Notes on lecture 3

Jiten Chandra Kalita and Shyamashree Upadhyay

1 Derivative of a vector function

It will be easy to understand this topic if we interpret the vector function in
terms of space curves.
Definition 1.0.1. Suppose ~r(t) is a vector function, where t ∈ [a, b]. Then
the derivative of ~r(t), denoted by d~
r
dt
is given by:
d~r ~r(t + h) − ~r(t)
= lim (+)
dt h→0 h
provided the limit exists. If the limit exists, then we say that the derivative
of the vector function ~r exists and it is given by (+). 

1.1 Geometric interpretation of the derivative

The secant approach: Suppose we are considering a space curve. Suppose
at time t, the particle is at the point P on the space curve. And suppose at
time t + h, the particle is at the point Q on the space curve. Then the vector
−→
P Q equals ~r(t + h) − ~r(t). See part (a) of Figure 1.1.1 for an illustration. If
we divide the vector ~r(t + h) −~r(t) by h (where h is a very small real number,
that is, h1 is a very big real number), then in fact, we are multiplying the
−→
vector P Q by a scalar λ (= h1 ), which is a very big real number. Then
definitely, the new vector ~r(t+h)−~
h
r(t)
is going to have the same direction as
−→ −→
P Q. But its length will be much much bigger than the length of P Q. This
vector ~r(t+h)−~
h
r(t)
passing through P and Q is called the secant vector. See
Figure 1.1.1 for an illustration.
As we let h → 0, the secant vector will approach the tangent to the curve
at the point P (See part (b) of Figure 1.1.1). This finishes the geometrical
interpretation.

1
 Figure 1.1.1: Derivative of a vector function

1.2 Notation
d~
r
dt
is also denoted by r~0 (t). If r~0 (t) 6= ~0, then the derivative defines a tangent
to the curve C at the point P .
We often denote r~0 (t) by T~ . The unit tangent vector along the direction
of the tangent is denoted by t̂, where

T~ r~0 (t)
t̂ = = .
|T~ | |r~0 (t)|

Recall that if
~r(t) =< f (t), g(t), h(t) >
is a vector function, then its derivative r~0 (t) is given by

d~r ~r(t + h) − ~r(t)

r~0 (t) = = lim . (∗)
dt h→0 h
Theorem 1.2.1. If ~r(t) =< f (t), g(t), h(t) > is a vector function, then

r~0 (t) =< f 0 (t), g 0 (t), h0 (t) >,

provided each of the component functions of ~r(t) is a differentiable function.

Proof: It follows from (∗) above that

f (t + h) − f (t) g(t + h) − g(t) h(t + h) − h(t)

r~0 (t) =< lim , lim , lim >.
h→0 h h→0 h h→0 h

2
But now, it follows from single variable calculus that the right hand side of
the above equation is equal to < f 0 (t), g 0 (t), h0 (t) >, provided the functions
f, g, h are differentiable at t. 

Remark 1.2.2. Whenever we need to find out the derivative of a vector function
in terms of its components, we simply need to find out the derivatives of its
components and place them as the components of the derivative vector function,
provided the component functions are differentiable.

Example 1.2.3. If ~r(t) =< cos t, sin t, t >, then r~0 (t) =< − sin t, cos t, 1 >.


Some important properties of the derivative are listed in Theorem 1.2.4

below.

Theorem 1.2.4. Suppose u and v are differentiable vector functions, c is a

scalar, and f is a real-valued function. Then

Proof of (5) in Theorem 1.2.4: By definition, we have

 
d u(t + h) × v(t + h) − u(t) × v(t)
[u(t) × v(t)] = lim .
dt h→0 h

3
It is easy to see that the right hand side can be rewritten as
 
u(t + h) × v(t + h) − u(t + h) × v(t) + u(t + h) × v(t) − u(t) × v(t)
lim .
h→0 h
If we now bring the limit inside, then from the properties of cross product
(studied in earlier lectures), we can say that the above expression equals
v(t + h) − v(t) u(t + h) − u(t)
lim u(t + h) × lim + lim × v(t),
h→0 h→0 h h→0 h
which in turn equals u(t) × v0 (t) + u0 (t) × v(t).

We will now use part (4) of Theorem 1.2.4 above to derive an important
result, which is going to be very useful later on.
Result: If u is a vector of constant length, then u and u0 are orthogonal to
each other.
Proof of the result: Let |u| = a, where a is a constant. Then |u|2 = a2 .
This implies that u · u = a2 . Now taking derivative, we get
d
[u · u] = 0.
dt
By part (4) of Theorem 1.2.4, we then have:
u · u0 + u0 · u = 0.
But we know that vector dot product is commutative. So the above equation
implies
2u · u0 = 0 ⇒ u · u0 = 0.
This implies that u and u0 are orthogonal to each other.

2 Vector integration
Rb
Recall from single variable calculus that to compute the integral a f (t)dt of a
real valued function f over an interval [a, b], we first divided the interval [a, b]
into sub intervals, then we considered the areas of the strips or rectangles
formed by taking certain heights (dependent on the value of f ) over those
sub intervals, and then we added those areas. And as the number of sub
intervals of the interval [a, b] increases (or tend to ∞), the limit of the sum
Rb
of the areas of all those strips was defined to be the integral a f (t)dt.
We now have the following definition for the integral of a vector function:

4
Definition 2.0.1. Let ~r(t) =< f (t), g(t), h(t) > be a vector function. Then
Z b Z b Z b Z b
~r(t)dt =< f (t)dt, g(t)dt, h(t)dt > .
a a a a


Let us now state (without proof) the fundamental theorem of integral
calculus for vector functions (as Theorem 2.0.2 below).
Theorem 2.0.2. Suppose ~r(t) is a continuous vector function such that
~r(t) = R ~ 0 (t) for some other vector function R(t).
~ ~
[In such a case, R(t)
is called an antiderivative of ~r(t).] Then
Z b
~
~r(t)dt = R(b) ~
− R(a).
a

3 Arc length and curvature

Let us first have a quick glimpse at how we are going to find arc length in
cartesian coordinates. Suppose we consider a curve C in a 2 dimensional
plane on an interval [a, b]. Assume that C is the graph of a function of one
variable y = y(x) on the interval [a, b]. Let x0 = a and xn = b. Let us divide
the interval [a, b] into some sub intervals of the form [xi−1 , xi ] (1 ≤ i ≤ n)
by introducing some points x1 , . . . , xn−1 in between the points a and b. Note
that these points in between need not have to be equidistant from each other.
Consider the interval [xi−1 , xi ] for any i ∈ {1, . . . , n}. For i ∈ {1, . . . , n}, let
Pi denote the point (xi , y(xi )) in the xy-plane. Join the points Pi−1 and Pi
for each i by a straight line. Then we can easily
Pn see that the length L of the
curve C can be approximated by the sum i=1 |Pi−1 Pi |. That is,
n
X
L = lim |Pi−1 Pi |.
n→∞
i=1

See Figure 3.0.1 for an illustration.

pLet ∆xi = xi − xi−1 and ∆yi = y(xi ) − y(xi−1 ). Observe that |Pi−1 Pi | =
(∆xi )2 + (∆yi )2 . Then we have from the above equation that
n
X p
L = lim (∆xi )2 + (∆yi )2 .
n→∞
i=1

5
 Figure 3.0.1: The arc length

Now if y is a differentiable function of the single variable x, then by mean

value theorem (learnt in single variable calculus), we get
∆yi = y(xi ) − y(xi−1 ) = (xi − xi−1 )y 0 (x∗i ) = ∆xi y 0 (x∗i ),
for some x∗i ∈ (xi−1 , xi ). Then we can write
n
X p
L = lim (∆xi )2 + (∆yi )2
n→∞
i=1
n
X p
= lim 1 + {y 0 (x∗i )}2 ∆xi ,
n→∞
i=1
Rbq dy 2
which in turn equals a 1 + ( dx ) dx. That is, we have
Z br
dy
L= 1 + ( )2 dx. (∗∗)
a dx
The formula given in equation (∗∗) above is known as the arc length for-
mula, provided y 0 is continuous on [a, b].

3.1 Arc length when the curve is parametrized by the

variable t
Consider the plane curve C in Figure 3.1.1. Compare this curve C in Fig-
ure 3.1.1 with the curve in Figure 3.0.1. Observe that the curve C in Fig-
ure 3.1.1 is NOT the graph of a function of one variable, but the curve in

6
Figure 3.0.1 is so. This is because, in Figure 3.1.1, the curve C is drawn in a
twisted way. If we draw a vertical line at the point where the curve C bends
(or takes twist), then that vertical line is going to cut the curve C at more
than one points. The curve C in Figure 3.1.1 is not the graph of a function
of one variable. But still, there is a way out for finding the length of such a
curve. And this is done by parametrizing the curve C by a new variable
t.
Assume that the curve C is drawn in the xy-plane in such a way that the

Figure 3.1.1: Arc length when the curve is parametrized by t

initial point P0 of the curve has x-coordinate a and the final point Pn has
x-coordinate b. Suppose we consider any point P on the curve C whose co-
−→
ordinates are (x, y) and we join the point P with the origin O. Say, ~r = OP .
We are going to define ~r as < x(t), y(t) > (in terms of the variable t). That
is,
~r(t) =< x(t), y(t) > .
Here, we are considering an interval [α, β] (where the variable t lies). Let us
assume that x and y are continuous functions of the single variable t.

Let us now divide the interval [α, β] (in which the parameter t lies) into
some sub intervals: Let t0 = α, tn = β, and t1 , . . . , tn−1 be some interme-
diate points (which are not necessarily equidistant) on the interval [α, β].
Through the points t0 , t1 , . . . , tn , we will get hold of some points on the curve
C in the following way: For each i ∈ {0, . . . , n}, let Pi = (x(ti ), y(ti )) de-
note the corresponding point on the curve C. So we will get the points
P0 , P1 , . . . , Pn on the curve C. And corresponding to the points Pi−1 and Pi ,

7
we have the interval [ti−1 , ti ] ⊆ [α, β]. Let us join each pair of consecutive
points (Pi−1 , Pi ), where i ∈ {1, . . . , n} by straight lines (that is, P0 and P1 ,
P1 and P2 , and so on). Then (as before), we will have
n
X n
X p
L = lim |Pi−1 Pi | = lim (∆xi )2 + (∆yi )2 ,
n→∞ n→∞
i=1 i=1

where L denotes the length of C. Now,

∆xi = xi − xi−1 = x(ti ) − x(ti−1 )

Applying the mean value theorem, we get

x(ti ) − x(ti−1 ) = (ti − ti−1 )x0 (t∗i ) = ∆ti x0 (t∗i ),

where t∗i ∈ (ti−1 , ti ) and ∆ti = ti − ti−1 . Similarly, we get

∆yi = ∆ti y 0 (t∗∗

i ),

where t∗∗
i ∈ (ti−1 , ti ). It hence follows that

n
X n
X
p p
L = lim (∆xi )2 + (∆yi )2 = lim (x0 (t∗i ))2 + (y 0 (t∗∗ 2
i )) ∆ti
n→∞ n→∞
i=1 i=1
s 2 2
Z β 
dx dy
= + dt.
α dt dt
Here x0 and y 0 should be continuous on [α, β] and the curve C should be
traversed exactly once as t increases from α to β.

Remark 3.1.1. Whenever we consider a space curve or a plane curve in terms

of a vector function, we should always consider the interval of the parameter t
in such a way that it is in the increasing direction, so that we are able to track
the curve.

In this way, we can find the length of a curve C in a plane, where the
curve C has been parametrized by the variable t.

8
3.2 Arc length of a curve in space
Suppose now that we have a curve in the 3 dimensional space. Say, an
arbitrary point P on the curve has coordinates (x(t), y(t), z(t)). So x, y, z are
3 functions from an interval [α, β] to the real line R. The position vectors of
the initial and the end points of the curve are then ~r(α) and ~r(β) respectively.
Then applying the same methodology (as in the 2 dimensional plane curve
case), we get
s
Z β  2  2  2
dx dy dz
L= + + dt. (+)
α dt dt dt
Rβ
Theorem 3.2.1. L = α |r~0 (t)|dt, L denotes the length of a curve in space
which is parametrized by ~r(t) =< x(t), y(t), z(t) > where α ≤ t ≤ β.

Proof: q Observe that r~0 (t) =< x0 (t), y 0 (t), z 0 (t) >. It is then clear that the
dx 2 dy 2 dz 2




quantity dt
+ dt
+ dt
in equation (+) is nothing but the magni-
~0 ~0
tude |r (t)| of r (t). So, we have (from (+)):
Z β
L= |r~0 (t)|dt.
α

—————-END——————