Concept of derivative

©2019 Anders Gustafsson

Introductory example

Suppose you’re travelling by car. Let s(t) be the distance travelled from the
beginning of the journey until the time moment t.

The distance covered in the time interval [t, t + h], is s(t + h) − s(t). The average
speed is
s(t + h) − s(t)
vavg =
h

37
©2019 Anders Gustafsson
Introductory example

Suppose you’re travelling by car. Let s(t) be the distance travelled from the
beginning of the journey until the time moment t.

The distance covered in the time interval [t, t + h], is s(t + h) − s(t). The average
speed is
s(t + h) − s(t)
vavg =
h

The average speed on a very small interval is close to the instant speed. We say
that the instant speed at a moment t is
s(t + h) − s(t)
v(t) = lim
h→0 h

37
©2019 Anders Gustafsson
Derivatives

Definition
Given a function f (x), it’s derivative at a point a is

f (a + h) − f (a)
f � (a) = lim
h→0 h
If the above limit exists, then we say that the function f is differentiable at the
point a. The set of functions differentiable at x = a is denoted by D(a), so if we
write f ∈ D(a), it means that the limit f � (a) exists.

Remark: When we say that the derivative f � (a) exists, we mean that the limit
above exists (it’s a real number), i.e. that f is differentiable at a.
Remark: By putting x = a + h, we can rewrite the expression as
f (x) − f (a)
f � (a) = lim 38
x→a x−a
©2019 Anders Gustafsson
Derivative as slope of the tangent

Let f (x) be a function that is differentiable at x = a.

The chord (line) passing through the points (a, f (a)) and (a + h, f (a + h)) on f ’s
graph, is given by
f (a + h) − f (a)
y= (x − a) + f (a).
h
As h tends to 0, this line approaches the tangent line at (a, f (a)).

39
©2019 Anders Gustafsson
Distance, time and speed

f (a+h)−f (a)
The slope of the chord is h .

Its limit is f � (a) as h → 0.

40
Hence, f � (a) is the slope of the tangent. ©2019 Anders Gustafsson
Tangent line

We get the equation for the tangent line:

y = f � (a) · (x − a) + f (a).

Note that this equation only makes sense if f � (a) is a finite real number, that is,
for f ∈ D(a).

For an interactive graph illustrating what we just did, go to

[Link]

41
©2019 Anders Gustafsson
Continuity and differentiability

©2019 Anders Gustafsson

Differentiability implies continuity

Theorem
If a function f is differentiable at a point a, then f is also continuous at the point
a. Briefly, we can write it as D(a) ⊂ C(a).

Proof:

42
©2019 Anders Gustafsson
Contrapositive statement

The last theorem is most often used in its contrapositive form, i.e.:

If f is not continuous at x = a, then it is also not differentiable at x = a.

43
©2019 Anders Gustafsson
Reverse implication?

On the other hand, if f is continuous at x = a, can we then conclude that it is

differentiable at x = a?

The answer is a big NO.

There are many ways this can fail.

44
©2019 Anders Gustafsson
Example

√
What about f (x) = 3
x at x = 0?

45
©2019 Anders Gustafsson
Vertical tangent

The graph of f (x) = x1/3 has a vertical tangent at (0, 0).

f (0 + h) − f (0)
f � (0) = lim , does not exist.
h→0 h
(The above limit diverges to +∞.) 46
©2019 Anders Gustafsson
Example

What about f (x) = sin |x| at x = 0?

47
©2019 Anders Gustafsson
A cusp

f (0 + h) − f (0)
f � (0) = lim , does not exist.
h→0 h
(The left and right hand limits are different.)
48
©2019 Anders Gustafsson
Example

What about 
 1
 x sin x , x �= 0

f (x) =



0, x=0
at x = 0?

49
©2019 Anders Gustafsson
Example

What about 
 1
 x sin x , x �= 0

f (x) =



0, x=0
at x = 0?

49
©2019 Anders Gustafsson
Wild behaviour

f (0 + h) − f (0)
f � (0) = lim , does not exist.
h→0 h
50
(Wild behaviour as h → 0.)
©2019 Anders Gustafsson
How to find derivatives

©2019 Anders Gustafsson

Differentiation rules

Theorem
Given two functions f and g, we have whenever the right hand side makes sense
that

Sum rule For any real constants c, d,

(cf + dg)� (x) = cf � (x) + dg� (x)

Product rule
(fg)� (x) = f � (x)g(x) + f (x)g� (x)

Ratio rule � ��
f f � (x)g(x) − f (x)g� (x)
(x) =
g (g(x))2
Chain rule
2
[f (g(x))]� = f � (g(x)) · g� (x)
©2019 Anders Gustafsson
Remark

Note the condition that the right hand side must make sense for the equation to
hold.

In particular the theorem tells us that

• If f and g are both differentiable at x, then so is cf + dg.
• If f and g are both differentiable at x, then so is fg.
• If f and g are both differentiable at x and g(x) �= 0, then f /g is differentiable
at x.
• If g is differentiable at x and f is differentiable at g(x), then f (g(x)) is
differentiable at x.

In case the right hand side does not make sense, the theorem gives us no
information.
3
©2019 Anders Gustafsson
Proof of the product rule

Suppose that the right hand side f � (x)g(x) + f (x)g� (x) makes sense, i.e. that both
f and g are differentiable at x.
Then:
f (x + h)g(x + h) − f (x)g(x)
(fg)� (x) = lim =
h→0 h

f (x + h)g(x + h) − f (x)g(x + h) + f (x)g(x + h) − f (x)g(x)

lim =
h→0 h

f (x + h) − f (x) g(x + h) − g(x)

lim g(x + h) + lim f (x) =
h→0 � �� �
� h
�� � h→0 � h
�� �
→g(x) why?
→f � (x) →g� (x)

f � (x)g(x) + f (x)g� (x). 4

©2019 Anders Gustafsson
Other proofs

Sum rule Exercise (use the definition of derivative and basic limit laws).
Quotient rule Exercise. First use the definition of derivative to find the derivative
of 1/g(x). Then apply the product rule (which we have already
1
proved) to f (x) · g(x) .
Chain rule You can look up the proof in a calculus textbook, such as “Calculus”,
4th ed., by Spivak, Ch. 10, Thm 9.

5
©2019 Anders Gustafsson
Derivatives of some elementary functions

Notation:
d
f (x) = f � (x).
dx

Theorem
The following formulae hold
d d α d x
dx C =0 dx x = αxα−1 dx e = ex

d 1 d d
dx ln x = x dx sin x = cos x dx cosx = − sin x

d
arcsin x = √ 1 d 1
arccos x = − √1−x d
arctan x = 1
dx 1−x2 dx 2 dx 1+x2

6
©2019 Anders Gustafsson
Derivatives of some elementary functions

Notation:
d
f (x) = f � (x).
dx

Theorem
The following formulae hold
d d α d x
dx C =0 dx x = αxα−1 dx e = ex

d 1 d d
dx ln x = x dx sin x = cos x dx cosx = − sin x

d
arcsin x = √ 1 d 1
arccos x = − √1−x d
arctan x = 1
dx 1−x2 dx 2 dx 1+x2

6
©2019 Anders Gustafsson
Proof that d
dx
C =0

If f (x) = C is a constant, then

C−C
f � (a) = lim =0
h→0 h
Done.

7
©2019 Anders Gustafsson
Proof that d n
dx
x = nxn−1 where n ∈ N

Recall, the binomial theorem:

n � � � �
� n n−k k n n!
(x + y)n = x y , where = .
k k k!(n − k)!
k=0

8
©2019 Anders Gustafsson
Proof that d n
dx
x = nxn−1 where n ∈ N

Recall, the binomial theorem:

n � � � �
� n n−k k n n!
(x + y)n = x y , where = .
k k k!(n − k)!
k=0

d n
Proof that dx x = nxn−1 when n ∈ N
If f (x) = xn , n = 1, 2, 3, . . ., then

(a + h)n − an (an + nan−1 h + n(n−1)

2 an−2 h2 + . . . + hn ) − an
f � (a) = lim = lim =
h→0 h h→0 h
n(n − 1) n−2
= lim (nan−1 + a h + . . . + hn−1 ) = nan−1 .
h→0 2
Done
8
©2019 Anders Gustafsson
Proof that d p
dx
x = pxp−1 where p ∈ Z

Already done for p ∈ Z, p ≥ 0.

For p ∈ Z, p < 0, we have n = −p ∈ N, so
1
f (x) = xp =
xn
and
f � (x) = exercise!

Use the quotient rule and the already established power rule for n ∈ N.

9
©2019 Anders Gustafsson
 1
Proof that d 1/q
dx
x = q1 x q −1 where q ∈ Z \ {0}

Now consider y = x1/q , q ∈ Z \ {0}.

10
©2019 Anders Gustafsson
 p
Proof that d p/q
dx
x = pq x q −1 , for p, q ∈ Z q �= 0

Exercise
Hint, notice that
xp/q = (x1/q )p ,

and use the chain rule.

11
©2019 Anders Gustafsson
d α
dx
x = αxα−1 , α ∈ R

We have now proved

d α
x = αxα−1 , for all α ∈ Q.
dx

The formula also holds for irrational α, as well. It is based on the fact (which we
haven’t mentioned until now) that for irrational powers r and real x > 0 the
relevant definition is
xr = er ln x .

12
©2019 Anders Gustafsson
Proof that d
dx
sin x = cos x

With f (x) = sin x. We have

sin(a + h) − sin a
f � (a) = lim =
h→0 h
sin a cos h + cos a sin h − sin a
lim =
h→0 h
 
 cos h − 1 sin h 
lim  sin a + cos a  = cos a.
h→0  h h 
� �� � � �� �
→0 because (*) →1

cos h − 1 cos2 h − 1 − sin2 h sin h sin h 0

(∗) = = =− · → −1· = 0 as h → 0.
h h(cos h + 1) h(cos h + 1) h cos h + 1 2
13
©2019 Anders Gustafsson
More exercises

Prove that
d
cos x = − sin x
dx
and that
d 1
tan x = .
dx cos2 x

14
©2019 Anders Gustafsson
A comment on the definition of cos θ and sin θ

Students often have trouble explaining how to find the values of cos θ and sin θ,
when θ �∈ (0, π/2) (i.e. when θ is not an angle from a right triangle).

A common answer is something like “use the trigonometric identities”, e.g. the
identity sin(π − θ) = sin θ gives us

5π π 1
sin = sin = .
6 6 2

But why is it true that sin(π − θ) = sin θ, and how do we remember this and
similar identities?

15
©2019 Anders Gustafsson
Definition of cos and sin

Instead of memorizing (or looking up) a big number of trigonometric formulas

that may not make much sense, a better strategy is to understand the following
definition, which defines cos θ and sin θ for an arbitrary θ ∈ R.

16
©2019 Anders Gustafsson
Definition of cos θ and sin θ for arbitrary θ ∈ R

Definition
Let θ ∈ R and consider the point (x, y) on the unit circle (circle with radius 1,
centred at the origin) whose angle with the positive x-axis is θ (counterclockwise
for θ > 0, clockwise for θ < 0). We define

cos θ = x, sin θ = y.

17
©2019 Anders Gustafsson
Identities for cos(−θ) and sin(−θ).

It becomes obvious why

cos(−θ) = cos θ, sin(−θ) = − sin θ.

18
©2019 Anders Gustafsson
Identities for cos(−θ) and sin(−θ).

It also becomes obvious why

cos(π − θ) = − cos θ, sin(π − θ) = sin θ.

19
©2019 Anders Gustafsson
Identities for cos(−θ) and sin(−θ).

And also why

cos(θ + π) = − cos θ, sin(θ + π) = − sin θ.

20
©2019 Anders Gustafsson
Proof that d
dx
arcsin x = √ 1
1−x2

For y = arcsin x, note that

� π π�
y = arcsin x ⇒ sin y = x, y∈ − ,
2 2

dy � π π� dy 1 1
⇒ cos y · = 1, y∈ − , ⇒ = =√ .
dx 2 2 dx cos y 1 − x2

21
©2019 Anders Gustafsson
Exercise

Prove that
d 1
arccos x = − √ .
dx 1 − x2

22
©2019 Anders Gustafsson
Proof that d
dx
arctan x = 1
1+x2

For y = arctan x, note that

� π π�
y = arctan x ⇒ tan y = x, x∈ − ,
2 2
y� 1
⇒ =1 ⇒ y � = cos2 y = .
cos2 y 1 + x2

23
©2019 Anders Gustafsson
Interpretation of derivative

In the sciences, derivative means rate of change.

Example: If s(t) is the distance travelled (m) at time t (s), then s� (t) is the velocity
(m/s) at time t and s�� (t) is the acceleration m/s2 at time t.

Example: If Q(t) is the amount of charge (C) that’s passed through a cross section
of a wire at time t (s), then Q� (t) is the current (C/s=A).

24
©2019 Anders Gustafsson
Example

A motorcyclist is travelling along a road. On an overhead bridge is a traffic police

with a radar gun, positioned 6 m above the motorcycle. When the motorcycle is
passing a particular lamp post, the police knows that the distance x from the
motorcycle to the bridge is 8 m. At this moment, his instrument tells him that the
distance d between him and the motorcycle is decreasing with 52 km/h. If the
speed limit on this particular road is 60 km/h, can the motorcyclist be fined?

25
©2019 Anders Gustafsson
Example (continued)

Both x = x(t) and d = d(t) depend on time t.

However, it’s true that
(x(t))2 + 62 = (d(t))2 ,
so differentiating (with respect to t) we have

2x · x� = 2d · d� ,

i.e.
d · d�
x� =
.
x
At the moment when x = 0.008 km, we have d = 0.01 km, so
0.01 · (−52)
x� = km/h = −65 km/h.
0.008

26
©2019 Anders Gustafsson
Kena?

This carries a fine of S$150 and 4 demerit points (as of May 2019).

27
©2019 Anders Gustafsson
Kena?

This carries a fine of S$150 and 4 demerit points (as of May 2019).

However, this police officer hasn’t taken Maths I.

27
©2019 Anders Gustafsson
Local maxima and minima

©2019 Anders Gustafsson

Optimization

Consider a natural problem: suppose we know that some process is described by

a function and we need to find its maximal value. For example, if a businessman
knows how the income depends on the price, it’s crucial to find the price that
maximizes the income.

28
©2019 Anders Gustafsson
Global maximum/minimum

Definition
Consider a function f : A → R. A point x0 ∈ A is called a

point of global maximum if

f (x) ≤ f (x0 ), for all x ∈ A.

point of global minimum if

f (x) ≥ f (x0 ), for all x ∈ A.

29
©2019 Anders Gustafsson
Local maximum/minimum

Definition
Consider a function f : A → R. A point x0 ∈ A is called a

point of local maximum if there exists some open interval (a, b) � x0 such that

f (x) ≤ f (x0 ), for all x ∈ (a, b) ∩ A.

point of local minimum if there exists some open interval (a, b) � x0 such that

f (x) ≥ f (x0 ), for all x ∈ (a, b) ∩ A.

30
©2019 Anders Gustafsson
Finding local maximum or minimum points

You probably recall that finding local extreme points has something to do with
solving the equation f � (x) = 0.

While true, this is only part of the story.

Here are some examples where solving the equation f � (x) = 0 will not give you all
local extreme points.

31
©2019 Anders Gustafsson
Example, f (x) = sin |x|

f (x) = sin |x| has a point of local minimum at x = 0 (why?).

f � (0) does not exist.

32
©2019 Anders Gustafsson
 �
sin x1 ; for x �= 0
Example, f (x) =
−1, ; for x = 0.


sin 1
x ; for x �= 0
f (x) =
−1, ; for x = 0,
has a point of global minimum at x = 0 (why?).

f � (0) does not exist (why?).

33
©2019 Anders Gustafsson
Example, f : [0, ∞) → R, f (x) = x − x3

f : [0, ∞) → R, f (x) = x − x3 , has a point of local minimum at x = 0 (why?).

f � (0) = 1 �= 0.

34
©2019 Anders Gustafsson
Derivatives and extreme points

Theorem
If a is a local max/min point of a function f whose domain contains some interval
(a − δ, a + δ) (for some δ > 0), and if f is differentiable at a, then f � (a) = 0.

35
©2019 Anders Gustafsson
Derivatives and extreme points

Theorem
If a is a local max/min point of a function f whose domain contains some interval
(a − δ, a + δ) (for some δ > 0), and if f is differentiable at a, then f � (a) = 0.

Proof (for a local minimum point a):

For sufficiently small h, we have f (a + h) ≥ f (a). Thus,

f (a + h) − f (a) f (a + h) − f (a)
≥ 0 for h > 0, ≤ 0 for h < 0
h h
Since the derivative exists, the quotient above has a limit as h → 0, and so
f (a + h) − f (a) f (a + h) − f (a)
lim = lim = f � (a).
h→0+ h h→0− h

We can conclude that f � (a) = 0 (why?).

35
©2019 Anders Gustafsson
Finding extreme points

By the previous theorem, points of local maxima / minima, may only occur at
points c where
1. f � (c) = 0.
2. f � (c) does not exist.
3. There is no δ > 0 such that (c − δ, c + δ) is contained in the domain of f .
Points where Case 1 or Case 2 happens, are called critical points.
Definition
A critical point of a function f : A → R is c ∈ A such that either f � (c) = 0 or f � (c)
does not exist.

Moreover, if the domain is a closed bounded interval [a, b], Case 3 above occurs
exactly at the endpoints a and b.
36
©2019 Anders Gustafsson
Finding the global max/min of a continuous function on [a, b].

So, for a continuous function f : [a, b] → R we know that

• Global minimum and maximum points exist in [a, b] (by the max/min thm).
• These points must be points c ∈ [a, b] where
1. c is a critical point of f (i.e f � (c) = 0, or f � (c) does not exist), or
2. Any of the endpoints a or b.
By comparing the function values at these points, we can deduce what the
global max and min values and points are.

37
©2019 Anders Gustafsson
Example

Find the smallest and the largest values of

x3
f (x) = − x2 − 3x,
3
on the interval [0, 5].

38
©2019 Anders Gustafsson
Example

Find if possible, the largest and smallest values of f (x) = 1/x on the interval (0, 1).

39
©2019 Anders Gustafsson
Example

Find the largest and smallest values of

3
f (x) = x2/3 − x,
2
on [−2, 2].

40
©2019 Anders Gustafsson
Zero derivative not sufficient

By our observations for a differentiable f : (a, b) → R, c ∈ (a, b):

f has a point of local extremum at x = c ⇒ f � (c) = 0.

Not however that the reverse implication does not hold.

Example
For f (x) = x3 , f is differentiable and f � (0) = 0 but x = 0 is not a point of local
extremum.

41
©2019 Anders Gustafsson