3 Differentiation

3.1 Definitions
Definition of limit

Consider the function f (x). If we can make f (x) as near as we want to a given number l by choos-
ing x sufficiently near to a number a, then l is said to be the limit of f (x) as x → a and it is written
as
lim f (x) = l .
x→a
The Derivative

The derivative of f (x) is the slope of, or the gradient of the tangent to, the function f (x) at x and is given
by

df f (x + δx) − f (x)
f ′ (x) ≡ = lim
dx δx→0 δx

Similarly, the second derivative is

d2 f f ′ (x + δx) − f ′ (x)
f ′′ (x) ≡ = lim
dx2 δx→0 δx

and generally

dn f f (n−1) (x + δx) − f (n−1) (x)

f (n) (x) ≡
= lim
dxn δx→0 δx
′
where f (x) ≡ f (x), etc and f (x) = f (x).
(1) (0)

3.2 Examples of derivations of derivatives

The Derivative of xn

From the above definition

d xn (x + δx)n − xn (x + h)n − xn
= lim = lim
dx δx→0 δx h→0 h
Now recall that
an − bn = (a − b)(a(n−1) + a(n−2) b + a(n−3) b2 ......a b(n−2) + b(n−1) )
Applying this last result with a = x + h and b = x we get
d xn ((x + h)(n−1) + (x + h)n−2 x + (x + h)n−3 x2 ........(x + h)x(n−2) + x(n−1) )
= lim h
dx h→0 h

1
Thus
d xn
= nx(n−1)
dx
Note that this derivation can be extended to negative and fractional values of n - we shall assume without
proof that the above result is valid for all values of n.

The derivative of y = ex

Consider first the more general function f (x) = ax . The definition of the derivative gives

d ax (a(x+δx) − ax ) (a(δx) − 1)
= lim = lim ax
dx δx→0 δx δx→0 δx
In words this result states that the derivative of ax equals ax times the slope of ax at x = 0 (this slope is
the (a δx−1) ie the (a δx −a ) term. That is:
(δx) (0+δx) 0

f ′ (x) = ax f ′ (0) (1)

Note also that although we recognise this term as the slope, we do not get an analytic expression for this
slope. So its natural to ask the question whether a value of a exists such that at x = 0 the slope is 1. The
number for which f ′ (0) = 1 is given the name ”e”, after the mathematician Euler. Thus the function ex
is defined such that its slope at x = 0 equals unity and hence from the expression for the derivative of ax
we have for f (x) = ex
dy
y= = ex
dx
ie ex is a function which equals its slope.

Derivative of sin θ and cos θ

If f = sin θ then from the fundamental definition of the derivative

df sin(θ + δθ) − sin θ

= lim
dθ δθ→0 δθ
sin θ cos δθ + cos θ sin δθ − sin θ
= lim
δθ→0 δθ

= cos θ .
( )
sin δθ
cos δθ → 1, → 1 as δθ → 0
δθ
If f = cos θ then

df cos(θ + δθ) − cos θ

= lim
dθ δθ→0 δθ
cos θ cos δθ − sin θ sin δθ − cos θ
= lim
δθ→0 δθ

2
 = − sin θ .

3.3 Derivatives of basic functions

Differentiation from first principles is time-consuming. What we usually do is use a list of derivatives of
basic functions as a basis for more complicated functions. A set of useful derivatives is the following:

dxn
= nxn−1 (2)
dx
deax
= aeax (3)
dx
d ln(x) 1
= (4)
dx x
d sin(ax)
= a cos ax (5)
dx
d cos(ax)
= −a sin(ax) (6)
dx
d sec(ax) 1
= a sec(ax) tan(ax) [reminder : sec x ≡ ] (7)
dx cos x
d tan(ax)
= a sec2 (ax) (8)
dx
dcosec(ax) 1
= −acosec(ax) cot ax [reminder : cosecx ≡ ] (9)
dx sin x
d cot(ax) cos x
= −acosec2 ax [reminder : cot x ≡ ] (10)
dx sin x
d sin−1 (x/a) 1
=√ (11)
dx a − x2
2

d cos−1 (x/a) −1
=√ (12)
dx a2 − x2
−1
d tan (x/a) a
= 2 (13)
dx a + x2

3.4 The product rule and derivative of quotients

The product rule

For
f (x) = u(x)v(x)
what is f ′ (x)? We start from the definition

df f (x + δx) − f (x)
= lim
dx δx→0 δx

3
Now

f (x + δx) − f (x) = u(x + δx)v(x + δx) − u(x)v(x) (14)

= u(x + δx) [v(x + δx) − v(x)] + [u(x + δx) − u(x)] v(x) (15)

and we have
[ ] [ ]
df v(x + δx) − v(x) u(x + δx) − u(x)
= lim {u(x + δx) + v(x)}
dx δx→0 δx δx
In the limit δx → 0 the factors in the square brackets become v ′ (x) and u′ (x) and u(x + δx) become u(x).
Thus we have

df dv du
=u +v . (16)
dx dx dx
or more compactly using a different notation

(uv)′ = uv ′ + u′ v (17)
This rule can be extended to the product of three or more functions:

(uvw)′ = u′ vw + uv ′ w + uvw′ (18)

An example
d ( 3 ) d d ( 3)
x sin x = x3 (sin x) + x sin x
dx dx dx

= x3 cos x + 3x2 sin x .

The derivative of a quotient
Applying the product rule to a quotient of two factors
u(x) 1
f (x) = or f (x) = u(x) .
v(x) v(x)
So

f ′ (x) = (uv −1 )′ = u′ v −1 + u(v −1 )′ = (19)

( )
u −v
= +u = (20)
v v2
u′ v − uv ′
= (21)
v2
Examples
sin θ
For f = tan θ use f = cos θ
.

4
Then
df cos θ cos θ + sin θ sin θ
=
dθ cos2 θ
sin2 θ
= + 1 = 1 + tan2 θ = sec2 θ .
cos2 θ
1
If f = cosecθ = sin θ
then
df (−1) cos θ
= = − csc θ cot θ
dθ (sin θ)2
Similarly if f = sec θ, then
df
= sec θ tan θ.
dθ

3.5 The chain rule

We may have a situation where we need to differentiate a function of a function, i.e. we have a situation
where f (x) can be expressed as f = f (u(x)). For example

f (x) = (3 + x2 )3 = u(x)3 (22)

where
u(x) = 3 + x2
To differentiate such functions we use the chain rule

df df du
= .
dx du dx
For our example
( )3 ( )
f (x) = 3 + x2 = f (u) , u = 3 + x2

df df du ( )2
= = 3u2 · 2x = 3 3 + x2 × 2x
dx du dx
( )2
= 6x 3 + x2 .
Three more examples:

• f (x) = eax [one of our standard functions from few pages ago]

f (u) = eu and u = ax

df du df
= eu and = a thus = aeax
du dx dx

5
 • f (x) = (sin x)3
We have

f (u) = u3 and u = sin x

df du df
= 3u2 and = cos x thus = 3 sin2 x cos x
du dx dx
• f (x) = sin(x3 )
We have

f (u) = sin(u) and u = x3

d du df
= cos u and = 3x2 thus = 3x2 cos x3
du dx dx

3.6 Implicit differentiation

So far we have only considered functions where y = f (x), i.e. only one variable on the right-hand side.
We may have a situation where it is not so easy to express y in terms of x, e.g.:

x3 − 3xy + y 3 = 2 .

We therefore differentiate term by term with respect to x which is called implicit differentiation.
d 3 d d d
(x ) − (3xy) + (y 3 ) = (2)
dx dx dx dx
[ ]
dy dy
(3x ) − 3x + 3y + 3y 2
2
=0
dx dx
dy y − x2
= 2
dx y −x

3.7 Parametric differentiation and inverse differentiation

Inverse differentiation

If y = f (x) and x = f −1 (y) are inverse functions, then

dy 1
= dx (23)
dx dy

Example: inverse trigonometric functions

Differentiate y = sin−1 x. We see that x = sin y.

6
 dy 1 1
= dx = (24)
dx dy
cos y
Now, we know sin2 y + cos2 y = 1, thus
√
cos y = 1 − sin2 y (25)

so
dy 1
=√ (26)
dx 1 − sin2 y
but x = sin y thus

d sin−1 x 1
=√ (27)
dx 1 − x2
Example: inverse hyperbolic functions
Hyperbolic functions are defined as:

1 x 1
sinh x = (e − e−x ) = (28)
2 cosech x
1 x 1
cosh x = (e + e−x ) = (29)
2 sech x
whence

ex − e−x 1
tanh x =
x −x
= .
e +e coth x
We readily find that cosh z + sinh z = ez and cosh2 z − sinh2 z = 1.

The derivatives of hyperbolic functions are easily calculated:

ex −e−x
If y = sinh x = 2
, then
d sinh x ex + e−x
= = cosh x .
dx 2
dy dy
Similarly if y = cosh x, dx
= sinh x and also if y = tanh x, dx
= sech2 x .

For inverse hyperbolic functions we take y = sinh−1 xa as an example. If y = sinh−1 x

a
then x = a sinh y
and
dy 1 1 1
= = √ =√ .
dx a cosh y a 1 + sinh y2 x2 + a2
Thus
d ( x) 1
sinh−1 =√ .
dx a x2 + a2

7
If we have
x dx dy 1 1 a
y = tanh−1 , ie x = a tanh y, = a sech2 y and = ( )= 2 .
a dy dx a 1 − tanh y
2
a − x2

Parametric differentiation

Often we have variables which are functions of a parameter, e.g. the time t: x = x(t) and y = y(t), but
dy
we need dx . We make use of the chain rule and the derivative of an inverse function:

dy
dy dy dt dt
= = dx
(30)
dx dt dx dt

Example: The coordinate of a moving vehicle are given by x = −t2 , y = (1/3)t3 , where t is time. Find
dy/dx when t = 2.

The direction at the point when t = 2 is given by the tangent to the trajectory at this point. Hence
we need: 
dy 
(31)
dx t=2
dy
dy dy dt dt t2 −t
= = = = (32)
dx dt dx dx
dt
−2t 2
which for t = 2 is 
dy 
= −1 . (33)
dx t=2

3.8 Stationary points

Derivatives give the rate of change of a function. An important application is finding the maximum and
minimum of functions. If at some point, x0 we have

f ′ (x0 ) = 0

then this is a stationary point. For maxima and minima, f ′ (x) changes sign around x0 . In summary, at a
stationary point, where f ′ (x0 ) = 0, we have three possibilities

1. Minimum: f ′ (x0 ) = 0 and f ′′ (x0 ) > 0.

2. Maximum: f ′ (x0 ) = 0 and f ′′ (x0 ) < 0.

3. Point of inflection: f ′ (x0 ) = 0 and f ′′ (x0 ) = 0 and f ′′ changes sign through the point.

Example

Find the stationary point(s) of

f (x) = x ln x

8
and determine its (their) nature.
1
f ′ (x) = ln x + x = ln x + 1
x
we need to find x for which f ′ (x) = 0, i.e. ln x = −1
1
x = e−1 = .
e
To determine the nature of the stationary point we have to calculate the second derivative at x0 = e−1 .
1
f ′′ (x) = therefore f ′′ (x0 ) = e > 0
x
Thus the stationary point is a minimum.

Example

Find the stationary points of f (x) = x4 .

we start by calculating the derivative:

f ′ (x) = 4x3
thus x = 0 is a stationary point. To determine its nature, we calculate the second-order derivative

f ′′ (x) = 12x2 thus f ′′ (0) = 0

However f ′′ does not change sign around x = 0. So I cannot conclude that x = 0 is a point of inflection.
It is actually a minimum (prove it as an exercise).

3.9 Differentiation of a vector [see Riley, section 10.1]

The derivative of a vector function a(t) with respect to t is defined as
da a(t + h) − a(t)
= lim . (34)
dt h→0 h
In cartesian coordinates:
a = ax i + ay j + az k
where ax , ay , az are functions of t. Then
da dax day daz
= i+ j+ k.
dt dt dt dt
Some properties:
dca da
=c (c is a constant)
dt dt
d(a + b) da db
= +
dt dt dt
9
 d(a · b) db da
=a· + ·b
dt dt dt
d(a × b) db da
=a× + ×b
dt dt dt
For example, if
r = xi + yj + zk
is the position vector of a particle as a function of time, then the velocity and acceleration vectors are
given by:
d(r) dx dy dz
v= = i+ j+ k
dt dt dt dt
d(v) d2 r d2 x d2 y d2 z
a= = 2 = 2i+ 2j+ 2k
dt dt dt dt dt
Example

Consider the motion of a particle in a circle at constant speed. Show that the velocity vector, v, is
perpendicular to the position vector r of the particle; that the acceleration vector is perpendicular to v;
and that
|v|2
|a| = .
|r|
We notice that r and v are not constants, but their magnitudes are. So
|r|2 = r · r = constant

|v|2 = v · v = constant
This implies
d dr dr
(r · r) = 0 ⇒ ·r+r· =0
dt dt dt
or
r·v =0 ⇒ r, v perpendicular
In the same way:
d dv dv
(v · v) = 0 ⇒ ·v+v· =0
dt dt dt
or
v·a=0 ⇒ v, a perpendicular
Now we do
d dr dv
(r · v) = ·v+r· =v·v+r·a=0
dt dt dt
This implies that r and a are anti-parallel (they could only be parallel or anti-parallel from what demon-
strated earlier), i.e. r · a = −|r||a|. Therefore
|v|2
0 = v · v + r · a = |v|2 − |r||a| ⇒ |a| =
|r|

10