Scribd
Upload
18 views

Modul 5 Limit Cycle

This document discusses limit cycles and bifurcations in two-dimensional dynamical systems. It provides examples of limit cycles and describes how they can be analyzed using tools like the Poincare-Bendixson theorem and nullclines. The document also discusses different types of bifurcations that can occur in two dimensions like saddle-node, Hopf, and transcritical bifurcations.

Uploaded by

yuriska destania
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
18 views

Modul 5 Limit Cycle

This document discusses limit cycles and bifurcations in two-dimensional dynamical systems. It provides examples of limit cycles and describes how they can be analyzed using tools like the Poincare-Bendixson theorem and nullclines. The document also discusses different types of bifurcations that can occur in two dimensions like saddle-node, Hopf, and transcritical bifurcations.

Uploaded by

yuriska destania
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 26

TWO DIMENSIONAL

FLOWS

Lecture 5: Limit Cycles and


Bifurcations
5. Limit cycles

A limit cycle is an isolated closed trajectory


[”isolated” means that neighbouring trajec-
tories are not closed]

Fig. 5.1.1

• Stable limit cycles are very important


scientifically, since they model systems
that exhibit self-sustained oscillations i.e.
systems which oscillate even in the ab-
sence of an external driving force (e.g.
beating of a heart, rhythms in body tem-
perature, hormone secretion, chemical re-
actions that oscillate spontaneously). If
the system is perturbed slightly, it always
returns to the stable limit cycle.

1
• Limit cycles only occur in nonlinear sys-
tems - i.e. a linear system ẋ = Ax can
have closed orbits, but they won’t be iso-
lated!

ṙ = r(1 − r2); r ≥ 0
(
Example 5.1.1
θ̇ = 1
r∗ = 0 is an unstable fixed point and r∗ = 1
is stable.

Fig. 5.1.2 Fig. 5.1.3

hence all trajectories in the phase plane (ex-


cept r∗ = 0) approach the unit circle r∗ = 1
monotonically.

2
Example 5.1.2 Van der Pol oscillator

ẍ + µ(x2 − 1)ẋ + x = 0 (1)


where µ(x2 − 1)ẋ is a nonlinear damping term
whose coefficient depends on position: posi-
tive for |x| > 1 and negative for |x| < 1.

Fig. 5.1.4

Solution for µ = 1.5, starting from (x, ẋ) =


(0.5, 0) at t = 0

Limit cycle is not generally a circle (signals


are not sinusoidal), but must generally be
found numerically.

3
The Poincaré-Bendixson Theorem says that
the dynamical possibilities in the 2-dimensional
phase plane are very limited:

• If a trajectory is confined to a closed,


bounded region that contains no fixed
points, then the trajectory eventually
must approach a closed orbit.

• The formal proof of this theorem is subtle


and requires advanced ideas from topol-
ogy - so will not be developed further
here!

• This theorem does not hold in higher-


dimensional systems n ≥ 3). Trajecto-
ries may then wander around forever in a
bounded region without settling down to
a fixed point or closed orbit.

• In some cases, the trajectories are at-


tracted to a complex geometrical object
4
called a strange attractor; a fractal set
on which the motion is aperiodic and sen-
sitive to tiny changes in the initial con-
ditions. This makes the motion unpre-
dictable in the long run i.e. CHAOS! -
but not in 2D

5.2 Applications of Poincaré-Bendixson


How can we tell if a trajectory is indeed con-
fined to a closed, bounded region?

This is essential if we want to show that


the solution is a closed orbit and not a fixed
point, for example. In order to do this we
need to show that there exists a region R
within which the trajectory will remain for all
t → ∞, and which excludes a nearby fixed
point.

The trick is to construct a trapping region


R, which is a closed, connected set such that
the vector field points “inwards” everywhere
on the boundary of R. Then all trajectories
in R must be confined.
If we can also arrange that there are no fixed
points contained inside R, then the Poincaré-
Bendixson theorem ensures that R must con-
tain a closed orbit.

Fig. 5.2.1

5
Example in polar coordinates Consider the sys-
tem
ṙ = r(1 − r2) + µr cos θ
θ̇ = 1.

When µ = 0 there is a stable limit cycle at


r = 1. Show that a closed orbit still exists
for µ > 0, as long as µ is sufficiently small.

• Solution: We need to find two concentric


circles with radii rmin and rmax such that
ṙ < 0 on the outer circle and ṙ > 0 on
the inner circle. Then the annulus 0 <
rmin ≤ r ≤ rmax is the trapping region.

• For r > 0 there are no fixed points, since


θ̇ = 1, so no fixed points inside the an-
nulus.

• Hence if rmax and rmin exist, then Poincaré-


Bendixson implies the existence of a closed
orbit.

6
• For rmin we require ṙ = r(1−r2)+µr cos θ >
0 for all θ. Since cos θ ≥ −1, any rmin <
√
1 − µ will work.

• For rmax we need ṙ = r(1−r2)+µr cos θ <


0 for all
√ θ. By a similar argument, any
rmax > 1 + µ will work.

• So provided µ < 1 then both rmin and


rmax can be found and a closed orbit ex-
ists.

• In fact the closed orbit can exist for µ > 1


too, though this is more complicated to
show.

Fig. 5.2.2
7
Example using nullclines Consider the system

ẋ = −x + ay + x2y
ẏ = b − ay − x2y
representing a biochemical process called gly-
colysis which cells use to obtain energy from
sugar. x and y are concentrations of the com-
pounds ADP and F6P, and a, b > 0 are pa-
rameters. Construct a trapping region for
this system.

• Solution: We cannot use simple coordi-


nates as for the first example above, but
can first find the nullclines, defined as the
curves on which either ẋ or ẏ = 0.

8
• ẋ = 0 on y = x/(a + x2).

• ẏ = 0 on y = b/(a + x2).

• Then use the fact that the trajectory is


exactly horizontal on the ẏ = 0 nullcline
and exactly vertical on the ẋ = 0 null-
cline, to sketch the vector field.

9
Fig. 5.2.3 & 4

• We can construct the dashed boundary


in Fig. 5.2.4 using information from Fig.
5.2.3, such that all trajectories are in-
wards → a trapping region. By showing
that the fixed point where the nullclines
cross is a repeller, the remaining region
must contain a closed orbit.

10
5.3 Bifurcations revisited
Just as for 1-dimensional systems, we find in
2-dimensional systems that fixed points can
be created or destroyed or destabilized as pa-
rameters are varied - but now the same is true
of closed orbits as well. Hence we can begin
to describe the ways in which oscillations can
be turned on or off.

Saddle-node, transcritical and pitchfork


bifurcations

The bifurcations of fixed points discussed above


for 1-dimensional systems have analogues in
all higher dimensions, including n = 2.

Nothing really new happens when more di-


mensions are added - all the action is con-
fined to a 1-dimensional subspace along which
the bifurcations occur, while in the extra di-
mensions the flow is either simple attraction
to or repulsion from that subspace.

 ẋ = −ax + y
e.g. x2 − by as a model for a ge-
 ẏ =
1+x2
netic control system (a, b > 0)
12
Phase portrait

Fig. 5.3.1

Keep a constant and vary b ⇒ bifurcations


occur along the line y = ax.

13
Hopf bifurcations

Suppose a 2-dimensional system has a stable


fixed point. What are all the possible ways it
could lose stability as a parameter µ varies?

Consider the eigenvalues of the Jacobian λ1


and λ2

Fig. 5.3.2a

λ1 or λ2 passes through λ = 0 on varying


µ ⇒ saddle-node, transcritical and pitchfork
bifurcations.
14
Fig. 5.3.2b

λ1 and λ2 cross Im(λ) axis onto Re(λ) > 0


plane as µ is varied ⇒ Hopf bifurcation

Figs 5.3.2a and b are the only two possible


scenarios for λ1 and λ2, since the λs satisfy
a quadratic equation with real coefficients.

NB The Hopf Bifurcation has no coun-


terpart in 1-dimensionsal systems

15
Supercritical Hopf bifurcation Leads from a
decaying oscillation to growth and saturation
of a sustained oscillation.

Fig. 5.3.3

ṙ = µr − r3
(
Example:
θ̇ = ω + br2

Phase portraits

Fig. 5.3.4
16
Subcritical Hopf bifurcation

Much more dramatic...and potentially dan-


gerous in engineering! After the bifurcation,
the trajectories jump to a distant attractor,
which could be a fixed point, another limit
cycle, infinity or - for n ≥ 3 - a chaotic at-
tractor (e.g. the Lorenz equations in Lecture
6).

The question as to whether a Hopf bifurca-


tion is sub- or super-critical is difficult to an-
swer without an accurate numerical solution
e.g. via a computer.

17
Oscillating chemical reactions

These are a famous example of the occur-


rence of Hopf bifurcations: the Belousov-
Zhabotinsky reaction. This is a complex set
of more than 20 reactions amongst reagents
which include malonic acid, bromate ions,
sulphuric acid and a cerium catalyst. The
system undergoes periodic oscillations which
appear as changes in colour of the solution.

18
The kinetic rate equations for the chemical
reactions can be written in terms of the con-
centrations ci of the ith reagent as
dci
= fi(c1, c2...cN )
dt

19
20
Example
Lengyel et al. (1990) derived a simple model
for another oscillating chemical reaction be-
tween ClO2, iodine (I2) and malonic acid
(M A)
M A + I2 → IM A + I − + H +
1
ClO2 + I − → ClO2− + I2
2
ClO2 + 4I − + 4H + → Cl− + 2I2 + 2H2O
from which equations for the rate of change
of concentration of I2, ClO2 and M A can be
derived which depend on products of the con-
centrations of the other reagents with rate
constants ki. After suitable nondimensional-
isation (see Strogatz Ch. 8 for details), this
reduces to the dynamical system
4xy
ẋ = a − x − 2
1 + x
y
 
ẏ = bx 1 −
1 + x2
where a and b are constants. This turns out
to be a 2-dimensional nonlinear autonomous
dynamical system which can exhibit periodic
oscillations....
21
Closely related are waves of excitation in neu-
ral or cardiac tissue.

22
5.4 Poincaré Maps
These are useful for studying swirling flows,
such as that near a periodic or quasi-periodic
orbit or, as we shall see later, the flow in some
chaotic systems.

Consider an n-dimensional system Ẋ = f (X).


Let S be an n − 1 dimensional surface of sec-
tion.

Fig. 5.4.1

S is required to be transverse to the flow i.e.


all trajectories starting on S flow through it,
not parallel to it.

23
The Poincaré map P is a mapping from S
to itself, obtained by following trajectories
from one intersection with S to the next. If
Xk ⊂ S denotes the kth intersection, then the
Poincaré map is defined by

Xk+1 = P (Xk )
Suppose that X∗ is a fixed point of P i.e.
P (X∗ ) = X∗. Then a trajectory starting at
X∗ returns to X∗ after some time T and is
therefore a closed orbit for the original sys-
tem Ẋ = f (X).

Hence the Poincaré map converts problems


about closed orbits into problems about fixed
points of a mapping.

The snag is that it is typically impossible to


find an analytical formula for P !

24
ṙ = r(1 − r2)
(
Example of a rare exception
θ̇ = 1
Let S be the positive x-axis and r0 is an initial
condition on S. Since θ̇ = 1, the first return
to S occurs after a time of flight t = 2π.
Then r1 = P (r0 ), where r1 satisfies
Z r
1 dr
Z2π
2
= dt = 2π ⇒ r1
r0 r(1 − r ) 0

Hence P (r) = [1 + e−4π (r−2 − 1)]−1/2

Fig. 5.4.2

“Cobweb” construction enables us to iterate


the map graphically. The cobweb shows that
r∗ = 1 is stable and unique (as expected).
25

You might also like