Notes

M ATHEMATICAL M ETHODS AND M ODELLING
S EMESTER 2
APPM2021
Course Notes
by
Authors
2023
Contents
Contents i
1 Introduction 1
1.1 Assumed background for the course . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Review of the Modelling Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Some popular real-life systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.1 Physical systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.2 Biological Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Systems of differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4.1 Categories of ODE Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4.2 Solutions of ODE Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Basic terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Qualitative Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 First Order Systems 9

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Phase diagrams of 1D Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Stability of critical points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Stability condition for ID systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Solutions of ID systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Modelling Applications - 1D Systems 13

3.1 The Logistic Population Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Solutions of 1D systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2.1 Solution of the Logistic Model . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Plane (2D) Autonomous Systems 17

4.1 Equilibrium solutions (critical points) . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Analytical Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
i
4.2.1 Direction field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2.2 Simple Isoclines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.3 Separatrices (straight line trajectories) . . . . . . . . . . . . . . . . . . . 20
4.3 Phase Diagrams of 2D Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.4 Characterising Critical Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.5 Finding the shape of trajectories analytically . . . . . . . . . . . . . . . . . . . . 23
4.6 Obtaining solution curves from the phase diagram . . . . . . . . . . . . . . . . . 24
4.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5 Solutions of linear Systems 28

5.1 Method 1: Elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.2 Method 2: Systems Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6 Linear Stability 34
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.2 Types of equilibria (Linear systems) . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.3 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.4 Eigenvalues and linear stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
7 Nonlinear stability 43
7.1 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.1.1 Taylor’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.1.2 Linearization Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.2 Local stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.3 Limit Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
8 Modelling Applications - 2D Systems 51

8.1 The linear Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.1.1 Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.1.2 Qualitative analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
8.2 Nonlinear Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
8.2.1 Interacting Populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
8.2.2 The Nonlinear Oscillator (Simple pendulum) . . . . . . . . . . . . . . . 54
8.2.3 Electrical circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
8.2.4 A general epidemic model . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
ii
9 Discrete systems 60
9.1 Why Discrete Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
9.2 From Difference Equations to Discrete Systems . . . . . . . . . . . . . . . . . . . 60
9.3 Discrete First Order Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
9.3.1 Equilibrium Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.3.2 General Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.3.3 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
9.4 Discrete Second Order Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
9.4.1 Linear Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
9.4.2 Nonlinear Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 69
iii
Chapter 1
Introduction
Mathematical modelling is a topic that looks into the use of various models, presented in
the form of techniques to analyse problems that arise in real life. Such techniques include
differential equations, numerical techniques, optimisation techniques. etc. No matter what
technique is used the process of employing it is the same.
1.1 Assumed background for the course

Familiarity with and knowledge of the following concepts shall be assumed in the
presentation of this course.
1. Basic algebra
• Solution of equations - linear and nonlinear.
• Simultaneous equations - linear and nonlinear.
• Inequalities - linear and nonlinear.
2. Functions and their graphs - incl exponential, hyperbolic
3. The derivative and curve sketching
• Rate of change.
• Increasing and decreasing functions.
• Concavity.
4. Linear algebra
• Matrix algebra.
• Eigenvalues and eigenvectors.
1
5. Matlab Use
Reasonable use of Matlab will be assumed from your experience with the two topics done
in the first semester. In particular, you should know how to use Matlab to solve ODE’s
numerically. A brief review of the process is given in another section later.
1.2 Review of the Modelling Process

Given below is an overview of the general modelling process that applies to a wide range of
modelling problems.
Step 1: Problem definition
• Set objectives.
• Define variables.
• Specify assumptions, simple first.
Step 2: Select the modelling approach
• Technique to be used; e.g. differential equations, stochastic model, etc.
• Analytical or computational
Step 3: Formulate the mode]
• Define the functional relationship between variables and translate them to

mathematical expressions/equations.
• Check if the model agrees with assumptions and certain conditions that should be
satisfied.
Step 4: Solve the model
• Graphical sketches /solutions.
• Qualitative solution.
• Analytic solution.
Step 5: Analysis (and validation) of results
• Interpret the results.
• Re-examine model formulation.
2
• Does the model answer all the questions necessary to understand the
system/problem?
In this module, we shall focus on problems that can be modelled using differential or
difference equations. Because the approach is already selected for the particular problems,
we shall employ Steps 3 to 5 most of the time.
1.3 Some popular real-life systems

Many situations in real life are characterised by forces of change that act on one another
with time. Because of such a change in the systems with time, they are commonly known
as dynamic systems. Such systems can be modelled by a set of differential equations (in
continuous time) or difference equations (in discrete time) that describe the interrelations
ship between the forces. Described below are examples of some systems that have been
studied, which we shall also discuss in more detail later in the course.
1.3.1 Physical systems

Example 1.1 Spring-mass system)
A body mass system suspended on a spring that moves only in one direction in a
friction-free media is an example of the physical system known as a linear oscillator(See
figure 1.1). The process of understanding the behaviour of such a system involves
identifying the physical laws that govern the system as well as the parameters involved that
characterise the components of the system, such as the mass of the body, and the tension
of the spring.
Out of these, a model can then be formulated that represents the physical laws in
mathematical expressions. Once the model has been formulated the techniques to be
learned in this module can then be used to understand how the system is likely to behave
with time. Taking this further may lead to obtaining an analytic solution for the system, a
function that describes its behaviour.
3
Figure 1.1: Spring Mass
Example 1.2 The pendulum

A pendulum of mass m, attached to a string/rod of fixed length L is a typical example of a
non-linear oscillator(See figure 1.2). In this case, the forces are acting in a two-dimensional
space. Hence the force F and the displacement x are vectors. The behaviour of the system
is seen to depend on parameters such as the gravitational force, the mass of the body and
the length of the string. This is an example of a model that leads to a system of non-linear
equations. Our study will include an analysis of such systems.
Figure 1.2: Pendulum
4
1.3.2 Biological Systems
Example 1.3 Single species model
In an introductory module to mathematical modelling, you may have come across a solution
to the model
dN
= r N, (1.1)
dt
describing the growth of a population of a single species. At that point, you may have been
interested in finding a solution to the system without really understanding the meaning of
the solution as it relates to describing how population growth is expected to develop with
time. The conditions under which it will grow or decline. In this module, we shall revisit
such models and derive more information about the system’s behaviour.
Example 1.4 Interacting populations

In real life, population systems do not exist in isolation as a homogeneous species. Their
mere existence presents an interaction with the environment in which they live. In another
dimension, they coexist with other populations that share the same environment. Thus
some understanding of the interactions that take place in an ecosystem is naturally
desirable, as well as a critical analysis of the impact that certain conditions presented by
the ecosystem will have on their survival.
Of particular interest would be the modelling of the interaction of populations of two

species. There are four types of interactions between species:
1. Symbiosis ( or mutualism): where the existence of one population enhances the of

another.
2. Competition: where the existence of both populations negatively affects each other, as
when they compete for the same resources.
3. Predator-prey: where the existence of one population (prey) enhances the other
(predator) while threatening the very existence of the former.
Example 1.5 Infectious deceases

The application of mathematical models to understand the spread of infectious diseases
has gained importance in the past decade or so. The complexity of understanding the
behaviour of the diseases within their hosts ( at a micro level) as well] as externally in their
socio-economic environment has brought about the recognition of the use of mathematics
to simulate the processes that go on based on known data and facts. Such understanding
could inform major policy decisions on such issues as drug inventions and other aspects of
national importance. taken in response to epidemics that arise from time to time.
5
Discrete Systems
Example 1.6 (Discrete Model) Although many models are based on the assumption that
events occur over continuous time, there are many situations in which observations are
taken in discrete time. Population census is a typical example of this case. This makes it
necessary to use discrete models which can be analysed in ways similar to those used for
continuous systems. In some cases, the behaviour of such discrete models may be quite
different from their continuous counterparts, for the same system.
The above list of examples is not exhaustive, but illustrative of the many dynamical systems
that can be modelled with systems of differential equations. Although solutions of the
systems are the most desirable thing to obtain, we shall focus our study of such systems on
analyzing them in an effort to get some understanding of their behaviour that can put
some light on their future states. Such an analysis is often the only thing possible to do
since solutions may not be obtainable. We shall use specific models related to the above in
exercises in the future.
1.4 Systems of differential Equations

The most general system representation in continuous time is
d x1
= f 1 (t , x 1 , x 2 , ....., x n )
dt
d x2
= f 2 (t , x 1 , x 2 , ....., x n )
dt
.
d xn
= f n (t , x 1 , x 2 , ....., x n ), (1.2)
dt
where f i , i = 1, 2, ..., n are continuously differentiable functions. Another way to write the
above equation is
d x(t )
= f(x) or x0 = f. (1.3)
dt
Such a system may also arise as a transformation of an n-th order differential equation. At
any time t , x = (x 1 (t ), x 2 (t ), ...x n (t )) describes the state of the system.
6
1.4.1 Categories of ODE Systems
There are three broad categories of ODE systems. If the function f has no explicit
term/expression in the variable t then the system is said to be autonomous. Thus an
autonomous system has the general form
d x1
= f 1 (x 1 , x 2 , ....., x n )
dt
d x2
= f 2 (x 1 , x 2 , ....., x n )
dt
.
d xn
= f n (x 1 , x 2 , ....., x n ), (1.4)
dt
In such a case the subsequent state of the system is dependent only on the current state
and not on time. The functions f i may be linear or non-linear in the variables x i (t ). If there
exists at least one nonlinear term, then the system is said to be nonlinear. Some systems do
not have any constant terms appearing on the right-hand side. Such systems are said to be
homogeneous, otherwise, they are said to be non-homogeneous. In this module, we shall
be discussing autonomous systems, both linear and nonlinear.
1.4.2 Solutions of ODE Systems

Analytic solutions of some simple systems may be obtained by standard methods, but for
many systems, it is often not easy and sometimes impossible to solve using analytical
methods. Although explicit solutions of many systems, especially nonlinear systems,
cannot be found, a lot can be understood about the expected behaviour of such systems
with time, without actually finding the solutions explicitly. In this module, we shall be
interested in understanding the behaviour of the system in the vicinity of their critical
points or equilibrium points.
1.5 Basic terminology

Here we present some of the concepts that will recur right through our discussions.
Using the analogy of a particle in space, the state of a system describes the position of the
particle at any time t is (x 1 (t ), x 2 (t ), ..., x n (t )). As t changes the position changes also with
respect to t in the phase space. As this happens the position traces a path known as the
7
phase path or phase path. Corresponding to each initial state or position, there is a unique
trajectory. A collection of these phase paths generated by choosing different initial states
gives a phase portrait of the systems. No trajectories cross over each other in the phase
space.
An equilibrium solution is that of the system in which the system remains for any
subsequent time. At this point, the system is said to be in equilibrium. In general, this is the
solution of the system
dx
= f(x) = 0. (1.5)
dt
At a point x, displaced from the critical point, either x0 =f(x) > 0 or f(x) < 0. This in turn
means that the rate of change is either increasing or decreasing. Determination of the
equilibrium solution, x e , is normally the starting point when analysing a system. The
equilibrium solution is also commonly called a fixed point, stationery point, critical
point, or steady-state solution. All these terms will be used interchangeably.
At any point besides the critical point, the system state (the positions of the particle) is
changing in a specific direction. The net direction of change is given by x0 , a gradient (or
direction) vector whose components are x0i , i = 1, 2, ..., n. As t changes, this direction also
changes. A collection of direction vectors is called a direction field. Note that the direction
of change at a point is the direction of the tangent to the curve at that point.
If it is possible to solve the system for x i (t ), i = 1, 2, ..., n, these are called the time series
solutions of the system.
1.6 Qualitative Analysis

Quantitative analysis of a system would normally involve obtaining as much information as
possible about the behaviour of the solution of the given system without actually finding the
solution. This would involve the following steps:
• Identifying equilibrium solutions/ critical points.
• Analysing the behaviourof the system in the neighbourhood of the critical point(s).
• Sketch a phase diagram.
• Determining the stability of the critical points by analysing the phase diagrams and
using other algebraic techniques available.
8
Chapter 2
First Order Systems
2.1 Introduction
Reading : Zill and Cullen Chapter 2 (p39-49) (NB: Zill uses the notation y = y(x) instead of
x = x(t ).)
NB: Zill uses the notation y = y(x) instead of x = x(t ).
We begin with the study of first-order autonomous systems of the form
dx
x0 = = f (x) (2.1)
dt
dx
Only one solution x(t ) is sought for this system. If we recall from basic calculus, dt
dx
describes the rate of change in x with respect to t . At any point t, dt
can be evaluated. Thus
any value of t corresponds to a (phase) point (x(t ), x 0 ) = (x, f (x)). As t changes, the trace of
the points generates a phase path.
First, we are interested in determining the equilibrium solution x e , of the system.

According to (1.5) this is the solution of
x 0 = f (x) = 0. (2.2)
At a point x, displaced from the critical point, either x 0 = f (x) > 0 or x 0 = f (x) < 0.
9
Activity 2.1
Suppose x is a function of t . Illustrate by means of sketches the following cases:
1. When x is increasing at an increasing rate;
2. When x is increasing at a decreasing rate;
3. When x is decreasing at an increasing rate;
4. When x is decreasing at a decreasing rate
2.2 Phase diagrams of 1D Systems
We will associate with x 0 > 0 a movement from left to right, and x 0 < 0 a movement from
right to left along the plot of x against x 0 . The phase plane of a 1D system is that of x against
f (x). As such the phase plot can be drawn without using a direct dependence on t. Thus
the dependence on t can be eliminated and the resulting equation is called the phase plane
differential equation.
Example 2.1 (Linear systems)

Consider the system x 0 = ax. This is a simple system for which a solution can be obtained
explicitly. The equilibrium solution is the solution of ax = 0, which in this case is x = x e = 0.
For a phase diagram, we plot x against x 0 or f (x). The graph of f (x) is a straight line
through the origin, the critical point (0, 0) in this case. However, the orientation of the line
depends on whether a > 0 or a < 0
Case 1: a > 0
ax > 0 if x > 0 and ax < 0 if x < 0. So as x gets positively large, x 0 > 0, and hence a positive
move (left to right). Similarly, as we move away from 0 in the negative direction, x 0 < 0,
hence the phase point moves from right to left.
Case 2: a < 0
Activity 2.2
1. Draw a phase path similar to the one in Case 1.
2. Study the movements for x > 0 and x < 0.
3. Mark the directions on the diagram.
Later on, we will focus on the meaning of the direction of the phase path relative to the
critical point.
10
Activity 2.3
For each of the systems
(i ) x 0 = ax 3 ; (i i ) x 0 = x 2 − 1; (i i i ) x 0 = ax 2
1. Determine the equilibrium point(s).
2. Plot the phase diagrams associated. with the cases a > 0 and a < 0.
3. Specify the type of equilibria in each case.
2.3 Stability of critical points

A more formal discussion of the classification or characterisation of equilibrium points will
be presented later when we discuss the stability of systems. At this stage we will only classify
them in these simple terms:
• Attractor : if arrows directed is towards the critical point on both sides. (see Example
2.1, (Case 2) and Activity 2.3 (i i ) (x e = −1).
• Repeller: if arrows are directed away from the critical point on both sides.(see
Example 2.1, (Case 1) and Activity 2.3 (i i ) (x e = 1).
• Shunt: if arrows do not change direction on either side of the critical point; i.e.
towards it on one side and away from it on the other. (see Activity 2.3 (i i ) ).
In more formal terms, an attractor is said to be stable and a repeller is unstable.
2.4 Stability condition for ID systems

The following condition can be used to check the stability of the critical point x e of the
continuous system
dy
= f (x) : (2.3)
dt
• If f 0 (x e ) < 0, then x e is asymptotically stable. That is, for an initial point (t 0 , x 0 ), the
point will converge to the critical point as t → ∞.
• If f 0 (x e ) = 0, then analyse the problem further to determine the stability of x e .
• If f 0 (x e ) > 0, then x e is unstable.
11
2.5 Solutions of ID systems
Solutions of 1D systems can be obtained by the standard analytic methods of solving first
order differential equations. A plot of x as a function of t gives the graph of the required
solution, also called the time series solution.
dx 1
Example 2.2 The time series solution of dt
= x 2 is x = −t +c
. The particular solution
corresponds to a particular condition, x(t 0 ) = x 0
Activity 2.4
Consider again the phase portrait of the system x 0 = x 2 − 1 in Activity 2.3.
1. Use the phase portrait of the system to sketch the solution plot of the system
for x(0) > 1, 0 < x(0) < 1, x(0) < −1.
2. Now solve the differential equation x 0 = x 2 −1, for the cases x(0) = 2, x(0) = 0.5,
x(0) = −2. Plot these solutions on the same axes, for t > 0.
3. Compare the above plots.
(Try to use a computer to help you with this exercise.)
2.6 Exercises
1. For each autonomous system given below
dx
i) dt = x 2 − 3x
dx
ii) dt
= (x − 2)4
dx
iii) dt
= x ln (x + 2)
a) Find all critical points.

b) Sketch the phase portrait for the system.
c) Classify the critical points.
2. Use the phase portraits in 1. above to sketch the time series solution of the systems.
3. Solve the equations in 1(c)i and 1(c)ii above analytically. Plot them and compare them
with your sketches in 2.
4. Use MATLAB or other sofware to draw the phase portraits and solutions for the systems
in 1.
12
Chapter 3
Modelling Applications - 1D Systems
Reading: Zill and Cullen, Chapter 3
We now want to apply our knowledge of solving and analysing systems to real life
situations. Systems modelled by 1D systems occur in various disciplines. Here we start
with an application to population dynamics.
3.1 The Logistic Population Model

Recall the model
dN
= r N, N ≥ 0, (3.1)
dt
for a population growing at a constant intrinsic rate r, For this model no constraints are
imposed by external factors such as environment, limited resources, etc.
Activity 3.1
1. Find the equilibrium point(t) of the above system.
2. Draw a phase portrait for the system.
3. What is/are the type(s) of critical point (s).
Density-dependent Logistic Model
In reality no population will grow exponentially because as the population gets larger, there
is likely to be crowding, which in turn may result in greater competition for the limited
resources. The effect of this is that the growth rate r will now be density-dependent, and
will be more likely to diminish with time.
13
If the growth rate r is dependent on population size (density) (i.e. r = r (N )), then equation
(3.1) becomes
dN
= N r (N ). (3.2)
dt
First a few notes about r (N )
• Assume N is large enough to be modelled as a function of t
• As N increases r (N ) decreases
• For large N , r (N ) has been shown to be negative r The simplest form is a straight line:
r (N ) = a − bN where a > 0 is the growth rate without environmental influences, and
b > 0 is the effect of increased population density.
NB: When b = 0, then the density-independent model (3.1) hods. Thus the
density-dependent model (3.2) becomes
dN
= N (a − bN ) (3.3)
dt
Note that this is now a nonlinear system and the equilibrium population is attained when
the rate of change of the population is zero.
Activity 3.2
Draw a phase portrait, of the model as given in 3.3 for a, b > 0, showing clearly
the direction arrows on the path.
Density-dependence is a term associated with the concept of carrying capacity. If it is

supposed that the environment in which the particular population grows is capable of
sustaining no more than K individuals in its population, then K is said to be the carrying
capacity.
So we can let a = r, b = r /K , and, write the logistic equation in the alternative form
dN r
= N (r − N ) (3.4)
dt K
Note here that f (K ) = 0. i.e. when N = k, N 0 = 0.
3.2 Solutions of 1D systems

Solutions of 1D systems can be obtained by solving the given differential equation for x(t )
in terms of t using the standard techniques. For example (3.1) can be solved analytically to
obtain N (t ) = N0 e r t . Here use the logistic equation (3.3) as an example.
14
3.2.1 Solution of the Logistic Model
Activity 3.3
Show that the solution of the logistic equation is
aN0
N (t ) = (3.5)
bN0 + (a − b)e −at
A solution curve for 0 < N0 < K is shown on the diagram. Problem 1 in the next exercise
gives an opportunity to sketch the (time series) solutions of the logistic equation for different
values of N0 . Note that no matter how close N0 is chosen to 0, the solution will grow fast
at the early stages (away from N = ba , the equilibrium level. As t → ∞, e −at → 0. Hence
aN0
N (t ) → bN0 = ba .
3.3 Exercises
1. Computer Exercise Consider again the logistic model (3.4) with K = 100.
a) Find the solution of the density dependent model when r = 2, K = 100.

b) Compute Ṅ for N ∈ [N0 , 200] and N0 = 5, 20, 50, 150, 200.
c) Plot N against Ṅ .
d) Plot t against N .
e) Investigate how the plot of t against N could be sketched from the phase portrait.
2. The autonomous differential equation
dv
m = mg − kv
dt
where k is a positive constant of proportionality and g is the acceleration due to
gravity (a constant), governs the velocity v of a body of mass m falling under the
influence of gravity.
a) Interprete the term −kv.

b) Because of the term −kv the velocity of the body falling from a great height,does
not. increase without bound as time t increases. Use a phase portrait of the
differential equation to find the limiting, or terminal, velocity of the body.
Explain your reasoning. (Ans. mg /k )
c) Find the terminal velocity of the body if air resistance is proportional to v 2 . (Ans.
p
mg /k)
15
3. The model for the spread of a virus in an isolated university campus of 1000 students
is given by
dx
= kx(1000 − x)
dt
where k is a constant and x in the number of infected students. Suppose a student
carrying a flu virus returns to campus.
a) What meaning would you give to the constant k ?
b) If it is assumed that the rate at which the virus spreads is proportional not only
to the number x of infected students but also to the number of students not
infected, determine the number of infected students after 6 days if it is further
observed that after 4 days, 50 students are infected. (Ans: 276 students)
4. The number N (t ) of supermarkets throughout the country that use a computerised

checkout system is described by the initial-value problem
dN
= N (1 − 0.0005N ), N (0) = 1
dt
a) How many supermarkets are expected to adopt the new procedure over a long
period of time. Illustrate this on a phase portrait.
b) Sketch a solution curve of the given initial-value problem.
c) Solve the initial-value problem (you may use MATLAB) and graph its solution.
How many supermarkets are expected to adopt the new technology when t = 10?
16
Chapter 4
Plane (2D) Autonomous Systems
Reading: Zill and Cullen, Chapter 8, 10
Many physical systems that change with time are modelled by differential/difference
equations of order higher than one. A transformation of these equations lead to systems
consisting of a set of first-order differential equations (see equation(1.4)) to be solved
simultaneously for the dependent variables.
The general form of a 2D autonomous system is

dx dy
= f (x, y), = g (x, y). (4.1)
dt dt
Note that there is no explicit dependence on t .
The state of the system at any time t is defined by the point (x(t ), y(t )). Note that although
the variables x and y are functions of t , for autonomous systems there is no term that
involves t explicitly. Hence the subsequent state of the system depends on the current state
and not on t .
If for some constants a, b, c, d , k1 and k2, the functions f and g have the form
f (x, y) = ax + b y + k 1 , g (x, y) = c x + d y + k 2 , (4.2)
then the system is said to be linear. Otherwise, it is nonlinear. If in the expressions of f and
g , the constants, k1 and k2 are both zero, the system is said to be homogeneous, otherwise,
it is non-homogeneous.
A solution of the above system is a pair of functions x(t ) and y(t ) such that (4.1) is true.
The graph of such a solution in x y-space is a trajectory(or path) of the points (x(t ), y(t ))
17
generated by choosing different values of t . If we are given x(0) and y(0), we might expect
that x(t ) and y(t ) would be uniquely determined for all t . In our study, we will assume that
f , g ∈ C 1 functions. It is useful to think of x and y as the coordinates in an x y -plane. A
solution (x(t ), y(t )) then describes a parametric curve in this plane. For each value of the
parameter, t a point of the plane is given. Another way to think of these trajectories is that
they are contours of the surface defined by the solution of the 2D system. If t is time, then
(x(t ), y(t )) is a moving point that follows the trajectory.
If the moving point starts from the same point x 0 , y 0 but at a different time, it will follow the
same trajectory. There is exactly one trajectory through each point x 0 , y 0 in the region
where f and g are C 1 . Different trajectories do not intersect.
We now describe some features that arise in the analysis of systems: equilibrium solutions,
simple isoclines, and separatrices.
4.1 Equilibrium solutions (critical points)

The analysis of system (4.1) begins with the identification of the equilibrium solution(s),
(x e , y e ). The equilibrium solution, (x e , y e ) is the solution of the system
dx
= f (x, y) = 0, (4.3)
dt
dy
= g (x, y) = 0, (4.4)
dt
which can be done using standard algebraic techniques. Other terms used for the
equilibrium solution are critical point, stationary point, steady-state solution.
Several possibilities can arise in the search for critical points of 2D systems. The
equilibrium the equation may have no solution, one solution or multiple solutions.
For linear, homogeneous systems, the equilibrium equations have the form
ax + b y = 0
(4.5)
cx + d y = 0
or
" #" # " #
a b x 0
= (4.6)
c d y 0
18
which can be interpreted as the intersection of the two lines f (x, y) = 0 and g (x, y) = 0.
1. Exactly one critical point (at the intersection of the two lines), still exist if the
determinant ad − bc 6= 0.
2. Parallel lines, in which case there are no critical points.
3. Two lines coincide, which means the whole line consists of critical points.
4. a = b = 0 or c = d = 0 or both, which means the lines are degenerate.
Activity 4.1
For each system given below determine the critical point(s), if they exist.
dx dy
(a) dt
= −y; dt
=x
(b) x = 3x + 2y; y 0 = 2x − 2y
0
(c) x 0 = −x − y + 8; y 0 = 2x − y − 4
(d ) x 0 = y; y 0 = x 2
(e) x 0 = y; y 0 = 2x − x 2
Once the equilibrium solution(s) (x e , y e ) have been determined the process of analysing
the behaviour of the system for any displaced position (x, y) around it can be done. Let’s
discuss more concepts in the 2D context that will assist in doing this.
4.2 Analytical Concepts
4.2.1 Direction field

At any point (x, y) we can determine the direction of movement in the x- and y-directions
dx dy dx
using the signs of dt
and dt
, respectively. If dt
> 0, then the x-direction is positive (left to
dx
right) and if dt
< 0, then the x-direction is negative (right to left). Similarly, the y-direction
dy dy
is positive (up) if dt
> 0 and negative (down) if dt
< 0.
Recall that geometrically, the direction is a directed vector. Thus the x-direction is
horizontal and the y-direction is vertical. The direction of (x, y) is thus the resultant vector
dx dy
from a combination of dt
and dt
.
The net direction of x and y at the point (x, y) is the gradient vector of the solution curve at
that point, given by the sum of the x- and y- direction vectors. The possible combinations
of x 0 and y 0 are as shown on the diagram in Figure 4.1
19
Figure 4.1: Direction field
As the position of (x, y) changes with t , different gradient vectors can be determined at
each position, and a collection of these forms the direction field. Note that the overall
magnitude and direction are determined by f (x, y) and g (x, y) on the right-hand side of
(4.1). Hence it is sufficient to determine the direction from the signs of f (x, y) and g (x, y),
which are in turn determined by the signs of x and y. The phase plane in the 2D case
consists of the x- and y-axes.
Here is another way to determine the resultant direction/gradient vector. If we assume that
dx
dt
6= 0 so that t is an inverse function of x and y(t ) is a composite function of x, then we
have
dy
dt dy dt dy
dx
= = . (4.7)
dt dx dx
dt
Thus the direction/gradient of a phase path at the point (x, y) can also be obtained from
d y g (x, y)
= . (4.8)
d x f (x, y)
A similar relationship can be obtained by supposing x is a function of y. Thus if we are at a

point where f (x, y) and g (x, y) are both nonzero, then it doesn’t matter whether we consider
y as a function of x or vice versa. Thus (4.8) is equivalent to
d x f (x, y)
= . (4.9)
d y g (x, y)
4.2.2 Simple Isoclines

These are lines (or curves) where the gradient of the phase path (the direction vectors) is
dx
either vertical or horizontal, signifying no change in x or y, respectively. If dt = f (x, y) = 0
dy
(or dx
= ∞), then there is no change in x and the direction vectors will be vertical everywhere
dy
on that line (or curve). Similarly, the direction vectors will be horizontal where dt
= g (x, y) =
dy
0 (or dx
= 0).
4.2.3 Separatrices (straight line trajectories)

Special type of trajectories, called separatrices (or straight-line trajectories) exist for some
systems. On these lines the direction vectors coincide with the trajectories; i.e. the
direction vectors are exactly aligned with the trajectories. The trajectories do not cross the
20
separatrices, hence they are also called asymptotes.
The determination of the separatrices is well-defined for some linear homogeneous

systems. The rule may not apply to other cases. For homogeneous systems, the critical
point is at the origin. As such we can assume that these straight-line trajectories have the
dy
form y = mx, and hence a slope dx
= m. Thus their slope can be determined by solving for
m in
d y g (x, y)
= = m, f (x, y) 6= 0. (4.10)
d x f (x, y)
If m is real, then the given system has separatrices. Otherwise, it does not. For the
g (x,y
homogeneous system (4.2) the solution of f (x,y)
= m, using y = mx, yields (Verify this!)
p
(d − a) ± (a − d )2 + 4bc
m= , b 6= 0. (4.11)
2b
Note that b = 0 corresponds to a vertical isocline.
4.3 Phase Diagrams of 2D Systems

The behaviour or dynamics of the linear (homogeneous) systems (4.1) can be represented
geometrically in the form of a phase diagram. This is a portrait of the system in the x y−
plane. Drawing a phase diagram involves the following steps.
General steps
1. Identify equilibrium (or critical) points.
2. Determine the Direction field:
a) Find the simple isoclines where x 0 = 0 and y 0 = 0

dy
In the form of (4.10), this is equivalent to dx
= 0.
b) Determine the direction field on all sides of the simple isoclines; i.e Find where
x 0 > 0, x 0 < 0, y 0 > 0, y 0 < 0.
dy dy
In the form of (4.10), this determining where dx
> 0, dx
< 0.
3. Sketch the curves by sketching several paths that follow the net directions, starting at
different points. (The initial points represent the initial conditions of the system.)
For the linear system (4.2), the simple isoclines and the direction field are easily
determined as:
21
x-direction y-direction
ẋ = 0 when y = − ba x ẏ = 0 when y = − dc x
ẋ > 0 when y > − ba x ẏ > 0 when y > − dc x
ẋ < 0 when y < − ba x ẏ < 0 when y < − dc x
Example 4.1
Consider the system:
dx dy
= x − y; = 5x − y, (4.12)
dt dt
whose critical point is (0, 0). The simple isoclines, orientation and directions are
summarised in the following table:
Orient./dir. Line/region
ẋ = 0 y = x (| on this line)
ẋ > 0 y < x(→ below line)
ẋ < 0 y >x ( above line)
ẏ = 0 y = 5x(- on this line)
ẏ > 0 y < 5x ( ↑ below line)
ẏ < 0 y > 5x ( ↓ above line)
Activity 4.2
Draw the phase portrait for the systems
dx dy
(a) dt
= −y; dt
=x
dx dy
(b) dt
= x + y; dt
= −x
0 0
(c) x = −x − y + 8; y = 2x − y − 4
(d ) x 0 = x 2 − 2x; y0 = y
Identifying all critical points, isoclines and asymptotes, as well as showing the
directions, Also state the nature of the critical points (i.e. attractor,... etc).
Exercise 4.1
Draw the phase diagram of the system
dx dy
= x(1 + y); =x−y (4.13)
dt dt
4.4 Characterising Critical Points

Any critical point can be characterised by the shape of the trajectory pattern as well as the
direction field around it. There are common patterns that have been given names. These are
nodes, saddle points, spiral focus, and centre. Depending on arrows of the direction fields,
22
the critical points can be further characterised as stable or unstable. For now, based on
the phase portraits, we can characterise a critical point either as asymptotically stable if the
arrows flow towards the critical point (an attractor), or unstable if they flow outwards from
the critical point (a repeller). Apart from these two scenarios, the direction field may have
neither of them or a combination of them. The various types of critical points are shown on
a separate page in Figure 4.2. Later on, we will discuss how to characterise critical points
based on algebraic conditions when we discuss stability in more detail.
4.5 Finding the shape of trajectories analytically

So far we have used sketches of the solution curves based on the direction field. It is
sometimes useful to be able to determine the shape of these curves in order to understand
the phase plane over a wider region. We are going to discuss this in this section.
We have seen that the phase portrait of the system
x 0 = ax + b y, y0 = cx + d y (4.14)
can be obtained by plotting the direction field
dy f (x, y)
= , (4.15)
d x g (x, y)
which describes the relationship between x and y, This is a first-order differential equation
which in many cases, can be solved using analytical methods. The solution of this equation
has the form h(x, y) = 0. Different values of the parameters will give rise to different curves
for the above relationship. These are the trajectories of the system, hence giving the phase
portrait thereof.
It is important to note that plotting the trajectories from the solution of the equations may
not be easy because of their complexities. However, the phase portrait approach will
normally always work.
We consider again the previous examples and solve the phase equation to find the
equations for the trajectories.
Example 4.2
dy
For the system x 0 = y, y 0 = x we have dx
= xy . This is a separable equation and so we can
write it as yd y = xd x and solve it to find
y 2 x2
= + c 1 or y 2 − x 2 = c. (4.16)
2 2
23
This is the equation of a hyperbola centred at (0, 0). By varying c we obtain the family of
hyperbolae on the x y-plane.
Activity 4.3
Find the equation of the solution curves of the system in Example 4.1
Although it is possible to find solutions of the first-order phase differential equations’ the
plots of their solutions may not be so trivial. Hence phase diagrams are better sketched
from the techniques discussed earlier.
4.6 Obtaining solution curves from the phase diagram

For the 1D systems discussed earlier we were able to obtain a graph/sketch of t against the
solution, also known as the time series plot. Time series solutions of 2D systems consist of
two plots one for t against x and the other for t against y. The former yields the solution
curves of x for different initial points by transcribing the horizontal distance of a phase
point from the equilibrium x -coordinate, while the latter transcribes the vertical distance
of the same point from the equilibrium y-coordinate.
Example 4.3: Transcribing a phase diagram

Consider a system whose phase diagram is as shown in Figure 4.3. Sketch the time series
solutions of x(t ) and y(t ) ) from the phase diagram. (see Figure 4.4).
Exercise 4.2
Sketch the solution curves corresponding to a (i ) repelling spiral, (i i ) a saddle point.
24
4.7 Exercises
1. For each of the systems given below:
a) Find the equilibrium point(s).
b) Plot the phase diagram that clearly shows the trajectories and their directions.
(i) x 0 = x − 2y, y 0 = 3x − 4y
(ii) x 0 = 3x + 2y, y 0 = −2x − 2y
(iii) x 0 = −x − y + 8, y 0 = 2x − y − 4
(iv) x 0 = x − y, y 0 = x2 − 1
(v) x 0 = ye y , y 0 = 1 − x2
(vi) x 0 = −3x + 6x y, y 0 = 2y − x y
(vii) x 0 = x 2 − y 2 − 1, y 0 = 2y
2. Given below are the parameters a, b, c, and d associated with various linear systems. For
each system defined by this set of parameters,
a) Find the critical point(s) of the system and specify their type.
b) Sketch the phase diagram for the system.
(i) a = 0, b = 1, =¸ − 4, d = 0
(ii) a = 1, b = 3, c = 1, d = −1
(iii) a = 2, b = 1, c = 1, d = 2
(vi) a = −3, b = −2, c = 1, d = −5
(v) a = 1, b = 2, c = −1, d = −2
(vi) a = 3, b = −1, c = 1, d = 2
(vii) a = 3, b = −18, c = 2, d = −9
3. Discuss the nature and stability of the simple critical points of the system corresponding
to the following nonlinear equation
d 2x dx
2
+6 − x 2 + 4x = 0
dt dt
25
26
Figure 4.2: Phase diagrams

Figure 4.3: An attracting spiral
Figure 4.4: Time series solutions for the attracting spiral above
27
Chapter 5
Solutions of linear Systems
Reading from Zill Chapter 8, 10
Whenever possible it is desirable to find explicit solutions to systems. This is possible to do

for most linear systems and some nonlinear systems. We take a moment to look at explicit
solutions of linear systems, which will be handy when we discuss the stability of the critical
points in the next section.
It has been noted that the solution to the system (4.2) is a pair of functions (x(t ), y(t )). You
know by now that each initial value corresponds to a trajectory or integral curve generated
by the pair (x(t ), y(t )). The phase diagram depicts the evolution of the system state from
arbitrary initial states.
We now discuss two methods used to determine the solution of a given linear autonomous
system (4.2),
x 0 = ax + b y [1], y 0 = c x + d y [2]. (5.1)
We are seeking functions of the form
x(t ) = h 1 (t ), y(t ) = h 2 (t ), (5.2)
that satisfy equation [1] and [2] simultaneously.
5.1 Method 1: Elimination

The method of elimination is similar to the method of solving simultaneous equations by
elimination. We proceed as follows:
28
[2] ÷ c and solve for x :
1 dy d 1 dy d
= x + y, → x= − y, c 6= 0. (5.3)
c dt c c dt c
Differentiating the last equation with respect to t gives
dx 1 d2y d d y
= − . (5.4)
dt c dt2 c dt
Hence, [1] becomes
1 d2y d d y 1 dy d
µ ¶
− =a − y +by (5.5)
c dt2 c dt c dt c
or
d2y dy
2
− (a + d ) + (ad − bc)y = 0, (5.6)
dt dt
which is a second-order differential equation with constant coefficients. To solve it, the
characteristic equation is
λ2 − (a + d )λ + (ad − bc) = 0, (5.7)
whose roots are

p p
(a + d ) + (a + d )2 − 4(ad − bc) (a + d ) − (a + d )2 − 4(ad − bc)
λ1 = and λ2 = . (5.8)
2 2
Hence, the solutions
y(t ) = c 1 e λ1 t + c 2 e λ2 t , (5.9)
and
1 dy d
x(t ) = − y, c 6= 0
c dt c (5.10)
(λ1 − d ) λ1 t (λ2 − d ) λ2 t
= c1 e + c2 e
c c
Each of the above solutions can be plotted separately for a portrait of how each variable
changes with time. The phase plot is that of x(t ) against y(t ).
5.2 Method 2: Systems Method

This is basically a method based on eigenvalues and eigenvectors. We first write [1] and [2]
as
" # " #" #
d x a b x
= (5.11)
dt y c d y
or
du
= Au, (5.12)
dt
29
where we have let u 1 (t ) = x(t ) and u 2 (t ) = y(t ).
The advantage of using this format is that it is easily generalised to solve larger systems.
Solutions. of the above systems are sought in terms of exponentials as

" # " #
λt x x̂
u = û e , or = e λt , (5.13)
y ŷ
where û is a constant vector - the eigenvector corresponding to the eigenvalue λ, to be

determined.
Differentiating with respect to t yields
du
= ûλe λt . (5.14)
dt
Hence
du
= Au = A ûe λt = ûλe λt , (5.15)
dt
from which
" #" # " #
a b x̂ x̂
Au = ûλ, or =λ , (5.16)
c d ŷ ŷ
which we in turn write as

(A − λI )û = 0. (5.17)
Nontrivial solutions of the above system exist if and only if

¯ ¯
¯a − λ b ¯
|A − λI | = ¯ ¯=0 (5.18)
¯ ¯
¯ c d − λ¯
Expanding this determinant yields the quadratic equation
(a − λ)(d − λ) − bc = λ2 − (a + d )λ + ad − bc = 0,
the same characteristic equation obtained in the first method, whose two roots are given by
p
(a + d ) ± (a + d )2 − 4(ad − bc)
λ= .
2
In vector form the solution is
u = c 1 û1 e λ1 t + c 2 û2 e λ2 t , (5.19)
where û1 and û2 are the eigenvectors corresponding to λ1 and λ2 respectively. Also, c 1 and
c 2 are arbitrary constants (associated with the roots λi ).
30
Note that the two solutions given by (5.19) are explicitly
u 1 (t ) = x(t ) =c 1 û1,1 e λ1 t + c 2 û2,1 e λ2 t , (5.20)

λ1 t λ2 t
u 2 (t ) = y(t ) =c 1 û1,2 e + c 2 û2,2 e . (5.21)
Three types of general solutions associated with the various cases about the roots λ1 and λ2
arise:
Case 1: Real and distinct roots(λ1 6= λ2 )

The solution of the system is given by
u = c 1 û1 e λ1 t + c 2 û2 e λ2 t . (5.22)
Case 2: Repeated roots (λ1 = λ2 = λ)

Here the solution has the form
u = c 1 û1 e λt + c 2 û2 t e λt , (5.23)
when one eigenvector is always a scalar multiple of the other eigenvector.
Case 3: Complex conjugate roots (λ = α ± βi )

For this
u = c 1 û1 e (α+βi )t + c 2 û2 r e (α−βi )t . (5.24)
A more convenient form of the general solution, based on the Euler formula
e i ωt = cos (ωt ) + i sin (ωt ) , e −i ωt = cos (ωt ) − i sin (ωt )
is
u = c 1 û1 cos βt + i c 2 û2 sin βt .
¡ ¢ ¡ ¢
Actually, the fact that λ1 = λ2 leads to the fact that û1 and û2 are conjugate vectors, i.e. û2 =
û1 . If we let
1 i
v̂1 = (û1 + û2 ), v̂2 = (−û1 + û2 ),
2 2
then manipulation of the two solutions gives v̂1 and v̂2 as real vectors. This leads to a more
general solution,
u 1 = x = v̂1 cos βt − v̂2 sin βt e αt

£ ¡ ¢ ¡ ¢¤
(5.25)
u 2 = y = v̂2 cos βt + v̂1 sin βt e αt ,
£ ¡ ¢ ¡ ¢¤
(5.26)
where v̂i are defined above.
The vectors v̂1 and v̂2 are actually equivalent to
v̂1 = Re(û1 ) and v̂2 = I m(û1 )
Sketches of the various possible solutions are shown in figure 5.1
31
Figure 5.1: Solutions for different values of λ1 and λ2
32
5.3 Exercises
For each system below find the eigenvalues and their corresponding eigenvectors. Hence
deduce the solutions of the systems.
a) x 0 = 2x + 3y, y 0 = 2x + y
dx dy
b) = −x + 3y,
dt dt = −3x + 5y
Ã !
2 8
c) X0 = X
−1 −2
33
Chapter 6
Linear Stability
Reading from Zill: Chapter 10
6.1 Introduction
Two issues will be discussed here concerning the equilibrium solutions:
1. The type of field created by the trajectories around the critical points; i.e. the shape of
trajectories.
2. The stability of the critical points.
6.2 Types of equilibria (Linear systems)

So far we have used isoclines and direction fields to help us determine the type of
equilibrium point a system has. Sometimes it can be tricky to deduce from the sketch. A
more systematic way is based on characterising the system using eigenvalues derived from
the coefficients a, b, c, d of the autonomous linear system (review the concept of
eigenvalues and eigenvectors). Such characterisation summarises what happens to the
solutions 5.19 as t → ∞.
The eigenvalues are the same roots obtained in the previous section from solving the
characteristic equation.
λ2 − (a + d )λ + ad − bc = 0, (6.1)
as p
(a + d )2 − 4(ad − bc)
(a + d ) ±
λ1 , λ2 = .
2
In terms of the given matrix A, the expressions, a + d and ad − bc, are respectively the trace
and determinant of A. i.e. a + d = t r (A), and ad − bc = d et (A).
34
To simplify our notation let
T = a + d (the trace)
D = ad − bc (the determinant)
∆ = T 2 − 4D (the discriminant)
The types of roots (eigenvalues) for a given system are determined by the discriminant ∆:
If ∆ > 0, then the roots are real and distinct (i.e. λ1 6= λ2 )

If ∆ = 0, then there is one real root of multiplicity two (repeated root) (i.e. λ1 = λ2 = λ)
If ∆ < 0, then there is a pair of complex conjugate roots, (λ1 = α + βi and λ2 = α − βi )
The various types of eigenvalues lead to different equilibrium states of the system. These
equilibrium states are described by the limit of the solutions (5.19) at infinity. We want to
characterise the equilibrium states using the eigenvalues, even without knowledge of the
solutions.
The following broad categories of critical points can be associated with the various
eigenvalue types:
1. Real, distinct roots: (λ1 6= λ2 )

For these ∆ > 0 (or T 2 > 4D).
Two situations arise:
i) Case 1: λ1 and λ2 have the same sign; i.e. λ1 , λ2 > 0 or λ1 , λ2 < 0.

Here the exponential terms of the solutions x(t ) and y(t ) in 5.19 will either grow
exponentially if λ1 , λ2 > 0, or decrease exponentially to a point. The critical point, in this
case, is a node. The orientation of the trajectories depends on whether |λ1 | < |λ2 | or vice
versa. This relationship determines which term grows/decays faster.
Figure 6.1: Nodes: a) Stable (λ1 , λ2 < 0) , b) Unstable (λ1 , λ2 > 0).
35
ii) Case 2: λ1 and λ2 have opposite signs; i.e. λ1 < 0 < λ2 or λ2 < 0 < λ1 . In terms of (5.19),
one of the exponential terms will decrease while the other increases. Note that since the
decreasing term never vanishes, the trajectories will not reach the critical point, but will get
closer and then divert away. Hence the critical point is a saddle point.
Figure 6.2: Saddle: (a) λ1 < 0 < λ2 , (b) λ2 < 0 < λ1 .
2. Real, repeated roots: λ1 = λ2 = λ

Here ∆ = 0 (or T 2 = 4D). In (20) the two terms combine to give a single exponential
function. The coefficients in x(t ) will be a constant multiple of those in y(t ). The critical
point, in this case, is an improper node.
Figure 6.3: Degenerate nodes (λ1 = λ2 = λ) a) Unstable improper nodes b) Stable improper
nodes c) Unstable Star point d) Stable Star point.
36
3. Complex conjugate roots: λ1 , λ2 = α ± βi :
Here ∆ < 0 (or T 2 < 4D). Two types of critical points arise for this case:
i) Case 1: A Centre: arises when Re(λi ) = α = 0 (pure imaginary roots). this is better seen in
terms of (5.25)-(5.26). When α = 0, the exponential factor reduces to 1, leaving only the
trigonometric terms, Hence a periodic behaviour of the solutions will occur. Sometimes
this critical point is referred to as a focus.
Figure 6.4: Neutrally stable Centre
ii) Case 2: A Spiral focus occurs if α, β 6= 0 and the trajectories around the equilibrium point
have spiral shapes. In (5.25)-(5.26), depending on whether α > 0 or α < 0, the solutions will
get larger or smaller, respectively, due to the exponential term. Thus the amplitude will vary
with time. At the same time some periodic behaviour, attributed to the trigonometric
factors will occur. Thus spiral curves will be generated from different initial points.
The orientation of the spirals depends on whether α and β have the same or opposite signs.
Figure 6.5: Above are Spiral Foci (α, β 6= 0) a) Stable, b) Unstable; Below are degenerates c)
Stable, d) Unstable
37
6.3 Stability
The phase plane equation alone is not enough to determine the stability of the critical
points of a given system. Time dependence must be considered.
We want to understand the stability of the equilibrium solutions of the system (4.1).
Suppose a system is slightly displaced from its equilibrium state and then allowed to evolve
with time. Stability is about the behaviour of the solution if the system is displaced from its
equilibrium state.
In broad terms, a critical point of a given system may either be stable or unstable.
A formal definition of stability is given below:
Definition 6.1 The critical point P (x e , y e ) is called stable if for every radius r 0 > 0 we can
find another radius r 1 > 0 (possibly smaller) such that any trajectory that passes through a
point inside the circle with centre P and radius r 1 will thereafter stay inside the circle with
centre P and radius r 0 . An equilibrium point that is not stable is called unstable.
Intuitively, a system is:
1. Stable: if for an initial point x 0 within a radius R of the critical point, the solution will
remain within this radius for any future time.
2. Asymptotically stable: if with time, it will return to its equilibrium state. On the phase
diagram, the direction field (arrows) will be towards the critical point. Thus an attractor is a
stable critical point.
3. Unstable: if it will be driven away from its equilibrium state as time increases. Here the
direction field (arrows) will be outward, away from the critical point. Hence a repeller is
unstable.
4. Neutrally stable: if the system oscillates indefinitely about, the equilibrium values after
the system is displaced from the equilibrium position. These are illustrated in Figure (6.6).
38
Figure 6.6: Stability scenarios
The question at this point is whether we always have to drag the phase portrait in order
to deduce the stability of the critical points. In some situations even plotting the phase
diagram may not be straight forward, and yet we will still be interested in understanding
the behaviour of the system about the critical points. In this section, we discuss ways of
determining stability properties without drawing the phase portraits, in a similar way that
we were able to determine the type of critical points.
6.4 Eigenvalues and linear stability

In an earlier discussion, we obtained solutions for a system in terms of exponential
functions. A closer look at the exponential function e λt will reveals the following:
If λ < 0 and real, the function will decrease asymptotically to zero as t → ∞.

If λ > 0 and real, the function will increase unboundedly as t → ∞.
If λ = α ± βi is complex, the form e αt ( cos βt ± sin βt ) suggests that the solution will have
¡ ¢ ¡ ¢
pure (periodic) oscillations if α = 0, and growing or decreasing oscillations due to the

presence of the e αt factor if α 6= 0.
Since the exponent, λt , of the solutions is associated with the characteristic roots of the
system, it seems logical to associate the behaviour of the solutions with the type and sign of
these eigenvalues. Thus an analysis of the stability of the systems can be derived from
knowledge of these eigenvalues.
We recall that the eigenvalues of the system (4.1) are given by

p
(a + d ) ± (a + d )2 − 4(ad − bc)
λ=
2
39
Continuing with the notation
T = a + d, D = ad − bc, ∆ = (a + d )2 − 4(ad − bc) = T 2 − 4D,
used earlier to categorise the equilibrium points by type, we now use these same conditions
to determine stability. A summary of the conditions and characterisation of the equilibrium
points for linear systems is now presented:
1. 4 > 0, : real and distinct

Roots T and D Type Stability
A. same sign D >0 Node
(a) both positive T >0 unstable
(b) both negative T <0 stable
B. opposite signs D <0 Saddle point unstable
C. one root is zero D =0 Degenerate
(a) other root positive T >0 unstable
(2) other root negative T <0 neutrally stable
2. 4 = 0, : real and equal

A. both nonzero Degenerate node
(a) both positive T >0 unstable
(b) both negative T <0 stable
B. both zero T =0 algebraically unstable
3. 4 < 0, : complex conjugates

A. real part has same sign Spiral focus
(a) real part positive T >0 unstable
Direction - clockwise (c < 0)
(b) real part negative T <0 stable
Direction - anticlockwise (c > 0)
B. real part zero T =0 Centre neutrally stable
N B.: If D = 0 (i.e. ad = bc) then the system has an equilibrium point at any point along the
straight line ax + b y = 0. This is Case 1C in the above summary table.
The above conditions are summarised in the following theorem:
40
Theorem (Stability Criteria-Linear systems)
An equilibrium solution is
• Stable if the real parts of both roots are less than zero (negative).
• Unstable if the real part of at least one is greater than zero (positive).
• Neutrally stable if the real part of one root equals zero and the other is less than or
equal to zero (unless both roots are identically zero)
Thus the linear system is stable if the eigenvalues lie in the upper left part of the complex
plane. If the eigenvalues are purely imaginary, the linear system has a centre and the
critical point is neutrally stable, but not asymptotically stable.
The above characterisation (type and stability) of the critical points can be summarised on
a T D-plane as shown in Figure 6.7
Figure 6.7: Diagram from Zill Cullen nineth edition indicating types of critical points for the
various values of T, D, and ∆ of a linear system
41
From the diagram in Figure 6.7, we note that:
1. The case ∆ = 0 corresponds to the parabola D = T 2 /4. The other types of equilibria are
either below ∆ > 0 or above ∆ < 0 this parabola.
2. Saddle points appear on the part of the plane below the T -axis. The small axes indicate
that the roots are real and have opposite signs.
3. Roots falling on the space above the T -axis and below the parabola would yield nodes.
4. Roots between the D-axis and the parabola correspond to foci and centres.
It is useful to be able to read this summary and extract information about the type of roots.
6.5 Exercises
1. Determine the type and stability of the models in the exercises of the previous section.
2. Refer to the systems given in Problem 2 of the exercises in chapter 4.
a) Discuss the type and stability of the equilibrium points. Compare your findings
using the phase diagrams obtained earlier.
b) Deduce the solutions of the systems.
42
Chapter 7
Nonlinear stability
Reading from Zill: Chapter 10. Unlike linear systems, stability analysis of nonlinear systems
is complicated by the possible existence of more than one critical point. Thus for nonlinear
systems, it is normal to speak of local stability since the behaviour of the system around a
particularly critical point may be different from the behaviour around another critical
point. One critical point may be stable while another critical point of the same system may
be unstable.
A nonlinear system with the general form
x 0 = f (x, y)
(7.1)
y 0 = g (x, y)
will have the nice form

" #" #
x0 a b x
= (7.2)
y0 c d y
which is convenient for analysis using eigenvalues.
For nonlinear systems, the above form can be used only as an approximation of the given
system. The transformation of a nonlinear system to the above form is called the
linearization of the nonlinear system. Under certain conditions, the stability of the
nonlinear systems can then be deduced from that of its approximate linear system. It turns
out that under such suitable conditions, whatever stability condition is revealed by the
approximate linear system holds true for the nonlinear system it approximates. It is
important to point out that nonlinear systems can be approximated by such linearization
in some cases and not in others. hence this technique needs to be applied cautiously.
43
7.1 Linearization
The stability analysis of nonlinear systems is based on the following Taylor theorem for more
than one variable:
7.1.1 Taylor’s Theorem

We need to make use of the Taylor theorem for multivariable functions about a point (x 0 , y 0 ).
It is stated as follows:
f (x, y) = f (x 0 , y 0 ) + f x (x 0 , y 0 )(x − x 0 ) + f y (x 0 , y 0 )(y − y 0 ) + ..... (7.3)
or
f (x 0 + h, y 0 + k) = f (x 0 , y 0 ) + h f x (x 0 , y − 0) + k f y (x 0 , y 0 ) + .... (7.4)
We apply this result and find an approximation of the given nonlinear system by using a
Taylor expansion of f (x, y) and g (x, y) near an equilibrium point (x e , y e )
We begin by supposing an equilibrium point of the nonlinear system is (0, 0). Then for
points (x, y) near (0, 0) we can use Taylor’s theorem to write
f (x, y) = f (0, 0) + f x (0, 0)x + f y (0, 0)y + ....

g (x, y) = g (0, 0) + g x (0, 0)x + g y (0, 0)y + ....
Combining the above two expansions provides a way of writing the given nonlinear system
as
x 0 = ax + b y + Φ1 (x, y)
y 0 = c x + d y + Φ2 (x, y)
or
Ã ! Ã ! Ã !Ã ! Ã !
x0 x a b x Φ1 (x, y)
=A = + . (7.5)
y0 y c d y Φ2 (x, y)
Ã ! Ã !
x0 x
=A . (7.6)
y0 y
N.B. A is the Jacobian of f (x, y) and g (x, y) evaluated at the equilibrium point (x e , y e ).
44
Definition 7.1 An equilibrium point is said to be simple if its linearised system xr has no
zero eigenvalues; i.e. if d et (A) 6= 0
Definition 7.2 A simple critical point is called hyperbolic if A in its linearization of Ax has
no eigenvalues with zero real parts; i.e Re(λi ) 6= 0 for all i.
Example 7.1
Consider the system
x 0 = x + 2y + x 2 , y 0 = 2x + y + x y 2 .
Clearly (x e , y e ) = (0, 0).

f x (0, 0) = 1, f y (0, 0) = 2, g x (0, 0) = 2, g y (0, 0) = 1
The point (0, 0) is a simple point since λi are 3 and 1, and d et (A) = −3 6= 0. It is also
hyperbolic since both roots are real.
For a general equilibrium point (x e , y e ) we have the Taylor expansions
f (x, y) = f (x e , y e ) + f x (x e , y e )(x − x e ) + f y (x e , y e )(y − y e ) + .....

g (x, y) = g (x e , y e ) + g x (x e , y e )(x − x e ) + g y (x e , y e )(y − y e ) + .....
Thus we can let
a = f x (x e , y e ), b = f y (x e , y e ), c = g x (x e , y e ), and d = g y (x e , y e ) (7.7)
We can define new variables X = (x − x e ), Y = (y − y e ) and the above system becomes

Ã ! Ã ! Ã !Ã ! Ã !
x0 X0 a b x − xe Φ1 (X , Y )
= = + .
y0 Y0 c d y − ye Φ2 (X , Y )
Definition 7.3 The equilibrium point (x e , y e ) is called isolated if there is some circle around
which there are no other stationary points of the system. This is automatic if ad − bc 6= 0 for
the matrix. A of the linear approximation to the system near (x e , y e ).
Activity 7.1
Consider the system
x 0 = y, y 0 = 2x + x 2 (7.8)
1. Recall/Identify the critical points of the system. Determine whether they are
isolated. isolated.
2. write a linear approximation of the system about each of the critical points.
45
7.1.2 Linearization Theorem
If we are going to use an approximation of the given system, we need to be certain that any
conclusions about the type and stability of critical points derived from the linearized
system, will be related and informative of the type and stability of the critical points of the
original nonlinear system. This is, is the structure of the nonlinear system the same as that
of its linearization? The following theorem provides conditions for the equivalence of the
two systems.
Theorem 7.1 Let the nonlinear system (4.1) have a simple critical point (0, 0) (for
simplicity). Then the neighbourhood U of this equilibrium, the phase portraits of the
nonlinear system (4.1) and its linearization in (7.7) are equivalent.
In other words, if λi are the eigenvalues of the linearised system, then so long as the critical
point is simple, the nonlinear system and the corresponding linearised system are
equivalent.
In fact, provided ad − bc 6= 0 and the matrix A has distinct eigenvalues that are not purely
imaginary, the stationary points of the nonlinear system and its linear approximation will
always be of the same type.
Exceptions:
• ad − bc = 0: at least one root is 0 - the phase portrait of the system and its linear
approximation may differ.
• Linear approximation has a centre.
• Eigenvalues of A are equal and nonzero - stability may be the same but the type may
differ; e.g. Linear has a node and nonlinear has a spiral point.
Thus near a simple critical point the system is structurally stable - it is robust to small
perturbations. So if a critical point is simple it is enough to study the linearization Ax, as if
it was a linear system x 0 = Ax.
Because we have performed a transformation for points assumed to be near the particular
equilibrium point, we say that the stability analysis is local. Locally, we can use
characterisation of the stability using the theory of linear systems. Each point, if isolated,
has to be studied separately.
46
7.2 Local stability
Stability of nonlinear systems is determined by the linear approximation to the system at
the equilibrium point, provided that the coefficients of the linearised system are such that
ad − bc 6= 0 and the linear approximation does rot have a centre.
Let (x e , y e ) be an equilibrium point of the autonomous system

dx dy
= f (x, y), = g (x, y),
dt dt
and let
dx
= a(x − x e ) + b(y − y e ), (7.9)
dt
dy
= c(x − x e ) + d (y − y e ), (7.10)
dt
be the linear approximation to the system near (x e , y e ). Assume that ad − bc 6= 0. Then the
given
Ã equilibrium
! point is asymptotically stable if the real parts of the eigenvalues of
a b
A= are both negative and is unstable if at least one of those real parts is positive.
c d
Following the above synopsis we can summarise the nature of critical points for nonlinear
systems in Figure 7.1.
Example 7.2
The system
x 0 = y, y 0 = 2x − x 2 ,
has the points at (0, 0) and (2, 0). (0, 0) is a simple critical point (Φ(x, y) = x 2 and is a saddle
point.
For the critical point (2, 0) we make a change of variable z = x − 2 (i.e. x = z + 2) and obtain
the system
z 0 = y, y 0 = −2z − z 2 .
The point (0, 0) is a centre of the corresponding linear system. From this we have that (0, 0)
is either a centre or a spiral focus for the original system. We further observe that a
trajectory C passing through the positive x-axis near (0, 0) must intersect the negative
x-axis. But the last system is changed if we replace y by −y and t by −t . which implies that
C is closed. Therefore (0, 0) is a centre, so (2, 0) is a centre for the original system.
Activity 7.2
Use linearization to determine the type and stability of the system
x 0 = y − x 2 + 2, y 0 = x 2 − x y.
47
Figure 7.1: Diagram from Zill Cullen nineth edition indicating types of critical points for the
various values of T, D, and ∆ when analysing a nonlinear system
7.3 Limit Cycles

It is possible that the local stability of certain nonlinear systems will not be explained by
linearization.
Consider the following example.
Example 7.3
dx
= µx − ωy − x(x 2 + y 2 ) = f (x, y, µ, ω), (7.11)
dt
dy
= µy + ωx − y(x 2 + y 2 ) = g (x, y, µ, ω), (7.12)
dt
Where µ and ω are parameters. The equilibrium point ls determined from
µx − ωy − x(x 2 + y 2 ) = f (x, y, µ, ω) = 0, (7.13)
µy + ωx − y(x 2 + y 2 ) = 0, (7.14)
48
Multiplying (7.13) by y and (7.14) by x and subtracting leads to
µx y − ωy 2 = ωx 2 + µx y,
and the only critical point is (0, 0).

Ã !
µ −ω
The linear approximation is such that A = so that
ω µ
T = 2µ, D = µ2 + ω2 > 0, ∆ = 4µ2 − 4µ2 − 4ω2 = −4ω2 < 0,
and λ = µ ± ωi . Thus the type and stability of the critical point depends on the sign of µ:
If µ < 0, the system has a stable spiral focus at (0, 0).
If µ = 0, the system has a neutrally stable centre at (0, 0).
If µ < 0, the system has an unstable spiral focus.
The following points are important and they describe the nature of limit cycles:
1. Stability changes from stable to neutral to unstable as µ changes from negative to

positive.
2. µ = 0 corresponds to a neutral trajectory that separates the phase plane into two. This
phenomenon is known as a limit cycle or bifurcation. Unlike the orbits of an oscillatory
system, it is isolated (it is the only one in the neighbourhood).
The above points are typical of limit cycles in general. There is a change of stability is
certain parameters change from being negative to being positive. Also the limit cycles are
isolated in terms of the trajectories immediately around them.
The question may arise, how do we know whether we have a limit cycle or a periodic
solution?
Figure 7.2
49
7.4 Exercises
1. Use linearization, where appropriate, to determine the type and stability of the critical
points of systems below. Sketch the phase portraits.
a) x 0 = −4y + 2x y − 8, y 0 = 4y 2 − x 2
a) x 0 = x 2 − y 2 − 1, y 0 = 2y
a) x 0 = ye y , y 0 = 1 − x2
a) x 0 = x − y, y 0 = x2 − 1
a) x 0 = 1 − 2x y, y 0 = 2x y − y
a) x 0 = y − x 2 + 2, y 0 = x2 − x y
2. Discuss the nature and stability of the critical points of the system modelled by the
following nonlinear system obtained from the differential equation.
d 2x dx
2
+6 − x 2 + 4x = 0.
dt dt
3. Consider the system
x 0 = y + x(1 − x 2 − y 2 )
y 0 = −x + y(1 − x 2 − y 2 )
a) Show that the transformation x = r cos (θ), and y = r sin (θ) , leads to the system
r 0 = r (1 − r 2 ), θ 0 = −1. (7.15)
b) Use the transformed system to show that the system has a limit cycle. Describe the limit
cycle and determine whether it is stable.
c) Use linearization to determine stability of the system. Does this method help to determine
stability of the system? Explain.
50
Chapter 8
Modelling Applications - 2D Systems
Reading: Zill and Cullen, Chapter 5
8.1 The linear Oscillator

Recall the example of the linear oscillator (or spring-mass system) outlined earlier in the
course. The following steps will guide you in formulating a model for the system.
8.1.1 Model Formulation

The simplest model is based on the assumption that the spring system moves in a media
free from friction or air resistance.
1. Identify the two parameters involved in the system (the values that have to be known).
Identify also the equilibrium position of the system.
2. Now suppose the body is displaced slightly of its equilibrium position and then
released from rest. Identify and describe the two forces acting in the system.
3. What physical law can be used to describe the downward force?
4. Write a mathematical expression that represents the fact that the upward force is
opposite the downward force and proportional to the displacement of the body from
equilibrium.
5. Show that the system can be modelled as
d 2x
= −ω2 x (8.1)
dt2
51
8.1.2 Qualitative analysis
In order to apply the techniques we have discussed for sketching phase diagrams we need
to convert this second order differential equation to an equivalent system of linear first
order equations.
Activity 8.2
1. Derive the system of first order equations.
2. Draw a phase diagram of the system.
3. Use the phase diagram to deduce the oscillatory behaviour of the system.
4. Use algebraic techniques to analyse the system.
You should have realised that
• y = dd xt gives the velocity of the body. Thus y = 0 means the body is at rest.
dy 2
• dt = dd t x2 is acceleration, which we associate with the restoring force from the spring.
d 2x
¶
(NB: m 2 = F = −kx.
dt
8.2 Nonlinear Models
8.2.1 Interacting Populations

General Model
Models of interacting populations are examples of nonlinear models. Earlier in the module
the ways in which populations was mentioned. Now we look at related models more closely.
The most general form of a model for interating populations of sizes x and y is given by
dx
= x(a − bx − k y) (8.2)
dt
dy
= y(c − d y − σx) (8.3)
dt
where a, b, c, d , k, and σ are positive constants of proportionality. The following activity is
meant to spark some thoughts about the population system.
Activity 8.3
52
Suppose x, y 6= 0. (What does this assumption mean?)
1. Let b = k = d = σ = 0.
(a) What is the resulting population models?
(b) How do each species grow? Does the interaction affect the growth of the species?
(c) What is the physical meaning of the constants a and c?
2. Now suppose that only k = σ = 0.
(a) How does this affect the rate of growth of each species?
(b) Does the presence of one species affect the other species? Explain.
(c) Interpret the constants b and d .
(d) Give an example of two species that interact in this way.
3. Now suppose that b = d = 0.
(a) Describe how the species interact in this case and how their growth rate is
affected.
(b) Give an example of species that interact in this way.
4. Do the same for c = 0.
A Predator-prey Model
One of the popular models of a predator-prey intraction is due to Lotka and Volterra, and
has the general form:
dx
= x(a − bx − c y) (8.4)
dt
dy
= y(−k + λx) (8.5)
dt
Activity 8.4
1. Which of the variables x and y is associated with prey, and which one with predator.
2. Describe the meaning of each constant in the model. The Case of Unlimited Food
Source for the Prey (b = 0)
53
The Case of Unlimited Food Source for the Prey (b = 0)
The above model is further simplified by assuming that there resources are available in
abundance to support the prey population. Under this assumption the above system
becomes
dx
= x(a − c y) (8.6)
dt
dy
= y(−k + λx) (8.7)
dt
Activity 8.5
1. Describe the situations represented by x = 0 and y = 0 and their effect on the system.
2. Identify the equilibrium solutions.
3. The prey:
Describe the conditions under which the prey population is at equilibrium, increasing
and declining in terms of the constants and the predator population.
4. Do a similar investigation for the predator, stating the effect of an increase in the
constants k and λ.
5. Draw the phase portrait of the system.
6. Discuss the stability of the system using (a) The phase diagram; (b) Linearisation.
8.2.2 The Nonlinear Oscillator (Simple pendulum)

Let us now discuss the system of a simple pendulum cited earlier as an example of a physical
system.
Activity 8.6
1. Why is this a 2D system?
2. What do L and θ as shown in the diagram represent?
3. Identify the equilibrium position (physical) of the system.
4. Suppose the mass is displaced slightly from the equilibrium position and held in that
position. Describe all the forces acting in this system and express them in
mathematical form.
54
5. If the mass is released from its displaced position, show that the motion will be
described by the equation
d 2θ
+ q sin (θ) = 0, q > 0 (8.8)
dt2
6. Interprete each term in the model.
7. Use the substitution x = θ, ẋ = y, to transform the second order differential equation

to a system of first order equations.
8. Use a phase diagram to analyse the system behaviour.
9. Use linearisation to analyse the system.
10. Compare the two approaches to the analysis of the system. Which one was easier?
Which one revealed more information? You will have the opportunity to explore
pendulum further in the tutorials. Other variations of the system will also be
investigated.
There are a number of areas where sytems are modelled by nonlinear systems. Here a few
are presented.
8.2.3 Electrical circuits

Consider the RLC electrical circuit consisting of a capacitor (C ), a resistor (R), and an
inductor (L) in a simple closed loop. The effect of each component of the circuit is
measured in terms of the relationship between current, I , and voltage V on that branch of
the loop. An idealised physical model gives the relations
dVC
C = IC , (capacitor) (8.9)
dt
VR = f (I R ) , (resistor) (8.10)
d IL
L = VL , (inductor) (8.11)
dt
where VC represents the voltage across the capacitor, I R represents the current through the
resistor, and so on. The function f (x) is called the V − I characteristic of the resistor. In the
classical linear model of the RLC circuit we assume that f (x) = R x where R is the
resistance. Kirchoff’s current law states that the sum of the currents flowing into a node
equals the sum of the currents flowing out. Kirchoff’s voltage law states that the sum of the
voltage drops along a closed loop must add up to zero. Determine the behaviour of this
55
circuit over time in the case where L = 1,C = 1/3, and f (x) = x 3 + 4x.
Assumptions: IC = I R = I L , VC + VR + VL = 0 (from Krchoff’s law).

Note that there six state variables for the system, but Kirchoff’s law can be used to reduce
them to two.
Let x 1 = I R = IC = I L and x 2 = VC = VR = VL . The equations of the system are
ẋ 2 = x 1 , VR = x 13 + 4x 1 , ẋ 1 = VL , x 2 + VR + VL = 0
Substituting and rearranging we get the system
ẋ 1 = −x 13 − 4x 1 − x 2 , ẋ 2 = 3x 1
The system has an equilibrium point at (0, 0). We perform an analysis of the system based
! is done on the x 1 x 2 - plane (current vs.

on its linear approximation. Note that Ãthe analysis
−4 −1
voltage) The linearised system has A = , whose eigenvalues are r = −3, −1. Since
3 0
the eigenvalues are negative, the equilibrium is a stable node.
In terms of the physical system, the overall behaviour can be described in terms of two
quantities: the current through the resistor and the voltage drop across the capacitor.
Regardless of the initial state of the circuit, both quantities eventually tend to zero.
Furthermore it is eventually true that either voltage is positive and current is negative, or
vice versa. The behaviour of other quantities of interest can be described in terms of these
two variables (based on the equations of the system). For example x 1 actually represents
the current through any branch of the circuit loop.
8.2.4 A general epidemic model
Consider the spread of an infectious disease in a population of N individuals. Suppose that

at time t there are
x(t ) susceptibles: those so far not infected and therefore liable to infection;
y(t ) infectives: those who have the disease and are still at large;
z(t ) who are isolated, or who have recovered and are immune (or even dead)
Various models have been developed for the study of the epidemics. On of the simplest one
supposes that β is the rate at which susceptibles are infected, and γ is the rate at which
56
current infectives become isolated. Then the dynamics of the disease is described by
dx
= −βx y
dt
dy
= βx y − γy (8.12)
dt
dz
= γy
dt
57
8.3 Exercises
1. Consider the model of a simple pendulum governed by the system
θ̈ + 2k θ̇ + q sin (θ) = 0, k, q > 0
• What is the physical interpretation of the extra term (first derivative term)?
• Obtain a linear approximation of the system.
• Use the analysis of the linearised system to determine local stability near the
equilibrium points.
• Sketch a phase portrait of the system for −2π < θ < 2π.
2. A spring whose mathematical model incorporates a nonlinear restorative force is

called a nonlinear spring. Two such models for a hard spring and a soft spring are,
respectively
d 2x
2
+ x + x 3 = 0, and (8.13)
dt
d 2x
+ x − x3 = 0 (8.14)
dt2
• Use linearisation to determine the stability of the systems.
• Explain your observation of the behaviour of the system.
• Use a numerical solver to obtain the phase plot and solutions of the two systems
using the initial conditions x(0) = 2, ẋ(0) = 0 and x(0) = 2, ẋ(0) = −3 in turn.
Compare your results with the conclusions drawn in (a) above.
3. A rumour spreads through a closed population of constant size N + 1. At time t the

total population can be classified into three categories: x persons who are ignorant
ot the rumour; y persons who are actively spreading the rumour; z persons who have
heard the rumour but have stopped spreading it: if two people who are spreading the
rumour meet then they stop spreading it. The contact rate between any two categories
is µ. Show that the equations
ẋ = −µx y, ẏ = µ[x y − y(y − 1) − y z]
give a deterministic model of the problem. Find the equations of the phase paths and
sketch the phase diagram. Show that, when initially y = 1, and x = N , the number of
people who ultimately never hear the rumour is x 1 , where
³x ´
1
2N + 1 − 2x 1 + N log =0
N
58
4. In a simple model of a national economy, I˙ = I − αC , Ċ = β(I −C −G), where I is the
national income, C is the rate of consumer spending, and G the rate of government
expenditure; the constants α and β satisfy 1 < α < ∞, 1 ≤ β < ∞.
2
Discuss an economy in which G = G 0 + k I , and show that there are two equilibrium
states.
5. The blue whale and fin whale are two similar species that inhabit the same areas.
Hence, they are thought to compete. The intrinsic growth rate is estimated at 5% per
year for the blue whale and 8% per year for the fin whale. The environmental carrying
capacity (the maximum number of whales that the environment can support) is
estimated at 150000 blues and 400000 fins. The extent to which the whales compete
is unknown. In the last 100 years intense harvesting has reduced the whale
population to around 5000 blues and 70000 fins. Using the logistic growth of the
individual populations determine whether the blue whale become extinct?
59
Chapter 9
Discrete systems
9.1 Why Discrete Models

Up to now we have discussed the analysis of continuous-time systems. For many
organisms, births occur in regular, well-defined ’breeding seasons’. Hence models of their
growth cannot be reealistically based on continuous time. Most trees thower repeatedly on
a seasonal basis. Some flower once and die. Insects may have one or more generations per
year. The growth process involve several distinct stages which, when taken into account,
lead to various models. Some fish species reproduce seasonally, and more than that some
die after spawning. Breeding seasons for birds can be quite short. with marked egg laying
and hatching periods. Some mammal species ovulate once a year and produce a single
litter. While for some births can occur at different times for different individuals, for others
births are highly synchronised to within a day or two in a population. Mammals can be
modelled by age-structured models with discrete, nonoverlapping generations. The census
of such populations are observed at specific time intervals. hence a disctrete time frame
would be more suitable to model the dynamics in some cases.
9.2 From Difference Equations to Discrete Systems

The general form of a difference equation of order n is
F (x k , x k−1 , x k−2 , . . . , x k−n ) = 0
or
x k = f (x k−1 , x k−2 , . . . , x k−n ) (general) (9.1)
x k = a 1 x k−1 + a 2 x k−2 + . . . + a n x k−n linear (9.2)
60
with the initial conditions x 0 , x 1 , . . . , x n−1 given.
Thus the first and second-order difference equations have the respective general forms:
x k = f (x k−1 ) , x 0 given (9.3)
x k = f (x k−1 , x k−2 ) , x 0 , x 1 given (9.4)
The system (9.1) can be reduced to a system of first order difference equations by defining
variables as follows
y 1 (k) = x k−n = y 2 (k − 1)
y 2 (k) = x k−n+1 = y 3 (k − 1) (9.5)
..
.
y n−1 (k) = y n (k − 1)
(9.6)
y n (k) = x k
The resulting system has the form
y 1 (k) = y 2 (k − 1)
y 2 (k) = y 3 (k − 1)
.. (9.7)
.
¡
y n (k) = F y 1 (k − 1), . . . , y n (k − 1)
if nonlinear, and  
0 1 ... 0
   ..  
y 1 (k) 
 0 0 . 0  y 1 (k − 1)

 ..   .. ..
 .. 

 . =
 
. . 0 
 . 
 (9.8)
 
y n (k) 
 0 ... 0 1  y n (k − 1)

an . . . a2 a1
or y(k) = Ay(k − 1)
The analysis of discrete systems is analogous to that of systems of ODE’s.
9.3 Discrete First Order Equations

The general form of discrete first order equations is
x k+1 = f (x k )
Linear first order equations have the form
x k+1 = ax k , x 0 given (Homogeneous) (9.9)
61
x k+1 = ax k + b, x 0 given (Non-homogeneous) (9.10)
9.3.1 Equilibrium Solutions

For discrete first order equations, the equilibrium point satisfies
x k+1 − x k = 0 or x e = f (x e ) (9.11)
b
in particular for (9.10) the equilibrium point is x e = 1−a .
9.3.2 General Solutions

General solutions of the above equations are obtained by successive substitution of x k ,
starting with x 0 . Recognising a general pattern of the solution may be difficult for nonlinear
systems than it is for linear systems.
What follows is a derivation of the general solution of the linear equations.
First consider equation (9.9). Starting with a given x 0 , successive substitution yields
x 1 = ax 0
x 2 = ax 1 = a 2 x 0
:
xn = a n x0
Hence the general solution:

xk = a k x0 (9.12)
Thus the solution is a geometric sequence.
Equation (9.10) on the other hand yields
x 1 = ax 0 + b
x 2 = ax 1 + b = a (ax 0 + b) + b = a 2 x 0 + (a + 1)b
..
.
x n = a n x 0 + a n−1 + a n−2 + . . . + 1 b
¡ ¢
1 − an
= a n x0 + b
1−a
Hence the general solution for the non-homogeneous case is
k
xk = a k x 0 + 1−a
1−a
b
62
b b
µ ¶
k
=a x0 − + , a 6= 1 (9.13)
1−a 1−a
= x c (k) + x p (9.14)
where x c (k) and x p = x e denote the complementary and particular solution of the equation.
9.3.3 Stability Analysis

The solutions of the first order difference equations (65) and (66) for a given x 0 were found
to be respectively
xk = a k x0 , (x e = 0) (9.15)
x k = (x 0 − x e ) a k + x e (9.16)
Stability in this case means x k → x e at t → ∞. For this to happen, |a| < 1, must hold for
both solutions; i.e. −1 < a < 1. If 0 < a < 1, a k → 0 as t → ∞ in positive descending steps.
If −1 < a < 0, x k → x e with alternating steps on either side of x e (oscillations) (see Figure
-reffigdec). In either case the solution (or the system) is stable.
Figure 9.1: Geometric and Oscilaltory decay: |a| < 1
If |a| > 1, x k will diverge from x e , i.e. a k will climb up the steps indefinetely in larger and
larger steps if 1 < a or in alternating steps if a < −1 (see Fig 9.1). In either case the solution
(or system) is unstable.
63
Figure 9.2: Geometric and Oscillatory increase: |a|>1
Phase Diagrams
Constructing phase diagrams for first order discrete equations involves plotting x k against
x k+1 in the x k , x k+1 -plane. ’The equilibrium point x e in (67) is at the intersection of the
graphs of x k+1 = x k (a 45◦ -line) and x k+1 = f (x k ).
On a phase diagram stability analysis is equivalent to systematics movements between the

45◦ line and f (x k ). Starting from an arbitrary x(0) = x 0 , f (x k ) allows us to read x 1 = f (x 0 ).
The 45◦ -line then translates x 1 for t = 1 from the vertical to the horizontal axis where x 1 is
now taken as the new initial condition for t = 1.x 2 = f (x 1 ) then gives x 2 for t = 2 and so on.
The existence of the solution (the fixed point) depends on whether f (x k ) intersects the
45◦ -line. The resulting scinario is sometimes referred to as the Cobweb diagram
Further we note that in (65) and (66), f (x) = ax + b with b equal or not equal to zero. In
both cases f 0 (x) = f 0 (x e ) = a. Thus the requirement for stability that |a| < 1 is equivalent to
saying f 0 (x e ) |< 1
Also, phase diagrams can be used to sketch the graphs of the soiutions of the given
equations, in the same way that it is done for continuous time.
By observing that the coefficient a in the linear systems is the derivative of f (x k ) with
respect to :
64
Figure 9.3: Geometric and Oscillatory increase: |a|>1
x k , it is natural to generalise and say that stability of first order systems, linear or nonlinear
can be deduced from f 0 (x e ). More specifically, the system x k+1 = f (x k ) is stable if ¯ f 0 (x e )¯ <
¯ ¯
1 and unstable otherwise.
9.4 Discrete Second Order Systems

The general form is
¡ ¢
x k+1 = f x k , y k (9.17)
¡ ¢
y k+1 = g x k , y k (9.18)
which may be given as is, or obtained from converting a second order difference equation
x k+1 = f (x k−1 , x k )
65
The linear case is " # " #" #
y 1 (k + 1) a b y 1 (k)
= (9.19)
y 2 (k + 1) c d y 2 (k)
which may result from converting the linear difference equation
x k+1 = a 1 x k + a 2 x k−1 (9.20)

" #
y 1 (k)
We shall use y k = ; i.e. unsubscripted variables denote vectors.
y 2 (k)
The equilibrium solutions for second order systems satisfy
¡ ¢ ¡ ¢
xe = f xe , y e , y e = g xe , y e (9.21)
simultaneously.
9.4.1 Linear Stability Analysis

The stability analysis of a discrete second order system is similar to that of the counterpart
continuous systems.
Phase Diagrams - Plane Systems
Phase diagrams for discrete second order systems are constructed in the same way as for
the ODE systems. The only difference is that the points are discrete and connected as
continuous curves only as a visual aid (Figure 9.4).
Figure 9.4: Phase diagram and solution sketch for a 2-D system
Eigen values of system (9.19) can be obtained in the same way as in the continuous case.
The same eigenvalues can be obtained by solving the equation
λ2 − a 1 λ − a 2 = 0
66
if the former is equivalent to the later (by conversion). The stability condition for discrete
systems differs from that of continuous systems. Hence the following definition is
important to note.
Definition 9.1 An equilibrium point is called asymptotically stable if |λi | < 1 for all i ; i.e. all
eiqenvalues lie inside the unit circle. N.B.
• "All" means all cases of the roots of the characteristic equation, real and complex.
• For λ = α ± βi , this means ρ = α2 + β2 < 1

p
We can also use the notation used before. The characteristic equation is P (λ) = λ2 −T λ+D =
0 and 4 = T 2 − 4D. The stability conditions of discrete systems are equivalently
|P (0)| = |D| < 1, P (−1) > 0, and P (1) > 0
which imply
|D| < 1 and |T | < 1 + D
Types of Equilibrium Solutions
Equilibrium solutions of discrete systems can be classified in a way similar to continuous

systems. These are summarised below:
1. 1. Case 1: ∆ > 0, real and distinct roots
If |λi | < 1, i = 1, 2 the equilibrium point is a stable node.

If |λi | > 1, i = 1, 2 the equilibrium point is an unstable node.
If |λi | < 1 and ¯λ j ¯ > 1, i 6= j , the equilibrium point is a saddle point.
¯ ¯
2. 2. Case 2: 4 = 0, λ1 = λ2 = λ
If |λ| < 1, the equilibrium point is a stable improper node.

If |λ| > 1, the equilibrium point is an unstable improper node.
3. Case 3: 4 < 0, λ = α ± βi
If ρ = α2 + β2 < 1, the equilibrium point is a stable focus.

p
If ρ > 1, the equilibrium point is an unstable focus. If ρ = 1, the equilibrium point is a

centre.
Note that even if α, the real part of the eigenvalues, is zero, there can still be a spiral
focus.
67
Solutions of Linear Second Order Systems
The solution of
y k+1 = Ay k y(0) = y 0
is, by iteration
yk = Ak y0, for any t
By an analysis similar to ODEs, the systems have eigenvalues λ1 , . . . , λn and their

corresponding eigenvectors v 1 , . . . , v n . In terms of these the solution can be expressed as:
Case 1: Real distinct roots
yk = c 1 v1 λk1 + . . . + c n vn λkn (9.22)
The solution of the nonhomogeneous system
yk = Ayk−1 + b, y(0) = y0
is
yk = c 1 v1 λk1 + c 2 v2 λk2 + ye (9.23)
where y e is the equilibrium solution.
Case 2: Multiple roots:

³ ´
yk = C 1 +C 2 k λk (9.24)
Case 3:Complex roots(λ = α ± βi )

" #
k y 1 (0) cos (k) φ + y 2 (0) sin (k) φ
yk = ρ (9.25)
y 2 (0) cos (k) φ − y 1 (0) sin (k) φ
or ³ ´
yk = ρ k C 1 cos (k) φ +C 2 sin (k) φ (9.26)
³ ´´−1
β
³
by the superposition property, where ρ = α2 + β2 and φ = tan α
p
.
68
9.4.2 Nonlinear Stability Analysis
As for continuous systems, local stability can be determined from the linearised systems
near the equilibrium points.
Using the characteristic equation of the corresponding linearised system, it is suffiecient to

test whether the following condition is satisfied:
2 > 1 + D > |T |
¡ ¢
i.e. If both eigenvalues |λi | < 1, x e , y e is stable.
In the next section we discuss selected examples of disctrete systems and their analysis.
69
Appendix
A Conversion of higher order ODEs

In many cases systems are modelled using ODEs of order higher than one. Inorder to apply
the analysis techniques discussed for first order systems, these higher order ODEs have to
be converted to systems of first order ODEs.
Example A.1
The ODE
y 00 + 2y 0 − 5y = x 2
has the equivalent first order system
y 10 = y 2
y 20 = x 2 + 5y 1 − 2y 2
Any n-th order differential equation can be converted to an equivalent system of first order
equations by setting the dependent variable in the original equation to be a (preferably)
indexed variable and then taking derivatives of the new variables and assigning them names,
up to the highest required derivative. Suppose a system is represented by the n-th order ODE
a n (t )y (n) + a n−1 (t )y (n−1) + . . . + a 1 (t )y 0 + a 0 (t )y = f (t , y)
Then the system is converted as follows:
Let y = y 1 . So that
y 10 = y 2
y 20 = y 3
..
.
0
y n−1 = yn
y n0 = y n = f (t , y) − â 0 (t )y 1 − â 1 (t )y 2 − . . . − â n−1 (t )y n
ak
where â k = an
. Note that the system consists of n differetnial equations. Once the system is
in this form, standard numerical techniques can be used for analysis.
70
Example A.2 Conversion from ODE to linear systems
d2y dy
To convert the differential equation t 2 d t 2 − 7t d t + 16y = 0 we can let
d y d y1 d y2 d 2 y
y = y1, = = y2, =
dt dt dt dt2
This leads to the system
d y1
= y2
dt
d y2
¶
1 ¡ ¢ 1 1
= 2 −16y 1 + 7t y 2 = −16 2 y 1 + 7 y 2
dt t t t
Note
71

Notes

Uploaded by

Copyright:

Available Formats

Notes

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Notes

Uploaded by

Copyright:

Available Formats

M ATHEMATICAL M ETHODS AND M ODELLING

2 First Order Systems 9

3 Modelling Applications - 1D Systems 13

4 Plane (2D) Autonomous Systems 17

5 Solutions of linear Systems 28

8 Modelling Applications - 2D Systems 51

1.1 Assumed background for the course

• Solution of equations - linear and nonlinear.

• Simultaneous equations - linear and nonlinear.

• Inequalities - linear and nonlinear.

2. Functions and their graphs - incl exponential, hyperbolic

3. The derivative and curve sketching

• Increasing and decreasing functions.

• Eigenvalues and eigenvectors.

1.2 Review of the Modelling Process

• Specify assumptions, simple first.

Step 2: Select the modelling approach

• Technique to be used; e.g. differential equations, stochastic model, etc.

Step 3: Formulate the mode]

• Define the functional relationship between variables and translate them to

Step 4: Solve the model

• Graphical sketches /solutions.

Step 5: Analysis (and validation) of results

• Interpret the results.

• Re-examine model formulation.

1.3 Some popular real-life systems

1.3.1 Physical systems

Example 1.2 The pendulum

Figure 1.2: Pendulum

Example 1.4 Interacting populations

Of particular interest would be the modelling of the interaction of populations of two

1. Symbiosis ( or mutualism): where the existence of one population enhances the of

Example 1.5 Infectious deceases

1.4 Systems of differential Equations

1.4.2 Solutions of ODE Systems

1.5 Basic terminology

1.6 Qualitative Analysis

• Identifying equilibrium solutions/ critical points.

• Sketch a phase diagram.

First Order Systems

We begin with the study of first-order autonomous systems of the form

First, we are interested in determining the equilibrium solution x e , of the system.

2.2 Phase diagrams of 1D Systems

Example 2.1 (Linear systems)

2.3 Stability of critical points

In more formal terms, an attractor is said to be stable and a repeller is unstable.

2.4 Stability condition for ID systems

• If f 0 (x e ) = 0, then analyse the problem further to determine the stability of x e .

• If f 0 (x e ) > 0, then x e is unstable.

(Try to use a computer to help you with this exercise.)

a) Find all critical points.

Modelling Applications - 1D Systems

Reading: Zill and Cullen, Chapter 3

3.1 The Logistic Population Model

Density-dependent Logistic Model

• Assume N is large enough to be modelled as a function of t

Density-dependence is a term associated with the concept of carrying capacity. If it is

3.2 Solutions of 1D systems

a) Find the solution of the density dependent model when r = 2, K = 100.