Vector-Calculus Class Note Vector-Calculus Class Note

lOMoARcPSD|5686828
Vector-Calculus Class Note
Several Variable Calculas (Indian Institute of Technology Kanpur)
StuDocu is not sponsored or endorsed by any college or university

Downloaded by Suraj Gautam ([email protected])
lOMoARcPSD|5686828
Vector Calculus Lecture Note

MA - 121
Posenjit Das
Department of Mathematics,
Indian Institute of Space Science and Technology,
Valiamala P.O., Trivandrum 695 547, India
email: [email protected]

lOMoARcPSD|5686828
ii

lOMoARcPSD|5686828
Contents
1 The Space of Vectors 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Algebraic Structure . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Spatial Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Topology of the Euclidean Space Rm . . . . . . . . . . . . . . . . 7
1.5 Few Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5.1 Line passing though a point, and parallel to a vector . . . 10
1.5.2 Line passing through a two points . . . . . . . . . . . . . 10
1.5.3 Plane passing through a point, and perpendicular to a
vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Directional Derivative and Differentiability 11

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Directional Derivative and Differentiability . . . . . . . . . . . . 12
2.3 Application: Level sets . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Arc Length Function 21

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Parametric Curves . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Arc Lengh of Parametric Curves . . . . . . . . . . . . . . . . . . 23
3.4 Arc Length Function of Parametric Curves . . . . . . . . . . . . 24
4 Line Integration 27
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Line Integration of Scalar Fields . . . . . . . . . . . . . . . . . . 27
4.3 Line Integration of Vector Fields . . . . . . . . . . . . . . . . . . 28
4.4 Conservative Vector Fields . . . . . . . . . . . . . . . . . . . . . . 30
4.5 1st Structure Theorem of CVFs . . . . . . . . . . . . . . . . . . . 32
4.6 Fundamental Theorem in R2 : Green’s Theorem . . . . . . . . . . 35
4.6.1 Application: 2nd Strure Theorem of CVFs in R2 . . . . . 36
4.6.2 Application: Extended Green’s Theorem . . . . . . . . . . 36
4.6.3 Application: Recognizing CVF in R2 . . . . . . . . . . . . 38
4.6.4 Application: Finding Area enclosed by a loop . . . . . . . 38
5 Surface Integration 41
iii

lOMoARcPSD|5686828
iv CONTENTS

lOMoARcPSD|5686828
Chapter 1
The Space of Vectors
1.1 Introduction
Though the need of the conception of vectors was realized after the discovery of
physical quantities carrying multiple values together, e.g., velocity, force, e.t.c.,
significant development in the systematic study of vector is done in the recent
era, in the mid of 19th century. The primary development in the theory of
vectors is witnessed in the late 17th century, while studying the of structure
of complex numbers. By the passage of time the theory got enriched, as the
need grew, to understand the geometrical objects and physical phenomenons.
The modern theory vectors which is a quite sophisticated in nature is developed
between late 19th century and early 20th century.
In the early development, vectors were defined to be finite straight line seg-
ments with unique arrow heads, where the length of the straight line segments
represented the “magnitude” of the vectors and the arrow determined the “di-
rection” of the vector; but it failed to explain many properties of vectors which
should be satisfied naturally. Later, a corrected model of vectors was formalized
by defining vectors as pairs of points in the corresponding space; and subse-
quently more refined and formal definition represents vectors as points in the
ambient space.
The modern definition of vectors is axiomatic in nature; it defines vector by

enlisting all the properties which are wished to be imposed (manifest) upon the
vectors. Although these properties altogether endows the set of vectors with
two structures, one is algebraic structure and the other one is spatial structure,
the set of vectors, in general, is defined only through algebraic properties; and
afterwards the spatial structure is manifested on it.
1.2 Algebraic Structure of Vectors

Definition 1.2.1. A set of elements V = {v̄i | i ∈ λ} is said to be a space of
vectors (or a vector space) over the reals R (or complex numbers C) if it has
the following algebraic structure:

lOMoARcPSD|5686828
2 CHAPTER 1. THE SPACE OF VECTORS
(1) (Existence of binary operation) There exists a function + : V × V −→ V

which is called the “addition” operation. Through this binary operation
two elements v̄1 and v̄2 from V generate a third element which we denote
by v̄1 + v̄2 , in V.
(2) (Associativity of addition) Each v̄1 , v̄2 and v̄3 in V should satisfy (v̄1 +
v̄2 ) + v̄3 = v̄1 + (v̄2 + v̄3 ).
(3) (Existence of additive identity) There exists an element 0̄ such that v̄+ 0̄ =
0̄ + v̄ = v̄ for all v̄ in V. Such element 0̄ is called an additive identity of V.
(4) (Existence of additive inverse) For each v̄ in V there exists an element v̄ ′
such that v̄ + v̄ ′ = v̄ ′ + v̄ = 0̄. Such element v̄ ′ is called an additive inverse
of v̄ in V.
(5) (Commutativity of addition) Each v̄1 and v̄2 in V must satisfy v̄1 + v̄2 =
v̄2 + v̄1 .
(6) (Existence of outer multiplication) There exists a function · : R × V −→ V
which is called the “outer multiplication” operation. For an element α
form R and an element v̄ from V, this operation generates a new element
of V, which is denoted by α.v̄.
(7) (Distribution of addition) Any element α in R and v̄1 , v̄2 in V should
satisfy α.(v̄1 + v̄2 ) = α.v̄1 + α.v̄2 .
(8) (Distribution of multiplication) For α, β in R and for v̄ in V we must have
(α + β).v̄ = α.v̄ + β.v̄.
(9) (Compatibility of multiplication) Any α, β in R any v̄ in V must satisfy
(αβ).v̄ = α(β.v̄) = β.(αv̄).
(10) (Identity of multiplication) For any v̄ in V we should have 1.v̄ = v̄.
Exercise 1.2.2. Show that in a space of vectors

a. Additive identity element is unique, i.e., there exists one and only one additive
identity element in a space of vectors; and it is denoted by 0̄.
b. Each element v̄ has unique additive inverse which is denoted by −v̄.
c. For an element v̄ ∈ V, if we have v̄ + v̄ = v̄, then v̄ = 0̄.
d. For any element v̄, we must have 0.v̄ = 0̄.
e. For any element v̄, the element −1.v̄ is the additive inverse of v̄, i.e., −1.v̄ = −v̄.
One may ask whether the above definition make sense. The following ex-
amples will give an affirmative answer to the question and hence establish the
existence of the space of vectors.
Example 1.2.3.
(I) Let V = R2 = {(x, y)| x, y ∈ R}. Define the addition operation “+” by
(x1 , y1 ) + (x2 , y2 ) = (x1 + x2 , y1 + y2 ), i.e., coordinate wise addition. One
can easily check that the operation + is associative and commutative in nature;
(0, 0) is the additive identity; and for any vector (x, y) in R, (−x, −y) is the
additive inverse of it. Define the outer multiplication operation “.” by α.(x, y) =

lOMoARcPSD|5686828
1.2. ALGEBRAIC STRUCTURE 3
(αx, αy), i.e., coordinate wise multiplication and by definition it is obvious that
“.” satisfies the properties (6)-(10) of Definition 1.2.1. This shows that V is a
space of vectors over R.
(II) In a similar fashion Rn , n ≥ 1 forms space of vectors over R with respect to
coordinate wise addition and multiplication.
(III) Suppose (x0 , y0 ) ∈ R2 . Let L(x0 ,y0 ) = {(λx0 , λy0 ) | λ ∈ R}. One should be
able to see that L(x0 ,y0 ) is a space of vectors over R. Note that the space of
vectors L(x0 ,y0 ) is a subset of the space of vectors R2 .
(IV) Let V = C[0, 1] be the set of all real valued continuous functions over [0, 1].
For f, g ∈ V, define f + g by (f + g)(x) = f(x) + g(x) for all x ∈ [0, 1], i.e.,
point wise addition; and for α ∈ R and f ∈ V define α.f by (α.f)(x) = αf(x) for
all x ∈ [0, 1]. It is easy to see that with respect these two operations V forms
a space of vectors over R. The additive identity of this space is the constant
function zero, i.e., the function 0(x) = 0 for all x ∈ [0, 1]; and for each element
f ∈ V, the additive inverse of f is the function −f given by (−f)(x) = −f(x) for
all x ∈ [0, 1].
(V) Let V = Mm×n (R) be the set of all m×n matrices with real entries. Let “+”
denote the usual matrix addition operation, i.e., for A = (aij ) and B = (bij ) in
V, A + B is defined by A + B = (aij + bij ). Define an operation . : R × V −→ V
by α.(aij ) = (αaij ) where α ∈ R and A = (aij ) ∈ V. It is easy to check that
with respect to these two operations V forms a space of vectors over R.
Definition 1.2.4. Let V be a space of vectors over R and S := {v̄1 , v̄2 , · · · , v̄n }
be a collection of vectors from V. By a linear combination of the vectors in S
we mean an element of the form α1 v̄1 + α2 v̄2 + · · · + αn v̄n where αi ∈ R are
constants.
Let < S >:= {α1 v̄1 + α2 v̄2 + · · · + αn v̄n | αi ∈ R} be the collection of all
possible linear combinations of vectors from S. It can be proved that < S > forms
a vector space “being in V”. < S > is called the linear span/hull of vectors in
S. Sometimes < S > is denoted by L(S).
Example 1.2.5. 1. Let v̄ ∈ R2 . Then < {v̄} > in R2 is a line passing though
origin (0, 0) and v̄.
2. Let v̄1 , v̄2 ∈ R3 , then < {v̄1 , v̄2 } > is either a straight line passing though
origin or a plane containing (0, 0), v̄1 and v̄2 in R3 .
3. Let v̄1 , v̄2 ∈ R2 , then < {v̄1 , v̄2 } > is either a straight line passing though
origin or the whole plane R2 .
4. Let e1 = (1, 0) and e2 = (0, 1). It can be checked that < {e1 , e2 } >= R2 .
5. Let e1 = (1, 0, 0), e2 = (0, 1, 0) and e3 = (0, 0, 1). It can be checked that
< {e1 , e2 , e3 } >= R3 .
Definition 1.2.6. A subset S of a vector space V over R is called a basis, if S
generates V, i.e., < S >= V; and any proper subset of S can not generate V
Example 1.2.7. 1. {e1 , e2 } is a basis of R2 .
2. {(1, 1), (2, 2)} is not a basis of R2 .

lOMoARcPSD|5686828
3. {(1, 1), (2, 1)} is a basis of R2 . (Prove it!)
4. {(1, 1), (2, 0), (0, 3)} is not a basis of R2 .
Exercise 1.2.8. Show that any basis of Rm contains exactly m elements.
We are now ready to introduce the axiomatic notion of length of a vector

through a function called “norm” which will give a spatial structure on the space
of vectors. The concept of length of vector is closely related to the thought of
distance and hence endowing the space of vectors by this spatial structure will
enable us to talk about continuity, differentiability, integrability of functions on
the space of vectors.
1.3 The Spatial Structure of Vectors

Definition 1.3.1. Let V be a space of vectors over R. A function k.k : V −→ R
is called norm if it satisfies the following properties:
N-1: kv̄k ≥ 0 for all v̄ in V; and kv̄k = 0 if and only if v̄ = 0̄.
N-2: kαv̄k = |α|kv̄k for all α in R and for all v̄ in V.
N-3: kv¯1 + v¯2 k ≤ kv¯1 k + kv¯2 k for all v1 , v2 in V.
For any vector v̄ in V, kv̄k is called the norm or length of v̄. A space of
vectors endowed with a norm function is called a normed space. Geometrically
norm of a vector v̄ represents the length of line segment joining 0̄ and v̄.
Example 1.3.2.
(I) Let V = Rm where n is a natural number. We already have seen that V
is
q a vector space over R. For any v̄ = (v1 , v2 , · · · , vm ) in V define kv̄k =
v21 + v22 + · · · v2m . Then one can check that the function k.k : V −→ R is a
norm on V. This norm is called the Euclidean norm on Rm .
(II) Let V = C[0, 1]. Then V is a vector space over R. Define a function k.k :
V −→ R by kfk = sup {|f(x)| | x ∈ [0, 1]}. It can be seen that the function k.k is
a norm on V.
(III) For the same vector space as in the previous example, for any f ∈ C[0, 1]
Z 1 !1/2
define kfk = f2 . One can show that the defined function k.k : V −→ R
0
is a norm called the L2 -norm..
(IV) Let V = Mm×n (R). It has already been established that V is a vector space
over R. Define k.k : V −→ R by kAk = sup {kAx̄T k | x̄ ∈ Rn , kx̄k = 1} where A
is in V. The reader may take it as an exercise to show that the function k.k is
a norm on V.
Having a norm function k.k on a space of vectors we can define a metric,

i.e., a distance function on V which measures the distance between any two
elements in V such that the distance between 0̄ and any element v̄ in V remains
the desired one, kv̄k.

lOMoARcPSD|5686828
1.3. SPATIAL STRUCTURE 5
Definition 1.3.3. Let V be a space of vectors over R and k.k : V −→ R a norm.

Define a function d : V × V −→ R by d(x̄, ȳ) = kx̄ − ȳk for all x̄, ȳ in V. This
function d is called the metric or the distance function on V induced by the
norm k.k.
We see that this metric has some nice characteristics.
Proposition 1.3.4. Suppose V be a space of vectors endowed with a norm k.k
and d the metric induced by the norm k.k. Then the metric d : V × V −→ R
has the following properties:
M-1: (Non-negativity) d(x̄, ȳ) ≥ 0 for all x̄, ȳ ∈ Rm ; and d(x̄, ȳ) = 0 if and only
if x̄ = ȳ.
M-2: (Symmetricity) d(x̄, ȳ) = d(ȳ, x̄) for all x̄, ȳ ∈ V.
M-3: (Triangle Inequality) d(x̄, ȳ) + d(ȳ, z̄) ≥ d(x̄, z̄) for all x̄, ȳ, z̄ ∈ V; and
d(x̄, ȳ) + d(ȳ, z̄) = d(x̄, z̄) if and only if x̄, ȳ and z̄ are collinear.
N-1: d(x̄, ȳ) = d(0̄, x̄ − ȳ) for all x̄, ȳ ∈ V.
N-2: d(αx̄, αȳ) = |α|d(x̄, ȳ) for all x̄, ȳ ∈ V.
In this context it should be noted that if a metric is to be defined on a
space, not necessarily space of vectors, it is defined axiomatically satisfying the
properties M-1 to M-3 in Proposition 1.3.4.
Let us now compute some metrics corresponding to norms in the respective

spaces.
Example 1.3.5.
(I) In Example 1.3.2, (I) if we compute the metric induced by the corresponding
norm, then the metric is given by
q
d(x̄, ȳ) = (x1 − y1 )2 + (x2 − y2 )2 + · · · + (xm − ym )2
where x̄ = (x̄1 , x2 , · · · , xm ), ȳ = (y1 , y2 , · · · , ym ) ∈ Rm . This metric is on Rm

is called the Euclidean metric on Rm .
(II) In Example 1.3.2, (II) the induced metric will be given by
d(f, g) = sup {|(f − g)|(x) | x ∈ [0, 1]} for all f, g ∈ C[0, 1].
This metric is called the supremum metric or the uniform metric on C[0, 1].
In high school Mathematics we have learned the measurement of angle be-
tween two vectors in R2 is done by dot product of vectors. For general space
of vectors we introduce a notion called inner product of vectors which exactly
catches the notion of angle between two vectors.
Definition 1.3.6. Let V be a space of vectors over R. A function h. , .i :
V × V −→ R is called an inner product if it satisfies the following properties:
IP-1 (Bilinearity): hα1 v̄1 + α2 v̄2 , v̄3 i = α1 hv̄1 , v̄3 i + α2 hv̄2 , v̄3 i for all α1 , α2 in
R and v̄1 , v̄2 , v̄3 in V.

lOMoARcPSD|5686828
IP-2 (Symmetricity): hv̄1 , v̄2 i = hv̄2 , v̄1 i for all v̄1 , v̄2 in V.
IP-3 (Positive-definiteness): hv̄, v̄i ≥ 0 for all v̄ in V; and hv̄, v̄i = 0 if and only
if v̄ = 0̄.
Example 1.3.7.
(I) Define h. , .i : R2 × R2 −→ R by h(x1 , y1 ) , (x2 , y2 )i = x1 x2 + y1 y2 for
all (x1 , y1 ), (x2 , y2 ) in R2 . This function is the well known dot product on R2 .
One can easily verify that the function h. , .i is an inner product on R2 .
(II) In more generality, the function h. , .i : Rm × Rm −→ R defined by hx̄ , ȳi =
x1 y1 + x2 y2 + · · · + xm ym for all x̄ = (x1 , x2 , · · · , xm ), ȳ = (y1 , y2 , · · · , ym ) in
Rm can be verified to be an inner product on Rm , which called the Euclidean
inner product.
(III) On the space of vectors V = C[0, 1], define h. , .i : V × V −→ R by hf , gi =
Z1
fg for all f, g in V. It is east to check that the function h. , .i is an inner
0
product on V, and is called the L2 inner product.
Note that thep property IP-1pof Definition 1.3.6 implies that hαf , αfi =
α2 hf , fi, i.e., hαf , αfi = |α| hf , fi for all f in V and for all α in R. Thus
looking at IP-1 and IP-3 of Definition 1.3.6 one may naturally ask whether inner
product leads to giving a norm in the corresponding space. We shall see that
the answer, in fact, is affirmative. To see this we shall observe some properties
of inner product.
Proposition 1.3.8 (Cauchy-Schwarz inequality). Let V be a space of vectors

and h. , .i : V × V −→ R an inner product on it. Then hf , gi2 ≤ hf , fihg , gi
for all f, g in V.
Proof. Let f, g in V. Then hλf + g , λf + gi = λ2 hf , fi + 2λhf , gi + hg , gi.

Since hλf + g , λf + gi ≥ 0 for all λ in R, we have hf , gi2 ≤ hf , fihg , gi.
Corollary 1.3.9. Let

pV be a space of vectors
p and h. p
, .i : V × V −→ R an inner
product on it. Then hf + g , f + gi ≤ hf , fi + hg , gi.
Proof. Let f, g in V. Then hf + g , f + gi = hf , fi + 2hf , gip + hg , gi. Using

Cauchy-Schwarz
p inequality
p we get hf + g , f + gip≤ hf , fi + 2 hf , fihg
p , gi +
2
hg
p , gi = ( hf , fi + hg , gi) . This proves hf + g , f + gi ≤ hf , fi +
hg , gi.
Now we are ready to show that an inner product on a space of vectors defines
a norm on the corresponding space.
Corollary 1.3.10. Let V be a space if vectors and h. , .i : V × V −→pR be an

inner product on it. Then the function k.k : V −→ R defined by kfk = hf , fi
for all f in V is a norm on V.
Proof. By the definition we see that kfk ≥ 0 for all f in V, and kαfk = |α|kfk
for all f in V. By corollary 1.3.9 we see that kf + gk ≤ kfk + kgk for all f, g in
V. This completes the proof.
Example 1.3.11.

lOMoARcPSD|5686828
1.4. TOPOLOGY OF THE EUCLIDEAN SPACE RM 7
(I) Let Rm be endowed with Euclidean inner product. Then it can be seen easily
that the corresponding norm is the Euclidean norm on Rm .
(II) The L2 inner product in C[0, 1] induces the L2 norm on the corresponding
space.
Example 1.3.12.
(I) Let x̄, ȳ ∈ R3 be such that the line segments 0̄, x̄ and 0̄, ȳ are perpendicular
to each other. Then one can check that the Euclidean inner product of x̄ and ȳ
is zero.
(II) Let x̄, ȳ ∈ R3 be such that the line segments 0̄, x̄ and 0̄, ȳ are parallel to
each other. Then one can check that the absolute value of the Euclidean inner
product of x̄ and ȳ is product of the norm of x̄ and ȳ.
In view of the above examples, for a general space of vectors we define the
following:
Definition 1.3.13. Let V be a space of vectors with an inner product h. , .i.
Two vectors v̄1 and v̄2 in V are called
(I) Orthogonal, if hv̄1 , v̄2 i = 0.
(II) Parallel, if |hv̄1 , v̄2 i| = kv̄1 kkv̄2 k.
In this chapter so far we have discussed on general space of vectors. But in the
rest of topics we shall mainly restrict our discussion in the space of vectors Rm ,
specially R2 and R3 , with Euclidean inner product. In the next section we shall
continue to explore the spatial structure of Rm .
1.4 Topology of the Euclidean Space Rm

Roughly speaking, topology is a spatial structure to study continuous functions
which preserves “nearness” of points. Thus topology, actually, meant to define
the nearness between elements in a set. In real analysis at first we have measured
this “nearness” in terms of distance, e.g., if x is a point in R then the points
which are near to x by distance less than δ are given by the set (x − δ, x + δ) =
{y ∈ R | |x − y| < δ}; and later without specifying the distance we defined the
“nearness” by open sets (or open neighborhoods) , e.g., the points near to x ∈ R
are given by points in open sets containing x; smaller the open sets nearer the
points to x. We shall buildup this concept of “nearness” for the space of vectors
Rm , and later through which we shall study continuous functions on Rm .
Definition 1.4.1. An open m-ball in Rm centered at x̄0 with radius r > 0 is
denoted by B(x̄0 , r) and is defined by
B(x̄0 , r) = { x̄ ∈ Rm | d(x̄0 , x̄) < r }
A closed n-ball in Rm centered at x̄0 with radius r > 0 is denoted by B(x̄0 , r)

and is defined by
B(x̄0 , r) = { x̄ ∈ Rm | d(x̄0 , x̄) ≤ r }
If r = 1, the n-balls are called unit n-balls and are, generally, denoted by
B(x̄0 ) and B(x̄0 ) for open and closed respectively.

lOMoARcPSD|5686828
Example 1.4.2.
(A) In R, the open interval (0, 1) is the open 1-ball with center 12 and radius 12 ;
i.e., (0, 1) = B( 12 , 21 ). In general, for a, b ∈ R where a 6= b we have (a, b) =
B( a+b a−b a+b a−b
2 , 2 ) and [a, b] = B( 2 , 2 ).
The unit open 1-ball, i.e, the unit open ball in R, is given by B(0) = (−1, 1);
and similarly B(0) = [−1, 1].
(B) The unit closed 2-ball is the circular disk { (x, y) ∈ R2 | x2 + y2 ≤ 1 } where
0̄ = (0, 0) ∈ R2 , the and unit open 2-ball B(0̄) is the “boundary-less” circular
disk { (x, y) ∈ R2 | x2 + y2 < 1 } where 0̄ = (0, 0) ∈ R2 .
(C) The closed 3-ball with center (x0 , y0 , z0 ) and radius r is the solid sphere
{ (x, y, z) ∈ R3 | (x − x0 )2 + (y − y0 )2 + (z − z0 )2 ≤ r2 }. The open 3-ball
B((x0 , y0 , z0 ), r) is the “boundary-less” solid sphere { (x, y, z) ∈ R3 | (x − x0 )2 +
(y − y0 )2 + (z − z0 )2 < r2 }.
Definition 1.4.3. A set U ⊆ Rm is called open if for each point x̄ ∈ U there

exists an open ball B(x̄, r) such that B(x̄, r) ⊆ U. A set C ⊆ Rm is called closed
if R\C is open.
Example 1.4.4.
(A) Any open ball B(x̄, r) in Rm is open, as for each ȳ ∈ B(x̄, r), we have
B(ȳ, r0 ) ⊆ B(x̄, r) where r0 = min {d(x̄, ȳ), r − d(x̄, ȳ)}.
(B) Any closed interval [a, b] is not open, for the points a and b in [a, b] there
exists no open 1-ball, i.e., open intervals, centered at a and centered at b, which
lies in [a, b]. Note that for a = b, we have [a, b] = {a}.
(C) The closed interval [a, b] is a closed set. To see this we need to observe
that R\[a, b] = (−∞, a) ∪ (b, ∞) is open. Suppose x ∈ R\[a, b]. Let r0 =
min {d(x, a), d(x, b)}. Then it is easy to show that B(x, r0 ) ⊆ R\[a, b]. Since
x ∈ R\[a, b] was arbitrary, we get that R\[a, b] is open.
Proposition 1.4.5.
1. Finite intersection of open sets is open.
2. Arbitrary union of open sets is open.
Proof. Exercise
Proposition 1.4.6.
1. Finite union of closed sets is closed.
2. Arbitrary intersection of closed sets is closed.
Proof. The proof is very similar to that of Proposition 1.4.5.
Corollary 1.4.7. Any finite set in Rm is closed.
Proof. Follows from Proposition 1.4.6.
Example 1.4.8. Arbitrary intersection of open sets may not be open. Set
m
\1/n) ⊂ R . Then each Um is an open n-ball and hence an open
Un := B(0̄,
set. But Un = {0̄}, which is closed.
n≥1

lOMoARcPSD|5686828
1.4. TOPOLOGY OF THE EUCLIDEAN SPACE RM 9
Definition 1.4.9. Any function f : Rm −→ R is called a scalar field, and

any function g : Rm −→ Rn is called a vector field. The vector field g can be
represented by n-many scalar fields, in the following fashion:
g = (g1 , g2 , · · · , gn ) where gi : Rm −→ R are scalar fields.
Example 1.4.10. 1. f(x, y) = x2 + y2 for all (x, y) ∈ R2 is a scalar field.
2. g(x, y) = (x2 , y2 , 1) for all (x, y) ∈ R2 is a vector field.
3. h(x, y, z) = (x2 , y + z) for all (x, y, z) ∈ R3 is a vector field.
Definition 1.4.11. Let G be an open set in Rm , and f : G −→ R a scalar field.
f is called continuous at ā if for any ǫ > 0 there exists δ > 0 such that whenever
x̄ ∈ B(ā, δ) (, i.e., ||x̄ − ā|| < δ, i.e., d(x̄, ā) < δ) we have f(x̄) ∈ B(f(ā), ǫ) (i.e.,
|f(x̄) − f(ā)| < ǫ, i.e., d(f(x̄), f(ā)) < ǫ). Equivalently, f is continuus at ā, if
for every sequence {x̄n } in G converging to ā, the corresponding sequence f(x̄n )
converges to f(ā).
A vector field F : G −→ Rn is called continuous at ā, if each of the scalar

component functions/fields of F are continuous.
In general continuous functions are called C0 -type functions.

Definition 1.4.12. Let G ⊂ Rm be open and f : G −→ R a scalar field. f is
called a Cn -type function at ā ∈ G if all the nth partial derivatives of f at ā
exists, and are continuous at ā. f is called smooth if f is C1 -type and the partial
derivatives of f do not vanish simultaneously at the corresponding point/points.
If the partial derivatives of f of all order exists, and are continuous, f is called
to be of C∞ -type.
Similar definitions for vector fields, as it is defined in Definition 1.4.11,

exist.
Proposition 1.4.13. (I) Let G ⊂ Rm be open and f1 , f2 , · · · , fs : G −→ R
be Cn -type at ā (n ≥ 0), then α1 f1 + α2 f2 + · · · + αs fs : G −→ R and
f1 f2 · · · fs : G −→ R are Cn -type at ā where αi are constants in R.
(II) If f2 6= 0 in G, then f1 /f2 is Cn -type at ā.
(III) Let V ⊂ Rm be open. Suppose f : G −→ Rt and g : V −→ Rs are Cn -

f g
type (n ≥ 0), then g ◦ f : G −→ Rs , if exists, is Cn -type (G −→ Rt ⊃ V −→ Rs ).
Proof. Exercise! Please note that g ◦ f will have domain G iff f(G) ⊂ V.
By repeated applications of Proposition 1.4.13 it is obvious that
Example 1.4.14. 1. f(x, y, z) = x + yz + z3 for all (x, y, z) ∈ R3 is contin-
uous throughout.
2. f(x, y) = 2 + x exp(y) + sin(x + y) for all (x, y) ∈ R2 is continuous on R2 .
p
3. f(x, y) = x/ x2 + y2 for all (x, y) ∈ R2 \(0, 0) is continuous on R2 \(0, 0).
4. f(x) = |x| is C0 -type on R, but not of C1 -type on R.

lOMoARcPSD|5686828
5. f(x) = x|x| is C1 -type on R, but not of C2 -type on R.

6. f(x, y) = x + exp(xy) is C∞ -type function.
7. Any Cn -type function is also a Cn−1 -type function.
1.5 Few Applications

1.5.1 Line passing though a point, and parallel to a vector
Let P0 ∈ Rm a point and v̄ a vector in Rm . It is well known that < v̄ >=:
Lv̄ (0̄) = {αv̄ ∈ Rm | α ∈ R} is a straight line passing though origin parallel to
v̄ (i.e., having direction as v̄). Suppose Lv̄ (P0 ) be the line passing though P0
and parallel to vector v̄. We shall show that Lv̄ (P0 ) = P0 + Lv̄ (0̄). It is easy to
check that for any w̄ ∈ Lv̄ (0̄), the point P0 + w̄ is in Lv̄ (P0 ). Which shows that
P0 + Lv̄ (0̄) ⊂ Lv̄ (P0 ). Next one may prove Lv̄ (P0 ) ⊂ P0 + Lv̄ (0̄) in the following
fashion: suppose P be any point in Lv̄ (P0 ). Then, P0 being a special point in
Lv̄ (P0 ), we have (P − P0 )||v̄, i.e., P = P0 + αv̄ for some α ∈ R. This shows that
Lv̄ (P0 ) ⊂ P0 + Lv̄ (0̄) = {P0 + αv̄ ∈ Rm | α ∈ R}. This completes the description
of Lv̄ (P0 ).
Another approach would be to see that any kind of translation of the line
Lv̄ (0̄) look like P∗ + Lv̄ ((0, 0)) for some P∗ ∈ Rm , and which is a straight line.
Note that this straight line is not a vector space as it does not contain 0̄. It is
easy to see that this straight line is passing though P∗ and is parallel to v̄, i.e.,
to the straight line Lv̄ (0̄); and therefore the straight line passing though P0 and
parallel to v̄ will be P0 + Lv̄ ((0, 0)).
As Lv̄ (P0 ) = P0 + Lv̄ (0̄) = {P0 + αv̄ ∈ Rm | α ∈ R}, the parametric rep-
resentation of the line Lv̄ (P0 ), when m = 3, i.e., in R3 would be (x, y, z) =
(x0 + αv1 , y0 + αv2 , z0 + αv3 ) where P0 = (x0 , y0 , z0 ) and v̄ = (v1 , v2 , v3 ).
1.5.2 Line passing through a two points

Let P1 and P2 be two points in Rm , and L(P1 , P2 ) be the line passing though
P1 and P2 . One may check that
L(P1 , P2 ) = P1 + LP2 −P1 (0̄) = LP2 −P1 (P1 ) = {P1 + α(P2 − P1 ) | α ∈ R}.
1.5.3 Plane passing through a point, and perpendicular to

a vector
Let v̄ be a vector in R3 , and P0 a point. Let Sv̄ (P0 ) denote the plane passing
though P0 and orthogonal to v̄. One may check that
Sv̄ (P0 ) = {P ∈ R3 | < P − P0 , v̄ >= 0}.

lOMoARcPSD|5686828
Chapter 2
Directional Derivative and

Differentiability
2.1 Introduction
Let G be an open set in R and f : G −→ R a real valued function of real variable.
f(x0 + h) − f(x0 )
We call f to be differentiable at x0 ∈ G if lim exists, and the
h→0 h
d ′ ′
limit is termed as dx (f(x))|x0 or f (x0 ). f (x0 ) represents the rate of change of
f along x-axis at the point x0 . We already have learned that if f is differentiable
at x0 , it also continuous at x0 .
If G be open in R2 and f : G −→ R a scalar field, then we defined the limits

f(x0 + h, y0 ) − f(x0 , y0 ) f(x0 , y0 + h) − f(x0 , y0 )
lim and lim , if exist, partial
h→0 h h→0 h
derivatives of f with respect to x and y respectively at (x0 , y0 ) ∈ G. We de-
∂ ∂
noted the value of the limit as ∂x (f)|(x0 ,y0 ) and ∂y (f)|(x0 ,y0 ) respectively; and
they represent the rate of change of f along x-axis and y-axis respectively. It is
well known that existence of partial derivatives of a function can not ensure the
continuity of the function. For example let f : R2 −→ R be defined by

x2 y
x4 +y2
if (x, y) 6= (0, 0)
f(x, y) =
0 if (x, y) = (0, 0)
∂ ∂
One may check that ∂x (f)|(0,0) and ∂y (f)|(0,0) exist, but f is not continuous
at (0, 0).
This tells us the concept of partial differentiability for several variable func-
tions is too weak to be defined as the equivalent concept of differentiability as
we have for real variable function.
We further have learned that a scalar field f : G −→ R is said to be differen-

tiable at P0 = (x0 , y0 ) ∈ G if there exists α, β ∈ R such that
f(x0 + h, y0 + k) − f(x0 , y0 ) − αh − βk
lim √ = 0.
(h,k)→(0,0) h2 + k2
∂ ∂
And, in that case it can be verified that α = ∂x (f)|P0 and β = ∂y (f)|P0 .
11

lOMoARcPSD|5686828
12CHAPTER 2. DIRECTIONAL DERIVATIVE AND DIFFERENTIABILITY
One should be careful that the mere existence of the partial deriva-
tives at P0 does not ensure differentiability of the given function. Fur-
ther, if the partial derivatives of the function at the point P0 do not
exists, the function is not differentiable at P0 . One may check that
that if f is differentiable at P0 , the it is continuous at P0 (Prove it!).

x + y if x = 0 or y = 0
Exercise 2.1.1. 1. Let f(x, y) = . Show that f
1 otherwise
has partial derivatives at (0, 0), but f is not continuous at (0, 0).
2. Show that the function f(x, y) = |x| + |y| for all (x, y) ∈ R2 is continuous,
but partial derivatives pf f at (0, 0) does not exists, and therefore f is not
differentiable at (0, 0).
3. Let f(x, y) = x2 + y2 for all (x, y) ∈ R2 . Show that partial derivatives of

f exists at any point of R2 . Is f differentiable at every point of R2 ?
p
4. Let f(x, y) = x2 + y2 for all (x, y) ∈ R2 . Show that partial derivatives
of f does not exists at (0, 0).
xy
x2 +y2
if (x, y) 6= (0, 0)
5. Let f(x, y) = . Show that f has partial deriva-
0 if (x, y) = (0, 0)
tives at (0, 0), but f is not continuous at (0, 0).
At this point we ask
Problem 2.1.2. 1. As partial differentiation represents the rate of change of

a function along respective axes, can we talk of rate of change of function
along any given straight line or, in general, along any given curve? If yes,
whether it has any relation with the partial derivatives.
2. Under what condition the partial differentiability can replace the concept
of differentiability of a scalar field?
In this chapter we mainly try to get the answer to the above questions.
2.2 Directional Derivative and Differentiability

∂
Let G be open in R3 , P0 ∈ G, and f : G −→ R a scalar field. Suppose ∂x (f)|P0
exists. If we look at the definition of partial derivatives carefully we shall see
that it is the rate of change of f along the line passing though P0 and parallel to
the x-axis, i.e, the vector e1 = (1, 0, 0). Note that the parametric form of line
Le1 (P0 ) is P0 + he1 where h ∈ R; and
∂ f(x0 + h, y0 , z0 ) − f(x0 , y0 , z0 )
∂x (f)|P0 = lim
h→0 h
f((x0 , y0 , z0 ) + t(1, 0, 0)) − f((x0 , y0 , z0 ))
= lim
t→0 t
f(P0 + te1 ) − f(P0 )
= lim
t→0 t

lOMoARcPSD|5686828
2.2. DIRECTIONAL DERIVATIVE AND DIFFERENTIABILITY 13
Imitating this we define the rate of change of f along any line passing though
P0 and parallel to a unit vector v̄ (i.e., ||v̄|| = 1), i.e., along the line Lv̄ (P0 ) as
f(P0 + tv̄) − f(P0 )
lim , if the limit exists. We define this quantity the direc-
t→0 t
tional derivative of f at P0 along the unit vector v̄ and is denoted by
f(P0 + tv̄) − f(P0 )
Dv̄ (f)|P0 , i.e., Dv̄ (f)|P0 = lim , if the limit exists.
t→0 t
d
Exercise 2.2.1. 1. Show that Dv̄ (f)|P0 = dt (f(P0 + tv̄))|t=0 .
2. For any direction v̄, find the directional derivative Dv̄ (f)|(0,0) where f(x, y) =
x2 + y2 .

x + y if x = 0 or y = 0
3. Let f(x, y) = . Find directions v̄ such that
1 otherwise
Dv̄ (f)|(0,0) exists.
p
4. Let f(x, y) = x2 + y2 for all (x, y) ∈ R2 . Show that Dv̄ (f)|(x0 ,y0 ) exists
for all direction v̄ if (x0 , y0 ) 6= (0, 0).
It is obvious that if DP0 (f)|v̄ exists for all direction v̄ (unit vector), then all
∂
the partial derivatives of f exists at P0 , as De1 (f)|P0 = ∂x (f)|P0 , De2 (f)|P0 =
∂ ∂
∂y (f)| P0
, and D e3
(f)| P0
= ∂z (f)| P0
. So, we ask
Problem 2.2.2. If all the partial derivatives of f exists at P0 , does Dv̄ (f)|P0
exit for all direction v̄?
This question comes naturally as we know that any vector v̄ can be written
as linear combinations of e1 , e2 and e3 in R3 , i.e., in R3 we can reach any point
v̄ from 0̄ by moving along x-axis, y-axis and z-axis.
Let P0 = (x0 , y0 , z0 ) and v̄ = (v1 , v2 , v3 ) where ||v̄|| = 1. Note that vi 6= 0

for i = 1, 2, 3 as ||v̄|| = 1. Now, Write f(P0 + tv̄) − f(P0 ) as
f(P0 + t(v1 , v2 , v3 )) − f(P0 ) = f(P0 + t(v1 , v2 , v3 )) − f(P0 + t(v1 , v2 , 0))+
f(P0 + t(v1 , v2 , 0)) − f(P0 + t(v1 , 0, 0))
+f(P0 + t(v1 , 0, 0)) − f(P0 + t(0, 0, 0)).
= f([P0 + t(v1 , v2 , 0)] + tv3 e1 ) − f(P0 + t(v1 , v2 , 0))+

f([P0 + t(v1 , 0, 0)] + tv2 e2 ) − f(P0 + t(v1 , 0, 0))
+f([P0 + t(0, 0, 0)] + tv1 e1 ) − f(P0 + t(0, 0, 0)).
And, therefore,
f(P0 + tv̄) − f(P0 )

lim
t→0 t
f(P0 + t(v1 , v2 , 0) + tv3 e3 ) − f(P0 + t(v1 , v2 , 0))
= lim +
t→0 t
f(P0 + t(v1 , 0, 0) + tv2 e2 ) − f(P0 + t(v1 , 0, 0))
lim +
t→0 t
f(P0 + t(0, 0, 0) + tv1 e1 ) − f(P0 + t(0, 0, 0))
lim
t→0 t
f(P0 + t(v1 , v2 , 0) + tv3 e3 ) − f(P0 + t(v1 , v2 , 0)) v3
= lim +
t→0 t v3

lOMoARcPSD|5686828
f(P0 + t(v1 , 0, 0) + tv2 e2 ) − f(P0 + t(v1 , 0, 0)) v2

lim +
t→0 t v2
f(P0 + t(0, 0, 0) + tv1 e1 ) − f(P0 + t(0, 0, 0)) v1
lim
t→0 t v1
Now, if partial derivative of f with respect to x exists at P0 , with respect
to y at P0 + t(v1 , 0, 0) and with respect to z at P0 + t(v1 , v2 , 0), then we can
calculate the above limit and we get the following.
f(P0 + tv̄) − f(P0 ) ∂ ∂ ∂
lim = v3 (f)|P0 +t(v1 ,v2 ,0) +v2 (f)|P0 +t(v1 ,0,0) +v1 (f)|P0 .
t→0 t ∂z ∂y ∂x
∂ ∂ ∂ ∂
Which will be equal to v3 ∂z (f)|P0 + v2 ∂y (f)|P0 + v1 ∂x (f)|P0 if ∂z (f)|(x,y,z)
∂
and ∂y (f)|(x,y,z) are continuous at P0 .
Therefore, we have the following:
Theorem 2.2.3. Let G be open in R3 , P0 ∈ G and f : G −→ R an scalar field.

Suppose partial derivatives of f exists in an open ball around P0 and, two of the
three partial derivatives are continuous at P0 , then for any unit vector v̄ in R2 ,
(I) Dv̄ (f)|P0 exists and
∂ ∂ ∂
(II) Dv̄ (f)|P0 =< ∇f|P0 , v̄ >= v1 ∂x (f)|P0 + v2 ∂y (f)|P0 + v3 ∂z (f)|P0 .
It is well known that if f is differentiable at P0 , then all the partial derivatives

exist at P0 . It is worth exploring whether Dv̄ (f)|P0 exists for all direction v̄ if f
is differentiable at P0 . For this we first revisit the definition of differentiability
of a real valued fucntion g : G −→ R where G ⊂ R is open. Suppose a ∈ G. We
say that g is differentiable at a if
g(x) − g(a)
lim exists. If this limit exists we call this value to be g ′ (a),
x→a x−a
g(x) − g(a)
i.e., g ′ (a) = lim . Now define a function g1 : G −→ R by
x→a x−a
f(x)−f(a)
if x 6= a
g1 (x) = x−a
α if x = a
Note that if we set α = g ′ (a), then g1 is continuous at x = a, as g is
differentiable at x = a. Now suppose it is not known whether g is differentiable
at a. If we assume that g1 is continuous at a, then it is easy to see that g is
differentiable at a, and α = f ′ (a). Thus, we have
Theorem 2.2.4 (Caratheodory’s Lemma). Let G ⊂ R be open, a ∈ G, and

f : G −→ R a function. Then, f is differentiable at a if and only if there exists
a function f1 : G −→ R such that
(I) f1 is continuous at a, and
(II) f(x) − f(a) = (x − a)f1 (x) for all x ∈ G.
f(x) − f(a) d
Specifically, f1 (x) = for x ∈ G\{a}; and f1 (a) = dx (f)|a = f ′ (a).
x−a
It can be seen that the above result also holds true for functions of several
variables.
Theorem 2.2.5 (Caratheodory). Let G ⊂ R3 be open, ā ∈ G, and f : G −→ R a

function. Then, f is differentiable at ā if and only if there exists three functions
f1 , f2 , f3 : G −→ R such that

lOMoARcPSD|5686828
(I) f1 , f2 and f3 are continuous at ā, and

(II) f(x, y, z) − f(a1 , a2 , a3 ) = (x − a1 )f1 (x, y, z) + (y − a2 )f2 (x, y, z) + (z −
a3 )f3 (x, y, z) for all (x, y, z) ∈ G.
∂ ∂
Further, it can be seen that f1 (ā) = ∂x (f)|ā , f2 (ā) = ∂y (f)|ā , and, f3 (ā) =
∂
∂z (f)|ā .
Proof. Assignment!
Now, we assume that f : G −→ R be differentiable at P0 where G ⊂ R3
is open and P0 ∈ G. Then by Theorem 2.2.5 there exists functions f1 , f2 , f3 :
G −→ R continuous at P0 such that
f(x, y, z)−f(x0 , y0 , z0 ) = (x−x0 )f1 (x, y, z)+(y−y0 )f2 (x, y, z)+(z−z0 )f3 (x, y, z)
for all (x, y, z) ∈ G. Now, let v̄ be any unit vector in R3 and we choose the
points (x, y, z) ∈ G such that they lie on the straight line P0 + tv̄ where t ∈ R.
Then we get
f(P0 +tv̄)−f(P0 ) = (x0 +tv1 −x0 )f1 (P0 +tv̄)+(y0 +tv2 −y0 )f2 (P0 +tv̄)+(z0 +tv3 −z0 )f3 (P0 +tv̄)
for all t ∈ R, i.e.,
f(P0 + tv̄) − f(P0 ) (x0 + tv1 − x0 )f1 (P0 + tv̄) (y0 + tv2 − y0 )f2 (P0 + tv̄) (z0 + tv3 − z0 )f3 (P0 + tv̄)
= + +
t t t t
for all t ∈ R\{0}.
Since f1 , f2 and f3 are continuous at P0 , as t −→ 0, we see the limits in the

right side exists and therefore, the limit in the left side exists, i.e., Dv̄ (f)|P0 exist.
Since v̄ was arbitrary, we see that Dv̄ (f)|P0 exists for all direction v̄. Further,
from the above equality we get, under limit,
∂ ∂ ∂
Dv̄ (f)|P0 = v1 (f)|P0 + v2 (f)|P0 + v3 (f)|P0 =< ∇f|P0 , v̄ > .
∂x ∂y ∂z
From this we conclude the following
Theorem 2.2.6. Let G ⊂ R3 be open, P0 ∈ G and f : G −→ R differentiable at
P0 , then
(I) Dv̄ (f)|P0 exists for all direction v̄, and
(II) Dv̄ (f)|P0 =< ∇f|P0 , v̄ > .
Now, if we follow the statements of Theorem 2.2.3 and Theorem 2.2.6, we
shall see that two different “hypotheses” gives the same result. So, naturally
one asks whether those two “hypotheses” in Theorem 2.2.3 and Theorem 2.2.6
have any relation. Clearly, mere differentiability of a function can not ensure
continuity of partial derivatives or even the existence of partial derivatives in
neighbourhood points. So, we should check whether the “conditions” of partial
differentiations in Theorem 2.2.3 ensure differentiability of the function in the

lOMoARcPSD|5686828
corresponding point. The answer is, in fact, “Yes”! From the technique applied
in Theorem 2.2.3, and the technique of forming the increment function f1 in
Theorem 2.2.4 we see a scope/ we sense our intuition of proving differentiability
of a scalar field under the condition the partial derivatives are continuous.
Let G ⊂ Rn be open and P0 = (x0 , y0 , z0 ) ∈ G. Suppose f : G −→ R a
scalar field. Then,
f(x, y, z) − f(x0 , y0 , z0 ) = f(x, y, z) − f(x, y, z0 )+
f(x, y, z0 ) − f(x, y0 , z0 )+
f(x, y0 , z0 ) − f(x, y, z)
Therefore, if we define
f(x,y,z)−f(x,y,z0 ) f(x,y,z0 )−f(x,y0 ,z0 )
x−x0 if z 6
= z0 y−y0 if y 6= y0
f1 (x, y, z) = ∂ f2 (x, y, z) = ∂
∂z (f)| (x,y,z0 ) if z = z0 ∂y (f)| (x,y ,z ) if y = y0
0 0
f(x,y0 ,z0 )−f(x0 ,y0 ,z0 )

if x 6= x0 ∂ ∂
f3 (x, y, z) = ∂
x−x0 , provided ∂z (f)|(x,y,z0 ) , ∂y (f)|(x,y0 ,z0 )
∂x (f)| (x0 ,y ,z
0 0 ) if x = x 0
∂
and ∂x (f)|(x0 ,y0 ,z0 ) exist.
∂ ∂
If we further assume that ∂z (f)|(x,y,z) , and ∂y (f)|(x,y,z) are continuous at
P0 , then we see that the functions f1 , f2 and f3 are continuous at P0 . And, it
is easy to check that
f(x, y, z)−f(x0 , y0 , z0 ) = (z−z0 )f1 (x, y, z)+(y−y0 )f2 (x, y, z)+(x−x0 )f1 (x, y, z)
for all (x, y, z) ∈ G. Therefore, by Theorem 2.2.5, we have f is differentiable at

P0 . This observation leads to the following
Theorem 2.2.7. Let G ⊂ R3 be open and P0 ∈ G. Suppose f : G −→ R a scalar

field such that all the partial derivatives exists in an open ball around P0 , and
two of the three partial derivatives are continuous at P0 , then f is differentiable
at P0 .
Thus, in view of Theorem 2.2.7 and Theorem 2.2.6, we see that Theorem
2.2.3 follows as a corollary; and combining those two results we can write the
following for G ⊂ R3 open, P0 ∈ G and f : G −→ R:
Remark 2.2.8.

A1 : all the partial derivatives of f exist in an open ball araound P0
A:
A2 : two of the three partail derivatibes of fare continuous at P0
⇓

B: f is differentiable at P0
⇓

C1 : Dv̄ (f)|P0 exists for all direction v̄
C:
C2 : Dv̄ (f)|P0 =< ∇f|P0 , v̄ >
Since A ⇒ B ⇒ C follows, contrapositively, we have (∼ C) ⇒ (∼ B) ⇒ (∼ A).

These results will be useful to prove or disprove a scalar field is differentiable.

lOMoARcPSD|5686828
Example 2.2.9. 1. Let f(x, y) = exp(x + y) for all (x, y) ∈ R2 . Since f is

a composition of C∞ -type function, f is a C∞ -type function on R2 (and
therefore it satisfies all the conditions of “A” in Remark 2.2.8 or the hy-
pothesis of Theorem 2.2.7), we see that f is differentiable on R2 .
p
2. Let f(x, y) = x2 + y2 for all (x, y) ∈ R2 . It can be checked that the
partial derivatives of f do not exist at (0, 0), and therefore (“C1” fails to
hold in Remark 2.2.8) f is not differentiable at (0, 0).

x2 y
if (x, y) 6= (0, 0)
3. Let f(x, y) = x4 +y2 . It can be shown that for any
0 if (x, y) = (0, 0)
direction v̄, Dv̄ (f)|(0,0) exits; but Dv̄ (f)|(0,0) 6=< ∇(f)|(0,0) , v̄ > (i.e., “C2”
of Remark 2.2.8 fails to hold) and therefore f is not differentiable at (0, 0).
4. Let f(x, y) = |x| + |y| for all (x, y) ∈ R2 . It can be seen that Dv̄ (f)|(0,0)
does not exist for any direction v̄, and therefore f is not differentiable at
(0, 0).
2 2
x y
if (x, y) 6= (0, 0)
Exercise 2.2.10. 1. f(x, y) = x4 +y2 . Show that f is
0 if (x, y) = (0, 0)
differentiable at (0, 0), though fy is not continuous at (0, 0).
2. Let f(x, y) = |xy| for all (x, y ∈ R2 ). Show that f is differentiable at (0, 0).
p
3. let f(x, y) = |xy| for all (x, y) ∈ R2 . Show that f is not differentiable at
(0, 0).

x3 y
if (x, y) 6= (0, 0)
4. f(x, y) = x4 +y2 . Show that f fails to be differen-
0 if (x, y) = (0, 0)
tiable at (0, 0), though Dv̄ (f)|(0,0) =< ∇(f)|(0,0) , v̄ > for any direction
v̄.

x2 y
if (x, y) 6= (0, 0)
5. f(x, y) = x2 +y2 . Show that f is continuous at (0, 0);
0 if (x, y) = (0, 0)
Dv̄ (f)|P0 exists for all direction v̄; but f is not differentiable at (0, 0).
Remark 2.2.11. Let G ⊂ R2 be open, P0 ∈ G, and f : G −→ R a scalar field.

Let f be such that for any direction v̄ we have
Dv̄ (f)|P0 =< ∇(f)|P0 , v̄ >= ||∇(f)|P0 || ||v̄|| cos(θ) = ||∇(f)|P0 || cos(θ).
Where θ is the angle inscribed by ∇(f)|P0 and v̄ with 0̄. Then we have
∇(f)|P0
1. Dv̄ (f)|P0 is maximum if cos(θ) = 1, i.e., if v̄ = ||∇(f)|P0 || .
∇(f)|
2. Dv̄ (f)|P0 is mimimum if cos(θ) = −1, i.e., v̄ = − ||∇(f)|PP0 || .
0
∂ ∂
3. Dv̄ (f)|P0 = 0 if cos(θ) = π/2, i.e., v̄ = ±( ∂y (f)|P0 , − ∂x (f)|P0 ).

lOMoARcPSD|5686828
2.3 Application: Level sets

Let U ⊂ Rn , P0 ∈ U, and f : U −→ R a scalar field. For each c ∈ R, the level
set of f of the level c is donoted by Sc (f) (or by Lc (f)), and is defined by
Sc (f) := {P ∈ U | f(P) = c}.
When n = 2, we call level sets to be level curves, as geometrically those

represent a curves; and when n = 3, we call level sets to be level surfaces as
geometrically those represent surfaces.
Example 2.3.1. 1. Let f(x, y) = x2 + y2 for all (x, y) ∈ R2 . Then for each
c > 0, the level curve Sc (f) is the circle x2 + y2 = c.
2. Let f(x, y, z) = x2 + y2 + z2 for all (x, y, z) ∈ R3 . Then for each c > 0,
the level surface Sc (f) is the sphere x2 + y2 + z2 = c.
Let n = 3, and suppose that the scalar field f is smooth. Consider the
level surface Sc (f). Since Sc (f) is a surface, it is union of curves. Let γ(t) =
(x(t), y(t), z(t)) for all t ∈ I ⊂ R be any smooth curve on Sc (f) passing though
the point P0 . Then f(γ(t)) = c. From which, on differentiating with respect to
t, we get
∂ d ∂ d ∂ d
(f)|P0 (x)|t0 + (f)|P0 (y)|t0 + (f)|P0 (z)|t0 = 0 , i.e.,
∂x dt ∂y dt ∂z dt
< ∇(f)|P0 , γ ′ (t)|t0 >= 0,

where γ(t0 ) = P0 , γ ′ (t)|t=t0 = (x ′ (t), y ′ (t), z ′ (t))t=t0 is tangent to the curve γ
at the point P0 .
This shows that ∇(f)|P0 is normal to the curve γ at the point P0 . Since
P0 is arbitrary, we see that at any point, the gradient of a scalar field is
normal to its corresponding level set at that point, i.e., ∇(f)|P0 ⊥ Sc (f)
at P0 .
Exercise 2.3.2. 1. Draw the level surfaces of f(x, y, z) = x2 + y2 − z2 for
all (x, y, z) ∈ R3 .
2. Draw the level curves of f(x, y) = x2 − y2 for all (x, y) ∈ R2 .
2 2
3. Let h(x, y) = 2e−x + e−3y denote the height on a mountain at position
(x, y). In what direction from (1, 0) should one begin walking in order to
climb the fastest?
4. Find a unit normal vector to √ the√ following surfaces at the specified point
(i) x2 + y2 + z2 = 9 at (0, 3, 3), (ii) x3 y3 + y − z + 2 = 0 at (0, 0, 2),
(iii) z = 1/(x2 + y2 ) at (1, 1, 1/2).
5. Suppose that a √ particle is ejected from the surface x2 + y2 − z2 = −1 at
the point (1, 1, 3) along the normal directed toward the xy plane to the
surface at time t = 0 with a speed of 10 units per second. When and where
does it cross the xy plane?

lOMoARcPSD|5686828
2.3. APPLICATION: LEVEL SETS 19
6. Starting from (1, 1), in which direction should one travel in order to obtain
the most rapid rate of decrease of the function f : R2 −→ R defined by
f(x, y) := (x + y − 2)2 + (3x − y − 6)2 ?
p
7. About how much will the function f(x, y) := ln x2 + y2 change if the
point (x, y) is moved from (3, 4) a distance 0.1 unit straight toward (3, 6)?

lOMoARcPSD|5686828

lOMoARcPSD|5686828
Chapter 3
Arc Length Function
3.1 Introduction
If L is a straight line segment (finite) or a finite union of straight line segments in
R3 , it is well known to us how to measure the length of the segment or totality
of of the length of the segments. In this chapter we shall drive a tool to measure
the length of a segment of a parametric curve.
3.2 Parametric Curves

Definition 3.2.1. Let I be an interval. Any continuous function γ : I −→ R3
is called a parametric curve. Note that for each t ∈ I, we have γ(t) =
(x(t), y(t), z(t)) ∈ R3 . We shall use the notation {γ} to denote the trace of γ,
and is defined by {γ} = {γ(t) | t ∈ I} ⊂ R3 along with an ordering induced from
the ordering in I (i.e., if is the induced ordering on {γ}, then is defined
by γ(t1 ) γ(t2 ) iff t1 ≤ t2 ). This ordering on the points of γ(I) is called
orientation of γ. If I = [a, b], then γ(a) is called the initial point of γ,
and γ(b) is called the final point of γ. If initial point and final point of a
parametric curve are the same point, we call the curve to be a loop/closed
curve. Let C ⊂ R3 be a set of ordered points. If C can be seen as a trace of a
parametric curve say, δ : I −→ R3 , we say that C is a parametric curve with a
parametrisation δ, and write C = {δ : I −→ R3 }.
The following example will clarify the terminologies.
Example 3.2.2. 1. Let γ1 : [0, 1] −→ R2 be defined by γ1 (t) = exp(2πit)
for all t ∈ [0, 1].
2. Let γ2 : [0, 2π] −→ R2 be defined by γ1 (t) = exp(it) for all t ∈ [0, 2π].
In the above example it can be seen that though γ1 6= γ2 , we see that
{γ1 (t) | t ∈ [0, 1]} = {γ2 (t) | t ∈ [0, 2π]} = {(x, y) | x2 + y2 = 1} = C, say. Then
C is a trace of a parametric curve which has, at least, two parametrisation. In
this case we simply say that C is a parametric curve.
Remark 3.2.3. A parametric curve γ : [a, b] −→ R3 can be viewed as a
weighted line segment; i.e., each point x of the segment [a, b] is given a weightage
γ(x).
21

lOMoARcPSD|5686828
22 CHAPTER 3. ARC LENGTH FUNCTION
Now we shall show that a parametric curve has infinitely many parametrisa-
tion. Suppose [a, b] be an intervals in R. Define a maps Φ[a,b] : [a, b] −→ [0, 1],
and Ψ[a,b] : [0, 1] −→ [a, b] by
b−t
Φ[a,b] (t) = for all t ∈ [0, 1] (this function scales the interval [a, b] onto
b−a
[0, 1])
and
Ψ[a,b] (s) = bs + a(1 − s) for all s ∈ [0, 1] (this function scales the interval [0, 1]
onto [a, b]).
Note that Φ[a,b] (a) = 0, Φ[a,b] (b) = 1, Ψ[a,b] (0) = a, Ψ[a,b] (1) = b,
Φ[a,b] ◦ Ψ[a,b] (s) = s for all s ∈ [0, 1], Ψ[a,b] ◦ Φ[a,b] (t) = t for all t ∈ [a, b],
(i.e., Φ[a,b] = Ψ−1 −1
[a,b] , and Ψ[a,b] = Φ[a,b] ) and both the functions are infinitely
many differentiable . Now, for given any pair of interval [a, b] and [c, d], we
[a,b]
shall define a map T[c,d] : [c, d] −→ [a, b], using these properties of Φ and Ψ as
follows
Φ[c,d] Ψ[a,b]
[c, d] −→−→−→ [0, 1] −→−→−→ [a, b], i.e.,
[a,b]
T[c,d] (t) = Ψ[a,b] ◦ Φ[c,d] (t) for all t ∈ [c, d]. Now for any parametric
[a,b]
curve γ : [a, b] −→ R3 , consider the map γ ◦ T[c,d] : [c, d] −→ R3 . Check
[a,b] [a,b]
that {γ} = {γ ◦ T[c,d] }, and therefore both γ and γ ◦ T[c,d] represent the same
parametric curve. If we assume that {γ} = C, then the above discussion shows
that C has infinitely many parametrisation. The developed theory will show
that it is enough to consider one favorable parametrisation of a parametric
curve, and study the topics under discussion.
Definition 3.2.4. We call a parametric curve to be Cn -type/C∞ -type/smooth

(n ≥ 1) if it has a parametrisation which is Cn /C∞ /smooth. If the domain of
the curve is a closed interval, then, in this case, it will mean that the curve
has an extension in an open interval containing the closed interval and we are
taking about Cn -ness/C∞ -ness/smoothness of the extension of the curve without
mentioning the extension.
Let [a, b] ⊂ R. Define a map id : [a, b] −→ [a, b] by id(x) = a + b − x

for all x ∈ [a, b]. Note that the function id reverses the order of the elements
of the interval [a, b]; and therefore, for any parametric curve γ : [a, b] −→ R3 ,
the function γ ◦ id : [a, b] −→ R3 represents again a parametric curve such
that γ([a, b]) = γ ◦ id([a, b]), and the ordering of {γ ◦ id} is the reverse of the
ordering of {γ}, i.e., γ and γ ◦ id represents the same curve (set of points), but
the orientation of the curves are opposite to each other; specifically, γ ◦ id(t) =
γ(a + b − t) for all t ∈ [a, b].
Definition 3.2.5. For any parametric curve γ : [a, b] −→ R3 , the corresponding

parametric curve γ−1 : [a, b] −→ R3 defined by γ−1 (t) = γ ◦ id(t) = γ(a + b − t)
for all t ∈ [a, b] is called the inverse of γ.

lOMoARcPSD|5686828
3.3. ARC LENGH OF PARAMETRIC CURVES 23
Exercise 3.2.6. 1. Show that if a parametric curve is Cn -type/C∞ -type/smooth,

then the inverse of the parametric curve is also Cn -type/C∞ -type/smooth
(n ≥ 0).
[a,b] [a,b]
2. Show that τ[c,d] is a smooth function, and conclude that γ◦τ[c,d] is smooth
if γ is smooth.
Definition 3.2.7. Let γ1 : [a, b] −→ R3 and γ2 : [c, d] −→ R3 be two para-

metric curves such that the final point of γ1 is the initial point of γ2 , i.e.,
γ1 (b) = γ2 (c), then the composition of γ1 and γ2 , denoted by γ1 ∗ γ2 is
defined by
 [a,b] 1
 γ1 ◦ T[0, 1 ] (t) if t ∈ [0, 2 ]

2
γ1 ∗ γ2 (t) =

 γ ◦ T [c,d] (t) if t ∈ [ 1 , 1]
2 [ 1 ,1] 2
2
Note that γ1 ◦ γ2 is also a parametric curve.
3.3 Arc Lengh of Parametric Curves

Let γ : [a, b] −→ R3 be a curve and we wish to calculate the length of the curve,
if possible. For this we imitate the theory of Riemann integration to calculate
area under a curve defined in an finite interval. Let P([a, b]) be the set of all
partitions of [a, b]. We first take a partition P : {a = t0 , t1 , t2 , · · · , tn = b} from
P([a, b]) and, though linear approximation find an approximate value ℓP (γ) of
the length of the curve γ for the chosen partition P as
n
X
ℓP (γ) = ||γ(ti ) − γ(ti−1 )||.
i=1
If lim ℓP (γ) exists we say that length of γ exists, and the value of the limit is
n→∞
defined as the length of γ, and is denoted by ℓ(γ); i.e., ℓ(γ) = lim ℓP (γ), if the
n→∞
limit exists.
If we let ||P|| := max |tn − tn−1 |, we see that lim ℓP (γ) = lim ℓP (γ).
n n→∞ ||P||→0
Let Peq ([a, b]) = {P ∈ P([a, b]) | points in P are equidistant}. As can be
seen that lim ℓP (γ) = lim {ℓP (γ) | P ∈ Peq ([a, b])}, for the context under
||P||→0 ||P||→0
discussion it is enough to take partitions of [a, b] from Peq ([a, b]). Note that
any partition in Peq ([a, b]) is uniquely determined by the number of points in
it; and therefore, instead of notation ℓP (γ) it is enough to use the notation
ℓn (γ) to indicate that we have taken a partition from Peq ([a, b]) having n + 1
equidistant points. Thus, we shall write ℓ(γ) = lim ℓn (γ), if the limit exists.
n→∞
Let P ∈ Peq ([a, b]) with n + 1 points. Assume that ||P|| = ∆t. Clearly, it
implies that for any two consecutive points ti and ti+1 from P we have ti+1 =
ti + ∆t. We see that lim ℓP (γ) = lim ℓn (γ) = lim ℓn (γ). If we write
n→∞ ||P||→0 ∆t→0
γ(t) = (x(t), y(t), z(t)), then we see that

lOMoARcPSD|5686828
n
X q
lim ℓn (γ) = lim (x(ti ) − x(ti−1 ))2 + (y(ti ) − y(ti−1 ))2 + (z(ti ) − z(ti−1 ))2
∆t→0 ∆t→0
i=1
Xn p
(x(ti ) − x(ti−1 ))2 + (y(ti ) − y(ti−1 ))2 + (z(ti ) − z(ti−1 ))2
= lim ∆t
∆t→0 ∆t
i=1
Xn
r
(x(ti ) − x(ti−1 ))2 + (y(ti ) − y(ti−1 ))2 + (z(ti ) − z(ti−1 ))2
= lim ∆t
∆t→0 ∆t2
i=1 s
Xn 2 2 2
x(ti ) − x(ti−1 ) y(ti ) − y(ti−1 ) z(ti ) − z(ti−1 )
= lim + + ∆t
∆t→0 ∆t ∆t ∆t
i=1 s
Xn 2 2 2
d d d
= lim (x(t))|ti−1 + (y(t))|ti−1 + (z(t))|ti−1 ∆t
∆t→0 dt dt dt
i=1
if γ is differentiable, i.e., if γ ′ (t) exists
Xn
= lim ||γ ′ (t)||ti−1 ∆t
∆t→0
Zb i=1
= ||γ ′ (t)||dt
a
if γ ′ is continuous, i.e., if γ is C1 -type
Thus, we get the following theorem
Theorem 3.3.1. Let γ : [a, b] −→ R3 be a C1 -type curve. Then, ℓ(γ) exists,

Zb
and ℓ(γ) = ||γ ′ ||.
a
Exercise 3.3.2. 1. Let γ : [a, b] −→ R3 be a C1 -type curve. Show that

[a,b] [a,b]
γ ◦ T[c,d] : [c, d] −→ R3 is also C1 -type curve; and ℓ(γ) = ℓ(γ ◦ T[c,d] ).
2.
3.4 Arc Length Function of Parametric Curves

Let γ : [a, b] −→ R3 be a C1 -type curve. It is well know that the length of
Zb
the curve γ from γ(a) to γ(b) is given by ℓ(γ) = ||γ ′ ||; and therefore for any
a Zx
x ∈ [a, b], the length of the curve γ from γ(a) ot γ(x) is give by ||γ ′ ||. This
Zx a
defines a map s : [a, b] −→ R by s(x) = ||γ ′ ||. This function is called the arc
a
length function of the curve γ. We shall study the properties of this function
after we observe few examples.
Example 3.4.1. Let γ(t) = (cos(2πt), sin(2πt)) for all t ∈ [0, 1]. Clearly γ
is a C1 -type curve, and ||γ ′ (t)|| = ||(−2π sin(2πt), 2π cos(2πt))|| = 2π for all
t ∈ [0, 1]. Therefore, the arc length function of γ is given by

lOMoARcPSD|5686828
3.4. ARC LENGTH FUNCTION OF PARAMETRIC CURVES 25

Zx
s(x) = ||γ(t)||dt
0
Zx
= 2πdt
0
= 2πx
We know that the curve {γ} has another parametrisation δ : [0, 2π] −→ R3
defined by δ(t) = (cos(t), sin(t)) for all t ∈ [0, 2π]. For this parametrisation it
can be seen that s(x) = x for all x ∈ [0, 2π].
Remark 3.4.2. The above example clearly shows that the different parametri-
sation to the same curve gives different arc length function, i.e., So, we ask
for a given parametric curve C whether there is any parametrisation η defined
on an interval I for which s(x) = x for all x ∈ I. Note that in that case, for
that special paramerisation,
Z say η, we shall get ||η ′ (t)||
Z = 1 for all t ∈ I. Sup-
x
pose I = [c, d]. Since ||η ′ (t)||dt = ℓ(C), and s(x) = ||η ′ (t)||dt it is clear that
I c
c = 0 and d = ℓ(C). Since the arc length function s is a surjective function from
the domain of the parametrised curve C to the interval [0, ℓ(C)], we naturally
ask whether the function s−1 can be used to convert domain of the parametric
curve to [0, ℓ(C)]. The following properties will help us to get the answer to this
question.
Before going into the properties let us recall a fact: “Any real valued
bijective continuous function defined on a closed interval has contin-
uous inverse. Further, if the function is smooth, the inverse is also
smooth.” (Prove it!!!)
Theorem 3.4.3. Let s : [a, b] −→ R be the arc length function of the C1 -type
curve γ : [a, b] −→ R3 . Then the following holds
(I) s(x) ≥ 0 for all x ∈ [a, b], s(a) = 0, and s(b) = ℓ(γ).
(II) s is non-decreasing on [a, b].
(III) s 6≡ 0 if and only γ ′ (t) 6= 0̄ for some t ∈ [a, b], i.e., s ≡ 0 if and only if
γ ′ = 0̄ on [a, b].
(IV) If γ ′ (t) 6= 0̄ for all t ∈ [a, b] (i.e., γ is smooth), then s(x) = 0 if and only
if x = a.
(V) If γ ′ (t) 6= 0̄ for all t ∈ [a, b] (i.e., γ is smooth), then s is strictly increasing
on [a, b].
(VI) s is continuous on [a, b].
(VII) s is differentiable on [a, b].
d
(VIII) s is C1 -type on [a, b], and dx s(x) = ||γ ′ (x)|| for all x ∈ [a, b].
(IX) s is smooth if γ is smooth.
Proof. Assignment!

lOMoARcPSD|5686828
Definition 3.4.4. Suppose s : [a, b] −→ R be the arc length function of the

C1 -type curve γ : [a, b] −→ R3 . Then due to (VIII) of Theorem 3.4.3 we write
Zb Zb Z ℓ(γ)
||γ ′ (t)||dt as ds or ds; as if ds = ||γ ′ (t)||dt.
a a 0
Zb Zb
Remark 3.4.5. It is well known that dx = 1 dx = the length of the line
a a
segment (a, 0), (b, 0) whose parametric equations is give by x̄(t) = (t, 0) for all
t ∈ [a, b]. Note that in this case the arc length function of the curve x̄ is given
d
by s(x) = x − a for all x ∈ [a, b], and s ′ (x) = 1 = dx (x) for all x ∈ [a, b]. Now,
Zb
for any parametric curve γ : [a, b] −→ R, the length of the curve is ||γ ′ (t)||dt
Zb a
which does not look like 1 dx, rather it is weighted, i.e., instead of the fixed
a
unit weight 1 at each point t, a weight ||γ ′ (t)|| being carried by each point t. So,
one may ask what should replace “dx”, and which will look like “dx” so that
Zb
d
‘‘dx ′′ = length of the curve. Now, recall that if dx (f(x)) = g(x), we write it
a Z
in terms of language of integration/ symbol of integration as g(x)dx = f(x) or
Z Z Z
′
f (x)dx = f(x), or sometimes in short d(f(x)) = f(x), or simply df = f;
and therefore, in the light of (VIII) it is clear that “dx” should be replaced by
Z Zb
“ds” so that ds = s, and hence ds = s(b) − s(a) = length of the curve.
a
′
Further, if γ (t) 6= 0̄ for all t ∈ [a, b], clearly, by (V)th property we see that
s has its inverse s1 : [0, ℓ(γ)] −→ [a, b], and (VIII)th property ensures C1 -type-
ness of it. So, we consider the C1 -type curve γ ◦ s−1 : [0, ℓ(γ)] −→ R3 . It can be
d
checked that {γ} = {γ ◦ s−1 }, and || dt (γ ◦ s−1 (t))|| = 1, and therefore for γ ◦ s−1
Zx Zx
d −1
the arc length function is s(x) = || (γ ◦ s (t))||dt = 1 dt = x for all
0 dt 0
−1
x ∈ [0, ℓ(γ)]. The parametrisation γ◦s is called parametrisation of the
curve γ by its arc length function, or the arc-length parametrisation
of the curve γ.
Exercise 3.4.6. 1. pareametrise the curve C : x2 = y3 , y = z2 , z ≥ 0, x ≥ 0

and thereby find the arc length function of the curve C.
2. Let γ(t) = (a cos(t), a sin(t), bt) for all t ∈ [0, 2π]. Find arc length
function of the curve γ; and also parametrize the curve by arc length
function, if possible.
3. Show that for a C1 -type parametric curve γ on a finite interval, we have

ℓ(γ) = ℓ(γ−1 ).
4. Show that for any smooth increasing function φ : [c, d] −→ [a, b] and
for any C1 -type curve γ : [a, b] −→ R3 , we have {γ ◦ φ} = {γ}, and
ℓ(γ) = ℓ(γ ◦ φ).

lOMoARcPSD|5686828
Chapter 4
Line Integration
4.1 Introduction
In this chapter we shall study theory of integration of vector fields along a curve
aiming to classify a certain kind of vector fields called conservative vector fields.
4.2 Line Integration of Scalar Fields

Let G ⊂ R3 and f : G −→ R3 a scalar field. Suppose γ : [a, b] −→ R3 is a
parametric curve such that Im(γ) ⊂ G. We define line integration of f along the
n
X n
X
length of γ as the limit lim f(γ(ξi )) γ(ti ), γ(ti−1 ) = lim f(γ(ξi )) ||γ(ti )−
n→∞ n→∞
i=1 i=1
γ(ti−1 )||, if exists; where P = {a = t0 < t1 < · · · < tn = b} is an equidistant
partition of [a,Zb] and ξZ i is an arbitrary point in [ti−1 , ti ]; and the limiting value
is denoted by f or f ds. As in the previous chapter, one may check that
γ γ
the limit will exists if f is continuous, and γ is C1 -type, which we state as a
theorem below.
Theorem 4.2.1. Let G ⊂ R3 and f : G −→ R3 a continuous Zscalar field and
γ : [a, b] −→ R3 a C1 -type curve such that Im(γ) ⊂ G. Then f exists, and
γ
Z Zb Zb
f= f(γ(t)) ||γ ′ (t)|| dt = f(γ(t)) ds.
γ a a
Exercise 4.2.2. Let G ⊂ R and f : G −→ R3 a continuous scalar field and

3
γ : [a, b] −→ R3 a C1 -type curve such that Im(γ) ⊂ G

Z Z
1. Show that f= f.
γ γ−1
2. If
Z φ : Z[c, d] −→ [a, Z b] is a smooth increasing function, then show that
f= f, i.e., f is independent of parametrisation of γ.
γ γ◦φ γ
Remark 4.2.3. Let γ be a piece-wise Cr -type/smooth (r ≥ 0) parametric curve,

i.e., γ is a Cr -type curve expect finitely many points, say m-many points, then
27

lOMoARcPSD|5686828
28 CHAPTER 4. LINE INTEGRATION
clearly γ can be realized as a union (or addition) of m + 1 parametric curves

which are Cr -type/smooth, i.e., there exists γ1 , γ2 , · · · , γm+1 Cm -type/smooth
parametric curves such that γ = γ1 ∗ γ2 ∗ · · · ∗ γm+1 .
Now, we shall define line integration of a continuous scalar field along the
length of a piece-wise C1 -type curve.
Definition 4.2.4. Let G ⊂ R3 , f : G −→ R3 a scalar field, and γ a piecewise
continuous parametric curve, and suppose γ = γ1 ∗ γ2 ∗ · · · ∗ γs where each γi
Z Xs Z
is continuous. Then we define f= f.
γ i=1 γi
4.3 Line Integration of Vector Fields

Let G ⊂ R3 and F̄ : G −→ R3 a vector field. Suppose γ : [a, b] −→ R3 a
smooth curve such that Im(γ) ⊂ G. Since γ is smooth, at each point of γ
γ ′ (t)
we have a unit tangent T̄ (t) defined by T̄ (t) = . We define the line
||γ ′ (t)||
integration of the vector field F̄ along γ by the line integrationZ of the scalarZ field
< F̄(γ(t)), T̄ (t) > along the length of γ; and denoted by F̄ or by F̄.ds̄,
Z Z γ γ
i.e., F̄.ds̄ = < F̄(γ(t)), T̄ (t) > ds. Note that if F̄ is continuous, then the
γ γ
function < F̄(γ(t)), T̄ (t) > is also continuous, and therefore by Theorem 4.2.1,
we have
Z Z
F̄.ds̄ = < F̄(γ(t)), T̄ (t) > ds
γ γ
Zb
= < F̄(γ(t)), T̄ (t) > ||γ ′ (t)|| dt
a
Zb
γ ′ (t)
= < F̄(γ(t)), > ||γ ′ (t)|| dt
a ||γ ′ (t)||
Zb
< F̄(γ(t)), γ ′ (t) > ′
= ||γ (t)|| dt
a ||γ ′ (t)||
Zb
= < F̄(γ(t)), γ ′ (t) > dt, as γ ′ (t) 6= 0̄.
a
Thus we have the following theorem.
Theorem 4.3.1. Let G ⊂ R3 and F̄ : G −→ R3 a continuous vector field.
Z
Suppose γ : [a, b] −→ R3 a smooth curve such that Im(γ) ⊂ G. Then F̄
γ
Z Zb
exists, and F̄ = < F̄(γ(t)), γ ′ (t) > dt.
γ a
Exercise 4.3.2. Let G ⊂ R3 and F̄ : G −→ R3 a continuous vector field and

γ : [a, b] −→ R3 a smooth curve such that Im(γ) ⊂ G

lOMoARcPSD|5686828
4.3. LINE INTEGRATION OF VECTOR FIELDS 29

Z Z
1. Show that F̄ = − F̄.
γ γ−1
2. ZIf φ : Z[c, d] −→ [a, Z b] is a smooth increasing function, then show that

F̄ = F̄, i.e., F̄ is independent of parametrisation of γ.
γ γ◦φ γ
Definition 4.3.3. Let G ⊂ R3 , F̄ : G −→ R3 a vector field, and γ a piecewise

continuous parametric curve, and suppose γ = γ1 ∗ γ2 ∗ · · · ∗ γs where each γi
Z Xs Z
is continuous. Then we define F̄ = F̄.
γ i=1 γi
Example 4.3.4. 1. Let F̄(x, y) = (x2 , y2 ) for all (x, y) ∈ R2 , and γ(t) =
(r cos(t), r sin(t)) for all t ∈ [0, 2π]. Clearly F̄ is continuous and γ is
C1 -type, and ′
Z check that γ (t) 6= 0 for all t ∈ [0, 2π] (i.e., γ is smooth),
therefore, F̄ exists and,
γ
Z Z 2π
F̄ = < F̄(γ(t)), γ ′ (t) > dt
γ 0
Z 2π
= r3 < (cos(t)2 , sin(t)2 ), (− sin(t), cos(t)) > dt
0
Z 2π
= r3 < (cos(t)2 , sin(t)2 ), (− sin(t), cos(t)) > dt
0
=0

−y x
2. Let F̄(x, y) = , for all (x, y) ∈ R2 \(0, 0), and γ(t) =
x2 + y2 x2 + y2
(r cos(t), r sin(t)) for all
Z t ∈ [0, 2π]. Clearly F̄ is continuous and γ is
C1 -type, and therefore, F̄ exists and,
γ
Z Z 2π
F̄ = < F̄(γ(t)), γ ′ (t) > dt
γ 0
Z 2π
= < (− sin(t), cos(t)), (− sin(t), cos(t)) > dt
0
Z 2π
= 1 dt
0
= 2π
3. Let F̄(x, y) = (f1 (x), f2 (y)) for all (x, y) ∈ R2 where f1 , f2 are continuous
functions such that g1′ = f1 , and g2′ = f2 ; and γ(t) = (r cos(t), r sin(t))
1
Zfor all t ∈ [0, 2π]. Clearly F̄ is continuous and γ is C -type, and therefore,
F̄ exists and,
γ

lOMoARcPSD|5686828
Z Z 2π
F̄ = < F̄(γ(t)), γ ′ (t) > dt
γ 0
= 0(Check!!)
Since we know that for a C1 -type scalar field f, the vector field ∇(f) is
continuous, and directional derivatives Dv̄ (f)|P exists for all direction v̄, and
for all point P; and Dv̄ (f)|P =< ∇(f)|P , v̄ >. And since for a continuous
vector field F̄ in R3 and a smooth parametric curve γ : [a, b] −→ R3 , we
Z Zb
have F̄ = < F̄(γ(t)), γ ′ (t) > dt, one naturally ask what happens if
γ a
< F̄(γ(t)), γ ′ (t) > represents aZdirectional derivatives of a C1 scalar field, i.e.,
we are asking what happens to F̄, if F̄ = ∇(f) for some C1 -type scalar field f.
γ
Note that thought γ ′ (t) is not a unit vector, we can choose a parametrisation of
the given curve such that the derivative of it a unit vector (e.g., consider γ◦s−1 ).
Z Zb Zb
′
So, ∇(f) = < ∇(f)(γ(t)), γ (t) > dt = Dγ ′ (t) (f)|γ(t) dt
γ a a
Zb
= < ∇f(γ(t)), γ ′ (t) > dt
a
Zb
∂ d ∂ d ∂ d
= f|γ(t) x(t) + f|γ(t) y(t) + f|γ(t) z(t) dt
a ∂x dt ∂y dt ∂z dt
Zb
d
= f(γ(t))
a dt
= f(γ(b)) − f(γ(a)).
Zb Z
Thus, Dγ ′ (t) (f)|γ(t) dt = ∇(f) = f(γ(b)) − f(γ(a)). This leads to the
a γ
following theorem called the Fundamental Theorem of Line Integration.
Theorem 4.3.5. Let G ⊂ R3 and F̄ : G −→ R3 a continuous vector field. If
F̄ = ∇(f) for some scalar field f : G −→ R, then for Zany smooth parametric
curve γ with initial point A and final point B we have F̄ = f(B) − f(A).
γ
4.4 Conservative Vector Fields

As any vector field which is of the type ∇(f) is special, as it is known from
previous result; and for which we shall set a definition for it.
Definition 4.4.1. Let G ⊂ R3 , and F̄ : G −→ R3 a vector field. F̄ is called
conservative vector field (CVF) if there exists a scalar field f : G −→ R
such that F̄(x̄) = ∇(f(x̄)) for all x̄ ∈ G, or simply F̄ = ∇(f) on G. The function
f is called of the gradient function of F̄.
Thus, in view of Theorem 4.3.5 we can say that line integration of conser-
vative vector field along a path does not depend upon the path, but upon the

lOMoARcPSD|5686828
4.4. CONSERVATIVE VECTOR FIELDS 31
end points of the curve. If the end points of the curve are the same point, i.e.,
the curve is a loop, then clearly the value of the line integration is zero.
Definition 4.4.2. Let F̄ : G −→ R3 a continuousZ vector
Z field. We shall say that
line integration of F̄ is path independent, if F̄ = F̄ for any two smooth
γ1 γ2
curves γ1 and γ2 with same initial and final points.
Clearly, by Theorem 4.3.5, line integration of a continuous conser-
vative vector field is path independent. Now, suppose F̄ is a general
continuous vector field (non necessarily conservative). If line
Z integration
Z of F
is path independent, for any smooth loop γ, we shall have F̄ = F̄, as γ
γ γ−1
and γ−1 has same initial and final points,
Z and therefore,
Z sinceZ by the properties
Z
of line integration of vector fields F̄ = − F̄, we get F̄ = − F̄, i.e.,
Z γ−1 γ γ γ
F̄ = 0. So, naturally one asks whether the converse is true, i.e., suppose F̄ is a
γ Z
continuous vector field such that F̄ = 0 for all smooth loop γ, is then the line
γ
integration of F̄ path independent? The answer is, in fact, “yes”!. We describe
it in the theorem below.
Theorem 4.4.3. Let G ⊂ R3 and F̄ : G −→ R3 a continuous vector field. Then
the following are equivalent
(I) Line integration of F̄ is independent of path.
Z
(II) F̄ = 0 for all smooth loop γ in G.
γ
Proof. (I) =⇒ (II): Clear from the discussion above.
(II) =⇒ (I): Let γ1 and γ2 are two paths in G with same initial point A
Z Z
and final point B. We shall show that F̄ = F̄. Consider γ = γ1 ∗ γ−1
2 .
γ1 γ2
Clearly,
Z γ is a loop with initial and final point A, and therefore, by hypothesis,
F = 0. From which we see
γ Z
0 = F̄
γ
Z
= F̄
γ1 ∗γ−1
2
Z Z
= F̄ + F̄ .
γ1 γ−1
2
Z Z
= F̄ + − F̄
γ1 γ2
Z Z
= F̄ − F̄
γ1 γ2

lOMoARcPSD|5686828
Z Z
This shows that F̄ = F̄, which completes the proof.
γ1 γ2
From Theorem 4.3.5 and Theorem 4.4.3 we see the following implications

A: F̄ is a continuous CVF
⇓

B: Line integration of F is path independent
m
Z
C: F̄ = 0 for all smooth loop γ
γ
At this point we should check whether B =⇒ A (same as C =⇒ A)

holds true. In that case we shall be able to describe the conservative vector
fields in terms of its properties of line integration. In the theorem below, called
1st Structure Theorem of CVF, we shall see that, in fact, B =⇒ A
holds under certain conditions; and from which we get the structure of
conservative vector fields in terms of theory of line integration.
4.5 1st Structure Theorem of CVFs

Definition 4.5.1. G ⊂ R3 is called path-connected, if for every two points A
and B in G, there exists a path (continuous) γ : [0, 1] −→ G from A to B, i.e.,
γ(0) = A, and γ(1) = B.
It can be seen that if G is open and path-connected, then every two points in
G can be connected by a piece-wise smooth path.
Theorem 4.5.2. Let G ⊂ R3 be open and path-connected. F̄ : G −→ R3 a

continuous vector field. Then the following are equivalent
(I) Line integration of F̄ is independent of path.

Z
(II) F̄ = 0 for all smooth loop γ in G.
γ
(III) F is a CVF.
Proof. We already have completed “(I) ⇔ (II) ⇐= (III)”. So, to complete

the proof we need to show “(II) =⇒ (III)”. So we suppose that (II) holds or
equivalently (I) holds. Shall show that (III) holds, though the proof is out of
the syllabus.
Step 1: Fix a point A in G and take a variable point P = (x, y, z) in G. Since

G is open and path connected,Z there exists a polygon γP (piece-wise smooth
path) from A to P. Consider F̄.
γP

lOMoARcPSD|5686828
4.5. 1ST STRUCTURE THEOREM OF CVFS 33
Step 2: Since line integration of F̄ is independent

Z ofZ path, for another piece-
wise smooth path δP from A to P, we have F̄ = F̄. Therefore, we can
Z δP γP
define a function f : G −→ R by f(P) = F̄

γP
Step 3: It can be shown that partial derivatives of f exists, and ∇(f) = F̄ on

G.
This completes the proof.
Remark 4.5.3. Note that we can not prove a vector field to be conservative
using Theorem 4.5.2; rather it will be extremely helpful to disprove a VF is
conservative, if the VF satisfies all the hypotheses of the theorem.

−y x
Example 4.5.4. Let F̄(x, y) = , for all (x, y) ∈ R2 \(0, 0).
x2 + y2 x2 + y2
NoteZthat the domain of F̄ is path-connected, and F̄ is continuous. We have seen
that F̄ = 2π 6= 0 for any circle γ around (0, 0), and therefore by 1st Structure
γ
Theorem of CVF we get F̄ is not conservative!
Since computation line integration of a vector field becomes extremely easy

if the vector field is conservative, we, for sure, want to build up a tool which
affirmatively detects a vector is conservative, if it is so.

−y x
Example 4.5.5. Let F̄(x, y) = , for all (x, y) ∈ R2 except
x2 + y2 x2 + y2
the “non negative part of x-axis”. We shall see that F̄ is conservative, but right
now we do not have any tool to detect it.
One approach would be trying to find out a scalar field, on using some algo-
rithm, for which we can write the vector field as a gradient of the scalar field.
There are disadvantages of this approach, which will be discussing after we dis-
cuss the algorithm. Suppose F̄ : G −→ R3 a conservative vector field. We shall
try to find f : G −→ R such that F̄ = ∇(f) on G.
Step 1: Suppose F̄ = (F1 , F2 , F3 ), and suppose we have found f : G −→ R such

that F̄ = ∇(f). In that case we have three equations.
 ∂
 ∂x f = F1 − − (1a)
∂
(1) ∂y f = F2 − − (1b)
 ∂
∂z f = F3 − − (1c)
Step 2: From (1a) we get

Z
f= F1 + φ1 (y, z) − −(2)
x
where φ is an arbitrary function of y and z.

lOMoARcPSD|5686828
Step 3: Differentiating (2) with respect to y and comparing with (1b) we get
Z
∂ ∂
φ1 (y, z) = F2 − F1 ,
∂y ∂y x
and therefore,
Z Z
∂
φ1 (y, z) = F2 − F1 + φ2 (z) − −(3)
y ∂y x
Plugging in the value of φ1 from (3) in (2) we get

Z Z Z
∂
f= F1 + F2 − F1 + φ2 (z) − −(4)
x y ∂y x
where φ2 is an arbitrary function of z.
Step 4: Again differentiating (4) with respect to z and comparing with (1c) we
get
Z Z Z Z
∂ ∂
φ2 (z) = F3 − F1 + F2 − F1 + c − −(5)
z ∂z x y ∂y x
where c is an arbitrary pure constant. Plugging in the value of φ2 from (5) in

(4) we get the function f, i.e.,
Z Z Z Z Z Z Z
∂ ∂ ∂
f= F1 + F2 − F1 + F3 − F1 + F2 − F1 +c
x y ∂y x z ∂z x y ∂y x
Since we do not have, so far, a definitive tool which tells us a given vector
field is conservative, we need check whether ∇(f) = F̄ holds after finding f in
this process.
Example 4.5.6. 1. Let F(x, y, z) = (sin(z)ex , y2 , cos(z)ex + z3 ). F1 =

sin(z)e , F2 = y2 , and F3 = cos(z)ex + z3 .
x
Z
So, f = sin(x)ex + φ1 (y, z) = sin(z)ex + φ( y, z).
x
Z
We know that ∂
∂y φ1 = F2 − ∂
∂y sin(z)ex = y2 , and therefore φ1 = y2 +
y
φ1 (y, z) = 1/3z3 + φ2 (z).
Thus, f = sin(z)ex + 1/3y3 + φ2 (z).

∂ ∂
Again, ∂z φ2 = F3 − ∂z sin(z)ex +1/3y3 = z3 , and therefore φ2 = 1/4z4 +c
where c is an arbitrary constant. Which gives us
f = sin(z)ex + 1/3y3 + 1/4z4 + c.
In this case we see that ∇(f) = F̄ on the domain of F̄, and therefore F̄ is a
conservative vector field.

lOMoARcPSD|5686828
4.6. FUNDAMENTAL THEOREM IN R2 : GREEN’S THEOREM 35
2. Let F̄(x, y) = (ey , x). Following the above process we see that f = xey +
xy − ey + c, but it is easy to note that ∇(f) 6= F̄.
Let G ⊂ R2 be open and F̄ = (F1 , F2 ) : G −→ R2 a conservative vector
∂
field. Then there exists f : G −→ R such that ∇(f) = F̄ in G, i.e, ∂x f = F1
2 2
∂ ∂ ∂ ∂ ∂
and ∂y f = F2 ; and therefore ∂x∂y f = ∂x F2 , and ∂y∂x f = ∂y F1 . Thus, if
∂2 ∂2 ∂ ∂
∂x∂y f = ∂y∂x f, we shall have ∂x F2 = ∂y F1 .
Fact: (Schwarz’s theorem or Clairaut’s theorem, or Young’s theorem)

Let G be an open set in R2 and P0 ∈ G. Suppose f : G −→ R be such that (i)
fx and fy exists in G, and (ii) either fxy or fyx exists and is continuous at P0 ,
then both fxy (P0 ) and fyx (P0 ) exist and fxy (P0 ) = fyx (P0 ).
i.e., in particular, if f is C2 -type, then all the partial derivatives of f of second

order commute; and therefore for a C1 -conservative vector field F̄ = (F1 , F2 ), we
∂ ∂
have, by our previous discussion, ∂x F2 = ∂y F1 on G.
Theorem 4.5.7. Let G be an open set in R2 and P0 ∈ G. Suppose F̄ = (F1 , F2 ) :
∂ ∂
G −→ R2 be a C1 -type conservative vector field, then ∂x F2 = ∂y F1 on G.
This result will help us quite a bit to quickly detect non-CVF which are
C1 -type.
Example 4.5.8. Consider the previous examples

−y x
1. F̄(x, y) = , for all (x, y) ∈ R2 \(0, 0). It can be seen,
x2 + y2 x2 + y2
∂ ∂
in this example, that dom(F̄) is open, F̄ is C1 -type, and ∂x F2 − ∂y F1 = 0;
but from this we can not have any conclusion.
∂ ∂
2. F̄ = (ey , x). In this case it is easy to see that ∂x F2 − ∂y F1 6= 0, and
therefore F̄ is not conservative.
4.6 Fundamental Theorem in R2 : Green’s The-

orem
We can realize the fundamental theorem of line integration or fundamental the-
orem integral calculus as evaluation of an integration of the “differentiation” of
a function over a “one dimensional” bounded region with “no hole” by evalua-
tion of the function at the “boundary” points. Naturally one asks what could
be its higher dimensional analogue! Essentially it asks about formulation of a
possible version of fundamental theorem of integration for higher dimensional
cases. Green’s theorem gives answer to this question for the case “dimension”
is two.
Definition 4.6.1. G ⊂ R3 is called simply connected if G is path-connected
and “hole-less”(:every loop in G can be continuously deformed, being in G, to a
point in G).
Theorem 4.6.2. Let G ⊂ R2 be a simply connected bounded set with a piece-
wise smooth boundary ∂G which is oriented positively
ZZ (anti-clock), Zand F̄ =
∂ ∂
(F1 , F2 ) : G −→ R2 a C1 -type vector field. Then ( F2 − F1 ) = F̄.
G ∂x ∂y ∂G

lOMoARcPSD|5686828

Z Z
Note that, the integral F̄ sometimes is written as (F1 dx + F2 dy)
∂G ∂G
with the understanding that ∂G := γ(t) = (x(t), y(t)) for t ∈ I, leading to
γ ′ (t)Z= (x ′ (t), Zy ′ (t)), and therefore,
F̄ = < (F1 (γ(t)), F2 (γ(t))), (x ′ (t), y ′ (t)) > dt
∂G ∂G
R
(F1 (γ(t)) x ′ (t) + F2 (γ(t)) y ′ (t)) dt
= ∂G Z
which symbolically written as (F1 dx + F2 dy).
∂G
4.6.1 Application: 2nd Strure Theorem of CVFs in R2

Let G ⊂ R2 be open and simply connected, and F̄ = (F1 , F2 ) : G −→ R2 a
∂ ∂
C1 -type vector field. Suppose ∂x F2 − ∂y F1 = 0. We take any loop C, oriented
positively in G, and consider the region GC enclosed by C in G. Clearly ∂GC =
C. Since G is simply connected, GC is also simply connected, and therefore, we
can apply Green’s theorem on GC for the function F̄, and get
ZZ Z Z
∂ ∂
0= F2 − F1 dx dy = F̄ = F̄.
G ∂x ∂y ∂GC C
This shows that line integration of F̄ along any loop vanishes and which,
from 1st Structure Theorem of CVF we see that F̄ is conservative.
Theorem 4.6.3. Let G ⊂ R2 be open and simply connected, and F̄ = (F1 , F2 ) :

G −→ R2 a C1 -type vector field. Then the following are equivalent
1. F̄ is a conservative vector field.

∂ ∂
2. ∂x F2 − ∂y F1 = 0.
3. Line Integration of F̄ is path independent.

Z
4. F̄ = 0 for all loop C.
C
Remark 4.6.4. 1. Note that one never achieves 1st structure theorem of
CVF from the 2nd structure theorem, and vice versa.
2. Suppose F̄ = (F1 , F2 ) : G −→ R2 is a C1 -type vector field with an open

∂ ∂
domain G. Suppose ∂x F2 − ∂y F1 = 0. Suppose C be a loop in G. Consider
the regionZ GC in G enclosed by the loop C. If GC has no hole, byZ Green’s
theorem, F̄ = 0; and if GC has a hole (due to hole in G), then F̄ may
C C
not be zero. This shows that for C1 -type vector fields having “curl” zero,
only the holes in its domain creates obstructions in its way to be CVF.
4.6.2 Application: Extended Green’s Theorem

Theorem 4.6.5 (Extended Green’s Theorem). Let F̄ = (F1 , F2 ) be a C1
type vector field with an open domain G ⊂ R2 having n holes H1 , H2 , · · · , Hn .
Suppose C1 is a loop in G such that the region of G entangled by C contains all

lOMoARcPSD|5686828
the holes. Further suppose C2 , C3 , · · · , Cn+1 are loops in G such that the region
of G entangled by Ci contains only one hole Hi for all i = 1, 2, · · · , n, and the
entangled region is inside the region entangled by C1 (i.e., Ci ’s are smaller loops
than C1 ). Orientation of all the loops are positive. Then
ZZ Z XZ
n+1
∂ ∂
F2 − F1 dx dy = F̄ − F̄.
G ∂x ∂y C1 Ci
i=2
Proof. Look at Apostol’s Calculus Vol-II

∂
Remark 4.6.6. As a consequence of this we see that if ∂x F2 − ∂ F1 = 0, and if
ZZ ∂y
∂ ∂
C1 and C2 are any two loops around a hole in G, then 0 = F2 − F1 dx dy =
Z Z Z Z G ∂x ∂y
F̄ − F̄, i.e., F̄ = F̄.
C1 C2 C1 C2

−y x
Example 4.6.7. Let F̄(x, y) = , 2 for all (x, y) ∈ G :=
x + y Z x + y2
2 2
R2 \(0, 0). Suppose C be a loop in G. Then F̄ = 0 or 2π:

C
∂ ∂
We all know that for this vector field ∂x F2 − ∂y F1 = 0, and F̄ is C1 -type.
Case-I: Suppose C be such that the region entangled by Z it in ZG contains the

hole (i.e., missing (0, 0)), i.e., C is around (0, 0), then F̄ = F̄ where C∗
C C∗
is any another loop around (0, 0). We choose C∗ to be x2 + y2 Z= r2 where r is
such that C∗ is inside the region entangled by C. We know that F̄ = 2π, and
Z C∗
therefore we have F̄ = 2π.

C
Case-II: Suppose the region entangled by ZC in G do not contain the hole.

Then, we can apply Green’s theorem and get F̄ = 0.
C

1 y 1
Example 4.6.8. Let F̄(x, y) = e , − ey
for all (x, y) ∈ R2 \ y − axis.
x2 x Z
Let γ be a path from (1, 1) to (10, 30). We shall calculate F̄.
γ
2
Since dom(F̄) = R \y − axis, it is not path-connected. But the domain has
two path-connected components GL , and GR , where GL = {(x, y) | y < 0}, and
GR = {(x, y) | y > 0}. Since the points (1, 1) & (10, 30) lies in GR , and γ is path
from (1, 1) to (10, 30), the path γ lies onZ GR ; and therefore, we consider the
∂ ∂
function F̄ on GR for the computation of F̄. We see that ∂x F2 − ∂y F1 = 0.
γ
Since GR is open and simply connected, and F̄ is C1 -type, by 2nd structure
Z
theorem of CVF we get F̄ : GR −→ R2 is conservative on GR , and hence F̄ =
γ
1
f((10, 30)) − f((1, 1)) where ∇(f) = F̄ on GR . It is easy to see that f = − ey ,
x

lOMoARcPSD|5686828

Z
1 30
and therefore F̄ = − e + e.
γ 10

−y x
Exercise 4.6.9. Let F̄(x, y) = for all (x, y) ∈ G :=
,
Z x2 + y2 x2 + y2
2
R \(0, 0). Let γ be a path from (1, 1) to (10, 30). Calculate F̄.
γ
[Hint: Use the technique applied in the previous example.]
4.6.3 Application: Recognizing CVF in R2

Suppose G ⊂ R2 be open and F̄ = (F1 , F2 ) : G −→ R2 a continuous vector field.
∂ ∂
If F̄ is C1 -type, we check ∂x F2 − ∂y F1 .
∂ ∂
♣ If ∂x F2 − ∂y F1 6= 0 we conclude that F̄ is not a CVF.
∂ ∂
♣ If ∂x F2 − ∂y F1 = 0, we check whether dom(F̄).
If dom(F̄) is simply connected, we conclude that F̄ is a CVF.
If dom(
Z F̄) has holes, around each hole we take loops C for
which we can compute F̄.
C Z
N If F̄ = 0 for each hole, we conclude that F̄ is a CVF.
ZC
N If F̄ 6= 0 for some hole, we conclude that F̄ is not a
C
CVF.
If F̄ is not of C1 -type, but continuous, we try to get f : dom(F̄) −→ R such
that ∇(f) = F̄ on dom(F̄) by solving equation “∇(f) = F̄”.
♣ If this method gives affirmative answer, for sure, F̄ is aZCVF.
♣ If this method fails, we try to find a loop C such that F̄ 6= 0.
C
4.6.4 Application: Finding Area enclosed by a loop

Let C be a piecewise smooth loop in R2 and G be the region in R2 enclosed by
the loop C, i.e., C = ZZ
∂G. We shall find Area(G) using ZGreen’s theorem. We
know that Area(G) = 1 dx dy which we can write as F̄, on using Green’s
G C
theorem if we get suitable C1 -type vector field F̄ on G and orient C = ∂C posi-
tively.
∂ ∂
So, we orient C positively and choose F̄ for which ∂x F2 − ∂y F1 = 1. It
is to be noted that there are infinite number of such F̄, e.g., (0, x); (−y, 0);
(−y/2, x/2); (−y/n, (n − 1)x/n), etc. We choose F̄ = (0, x), which is C1 -type.
Since G is simply connected, C = ∂G is positively oriented, and F̄ is of C1 -type,
we get applying Green’s theorem
ZZ ZZ Z Z
∂ ∂
Area(G) = 1 dx dy = ( F2 − G2 ) dx dy = F̄ = (0, x).
G G ∂x ∂y ∂G C
Suppose C has a parametric form γ(t) = (x(t), y(t)) for all t ∈ [a, b]. Then,
Zb Zb
′ ′
Area(G) = < (0, x(t), (x (t), y (t))) > dt = x(t) y ′ (t) dt.
a a

lOMoARcPSD|5686828
x2 y2
Example 4.6.10. We shall find the area enclosed by C : 2 + 2 = 1, where
a b
a, b > 0 by Green’s theorem.
Let G be the region enclosed by COn Choosing F = (0, x) we see that

ZZ ZZ Z Z
∂ ∂
Area(G) = 1 dx dy = ( F2 − G2 ) dx dy = F̄ = (0, x).
G G ∂x ∂y ∂G C
We parametrize C as C : γ(t) = (a cos(t), b sin(t)) for all t ∈ [0, 2π]. And

therefore we have,
Z 2π Z 2π
Area(G) = < (0, a cos(t)), (−a sin(t), b cos(t)) > dt = ab cos2 (t) dt = πab.
0 0

lOMoARcPSD|5686828

lOMoARcPSD|5686828
Chapter 5
Surface Integration
Definition 5.0.1. Let S
41

lOMoARcPSD|5686828
42 CHAPTER 5. SURFACE INTEGRATION

Vector-Calculus Class Note Vector-Calculus Class Note

Uploaded by

Copyright:

Available Formats

Vector-Calculus Class Note Vector-Calculus Class Note

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vector-Calculus Class Note Vector-Calculus Class Note

Uploaded by

Copyright:

Available Formats

lOMoARcPSD|5686828

Vector-Calculus Class Note

Several Variable Calculas (Indian Institute of Technology Kanpur)

StuDocu is not sponsored or endorsed by any college or university

Vector Calculus Lecture Note

Downloaded by Suraj Gautam ([email protected])

Downloaded by Suraj Gautam ([email protected])

1 The Space of Vectors 1

2 Directional Derivative and Differentiability 11

3 Arc Length Function 21

Downloaded by Suraj Gautam ([email protected])

Downloaded by Suraj Gautam ([email protected])

The Space of Vectors

The modern definition of vectors is axiomatic in nature; it defines vector by

1.2 Algebraic Structure of Vectors

Downloaded by Suraj Gautam ([email protected])

2 CHAPTER 1. THE SPACE OF VECTORS

(1) (Existence of binary operation) There exists a function + : V × V −→ V

Exercise 1.2.2. Show that in a space of vectors

Downloaded by Suraj Gautam ([email protected])

1.2. ALGEBRAIC STRUCTURE 3

Downloaded by Suraj Gautam ([email protected])

4 CHAPTER 1. THE SPACE OF VECTORS

3. {(1, 1), (2, 1)} is a basis of R2 . (Prove it!)

4. {(1, 1), (2, 0), (0, 3)} is not a basis of R2 .

Exercise 1.2.8. Show that any basis of Rm contains exactly m elements.

We are now ready to introduce the axiomatic notion of length of a vector

1.3 The Spatial Structure of Vectors

N-1: kv̄k ≥ 0 for all v̄ in V; and kv̄k = 0 if and only if v̄ = 0̄.

N-2: kαv̄k = |α|kv̄k for all α in R and for all v̄ in V.

N-3: kv¯1 + v¯2 k ≤ kv¯1 k + kv¯2 k for all v1 , v2 in V.

Having a norm function k.k on a space of vectors we can define a metric,

Downloaded by Suraj Gautam ([email protected])

1.3. SPATIAL STRUCTURE 5

Definition 1.3.3. Let V be a space of vectors over R and k.k : V −→ R a norm.

Let us now compute some metrics corresponding to norms in the respective

where x̄ = (x̄1 , x2 , · · · , xm ), ȳ = (y1 , y2 , · · · , ym ) ∈ Rm . This metric is on Rm

Downloaded by Suraj Gautam ([email protected])

6 CHAPTER 1. THE SPACE OF VECTORS

Proposition 1.3.8 (Cauchy-Schwarz inequality). Let V be a space of vectors

Proof. Let f, g in V. Then hλf + g , λf + gi = λ2 hf , fi + 2λhf , gi + hg , gi.

Corollary 1.3.9. Let

Proof. Let f, g in V. Then hf + g , f + gi = hf , fi + 2hf , gip + hg , gi. Using

Corollary 1.3.10. Let V be a space if vectors and h. , .i : V × V −→pR be an

Downloaded by Suraj Gautam ([email protected])

1.4. TOPOLOGY OF THE EUCLIDEAN SPACE RM 7

1.4 Topology of the Euclidean Space Rm

B(x̄0 , r) = { x̄ ∈ Rm | d(x̄0 , x̄) < r }

A closed n-ball in Rm centered at x̄0 with radius r > 0 is denoted by B(x̄0 , r)

Downloaded by Suraj Gautam ([email protected])

8 CHAPTER 1. THE SPACE OF VECTORS

Definition 1.4.3. A set U ⊆ Rm is called open if for each point x̄ ∈ U there

Proof. The proof is very similar to that of Proposition 1.4.5.

Corollary 1.4.7. Any finite set in Rm is closed.

Proof. Follows from Proposition 1.4.6.

Downloaded by Suraj Gautam ([email protected])

1.4. TOPOLOGY OF THE EUCLIDEAN SPACE RM 9

Definition 1.4.9. Any function f : Rm −→ R is called a scalar field, and

A vector field F : G −→ Rn is called continuous at ā, if each of the scalar