Académique Documents
Professionnel Documents
Culture Documents
a)
x =k 1 axk 2 x 2
A fixed (or stationary) point is a point at which x =0 . This system has two fixed
points at
x =0 and x=k a / k
1
Graphically (fig 1) it can be demonstrated that x =0 is an unstable fixed point and that
x =k 1 a / k 2 is stable
5
4 .5
4
3 .5
3
2 .5
2
1 .5
1
0 .5
0
1 0
2 0
3 0
4 0
5 0
t
6 0
7 0
8 0
9 0
1 0 0
Figure 2
c)
Adding a third reaction to the system:
A X k 1= 2X
A X k 2= 2X
k
B X 3 =C
c 1=k 1 ak 3 b
Figure 1
c 2=k 2
d)
This chemical system has fixed points at
x =0 and
. x = k 1 ak 3 b/k 2 or
x = k 1 ak 3 b/k 2
Figure 3
Question 2
a)
The system
x = xhy
y = x y
x = 1 h
y
1 1
x
y
1 h
1 1
=h1
Using these facts it is possible to determine the nature of the fixed points (Fig 4)
Figure 4
If is negative (0,0) will be a saddle, if h is zero there will be a non isolated fixed point,
that is a point on a manifold line of fixed points.
However if (0,0) is not a saddle point it will be a stable point as the trace is independent of
h.
The red line on the graph (fig 4) denotes the points at which T 2=4 , in this case
when h =0 . If is greater than one ,(0,0) will be a spiral node, otherwise it will be a stable
sink node. Should h exactly equal one, the fixed point will be a star node.
Given that the trace cannot equal zero there is no possibility that the node is a
centre of a fixed orbit.
To summarise (Chart 1)
h
h1
h=1
1h0
h=0
Node Saddle Point Non isolated fixed point Stable Node Stable Star
h0
Stable Spiral
Chart 1
b)
Taking the system
x = yhx x 2 y 2
2
2
y = xhy x y ?
( a ) If h is set to h=0 then the system becomes decoupled. It is clear from
inspection that there exists a fixed point at the origin (0,0)
xy = 10 10 xy
This matrix has a determinant of -1 so the point will be a saddle point. As the matrix
0
1
is diagonalised it is simple to see the eigenvectors and values are 1 1 and -1 0 ,
essentially showing that the x-axis is a stable manifold, and that the y-axis is unstable.
( b ) Setting:
r 2=x 2 y 2
dr
=2 x x2 y y
dt
dr
=2x yhx x 2 y 2 2y hx x 2 y 2 x
dt
( c ) By separation of variables
dr
=hr 3
dt
r 3 dr
1 dt
2r 2 = tc
r=
1
2t K
As t the denominator will also tend towards infinity, regardless of the magnitude
of K, therefore r will tend to 0. Given that the radius of trajectories is decreasing to zero,
this would suggest that (0,0) is now a stable spiral.
c)
Consider the system
dx
= y 2 1
dt
dy
=x 3 y
dt
y=x 3
1 ,1
Graphically:
Figure 5
To analyse the nature of these fixed points the system is linearised about the saddle
points, generating the Jacobian matrix:
2 x
x t
2 y
x t
2 x
y t
2 y
y t
0 2y
2
3x 1
Both the determinant and trace are positive, and T 24 = 23 0 . This is evidence
that (-1 , -1) is a stable spiral. To determine the nature of this spiral it is necessary to
consider some trial points
Point
dx
= y 21
dt
dy
3
=x y
dt
( -1 ,-0.9 )
( -1, -1.1 )
(- 0.9 , -1)
(-1.1 , -1)
-0.19
0.21
-0.1
0.1
0.271
0.331
Sketching these points (Fig 6) and the associated changes in x and y shows that this in
an anti clockwise sink spiral.
Figure 6
Whereas at point (1,1) , the determinant is -6, which is evidence of a saddle node.
Recall the Jacobian matrix, at (1,1):
7
J = 0 2
3 1
11 , 2 and 23 , -3 which
correspond to the manifolds of the saddle node. Given that the first eigenvalue is positive
1
1 is the unstable manifold.
Figure 7
Closed orbits
To exclude the possibility of closed orbits, by Dulac's criterion it is sufficient to find a
continuous and differentiable Lyapunov function G x , y such that
dx
dy
G
G
x
dt
y
dt
2 dx
2 dy
2
2
2
C
C
=C 0C 1=C
x
dt
y
dt
This is clearly negative, for all values of x and y, therefore there are no closed orbits in
this system.
Question 3
a)
a chaotic dynamical system
A chaotic dynamical system is a system in which relatively minor differences is the
initial conditions of a system lead to vastly different behaviour of the system. The system is
thus hard to predict over a long period of time. ( True: a computerised system can perform
several iterations very fast, but there is no way to predict the outcome without running
every iteration , equally by its nature a chaotic system is very susceptible to computational
rounding errors).
However a chaotic system should not be confused with a random system (although
it may appear so). These systems are deterministic, in that if the initial conditions and
nature of the system is known then it is possible to know the system at any time t. Crucially
if an experiment with a chaotic system was repeated with exactly the same initial
conditions it would reach the same state.
A classical example of a chaotic system that was studied by Poincar is the three
mass problem. Illustrated (Fig 8) a diagram by DM Harrison of Toronto University1
Figure 8
Figure 8
This shows the motion of a planet, being affected by the gravity of two similar suns,
following standard Newtonian physics. If this simulation does have a periodic path, it could
not be predicted how long it is. These two simulations differ only in the initial position of the
orbiting plant, but are markedly different.
b)
A phase plane is a system in which only two dimensions are considered, such as
the problems considered earlier in this piece of the form:
xy = ac bd xy
1 DM Harrison , Physics Flash Animations , University of Toronto
10
Figure 9
2 Sprott, J.C (2010) Elegant Chaos . Singapore : World Scientific pp 109 - 115
11
Bibliography
Arrowsmith, D.K. & Place, C.M. (1992) Dynamical Systems. London: Chapman and Hall
Glendinning, P. (1994) Stability, Instability and Chaos. Cambridge: Cambridge University
Press
Harrison, D.M.(2011) Flash Animations for Physics Available at:
http://www.upscale.utoronto.ca/GeneralInterest/Harrison/Flash/#chaos
Accessed 13/03/2012
Perko, L (1991) Differential Equations and Dynamical Systems (2nd ed). New York;
Springer
Sprott, J.C (2010) Elegant Chaos . Singapore : World Scientific
12
Appendix
%%%% Code to see how the chemical system in Q1 tends to a fixed point
k1a = 0.5;
k2 =0.25;
for X0 = 0:0.1:5
x(1) = X0;
t(1) = 0;
for j = 2:1:100
t(j) = t(j-1) + 0.1
x(j) = x(j-1) + 0.1*(k1a*(x(j-1)) - k2*(x(j-1))^2)
end
hold on
plot(x,'r')
end
13