MATH 275: Calculus III

Lecture Notes by Angel V. Kumchev

 Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Lecture 1. Three-Dimensional Coordinate Systems . . . . . . . . . . . . . . . . . . . . . 1
Lecture 2. Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Lecture 3. Lines and Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Lecture 4. Vector Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Lecture 5. Space Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Lecture 6. Multivariable Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Lecture 7. Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Lecture 8. Partial Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Lecture 9. Directional Derivatives and Gradients . . . . . . . . . . . . . . . . . . . . . . . 73
Lecture 10. Tangent Planes and Normal Vectors . . . . . . . . . . . . . . . . . . . . . . . 77
Lecture 11. Extremal Values of Multivariable Functions . . . . . . . . . . . . . . . . . . . 87
Lecture 12. Lagrange Multipliers* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Lecture 13. Double Integrals over Rectangles . . . . . . . . . . . . . . . . . . . . . . . . 101
Lecture 14. Double Integrals over General Regions . . . . . . . . . . . . . . . . . . . . . 107
Lecture 15. Double Integrals in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . 117
Lecture 16. Triple Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Lecture 17. Triple Integrals in Cylindrical and Spherical Coordinates . . . . . . . . . . . . 129
Lecture 18. Applications of Double and Triple Integrals . . . . . . . . . . . . . . . . . . . 135
Lecture 19. Change of Variables in Multiple Integrals* . . . . . . . . . . . . . . . . . . . 145
Lecture 20. Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Lecture 21. Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Lecture 22. The Fundamental Theorem for Line Integrals . . . . . . . . . . . . . . . . . . 169
Lecture 23. Green’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
i
Lecture 24. Surface Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Lecture 25. Stokes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Lecture 26. Gauss’ Divergence Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Appendix A. Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Appendix B. Definite Integrals of Single-Variable Functions . . . . . . . . . . . . . . . . 207
Appendix C. Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Appendix D. Introduction to Mathematica . . . . . . . . . . . . . . . . . . . . . . . . . . 215
Appendix E. Answers to Selected Exercises . . . . . . . . . . . . . . . . . . . . . . . . . 233

ii
 Preface

These lecture notes originated from a set of class notes that I used to suplement my lectures
for M408D Sequences, Series, and Multivariable Calculus at UT-Austin, and later for MATH 275
here, at TU. About 2006, the TU mathematics department set a goal to include in MATH 275 some
topics that had previously been excluded “for lack of class time.” I found that difficult to achieve
within the constraints imposed by the organization of the standard calculus text that we were using
at the time, yet quite manageable if I moved some material around and kept the applications more
focused on geometry. Those changes were significant enough to diminish the value of the textbook
as a primary reference for the course—it became a (rather expensive) source of homework assign-
ments and pretty pictures. That motivated me to add some exercises and illustrations to the class
notes and to turn them in a self-contained resource—these lecture notes.
I have not written these notes in vacuum, so there are several people whose help and suggestions
I gratefully acknowledge. The earliest version of my UT-Austin class notes benefited greatly from
numerous exchanges with Michael Tehranchi and Vrej Zarikian, and my good friend and colleague
Alexei Kolesnikov helped me with the addition of some new material to the current version. My
colleagues John Chollet and Tatyana Sorokina have used these notes in their own teaching and in
the process have not only helped me find various typos, but also have provided me with insights
regarding possible improvements—as have many of their and my own former students.
The present version of these notes differs from earlier versions in two main ways. First, I added
a couple of new topics: the method of Lagrange multipliers (Lecture #12) and the general theory
of change of variables in multiple integrals (Lecture #19). While those topics are not officially a
part of the MATH 275 curriculum, I was convinced over time that they are too important not to be
mentioned. The second major addition to this version is the inclusion of more visual illustrations
and of several Mathematica tutorials. Among the latter, I want to mention specifically Appendix D,
which is a version of the introduction to Mathematica that is part of the computer laboratories used
in calculus courses at TU. To that end, I thank my colleagues Raouf Boules, Geoffrey Goodson,
Ohoe Kim and Michael O’Leary for allowing me to use their work.

iii
 LECTURE 1

Three-Dimensional Coordinate Systems

You are familiar with coordinate systems in the plane from single-variable calculus. The first
coordinate system in the plane that you have seen is most likely the Cartesian system, which
consists of two perpendicular axes. These axes are usually positioned so that one of them, called
the x-axis, is horizontal and oriented from left to right and the other, called the y-axis, is vertical
and oriented upward (see Figure 1.1). The intersection point of the two axes is called the origin
and is commonly denoted by O.

x
O

Figure 1.1. Two-dimensional Cartesian coordinate system

1.1. Cartesian coordinates in space

The three-dimensional Cartesian coordinate system consists of a point O in space, called the
origin, and of three mutually perpendicular axes through O, called the coordinate axes and labeled
the x-axis, the y-axis, and the z-axis. The labeling of the axes is usually done according to the
right-hand rule: if one places one’s right hand in space so that the origin lies on one’s palm, the
y-axis points toward one’s arm, and the x-axis points toward one’s curled fingers, then the z-axis
points toward one’s thumb.
There are several standards for drawing two-dimensional models of three-dimensional objects.
In particular, there are three somewhat standard ways to draw a three-dimensional coordinate sys-
tem. These are shown on Figure 1.2. Model (c) is the one commonly used in manual sketches.
Model (a) is a version of model (c) that is commonly used in textbooks and will be the one we
shall use in these notes; unlike model (c), it is based on real mathematics that transforms three-
dimensional coordinates to two-dimensional ones. Model (b) is the default model preferred by
computer algebra systems such as Mathematica.
The coordinate axes determine several notable objects in space. First, we have the coordinate
planes: these are the unique planes determined by the three different pairs of coordinate axes. The
coordinate planes determined by the x- and y-axes, by the x- and z-axes, and by the y- and z-axes
are called the xy-plane, the xz-plane, and the yz-plane, respectively. The three coordinate planes
divide space into eight “equal” parts, called octants; the octants are similar to the four quadrants
1
 z z z

y y
x
x x
(a) (b) (c)

Figure 1.2. Three-dimensional Cartesian coordinate systems

that the two coordinate axes divide the xy-plane into. The octants are labeled I through VIII, so
that octants I through IV lie above the respective quadrants in the xy-plane and octants V through
VIII lie below those quadrants.
Next, we define the Cartesian coordinates of a point P in space. We start by finding its orthog-
onal projections onto the yz-, the xz-, and the xy-planes. Let us denote those points by P x , Py , and
Pz , respectively (see Figure 1.3). Let a be the directed distance from P x to P: that is,
(
|PP x | if the direction from P x to P is the same as that of the x-axis,
a=
−|PP x | if the direction from P x to P is opposite to that of the x-axis.

Further, let b be the directed distance from Py to P, and let c be the directed distance from Pz to P.
The three-dimensional Cartesian coordinates of P are (a, b, c) and we write P(a, b, c).

Example 1.1. Plot the points A(4, 3, 0), B(2, −2, 2), and C(2, 2, −2).

Example 1.2. Determine the sets of points described by the equations: z = 1; x = −2; and
y = 3, x = −1.

Answer. The equation z = 1 represents a plane parallel to the xy-plane. The equation x = −2
represents a plane parallel to the yz-plane. The equations y = 3, x = −1 represent a vertical line
perpendicular to the xy-plane and passing through the point (3, −1, 0). 

Py b a Px
P
c
y
x Pz

Figure 1.3. Cartesian coordinates of a point in space

2
 1.2. Distance in space
Let P1 = (x1 , y1 , z1 ) and P2 = (x2 , y2 , z2 ) be two points in space. Then the distance between P1
and P2 is given by p
|P1 P2 | = (x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 . (1.1)
Example 1.3. The distance between the points A(4, 3, 0) and B(2, −2, 2) is
p √
|AB| = (4 − 2)2 + (3 + 2)2 + (0 − 2)2 = 33.

Example 1.4. Find the equation of the sphere of radius 3 centered at C(1, −2, 0).
Solution. The sphere of radius 3 centered at C(1, −2, 0) is the set of points P(x, y, z) such that
p
|PC| = (x − 1)2 + (y + 2)2 + (z − 0)2 = 3,
that is, the points P(x, y, z), whose coordinates satisfy the equation
(x − 1)2 + (y + 2)2 + z2 = 9. 
By generalizing this example, we obtain the following proposition.
Proposition 1.1. The sphere with center C(a, b, c) and radius r > 0 has an equation
(x − a)2 + (y − b)2 + (z − c)2 = r2 . (1.2)
Conversely, the set of points (x, y, z) satisfying (1.2) is a sphere in space with center (a, b, c) and
radius r.
Example 1.5. The equation
x2 + y2 + z2 − 4x + 6y + z = 7
describes a sphere. Find its center and radius.
Solution. We can rewrite the above equation as
x2 − 4x + 4 + y2 + 6y + 9 + z2 + z + 1
4
=7+4+9+ 1
4
(x − 2)2 + (y + 3)2 + (z + 21 )2 = 81
4
.
The last equation is of the form (1.2) with a = 2, b = −3, c = − 12 , and r = 92 . Hence, the given
equation describes a sphere with center (2, −3, − 12 ) and radius 92 . 

1.3. Cylindrical coordinates

You should be familiar with the polar coordinate system in the plane from single-variable
calculus (see also Appendix C). In the remainder of this lecture, we define two sets of coordinates
in space that extend the polar coordinates in the plane to the three-dimensional setting. We start
with the so-called “cylindrical coordinates”. The basic idea of the cylindrical coordinate system
is to replace the Cartesian coordinates x, y in the xy-plane by the corresponding polar coordinates
and to keep the Cartesian z-coordinate.
Let P be a point in space with Cartesian coordinates (x, y, z). The cylindrical coordinates of
P are the triple (r, θ, z), where (r, θ) are the polar coordinates of the point P0 (x, y) in the xy-plane
(see Figure 1.4). In other words, we pass between Cartesian and cylindrical coordinates in space
3
 z

P(x, y, z)

z
y
x θ r
P0 (x, y)

Figure 1.4. Cylindrical coordinates of a point in space

by passing between Cartesian and polar coordinates in the xy-plane and keeping the z-coordinate
unchanged. The formulas for conversion from cylindrical to Cartesian coordinates are
x = r cos θ, y = r sin θ, z = z, (1.3)
and those for conversion from Cartesian to cylindrical coordinates are
p
r = x2 + y2 , θ = arctan(y/x), z = z. (1.4)
The formula θ = arctan(y/x) in (1.4) comes with a caveat: unless the point (x, y) lies in the first
quadrant of the xy-plane, this formula yields a reference angle which needs to be adjusted to obtain
the actual value of θ. (This is similar to the conversion from Cartesian to polar coordinates in the
plane: see §C.2 in Appendix C).
Example 1.6. Find the Cartesian coordinates of the points P(2, π, e) and Q(4, arctan(−2), 2.7).
Solution. The Cartesian coordinates of P are (x, y, e), where
x = 2 cos π = −2, y = 2 sin π = 0.
That is, P(−2, 0, e). The Cartesian coordinates of Q are (x, y, 2.7), where
4 8
x = 4 cos(arctan(−2)) = √ , y = 4 sin(arctan(−2)) = − √ .
5 5
That is, Q( √5 , − √5 , 2.7).
4 8

Example 1.7. Find the cylindrical coordinates of the points P(3, 4, 5) and Q(−2, 2, −1).
Solution. The cylindrical coordinates of P are (r, θ, 5), where
√ 4
r = 32 + 42 = 5, tan θ = ,
3
and θ lies in the first quadrant. Hence, θ = arctan( 3 ) and the cylindrical coordinates of P are
4

(5, arctan( 34 ), 5). The cylindrical coordinates of Q are (r, θ, −1), where
p √
r = 22 + (−2)2 = 2 2, tan θ = −1,
and θ lies in the second quadrant. Hence, θ = 3π
and the cylindrical coordinates of P are
√ 4
(2 2, 3π
4
, −1). 

4
 z y

x
y ( 21 , 0)
x

Figure 1.5. The cylinder x2 + y2 = x and its cross-section

Example 1.8. Find the Cartesian equation of the surface r = cos θ. What kind of surface is
this?
Solution. We have
r = cos θ ⇐⇒ r2 = r cos θ ⇐⇒ x2 + y2 = x.
The last equation is the equation of a circle of radius 1
2
centered at the point ( 12 , 0) in the xy-plane:
x2 + y2 = x ⇐⇒ x2 − x + 14 + y2 = 1
4
⇐⇒ (x − 12 )2 + y2 = 14 .
Here, however, we use the equation x2 + y2 = x to describe a surface in space, so we view it as an
equation in x, y, z. Since z does not appear in the equation, if the x- and y-coordinates of a point
satisfy the equation x2 + y2 = x, then the point is on the surface. In other words, if a2 + b2 = a, then
all points (a, b, z), whose x-coordinate is a and y-coordinate is b, are on the surface. We also know
that all the points (x, y, 0), with x2 + y2 = x, are on the surface: these are the points on the circle
in the xy-plane with the same equation. Putting these two observations together, we conclude that
the given surface is the cylinder that consists of all the vertical lines passing through the points of
the circle x2 + y2 = x in the xy-plane (see Figure 1.5). 

1.4. Spherical coordinates

We now introduce the so-called “spherical coordinates” in space. In order to define this set
of coordinates, we choose a special point O, the origin, a plane that contains O, and a pair of
mutually perpendicular axes passing through O—one that lies in the chosen plane and another that
is orthogonal to the chosen plane. Since we want to relate the spherical coordinates of a point
to its Cartesian coordinates, it is common to choose O to be the origin O(0, 0, 0) of the Cartesian
coordinate system, the special plane to be the xy-plane, and the two axes to be the x-axis and the
z-axis. With these choices, the spherical coordinates of a point P(x, y, z) in space are (ρ, θ, φ),
where ρ = |OP|, θ is the angular polar coordinate of the point P0 (x, y) in the xy-plane, and φ is the
angle between OP and the positive direction of the z-axis (see Figure 1.6); it is common to restrict
φ to the range 0 ≤ φ ≤ π.
The angles θ and φ are often called the longitude and the latitude of P, because with the right
choice of the spherical coordinate system, the angles θ and 90◦ − φ for a point on the surface of
the Earth are its geographic longitude and latitude.
5
 z

P(x, y, z)
ρ
z = ρ cos φ
y
x θ
r = ρ sin φ P0 (x, y)

Figure 1.6. Spherical coordinates of a point in space

Consider a point P(x, y, z) in space and let P0 (x, y, 0). Also, let (ρ, θ, φ) be the spherical coor-
dinates of P. From the right triangle 4OPP0 , we find that
z = ±|P0 P| = |OP| cos φ = ρ cos φ, |OP0 | = |OP| sin φ = ρ sin φ.
On the other hand, using polar coordinates in the xy-plane, we find that
x = |OP0 | cos θ = ρ sin φ cos θ, y = |OP0 | sin θ = ρ sin φ sin θ.
Thus, the formulas for conversion from spherical to Cartesian coordinates are
x = ρ sin φ cos θ, y = ρ sin φ sin θ, z = ρ cos φ. (1.5)
The formulas for conversion from Cartesian to spherical coordinates are
p
ρ = x2 + y2 + z2 , θ = arctan(y/x), φ = arccos(z/ρ), (1.6)
where the formula for θ comes with some strings attached (the same as in the cases of polar and
cylindrical coordinates).
Example 1.9. Find the Cartesian coordinates of the point P(2, π/3, 3π/4).
Solution. The Cartesian coordinates of P are
√
1 3 √
x = 2 sin(3π/4) cos(π/3) = √ , y = 2 sin(3π/4) sin(π/3) = √ , z = 2 cos(3π/4) = − 2.
2 2

√
Example 1.10. Find the spherical coordinates of the points P(1, −1, − 6) and Q(−2, 2, −1).
Solution. The spherical coordinates of P are
√ √ √ √
√
5π
ρ = 1 + 1 + 6 = 8, φ = arccos − 6/ 8 = arccos − 3/2 = ,
6
and the solution θ of
tan θ = −1, 3π/2 < θ < 2π.
√
Hence, θ = 7π/4 and the spherical coordinates of P are ( 8, 7π/4, 5π/6).
The spherical coordinates of Q are
√
ρ = 4 + 4 + 1 = 3, φ = arccos(−1/3),
and the solution θ of
tan θ = −1, π/2 < θ < π.
6
Hence, θ = 3π/4 and the spherical coordinates of Q are (3, 3π/4, arccos(−1/3)). 
Example 1.11. Find the Cartesian equation of the surface ρ2 (sin2 φ − 2 cos2 φ) = 1.
Solution. Since
x2 + y2 = ρ2 sin2 φ cos2 θ + ρ2 sin2 φ sin2 θ = ρ2 sin2 φ,
we have
ρ2 sin2 φ − 2ρ2 cos2 φ = 1 ⇐⇒ x2 + y2 − 2z2 = 1.
The last equation represents a surface known as “one-sheet hyperboloid”; it is shown on Figure 1.7.


z
z

y
x y
x

Figure 1.7. The hyperboloid x2 + y2 − 2z2 = 1 Figure 1.8. The region 2 ≤ ρ ≤ 3, π/2 ≤ φ ≤ π

Example 1.12. Describe the solid defined by the inequalities

2 ≤ ρ ≤ 3, π/2 ≤ φ ≤ π.
Solution. The second inequality restricts the points of the solid to the lower half-space. The
inequality ρ ≤ 3 describes the points in space that are within distance of 3 from the origin, that is,
the interior of the sphere x2 + y2 + z2 = 9. Similarly the inequality ρ ≥ 2 describes the exterior of
the sphere x2 + y2 + z2 = 4. Thus, the given solid is obtained from the lower half of the interior of
the sphere x2 + y2 + z2 = 9 by removing from it the interior of the smaller sphere x2 + y2 + z2 = 4
(see Figure 1.8). 
Example 1.13. Find the cylindrical and spherical equations of the sphere x2 + y2 + z2 + 4z = 0.
Solution. By (1.4), x2 + y2 = r2 , so the cylindrical form of the given equation is
x2 + y2 + z2 + 4z = 0 ⇐⇒ r2 + z2 + 4z = 0.
By (1.6), x2 + y2 + z2 = ρ2 , so the spherical form of the given equation is
ρ2 + 4ρ cos φ = 0 ⇐⇒ ρ = −4 cos φ. 
7
 Exercises
1.1. Which of the points P(1, 3, 2), Q(0, −2, −1), and R(5, 3, −3): is closest to the xy-plane; lies in the xz-plane; lies
in the yz-plane; is farthest from the origin?
1.2. Describe the set in R3 represented by the equation x + 2y = 2.
1.3. Find the side lengths of the triangle with vertices A(0, 6, 2), B(3, 4, 1) and C(1, 3, 4). Is this a special kind of a
triangle: equilateral, isosceles, right?
1.4. Find an equation of the sphere that passes through P(4, 2, 1) and has center C(1, 1, 1).
1.5. Show that the equation x2 + y2 + z2 = x + 2y + 4z represents a sphere and find its center and radius.
Describe verbally the set in R3 represented by the given equation or inequality.
1.6. x = −3 1.8. −1 ≤ y ≤ 2 1.10. 4 < x2 + y2 + z2 < 9
1.7. z ≤ 2 1.9. (x − 1) + y + z ≥ 4
2 2 2
1.11. xy = 0

1.12. Write inequalities to describe the half-space consisting of all points to the right of the plane y = −2.
1.13. Write inequalities to describe the interior of the upper hemisphere of radius 2 centered at the origin.
Find the Cartesian coordinates of the point with the given cylindrical coordinates.
1.14. (2, 0, 3) 1.15. (3, π/4, −3) 1.16. (3, 7π/6, 2) 1.17. (3, 3π/2, −1)

Find the cylindrical coordinates of the point with the given Cartesian coordinates.
√
1.18. (1, −1, 4) 1.19. (−2, 2, 2) 1.20. ( 3, 1, −2) 1.21. (3, 3, 2)

Find the Cartesian coordinates of the point with the given spherical coordinates.
1.22. (1, 0, 0) 1.23. (3, π/4, 3π/4) 1.24. (3, 7π/6, π/2) 1.25. (2, π/2, π/4)

Find the spherical coordinates of the point with the given Cartesian coordinates.
√ √ √
1.26. (1, 1, 2) 1.27. (−1, 1, − 2) 1.28. ( 3, 1, 0) 1.29. ( 2, 1, −1)

Find the cylindrical and spherical equations of the surface represented by the given Cartesian equation.
1.30. x2 + y2 = 2z 1.31. x2 + 2x + y2 + z2 = 1 1.32. z = x2 + y2
p

Find the Cartesian equation of the surface represented by the given cylindrical or spherical equation. If possible,
describe the surface verbally.
1.33. ρ cos φ = −2 1.35. r2 + z2 = 4 1.37. ρ sin2 φ − 2 cos φ = 0
1.34. r = 2 sin θ 1.36. ρ = 2 sin φ sin θ 1.38. r cos θ − 2r sin θ = 3

Describe verbally the set in R3 represented by the given conditions.

1.39. θ = π/4 1.41. r = 1, 0 ≤ z ≤ 1 1.43. ρ ≤ 1, −π/2 ≤ θ ≤ π/2
1.40. ρ = 2 1.42. 2 ≤ ρ ≤ 4, 0 ≤ φ ≤ π/2 1.44. ρ sin φ ≤ 1, π/2 ≤ φ ≤ π

8
 LECTURE 2

Vectors

2.1. Geometric vectors

A vector is a quantity (such as velocity, acceleration, force, impulse, etc.) that has both mag-
nitude and direction. In mathematics, we consider two types of vectors: geometric and algebraic.
A geometric vector is a directed line segment, represented visually as an arrow: the length of the
arrow represents the magnitude of the vector and the arrow points in the direction of the vector.
We shall denote vectors by small boldface letters: a, b, . . . , z. If v is a vector represented by an
−−→
arrow with initial point A and terminal point B, we may also write v = AB; this is consistent with
an alternative notation for vectors: ~a, ~b, . . . ,~z.
We say that two vectors u and v are equal, and write u = v, if u and v have equal magnitudes
(lengths) and the same direction. The vector of zero length is called the zero vector and is denoted
0; it is the only vector with no specific direction.

2.2. Vector addition

−−→ −−→
Given two vectors u = AB and v = BC (represented so that the terminal point of the first vector
is the initial point of the second), we define their sum u + v by (see Figure 2.1(a))
−−→ −−→ −−→
u + v = AB + BC = AC.
−−→ −−→
When u = AB and v = CD are two vectors in general position, with C , B, we first represent v by
−−−→ −−−→
an equal arrow BD0 and then apply the above definition: u + v = AD0 , as shown on Figure 2.1(b).

C D0
D
v v
v
v

+
+

v
u
u

B B
C u
u
A A
(a) (b)

Figure 2.1. Vector addition

9
 2.3. Multiplication by scalars
Let v be a vector and let c be a real number (called also a scalar). The scalar multiple cv is the
vector whose length is |c| times the length of v and whose direction is the same as the direction of
v when c > 0, or opposite to the direction of v when c < 0. If c = 0 or v = 0, then cv = 0.
The vector (−1)v has the same length as v but the opposite direction. We call this vector the
negative (or opposite) of v and denote it by −v. For two vectors u and v, we define their difference
u − v by
u − v = u + (−v).
We visualize the sum and the difference of two vectors u and v by drawing them so that they
share the same initial point. We then consider the parallelogram they define. The sum u + v is
represented by the diagonal of the parallelogram that passes through the common initial point of u
and v and the difference u − v is represented by the other diagonal (see Figure 2.2).

u−
v

v
+
u

v v
u u
(a) (b)
Figure 2.2. The sum and the difference of u and v

2.4. Coordinate representation of vectors. Algebraic vectors

Often, it is convenient to introduce a coordinate system and to work with vectors algebraically
−−→
instead of geometrically. If we represent a (geometric) vector a by the arrow OA, where O is
the origin of a Cartesian coordinate system, the terminal point A of a has coordinates (a1 , a2 ) or
(a1 , a2 , a3 ), depending on the dimension. We call these coordinates the coordinates or the compo-
nents of the algebraic vector a and write
a = ha1 , a2 i or a = ha1 , a2 , a3 i.
−−→
We will use the same algebraic vector to represent all the geometric vectors equal to OA. For
example, if A(1, −1), B(−1, 3), and C(0, 2), then
−−→ −−→
BC = OA = h1, −1i.
Thus, for any two points A(a1 , a2 , a3 ) and B(b1 , b2 , b3 ), the algebraic vector v representing the arrow
−−→
AB is
v = hb1 − a1 , b2 − a2 , b3 − a3 i.
From now on, we shall state all the formulas for three-dimensional vectors. One usually obtains
the respective two-dimensional definition or result simply by dropping the last component of the
vectors; any exceptions to this rule will be mentioned explicitly when they occur.
10
 2.5. Operations with algebraic vectors
Note that when we write a vector a algebraically, the distance formula provides a convenient
tool for computing the length (also called the norm) kak of that vector: if a = ha1 , a2 , a3 i, then

1/2
kak = a21 + a22 + a23 . (2.1)

Furthermore, if a = ha1 , a2 , a3 i and b = hb1 , b2 , b3 i are vectors and c is a scalar, we have the
following coordinate formulas for a + b, a − b, and ca:

a + b = ha1 + b1 , a2 + b2 , a3 + b3 i, (2.2)
a − b = ha1 − b1 , a2 − b2 , a3 − b3 i, (2.3)
ca = hca1 , ca2 , ca3 i. (2.4)

In other words, we add vectors, subtract vectors, and multiply vectors by scalars by simply per-
forming the respective operations componentwise.

Example 2.1. Consider the vectors a = h1, −2, 3i and b = h0, 1, −2i. Then

2a + b = h2, −4, 6i + h0, 1, −2i = h2, −3, 4i ,

p √
k2a + bk = 22 + (−3)2 + 42 = 29.

The algebraic operations with vectors have several properties that resemble familiar properties
of the algebraic operations with numbers.

Theorem 2.1. Let a, b, and c denote vectors and let c, c1 , c2 denote scalars. Then:
i) a+b=b+a
ii) a + (b + c) = (a + b) + c
iii) a+0=a
iv) a + (−a) = 0
v) c(a + b) = ca + cb
vi) (c1 + c2 )a = c1 a + c2 a
vii) (c1 c2 )a = c1 (c2 a)
viii) 1a = a
ix) kak ≥ 0, with kak = 0 if and only if a = 0
x) kcak = |c| · kak
xi) ka + bk ≤ kak + kbk (the triangle inequality)

Proof. The proofs of parts i)–viii) are all similar: they use (2.2)–(2.4) to deduce i)–viii) from
the respective properties of numbers. For example, to prove the distributive law vi), we write
a = ha1 , a2 , a3 i and argue as follows:

(c1 + c2 )a = h(c1 + c2 )a1 , (c1 + c2 )a2 , (c1 + c2 )a3 i by (2.4)

= hc1 a1 + c2 a1 , c1 a2 + c2 a2 , c1 a3 + c2 a3 i by the distributive law for numbers
= hc1 a1 , c1 a2 , c1 a3 i + hc2 a1 , c2 a2 , c2 a3 i by (2.2)
= c1 a + c2 a by (2.3).
11
The proofs of ix) and x) are also easy. For instance,

1/2
kcak = (ca1 )2 + (ca2 )2 + (ca3 )2


1/2 √
1/2
= c2 a21 + a22 + a23 = c2 a21 + a22 + a23 = |c| · kak.
The proof of xi) is slightly more involved, so we leave it to the exercises. 

2.6. Unit vectors and the standard basis

The simplest nonzero vectors are:
e1 = h1, 0, 0i , e2 = h0, 1, 0i , e3 = h0, 0, 1i .
These are called the standard basis vectors. Sometimes, especially in physics and engineering,
the vectors e1 , e2 , and e3 are denoted by i, j, and k, respectively, but we shall stick with the above
notation, because it is more common in mathematics (it is also more consistent with generalizations
to higher dimensions).
Because of (2.2)–(2.4), it is always possible to express a given vector as a combination of the
three standard basis vectors. For example,
h1, 2, 4i = h1, 0, 0i + h0, 2, 0i + h0, 0, 4i = h1, 0, 0i + 2 h0, 1, 0i + 4 h0, 0, 1i = e1 + 2e2 + 4e3 .
More generally, if a = ha1 , a2 , a3 i, we have
a = ha1 , 0, 0i + h0, a2 , 0i + h0, 0, a3 i = a1 e1 + a2 e2 + a3 e3 .
Notice that e1 , e2 , and e3 have length 1. In general, vectors of length 1 are called unit vectors.
We remark that if a is any nonzero vector (i.e., a , 0), then there is a unique unit vector that points
in the same direction as a—namely, the vector
1 a
u= a= .
kak kak
Indeed, if c = kak−1 , then a and u = ca point in the same direction (since c > 0) and
kuk = kcak = |c| · kak = kak−1 · kak = 1.
This simple computation is known as normalization, and u is called the normalization of a.
Example 2.2. Consider the vector a = h2, 1, 2i. Its length is
√
kak = 22 + 12 + 22 = 3,
so the normalization of a is the vector
= , 1, 2 .
1

2 
3
a 3 3 3

2.7. The dot product

We can now add two vectors and we can multiply a vector by a scalar. Is there a way to multiply
two vectors? The first way that comes to mind is to multiply the vectors componentwise:
ab = ha1 , a2 , a3 i hb1 , b2 , b3 i = ha1 b1 , a2 b2 , a3 b3 i ,
but this way of multiplying two vectors turns out to be of little practical use. Instead, through the
remainder of this lecture, we shall talk about two meaningful ways to multiply vectors: the dot
product and the cross product.
12
 Definition. Given vectors a = ha1 , a2 , a3 i and b = hb1 , b2 , b3 i, we define their dot product a · b
to be
a · b = a1 b1 + a2 b2 + a3 b3 . (2.5)
Notice that the dot product of two vectors is a scalar, not a vector. Sometimes, the dot product
is called also the inner product or the scalar product of a and b. Because of this, you should avoid
the temptation to refer to the scalar multiple ca as the “scalar product of c and a”.
Example 2.3. The dot product of a = h4, 2, −1i and b = h−2, 2, 1i is
a · b = (4)(−2) + (2)(2) + (−1)(1) = −5.
The dot product of u = 2e1 + 3e2 + e3 and v = e1 + 4e2 is
u · v = (2)(1) + (3)(4) + (1)(0) = 14.
The next theorem summarizes some important algebraic properties of the dot product.
Theorem 2.2. Let a, b, and c denote vectors and let c1 , c2 denote scalars. Then:
i) a · a = kak2
ii) a · 0 = 0
iii) a · b = b · a
iv) (c1 a) · (c2 b) = c1 c2 (a · b)
v) (a + b) · c = a · c + b · c
Proof. These follow from the definition of the dot product and from the properties of the real
numbers. For example, if a = ha1 , a2 , a3 i and b = hb1 , b2 , b3 i, we have
a · b = a1 b1 + a2 b2 + a3 b3 by (2.5)
= b1 a1 + b2 a2 + b3 a3 by the commutative law for real numbers
=b·a by (2.5).
That is, iii) holds. 
We said earlier that our definition of the dot product is motivated by applications. The next
theorem relates the dot product of two vectors to their geometric representations.
Theorem 2.3. If a and b are two vectors, then
a · b = kak kbk cos θ,
where θ is the angle between a and b.
Corollary 2.4. If a and b are two nonzero vectors and θ is the angle between them, then
a·b
cos θ = .
kak kbk
Example 2.4. Find the angle between a = h1, 2, −1i and b = h0, 2, 1i.
Solution. The angle θ between a and b satisfies
(1)(0) + (2)(2) + (−1)(1) 3 3
cos θ = p √ = √ √ = √ ,
12 + 22 + (−1)2 02 + 22 + 12 6 5 30
so θ = arccos √330 = 0.9911 . . . .



13
 u

a − proju a
a

proju a

Figure 2.3. The projection of a vector onto a unit vector

We say that two vectors a and b are orthogonal (or perpendicular) if a · b = 0. By the corollary,
a · b = 0 implies that the angle between a and b is 90◦ , so this definition makes perfect sense.

Note that if a = ha1 , a2 , a3 i, then we can use the dot products of a and the standard basis vectors
e j to extract the components of a:
a1 = a · e1 , a2 = a · e2 , a3 = a · e3 .
More generally, suppose that u is a unit vector. Then a · u is a scalar and
proju a = (a · u)u
is a vector that is parallel to u. This vector is called the (orthogonal) projection of a onto u. It has
the property that the difference between it and the original vector a is perpendicular to u:
a − proju a · u = 0.

The two-dimensional case of this is illustrated on Figure 2.3. Note that the size of the dot product
a·u equals the length of the projection; the sign of that dot product indicates whether the projection
and u have the same direction (positive sign) or opposite directions (negative sign).

2.8. The cross product

The dot product of two vectors is a scalar. In this section, we define another product of two
vectors: their “cross product”, which is a vector. Since our discussion requires basic familiarity
with 2 × 2 and 3 × 3 determinants, a quick look at Appendix A may be helpful before proceeding
further.
Definition. Let a = ha1 , a2 , a3 i and b = hb1 , b2 , b3 i be two three-dimensional vectors. Then
their cross product a × b is the vector
     
a2 a3  a1 a3  a1 a2 
a × b =  e −  e +  e .
b2 b3  1 b1 b3  2 b1 b2  3
Remarks. There are a couple of things we should point out:
1. The cross product is defined only for three-dimensional vectors. There is no analog of
this operation for two-dimensional vectors.
2. The defining formula may be confusing at first. On the other hand, following the exact
labeling is important because of the properties of the cross product (see Theorem 2.5
14
 below). Another representation of the cross product that is much easier to remember is
 
e1 e2 e3 
a × b = a1 a2 a3  .
 
(2.6)
b1 b2 b3 
This expression is not exactly a determinant in the sense described in Appendix A (since
its first row consists of vectors instead of numbers), but if we expand this “determinant”
as if we were dealing with the determinant of a regular 3 × 3 matrix, we get exactly the
cross product of a and b.
The following theorem summarizes the algebraic properties of the cross product.
Theorem 2.5. Let a, b, and c denote three-dimensional vectors and let c, c1 , c2 denote scalars.
Then
i) b × a = −(a × b)
ii) (c1 a) × (c2 b) = c1 c2 (a × b)
iii) a × (b + c) = a × b + a × c
iv) (a + b) × c = a × c + b × c
v) a × a = 0 and a × 0 = 0
vi) a × (b × c) = (a × b) × c may fail.
Example 2.5. We have
 
e1 e2 e3       
 0 0  1 0  1 0 
e1 × e2 =  1 0 0  =   e1 − 0 0 e2 + 0 1 e3 = e3 ,

       
0 1 0 1 0

and similarly,
e2 × e3 = e1 , e3 × e1 = e2 .
By property i), we also have
e2 × e1 = −e3 , e1 × e3 = −e2 , e3 × e2 = −e1 .
Note that
e1 × (e2 × e1 ) = e1 × (−e3 ) = −(e1 × e3 ) = e2
and
(e1 × e2 ) × e1 = e3 × e1 = e2 ,
so e1 × (e2 × e1 ) = (e1 × e2 ) × e1 . On the other hand,
e1 × (e1 × e2 ) = e1 × e3 = −e2 ,
whereas by property v),
(e1 × e1 ) × e2 = 0 × e2 = 0.
Thus, e1 × (e1 × e2 ) , (e1 × e1 ) × e2 ; this illustrates property vi) above. 

15
 Example 2.6. Compute the cross product of a = h3, −2, 1i and b = h1, −1, 1i.
Solution. By (2.6),
 
e1 e2 e3 
a × b =  3 −2 1 
 
 1 −1 1 
     
−2 1 3 1  3 −2
=  e −  e +  e
−1 1 1 1 1 2 1 −1 3
= −e1 − 2e2 − e3 = h−1, −2, −1i . 
Second solution. We can also argue by using the algebraic properties in Theorem 2.5 and the
values of the cross products of e1 , e2 , and e3 :
(3e1 − 2e2 + e3 ) × (e1 − e2 + e3 )
= 3e1 × e1 − 2e2 × e1 + e3 × e1 − 3e1 × e2 + 2e2 × e2 − e3 × e2 + 3e1 × e3 − 2e2 × e3 + e3 × e3
= 0 + 2e3 + e2 − 3e3 + 0 + e1 − 3e2 − 2e1 + 0 = −e1 − 2e2 − e3 . 
Finally, we discuss the geometric meaning of the cross product.
Theorem 2.6. Let a = ha1 , a2 , a3 i and b = hb1 , b2 , b3 i be three-dimensional vectors. Then:
i) a × b is perpendicular to both a and b. The direction of a × b is determined by the
right-hand rule.
ii) ka × bk = kak kbk | sin θ|, where θ is the angle between a and b. That is, the length of a × b
is equal to the area of the parallelogram determined by a and b.
iii) a × b = 0 if and only if a and b are parallel.
iv) The volume V of the parallelepiped determined by three vectors a, b, and c which are not
coplanar (that is, they do not lie in one plane) is given by the formula
V = a · (b × c).
 

Furthermore, three vectors a, b, and c are coplanar if and only if a · (b × c) = 0.

Example 2.7. Find two unit vectors perpendicular to a = h1, 2, 3i and b = h2, 1, 1i.
Solution. We know from part i) of Theorem 2.6 that a × b is orthogonal to both a and b. We
find that
 
e1 e2 e3       
 2 3  1 3  1 2 
a × b =  1 2 3  =  2 1 e2 + 2 1 e3 = −e1 + 5e2 − 3e3 .

 e1 −    
2 1 1 1 1 

Hence, p √
ka × bk = (−1)2 + 52 + (−3)2 = 35,
and two unit vectors orthogonal to a and b are the normalization of a × b and its opposite:
   
a×b −1 5 −3 1 −5 3
u= = √ , √ , √ and −u= √ , √ , √ .
ka × bk 35 35 35 35 35 35


16
 Example 2.8. Find the area of the triangle with vertices P(1, 2, 3), Q(−1, 3, 2), and R(3, −1, 2).
Solution. The area of 4PQR is one half of the area of the parallelogram determined by the
−−→ −−→
vectors PQ = h−2, 1, −1i and PR = h2, −3, −1i. We know from part ii) of Theorem 2.6 that the
area of the parallelogram is the length of
 
 e1 e2 e3       
−−→ −−→    1 −1 −2 −1 −2 1 
PQ × PR = −2 1 −1 =   
 e1 −  2 −1 e2 +  2 −3 e3 = −4e1 − 4e2 + 4e3 .
    
 2 −3 −1 −3 −1
√ √
Therefore, the area of the triangle is 12 16 + 16 + 16 = 2 3. 
2.9. Vectors in Mathematica
2.9.1. Representation. We represent a vector in Mathematica as a list of numbers (the com-
ponents of the vector). For example, to define a vector a = ha1 , a2 , a3 i, we write
a = {a1, a2, a3}
To refer to a particular component of a vector (or a list), we use double square brackets. For exam-
ple, the second component of the vector a defined above can be accessed as a[[2]]. Although we
are only interested in two- and three-dimensional vectors here, we should note that Mathematica
can handle vectors of higher dimensions just as easily.
On the geometric side, Mathematica has a built-in command Arrow, which produces a geo-
metric vector in two or three dimensions. In three dimensions, its basic syntax is
Arrow[{{x1, y1, z1}, {x2, y2, z2}}]
This results in an arrow with initial point (x1 , y1 , z1 ) and terminal point (x2 , y2 , z2 ). (See Mathe-
matica’s help for various formatting options and more advanced usage.) To get a two-dimensional
vector, we simply use two-dimensional points as arguments:
Arrow[{{x1, y1}, {x2, y2}}]
For technical reasons, the Arrow command produces a “graphics primitive” instead of an actual
plot of the geometric vector. For example, if we try to execute the command
Arrow[{{1,2}, {4,-1}}]
Mathematica will simply repeat it in response. To see the result of the above command, we need
to “wrap” it in a Graphics command like this:
Graphics[Arrow[{{1,2}, {4,-1}}]]
Similarly, to display a three-dimensional geometric vector, we use a Graphics3D command. Here
is an example:
Graphics3D[Arrow[{{1,2,0}, {1,-1,2}}]]
2.9.2. Vector addition and scalar multiplication. Vector addition, vector subtraction and
scalar multiplication are denoted in the usual way: if c is a scalar and a and b are two vectors (of
equal dimensions), then

a+b denotes the vector sum a + b

a-b denotes the vector difference a − b
c*a, c a denote the scalar multiple ca

17
 For example, let us define the vectors a = h2, 7, −3i and b = h−3, 0, 5i.
a = {2,7,-3}
b = {-3,0,5}
We can calculate 2a, a − b, and a + 3b as follows:
2a
a-b
a+3*b

{4,14,-6}
{5,7,-8}
{-7,7,12}
2.9.3. The dot and cross products. Mathematica has built-in commands for calculating the
dot and cross products: Dot and Cross, respectively. These commands also have infix operator
forms. A period between vectors performs a dot product, and a × symbol between vectors performs
a cross product. To enter the symbol ×, first hit the Esc key, type “cross”, and then hit again the
Esc key. As an example, let us use Mathematica to compute the dot and cross products of two
generic three-dimensional vectors:
a = {a1, a2, a3}
b = {b1, b2, b3}
Here are the two versions of the dot product:
a.b
Dot[a,b]

a1 b1 + a2 b2 + a3 b3
a1 b1 + a2 b2 + a3 b3
And one of the versions of the cross product (see the Mathematica notebook file for the other
version):
Cross[a,b]

{-a3 b2 + a2 b3, a3 b1 - a1 b3, -a2 b1 + a1 b2}

2.9.4. Graphical representation revisited. Using the Arrow command can be cumbersome
when only basic drawing functionality is needed. In such situations, it is much simpler to use a
user-defined command that “wraps around” Arrow and makes it more user-friendly (but cuts some
of its functionality). The command vector defined below presents one such option.
vector[a_, b_, rgb_:{0,0,0}] := Which[
Length[a] == Length[b] == 2,
Graphics[{RGBColor[rgb[[1]],rgb[[2]],rgb[[3]]], Arrow[{a,b}]}],
Length[a] == Length[b] == 3,
Graphics3D[{RGBColor[rgb[[1]],rgb[[2]],rgb[[3]]], Arrow[{a,b}]}],
True, Print["Bad dimensions!"]
]

18
 The command vector can be called in two ways. In its more basic form,
vector[{x1, y1}, {x2, y2}]
vector[{x1, y1, z1}, {x2, y2, z2}]
the vector command returns a graphical representation of a two- or three-dimensional arrow with
the given initial and terminal points (similarly to Arrow). For example, try
vector[{1,2}, {-1,0}]
vector[{1,1,0}, {1,2,-1}]
When the dimensions of the two inputs do not match, or if they match but are neither 2 nor 3, the
vector command prints out the error message “Bad dimensions!”:
vector[{1,2}, {1,2,-1}]

Bad dimensions!
The vector command allows the user to include a third, optional argument in the form of a pre-
defined color or a three-dimensional vector to be interpreted as an RGB color-specification. For
example, the following four commands will render their outputs in red, red, blue, and teal, respec-
tively:
vector[{1,2}, {-1,0}, Red]
vector[{1,2}, {-1,0}, {1,0,0}]
vector[{1,2}, {-1,0}, {0,0,1}]
vector[{1,2}, {-1,0}, {0,0.5,0.5}]

Exercises
−−→ −−→
2.1. Let O(0, 0), A(1, 2), B(3, 0), and M(−2, 1) be points in R2 . Draw the geometric vectors v = AB and w = MO and
use them to draw the following vectors v + 2w, w − v, and w + 21 v.
Find kbk, a − 3b, ka + 12 bk, and 2a + b − 3e2 for the given vectors a and b.
2.2. a = h1, −2i , b = h2, 0i 2.4. a = h2, 0, −1i , b = h−1, 1, 1i
2.3. a = 3e1 + 2e2 , b = −e1 + e2 2.5. a = e2 + 21 e3 , b = −2e1 + e2 + e3

2.6. Find a vector b that has the same direction as a = h1, −1i and length kbk = 3.
Find a · b for the given vectors a and b.
2.7. a = h3, −2, 1i , b = h1, 0, 2i 2.9. a = 3e1 + 2e2 + 4e3 , b = −e1 + 2e2 + 21 e3
2.8. a = h2, 0, −1i , b = h2, 1, 4i 2.10. kak = 2, kbk = 3, ^(a, b) = 60◦

Find the angle between the given vectors.

2.11. a = h2, 0, −1i , b
= h2,
√ 1, 4i 2.13. a = 3e1 + 2e2 , b = −e1 + e3
2.12. a = h2, 0i , b = 1, 3 2.14. a = e2 + e3 , b = e1 + 2e2 − 3e3

2.15. Find the three angles of the triangle with vertices A(1, 0, 0), B(2, 1, 0), C(0, 1, 2).
Determine whether the given vectors are parallel, perpendicular, or neither.

√ √ 
2.16. a = h2, 1, −1i , b = 12 , 1, 2 2.18. a = h1, 2, 3i , b = 1, 2, 3



2.17. a = h1, −2, 2i , b = h−2, 4, −4i

2.19. Find the projection of a = h1, −1, 4i onto u = 3, 3, −3

1 2 2

.
2.20. Find the projection of a = h2, −1, 2i onto the direction of b = h2, −1, 0i.

19
Find a × b for the given vectors a and b.
2.21. a = h2, −3, 1i , b = h1, 1, 2i 2.23. a = e1 + 2e2 , b = 3e2 + e3
2.22. a = 3e1 + 2e2 + 2e3 , b = −e1 + 2e2 + e3 2.24. a = h1, −1, 2i , b = h1, 0, −3i

Let a = h−1, 2, −4i, b = h7, 3, −4i, and c = h−2, 1, 0i. Evaluate the given expression or explain why it is undefined.
2.25. a · (2b + c) 2.27. b × (c × a) 2.29. k2ck (a × (a + 3b))
2.26. kb × (2a)k 2.28. c · (−2a × b) 2.30. a × (c · b)

2.31. Find a unit vector u orthogonal to both x = h−2, 3, −6i and y = h2, 2, −1i.
(a) by using the cross product; (b) without using the cross product.

2.32. Find the area of the triangle with vertices A(3, 2, 1), B(2, 4, 5), and C(3, 1, 4).
2.33. Find the volume of the parallelepiped with adjacent edges OA, OB, and OC, where A(2, 3, 0), B(3, −1, 2), and
C(1, 0, −4).
2.34. Use the method from the second solution of Example 2.6 to find r × s, where
r = cos φ cos θe1 + cos φ sin θe2 − sin φe3 , s = − sin φ sin θe1 + sin φ cos θe2 .
2.35. Use the method from the second solution of Example 2.6 to find u × v, where
u = e1 + a cos φe2 + a sin φe3 , v = −b sin φe2 + b cos φe3 .
2.36. Let r = hx, y, zi and a = ha, b, ci. Describe the set of points P(x, y, z) such that kr − ak = R.
2.37. Prove the Cauchy–Schwarz inequality: If a, b are three-dimensional vectors, then
|a · b| ≤ kak kbk.
2.38. Prove the triangle inequality: If a, b are three-dimensional vectors, then
ka + bk ≤ kak + kbk.
2.39. Are the following statements true or false?
(a) If a , 0 and a · b = a · c, then it follows that b = c.
(b) If a , 0 and a × b = a × c, then it follows that b = c.
(c) If a , 0, a · b = a · c, and a × b = a × c, then it follows that b = c.

20
 LECTURE 3

Lines and Planes

3.1. Equations of a straight line in space

In our discussion of lines and planes, we shall use the notion of a position vector. A position
vector is a vector whose initial point is at the origin. Thus, the components of a position vector are
the same as the coordinates of its terminal point.
We shall specify a line in space using a point on the line and the direction of the line. If r0 is
the position vector of a point P0 (x0 , y0 , z0 ) on the line and if v is a vector parallel to the line, then
the terminal point of a position vector r is on the line if and only if
r = r0 + tv for some t ∈ R. (3.1)
We call (3.1) a parametric vector equation of the line ` through P0 and parallel to v. In this
equation, we construct a generic point P on ` in two steps: first we get on ` by going to the given
point P0 ; then we slide along ` by a multiple of the given vector v until we reach P (see Figure 3.1).
The value of the parameter t indicates the distance along ` (measured in units equal to kvk) that we
need to move to reach P: for example, t = 0 corresponds to P = P0 and t = −2 corresponds to a
point that is at a distance 2kvk from P0 in the direction opposite to v.

P `
tv
P0
r
r0

Figure 3.1. The parametric vector equation of a line

Writing v and r in component form, v = ha, b, ci and r = hx, y, zi, we can express (3.1) in terms
of the components of the vectors involved:
hx, y, zi = hx0 , y0 , z0 i + t ha, b, ci = hx0 + ta, y0 + tb, z0 + tci ,
or equivalently:
x = x0 + ta, y = y0 + tb, z = z0 + tc (t ∈ R). (3.2)
These are the parametric scalar equations of `. In general, the vector v = ha, b, ci is determined
up to multiplication by a nonzero scalar, that is, we can use any nonzero scalar multiple of v in
place of v. This will change the coefficients in equations (3.1) and (3.2), but not the underlying set
21
of points. We shall refer to the coordinates a, b, c of any vector v parallel to ` as direction numbers
of `.
Another common way to specify a line ` in space are its symmetric equations. Given the scalar
equations (3.2), with abc , 0, we can eliminate t to obtain the equations
x − x0 y − y0 z − z0
= = . (3.3)
a b c
These are the symmetric equations of `. If one of a, b, or c is zero—say, b = 0, we obtain the
symmetric equations of ` by keeping that scalar equation and eliminating t from the other two:
x − x0 z − z0
y = y0 , = .
a c
If two among a, b, c are equal to zero, the corresponding scalar equations represent the symmetric
equations of ` after discarding the third parametric equation: if a = b = 0, the symmetric equations
of ` are
x = x0 , y = y0 .
Finally, we can describe a line by the coordinates of any two distinct points P1 (x1 , y1 , z1 ) and
−−−→
P2 (x2 , y2 , z2 ) on the line. If P1 and P2 are such points, the vector P1 P2 = hx2 − x1 , y2 − y1 , z2 − z1 i
is parallel to the line, so we can write the symmetric equations (3.3) in the form
x − x1 y − y1 z − z1
= = , (3.4)
x2 − x1 y2 − y1 z2 − z1
(or something like it, for other types of symmetric equations).
Example 3.1. Find the equations of the line ` through the points P(1, 0, 3) and Q(2, −1, 4).
−−→
Solution. To obtain the symmetric equations, we note that PQ = h1, −1, 1i is a vector parallel
to `, so (3.3) can be written as
x−1 y−0 z−3
= = ⇐⇒ x − 1 = −y = z − 3.
1 −1 1
−−→
The parametric scalar equations (based again on PQ) are
x = 1 + t, y = 0 + t(−1) = −t, z=3+t (t ∈ R),
and the parametric vector equation is
r(t) = h1, 0, 3i + t h1, −1, 1i = h1 + t, −t, 3 + ti (t ∈ R). 
Example 3.2. Find a parametric vector equation of the line ` passing through P(−1, 2, −3) and
parallel to the line 2(x + 1) = 4(y − 3) = z.
Solution. We can write the equations of the second line as
x+1 y−3
1 = 1 = z,
2 4

so v = , 1, 1
is a vector parallel to the second line. Thus, v is also parallel to `. That is, ` passes

1 
2 4
through P and is parallel to v. Its vector equation then is
r(t) = h−1, 2, −3i + t 21 , 14 , 1 = −1 + 12 t, 2 + 41 t, −3 + t



(t ∈ R). 
22
 Example 3.3. Find the point of intersection and the angle between the lines with the vector
parametrizations
r1 (t) = h1 + t, −1 − t, −4 + 2ti and r2 (u) = h1 − u, 1 + 3u, 2ui .
Solution. If it exists, the point of intersection of the given lines will have coordinates (x, y, z)
such that
x = 1 + t = 1 − u, y = −1 − t = 1 + 3u, z = −4 + 2t = 2u,
for some t, u. From the equations for x, we find that t = −u. Hence, the other two equations
between t and u become
−1 + u = 1 + 3u, −4 − 2u = 2u =⇒ u = −1, t = −u = 1.
Therefore, the point of intersection is (2, −2, −2).
To find the angle between the lines, we find the angle θ between their parallel vectors: v =
h1, −1, 2i for the first line and u = h−1, 3, 2i for the second. Thus, by Corollary 2.4,
u·v −1 − 3 + 4
cos θ = = √ √ =0 =⇒ θ = 90◦ .
kuk kvk 6 14


3.2. Equations of a plane

We shall specify a plane in space using a point on the plane and a direction perpendicular to the
plane. If r0 is the position vector of a point P0 (x0 , y0 , z0 ) on the plane and n = ha, b, ci is a vector
perpendicular (also called normal) to the plane, then a generic point P(x, y, z) is on the plane if and
−−→
only if P0 P is perpendicular to n (see Fig. 3.2). Introducing the position vector r of P, we can
express this condition as
−−→
P0 P · n = 0 ⇐⇒ (r − r0 ) · n = 0, (3.5)
or equivalently,
n · r = n · r0 . (3.6)
These are the vector equations of the given plane.

P0 n
r − r0
P

O y
x

Figure 3.2. The vector equation of a plane

If we express the dot product in (3.5) in terms of the components of the vectors, we get
a(x − x0 ) + b(y − y0 ) + c(z − z0 ) = 0, (3.7)
23
or the scalar equation of the given plane. Given the last equation, we can easily obtain the plane’s
linear equation:
ax + by + cz + d = 0, (3.8)
simply by writing d = −ax0 − by0 − cz0 .
Example 3.4. Find the equation of the plane that contains P(1, −2, 3) and is perpendicular to
e2 + 2e3 .
Solution. The scalar equation of this plane is
0(x − 1) + 1(y + 2) + 2(z − 3) = 0 ⇐⇒ y + 2z − 4 = 0. 
Example 3.5. Find the equation of the plane π that contains P(3, −1, 5) and is parallel to the
plane with equation 4x + 2y − 7z + 5 = 0.
Solution. A normal vector for the given plane is n = h4, 2, −7i. Hence, the scalar equation of
π, which also has n as its normal vector, is
4(x − 3) + 2(y + 1) − 7(z − 5) = 0 ⇐⇒ 4x + 2y − 7z + 25 = 0. 
Example 3.6. Find the equation of the plane that contains the points P(1, 2, 3), Q(−1, 3, 2), and
R(3, −1, 2).
−−→ −−→
Solution. The cross product PQ × PR is a normal vector for this plane. By Example 2.8,
−−→ −−→
PQ × PR = −4e1 − 4e2 + 4e3 ,
so we can use the normal vector n = h1, 1, −1i. We obtain the scalar equation
(x − 1) + (y − 2) − (z − 3) = 0 ⇐⇒ x + y − z = 0. 
Example 3.7. Find the angle between the planes 2x − y + 3z = 5 and 5x + 5y − z = 1.
Solution. The angle between the two planes equals the angle between their normal vectors:
n1 = h2, −1, 3i and n2 = h5, 5, −1i, respectively. If θ is the angle between n1 and n2 , we use the
dot product to find
n1 · n2 10 − 5 − 3 2
cos θ = = √ √ = √ .
kn1 k kn2 k 14 51 714

Example 3.8. Find the line of intersection of the planes x + y + z + 1 = 0 and x − y + z + 2 = 0.
Solution. We shall compute the symmetric equations of the given line. First, by adding the
two equations, we obtain the equation
z + 3/2
2x + 2z + 3 = 0 ⇐⇒ x = −z − 3/2 ⇐⇒ x = .
−1
On the other hand, subtracting the two given equations, we get
2y − 1 = 0 ⇐⇒ y = 1/2.
Hence, the symmetric equations of the line of intersection are
z + 3/2
x= , y = 1/2.
−1
24
Note that these symmetric equations lead to the parametric scalar equations
x = t, y = 1/2, z = −3/2 − t (t ∈ R). 
Example 3.9. Find the distance from P(3, −5, 2) to the plane x − 2y + z = 5.
First solution. The distance from P to the given plane is equal to the distance |PQ|, where Q
is the intersection point of the plane and the line through P that is perpendicular to the plane. The
normal vector to the plane n = h1, −2, 1i is parallel to the line PQ, so the symmetric equations of
that line are
x−3 y+5 z−2
= = .
1 −2 1
Combining these equations with the equation of the plane, we obtain a linear system
y+5 y+5
x−3= , z−2= , x − 2y + z = 5,
−2 −2
whose solution ( 43 , − 53 , 31 ) is the point Q. Therefore,

1/2 √
|PQ| = (− 53 )2 + ( 103 )2 + (− 53 )2 = 5
3
6. 

We now give a second solution, which is somewhat trickier, but has the advantage that can be
easily generalized (see the next example).

Second solution. Let Q(u, v, w) be the same point as before. In this solution, we avoid the
explicit computation of u, v, w.
−−→ −−→
We have PQ = hu − 3, v + 5, w − 2i. By the definition of Q, we know that PQ is normal to the
−−→
given plane, so PQ is parallel to the normal vector n = h1, −2, 1i. Hence, by the properties of the
dot product,
 −−→ −→  √
n · PQ = knk |PQ|  cos ^(n, −
PQ) = 6 |PQ|,


since the angle between the two vectors is 0◦ or 180◦ . On the other hand, using the definition of
the dot product, we have
 −−→
n · PQ = |(u − 3) − 2(v + 5) + (w − 2)| = |u − 2v + w − 15| = |5 − 15| = 10.

Here, we have used that u − 2v + w = 5, because Q lies on the given plane. Comparing the two
 −−→ √
expressions for n · PQ, we conclude that |PQ| = 10/ 6. 
Example 3.10. Find the distance from P(x0 , y0 , z0 ) to the plane ax + by + cz + d = 0.
Solution*. We essentially repeat the second solution of the previous example. Let Q(u, v, w)
be the intersection point of the plane and the line through P that is perpendicular to the plane. Then
−−→ −−→
PQ = hu − x0 , v − y0 , w − z0 i. Since PQ is normal to the given plane, it is parallel to its normal
vector n = ha, b, ci. Hence, by the properties of the dot product,
 −−→
n · PQ = knk |PQ|  cos ^(n, −−→  √
PQ) = a2 + b2 + c2 |PQ|,


25
since the angle between the two vectors is 0◦ or 180◦ . On the other hand, using the definition of
the dot product, we have
 −−→
n · PQ = |a(u − x0 ) + b(v − y0 ) + c(w − z0 )|
= |au + bv + cw − ax0 − by0 − cz0 |
= | − d − ax0 − by0 − cz0 | = |ax0 + by0 + cz0 + d|.
Here, we have used that au + bv + cw = −d, because Q lies on the given plane. Comparing the two
 −−→
expressions for n · PQ, we conclude that the desired distance is
|ax0 + by0 + cz0 + d|
|PQ| = √ .
a2 + b2 + c2


Exercises
3.1. Find the equations of the line that passes through the point (1, −2, 3) and is perpendicular to the plane with
equation −2x + y − z + 4 = 0.
3.2. Find the equations of the line that passes through the points (0, −2, 3) and (1, 1, 5).
3.3. Let `1 be the line with equations x = −1 + 2t, y = 3t, z = 2 and let `2 be the line through the points (4, 3, 2) and
(6, −1, 4). Determine whether these two lines are parallel, intersecting, or skew. If they intersect, find their intersection
point.
x−3 y+1
3.4. Find the intersection points of the line with equations = = z and the coordinate planes.
2 5
3.5. Find the angle between the lines whose respective equations are
x+1 y+2 z−2 x+1 y+2 z−2
= = and = = .
−3 2 5 2 1 −2
3.6. Explain why the two lines from Exercise 3.5 intersect.
3.7. Find an equation of the plane that passes through the point (1, −2, 3) and is perpendicular to the vector h3, 0, −2i.
x−1 y+2 z−2
3.8. Find an equation of the plane that contains the line with equations = = and is parallel to the
2 3 4
plane 3x − 2y = 5.
x−1 y+2 z
3.9. Find an equation of the plane that contains the line with equations = = and is perpendicular to the
2 −1 4
line x = 1 − 2t, y = 4t, z = −3 + 2t.
x−1 y+2 z
3.10. Find an equation of the plane that contains the line with equations = = and is perpendicular to
2 3 4
the line x = 1 − 2t, y = 3t, z = −3 + 2t.
3.11. Find an equation of the plane through the points (1, −1, 0), (2, 0, 3), and (−2, 2, 1).
3.12. Find an equation of the plane through the points (1, 2, −1), (−1, 0, 1), and (2, 3, 0).
3.13. Find the equations of the intersection line of the planes x + y − z = 2 and 2x − 3y + z = 6.
3.14. Find an equation of the plane that contains both the line `1 with equations x = 1 + 2t, y = −1 + 3t, z = −2t and
the line `2 with equations x = −2 − 3t, y = t, z = 4 + 4t.
3.15. Find an equation of the plane that contains both the line `1 with equations x = 1 − 2t, y = 2 + t, z = −1 − t and
the line `2 with equations x = 1 + t, y = 2 − 3t, z = −1.

26
Find the angle between the given planes.
3.16. 2x − y + 3z = 0, x + 2y + z = 4 3.17. x + y + z = 4, x + 2y − 3z = 6

3.18. The planes 2x − 2y − z = 1 and 6x − 6y − 3z = −5 are parallel. Find the distance between them.

27
 LECTURE 4

Vector Functions

4.1. Definition
A vector-valued function (or a vector function) is a function whose domain is a set of real
numbers and whose range is a set of vectors. In general, a vector function has the form
f(t) = h f1 (t), f2 (t), f3 (t)i = f1 (t)e1 + f2 (t)e2 + f3 (t)e3 .
where f1 , f2 , and f3 are functions in the regular sense (functions for which both the inputs and out-
puts are real numbers). The functions f1 , f2 , and f3 are sometimes called the component functions
of the vector function f. When we want to distinguish that the values of a function f are numbers
(as opposed to vectors), we will refer to f as a scalar function.
Example 4.1. As we know from Lecture #3, the parametric equation of the line in space passing
through (1, 0, −5) and parallel to the vector h1, 2, 1i is, in fact, a vector function:
r(t) = h1 + t, 2t, −5 + ti (t ∈ R).
The line is the set of values of this function, not the graph (the graph is four-dimensional).
Here is another example of a vector function:
f(t) = cos t, (t − 5)−1 , ln t ;



its domain is (0, 5) ∪ (5, ∞). 

4.2. Calculus of vector functions: Limits

Vector functions can have limits, derivatives, antiderivatives, and definite integrals—just like
scalar functions. As we will see, all of these can be computed componentwise. Because of that,
once one is proficient in the calculus of scalar-valued functions, one can easily master the calculus
of vector functions. We begin with limits and continuity. The limit of a vector function is a vector.
Definition. If f(t) = h f1 (t), f2 (t), f3 (t)i is defined near a, then the limit of f as t → a is defined
by D E
lim f(t) = lim f1 (t), lim f2 (t), lim f3 (t) ,
t→a t→a t→a t→a
provided that the limits of all three component functions exist.
Definition. A vector function f(t) = h f1 (t), f2 (t), f3 (t)i, defined near a, is continuous at a if
lim f(t) = f(a).
t→a

Example 4.2. We have

 D E

lim cos t, (t − 5)−1 , ln t = lim cos t, lim(t − 5)−1 , lim ln t = cos 7, 12 , ln 7 .


t→7 t→7 t→7 t→7

29
 4.3. Calculus of vector functions: Derivatives
Definition. If f(t) = h f1 (t), f2 (t), f3 (t)i is continuous at a, then its derivative f 0 (a) at a is defined
by
f(t) − f(a) f(a + h) − f(a)
f 0 (a) = lim = lim ,
t−a
t→a h→0 h
provided that the limit on the right exists. If f 0 (a) exists, we say that f is differentiable at a.
Theorem 4.1. A vector function f(t) = h f1 (t), f2 (t), f3 (t)i is differentiable at a if and only if all
three component functions f1 , f2 , f3 are. When f 0 (a) exists, we have
f 0 (a) = f10 (a), f20 (a), f30 (a) .



In other words, the derivative of a vector function can be computed using componentwise differen-
tiation.
Proof. We have
f(t) − f(a)
f 0 (a) = lim
t→a t−a
h f1 (t) − f1 (a), f2 (t) − f2 (a), f3 (t) − f3 (a)i
= lim
t→a
 t−a 
f1 (t) − f1 (a) f2 (t) − f2 (a) f3 (t) − f3 (a)
= lim , ,
t→a t−a t−a t−a
= f1 (a), f2 (a), f3 (a) .

0 0 0


Example 4.3. If f(t) = cos t, (t − 5)−1 , ln t , then



f 0 (t) = − sin t, −(t − 5)−2 , t−1 .



Theorem 4.2. Let f and g be differentiable vector functions, let u be a differentiable scalar
function, and let c be a scalar. Then the following formulas hold:
0
i) f(t) + g(t) = f 0 (t) + g0 (t);

0
ii) cf(t) = cf 0 (t);

0
iii) u(t)f(t) = u(t)f 0 (t) + u0 (t)f(t);

0
iv) f(t) · g(t) = f(t) · g0 (t) + f 0 (t) · g(t);

0
v) f(t) × g(t) = f(t) × g0 (t) + f 0 (t) × g(t);

0
vi) f(u(t)) = f 0 (u(t))u0 (t).


Proof. We shall prove iv). If f(t) = h f1 (t), f2 (t), f3 (t)i and g(t) = hg1 (t), g2 (t), g3 (t)i, then
f(t) · g(t) = f1 (t)g1 (t) + f2 (t)g2 (t) + f3 (t)g3 (t).
Hence, the left side of iv) equals
0
0
f(t) · g(t) = f1 (t)g1 (t) + f2 (t)g2 (t) + f3 (t)g3 (t)


= f10 (t)g1 (t) + f1 (t)g01 (t) + f20 (t)g2 (t) + f2 (t)g02 (t) + f30 (t)g3 (t) + f3 (t)g03 (t).
On the other hand, we have
f 0 (t) · g(t) = f10 (t), f20 (t), f30 (t) · hg1 (t), g2 (t), g3 (t)i = f10 (t)g1 (t) + f20 (t)g2 (t) + f30 (t)g3 (t),



f(t) · g0 (t) = h f1 (t), f2 (t), f3 (t)i · g01 (t), g02 (t), g03 (t) = f1 (t)g01 (t) + f2 (t)g02 (t) + f3 (t)g03 (t),



30
so the right side of iv) equals
f10 (t)g1 (t) + f20 (t)g2 (t) + f30 (t)g3 (t) + f1 (t)g01 (t) + f2 (t)g02 (t) + f3 (t)g03 (t).
This establishes iv). 

4.4. Calculus of vector functions: Integrals

The definition of the definite integral of a continuous function via Riemann sums (see Appen-
dix B) can also be extended to vector functions. Here, we will adopt an equivalent definition, which
is a little less transparent but considerably easier to use.
Rb
Definition. If f(t) = h f1 (t), f2 (t), f3 (t)i is continuous in [a, b], then its definite integral a f(t) dt
is defined by
Z b Z b Z b Z b 
f(t) dt = f1 (t) dt, f2 (t) dt, f3 (t) dt .
a a a a
Rb
Remark. Note that f(t) dt is a vector (not a scalar and not a vector function).
a

Example 4.4. If f(t) = cos t, (t − 5)−1 , ln t , then



Z 2 Z 2 Z 2 Z 2 
f(t) dt = cos t dt, −1
(t − 5) dt, ln t dt
1 1 1 1
D 2  2  2 E
= sin t 1 , ln |t − 5| 1 , t ln t − t 1
= sin 2 − sin 1, ln( 34 ), 2 ln 2 − 1 .



Theorem 4.3. Let f and g be continuous vector functions, let u be a vector, and let c be a scalar.
Then the following formulas hold:
Rb Rb Rb
i) a f(t) + g(t) dt = a f(t) dt + a g(t) dt;

Rb Rb
ii) a cf(t) dt = c a f(t) dt;

Rb Rb
iii) a u · f(t) dt = u · a f(t) dt;

Rb Rb
iv) a u × f(t) dt = u × a f(t) dt;

Rb
v) a f 0 (t) dt = f(b) − f(a).
Proof. We shall prove iv). Both sides of the identity are three-dimensional vectors, so it suffices
to show that their respective components are equal. If u = hu1 , u2 , u3 i and f(t) = h f1 (t), f2 (t), f3 (t)i,
then
u × f(t) = hu2 f3 (t) − u3 f2 (t), u3 f1 (t) − u1 f3 (t), u1 f2 (t) − u2 f1 (t)i .
Thus, the first component of the left side of iv) is
Z b Z b Z b
u2 f3 (t) − u3 f2 (t) dt = u2


f3 (t) dt − u3 f2 (t) dt.
a a a

Since the right side of this identity is also the first component of the right side of iv), this shows that
the first components of the two sides of iv) match. The proof that the remaining two components
of the two sides of iv) match is similar. 
31
 4.5. Vector functions in Mathematica
Defining and operating with vector functions in Mathematica is no different

from defining and
operating with scalar functions. For example, to define the function f(t) = cos t, (t − 5) , ln t
−1

from Examples 4.2–4.4, we simply use the command

f[t_] := {Cos[t], 1/(t-5), Log[t]}
We can then use the commands Limit, D, Integrate and NIntegrate on the vector function
f[t] just as we use them on scalar-valued functions:
In[1]:= Limit[f[t], t->1]
Out[1]:= {Cos[1], -(1/4), 0}

In[2]:= f’[t]
Out[2]:= {-Sin[t], -(1/(-5+t)ˆ2), 1/t}

In[3]:= D[f[2t], t]
Out[3]:= {-2Sin[2t], -(2/(-5+2t)ˆ2), 1/t}

In[4]:= NIntegrate[f[t], {t,1,4}]

Out[4]:= {-1.59827, -1.38629, 2.54518}

In[5]:= Integrate[f[t], {t,0,1}]

Out[5]:= {Sin[1], -Log[5/4], -1}
Exercises
, sint 2t , e , where: (a) a = 1; (b) a = 0.

ln(t+1) t

4.1. Evaluate lim t
t→a

Find the derivative of the given vector function.

4.2. f(t) = ln t2 + 1 , arcsin 2t, et 4.4. f(t) = t√


 3
e1 − 3e2t e3
2 2
4.3. f(t) = 5t + 2t, et sin t, et cos t 4.5. f(t) = t2 + 1e1 − cos(2t)e2 + sin(2t)e3

3 

Evaluate the given integral.

R1
−1
−1
R1
3
4.6. 0 t2 + 4 e1 + t t2 − 2 e3 dt −1 5t + 2t, e , sin t cos t dt
2t 4

4.7.

4.8. Find f(t), if f 0 (t) = 2te1 + et e2 + (ln t) e3 and f(1) = e1 + e2 .

32
 LECTURE 5

Space Curves

Vector functions are a great tool for dealing with curves, both in the plane and in space. In this
lecture, we use vector functions to give a unified treatment of plane curves and space curves.

5.1. Review of plane curves

Recall that a curve γ in the plane is the set
γ = (x, y) : x = x(t), y = y(t), t ∈ I ,


where I is some interval in R (finite or infinite; open, closed, or semiopen) and x(t) and y(t) are
continuous functions of t. We refer to the pair of equations
x = x(t), y = y(t) (t ∈ I) (5.1)
as parametric equations or a parametrization of γ. A curve γ has many parametrizations, and some-
times replacing one parametrization by another can change a particular property of the curve, even
if the set of points stays the same. Thus, when we want to distinguish a particular parametrization,
we talk about the parametric curve γ with equations (5.1).

5.2. Definition of space curve

Next, we extend the above definitions to three dimensions.
Definition. If x(t), y(t), z(t) are continuous functions defined on an interval I, then the set
γ = (x, y, z) : x = x(t), y = y(t), z = z(t), t ∈ I


is called a space curve. The equations

x = x(t), y = y(t), z = z(t) (t ∈ I) (5.2)
are called parametric equations or a parametrization of γ; t is the parameter.
Alternatively, we can think of the curve γ with equations (5.2) as the range of the vector func-
tion
r(t) = x(t)e1 + y(t)e2 + z(t)e3 (t ∈ I). (5.3)
We shall refer to (5.3) as the parametric (vector) equation of γ. Note that the plane curve (5.1) can
be treated as a special case of a space curve given by the parametric equation
r(t) = x(t)e1 + y(t)e2 + 0e3 (t ∈ I).
A good way to visualize the connection between a space curve γ and its parametrization r(t) is to
think of the vector r(t) as a position vector which moves as t varies. The terminal point of r(t) then
traces the points on the curve (see Figure 5.1(a), or execute the Mathematica code on Figure 5.1(b)
for an interactive demonstration).
33
 z
γ f[t_] := {t*Cos[t], t*Sin[t], t-1}
Manipulate[
r(t) Show[{
ParametricPlot3D[f[t], {t, 0, 2Pi}],
Graphics3D[{Red, Arrow[{{0,0,0}, f[t]}]}]
}],
O y {t, 0, 2Pi}]
x
(a) (b)

Figure 5.1. Vector functions and curves in space

Example 5.1. Consider the space curve γ with parametric equation

r(t) = (cos t)e1 + (sin t)e2 + te3 (0 ≤ t < ∞).
The point (cos t, sin t, 0) traces the unit circle x + y = 1 in the xy-plane in the counterclockwise
2 2

direction, so the point (cos t, sin t, t) on γ moves in the positive direction of the z-axis along a spiral
lying above that circle. This curve is known as the helix; it is shown on Figure 5.2. 

z z

O y O y
x x

Figure 5.2. The helix Figure 5.3. r(t) = ht cos t, t sin t, ti

Example 5.2. Consider the space curve γ with parametric equation

r(t) = (t cos t)e1 + (t sin t)e2 + te3 (0 ≤ t < ∞).
The initial point of the curve is the origin. For any given t, we have
(t cos t)2 + (t sin t)2 = t2 ,
so the x, y, and z-coordinates of a point on γ always satisfy the equation x2 + y2 = z2 . Thus, γ is an
infinite spiral that starts at the origin and unwinds upward and outward; it is shown on Figure 5.3.

Example 5.3. Given two points P1 (x1 , y1 , z1 ) and P2 (x2 , y2 , z2 ), the space curve
r(t) = (1 − t) hx1 , y1 , z1 i + t hx2 , y2 , z2 i (0 ≤ t ≤ 1)
34
is the line segment with endpoints P1 and P2 . Indeed, we can express r(t) in the form

r(t) = hx1 + t(x2 − x1 ), y1 + t(y2 − y1 ), z1 + t(z2 − z1 )i ,

−−−→
which is the parametric equation of the line through P1 that is parallel to P1 P2 . Thus, if the domain
for t was the whole real line R, the curve would be the straight line through P1 and P2 . Restricting
the domain for t to the closed interval [0, 1], we obtain the piece of the line between the endpoints
of r(0) = hx1 , y1 , z1 i and r(1) = hx2 , y2 , z2 i; that is precisely the line segment P1 P2 . 

5.3. Smooth curves

At first glance, continuous curves appear to be a good model for the geometric objects that we
think of as “curves”: that is, two- or three-dimensional objects that are essentially one-dimensional.
However, it turns out that there are some rather peculiar continuous parametric curves. For exam-
ple, in 1890 the Italian mathematician Giuseppe Peano stunned the mathematical community with
his construction of a continuous plane curve that fills the entire unit square (see §5.7); there are also
examples of continuous space curves that fill an entire cube. In order to rule out such pathologi-
cal curves, we shall restrict our attention to a special class of parametric curves called “piecewise
smooth”.

Definition. A curve γ given by a vector function r(t), t ∈ I, is called smooth if r0 (t) is contin-
uous and r0 (t) , 0 (except possibly at the endpoints of I). A curve γ consisting of a finite number
of smooth pieces is called piecewise smooth.

Example 5.4. For the helix r(t) = (cos t)e1 + (sin t)e2 + te3 , t ≥ 0, we have

r0 (t) = (− sin t)e1 + (cos t)e2 + e3 , 0.

Hence, the helix is a smooth curve. 

Example 5.5. Determine whether the curve γ defined by the vector function

r(t) = t2 e1 + t3 e2 + (cos t)e3 (t ∈ R)

is smooth, piecewise smooth, or neither.

Solution. The derivative

r0 (t) = 2te1 + 3t2 e2 − sin te3
vanishes when t = 0. Thus, γ is not smooth. On the other hand, the curves

γ1 : r(t) = t2 e1 + t3 e2 + (cos t)e3 (t ≥ 0)

and
γ2 : r(t) = t2 e1 + t3 e2 + (cos t)e3 (t ≤ 0)
are smooth, since they fail the condition r0 (t) , 0 only at an endpoint of the intervals on which
they are defined. Since γ is the union of γ1 and γ2 , it follows that γ is piecewise smooth. 
35
 z

P
r0 (t)

O y
x

Figure 5.4. The tangent vector r0 (t) at P = r(t)

5.4. Tangent lines

The requirement that r0 (t) , 0 for a smooth curve is not arbitrary. To explain where it comes
from, we now consider the following problem.
Problem. Let γ be a smooth parametric curve with parametric equation
r(t) = x(t)e1 + y(t)e2 + z(t)e3 (t ∈ I).
Find an equation of the tangent line to γ at the point r(t0 ) = (x0 , y0 , z0 ), where t0 is an internal
point of the interval I.
Idea of solution. The tangent line will have the parametric equations
x = x0 + at, y = y0 + bt, z = z0 + ct (t ∈ R),
where u = ha, b, ci is any vector parallel to the tangent line. In other words, we need to determine
the direction of the tangent line.
Recall that the tangent line through P(x0 , y0 , z0 ) is the limit of the secant lines through P and a
point Q(x, y, z) on the curve as Q → P. If Q = r(t), then a vector parallel to the secant line through
P and Q is
−−→
PQ = hx − x0 , y − y0 , z − z0 i = r(t) − r(t0 ).
Thus,
direction of tangent line at P = lim direction of secant line through P, Q
 
t→t0

= lim direction of r(t) − r(t0 )

 
t→t0

= lim direction of normalized r(t) − r(t0 )

 
t→t0

r(t) − r(t0 ) /(t − t0 )

r(t) − r(t0 )
= lim = lim .
t→t0 kr(t) − r(t0 )k t→t0 kr(t) − r(t0 )k/(t − t0 )

By the definition of derivative, the limit of the numerator is r0 (a). By a slightly more complicated
argument (which we omit), the limit of the denominator is kr0 (a)k. Thus, a vector parallel to the
tangent line is r0 (t0 )/kr0 (t0 )k. 
36
 We now see the reason for the requirement of non-vanishing of the derivative in the definition of
smoothness: with that requirement, a smooth curve has a tangent line at every point. Furthermore,
the vector r0 (t) is parallel to the tangent line to the curve at the endpoint P of r(t). For this reason,
r0 (t) is called also the tangent vector to the curve r(t) (see Figure 5.4). The unit tangent vector to
r(t) is merely the normalization of r0 (t), and hence, is given by the formula
r0 (t)
T(t) = .
kr0 (t)k
Example 5.6. For the helix, we have
r0 (t) = (− sin t)e1 + (cos t)e2 + e3 ,
so the unit tangent vector is
r0 (t)
 
h− sin t, cos t, 1i − sin t cos t 1
T(t) = 0 = p = √ , √ , √ .
kr (t)k (− sin t)2 + (cos t)2 + 12 2 2 2

Example 5.7. Find the equation of the tangent line to the helix at the point P corresponding to
the value of the parameter t = π.
Solution. We have P(−1, 0, π). The tangent line passes through P and is parallel to the tangent
vector r0 (π) = h0, −1, 1i (see Example 5.6). Hence, its parametric equation is
r(t) = h−1, 0, πi + t h0, −1, 1i = h−1 + t, −t, π + ti (t ∈ R);
y z−π
the symmetric equations of the tangent line are x = −1, = .
−1 1

Example 5.8. Find the angle between the helix and the curve γ defined by the vector function
r2 (t) = −te1 + (t − 1)2 e2 + πte3 (t ≥ 0)
at P(−1, 0, π).
Solution. Let σ denote the helix. If P lies on both γ and σ, the angle between γ and σ at P
is the angle between their tangent vectors. It is easy to see that both curves do pass through P:
r1 (π) = r2 (1) = h−1, 0, πi. We know from Example 5.7 that the tangent vector to the helix at P is
v1 = h0, −1, 1i. The tangent vector to γ at P is
v2 = r02 (1) = h−1, 2(t − 1), πit=1 = h−1, 0, πi .
Hence, if α is the angle between γ and σ at P, we have
v1 · v2 0 · (−1) + (−1) · 0 + 1 · π π
cos α = = p = √ .
02 + (−1)2 + 12 (−1)2 + 02 + π2 2 + 2π2
p
kv1 k kv2 k
Thus,
π
 
α = arccos √ ≈ 0.8315.
2 + 2π2

37
 5.5. Normal vectors*
We now want to find vectors that are perpendicular to the curve r(t), that is, we want to find
vectors orthogonal to the tangent vector r0 (t), or equivalently, to the unit tangent vector T(t). One
such vector is the unit normal vector
T0 (t)
N(t) = 0 .
kT (t)k
In order to prove that N(t) and T(t) are orthogonal, we differentiate both sides of the identity
T(t) · T(t) = kT(t)k2 = 1.
We obtain
T0 (t) · T(t) + T(t) · T0 (t) = 0 =⇒ 2T(t) · T0 (t) = 0 =⇒ T(t) · T0 (t) = 0.
That is, T0 (t) is orthogonal to T(t).
Another vector orthogonal to T(t) is the unit binormal vector
B(t) = T(t) × N(t).
By the properties of the cross product, B(t) is perpendicular to both T(t) and N(t) and has length
equal to the area of the parallelogram with sides T(t) and N(t). Since this parallelogram is a square
of side length 1, B(t) is a unit vector.
Remark. Note that it is important to use T(t) and not r0 (t) when we compute the normal vector.
In fact,
r00 (t) need not be orthogonal to r0 (t), and in most cases it will not be. For example, if
r(t) = 1, t, t2 , we have r0 (t) = h0, 1, 2ti and r00 (t) = h0, 0, 2i, so
r0 (t) · r00 (t) = 4t , 0 unless t = 0.
On the other hand,
T(t) = 0, (1 + 4t2 )−1/2 , 2t(1 + 4t2 )−1/2 , T0 (t) = 0, −4t(1 + 4t2 )−3/2 , 2(1 + 4t2 )−3/2 ,




and these are orthogonal.

5.6. Arc length

You should be familiar with the formula for arc length of a parametric plane curve: if γ is a
parametric curve in the xy-plane with parametric equations (5.1), where I = [a, b], then its length
L(γ) is given by the formula
Z bp
L(γ) = x0 (t)2 + y0 (t)2 dt. (5.4)
a
In this section, we generalize this formula to space curves. We argue from scratch, so even if (5.4)
appears new, this should not be an obstacle.
First, we must say what we mean by “length” of a curve. We shall focus on piecewise smooth
curves given by parametric equations of the form (5.3). The intuitive idea of length is that we
can take the curve, “straighten” it into a line segment, and then measure the length of that line
segment. Now, imagine that we live in a “rigid” world, where we cannot straighten our curve (or
bend a thread so it would lay on top of the curve). We could try to cut a number of short straight
line segments and piece them together so that the resulting polygon lies close to our original curve;
we can then measure the total length of the polygon and argue that the length of the curve is about
the same. This is the basic idea behind our definition of arc length.
38
 z

O y
x

Figure 5.5. Polygonal approximation to a curve by a 13-gon

Let γ be a parametric curve given by (5.3) with I = [a, b]. For each n ≥ 1, set ∆n = (b − a)/n
and define the numbers that partition [a, b] into n subintervals of equal lengths:
t0 = a, t1 = a + ∆ n , t2 = a + 2∆n , ..., tn = a + n∆n = b.
We write xi = x(ti ), yi = y(ti ), zi = z(ti ) for the coordinates of the point Pi (xi , yi , zi ) that corresponds
to the value t = ti of the parameter. The points P0 , P1 , P2 , . . . , Pn partition γ into n arcs (see
Figure 5.5). Let Ln be the length of the polygon with vertices at those points, that is,
n
X n
X p
Ln = |Pi−1 Pi | = (xi − xi−1 )2 + (yi − yi−1 )2 + (zi − zi−1 )2 .
i=1 i=1

If the sequence converges and lim Ln = L, we call L the arc length of the given curve.
{Ln }∞
n=1
n→∞
Using the above definition and the properties of smooth functions and of definite integrals, we
can prove the following formula for the length of a piecewise smooth curve.
Theorem 5.1. Let γ be a piecewise smooth parametric curve with equation
r(t) = x(t)e1 + y(t)e2 + z(t)e3 (t ∈ [a, b]).
Then the arc length of γ is
Z b p Z b
L(γ) = x0 (t)2 + y0 (t)2 + z0 (t)2 dt = kr0 (t)k dt. (5.5)
a a

Remarks. 1. Note that (5.4) is a special case of (5.5): namely, the case when z(t) = 0.
2. Note that (5.5) gives the length of the parametric curve and not the length of the graph. For
example, the parametric equation
r(t) = cos te1 + sin te2 + 0e3 (0 ≤ t ≤ 4π)
represents the unit circle in the xy-plane. If we apply (5.5) to this curve, we will find that its length
is 4π (and not 2π). The explanation for this is that the above curve traces the unit circle twice.
Example 5.9. If f 0 is continuous on [a, b], the length of the curve y = f (x) is
Z bp
L= 1 + f 0 (x)2 dx.
a
39
This formula follows from the theorem by representing the given graph as a parametric curve with
parametrization r(t) = te1 + f (t)e2 + 0e3 , a ≤ t ≤ b. 

Example 5.10. Find the length of the curve γ given by r(t) = 2t3/2 , cos 2t, sin 2t , where


0 ≤ t ≤ 1.

Solution. We have
r0 (t) = 3t1/2 e1 − 2 sin 2te2 + 2 cos 2te3 ,
so (5.5) yields
Z 1 Z 1 q
2
L(γ) = kr (t)k dt =
0
+ (−2 sin 2t)2 + (2 cos 2t)2 dt
3t1/2
0 0
Z 1p Z 1√
= 9t + 4 sin 2t + 4 cos 2t dt =
2 2 9t + 4 dt
0 0
√

1 13 1/2 2 13 13 − 8
Z
= u du = ≈ 2.879.
9 4 27


5.7. Space-filling curves*

In this section, we describe the general idea of the construction of Peano’s space-filling curve.
We start with a piecewise linear curve contained in the unit square. This curve is then shrunk to
a fraction of its original size and several copies of it are placed throughout the unit square in such
a way that they do not intersect and can be easily connected one to another with straight lines.
This yields a more complicated polygon that also lies within the unit square. The whole process
(shrinking, copying, connecting) is then repeated iteratively. The first two iterations of Peano’s
original constructions are displayed on Figure 5.6. The result of the iterative process is a sequence
of curves. It can be proved that the sequence of continuous functions underlying this sequence of
curves has a limit, which is also a continuous function. It should be intuitively clear (and this can
be proved rigorously) that the limit curve fills the entire square.

Figure 5.6. The first two iterations in Peano’s construction

40
 5.8. Space curves in Mathematica
Mathematica can be quite useful when trying to visualize space curves. Mathematica’s main
built-in command for plotting a space curve is ParametricPlot3D. Its basic syntax is
ParametricPlot3D[f[t], {t,a,b}]
where f is a three-dimensional vector function and a and b are two numerical expressions de-
scribing the endpoints of the interval for the parameter t. The function f can be given by its
name as above, or explicitly, by a vector whose components are functions of t. The command
ParametricPlot3D has many additional options which allow the user to format its output. The
best source to learn about all of those is Mathematica’s help system. Here, we include only a few
examples illustrating the basic syntax and some of the most common options:
• To plot three loops of a helix, having already defined the function r(t) = hcos t, sin t, t/6i:
r[t_] := {Cos[t], Sin[t], t/6}
ParametricPlot3D[r[t], {t,0,6Pi}]
• To plot two thick, black loops of a helix, using the explicit definition:
ParametricPlot3D[{Cos[t], Sin[t], t/6}, {t,0,4Pi},
PlotStyle->{Black,Thick}]
• To plot the spiral on Figure 5.3, without a coordinate frame, thick and in red:
ParametricPlot3D[{t*Cos[t], t*Sin[t], t}, {t,0,8.5Pi},
PlotStyle->{Red,Thick}, Axes->False, Boxed->False]

Exercises
5.1. Does the point P(0, π, 1) lie on the curve parametrized by the given vector function r(t)? If so, for what value(s)
of t?
(a) r(t) = hsin t, 2 arccos t, t + 1i (b) r(t) = hsin t, t, cos 2ti (c) r(t) = te2 + tan te3

5.2. Do the curves r1 (t) = t, t2 , t3 and r2 (t) = t − 2, t2 − 3t + 2, t2 + t − 6 intersect?




= t, t , t and r2 (t) =

2 3
5.3. Let two objects move in space along trajectories described by the vector functions r1 (t)
t − 2, t − 3t + 2, t + t − 6 . Will the objects ever collide? That is, will they ever be at the same point in space at the

2 2


same time?
5.4. Find the unit tangent vector T(t) to the curve r(t) = hsin 3t, cos 3t, ln(t + 1)i at the point with t = 0.
5.5. Find the equations of the tangent line to the curve r(t) = cos(πt), 3 sin(πt), t3 at the point (−1, 0, 1).



5.6. Find the equations of the tangent line to the curve r(t) = t3 − 2, t2 + 1, 3t + 1 at the point (−2, 1, 1).



5.7. Find the equations of the tangent line to the curve r(t) = h5 cos t, 5 sin t, 2 cos 2ti at the point (−5, 0, 2).
5.8. Find the points on the curve r(t) = h5 cos t, 5 sin t, 2 cos 2ti where the tangent line is parallel to the x-axis.
Is the given curve smooth? If not, then is it piecewise smooth?
5.9. r(t) = t
3 + t, sin 2t, 3t2 − 1 5.11. r(t) = cos t, t2 , cos 2t




5.10. r(t) = 1, t2 , t3 , −1 ≤ t ≤ 1 5.12. r(t) = hcos t, cos 2t, cos 3ti

5.13. Find the intersection point of the curves r1 (t) = 2t, t2 + t, t3 and r2 (t) = t, t2 − 3t + 4, t2 + t − 5 . Find the




angle between the curves at that point, that is, find the angle between the tangent vectors at the point.
5.14. The curves r1 (t) = t2 , 2t, ln(t2 − 3) and r2 (t) = t + 1, t2 − 5, t2 − 2t − 3 intersect at the point (4, 4, 0). Find the




cosine of the angle between the two curves at that point.

41
Find the arc length of the curve r(t).
√ 
5.15. r(t) = 1, t2 , t3 , −1 ≤ t ≤ 1 5.16. r(t) = t, ln t, 2 2t , 1 ≤ t ≤ 2



Rb
5.17. Let r(t), a ≤ t ≤ b, be a parametrization of a smooth curve γ. Let L = a kr0 (t)k dt be the length of γ and define
the new function Z t
s(t) = kr0 (u)k du (a ≤ t ≤ b).
a
(a) Evaluate s(a) and s(b).
(b) Use the Fundamental Theorem of Calculus to show that s0 (t) > 0 for all t with a < t < b.
(c) Let t = t(s), 0 ≤ s ≤ L, be the inverse function of the function s(t), and define the vector function
f(s) = r(t(s)), 0 ≤ s ≤ L. Show that kf 0 (s)k = 1 for all s with 0 < s < L.
(d) Convince yourself that the vector function f(s) from part (c) is another parametrization of the curve γ. This
parametrization has a special property: f(s) is the position vector of the point on γ that is s units along the
curve after the initial point. This parametrization is known as the parametrization with respect to arc length
or the natural parametrization and plays an important role
in the study of space
 curves.
(e) Parametrize with respect to arc length the curve r(t) = e3t sin 4t, e3t cos 4t, 5 , 0 ≤ t ≤ 13 ln 4.

42
 LECTURE 6

Multivariable Functions

6.1. Functions of two variables

In this lecture we introduce multivariable functions. These are functions that depend on mul-
tiple inputs (usually 2, 3, or 4), but produce only a single output. We start our discussion with the
simplest example: functions of two variables. For example, the temperature T at any point on the
surface of the Earth at this very moment depends on two pieces of information: the longitude x
and the latitude y of the point. Thus, we can think of T as a function T (x, y) of two variables x and
y or as a function of the ordered pair (x, y). We now give a formal definition.
Definition. A function f of two variables is a rule that assigns to each ordered pair of real
numbers (x, y) in a set D a unique real number denoted f (x, y). The set D is the domain of f and
the set of all values that f takes on is its range, that is, the range of f is { f (x, y) : (x, y) ∈ D}.

6.2. Functions of three and more variables

The example that we used above to motivate the introduction of functions of two variables is
somewhat artificial. A more natural (and also more useful) quantity is the temperature T at a point
(x, y, z) in space at this very moment. Then T is a function of three variables—the coordinates
(x, y, z) of the point. Note that with this function in hand we don’t have to assume that the temper-
ature in a classroom and the temperature on the wing of an airplane flying 39,000 feet above that
classroom are the same. An even more realistic example is that of the temperature T at the point
(x, y, z) in space at time t. This quantity is a function T (x, y, z, t) of four variables.
Definition. A function f of n variables x1 , x2 , . . . , xn is a rule that assigns to each ordered n-
tuple of real numbers (x1 , x2 , . . . , xn ) in a set D a unique real number denoted f (x1 , x2 , . . . , xn ). The
set D is the domain of f and the set of all values that f takes on is its range. In particular, a function
f of three variables is a rule that assigns to each ordered triple of real numbers (x, y, z) in a set D
(in R3 ) a unique real number denoted f (x, y, z).

6.3. Examples of domains and ranges of multivariable functions

√
Example 6.1. The function f (x, y) = 1 − xy is defined whenever
1 − xy ≥ 0 ⇐⇒ xy ≤ 1.
If x > 0, the last inequality is satisfied for y ≤ x−1 ; if x < 0, the inequality is satisfied for y ≥ x−1 ;
and if x = 0, it is satisfied for all y. That is, the domain of f contains the points in I quadrant on
and below the graph of y = x−1 , the points in III quadrant on and above the graph of y = x−1 , and
all the points in II and IV quadrants, including the x- and y-axes (see Figure 6.1).
The range of f is [0, ∞). Clearly, f cannot attain negative values. On the other hand, if k ≥ 0,
1 − xy = k ⇐⇒ 1 − xy = k2 ⇐⇒ xy = 1 − k2 .
p

43
Hence, if k ≥ 0, k , 1, f (x, y) = k for all points (x, y) on the hyperbola xy = 1 − k2 ; also, f (x, y) = 1
for all points (x, y) on the coordinate axes. That is, every number k ∈ [0, ∞) does belong to the
range of f . 

y z

y
x

√
Figure 6.1. The domain of z = 1 − xy Figure 6.2. z = 3 − x2 − y2

Example 6.2. The domain of the function f (x, y, z) = ln(16 − x2 − y2 − 9z2 ) consists of the points
(x, y, z) in space whose coordinates satisfy
x2 y2 z2
16 − x2 − y2 − 9z2 > 0 ⇐⇒ + + < 1.
16 16 16/9
Since ln x is an increasing function, the range of f consists of the logarithms of the positive num-
bers in the range of the function w = 16 − x2 − y2 − 9z2 . Because x2 + y2 + 9z2 can attain any
non-negative value and cannot attain any negative value, it follows that w takes on the numbers in
the interval (−∞, 16]. Thus, the range of f consists of the logarithms of the numbers in (0, 16], that
is, the range of f is (−∞, ln 16]. 

6.4. Graphs
Definition. If f is a function of n variables x1 , x2 , . . . , xn with domain D, then the graph of f
is the set of all points (x1 , x2 , . . . , xn , z) in Rn+1 with (x1 , x2 , . . . , xn ) ∈ D and z = f (x1 , x2 , . . . , xn ),
that is, the graph of f is the set
(x1 , x2 , . . . , xn , f (x1 , x2 , . . . , xn )) : (x1 , x2 , . . . , xn ) ∈ D .


In particular, the graph of a function f (x, y) defined on a set D in R2 is the surface given by
(x, y, f (x, y)) : (x, y) ∈ D .


Example 6.3. The graph of the function f (x, y) = 3 − x2 − y2 is the surface

z = 3 − x2 − y2 .
This surface is displayed on Figure 6.2. 
44
 6.5. Level curves and surfaces
While the graph of a function of two variables provides an excellent visualization of the func-
tion, it might be quite difficult to sketch without advanced technology. What can we do using
only manual (and mental) labor? One answer is well-known to cartographers. When drawing
topographic or termal maps, cartographers often produce two-dimensional drawings marked with
curves of constant elevation or constant temperature. We can do the same for any function f of
two variables: we can graph f in the xy-plane by marking its “level curves”.
Definition. Let f be a function of two veriables, denoted x and y, and let k be a real number.
The k-level curve of f is the set of all points (x, y) in the domain D of f which satisfy the equation
f (x, y) = k.
This idea becomes even more important in the case of functions of three variables. Note that
the graph w = f (x, y, z) of such a function is a four-dimensional object, which we can study
analytically but cannot represent graphically in our three-dimensional universe. However, we can
draw the “level surfaces” of f .
Definition. Let f be a function of three veriables, denoted x, y and z, and let k be a real number.
The k-level surface of f is the set of all points (x, y, z) in the domain D of f which satisfy the
equation f (x, y, z) = k.

x
− 13 − 12 −1 −2 2 1 1 1
2 3

Figure 6.3. The level curves of f (x, y) = x/(x2 + y2 )

Example 6.4. Describe the level curves of the function

x
f (x, y) = 2 .
x + y2
Solution. We want to draw the curve
x
=k
x2 + y2
for all k. If we exclude the origin from consideration (since the curve is undefined there), we can
represent this curve by the equation
kx2 + ky2 − x = 0.
When k = 0, this equation represents the y-axis x = 0, so the level curve f (x, y) = 0 is the y-axis
with the origin removed from it. When k , 0, we can transform the equation of the level curve
45
further to  2
x 1 1
x − + y2 = 0
2
⇐⇒ x− + y2 = ,
k 2k 4k2
which is the equation of a circle centered at ( 2k1 , 0) of radius 2|k|
1
(k can be negative). Since the origin
does lie on this circle, the level curve f (x, y) = k is the circle with the origin removed from it. A
contour map of f showing the level curves with k = 0, ± 13 , ± 12 , ±1, ±2 is shown on Figure 6.3. 

6.6. Limits of multivariable functions

Since all the important concepts in single-variable calculus were defined in terms of limits, if
we are to generalize those concepts to functions of two and more variables, we must first define
the notion of a limit of a multivariable function. The following definition generalizes the formal
definition of limit to functions of two variables.
Definition. Let f be a function of two variables whose domain D includes points arbitrarily
close to (a, b). We say that the limit of f (x, y) as (x, y) approaches (a, b) is L and write
lim f (x, y) = L or f (x, y) → L as (x, y) → (a, b),
(x,y)→(a,b)

if for any given number ε > 0, there is a corresponding number δ = δ(ε) > 0 such that
p
0 < (x − a)2 + (y − b)2 < δ =⇒ | f (x, y) − L| < ε.
This definition is very formal. It is also very useful in proofs (something we will not be con-
cerned with) and very imposing at first sight. A more intuitive definition reads as follows.
Definition. Let f be a function of two variables whose domain D includes points arbitrarily
close to (a, b). We say that the limit of f (x, y) as (x, y) approaches (a, b) is L and write
lim f (x, y) = L or f (x, y) → L as (x, y) → (a, b),
(x,y)→(a,b)

if f (x, y) approaches L as (x, y) approaches (a, b), independent of how (x, y) approaches (a, b).
We can define limits of functions of three or more variables in a similar fashion.
Definition. Let f be a function of three variables whose domain D includes points arbitrarily
close to (a, b, c). We say that the limit of f (x, y, z) as (x, y, z) approaches (a, b, c) is L and write
lim f (x, y, z) = L or f (x, y, z) → L as (x, y, z) → (a, b, c),
(x,y,z)→(a,b,c)

if f (x, y, z) approaches L as (x, y, z) approaches (a, b, c), independent of how (x, y, z) approaches
(a, b, c). More precisely, the limit of f (x, y, z) as (x, y, z) approaches (a, b, c) is L, if for any given
number ε > 0, there is a corresponding number δ = δ(ε) > 0 such that
p
0 < (x − a)2 + (y − b)2 + (z − c)2 < δ =⇒ | f (x, y, z) − L| < ε.

Limits of multivariable functions can be much trickier than those of single-variable functions.
The primary reason for that is the infinitude of ways in which the arguments (x, y) (or (x, y, z)) can
approach the point (a, b) (or (a, b, c)). This point is best illustrated by examples of limits that do
not exist.
46
 Example 6.5. The limit
x
lim
+ y2
(x,y)→(0,0) x2
does not exist. Indeed, we know from Example 6.4 (see also Figure 6.3) that the level curves of
this function all want to pass through (0, 0) (but since the point is not in the domain of f , they
can’t). Thus, if we approach (0, 0) along two distinct level curves (say, along f (x, y) = 1 and
f (x, y) = −1), the function values will approach distinct values (1 and −1, respectively). However,
if the limit existed, f (x, y) would have to approach the same value independent of the way (x, y)
approaches (0, 0). Therefore, the limit does not exist. 
xy4
Example 6.6. Show that the limit lim does not exist.
(x,y)→(0,0) x2 + y8
Solution. First, consider the behavior of the function as (x, y) → (0, 0) along the straight line
y = x. We have
x5 x3
lim f (x, x) = lim 2 = lim = 0.
x→0 x→0 x + x8 x→0 1 + x6
On the other hand, if we let (x, y) → (0, 0) along the curve x = y4 , we get
y8 1
= .
lim f (y4 , y) = lim
y→0 y→0 y + y
8 8 2
Since different approach paths lead to different limiting values, the limit does not exist. 
Example 6.7. For any numbers a, b, c, we have
lim x = a, lim y = b, lim c = c.
(x,y)→(a,b) (x,y)→(a,b) (x,y)→(a,b)

Using these facts and the two-variable versions of the limit laws:
lim [ f (x, y) ± g(x, y)] = lim f (x, y) ± lim g(x, y), etc.,
(x,y)→(a,b) (x,y)→(a,b) (x,y)→(a,b)

we can compute the limit of any polynomial in x and y at any point (a, b) in the plane. For example,
lim xy = lim y = 2 · 3 = 6.



lim x
(x,y)→(2,3) (x,y)→(2,3) (x,y)→(2,3)

Next, we look at some examples of evaluation of limits of functions of two and three variables.
sin(x2 + y2 )
Example 6.8. Evaluate lim .
(x,y)→(0,0) x2 + y2
Solution. If we change the coordinates of the point (x, y) from Cartesian to polar, the condition
(x, y) → (0, 0) is equivalent to r → 0 and the limit takes the form
sin(x2 + y2 ) sin(r2 ) 2r cos(r2 )
lim = lim = lim = 1,
(x,y)→(0,0) x 2 + y2 r→0 r2 r→0 2r
where we used l’Hospital’s rule to pass from the second limit to the third. 

47
 (x2 + y2 )z
Example 6.9. Evaluate lim .
(x,y,z)→(0,0,0) x2 + y2 + z2
Solution. This time, we change the coordinates of the point (x, y, z) from Cartesian to spheri-
cal, and the condition (x, y, z) → (0, 0, 0) becomes ρ → 0. Hence,
(x2 + y2 )z (ρ sin φ)2 (ρ cos φ)
lim = lim
(x,y,z)→(0,0,0) x2 + y2 + z2 ρ→0 ρ2
= lim ρ sin2 φ cos φ = 0. 
ρ→0

x2 y − 2x2 − 2xy + 6x + 5y − 12
Example 6.10. Evaluate lim .
(x,y)→(1,2) x2 + y2 − 2x − 4y + 5
Solution. Both the numerator and the denominator of the function vanish at the point (1, 2),
so the limit is an indeterminate form of type 0/0. We will try to argue similarly to the previous
example, but before we can do so, we need to change the limit from one at the point (1, 2) to one
at the origin. Let us define new variables u and v by
x = 1 + u, y = 2 + v.
The condition (x, y) → (1, 2) is equivalent to (u, v) → (0, 0), so we can rewrite the original limit as
x2 y − 2x2 − 2xy + 6x + 5y − 12
lim
(x,y)→(1,2) x2 + y2 − 2x − 4y + 5
(1 + u)2 (2 + v) − 2(1 + u)2 − 2(1 + u)(2 + v) + 6(1 + u) + 5(2 + v) − 12
= lim
(u,v)→(0,0) (1 + u)2 + (2 + v)2 − 2(1 + u) − 4(2 + v) + 5
u2 v + 2u + 4v
= lim .
(u,v)→(0,0) u2 + v2
We can now change the coordinates of the point (u, v) from Cartesian to polar. We find that
u2 v + 2u + 4v r3 cos θ sin θ + 2r cos θ + 4r sin θ
lim = lim
(u,v)→(0,0) u2 + v2 r→0 r2
= lim r cos θ sin θ + (2 cos θ + 4 sin θ)r−1


r→0

= lim(2 cos θ + 4 sin θ)r−1

r→0

+∞ when 2 cos θ + 4 sin θ > 0,

= 0 when 2 cos θ + 4 sin θ = 0,
−∞ when 2 cos θ + 4 sin θ < 0.


Hence, the given limit does not exist. 

6.7. Continuity of multivariable functions

We can use the notion of limit of a multivariable function to define the notion of continuity.
Definition. Let f be a function of two variables whose domain D includes (a, b) and points
arbitrarily close to (a, b). We say that f (x, y) is continuous at (a, b), if
lim f (x, y) = f (a, b).
(x,y)→(a,b)

48
We say that f is continuous in D, if it is continuous at every point (a, b) ∈ D.
Definition. Let f be a function of three variables whose domain D includes (a, b, c) and points
arbitrarily close to (a, b, c). We say that f is continuous at (a, b, c), if
lim f (x, y, z) = f (a, b, c).
(x,y,z)→(a,b,c)

We say that f is continuous in D, if it is continuous at every point (a, b, c) ∈ D.

We want also to extend the notion of piecewise continuity to multivariable functions, and in
particular, to functions of two variables. Recall that a single-variable function f is called piecewise
continuous on [a, b] if it is continuous everywhere in [a, b] except for a finite number of points,
where it can have only jump or removable discontinuities. The two-dimensional version of this
definition reads as follows.
Definition. Let f be a function of two variables with domain D. We say that f is piecewise
continuous in D, if it is continuous at every point (a, b) ∈ D, except possibly for the points on a
finite number of smooth curves in D, say γ1 , γ2 , . . . , γn , where f may have “jump discontinuities.”
Remark. As in the case of single-variable functions, polynomials in several variables are con-
tinuous. Likewise, rational functions in several variables are continuous except for the points where
the denominator is 0. Finally, if f is a continuous function (of a single variable) and g is a con-
tinuous function of the variables x, y, . . . , the composition f (g(x, y, . . . )) is continuous at all points
(x, y, . . . ) where it is defined.
Example 6.11. The function
x2 − y2
 
f (x, y) = cos
x2 + y2
is continuous everywhere in its domain, which is all the points in R2 except (0, 0). Indeed, f is the
composition of the continuous function w = cos t and the rational function
x2 − y2
t= .
x2 + y2
Example 6.12. Show that the function
 sin(x + y )
2 2

if (x, y) , (0, 0),
f (x, y) = x2 + y2
1 if (x, y) = (0, 0),


is continuous everywhere.
Solution. Recall that
sin x
lim = 1.
x→0 x
Hence, the function 
 sin x
if x , 0,
g(x) = x
1 if x = 0,
is continuous everywhere. Furthermore, the polynomial x2 + y2 is also continuous everywhere.
Therefore, the function f (x, y) = g(x2 + y2 ) is also continuous everywhere. 
49
 Exercises
Find the domain and the range of the given function.
z
6.1. f (x, y) = y2 e−x /2 6.2. f (x, y) = ln x + y2
2


6.3. f (x, y, z) =
x2 − y2 − 1

Sketch the domain of the given function.

√
6.4. f (x, y) = ln x + y2 x + 2y


6.5. f (x, y) =
x 2 + y2 − 1

For each function, draw the level curves for levels −1, 0, 1, 2, and 3; then sketch the graph z = f (x, y).
6.6. f (x, y) = 1 + y2 6.7. f (x, y) = 2x + 3y − 5 6.8. f (x, y) = x2 + y2

Find the given limit or show that it does not exist.

sin(2xy) xy
x4 − 4xy2 + 2x3 y


6.9. lim 6.11. lim 6.13. lim
x2 + y2
p
(x,y)→(1,3) (x,y)→(0,0)
(x,y,z)→(1,0,2) 1 − zy2
xy xy
6.10. lim 3x2 y 6.14. lim √
(x,y)→(0,0) x2 + y2 6.12. lim
(x,y)→(0,0) x4 + y2
(x,y)→(0,0) 1 + xy − 1

Find the set of points in R2 or R3 where the given function is continuous.

6.15. f (x, y) = sin
 x +x y 6.17. f (x, y, z) =
4 2 2

2 2 2


ln 4 −xyx − y − 4z
 xy if (x, y) , (0, 0), √ if (x, y) , (0, 0),
6.16. f (x, y) = x2 + y2 6.18. f (x, y) = 1 + xy − 1
1 if (x, y) = (0, 0). 1 if (x, y) = (0, 0).

50
 LECTURE 7

Surfaces

7.1. Introduction
In single-variable calculus, the interplay between curves and functions is essential: we use
geometrical questions about curves to motivate the study of calculus and then apply our knowledge
of calculus to answer questions about plane curves. A similar symbiosis occurs in multivariable
calculus between functions and surfaces. However, while we are familiar with a fair number of
plane curves, our supply of known surfaces is limited essentially to planes, spheres, and graphs
z = f (x, y). We introduced also the level surfaces of functions of three variables: these are the sets
Σ = (x, y, z) ∈ R3 : F(x, y, z) = k .


However, there are hardly any level surfaces that we recognize. The main goal of this lecture is to
introduce some other surfaces that are commonly used in multivariable calculus.

7.2. Cylinders
Definition. A cylinder is a surface that consists of all lines—called rullings—that are parallel
to a given line and pass through a given plane curve.
Example 7.1. We have seen already one example of a cylinder: in Example 1.8, we explained
that the equation x2 +y2 = x describes the surface Σ in space that consists of all the points on vertical
lines that pass through a circle in the xy-plane. Note that by a similar argument, any equation of
one of the forms
f (x, y) = 0, f (x, z) = 0, or f (y, z) = 0
represents a cylinder in space. For example, the equation 2y2 + z2 = 1 represents the cylinder that
passes through the ellipse 2y2 + z2 = 1 in the yz-plane and has rullings parallel to the x-axis. Note
that the three-dimensional equations of the ellipse are
2y2 + z2 = 1, x = 0.
Similarly, in space, the equation x = y2 represents the cylinder that passes through the parabola
x = y2 in the xy-plane and has rullings parallel to the z-axis. 

7.3. Quadric surfaces

Quadric surfaces are generalizations of the conic sections to three dimensions.
Definition. A quadric surface is the set of all points (x, y, z) ∈ R3 that satisfy a second-degree
equation in the variables x, y, and z:
A11 x2 + A12 xy + A13 xz + A22 y2 + A23 yz + A33 z2 + B1 x + B2 y + B3 z + B0 = 0,
where A11 , . . . , A33 , B0 , . . . , B3 are constants.
51
 In other words, a quadric surface is a level surface of a quadratic polynomial in three variables.
In the next couple of examples, we visualize some quadric surfaces by sketching their traces in
planes parallel to the coordinate planes.
Example 7.2. Consider the quadric surface
x2
+ y2 + z2 = 1.
4
The trace of this surface in the horizontal plane z = k has equations
x2
+ y2 = 1 − k2 , z = k,
4
which represent an ellipse, provided that
1 − k2 > 0 ⇐⇒ −1 < k < 1.
Similarly, the traces of the surface in the planes y = k, −1 < k < 1, and x = k, −2 < k < 2, are also
ellipses. Thus, the given surface is called an ellipsoid. 
Example 7.3. Consider the quadric surface
z = x2 − 2y2 .
The trace of this surface in a plane x = k is
z = k2 − 2y2 , x = k,
which is a parabola open downward. Likewise, the trace in a plane y = k is a parabola open
upward,
z = x2 − 2k2 , y = k.
Lastly, the trace in a vertical plane z = k is the hyperbola
x2 − 2y2 = k, z = k.
This surface is called a hyperbolic paraboloid. 
Remark. At first, one might think that there is no neat classification of all quadric surfaces, but
in fact there is. First of all, by a change of the coordinates, every quadric surface can be brought
into the form
Ax2 + By2 + Cz2 + Dz + E = 0,
where either D = 0 or C = E = 0. If one of the variables is missing—say, if A = 0 or B = 0,
then the quadric surface is a cylinder parallel to the respective coordinate axis. This leaves only six
possibilities, which lead to the following six surfaces: cone, ellipsoid, elliptic paraboloid, hyper-
bolic paraboloid, one-sheet hyperboloid, and two-sheet hyperboloid. These surfaces in standard
positions are displayed on Figure 7.1 and some basic facts about them are summarized in Table 7.1.

Example 7.4. Classify the quadric surface Σ with equation

4y2 + z2 − x + 16y − 4z + 19 = 0.
52
 Surface Equation Traces Axis
xy yz xz
Elliptic z x2 y2
= + Ellipses Parabolas Parabolas z-axis
Paraboloid c a2 b2
2
Hyperbolic z x y2
= 2− 2 Hyperbolas Parabolas Parabolas z-axis
Paraboloid c a b
2 2
x y z2
Ellipsoid + + =1 Ellipses Ellipses Ellipses
a2 b2 c2
One-Sheet x2 y2 z2
+ − =1 Ellipses Hyperbolas Hyperbolas z-axis
Hyperboloid a2 b2 c2
x2 y2 z2
Cone + − =0 Ellipses Hyperbolas Hyperbolas z-axis
a2 b2 c2
Two-Sheet x2 y2 z2
+ − = −1 Ellipses Hyperbolas Hyperbolas z-axis
Hyperboloid a2 b2 c2
Table 7.1. The non-degenerate quadric surfaces

z z z

y
y
y x x
x

(a) Elliptic Paraboloid (b) Hyperbolic Paraboloid (c) Ellipsoid

z z z

y y y
x x x

(d) One-Sheet Hyperboloid (e) Cone (f) Two-Sheet Hyperboloid

Figure 7.1. The non-degenerate quadric surfaces

53
 Solution. We can rewrite this equation in the form
4(y + 2)2 + (z − 2)2 − (x + 1) = 0.
If we now change the coordinates to
X = y + 2, Y = z − 2, Z = x + 1,
we find that a point in P(x, y, z) in the xyz-space is on Σ if and only if P(X, Y, Z) is on the surface
Z = 4X 2 + Y 2 in the XYZ-space. The latter surface is an elliptic paraboloid with a = 14 , b = 1,
and c = 1. Note that in the xyz-space the vertex of the paraboloid is at (−1, −2, 2) (these are the
xyz-coordinates of the origin of the XYZ-coordinate system) and its axis is the line y = −2, z = 2
(these are equations of the Z-axis in xyz-coordinates). 

7.4. Parametric surfaces

By now, you have probably come to the realization that a “surface” is a three-dimensional
geometric object that is essentially two-dimensional. In the case of graphs z = f (x, y) this is
quite obvious: the points (x, y, z) on the graph are in one-to-one correspondence with the points
(x, y) in the domain of f (x, y). For example, the points on the elliptic paraboloid z = x2 + 2y2
(see Figure 7.1(a)) are in a one-to-one correspondence with the points in the xy-plane. On the
other hand, the two-dimensionality of a level surface, such as the ellipsoid x2 + 2y2 + z2 = 4,
is less obvious from its equation (though it is quite intuitive from Figure 7.1(c)). Level surfaces
are also hard to plot, because there is no easy way to generate points on a level surface. On the
other hand, level surfaces are the natural analytic representation for some basic examples: the cone
x2 + y2 − z2 = 0, the cylinder x2 + y2 = 1, and the sphere x2 + y2 + z2 = 1 are all level surfaces
that cannot be represented as graphs of functions of two variables. In this section, we introduce the
notion of a “parametric surface”.
Definition. If x = x(u, v), y = y(u, v) and z = z(u, v) are three continuous functions defined on
a set D in R2 , then the set Σ of the points (x, y, z) satisfying the parametric equations
x = x(u, v), y = y(u, v), z = z(u, v) ((u, v) ∈ D), (7.1)
is called a parametric surface; the variables u, v are called the parameters. Equations (7.1) and the
vector function r : R2 → R3 , given by
r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 ((u, v) ∈ D), (7.2)
are referred to as parametrizations of Σ.
Remark. Like level surfaces, parametric surfaces generalize of the graphs z = f (x, y) of func-
tions of two variables. Indeed, suppose that Σ is the surface z = f (x, y), (x, y) ∈ D. Then we can
parametrize Σ by the equations
x = u, y = v, z = f (u, v) ((u, v) ∈ D).
On the other hand, a parametric surface is almost as easy to plot as the graph of a function: to plot
z = f (x, y), we let the point (x, y) roam through the domain D and plot the points (x, y, f (x, y)); to
plot the surface Σ parametrized by (7.2), we let (u, v) roam through D and plot the endpoints of the
position vectors r(u, v).
54
 Example 7.5. Identify the surface Σ, parametrized by the vector function
r(u, v) = ue1 + u cos ve2 + u sin ve3 ,
where −∞ < u < ∞ and 0 ≤ v ≤ 2π.
Solution. For a point (x, y, z) on Σ, we have
y2 + z2 = (u cos v)2 + (u sin v)2 = u2 = x2 .
This is the equation of a cone whose axis of symmetry is the x-axis. 
Example 7.6. Find a parametrization of the plane Π through the points A(1, 2, 3), B(1, −1, −2),
and C(−2, 0, 1).
−−→
Solution. Let us first find an equation of the plane. Two vectors parallel to Π are AB =
−−→ −−→ −−→
h0, −3, −5i and BC = h−3, 1, 3i. Hence, the vector n = AB × BC is normal to Π. We have
n = (−3e2 − 5e3 ) × (−3e1 + e2 + 3e3 )
= 9(e2 × e1 ) − 9(e2 × e3 ) + 15(e3 × e1 ) − 5(e3 × e2 ) = −4e1 + 15e2 − 9e3 ,
so an equation for Π is
−4(x − 1) + 15(y − 2) − 9(z − 3) = 0.
Let us solve this equation for one of the variables, say for x:
−4(x − 1) + 15(y − 2) − 9(z − 3) = 0 ⇐⇒ x−1= 15
4
(y − 2) − 94 (z − 3)
⇐⇒ x= 15
4
y − 49 z + 14 .
We see that Π is the graph of the function
x= 15
4
y − 49 z + 14 ,
so it is parametrized by the vector function
r(y, z) = − 94 z + e1 + ye2 + ze3 ,
15 1


4
y 4
where −∞ < y, z < ∞. 
Example 7.7. Let a = ha1 , a2 , a3 i and b = hb1 , b2 , b3 i be two vectors that are not parallel to
each other. Find a parametrization of the plane Π through the origin that is parallel to a and b.
Solution. We can try to argue similarly to the previous problem. However, the equation of Π
is
a2 b3 − a3 b2 x + a3 b1 − a1 b3 y + a1 b2 − a2 b1 z = 0,

and it is not clear which variable to solve for: any of the coefficients can be zero. Thus, we will
use a different approach.
Let P(x, y, z) be an arbitrary point in Π. We consider a parallelogram OAPB in Π, whose sides
−−→ −−→
OA and OB are parallel to a and b, respectively (see Figure 7.2). Then OA = ua and OB = vb for
some numbers u, v. If we denote the position vector of P by r, we conclude that r = ua + vb. By
varying u, v we get the entire plane Π, so one possible parametrization is
r(u, v) = (a1 u + b1 v)e1 + (a2 u + b2 v)e2 + (a3 u + b3 v)e3 (−∞ < u, v < ∞). 

55
 P

B
b r
A
a
O

Figure 7.2. The parallelogram OAPB

Example 7.8. Find a parametrization of the elliptic cylinder

x2 + 4z2 = 1, y ≥ 5.
Solution. A parametrization of the ellipse x2 + 4z2 = 1 in the xz-plane is given by the functions
x = cos u, z = 12 sin u (0 ≤ u ≤ 2π).
The points on the given cylinder must have x- and z-coordinates that satisfy these two equations
and a y-coordinate with y ≥ 5. Thus, we can parametrize the cylinder by
r(u, v) = cos ue1 + ve2 + 12 sin ue3 ,
where 0 ≤ u ≤ 2π and v ≥ 5. 
Example 7.9. Find a parametrization of the part of the plane 3x + 2y + z = 4 that lies within the
paraboloid z = x2 + 2y2 + 1.
Solution. By solving the equation of the plane for z, z = 4 − 3x − 2y, we can easily obtain a
parametrization of the whole plane:
r(u, v) = ue1 + ve2 + (4 − 3u − 2v)e3 (−∞ < u, v < ∞).
To parametrize just the given part of the plane, we restrict u, v to its projection onto the xy-plane.
The plane and the paraboloid intersect along the curve that contains the simultaneous solutions of
the equations
3x + 2y + z = 4, z = x2 + 2y2 + 1.
Eliminating z, we find that the x- and y-coordinates of the points on this curve satisfy

2
2
4 − 3x − 2y = x2 + 2y2 + 1 =⇒ x + 32 + 2 y + 12 = 23 4
.
Thus, the intersection curve of the plane and the ellipsoid consists of all points on the plane whose
x- and y-coordinates satisfy the last equation. Consequently, the intersection curve projects onto an
ellipse in the xy-plane, and the part of the plane 3x + 2y + z = 4 that lies within the given paraboloid
projects onto the interior of that ellipse. We obtain the parametrization

2
2
r(u, v) = ue1 + ve2 + (4 − 3u − 2v)e3 , u + 23 + 2 v + 12 ≤ 23 4
. 

7.5. Plotting surfaces using Mathematica

Mathematica provides several commands for plotting surfaces, which we survey in this section.
56
 7.5.1. Graphs of functions of two variables: Plot3D. To plot the surface z = f (x, y), where
f is a function of two variables, we use the command Plot3D. The basic form of this command is
Plot3D[f[x,y], {x,a,b}, {y,c,d}]
where a, b, c and d are numbers or numerical expressions. This will generate the part of the graph
of f where a ≤ x ≤ b and c ≤ y ≤ d. For example, the following command will plot the surface
z = x2 − y2 over the rectangular region given by −1 ≤ x ≤ 1 and −1 ≤ y ≤ 1:
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1}]
Similarly to the Plot command, Plot3D has many options. You can get a list of these options
and their default settings by executing
Options[Plot3D]
or by reading the help on Plot3D. Here we demonstrate a few of the more commonly used options.
• To label the axes, we use the AxesLabel option:
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1}, AxesLabel->{"x","y","z"}]
• To change the aspect ratio of the plot, we specify values to the parameter BoxRatios.
The option BoxRatios specifies the ratio of the side lengths for the box that encloses the
graph. The setting BoxRatios->Automatic gives identical scales on all three axes.
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1}, BoxRatios->{1,2,3}]
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1}, BoxRatios->Automatic]
• To suppress the box surrounding the surface we set the option Boxed to False. Similarly,
the axes can be suppressed by setting Axes to False.
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1}, Boxed->False, Axes->False]
• The colors of the surface may be changed using the options PlotStyle and ColorFunction
and the color-specifying commands RGBColor or Hue as shown in the examples below.
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1}, PlotStyle->Red]
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1}, PlotStyle->RGBColor[1,0,0]]
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1}, PlotStyle->Hue[0.5]]
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1},
ColorFunction->Function[{x,y,z}, RGBColor[xˆ2,0,yˆ2]]]
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1},
ColorFunction->Function[{x,y,z}, RGBColor[xˆ2,0,yˆ2]],
ColorFunctionScaling->False]
• To eliminate the colors altogether, we set the option ColorFunction to (White&). We
can also suppress the mesh that Mathematica lays on the plot by setting the option Mesh
to False.
Plot3D[xˆ2-yˆ2, {x,-1,1}, {y,-1,1},
ColorFunction->(White&), Mesh->False]
Rotating the surface around can, sometimes, reveal its shape and characteristics more clearly.
To rotate a surface plot in Mathematica, simply left-click on it and start moving the mouse without
releasing the mouse button. Sliding the mouse up and down while holding the Ctrl key allows
you to zoom in and out the surface.
57
 7.5.2. Level surfaces: ContourPlot3D. The basic syntax for plotting a level surface using
Mathematica is
ContourPlot3D[f[x,y,z]==k, {x,x1,x2}, {y,y1,y2}, {z,z1,z2}]
This command plots the part of the surface f (x, y, z) = k that lies within the box given by the
inequalities
x1 ≤ x ≤ x2 , y1 ≤ y ≤ y2 , z1 ≤ z ≤ z2 .
For convenience, the command ContourPlot3D also accepts surfaces given by the “more general”
type of equation f (x, y, z) = g(x, y, z):
ContourPlot3D[f[x,y,z]==g[x,y,z], {x,x1,x2}, {y,y1,y2}, {z,z1,z2}]
Finally, this command can plot simultaneously several level surfaces of the same function f . For
example, to plot the −1-, 3- and 6-level surfaces of the function x2 − y2 − 3z (in blue, red and green,
respectively), we use
ContourPlot3D[xˆ2-yˆ2-3z, {x,-2,2}, {y,-2,2}, {z,-2,2},
Contours->{{-1,Blue}, {3,Red}, {6,Green}}
7.5.3. Parametric surfaces: ParametricPlot3D. To plot the parametric surface with pa-
rametrization (7.1), we use a version of the command ParametricPlot3D, which we used in
Lecture #5 to plot a parametric space curve. When used to plot a parametric surface, the basic
syntax of this command is
ParametricPlot3D[{x[u,v],y[u,v],z[u,v]}, {u,u1,u2}, {v,v1,v2}]
Alternatively, if the surface is described in the vector form (7.2) and we have defined the vector
function r(u, v) in Mathematica as r[u,v], we can use the form
ParametricPlot3D[r[u,v], {u,u1,u2}, {v,v1,v2}]
Below we give examples of both forms of ParametricPlot3D. Both commands plot the surface
with parametrization r(u, v) = u cos ve1 + u sin ve2 + (v/2)e3 , where 0 ≤ u ≤ 1, 0 ≤ v ≤ 2π.
ParametricPlot3D[{u*Cos[v], u*Sin[v], v/2}, {u,0,1}, {v,0,2Pi},
PlotStyle->Red]

r[u_,v_] := {u*Cos[v], u*Sin[v], v/2}

ParametricPlot3D[r[u,v], {u,0,1}, {v,0,2Pi}, Mesh->None]
The above examples illustrate that ContourPlot3D and ParametricPlot3D accept some op-
tions that resemble those for Plot and Plot3D. This is true for most plotting options and the best
way to learn about such matters is to take a look at Mathematica’s help and the examples there.
7.5.4. Non-rectangular domains: RegionFunction. Sometimes, one wants to plot a func-
tion or a parametric surface with a domain other than a rectangular box. Or, perhaps, one wants to
plot a part of a level surface, but that part is restricted by a shape other than a box (for example, an
ellipsoid). In such situation, one can use the option RegionFunction of the above plotting com-
mands to restrict the variables to regions other than rectangular boxes. For example, the following
command restricts the plot of the function z = x2 − y2 to the interior of the ellipse x2 + y2 /4 ≤ 1:
Plot3D[xˆ2-yˆ2, {x,-3,3}, {y,-3,3},
RegionFunction->Function[{x,y,z}, xˆ2+yˆ2/4<=1],
PlotStyle->Red, Mesh->None]
58
Note that the value of RegionFunction is set to Function[{x,y,z}, ...] where in place of
the dots we have included the restriction on the variables. That restriction can involve double
inequalities or even several simultaneous inequalities (see the next example).
Here is another example of use of RegionFunction, this time inside a ContourPlot3D com-
mand:
ContourPlot3D[xˆ2+2yˆ2-zˆ2==1, {x,-3,3}, {y,-3,3}, {z,-3,3},
RegionFunction->Function[{x,y,z}, 1<xˆ2+yˆ2<4 && x+z>-2], Mesh->None]
This command plots the part of the hyporboloid x2 + 2y2 − z2 = 1 that lies between the cylinders
x2 + y2 = 1 and x2 + y2 = 4 and above the plane x + z = −2.
The usage of RegionFunction to restrict the parameters of a parametric surface is similar, but
somewhat more subtle: the region function generally takes the form
Function[{x,y,z,u,v}, ...]
with a restriction on the parameters u and v only. (Though it is possible to include also restrictions
on x, y and z, which would have the same effect as described above.) The dynamic command below
plots the parametric surface
r(u, v) = u cos ve1 + u sin ve2 + (v/2)e3 ,
where 0 ≤ u ≤ 1, 0 ≤ v ≤ 2π and u2 + v2 /4 ≤ a, with a allowed to change between 0.1 and 10.
Manipulate[
ParametricPlot3D[{u*Cos[v], u*Sin[v], v/2}, {u,0,1}, {v,0,2Pi},
RegionFunction->Function[{x,y,z,u,v}, uˆ2+vˆ2/4<=a],
PlotRange->{{-1,1}, {-1,1}, {0,Pi}}, Mesh->None],
{a,1/10,10}]
It is instructive to execute it and experiment with various values of a in order to see the effect on
the plot from changing the restrictions on the parameters.
Exercises
Find the traces of the given quadric surface in the planes x = k, y = k, and z = k. Use those traces and Table 7.1 to
classify the surface.
7.1. z = x2 − y2 7.2. 4y2 − x2 − z2 = 1 7.3. y2 + 4z2 = 5 − x2 7.4. y2 + 4z2 = 5 + x2

Rewrite the equation of the given quadric surface in a standard form and use the result and Table 7.1 to classify the
surface.
7.5. 3x + 6y2 − 3z2 = 9 7.6. x2 + 4y2 + z2 − 4x + 2z = 4

7.7. Parametrize the curve of intersection of the cylinder x2 + y2 = 9 and the plane z = x + y + 2.
Identify the given parametric surface.
7.8. r(u, v) = u cos ve1 + u sin ve2 + u2 e3 7.9. r(θ, z) = 2 cos θe1 + sin θe2 + ze3
x−1 y+1 z x+2 z−4
7.10. Parametrize the plane that contains the lines = = and =y= .
2 3 −2 −3 4
7.11. Parametrize the half of the hyperboloid x2 − 4y2 − 4z2 = 16 that lies in the half-space x ≤ 0.
7.12. Parametrize the part of the hemisphere x2 + y2 + z2 = 5, z ≥ 0 that lies inside the cylinder x2 + y2 = 1.
7.13. Parametrize the part of the paraboloid z = 4 − x2 − y2 that lies above the xy-plane by using:
(a) x and y as parameters; (b) the polar coordinates in the plane as parameters.
7.14. Let 0 < a < b. A torus of radii a and b is the surface of revolution obtained by rotating about z-axis the circle
(x − b)2 + z2 = a2 in the xz-plane. Parametrize the torus of radii a and b.

59
 LECTURE 8

Partial Derivatives

In this lecture, we introduce the derivatives of multivariable functions—the so-called “partial

derivatives.”

8.1. First-order partial derivatives

Definition. If f is a function of two variables, its partial derivatives f x and fy at (a, b) are
defined by
∂f f (a + h, b) − f (a, b)
f x (a, b) = (a, b) = ∂ x f (a, b) = lim
∂x h→0 h
and
∂f f (a, b + h) − f (a, b)
fy (a, b) = (a, b) = ∂y f (a, b) = lim .
∂y h→0 h
If f is a function of n variables x1 , x2 , . . . , xn , its partial derivatives f x1 , f x2 , . . . , f xn at (a1 , a2 , . . . , an )
are defined by
∂f
f x j (a1 , a2 , . . . , an ) = (a1 , a2 , . . . , an ) = ∂ x j f (a1 , a2 , . . . , an )
∂x j
f (a1 , . . . , a j + h, . . . , an ) − f (a1 , . . . , a j , . . . , an )
= lim ( j = 1, 2, . . . , n).
h→0 h
To illustrate the meaning of the partial derivatives, we take a closer look at the two-dimensional
case. Given a function f (x, y) and a point (a, b) in its domain, we obtain two (single-variable)
functions
F1 (x) = f (x, b) and F2 (y) = f (a, y),
by fixing, respectively, the variables y and x in f (x, y). The points on the surface z = f (x, y) with
y = b lie on the curve of intersection of that surface and the vertical plane y = b (see the blue curve

Figure 8.1. The partial derivatives as slopes

61
on Figure 8.1). That curve has exactly the same shape as the graph of the function F1 (x) in the
plane. Likewise, the graph of F2 (y) can be visualized as the curve of intersection of the surface
z = f (x, y) and the vertical plane x = a (see the red curve on Figure 8.1). Using the functions F1 (x)
and F2 (y), we can express the partial derivatives of f (x, y) at (a, b) as
f (a + h, b) − f (a, b) F1 (a + h) − F1 (a)
f x (a, b) = lim = lim = F10 (a) (8.1)
h→0 h h→0 h
and
f (a, b + h) − f (a, b) F2 (b + h) − F2 (b)
fy (a, b) = lim = lim = F20 (b). (8.2)
h→0 h h→0 h
That is, the partial derivatives f x and fy of f are the derivatives of the functions F1 (x) and F2 (y).
Recall that F10 (a) and F20 (b) are the slopes of the graphs of those functions at x = a and y = b,
respectively. Looking at the copies of those graphs on the surface z = f (x, y) (the blue and red
curves on Figure 8.1), we see that:
The partial derivatives f x (a, b) and fy (a, b) represent the slopes at (a, b) of the
sections of the graph z = f (x, y) with the vertical planes y = b and x = a,
respectively. That is, f x (a, b) and fy (a, b) represent the rates of change of f at
(a, b) in the directions of the x- and the y-axes, respectively.
Furthermore, equations (8.1) and (8.2) can be extended to functions of more than two variables.
In that more general form, they lead to the following practical observation.
To compute the partial derivative ∂ x j f (x1 , . . . , xn ) with respect to one of the vari-
ables, one can treat all other variables as constants and differentiate the result-
ing single-variable function F(x j ) using the familiar rules from single-variable
calculus.
The following example demonstrates how we use this simple observation to calculation of partial
derivatives.

Example 8.1. Find the first-order partials of the functions

x
f (x, y) = , g(x, y, z) = e xy sin xz.
x2 + y2
Solution. The partial derivatives of f are
1 · (x2 + y2 ) − x · (2x) y2 − x2
f x (x, y) = =
(x2 + y2 )2 (x2 + y2 )2
and
−(2y) −2xy
fy (x, y) = x · = .
(x2 + y2 )2 (x2 + y2 )2
The partials of g are

g x (x, y, z) = ye xy sin xz + e xy (z cos xz),

gy (x, y, z) = xe xy sin xz, gz (x, y, z) = e xy (x cos xz). 
62
 8.2. Higher-order partial derivatives
Just as we can differentiate the derivative f 0 (x) of a single-variable function to obtain its sec-
ond derivative, we can differentiate the partial derivatives of a multivariable function to obtain its
higher-order partials. For example, a function f of two variables has four second-order partial
derivatives f xx , f xy , fyx , and fyy :
∂2 f ∂ ∂f ∂2 f ∂ ∂f
   
f xx = 2 = , f xy = = ,
∂x ∂x ∂x ∂y∂x ∂y ∂x
∂2 f ∂ ∂f ∂2 f ∂ ∂f
   
fyx = = , fyy = 2 = .
∂x∂y ∂x ∂y ∂y ∂y ∂y
Example 8.2. Find the second-order partials of the function f (x, y) = x2 cos y + y2 sin x.
Solution. We have
f x (x, y) = 2x cos y + y2 cos x, fy (x, y) = −x2 sin y + 2y sin x.
The second-order partials of f then are:
f xx = 2 cos y − y2 sin x, f xy = −2x sin y + 2y cos x,
fyx = −2x sin y + 2y cos x, fyy = −x2 cos y + 2 sin x. 
Remark. Notice that the second-order partials of the function in the example satisfy f xy = fyx .
In other words, for this function it is not important whether we differentiate first with respect to x
and then with respect to y or vice versa. One might ask whether this is always true. The answer is:
“only for ‘nice’ functions”. The following theorem gives a necessary condition for this to hold.
Theorem 8.1. Suppose that f is a function of two variables defined near (a, b) and having
second-order partials f xy and fyx that are continuous at (a, b). Then
f xy (a, b) = fyx (a, b).
The following example illustrates the usefulness of the above theorem.
Example 8.3. Find the second-order partials of
f (x, y, z) = sin(x2 + 2y2 − z).
Solution. The first-order partials of f are:
f x (x, y, z) = 2x cos(x2 + 2y2 − z),
fy (x, y, z) = 4y cos(x2 + 2y2 − z),
fz (x, y, z) = − cos(x2 + 2y2 − z).
There are nine second-order partials, but we really have to compute only six of them. We have
f xx (x, y, z) = 2 cos(x2 + 2y2 − z) + 2x − 2x sin(x2 + 2y2 − z) ,

fyy (x, y, z) = 4 cos(x2 + 2y2 − z) + 4y − 4y sin(x2 + 2y2 − z) ,

fzz (x, y, z) = − sin(x2 + 2y2 − z).

63
For the mixed partials, we have
f xy (x, y, z) = 2x − 4y sin(x2 + 2y2 − z) = −8xy sin(x2 + 2y2 − z),

fzy (x, y, z) = − − 4y sin(x2 + 2y2 − z) = 4y sin(x2 + 2y2 − z),

fzx (x, y, z) = − − 2x sin(x2 + 2y2 − z) = 2x sin(x2 + 2y2 − z).

Note that the last three functions are continuous (they are products of polynomials in x, y, z and
compositions of sine functions and polynomials in x, y, z). Furthermore, it is clear that the partials
fyx , f xz , and fyz will also be continuous. But then the theorem implies that, in fact, we have
fyx = f xy , f xz = fzx , fyz = fzy . 
Example 8.4. Find f xy (0, 0) and fyx (0, 0) for
2 2

 xy(y − x ) if (x, y) , (0, 0),
f (x, y) = x2 + y2
0 if (x, y) = (0, 0).


Solution. The first order partials of f are:

 5 4 2 3
 y − x y − 4x y
if (x, y) , (0, 0),
f x (x, y) = (x2 + y2 )2
0 if (x, y) = (0, 0),


and
 xy + 4x y − x
 4 3 2 5
if (x, y) , (0, 0),
fy (x, y) = (x2 + y2 )2
0 if (x, y) = (0, 0).


These formulas are straightforward everywhere except at (x, y) = (0, 0), where we use the definition
of partial derivatives:
f (h, 0) − f (0, 0) 0−0
f x (0, 0) = lim = lim =0
h→0 h h→0 h
and
f (0, h) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
h→0 h h→0 h
Then
f x (0, h) − f x (0, 0) h−0
f xy (0, 0) = lim = lim =1
h→0 h h→0 h
and
fy (h, 0) − fy (0, 0) −h − 0
fyx (0, 0) = lim = lim = −1,
h→0 h h→0 h
that is, f fails to satisfy f xy (0, 0) = fyx (0, 0). On the other hand, when (x, y) , (0, 0), we have
y6 + 9x2 y4 − 9x4 y2 − x6
f xy (x, y) = fyx (x, y) = .
(x2 + y2 )3
This is not surprising, considering that the function
y6 + 9x2 y4 − 9x4 y2 − x6
(x2 + y2 )3
64
is continuous everywhere except at (0, 0) (since it is a rational function in x and y, whose denomi-
nator vanishes only at (0, 0)). 
8.3. Tangent planes
Recall that the equation of the tangent line to the curve y = f (x) at a point P(x0 , y0 ) on the
curve is given by
y − y0 = f 0 (x0 )(x − x0 ).
The following theorem is the analog of this formula for functions of two variables. It describes
the equation of the tangent plane to the surface z = f (x, y) at a point P(x0 , y0 , z0 ) on that surface in
terms of the partials of f . Strictly speaking, we should first define what a tangent plane is, but we
will leave that for Lecture #10. The same applies to the proof of the theorem.
Theorem 8.2. Suppose that f has continuous partial derivatives at (x0 , y0 ). Then the equation
of the tangent plane to the surface z = f (x, y) at the point P(x0 , y0 , z0 ) is
z − z0 = f x (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ).
Example 8.5. Find the tangent plane to the surface z = e x −y at (−1, 0, e).
2 2

Solution. First, observe that the point is indeed on the surface:

2 2
e = e(−1) −0 .
The partials of the given function are
∂z 2 2 ∂z 2 2
= 2xe x −y , = −2ye x −y ,
∂x ∂y
whence
∂z ∂z
(−1, 0) = −2e, (−1, 0) = 0.
∂x ∂y
Applying the theorem, we find that the equation of the tangent plane is
z − e = −2e(x − (−1)) + 0(y − 0) ⇐⇒ 2ex + z + e = 0. 
8.4. Linear approximations and differentiable functions*
Recall that for a single-variable function f which has a derivative f 0 (a) at x = a (we called
such functions “differentiable”), the tangent line at (a, f (a)) always exists and has equation
y = f (a) + f 0 (a)(x − a).
That is, the linear function f (a) + f 0 (a)(x − a) provides a very good approximation to f (x) when x
is near a:
f (x) ≈ f (a) + f 0 (a)(x − a).
When we deal with functions of two variables, we may encounter functions f (x, y) for which both
partial derivatives f x (a, b) and fy (a, b) exist, but the plane
z = f (a, b) + f x (a, b)(x − a) + fy (a, b)(y − b)
is not tangent to the surface z = f (x, y) and does not provide a good approximation to f (x, y) for
(x, y) near (a, b). Thus, in the case of functions of two or more variables, we reserve the term
“differentiable function” for functions for which this type of approximation (known as the linear
approximation to f at (a, b)) does hold.
65
 Definition. A function f of two variables is called differentiable at (a, b), if near (a, b) we have
f (x, y) = f (a, b) + f x (a, b)(x − a) + fy (a, b)(y − b) + ε1 (x, y)(x − a) + ε2 (x, y)(y − b),
where
lim ε1 (x, y) = lim ε2 (x, y) = 0.
(x,y)→(a,b) (x,y)→(a,b)

Similarly, a function f of three variables is called differentiable at (a, b, c), if near (a, b, c) we have
f (x, y, z) = f (a, b, c) + f x (a, b, c)(x − a) + fy (a, b, c)(y − b) + fz (a, b, c)(z − c)
+ ε1 (x, y, z)(x − a) + ε2 (x, y, z)(y − b) + ε3 (x, y, z)(z − c),
where
lim ε1 (x, y, z) = lim ε2 (x, y, z) = lim ε3 (x, y, z) = 0.
(x,y,z)→(a,b,c) (x,y,z)→(a,b,c) (x,y,z)→(a,b,c)

Example 8.6. Show that the function f (x, y) = x3 + xy − 2y3 is differentiable at (1, −1).
Solution. We have
f (1, −1) = 2, f x (1, −1) = 2, fy (1, −1) = −5.
Hence, we are looking for functions 1 (x, y) and 2 (x, y) such that
f (x, y) = 2 + 2(x − 1) − 5(y + 1) + 1 (x, y)(x − 1) + 2 (x, y)(y + 1),
lim 1 (x, y) = lim 2 (x, y) = 0.
(x,y)→(1,−1) (x,y)→(1,−1)

To find such functions 1 (x, y) and 2 (x, y), we will transform the expression
f (x, y) − 2 − 2(x − 1) + 5(y + 1).
Similarly to Example 6.10, it is convenient to rewrite this expression in terms of the variables
u = x − 1 and v = y + 1. We have
f (u + 1, v − 1) − 2 − 2u + 5v = u3 − 2v3 + 3u2 + uv + 6v2
= (u2 + 3u + v)u + (−2v2 + 6v)v.
Returning back to the original variables x and y, we now obtain
f (x, y) − 2 − 2(x − 1) + 5(y + 1) = 1 (x, y)(x − 1) + 2 (x, y)(y + 1),
where
1 (x, y) = (x − 1)2 + 3(x − 1) + (y + 1), 2 (x, y) = −2(y + 1)2 + 6(y + 1).
Thus, we have now found functions 1 (x, y) and 2 (x, y) satisfying (8.4). Moreover, it is easy to see
that these functions satisfy also (8.4). Therefore, the function f (x, y) is differentiable at (1, −1). 
Next, we state an alternative definition of differentiability, which is sometimes more convenient
to use. For example, this definition is easier to apply when one aims to show that a given function
is not differentiable (see Exercises 8.30 and 8.31).
Definition. A function f of two variables is called differentiable at (a, b), if
f (x, y) − f (a, b) − f x (a, b)(x − a) − fy (a, b)(y − b)
lim = 0.
(x − a)2 + (y − b)2
p
(x,y)→(a,b)

66
Similarly, a function f of three variables is called differentiable at (a, b, c), if
f (x, y, z) − f (a, b, c) − f x (a, b, c)(x − a) − fy (a, b, c)(y − b) − fz (a, b, c)(z − c)
lim = 0.
(x − a)2 + (y − b)2 + (z − c)2
p
(x,y,z)→(a,b,c)

As we mentioned above (and as Exercises 8.30 and 8.31 demonstrate), the existence of partial
derivatives does not guarantee the differentiability of f . It is therefore natural to ask what condition
does? It turns out that the existence of continuous partials suffices. For the sake of simplicity, we
state the relevant theorem only for functions of two variables.
Theorem 8.3. If the partial derivatives f x and fy exist near (a, b) and are continuous at (a, b),
then f is differentiable at (a, b).
8.5. The chain rule
Recall that the chain rule in single-variable calculus gives a formula for the derivative of a
composite function: if y = f (x(t)), then
dy d f dx
= .
dt dx dt
In this section, we discuss the generalizations of the chain rule to functions of two and more
variables. Our first result concerns composite functions of the form F(t) = f (x1 (t), . . . , xn (t)),
where f is a differentiable function of n variables and x j = x j (t), 1 ≤ j ≤ n, are differentiable
single-variable functions.
Theorem 8.4. Let f be a differentiable function of n variables and let x1 (t), . . . , xn (t) be differ-
entiable single-variable functions. Then z = f (x1 (t), . . . , xn (t)) is a differentiable function of t, and
its derivative is
dz ∂ f dx1 ∂ f dxn
= + ··· + .
dt ∂x1 dt ∂xn dt
Remark. Take a note that we use “partial d’s,” ∂x∂ j , to denote derivatives of functions of two
variables, but we use “straight d’s,” dtd , to denote derivatives of functions of a single variable. It
is important to follow this convention when dealing with a mixture of multivariable and single-
variable functions.
Example 8.7. Let z = f (x, y) = x ln(x + 2y), x = cos t, and y = sin t. Find z0 (0).
Solution. The partials of f are
x 2x
f x (x, y) = ln(x + 2y) + , fy (x, y) = ,
x + 2y x + 2y
and the derivatives of x(t) and y(t) are
x0 (t) = − sin t, y0 (t) = cos t.
Hence, by the above theorem, the derivative z0 (0) of the composite function z = f (x(t), y(t)) at t = 0
is
dz ∂z dx ∂z dy
(0) = (1, 0) (0) + (1, 0) (0)
dt ∂x dt ∂y dt
   
1 2
= ln(1 + 0) + (−0) + (1) = 2.
1+0 1+0
67
In general, we have
   
dz cos t 2 cos t
= ln(cos(t) + 2 sin t) + (− sin t) + (cos t).
dt cos t + 2 sin t cos t + 2 sin t

We now state the general version of the chain rule for compositions of multivariable functions.
Theorem 8.5. Let f be a differentiable function of n variables and let each x j = x j (t1 , . . . , tm ),
1 ≤ j ≤ n, be a differentiable function of (the same) m variables. Then
u = f (x1 (t1 , . . . , tm ), . . . , xn (t1 , . . . , tm ))
is a differentiable function of t1 , . . . , tm , with partial derivatives
∂u ∂ f ∂x1 ∂ f ∂xn
= + ··· + (i = 1, . . . , m).
∂ti ∂x1 ∂ti ∂xn ∂ti
Example 8.8. Let z = x2 − xy, x = s sin t, and y = t sin s. Find ∂z/∂s and ∂z/∂t.
Solution. We have
z x (x, y) = 2x − y, zy (x, y) = −x,
and
x s (s, t) = sin t, xt (s, t) = s cos t, y s (s, t) = t cos s, yt (s, t) = sin s.
Hence, the partials of the composite function z(x(s, t), y(s, t)) are
z s (s, t) = z x (x(s, t), y(s, t))x s (s, t) + zy (x(s, t), y(s, t))y s (s, t)
= (2s sin t − t sin s) sin s + (−s sin t)(t cos s)
and
zt (s, t) = z x (x(s, t), y(s, t))xt (s, t) + zy (x(s, t), y(s, t))t s (s, t)
= (2s sin t − t sin s)(s cos t) + (−s sin t) sin s. 
Remark. Often, we do not substitute x(s, t) and y(s, t) in the formulas for the partials of z(s, t).
In such cases, the answer in the above example would take the form
z s (s, t) = (2x − y) sin s − xt cos s, zt (s, t) = (2x − y)s cos t − x sin s.
Example 8.9. Find the partial derivatives of w(x, y, t), where
√
w = u2 + v2 , u = y + x cos t, v = x + y sin t.
Solution. The partial derivatives of these functions are:
∂w u ∂w v
= √ , = √ ,
∂u u2 + v2 ∂v u2 + v2
∂u ∂u ∂u
= cos t, = 1, = −x sin t,
∂x ∂y ∂t
∂v ∂v ∂v
= 1, = sin t, = x cos t.
∂x ∂y ∂t
68
Hence, the partial derivatives of the composite function w(x, y, t) are:
∂w ∂w ∂u ∂w ∂v u cos t v u cos t + v
= + = √ + √ = √ ,
∂x ∂u ∂x ∂v ∂x u2 + v2 u2 + v2 u2 + v2
∂w ∂w ∂u ∂w ∂v u v sin t u + v sin t
= + = √ + √ = √ ,
∂y ∂u ∂y ∂v ∂y u2 + v2 u2 + v2 u2 + v2
∂w ∂w ∂u ∂w ∂v −ux sin t vx cos t −ux sin t + vx cos t
= + = √ + √ = √ .
∂t ∂u ∂t ∂v ∂t u2 + v2 u2 + v2 u2 + v2

Example 8.10. Find the partial derivative wyx of w(x, y, t), where
√
w = u2 + v2 , u = y + x cos t, v = x + y sin t.
Solution. To find a higher-order partial derivative of a composite function, we “mix” the chain
rule with the other rules for partial differentiation. We have
∂2 w ∂ ∂w ∂ u + v sin t
   
= = √
∂x∂y ∂x ∂y ∂x u2 + v2
∂ u + v sin t ∂u ∂ u + v sin t ∂v
   
= √ + √ .
∂u u2 + v2 ∂x ∂v u2 + v2 ∂x
We now compute the partials with respect to u and v:
∂ u + v sin t ∂
 
= (u + v sin t)(u2 + v2 )−1/2


√
∂u u2 + v2 ∂u
= (u2 + v2 )−1/2 + (u + v sin t)(−u)(u2 + v2 )−3/2
 v2 − uv sin t
= (u2 + v2 )−3/2 (u2 + v2 ) − u(u + v sin t) = 2 ,

(u + v2 )3/2
∂ u + v sin t ∂
 
= (u + v sin t)(u2 + v2 )−1/2


√
∂v u2 + v2 ∂v
= (sin t)(u2 + v2 )−1/2 + (u + v sin t)(−v)(u2 + v2 )−3/2
 u2 sin t − uv
= (u2 + v2 )−3/2 (u2 + v2 ) sin t − v(u + v sin t) = 2 .

(u + v2 )3/2
Substituting these partials into the expression for wyx , we get
∂2 w ∂ u + v sin t ∂u ∂ u + v sin t ∂v
   
= √ + √
∂x∂y ∂u u2 + v2 ∂x ∂v u2 + v2 ∂x
2 2
v − uv sin t u sin t − uv
= 2 · cos t + 2 ·1
(u + v ) 2 3/2 (u + v2 )3/2
u2 sin t − uv(1 + sin t cos t) + v2 cos t
= .
(u2 + v2 )3/2

69
 8.6. Partial derivatives in Mathematica
Recall that if f (x) is a single-variable function, we can calculate its derivative in Mathematica
using either of the following two commands:
f’[x]
D[f[x],x]
The D command that appears in the latter option is very versatile and is used also to calculate partial
derivatives. Its basic form for multivariable functions is
D[f[x1 , x2 , . . . , xn ], xi , x j , . . . ]
For a function f (x1 , . . . , xn ), this command will calculate the partial derivative f xi ,x j ,... (x1 , . . . , xn ).
Here are some examples:
D[f[x,y,z],x] f x (x, y, z)
D[f[x,y,z],x,z] f xz (x, y, z)
D[f[x,y],x,x,y] f xxy (x, y)
D[f[x,u,v],v,v,x,u] fvvxu (x, u, v)
The D command has also two other forms that provides useful shortcuts for special kinds of
partial derivatives. If f is a function of several variables, one of which is x, then
D[f[x, . . . ,], x, n]
is a quick way to obtain the n-th order partial ∂n f /∂xn . Furthermore, if f is a multivariable function,
the command
D[f[x1 , x2 , . . . , xn ], x1 , x2 , . . . , xn ]
will calculate its gradient ∇ f (x1 , . . . , xn ) (see Lecture #9).
There is one use of the D command that requires extra care. Suppose that we have a function
f (x, y) = x2 − y2 and we want to define a new function g by g(x, y) = f x (x, y) + fy (x, y). It may feel
natural to use the following Mathematica commands:
f[x_,y_] := xˆ2 - yˆ2
g[x_,y_] := D[f[x,y],x] + D[f[x,y],y]
However, the function g defined this way behaves rather bizarrely. For instance, asking Mathemat-
ica to execute simple commands like
g[0,0]
Plot3D[g[x,y], {x,-1,1}, {y,-1,1}]
results in cryptic error messages like “General::ivar: 0 is not a valid variable.” The proper way to
define g in this case is to use “=” instead of “:=” in the above code
g[x_,y_] = D[f[x,y],x] + D[f[x,y],y]
Try the above commands with g defined this way.

Exercises
Find the first-order partial derivatives of the given function.
√
8.1. f (x, y) = x4 − 4x2 y3/2 8.3. f (x, y, z) = ln(3x + 7yz)
8.5. f (x, z) = x2 + z2 + 1
x
8.2. f (x, y, z) = e2xy sin xz 8.4. f (x, y) = x2 + xy + y2 8.6. f (r, θ) = r tan2 θ

70
Find the second-order partial derivatives of the given function. √
8.7. f (x, y) = x4 − 4x2 y3/2 8.8. f (x, y) = cos x2 + xy + y2 8.9. f (x, z) = x2 + z2 + 1

Find the indicated partial derivative of the given function.

8.10. f (x, y) = x4 − 4x2 y3/2 ; f xyx 8.12. f (x, y) = cos x2 + xy + y2 ; f xxy


√
8.11. f (x, z) = x2 + z2 + 1; ∂3 f /∂x∂2 z 8.13. f (x, y, z) = xz2 e2yz ; f xyz , fzzy

Use the chain rule to calculate dF/dt for the given function F.
8.14. F(t) = f (x(t), y(t)), where f (x, y) = 2xy + y3 , x(t) = t2 + 3t + 4, y(t) = t2
8.15. F(t) = f (x(t), y(t)), where f (x, y) = xye x +4y , x(t) = 2 cos t, y(t) = sin t
2 2

8.16. F(t) = f (x(t), y(t), z(t)), where f (x, y, z) = x2 + y2 − 3xyz, x(t) = t sin t, y(t) = t cos t, z(t) = 12 t2
8.17. F(t) = f (x(t), y(t), z(t)), where f (x, y, z) = ln(xy + 2yz − 3xz), x(t) = 2et , y(t) = 3e−t , z(t) = e2t

Use the chain rule to calculate ∂F/∂s and ∂F/∂t for the given function F.
8.18. F(s, t) = f (x(s, t), y(s, t)), where f (x, y) = 2x2 y − xy2 , x(s,
t) = t2 + 3s2 , y(s, t) = 3st
8.19. F(s, t) = f (x(s, t), y(s, t)), where f (x, y) = arctan x2 − 4y2 , x(s, t) = 2se
t + 2se−t , y(s, t) = te s − te−s
8.20. F(s, t) = f (x(s, t), y(s, t), z(s, t)), where f (x, y, z) = ln x2 + 2y2 + 3z2 , x(s, t) = s cos t, y(s, t) = s sin t,
z(s, t) = t2
8.21. F(s, t) = f (x(s, t), y(s, t), z(s, t)), where f (x, y, z) = cos xy + y3 x + z2 , x(s, t) = s2 + t, y(s, t) = s − t2 ,
z(s, t) = st + s + t

Use the chain rule to calculate the given partial derivative at the given point.
8.22. gr (r, θ) at (r, θ) = (2, π), where g(x, y) = xy2 − x2 y, x = r cos θ, y = r sin θ
2 2 2
8.23. hu (u, v) at (u, v) = (2, −3), where h(x, y, z) = e x −y −z , x = 2u + v, y = u − v, z = u + v + 1
8.24. ∂R/∂x at (x, y) = (1, 1), where R(u, v, w) = ln(u + v2 + w2 ), u = x + 2y, v = 2x − y, w = 2xy
2

8.25. Show that the function u(x, y) = aey cos x + bey sin x is a solution of the differential equation u xx + uyy = 0.
8.26. Show that any function u(x, t) of the form
u(x, t) = f (x + at) + g(x − at)
is a solution of the differential equation utt = a2 u xx .
Find the linear approximation to the given differentiable function near the given point.
8.27. f (x, y) = ln(2xy + x + y), (2, 1) 8.28. f (x, y, z) = arctan(xy + xz + 2yz), (3, 2, 1)

8.29. Show that the function f (x, y) = xy is differentiable at (1, 2) by finding functions 1 (x, y) and 2 (x, y) that satisfy
the definition of differentiability in §8.4.
8.30. (a) Calculate f x (0, 0) and fy (0, 0) for the function
 xy

if (x, y) , (0, 0),
f (x, y) = x2 + y2
0 if (x, y) = (0, 0).
(b) Show that f (x, y) is not continuous at (0, 0).
(c) Show that f (x, y) is not differentiable at (0, 0).
8.31. Let a, b, c, and d be real numbers.

(a) Calculate f x (0, 0) and fy (0, 0) for the function

 ax + bx y + cxy + dy
 3 2 2 3
if (x, y) , (0, 0),
f (x, y) = x 2 + y2
0 if (x, y) = (0, 0).


(b) Show that f (x, y) is continuous at (0, 0).

(c) Show that if a , c or b , d, then f (x, y) is not differentiable at (0, 0).

71
 LECTURE 9

Directional Derivatives and Gradients

9.1. Directional derivatives

Given a function of two variables f (x, y), we know how to compute its rate of change in the
x-direction and in the y-direction: the rate of change in the x-direction is given by the partial
derivative with respect to x,
∂f f (x + h, y) − f (x, y)
f x (x, y) = (x, y) = lim ,
∂x h→0 h
and the rate of change in the y-direction is given by the partial derivative with respect to y,
∂f f (x, y + h) − f (x, y)
fy (x, y) = (x, y) = lim .
∂y h→0 h
A natural question then is how to compute the rate of change of f in other directions. This question
leads to the notion of a “directional derivative.”
Definition. If f is a function of two variables and if u = ha, bi is a unit vector, the directional
derivative of f in the u-direction is
f (x + ha, y + hb) − f (x, y)
Du f (x, y) = lim , (9.1)
h→0 h
if the limit exists. If f is a function of three variables and if u = ha, b, ci is a unit vector, the
directional derivative of f in the u-direction is
f (x + ha, y + hb, z + hc) − f (x, y, z)
Du f (x, y, z) = lim . (9.2)
h→0 h
Remark. Note that if we introduce the function g(t) = f (x + ta, y + tb), the limit on the right
side of (9.1) is also
g(h) − g(0)
lim = g0 (0);
h→0 h
that is, Du f (x, y) = g0 (0). On the other hand, by the chain rule,
g0 (t) = f x (x + ta, y + tb) · a + fy (x + ta, y + tb) · b,
g0 (0) = f x (x, y)a + fy (x, y)b.
This is (essentially) how one proves the following result.
Theorem 9.1. If f is a differentiable function of two variables and u = ha, bi is a unit vector,
then
Du f (x, y) = f x (x, y)a + fy (x, y)b.
If f is a differentiable function of three variables and u = ha, b, ci is a unit vector, then
Du f (x, y, z) = f x (x, y, z)a + fy (x, y, z)b + fz (x, y, z)c.
73
 2x
Example 9.1. Find the directional derivative of f (x, y) = at the point (1, 0) in the direction
√ x−y
of v = e1 − 3e2 .
Solution. First, notice that v is not unit and that the unit vector in the direction of v is
√  √ 
v h1, − 3i 1 3
u= = √ = ,− .
kvk 1+3 2 2
The partials of f are
2(x − y) − 2x −2y 2x
f x (x, y) = 2
= , fy (x, y) = ,
(x − y) (x − y)2 (x − y)2
so the theorem gives
√ √
1 −2y − 3 2x −y − 3x √
Du f (x, y) = + = , Du f (1, 0) = − 3.
2 (x − y)2 2 (x − y)2 (x − y)2

x
Example 9.2. Find the directional derivative of f (x, y, z) = at the point (2, 1, 1) in the
y+z
direction of v = h1, 2, 3i.
Solution. The unit vector in the direction of v is
 
v h1, 2, 3i 1 2 3
u= = √ = √ , √ , √ ,
kvk 1+4+9 14 14 14
and the partial derivatives of f are
1 −x −x
f x (x, y, z) = , fy (x, y, z) = , fz (x, y, z) = .
y+z (y + z) 2 (y + z)2
Hence,
1 1 1 2 1 3 −2
Du f (2, 1, 1) = √ − √ − √ = √ .
2 14 2 14 2 14 14


9.2. The gradient vector

Definition. If f is a function of two variables x and y, its gradient, denoted ∇ f or grad f , is the
vector function
∂f ∂f
∇ f = h f x (x, y), fy (x, y)i = e1 + e2 .
∂x ∂y
If f is a function of three variables x, y, and z, its gradient, ∇ f or grad f , is
∂f ∂f ∂f
∇ f = h f x (x, y, z), fy (x, y, z), fz (x, y, z)i = e1 + e2 + e3 .
∂x ∂y ∂z
Remark. Note that using the gradient vector, we can restate both formulas in Theorem 9.1 in
the form
Du f (x) = ∇ f (x) · u, (9.3)
where x = (x, y) or x = (x, y, z).
74
 Theorem 9.2. Let f be a differentiable function of two or three variables. The maximum value
of the directional derivative Du f (x) is k∇ f (x)k; it occurs when u has the same direction as ∇ f (x).
Proof. By (9.3) and the properties of the dot product,
Du f (x) = ∇ f (x) · u = k∇ f (x)k kuk cos θ = k∇ f (x)k cos θ,
where θ is the angle between ∇ f (x) and u. Since cos θ ≤ 1, with equality when θ = 0 (i.e., when
∇ f (x) and u point in the same direction), we find that
Du f (x) ≤ k∇ f (x)k,
with an equality when u points in the direction of the gradient. 
Example 9.3. By Example 9.2, the gradient of the function
x
f (x, y, z) =
y+z
is  
1 −x −x
∇ f (x, y, z) = , , .
y + z (y + z)2 (y + z)2
Hence, the largest directional derivative at (2, 1, 1) occurs in the direction of
 
1 −1 −1
∇ f (2, 1, 1) = , , .
2 2 2
√
Its value is k∇ f (2, 1, 1)k = 2
3
. 
Exercises
Find the directional derivative of the given function in the given direction.
9.1. f (x, y) = x2 + y2 e2xy
, in the direction of h1, 2i

9.2. f (x, y, z) = ln xy + z2 , in the direction of e2 − 2e3

9.3. Find the directional derivative of F(x, y, z) = 4x2 − y2 − z2 at P(2, −1, 3) in the direction of v = h−1, −2, 2i.
p

9.4. Find the directional derivative of

 
2x
G(x, y, z) = arcsin
4x + y2 + z2
2

at the point P(1, 2, 0) in the direction of the point Q(−1, 3, 1).

Find the maximum rate of change of f at the given point and the direction in which it occurs.
9.5. f (x, y) = tan(xy + x + y), P(2, −1) 9.6. f (x, y, z) = xy + yz2 , Q(1, −3, 1)

75
 LECTURE 10

Tangent Planes and Normal Vectors

10.1. Tangent planes and normal vectors to level surfaces

We now fulfill a promise made in Lecture #8. We start by defining the tangent plane to a level
surface.
Definition. Let F(x, y, z) be a differentiable function, and let P(x0 , y0 , z0 ) be a point on the
surface Σ with equation F(x, y, z) = k. The tangent plane to Σ at P is the plane, if it exists, that
contains P and the tangent vectors to any smooth curve γ on Σ that passes through P.
The following theorem establishes a formula for the equation of the tangent plane to a level
surface.
Theorem 10.1. Let F(x, y, z) be a differentiable function, and let P(x0 , y0 , z0 ) be a point on the
surface Σ with equation F(x, y, z) = k. If ∇F(x0 , y0 , z0 ) , 0, then the tangent plane to Σ at the point
P exists and is perpendicular to ∇F(x0 , y0 , z0 ). In particular, the tangent plane has equation
F x (x0 , y0 , z0 )(x − x0 ) + Fy (x0 , y0 , z0 )(y − y0 ) + Fz (x0 , y0 , z0 )(z − z0 ) = 0. (10.1)
Proof. Note that if γ is a curve on Σ given by the vector function r(t) = x(t)e1 + y(t)e2 + z(t)e3 ,
then the component functions of r(t) must satisfy the equation of Σ:
F(x(t), y(t), z(t)) = k.
When we differentiate this equation using the chain rule, we get
∂F dx ∂F dy ∂F dz
+ + = 0 ⇐⇒ ∇F · r0 (t) = 0.
∂x dt ∂y dt ∂z dt
In other words, the tangent vector to the curve, r0 (t), is perpendicular to the gradient vector ∇F.
First of all, this proves the existence of the tangent plane—it is the plane through P that is perpen-
dicular to ∇F(x0 , y0 , z0 ). Second, we get that (10.1) is the equation of the tangent plane. 
Example 10.1. Find the equation of the tangent plane at a point P(x0 , y0 , z0 ) of the one-sheet
hyperboloid
x2 y2 z2
F(x, y, z) = 2 + 2 − 2 = 1.
a b c
Solution. We have
2x0 2y0 2z0
F x (x0 , y0 , z0 ) = 2 , Fy (x0 , y0 , z0 ) = 2 , Fz (x0 , y0 , z0 ) = − 2 ,
a b c
so the equation of the tangent plane to the hyperboloid at a point P(x0 , y0 , z0 ) is
2x0 2y0 2z0
2
(x − x0 ) + 2 (y − y0 ) − 2 (z − z0 ) = 0,
a b c
77
which is equivalent to
2xx0 2yy0 2zz0 2x02 2y20 2z20
+ 2 − 2 = 2 + 2 − 2.
a2 b c a b c
Since P lies on the hyperboloid, the right side is equal to 2 and we can rewrite the last equation in
the form
xx0 yy0 zz0
+ 2 − 2 = 1.
a2 b c

Example 10.2. Find the equation of the tangent plane at a point P(x0 , y0 , z0 ) of the surface
z = f (x, y), where f (x, y) is a differentiable function.
Solution. We can rewrite the equation of the surface as
f (x, y) − z = 0,
which is of the form considered above with F(x, y, z) = f (x, y) − z and k = 0. Since
F x (x0 , y0 , z0 ) = f x (x0 , y0 ), Fy (x0 , y0 , z0 ) = fy (x0 , y0 ), Fz (x0 , y0 , z0 ) = −1,
we deduce that the equation of the tangent plane to the surface at (x0 , y0 , z0 ) is
f x (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ) − (z − z0 ) = 0,
which is equivalent to the equation stated without proof in §8.3. Therefore, Theorem 8.2 is merely
special case of Theorem 10.1. 
Finally, we define the normal line to a surface.
Definition. Let F(x, y, z) be a differentiable function and let P(x0 , y0 , z0 ) be a point of the sur-
face Σ with equation F(x, y, z) = k. The normal line to Σ at P is the line that passes through P and
is orthogonal to the tangent plane at P.
The same reasoning that we used to establish (10.1) shows that the symmetric equations of the
normal line to Σ at P is
x − x0 y − y0 z − z0
= = ,
F x (x0 , y0 , z0 ) Fy (x0 , y0 , z0 ) Fz (x0 , y0 , z0 )
with the proper modifications, if some of the partials vanish at (x0 , y0 , z0 ).

10.2. Tangent planes and normal vectors to parametric surfaces

Suppose that Σ is a surface with parametrization
r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 ((u, v) ∈ D)
and P0 (x0 , y0 , z0 ) = r(u0 , v0 ) is a point on Σ. We shall focus on curves γ on Σ that are the images of
parametric curves
u = u(t), v = v(t) (t ∈ I)
that lie in D and pass through the point (u0 , v0 ).
Definition. The tangent plane to Σ at P is the plane, if it exists, that contains P and the tangent
vectors to any smooth curve γ on Σ of the above type.
78
 We remark that this definition, at least at first glance, seems more restrictive than the definition
we gave in §8.3 for the tangent plane to a level surface: we look only at some curves on Σ instead
of looking at all such curves. However, it can be shown that our definition and the definition
that refers to all the curves on Σ yield the same notion of tangent plane. The following theorem
establishes the existence of the tangent plane at a “smooth” point of a parametric surface.
Theorem 10.2. Let r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 , (u, v) ∈ D, be a differentiable vector
function with partials
ru = xu e1 + yu e2 + zu e3 , rv = xv e1 + yv e2 + zv e3 ,
and let P(x0 , y0 , z0 ) = r(u0 , v0 ) be a point on the parametric surface Σ given by r(u, v). If
n = ru (u0 , v0 ) × rv (u0 , v0 ) , 0,
then the tangent plane to Σ at the point P exists and is perpendicular to the vector n.
Proof. Let γ be a curve on Σ of the type considered in the above definition. It can be parametrized
by means of the vector function
f(t) = r(u(t), v(t)) = x(u(t), v(t))e1 + y(u(t), v(t))e2 + z(u(t), v(t))e3 (t ∈ I).
Thus, we can use the chain rule to express its tangent vector f 0 (t) in terms of the derivatives of
r(u, v), u(t), and v(t):
df dx dy dz
= e1 + e2 + e3
dt dt dt dt
∂x du ∂x dv ∂y du ∂y dv ∂z du ∂z dv
    
= + e1 + + e2 + + e3
∂u dt ∂v dt ∂u dt ∂v dt ∂u dt ∂v dt
∂x du ∂y du ∂z du ∂x dv ∂y dv ∂z dv
   
= e1 + e2 + e3 + e1 + e2 + e3
∂u dt ∂u dt ∂u dt ∂v dt ∂v dt ∂v dt
du ∂x ∂y ∂z dv ∂x ∂y ∂z
   
= e1 + e2 + e3 + e1 + e2 + e3 = u0 (t)ru + v0 (t)rv .
dt ∂u ∂u ∂u dt ∂v ∂v ∂v
In particular, the tangent vector f 0 (t0 ) at the point P0 of γ is
f 0 (t0 ) = u0 (t0 )ru (u0 , v0 ) + v0 (t0 )rv (u0 , v0 ).
In other words, the tangent vector is the sum of a vector parallel to ru (u0 , v0 ) and a vector parallel
to rv (u0 , v0 ). It follows that the cross product n, which is perpendicular to both ru (u0 , v0 ) and
rv (u0 , v0 ), will be perpendicular to the tangent vector f 0 (t0 ) to the curve γ. Note that the last vector
is independent of the curve and is completely determined by the parametrization of the surface Σ.
This establishes the existence of the tangent plane: it is the plane through the point P0 with normal
vector n. 
Example 10.3. Find an equation of the tangent plane to the surface Σ with parametrization
r(u, v) = cos u(4 + 2 cos v)e1 + sin u(4 + 2 cos v)e2 + 2 sin ve3 (0 ≤ u, v ≤ 2π)
at the point (4, 0, 2) = r(0, π/2).
79
 Solution. We have
ru (u, v) = − sin u(4 + 2 cos v)e1 + cos u(4 + 2 cos v)e2 ,
rv (u, v) = −2 cos u sin ve1 − 2 sin u sin ve2 + 2 cos ve3 ,
so the normal vector to Σ at (4, 0, 2) is
n = (ru × rv )(0, π/2) = h0, 4, 0i × h−2, 0, 0i = (4e2 ) × (−2e1 ) = −8(e2 × e1 ) = 8e3 .
Thus, the equation of the tangent plane is
8(z − 2) = 0 ⇐⇒ z = 2. 
Example 10.4. Find the equation of the tangent plane at a point P(x0 , y0 , z0 ) of the surface
z = f (x, y), where f (x, y) is a differentiable function.
Solution. We can parametrize the given surface as
r(x, y) = xe1 + ye2 + f (x, y)e3 ((x, y) ∈ D),
so the normal vector n to it is
 
e1 e2 e3       
 0 f x  1 f x  1 0 
n = r x × ry =  1 0 f x  =  e +   e = − f x e1 − fy e2 + e3 .

e − 
 0 1 fy  1 f y  1 0 f y  2 0 1  3

Thus, the equation of the tangent plane at (x0 , y0 , z0 ) is

− f x (x0 , y0 )(x − x0 ) − fy (x0 , y0 )(y − y0 ) + (z − z0 ) = 0,
which is equivalent to the equation we gave earlier in Theorem 8.2 and Example 10.2. 

10.3. Smooth and piecewise smooth surfaces

Recall that we called the parametric curve γ given by the vector function r(t), t ∈ I, smooth
if the tangent vector to the curve, r0 (t), does not vanish for t ∈ I. We later saw that this condition
ensures that γ has a tangent line at each point. In the next definition, we define a “smooth surface”
so that it has a tangent plane at each point.
Definition. The level surface Σ with equation F(x, y, z) = k is called smooth, if the function
F(x, y, z) is differentiable at every point of Σ and ∇F(x, y, z) , 0, except possibly at the boundary
of Σ. The parametric surface Σ, given by
r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 ((u, v) ∈ D), (10.2)
is called smooth, if the partials ru and rv are defined for all (u, v) ∈ D and ru × rv , 0, except
possibly at the boundary of D.
Note that it is an immediate consequence of this definition and Theorems 10.1 and 10.2 that a
smooth surface (either level or parametric) has a tangent plane at each point.
Definition. The level surface Σ with equation F(x, y, z) = k is called piecewise smooth, if the
function F(x, y, z) is continuous and we can represent Σ as the union of a finite number of smooth
surfaces, say Σ = Σ1 ∪ Σ2 ∪ · · · ∪ Σn . The parametric surface Σ, given by (10.2) is called piecewise
smooth, if r(u, v) is continuous and we can represent Σ as the union of a finite number of smooth
surfaces, say Σ = Σ1 ∪ Σ2 ∪ · · · ∪ Σn .
80
 Example 10.5. Determine whether the surface Σ is smooth, piecewise smooth, or neither:
(a) Σ is the elliptic paraboloid z = x2 + 4y2 .
(b) Σ is the cone z2 = x2 + y2 .
First solution. (a) We view Σ as a level surface of F(x, y, z) = z − x2 − 4y2 . Since
∇F(x, y, z) = h−2x, −8y, 1i , 0,
the paraboloid is smooth.
(b) We view Σ as a level surface of F(x, y, z) = z2 − x2 − y2 . Since
∇F(x, y, z) = h−2x, −2y, 2zi = 0 ⇐⇒ x = y = z = 0,
the cone is not smooth. However, we can break Σ into two surfaces: the level surface Σ1 of
F(x, y, z) = z2 − x2 − y2 (z ≥ 0),
and the level surface Σ2 of
F(x, y, z) = z2 − x2 − y2 (z ≤ 0),
each of which is smooth, because the origin is a boundary point for them. Thus, the cone is
piecewise smooth. 

Second solution. (b) We view Σ as a parametric surface given by

r(u, v) = v cos ue1 + v sin ue2 + ve3 ,
where 0 ≤ u ≤ 2π, −∞ < v < ∞. Then
ru = −v sin ue1 + v cos ue2 , rv = cos ue1 + sin ue2 + e3 ,
and
ru × rv = (−v sin ue1 + v cos ue2 ) × (cos ue1 + sin ue2 + e3 )
= −v sin2 u(e1 × e2 ) − v sin u(e1 × e3 ) + v cos2 u(e2 × e1 ) + v cos u(e2 × e3 )
= −v sin2 ue3 + v sin ue2 − v cos2 ue3 + v cos ue1 = v cos ue1 + v sin ue2 − ve3 .
Hence, ru × rv = 0 when v = 0 and the cone is not smooth. However, we can break Σ into two
surfaces: the parametric surface Σ1 given by
r(u, v) = v cos ue1 + v sin ue2 + ve3 (0 ≤ u ≤ 2π, v ≥ 0),
and the parametric surface Σ2 given by
r(u, v) = v cos ue1 + v sin ue2 + ve3 (0 ≤ u ≤ 2π, v ≤ 0),
each of which is smooth, because the origin is a boundary point for them. Thus, the cone is
piecewise smooth. 
81
 10.4. Orientable surfaces
We conclude this lecture with the notion of orientation of a surface, which we shall need in
our discussion of surface integrals later in these notes. Intuitively, we think of a surface as a two-
dimensional object, which is placed in space and thus has an “upper side” and a “lower side.”
However, this is not always the case, even for nice, smooth surfaces. We use the concept of orien-
tation of a surface to separate the intuitive, two-sided surfaces from the one-sided ones. However,
before giving a formal definition, we present an example of a surprisingly simple one-sided surface.

C B

D A

Figure 10.1. The Möbius strip Figure 10.2. The Möbius strip
before gluing after gluing

Example 10.6. Take a long strip of paper in the form of a rectangle ABCD, the vertices listed
counterclockwise so that AB and CD are the two short sides of the strip (see Figure 10.1). Then
twist the paper half a turn and glue the short sides so that A gets glued to C and B to D. The
resulting surface (see Figure 10.2) has only one side! Indeed, if you take a paintbrush, started
at the cut, and moved along the surface parallel to its edges, you will be able to paint the entire
strip without lifting the paintbrush. This surface is known as the Möbius strip. You may think at
first that the Möbius strip is an esoteric example and that “normal” surfaces, given by nice, simple
formulas, don’t behave this way. Think again! The Möbius strip is given by a nice, simple formula:
its parametric equations are
x = (2 + v cos(u/2)) cos u, y = (2 + v cos(u/2)) sin u, z = v sin(u/2), (10.3)
where 0 ≤ u ≤ 2π and − 21 ≤ v ≤ 12 . 
Definition. Let Σ be a smooth surface in R3 . For each nonboundary point P(x, y, z) on the
surface, let n1 (x, y, z) and n2 (x, y, z) be the two unit vectors at P that are normal to Σ. We say that
Σ is orientable, if it is possible to select a unit normal vector n(x, y, z) at every nonboundary point
of Σ in such a way that n(x, y, z) varies continuously as (x, y, z) varies over Σ. A choice of the unit
normal vector n as described above is called an orientation of Σ.
Remark. If a surface Σ is orientable, then there are two distinct orientations of that surface. For
example, let us choose a distinguished point P0 (x0 , y0 , z0 ) on Σ. Let us select and fix one of the two
unit normal vectors at P0 ; we denote it by n0 . Then for each nonboundary point P near P0 only
one of the unit normal vectors at P will be close to n0 , the other will be close to −n0 . Thus, the
continuity requirement on our choice of normal vectors settles the orientation of all normal vectors
near P0 . Of course, once we have decided on the choice of normal vectors at points near P0 , we
82
can repeat the same procedure to settle the orientation of the normal vectors at points that may not
be near P0 but are near points that are near P0 ; after that, we can settle the choices at points that
may not be near P0 or near points that are near P0 , but are near points that are near points that are
near P0 ; etc. Ultimately, after we choose between the two unit normal vectors n0 and −n0 at P0 ,
we set a domino effect that yields a consistent choice between the unit normal vectors at all other
nonboundary points of Σ.

r[u_,v_] := {2Cos[u]Sin[v], 2Sin[u]Sin[v], 2Cos[v]}

n[u_,v_] := 0.5*r[u,v]
p = ParametricPlot3D[r[u,v], {u,0,2Pi}, {v,0,Pi},
Boxed -> False, Axes -> False, Mesh -> None, ViewPoint -> {6,1,4}];
Manipulate[
Show[{p, vector[r[u,v], r[u,v] + n[u,v], Red]}, PlotRange -> All],
{{u,0}, 0, 4Pi}, {{v,0}, 0, Pi}]

r[u_,v_] := {Cos[u](2 + v*Cos[u/2]), Sin[u](2 + v*Cos[u/2]), v*Sin[u/2]}

n[u_,v_] = 0.5*Cross[D[r[u,v], u], D[r[u,v], v]]
p = ParametricPlot3D[r[u,v], {u,0,2Pi}, {v,-1/2,1/2},
Boxed -> False, Axes -> False, Mesh -> None, ViewPoint -> {6,1,4}];
Manipulate[
Show[{p, vector[r[u,v], r[u,v] + n[u,v], Red]}, PlotRange -> All],
{{u,0}, 0, 4Pi}, {{v,0}, -1/2, 1/2}]

Figure 10.3. Normal vectors of orientable and non-orientable surfaces

Example 10.7. A good way to illustrate the concept of orientability is to compare the normal
vectors on an orientable surface to those on a non-orientable surface. Such a comparison is most
83
successful, if it is undertaken using graphical software like Mathematica, so that is what we will
do. In this example, we will compare the normal vectors to the sphere x2 + y2 + z2 = 4 and the
normal vectors to the Möbius strip with parametric equations (10.3).
The first block of Mathematica code in Figure 10.3 produces a dynamic plot of the sphere
(parametrized as in Example 10.9 below) and of the outward unit normal vector at its north pole.
One can then use the sliders to move the normal vector on the surface of the sphere. Execution that
code1 in Mathematica and experimentation with the sliders will demonstrate that as one moves the
normal vector on the sphere, its direction changes but always points away from the center of the
sphere. Furthermore, no matter how the normal vector meanders around, every time it returns to a
specific point it points out. This is because the sphere is orientable (see Example 10.9 below); the
normal vectors Mathematica displays are the vectors of its “outward orientation.”
The second block of code in Figure 10.3 produces a similar dynamic plot of the Möbius strip
and one of its normal vectors. However, experimentation with its output will demonstrate that as
one moves the normal vector on the Möbius strip, the vector may be positioned at the same location
with opposite directions: e.g., when (u, v) = (0, 0) and when (u, v) = (2π, 0) (see the screenshots in
Figure 10.3). This is because the Möbius strip is non-orientable. 
Example 10.8. Let Σ be the graph of a function z = f (x, y), (x, y) ∈ D. In Example 10.4, we
used the parametrization
r(x, y) = xe1 + ye2 + f (x, y)e3 ((x, y) ∈ D)
to show that
r x × ry = − f x e1 − fy e2 + e3
is a normal vector to the surface at the point (x, y, z) of the surface. Note that this vector always
has a positive third component. Thus, the normalized cross product
r x × ry − f x e1 − fy e2 + e3
n= =
1/2 ,
kr x × ry k 1 + f x2 + fy2
will pick consistently one of the two unit normal vectors to the surface: namely, it will always pick
the unit normal that points upward. Consequently, n represents an orientation of the surface; this
orientation is often referred to as the upward orientation of the surface. 
Example 10.9. Let Σ be the sphere x2 + y2 + z2 = 4. Its spherical equation is ρ = 2, so we
can describe Σ as the set of points (x, y, z) in space whose Cartesian coordinates are related to their
spherical coordinates via the equations
x = 2 cos θ sin φ, y = 2 sin θ sin φ, z = 2 cos φ,
where 0 ≤ θ ≤ 2π, 0 ≤ φ ≤ π. This provides a parametrization of Σ:
r(θ, φ) = 2 cos θ sin φ e1 + 2 sin θ sin φ e2 + 2 cos φ e3 ,
where 0 ≤ θ ≤ 2π, 0 ≤ φ ≤ π. Then
rθ = −2 sin θ sin φ e1 + 2 cos θ sin φ e2 , rφ = 2 cos θ cos φ e1 + 2 sin θ cos φ e2 − 2 sin φ e3 ,

1
The code uses the vector command described in §2.9, so that code needs to be included as well.
84
and
rθ × rφ = (−2 sin θ sin φ e1 + 2 cos θ sin φ e2 ) × (2 cos θ cos φ e1 + 2 sin θ cos φ e2 − 2 sin φ e3 )
= −4 sin2 θ sin φ cos φ(e1 × e2 ) + 4 sin θ sin2 φ(e1 × e3 )
+ 4 cos2 θ sin φ cos φ(e2 × e1 ) − 4 cos θ sin2 φ(e2 × e3 )
= −4 cos θ sin2 φ e1 − 4 sin θ sin2 φ e2 − 4 sin2 θ + cos2 θ sin φ cos φ e3

= −4 sin φ cos θ sin φ e1 + sin θ sin φ e2 + cos φ e3 = −2 sin φ r.

Hence, the normal vectors at a point of the sphere are parallel to the position vector r of the point.
Since the length of r is the distance from the origin to a point on Σ, we have krk = 2. Thus, n = 12 r
is a unit normal vector at the point P = r(u, v) of Σ. Note that since n has the same direction as
−−→
the OP, it will consistently be the unit normal vector to the sphere that is pointed away from the
origin. Therefore, n yields an orientation of Σ that is often called the outward orientation of the
sphere. 

Exercises
Find the equations of the tangent plane and of the normal line to the given surface at the given point.
10.1. z = xye x+2y , P(2, −1,

−2) 10.3. x2 − 2y2 + 2z2 = 10, P(2, −1, −2)
10.2. z = log2 xy2 + x2 y , P(1, −2, 1) 10.4. ln(xyz) = x + yz−1 , P(1, e, e)

10.5. (a) Find the equation of the normal line to the sphere x2 + y2 + z2 = r2 at a point (x0 , y0 , z0 ) of the sphere.
(b) Show that the normal line from part (a) passes through the origin.
√ √ √
10.6. (a) Find the equation of the tangent plane to the surface x + y + z = 4 at a point P(x0 , y0 , z0 ) on this surface.
(b) Show that the sum of the lengths of the line segments that the tangent plane from part (a) cuts from the
coordinate axes is independent of the choice of P. (The line segment that a plane cuts from a coordinate
axis is the segment whose endpoints are the origin and the intersection point of the plane and the axis.)
10.7. Let P(x0 , y0 , z0 ) be a point on the surface xyz = 6. Show that the volume of the tetrahedron that the tangent plane
at P cuts from the first octant is independent of the choice of P.
10.8. Find the points on the hyperboloid x2 − 2y2 − z2 = −2 at which the tangent plane is parallel to the plane
2x − 3y + 2z + 7 = 0.
10.9. Find the points on the hyperbolic paraboloid x = y2 − 4z2 at which the normal line is perpendicular to the plane
x + y + z = 1.

2
2
2
10.10. Show that the hyperboloid x2 − 2y2 − z2 = −2 and the ellipsoid x + 43 + y + 13 + 14 z + 13 = 1 are tangent
to each other at the point (−1, −1, 1), that is, show that they have the same tangent plane at that point.
Find an equation of the tangent plane to the given parametric surface at the given point.
10.11. r(u, v) =
huv,
p u sin v, u cos vi, (u, v) = (1, π/2) 10.14. r(u, v) = u2 − v2 , u + v, u2 + 2v , (0, 2, 3)


√
10.12. r(y, z) =
4 + y2 + z2 , y, z , (y, z) = (1, 1) 10.15. r(u, v) = hu cos 2v, u sin 2v, vi , (1, 3, π/6)
10.13. r(u, v) = uv, u2 + v2 , u + v + 2 , (u, v) = (−1, 0)


10.16. Let Σ be the parametric surface given by

r(u, v) = (u + v)e1 + (uv)e2 + (u3 − v3 )e3 (−∞ < u, v < ∞).
(a) Find the point P on Σ that corresponds to the values u = 1 and v = 1 of the parameters.
(b) Find the equation of the tangent plane to Σ at the point P found in (a).

85
10.17. Let Σ be the parametric surface given by
r(u, v) = (u + v)e1 + (u2 + v2 )e2 + (u3 + v3 )e3 (−∞ < u, v < ∞).
Find the point(s) on Σ where the tangent plane is perpendicular to the line x = y = −z.
Determine whether the given surface is smooth, piecewise smooth, or neither.
10.18. x2 + 3y2 −

2xyz2 = 6 3 10.20. r(u, v) = hu, uv, sin uvi , (u, v) ∈ R2
10.19. r(u, v) = 3u, u − 2v, u + v , (u, v) ∈ R 10.21. z2 + sin(xy) = 1
2
 2

10.22. Find the point(s) where the parametric surface given by

r(u, v) = (u + v)e1 + (uv)e2 + (u2 − v2 )e3 (−∞ < u, v < ∞)
is not smooth.

86
 LECTURE 11

Extremal Values of Multivariable Functions

In this lecture, we shall discuss the local and absolute (or global) maximum and minimum
values of a function of two variables f (x, y).

11.1. Review of extremal values of single-variable functions

We start with a review of the extrema of single-variable functions. For example, let us classify
the local and global extrema of the function
f (x) = x4 − 8x3 + 18x2 − 10
in the interval [−1, 2]. First, we find the critical points of f : these are the zeros of
f 0 (x) = 4x3 − 24x2 + 36x = 4x(x2 − 6x + 9) = 4x(x − 3)2 .
We have
4x(x − 3)2 = 0 ⇐⇒ x = 0 or x = 3.
We determine whether the critical points yield local extrema by the second-derivative test:
f 00 (x) = 12x2 − 48x + 36 =⇒ f 00 (0) = 36 > 0, f 00 (3) = 0,
that is, f has a local minimum at x = 0, (0, −10). (We can also show that f has no extremum at
x = 3, but that does not follow from the second derivative test.) Finally, to find the absolute extrema
of f in [−1, 2], we compare the local extrema and the boundary values: f (−1) = 17, f (0) = −10,
and f (2) = 14. We find that the absolute maximum of f is 17 and its absolute minimum is −10.

11.2. Local extrema of a function of two variables

Definition. Let f be a function of two variables. If (a, b) is a point in the domain of f , then
f has a local maximum at (a, b) if f (x, y) ≤ f (a, b) for all points (x, y) near (a, b); f has a local
minimum at (a, b) if f (x, y) ≥ f (a, b) for all points (x, y) near (a, b). A local maximum or minimum
is called also a local extremum.
Theorem 11.1. If f has a local extremum at (a, b) and its first-order partials f x (a, b) and fy (a, b)
exist, then
f x (a, b) = fy (a, b) = 0.
Proof. If f (x, y) has a local maximum/minimum at (a, b), then the single-variable function
g(x) = f (x, b) also has a local maximum/minimum at x = a. Hence,
g0 (a) = 0 =⇒ f x (a, b) = 0.
Similarly, fy (a, b) = 0 by considering h(y) = f (a, y). 
87
 Definition. Let f be a function of two variables. A critical point of f is a point (a, b) such that
either f x (a, b) = fy (a, b) = 0, or one of these partial derivatives does not exist. Those critical points
(a, b) of f for which f x (a, b) = fy (a, b) = 0 are sometimes called stationary points.
Example 11.1. Find all critical points of the function
f (x, y) = x3 + y2 − 6xy + 6x + 3y − 2.
Solution. This function has first-order (and also second-, third-, and higher-order) partial
derivatives at any point (x, y) in the R2 , so the only critical points will be the stationary points
of f . Since
f x (x, y) = 3x2 − 6y + 6 and fy (x, y) = 2y − 6x + 3,
the stationary points of f are the solutions of the system
( (
3x2 − 6y + 6 = 0 x2 − 2y + 2 = 0
⇐⇒ .
2y − 6x + 3 = 0 2y = 6x − 3
Substituting 2y from the second equation into the first, we find that x must satisfy the equation
x2 − (6x − 3) + 2 = 0 ⇐⇒ x2 − 6x + 5 = 0 ⇐⇒ x = 1 or x = 5.
When we substitute x = 1 in the second equation of the system, we obtain 2y = 3 so y = 1.5; when
we substitute x = 5, we obtain 2y = 27 so y = 13.5. Thus, the critical points of f are (1, 1.5) and
(5, 13.5). 

11.3. The second-derivative test

Note that the theorem in the previous section says that (just as in the case of functions of a
single variable) a function of two variables can have a local extremum only at a critical point.
However, we know that for a function of one variable a critical point is not necessarily an extremal
point. Is this also the case for functions of two variables? And if so, then how can we determine
whether a critical point is an extremal point or not? The following theorem is often sufficient to
answer these questions. It is the two-variable version of the second-derivative test.
Theorem 11.2 (Second-derivative test). Suppose that f (x, y) has continuous second-order par-
tial derivatives near the stationary point (a, b) and define
 
f (a, b) f (a, b)
D = D(a, b) = f xx (a, b) fyy (a, b) − f xy (a, b)2 =  xx xy .
 
fyx (a, b) fyy (a, b)
Then:
i) if D > 0 and f xx (a, b) > 0, f has a local minimum at (a, b);
ii) if D > 0 and f xx (a, b) < 0, f has a local maximum at (a, b);
iii) if D < 0, f does not have a local extremum at (a, b).
In case iii), the point (a, b) is called a saddle point, since under those conditions the graph of f
near (a, b) resembles a saddle.
Example 11.2. Find the local extrema of the function
f (x, y) = x3 + y2 − 6xy + 6x + 3y − 2.
88
 Solution. We know from Example 9.1 that the critical points of f are (1, 1.5) and (5, 13.5).
The second-order partials of f are:

f xx (x, y) = 6x, fyy (x, y) = 2, f xy (x, y) = −6,

so f xx (5, 13.5) = 30 and

   
 6 −6  30 −6
D(1, 1.5) =   = −24, D(5, 13.5) =   = 24.
−6 2  −6 2 
Hence, by the second-derivative test, (1, 1.5) is a saddle point of f and f (5, 13.5) = −29 is a local
minimum. 

Example 11.3. Find the point on the plane 2x − y + 2z − 3 = 0 that is closest to the origin. What
is the distance between the plane and the origin?

Solution. We want to minimize the function

d(x, y, z) = x2 + y2 + z2

for (x, y, z) subject to the constraint

2x − y + 2z − 3 = 0.

This is the same as minimizing the function

f (x, z) = x2 + (2x + 2z − 3)2 + z2 = 5x2 + 8xz + 5z2 − 12x − 12z + 9

in the xz-plane. The first-order partials of f are

f x (x, z) = 10x + 8z − 12, fz (x, z) = 8x + 10z − 12,

so the critical points are the solutions of the system

( (
10x + 8z − 12 = 0 5x + 4z = 6
⇐⇒ .
8x + 10z − 12 = 0 4x + 5z = 6

Solving the last system, we find that the only critical point of f is ( 32 , 23 ). The second-order partials
of f are
f xx (x, z) = fzz (x, z) = 10, f xz (x, z) = fzx (x, z) = 8,
so
 
10 8 
D= D( 32 , 23 ) =   = 36.
8 10
Since D > 0 and f xx > 0, we conclude that f ( 23 , 23 ) = 43 is a local minimum of f . It is intuitively
clear that this local minimum must actually be an absolute minimum of f , so the minimum distance
occurs when the point (x, y, z) on the plane is ( 23 , − 13 , 23 ). The distance from the origin to that point
(which is the distance from the origin to the plane) is (2/3)2 + (−1/3)2 + (2/3)2 = 1.
p

89
 11.4. Absolute extrema of a function of two variables
In real life, we are more likely to be interested in absolute (or global) extrema of functions than
in local extrema. In the last example, we used practical considerations (i.e., intuition) to make the
jump from the local to the global minimum. Is there another (hopefully, more rigorous) way to
approach global extrema? The answer is “yes,” if the function is defined on a “closed, bounded
set.”
First, we must say what does it mean for a set in the plane to be “closed and bounded.” Instead
of formal definitions we will be content with explanations by example. A set R is closed if it con-
tains all its “boundary points.” For example, among the sets on Figure 11.1 (where solid boundary
curve belongs to the set and a dotted curve does not), the sets in (a) and (d) are closed and the sets
in (b), (c), and (e) are not. A set R is bounded if it is contained inside some disk centered at the
origin. For example, of the sets on Figure 11.1, only (d) is not bounded.

(a) (b) (c)

(d) (e)

Figure 11.1. Examples of closed and bounded sets

We can now state the algorithm for finding the absolute extrema of a function of two variables
on a closed, bounded set. It is a generalization of the closed interval method for finding absolute
extrema of functions of a single variable on a closed interval. To find the absolute extrema of
f (x, y) on a closed, bounded set R:
1. Find the critical points of f inside R and the respective values of f .
2. Find the absolute extrema of f on the boundary of R.
90
 3. The absolute maximum of f in R is the largest among the numbers found in Steps 1 and 2;
the absolute minimum of f in R is the smallest among the numbers found in Steps 1 and 2.
Example 11.4. Find the absolute maximum and minimum values of the function
f (x, y) = x3 − y3 − 3x + 12y
in the closed and bounded region
R = (x, y) : 0 ≤ x ≤ 2, −3x ≤ y ≤ 0 .


Solution. 1. The first-order partials of f are

f x (x, y) = 3x2 − 3, fy (x, y) = −3y2 + 12,
so the critical points of f are the solutions of the system
( ( (
3x2 − 3 = 0 x2 = 1 x = ±1
⇐⇒ ⇐⇒ .
−3y2 + 12 = 0 y2 = 4 y = ±2
That is, f has four critical points: (1, 2), (1, −2), (−1, 2), and (−1, −2). However, of these only
P(1, −2) lies inside R. The respective value of f is f (1, −2) = −18.
2. The boundary of R consists of three line segments (see Figure 11.2):
`1 : (x, y) : 0 ≤ x ≤ 2, y = 0 ,


`2 : (x, y) : 0 ≤ x ≤ 2, y = −3x ,


`3 : (x, y) : x = 2, −6 ≤ y ≤ 0 .


2.1. When (x, y) lies on `1 , f (x, y) is really only a function of x:

g1 (x) = f (x, 0) = x3 − 3x (0 ≤ x ≤ 2).
The critical values of g1 are the solutions of 3x2 − 3 = 0. These are x = ±1, but we are interested
only in x = 1 (since −1 lies outside [0, 2]). Thus, we need compare
g1 (0) = 0, g1 (1) = −2, and g1 (2) = 2.
We find that the absolute minimum and maximum of f on `1 are f (1, 0) = −2 and f (2, 0) = 2.

y
`1
x
P
`3 `2

Figure 11.2. The boundary of R

91
 2.2. When (x, y) lies on `2 , f (x, y) is again a function of x only:
g2 (x) = f (x, −3x) = 28x3 − 39x (0 ≤ x ≤ 2).
√
The critical values of g2 are the solutions√of 84x2 −39 = 0: x = ± 13/28. Again, we are interested
only in one of these critical points: x = 13/28. We compare
p p
g2 (0) = 0, 13/28 = −26 13/28 = −17.71 . . . , g2 (2) = 146.


g2 and
We find that the absolute minimum and maximum of f on `2 are
p p
f ( 13/28, −3 13/28) = −17.71 . . . and f (2, −6) = 146.
2.3. When (x, y) lies on `3 , f (x, y) is only a function of y:
g3 (y) = f (2, y) = −y3 + 12y + 2 (−6 ≤ y ≤ 0).
The critical values of g3 are the solutions of −3y2 + 12 = 0, y = ±2. After dropping y = 2, we
compare
g3 (0) = 2, g3 (−2) = −14, and g3 (−6) = 146.
We find that the absolute minimum and maximum of f on `3 are, respectively, f (2, −2) = −14 and
f (2, −6) = 146.
3. Comparing all the values found in Steps 1 and 2, we conclude that
min f (x, y) = min{−18, −2, 2, −17.71, 146, −14} = −18 = f (1, −2),
(x,y)∈R
max f (x, y) = max{−18, −2, 2, −17.71, 146, −14} = 146 = f (2, −6). 
(x,y)∈R

Example 11.5. Find the absolute maximum and minimum values of the function
f (x, y) = x2 − 2xy − 2x + 4y
in the closed region bounded by the ellipse γ with parametric equations
x = 3 cos t, y = 2 sin t (0 ≤ t ≤ 2π).
Solution. 1. The first-order partials of f are
f x (x, y) = 2x − 2y − 2, fy (x, y) = −2x + 4,
so the critical points of f are the solutions of the system
( ( (
2x − 2y − 2 = 0 x=y+1 y= x−1=1
⇐⇒ ⇐⇒
−2x + 4 = 0 x−2=0 x=2
Hence, P(2, 1) is the only critical point of f (x, y). It lies inside γ (see Figure 11.3) and f (2, 1) = 4.

2. On γ, we have
f (x, y) = f (3 cos t, 2 sin t) = 9 cos2 t − 12 cos t sin t − 6 cos t + 8 sin t,
so we have to find the maximum and the minimum of
g(t) = 9 cos2 t − 12 cos t sin t − 6 cos t + 8 sin t (0 ≤ t ≤ 2π).
92
 y
γ
P
x

Figure 11.3. The ellipse γ and the critical point P

We have
g0 (t) = −18 cos t sin t + 12 sin2 t − 12 cos2 t + 6 sin t + 8 cos t
= 24 sin2 t + 6 sin t − 12 + 2 cos t(4 − 9 sin t).
√
Substituting u = sin t and cos t = ± 1 − u2 , we can rewrite the equation g0 (t) = 0 as
√
24u2 + 6u − 12 ± 2 1 − u2 (4 − 9u) = 0
√
12u2 + 3u − 6 = ± 1 − u2 (9u − 4)

2
12u2 + 3u − 6 = (1 − u2 )(9u − 4)2
225u4 − 200u2 + 36u + 20 = 0.
The last equation has four real roots, whose approximate values are:
α1 ≈ −0.980, α2 ≈ −0.245, α3 ≈ 0.547, α4 ≈ 0.678.
These are the values of sin t at the critical numbers of g(t). They yield a total of eight possible
critical numbers: four pairs that have the same sines and opposite cosines. In fact, only one of
the numbers in each pair is a critical number of g, but we will ignore this and will treat all eight
numbers as critical. The values of all these critical (and pseudo-critical) numbers, their sines and
cosines, and the respective values of g(t) are listed in Table 11.1.

αk t cos t sin t g(t)

−0.980 4.510 −0.201 −0.980 −8.630
−0.980 4.915 0.201 −0.980 −6.316
−0.245 3.389 −0.970 −0.245 9.470
−0.245 6.036 0.970 −0.245 3.533
0.547 0.579 0.837 0.547 0.166
0.547 2.563 −0.837 0.547 21.201
0.678 0.745 0.735 0.678 −0.104
0.678 2.397 −0.735 0.678 20.681
Table 11.1. Values of g(t) at its critical numbers

93
We also have g(0) = g(2π) = 3, so
max g(t) = g(π − arcsin α3 ) ≈ 21.201, min g(t) = g(π − arcsin α1 ) ≈ −8.630.
0≤t≤2π 0≤t≤2π
3. Comparing all the values found in Steps 1 and 2, we conclude that the absolute minimum of
f inside γ is
min{4, −8.630, 21.201} = −8.630,
and that the absolute maximum of f is
max{4, −8.630, 21.201} = 21.201. 
Exercises
Find and classify the critical points (as local maxima, local minima, or saddle points) of the given function.
11.1. f (x, y) = 9 + x + 2y − x2 + 12 y2 11.4. f (x, y) = xy(x + y + 1) 11.7. f (x, y) = x2 − 6xy + 4y3
11.2. f (x, y) = 2x − 4y − x − 2y
2 2 11.5. f (x, y) = e2y
sin(xy) 11.8. f (x, y) = x3 − 6xy + 8y3
11.9. f (x, y) = x3 + y4 − 3xy + 4
2 2
11.3. f (x, y) = x4 + y4 − 4xy + 2 11.6. f (x, y) = xye−x −y

Find the absolute maxima and minima of the given function on the given set.
11.10. f (x, y) = 1 + 2x + 3y on the triangle with vertices (0, 3), (2, 1), and (5, 3)
11.11. f (x, y) = x2 + 4y2 − xy on the square −1 ≤ x ≤ 1, −1 ≤ y ≤ 1
11.12. f (x, y) = x3 + 2xy2 − 2x on the disk x2 + y2 ≤ 4
11.13. f (x, y) = x3 + y2 − 2xy + 3 on the square 0 ≤ x ≤ 1, 0 ≤ y ≤ 1
11.14. Find the shortest distance from the point (3, 2, −1) to the plane x − y + 2z = 1.
11.15. Find the point(s) on the cone (z − 2)2 = x2 + y2 that is closest to the point (1, 1, 2).
11.16. Among all the rectangular boxes of volume 125 ft3 , find the dimensions of that which has the least surface
area.
11.17. Find the volume of the largest rectangular box with faces parallel to the coordinate planes that can be inscribed
in the ellipsoid x2 + 4y2 + 4z2 = 16.

94
 LECTURE 12

Lagrange Multipliers*

12.1. Introduction
Recall Example 11.3. In that example, we had to minimize the function
f (x, y, z) = x2 + y2 + z2
subject to the constraint
2x − y + 2z − 3 = 0.
We used the constraint to eliminate one of the variables and reduced the problem to that of min-
imizing a related function of two variables (without any constraints). On an abstract level, one
expects that the same idea should work if one wants to minimize or maximize any “nice” function
f (x, y, z) subject to the constraint that g(x, y, z) = k, where g is another “nice” function and k is a
constant. However, while such an approach is theoretically possible, for any practical purpose it is
pretty much doomed unless g is a very simple function (such as the linear function that appears in
the above example). In this lecture, we describe the method of Lagrange multipliers, which allows
us to solve the general problem.

12.2. Basic form of the method of Lagrange multipliers

Suppose that P(x0 , y0 , z0 ) is a point at which the function f (x, y, z) is maximal subject to the
constraint g(x, y, z) = k. This constraint describes a level surface, which we will denote by Σ. Thus,
P is a point on the surface Σ such that f (x0 , y0 , z0 ) is maximal. Let γ be a smooth parametric curve
on Σ parametrized by the vector function r(t) = hx(t), y(t), z(t)i, t ∈ I, and such that r(t0 ) = P (so
that, γ passes through P). Since f (x0 , y0 , z0 ) is maximal on Σ and γ lies on Σ, we have
f (x0 , y0 , z0 ) ≥ f (x, y, z) for all (x, y, z) on γ.
Using the parametrization of γ, we get
f (x(t0 ), y(t0 ), z(t0 )) ≥ f (x(t), y(t), z(t)) for all t ∈ I,
so t = t0 is a local maximum of the single-variable function h(t) = f (x(t), y(t), z(t)), t ∈ I. There-
fore, h0 (t0 ) = 0. On the other hand, by the chain rule,
h0 (t) = f x (x(t), y(t), z(t))x0 (t) + fy (x(t), y(t), z(t))y0 (t) + fz (x(t), y(t), z(t))z0 (t)
= ∇ f (x(t), y(t), z(t)) · r0 (t),
so the equation h0 (t0 ) = 0 can be rewritten as
∇ f (x0 , y0 , z0 ) · r0 (t0 ) = 0.
That is, the gradient of f at the point P is perpendicular to the tangent vector r0 (t0 ) to γ at P. Since
γ was an arbitrary curve on Σ, this shows that the gradient ∇ f (P) is perpendicular to all tangent
95
vectors to Σ. Hence, ∇ f (P) is parallel to the normal vector to Σ at P—which is ∇g(P). We conclude
that if P(x0 , y0 , z0 ) is on Σ and f (x0 , y0 , z0 ) is maximal, then
∇ f (x0 , y0 , z0 ) = λ∇g(x0 , y0 , z0 ) for some number λ.
We note that if we had started with a point P(x0 , y0 , z0 ) on Σ such that f (x0 , y0 , z0 ) is minimal, we
could have used a similar argument to reach the same conclusion. These observation motivate the
following approach towards constrained optimization.
To find the extrema of a function f (x, y, z) subject to the constraint g(x, y, z) = k (when ∇g , 0
on the surface g(x, y, z) = k):
1. Find all the values x, y, z, and λ such that
(
∇ f (x, y, z) = λ∇g(x, y, z),
g(x, y, z) = k.
That is, solve this system of four equations in the four unknowns x, y, z, λ.
2. Evaluate f at all the points (x, y, z) found in Step 1. The maximum of f is the largest
among these numbers; the minimum of f is the smallest among them.
Example 12.1. Find the maximum and minimum values of the function f (x, y, z) = 8x − 4z on
the ellipsoid x2 + 10y2 + z2 = 5.
Solution. Let g(x, y, z) = x2 + 10y2 + z2 . We have
f x (x, y, z) = 8, fy (x, y, z) = 0, fz (x, y, z) = −4,
and
g x (x, y, z) = 2x, gy (x, y, z) = 20y, gz (x, y, z) = 2z,
so the system in Step 1 of the method of Lagrange multipliers is
8 = 2λx x = 4/λ
 

 

0 = 20λy
 y = 0

⇐⇒
−4 = 2λz
 z = −2/λ

x + 10y + z = 5 x + 10y2 + z2 = 5

 2 
2 2  2

Here we have used that λ , 0, because of the first or third equations. Substituting the expressions
for x, y, and z in the last equation of the system, we find that λ must be a solution of
16/λ2 + 0 + 4/λ2 = 5 ⇐⇒ λ2 = 4 ⇐⇒ λ = ±2.
The respective points (x, y, z) are (2, 0, −1) and (−2, 0, 1), so the conditional maximum of f is
max{ f (2, 0, −1), f (−2, 0, 1)} = max{20, −20} = 20 = f (2, 0, −1),
and the conditional minimum of f is
min{ f (2, 0, −1), f (−2, 0, 1)} = min{20, −20} = −20 = f (−2, 0, 1). 
Example 12.2. Find the maximum and minimum values of the function f (x, y, z) = x3 + y3 + z3
on the sphere x2 + y2 + z2 = 1.
96
 Solution. Let g(x, y, z) = x2 + y2 + z2 . We have
f x (x, y, z) = 3x2 , fy (x, y, z) = 3y2 , fz (x, y, z) = 3z2 ,
and
g x (x, y, z) = 2x, gy (x, y, z) = 2y, gz (x, y, z) = 2z,
so the system in Step 1 of the method of Lagrange multipliers is
= x(x − 2λ/3) = 0 x = 0, 2λ/3
  
2

 3x 2λx 
 

3y2 = 2λy
 y(y − 2λ/3) = 0
 y = 0, 2λ/3

⇐⇒ ⇐⇒

 3z2 = 2λz 
 z(z − 2λ/3) = 0 
 z = 0, 2λ/3
x + y2 + z2 = 1 x + y2 + z2 = 1 x + y2 + z2 = 1

 2 
 2 
 2

For each of the eight choices for x, y, z from the first three equations, we now solve the fourth
equation for λ:
1. x = y = z = 0. Then the fourth equation fails and we get no solution of the system.
2. x = y = 0, z = 2λ/3. Then the fourth equation becomes
4λ2 /9 = 1 ⇐⇒ λ = ±3/2,
and we obtain the points (0, 0, 1) and (0, 0, −1).
3. x = z = 0, y = 2λ/3. Then the fourth equation becomes
4λ2 /9 = 1 ⇐⇒ λ = ±3/2,
and we obtain the points (0, 1, 0) and (0, −1, 0).
4. y = z = 0, x = 2λ/3. Then the fourth equation becomes
4λ2 /9 = 1 ⇐⇒ λ = ±3/2,
and we obtain the points (1, 0, 0) and (−1, 0, 0).
5. x = 0, y = z = 2λ/3. Then the fourth equation becomes
√
8λ2 /9 = 1 ⇐⇒ λ = ±3 2/4,
√ √ √ √
and we obtain the points (0, 2/2, 2/2) and (0, − 2/2, − 2/2).
6. y = 0, x = z = 2λ/3. Then the fourth equation becomes
√
8λ2 /9 = 1 ⇐⇒ λ = ±3 2/4,
√ √ √ √
and we obtain the points ( 2/2, 0, 2/2) and (− 2/2, 0, − 2/2).
7. z = 0, x = y = 2λ/3. Then the fourth equation becomes
√
8λ2 /9 = 1 ⇐⇒ λ = ±3 2/4,
√ √ √ √
and we obtain the points ( 2/2, 2/2, 0) and (− 2/2, − 2/2, 0).
8. x = y = z = 2λ/3. Then the fourth equation becomes
√
12λ2 /9 = 1 ⇐⇒ λ = ± 3/2,
√ √ √ √ √ √
and we obtain the points ( 3/3, 3/3, 3/3) and (− 3/3, − 3/3, − 3/3).
97
A list of the values of f at the above points is
√ √ √ √
1, −1, 1/ 2, −1/ 2, 1/ 3, −1/ 3,
so the minimum and the maximum of f on the sphere are −1 and 1, respectively. The minimum
is attained at the points of intersection of the sphere with the negative directions of the coordinate
axes and the maximum is attained at the points of intersection of the sphere with the positive
directions of the coordinate axes. 

12.3. Lagrange multipliers in the case of two constraints**

Sometimes, one needs to find the extrema of a function f (x, y, z) subject to multiple constraints:
say, g(x, y, z) = k, h(x, y, z) = m. The variant of the method of Lagrange multipliers that addresses
this question is as follows:
1. Find all the values x, y, z, λ, and µ such that

∇ f (x, y, z) = λ∇g(x, y, z) + µ∇h(x, y, z)

g(x, y, z) = k .
h(x, y, z) = m


That is, solve this system of five equations in the five unknowns x, y, z, λ, µ.
2. Evaluate f at all the points (x, y, z) found in Step 1. The maximum of f is the largest
among these numbers; the minimum of f is the smallest among them.
Example 12.3. Find the maximum and minimum values of
f (x, y, z) = 3x − y − 3z
on the intersection curve of the plane x + y − z = 0 and the elliptic cylinder x2 + 2z2 = 1.
Solution. Let
g(x, y, z) = x + y − z, h(x, y, z) = x2 + 2z2 .
The partials of f, g, h are
f x = 3, fy = −1, fz = −3,
g x = 1, gy = 1, gz = −1,
h x = 2x, hy = 0, hz = 4z,
so we have to solve the system
3 = λ + 2µx λ = −1 λ = −1
  

  
−1 = λ 2µx = 4  x = 2µ








 −1

−3 = −λ + 4µz ⇐⇒ 4µz = −4 ⇐⇒ z = −µ−1

x+y−z=0 y=z−x y = −3µ−1

 
 


 
 

  
x + 2z2 = 1 x + 2z2 = 1 x + 2z2 = 1
 2  2  2

Substituting the expressions for x and z into the fifth equation, we find that µ must satisfy the
equation
√
4µ−2 + 2µ−2 = 1 ⇐⇒ µ2 = 6 ⇐⇒ µ = ± 6.
98
 √ √ √ √ √ √
These give the points (2/ 6, −3/ 6, −1/ 6) and (−2/ 6, 3/ 6, 1/ 6). Since
√ √
   
2 −3 −1 12 −2 3 1
f √ , √ , √ = √ =2 6 and f √ , √ , √ = −2 6,
6 6 6 6 6 6 6
√ √
we conclude that the maximum of f on the given curve is 2 6 and that its minimum is −2 6. 
Exercises
Use the method of Lagrange multipliers to find the maximum and minimum values of the function f subject to the
given constraint.
12.1. f (x, y) = x2 y; x2 + y4 = 1 12.3. f (x, y, z) = xyz; x2 + 2y2 + z2 = 3
12.2. f (x, y, z) = 3x − 2y + z; 3x + 2y + z = 24
2 2 2

12.4. Use the method of Lagrange multipliers to find the minimum of the function f (x, y, z) = xy + yz + zx subject to
the constraints xy = 1, z = x2 + y2 + 1, and x > 0. What can you say about the maximum of f subject to the same
constraints?
12.5. Use the method of Lagrange multipliers to find the point(s) on the cone (z − 2)2 = x2 + y2 that is closest to the
point (1, 1, 2).
12.6. Use the method of Lagrange multipliers to find three positive numbers whose sum is S , where S > 0, and whose
product is maximal.
12.7. Maximize ni=1 xi yi subject to the constraints ni=1 xi2 = 1 and ni=1 y2i = 1.
P P P

99
 LECTURE 13

Double Integrals over Rectangles

In this lecture, we generalize the idea of the definite integral (see Appendix B) to functions of
two variables.

13.1. Definition
Let f (x, y) be a bounded function on the rectangle R,
R = [a, b] × [c, d] = (x, y) : a ≤ x ≤ b, c ≤ y ≤ d .


For each n ≥ 1, let

x0 = a, x1 = a + ∆x, x2 = a + 2∆x, ..., xn = a + n∆x = b, ∆x = (b − a)/n,
be the points that divide the interval [a, b] into n subintervals of equal lengths; and let
y0 = c, y1 = c + ∆y, y2 = c + 2∆y, ..., yn = c + n∆y = d, ∆y = (d − c)/n,
be the points that divide the interval [c, d] into n subintervals of equal lengths. Also, pick n2 sample
points (xi∗j , y∗i j ) so that
xi−1 ≤ xi∗j ≤ xi , y j−1 ≤ y∗i j ≤ y j (1 ≤ i, j ≤ n).
Then the double integral of f over R is defined by
ZZ Xn X n
f (x, y) dA = lim f (xi∗j , y∗i j )∆A, ∆A = ∆x∆y, (13.1)
R n→∞
i=1 j=1

provided that the limit exists. The sums on the right side of (13.1) are called Riemann sums of f
on R.
Remarks. 1. Note that the vertical lines x = xi and y = y j split the rectangle R into n2 congruent
subrectangles, which we may label according to their upper right corner points: if 1 ≤ i, j ≤ n,
then Ri j is the subrectangle whose upper right corner is the point (xi , y j ). With this labelling of
the subrectangles, we can report that each sample point (xi∗j , y∗i j ) lies in the subrectangle Ri j with
the same indices and there is exactly one sample point per subrectangle (see the left side of Figure
13.1). Furthermore, the quantity ∆A in (13.1) is equal to the common area of the rectangles Ri j :
one n2 th part of the area of R. Thus, each term in the Riemann sum of f is the product of a sample
value of f in one of the rectangles Ri j and the area of that rectangle. As the latter product is the
volume of a rectangular box with base Ri j and height f (xi∗j , y∗i j ), the entire Riemann sum equals the
volume of a solid built by sticking together n2 rectangular boxes (see the right side of Figure 13.1).

2. The limit in (13.1) is supposed to exist and be independent of the choice of the sample
points. This is not necessarily true for any old function, but it does hold for functions that are
piecewise continuous in R (see §6.7 for the definition of piecewise continuity).
101
 y
d

yj
(xi∗j , y∗i j )

c
x
a xi b

Figure 13.1. The sample points and solid volume of a Riemann sum
RR
3. If f (x, y) ≥ 0 on R, then R f (x, y) dA represents the volume of the solid whose upper
boundary is the surface z = f (x, y) and whose lower boundary is R. That is,
ZZ
f (x, y) dA = vol. (x, y, z) : (x, y) ∈ R, 0 ≤ z ≤ f (x, y)

R
= vol. (x, y, z) : a ≤ x ≤ b, c ≤ y ≤ d, 0 ≤ z ≤ f (x, y) .


4. From now on, we shall use the notation |D| to denote the area of a plane region D. Let us
divide both sides of (13.1) by |R| = (b − a)(c − d). Since ∆A = |R|/n2 , we have ∆A/|R| = 1/n2 , so
we get
ZZ n n
1 1 XX
f (x, y) dA = lim 2 f (xi∗j , y∗i j ).
|R| R n→∞ n
i=1 j=1

Since n2 i j f (xi j , yi j ) is the average of the sample values f (xi∗j , y∗i j ), we interpret the last equa-
1 ∗ ∗
PP
tion in terms of the average value of f :
ZZ
1
average f (x, y) = f (x, y) dA. (13.2)
(x,y)∈R |R| R
Example 13.1. Evaluate the double integral of the function
(
5 − 2x − 3y if 5 − 2x − 3y ≥ 0,
f (x, y) =
0 if 5 − 2x − 3y ≤ 0,
over the rectangle R = [0, 3] × [0, 2].
Solution. We shall use the interpretation of the double integral as volume. By Remark 3 above,
the value of the integral ZZ
f (x, y) dA
R
is the volume of the solid E that lies between R and the part of the surface z = f (x, y) with (x, y) ∈ R.
The graph of f (x, y) (before the restriction to R) consists of the part of the plane z = 5 − 2x − 3y
with 5−2x−3y ≥ 0 and the part of the xy-plane z = 0 with 5−2x−3y ≤ 0. The line ` with equation
5 − 2x − 3y = 0 splits the rectangle R in two parts: a triangle with vertices (0, 0), ( 25 , 0), and (0, 53 ),
102
 (0, 2) (3, 2)
(0, 35 )
`

(0, 0) (3, 0)
( 52 , 0)

Figure 13.2. The pyramid E and its base

where 5 − 2x − 3y ≥ 0; and a pentagon with vertices ( 52 , 0), (3, 0), (3, 2), (0, 2), and (0, 35 ), where
5 − 2x − 3y ≤ 0 (see Figure 13.2). For points (x, y) lying in the pentagon, the graph z = f (x, y) and
the xy-plane coincide, and for (x, y) in the triangle, the graph z = f (x, y) is a triangle in the plane
z = 5 − 2x − 3y. Thus, the solid E is the pyramid with vertices O(0, 0, 0), A( 25 , 0, 0), B(0, 53 , 0), and
C(0, 0, 5). Its volume is
1 1 1 1 1 5 5 125
|OC| · area(4OAB) = |OC| · |OA| |OB| = 5 · = .
3 3 2 3 223 36
Hence, ZZ
125
f (x, y) dA = .
R 36


13.2. Iterated integrals

In practice, we want to reverse the order of things in the above example, so that we are able to
compute volumes, averages, etc. by interpreting them as double integrals and then evaluating the
latter using some “standard” machinery independent of our interpretation. One such method is the
method of iterated integrals. An iterated integral is an expression of one of the forms
Z d Z b  Z bZ d 
f (x, y) dx dy or f (x, y) dy dx.
c a a c
Sometimes, we write these integrals as
Z dZ b Z bZ d
f (x, y) dxdy and f (x, y) dydx,
c a a c
respectively. Note that these are definite integrals of functions defined by means of definite inte-
grals and not double integrals. Thus, we can evaluate such integrals using the familiar methods
from single-variable calculus.
Example 13.2. Evaluate
Z 1Z 3 Z 3Z 1
3
x y dydx and x3 y dxdy.
0 0 0 0

Solution. We have
Z 1Z 3 Z 1 3 1 1
x3 y2 9x4
Z 
9 9
x y dydx =
3
dx = x dx =
3
= ,
0 0 0 2 0 2 0 8 0 8
103
and Z 3Z 1 3
1 Z 3 3
x4 y y2
Z 
1 9
x y dxdy =
3
dy = y dy = = .
0 0 0 4 0 4 0 8 0 8

13.3. Fubini’s theorem
Notice that in the last example, we evaluated the two possible iterated integrals of x3 y over
[0, 1] × [0, 3] and obtained the same value for both despite following different routes. Is this always
the case? The answer to this question is in the affirmative. Moreover, the common value of the two
iterated integrals is the value of the double integral of f over the rectangle. This fact is known as
Fubini’s theorem.
Theorem 13.1 (Fubini). Let f (x, y) be piecewise continuous in the rectangle R = [a, b] × [c, d].
Then ZZ Z Z b Z Z d d b
f (x, y) dA = f (x, y) dydx = f (x, y) dxdy.
R a c c a
Example 13.3. Find the volume of the solid bounded above by the graph of f (x, y) = x3 y and
below by the rectangle R = [0, 1] × [0, 3].
Solution. We can express this volume by the double integral
ZZ
x3 y dA.
R
By Fubini’s theorem and by Example 13.2,
ZZ Z 1Z 3 Z 3Z 1
9
x y dA =
3
x y dydx =
3
x3 y dxdy = ,
R 0 0 0 0 8
so the volume of the solid is 9/8. 
Example 13.4. Evaluate
√
ZZ
ey x + ey dA, where R = [0, 4] × [0, 1].
R
Solution. By Fubini’s theorem,
Z 4Z 1
√ √
ZZ
e x + e dA =
y y ey x + ey dydx.
R 0 0
By the substitution u = x + ey , du = ey dy, the inner integral is
Z 1 Z e+x  3/2 e+x
√ u 2
e x + e dy =
y
u du =
1/2
= (x + e)3/2 − (x + 1)3/2 .


y
0 1+x 3/2 1+x 3
Hence,
√ 2 4
ZZ Z
e x + e dA =
y
(x + e)3/2 − (x + 1)3/2 dx


y
R 3 0
22  4
= (x + e)5/2 − (x + 1)5/2 0
35
4
= (e + 4)5/2 − e5/2 − 55/2 + 1 ≈ 13.308.



15
104
 Second solution. Let us try to see how things change, if we use the other iterated integral. We
have Z 4 Z 4+ey  3/2 4+ey
√ √ u 2
x + ey dx = u du = = (4 + ey )3/2 − e3y/2 .


0 ey 3/2 ey 3
Hence, Fubini’s theorem yields
Z 1Z 4 Z 1 Z 4
√ √ √
ZZ 
e x + e dA =
y y e x + e dxdy =
y y ey
x + e dx dy
y
R 0 0 0 0
1
2 e
Z Z
2
= e (4 + e ) − e
y y 3/2 3y/2
dy = (4 + u)3/2 − u3/2 du



3 0 3 1
22 e 4
= (4 + u)5/2 − u5/2 1 = (4 + e)5/2 − e5/2 − 55/2 + 1 .
 

35 15

Exercises
13.1. Use a Riemann sum with n = 3 to estimate the value of xy + y2 dA, where R = [1, 2] × [−1, 1]. Take the
RR

R
sample points to be:
(a) the lower right corners of the subrectangles; (b) the centers of the subrectangles.

Evaluate the given double integral by identifying it as the volume of a solid and calculating that volume using geometric
means.ZZ ZZ
13.2. 2 dA, R = [−1, 2] × [0, 3] 13.3. (3 − x) dA, R = [0, 1] × [0, 1]
R R

Evaluate the given iterated integral.

Z 2Z 3 Z 4 Z 3   Z 1 Z 1
x
1 + 2xy2 dxdy + y ln x dydx 3 x+2y dydx


13.4. 13.5. 13.6.
1 1 1 1 y 0 0

Evaluate
ZZ the given double integral.
2+y
ZZ
xy + y2 dA, R = [1, 2] × [−1, 1] dA, R = [0, 1] × [−2, 3]


13.7. 13.10.
ZZR + x
1 2
ZZR
xy
13.8. 2 + y2 + 1
dA, R = [0, 1] × [0, 1] 13.11. (x + y)e2xy dA, R = [1, 2] × [0, 1]
ZZR x ZZR
 √   √ 
sin(3x + 2y) dA, R = [−π, 0] × [− 21 π, 14 π] xy cos x2 + y2 dA, R = 0, π × 0, π


13.9. 13.12.
R R

105
 LECTURE 14

Double Integrals over General Regions

To define the double integral of a bounded function f (x, y) on a bounded region D in the plane,
we use that the definition given in §13.1 applies to any piecewise continuous function F(x, y)
defined on a rectangle (recall Remark 2 in §13.1).

14.1. Definition
Let D be the closed region in R2 bounded by a piecewise smooth curve γ and let f (x, y) be
piecewise continuous inside D. Define the function
(
f (x, y) if (x, y) ∈ D,
F(x, y) =
0 otherwise.
Then the double integral of f over D is defined by
ZZ ZZ
f (x, y) dA = F(x, y) dA,
D R

where R is any rectangle containing D.

Remark. Note that with the assumption that f (x, y) is piecewise continuous in D, F(x, y) is
continuous everywhere except possibly at the points of γ and the (possible) points inside D where
f (x, y) is discontinuous. Thus, if f (x, y) is actually continuous in D, the only discontinuities of
F(x, y) lie on γ. If f (x, y) is a genuine piecewise continuous function, with discontinuities on some
smooth curves γ1 , . . . , γn inside D, then the discontinuities
RR of F(x, y) lie on the curves γ, γ1 , . . . , γn .
In either case, F(x, y) is piecewise continuous and R F(x, y) dA is defined by (13.1).

14.2. Properties of the double integral

We now state several properties of the double integral which are two-dimensional versions of
the properties of the definite integral listed in Appendix B.
Theorem 14.1. Let f (x, y), g(x, y) be piecewise continuous functions and a, b denote constants.
Then ZZ ZZ ZZ
i) [a f (x, y) + bg(x, y)] dA = a f (x, y) dA + b g(x, y) dA.
D D D

ii) If f (x, y) ≤ g(x, y), then

ZZ ZZ
f (x, y) dA ≤ g(x, y) dA.
D D

107
 iii) If D = D1 ∪ D2 , D1 ∩ D2 = ∅, then
ZZ ZZ ZZ
f (x, y) dA = f (x, y) dA + f (x, y) dA.
D D1 D2
ZZ
iv) 1 dA = |D|, |D| being the area of D.
D

Furthermore, similar to the double integral over a rectangle, the double integral over a general
region D can be interpreted as the volume of a solid or the average value of the function. When
f (x, y) ≥ 0, ZZ
f (x, y) dA = vol. (x, y, z) : (x, y) ∈ D, 0 ≤ z ≤ f (x, y) .

D
Also, for any function f (x, y) that is piecewise continuous on D, we have
ZZ
1
average f (x, y) = f (x, y) dA. (14.1)
(x,y)∈D |D| D

14.3. Evaluation of double integrals

We now state two formulas which follow from Fubini’s theorem and which we shall use to
evaluate double integrals. First, consider a region D of the form
D = (x, y) : a ≤ x ≤ b, g1 (x) ≤ y ≤ g2 (x) ,


where g1 and g2 are continuous functions of x. We call such regions type I regions. If f is a
continuous function on D, we have
ZZ Z b  Z g2 (x) 
f (x, y) dA = f (x, y) dy dx. (14.2)
D a g1 (x)

We also define type II regions, which a regions of the form

D = (x, y) : c ≤ y ≤ d, g1 (y) ≤ x ≤ g2 (y) ,


where g1 and g2 are continuous functions of y. If f is a continuous function on such a region, we

have ZZ Z d  Z g2 (y) 
f (x, y) dA = f (x, y) dx dy. (14.3)
D c g1 (y)

Example 14.1. Evaluate ZZ

ye x dA,
√ D
where D = (x, y) : 0 ≤ x ≤ 1, x ≤ y ≤ 1 .


Solution. Since D is a type I region, (14.2) yields

Z 1Z 1 Z 1  2 x 1
1 1
ZZ Z
ye
ye dA =
x
√
ye dydx =
x
dx = (1 − x)e x dx
D 0 x 0 2 √x 2 0
 Z 1 
1  x 1 1  1
e − 2
= (1 − x)e 0 + e dx =
x
− 1 + ex 0 = .

2 0 2 2

108
 (0, 1) (1, 1)
D
√
y= x

(0, 0)

Figure 14.1. The region D

Second solution. Alternatively, we can try to describe D as a type II region and then use (14.3).
We start by sketching D: see Figure 14.1. In our type I description of D,√we first let x vary in the
range 0 ≤ x ≤ 1, and then for each fixed x, we restrict y to the range x ≤ y ≤ 1. To obtain a
type II description, we interchange the roles of x and y: first, we let y vary in 0 ≤ y ≤ 1, and then
2 2
√ restrict x to the range 0 ≤ x ≤ y . Here, the limit y comes from the description
for each fixed y, we
of the curve y = x in the form x = x(y):
√
y = x ⇐⇒ x = y2 .

Thus,
D = (x, y) : 0 ≤ y ≤ 1, 0 ≤ x ≤ y2 ,


and (14.3) yields

ZZ Z 1Z y2 Z 1  y2
Z 1
2
f (x, y) dA = ye dxdy =
x
y e x 0 dy = y ey − 1 dy


D 0 0 0 0
Z 1  u
1
e −u e−2
= eu − 1 = = .

1


2
du
0 2 0 2

Example 14.2. Evaluate the double integral
ZZ
2
ey dA,
D

where D = (x, y) : 0 ≤ x ≤ 1, x ≤ y ≤ 1 .


Solution. Our initial impulse is to write

ZZ Z 1Z 1
y2 2
e dA = ey dydx,
D 0 x
R 2
but this approach is doomed, because the indefinite integral ey dy is not solvable in elementary
functions. On the other hand, if we observe that
D = (x, y) : 0 ≤ y ≤ 1, 0 ≤ x ≤ y ,


109
we have
ZZ Z 1Z y Z 1 Z 1
y2 y2
 y
y2 2
e dA = e dxdy = e x 0 dy = ey y dy
D 0 0 0 0
Z 1  u
1
e e−1
= eu = = .
1


2
du
0 2 0 2

Example 14.3. Evaluate the double integral
ZZ
2xy dA,
D
where D is the triangle with vertices (0, 0), (0, 3), and (1, 2).

(0, 3)
y=3−x
D (1, 2)

y = 2x

(0, 0)

Figure 14.2. The triangle D

Solution. The sides of the triangle D are the lines in the plane determined by the pairs of points
(0, 0) and (0, 3), (0, 0) and (1, 2), and (0, 3) and (1, 2). These are the lines with equations
x = 0, y = 2x, and y = 3 − x,
respectively (see Figure 14.2). Hence, we can express D as a type I region:
D = (x, y) : 0 ≤ x ≤ 1, 2x ≤ y ≤ 3 − x .


Thus, the given double integral is

ZZ Z 1Z 3−x Z 1
2xy dA = 2xy dydx = x (3 − x)2 − 4x2 dx
 
D 0 2x 0
1 1
9x2 3x4
Z 
7
= 2 3
dx = − 2x3 − = .


9x − 6x − 3x
0 2 4 0 4

Example 14.4. Find the volume of the solid E in the first octant that is bounded by the cylinder
y2 + z2 = 4 and by the plane x = 2y.
Solution. We start by sketching the solid E: the left side of Figure 14.3 shows a joint display
of the parts of the cylinder y2 + z2 = 4 and the plane x = 2y that lie in the first octant and how they
intersect. From that display, we see that E is bounded from below by a triangle D in the xy-plane
and from above by the part of the cylinder that lies above D. The triangle D (see the right side of
110
 (0, 2) (4, 2)
D
x = 2y

(0, 0)

Figure 14.3. The solid E and its base

Figure 14.3) is the region in the xy-plane bounded by the y-axis, by the line x = 2y, and by the line
y = 2, along which the cylinder intersects tha xy-plane. Thus, we see that
D = (x, y) : 0 ≤ x ≤ 4, 12 x ≤ y ≤ 2 = (x, y) : 0 ≤ y ≤ 2, 0 ≤ x ≤ 2y .
 

Since the surface of the cylinder is the graph of z = 4 − y2 , it follows that

p
p
E = (x, y, z) : (x, y) ∈ D, 0 ≤ z ≤ 4 − y2 .


Hence, by Remark 3 and Fubini’s theorem,

ZZ p Z 4Z 2 p Z 2Z 2y p
vol.(E) = 4 − y dA =
2 4 − y dydx =
2 4 − y2 dxdy.
D 0 x/2 0 0
We shall use the latter iterated integral, because its evaluation is less technical:
Z 2 Z 2y p Z 2 p Z 0  3/2 4
√ u 16
vol.(E) = 4 − y dxdy =
2 2y 4 − y dy =
2 u (−du) = = .
0 0 0 4 3/2 0 3

Example 14.5. Reverse the order of integration in the iterated integral
Z 1 Z arctan x
f (x, y) dydx.
0 0

First solution. We have

Z 1Z arctan x ZZ
f (x, y) dydx = f (x, y) dA,
0 0 D

where D = (x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ arctan x is the region displayed on Figure 14.4.



(1, π/4)
y = arctan x
D
(0, 0) (1, 0)

Figure 14.4. The region D

111
Since
y = arctan x ⇐⇒ x = tan y,
a reversal of the roles of x and y yields a representation of D as a type II region:
D = (x, y) : 0 ≤ y ≤ π/4, tan y ≤ x ≤ 1 .


Thus, ZZ Z π/4 Z 1
f (x, y) dA = f (x, y) dxdy.
D 0 tan y

Second solution. We now demonstrate an alternative way for changing D into a type II set.
Initially, D is described by the inequalities
0 ≤ x ≤ 1, 0 ≤ y ≤ arctan x.
Since arctan x ≤ π/4 for all x, 0 ≤ x ≤ 1, we can add to the above inequalities the redundant
inequality y ≤ π/4. Thus, we now have D described by
0 ≤ x ≤ 1, 0 ≤ y ≤ arctan x, y ≤ π/4,
or equivalently, by
0 ≤ x ≤ 1, 0 ≤ y ≤ π/4, y ≤ arctan x.
We observe that
y ≤ arctan x ⇐⇒ tan y ≤ x ⇐⇒ x ≥ tan y,
so we can now replace the definition of D by
0 ≤ x ≤ 1, 0 ≤ y ≤ π/4, x ≥ tan y.
Since tan y ≥ 0 for all y in the range 0 ≤ y < π/2, we find that the condition x ≥ 0 is superfluous
and can be dropped. This yields
D = (x, y) : 0 ≤ y ≤ π/4, tan y ≤ x ≤ 1 ,


whence Z 1Z arctan x Z π/4 Z 1

f (x, y) dydx = f (x, y) dxdy.
0 0 0 tan y

Example 14.6. Reverse the order of integration in the iterated integral
Z 2 Z 3−x
f (x, y) dydx.
0 x/2

First solution. We have

Z 2Z 3−x ZZ
f (x, y) dydx = f (x, y) dA,
0 x/2 D

where D = (x, y) : 0 ≤ x ≤ 2, x/2 ≤ y ≤ 3 − x is the region displayed on Figure 14.5. We


can easily convert the equations of the lines to the form x = g(y) (see the figure). However, this
time we have to be careful, because the function x = g(y) that bounds D from the right is neither
x = 3 − y nor x = 2y. The precise representation of D as a type II region is
D = (x, y) : 0 ≤ y ≤ 3, 0 ≤ x ≤ g(y) ,


112
 (0, 3)

y=3−x↔ x=3−y

D (2, 1)
y = x/2 ↔ x = 2y
(0, 0)

Figure 14.5. The region D

where (
2y if 0 ≤ y ≤ 1,
g(y) =
3 − y if 1 ≤ y ≤ 3.
However, this function is not particularly convenient to work with. To obtain a “friendlier” answer,
we split D in two using the horizontal line y = 1. Then each of the two subregions we get is
bounded from the right by a single line, say:
D1 = (x, y) : 0 ≤ y ≤ 1, 0 ≤ x ≤ 2y , D2 = (x, y) : 1 ≤ y ≤ 3, 0 ≤ x ≤ 3 − y .
 

Thus, Z 2Z 3−x Z 1Z 2y Z 3Z 3−y

f (x, y) dydx = f (x, y) dxdy + f (x, y) dxdy.
0 x/2 0 0 1 0

Second solution. Initially, D is described by the inequalities
0 ≤ x ≤ 2, x/2 ≤ y ≤ 3 − x.
On noting that 3 − x ≤ 3 and x/2 ≥ 0, we can add to the above inequalities the redundant condition
0 ≤ y ≤ 3, that is, we have D described by
0 ≤ x ≤ 2, 0 ≤ y ≤ 3, x/2 ≤ y ≤ 3 − x.
Next, we observe that
x/2 ≤ y ⇐⇒ x ≤ 2y, y≤3−x ⇐⇒ x ≤ 3 − y,
so we can further transform the definition of D to
0 ≤ y ≤ 3, 0 ≤ x ≤ 2, x ≤ 2y, x ≤ 3 − y,
or equivalently to
0 ≤ y ≤ 3, 0 ≤ x ≤ min(2, 2y, 3 − y).
This is the equivalent of the description of D using g(y) in our first solution. By comparing 2, 2y,
and 3 − y, we find that the minimum equals 2y or 3 − y according as y ≤ 1 or y ≥ 1. Therefore, D
is the union of the region D1 described by
0 ≤ y ≤ 1, 0 ≤ x ≤ 2y
and the region D2 described by
1 ≤ y ≤ 3, 0 ≤ x ≤ 3 − y.
113
Thus, Z 2Z 3−x Z 1Z 2y Z 3Z 3−y
f (x, y) dydx = f (x, y) dxdy + f (x, y) dxdy.
0 x/2 0 0 1 0


Exercises
Evaluate the given iterated integral.
Z 2 Z 4−y Z 2 Z 2x √3 Z 1 Z y
y2 − 3x dxdy


14.1. 14.2. 4− x2 dydx 14.3. y cos(x − y) dxdy
0 y 1 0 0 0

Evaluate
ZZ the given double integral.
yx2 dA, D = (x, y) : 1 ≤ y ≤ 3, y ≤ x ≤ y2

14.4.
ZZD
3x sin 2y dA, D = (x, y) : 0 ≤ x ≤ π, x/2 ≤ y ≤ x

14.5.
ZZD
2y √
dA, D = (x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ x

14.6.
ZZD x + 1
2

e x/y dA, D = (x, y) : 0 ≤ y ≤ 1, y2 ≤ x ≤ y


14.7.
ZZD
x sin y dA, D = (x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ x2

14.8.
ZZD
x2 + y3 dA, D is the region bounded by y = 2x and y = x2


14.9.
ZZD
14.10. (3x − 4y) dA, D is the triangle with vertices (1, −1), (3, 1), and (1, 2)
ZZD
14.11. xy2 dA, D is the right half of the disk x2 + y2 ≤ 4
ZZD
14.12. x2 dA, D is the triangle with vertices (0, −1), (1, 1), and (0, 2)
ZZD
6x2
dA, D = (x, y) : 1 ≤ y ≤ 2, 0 ≤ x ≤ 2y

14.13. 4+2
D y

Sketch the region of integration and reverse the order of integration.

Z 1Z 1 Z 2 Z 4−x Z 2 Z ln y
14.14. √
f (x, y) dydx 14.16. f (x, y) dydx 14.18. f (x, y) dxdy
0 √x 0 x2 /2 Z1 π/2 0Z
Z 2 Z 4−x2 Z 2 Z 4−y 1

14.15. f (x, y) dydx 14.17. f (x, y) dxdy 14.19. f (x, y) dydx

√ 0 2 0 sin x
0 − 4−x2

Evaluate the given integral by reversing the order of integration.

Z 1Z 1 Z 8Z 2 Z 4 Z 2
2 dxdy 3
14.20. ye−x dxdy 14.21. 14.22. ey dydx
3y x + 1
√ 4 √
0 y2 0 0 x

14.23. Find the average value of the function f (x, y) = y3 over the triangle with vertices (0, 2), (1, 1), and (3, 2).
14.24. Find the volume of the solid lying below the hyperbolic paraboloid z = 4 + x2 − y2 and above the rectangle
R = [0, 1] × [0, 2].
14.25. Find the volume of the solid in the first octant bounded by the cylinder x2 + z2 = 1, the plane y = 3, and the
coordinate planes.
14.26. Find the volume of the solid lying below the plane 4x−y+2z = 6 and above the region bounded by the parabola
x = y2 − 1 and the right half of the circle x2 + y2 = 1.

114
14.27. Find the volume of the solid lying below the paraboloid z = 3xy + 2y + 1 and above the region bounded by the
parabola y = x2 and the line y = 2x + 3.
14.28. Find the volume of the solid lying below the parabolic cylinder z = y2 and above the region bounded by the
parabola y = x2 and the line y = x + 2.
14.29. Find the volume of the solid bounded by the cylinders x2 + y2 = 1 and x2 + z2 = 1.
14.30. Evaluate D 1 − x2 − y2 dA, where D is the unit disk x2 + y2 ≤ 1, by identifying it as the volume of a solid
RR p

and calculating that volume using geometric means.

14.31. Evaluate D x8 sin x + y9 + 1 dA, where D is the disk x2 + y2 ≤ 4.
RR

115
 LECTURE 15

Double Integrals in Polar Coordinates

15.1. The formula

Theorem 15.1. Let f (x, y) be piecewise continuous in a region D in the (Cartesian) xy-plane
and let R be the same region in polar coordinates, that is,
R = (r, θ) : (r cos θ, r sin θ) ∈ D .


Then ZZ ZZ
f (x, y) dA = f (r cos θ, r sin θ)r dA. (15.1)
D R

The formal proof of (15.1) lies beyond the scope of these notes, but we should give at least
some motivation for this formula. Let a, b, α, β be non-negative numbers such that a < b and
0 ≤ α < β ≤ 2π. Consider the case when D is the region bounded by the circles x2 + y2 = a2 and
x2 + y2 = b2 and by the half-lines through the origin forming angles α and β with the positive x-axis
(see Figure 15.1). Note that D contains the points (x, y) in the plane whose polar coordinates (r, θ)
satisfy the inequalities a ≤ r ≤ b and α ≤ θ ≤ β. We recall that the area of D is given by the
formula
b+a
 
area(D) = 2 (β − α) b − a = (β − α)(b − a)
2 2
.
1


(15.2)
2
For each integer n ≥ 1, let
r0 = a, r1 = a + h, r2 = a + 2h, ..., rn = a + nh = b, h = (b − a)/n,
be the points that divide the interval [a, b] into n subintervals of equal lengths; and let
θ0 = α, θ1 = α + δ, θ2 = α + 2δ, ..., θn = α + nδ = β, δ = (β − α)/n,

θ=β θ
r=b β
θ = θj
r = ri
θj
(xi∗j , y∗i j ) (ri∗ , θ∗j )

r=a α
θ=α
r
a ri b

Figure 15.1. The regions D and R, with partitions and sample points
117
be the points that divide the interval [α, β] into n subintervals of equal lengths. We now subdivide
the region D into n2 subregions of similar shapes bounded by the circles of radii a, r1 , r2 , . . . , rn−1 , b
centered at the origin and the half-lines through the origin forming angles α, θ1 , θ2 , . . . , θn−1 , β with
the positive x-axis (see Figure 15.1). Let Di j denote the subregion bounded by the circles of radii
ri−1 and ri and by the half-lines at angles θ j−1 and θ j with the positive x-axis: in polar coordinates,
Di j is given by the inequalities
ri−1 ≤ r ≤ ri , θ j−1 ≤ θ ≤ θ j .
We now pick a sample point (xi∗j , y∗i j ) in each subregion Di j : we choose (xi∗j , y∗i j ) to be the point in
the plane whose polar coordinates are ri∗ = (ri−1 +ri )/2 and θ∗j = (θ j−1 +θ j )/2. Using the subregions
Di j and the sample points (xi∗j , y∗i j ), we form the sum
n X
X n
f (xi∗j , y∗i j )|Di j |, |Di j | = area(Di j ).
i=1 j=1

Note that this sum bears a strong resemblance to a Riemann sum for f (x, y). Indeed, it can be
proved that if f (x, y) is continuous, then
ZZ n X
X n
f (x, y) dA = lim f (xi∗j , y∗i j )|Di j |. (15.3)
D n→∞
i=1 j=1

We now express the double sum in (15.3) in terms of the ri ’s and the θ j ’s. By (15.2) with a = ri−1 ,
b = ri , α = θ j−1 and β = θ j , we have
r + r 
i i−1
|Di j | = (θ j − θ j−1 )(ri − ri−1 ) = δhri∗ .
2
Furthermore, since the polar coordinates of (xi∗j , y∗i j ) are (ri∗ , θ∗j ), we have
xi∗j = ri∗ cos θ∗j , y∗i j = ri∗ sin θ∗j ,
whence f (xi∗j , y∗i j ) = f (ri∗ cos θ∗j , ri∗ sin θ∗j ). Therefore,
n X
X n n X
X n
f (xi∗j , y∗i j )|Di j | = f (ri∗ cos θ∗j , ri∗ sin θ∗j )ri∗ δh.
i=1 j=1 i=1 j=1

Substituting this expression into the right side of (15.3), we obtain

ZZ X n X n
f (x, y) dA = lim f (ri∗ cos θ∗j , ri∗ sin θ∗j )ri∗ δh. (15.4)
D n→∞
i=1 j=1

The double sum on the right side of (15.4) is a Riemann sum for the function f (r cos θ, r sin θ)r on
the rectangle R = [a, b] × [α, β] when the sample points (ri∗ , θ∗j ) are always chosen at the centers of
the subrectangles (see Figure 15.1). Hence,
X n X n ZZ
lim f (ri cos θ j , ri sin θ j )ri δh =
∗ ∗ ∗ ∗ ∗
f (r cos θ, r sin θ)r dA.
n→∞ R
i=1 j=1

The latter equality and (15.4) justify the (15.1) in the special case when D is a sector of the above
kind.
118
 15.2. Examples
Example 15.1. Evaluate the integral
ZZ
(x + y) dA,
D

where D is the region lying between the circles x2 + y2 = 1 and x2 + y2 = 4 and to the left of the
y-axis.

Solution. The polar form of D is

R = (r, θ) : 1 ≤ r ≤ 2, π/2 ≤ θ ≤ 3π/2 .



Hence,
ZZ ZZ Z 2Z 3π/2 
(x + y) dA = (r cos θ + r sin θ)r dA = r cos θ + r sin θ) dθ dr
2 2
D R 1 π/2
Z 2 Z 2
3π/2 14
= r2 sin θ − cos θ π/2 dr = −2 r dr = −
2
.

1 1 3


Example 15.2. Find the area enclosed by the curve r = 4 + 3 cos θ.

Solution. Since 4 + 3 cos θ ≥ 1 for all θ, the given curve “loops” around the origin, enclosing
the region (see Figure 15.2)

R = (r, θ) : 0 ≤ θ ≤ 2π, 0 ≤ r ≤ 4 + 3 cos θ .



If D is the region R in Cartesian coordinates, then we know from property (iv) in Theorem 14.1
that its area is
ZZ
|D| = 1 dA.
D

D
O

Figure 15.2. The region r ≤ 4 + 3 cos θ Figure 15.3. The solid in Example 15.3
119
Hence, by (15.1),
ZZ ZZ Z 2π Z 4+3 cos θ 
|D| = 1 dA = r dA = r dr dθ
D R 0 0
Z 2π Z 2π
1 1
= (4 + 3 cos θ)2 dθ = 16 + 24 cos θ + 9 cos2 θ dθ


2 0 2 0
Z 2π    2π
9 41θ 9 41π
= 8 + 12 cos θ + (1 + cos 2θ) dθ = + 12 sin θ + sin 2θ = .
0 4 4 8 0 2

Example 15.3. Find the volume of the solid that lies inside the sphere x2 + y2 + z2 = 16 and
outside the cylinder x2 + y2 = 4.
Solution. The given solid is symmetric with respect to the xy-plane, so it is twice as large as
the solid E that lies inside the sphere, outside the cylinder, and above the xy-plane (see Figure 15.3).
This solid is the part of the space between the upper hemisphere and the annular region D in the
xy-plane bound by the sphere and the cylinder:
D = (x, y) : 4 ≤ x2 + y2 ≤ 16 .


Sincepthe upper hemisphere of x2 + y2 + z2 = 16 can be described as the graph of the function

z = 16 − x2 − y2 , we can express E as
p
E = (x, y, z) : (x, y) ∈ D, 0 ≤ z ≤ 16 − x2 − y2 .


Hence, ZZ p
vol.(E) = 16 − x2 − y2 dA.
D
The polar form of D is
R = (r, θ) : 2 ≤ r ≤ 4, 0 ≤ θ ≤ 2π ,


and p p √
16 − (r cos θ)2 − (r sin θ)2 = 16 − r2 (cos2 θ + sin2 θ) = 16 − r2 .
Hence, (15.1) yields
ZZ p √ ZZ
vol.(E) = dA =
16 − x2
r 16 − r2 dA
− y2
D R
Z 4  Z 2π √  Z 4 √
= r 16 − r dθ dr = 2π
2 r 16 − r2 dr.
2 0 2

We evaluate the last integral using the substitution u = 16 − r . Then 2

du = −2r dr, u(2) = 12, u(4) = 0,

so 12
Z 0 Z 12 
√
2π 3/2
vol.(E) = 2π 1/2
=π 1/2
du = = 16 3 π.
1


u − 2
du u u
12 0 3 0
√
Thus, the volume of the original solid is 32 3 π. 
120
 Exercises
Evaluate
ZZ the given integral by changing to polar coordinates.
15.1. (x + 3y) dA, D is the disk x2 + y2 ≤ 4
ZZD
e x +y dA, D is the right half of the disk x2 + y2 ≤ 1
2 2
15.2.
ZZD p
15.3. 16 − x2 − y2 dA, D is the part of the annulus 1 ≤ x2 + y2 ≤ 9 lying in the first quadrant
ZZD
xy
15.4. dA, D is the part of the disk x2 + y2 ≤ 9 lying below the line y = x and above the x-axis
ZZD 4 + x 2 + y2
2 2
xye−x −y dA, D is the larger region bounded by the y-axis, the line x+y = 0, and the semicircle x = 4 − y2
p
15.5.
D

15.6. Find the volume of the solid enclosed by the paraboloid z = 4 − x2 − y2 and the xy-plane.
15.7. Find the volume of the solid enclosed by the cone z2 = x2 + y2 and the plane z = 3.
15.8. Find the volume of the solid lying inside both the cylinder y2 + z2 = 4 and the sphere x2 + y2 + z2 = 9.
15.9. Find the volume lying above the paraboloid z = 12 x2 + y2 and inside the ellipsoid x2 + y2 + 4z2 = 2.

15.10. Find the volume lying below the hemisphere z = 4 − x2 − y2 and above the region D in the xy-plane where
p

x2 + y2 ≤ 2x.
15.11. Find the volume of the solid lying below the plane z = x + y and above the disk x2 + y2 ≤ x + y.
15.12. Find the volume of the solid lying inside all three cylinders x2 + y2 = 1, x2 + z2 = 1, and y2 + z2 = 1.

121
 LECTURE 16

Triple Integrals

16.1. Definition
Let f (x, y, z) be a bounded function on the rectangular box B,
[a1 , b1 ] × [a2 , b2 ] × [a3 , b3 ] = (x, y, z) : a1 ≤ x ≤ b1 , a2 ≤ y ≤ b2 , a3 ≤ z ≤ b3 .


For each n ≥ 1, let

x0 = a1 , x1 = a1 + ∆x, x2 = a1 + 2∆x, ..., xn = a1 + n∆x = b1 , ∆x = (b1 − a1 )/n,
be the points that divide the interval [a1 , b1 ] into n subintervals of equal lengths; let
y0 = a2 , y1 = a2 + ∆y, y2 = a2 + 2∆y, ..., yn = a2 + n∆y = b2 , ∆y = (b2 − a2 )/n,
be the points that divide the interval [a2 , b2 ] into n subintervals of equal lengths; and let
z0 = a3 , z1 = a3 + ∆z, z2 = a3 + 2∆z, ..., zn = a3 + n∆z = b3 , ∆z = (b3 − a3 )/n,
be the points that divide the interval [a3 , b3 ] into n subintervals of equal lengths. Also, pick n3
sample points (xi∗jk , y∗i jk , z∗i jk ) so that
xi−1 ≤ xi∗jk ≤ xi , y j−1 ≤ y∗i jk ≤ y j , zk−1 ≤ z∗i jk ≤ zk (1 ≤ i, j, k ≤ n).
Then the triple integral of f over B is defined by
ZZZ n X
X n X n
f (x, y, z) dV = lim f (xi∗jk , y∗i jk , z∗i jk )∆V, ∆V = ∆x∆y∆z, (16.1)
B n→∞
i=1 j=1 k=1

provided that the limit exists. The sums on the right side of (16.1) are called Riemann sums of f
on B.
More generally, if S is a solid region in R3 bounded by a piecewise smooth surface Σ and f is
a piecewise continuous function on S , we define the triple integral of f over S as
ZZZ ZZZ
f (x, y, z) dV = F(x, y, z) dV,
S B

where B is any rectangular box containing S and F is the function

(
f (x, y, z) if (x, y, z) ∈ S ,
F(x, y, z) =
0 otherwise.

Remark. As in the cases of definite and double integrals, the above definitions define the triple
integral for all piecewise continuous functions over solids bounded by piecewise smooth surfaces.
123
Furthermore, similarly to the definite and double integrals, the triple integral of a function is related
to its average in the region of integration:
ZZZ
1
average f (x, y, z) = f (x, y, z) dV. (16.2)
(x,y,z)∈S vol.(S ) S

16.2. Properties of the triple integral

The following theorem lists the three-dimensional versions of the properties of double integrals
from Theorem 14.1.
Theorem 16.1. Let f (x, y, z), g(x, y, z) be piecewise continuous functions and a, b denote con-
stants. Then
ZZZ ZZZ ZZZ
i) [a f (x, y, z) + bg(x, y, z)] dV = a f (x, y, z) dV + b g(x, y, z) dV.
S S S
ii) If f (x, y, z) ≤ g(x, y, z), then
ZZZ ZZZ
f (x, y, z) dV ≤ g(x, y, z) dV.
S S
iii) If S = S 1 ∪ S 2 , S 1 ∩ S 2 = ∅, then
ZZZ ZZZ ZZZ
f (x, y, z) dV = f (x, y, z) dV + f (x, y, z) dV.
S S1 S2
ZZZ
iv) 1 dV = vol.(S ).
S

There is also a three-dimensional version of Fubini’s theorem, which we shall use to evaluate
triple integrals. Before stating it, we need to introduce some terminology. We say that a solid
region S in R3 is of type I, if it can be written in the form
S = (x, y, z) : (x, y) ∈ D, g1 (x, y) ≤ z ≤ g2 (x, y)


for some region D in the xy-plane and some continuous functions g1 and g2 on D. We say that S is
a type II region, if it can be written in the form
S = (x, y, z) : (y, z) ∈ D, g1 (y, z) ≤ x ≤ g2 (y, z)


for some region D in the yz-plane. We say that S is a type III region, if it can be written in the form
S = (x, y, z) : (x, z) ∈ D, g1 (x, z) ≤ y ≤ g2 (x, z)


for some region D in the xz-plane.

Theorem 16.2. Let f (x, y, z) be a piecewise continuous function on a solid region S in R3 .
i) If S is of type I, then
ZZZ ZZ  Z g2 (x,y) 
f (x, y, z) dV = f (x, y, z) dz dA; (16.3)
S D g1 (x,y)

ii) If S is of type II, then

ZZZ ZZ  Z g2 (y,z) 
f (x, y, z) dV = f (x, y, z) dx dA; (16.4)
S D g1 (y,z)

124
 iii) If S is of type III, then
ZZZ ZZ  Z g2 (x,z) 
f (x, y, z) dV = f (x, y, z) dy dA. (16.5)
S D g1 (x,z)

16.3. Evaluation of triple integrals

Example 16.1. Evaluate the triple integral
ZZZ
yz cos(x5 ) dV,
S

where S = (x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ x, x ≤ z ≤ 2x .


Solution. Let D = (x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ x . By (16.3),


ZZZ ZZ  Z 2x 
yz cos(x ) dV =
5 5
yz cos(x ) dz dA
S D x
ZZ ZZ
1 3
= y cos(x ) (2x) − x dA =
5 2 2
yx2 cos(x5 ) dA
 
2 D 2 D
3 1 3 1 4
Z Z x  Z
= yx cos(x ) dy dx =
2 5
x cos(x5 ) dx
2 0 0 4 0
Z 1
3 3 1 3
= cos u du = sin u 0 = sin 1 ≈ 0.126.
20 0 20 20

Example 16.2. Evaluate the triple integral
ZZZ p
x2 + z2 dV,
S

where S is the solid bounded by the paraboloid y = x2 + z2 and the plane y = 4.

Solution. We can write S as a type III region:
S = (x, y, z) : x2 + z2 ≤ 4, x2 + z2 ≤ y ≤ 4 .


Hence, by (16.5),
ZZZ p ZZ  Z 4 p  ZZ

p
x2 + z2 dV = x2 + z2 dy dA = 4 − x 2 − z2 x2 + z2 dA.
S x2 +z2
x2 +z2 ≤4 x2 +z2 ≤4

We can easily compute the last integral by changing to polar coordinates in the xz-plane:
ZZ ZZ Z 2 Z 2π

p
2 2
x2 + z2 dA = 4 − r r r dA =
2
r2 4 − r2 dθdr



4−x −z
0 0
x2 +z2 ≤4 0≤r≤2
2 2
4r3 r5
Z 
128π
= 2π 2 4
dr = 2π = .


4r − r −
0 3 5 0 15

125
 Example 16.3. Compute the volume of the solid T described by the inequalities
x ≥ 0, z ≥ 0, y ≥ 2x, x + 2y + 2z ≤ 2.
Solution. In fact, T is the tetrahedron with vertices (0, 0, 0), (0, 1, 0), (0, 0, 1), and ( 52 , 45 , 0), but
we shall ignore this fact and make a point of using triple integrals and inequalities. By solving the
inequlity x + 2y + 2z ≤ 2 for z, we can rewrite the definition of T as
x ≥ 0, y ≥ 2x, 0 ≤ z ≤ 1 − x/2 − y. (16.6)
Note that the implicit in the inequality for z is the condition
1 − x/2 − y ≥ 0 ⇐⇒ y ≤ 1 − x/2.
Adding the latter inequality to those in (16.6), we get
x ≥ 0, 2x ≤ y ≤ 1 − x/2, 0 ≤ z ≤ 1 − x/2 − y
We now note that the condition x ≤ 52 is implicit in the inequalities for y. Making this condition
explicit, we finally have T described as a type I set:
T = (x, y, z) : 0 ≤ x ≤ 25 , 2x ≤ y ≤ 1 − x/2, 0 ≤ z ≤ 1 − x/2 − y .


We can now compute the volume of T :

ZZZ Z 2/5 Z 1−x/2 Z 1−x/2−y
vol.(T ) = 1 dV = 1 dzdydx
T 0 2x 0
Z 2/5 Z 1−x/2 Z 2/5
1−x/2
= (1 − x/2 − y) dydx = (1 − x/2)y − 12 y2 2x dx

0 2x 0
Z 2/5
= (1 − x/2)2 − 12 (1 − x/2)2 − 2x(1 − x/2) + 2x2 dx


0
Z 2/5 Z 1
1
= + 25 2
= + 1 2
du = .
1 5

1

2

2
− 2
x 8
x dx 2
− u 2
u 5
0 0 15


16.4. Double and triple integrals in Mathematica

Mathematica’s built-in commands Integrate and NIntegrate can evaluate not only defini-
tite integrals in one variable, but also double, triple, and even higher-dimensional integrals. There
are two main ways to use these commands to evaluate double and triple integrals.
Suppose first that the integral can be easily converted to an iterated one: say,
Z b Z g2 (x) Z b Z g2 (z) Z h2 (y,z)
f (x, y) dydx or f (x, y, z) dxdydz.
a g1 (x) a g1 (z) h1 (y,z)

If the integrations are within Mathematica’s reach, we can use Integrate as follows
Integrate[f[x,y], {x,a,b}, {y,g1[x],g2[x]}]
Integrate[f[x,y,z], {z,a,b}, {y,g1[z],g2[z]}, {x,h1[y,z],h2[y,z]}]
If the functions are such that Mathematica cannot evaluate the integrals exactly, we can use
NIntegrate in a similar fashion to obtain a numeric approximation. Here are some examples:
126
 In[1]:= Integrate[x*yˆ2, {x,0,1}, {y,0,1}]
Out[1]:= 1/6

In[2]:= Integrate[x*yˆ2, {x,0,1}, {y,0,x}]

Out[2]:= 1/15

In[3]:= Integrate[y*z*Cos[xˆ5], {x,0,1}, {y,0,x}, {z,x,2x}]

Out[3]:= 3Sin[1]/20

In[4]:= Integrate[Sin[x+y]/(x+y), {x,1,2}, {y,1,2}]

Out[4]:= Cos[2] - 2Cos[3] + Cos[4] + 2SinIntegral[2]
- 6SinIntegral[3] + 4SinIntegral[4]

In[5]:= NIntegrate[Sin[x+y]/(x+y), {x,1,2}, {y,1,2}]

Out[5]:= 0.0619179
Next, suppose that we are interested in an integral over two- or three-dimensional domain D
that is described by one or more inequalities. In such situations, we use a familiar trick: we replace
the original function f by another which is zero outside D. Mathematica provides two commands
to execute this trick: If[] and Boole[]. For example, suppose that we want to integrate the
function x2 − xy + y2 over the region D above the x-axis and the line y = −x and inside the ellipse
2x2 + y2 = 3 (see Figure 16.1). To define the function
(
x2 − xy + y2 if (x, y) ∈ D,
f (x, y) =
0 if (x, y) < D,
we can use either of the following two Mathematica syntaxes:
f[x_,y_] := If[2xˆ2+yˆ2<=3 && x+y>=0 && y>=0, xˆ2-x*y+yˆ2, 0]
f[x_,y_] := (xˆ2-x*y+yˆ2)Boole[2xˆ2+yˆ2<=3 && x+y>=0 && y>=0]
Thus, to evaluate D (x2 − xy + y2 ) dA, we can use (the condition y ≥ 0 is incorporated in the range
RR
of y)
Integrate[(xˆ2-x*y+yˆ2)Boole[2xˆ2+yˆ2<=3 && x+y>=0], {x,-1,2}, {y,0,2}]

1.5

1.0

0.5

0.0
-1.0 -0.5 0.0 0.5 1.0

Figure 16.1. The region above y = 0 and y = −x and insider 2x2 + y2 = 3

127
Here are some examples:
In[6]:= Integrate[(x*y+yˆ2)Boole[xˆ2+yˆ2<1], {x,-1,1}, {y,-1,1}]
Out[6]:= Pi/4

In[7]:= Integrate[x*y*z*Boole[xˆ2+2yˆ2+zˆ2<4],
{x,0,2}, {y,0,2}, {z,0,2}]
Out[7]:= 2/3

In[8]:= Integrate[x*y*z*Boole[x+2yˆ2+3zˆ3<2],
{x,0,2}, {y,0,2}, {z,0,2}]
Out[8]:= "Some gibberish"

In[9]:= NIntegrate[x*y*z*Boole[x+2yˆ2+3zˆ3<2],
{x,0,2}, {y,0,2}, {z,0,2}]
Out[9]:= 0.0468292
Exercises
Evaluate the given iterated integral.
Z 1 Z 1 Z xy+1 Z 2 Z zZ yz
4
16.1. √
6xyz dzdydx 16.2. 4e−z dxdydz
0 x x+y 0 0 0

Evaluate
ZZZthe given triple integral.
2x2 + 3y + 2z dV, S = (x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ x, y ≤ z ≤ 2y


16.3.
ZZZS
e−y dV, S = (x, y, z) : 1 ≤ z ≤ 2, 0 ≤ x ≤ 1/z, 0 ≤ y ≤ xz

16.4.
ZZZS
16.5. xy dV, S lies below the plane z = x + y + 2 and above the right half of the disk x2 + y2 ≤ 4
ZZZ S

16.6. 4xz dV, S is the tetrahedron bounded by the plane 3x + 2y + z = 6 and the coordinate planes
ZZZS
16.7. x cos z dV, S is the solid bounded by the cylinder y = x2 and the planes z = 0, y = 0, z = y, x = 1
ZZZS
y x2 + z2 dV, S is the solid bounded by the cylinder x2 + z2 = 2 and the planes y = 1 and y = 3


16.8.
S

16.9. Use a triple integral to find the volume of the solid bounded by the cylinder 2y = x2 and the planes z = −1,
z = 12 x, and y = 2.

128
 LECTURE 17

Triple Integrals in Cylindrical and Spherical Coordinates

17.1. Triple integrals in cylindrical coordinates

Theorem 17.1. Let f (x, y, z) be piecewise continuous in a solid region S in (Cartesian) space
and let R be the same region in cylindrical coordinates, that is,
R = (r, θ, z) : (r cos θ, r sin θ, z) ∈ S .


Then ZZZ ZZZ

f (x, y, z) dV = f (r cos θ, r sin θ, z)r dV. (17.1)
S R

The target applications of this theorem are triple integrals over type I regions
S = (x, y, z) : (x, y) ∈ D, u1 (x, y) ≤ z ≤ u2 (x, y) ,


where D is a two-dimensional region with a simple polar form. In such situations, we can appeal
to (16.3) to get
ZZZ ZZ  Z u2 (x,y) 
f (x, y, z) dV = f (x, y, z) dz dA.
S D u1 (x,y)
Then, after evaluating the integral over z, we can try to evaluate the resulting double integral using
polar coordinates in the xy-plane. What Theorem 17.1 does is change coordinates in the xy-plane
from Cartesian to polar before the integration over z has been performed.
Example 17.1. Find the volume of the solid S that lies within the cylinder x2 + y2 = 4 and the
sphere x2 + y2 + z2 = 9.
Solution. The given solid lies within the cylinder, below the upper half of the sphere, and
above the lower half of the sphere (see Figure 17.1). Since the upper and lower hemispheres are
the respective graphs of the functions
p p
z = 9 − x2 − y2 and z = − 9 − x2 − y2 ,
we can write S as a type I region:
p p
S = (x, y, z) : x2 + y2 ≤ 4, − 9 − x2 − y2 ≤ z ≤ 9 − x2 − y2 .


The cylindrical coordinates description of the solid then is

√ √
R = (r, θ, z) : 0 ≤ r ≤ 2, − 9 − r2 ≤ z ≤ 9 − r2 .


Therefore, by (17.1),
√
ZZZ ZZZ ZZ  Z 9−r2  ZZ √
vol.(S ) = 1 dV = r dV = √ r dz dA = 2r 9 − r2 dA,
S R D − 9−r2 D
129
 Figure 17.1. The solid S in Example 17.1

where D is the polar rectangle (that is, a disk) D = [0, 2] × [0, 2π]. Applying Fubini’s theorem, we
get
ZZ √ Z 2 Z 2π √ Z 2 √
2r 9 − r dA =
2 2r 9 − r dθdr =
2 4πr 9 − r2 dr.
D 0 0 0

We now make a substitution u = 9 − r . We have2

du = −2r dr, u(0) = 9, u(2) = 5,

so the volume of S is
9 √
Z 2 √ Z 5
2πu3/2

4π(27 − 5 5)
4πr 9 − r dr =
2 2πu (−du) =
1/2
= ≈ 66.265.
0 9 3/2 5 3

Example 17.2. Evaluate ZZZ p
x2 + y2 dV,
B

where B is the ball x2 + y2 + z2 ≤ 4.

Solution. In cylindrical coordinates, the ball is
R = (r, θ, z) : r2 + z2 ≤ 4

√ √
= (r, θ, z) : 0 ≤ r ≤ 2, − 4 − r2 ≤ z ≤ 4 − r2 .


Thus, by (17.1) and (16.3),

√
ZZZ p ZZZ ZZ  Z 4−r2  ZZ √
x2 + y2 dV = r r dV = √ r dz dA =
2
2r2 4 − r2 dA
B R D − 4−r2 D
Z 2Z 2π √ Z 2 √
= 2r2 4 − r2 dθdr = 2π 2r2 4 − r2 dr.
0 0 0

We now substitute r = 2 sin t. Since

√ p
4 − r2 = 4 − 4 sin2 t = 2 cos t, dr = (2 cos t)dt, t(0) = 0, t(2) = π/2,
130
we get
Z 2 √ Z π/2
2π 2r 4 − r2 dr = 4π
2
(2 sin t)2 (2 cos t) (2 cos t)dt
0 0
Z π/2 Z π/2
= 16π (2 sin t cos t) dt = 16π
2
(sin 2t)2 dt
0 0
Z π/2
π/2
= 8π (1 − cos 4t) dt = 8π t − 4 sin 4t 0 = 4π2 .
 1
0


17.2. Triple integrals in spherical coordinates

Theorem 17.2. Let f (x, y, z) be piecewise continuous in a solid region S in (Cartesian) space
and let R be the same region in spherical coordinates, that is,
R = (ρ, θ, φ) : (ρ cos θ sin φ, ρ sin θ sin φ, ρ cos φ) ∈ S .


Then
ZZZ ZZZ
f (x, y, z) dV = f (ρ cos θ sin φ, ρ sin θ sin φ, ρ cos φ)ρ2 sin φ dV. (17.2)
S R

Example 17.3. Evaluate

ZZZ p
2 2 2
x2 + y2 + z2 e−x −y −z dV,
B

where B is the ball x + y + z ≤ a .2 2 2 2

Solution. In spherical coordinates, x2 + y2 + z2 = ρ2 and

B = (ρ, θ, φ) : ρ ≤ a = [0, a] × [0, 2π] × [0, π],


so (17.2) yields
ZZZ p Z aZ 2π Z π
2 2 2 2
x2 + y2 + z2 e−x −y −z dV = ρe−ρ ρ2 sin φ dφdθdρ
B 0 0 0
Z aZ 2π π
2
= ρ3 e−ρ − cos φ 0 dθdρ
0 0
Z a Z 2π Z a
3 −ρ2 2
= 2ρ e dθdρ = 4π ρ3 e−ρ dρ.
0 0 0

We substitute u = ρ in the integral over ρ and get

2

ZZZ p Z a2 Z a2
−x2 −y2 −z2
x +y +z e dV = 2π ue du = 2π
−u
u d − e−u


2 2 2
B 0 0
2
Z a2
−u a 2 2

= − 2πue + 2π e−u du = 2π 1 − e−a − a2 e−a .

0
0

131
 Example 17.4. Evaluate ZZZ
y2 z dV,
S
where S is the solid that lies between the spheres ρ = 1 and ρ = 2 and below the cone φ = π/4
(see Figure 17.2).
Solution. In spherical coordinates, the conditions describing S can be expressed as
1 ≤ ρ ≤ 2, π/4 ≤ φ ≤ π,
so in spherical coordinates S is the rectangular box [1, 2] × [0, 2π] × [π/4, π]. Then, by (17.2),
ZZZ Z 2 Z 2π Z π
y z dV =
2
(ρ sin φ sin θ)2 (ρ cos φ)ρ2 sin φ dφdθdρ
S 1 0 π/4
Z 2π Z π  Z 2 
= 3
sin φ cos φ sin θ 2 5
ρ dρ dφdθ
0 π/4 1
Z 2π Z π  2
= sin3 φ cos φ sin2 θ 61 ρ6 1 dφdθ
0 π/4
Z 2π  Z π 
21
= 2
sin θ 3
sin φ cos φ dφ dθ.
2 0 π/4
We substitute u = sin φ in the integral over φ and get
Z π Z 0  4 1/ √2
u 1
sin3 φ cos φ dφ = √ u du = −
3
=− .
π/4 1/ 2 4 0 16
Hence,
21 2π 2 21 2π
ZZZ Z Z
21π
y z dV = −
2
sin θ dθ = − (1 − cos 2θ) dθ = − .
S 32 0 64 0 32


Figure 17.2. The solid in Example 17.4 Figure 17.3. The solid in Example 17.5
132
 the solid S that lies inside the sphere x2 + y2 + z2 = 4, above
Example 17.5. Find the volume ofp
the xy-plane, and below the cone z = x2 + y2 .
Solution. We first derive the spherical description of S . In Cartesian coordinates, the solid is
defined by the inequalities
p
x2 + y2 + z2 ≤ 4, 0 ≤ z ≤ x2 + y2 .
In spherical coordinates, these become
ρ2 ≤ 4, 0 ≤ ρ cos φ ≤ ρ sin φ,
or equivalently,
0 ≤ ρ ≤ 2, 0 ≤ cos φ ≤ sin φ.
The conditions on φ restrict it to the range π/4 ≤ φ ≤ π/2, so the spherical description of S is
[0, 2] × [0, 2π] × [π/4, π/2]. The solid is displayed at Figure 17.3. We now use Theorem 16.1 to
express the volume of S as a triple integral and then appeal to (17.2) to convert the triple integral
to spherical coordinates. We obtain
ZZZ Z π/2 Z 2π Z 2
vol.(S ) = 1 dV = ρ2 sin φ dρdθdφ
S π/4 0 0
Z π/2 Z 2π Z π/2
8 16π
= sin φ dθdφ = sin φ dφ
3 π/4 0 3 π/4
√
16π  π/2 8 2π
= − cos φ π/4 = .
3 3

Exercises
Evaluate
ZZZthe given integral by changing to cylindrical coordinates.
p
17.1. x2 + y2 dV, S is bounded by the xy-plane, the cylinder x2 + y2 = 1, and by the paraboloid z = 4 − x2 − y2
ZZZ S

17.2. y dV, S lies to the right of the xz-plane and is bounded by the cylinders x2 + y2 = 1 and x2 + y2 = 9 and
S
planes z = 1 and z = 4 + x
by theZZZ
17.3. 1 dV, S is the smaller solid bounded by the plane z = 1 and the sphere x2 + y2 + z2 = 2
S

Evaluate
ZZZthe given integral by changing to spherical coordinates.
p
17.4. x2 + y2 dV, S lies above the xy-plane and within the sphere x2 + y2 + z2 = 4
ZZZS
17.5. xz dV, S lies between the spheres x2 + y2 + z2 = a2 and x2 + y2 + z2 = 2a2 in the first octant
ZZZS
17.6. 1 dV, S lies above the cone z2 = x2 + y2 and below the sphere x2 + y2 + (z − 2)2 = 4
ZZZS
xy + z dV, S lies above the cone z = x2 + y2 and below the sphere x2 + y2 + z2 = 4

p
17.7.
S

Evaluate the given iterated integral by changing to cylindrical or spherical coordinates.

Z 1 Z √1−x2 Z 2−x2 −y2 Z 2 Z √4−x2 Z √9−x2 −y2
2 3/2
x +y
2


17.8. dzdydx 17.10. √ 2 2 xyz dzdydx
0 0 x2 +y2 0 0 4−x −y
√ 2
√ 2 2
Z Z 2 Z 4−y 4−x −y
17.9. z dzdxdy
0 0 0
133
Use a triple integral to find the volume of the given solid.
17.11. The solid bounded by the paraboloid y = −x2 − z2 and the plane y = −9.
17.12. The smaller solid bounded by the plane z = 1 and the sphere x2 + y2 + z2 = 2.
17.13. The solid above the cone φ = π/6 and below the sphere ρ = 4 cos φ.

134
 LECTURE 18

Applications of Double and Triple Integrals

We have already seen one application of the double and triple integrals: both can be used to
compute volumes of solids. In this lecture, we describe some other applications.

18.1. Surface area

Let Σ be a smooth parametric surface, given by
r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 ((u, v) ∈ D).
We can use a double integral to calculate the surface area of Σ. However, we first should define
what “surface area” is. For that, we draw inspiration from the definition of arc length in §5.6.
For simplicity, we shall assume that D is a rectangle in the uv-plane, say: [a, b] × [c, d]. We
can treat the general case similarly, but there are more technical details to pay attention to and little
extra insight to gain. For each n ≥ 1, let a = u0 < u1 < · · · < un = b be the points that divide the
interval [a, b] into n subintervals of equal lengths; and let c = v0 < v1 < · · · < vn = d be the points
that divide the interval [c, d] into n subintervals of equal lengths. The points
(un , v0 ) (un , v1 ) ··· (un , v j ) · · · (un , vn ),
.. .. .. ..
. . . .
(ui , v0 ) (ui , v1 ) ··· (ui , v j ) · · · (ui , vn ),
.. .. .. ..
. . . .
(u1 , v0 ) (u1 , v1 ) ··· (u1 , v j ) · · · (u1 , vn ),
(u0 , v0 ) (u0 , v1 ) ··· (u0 , v j ) · · · (u0 , vn ),
define a partition of the rectangle D into n2 congruent subrectangles. The vector function r maps
each of those small rectangles to a small “patch” on Σ and each grid point (ui , v j ) to a point Pi j on
the surface. The patches form a partition of Σ into n2 pieces, which we label as Σi j , 1 ≤ i, j ≤ n,
so that Σi j is the image of the rectangle that has (ui , v j ) as its upper right corner point. We can
approximate the area of Σi j by the area of the parallelogram Πi j three of whose vertices are the
points Pi−1, j , Pi, j−1 , and Pi j . Recall that the area of the parallelogram Πi j is given by
−−−→ −−−→
|Πi j | = Pi j Pi−1, j × Pi j Pi, j−1 .
It follows that the total area An of all the parallelograms Πi j is
n X
n n X
n
−−−→
Pi j Pi−1, j × −
−−→
X X
An = |Πi j | = Pi j Pi, j−1 .
i=1 j=1 i=1 j=1

We define the surface area of Σ as the limit of the sequence {An }∞

n=1 , if it exists. Using the above
definition and the properties of differentiable functions and double integrals, we can prove the
following formula for the surface area of a piecewise smooth surface.
135
 Theorem 18.1. Let Σ be a smooth parametric surface, given by

r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 ((u, v) ∈ D).

The surface area of Σ is given by

ZZ
area(Σ) = kru × rv k dA. (18.1)
D

If Σ is the graph of z = f (x, y), (x, y) ∈ D, then using the parametrization

r(x, y) = xe1 + ye2 + f (x, y)e3 ((x, y) ∈ D),

we can express the area of Σ as

ZZ q
area(Σ) = 1 + f x2 + fy2 dA. (18.2)
D

Example 18.1. Find the area of the part of the plane 2x + 5y + z = 10 that lies inside the cylinder
x2 + y2 = 9.
Solution. We apply (18.2) with f (x, y) = 10 − 2x − 5y and D the disk x2 + y2 ≤ 9. The desired
area is then
ZZ p ZZ √ √
A= 1 + (−2)2 + (−5)2 dA = 30 dA = 30 · 9π,
D D

where 9π is the area of the disk x2 + y2 ≤ 9. 

Example 18.2. Find the area of the surface Σ with parametrization

r(u, v) = cos u(4 + 2 cos v)e1 + sin u(4 + 2 cos v)e2 + 2 sin ve3 (0 ≤ u, v ≤ 2π).
This surface is called a torus and resembles a donut; it is displayed on Figure 18.1

Figure 18.1. A torus

Solution. We have

ru (u, v) = − sin u(4 + 2 cos v)e1 + cos u(4 + 2 cos v)e2 ,

rv (u, v) = −2 cos u sin ve1 − 2 sin u sin ve2 + 2 cos ve3 ,
136
so
ru × rv = 2 sin2 u sin v(4 + 2 cos v)(e1 × e2 ) − 2 sin u cos v(4 + 2 cos v)(e1 × e3 )
− 2 cos2 u sin v(4 + 2 cos v)(e2 × e1 ) + 2 cos u cos v(4 + 2 cos v)(e2 × e3 )
= 2 cos u cos v(4 + 2 cos v)e1 + 2 sin u cos v(4 + 2 cos v)e2 + 2 sin v(4 + 2 cos v)e3
= (8 + 4 cos v) cos u cos ve1 + sin u cos ve2 + sin ve3 ,


p
kru × rv k = |8 + 4 cos v| cos2 u cos2 v + sin2 u cos2 v + sin2 v = 8 + 4 cos v.
Thus, the surface area of Σ is
Z 2π Z 2π Z 2π
(8 + 4 cos v) dudv = 2π (8 + 4 cos v) dv = 32π2 .
0 0 0

Example 18.3. Express the area of the ellipsoid Σ with equation x2 + y2 + 4z2 = 1 in terms of
the number √
+
Z 1√
ln(2 3)
µ= 1 + 3w2 dw = 1 + √ .
0 2 3
Solution. First, we have to parametrize Σ. The unit sphere x2 + y2 + z2 = 1 is the set of all
points (x, y, z) whose spherical coordinate ρ is 1:
x = cos u sin v, y = sin u sin v, z = cos v,
where 0 ≤ u ≤ 2π and 0 ≤ v ≤ π (u and v play the roles of the spherical coordinates θ and
φ). Since (x, y, z) lies on Σ exactly when (x, y, 2z) lies on the unit sphere, we obtain the following
parametrization for the ellipsoid:
r(u, v) = cos u sin ve1 + sin u sin ve2 + 21 cos ve3 ,
where 0 ≤ u ≤ 2π and 0 ≤ v ≤ π. Then
ru (u, v) = − sin u sin ve1 + cos u sin ve2 + 0e3 ,
rv (u, v) = cos u cos ve1 + sin u cos ve2 − 21 sin ve3 ;
ru × rv = − sin2 u sin v cos v(e1 × e2 ) + 12 sin u sin2 v(e1 × e3 )
+ cos2 u sin v cos v(e2 × e1 ) − 21 cos u sin2 v(e2 × e3 )
= − 12 cos u sin2 ve1 − 21 sin u sin2 ve2 − sin v cos ve3
= − 21 sin v cos u sin ve1 + sin u sin ve2 + 2 cos ve3 ;


p
kru × rv k = 12 | sin v| cos2 u sin2 v + sin2 u sin2 v + 4 cos2 v
p √
= 2 sin v sin2 v + 4 cos2 v = 21 sin v 1 + 3 cos2 v.
1

Thus, the surface area of Σ is

Z π Z 2π √ Z π √
A= 1
2
sin v 1 + 3 cos2 v dudv = π sin v 1 + 3 cos2 v dv
0 0 0
Z −1 √ Z 1√
=π 1 + 3w2 (−dw) = 2π 1 + 3w2 dw = 2πµ. 
1 0

137
 Example 18.4. Find the volume and the surface area of the solid bounded by the paraboloids
z = 3x2 + 3y2 and z = 4 − x2 − y2 .

Solution. The given solid has the paraboloid z = 4 − x2 − y2 as its upper boundary and the
paraboloid z = 3x2 + 3y2 as its lower boundary, so it can be described as

S = (x, y, z) : (x, y) ∈ D, 3x2 + 3y2 ≤ z ≤ 4 − x2 − y2 ,



where D is the projection of the solid on the xy-plane (see Figure 18.2). Since the intersection
curve of the two paraboloids is

γ = (x, y, z) : z = 3x2 + 3y2 = 4 − x2 − y2



= (x, y, z) : x2 + y2 = 1, z = 3 ,


we find that D is the unit disk x2 + y2 ≤ 1. Hence, the volume of E is

ZZZ ZZ  Z 4−x2 −y2  ZZ

vol.(S ) = 1 dV = 1 dz dA = 4 − 4x2 − 4y2 dA.


S D 3x2 +3y2 D

The surface area A of the solid is the sum of the surface areas of the two paraboloids, each of which
can be computed using (18.2):
ZZ p ZZ p
A= 1 + (−2x) + (−2y) dA +
2 2 1 + (6x)2 + (6y)2 dA.
D D

Figure 18.2. The solid in Example 18.4

138
Let R = [0, 1] × [0, 2π] be the polar form of the disk D. Converting all three integrals to polar
coordinates, we obtain
ZZ ZZ
vol.(S ) = 4 − 4(x + y ) dA =
2 2
r 4 − 4r2 dA



D R
Z 1  Z 2π  Z 1
1
= 4r − 4r dθ dr = 2π
3
4r − 4r3 dr = 2π 2r2 − r4 0 = 2π,



0 0 0
ZZ p p
A= 1 + 4(x2 + y2 ) + 1 + 36(x2 + y2 ) dA


D
√ √ Z 1  Z 2π √ √
ZZ 
= r 1 + 4r2 + 1 + 36r2 dA = r 1 + 4r2 + 1 + 36r2 dθ dr



R 0 0
Z 1 √ √
Z 5 Z 37 
= 2π r 1 + 4r + 1 + 36r dr = 2π 1/2 1
du + 1/2 1




2 2 u u du
8 72
0 1 1
 3/2 5  3/2 37 
u u π √ √
= 2π + = 37 37 + 45 5 − 10 ≈ 18.366.


12 1 108 1 54


18.2. Density and mass

Double and triple integrals also represent various physical quantities. In this section, we discuss
applications related to the computation of mass and the coordinates of the center of mass of laminas
and solids. Suppose that a lamina occupies a region D in the xy-plane and has density δ(x, y) at a
point (x, y) ∈ D. If δ(x, y) is the constant function δ(x, y) = c, the total mass of the lamina is c|D|,
where |D| denotes the area of D. Note that we can think of this number also as
ZZ ZZ
c dA = δ(x, y) dA.
D D

Using the definition of the double integral as the limit of Riemann sums, we can prove that the latter
formula applies to any continuous function δ(x, y). That is, if the density function δ is continuous
in D, then the mass m of the lamina is given by the formula
ZZ
m= δ(x, y) dA.
D

We can also give formulas for the coordinates of the center of mass of the lamina: if ( x̄, ȳ) is the
center of mass, then
ZZ ZZ
1 1
x̄ = xδ(x, y) dA, ȳ = yδ(x, y) dA.
m D m D
Similarly, if a solid occupies a region S in space and has density δ(x, y, z) at a point (x, y, z) ∈ S ,
then the mass m of the solid is given by the integral
ZZZ
m= δ(x, y, z) dV.
S
139
We can also give formulas for the coordinates of the center of mass of the solid: if ( x̄, ȳ, z̄) is the
center of mass of S , then
ZZZ ZZZ ZZZ
1 1 1
x̄ = xδ(x, y, z) dV, ȳ = yδ(x, y, z) dV, z̄ = zδ(x, y, z) dV.
m S m S m S

Example 18.5. A flat lamina occupies the disk x2 +y2 ≤ 4 and has density δ(x, y) = x+y+x2 +y2 .
Find its mass and the coordinates of its center of mass.
Solution. Let R denote the polar form of the disk D, that is, R = [0, 2] × [0, 2π]. By (15.1), the
mass m of the lamina is
ZZ ZZ
m= x + y + x + y dA =
2 2
r r cos θ + r sin θ + r2 dA



D R
Z 2π  Z 2  Z 2π
= r (cos θ + sin θ) + r dr dθ =
2 3
(cos θ + sin θ) + 4 dθ = 8π.

8


3
0 0 0
Similarly, the coordinates of the center of mass are:
ZZ
1
x̄ = x(x + y + x2 + y2 ) dA
m D
ZZ
1
= r2 cos θ r cos θ + r sin θ + r2 dA


8π R
Z 2π  Z 2 
1
= cos θ r (cos θ + sin θ) + r dr dθ
3 4


8π 0 0
Z 2π
1
= cos θ 4(cos θ + sin θ) + 32


5
dθ
8π 0
Z 2π
1 1
= 2(1 + cos 2θ) + 2 sin 2θ + 325 cos θ dθ = ,


8π 0 2
ZZ
1
ȳ = y(x + y + x2 + y2 ) dA
m D
ZZ
1
= r2 sin θ r cos θ + r sin θ + r2 dA


8π R
Z 2π  Z 2 
1
= sin θ r (cos θ + sin θ) + r dr dθ
3 4


8π 0 0
Z 2π
1
= sin θ 4(cos θ + sin θ) + 325 dθ


8π 0
Z 2π
1 1
= 2 sin 2θ + 2(1 − cos 2θ) + 325 sin θ dθ = .


8π 0 2

Example 18.6. Find the center of mass ofpthe solid H that occupies the upper half of the ball
x + y2 + z2 ≤ 4 and has density δ(x, y, z) = z x2 + y2 + z2 .
2

Solution. The solid H is described by inequalities

x2 + y2 + z2 ≤ 4, z ≥ 0.
140
In spherical coordinates these become ρ ≤ 2 and 0 ≤ φ ≤ π/2, respectively. Thus, by (17.2), the
mass of the solid is
ZZZ p Z 2 Z 2π Z π/2
m= z x2 + y2 + z2 dV = (ρ2 cos φ)(ρ2 sin φ) dφdθdρ
H 0 0 0
Z π/2 Z 2π
32 π/2 2π
Z 2  Z Z
= sin φ cos φ ρ dρ dθdφ =
4
sin φ cos φ dθdφ
0 0 0 5 0 0
64π π/2 64π 1
Z Z
32π
= sin φ cos φ dφ = u du = .
5 0 5 0 5
The coordinates of the center of mass then are
ZZZ
1 p
x̄ = xz x2 + y2 + z2 dV
m H
Z 2 Z 2π Z π/2
5
= (ρ3 cos θ sin φ cos φ)(ρ2 sin φ) dφdθdρ
32π 0 0 0
Z 2 Z π/2  Z 2π 
5
= 5 2
ρ sin φ cos φ cos θ dθ dφdρ = 0,
32π 0 0 0
ZZZ
1 p
ȳ = yz x2 + y2 + z2 dV
m H
Z 2 Z 2π Z π/2
5
= (ρ3 sin θ sin φ cos φ)(ρ2 sin φ) dφdθdρ
32π 0 0 0
Z 2 Z π/2  Z 2π 
5
= 5 2
ρ sin φ cos φ sin θ dθ dφdρ = 0,
32π 0 0 0
ZZZ
1 p
z̄ = z2 x2 + y2 + z2 dV
m H
Z 2 Z 2π Z π/2
5
= (ρ3 cos2 φ)(ρ2 sin φ) dφdθdρ
32π 0 0 0
Z π/2 Z 2π Z 2 
5
= 2
sin φ cos φ 5
ρ dρ dθdφ
32π 0 0 0
Z π/2 Z 2π
5
= sin φ cos2 φ dθdφ
3π 0 0
Z π/2
10 0 2
Z
10 10
= sin φ cos φ dφ =
2
u (−du) = .
3 0 3 1 9


18.3. An improper integral*

In this section, we use double integrals and polar coordinates to show that
√
π
Z ∞
−x2
e dx = .
0 2
141
 Proof. Let Db denote the disk x2 + y2 ≤ b2 and let S b denote the square with vertices (±b, ±b)
(see Figure 18.3). As b → ∞, both Db and S b expand to fill the entire xy-plane, so we figure that
ZZ ZZ ZZ
−x2 −y2 −x2 −y2 2 2
lim e dA = lim e dA = e−x −y dA. (18.3)
b→∞ Db b→∞ Sb R2
(This can be proved rigorously, but it is also intuitively clear, so we won’t dwell on it.) We now
consider each limit separately.
y
Sb (b, b)
Db
x

(−b, −b)

Figure 18.3. The disk Db and the square S b

First, we consider the integral over the disk Db . In polar coordinates, Db is [0, b] × [0, 2π], so
ZZ Z 2π  Z b  Z 2π  Z b2 
−x2 −y2 −r2 1 2

e dA = re dr dθ = e du dθ = π 1 − e−b .
−u
Db 0 0 0 2 0
Taking limits as b → ∞, we deduce that
ZZ
2 2 2

lim e−x −y dA = lim π 1 − e−b = π. (18.4)
b→∞ Db b→∞

On the other hand, if we define Z b

2
I(b) = e−x dx,
−b
we have
ZZ Z bZ b  Z b
−x2 −y2 −x2 −y2 2
e dA = e dx dy = I(b)e−y dy = I(b)I(b) = I(b)2 .
Sb −b −b −b
−x2
Furthermore, since the function e is even, we have
Z b Z b
−x2 2
I(b) = e dx = 2 e−x dx.
−b 0
Hence, 2 2
ZZ  Z b Z ∞
−x2 −y2 −x2 −x2
lim e dA = lim 2 e dx = 4 e dx . (18.5)
b→∞ Sb b→∞ 0 0
Combining (18.3)–(18.5), we obtain that
Z ∞ 2
−x2
4 e dx = π,
0
which is equivalent to desired conclusion. 
142
 Remark. Note that in the above proof we used that the limit
Z b
2
lim e−x dx
b→∞ 0
R∞ 2
exists, or equivalently, that the improper integral 0 e−x dx converges. This can be proved using
the comparison test for improper integrals.
Exercises
Find the area of the given surface Σ.
18.1. Σ is the part of the plane x + 2y + z = 3 that lies in the first octant
18.2. Σ is the part of the paraboloid z = x2 − y2 that lies between the cylinders x2 + y2 = 1 and x2 + y2 = 4
18.3. Σ is the part of the sphere x2 + y2 + z2 = 4 that lies inside the cylinder x2 + y2 = 2
18.4. Σ is the part of the surface z = xy that lies inside the cylinder x2 + y2 = 1
18.5. Σ is the parametric surface given by r(u, v) = (uv)e1 + (u + v)e2 + (u − v)e3 , u2 + v2 ≤ 1
18.6. Σ is the parametric surface given by r(u, v) = (u cos v)e1 + (u sin v)e2 + ve3 , 0 ≤ u ≤ 1, 0 ≤ v ≤ π

Find the mass and the center of mass of the lamina that occupies the region D and has density given by δ(x, y).
18.7. D is bounded by the curve y = e x , the line x = 1, and the coordinate axes; δ(x, y) = y
18.8. D is the disk x2 + y2 ≤ 2y; δ(x, y) is twice the distance from (x, y) to the origin

Find the mass and the center of mass of the given solid S if its density is given by δ(x, y, z).
18.9. S is the cube 0 ≤ x, y, z ≤ 2; δ(x, y, z) = x2 + y2 + z2
18.10. S is the hemisphere x2 + y2 + z2 ≤ 4, z ≥ 0; δ(x, y, z) is half the distance from (x, y, z) to the z-axis
18.11. S is the hemisphere x2 + y2 + z2 ≤ 1, z ≥ 0; δ(x, y, z) = 4z

18.12. Use the value of the integral computed in §18.3 to evaluate the integrals
Z ∞
2
In = x2n e−x dx (n = 1, 2, 3, . . . ).
0
18.13. Use polar coordinates and the value of the integral computed in §18.3 to evaluate
Z ∞Z ∞
|ax + by|e−(x +y )/2 dA.
2 2

−∞ −∞

143
 LECTURE 19

Change of Variables in Multiple Integrals*

In this lecture, we describe the general theory behind the transformations of a double integral
from Cartesian to polar coordinates and of a triple integral from Cartesian to cylindrical or spherical
coordinates. We also show how the “u-substitution” from single-variable calculus fits within the
same general theory.

19.1. Transformations of R2 and R3

Definition. Let D and R be regions in R2 . A function T : D → R is called a transformation of
D onto R if R = T (D), the range of T . That is, a transformation of D is given by a pair of functions
x = f (u, v), y = g(u, v) ((u, v) ∈ D) (19.1)
such that each point (x, y) ∈ R satisfies (x, y) = T (u, v) for some (u, v) ∈ D. The transformation T
is called one-to-one if for any choice of (x, y) ∈ R the equations (19.1) have exactly one solution
(u, v).
Let D and R be solid regions in R3 . A function T : D → R is called a transformation of D onto
R if R = T (D), the range of T . That is, a transformation of D is given by a triple of functions
x = f (u, v, w), y = g(u, v, w), z = h(u, v, w) ((u, v, w) ∈ D) (19.2)
such that each point (x, y, z) ∈ R satisfies (x, y, z) = T (u, v, w) for some (u, v, w) ∈ D. The trans-
formation T is called one-to-one if for any choice of (x, y, z) ∈ R the equations (19.2) have exactly
one solution (u, v, w).
Example 19.1. The functions
x = u + v, y = 2u − 3v ((u, v) ∈ R2 )
define a transformation of R2 onto itself. This transformation is one-to-one. Indeed, if we solve
the above equations for u and v, we find that
u = 0.6x + 0.2y, v = 0.4x − 0.2y.
Hence, in this case, equations (19.1) have exactly one solution for every choice of (x, y). 
Example 19.2. The functions
x = u cos v, y = u sin v ((u, v) ∈ R2 ) (19.3)
define another transformation of R2 onto itself. However, this transformation is not one-to-one,
because for any givenp point (x, y), we can find multiple solutions (u, v) of the above equations.
Indeed, let r = x2 + y2 and θ ∈ [0, 2π) be the polar coordinates of the point (x, y). Then
(u, v) = (r, θ), (u, v) = (r, θ + 4π), and (u, v) = (−r, θ + π) are three different solutions of (19.3).
145
 v
2π

Figure 19.1. Polar coordinates as the domain of a one-to-one transformation

On the other hand, if we restrict (u, v) to the set (see Figure 19.1: on that figure, the origin is
part of the shaded area but the rest of the v-axis is not)
D = (u, v) ∈ R2 : u > 0, 0 ≤ v < 2π ∪ {(0, 0)},


then the functions in (19.3) do define a one-to-one transformation of D onto R2 . Indeed, for any
(x, y) , (0, 0), the only solution of (19.3) that lies in D are the (unique) polar coordinates (r, θ) of
the point (x, y). If (x, y) = (0, 0), then any solution of (19.3) must have u = 0, and the only such
point in D is the origin. 

Example 19.3. The functions

x = u cos v sin w, y = u sin v sin w, z = u cos w (19.4)
define a transformation T of the solid region (the set of spherical coordinates (u, v, w) of points in
space)
D = (u, v, w) ∈ R3 : u ≥ 0, 0 ≤ v ≤ 2π, 0 ≤ w ≤ π


onto R3 . Note that this transformation is not one-to-one, because any point (0, v, w) is a solution of
(19.4) when x = y = z = 0 and any point (z, v, 0) is a solution of (19.4) when x = y = 0 and z > 0.
The trasformation is one-to-one on the region
D0 = (u, v, w) ∈ R3 : u > 0, 0 < v < 2π, 0 < w < π ,


but it is then not a transformation onto R3 . Indeed, the points on the half-plane y = 0, x ≥ 0 are then
not part of the range of T ; any other point in space has unique spherical coordinates that belong to
D0 . Hence, T is a one-to-one transformation of D0 onto R3 with the above half-plane removed. 

19.2. The Jacobian

For the remainder of this lecture, we shall focus on transformations whose component functions
have continuous partial derivatives in the inside their domain (that is, we exclude any boundary
points of D from this requirement).
Definition. Let T : D → R be a (two-dimensional) transformation given by the functions
(19.1). If the functions f and g have continuous first-order partials inside D, then the Jacobian of
146
T is the determinant
∂(x, y) ∂u f (u, v) ∂v f (u, v)
 
= .
∂(u, v)  ∂u g(u, v) ∂v g(u, v) 
Let T : D → R be a (three-dimensional) transformation given by the functions (19.2). If the
functions f, g and h have continuous first-order partials inside D, then the Jacobian of T is the
determinant
∂u f (u, v, w) ∂v f (u, v, w) ∂w f (u, v, w)
 
∂(x, y, z) 
=  ∂u g(u, v, w) ∂v g(u, v, w) ∂w g(u, v, w)  .

∂(u, v, w)  ∂ h(u, v, w) ∂ h(u, v, w) ∂ h(u, v, w) 
u v w

Example 19.4. Compute the Jacobians of the transformations in Examples 19.2 and 19.3.
Solution. For the transformation (19.3), we have
∂(x, y) ∂u (u cos v) ∂v (u cos v) cos v −u sin v 
   
= = = u(cos v)2 + u(sin v)2 = u.
∂(u, v)  ∂u (u sin v) ∂v (u sin v)   sin v u cos v
For the transformation (19.4), we have
 
cos v sin w −u sin v sin w u cos v cos w
∂(x, y, z) 
=  sin v sin w u cos v sin w u sin v cos w  = −u2 sin w.
 
∂(u, v, w)  cos w 0 −u sin w 


19.3. Change of variables in double and triple integrals

In this section, we reach the focal point the lecture: the substitution rules for double and triple
integrals. However, before we state the main results, we need to adjust slightly our notation. So
far in the lecture, we wrote our transformations as T : D → R, which was a rather reasonable
choice of names for a transformation and its domain and range. However, in most applications to
double and triple integrals, we will be interested in transformations T with a predetermined range,
which will already be named D! Hence, from now on, we will reverse the roles of the letters D
and R and write our transformations as T : R → D. Our first theorem states the general form of
the substitution rule for double integrals.
Theorem 19.1. Let D and R be regions in R2 , let F : D → R be a continuous function, and let
T : R → D be a transformation of R onto D given by functions (19.1) with continuous first-order
partials inside R. Suppose also that T is one-to-one inside R.1 Then
 ∂(x, y) 
ZZ ZZ  
F(x, y) dA = F( f (u, v), g(u, v))   dA. (19.5)
D R ∂(u, v) 
Remark. The expression |∂(x, y)/∂(u, v)| in (19.5) represents the absolute value of the Jacobian
of the transformation T .
Example 19.5. Let us apply formula (19.5) to the transformation (19.3) from Example 19.2.
By Example 19.4, we have
 ∂(x, y) 
 
 ∂(u, v)  = |u|.
 

1
This restriction is allowed to fail at boundary points of R.
147
Starting with a set D in the xy-plane, let R be the set of polar coordinates of points in D (in
Lecture #15, we called R the polar description of D). Then R is essentially a subset of the region
D in Figure 19.1 where the transformation (19.3) is one-to-one (R may have boundary points that
are excluded on Figure 19.1). Hence, T is a transformation of R onto D that is one-to-one inside
R. Since u ≥ 0 for (u, v) ∈ R, formula (19.5) becomes
ZZ ZZ
F(x, y) dA = F(u cos v, u sin v)u dA.
D R
Thus, the formula (15.1) for changing double integrals to polar coordinates is a special case of
Theorem 19.1. 
Example 19.6. Evaluate the integral
ZZ
e(x+y)/(x−y) dA,
D
Where D is the quadrilateral with vertices (1, 0), (2, 0), (0, −2), and (0, −1).
Solution. The substitution
u = x + y, v = x − y
will simplify the integrand. With this in mind, we solve the above equations for x and y to obtain
the transformation T given by
x = (u + v)/2, y = (u − v)/2.
The Jacobian of T is
∂(x, y)  12
 1

= 1 1 = − 4 − 4 = −2.
1 1

1
2 
∂(u, v) 2
− 2
Hence, by (19.5), ZZ ZZ
(x+y)/(x−y)
dA = eu/v 1


e 2
dA,
D R
where R is the set in the uv-plane whose image under T is the original quadrilateral D.
Next, we figure the set R. The quadrilateral D in the xy-plane is shown on the right in Fig-
ure 19.2: it is bounded by the x- and y-axes and by the lines y = x − 2 and y = x − 1. The equations
of the two oblique lines are easily described in uv-coordinates. For example, since
y= x−2 ⇐⇒ x−y=2 ⇐⇒ v = 2,
the line y = x − 2 is described by the equation v = 2; y = x − 1 is described by v = 1. Further, the
x-axis has equation
y=0 ⇐⇒ (u − v)/2 = 0 ⇐⇒ v = u,
and the y-axis has equation
x=0 ⇐⇒ (u + v)/2 = 0 ⇐⇒ v = −u.
Hence, R is the quadrilateral in the uv-plane bounded by the lines v = 1, v = 2, v = u, and v = −u.
This quadrilateral is shown on the left in Figure 19.2.
For the purpose of evaluating the double integral over R, we now express R as a type II region
in the uv-plane:
R = (u, v) : 1 ≤ v ≤ 2, −v ≤ u ≤ v .


148
 v y
y= x−1
v=2
R 1 2
x
v=1
v = −u v=u −→ D
−1 y= x−2
u
−2

Figure 19.2. The quadrilateral D and its uv-description R

Hence,
1 2 v u/v
ZZ ZZ Z Z
1
e(x+y)/(x−y)
dA = e dA =
u/v
e dudv
D 2 R 2 1 −v
1 2  u/v u=v 1 2
Z Z
= ve u=−v dv = ve − ve−1 dv


2 1 2 1
2
e − e−1 e − e−1  1 2 2 3(e − e−1 )
Z
= v dv = 2
v 1= .
2 1 2 4

Next, we state the version of Theorem 19.1 for triple integrals.
Theorem 19.2. Let D and R be solid regions in R3 , let F : D → R be a continuous function,
and let T : R → D be a transformation of R onto D given by functions (19.2) with continuous first
order partials inside R. Suppose also that T is one-to-one inside R.2 Then
 ∂(x, y, z) 
ZZZ ZZZ  
F(x, y, z) dV = F( f (u, v, w), g(u, v, w), h(u, v, w))   dV. (19.6)
D R ∂(u, v, w) 
Example 19.7. Let us apply formula (19.6) to the transformation (19.4) from Example 19.3.
By Example 19.4, we have
 ∂(x, y, z)  
 
 ∂(u, v, w)  = − u sin w = u | sin w|.
2
 2
   

Starting with a set D in the xyz-space, let R be the set of spherical coordinates of points in D (the
spherical description of D). With the possible exception of some boundary points, R is a subset
of the region D0 in Example 19.3, where the transformation is one-to-one. Furthermore, since
0 ≤ w ≤ π, we have | sin w| = sin w. Hence, formula (19.6) becomes
ZZZ ZZZ
F(x, y, z) dV = F(u cos v sin w, u sin v sin w, u cos w)u2 sin w dV.
D R
Save for the labeling of the solids and of the variables on the right side of this formula, this is
equation (17.2). We see that the formula for changing triple integrals to spherical coordinates is a
special case of (19.6). 
2
As in Theorem 19.1, this restriction is allowed to fail at boundary points of R.
149
 Figure 19.3. The solid 4x2 + 9y2 + z4 ≤ 1, x ≥ 0, y ≥ 0, z ≥ 0

Example 19.8. Use a triple integral to find the volume of the solid E (see Figure 19.3) in the
first octant bounded by the coordinate planes and the surface 4x2 + 9y2 + z4 = 1.

Solution. The solid E is described by the inequalities

x ≥ 0, y ≥ 0, z ≥ 0, 4x2 + 9y2 + z4 ≤ 1. (19.7)

The substitution
u = 2x, v = 3y, w = z2

transforms the above inequalities to

u ≥ 0, v ≥ 0, w ≥ 0, u2 + v2 + w2 ≤ 1. (19.8)

The latter inequalities can be written quite simply by using spherical coordinates in the uvw space:
if (ρ, θ, φ) are the spherical coordinates of the Cartesian point (u, v, w), then we can rewrite (19.8)
as
0 ≤ θ, φ ≤ π/2, 0 ≤ ρ ≤ 1. (19.9)

Hence, solving (19.7) for x, y, z and replacing (u, v, w) by their spherical coordinates, we find that
the transformation T given by

x = 12 ρ sin φ cos θ, y = 13 ρ sin φ sin θ, z = (ρ cos φ)1/2

maps the rectangular box R described by (19.9) onto E. Furthermore, as before, it follows from
the properties of the spherical coordinates that this transformation is one-to-one inside R.
Next we calculate the Jacobian of the above transformation. We have

∂ρ x = xρ−1 , ∂θ x = −x tan θ, ∂φ x = x cot φ,

∂ρ y = yρ−1 , ∂θ y = y cot θ, ∂φ y = 32 y cot φ,
∂ρ z = 12 zρ−1 , ∂θ y = 0, ∂φ z = − 12 z tan φ.
150
Thus, using some properties of determinants from linear algebra, we find
 −1 
 xρ −x tan θ x cot φ
∂(x, y, z)  −1

=  yρ

y cot θ y cot φ 
∂(ρ, θ, φ)  1 zρ−1 0 − 12 z tan φ
2
   
1 − tan θ cot φ  1 − tan θ cot φ 
= 2 xyzρ 1 cot θ cot φ  = 2 xyzρ 0 cot θ + tan θ
1 −1 
  1 
−1 

 0 
1 0 − tan φ 1 0 − tan φ
 
1 cot φ  1
= 21 xyzρ−1 (cot θ + tan θ)   = xyzρ−1 (cot θ + tan θ)(− tan φ − cot φ)
1 − tan φ 2
= − 12 xyzρ−1 (sin θ cos θ)−1 (sin φ cos φ)−1 = − 181 ρ3/2 sin φ(cos φ)−1/2 .
Hence,
ZZZ ZZZ
1
vol.(E) = 1 dV = ρ3/2 sin φ(cos φ)−1/2 dV
E 18 R
Z π/2 Z π/2 Z 1
1
= ρ3/2 sin φ(cos φ)−1/2 dρdθdφ
18 0 0 0
Z π/2 Z π/2 Z π/2
1 −1/2 2 5/2 1 π
= ρ 0 dθdφ = sin φ(cos φ)−1/2 dφ.
 
sin φ(cos φ) 5
18 0 0 90 0
Note that the last integral is improper near φ = π/2. The substitution u = cos φ changes it to
π π
Z 0 Z 1
vol.(E) = u −1/2
(−du) = lim u−1/2 du
90 1 90 b→0+ b
π 1 π π
= lim+ 2u1/2 b = lim+ 2 − 2b1/2 = .


90 b→0 90 b→0 45

Exercises
Find the region R such that the given funcion T : D → R2 is a transformation of D onto R.
19.1. T (u, v) = (2u, v), D : u2 + v2 ≤ 1 19.4. T (u, v) = (u + 2v, 2u − v), D : 2u2 − v2 ≥ 1
19.2. T (u, v) = (v, u/2), D : u − 2v ≤ 2
2 2
19.5. T (u, v) = (u + v, u − 3v), D : 0 ≤ u, v ≤ 1
19.3. T (u, v) = (u + v, v), D : u2 + v2 ≤ 2

19.6. Let D be the triangle in the xy-plane enclosed by the coordinate axes and the line x + y = 1, and let S be the
square in the uv-plane where 0 ≤ u, v ≤ 1.
(a) Show that the transformation T given by x = u − uv, y = uv transforms S onto D. [Hint: x + y = u.]
(b) Show that T is one-to-one inside S (that is, when 0 < u, v < 1).
(c) Find the Jacobian of T .
Use Theorem 19.1 or 19.2 to evaluate the given integral using the suggested change of variables. You may have to
solve the given equations for x and y to obtain the transformation explicitly.
ZZ
19.7. x2 y dA, where D is the quadrilateral with vertices (0, 0), (1, 1), (1, −3), and (2, −2); x = u + v, y = u − 3v
ZZD
√
19.8. cos x + 2y dA, where D is in the first quadrant and below x + 2y = 2; x + 2y = u2 , y = v
ZZZD

19.9. (1 + x + y + z)−3 dV, where D is in the first octant and below x + y + z = 1; x + y + z = u, y = v, z = w

D

151
19.10. Let D be the quadrilateral in the xy-plane with vertices (0, 0), (1, 1), (5, −1), and (2, −1).
(a) Find a transformation T : R2 → R2 such that T maps the axis u = 0 to the line y = x and the axis v = 0 to
the line in the xy-plane that passes through the origin and the point (2, −1).
(b) Let T be the transformation T found in part (a). Find the region R in the uv-plane that T maps onto the
quadrilateral D.
f : D → R be a continuous function. Use RR
(c) Let T and R be as in parts (a) and (b), and let RR Theorem 19.1 and
the transformation T to RRtransform the integral D f (x, y) dA into an integral of the form R g(u, v) dA.
(d) Use part (c) to evaluate D ln(x + 2y + 1) dA.
19.11. Let R be the region in the first quadrant of the xy-plane bounded by the curves xy = 1, xy = 4, 2x2 − y2 = 1,
and 2x2 − y2 = 6.
(a) Find a region D in the uv-plane such that the functions u = xy and v = 2x2 − y2 define a tranformation T of
R onto D.
(b) Let u > 0 and v > 0. Show that if (x, y) is a solution of the system of equations
xy = u, 2x2 − y2 = v,
then x must be a solution of the equation 2x4 − vx2 − u2 = 0. Show that the latter equation has a unique
positive real solution, and deduce that T is a one-to-one transformation of R onto D.
(c) Find the Jacobian of T .
(d) Use the transformation T to evaluate R (2x2 + y2 ) dA. [Hint: Switch the usual roles of x, y and u, v.]
RR

19.12. In this exercise, we will find the volume of the solid E enclosed by the surface Σ : x2/3 + y2/3 + z2/3 = 4.
(a) Let x = u3 , y = v3 , and z = w3 . Observe that the point (x, y, z) is on/inside Σ if and only if the point (u, v, w)
is on/inside the sphere u2 + v2 + w2 = 4. Use this observation and spherical coordinates in the uvw-space to
show that the formulas
x = ρ3 cos3 θ sin3 φ, y = ρ3 sin3 θ sin3 φ, z = ρ3 cos3 φ
define a transformation T of B onto E, where B is the rectangular box 0 ≤ ρ ≤ 2, 0 ≤ θ ≤ 2π, 0 ≤ φ ≤ π.
Show that T is one-to-one inside B.
(b) Find the Jacobian of the transformation T from part (a).
(c) Express the volume of E as a triple integral and use the transformation T above to evaluate that triple
integral.
19.13. Let p, q, r, s be non-negative integers, and let S be the triangular pyramid (called also simplex) defined by the
inequalities x ≥ 0, y ≥ 0, z ≥ 0, and x + y + z ≤ 1.
(a) Let U denote the cube of side length 1 in the uvw-space defined by the inequalities 0 ≤ u, v, w ≤ 1. Show
that the equations
x + y + z = u, y + z = uv, z = uvw
define a transformation T of U onto S and that this transformation is one-to-one inside U. Is T one-to-one
on U; if not, then at which boundary points does T fail to be one-to-one?
(b) Find the Jacobian of the transformation T from part (a).
(c) Use Theorem 19.2 and the transformation T above to evaluate
ZZZ
x p yq zr (1 − x − y − z) s dV.
S
R1
[Hint: You may want first to evaluate 0 ta (1 − t)b dt, where a, b are non-negative integers.]

152
 LECTURE 20

Vector Fields

20.1. Definition
If D is a region in R2 , a vector field on D is a function F that assigns to each point (x, y) in D a
two-dimensional vector F(x, y). If D is a solid region in R3 , a vector field on D is a function F that
assigns to each point (x, y, z) ∈ D a three-dimensional vector F(x, y, z).
Note that we can express each two-dimensional vector field F in terms of a pair of functions of
two variables:
F(x, y) = hP(x, y), Q(x, y)i = P(x, y)e1 + Q(x, y)e2 .
Similarly, we can express each three-dimensional vector field F in terms of a triple of functions of
three variables:
F(x, y, z) = hP(x, y, z), Q(x, y, z), R(x, y, z)i = P(x, y, z)e1 + Q(x, y, z)e2 + R(x, y, z)e3 .
Example 20.1. The function F(x, y, z) = x2 e1 + y2 e2 + z2 e3 is a vector field on R3 . The function
F(x, y) = cos(xy)e1 + sin(xy)e2 is a vector field on R2 . 
The standard way to visualize a vector field is to draw the arrows representing F(x, y) (or
F(x, y, z)) applied at the points (x, y) (or (x, y, z)) for a number of choices for (x, y) (or (x, y, z))
from D.
Example 20.2. Sketch twenty vectors of the vector field F(x, y) = 12 e1 − xe2 .
Answer. See Figure 20.1. 

1
1 x
O

Figure 20.1. The vector field F(x, y) = 21 e1 − xe2

153
 20.2. Gradient fields and potential functions
We are already familiar with an important class of vector fields. Suppose that f (x, y) is a
differentiable function on D. In Lecture #9, we defined the gradient ∇ f of f :
∇ f (x, y) = f x (x, y)e1 + fy (x, y)e2 .
We now see that ∇ f is a two-dimensional vector field on D. Similarly, if f (x, y, z) is a differentiable
function on a solid D in R3 , then ∇ f (x, y, z) is a three-dimensional vector field on D. From now
on, we will refer to the gradient of a function f as the gradient vector field of f .
Example 20.3. Find the gradient vector field of the function f (x, y) = x2 + y2 .
Solution. We have
∇ f (x, y) = f x (x, y)e1 + fy (x, y)e2 = 2xe1 + 2ye2 . 
In the subsequent lectures, we will see that vector fields F which are the gradient fields of some
function play an important role in vector calculus. Such vector fields are known as conservative,
that is, F is conservative if there is a function f such that F = ∇ f . If F = ∇ f , then f is called a
potential function of F, or simply a potential.
Example 20.4. By Example 20.3, the vector field F(x, y) = 2xe1 +2ye2 is conservative: F = ∇ f ,
where f (x, y) = x2 + y2 . Similarly, F(x, y, z) = 2ye1 + 2xe2 + e3 is conservative, because F = ∇g,
where g(x, y, z) = 2xy + z. 
Not every vector field is conservative, and it is not very difficult to give an example of a vector
field that is not conservative.
Example 20.5. Show that the vector field F(x, y) = (x2 + y)e1 + y3 e2 is not conservative.
Solution. Suppose that F is conservative, that is, there is some function f (x, y) such that
f x (x, y) = x2 + y, fy (x, y) = y3 .
Then
f xy (x, y) = ∂y (x2 + y) = 1, fyx (x, y) = ∂ x (y3 ) = 0.
Since both mixed partials are constants, they are continuous everywhere. Thus, by Theorem 8.1,
we must have f xy = fyx . However, they are not equal, a contradiction. Therefore, our assumption
that F is conservative must be false. 
We shall use this example as a starting point of our discussion of conservative vector fields.
However, before delving into that, we need to fix some terminology.

20.3. Terminology
Recall that in §11.4 we defined a closed region in the plane to be a region containing all of its
boundary. We call a region D in the plane or in space open, if it contains none of the points on
its boundary. Some examples of open and non-open sets in the plane are shown on Figure 20.2.
On the figure, solid and dashed curves indicate boundary that does and does not belong to the set,
respectively. The first set is open, because it does not contain any of its boundary; the second set is
not open, because it contains all of its boundary (it is closed); and the last set is not open, because
it contains a part of its boundary.
154
 not open open not open
Figure 20.2. Open and non-open sets

We say that a set D, in the plane or in space, is connected, if we can connect every two points
of D with a continuous path that lies entirely in D. Geometrically, this means that D “has only one
piece”. Several examples of connected and disconnected plane sets are displayed on Figure 20.3.

disconnected connected connected

disconnected disconnected connected

Figure 20.3. Connected and disconnected sets

We say that a set D in the plane is simply-connected, if it is connected and every curve γ that
lies entirely in D can be shrunk to a point in D without leaving D. We say that a set D in space is
simply-connected, if it is connected and every surface Σ that lies entirely in D can be shrunk to a
point in D without leaving D. In both cases, it is “closed” curves and surfaces that we really need to
test. Both in two and in three dimensions, simple-connectedness means that D “has only one piece
and no holes”. Also, in both cases, we can check whether a connected set is simply-connected by
looking at the compliment of D to the whole plane or the whole space. If we denote that set by D0 ,
then we must be able to connect each point of D0 to any other point of D0 and to “infinity” by a
path that lies entirely in D0 . Some examples of simply-connected and not simply-connected plane
sets are displayed on Figure 20.4.

simply-connected not simply-connected not simply-connected

Figure 20.4. Simply-connected and not simply-connected sets

155
 20.4. Conservative vector fields in R2
We can easily generalize the solution of Example 20.5. Let F(x, y) = P(x, y)e1 + Q(x, y)e2 be a
conservative vector field and let f (x, y) be its potential, that is,
P(x, y) = f x (x, y), Q(x, y) = fy (x, y).
Then
Py = f xy and Q x = fyx =⇒ Py = Q x .
This is the basic idea underlying the following theorem.
Theorem 20.1. If F(x, y) = P(x, y)e1 + Q(x, y)e2 is a conservative vector field and P(x, y) and
Q(x, y) have continuous first order partials in a domain D, then throughout D we have
∂y P = ∂ x Q. (20.1)
Note that we can use this theorem to show that a given vector field is not conservative. How-
ever, we would rather have a tool for establishing that a vector field is conservative. The next
theorem says that equation (20.1) implies that F is conservative, provided that the region D has a
special property.
Theorem 20.2. If F(x, y) = P(x, y)e1 + Q(x, y)e2 is a vector field that satisfies (20.1) on an open,
simply-connected region D, then F is conservative.
Example 20.6. Determine whether or not the vector field is conservative:
(a) F(x, y) = (2x + y2 )e1 + (2y + x2 )e2 ;
(b) F(x, y) = (x3 + 3xy2 )e1 + (y3 + 3x2 y)e2 ;
−y x
(c) F(x, y) = 2 e1 + 2 e2 .
x +y 2 x + y2
Solution. (a) Let P(x, y) = 2x + y2 and Q(x, y) = 2y + x2 . Then
Py = 2y , 2x = Q x ,
so the vector field is not conservative.
(b) Let P(x, y) = x3 + 3xy2 and Q(x, y) = y3 + 3x2 y. Then
Py = 6xy = Q x for all (x, y) ∈ R2 .
Since R2 is open and simply-connected, it follows that the vector field is conservative.
(c) We have
∂ (−1)(x2 + y2 ) − (−y)(2y) y2 − x2
 
−y
= = ,
∂y x2 + y2 (x2 + y2 )2 (x2 + y2 )2
∂ (1)(x2 + y2 ) − (x)(2x) y2 − x2
 
x
= = 2 ,
∂x x2 + y2 (x2 + y2 )2 (x + y2 )2
for all (x, y) , (0, 0). However, the set of all (x, y) , (0, 0) is not simply-connected. Thus,
Theorem 20.2 says nothing about F. On the other hand, if we consider F on the set D obtained
by cutting the xy-plane along a half-line containing the origin (e.g., we can cut the plane along the
positive x-axis, or the negative y-axis, or the part of the line y = 2x with x ≤ 1), then the theorem
does say that F is conservative on D (which is open and simply-connected). Later we shall show
that this vector field is not conservative on the set of all (x, y) , 0. 
156
 20.5. Curl and conservative vector fields in R3
Definition. If F(x, y, z) = P(x, y, z)e1 + Q(x, y, z)e2 + R(x, y, z)e3 is a vector field on R3 such that
P, Q, R are differentiable, then the curl of F is the vector field on R3 defined by
∂R ∂Q ∂P ∂R ∂Q ∂P
     
curl F = − e1 + − e2 + − e3 .
∂y ∂z ∂z ∂x ∂x ∂y
Remark. A formula for curl F that is easier to remember is that
 
 e1 e2 e3 
curl F = ∇ × F = ∂ x ∂y ∂z  .
 
 P Q R
In reality, this is a terrible abuse of notation, but it helps remember the formula, just as (2.6) helps
us remember the definition of the cross product.
Example 20.7. Find curl F for F(x, y, z) = xez e1 + yez e2 + (x2 z + y2 z)e3 .
Solution. We have
 
 e1 e2 e3 
curl F =  ∂ x ∂y ∂z 

 xe ye x z + y2 z
z z 2

 ∂y ∂z   ∂x ∂z   ∂ x ∂y 
     
= z 2 e − e + e
ye x z + y2 z 1  xez x2 z + y2 z 2  xez yez  3
  

= ∂y (x2 z + y2 z) − ∂z (yez ) e1 − ∂ x (x2 z + y2 z) − ∂z (xez ) e2 + ∂ x (yez ) − ∂y (xez ) e3

= (2yz − yez )e1 − (2xz − xez )e2 + (0 − 0)e3 = y(2z − ez )e1 + x(ez − 2z)e2 . 
We can use the curl of a vector field to state three-dimensional versions of the theorems in the
previous section.
Theorem 20.3. If f (x, y, z) has continuous second-order partials, then
curl(∇ f ) = 0.
In particular, if F is a conservative three-dimensional vector field, then curl F = 0.
Proof. The proof is straightforward:
 
 e1 e2 e3  
  ∂y ∂z   ∂ x ∂z   ∂ x ∂y 
    
curl(∇ f ) =  ∂ x ∂y ∂z  =   e1 − ∂ x f ∂z f  e2 + ∂ x f ∂y f  e3

∂ ∂
       
∂ x f ∂y f ∂z f  y f z f

∂ f ∂2 f ∂ f ∂2 f ∂ f ∂2 f
 2   2   2 
= − e1 − − e2 + − e3 = 0.
∂y∂z ∂z∂y ∂x∂z ∂z∂x ∂x∂y ∂y∂x

The converse theorem is not universally true, but as in the case of two-dimensional fields, it
can be proved for special sets.
Theorem 20.4. If F is a three-dimensional vector field that has continuous partials and satisfies
curl F = 0
on an open, simply-connected set in R3 , then F is conservative.
157
 Example 20.8. Determine whether the vector field F = ye2xy+z e1 + xe2xy+z e2 + ze2xy+z e3 is
2 2 2

conservative.
Solution. We have
 
 e1 e2 e3 
curl F =  ∂ x ∂y ∂z 

2xy+z 2 2xy+z 2 2xy+z2
ye xe ze 
 ∂y ∂z   ∂x ∂z   ∂x ∂y 
     
=  2xy+z2 2 e − 2 2 e + 2 2 e
ze2xy+z  1 ye2xy+z ze2xy+z  2 ye2xy+z xe2xy+z  3
  
xe
2 2 2
= (z(2x) − x(2z))e2xy+z e1 − (z(2y) − y(2z))e2xy+z e2 + (x(2y) − y(2x))e2xy+z e3 = 0.
Since the partials of F are continuous everywhere in R3 (a simply-connected set), it follows that F
is conservative. 
Remark. Note that when F(x, y, z) = P(x, y)e1 + Q(x, y)e2 + 0e3 (that is, when F(x, y, z) is a
two-dimensional vector field that “pretends” to be three-dimensional), then
∂Q ∂P
 
curl F = − e3 ,
∂x ∂y
and the condition curl F = 0 is equivalent to the condition discussed in §20.4.
20.6. Finding potentials
Example 20.9. Find a potential for the vector field F(x, y) = (x3 + 3xy2 )e1 + (y3 + 3x2 y)e2 , if it
exists.
Solution. We know that this field is conservative from Example 20.6, so a potential exists. If
we denote it by f (x, y), then f must satisfy the conditions
f x (x, y) = x3 + 3xy2 , fy (x, y) = y3 + 3x2 y.
In other words, for a fixed y, f (x, y) is an antiderivative of x3 + 3xy2 , and for a fixed x, f (x, y) is an
antiderivative of y3 + 3x2 y. Hence,
x4 3x2 y2
Z
f (x, y) = x3 + 3xy2 dx = + + g(y).
4 2
Note that our answer infolves an arbtrary function of y instead of an arbitrary constant. In order to
find out more about g(y), we differentiate the above expression for f (x, y) with respect to y:
fy (x, y) = ∂y 14 x4 + 23 x2 y2 + g(y) = 3x2 y + g0 (y).

Comparing this with the equality fy = y3 + 3x2 y, we find that

y4
3x2 y + g0 (y) = y3 + 3x2 y ⇐⇒ g0 (y) = y3 ⇐⇒ g(y) = + C.
4
We conclude that f (x, y) = 41 x4 + 32 x2 y2 + 41 y4 + C. 
Note that in the above example we can easily check our answer: direct differentiation gives
∇ f (x, y) = F(x, y), so f (x, y) is potential for F(x, y) by the definition of potential.
x y
Example 20.10. Find a potential for the vector field F(x, y) = 2 e1 + e2 , if it exists.
x + y2 x2 + y2
158
 Solution. The potential f (x, y), if it exists, must satisfy the conditions
x y
f x (x, y) = 2 , f y (x, y) = .
x + y2 x2 + y2
Thus,
d(x2 + y2 ) ln(x2 + y2 )
Z Z
x 1
f (x, y) = dx = = + g(y).
x2 + y2 2 x2 + y2 2
To find g(y), we use fy (x, y):
1 2y y
fy (x, y) = + g0 (y) = 2 ⇐⇒ g0 (y) = 0 ⇐⇒ g(y) = C.
2 x +y
2 2 x + y2
We conclude that f (x, y) = 12 ln(x2 + y2 ) + C. 
Remark. Note that the function f (x, y) = 12 ln(x2 + y2 ) +C in the previous example is a potential
for the vector field F(x, y) for all (x, y) , (0, 0). Therefore, F is conservative in the set R2 − {(0, 0)}.
On the other hand, if we tried to apply Theorem 20.2 to it, we would run into the same trouble as
in the solution of Example 20.6, part (c): this set is not simply-connected.
Example 20.11. Find a potential for the vector field F(x, y, z) = (2xz+y2 )e1 +2xye2 +(x2 +3z2 )e3 ,
if it exists.
Solution. The potential f (x, y, z), if it exists, must satisfy the conditions
f x (x, y, z) = 2xz + y2 , fy (x, y, z) = 2xy, fz (x, y, z) = x2 + 3z2 .
Thus, arguing similarly to the last two examples, we find that
Z
f (x, y, z) = (2xz + y2 ) dx = x2 z + y2 x + g(y, z),
fy (x, y, z) = 2yx + gy (y, z) = 2xy ⇐⇒ gy (y, z) = 0 ⇐⇒ g(y, z) = h(z),
fz (x, y, z) = x2 + h0 (z) = x2 + 3z2 ⇐⇒ h0 (z) = 3z2 ⇐⇒ h(z) = z3 + C.
We conclude that f (x, y, z) = x2 z + y2 x + z3 + C. 
Example 20.12. Find a potential for the vector field F(x, y, z) = e2xy e1 + e2xy e2 + z2 e3 , if it exists.
Solution. Any potential f (x, y, z) must satisfy
f x (x, y, z) = e2xy , fy (x, y, z) = e2xy , fz (x, y, z) = z2 .
From the first of these equations, we find
e2yx
Z
f (x, y, z) = e2xy dx =
+ g(y, z).
2y
Next, we differentiate the last function with respect to y and get
2xe2xy e2xy
 
x 1
fy (x, y, z) = − 2 + gy (y, z) = e2xy
⇐⇒ gy (y, z) = e 2xy
1− + 2 .
2y 2y y 2y
We seem to be running in circles, and we indeed are, because the field is non-conservative. For
example,
∂ x e2xy = 2ye2xy , 2xe2xy = ∂y e2xy ,

so the third component of curl F is non-zero. 

159
 20.7. The divergence of a vector field
Definition. If F(x, y, z) = P(x, y, z)e1 + Q(x, y, z)e2 + R(x, y, z)e3 is a vector field on R3 such that
P, Q, R are differentiable, then the divergence of F is the function defined by
∂P ∂Q ∂R
div F = + + .
∂x ∂y ∂z
Example 20.13. Find div F for F = ez sin(xy)e1 + ez cos(xy)e2 + xyze3 .
Solution. We have
div F = ∂ x (ez sin(xy)) + ∂y (ez cos(xy)) + ∂z (xyz) = yez cos(xy) − xez sin(xy) + xy. 
Theorem 20.5. If F = Pe1 + Qe2 + Re3 has continuous second-order partials, then
div curl F = 0.
Proof. The proof is straightforward:
∂ ∂R ∂Q ∂ ∂P ∂R ∂ ∂Q ∂P
     
div curl F = − + − + −
∂x ∂y ∂z ∂y ∂z ∂x ∂z ∂x ∂y
∂R
2
∂ Q
2
∂P
2
∂R
2
∂Q
2
∂P
2
= − + − + − = 0.
∂x∂y ∂x∂z ∂y∂z ∂y∂x ∂z∂x ∂z∂y

Exercises
20.1. Sketch the vector field F(x, y) = (x − y)e1 + xe2 by drawing arrows in a similar fashion to Example 20.2.
Find the gradient vector field of the given function.
20.2. f (x, y) = e2y sin(xy) 20.3. f (x, y, z) = x2 + y2 + z2 20.4. f (x, y, z) = ln x2 + yz
p

Determine whether or not the given set is open, connected, and simply-connected.
20.5. (x, y) : x > 0, y > 0 20.8. (x, y) : x2 + y2 ≤ 1
 

20.6. (x, y) : x > 0, y ≥ 0 20.9. (x, y) : x2 + y2 < + >

2 2

1 or x y 2
20.7. (x, y) : 1 < x2 + y2 < 2

20.10. (x, y) : xy ≤ 0

Determine whether the vector field F(x, y) is conservative. If it is, find a potential function for it.
20.11. F(x, y) = (x − y)e1 + (x + y)e2 y2
20.14. F(x, y) = e1 + (2y arctan x)e2
20.12. F(x, y) = ye1 + (x + 3y)e2 1 + x2
20.13. F(x, y) = ey e1 + xey e2 20.15. F(x, y) = (4x + sin y)e1 + (sin y + x cos y)e2

Find the curl and the divergence of the given vector field.
20.16. F(x, y, z) = x2 yze1 − 4xyz2 e3 20.18. F(x, y, z) = cos xze1 + 2 sin xye2 + ze2xy e3
2x
20.17. F(x, y, z) = 2 arcsin ye1 + p e2 − 3z2 e3
1 − y2

Determine whether the vector field F(x, y, z) is conservative. If it is, find a potential function for it.
20.19. F(x, y, z) = (yz + x2 )e1 + (xz + 4y)e2 + (xy − 3z2 )e3 20.22. F(x, y, z) = 2 arcsin ye1 + p 2x e2 − 3z2 e3
20.20. F(x, y, z) = y2 cos ze1 + 2xy cos ze2 − xy2 sin ze3 1 − y2
20.21. F(x, y, z) = xey e1 + ye x e2 + xyze3

20.23. Is there a vector field F on R3 such that curl F = xy2 e1 + yz2 e2 + zx2 e3 ? Explain.

160
 LECTURE 21

Line Integrals

21.1. Line integrals of scalar functions

We start with a smooth parametric curve γ, given by a vector function
r(t) = x(t)e1 + y(t)e2 + z(t)e3 (a ≤ t ≤ b).
Recall that the smoothness of γ means that r0 (t) = x0 (t)e1 + y0 (t)e2 + z0 (t)e3 is continuous and
r0 (t) , 0 when a < t < b (see §5.3). Further, suppose that f (x, y, z) is a continuous function on a
set S in R3 that contains the curve γ. In particular, f (x, y, z) is defined at each point (x, y, z) on γ.
For each n ≥ 1, set ∆n = (b−a)/n and define the numbers that partition [a, b] into n subintervals
of equal lengths:
t0 = a, t1 = a + ∆ n , t2 = a + 2∆n , ..., tn = a + n∆n = b.
We write xi = x(ti ), yi = y(ti ), zi = z(ti ) for the coordinates of the point Pi (xi , yi , zi ) that corre-
sponds to the value t = ti of the parameter. The points P0 , P1 , P2 , . . . , Pn partition γ into n arcs:
γ1 , γ2 , . . . , γn , where γi has Pi−1 and Pi as its endpoints. Further, we pick n values of the parameter
t1∗ , t2∗ , . . . , tn∗ so that
t0 ≤ t1∗ ≤ t1 , t1 ≤ t2∗ ≤ t2 , . . . , tn−1 ≤ tn∗ ≤ tn .
These yield n sample points P∗i (xi∗ , y∗i , z∗i ) on the curve with coordinates xi∗ = x(ti∗ ), y∗i = y(ti∗ ),
z∗i = z(ti∗ ). Then the line integral of f along γ is defined by
Z X n
f (x, y, z) ds = lim f (xi∗ , y∗i , z∗i )∆si , ∆si = length(γi ), (21.1)
γ n→∞
i=1

provided that the limit exists.

Recall that, by Theorem 5.1,
Z ti p
∆si = x0 (t)2 + y0 (t)2 + z0 (t)2 dt.
ti−1

Using this formula, the above definition, the properties of continuous functions, and the definition
of the definite integral, it can be shown that for continuous functions f the limit (21.1) always
exists and the line integral can be expressed as a definite integral.
Theorem 21.1. Let γ be a smooth parametric curve, given by the vector function
r(t) = x(t)e1 + y(t)e2 + z(t)e3 (a ≤ t ≤ b).
If f (x, y, z) is continuous on γ, then
Z Z b p
f (x, y, z) ds = f (x(t), y(t), z(t)) x0 (t)2 + y0 (t)2 + z0 (t)2 dt. (21.2)
γ a

161
 Remarks. 1. Using more extensively the language of vector functions, we can write (21.2) in
the form
Z Z b
f (x, y, z) ds = f (r(t))kr0 (t)k dt. (21.3)
γ a

2. As in the cases of definite and multiple integrals, the line integral of a continuous function
along a curve is related to its average value on the points of the curve:
Z
1
average f (x, y, z) = f (x, y, z) ds.
(x,y,z)∈γ length(γ) γ
Example 21.1. Evaluate Z

−1
x2 + y2 + z2 ds
γ

where γ is the helix

x = cos t, y = sin t, z=t (0 ≤ t ≤ T ).
Solution. We have
Z Z T
2 −1

−1 p
x +y +z
2 2
ds = (cos t)2 + (sin t)2 + t2 (− sin t)2 + (cos t)2 + 12 dt


γ 0
√
Z T √ Z T
2 √
= (1 + t ) 1 + 1 dt =
2 −1
dt = 2 arctan T.
0 0 1+t 2

We have already seen on several instances that properties of plane curves or two-dimensional
vectors often appear as special cases of analogous properties of space curves or three-dimensional
vectors. This general principle also applies to line integrals. We just defined the line integral of a
function of three variables along a space curve. Is it possible to define the line integral of a function
of two variables along a plane curve? The answer to this question is in the affirmative. One way
to develop the theory of such line integrals is by repeating the above definitions and reproving the
above theorems with one less variable. Alternatively, we can define two-dimensional line integrals
as a special case of three-dimensional line integrals. We take the latter route.
Observe that a two-variable function f (x, y) is also a three-variable function f (x, y, z) (which
does not really depend on z) and that a plane curve, given by
r(t) = x(t)e1 + y(t)e2 (a ≤ t ≤ b), (21.4)
is also a space curve, given by
r0 (t) = x(t)e1 + y(t)e2 + 0e3 (a ≤ t ≤ b). (21.5)
If γ is the plane curve (21.4) and f (x, y) is continuous on γ, we define the line integral of f along
γ by
Z Z
f (x, y) ds = g(x, y, z) ds,
γ γ0
162
where γ0 is the space curve (21.5) and g(x, y, z) = f (x, y) for all (x, y) in the domain of f and all z.
In particular, by (21.2), we have
Z Z b p
f (x, y) ds = g(x(t), y(t), 0) x0 (t)2 + y0 (t)2 + 02 dt
γ a
Z b p
= f (x(t), y(t)) x0 (t)2 + y0 (t)2 dt.
a
That is, we have the formula
Z Z b Z b p
f (x, y) ds = f (r(t))kr (t)k dt =
0
f (x(t), y(t)) x0 (t)2 + y0 (t)2 dt. (21.6)
γ a a
Note that the vector form of (21.6) and the vector form (21.3) of (21.2) look identical; the only
difference is that the vector function r(t) is two-dimensional in (21.6) and three-dimensional in
(21.3).
Example 21.2. Evaluate Z
xey ds,
γ
where γ is the arc of the curve x = ey from (1, 0) to (e, 1).
Solution. The curve can be parametrized by the equations
x = et , y=t (0 ≤ t ≤ 1).
Hence, by (21.6),
Z Z 1 p Z 1 √
xe ds =
y
e e (e ) + 1 dt =
t t t 2 2 e2t e2t + 1 dt.
γ 0 0

Using the substitution u = e + 1, we find that 2t

du = 2e2t dt, u(0) = 2, u(1) = e2 + 1,

and
e2 +1
1 √ (e2 + 1)3/2 − 23/2
Z Z
2t
+ 1 dt = u1/2 =
1


e e2t 2
du ≈ 7.157.
0 2 3

We can easily extend the defintion of line integrals to piecewise smooth curves. If γ is a
piecewise smooth parametric curve that is the union of the smooth curves γ1 , γ2 , . . . , γn , γ = γ1 ∪
γ2 ∪ · · · ∪ γn , we define
Z Z Z Z
f (x, y, z) ds = f (x, y, z) ds + f (x, y, z) ds + · · · + f (x, y, z) ds.
γ γ1 γ2 γn
Note that all the line integrals on the right side of the formula are along smooth curves and can be
evaluated via the limit in (21.1) or via the integral formulas (21.2), (21.3), or (21.6).
Example 21.3. Evaluate the line integral
Z
xy2 ds
γ
where γ is the triangle with vertices (0, 1, 2), (1, 0, 3), and (0, −1, 0).
163
 Solution. We split γ into three line segments:
• γ1 is the segment connecting (0, 1, 2) and (1, 0, 3):
r(t) = t h0, 1, 2i + (1 − t) h1, 0, 3i = h1 − t, t, 3 − ti (0 ≤ t ≤ 1);
• γ2 is the segment connecting (0, 1, 2) and (0, −1, 0):
r(t) = t h0, 1, 2i + (1 − t) h0, −1, 0i = h0, 2t − 1, 2ti (0 ≤ t ≤ 1);
• γ3 is the segment connecting (1, 0, 3) and (0, −1, 0):
r(t) = t h1, 0, 3i + (1 − t) h0, −1, 0i = ht, t − 1, 3ti (0 ≤ t ≤ 1).
By (21.2),
1 1
√
Z Z p √ Z 3
xy2 ds = 2
(−1) + 1 + (−1) dt = 3 t2 − t 3
dt =


(1 − t)t 2 2 2 ;
γ1 0 0 12
Z Z 1 √
xy2 ds = 0(2t − 1)2 02 + 22 + 22 dt = 0;
γ2 0
√
Z Z 1 √ √ Z 1
11
xy2 ds = 2
12 + 12 + 32 dt = t3 − 2t2 + t dt = .


t(t − 1) 11
γ3 0 0 12
Hence,
Z Z Z √ √ Z
3 11
xy ds =
2
xy ds + xy ds +
2
xy ds = +2
≈ 0.421. 2
γ γ1 γ2 γ3 12 12


21.2. Line integrals of vector fields

Let F be a continuous vector field defined on a piecewise smooth curve γ given by a vector
function r(t), a ≤ t ≤ b. We define the line integral of F along γ by
Z Z
F · dr = (F · T) ds,
γ γ

where T is the unit tangent vector to the curve γ. Recall from §5.4 that the unit tangent vector T(t)
at the point of γ corresponding to the value t of the parameter can be expressed in terms of r(t) by
the formula
r0 (t)
T(t) = 0 .
kr (t)k
Therefore, by (21.3) or (21.6) (depending on the dimension of F and γ),
Z b Z b
r0 (t)
Z 
(F · T) ds = F(r(t)) · 0 kr (t)k dt =
0
F(r(t)) · r0 (t) dt.
γ a kr (t)k a

That is, we have

Z Z Z b
F · dr = (F · T) ds = F(r(t)) · r0 (t) dt. (21.7)
γ γ a
164
 What is the meaning of this kind of line integral? Based on the interpretation that we gave to
the line integral of a function, we can say that
Z Z
1 1
F · dr = (F · T) ds = average (F · T).
length(γ) γ length(γ) γ γ

Recall from §2.7 that the dot product F · T is the size of the projection of F onto the direction of T,
taken with positive or negative sign according as the projection and T point in the same direction
or in opposite directions. Since T(t) is the direction of the curve at the point corresponding to the
value t of the parameter, we conclude that
Z
1
F · dr
length(γ) γ
is the “average projection of F in the direction of the curve”.
Example 21.4. Evaluate Z
ye1 + ze2 + xe3 · dr,


γ
where γ is the helix r(t) = h2 cos t, 2 sin t, 3ti, 0 ≤ t ≤ 2π.
Solution. By (21.7),
Z Z 2π
ye1 + ze2 + xe3 · dr = h2 sin t, 3t, 2 cos ti · h(2 cos t)0 , (2 sin t)0 , (3t)0 i dt


γ 0
Z 2π
= h2 sin t, 3t, 2 cos ti · h−2 sin t, 2 cos t, 3i dt
0
Z 2π
= − 4 sin2 t + 6t cos t + 6 cos t dt


0
Z 2π Z 2π
= −2 (1 − cos 2t) dt + 6 (t + 1) d(sin t)
0 0
Z 2π
2π  2π
= − 2t + sin 2t 0 + 6(t + 1) sin t 0 − 6 sin t dt = −4π.

0

In the above example, the parametrization of γ was given explicitly. However, we often have
only a geometric description of the curve and have to obtain the parametrization ourselves. In such
situations, it is important to remember that (21.7) is sensitive to the orientation of the curve, that
is, the direction in which r(t) traces the curve as t increases. Indeed, if γ1 and γ2 are two parametric
curves that represent the same geometric curve γ but with opposite orientations, then
Z Z
F · dr = − F · dr. (21.8)
γ1 γ2
That is, if we change the orientation of the curve, the line integral changes sign. This fact goes
back to the property of definite integrals that
Z a Z b
f (x) dx = − f (x) dx.
b a
Because of (21.8), a geometric description of the curve must always specify its orientation.
165
 Example 21.5. Evaluate Z
x2 e1 + y2 e2 · dr,


γ

where γ is the right half of the ellipse x + 4y2 = 4 oriented in the counterclockwise direction.
2

Solution. The point (a, b) is on the given ellipse exactly when the point (a, 2b) is on the circle
x + y2 = 4. Thus, we can parametrize γ by
2

x = 2 cos t, y = sin t (−π/2 ≤ t ≤ π/2).

Note that as t increases from −π/2 to π/2, the point (x(t), y(t)) moves along γ in the counterclock-
wise direction. Applying (21.7) to the above parametrization, we get
Z Z π/2
x e1 + y e2 · dr =
2 2
(2 cos t)2 e1 + (sin t)2 e2 · (2 cos t)0 e1 + (sin t)0 e2 dt




γ −π/2
Zπ/2
= 4 cos2 te1 + sin2 te2 · − 2 sin te1 + cos te2 dt



−π/2
Zπ/2
= − 8 cos2 t sin t + sin2 t cos t dt


−π/2
Z π/2 Z π/2
= −8 cos t sin t dt +
2
sin2 t cos t dt
−π/2 −π/2
Z π/2 Z 1
2
=0+2 sin t cos t dt = 2
2
u2 du = (u = sin t).
0 0 3


Suppose that F(x, y, z) = f (x, y, z)e1 . Then (21.7) yields

Z Z b Z b
F · dr = f (r(t))e1 · x (t)e1 + y (t)e2 + z (t)e3 dt =
0 0 0
f (r(t))x0 (t) dt.



γ a a
R
Another notation for this integral is γ f (x, y, z) dx, that is,
Z Z Z b
f (x, y, z) dx = ( f e1 ) · dr = f (r(t))x0 (t) dt. (21.9)
γ γ a

Similarly, applying (21.7) to the vector fields F(x, y, z) = f (x, y, z)e2 and F(x, y, z) = f (x, y, z)e3 ,
we obtain
Z Z Z b
f (x, y, z) dy = ( f e2 ) · dr = f (r(t))y0 (t) dt (21.10)
γ γ a

and
Z Z Z b
f (x, y, z) dz = ( f e3 ) · dr = f (r(t))z0 (t) dt. (21.11)
γ γ a

These three integrals are called the line integrals of f with respect to x, y, and z, respectively. Since
they are derived from (21.7), these integrals are also sensitive to the orientation of the curve γ
166
and change signs when we reverse the orientation of the curve. For example, if γ1 and γ2 are two
parametric curves that represent the same geometric curve with opposite orientations, then
Z Z
f (x, y, z) dx = − f (x, y, z) dx.
γ1 γ2

Note that if F(x, y, z) = P(x, y, z)e1 + Q(x, y, z)e2 + R(x, y, z)e3 , then we can use (21.9)–(21.11)
to write Z Z Z Z
F · dr = P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz.
γ γ γ γ
In other words, the line integral of a vector field decomposes into a sum of the line integrals of its
componentsR with respect to the corresponding variables. The right side of the last identity is often
written as γ P dx + Q dy + R dz. With this notation, we can rewrite that identity in the form
Z Z
(Pe1 + Qe2 + Re3 ) · dr = P dx + Q dy + R dz. (21.12)
γ γ

Example 21.6. Evaluate Z

y dx + x dy + xyz dz,
γ
where γ is the space curve
x = et + e−t , y = et − e−t , z = t2 (0 ≤ t ≤ 1).
Solution. By (21.9)–(21.11),
Z Z 1

0
0
y dx + x dy + xyz dz = e − e−t et + e−t + et + e−t et − e−t + e2t − e−2t t2 (2t) dt
 t



γ 0
Z 1
2
2
= et − e−t + et + e−t
+ 2t3 e2t − e−2t dt


0
Z 1 Z 1
= e −2+e +e +2+e
2t −2t 2t −2t
dt + t3 d e2t + e−2t



0 0
Z 1 Z 1
1
=2 e2t + e−2t dt + t3 e2t + e−2t 0 − e2t + e−2t (3t2 ) dt




0 0
Z 1
1 3
= e2t − e−2t 0 + e2 + e−2 − t2 d e2t − e−2t


2 0
3 1 2t
Z
3  2 2t −2t 1
= 2e − t e − e
2
+ −2t



0
e − e (2t) dt
2 2 0
e2 + 3e−2 3 1
Z
= + t d e2t + e−2t


2 2 0
e + 3e
2 −2
3 1 2t
Z
3  2t −2t 1
= + t e +e + −2t



0
− e e dt
2 2 2 0
3 1 5e2 + 15e−2
= 2e2 + 3e−2 − e2t − e−2t 0 = .
4 4

167
 Example 21.7. Evaluate Z
y2 − xy dx,


γ
where γ is the curve y = ln x from (1, 0) towards (e, 1).
Solution. We can parametrize γ as
x = t, y = ln t (1 ≤ t ≤ e).
Hence, (21.9) gives Z Z e
y − xy dx =
2
ln2 t − t ln t dt.



γ 1
Using integration by parts, we find that
Z e Z e Z e Z e
 2 e
ln t dt = t ln t 1 −
2
t d ln t = e −
2
t 2t ln t dt = e − 2
−1



ln t dt
1 1 1 1
Z e
e
= e − 2 t ln t 1 + 2 t t−1 dt = −e + 2(e − 1) = e − 2;


1
Z e Z e Z e Z e
e
t ln t dt = ln t d 2 t = 2 t ln t 1 −
1 2 1 2
t t dt = 2 e − 2
1 2 −1 1 2
t dt = 41 e2 + 1 .


1


2
1 1 1 1

Thus, the given line integral equals e − 2 − 4 e + 1 ≈ −1.379.

1 2



Exercises
Evaluate the given line integral.
Z
21.1. x3 y ds; γ is given by r(t) = te1 + t2 e2 , 0 ≤ t ≤ 1
Zγ

21.2. x2 y ds; γ is the upper half of the circle x2 + y2 = 4

Zγ

21.3. (xy + ln x) dy; γ is the arc of the curve y = x3/2 from (1, 1) to (0, 0)
Zγ
21.4. xy dx − x2 dy; γ is the arc of the circle x2 + y2 = 4 from (2, 0) to (0, 2)
Zγ
xy + z2 ds; γ is given by r(t) = cos 2te1 + sin 2te2 + 3te3 , 0 ≤ t ≤ π.


21.5.
Zγ
21.6. z2 dx + y dz; γ is given by r(t) = (t − sin t)e1 + (1 − cos t)e2 + sin(t/2)e3 , 0 ≤ t ≤ 2π
Zγ

21.7. yz dx + 2xz dy − xy dz; γ consists of the line segments from (1, 0, 0) to (1, 0, 2) to (2, 2, 2)
Zγ
21.8. y2 dx + x2 dy + xyz dz; γ consists of the upper half of the circle x2 + y2 = 1 in the xy-plane taken clockwise
γ
and the line segment from (1, 0, 0) to (0, 1, 1)
Z
Evaluate the line integral F · dr for the given vector field F and the given curve γ.
γ
21.9. F(x, y, z) = x2 z e2 − (x + y)e3 ; γ is the ellipse r(t) = 2e1 + 3 sin te2 + cos te3 ,√0 ≤ t ≤ 2π
21.10. F(x, y, z) = sin x e1 + cos y e2 − 2xyze3 ; γ is given by r(t) = t3 e1 + 2t2 e2 + te3 , 0 ≤ t ≤ 1
21.11. F(x, y) = x sin ye1 + ye2 ; γ is the arc of the parabola x = y2 from (1, −1) to (4, 2)

168
 LECTURE 22

The Fundamental Theorem for Line Integrals

22.1. The main theorem

So far, we have been reducing line integrals to definite integrals. In the case of conservative
vector fields, there is another, faster way to evaluate their line integrals. That alternative approach
is based on the following theorem, known as the fundamental theorem for line integrals.
Theorem 22.1. Let γ be a piecewise smooth curve given by a vector function r(t), a ≤ t ≤ b,
and let f (x, y, z) be a differentiable function, whose gradient vector field ∇ f is continuous on γ.
Then Z
∇ f · dr = f (r(b)) − f (r(a)). (22.1)
γ

Proof. We have
Z Z
∇ f · dr = f x dx + fy dy + fz dz by (21.12)
γ γ
Z b
= f x (r(t))x0 (t) + fy (r(t))y0 (t) + fz (r(t))z0 (t) dt


by (21.9)–(21.11)
a
Z b
d
=

f (r(t)) dt by the chain rule
a dt
= f (r(b)) − f (r(a)) by the FTC. 
Suppose that F(x, y, z) is a conservative vector field with potential function f (x, y, z), that is,
F = ∇ f . Theorem 22.1 then gives
Z Z
F · dr = ∇ f · dr = f (r(b)) − f (r(a)).
γ γ

This demonstrates the importance of conservative vector fields. If F is conservative and its potential
f is known, the evaluation of the line integral of F is rather easy: all we need to do is calculate the
net change of f along the curve. The next example illustrates this.
Example 22.1. Evaluate Z
F · dr,
γ
where F(x, y, z) = 2ye1 + 2xe2 + e3 and γ is the curve given by r(t) = hsin 2t, cos 2t, ln(t + 1)i,
0 ≤ t ≤ π.
Solution. We observe that F = ∇g, where g(x, y, z) = 2xy + z. Hence, by (22.1),
Z Z
F · dr = ∇g · dr = g(r(π)) − g(r(0)) = g(0, 1, ln(π + 1)) − g(0, 1, 0) = ln(π + 1).
γ γ
169
Observe that the answer will stay the same, if we replace γ by any other piecewise smooth curve
connecting the points (0, 1, 0) and (0, 1, ln(π + 1))! For example, the line integral of F along the
curve displayed on Figure 22.1 (I have no idea what its parametrization may be!) will also equal
ln(π + 1). 

γ
(0, 1, ln(π + 1))

(0, 1, 0)

+ 2xe2 + e3 ) · dr = ln(π + 1)
R
Figure 22.1. A curve with γ (2ye1

22.2. Path independence

R saw that if F(x, y, z) is a conservative vector field, Theorem 22.1 implies that the line
We just
integral γ F·dr depends only on the endpoints of γ. Thus, if γ1 and γ2 are two curves in the domain
of F that share the same endpoints, then
Z Z
F · dr = F · dr,
γ1 γ2
R
that is, γ F · dr is path independent. The next theorem says that the converse of this observation is
also true: if the line integrals of F are path independent, then F is conservative.
Theorem R22.2. Let F be a vector field that is continuous in a open, connected region D. If the
line integral γ F · dr is path independent in D, then F is conservative in D.
Note that unlike Theorems 20.2 and 20.4, this theorem requires less of the set D: only that
it be open and connected (as opposed to simply connected). On the other hand, the test that it
provides is much more elusive: we must determine that all (infinitely many) line integrals of F
are path independent. This hardly seems to be a practical test. And yet, if you look up the proofs
of Theorems 20.2 and 20.4 in an advanced calculus text, you will discover that those results are
deduced from Theorem 22.2. Thus, every time we show that some vector field is conservative by
means of Theorems 20.2 or 20.4, we implicitly make use of Theorem 22.2 too.
We now return to the vector field
−y x
F(x, y) = e1 + 2 e2 ,
+y 2x2 x + y2
which we studied in Example 20.6(c). Recall that this field is defined and differentiable everywhere
but at the origin. Furthermore, we have
∂ ∂ y2 − x2
   
−y x
= = ,
∂y x2 + y2 ∂x x2 + y2 (x2 + y2 )2
170
also for all (x, y) , (0, 0). Back in Lecture #20, we showed that this vector field is conservative
on the plane cut along a half-line passing through the origin, but were unable to decide whether F
is conservative in its entire domain. The next example shows that, in fact, this vector field is not
conservative in its full domain.
−y x
Example 22.2. The vector field F(x, y) = 2 e1 + 2 e2 is not conservative.
x +y 2 x + y2
Solution. Let γ be the unit circle
x = cos t, y = sin t (0 ≤ t ≤ 2π).
By (21.9) and (21.10),
Z Z
−y x
F · dr = dx + dy
γ x +y x2 + y2
2 2
γ
Z 2π 
(− sin t)(cos t)0 (cos t)(sin t)0

= + dt
0 cos2 t + sin2 t cos2 t + sin2 t
(− sin t)2 + (cos t)2
Z 2π
= dt = 2π.
0 cos2 t + sin2 t
On the other hand, if F was conservative, with potential function, say f (x, y), Theorem 22.1 would
give
Z
F · dr = f (cos 2π, sin 2π) − f (cos 0, sin 0) = f (1, 0) − f (1, 0) = 0.
γ

Since this value differs from the value of the integral we got via direct calculation, F must be
non-conservative. 
Example 22.3. Evaluate
Z
−y x
dx + 2 dy,
γ x2 +y 2 x + y2
where γ is the circle (x − 2)2 + (y + 1)2 = 25, oriented counterclockwise.
First solution (not recommended). We just showed that the vector field
−y x
F(x, y) = e1 + 2 e2
x2 +y 2 x + y2
is conservative on the plane with a half-line removed and non-conservative on the plane with just
the origin removed. Thus, F is conservative on curves that don’t loop around the origin and non-
conservative on curves that do. In particular, F is not conservative on γ, since the origin lies inside
it. This means that we can’t use Theorem 22.1 to evaluate the given integral, so we go back to the
methods from the last lecture.
A point (a, b) is on γ exactly when (a − 2, b + 1) is on the circle x2 + y2 = 25. Thus, we can
derive the parametrization of γ from that of the latter circle: the parametric equations of γ are
x = 2 + 5 cos t, y = −1 + 5 sin t (0 ≤ t ≤ 2π).
171
Denote the given line integral by I. By (21.9) and (21.10),
(1 − 5 sin t)(2 + 5 cos t)0 (2 + 5 cos t)(−1 + 5 sin t)0
Z 2π  
I= + dt
0 (2 + 5 cos t)2 + (−1 + 5 sin t)2 (2 + 5 cos t)2 + (−1 + 5 sin t)2
(1 − 5 sin t)(−5 sin t) + (2 + 5 cos t)(5 cos t)
Z 2π
= dt
0 4 + 20 cos t + 25 cos2 t + 1 − 10 sin t + 25 sin2 t
25 − 5 sin t + 10 cos t 5 − sin t + 2 cos t
Z 2π Z 2π
= dt = dt.
0 30 − 10 sin t + 20 cos t 0 6 − 2 sin t + 4 cos t
Most likely, you have not seen integrals like the last in single-variable calculus. It can be
reduced to an integral of a rational function by the substitution u = tan(t/2), but we must be
extremely careful, because this substitution is not “well-behaved”. We must even be careful, if we
use mathematical software or an integral table. For example, Mathematica reports that
5 − sin t + 2 cos t
Z  
t 2
dt = + arctan + C,
6 − 2 sin t + 4 cos t 2 1 − tan(t/2)
which would suggest that
2π
5 − sin t + 2 cos t
Z 2π  
t 2
dt = + arctan = π,
0 6 − 2 sin t + 4 cos t 2 1 − tan(t/2) 0
... if it wasn’t for the fact that Mathematica also proudly proclaims that
5 − sin t + 2 cos t
Z 2π
dt = 2π.
0 6 − 2 sin t + 4 cos t
In fact, both answers that Mathematica returns are correct and our calculation between them is
flawed, but this comes to show that we may want to avoid this approach... 
Second solution (recommended). Recall that in the solution of the last example, we showed
that Z
−y x
dx + 2 dy = 2π,
γ0 x + y x + y2
2 2

where γ0 is the unit circle taken in the counterclockwise direction. It turns out that, although we
cannot appeal to Theorem 22.1 to evaluate the given line integral directly, we can use it to show
that Z Z
−y x −y x
dx + 2 dy = dx + 2 dy.
γ x +y x +y γ0 x + y x + y2
2 2 2 2 2

Let γ+ be the part√of γ that lies above the x-axis and let τ+ be the√curve that has the same
endpoints as γ+ : (2 ± 24, 0), but consists of the line segment from (2 + √24, 0) to (1, 0), the upper
half γ+0 of the unit circle γ0 , and the line segment from (−1, 0) to (2 − 24, 0) (see Figure 22.2).
Also, let D+ be the plane cut along the negative y-axis (dashed line). Since γ+ and τ+ lie in D+ and
F is conservative in D+ , Theorem 22.2 gives
Z Z
−y x −y x
dx + 2 dy = dx + 2 dy.
γ+ x + y x +y τ+ x + y x + y2
2 2 2 2 2

√ γ that lies below the x-axis and −let τ be the curve that consists of
Next, let γ− be the part of −

the line segment from (2 − 24, 0) to (−1, 0), the lower half γ0 of the unit circle γ0 , and the line
172
 y y
τ− x
γ+

x
+
τ
γ−

Figure 22.2. The curves γ+ and Figure 22.3. The curves γ− and
τ+ in D+ τ− in D−

√
segment from (1, 0) to (2 + 24, 0) (see Figure 22.3). Also, let D− be the plane cut along the
positive y-axis (dashed line). Since γ− and τ− lie in D− and F is conservative in D− , Theorem 22.2
gives
Z Z
−y x −y x
dx + 2 dy = dx + 2 dy.
γ− x + y x +y τ− x + y x + y2
2 2 2 2 2

Combining this identity with the one before it, we get

Z Z Z Z Z
F · dr = F · dr + F · dr = F · dr + F · dr.
γ γ+ γ− τ+ τ−

We break the integrals over τ+ and τ− into six integrals over γ+0 , γ−0 , and the two horizontal edges.
Each horizontal edge appears in the sum twice: once oriented from left to right and once oriented
from right to left. Thus, those contributions cancel out (recall (21.8)) and we get
Z Z Z Z Z
F · dr + F · dr = F · dr + F · dr = F · dr.
τ+ τ− γ+0 γ−0 γ0

It follows that, as promised,

Z Z
−y x −y x
dx + 2 dy = dx + 2 dy = 2π.
γ x +y x +y γ0 x + y x + y2
2 2 2 2 2

Exercises
Z
Find a potential for the given vector field and then use the fundamental theorem to evaluate F · dr.
γ
22.1. F(x, y) = ey e1 + xey e2 ; γ is a smooth curve connecting (1, 0) and (2, −1)
22.2. F(x, y) = (4x + sin y)e1 + (sin y + x cos y)e2 ; γ is given by r(t) = (t − sin t)e1 + (1
√ − cos t)e2 , 0 ≤ t ≤ 2π
22.3. F(x,
√ y) = ye 1 + (x + 3y)e2 ; γ follows the ellipse x 2
+ 2y 2
= 5 from (1, 2) to (− 5, 0) and then the line segment
from (− 5, 0) to (1, −2)
22.4. F(x, y, z) = (yz + x2 )e1 + (xz + 4y)e2 + (xy − 3z2 )e3 ; γ consists of the line segments from (1, 2, 3) to (0, 2, 1) to
(−1, 3, 0) to (3, 2, 1)
22.5. F(x, y, z) = y2 cos ze1 + 2xy cos ze2 − xy2 sin ze3 ; γ is given by r(t) = t2 e1 + t3 e2 − 2t2 e3 , 0 ≤ t ≤ 2
173
 y

γ
22.6. Evaluate Z
−y x
dx + 2 dy,
γ x2+y 2 x + y2
x
where γ is the curve from the figure to the right.

174
 LECTURE 23

Green’s Theorem

In this lecture we discuss the first of the three major theorems of vector calculus. It is known
as Green’s theorem and provides an alternative technique for evaluation of certain line integrals.

23.1. Some definitions

First, we need to fix some terminology. A curve γ, given by r(t), a ≤ t ≤ b, is called closed,
if r(a) = r(b). In other words, the curve is closed, if its initital point is also its terminal point
(see Figure 23.1). A curve γ is called simple, if r(t1 ) , r(t2 ) whenever t1 , t2 , except that we

r(a) = r(b)

Figure 23.1. A closed, positively oriented curve

allow r(a) = r(b). In other words, a simple curve does not self-intersect and does not touch itself,
but can be closed. For example, among the curves displayed on Figure 23.2, those in (a) and (e)
are simple and those in (b)–(d) are not: the curve in (b) self-intersects at point A; the curve in (c)
passes through its initial point before reaching its terminal point; and the curve in (d) touches itself
at point B.
It is intuitively clear (though quite difficult to prove) that a piecewise smooth, simple, closed
plane curve γ divides the plane into two sets, which we can call intuitively the interior and the
exterior of γ, the interior being the bounded set and the exterior the unbounded. This fact is known
as Jordan’s theorem, named after the French mathematician Camille Jordan for his work on plane
curves, including this result. However, the first full proof of Jordan’s theorem was given by an
American: Oswald Veblen.
Let γ be a piecewise smooth, simple, closed curve. We say that γ is positively oriented, if its
interior lies always to the left of the point r(t) as it traces the curve in the direction of increasing of
t. For “round” curves, such as a circle or the curve on Figure 23.2(e), the positive orientation of the
curve is the counterclockwise orientation. However, this is not the case for all curves: the positive
orientation goes clockwise for the part of the curve on Figure 23.1 that follows immediately the
initial point r(a).
175
 A

r(a)
(a) (b) (c)
B

(d) (e)

Figure 23.2. Simple and non-simple curves

23.2. Green’s theorem

Theorem 23.1 (Green). Let P(x, y) and Q(x, y) have continuous first-order partials on an open
set U. Let γ be a positively oriented, piecewise smooth, simple, closed plane curve such that γ and
all of its interior D are contained inside U. Then
∂Q ∂P
Z ZZ  
P dx + Q dy = − dA. (23.1)
γ D ∂x ∂y
In other words, Green’s theorem relates a line integral along a simple, closed plane curves to
a double integral over the interior of the curve. Since we know how to evaluate double integrals
this yields another possible integration technique for line integrals along closed curves. However,
there is a catch: we absolutely must check the conditions of Green’s theorem before we use it. For
example, we showed in Example 20.6(c) that the functions
−y x
P(x, y) = 2 , Q(x, y) = 2
x +y 2 x + y2
are differentiable and satisfy the condition ∂P/∂y = ∂Q/∂x everywhere except at the origin. Thus,
if γ is a simple, closed curve whose interior D does not contain the origin, Green’s theorem gives
∂Q ∂P
Z ZZ  
P dx + Q dy = − dA = 0.
γ D ∂x ∂y
Note that for such a curve the application of Green’s theorem is legitimate. On the other hand, if
the origin is in the interior of the curve γ, the conditions of Green’s theorem are violated (at one
point) and the theorem is not applicable. Were we to apply it anyways, the result would again be
that the line integral along γ is 0. For example, such an illegitimate application of Green’s theorem
yields
Z
−y x
dx + 2 dy = 0
γ0 x + y x + y2
2 2

176
when γ0 is the unit circle x2 + y2 = 1. However, we know from Example 22.2 that the last integral
equals 2π, not 0. We see that if the functions P(x, y) and Q(x, y) or some of their derivatives are
not defined at even a single point in the interior of the curve, Green’s theorem is not applicable and
should not be used. Now, let us take a look at some of its legitimate uses.
Example 23.1. Evaluate
Z
e4x sin x + 2y dx + x2 + arctan y dy,



γ
where γ is the rectangle with vertices (1, 2), (5, 2), (5, 4), (1, 4).
Solution. We have P(x, y) = e4x sin x + 2y and Q(x, y) = x2 + arctan y. The partials of these
functions are:
1
P x (x, y) = (4 sin x + cos x)e4x , Py (x, y) = 2, Q x (x, y) = 2x, Qy (x, y) = .
1 + y2
Since these are all continuous everywhere, the open set U in the statement of Green’s theorem can
be taken to be R2 . We can therefore apply Green’s theorem. We get
Z ZZ Z 5Z 4
e sin x + 2y dx + x + arctan y dy =
4x 2
(2x − 2) dA =



(2x − 2) dydx
C D 1 2
Z 5
5
= 2(2x − 2) dx = 2 x2 − 2x 1 = 32.

1
Here, D denotes the rectangle [1, 5] × [2, 4]. 
Example 23.2. Evaluate
Z
e4x sin x + 2y dx + x2 + arctan y dy,



γ
where γ is the boundary of the semiannular region shown on Figure 23.3 (the radii of the two
semicircles are 2 and 3).

y
γ
(0, 2)
x

(0, −3)

Figure 23.3. The region D

Solution. The integrand is the same as in the previous problem, but the curve is different. We
can appeal to Green’s theorem for the same reasons as in the previous example. We get
Z ZZ
e sin x + 2y dx + x + arctan y dy =
4x 2



(2x − 2) dA,
γ D

177
where D is the semiannular region shown on Figure 23.3. Note that this region is conveniently
described in polar coordinates:
D = (r, θ) : 2 ≤ r ≤ 3, −π/2 ≤ θ ≤ π/2 .


Thus, by converting the double integral to polar coordinates, we obtain

Z Z 3 Z π/2
e sin x + 2y dx + x + arctan y dy =
4x 2



(2r cos θ − 2)r dθdr
γ 2 −π/2
Z 3 π/2
=

(2r sin θ − 2θ)r −π/2
dr
2
Z 3
= 4r2 − 2πr dr = 25 13 − 5π.


2

In the above examples, we used Green’s theorem to evaluate line integrals by means of double
integrals. However, Green’s theorem is a relation between the line integral and the double integral
appearing in (23.1) that can be used in either direction. Sometimes, we want to use Green’s theorem
to evaluate a double integral by means of a line integral. For example, under the assumption of
Theorem 23.1, we have
Z ZZ ZZ
x dy = ∂ x (x) − ∂y (0) dA = 1 dA = area(D),


γ D D
Z ZZ ZZ
(−y) dx = ∂ x (0) − ∂y (−y) dA = 1 dA = area(D).


γ D D

In other words, we have the following new formulas for the area enclosed by a simple, closed
curve.
Theorem 23.2. Let D be a plane region, bounded by a piecewise smooth, simple, closed curve
γ. Then
Z Z Z
1
area(D) = x dy = (−y) dx = x dy − y dx. (23.2)
γ γ 2 γ
x2 y2
Example 23.3. Find the area enclosed by the ellipse 2 + 2 = 1.
a b
Solution. We will use formula (23.2) and the parametric representation of the ellipse:
x = a cos t, y = b sin t (0 ≤ θ ≤ 2π).
By (23.2),
1 2π 
Z Z
1
area(D) = x dy − y dx =

(a cos t)(b cos t) − (b sin t)(−a sin t) dt
2 C 2 0
Z 2π
1 2π
Z
1
= ab cos t + sin t dt =
2 2
ab dt = πab.


2 0 2 0

178
 23.3. Vector forms of Green’s theorem*
We can use the curl and divergence of a vector field to give alternative formulations of Green’s
theorem. Suppose that γ is a positively oriented, piecewise smooth, simple, closed curve with
parametrization
r(t) = x(t)e1 + y(t)e2 + 0e3 (a ≤ t ≤ b),
that is, γ is a plane curve viewed as a space curve. Furhter, consider the vector field
F(x, y, z) = P(x, y)e1 + Q(x, y)e2 + 0e3 .
By (21.12), Z Z
F · dr = P(x, y) dx + Q(x, y) dy.
γ γ
Also, by the definition of curl,
∂0 ∂Q ∂P ∂0 ∂Q ∂P ∂Q ∂P
       
curl F = − e1 + − e2 + − e3 = − e3 ,
∂y ∂z ∂z ∂x ∂x ∂y ∂x ∂y
because P(x, y) and Q(x, y) do not depend on z. Thus, we can rewrite (23.1) as
Z ZZ
F · dr = (curl F) · e3 dA. (23.3)
γ D

Another formulation of Green’s theorem involves the vector field

G(x, y, z) = Q(x, y)e1 − P(x, y)e2 + 0e3 ,
which is orthogonal to the vector field F that we used in (23.3):
F · G = P(x, y)Q(x, y) + Q(x, y)(−P(x, y)) + 0 = 0.
Note that
∂Q ∂P
div F = − .
∂x ∂y
Also, by (21.7), Z Z
F·r= (F · T) ds,
γ γ
where T is the unit tangent vector to γ. It is a nice exercise to try to use the properties of the dot
product to convince yourself that
y0 (t)e1 − x0 (t)e2
n(t) = p
x0 (t)2 + y0 (t)2
is a unit vector that is normal to the curve γ and points towards the exterior of the curve; it is called
the outward normal vector of γ. Since F · T = G · n, we get
Z Z
F · r = (G · n) ds,
γ γ

and we can restate (23.1) as Z ZZ

(G · n) ds = div G dA. (23.4)
γ D

179
 Exercises
23.1. (a) Let γ be the square with vertices (1, 1), (2, 1), (2, 2), and (1, 2). Parametrize γ and use the formulas from
Lecture #21 to evaluate
Z
x2 y dx + xy2 dy.
γ

(b) Evaluate the integral from part (a) by means of Green’s theorem.

Use Green’s theorem to evaluate the given line integral (all the curves are assumed positively oriented).
Z
23.2. y3 dx − x3 dy; γ is the circle x2 + y2 = 4
Zγ

23.3. (xy + tan x) dx + x2 + ecos y dy; γ is the boundary of the region bounded by y = x and y = x2


Zγ
xy3 dx + 4x2 y2 dy; γ consists of the semicircle x = 9 − y2 and the line segment from (0, 3) to (0, −3)
p
23.4.
Zγ
xy2 + sin x2 dx + (x2 + y2 )3/2 + arctan y dy; γ is the boundary of region bounded by the semicircle



23.5.
√ γ √
y= Z 4 − x2 and the lines y = x and y = − 3x
23.6. e x sin y dx + e x cos y dy; γ is the ellipse (x + 1)2 + 5(y − 2)2 = 7
γ
Z
Use Green’s theorem to evaluate the line integral F · dr for the given field F(x, y) and the given curve γ.
γ

23.7. F(x, y) = y3/2 + x e1 + x2 + y e2 ; γ consists of the graph of y = sin2 x from (0, 0) to (π, 0) and the line segment

from (π, 0) to (0, 0)

23.8. F(x, y) = x3 + y e1 + 2xy
+ y4 e2 ; γ is the cardioid r = 2 + cos θ, positively oriented

23.9. F(x, y) = xe1 + x3 + 3xy2 e2 ; γ consists of the graph of y = 4 − x2 from (−2, 0) to (2, 0) and the line segment
from (2, 0) to (−2, 0)

23.10. When the circle x2 + (y − 1)2 = 1 is rolled along the x-axis in the positive direction, the point whose original
position is the origin traces a curve called a cycloid (see Figure 23.4). This curve can be parametrized by the vector
function

r(t) = (t − sin t)e1 + (1 − cos t)e2 (t ≥ 0).

Use Theorem 23.2 to find the area bounded by the x-axis and the first arc of the cycloid (that is, the part of the cycloid
with 0 ≤ t ≤ 2π).

Figure 23.4. Cycloid

23.11. When a circle of radius 1 is rolled along the outside of the circle x2 + y2 = 4 a fixed point on the smaller circle
traces a curve called an epicycloid (see Figure 23.5). This curve can be parametrized by the vector function

r(t) = (3 cos t − cos 3t)e1 + (3 sin t − sin 3t)e2 (0 ≤ t ≤ 2π).

Use Theorem 23.2 to find the area enclosed by this epicycloid.

180
 Figure 23.5. Epicycloid Figure 23.6. Hypocycloid Figure 23.7. Astroid
23.12. When a circle of radius 1 is rolled along the inside of the circle x2 + y2 = 9 a fixed point on the smaller circle
traces a curve called a hypocycloid (see Figure 23.6). This curve can be parametrized by the vector function
r(t) = (2 cos t + cos 2t)e1 + (2 sin t − sin 2t)e2 (0 ≤ t ≤ 2π).
Use Theorem 23.2 to find the area enclosed by this hypocycloid.
23.13. When a circle of radius 1 is rolled along the inside of the circle x2 + y2 = 16 a fixed point on the smaller circle
traces a different kind of hypocycloid called also an astroid (see Figure 23.7). This curve can be parametrized by the
vector function
r(t) = (3 cos t + cos 3t)e1 + (3 sin t − sin 3t)e2 (0 ≤ t ≤ 2π).
Use Theorem 23.2 to find the area enclosed by this astroid.

181
 LECTURE 24

Surface Integrals

24.1. Surface integrals of scalar functions

We start with a smooth parametric surface Σ, given by the vector function
r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 ((u, v) ∈ D).
Recall that the smoothness of Σ means that ru (u, v) and rv (u, v) are continuous and (ru ×rv )(u, v) , 0
in D (see §10.3). Further, suppose that f (x, y, z) is a continuous function on a set S in R3 that
contains the surface Σ. In particular, f (x, y, z) is defined at each point (x, y, z) on Σ.
For simplicity, we shall assume that D is a rectangle in the uv-plane: D = [a, b] × [c, d]. We
can treat the general case similarly, but there are more technical details to pay attention to and little
extra insight to gain. For each n ≥ 1, let a = u0 < u1 < · · · < un = b be the points that divide the
interval [a, b] into n subintervals of equal lengths; let c = v0 < v1 < · · · < vn = d be the points that
divide the interval [c, d] into n subintervals of equal lengths. The points
(un , v0 ) (un , v1 ) · · · (un , v j ) · · · (un , vn ),
.. .. .. ..
. . . .
(ui , v0 ) (ui , v1 ) · · · (ui , v j ) · · · (ui , vn ),
.. .. .. ..
. . . .
(u1 , v0 ) (u1 , v1 ) · · · (u1 , v j ) · · · (u1 , vn ),
(u0 , v0 ) (u0 , v1 ) · · · (u0 , v j ) · · · (u0 , vn ),
define a partition of the rectangle D into n2 congruent subrectangles. The vector function r maps
each of those small rectangles to a small “patch” on Σ and each grid point (ui , v j ) to a point Pi j on
the surface. The patches form a partition of Σ into n2 pieces, which we label as Σi j , 1 ≤ i, j ≤ n,
so that Σi j is the image of the rectangle that has (ui , v j ) as its upper right corner point. Further, we
pick n2 sample points (u∗i j , v∗i j ) so that
ui−1 ≤ u∗i j ≤ ui , v j−1 ≤ v∗i j ≤ v j .
These yield n2 sample points P∗i j on the surface, with coordinates xi∗j = x(u∗i j , v∗i j ), y∗i j = y(u∗i j , v∗i j ),
z∗i j = z(u∗i j , v∗i j ). Then the surface integral of f over Σ is defined by
ZZ n X
X n
f (x, y, z) dS = lim f (xi∗j , y∗i j , z∗i j )∆S i j , ∆S i j = area(Σi j ), (24.1)
Σ n→∞
i=1 j=1

provided that the limit exists.

Recall that, by Theorem 18.1,
ZZ
∆S i j = kru × rv k dA,
Di j

183
where Di j is the rectangle in the uv-plane whose image is Σi j . Using this formula, the above
definition, the properties of continuous functions, and the definition of the double integral, it can
be shown that for continuous functions f the limit (24.1) always exists and the surface integral can
be expressed as a double integral.
Theorem 24.1. Let Σ be a smooth parametric surface, given by the vector function
r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 ((u, v) ∈ D).
If f (x, y, z) is continuous on Σ, then
ZZ ZZ
f (x, y, z) dS = f (r(u, v))kru × rv k dA. (24.2)
Σ D

Similarly to line integrals, we can extend the above definition of the surface integral to piece-
wise smooth surfaces. If Σ is a piecewise smooth surface that is the union of the smooth surfaces
Σ1 , Σ2 , . . . , Σn , Σ = Σ1 ∪ Σ2 ∪ · · · ∪ Σn , then we define
ZZ ZZ ZZ ZZ
f (x, y, z) dS = f (x, y, z) dS + f (x, y, z) dS + · · · + f (x, y, z) dS .
Σ Σ1 Σ2 Σn

Remark. Like all other types of integrals we have encountered so far, the surface integral can
be interpreted in terms of averages. In this case, we are dealing with the average value of the
function on the surface:
ZZ
1
average f (x, y, z) = f (x, y, z) dS . (24.3)
(x,y,z)∈Σ area(Σ) Σ
Example 24.1. Evaluate ZZ
xy dS ,
Σ

where Σ is the graph of z = x3/2 + y3/2 over the rectangle 0 ≤ x, y ≤ 1.

2


3

Solution. The surface can be parametrized by the vector function

r(u, v) = ue1 + ve2 + 23 u3/2 + v3/2 e3


(0 ≤ u, v ≤ 1).
Hence,
ru (u, v) = e1 + 0e2 + u1/2 e3 , rv (u, v) = 0e1 + e2 + v1/2 e3 ,
and
 
e1 e2 e3       
 0 u1/2  1 u1/2  1 0 
ru × rv =  1 0 u  =  1/2  e2 + 
 e3 = −u1/2 e1 − v1/2 e2 + e3 .
 1/2 
1/2  e1 − 
     
1/2
0 1 v  1 v 0 v 0 1 

Applying (24.2), we get

ZZ ZZ √ Z 1 Z 1 √ 
xy dS = uv u + v + 1 dA = v u u + v + 1 du dv. (24.4)
S 0 0
0≤u,v≤1

184
Using the integration by parts, we find that
Z 1 √ Z 1
u u + v + 1 du = 3 2
u d(u + v + 1)3/2
0 0
1
Z 1
= u(u + v + 1)w3/2 0 (u + v + 1)3/2 du
2 2
3
− 3
0
5/2 1
= 23 (v + 2)3/2 − + +

2 2

3 5
(u v 1) 0
= 23 (v + 2)3/2 − 4
15
(v + 2) + 15 (v +
5/2 4
1)5/2 .
Inserting the result back into (24.4) and using another integration by parts, we obtain
ZZ Z 1
xy dS = v 32 (v + 2)3/2 − 15 (v + 2)5/2 + 154 (v + 1)5/2 dv
 4

S 0
Z 1
= (v + 2)5/2 − 105 (v + 2)7/2 + 105 (v + 1)7/2
4 8 8


vd 15
0

1
= v (v + 2)5/2 − + 2)7/2 + + 1)7/2
 4 8 8
15 105
(v 105
(v 0
Z 1
+ 2)5/2 − + 2)7/2 + + 1)7/2 dv
4 8 8


− 15
(v 105
(v 105
(v
0
1
= − 27 37/2 + 27 27/2 − 154 72 (v + 2)7/2 − 634 (v + 2)9/2 +
5/2
+ 1)9/2
4

 4
15
3 63
(v 0
√ √
√ √
= 9 3 + 16 2 − 105
4 8 1


105 9
81 3 − 8 2 − 2 ≈ 0.372. 

24.2. Surface integrals of vector fields

Definition. Let F be a continuous vector field defined on an oriented surface Σ with unit normal
vector n. Then the surface integral of F over Σ is
ZZ ZZ
F · dS = (F · n) dS . (24.5)
Σ Σ

Remark. Since F · n is the size of the projection of F in the direction of n, we can interpret the
surface integral of F as the average size of its projection onto the direction of orientation of Σ.
Suppose that the surface is given by the vector function
r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3 ((u, v) ∈ D),
and that it is oriented using the unit normal vector
ru × rv
n(u, v) = .
kru × rv k
Combining (24.5) and (24.2), we get
ZZ ZZ ZZ  
ru × rv
F · dS = (F · n) dS = F· dS
Σ Σ Σ kru × rv k
ZZ   ZZ
ru × rv
= kru × rv k dA =


F· F · (ru × rv ) dA. (24.6)
D kru × rv k D

185
In particular, when Σ is the graph z = f (x, y), (x, y) ∈ D, with the upward orientation (recall
Example 10.8), we obtain
ZZ ZZ
(Pe1 + Qe2 + Re3 ) · dS = − P(x, y) f x (x, y) − Q(x, y) fy (x, y) + R(x, y) dA.


(24.7)
Σ D

Example 24.2. Evaluate ZZ

(xe1 − 2ye2 + ze3 ) · dS,
Σ

where Σ is the elliptic paraboloid z = 4 − 2x2 − 2y2 , z ≥ 0, with the upward orientation.
Solution. Condition z ≥ 0 restricts x, y to the circle
4 − 2x2 − 2y2 ≥ 0 ⇐⇒ x2 + y2 ≤ 2.
Since we are using the upward orientation, we can appeal to (24.7). We have
∂ x 4 − 2x2 − 2y2 = −4x, ∂y 4 − 2x2 − 2y2 = −4y,

so (24.7) gives
ZZ ZZ
(xe1 − 2ye2 + ze3 ) · dS = 4x2 − 8y2 + z(x, y) dA


Σ
x2 +y2 ≤2
ZZ
= 4x2 − 8y2 + 4 − 2x2 − 2y2 dA

x2 +y2 ≤2
ZZ
= 4 + 2x2 − 10y2 dA.

x2 +y2 ≤2

Converting the double integral to polar coordinates, we obtain

ZZ Z 2π Z √2
(xe1 − 2ye2 + ze3 ) · dS = 4 + 2r2 cos2 θ − 10r2 sin2 θ r drdθ


Σ 0 0
 √2
Z 2π
= 2r + 2 r cos2 θ − 52 r4 sin2 θ 0 dθ
 2 1 4
0
Z 2π
= 4 + 2 cos2 θ − 10 sin2 θ dθ


0
Z 2π
= 4 + (1 + cos 2θ) − 5(1 − cos 2θ) dθ = 0.


0

Example 24.3. Evaluate ZZ
(xze1 + xe2 + ye3 ) · dS,
Σ

where Σ is the hemisphere x2 + y2 + z2 = 25, y ≥ 0, oriented in the direction of the positive y-axis.
186
 Solution. It is best to represent the hemisphere in spherical coordinates:
r(θ, φ) = 5 cos θ sin φe1 + 5 sin θ sin φe2 + 5 cos φe3 (0 ≤ θ, φ ≤ π).
Then
rθ = −5 sin θ sin φe1 + 5 cos θ sin φe2 ,
rφ = 5 cos θ cos φe1 + 5 sin θ cos φe2 − 5 sin φe3 ;
rθ × rφ = −25 sin2 θ sin φ cos φ(e1 × e2 ) + 25 sin θ sin2 φ(e1 × e3 )
+ 25 cos2 θ sin φ cos φ(e2 × e1 ) − 25 cos θ sin2 φ(e2 × e3 )
= −25 cos θ sin2 φe1 − 25 sin θ sin2 φe2 − 25 sin φ cos φe3
= −25 sin φ(cos θ sin φe1 + sin θ sin φe2 + cos φe3 ).

Note that the vector rθ × rφ has a negative second component for all θ, φ with 0 ≤ θ, φ ≤ π.
Therefore, we can orient Σ in the direction of the positive y-axis by taking
−(rθ × rφ ) rφ × rθ
n= = .
krθ × rφ k krφ × rθ k
With this choice, we have
F(r(θ, φ)) = 25 cos θ sin φ cos φe1 + 5 cos θ sin φe2 + 5 sin θ sin φe3
= 5 sin φ(5 cos θ cos φe1 + cos θe2 + sin θe3 ),
F(r(θ, φ)) · rφ × rθ = 125 sin2 φ 5 cos2 θ sin φ cos φ + sin θ cos θ sin φ + sin θ cos φ .

Thus, (24.6) yields

ZZ ZZ
F · dS = 125 sin2 φ(5 cos2 θ sin φ cos φ + sin θ cos θ sin φ + sin θ cos φ) dA
Σ
0≤θ,φ≤π
Z π Z π
= 125 5 cos2 θ sin3 φ cos φ + sin θ cos θ sin3 φ + sin θ sin2 φ cos φ dθdφ.


0 0
We now use that
π π π
π
Z Z Z
cos θ dθ = ,
2
sin θ cos θ dθ = 0, sin θ dθ = 2.
0 2 0 0
Substituting these values in the iterated integral above, we obtain
ZZ Z π
F · dS = 125 (2.5π sin3 φ cos φ + 2 sin2 φ cos φ) dφ
S 0
Z 0
= 125 (2.5πu3 + 2u2 ) du = 0.
0


Example 24.4. The surface Σ with vector equation

r(u, v) = u cos ve1 + u sin ve2 + ve3 ,
187
 Figure 24.1. Two views of the helicoid

where 0 ≤ u ≤ 1, 0 ≤ v ≤ π, is called helicoid (see Figure 24.1). Evaluate the surface integral of
p p
F(x, y, z) = x 1 + x2 + y2 e1 + y 1 + x2 + y2 e2 + (x + y)ze3
over Σ with upward orientation.
Solution. We have
ru = cos ve1 + sin ve2 , rv = −u sin ve1 + u cos ve2 + e3 ;
ru × rv = u cos2 v(e1 × e2 ) + cos v(e1 × e3 ) − u sin2 v(e2 × e1 ) + sin v(e2 × e3 )
= sin ve1 − cos ve2 + ue3 .
Note that the vector ru × rv has a positive third component for all u with 0 ≤ u ≤ 1. Therefore, we
can orient Σ in the upward direction by the unit normal vector
ru × rv
n= .
kru × rv k
With this choice, we have
√ √
F(r(u, v)) = u cos v 1 + u2 e1 + u sin v 1 + u2 e2 + uv(cos v + sin v)e3
F(r(u, v)) · (ru × rv ) = u2 v(cos v + sin v).
Thus, (24.6) yields
ZZ Z π Z 1
F · dS = u2 v(cos v + sin v) dudv
Σ 0
Z 0
1 π 1 π
Z
= v(cos v + sin v) dv = v d(sin v − cos v)
3 0 3 0
π 1 π π 2
Z
1
= v(sin v − cos v) 0 − (sin v − cos v) dv = − . 
3 3 0 3 3
188
 Exercises
ZZ
24.1. Evaluate the surface integral x2 yz dS , where Σ is the surface
Σ
z = 1 + 2x + 3y, 0 ≤ x ≤ 3, 0 ≤ y ≤ 2,
that is, the part of the plane z = 1 + 2x + 3y that lies above the rectangle [0, 3] × [0, 2].
ZZ
24.2. Evaluate the surface integral x2 z2 dS , where Σ is the surface
Σ
p
z = x2 + y2 , 1 ≤ x2 + y2 ≤ 4,
that is, the part of the cone z2 = x2 + y2 that lies between the planes z = 1 and z = 2.
ZZ
x2 z + y2 z dS , where Σ is the hemisphere x2 + y2 + z2 = 4, z ≥ 0, that is, the


24.3. Evaluate the surface integral
Σ
parametric surface given by
r(u, v) = 2 cos u sin ve1 + 2 sin u sin ve2 + 2 cos ve3 , 0 ≤ u ≤ 2π, 0 ≤ v ≤ π/2.
ZZ
x2 y + z2 dS , where Σ is the part of the cylinder x2 + y2 = 9 that lies between


24.4. Evaluate the surface integral
Σ
the planes z = 0 and z = 2, that is, the parametric surface given by
r(u, v) = 3 cos ue1 + 3 sin ue2 + ve3 , 0 ≤ u ≤ 2π, 0 ≤ v ≤ 2.
ZZ
24.5. Evaluate the surface integral F · dS, where F(x, y, z) = xye1 + yze2 + zxe3 and Σ is the part of the paraboloid
Σ
z = 4 − x2 − y2 that lies above the square 0 ≤ x, y ≤ 1 and is oriented upward.
ZZ
24.6. Evaluate the surface integral F · dS, where F(x, y, z) = xzey e1 − xzey e2 + ze3 and Σ is the part of the plane
Σ
x + y + z = 1 in the first octant with the upward orientation.
ZZ
24.7. Evaluate the surface integral F · dS, where F(x, y, z) = xze1 + 2yze3 and Σ is the paraboloid y = x2 + z2 ,
Σ
0 ≤ y ≤ 1, oriented in the direction of the positive y-axis.
ZZ
24.8. Evaluate the surface integral F · dS, where F(x, y, z) = xe1 + ye2 + xyze3 and Σ is the helicoid given by
Σ
r(u, v) = u cos ve1 + u sin ve2 + ve3 , 0 ≤ u ≤ 1, 0 ≤ v ≤ π,
and oriented by n = ru × rv .

189
 LECTURE 25

Stokes’ Theorem

In the last two lectures, we discuss the other two major integral theorems of vector calculus:
Stokes’ theorem and Gauss’ divergence theorem. Stokes’ theorem is a generalization of Green’s
theorem to closed curves on surfaces in space. Before we state it, we introduce the notion of
orientation of curves on an oriented surface.
Definition. Suppose that Σ is an oriented surface with unit normal vector n. We say that a
closed curve γ on Σ is positively oriented, if the vector n × T points towards the interior of the
curve; here T is the unit tangent vector to γ.
Theorem 25.1 (Stokes). Let F be a vector field that has continuous partial derivatives on an
open set U in R3 . Let Σ be an oriented, piecewise smooth surface that is contained in U and is
bounded by a simple, closed, piecewise smooth, positively oriented curve γ. Then
Z ZZ
F · dr = curl F · dS. (25.1)
γ Σ

We said above that Stokes’ theorem is a generalization of Green’s. It would be nice to corrob-
orate that. Consider the parametric surface Σ given by
r(x, y) = xe1 + ye2 + 0e3 ((x, y) ∈ D),
that is, Σ is the region D in the xy-plane turned into a “three-dimensional surface”. If we give this
surface an upward orientation, then the unit normal vector n is e3 , and Stokes’ theorem gives
Z ZZ ZZ
F · dr = curl F · dS = (curl F) · e3 dS .
γ Σ Σ

Converting the latter surface integral into a double integral, we find

r x (x, y) = e1 , ry (x, y) = e2 , r x × ry = e3 ,
and hence, Z ZZ ZZ
F · dr = (curl F) · e3 dS = (curl F) · e3 dA.
γ Σ D
This is exactly one of the vector forms of Green’s theorem (recall (23.3)).
We now move to describe the main applications of Stokes’ theorem. They are similar to the
applications of Green’s theorem. First, we can apply Stokes’s theorem to convert a line integral
over a closed space curve γ to a surface integral over a surface Σ that has γ as its boundary.
Example 25.1. Evaluate
Z
 x3 +x
+ z2 e1 + arctan y + x2 e2 + ln(z2 + 1) + y2 e3 · dr,




e
γ
191
 Figure 25.1. The spherical patch x2 + y2 + z2 = 9, x, y, z ≥ 0, and its boundary

where γ is the positively oriented boundary of the part of the sphere x2 + y2 + z2 = 9 that lies in the
first octant, with downward orientation (see Figure 25.1.

Solution. All the partials of F will be continuous everywhere, so we can apply Stokes’ theo-
rem. We get
Z ZZ
F · dr = curl F · dS,
γ Σ

where Σ is the given surface and

F(x, y, z) = e x +x + z2 e1 + arctan y + x2 e2 + ln(z2 + 1) + y2 e3 .

3

As we do usually when dealing with spheres, we parametrize Σ using spherical coordinates:

r(θ, φ) = 3 cos θ sin φe1 + 3 sin θ sin φe2 + 3 cos φe3 (0 ≤ θ, φ ≤ π/2).

Then, similarly to Example 24.3,

rθ = −3 sin θ sin φe1 + 3 cos θ sin φe2 ,

rφ = 3 cos θ cos φe1 + 3 sin θ cos φe2 − 3 sin φe3 ;
rθ × rφ = −9 sin φ(cos θ sin φe1 + sin θ sin φe2 + cos φe3 ).

Since −9 sin φ cos φ ≤ 0 for all φ, 0 ≤ φ ≤ π/2, the downward orientation of Σ is given by
rθ × rφ
n= .
krθ × rφ k

Furthermore, we have

curl F = (2y − 0)e1 − (0 − 2z)e2 + (2x − 0)e3 = 2ye1 + 2ze2 + 2xe1 ;

curl F(r(θ, φ)) = 6 sin θ sin φe1 + 6 cos φe2 + 6 cos θ sin φe3 ,
curl F(r(θ, φ)) · (rθ × rφ ) = −54 sin2 φ sin θ cos θ sin φ + (sin θ + cos θ) cos φ .

192
We conclude that
Z ZZ
F · dr =


curl F(r(θ, φ)) · (rθ × rφ ) dA
γ
0≤θ,φ≤π/2
Z π/2 Z π/2
= −54 sin2 φ sin θ cos θ sin φ + (sin θ + cos θ) cos φ dθdφ


0 0
Z π/2 π/2
= −54 sin2 φ sin2 θ sin φ + (− cos θ + sin θ) cos φ 0 dφ
1
2
0
Z π/2
= −54 sin2 φ sin φ + 2 cos φ dφ
1


2
0
Z π/2 Z π/2
= −27 2
sin2 φ cos φ dφ


1 − cos φ sin φ dφ − 108
0 0
Z 1 Z 1
= −27 1 − u2 du − 108 u2 du = −54.


0 0


Example 25.2. Evaluate Z
(yze1 + 2xze2 + 3xye3 ) · dr,
γ

where γ is the intersection curve of the ellipsoid x2 + 2y2 + z2 = 4 and the plane x + z = 2, oriented
positively relative to the outward orientation of the ellipsoid.

Solution. The intersection curve γ of the ellipsoid and the plane is displayed on Figure 25.2.
It can be seen from the figure that the positive orientation of γ relative to the outward orientation
of the ellipsoid is in the counterclockwise direction. Note that this is also the positive orientation
of γ relative to the upward orientation of the plane x + z = 2. The equations of γ are
x2 + 2y2 + z2 = 4, x+z=2 ⇐⇒ x2 + 2y2 + (2 − x)2 = 4, z = 2 − x.

Figure 25.2. The intersection curve of x2 + 2y2 + z2 = 4 and x + z = 2

193
Some basic algebra shows that
x2 + 2y2 + (2 − x)2 = 4 ⇐⇒ x2 + 2y2 + 4 − 4x + x2 = 4 ⇐⇒ 2x2 − 4x + 2y2 = 0
⇐⇒ x2 − 2x + y2 = 0 ⇐⇒ (x − 1)2 + y2 = 1,
so γ is described by the equations
(x − 1)2 + y2 = 1, z = 2 − x.
That is, γ contains the points (x, y, z) on the plane z = 2 − x for which the point (x, y) lies on the
circle (x − 1)2 + y2 = 1 in the xy-plane. It is possible to parametrize γ and to evaluate the given line
integral directly (see the exercises), but we want to use Stokes’ theorem.
Let Σ be the portion of the plane z = 2 − x that lies within γ, with the upward orientation. Then,
by Stokes’ theorem,
Z ZZ
(yze1 + 2xze2 + 3xye3 ) · dr = curl(yze1 + 2xze2 + 3xye3 ) · dS.
γ Σ

We have
 
 e1 e2 e3  
  ∂y ∂z  ∂ x ∂z  ∂ x ∂y 
    
curl(yze1 + 2xze2 + 3xye3 ) = ∂ x ∂y ∂z  =  e + 

e −  e
yz 2xz 3xy 2xz 3xy 1 yz 3xy 2 yz 2xz 3

= (3x − 2x)e1 − (3y − y)e2 + (2z − z)e3 = xe1 − 2ye2 + ze3 .

Further, Σ is the graph of z = 2 − x for (x, y) in the disk D with equation (x − 1)2 + y2 ≤ 1. Thus,
(24.7) gives
ZZ ZZ
curl(yze1 + 2xze2 + 3xye3 ) · dS = (xe1 − 2ye2 + ze3 ) · dS
Σ Σ
ZZ
= (−x∂ x (2 − x) − (−2y)∂y (2 − x) + z(x, y)) dA
D
ZZ ZZ
= (x + 0 + (2 − x)) dA = 2 dA.
D D

The last double integral represents twice the area of D, which is a disk of radius 1. Consequently,
Z ZZ
(yze1 + 2xze2 + 3xye3 ) · dr = 2 dA = 2area(D) = 2π. 
γ D

In the last two examples, we used Stokes’ theorem to reduce a line integral of a vector field to a
surface integral of its curl. In the next couple of examples, we describe the other main application
of Stokes’ theorem: to simplify the evaluation of surface integrals. To illustrate the first flavor of
such applications, we take a look at a surface integral that we evaluated in Example 24.2 by writing
it as a double integral. We now give an alternative solution, which uses Stokes’ theorem.
Example 25.3. Evaluate ZZ
(xe1 − 2ye2 + ze3 ) · dS,
Σ
where Σ is the part of the paraboloid z = 4−2x2 −2y2 that lies above the xy-plane, oriented upward.

194
 Figure 25.3. The paraboloid z = 4 − 2x2 − 2y2 , z ≥ 0 and its boundary

Solution. The main idea is to replace the given surface integral (which is a two-dimensional
object) by a line integral along the surface boundary (which is a one-dimensional object). Recall
from the previous example that
xe1 − 2ye2 + ze3 = curl F, F = yze1 + 2xze2 + 3xye3 .
Hence, when we apply Stokes’ theorem to F on Σ, we get
ZZ ZZ Z
(xe1 − 2ye2 + ze3 ) · dS = curl F · dS = F · dr,
Σ Σ γ

where γ is the boundary of the given surface, positively oriented. The curve γ is the intersection
curve of the paraboloid and the xy-plane, that is, the circle x2 + y2 = 2, z = 0. We can parametrize
γ using
√ √
r(t) = 2 cos te1 + 2 sin te2 + 0e3 (0 ≤ t ≤ 2π).
Note that with this parametrization, γ is positively oriented relative to the upward orientation of Σ
(see Figure 25.3). Thus,
Z Z 2π
F · dr = F(r(t)) · r0 (t) dt
γ 0
Z 2π √ √
= 0e1 + 0e2 + 6 cos t sin te3 · − 2 sin te1 + 2 cos te2 dt = 0.




0

In the above solution, the vector field F resulted from a “happy coincidence”. Had we not
just worked out Example 25.2, we would have started the above solution of Example 25.3 by
“observing” that
xe1 − 2ye2 + ze3 = curl yze1 + 2xze2 + 3xye3 .

Such an “observation”
RR falls just short of reliance upon “divine intervention”. In general, given a
surface integral Σ F·dS, in order to argue similarly to above, we need to find a vector field G such
that curl G = F is some given vector field. This is a difficult problem, which limits the applicability
of the method. Still, the above example is not a complete waste of time: sometimes, one has to
calculate the surface integral of the curl of a given field, and in those situations one can argue as
above.
195
 Figure 25.4. The surface Σ from Example 25.4

Example 25.4. Evaluate

ZZ
curl yze1 + 2xze2 + 3xye3 · dS,


Σ

where Σ is the piecewise smooth surface displayed on Figure 25.41, oriented upward (the boundary
curve of Σ is the square formed by the lines x = ±2 and y = ±2 in the xy-plane).

Solution. The main idea of this solution is to replace the given surface integral by another
surface integral that is easier to evaluate. By Stokes’ theorem,
ZZ Z
curl yze1 + 2xze2 + 3xye3 · dS = yze1 + 2xze2 + 3xye3 · dr,



Σ γ

where γ is the boundary square of Σ, oriented positively relative to the upward orientation of the
surface (that is, γ is oriented counterclockwise in the xy-plane). We now observe that γ is also the
positively oriented boundary of the surface Σ0 with parametrization
r(x, y) = xe1 + ye2 + 0e3 (−2 ≤ x, y ≤ 2)
and with the upward orientation. Hence, another application of Stokes’ theorem gives
Z ZZ
yze1 + 2xze2 + 3xye3 · dr = curl yze1 + 2xze2 + 3xye3 · dS.



γ Σ0

We have
curl yze1 + 2xze2 + 3xye3 = xe1 − 2ye2 + ze3 ,

r x × ry = e1 × e2 = e3 ,
Hence, by (24.7),
ZZ ZZ
curl yze1 + 2xze2 + 3xye3 · dS = xe1 − 2ye2 + ze3 · dS



Σ
Z ΣZ0
= xe1 − 2ye2 + 0e3 · e3 dA = 0.



−2≤x,y≤2
√
1
In fact, the surface is the graph of the function z = min 4 − x2 ,
p

4 − y2 , but that is unimportant for the
purposes of the example.
196
 Exercises
Z
Use Stokes’ theorem to evaluate F · dr for the given field F(x, y, z) and the given curve γ.
γ
25.1. F(x, y, z) = 2ze1 + xe2 +3ye3 ; γ is the ellipse that is the intersection of the plane z = x and the cylinder x2 +y2 = 4,
oriented clockwise
25.2. F(x, y, z) = (z − y)e1 + (z − x)e2 + (x + y)e3 ; γ is the triangle with vertices (0, 1, 0), (0, 2, 2) and (1, 2, 2), oriented
positively as viewed from above (i.e., relative to the upward orientation of its plane)
25.3. F(x, y, z) = x2 e1 −2xye2 +yz2 e3 ; γ is the boundary of the surface z = x2 y, 0 ≤ x, y ≤ 1, oriented counterclockwise
25.4. F(x, y, z) = x2 ye1 + 31 x3 e2 + xye3 ; γ is the curve of intersection of the hyperbolic paraboloid z = y2 − x2 and the
cylinder x2 + y2 = 1, oriented counterclockwise as viewed from above
ZZ
Use Stokes’ theorem to evaluate curl F · dS for the given field F(x, y, z) and the given surface Σ.
Σ
25.5. F(x, y, z) = 3ye1 − xze2 − yz2 e3 ; Σ is the part of the elliptic paraboloid 2z = x2 + y2 below z = 2, oriented upward
25.6. F(x, y, z) = yze1 + xze2 + xye3 ; Σ is the part of the paraboloid z = 9 − x2 − y2 above z = 8, oriented downward
25.7. F(x, y, z) = x2 e1 − 2xye2 + yz2 e3 ; Σ is the the surface z = x2 y, 0 ≤ x, y ≤ 1, oriented upward
25.8. F(x, y, z) = ye1 + xze2 + 2xe3 ; Σ is the hemisphere x2 + y2 + z2 = 1, y ≥ 0, oriented in the direction of the positive
y-axis

197
 LECTURE 26

Gauss’ Divergence Theorem

We now turn to another generalization of Green’s theorem, known as Gauss’ divergence theo-
rem. It relates the surface integral over the boundary of a simply-connected solid to a triple integral
over the entire solid, similarly to how Green’s theorem relates a line integral over the boundary of
a simply-connected plane region to the double integral over the entire region.
Like Green’s theorem, the divergence theorem requires that we pay attention to terminology.
Recall the definition of a simply-connected region from §20.3. By a simply-connected solid, we
mean a simply-connected region in space that is closed (contains all its boundary points). For
such a solid, its boundary consists of the points on its “outer” (visible) surface. The boundary
of a simply-connected solid is the rightful three-dimensional analog of the simple, closed curve
appearing in Green’s theorem. We also define the positive orientation of the boundary of a solid
(assuming that it is orientable) to be the outward orientation.
Theorem 26.1 (Gauss’ divergence theorem). Let S be a bounded, simply-connected solid in
R , whose boundary is the piecewise smooth, positively oriented surface Σ. Let F be a vector field
3

whose first-order partials are continuous on an open set U containing S . Then

ZZ ZZZ
F · dS = div F dV. (26.1)
Σ S

Example 26.1. Evaluate

ZZ
(x3 + y sin z)e1 + (y3 + z sin x)e2 + 3ze3 · dS,


Σ
where Σ is the positively oriented boundary of the solid S bounded by the xy-plane and by the
hemispheres p p
z = 1 − x2 − y2 , z = 4 − x2 − y2 .

Figure 26.1. The solid 1 ≤ x2 + y2 + z2 ≤ 4, z ≥ 0

199
 Solution. We write F = (x3 + y sin z)e1 + (y3 + z sin x)e2 + 3ze3 . Then, by Gauss’ theorem,
ZZ ZZZ ZZZ
F · dS = div F dV = 3x2 + 3y2 + 3 dV.


Σ S S
The solid S is displayed on Figure 26.1. It is described by the inequalities
1 ≤ x2 + y2 + z2 ≤ 4, z ≥ 0.
These inequalities are translated in spherical coordinates as
1 ≤ ρ ≤ 2, 0 ≤ θ ≤ 2π, 0 ≤ φ ≤ π/2.
Hence, converting the triple integral over S to spherical coordinates, we obtain
ZZ Z 2 Z π/2 Z 2π
F · dS = (3ρ2 sin2 φ + 3)ρ2 sin φ dθdφdρ
Σ 1 0 0
Z 2 Z π/2
= 6π (ρ4 sin2 φ + ρ2 ) sin φ dφdρ
1 0
Z 2  Z π/2 
= 6π ρ (1 − cos φ) + ρ sin φ dφ dρ
4 2 2


1 0
Z 2Z 1 
= 6π ρ (1 − u ) + ρ du dρ
4 2 2


1 0
Z 2
= 6π 2 4
+ ρ2 dρ = π.

194
3
ρ 5
1

Example 26.2. Evaluate ZZ
x2 e1 + y2 e2 + z2 e3 · dS,


Σ
where Σ is the positively oriented surface of the solid S shown on Figure 26.2 (a cube of side length
2 with a corner cube of side length 1 removed).

(0, 2, 2)
(2, 0, 2)

(2, 2, 0)

Figure 26.2. The solid E

Solution. It is possible to evaluate the surface integral using double integrals. However, were
we to do that, we would have to parametrize each face separately, and there are nine faces. Fur-
thermore, we would have to split the three double integrals corresponding to the L-shaped faces in
200
two, so we would have to evaluate twelve double integrals altogether. On the other hand, S is the
difference of two cubes, so we can calculate a triple integral over S by writing it as the difference
of two triple integrals over those cubes. This suggests that we use the divergence theorem.
By Gauss’ theorem,
ZZ ZZZ
x e1 + y e2 + z e3 · dS =
2 2 2
(2x + 2y + 2z) dV.


Σ S
3 3
Since E is the difference of the cubes [0, 2] and [1, 2] , it is easier to calculate the triple integral
over E as the difference of the triple integrals over those two cubes:
ZZZ ZZZ ZZZ
(2x + 2y + 2z) dV = (2x + 2y + 2z) dV − (2x + 2y + 2z) dV
S
[0,2]3 [1,2]3
Z 2Z 2Z 2 Z 2Z 2Z 2
= (2x + 2y + 2z) dzdydx − (2x + 2y + 2z) dzdydx
0 0 0 1 1 1
Z 2Z 2 Z 2Z 2
2 2
2
= 2xz + 2yz + z 0 dydx − 2xz + 2yz + z2 1 dydx
  
0 0 1 1
Z 2Z 2 Z 2Z 2
= (4x + 4y + 4) dydx − (2x + 2y + 3) dydx
0 0 1 1
Z 2 Z 2
2 2
= 4xy + 2y + 4y 0 dx −
2
2xy + y2 + 3y 1 dx
 
0 1
Z 2 Z 2
2  2
= (8x + 16) dx − (2x + 6) dx = 4x2 + 16x 0 − x2 + 6x 1 = 39.

0 1

With a little bit of extra effort, we can also use the divergence theorem to evaluate surface
integrals of scalar functions.
Example 26.3. Evaluate ZZ
x2 z + y2 z dS ,


Σ
where Σ is the sphere x2 + y2 + z2 = 4.
Solution. Note that we have to deal with the surface integral of a scalar function and not of a
vector field. Thus, Gauss’ theorem is not directly applicable. However, we know that the sphere Σ
can be parametrized using spherical coordinates by the vector function
r(θ, φ) = 2 cos θ sin φe1 + 2 sin θ sin φe2 + 2 cos φe3 ,
where 0 ≤ θ ≤ 2π and 0 ≤ φ ≤ π. We also know (we went through this calculation several times
already) that the (outward) unit normal vector is then
n = rφ × rθ = cos θ sin φe1 + sin θ sin φe2 + cos φe3 .
Note that in Cartesian coordinates n = 12 xe1 + ye2 + ze3 . We can use this observation to represent


the integrand in the given integral in the form F · n, with F some nice vector field; we can then
appeal to Gauss’ theorem. For example, if we choose F = 2(x2 + y2 )e3 , we obtain
F · n = 0 21 x + 0 12 y + 2(x2 + y2 ) 21 z = x2 z + y2 z,

201
whence ZZ ZZ ZZ ZZZ
x z + y z dS =
2 2
(F · n) dS = F · dS = div F dV = 0.


Σ Σ Σ B
Here, B is the ball x2 + y2 + z2 ≤ 4.
Note that the above choice of F is not unique. We could have chosen also
F = 2xze1 + 2yze2 , F = xze1 + yze2 + (x2 + y2 )e3 ,
or any other vector field F that satisfies
F · n = x2 z + y2 z.
However, in most cases, those choices would yield triple integrals that equal 0 for less obvious
reasons than the one we encountered. For example, the choice F = 2xze1 + 2yze2 yields
ZZ ZZZ ZZZ
x z + y z dS =
2 2
div F dV =


4z dV.
Σ B B
The last integral is still 0, but this requires some thought (try to convince yourself by interpreting
it in terms of the average value of f (x, y, z) = 4z). 
We can also use the divergence theorem to calculate volumes by means of surface integrals
similarly to how we earlier used Green’s theorem to calculate areas by means of line integrals. For
example, if S is a simply-connected solid and Σ is its boundary, then
ZZ ZZZ ZZZ
xe1 · dS = div(xe1 ) dV = 1 dV = vol.(S ).
Σ S S
Similarly, we obtain formulas for the volume of S that use the vector fields ye2 and ze3 . Here are
some of the surface integrals representing vol.(S ):
ZZ ZZ ZZ ZZ
1
vol.(E) = xe1 · dS = ye2 · dS = ze3 · dS = (xe1 + ye2 + ze3 ) · dS. (26.2)
Σ Σ Σ 3 Σ
Example 26.4. Find the volume of the spiral pipe shown on Figure 26.3. The outer surface of
the pipe is parametrized by the vector function
r(u, v) = (2 + sin v) cos 2ue1 + (2 + sin v) sin 2ue2 + (u + cos v)e3 (0 ≤ u, v ≤ 2π),
and its two openings are the disks (x − 2) + z ≤ 1 and (x − 2) + (z − 2π) ≤ 1 in the xz-plane.
2 2 2 2

Figure 26.3. The surface r(u, v) = (2 + sin v) cos 2ue1 + (2 + sin v) sin 2ue2 + (u + cos v)e3

202
 Solution. Let Σ be the positively oriented, piecewise smooth surface consisting of the paramet-
ric surface Σ0 that is the outer surface of the pipe and the two disks representing its two openings.
We denote the disks Σ1 and Σ2 . Then Σ is the boundary of the simply-connected solid S whose
volume we want to find. By (26.2),
ZZ  ZZ ZZ ZZ 
vol.(S ) = ze3 · dS = + + ze3 · dS.
Σ Σ0 Σ1 Σ2

The unit normal vectors to Σ1 are e2 and −e2 , and of those two, it is −e2 that points out of the solid
S . Thus, by (26.2), ZZ ZZ
ze3 · dS = (ze3 ) · (−e2 ) dS = 0.


Σ1 Σ1
Similarly, ZZ ZZ
ze3 · dS = (ze3 ) · (e2 ) dS = 0.


Σ2 Σ2
Hence, ZZ
vol.(S ) = ze3 · dS.
Σ0
We evaluate the last surface integral using the orientation
ru × rv
n= .
kru × rv k
Thus, by (24.6),
ZZ ZZ
ze3 · dS =


(z(u, v)e3 ) · (ru × rv ) dA.
Σ0
0≤u,v≤2π

Note that the dot product depends only on the third component of ru × rv . We have
ru (u, v) = −2(2 + sin v) sin 2ue1 + 2(2 + sin v) cos 2ue2 + e3 ,
rv (u, v) = cos v cos 2ue1 + cos v sin 2ue2 − sin ve3 .
Thus, the cross product ru × rv is the sum of six nonzero terms:
ru × rv = −2 cos v(2 + sin v) sin2 2u(e1 × e2 ) + 2 cos v(2 + sin v) cos2 2u(e2 × e1 ) + other terms
= −2 cos v(2 + sin v) sin2 2u + cos2 2u e3 + other terms

= −2 cos v(2 + sin v)e3 + other terms.

The omitted terms are cross products involving e1 × e3 , e2 × e3 , etc., which do not contribute to the
coefficient of e3 in ru × rv . Thus,
(z(u, v)e3 ) · (ru × rv ) = ((u + cos v)e3 ) · (ru × rv ) = −2 cos v(u + cos v)(2 + sin v).
We conclude that
ZZ ZZ
ze3 · dS = − 2 cos v(u + cos v)(2 + sin v) dA


Σ0
0≤u,v≤2π
Z 2π Z 2π 
= − 2u cos v(2 + sin v) − 2 cos v sin v − 4 cos v dv du.
2 2


0 0

203
We have
Z 2π Z 2π Z 2π
(2 + sin v) cos v dv = 0, cos v sin v dv = 0,
2
cos2 v dv = π,
0 0 0
so
ZZ Z 2π
ze3 · dS = − 2u(0) − 2(0) − 4π du = −8π2 .


Σ0 0
We obtained a negative volume, because we did not check whether the unit vector n gives the
inward or the outward orientation of the surface. It turns out that it gives the inward (negative)
orientation; hence, the negative volume. Consequently, the volume of the pipe is 8π2 . 
Exercises
ZZ
Use Gauss’ theorem to evaluate F · dS for the given field F(x, y, z) and the given surface Σ oriented outward.
Σ
26.1. F(x, y, z) = y3 e1 + z3 e2 + x3 e3 ; Σ is the boundary of the solid S defined by x2 + y2 + z2 ≤ 3, x ≥ 0, y ≥ 0
26.2. F(x, y, z) = x3 e1 + y3 e2 + z3 e3 ; Σ as in Exercise 26.1
26.3. F(x, y, z) = x2 z3 e1 + x2 ye2 + y2 ze3 ; Σ is the boundary of the solid S enclosed by the sphere x2 + y2 + z2 = 4
inside the cylinder x2 + y2 = 1
26.4. F(x, y, z) = xy2 e1 + yz2 e2 + zx2 e3 ; Σ is the boundary of the solid shown on Figure 26.4, that is, the part of the
solid sphere x2 + y2 + z2 ≤ 4 outside
√ the first octant
26.5. F(x, y, z) = y3 ez e1 + y 4 − x2 e2 + ey cos xe3 ; Σ is the boundary of the solid S that lies below the surface
z = 2 − x3 − y3 , −1 ≤ x, y ≤ 1, and above the xy-plane

Figure 26.4. Seven eights of the sphere x2 + y2 + z2 ≤ 4

ZZ
x2 + y2 + z3 dS , where Σ is the sphere x2 + y2 + z2 = 9.


26.6. Use Gauss’ theorem to evaluate
Σ

Use the volume formulas (26.2) to calculate the volume of the given solid.
26.7. The solid enclosed by the ellipsoid x2 + 2y2 + 4z2 = 4, parametrized by
√
r(θ, φ) = 2 cos θ sin φe1 + 2 sin θ sin φe2 + cos φe3 (0 ≤ θ ≤ 2π, 0 ≤ φ ≤ π)
26.8. The solid bounded by the torus
r(u, v) = cos u(4 + 2 cos v)e1 + sin u(4 + 2 cos v)e2 + 2 sin ve3 (0 ≤ u, v ≤ 2π)

204
 APPENDIX A

Determinants

An n × n matrix is an array of real numbers with n rows and n columns. For example, A below
is a 2 × 2 matrix and B is a 3 × 3 matrix:
 
  1 −3 1
1 2
A= , B = 0 2 −1 .
3 4
1 1 −2
The determinant of a 2 × 2 matrix  
a11 a12
A=
a21 a22
is the following number (the product of its diagonal entries minus the product of its off-diagonal
entries):  
a11 a12 
det(A) = 
  = a11 a22 − a12 a21 .
a21 a22 
The determinant of a 3 × 3 matrix
 
a11 a12 a13
A = a21 a22 a23 
a31 a32 a33
is given by expansion along the first row, a process which requires the computation of three 2 × 2
determinants:
 
a11 a12 a13       
a22 a23  a21 a23  a21 a22 
det(A) = a21 a22 a23  = a11  a31 a33  + a13 a31 a32  .
 
 − a12    
a31 a32 a33  a32 a33 

Example A.1. Compute the determinant of the matrix B above.

Solution. We have
     
2 −1 0 −1 0 2 
det(B) = 1 
  − (−3) 
  + 1 
1 −2 1 −2 1 1
= (2 · (−2) − 1 · (−1)) + 3(0 · (−2) − 1 · (−1)) + (0 · 1 − 2 · 1) = −2. 

205
 APPENDIX B

Definite Integrals of Single-Variable Functions

B.1. Definition of the definite integral

Definition. Let f (x) be a bounded function on [a, b]. For each n = 1, 2, . . . , set ∆n = (b − a)/n
and define the numbers
x0 = a, x1 = a + ∆n , x2 = a + 2∆n , ..., xn = a + n∆n = b.
Also, choose sample points x1∗ , x2∗ , . . . , xn∗ , with
x0 ≤ x1∗ ≤ x1 , x1 ≤ x2∗ ≤ x2 , ..., xn−1 ≤ xn∗ ≤ xn .
The definite integral of f from a to b is
Z b n
X
f (x) dx = lim f (xi∗ )∆n , (B.1)
a n→∞
i=1
provided that the limit exists. The sums on the right side of (B.1) are called Riemann sums of f on
[a, b].
Remarks. 1. In the above definition, the limit is supposed to exist and be independent of the
choice of the sample points. This is not true for any old function, but it does hold for functions
that are continuous on [a, b] or have only a finite number of jump discontinuities inside [a, b]. For
such functions, we may evaluate the integral using any choice of the sample points (e.g., we can
take each xi∗ to be the midpoint of the respective subinterval).

y = f (x)

a b

Figure B.1. The definite integral as area

2. In single-variable calculus, the definite integral of a continuous function is usually motivated

by the area problem: If f (x) is a continuous function on [a, b], with f (x) ≥ 0, calculate the area of
the region (see Figure B.1)
R = (x, y) : a ≤ x ≤ b, 0 ≤ y ≤ f (x) .


207
The answer turns out to be the definite integral of f (x):
Z b
area(R) = f (x) dx.
a

3. A different interpretation of the definite integral arises, if we divide both sides of (B.1) by
b − a. Since ∆n = (b − a)/n, we have
n n n
1 X ∆n X 1X
f (xi∗ )∆n = f (xi∗ ) = f (xi∗ ).
b − a i=1 b − a i=1 n i=1
Hence, dividing both sides of (B.1) by b − a, we derive
Z b n
1 1X
f (x) dx = lim f (xi∗ ). (B.2)
b−a a n→∞ n i=1

Note that the expression 1n i f (xi∗ ) is the average of the sample values f (x1∗ ), f (x2∗ ), . . . , f (xn∗ ),
P
which are about equally spaced. Thus, the limit on the right side of (B.2) represents the average of
“infinitely many, equally spaced” sample values of f (x). Consequently, we may argue that the left
side of (B.2) is the average value of f (x):
Z b
1
average f (x) = f (x) dx. (B.3)
x∈[a,b] b−a a
This alternative interpretation of the definite integral is rarely emphasized in single-variable
calculus, but is much more appropriate for the purpose of these notes, because we want to draw
parallels between the properties of definite integrals and the properties of other types of integrals
(double, triple, line, surface) whose geometric meaning may be less obvious.

B.2. Properties of the definite integral

The following is a list of some general properties of definite integrals:
Z b Z b Z b
i) [ f (x) + g(x)] dx = f (x) dx + g(x) dx
a a a
Z b Z b
ii) c f (x) dx = c f (x) dx
a a
Z c Z b Z b
iii) f (x) dx + f (x) dx = f (x) dx
a c a
Z b Z b
iv) If f (x) ≤ g(x), then f (x) dx ≤ g(x) dx
a a
Z b
v) If m ≤ f (x) ≤ M, then m(b − a) ≤ f (x) dx ≤ M(b − a).
a
Z a Z b Z a
vi) f (x) dx = − f (x) dx, f (x) dx = 0.
b a a
Of course, the most important property of definite integrals is the Fundamental Theorem of
Calculus, also known as the Newton–Leibnitz theorem.
208
 Theorem B.1 (Fundamental theorem of calculus). Suppose that f (x) is a continuous function
on [a, b].
i) We have Z x
d
f (t) dt = f (x).
dx a
Rx
that is, a f (t) dt is an antiderivative of f (x).
ii) If F(x) is any antiderivative of f (x), then
Z b
f (x) dx = F(b) − F(a).
a

209
 APPENDIX C

Polar Coordinates

C.1. Definition
Consider a point P in the plane. Recall that its Cartesian coordinates (x, y) are the signed
distances from P to two perpendicular axes (the x- and y-axes). In this appendix, we recall another
set of coordinates: the polar coordinates of P. The polar coordinate system consists of a point O
called the pole (or the origin) and a ray starting at O called the polar axis. The polar coordinates of
a point P , O are a pair of real numbers (r, θ), where r is the distance from O to P and θ ∈ [0, 2π)
is the oriented angle between the polar axis and the line OP (see Figure C.1); the polar coordinates
of the pole O are (0, θ), where θ is any number in [0, 2π).

y
P(x, y)
2
y
+
P(r, θ) p x2
r=
r θ
θ
x
O O Q(x, 0)

Figure C.1. Polar coordinates Figure C.2. Polar and Cartesian coordinates

C.2. Relations between polar and Cartesian coordinates

We want to be able to convert polar coordinates to Cartesian and vice versa. In order to do
so, we need to set a Cartesian and a polar coordinate systems in the same copy of the plane. We
choose the two coordinate systems so that they share the same origin and the polar axis coincides
with the positive x-axis (see Figure C.2). Since the distance between the origin and P(x, y) is the
same no matter what coordinate system we use, we have
p
r = |OP| = x2 + y2 .
The relation between θ and (x, y) is more technical, but not much more complicated. First, consider
a point P(x, y) in the first quadrant. From 4OPQ on Figure C.2,
y
x = r cos θ, y = r sin θ, tan θ = .
x
211
The first two of these equations hold for all points P(x, y), even if they do not lie in the first
quadrant, that is, the Cartesian coordinates of a point P(r, θ) are obtained from the formulas
x = r cos θ, y = r sin θ. (C.1)
On the other hand, the relation tan θ = y/x extends only to the points with x , 0 and determines θ
uniquely only on the half-planes x > 0 or x < 0. Thus, when we write the formulas for converting
Cartesian into polar coordinates
p y
r = x2 + y2 , tan θ = , (C.2)
x
we must remember that the latter formula must be applied cautiously.
Example C.1. The point P whose polar coordinates are (2, 32 π) has Cartesian coordinates

√
x = 2 cos 23 π = −1, y = 2 sin 32 π = 3.


Example C.2. The Cartesian point P(2, 2) has polar coordinates
√ √ π
 
2
r = 22 + 22 = 2 2, θ = arctan = .
2 4
Here, we used that since P lies in the first quadrant, the solution of the second equation in (C.2) is
given by θ = arctan(y/x). The Cartesian point P(2, −2) has polar coordinates (r, θ), where
p √
r = 22 + (−2)2 = 2 2
and θ is such that
tan θ = −1, π, 2π ,
3


θ∈ 2
√
that is, P(2 2, 74 π). 
Example C.3. The Cartesian point P(−3, 4) has polar coordinates(r, θ), where
p
r = (−3)2 + 42 = 5
and θ is such that
4
tan θ = − , π, π .
1


θ∈ 2
3
Thus, P(5, arctan(− 43 ) + π). 
C.3. Polar curves
A polar curve is the set of points in the plane whose (polar) coordinates (r, θ) satisfy an equation
r = f (θ), where f is some function. Note that using equations (C.1), we can write every polar curve
in parametric form:
x = r cos θ = f (θ) cos θ, y = r sin θ = f (θ) sin θ.
Example C.4. The curve r = 3 is the set of points at distance 3 from the origin, that is, a circle
of radius 3 centered at O (whose Cartesian equation is x2 + y2 = 9, of course). Notice that we reach
the same conclusion if we use (C.2) to convert the given polar equation into a Cartesian equation:
p
r = 3 ⇐⇒ x2 + y2 = 3 ⇐⇒ x2 + y2 = 9.

212
 O

Figure C.3. The cardioid r = 1 + cos θ

Example C.5. The curve θ = 34 π is the ray starting at the origin and intersecting the polar axis
at an angle of 135◦ . 
Example C.6. The curve r = −2 cos θ is a circle. To see this, we pass to Cartesian coordinates:
r = −2 cos θ ⇐⇒ r2 = −2r cos θ ⇐⇒ x2 + y2 = −2x.
The last equation is the Cartesian equation of a circle of radius 1 centered at (−1, 0):
x2 + y2 = −2x ⇐⇒ x2 + 2x + 1 + y2 = 1 ⇐⇒ (x + 1)2 + y2 = 1.

Example C.7. The curve r = 1 + cos θ is known as a cardioid, because its graph is shaped like
a heart (see Figure C.3). Unlike the curves in the previous examples, the cardioid does not have a
simple Cartesian equation which we can use to sketch its graph. 

C.4. Area and arc length in polar coordinates

Theorem C.1. Let 0 ≤ a < b ≤ 2π and let f be a positive continuous function defined on [a, b].
Then the area A of the region in the plane bounded by the curves
θ = a, θ = b, r = f (θ),
is given by
Z b
1
A= [ f (θ)]2 dθ.
2 a

Example C.8. Find the area enclosed by the cardioid (see Example C.7).
Solution. By Theorem C.1, the area is given by the integral
1 2π 1 2π
Z Z
(1 + cos θ) dθ =
2
1 + 2 cos θ + cos2 θ dθ


2 0 2 0
1 2π 
Z
= 1 + 2 cos θ + 21 (1 + cos 2θ) dθ

2 0
1 2π 3π
= θ + 2 sin θ + 21 θ + 14 sin 2θ 0 = .
2 2

213
 Theorem C.2. Let 0 ≤ a < b ≤ 2π and let f be a positive continuous function defined on [a, b].
Then the length L of the curve r = f (θ) is given by
Z bp
L= f (θ)2 + f 0 (θ)2 dθ.
a

Example C.9. Find the length L of the cardioid.

Solution. By Theorem C.2,
Z 2π p Z 2π √
L= (1 + cos θ) + (− sin θ) dθ =
2 2 2 + 2 cos θ dθ
0 0
Z 2π p Z 2π Z π
= 4 cos2 (θ/2) dθ = 2| cos(θ/2)| dθ = 4 | cos u| du
0 0 0
Z π/2 Z π
=4 cos u du − 4 cos u du = 8.
0 π/2


214
 APPENDIX D

Introduction to Mathematica

D.1. Mathematica notebooks

A standard Mathematica file is called a notebook and has extension .nb. To create a new
notebook, select File→New→Notebook from the main menu or press Ctrl+N. That will result in
a new window filled with white space and named Untitled-1, Untitled-2, or similar.
Notebooks consist of one or more cells. Each cell represents a logical unit of the notebook and
is shown by a blue square bracket on the right side of the screen. There are several different types
of cells, but the three main types are: input cells, output cells, and text cells. In addition to these,
there are types of cells that provide special formatting: title cells, subtitle cells, section cells, etc.

D.1.1. Selecting and editing cells. To select an existing cell, click on its cell bracket. A
selected cell is indicated by a thick black bracket. To select more than one consecutive cell, click
on one cell bracket and then drag over the others or click on the desired cells while pressing the
Ctrl key down. To select cells that are not next to each other, click on the first one then, while
holding the Ctrl key down, point and click on the next one and so on. This is particularly useful
in printing a selection of cells. One can also copy and paste selected cells.
To edit the content of a cell, simply click on the desired location and start editing. One can also
change an entire cell by selecting it and choosing a different style type from the pull-down menu
in the toolbar.

D.1.2. Creating a new cell. If you position the mouse between two existing cells or after the
last cell and then click, a horizontal line (called the cell insertion point or line) will be shown
across the screen. This indicates where the next new cell will be inserted. Now, if you begin to
type, Mathematica will create a new cell containing the text that you type. If you type several
lines (using the Enter key to begin new lines), the cell bracket will grow to enclose everything
you type. Note that, by default, every new cell you type in is created as an input cell. That means
that Mathematica treats the symbols inside that cell as Mathematica code and may try to execute
it. That is not always desirable, so you must always convert the style of cells you consider text
from Input to Text. For example, a cell containing your name should have type Text, Subtitle or
Subsubtitle, not Input.

D.1.3. Executing cells. To execute a cell (to get an answer to an arithmetic operation, for ex-
ample), placing the cursor anywhere within the cell and then press Shift+Enter simultaneously
or press the Enter key on the numeric keypad. As soon as a cell is executed, it will be labeled
with In[1] for the first cell. The result will be shown in a new cell labeled Out[1]. The next cell
executed will be labeled In[2] regardless of its actual location relative to the first cell. Mathemat-
ica pays attention only to the order in which the cell is executed rather than to its relative position
with respect to other cells. Note that Mathematica will execute only input cells.
215
 Sometimes, one wants to execute more than one cell. That can be done by selecting the de-
sired cells and then pressing Shift+Enter. One can also execute the entire notebook by select-
ing Evaluation→Evaluate Notebook from the main menu. This will make Mathematica run
through the entire notebook and execute all input cells in the order of appearance. Note that if there
is text contained in an input cell, Mathematica will try to execute that text. The results of the latter
can be comical—for example, executing an input cell containing the text “Today is 10/10/2010.”
results in the following:
In[1]:= Today is 10/10/2010.
Out[1]:= 0.000497512 is Today

D.2. Basic syntax

D.2.1. Arithmetic operations and numerical values. Mathematica supports the basic arith-
metic operations. The following symbols are used:
+ indicates addition
- indicates subtraction
* indicates multiplication
/ indicates division
ˆ indicates exponentiation
Here are some examples of those rules in action:
In[1]:= 2+2*3
Out[1]:= 8

In[2]:= 2+2 3
Out[2]:= 8

In[3]:= 5ˆ3
Out[3]:= 125

In[4]:= 2/3-1/6
Out[4]:= 1/2
Note the second example in this list: the space between the numbers 2 and 3 is interpreted as
a multiplication. A space between two expressions is another way to indicate multiplication in
Mathematica. Sometimes, even the space is unnecessary: e.g., “2*x”, “2 x” and “2x” are all
interpreted by Mathematica as “2x”.
Notice that Mathematica returns the sum of two rational numbers as a rational number. That
is because by default, Mathematica uses exact arithmetic. This means that a rational number,
such as 12 , will normally be displayed as a fraction and not as the decimal 0.5. In fact, the latter
is considered an approximation having a limited precision. If we want the numerical value of a
fraction, we can use the built-in command N[]. In general, the command
N[..., n]
returns the numerical value of the expression in place of the ..., with n correct decimal digits. When
the value of n is missing, the default precision is used. For example, to get the numeric value of
7
94
, we can try one of the following:
216
 In[1]:= N[7/94]
Out[1]:= 0.0744681

In[2]:= N[7/94, 30]

Out[2]:= 0.0744680851063829787234042553191
An alternative way to get the numeric value of an expression is to add a left apostrophe ‘ after it.
For example,
In[1]:= 7/94‘
Out[1]:= 0.0744681
It is important to remember that Mathematica performs arithmetic computations in the usual
mathematical order: exponentiations are performed first; then multiplications and divisions, from
left to right; finally, additions and subtractions, also from left to right. When this natural order is
to be changed, we use parentheses. For example:
2ˆ3 - 1/3*4 + 13 represents 23 − 31 ∗ 4 + 13 = 8 − 34 + 13 = 19 23 ;
23 − 1
(2ˆ3 - 1)/(3*4 + 13) represents = 94
7
.
3 + 13
4

D.2.2. Elementary built-in functions and constants. Some of Mathematica’s built-in con-
stants are:
Pi π = 3.14159265358979 . . .
E e = 2.71828182845904
√ ...
I i = −1
Note that Mathematica is case-sensitive, so pi is not the number π! All built-in constants and
functions in Mathematica begin with an upper-case letter. Furthermore, function arguments in
Mathematica are enclosed in square brackets [], as opposed to the parentheses used in usual math-
ematical notation. Hence, the input Sqrt[16] returns the value 4, whereas sqrt[16], Sqrt(16)
and sqrt(16) do not work. Table D.1 below lists some of the elementary built-in functions that
you will use often.

Mathematica function Usual notation/Description

Sin[x],Cos[x],Tan[x],Sec[x] sin(x), cos(x), tan(x), sec(x)
ArcSin[x],ArcCos[x],ArcTan[x] arcsin(x),
√ arccos(x), arctan(x)
Sqrt[x] x
Abs[x] |x|
Log[x] ln(x)
Log10[x] log(x), log10 (x)
Log[b,x] logb (x)
N[x,n] Approximate value of x with n digits
Exp[x], Eˆx e x , exponential function base e
n! Factorial of n: n! = 1 · 2 · 3 · · · n
Table D.1. Mathematica syntax of some basic functions

217
 Mathematica gives exact results whenever possible. So, typing Pi as input will return π. If it
is a numerical approximation to π that we want, we should ask for it by using the N[]-command,
as shown in the following examples:
In[1]:= N[Pi]
Out[1]:= 3.14159

In[1]:= N[Pi,50]
Out[2]:= 3.1415926535897932384626433832795028841971693993751
Here are some more examples of how Mathematica tries to use exact calculations:
In[1]:= Sin[Pi/2]
Out[1]:= 1

In[2]:= Log10[1000]
Out[2]:= 3

In[3]:= Log10[200]
Out[3]:= Log[200]/Log[10]

In[4]:= N[Log10[1000]]
Out[4]:= 2.30103

In[5]:= 5!
Out[5]:= 120
As we mentioned earlier, Mathematica considers decimals, such as 0.5, approximate numbers.
Therefore, most numerical calculations using such decimals are limited in precision. For example,
compare the outputs of the following two calculations:
N[Eˆ(1/2),50]
1.6487212707001281468486507878141635716537761007101

N[Eˆ0.5,50]
1.64872
√
In the first calculation, we ask Mathematica to evaluate e1/2 = e, which is an exact number
having an infinite precision. The second calculation, on the other hand, is limited in precision by
the presence of the exponent 0.5, which Mathematica treats differently from 12 . For this reason, it
is preferred to represent decimals by fractions in numerical calculations.
D.2.3. Defining functions of one variable. To define a new function f (x) in Mathematica,
we typically use syntax of the form
f[x_] := ****
where the stars should be replaced by the appropriate formula. The symbols := (the delayed as-
signment operator) separate the definition in two. The left-hand side of the definition (f[x_]
above) specifies the name and the variable of the function; the right-hand side describes the func-
tion, usually by a formula involving the variable. For example, the command
f[x_] := xˆ2
218
defines the function f (x) = x2 . Note that we enclose the argument between square brackets and that
we use an underscore character after the argument’s name on the left-hand side of the definition.
Mathematica becomes aware of the definition of a user-defined function not when one has
typed in the definition of the function, but only after the cell containing that definition has been
executed. A common mistake is to type the definition of a function one wants to use and then
to move on to using that function without executing the definition first. If a function has not
been previously defined, Mathematica is likely to simply repeat any input that uses the undefined
function. For example,
In[1]:= someNewFunction[5]
Out[1]:= someNewFunction[5]
Note that when Mathematica does not recognize the name of a function or a variable, it displays
that name in blue: e.g., someNewFunction[]. Once the definition of the function has been loaded
(so that Mathematica knows what that name means), the highlighting changes from blue to black
and calls to the function result in its evaluation. For example,
In[1]:= someNewFunction[x_] := Sin[x]

In[2]:= someNewFunction[5]
Out[2]:= Sin[5]

In[3]:= N[someNewFunction[5],10]
Out[3]:= -0.9589242747
Remark. There are two conventions that you should follow when defining your own functions:
• To avoid conflicts with Mathematica’s built-in commands, always start the name of your
functions with a small letter. Since all built-in commands start with a capital letter, this
convention avoids the accidental overwriting of a built-in command.
• To avoid conflicts with the definitions of namesake functions, clear the name of the func-
tion before you define it. Mathematica remembers old definitions even if we delete the
cells containing them (or even close the notebook in which they were introduced), and this
can occasionally cause problems. To force Mathematica to forget such old definitions and
start afresh, we use of the built-in function Clear. It is best to precede each function’s
definition with a call to Clear to erase potential prior definitions. For example, to define
the function g(x) = x3 + 2x, we should use
Clear[g]
g[x_] := xˆ3+2x
D.2.4. Lists. Many of Mathematica’s more sophisticated commands have inputs and outputs
in the form of lists. A list is a collection of expressions (numbers, variables, functions, other lists,
etc.) separated by commas and enclosed in a pair of curly braces. For example, {x,0,2} is a list
of three elements: the variable x and the numbers 0 and 2. Here is another example:
L = {1,3,5,7,9,11}
This defines a list L whose elements are the odd integers 1, 3, . . . , 11. When a list has been given a
name, we can use that name followed by a number in double brackets ([[ ]]) to access a specific
element of the list. For example, the fourth element of the list L above can be obtained by L[[4]].
219
 Thus, we have now encountered four different types of brackets in Mathematica, each with its
own dedicated use. These types are:
() Round parentheses indicate the order of operations.
Do not use braces or brackets instead!
[] Square brackets enclose the argument of a function.
Do not use round parentheses instead!
{} Curly brackets (braces) enclose a list.
[[]] Double square brackets refer to a particular element of a list.

D.3. Plotting commands

D.3.1. Plotting functions of one variable.
D.3.1.1. Basic plots. To graph a function (or an expression) of one variable, we use the Plot
command. The command takes at least two arguments, a function (or an expression), and a range.
The range is a list of three elements: the variable in the expression, for example x, a minimum
value for x, and a maximum value for x. For example, the command
Plot[Sin[x], {x,0,2Pi}]
will plot the sine function over the range 0 ≤ x ≤ 2π. One can also graph a previously defined
function by using its name as the first argument of the Plot command as shown in the following
example.
Clear[f]
f[x_] := Cos[Eˆx]
Plot[f[x], {x,0,3}]
1.0

0.5

0.5 1.0 1.5 2.0 2.5 3.0

-0.5

-1.0

D.3.1.2. Graphics options. An option is a rule of the form OptionName->OptionValue,

where OptionName is the name of the option and OptionValue is the value that it is assigned. If
an option is not supplied explicitly, the default value is used. The following command will display
all the possible plotting options in Mathematica together with their default settings.
Options[Plot]
{AlignmentPoint->Center, AspectRatio->1/GoldenRatio,
Axes->True, AxesLabel->None, AxesOrigin->Automatic,
...
Ticks->Automatic, TicksStyle->{},
WorkingPrecision->MachinePrecision}
However, a much better way to learn about the various options of the Plot command is to read the
documentation entry on it in Mathematica’s Documentation Center (under the Help menu). Here,
220
we present only a few of the more commonly used options.

GridLines. We can use this option to simulate graphing paper:

Plot[Sin[x], {x,-4,4}, GridLines->Automatic]
1.0

0.5

-4 -2 2 4

-0.5

-1.0

The two basic values for this option are None (the default) and Automatic, which makes Math-
ematica generate the most reasonable rectangular grid for the given graph. There are also more
sophisticated alternatives that can be used to control the size and formatting of the grid—see the
documentation entry on GridLines.

PlotStyle. We can use this option to change the appearance of the plot. The value of PlotStyle
can be a single directive (e.g., Blue or Thick) or a list of such directives. Often used styles
are: Thin or Thick, Dashed or Dotted, and various colors. In particular, there are several ways
to specify the color of the plot. The most common is by the function RGBColor, which takes
three arguments: the amount of red, the amount of green, and the amount of blue. These argu-
ments must be numbers between 0 and 1, and they need not add up to 1. There are also sev-
eral predefined colors that can be assigned to PlotStyle; those include: Red, Green, Blue,
Yellow, Cyan, Magenta, LightOrange, Lighter[Purple], Darker[Brown], and simi-
lar modifications. Here are a couple of examples using different styling options:
Plot[Sin[x], {x,-4,4},
PlotStyle->{RGBColor[0.8,0.2,0.2], Dashed}]
Plot[Sin[x], {x,-4,4},
PlotStyle->{Thick, Darker[Green], Dotted}]
1.0 1.0

0.5 0.5

-4 -2 2 4 -4 -2 2 4

-0.5 -0.5

-1.0 -1.0

221
PlotRange. We can use this option to specify the range of y-coordinates to include in the plot.
The most often used variants of this option are
PlotRange->All
PlotRange->{a,b}
When PlotRange is set to All, Mathematica plots all the points on the graph. When PlotRange
is set to an actual range {a,b}, Mathematica plots only those points on the graph that have y-
coordinates in the range a ≤ y ≤ b. When this option is not set, Mathematica displays only
points where it thinks the function is “interesting”. Sometimes, this does not work very well and
Mathematica needs to be told to plot all points. For example, compare the outputs of the following
two attempts to plot (sin x)/x:
Plot[Sin[x]/x, {x,-20,20}]
Plot[Sin[x]/x, {x,-20,20}, PlotRange->All]
0.4 1.0

0.3 0.8

0.2 0.6

0.1 0.4

0.2
-20 -10 10 20

-0.1
-20 -10 10 20

-0.2 -0.2

D.3.1.3. Superimposing several graphs. One can superimpose the graphs of several functions
by supplying the functions as a list in the first argument of a Plot command. The following
command will plot sin x and x − x3 /6 + x5 /120.
Plot[{Sin[x], x-xˆ3/6+xˆ5/120}, {x,-5,5}]
One can use the PlotStyle option to get more easily distinguishable curves as seen in the follow-
ing example. Notice that the argument of PlotStyle is a list of two lists: the first list is the list
of plotting options applied to the graph of sin x and the second list is the list of plotting options
applied to the graph of x − x3 /6 + x5 /120.
Plot[{Sin[x], x-xˆ3/6+xˆ5/120}, {x,-5,5},
PlotStyle->{{Thick, Red}, {Blue}}]
3

-4 -2 2 4

-1

-2

-3

222
 D.4. Dynamic output
D.4.1. Manipulate. A major strength of Mathematica is its ability to handle dynamically
updated variables. The basic command to access Mathematica’s dynamic features is Manipulate.
It takes two or more inputs: the first input is the expression that we want to update dynamically
and the subsequent inputs are lists indicating ranges of parameters that appear in the expression.
Here are two examples of dynamic plots:
Manipulate[
Plot[Cos[x], {x,-a,a}, PlotRange->{-1,1}], {a,5,0.001}]
Manipulate[
Plot[Sin[x+c]+d, {x,-7,7}, PlotRange->{-3.5,3.5}],
{c,0,3}, {d,-2,2}]

c
a
d

1.0

0.5 2

-4 -2 2 4
-6 -4 -2 2 4 6

-1

-0.5
-2

-3
-1.0

Note the little + buttons next to the sliders in the above dynamic outputs. Clicking on one of
those opens a navigation menu for that particular slider. The navigation menu allows us to see
animated versions of the output, to find the current value of the parameter, or to set the parameter
equal to a particular value. For example, we can see that setting the two parameters to c = 1.57
and d = −1 in the second dynamic plot results in a sine curve that looks exactly as the graph
y = cos x − 1.

1.57
d

-1

-6 -4 -2 2 4 6

-1

-2

-3

223
 Although dynamic evaluations are a wonderful tool, they also require some care to avoid falling
in certain traps. Consider the following dynamic plot.
Clear[f]
f[x_] := Exp[-xˆ2]
Manipulate[Plot[f[x], {x,-a,a}, PlotRange->{0,1}], {a,2,4}]

1.0

0.8

0.6

0.4

0.2

-2 -1 0 1 2

We see a nice bell-shaped curve. Now, let us take a look at another dynamic plot.
Clear[f]
f[x_] := x*Exp[-x]
Manipulate[Plot[f[x], {x,-a,a}, PlotRange->{-1,1}], {a,2,4}]
You may expect that after executing these commands, the pair of plots will look something like this:

a
a

1.0
1.0

0.8

0.5

0.6

-2 -1 1 2
0.4

0.2 -0.5

-2 -1 0 1 2 -1.0

224
However, the reality will look more like this:

a a

1.0 1.0

0.5 0.5

-2 -1 1 2 -2 -1 1 2

-0.5 -0.5

-1.0 -1.0

This is not a bug! Since both plots are supposed to present the graph of the function f (x) and are to
be updated dynamically, changing the definition of f (x) will result in both plots changing to match
the new definition of f (x). For example, if we execute the commands
Clear[f]
f[x_] := xˆ2
both plots above will change to dynamic plots of the parabola y = x2 . The safest way to avoid such
confusion is not to reuse function names when using Manipulate. For example, the dynamic plots
2
of the functions e−x and xe−x above should have been entered as
Clear[f1, f2]
f1[x_] := Exp[-xˆ2]
Manipulate[Plot[f1[x], {x,-a,a}, PlotRange->{0,1}], {a,2,4}]

f2[x_] := x*Exp[-x]
Manipulate[Plot[f2[x], {x,-a,a}, PlotRange->{-1,1}], {a,2,4}]

D.5. Calculus-related Mathematica commands

Mathematica has built-in commands for many of the operations that you will learn in calculus.
In this section, we list some of those commands.

D.5.1. Limits. To evaluate a limit in Mathematica, we use the built-in command Limit is
easy. For example, to evaluate
e x − 1 − x − x2 /2
lim ,
x→0 x3
we just execute the following the command.
In[1]:= Limit[(Eˆx-1-x-xˆ2/2)/xˆ3, x->0]
Out[1]:= 1/6
(Recall that Mathematica uses E for e = 2.71828 . . . )
One-sided limits can be evaluated using the option Direction. The value assigned to this
option can be any nonzero number, but only the sign of the number really matters: if a > 0, then
225
Direction->a specifies a left limit; if a < 0, then Direction->a specifies a right limit. Here are
several examples:
In[1]:= Limit[Abs[x]/x, x->0, Direction->1]
Out[1]:= -1

In[2]:= Limit[Abs[x]/x, x->0, Direction->-1]

Out[2]:= 1

In[3]:= Limit[Exp[1/(x-1)], x->1, Direction->1]

Out[3]:= 0

In[4]:= Limit[Exp[1/(x-1)], x->1, Direction->-1]

Out[4]:= Infinity
In particular, the last input/output pair demonstrates what happens when a limit is infinite. It should
be noted that if no direction is specified and the two-sided limit does not exist, then Mathematica
automatically reverts to evaluation of the right limit (Direction->-1). This explains, for example,
the following evaluation.
In[1]:= Limit[Abs[x]/x, x->0]
Out[1]:= 1
Mathematica can also handle limits at ±∞. To evaluate a limit as x → ∞, we simply replace
the number in the second argument of Limit by the word Infinitiy (or by -Infinity, for limits
as x → −∞). For example,
In[1]:= Limit[Log[x]/x, x->Infinity]
Out[1]:= 0

In[2]:= Limit[\Sqrt[xˆ2+4x]/(3x - 7), x->-Infinity]

Out[2]:= -1/3
Finally, a few words about limits which do not exist. We already know what happens when a
limit is infinite. What if the limit fails to exist for a different reason? For example, it is known from
calculus that the limit lim sin(1/x) does not exist. Mathematica reports that fact as follows:
x→0
In[1]:= Limit[Sin[1/x], x->0]
Out[1]:= Interval[{-1,1}]
So, in this case Mathematica returns a whole range as the value of the limit instead of a single
number. That means that the limit does not exist.
D.5.2. Differentiation. Mathematica has a built-in command to calculate the derivative of a
function. To find f 0 (x), given f (x), we simply execute one of the following two commands
In[1]:= f’[x]
Out[1]:= D[f[x],x]
For example, if f (x) = x3 + cos(x2 ), then
In[1]:= f’[x]
Out[1]:= 2x - 2x Sin[xˆ2]

226
 In[2]:= D[f[x],x]
Out[2]:= 2x - 2x Sin[xˆ2]
Note that the latter version of the differentiation command works also for partial derivatives of
multivariable functions. For example, if
f (x, y) = sin(x + 2y) + (x2 − y2 ) exp(x2 + 3y2 ),
then D[f[x,y],x] and D[f[x,y],y] calculate, respectively, the x- and y-partials of f (x, y).
Clear[f]
In[1]:= f[x_,y_] := Sin[x+2y] + (xˆ2-yˆ2)Exp[xˆ2+3yˆ2]
D[f[x,y], x]
Out[2]:= (2xˆ3-2xyˆ2+2x)Eˆ(xˆ2+3yˆ2) + Cos[x+2y]

In[3]:= D[Sin[x+2y] + (xˆ2-yˆ2)Exp[xˆ2+3yˆ2], y]

Out[3]:= (6yxˆ2-6yˆ3-2y)Eˆ(xˆ2+3yˆ2) + 2Cos[x+2y]
D.5.3. Integration. Mathematica’s basic command for evaluation of integrals is Integrate.
It can evaluate both definite and indefinite integrals. In its two basic forms, this command looks
like this R
Integrate[f[x], x] evaluates f (x) dx
Rb
Integrate[f[x], {x,a,b}] evaluates a f (x) dx
The second version of Intergate tries to evaluate the given integral exactly in “closed form,”
but that is not always possible. In such cases, it may be preferable to settle for a very close numeric
approximation to the value of the definite integral. Mathematica provides another command for
this purpose. The command
NIntegrate[f[x], {x,a,b}]
Rb
uses sophisticated algorithms to approximate the definite integral a f (x) dx and the result is usu-
ally fairly close to the exact answer. By default, the answer that you obtain by executing this
command is accurate to at least 6 digits; for higher precision, the options PrecisionGoal and
WorkingPrecision can be used (see the documentation entry on NIntegrate). Here are some
examples of integral evaluations:
In[1]:= Integrate[ArcTan[2x+3], x]
Out[1]:= (3/2)ArcTan[3+2x] + x*ArcTan[3+2x] - (1/4)Log[5+6x+2xˆ2]

In[2]:= Integrate[Exp[-xˆ2], {x,0,2}]

Out[2]:= (1/2) Sqrt[Pi] Erf[2]

In[3]:= NIntegrate[Exp[-xˆ2], {x,0,2}]

Out[3]:= 0.882081

In[4]:= NIntegrate[Exp[-xˆ2], {x,0,2}, PrecisionGoal->10]

Out[4]:= 0.882081

In[5]:= NIntegrate[Exp[-xˆ2], {x,0,2}, WorkingPrecision->10]

Out[5]:= 0.8820813907

227
 In[6]:= NIntegrate[Exp[-xˆ2], {x,0,2},
WorkingPrecision->10, PrecisionGoal->10]
Out[6]:= 0.8820813908
The last three examples in this series are meant as a warning that use of the high-precision op-
tions of NIntegrate requires some care. The safest solution is to set both PrecisionGoal and
WorkingPrecision to the same number (as in the last example). What exactly goes wrong in the
two earlier examples is beyond the reach of this introduction to Mathematica; suffice it to say that
if you need to use these options, you should first study carefully the documentation.
D.6. User-defined commands
Mathematica’s includes some programming features which allow the sophisticated user to de-
fine his/her own commands. User-defined commands share features with both built-in commands
and user-defined functions. Similar to built-in commands, user-defined commands may have mul-
tiple inputs (of varying types) and may produce all kinds of output, including numeric answers,
text, plots, etc. On the other hand, similar to user-defined functions, we must type and execute
the definition of a user-defined command before we can use it. In the laboratories, we will make
extensive use of such custom commands, so it is important to learn how to apply them properly,
even if the actual “nuts and bolts” of the command seem mysterious.
The definition of the command newCommand always takes one of the following two forms:
newCommand[*list of arguments*] := Module[
*some Mathematica code*
]
newCommand[*list of arguments*] := DynamicModule[
*some Mathematica code*
]
Here, *list of arguments* and *some Mathematica code* are placeholders for the actual
list of arguments and the definition of the command. In order to be able to use such a command,
we must do two things:
• We must familiarize ourselves with the arguments the command accepts and the output
that it returns. These are usually described in the manual, shortly before or after the
definition of the command.
• We must execute the cell containing the definition of the command. Until then all occur-
rences of the name of the command will appear in blue (e.g., newCommand) and Mathe-
matica will simply repeat any command containing a call to the new command.
The next three pages demonstrate these points. They present printouts of the definition of a
module that appears in an actual TU calculus laboratory assignment and several calls to it. On the
first printout, the definition of newton has not been executed and the “output” merely repeats the
input. On the second printout, the definition has been executed but the arguments do not match
the expected ones—in one case two inputs are missing, and in the other n is not an integer. On the
third printout, all the rules have been followed and the call to newton has resulted in an actual plot.

228
12.3. Geometric Interpretation
Newton's method for solving f HxL = 0 is equivalent to the following geometric process. Pick a starting number
x0 near the root r. Draw the tangent line to the graph of f at the point Hx0 , f Hx0 L L. The x-intercept, x1 , of the
tangent line is generally closer to r than x0 was. The process is then repeated until some prescribed accuracy is
attained.

The module below, called newton, repeats the above construction n times, where n is a given positive integer.
It stores the values of x0 , x1 , ..., xn in a list named data and produces a plot of the iteration. The calling
parameters of the module are:
f = the function f ;
a = the left end of some interval that contains r;
b = the right end of the interval;
x0 = the initial value to start the iteration;
n = the number of iterations desired.

newtonAf_, a_, b_, x0_, n_E := ModuleA8g<, gAx_E = x - f@xD  f'@xD;

data = NestList@g, x0, nD; data2 = Flatten@Table@
8data@@iDD, 0, data@@iDD, f@data@@iDDD<, 8i, Length@dataD<DD;
data3 = Partition@data2, 2D;
data4 = Drop@data3, -1D;
p1 = Plot@f@xD, 8x, a, b<, PlotStyle -> 8Red, Thick<,
AxesLabel -> 8"x", "y"<, PlotRange -> AllD;
p2 = Graphics@8Hue@.6D, Line@data4D<D;
Show@8p1, p2<DE

We now apply this module to the function f HxL = x2 - 2 with x0 = 0.5 and the interval @0, 2.5D in order to
obtain a geometric illustration of the first iteration (n = 1).
Clear@fD;
fAx_E := x^2 - 2
In[1]:= newton@f, 0, 2.5, .5, 1D
Out[1]= newton@f, 0, 2.5, 0.5, 1D
12.3. Geometric Interpretation
Newton's method for solving f HxL = 0 is equivalent to the following geometric process. Pick a starting number
x0 near the root r. Draw the tangent line to the graph of f at the point Hx0 , f Hx0 L L. The x-intercept, x1 , of the
tangent line is generally closer to r than x0 was. The process is then repeated until some prescribed accuracy is
attained.

The module below, called newton, repeats the above construction n times, where n is a given positive integer.
It stores the values of x0 , x1 , ..., xn in a list named data and produces a plot of the iteration. The calling
parameters of the module are:
f = the function f ;
a = the left end of some interval that contains r;
b = the right end of the interval;
x0 = the initial value to start the iteration;
n = the number of iterations desired.
In[1]:= newtonAf_, a_, b_, x0_, n_E := ModuleA8g<, gAx_E = x - f@xD  f'@xD;
data = NestList@g, x0, nD; data2 = Flatten@Table@
8data@@iDD, 0, data@@iDD, f@data@@iDDD<, 8i, Length@dataD<DD;
data3 = Partition@data2, 2D;
data4 = Drop@data3, -1D;
p1 = Plot@f@xD, 8x, a, b<, PlotStyle -> 8Red, Thick<,
AxesLabel -> 8"x", "y"<, PlotRange -> AllD;
p2 = Graphics@8Hue@.6D, Line@data4D<D;
Show@8p1, p2<DE

We now apply this module to the function f HxL = x2 - 2 with x0 = 0.5 and the interval @0, 2.5D in order to
obtain a geometric illustration of the first iteration (n = 1).
In[2]:= Clear@fD;
fAx_E := x^2 - 2
In[4]:= newton@f, .5, 2D
Out[4]= newton@f, 0.5, 2D
In[5]:= newton@f, 0, 2.25, .5, 1.5D
NestList::intnm : Non-negative machine-sized integer expected at position 3 in NestList@g$116, 0.5, 1.5D. ‡
12.3. Geometric Interpretation
Newton's method for solving f HxL = 0 is equivalent to the following geometric process. Pick a starting number
x0 near the root r. Draw the tangent line to the graph of f at the point Hx0 , f Hx0 L L. The x-intercept, x1 , of the
tangent line is generally closer to r than x0 was. The process is then repeated until some prescribed accuracy is
attained.

The module below, called newton, repeats the above construction n times, where n is a given positive integer.
It stores the values of x0 , x1 , ..., xn in a list named data and produces a plot of the iteration. The calling
parameters of the module are:
f = the function f ;
a = the left end of some interval that contains r;
b = the right end of the interval;
x0 = the initial value to start the iteration;
n = the number of iterations desired.
In[2]:= newtonAf_, a_, b_, x0_, n_E := ModuleA8g<, gAx_E = x - f@xD  f'@xD;
data = NestList@g, x0, nD; data2 = Flatten@Table@
8data@@iDD, 0, data@@iDD, f@data@@iDDD<, 8i, Length@dataD<DD;
data3 = Partition@data2, 2D;
data4 = Drop@data3, -1D;
p1 = Plot@f@xD, 8x, a, b<, PlotStyle -> 8Red, Thick<,
AxesLabel -> 8"x", "y"<, PlotRange -> AllD;
p2 = Graphics@8Hue@.6D, Line@data4D<D;
Show@8p1, p2<DE

We now apply this module to the function f HxL = x2 - 2 with x0 = 0.5 and the interval @0, 2.5D in order to
obtain a geometric illustration of the first iteration (n = 1).
In[3]:= Clear@fD;
fAx_E := x^2 - 2
In[5]:= newton@f, 0, 2.5, .5, 1D
y

2
Out[5]=
1

x
0.5 1.0 1.5 2.0 2.5
-1

-2
 APPENDIX E

Answers to Selected Exercises

√ √ √
§1. 1. Q; none; Q; R. 3. 14, 14, 14; equilateral. 4. (x −1)2 +(y−1)2 +(z−1)2 = 10. 6. A plane parallel to
the yz-plane and 3 units behind it. 8. All points on or between the planes y = −1 and √y = 2.√ 10. All points between
the spheres centered at O that have radii 2 and 3. 13. x2 + y2 + z2 < 4, z > 0. 15. ( 32 2, 32 2, −3). 17. (0, −3, −1).
√ √ √ √ √ √
18. ( 2, 7π/4, 4). 20. (2, π/6, −2). 24. (− 32 3, − 32 , 0). 25. (0, 2, 2). 26. ( 6, π/4, arccos(2/ 6)).
27. (2, 3π/4, 3π/4). 30. r2 = 2z, ρ2 sin2 φ = 2ρ cos φ. 32. z = r, φ = π/4. 33. z = −2, a horizontal plane.
35. x + y + z = 4, a sphere with radius 2 and center O. 37. x2 + y2 − 2z = 0.
2 2 2
40. The sphere with radius 2
and center O. 42. All points above the xy-plane that lie on or between the spheres with center O and radii 2 and 4.
44. All points below the xy-plane that lie on or inside the infinite cylinder with radius 1 and axis the z-axis.
√ √ √
§2. 2. 2, h−5, −2i , 2 2, h4, −7i. 4. 3, h5, −3, −4i , 21 11, h3, −2, −1i. 8. 0. 10. 3. 12. π/3.
√
14. arccos(−1/
√ 28) ≈ 101 ◦
. 15. π/2, π/3, π/6.√ 18. Neither. 19. h−1, −2, 2i. 21. h−7, −3, 5i. 23. h2, −1, 3i.
26. 2 1329. 28. 80. 30. Undefined. 32. 12 110. 35. abe1 − b cos φe2 − b sin φe3 .

§3. 2. x = t, y = −2 + 3t, z = 3 + 2t (t ∈ R); x = y+2 3 = 2 .

z−3
5. arccos − 3 √1438 ≈ 139◦ .


3. Skew.
7. 3x − 2z + 3 = 0. 9. −x + 2y + z + 5 = 0. 10. There is no such plane. 11. 4x + 5y − 3z + 1 = 0. 13. x−2.4 2 = y+0.4
3 = 5z .
16. π/2. 17. 89 .

§4. 1. (a) hln 2, sin 2, ei; (b) h1, 2, 1i. 2. t22t+1 , √1−4t
2
, et . 4. 3t2 , 0, −6e2t . 7. 0, 21 e2 − e−2 , 0 .





2

§5. 1. (a) Yes, t = 0; ; (c) no. 2. Yes, r1 (0) = r2 (2) = h0, 0, 0i. 4. √3 , 0, √1 . 5. x = −1, y = 1. 7. x = −5,


10 10
z = 2. 8. (0, ±5, −2). 9. Smooth. 11. Piecewise smooth. 16. 1 + ln 2. 17. (a) s(a) = 0, s(b) = L;
14. √16 .

 3 53
(e) f(s) = 1 + 5 sin 4 ln(1 + 5 ) , 1 + 5 cos 4 ln(1 + 5 ) , 5 , 0 ≤ t ≤ 5.
3s 3s 3s 3s

§6. 1. Domain: all points (x, y); range: [0, ∞). 3. Domain: the points (x, y, z) for which (x, y) is not on the
hyperbola x2 − y2 = 1; range: all real numbers. 5. The points on and above the line y = −x/2, except for those that
lie on the circle x2 + y2 = 1. 6.√Levels −1 and √ 0: no level curves; level 1: the line y = 0; level 2: the lines y = 1 and
y = −1; level 3: the lines y = 2 and y = − 2. 8. Level −1: no level curve; level 0: the level curve consists of a
single point, (0, 0); for k = 1, 2, 3, the k-level curve is the circle x2 + y2 = k. 9. −29. 11. 0. 12. D.N.E. 14. 2.
15. All (x, y). 16. All (x, y) , (0, 0). 17. The points with x2 + y2 + 4z2 ≤ 4.

§7. 1. Parabolas, parabolas, hyperbolas, hyperbolic paraboloid. 3. Ellipses, ellipses, ellipses, ellipsoid.
6. (x − 2)2 + 4y2 + (z + 1)2 = 9, an ellipsoid centered at (2, 0, −1). 7.
r(t) =ph3 cos t, 3 sin t, 3 cos t + 3 sin t + 2i,
 2 z 2= x + y .
z) = − 2 y + z2  + 4, y, z , −∞ < y, z < ∞.
2 2 2 2
0 ≤ t ≤ 2π. 8.
The elliptic paraboloid 11. r(y,
13. (a) r(x, y) = x, y, 4 − x − y , x + y ≤ 4; (b) r(u, v) = u cos v, u sin v, 4 − u , 0 ≤ u ≤ 2, 0 ≤ v ≤ 2π.
2 2

14. r(u, v) = h(b + a cos v) cos u, (b + a cos v) sin u, a sin vi, 0 ≤ u, v ≤ 2π.

§8. 1. 4x3 − 8xy3/2 , −6x2 y1/2 . 3. 3(3x + 7yz)−1 , 7z(3x + 7yz)−1 , 7y(3x + 7yz)−1 .

x−1
x
x−1
4. 2x2 + xy x2 + xy + y2 + x2 + xy + y
2 ln x2 + xy + y2 ,
2x2 + xy x2 + xy + y2



2 3/2 1/2
. 7. 12x
− 8y 2, −12xy 2
,
2 −1/2
−3x y . 8. −(2x + y) cos
2

x + xy +2y − 2 sin
2 2

x + xy + y 1/2, −(2x + y)(x2+ 2y) cos
2 2

x + xy + y − sin2 x + 2xy3+
y2yz ,
2 2

−(x + 2y) cos x + xy + y − 2 sin x + xy + y . 10. −12y . 13. 6z + 4yz e , 12xz + 24xyz + 8xy z e .
2 2 2 2 3 2yz

2(y − 3z)et − 3(x + 2z)e−t + 2(2y − 3y)e2t

14. 2y(2t + 3) + (2x + 3y2 )(2t). 17. . 18. (4xy − y2 )6s + (2x2 − 2xy)3t,
xy + 2yz − 3xz
(4xy − y2 )2t + (2x2 − 2xy)3s. 21. (−y sin xy + y3 x ln 3)2s + (−x sin xy + 3 x ) + 2z(t + 1),
233
(−y sin xy + y3 x ln 3) + (−x sin xy + 3 x )(−2t) + 2z(sg + 1). 23. −6e−24 . 24. 79 . 27. ln 7 + 37 (x − 2) + 57 (y − 1).
29. 1 (x, y) = (y − 2), 2 (x, y) = 0. 30. (a) 0, 0. 31. (a) a, d.
√
§9. 1. √25 e2xy (x + 2y + x2 y + 2xy2 + y3 + 2x3 ). 4. 60 1
10. 6. 7, in the direction of h−3, 2, −6i.

x−2 y+1 z+2

§10. 1. 3x + 2y + z − 2 = 0 and = = z + 2. 3. x + y − 2z − 5 = 0 and x − 2 = y + 1 = .
3 2 −2
5. (a) x = x0 + tx0 , y = y0 + ty0 , z = z0 + tz0 . 9. 16 , − 2 , 8 . 10. The tangent plane to both is x − 2y + z = 2.
3 1 1

11. −x + π2 y − z = 0. 14. −2y + z + 1 = 0. 17. (− 32 , 10 9 , − 27 ). 18. Smooth. 20. Piecewise smooth. 21. Not even
26
π
piecewise smooth: it is non-smooth on each curve xy = − 2 + 2kπ.

§11. 1. A saddle point at ( 21 , −2). 3. Local minima at (±1, ±1) and a saddle point at (0, 0). 6. Local maxima
at √ , ±1
( ±1 √ ), local minima at ( ±1√ , ∓1 √ ), and a saddle point at (0, 0). 8. A local minimum at (1, 12 ) and a saddle point
2 2 √ 2 √2 √
at (0, 0). 10. 8, 20. 12. − 34 √22 , 43 √22 . 14. 2/3. 16. The cube of side length 5.

§12. 1. ± 54 5−1/4 . 4. The minimum is 7; there is no maximum. 6. S /3, S /3, S /3.

8 π.
8
§13. 1. (a) − 27 . 3. 2.5. 4. 20 23 . 6. 8(ln 3)−2 . 8. 3
4 ln 3 − ln 2. 10. 25
12. −1.
R 1 R y2
§14. 1. −21 31 . 3. sin 1−cos 1. 4. 257.2. 6. 12 ln 2. 8. 12 (1−sin 1). 10. 7. 13. 4 ln 6. 14. 0 0 f (x, y) dxdy.
√
R 2 R 2y R 4 R 4−y R ln 2 R 2
16. 0 0 f (x, y) dxdy + 2 0 f (x, y) dxdy. 18. 0 20. 14 1 − e−1 . 21. 14 ln 17. 23. 4.9.


e x f (x, y) dydx.
24. 6. 26. 23 π + 15
56
3 . 30. 3 π.
. 27. 243.2. 29. 16 2

√ √
√

§15. 2. (e − 1) π2 . 3. 15 15 − 7 7 π6 . 5. 18 1 − 5e−4 . 6. 8π. 8. 36 − 20 5 π. 10. 38 π − 32 π


3 9 . 11. 2 .

§16. 2. 12 1 − e−16 . 3. 0.7. 5. 64 1 16

15 . 7. 2 (1 − sin 1). 9. 3 .
√
15 π. 3. 3 2 2 − 1 − π. 4. 2π . 6. 8π. 8. 35 .
§17. 1. 34 2 π. 13. 3 π.
2π 2π
11. 81 14

2

√ √
√ √
2n−1 π
§18. 1. 94 6. 3. 8π 2 − 2 . 5. π 2 6 − 83 . 8. 64 6
, , ,
7 7 7
= 13



9 , ( 5 0). 9. 32, 6 6 6 . 12. In 22 · · · 2 2 .

§19. 1. R : x2 + 4y2 ≤ 4. 3. R : x2 − 2xy + 2y2 ≤ 2. 5. R is a parallelogram with vertices (0, 0), (1, 1), (1, −3),
and (2, −2). 6. (c) u. 7. −6. 9. 12 ln 2 − 16 5
. 11. (a) D : [1, 4] × [1, 6]; (c) −4x2 − 2y2 ; (d) 7.5. 13. (a) T is not
one-to-one on the planes u = 0 and v = 0; (b) u2 v; (c) p!q!r!s!/(p + q + r + s + 3)!.
 
2x z y
§20. 2. ye2y cos xy, xe2y cos xy + 2e2y sin xy . 4. , ,


. 5. Open and simply-connected.
x2 + yz x2 + yz x2 + yz
6. Simply-connected, but not open. 7. Open and connected, but not simply-connected. 9.
Open, but not connected.
11. No. 13. Yes, f (x, y) = xe
y + c. 14. Yes, f (x, y) = y2 arctan x+ c. 16. curl F = −4xz2 , x2 y + 4yz2 , −x2 z ,


div F = −6xyz. 18. curl F = 2xze2xy , −2yze2xy − x sin xz, 2y cos xy , div F = e2xy + 2x cos xy − z sin xz. 20. Yes,
f (x, y, z) = xy2 cos z + c. 21. No. 23. No, because div(xy2 e1 + yz2 e2 + zx2 e3 ) , 0.
√ √
1 7
. 5. 3π3 13. 7. 14. 9. 15π. 10. 32 − cos 1 + sin 2.


§21. 1. 840 125 5 − 1 . 3. 24
§22. 1. −1 + 2e−1 . 3. −4. 5. 256 cos 8. 6. 4π.
§23. 2. −24π. 4. 162. 6. 0. 7. 43 − 12 π2 . 9. 91 73 . 10. 3π. 12. 2π.
√
§24. 1. 171 14. 3. 16π. 5. 713/180. 7. − 23 π.
13
§25. 1. 8π. 3. − 12 . 5. −20π. 7. − 13
12 .
√ √ √
§26. 1. 0. 2. 275 3 π. 5. 83 π + 4 3. 6. 216π. 7. 3 π.
8 2
8. 32π2 .

234