Duffing equation (Poincaré section)
This simulation explores the Duffing equation, which reads (in
dimensionless variables) as follows:
x'' + 2 γ x' - x
(1-x2) = f cos ωt
where each '
denotes a time derivative.
-
You can select below the parameters γ and f, as
well as the initial conditions for the elongation x and the
velocity v = x'.
-
The unit time is 1/ ω (so that ω = 1).
-
For information on other elements, put over them the mouse pointer to
get a tooltip.
-
The simulation will display a stroboscopic Poincaré section defined by
a condition in the form
ωt mod 2π = φ.
Activities
-
Start the simulation by pressing Start. By default the section
will be erased after the first 10 cycles to skip the transitory until
the system is on the attractor (well, very very close to it).
-
Use different initial values for x and v. Is there any
difference in the displayed section? Discuss your answer.
-
Use the mouse to select smaller and smaller parts of the attractor
section. The increased resolution will show that the attractor has
structure at all magnification levels: it is a fractal, much like a
Cantor set.
-
Notice that at higher magnifications you will have to wait longer to
get a number of solution points high enough to see the attractor.
-
Press Window to recover the original magnification.
-
Change φ by steps from 0 to 2π to get an idea of the full
attractor in the three-dimensional space (t,x,v).
You can see at work one of the mechanisms of chaos: the so-called baker
transformation, which is better shown in Duffing_baker.
-
You can run the examples Duffing1, Duffing2, Duffing3
and Duffing4 of Dynamics
Solver for faster simulations, which allow exploring the Cantor
set structure of the attractor, as well as the baker transformation.
The sensible dependence on initial conditions is shown in Duffing5.
This is an English translation of the Basque original for a course
on mechanics, oscillations and waves.
It requires Java
1.5 or newer and was created by Juan
M. Aguirregabiria with Easy
Java Simulations (Ejs) by Francisco
Esquembre. I thank Wolfgang Christian and Francisco Esquembre for
their help.