Pages 16 To 23 (Annotated)

Stable nodes and spirals are known as attractors, whilst unstable nodes and spirals are
called repellers. We have only considered the case of nodes at the origin in this general
linear system, but if we had a system like
(12)
=
X AX + C
for a constant vector C, then the critical point could occur at (xe , ye ) which is not (0,0) . All
we need do is to choose new variables which shift the critical point to the origin:
=
X xe and =
Y ye
Since the first critical point of (12) is given by
a1 xe + b1 ye + c1 =
0
then changing variables in (12) yields
d dX
=
= a1 ( + xe ) + b1 ( + ye ) + c1
dt dt
= ( a1 xe + b1 ye + c1 ) + a1 + b1
(13)
= a1 + b1
Similarly
d dY
= = a2 + b2
dt dt
(14)
So the system (13) and (14) arising from the change of variable is again one with the critical
point at the new origin and all of our earlier classification of behaviour is still valid.
Nonlinear Systems and Linearization

Consider a general system of two nonlinear ODEs
dX
= F ( X ,Y )
dt
dY
= G ( X ,Y )
dt
(15)
Let (xe , ye ) be a critical point then we know=

that F (xe , ye ) 0=
and G(xe , ye ) 0 .
Consider solutions close to the critical point
=
X (t ) xe + (t ) and Y(t)=ye + (t )
where (t ) and (t ) are small (called perturbations) and approach zero as X and Y approach
the critical point. Since xe and y e are constant we have from the above
MA6100
Phase Plane Analysis
dX
=
dt
dY
=
dt
d
= F ( x e + , y e + )
dt
d
= G ( x e + , y e + )
dt
We then expand the functions on the right of each as a Taylor series in 2 variables about the
critical point (xe , ye ) and take the linear approximation to this (a tangent plane) giving
d
F ( xe , ye ) + F ( xe , ye ) + F ( xe , ye )
dt
d
G ( xe , ye ) + G ( xe , ye ) + G ( xe , ye )
dt
However, the first term on each of the right hand sides is zero by definition of the critical
point and we are left with the linear approximation system (in matrix form) of
F
=
G
(16)
where the partial derivatives are evaluated at the critical point (xe , ye ) .
Classification
As (16) is a linear system we can find its eigenvalues in the same way as earlier from the
characteristic equation
2 (trace A) + det(A) =
0
where
trace(A=) F + G
det(
A) F G F G
=
Also from the chain rule,

F F dX F dY F d ( xe + )
F
=
+ =
=
+0
X d Y d X
X
d
and similarly for the other 3 partial derivatives,
F F
G G
G G
= =
=
,
,
Y
X
Y
Then
F FX
=
G GX
FY
and so system (16) can also be written as
GY
X FX
=
Y GX
MA6100
FY X xe
GY Y ye
(17)
2
In this form the matrix of first order partial derivatives in (17) is call the Jacobian of the
system
F
J = X
GX
FY
GY
(18)
So in summary, establish the critical points of the full non-linear system, take a linear
approximation to this system and then evaluate the Jacobian at each point of interest.
Finally, apply the linear theory to classify the behaviour at these points.
It can be shown that in most cases the behaviour of the non-linear system close to each
critical point will be the same as for the linear approximation system. Be aware though that
if the critical point is a centre in the linearized system then it could be a spiral or a centre in
the non-linear system. In this case the non-linear terms are not small enough to be
neglected in predicting the full dynamics.
Predator-Prey Interaction Revisited
We remember from earlier that we have a non-linear system
dX
= 1 X c1 XY
dt
dY
= c2 XY 2Y
dt
(8)

with all parameters positive, and we found there were critical points at (0,0) and 2 , 1
c2 c1
Using the Jacobian, classify the behaviour (8) near to these two points.
MA6100
So the linearized model has a centre, but we are already aware that this is a case which
might not be true of the full non-linear model - it might be a spiral which would mean a very
different outcome in the long-term. We can however do some further analysis now to show
that regardless of the parameter choices the non-linear model cannot have a spiral at
2 1
, . Firstly, we eliminate time from system (8) and solve the resulting ODE to yield an
c2 c1
implicit equation in X and Y.
MA6100
We arrive at the implicit equation
1 lnY c1Y = c2 X 2 ln X + K
(19)
where K is some constant. Consider a single trajectory (just one initial value), and on that
phase plane a vertical line drawn at some value X = X1 . What difference you would notice
between the intersections in the case of a centre compared to that of a spiral?
MA6100
One final analytical result we can obtain relates to the average population, but first
remember that the average value of a function f (x) between a vertical line drawn at some
value
=
x a=
and x b is
1
f (x)dx . We have shown that any trajectory is closed, so lets
b a a
t0 ) X (t0 + T ) and similarly for

assume it takes a time T to return to its position at time t 0 , then X (=
Y (t ) .
MA6100
Exercise
1.
1 1,=
c2 0.005 and with 3 initial values for X and Y of (100,80),
Using the parameters=
(50,50) and (100,170), produce in the PC lab similar graphs to the following for the
Lotka-Volterra predator-prey model whose centre is at (100,100). Hint: you can enter
the subscripted parameters in Maple using the code beta[1]=1.0; etc.
6. Suppose a commercially important crop is endangered by the accidental introduction

from abroad of an insect which damages it. To combat this its natural predator insect is
also imported by the crop farmers, but this does not solve the problem. As a further
measure the farmers are considering using an insecticide which will have per capita
death rates of p1 and p2 for the prey and predator populations respectively.
(a)
Adapt the Lotka-Volterra model (8) to include the effects of the pesticide
(b)
Now run this adapted model on Maple

with all the parameters the same from
question 5, but with p1 =p2 = 0.1 and
reproduce the graph shown opposite.
(c)
Briefly discuss what this suggests for the

average populations in comparison to the
corresponding graph in question 5.
Confirm this by showing algebraically that
the new centre of the model is at
2 + p2 1 p1
,
.
c1
c2
MA6100
7.
An obvious limitation to the earlier predator-prey models is the assumption that the
prey population increases exponential.
(a)
Explain how the insecticide model of question 6 has been enhanced so that it is
modelled by the system
dX
X
= 1 X 1 c1 XY p1 X
dt
K
dY
= c2 XY 2Y p2Y
dt
8.
(b)
Using all the earlier parameter values along with K = 1000 , generate timedependent graphs for each population up to time t = 40 .
(c)
Find the equilibrium populations.
Another simple ecosystem model is that for competing species, where competition for
limited resources such as food or territory key to survival.
(a)
Explain the terms and the assumptions made in this mathematical model of the
competing species:
dX
= 1 X c1 XY
dt
dY
= 2Y c2 XY
dt
(b)
=
1 0.22,
=
2 0.06,
=
c1 0.053
=
c2 0.0046 investigate with Maple the
Taking
behaviour of the system over time; in particular how different initial conditions
can affect the long-term outcome.
MA6100

Pages 16 To 23 (Annotated)

Uploaded by

Copyright:

Available Formats

Pages 16 To 23 (Annotated)

Uploaded by

Document Information

Original Description:

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

Pages 16 To 23 (Annotated)

Uploaded by

Copyright:

Available Formats

Stable nodes and spirals are known as attractors, whilst unstable nodes and spirals are

Nonlinear Systems and Linearization

Let (xe , ye ) be a critical point then we know=

Phase Plane Analysis

Also from the chain rule,

and similarly for the other 3 partial derivatives,

Phase Plane Analysis

Phase Plane Analysis

Phase Plane Analysis

We arrive at the implicit equation

Phase Plane Analysis

t0 ) X (t0 + T ) and similarly for

Phase Plane Analysis

6. Suppose a commercially important crop is endangered by the accidental introduction

Now run this adapted model on Maple

Briefly discuss what this suggests for the

Phase Plane Analysis

Find the equilibrium populations.

Phase Plane Analysis

You might also like