Simple examples of strange nonchaotic attractors
In modern nonlinear dynamics, systems with continuous and discrete time are
studied on equal rights. The first are those defined by differential equations, and the second by iterative maps.
In a continuous-time system appearance of SNA becomes possible in a case of
quasiperiodic time dependence of coefficients in the differential equation, which demand
presence at least two incommensurate frequencies. In iterative maps there is an internal
own 'rhythm', the steps of discrete time, so, it is sufficient to have one
frequency of driving incommensurate with the frequency of steps.
The simplest class of systems manifesting SNA is represented by mappings with one coordinate variable and one phase.
The external driving is characterized by one parameter of frequency,
which is often taken to be equal to the 'golden mean',
the number
(51/2-1)/2.
Remark.
This selection may be explained not only by simplification of the theoretical analysis, as
usually mentioned, but also by the fact that the subtle structures examined in
experiments and numerics are distinguishable better that at other frequency ratios.
Let us take a map
xn+1=lth xn,
which demonstrates a pitchfork bifurcation: at
l<1
a single attractor is a fixed point at origin; at
l>1
it becomes unstable, but instead two symmetric stable points appear
x=±x0(l).
Following Grebogi et al., let us modify the map introducing a quasiperiodic
modulation of the coefficient:
where q is a phase variable.
Parameter w=(51/2-1)/2 corresponds to a frequency of the external driving,
which is of multiplicative nature.
It is easy to estimate Lyapunov exponent for the solution x=0.
In linear approximation,
and
Due to irrationality of w, the phase values
qn at
N® ¥
are uniformly and dense distributed over the unit interval.
Therefore, we change the sum to the integral, and for the Lyapunov exponent get
At
l<1
we have
L<0, i.e. the solution is stable,
and the line x=0 on the plane
(q,x)
is an attractor.
For l>1
it is unstable, has a positive Lyapunov exponent, and cannot be an attractor.
What can one say about an attractor at
l>1?
First, it must contain points with |x|>0.
Second, due to the relation |th x|<1, the attractor is placed entirely
in a band
|x|<2l. Third, the axis x=0 contains an infinite
set of points of the attractor. Indeed, at
q=1/4 cosine vanishes, therefore, for all
qn=1/4+nw,
where n=1, 2,...,¥
we have xn=0. This subset of the attractor is infinite and dense,
so, the attractor obviously is a nontrivial geometric object, the 'strange set'.
In figure, portraits of SNA in the model under consideration are shown
on the phase plane
Lyapunov exponent of this attractor is negative. Indeed, for the function
f(x)=th x the relation is valid:
|f'(x)|£|f(x)/x|,
and the equality is reached only at x=0.
Hence, for the Lyapunov exponent we have
Lyapunov exponent of this attractor is negative. Indeed, for the function
f(x)=th x the relation is valid:
|f'(x)|£|f(x)/x|,
and the equality is riched only at x=0.
Hence, for the Lyapunov exponent we have
Driven quadratic map
One of popular and rich models in nonlinear dynamics is a quadratic map
xn+1=l-xn2,
manifesting a transition to chaos via period doubling bifurcation cascade under
increase of the control parameter l. Introducing
modulation of the parameter with an irrational frequency w
and amplitude e, we get the model
which can demonstrate SNA.
In the following figure, we show a chart of regimes in the parameter plane
(l, e),
and on periphery examples of iteration diagrams
for attractors at several characteristic points.
Yellow, orange and light blue areas are those of torus of type Т1, Т2, Т4,
which are depicted by a respective number of closed curves in iteration diagrams.
Green designates SNA, and dark blue - chaos.
Model with tangent bifurcation
Let us consider a map
xn+1=f(xn)+b,
where a function f is selected in such way that a tangent
bifurcation occurs: in a course of increase of
b two fixed points, one stable and another unstable,
get closer, collide and disappear, leaving a narrow 'channel', through which
an orbit travels for a long time. If the dynamics after the passage
is of such kind that a return to the input of the channel is ensured, a type of
dynamical behavior known as intermittency (Pomeau and Manneville, 1980) takes place.
To perform a local analysis of the bifurcation, we may exclude the reinjection mechanism and
consider function f(x)=x/(1–x).
Assuming that the parameter b oscillates quasiperiodically with amplitude
e, we come to the model
It is appropriate for a study of the transition, which is an analog of the tangent
bifurcation al gives birth to SNA, but for description of the SNA itself
it is needed in a modification introducing the reinjection. It is interesting
that a variable change
xn=1–Yn/Yn–1
reduces the equation to a linear difference equation of Harper
known in solid-state physics in the context of analysis
of localization and delocalization of quantum states in one-dimensional model
of quasiperiodic medium.
Circle map with quasiperiodic driving
The circle map is defined as
xn+1=xn+r+(K/2p)sin2pxn,
where the variable xn is defined up to an integer part ('modulo 1').
It may be interpreted as an equation for self-oscillator under periodic pulses: xn
is a phase for the oscillations just before the n-th kick, K
characterizes intensity of the kicks, and r – detuning of the frequency of pulses and
of self-oscillations in the device. With a term corresponding to an additional external
driving with incommensurate frequency we get a model
Without the additional term,
in the parameter plane (r,K) one can see the Arnold tongues, the regions of periodic
dynamics, or 'mode locking' (gray), quasiperiodicity between them (white), and chaos (blue),
see the left diagram.
The bottom part of the diagram, where the regular dynamics take place, and the top one, where chaos is
possible, are separated by the critical line K=1 (dotted).
In presence of the quasiperiodic force, instead of the critical line one observes a critical zone
(Ding et al., Phys.Rev. A39, 1989, 2593). Its width grows with the amplitude of the
driving (see the right diagram). The Arnold tongues transform into zones of
two-frequency quasiperiodic dynamics (one is the frequency of steps of discrete time, and the
second - the frequency of driving). Regimes between the tongues correspond now to
the three-frequency quasiperiodicity, or three-dimensional tori (a frequency of 'rotation' of the cyclic
variable x is the third one).
These two types of regular dynamics take place below the critical zone.
Above it the chaotic dynamics occur. Finally, in the critical zone itself, in dependence of
parameters one can observe either two-frequency quasiperiodicity, or SNA (see an example in the
figure below
for e=0.6, K=1, r=0.3,
w=(51/2-1)/2).