Calculus III

Nazarbayev University

Lecture notes based on Calculus: Early Transcendentals, 10th Ed.

by Anton, Bivens, and Davis

Daniel Oliveira da Silva

Last updated: August 20, 2021

Please send any corrections to [email protected].

Contents

11 Three Dimensional Space and Vectors 3

11.1 Rectangular Coordinates in Three Dimensions . . . . . . . . . . . . . . . . . 3
11.2 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
11.3 Dot Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
11.4 Cross Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
11.5 Parametric Equations of Lines . . . . . . . . . . . . . . . . . . . . . . . . . . 16
11.6 Planes in 3-space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
11.7 Quadric Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
11.8 Cylindrical and Spherical Coordinates . . . . . . . . . . . . . . . . . . . . . . 22

12 Vector-valued Functions 26
12.1 Vector-valued Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
12.2 Calculus of Vector-Valued Functions . . . . . . . . . . . . . . . . . . . . . . 28
12.3 Arc Length Parametrizations . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
12.4 Unit Tangent, Normal, and Binormal Vectors . . . . . . . . . . . . . . . . . 35
12.5 Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
12.6 Motion Along a Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

13 Functions of Several Variables 42

13.1 Introduction to Functions of Several Variables . . . . . . . . . . . . . . . . . 42
13.2 Limits and Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
13.3 Partial Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
13.4 Differentiability, Differentials, and Local Linearity . . . . . . . . . . . . . . . 56
13.5 The Chain Rule for Functions of Several Variables . . . . . . . . . . . . . . . 59
13.6 Directional Derivatives and Gradients . . . . . . . . . . . . . . . . . . . . . . 62
13.7 Tangent Planes and Normal Vectors . . . . . . . . . . . . . . . . . . . . . . . 67
13.8 Extreme Values of Functions of Two Variables . . . . . . . . . . . . . . . . . 69

14 Multiple Integrals 75
14.1 Double Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
14.2 Double Integrals over Non-Rectangular Regions . . . . . . . . . . . . . . . . 78
14.5 Triple Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
14.7 Change of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

1
 14.3 and 14.6 Integrals in Cylindrical and Spherical Coordinates . . . . . . . . . 91
14.4 Surface Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

15 Vector Calculus 99
15.1 Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
15.2 Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
15.3 Conservative Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
15.4 Green’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
15.5 Surface Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
15.6 Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
15.7 The Divergence Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
15.8 Stokes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

2
Chapter 11

Three Dimensional Space and Vectors

Now that you have learned the basics of differentiation and integration, we will extend these
ideas to functions of several variables. To do this, we will spend some time introducing
some ideas which extend the tools used to study functions of a single variable to functions
of several variables. For simplicity, we will focus our attention on functions of two or three
variables (mostly two variables), but most of the concepts we introduce (though not all) will
apply to any number of variables. When they do not apply to any number of variables, this
will be stated explicitly. We begin our discussion with an introduction to three dimensional
space.

11.1 Rectangular Coordinates in Three Dimensions

As in the case of functions of a single variable, it is useful to represent functions of sev-
eral variables graphically. Recall that for functions of the form y = f (x), we graph using
a coordinate system with two coordinates (x, y). For functions of two variables, we will
typically use x and y to denote the variables, and the function will typically be written in
the form z = f (x, y). Thus, to graph these functions, we will use a rectangular coordinate
system where the position of a point is specified by a triple of numbers (x, y, z), which tell
you the distance from the origin along three axes, known as the x-axis, y-axis, and z-axis,
respectively. For simplicity, we will refer to this coordinate system as 3-space.

Right-handed Coordinate Systems

As you may expect, 3-space will be very similar to the xy-plane (which will now be referred
to as 2-space), but with an additional axis. To avoid any ambiguities as to how these axes
should be arranged, we will use what is called a right-handed coordinate system. In this
system, the axes are depicted as in the following diagram:

3
The reason for the term “right-handed” is because you can determine the relationship of
the axes to each other by using your right hand and pointing your thumb, index, and
middle fingers, respectively, in perpendicular directions, you will obtain the x, y, and z axes,
respectively. Try it now.
To make sure we understand how this works, let us plot a few points in 3-space.

Example 1. Plot the following points in 3-space:

1. A = (1, 0, 0)

2. B = (0, −1, 0)

3. C = (0, 0, −1)

4. D = (1, −1, −1)

Solution. The points should appear as in the following graph:

4
Distances in 3-space
As in 2-space, it is useful to have a way of measuring the distance between points in 3-space.
This is given by the following formula:

Definition 1. Consider two points P = (x1 , y1 , z1 ) and Q = (x2 , y2 , z2 ) in 3-space. The

distance between these two points, denoted by d(P, Q), is
p
d(P, Q) = (x1 − x2 )2 + (y1 − y2 )2 + (z1 − z2 )2 .

Note the similarity with the formula you learned for 2-space.

Example 2. Find the distances between the following points:

1. P = (2, 3, −1) and Q = (4, −1, 3)

p √
Solution. d(P, Q) = (2 − 4)2 + (3 + 1)2 + (−1 − 3)2 = 36 = 6

2. A = (1, 3, 1) and B = (−1, 3, −1)

p √
Solution. d(A, B) = (1 + 1)2 + (3 − 3)2 + (1 + 1)2 = 2 2

5
11.2 Vectors
The next tool we need for our discussion of functions of several variables is that of a vector.
Vectors are probably one of the most useful concepts to arise from mathematics, so we will
give you a brief introduction to them here. We begin with the following definition:

Definition 2. A vector is a quantity with an associated direction.

An associated definition which we will need is the following:

Definition 3. A scalar is a quantity without an associated direction.

These definitions may seem strange, but here are some examples to illustrate them:

• Velocity: Recall that a velocity is a speed in a specified direction, such as travelling

north at 100 km/hr. Thus, velocity is a vector.

• Forces: When an object is subjected to a force, the force has to be applied in a

particular direction. For example, we can lift a box up (direction) with a force of a
certain magnitude (quantity). Thus, force is a vector.

• Mass: When we discuss mass, we state only a number; an object does not have a mass
in a particular direction. Thus, mass is a scalar.

• Time: When we measure time, it is not dependent on any direction. Thus, time is a
scalar.

Note that scalars are nothing more than just numbers. Thus, we will use the terms scalars
and numbers interchangeably.
In our discussions, we will use vectors in both 2-space and 3-space. Mathematically, we
represent a vector by a tuple hx, yi or hx, y, zi in two or three dimensions, respectively. The
numbers x, y, and z are known as the components of the vector. As you can imagine, writing
vectors this way repeatedly is a bit cumbersome, and so we often refer to vectors by using a
single letter; in textbooks, it is customary to denote vectors by lower-case boldface letters,
such as v = hx, yi. However, when writing by hand, it is customary to denote vectors by
letters with arrows above them, such as → −v = hx, yi. We will follow the standard convention
in this course.
This method of representation has a simple geometric interpretation: a vector can be
thought of as an arrow in 2-space or 3-space, as appropriate, with tail fixed to the origin,
and tip located at the coordinate (x, y) or (x, y, z).

Example 3. Plot the following vectors:

1. h1, 1i

Solution. The desired vector is

6
2. h−1, 3i

Solution. The desired vector is

3. h1, 0, 1i

Solution. The desired vector is

7
4. h2, −1, 1i

Solution. The desired vector is

8
Vector Algebra
To apply vectors to various problems in science and engineering, it is necessary to define
algebraic operations on vectors, just as we do for numbers. For the moment, we will introduce
only three operations on vectors: addition, multiplication by a scalar, and subtraction.

Definition 4. Let v = hx1 , y1 i and w = hx2 , y2 i. We define the vector sum v + w by

v + w = hx1 + x2 , y1 + y2 i.

Remark 1. Recall that we follow the convention that unless stated otherwise, the definition
applies to three or more variables in an analogous way.

Example 4. Compute v + w for the following vectors.

1. v = h1, 2i, w = h5, −1i.

Solution. v + w = h6, 1i

2. v = h−3, 4, 5i, w = h2, 1, −3i.

Solution. v + w = h−1, 5, 2i

Definition 5. Let v = hx, yi and k be a scalar. We define the product kv of the vector v
and the scalar k by
kv = hkx, kyi.
In the special case where k = −1, we may write (−1)v = −v.

Example 5. Let k and v be as given below. Compute kv.

1. k = 2, v = h2, 5i

Solution. kv = h4, 10i

2. k = −1/2, v = h2, −10, 3i

Solution. kv = h−1, 5, −3/2i

Definition 6. Let v and w be vectors. We define vector subtraction by

v − w = v + (−w)

Example 6. Compute v − w for the following vectors.

1. v = h1, 2i, w = h5, −1i.

9
 Solution. v − w = h−4, 3i

2. v = h−3, 4, 5i, w = h2, 1, −3i.

Solution. v − w = h−5, 3, 8i

In addition to the basic operations above, we will also need to introduce some operations
which are unique to vectors.

Definition 7. Let v = hx, yi. We define the norm kvk of v as the length of the vector when
drawn in the standard way. By the distance formula, this is given by
p
kvk = x2 + y 2

in 2-space, and p
kvk = x2 + y 2 + z 2
in 3-space.

Example 7. Let v be given below. Compute kvk.

1. v = h1, 2i
√
Solution. kvk = 5

2. v = h3, 0, 4i

Solution. kvk = 5

Definition 8. A unit vector is a vector vector whose norm is 1.

Note that any vector can be scaled to be a unit vector. That is, if v is ANY vector, then
1
u= v
kvk

is a unit vector. The process of scaling a vector so that it is a unit vector is called normalizing.
While there are many, MANY unit vectors in 2-space and 3-space, we want to pay
attention to some particular unit vectors, for which we have special notation. In 3-space, we
denote
i = h1, 0, 0i
j = h0, 1, 0i
k = h0, 0, 1i.
The notation in 2-space is analogous.

10
11.3 Dot Products
In addition to the operation of scalar multiplication, where we multiply a vector and a scalar,
it will also be useful to multiply vectors by vectors. As we will see, there are multiple ways
of defining such a product. One method, which will be discussed here, is known as the dot
product.

Definition 9. Let u = hu1 , u2 i, v = hv1 , v2 i be vectors in 2-space. The dot product u · v of

u and v is defined as
u · v = u1 v1 + u2 v2 .

Example 8. Compute the dot product u · v for the following vectors.

1. u = h1, 2i, v = h3, 4i

Solution. u · v = 1 · 3 + 2 · 4 = 11

2. u = h7, 2, 1i, v = h−1, 3, 3i

Solution. u · v = 7 · (−1) + 2 · 3 + 1 · 3 = 2

If we represent vectors as arrows on the plane or in space, then it can sometimes be

useful in applications to be able to determine the angle between them. For this, we have the
following theorem.

Theorem 1. If u and v are non-zero vectors in 2- or 3-space, and if θ is the angle between
them, then
u · v = kukkvk cos θ.
In the special case where u ⊥ v, then u · v = 0.

Another useful fact about norms is that there is a connection between norms and dot
products, as we see in the following theorem.

Theorem 2. If v is a vector in 2-space or 3-space, then

v · v = kvk2 .

Orthogonal Decomposition
Consider three vectors v, w1 , and w2 , where w1 ⊥ w2 . As you will see later, it is often
useful to write one vector as a combination of other vectors. More specifically, suppose we
wanted to write
v = aw1 + bw2

11
for some scalars a and b. If this can be done (in many cases, it cannot), then we can try to
find the coefficients in the following way: take the equation above and take the dot product
of both sides with the vector w1 . You should get
v · w1 = aw1 · w1 + bw2 · w1
= akw1 k2 .
Solving for a, we see that
v · w1
a= .
kw1 k2
By a similar computation, we can see that
v · w2
b= .
kw2 k2
We can interpret this result in the following way: the scalar a tells us how much of the vector
v points in the direction of the vector w1 , and b tells us how much the vector v points in
the direction of w2 . This leads us to the following definition:
Definition 10. Let v and w be vectors. We define the projection of v along w as the vector
v·w
projw v = w.
kwk2
Example 9. Let v = h3, 2i, w1 = h1, 1i, and w2 = h1, −1i. Compute projw1 v and projw2 v,
and write the vector v as a combination of w1 and w2 .
Solution. Using the formula in the definition, we have
v · w1
projw1 v = w1
kw1 k2
5
= h1, 1i
2
 
5 5
= ,
2 2
and
v · w2
projw2 v = w2
kw2 k2
1
= h1, −1i
2
 
1 1
= ,− .
2 2
We can then write v as
v = projw1 v + projw2 v
   
5 5 1 1
= , + ,−
2 2 2 2
= h3, 2i .

12
11.4 Cross Products
Warning! The topic covered in this section applies only in 3-space!

We will now discuss a second way of multiplying vectors, called the cross product of two
vectors. It is a bit more complicated than the dot product we discussed previously, but the
result is a vector, unlike the dot product. Before we can define the cross product, we need
some preliminary definitions.

Determinants
Definition 11. Consider a 2 × 2 matrix of numbers, such as
 
a1 a2
.
b1 b2

The determinant of this matrix is defined as

a1 a2
= a1 b 2 − a2 b 1 .
b1 b2

Example 10. Compute the determinant

1 2
.
3 4

Solution. According to the formula above, the determinant is

1 2
= 1 · 4 − 2 · 3 = −2.
3 4

In addition to determinants of 2 × 2 matrices, we will also need determinants of 3 × 3

matrices.

Definition 12. Consider a 3 × 3 matrix of numbers, such as

 
a1 a2 a3
 b1 b2 b3  .
c1 c2 c3

The determinant of this matrix is defined as

a1 a2 a3
b b b b b b
b 1 b 2 b 3 = a1 2 3 − a2 1 3 + a3 1 2 .
c2 c3 c1 c3 c1 c2
c1 c2 c3

13
Example 11. Compute the determinant

1 0 2
0 3 −1 .
−1 2 1

Solution. Using the definition, we have

1 0 2
3 −1 0 −1 0 3
0 3 −1 = 1 −0 +2
2 1 −1 1 −1 2
−1 2 1
= 1 · (3 + 2) − 0 · (0 + 1) + 2 · (0 + 3)
= 11

Cross Products
Using determinants, we can now define a new method of multiplying vectors.

Definition 13. Let v = ha1 , a2 , a3 i and w = hb1 , b2 , b3 i be vectors in 3-space. The cross
product of v and w is the vector v × w defined by

i j k
v × w = a1 a2 a3
b 1 b2 b3
a2 a3 a a a a
=i −j 1 3 +k 1 2 .
b2 b3 b1 b3 b1 b2

Example 12. Let v = h0, 1, 2i, and w = h1, 2, 0i. Compute v × w.

Solution. Using the formula above, we obtain

i j k
v×w = 0 1 2
1 2 0
1 2 0 2 0 1
=i −j +k
2 0 1 0 1 2
= −4i + 2j − k
= h−4, 2, −1i.

Cross products have several useful properties, some of which may be a little unexpected.

14
Theorem 3. Let v and w be vectors in 3-space. The following properties hold:
1. v × w = −w × v;
2. v · (v × w) = 0 and w · (v × w) = 0;
3. If v and w are vectors in 3-space with angle θ between them, then kv×wk = kvkkwk sin θ;
4. v × w = 0 if and only if v = cw for some scalar c.
Example 13. Let v and w be as in Example 12. Compute w × v.
Solution. We could compute this using the definition, but it is easier to apply Theorem 3.
We saw previously that
v × w = h−4, 2, −1i.
By Theorem 3, we must then have that
w × v = −h−4, 2, −1i
= h4, −2, 1i.

The Triple Scalar Product

We have now defined two methods for multiplying vectors: the dot product, which yields a
scalar, and the cross product, which yields a vector. In applications of calculus, it is often
useful to combine the two operations, as in the following definition.
Definition 14. Let u, v, and w be vectors in 3-space. The triple scalar product of u, v,
and w is the scalar given by
u · (v × w).
While you could compute this directly using only the definitions you have learned, there
is a slight shortcut, which is the content of the following theorem.
Theorem 4. Let u = hu1 , u2 , u3 i, v = hv1 , v2 , v3 i, and w = hw1 , w2 , w3 i. Then
u1 u2 u3
u · (v × w) = v1 v2 v3 .
w1 w2 w3
Example 14. Let u = h3, −2, −5i, v = h1, 4, −4i, and w = h0, 3, 2i. Compute u · (v × w).
Solution. Using Theorem 4, we have
3 −2 −5
u · (v × w) = 1 4 −4
0 3 2
= 49.

15
11.5 Parametric Equations of Lines
In calculus II, you spent some time discussing parametric equations for lines in 2-space. This
is not particularly useful in the discussion of functions of one variable, but it is much more
useful when discussing functions of several variables. We will thus discuss this topic again,
but in three dimensions. As always, what we say will also apply for any other number of
dimensions.
Consider two points P = (x, y, z) and P0 = (x0 , y0 , z0 ) in 3-space. We define the vector
−−→
P0 P to be the vector
−−→
P0 P = hx − x0 , y − y0 , z − z0 i.
Graphically, this corresponds to a vector starting at P0 , and ending at P . Since we represent
−−→
vectors graphically as line segments, it follows that the vector P0 P can be interpreted as the
line segment between P0 and P .
This is fine, but we want equations for lines, which are supposed to be infinitely long in
both directions. To fix this, we make the following observation:
−−→ −−→
Fact: If P , P0 , and P1 lie on the same line, then P0 P = tP0 P1 for some scalar t.

This suggests that a line is completely determined by a point P0 , and a direction vector v.
We express this by
−−→
P0 P = tv.
If v = ha, b, ci, then we can write this explicitly as
hx − x0 , y − y0 , z − z0 i = hta, tb, tci.
The only way these vectors can be equal is if each component is equal, which gives us the
equations
x = x0 + at y = y0 + bt z = z0 + ct.
This leads us to the following definition:
Definition 15. Let P0 = (x0 , y0 , z0 ), and let v = ha, b, ci be a vector in 3-space. Then the
parametric equations for the line through P0 in the direction of v are
x = x0 + at, y = y0 + bt, z = z0 + ct.
Example 15. Find the parametric equations for the lines through the points P0 and in the
direction of the vectors v given below.
1. P0 = (4, 2), v = h−1, 5i.

Solution. Using the formulas above, you should obtain

x = 4 − t, y = 2 + 5t.

16
 2. P0 = (1, 2, 3), v = h4, 5, −7i.

Solution. Using the formulas above, you should obtain

x = 1 + 4t, y = 2 + 5t, z = 3 − 7t.

Example 16. Let P0 = (2, 4, −1) and P1 = (5, 0, 7).

1. Find the parametric equations for the line through P0 and P1 .

−−→
Solution. We first need a direction vector. Note that P0 P1 will point in the right
direction, so we use that for v. Thus

v = h3, −4, 8i.

From this, we get the equations

x = 2 + 3t, y = 4 − 4t, z = −1 + 8t.

2. Find the coordinates where this line passes through the xy-plane.

Solution. This line passes through the xy-plane when the z-coordinate is zero. It is
easy to see that this happens when
1
t=− .
8
It follows that
13 7
x= , y= .
8 2
Thus, the line crosses at (13/8, 7/2, 0).

Notation: It is often cumbersome to write out equations like this. Thus, if we define

r = hx, y, zi, r0 = hx0 , y0 , z0 i,

we can combine these equations into the vector equation

r = r0 + tv.

Note the similarity to the standard equation for a line in the plane y = mx + b.

17
11.6 Planes in 3-space
Just as we can represent a line in 2-space (or 3-space) by an algebraic equation, we would
like to do the same thing for planes in 3-space. To do this, we need a preliminary definition.

Definition 16. A vector that is perpendicular to a plane is called a normal vector for that
plane.

With this definition in mind, consider two points P = (x, y, z) and P0 = (x0 , y0 , z0 ) on
−−→
a plane, and let P0 P = hx − x0 , y − y0 , z − z0 i be the corresponding vector from P0 to P .
Suppose that n is a normal vector for this plane. By Theorem 1, we must have
−−→
n · P0 P = 0.

Let r = hx, y, zi and r0 = hx0 , y0 , z0 i. We can then rewrite the equation above as

n · (r − r0 ) = 0.

If we write n more explicitly as n = ha, b, ci, it is easy to see that this dot product becomes

a(x − x0 ) + b(y − y0 ) + c(z − z0 ) = 0.

Thus, we have just arrived at the following theorem:

Theorem 5. The equation of the plane containing the point (x0 , y0 , z0 ) whose orientation is
determined by the normal vector n = ha, b, ci is

a(x − x0 ) + b(y − y0 ) + c(z − z0 ) = 0.

Example 17. Find the equation of the plane through the point P0 = (0, 2, −1) and perpen-
dicular to the vector n = h2, 3, 4i.

Solution. Using the formula above, we get

2x + 3(y − 2) + 4(z + 1) = 0.

Example 18. Find the equation of the plane determined by the points P0 = (1, 2, −1),
P1 = (2, 3, 1) and (3, −1, 2).

Solution. To apply the equation in Theorem 5, we need a point and a normal vector, which
we are not given. However, we are given three points on the plane. We can then find two
vectors on the plane, given by
−−→ −−→
u = P0 P1 = h1, 1, 2i and v = P0 P2 = h2, −3, 3i.

18
Recall from Theorem 3 that the cross product u × v will be perpendicular to both u and v.
Thus, let us choose
n = u × v = h9, 1, −5i.
With this, we can now write the desired equation using the point P0 , which gives us

9(x − 1) + y − 2 − 5(z + 1) = 0.

Example 19. Find the angle of intersection between the planes 2x − 4y + 4z = 6 and
6x + 2y − 3z = 4.

Solution. The only time we have discussed angles in this course was when discussing prop-
erties of the dot product and cross product. Recall from Theorem 1 and Theorem 3 that for
two vectors in 3-space, we have

u · v = kukkvk cos θ and u × v = kukkvk sin θ,

where θ is the angle between the two vectors.

To apply this, we must find the normal vectors for the planes, as the angle between the
planes is determined by the angles between the normal vectors. We can do this by simply
looking at the coefficients in front of the variables; thus, we can let

u = h2, −4, 4i and v = h6, 2, −3i.

If we use the dot product, we can see that

u·v −8 4
cos θ = = =− .
kukkvk 6·7 21

From this, we get that θ ≈ 79◦ . Note that this is not the only solution. (Why?)

11.7 Quadric Surfaces

Previously, you have learned about graphing curves in space. Some special curves which
you learned about in calculus II were the conic sections. These were curves described by
equations of the form
Ax2 + Bxy + Cy 2 + Dx + Ey + F = 0.
We now want to discuss the analogous surfaces in three dimensions, which are known as
quadric surfaces.

Definition 17. A quadric surface is a surface in 3-space which is described by a quadratic

equation in the variables x, y, and z.

19
 As an example, recall that an ellipse is a curve described by the equation
x2 y 2
+ 2 = 1.
a2 b
Such a curve will have the following form:

The three-dimensional analogue of this is called an ellipsoid, and it is defined by an equation

of the form
x2 y 2 z 2
+ 2 + 2 = 1,
a2 b c
and it will have an appearance of the following form:

Notice the similarity with the two-dimensional version.

Graphs of functions convey a lot of information about functions, and so it is useful to
know how to sketch them. To sketch quadric surfaces (and functions of several variables

20
later), we apply the following simple trick: Choose a fixed number for one of the variables
in the equation that defines the surface. This makes the equation have only two variables,
which is now a two-dimensional curve. This curve can be sketched using what you know
from calculus I and II. Such a curve is called a trace of the surface, and is analogous to
plotting points to find graphs of functions. Once several traces are sketched, they can be
stacked together to form a graph of the surface. We demonstrate this with an example.

Example 20. Sketch a graph of the surface z 2 = x2 + y 2 .

Solution. We begin by first drawing some traces of the surface. To start, let us pick some
values of z to substitute into the equation. For example, we can pick the values z = ±1,
z = ±2, z = ±3, which give us the equations x2 + y 2 = 1, x2 + y 2 = 4, and x2 + y 2 = 9,
respectively, which define circles of radii 1, 2, and 3, respectively, at height z = ±1, z = ±2,
and z = ±3, respectively. If you then draw a smooth surface connecting all these curves
together, you should obtain something similar to the following diagram:

Note that the picture should be symmetric. The asymmetric appearance of the surface in
the diagram above is due to limitations of the software used to generate it.

Remark 2. We will not devote any more time to the discussion of quadric surfaces. The
important idea from this section is the idea of sketching surfaces using traces, which we will
see again later in the semester.

21
11.8 Cylindrical and Spherical Coordinates
At the end of calculus II, you saw that there are other ways of writing down locations in
2-space mathematically. In addition to denoting locations on the plane using rectangular
coordinates (x, y), you also learned about the polar coordinate system (r, θ), where coordi-
nates are specified in terms of a distance from the origin r and an angle θ with respect to
the x-axis. We will now do something similar in three dimensions. As we will see, there are
multiple ways of doing this, and which is more useful will depend on the context.

Cylindrical Coordinates
The first method of describing points in 3-space is the cylindrical coordinate system, and is
an extension of the polar coordinate system in 2-space. Simply put, we specify a point in
3-space by first specifying its location on the xy-plane, and then specifying its height above
or below the plane, as in the following diagram:

Thus, we typically write points in space using the coordinates (r, θ, z). Based on what you
have already learned in calculus II, you should be able to see that we can convert from one
coordinate system to the other using the following formulas: To see how this works, let us

Cylindrical to Rectangular Rectangular

p to Cylindrical
x = r cos θ r = x +y2 2

y = r sin θ tan θ = xy
z=z z=z

consider some examples.

Example 21. Convert the rectangular coordinate (1, −1, 2) to cylindrical coordinates.

Solution. Using the formulas above, we compute r by

p √
r = 12 + (−1)2 = 2.

22
For the polar angle θ, we must solve the equation
tan θ = −1.
This occurs when θ = 5π/4 + nπ for n = 0, 1, 2, 3, . . .. Since we are in the fourth quadrant
on the xy-plane, we choose n = 1, so√that θ = 7π/4. Finally, z = 2, so that our coordinate
in cylindrical coordinates becomes ( 2, 5π/4, 2).
Example 22. Convert the following cylindrical coordinates to rectangular coordinates, and
plot the points in 3-space.
1. P = (1, π/4, 2)

Solution. Using the formulas above, we obtain

√ √
2 2
x= , y= , z = 2.
2 2

√
2. Q = ( 2, 5π/4, 3)

Solution. Using the formulas above, we obtain

x = 1, y = −1, z = 3.

We plot the points on the following axes:

23
Spherical Coordinates
The second method we use to denote coordinates in 3-space is known as the spherical coor-
dinate system. In this system, position is specified by three coordinates (ρ, θ, φ). Here, ρ is
the distance from the origin, θ is the angle formed with respect to the x-axis, and φ is the
angle formed with respect to the z-axis, as in the following diagram:

By using some basic geometry, it is easy to see that we can convert spherical coordinates to
rectangular coordinates by using the following formulas: Let us look at some examples.

Spherical to Rectangular Rectangular

p to Spherical
x = ρ sin φ cos θ ρ = x2 + y 2 + z 2
y = ρ sin φ sin θ tan θ = xy
z = ρ cos φ cos φ = √ 2 z 2 2
x +y +z

√
Example 23. Convert the rectangular coordinate (1, −1, 2) to spherical coordinates.
Solution. Using the formulas above, we compute ρ by
q √
ρ = 12 + (−1)2 + ( 2)2 = 2.

For θ, we must solve the equation

tan θ = −1.
As we saw in Example 21, we can choose the solution θ = 7π/4. Finally, for φ, we must
solve the equation √ √
2 2
cos φ = q √ = .
2
12 + (−1)2 + ( 2)2
Note that the point lies in the fourth quadrant on√the xy-plane, so that we must choose
φ = 7π/4. We thus conclude that our point (1, −1, 2) in rectangular coordinates becomes
(2, 7π/4, 7π/4) in spherical coordinates.

24
Surfaces in Cylindrical and Spherical Coordinates
One of the things which makes these new coordinate systems so useful is that some surfaces
in 3-space which are represented by very complicated formulas can sometimes be represented
very simply by switching coordinate systems. As an example, consider the sphere or radius
3 centered at the origin. In rectangular coordinates, the equation for such a circle is

x2 + y 2 + z 2 = 9.

If we use the definition of ρ, we see that this equation can be rewritten in terms of ρ in the
simple form
ρ2 = 9,
which you can reduce to the even simpler form ρ = 3.
p
Example 24. Consider the cone z = x2 + y 2 . Rewrite this equation in terms of cylindrical
coordinates.
p
Solution. Recall the definition r = x2 + y 2 in cylindrical coordinates. This equation then
takes the simple form z = r.

Example 25. Sketch a graph of the surface defined by the equation r = 1.

Solution. Rewriting this in rectangular coordinates gives us x2 + y 2 = 1. (How do you

know the equation is in cylindrical coordinates at the beginning?) This equation describes
a circle centered at the origin on the xy-plane. Note that this equation is independent of
the z-coordinate. Thus, at any height z, there will be a circle of radius 1. This describes a
cylinder centered at the origin, as in the following diagram:

25
Chapter 12

Vector-valued Functions

At the end of calculus II, you learned about curves defined by parametric equations. These
are equations of the form
x = f (t),
y = g(t).
That is, instead of defining y in terms of x, we define both x and y in terms of another
variable, which is often denoted by t.
Because each equation depends on another paramater t, it is a simple matter to generalize
this concept to any number of equations. For our purposes, we will only consider curves in
2-space or 3-space, so in addition to the above system of equations, we will also consider a
set of three equations, such as
x = f (t),
y = g(t),
z = h(t).
As we will see, this concept is much more useful when discussion functions of several variables
than functions of single variables.

12.1 Vector-valued Functions

Consider a set of parametric equations, such as

x = f (t),
y = g(t), (12.1)
z = h(t).

Just as in the case of two equations, we can make a table of values for x, y, and z for given
values of t which will give us a set of coordinates in 3-space. We can then draw a smooth
curve through all the points, giving us a curve in 3-space, where each point has coordinate
(f (t), g(t), h(t)) for some t. Since we may think of a point (x, y, z) as a vector hx, y, zi, then

26
the equations in (12.1) define a new object, which is a function which outputs a vector in
2-space or 3-space. We define this formally below.

Definition 18. A vector-valued function r(t) is a function which, for each number t, deter-
mines a vector r(t) = hf (t), g(t), h(t)i. The set of all t for which r(t) is defined is the domain
of r(t), and it is the domain of the three functions f , g, and h.

As always, this definition works for any number of equations. When we state vector
functions, it is customary to simply write them in the form

r(t) = hf (t), g(t), h(t)i

rather then writing out the whole system as we did in equation (12.1).

Example 26. Sketch a graph of the vector function r(t) = ht, t2 i.

Solution. As was discussed above, a good place to start is with a table of values.

t 0 1 2 3 4
x 0 1 2 3 4 etc.
y 0 1 4 9 16

Plotting these in 2-space, we obtain the following graph:

27
Example 27. Sketch a graph of the vector function r(t) = hcos t, sin t, t/5i.
Solution. As before, we start with a table of values.

t 0 π/6
√ π/4
√ π/3 π/2
x 1 3/2 √2/2 1/2
√ 1
etc.
y 0 1/2 2/2 3/2 0
z 0 π/30 π/20 π/15 π/10

Plotting these in 3-space (not a simple task, but worth trying), we obtain the following
graph:

12.2 Calculus of Vector-Valued Functions

Now that we have defined vector functions, we would like to do calculus with them. By this,
we mean do things like compute derivatives, integrals, etc. As we will see, this will require
modifying definitions slightly so that they apply to vector-valued functions.

Limits
Recall that in calculus I, we stated the following (informal) definition of a limit:

28
 We say that the function f (x) has limit L as x approaches a if, the value of f (x)
gets closer to L as the value of x gets closer to a.

There is a more precise statement of the definition in section 1.4 of the textbook, but we
will not need it here. The only thing we need to know is how to compute limits, which you
learned in calculus I. Assuming you know how to compute limits of scalar functions, we can
state the following definition for vector functions:

Definition 19. Let r(t) be defined on an interval containing t = a, except maybe at t = a.

We say that the vector L is the limit of the function r(t) at t = a if

lim kr(t) − Lk = 0.
t→a

In this case, we write

lim r(t) = L.
t→a

Note that the function in the first limit above is a scalar function; the norm k · k turns the
vector into a scalar. Thus, we should be able to compute that limit using what we already
know. There is one problem with this definition: how do you find the vector L? To find this
vector, we may use the following theorem:

Theorem 6. If r(t) = hx(t), y(t), z(t)i, then

D E
lim r(t) = lim x(t), lim y(t), lim z(t) ,
t→a t→a t→a t→a

if these limits exist.

Example 28. Compute the following limits:

1. lim ht, cos ti

t→0

Solution. By using Theorem 6, we have

lim ht, cos ti = h0, cos 0i = h0, 1i.

t→0

2. lim t2 , t, sin t
t→π

Solution. By using Theorem 6 again, we have

lim t2 , t, sin t = π 2 , π, sin π = π 2 , π, 0 .

t→π

29
Derivatives
Now that we understand limits, we may use them to compute derivatives. We do this using
the following definition:

Definition 20. Let r(t) be a vector function. We define the derivative r0 (t) of r(t) to be
the vector function defined by

r(t + h) − r(t)
r0 (t) = lim ,
h→0 h
if this limit exists. In this case, we say that r(t) is differentiable at t. Additionally, if
r0 (a) 6= 0, we refer to the resulting vector as the tangent vector to r(t) at t = a.

From this definition and Theorem 6, we have the following simple trick to computing
derivatives of vector functions:

Theorem 7. Let r(t) = hf (t), g(t), h(t)i. If f , g, and h, are differentiable, then

r0 (t) = hf 0 (t), g 0 (t), h0 (t)i.

Example 29. Compute the derivatives of the following vector functions.

1. r(t) = ht2 , cos ti

Solution. Using Definition 20 and Theorem 7, and what we know of derivatives of

scalar functions, we have
r0 (t) = h2t, − sin ti.

2. r(t) = ht2 , cos t, 5i

Solution. Using Definition 20 and Theorem 7, and what we know of derivatives of

scalar functions, we have
r0 (t) = h2t, − sin t, 0i.

In many ways, vector functions behave a lot like scalar functions, as we see in the following
theorem:

Theorem 8. Let r1 (t) and r2 (t) be differentiable, let c be a constant vector, and let k be a
scalar. Then the following derivative formulas hold:
d
1. (c) = 0
dt

30
 d
2. (kr1 (t)) = kr01 (t)
dt
d
3. (r1 (t) ± r2 (t)) = r01 (t) ± r02 (t)
dt
These formulas should look familiar; they are identical to the formulas for scalar functions.
In addition to the above formulas, we have some new formulas that are really unique to
vector functions:
Theorem 9. Let f be a differentiable scalar function, and let r1 (t) and r2 (t) be differentiable
vector functions. Then the following formulas hold:
d
1. (f (t)r1 (t)) = f 0 (t)r1 (t) + f (t)r01 (t)
dt
d
2. (r1 (t) · r2 (t)) = r01 (t) · r2 (t) + r1 (t) · r02 (t)
dt
d
3. (r1 (t) × r2 (t)) = r01 (t) × r2 (t) + r1 (t) × r02 (t)
dt
While these equations apply only to vector functions, you should observe that they still
resemble the product rule that you already know.

Integrals
Just as we defined derivatives of vector functions, we would like to define integrals of vector
functions. This is done in largely the same way as it was done in calculus I; thus, we will not
go into the details. Instead, we will state the main result for integrals of vector functions:
Theorem 10. If r(t) = hf (t), g(t)i, then the definite integral of r(t) from t = a to t = b is
Z b Z b Z b 
r(t) dt = f (t) dt, g(t) dt .
a a a

The corresponding expression in higher dimensions is analogous.

Example 30. Compute the following definite integrals.
Z 1
1. h2t, (t + 1)2 i dt
0

Solution. Using Theorem 10, we have

Z 1 Z 1 Z 1 
2 2
h2t, (t + 1) i dt = 2t dt, (t + 1) dt
0 0 0
 
7
= 1, .
3

31
 Z π
2. ht sin t, cos t, ti dt
0

Solution. Using Theorem 10 again, we have that

Z π Z π Z π Z π 
ht sin t, cos t, ti dt = t sin t dt, cos t dt, t dt
0 0 0 0
π2
 
= π, 0, .
2

As you might expect, vector integrals have various properties which are analogous to the
scalar case.

Theorem 11. Let r1 (t) and r1 (t) be vector-valued functions that are continuous on the
interval [a, b]. Let k be a scalar. Then the following properties hold:
Z b Z b
1. kr1 (t) dt = k r1 (t) dt
a a
Z b Z b Z b
2. r1 (t) ± r2 (t) dt = r1 (t) dt ± r2 (t) dt
a a a

In addition, to definite integrals, it will sometimes be useful to have a notion of indefinite

integrals for vector functions. First, we need a definition:

Definition 21. Let r(t) and R(t) be vector functions. We say R is the antiderivative of r
if R0 (t) = r(t). We denote this with the notation
Z
r(t) dt = R(t).

As with scalar functions, antiderivatives of vector functions are not unique.

Theorem 12. Let R(t) be an antiderivative of r(t). Then so is R(t) + C for any constant
vector C.

12.3 Arc Length Parametrizations

Frequently, when performing computations, the difficulty of the computations can be reduced
by changing the parameters involved. For example, in calculus I, you learned the technique
of u-substitution, which you may use to transform the indefinite integral
Z
x2
2
 2
2xex cos ex esin e dx

32
into the much simpler integral Z
eu du.

This is an example of a change of variables used to simplify a computation. When discussing

vector functions, we often refer to the independent variable as a parameter, and the resulting
change in variables as a change of parameters. In this section, we will discuss how to
handle changes of parameters for vector functions. We will then discuss a specific change
of parameter, known as the arc length parametrization, which will often take complicated
equations and reduce them to a very simple form.

Smooth Parametrizations
To avoid certain complications, it is necessary to restrict our attention to certain classes of
vector functions which don’t behave too wildly. The appropriate functions are specified in
the following definition:

Definition 22. We say that the vector function r(t) is smooth if the derivative r0 (t) exists
and is continuous, with r0 (t) 6= 0 for any t.

Example 31. Determine if the following vector functions are smooth.

1. r(t) = ht, t2 i

Solution. The derivative of r(t) is r0 (t) = h1, 2ti. Since the first entry in this vector
cannot be equal to zero for any value of t, it follows that r0 (t) 6= 0 for any t. Moreover,
1 and 2t are continuous functions of t. Therefore r(t) is smooth.

2. r(t) = cos ti + sin tj

Solution. The derivative of r(t) is r0 (t) = − sin ti + cos tj. Since sin t and cos t cannot
both be equal to zero for the same value of t, it follows that r0 (t) 6= 0 for any t.
Moreover, sin t and cos t are continuous functions of t. Therefore r(t) is smooth.

3. r(t) = h1, t2 i

Solution. This function is not smooth, because r0 (t) = h0, 2ti = 0 when t = 0. How-
ever, if we consider r(t) only on some interval which does not contain t = 0, then it is
smooth.

33
Arc Length
Recall that for a vector function r(t) = hx(t), y(t)i in two dimensions, the arc length of the
curve generated by r(t) from t = a to t = b is given by the integral
s
Z b  2  2
dx dy
s= + dt.
a dt dt

Using the norm notation, we can rewrite this as

Z b
s= kr0 (t)k dt.
a

It can be shown that this equation can be used for a function r(t) in any number of dimen-
sions. Thus, we will simply use this expression from now on.

Example 32. Compute the arc length for the curve defined by the expression r = cos ti +
sin tj + tk from t = 0 to t = π.

Solution. First, observe that

p √
kr0 (t)k = cos2 t + sin2 t + 1 = 2.

Thus Z π √ √
s= 2 dt = 2π.
0

Arc Length as a Parameter

As we will see a bit later, it is often quite convenient to make a change of parameters where
the new parameter is the arc length. There is one issue with this, though: as we saw in the
formula above, the arc length depends on both a starting point and an end point. So to use
arc length as a parameter, we do the following:

1. Choose a convenient but arbitrary starting point r0 = r(t0 ).

2. Use the formula above to compute the arc length from t0 to t. This should give you
an expression s = f (t) that (hopefully) can be solved for t in terms of s.

3. Rewrite the vector function in terms of s by substituting for t in the function.

4. If the curve needs to be oriented in the opposite direction, replace s with −s.

Example 33. Find the arc length parametrization for r(t) = ti + (2t + 2)j with t0 = 0 as
starting point.

34
Solution. Since we are given the starting point t0 = 0, we begin by computing the arc length:
Z t
0 0 0
Z t
√ √
s= kr (t )k dt = 1 + 4 dt0 = 5t.
0 0

We may solve this for t, obtaining

s
t= √ .
5
Substituting this into the function, we get
 
s 2s
r(s) = √ i + √ + 2 j.
5 5

Example 34. Parametrize the circle r(t) = cos ti + sin tj in terms of arc length, with t0 = 0,
so that the circle is oriented clockwise.
Solution. First, we observe that kr0 (t)k = 1, so that
Z t
s= 1 dt0 = t.
0

Thus, we can simply write r(s) = cos si + sin sj. However, you should check that this
parametrization gives a counterclockwise orientation. (Why?) Thus, we switch s with −s,
so that
r(s) = cos(−s)i + sin(−s)j.

12.4 Unit Tangent, Normal, and Binormal Vectors

For applications of calculus, there are several special vectors that are related to vector func-
tions which we use.

Unit Tangent Vectors

Previously, we discussed how the vector r0 (t) defines a vector that is tangent to the curve
r(t), in much the same way as the derivative f 0 (x) gives you the tangent line (or, it’s slope,
at least) to the function f (x). When we need to use tangent vectors in applications, it is
often necessary to use a unit vector. This leads us to the following definition:
Definition 23. If r(t) is a smooth function, then we define the unit tangent vector to be
the vector T(t) given by
r0 (t)
T(t) = 0 .
kr (t)k

35
Remark 3. Note that it is necessary that r(t) is smooth, so that the denominator in this
expression is never 0.
Example 35. Compute T(t) for the vector function r(t) = ht2 , t3 i at t = 2.
Solution. First, we compute that
r0 (t) = h2t, 3t2 i.
At t = 2, this becomes
r0 (2) = h4, 12i,
which has norm √ √ √
kr0 (2)k = 16 + 144 = 160 = 4 10.
From this, we see that
h4, 12i 1
T(2) = √ = √ h1, 3i.
4 10 10

Unit Normal Vectors

In addition to the unit tangent vector, there are two other special unit vectors related to a
smooth function r(t) which do not have analogs in the scalar case. The first of these is the
unit normal vector, which is defined below:
Definition 24. Let r(t) be smooth, and let T(t) be the corresponding unit tangent vector.
The unit normal vector is the vector N(t) defined by
T0 (t)
N(t) = .
kT0 (t)k
Of course, this requires that T(t) is smooth.
The reason for the terminology “normal” in this context is because it can be shown that
that T(t) and N(t) are perpendicular to each other, so that N(t) is normal to T(t).
Example 36. Let r(t) be as in Example 35. Find N(t) at t = 2.
Solution. In Example 35, we saw that

r0 (t) = h2t, 3t2 i,

so that
1
T(t) = p h2t, 3t2 i
2
(2t) + (3t ) 2 2

1
= √ h2t, 3t2 i
t 4 + 9t2
1
=√ h2, 3ti.
4 + 9t2

36
Differentiating this, we obtain
9t 1
T0 (t) = − 2 3/2
h2, 3ti + √ h0, 3i
(4 + 9t ) 4 + 9t2
1
= h−18t, 12i .
(4 + 9t2 )3/2
When t = 2, this becomes
1
T0 (2) = h−36, 12i.
403/2
Observe that
3
kT0 (2)k = ,
20
so that
1 1
N(2) = √ h−36, 12i = √ h−3, 1i.
12 10 10
You should check that this vector is perpendicular to T(2), and that it is a unit vector.

Unit Binormal Vectors

The next special vector of interest to us exists only in 3-space.
Definition 25. If r(t) is a smooth function so that T(t) is also smooth, then we define the
unit binormal vector as the vector B(t) such that

B(t) = T(t) × N(t).

The definition of this vector should make it clear why we call it a bi normal vector: it is
normal to both the unit tangent and the unit normal. This vector tends to not appear as
often as the others, and so we will not devote any more time to it.

Unit Tangent and Unit Normal Vectors and Arc Length

As we discussed in section 12.3, many expressions simplify significantly when expressed using
the arc length parametrization. For example, it can be shown that, when expressed in terms
of arc length, the expressions for T and N reduce to
r00 (s)
T(s) = r0 (s) and N(s) = .
kr00 (s)k

12.5 Curvature
In calculus 1, you learned about concavity, which gives you information about the way
in which a curve y = f (x) behaves graphically on the xy-plane. In this section, we will
generalize this concept to curves represented by vector functions in 2- or 3-space.

37
Definition 26. Suppose C is a smooth curve in 2- or 3-space that is defined by the vector
function r(s), which is parametrized with respect to arc length. Then the curvature κ(s) at
the point s is given by
dT
κ(s) = .
ds
Example 37. Consider a circle of radius a centered at the origin in 2-space. Compute the
curvature κ(s) for any s.

Solution. Recall that the circle can be parametrized by

r(t) = a cos ti + a sin tj.

To compute the curvature, we must first parametrize this in terms of arc length. For this,
choose t0 = 0, and set Z t
s= kr(t0 )k dt0
Z0 t
= a dt0
0
= at.
Thus, in terms of arc length, we have
s
t= .
a
Substituting this in the function, we have
s s
r(s) = a cos i + a sin j
a a
With this, the next step is to compute T and T0 . Recall that, in terms of arc length, we
have
T(s) = r0 (s),
so that
dT 00 1 s 1 s
= r (s) = − cos i − sin j,
ds a a a a
from which we deduce that
dT 1
= .
ds a
Notice that this expression does not depend on s.

Example 38. Compute the curvature κ(s) of the helix

r(t) = cos ti + sin tj + tk.

38
Solution. The first step is to reparametrize in terms of arc length. Recall from Example 32
that the arc length from t0 = 0 to t is given by
Z t√ √
s= 2 dt0 = 2t,
0

so that
s
t= √ .
2
We can thus write      
s s s
r(s) = cos √ i + sin √ j+ √ k.
2 2 2
Thus    
00 1 s 1 s
r (s) = − cos √ i − sin √ j + 0 k.
2 2 2 2
It follows that
1
κ(s) = kr00 (s)k = .
2

Example 39. Compute the curvature of the line r(s) = r0 + sv for any constant vectors r0
and v.

Solution. Since we are given the line in terms of arc length, we can compute the curvature
immediately:
κ(s) = kr00 (s)k
= k0k
= 0.
So, lines have zero curvature.

12.6 Motion Along a Curve

Recall from calculus I that if an object’s position is described by a function s(t), then its
velocity is v(t) = s0 (t), and its acceleration is given by a(t) = v 0 (t) = s00 (t). In addition, the
speed of the object is given by |v(t)|. Suppose now that an object’s motion in 2- or 3-space
is described by a position function r(t). Then the analogous expressions for velocity and
acceleration are
v(t) = r0 (t) and a(t) = r00 (t).
Note that all of these are vectors; if we know the object’s position at any time, then we know
it’s velocity (which here means speed and direction of travel), and its acceleration (which
includes rate and direction). For the speed, the corresponding expression becomes kv(t)k.

39
Example 40. Suppose that a particle’s position is described by the vector function r(t) =
h2 cos t, 2 sin ti.
1. Find the instantaneous velocity and speed of the particle at time t.

Solution. By definition, we have

v(t) = h−2 sin t, 2 cos ti,

from which we may deduce that kv(t)k = 2.

2. Show that the acceleration and velocity at any moment t are perpendicular to each
other.

Solution. First, we compute the acceleration:

a(t) = v0 (t) = h−2 cos t, −2 sin ti.

From this, you can check that

v(t) · a(t) = 0,
so that the two vectors are always perpendicular.

Another important concept when discussing motion in space is that of displacement.

Recall that the displacement of a particle whose motion is described by a scalar function
s(t) over the time interval [a, b] is

∆s = s(b) − s(a).

Note that this can be expressed in terms of velocity:

Z b
∆s = v(t) dt.
a

For vector functions, the expressions are analogous:

Z b
∆r = r(b) − r(a) = v(t) dt.
a

Example 41. Find the total displacement for an object moving according to r(t) = ht, t2 i
from t = 0 to t = 4.
Solution. Using the definition above, we have
Z b
∆r = h1, 2ti dt
a
= h4, 16i.

40
Projectile Motion
As an application of this theory we have developed, we want to consider the following
situation. Suppose a cannonball of mass m is shot out of a cannon so that its motion is
restricted to the xy-plane, with initial velocity vector v0 . How can we describe its motion?
Recall that Newton’s Second Law of Motion states that

F = ma.

In this example, let us make the following assumption: after the cannon is fired, the only
force acting on the cannonball is gravity, which we assume acts downward with acceleration
g ≈ 9.8 m/s2 . Then
F = −mgj.
We can thus say that
a = −gj.
Integrating, we get Z
v(t) = a dt = −gtj + c.

We are given that the moment the cannonball leaves the cannon, it has velocity v0 . Thus

v0 = v(0) = c,

so that
v(t) = −gtj + v0 .
To find the position vector r(t), let us assume the mouth of the cannon is located at r0 . If
we integrate again, we will obtain

gt2
Z
r(t) = v(t) dt = − j + tv0 + c.
2

Using the fact that r(0) = r0 , we see that

gt2
Z
r(t) = v(t) dt = − j + tv0 + r0 .
2
Thus, if we know the initial position, velocity, and acceleration, we can completely describe
the motion of the cannonball.

41
Chapter 13

Functions of Several Variables

In the previous chapter, we considered vector-valued functions of a single parameter t. We

now want to return to scalar functions, but which are functions of two or more variables.

13.1 Introduction to Functions of Several Variables

We begin with a simple (but informal) definition of functions of several variables.

Definition 27. A function of n variables is a rule that assigns a unique number f to points
(x1 , x2 , . . . , xn ) in n-space The resulting unique number is denoted f (x1 , x2 , . . . , xn ).

Some examples of such functions are

• f (x, y) = xy

• f (x, y, z) = x + y + z 2

• f (x, y, z) = cos(xy) − xy

As with functions of a single variable, there are several sets associated with a function that
are important.

Definition 28. If f is a function of n variables, then its domain is the set of all (x1 , . . . , xn )
for which f (x1 , . . . , xn ) is defined.

Example 42. Find the domains of the following functions.

1. f (x, y) = x2 + y 2

Solution. Since any value of x and y can be substituted, then the domain is the whole
xy-plane.
p
2. f (x, y) = 1 − x2 − y 2

42
 Solution. For this to be defined, we must have
x2 + y 2 ≤ 1.
Thus, the domain is the disc x2 + y 2 ≤ 1.

Definition 29. If f is a function of n variables, then the range of f is the set of all possible
values of f (x1 , . . . , xn ) for all (x1 , . . . , xn ) in the domain of f .
For functions of several variables, the range is often difficult to determine, and not nec-
essarily very useful to think about, so we will not devote any more time to it.

Graphing Functions of Two Variables

As with functions of a single variable, a lot of information about a function can be conveyed
by creating a graph of a function of two variables. To graph such functions, we follow the
procedure we used in section 11.7 to graph quadric surfaces, but in a slightly modified form.
For this, we need a new definition.
Definition 30. Let z = f (x, y) be a function of two variables, and let k be a real number.
The level curve at height k is the set of points (x, y) satisfying f (x, y) = k.
Example 43. Sketch the level curves of the function z = y − x2 at heights k = 0, 1, 2, 3.
Solution. By substituting the given numbers into z, we obtain the parabolas y = x2 , y =
x2 + 1, y = x2 + 2, and y = x2 + 3. Graphing these, we obtain the following graphs:

43
Note that level curves are similar to traces discussed in section 11.7, except that we always
choose the variable z to substitute numbers.
Once we have the level curves of a function f (x, y), we may proceed as before to graph
it.
Example 44. Sketch a graph of the function z = y − x2 .
Solution. Since we have already sketched the level curves, we simply draw them in 3-space
now:

If we connect these curves with a smooth surface, then end result is:

44
Example 45. Use level curves to sketch a graph of f (x, y) = y 2 − x2 .
Solution. We first graph the level curves at heights k = −2, −1, 0, 1, 2:

We then connect them with a smooth surface, giving us the following graph:

45
13.2 Limits and Continuity
We now want to discuss the concepts of limits and continuity for functions of several variables.
Recall that for functions of a single variable, we said that the limit as x → x0 exists if and
only if the right- and left-handed limits exist. That is, it cannot matter from which direction
along the number line you approach x0 . For functions of several variables, the situation is
a little more complicated; if we want to examine the value of f (x, y) as (x, y) approaches
(x0 , y0 ), there are infinitely many different ways we can approach the point (x0 , y0 ); any curve
r(t) which passes through (x0 , y0 ) can define a path which approaches (x0 , y0 ). Thus, our
definition of limits should somehow take this into consideration. For this reason, we begin
by defining limits first along specific curves. This will then be extended to a more general
type of limit.

Limits
Definition 31. Let C be a smooth curve represented by a vector function r(t) = hx(t), y(t)i
such that r(t0 ) = hx0 , y0 i, and let f (x, y) be a function of two variables. We define the limit
of f as (x, y) → (x0 , y0 ) along the curve C as

lim f (x, y) = lim f (x(t), y(t)).

(x,y)→(x0 ,y0 ) t→t0
along C

Remark 4. You should pay attention to the fact that the function in the limit on the right-
hand side is a scalar function of only the variable t. Thus, you should be able to evaluate
that limit using only what you have learned in calculus I.

Example 46. Let C be the curve defined by r(t) = hcos t, sin ti. Compute the limit of the
function f (x, y) = xy as (x, y) → (1, 0) along C.

Solution. First, observe that as t → 0, then (x(t), y(t)) → (1, 0). By definition, we have

lim xy = lim cos t sin t

(x,y)→(1,0) t→0
along C

=1·0
= 0.

Example 47. Let C1 be the curve defined by r(t) = ht, 0i, and let C2 be the curve defined
by r(t) = ht, ti. Compute the limits of the function
xy
f (x, y) =
x2 + y2

as (x, y) → (0, 0) along C1 and C2 .

46
Solution. First, observe that as t → 0, then (x(t), y(t)) → (0, 0) for both C1 and C2 . For C1 ,
we then have
xy 0
lim 2 2
= lim 2
(x,y)→(0,0) x + y t→0 t
along C1

= 0.
On the other hand, for C2 , we have

xy t2
lim = lim
(x,y)→(0,0) x2 + y 2 t→0 2t2
along C2
1
= .
2
Note that the two limits do not coincide.
These two examples should give you an idea that the concept of a limit becomes a bit more
tricky for functions of several variables. Thus, to define it properly, we need to introduce
some new terminology.

Definition
p32. The open ball of radius r centered at (x0 , y0 ) is the set of all points (x, y)
such that (x − x0 )2 + (y − y0 )2 < r.

The open ball is used to define more general open sets:

Definition 33. We say a subset D of the 2-space is open if, for each (x, y) in D, there is an
r > 0 such that the ball centered at (x, y) of radius r is contained inside D.

Example 48. Consider the subset D of 2-space shown in the diagram below:

In this diagram, we see that, for the point (x, y), the ball of radius r centered at (x, y) will
fit completely inside D. You should be able to see that you can do this for any point (x, y)
in D (though r may have to be adjusted for different coordinates).

47
 Just as there are open sets, we also have closed sets:

Definition 34. We say a subset C of 2-space is closed if it contains its boundary.

With these definitions, we can give a proper definition of limits:

Definition 35. Let f be a function of two variables (x, y), and assume that f is defined
everywhere in an open ball centered at (x0 , y0 ), except maybe at (x0 , y0 ). We say that

lim f (x, y) = L
(x,y)→(x0 ,y0 )

if, given any number  > 0, we can find another number δ > 0 such that
p
|f (x, y) − L| <  whenever 0 < (x − x0 )2 + (y − y0 )2 < δ.

Remark 5. This may sound more complicated than the definition you are used to for limits,
but it is saying the same thing: the closer you make (x, y) and (x0 , y0 ), the closer f (x, y))
and L will be. You should remember this, rather than the formal definition given above.

The obvious question to ask now is the following: what is the connection between limits
and limits along specific curves? A partial answer is contained in the following theorem:

Theorem 13. If f (x, y) → L as (x, y) → (x0 , y0 ), then f (x, y) → L as (x, y) → (x0 , y0 )

along any smooth curve. Conversely, if the limit fails to exist along any curve, or if the
limits along any two curves differ, then the limit does not exist.

Example 49. Show that the function

xy
f (x, y) =
x2 + y2

from Example 47 does not have a limit at (0, 0).

Solution. Based on the results of Theorem 13, it suffices to show that there are two smooth
curves C1 and C2 such that the limits along C1 and C2 differ. If we define C1 and C2 as in
Example 47, then we have
xy xy 1
lim = 0 and lim = .
(x,y)→(0,0) x2 + y2 (x,y)→(0,0) x2 +y 2 2
along C1 along C2

It follows that the limit of f as (x, y) → (0, 0) does not exist.

48
Continuity
Now that we have defined limits, we may use this to define continuity for functions of several
variables. This will be done largely in the same way as was done for functions of single
variables.

Definition 36. A function f (x, y) is said to be continuous at (x0 , y0 ) if

lim f (x, y) = f (x0 , y0 ).

(x,y)→(x0 ,y0 )

If D is an open set, we say that f is continuous on D if it is continuous at every point of D.

If D is all of 2-space, then we say that f is continuous everywhere.

This definition is analogous to the single-variable version. However, we do not yet know
how to apply it because we have not yet discussed which functions are continuous. For this,
we have the following theorem:

Theorem 14. The following facts about limits hold:

1. If g(x) is continuous at x0 and h(y) is continuous at y0 , then f (x, y) = g(x)h(y) is

continuous at (x0 , y0 ).

2. If h(x, y) is continuous at (x0 , y0 ) and g(u) is continuous at u = h(x0 , y0 ), then the

composition f (x, y) = g(h(x, y)) is continuous at (x0 , y0 ).

3. If f (x, y) is continuous at (x0 , y0 ) and the functions x(t) and y(t) are continuous at t0
with x(t0 ) = x0 and y(t0 ) = y0 , then the composition f (x(t), y(t)) is continuous at t0 .

This theorem is saying something pretty simple: if a function of several variables consists
of products or compositions of continuous functions of a single variable which are continuous,
then the original function is itself continuous as a function of several variables.

Example 50. Compute the following limits.

cos(xy)
(a) lim
(x,y)→(0,0) 1 + sin(xy)

Solution. By Theorem 14, the function is continuous, so that

cos(xy) cos(0)
lim = = 0.
(x,y)→(0,0) 1 + sin(xy) 1 + sin(0)

xy 2 + 2x2 y + 1
(b) lim
(x,y)→(2,1) y 2 − x2

49
 Solution. Since the denominator of the function is defined at (2, 1), it follows from
Theorem 14 that the function is continuous. Thus
xy 2 + 2x2 y + 1 2·1+2·4·1+1 11
lim 2 2
= = .
(x,y)→(2,1) y −x 4−1 3

(c) Let (
(1+x) sin y
y
if (x, y) 6= (0, 0),
f (x, y) =
1 if (x, y) = (0, 0).
Compute
lim f (x, y).
(x,y)→(0,0)

Solution. First, observe that the function

(
sin y
y
if y 6= 0,
h(y) =
1 if y = 0.

is continuous at y = 0 (why?), as is the function g(x) = 1 + x. By Theorem 14, it follows

that f (x, y) is also continuous. Thus, we have that

lim f (x, y) = 1.
(x,y)→(0,0)

Limits and Changes of Variables

Recall from calculus I that sometimes a complicated limit computation can be simplified by
a clever change of variables. The same can be said for functions of several variables. As an
example of this, let us consider the following example.

Example 51. Compute the limit

lim (x2 + y 2 ) ln(x2 + y 2 ).

(x,y)→(0,0)

Solution. To begin, we note that we can rewrite the function using cylindrical coordinates
as
(x2 + y 2 ) ln(x2 + y 2 ) = r2 ln r2 .
Using properties of logarithms, we can further simplify this as

(x2 + y 2 ) ln(x2 + y 2 ) = 2r2 ln r.

50
This lets us rewrite the desired limit as

lim (x2 + y 2 ) ln(x2 + y 2 ) = lim 2r2 ln r.

(x,y)→(0,0) r→0

To compute this, we apply L’Hospital’s rule:

ln r
lim 2r2 ln r = 2 lim 1
r→0 r→0
r2
1
r
= 2 lim
r→0 − r23
= − lim r2
r→0
= 0.

Thus, we conclude that

lim (x2 + y 2 ) ln(x2 + y 2 ) = 0.
(x,y)→(0,0)

13.3 Partial Derivatives

For a function of a single variable f (x), we defined the derivative of f at x0 as

f (x0 + h) − f (x0 )
f 0 (x0 ) = lim .
h→0 h
We interpret this as a rate of change: a change in x causes a corresponding change in
y = f (x). Suppose now we consider a function of two variables f (x, y). What do we mean
by a derivative in this case? We now have two variables which can be changed independently
of each other. What do we mean by the rate of change in this situation, if there are multiple
ways of causing such a change?
To get around this ambiguity, we introduce a new type of derivative, called a partial
derivative, according to the definition below.

Definition 37. Let f (x, y) be a function of two variables with domain D. Let (x0 , y0 ) in D.
We define the partial derivative of f with respect to x at (x0 , y0 ) as

∂f f (x0 + h, y0 ) − f (x0 , y0 )
fx (x0 , y0 ) = = lim .
∂x h→0 h
In the same way, we define the partial derivative of f with respect to y at (x0 , y0 ) as

∂f f (x0 , y0 + h) − f (x0 , y0 )
fy (x0 , y0 ) = = lim .
∂y h→0 h

51
 The idea behind the definition is simple: we are computing the rate of change of the
function with only one variable allowed to change. All others must remain fixed.
Example 52. Compute the partial derivatives fx (0, 0) and fy (0, 0) for the function

f (x, y) = 2x2 y + x cos(y).

Solution. To compute fx , we apply the definition:

f (h, 0) − f (0, 0)
fx (0, 0) = lim
h→0 h
h−0
=
h
= 1.
To compute the partial derivative with respect to y, we will use the following trick: in
the definition of fy , the variable x is fixed to x0 . Thus, we may think of fy (x0 , y0 ) as the
derivative of the function f (y) = 2x20 y + x0 cos y at y0 . The partial derivative is then

fy (x0 , y0 ) = f 0 (y0 )
= 2x20 − x0 sin y0 .

Using (x0 , y0 ) = (0, 0), we get

fy (0, 0) = 0.

Partial Derivatives as Functions

In applications of derivatives (whether ordinary or partial derivatives), it is often useful to
think of them as a new function which is derived from another function. For example, in
Example 52, we could have left the variables x and y and completed the computation, giving
us the new function
f (x + h, y) − f (x, y) 2(x + h)2 y + (x + h) cos y − (2x2 y + x cos y)
lim = lim
h→0 h h→0 h
2(2xh + h2 )y + h cos y
= lim
h→0 h
= lim (4x + h)y + cos y
h→0
= 4xy + cos y.
This gives us a new function, which we refer to as simply the partial derivative with respect
to x, and we denote as
∂f
or fx (x, y).
∂x
An analogous statement can be made for partial derivatives with respect to any variable.

52
Example 53. Compute the following partial derivatives:
∂ 2 2 2
1. (x y z )
∂x
Solution. Using the trick from the second part of the previous example, we have
∂ 2 2 2
(x y z ) = 2xy 2 z 2 .
∂x

∂
2. (xw2 + 2z 3 sin y)
∂y

∂
Solution. (xw2 + 2z 3 sin y) = 2z 3 cos y
∂y

Implicit Partial Differentiation

Just as we had implicit differentiation for functions of a single variable, we can generalize
this to functions of several variables in the form of implicit partial derivatives. To see how
this works, consider the following problem: suppose we wanted to compute √ the√rate of the
change in the y-direction of the sphere x2 + y 2 + z 2 = 1 at the point (0, 2/2, 2/2). One
way to proceed is to solve for z (the height of the sphere):
p
z = 1 − x2 − y 2 .

We can then compute the partial derivative with respect to y, and continue as above.
Alternatively, we can proceed implicitly in the following way: if we assume that z is a
function of x and y, we can then compute the partial derivative implicitly using the chain
rule:
∂ ∂
x2 + y 2 + z 2 =


(1).
∂y ∂y
This expression becomes
∂z
2y + 2z = 0,
∂y
or
∂z y
=− .
∂y z
√ √
At the point (0, 2/2, 2/2), this becomes
∂z
= −1.
∂y
Example 54. Suppose z satisfies z 2 = x2 y 2 . Find the rate of change of z in the x-direction
at (1, 1, 1).

53
Solution. As before, we differentiate implicitly:
∂ 2 ∂ 2 2
(z ) = (x y ),
∂x ∂x
which becomes
∂z
2z = 2xy 2 .
∂x
Thus
∂z xy 2
= .
∂x z
At (1, 1, 1), this becomes
∂z
= 1.
∂x

Partial Derivatives and Continuity

When learning about continuity and functions of a single variable, one of the most important
facts you learned is that if a function is differentiable, then it is automatically continuous.
We will see in the following example that this is not true for functions of more than one
variable.
Consider the function from Example 49:
(
xy
x2 +y 2
if (x, y) 6= (0, 0),
f (x, y) =
0 if (x, y) = (0, 0).

As we saw before, this function is not continuous. However, its partial derivatives exist:

f (0 + h, 0) − f (0, 0)
fx (0, 0) = lim
h→0 h
0
= lim
h→0 h
= 0.

f (0, 0 + h) − f (0, 0)
fy (0, 0) = lim
h→0 h
0
= lim
h→0 h
= 0.
Thus, functions of more than one variable do behave differently than functions of single
variables, at least in one way.

54
Higher-Order Partial Derivatives
Since we defined (partial) derivatives for functions of several variables, you may be wondering
if there is an equivalent notion of second or third derivatives for these functions. There is
such a notion, and it works in the obvious way that you would expect.
For example, consider a function f (x, y) with partial derivative fx (x, y). We may again
differentiate with respect to x by using the definition:

fx (x + h, y) − fx (x, y)
fxx (x, y) = lim ,
h→0 h
if this limit exists. We could, however, differentiate fx with respect to y. This is done in the
obvious way:
fx (x, y + h) − fx (x, y)
fxy (x, y) = lim .
h→0 h
Note, however, that in the other notation, there is a slight difference in how we keep track
of the partial derivatives
∂ 2f ∂ 2f
fxy = , fyx = .
∂y∂x ∂x∂y
In the case where we differentiate twice with respect to the same variable, we use the notation

∂ 2f
fxx = .
∂x2
Example 55. Compute fxx , fyy , fxy and fyx for the following functions:

1. f (x, y) = x2 y 2

Solution. fxx = 2y 2 , fyy = 2x2 , fxy = fyx = 4xy

2. f (x, y, z) = 2x cos y sin z − yez sin x

Solution. fxx = yez sin x, fyy = −2x cos y sin z, and

fxy = fyx = −2 sin y sin z − ez cos x

In the examples above, you should obtain that in both cases, we have fxy = fyx . You
may be wondering if this is always the case. As the next theorem indicates, under many
(though not all!) circumstances, this is the case.

Theorem 15. Let f be a function of two variables such that fxy and fyx are continuous on
some open disc. Then fxy = fyx is continuous on that disc.

55
13.4 Differentiability, Differentials, and Local Linear-
ity
In this section, we will discuss some more advanced aspects of partial derivatives.

Differentiability
In calculus I, we defined a function f (x) to be differentiable at a point x0 if the derivative
f 0 (x) exists at x0 . From this simple definition, we were able to deduce the following facts
about f (x):

• f (x) must be continuous at x0 ;

• f (x) cannot have a vertical tangent line at x0 ;

• f (x) can be approximated by a linear function for values of x near x0 .

However, recall that the function

(
xy
x2 +y 2
if (x, y) 6= (0, 0),
f (x, y) =
0 if (x, y) = (0, 0).

from Example 49 is not continuous at (0, 0), yet its partial derivatives exist at (0, 0). Thus,
for functions of several variables, the existence of derivatives is not enough to guarantee
the above three statements hold. Thus, we must modify what we mean by a “differentiable
function” when discussing functions of several variables.
To define differentiability in this context, we first need the following definition:

Definition 38. For a function f (x, y), we define the increment ∆f at (x0 , y0 ) is

∆f = f (x0 + ∆x, y0 + ∆y) − f (x0 , y0 ).

With this, we can now state a proper definition of differentiability:

Definition 39. A function f (x, y) is said to be differentiable at (x0 , y0 ) if both fx (x0 , y0 )

and fy (x0 , y0 ) exist, and

∆f − fx (x0 , y0 )∆x − fy (x0 , y0 )∆y

lim p = 0.
(∆x,∆y)→(0,0) (∆x)2 + (∆y)2

The definition for functions of more than two variables is analogous. Using this definition,
we may prove the following theorem about differentiable functions:

Theorem 16. If a function f (x, y) is differentiable at (x0 , y0 ), then it is continuous at

(x0 , y0 ).

56
Note that in this theorem, it is necessary that we use the term “differentiable” in the sense
of Definition 39.
To apply this theorem, one would need to show that, for a particular function, Definition
39 is satisfied, which could be a difficult computation. Fortunately, the following theorem
gives us a way around this.
Theorem 17. If fx (x0 , y0 ) and fy (x0 , y0 ) exist and are continuous at (x0 , y0 ), then f is
differentiable at (x0 , y0 ).

Differentials
In calculus I, for a differentiable function f (x), we defined its differential df at x0 as
df = f 0 (x0 )dx.
Here, the term dx should be interpreted as ∆x = x − x0 . We can generalize this concept to
functions of several variables as
df = fx (x0 , y0 )dx + fy (x0 , y0 )dy,
and similarly for three or more variables. This expression is known as the total differential
of f at (x0 , y0 ).
As was discussed in calculus I, differentials are used to approximate errors in experimental
results due to inaccuracies in measurements. For quantities which depend on several different
measurements which can be expressed in the form f (x, y), the error ∆f is approximated by
∆f ≈ df = fx (x0 , y0 )dx + fy (x0 , y0 )dy,
where we assume dx = ∆x is the error in the measurement of x, and dy = ∆y is the error
in the measurement of y. In the following example, we will do an error analysis for a simple
experiment.
Example 56. Suppose you want to measure the average velocity of a ball rolling across a
track of length ` = 2 m ±.1 m. Suppose you measure the time with a stopwatch to be t = 5s
±.5s. What is the maximum possible error in your computation of the ball’s velocity?
Solution. Recall that the average velocity of an object travelling a distance ` for a time t is
`
v(`, t) = .
t
The maximum error in our measurement is given by the total differential:
∆v = dv
= v` (`0 , t0 )d` + vt (`0 , t0 )dt
1 `0
= ∆` − 2 ∆t
t0 t0
0.1 m (2 m)(0.5 s)
= −
5s 25 s2
1 m
= .
25 s
57
Local Linearity
The final topic in this section relates to linear approximations. Recall from calculus I that
if f (x) is differentiable at x0 , then we may approximate it by a line of the form
f (x) ≈ f (x0 ) + f 0 (x0 )(x − x0 ).
That is, we approximate f (x) for values of x near x0 by the tangent line through x0 . As you
might expect, we can do this for functions of two or more variables, as well. We will focus
our attention on the case where f is a function of two variables. In this case, we approximate
by a plane.
Recall from the discussion of differentials that
∆f ≈ df = fx (x0 , y0 )dx + fy (x0 , y0 )dy = fx (x0 , y0 )∆x + fy (x0 , y0 )∆y.
Expanding the ∆f , ∆x and ∆y terms, we obtain
f (x, y) − f (x0 , y0 ) ≈ fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ),
which becomes
f (x, y) ≈ f (x0 , y0 ) + fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ).
If write z = f (x, y), then this becomes
z ≈ z0 + fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ),
which is the equation of a plane through (x0 , y0 , f (x0 , y0 )). Note that the expression on the
right-hand side is linear; that is, the variables x and y appear only to the first power, and
there is no product involving x and y. This leads us to the following definition:
Definition 40. Let f (x, y) be differentiable at (x0 , y0 ). Then the local linear approximation
to f at (x0 , y0 ) is
L(x, y) = f (x0 , y0 ) + fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ).
√
Example 57. Use a local linear approximation to approximate 25.1402.
Solution. To begin, we first observe that 25.1402 = (3.01)2 + (4.01)2 . With this in mind,
define p
f (x, y) = x2 + y 2 .
The local linear approximation to f at (3, 4) is
L(x, y) = f (3, 4) + fx (3, 4)(x − 3) + fy (3, 4)(y − 4)
   
3 4
=5+ (x − 3) + (y − 4).
5 5

58
If we evaluate this at (3.01, 4.01), we have

L(3.01, 4.01) = 5 + (0.6)(0.01) + (0.8)(0.01) = 5.014.

√
Thus 25.1402 ≈ 5.014. Compare this with the actual value of
√
25.1402 = 5.0140003988831113773697106012037...

13.5 The Chain Rule for Functions of Several Variables

Recall that when differentiating compositions of functions of a single variable, we have the
following formula, known as the chain rule:
d
(f ◦ g(x)) = f 0 (g(x))g 0 (x).
dx
When considering functions of several variables, there are many ways in which we may
compose functions together. To determine how the chain rule will work, we consider each of
these cases separately. In all cases, assume that z = f (x, y).

Case 1: x, y are parametric functions.

Assume that x = x(t) and y = y(t). In this case, the function we wish to differentiate is of
the form
z = f (x(t), y(t)),
which is a function of only the variable t. For such functions, the chain rule takes the form
dz ∂z dx ∂z dy
= + .
dt ∂x dt ∂y dt
Example 58. Differentiate the following functions.

1. z = x2 + y 2 , x(t) = t2 + t, y(t) = et .

Solution. Applying the chain rule, we obtain

dz dx dy
= 2x(t) + 2y(t)
dt dt dt
= 2(t + t)(2t + 1) + 2e2t .
2

2. z = sin(xy), x(t) = sin t, y(t) = cos t.

59
 Solution. Applying the chain rule, we obtain
dz dx dy
= y(t) cos(x(t)y(t)) + x(t) cos(x(t)y(t))
dt dt dt
2 2
= cos (t) cos(sin(t) cos(t)) − sin (t) cos(sin(t) cos(t))
= cos(2t) cos(sin(t) cos(t)).

Case 2: x and y are functions of several variables

Let us now assume we have a function z = f (x, y), where x = x(u, v) and y = y(u, v). In
this case, the composite function z = f (x(u, v), y(u, v)) depends on the two variables u and
v, and so it is appropriate to compute partial derivatives. In such a situation, the chain rule
takes the form
∂z ∂z ∂x ∂z ∂y
= + ,
∂u ∂x ∂u ∂y ∂u
and
∂z ∂z ∂x ∂z ∂y
= + .
∂v ∂x ∂v ∂y ∂v
Example 59. For each of the functions below, compute fu and fv .
1. f (x, y) = ex−y , x(u, v) = u + v, y(u, v) = u − v.

Solution. Applying the chain rule for derivatives, we obtain

∂f
= ex(u,v)−y(u,v) − ex(u,v)−y(u,v)
∂u
= 0,
∂f
= ex(u,v)−y(u,v) (1) − ex(u,v)−y(u,v) (−1)
∂v
= 2e2v .

2. f (x, y) = x2 y 2 + cos y, x(u, v) = u2 + v 2 , y(u, v) = eu .

Solution. Applying the chain rule for partial derivatives, we obtain

∂f
= (2x(u, v)y 2 (u, v))(2u) + (2x2 (u, v)y(u, v) − sin y(u, v))eu
∂u
= 4u(u2 + v 2 )e2u + 2(u2 + v 2 )2 − sin eu e2u ,
 

∂f
= 2x(u, v)y 2 (u, v)2v
∂v
= 4v(u2 + v 2 )e2u .

60
Implicit Differentiation
As a special application of the chain rule, consider an implicit function of the form

f (x, y) = c

for some function f (x, y). Suppose we want to compute the derivative y 0 . If we assume that
y is the dependent variable, then we can apply the chain rule to say
df ∂f dx ∂f dy
= + .
dx ∂x dx ∂y dx
Since
dx
= 1,
dx
this reduces to
df ∂f ∂f dy
= + .
dx ∂x ∂y dx
Since f (x, y) = c, we should have
df
= 0,
dx
so that
∂f ∂f dy
0= + .
∂x ∂y dx
Solving for the desired quantity, we obtain
∂f
dy ∂x
= − ∂f .
dx ∂y

Thus, we have arrived at the following theorem:

Theorem 18. Suppose that y is defined implicitly by an equation of the form f (x, y) = c.
Then
∂f
dy ∂x
= − ∂f .
dx ∂y

Example 60. Suppose y is an implicitly defined function defined by the equation

x3 + y 3 − xy = 1.

Compute y 0 .
Solution. Based on the above computation, we have
dy 3x2 − y
=− 2 .
dx 3y − x

61
13.6 Directional Derivatives and Gradients
Consider the function f (x, y) = cos(x) cos(y), depicted in the graph below:

Recall Definition 37 for partial derivatives:

f (x0 + h, y0 ) − f (x0 , y0 )
fx (x0 , y0 ) = lim ,
h→0 h
f (x0 , y0 + h) − f (x0 , y0 )
fy (x0 , y0 ) = lim .
h→0 h
In the case of the partial derivative with respect to y, we stated previously that we can
interpret this derivative as the derivative obtained by fixing a value of x0 , and then consid-
ering the function as a function of only the variable y. For example, if we want to compute
fy (0, π/2), we first fix x = 0, which gives us the black curve in the following diagram:

62
We then compute the (ordinary) derivative of the function f (0, y) at y = π/2, which gives
us the slope of the line depicted in the following diagram in green:

Thus, we may think of the partial derivative with respect to y at (0, π/2) as representing
the slope of the surface z = f (x, y) at (0, π/2) in the direction parallel to the y-axis. An
analogous statement can be made regarding fx (0, π/2).
This leads us to an obvious question: what if we want to consider the slope of our surface
along some direction that is not parallel to the x- or y-axes? The answer to this comes in
the following definition:

Definition 41. Let u = hu1 , u2 i be a unit vector. The directional derivative of f in the
direction of u at (x0 , y0 ), denoted by Du f (x0 , y0 ), is defined by

d
Du f (x0 , y0 ) = [f (x0 + su1 , y0 + su2 )]s=0 ,
ds
if this derivative exists.

Remark 6. You should check that

Di f (x0 , y0 ) = fx (x0 , y0 )
Dj f (x0 , y0 ) = fy (x0 , y0 ).

In the case of a function of three variables f (x, y, z), we would also have

Dk f (x0 , y0 , z0 ) = fz (x0 , y0 , z0 ).

Using the function f√(x, y) √= cos(x) cos(y) as an example, the directional derivative
Du f (0, π/2) for u = (1/ 2, 1/ 2) tells us the slope of the tangent line of the curve repre-
sented by the black line in the following diagram:

63
Example 61. Compute the directional derivatives of the following functions at the indicated
points in the indicated directions:

1. f (x, y) = xy at (1, 1) for u = h √12 , √12 i

Solution. Using the definition, we have

  
d s s
Du f (1, 1) = f 1 + √ ,1 + √
ds 2 2 s=0
" 2 #
d s
= 1+ √
ds 2
  s=0
2 s
= √ 1+ √
2 2 s=0
2
=√
2

√
3
2. f (x, y, z) = cos(x + y + z) at (0, π/4, π/4) for u = h 21 , 2
, 0i

Solution. Using the definition, we have

d h π s  √ i
Du f (0, π/4, π/4) = cos + 1+ 3
ds √ 2 2 s=0
1+ 3 π s  √ 
=− sin + 1+ 3
2√ 2 2 s=0
1+ 3
=− .
2

64
The Gradient Vector
As is always the case in calculus, we would like to find a shortcut to our lengthy computations.
In the case of directional derivatives, we have a simple shortcut. To apply this shortcut, we
must first introduce the following special vector.

Definition 42. If f (x, y) is a function of two variables, then the gradient of f is the vector
denoted by ∇f (x, y), given by

∇f (x, y) = hfx (x, y), fy (x, y)i.

The symbol ∇ is referred to as either “del” or “nabla”. When there is no confusion, we

will simply write ∇f to denote the gradient of f . With this definition, we have the following
shortcut to computing directional derivatives:

Theorem 19. Let f (x, y) be differentiable at (x0 , y0 ), and let u = u1 i + u2 j. Then the
directional derivative Du f (x0 , y0 ) exists and is given by

Du f (x0 , y0 ) = ∇f (x0 , y0 ) · u.

Example 62. Repeat the computations of Example 61 using the gradient vector.

Solution. For f (x, y) = xy and u = h √12 , √12 i at (1, 1), we have

∇f (x, y) = hy, xi,

so that
1 1 2
Du f (1, 1) = h1, 1i · h √ , √ i = √ .
2 2 2
√
3
In the second example, we had f (x, y, z) = cos(x + y + z), u = h 12 , 2
, 0i. A simple
computation will show that

∇f (x, y, z) = h− sin(x + y + z), − sin(x + y + z), − sin(x + y + z)i

= − sin(x + y + z)h1, 1, 1i.

Thus, we have
* √ + √
1 3 1+ 3
Du f (0, π/4, π/4) = h−1, −1, −1i · , ,0 = − .
2 2 2

Note that both of these computations agree with the results obtained above.
x+y
Example 63. Compute q the directional derivative of f (x, y) = e at (ln 2, ln 2) in the
1 2
direction of u = h− √3 , 3 i.

65
Solution. We first compute the gradient:

∇f (x, y) = ex+y h1, 1i.

Using Theorem 19, we get

* r +
1 2
Du f (ln 2, ln 2) = e2 ln 2 h1, 1i · − √ ,
3 3
* r +
1 2
= h4, 4i · − √ ,
3 3
√
4 2−4
= √ .
3

Properties of the Gradient

The gradient vector has several useful properties. The first useful property is useful for
problems involving maximization and minimization.

Theorem 20. Let f (x, y) be a function of two variables. If f is differentiable at (x0 , y0 ),

then the following statements hold:

• If ∇f (x0 , y0 ) = 0, then all directional derivatives of f are zero at (x0 , y0 ).

• If ∇f (x0 , y0 ) 6= 0, then the directional derivative of f has its largest value in the direc-
tion of ∇f (x0 , y0 ). Moreover, the value of the directional derivative in this direction is
k∇f (x0 , y0 )k.

• If ∇f (x0 , y0 ) 6= 0, then the directional derivative of f has its smallest value in the
direction opposite to ∇f (x0 , y0 ). Moreover, the value of the directional derivative in
this direction is −k∇f (x0 , y0 )k.

Example 64. Let z = x2 + y 2 . Find the direction which gives you the maximum value of
the directional derivative at (1, 2), and compute the directional derivative in this direction.

Solution. By Theorem 20, the first step is to compute the gradient vector:

∇f (x, y) = h2x, 2yi.

At (1, 2), this becomes

∇f (1, 2) = h2, 4i.

66
This gives us the direction of the maximum value of the directional derivative; however, it
is not a unit vector. To make it into a unit vector, we normalize:

h2, 4i
u=
kh2, 4ik
1
= √ h2, 4i
20
1
= √ h1, 2i.
5
In this direction, the directional derivative will be
√
Du f (1, 2) = k∇f (1, 2)k = 20.

The second useful property of gradients is that they give us a simple way of finding a
vector normal to certain types of curves.

Theorem 21. Assume that f (x, y) has continuous first-order partial derivatives in an open
disc centered at (x0 , y0 ), and that ∇f (x0 , y0 ) 6= 0. Then ∇f (x0 , y0 ) is normal to the level
curve through f (x0 , y0 ).

13.7 Tangent Planes and Normal Vectors

An important problem in calculus is the problem of finding planes that are tangent to a
surface z = f (x, y). This is analogous to the problem of finding tangent lines to curves of
the form y = f (x). In that situation, we saw that the tangent line to y = f (x) at x0 has
equation
y = f (x0 ) + f 0 (x0 )(x − x0 ).
To find the corresponding expression in three dimensions, we make the following observation:
in Theorem 21, we claimed that if C is a level curve for the function z = f (x, y), then the
gradient ∇f defines a vector which is normal to C at any point in C (assuming that ∇f 6= 0).
By a similar argument, one can show the following more general result:

Theorem 22. Let S be a surface defined by an equation of the form F (x, y, z) = C. Then
the equation for the tangent plane at (x0 , y0 , z0 ) is given by

Fx (x0 , y0 , z0 )(x − x0 ) + Fy (x0 , y0 , z0 )(y − y0 ) + Fz (x0 , y0 , z0 )(z − z0 ) = 0.

Moreover, if ∇F (x0 , y0 , z0 ) 6= 0, then nablaF (x0 , y0 , z0 ) is normal to S at (x0 , y0 , z0 ). In

the case where S is defined by a function of the form z = f (x, y), we obtain the simplified
expression
z = f (x0 , y0 ) + fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ).

67
 2 2 2
Example√65. Find
√ the √ equation of the tangent plane to the sphere x + y + z = 1 at the
point (1/ 3, 1/ 3, 1/ 3).
Solution. Using Theorem 22, we see that we must first compute the gradient of the function

F (x, y, z) = x2 + y 2 + z 2 .

This is given by
√ √ √ 2
∇F (1/ 3, 1/ 3, 1/ 3) = √ h1, 1, 1i.
3
With this, the equation becomes
     
2 1 2 1 2 1
√ x− √ +√ y−√ +√ z−√ = 0.
3 3 3 3 3 3
If we like, we may simplify this to
     
1 1 1
x− √ + y−√ + z−√ = 0.
3 3 3

Example 66. Let S be the surface defined by the equation z = x2 + y 2 .

1. Find the equation of the plane tangent to S at (1, 1, 2).

Solution. By Theorem 22, the equation for the tangent plane can be written as
z = f (1, 1) + fx (1, 1)(x − 1) + fy (1, 1)(y − 1)
= 2 + 2(x − 1) + 2(y − 1).

2. Find the parametric equations for the line that is normal to S at (1, 1, 2).

Solution. Recall that the parametric equations for a line can be written in vector form
as r(t) = r0 + tv, where r0 is a point on the line (in vector form), and v is a vector
pointing in the direction of the line. For r0 , we may use h1, 1, 2i, since our line should
contain the point to which it is normal.
For the direction vector, we first rewrite our expression in the form z − x2 − y 2 = 0,
so that F (x, y, z) = 0. Using the second part of Theorem 22, the normal vector at
(1, 1, 2) is given by
∇F (1, 1, 2) = h−2, −2, 1i.
Thus, we obtain
r(t) = h1, 1, 2i + th−2, −2, 1i
= h1 − 2t, 1 − 2t, 2 + ti.

68
 Thus, the desired parametric equations are

x(t) = 1 − 2t, y(t) = 1 − 2t, z(t) = 2 + t.

3. Find the angle between the tangent plane and the xz-plane.

Solution. Recall that if two planes have normal vectors v1 and v2 , then Theorem 1
tells us that the angle θ between the two vectors satisfies
v1 · v2
cos θ =
kv1 kkv2 k

In part 2 above, we saw that the normal vector to the tangent plane is v1 = h−2, −2, 1i.
For the xz-plane, the normal vector is v2 = h0, 1, 0i. Thus, we obtain
2
cos θ = − ,
3
so that θ ≈ 131◦ .

13.8 Extreme Values of Functions of Two Variables

One of the major uses of calculus is to find maximum or minimum values of functions. Recall
that in calculus I, we considered the problem of finding the absolute maximum and minimum
values of functions defined on closed, bounded intervals. In this section, we will generalize
those ideas to the case of functions of two variables.
To begin, we should first state precisely what we are looking for. This is the content of
the following two definitions:

Definition 43. A function f of two variables with domain D is said to have a relative
maximum at a point (x0 , y0 ) if there is a disc centered at (x0 , y0 ) contained in D such that
f (x0 , y0 ) ≥ f (x, y) for all points (x, y) that lie inside the disc, and f is said to have an absolute
maximum at (x0 , y0 ) if f (x0 , y0 ) ≥ f (x, y) for all points (x, y) in D.

Definition 44. A function f of two variables with domain D is said to have a relative
minimum at a point (x0 , y0 ) if there is a disc centered at (x0 , y0 ) contained in D such that
f (x0 , y0 ) ≤ f (x, y) for all points (x, y) that lie inside the disc, and f is said to have an absolute
minimum at (x0 , y0 ) if f (x0 , y0 ) ≤ f (x, y) for all points (x, y) in D.

Points that are either maxima or minima of functions are collectively known as extrema.
Next, we generalize the definition of a closed and bounded interval. This will be important
when we discuss absolute extrema. Recall that a set is closed if it contains its boundary. For
the term bounded, we will use the following definition:

69
Definition 45. A set of points in 2-space is called bounded if the entire set can be contained
within some rectangle. Otherwise, it is called unbounded.

Example 67. The set D, shown in blue in the diagram,

is a bounded set because it is contained inside the black rectangle in the diagram.

The reason for discussing functions defined on closed and bounded sets is so that we may
apply the following theorem:

Theorem 23 (The Extreme Value Theorem). If f (x, y) is continuous on a closed and

bounded set R, then f has both an absolute maximum and an absolute minimum in R.

This theorem is important, but not very useful; it does not help you in finding the
maximum or minimum. It merely tells you whether you can expect to succeed in finding
the extrema. To actually find them, we make the following obvious observation: an extreme
value of a function f (x, y) defined on a closed and bounded set D will occur either on the
boundary of D, or inside the boundary of D. We refer to anything within the boundary of
a set as the interior of the set.
To check the interior of D for extrema, we first need some familiar definitions:

Definition 46. A point (x0 , y0 ) in the domain D of a function f (x, y) is called a critical
point of f if either both fx (x0 , y0 ) = 0 and fy (x0 , y0 ) = 0 or if one or both partial derivatives
do not exist at (x0 , y0 ).

Example 68. Find all critical points for the following functions:

1. f (x, y) = x2 + y 2

70
 Solution. We first compute the partial derivatives:

fx (x, y) = 2x,
fy (x, y) = 2y.

From these expressions, we see that fx (x, y) = 0 when x = 0, and fy (x, y) = 0 when
y = 0. Thus, the only critical point is (0, 0).

2. f (x, y) = cos(x + y)

Solution. The partial derivatives of this function are

fx (x, y) = fy (x, y) = − sin(x + y).

Observe that
sin(x + y) = 0 whenever x + y = nπ.
Thus, any point of the form (x, nπ − x) is a critical point of this function.

Critical points are important because of the following theorem:

Theorem 24. If a function f of two variables has an absolute extremum (either an absolute
maximum or an absolute minimum) at an interior point of its domain, then this extremum
occurs at a critical point.
More generally, we can say the following:
Theorem 25. If f (x, y) has a relative extremum at a point (x0 , y0 ), and if the first-order
partial derivatives of f exist at this point, then fx (x0 , y0 ) = 0 and fy (x0 , y0 ) = 0.
These results suggest that we can proceed as we did in calculus I: look for critical points in
the interior of the domain, and compare the value of the function at critical points with the
values on the edges of D.
But beware! As you might remember from calculus I, the derivative of a function can be
zero at a point which is not a maximum or a minimum, such as in the function f (x) = x3 :

71
For this function, f 0 (0) = 0, but x = 0 is not a local maximum or minimum. An analogous
situation that can occur for functions of two variables is something known as a saddle point,
shown in the following diagram:

This is the function f (x, y) = y 2 − x2 . This function satisfies fx (0, 0) = 0 and fy (0, 0) = 0,
but (0, 0) is not a maximum or minimum.
This leads us to an obvious question: once we find critical points, how do we know if
they are a maximum, or minimum, or neither? To answer this, we may apply the following
theorem:

Theorem 26 (Second Derivative Test). Let f be a function of two variables with continuous
second-order partial derivatives in some disc centered at a critical point (x0 , y0 ), and let
2
D = fxx (x0 , y0 )fyy (x0 , y0 ) − fxy (x0 , y0 ).

1. If D > 0 and fxx (x0 , y0 ) > 0, then f has a relative minimum at (x0 , y0 ).

2. If D > 0 and fxx (x0 , y0 ) < 0, then f has a relative maximum at (x0 , y0 ).

3. If D < 0, then f has a saddle point at (x0 , y0 ).

4. If D = 0, then no conclusion can be drawn.

Example 69. Find the absolute extrema for the function f (x, y) = x2 + y 2 defined in the
closed unit disc {(x, y) : x2 + y 2 ≤ 1}.

Solution. Based on the above discussion, we should first find the value of f at any critical
points. In Example 68 part (1), we saw that the only critical point of this function in the
interior of the closed unit disc is (0, 0) . Evaluating f at this point, we see that f (0, 0) = 0.
The next step is to find the maximum and minimum of f on the boundary x2 + y 2 = 1.
Look very carefully at the equation that defines the boundary: it implies that for any point
on the boundary, we will have f (x, y) = 1. Thus, the function is constant on the boundary
of the closed unit disc, with value 1. Thus, we see that the function attains an absolute
maximum value of 1 at any point on this boundary, and an absolute minimum at (0, 0).

72
Example 70. Find the maximum and minimum values of the function

f (x, y) = 2xy − 6y − 4x

defined inside the closed square with vertices (0, 0), (0, 4), (4, 0), (4, 4).

Solution. For reference, let us sketch the domain D:

Next, we look for critical points in the interior of D:

fx (x, y) = 2y − 4 = 0 when y = 2

and
fy (x, y) = 2x − 6 = 0 when x = 3.
So the only critical point is (3, 2), which is in the interior of D. We note that f (3, 2) = −12.
Next, we check the boundary of D. The boundary consists of the four lines determined
by x = 0, y = 0, x = 4 and y = 4. We must check each one separately.
Case 1: x = 0 In this case, the function takes the form f (y) = −6y for 0 ≤ y ≤ 4. This is a
function of one variable, so we may use the techniques from calculus 1 to see that there are
no critical points in the interval (0, 4), and f (0) = 0, f (4) = −24.
Case 2: y = 0 In this case, the function takes the form f (x) = −4x for 0 ≤ x ≤ 4. This is
a function of one variable, so we may use the techniques from calculus 1 to see that there
are no critical points in the interval (0, 4), and f (0) = 0, f (4) = −16.
Case 3: x = 4 In this case, the function takes the form f (y) = 2y − 16 for 0 ≤ y ≤ 4. This
is a function of one variable, so we may use the techniques from calculus 1 to see that there
are no critical points in the interval (0, 4), and f (0) = −16, f (4) = −8.

73
Case 4: y = 4 In this case, the function takes the form f (x) = 4x − 24 for 0 ≤ x ≤ 4. This
is a function of one variable, so we may use the techniques from calculus 1 to see that there
are no critical points in the interval (0, 4), and f (0) = −24, f (4) = −8.
Based on these computations, we see that the maximum value of f in D is f (0, 0) = 0,
and the minimum is f (0, 4) = −24.

74
Chapter 14

Multiple Integrals

In this chapter, we consider the problem of integrating functions of several variables. As

we will see, things are significantly more complicated for functions of several variables; even
more so than for vector functions. We begin by discussing what we mean by “integral” for
functions of several variables.

14.1 Double Integrals

In calculus I, we considered the area problem: given a function f (x), compute the area under
the curve defined by y = f (x), between x = a and x = b, and the x-axis. For functions
of several variables, the analogous problem is the volume problem: given a function f (x, y)
defined on some subset D of the xy-plane, what is the volume of the solid region between the
xy-plane and the surface z = f (x, y) over the region D? Here, we briefly sketch the process
of computing this volume:

• Divide the region D into rectangular pieces, each of which has area ∆Ak .

• In each rectangle, choose a coordinate (x∗k , yk∗ ) and compute zk = f (x∗k , yk∗ ).

• The rectangle containing the point (x∗k , yk∗ ) and the height zk determine a rectangular
parallelepiped with volume Vk = f (x∗k , yk∗ )∆Ak .

• Approximate the volume of the solid region by the sum of the volumes of the paral-
lelepipeds:
XN
V ≈ f (x∗k , yk∗ )∆Ak .
k=1

• To improve the approximation, we let the number of rectangles N go to infinity:

N
X
V = lim f (x∗k , yk∗ )∆Ak .
N →∞
k=1

75
The expression above is nothing more than a two-dimensional Riemann sum, and suggests
the following definition:
Definition 47. Let f (x, y) be a continuous and non-negative function of two variables, and
let R be a region in the xy-plane. Then the volume under the surface z = f (x, y) is
N
X
V = lim f (x∗k , yk∗ )∆Ak ,
N →∞
k=1

if this limit exists. In this case, we denote the result by

N
X ZZ
lim f (x∗k , yk∗ )∆Ak = f (x, y) dA.
N →∞ R
k=1

The notation on the right is read “the double integral of f (x, y) over the region R”.
As we have seen repeatedly in calculus, it is tedious (and sometimes, quite difficult) to
perform computations using the definition directly. A more efficient method is to use the
method of iterated integrals, which we describe below.

Iterated Integrals
Consider a function f (x, y). If we hold the variable y fixed, then we can think of f (x, y) as
a function of only one variable. In this case, we can compute its definite integral:
Z b x=b
f (x, y) dx = F (x, y) .
a x=a

The end result of this computation will clearly be a function of y. Since this is the case, we
can compute its definite integral:
Z dZ b Z d x=b

f (x, y) dxdy = F (x, y) dy.
c a c x=a

The expression on the left in this equation is called an iterated integral; we integrate twice,
first in one variable, then in the other. The expression on the right side explains how to
actually compute it.
Example 71. Compute the following iterated integrals:
Z 1Z 2
1. xy dxdy
0 1

Solution. As explained above, we first integrate in the variable x:

Z 2
x2
 
x=2 4 1 3y
xy dx = ·y = − y= .
1 2 x=1 2 2 2

76
 We then integrate in the y variable:
Z 1Z 2 Z 1
3y 3
xy dxdy = dy = .
0 1 0 2 4

Z 1 Z π/2
2. y 2 cos(x) dxdy
0 0

Solution. Based on the above discussion, we first integrate in x:

Z π/2  π  
y 2 cos(x) dx = y 2 sin − sin 0 = y 2 .
0 2
We then integrate in y:
Z 1 Z π/2 Z 1
2 1
y cos(x) dxdy = y 2 dy = .
0 0 0 3

You might be wondering if you can integrate in the opposite order. The answer is, yes
you can! And, it is done in the analogous way, as we see in the following examples:

Example 72. Compute the following iterated integrals (Pay attention to the order!):
Z 2Z 1
1. xy dydx
1 0

Solution. This time, begin with the integral in y:

Z 1
x
xy dy =
0 2
Then integrate in x: Z 2 Z 1 Z 2
x 3
xy dydx = dx = .
1 0 1 2 4

Z π/2 Z 1
2. y 2 cos(x) dydx
0 0

77
 Solution. Again, begin with the y integral:
Z 1
cos(x)
y 2 cos(x) dy =
0 3
Then integrate in x:
Z π/2 Z 1 Z π/2
2 cos(x) 1
y cos(x) dydx = dx = .
0 0 0 3 3

You may have noticed that the integrals give the same answer, regardless of the order in
which you integrate. The following theorem indicates that this is not merely a coincidence.
Theorem 27 (Fubini’s Theorem). Let f (x, y) be continuous in the rectangular region R
defined by a ≤ x ≤ b and c ≤ y ≤ d. Then
ZZ Z bZ d Z dZ b
f (x, y) dA = f (x, y) dxdy = f (x, y) dydx.
R a c c a

Thus, when computing a double integral over a rectangular region R, Fubini’s theorem tells
us that we can compute the integral using any of the iterated integrals above.

14.2 Double Integrals over Non-Rectangular Regions

In the previous section, we introduced the concept of a double integral, and we discussed
how they can be computed in the simple case where we integrate over a rectangular region
R. We now want to consider the possibility that R may not be a simple rectangle. To make
the problem more manageable, we will consider only two types of regions:
Type I Regions: A type I region R is a subset of the xy-plane bounded on the left by the
line x = a, on the right by the line x = b, on the top by a function of the form y = h(x),
and on the bottom by a function of the form y = g(x):

78
Type II Regions: A type II region R is a subset of the xy-plane bounded below by a line
y = c, above by a line y = d, on the left by a function x = g(y), and on the right by a
function x = h(y):

To compute such integrals, we apply the following theorem:

Theorem 28. Let f (x, y) be continuous over a subset R of the plane.
• If R is a type I region, then
ZZ Z bZ h(x)
f (x, y) dA = f (x, y) dydx.
R a g(x)

• If R is a type II region, then

ZZ Z d Z h(y)
f (x, y) dA = f (x, y) dxdy.
R c g(y)

Example 73. Let R be the region defined bounded by the graphs of the functions y = x2 ,
y = x3 , x = 2 and x = 3. Compute the following integrals.
ZZ
1. x dA
R

Solution. In this example, R is a type I region. Thus, Theorem 28 tells us that

ZZ Z 3 Z x3
x dA = x dydx
R 2 x2
Z 3 y=x3
= xy dx
2 y=x2
Z 3
= x4 − x3 dx.
2

79
 This now just a basic integral that you have seen in calculus I, so we can compute it:
Z 3
519
x4 − x3 dx = .
2 20
Thus, we have that ZZ
519
x dA = .
R 20

ZZ
2. xy dA
R

Solution. Since we are again integrating over a type I region, we use Theorem 28:
ZZ Z 3 Z x3
xy dA = xy dydx
R 2 x2
Z 3
y 2 y=x3
= x· dx
2 2 y=x2
1 3 7
Z
= x − x5 dx
2 2
16255
= .
48

Reversing the Order of Integration

Sometimes, computing a given integral can be difficult, if not impossible. As an example of
this, consider the region R bounded by the line y = 2x, x = 1, and y = 0, as shown in the
following diagram:

80
Suppose that we wanted to compute the integral
ZZ
2
ex dA.
R

As a first attempt, we could try to compute it as a type II region. Thus, the integral becomes
ZZ Z 2Z 1
x2 2
e dA = ex dxdy.
R 0 y/2
2
Right away, we have run into a problem: the antiderivative of ex cannot be written in closed
form. (Remember that we had to compute this using power series before.) So computing
the integral this way will not work. Instead, let us try to consider this problem by treating
R as a type I region. In this case, the integral becomes
ZZ Z 1 Z 2x
x2 2
e dA = ex dydx
R
Z0 1 0 2x
2
= ex dx
0 0
Z 1
2
= 2xex dx
0
= e − 1.
Thus, by changing the order of integration, an integral can be significantly easier to compute.
Note that Fubini’s Theorem guarantees that the answer we obtain will not depend on the
order of integration.
Example 74. Let R be the region bounded by y = x and y = x2 . Compute the integral
ZZ
2xy − x2 dA
R

two different ways, by treating it first as a type I region, and then as a type II integral.
Solution. The region R is given in the following diagram:

81
Treating this integral as an integral over a type I region, we obtain
ZZ Z 1Z x
2
2xy − x dA = 2xy − x2 dydx
R 2
Z0 1 x x
= xy 2 − x2 y dx
x2
Z0 1
= x4 − x5 dx
0
1
= .
30
Treating R as a type II region, the integral becomes
ZZ Z 1 Z √y
2
2xy − x dA = 2xy − x2 dxdy
R 0 y
1 √
x3 x= y
Z
2
= x y− dy
0 3 x=y
Z 1
y 3/2 2y 3
= y2 − − dy
0 3 3
1
= .
30

Areas and Double Integrals

While double integrals are intended to compute volumes, they can be used to compute areas,
as well. For this, we have the following theorem:
Theorem 29. Let R be a region on the xy-plane with area A. Then
ZZ
A= 1 dA.
R
√
Example 75. Compute the area of the region R bounded by y = x2 and y = x.
Solution. Using Theorem 29, we have
ZZ
A= 1 dA
R
√
Z 1Z x
= 1 dydx
0 x2
Z 1 √
= x − x2 dx
0
1
= .
3

82
14.5 Triple Integrals
We now want consider integrals for functions of three variables f (x, y, z). It is not really
possible to visualize such an integral, as this would represent the “hyper-volume” of a four-
dimensional object. However, such integrals arise in various contexts, so it is important to
discuss them. The basic idea is the same as the two-dimensional case: we have a function
f (x, y, z) defined in some solid region G in 3-space. We approximate the region G with small
cubes of volume ∆Vk . Inside each cube, we select a point (x∗k , yk∗ , zk∗ ), which is then used to
compute the hyper-volume Hk by

Hk = f (x∗k , yk∗ , zk∗ )∆Vk .

The total hyper-volume H is then approximately equal to

N
X
H≈ f (x∗k , yk∗ , zk∗ )∆Vk .
k=1

Letting the number of cubes N → ∞, this becomes exact:

N
X
H = lim f (x∗k , yk∗ , zk∗ )∆Vk .
N →∞
k=1

If this limit exists, we denote the result by

N
X ZZZ
lim f (x∗k , yk∗ , zk∗ )∆Vk = f (x, y, z) dV.
N →∞ G
k=1

To actually compute such integrals, we begin with the simplest case: triple integrals over
rectangular regions.

Integrals over Rectangular Regions

Suppose we have a function f (x, y, z) defined on a rectangular box G defined by a ≤ x ≤ b,
c ≤ y ≤ d, and k ≤ z ≤ l. To integrate over G, we apply the following 3D analog of Fubini’s
theorem:

Theorem 30 (3D Fubini’s Theorem). Let G be the solid region satisfying a ≤ x ≤ b,

c ≤ y ≤ d, and k ≤ z ≤ l. If f is continuous in G, then
ZZZ Z bZ dZ l
f (x, y, z) dV = f (x, y, z) dzdydx.
G a c k

Moreover, this result does not change if you change the order of integration on the right-hand
side.

83
Example 76. Let G be the rectangular box defined by 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1.
Compute the following integrals.
ZZZ
1. xyz dV
G

Solution. Using Theorem 30, we have

ZZZ Z 1Z 1Z 1
xyz dV = xyz dxdydz
G 0 0 0
Z 1  Z 1  Z 1 
= x dx y dy z dz
0 0 0
 3
1
=
2
1
= .
8

ZZZ
2. x + y + z dV
G

Solution. Using Theorem 30, we have

ZZZ Z 1Z 1Z 1
x + y + z dV = x + y + z dxdydz
G 0 0 0
Z 1Z 1 2
x x=1
= + xy + xz dydz
0 0 2 x=0
Z 1Z 1
1
= + y + z dydz
0 0 2
Z 1
y y2 y=1
= + + zy dz
0 2 2 y=0
Z 1
= 1 + z dz
0
z2 z=1
=z+
2 z=0
3
= .
2

84
Integrals over Non-Rectangular Regions
In most applications of triple integrals, the solid region G will not be a simple rectangular
box. So, we need to be able to handle more general regions. The next simplest type of region
which we will discuss are regions known as simple xy-solid.
Definition 48. A solid region G is a simple xy-solid if it can be generated by translating
a region R on the xy-plane along a line parallel to the z-axis, but bounded below by the
surface z = g1 (x, y) and above by z = g2 (x, y).
For such solid regions, we apply the following theorem to compute integrals:
Theorem 31. Let G be a simple xy-solid with upper surface z = g2 (x, y) and lower surface
z = g1 (x, y), and let R be its projection onto the xy-plane. If f (x, y, z) is continuous in G,
then !
ZZZ ZZ Z g2 (x,y)
f (x, y, z) dV = f (x, y, z) dz dA.
G R g1 (x,y)

Remark 7. We can define simple xz-solids and simple yz-solids analogously. For such solid
regions, an analogous version of Theorem 31 holds, as well.
Example 77. Let G be the simple xy-solid defined by the region
R = {(x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1}
bounded above by the plane z = x + 1 and below by the xy-plane. Compute
ZZZ
x + y dV.
G

Solution. Using Theorems 31 and 27, we get

ZZZ Z 1 Z 1 Z x+1
x + y dV = x + y dzdydx
G 0 0 0
Z 1Z 1 z=x+1
= (x + y)z dydx
0 0 z=0
Z 1Z 1
= (x + y)(x + 1) dydx
0 0
Z 1Z 1
= x2 + x + y(x + 1) dydx
Z0 1 0
(x + 1)y 2 y=1
= (x2 + x)y + dx
0 2 y=0
Z 1
3 1
= x2 + x + dx
0 2 2
19
= .
12

85
Example 78. Let G be the simple xz-region defined by the region on the xz-plane bounded
by x + z = 1, x = 0, z = 0, as shown in the following diagram,

bounded on the right by y = x + 1 and on the left by y = x − 1. Compute the integral

ZZZ
z dV.
G

Solution. Using the analogous version of Theorem 31 for simple xz-regions, we get
ZZZ Z Z Z x+1 
z dV = z dy dA
G R x−1
ZZ
= 2zdA.
R

To proceed, let us choose an order of integration. If we integrate first in x, we must fix z, in

which case the integral in the x-direction will begin at x = 0 and end at x = 1 − z, as shown
in the following diagram:

86
In the z-direction, we integrate from z = 0 to z = 1. Thus, the desired integral can be
written as ZZ Z 1 Z 1−z
2zdA = 2z dxdz
R 0 0
Z 1
= 2z(1 − z) dz
0
1
= .
3

Volumes and Triple Integrals

In the previous section, we saw that it is possible to compute the area of a region R on
the xy-plane by integrating the function 1 over the region R. An analogous result holds for
volumes:

Theorem 32. Let G be a solid region in 3-space. Then the volume V of G is given by
ZZZ
V = 1 dV.
G

Example 79. Find the volume of the portion of the cylinder x2 + y 2 = 1 bounded above by
the plane z = x + 1 and below by the plane z = x − 1.

Solution. It is easy to see from the description that the solid region G in this question is a
simple xy-region. Thus, Theorem 32 says that
ZZZ
V = 1 dV.
G

By Theorem 31, this can be rewritten as

Z Z Z x+1 
V = 1 dz dA
R x−1
ZZ
= 2 dA
R
ZZ
=2 1 dA,
R

where R is the disc x2 + y 2 ≤ 1. By Theorem 29, the double integral is nothing but the area
of the disc. It follows that
V = 2π.

87
14.7 Change of Variables
Recall that u-substitution is a technique for evaluating integrals of functions of one variable
that can often greatly simplify computations. More specifically, we can write
Z b Z u(b)
0
f (u(x))u (x) dx = f (u) du.
a u(a)

This effectively changes the variable from x to u. Because of this, we refer to such a rewriting
of the integral as a change of variable.
For functions of several variables, we have an analogous trick. We will consider the
two-dimensional case, but for higher dimensions the trick is similar. We begin with some
definitions.
Definition 49. Suppose we have a change of variables defined by a function T (u, v) =
(x(u, v), y(u, v)). If S is a subset of the uv-plane contained in the domain of T , then the
image of S, denoted by T (S), is the set

T (S) = {(x, y) : (x, y) = T (u, v) for some (u, v) in S}.

The function T is referred to as a transformation of the uv-plane to the xy-plane.

Once we have such a transformation, we would like to know how it will affect a given
integral. For this, we need another definition:
Definition 50. Let T be a transformation of the uv-plane to the xy-plane defined by x =
x(u, v) and y = y(u, v). Then the Jacobian of T is the determinant
∂x ∂x
∂(x, y) ∂u ∂v
J(u, v) = = ∂y ∂y .
∂(u, v) ∂u ∂v

This definition generalizes to higher dimensions in the obvious way.

Example 80. Consider a transformation T defined by
1 1
x = (u + v) and y = (u − v).
2 2
Compute the Jacobian J(u, v) of T .
Solution. Using the formula in Definition 50, we have
∂(x, y) 1 1
= 2 2
1
∂(u, v) 2
− 12
1
=− .
2

88
Example 81. Let T be the transformation of uvw-space to xyz-space given by

x = u − uv, y = uv − uvw, and z = uvw.

Compute the Jacobian of T .

Solution. We again use Definition 50:

1−v −u 0
∂(x, y, z)
= v − vw u − uw −uv
∂(u, v, w)
vw uw uv
= (1 − v)u2 v + u2 v 2 .

With this definition in mind, we can state the change of variables formula for integrals
of functions of several variables:

Theorem 33. Let T be a transformation of the uv-plane into the xy-plane determined by
T (u, v) = (x(u, v), y(u, v)). If R is the image in the xy-plane of the region S in the uv-plane
under the transformation T , and if the Jacobian J(u, v) is non-zero and does not change
sign in S, then
ZZ ZZ
∂(x, y)
f (x, y) dAxy = f (x(u, v), y(u, v)) dAuv .
R S ∂(u, v)

The result in higher dimensions in analogous.

Example 82. Compute the integral

x−y
ZZ
dA,
R x+y
where R is the region enclosed by the lines x − y = 0, x − y = 1, x + y = 1, and x + y = 3.

Solution. Attempting to do this integral as written is quite difficult (though not necessar-
ily impossible). To simplify our computations, let us make the following observation: the
boundaries of the region R are given using equations that look like the terms in the integral.
This suggests that we try making the following change of variables:

u=x+y and v = x − y.

Solving for x and y, we get

1 1
x = (u + v) and y = (u − v).
2 2

89
We have encountered this coordinate transformation before in Example 80; its Jacobian is
∂(x, y) 1
=− .
∂(u, v) 2
Moreover, we can rewrite the boundaries of R in terms of u and v as v = 0, v = 1, u = 1
and u = 3. Using this in the formula from Theorem 33, we get that
Z 1Z 3
x−y
ZZ
v 1
dA = − dudv
R x+y 0 1 u 2
Z 1  Z 3 
1 1
= v dv du
2 0 1 u
ln 3
= .
4

Example 83. Let G be the solid region in 3-space bounded by the surfaces x = 1, x = 3,
z = y, z = y + 1, xy = 2, and xy = 4. Compute
ZZZ
(z − y)2 xy dV.
G

Solution. From the information given in the question, it is clear that we should make the
transformation
u = x, v = z − y, and w = xy.
Then the region G becomes the rectangular region 1 ≤ u ≤ 3, 0 ≤ v ≤ 1, and 2 ≤ w ≤ 4.
Solving for x, y, and z gives us
w w
x = u, y = , and z = v + .
u u
The Jacobian for this is
1 0 0
∂(x, y, z)
= − uw2 0 u1
∂(u, v, w)
− uw2 1 u1
1
=− .
u
With all these facts, we can rewrite the integral as
ZZZ Z 3Z 1Z 4 2
2 v w
(z − y) xy dVxyz = dwdvdu
G 1 0 2 u
Z 4 Z 1 Z 3
2 du
= w dw v dv
2 0 1 u
1
= 6 · · ln 3
3
= 2 ln 3.

90
14.3 and 14.6 Integrals in Cylindrical and Spherical
Coordinates
Now that we have discussed how to change from one coordinate system to another, we should
consider some special cases that appear frequently.

Polar Coordinates
In two dimensions, a useful coordinate system was the polar coordinate system (r, θ), whose
coordinates transform according to

x = r cos θ and y = r sin θ.

The Jacobian of this transformation is

∂(x, y) cos θ sin θ
=
∂(r, θ) −r sin θ r cos θ
= r.

Applying Theorem 33, we see that if R is the image of S under the transformation above,
then ZZ ZZ
f (x, y) dAxy = f (r, θ) rdArθ .
R S

Example 84. Let R be the unit disc x2 + y 2 ≤ 1 in the xy-plane. Compute the integral
ZZ p
x2 + y 2 dA.
R

Solution. To simplify the computation, we will switch to polar coordinates. First, we observe
that the unit disc x2 + y 2 ≤ 1 can be defined as the image of a set S in polar coordinates as

S = {(r, θ) : 0 ≤ r ≤ 1, 0 ≤ θ ≤ 2π}.

Thus, using the formula derived above, we may rewrite

ZZ p ZZ p
2 2
x + y dA = (r cos θ)2 + (r sin θ)2 rdrθ
R S
Z 2π Z 1
= r2 drθ
0 0
2π
= .
3

91
Example 85. Compute the integral
ZZ
xy dA,
R

where R is the quarter annulus

R = {(x, y) : 1 ≤ x2 + y 2 ≤ 2, x ≥ 0, y ≥ 0}.
Solution. As before, we can specify R as the image of the polar set
√
S = {(r, θ) : 1 ≤ r ≤ 2, 0 ≤ θ ≤ π/2}.
Thus, we may rewrite the given integral as
ZZ ZZ
xy dAxy = (r cos θ)(r sin θ) rdArθ
R S
Z π/2 Z 2
= r3 sin θ cos θ drdθ
0 1
Z π/2 ! Z
2 
1 3
= sin(2θ) dθ r dr
0 2 1
1 15
= ·
2 4
15
= .
8

Cylindrical Coordinates
In three dimensions, we have discussed two useful coordinate systems. The first of these is
the cylindrical coordinate system. Recall that the coordinate transformations for this system
are
x = r cos θ, y = r sin θ, and z = z.
In a computation almost identical to the polar coordinate case, the Jacobian becomes
cos θ sin θ 0
∂(x, y, z)
= −r sin θ r cos θ 0
∂(r, θ, z)
0 0 1
= r.
Example 86. Compute the integral
ZZZ
z dV,
G

where G is the hollow cylinder

G = {(x, y, z) : 1 ≤ x2 + y 2 ≤ 2, 0 ≤ z ≤ 1}.

92
Solution. Using Theorem 33, we have
ZZZ ZZZ
z dVxyz = z rdVrθz ,
G H

where H is the set

H = {(r, θ, z) : 1 ≤ r ≤ 2, 0 ≤ θ ≤ 2π, 0 ≤ z ≤ 1}.
Since this is a rectangle, we may rewrite the integral as an interated integral:
ZZZ Z 1 Z 2π Z 2
z rdVrθz = rz drdθdz
H 0 0 1
Z 1  Z 2π  Z 2 
= z dz dθ r dr
0 0 1
1 3
= · 2π ·
2 2
3π
= .
2

Spherical Coordinates
The last special coordinate system we discuss is the spherical coordinate system. Recall that
this is given the by transformations
x = ρ sin φ cos θ, y = ρ sin φ sin θ, and z = ρ cos φ.
It is easy to see that the Jacobian in this case becomes
sin φ cos θ −ρ sin φ sin θ ρ cos φ cos θ
∂(x, y, z)
= sin φ sin θ ρ sin φ cos θ ρ cos φ sin θ
∂(ρ, θ, φ)
cos φ 0 −ρ sin φ
= −ρ2 sin φ.
Remark 8. Most books (including our textbook) will say that the Jacobian should be
ρ2 sin φ. The reason why we obtain a negative here is because of the order of the variables
in our case, we have computed J(r, θ, φ), while the book computes J(r, φ, θ). Changing the
order in which we list the variables causes two columns in our determinant to switch places,
which introduces the negative that appears in our final solution. Since we use the absolute
value anyway, this does not matter.
Example 87. Compute the integral
ZZZ
x dV,
G

where G is the hollow sphere 1 ≤ x2 + y 2 + z 2 ≤ 4.

93
Solution. Using Theorem 33, the integral becomes
ZZZ ZZZ
x dVxyz = ρ sin φ cos θ ρ2 sin φdVρθφ ,
G H

where H = {(ρ, θ, φ) : 1 ≤ ρ ≤ 2, 0 ≤ θ ≤ 2π, 0 ≤ φ ≤ π}. The integral then becomes

ZZZ Z π Z 2π Z 2
2
ρ sin φ cos θ ρ sin φdVρθφ = ρ3 sin2 φ cos θ dρdθdφ
H
0Z 2 0 1
 Z 2π 
3
= ρ dρ cos θ dθ ×
1 0
Z π 
2
× sin φ dφ
0
15 π
= ·0·
4 2
= 0.

14.4 Surface Area

In this section, we want to consider the problem of computing areas of surfaces. We consider
two different cases.

Surfaces of the form z = f (x, y)

Recall from calculus II the arc length of a curve specified by y = f (x) from x = a to x = b
is given by the formula Z bp
L= 1 + (f 0 (x))2 dx.
a
The three-dimensional analogue of this situation is the following: suppose we have a surface
given by the function z = f (x, y), where (x, y) are elements of some region R of the xy-plane.
The surface area of this surface is then given by the following theorem.

Theorem 34. Let R be a region in the xy-plane, and let σ be a surface determined by the
equation z = f (x, y) for (x, y) in R. Then the surface area S of σ is given by
s 
ZZ 2  2
∂z ∂z
S= + + 1 dA.
R ∂x ∂y

Example 88. Find the surface area of the portion of the plane z = 1 − x − y over the region
R satisfying 0 ≤ x ≤ 1 and 0 ≤ y ≤ 1.

94
Solution. Applying the formula in Theorem 34, we obtain
ZZ √
S= 3 dA
R
Z 1Z 1√
= 3 dxdy
0 0
√
= 3.

Example 89. Compute the surface area of the upper hemisphere of a sphere of radius R.
Solution. The unit sphere is determined by the equation x2 + y 2 + z 2 = R2 . This is not a
function of the form z = f (x, y), but we can make it into one by solving for z:
p
z = ± R 2 − x2 − y 2 .
Since we are interested only in the upper hemisphere, we can simplify this to
p
z = R 2 − x2 − y 2 .
To apply the formula in Theorem 34, we compute the partial derivatives
∂z x
= −p ,
∂x R − x2 − y 2
2

∂z y
= −p .
∂y R 2 − x2 − y 2
It follows from this that
s 2  2
∂z ∂z R
+ +1= p ,
∂x ∂y R 2 − x2 − y 2
so that ZZ
R
S= p dA,
R R 2 − x2 − y 2
where R is the unit disc x2 +y 2 ≤ R. To compute this integral, we switch to polar coordinates:
ZZ Z 2π Z R
R R
p dA = √ rdrdθ
R R 2 − x2 − y 2 0 0 R − r2
2
Z 2π Z R
r
=R √ drdθ
0 0 R2 − r 2
Z 2π
=R R dθ
0
= 2πR2 .
Note that this is what one would expect from using basic geometry.

95
Parametric Surfaces
Another type of surface that we will consider are surfaces known as parametric surfaces.
These are surfaces in 3-space which are defined by parametric equations of the form

x = x(u, v), y = y(u, v), z = z(u, v),

where u and v are the parameters. This is analogous to parametric curves (in any dimension).

Example 90. Find a parametric function r(u, v) for the unit sphere x2 + y 2 + z 2 = 1 using
spherical coordinates.

Solution. If we switch to spherical coordinates, we have

x = ρ sin φ cos θ, y = ρ sin φ sin θ, z = ρ cos φ.

For the unit sphere, we have ρ = 1, so that these equations become

x = sin φ cos θ, y = sin φ sin θ, z = cos φ.

Example 91. Consider a function z = f (x, y). Write the corresponding surface σ as a
parametric surface.

Solution. For any function z = f (x, y), we may may parametrize the surface as

x = u, y = v, z = f (u, v).

To compute the surface area for a parametric surface, we proceed as we did for parametric
curves: we first consider the surface as defined by a vector function

r(u, v) = hx(u, v), y(u, v), z(u, v)i.

We then apply the following theorem:

Theorem 35. Let σ be a parametric surface defined by the vector function

r(u, v) = hx(u, v), y(u, v), z(u, v)i,

where (u, v) are in some region R of the uv-plane. Assume that the partial derivatives of r
exist and are not equal to zero. Then
ZZ
∂r ∂r
A= × dA.
R ∂u ∂u

96
Example 92. Compute the surface area of the portion of the unit sphere in the octant
0 ≤ x, 0 ≤ y, 0 ≤ z.
Solution. In Example 90, we saw that we may write the sphere as a parametric function by

r(θ, φ) = hsin φ cos θ, sin φ sin θ, cos φi.

To apply Theorem 35, we need

∂r
= h− sin φ sin θ, sin φ cos θ, 0i,
∂θ
∂r
= hcos φ cos θ, cos φ sin θ, − sin φi.
∂φ
A simple computation will show that
∂r ∂r
× = −hsin2 φ cos θ, sin2 φ sin θ, sin φ cos φi,
∂θ ∂φ
and
∂r ∂r
× = sin φ.
∂θ ∂φ
Thus, the surface area becomes
Z π/2 Z π/2
A= sin φ dφdθ
0 0
π
= .
2

Example 93. Find the surface area of the portion of the paraboloid z = x2 + y 2 between
z = 0 and z = 1.
Solution. Since this surface is specified by a function z = f (x, y), we parametrize it as

r(u, v) = hu, v, u2 + v 2 i,

where 0 ≤ u2 + v 2 ≤ 1. For this, we need

∂r ∂r
× = h−2u, −2v, 1i,
∂u ∂v
which satisfies
∂r ∂r √
× = 1 + 4u2 + 4v 2 .
∂u ∂v
Thus, the surface area is ZZ √
A= 1 + 4u2 + 4v 2 dA,
R

97
where R is the unit disc. To compute this integral, we switch to polar coordinates: u =
r cos θ, v = r sin θ. Using Theorem 33, this becomes
ZZ √ Z 2π Z 1 √
1 + 4u2 + 4v 2 dA = 1 + 4r2 rdrdθ.
R 0 0

By making the substitution w = 1 + 4r2 , we may rewrite this as

Z 2π Z 1 √
1 2π 5 √
Z Z
2
1 + 4r rdrdθ = w dwdθ
0 0 8 0 1
π 3/2

= 5 −1 .
6

98
Chapter 15

Vector Calculus

In the final chapter of the book, we take several of the topics we have already discussed and
combine them with some new ideas which are very useful. The first of these is the idea of a
vector field.

15.1 Vector Fields

We begin with a definition.

Definition 51. A vector field is a function which associates to each point P = (x, y) a
vector F(x, y) = hf (x, y), g(x, y)i.

Example 94. Here are some examples of vector fields in 2, 3, and 4 dimensions, respectively.

1. F(x, y) = hxy, x − yi

2. F(x, y, z) = hy, z, xi

3. F(x, y, z, w) = hx − y, z − w, x + z, w − yi

Remark 9. You should observe that the number of variables is always the same as the
number of components in the resulting vector. Thus, vector fields are not the same as vector
functions r(t) that we have considered before.

Remark 10. To simplify the notation a bit, we often write our coordinates (x, y) or (x, y, z)
as the vector r. Thus, a vector field can be written in the shorter form F(r).

Graphical Representations
A convenient way to represent vector fields is using the following method: suppose you have
a vector field in two dimensions, say, F(x, y) = hf (x, y), g(x, y)i. If we fill up the plane with
an imaginary grid, then at each grid point, we can draw the vector that corresponds to the
vector field at that point. If we do this for each point, then we can fill the plane with vectors

99
that tell you which direction, and with what magnitude, the vector field is pointing at that
point on the plane.

Example 95. Sketch the vector field F(x, y) = hx, yi.

Solution. By using the procedure described above, you should obtain something similar to
the following diagram:

As an example of how we could interpret such a diagram, suppose that the vectors depicted
here correspond to a force acting on an object located at the point where a particular vector
is located. Such an object would be pushed by the force in the direction of the arrow. As
the object moves, so does the force pushing it, possibly causing it to change directions. In
the diagram above, all arrows point outwards. Thus, an object being pushed by this vector
field would simply be pushed away from the origin.

Example 96. Sketch the vector field F(x, y) = h−y, xi.

Solution. Following the procedure above, you should obtain something similar to the follow-
ing diagram:

100
In this example, an object being pushed by this vector field would simply move around in a
circle or spiral centered at the origin.

Conservative Vector Fields

An important class of vector fields are those which are given by the gradient of a function.
Recall that for a function f (x, y), the gradient vector is the vector

∇φ = hφx (x, y), φy (x, y)i,

and similarly in 3 dimensions. This leads us to the following definition:

Definition 52. Let F(r) be a vector field. We say that F is conservative if there exists a
scalar function φ such that
F(r) = ∇φ.
In this case, the function φ is called the potential function.

101
Example 97. Show that φ(x, y) = x2 + y 2 is the potential function for the vector field
F(x, y) = h2x, 2yi.
Solution. It is a simple matter to compute
∂φ ∂φ
= 2x and = 2y.
∂x ∂y
Thus ∇φ = h2x, 2yi = F(x, y).

Divergence pf a Vector Field

In addition to gradient vector fields, there are two operations involving scalar functions and
vector fields which are very useful in applications of calculus.
Definition 53. Let F = hf (x, y), g(x, y)i be a vector field. We define the divergence of F
to be the scalar quantity
∂f ∂g
div F = + .
∂x ∂y
The three-dimensional case is defined analogously.
Example 98. Compute the divergence of the following vector fields.
1. F(x, y) = hx, yi
∂ ∂
Solution. div F = ∂x
(x) + ∂y
(y) = 2.

2. F(x, y) = h−y, xi
∂ ∂
Solution. div F = ∂x
(−y) + ∂y
(x) = 0.

3. F(x, y, z) = h2x2 − z, y + x, z − yi
∂ ∂ ∂
Solution. div F = ∂x
(2x2 − z) + ∂y
(y + x) + ∂z
(z − y) = 2x2 + 2.

To understand the meaning of the divergence, consider the results of parts (1) and (2)
of Example 98 with the vector fields sketched in Examples 95 and 96. The arrows which
represent the vector field F = hx, yi all “diverge” outwards from the origin, and the resulting
divergence is a positive number. On the other hand, the arrows which represent the vector
field F = h−y, xi seem to just circulate around the origin, not really beginning or ending
anywhere. As we saw, the divergence for this vector field is zero. Thus, we can make the
following observation about divergence: a point on the plane or in space which has only
arrows coming out of it will have a positive divergence. A point for which the arrows are all
going into it will have negative divergence. Such points are often referred to as “sources” and
“sinks”, respectively. If the arrows neither go into or come out of the point, the divergence
will be zero.

102
The Curl of a Vector Field
The final operation on vector fields we will discuss applies only to vector fields in 3-space.

Definition 54. Let F(x, y, z) = hf (x, y, z), g(x, y, z), h(x, y, z)i be a vector field. We define
the curl of F as the vector field defined by
 
∂h ∂g ∂f ∂h ∂g ∂f
curl F = − , − , − .
∂y ∂z ∂z ∂x ∂x ∂y

Example 99. Compute the curl of the following vector fields.

1. F = hx, y, 0i

Solution. Using the formula above, we have

curl F = h0, 0, 0i.

2. F = h−y, x, 0i

Solution. Using the formula above, we have

curl F = h0, 0, 2i.

To understand the significance of the curl vector, it is useful to look at graphical examples.
In lecture, we will consider some simple examples which will allow us to make the following
observation: the curl vector gives us a direction vector for a line which vectors will literally
“curl” around.

The Del Vector

To write the gradient, divergence, and curl using a simple, unified notation, let us introduce
the vector  
∂ ∂ ∂
∇= , , .
∂x ∂y ∂z
If we then denote the gradient of the function φ by grad φ, we can then write all three in
the easy-to-remember forms
grad φ = ∇φ,
div F = ∇ · F
curl F = ∇ × F.

103
Another special operator which you may run into in some advanced science or engineering
courses is the Laplacian operator

∂ 2φ ∂ 2φ ∂ 2φ
∆φ = + + 2.
∂x2 ∂y 2 ∂z
Using the notation above, we may write

∆φ = ∇ · ∇φ.

Because of this, some books (particularly physics textbooks) will write that

∆φ = ∇2 φ.

While this notation is incorrect (why?), it is used frequently, so you should be aware of what
it means.

15.2 Line Integrals

We now want to introduce a new type of integral which combines integral of vector functions
with vector fields. Consider the following problem:
Suppose we have a smooth curve C in 2-space defined by the vector function r(t),
with a ≤ t ≤ b, and suppose we have a function f defined on some region which
includes the curve C. Then the function f (x(t), y(t)) defines a curve in 3-space.
What is the area under this curve?
To answer this question, we may proceed as we always do: begin with an approximation using
N rectangles, compute the area of each rectangle, sum all the areas of all the rectangles, and
then let the number of rectangles N go to infinity. However, because the function is defined
along a curve, we must do the following:
• Approximate the curve C with line segments, each of length ∆sk .

• In each of the line segments from the previous step, we choose t∗k such that (x(t∗k ), y(t∗k ))
is between the endpoints of the line segment.

• The height of the rectangle in this line segment will be given by f (x(t∗k ), y(t∗k )). Thus,
the area of each approximating rectangle will be

Ak = f (x(t∗k ), y(t∗k ))∆sk .

We then sum all the areas:

N
X
A≈ f (x(t∗k ), y(t∗k ))∆sk .
k=1

104
 • The resulting sum looks like a Riemann sum. Thus, we can take a limit as N → ∞.
If this limit exists, we denote it by
N
X Z
lim f (x(t∗k ), y(t∗k ))∆sk = f (x, y) ds.
N →∞ C
k=1

This is called the line integral of f with respect to arc length over the curve C.
We can, of course, do the same thing in three dimensions, in which case the expression
becomes Z
A= f (x, y, z) ds.
C
This is nice and all, but it doesn’t tells us much about how to compute these integrals. To
compute them, note that ∆sk is the length of the line segment between two coordinates
(xk−1 , yk−1 ) and (xk , yk ). Thus, we can say that
p
∆sk = (xk − xk−1 )2 + (yk − yk−1 )2
p
= (∆xk )2 + (∆yk )2
s 2  2
∆xk ∆yk
= + ∆tk .
∆tk ∆tk
Substituting this above gives us
Z N
X
f (x, y) ds = lim f (x(t∗k ), y(t∗k ))∆sk
C N →∞
k=1
s
N 2  2
X
∗ ∗ ∆xk ∆yk
= lim f (x(tk ), y(tk )) + ∆tk .
N →∞
k=1
∆tk ∆tk

Observe that this is now a Riemann sum in the variable t. Thus, the limit becomes
Z Z b p
f (x, y) ds = f (x(t), y(t)) (x0 (t))2 + (y 0 (t))2 dt.
C a

If we recall that r(t) = hx(t), y(t)i, then this expression becomes

Z Z b
f (x, y) ds = f (r(t))kr0 (t)k dt.
C a

Thus, we have just discovered the following:

Theorem 36. Let C be a smooth curve parametrized by the function r(t) with a ≤ t ≤ b,
and let f be a function. Then the line integral of f over the curve C is
Z Z b
f (x, y) ds = f (r(t))kr0 (t)k dt.
C a

105
Remark 11. You should note that the equation above is independent of the number of
variables. Thus, it applies in any number of dimensions.
Example 100. Let f (x, y) = x2 + y 2 , and let C be the half circle parametrized by r(t) =
hcos t, sin ti where 0 ≤ t ≤ π. Compute
Z
f (x, y) ds.
C

Solution. By definition, we have

Z Z π
 2 p
f (x, y) ds = cos t + sin2 t (− sin t)2 + (cos t)2 dt
C
Z0 π
= 1 dt
0
= π.

Example 101. Let f (x, y, z) = x − y + z, and let C be the helix

r(t) = hcos t, sin t, ti

where 0 ≤ t ≤ 4π. Compute Z

f (x, y, z) ds.
C

Solution. By definition, we have

Z Z 4π p
f (x, y, z) ds = [cos t − sin t + t] (− sin t)2 + (cos t)2 + 1 dt
C
Z0 4π √
= 2 [cos t − sin t + t] dt
0
√
= 8 2π 2

Line Integrals with Respect to a Variable

A modified form of line integrals which is also useful involves integrating along a curve, but
with respect to a single variable. This requires that we pay attention to a few details. The
first of these details is the concept of orientation of a curve.
Definition 55. Let C be a curve on the plane or in space. We say that C is oriented if
there exists a vector function r(t) such that all points of C are specified by r(t) for some
value of t. The direction in which the curve is drawn as the parameter t increases is called
the orientation of the curve.

106
 The reasons why we have to pay attention to this are a bit technical, so we will not really
discuss them. Instead, we will simply assume all curves under discussion are oriented, so
that these details can be safely ignored.
Definition 56. Suppose we have a smooth, oriented curve C in 2-space, and a function
f (x, y). We define the line integral of f with respect to the variable x as
Z N
X
f (x, y) dx = lim f (x(t∗k ), y(t∗k ))∆xk
C N →∞
k=1

The line integral with respect to any other variables are defined analogously.
As before, this definition isn’t helpful in computing these integrals. However, by following
the derivation in the previous section, it is easy to obtain the following useful theorem:
Theorem 37. Let C be an oriented curve parametrized by the vector function r(t) with
a ≤ t ≤ b, and let f be a function defined on C. Then the line integral of f with respect to
x over the curve C is given by
Z Z b
dx
f (x, y) dx = f (x(t), y(t)) dt.
C a dt
An analogous equation holds for integrals with respect to other variables.
Example 102. Consider the oriented curve C defined by r(t) = ht, t2 − 2t + 5i from t = 0
to t = 1. Compute the line integral
Z
2x − y dy.
C

Solution. Using the formula above, we have

Z Z 1
2t − (t2 − 2t + 5) (2t − 2) dt
 
2x − y dx =
C 0
23
= .
6

Example 103. Compute the line integral

Z
xy dz,
C

where C is the curve parametrized by

r(t) = hcos t, sin t, sin ti,

for 0 ≤ t ≤ π.

107
Solution. Using the formula above, we have
Z Z π
xy dz = cos t · sin t · cos t dt
C 0
Z π
= cos2 t sin t dt.
0

By using the substitution u = cos t, we may rewrite this as

Z Z 1
xy dz = u2 du
C −1
2
= .
3

Example 104. Compute the line integral

Z
cos(πx) dy,
C

where C is the line segment from P0 = (1, 2) to P1 = (3, 5).

Solution. First, we need a parametrization of the given line segment. Recall that lines are
given by equations of the form
r(t) = r0 + tv0 ,
where r0 is a point on the line and v0 is the direction vector. In this example, we may use
−−→
the point P0 for r0 . For v0 , we use the vector P0 P1 , so that

r(t) = h1 + 2t, 2 + 3ti.

In particular, note that r(0) = P0 and r(1) = P1 . Thus, we can rewrite the integral as
Z Z 1
cos(πx) dy = cos π(1 + 2t) · 3 dt
C 0
= 0.

Line Integrals of Vector Fields

Using the material we have just developed, we can extend the definition of line integrals to
vector fields.

108
Definition 57. Let F = hf (x, y), g(x, y)i be a vector field, and let C be a curve parametrized
by the smooth vector function r(t) = hx(t), y(t)i from t = a to t = b. Then the line integral
of F over C is defined as
Z Z
F · dr = f (x, y) dx + g(x, y) dy.
C C

The definition for the three-dimensional case is analogous.

Example 105. Let F(x, y) = hx, yi, and let C be the unit circle. Compute
Z
F · dr.
C

Solution. Recall that we can parametrize the unit circle by x(t) = cos t and y = sin t. From
our definition, we have
Z Z
F · dr = x dx + y dy
C C
Z 2π
= − cos t · sin t + sin t · cos t dt
0
= 0.

Line Integrals Along Piecewise Smooth Curves and Negative Curves

To conclude our discussion of line integrals, we introduce some notation. Suppose that a
curve C is not smooth, but a union of several pieces C1 , . . . , Cn , each of which is smooth.
Such curves are called piecewise smooth. We can then define
Z n Z
X
F · dr = F · dr. (15.1)
C k=1 Ck

An analogous definition holds for the other types of integrals we discussed.

Another notational convenience is that of a “negative curve”. Suppose we have an ori-
ented curve C defined by the parametric equations x = f (t) and y = g(t) for a ≤ t ≤ b. We
define the negative curve −C as the curve parametrized by x = f (a+b−t) and y = g(a+b−t)
for a ≤ t ≤ b. Integrals over negative curves then satisfy the following property:

Theorem 38. Let C be a piecewise smooth curve, and let f be defined on C. Then
Z Z
F · dr = − F · dr.
−C C

109
15.3 Conservative Vector Fields
Consider the vector field F(x, y) = hx, yi, and let C1 be the line segment from (0, 0) to (1, 1)
parametrized by
r1 (t) = ht, ti.
A simple computation will show that
Z
F · dr = 1.
C1

Suppose now that we have the curve C2 parametrized by

r2 (t) = ht, t2 i.
Then another simple computation will show that
Z
F · dr = 1.
C2

The thing to observe here is the following: in both cases, we computed the line integral
for a vector field F over two different curves, both of which begin at the origin (0, 0) and
ended at the point (1, 1), and in both cases the result was the same. In addition, a simple
computation will show that
F = ∇φ,
where φ = (x2 + y 2 )/2. As we will see, this is no coincidence.
Recall that a vector field F is conservative if there exists a scalar function φ such that
∇φ = F.
These vector fields are particularly simple and useful due to the following theorem:
Theorem 39 (Fundamental Theorem of Line Integrals). Let F be the vector field
F = hf (x, y), g(x, y)i
where f and g are both defined and continuous in some region D containing the points (x0 , y0 )
and (x1 , y1 ). Assume that F is conservative, so that there exists a function φ(x, y) such that
F = ∇φ
in D. Then for any piecewise smooth curve C in D such that C starts at (x0 , y0 ) and ends
at (x1 , y1 ), we have Z
F · dr = φ(x1 , y1 ) − φ(x0 , y0 ).
C
Alternatively, we can write
Z
∇φ · dr = φ(x1 , y1 ) − φ(x0 , y0 ).
C

110
This theorem can be interpreted in the following way: if we integrate a conservative vector
field F over a curve from (x0 , y0 ) to (x1 , y1 ), then the curve itself does not matter; the only
thing that affects your answer is the endpoints.
Example 106. Let φ(x, y) = x2 y + xy 2 . Compute
Z
∇φ · dr
C

for the following parametric curves C without using Theorem 39.

1. r(t) = ht, ti for 0 ≤ t ≤ 1.

Solution. By definition, we have

Z Z 1 
2 dx 2 dy
∇φ · dr = (2x(t)y(t) + y (t)) + (x (t) + 2x(t)y(t)) dt
C 0 dt dt
Z 1
= 3t2 + 3t2 dt
0
1
= 2t3
0
= 2.

(
h2t, 0i for 0 ≤ t ≤ 1/2;
2. r(t) =
h1, 2ti for 1/2 ≤ t ≤ 1.

Solution. Using equation (15.1), we have

Z Z Z
∇φ · dr = ∇φ · dr + ∇φ · dr,
C C1 C2

where C1 is parametrized by r1 (t) = h2t, 0i for 0 ≤ t ≤ 1/2, and C2 is parametrized by

r2 (t) = h1, 2t − 1i for 1/2 ≤ t ≤ 1. For C1 , we have
Z Z 1/2  
2 dx 2 dy
∇φ · dr = (2x(t)y(t) + y (t)) + (x (t) + 2x(t)y(t)) dt
C1 0 dt dt
= 0.
For C2 , we have
Z Z 1  
dx dy
∇φ · dr = (2x(t)y(t) + y (t)) + (x2 (t) + 2x(t)y(t))
2
dt
C2 1/2 dt dt
Z 1
= (4t − 1) · 2 dt
1/2

= 2.

111
 Thus Z Z Z
∇φ · dr = ∇φ · dr + ∇φ · dr
C C1 C2
=0+2
= 2.

Integrals along Closed Curves

We now want to pay particular attention to a situation that comes up very often: the curve
C begins and ends at the same point. Based on the results of Theorem 39, we would expect
that if F is conservative and C is a curve which begins and ends at the same point, then
Z
F · dr = 0.
C

It turns out that this is almost true, but there are some potential problems that can arise.
To avoid these problems, we must make some additional assumptions. This requires the
following definitions:
Definition 58. Let C be a curve parametrized by r(t) for a ≤ t ≤ b. We say that C is
closed if r(a) = r(b).
Definition 59. Let D be a set in 2- or 3-space. We say that D is connected if we can draw
a curve between any two points in D without having to leave D.
Example 107. The set D shown here is connected:

112
The set D shown here is not connected:

Now we may state the correct form of the expected result.

Theorem 40. Let f (x, y) and g(x, y) be continuous on some open, connected region D.
Then the following are equivalent:

1. F(x, y) = f (x, y)i + g(x, y)j is a conservative vector field on D.

Z
2. F · dr = 0 for every piecewise smooth closed curve C in D.
C
Z
3. F · dr is independent of the path C chosen from (x0 , y0 ) to (x1 , y1 ).
C

This theorem is very helpful if you already know that F is a conservative vector field.
To check if a given vector field F is conservative, we have a test we can use, which will be
stated below; however, it requires the following definitions:

Definition 60. Let C be a curve parametrized by r(t). We say C is simple if it does not
intersect itself, except at its endpoints.

Definition 61. Let D be a connected set in 2-space. We say D is simply connected if there
are no simple closed curves in D which enclose points not in D.

Example 108. The set D in this diagram is simply connected:

113
The set D in this diagram is not simply connected:

With these definitions, we may finally state our test for conservative vector fields.

Theorem 41 (Conservative Vector Field Test). Let f (x, y) and g(x, y) be continuous and
have continuous first partial derivatives in some open region D. If F = f (x, y)i + g(x, y)j is

114
a conservative vector field on D, then
∂f ∂g
= (15.2)
∂y ∂x

at each point in D. Conversely, if D is simply connected an (15.2) holds at each point in D,

then F is a conservative vector field.

Example 109. Determine if the following vector fields are conservative on any open set D:

1. F = hy, −xi

Solution. To apply equation (15.2), we compute

∂f ∂g
= 1, and = −1.
∂y ∂x
Since this holds on the whole xy-plane, it follows that that the vector field is not
conservative anywhere.

2. F = x2 y + 31 y 3 , 13 x3 + xy 2 .

Solution. It is easy to check that

∂f ∂g
= = x2 + y 2 .
∂y ∂x

Since this holds for any (x, y), it follows that F is conservative on the whole xy-
plane.

Computing Potential Functions

Theorem 41 not only gives us a way to determine if a given vector field is conservative, but
it also gives us a way to compute the potential function φ for a conservative vector field,
according to the following procedure:
R
1. Integrate f with respect to x to obtain φ(x, y) = f (x, y) dx + F (y).
R

2. Differentiate with respect to y to obtain φy (x, y) = f (x, y) dx y + Fy (y). Set this
equal to g, and use this to find F (y).

Example 110. Let F(x, y) = 2xy 3 i + (1 + 3x2 y 2 )j. Find the potential function φ(x, y).

115
Solution. As the procedure says, we integrated f (x, y):
Z
φ(x, y) = f (x, y) dx = x2 y 3 + F (y).

Differentiating with respect to y gives us

∂φ
= 3x2 y 2 + F 0 (y) = 1 + 3x2 y 2 ,
∂y
so that F 0 (y) = 1. Integrating in y tells us that F (y) = y + C. Since the constant C doesn’t
matter, we choose C = 0. Thus
φ(x, y) = x2 y 3 + y.

Example 111. Compute the potential function for the vector field
 
1 x
F= ,−
y y2
and compute the integral Z
F · dr,
C
where C is the unit circle r(t) = hcos t, sin ti from t = π/4 to t = 3π/4.
Solution. As before, we first integrate f (x, y) in x:
Z
1 x
φ(x, y) = dx = + F (y).
y y
Next, we differentiate in y:
∂φ x x
= − 2 + F 0 (y) = − 2 .
∂y y y
From this, we see that
F 0 (y) = 0,
so that F (y) = C. As before, we may choose C = 0, which gives us
x
φ(x, y) = .
y
To evaluate the line integral, we apply the Fundamental Theorem of Line Integrals:
Z √ √ ! √ √ !
2 2 2 2
F · dr = φ − , −φ ,
C 2 2 2 2
= −1 − 1
= −2.

116
15.4 Green’s Theorem
In this section, we will discuss an important theorem which connects line integrals with area
integrals. We remark that this result holds only in two dimensions.

Theorem 42 (Green’s Theorem). Let R be a simply connected plane region whose boundary
is a simple closed piecewise smooth curve C oriented counterclockwise. If f (x, y) and g(x, y)
are continuous and have continuous first partial derivatives on some open set containing R,
then Z ZZ  
∂g ∂f
f (x, y) dx + g(x, y) dy = − dA. (15.3)
C R ∂x ∂y
This theorem says something significant: instead of doing a line integral, you can instead
replace it with an area integral, and vice-versa. Depending on the circumstances, one may
be significantly more easier than the other (say, if the derivatives on the right side are equal
...)

Example 112. Evaluate the integral

Z
x2 y dx + xdy,
C

where C is the path consisting of a triangle with vertices at (0, 0), (1, 0) and (1, 1), using
Green’s Theorem.

Solution. In this example, we have f (x, y) = x2 y and g(x, y) = x. By Green’s Theorem, we

have Z ZZ
2
x y dx + xdy = 1 − x2 dA
C
Z 1RZ x
= 1 − x2 dydx
0 0
Z 1
= x − x3 dx
0
1
= .
4

Example 113. Compute the integral

Z
x − y dx + y − x dy
C

where C is any simple closed piecewise smooth curve on the plane.

117
Solution. Since
∂g ∂f
= ,
∂x ∂y
it follows from Green’s Theorem that
Z ZZ
x − y dx + y − x dy = 0 dA = 0,
C R

regardless of the choice of C. Here, we assume that R is the region enclosed by C.

Remark 12. If C is a simple closed curve, we use the notation
I
f (x, y) dx + g(x, y) dy

to emphasize the fact that the integral is over a closed curve.

15.5 Surface Integrals

We now want to consider the following problem: suppose we have a function f (x, y, z) defined
on some region of 3-space, and a surface σ which is inside that region. How do we integrate
a function over only the values of (x, y, z) that are contained in the surface σ?
As always, we proceed by first approximating with rectangles. Since the surface σ is a
two-dimensional object, we divide it into rectangular pieces, each which has surface area ∆Sk .
We then choose a coordinate (x∗k , yk∗ , zk∗ ) inside this rectangle, and we create a rectangular
cube whose volume is
Vk = f (x∗k , yk∗ , zk∗ )∆Sk .
Then the total volume of this solid region is approximately given by
N
X
V = f (x∗k , yk∗ , zk∗ )∆Sk .
k=1

As N → ∞, this approximation becomes an exact equality. Thus, we have that

N
X
V = lim f (x∗k , yk∗ , zk∗ )∆Sk ,
N →∞
k=1

if this limit exists. If it does exist, we denote it by

N
X ZZ
lim f (x∗k , yk∗ , zk∗ )∆Sk = f (x, y, z) dS.
N →∞ σ
k=1

This is known as a surface integral over the surface σ.

As is usually the case, the definition does not give us much information about how to
compute such a thing. Luckily for us, the following theorem tells us how to compute such
integrals.

118
Theorem 43. Let σ be a smooth parametric surface defined by the vector function
r(u, v) = hx(u, v), y(u, v), z(u, v)i,
where (u, v) varies over a region R in the uv-plane. If f (x, y, z) is continuous on σ, then
ZZ ZZ
∂r ∂r
f (x, y, z) dS = f (x(u, v), y(u, v), z(u, v)) × dA.
σ R ∂u ∂v
Example 114. Let σ be the surface defined by r(u, v) = hu, v, 0i for 0 ≤ u ≤ 1 and
0 ≤ v ≤ 1. Compute ZZ
x + y + z dS.
σ

Solution. To apply Theorem 43, we first compute

∂r ∂r
× = h0, 0, 1i.
∂u ∂v
Thus, we have ZZ ZZ
x + y + z dS = (u + v) kh0, 0, 1ik dA
σ R
Z 1Z 1
= u + v dudv
0 0
Z 1 2
u 1
= + uv dv
2 0
Z0 1
1
= + v dv
0 2
= 1.

Example 115. Evaluate ZZ

x2 + y 2 + z 2 dS,
σ
where σ is the unit sphere x2 + y 2 + z 2 = 1.
Solution. If we switch to spherical coordinates
x = sin φ cos θ, y = sin φ sin θ, z = cos φ,
then we can parametrize σ by
r(θ, φ) = hsin φ cos θ, sin φ sin θ, cos φi,
where 0 ≤ θ ≤ 2π and 0 ≤ φ ≤ π. A lengthy computation will show that
∂r ∂r
× = h− sin2 φ cos θ, − sin2 φ sin θ, − sin φ cos φi.
∂θ ∂φ

119
From this, it is easy to see that
∂r ∂r
× = sin φ.
∂θ ∂φ
Moreover, we also have that x2 + y 2 + z 2 = 1. Thus, applying Thorem 43, we get
ZZ Z π Z 2π
2 2 2
x + y + z dS = 1 · sin φ dθdφ
σ 0 0
= 4π.

A Special Case: Surfaces Defined by Functions

In the case where σ is defined by functions, it can be shown that the formula in Theorem 43
reduces to a much simpler formula. As an example, suppose that σ is defined by a function
z = g(x, y). In this case, the formula reduces to
s 
ZZ ZZ 2  2
∂z ∂z
f (x, y, z) dS = f (x, y, g(x, y)) + + 1 dA.
σ R ∂x ∂y
The formula is analogous in the cases y = g(x, z) and x = g(y, z).
Example 116. Evaluate ZZ
xz dS,
σ
where σ is the portion of the plane x + y + z = 1 in the first octant.
Solution. We may rewrite the equation for the surface in the form z = 1 − x − y, so that it
is defined by a function. Applying our new formula, we have that
ZZ ZZ √
xz dS = x(1 − x − y) 3 dA,
σ R

where R is the region in the xy-plane bounded by x = 0, y = 0, and x + y = 1 (why?). Thus,

we may rewrite the integral as
ZZ √ Z 1 Z 1−x
xz dS = 3 x − x2 − xy dydx
σ 0 0
√ Z 1 xy 2 1−x
= 3 xy − x2 y − dx
0 2 0
√ Z 1 1
= 3 x − x2 − x2 + x3 − (x3 − 2x2 + x) dx
2
√ 0
3
= .
24

120
15.6 Flux
An important concept which is useful in applications of vector calculus is the concept of flux.
Before we can explain what it means, we must define some new concepts which are necessary
in the description of flux.

Orientable Surfaces
Let σ be a surface in 3-space. Roughly speaking, we say σ is orientable if it has two distinct
sides. We can make this more mathematically rigorous by specifying each side of the surface
by a continuous unit normal vector n(x, y, z), defined for each (x, y, z) on σ.
Example 117. Let σ be the unit sphere x2 + y 2 + z 2 = 1. Then for each (x, y, z) in σ, the
vector n = hx, y, zi defines a unit normal vector on σ. Thus it is an orientation of σ.
Example 118. Let σ be a smooth parametric surface parametrized by r(u, v). Then
∂r ∂r
∂u
× ∂v
n(u, v) = ∂r ∂r
∂u
× ∂v

defines a unit normal vector on σ. This orientation is called the positive orientation for σ.

Flux
Consider a vector field F(x, y, z) defined in some region of 3-space which contains the surface
σ. Let us assume that the vector field F(x, y, z) represents the motion of some fluid (where
fluid should be interpreted to mean anything that can be represented by a vector field) that
passes through the point (x, y, z) per unit of time. A question of interest in many different
subjects is the following: what is the total amount of fluid that passes through the surface
σ per unit of time? This quantity is referred to as the flux of the vector field F the surface
σ, and is denoted by Φ.
The obvious question to ask ourselves at this point is the following: how do we compute
Φ? The answer to this question is contained in the following theorem:
Theorem 44. Let σ be a smooth parametric surface defined by r(u, v), where (u, v) vary
over a region R of the uv-plane, and let
F(x, y, z) = hf (x, y, z), g(x, y, z), h(x, y, z)i
be a vector field such that f , g, and h are continuous on σ. Then
ZZ ZZ  
∂r ∂r
Φ= F · n dS = F· × dA.
σ R ∂u ∂v
A simple but lengthy computation will show that
∂r ∂r
× =
∂θ ∂φ

121
Example 119. Find the flux of the vector field F(x, y, z) = hy, z, xi across the surface σ
parametrized by r(u, v) = hu, v, 0i, where 0 ≤ u ≤ 1 and 0 ≤ v ≤ 1, oriented upwards.
Solution. To apply Theorem 44, we first compute
∂r ∂r
× = i × j = k.
∂u ∂v
This points upwards, so that it is the correct orientation. Thus
 
∂r ∂r
F· × = u,
∂u ∂v
and ZZ
Φ= u dA
R
Z 1Z 1
= u dudv
0 0
1
= .
2

Example 120. Find the flux of the vector field F(x, y, z) = h0, 0, zi across the outward
oriented unit sphere x2 + y 2 + z 2 = 1.
Solution. To apply Theorem 44, we need a parametrization of the surface. For this, we turn
to the spherical coordinate representation
r(θ, φ) = hsin φ cos θ, sin φ sin θ, cos φi.
In Example 115, we saw that
∂r ∂r
× = h− sin2 φ cos θ, − sin2 φ sin θ, − sin φ cos φi.
∂θ ∂φ
Note that this points inward, so that we must choose the opposite orientation:
∂r ∂r
− × = hsin2 φ cos θ, sin2 φ sin θ, sin φ cos φi.
∂θ ∂φ
Thus  
∂r ∂r
F· − × = sin φ cos2 φ.
∂θ ∂φ
From this, we see that ZZ  
∂r ∂r
Φ= F· − × dA
R ∂θ ∂φ
Z π Z 2π
= sin φ cos2 φ dA
0 0
4π
= .
3

122
15.7 The Divergence Theorem
The next of the big theorems of vector calculus is a theorem known as the Divergence
Theorem. Before we can state it, we need more preliminary definitions.
Definition 62. We say that a surface σ in 3-space is closed if the surface is the boundary
of a solid region G of 3-space.
With this definition, we can state the main theorem of this section.
Theorem 45 (Divergence Theorem). Let G be a solid region of 3-space with boundary surface
σ oriented outwards from the solid. If

F(x, y, z) = hf (x, y, z), g(x, y, z), h(x, y, z)i

is a vector field whose components f , g, and h have continuous first partial derivatives on
some open set containing G, and if n is the outward unit normal vector on σ, then
ZZ ZZZ
F · n dS = ∇ · F dV.
σ G

Example 121. Use the Divergence Theorem to compute the flux across the outward oriented
unit cube G = {(x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1} of the vector field

F(x, y, z) = h2x, 3y, z 2 i.

Solution. Recall that the flux Φ is computed by

ZZ
Φ= F · n dS,
σ

where σ is the surface of the unit cube. By the Divergence Theorem, we may write
ZZZ
Φ= ∇ · F dV
ZZZ G

= 5 + 2z dV
G
Z 1Z 1Z 1
= 5 + 2z dzdydx
0 0 0
Z 1Z 1 z=1
= 5z + z 2 dydx
0 0 z=0
Z 1Z 1
= 6 dydx
0 0
= 6.

123
Example 122. Let F be the vector field

F(x, y, z) = hx + z 2 sin(yz), y − x2 z 3 cos(xz), zi.

Find the flux Φ of F across the surface σ of the outward oriented cylinder G bounded by
the surfaces x2 + y 2 = 1, z = 0 and z = 1.
Solution. Trying to compute the flux directly using Theorem 44 will result in some very
complicated integrals. Thus, if we make the observation that

∇ · F = 3,

then the Divergence Theorem tells us that

ZZ
Φ= F · n dS
ZZZσ

= 3 dV
Z ZGZ
=3 dV.
G

Since the integral above is just the volume of the cylinder, we have

Φ = 3π.

15.8 Stokes’ Theorem

In the final chapter of the course, we will discuss the last of the great theorems of vector
calculus, a result known as Stokes’ Theorem. To properly state it, we must define a more
specific form of the term orientation for curves.
Definition 63. Consider a surface σ with orientation vector n. Let C be the curve that
defines the boundary of σ. Imagine a person standing on the curve C so that the direction
from their feet to their head is the same as that of n. (See the diagram in the textbook.) If
the person is walking along the curve C so that the surface is located to their left, we say
the curve has positive orientation. If the person is walking on the curve so that the surface
is to their right, we say the curve has negative orientation.
With this definition, we may now state Stokes’ Theorem.
Theorem 46 (Stokes’ Theorem). Let σ be a piecewise smooth oriented surface that has a
boundary which is a simple, closed, piecewise smooth curve C with positive orientation. If
the components of the vector field

F(x, y, z) = hf (x, y, z), g(x, y, z), h(x, y, z)i

124
are continuous and have continuous first partial derivatives on some open set containing σ,
then I ZZ
F · dr = (∇ × F) · n dS.
C σ

Example 123. Use Stokes’ Theorem to compute

I
F · dr
C

for the vector field F = hy, z, xi, where C is the boundary curve of the surface defined by
x2 + y 2 ≤ 1, z = 0, oriented upwards.

Solution. The surface σ is simply the unit disc on the plane z = 0, which we oriented with
the unit normal vector n = h0, 0, 1i. To apply Stokes’ Theorem, we instead compute a
surface integral over the parametric surface

s(r, θ) = hr cos θ, r sin θ, 0i,

where 0 ≤ r ≤ 1 and 0 ≤ θ ≤ 2π. A simple computation will show that

∇ × F = h1, −1, 1i,

from which we see that

∇ × F · n = 1.
Thus, Stokes’ Theorem says that
I ZZ
F · dr = (∇ × F) · n dS
C Z Zσ
= dS.
σ

Recall that the integral in the final line above is nothing but the surface area of σ. Since σ
is simply the unit disc x2 + y 2 ≤ 1 on the xy-plane, its surface are is given by S = π, so that
I
F · dr = π.
C

125