3.

3 Vector Fields: An Introduction

Let X be an open subset of R
n
. A Vector Field (VF) F on X is a mapping
from X into R
n
. I.e.,
F : X R
n
.
We also call n the dimension of the VF F. Usually n is 3 or 2, but it can be
1 or even 4,5,6,. . . . Why n = 3 is the most meaningful? The answer is
simple: It seems that our nature is 3-dimensional. So it seems that the most
meaningful (or the most applicable (to our physical world)) mathematical
entities are necessarily 3-dimensional ones.
Note that the dimension of domain X is equal to that of codomain R
n
. Why
is it so? Assume that we interpret x X as a position of a given object, and
F(x) an information attached to the position (of the object). Then certainly
most important information about the object must be denoted as a vector
(of the same dimension with the domain space) such as velocity or force; or,
as a scalar (real number) such as temperature, humidity, altitude, pressure,
. . . . In fact, these two kinds of information are closely connected (see
Gradient Fields and Potentials below). We call a real valued function f on
X R
n
a scalar eld. I.e., f : X R
n
R is a scalar eld (we usually
call it a real valued function in other contexts though).
The most general way to visualize a vector eld F (of two or three dimension)
is this: Choose a proper number of points from the domain, and draw a vector
F(x), which is emanating from x, for each chosen point x. From now on, we
will often call a point of X a position vector or a radial vector.
Example 1. Let F : R
2
R
2
be a VF such that F(x) = (1, 1). Draw a
picture for the VF F.
See Figure 3.27 on p. 208.
1
Example 2. Depict G : R
2
R
2
, G(x, y) = (y, x).
See Figure 3.28 on p. 209.
Some simple mathematics helps us understand the picture better. Note that
||G(x, y)|| = ||(y, x)|| =
_
y
2
+ x
2
= ||(x, y)||, and that r G(r) = (x, y)
(y, x) = 0. Recall that || r||, called the norm of a vector r, denotes the
size of the vector. Hence, on each concentric circle (with center 0), G(r) is
tangent to the circle, and the size of G(r) is equal to the radius of the circle.
Then, for each r = 0, there are only two such vectors G(r) and G(r). By
inspecting the equation G(1, 0) = (0, 1), we easily see that G(1, 0) is a
clockwise tangent vector to the concentric circle on which the position vector
(1, 0) lie. Then it is our convention to conclude that, for any r = 0, G(r) is
clockwise tangent to the corresponding concentric circle. A natural but not
widely asked question is this. How do we justify the convention?
These are examples of 2-dimensional vector elds. Now we introduces a
3-dimensional VF.
Example 3. (Gravitational vector eld)
F(x, y, z) =
GMm
(x
2
+ y
2
+ z
3
)
3/2
(x, y, z), (x, y, z) = 0.
From the equation, it follows easy that (i) the magnitude of F is inversely
proportional to the square of the distance from the origin (to the given po-
sition), (ii) F points toward the origin. We sometimes call the gravitational
VF an inverse square VF.
See Figure 3.29 on p. 209.
2
As we see, we can understand 2-dimensional examples more easily than 3-
dimensional examples. Yet, it seems certain that our physical world, where
we live, is 3-D. Note that even if our physical world is 3-D, our sense data (see
http://en.wikipedia.org/wiki/Sense data) are in fact largely 2-dimensional.
Hence, this can explain why 2-dimensional examples are more easily under-
standable than 3-dimensional ones.
Gradient Fields and Potentials
A VF F : X R
n
is called a gradient vector eld if F(x) = f(x) for
some scalar valued function f : X R. In such a case we call the scalar
valued function f the potential of the VF F.
Example. Show that the above gravitational VF is a gradient VF.
A level set of a real valued function f : X R
n
R is a set such that
{x | f(x) = f(x
0
)} for some x
0
X. A level set is called a level line when
n = 2, and a level surface when n = 3. Some physical examples of level sets
are contour lines, isobars, and isotherms. We often call a physical level line
(resp. surface) as an equipotential line (resp. surface).
Recall that since
D
u
f(x) = f(x) u,
it follows that the vector f(x) is perpendicular to the corresponding level
set (see Figure 3.30 on p. 211).
Flow Lines of Vector Fields
Denition 3.2. A ow line of a VF F : X R
n
is a path : I R
n
such that

(t) = F((t)).
A ow line of a VF F usually represents a path (in a dynamic sense) of
a particle that moves in the given vector eld F. If so, the derivative of
(i.e., the velocity of the moving particle) at time t is given by the value of
3
the VF F at the corresponding point (see Figure 3.33 on p. 211). Needless
to say calculating a ow line passing through a given point of a VF is a very
important problem in applied mathematics.
Example 4. Calculate the ow lines of the VF F(x, y, z) = (2, 3, 1).
See Figure 3.34 on p. 212.
Example 5. Calculate the ow lines of the VF F(x, y) = (y, x).
See Figure 3.35 on p. 212.
Suppose that a particle is in a VF, and it initially stands still. Then it is our
convention that the movement of the particle is completely subject to the
VF. In mathematical terms, the initial problem

(t) = F((t)), (0) = r

0
has a unique solution. Hence, it follows that the ow lines cannot cross each
other.
4
3.4 Gradient, Divergence, Curl, and the Del Operator
Notations
(x, y): Cartesian (or rectangular) coordinates in R
2
.
(r, ): Polar coordinates in R
2
.
(x, y, z): Cartesian coordinates in R
3
.
(r, , z): Cylindrical coordinates in R
3
.
(, , ): Spherical coordinates in R
3
.
i = (1, 0, 0), j = (0, 1, 0), k = (0, 0, 1) are unit vectors that point in the
directions of increasing x, y, z coordinates, respectively.
Similarly, e
r
, e

, e
z
(= k) are unit vectors that point in the directions of
increasing r, , z coordinates, respectively.
See Figure 1.108 on p. 70.
And e

, e

, e

are unit vectors that point in the directions of increasing

, , coordinates, respectively.
See Figure 1.109 on p. 70.
Recall that i, j, k are orthonormal; i.e., (i) they are unit vectors (||i|| = ||j|| =
||k|| = 1) and (ii) they are mutually perpendicular (i j, i k, j k). It is
not dicult to check that e
r
, e

, e
z
as well as e

, e

, e

are also orthonormal.

Note that i (and j, k as well) points the same direction at every point in R
3
.
On the other hand, the direction, which e
r
(and e

, e

, e

as well) points, can

be dierent from point to point in R
3
.
5
More precisely,
e
r
(and e

as well) points the same direction only on each half-plane =

0
[0, 2), < z < , and
e

(and e

as well) points the same direction only on each ray =

0

[0, 2), =
0
[0, ].
Suppose F is a coordinate free 3-D VF. Then we write
F i = F
1
, F j = F
2
, F k = F
3
,
and F = (F
1
, F
2
, F
3
) = F
1
i + F
2
j + F
3
k in Cartesian coordinates,
F e
r
= F
r
, F e

= F

, F e
z
= F
z
,
and F = (F
r
, F

, F
z
) = F
r
e
r
+ F

+ F
z
e
z
in cylindrical coordinates, and
F e

= F

, F e

= F

, F e

= F

,
and F = (F

, F

, F

) = F

+ F

+ F

in spherical coordinates.
Example 1. Let F(x, y, z) = (x, y, 0) be a 3-D VF. Represent F in cylindrical
coordinates.
Example 2. Let G(x, y, z) = (y, x, 0) be a 3-D VF. Represent G in cylin-
drical coordinates.
Example 3. Let H(x, y, z) = (x, y, z) be a 3-D VF. Represent H in spherical
coordinates.
Gradient
Let f be a 3-D scalar eld. Then
f i = D
i
f =
f
x
, f j = D
j
f =
f
y
, f k = D
k
f =
f
z
;
and thus,
f = (
f
x
,
f
y
,
f
z
).
6
Similarly, we have
f e
r
= D
e
r
f =
f
r
, f e

= D
e

f =
1
r
f

,
f =
f
r
e
r
+
1
r
f

+
f
z
e
z
;
and
f e

= D
e

f =
f

, f e

= D
e

f =
1

,
f =
f

+
1

+
1
sin
f

.
Note that a -interval (resp. -interval) of length l generates an arc of length
rl (resp. l), and that r = sin .
Divergence
Let F be a 3-D vector eld. Then the divergence of F can be expressed in
three dierent ways as follows.
(1) F = F
1
/x + F
2
/y + F
3
/z,
(2) F = F
r
/r + F
r
/r + (1/r)F

/ + F
z
/z,
(3) F = 2F

/+F

/+(1/( sin ))(sin F

)/+(1/( sin ))F

/.
(2) and (3) can be deduced from (1) by the chain rule. However, we omit
the proofs in this stage since the proofs are a little tedious and we will prove
them in a later section. In any case it is highly recommended to verify (2)
and (3) for some concrete VFs.
Example 4. Verify (2) for F(x, y, z) = (x, y, 0) and G(x, y, z) = (y, x, 0).
Example 5. Verify (3) for H(x, y, z) = (x, y, z).
7
Curl
Let F be a 3-D vector eld. Then the curl of F (F) can be expressed in
three dierent ways as follows.
(1)

i j k
/x /y /z
F
1
F
2
F
3

(2)
(
1
r
F
z

1
r
(rF

)
z
,
F
r
z

F
z
r
,
1
r
(rF

)
r

1
r
F
r

)
(3)
1

2
sin

(
( sin F

(F

,
F

( sin F

, sin
(F

sin
F

)
(2) and (3) can be deduced from (1) by the chain rule. However, we omit
the proofs in this stage since the proofs are a little tedious and we will prove
them in a later section. In any case it is highly recommended to verify (2)
and (3) for some concrete VFs.
Example 6. Verify (2) for F(x, y, z) = (x, y, 0) and G(x, y, z) = (y, x, 0).
Example 7. Verify (3) for H(x, y, z) = (x, y, z).
8
4.3 Lagrange Multipliers
Constrained Extrema
Theorem 3.1 Let f, g be scalar functions on an open set X in R
n
. Let S
denote the level set of g at height c, i.e. S = {x X | g(x) = c}. Then if f|
S
has an extremum (maximum or minimum) at x
0
S such that g(x
0
) = 0,
then there exists some scalar such that
f(x
0
) = g(x
0
).
Example. Let T = T(x, y) = 16x
2
+ 24x + 40y
2
. Find the greatest and the
least values of T on the closed unit disk.
Example. Find the shortest distance from (1, 0) to the parabola y
2
= 4x.
Example 1. An open rectangular box is to be manufactured having a xed
volume of 4 m
3
. What dimensions should the box have so as to minimize the
amount of material used to make it?
The Case of More than One Constraint
Theorem 3.2 Let f, g
1
, . . . , g
k
be scalar functions on an open set X in
R
n
(k < n). Let S denote the intersection of level sets of g
1
, . . . , g
k
, i.e.
S = {x X | g
1
(x) = c
1
, . . . , g
k
(x) = c
k
}. Then if f|
S
has an extremum
at x
0
S and if g
1
(x
0
), . . . , g
k
(x
0
) are linearly independent, then there
exist some scalars
1
, . . . ,
k
such that
f(x
0
) =
1
g
1
(x
0
) + +
k
g
k
(x
0
).
Example. Find the minimum of the function f(x, y, z) = x
2
+y
2
+z
2
subject
to x + 2y + 3z = 6 and x + 3y + 6z = 9.
9
Some Applications
Exercise 13. (Cobb-Douglas Model) John has decided to invest $360, 000 in
his factory. His economic analysts have noted that the output of this factory
is modeled by the function
Q(K, L) = 60K
1/3
L
2/3
,
where K represents the amount money spent on equipment and L represents
the amount spent on labor.
(a) How should John allocate the $360, 000 between labor and equipment?
(b) Check that Q/K = Q/L at the optimal value for K and L.
Exercise 14. Let Q(K, L) be a production function for a company where K
and L represent the respective amounts spent on equipment and labor. Let
p denote the price of equipment per unit and w the cost of labor per unit.
Show that, subject to a xed production Q(K, L) = c, the total cost M of
production is minimized when K and L are such that
1
p
Q
K
=
1
w
Q
L
.
Exercise 20. (Snells law of refraction) A ray of light travels at a constant
speed in a uniform medium, but in dierent media light can travel at dierent
speeds. For example, if a ray of light passes from air to water, it is refracted
as shown in Figure 4.44.
See Figure 4.44 on p. 287.
Suppose the speed of light in medium 1 is v
1
and in medium 2 is v
2
. Then,
by Fermats principle of least time, the light strike the boundary between
medium 1 and medium 2 at a point P so that the total time the light travels
is minimized.
(b) Show that the total travel time is minimized when
sin
1
sin
2
=
v
1
v
2
.
10
6.1 Scalar and Vector Line Integrals
Vector Line Integrals
Denition 1.2 The (vector line) integral of (the VF) F along (the path)
x : [a, b] R
3
, denoted by
_
x
F ds, is
_
x
F ds =
_
b
a
F(x(t)) x

(t) dt.
Note that ds = (dx, dy, dz), and that since x(t) = (x(t), y(t), z(t)), x

(t)dt =
(x

(t), y

(t), z

(t))dt = (dx/dt, dy/dt, dz/dt)dt = (dx, dy, dz). Hence, ds =

x

(t)dt.
Observe that if F represents a force eld, then
_
x
F ds is in fact the work
done by F on a particle as the particle moves along the path x.
Example 2. Let F(x, y, z) = (x, y, z), and let x : [0, 1] R
3
be a path such
that x(t) = (t, 3t
2
, 2t
3
). Calculate
_
x
F ds.
Example 5. Evaluate
_
x
(y + z)dx + (x + z)dy + (x + y)dz,
where x(t) = (t, t
2
, t
3
) for 0 t 1.
Exercise 17. Find the work done by the force eld F(x, y, z) = (2z
5

3xy, x
2
, x
2
z) on a particle as the particle moves around the perimeter of the
square with vertices (1, 1, 3), (1, 1, 3), (1, 1, 3), (1, 1, 3), oriented coun-
terclockwise.
Exercise 22. Calculate
_
C
zdx + xdy + ydz,
where C is the curve obtained by intersecting the surfaces z = x
2
and x
2
+
y
2
= 4 and oriented clockwise.
Exercise 29. Let C be a curve on the surface {(x, y, z) | f(x, y, z) = 2010}.
Show that
_
C
f ds = 0.
11
6.2 Greens Theorem
Theorem 2.1 (Greens Theorem) Let D R
2
be a bounded and connected
domain. Assume D (boundary of D) is a closed curve, i.e., D = C; or, is
consisted of a few (separated) closed curves, i.e., D = C
1
C
l
. Let
F = (M(x, y), N(x, y)) be a VF on D. Then we have
_
C
M(x, y)dx + N(x, y)dy =
_ _
D
N
x

M
y
dxdy.
Example 1. Verify Greens theorem when F = (xy, y
2
) and D : x
2
y x.
Example. Show that
1
2
_
D
ydx + xdy =
_ _
D
1 dxdy (area of D),
and verify it for D : 1 x
2
+ y
2
4.
Example. Use Greens theorem to recover the formula
Area =
1
2
_
b
a
r
2
d.
Example. Let F = (y/(x
2
+ y
2
), x/(x
2
+ y
2
)), and let D : x
2
+ y
2
1.
Investigate why Greens theorem seems to fail for this F.
Proof of Greens theorem We rst prove Greens theorem for an ele-
mentary region, and then generalize it to more general regions such as an
annulus.
Denition. A region D R
2
is called an elementary region if there exist
functions , , , such that
D = {(x, y) | (y) x (y), c y d},
D = {(x, y) | (x) y (x), a x b}.
12
6.3 Conservative Vector Fields
Denition 3.1. We say a VF F has path-independent line integrals if
_
C
1
F ds =
_
C
2
F ds,
for any pair of curves C
1
, C
2
having the same initial and terminal points.
Theorem 3.2. F has path-independent line integrals if and only if
_
C
F ds = 0,
for any closed curve.
Example 1. Show that F = (y, x) does not have path-independent line
integrals.
Theorem 3.3. F has path-independent line integrals i F is a gradient VF,
i.e., F = f for some scalar function f (called a potential). Moreover, if A
is the initial point and B is the terminal point of a curve C, then
_
C
F ds = f(B) f(A).
Example. Let x(t) = (t
9
, sin
9
(t/2)) (0 t 1). Evaluate
_
x
ydx + xdy.
Example 7. Show that F = (2xy + cos 2y, x
2
2x sin 2y) is a gradient VF,
and nd a potential of F.
Theorem 3.5. F = (M, N) is a gradient VF in a simply-connected region if
and only if (M, N, 0) = 0.
Example. Investigate this theorem for F = (y/(x
2
+ y
2
), x/(x
2
+ y
2
)).
13
Chapter 7. Surface Integrals and Vector Analysis
7.1 Parametrized Surfaces
In Calculus 2, we studied that surfaces in space can be mathematically rep-
resented either as a graph of a function, i.e., as points (x, y, z) in R
3
satis-
fying z = f(x, y); or as a level set, i.e., as points (x, y, z) in R
3
satisfying
F(x, y, z) = c.
We also studied that a curve (called a parametrized curve) in space can
be represented by using parameters. A parametrized surface in space can be
dened as follows.
Denition 1.1. A parametrized surface in R
3
is a continuous function X :
D R
2
R
3
that is (nearly) 1:1. We call the image X(D) (i.e., geometric
surface) the underlying surface of X and denote it by S. The variables in D
are called parameters.
NOTE that every denition (at least in mathematics) is NOT perfect.
Often, from case to case, it needs to be modied.
Example 4. A surface represented as the graph of a function z = f(x, y) can
also be represented as a parametrized surface.
Example 5 (a). A doughnut-shaped surface generated by rotating a circle
with respect to an axis outside the circle is called a torus. Parametrize a
torus.
Coordinate Curves, Normal Vectors, and Tangent Planes
Let S be a surface parametrized by X=X(s, t). If we x t = t
0
and let only
s vary, we obtain a continuous map
s X(s, t
0
),
whose image is a curve lying in S. We call this curve the s-coordinate curve
at t = t
0
. Similarly, we can x s = s
0
and obtain a map
t X(s
0
, t),
whose image is the t-coordinate curve at s = s
0
.
14
Example 5 (b). Assume a torus is parametrized as follows:
x = (a + b cos t) cos s
y = (a + b cos t) sin s
z = b sin t,
where 0 s, t 2, and 0 < b < a. Find s-coordinate curves at t =
0, /2, , 3/2, and nd t-coordinate curves at s = 0, /2, , 3/2.
Denition 1.2. Let T
s
=
X
s
, T
t
=
X
t
. If
N(s, t) = T
s
(s, t) T
t
(s, t) = 0
for every (s, t) D, then we say that parametrized surface S = X(D) is
smooth. If S is a smooth parametrized surface, we call the nonzero vector
N = T
s
T
t
the standard normal vector arising from the parametrization
X. We also call T
s
, T
t
tangent vectors to the coordinate curves (or to the
surface S).
Recall that if a nonzero vector N = (A, B, C) is normal to the surface S at
(x
0
, y
0
, z
0
), then the tangent plane to the surface S at (x
0
, y
0
, z
0
) is dened
as
A(x x
0
) + B(y y
0
) + C(z z
0
) = 0.
Example. Find an equation to the plane tangent to the ellipsoid
F(x, y, z) = 9x
2
+ 4y
2
+ z
2
29 = 0
at (1, 2, 2).
Example. Find the tangent plane to the surface
z = x
2
+ y
2
at (1, 2, 5).
15
Area of a Smooth Parametrized Surface
Let X : D R
2
R
3
be a parametrization of the geometric surface
S (= X(D)). The key geometric observation (to calculate the area of S)
is as follows: Consider an innitely small (innitesimal) rectangle dA =
[s
0
, s
0
+ ds] [t
0
, t
0
+ dt] D. The image of the rectangle under X is an
innitesimal portion of S that is technically equal to the parallelogram with
a corner at X(s
0
, t
0
) and spanned by the vectors T
s
(s
0
, t
0
)ds and T
t
(s
0
, t
0
)dt.
The area of the parallelogram (and thus X(dA)) is equal to
|| T
s
(s
0
, t
0
)ds T
t
(s
0
, t
0
)dt|| = || T
s
(s
0
, t
0
) T
t
(s
0
, t
0
)|| dsdt.
From this observation we have
Surface area of S =
_ _
D
|| T
s
(s, t) T
t
(s, t)|| dsdt.
Example 12. Calculate the surface area of a sphere of radius a (i) when the
sphere is represented as a parametrized surface, or (ii) when the sphere is
represented as a graph.
Check that the surface area of the graph of f(x, y) over D is equal to
_ _
D
_
f
2
x
+ f
2
y
+ 1 dxdy.
16
7.2 Surface Integrals
Vector Surface Integrals
Since we will not deal with scalar surface integrals in this course, we will call
a vector surface integral just a surface integral.
Denition 2.2. Let X : D R
3
be a smooth parametrized surface, and let
F(x, y, z) be a vector eld. Then the surface integral of F over X, denoted
as
_ _
x
F dS, is dened by
_ _
x
F dS =
_ _
D
F(X(s, t)) N(s, t) dsdt,
where N(s, t) = X
s
X
t
.
Note that as yet the term dS is not explicitly dened.
Recall that the standard normal vector N is not a zero vector, i.e., || N|| = 0.
Hence, by letting
n(s, t) =
N(s, t)
|| N(s, t)||
,
we obtain a unit vector normal to the surface S at X(s, t). Now we dene
dS as n dS; and thus, it follows that
_ _
x
F dS =
_ _
x
(F n) dS,
that is a so-called a scalar surface integral. (Note that F n is a scalar
function.) This implies that a vector surface (resp. line) integral can be
transformed into a scalar surface (resp. line) integral. Moreover, a vector
surface (resp. line) integral must be transformed into a scalar surface (resp.
line) integral in order to be calculated.
Example 4. Let F = (x, y, z 2y). Evaluate
_ _
x
F dS, where X is the
helicoid
X(s, t) = (s cos t, s sin t, t), 0 s 1, 0 t 2.
The helicoid is a round-stairs-shaped surface. (see Figure 7.18 on p. 424).
17
Let X(s, t) (0 s, t 1) be a parametrization of a surface S. Then a map-
ping Y : [0, 1] [0, 1] R
3
such that Y(s, t) = X(t, s) also parametrizes
S. Note that the two unit vectors n
X
(t, s) and n
Y
(s, t) are opposite in direc-
tion. Note also that both n
X
(t, s) and n
Y
(s, t) are normal to S at the same
point (i.e., at Y(s, t) = X(t, s)). This means that two unit normal vectors
are possible at each point on a surface.
Denition 2.6. A surface S is orientable (or two-sided) if it is possible to
choose a single unit normal vector (of the two) at each point of S so that
the collection of these normal vectors can vary continuously over the whole
S. Otherwise, S is called nonorientable (or one-sided).
Note that it is always a simple matter to give an orientation to a curve;
but, surprisingly, it is not always possible to give an orientation to even a
simple surface. A notorious such example is the Mobius strip.
Remark. Let X and Y be parametrizations of an orientable surface S. Let
n
X
(resp. n
Y
) denote the unit normal vector generated by X (resp. Y).
Note that n
X
= n
Y
= 0 or n
X
= n
Y
= 0 at any point on S. Then,
since n
X
and n
Y
are both continuous, n
X
= n
Y
on the whole surface, or
n
X
= n
Y
on the whole surface. We say that parametrizations X and Y are
orientation-preserving if n
X
= n
Y
, and orientation-reversing if n
X
= n
Y
.
Theorem 2.5. Let X and Y be parametrizations of the surface S. Let F
be a VF. Then
_ _
X
F dS =
_ _
Y
F dS,
if X and Y are orientation preserving, or
_ _
X
F dS =
_ _
Y
F dS,
if X and Y are orientation reversing.
Example 8. The Mobius strip can be parametrized by
X(s, t) = (x(s, t), y(s, t), z(s, t)) (0 s 2, 1/2 t 1/2),
where x = (1 + t cos
s
2
) cos s, y = (1 + t cos
s
2
) sin s, z = t sin
s
2
.
18
Exercise 24. The glass dome of a futuristic greenhouse is shaped like the
surface z = 8 2x
2
2y
2
. The greenhouse has a at dirt oor at z = 0.
Suppose that the temperature T, at points in and around the greenhouse,
varies as
T(x, y, z) = x
2
+ y
2
+ 3(z 2)
2
.
Then the temperature gives rise to a heat ux density eld H given by H =
k T. (Here k is a positive constant that depends on the insulating
properties of the particular medium.) Find the total heat ux outward across
the dome and the surface of the ground if k = 1 on the glass and k = 3 on
the ground.
19
7.3 Stokess and Gausss Theorems
Stokess theorem
Theorem 3.2 (Stokess Theorem) Let S be a bounded and orientable surface
in R
3
, and let S denote the boundary of S. Let F be a vector eld. Then
_ _
S
F dS =
_
S
F ds.
Here, we assume that S and S are consistently oriented.
Denition. We say that S and S are consistently oriented if the ngers curl
in the direction of the orientation of S when the thumb of your right hand
points along n. We assume that the surface and its boundary are consistently
oriented if not mentioned otherwise.
As shown below the Stokes theorem implies the Green theorem.
Example. Let F(x, y, z) = (M(x, y), N(x, y), 0). Here, 0 denotes the zero
function. Assume that the surface S is included in the xy-plane; i.e., S
R
2
xy
{0}. Then, it follows that
F = (0, 0,
N
x
(x, y)
M
y
(x, y)),
and that
dS = n dS = k dxdy.
Since S R
2
xy
{0}, the path x of S must be of the form x = (x(t), y(t), 0).
If we denote the surface S as D (it seems natural to do so since S is included
in the xy-plane), then the Stokes theorem can be rewritten as follows:
_ _
D
N
x
(x, y)
M
y
(x, y)dA =
_
D
M(x, y)dx + N(x, y)dy.
That is in fact Greens theorem.
Example 1. Let S be the paraboloid z = 9 x
2
y
2
; z 0, and let
F = (2z y, x + z, 3x 2y). Verify the Stokes theorem.
20
Example. Evaluate the surface integral
_ _
S
F dS,
where S : x
2
+ y
2
+ z
2
= 1, x + y + z 1; and F = (y z, z x, x y).
The orientation of S is taken to be that of the upward normal.
We assume that the surface is oriented in the outward (or upward) direc-
tion if not mentioned otherwise.
Gausss theorem
Theorem 3.3 (Gausss Theorem) Let W be a bounded solid region in R
3
,
and let W denote the boundary of W. Let F be a vector eld in R
3
. Then
_ _ _
W
F dV =
_ _
W
F dS.
Here, we assume that W is oriented in the proper (i.e., outward) direction.
Gausss theorem, often called the divergence theorem, can be easily gen-
eralized to any dimensional Euclidean space.
Theorem (N-dimensional Divergence Theorem) Let be a bounded solid
region in R
n
(n = 1, 2, 3, . . . ), and let denote the boundary of . Let F
be a vector eld in R
n
. Then
_

_

F d =
_

_

F d.
Here, we assume that is oriented in the proper (i.e., outward) direction.
Note that Gausss theorem also implies Greens theorem. Moreover, it
also implies the fundamental theorem of calculus. In fact, 1-dimensional di-
vergence theorem is the fundamental theorem of calculus, and 2-dimensional
divergence theorem is Greens theorem. Recall that the fundamental theorem
of calculus says that
_
[a, b]
dF
dx
dx = F(b) F(a).
21
Example 3. Verify the Gauss theorem when F = (x, y, z) and
W : x
2
+ y
2
a
2
, 0 z b (a, b > 0).
Example. Evaluate the surface integral
_ _
S
F dS ;
where
S : z = (1 x
2
y
2
)e
1x
2
3y
2
, x
2
+ y
2
1,
and
F = (e
y
cos z,

x
3
+ 1 sin z, x
2
+ y
2
+ 3).
The Meaning of Divergence and Curl
Proposition 3.4 (The Meaning of Divergence) Let P be a point in R
3
, and
F be a VF in some neighborhood of P. Then
F(P) = lim
0
+
3
4
3
_ _
S

F dS,
where S

denotes a sphere of radius centered at P. This means that since

_ _
S

FdS represents the ux of F across the sphere S

, lim
1
V ol(S

)
_ _
S

FdS
(i.e., F(P)) represents the limit of the ux per unit volume, in other words
the ux density of F at P.
Proposition 3.5 (The Meaning of Curl) Let P be a point in R
3
, and F be a
VF near P. Let D

be a circle of radius centered at P, and that is normal

to a given unit vector n. Then
n F(P) = lim
0
+
1

2
_
C

F ds,
where C

= D

. This means that since

_
C

F ds represents the circulation

of F along the circle C

, lim
1
Area(D

)
_
C

Fds (i.e., nF(P)) represents the

limit of the circulation per unit area around n, in other words the circulation
density of F at P around n.
22
Proof of Stokess Theorem
Step 1. We rst assume that (i) F = (M(x, y, z), 0, 0); (ii) S : z = f(x, y);
(iii) D : a x b, (x) y (x), where D denotes the domain of f.
Then we prove the following (on the above assumption):
_
S
F ds =
_
b
a
M(x, (x), f(x, (x))) M(x, (x), f(x, (x))) dx.
Next, we prove the following:
_ _
S
F dS =
_
b
a
M(x, (x), f(x, (x))) M(x, (x), f(x, (x))) dx.
Finally, we extend this result to the case that D is a somewhat arbitrary
region of R
2
. That means Stokess theorem holds when F = (M(x, y, z), 0, 0),
and when (ii) S : z = f(x, y).
Step 2. We rst prove that Stokess theorem holds when F = (M(x, y, z), 0, 0),
and when the surface S is one of the following types:
(i) S : z = f(x, y),
(ii) S : y = f(x, z),
(iii) S is a subset of a plane parallel to the yz-plane.
Then we show that most surfaces of practical importance can be decom-
posed into as a nite union of the surfaces of those types. Hence, in a loose
sense, Stokess theorem holds when F is of the form M((x, y, z), 0, 0).
Step 3. We prove Stokess theorem for F = (0, N(x, y, z), 0) (and thus
for F = (0, N(x, y, z), 0) similarly). This (nearly) completes the proof of
Stokess theorem due to the linearity of the curl and integral operators.
23
Proof of Gausss Theorem
Step 1. We rst assume that (i) F = (0, 0, P(x, y, z)); (ii) W : (x, y)
z (x, y); (iii) R : a x b, (x) y (x), where R denotes the
domain of and . In this case we call W an elementary region of type 1. Let
W = S; and let (i) S
1
denote the bottom surface of W, i.e., S
1
: z = (x, y),
(ii) S
2
denote the top surface of W, i.e., S
2
: z = (x, y), and (ii) S
3
denote
the lateral surface of W, i.e., S
3
: (x, y) z (x, y), (x, y) R. Note
that S = S
1
S
2
S
3
and S
1
, S
2
, S
3
are mutually disjoint.
Then we have the following:
_ _
S
1
F dS =
_ _
R
P(x, y, (x, y)) dxdy,
_ _
S
2
F dS =
_ _
R
P(x, y, (x, y)) dxdy,
_ _
S
3
F dS = 0;
and thus,
_ _
S
F dS =
_ _
R
P(x, y, (x, y)) P(x, y, (x, y)) dxdy.
On the other hand, since F = P/z, we have the following:
_ _ _
W
FdV =
_ _
R
P(x, y, (x, y)) P(x, y, (x, y)) dxdy.
Step 2. We similarly dene elementary regions of type 2 and of type 3. A
region simultaneously of types 1, 2, and 3 is called an elementary region.
If W is an elementary region, then we have
_ _
W=S
F dS =
_ _ _
W
F dV.
Step 3. We extend the result in Step 2 to the case that W is a more general
region of R
3
. Note that a lled torus can be such an example.
24
The Divergence in Cylindrical Coordinates
Exercise 23. Write
F = F
r
e
r
+ F

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

= F

, F e
z
= F
z
.
Consider the innitesimal cylindrical coordinate cuboid S shown in Figure
7.45 on p. 454, in which the pairs of opposite faces correspond to values
r dr/2 and r + dr/2;
d/2 and + d/2;
z dz/2 and z + dz/2.
Since
F(r, , z) =
1
r dr d dz
_ _
S
F dS,
we have
F = F
r
/r + F
r
/r + (1/r)F

/ + F
z
/z.
The Divergence in Spherical Coordinates
Exercise 24. Write
F = F

+ F

+ F

,
where F e

= F

, F e

= F

, F e

= F

.
Consider the innitely small spherical coordinate cuboid S shown in Figure
7.46 on p. 455, in which the pairs of opposite faces correspond to values
d/2 and + d/2;
d/2 and + d/2;
d/2 and + d/2.
Since
F(, , ) =
1

2
sin d d d
_ _
S
F dS,
we have
F = 2F

/ +F

/ + (1/( sin ))(sin F

)/ + (1/( sin ))F

/.
25
The Curl in Cylindrical Coordinates
Exercise 26. Write
F = F
r
e
r
+ F

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

= F

, F e
z
= F
z
.
(a) Consider the innitesimal cylindrical coordinate rectangle shown in Figure
7.48 on p. 455 dened by
r
0
r r
0
+ dr;

0

0
+ d.
Show that
e
z
F =
1
r
(rF

)
r

1
r
F
r

.
(b) Consider the innitesimal cylindrical coordinate rectangle shown in Figure
7.49 on p. 455 dened by
z
0
z z
0
+ dz;

0

0
+ d.
Show that
e
r
F =
1
r
F
z

z
.
(c) Consider the innitesimal cylindrical coordinate rectangle shown in Figure
7.50 on p. 456 dened by
r
0
r r
0
+ dr;
z
0
z z
0
+ dz.
Show that
e

F =
F
r
z

F
z
r
.
26
The Curl in Spherical Coordinates
Exercise 27. Write
F = F

+ F

+ F

,
where F e

= F

, F e

= F

, F e

= F

.
(a) Consider the innitesimal spherical coordinate rectangle shown in Figure
7.51 on p. 456 dened by

0

0
+ d;

0

0
+ d.
Show that
e

F =
1
sin
[
(sin F

].
(b) Consider the innitesimal spherical coordinate rectangle shown in Figure
7.51 on p. 456 dened by

0

0
+ d;

0

0
+ d.
Show that
e

F =
1

[
1
sin
F

(F

].
(c) Consider the innitesimal spherical coordinate rectangle shown in Figure
7.51 on p. 456 dened by

0

0
+ d;

0

0
+ d.
Show that
e

F =
1

[
(F

].
27
7.4 Further Vector Analysis; Maxwells Equations
Greens formulas
Greens rst formula (identity):
_ _
S=W
f g dS =
_ _ _
W
(f
2
g + f g) dV.
Greens second formula (identity):
_ _
S=W
(f g g f) dS =
_ _ _
W
(f
2
g g
2
f) dV.
Maxwells Equations
Gausss law for electric elds
It is known that the electric eld E arisen from a single point charge q at 0
is given by
E(r) =
q
4
r
r
3
.
If E is arisen from a single point charge q at 0, then
_ _
S
E dS = 0 if 0 / W,
and
_ _
S
E dS = q if 0 W/W.
If E is arisen from a nite number of point charges q
j
at position r
j
(j = 1, 2, . . . , n), then
_ _
S
E dS =

r
j
W
q
j
.
If E is arisen from a continuous charge distribution given by a charge
density (r), then
_ _
S
E dS =
_ _ _
W
dV.
28
It follows from the above equation that
E = .
Amp`ere circuital law
Let the vector eld J describe the electric current density, and let B denote
the induced magnetic eld. Let S be a non-closed surface. It is empirically
known that
_
S
B ds =
_ _
S
J dS.
Then it follows that
B = J.
Faradays law of induction
Let E(r, t) and B(r, t) be time-dependent electric and magnetic elds re-
spectively. Let S be a non-closed surface. It is empirically known that the
voltage around S equals the negative rate of change of magnetic ux across
S. Then it follows that
E =
B
t
.
29
Fourier Series
The Laplace equation in two dimensions is as follows:
(1)
2
u(x, y) =

2
u
x
2
+

2
u
y
2
= 0.
Example 1 Verify that u in (1) represents a steady state temperature dis-
tribution in two dimensions.
(Solution) Let M = [x
0
, x
1
] [y
0
, y
1
]. Since there is no net heat ow into M
or out of it, we have
_
y
1
y
0
u
x
(x
1
, y) dy
_
y
1
y
0
u
x
(x
0
, y) dy +
_
x
1
x
0
u
y
(x, y
1
) dx
_
x
1
x
0
u
y
(x, y
0
) dx
=
_
y
1
y
0
u
x
(x
1
, y)
u
x
(x
0
, y) dy +
_
x
1
x
0
u
y
(x, y
1
) dx
_
x
1
x
0
u
y
(x, y
0
) dx = 0.
Divide this equation by x
1
x
0
and let x
1
x
0
. Then we have
_
y
1
y
0

2
u
x
2
(x
0
, y) dy +
u
y
(x
0
, y
1
)
u
y
(x
0
, y
0
) = 0.
Now divide by y
1
y
0
and let y
1
y
0
. This yields

2
u
x
2
(x
0
, y
0
) +

2
u
y
2
(x
0
, y
0
) = 0.
Example 2 Derive

2
u(r, ) =

2
u
r
2
(r, ) +
1
r
u
r
(r, ) +
1
r
2

2
u

2
(r, ) = 0
directly or from (1) by a change of variables.
30
Example 3 [Dirichlets Problem on the Unit Circle] Solve
_

2
u(x, y) = 0 on x
2
+ y
2
< 1,
u(x, y) = f(x, y) on x
2
+ y
2
= 1
(
_

2
u
r
2
(r, ) +
1
r
u
r
(r, ) +
1
r
2

2
u

2
(r, ) = 0 (0 < r < 1)
u(1, ) = f()
u(r, + 2) = u(r, )
u(r, ) is continuous at r = 0)
Example 4 Show that the following periodic functions
1, re
i
, r
2
e
i2
, r
3
e
i3
, . . .
or
1, r cos , r sin , r
2
cos 2, r
2
sin 2, r
3
cos 3, r
3
sin 3, . . .
satisfy

2
u
r
2
(r, ) +
1
r
u
r
(r, ) +
1
r
2

2
u

2
(r, ) = 0.
Questions 1. Does a convergent innite series

k=
c
k
r
|k|
e
ik
or

k=0
(a
k
r
k
cos k + b
k
r
k
sin k)
also satisfy the equation
u
rr
+
1
r
u
r
+
1
r
2
u

= 0?
2. Can the boundary condition (i.e., u(1, ) = f()) be satised by some c
k
(or a
k
and b
k
)? If so, how can we nd them?
31
Example 5 Show that a function f on R can be represented as the sum of
an odd function and an even function.
Example 6 Show that
(i)
_

cos kx cos jx dx =
j
k
(k, j = 1, 2, 3, . . . );
(ii)
_

sin kx sin jx dx =
j
k
(k, j = 1, 2, 3, . . . );
(iii)
_

cos kx sin jx dx = 0 (k, j = 1, 2, 3, . . . );

(iv)
_

e
ikx
e
ijx
dx = 2
j
k
(k, j = 1, 2, 3, . . . ).
Recall that the Kronecker delta
j
k
is dened by

j
k
=
_
1 if k = j,
0 if k = j.
Example 7 Find a
k
, b
k
, c
k
such that
f(x)

k=
c
k
e
ikx
= a
0
/2 +

k=1
(a
k
cos kx + b
k
sin kx)
where f is a periodic function with period 2.
We say that

k=
c
k
e
ikx
(c
k
=
1
2
_

f(x) e
ikx
dx)
is the Fourier series of f in its exponential form, and
a
0
2
+

k=1
(a
k
cos kx+b
k
sin kx) (a
k
=
1

f(x) cos kx dx, b

k
=
1

f(x) sin kx dx)

is the Fourier series of f in its trigonometric form.
32
Example 8 Show that
a
n
= c
n
+ c
n
, b
n
= i(c
n
c
n
), c
n
=
a
n
ib
n
2
, c
n
=
a
n
+ ib
n
2
.
Theorem 9 Let f be a periodic function with period 2. Then the Fourier
series
a
0
2
+

k=1
(a
k
cos kx + b
k
sin kx) (or

k=
c
k
e
ikx
) converges to f(x)
when f is continuous at x, and it converges to
f(x+)+f(x)
2
otherwise.
Example 10 (Pulse function) Let f : R R be a periodic function with
period 2 such that
f(x) =
_

_
1 ( < x < 0),
0 (x = 0, ),
1 (0 < x < ).
Notice that f is an odd function. Find the Fourier series of f.
Example 11 (Triangular waves) Let f : R R be a periodic function with
period 2 such that
f(x) =
_
/2 + x ( x < 0),
/2 x (0 x < ).
Notice that f is an even function. Find the Fourier series of f and evaluate
(2) =

k=1
1
k
2
.
Example 12 [Euler] The formula

k=1
1
k
2
=

2
6
was rst found by Euler. His original proof is as follows.
Heuristic Proof. Let f(x) = sin x/x = 1x
2
/3!+x
5
/5! . Then, f(0) = 1
and f(k) = 0 for k Z. Note that if P(x) = a
n
x
n
+ +a
1
x+a
0
(a
n
= 0)
is a polynomial such that a
0
= 1, P(
k
) = 0 (k = 1, 2, . . . , n), then
P(x) = (1
x

1
)(1
x

2
) (1
x

n
).
33
Hence, it seems natural to assume that
f(x) = (1 x/)(1 + x/)(1 x/2)(1 + x/2)
= (1 x
2
/
2
)(1 x
2
/(2)
2
)
= 1 (1/
2
+ 1/(2)
2
+ 1/(3)
2
+ )x
2
+ .
By comparing coecients of x
2
for the two expansions of f(x), we obtain
that
1/3! = 1/
2
+ 1/(2)
2
+ 1/(3)
2
+ .
In other words,

k=1
1
k
2
=

2
6
.
Example 13 [Dirichlet Kernel] Let
D
n
(t) =
1
2
n

k=n
e
ikt
.
Then,
(i) D
n
(t) = D
n
(t); (ii)
_

D
n
(t)dt = 1; (iii) D
n
(t) =
sin(n + 1/2)t
2 sin(t/2)
.
Example 14 [Fejer Kernel] Let
K
n
(t) =
1
n
n1

k=0
D
k
(t).
Then,
(i) K
n
(t) = K
n
(t); (ii)
_

K
n
(t)dt = 1; (iii) K
n
(t) 0;
(iv) K
n
(t) =
1
2n
sin
2
(nt/2)
sin
2
(t/2)
=
1
2n
1 cos(nt)
1 cos t
.
34
Example 15 Show that
sin x
x
=

n=1
(1
x
2
n
2
) (1 x 1).
Hints: 1. Find the Fourier series of f
t
(x) = cos tx (0 < t < 1) and 2.
d/dt(log(sin t/t)) = cot t 1/t, d/dt(log(1 t
2
/n
2
)) = 2t/(t
2
n
2
).
Example 16 Deduce the Wallis formula

2
=
2
1
2
3
4
3
4
5
6
5
6
7

from the above formula
sin x
x
=

n=1
(1
x
2
n
2
).
Example 17 Show that
1
sin x
=
1
x
+

n=1
(1)
n
(
1
x n
+
1
x + n
) (0 < x < ).
Hints: Find the Fourier series of f
t
(x) = cos tx (0 < t < 1).
Example 18 Show that
_

0
sin t
t
dt =

2
.
Example 19 Using Example 18, show that
_

0
sin t cos t
t
dt =

4
;
_

0
sin
2
t
t
2
dt =

2
.
Example 20 Show the Vi`ete product formula
2

2
2
_
2 +

2
2
_
2 +
_
2 +

2
2
.
35
Lemma 1 Let f
n
(x) = a
0
/2 +a
1
cos x +b
1
sin x + +a
n
cos nx +b
n
sin nx
be a partial sum of the F-series of f(x). Then,
1

f(x)f
n
(x)dx =
a
0
2
2
+
n

k=1
(a
k
2
+ b
k
2
) =
1

f
n
2
(x)dx;
and
1

(f(x) f
n
(x))
2
dx
1

(f(x) g
n
(x))
2
dx
for any g
n
(x) = p
0
/2 + p
1
cos x + q
1
sin x + + p
n
cos nx + q
n
sin nx.
Theorem 2 [Bessels Inequality] Let a
0
/2 + a
1
cos x + b
1
sin x + be the
F-series of f(x). Then,
a
0
2
2
+

k=1
(a
k
2
+ b
k
2
)
1

f
2
(x)dx.
Corollary 3 [Riemann-Lebesgue Lemma] Let a
0
/2 +a
1
cos x +b
1
sin x +
be the F-series of f(x). Then,
lim
k
a
k
= lim
k
b
k
= 0.
In other words,
lim
k
_

f(x) cos kxdx = lim

k
_

f(x) sin kxdx = 0.

Theorem 4 [Plancherels equality] Let a
0
/2 +a
1
cos x +b
1
sin x + be the
F-series of f(x). Then,
a
0
2
2
+

k=1
(a
k
2
+ b
k
2
) =
1

f
2
(x)dx.
36
Theorem 5 [Riesz-Fischer Theorem]

k=
c
k

d
k
=
1
2
_

f(x)g(x) dx (Parseval

s equality),
where c
k
=
1
2
_

f(x) e
ikx
dx, d
k
=
1
2
_

g(x) e
ikx
dx.
Note: The Riesz-Fischer theorem implies that L
2
and l
2
are isomorphic. This
result is very important, in particular, in Quantum Mechanics. In fact, John
von Neumann showed the equivalence between the wave mechanics and
the matrix mechanics based on this fact.
Legendres Equation, Legendre Polynomials, and Legendre-Fourier
Series
Solving the following equation
(1 x
2
)y

2xy

+ y = 0 (1 x 1)
is equivalent to nding the eigenvalues and eigenvectors of the following equa-
tion
((x
2
1)D
2
+ 2xD)y = y (1 x 1),
where D = d/dx.
Fact 1. Eigenvalues are
n
= n(n + 1) (n = 0, 1, 2, . . . ). Note that
lim
n

n
= .
Fact 2. The eigenfunction space {y
n
(x)} corresponding to each eingenvalue

n
is one dimensional.
Fact 3. y
n+1
(x) is linearly independent from y
1
(x), y
2
(x), . . . , y
n
(x).
37
If we substitude y(x) =

k=0
a
k
x
k
into the Legendres equation
(1 x
2
)y

2xy

+ n(n + 1)y = 0 (1 x 1),

then we have a recurrence relation
a
k+2
=
(n + k + 1)(n k)
(k + 1)(k + 2)
a
k
(k = 0, 1, 2, . . . ).
Let P
n
(x) = y
n
(x) with a
n
=
(2n)!
2
n
(n!)
2
for n = 0, 1, 2, . . . (note that P
0
(x) =
1). Then we have
P
n
(x) =

0kn/2
(1)
k
(2n 2k)!
2
n
k!(n k)!(n 2k)!
x
n2k
.
We also have
P
n
(x) =
1
2
n
n!
d
n
dx
n
[(x
2
1)
n
] (Rodrigues

s formula)
and
G(u, x) =

n=0
P
n
(x)u
n
=
1

1 2xu + u
2
(generating function).
Example Show that
(P
i
, P
j
) =
_
1
1
P
i
(x)P
j
(x)dx =
2
2j + 1

j
i
.
Hint: Letting () =
_

0
e
t
t
1
dt and B(x, y) =
_
1
0
t
x1
(1 t)
y1
dt, then
(x + y)B(x, y) = (x)(y) (, x, y > 0).
38
How are Fourier Series and Taylor Series related
Let f(z) =
1
1z
, z = re
i
, and let f = u + iv, where u and v are real valued
functions. Then
1
1 z
=
1
1 z

1 z
1 z
=
1 z
1 (z + z) + z z
=
1 r cos + ir sin
1 2r cos + r
2
so that
u(z) =
1 r cos
1 2r cos + r
2
and v(z) =
r sin
1 2r cos + r
2
.
Since f(z) = 1/(1 z) = 1 + z + z
2
+ z
3
+ (|z| < 1), f(re
i
) =
1 + (re
i
) + (re
i
)
2
+ (re
i
)
3
+ (|r| < 1). In particular,
u(re
i
) = 1 + r cos + r
2
cos 2 + r
3
cos 3 + ,
v(re
i
) = 0 + r sin + r
2
sin 2 + r
3
sin 3 + .
Letting r = 1/2 and = /4, then we have rspectively
2 sin
5 4 cos
= (
1
2
) sin + (
1
2
)
2
sin 2 + (
1
2
)
3
sin 3 + (< < ),
r

2(1 + r
2
) 2r
=
1

2
r+r
2
+
1

2
r
3

2
r
5
r
6

2
r
7
+
1

2
r
9
+ (|r| < 1).
As an application, we have
_
2
0
2 sin sin k
5 4 cos
d =

2
k
.
39
Let f(z) = e
z
, z = re
i
, and let f = u + iv, where u and v are real valued
functions. Then
e
z
= e
r cos +ir sin
= e
r cos
(cos(r sin ) + i sin(r sin )).
so that
u(z) = u(re
i
) = e
r cos
cos(r sin ) and v(z) = v(re
i
) = e
r cos
sin(r sin ).
Since f(z) = e
z
= 1 +z/1! +z
2
/2! +z
3
/3! + , f(re
i
) = 1 +(re
i
)/1! +
(re
i
)
2
/2! + (re
i
)
3
/3! + . In particular,
u(re
i
) = 1 + r cos /1! + r
2
cos 2/2! + r
3
cos 3/3! + ,
v(re
i
) = 0 + r sin /1! + r
2
sin 2/2! + r
3
sin 3/3! + .
Letting r = 1, then we have
e
cos
cos(sin ) = 1 + cos /1! + cos 2/2! + cos 3/3! + ,
e
cos
sin(sin ) = 0 + sin /1! + sin 2/2! + sin 3/3! + .
As an application, we have
_
2
0
e
cos
cos(sin ) cos kd =

k!
.
40