Example 8.1: Let
, and let
satisfy
(2)
, for
, and
(3)
is strictly increasing on
.
The set
is an attractor satisfying
Proof: It is a standard exercise to show that
is an attractor with
. Let
Note that
is an attractor block associated with
for
and that
. Therefore,
, which establishes the inequality
Now consider the sequence
, where
,
and
, for
. If
,
then
, which means that
. Let
satisfy
, for
, and
.
Define
by
Note that
is an
-pseudo-orbit for any
and that
leaves any compact subset of
. Since
can be chosen arbitrarily close to
,
this statement implies that
and the proof is complete.
Example 8.4: Let
, and let
be the general quadratic map in
standard form:
There is a unique value of
, which happens to be close to
, for
which there is a superattracting orbit of period
with itinerary
. In other words, there is a periodic orbit
, with
,
and
,
for
.
One can show that
Note that the intensity is of the order
when
is
.
Therefore, this particular family of attractors will be extremely
difficult to detect by direct computer simulation for even relatively
modest periods.
Example 8.5: Let
, and let
be the time 1 map of the vector field
where
are polar coordinates on
. For
positive values of
, this map has two attractors,
where
. Note that
is an invariant circle,
is a
set of
fixed points, and
.
Here one can estimate that
for small
.
This example appears to be artficial, but it is related to supercritical
Hopf bifurcation for maps of the plane. The attractor
corresponds to
the invariant circle, while the attractor
corresponds to the periodic
sink with rotation number
. One can see that, while the invariant
circle is not too difficult to detect with direct computer simulations,
even modestly high resonances pose a problem. For example, with 64 bit
arithmetic, one can reasonably expect to detect an invariant circle with a
radius of
. However, one would expect to have difficulty
detecting a periodic sink of period 33 for a radius less than
.
Experience has shown that these resonances are indeed difficult to find
with direct computer simulations [1].
Copyright (c) 1998 by Richard
McGehee, all rights reserved.