Dynamical Systems and Chaos

CAS Spring 2008

 Introduction to Dynamical Systems
Basic Concepts of Dynamics
• A dynamical system:
– Has a notion of state, which contains all the information upon which the dynamical
system acts.
– A simple set of deterministic rules for moving between states. (Minor exception:
stochastic dynamical systems).
– Iterators. State diagrams that plot x t vs. xt+1 characterize a dynamical system.
• Example asymptotic behaviors
– Fixed point
– Limit cycles and quasi-periodicity
– Chaotic
• Limit sets: The set of points in the asymptotic limit.
• Goal: Make quantitative or qualitative predictions of the asymptotic
behavior of a system.
 Introduction
• Two main types of dynamical systems:
– Differential equations
– Iterated maps (difference equations)
• Primarily concerned with how systems change over time, so focus on
ordinary differential equations (one independent variable).
• Framework for ODE: x ˙ i ≡ dx i / dt
x˙1 = f1 (x1, x 2 ,..., x n )
M
x˙ n = f€
n (x1, x 2 ,..., x n )

• Phase space is the space with coordinates <x1, … xn>

• We call this a n-dimensional system or an n-th order system.
€
 Linear vs. Nonlinear

• A system is said to be linear if all xi on the right-hand side appear

to the first power only.
• Typical nonlinear terms are products, powers, and functions of xi,
e.g.,
– x 1x 2
– (x1)3
– cos x2
• Why are nonlinear systems difficult to solve?
– Linear systems can be broken into parts and nonlinear systems cannot.
• In many cases, we can use geometric reasoning to draw
trajectories through phase space without actually solving the
system.
 Chaos

• Chaotic dynamical systems

– Have complicated, often apparently random behavior
– Are deterministic
– Are predictable in the short term
– Are not predictable in the long term
– Are everywhere
• Turbulence
• Planetary orbits
• Weather
• ?disease dynamics, stock markets and other CAS?
 Example Chaotic Dynamical System
The Logistic Map

• Consider the following iterative equation:

x t +1 = 4rx t (1− x t ) x t ,r ∈ [0,1]

• We are interested in the following questions:
– What are the possible asymptotic trajectories given different x 0 for fixed r?
€ • Fixed points
• Limit cycles
€
• Chaos
– How do these trajectories change with small perturbations?
• Stable
• Unstable
– What happens as we vary r?
 The Logistic Map cont.
• The logistic map:
x t +1 = 4rx t (1− x t )
• What is the behavior of this equation for different values of r and x0?
1
– For r≤ x t ==> 0 (stable fixed point)
€ 4
1 3
– For <r< x t ==> stable fixed point attractor (prev. slide)
4 4
3 note: xt = 0 is a second fixed point (unstable)
€
– For
r> x t ==> periodic with unstable points
4
€

• If r < 1/4 then xt+1 < xt

• However,
€ consider what happens as r increases, between 1/4 and 3/4:
– For an given r, system settles into a limit cycle (period)
– Successive period doublings (called bifurcations) as r increases
Logistic Map
State Diagram
Figure 10.2 goes here.
Transition to Chaos
 Characteristics of Chaos
• Deterministic.
• Unpredictable:
– Behavior of a trajectory is unpredictable in long run.
– Sensitive dependence on initial conditions.
• Mixing : The points of an arbitrary small interval eventually become
spread over the whole unit interval.
– Ergodic (every state space trajectory will return to the local region of a previous point
in the trajectory, for an arbitrarily small local region).
– Chaotic orbits densely cover the unit interval.
• Embedded (infinite number of unstable periodic orbits within a chaotic
attractor).
– In a system with sensitivity there is no possibility of detecting a periodic orbit by
running the time series on a computer (limited precision, round-off error).
• Bifurcations.
 Predicting chaos
• The cascade of bifurcations can be predicted from the
Feigenbaum constant
• The value of r at which logistic map bifurcates into period 2n limit
cycle is an
• dk = (ak - ak -1)/(ak+1 -ak)
• d approaches 4.669…so that the rate of time between
bifurcations approaches a constant.
 Information Loss

• Chaos as Mixing and Folding

• Information loss as loss of correlation from initial conditions
Shadowing Lemma to the rescue
 Attractors
• Initially, a trajectory through a dynamical
system may be erratic. This is known as
the initial transient, or start-up transient.
• The asymptotic behavior of the system is
known as equilibrium, steady state, or
dynamic equilibrium.
– This does not necessarily imply a static equilibrium
or static state.
– The only equilibrium states which can be observed
experimentally are those modeled by limit sets
which receive most of the trajectories.
• These are called attractors.
• Flake, Chapter 11
– Henon Map
– Lorenz Equations

The Lorenz Attractor

Ruele and Takens (1971)
 Attractor Basins
(from Abraham and Shaw, 1984)
• Basin of attraction: The points of all trajectories that converge to a given
attractor.
– In a typical phase portrait, there will be more than one attractor.
• The dividing boundaries (or regions) between different attractor regions (basins)
are called separatrices.
– Any point not in a basin of attraction belongs to a separatrix.
Example Trajectories
Linear Vector Fields

Wikipedia, 2007
Henon Map
xt+1= a - xt2 +byt
yt+1 = xt

x0, y0 between -1 and 1

Time series: xt vs t ?
 State Space for the Henon Map
Henon Map
xt+1= a - xt2 +byt
yt+1 = xt
yt

xt
Lorenz Equations
 Chaos and Scientific Modeling

• Boiling water
• Predicting weather
• Predicting climate change

• How can we make the notion of chaos precise?

• How can we be sure that what looks like chaos is really chaotic
and not just very complicated but perfectly predictable?
– E.g., When we see a seemingly chaotic time series, how can we be sure that
it is not periodic, with an extremely long period?
• How can chaos be measured?
• What is the value of numerical calculations in the presence of
chaos?
 Reading & References

Reading For Monday, Flake Chapter 12

References
• Chaos and Fractals by by H. Peitgen, H. Jurgens, and D. Saupe.
Springer-Verlag (1992).
• Nonlinear Dynamics and Chaos by S. H. Strogatz. Westview (1994).
• J. Gleick Chaos. Viking (1987).
• Robert L. Devaney An Introduction to Chaotic Dynamical Systems.
Addison-Wesley (1989).
• Ralph Abraham and Christopher D. Shaw Dynamics-The Geometry of
Behavior Vol. 1-3 (1984).
 Review: Chaos and Strange Attractors

• Bifurcations leading to chaos:

– In the 1 D logistic map, the amount by which r must be increased to
get new period doublings gets smaller and smaller for each new bifurcation.
– This continues until the critical point is reached (transition to chaos).
• Why is chaos important?
– Seemingly random behavior may have a simple, deterministic explanation.
– Contrast with world view based on probability distributions.
• A formal definition of chaos:
– Chaos is defined by the presence of positive Lyapunov exponents.
• Working definition (Strogatz, 1994)
– “Chaos is aperiodic long-term behavior in a deterministic system that exhibits
sensitive dependence on initial conditions.”
• Strange Attractors: chaotic systems with an asymptotic dynamic equilibrium.
The system comes close to previous states, but never repeats them.
 State Spaces: A Geometric Approach
(Abraham and Shaw, 1984)

• An system of interest is observed in

different states.
• These observed states are the target of
modeling activity.
• State space: a geometric model of the
set of all modeled states.
• Trajectory: A curve in the state space,
connecting subsequent observations.
• Time series: A graph of the trajectory.
• The history of a real system is
represented graphically.
• Example: Lotka-Volterra equations:
population growth of 2 linked populations
dF/dt = F(a-bS)
dS/dt = S(cF-d)
 Lotka Volterra

2 spp Lotka Voltera

dF/dt = F(a-bS)
dS/dt = S(cF-d)
a is reproduction rate of Fish
b is # of Fish a Shark can eat
c is the energy of a Fish (fraction of a new shark)
d is death rate of a shark

Compare to single population logistic map

x
Where
= 4rx (1− x )
t +1 is the equilibrium?
t t

Carrying capacity (density dependence, negative feedbacks) set by the environment vs

another population

Discrete vs continuous equations

stochastic extinction with discrete equations dxi/dt = xiAj(1-xj)
continuous chaos requires 3 dimensions (3 populations) A = A11 A12 A13
matrix of coefficients that spp j has on spp I A21 A22 A23 A31 A32 A33
Lotka Volterra Time Series
 Individual Models
• Implementing chaos as a quasi CA
– Each individual is represented explicitly
– Compare the sizes of the state spaces
• 3 floating point numbers vs
• 2 bits per individual x the number of individuals
– What can we hope to predict in such a complicated system?
– How can we hope to find ecosystem stability
– Relate to Wolfram’s CA classes
 Just one more little complication

• We’ve gone from simple 1 species population model

To a model where multiple populations interact
To a model where each individual is represented

• What if there are differences between individuals?

• Natural selection
– Geometric increases in population sizes
– Carrying capacity (density dependence) that limits growth
– Heritable variation in individuals that results in differential survival
– Populations become dynamical complex adaptive systems
 Reading for MONDAY March 10

1. Flake, chapter 20 -20.02 (p 339-343)

2. Darwin, The Origin of Species, Introduction chapter
http://www.literature.org/authors/darwin-charles/the-origin-of-species-6th
-edition/introduction.html
3. The Central Dogma of Biology for Computer Scientists
http://cnx.org/content/m11415/latest/