lOMoARcPSD|3471321

MATH0011 - Lecture notes Second half of term (multivariable

calculus)
Mathematical Methods 2 (University College London)

StuDocu is not sponsored or endorsed by any college or university

Downloaded by mick dundee ([email protected])
 lOMoARcPSD|3471321

Chapter 1

INTRODUCTION

1.1 Review from Methods I

1.1.1 Partial derivatives
To find a partial derivative we hold all but one of the independent variables constant and differentiate
with respect to that one variable using the ordinary rules for one-variable calculus. So for a function
f (x, y), we have

∂f f (x + h, y) − f (x, y)
= fx = lim
∂x h→0 h
∂f f (x, y + h) − f (x, y)
= fy = lim
∂y h→0 h
You form second, and higher, order partial derivatives in the same way:
� �
∂ ∂f fx (x, y + h) − fx (x, y)
= fxy = lim
∂y ∂x h→0 h

and for sufficiently smooth functions it doesn’t matter what order you take mixed derivatives in:

fxxy = fxyx = fyxx .

1.1.2 Gradient and directional derivative

For a scalar function f (x, y, z), we define the gradient of f (often pronounced “grad f ”) as

∂f ∂f ∂f
∇f = i+ j+ k.
∂x ∂y ∂z

The equivalent for functions of two variables is just what you’d expect: since ∂f /∂z = 0, we would have

∂f ∂f
∇f = i+ j.
∂x ∂y
The gradient vector
• points in the direction in which f increases most rapidly
• has magnitude which is the rate of change of f in that direction
1

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

• is perpendicular to level sets of f .

For any unit vector û, we define the directional derivative in the direction of û as
Dû f = û · ∇f.
Note that if û lies within a level set of f (pointing along a contour line in 2D, or lying within a tangent
plane in 3D) then Dû f will be zero. The gradient vector itself is perpendicular to these directions.

1.1.3 Geometric interpretation of functions of several variables

One way to interpret a function of two variables is to think of the function f (x, y) as the height of a
surface z = f (x, y).
Example: f (x, y) = 6 − 2x − 3y
The surface z = 6 − 2x − 3y, i.e. 2x + 3y + z = 6, is a plane
which intersects the x-axis where y = z = 0, i.e. x = 3;
which intersects the y-axis where x = z = 0, i.e. y = 2;
which intersects the z-axis where x = y = 0, i.e. z = 6.
Example: f (x, y) = x2 − y 2 10
8
6
4
2
0

In the plane x = 0, there is a maximum at y = 0; in the plane -2

-4
-6
-8
-10

y = 0, there is a minimum at x = 0. The whole surface is 1

2
3

shaped like a horse’s saddle; and the picture shows a structure

-3 0
-2 y
-1 -1
0
1 -2
x 2
3 -3

for which (0, 0) is called a saddle point.

Example : f (x, y) = x2 + y 2 18
16
14
12
10

In the plane x = 0, there is a minimum at y = 0; in the plane 8

6
4
2
0

y = 0, there is a minimum at x = 0. Any vertical section 1

2
3

through this surface gives a parabola: this is a paraboloid bowl.

-3 0
-2 y
-1 -1
0
1 -2
x 2
3 -3

Another way to picture the same surface is to do as map-makers or weather forecasters do and draw
contour lines (or level curves) – produced by taking a section, using a plane z = const. and projecting it
onto the xy-plane. For f (x, y) = x2 + y 2 as above, the contour lines are concentric circles.
For a function of three variables, g(x, y, z), one of the most intuitive ways to understand its meaning
is to think about its level surfaces. For example, the function g(x, y, z) = x2 + y 2 + z 2 has levels sets
which are concentric spheres; the function g(x, y, z) = x + y + z has level sets which are planes.
A surface z = f (x, y) can also be thought of in this way by setting g(x, y, z) = z − f (x, y). Then our
surface is the level set g(x, y, z) = 0.
Example
Find a unit vector perpendicular to the surface z = x2 + y 2 at the point A(1, 2, 5). Deduce the
equation of the tangent plane to the surface at A.
Solution
The normal to the surface g = constant at any point is in the direction of ∇g at that point. In this
case the formulation g = constant is given as g(x, y, z) = z − x2 − y 2 = 0. ∇g = −2xi − 2yj + k so
at the point A we have ∇g = −2i − 4j + k. The unit normal is n = (21)−1/2 (−2i − 4j + k).
The tangent plane at A is perpendicular to the normal at A. The equation of the plane is
r · n = const. where r = xi + yj + zk, i.e. in this case (x, y, z) · (21)−1/2 (−2, −4, 1) = constant.
Substituting in the point A(1, 2, 5) to find the constant, and tidying up, we find the tangent plane
is 2x + 4y − z = 5.
Lecture 1

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 3

r
y
θ
x

Figure 1.1: Polar coordinates

1.2 Coordinate systems

1.2.1 Cartesian coordinates
We are familiar with the rectangular Cartesian coordinates x, y in a plane, or x, y, z in space. If the
unit vectors in the x, y and z directions are given by i, j and k then we can write the position vector r
of a point as
r = (x, y, z) = xi + yj + zk.
These two notations will be used interchangeably in this course.

1.2.2 Plane polar coordinates

In a fixed frame of reference the rectangular coordinates (x, y) are related to the polar coordinates (r, θ)
through the relations
x = r cos θ, y = r sin θ, r > 0, θ ∈ [0, 2π) (1.1)
�y�
r = (x2 + y 2 )1/2 , θ = arctan . (1.2)
x
Example
Express in polar coordinates the portion of the unit disc that lies in the first quadrant.

Solution
The region may be expressed in polar coordinates as 0 ≤ r ≤ 1 and 0 ≤ θ ≤ π/2.

1.2.3 Cylindrical polar coordinates

In a fixed frame of reference the Cartesian coordinates (x, y, z) are related to the cylindrical coordinates
(r, θ, z) through the relations:

x = r cos θ, y = r sin θ, z = z, r > 0, θ ∈ [0, 2π) (1.3)

�y�
r = (x2 + y 2 )1/2 , θ = arctan , z = z. (1.4)
x
Example
Express in cylindrical coordinates the solid bounded above by the cone z = 2 − (x2 + y 2 )1/2 and
bounded below by the disc Ω : (x − 1)2 + y 2 ≤ 1.

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

Solution
The region Ω may be expressed as (r cos θ − 1)2 + r2 sin2 θ ≤ 1 which is −2r cos θ + r2 ≤ 0 which
becomes r ≤ 2 cos θ, and −π/2 ≤ θ ≤ π/2.
The z-limits are 0 ≤ z ≤ 2 − (x2 + y 2 )1/2 which become 0 ≤ z ≤ 2 − r.
We can see that cylindrical polar coordinates are just plane polar coordinates with the usual Cartesian
z added on to make them 3D.

1.2.4 Spherical coordinates

ρ
θ

x φ

Figure 1.2: Spherical polar coordinates

In a fixed frame of reference the Cartesian coordinates (x, y, z) are related to the spherical coordinates
(ρ, θ, φ) through the relations:
x = ρ sin θ cos φ, y = ρ sin θ sin φ, z = ρ cos θ, (1.5)
ρ > 0, θ ∈ [0, π], φ ∈ [0, 2π) (1.6)
�y�
ρ = (x2 + y 2 + z 2 )1/2 , φ = arctan , θ = arccos [z(x2 + y 2 + z 2 )−1/2 ] (1.7)
x
In the above equations, θ is the latitude or polar angle, and φ is the longitude.
Remark: Spherical coordinates are commonly used in applications where there is a centre of symmetry.
The centre of symmetry is then taken as the origin.
Example
Express in spherical polar coordinates the solid T that is bounded above by the cone z 2 = x2 + y 2 ,
below by the xy-plane, and on the sides by the hemisphere z = (4 − x2 − y 2 )1/2 .
Solution
The solid is defined by the following inequalities:
z2 ≤ x2 + y 2
z ≥ 0
x + y + z2
2 2
≤ 4

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 5

Substituting the definitions of x, y and z in terms of ρ, θ and φ gives:

ρ2 cos2 θ ≤ ρ2 sin2 θ cos2 φ + ρ2 sin2 θ sin2 φ

ρ cos θ ≥ 0
ρ2 sin2 θ cos2 φ + ρ2 sin2 θ sin2 φ + ρ2 cos2 θ ≤ 4

so

ρ2 cos2 θ ≤ ρ2 sin2 θ
ρ cos θ ≥ 0
ρ2 ≤ 4

and we can use the fact that ρ ≥ 0 and sin θ ≥ 0 to deduce

ρ ≤ 2
cos θ ≥ 0
1 ≤ tan θ.

Given that 0 ≤ θ ≤ π, this reduces to:

0≤ ρ ≤2
π/4 ≤ θ ≤ π/2.

In this case, where there is no information about φ contained in our limits, we use the whole
permitted range:
0 ≤ φ < 2π.

Note: The radial coordinate in spherical coordinates, which we’ve denoted ρ, is often called r (because
it is a radius). Either notation is fine, but if you use r then make sure you specify which coordinates you
are using to avoid confusion.

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

Chapter 2

CALCULUS OF VECTOR
FUNCTIONS

2.1 Vector Functions

A vector-valued function f is a vector function whose components are single-valued functions (scalar
valued functions).
For example, given three single-valued functions f1 (t), f2 (t), f3 (t) we can form the vector-valued
function
f (t) = (f1 (t), f2 (t), f3 (t)) = f1 (t)i + f2 (t)j + f3 (t)k (2.1)
The magnitude of the vector-valued function f (t) is a scalar-valued function and is defined by
� � � �
�f (t)� = f12 (t) + f22 (t) + f32 (t) 1/2 (2.2)

Examples

(a) f1 (t) = cos(t), f2 (t) = sin(t), f3 (t) = 0.

(b) f1 (t) = t, f2 (t) = g(t), f3 (t) = 0 for t ∈ [a, b].

In general, the graph of the vector function f (t) = f1 (t)i + f2 (t)j + f3 (t)k is a curve C, in the sense that,
as t varies, the tip of the position vector f (t) traces out C. The equations

x = f1 (t), y = f2 (t), z = f3 (t) (2.3)

corresponding to the components of f are the parametric equations of C. If one of the components is
zero, e.g. f (t) = f1 (t)i + f2 (t)j, then C is said to be a planar curve, otherwise C is a space curve.

2.1.1 Derivative of a vector function

Given the vector function f (t) = (f1 (t), f2 (t), f3 (t)) the derivative of f is defined by

f � (t) = (f1� (t), f2� (t), f3� (t)).

Example
Find f ��� (t) for f = (t sin(t), e−t , t).

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 7

Solution
f (t) = (t sin(t), e−t , t)
f � (t) = (sin(t) + t cos(t), −e−t , 1)
f �� (t) = (2 cos(t) − t sin(t), e−t , 0)
f ��� (t) = (−3 sin(t) − t cos(t), −e−t , 0).

Properties of the derivative

• (f + g)� (t) = f � (t) + g � (t)

• (αf )� (t) = αf � (t) where α is a constant scalar

• (uf )� (t) = u(t)f � (t) + u� (t)f (t), where u is a scalar function

• (f · g)� (t) = f (t) · g � (t) + f � (t) · g(t).

• (f × g)� (t) = f (t) × g � (t) + f � (t) × g(t).

• (f (u))� (t) = f � (u(t))u� (t).

Velocity
If the function r(t) represents a position vector, as a function of time, then its derivative gives us the
velocity of the point:
dr
v= = r� (t).
dt
Lecture 2

2.1.2 Integral of a vector function

�b
Given the vector function f (t) = (f1 (t), f2 (t), f3 (t)), the integral of f is defined by a
f (t) dt =
�b �b �b
( a f1 (t) dt, a f2 (t) dt, a f3 (t) dt).

Example �
1
Find 0 f (t) dt for f (t) = (t, (t + 1)1/2 , −et ).

Solution
� 1 � 1 � �
f (t) dt = t, (t + 1)1/2 , −et dt
0 0
�� 1 � 1 � 1 �
1/2 t
= t dt, (t + 1) dt, −e dt
0 0 0
� �
� 1 2 �1 � 2 3/2
�1 � �
t 1
= 2 t 0 , 3 (t + 1) , −e 0
0
� �
= 1
2− 0, 32 (2)3/2 − 32 (1)3/2 , −e + 1
� �
1 2 3/2
= 2 , 3 [2 − 1], 1 − e

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

Properties of the integral

�b �b �b
• a (f (t) + g(t)) dt = a f (t) dt + a g(t) dt
�b �b
• a (αf )(t) dt = α a f (t) dt, where α is a constant scalar
�b �b
• a (c · f )(t) dt = c(t) · a f (t) dt, where c is a constant vector.
�b �b
• a (c × f )(t) = c(t) × a f (t) dt, where c is a constant vector.

2.2 Curves
1. The equation of a straight line is parametrised by
r(t) = r0 + td, t ∈ (−∞, ∞) (2.4)

2. More generally, every vector function

r(t) = x(t)i + y(t)j + z(t)k (2.5)
parametrises a curve in space (or a curve in plane if one of the components x(t), y(t) or z(t) is
zero). It is also important to understand that a parametrised curve C is an oriented curve in the
sense that as t increases on some interval of definition I, the tip of the position vector r(t) traces
out C in a certain direction. For example, the unit circle parametrised by
r(t) = cos(t)i + sin(t)j, t ∈ [0, 2π)
is traversed in the anticlockwise direction, starting at the point (1,0).

2.2.1 Tangent vector, tangent line

For a given curve C parametrised by
r(t) = (x(t), y(t), z(t)),
the derivative vector
r� (t) = (x� (t), y � (t), z � (t))
is called the tangent vector to the curve C at the point P (x(t), y(t), z(t)), and r� (t) points out in the
direction of increasing t.
Example
In the case of a circle of radius a parametrised as r(t) = a cos(t)i + a sin(t)j, show that the tangent
vector r� (t) is perpendicular to the radius vector r(t).
Solution
The radius vector is r(t) = (a cos(t), a sin(t)) which has |r| = a. We differentiate to get
r� (t) = −a sin(t)i + a cos(t)j
which also has magnitude |r� | = a. Now
r(t) · r� (t) = (a cos(t), a sin(t)) · (−a sin(t), a cos(t))
= a cos(t)(−a sin(t)) + a sin(t)a cos(t) = 0
and since we know that if |c| �= 0 and |d| �= 0 but c · d = 0 then c is perpendicular to d, so in this
case r� (t) is perpendicular to r(t).

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 9

For a given curve C parametrised by r(t) = (x(t), y(t), z(t)) the tangent line at a point t is the vector
function
R(λ) = r(t) + λr� (t), λ ∈ (−∞, ∞) (2.6)

Example
Find a vector tangent to the twisted cubic

r(t) = ti + t2 j + t3 k

at the point P (2, 4, 8), and then parametrise the tangent line.

Solution
The derivative along the curve is
r� (t) = i + 2tj + 3t2 k
and at the point P (2, 4, 8) we have t = 2 and so

r� = i + 4j + 12k

and the tangent line at the point P is

R(λ) = (2 + λ, 4 + 4λ, 8 + 12λ).

2.2.2 Taylor series

The tangent line equation at the point t = a:

R(λ) = r(a) + λr� (a), (2.7)

is in fact an approximation to the curve at that point:

r(t) ≈ r(a) + (t − a)r� (a). (2.8)

It is the best linear approximation to the curve at t = a, and is the beginning of a Taylor series for r
about the point r(a):
∞
� (t − a)n (n)
r(t) = r(a) + r (a). (2.9)
n=1
n!
Lecture 3
2.2.3 Arc length
The length of a continuously differentiable curve (C): r = r(t), t ∈ [a, b] is given by
� b
L(C) = |r� (t)| dt. (2.10)
a

Example Find the length of the curve r(t) = 2 cos(t)i + 2 sin(t)j + t2 k from t = 0 to t = π/2.

Solution The derivative vector along the curve is

r� (t) = −2 sin(t)i + 2 cos(t)j + 2tk

which has magnitude

|r� (t)| = (4 sin2 t + 4 cos2 t + 4t2 )1/2 = 2(1 + t2 )1/2 .

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

10

Thus the length of the curve is

� π/2 � π/2
2 1/2
L(C) = 2(1 + t ) dt = 2 (1 + t2 )1/2 dt
0 0

This integral is soluble by integrating by parts, then using a substitution t = sinh x:

� π/2
L(C) = 2(1 + t2 )1/2 dt
0
� �π/2 � π/2
= 2t(1 + t2 )1/2 − (1 + t2 )−1/2 (2t)t dt
0 0
� π/2
2(t2 + 1 − 1)
= π(1 + π 2 /4)1/2 − dt
0 (1 + t2 )1/2
� � π/2
π/2
2(t2 + 1) 2
= π(1 + π 2 /4)1/2 − 2 1/2
dt + dt
(1 + t ) (1 + t2 )1/2
�0 �� � �0 �� �
L(C) substitution
� arcsinh (π/2)
2 1/2 2
2L(C) = π(1 + π /4) + cosh x dx
0 (1 + sinh2 x)1/2
� arcsinh (π/2)
= π(1 + π 2 /4)1/2 + 2 dx
0
= π(1 + π 2 /4)1/2 + 2 arcsinh (π/2)
� �1/2
π π2 �π�
L(C) = 1+ + arcsinh
2 4 2

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

Chapter 3

SCALAR FUNCTIONS OF
SEVERAL VARIABLES

3.1 Functions of two variables

We think of a function of two variables, f (x, y), as defining a surface z = f (x, y) above the xy plane.
In this section we will look at the properties of that surface, as we can extract them from the function
f (x, y).

3.1.1 Tangent plane

The tangent plane at a point A = (a, b, f (a, b)) is just the linear approximation to the surface at
that point. The tangent plane at A is perpendicular to the normal at A, i.e. is perpendicular to the
gradient vector at A. Now the surface is f (x, y) − z = 0, or F (x, y, z) = const. = 0, whose gradient is
∇F = ∂f ∂f
∂x i + ∂y j − k. Thus the tangent plane r · n = const. at A is
� � � �
∂f ∂f
x +y − z = const. (3.1)
∂x A ∂y A
� � � �
∂f ∂f
This plane passes through A(a, b) so the constant in eq. (3.1) has the value a ∂x +b ∂y − f (a, b)
A A
and thus (3.1) becomes
� � � �
∂f ∂f
z = f (a, b) + (x − a) + (y − b) . (3.2)
∂x A ∂y A

Thus the linear approximation to the surface close to A(a, b) is given by the tangent plane (3.2), so
� � � �
∂f ∂f
f (x, y) ≈ f (a, b) + (x − a) + (y − b) (3.3)
∂x A ∂y A

and we note that we can rewrite this using the gradient ∇f as

f (x, y) = f (a, b) + ((x − a), (y − b)) · (∇f )A . (3.4)

Example
Find the linear approximation to f (x, y) = x2 + y 2 near the point (1, 2).

11

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

12

Solution
If f (x, y) = x2 + y 2 then ∂f /∂x = 2x and ∂f /∂y = 2y. At the point (1, 2) we have f = 5,
∂f /∂x = 2 and ∂f /∂y = 4. Thus

f (x, y) ≈ 5 + (x − 1)2 + (y − 2)4 = 2x + 4y − 5.

3.1.2 Taylor series

The tangent plane equation is a Taylor series, truncated at linear order:
� �� � ��
x−a fx �
f (x, y) ≈ f (a, b) + � .
y−b fy �
A

The next term involves all the partial second derivatives of f :

� �� � �� � �� � �� � �
x−a fx � +1 x−a fxx fxy x−a
� �
f (x, y) ≈ f (a, b) + � .
y−b fy �
A 2 y − b fxy fyy �
A
y−b

The matrix � ��
fxx fxy �
H(a, b) = �
fxy fyy �
A

is called the Hessian matrix of the function f at the point (a, b).
Lecture 4

3.1.3 Critical points of functions of two variables

Suppose we want to understand the shape of our surface z = f (x, y). In most places, the gradient vector
is perpendicular to contour lines and points “uphill”; but what can we say at places where the gradient
is zero?
Points where ∇f = 0 are called critical points and to categorise the behaviour of the function near
them, we use the Taylor series.
We saw that the quadratic-order Taylor series for f (x, y) about a point (a, b) is given by
� �� � �� � �� � �� � �
x−a fx � +1 x−a fxx fxy x−a
� �
f (x, y) ≈ f (a, b) + � .
y−b fy �
A 2 y − b fxy fyy �
A
y−b

At a critical point, the middle term is zero (because ∇f = 0) so we have

� �� � �� � �
1 x−a fxx fxy � x−a 1
f (x, y) ≈ f (a, b) + � = f (a, b) + (x − a)� H(a, b)(x − a).
2 y−b fxy fyy �
A
y − b 2

So how does the function behave as we move a small distance away from the point (a, b)? We have

1
f (x) − f (a) = (x − a)� H(a)(x − a)
2

so the behaviour all depends on the matrix H.

[breakable, enhanced]

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 13

Aside: Eigenvalues and Eigenvectors

Definition Suppose we have a square matrix A. Then we say λ is an eigenvalue of A with corresponding
eigenvector v, if:
Av = λv.

How to find eigenvalues The relation

Av = λv =⇒ Av = λIv =⇒ (A − λI) v = 0.

For this to be true for a nonzero vector v, we need the determinant

|A − λI| = 0.

If A is an n × n matrix, this is an nth-order polynomial to solve for λ. In general it will have n

roots, though some of them may be repeated.
How to find an eigenvector Once we know an eigenvalue λ, we can use the relation

(A − λI) v = 0

to find the eigenvector v by row reduction.

Key facts we will need For a real, symmetric n × n matrix:
• The eigenvalues are real
• The eigenvectors of distinct eigenvalues are mutually orthogonal
• You can use the set of eigenvectors v 1 , v 2 , . . . as a basis for Rn , i.e. any vector x can be
written
x = αv 1 + βv 2 + · · ·
• The determinant of the matrix is the product of the eigenvalues
• The trace of the matrix (the sum of the diagonal elements) is the sum of the eigenvalues.

Now, H (by construction) is a real, symmetric matrix, so we know

• H has real eigenvalues: call them λ1 , λ2
• The corresponding eigenvectors v 1 and v 2 are orthogonal
• The trace of H is the sum λ1 + λ2
• The determinant of H is the product λ1 λ2
Now let’s look at our function approximation. Use the unit eigenvectors of H to decompose

x − a = αv 1 + βv 2

which gives
� �
f (x) − f (a) = 21 (αv 1 + βv 2 )� H(a)(αv 1 + βv 2 ) = 21 (αv 1 + βv 2 )� (αλ1 v 1 + βλ2 v 2 ) = 1
2 α 2 λ 1 + β 2 λ2 .

We can see immediately that:

• If λ1 > 0 and λ2 > 0 then f (x) − f (a) > 0 so the point a is a local minimum of the surface.

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

14

• If λ1 < 0 and λ2 < 0 then f (x) − f (a) < 0 so the point a is a local maximum of the surface.

• If λ1 and λ2 have different signs, the sign of f (x) − f (a) depends on the direction in which we
moved away from a. This is called a saddle point.

We don’t have to determine the eigenvalues to classify the critical points, we just need to find the Hessian
matrix H and its determinant |H|, called the Hessian.

Saddle points occur wherever |H| < 0.

Local minima occur where |H| > 0 and fxx > 0.

Local maxima occur where |H| > 0 and fxx < 0.

If |H| = 0 then at least one eigenvalue is zero and the second-order approximation is not enough to
classify this critical point. We call this a degenerate critical point.

Example Find and classify all the critical points of the function

f (x, y) = x(x2 − 1)y

Solution We start by finding points where ∇f = 0.

∂f ∂f
= 3x2 y − y = y(3x2 − 1) = 2x(x2 − 1)
∂x ∂y

We can satisfy the second of these if either x = 0 or x = ±1. In either of these cases, the bracket
(3x2 − 1) is nonzero so we can only satisfy the first condition by setting y = 0. Our three critical
points are (0, 0), (−1, 0) and (1, 0).
Next we find the Hessian matrix
� � � �
fxx fxy 6xy 3x2 − 1
H= = .
fxy fyy 3x2 − 1 0
� � � � � �
0 −1 0 2 0 2
At (0, 0), H = ; at (−1, 0), H = ; at (1, 0), H = ;
−1 0 2 0 2 0
so all three are saddle points.

Lecture 5

3.2 The Chain Rule

For a function of one variable, f (x), if x also depends on another variable t, then f depends on t and
the chain rule gives
df df dx
= (3.5)
dt dx dt
We now generalise this result for a function of several variables. For f (x, y), suppose x and y depend on
t. Let t increase by ∆t and so x and y increase by ∆x and ∆y. Then from our earlier work on linear
approximations we obtain � � � �
∂f ∂f
f (t + ∆t) ≈ f (t) + ∆x + ∆y (3.6)
∂x ∂y

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 15

so if we rearrange:
� �
� �
∂f ∂f
f (t + ∆t) − f (t) ≈ ∆x + ∆y
∂x ∂y
� � � �
f (t + ∆t) − f (t) ∆x ∂f ∆y ∂f
≈ +
∆t ∆t ∂x ∆t ∂y
then in the limit ∆t → 0 we obtain the chain rule for a function of two variables, which is,
� � � �
df ∂f dx ∂f dy
= + . (3.7)
dt ∂x dt ∂y dt

Note that f depends on x and y [so partial derivatives ∂f /∂x, ∂f /∂y] whilst x and y depend on the
single variable t [so ordinary derivatives dx/dt, dy/dt]. Thus f depends on t and has the derivative
df /dt given by the chain rule (3.7).

Example
If f (x, y) = x2 + y 2 , where x = sin(t), y = t3 , then
� � � �
df ∂f dx ∂f dy
= + = 2x cos(t) + 2y3t2 = 2 sin(t) cos(t) + 6t5 .
dt ∂x dt ∂y dt

Of course in this simple example we can check the result by substituting for x and y before differentiation
to give f (t) = (sin(t))2 + (t3 )2 , so df 5
dt = 2 sin(t) cos(t) + 6t as before.
The chain rule extends directly to functions of three or more variables.
Example
If f (x, y, z) = ln (2x − 3y + 4z), where x = et , y = ln t, z = cosh t, then
� � � � � �
df ∂f dx ∂f dy ∂f dz
= + +
dt ∂x dt ∂y dt ∂z dt
t
2e 3(1/t) 4 sinh t
= − +
2x − 3y + 4z 2x − 3y + 4z 2x − 3y + 4z
2et − 3/t + 4 sinh t
= .
2et − 3 ln t + 4 cosh t

3.2.1 Extended chain rule

For f (x, y) suppose that x and y depend on two variables s and t. Then changing either s or t changes
x and y, so changes f , i.e. producing ∂f ∂f
∂s and ∂t according to the extended chain rule

∂f ∂f ∂x ∂f ∂y ∂f ∂f ∂x ∂f ∂y
= + , = + . (3.8)
∂s ∂x ∂s ∂y ∂s ∂t ∂x ∂t ∂y ∂t
Example
f (x, y) = x2 y 3 , where x = s − t2 , y = s + 2t. Then
∂f ∂f
= 2xy 3 and = 3x2 y 2
∂x ∂y
and
∂f ∂f ∂x ∂f ∂y
= + = 2xy 3 + 3x2 y 2 = (s − t2 )(s + 2t)2 (5s + 4t − 3t2 )
∂s ∂x ∂s ∂y ∂s

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

16

∂f ∂f ∂x ∂f ∂y
= + = 2xy 3 (−2t) + 3x2 y 2 (2)
∂t ∂x ∂t ∂y ∂t
= 2(s − t2 )(s + 2t)2 (3s − 2st − 7t2 ).

Example If f is a function of x and y, where x = es cos t, y = es sin t, prove that sin t ∂f ∂f s ∂f

∂s +cos t ∂t = e ∂y .

Solution If x = es cos t and y = es sin t then

∂x ∂y
= es cos t = es sin t
∂s ∂s
∂x ∂y
= −es sin t = es cos t.
∂t ∂t
It follows that
∂f ∂f s ∂f s
= e cos t + e sin t
∂s ∂x ∂y
∂f ∂f ∂f s
= − es sin t + e cos t
∂t ∂x ∂y

Combining these two equations we have

∂f ∂f ∂f s
sin t + cos t = e .
∂s ∂t ∂y

3.2.2 Matrix form

We can write the extended chain rule:
∂f ∂f ∂x ∂f ∂y ∂f ∂f ∂x ∂f ∂y
= + , = + (3.9)
∂s ∂x ∂s ∂y ∂s ∂t ∂x ∂t ∂y ∂t

in matrix notation as � � � �� �
∂f /∂s ∂x/∂s ∂y/∂s ∂f /∂x
= (3.10)
∂f /∂t ∂x/∂t ∂y/∂t ∂f /∂y

The matrix we have introduced here (well, actually, its transpose) is called the Jacobian matrix of the
transformation between (x, y) and (s, t).

3.2.3 Jacobian
The Jacobian of the transformation x = x(s, t), y = y(s, t) is the determinant
� �
∂(x, y) �� ∂x/∂s ∂y/∂s �� ∂x ∂y ∂x ∂y
=� = − (3.11)
∂(s, t) ∂x/∂t ∂y/∂t � ∂s ∂t ∂t ∂s

of the Jacobian matrix. (Recall that the determinant of a matrix is the same as the determinant of its
transpose, so we don’t need to worry about which version we use for this.)

Example
Find the Jacobian of the coordinate transformation x = s2 /t, y = t2 /s.

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 17

Solution
The matrix we need to find the determinant is
� � � �
∂x/∂s ∂y/∂s 2s/t −t2 /s2
=
∂x/∂t ∂y/∂t −s2 /t2 2t/s
Lecture 6 and this has determinant
� �
∂(x, y) �� 2s/t −t2 /s2 ��
=� = 4 − 1 = 3.
∂(s, t) −s2 /t2 2t/s �

so the Jacobian is 3.

Question: If x = x(s, t) and y = y(s, t), can we solve for s and t in terms of x and y?
∂(x, y)
Answer: Yes, provided that the Jacobian �= 0. This is the implicit function theorem.
∂(s, t)
When this is possible, comparison of the two equations
� � � �� �
∂f /∂s ∂x/∂s ∂y/∂s ∂f /∂x
=
∂f /∂t ∂x/∂t ∂y/∂t ∂f /∂y
� � � �� �
∂f /∂x ∂s/∂x ∂t/∂x ∂f /∂s
=
∂f /∂y ∂s/∂y ∂t/∂y ∂f /∂t
shows us that the Jacobian matrices for a tranformation and its reverse are each other’s inverses:
� � � �−1
∂s/∂x ∂t/∂x ∂x/∂s ∂y/∂s
= .
∂s/∂y ∂t/∂y ∂x/∂t ∂y/∂t
� � � �−1
∂s/∂x ∂s/∂y ∂x/∂s ∂x/∂t
= .
∂t/∂x ∂t/∂y ∂y/∂s ∂y/∂t
It then follows that the determinants – the Jacobians – must also be related as inverses:
∂(x, y) 1
= ∂(s,t)
.
∂(s, t)
∂(x,y)

Exercise Using x = s2 /t and y = t2 /s as above, solve for s and t in terms of x and y and verify that
the Jacobian of the reverse transformation is 1/3.

3.2.4 The Laplacian

From the operator ∇ we can construct other operators. The most important of these is the Laplacian
(or the Laplace operator) ∇2 = ∇ · ∇ which operates on scalar functions f (x, y, z) according to the rule:

∂2f ∂2f ∂2f

∇2 f = ∇ · (∇f ) = 2
+ 2 + 2. (3.12)
∂x ∂y ∂z
Example
Calculate the Laplacian for f (x, y, z) = x2 + y 2 + z 2 .

Solution
∂2f ∂2f ∂2f ∂ ∂ ∂
∇2 f = + + = 2x + 2y + 2z = 2 + 2 + 2 = 6.
∂x2 ∂y 2 ∂z 2 ∂x ∂y ∂z

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

18

3.3 Reference results for polar coordinates

Jacobians
The following Jacobians may be quoted as standard results:

∂(x, y)
= r for plane polar coordinates
∂(r, θ)
∂(x, y, z)
= r for cylindrical polar coordinates
∂(r, θ, z)
∂(x, y, z)
= ρ2 sin θ for spherical polar coordinates.
∂(ρ, θ, φ)

Grad in polar coordinates

The form of grad in polar coordinates is not examinable in this module, but you may find them useful
in later years.
In plane polar coordinates we denote the unit vector in the r direction by er , and the unit vector
in the θ direction by eθ . The values of these two vectors vary according to location. Then we have

∂f 1 ∂f
∇f = er + e
∂r r ∂θ θ
Similarly, for cylindrical polar coordinates, we have
∂f 1 ∂f ∂f
∇f = e + e + e
∂r r r ∂θ θ ∂z z

Spherical polar coordinates

Finally, in spherical polar coordinates we denote the unit vector in the ρ direction as eρ and similar
for θ and φ. Grad is given by:
∂f 1 ∂f 1 ∂f
∇f = e + e + e .
∂ρ ρ ρ ∂θ θ ρ sin θ ∂φ φ

Laplacian in polar coordinates

The Laplacian in Cartesian coordinates is given by

∂2f ∂2f ∂2f

∇2 f = 2
+ 2 + 2.
∂x ∂y ∂z
The forms for the Laplacian in different coordinate systems are more complex. They are given by:
� �
2 1 ∂ ∂f 1 ∂2f
∇ f = r + 2 2 in plane polar coordinates
r ∂r ∂r r ∂θ
� �
1 ∂ ∂f 1 ∂2f ∂2f
∇2 f = r + 2 2 + 2 in cylindrical polar coordinates
r ∂r ∂r r ∂θ ∂z
� � � �
2 1 ∂ 2 ∂f 1 ∂ ∂f 1 ∂2f
∇ f = ρ + sin θ + in spherical polar coordinates
ρ2 ∂ρ ∂ρ ρ2 sin θ ∂θ ∂θ ρ2 sin2 θ ∂φ2

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

Chapter 4

DOUBLE AND TRIPLE

INTEGRALS

4.1 Multiple-Integral Notation

Previously ordinary integrals of the form
� � b
f (x) dx = f (x) dx (4.1)
J a

where J = [a, b] is an interval on the real line, have been studied.

y = f (x)

a b
Here we study double integrals ��
f (x, y) dx dy (4.2)
Ω

where Ω is some region in the xy-plane, and a little later we will study triple integrals
���
f (x, y, z) dx dy dz (4.3)
T

where T is a solid (volume) in the xyz-space.

4.2 Double Integrals

4.2.1 Properties
(1) Area property ��
dx dy = Area of Ω.
Ω
��
In particular if Ω is the rectangle Ω = [a, b] × [c, d] then Ω
dx dy = (b − a)(d − c).

19

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

20

(2) Linearity
�� �� ��
[αf (x, y) + βg(x, y)] dx dy = α f (x, y) dx dy + β g(x, y) dx dy (4.4)
Ω Ω Ω
where α and β are constants.
(3) Additivity
If Ω is broken up into a finite number of nonoverlapping basic regions Ω1 ,. . . ,Ωn , then
�� �� ��
f (x, y) dx dy = f (x, y) dx dy + . . . + f (x, y) dx dy. (4.5)
Ω Ω1 Ωn

4.2.2 Geometric Interpretation

z = f (x, y)

The double integral over Ω gives the volume of the solid T whose upper boundary is the surface
z = f (x, y) and whose lower boundary is the region Ω in the xy-plane:
��
f (x, y) dx dy = volume of T. (4.6)
Ω

4.2.3 The Evaluation of Double Integrals by Repeated Integrals

�b
If an ordinary integral a f (x) dx proves difficult to evaluate, it is not
�� because of the interval [a, b] but
because of the integrand f . Difficulty in evaluating a double integral Ω f (x, y) dx dy can come
��from two
sources: from the integrand f or from the domain Ω. Even such a simple looking integral as Ω 1 dx dy
is difficult to evaluate if Ω is complicated.
In this section we introduce a technique for evaluating double integrals over domains that have
special shapes. The key idea is that double integrals over such special domains can be reduced to a pair
of ordinary integrals.

Horizontally simple domain

The projection of the domain Ω onto the x-axis is a closed interval [a, b]
and Ω consists of all points (x, y) with

a ≤ x ≤ b, and φ1 (x) ≤ y ≤ φ2 (x). (4.7)

Then �� �
a b �� � b φ2 (x)
f (x, y) dx dy = f (x, y) dy dx. (4.8)
Ω a φ1 (x)
� φ2 (x)
Here we first calculate φ1 (x)
f (x, y) dy by integrating f (x, y) with respect to y from y = φ1 (x) to

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 21

y = φ2 (x). The resulting expression is a function of x alone, which we then integrate with respect to x
from x = a to x = b.

Example ��
Evaluate Ω (x4 − 2y) dx dy, where the domain Ω consists of all points (x, y) with −1 ≤ x ≤ 1 and
−x2 ≤ y ≤ x2 .

Lecture 7
Solution
�� � x=1 � y=x2 � x=1 2
4
(x − 2y) dx dy = 4
(x − 2y) dy dx = [x4 y − y 2 ]y=x
y=−x2 dx
Ω x=−1 y=−x2 x=−1
� x=1
= 2x6 dx = [2x7 /7]x=1
x=−1 = 4/7
x=−1

Vertically simple domain

d The projection of the domain Ω onto the y-axis is a closed interval [c, d]
and Ω consists of all points (x, y) with

c c ≤ y ≤ d, and ψ1 (y) ≤ x ≤ ψ2 (y). (4.9)

Then
�� � �� �
d ψ2 (y)
f (x, y) dx dy = f (x, y) dx dy. (4.10)
Ω c ψ1 (y)
� ψ (y)
Here we first calculate ψ12(y) f (x, y) dx by integrating f (x, y) with respect to x from x = ψ1 (y) to
x = ψ2 (y). The resulting expression is a function of y alone, which we then integrate with respect to y
from y = c to y = d.
The integrals in the right-hand sides of formulae (4.8) and (4.10) are called iterated integrals.

Remark 1
Sometimes a domain can be expressed both as a horizontally simple domain: a ≤ x ≤ b, φ1 (x) ≤ y ≤
φ2 (x), and as a vertically simple domain: c ≤ y ≤ d, ψ1 (y) ≤ x ≤ ψ2 (y).

Example
The triangular region with corners at (0, 0), (1, 0) and (1, 1):

�
�
�
�
�
�
�
�
�
�

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

22

can be written either as the horizontally simple region

0≤x≤1 0≤y≤x

or the vertically simple region

0≤y≤1 y ≤ x ≤ 1.

Remark 2

Finally, if Ω, the domain of integration, is neither horizontally nor vertically simple, then it is usually
possible to break it up into a finite number of domains, say Ω1 , . . . , Ωn , each of which is either horizontally
or vertically simple. Then we can use the additivity property given above.

4.2.4 Evaluating Double Integrals Using

Polar Coordinates
Let Ω be a domain formed with all points (x, y) that have polar coordinates (r, θ) in the set

Γ : α ≤ θ ≤ β, ρ1 (θ) ≤ r ≤ ρ2 (θ) (4.11)

where β ≤ α + 2π.
�� ��
f (x, y) dx dy = f (r cos(θ), r sin(θ))r dr dθ
Ω Γ
� β � ρ2 (θ)
= f (r cos(θ), r sin(θ))r dr dθ. (4.12)
α ρ1 (θ)

Recall that the standard conversion from cartesian to polar coordinates is

x = r cos θ, y = r sin θ.

Example ��
Use polar coordinates to evaluate Ω xy dx dy, where Ω is the portion of the unit disc that lies in
the first quadrant.

Solution
The region may be expressed in polar coordinates as 0 ≤ r ≤ 1 and 0 ≤ θ ≤ π/2. Thus we have
�� � π/2 � 1 � π/2
xy dx dy = r cos θr sin θr dr dθ = cos θ sin θ[r4 /4]10 dθ
Ω 0 0 0
� π/2 � sin π/2
= 1
4 cos θ sin θ dθ = 1
4 u du = 41 [u2 /2]10 = 1
8
0 0

in which we used the substitution u = sin θ with du = cos θ dθ.

Example
Calculate the volume of the solid bounded above by the cone z = 2 − (x2 + y 2 )1/2 and bounded
below by the disc Ω : (x − 1)2 + y 2 ≤ 1.

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 23

Solution
The region Ω may be expressed as 0 ≤ r ≤ 2 cos θ and −π/2 ≤ θ ≤ π/2 (we did this exercise in
chapter 1). The integral may be expressed as
��
2 − (x2 + y 2 )1/2 dx dy
Ω

which then becomes

�� � π/2 � 2 cos θ � π/2
2 − (x2 + y 2 )1/2 dx dy = (2 − r)r dr dθ = [r2 − r3 /3]20 cos θ dθ
Ω −π/2 0 −π/2
� π/2
= [4 cos2 θ − 8 cos3 θ/3] dθ
−π/2

We use the fact that cos2 θ = 12 (1 + cos 2θ) and substitute u = sin θ, du = cos θ dθ:
�� � π/2 � π/2
2 − (x2 + y 2 )1/2 dx dy = 4 cos2 θ dθ − 8
3 cos3 θ dθ
Ω −π/2 −π/2
� π/2 � 1
= 2 1 + cos 2θ dθ − 8
3 (1 − u2 )du
−π/2 −1
π/2
= 2[θ + 1
2 sin 2θ]−π/2 − 83 [u − 31 u3 ]1−1 = 2π − 32
9 .

4.3 Triple Integrals

4.3.1 Properties
(1) Volume property
���
dx dy dz = Volume of T.
T
���
In particular if T is the box T = [a, b] × [c, d] × [e, f ] then T
dx dy dz = (b − a)(d − c)(f − e).

(2) Linearity
���
[αf (x, y, z) + βg(x, y, z)] dx dy dz
T
��� ���
=α f (x, y, z) dx dy dz + β g(x, y, z) dx dy dz (4.13)
T T

where α and β are constants.

(3) Additivity If T is broken up into a finite number of nonoverlapping basic regions T1 ,. . . ,Tn , then
��� ��� ���
f (x, y, z) dx dy dz = f (x, y, z) dx dy dz + . . . + f (x, y, z) dx dy dz. (4.14)
T T1 Tn

Lecture 8

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

24

4.3.2 The Evaluation of Triple Integrals by Repeated Integrals

Let T be a solid whose projection onto the xy-plane is labelled Ωxy . Then the solid T is the set of all
points (x, y, z) satisfying
(x, y) ∈ Ωxy , χ1 (x, y) ≤ z ≤ χ2 (x, y). (4.15)
The triple integral over T can be evaluated by setting
��� �� �� �
χ2 (x,y)
f (x, y, z) dx dy dz = f (x, y, z) dz dx dy. (4.16)
T Ωxy χ1 (x,y)

In eq. (4.16) we can evaluate the integration with respect to z first and then evaluate the double integral
over the domain Ωxy as studied for double integrals. In particular if Ωxy is horizontally simple, say
a ≤ x ≤ b, φ1 (x) ≤ y ≤ φ2 (x). (4.17)
then the solid T itself is the set of all points (x, y, z) such that
a ≤ x ≤ b, φ1 (x) ≤ y ≤ φ2 (x), χ1 (x, y) ≤ z ≤ χ2 (x, y) (4.18)
and the triple integral over T can be expressed by three ordinary integrals as:
��� � b �� φ2 (x) �� χ2 (x,y) � �
f (x, y, z) dx dy dz = f (x, y, z) dz dy dx. (4.19)
T a φ1 (x) χ1 (x,y)

Here we first integrate with z [from z = χ1 (x, y) to z = χ2 (x, y)], then with respect to y [from y = φ1 (x)
to y = φ2 (x)], and finally with respect to x [from x = a to x = b].
There is nothing special about this order of integration. Other orders of integration are possible and
in some cases more convenient. Suppose for example that the projection of T onto the xz-plane is a
domain Ωxz of the form
z1 ≤ z ≤ z2 , φ1 (z) ≤ x ≤ φ2 (z). (4.20)
If T is the set of all (x, y, z) with
z1 ≤ z ≤ z2 , φ1 (z) ≤ x ≤ φ2 (z), ψ1 (x, z) ≤ y ≤ ψ2 (x, z) (4.21)
then ��� � �� �� � �
z2 φ2 (z) ψ2 (x,z)
f (x, y, z) dx dy dz = f (x, y, z) dy dx dz. (4.22)
T z1 φ1 (z) ψ1 (x,z)

In this case we integrate first with respect to y, then with respect to x, and finally with respect to z.
Still four other orders of integration are possible.
Example
� 2 � x � 4−x2
Evaluate the triple integral 0 0 0 xyz dz dy dx.
Solution
� 2 � x � 4−x2 � 2 � x � 4−x2 � 2 � x
2
xyz dz dy dx = x y z dz dy dx = x y[ 21 z 2 ]4−x
0 dy dx
0 0 0 0 0 0 0 0
� 2 � x � 2
2 2
= 1
2 x(4 − x ) y dy dx = 1
2 x(4 − x2 )2 [ 12 y 2 ]x0 dx
0 0 0
� 2 � 2
= 1
4 x3 (4 − x2 )2 dx = 1
4 16x3 − 8x5 + x7 dx
0 0
4
= 1
4 [4x − 4x6 /3 + x8 /8]20 = 14 [64 − 256/3 + 32] = 8/3

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 25

Example
Use triple integration to find the volume of the tetrahedron T shown in the figure.

z
✻
(0, 0, 1)
✄❅
✄ ❅
✄ ❅
✄ ❅
✄ ❅ (0, 1, 0)
✄ ✔ ✟ ✲
✟
✄ ✔ ✟✟✟
y
✄✔ ✟
✄ ✟
✟
✔
x (1, 0, 0)
✔
☛

Solution
The region of integration is bounded above by the plane x + y + z = 1 so we need

0≤z ≤1−x−y

Then the limits on x and y are that they are both positive, and that the z interval above should
exist:
x≥0 y≥0 1 − x − y ≥ 0.

The y limits can therefore be written

0≤y ≤1−x

and the upper x limit is defined by the requirement that this y interval exists: 1 − x ≥ 0 so

0 ≤ x ≤ 1.

The volume may then be expressed as

� 1 � 1−z � 1−z−y � 1 � 1−z
Volume of T = dx dy dz = (1 − z − y) dy dz
z=0 y=0 x=0 z=0 y=0
� 1 � 1
1−z
= [(1 − z)y − 21 y 2 ]y=0 dz = 1
2 (1 − z)
2
dz
z=0 z=0
� 1
= 1
2 (z 2 − 2z + 1) dz = 12 [ 13 z 3 − z 2 + z]1z=0 = 16 .
z=0

Meaning
We have seen that integrating 1 over a region gives its volume. But what about integrating a general
function?
If we consider the function f (x, y, z) to represent the density of our object at each point inside it,
then ���
f (x, v, z) dx dy dz
T

gives the mass of the object.

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

26

4.3.3 Evaluating Triple Integrals Using Cylindrical Coordinates

Let T be a solid whose projection onto the xy-plane is labelled Ωxy . Then the solid T is the set of all
points (x, y, z) satisfying
(x, y) ∈ Ωxy , χ1 (x, y) ≤ z ≤ χ2 (x, y). (4.23)
The domain Ωxy has polar coordinates in some set Ωrθ and then the solid T in cylindrical coordinates
is some solid S satisfying

(r, θ) ∈ Ωrθ , χ1 (r cos(θ), r sin(θ)) ≤ z ≤ χ2 (r cos(θ), r sin(θ)). (4.24)

Then
��� �� �� �
χ2 (x,y)
f (x, y, z) dx dy dz = f (x, y, z) dz dx dy
T Ωxy χ1 (x,y)
�� �� �
χ2 (r cos(θ),r sin(θ))
= f (r cos(θ), r sin(θ), z) dz r dr dθ =
Ωrθ χ1 (r cos(θ),r sin(θ))
���
f (r cos(θ), r sin(θ), z)r dr dθ dz. (4.25)
S

Example
� 2 � (4−x2 )1/2 � 4−x2 −y2
Evaluate using cylindrical coordinates the triple integral −2 −(4−x2 )1/2 0
(x2 +y 2 ) dz dy dx.

Solution
The region

0< z < 4 − x2 − y 2
2 1/2
−(4 − x ) < y < (4 − x2 )1/2
−2 < x <2

is the same as the region

0< z < 4 − r2
r2 < 4.

Substituting in the polar coordinates we have

��� � 2π � 2 � 4−r 2 � 2π � 2
2
(x2 + y 2 ) dz dy dx = r3 dz dr dθ = [r3 z]4−r
z=0 dr dθ
0 0 0 0 0
� 2π � 2 � 2π
= r3 (4 − r2 ) dr dθ = [r4 − r6 /6]20 dθ
0 0 0
� 2π
= [16 − 32/3] dθ = 2π[16 − 32/3] = 32π/3
0

Lecture 9
Example
Use cylindrical coordinates to find the volume of the solid T bounded above by the plane z = y
and below by the paraboloid z = x2 + y 2 .

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 27

Solution
The region x2 + y 2 < z < y only exists where x2 + y 2 < y, i.e. r2 < r sin θ, which means r < sin θ
which can only happen in the quadrants where sin θ is positive, 0 ≤ θ ≤ π. Thus the integral may
be written as � � � π sin θ r sin θ
VT = r dz dr dθ
θ=0 r=0 z=r 2

which is easy to solve by repeated integration, giving the final answer VT = π/32.

4.3.4 Evaluating Triple Integrals Using Spherical Coordinates

Let T be a solid in xyz-space with spherical coordinates in the solid S of ρθφ-space. Then, recalling the
coordinate transformation

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

we can write
��� ���
f (x, y, z) dx dy dz = f (ρ sin θ cos φ, ρ sin θ sin φ, ρ cos θ) ρ2 sin θdρ dθ dφ. (4.26)
T S

Example
Find the volume of the solid T that is bounded above by the cone z 2 = x2 + y 2 , below by the
xy-plane, and on the sides by the hemisphere z = (4 − x2 − y 2 )1/2 .

Solution
We saw this region in chapter 1. It can be expressed as

0 ≤ ρ ≤ 2, π/4 < θ < π/2 and 0 ≤ φ < 2π.

��� � π/2 � 2 � 2π � π/2 � 2

2 2
ρ sin θ dρ dθ dφ = sin θ ρ dφ dρ dθ = 2π sin θ ρ2 dρ dθ
T π/4 0 0 π/4 0
� π/2 � π/2
= 2π sin θ[ρ3 /3]20 dθ = 16
3 π sin θ dθ
π/4 π/4
16 π/2 16
√
= 3 π[− cos θ]π/4 = 3 π(2
−1/2
) = 38 π 2

4.4 Jacobians and changing variables in multiple integration

During the course of the last few sections you have met several formulae for changing variables in multiple
integration: to polar coordinates, to cylindrical coordinates, to spherical coordinates. The purpose of
this section is to bring some unity to that material and provide a general description for other changes
of variable.

4.4.1 Change of variables for double integrals

Consider the change of variables x = x(u, v) and y = y(u, v), which maps the points (u, v) of some
domain Γ into the points (x, y) of some other domain Ω. Then
�� � �
� ∂(x, y) �
The area of Ω = � � du dv. (4.27)
Γ ∂(u, v)
� �

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

28

Suppose now that we want to integrate some function f (x, y) over Ω. If this proves difficult to do directly,
then we can change variables (x, y) to (u, v) and try to integrate over Γ instead. Then
�� �� � �
� ∂(x, y) �
f (x, y) dx dy = f (x(u, v), y(u, v)) �� � du dv. (4.28)
Ω Γ ∂(u, v) �
Note: since the Jacobian ∂(x, y)/∂(u, v) is already the determinant, the bars here mean take the absolute
value of the Jacobian.

Example ��
Evaluate the integral Ω xy dx dy, where Ω is the first-quadrant region bounded by the curves
x2 + y 2 = 4, x2 + y 2 = 9, x2 − y 2 = 1 and x2 − y 2 = 4.

Solution
The region Ω consists of the intersection of x > 0 and y > 0 with the four conditions above. We
try the substitution u = x2 + y 2 , v = x2 − y 2 . The Jacobian is
� �
∂(u, v) �� 2x 2x ��
= = −8xy
∂(x, y) � 2y −2y �
and hence the inverse Jacobian is
∂(x, y)
= − 81 x−1 y −1 .
∂(u, v)
The integral becomes
�� � 9 � 4
1
xy dx dy = 8 du dv = 18 [(9 − 4)(4 − 1)] = 15
8 .
Ω u=4 v=1

4.4.2 Change of variables for triple integrals

Consider the change of variables x = x(u, v, w), y = y(u, v, w), z = z(u, v, w) which maps the points
(u, v, w) of some solid S into the points (x, y, z) of some other solid T . Then
��� � �
� ∂(x, y, z) �
The volume of T = � � du dv dw. (4.29)
S ∂(u, v, w)
� �

Suppose now that we want to integrate some function f (x, y, z) over T . If this proves difficult to do
directly, then we can change variables (x, y, z) to (u, v, w) and try to integrate over S instead. Then
��� ��� � �
� ∂(x, y, z) �
f (x, y, z) dx dy dz = f (x(u, v, w), y(u, v, w), z(u, v, w)) �
� � du dv dw. (4.30)
T S ∂(u, v, w) �
Referring back to equations (4.25) and (4.26), and the Jacobians given at the end of §3.2.3, we can
verify that this formula is correct for a change from Cartesian to cylindrical coordinates (Jacobian is r)
and for a change from Cartesian to spherical coordinates (Jacobian is ρ2 sin θ).

Example
Calculate the volume of the ellipsoid x2 /a2 + y 2 /b2 + z 2 /c2 ≤ 1.

Solution
We change variables to set u = x/a, v = y/b and w = z/c. This has Jacobian ∂(x, y, z)/∂(u, v, w) =
abc. Then ��� ���
dx dy dz = abc du dv dw.
T sphereofradius1

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 29

We can then either use the volume of a sphere of radius 1 (which is 4π/3) or calculate it using a
second change of variables is into spherical coordinates relative to the Cartesian axes u, v and w
(similar to the final exercise on example sheet 3).
The final result is
4
Volume = πabc.
3
Lecture 10

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

Chapter 5

VECTOR OPERATORS

We’ve looked at vector functions of a single variable, r(t). We’ve looked at scalar functions of several
variables, f (x). To complete the set, in this final chapter we will consider vector fields: vector functions
of several variables F (x) = F1 i + F2 j + F3 k, each of whose components Fi depends on position.

5.1 The vector differential operator ∇ (grad)

The vector differential operator ∇ is defined formally by setting
� �
∂ ∂ ∂
∇= , , . (5.1)
∂x ∂y ∂z

By ‘formally’ we mean that this is not an ordinary vector but its ‘components’ are differentiation symbols.
As the term ‘operator’ suggests, ∇ is thought of as something that ‘operates’ on things. What sort of
things? Scalar functions and vector functions.
Suppose that f (x, y, z) is a scalar function. Then ∇ operates on f as follows:
� � � �
∂ ∂ ∂ ∂f ∂f ∂f
∇f = , , f= , , . (5.2)
∂x ∂y ∂z ∂x ∂y ∂z

This is the gradient of f , which we have already seen.

Question How does ∇ operate on vector functions?

Answer Using either of the vector products we already know – dot or cross.

5.1.1 Divergence
If q(x, y, z) = (q1 (x, y, z), q2 (x, y, z), q3 (x, y, z)) is a vector function, then by definition

∂q1 ∂q2 ∂q3

div (q) = ∇ · q = + + . (5.3)
∂x ∂y ∂z
This “product”, ∇ · q, defined in imitation of the ordinary dot product, is a scalar called the divergence
of q.

Example
Calculate ∇ · q for the vector function q(x, y, z) = (x − y, x + y, z).

30

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 31

Solution
∂(x − y) ∂(x + y) ∂z
∇·q = + + = 1 + 1 + 1 = 3.
∂x ∂y ∂z

5.1.2 Curl
For the vector q = q1 i + q2 j + q3 k, we also have
 
i j k
∂ ∂ ∂
curl (q) = ∇ × q = det ∂x ∂y ∂z

q1 q2 q3
� �
∂q3 ∂q2 ∂q1 ∂q3 ∂q2 ∂q1
= − , − , − . (5.4)
∂y ∂z ∂z ∂x ∂x ∂y

This second “product”, ∇ × q, defined in imitation of the ordinary cross product, is a vector called the
curl of q.

Example
Calculate ∇ × q for (from above) q(x, y, z) = (x − y, x + y, z).

Solution � �
� i j k �
� �
∇ × q = �� ∂/∂x ∂/∂y ∂/∂z � = i(0) + j(0) + k(1 + 1) = 2k.
�
� x−y x+y z �

5.1.3 Helmholtz decomposition

Any sufficiently nice vector field F (x) can be written as the sum of a gradient part and a curl part:

F = ∇f + ∇ × A.

The functions f and A are called the scalar potential and vector potential respectively.

Question How can we tell whether a given vector field is the gradient of a scalar function?
Answer Suppose it is:
∂f ∂f ∂f
F1 = F2 = F3 =
∂x ∂y ∂z
Then, using the mixed derivatives theorem, we can show that

∂F1 ∂2f ∂F2 ∂F2 ∂2f ∂F3 ∂F3 ∂2f ∂F1

= = = = = =
∂y ∂x∂y ∂x ∂z ∂y∂z ∂y ∂x ∂z∂x ∂z
which is equivalent to saying
∇ × F = 0.
In fact this is sufficient: a sufficiently smooth vector field which has zero curl is the gradient of a
scalar function.

A vector field which has zero curl is called irrotational; a vector field which has zero divergence is
called solenoidal.
If we have an irrotational vector field, we can find the corresponding scalar potential by repeated
integration (remembering to treat the “constant of integration” properly when using partial derivatives).

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

32

Example Find the scalar potential corresponding to the vector field

F = (2x + z)i + (2y + z)j + (x + y + z)k

Solution We have three simultaneous equations:

∂f ∂f ∂f
= 2x + z = 2y + z = x + y + z.
∂x ∂y ∂z
Integrating the first gives
f (x, y, z) = x2 + xz + c1 (y, z)
which we can then differentiate and substitute into the other two:
∂c1 ∂c1
= 2y + z x+ = x + y + z.
∂y ∂z

Integrate the first (noting that c1 has no x dependence so the new “constant” can only depend on
z):
c1 (y, z) = y 2 + yz + c2 (z)
and differentiate to substitute into the final equation:
∂c2
x+y+ =x+y+z c2 = 21 z 2 + C
∂z
where now C is a genuine (undetermined) constant. Putting it all back together,

F = ∇f where f = x2 + xz + y 2 + yz + 12 z 2 + C.

Lecture 11

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

Chapter 6

LINE AND SURFACE

INTEGRALS

6.1 Line integrals

Let h(x, y, z) = (h1 (x, y, z), h2 (x, y, z), h3 (x, y, z)) be a vector function that is continuous over a smooth
curve C parametrised by C : r(u) = (x(u), y(u), z(u)) with u ∈ [a, b]. The line integral of h over C is the
number � � b
h(r) · dr = [h(r(u)) · r� (u)] du. (6.1)
C a

Although we stated this definition in terms of three-dimensional vectorial functions h(x, y, z) and curves
in space r(u) = (x(u), y(u), z(u)), it also includes the two-dimensional case: h(x, y) and plane curves
r(u) = (x(u), y(u)).

Example �
Calculate C h(r) · dr given that h(x, y) = (xy, y 2 ) and the plane curve C is parametrised by
r(u) = (u, u2 ) with u ∈ [0, 1].

Solution
The derivative of the vector r is
r� (u) = (1, 2u)
and on the path of integration, x = u and y = u2 so our integral becomes
� � 1 � 1
2
h(r) · dr = (xy, y ) · r (u) du = �
(u3 , u4 ) · (1, 2u) du
C u=0 u=0
� 1
= (u3 + 2u5 ) du = [u4 /4 + 2u6 /6]10 = 7
12 .
u=0

If the curve C is not smooth but is made up of a finite number of adjoining smooth pieces C1 , . . . , Cn ,
i.e. it is piecewise smooth,
� � then we define
� the integral over C as the sum of the integrals over Ci for
i = 1, . . . , n, that is C = C1 + · · · + Cn . All polygonal paths are piecewise smooth.
In the next example we integrate over a triangle. We do this by integrating over each side and then
adding up the results. Observe that the directed line segment that begins at a = (a1 , a2 ) and ends at
b = (b1 , b2 ) can be parametrised by setting r(u) = (1 − u)a + ub with u ∈ [0, 1].

33

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

34

Example �
Evaluate the line integral C h(r) · dr, if h(x, y) = (ey , − sin(πx)) and C is the triangle with vertices
(1, 0), (0, 1), (−1, 0) traversed anticlockwise.

Solution
We parametrise the three lines of the triangle:

r1 (u) = (1 − u)(1, 0) + u(0, 1) = (1 − u, u)

r2 (v) = (1 − v)(0, 1) + v(−1, 0) = (−v, 1 − v)
r3 (w) = (1 − w)(−1, 0) + w(1, 0) = (2w − 1, 0)

which are covered in the order above, with u, v, w all increasing over the range [0, 1]. The derivative
vectors are

r�1 (u) = (−1, 1)

r�2 (v) = (−1, −1)
r�3 (w) = (2, 0)

so our integral is
� � � �
h(r) · dr = h(r) · dr + h(r) · dr + h(r) · dr
C C1 C2 C3
� 1 � 1
= (eu , − sin(π(1 − u))) · (−1, 1) du + (e(1−v) , − sin(π(−v))) · (−1, −1) dv
0 0
� 1
+ (1, − sin(π(2w − 1))) · (2, 0) dw
0
� 1 � 1 � 1
= −eu − sin(π(1 − u)) du + −e(1−v) + sin(π(−v)) dv + 2 dw
0 0 0
� �1 � �1
cos(π(1 − u))
u (1−v) cos(πv)
= −e + + e − + [2w]10
π 0 π 0
= 4 − 2e − 4/π.

When we integrate over a parametrised curve, we integrate in the direction determined by the
parametrisation.
� � If we integrate in the opposite direction, our answer is altered by a factor of −1,
that is −C = − C .

6.1.1 Another notation for line integrals

If h(x, y, z) = (h1 (x, y, z), h2 (x, y, z), h3 (x, y, z)) then the line integral over a curve C can be written as
� �
h(r) · dr = {h1 (x, y, z) dx + h2 (x, y, z) dy + h3 (x, y, z) dz} . (6.2)
C C

Example �
Evaluate C {x2 y dx + xy dy}, where C is the straight-line path connecting (1, 0) to (0, 1).

Solution
The straight-line path in question is given by

r(t) = (1 − t)(1, 0) + t(0, 1) = (1 − t, t)

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 35

with t ∈ [0, 1], whose derivative vector is

(dx/dt, dy/dt) = r� (t) = (−1, 1).

Thus our integral is

� � � 1
{x2 y dx + xy dy} = (x2 y, xy) · dr = ((1 − t)2 t, (1 − t)t) · (−1, 1) dt
C C t=0
� 1
= t − t dt = [t /3 − t4 /4]10 =
2 3 3 1
3 − 1
4 = 1
12
t=0

Lecture 12

6.2 The Fundamental Theorem for Line Integrals

In general, if we integrate a vector function h from one point to another, the value of the line integral
depends on the path chosen. There is, however, an important exception. If the vector function h is
a gradient, i.e. there exists a scalar function f such that h = ∇f , then the value of the line integral
depends only on the endpoints of the path and not on the path itself. The details are spelled out in the
following theorem.

Theorem
Let C, parametrised by r = r(u) with u ∈ [a, b], be a piecewise smooth curve that begins at
α = r(a) and ends at β = r(b). Then if the vector function h is a gradient, i.e. h = ∇f , we have
� �
h(r) · dr = ∇f (r) · dr = f (β) − f (α). (6.3)
C C

NOTE: It is important to see that this result is an extension of the fundamental theorem of
�b
integral calculus: a f � (x) dx = f (b) − f (a).

Corollary
If the curve C is closed, i.e. α = β, then f (α) = f (β) and
�
∇f (r) · dr = 0.
C

Example
Integrate the vector function h(x, y) = (y 2 , 2xy − e2y ) over the circular arc C : r(u) = (cos u, sin u)
with u ∈ [0, π/2].

Solution
We note that h is a gradient:

h = (y 2 , 2xy − e2y ) = ∇(xy 2 − 21 e2y ).

Thus we only need to look at the endpoints:

�
(0,1)
h · dr = [xy 2 − 21 e2y ](1,0) = 12 (1 − e2 ).
C

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

36

6.3 Green’s Theorem

If P (x, y) and Q(x, y) are scalar functions defined over a domain Ω with piecewise smooth closed boundary
C, then
�� � � �
∂Q ∂P
(x, y) − (x, y) dx dy = P (x, y) dx + Q(x, y) dy (6.4)
Ω ∂x ∂y C

where the integral on the right is a line integral over C taken in the anticlockwise direction.
�
Remark As indicated, the symbol is used to denote the line integral over a simple closed curve C
taken in the anticlockwise direction. The quantity
�
h(r) · dr
C

is called the circulation of h around C.

Example �
Use Green’s theorem to evaluate C (3x2 + y) dx + (2x + y 3 ) dy, where C is the circle x2 + y 2 = a2 .

Solution
In order to fit into the format of Green’s theorem, we need

P (x, y) = 3x2 + y and Q(x, y) = 2x + y 3 .

This gives
∂Q ∂P
= 2 and =1
∂x ∂y
and so
� ��
(3x2 + y) dx + (2x + y 3 ) dy = [2 − 1] dx dy
C x2 +y 2 <a2
� a � 2π
= r dθ dr = 2π[r2 /2]a0 = πa2 .
r=0 θ=0

6.3.1 The 2D Divergence (Gauss) Theorem

We start from Green’s Theorem:
�� � � �
∂Q ∂P
(x, y) − (x, y) dx dy = P (x, y) dx + Q(x, y) dy. (6.5)
Ω ∂x ∂y C

If we introduce a vector field q = (Q, −P ) we obtain the divergence theorem in two dimensions:
Let Ω be a two-dimensional domain bounded by a piecewise smooth closed curve C. Then for any
(continuously differentiable) vector function q(x, y) we have that
�� �
(∇ · q) dx dy = (q · n) ds (6.6)
Ω C

where n is the outer unit normal and the integral on the right is taken with respect to arc length (which
we will see next time)
Lecture 13

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 37

6.4 Line integrals with respect to arc length

Suppose that f is a scalar function continuous on a piecewise smooth curve C parametrised by r = r(u)
with u ∈ [a, b]. If s(u) is the length of the curve from the tip of r(a) to the tip of r(u), then, as we have
seen in section 2.2.3, s� (u) = |r� (u)|. The integral of f over C with respect to arc length s is defined by
setting
� � b
f (r) ds = f (r(u))s� (u) du. (6.7)
C a
If the curve represents the position of a wire, and the function f is its mass per unit length, then the
integral w.r.t. arc length gives us the total mass of the wire.

Example Ants are distributed on the spiral

r(θ) = (2θ cos θ, 2θ sin θ), 0 ≤ θ ≤ 4π

with density (ants per unit length)

a(θ) = θ.
Find the total number of ants on the curve.

Solution We use an integral with respect to arc length:

� � 4π
Ants = a(θ) ds = θ |r� (θ)| dθ
C 0

The derivative vector is

r� (θ) = (2 cos θ − 2θ sin θ, 2 sin θ + 2θ cos θ)

� �1/2
|r� (θ)| = 2 (cos θ − θ sin θ)2 + (sin θ + θ cos θ)2
� �1/2 � �1/2
= 2 cos2 θ − 2θ sin θ cos θ + θ2 sin2 θ + sin2 θ + 2θ sin θ cos θ + θ2 cos2 θ = 2 1 + θ2

so � � �4π
4π � �
2 1/2 2� �3/2 2 �� �3/2 �
Ants = 2θ 1 + θ dθ = 1 + θ2 = 1 + 16π 2 −1 .
0 3 0 3

Example: 2D Divergence Theorem

Verify the 2D divergence theorem for the vector function q = xi and the region

Ω : x2 + y 2 ≤ a2 .

Solution Recall the 2D divergence theorem states

�� �
(∇ · q) dx dy = (q · n) ds (6.8)
Ω C

Let’s calculate the left hand side first. We have

∂x ∂0
∇·q = + =1
∂x ∂y
so �� ��
(∇ · q) dx dy = 1 dx dy = Area of Ω = πa2 .
Ω Ω

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

38

Now the right hand side. The curve C is a circle of radius a traversed anticlockwise:

C : r(t) = a cos ti + a sin tj 0 ≤ t < 2π.

The arc length element is

r� (t) = −a sin ti + a cos tj |r� (t)| = a ds = a dt.

We substitute x = a cos t into the definition of q and note that the outer unit normal to a circle is
just the unit radial vector, to obtain

q = a cos ti n = cos ti + sin tj q · n = a cos2 t.

Then the line integral becomes

� � 2π � 2π � �2π
2 2 1 a2 1 a2
(q · n) ds = a cos t a dt = a (1 + cos 2t) dt = t + sin 2t = (2π) = πa2 .
C 0 0 2 2 2 0 2

These two results are the same, as required.

6.5 Parametrised Surfaces; Surface Area

We have seen that a space curve C can be parametrised by a vector function r = r(u) where u ranges
over some interval I of the u-axis. In an analogous manner, we can parametrise a surface S in space by
a vector function r = r(u, v) where (u, v) ranges over some domain Ω of the uv-plane.

Example (The graph of a function)

The graph of a function y = f (x), x ∈ [a, b] can be parametrised by setting r(u) = (u, f (u)),
u ∈ [a, b].
Similarly, the graph of a function z = f (x, y) with (x, y) ∈ Ω can be parametrised by setting
r(u, v) = (u, v, f (u, v)), (u, v) ∈ Ω.

Example (A plane)
If two vectors a and b are not parallel, then the set of all combinations ua + vb generates a plane
P0 that passes through the origin. We can parametrise this plane by setting r(u, v) = ua + vb,
with u and v real numbers.
The plane P that is parallel to P0 and passes through the tip of a vector c can be parametrised by
setting r(u, v) = ua + vb + c, with u and v real numbers.

Lecture 14
Example (A sphere)
The sphere of radius a centred at the origin can be parametrised by setting

r(u, v) = (a sin(u) cos(v), a sin(u) sin(v), a cos(u)), (6.9)

(u, v) ∈ [0, π] × [0, 2π).

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 39

6.5.1 The fundamental vector product

Let S be a surface parametrised by r = r(u, v), (u, v) ∈ Ω. The cross product

N = r�u × r�v (6.10)

is called the fundamental vector product of the surface S.

The vector N (u, v) is perpendicular to the surface S at the point with position vector r(u, v) and, if
different from zero, can be taken as the normal to the surface S at that point.

Example
For the plane r(u, v) = ua + vb + c, the vector a × b is normal to the plane.

Example
The fundamental vector product for a sphere is parallel to the radius vector r(u, v). (Using the
parametrisation given above, N = a sin(u)r.)

6.5.2 The area of a parametrised surface

The area of a surface S parametrised by r = r(u, v), (u, v) ∈ Ω, is given by
��
Area of S = |N (u, v)| du dv. (6.11)
S

Example (The surface area of a sphere)

Using the parametrisation given by equation (6.9), we had N = a sin(u)r so |N | = a2 sin(u) and
the area is
�� � π � 2π
|N (u, v)| du dv = a2 sin(u) dv du = 2πa2 [− cos(u)]π0 = 4πa2 .
S u=0 v=0

Example (The area of a plane domain)

A plane domain may be parameterised as r = (u, v, 0) for (u, v) ∈ Ω. Then r�u = (1, 0, 0) and
r�v = (0, 1, 0) and so the fundamental vector product is N = (0, 0, 1) which has magnitude 1.
��
1 du dv = Area of Ω.
Ω

6.5.3 The area of a surface z = f (x, y)

Let the surface S be the graph of the function z = f (x, y) with (x, y) ∈ Ω. Then

� � �� �2 � �2 �1/2
∂f ∂f
Area of S = + +1 dx dy. (6.12)
Ω ∂x ∂y

In this case the parametrisation of S is

r(u, v) = (u, v, f (u, v)), (u, v) ∈ Ω

and therefore N = (−fx , −fy , 1). The unit vector n = N / |N | is called the upper unit normal because it
points upwards.

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

40

Example
Find the surface area of that part of the parabolic cylinder z = y 2 that lies over the triangle with
vertices (0, 0), (0, 1), (1, 1) in the xy-plane.
Solution
The surface is the graph of the function f (x, y) = y 2 , and the plane region Ω is 0 ≤ y ≤ 1 and
0 ≤ x ≤ y. Using equation (6.12) we have
�� � 1 � 1
Area of S = [(2y)2 + 1]1/2 dx dy = (1 + 4y 2 )1/2 [x]yx=0 dy = y(1 + 4y 2 )1/2 dy
Ω y=0 y=0
� 1
= 1
8 8y(1 + 4y 2 )1/2 dy = 18 [ 32 (1 + 4y 2 )3/2 ]1y=0 = 1
12 [5
3/2
− 1].
y=0

6.6 Surface Integrals

Let H(x, y, z) be a scalar function, continuous over a surface S parametrised by r = r(u, v), (u, v) ∈ Ω.
The surface integral of H over S is the number
�� ��
H(x, y, z) dσ = H(r(u, v)) |N (u, v)| du dv. (6.13)
S Ω

Taking H ≡ 1 and referring back to eq. (6.11) we get

��
dσ = Area of S. (6.14)
S

Example ��
If a = (a1 , a2 , a3 ) and b = (b1 , b2 , b3 ), calculate S xy dσ, where S : r(u, v) = ua + vb with
(u, v) ∈ [0, 1] × [0, 1].
Solution
The fundamental vector product is N = r�u × r�v = a × b. Thus
� � � 1� 1
xy dσ = (a1 u + b1 v)(a2 u + b2 v) |a × b| du dv
S 0 0
� 1 � 1
= |a × b| (a1 a2 u2 + (a1 b2 + a2 b1 )uv + b1 b2 v 2 ) du dv
0 0
� 1
= |a × b| [ 31 a1 a2 u3 + 21 (a1 b2 + a2 b1 )u2 v + b1 b2 uv 2 ]1u=0 dv
0
� 1
= |a × b| ( 31 a1 a2 + 12 (a1 b2 + a2 b1 )v + b1 b2 v 2 ) dv
0
= |a × b| [ 31 a1 a2 v + 41 (a1 b2 + a2 b1 )v 2 + 13 b1 b2 v 3 ]1v=0
= |a × b| ( 31 (a1 a2 + b1 b2 ) + 41 (a1 b2 + a2 b1 ))

Lecture 15
6.6.1 Flux of a vector function
Let q(x, y, z) be a vector function that is continuous over a smooth surface S parametrised by r = r(u, v),
(u, v) ∈ Ω. The flux of q across S in the direction of the unit normal n to the surface S is the number
��
q · n dσ (6.15)
S

Downloaded by mick dundee ([email protected])

 lOMoARcPSD|3471321

MATH0011: Multivariable Calculus 41

which can be calculated as

�� �� ��
q · n dσ = q(r(u, v)) · n ||N || du dv = q(r(u, v)) · N du dv. (6.16)
S Ω Ω
Example
Calculate the flux of the vector function q(x, y, z) = (x, y, 0) across the sphere S: x2 + y 2 + z 2 = a2
in the outward direction.
Solution
We parametrise the sphere using equation (6.9) which gives N = a sin ur and so the scalar function
q · N = a sin u(x, y, 0) · (x, y, z) = a sin u(x2 + y 2 ) = a3 sin3 u. Thus
� 2π � π � 2π � π
3 3 3
Flux = a sin u du dv = a sin u(1 − cos2 u) du dv
v=0 u=0 v=0 u=0
� 2π � −1 � 2π � 1
= a3 (1 − c2 )(−1) dc dv = a3 (1 − c2 ) dc dv
v=0 c=1 v=0 c=−1
� 2π � 2π
= a3 [(c − c3 /3)]1c=−1 dv = a3 (4/3) dv = 38 πa3 .
v=0 v=0
in which we used the substitution c = cos u.

Proposition
If S is the graph of a function z = f (x, y) with (x, y) ∈ Ω and n is the upper unit normal, the the flux
of the vector function q = (q1 (x, y, z), q2 (x, y, z), q3 (x, y, z)) across S in the direction of n is
�� ��
q · n dσ = (−q1 fx − q2 fy + q3 ) dx dy. (6.17)
S Ω

Proof
We can parametrise the surface by r = (u, v, f (u, v)) with (u, v) ∈ Ω. Then the fundamental vector
product is
N = r�u × r�v = (1, 0, fu ) × (0, 1, fv ) = (−fu , −fv , 1)
and we have
�� ��
q · n dσ = (q · N ) du dv
S
� �٠��
= (−q1 fu − q2 fv + q3 ) du dv = (−q1 fx − q2 fy + q3 ) dx dy.
Ω Ω
where we have simply changed the names of the variables at the end.
Example
Let S be the portion of the paraboloid z = 1 − x2 − y 2 that lies above the unit disc Ω. Calculate
the flux of q(x, y, z) = (x, y, z) across this surface in the direction of the upper unit normal.
Solution
Using the formulation above, f (x, y) = 1 − x2 − y 2 , fx = −2x and fy = −2y. Then
��
Flux = (−x(−2x) − y(−2y) + (1 − x2 − y 2 )) dx dy
Ω
�� � 1 � 2π
2 2
= (x + y + 1) dx dy = (r2 + 1)r dθ dr = 32 π.
Ω r=0 θ=0

Lecture 16

Downloaded by mick dundee ([email protected])