Differentiation − Dr A. J.

Bevan 1

October 20, 2010

Mathematical Techniques: Revision Notes

Dr A. J. Bevan,

These notes contain the core of the information conveyed in the lectures. They are not a substitute for
attending the lectures and none of the examples covered are reproduced here. Worked examples of the
techniques described in this note can be found in the tutorial question/solution material provided on the
course web site.

1 Rates of change

The change in y with respect to x between these two points on the curve is given by the ratio

δy y2 − y1
= ,
δx x2 − x1
y(x + δx) − y(x)
= . (1.1)
δx

0.55

0.5

0.45

y
0.4
2
y

0.35 δy

0.3
y
1
x1 x2
0.25
δx

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Figure 1: The function y = f (x).

If we take the limit as δx tends to zero we obtain

! "
δy dy
lim = ,
δx→0 δx dx
! "
y(x + δx) − y(x)
= lim . (1.2)
δx→0 δx
Differentiation − Dr A. J. Bevan 2

which is exact. In general the derivative obtained will be a function of x and can itself be differentiated in
order to obtain higher order derivatives.

1.1 Standard Derivatives

It is tedious to work out the derivatives of common functions using Eq. (1.2) each time, and in practice we
rely on tables or books of ’standard derivatives’. Table 1 lists a number of useful results.

Table 1: Table of standard derivatives.

dy
y = f (x) dx
xn nx n−1

ex ex
ekx kekx
ax ax ln a
1
ln x x
1
loga x x ln a
sin x cos x
cos x − sin x
tan x sec2 x
cot x −cosec2 x
sec x sec x tan x
cosecx −cosecx cot x
sinh x cosh x
cosh x sinh x
tanh x 1/ cosh2 x
sin−1 x √ 1
1−x2
cos−1 x √ −1
1−x2
1
tan−1 x 1+x2
sinh−1 x √ 1
x2 +1
cosh−1 x √ 1
x2 −1
tanh−1 x 1
1−x2

2 Chain Rule

Consider the function y = f (u), where u = g(x). Here y is a function of u, which itself is a function of some
other variable x. The chain rule states that for such a function

dy dy du
= . (2.1)
dx du dx
Differentiation − Dr A. J. Bevan 3

3 Differentiation of a Products

The rule for the differentiation of the product of two functions u(x) and v(x) is

d du dv
(uv) = v +u , (3.1)
dx dx dx

where it is assumed that both u and v are differentiable with respect to x.

4 Differentiation of Quotients

The rule for the differentiation of the ratio of two functions u(x) and v(x) is

d # u $ v du − u dv
= dx 2 dx , (4.1)
dx v v

where it is assumed that both u and v are differentiable with respect to x.

5 Logarithmic Differentiation

If one has a more complicated function, such as

un (x)v m (x)
y= , (5.1)
hp (x)

while it is possible to recursively use the quotient and product rules to compute the derivative, it is possible,
and often convenient to use natural logarithms to simplify the problem at hand. Taking the log of both sides
of Eq. (5.2) one obtains

ln y = n ln u(x) + m ln v(x) − p ln h(x), (5.2)

which can be differentiated, noting that y is a function of x so that

d 1 dy
(ln y) = , (5.3)
dx y dx

thus

1 dy n du m dv p dh
= + − , (5.4)
y dx u dx v dx h dx

which can be used to determine the derivative as a function of x given the form of u, v and h.

6 Implicit Differentiation

It is possible to differentiate functions that are implicitly relating y to x using the chain rule (as with the
previous example). For example, consider

x2 + y 2 = R2 , (6.1)
Differentiation − Dr A. J. Bevan 4

which describes a circle. The derivative of this equation is

dy
2x + 2y = 0, (6.2)
dx

so

dy x
=− . (6.3)
dx y

7 Parametric Functions

If we have a function defined by

y = h(θ) and x = g(θ), (7.1)

then we can compute the derivative of y with respect to x by using the chain rule, and noting that

dy dy dθ
= × . (7.2)
dx dθ dx

assuming that it is possible do differentiate both h and g with respect to θ.

8 Tangents and normals to a curve

Consider a straight line where

y = mx + C, (8.1)

and

dy
m = , (8.2)
dx

which is the slope of the line. Given a point on the line (x1 , y1 ) it is possible to compute the equation of the
line by noting that

y − y1 = m(x − x1 ), (8.3)
y = mx − mx1 + y1 , (8.4)

which is the equation of the tangent to a curve with gradient m at the point (x1 , y1 ).
The equation of the normal to a curve at (x1 , y1 ) can be obtained by recalling that the gradient of the normal
to a curve is −1/m. So the equation of the normal to a curve is just

1 1
y=− x + x1 + y1 . (8.5)
m m
Differentiation − Dr A. J. Bevan 5

9 Differentiating inverse trigonometric functions

The table of standard derivatives (Table 1) lists the derivatives of the trigonometric functions. These can
be used in order to derive the rules for differentiating inverse trigonometric functions. If we consider the
equation

y = arcsin(x),

we are able to take the sine of both sides to obtain

sin(y) = x.

We can now differentiate x with respect to y which yields

dx
= cos(y),
dy
%
= 1 − x2 ,

using sin2 (y) + cos2 (y) = 1. We can now readily obtain the derivative of arcsin(x) which is

dy 1
= √
dx 1 − x2

It is possible to show that the derivative of y = arccos(x) is

dy 1
= −√
dx 1 − x2

and similar steps can be taken in order to show that for y = arctan(x)

dy 1
= .
dx 1 + x2

10 Radius of curvature

A useful property of a curve is the radius of curvature R. The radius of curvature is related to the change
in direction δθ of the curve as a function of the change in position δs along the curve. The meaning of this
quantity can be seen from the following: In the limit that δs tends to zero, the arc segment of length ds for
the function y = f (x) forms part of the circumference of a circle (See Fig. 2). The radius of this circle is R,
and the center of the radius of curvature is the point (xc , yc ). We can relate ds to R via the angle subtended
by the arc

ds
dθ = ,
R

so,

dθ 1
= .
ds R
Differentiation − Dr A. J. Bevan 6

2.2
(xc, yc)
2

1.8

1.6

δθ
y

1.4 R

1.2
δs
1 (x , y )
1 1

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

x

Figure 2: The function y = f (x) showing the radius of curvature R and center of curvature (xc , yc ) of the
function for a point (x1 , y1 ) on the curve.

Recall that

dy
= tan θ.
dx

If we differentiate both sides of the last equation with respect to the position along the curve s we obtain
! "
d dy d
= [tan θ]
ds dx ds
! "
d dy dx d dθ
= [tan θ]
dx dx ds dθ ds
d2 y dθ
cos θ = sec2 θ
dx2 ds
d2 y dθ
= sec3 θ
dx2 ds

If we recall that 1 + tan2 θ = sec2 θ and that tan θ = y $ , we can substitute for sec3 θ to obtain

d2 y dθ
= (1 + tan2 θ)3/2
dx2 ds
& ! "2 '3/2
d2 y dy dθ
= 1+ .
dx2 dx ds

dθ
We can re-arrange the equation and substitute ds for R to obtain

( # $2 )3/2
dy
1+ dx
R= d2 y
(10.1)
dx2
Differentiation − Dr A. J. Bevan 7

On inspection of Eq. (10.1) we can see that the sign of R is given by the sign of the second derivative of
the function y. If the second derivative is negative at a point (x, y), then the curve is convex at that point.
Conversely if the second derivative is positive at (x, y) then the curve is concave at that point.
The point (xc , yc ) corresponding to the center of the circle with radius of curvature R is given by

xc = x1 − R sin θ,
yc = y1 + R cos θ.

11 Stationary points: Maxima, Minima and Inflections

We are able to use the derivatives of a function to obtain more information about features of how the
value of a function f changes with x. In particular we can identify the positions where the function f is
locally at a maximum or minimum value as illustrated in Figure 3. Generically we call maxima and minima
turning points. Maxima have a positive gradient of f (x) for x < xmaximum , and a negative gradient for
x > xmaximum . At the point x = xmaximum the gradient is zero. The minima of f (x) can be identified by
noting that the gradient is negative for x < xminimum , zero for x = xminimum and positive for x > xminimum .
It is not sufficient to identify minima and maximum solely by y $ = 0, as we have not considered all of the
possibilities as to how the gradient changes. We know that the gradient is changing as we scan through
a turning point, and that the sign of the gradient changes sign as we do this. Thus, the value of y $$ at a
turning point is non-zero. In fact the value of y $$ at a turning point is negative for a maximum and positive
for a minimum.

3.06

3.05

3.04 (x,y)max

3.03
(x,y)
inflection
3.02
y

3.01

3
(x,y)min
2.99

2.98
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2
x

Figure 3: The function y = f (x) illustrating a local maximum, minimum, and point of inflection.

If y $$ (x) = 0 for some value of x, then this point is called a point of inflection. In order for the second
derivative to be zero at a point of inflection xI the value of the second derivative has to change sign when we
scan from values of x < xI through x = xI to x > xI . Collectively turning points and points of inflection are
called stationary points. The values of y $ (x) and y $$ (x) in the vicinity of stationary points are summarised
in Table 2.
Differentiation − Dr A. J. Bevan 8

Table 2: Derivative values near stationary points: positive values are indicated by + and negative values are
indicated by −.

y$ y $$
Maximum
x < turning point + + or −
x at the turning point 0 −
x > turning point − + or −
Maximum
x < turning point − + or −
x at the turning point 0 +
x > turning point + + or −
Inflection
x < turning point − or + + or −
x at the turning point 0 0
x > turning point + or − + or −

12 Partial Differentiation

Consider the function z = f (x, y), which can be represented as a surface in three dimensions (See Fig. 4).
The variables x and y are orthogonal, which is another way of saying that they are independent. This means
if we want to differentiate z with respect to x, we can consider y as a constant while doing so. Similarly,
if we want to differentiate z with respect to y, we can consider x as a constant. When we do this, we use
slightly different notation than when differentiating a function of a single variable. We use the curly d symbol
’∂’ called ’del’ to indicate a partial derivative, i.e. that we are differentiating a function of more than one
variable. So we can write
∂z
∂x
as the partial derivative of z with respect to x. This quantity is the rate of change of z with respect to x for
a constant y. Similarly
∂z
∂y
is the partial derivative of z with respect to y, or the rate of change of z with respect to y for a constant x.
Higher derivatives of z can be calculated, but it is possible to differentiate either with respect to y or with
respect to x, so we have four choices of the second derivative of z. If we differentiate ∂z/∂x we can obtain

∂2z
,
∂x2
∂2z
,
∂y∂x

by differentiating with respect to x or y, respectively. Similarly we obtain

∂2z
,
∂y 2
∂2z
,
∂x∂y
Differentiation − Dr A. J. Bevan 9

2
1.8
1.6
1.4
1.2
z

1
0.8
0.6
0.4
0.2
0
1
0.8
0.6
0.4
0.2
-0
y -0.2 1
-0.4 0.6 0.8
-0.6 -0 0.2 0.4
-0.8 -0.6 -0.4 -0.2
-1 -1 -0.8 x

Figure 4: The curve f (x, y) about a saddle point.

by differentiating ∂z/∂y with respect to y or x, respectively. A consequence of the orthogonality of y and x

is that the same result is obtained when differentiating z by y then x, and when differentiating z by x then
y, to put this another way:

∂2z ∂2z
= .
∂y∂x ∂x∂y

NOTE: With the function y = f (x) it is possible to invert a derivative using the relation

dy 1
= dx .
dx dy

It can be convenient to use this relation if it is simpler to invert and differentiate a function with respect to
y than with respect to x. The same is not true with partial differentiation! We can see this by considering
derivatives of the following function z = x2 + y. The derivative of z with respect to x is simply 2x. However
√
in order to differentiate x with respect to z, we must first re-write the function as x = z − y. The derivative
√
of x with respect to z is 1/(2 z − y).

13 Revisiting stationary points

We can extend the discussion in Section 11 to functions of more than one variable. For a function f (x, y) it
is possible to use the tests summarised in Table 2 to identify stationary point either in x or in y. In order
to identify local maximum we require that a function f (x, y) has a maximum in both x and y. Similarly for
a point to be at a local minimum of the function we require that both x and y are at minimum. As shown
in Figure 4 it is possible that a function has a minimum in x and a maximum in y (or visa versa) at a given
point (xs , ys ). When this situation arises we call the point (xs , ys ) a saddle point.
Differentiation − Dr A. J. Bevan 10

14 Total Derivative

For some function z = f (x, y), it can be useful to know how z changes for small changes in both x and y.
If we recall that x and y are independent variables, any small change δx in the variable x is independent of
a change δy in the variable y. Let us consider the case where y is constant, and x changes by δx. We can
write the change in z simply as
∂z
δz = δx.
∂x
If we now consider x to be constant, and allow y to change by some small amount δy, we are able to write
down the change in z as
∂z
δz = δy.
∂y

Each of these solutions results from the fact that x and y are independent, and z = f (x, y) can be reduced into
a one-dimensional problem. Now we are ready to consider the general case where both x and y independently
change by some small amount, and calculate the change in z which is
∂z ∂z
δz = δx + δy. (14.1)
∂x ∂y
This is called the total derivative, and it is the sum of the change in z expected as a result of a small change
in x, and the change in z expected for a small change in y.

14.0.1 Rates of Change with Respect to Time

Consider the change z when x and y change in an interval of time δt; Eq. (14.1) becomes (dividing by δt)

∂z ∂z dx ∂z dy
= + .
∂y ∂x dt ∂y dt
In the limit where δt → 0:
dz ∂z dx ∂z dy
= + . (14.2)
dt ∂x dt ∂y dt

15 Change of Variables

So in general for some function z = f (x, y), where x = x(u), and y = y(u)

dz ∂z dx ∂z dy
= + ,
du ∂x du ∂y du

We can go a step further. If some function is given by z = f (x, y), where x = h(u, v), and y = g(u, v), where
u and v are two orthogonal variables, then z is also a function of u and v and we can write:

∂z ∂z ∂x ∂z ∂y
= + , (15.1)
∂u ∂x ∂u ∂y ∂u
∂z ∂z ∂x ∂z ∂y
= + . (15.2)
∂v ∂x ∂v ∂y ∂u
Differentiation − Dr A. J. Bevan 11

15.1 Changing between polar and cartesian co-ordinates

We are able to write a function z = f (x, y) in terms of polar coordinates in terms of the distance from the
origin r, and an angle subtended between the radial distance to a point on the function, the origin and the
x−axis θ. The angle θ has a value between zero and 2π. We can determine expressions for the cartesian
coordinates x and y in terms of r and θ. These are simply

x = r cos(θ),
y = r sin(θ).

We can write the derivative of z in terms of r and θ by noting that

∂f ∂f ∂x ∂f ∂y
= + ,
∂r ∂x ∂r ∂y ∂r
∂f ∂f ∂x ∂f ∂y
= + .
∂θ ∂x ∂θ ∂y ∂θ

Where,

∂x
= cos θ,
∂r
∂x
= −r sin θ,
∂θ

and,

∂y
= sin θ,
∂r
∂y
= r cos θ.
∂θ

Using these results we can write

∂f ∂f ∂f
= cos θ + sin θ ,
∂r ∂x ∂y
∂f ∂f ∂f
= −r sin θ + r cos θ .
∂θ ∂x ∂y

We are able to use this result to determine the derivative of f with respect to r and θ for any differentiable
function f = f (x, y).