Link

Examples of nonlinear equations

Simple harmonic oscillator (linear ODE)

Nonlinear Differential Equations

d 2 x (t )
= kx(t )
dt

More complicated motion (nonlinear ODE)

and The Beauty of Chaos

d 2 x (t )
= kx (t )(1 x (t ))
dt

Other examples: weather patters, the turbulent motion

of fluids
Most natural phenomena are essentially nonlinear.
1

A simple pendulum

What is special about nonlinear ODE?

For solving nonlinear ODE we can use the same

Model: 3 forces

methods we use for solving linear differential equations

gravitational force

What is the difference?

Solutions of nonlinear ODE may be simple, complicated,

frictional force is proportional

to velocity

or chaotic

periodic external force

Nonlinear ODE is a tool to study nonlinear dynamic:

chaos, fractals, solitons, attractors

Equations

d 2
= g + f + ext
dt 2

g = mgL sin( ), f =

d
,. ext = F cos(t )
4
dt

Case 1: A very simple pendulum

d 2
d
= 02 sin( )
+ f cos(t )
dt 2
dt
F
mgL g

= , =
, f =
02 =
L
mL2
mL2
I

d 2
= 02 sin( )
dt 2
code

Computer simulation: there are very many web sites

with Java animation for the
simple pendulum

Is there any difference between the nonlinear

pendulum

1.2

(0)=1.0
0.8

1.0
0.8

d 2
= 02 sin( )
dt 2

0.6
0.4

(t)

0.2
0.0

and the linear pendulum?

-0.2
-0.4
-0.6

d 2
= 02
dt 2

-0.8
-1.0
-1.2

10

20

30

40

50

60

time
7

Amplitude dependence of frequency

1.2

(0)=1.0
0.8
(0)=0.2

1.0
0.8

For small oscillations the solution for the nonlinear

pendulum is periodic with

0.6
0.4

= 0 = g L

(t)

0.2
0.0

For large oscillations the solution is still periodic but with

-0.2
-0.4

frequency

-0.6
-0.8

< 0 = g L

-1.0
-1.2

10

20

30

40

50

60

explanation:

time

1
sin( ) 2 + K
2
10
sin( ) <

Phase-Space Plot
velocity versus position

Case 2: The pendulum with dissipation

0.8
(0)=1.0
(0)=0.2

0.8
0.6

E1

0.4

d/dt

0.2
0.0

phase-space plot is
a very good way to
explore the dynamic
of oscillations

E2

-0.2

d 2
d
= 02 sin( )
dt
dt 2

1.2
1.0
0.6
0.4
0.2

code

0.0
-0.2
-0.4
-0.6

-0.4

-0.8
-1.0

-0.6

-1.2

-0.8
-1.0
-1.0

(0)=1.0, =0.1

0.8

(t)

1.0

10

20

30

40

50

time

-0.5

0.0

0.5

1.0

11

How about frequency in this case?

12

Phase-space plot

Case 3: Resonance and beats

for the pendulum with dissipation

d 2
= 02 sin( ) + f cos(t )
dt 2

0.8
0.6
0.4

1.2
1.0

When the magnitude of the force is very large the

system is overwhelmed by the driven force (mode
locking) and the are no beats

0.6
0.4

-0.2

0.2

(t)

-0.4
-0.6
-0.8

(0)=1.0, =0.1

0.8

0.0

0.0
-0.2

When the magnitude of the force is comparable with

the magnitude of the natural restoring force the beats
may occur

-0.4

-0.5

0.0

-0.6

0.5

-0.8

-1.0
-1.2

10

20

30

40

code

50

time

13

14

Beats

Example: beats

In beating, the natural response and the driven

response add:

2.5

(0)=0.2, =0.0, f=0.2, =1.1

2.0

t sin

1.5

1.0
0.5

(t)

0 sin(t ) + 0 sin(0t ) = 2 0 cos

+ 0

mass is oscillating at the average frequency

( + 0 ) 2 and an amplitude is varying at the slow

0.0
-0.5
-1.0
-1.5

frequency ( 0 ) 2

-2.0
-2.5

20

40

60

80

100

120

d/dt

d/dt

0.2

time

code
15

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
-1.2
-1.4
-1.6
-1.8
-2.0
-2.0 -1.5 -1.0 -0.5

0.0

0.5

1.0

1.516 2.0

code

Resonance
2.5

(0)=0.2, =0.0, f=0.1, =1.0

2.0
1.5
1.0

Case 4: Complex Motion

For a simple harmonic oscillator

the amplitude of oscillations
increases without bound

(t)

0.5
0.0
-0.5
-1.0

-2.0
-2.5

20

40

60

80

100

120

time

That is not true for the nonlinear

oscillator

d/dt

-1.5

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
-1.2
-1.4
-1.6
-1.8
-2.0
-2.0 -1.5 -1.0 -0.5

d 2
d
= 02 sin( )
+ f cos(t )
dt 2
dt
We have to compare the relative magnitude of the
natural restoring force, the driven force and the
frictional force
The most complex motion one would expect when the
three forces are comparable

0.0

0.5

1.0

1.5

17

2.0

code

18

Case 4: Chaotic Motion

15
10

(0)=1.0, =0.2, f=0.7, =0.66

Case 4: Chaotic Motion

Chaotic motion is not random!

15
10

-5

(t)

-10

-10

d/dt

-15
-20
-25

0 20 40 60 80 100120140160180200220240

time

Chaos is the deterministic

behavior of a system displaying no
discernable regularity

-5

-15

-20

-1

-25

-3

0 20 40 60 80 100120140160180200220240

time

-2

-24-22-20-18-16-14-12-10-8 -6 -4 -2 0 2 4 6 8 10

A chaotic system is one with

an extremely high sensitivity
to parameters or initial
conditions
The sensitivity to even
miniscule changes is so high
that, in practice, it is
impossible to predict the
long range behavior unless
the parameters are known to
infinite precision (which they
never are in practice)

19

20

How can we quantify this lack of

2
1

d/dt

Measuring Chaos

predictably?

-1
-2

How do we know if a system is chaotic?

-3

-24-22-20-18-16-14-12-10-8 -6 -4 -2 0 2 4 6 8 10

This divergence of the trajectories can be described by the

Lyapunov exponent , which is defined by the relation:

The most important characteristic of chaos is sensitivity

to initial conditions.

xn = x0 en

Sensitivity to initial conditions implies that our ability to

make numerical predictions of its trajectory is limited.

where xn is the difference between the trajectories at time n.

If the Lyapunov exponent is positive, then nearby
trajectories diverge exponentially.
Chaotic behavior is characterized by the exponential
divergence of nearby trajectories.
21

22

3
2

xn = x0 e

Chaotic structure in phase space

d/dt

(t)

(0)=1.0,
=0.2, f=0.7, =0.66
(0)=1.00001, =0.2, f=0.7, =0.66

0
-1
-2

1. Limit cycles: ellipse-like figures with

frequencies greater then 0

-3

-24-22-20-18-16-14-12-10-8 -6 -4 -2 0 2 4 6 8 10

2. Strange attractors: well-defined, yet

complicated semi-periodic behavior. Those
are highly sensitive to initial conditions. Even
after millions of observations, the motion
remains attracted to those paths
3. Predictable attractors: well-defined, yet fairy
simple periodic behaviors that not
particularly sensitive to initial conditions

23

4. Chaotic paths: regions of phase space that

appear as filled-in bands rather then lines

24

The Lorenz Model & the butterfly effect

Example

In 1962 Lorenz was looking for a simple model for

weather predictions and simplified the heat-transport
equations to the three equations

dx
dt
dy
dt
dz
dt

= 10( y x )
= xz + 28 x y
=

8
xy z
3

The solution of these simple nonlinear equations gave

the complicated behavior that has led to the modern
interest in chaos
25

26

Hamiltonian Chaos

Summary

The Hamiltonian for a particle in a potential

The simple systems can exhibit complex behavior

Chaotic systems exhibit extreme sensitivity to initial

1
H=
( px2 + p 2y + pz2 ) + V ( x, y , z )
2m

conditions.

for N particles 3N degrees of freedom

Examples: the solar system, particles in EM fields, ...
more specific example: the rings of Saturn
Attention: no dissipation!
Constants of motion: Energy, Momentum (linear, angular)
When a number of degrees of freedom becomes large,
the possibility of chaotic behavior becomes more likely.

27

Practice

28

Fourier Analysis of Nonlinear Oscillations

Duffing Oscillator

d 2x
dx 1
+
x (1 x 2 ) = f cos(t )
dt 2
dt 2

The traditional tool for decomposing both periodic

and non-periodic motions into an infinite number of
harmonic functions

Write a program to solve the Duffing model. Is there a

parametric region in ( , f , ) where the system is
chaotic

It has the distinguishing characteristic of generating

a periodic approximations

29

30

Fourier series

Coefficients:

For a periodic function

the coefficients are determined by the standard

technique for orthogonal function expansion

y (t + T ) = y (t )

one may write

y (t ) =

a0
+ (an cos( nt ) + bn sin( nt ) ),
2 n =1

2
T

an =

2
cos(nt ) y (t )dt ,
T 0

bn =

2
sin( nt ) y (t )dt ,
T 0

2
T

The Fourier series is a best fit in the least square

sense of data fitting
A general function may contain infinite number of
components. In practice a good approximation is
possible with about 10 harmonics
31

32

Fourier transform

Spectral function

The right tool for non-periodic functions

If

y (t ) =

1
2

Y ( )e

it

measures the amount of frequency

1
2

y (t )e

it

making up this

response

and the inverse transform is

Y ( ) =

y (t ) represent the response of some system as a

Y ( ) is a spectral function that

function of time,

dt

2
a plot of Y ( ) versus is called the power spectrum

33

Methods to calculate Fourier transform

34

Discrete Fourier transform

Analytically

Assume that a function y(t) is sampled at a discrete

Direct numerical integration

number of N+1 points, and these times are evenly

Discrete Fourier transform

spaced

(for functions that are known only for a finite number

Let T is the time period for the sampling:

of times tk

a function y(t) is periodic with T, y(t+T)=y(t)

The largest frequency for this time interval is

Fast Fourier transform (FFT)

1 = 2 / T and n = n1 = n 2 / T = n 2 /( Nh )

35

36

Discrete Fourier transform

DFT in terms of separate real and

The discrete Fourier transform, after applying a trapezoid

imaginary parts

rule

Y (n ) =

y (t ) =

1
2
1
2

in t

in t

h
y (t )dt =
2

2
hN

Y ( )d =

2kn
N

eix = cos( x ) + i sin( x )

yk

h N
[(cos(2kn / N ) Re( yk )
2 k =1
+ sin(2kn / N ) Im( yk ))
+ i (cos( 2kn / N ) Im( yk )

Y (n ) =

k =1

n =1

2nt
hN

Y (n )

sin(2kn / N ) Re( yk ))]

37

38

Practice for the simple pendulum

Solve the simple pendulum for harmonic motion, beats,
and chaotic motion (the dissipation and driven forces
are included)
Decompose your numerical solutions into a Fourier
series. Evaluate contribution from the first 10 terms
Evaluate the power spectrum from your numerical
solutions

39