Lecture 17

Surface Integrals

Relevant section of AMATH 231 Course Notes: Section 3.1.1

Surfaces are 2-dimensional subsets of R2 .

We wish to perform two fundamental integration

procedures over such surfaces.

1. Integration of scalar functions f (x, y, z) over S

Z
“ f dS ”
S

Examples:
R
1. If f (x, y, z) = 1, then S f dS is the area of S.

2. If f (x, y, z) is the charge density (per unit area), then dq = f dS is the amount of charge in
element dS. Z Z
total charge Q = dq = f dS
S S

If f is the mass density (per unit area) then dm = f dS, is the amount of charge in element dS,
Z Z
total mass M = dm = f dS
S S

In fact, to illustrate, we can easily compute the surface area of a sphere SR of radius R from our
knowledge of spherical polar coordinates. We’ll simply keep r = R fixed to remain on the sphere. The
volume integral then becomes a surface integral:
Z Z 2π Z π
dS = R2 sin φ dφ dθ
SR 0 0

1
 Z 2π Z π
= R2 sin φ dφ dθ
0 0
2
= 4πR . (1)

We’ll derive this result more mathematically later in this section.

2. Integration of vector functions F(x, y, z) over S

At each element dS, compute F · N̂ where N̂ is the

outward unit normal to S.
Z
“ F · N̂ dS ”
S

is the total outward flux of F through S.

In what follows, we shall develop a method to perform integrations over surfaces by using parametriza-
tions of surfaces, much as we did for integrations over curves. But we mention that it may be possible
to avoid such detailed work if our problem has some nice symmetries. In some special cases that
occur in physics such as in gravity and E&M (e.g., spherically symmetric vector fluids and spherically
symmetric surfaces), we can compute fluxes in a rather straight forward way using the definition of
the surface integral.

2
Surface integration via parametrization of surfaces

In general, we parametrize the surface S and then express the surface integrals from (1.) and (2.)
above as integrations over these parameters. We shall need two parameters, say u and v, to define S,
because S is 2-dimensional.

D is the set of parameter values (u, v) needed to define S.

The parameterization will be denoted by (to conform with the
AMATH 231 Course Notes)

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

(x0 , y0 , z0 ) = (x(u0 , v0 ), y(u0 , v0 ), z(u0 , v0 ))

Our goal: We want to write

Z Z Z
f dS = f (g(u, v)) − − −{z− −−} dudv
S D | {z } |
f evaluated may need

on surface S something
here
and
Z Z Z
F · N̂ dS = F(g(u, v)) · N̂(u, v) − − − − − dudv
S D

3
Some examples of surface parameterizations

1. Sphere with radius R, x2 + y 2 + z 2 = R2 .

Use the angles from spherical polar coordi-

nates to identify a point on the sphere:
Let u = θ, the azimuthal angle (angle with
respect to xz plane).
Let v = φ, the polar angle (angle with re-
spect to z axis).

Therefore,

x(u, v) = R sin v cos u

y(u, v) = R sin v sin u

z(u, v) = R cos v,

so that our parametrization may be expressed as

g(u, v) = (R cos u sin v, R sin u sin v, R cos v). (2)

The parameter space D that defines the sphere:

0 ≤ u ≤ 2π, 0 ≤ v ≤ π

4
2. Cylinder with x2 + y 2 = a2 , 0 ≤ z ≤ b

x(u, v) = a cos u
y(u, v) = a sin u
z(u, v) = v
0 ≤ u ≤ 2π,
0≤v≤b

The resulting parametrization:

g(u, v) = (a cos u, a sin u, v). (3)

The region D in parameter space defining the cylinder is shown at the upper right.

3. Cone with z 2 = x2 + y 2 , 0 ≤ z ≤ b

If we set z = v (then 0 ≤ v ≤ b), then

x = v cos u

y = v sin u

0 ≤ u ≤ 2π, 0 ≤ v ≤ b

The resulting parametrization:

g(u, v) = (v cos u, v sin u, v). (4)

The region D in parameter space defining the cone is shown below:

5
4. The graph of a function z = f (x, y) which, in general, describes a surface in R3 , as sketched
generically below.
z

z = f (x, y)

O
y
.
x

We could simply use x and y as the parameters – and will do so in some cases in the future. But
for now, we’ll let u and v be the parameters and let x = u and v = y.

For example, consider the graph of f (x, y) = x2 + y 2 over the region −1 ≤ x ≤ 1, −1 ≤ y ≤ 1 in

the plane, as sketched below.
z

z = x2 + y 2

O
2 y
1
x

The parametrization of this surface is

x(u, v) = u , y(u, v) = v z(u, v) = u2 + v 2 , (5)

which can be expressed in vector form as

g(u, v) = (u, v, u2 + v 2 ) . (6)

The parameter space D for this example is

−1 ≤ u ≤ 1 , − 1 ≤ v ≤ 1, (7)

the square of sides of length 2 centered at the origin.

6
5. Planes are the simplest objects because they are flat.

(a) Any plane parallel to a coordinate axis is trivial to parametrize. Consider, for example,
the plane z = c. This means that z(u, v) = c for all (u, v). And x and y can simply be
parametrized as x(u, v) = u and y(u, v) = v. The result:

g(u, v) = (u, v, c). (8)

The other planes, e.g., x = a, are treated in similar ways.

(b) As for the more general plane, Ax + By + Cz = D, we simply express one coordinate in
terms of the other two, e.g.,

D A B
z= − x− y (assuming that C 6= 0). (9)
C C C

Once again, we let x and y be parametrized by u and v, respectively. The resulting

parametrization of the general plane is
 
D A B
g(u, v) = u, v, − u − v . (10)
C C C

Please see Example 3.1 and Exercise 3.1 on pages 66-67 of the AMATH 231 Course Notes
for further discussion and another method of generating planes.

7
Normal vector: N = (1,1,0) X (1,-1,1) = (1,-1,-2)
Vector from (1,2,3) to (x,y,z) on plane is v = (x-1,y-2,z-3)
N . v = 0 => 1(x-1) + (-1)(y-2) + (-2)(z-3) = 0

z = 2-x+3y=2-u+3v
so (x,y,z) = (u,v,2-u+3v)
Normal vectors to surfaces from parametrizations

Relevant sections from AMATH 231 Course Notes: Sections 3.1.2 and 3.1.3

Curve C1 on the surface is obtained by fixing v = v0 and letting u vary:

cv0 (u) = g(u, v0 ).

Curve C2 on the surface is obtained by fixing u = u0 and letting v vary:

cu0 (v) = g(u0 , v).

Goal: To find N(u0 , v0 ), the normal vector to S at (x0 , y0 , z0 ) = g(u0 , v0 ). N is perpendicular to the
tangent vectors Tu and Tv to curves cv0 (u) and cu0 (v), respectively.
We can choose either of N = ±Tu × Tv .
Question: How do we find Tu and Tv ?
Answer: We know how to obtain tangent vectors to curves:

Tu = cv0 ′ (u)
∂
= g(u, v)v=v0
∂u

∂
⇒ Tu (u0 , v0 ) = g(u, v)
∂u
(u, v) = (u0 , v0 )
 
 ∂x ∂y ∂z 
 
=  , , 
 ∂u ∂u ∂u 
(u0 , v0 ) (u0 , v0 ) (u0 , v0 )

8
 Likewise,  
 ∂x ∂y ∂z 
 
Tv (u0 , v0 ) =  , , 
 ∂v ∂v ∂v 
(u0 , v0 ) (u0 , v0 ) (u0 , v0 )
∂ ∂
(You require two vectors – one involves , the other – because you have two parameters u, v.)
∂u ∂v

N(u0 , v0 ) = ± Tu (u0 , v0 ) × Tv (u0 , v0 )

î ĵ k̂
= ± ∂x ∂y ∂z
∂u ∂u ∂u
∂x ∂y ∂z
∂v ∂v ∂v (u0 ,v0 )

Note that all derivatives are evaluated at (u0 , v0 ).

Examples:

1. The cylinder g(u, v) = (cos u, sin u, v), 0 ≤ u ≤ 2π, 0 ≤ v ≤ 1.

Step 1: Parameterization: Given.

Step 2: Compute N(u, v) via Tu , Tv .

 
∂ ∂x ∂y ∂z
Tu = g(u, v) = , , = (− sin u, cos u, 0)
∂u ∂u ∂u ∂u
 
∂ ∂x ∂y ∂z
Tv = g(u, v) = , , = (0, 0, 1)
∂v ∂v ∂v ∂v
i j k
⇒ Tu × Tv = − sin u cos u 0 = (cos u, sin u, 0) = N(u, v)
0 0 1

Top view:

9
 As expected, the normal vector N is parallel to the xy-plane (since it has zero z-component)
and points radially outward from the cylindrical surface. Note also that kN(u, v)k = 1.

2. The general plane Ax + By + Cz = D which was discussed in the previous section. (Of course,
we know the answer: A normal vector to the plane is given by N = (A, B, C). But let’s see what
our parametrization method yields.)

By expressing z in terms of x and y, we arrived at the following parametrization,

 
D A B
g(u, v) = u, v, − u − v . (11)
C C C

We first compute the tangent vectors Tu and Tv :

   
∂ ∂x ∂y ∂z A
Tu = g(u, v) = , , = 1, 0, − (12)
∂u ∂u ∂u ∂u C
   
∂ ∂x ∂y ∂z B
Tv = g(u, v) = , , = 0, 1, − (13)
∂v ∂v ∂v ∂v C
i j k  
A
A B
⇒ Tu × Tv = 1 0 −C = , , 1 = N(u, v) . (14)
C C
B
0 1 = C

This result is consistent with the well-known result that the vector (A, B, C) (or any constant
multiple of it) is normal to the plane Ax + By + Cz = D.

3. Sphere with radius R, x2 + y 2 + z 2 = R2 . Parameterize as (u = θ, v = φ) (spherical coordinates).

g(u, v) = (R cos
| u{zsin v}, R |sin u{zsin v}, R cos
| {z v})
x y z

0 ≤ u ≤ 2π, 0≤v≤π
Need to compute Tu and Tv :

i j k
Tu × Tv = −R sin u sin v R cos u sin v 0
R cos u cos v R sin u cos v −R sin v

= (−R2 cos u sin2 v, −R2 sin u sin2 v, |−R2 sin2 u sin v cos v{z
− R2 cos2 u sin v cos v})
simplif ies to −R2 sin v cos v
= −R sin v (R cos u sin v, R sin u sin v, R cos v)
| {z }
g(u,v)
= −R sin v g(u, v)

10
 For the outward normal, choose

N = − Tu × Tv (or Tv × Tu )

= R sin v g(u, v)

Note that sin v ≥ 0 for v ∈ [0, π] as shown below on the left. The resulting outward normal N is
sketched on the right.

0≤v≤π

Let us now compute the magnitude of this normal vector:

k N k= R| sin v| k g(u, v) k= R2 sin v, (15)

where we have used the fact that k g(u, v) k= R since the point g(u, v), by construction, lies on
a surface of sphere R. You may now be wondering why the magnitude of this normal vector is not
constant; in particular, why would it be changing with the spherical polar angle v (angle between radial
vector g and the z-axis). The answer is that the R2 sin v factor provides another piece of information
that necessary to translate the integration over the sphere to an integration in uv parameter space.
It is the “conversion factor,” or Jacobian between an infinitesimal element of area dA in the uv
parameter space (shown at the left below) and the actual element of area dS on the sphere (shown at
the right). We’ll show this below.

11
Lecture 18

Surface integrals of scalar functions

Relevant section from AMATH 231 Course Notes: 3.1.2, 3.1.3, 3.2.1

Let f (x, y, z) be defined over a surface S. We

want to integrate f over S.

In the “spirit of calculus”:

The partitioning of D on the left defines a partitioning of S shown on the right.

Pick sample points (u∗k , vk∗ ), a subset of an appropriate square. Each of these points defines a
∗ , z ∗ ) = g(u∗ , v ∗ ).
point (x∗kl , ykl kl kl kl

Evaluate f at (x∗kl , ykl

∗ , z ∗ ).
kl

Goal:
X
f (x∗kl , ykl
∗ ∗
, zkl )∆Skl
kl

But we don’t know what ∆Skl is!

Approximate ∆Skl by ∆Skl ′ – the area of a flat plane that “covers” ∆Skl that is tangent to S
somewhere in ∆Skl . So

12
 X
∼
= ∗
f (x∗kl , ykl ∗
, zkl )∆Skl ′

So what is ∆Skl ′ ? The area of the flat tangent plane piece.

kP ′ Q′ k = ( speed w.r.t. u ) × ∆u = kTu (u0 , v0 )k∆u (cat)

Similarly,
kP ′ S ′ k = kTv (u0 , v0 )k∆v (dog)

So the area of the parallelogram kP ′ Q′ R′ S ′ k is

= (cat) · (dog) · sin θ

= kTu kkTv k sin θ∆u∆v

= kTu × Tv k∆u∆v

= kN(u0 , v0 )k∆u∆v
The final result is
X X
f (. . .)∆S ′ ∼
= f (x∗ , y ∗ , z ∗ )kN(u∗ , v ∗ )k∆u∆v
|{z}
n, m → ∞
partition
f evaluated

Z Z Z on surface
z }| {
⇒ f dS = f (g(u, v)) kN(u, v)k du
| {zdv}
S u,v
dv du
| {z }
dS element of area

13
This is the formula for computing surface integrals of scalar-valued functions. Let us now work out a
few surface integrations using the result stated at the end of the previous section.

Example 1: Surface area of a sphere with radius R.

We shall use the parametrization introduced previously:

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

= (R cos u sin v, R sin u sin v, R cos v)

where v = θ, u = φ, 0 ≤ v ≤ π, 0 ≤ u ≤ 2π. Recall that

N = ±Tu × Tv

We choose Tv × Tu since it is the outward normal. Previously, we computed the magnitude of this
normal vector to be
kN(u, v)k = R2 sin v. (16)

Let us now investigate the signifance of this result. In the figure below, on the left, the infintesimal
element of area dA produced by infinitesimals du and dv does not vary as we move around the rectangle
in uv space. But the resulting element of area dS does vary on the sphere. For example, the closer you
are to the top of the sphere (i.e., the “North Pole”), the smaller the element ds because the lines of
constant v, i.e., the lines of longitude, are coming together as we approach the pole. This is reflected
in the sin v term. As v → 0, the element of area dS gets smaller and smaller.

dA = dudv

→ dS = kN(u, v)k dudv

| {z }
Jacobian

In fact, you have already seen this factor: it appears in the volume element dV in spherical polar
coordinates. The only difference is that the R above, which is a fixed value – since we are considering
a spherical shell – becomes the lower case r in 3D spherical polar coordinates, since we also integrate

14
over this variable. In order to compute the surface area of the sphere, we choose f (x, y, z) = 1. Then
Z Z Z Z
f dS = dS = A
S S

In general, the integration of a function f over a surface will be computed as follows,

Z Z Z Z
f dS → f (g(u, v)) kN(u, v)k |{z}
dA
S D
dudv
|{z}
dvdu

Ṫhis implies that the surface area A is given by (f = 1),

Z Z
kN(u, v)kdA .
D

For our problem,

kN(u, v)k = R2 sin v .

We have to integrate this function over the region, 0 ≤ u ≤ 2π, 0 ≤ v ≤ π, i.e.,

Z Z Z 2π Z π
2
A= kN(u, v)kdA = R
| {zsin v} dvdu
| {z }
D 0 0
| {z } kN(u,v)k dA
parameters
that generate
S
Inner Integral:
Z π
π
R2 sin v dv = R2 (− cos v) = 2R2
0
0
Outer Integral:
Z 2π
2π
2 2
2R du = R (u) = 4πR2
0
0
.
This agrees with our result for the area of a spherical surface of radius R.

15
Example 2: Given a thin hemispherical shell, radius R, positioned as shown below. Find its centroid
(0, 0, z).

RR
z dS
z = R RS
S dS
RR
Here, the denominator S dS is the surface

area of the hemisphere which is 2πR2 .

RR
We need to compute S z dS.
Z Z Z Z
z dS =
|{z} z(u, v)kN(u, v)kdA
S D
parametrization
g(u, v)

For z(u, v), we now have to express z on the surface in terms of u and v.

⇒ z(u, v) = R cos v

Note that we are integrating over the upper hemisphere, so the ranges of the parameters will be
0 ≤ u ≤ 2π (as for the sphere) but 0 ≤ v ≤ π/2 (unlike the sphere).
Z Z π
2π 2
2
= | cos
R sin v} dvdu
{z v} |R {z
0 0 z kN(u, v)k
Jacobian
Inner Integral:
π
Z 2π 2
1 1 3
R3 cos v sin v dv = R3 ( sin2 v) = R
0 2 2
0
Outer Integral:
Z 2π
2π
1 3 1
R du = R3 (u) = πR3
2 0 2
0

16
 πR3 R
⇒z= 2
= (half-way up)
2πR 2

Note: You may wish to compare this result to the centroid of the solid hemisphere, z = 83 R, which
can be found by integration in 3D spherical polar coordinates.

RR
Example 3: Compute the integral S x2 z dS over the cylinder g(u, v) = (cos u, sin u, v), 0 ≤ u ≤ 2π,
0 ≤ v ≤ 1.

Step 1: Parameterization: Given.

Step 2: Compute N(u, v) via Tu , Tv .

 
∂r ∂x ∂y ∂z
Tu = = , , = (− sin u, cos u, 0)
∂u ∂u ∂u ∂u
∂r
Tv = = (0, 0, 1)
∂v
i j k
⇒ Tu × Tv = − sin u cos u 0 = (cos u, sin u, 0) = N(u, v)
0 0 1

Top view:

As expected, the normal vector N is parallel to the xy-plane (since it has zero z-component) and
points radially outward from the cylindrical surface. Note also that kN(u, v)k = 1.

17
We now use the parametrizations x = cos u and z = v in the integrand to give
Z Z Z 1 Z 2π
2 π
x z dS = cos2 u v dudv = . (17)
S 0 0 2

The above integral is particularly convenient to compute since it is separable, i.e., the u- and v-
dependent terms can be separated so that that the double integral over u and v can be expressed as
a product of integrals over u and v:
Z 1 Z 2π Z 1  Z 2π   
2 2 1 π
cos u v dudv = v dv cos u du = (π) = . (18)
0 0 0 0 2 2

Example 4: Compute the area of the cylinder used in Example 3.

RR
We simply have to compute the surface integral S dS. Using the parametrization from Example 3

and the fact that kN(u, v)k = 1, the surface area is given by
Z Z Z 1 Z 2π
dS = dudv = 2π. (19)
S 0 0

This result is confirmed if we take the cylinder, cut it open by means of a vertical cut, and flatten it
out. The result is a rectangle with dimensions 2π and 1. The area of this rectangle is 2π.

Example 5: Compute the surface area of a right circular cone (without the base circle) of height h
and base radius b. This problem is done in Example 3.4 of the AMATH 231 Course Notes and was
covered in class. The appropriate pages of the Course Notes are reproduced below.

18
 From AMATH 231 Course Notes
In order to approximate the area ∆S of a surface element we need to approximate the
vectors A and B that define the sides of the surface element in Figure 3.6. Using the linear
approximation (3.11) we obtain

∂g ∂g
A ≈ (∆u) (u0 , v0 ), B ≈ (∆v) (u0 , v0 ).
∂u ∂v
Thus, simplifying ∆S ≈k A × B k gives

∂g ∂g
∆S ≈ (u0 , v0 ) × (u0 , v0 ) ∆u∆v. (3.16)
∂u ∂v

To obtain the total area S we have to sum over all surface elements determined by the
partition of Duv and take the limit as N → ∞ and max(∆u), max(∆v) → 0. This process
leads to a double integral over the set Duv in the uv-plane. With the above as motivation
we make the following.

Definition:
The surface area of the surface by x = g(u, v), (u, v) ∈ Duv , where g is C 1 , is defined by
ZZ
∂g ∂g
S= × du dv. (3.17)
∂u ∂v
Duv

Example 3.4:
Calculate the surface area of a cone of radius b and height h.

Solution: The cone is given by

x2 + y 2 z2
= , 0 ≤ z ≤ h.
b2 h2
A suitable vector-valued function is

x = g(u, v) = (bu cos v, bu sin v, hu), (u, v) ∈ Duv ,

with Duv = [0, 1] × [0, 2π]. The tangent vectors are

∂g ∂g
= (b cos v, b sin v, h), = (−bu sin v, bu cos v, 0),
∂u ∂v
giving
i j k
∂g ∂g
× = b cos v b sin v h = (−bhu cos v, −bhu sin v, b2 u),
∂u ∂v
−bu sin v bu cos v 0
and
∂g ∂g √
× = b2 + h2 bu.
∂u ∂v

72
Thus by (3.17) the surface area is
ZZ √
S= b2 + h2 bu du dv
Duv
 
√ Z1 Z2π
= b2 + h2 b  u dv  du
u=0 v=0
√
= · · · = πb b2 + h2 .

v
2π D uv

0 1 u

Comment:
This example is simply intended to show you how the “machinery” works. The surface
area of a cone can be calculated by simple geometry. 

The most important part of this subsection is the approximation (3.16) for the area of a
surface element, which we shall use when introducing surface integrals, the main goal of this
chapter.

3.1.4 Orientation of a surface

Consider a surface Σ given by

x = g(u, v), (u, v) ∈ Duv ,

where g is one-to-one and of class C 1 , and the normal vector

∂g ∂g
N= ×
∂u ∂v
is non-zero for all (u, v) ∈ Duv . Since g is one-to-one the surface does not intersect itself (see
Figure 3.7), and since g is of class C 1 , N varies continuously over Σ. In the sequel we shall
need to work with a unit normal n on Σ. There are two choices for n at each point, namely
1
n=± N.
kNk

73
Surface integrals (cont’d)

Surface integrals of vector functions and “flux”

Relevant section from AMATH 231 Course Notes: 3.2.2, p. 77

We return to the concept of the “flux” of a

vector field F. Now, however, we are
concerned with the flux of F through s
surface S.

The unit normal N̂ to S at (x, y, z) is usually

directed “outward” (as dictated by the
physical problem).

(x, y, z) ∈ S is centered in an infinitesimal

element dS of area on S.

The “amount” of F pointing in the direction of N̂ is the projection of F in the direction of N̂,
i.e., F · N̂. The (infinitesimal) flux of F through the surface element dS is F · N̂ dS. To obtain the
total flux through the surface S, we must integrate over all elements dS centered at (x, y, z) ∈ S. We
denote this vector surface integral as follows
Z Z
F · N̂ dS. (11)
S

In order to understand this idea better, let us examine a particular physical application of the flux
integral.
Flux in terms of fluid flow

First, consider a region D that lies in the xy-plane as sketched below. Suppose that a fluid is passing
through this region. For the moment, we assume that motion of the fluid is perpendicular to region
D, travelling in the direction of the positive z-axis. Moreover, we assume that the speed of the fluid
particles crossing D is constant throughout the region. As such, we are assuming that the velocity
field of the fluid is
F = vk, v > 0 (constant) . (12)

F = vk

y
D
x

We first ask the question: How much fluid flows through region D during a time interval ∆t?
Consider a tiny rectangular element of area ∆A = ∆x∆y centered at a point (x, y) in D. After a time
∆t, the fluid particles situated in this element will have moved a distance v∆t upward. The volume
of fluid that has passed through this element ∆A on D is the volume of the box of base area ∆A and
height v∆t:
v∆t∆A. (13)

This box is sketched below.

z

F = vk

y
v∆t
D
x

∆A

The total volume ∆V of fluid that has passed through region D over the time interval ∆t is
obtained by summing up over all area elements ∆A in D:
Z Z
∆V = v∆t dA = v∆tA(D), (14)
D

where A(D) denotes the area of D. Of course, this is a rather trivial result: the volume of fluid passing
through D is simply the volume of the solid of base area A(D) and height v∆t. Dividing both sides
by ∆t, we have
∆V
= vA(D). (15)
∆t
In the limit ∆t → 0, we have the instantaneous rate of change of the volume of fluid passing through
region D, or simply the rate of fluid flow through region D:

V ′ (t) = vA(D). (16)

This quantity is the flux of the vector field v through region D.

Now suppose that the fluid is now moving at a constant speed v through region D but not
necessarily at right angles to it, i.e., not necessarily parallel to its normal vector k. We shall suppose
that
v = v1 i + v2 j + v3 k, k v k= v (17)

and let γ denote the angle between v and the normal vector k.
In this case, the fluid particles that pass through the tiny element ∆A after a time interval ∆t
form a parallelopiped of base area ∆A and height v cos γ∆t, as sketched below.
z

γ y
v cos γ
x

∆A

The volume of this box is

v cos γ∆t∆A. (18)

(Think of this tower of fluid as a deck of playing cards that has been somewhat sheared. When you
slide the cards back to form a rectangular arrangement, the height of the deck is vδt cos γ.)
 The total volume ∆V of fluid that has passed through region D over the time interval ∆t is
obtained by summing up over all area elements ∆A in D:
Z Z
∆V = v cos γ∆t dA = v cos γ∆tA(D), (19)
D

Dividing both sides by ∆t, we have

∆V
= v cos γA(D). (20)
∆t
In the limit ∆t → 0, we obtain the flux of the vector field v through region D:

V ′ (t) = v cos γA(D). (21)

We shall rewrite this flux as follows,

V ′ (t) = v · N̂ A(D), (22)

since N̂ = k is the normal vector to surface D. Note that this general case includes the first case,
γ = 0. And in the case that γ = π/2, there is no flow through the region D, so the flux is zero.
Of course, the above results have been rather trivially obtained since (i) the vector fields are
constant and (ii) the region D is flat. Let us now generalize the first case, i.e., the vector field v is
assumed to be nonconstant over region D, i.e.,

v(x, y) = v1 (x, y)i + v2 (x, y)j + v3 (x, y)k. (23)

In this case, the total volume ∆V of fluid that has passed through region D over the time interval ∆t
is obtained by summing up over all area elements ∆A in D:
Z Z Z Z
∆V = ∆t v(x, y) · N̂ dA = ∆t v3 (x, y) dA. (24)
D D

Once again dividing by ∆t and taking the limit ∆t → 0, we obtain the total flux of v through region
D: Z Z
′
V (t) = v(x, y) · N̂ dA (25)
D
Now suppose that we were concerned with rate of mass flow through region D. The amounts/volumes
of fluid examined earlier would be replaced by amounts of mass flowing through a surface element.
This means replacing the velocity vector field v by the momentum field F = ρ v, where ρ is the mass
density. The rate of transport of mass through region D would then be given by
Z Z Z Z
′
M (t) = F(x, y) · N̂ dA = ρ(x, y) v(x, y) · N̂ dA (26)
D D

This concludes our discussion of this simple problem involving fluid flow through a flat surface.
Generalization to arbitrary surfaces

We now wish to generalize the above result to general surfaces in R3 . In other words, we do not
require the surface to be flat, as was region D in the plane, but rather a general surface S in R3 –
for example, a portion of a sphere, or perhaps the entire sphere. In the “spirit of calculus,” we divide
the surface S into tiny infinitesimal pieces dS. We then construct a normal vector N̂ to each surface
element dS at a point in dS, as sketched below.

We then form the dot product of the vector field F at that point with the normal vector N̂. This
will represent the local flux of F through the surface element dS. To obtain the total flux through the
surface S, we add up the fluxes of all elements dS – an integration over S that is denoted as
Z
F · N̂ dS. (27)
S

This is the “flux” of F through surface S. Note that in some books, especially Physics books, the
vector surface integral is denoted as
Z Z
F · dS. (28)
S

Here, the infinitesimal surface area element is a vector that is defined as

dS = N̂ dS, (29)

where dS is the infinitesimal surface element and N̂ is the unit normal vector to the surface element.
In other books, the infinitesimal surface element is denoted as dA = N̂dS, so that the flux integral
is denoted as
Z Z
F · dA. (30)
S
Practical computation of flux integrals

The integrand in a vector surface integral, F · N̂, will generally depend on the x, y, z. Since these
coordinates are restrcted to the surface S of interest, they, hence the integrand, will depend on u and
v (i.e., (x(u, v), y(u, v), z(u, v))).
As we saw in the introduction to this section, in some special cases that occur in physics, e.g.
gravity, electricity/magnetism, the vector fields and the surfaces are spherically symmetric. In such
cases, we can compute fluxes in a rather straightforward way using the definition of the flux integral.
In general, however, the computation of flux integrals may not be as straightforward as for these
elementary examples. As in the case of surface integrals for scalar functions, we’ll have to use the
parametrization of the surface.

Notice that in the total flux integral, Z Z

F · N̂ dS
S

the integrand F · N̂ is a scalar –call it f (x, y, z) – so that the flux integral becomes
Z Z
= f dS
S

But we know how to compute surface integrals of scalar functions! If we parametrize the surface S,
(g(u, v), (u, v) ∈ Duv ), then
Z Z Z Z
f dS = f (g(u, v))kN(u, v)kdA (31)
S Duv

Here, f is the dot product F · N̂ on surface S:

f (g(u, v)) = F(g(u, v)) · N̂(u, v)

| {z } | {z }
F on surf ace appropriate
unit
normal on
S

What is N̂?
N(u, v)
N̂(u, v) =
kN(u, v)k
So Equation (31) becomes
Z Z
N(u, v)
= F(g(u, v)) · kN(u, v)kdA
S kN(u, v)k
The final result is:
Z Z Z Z
F · N̂ dS = F(g(u, v)) · N(u, v) dA
S Duv

The vector surface integral is translated into a surface integral involving the scalar function F · N̂.
The integration on the right-hand side is performed over the region in uv parameter space, Duv , that
generates the surface S.

Some of the examples which are presented below will involve either spherical surface SR – either
a part of it or its entirety – of radius R and centered at (0, 0, 0). Recall that the parametrization of
this surface is given by

g(u, v) = (R sin v cos u, R sin v sin u, R cos v), 0 ≤ u ≤ 2π, 0 ≤ v ≤ π. (32)

Also recall from our earlier work that the outward normal associated with this parametrization was
given by

N = Tv × Tu = R sin v g(u, v). (33)

1. As a kind of warmup, let us first compute the total flux of the vector field F = k = (0, 0, 1)
through the circular planar region S on the xy-plane,

S = {(x, y, 0) | x2 + y 2 ≤ R2 } . (34)

The boundary of this region is the intersection of the spherical surface SR with the xy-plane.
The situation is sketched below.
z

y
x
 The computation of the flux of F through this region is quite straightforward since the surface
is a plane and the vector field is constant. In fact, the surface lies on the xy-plane which implies
that its unit normal vector N̂ = k. We can simply use Eq. (21) from the earlier part of this
lecture:
Z Z Z Z Z Z
F · N̂ dA = (0, 0, 1) · (0, 0, 1) dA = dA = πR2 . (35)
S S S

2. Now let us compute the total flux of the same vector field, F = k = (0, 0, 1), but through the
upper hemispherical surface of SR , as sketched below.
z

upper hemisphere of
SR

y
x

We must compute the surface flux integral,

Z Z Z Z
F · N̂ dS = F(g(u, v)) · N(u, v) dA
S Duv

The integrand must be evaluated over the surface S. Here, F(g(u, v)) = (0, 0, 1) so that

F(g(u, v)) · N(u, v) = (0, 0, 1) · R sin v(R sin v cos u, R sin v sin u, R cos v)

= R2 sin v cos v.

Now integrate over surface, keeping in mind that we’re integrating only over the upper hemi-
sphere, so that 0 ≤ v ≤ 12 π:
Z Z Z π/2 Z 2π
F · N̂ dS = R2 sin v cos v dudv
S 0 0
Z π/2
= 2πR2 sin v cos v dv
0
1
2π
1
= (2πR2 ) sin2 v
2
0
| {z }
2
= πR .
 Note that we obtain the same result as in Example 1. The net flux of the fluid passing through
the hemispherical surface is the same as the flux of the fluid passing through the projection of
this surface onto the xy-plane.

3. Let us now compute the total outward flux of the vector field F = k = (0, 0, 1) of Examples
1 and 2 through the entire spherical surface SR .

It appears that the amount of vector field entering the

sphere from the bottom is equal to the amount that is
exiting at the top. One would conjecture, then, that the
total flux across the sphere is zero.

Once again, we must compute the flux integral,

Z Z Z Z
F · N̂ dS = F(g(u, v)) · N(u, v) dA
S Duv

The integrand must be evaluated over the surface. Here, F(g(u, v)) = (0, 0, 1), so that

F(g(u, v)) · N(u, v) = (0, 0, 1) · R sin v(R sin v cos u, R sin v sin u, R cos v)

= R2 sin v cos v.

Now integrate over entire spherical surface: In this case 0 ≤ v ≤ π:

Z Z Z π Z 2π
F · N̂ dS = R2 sin v cos v dudv
S 0 0
Z π
2
= 2πR sin v cos v dv
0

π
1 2
= (2πR ) sin2 v
2
0
| {z }
=0
= 0.

This was as expected – net outward flow through the top is equal to the net inward flow through
the bottom.

4. Compute the total outward flux of the vector field F = zk = (0, 0, z) through the spherical
surface SR .
 The arrows in the vector field get longer as you move
away from the xy-plane. Moreover, for z > 0, they point
upward and for z < 0, they point downward. Therefore,
there is a net outward flow from the xy-plane through
the surface. We expect a nonzero, in fact, positive
outward flux here.
On the surface,

F(g(u, v)) = (0, 0, z(u, v))

= (0, 0, R cos v),

so that

F(g(u, v)) · N(u, v) = (0, 0, R cos v) · R sin v(R sin v cos u, R sin v sin u, R cos v)

= R3 cos2 v sin v.

The flux is therefore given by

Z Z Z π Z 2π
F · N̂ dS = R3 cos2 v sin v dudv
S 0 0
Z π
3
= 2πR cos2 v sin v dv
0
 
π
 1 
= (2πR3 ) − cos3 v
 

 3 
0
2
= (2πR3 )( )
3
4 3
= πR .
3

By symmetry, you will find the same result for F = xi or F = yj (i.e., same type of vector field,
just in a different direction).

Note: The fact that the total flux for these vector fields is equal to the volume of the sphere
is not a coincidence, as we’ll see in the next lecture. It is a consequence of the celebrated
Divergence Theorem.
Gauss Divergence Theorem in R3

Relevant section of AMATH 231 Course Notes: Section 4.2.1

In this section, we are concerned with the outward flux of a vector field (through a smooth/piece-
wise smooth) surface S that encloses a region V ⊂ R3 . Typically, in physics such surfaces are spheres,
boxes, parallelpipeds or cylinders. This is the subject of the celebrated Gauss Divergence Theorem,
the three-dimensional version of the Divergence Theorem in the Plane of a previous lecture.
The Gauss Divergence Theorem is one of the most important results of vector calculus. It is not
as important for computational purposes as for conceptual developments. It provides the basis for the
important equations in electromagnetism (Maxwell’s equations), fluid mechanics (continuity equation)
and continuum mechanics in general (heat equation, diffusion equation).
It is sufficient to consider a somewhat simplified version of the general Divergence Theorem.

A simplified version of the Divergence Theorem:

Let S be a “nice” (i.e., piecewise smooth) closed and nonintersecting surface that encloses a region
D ⊂ R3 , such that an outward unit normal vector N̂ exists at all points on S. Also assume that a
vector field F and its derivatives are defined over region D and its boundary S.

The Divergence Theorem states that:

Z Z Z Z Z
F · N̂ dS = div F dV . (36)
S D
| {z } | {z }
surf ace integral volume integral

Once again, we have assumed that div F exists at all points in V .

You have already seen a version of this theorem – the two-dimensional Divergence Theorem in
the plane. It expressed the total outward flux of a 2D vector field F through a closed curve C in the
plane as an integral of the divergence of F over the region D enclosed by C:
I Z Z
F · N̂ ds = div F dA. (37)
C D

Examples: In what follows, unless otherwise indicated, the surface S is an arbitrary surface in R3
satisfying the conditions of the Divergence Theorem.
1. The vector field F = k = (0, 0, 1). This vector field could be viewed as the velocity field of a
fluid that is travelling with constant speed in the positive z-direction:
y

The divergence of this vector field is zero:

∂ ∂ ∂
div F = (0) + (0) + (1) = 0. (38)
∂x ∂y ∂z

More importantly, it exists at all points in R3 , i.e., there are no singularities, so that we may
employ the Divergence Theorem. Therefore, for any surface S enclosing a region D, we have
Z Z Z Z Z Z Z Z
F · N̂ dS = div F dV = 0 dV = 0. (39)
S D D

In other words, the total outward flux of F over the surface S is zero. In terms of the fluid
analogy, fluid is entering the region through surface S from the bottom at the same rate that it
is leaving it at the top. There is no creation of extra fluid anywhere inside region D that would
cause a nonzero flux.

2. The vector field F = zk = (0, 0, z). A sketch of the vector field is given below.
y

This field could be viewed as the velocity field of a liquid that originates from the xy-plane and
travels upward and downward away from it. As it moves away, it accelerates, since the velocity
is proportional to the distance from the xy-plane.
 The divergence of this field is

∂ ∂ ∂
div F = (0) + (0) + (z) = 1. (40)
∂x ∂y ∂z

Once again, the divergence exists at all points in R3 . Therefore, by the Divergence Theorem
Z Z Z Z Z Z Z Z
F · N̂ dS = div F dV = 1 dV = V (D), (41)
SR D D

the volume of region D.

Note that the same result for the flux, i.e., Eq. (41), would be obtained for the following vector
fields:
(i) F = xi, (ii) F = yj, (42)

since the divergence of each of these vector fields is 1. And the list does not stop here. Consider
the set of all vector fields of the form

F = c1 xi + c2 yj + c3 zk, where c1 + c2 + c3 = 1. (43)

In all cases, we have div F = 1, so that the total outward flux of each of these fields through the
surface S is V (D), the volume of region D.

3. The vector field F = z 2 k = (0, 0, z 2 ). Here, all arrows of F point upward, as sketched below.
y

This could be visualized as fluid that emanates from the xy-plane to travel upward, accelerating
as it moves away from the plane, along with fluid that approaches the xy-plane from below,
decelerating as it gets closer.

The divergence of F is
∂ ∂ ∂ 2
div F = (0) + (0) + (z ) = 2z (44)
∂x ∂y ∂z
Therefore, by the Divergence Theorem
Z Z Z Z Z Z Z Z
F · N̂ dS = div F dV = 2 z dV. (45)
S D D

The value of this integral will depend on the region D. In principle, if we knew the region, we
could integrate over it, using the techniques for integration in R3 developed earlier in the course.

There is one interesting point regarding this integral: It is related to the z coordinate of the
centroid of region D. Recall that
RRR RRR
D z dV D z dV
z̄ = R R R .= , (46)
D dV V (D)

implying that Z Z Z
z dV = V (D)z̄. (47)
D

Therefore, Eq. (45) becomes

Z Z
F · N̂ dS = 2z̄V (D). (48)
S

Note that if the surface S is located in the upper half-plane, i.e., z > 0, then z̄ > 0, implying that
the total outward flux is positive. However, if the surface S is located in the lower half-plane,
i.e., z < 0, the total outward flux is negative. Why is this so? And why would the total outward
flux be directly proportional to the volume V (D) of the region D enclosed by the surface S?