. ∂x ∂y ∂z ∂ρ ∂θ ∂F . ∂F . ∂ρ ∂θ By the Chain Rule. we get: ∂F 1 ∂F ∂F ∂F = − sin θ + sin φ cos φ cos θ ρ sin2 φ cos θ ∂x ρ sin φ ∂ρ ∂θ ∂φ ∂F 1 ∂F ∂F ∂F ρ sin2 φ sin θ = + cos θ + sin φ cos φ sin θ ∂y ρ sin φ ∂ρ ∂θ ∂φ ∂F 1 ∂F ∂F ρ cos φ = − sin φ ∂z ρ ∂ρ ∂φ . ∂θ ∂x ∂θ ∂y ∂θ ∂z ∂θ ∂F ∂F ∂x ∂F ∂y ∂F ∂z = + + . and so: i = sin φ cos θ eρ − sin θ eθ + cos φ cos θ eφ Lastly. ∂y . ∂φ ∂x ∂φ ∂y ∂φ ∂z ∂φ which yields: ∂F ∂F ∂F ∂F = sin φ cos θ + sin φ sin θ + cos φ ∂ρ ∂x ∂y ∂z ∂F ∂F ∂F = −ρ sin φ sin θ + ρ sin φ cos θ ∂θ ∂x ∂y ∂F ∂F ∂F ∂F = ρ cos φ cos θ + ρ cos φ sin θ − ρ sin φ ∂φ ∂x ∂y ∂z Step 4: Use the three formulas from Step 3 to solve for ∂F . and so: j = sin φ sin θ eρ + cos θ eθ + cos φ sin θ eφ Likewise. this involves solving a system of three equations in three unknowns. we have ∂F ∂φ in terms of ∂F ∂F ∂F ∂x . ∂F in terms of ∂F . ∂z . ∂F ∂F ∂x ∂F ∂y ∂F ∂z = + + . ∂F . LINE AND SURFACE INTEGRALS sin θ (sin φ eρ + cos φ eφ ) + cos θ eθ = (sin2 θ + cos2 θ)j = j . ∂φ Again.

where r(x. Verify that we get the same answers if we switch to spherical coordinates. y.19. φ) = ρ2 (so that F(ρ.17 we showed that ∇ r 2 = 2 r and ∆ r 2 = 6. Doing this last step is perhaps the most tedious. and 8 terms involving eφ . as we showed earlier. ∂F . since it involves simplifying 3 × 3 + 3 × 3 + 2 × 2 = 22 terms! Namely. as expected. θ. y.4. And the Laplacian is ρ ∂ 1 ∂2 F 1 ∂F 1 ∂ 2 ∂F + 2 ρ + sin φ ∆F = 2 2 2 ∂ρ ∂φ ρ ∂ρ ρ sin φ ∂φ ρ2 sin φ ∂θ ∂ 1 1 ∂ 2 1 (sin φ (0)) (0) + 2 = 2 (ρ 2ρ) + 2 ρ ∂ρ ρ sin φ ρ sin φ ∂φ 1 ∂ (2ρ3 ) + 0 + 0 = 2 ρ ∂ρ 1 = 2 (6ρ2 ) = 6 . Curl and Laplacian 185 Step 5: Substitute the formulas for i. But the algebra is straightforward and yields the desired result: ∇F = ∂F 1 ∂F 1 ∂F eρ + eθ + eφ ∂ρ ρ sin φ ∂θ ρ ∂φ Example 4. k from Step 2 and the formulas for ∂F . so r r = 2ρ = 2 r .6 Gradient. φ) = r 2 ). j. let F(ρ. z) = ∂x i + ∂y j + ∂z k. ρ ∇F = . ∇F = 1 ∂F ∂F ∂F − sin θ + sin φ cos φ cos θ ρ sin2 φ cos θ (sin φ cos θ eρ − sin θ eθ ρ sin φ ∂ρ ∂θ ∂φ + cos φ cos θ eφ ) + 1 ∂F ∂F ∂F ρ sin2 φ sin θ (sin φ sin θ eρ + cos θ eθ + cos θ + sin φ cos φ sin θ ρ sin φ ∂ρ ∂θ ∂φ + cos φ sin θ eφ ) + 1 ∂F ∂F − sin φ ρ cos φ (cos φ eρ − sin φ eφ ) . θ. In Example 4. as expected. The gradient of F in spherical coordinates is ∂F 1 ∂F 1 ∂F eρ + eθ + eφ ∂ρ ρ sin φ ∂θ ρ ∂φ 1 1 (0) eθ + (0) eφ = 2ρ eρ + ρ sin φ ρ r = 2ρ eρ = 2ρ . Divergence. ∂F ∂x ∂y ∂z ∂F ∂F ∂F from Step 4 into the Cartesian gradient formula ∇F(x. ρ ∂ρ ∂φ which we see has 8 terms involving eρ . z) = x i + y j + z k in Cartesian coordinates. Solution: Since r 2 = x2 + y2 + z2 = ρ2 in spherical coordinates. 6 terms involving eθ .

11. Define the normal derivative ∂n of u over a closed surface Σ with outward unit normal vector n by ∂u = Dn u = n · ∇u.186 CHAPTER 4.) ∂n Σ . Find the Laplacian of the function in Exercise 3 in spherical coordinates. ∆ (1/r) = 0 14.e. Use f = u ∇v in the Divergence Theorem to prove: (a) Green’s first identity: S ∂F ∂r er + 1 r ∂F ∂θ eθ + ∂F ∂z ez (u ∆ v + (∇u) · (∇v)) dV = S Σ (u ∇v) · dσ (b) Green’s second identity: (u ∆ v − v ∆ u) dV = Σ (u ∇v − v ∇u) · dσ ∂u 27. y. ∂n ∂u Show that dσ = 0. prove the given formula (r = r is the length of the position vector field r(x. y. 12. z) = x3 + y3 + z3 7. f (x. find div f and curl f. θ. 9. find div f and curl f. y. z) = x i + y j + z k). y. ∇ (1/r) = −r/r3 13.17. div (F + G) = div F + div G 18. Suppose that ∆ u = 0 (i. z) = 2 x + y2 10. y. z) = x + y + z 4. ∇ (ln r) = r/r2 16. (Hint: Use Green’s second identity. div (∇ f × ∇g) = 0 22. 3. θ. 25. div (F × G) = G · curl F − F · curl G 21. curl (F + G) = curl F + curl G 19. z) in Cartesian coordinates. find the Laplacian of the function f (x. B For Exercises 12-23. z) = x5 ¨ For Exercises 1-6. y. ∇ · (r/r3 ) = 0 15. y. Let f (x. Prove Theorem 4. f (x. y. z in Cartesian coordinates. 8. f (x. div ( f F) = f div F + F · ∇ f 20. Find ∇ f in cylindrical coordinates. LINE AND SURFACE INTEGRALS  A 1. z) = e x+y+z Exercises ©  2. φ) = eρ + ρ cos θ eθ + ρ eφ in spherical coordinates. z) = (x2 + y2 + z2 )3/2 6. ∆ ( f g) = f ∆ g + g ∆ f + 2(∇ f · ∇g) C 24. Find the Laplacian of the function in Exercise 6 in spherical coordinates. z) = r er + z sin θ eθ + rz ez in cylindrical coordinates. Derive the gradient formula in cylindrical coordinates: ∇F = 26. u is harmonic) over 3 . f (x. curl ( f F) = f curl F + (∇ f ) × F 23. curl (curl F) = ∇(div F) − ∆ F 17. f (x. For f(r. f (x. y. z) = e−x 2 −y2 −z2 5. For f(ρ.

. Flatland. P.S.. Hoel..J. San Diego..G. Sherali and C. Nonlinear Programming: Theory and Algorithms. with a modern approach based on differential forms.D. New York: John Wiley & Sons. Very thorough. Elementary Linear Algebra: Applications Version. G. 1970 Standard intermediate-level treatment of classical mechanics. Most of the mathematics will be understandable after reading the present book. Jackson. H. New York: Dover Publications. 2000 Standard treatment of elementary linear algebra. famous for being intimidating. New York: John Wiley & Sons. covering a wide range of topics. and C.. Hecht. J. New York: Academic Press.. 1975 An advanced book on electromagnetism. Introduction to Probability Theory... Curves and Surfaces for Computer Aided Geometric Design: A Practical Guide. 2nd edition. 2nd edition. E. Farin. 2nd edition. H. 187 . Classical Electrodynamics. CA: Academic Press. 7th edition. O’Neill. calculus-based probability theory. Shetty. New York: John Wiley & Sons. 1952 Classic tale about a creature living in a 2-dimensional world who encounters a higherdimensional creature. New York: Academic Press. S. Marion.C.M. M. Elementary Differential Geometry. MA: Houghton Mifﬂin Co.. Anton. Classical Dynamics of Particles and Systems. 1966 Intermediate-level book on differential geometry. 1993 Thorough treatment of nonlinear optimization. Bazaraa.B. Stone. J. Rorres. 2nd edition. 1987 An intermediate-level book on optics. 2nd edition. Port and C. with lots of humor thrown in. Reading.. Optics. Boston. 1990 An intermediate-level book on curve and surface design..A. E. B.D. Lots of good exercises. Inc.Bibliography Abbott. MA: Addison-Wesley Publishing Co. 1971 An excellent introduction to elementary. 8th edition.

4◦ 5. 4x − 4y + 3z − 10 = 0 13. 3) 11. x = 5t. z = −7t 21. 57) 1. π . (−5. 1.5 (p. 6 . z = t 5. 24. since v · w = 0. 41 (e) (d) 2 √ 41 2 Section 1. −1) 5. center: (−1. √ 30 30 30 (b) (2. 14) 1. 1 − cos t. −4. (a) r2 + z2 = 25 (b) ρ = 5 7. y = 1 5. x − 2y − z + 2 = 0 √ 15. π ) 6 2 5. z = 0 and a = − b . −5) 5. Section 1. 4. √ 7. y = 2t. y = z = 1 3. −10. sin t. 3.6. cos t) 9. θ. 2) 15. v(t) = (1. 1.65 9. radius: 5. 0 and (8. 2 cos 2t. 90◦ 7. 0 √ 7. −1) (b) ( 17. No intersection. Hint: Use the distance formula for Cartesian coordinates. 1). −3) −2 −1 (c) √ . −3) y = 3 + 4t. y = 2+3t. (a. 7. a cot φ) 12. −2. 73. x = 1 + t. No (d) 1 (b) x = 2 + t.Appendix A Answers and Hints to Selected Exercises 3. −23. Hint: use Definition 1. 14 Section 1. circle x2 + y2 = 4 in the planes z = ± 5 y y x x 9. (a) (2 7. −4) (h) (−1. 4. z = −3 + 8t 7. 2) 3. 2−c . 3) + t(1. radius: 1. = y−3 = z−2 4 −3 189 . 0) (b) (2 7. 10 3. 0◦ 9. π . 3t2 ). 2t. 4 5 11. Yes 3. x = 1. 29) 1. 11π . 11x − 24y + 21z − 26 = 0 17. −1. 2−c . (a) (4. (j) No. (a) (2. (a) (2. f ′ (t) = (−2 sin 2t. 4. (8. 50) √ 1. 0.816) 3 3 √ √ 11π 3. (10. √ √ 11. |v · w| = 0 < 21 √5 = v w √ √ 13. y = −2 + 7t. (a) (−4. 8) √ 1. 8) (g) (−7. 39) 1. (a) r2 + 9z2 = 36 (b) ρ2 (1 + 8 cos2 φ) = 36 10. 9/ 35 19. −24) 3. −6. (a) Line parallel to c (b) Half-line Section 1. z = 3 + t (c) x − 2 = z − 3. 18) 1. a(t) = (0. v + w = 26 < 21 + 5 = v + w 15. z = 2 − 3t (c) x−2 5 (b) x = 2 + 5t. 1) Chapter 1 Section 1. center: (2. 2) + t(5. sin t). 9 13.2 (p. 0 (f) (14. 5) 3. lines a = b . x = 1 + 2t. y = 1. Section 1. v + w is larger. (1. −6. 1) Section 1. No.72 9.4 (p. Yes.10(c).3 (p.8 (p. 3. Section 1. 2.1 (p. 16.√ −1) 6. √5 .7 (p. 46) 1. f ′ (t) = (1. (a) 5 √ (e) 2 17 √ √ (b) 5 (c) 17 2.6 (p. 1) (i) (−2. Hint: See Theorem 1. 3. z = 0 2a 2b 13.

(1. 2y. − cos t. Hint: Theorem 1. ∂ 2 = x2 e xy . 0). domain: 9. y) : x2 + y2 ≥ 4}. 1) : → (1. then write 11 1 5. 70) 2. ∂ f = 12x2 . 1). − sin t. saddle pt.2858. 0). 2 13. domain: 3 . ∂x ∂y 2 ∂2 f xy + 1 23. ∂y 3 ∂x ∂f ∂x = 2x. ∂ f = ye xy + y. √ 3π 5 2 3.4 (p. y0 ) = (0. ∂ 2 = y2 e xy . saddle pt.9 (p. ( √ x .3998). range: [0. 1] 7. local min. and Theorem 1. ∂f ∂x ∂f −(x2 +y2 ) 15. 2 15. 1 decrease: (−45. 0) 7. 1). local min. 0) = xe xy + x 7. (0. ∂x ∂y = 0 25.16 7. 20). xy cos(xyz)) √ 1 1. 1) 5. ∂f 1 2 −1/2 5. ∂x 3 2 2 ∂f = 1 (x2 + y + 4)−2/3 13. 3x + 4y − 5z = 0 2 1 √ (sin t. (x0 . 88) Section 2. Hint: Use f ′ (t) = f(t) T. 3. x = y = 4. increase: (45. 77) 9. ∂2 f ∂2 f 2 −3/2 . ∂ f = 0 ∂y ∂x ∂y 9. (2x.190 Appendix A: Answers and Hints to Selected Exercises ∂2 f 1 = − 4 (x2 + y + 4)−3/2 . x + 2y = z 7. ∂ f = y cos(xy). ∂x2 = (y + 4)(x + y + 4) Section 2. 0) = x(x2 + y + 4)−1/2 . 63) 1. does not exist 11. (1. ∂ f = 4x3 . ∂x ∂y = 0 ∂y2 parallel to c (c) Hint: Think of the functions as position vectors. 2 2 13. 5. √3 √ {(x. (−1.6 (p. ∂x ∂y = 0 19. 3 ) = 0 9. T(t) = √ 2 11 1 √ (− sin t.16 Section 1. Hint: Use Exercise 6. −1). ∂y2 = 2. 2y) 3. −1. t by 27s+16 2 2 3/2 27 (13 2/3 − 8) 5. saddle pts. local min.37. ∞) 15. xz cos(xyz). x2 +y2 +4 √ y x2 +y2 +4 ) 5. 2 −4 9 6. 2 (x − 1) + 4 (y − 2) + 12 (z − ′ (t) in terms of N(t). ∂y = −2ye ∂x ∂2 f ∂2 f ∂f ∂y = x cos(xy) 17. −0. differentiate that to get f ′′ (t). ∂y2 ∂2 f 1 2 −3/2 ∂x ∂y = − 2 x(x + y + 4) 2f 2f 21. local max. 0 17. ∂f ∂y = 2y 3. −1). (−1. Replace B(t) = Theorem 1. ∂x ∂y = (1 + xy)e ∂x2 2f ∂2 f ∂2 f = 0. ∞) 3.20(e). domain: range: [−1. N(t) = (− cos t.1 (p. ∂y = 2 (x + y + 4) ∂x ∂f (1.94037) . (yz cos(xyz). (−1. 9 T 11. Section 2. 82) 1. Example 1. 95) 2. put those ex1. 2z) 11. 15. −2x + y − z − 2 = 0 √ pressions into f ′ (t) × f ′′ (t). 1. ∂ f = y(x2 + y2 )−1/2 ∂x ∂y = depth=10 13. 74) 1. 3 cos(1) 17. cos t. Section 2. width = height 9. local min. does not exist Section 2. ∂ 2 = −x−2 .2 (p.03256. ∂ f = 2x (x2 + y + 4)−2/3 . 2x + 3y − z − 3 = 0 3.3 (p. ∂ f = −2xe−(x +y ) . −20) 9. local min. 0) : → (0. ∂x2 = 2. z = 2 11. −1). 1). Hint: Use κ(t) = 1/2 Chapter 2 Section 2.5 (p. y0 ) = (1. (2x. (1/x. 1/y) 7. 1/2) 11. (0. ∂y2 ∂x ∂2 f ∂2 f = −y−2 . (−1. (1. (x0 . range: [−1. ∂ f = x(x2 + y2 )−1/2 .

F(x. 7/12. 1 n 20 30 √ . F(x. √ 13 13 −9 2 √ . 67/15 9. 163) 1. − √ . Other languages have similar functions.2 (p. max. min. 2π Chapter 3 Section 3. 112) 1. No 9. 2π(π − 1) 7. Section 3. 2 9. F(x. (17 17 − 5 5)/3 3. 0. 0. √ π 2. 1 4 7. 7 12 7 6 1 2 Section 4. 12π/5 7. (2 cos(π2 ) + π4 − 2)/4 5. − √ .5 (p. min. 4 . Both are n (n+1)2 (n+2) 7. Hint: Start by showing that er = cos θ i + sin θ j.7 (p. 3π/16) 7. y) j in 2 can be extended in a natural way to be a vector field in 3 .191 Section 2. 23 5. 2 2 π2 2.168 Section 4.318. 1. Yes. Hint: Think of how F is defined. 0 3.6 (p. Yes. 142) 1. 3π ) 5. 8π 3. Yes 13. (0. min. 15/4 Section 3.3 (p. ez = k. (0. 216π 2. y) i+ Q(x.7 (p.3 (p. 7/12) 1. 104) 1. 6(x + y + z) 2 7. 10.6 (p. (4ρ2 − 6)e−ρ 9. max. 9 8. 116) 1. 8 ln 2 − 3 5. 16/15 3. 1 − sin 2 2 − 33/2 ) 9. 8/3) 3.4 (p. 0 3. div f = ρ − sin φ + cot φ. 6 11. 175) √ √ √ 1. 4.5 (p. 149) 11. 1 3. 5 9. 12ρ 8. y) = P(x. Section 4. Yes. −2π 9. √ 5 5 Section 3. 5a/12) 9. 1 6. 15 1. ≈ 1. curl f = cot φ cos θ eρ + 2eθ − 2 cos θ eφ 25. 8abc √ 3 3 −4 −2 √ . 12 x2 + y2 + z2 5.1 (p. (1. 2 10. − 2z er + r12 ez r3 sin θ 2 11. 1 5. 1 3. otherwise use e = 2. (0. Hint: Think of how a vector field f(x. ≈ 0. √ 5 5 30 20 . eθ = − sin θ i + cos θ j. (Hint: In Java the exponential function e x can be obtained with Math. No 4. 109) 1. 3 3. 123) 1. 127) 4a 1. y) = xy2 + x3 7. y) = 4x2 y + 2y2 + 3x Section 3. 1 3 1 6 7. . No 19. 4π 7. The values should converge to ≈ 1. 186) 7. 134) 2 4 √ . Yes. F(x. 4π 3 (8 Section 4.exp(x). Section 3. max. 6 Section 4. y) = x − y2 2 5.146 3. (7/12. 155) 1.4 (p. 5.7182818284590455 in your program. 3. 1/2 3.705 4. 24π 11. (b) No. 10. 2πab Section 3.) 2. 100) 1. −5π 5. ≈ 0. 6 5 π 4 6. y) = axy + bx + cy + d 1 6 2 2 5. 7.2 (p. √ 5 5 13 13 √ 59 −1 4 . 9 3. 2/5 4. Chapter 4 Section 4.1 (p.

2. then n(v. k. then n(v. the angle θ between av and bw is 90◦ . Also.4. w). This was already shown in Example 1. We will consider the case when a > 0 and b > 0 (the other three possibilities are handled similarly). w) is perpendicular to the plane containing v and w. Hence the magnitude of n(av. by definition. then n(v. For example. 192 . For av = ai and bw = bk. j. k. k. bw) = ab(v × w) for any scalars a. bw) = 0 = ab(v × w). then n(av. If either a = 0 or b = 0 then n(av. and θ is the angle between them. If v and w are nonzero and parallel. There are four possibilities for the combinations of signs for a and b. bw) must be either |ab|j or −|ab|j. Let v and w be any two of the basis vectors i. w) is v w sin θ. as follows: 1. w) = 0. If either v or w is 0. Step 2: Show that n(av. namely.Appendix B We will prove the right-hand rule for the cross product of two vectors in For any vectors v and w in 3. bw) is perpendicular to the plane containing ai and bk. Since its magnitude is |ab|. w) = v × w if v and w are any two of the basis vectors i. So assume that a 0 and b 0. define a new vector. n(v.11 in Section 1. bw). Thus. 3. w) is the vector in 3 such that: (a) the magnitude of n(v. w in 3 . so the result holds. 3. is ai bk sin 90◦ = |ab|. which would prove the right-hand rule for the cross product (by part 1(c) of our definition). bw) must be a scalar multiple of j. b if v and w are any two of the basis vectors i. The goal is to show that n(v. the xz-plane. (b) n(v. w. j. we will perform the following steps: Step 1: Show that n(v. w) = 0. w) = v × w for all v. by definition. and (c) v. w) form a right-handed system. n(av. n(v. To do this. j. we will show that the result holds for v = i and w = k (the other possibilities follow in a similar fashion). If v and w are nonzero and not parallel. n(av.

which is shown in the figure below. n(ai. bk) = −abj. So since. ∴ n(av. which is what we would expect. n(ai. v) has magnitude 0. If θ is the angle between u and v. Now. Thus. bw) must be either abj or −abj. w) = n(u. this means that we must have n(ai. If u = 0 then the result holds trivially since n(u. then the projection vector . Therefore. k form a righthanded system.193 In this case. u v θ v θ P n(u. bk) form a right-handed system. which is what we needed to show. j form a left-handed system. ai. then the result follows easily since n(u. w) = n(u. since i. v and this vector form a right-handed system. b > 0. v + w) = n(u. v + w). w) and n(u. v). j. then we see that pro jP u v has magnitude u v sin θ. −j form a righthanded system. bk) = ab(i × k). Multiply the vector v by the positive scalar u . Since u (v + w) is the sum of the vectors u v and u w. But we know that ai × bk = ab(i × k) = ab(−j) = −abj. v) + n(u. v). w). v + w). bk) has to be either abj or −abj. v + w) = n(u. which is the magnitude of n(u. v) u pro jP u v Now apply this same geometric construction to get n(u. and since n(ai. then i. v). w) = 0 + n(u. −abj form a right-handed system (since a > 0. Let P be a plane perpendicular to u. bk. bw) = ab(v × w) Step 3: Show that n(u. ai. So now assume that u. If v = 0. since in that case θ = 0◦ and so sin θ = 0 which means that n(u. and ab > 0). w) are all the zero vector. We will describe a geometric construction of n(u. v) and which is perpendicular to pro jP u v (and hence perpendicular to v). v and w are all nonzero vectors. 0 + w) = n(u. and so i. n(av. You can think of this projection vector (denoted by pro jP u v) as the shadow of the vector u v on the plane P. v. v) + n(u. by definition. 0) = n(u. w) for any vectors u. Hence this vector must be n(u. bk. k. with the light source directly overhead the terminal point of u v. n(u. v) and n(u. A similar argument shows that the result holds if w = 0. So rotating pro jP u v by 90◦ in a counter-clockwise direction in the plane P gives a vector whose magnitude is the same as that of n(u. k. w. then project the vector u v straight down onto the plane P. Since this vector is in P then it is also perpendicular to u. Note that this holds even if u v. And we can see that u.

we have shown that −n(v. w) for any vectors v. Write v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k. n(v. By definition. n(v.194 Appendix B: Proof of the Right-Hand Rule for the Cross Product pro jP u (v + w) is the sum of the projection vectors pro jP u v and pro jP u w (to see this. v) + n(u. w. w) form a right-handed system. w) is a vector with the same magnitude as n(w. w). w) is perpendicular to the plane containing w and v. and that w. w. v. v) = −n(v. so the result holds. w) form a left-handed system. and hence so is −n(v. v. using the shadow analogy again and the parallelogram rule for vector addition. −n(v. which is the same as the magnitude of n(v. Then n(w. n(v. w). w) P n(u. or if either is 0. and hence is the same as the magnitude of −n(v. Step 5: Show that n(v. w. think of how projecting a parallelogram onto a plane gives you a parallelogram in that plane). then n(w. v. and so w. If v and w are nonzero and parallel. Then by Steps 3 and 4. w) = v × w for all vectors v. So assume that v and w are nonzero and not parallel. So then rotating all three projection vectors by 90◦ in a counter-clockwise direction in the plane P preserves that sum (see the figure below). Thus. we have . v) n(u. which means that n(u. v). w). v) = 0 = −n(v. w) must be n(w. u v u (v + w) v+w pro jP u v pro jP u (v + w) u w w θ θ v u pro jP u w n(u. v + w) Step 4: Show that n(w. w) form a right-handed system. −n(v. v. v) has magnitude w v sin θ. v + w) = n(u. Also. and hence w. So by definition this means that −n(v. v) and is perpendicular to the plane containing w and v. w). w). w) form a right-handed system.

v1 i + v2 j + v3 k) = −n(w1 i. v1 i + v2 j + v3 k) + −n(w2 j. v1 i + v2 j + v3 k) = v1 w2 k − v3 w2 i and −n(w3 j. v × w form a righthanded system. We can use Steps 1 and 2 to evaluate the three terms on the right side of the last equation above: −n(w1 i. w) = n(v1 i + v2 j + v3 k. v1 i + v2 j + v3 k) = −v2 w1 k + v3 w1 j Similarly. w) form a right-handed system. w. v2 j) + −n(w1 i.195 n(v. So since v. i) + −v2 w1 n(i. v3 k) = −v1 w1 n(i. j) + −v3 w1 n(i. ∴ n(v. putting it all together. v1 i + v2 j + v3 k). w1 i + w2 j + w3 k) = n(v1 i + v2 j + v3 k. w. which completes the proof. v1 i + v2 j + v3 k) + −n(w3 k. . w3 k) = −n(w1 i. Thus. w2 j + w3 k) = n(v1 i + v2 j + v3 k. w) = v × w for all vectors v. w2 j) + n(v1 i + v2 j + v3 k. v1 i) + −n(w1 i. we have n(v. then v. n(v. v1 i + v2 j + v3 k) = −v1 w3 j + v2 w3 i . w1 i) + n(v1 i + v2 j + v3 k. w) = −v2 w1 k + v3 w1 j + v1 w2 k − v3 w2 i − v1 w3 j + v2 w3 i = (v2 w3 − v3 w2 )i + (v3 w1 − v1 w3 )j + (v1 w2 − v2 w1 )k = v × w by definition of the cross product. w1 i) + n(v1 i + v2 j + v3 k. w. k) = −v1 w1 (i × i) + −v2 w1 (i × j) + −v3 w1 (i × k) = −v1 w1 0 + −v2 w1 k + −v3 w1 (−j) −n(w1 i. we can calculate −n(w2 j.

This will cause the graph to be plotted for a ≤ x ≤ b and c ≤ y ≤ d.5 0 0. use an expression of the form [a : b][c : d]. for some numbers a < b and c < d.5 0 . To specify an x range and a y range.5 1 -1 -0. <range> is the range of x and y values (and optionally the range of z values) over which to plot. type this at the gnuplot> prompt: splot [−1 : 1][−2 : 2] 2*x**2 + y**2 The result is shown below: 2*(x**2) + y**2 7 6 5 4 3 2 1 0 2 1. y).5 1 -2 -1 -0. To graph the function z = 2x2 + y2 from x = −1 to x = 1 and from y = −2 to y = 2.197 For a function z = f (x.1.5 0. listed below: Symbol + − * / ** exp(x) log(x) sin(x) cos(x) tan(x) Operation Addition Subtraction Multiplication Division Power ex ln x sin x cos x tan x Example 2+3 3−2 2*3 4/2 2**3 exp(2) log(2) sin(pi/2) cos(pi) tan(pi/4) Result 5 1 6 2 23 = 8 e2 ln 2 1 −1 1 Example C. Function definitions use the x and y variables in combination with mathematical operators.5 -1.

1 set xlabel "x" set ylabel "y" set zlabel "z" set contour both set isosamples 25 splot [−1 : 1][−2 : 2] 2*(x**2) + y**2. 120. For clarity. 1. 1. use this command: set contour both The default mesh size for the grid on the surface is 10 units. 120. 25) like this: set isosamples 25 Putting all this together. exp(x+y) By default. put a comma after the first function then append the new function: splot [−1 : 1][−2 : 2] 2*(x**2) + y**2. y) on both the surface and projected onto the xy-plane. use this command before the splot command: set zeroaxis Also. To get more of a colored/shaded surface. To display the axes. increase the mesh size (to. say. parentheses can be used to make sure the operations are being performed in the correct order: splot [−1 : 1][−2 : 2] 2*(x**2) + y**2 In the above example.and y-axes are switched from their usual position. to also plot the function z = e x+y on the same graph. to label the axes. the x-axis and y-axis are not shown in the graph. use this command: set view 60. by default the x. To show the axes with the orientation which we have used throughout the text.198 Appendix C: 3D Graphing with Gnuplot Note that we had to type 2*x**2 to multiply 2 times x2 . 1 Also. use these commands: set xlabel "x" set ylabel "y" set zlabel "z" To show the level curves of the surface z = f (x. we get the following graph with these commands: set zeroaxis set view 60. exp(x+y) .

2.199 25 20 15 z 10 5 0 2*(x**2) + y**2 6 5 4 3 2 1 exp(x+y) 20 15 10 5 -1 -0.v*sin(u).5 -2 -1. y = r sin θ . for a surface parametrized in cylindrical coordinates x = r cos θ .7 (p. That is. the variable v represents r.34 from Section 1. Example C.5 y 0 0. z=z you would do the following: set mapping cylindrical set parametric splot [a : b][c : d] v*cos(u).f(u. For example. with c ≤ v ≤ d. and z = f (u. with a ≤ u ≤ b. 49) was created using the following commands: . y) = c.5 2 1 x The numbers listed below the functions in the key in the upper right corner of the graph are the “levels” of the level curves of the corresponding surface. If you do not want the function key displayed.v) where the variable u represents θ. it can be turned off with this command: unset key PARAMETRIC FUNCTIONS Gnuplot has the ability to graph surfaces given in various parametric forms.5 1 0.5 -1 0 -0. The graph of the helicoid z = θ in Example 1. they are the numbers c such that f (x.5 1. v) is some function of u and v.

”. PRINTING AND SAVING In Windows.v*sin(u). type quit at the gnuplot> command prompt.. and enter png in the Terminal type? textfield. you will see that r varies from 0 to 2. select “Options” and then the “Print. Then. to save the graph as a file called graph.” option in the File menu). and θ varies from 0 to 4π. graph. 120.. to print a graph from Gnuplot right-click on the titlebar of the graph’s window.png in the current directory (usually the directory where wgnuplot.png) in the Output filename? textfield. since the print quality is high and there are many PostScript viewers available. in the File menu again... select “Output Device . as a PNG file. 1 set xyplane 0 set xlabel "x" set ylabel "y" set zlabel "z" unset key set isosamples 15 splot [0 : 4*pi][0 : 2] v*cos(u).. hit OK. though you can change that setting using the “Change Directory .png. In Linux. There are many terminal types (which determine the output format). the postscript terminal type is popular.” option and enter a filename (say. Looking at the graph.u The command set xyplane 0 moves the z-axis so that z = 0 aligns with the xy-plane (which is not the default in Gnuplot). hit OK. 1.. Run the command set terminal to see all the possible types. To save a graph. In Linux. you would issue the following commands: set terminal png set output ’graph.. To quit Gnuplot.exe is located.200 Appendix C: 3D Graphing with Gnuplot set mapping cylindrical set parametric set view 60.png’ and then run your splot command. go to the File menu on the main Gnuplot menubar. Now run your splot command again and you should see a file called graph. select the “Output . .” option. say.

Opaque formats include proprietary formats that can be read and edited only by proprietary word processors. (Thus. commercial. or absence of markup. either copied verbatim. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. "Title Page" . A "Transparent" copy of the Document means a machine-readable copy. A copy made in an otherwise Transparent file format whose markup. the material this License requires to appear in the title page. in the notice that says that the Document is released under this License. The "Cover Texts" are certain short passages of text that are listed. The Document may contain zero Invariant Sections. For works in formats which do not have any title page as such. A copy that is not "Transparent" is called "Opaque". PostScript or PDF produced by some word processors for output purposes only. A Front-Cover Text may be at most 5 words. Examples of transparent image formats include PNG.202 GNU Free Documentation License A "Modified Version" of the Document means any work containing the Document or a portion of it. as Front-Cover Texts or Back-Cover Texts. SGML or XML using a publicly available DTD. or of legal. legibly. and a Back-Cover Text may be at most 25 words. The "Invariant Sections" are certain Secondary Sections whose titles are designated. or with modifications and/or translated into another language. as being those of Invariant Sections. a Secondary Section may not explain any mathematics. Texinfo input format. The "Title Page" means. and the machine-generated HTML. and standard-conforming simple HTML. for a printed book. plus such following pages as are needed to hold. Examples of suitable formats for Transparent copies include plain ASCII without markup. A "Secondary Section" is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document’s overall subject (or to related matters) and contains nothing that could fall directly within that overall subject.) The relationship could be a matter of historical connection with the subject or with related matters. philosophical. LaTeX input format. if the Document is in part a textbook of mathematics. XCF and JPG. PostScript or PDF designed for human modification. If the Document does not identify any Invariant Sections then there are none. SGML or XML for which the DTD and/or processing tools are not generally available. the title page itself. ethical or political position regarding them. that is suitable for revising the document straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor. has been arranged to thwart or discourage subsequent modification by readers is not Transparent. and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. represented in a format whose specification is available to the general public. in the notice that says that the Document is released under this License. An image format is not Transparent if used for any substantial amount of text.

1. For persons making modifications to the book. VERSION: 1. please record the pertinent information here. following the format in the first item below.0 Date: 2008-01-04 Author(s): Michael Corral Title: Vector Calculus Modification(s): Initial version 209 .History This section contains the revision history of the book.

. . . . . . . . . 121 rectangular . . . . . . . . . . . . . . . . . . . . . . . . v. . . . . . . . . . . . . . . .124 ¯ y . . . . . k . . . . . . 71 ∂x . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 area element . . . . . . . . . . . . . . . . . . . . . 105 R C C1 . . . . . . . . . . . 174 closed curve . . . . . . . . . . . . . . . . . . . . . . . . 15 annulus . . . . . . . . . . . . . 167 conservative field . . . . . . . . . . . . . . . 12 A acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 Chain Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 er . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 i. . . . . . . . . . . . . . . . . . 1 Cartesian . . . . . . . . 178 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 C capping surface . . . . . 145 closed surface . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 conical helix . . . . . . . . . . . . . 178 . . . . . . . . . . . . . . . . . . . . . . . . . . . w) ∂f . 105 average value . . . 124 M xy . . . . . . . . . . . 56 Beta function . . . . 1 curvilinear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . y) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . My . . . . . . . . . . . . 126 ¯ δ(x. . . . . . . . . . . 47. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59. . . . 80. . . . . . . . 177 ∇2 . . . . . . 71 Dv f . . 47 cylindrical . 96 continuity . . . . . . . . . . . . . . . . 175 Cauchy-Schwarz Inequality . . . 124 centroid . . . . . . . 164 left-handed .136. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .117. . . . . . . . . . . . . . . . . . . . . . . . y. . . . . . . . . . . . . . . . . . 47. . . . . . . 139 C∞ . . . . . eρ . . . . . . . . 119 ∂(u. . . . . . . . . . . . . M xz . . . . . . . . . . . . . . . . . . 52. . . . . . . . . . . . 84 M x . . . . . . . . . . . . . . . . . . . z) . . . . . . . . . . . . . . . 102 . . . . . 163 Σ C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 3 . . . . . . . . . . . . . . . . . . . j. 181 dr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 ¯ z . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 continuously differentiable . . . . . . . . . . . . . . eφ . . . . . . . . . . . 80 coordinates .Index Symbols D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 ∇ . . . .145 ∂ . . . . . . . . . . . . . . . . 1 . . . ez . . 17 center of mass . . . . . . . . . . . 55 210 . . . . . . . . . . . . . . . . . . . . . . . . . . 124 ∂(x. . . . . . . . . . . . . . . . . . . . 182 ellipsoidal . . . . . . . . . . . Myz . . . . . . . . . 2 polar . . 161 collinear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 ∆ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 constrained critical point . . . . . . . . . . . . . . . . . . . . . . . . 153 arc length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . eθ . . . . . . . . . . . . . 60. . . . . . . . . . . . . . . . . . . . . . . . . .1 x . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 B Bézier curve . . . . . . . . . . . . . . . . . . . . 110 S angle . . . 147 change of variable. . . . . . . . . . . . . . . . . . . . . . . .

43 hypersurface . . . . . . . . . . . . . . . . . 62 cylinder . . . . . 182 coplanar . . 177. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 expected value . . . . . 182 Green’s identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 elliptic cone . . 108 integral . . . . . . . . . . . . . . 144 direction angles . . . 169. . . 26 differential . . 182 Divergence Theorem . . . . . . . . . . . . . . . . . . 43 two sheets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 one sheet .1 continuous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 F flux . 110 I improper integral . . 15 double integral . . . . . . . . . . . . . 102 multiple . . . . . . . . . . . . . . . . . . . . . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 point to plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 H harmonic . . . . . . . . . . . . . . . . . . . . . . 132 extreme point . . . . . . . . . . . 52 determinant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80. . . . . . . . . . . . . . . . . . . . . . . . . . . 83 global minimum . . . . . . . . . . . 130 divergence . . . . . . . . . . . . . . . . . . . . . . . 162. . . . . . . . . . . . . . . . . . . . . . . . . . 134 covariance. . 44 hyperboloid . . . . . . . . . . . . . . . . . . 2 spherical . . . . .Index right-handed . . . . . . . . . . 110 hypervolume . . . . . . . 162 dot product . . . 139 directed curve . 131 normal . . . . . . . . . . . . . . . 186 Green’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 distance . . . . . . . . . . . . . . . . . . 69 scalar. . . . . . 178. . . . . . . . . . 105 improper . . . . . . . . . . . . . . . . . . . . . . 1 exact differential form . . . . . . . . . . . . . . . . . . . . . 83 gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 between points . . . . . . . . 124 derivative . . . . . . . . . . . 7 from point to line . . . . . . . . . . . . . . 45 . . . . . . 156. . . . . 123. . . . .