6.

Mathematical Modelling
6.1. What is a mathematical model ?
A model is a simplified representation of a real system. There are two important concepts in
this definition.
Firstly, system is a group of objects (components or factors) that interact with one another
in a organized manner, and the net result of their interactions produces the systems
behavior, sunction and purpose. Examples of systems in chemical engineering are reaction
kinetics, equilibrium, industrial units, and transport phenomena.
Secondly, when a system is represented or described in a simpler form known as a model,
the model becomes a tool for us to understand the system by helping us to sift through its
complexity and to focus on the important, relevant aspects.
Models can be of many types:
1. pictorial (e.g. illustrations, diagrams and flowcharts)
2. conceptual or verbal (descriptions in a natural language)
3. physical (replica or mock-up of the system such as a scale airplane model for the wind
tunnel experiments, and the bench-top distillation colums)
4. mathematical.
Consequently, a mathematical model is one type of simplified representation of a real
system. It describes the system using mathematical principles in the form of an equation or
a set of equations. For example, in a controlled agricultural experiment, the crop yield (Y)
responds to the nitrogen fertilizer rate (N) by the following mathematical relationship:
Y = 1143 + 31.7 N - 0.084 N2
Which reveals that with increasing nitrogen rate, crop yield increases but with diminishing
returns, and that past a certain threshold of nitrogen (189) the yield will start to decrease
(probably due to nitrogen toxicity).

6.2. Uses of mathematical models

Mathematical models are developed primarily because we want to solve one or more
problems in a real system. For example: how do we increase the growth and yield of crops ?
how do we increase the rate and yield of chemical reactions ? Why are the current
practices not effective in controlling global warming ?

As compared to other model types ( e.g. pictorial, conceptual and physical), mathematical
models are often the most useful type. This is because the not only tell us what are the
essential components of a real system, but more importantly, they also tell us how and to
what degree do each of these components interact with each other. Besides helping us to
understand the system (i.e. the what, why, and how of a system) mathematical models
exclusively allow us, in an organized manner, to predict and control the system.
Consequently, mathematical models provide us a tool to solve problems.

6.3 Characteristics of mathematical models

1) Models are incomplete description of whole systems.
Because a model is a simpler representation of a system, it is actually an incomplete or
imperfect description of the system. There is always some loss of information when a real
system is translated into a model. The model does not tell us the whole picture of the
system, only the essential parts that are important to explain the systems behavior,
function, and purpose. This incomplete depiction of a system means that the estimates or
simulations from a mathematical model are subject to error.
The degree of modeling error depends on how much information is lost; that is, how much of
the essential parts of the system that we have ignored in our models. It also depends on
how much we have under- or over- estimated the effects of the system components. The
key here is then for us to be able to capture and summarize the essential information of the
real system into a model. We have to identify the essential components of the real system,
and to describe adequately how and to what degree each of these components interacts
with one another into our model. What we aim to achieve at the end is a model that is
simpler than the real system but immensely useful because it helps us to study the system
and additionally, to control and anticipate the systems behavior with the desired detail and
accuracy.
We might then be tempted to develop a model that is complete, that perfectly represents a
real system in every way. But such a model, perfect though it may be, is of no use to us
because what we have actually done is not the development of a tool to aid our
understanding but the perfect duplication or cloning of the actual system itself. We would
not be better using such a model to understand the system.
In the other extreme, we must be careful too that we do not develop models that are too
simple or incomplete, that vital aspects of the system are ignored or described
inadequately, and that our understanding of the system becomes deficient, or worse, or
wrong.

2) Models are built from assumptions

Real systems are often complex and we need assumptions to simplify them for our
understanding. Without the use of assumptions, we might fail to develop a model because

there is still too much complexity or uncertainty about the system. In other words,
assumptions help us to grasp the system so that we are able to summarize the system into a
model.
As we will later see, it is vital that we test our mathematical models against reality to
ensure that our models are accurate despite the use of assumptions.

3) Model simplicity versus model accuracy

In the development of a mathematical model, there is often a trade-off between model
simplicity and model accuracy. The more holistic and in-depth we want to represent the
system, the more complex and accurate the model becomes. It becomes increasingly more
accurate because more information about the system is summarized and fewer assumptions
are used in the model. In contrast, the simpler the model, the less accurate our model
becomes because the model tells us increasingly less about the system, resulting in greater
loss of information.
Complex models, however are often cumbersome to use, often requiring intensive and
repetitive calculations and a lot of data, some of which may not be easily available. How to
decide between a complex and a simple model depends very much on the purpose of our
study. At times, only a rough estimation will suffice, so we will be better off with a simpler
model. But when high accuracy is required or in critical operations (e.g. development of a
new airplane or nuclear reactor), the use of a complex model is mandatory.
4) No one best model for all circumstances
A real system is often described by many alternative models. This because one model may
focus on one aspect of the system, not covered or covered with insufficient depth by the
other models. These models also differ from each other in terms of their model simplicity
and accuracy. Therefore, there is no one model that is suitable for all circumstances. The
best model is one that meets our interests and purpose of study, not necessarily always
being the most accurate or the most simple model.

6.4.

Functions and Equations

A function describes the relationship between two sets of variables such that for any
value of one or more independent variables, there is only one corresponding value for
the dependent variable. A function is simply a rule that takes one or more values as input
and produces one, and only one, output (Fig. 6.1). For example, the function f(x):

f ( x) 4 x 1

[6.1]

has the following rule: for every value of x (input), it is multiplied by four, and then
subtracted by one which will produce a single answer (output). If x is 5, the output is (45
1) or 19. In this case, for every value of x, there is only one output. But the following
expression expression

f ( x) x
is not a function because for a given value of x, there are two (not one) possible output. If x
is 9, for instance, the output can either be -3 or 3.

Fig. 6.1. For every input x, a function produces one and only one output y. The vertical line
shows that A is a function, but B is not.
A function may also have two or more arguments such as

g ( x, y ) x 2 y
which shows that the g function has two arguments, x and y, and the rule of the function is
to add y to x squared. The arguments of a function are also called independent variables
because their values are independent of each other and they are not modified within the
function. The output of a function is instead known as the dependent variable because its
value depends on the independent variables. Eq. 6.1 can be expressed alternatively as

y f ( x), y f ( x ) 4 x 1, or simply y 4 x 1
where the output of the function f(x) is y, and y, the dependent variable, is a function of
the independent variable, x, computed by 4x - 1.
It is important to realize that not all functions are equations. A function, as stated
previously, describes the relationship between the dependent and independent variables.
And to describe their relationship, an equation is very often (but not always) used. In this

chapter, however, both terms, functions and equations, will be treated as synonymous to
each other. Both terms will be used to refer to the mathematical relationship between the
dependent and independent variables.

6.4.1. Function types

6.4.1.1. Polynomial functions
Polynomial functions are the simplest elementary functions, and they have the
following form:
n

f ( x) b0 b1x b2 x ... bn x bi xi
2

i 0

where b0, b1, , bn are the function parameters with bn 0; and n is called the degree of
the polynomial function, and n is either zero or a positive integer. The expressions of the
first six members of the polynomial functions are summarized below:

Degree, n

Equation

Special name

f ( x ) b0

Constant

f ( x ) b0 b1x

Linear

f ( x ) b0 b1x b2 x 2

Quadratic

f ( x ) b0 b1x b2 x 2 b3 x 3

Cubic

f ( x ) b0 b1x b2 x 2 b3 x 3 b4 x 4

Quartic

f ( x ) b0 b1x b2 x 2 b3 x 3 b4 x 4 b5 x 5

Quintic

In engineering studies, polynomial functions having degrees up to three are often used, of
which the linear function is one of the most widely used because of its simplicity, and it
depicts several engineering processes well. The linear function reveals that for every unit
change in x, the amount of change in y remains constant (Fig. 6.2.a). In other words, the
rate of change in y is constant. This is similar to the constant polynomial function (n = 0)
where the rate of change in y is constantly at zero (i.e., no change in y).

(a)

(b)

(c)

(d)

Fig. 6.2. Polynomial functions: a) linear (b0 = 0, b1 = 1), b) quadratic (b0 = 0, b1 = 3.75, b2 =
-0.25), c) cubic (b0 = 0, b1 = 7.5, b2 = -1.2, b3 = 0.05), and d) quartic (b0 = 0, b1 = 13, b2 = -3.9,
b3 = 0.39, b4 = -0.0125)
At higher degrees (n > 1), polynomial functions consist of smooth curves with a series of
hills and valleys (Fig. 6.2. b-d). These hills and valleys are called turning points. The
maximum possible number of turning points is one less than the degree of the polynomial. If
f is a polynomial function of degree n, then there are at most (n-1) turning points. For
instance, a fourth-degree polynomial function has a maximum of three turning points.
The roots of a function are the values of x when the curve intersects the x-axis (more
specifically, the values of x when y = 0). A linear function has only one root, and it is
determined easily by

b0
b1

For quadratic functions, their roots are determined using the well known formula:

b1 b12 4b0b2

[6.2]

2b2

But for the cubic and quartic functions, their roots are calculated in a much more complex
manner which will not be described here.

6.4.1.2. Rational functions

A rational function is quotient (ratio) of two polynomial functions p(x) and q(x):

f ( x)

p( x)
,
q ( x)

q( x) 0

provided that the value of x is not such that the function q(x) = 0, else division by zero
occurs and the result is undefined. Some examples of rational functions are

f ( x)

2x 3
,
x2

g ( x)

x3 2 x 2 5 x 9
,
x 1

x0

x 1

One important example of a rational function is the rectangular hyperbola, in which the
asymptotes intersect at right angles. An asymptote is a line that forms the limiting value of
a curve. Consider an example of a rectangular hyperbola: the Michaelis-Menten function
(Fig. 6.3).

Fig. 6.3. Michaelis-Menten function for chemical reaction rates

The Michaelis-Menten function is often used to describe the rate of a chemical reaction
catalyzed by an enzyme, in which the reaction rate depends on the concentration of a
substrate, and it approaches maximum reaction rate for large enough substrate
concentration. This function is given as

v(C )

Vmax C
K C

[6.3]

where the reaction rate is v(C) which depends on the substrate concentration, C; Vmax is the
maximum reaction rate; and K is a constant. The value of K is such that v(K) = 0.5Vmax; that
is, K is the substrate concentration to achieve half maximum reaction rate. For this reason,
K is known as the half saturation constant. In Eq. 6.3, C can (theoretically) decrease to a
value as near as K but never reaching K, else division by zero occurs. Likewise, the reaction
rate v(C) can be increased to a value as near asbut never reachingthe maximum rate,
Vmax. Thus, Eq. 6.3 has two asymptotes: v(C) = Vmax and C = -K, and these two asymptotes are
perpendicular to each other.
A similar rectangular hyperbola function is one that describes the leaf photosynthesis rate,
A(L):

Am kI 0e kL
A L
Am kI 0e kL

[6.4]

where Am is the maximum photosynthesis rate; is the photosynthetic efficiency factor; k

is the canopy extinction coefficient for solar radiation; I0 is the solar irradiance above
canopy; and L is the cumulative leaf area index from the canopy top to the canopy height
being considered. Comparing to Eq. 6.3, Am is analogous to Vmax, and the entire term kI0e-kL
(i.e., incident solar irradiance on leaf) is analogous to C. Eq. 6.4 reveals that with increasing
incident solar irradiance on the leaf, leaf photosynthesis rate gradually approaches Vmax but
at a decreasing rate.

6.4.1.3. Power functions

A power function has the following form:

f ( x) b0 xb1
where b0 and b1 are the function parameters (Fig. .4). The power function can be expressed
alternatively as

f ( x) xb1
where f (x) is proportional to x raised to the power of b1. The parameter b1 is of particular
importance, and it is sometimes known as the scaling or allometric factor.

Many natural phenomena or processes follow the so-called power scaling law. The size and
frequency of earthquakes, for example, follow a power law relationship. Earthquakes vary in
their size or magnitude: from slight earth tremors to large, destructive ones, sufficient to
destroy parts or even a whole city. Large earthquakes are rare, but slight earth tremors are
more frequent. It is found that the number N of earthquakes of a given size is related to
the size of the earthquake, S, through an inverse power law:

N S f

[6.5]

where the scaling factor, f, is nearly equal to one for many regions in the world. Eq. 6.5 is
known as the Gutenberg-Richter law for earthquakes.

Fig. .4. Power function

A creatures heartbeat rate is widely known to decrease as its body size increases. The
heartbeat rate for a human adult, for instance, is typically 60-90 beats per minute which is
slower than that for an infant (90-140 beats per minute). Similarly, an elephant is about 100
times larger than a mouse, but its heartbeat rate is about 17 times slower than the mouses
heartbeat rate. For a wide variety of organisms, the heartbeat rate is found to be
proportional to the body mass, W, raised to the power of

heartbeat rate W

14 , or

In addition, the total metabolic rate for most organisms, ranging from small organisms such
as bacteria to the large animals such as whales, is also proportional to the body mass but
raised to power of

3
4

, or

total metabolic rate W

This scaling law can also be applied to plants, where it was recently found by Reich et al.
(2006) that plant respiration (metabolic rate) scaled isometrically (one-to-one) with plant
mass (i.e., the scaling factor is 1):

respiration W
The above relationship was revealed after measurements involving about 500 individual
plants, across 43 species, from varying environments, and covering six orders of magnitude
variation in plant mass. Reich et al. (2006) further noted that plant respiration had a
stronger isometric relationship with plant nitrogen content than with plant mass, and that
this respiration-nitrogen relationship was independent of whether the plant is located
indoors (such as in a laboratory or glass house) or in the field.
Other interesting scaling relationships also exist: the diameter of a tree trunk is
proportional to the tree mass raised to the power of

tree trunk diameter W

3
4

and that the average distance between trees of the same mass in a mature forest is
proportional to the diameter of their trunks:

mean distance of tree with same mass trunk diameter

Using a mechanistic model, Enquist (2002) estimated the power law relationship between
plant mass (W) and several plant properties as

W X

where plant property, X, and its corresponding scaling factor, f, are as shown in Table
Error: Reference source not found.
Power law relationship between plant mass, W, and plant property, X:
Property, X

Scaling factor,
f

Property, X

W X

Scaling factor,
f

No. leaves

3/4

Fluid flow rate

3/4

No. branches

3/4

Respiration rate

3/4

No. tubes

3/4

Pressure gradient

1/4

Branch length

1/4

Fluid velocity

1/8

Branch radius

3/8

Branch resistance

3/4

Area of conducting
tissue

7/8

Tree height

1/4

Tube radius

> 1/6

Reproductive biomass

3/4

Conductivity

Fluid volume

25/24

10

Leaf-specific
conductivity

1/4

6.3.1.4. Exponential and logarithmic functions

Exponential functions look similar to the power functions which were discussed earlier, with
one important difference: that the variable x in exponential functions is now the power,
rather than the base. Previously, power functions like f (x) = x3 means that the base is x,
and it is raised to the power of 3. But in the case of exponential functions, they have the
following general form:

f ( x) b0 a b1 x

[6.6]

where b0 and b1 are the function parameters; and the base a is a positive constant other
than 1. The exponential function is often expressed in terms of base e, where e is the
natural logarithm base, and it has a value of about 2.71828. The exponential function to
base e is alternatively written as exp( ) such as

f ( x) b0 exp b1x
where exp(b1x) is equivalent to

eb1x (Fig. 6.5).

Fig. 6.5. Exponential and logarithmic functions

The inverse of exponential functions is logarithmic functions (Fig. 6.5). Consider the
following exponential equation:

y c a bx

[6.7]

11

Referring to Eq. 6.6, the base in Eq. 6.7 is a, and the parameters c = b0 and b = b1.
Rearranging Eq. 6.7, we get
1

y b
x
a
c
To solve for x, we take the logarithm to base a on both sides to obtain
1

y b
x log a
c
1
1
1
log a log a y
b
b
c

[.8]

We see that the roles of x and y in the exponential function (Eq. 6.7) are reversed in the
logarithmic function (Eq. .8). In Eq. 6.7, we determine y, given x. But in Eq. .8, we instead
determine x, given y. Hence, this is why logarithmic functions are said to be the inverse of
exponential functions. We can express logarithmic functions more briefly in the following
form:

y b0 b1 log a x

[6.9]

where, comparing Eq. 6.9 to .8,

1
1
b0 log a
b
c

and b1

1
b

and the roles of x and y have been swapped; that is, the x and y in Eq. 6.9 are the y and x in
Eq. .8, respectively. For natural logarithms, they have the following form:

y b0 b1 ln x
x

In Fig. 6.5, e climbs very quickly for large values of x, but ln x increases very slowly for
x

large values of x. Looking at both curves, e and ln x are the reflection of each other about
the line y = x.

6.3.1.5. Trigonometric functions

Certain natural processes follow cyclical or periodic variations such as the diurnal variation
in air temperature and solar irradiance with time. In this case, trigonometric functions are
useful to describe such periodic patterns. Trigonometric functions actually refer to a family
of six functions: sine, cosine, tangent, cosecant, secant and cotangent, of which the sine and
cosine functions are two of the most widely used functions. The sine and cosine functions
have the following forms:
Sine:

f ( x) A sin x

C
12

f ( x) A cos x

Cosine:

where A is the amplitude (the height of each peak above the baseline); C is the vertical
offset (the height of the baseline); is the angular frequency, given by 2/P, where P is the
period of the cycle (the length of the cycle before the variation repeats); and is the
phase shift (the horizontal offset of the curve) (Fig. 6.6).

Fig. 6.6. Sine function

For stable weather and uniform skies throughout the day, the diurnal variation in
solar irradiance can be described as following a sinusoidal pattern. In Fig. 6.14, the variation
of solar irradiance begins from sunrise (6:00 hour) to sunset (18:00 hour). Exactly at
sunrise, solar irradiance is 0 W m-2, and it increases steadily to 600 W m-2 at 12:00 hour,
then declines with time, finally reaching 0 W m-2 again at sunset. This diurnal pattern can be
described by a sine function:

2
t 6 ,
24

f (t ) 600 sin

6 t 18

where t is the hour; the amplitude (A) is 600 W m-2; the period (P) is 24 hours; and the
phase shift () is 6 hours since sunrise was at 6:00 hour; this would ensure that the solar
irradiance would begin at 0 W m-2 at sunrise.

6.3.2. Some properties of functions

6.3.2.1. Linear and nonlinear functions
A function is linear if each term in the equation is either a constant or the product of a
constant and a variable to the first power. Examples of linear equations are

13

x 1 0

x 2 y 9 11z
2 x1 x2 4 x3 3
When plotted, linear equations describe straight lines in the Cartesian coordinates, and the
rate of change in the dependent variable (i.e., its slope) remains a constant.
Any function which is not linear is a nonlinear function. Unlike a linear equation, a nonlinear
equation is not a linear sum of its independent variables, and the rate of change of the
dependent variable is not a constant. Equations such as these

x2 x 1 0
x 2 xy 9 0
y

2 x1 x2 4 x3
6 x1 z

are examples of nonlinear equations.

6.3.2.2. Slope
Consider a curve which is described by the function y = f (x). Its slope or gradient at point
x is of interest because it gives us the rate of change in y per unit change in x. To
determine the slope of a curve at point x, we differentiate the function f at point x; that is,

slope

df x
dx

The slope of the curve at point x, which is the first derivative of the function f, is
sometimes expressed alternatively as dy/dx or D(x). For a curve, its slope varies from point
to point, but for a linear line, its slope is equal at every point along the line. Its slope, m, can
be determined simply by

y2 y1
x2 x1

where (x1, y1) and (x2, y2) are any two pairs of points on the line.
Example 6.1
Determine the slope of the curve at x = 3 for the following functions:

14

a)

f ( x) 2 x 2 5 x 9

b)
c)

f ( x) 4 x 5
f ( x) 4 ln 5 x

d)

f ( x) 600cos

x 5 9
12

Solution
a)
b)
c)
d)

df (3)
dx
df (3)
dx
df (3)
dx
df (3)
dx

4 x 5 12
20 x 6

20
729

1
3

50 sin
x 5 25
12

x 1

Taking the derivative of a single-argument function, such as f(x), gives us the slope at x, but
when a function has two or more arguments, such as f(x, y), the slope in the x-direction is
the derivative of f(x, y) with respect to x, while holding the other argument y constant. And
the slope in the y-direction is the derivative of f(x, y) with respect to y, while holding x
constant.
To better illustrate this point, consider the function, f(x, y) = 2x + 3y. It has two
arguments, x and y, and this function describes a surface in the three-dimensional space
(Fig. 6.7). Suppose that we wish to determine the slope of any point on the surface in the xdirection along y = 1. This means that f(x, 1) = 2x + 3, which describes a linear line (Fig.
6.8a), and its derivative is

df x,1
2
dx
which shows that along y = 1, the slope in the x-direction is 2. It can be shown that
regardless of what value we fix y, the slope at any point on the surface in the x-direction is
always 2. Now suppose that we wish to determine the slope of any point on the surface in
the y-direction along x = 1. So the function is f(1, y) = 2 + 3y, which again describes a linear
line (Fig. 6.8b), and its derivative is

df 1, y
3
dy
which shows that along x = 1, the slope in the y-direction is 3. It can also be shown that
regardless of the value of x the slope at any point on the surface in the y-direction is
always 3.

15

Fig. 6.7. The surface of the function f(x, y) = 2x + 3y

Differentiating a multiple-argument function with respect to one argument while holding the
rest constant is the concept of partial differentiation. In the above example, the partial
derivative of f(x, y) with respect to x, holding y constant, is the slope in the x-direction.
And the partial derivative of f(x, y) with respect to y, holding x constant, is instead the
slope in the y-direction.
The partial derivatives of a function are expressed using the dell () notation rather than
d notation which is often reserved to mean total derivative. Consequently, the partial
derivatives of the function f(x, y) = 2x + 3y in the above example are

f x, y
f x, y
3
2 and
y
x

16

a)

b)

Fig. 6.8. The function f(x, y) = 2x + 3y when: a) y is fixed at 1, and b) x is fixed at 1. For
case: a) the slope is constant at 2, and b) the slope is constant at 3. Partial derivatives show
that the slopes at any point on the surface in the x- and y-direction are always 2 and 3,
respectively.
Example 6.2
Determine the slope at (1, -1) for the function

f x, y x 2 cos 2 x 3 y in both the x-

and y-directions. Note: all the (x, y) points are in unit radians.

Solution
Fig. .9 illustrates the surface for the function

f x, y x 2 cos 2 x 3 y , and Fig. .10a

and b show the function curve, fixing y = -1 and x = 1, respectively. Fixing y = -1, and
differentiating f(x, -1) with respect to x gives

df x, 1
2 x 2sin 2 x 3
dx
so the slope at x = 1 when y = -1 in the x-direction is

2 1 2sin 2 1 3 0.3171 .

Now fixing x = 1, and differentiating f(1, y) with respect to y gives

df 1, y
3sin 2 3 y
dy
so the slope at y = -1 when x = 1 in the y-direction is

3sin 2 3 1 2.5244 .

17

Fig. .9. The surface of the function

a)

f x, y x 2 cos 2 x 3 y

b)

Fig. .10. The function

f x, y x 2 cos 2 x 3 y when: a) y is fixed at -1, and b) x is

fixed at 1

18

A faster way to obtain the solution is to determine the partial derivatives directly for the
function

f x, y x 2 cos 2 x 3 y which are

f x, y
f x, y
3sin 2 x 3 y
2 x 2sin 2 x 3 y and
y
x
Consequently, the slope at point (1, -1) in the x-direction and y-direction are

2 1 2sin 2 1 3 1 0.3171 and 3sin 2 1 3 1 2.5244 , respectively.

6.3.2.3. Maximum and minimum

The maximum and minimum, collectively called extremum, are the largest and smallest values
of a function, respectively. We can imagine a function curve as a series of hills and valleys,
where the height of each peak and depth of each valley are known as the local maximum and
local minimum, respectively (Fig. 6.11). A function has a global maximum at xmax only if f(xmax)
f(x) for all x. Likewise, a function has a global minimum at xmin only if f(xmin) f(x) for all x.

Fig. 6.11. Local and global extrema

A function may have one or more global and local maxima, or even lack thereof. For instance,
the function, f(x) = x2, has only one global minimum at x = 0 (Fig. 6.12a), but functions, f(x) =
x and f(x) = x3, have no local or global maxima (Fig. 6.12b and c). The function, f(x) = x3 x,
has only one local extremum but no global extremum, where the local maximum and minimum
occur at

x 1

3 and x 1

3 , respectively (Fig. 6.12d). The trigonometric function,

f(x) = cos(x), has infinitely many global maxima, occurring at x = 0, 2, 4, , and infinitely

19

many global minima at x = , 3, 5, (Fig. 6.12e). However, the function, f(x) = 4cos(x) x, has infinitely many local extrema but no global extrema (Fig. 6.12f).
Calculus is required to determine exactly the extrema of any given function. We have seen
that a function curve may have one or more turning points. At a turning point, its slope must
be zero. Recall that the slope of a curve is determined by differentiating the curve function
f with respect to x. Suppose that (xm, ym) represents either the maximum or minimum point
on a slope. The slope (dy/dx) at the turning point (xm, ym) must be zero. Hence, we can
determine the value of xm when dy/dx = 0. We next substitute x = xm into the function to
determine the value of ym.
Consider the experiment by Pandey et al. (2000) who found that maize yield (kg ha-1)
corresponded to nitrogen fertilizer rate (kg ha-1) in the following quadratic manner:

Y 1143 31.7 N 0.084 N 2

[6.10]

We differentiate Eq. 6.10 with respect to nitrogen to obtain the slope of the curve at a
given N:

dY
31.7 0.168 N
dN
At maximum yield (turning point), the slope of the curve must be zero ( dY/dN = 0), so

dY
31.7 0.168 N 0
dN
N 188.7
which shows that maximum yield is achieved at nitrogen rate of 188.7 kg ha -1. To determine
the maximum yield, we substitute N = 188.7 kg ha-1 into Eq. 6.10 to obtain Y = 4133.7 kg ha-1.

20

a)

b)

c)

d)

e)

f)

Fig. 6.12. Extrema (or lack thereof) of several functions: a) f(x) = x2, b) f(x) = x, c) f(x) =
x3, d) f(x) = x3 x, e) f(x) = cos(x), and f) f(x) = 4cos(x) - x
It is useful if we can determine if a given turning point is a maximum or minimum point
without having to plot the function. We do this by performing a double differentiation,
d2y/dx2, on the function. Double differentiation measures the rate of change in the slope.
So whenever the double derivative is positive, the slope of the curve is increasing, and the
function curve concave upwards (Fig. 6.13a). This means that d2y/dx2 is positive at the
minimum point. In contrast, whenever the double derivative is negative the slope of the

21

curve is decreasing so the function curve concaves downwards (Fig. 6.13b). This gives a
negative d2y/dx2 at the maximum point. Note that the d2y/dx2 of a linear function is always
zero because the slope at all points along the line is a constant ( i.e., no change in slope).
a)

b)

Fig. 6.13. a) Increasing slope (curve concaves upwards, or a convex curve), and b)
decreasing slope (curve concaves downwards, or a concave curve)
Example 6.3.
Determine the extrema for the function y = 2x3 3x2 12x + 1.
Solution
Let us determine the single derivative of the function first:

dy
6 x 2 6 x 12
dx
At a turning point, its slope must be zero. Thus,

dy
6 x 2 6 x 12 0
dx
Using Eq. 6.2 to determine the roots means that x is -1 or 2. Substituting x = -1 and x = 2
into the original function gives:
for x = -1:

y 2 1 3 1 12 1 1 8

for x = 2:

y 2 2 3 2 12 2 1 19

which means that (-1, 8) and (2, -19) are the turning points of the function. To determine
which points are the minimum and maximum, we take the double derivative of the original
function:

d2y
12 x 6
dx 2
and substituting x = -1 and x = 2 into the double derivative gives
for x = -1:

d2y
12 1 6 18 (negative)
dx 2
22

for x = 2:

d2y
12 2 6 18 (positive)
dx 2

which means that (-1, 8) and (2, -19) are the maximum and minimum points, respectively.

Fig. 6.14. Diurnal solar irradiance during stable weather and clear skies

23