UNIVERSITY OF WATERLOO

Faculty of Mathematics

Chaos Theory and the Lorenz Equation: History, Analysis, and Application

April 4, 2012

prepared by
 Table of Contents

1.0 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2.0 Edward Lorenz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

3.0 The Lorenz Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

3.2 Fixed Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3.3 The Lorenz Attractor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

4.0 Graphical Interpretations of the Lorenz Equation . . . . . . . . . . . . . . . . . . . . 6

5.0 The Buttery Eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

6.0 Applications of Chaos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

7.0 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Appendix A  Lorenz Equation Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

List of Figures

Figure 1  XZ plane, ρ = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Figure 2  Time vs X, ρ = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Figure 3  XZ plane, ρ = 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Figure 4  Time vs X, ρ = 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Figure 5  XZ plane, ρ = 26 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Figure 6  Time vs X, ρ = 26 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

ii
1.0 Introduction

The purpose of this paper is to provide the reader with an introduction to the concepts of chaos theory

and sensitive dependence on initial conditions. One of the rst and most famous equations regarding

these two topics, the Lorenz Equation, will be discussed. The equation's history, properties, and graphical

interpretations will be examined. This paper concludes with a brief discussion of chaos in our world today.

To begin the study of chaos and sensitive dependence on initial conditions, one must dene these terms.

Chaos can be described as long term, aperiodic behaviour that exhibits sensitive dependence on initial

conditions. Sensitive dependence on initial conditions implies that nearby trajectories diverge exponentially

fast over time, (Strogatz 320). This means if one begins with two near identical states, they will arrive at

two drastically dierent states after iterating them over some chaotic function. Mathematicians of the 1800s

James Maxwell and Henri Poincare developed ideas regarding sensitive dependence on initial conditions,

but it was not until Edward Lorenz that formal mathematics was used to explore this new and exciting

eld.

2.0 Edward Lorenz

The Lorenz Equation was named for its inventor Edward Norton Lorenz, an American mathematician and

meteorologist. Born in 1918, he studied mathematics at Dartmouth and Harvard, before breaking from his

work to serve as a weather forecaster for the Air Force during World War II. As the war concluded, he

entered the Massachusetts Institute of Technology (MIT), where he earned a doctoral degree in meteorology.

He was immediately hired as sta at MIT and was made an associate professor in 1955 (Telegraph).

In the mid-1900s, the eld of meteorology was still very much in its infancy. Lorenz programmed an

existing computer, the Royal McBee, to aide in his research of atmosphere equations and forecasting. His

program allowed him control the initial conditions of a weather system based on 12 dierential equations.

Furthermore, his program could perform up to sixty calculations per second, and was able to run for a very

long time without stopping. Lorenz would input data and his program would return the resulting weather

and atmosphere information (Elert, Telegraph).

1
3.0 The Lorenz Equation

The idea behind the Lorenz Equation came in 1961, when Lorenz ran his program with data rounded o

from a previous experiment. He recalls the moment he realised the `chaos' present in weather systems:

I typed in some of the intermediate conditions which the computer had printed out as new
initial conditions to start another computation and then went out for a while. Afterwards, I
found that the solution was not the same as the one I had before. I soon found that the reason
was that the numbers I had typed in were not the same, but were rounded o numbers. The
small dierence between something retained to six decimal places and rounded o to three had
amplied in the course of two months of simulated weather until the dierence was as big as the
signal itself. And to me this implied that if the real atmosphere behaved in this method then
we simply couldn't make forecasts two months ahead. The small errors in observation would
amplify until they became large. (Peitgen)

This occurrence would become the basis for the Lorenz Equation. The fact that two weather conditions,

which dier by less than 0.1%, can produce drastically dierent results underscores how chaos and sensitive

dependence on initial conditions are integrated into meteorology.

In March 1963, Lorenz wrote that he wanted to introduce, ordinary dierential equations whose solutions

aord the simplest example of deterministic non periodic ow and nite amplitude convection, (Lorenz

134). In his paper, he examines the work of meteorologist Barry Saltzman and physicist John Rayleigh

while incorporating several physical phenomena (Bradley, Viswanath). Lorenz found that when applying

the Fourier Series to one of Rayleigh's convection equations that,  . . . all except three variables tended to

zero, and that these three variables underwent irregular, apparently non periodic functions, (Lorenz 135).

He then used these variables to construct a simple model based on the 2-dimensional representation of the

earth's atmosphere. The following is known as the Lorenz Equation:

dx
= Ẋ = σ(y − x)
dt
dy
= Ẏ = ρx − y − xz
dt
dz
= Ż = xy − βz
dt

2
3.1 Parameter Explanation

Here x, y, z do not refer to coordinates in space (Gulick 276). In fact, x represents the convective over-

turning on the plane, while y and z are the horizontal and vertical temperature variation respectively.

The parameters of this model are σ, which represents the Prandtl number, or the ratio between the uid

viscosity to its thermal conductivity, ρ, which represents the dierence in temperature between the top and

bottom of the atmosphere plane, and β, which is the ratio of the width to the height of the plane (Bradley,

Gulick). Lorenz found the values of σ = 10 and β = 8/3, and initial conditions of (x0 , y0 , z0 ) = (0, 1, 0) to

be the best representation of the earth's atmosphere (Lorenz 136-137). For this project, we assume these

values.

3.2 Fixed Points

The following proof is based on Record 2003 and Rothmayer 1993

While Lorenz found chaos to be a large factor in meteorology, the equation he created does not exhibit

chaos for all parameters. In fact, there are many parameter values where the function is stable and contain

xed points. We will now explore how Lorenz came to realize xed points of his system, as well as for what

values the equation exhibits chaos.

To nd xed points of the Lorenz Equation, we will rst solve for its equilibria. To nd these equilibrium

points, we will set Ẋ , Ẏ , and Ż to 0.

σ (y - x) = 0 (1)

ρx - y - xz = 0 (2)

xy - βz = 0 (3)

Solving:

x = y (1)

y = ρx - xz (2)

x = x(ρ - z) ⇒ z = ρ - 1 (1) (2)

x2 - (ρ - 1)β = 0 (1)(3)

3
 p
x = y = ± β(ρ − 1)

Our three equilibrium points are

p p p p
(0, 0, 0), K1 = ( - β(ρ − 1), - β(ρ − 1), ρ - 1), K2 = ( β(ρ − 1), β(ρ − 1), ρ - 1)

The behaviour of the Lorenz Equation is complex, so we consider cases of the parameters. We established

that Lorenz preferred σ = 10 and β = 8/3, and thus, we will only concern ourselves with ρ.

Case 1: 0< <1 ρ

(0, 0, 0) yields the only real xed point of the equilibrium points. There is no chaos when 0 < ρ < 1

Case 2: ρ = 1

We identify ρ = 1 as a bifurcation point, as the other two equilibrium points will appear when ρ > 1. There

is no chaos when ρ = 1

Case 3: ρ > 1

p p p p
We have two new xed points K1 = ( - β(ρ − 1), - β(ρ − 1), ρ - 1), K2 = ( β(ρ − 1), β(ρ − 1), ρ -

1)

We will need to check the stability of these points by linearizing the Lorenz Equation and nding its

eigenvalues. A point is stable when its eigenvalues are all negative. We will linearize the system near an

already established equilibrium point from above, call it (X , Y , Z) using the Jacobian Matrix. This gives

    
Ẋ   −σ σ 0  X 
    
 Ẏ  = ρ − Z −1 X   Y 
    
    
Ż Y X −β Z

To get the eigenvalues of the above 3 x 3 matrix (call it A), we solve det(A - λ I) = 0

This yields
 
−σ − λ σ 0 
 
det 
 ρ − Z −1 − λ X =0

 
Y X −β − λ

4
 2 2
λ3 + (β + σ + 1)λ2 + (β + βσ + σ − ρσ + σZ + X )λ + βσ(1 − ρ) + σ(XY + X + βZ) = 0(4)

If we take (X , Y , Z) to be the equilibrium point (0, 0, 0), we get

λ3 + (β + σ + 1)λ2 + (β + βσ + σ − ρσ)λ + βσ(1 − ρ) = 0

-β is a solution, so we can factor to get

(λ + β)(λ2 + (σ + 1)λ + σ(1 − ρ)) = 0

Thus the eigenvalues are

p
−σ − 1 ± (σ + 1)2 + 4σ(ρ − 1)
λ1 , λ2 = , λ3 = −β
2

When ρ > 1 however, λ1 > 0 λ2 , λ3 < 0 and (0, 0 , 0) is not stable and thus not a xed point for ρ > 1

If we take (X , Y , Z) to be either K1 or K2 and plug them into (4), we end up with eigenvalues

µ3 + (β + σ + 1)µ2 + (σ + ρ)βµ + (1 − ρ)2σβ = 0

All three eigenvalues µ 1 , µ2 , µ3 will be negative when

σ(σ + β + 3)
ρ< = ρc
σ−β−1

470
Substituting σ = 10, β = 8/3, we get ρ <
19 ≈ 24.74 = ρc

Thus, K1 , K2 are stable and xed points when 1 < ρ < 24.74. When ρ≥ 24.74, not all of µ1 , µ2 , µ3 will

be negative, and K1 , K2 will not be stable and thus not xed points.

At ρ > ρc , the three equilibrium points ( (0, 0, 0), K1 , K2 ) are unstable and are not xed points. They do

not approach innity, but rather enter a region around the origin. This is where we begin to see chaotic

behaviour of the Lorenz Equation.

5
To recap the xed points of the Lorenz Equation are

ρ Fixed Points

[0-1] (0, 0, 0)

(1, 24.74) K1 , K2

[24.74-30.1) None, chaos occurs

[30.1 - ∞) intermittency (not proven)

3.3 The Lorenz Attractor

As shown above, when 24.74 ≤ρ < 30.1 the Lorenz Equation displays chaos. This condition on ρ gives the

equation a `nickname': The Lorenz Attractor. The Lorenz Attractor is a strange attractor, which means the

equation is non-periodic, as thus never repeats itself. Strange attractors are also coupled with the notion of

chaos and sensitive independance on initial conditions, in that one cannot predict where on the attractor

the system will be in the future.

4.0 Graphical Interpretations of the Lorenz Equation

The following graphs are based on the Lorenz Equation using initial conditions (x0 , y0 , z0 ) = (0, 1, 0),

σ = 10, β = 8/3, and time = 60 seconds; only ρ is changed. Two slightly dierent starting points (one

blue, the other red) and their subsequent orbits are plotted on the x-z plane as well as the x-time plane.

It should be noted that another property of the Lorenz Equation, is that it is deterministic, which means

each set of initial conditions produces a unique graph (Carriuolo 2006). One would expect dierent graphs

if we had used the initial conditions (x0 , y0 , z0 )=(1, 0, 1), though our results and conclusions would remain

unchanged

6
 xz.jpg

Figure 1: XZ plane, ρ = 1

time.jpg

Figure 2: Time vs X, ρ = 1

The red and blue orbits appear to behave similarly over time; their paths are very much the same. Accord-

ingly, no chaos is present in this equation. This is to be expected as ρ = 1 produces a stable system.

7
 xz.jpg

Figure 3: XZ plane, ρ = 10

time.jpg

Figure 4: Time vs X, ρ = 10

Again, the blue and red orbits behave very similarly over time and over much iteration; no chaos is present.

This is to be expected as ρ = 10 < 24.74

8
 xz.jpg

Figure 5: XZ plane, ρ = 26

time.jpg

Figure 6: Time vs X, ρ = 26

These two graphs of the Lorenz Equation demonstrate chaos. Small, initial distances between the blue and

red points lead to large, unpredictable distances between them after numerous iterations on the x-z plane

and over time . Hence, sensitive dependence on initial conditions and chaos are present in this equation.

This is expected, as 24.74≤ρ < 30.31.

It is dicult to notice on the two-dimensional x-z plane, but orbits on Figure 5 (and for all Lorenz Attractor

graphs) do not intersect. The red orbit never overlaps itself nor the blue orbit and vice versa. This

emphasizes the idea that these orbits are non-periodic. If the orbits were to cross one another, cycles would

exist, and therefore chaos would be absent (Gulick 282).

Lorenz himself was able to produce the x-z plane and time series images on his computer. In conclusion to

9
his famous paper, he declared, when our result concerning the instability of non periodic ow are applied

to the atmosphere, they indicate that prediction of the suciently distant future is impossible by any

method, (Lorenz 141). Here, Lorenz states that sensitive dependence on initial conditions make it dicult

to predict the weather, or any chaotic system for that matter.

5.0 The Buttery Eect

One can see that Figure 5 resembles a buttery, and thus, the Lorenz Attractor and chaos in general are

nicknamed, The Buttery Eect. This draws upon Lorenz's ndings that two seemingly identical weather

systems could produce two very dierent weather systems in the near future. Thus, a buttery apping

its wings could alter the atmosphere ever so slightly, so as to deviate from the initial conditions, and

accordingly alter the course of weather forever. Lorenz rst used the example of a seagull's wings, though

the analogy has morphed into using a buttery (Telegraph).

6.0 Applications of Chaos

Chaotic systems have many applications, in particular to nonlinear equations. Chaotic systems can be used

to explain topics in engineering, geography, and even the stock market. In 1991, Edgar E Peters, wrote,

for over 30 years, there were many people who thought the stock market follows regular cycles. Recent

research, however, suggests the S&P 500 has non periodic cycles, governed by attractors. Peters goes on

to formulate several equations proving his claim and explains how one can determine whether a system is

a chaotic. To do so, one must use system equations to construct a phase space (this can be very dicult).

Also, this phase space needs to have fractal dimension and sensitive dependence on initial conditions. Peters

states that the need for chaotic systems to exist is because linear nancial models have failed many times.

As with Lorenz's conclusion about weather prediction, Peters states, long range economic forecast is not

feasible beyond a short time frame," (Peters 62). This example highlights one of the many ways our lives

and the world around us are lled with chaos.

10
7.0 Conclusion

Though discovered by accident, the Lorenz Equation has had a signicant contribution to mathematics

and many other disciplines. In particular, the Lorenz Equation helped pioneer the study of chaos and

sensitive dependence on initial conditions. This equation behaves like any other family of equations, in that

it has xed and bifurcation points, which can be graphed accordingly. Unlike other families, however, this

equation is chaotic for certain parameter values. When it is chaotic, it is known as the Lorenz Attractor,

which has certain chaotic properties. The Lorenz Equation has made quantifying chaos possible which has

inspired many mathematicians to research and study chaos.

11
References

[1] Bradley, Larry. Chaos and Fractals. 2010. Web <http://www.stsci.edu/~lbradley/seminar/

[2] Carriuolo, Matthew. The Lorenz Attractor, Chaos, and Fluid Flow 2005. Web <http://physics.
brown.edu/physics/undergradpages/theses/matthew_carriuolo_thesis.pdf.

[3] Elert, Glenn. The Chaos Hypertextbook 1995. Web <http://www.hypertextbook.com/chaos/21.

[4] Gulick, Denny. Encounters with Chaos and Fractals Maryland, USA. McGraw-Hill Education, 1992

[5] Lorenz, Edward N. Deterministic Nonperiodic Flow Journal of Atmospheric Sciences, Volume 20.

Boston, USA. 1963

[6] Peitgen, H. O., H. Jürgens, D. Saupe, & C. Zahlten. Fractals: An Animated Discussion. Videocassette.

New York: Freeman, 1990.

[7] Peters, Edgar E. A Chaotic Attractor for the S & P500. Financial Analysis Journal, Vol 47, No. 2.

CFA Institute, 1991.

[8] Record, Nicholas. Introduction to Lorenz's System of Equations 2006. Web. <http://www.math.uni.
wroc.pl/~karch/B2/lorenz-system.pdf.

[9] Rothmayer, A.P, D.W. Black. Ensembles of the Lorenz Attractor. Mathematical and Physical Sciences,

Vol 441, No 1912. 1993

[10] Strogatz, Stephen Henry. Nonlinear Dynamics and Chaos. Da Capo Press, 1994

[11] The Telegraph Obituaries: Professor Edward Lorenz 2008. Web <http://www.telegraph.co.uk/
news/obituaries/1895916/Professor-Edward-Lorenz.html>

[12] Viswanath, Divakar. The Fractal Property of the Lorenz Attractor. Physica D: Nonlinear Phenomena,

Volume 190, Issues 12, March 2004.

12
Appendix A  Lorenz Equation Graphs

The gures in this project were made using the website <http://www.cmp.caltech.edu/~mcc/Chaos_
Course/Lesson1/Demo8.html>

13