Mathematical Methods For Physics 2

Download as pdf or txt
Download as pdf or txt
You are on page 1of 62

Maths Methods 2

Spring 2013
1 Useful Mathematical Results
1.1 Trigonometric Identities
We recall the following useful trigonometric identities:
q
H
O
A

Figure 1:
tan =
sin
cos
, cosec =
1
sin
, sec =
1
cos
, cot =
1
tan
=
cos
sin
.
From the well known identity
cos
2
+ sin
2
= 1 (1.1)
we have
1 + tan
2
= sec
2

and
cot
2
+ 1 = cosec
2
.
Addition and subtraction formulae:
sin ( ) = sin cos cos sin
and
cos ( ) = cos cos sin sin
If = then from the above addition and subtraction formulae we have
sin 2 = 2 sin cos
and
cos 2 = cos
2
sin
2
. (1.2)
1
Using (1.1) in (1.2) we have
cos 2 = 1 2 sin
2

sin
2
=
1
2
(1 cos 2) . (1.3)
Alternatively we have
cos 2 = 2 cos
2
1
cos
2
=
1
2
(1 + cos 2) . (1.4)
The formulas (1.3) and (1.4) are useful when trying to integrate cos
2
or sin
2
.
1.2 Trigonometric functions
In Figure 2 we display graphs of the trigonometric functions sin , cos , tan , csc , sec and cot .
p p
2
3p
2
2p
q
tan(q)
0
10
-10
p p
2
3p
2
2p
sin(q)
0
1
-1
q
p p
2
3p
2
2p
cos(q)
0
1
-1
q
p p
2
3p
2
2p
cot(q)
0
10
-10
q
p p
2
3p
2
2p
csc(q)
0
10
-10
q
p
p
2
3p
2
2p
sec(q)
0
10
-10
q
Figure 2:
1.3 Determinants of Matrices
For a 2 2 matrix we have
A =
_
a b
c d
_
det A = |A| = ad bc.
For a 3 3 matrix we have
A =
_
_
a b c
d e f
g h i
_
_
2
det A = |A| = a

e f
h i

d f
g i

+c

d e
g h

= a (ei hf) b (di gf) +c (dh ge) .


1.4 Vectors
Vectors have magnitude and direction.
The vector i has magnitude 1 and points in the increasing x direction. The vector j has magnitude 1
and points in the increasing y direction. The vector k has magnitude 1 and points in the increasing z
direction.
The vectors i, j and k are unit vectors. All unit vectors have magnitude 1 and usually they are written
with hats, e.g. u or v.
A standard vector
F = F
x
i +F
y
j +F
z
k
has 3 components, F
x
, F
y
and F
z
that are all scalars. F
x
is the x component of F, F
y
is the y component
of F and F
z
is the z component of F.
The magnitude of the vector F is denoted by |F| and is dened by
|F| =
_
F
2
x
+F
2
y
+F
2
z
.
A unit vector in the direction of F has magnitude 1 and points in the same direction as F, it is denoted
by

F =
1
|F|
F =
1
|F|
(F
x
i +F
y
j +F
z
k) .
Example 1.1 For the vector
G = 3i + 2j
nd the three components G
x
, G
y
and G
z
, then determine the magnitude of G and the unit vector that
points in the same direction as G.
Solution:
The three components G
x
, G
y
and G
z
are
G
x
= 3, G
y
= 2 and G
z
= 0.
Hence the magnitude of G is
|G| =
_
G
2
x
+G
2
y
+G
2
z
=
_
3
2
+ 2
2
+ 0
2
=

13
and the unit vector that points in the same direction as G is

G =
1
|G|
G =
1

13
(3i + 2j) .
Example 1.2 For the vector
F = 2yi 3zj + 9z
2
k
nd the three components F
x
, F
y
and F
z
and then determine the magnitude of F.
3
Solution:
The three components F
x
, F
y
and F
z
are
F
x
= 2y, F
y
= 3z, and F
z
= 9z
2
and hence the magnitude of F is
|F| =
_
(2y)
2
+ (3z)
2
+ (9z
2
)
2
=
_
4y
2
+ 9z
2
+ 81z
4
.
1.4.1 The scalar (or dot) product
The scalar product of 2 vectors A and B is given by
A B = |A| |B| cos
where is the angle between the two vectors.
Since cos 0 = 1 and cos

2
= 0 we have
i i = |i| |i| cos 0 = 1 1 1 = 1, j j = 1 and k k = 1.
Furthermore
i j = |i| |j| cos

2
= 1 1 0 = 0, i k = 0 and j k = 0.
If
A = A
x
i +A
y
j +A
z
k and B = B
x
i +B
y
j +B
z
k
A B = (A
x
i +A
y
j +A
z
k) (B
x
i +B
y
j +B
z
k)
= A
x
iB
x
i+A
x
iB
y
i+A
x
iB
z
i
+A
y
iB
x
i+A
y
iB
y
i+A
y
iB
z
i
+A
z
iB
x
i+A
z
iB
y
i+A
z
iB
z
i
= A
x
B
x
(i i) +A
x
B
y
(i j) +A
x
B
z
(i k)
+A
y
B
x
(j i) +A
y
B
y
(j j) +A
y
B
z
(j k)
+A
z
B
x
(k i) +A
z
B
y
(k j) +A
z
B
z
(k k)
= A
x
B
x
+A
y
B
y
+A
z
B
z
.
Example 1.3 For
F = 3yi 2xk and G = i + 2zj 3k
calculate F G.
Solution: We have
F G = 3y 1 + 0 2z + (2x) (3)
= 3y + 6x.
4
1.4.2 The vector (or cross) product
The vector product of 2 vectors A and B is given by
AB = |A| |B| sin n
where n is a unit vector perpendicular to A and B, pointing in the direction given by the right hand
screw rule (i.e. the direction in which a screw would advance if it were turned from A through the angle
to B. The cross product can be evaluated by calculating the determinant of the 3 3 matrix
AB =

i j k
A
x
A
y
A
z
B
x
B
y
B
z

= i

A
y
A
z
B
y
B
z

A
x
A
z
B
x
B
z

+k

A
x
A
y
B
x
B
y

= i (A
y
B
z
B
y
A
z
) j (A
x
B
z
B
x
A
z
) +k(A
x
B
y
B
x
A
y
) .
Example 1.4 For
A = 3xi 4k and B = 2i 9xj +k
evaluate AB.
Solution: We have
AB =

i j k
3x 0 4
2 9x 1

= i

0 4
9x 1

3x 4
2 1

+k

3x 0
2 9x

= i ((0) (1) (9x) (4)) j ((3x) (1) (2) (4)) +k((3x) (9x) (2) (0))
= i (36x) j (3x + 8) +k
_
27x
2
_
= 36xi (8 + 3x) j 27x
2
k.
Remark: AB = BA
1.5 Dierentiation
Given a function f(x), (or f(t)) of a single variable x (or t), we denote the derivative of f(x) by f

(x)
or
df
dx
(or the derivative of f(t) by

f(t) or
df
dt
) The following table gives the derivatives of some common
functions.
f(x) f

(x)
x
n
nx
n1
sin x cos x
cos x sin x
tan x sec
2
x
ln |x|
1
x
e
x
e
x
1.5.1 The product rule
If u = u(x) and v = v(x) then
d (uv)
dx
=
du
dx
v +u
dv
dx
= u

v +uv

.
5
1.5.2 The quotient rule
If u = u(x) and v = v(x) then
d
dx
_
u
v
_
=
u

v uv

v
2
.
1.5.3 The chain rule
If f = f (u(x)) then
df
dx
=
df
du
du
dx
.
Example 1.5 For
f(x) = x
3
cos x
calculate f

(x).
Solution: Sincef(x) = x
3
cos x we set
u = x
3
and v = cos x
and hence
du
dx
= 3x
2
and
dv
dx
= sin x
so from the product rule we have
df
dx
=
d (uv)
dx
=
du
dx
v +u
dv
dx
= 3x
2
cos x +x
3
(sin x)
= 3x
2
cos x x
3
sin x.
Example 1.6 For
f(x) =
e
x
sin x
calculate f

(x).
Solution: Setting
u = e
x
and v = sin x
we have
du
dx
= e
x
and
dv
dx
= cos x.
Hence from the quotient rule we have
df
dx
=
d
_
u
v
_
dx
=
du
dx
v u
dv
dx
v
2
=
e
x
sin x e
x
cos x
sin
2
x
.
Example 1.7 For
f(x) = cos 3x
2
calculate f

(x).
6
Solution: Setting u = 3x
2
we have f(x) = cos 3x
2
= cos u and hence
du
dx
= 6x and
df
du
=
d cos u
du
= sin u.
Thus from the chain rule we have
df
dx
=
df
du
du
dx
= sin u 6x = 6xsin u = 6xsin 3x
2
.
Remarks
We note the following useful identities
e
a+b
= e
a
e
b
, e
ab
= (e
a
)
b
and for a, b > 0
ln(ab) = ln a + ln b, ln
_
a
b
_
= ln a ln b and ln b
a
= a ln b.
1.6 Integration
The following table gives the integrals of some common functions.
f(x)
_
f(x)dx
kx
n k
n+1
x
n+1
+c, n = 1
sin kx -
1
k
cos kx +c
cos kx
1
k
sin kx +c
e
kx 1
k
e
kx
+c
k
x
k ln |x| +c
1.7 Integration by parts
Recalling the product rule we have
d
dx
(fg) = f

g +fg

and hence by integration we obtain


_
b
a
d
dx
(fg) dx =
_
b
a
f

gdx +
_
b
a
fg

dx
[fg]
b
a
=
_
b
a
f

gdx +
_
b
a
fg

dx.
Manipulating the above equation gives the standard form of integration by parts formula
_
b
a
f

gdx = [fg]
b
a

_
b
a
fg

dx.
Example 1.8 Use integration by parts to integrate
_
2
1
xe
x
dx.
7
Solution: We have
_
2
1
xe
x
dx = [xe
x
]
2
1

_
2
1
e
x
dx, g = x g

= 1, f

= e
x
f = e
x
= [xe
x
]
2
1
[e
x
]
2
1
= 2e
2
e
_
e
2
1
_
= e
2
.
1.7.1 Integration using substitution
_
b
a
f(x)dx =
_
u(b)
u(a)
f(u)du
dx
du
=
_
u(b)
u(a)
_
f(u)
dx
du
_
du.
Example 1.9 Use integration using substitution to evaluate the following integral
_
1
0
x
_
1 +x
2
_
3
dx.
Solution: Setting u = 1 +x
2
we have
du
dx
= 2x
dx
du
=
1
2x
, u(0) = 1 + 0
2
= 1 and u(1) = 1 + 1
2
= 2
and hence
_
1
0
x
_
1 +x
2
_
3
dx =
_
u(1)
u(0)
xu
3
dx
du
du,
=
_
2
1
xu
3
1
2x
du
=
1
2
_
2
1
u
3
du
=
1
2
_
u
4
4
_
2
1
=
1
8
_
2
4
1
4

=
15
8
.
1.8 Polar coordinates
There are two unit vectors associated with polar coordinates: e

and e

. They have magnitude 1 and


point in the directions of increasing and respectively, also they are perpendicular so that
e

.
Remark 1 We note that e

and e

are position dependent unlike i, j and k (i.e. the direction in which


they point depends on where they are in the (x, y) plane).
8
r
f
x
y
i
j
= cos
= sin
=

lan =

Figure 3:
1.9 3-D coordinate systems
In 3-d there are three commonly used coordinate systems:
coordinate system variables unit vectors
Cartesian (x, y, z) (i, j, k)
Cylindrical (, , z) (e

, e

, e
z
)
Spherical (r, , ) (e
r
, e

, e

)
P(r,f,z)
= cos
= sin
=
lan =

r
z
x
y
f
z
Figure 4:
1.9.1 Cylindrical polar coordinates
A point P(x, y, z) can be written in terms of the cylindrical polar coordinates , , z such that P(x, y, z) =
P(, , z). One can think of P (, , z) as being a point on the surface of a cylinder with radius , see
Figure 4.
9
Remark 2 The unit vector e
z
points in the increasing z-direction (and is in fact k !).
f
x
y
z
r
r sinq
r sinq sinf
r sinq cosf
rcosq
P(r,q,f)



q
Figure 5:
1.9.2 Spherical polar coordinates
A point P(x, y, z) can be written in terms of the spherical polar coordinates r, , such that P(x, y, z) =
P(r, , ). One can think of P (r, , ) as being a point on the surface of a sphere with radius r, see Figure
5.
Remark 3 The unit vectors are such that e
r
points in the increasing r-direction, e

points in the in-


creasing -direction, e

points in the increasing -direction.


10
2 2D Surface Integrals
2.1 Cartesian coordinates
The area A under a graph y = f(x) between two points x = a and x = b (see Figure 6) can be calculated
by evaluating the integral
A =
_
b
a
f(x)dx.
The area A can be approximated by the sum of the areas of N thin strips S
i
centred at x
i
with width
x and height f(x
i
) (see Figure 6)) hence we have
A

area of strips S
i
=
N

i=1
f(x
i
)x.
In the limit as the x 0 i.e. as the number of strips tends to innity we have
N

i=1
f(x
i
)x
_
b
a
f(x)dx = A.
Alternatively we could divide each strip into blocks with centre (x
i
, y
j
), width x and height y
A
N

i=1
(area of S
i
)
where
area of S
i
=
Mi

j=1
(area of B
j
).
Hence
A
N

i=1
(area of S
i
)
=
N

i=1
Mi

j=1
(area of B
j
)
=
N

i=1
Mi

j=1
yx
. .
area of Bj
.
Note that the number of blocks M
i
, depends on the height of S
i
, i.e. the value of f(x
i
). In the limit as
y=f(x)
a b
A
f(x
i
)
x
i
b a
y=f(x)
Figure 6:
11
the number of blocks tends to innity and x and y tend to zero we have
N

i=1
Mi

j=1
yx
. .
area of Bj

_
b
a
_
f(x)
0
dydx = A.
In general the area of a 2D surface S can be approximated by
A =

(area of the surface elements)


=

dA.
In the limit as the area of the surface element tends to zero we have
A =
_
S
dA.
Here the integral
_
S
is a double integral i.e.
__
, and the dA is a double integrand i.e. dxdy
(1,0)
(0,1)
(0,0)
T
Figure 7:
Example 2.1 Calculate the area of the triangle T bounded by the lines x = 0, y = 0 and y = 1 x.
Solution:
First we draw the triangle T, see Figure 7. We have that
area of T =
_
T
dA =
_ __
dy
_
dx
The inner integral relates to the area of a strip with width x. The base of this strip is at y = 0, the top
varies depending on the position that the strip is along the x-axis (i.e. it is a function of x).
From Figure 7 we see that the top of the strip is 1 x, so the limits on the inner integral are
area of T =
_ __
1x
0
dy
_
dx.
For the limits on the outer integral we recall that this integral is associated with the sum of the strips.
The left hand strip is centred at x = 0 and the right hand one is at x = 1, so we have
area of T =
_
1
0
_
1x
0
dydx.
To evaluate a double integral we rst evaluate the inner integral and then we evaluate the resulting
integral
12
A =
_
1
0
__
1x
0
dy
_
dx =
_
1
0
_
y

1x
0
dx
=
_
1
0
[(1 x) 0] dx
=
_
1
0
(1 x)dx =
_
x
x
2
2
_
1
0
=
1
2
.
2.1.1 Polar coordinates
rDf
r
Dr
Df
Figure 8:
If we want to nd the area of a circular or part circular surface we use surface elements derived from
polar coordinates.
The area of a surface element can be approximated by (see Figure 8)
A .
Thus the area of the surface
A

(area of the surface elements)

.
In the limit as the area of each surface element tends to zero we have
A =
_
S
dA =
_ _
. .
S
dd
. .
dA
.
The limits on the integrals depend on the surface that you are integrating over.
13
Example 2.2 Calculate the area of the annular region S by 0 /2 and 1 2.
Solution: First we draw the annular region S, see Figure 9.
(1,0) (2,0)
(0,2)
(0,1) S
Figure 9:
We see that
area of S =
_
S
dA =
_ __
d
_
d
From Figure 9 we see that the limits for the integral are 1 and 2, while the limits for the integral are
/2 and /2. Thus
area of S =
_
2
0
_
2
1
dd
=
_
2
0
__
2
1
d
_
d
=
_
2
0
_

2
2
_
2
1
d
=
_
2
0
_
2
1
2
_
d
=
_
2
0
3
2
d
=
3
2
[]

2
0
=
3
2
( 0)
=
3
4
.
2.2 Calculating mass and charge density
The techniques that we have used to calculate the area of surfaces can be applied to calculate the mass
of a 2D object or the charge on a 2D object.
14
If the density of the rectangular plate S bounded by 0 x 2 and 0 y 1 is given by f(x, y) =
(1 + x
2
)(1 + y) kgm
2
we can approximate the mass of S by dividing it up into small surface elements
and summing the masses of each of these elements.
Since for an object with constant density, mass = area x density, we can approximate the mass of a
surface element by its area multiplied by the density at its mid-point
i.e. (mass of S.E.) xyf(x
i
, y
j
)
Thus we have
S =

(mass of S.E.) =

i,j
xyf(x
i
, y
j
)
_ _
f(x, y)dydx.
Adding the required limits for x and y gives
Mass of S =
_
2
0
_
1
0
(1 +x
2
)(1 +y)dydx
=
_
2
0
__
1
0
(1 +x
2
)(1 +y)dy
_
dx
=
_
2
0
_
(1 +x
2
)
_
1
0
(1 +y)dy
_
dx
=
_
2
0
(1 +x
2
)
_
y +
y
2
2
_
1
0
dx
=
_
2
0
(1 +x
2
)
__
1 +
1
2
_
0
_
dx
=
_
2
0
3
2
(1 +x
2
)dx
=
3
2
_
x +
x
3
3
_
2
0
= 7 kg.
Example 2.3 If the charge density on the semi circular disc S in Figure 10 is f(, ) = 3 Cm
2
,
calculate the total electric charge on S.
Solution: We have
total charge onS =

i
charge on S.E.
charge on S.E. area of S.E. charge density at the mid-point of S.E.(
i
,
j
)
= f(
i
,
j
).
So
15
(0,1)
(0,1)
(1,0)
S
Figure 10:
total charge onS =
_
2

2
_
1
0
f(, )dd
=
_
2

2
_
1
0
3dd
=
_
2

2
__
1
0
3
2
d
_
d
=
_
2

2
_

1
0
d
=
_
2

2
d
= []

2
= C.
2.3 General Formulas
We recall the following useful formulas.
Area of S =
_
S
dA,
Mass of S =
_
S
fdA where f = density,
charge on S =
_
S
fdA where f = charge density .
16
2.4 Double integrals from a mathematical point of view
So far we have been looking at things from a physical point of view. We could have asked the previous
questions in the following mathematical ways:
Example 2.4 Evaluate the integral of the function f(x, y) = (1 + x
2
)(1 + y) over the two dimensional
rectangular region with 0 x 2, 0 y 1.
Solution:
_
2
0
_
1
0
f(x, y)dydx =
_
2
0
_
1
0
(1 +x
2
)(1 +y)dydx = 7.
Example 2.5 Find the integral of f(, ) = 3 over the semi-circular region 0 1,

2


2
.
Solution:
_
2

2
_
1
0
3dd = .
17
3 Surface Areas in 3D
To evaluate the area of a 3D surface we use
A =
_
S
dA.
As in the case of 2D surface integrals
_
S
is a double integral i.e.
__
, and dA is a double integrand i.e.
dxdy.
3.1 Cartesian coordinates
If we wish to evaluate the surface area (or the mass, or the charge) of the hollow block 0 x 7,
0 y 5 and 1 z 3, we need to break the block up into six 2D surfaces, i.e. its six faces.
On the top and bottom faces of the block x and y vary, and z is constant, so we have
area of top =
_
5
0
_
7
0
dxdyand area of bottom =
_
5
0
_
7
0
dxdy.
On the left and right hand faces x and z vary and y is constant, so we have
area of left hand face =
_
3
1
_
7
0
dxdzand area of right hand face =
_
3
1
_
7
0
dxdz.
On the remaining two faces y and z vary, and x is constant, so
area of front =
_
3
1
_
5
0
dydzand area of back =
_
3
1
_
5
0
dydz.
Example 3.1 Find the integral of f(x, y, z) = 1 +x +y +z over the surface of the hollow block S.
18
Solution: We have
_
S
fdA =
_
top
fdA+
_
bottom
fdA+
_
left side
fdA+
_
right side
fdA+
_
front
fdA+
_
back
fdA
=
_
5
0
_
7
0
_
1 +x +y + z
(=1)
_
dxdy +
_
5
0
_
7
0
_
1 +x +y + z
(=3)
_
dxdy
+
_
3
1
_
7
0
_
1 +x + y
(=0)
+z
_
dxdz +
_
3
1
_
7
0
_
1 +x + y
(=5)
+z
_
dxdz
+
_
3
1
_
5
0
_
1 + x
(=0)
+y +z
_
dydz +
_
3
1
_
5
0
_
1 + x
(=7)
+y +z
_
dydz
=
_
5
0
_
7
0
(1 +x +y + 1) dxdy +
_
5
0
_
7
0
(1 +x +y + 3) dxdy
+
_
3
1
_
7
0
(1 +x +z) dxdz +
_
3
1
_
7
0
(1 +x + 5 +z) dxdz
+
_
3
1
_
5
0
(1 +y +z) dydz +
_
3
1
_
5
0
(1 + 7 +y +z) dydz
=
_
5
0
_
7
0
(6 + 2x + 2y) dxdy +
_
3
1
_
7
0
(7 + 2x + 2z) dxdz +
_
3
1
_
5
0
(9 + 2y + 2z) dydz
=
_
5
0
_
6x +x
2
+ 2yx

7
0
dy +
_
3
1
_
7x +x
2
+ 2zx

7
0
dz +
_
3
1
_
9y +y
2
+ 2zy

5
0
dz
=
_
5
0
(91 + 14y) dy +
_
3
1
(98 + 14z) dz +
_
3
1
(70 + 10z) dz
=
_
91y + 7y
2

5
0
+
_
91z + 7z
2

3
1
+
_
70z + 5z
2

3
1
= 630 + 336 98 + 255 75 = 1048.
3.2 Cylindrical polar coordinates
a
x
y
z
4
0
Figure 11:
What about if we wish to nd the mass of a hollow cylinder S in Figure 11 whose surface density is
f(, , z) = 1 +z
2
?
We use Mass =
_
S
fdA, but what is dA? First we split the surface of the cylinder into three parts; the
top, the base and the curved surface (C.S.), so
mass =
_
top
fdA+
_
base
fdA+
_
C.S.
fdA.
19
The top and the base are 2D surfaces, so as in Section 2 we have (see Figure 12)
_
top
fdA =
_
2
0
_
a
0
fdd and
_
base
fdA =
_
2
0
_
a
0
fdd
with z constant on both surfaces.
On the curved surface we have is constant, and in fact = a where a is the radius of the cylinder.
DA1 = (r Df) (Dz)
dA1 = r df dz
DA1
DA2
DA2 = (r Df) (Dr)
dA2 = r df dr
Df
r Df
Dr
r
z
x
y
Dz
r Df
Figure 12:
We know that dA is related to the area of a surface element on the curved surface of the cylinder see
gure 12.
So area = az and hence dA = addz.
Thus
_
C.S.
fdA =
_
4
0
_
2
0
faddz.
So
mass =
_
2
0
_
a
0
fdd +
_
2
0
_
a
0
fdd +
_
4
0
_
2
0
faddz.
Since f = 1 +z
2
we have
mass =
_
2
0
_
a
0
_
1 + z
2
(=16)
_
dd +
_
2
0
_
a
0
_
1 + z
2
(=0)
_
dd +
_
4
0
_
2
0
_
1 +z
2
_
addz
=
_
2
0
_
a
0
17dd +
_
2
0
_
a
0
dd +
_
4
0
_
2
0
_
1 +z
2
_
addz
=
_
2
0
17
2
a
2
d +
_
2
0
a
2
2
d +
_
4
0
2a
_
1 +z
2
_
dz
= 17a
2
+a
2
+ 2a
_
z +
z
3
3
_
4
0
= 18a
2
+ 50
2
3
a.
20
3.3 Spherical polar coordinates
DA= (r sinq Df) (r Dq)
dA= r
2
sinq dq df
Df
x
y
z
r Dq
r sinq Df
Dq
Figure 13:
If we want to evaluate the integral of f(r, , ) over the surface of a sphere with radius R we note that
since r is constant (= R) and and vary on the surface of the sphere we have (see Figure 13)
area of S.E. = R Rsin
= R
2
sin
and hence
dA = R
2
sin dd.
Remark 4 The angle only varies from 0 to , while the angle varies from 0 to 2 (for the whole
surface of a sphere).
Example 3.2 Given a hollow sphere with centre at the origin, radius 2, with charge density f(R, , ) =
sin
2
+ cos
2
, calculate the total charge on the surface of the sphere.
Solution: We have
total charge =
_
S
fdA (f is charge density)
=
_

0
_
2
0
_
sin
2
+ cos
2

_
R
2
sin dd
=
_

0
_
2
0
1 2
2
sin dd
=
_

0
_
2
0
4 sin []
2
0
d
= 8
_

0
sin d
= 16.
21
3.4 Scalar and vector elds
A scalar (vector) eld is a distribution of scalar values (vector values) on/in a specied surface/region in
space, such that there is a unique scalar (vector) associated with each point on/in the surface/region.
Examples of scalar elds are temperature in a room or pressure in a room.
Examples of vector elds are magnetic ux around a bar magnet or water velocity on the surface of a
river.
3.5 Flux of a vector eld
The ux of a vector eld F = F
x
i+F
y
j+F
z
k (F = F

+F

+F
z
e
z
, F = F
r
e
r
+F

+F

),
through a surface S is equal to the surface integral of the component of F normal to the surface.
Mathematically we have that
Flux = =
_
S
F
n
dA =
_
S
F ndA
where F
n
denotes the component of F that points in the direction of n.
If n points out of the surface we have outward ux, if n point in to the surface we have the inward
ux.
Example 3.3 Evaluate the outward ux of F = 3e

2e
z
through the base of a cylinder centred at the
origin, with height H and radius R.
Solution:
On the base of the cylinder n =e
z
and hence
F n = (3e

2e
z
) (e
z
)
= (3)(0) + (0)(0) + (2)(1) = 2
Since
Flux = =
_
S
F ndA
we have
=
_
S
2dA.
But what is dA? On the base of a cylinder and vary and we have dA = dd so,
=
_
2
0
_
R
0
2dd =
_
2
0
_
2

2
2
_
R
0
d = 2R
2
.
Alternatively, since we know that the area of S =
_
S
dA, we have = 2 area of S = 2R
2
.
If we want to calculate the outward ux of G = ze

+ cos()e

e
z
through the curved surface of the
cylinder in Example 3.3, we have n = e

and so
G n = (2ze

+ cos()e

e
z
) e

= (2z)(1) + cos()(0) + (1)(0)


= 2z
22
and hence the outward ux
=
_
S
2zdA
=
_ H
2

H
2
_
2
0
2zRddz
=
_ H
2

H
2
_
2
0
2RzRddz
=
_ H
2

H
2
_
2
0
2R
2
zddz
=
_ H
2

H
2
_
2R
2
z

2
0
dz
=
_ H
2

H
2
4R
2
zdz
= 4R
2
_
z
2
2
_
H
2

H
2
= 0.
23
4 Volume Integrals
4.1 Cartesian, cylindrical polar and spherical polar coordinates
DV= (r Df) (Dr) (Dz)
dV= rdr df dz
Df
rDf
Dr
Dz
r
z
x
y
Figure 14:
The integral of a function g(x, y, z), or g(, , z) or g(r, , ), over a 3D object is given by
_
object
gdV
where dV is the limit of the volume of a small volume element in the object.
In Cartesian coordinates dV = dxdydz.
In cylindrical polar coordinates dV = dddz (see Figure 14).
In spherical polar coordinates dV = r
2
sin()dddr (see Figure 15).
24
DV= (r sinq Df) (r Dq) (Dr)
dV= r
2
sinq dr dq df
x
y
z
Dr
r Dq
r sinq Df
Dq
Df
Figure 15:
Volume integrals can be used to calculate the volume of 3D objects, or the mass or charge of/in a 3D
object. Also they can be used to calculate the moment of inertia of a 3D object. For the objects in Figure
16 we have:
y
z
x
R
0
y
z
x
a
H
0
y
z
x
a
0
b
c
Figure 16:
In cartesian coordinates
_
V
gdV =
_
c
0
_
b
0
_
a
0
gdxdydz.
In cylindrical polar coordinates
_
V
gdV =
_
H
0
_
2
0
_
a
0
gdddz.
In spherical polar coordinates
_
V
gdV =
_
2
0
_

0
_
R
0
gr
2
sin()drdd.
Example 4.1 Evaluate the volume integral of g(r, , ) = 4r cos(/8) over a sphere centred at the origin
with radius 2.
Solution: We have
_
sphere
f(, , z)dV =
_
2
0
_

0
_
2
0
g(r, , )r
2
sin()drdd
25
with g(r, , ) = 4r cos(/8). Thus
_
sphere
f(, , z)dV =
_
2
0
_

0
_
2
0
4r
3
cos(/8) sin()drdd
=
__
2
0
cos(/8)d
___

0
sin()d
___
2
0
4r
3
dr
_
= [8 sin(/8)]
2
0
[cos()]

0
_
r
4

2
0
= 8
_
1

2
0
_
((1 1)) (16 0)
=
256

2
.
4.2 Moment of Inertia
The moment of inertia of a mass m about the z-axis is given by I = m
2
, where =
_
x
2
+y
2
is the
distance from the z-axis.
The moment of inertia of a solid object V about the z-axis is equal to the sum of the moments of inertia
of the individual volume elements in the object.
Since the volume elements are small we can approximate their mass by their volume multiplied by the
density at their mid points p
i
. Also we can approximate their distance from the z-axis by the distance of
their mid points p
i
from the z-axis.
Hence the moment of inertia of a volume element is approximated by
I
i
=
2
Mass f(p
i
)V
i

2
i
.
Thus we have that the moment of inertia I is such that
I
N

i=1
I
i
=
N

i=1
f(p
i
)V
i

2
i

_
object
f
2
dV.
Example 4.2 Calculate the moment of inertia of a cylinder with radius 1, height H, centre the origin
and density f = z
2
.
Solution: We have
I =
_ H
2

H
2
_
2
0
_
1
0

2
z
2
dddz
=
_ H
2

H
2
_
2
0
_
1
0

4
z
2
dddz
=
_ H
2

H
2
_
2
0
_
1
0
_

5
5
_
1
0
z
2
ddz
=
_ H
2

H
2
_
2
0
1
5
z
2
ddz
=
_ H
2

H
2
2
5
z
2
dz
=
_
2
15
z
3
_H
2

H
2
=
H
3
120
.
26
5 Line Integrals
5.1 Work done by a force
A
B
C
Figure 17:
Line integrals are used to calculate the work done by a force F (or a vector eld) F = F
x
i+F
y
j+F
z
k, on
a particle in moving it along a curve C =

AB see Figure 17.


The work done by a constant force F on a body undergoing linear displacement is given by the distance
|AB| multiplied by the component of F in the direction AB, i,e, in the direction S, see Figure 18.
S
e
q
A B
S
F
Figure 18:
Thus we have
W = F S = |F| |S| cos()
and since
F = F
s

S +F
s

with
F
s
= |F| cos()
we have
W = F
s
|S| .
For a non-constant force F acting on a particle moving along a non-linear path C we can approximate
the work done by approximating the path C by small linear path segments, and then on each of these
27
Figure 19:
segments we can approximate the force F by the value of F at the start of each linear segment (i.e. by a
constant value).
On the i
th
segment along the curve that has start point at r
i
= x
i
i + y
i
j + z
i
k and at end point
r
i+1
= x
i+1
i + y
i+1
j + z
i+1
k (see gure 19), we have that the work done by the constant force F(r
i
) in
moving a particle along the segment is
W
i
= F(r
i
) r
i
.
So
W

i
W
i
=

i
F(r
i
) r
i
.
In the limit as the size of the line segments tends to zero, we have
W =
_
C
F(r) dr.
Since F = F
x
i+F
y
j+F
z
k and r = xi +yj +zk we have
F dr = (F
x
i+F
y
j+F
z
k) (dxi +dyj +dzk)
= F
x
dx +F
y
dy +F
z
dz.
Hence
W =
_
C
F
x
dx +F
y
dy +F
z
dz. (5.5)
We need to write (5.5) as a denite integral involving a single variable, say x, y, z, t or .
5.2 Line integrals in the x-y plane
Example 5.1 Calculate the work done by F = 3yi 5xj+100xyk on a particle moving along the path
AB shown in Figure 20.
Solution: On the path AB we have y =
1
2
x and z = 3 and hence
dy
dx
=
1
2
and
dz
dx
= 0.
We can write (5.5) as a denite integral in x:
W =
_
C
F
x
dx +F
y
dy +F
z
dz =
_
X
B
X
A
(F
x
(x)
dx
dx
+F
y
(x)
dy
dx
+F
z
(x)
dz
dx
)dx
28
y=x/2
plane z=3
A
B
4
x
y
2
Figure 20:
where X
A
is the x coordinate at the start point A and X
B
is the x coordinate at the end B.
Thus X
A
= 0 and X
B
= 4. Furthermore since
F
x
(x) = 3y =
3
2
x, F
y
(x) = 5x, F
z
(x) = 100xy = 50x
and
dx
dx
= 1,
dy
dx
=
1
2
and
dz
dx
= 0
we have
W =
_
4
0
__
3
2
x
_
(1) (5x)
_
1
2
_
+ (50x)(0)
_
dx
=
_
4
0
xdx =
_

x
2
2
_
4
0
= 8.
Remark 5
1. The line integral of a vector eld F along a path P = P
1
+P
2
can be written as
_
P
F dr =
_
P1
F dr+
_
P2
F dr.
2.
_
B
A
F dr =
_
A
B
F dr.
Example 5.2 Calculate the work done by F = 3xi + 5xj 2k on a particle moving along the path
Ac = AB +BC shown in Figure 21.
Solution: We have
_
C
A
F dr =
_
B
A
F dr+
_
C
B
F dr
=
_
B
A
F
x
dx +F
y
dy +F
z
dz +
_
C
B
F
x
dx +F
y
dy +F
z
dz.
On path P
1
we have
x = 1, z = 0
dx
dy
= 0 and
dz
dy
= 0
29
plane x=-1
A(0,0) B(2,0)
C(2,1)
P
1
P
2
y
z
Figure 21:
and on path P
2
we have
x = 1, y = 2
dx
dz
= 0 and
dy
dz
= 0.
Thus we have
_
B
A
F dr =
_
Y
B
Y
A
(F
x
(y)
dx
dy
+F
y
(y)
dy
dy
+F
z
(y)
dz
dy
)dy
_
C
B
F dr =
_
Z
C
Z
B
(F
x
(z)
dx
dz
+F
y
(z)
dy
dz
+F
z
(z)
dz
dz
)dz

_
B
A
F dr =
_
2
0
((3x)(0) + (5x)(1) (2)(0))dy =
_
2
0
5dy = 10

_
C
B
F dr =
_
1
0
((3x)(0) + (5x)(0) (2)(1))dz =
_
1
0
2dz = 2
and hence
_
C
A
F dr =10 2 = 12.
Example 5.3 Calculate the work done by F = yi +xj on moving a particle along the path AB shown
in Figure 22.
Solution: The work done is given by
W =
_
B
A
F dr =
_
B
A
F
x
dx +F
y
dy =
_
X
B
X
A
(F
x
(x)
dx
dx
+F
y
(x)
dy
dx
)dx
and since
y = (4 x
2
)
1
2

dy
dx
=
1
2
(4 x
2
)(2x) =
x
(4 x
2
)
1
2
we have
W =
_
X
B
X
A
_
(4 x
2
)
1
2
(1) + (x)
_
x
(4 x
2
)
1
2
__
dx
=
_
2
0
_
(4 x
2
)
1
2
x
2
(4 x
2
)

1
2
_
dx.
This is not an easy integral to solve, so hopefully there is a better way of calculating W.
30
x
A(2,0)
B(0,2)
z=constant
y
Figure 22:
P
1
=(x
1
,y
1
,z
1
)
P
2
=(x
2
,y
2
,z
2
)
Figure 23:
5.3 Parametric Equations
Any path in 3D may be expressed in terms of three parametric equations x(t), y(t) and z(t) that involve
a parameter t that varies from t
1
to t
2
such that
x(t
1
) = x
1
, y(t
1
) = y
1
, z(t
1
) = z
1
, x(t
2
) = x
2
, y(t
2
) = y
2
and z(t
2
) = z
2
.
We have (see Figure 23)
_
P2
P1
F dr =
_
t2
t1
(F
x
(t)
dx
dt
+F
y
(t)
dy
dt
+F
z
(t)
dz
dt
)dt.
Example 5.3 (continued)
Using polar coordinates we can write the curve C = AB in terms of three parametric equations
x = 2 cos , y = 2 sin , z = constant

dx
d
= 2 sin ,
dy
d
= 2 cos ,
dz
d
= 0.
When = 0 x = 2, y = 0 and z = constant and hence we are at point A. While when =

2
x = 0,
y = 2 and z = constant hence we are at point B.
Thus we have
31
_
B
A
F dr =
_

B

A
(y
dx
d
+x
dy
d
)d
=
_
2
0
[(2 sin )(2 sin ) + (2 cos )(2 cos )] d
=
_
2
0
(4 sin
2
+ 4 cos
2
)d
=
_
2
0
4d = 2.
Example 5.4 Evaluate the integral of F = 2xi + 3zj 5k along the curve given parametrically by
x(t) = t, y(t) = 2t, z(t) = t
2
where t varies from 0 to 1.
Solution: Since
x(t) = t, y(t) = 2t, z(t) = t
2

dx
dt
= 1,
dy
dt
= 2,
dz
dt
= 2t
we have
_
C
F dr =
_
t2
t1
(F
x
(t)
dx
dt
+F
y
(t)
dy
dt
+F
z
(t)
dz
dt
)dt
=
_
1
0
[(2x)(1) + (3z)(2) + (5)(2t)] dt
=
_
1
0
(2t 6t
2
+ 10t)dt
=
_
1
0
(12t 6t
2
)dt
=
_
6t
2
2t
3

1
0
= 4.
5.4 Path dependence/independence
A line integral of a eld/force F from a point A to a point B is said to be path dependent if it depends
on the path taken to get from A to B, otherwise it is said to be path independent. If the line integral of
F is path dependent F is non conservative.
Example 5.5 By considering two paths that start from A = (2, 0) and end at B = (0, 2) (see Figure 24),
show that the line integral of the vector eld F = yi +xj is path dependent.
Solution: From Example 5.3 we have that
_
P2
F dr = 2.
While on P
2
we have
_
P2
F dr =
_
B
A
F
x
dx +F
y
dy =
_
X
B
X
A
(F
x
(x)
dx
dx
+F
y
(x)
dy
dx
)dx.
32
x
A(2,0)
B(0,2)
y
P
1
P
2
Figure 24:
Since the path P
2
is dened by y = 2 x we have
dy
dx
= 1 and hence
_
P2
F dr =
_
0
2
[(y)(1) + (x)(1)] dx
=
_
0
2
[(2 x) x] dx
=
_
0
2
2dx = [2x]
0
2
= 4.
So
_
P1
F dr =
_
P2
F dr and hence the integral of F is path dependent which tells us that F is non-
conservative.
5.5 Closed curves (loops)
rA
rB
C
1
C
2
Figure 25:
For a closed curve C that comprises of the path C
1
from r
A
to r
B
(see Figure 25) and the path C
2
from
r
B
to r
A
, we have that the line integral of F around the closed curve (loop) C is given by
_
C
F dr=
_
C1
F dr+
_
C2
F dr.
The circle around the integral of C indicates that C is a closed curve.
33
5.6 Conservative elds (or forces)
Remark 6 1. The line integral of a conservative eld from a point A to a point B is path independent,
i.e. it only depends on the initial point A and the end point B.
2. The line integral of a conservative eld around any closed loop is always equal to zero, i.e.
_
C
F dr =0 (5.6)
for a conservative eld F and for any closed curve C.
3. If
_
C
G dr =0 then G is non conservative.
4. If
_
C
G dr =0 then G is might be conservative.
Example 5.6 Show that a conservative force F satises (5.6).
Solution:
Since
_
C
F dr=
_
C1
F dr+
_
C2
F dr and
_
B
A
F dr =
_
A
B
F dr, and for a conservative eld
_
C1
F
dr =
_
C2
F dr, we have the following for any conservative eld.
_
C
F dr =
_
C1
F dr+
_
C2
F dr
=
_
C2
F dr+
_
C2
F dr = 0.
x
y
A(0,0)
B(1,1)
C
2
P(1,0)
Figure 26:
Example 5.7 Calculate the line integral of F = x
2
i + yj along the two paths C
1
= AP + PB and C
2
shown in Figure 26. Say if the eld F could be a conservative eld.
Solution:
The path C
1
consists of the two straight line paths AP and PB. On the rst path AP, y = 0 and x
varies from 0 to 1
dy
dx
= 0.
On the second path PB, x = 1 and y varies from 0 to 1
dx
dy
= 0.
Hence we have
34
_
C1
F dr =
_
P
A
F dr+
_
B
P
F dr
=
_
X
P
X
A
(F
x
(x)
dx
dx
+F
y
(x)
dy
dx
)dx +
_
Y
B
Y
P
(F
x
(y)
dx
dy
+F
y
(y)
dy
dy
)dy
=
_
1
0
_
(x
2
)(1) + (y)(0)

dx +
_
1
0
_
(x
2
)(0) + (y)(1)

dy
=
_
1
0
x
2
dx +
_
1
0
ydy
=
_
x
3
3
_
1
0
+
_
y
2
2
_
1
0
=
1
3
+
1
2
=
5
6
.
On C
2
, y = x
dy
dx
= 1
_
C2
F dr =
_
A
B
F
x
dx +F
y
dy
=
_
X
A
X
B
(F
x
(x)
dx
dx
+F
y
(x)
dy
dx
)dx
=
_
0
1
_
(x
2
)(1) + (y)(1)

dx
=
_
0
1
(x
2
+x)dx
=
_
x
3
3
+
x
2
2
_
0
1
=
5
6
.
Since the path C
1
+C
2
is a closed curve with
_
C1+C2
F dr =
_
C1
F dr+
_
C2
F dr
=
5
6

5
6
= 0
it follows that F may be conservative.
35
6 Partial derivatives
t
x
F(x,t)
(x,t)
(x,t+h)
(x+h,t)
Figure 27:
The partial derivative of a function F(x, y, z, t, ...) with respect to x is dened by
F
x
and is evaluated by
treating the other variables as constant and dierentiating with respect to x. Consider a function of two
variables F(x, t), the partial derivatives
F
x
and
F
t
at a given point (x, t) of a function F(x, t), represent
the slopes of the 3D graph y = F(x, t) in the direction of the x-axis and the t-axis at the given point
(x, t) (see gure 27). We dene
F
x
= lim
h0
F(x +h, t) F(x, t)
h
(6.7)
and
F
t
= lim
h0
F(x, t +h) F(x, t)
h
. (6.8)
Example 6.1 Given that f(x, y, t) = 3x
2
te
y
nd
f
x
,
f
t
and
f
y
.
Solution: We have
f
x
= 3te
y
d
dx
x
2
= (3te
y
)(2x) = 6te
y
x
f
t
= 3x
2
e
y
d
dt
t = 3x
2
e
y
f
x
= 3x
2
t
d
dy
e
y
= 3x
2
te
y
.
6.1 The chain rule for partial derivatives
For a function f(u(x, y, t)) we have

x
f(u(x, y, t)) =
df
du
u
x
,

y
f(u(x, y, t)) =
df
du
u
y
and

t
f(u(x, y, t)) =
df
du
u
t
. (6.9)
Example 6.2 For f = cos(3x
2
y) nd
f
x
.
Solution: Setting u = 3x
2
y we have
cos(3x
2
y) = cos u with
df
du
= sin u, and
u
x
= 6xy.
36
Hence from the chain rule (6.9) we have
f
x
= 6xy sin(3x
2
y).
6.2 Higher order partial derivatives
From the example above we see that partial derivatives of a function F(x, y, t) are also functions of x, y
and t, and so can be partially dierentiated themselves.
Second order partial derivatives take the following form

2
F
x
2
=

x
F
x

2
F
yx
=

y
F
x

2
F
y
2
=

y
F
y

2
F
xy
=

x
F
y
.
Example 6.3 For F(x, y) = 3e
x
cos 2y calculate
F
x
,
F
y
,

2
F
x
2
,

2
F
yx
,

2
F
y
2
and

2
F
xy
.
Solution: We have
F
x
= 3e
x
cos 2y,
F
y
= 6e
x
sin 2y
and

2
F
x
2
=

x
F
x
=

x
(3e
x
cos 2y) = 3e
x
cos 2y

2
F
yx
=

y
F
x
=

y
(3e
x
cos 2y) = 6e
x
sin 2y

2
F
y
2
=

y
F
y
=

y
(6e
x
sin 2y) = 12e
x
cos 2y

2
F
xy
=

x
F
y
=

x
(6e
x
sin 2y) = 6e
x
sin 2y.
Remark 7 It is always the case that

2
F
xy
=

2
F
yx
.
37
7 Directional derivatives
u
P
Q
R
0 cm 20 cm 40 cm
0
o
C
10
o
C 20
o
C
q
n
Figure 28:
The spatial rate of change of a scalar eld in a specied direction is known as a directional derivative.
Example 7.1
Consider a 40 cm thick wall that has a temperature of 0

C on one side and 20

C on the other, with the


temperature varying uniformly within the wall (see Figure 28).
The rate of change of temperature (T) in the direction specied by u is given by
change in T
distance in specied direction
=
(10 5)

C
PQ
=
5 cos
10

C
cm
= 0.5 cos

C
cm
. (7.10)
Since PR = PQcos we have PQ =
PR
cos
=
10
cos
cm.
The equation (7.10) is an example of a directional derivative. At any point P there is an innity of
directional derivatives.
A vector at P with magnitude equal to the largest directional derivative, and pointing in the direction
in which this largest directional derivative occurs, is known as the gradient vector or more commonly as
grad.
Remark 8 The directional derivative at P in the direction u can be obtained by taking the scalar product
of u with the gradient vector.
Example 7.2 What is grad T at P in the previous example? Also show that
u grad T = 0.5

C
cm
.
Solution: Since grad T has magnitude equal to the largest directional derivative of T we see that grad
T at P is when cos = 1 i.e. when = 0 and grad T points in the direction of n. So from (7.10) we have
grad T = 0.5n

C
cm
.
Since
u n =| u| |n| cos = cos
we have that
u grad T = 0.5 u n

C
cm
= 0.5 cos

C
cm
.
38
7.1 Directional derivative at a point
Dr
u ^
r
Q=r+Dr u ^
Figure 29:
The directional derivative of a scalar eld at a point at r in the direction u (see Figure 29) is given by

u
(r) = lim
r0
(r + r u) (r)
r
.
If we take the limit as r tends to zero we have the directional derivative of at the point at r in the
direction u.
The directional derivative of at a point at r = (x, y, z) in the direction of the x-axis, i.e. in the direction
of i is

i
(r) = lim
x0
(x + x, y, z) (x, y, z)
x
.
This is the partial derivative of with respect to x. Hence

i
(r) =
(r)
x
Similarly we have

j
(r) =
(r)
y
and

k
(r) =
(r)
z
7.2 The gradient vector, grad
We recall that any directional derivative in the direction u can be calculated by evaluating

u
= u grad .
So we have

i
= i grad ,

j
= j grad ,

k
= k grad
Since grad is a vector we can write it as
grad = ai +bj +ck
and hence

i
= i (ai +bj +ck) = a =

x

j
= j (ai +bj +ck) = b =

y

k
= k (ai +bj +ck) = c =

z
.
39
Thus we have
grad =

x
i +

y
j +

z
k. (7.11)
Example 7.3
1. Find the gradient of the scaler eld U =
1
2
(x
2
+y
2
+z
2
).
2. Evaluate the magnitude of grad U at the point P = (1, 2, 3).
3. Calculate the directional derivative of U at P in the direction s = i +j.
Solution:
1. Using (7.11) we have
grad U =
U
x
i +
U
y
j +
U
z
k =xi +yj +zk.
2.
|grad U(P)| =
_
1
2
+ 2
2
+ 3
2
=

14.
3. The directional derivative of U at P in the direction s = i +j is given by
U

s
(P) = grad U(P) s = (i + 2j + 3k)
i +j

2
=
1

2
+
2

2
=
3

2
.
7.3 Gradient and physical laws
The following physical laws involve the gradient vector.
F = grad U F - conservative force eld U - potential energy eld (7.12)
E = grad V E - electric eld V - electric potential (7.13)
h = k grad T h - heat ow T - temperature k - thermal conductivity (7.14)
Example 7.4 The electric potential V at a point is given by V (x, y, z) = x
2
e
y
(z+1). Find the magnitude
and direction of the electric eld vector E at the point (1, 0, 0).
Solution: Since
E = grad V =
V
x
i
V
y
j
V
z
k
we have
grad V = 2xe
y
(z + 1)i x
2
e
y
(z + 1)j x
2
e
y
k
and hence the direction of the electric eld E at the point (1, 0, 0) is
E(1, 0, 0) = grad V (1, 0, 0) = 2i j k
and its magnitude is
|E(1, 0, 0)| =
_
(2)
2
+ (1)
2
+ (1)
2
=

6.
40
7.4 Dierential operators
The dierential operator

x
is an example of a scaler dierential operator, it operates on a scalar eld T
to give another scalar eld
T
x
.
The dierential operator i

x
is an example of a vector dierential operator, it operates on a scalar eld
T to give a vector eld i
T
x
.
The expression for the gradient of a scalar eld involves a dierential operator.
grad f =
f
x
i +
f
y
j +
f
z
k
=
_
_
i

x
+j

y
+k

z
. .
_
_
(1)
f.
Here (1) =
_
i

x
+j

y
+k

z
_
is a vector dierential operator, it acts on the scalar eld f to give a
vector eld grad f
The dierential operator (1) is commonly known as del or nabla, and is denoted by the symbol
grad f = f =
f
x
i +
f
y
j +
f
z
k.
We can write the physical laws (7.12)-(7.14) in terms of dierential operator :
h = kT, E = V , F = U.
7.5 Other coordinate systems
We have seen that it is often easier to work in coordinate systems other than rectangular Cartesian. We
will, in fact, concentrate on two others, cylindrical polars and spherical polars. We will indicate here,
however, a general approach for nding dierential operators in other systems.
Suppose the new system (q
1
, q
2
, q
3
) has orthogonal vectors e
1
, e
2
, e
3
at each point ( eg e

, e

, e
z
). Then
we can write a small distance ds as ds
2
= dx
2
+ dy
2
+ dz
2
= h
2
1
dq
2
1
+ h
2
2
dq
2
2
+ h
2
3
dq
2
3
. For example,
in 2-d, we have ds
2
= d
2
+
2
d
2
. The coecients h
i
are called the metric, and generally in 3-d
space they will form a 3x3 matrix. They are a key component of Einsteins theory of General Rel-
ativity, in which the metric of 4-dimensional space-time is determined by the distribution of matter.
Famously, the metric for the space-time outside a point mass M is called the Schwarzschild metric, given
by ds
2
= (1
2GM
rc
2
)dt
2
+ (1
2GM
rc
2
)
1
dr
2
+ r
2
d
2
+ r
2
sin
2
d
2
. Notice that for large r, this reduces
to ds
2
= dt
2
+dr
2
+r
2
d
2
+r
2
sin
2
d
2
, which is the (Minkowski) metric for at space-time, in which
we can see embedded the familiar 3-d spherical polar distance.
We now use the h
i
coecients to nd a general denition of the gradient operator. We can write the
small change in any quantity as d =

q1
dq
1
+

q2
dq
2
+

q3
dq
3
, but we also have that
d = .(h
1
e
1
dq
1
+h
2
e
2
dq
2
+h
3
e
3
dq
3
). This shows that h
1
.e
1
=

q1
etc, and hence
=
1
h1

q1
e
1
+
1
h2

q2
e
2
+
1
h3

q3
e
3
.
Our two important cases give
i) Cylindrical Polars =

+
1

+

z
e
z
, and
ii) Spherical Polars =

r
e
r
+
1
r

+
1
r sin

.
41
8 The divergence of a vector eld
So far we have only looked at the spatial variations of scalar elds, what about the spatial variations of
vector elds?
There are two elds that are used to describe spatial variations of vectors elds F, one is a scalar eld
known as the divergence of F, and the other is a vector eld known as the curl of F.
8.1 Mathematical approach to divergence
The divergence of a vector eld F is the scalar eld
div F = F =
_
i

x
+j

y
+k

z
_
(F
x
i+F
y
j+F
z
k)
=
F
x
x
+
F
y
y
+
F
z
z
.
Example 8.1 Calculate the divergence of the vector eld F = 3xe
y
i+2yj+z
2
k and evaluate F(2, 0, 4).
Solution: We have
div F = F =
(3xe
y
)
x
+
(2y)
y
+
(z
2
)
z
= 3e
y
+ 2 + 2z
F(2, 0, 4) = 3 + 2 + 8 = 13.
8.2 Physical approach to divergence
Physically the divergence of a vector eld F is related to the ow of energy or matter, e.g. ow of water
in a pipe or ow of heat through some conducting material.
Consider the case of heat ow in a rod. Heat generated (by ssion of uranium nuclei say) at a rate (r),
ows via conduction to the cooler outer edge of the rod. If we dene the heat ow rate by h(r) then the
two elds (r) and h(r) are related by conservation of energy.
Consider a very small region R of the rod with volume V and surface S, such that R contains the point
P with position vector r
p
. Since R is small we can approximate the total rate at which heat is generated
in R by (r
p
)V. Since the temperature of the region does not change we must have that the amount of
heat generated in R must be equal to the rate of ow of heat out of R through its surface S.
Thus we have
net outward heat ow rate across S (r
p
)V

net outward heat ow rate across S


V
(r
p
).
If we let V tend to zero then the left hand side describes the net outward ow rate per unit volume at
the point P. We dene this to be the divergence of h at the point P.
Thus
div h(P) = lim
V 0
_
net outward heat ow
V
_
= (r
p
).
Since this equation holds for all points P we have
div h = .
42
8.3 Equivalence of the two denitions of divergence
L
Q`
Q
i
j
k
(5)
(6)
Figure 30:
If we take the small region R in the heated rod to be a cube with centre C = (X +
L
2
, Y, Z) and sides of
length L, then we can calculate the net outward heat ow through R by using
Flux = =
_
R
h ndA =
_
R
h
n
dA
=
6

i=1
__
face i
h ndA
_
where h
n
denotes the component of h pointing in the direction of n and n is the outward pointing normal
to R.
If we set (1) to be the face of the cube lying in the plane x = X (see Figure 30) and (2) to be the face of
the cube lying in the plane x = X +L, then the outer unit normal to (1) is i and the outer unit normal
to (2) is i. Also Q = (X, Y, Z) is the mid point of (1) and Q

= (X +L, Y, Z) is the mid point of (2) and


the volume V = L
3
.
Since the cube is small we can approximate the heat ow vector h on each face of the cube by its value
at the mid points, hence on face (1) we can approximate h by h(Q) = h(X, Y, Z) and on face (2) we can
approximate h by h(Q

) = h(X +L, Y, Z).


Hence on (1) we have h n = h (i) = h i = h
x
(X, Y, Z) (i.e. minus the x component of h at Q),
and on (2) we have h n = h i = h
x
(X +L, Y, Z).
So _
face 1
h ndA
_
face 1
h
x
(X, Y, Z)dA = h
x
(X, Y, Z)
_
face 1
dA = h
x
(X, Y, Z)L
2
_
face 2
h ndA
_
face 2
h
x
(X +L, Y, Z)dA = h
x
(X +L, Y, Z)L
2
.
So
net outward heat ow across (1)+(2)
V
=
L
2
[h
x
(X +L, Y, Z) h
x
(X, Y, Z)]
L
3
=
h
x
(X +L, Y, Z) h
x
(X, Y, Z)
L
lim
V 0
_
net outward heat ow across (1)+(2)
V
_
= lim
L0
_
h
x
(X +L, Y, Z) h
x
(X, Y, Z)
L
_
=
h
x
x
.
43
Similarly
lim
V 0
_
net outward heat ow across (3)+(4)
V
_
=
h
y
y
and
lim
V 0
_
net outward heat ow across (5)+(6)
V
_
=
h
z
z
.
Thus
div h = lim
V 0
_
net outward heat ow across S
V
_
=
h
x
x
+
h
y
y
+
h
z
z
.
8.4 Interpretation of divergence in terms of sources and sinks
+
+
+
+
+
+
Figure 31:
We have seen that the divergence of a steady state heat ow eld at any point is the heat source density
at that point. Similar ideas can be applied to other ow elds.
We can regard positive electric charge as being the source of an electric eld E
div E =

0
(one of Maxwells equations).
Here is the electric charge density and
0
is the permittivity of free space.
The interpretation of divergence in terms of sources can be illustrated in eld line diagrams.
For example the eld lines in the region outside a positively charged sphere are regularly spaced continuous
(i.e. unbroken) radial lines directed outwards see Figure 31. In 3D the density of these lines represents
the magnitude of the eld.
You can only make the association of eld magnitude and density of continuous eld lines in regions
where the divergence of the vector eld is zero.
Field lines have sources (i.e. starting point) in regions of positive divergence, and sinks (end points) in
regions of negative divergence.
If a vector eld has zero divergence everywhere, then there are no sources or sinks anywhere and the eld
lines are closed loops.
Magnetic elds are examples of divergence free elds see Figure 32.
div B = 0 Maxwells equation, B - magnetic eld.
44
Figure 32:
8.5 The divergence and physical laws
The following are examples of physical laws that involve the divergence of vector elds:
divh = , h steady state heat ow, - heat source density,
divE =

0
, E electric eld, - electric charge density,
0
- permittivity of free space,
and
divB = 0, B magnetic eld.
Example 8.2 Calculate the divergence of the vector eld F = y cos
2
xi+yz
2
j+e
z
xk at the point (, 0, 2).
Solution: We have
div F = F =
F
x
x
+
F
y
y
+
F
z
z
= 2y cos xsin x +z
2
+e
z
x
and hence
div F(, 0, 2) = 2(0)(cos sin ) + 2
2
+e
2
= 4 +e
2
.
Example 8.3 Could either of the following vector elds F or G in principle represent a magnetic eld
F = xy
2
i +y
3
j + (x
2
y 4y
2
z)k or G = 2yi 3xyj (9 + 5z)k?
Solution: Since
div F =
F
x
x
+
F
y
y
+
F
z
z
= y
2
+ 3y
2
4y
2
= 0
and
div G =
G
x
x
+
G
y
y
+
G
z
z
= 3x 5
it follows that F could in principle represent a magnetic eld and G could not.
Example 8.4 If the scalar eld f is equal to the divergence of the vector eld G = xcos yi+x
2
e
y
j+z
2
yk,
calculate H = grad f.
45
Solution: We have
f = divG =
G
x
x
+
G
y
y
+
G
z
z
= cos y +x
2
e
y
+ 2zy
and hence
H = grad f =
f
x
i +
f
y
j +
f
z
k
= 2xe
y
i + (sin y +x
2
e
y
+ 2z)j + 2yk
= 2xe
y
i + (x
2
e
y
+ 2z sin y)j + 2yk.
Remark In operator terms the vector eld H in the above example takes the form
H = f = ( G).
Example 8.5 If f = 3x
2
y +y cos z +y
2
calculate div F, where F = grad f.
Solution: We have
F = grad f =
f
x
i +
f
y
j +
f
z
k
and hence
F = 6xyi + (3x
2
+ cos z + 2y)j y sin zk
div F =
F
x
x
+
F
y
y
+
F
z
z
= 6y + 2 y cos z.
8.6 Other coordinate systems
Using the same notation as earlier, we can nd a general expression for the divergence of a vector eld
from the denition above
div h(P) = lim
V 0
_
net outward heat ow
V
_
.
We have V = h
1
h
2
h
3
dq
1
dq
2
dq
3
, and we nd, for example, that the ux of a vector eld u across two
opposite faces q
1
=constant is given by

q1
(u
1
h
2
h
3
)dq
1
dq
2
dq
3
. Then, taking the limit, we get
div u = u =
1
h
1
h
2
h
3
(

q
1
(u
1
h
2
h
3
) +

q
2
(u
2
h
1
h
3
) +

q
3
(u
3
h
1
h
2
)).
This gives
i) Cylindrical polars
F =
1

(F

) +
1

+
F
z
z
=
F

+
F

+
1

+
F
z
z
ii) Spherical polars
F =
1
r
2

r
(r
2
F
r
) +
1
r sin

(F

sin ) +
1
r sin
F

.
46
8.7 The scalar operator del squared
In operator terms we have div (grad f) = f =
2
f, here
2
is the scalar operator del squared, such
that

2
=

2
x
2
+

2
y
2
+

2
z
2
with

2
f =

2
f
x
2
+

2
f
y
2
+

2
f
z
2
.
We can also have del squared acting on a vector eld, such that

2
F =
2
(F
x
i+F
y
j+F
z
k)
=
2
F
x
i+
2
F
y
j+
2
F
z
k.
We can use the results from the previous sections to nd the expression for the Laplacian operator in
other coordinate systems. We nd

2
=
1
h
1
h
2
h
3
_

q
1
_
h
2
h
3
h
1

q
1
_
+

q
2
_
h
3
h
1
h
2

q
2
_
+

q
3
_
h
1
h
2
h
3

q
3
__
.
This gives
i) Cylindrical polars

2
=
1

_
+
1

2
+

2

z
2
ii) Spherical polars

2
=
1
r
2
sin
_
sin

r
_
r
2

r
_
+

_
sin

_
+
1
sin

2
_
.
47
9 The curl of a vector eld
9.1 Mathematical denition of curl
The mathematical denition of the curl of a vector eld F is obtained by taking the cross product of del
(nabla) with F i.e.
curl F = F =

i j k

z
F
x
F
y
F
z

= i
_
F
z
y

F
y
z
_
j
_
F
z
x

F
x
z
_
+k
_
F
y
x

F
x
y
_
.
Example 9.1 Calculate the curl of the vector eld F = x
2
yi +y
2
zj +z
2
xk and then evaluate curl F at
P = (1, 2, 3)
Solution: We have
curl F =

i j k

z
x
2
y y
2
z z
2
x

= i
_

y
_
z
2
x
_


z
_
y
2
z
_
_
j
_

x
_
z
2
x
_


z
_
x
2
y
_
_
+k
_

x
_
y
2
z
_


y
_
x
2
y
_
_
= y
2
i z
2
j x
2
k.
Hence
curl F(1, 2, 3) = 4i 9j k.
9.2 Physical denition of curl
The physical denition of curl F is given in terms of the work done by F during one complete traversal
of a closed loop. The work done depends on the orientation of the loop. Using the right hand screw rule
we can specify the orientation of the loop in terms of a unit vector n at right angles to the plane of the
loop. If we consider a loop that lies in the (x, y) plane like in (Figure 34) then an anti-clockwise traversal
of the loop is associated with n = k, whereas a clockwise traversal is associated with n = k.
Assume that we have calculated W =
_
L
F dr then as the loop gets smaller and smaller so does W.
However, if we divide W by A =area enclosed by L, then we have a quantity that remains nite as the
loop gets smaller. If we dene
C = lim
A0
W
A
then C is in fact the component of curl F in the direction specied by n
i.e.
C = curl F n.
In our example n = k and hence C is the z-component of curl F with C = curl F k.
9.3 Local rotation
The curl of a vector eld F (or rot F as it is sometimes called) can be thought of in terms of the local
rotation caused by F.
Consider the case of the water velocity on the surface of a river. In fast owing rivers small oating
objects rotate as they ow downstream (and this rotation is faster nearer the banks than it is near the
middle of the river.) This is because the velocity of the river is greater in the middle of the river than at
the edges (see gure 33). The magnitude of these rotations give the z-component of curl v.
48
x
y
z
v(x,y)
(curl v)
z
> 0
(curl v)
z
< 0
river bank
river bank
Figure 33:
H
H
B
R
S
Q
C
T
D
A
P
x
y
Figure 34:
9.4 Equivalence of the physical and mathematical denitions of curl
We recall that curl F n = lim
A0
W
A
where W is the work done by F during one complete traversal of
a closed loop L that encloses an area A. We take L to be a square loop in the (x, y) plane with centre
P = (X, Y, Z) (see Figure 34)). For the loop L we have
W =
_
L
F dr =
_
AB
F dr+
_
BC
F dr+
_
CD
F dr+
_
DA
F dr.
Furthermore
_
AB
F dr =
_
AB
F
x
dx +F
y
dy +F
z
dz =
_
x(B)
x(A)
_
F
x
dx
dx
+F
y
dy
dx
+F
z
dz
dx
_
dx
where x(A) and x(B) respectively denote the x-coordinates of the points A and B. Since y = Y
H
2
and
z = constant on AB we have
dy
dx
=
dz
dx
= 0 and hence
_
AB
F dr =
_
X+
H
2
X
H
2
F
x
dx.
Since H is small we can approximate the x component of F on AB by its value at the midpoint S and
hence we have
_
AB
F dr =
_
X+
H
2
X
H
2
F
x
dx HF
x
(S)
49
Similarly
_
CD
F dr =
_
x(D)
x(C)
_
F
x
dx
dx
+F
y
dy
dx
+F
z
dz
dx
_
dx =
_
X
H
2
X+
H
2
F
x
dx HF
x
(T)
_
BC
F dr =
_
y(C)
y(B)
_
F
x
dx
dy
+F
y
dy
dy
+F
z
dz
dy
_
dy =
_
Y +
H
2
Y
H
2
F
y
dy HF
y
(Q)
_
DA
F dr =
_
y(A)
y(D)
_
F
x
dx
dy
+F
y
dy
dy
+F
z
dz
dy
_
dy =
_
Y
H
2
Y +
H
2
F
y
dy HF
y
(R).
Thus we have
W =
_
L
F dr = HF
x
(S) HF
x
(T) +HF
y
(Q) HF
y
(R)
= H(F
x
(S) F
x
(T)) +H(F
y
(Q) F
y
(R))
and so
curl F n = lim
A0
W
A
= lim
H0
(H(F
x
(S) F
x
(T)) +H(F
y
(Q) F
y
(R)))
H
2
= lim
H0
F
x
(X, Y
H
2
, Z) F
x
(X, Y +
H
2
, Z)
H
+ lim
H0
F
y
(X +
H
2
, Y, Z) F
y
(X
H
2
, Y, Z)
H
=
F
x
y
+
F
y
x
.
As we have calculated the work done by traversing the loop in an anti-clockwise direction, from the right
hand screw rule we have that n = k and hence we have that the z-component of curl F is given by
_
Fy
x

Fx
y
_
. For the x and y components of curl F we would need to consider loops in the yz and xz
planes.
9.5 Conservative elds
When the work done by a vector elds F for all possible loops in the domain of F is equal to zero, then
the curl of F equals zero everywhere and we dene F to be conservative.
This gives us a test for conservative elds
curl F = 0 for any conservative eld. (9.15)
The term conservative is used for any vector eld that satises (9.15).
Example 9.2 Is the vector eld F =
_
x
2
35
_
j + 2zk conservative?
Solution: We have
curl F =

i j k

z
0 x
2
35 2z

= i
_

y
(2z)

z
_
x
2
35
_
_
j
_

x
(2z)

z
(0)
_
+k
_

x
_
x
2
35
_


y
(0)
_
= 0i + 0j + 2xk.
Since curl F = 0 for all values of x, y, z F is non conservative.
50
9.5.1 Maxwells equations in dierential form
Maxwells equations involve div and curl:
curl E =
B
t
Faradays law for induction
curl B =
0
J +
0

0
curl
E
t
Amperes law
div B = 0 Gausss law for magnetism
div E =

0
Gausss law for electricity.
Here E - electric eld, B - magnetic eld, J - electric current density,
0
- permeability of free space,
0
- permittivity of free space, - electric charge density.
9.5.2 Operator identities
Recall the following useful operator identities:
= del or nabla = i

x
+j

y
+k

z
f = grad f = i
f
x
+j
f
y
+k
f
z
div F = F =
F
x
x
+
F
y
y
+
F
z
z
curl F = F =

i j k

z
F
x
F
y
F
z

= i
_
F
z
y

F
y
z
_
j
_
F
z
x

F
x
z
_
+k
_
F
y
x

F
x
y
_
div (grad f) = f =
2
f =

2
f
x
2
+

2
f
y
2
+

2
f
z
2
.
Also note that for all vector elds F we have
div (curl F) = (F) = 0
curl (grad f) = f = 0.
Example 9.3 Verify that div (curl F) = 0 for the vector eld F = ae
x
i z cos yj + 3xyzk.
Solution: Set
G = curl F =

i j k

z
ae
x
z cos y 3xyz

= i (3xz + cos y) j (3yz 0) +k(0 0)


= (3xz + cos y)i 3yzj
and hence
div G = div(curl F) =
G
x
+
G
y
+
G
z
= 3z 3z + 0 = 0.
51
9.6 Other coordinate systems
We can use Stokes Theorem ( see next section) to nd an expression for the curl of a vector eld in a
general coordinate system. We nd
curl F = F =
1
h
1
h
2
h
3

e
1
h
1
e
2
h
2
e
3
h
3

q1

q2

q3
h
1
F
1
h
2
F
2
h
3
F
3

So we get
i) Cylindrical polars
curl F = F =
1

e
z

z
F

F
z

ii) Spherical polars


curl F = F =
1
r
2
sin

e
r
e

r e
3
h
3

q1

q2

q3
h
1
F
1
h
2
F
2
h
3
F
3

52
10 Stokess and Gausss theorems
10.1 Stokess theorem
The surface integral (ux) of curl F across any simple surface S is equal to the line integral of F around
the closed boundary curve C.
_
S
curl F ndA =
_
C
F dr.
The sense of the traversal of the line integral is related to the unit normal of the surface by the right
hand screw rule.
10.2 Gausss theorem (divergence theorem)
The volume integral of div F over a region B is equal to the surface integral (outward ux) of F across
the surface enclosing B.
_
B
div FdV =
_
S
F ndA where n is the outward unit normal to S.
x
y
x
y
z
x
y
z
Figure 35:
Remark 9 In Stokess theorem we have that
_
S
curl F ndA =
_
C
F dr
where S is the surface enclosed by the closed curve C. For any closed curve C there are innitely many
surfaces S that are enclosed by C.
If C is the closed curve x
2
+y
2
= 1, z = 0, one possible S could be the circular disk x
2
+y
2
= 1, z = 0
(see Figure 35). Alternatively S could be the hemisphere x
2
+ y
2
+ z
2
= 1, z 0. Or S could be the
surface of the cylinder = 1, 0 2, 0 < z 2 (i.e. the curved surface and top, but not the base).
Example 10.1 (of Gausss theorem)
Verify that Gausss theorem holds for the vector eld F = 3xyi 2zxk, and the brick B bounded by
1 x 3, 0 y 2, 2 z 5.
Solution: We need to show that
_
B
div FdV =
_
S
F ndA where S denotes the surface of the brick B.
For the right hand side we have
_
S
F ndA =
6

i=1
(
_
face i
F ndA)
53
such that
on face 1, x = 1, 0 y 2, 2 z 5 and n = i
on face 2, x = 3, 0 y 2, 2 z 5 and n = i
on face 3, y = 0, 1 x 3, 2 z 5 and n = j
on face 4, y = 2, 1 x 3, 2 z 5 and n = j
on face 5, z = 2, 1 x 3, 0 y 2 and n = k
on face 6, z = 5, 1 x 3, 0 y 2 and n = k.
Hence we have
_
face 1
F ndA =
2
_
0
5
_
2
(3xyi 2zxk) (i)dzdy
=
2
_
0
5
_
2
3ydzdy
=
_
_
2
_
0
3ydy
_
_
_
_
5
_
2
dz
_
_
=
_
3
2
y
2
_
2
0
[z]
5
2
= 18
_
face 2
F ndA =
2
_
0
5
_
2
(3xyi 2zxk) idzdy
=
2
_
0
5
_
2
3xydzdy
=
2
_
0
5
_
2
9ydzdy
=
_
9
2
y
2
_
2
0
[z]
5
2
= 54
_
face 3
F ndA =
3
_
1
5
_
2
(3xyi 2zxk) (j)dzdx
=
3
_
1
5
_
2
0dzdx = 0
_
face 4
F ndA =
3
_
1
5
_
2
(3xyi 2zxk) jdzdx = 0
54
_
face 5
F ndA =
3
_
1
2
_
0
(3xyi 2zxk) (k)dydx
=
3
_
1
2
_
0
2zxdydx
=
3
_
1
2
_
0
4xdydx
=
_
_
3
_
1
xdx
_
_
_
_
2
_
0
4dy
_
_
= 32
and
_
face 6
F ndA =
3
_
1
2
_
0
(3xyi 2zxk) kdydx
=
3
_
1
2
_
0
2zxdydx
=
3
_
1
2
_
0
10xdydx = 80.
So _
S
F ndA = 18 + 54 + 0 + 0 + 0 + 32 80 = 12.
For the left hand side we have
_
B
div FdV =
F
x
x
+
F
y
y
+
F
z
z
= 3y 2x.
55
So
_
B
div FdV =
5
_
2
2
_
0
3
_
1
(3y 2x) dxdydz
=
5
_
2
2
_
0
_
3yx 2x
2

3
1
dydz
=
5
_
2
2
_
0
[(9y 9) (3y 1)] dydz
=
5
_
2
2
_
0
(6y 8)dydz
=
5
_
2
_
3y
2
8y

2
0
dz
=
5
_
2
(12 16) dz
=
5
_
2
4dz = 4 [z]
5
2
= 12.
Since the left hand side equals the right hand side, the theorem is veried.
Example 10.2 (of Gausss theorem)
Verify that Gausss theorem holds for the vector eld F = r
2
e
r
+ r sin cos e

and the sphere B with


centre the origin and radius 1.
Solution: We need to show that
_
B
div FdV =
_
S
F ndA where S denotes the surface of the sphere B.
For the right hand side we have
_
S
F ndA =
_
S
F e
r
dA =
_
S
r
2
dA.
Since r = 1 on the surface of the sphere and dA = R
2
sin dd = sin dd we have
_
S
F ndA =
_
S
r
2
dA =
_
2
0
_

0
sin dd
=
__
2
0
d
___

0
sin d
_
= []
2
0
[cos ]

0
= (2 0)((1 1)) = 4.
For the left hand side we note that
div F = F =
1
r
2

r
(r
2
F
r
) +
1
r sin

(F

sin ) +
1
r sin
F

and hence
div (r
2
e
r
+r sin cos e

) =
1
r
2

r
(r
2
r
2
) +
1
r sin

(0 sin ) +
1
r sin
(r sin cos )

=
1
r
2

r
(r
4
) +
1
r sin

(0) +
r sin
r sin
(cos )

= 4r sin .
56
Thus since dV = r
2
sin drdd we have
_
B
divFdV =
_
B
(4r sin )dV
=
_
2
0
_

0
_
1
0
(4r sin )r
2
sin drdd
=
_
2
0
_

0
_
1
0
(4r
3
r
2
sin ) sin drdd
=
_
2
0
_

0
_
r
4

1
3
r
3
sin
_
1
0
sin dd
=
_
2
0
_

0
_
1
1
3
sin
_
sin dd
=
_
2
0
_
1
1
3
sin
___

0
sin d
_
d
=
_
2
0
_
1
1
3
sin
_
[cos ]

0
d
= 2
_
2
0
_
1
1
3
sin
_
d
= 2
_
+
1
3
cos
_
2
0
= 4.
Example 10.3 (of Stokess theorem)
Verify that Stokess theorem holds for the vector eld F = yzi xzj + zk and the closed curve C given
by x
2
+y
2
= 1, z = 3.
Solution: We need to show that _
S
curl F ndA =
_
C
F dr
where n is related to the direction of traversal of the curve C by the right hand screw rule, and S is a
surface enclosed by C.
We choose our surface S to be the disc x
2
+y
2
1, z = 3, and we take a clockwise traversal of the curve
C, so that n = k.
For the right hand side we have
_
C
F dr =
_
C
F
x
dx +F
y
dy +F
z
dz
=
_
2
1
(F
x
dx
d
+F
y
dy
d
+F
z
dz
d
)d
where
1
= 2,
2
= 0, x = cos , y = sin , z = 3
dx
d
= sin ,
dy
d
= cos ,
dz
d
= 0.
Hence
_
C
F dr =
0
_
2
(yz sin xz cos ) d
=
0
_
2
_
3 sin
2
3 cos
2

_
d
=
0
_
2
3d = [3]
0
2
= 3 (0 2) = 6.
57
For the left hand side we have
_
S
curl F ndA =
_
2
0
_
1
0
curl F ndd
with
curl F =

i j k

z
F
x
F
y
F
z

i j k

z
yz xz z

= i
_
z
y

(xz)
z
_
j
_
z
x

(yz)
z
_
+k
_
(xz)
x

(yz)
y
_
= xi +yj 2zk.
Hence
curl F n = (xi +yj 2zk) (k)
= 2z = 6 (since z = 3).
So
_
S
curl F ndA =
_
2
0
_
1
0
6dd
=
__
1
0
6d
___
2
0
d
_
=
_
3
2

1
0
[]
2
0
= 6.
Since the left hand side equals the right hand side, the theorem is veried.
Example 10.4 (of Stokess theorem)
Verify that Stokess theorem holds for the vector eld F = yzi xzj + zk and the closed curve C that
lies in the plane x = 1 and consists of the three lines L
1
, L
2
and L
3
where on L
1
we have z = 0 and
0 y 1, on L
2
we have y = 1 and 0 z 1 and on L
3
we have z = y and 0 y 1.
Solution: We need to show that _
S
curl F ndA =
_
C
F dr
where n is related to the direction of traversal of the curve C by the right hand screw rule, and S is a
surface enclosed by C.
We choose our surface S to be the two dimensional triangular surface lying in the plane x = 1 that is
bounded by the three lines L
1
, L
2
and L
3
, and we take an anticlockwise traversal of the curve C, so that
n = i.
For the right hand side we have
_
C
F dr =
_
L1
F dr +
_
L2
F dr +
_
L3
F dr.
Since x = 1 and z = 0 on L
1
we have
dx
dy
=
dz
dy
= 0
and since y varies from 0 to 1 we have
_
L1
F dr =
_
L1
yzdx xzdy +zdz
=
_
1
0
_
yz
dx
dy
xz
dy
dy
+z
dz
dy
_
dy
=
_
1
0
xzdy =
_
1
0
0dy = 0.
58
Since x = 1 and y = 1 on L
2
we have
dx
dz
=
dy
dz
= 0
and since z varies from 0 to 1 we have
_
L2
F dr =
_
L2
yzdx xzdy +zdz
=
_
1
0
_
yz
dx
dz
xz
dy
dz
+z
dz
dz
_
dz
=
_
1
0
zdz =
_
z
2
2
_
1
0
= 1/2.
Since x = 1 and z = y on L
3
we have
dx
dy
= 0 and
dz
dy
= 1
and since y varies from 0 to 1 we have
_
L3
F dr =
_
L3
yzdx xzdy +zdz
=
_
1
0
_
yz
dx
dy
xz
dy
dy
+z
dz
dy
_
dy
=
_
1
0
(xz +z)dy
=
_
1
0
(y +y)dy =
_
1
0
0dy = 0.
Hence _
C
F dr =
_
L1
F dr +
_
L2
F dr +
_
L3
F dr = 0 + 1/2 + 0 = 1/2.
From the previous example we have that curlF = xi +yj 2zk and hence for the left hand side we have
_
S
curl F ndA =
_
S
curl F idA =
_
S
xdA.
Since S lies in the plane x = 1 we have
_
S
xdA =
_
S
dA = area of S = 1/2.
Since the left hand side equals the right hand side, the theorem is veried.
59
11 The calculus of variations
11.1 The shortest distance
Suppose we want to nd the shortest distance between two points in 2-d space. If they are (x
1
, y
1
), (x
2
, y
2
),
then any distance between the two is given by
I =
_
x2
x1
_
dx
2
+dy
2
=
_
x2
x1
_
1 +y
2
dx
where we do the integration along the route y = y(x). We wish to nd the route that minimizes the
integral.
Suppose that y(x) is the required answer, and that we write any other curve as Y (x) = y(x) + h(x),
with h(x
1
) = h(x
2
) = 0. Thinking of I as I(), we want to nd the minimum value of I, so we want
dI
d
= 0 when = 0.
We nd
dI
d
=
_
x2
x1
1
2
1

1 +Y
2
2Y

dY

d
dx. Using Y

(x) = y

(x) +h

(x) gives
dY

d
= h

(x), so that
dI
d =0
=
_
x2
x1
y

(x)h

(x)
_
1 +y
2
dx = 0.
We now integrate by parts, to get
dI
d =0
=
_
y

h(x)
_
1 +y
2
_
x2
x1

_
x2
x1
h(x)
d
dx
_
y

_
1 +y
2
_
dx.
The rst term is zero, because h(x
1
) = h(x
2
) = 0, so to achieve a minimum we need
d
dx
_
y

1+y
2
_
= 0.
This means that
_
y

1+y
2
_
= constant, so y

= constant. Thus, the required curve y(x) is a straight line,


as expected.
11.2 Eulers Equation
Suppose we now generalize the problem, to ask which curve y(x) minimizes the integral I =
_
x2
x1
F(x, y, y

)dx,
where F is a given function.
As before, consider any curve Y (x) = y(x) + h(x), with h(x
1
) = h(x
2
) = 0. Thinking of I as I(), we
want to nd the minimum value of I, so we want
dI
d
= 0 when = 0.
We nd
dI
d
=
_
x2
x1
_
F
Y
dY
d
+
F
Y

dY

d
_
dx. Using Y

(x) = y

(x) +h

(x) again gives


dY

d
= h

(x), so that (
dI
d
)
=0
=
_
x2
x1
_
F
y
h +
F
y

_
dx = 0.
Using parts
_
x2
x1
F
y

dx =
_
F
y

h
_
x2
x1

_
x2
x1
d
dx
_
F
y

_
hdx. The integrated term is zero, again because
of the boundary conditions, so we nally get
_
dI
d
_
=0
=
_
x2
x1
_
F
y

d
dx
F
y

_
hdx = 0.
Since h is arbitrary, we have
F
y

d
dx
F
y

= 0,
which is the Euler ( sometimes known as Euler-Lagrange) equation.
11.3 The Brachistochrone Problem
This is a famous problem, in which we must nd the shape of a smooth wire, joining the points (x
1
, y
1
)
and (x
2
, y
2
), so that a bead will slide down the wire in the shortest time.
Clearly, we need to minimize
_
dt.Taking initial speed v = 0, and the reference level for gravitational
potential energy as y = 0, then at the point (x, y) we have that the kinetic energy =
1
2
mv
2
and the
60
potential energy = mgy ( we are measuring y as positive in the down direction). Conservation of
energy then gives
1
2
mv
2
mgy = 0, so that v =

2gy.
So we need to minimize the integral
_
dt =
_
ds
v
=
_
ds

2gy
=
1

2g
_
_
1 +y
2

y
dx.
It turns out that the easiest way (because F does not explicitly contain x) to use the Euler equation is
by using ds =
_
dx
2
+dy
2
= dy

1 +x
2
, so that we try to minimize
1

2g
_

1 +x
2

y
dy. We now use
the alternate version of the Euler equation:
F
x

d
dy
F
x

= 0, because the integration is with respect to


y, and this, when applied to the y-integral, gives
(

1+x
2

y
)
x

d
dy
(

1+x
2

y
)
x

= 0.
The rst term is zero, since F has no explicit dependence on x, so we get
d
dy
(

1+x
2

y
)
x

= 0. This
integrates immediately to
x

1 +x
2

y
= constant, which we can rearrange as
dx
dy
=
_
cy
1 cy
. This
integrates, by separation of variables, to give x =
_
y
c
y
2
+
1
2c
cos
1
(1 2cy) +c

. If the curve passes


through the origin, then we have c

= 0. This curve is in fact the equation of a cycloid, and it is more


simply written using a parameter , when we get
x =
1
2c
( sin ), y =
1
2c
(1 cos ).
A cycloid is well known as the locus of a point on the circumference of a circle as that circle rolls along
a level surface.
11.4 Hamiltons Principle
We may generalize our original question, of minimizing I =
_
x2
x1
F(x, y, y

)dx, to the case where F depends


on several variables y
1
, y
2
, ..., each of which depend on x, so that we have I =
_
x2
x1
F(x, y
1
, y
2
, ..., y

1
, y

2
, ....)dx.
Using the same method as before, we now get a set of Euler equations, of the form
F
y
i

d
dx
F
y

i
= 0, i = 1, 2, ....., n.
This is particularly useful in analyzing systems which depend on several coordinates. If we dene the
Lagrangian L of a system as the dierence of its kinetic and potential energies, so L = T V , then
Hamiltons principle says that a system evolves in such a way that the time integral of the Lagrangian
is a minimum. This leads to the Lagrange equations
L
x
i

d
dt
F
x

i
= 0, i = 1, 2, ....., n, where now the
variables are labeled x
i
, and they all depend on t.
This formulation of mechanics is equivalent to the Newtonian one, but it involves only scalars, and it is
more amenable to choosing dierent coordinate systems to the usual Cartesian x, y, z. We can easily see
that in one dimension, it reduces to the usual form, since if we take T =
1
2
mx
2
and potential V (x) ( so
that the force in the x-direction is F =
dV
dx
), we get
(V )
x

d
dt
T
x

= 0, or F = m x, which of course is
just Newtonss second law.
11.5 Noethers Theorem
Hamiltons principle can be used in non-mechanical contexts, such as electromagnetism, with an appro-
priate denition of the Lagrangian. In general, in known theories, it simply reproduces results already
know, albeit sometimes in a simpler and more elegant way. However, where a theory is incomplete, a
postulated Lagrangian can be a useful way forward. Such an approach has been widely used in particle
physics and string theory.
61
An example of this is Noethers Theorem, which states that the invariance of the Lagrangian with respect
to a variable ( such as position or time) implies the conservation of a particular quantity (linear momentum
and energy, respectively, for the two examples mentioned). We can see this in the very simplest example,
with L =
1
2
mx
2
V (x). Applying Eulers equation gives
L
x

d
dt
F
x

= 0, or
L
x
=
d
dt
L
x

.
Now consider
dL
dt
=
L
x
x

+
L
x

+
L
t
, using the chain rule of partial dierentiation, and assuming
that L = L(x, x

, t). If, in fact, L does not explicitly contain t, then


L
t
= 0. Using the result in the
previous paragraph, we then get
dL
dt
=
d
dt
(
L
x

)x

+
L
x

, which we can see is the result of the product


rule
dL
dt
=
d
dt
(
L
x

).
Therefore we have
d
dt
(L
L
x

) = 0, and so L
L
x

= constant. But
L
L
x

=
1
2
mx
2
V (x) mx
2
=
1
2
mx
2
V (x), so we have
1
2
mx
2
+ V (x) = constant ie the
conservation of energy.
62

You might also like