Vous êtes sur la page 1sur 14

Chapter 16

Curvilinear Vector
Analysis
All the vector analytical quantities discussed in the previous chapters can
also be calculated in other coordinate systems. The general procedure is to
start with denitions of quantities in a coordinate-free way and substitute the
known quantities in terms of the particular coordinates we are interested in
and read o the vector analytic quantity. Instead of treating cylindrical and
spherical coordinate systems separately, we lump them together and derive re-
lations that hold not only in the three familiar coordinate systems, but also in
all coordinate systems whose unit vectors form a set of right-handed mutually
perpendicular vectors. Since the geometric denitions of all vector-analytic
quantities involve elements of length, we start with the length elements.
16.1 Elements of Length
Consider curvilinear coordinates
1
(q
1
, q
2
, q
3
) in which the primary line curvilinear
coordinates
elements are given by
dl
1
= h
1
(q
1
, q
2
, q
3
) dq
1
, dl
2
= h
2
(q
1
, q
2
, q
3
) dq
2
, dl
3
= h
3
(q
1
, q
2
, q
3
) dq
3
,
where h
1
, h
2
, and h
3
are some functions of coordinates. By examining the
primary line elements in Cartesian, spherical, and cylindrical coordinates, we
can come up with Table 16.1.
Denoting the unit vectors in curvilinear coordinate systems by e
1
, e
2
, and
e
3
, we can combine all the equations for the elements of length and write
them as a single vector equation:
dr = d

l = e
1
dl
1
+ e
2
dl
2
+ e
3
dl
3
= e
1
h
1
dq
1
+ e
2
h
2
dq
2
+ e
3
h
3
dq
3
. (16.1)
1
As will be seen shortly, Cartesian coordinates are also included in such curvilinear
coordinates. The former have lines (and planes) as their primary lengths and surfaces, thus
the word linear in the name of the latter.
424 Curvilinear Vector Analysis
Curvilinear Cartesian Spherical Cylindrical
q
1
x r
q
2
y
q
3
z z
h
1
1 1 1
h
2
1 r
h
3
1 r sin 1
Table 16.1: The specications of the three coordinate systems in terms of curvilinear
coordinates.
This equation is useful in its own right. For example, we can obtain the curvi-
linear unit vectors as follows. Rewrite Equation (16.1) in terms of increments:
r e
1
h
1
q
1
+ e
2
h
2
q
2
+ e
3
h
3
q
3
.
Keeping q
2
and q
3
constant (so that q
2
= 0 = q
3
), divide both sides by
q
1
to obtain
r
q
1
e
1
h
1
.
In the limit, the LHS becomes a partial derivative and we get
e
1
=
1
h
1
r
q
1
. (16.2)
The other two unit vectors can be obtained similarly. We thus have
Box 16.1.1. The ith unit vector of a curvilinear coordinate system is
given by
e
i
=
1
h
i
r
q
i
, i = 1, 2, 3. (16.3)
This is a useful formula for obtaining the Cartesian components of curvilinear
unit vectors, when the Cartesian components of the position vector are given
in terms of curvilinear coordinates.
Example 16.1.1. As an illustration of the above procedure, we calculate the unit
vectors in spherical coordinates. First we write
r = x e
x
+ y e
y
+ z e
z
= e
x
r sin cos + e
y
r sin sin + e
z
r cos .
Now we dierentiate with respect to r to get
e
1
e
r
=
r
r
= e
x
sin cos + e
y
sin sin + e
z
cos .
16.2 The Gradient 425
Similarly,
e
2
e

=
1
r
r

= e
x
cos cos + e
y
cos sin e
z
sin ,
e
3
e

=
1
r sin
r

= e
x
sin + e
y
cos ,
where we have used Table 16.1. These are the results we obtained in Chapter 1 from
purely geometric arguments.

We are now in a position to nd the gradient, divergence, and curl of
a vector eld in general curvilinear coordinates. Once these are obtained,
nding their specic forms in cylindrical and spherical coordinates entails
simply substituting the appropriate expressions for q
1
, q
2
, and q
3
and h
1
, h
2
,
and h
3
.
16.2 The Gradient
The gradient is found by equating
df =
f
q
1
dq
1
+
f
q
2
dq
2
+
f
q
3
dq
3
to the dierential of f in terms of the gradient:
df = f dr = (f)
1
h
1
dq
1
+ (f)
2
h
2
dq
2
+ (f)
3
h
3
dq
3
.
The last two equations yield
(f)
1
h
1
=
f
q
1
, (f)
2
h
2
=
f
q
2
, (f)
3
h
3
=
f
q
3
,
which gives gradient in
curvilinear
coordinates
Box 16.2.1. The gradient of a function f in a curvilinear coordinate
system is given by
f = e
1
1
h
1
f
q
1
+ e
2
1
h
2
f
q
2
+ e
3
1
h
3
f
q
3
. (16.4)
This result, in conjunction with Table 16.1, agrees with the expression ob-
tained for the gradient in the Cartesian coordinate system. In cylindrical
coordinates, we obtain
f = e

+ e

+ e
z
f
z
, (16.5)
426 Curvilinear Vector Analysis
so that the operator in cylindrical coordinates is given by
= e

+ e

+ e
z

z
. (16.6)
Similarly, in spherical coordinates, we get gradient of a
function in
spherical
coordinates
f = e
r
f
r
+ e

1
r
f

+ e

1
r sin
f

(16.7)
with the operator given by
= e
r

r
+ e

1
r

+ e

1
r sin

. (16.8)
Example 16.2.1. The electrostatic potential of an electric dipole was given in
Example 10.5.1 in spherical coordinates. With the expression for the gradient given
above, we can nd the electric eld E = of a dipole in spherical coordinates:
E
r
=

dip
r
=

r

k
e
p cos
r
2

=
2k
e
p cos
r
3
,
E

=
1
r

dip

=
1
r

k
e
p cos
r
2

=
k
e
p sin
r
3
,
E

=
1
r sin

dip

=
1
r sin

k
e
p cos
r
2

= 0.
Summarizing, we have electric eld of an
electric dipole
E
dip
=
k
e
p
r
3
(2 e
r
cos + e

sin ). (16.9)
This is the characteristic eld of a dipole.

Example 16.2.2. Just as electric charges can produce electric dipoles, electric
currents can produce magnetic dipoles. We saw this in Subsection 15.2. In this
example, we will calculate the magnetic eld of a dipole directly. Consider the
magnetic eld of a circular loop of current as given in Equations (4.24) and (4.26).
We change the coordinates of the eld point P to spherical and assume that P is far
away from the loop, i.e., that a is very small compared to r. Writing r
2
for
2
+ z
2
and r sin for , we expand the integrands of (4.24) and (4.26) in powers of a/r
keeping only the rst nonzero power. Thus,
1
(r
2
+a
2
2ra sin cos t)
3/2
=
1
r
3

1 +

a
r

2
2

a
r

sin cos t

3/2
=
1
r
3

1 + 3

a
r

sin cos t

+ ,
r sin cos ta
(r
2
+a
2
2ra sin cos t)
3/2
=
1
r
2

sin cos t
a
r

1 +

a
r

2
2

a
r

sin cos t

3/2
=
1
r
2

sin cos t
a
r

1 + 3

a
r

sin cos t

+
=
1
r
2

sin cos t
a
r
+
3a
r
sin
2
cos
2
t

.
16.3 The Divergence 427
Substituting these in the integrals of (4.24) and (4.26) yields
B

=
k
m
Iaz
r
3

2
0
cos t

1 +
3a
r
sin cos t

dt =
3k
m
Ia
2
cos sin
r
3
,
where we substituted r cos for z. In an analogous way, we also obtain
B
z
=
k
m
Ia
r
2

2
0

sin cos t
a
r
+
3a
r
sin
2
cos
2
t

dt
=
k
m
Ia
r
2

2a
r
+
3a
r
sin
2

.
We are interested in the spherical components of the magnetic eld. To nd
these components, we rst write
B = B

+ B
z
e
z
and take the dot product with appropriate unit vectors:
B
r
= B e
r
= B

e
r
+ B
z
e
z
e
r
= B

sin + B
z
cos
=
3k
m
Ia
2
cos sin
r
3
sin +
k
m
Ia
r
2

2a
r

3a
r
sin
2

cos
=
2k
m
Ia
2
r
3
cos .
Similarly,
B

= B e

= B

+ B
z
e
z
e

= B

cos B
z
sin
=
3k
m
Ia
2
cos sin
r
3
cos
k
m
Ia
r
2

2a
r

3a
r
sin
2

sin
=
k
m
Ia
2
r
3
sin .
Summarizing, we write magnetic eld of a
magnetic dipole
B =
k
m
Ia
2
r
3
(2 e
r
cos + e

sin ). (16.10)
This has a striking resemblance to Equation (16.9). In fact once we identify Ia
2
as the magnetic dipole of the loop, and change all magnetic labels to electric ones,
we recover Equation (16.9).

16.3 The Divergence


To nd the divergence of a vector A, we consider the volume element of
Figure 16.1 and nd the outward ux through the sides of the volume. For
the front face we have

f
= A
f
e
1
a
f
,
428 Curvilinear Vector Analysis
A
A
1

e
1
A
2

e 2
A
3

e
3
P(q
1
, q
2
, q
3
)

l
1
l
3
l
2
Figure 16.1: Point P and the surrounding volume element in curvilinear coordinates.
Note that the midpoints of the front and back faces are q
1
/2 away from P in the
positive and negative e
1
directions, respectively. Similarly for the other four faces.
where A
f
means the value of A at the center of the front face and a
f
is the
area of the front face. Following the arguments presented for the Cartesian
case, we write

f
A
f
e
1
a
f
= A
1f
l
2f
l
3f
= A
1f
(h
2
q
2
)
f
(h
3
q
3
)
f
= A
1f
h
2f
h
3f
q
2
q
3
The subscript 1 in A
1f
, for example, means component of A in the direction
of the rst coordinate. The subscript f implies evaluationat the midpoint
on the front side whose second and third coordinates are the same as P, and
whose rst coordinate is q
1
+ q
1
/2. Thus, we have

f
A
1

q
1
+
q
1
2
, q
2
, q
3

h
2

q
1
+
q
1
2
, q
2
, q
3

h
3

q
1
+
q
1
2
, q
2
, q
3

q
2
q
3
because, unlike the Cartesian case, h
1
, h
2
, and h
3
are functions of the co-
ordinates. Using Taylor series expansion for the functions A
1
, h
2
, and h
3
yields

A
1
(q
1
, q
2
, q
3
) +
q
1
2
A
1
q
1

h
2
(q
1
, q
2
, q
3
) +
q
1
2
h
2
q
1

h
3
(q
1
, q
2
, q
3
) +
q
1
2
h
3
q
1

q
2
q
3
.
16.3 The Divergence 429
Multiplying out and keeping terms up to the third order (corresponding to
the order of a volume element by which we shall divide shortly), we obtain

A
1
h
2
h
3
+ A
1
h
2
h
3
q
1
+ A
1
h
3
h
2
q
1
+ h
2
h
3
A
1
q
1

q
1
2
q
2
q
3
=

A
1
h
2
h
3
+

q
1
(h
2
h
3
A
1
)

q
1
2
q
2
q
3
,
where we left out the explicit dependence of the functions on their independent
coordinate variables. For the back face we have

b
A
b
( e
1
a
b
) = A
1b
l
2b
l
3b
= A
1b
(h
2
q
2
)
b
(h
3
q
3
)
b
= A
1

q
1

q
1
2
, q
2
, q
3

h
2

q
1

q
1
2
, q
2
, q
3

h
3

q
1

q
1
2
, q
2
, q
3

q
2
q
3
.
Taylor expanding the three functions A
1
, h
2
, and h
3
as above, and multiplying
out yields

A
1
h
2
h
3


q
1
(h
2
h
3
A
1
)

q
1
2
q
2
q
3
.
Adding the front and back contributions, we obtain

1

f
+
b


q
1
(h
2
h
3
A
1
) q
1
q
2
q
3
.
Similarly, the uxes through the faces perpendicular to e
2
and e
3
are

2


q
2
(h
1
h
3
A
2
) q
1
q
2
q
3
,

3


q
3
(h
1
h
2
A
3
) q
1
q
2
q
3
. (16.11)
Adding the three contributions and dividing by the volume
V = l
1
l
2
l
3
= h
1
h
2
h
3
q
1
q
2
q
3
and nally taking the limit of smaller and smaller volumeswhich turns all
approximations into equalitieswe get divergence in
curvilinear
coordinates
Theorem 16.3.1. The divergence of a vector eld A in a curvilinear coordi-
nate system is given by
A =
1
h
1
h
2
h
3


q
1
(h
2
h
3
A
1
) +

q
2
(h
1
h
3
A
2
) +

q
3
(h
1
h
2
A
3
)

.
430 Curvilinear Vector Analysis
Now that we have a general formula for the divergence, we can use Table
16.1 to write the divergence in a specic coordinate system. For instance,
substituting the entries of the second column gives the formula in Theorem
13.2.1, and the third column yields divergence of a
vector eld in
spherical
coordinates
A =
1
r
2
sin

r
2
sin A
r

(r sinA

) +

(rA

=
1
r
2

r
2
A
r

+
1
r sin

(sinA

) +
A

. (16.12)
To obtain the divergence in cylindrical coordinates, we use the last column
and obtain
A =
1

(A

) +

(A

) +

z
(A
z
)

=
1

(A

) +
1

+
A
z
z
. (16.13)
Example 16.3.2. Consider the vector eld dened by
A = kr

e
r
,
where k and are constants. Let us verify the divergence theorem for a spherical
surface of radius R (see Figure 16.2). The total ux is obtained by integrating over
the surface of the sphere:
=

S
A da =

S
kR

e
r
e
n
R
2
sin d d
= kR
+2

S
sin d d = 4kR
+2
.
d
d

e
n =

e
r
R
x
y
z
Figure 16.2: The element of area and its unit normal for a sphere.
16.4 The Curl 431
On the other hand, using the expression for divergence in the spherical coordinate
system and noting that A

= 0 = A

, we obtain
A =
1
r
2

r
(r
2
A
r
) =
1
r
2
d
dr

kr
+2

= ( + 2)kr
1
,
where we have assumed that = 2. Therefore,

V

AdV =

R
0
( + 2)kr
1
r
2
dr


0
sin d

2
0
d = 4kR
+2
which agrees with the surface integration.
For = 2 the divergence appears to vanish everywhere. However, a closer
examination reveals that the statement is true only if r = 0. In fact, as we discussed
before, the divergence of A is proportional to the Dirac delta function, (r) in this
case [see Equation (15.2)].

16.4 The Curl
To calculate the curl, we choose a closed path perpendicular to one of the unit
vectors, say e
1
and calculate the line integral of A around it. The situation is
depicted in Figure 16.3. We calculate the contribution to the line integral from
path (1) in detail and leave calculation of contributions from the remaining
three paths to the reader. In all calculations, terms of higher order than the
second will be omitted

(1)
A dr A
l
r
l
= A
l
( e
3
l
l
) = A
3l
l
l
= A
3l
h
3l
q
3
= A
3

q
1
, q
2

q
2
2
, q
3

h
3

q
1
, q
2

q
2
2
, q
3

q
3
=

A
3

q
2
2
A
3
q
2

h
3

q
2
2
h
3
q
2

q
3
A
3
h
3
q
3
+

q
2
(h
3
A
3
)
q
2
2
q
3
.
P(q
1
, q
2
, q
3
)
(1)
(2)
(3)
(4)
h
2 b
dq
2
h
2 t
dq
2
h
3l
dq
3
h
3r
dq
3

e
2

e
3

e
1
Figure 16.3: Path of integration for the rst component of the curl of A in curvilinear
coordinates.
432 Curvilinear Vector Analysis
Following similar steps, the reader may check that

(2)
A dr A
3
h
3
q
3
+

q
2
(h
3
A
3
)
q
2
2
q
3
,

(3)
A dr A
2
h
2
q
2


q
3
(h
2
A
2
)
q
3
2
q
2
, (16.14)

(4)
A dr A
2
h
2
q
2


q
3
(h
2
A
2
)
q
3
2
q
2
.
Summing up all these contributions, we obtain

A dr


q
2
(h
3
A
3
)

q
3
(h
2
A
2
)

q
2
q
3
.
Dividing this by the area enclosed by the path
a = l
2
l
3
= h
2
h
3
q
2
q
3
we obtain the rst component, the component along the unit normal to the
area:
(A)
1
=
1
h
2
h
3


q
2
(h
3
A
3
)

q
3
(h
2
A
2
)

.
Corresponding expressions for the other two components of the curl can
be found by proceeding as above. We can put all of the components together
in a mnemonic determinant form: curl in curvilinear
coordinates
Theorem 16.4.1. The curl of a vector eld A in a curvilinear coordinate
system is given by
A =
1
h
1
h
2
h
3

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

q
1

q
2

q
3
h
1
A
1
h
2
A
2
h
3
A
3

. (16.15)
Note that A is not a cross product (except in Cartesian coordinates), warning! A is
not a cross
product in general
curvilinear
coordinates!
but a vector dened by the determinant on the RHS of (16.15).
If we substitute the appropriate values for hs and qs in spherical coordi-
nates, we obtain
A =
1
r
2
sin

e
r
e

r e

r sin

A
r
rA

r sin A

. (16.16)
16.4 The Curl 433
In cylindrical coordinates we get
A =
1

e
z

z
A

A
z

. (16.17)
Example 16.4.2. We have already calculated the magnetic eld of a dipole in
Example 16.2.2. Here we want to obtain the same result using the vector potential
of a dipole given in Equation (15.12). We take to be along the z-axis. Then
= e
z
= ( e
r
cos e

sin )
and
e
r
= (sin e

e
r
) = sin e

.
Therefore,
B = A =

k
m
e
r
r
2

k
m
sin e

r
2

=
k
m

r
2
sin

e
r
e

r e

r sin

0 0 r sin
sin
r
2

=
k
m

r
2
sin

e
r
e

r e

r sin

0 0
sin
2

=
k
m

r
2
sin

e
r

2 sin cos
r

r e

sin
2

r
2

=
k
m

r
3
(2 cos e
r
+ sin e

),
which is the expression obtained in Example 16.2.2.

Example 16.4.3. Consider the vector eld B described in cylindrical coordinates
as
B =
k

,
where k is a constant. The curl of B is easily found to be zero:
B =
1

e
z

z
0 (k/) 0

= 0.
However, for any circle (of radius a, for example) centered at the origin and located
in the xy-plane, we get
2

C
B dr =

2
0
k
a
e

( e

a d) = 2k = 0.
2
See also Example 14.3.3 which discusses this same vector eld in Cartesian coordinates.
434 Curvilinear Vector Analysis
The reason for this result is that the circle is not contractible to zero: At the
originwhich is inside the circle and at which = 0B is not dened.
This vector eld should look familiar. It is the magnetic eld due to a long
straight wire carrying a current along the z-axis. According to Amp`eres circuital
law, the line integral of B along any closed curve encircling the wire, such as the
above circle, gives, up to a multiplicative constant, the current in the wire, and this
current is not zero.
Example 16.4.4. A vector eld that can be written as central force elds
are conservative
F = f(r)r,
where r is the displacement vector from the origin, is conservative. It is instructive
to show this using both Cartesian and spherical coordinate systems.
First, in Cartesian coordinates
F = xf(r) e
x
+ yf(r) e
y
+ zf(r) e
z
and the curl is
F =

e
x
e
y
e
z

z
xf yf zf

= e
x


y
(zf)

z
(yf)

+ e
y


z
(xf)

x
(zf)

+ e
z


x
(yf)

y
(xf)

.
Concentrating on the x-component rst and using the chain rule, we have

y
(zf) = z
f
y
= z
df
dr
r
y
= zf

r
y
.
But
r
y
=

y

x
2
+ y
2
+ z
2
=
y
r
.
Thus,

y
(zf) = yzf

.
Similarly,

z
(yf) = yzf

.
Therefore, the x-component of F is zero. The y- and z-components can also be
shown to be zero, and we get F = 0.
On the other hand, using spherical coordinates, we easily obtain
F =
1
r
2
sin

e
r
e

r e

r sin

rf(r) 0 0

= 0.
Obviously, the use of spherical coordinates simplies the calculation consi-
derably.
16.4 The Curl 435
The preceding example shows that
Box 16.4.1. Any well-behaved vector eld whose magnitude is only a
function of radial distance, r, and whose direction is along r is conserva-
tive. Such vector elds are generally known as central vector elds.
16.4.1 The Laplacian
Combining divergence and the gradient gives the Laplacian. Using Equation
(16.4) in Theorem 16.3.1, we get Laplacian in
curvilinear
coordinates
Theorem 16.4.5. The Laplacian of a function f is the divergence of gradient
of f andin a curvilinear coordinate systemis given by

2
f =
1
h
1
h
2
h
3


q
1

h
2
h
3
h
1
f
q
1

+

q
2

h
1
h
3
h
2
f
q
2

+

q
3

h
1
h
2
h
3
f
q
3

.
For cylindrical coordinates the Laplacian is

2
f =
1

+
1

2
f

2
+

2
f
z
2
(16.18)
and for spherical coordinates it is

2
f =
1
r
2

r
2
f
r

+
1
r
2
sin

sin
f

+
1
sin

2
f

. (16.19)
Equations (16.7) and (16.19) allow us to write the angular momentum
dierential operator derived in Example 15.3.1 in spherical coordinates, which
is the most common way of writing it. We note that

r
2
f
r

= 2r
f
r
+ r
2

2
f
r
2
,
and
r (f) = r
f
r
,
and
(r )
2
f = r

r

f
r

= r
f
r
+ r
2

2
f
r
2
.
Substituting these plus (16.19) in (15.22) yields
L
2
f =
1
sin

sin
f

+
1
sin

2
f

. (16.20)
Therefore, the angular momentum operator depends only on angles in spher-
ical coordinates.
436 Curvilinear Vector Analysis
16.5 Problems
16.1. The divergence of a vector can be obtained in any coordinate system
by brute force calculation. In this problem you are asked to nd A in
cylindrical coordinates.
(a) Express A
x
in terms of cylindrical coordinates and components. Hint:
Write A in cylindrical ccordinates and take the dot product with e
x
expressing
everything in terms of cylindrical ccordinates.
(b) Use the chain rule
A
x
x
=
A
x

x
+
A
x

x
+
A
x
z
z
x
where A
x
is what you found in (a).
(c) Do the same with A
y
and A
z
, and add the three terms to obtain the
divergence in cylindrical coordinates.
16.2. Find the divergence of a vector in spherical coordinates following the
procedure outlined in Problem 16.1.
16.3. Find the gradient of a function in cylindrical and spherical coordinates
following a procedure similar to the one outlined in Problem 16.1.
16.4. Find the curl of a vector in cylindrical and spherical coordinates fol-
lowing a procedure similar to the one outlined in Problem 16.1.
16.5. Start with the Laplacian in Cartesian coordinates.
(a) By using the chain rule and expressing the second derivatives in cylindrical
coordinates, nd the Laplacian in cylindrical coordinates.
(b) Do the same for spherical coordinates.
16.6. The elliptic cylindrical coordinates (u, , z)are given by
x = a coshu cos
y = a sinh u sin
z = z
where a is a constant.
(a) What is the expression for the gradient of a function f in elliptic cylindri-
cal coordinates?
(b) What is the expression for the divergence of a vector A in elliptic cylin-
drical coordinates?
(c) What is the expression for the curl of a vector A in elliptic cylindrical
coordinates?
(d) What is the expression for the Laplacian of a function f in elliptic cylin-
drical coordinates?