19.

The Divergence Theorem (Gauss Theorem)

If a is a vector field in a volume V , and S is the closed surface bounding V , then

Z

a dV =
V

a dS
S

Proof : We derive the divergence theorem by making use of the integral definition of a
Z
1
a dS.
a = lim
V 0 V
S
Since this definition of a is valid for volumes of arbitrary shape, we can build a smooth
surface S from a large number, N, of blocks of volume V (i) and surface S (i) . We have
Z
1
a dS + O((i) )
a(r(i) ) =
(i)
V
S (i)
where (i) 0 as V (i) 0. Now multiply both sides by V (i) and sum over all i
N
X

a(r (i) ) V (i) =

i=1

N Z
X

a dS +

S (i)

i=1

N
X

Examples of the use of the Divergence Theorem

Consider the volume of a body:

V =

dV

Recalling that r = 3 we can write

V =

1
3

r dV
V

which using the divergence theorem becomes

V =

1
3

r dS
S

81

dS = R2 sin d d er .
On the hemisphere r dS = R3 sin d d so that
Z
Z /2
Z
sin d
r dS = R3
V =

19.4

d = 2R3

i=1

Taking the limit N we have, as claimed,

Z
Z
a dS .
a dV =

Volume of a body:

We can evaluate this by using spherical polars for the surface integral. As was derived
previously, for a hemisphere of radius R

giving the anticipated result

butions to the respective integrals cancel.

19.3

On the bottom face dS = ez dS so that r dS = z dS = 0 since z = 0. Hence the only

contribution comes from the (open) surface of the hemisphere and we see that
Z
1
V =
r dS .
3 hemisphere

(i) V (i)

On rhs the contributions from surface elements interior to S cancel. This is because where
two blocks touch, the outward normals are in opposite directions, implying that the contri-

Example: Consider the hemisphere x2 + y 2 + z 2 R2 centered on e3 with bottom face at

z = 0. Recalling that the divergence theorem holds for a closed surface, the above equation
for the volume of the hemisphere tells us
Z

Z
1
r dS +
r dS .
V =
3 hemisphere
bottom

2R3
.
3

The Continuity Equation

Consider a fluid with density field (r) and velocity field v(r). We have seen previously that
R
the volume flux (volume per unit time) flowing across a surface is given by S v dS. The
corresponding mass flux (mass per unit time) is given by
Z
Z
J dS
v dS
S

where J = v is called the mass current.

Now consider a volume V bounded by the closed surface S containing no sources or sinks of
fluid. Conservation of mass means that the outward mass flux through the surface S must
be equal to the rate of decrease of mass contained in the volume V .
Z
M
J dS =
.
t
S
R
The mass in V may be written as M = V dV . Therefore we have
Z
Z

dV + J dS = 0 .
t V
S
We now use the divergence theorem to rewrite the second term as a volume integral and we

obtain
Z 

+ J dV = 0
t
V
Now since this holds for arbitrary V we must have that
82

Then

+J =0.
t

E =

x 
q
q
i
i 3 =
4
r
4

ii 3 5
r 2xi xi
r3
2

=0

r 6= 0

We also know that

This equation, known as the continuity equation, appears in many different contexts since
it holds for any conserved quantity. Here we considered mass density and mass current J
of a fluid; but equally it could have been number density of molecules in a gas and current
of molecules; electric charge density and electric current vector; thermal energy density and
heat current vector; or even more abstract conserved quantities such as probability density!
To understand better the divergence of a vector field consider the divergence of the current
in the continuity equation

if J(r) > 0 then

< 0 and the mass density at r decreases
t

> 0 and the mass density at r increases

if J (r) < 0 then
t

19.5

Sources and Sinks

= 0. The continuity equation

Static case: Consider time independent behaviour where
t
tells us that for the density to be constant in time we must have J = 0 so that flux into
a point equals flux out.
However if we have a source or a sink of the field, the divergence is not zero at that point.
Z
In general the quantity
1
a dS
V S
tells us whether there are sources or sinks of the vector field a within V : if V contains
Z
Z
a source, then
a dS =
a dV > 0
Z

a dS =

E dS =

sphere

q
4

r dS
q
q
=
4 =
r3
4
0

(where the surface integral has been evaluated in section 18.1). So for any volume V enclosed
by a surface S we have
Z
Z
Z
Z
q
E dS =
E dV =
E dV =
E dS =
0
S
V
sphere
sphere
(where the first result is because E = 0).
P
We can replace one charge by a collection of charges i qi or a charge density (r). Hence
 Z
Z
Z
1
E dS =
E dV =
(r) dV
0 V
S
V
This recovers Gauss Law of electrostatics.
Since this must hold for arbitrary V we see
(r)
0

E =

which holds for all r and is one of Maxwells equations of Electromagnetism.

19.6

Corollaries of the divergence theorem

a sink, then

From
a dV < 0

If S contains neither sources nor sinks, then

a dV =

a dS

a dS = 0.

we have several immediate consequences

As an example consider electrostatics. You will have learned that electric field lines are
conserved and can only start and stop at charges. A positive charge is a source of electric
field (i.e. creates a positive flux) and a negative charge is a sink (i.e. absorbs flux or creates
a negative flux).
The electric field due to a charge q at the origin is
E=

If a = c where c is constant then

c dS = 0 and as this is true for arbitrary c then

Z
dS = 0
S

as previously stated.

q r
.
40 r 2
83

84

20

Take V to be the cylinder 0 < z < 1 with

a = (P (x, y), Q(x, y), 0). Then


Z
Z
P
Q
a dV =
dxdydz
+
x
y
V
cylinder

Z 
P
Q
=
+
dxdy
x
y
A

20.1

Line Integral Definition of Curl; Stokes Theorem

and Applications
Line Integral Definition of Curl
.

_n

and

y
A

a dS = 0|ends +

a dS

curved
x

But dS|curved = (dxex + dyey ) ez dz =

(dyex dxey ) dz and hence

Z 
A

P
Q
+
x
y

dxdy =

n a

(a c) = i (ijk aj ck ) = ck kjii aj = c ( a)

(a c) dV =

dS (c a) = c

a dr
C

_n = _e z
_e y

So for a tensor T , then regarding one index as just being there

Z
Z
Tij dSi
i Tij dV =
S

This is the generalised divergence theorem. In particular with Tij = ijk ak we recover
the previous result.

x0

85

S0

1
S

_e z

dS a

Using indices the divergence theorem becomes

Z
Z
i ai dV =
ai dSi

= lim

This definition of curl is independent of the choice of basis. The usual Cartesian form
for curl can be recovered from this general definition by considering small rectangles in the
(ex ey ), (ey ez ) and (ez ex ) planes respectively.

NB: the integral around C is taken in the right-hand sense with respect to the normal n to
the surface as in the figure above.

Again as this is true for all arbitrary constant c then

Z
Z
dS a
a dV =
V

Let P be a point with Cartesian coordinates (x0 , y0 , z0 ) situated at the centre of a small
rectangle C = ABCD of size x y , area S = x y , in the (ex ey ) plane.

we have
( a) dV =

The component of a parallel to n is defined to be

Apply the divergence theorem to a c (with c again constant). Then as

(P dy Qdx)

which is Greens theorem in a plane (sometimes called the two dimensional divergence
theorem) relating the integral over a planar surface A to the line integral C enclosing
this area.

Let S be a small planar surface containing the

point P , bounded by a closed curve C, with unit
normal n and (scalar) area S. Let a be a vector
field defined on S.

_ex

The line integral around C is given by the sum of four terms

I
Z B
Z C
Z D
Z
a dr =
a dr +
a dr +
a dr +
C

86

a dr
D

Since r = xex + yey + zez , we have dr = ex dx along D A and C B, and dr = ey dy

along A B and D C. Therefore
Z B
I
Z
a dr =
ay dy
A

ax dx
C

ay dy +

20.2

ax dx =

ax dx =

If S is an open surface, bounded by a simple closed

curve C, and a is a vector field defined on S, then

_n

ax dx

For small x & y , we can Taylor expand the integrands, viz

Z

Stokes Theorem

a dr =

d_S

a dS

ax (x, y0 y /2, z0 ) dx
D
x0 +x /2
x0 x /2


y ax
(x, y0 , z0 ) + O(y2) dx
ax (x, y0 , z0 )
2 y

ax (x, y0 + y /2, z0 ) dx
C
x0 +y /2
x0 x /2

ax (x, y0 , z0 ) +


y ax
(x, y0 , z0 ) + O(y2) dx
2 y

where C is traversed in a right-hand sense about dS.

(As usual dS = ndS and n is the unit normal to S.)

Proof:
Divide the surface area S into N adjacent small surfaces as indicated in the diagram. Let
S (i) = S (i) n(i) be the vector element of area at r(i) . Using the integral definition of curl,
I


1
a dr
n a = lim
S0 S
C
we multiply by S (i) and sum over all i to get

so
1
S

Z

a dr +

a dr

=
=

1
x y

Z

1
x y

ax dx
D
x0 +x /2

x0 x /2

B
C

N
X


ax dx


ax
(x, y0 , z0 ) + O(y2 ) dx
y

i=1

N I
X


a(r (i) ) n(i) S (i) =
i=1

where C (i) is the curve enclosing the area S (i) .

.
1
_n

ax
(x0 , y0 , z0 ) as x , y 0
y

2
_n

A similar analysis of the line integrals along A B and C D gives


Z B
Z D
1
ay
a dr +
(x0 , y0 , z0 ) as x , y 0
a dr
S A
x
C
Adding the results gives for our line integral definition of curl yields






ay ax 

ez a = a z =
x
y  (x0 , y0 , z0 )
in agreement with our original definition in Cartesian coordinates.

(It can be shown that a, when defined in this way, is independent of the shape of the
infinitesimal area S.)

Since each small closed curve C (i) is traversed in the same sense, then, from the diagram,
N I
X
all contributions to
a dr cancel, except on those curves where part of C (i) lies on
C (i)

the curve C. For example, the line integrals along the common sections of the two small
closed curves C (1) and C (2) cancel exactly. Therefore
I
N I
X
a dr =
a dr
i=1

Hence

87

i=1

The other components of a can be obtained from similar rectangles in the (ey ez ) and
(ex ez ) planes, respectively.

a dr + . . .
C (i)

a dr =
C

C (i)

a dS
88

n a dS