Vous êtes sur la page 1sur 58

CHAPTER 2

VECTOR ANALYSIS

2.1 Prelude
Considering a function f (x), whenever there is a tiny change in x, ie., from x to x + dx
would cause a correspondent change in f (x), say from f (x) to f (x + dx). The ratio between
(f (x + dx) − f (x))/dx as dx → 0 defines the slope of the curve around the point x, and is
written as f ′ (x). But if f = f (x, y), the ’slope’ of f would depend on where you stand and the
way you want to move. Namely, the ’slope’ becomes directional dependent, or in other words,
it becomes a vector. So the name ’slope’ is no more proper to characterize the undulations of
a multi-variable function; the vectorial in nature must be specified in the new name.
Of course, if the multi-variable function is a vector rather than a scalar, the so called tiny-
change perturbation on the system must introduce different responses from what in a scalar
function, depending also on the way the perturbation acts.
In this chapter we shall investigate how a multi-variable function, scalar or vector, changes
whenever there are tiny changes in its variables. Since we are dealing with a quantity like
∆G(r)/∆r, and the term is just like the derivative of G, therefore we are just going to find
out how many different kinds of ’derivative operation’ can act on G and the correspondent
’response’ of G.
A professional terminology of what we want to do is called the vector derivative and we
assign it a symbol ∇, read ’del’. Remember, ∇ is a vector operator. In the following sections
we shall discuss four ways the ’del’ acts:

1. Gradient, ∇

2. Divergence, ∇·

3. Curl, ∇×

4. Laplacian, ∇2 = ∇ · ∇

1
2 CHAPTER 2. VECTOR ANALYSIS

2.2 Gradient, ∇
Consider a multi-variable function F (x, y), the total variation of F can be written as

dF (x, y) = F (x + dx, y + dy) − F (x, y)


= [F (x + dx, y + dy) − F (x, y + dy)] + [F (x, y + dy) − F (x, y)]
∂F ∂F
= dx + dy, (2.1)
∂x ∂y
as dx, dy → 0. So dF comes from the change of F both along x-direction and y-direction, or
equivalently along a way r = x + y.
Generalizing the above result to introduce an arbitrary scalar function ϕ(x1 , x2 , x3 ), then
an infinitesimal change dϕ can be expressed as the product of two vectors:

dϕ = ∇ϕ · dr, (2.2)

where
∂ ∂ ∂
∇≡ ê1 + ê2 + ê3 (2.3)
∂x1 ∂x2 ∂x3
is a vector called gradient.

Figure 2.1:

The geometric meaning of ∇ϕ shown in Fig. 2.1 can be interpreted as follows: Knowing
that a function ϕ(x1 , x2 , x3 ) = c, where c is a constant, expands a surface and upon it exists a
point P at the position r. If we move along the surface to a nearby point Q, located at r + dr,
then dr must lies in the tangent plane to the surface at P . But as long as the movement is on
the constant surface ϕ, dϕ must be zero, which means

dϕ = ∇ϕ · dr = 0. (2.4)

Thus, ∇ϕ must be in the direction perpendicular to dr therefore to the surface.


On the other hand, if we now permit dr to take us from one surface ϕ = c1 to an adjacent
surface ϕ = c2 as shown in Fig. 2.2, then dϕ = ∇ϕ · dr = c1 − c2 = ∆c. So for a given dϕ, dr is
2.2. GRADIENT, ∇ 3

Figure 2.2:

a minimum if it takes along the direction along ∇ϕ since dϕ = |∇ϕ||dr| cos θ herein cos θ = 1.
Or, for a given dr, dϕ is a maximum if dr is chosen to be parallel to ∇ϕ. From either aspect,
∇ϕ is identified as a vector having the direction of the maximum space rate of change of ϕ.
The component of ∇ϕ in the direction of a unit vector â is given by ∇ϕ · â and is called the
directional derivative of ϕ in the direction â. This quantity indicates the rate of change of ϕ at
the position (x1 , x2 , x3 ) in the direction â.

Figure 2.3: Geometrical properties of ∇ϕ. PQ gives the value of dϕ/ds in the direction a.

Example 2.2.1. For the function ϕ = x2 y + yz at the point (1, 2, −1), find its rate of change
with distance s in the direction a = x̂ + 2ŷ + 3ẑ. At this same point, what is the greatest
possible rate of change with distance s and in which direction does it occur?
4 CHAPTER 2. VECTOR ANALYSIS

(1.) A vector having the direction of the maximum space rate of change of the function ϕ
is ∇ϕ given by

∇ϕ = 2xy x̂ + (x2 + z) ŷ + y ẑ
= 4x̂ + 2ẑ, (2.5)

at (1, 2, −1). The unit vector in the direction of a is â = √114 (x̂ + 2ŷ + 3ẑ). Therefore, the
greatest possible rate of change with distance s in the direction a is
dϕ 1 10
= ∇ϕ · â = √ (4 + 6) = √ . (2.6)
ds 14 14
(2.) For a given dϕ, ds would be a minimum if it takes along the direction of√∇ϕ. Therefore,
dϕ/ds will be greatest in the direction of ∇ϕ = 4 x̂ + 2 ẑ and has value |∇ϕ| = 20 at the point
(1, 2, −1).

We can extend the above analysis to find the rate of change of a vector field (rather than
a scalar field as above) in a particular direction. The scalar differential operator â · ∇ can be
shown to give the rate of change with distance in the direction â of the quantity (vector or
scalar) on which it acts. In Cartesian coordinates it may be written as

∂ ∂ ∂
â · ∇ = ax + ay + az . (2.7)
∂x ∂y ∂z
Thus we can write the infinitesimal change in an electric field in moving from r to r + dr as
dE = (dr · ∇)E.
Physically, the non-contact force field can be represented by the gradient of the potential.
The relation is given by
E(r) = −∇V (r), (2.8)
in which the minus sign indicates the direction of flow and the conservation of energy as well.
For example, if V (r) is the Coulomb potential of a point charge measured at r = (x, y, z), then
dV = ∇V (r) · dr = −E(r) · dr. It means that the total work done by the electric field to move
a point charge from r to r + dr is equal to
∫ r+dr
∆W = qE(r) · dr = −q[V (r + dr) − V (r)]. (2.9)
r

Form work-kinetic energy theorem, the work done on the system by the electric field increases
the kinetic energy of the charge, moving a charge along the electric field downhill to somewhere
of lower potential, showing the conservation of mechanical energy.
In math language, a function Y is said to be conservative if and only if it can be defined as
the gradient of some function Q such the total derivative of Q, dQ, can be established. In this
way, the integral of dQ along a closed path vanishes as it should be from Calculus. Applying to
thermodynamics in which a function Y can be defined as a thermal dynamical state function
as long as dY exists. For example, internal energy U and entropy S, are state functions, but
heat Q is not. So we set up the relation dS = δQ/T , and precisely speaking, δQ ̸= dQ, but we
often use that as an approximate form. If the trajectory of the gas on the phase diagram forms
a closed path, it must be also reversible. Under this circumstance, neither the internal energy
nor the entropy of molecules changes when they back to the initial state point.
2.3. DIVERGENCE, ∇· 5

Example 2.2.2. GRADIENT OF rn



As r = x2 + y 2 + z 2 , the gradient of f (r) = rn is given by

∂f ∂f ∂f
∇f (r) = x̂ + ŷ + ẑ
∂x ∂y ∂z
df ∂r df ∂r df ∂r
= x̂ + ŷ + ẑ
dr ∂x dr ∂y dr ∂z
1 df
= (xx̂ + yŷ + zẑ)
r dr
= nrn−1 r̂. (2.10)

As a special case of Coulomb potential f (r) = 1/r, we obtain the

1
∇f (r) = − r̂. (2.11)
r2
Example 2.2.3. The GRADIENT OF A POTENTIAL V (r)

Consider the potential V (r) = V ( x2 + y 2 + z 2 ). The gradient of V is given by

∂V ∂V ∂V
∇V (r) = x̂ + ŷ + ẑ
∂x ∂y ∂z
dV ∂r dV ∂r dV ∂r
= x̂ + ŷ + ẑ
dr ∂x dr ∂y dr ∂z
1 dV
= (xx̂ + yŷ + zẑ)
r dr
dV
= r̂ . (2.12)
dr
Example 2.2.4. If a vector function F depends on both space coordinates (x, y, z) and time t
, show that
∂F
dF = (dr · ∇)F + dt.
∂t
For independent variables x, y, z, t,

∂F ∂F ∂F ∂F
dF(x, y, z, t) = dx + dy + dz + dt
∂x ∂y ∂z ∂t
( ) ( ) ( )
∂ ∂ ∂ ∂F
= dxx̂ · x̂ F + dyŷ · ŷ F + dzẑ · ẑ F+ dt
∂x ∂y ∂z ∂t
∂F
= (dr · ∇)F + dt. (2.13)
∂t

2.3 Divergence, ∇·
The divergence of a vector function F is defined as the vector differentiation of F. The operator
takes the form ( )
∂ ∂ ∂
∇· = x̂ + ŷ + ẑ · . (2.14)
∂x ∂y ∂z
6 CHAPTER 2. VECTOR ANALYSIS

Therefore, the divergence of F(x, y, z) is given by


( )
∂ ∂ ∂
∇ · F(x, y, z) = x̂ + ŷ + ẑ · (Fx x̂ + Fy ŷ + Fz ẑ)
∂x ∂y ∂z
∂Fx ∂Fy ∂Fz
= + + , (2.15)
∂x ∂y ∂z

resulting in a scalar function.

Example 2.3.1. DIVERGENCE OF COORDINATE VECTOR


( )
∂ ∂ ∂
∇·r = x̂ + ŷ + ẑ · (xx̂ + yŷ + zẑ) = 3. (2.16)
∂x ∂y ∂z

Example 2.3.2. DIVERGENCE OF CENTRAL FORCE FIELD


Taking a point O as a reference, the central force f (r) is always pointing either towards or away
from O, ie., it can be written as
f (r) = f (r)r̂. (2.17)
Let’s see a general case:

∂ ∂ ∂
∇ · [rf (r)] = [xf (r)] + [yf (r)] + [zf (r)]
∂x ∂y ∂z
( )
df (r) ∂r df (r) ∂r df (r) ∂r
= 3f (r) + x +y +z
dr ∂x dr ∂y dr ∂z
df (r)
= 3f (r) + r . (2.18)
dr
If f (r) = rn−1 , then
∇ · [rrn−1 ] = (n + 2)rn−1 . (2.19)
Therefore, we find the zero divergence at n = −2 except at r = 0. Physically, this is just the
result of the divergence of the electric field of a point source at the origin.

p + dp
dp
p
r+dr dr p p +dp

Figure 2.4: The motion in a central force field.


2.3. DIVERGENCE, ∇· 7

More on central force field can be discussed. First, whenever a particle suffers a central
field, its angular momentum in motion would be conserved. Since L = r × p, we have

dL dr dp
= ×p+r×
dt dt dt
dv
= v×p+r×m (2.20)
dt
For a central force,
f (r) = f (r)r̂ = mr̈ r̂. (2.21)
Therefore, r × m dv
dt
= 0, and
dL
= 0. (2.22)
dt
Second, whenever a particle moving in a central field, it suffers a conservative force. Since
the force exerted on the particle can be written by Eq. (2.21), the work done on the particle
over a closed path is given by
I I
d2
f (r) · ds = m 2 rr̂ · (drr̂ + rdθθ̂)
C dt C
I ( ) A
d2 d2 r2
= m 2 r dr = m 2 = 0. (2.23)
dt C dt 2 A

Therefore, f (r) is a conservative force.

Example 2.3.3. INTEGRATION BY PARTS OF DIVERGENCE


∫ ∫
Prove the formula f (r)∇ · A(r) dr = − A(r) · ∇f (r) dr

∵ ∇ · [f (r)A(r)] = f (r)∇ · A(r) + A(r)∇f (r). (2.24)

Applying Gauss’s theorem,


∫ I
∇ · [f (r)A(r)]dV = [f (r)A(r)] · n̂ da = 0 (2.25)
V S
∫ ∫
∴ f (r)∇ · A(r) dr = − A(r) · ∇f (r) dr. (2.26)

2.3.1 Physical Interpretation of Divergence


Suppose that a vector field v(r) represents the velocity of a fluid at the spatial points r, and
that ρ(r) represents the fluid density at r at a given time t . Then the direction and magnitude
of the flow rate at any point will be given by the product ρ(r)v(r). Our objective is to calculate
the net rate of change of the fluid density in a volume element at the point r. To do so, we
set up a parallelepiped of dimensions dx, dy, dz centered at r, see Fig. 2.5. To first order
(infinitesimal dr and dt), the density of fluid exiting the parallelepiped per unit time through
the yz face located at x − dx/2 will be

Flow out, face at x − dx/2 := −ρ(x, y, z)vx |x−dx/2,y,z dy dz. (2.27)


8 CHAPTER 2. VECTOR ANALYSIS

Figure 2.5:

For the opposite face,

Flow out, face at x + dx/2 := +ρ(x, y, z)vx )|x+dx/2,y,z dy dz. (2.28)

Combining these, we have for both yz faces


[ ]
−ρ(x′ , y, z)vx |x−dx/2,y,z + ρ(x′ , y, z)vx |x+dx/2,y,z dy dz
[ ( ) ( )]
∂[ρ(x, y, z)vx ] dx ∂[ρ(x, y, z)vx ] dx
= − ρ(x, y, z)vx − + ρ(x, y, z)vx + dy dz
∂x 2 ∂x 2
∂[ρ(x, y, z)vx ]
= dx dy dz. (2.29)
∂x
Similarly, the out-flow through xy and xz faces per unit time are
∂[ρ(x, y, z)vz ]
dx dy dz, (2.30)
∂z
and
∂[ρ(x, y, z)vy ]
dx dy dz, (2.31)
∂y
respectively. Therefore the net flow throughout the volume element per unit time is given by
( )
∂[ρ(x, y, z)vx ] ∂[ρ(x, y, z)vy ] ∂[ρ(x, y, z)vz ]
+ + dx dy dz
∂x ∂y ∂z
= ∇ · (ρv) dx dy dz. (2.32)

However, the number of particles should be conserved in the entire volume, ie., whenever
we count N particles outside the surface of the volume, we lose N particles inside the volume.
The rate of gaining and losing should be the same indicating
∂ρ
∇ · (ρv) = − , (2.33)
∂t
which, we call it continuity equation, demonstrates the essential of the divergence.
Applying to the electromagnetics, where a divergenceless (has zero divergence) field exists
in some region is called a steady flow within that region. For example, in a charge free region,
the Gauss’s law states ∇ · E = 0. The equation can be satisfied for a constant E field. On the
other hand, the E field produced by a point charge at the origin given by
q
E= r̂ (2.34)
4πϵ0 r2
2.4. CURL, ∇× 9

Figure 2.6:

Figure 2.7: Flow diagrams: (a) with source and sink; (b) solenoidal. The divergence vanishes
at volume elements A and C, but is negative at B.

has nonzero divergence, since the strength of the field is not spatially uniform but varies as
1/r2 .
As for B field, since there is no magnetic charge, ∇ · B = 0 is always satisfied. So for
a magnetic bar the magnetic field lines must start from N-pole and end at S-pole, forming a
closed loop. Nearby a steady current, the B field would circulate about the wire, like a solenoid.
In the next section we shall see a divergenceless field can be written as a curl of another vector,
saying the vector potential A here for B.
In brief, the difference between E field and B field characterizes the divergence of a vector
field, which has nonzero values if there is a source or a sink in the field, otherwise the field is
divergenceless.

2.4 Curl, ∇×
Another possible operation with the vector operator ∇ is to take its cross product with a vector
and is written as
( ) ( ) ( )
∂Vz ∂Vy ∂Vx ∂Vz ∂Vy ∂Vx
∇×V = − êx + − êy + − êz
∂y ∂z ∂z ∂x ∂x ∂y

êx êy êz
∂ ∂ ∂
= ∂x ∂y

∂z . (2.35)
Vx Vy Vz

This vector operation is called the curl of V.


10 CHAPTER 2. VECTOR ANALYSIS

Example 2.4.1. CURL OF A CENTRAL FORCE FIELD


Calculate ∇ × [f (r)r̂].

x y z

Writing r̂ = ê
|r| x
+ ê
|r| y
+ ê
|r| z
and r = x2 + y 2 + z 2 , we have
[( ) ( )] [ ( ) ( )]
∂ z ∂ y ∂ x ∂ z
∇ × [f (r)r̂] = f (r) − f (r) êx + f (r) − f (r) êy
∂y |r| ∂z |r| ∂z |r| ∂x |r|
[ ( ) ( )]
∂ y ∂ x
+ f (r) − f (r) êz
∂x |r| ∂y |r|
[ ( ) ( )] [ ( ) ( )]
y d z z d y z d z x d y
= f (r) − f (r) êx + f (r) − f (r) êy
r dr |r| r dr |r| r dr |r| r dr |r|
[ ( ) ( )]
x d z y d y
+ f (r) − f (r) êz
r dr |r| r dr |r|
= 0. (2.36)

Example 2.4.2. A NONZERO CURL


Calculate F = ∇ × (−yêx + xêy ).



êx êy êz

F = ∂
∂x

∂y

∂z
= 2êz .
(2.37)

−y x 0

Example 2.4.3. Prove ∇ · (a × b) = b · ∇ × a − a · ∇ × b.

( )
∂ ∂ ∂
∇ · (a × b) = x̂ + ŷ + ẑ · [(ay bz − az by )x̂ + (az bx − ax bz )ŷ + (ax by − ay bx )ẑ]
∂x ∂y ∂z
∂ay ∂bz ∂by ∂az
= bz + ay − az − by
∂x ∂x ∂x ∂x
∂bx ∂az ∂bz ∂ax
+az + bx − ax − bz
∂y ∂y ∂y ∂y
∂by ∂ax ∂bx ∂ay
+ax + by − ay − bx
∂z ∂z ∂z ∂z
= b·∇×a−a·∇×b (2.38)

Or, you can simply treat it as a scalar triple product by using



∇ · (a × b) = ϵijk ∇i (aj bk )
ijk

= ϵijk (aj ∇i bk + bk ∇i aj )
ijk
= ϵ123 (ay ∇x bz + bz ∇x ay ) + ϵ132 (az ∇x by + by ∇x az )
+.........
= b·∇×a−a·∇×b (2.39)
2.4. CURL, ∇× 11

Figure 2.8: A region threaded by B field due to a current passing normally throughout the
surface.

Figure 2.9: Path for computing circulation at (x0 , y0 ).

2.4.1 Geometrical Interpretation of Curl

If there is a straight wire with current passing through it, a B field would be generated and
directs circularly around the wire. We use the word ’Curl’ to represent a field V(x, y, z) that
is associative with the circulation of fluid around a loop. As what we show in Fig. 2.8, in
which you can imagine a region threaded by ’huge’ B field due to a current I passing normally
throughout the surface. To analyze the relation between B and I, we cut the surface into
small pieces such that we are able to analyze the ’small’ circulation around a differential loop
demonstrated in Fig. 2.9. And you can have an intuitive guess that I am going to tell you that
as you sum over all the ’small’ circulation, you get the ’huge’ circulation.

The so-called circulation is to integrate V along a closed path 1 → 2 → 3 → 4. Assume the


12 CHAPTER 2. VECTOR ANALYSIS

investigation is taken on xy plane, then


∫ ∫ ∫ ∫
circulation1234 = Vx (x, y)dlx + Vy (x, y)dly + Vx (x, y)dlx + Vy (x, y)dly
1 2 3 4
∫ x0 +∆x/2 ∫ y0 +∆y/2
= Vx (x, y0 − ∆y/2)dx + Vy (x0 + ∆x/2, y)dy
x0 −∆x/2 y0 −∆y/2
∫ x0 −∆x/2 ∫ y0 −∆y/2
+ Vx (x, y0 + ∆y/2)dx + Vy (x0 − ∆x/2, y)dy
x0 +∆x/2 y0 +∆y/2
≈ Vx (x0 , y0 − ∆y/2)∆x + Vy (x0 + ∆x/2, y0 )∆y
− Vx (x0 , y0 + ∆y/2)∆x − Vy (x0 − ∆x/2, y0 )∆y
[ ]
∂Vx (x, y) ∆y ∂Vx (x, y) ∆y
≈ Vx (x, y) − − Vx (x, y) − ∆x
∂y 2 ∂y 2 x=x0 ,y=y0
[ ]
∂Vy (x, y) ∆x ∂Vy (x, y) ∆x
+ Vy (x, y) + − Vy (x, y) + ∆y
∂y 2 ∂y 2 x=x0 ,y=y0

∂Vx (x, y) ∂Vy (x, y)
= − ∆x∆y + ∆x∆y (2.40)
∂y x0 ,y0 ∂x x0 ,y0
≡ [∇ × V(x, y)]z ∆x∆y. (2.41)

In this case we ’lock’ V at some point z0 and just trace the circulation of flow on xy plane. So
[∇ × V(x, y)]z denotes the z-component of ∇ × V(x, y).
So in general, for a vector F(x, y, z) passing through an area the circulation around its
perimeter yields

[∇ × F(x, y, z)]z ∆x∆y


[∇ × F(x, y, z)]x ∆y∆z
[∇ × F(x, y, z)]y ∆x∆z. (2.42)

Therefore, we obtain a general expression about the circulation:


I ∫
F(r) · dl = [∇ × F(r)] · n̂da. (2.43)
C S

Figure 2.10: Direction of normal for the shaded rectangle when perimeter of the surface is
traversed as indicated.

In fluid dynamics, ∇ × V denotes the ’vorticity’ of the flow. Whenever the curl of a vector
V vanishes, ie., ∇ × V = 0, V is said as an irrotational field. For example, gravitational and
2.4. CURL, ∇× 13

electrostatic fields are both irrotational. In each case


1
V=C r̂. (2.44)
r2

We shall prove later that whenever a vector is irrotational, the vector may be written as the
(negative) gradient of a scalar potential. For example, the electrostatic field satisfies

∇ × E = 0, (2.45)

indicating that
E = −∇ϕ, (2.46)

in which ϕ is the electrostatic potential.

Example 2.4.4. VECTOR POTENTIAL OF A CONSTANT B FIELD


From electrodynamics we know that ∇ · B = 0, which implies that B can be written as the
curl of another vector A, ie., B = ∇ × A, where A is called the vector potential. Because
a · (b × c) = b · (c × a) = c · (a × b), therefore

∇ · (∇ × A) = (∇ × ∇) · A = 0. (2.47)

For an arbitrary B, it is easily to show that

1
A = (B × r). (2.48)
2

For example, if a constant B field is applied in the z-direction, ie., B = B0 ẑ, then

1∑
A = ϵijk ei Bj rk (2.49)
2 ijk
1
= (−yx̂ + xŷ)B0 . (2.50)
2

You can verify that

∑ ( )

∇×A = ϵijk ei Ak
ijk
∂rj
( )
B0 ∂Ax ∂Ay
= ϵzyx ez + ϵzxy ez
2 ∂y ∂x
( )
B0 ∂Ax ∂Ay
= −ez + ez
2 ∂y ∂x
= B 0 ez . (2.51)

Property 2.4.1. BAC-CAB Rule


Verify the expansion of the triple vector product

A × (B × C) = B(A · C) − C(A · B)
14 CHAPTER 2. VECTOR ANALYSIS

A × (B × C) = (Ax î + Ay ĵ + Az k̂) × [(Bx î + By ĵ + Bz k̂) × (Cx î + Cy ĵ + Cz k̂)]


= (Ax î + Ay ĵ + Az k̂) × [Bx Cy k̂ − Bx Cz ĵ − By Cx k̂ + By Cz î + Bz Cx ĵ − Bz Cy î]
= Ax (Bz Cx − Bx Cz ) k̂ + Ax (By Cx − Bx Cy ) ĵ
+Ay (Bx Cy − By Cx ) î + Ay (Bz Cy − By Cz ) k̂
+Az (By Cz − Bz Cy ) ĵ + Az (Bx Cz − Bz Cx ) î (2.52)
= [Bx (Ay Cy + Az Cz ) − Cx (Ay By + Az Bz )] î
+[By (Az Cz + Ax Cx ) − Cy (Az Bz + Ax Bx )] ĵ
+[Bz (Ax Cx + Ay Cy ) − Cz (Ax Bx + Ay By )] k̂ (2.53)
= B(A · C) − C(A · B) (2.54)

To express these tedious equations in a more concise manner, we introduce a notation ϵijk to
simply rule out the permutable and cyclic properties of vector products. Since the unit vectors
ex , ey , ez that span the Euclidian space are mutually orthogonal, we have

ei × ej = ek , (2.55)

in which i, j, k = x, y, z may be placed in cyclic order. Thus we define a permutation symbol


to express the above result as
ei × ej = ϵijk ek , (2.56)

in which

 1, even permutation;
ϵijk = −1, odd permutation; (2.57)

0, if one index is equal to another one.

So, the vector product C = A × B would be expressed as

Ci = (ϵijk Aj Bk + ϵikj Ak Bj )ei


Cj = (ϵjik Ai Bk + ϵjki Ak Bj )ej
Ck = (ϵkij Ai Bj + ϵkji Aj Bi )ek
(2.58)

Explicitly, we just have



C= ϵijk ei Aj Bk . (2.59)
ijk

Then, let’s rewrite the BAC-CAB rule by using ϵijk . First, we have


D ≡ (B × C) = ϵijk ei Bj Ck . (2.60)
ijk
2.4. CURL, ∇× 15

So

A×D = ϵlmi Am Di el
(
lmi
)
∑ ∑
= ϵlmi ϵjki Am Bj Ck el
lm i
jk

= (δlj δmk − δlk δmj ) Am Bj Ck el (2.61)
lm
jk
∑ ∑
= Ak Bj Ck ej − Aj Bj Ck ek
jk kj
= B(A · C) − C(A · B), (2.62)

in which we have used an important property of ϵijk



ϵijk ϵlmk = δil δjm − δim δjl (2.63)
k

where δij is known as the Kronecker delta, defined as


{
1, i = j;
δij = (2.64)
0, i ̸= j.
16 CHAPTER 2. VECTOR ANALYSIS

2.5 Integral Theorems of Gradient, Divergence, & Curl

Figure 2.11: A square tube of cross section dx2 dx3 .

Figure 2.12: A volume consist of many cubes. The surface integral upon the interface between
two adjacent cubes cancels each other. So to sum up the contributions from each cube effectively
equals to evaluate the surface integral of the field with respect to the outmost surface.

2.5.1 Gauss’ Theorem


This theorem relates the surface integral of a given vector function and the volume integral
of the divergence of that vector. problems of double and triple integrals. If a continuous,
differentiable vector field A is defined in a simply connected region of volume V bounded by a
closed surface S, then the theorem states that
∫ I
∇ · A dV = A · n̂ da, (2.65)
V S

where dV = dx1 dx2 dx3 and n̂ is the normal vector of the surface element da.
To prove this, first we write
∫ ∫ ∑ 3
∂Ai
∇ · A dV = dV. (2.66)
V V i=1 ∂xi
2.5. INTEGRAL THEOREMS OF GRADIENT, DIVERGENCE, & CURL 17

According to Fig. 2.11, in which a volume element dV intersects the surface S at P and Q and
thus defines two surface elements daP and daQ . Therefore Eq. (2.66) has the form
∫ (∫ ′ ∫ ′′ ∫ Q )I
P P
∂A1 ∂A1
dV = + +..... + dx1 dx2 dx3
V ∂x1 P P′ P ∗ ∂x1
I
= dx2 dx3 [A1 (Q) − A1 (P )]
I I
= A1 (Q) daQ − A1 (P ) daP
I I
≈ A1 (Q) · n̂da + A1 (P ) · n̂da. (2.67)

So similarly, we can evaluate the x2 and x3 components. And finally we should obtain the
Gauss’s theorem ∫ I
[∇ · A] dV = A · n̂da. (2.68)
V S

Figure 2.13:

Example 2.5.1. Evaluate



I= (x3 dydz + x2 ydzdx + x2 zdxdy),
S

where S is the closed surface of the cylinder in Fig. 2.13.


In this case, each component of the vector field F reads: Fx = x3 , Fy = x2 y, Fz = x2 z.
Therefore ∇ · F = 5x2 . Then we can apply Gauss’s theorem to calculate the surface integral,
which gives
∫ ∫ z ∫ 2π ∫ a
I = (∇ · F) dV = 5x2 dV
V 0 0 0
∫ b ∫ 2π ∫ a
5π 4
= 5r2 cos2 ϕ rdr dϕ dz = a b. (2.69)
0 0 0 4


Example 2.5.2. Evaluate the surface integral I = S F·dS, where F = (y−x)î+x2 z ĵ+(z+x2 )k̂,
and S is the open surface of the hemisphere x2 + y 2 + z 2 = a2 , z > 0.
18 CHAPTER 2. VECTOR ANALYSIS

Figure 2.14: The surface of the hemisphere x2 + y 2 + z 2 = a2 , z > 0.

In this case, the surface includes a hemisphere and a circle on the ground. So we have
∫ ∫ ∫
(∇ · F) dV = F · n̂dS + F · n̂dS. (2.70)
V Sh Sc
∫ ∫
However, since ∇ · F = 0, which implies Sh
F · n̂dS = − Sc
F · n̂dS.
So for the circle, n̂ = −k̂, and dS = dxdy; for the hemisphere, n̂ = r̂, and dS = r2 sin θdθdϕ.
Then, the surface integral over the open hemisphere is given by
∫ ∫ ∫
F · n̂dS = − F · n̂ dS = x2 dxdy
Sh Sc Sc
∫ 2π ∫ a
= −r2 cos2 ϕ rdrdϕ
0 0
πa4
= . (2.71)
4

Figure 2.15: Demonstration of Stoke’s theorem.

2.5.2 Stoke’s Theorem


Stoke’s theorem is the ’curl analogue’ of the divergence theorem and relates the integral of the
curl of a vector field over an open surface S to the line integral of the vector field around the
2.5. INTEGRAL THEOREMS OF GRADIENT, DIVERGENCE, & CURL 19

perimeter C bounding the surface. The math expression is shown by


I ∫
F(r) · dl = [∇ × F(r)] · n̂da. (2.72)
C S

So it’s important to notice that if S is a closed surface, the line integral vanishes since there is
no perimeter in this case. Therefore,

[∇ × F(r)] · n̂da = 0, (2.73)
S

for S a closed surface.

Example 2.5.3. Given the vector field F = y î − xĵ + z k̂, verify Stoke’s theorem for the
hemispherical surface x2 + y 2 + z 2 = a2 , z ≥ 0.
In this case ∇ × F = −2k̂, and the surface element on the hemisphere is dS = a2 sin θdθdϕ.
So the surface integral gives
∫ ∫ 2π ∫ π/2
[∇ × F(r)] · n̂ dS = −2k̂ · r̂ a2 sin θdθdϕ
S 0 0
∫ 2π ∫ π/2
= −2 cos θ a2 sin θdθdϕ
0 0
= −2πa . 2
(2.74)

Whereas the line integral over the perimeter of the circle leads to
I I
F(r) · dl = (y î − xĵ + z k̂) · (dx î + dy ĵ) (2.75)
C
IC

= (y dx − x dy). (2.76)
C

It is convenient to evaluate the integration in polar coordinates where x = a cos ϕ and y =


a sin ϕ. Then the above equation becomes
∫ 2π
(−a2 sin2 ϕ − a2 cos2 ϕ) dϕ = −2πa2 . (2.77)
0

Figure 2.16: Direction of B given by Oersted’s law.


20 CHAPTER 2. VECTOR ANALYSIS

Example 2.5.4. OERSTED’S AND FARADAY’S LAWS

Consider the magnetic field generated by a long wire that carries a time-independent current
I. The relevant Maxwell equation takes the form ∇ × B = µ0 J. Integrating this equation over
a disk S perpendicular to and surrounding the wire according to Fig. 2.17, we have
∫ ∫
1
I= J · n̂da = ∇ × B · n̂da. (2.78)
S µ0 S
Applying Stoke’s theorem, now we have
I
1
I= B · dl. (2.79)
µ0 C

Similarly, we can integrate Maxwell’s equation for ∇ × E. While moving the magnet to the
left with the loop held still, Faraday found a current flow in the loop. This can be illustrated
by Stoke’s theorem: ∫ ∫
d dΦ
∇ × E · n̂da = − B · n̂da = − , (2.80)
S dt S dt
in which Φ is the magnetic flux through the area S, which means
I

E · dl = − . (2.81)
C dt
H
This is Faraday’s law stating that the motional emf E = C E · dl is induced in the wire loop
whenever the magnetic flux through the loop changes.

Figure 2.17: Electromagnetic induction caused by the change of magnetic flux.

2.6 SUCCESSIVE APPLICATIONS OF ∇


Letting ∇ operate on gradient, divergence, or curl we obtain
(a)∇ · ∇ϕ (b)∇ × ∇ϕ (c)∇∇ · V
(d)∇ · ∇ × V (e)∇ × ∇ × V (2.82)
(a) Laplacian of ϕ = ∇ · ∇ϕ ≡ ∇2 ϕ, which by definition is of the form
( ) ( )
∂ ∂ ∂ ∂ϕ ∂ϕ ∂ϕ
∇ · ∇ϕ = x̂ + ŷ + ẑ · x̂ + ŷ + ẑ
∂x ∂y ∂z ∂x ∂y ∂z
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ
= + + 2. (2.83)
∂x2 ∂y 2 ∂z
2.6. SUCCESSIVE APPLICATIONS OF ∇ 21

For a charge free system, the electrostatic potential ϕ satisfies Laplace’s equation which reads

∇2 ϕ = 0. (2.84)

On the other hand, whenever there are free charges, ϕ obeys Poisson’s equation given by
ρ
∇2 ϕ = − , (2.85)
ϵ0
in which ρ is the charge density distribution.

Example 2.6.1. LAPLACIAN OF A CENTRAL FIELD POTENTIAL

Calculate ∇2 g(r). By definition,


( ) ( )
∂ ∂ ∂ ∂g(r) ∂g(r) ∂g(r)
∇ g(r) =
2
x̂ + ŷ + ẑ · x̂ + ŷ + ẑ
∂x ∂y ∂z ∂x ∂y ∂z
( ) ( )
∂ ∂ ∂ dg(r) ∂r dg(r) ∂r dg(r) ∂r
= x̂ + ŷ + ẑ · x̂ + ŷ + ẑ
∂x ∂y ∂z dr ∂x dr ∂y dr ∂z
∂ ′ ∂ ′ ∂
= [g (r)r′ (x)] + [g (r)r′ (y)] + [g ′ (r)r′ (z)]
∂x ∂y ∂z
′ ′
∂g (r) ∂r (x)
= r′ (x) + g ′ (r) + ( y term) + ( z term)
∂x ∂x
2
= g ′′ (r) + g ′ (r). (2.86)
r
In special case, g(r) = rn , this reduces to

∇2 rn = n(n + 1)rn−2 . (2.87)

This vanishes for n = 0 or n = −1, implying ϕ = constant or ϕ = 1/r. Namely, ϕ = 1/r is a


solution of Laplace’s equation for r ̸= 0. Again, at r = 0, derivatives of 1/r is undefined and
∇2 (1/r) yields a Dirac delta function δ(0).

(b) ∇ × ∇ϕ is nothing but a curl operating upon a vector ∇ϕ. As a consequence, which
gives us

x̂ ŷ ẑ

∂ ∂ ∂

∇ × ∇ϕ = ∂x ∂y ∂z
∂ϕ ∂ϕ ∂ϕ


∂x ∂y ∂z

= ϵijk ∇j (∇ϕ)k êi
ijk
( ) ( ) ( )
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂2ϕ ∂ 2ϕ ∂ 2ϕ
= − x̂ + − ŷ + − ẑ
∂y∂z ∂z∂y ∂z∂x ∂x∂z ∂x∂y ∂y∂x
= 0. (2.88)

This result sets up a very important principle: the curl of the gradient of an arbitrary scalar
function ϕ must be zero. Because a nonzero ∇ϕ is essentially a divergent field V, which of
course must be curless, or irrotational.
22 CHAPTER 2. VECTOR ANALYSIS

Similarly, we have (d) ∇ · ∇ × V evaluated by


∂ ∂ ∂
∂x ∂y ∂z
∂ ∂ ∂
∇ · ∇ × V = ∂x ∂y ∂z
= 0,
(2.89)
V V V
x y z

since the determinant has two identical rows. This states that the divergence of a curl vanishes,
or all curls are solenoidal.

Figure 2.18: http://web.ncf.ca/ch865/englishdescr/Solenoid.html

According to BAC-CAB rule, expressions (c) and (e) can be associated with the so-called
vector Laplacian ∇2 V by

∇ × ∇ × V = ∇∇ · V − ∇ · ∇V. (2.90)

Example 2.6.2. MAXWELL’S EQUATIONS AND ELECTROMAGNETIC WAVE EQUA-


TION

In vacuum, where there are no free charges and currents, Maxwell’s equations are

∇ · E = 0, (2.91a)
∇ · B = 0, (2.91b)
∂B
∇×E=− , (2.91c)
∂t
∂E
∇ × B = ϵ0 µ0 . (2.91d)
∂t
Performing time derivative on Eq. (2.91d) and applying Eq. (2.91c), we obtain

∇ × ∇ × E = ∇∇ · E − ∇ · ∇ × E
∂ 2E
∴ ∇ · ∇E = ϵ0 µ0 2 . (2.92)
∂t
The solution of the wave equation, Eq. (2.92), takes the form

sin(k · r − ωt) or cos(k · r − ωt), (2.93)



from which we find ω/|k| = 1/ ϵ0 µ0 = c; in vacuum, the electromagnetic waves travel by speed
of light. Therefore, the general solution given by the linear combination of above functions
would be
E(r, t) = E0 e±i(k·r−ωt) , (2.94)
2.6. SUCCESSIVE APPLICATIONS OF ∇ 23

in which E0 is the amplitude, k is the wave vector, and +(−) indicates a right- (left-) going
wave. Since the electric field is a transverse field, it polarizes perpendicularly to the transport
direction. Therefore, assume the wave travelling along the z-direction, a linearly polarized wave
can be written as

Ex (z, t) = E0x ei(kz−ωt) ϵ̂x or


Ey (z, t) = E0y ei(kz−ωt) ϵ̂y , (2.95)

whereas a right circular wave looks like


1
ER (z, t) = − √ (ϵ̂x + iϵ̂y )E0 ei(kz−ωt) . (2.96)
2
From Eq. (2.91b) we can write B = ∇ × A, where A is called vector potential. Inserting it
into Eq. (2.91c) we obtain ( )
∂A
∇× E+ = 0. (2.97)
∂t
Again, we encounters a curless field and therefore we can set
∂A
E+ = −∇ϕ, (2.98)
∂t
here ϕ = ϕ(r, t), is defined as the (nonstatic) electric potential. In summary, we have the
general relationship for E and B for homogeneous Maxwell’s equations:

B = ∇×A (2.99)
∂A
E = −∇ϕ − . (2.100)
∂t
Next, it should be no doubt that the presence of external charges and currents complicate
the analysis on Maxwell’s equations. For Gauss’s law,
ρ
∇·E= , (2.101)
ϵ0
and for Amper-Maxwell’s law
1 ∂E
∇×B− = µ0 J, (2.102)
c2 ∂t
or by integral form ∫
1 ∂ΦE
(∇ × B) · n̂da = + µ0 Ienc , (2.103)
c2 ∂t
in which Ienc denotes the enclosed current inside a virtual loop. As shown in Fig. 2.19, when
currents pass by metal plates and charge on them would simultaneously induce an E field across
the plates. Thus a fictional displacement current Id appears accompanied with the time-varying
electric flux. During charging, a magnetic field is generated by both Ienc and Id .

According to Gauss’s law and Amper-Maxwell’s law, we know that

∇ × (∇ × A) = ∇∇ · A − ∇ · ∇A
( )
1 ∂ϕ 1 ∂ 2A
∴ −∇ A + ∇ ∇ · A + 2
2
+ 2 2 − µ0 J = 0 (2.104)
c ∂t c ∂t
24 CHAPTER 2. VECTOR ANALYSIS

Figure 2.19:

So if we choose Lorentz gauge such that


1 ∂ϕ
∇·A+ = 0,
c2 ∂t
then we obtain decoupled differential equations for both potentials:
1 ∂ 2A
2 2
− ∇2 A = µ0 J, (2.105)
c ∂t
and

∇ · E = −∇2 ϕ − ∇·A
∂t
1 ∂ 2ϕ ρ
= −∇2 ϕ + 2 2 = . (2.106)
c ∂t ϵ0
Moreover, if we perform ∇· on Amper-Maxwell’s equation, we obtain

∇ · (∇ × B) = 0, (2.107)

which yields
1 ∂
∇ · E = µ0 ∇ · J, (2.108)
c2 ∂t
that is just the continuity equation of the current density:
∂ρ
∇·J=− . (2.109)
∂t

2.7 POTENTIAL THEORY


2.7.1 Scalar Potential
If a force over a given simply connected region of space S (which means that it has no holes)
can be expressed as the negative gradient of a scalar function φ,

F = −∇φ, (2.110)
2.7. POTENTIAL THEORY 25

where φ is called the scalar potential. The force F appearing as the negative gradient of a
single-valued scalar potential is labeled a conservative force. Eq. (2.110) implies that

∇ · (∇ × F) = 0, (2.111)

and I I I
F · dl = − ∇φ · dl = − dφ = 0. (2.112)
C C C

The vanishment of line integral of F along a closed path showing no work done by the force
around the trip verifies F a conservative force.

Example 2.7.1. GRAVITATIONAL POTENTIAL


Find the scalar potential for the gravitational force on a unit mass m1 ,

Gm1 m2
F=− r̂.
r2
If there exists a scalar potential φ, then F = −∇φ. To find φ, we perform the line integral on
F by ∫ r
Gm1 m2
F · dr = φ(∞) − φ(r) = . (2.113)
∞ r
If set the gravitational potential is zero at r = ∞, then

Gm1 m2
φ(r) = − (2.114)
r
serves as an absolute potential at r.

Example 2.7.2. CENTRIFUGAL POTENTIAL


Calculate the scalar potential for the centrifugal force per unit mass, FC = ω 2 r r̂. Applying
line integral we obtain ∫ r
ω2
φ(r) − φ(0) = − F · dr = − r2 . (2.115)
0 2
Taking φ(0) = 0 we have
ω2 2
φ(r) = − r .
2
If the force is inward, ie., Fsho = −kr r̂, Fsho is a centripetal force. In this case, the potential
is of the form
1
φ(r) = kr2 ,
2
which serves the potential of a simple harmonic oscillator.

2.7.2 Vector Potential


Analogies of E and φ in electrostatics are B and A in magnetostatics, and the corresponding
formula are
E = −∇φ (2.116)
and
B = ∇ × A, (2.117)
26 CHAPTER 2. VECTOR ANALYSIS

respectively. If Eq. (2.117) holds, then ∇ · B = 0 and B is solenoidal. To prove A does exist
for a solenoidal B, we start by setting B = b1 x̂ + b2 ŷ + b3 ẑ and A = a1 x̂ + a2 ŷ + a3 ẑ. Then
Eq. (2.117) reads that
Bi = ϵijk ∇j Ak î, (2.118)
or equivalently

∂a3 ∂a2
b1 = −
∂y ∂z
∂a1 ∂a3
b2 = −
∂z ∂x
∂a2 ∂a1
b3 = − . (2.119)
∂x ∂y

If A is on xy plane, then

∂a2
b1 = −
∂z
∂a1
b2 =
∂z
∂a2 ∂a1
b3 = − , (2.120)
∂x ∂y

which means
∫ z
a1 = b2 dz + f1 (x, y) (2.121a)

z0
z
a2 = − b1 dz + f2 (x, y), (2.121b)
z0

and
∫ z ( )
∂a2 ∂a1 ∂b1 ∂b2 ∂f2 (x, y) ∂f1 (x, y)
− = − + dz + −
∂x ∂y z0 ∂x ∂y ∂x ∂y
∂f2 ∂f1
= b3 |zz0 + −
∂x ∂y
∂f2 ∂f1
= b3 (x, y, z) − b3 (x, y, z0 ) +
− . (2.122)
∂x ∂y
∫x
Without loss of generality, we can always set f1 = 0 and f2 = x0 b3 (x, y, z0 ) dx such that

∂a2 ∂a1
− = b3 (x, y, z). (2.123)
∂x ∂y

And eventually, we can construct


∫ z ∫ z ∫ x
A= b2 (x, y, z) dz x̂ − b1 (x, y, z) dz ŷ + b3 (x, y, z0 ) dx ŷ. (2.124)
z0 z0 x0

Example 2.7.3. AMAGNETIC VECTOR POTENTIAL FOR A CONSTANT MAGNETIC


FIELD
2.7. POTENTIAL THEORY 27

Consider a constant B field B = B0 ẑ applied along z direction. By Eq. (2.124) the vector
potential A can be constructed to be

A = B0 x ŷ. (2.125)

The expression of Eq. (2.125) is not unique and is often called the Landau gauge. To seek for
another possible solution, we come to

∂a2
= 0,
∂z
∂a1
= 0,
∂z
∂a2 ∂a1
− = B0 . (2.126)
∂x ∂y

Then we may set a1 = a1 (y) and a2 = a2 (x) due to Eq. (2.126), which results in

B0 1
A= (−y x̂ + x ŷ) = (B × r). (2.127)
2 2
In this expression A is said to be in symmetric gauge.
Physically speaking, gauge in Hilbert space is something like coordinates in Euclidian space.
For the later, you can describe a ’math’ vector in different orthonormal coordinates without
changing its length by making a coordinate transformation. Analogically for the former, you can
measure some physical quantities in different gauges with no change of their own by making a
gauge transformation. For example, when we talk about B = ∇×A we know that any gradient
field would be curless. That is, we may safely set

A′ = A + ∇ϕ. (2.128)

Adding ∇ϕ in A does not affect B; this is called a gauge transformation. In other words, B is
said to be gauge invariant. In quantum mechanics, if an electron suffers a magnetic field the
kinetic term in the Hamiltonian would be written as
p2 1 [ e ] [ e ]
ψ → −i~∇ + A · −i~∇ + A ψ
2m 2m [ c c ]
1 i~e i~e e2 2
= ~ ∇ ψ−
2 2
∇ · (Aψ) − A · ∇ψ + 2 A ψ . (2.129)
2m c c c

In this case, Coulomb gauge is often chosen to let ∇ · A = 0. And then we have
[ ]
p2 1 2i~e e2 2
ψ → ~ ∇ ψ−
2 2
A · ∇ψ + 2 A ψ . (2.130)
2m 2m c c

You may ask that why it is A but not B in the equation? Yes, I do not make a mistake.
When you learn Lagrangian mechanics you may derive the Lagrange of the system from the
Hamiltonian. And then you proceed to derive the Euler equation to obtain the equation of
motion of the electron in the EM fields. You will see that the dynamics of the electron is
governed by Lorentz force F = −e(E+v×B) as it should be. This is a beautiful example, in my
opinion, to demonstrate the consistency between classical mechanics and quantum mechanics;
furthermore, it shades light on the quantum mechanics.
28 CHAPTER 2. VECTOR ANALYSIS

2.8 VECTOR ANALYSIS IN CURVED COORDINATES


In applying physics laws in most situations the results should be coordinate independent. Also,
the results should be independent of choice of origin of the coordinates. The use of vectors
gives us this independence.
Quantities that invariant under coordinate transformation are called scalars. For example,
we have mass or temperature that can be described by
M (x, y) = M (x′ , y ′ ). (2.131)
However, to describe more complicated phenomena we need vectors.

2.8.1 Rotation Matrix

x 2 -axis
x'2 -axis
x2 P
x'1 x'1 -axis
d
c
x'2 a b
e θ

θ
x1 -axis
O x1

Figure 2.20:

Consider a point P with coordinates (x1 , x2 , x3 ) with respect to some coordinate system.
By rotating the system with respect to a certain axis by an angle θ, how to point P in the new
coordinate? The situation can be illustrated by a 2D case as shown in Fig. 2.20.
x′1 = x1 cos θ + x2 sin θ
= x1 cos θ + x2 cos(π/2 − θ). (2.132)

x′2 = −x1 sin θ + x2 cos θ


= x1 cos(π/2 + θ) + x2 cos θ. (2.133)
If we set the angle between x′1 -axis and x1 -axis as (x′1 , x1 ), then we can define, in general,
the cosine of (x′i , xj ) to the λij :
λij ≡ cos(x′i , xj ). (2.134)
So from Fig. 2.20 we know that
λ11 = cos(x′1 , x1 ) = cos θ
λ12 = cos(x′1 , x2 ) = cos(π/2 − θ) = sin θ
λ21 = cos(x′2 , x1 ) = cos(π/2 + θ) = − sin θ
λ22 = cos(x′2 , x2 ) = cos(θ). (2.135)
2.8. VECTOR ANALYSIS IN CURVED COORDINATES 29

The position of P in new coordinates is specified by

x′1 = x1 cos(x′1 , x1 ) + x2 cos(x′1 , x2 )


= λ11 x1 + λ12 x2

x2 = x1 cos(x′1 , x1 ) + x2 cos(x′2 , x2 )
= λ21 x1 + λ22 x2 . (2.136)

Of course, we can make an extension to 3D case:

x′1 = λ11 x1 + λ12 x2 + λ13 x3


x′2 = λ21 x1 + λ22 x2 + λ23 x3
x′3 = λ31 x1 + λ32 x2 + λ33 x3 . (2.137)

Or in brief, ∑
x′i = λij xj , i = 1, 2, 3. (2.138)
j=1,3

Similarly, the inverse transformation is

x1 = x′1 cos(x′1 , x1 ) + x′2 cos(x′2 , x1 ) + x′3 cos(x′3 , x1 )


= λ11 x′1 + λ21 x′2 + λ31 x′3 , (2.139)

Or in a compact form, ∑
xi = λij x′j , i = 1, 2, 3. (2.140)
j=1,3

The quantity λij is called the directional cosine of the x′i -axis relative to the xj -axis. It is
convenient to write λ in the matrix form:
 
λ11 λ12 λ13
λ =  λ21 λ22 λ23  (2.141)
λ31 λ32 λ33

When λ is defined and specified for the transformation properties of coordinates, it is called a
transformation matrix or a rotation matrix.
Example 2.8.1. For example, if a point P in Fig. 2.21 is represented in (x1 , x2 , x3 ) system
by P (2, 1, 3), then what is the rotation matrix that determines the same point as P (x′1 , x′2 , x′3 ),
when x2 has been rotated toward x3 around the x1 -axis by an angle of 30◦ ?

λ11 = cos(x′1 , x1 ) = 1
λ12 = cos(x′1 , x2 ) = 0
λ13 = cos(x′1 , x3 ) = 0
λ21 = cos(x′2 , x1 ) = 0

λ22 = cos(x′2 , x2 ) = 3/2
λ23 = cos(x′2 , x3 ) = 1/2
λ31 = cos(x′3 , x1 ) = 0
λ32 = cos(x′3 , x2 ) = −1/2

λ33 = cos(x′3 , x3 ) = 3/2 (2.142)
30 CHAPTER 2. VECTOR ANALYSIS

x3
x '3

x '2

30°
P x2

x1
x'1

Figure 2.21:

So, P (x′1 , x′2 , x′3 ) is specified by

x′1 = x1 = 2

x′2 = 3/2x2 + 1/2x3 = 2.37


x3 = −1/2x2 + 3/x3 = 2.10 (2.143)

It should be notice that the rotation operator preserves the length of the position vector, that
is, √ √
r = x1 + x2 + x3 = x′12 + x′22 + x′32 = 3.74
2 2 2
(2.144)
So we can redefine a scalar and a vector in terms of the transformation property. Consider
a transformation as Eq. (2.138), if a quantity ϕ is unaffected, then ϕ is called a scalar. On the
other hand if quantities (A1 , A2 , A3 ) is transformed under the rotation of the coordinate, then
A is termed a vector.

λA λB

90° rotation 90° rotation


about x3 -axis about x2 -axis
x3

x2

x1

λB λA

90° rotation 90° rotation


about x2 -axis about x3 -axis

Figure 2.22:
2.8. VECTOR ANALYSIS IN CURVED COORDINATES 31

2.8.2 Exmaples of Derivatives in Polar Coordinates

s(t)

P (2)

dr
ds

r dθ P (1)
r2(t )

r1(t )
e(2)
r
der

eθ(1) e(1)
r

deθ
dθ θ
eθ(2)

Figure 2.23:

Consider a particle moving from P (1) to P (2) along the curve s(t) in the time interval dt
as shown in Fig. 2.23. In polar coordinates, we specify the position of the particle by r and θ
and label the unit vectors by er and eθ . A displacement ds would induce change in er and eθ
corresponding to
r − er
der = e(2) (1)
(2.145)
and
(2) (1)
deθ = eθ − eθ , (2.146)
which also imply that
der = dθ eθ (2.147)
deθ = −dθ er . (2.148)
Applying time derivative on both sides we obtain
ėr = θ̇ eθ (2.149)
ėθ = −θ̇ er . (2.150)
Now, the velocity of particle can be expressed as
dr d(r er )
v = =
dt dt
= ṙ er + rθ̇ eθ . (2.151)
That is, the velocity of a roaming particle can be resolved into a radial component ṙ er and an
angular component rθ̇ eθ .
Similarly, the acceleration can be analyzed as follows:
dv d
a = = (ṙ er + rθ̇ eθ )
dt dt
= (r̈ − rθ̇2 ) er + (2ṙθ̇ + rθ̈) eθ . (2.152)
32 CHAPTER 2. VECTOR ANALYSIS

2.8.3 Cylindrical Coordinates

z = x3

ez
r eφ

z er
x2
φ

x1

Figure 2.24: 5

The cylindrical coordinates (r, ϕ, z) of a point P are defined in terms of r: the distance
from the z axis; ϕ: angle between the vector, relying on x y plane and the x axis; and z: the
projection of the vector on z axis. The relation to Cartesian coordinates (x, y, z) is casted by

x = r cos ϕ y = r sin ϕ z = z (2.153)

The unit vectors r̂, ϕ̂, ẑ point at the direction of increase of the corresponding coordinates.
They constitute an orthonormal basis and any vector A can be expressed as

A = Ar r̂ + Aϕ ϕ̂ + Az ẑ. (2.154)

Since

l = x x̂ + y ŷ + z ẑ
= lr r̂ + lϕ ϕ̂ + lz ẑ, (2.155)

what is the relation between (x̂, ŷ, ẑ) and r̂, ϕ̂, ẑ? From equations

l = r cos ϕ x̂ + r sin ϕ ŷ + z ẑ (2.156)

and

dl = dlr r̂ + dlϕ ϕ̂ + dlz ẑ


= dr r̂ + rdϕ ϕ̂ + dz ẑ, (2.157)

then
lr dlr dlr
r̂ = = =
r dr dr
lϕ dlϕ dlϕ
ϕ̂ = = =
lϕ dlϕ rdϕ
lz dlz dlz
ẑ = = = . (2.158)
lz dlz dz
2.8. VECTOR ANALYSIS IN CURVED COORDINATES 33

From Eq. (2.156) we know

∂l
dlr = dr = ds cos ϕ x̂ + ds sin ϕ ŷ (2.159)
∂r
∂l
dlϕ = dϕ = −rdϕ sin ϕ x̂ + rdϕ cos ϕ ŷ (2.160)
∂ϕ
∂l
dlz = dz = dz ẑ (2.161)
∂z
Therefore, the unit vectors in cylindrical coordinates are
    
r̂ cos ϕ sin ϕ 0 x̂
 ϕ̂  =  − sin ϕ cos ϕ 0   ŷ  , (2.162)
ẑ 0 0 1 ẑ

and the inverse matrix is


    
x̂ cos ϕ − sin ϕ 0 r̂
 ŷ  =  sin ϕ cos ϕ 0   ϕ̂  . (2.163)
ẑ 0 0 1 ẑ

From Eq. (2.157) we can further determine an infinitesimal volume and surface in the
cylindrical coordinates:
dτ = dlr dlϕ dlz = rdrdϕ dz (2.164)
dar = r̂ drϕ dz = rdϕ dz r̂ (2.165)
If the surface lies in the x y plane,

daz = ẑ dlr dlϕ = rdrdϕ ẑ (2.166)

Cylindrical coordinates:
dv = r dr d φ dz
r dφ
dr

r
dz

Plane polar
φ coordinates:
r dφ da = r dr dφ

dr

Figure 2.25:
34 CHAPTER 2. VECTOR ANALYSIS

Example 2.8.2. Evaluate the surface and the volume of a cylinder with radius R and height
h.

∫ ∫
Sr = n̂ · dar = Rdϕ dz r̂ · r̂ = 2πRh.
∫ ∫ ∫
Sz = n̂ · daz = rdrdϕ|z=h ẑ · ẑ + rdrdϕ|z=0 (−ẑ) · (−ẑ) = 2πR2 .
∴ S = Sr + Sz = 2πR(h + R). (2.167)

V = r dr dϕ dz = πR2 h. (2.168)

2.8.4 Spherical Polar Coordinates

x3
er

r
θ eθ

x2
φ

x1

Figure 2.26:

The spherical polar coordinates (r, θ, ϕ) of a point P are defined in terms of r: the distance
from the origin; θ: the polar angle; and ϕ: the azimuthal angle. The relation to Cartesian
coordinates (x, y, z) is casted by

x = r sin θ cos ϕ y = r sin θ sin ϕ z = r cos θ (2.169)

The unit vectors r̂, θ̂, ϕ̂ point at the direction of increase of the corresponding coordinates.
They constitute an orthonormal basis and any vector A can be expressed as

A = Ar r̂ + Aθ θ̂ + Aϕ ϕ̂. (2.170)

Since

r = x x̂ + y ŷ + z ẑ
= rr r̂ + rθ θ̂ + rϕ ϕ̂, (2.171)

what is the relation between (x̂, ŷ, ẑ) and (r̂, θ̂, ϕ̂)? From equations

r = r sin θ cos ϕ x̂ + r sin θ sin ϕ ŷ + r cos θ ẑ (2.172)


2.8. VECTOR ANALYSIS IN CURVED COORDINATES 35

Spherical coordinates:
dv = r 2 sin θ dr dθ dφ

r sin θ

dr
da = r 2 sin θ dθ dφ

r
dθ r sin θ dφ
θ
r dθ

φ

Figure 2.27:

and
dr = dr r̂ + rdθ θ̂ + r sin θdϕ ϕ̂, (2.173)
Then
rr drr drr
r̂ = = =
r dr dr
rθ drθ drθ
θ̂ = = =
rθ drθ rdθ
rϕ drϕ drϕ
ϕ̂ = = = . (2.174)
rϕ drϕ r sin θdϕ

From Eq. (2.172) we know

∂r
drr = dr = dr sin θ cos ϕ x̂ + dr sin θ sin ϕ ŷ + dr cos θ ẑ (2.175)
∂r
∂r
drθ = dθ = rdθ cos θ cos ϕ x̂ + rdθ cos θ sin ϕ ŷ − rdθ sin θ ẑ (2.176)
∂θ
∂r
drϕ = dϕ = −rdϕ sin θ sin ϕ x̂ + rdϕ sin θ cos ϕ ŷ (2.177)
∂ϕ

Therefore, the unit vectors in spherical coordinates are


    
r̂ sin θ cos ϕ sin θ sin ϕ cos θ x̂
 θ̂  =  cos θ cos ϕ cos θ sin ϕ − sin θ   ŷ  , (2.178)
ϕ̂ − sin ϕ cos ϕ 0 ẑ

and the inverse matrix is


    
x̂ sin θ cos ϕ cos θ cos ϕ − sin ϕ r̂
 ŷ  =  sin θ sin ϕ cos θ sin ϕ cos ϕ   θ̂  . (2.179)
ẑ cos θ − sin θ 0 ϕ̂
36 CHAPTER 2. VECTOR ANALYSIS

From Eq. (2.173) we can further determine an infinitesimal volume and surface in the
spherical coordinates:
dτ = drr drθ drϕ = r2 sin θdrdθdϕ (2.180)
dar = r̂ drθ drϕ = r2 sin θdθdϕ r̂ (2.181)
If the surface lies in the x y plane and θ = π/2, then
daθ = θ̂ drr drϕ = rdrdϕ θ̂. (2.182)
Example 2.8.3. Find the surface and the volume of a sphere of radius R.
∫ ∫
Sr = n̂ · dar = R2 sin θ dθ dϕ r̂ · r̂ = 4πR2 . (2.183)

4π 3
V = r2 dr sin θ dθ dϕ = R . (2.184)
3

2.8.5 Orthogonal Coordinates


• Goal: To understand transformation relations between curvilinear coordinates and carte-
sian coordinates.
• Steps:
1. Write general curvilinear coordinates q1 , q2 , q3 , for example: r, θ, ϕ and know that
x = x(q1 , q2 , q3 )
y = y(q1 , q2 , q3 )
z = z(q1 , q2 , q3 ). (2.185)
Of course the inverse relations would be
q1 = q1 (x, y, z)
q2 = q2 (x, y, z)
q3 = q3 (x, y, z). (2.186)

2. Knowing that the magnitude of a vector would be coordinate invariant. In Cartesian


coordinates,
ds2 = dx2 + dy 2 + dz 2 , (2.187)
and in curvilinear coordinates,
ds2 = g11 dq12 + g12 dq1 dq2 + g13 dq1 dq3
+g21 dq2 dq1 + g22 dq22 + g23 dq2 dq3
+g31 dq3 dq1 + g32 dq3 dq2 + g33 dq32

= gij dqi dqj (2.188)
ij

To see what is gij we need another formulas


∂x ∂x ∂x
dx = dq1 + dq2 + dq3
∂q1 ∂q2 ∂q3
dy = . . .
dz = . . . (2.189)
2.8. VECTOR ANALYSIS IN CURVED COORDINATES 37

Then
( )2
2 ∂x ∂x ∂x
dx = dq1 + dq2 + dq3
∂q1 ∂q2 ∂q3
∑ ∂x ∂x
= dqi dqj
ij
∂qi ∂qj
∑ ∂y ∂y
dy 2 = ... = dqi dqj
ij
∂q i ∂q j
∑ ∂z ∂z
dz 2 = ... = dqi dqj (2.190)
ij
∂q i ∂q j

Therefore,
∂x ∂x ∂y ∂y ∂z ∂z
gij = + + (2.191)
∂qi ∂qj ∂qi ∂qj ∂qi ∂qj
Here we use without proving that gij = 0 if i ̸= j, in this way we may simply set
gii = h2i , in which
( )2 ( )2 ( )2
2 ∂x ∂y ∂z
hi = + + . (2.192)
∂qi ∂qi ∂qi
So now,
ds2 = (h1 dq1 )2 + (h2 dq2 )2 + (h3 dq3 )2 . (2.193)
The role of hi is actually a scale factor or a weight, with which each length component
in one coordinate can be related with that in another coordinate by

dsi = hi dqi . (2.194)

Therefore any infinitesimal displacement vector dr can be written as



dr = h1 dq1 ê1 + h2 dq2 ê2 + h3 dq3 ê3 = hi dqi êi (2.195)
i

Further the area element and the volume element can be expressed as

daij = dsi dsj = hi hj dqi dqj (2.196)

dτ = ds1 ds2 ds3 = h1 h2 h3 dq1 dq2 dq3 (2.197)


One thing should be noticed that the each area component must be specified by its
normal direction, ie.,

da = dai êi
i
= ds2 ds3 ê1 + ds1 ds3 ê2 + ds1 ds2 ê3
= h2 h3 dq2 dq3 ê1 + h1 h3 dq1 dq3 ê2 + h1 h2 dq1 dq2 ê3 (2.198)

3. Remember the weight for each coordinate components.


(a) cartesian: h1 = hx = 1, h2 = hy = 1, h3 = hz = 1
(b) spherical: h1 = hr = 1, h2 = hθ = r, h3 = hϕ = r sin θ
(c) cylindrical: h1 = hρ = 1, h2 = hϕ = ρ, h3 = hz = 1
38 CHAPTER 2. VECTOR ANALYSIS

2.8.6 Differential Operators in Curvilinear Coordinates


Gradient
Because the curvilinear coordinates are orthogonal, the gradient takes the same form as in
Cartesian coordinates, providing we use the differential displacements dri = hi dqi in the for-
mula. Thus, we have
∂ϕ ∂ϕ ∂ϕ
∇φ = ê1 + ê2 + ê3 . (2.199)
h1 ∂q1 h2 ∂q2 h3 ∂q3
Therefore the gradient operator is of the form

∂ ∂ ∂
∇= ê1 + ê2 + ê3 . (2.200)
h1 ∂q1 h2 ∂q2 h3 ∂q3

Divergence

Figure 2.28: Outflow of B1 in the q1 direction from a curvilinear volume element.

To derive the divergence in curvilinear coordinates, we apply Gauss’s theorem,


∫ I
(∇ · B) dτ = B · da, (2.201)
τ s

since ∇ · B is a scalar, we can equivalently to say,



B · da
∇·B= s ∫ , (2.202)
limdτ →0 τ dτ

if dτ is small enough over which the variation of ∇ · B is adiabatic. So first we calculate the
integral over a closed surface of a volume:
∫ ∫
ê1
B1 da1 = B1 h2 h3 dq2 dq3
∫ ∫
= − B1 (q1 − dq1 /2)h2 h3 dq2 dq3 + B1 (q1 + dq1 /2)h2 h3 dq2 dq3
∫ ( )
∂(B1 h2 h3 ) dq1
= − B1 (q1 )h2 h3 − dq2 dq3
∂q1 2
∫ ( )
∂(B1 h2 h3 ) dq1
+ B1 (q1 )h2 h3 + dq2 dq3
∂q1 2

∂(B1 h2 h3 )
= dq1 dq2 dq3 . (2.203)
∂q1
2.8. VECTOR ANALYSIS IN CURVED COORDINATES 39

Similarly,
∫ ∫
ê2
B2 da2 = B2 h1 h3 dq1 dq3
∫ ( )
∂(B2 h1 h3 ) dq2
= − B2 (q2 )h1 h3 − dq1 dq3
∂q2 2
∫ ( )
∂(B2 h1 h3 ) dq2
+ B2 (q2 )h1 h3 + dq1 dq3
∂q2 2

∂(B2 h1 h3 )
= dq1 dq2 dq3 . (2.204)
∂q2

and
∫ ∫
ê3
B3 da3 = B3 h1 h2 dq1 dq2
∫ ( )
∂(B3 h1 h2 ) dq3
= − B3 (q3 )h1 h2 − dq1 dq2
∂q3 2
∫ ( )
∂(B3 h1 h2 ) dq3
+ B3 (q3 )h1 h2 + dq1 dq2
∂q3 2

∂(B3 h1 h2 )
= dq1 dq2 dq3 . (2.205)
∂q3

So following Eq. (2.202), ∇ · B is given by


[∫ ∫ ∫ ]
∂(B1 h2 h3 ) ∂(B2 h1 h3 ) ∂(B3 h1 h2 )
∂q1
dq1 dq2 dq3 + ∂q2
dq1 dq2 dq3 + ∂q3
dq1 dq2 dq3
∇·B = ∫
limdτ →0
h1 h2 h3 dq1 dq2 dq3
[ ]
1 ∂(B1 h2 h3 ) ∂(B2 h1 h3 ) ∂(B3 h1 h2 )
= + + . (2.206)
h1 h2 h3 ∂q1 ∂q2 ∂q3

Laplacian

The Laplacian operator is still defined as the divergence of a gradient, and therefore ∇2 φ is
given by
[ ( ) ( ) ( )]
1 ∂ h2 h3 ∂φ ∂ h1 h3 ∂φ ∂ h1 h2 ∂φ
∇ φ=2
+ + . (2.207)
h1 h2 h3 ∂q1 h1 ∂q1 ∂q2 h2 ∂q2 ∂q3 h3 ∂q3

Curl

To see how the curl operator looks like, we apply Stoke’s theorem:
∫ I
(∇ × B) · da = B · dl. (2.208)
S C
40 CHAPTER 2. VECTOR ANALYSIS

Figure 2.29: Circulation around curvilinear element of area on a surface of constant q3 .

Consider a clockwise flow starting from point (q1 − dq1 /2, q2 − dq2 /2), the circulation around
the loop that follows the path 1 → 2 → 3 → 4 can be expressed as
∫ ∫
circulation = B1 (q2 − dq2 /2)h1 dq1 + B2 (q1 + dq1 /2)h2 dq2
1
∫ 2

+ B1 (q2 + dq2 /2)h1 dq1 + B2 (q1 − dq1 /2)h2 dq2 (2.209)
∫ [ 3
] 4
∫ [ ]
∂B1 ∂B1
≈ B1 (q2 ) − dq1 h1 dq1 − B1 (q2 ) + dq1 h1 dq1
∂q2 ∂q2
∫ [ ] ∫ [ ]
∂B2 ∂B2
+ B2 (q1 ) + dq1 h2 dq2 − B2 (q1 ) − dq1 h2 dq2
∂q1 ∂q1
[ ]
∂(B2 h2 ) ∂(B1 h1 )
= − dq1 dq2 (2.210)
∂q1 ∂q2
This implies the circulation per unit area on q1 q2 plane is ∇ × B|q3 . So, when we consider a
differential surface element in the curvilinear surface q3 = constant.
∫ ∫
(∇ × B) · da = ê3 · (∇ × B)h1 h2 dq1 dq2 (2.211)
S
I
= B · dl. (2.212)
C

Again, assume ∇ × B varying very slowly over the surface element, it can be evaluated by
H
B · dl
∇×B = C
(2.213)
limda→0 da
[ ]
1 ∂(B3 h3 ) ∂(B2 h2 )
∴ (∇ × B)1 = − ê1
h2 h3 ∂q2 ∂q3
[ ]
1 ∂(B1 h1 ) ∂(B3 h3 )
(∇ × B)2 = − ê2
h1 h3 ∂q3 ∂q1
[ ]
1 ∂(B2 h2 ) ∂(B1 h1 )
(∇ × B)3 = − ê3 . (2.214)
h1 h2 ∂q1 ∂q2

2.8.7 Differential Calculus In Cylindrical Coordinates


In cylindrical coordinates, the weight for each component is h1 = hρ = 1, h2 = hϕ = ρ, h3 =
hz = 1.
2.8. VECTOR ANALYSIS IN CURVED COORDINATES 41

• Gradient:

∂f ∂f ∂f
∇f = x̂ + ŷ + ẑ (2.215)
∂x ∂y ∂z

∂f ∂f ∂f
êρ · ∇f = |(∇f )ρ | = = = (2.216)
∂sρ hρ ∂ρ ∂ρ
∂f ∂f 1 ∂f
êϕ · ∇f = |(∇f )ϕ | = = = (2.217)
∂sϕ hϕ ∂ϕ ρ ∂ϕ
∂f ∂f ∂f
êz · ∇f = |(∇f )z | = = = . (2.218)
∂sz hz ∂z ∂z

∂f 1 ∂f ∂f
∇f = ρ̂ + ϕ̂ + ẑ. (2.219)
∂ρ ρ ∂ϕ ∂z
• Divergence:

( )
∂ ∂ ∂
∇ · v(r) = x̂ + ŷ + ẑ · (vx x̂ + vy ŷ + vz ẑ)
∂x ∂y ∂z
∂vx ∂vy ∂vz
= + + . (2.220)
∂x ∂y ∂z
In cylindrical coordinates, we follow Eq. (2.206) to obtain
[ ]
1 ∂ ∂ ∂
∇·v = (v1 h2 h3 ) + (v2 h1 h3 ) + (v3 h1 h2 ) (2.221)
h1 h2 h3 ∂q1 ∂q2 ∂q3
[ ]
1 ∂ ∂ ∂
= (ρvρ ) + (vϕ ) + (ρvz ) (2.222)
ρ ∂ρ ∂ϕ ∂z
1 ∂(ρvρ ) 1 ∂vϕ ∂vz
= + + . (2.223)
ρ ∂ρ ρ ∂ϕ ∂z

1 ∂(ρvρ ) 1 ∂vϕ ∂vz


∇·v = + + . (2.224)
ρ ∂ρ ρ ∂ϕ ∂z

• Laplacian:

∂ 2f ∂2f ∂ 2f
∇2 f = ∇ · (∇f ) = + + . (2.225)
∂x2 ∂y 2 ∂z 2
In cylindrical coordinates, we follow Eq. (2.207) to obtain
[ ( ) ( ) ( )]
1 ∂ ∂φ ∂ 1 ∂φ ∂ ∂φ
∇ φ=
2
ρ + + ρ . (2.226)
ρ ∂ρ ∂ρ ∂ϕ ρ ∂ϕ ∂z ∂z

( )
1 ∂ ∂φ 1 ∂2φ ∂2φ
∇ φ=
2
ρ + 2 2 + 2. (2.227)
ρ ∂ρ ∂ρ ρ ∂ϕ ∂z
42 CHAPTER 2. VECTOR ANALYSIS

• Curl:



x̂ ŷ ẑ

∇ × v = ∂
∂x

∂y

∂z

(2.228)

vx vy vz

In cylindrical coordinates, we follow Eq. (2.214) to obtain


[ ( )
1 ∂(v3 h3 ) ∂(v2 h2 )
∇×v = h1 − ê1
h1 h2 h3 ∂q2 ∂q3
( )
∂(v1 h1 ) ∂(v3 h3 )
+h2 − ê2
∂q3 ∂q1
( )]
∂(v2 h2 ) ∂(v1 h1 )
+h3 − ê3 (2.229)
∂q1 ∂q2


ê1 h1 ê2 h2 ê3 h3
1 ∂
= ∂ ∂ (2.230)
h1 h2 h3 ∂q1 ∂q2 ∂q3

h1 v1 h2 v2 h3 v3


ρ̂ ρϕ̂ ẑ
1 ∂ ∂ ∂
= ∂ρ ∂ϕ ∂z . (2.231)
ρ
vρ ρ vϕ vz

( ) ( ) ( )
1 ∂vz ∂(ρvϕ ) ∂vρ ∂vz 1 ∂(ρvϕ ) ∂vρ
∇×v = − êρ + − êϕ + − êz . (2.232)
ρ ∂ϕ ∂z ∂z ∂ρ ρ ∂ρ ∂ϕ

2.8.8 Differential Calculus In Spherical Coordinates


In spherical coordinates, the weight for each component is h1 = hr = 1, h2 = hθ = r, h3 =
hϕ = r sin θ.

• Gradient:

∂f ∂f ∂f
∇f = x̂ + ŷ + ẑ (2.233)
∂x ∂y ∂z

∂f ∂f ∂f
êr · ∇f = |(∇f )r | = = = (2.234)
∂sr hr ∂r ∂r
∂f ∂f ∂f
êθ · ∇f = |(∇f )θ | = = = (2.235)
∂sθ hθ ∂θ r∂θ
∂f ∂f ∂f
êϕ · ∇f = |(∇f )ϕ | = = = . (2.236)
∂sϕ hϕ ∂ϕ r sin θ∂ϕ
2.8. VECTOR ANALYSIS IN CURVED COORDINATES 43

∂f 1 ∂f 1 ∂f
∇f = r̂ + θ̂ + ϕ̂. (2.237)
∂r r ∂θ r sin θ ∂ϕ

• Divergence:

( )
∂ ∂ ∂
∇·v = x̂ + ŷ + ẑ · (vx x̂ + vy ŷ + vz ẑ)
∂x ∂y ∂z
∂vx ∂vy ∂vz
= + + (2.238)
∂x ∂y ∂z

In cylindrical coordinates, we follow Eq. (2.206) to obtain


[ ]
1 ∂ ∂ ∂
∇·v = (v1 h2 h3 ) + (v2 h1 h3 ) + (v3 h1 h2 )
h1 h2 h3 ∂q1 ∂q2 ∂q3
[ ]
1 ∂ 2 ∂ ∂
= 2 (vr r sin θ) + (vθ r sin θ) + (vϕ r)
r sin θ ∂r ∂θ ∂ϕ
1 ∂ 1 ∂ 1 ∂vϕ
= 2 (vr r2 ) + (vθ sin θ) + (2.239)
r ∂r r sin θ ∂θ r sin θ ∂ϕ

1 ∂ 1 ∂ 1 ∂vϕ
∇·v = 2
(vr r2 ) + (vθ sin θ) + . (2.240)
r ∂r r sin θ ∂θ r sin θ ∂ϕ

• Laplacian:

∂ 2f ∂ 2f ∂2f
∇2 f = ∇ · (∇f ) = + +
∂x2 ∂y 2 ∂z 2
( ) ( ) 1 ∂f
1 ∂ 2 ∂f 1 ∂ 1 ∂f 1 ∂( r sin θ ∂ϕ )
= 2 r + sin θ +
r ∂r ∂r r sin θ ∂θ r ∂θ r sin θ ∂ϕ
( ) ( ) 2
1 ∂ ∂f 1 ∂ ∂f 1 ∂ f
= 2 r2 + 2 sin θ + 2 2 . (2.241)
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2

( ) ( )
1 ∂ 2 ∂f 1 ∂ ∂f 1 ∂ 2f
∇f= 2
2
r + 2 sin θ + 2 2 . (2.242)
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2

• Curl:



x̂ ŷ ẑ

∇ × v = ∂
∂x

∂y

∂z

(2.243)

vx vy vz
44 CHAPTER 2. VECTOR ANALYSIS

In spherical coordinates, we follow Eq. (2.214) to obtain



ê1 h1 ê2 h2 ê3 h3
1 ∂ ∂ ∂

∇×v = (2.244)
h1 h2 h3 ∂q1 ∂q2 ∂q3

h1 v1 h2 v2 h3 v3


r̂ rθ̂ r sin θϕ̂
1 ∂ ∂ ∂
.
= 2 (2.245)
r sin θ ∂r ∂θ ∂ϕ

vr rvθ r sin θvϕ

Example 2.8.4. A rigid body is rotating about a fixed axis with a constant angular velocity ω.
Take ω to lie along the z-axis. Express the position vector r in circular cylindrical coordinates
and using circular cylindrical coordinates, (a) calculate v = ω × r, (b) calculate ∇ × v.
Consider a particle moving around a circle of radius R about an axis perpendicular to the
plane of motion. The velocity v = dr/dt = ṙ in circular motion can be expressed as

v=R = Rω = ω × R. (2.246)
dt

R
v

α r

Figure 2.30:

In cylindrical coordinates, r = ρ + z. Therefore,


d
v = = (ρ ρ̂ + z ẑ)
dt
= ρ̇ ρ̂ + ρϕ̇ ϕ̂ + ż ẑ, (2.247)
in which ρ̇ and ż must vanish for a rotational rigid body. Finally, the instantaneous velocity
reduces to
v = ρω ϕ̂. (2.248)

1 ∂(ρvϕ )
∇ × v = ∇ × (ρω ϕ̂) = ẑ
ρ ∂ρ
= 2ω ẑ. (2.249)
2.8. VECTOR ANALYSIS IN CURVED COORDINATES 45

Example 2.8.5. Solve Laplaces equation, ∇2 ψ = 0, in cylindrical coordinates for ψ = ψ(ρ).


( )
1 ∂ ∂ψ 1 ∂ 2ψ ∂ 2ψ
∇ψ =
2
ρ + 2 2 + 2
ρ ∂ρ ∂ρ ρ ∂ϕ ∂z
( )
1 ∂ ∂ψ
= ρ for ψ = ψ(ρ) (2.250)
ρ ∂ρ ∂ρ
If ∇2 ψ = 0 indicates that
∂ψ
ρ = c, (2.251)
∂ρ
as well as that ∫ ∫
ρ ψ(ρ)

c = dψ. (2.252)
a ρ ψ(a)
We may set ψ(a) = 0 at ρ = a such that
ρ
ψ(ρ) = c ln . (2.253)
a
Example 2.8.6. Find the circular cylindrical components of the velocity and acceleration of
a moving particle.
Since the position vector can be written as
r(t) = ρ(t) ρ̂ + z(t) ẑ = ρ(t)[cos ϕ x̂ + sin ϕ ŷ] + z(t) ẑ, (2.254)
the velocity is
dr(t)
= ρ̇(t)ρ̂ + ρ(t)ρ̂˙ + ż(t)ẑ
dt
= ρ̇(t)ρ̂ + ρ(t)ϕ̇(t)ϕ̂ + ż(t)ẑ
= vρ ρ̂ + vϕ ϕ̂ + vz ẑ, (2.255)
and the acceleration is
dv(t) d
= [vρ ρ̂ + vϕ ϕ̂ + vz ẑ]
dt dt
˙
= v̇ρ ρ̂ + vρ ρ̂˙ + v̇ϕ ϕ̂ + vϕ ϕ̂ + v̇z ẑ
= ρ̈ ρ̂ + 2ρ̇ϕ̇ ϕ̂ + ρϕ̈ϕ̂ − ρϕ̇2 ρ̂ + z̈ ẑ
= aρ ρ̂ + aϕ ϕ̂ + az ẑ. (2.256)
Example 2.8.7. Express ∂/∂x, ∂/∂y, ∂/∂z in spherical coordinates.
∂f ∂f ∂f ∂f 1 ∂f 1 ∂f
∵ x̂ + ŷ + ẑ = r̂ + θ̂ + ϕ̂ (2.257)
∂x ∂y ∂z ∂r r ∂θ r sin θ ∂ϕ
∂f
= (sin θ cos ϕx̂ + sin θ sin ϕŷ + cos θẑ)
∂r
1 ∂f
+ (cos θ cos ϕx̂ + cos θ sin ϕŷ − sin θẑ)
r ∂θ
1 ∂f
+ (− sin ϕx̂ + cos ϕŷ) (2.258)
r sin θ ∂ϕ
 ∂f     ∂f 
∂x sin θ cos ϕ 1r cos θ cos ϕ −r sin sin ϕ
θ ∂r
 ∂f     
∴  =  sin θ sin ϕ 1 cos θ sin ϕ cos ϕ  
 ∂y   r
∂f
r sin θ   ∂θ 
. (2.259)
∂f
∂z
cos θ − 1r sin θ 0 ∂f
∂ϕ
46 CHAPTER 2. VECTOR ANALYSIS

2.9 Review of Calculus


2.9.1 Vector Operation

Figure 2.31: http : //www.iap.tuwien.ac.at/www/surf ace/stmg allery/f ec rystallography

Figure 2.32:

2.9.2 Line Integrals


Vector calculus has important applications to curves and surfaces in physics and geometry. The
application of vector calculus to geometry is a field known as differential geometry. To construct
a curve, we may imagine a particle roaming not by its mind but by a rule r(t), in which r is the
positions vector and t is an arbitrary parameter. So the truth is that whenever t varies, r(t)
varies, bringing the particle to a new position. Then, during an interval a < t < b, the particle
depicts a trajectory, or a curve of its footprints. We call r(t) a parameter representation, given
by
r(t) = [x(t) î + y(t) ĵ + z(t) k̂]. (2.260)
2.9. REVIEW OF CALCULUS 47

Figure 2.33: Parametric representation of a curve

Example 2.9.1. Parametric Representation of a Circle


The circle x2 + y 2 = 4, z = 0, in the xy-plane can be represented by

r(t) = [2 cos t, 2 sin t, 0], (2.261)

in which 0 ≤ t ≤ 2π.

Figure 2.34: Right-handed circular helix

Example 2.9.2. Circular Helix


The right-handed screw curve C can be represented by the vector function

r(t) = [a cos t, a sin t, ct]. (2.262)

As we have known the parameter representation of a curve, let’s see how to apply it in the
line integral of a vector function. A line integral of a vector function over a curve C is defined
by ∫
F(r) · dr. (2.263)
C
48 CHAPTER 2. VECTOR ANALYSIS

Applying parameter representation and mapping the line integral to the parameter t-space, we
are equivalently dealing with the problem
∫ b
dr
F(r(t)) · dt. (2.264)
a dt
Of course you can recast the above equation into a expanded form:
∫ ∫ b
[Fx dx + Fy dy + Fz dz] = [Fx x′ (t) + Fy y ′ (t) + Fz z ′ (t)]dt. (2.265)
C a

Figure 2.35: Right-handed circular helix

Example 2.9.3. Evaluate the line integral of F(r) = −y î − xy ĵ along the curve C, from A to
B.
The parameter representation of C is given by
r(t) = cos t î + sin t ĵ, (2.266)
where 0 ≤ t ≤ π/2.
∫ π/2
F(r) · dr = (− sin t î − cos t sin t ĵ) · (− sin t î + cos t ĵ) dt
0
= π/4 − 1/3. (2.267)

Figure 2.36: Right-handed circular helix

Example 2.9.4. Evaluate the line integral of F(r) = [z, x, y] along the helix curve given by
r(t) = [cos t, sin t, 3t].
∫ 2π
F(r) · dr = (3t, cos t, sin t) · (− sin t, cos t, 3) dt
0
= 7π. (2.268)
2.9. REVIEW OF CALCULUS 49

2.9.3 Path Independence of Line Integrals

Figure 2.37:

Property 2.9.1. A line integral with continuous function F in a domain D in space is path
independent in D if and only if F is the gradient of some function G in D, ie.,

F = ∇G. (2.269)

So the line integral of F equals to


∫ ∫ ∫ b
F · dr = ∇G · dr = dG = G(b) − G(a). (2.270)
C C a

Eq. (2.270) implies two important results: firstly, the line integral of F depends only on initial
and final states and is path independent; and secondly, if C is a closed path, the line integral
of F vanishes. Physically, F is said to be a conservative field and G serves as a state function.
∫ 2,2,2
Example 2.9.5. For F = [2x, 2y, 2z], evaluate the integral 0,0,0 F · dr.
∫ 2,2,2
Since F = ∇G, where G = x2 + y 2 + z 2 , then 0,0,0 F · dr = G(2, 2, 2) − G(0, 0, 0) = 12.
Of course, you can evaluate the line integral directly by setting a curve C on which r(t) =
[t, t, t]. Then, for 0 ≤ t ≤ 2 we have
∫ 2 ∫ 2

F · r (t) dt = [2t, 2t, 2t] · [1, 1, 1] dt = 12. (2.271)
0 0
∫ 1,−1,7
Example 2.9.6. Evaluate the line integral I = 0,1,2 (3x2 dx + 2yz dy + y 2 dz).
What we need to do is to find a function G such that 3x2 = ∂G ∂x
, 2yz = ∂G
∂y
, and y 2 = ∂G
∂z
.
3 2
From Fx , we may set G(x, y, z) = x +h(y, z). Then from Fy , we may set h(y, z) = y z+k(z).
Finally, from Fz we should have k(z) = 0 such that ∂h(y, z)/∂z = y 2 . As a consequence, we
have G(x, y, z) = x3 + y 2 z. Since the integrand F can be the gradient of a function G, the line
integral of F would be path independent and would equal to G(1, −1, 7) − G(0, 1, 2) = 6.

Path Independence and Exactness of Differential Forms


Property 2.9.2. The relation between path independence and the exactness of differential
form is established by
F · dr = Fx dx + Fy dy + Fz dz. (2.272)
50 CHAPTER 2. VECTOR ANALYSIS

The form is called exact in a domain D if it is a differential of a function G, ie.,

F · dr = dG (2.273)
∂G ∂G ∂G
= dx + dy + dz
∂x ∂y ∂z
= ∇G · dr. (2.274)

Since F = ∇G, we know ∇ × F = 0, which means


∂Fz ∂Fy
= ;
∂y ∂z
∂Fx ∂Fz
= ;
∂z ∂x
∂Fy ∂Fx
= . (2.275)
∂x ∂y
If Eq. (2.275) holds in a simply connected region D, we say Eq. (2.272) is exact, and its line
integral must be path independent.
∫ 1,π/4,2
Example 2.9.7. Evaluate I = 0,0,1 [2xyz 2 dx + (x2 z 2 + z cos yz) dy + (2x2 yz + y cos yz) dz].

First, we check if the integrand is exact or not by


∂Fz ∂Fy
= 2x2 z + cos yz − yz sin yz = ;
∂y ∂z
∂Fx ∂Fz
= 4xyz = ;
∂z ∂x
∂Fy ∂Fx
= 2xz 2 = . (2.276)
∂x ∂y

Since it is exact, it can be the gradient of a function G. So we may set G(x, y, z) = x2 yz 2 +


h(y, z), and ∂h/∂y = z cos yz, indicating that h(y, z) = sin yz +k(z). From Fz we find k(z) = 0.
Eventually, G(x, y, z) = x2 yz 2 + sin yz. Therefore the line integral I equals to

I = G(1, π/4, 2) − G(0, 0, 1) = π + 1. (2.277)

Figure 2.38:
2.9. REVIEW OF CALCULUS 51

Example 2.9.8. Three components of a function F are


y x
Fx = − 2 , Fy = , Fz = 0.
x + y2 x2 + y 2
If
√ we want to calculate its line integral along a curve in the domain D defined as 1/2 <
x2 + y 2 < 3/2 as shown in Fig. 2.38, we will find something weird - if we apply previous
concepts, such as . So let’s see what’s wrong with that.
First, F · dr is exact because both Fx and Fy are not functions of z and
∂Fx y 2 − x2 ∂Fy
= 2 2 2
= . (2.278)
∂y (x + y ) ∂x
So the line integral of F along C in D would be
I I I
I= F · dr = ∇G · dr = dG = 0. (2.279)
C C C

YOU ARE DEADLY WRONG ! BECAUSE D IS NOT A SIMPLY CONNECTED REGION.


So you must realize that in the multiply-connected region the line integral of a function
along a CLOSED path would NOT vanish. Equivalently, the Stoke’s theorem
I ∫
F · dr = (∇ × F) · n̂ da (2.280)
C S

tells you that ∇ × F won’t be zero now. Instead, as in this geometry, F may Hplay a role of
superfluid velocity while it circulates around a singular point, a hole. And then C F · dr gives
the so-called winding number.
Now, let’s do the integration for this question.
I I I
−ydx + xdy
F · dr = Fx dx + Fy dy = . (2.281)
C C C x2 + y 2
For convenience, we may change to polar coordinates by setting x = a cos ϕ and y = a sin ϕ.
Thus the previous integration becomes
∫ 2π
I= dϕ = 2π. (2.282)
0

2.9.4 Double Integrals


2.9.5 Surfaces for Surface Integrals
Representation of Surfaces
A surface S in xyz-space can be represented by
z = f (x, y), or g(x, y, z) = 0. (2.283)
For example,
√ the upper plane of a hemisphere can be represented by x2 + y 2 + z 2 = a2 , z ≥ 0, or
z = a2 − x2 − y 2 . A surface can be thought of as the assemblage of curves {C} in space. The
introduction of a parameter r(t), where a ≤ t ≤ b, may help us describe a curve C by mapping
a ≤ t ≤ b located at t-axis onto C in xyz-space. So similarly we can introduce two parameters
{u, v} that spans a region R to map out the surface S in xyz-space by the parametric vector
r(u, v). A parametric representation of a surface S in space is of the form
r(u, v) = x(u, v)î + y(u, v)ĵ + z(u, v)k̂. (2.284)
52 CHAPTER 2. VECTOR ANALYSIS

Figure 2.39: Parametric representations of a curve and a surface

Figure 2.40:

Example 2.9.9. Parametric Representation of a Sphere

A sphere of radius a can be represented by


r(u, v) = a cos v cos u î + a cos v sin u ĵ + a sin v k̂, (2.285)
for 0 ≤ u ≤ 2π and −π/2 ≤ v ≤ π/2.
Example 2.9.10. Parametric Representation of a Cone
2.9. REVIEW OF CALCULUS 53

A cone z = x2 + y 2 , 0 ≤ z ≤ H can be represented by

r(u, v) = u cos v î + u sin v ĵ + uk̂, (2.286)

in which 0 ≤ u ≤ H and 0 ≤ v ≤ 2π.

Tangent Plane and Surface Normal

Figure 2.41: Tangent plane and normal vector

The tangent vectors of all the curves on a surface S through a point P of S form a plane,
called the tangent plane of S at P . Since S can be constructed from r(u, v), we can take u, v as
u(t), v(t), continuous functions of t thus differentiable t ot, from a curve C on S. In this case,
the position vector can be represented by

r(t) = r(u(t), v(t)), (2.287)

which means a tangent vector of C on S is given by

∂r ′ ∂r ′
r′ (t) = u (t) + v (t)
∂u ∂v
= ru u′ (t) + rv v ′ (t). (2.288)

So we know ru and rv are tangential to S at P . Since u and v spans a 2d-space, we may assume
ru and rv are linearly independent, ie., ru and rv may also spans a 2d-space termed by ru × rv
with the normal vector n̂ defined by

N ru × rv
n̂ = = . (2.289)
|N| |ru × rv |

Reminder : As we talked about the geometric meaning of the gradient vector of a function
f (x, y, z), we said that gradient is as the normal vector of the surface f (x, y, z) = c at point P .
To prove it we find a tangent vector r of C at P . Of course we can also another tangent vector
of C ′ that is also passing P . Then all these tangent vectors form a tangent plane on S around
P . So we would have
∂f ∂f ∂f
df = dx + dy + dz ≡ ∇f · dr = 0, (2.290)
∂x ∂y ∂z
54 CHAPTER 2. VECTOR ANALYSIS

Figure 2.42:

since we are still on the surface S, which immediately shows that ∇f is normal to the surface
at P since dr lies on S as well. So we may conclude that

∇f
n̂ = . (2.291)
|∇f |

Example 2.9.11. Unit Normal Vector of a Cone



The unit normal vector of a cone f (x, y, z) = −z + x2 + y 2 = 0 except at the apex is
given by [ ]
∇f 1 x y
n̂ = =√ √ î + √ ĵ − k̂ . (2.292)
|∇f | 2 x2 + y 2 x2 + y 2

Surface Integrals
To define a surface integral, we take a surface S, given by a parametric representation as just
discussed,
r(u, v) = x(u, v)î + y(u, v)ĵ + z(u, v)k̂. (2.293)
The surface is assumed piecewise smooth having the normal vector N = ru × rv at any point.
For an arbitrary vector function f , the surface integral over S can be defined as
∫ ∫
f · n̂ da ≡ f (r(u), r(v)) · N dudv. (2.294)
S R

Then we must have


n̂ da = n̂|N| dudv, (2.295)
ie., the surface element da = |N| dudv.
Since f · n̂ indicates
∫ the projection of f on n̂, the normal component of f , it serves as normal
flux across S, S ρv · n̂ da. Then the surface integral becomes the flux integral, counting the
number of particles across the surface per unit time.

Example 2.9.12. Compute the flux of water through the parabolic cylinder S: y = x2 ,
0 ≤ x ≤ 2, 0 ≤ z ≤ 3, if velocity v = [3z 2 , 6, 6xz] m/s and density ρ = 1g/cm3 .
Set u = x and v = z, then y = u2 , so the parameter presentation r = [u, u2 , v] and the
normal vector
N = ru × rv = [1, 2u, 0] × [0, 0, 1] = [2u, −1, 0]. (2.296)
2.9. REVIEW OF CALCULUS 55

Figure 2.43:

The surface integral is evaluated by


∫ ∫ 3 ∫ 2
v · N dudv = (6uv 2 − 6) dudv = 72 m3 /s. (2.297)
S 0 0

The question can also be solved easily if you know the unit normal vector must be along
[x̂, −ŷ, 0]. Then the surface integral can be represented by
∫ ∫ ∫ ∫
f · n̂ da ≡ fx dydz + fy dxdz + fz dxdy (2.298)
S S S S
∫ 3∫ 4 ∫ 3∫ 2
= 3z dydz − 6
2
dxdz = 72. (2.299)
0 0 0 0

Figure 2.44:

Example 2.9.13. Evaluate the surface integral of f = [x2 , 0, 3y 2 ] over the surface of x+y+z = 1
cut in the first octant.
We may set x = u and y = v, then z = 1 − u − v. So the parameter representation
r = [u, v, 1 − u − v]. Then the normal vector is given by
N = [1, 0, −1] × [0, 1, −1] = [1, 1, 1]. (2.300)
So the surface integral is
∫ ∫ 1 ∫ 1−v
1
f · N dudv = (u2 + 3v 2 ) dudv = . (2.301)
S 0 0 3
56 CHAPTER 2. VECTOR ANALYSIS

We can verify the Gauss’s theorem in this question by evaluating surface integral of each
surface and the volume integral. Throughout the xz-plane, n̂ = ŷ. However fy = 0 and
therefore the surface integral is zero. Throughout the yz-plane, n̂ = x̂. However fx = 0 at
x = 0 and therefore the surface integral is also zero. Throughout the xy-plane, n̂ = −ẑ. The
surface integral is given by
∫ ∫ 1 ∫ 1−y
− 3y dxdy = −
2
3y 2 dxdy = −(y 3 − 3y 4 /4)|10 = −1/4. (2.302)
S 0 0

So the total surface integral of the prism is 1/12. Now let’s evaluate the volume integral by
calculating the divergence of f first, which gives us 2x. Then,
∫ 1 ∫ 1−x ∫ 1−x−y
2x dxdydz = 1/12. (2.303)
0 0 0

The volume integral of ∇ · f do equal to the surface integral of f .

Figure 2.45: Möbius strip

From above examples we see that the value of the integral depends on the choice of the unit
normal vector n̂. We express this by saying that such an integral is an integral over an oriented
surface S, that is, over a surface S on which we have chosen one of the two possible unit normal
vectors in a continuous fashion. However, not a sufficiently small piece of a smooth surface is
always orientable. A well-known example is the Möbius strip, upon which one may travel from
P0 in the inner ring to the the outer ring when he backs to P0 .
Because of these ’nonorientable’ cases, we need to develop surface integrals without regard
to orientation, which is expressed by
∫ ∫
G(r) dA = G(r(u, v))|N(u, v)| dudv. (2.304)
S R

Here dA = |N(u, v)| dudv = |ru × rv | dudv is the surface element of S. If G(r) is the mass
density of S, then Eq. (2.304) is the total mass of S. If G(r) = 1 then Eq. (2.304) gives the
area A(S) of S, ∫ ∫
A(S) = dA = |N(u, v)| dudv. (2.305)
S R
The formula may have the evaluation of surface of a sphere or a donut, for example, to be
easier.

Example 2.9.14. The Surface Area of A Sphere


2.9. REVIEW OF CALCULUS 57

For a sphere of radius a, the parameter representation is r(u, v) = [a sin u cos v, a sin u sin v, a cos u],
0 ≤ u ≤ π and 0 ≤ v ≤ 2π. The normal vector is then
N(u, v) = ru × rv = [a cos u cos v, a cos u sin v, −a sin u] × [−a sin u sin v, a sin u cos v, 0]
= [a2 sin2 u cos v, a2 sin2 u sin v, a2 sin u cos u], (2.306)
and N(u, v) = a2 | sin u|. Therefore, the surface area is given by
∫ π ∫ 2π ∫ π
A(S) = a | sin u| dudv = 2πa
2 2
sin u du = 4πa2 . (2.307)
0 0 0

Example 2.9.15. Moment of Inertia of A Surface

Find the moment of inertia I of a spherical shell S: x2 + y 2 + z 2 = a2 of constant mass


density and total mass M about the z-axis.
By definition, the moment of inertia is

I = ρr2 dV, (2.308)

in which r is the distance from a tiny mass element to the rotation axis of reference. Now,
the distance from the shell to z-axis is a sin u. From the former example we know dA =
|N(u, v)| dudv = a2 | sin u| dudv. Therefore,
∫ 2π ∫ π
8π 4
I=ρ a4 sin3 u dudv = ρa . (2.309)
0 0 3

Since the mass of the shell is M = 4πρa2 , I = 3
M a2 .

Figure 2.46:

Finally, we show the surface integral over S can equivalently be mapped to integral over a
region R∗ on xy-plane. If a surface S is represented by z = f (x, y), then setting u = x, v = y,
we have parameter representation r = [u, v, f ]. The normal vector
N(u, v) = ru × rv = [1, 0, fu ] × [0, 1, fv ] = [−fu , −fv , 1]. (2.310)
The surface integral of a function G(r) over S can be expressed as
∫ ∫
G(r)dA = G(x, y, f (x, y))|[−fx , −fy , 1]| dxdy
R∗
S
√ ( )2 ( )2

∂f ∂f
= G(x, y, f (x, y)) 1 + + dxdy. (2.311)
R∗ ∂x ∂y
58 CHAPTER 2. VECTOR ANALYSIS

Example 2.9.16. Evaluate the surface integral of G = 3xy over a surface S : z = xy, 0 ≤ x ≤ 1,
0 ≤ y ≤ 1.
Applying Eq. (2.311) we have ∂f /∂x = y and ∂f /∂y = x.
∫ ∫ 1∫ 1 √
G dA = 3xy 1 + x2 + y 2 dxdy
S
∫0 1 0
1
= y(1 + x2 + y 2 )3/2 0 dy
0
1
= [(2 + y 2 )5/2 − (1 + y 2 )5/2 ]|10 = 1.055. (2.312)
5

Vous aimerez peut-être aussi